diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,487213 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.9996674979218625, + "global_step": 81198, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 8.210180623973727e-09, + "loss": 1.3529, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 1.6420361247947455e-08, + "loss": 1.2748, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 2.4630541871921184e-08, + "loss": 1.2929, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 3.284072249589491e-08, + "loss": 1.3449, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 4.105090311986864e-08, + "loss": 1.2153, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 4.926108374384237e-08, + "loss": 1.1738, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 5.7471264367816094e-08, + "loss": 1.2771, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 6.568144499178982e-08, + "loss": 1.2881, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 7.389162561576355e-08, + "loss": 1.425, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 8.210180623973728e-08, + "loss": 1.2565, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 9.0311986863711e-08, + "loss": 1.2989, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 9.852216748768474e-08, + "loss": 1.2727, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 1.0673234811165847e-07, + "loss": 1.438, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 1.1494252873563219e-07, + "loss": 1.2393, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 1.2315270935960593e-07, + "loss": 1.3218, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 1.3136288998357964e-07, + "loss": 1.3124, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.3957307060755338e-07, + "loss": 1.296, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.477832512315271e-07, + "loss": 1.3306, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 1.5599343185550083e-07, + "loss": 1.284, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 1.6420361247947457e-07, + "loss": 1.3571, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 1.7241379310344828e-07, + "loss": 1.4006, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 1.80623973727422e-07, + "loss": 1.313, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 1.8883415435139576e-07, + "loss": 1.3452, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 1.9704433497536947e-07, + "loss": 1.3528, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 2.0525451559934318e-07, + "loss": 1.2589, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 2.1346469622331695e-07, + "loss": 1.2113, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 2.2167487684729066e-07, + "loss": 1.3568, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 2.2988505747126437e-07, + "loss": 1.2717, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 2.3809523809523811e-07, + "loss": 1.274, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 2.4630541871921185e-07, + "loss": 1.306, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 2.5451559934318556e-07, + "loss": 1.2708, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 2.627257799671593e-07, + "loss": 1.2669, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 2.7093596059113304e-07, + "loss": 1.2226, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 2.7914614121510676e-07, + "loss": 1.3396, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 2.8735632183908047e-07, + "loss": 1.1855, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 2.955665024630542e-07, + "loss": 1.2978, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 3.0377668308702795e-07, + "loss": 1.3433, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 3.1198686371100166e-07, + "loss": 1.2657, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 3.201970443349754e-07, + "loss": 1.2795, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 3.2840722495894914e-07, + "loss": 1.2403, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 3.3661740558292285e-07, + "loss": 1.3992, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 3.4482758620689656e-07, + "loss": 1.2887, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 3.530377668308703e-07, + "loss": 1.2894, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 3.61247947454844e-07, + "loss": 1.4721, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 3.694581280788178e-07, + "loss": 1.3645, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 3.776683087027915e-07, + "loss": 1.2289, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 3.8587848932676523e-07, + "loss": 1.281, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 3.9408866995073894e-07, + "loss": 1.2876, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 4.0229885057471266e-07, + "loss": 1.2699, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 4.1050903119868637e-07, + "loss": 1.2833, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 4.1871921182266013e-07, + "loss": 1.2038, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 4.269293924466339e-07, + "loss": 1.2869, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 4.351395730706076e-07, + "loss": 1.3214, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 4.433497536945813e-07, + "loss": 1.2807, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 4.5155993431855504e-07, + "loss": 1.3728, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 4.5977011494252875e-07, + "loss": 1.3662, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 4.6798029556650246e-07, + "loss": 1.3504, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 4.7619047619047623e-07, + "loss": 1.2475, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 4.844006568144499e-07, + "loss": 1.2605, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 4.926108374384237e-07, + "loss": 1.3081, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 5.008210180623975e-07, + "loss": 1.3331, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 5.090311986863711e-07, + "loss": 1.2792, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 5.172413793103449e-07, + "loss": 1.2692, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 5.254515599343186e-07, + "loss": 1.3927, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 5.336617405582923e-07, + "loss": 1.206, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 5.418719211822661e-07, + "loss": 1.2989, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 5.500821018062399e-07, + "loss": 1.2834, + "step": 67 + }, + { + "epoch": 0.01, + "learning_rate": 5.582922824302135e-07, + "loss": 1.2454, + "step": 68 + }, + { + "epoch": 0.01, + "learning_rate": 5.665024630541873e-07, + "loss": 1.3024, + "step": 69 + }, + { + "epoch": 0.01, + "learning_rate": 5.747126436781609e-07, + "loss": 1.2601, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 5.829228243021347e-07, + "loss": 1.3885, + "step": 71 + }, + { + "epoch": 0.01, + "learning_rate": 5.911330049261084e-07, + "loss": 1.3489, + "step": 72 + }, + { + "epoch": 0.01, + "learning_rate": 5.993431855500822e-07, + "loss": 1.3637, + "step": 73 + }, + { + "epoch": 0.01, + "learning_rate": 6.075533661740559e-07, + "loss": 1.2922, + "step": 74 + }, + { + "epoch": 0.01, + "learning_rate": 6.157635467980297e-07, + "loss": 1.2083, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 6.239737274220033e-07, + "loss": 1.3628, + "step": 76 + }, + { + "epoch": 0.01, + "learning_rate": 6.321839080459771e-07, + "loss": 1.2029, + "step": 77 + }, + { + "epoch": 0.01, + "learning_rate": 6.403940886699508e-07, + "loss": 1.3244, + "step": 78 + }, + { + "epoch": 0.01, + "learning_rate": 6.486042692939245e-07, + "loss": 1.2332, + "step": 79 + }, + { + "epoch": 0.01, + "learning_rate": 6.568144499178983e-07, + "loss": 1.2153, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 6.65024630541872e-07, + "loss": 1.3931, + "step": 81 + }, + { + "epoch": 0.01, + "learning_rate": 6.732348111658457e-07, + "loss": 1.2675, + "step": 82 + }, + { + "epoch": 0.01, + "learning_rate": 6.814449917898195e-07, + "loss": 1.2873, + "step": 83 + }, + { + "epoch": 0.01, + "learning_rate": 6.896551724137931e-07, + "loss": 1.2074, + "step": 84 + }, + { + "epoch": 0.01, + "learning_rate": 6.978653530377669e-07, + "loss": 1.2863, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 7.060755336617405e-07, + "loss": 1.2507, + "step": 86 + }, + { + "epoch": 0.01, + "learning_rate": 7.142857142857143e-07, + "loss": 1.28, + "step": 87 + }, + { + "epoch": 0.01, + "learning_rate": 7.22495894909688e-07, + "loss": 1.2512, + "step": 88 + }, + { + "epoch": 0.01, + "learning_rate": 7.307060755336617e-07, + "loss": 1.3699, + "step": 89 + }, + { + "epoch": 0.01, + "learning_rate": 7.389162561576356e-07, + "loss": 1.1732, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 7.471264367816093e-07, + "loss": 1.216, + "step": 91 + }, + { + "epoch": 0.01, + "learning_rate": 7.55336617405583e-07, + "loss": 1.365, + "step": 92 + }, + { + "epoch": 0.01, + "learning_rate": 7.635467980295568e-07, + "loss": 1.2644, + "step": 93 + }, + { + "epoch": 0.01, + "learning_rate": 7.717569786535305e-07, + "loss": 1.2163, + "step": 94 + }, + { + "epoch": 0.01, + "learning_rate": 7.799671592775042e-07, + "loss": 1.2455, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 7.881773399014779e-07, + "loss": 1.3296, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 7.963875205254517e-07, + "loss": 1.2034, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 8.045977011494253e-07, + "loss": 1.3575, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 8.128078817733991e-07, + "loss": 1.2644, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 8.210180623973727e-07, + "loss": 1.2817, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 8.292282430213465e-07, + "loss": 1.158, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 8.374384236453203e-07, + "loss": 1.2557, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 8.456486042692939e-07, + "loss": 1.4313, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 8.538587848932678e-07, + "loss": 1.2218, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 8.620689655172415e-07, + "loss": 1.2374, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 8.702791461412152e-07, + "loss": 1.2176, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 8.78489326765189e-07, + "loss": 1.3144, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 8.866995073891626e-07, + "loss": 1.3128, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 8.949096880131364e-07, + "loss": 1.2492, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 9.031198686371101e-07, + "loss": 1.2451, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 9.113300492610838e-07, + "loss": 1.2775, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 9.195402298850575e-07, + "loss": 1.3226, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 9.277504105090313e-07, + "loss": 1.3012, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 9.359605911330049e-07, + "loss": 1.1834, + "step": 114 + }, + { + "epoch": 0.01, + "learning_rate": 9.441707717569787e-07, + "loss": 1.2362, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 9.523809523809525e-07, + "loss": 1.2615, + "step": 116 + }, + { + "epoch": 0.01, + "learning_rate": 9.605911330049262e-07, + "loss": 1.3468, + "step": 117 + }, + { + "epoch": 0.01, + "learning_rate": 9.688013136288999e-07, + "loss": 1.2032, + "step": 118 + }, + { + "epoch": 0.01, + "learning_rate": 9.770114942528738e-07, + "loss": 1.2958, + "step": 119 + }, + { + "epoch": 0.01, + "learning_rate": 9.852216748768474e-07, + "loss": 1.3524, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 9.93431855500821e-07, + "loss": 1.333, + "step": 121 + }, + { + "epoch": 0.01, + "learning_rate": 1.001642036124795e-06, + "loss": 1.2504, + "step": 122 + }, + { + "epoch": 0.01, + "learning_rate": 1.0098522167487686e-06, + "loss": 1.2535, + "step": 123 + }, + { + "epoch": 0.01, + "learning_rate": 1.0180623973727423e-06, + "loss": 1.2835, + "step": 124 + }, + { + "epoch": 0.01, + "learning_rate": 1.026272577996716e-06, + "loss": 1.2917, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 1.0344827586206898e-06, + "loss": 1.3416, + "step": 126 + }, + { + "epoch": 0.01, + "learning_rate": 1.0426929392446635e-06, + "loss": 1.3393, + "step": 127 + }, + { + "epoch": 0.01, + "learning_rate": 1.0509031198686371e-06, + "loss": 1.2399, + "step": 128 + }, + { + "epoch": 0.01, + "learning_rate": 1.059113300492611e-06, + "loss": 1.2399, + "step": 129 + }, + { + "epoch": 0.01, + "learning_rate": 1.0673234811165846e-06, + "loss": 1.2678, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 1.0755336617405583e-06, + "loss": 1.2024, + "step": 131 + }, + { + "epoch": 0.01, + "learning_rate": 1.0837438423645322e-06, + "loss": 1.2094, + "step": 132 + }, + { + "epoch": 0.01, + "learning_rate": 1.0919540229885058e-06, + "loss": 1.2965, + "step": 133 + }, + { + "epoch": 0.01, + "learning_rate": 1.1001642036124797e-06, + "loss": 1.3802, + "step": 134 + }, + { + "epoch": 0.01, + "learning_rate": 1.1083743842364534e-06, + "loss": 1.3311, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 1.116584564860427e-06, + "loss": 1.3136, + "step": 136 + }, + { + "epoch": 0.01, + "learning_rate": 1.1247947454844007e-06, + "loss": 1.2816, + "step": 137 + }, + { + "epoch": 0.01, + "learning_rate": 1.1330049261083746e-06, + "loss": 1.2369, + "step": 138 + }, + { + "epoch": 0.01, + "learning_rate": 1.1412151067323482e-06, + "loss": 1.3171, + "step": 139 + }, + { + "epoch": 0.01, + "learning_rate": 1.1494252873563219e-06, + "loss": 1.2337, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 1.1576354679802955e-06, + "loss": 1.2773, + "step": 141 + }, + { + "epoch": 0.01, + "learning_rate": 1.1658456486042694e-06, + "loss": 1.3736, + "step": 142 + }, + { + "epoch": 0.01, + "learning_rate": 1.174055829228243e-06, + "loss": 1.1971, + "step": 143 + }, + { + "epoch": 0.01, + "learning_rate": 1.1822660098522167e-06, + "loss": 1.2464, + "step": 144 + }, + { + "epoch": 0.01, + "learning_rate": 1.1904761904761906e-06, + "loss": 1.2851, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 1.1986863711001645e-06, + "loss": 1.2814, + "step": 146 + }, + { + "epoch": 0.01, + "learning_rate": 1.2068965517241381e-06, + "loss": 1.2441, + "step": 147 + }, + { + "epoch": 0.01, + "learning_rate": 1.2151067323481118e-06, + "loss": 1.1904, + "step": 148 + }, + { + "epoch": 0.01, + "learning_rate": 1.2233169129720854e-06, + "loss": 1.2877, + "step": 149 + }, + { + "epoch": 0.01, + "learning_rate": 1.2315270935960593e-06, + "loss": 1.3688, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 1.239737274220033e-06, + "loss": 1.3632, + "step": 151 + }, + { + "epoch": 0.01, + "learning_rate": 1.2479474548440066e-06, + "loss": 1.202, + "step": 152 + }, + { + "epoch": 0.01, + "learning_rate": 1.2561576354679805e-06, + "loss": 1.234, + "step": 153 + }, + { + "epoch": 0.01, + "learning_rate": 1.2643678160919542e-06, + "loss": 1.2715, + "step": 154 + }, + { + "epoch": 0.01, + "learning_rate": 1.272577996715928e-06, + "loss": 1.2097, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 1.2807881773399017e-06, + "loss": 1.2605, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 1.2889983579638754e-06, + "loss": 1.3171, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 1.297208538587849e-06, + "loss": 1.2133, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 1.3054187192118229e-06, + "loss": 1.1452, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 1.3136288998357965e-06, + "loss": 1.2238, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 1.3218390804597702e-06, + "loss": 1.3462, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 1.330049261083744e-06, + "loss": 1.262, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 1.3382594417077177e-06, + "loss": 1.2857, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 1.3464696223316914e-06, + "loss": 1.28, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 1.354679802955665e-06, + "loss": 1.2404, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 1.362889983579639e-06, + "loss": 1.2322, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 1.3711001642036126e-06, + "loss": 1.2051, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 1.3793103448275862e-06, + "loss": 1.267, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 1.3875205254515601e-06, + "loss": 1.3604, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 1.3957307060755338e-06, + "loss": 1.2656, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 1.4039408866995074e-06, + "loss": 1.2581, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 1.412151067323481e-06, + "loss": 1.3255, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 1.420361247947455e-06, + "loss": 1.2941, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 1.4285714285714286e-06, + "loss": 1.3556, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 1.4367816091954023e-06, + "loss": 1.1552, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 1.444991789819376e-06, + "loss": 1.208, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 1.4532019704433498e-06, + "loss": 1.2421, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 1.4614121510673235e-06, + "loss": 1.2265, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 1.4696223316912971e-06, + "loss": 1.1421, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 1.4778325123152712e-06, + "loss": 1.3112, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 1.4860426929392449e-06, + "loss": 1.2915, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 1.4942528735632185e-06, + "loss": 1.2922, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 1.5024630541871924e-06, + "loss": 1.2716, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 1.510673234811166e-06, + "loss": 1.259, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 1.5188834154351397e-06, + "loss": 1.333, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 1.5270935960591136e-06, + "loss": 1.3547, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 1.5353037766830873e-06, + "loss": 1.3425, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 1.543513957307061e-06, + "loss": 1.2619, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 1.5517241379310346e-06, + "loss": 1.2473, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 1.5599343185550085e-06, + "loss": 1.3747, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 1.5681444991789821e-06, + "loss": 1.2588, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 1.5763546798029558e-06, + "loss": 1.2804, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 1.5845648604269294e-06, + "loss": 1.3359, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 1.5927750410509033e-06, + "loss": 1.3584, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 1.600985221674877e-06, + "loss": 1.1715, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 1.6091954022988506e-06, + "loss": 1.2772, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 1.6174055829228245e-06, + "loss": 1.365, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 1.6256157635467982e-06, + "loss": 1.308, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 1.6338259441707718e-06, + "loss": 1.2293, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 1.6420361247947455e-06, + "loss": 1.1926, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 1.6502463054187193e-06, + "loss": 1.179, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 1.658456486042693e-06, + "loss": 1.2431, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 1.6666666666666667e-06, + "loss": 1.3027, + "step": 203 + }, + { + "epoch": 0.02, + "learning_rate": 1.6748768472906405e-06, + "loss": 1.2012, + "step": 204 + }, + { + "epoch": 0.02, + "learning_rate": 1.6830870279146142e-06, + "loss": 1.283, + "step": 205 + }, + { + "epoch": 0.02, + "learning_rate": 1.6912972085385879e-06, + "loss": 1.3206, + "step": 206 + }, + { + "epoch": 0.02, + "learning_rate": 1.6995073891625615e-06, + "loss": 1.3412, + "step": 207 + }, + { + "epoch": 0.02, + "learning_rate": 1.7077175697865356e-06, + "loss": 1.3156, + "step": 208 + }, + { + "epoch": 0.02, + "learning_rate": 1.7159277504105093e-06, + "loss": 1.3222, + "step": 209 + }, + { + "epoch": 0.02, + "learning_rate": 1.724137931034483e-06, + "loss": 1.2316, + "step": 210 + }, + { + "epoch": 0.02, + "learning_rate": 1.7323481116584568e-06, + "loss": 1.2065, + "step": 211 + }, + { + "epoch": 0.02, + "learning_rate": 1.7405582922824304e-06, + "loss": 1.2693, + "step": 212 + }, + { + "epoch": 0.02, + "learning_rate": 1.748768472906404e-06, + "loss": 1.2768, + "step": 213 + }, + { + "epoch": 0.02, + "learning_rate": 1.756978653530378e-06, + "loss": 1.3222, + "step": 214 + }, + { + "epoch": 0.02, + "learning_rate": 1.7651888341543516e-06, + "loss": 1.3052, + "step": 215 + }, + { + "epoch": 0.02, + "learning_rate": 1.7733990147783253e-06, + "loss": 1.1105, + "step": 216 + }, + { + "epoch": 0.02, + "learning_rate": 1.781609195402299e-06, + "loss": 1.2063, + "step": 217 + }, + { + "epoch": 0.02, + "learning_rate": 1.7898193760262728e-06, + "loss": 1.2637, + "step": 218 + }, + { + "epoch": 0.02, + "learning_rate": 1.7980295566502465e-06, + "loss": 1.2725, + "step": 219 + }, + { + "epoch": 0.02, + "learning_rate": 1.8062397372742201e-06, + "loss": 1.3958, + "step": 220 + }, + { + "epoch": 0.02, + "learning_rate": 1.814449917898194e-06, + "loss": 1.2354, + "step": 221 + }, + { + "epoch": 0.02, + "learning_rate": 1.8226600985221677e-06, + "loss": 1.1672, + "step": 222 + }, + { + "epoch": 0.02, + "learning_rate": 1.8308702791461413e-06, + "loss": 1.3864, + "step": 223 + }, + { + "epoch": 0.02, + "learning_rate": 1.839080459770115e-06, + "loss": 1.2985, + "step": 224 + }, + { + "epoch": 0.02, + "learning_rate": 1.8472906403940889e-06, + "loss": 1.2902, + "step": 225 + }, + { + "epoch": 0.02, + "learning_rate": 1.8555008210180625e-06, + "loss": 1.2534, + "step": 226 + }, + { + "epoch": 0.02, + "learning_rate": 1.8637110016420362e-06, + "loss": 1.1856, + "step": 227 + }, + { + "epoch": 0.02, + "learning_rate": 1.8719211822660098e-06, + "loss": 1.2558, + "step": 228 + }, + { + "epoch": 0.02, + "learning_rate": 1.8801313628899837e-06, + "loss": 1.3328, + "step": 229 + }, + { + "epoch": 0.02, + "learning_rate": 1.8883415435139574e-06, + "loss": 1.2827, + "step": 230 + }, + { + "epoch": 0.02, + "learning_rate": 1.896551724137931e-06, + "loss": 1.2387, + "step": 231 + }, + { + "epoch": 0.02, + "learning_rate": 1.904761904761905e-06, + "loss": 1.2304, + "step": 232 + }, + { + "epoch": 0.02, + "learning_rate": 1.9129720853858786e-06, + "loss": 1.1876, + "step": 233 + }, + { + "epoch": 0.02, + "learning_rate": 1.9211822660098524e-06, + "loss": 1.3143, + "step": 234 + }, + { + "epoch": 0.02, + "learning_rate": 1.929392446633826e-06, + "loss": 1.2963, + "step": 235 + }, + { + "epoch": 0.02, + "learning_rate": 1.9376026272577998e-06, + "loss": 1.3332, + "step": 236 + }, + { + "epoch": 0.02, + "learning_rate": 1.9458128078817736e-06, + "loss": 1.3246, + "step": 237 + }, + { + "epoch": 0.02, + "learning_rate": 1.9540229885057475e-06, + "loss": 1.3109, + "step": 238 + }, + { + "epoch": 0.02, + "learning_rate": 1.962233169129721e-06, + "loss": 1.1815, + "step": 239 + }, + { + "epoch": 0.02, + "learning_rate": 1.970443349753695e-06, + "loss": 1.2378, + "step": 240 + }, + { + "epoch": 0.02, + "learning_rate": 1.9786535303776687e-06, + "loss": 1.3524, + "step": 241 + }, + { + "epoch": 0.02, + "learning_rate": 1.986863711001642e-06, + "loss": 1.1702, + "step": 242 + }, + { + "epoch": 0.02, + "learning_rate": 1.995073891625616e-06, + "loss": 1.1918, + "step": 243 + }, + { + "epoch": 0.02, + "learning_rate": 2.00328407224959e-06, + "loss": 1.1794, + "step": 244 + }, + { + "epoch": 0.02, + "learning_rate": 2.0114942528735633e-06, + "loss": 1.1773, + "step": 245 + }, + { + "epoch": 0.02, + "learning_rate": 2.019704433497537e-06, + "loss": 1.288, + "step": 246 + }, + { + "epoch": 0.02, + "learning_rate": 2.027914614121511e-06, + "loss": 1.1907, + "step": 247 + }, + { + "epoch": 0.02, + "learning_rate": 2.0361247947454845e-06, + "loss": 1.1531, + "step": 248 + }, + { + "epoch": 0.02, + "learning_rate": 2.0443349753694584e-06, + "loss": 1.307, + "step": 249 + }, + { + "epoch": 0.02, + "learning_rate": 2.052545155993432e-06, + "loss": 1.2503, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 2.0607553366174057e-06, + "loss": 1.2547, + "step": 251 + }, + { + "epoch": 0.02, + "learning_rate": 2.0689655172413796e-06, + "loss": 1.1793, + "step": 252 + }, + { + "epoch": 0.02, + "learning_rate": 2.077175697865353e-06, + "loss": 1.2816, + "step": 253 + }, + { + "epoch": 0.02, + "learning_rate": 2.085385878489327e-06, + "loss": 1.3486, + "step": 254 + }, + { + "epoch": 0.02, + "learning_rate": 2.0935960591133008e-06, + "loss": 1.2942, + "step": 255 + }, + { + "epoch": 0.02, + "learning_rate": 2.1018062397372742e-06, + "loss": 1.2775, + "step": 256 + }, + { + "epoch": 0.02, + "learning_rate": 2.110016420361248e-06, + "loss": 1.2204, + "step": 257 + }, + { + "epoch": 0.02, + "learning_rate": 2.118226600985222e-06, + "loss": 1.2023, + "step": 258 + }, + { + "epoch": 0.02, + "learning_rate": 2.1264367816091954e-06, + "loss": 1.244, + "step": 259 + }, + { + "epoch": 0.02, + "learning_rate": 2.1346469622331693e-06, + "loss": 1.3044, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 2.1428571428571427e-06, + "loss": 1.3361, + "step": 261 + }, + { + "epoch": 0.02, + "learning_rate": 2.1510673234811166e-06, + "loss": 1.3175, + "step": 262 + }, + { + "epoch": 0.02, + "learning_rate": 2.1592775041050905e-06, + "loss": 1.3413, + "step": 263 + }, + { + "epoch": 0.02, + "learning_rate": 2.1674876847290643e-06, + "loss": 1.2898, + "step": 264 + }, + { + "epoch": 0.02, + "learning_rate": 2.1756978653530382e-06, + "loss": 1.3147, + "step": 265 + }, + { + "epoch": 0.02, + "learning_rate": 2.1839080459770117e-06, + "loss": 1.414, + "step": 266 + }, + { + "epoch": 0.02, + "learning_rate": 2.1921182266009855e-06, + "loss": 1.2005, + "step": 267 + }, + { + "epoch": 0.02, + "learning_rate": 2.2003284072249594e-06, + "loss": 1.2587, + "step": 268 + }, + { + "epoch": 0.02, + "learning_rate": 2.208538587848933e-06, + "loss": 1.3172, + "step": 269 + }, + { + "epoch": 0.02, + "learning_rate": 2.2167487684729067e-06, + "loss": 1.2294, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 2.22495894909688e-06, + "loss": 1.2437, + "step": 271 + }, + { + "epoch": 0.02, + "learning_rate": 2.233169129720854e-06, + "loss": 1.2525, + "step": 272 + }, + { + "epoch": 0.02, + "learning_rate": 2.241379310344828e-06, + "loss": 1.3093, + "step": 273 + }, + { + "epoch": 0.02, + "learning_rate": 2.2495894909688014e-06, + "loss": 1.3148, + "step": 274 + }, + { + "epoch": 0.02, + "learning_rate": 2.2577996715927752e-06, + "loss": 1.3188, + "step": 275 + }, + { + "epoch": 0.02, + "learning_rate": 2.266009852216749e-06, + "loss": 1.3171, + "step": 276 + }, + { + "epoch": 0.02, + "learning_rate": 2.2742200328407226e-06, + "loss": 1.2035, + "step": 277 + }, + { + "epoch": 0.02, + "learning_rate": 2.2824302134646964e-06, + "loss": 1.3175, + "step": 278 + }, + { + "epoch": 0.02, + "learning_rate": 2.2906403940886703e-06, + "loss": 1.3284, + "step": 279 + }, + { + "epoch": 0.02, + "learning_rate": 2.2988505747126437e-06, + "loss": 1.2924, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 2.3070607553366176e-06, + "loss": 1.2502, + "step": 281 + }, + { + "epoch": 0.02, + "learning_rate": 2.315270935960591e-06, + "loss": 1.3753, + "step": 282 + }, + { + "epoch": 0.02, + "learning_rate": 2.323481116584565e-06, + "loss": 1.2256, + "step": 283 + }, + { + "epoch": 0.02, + "learning_rate": 2.331691297208539e-06, + "loss": 1.3569, + "step": 284 + }, + { + "epoch": 0.02, + "learning_rate": 2.3399014778325123e-06, + "loss": 1.1733, + "step": 285 + }, + { + "epoch": 0.02, + "learning_rate": 2.348111658456486e-06, + "loss": 1.2439, + "step": 286 + }, + { + "epoch": 0.02, + "learning_rate": 2.35632183908046e-06, + "loss": 1.3348, + "step": 287 + }, + { + "epoch": 0.02, + "learning_rate": 2.3645320197044334e-06, + "loss": 1.2643, + "step": 288 + }, + { + "epoch": 0.02, + "learning_rate": 2.3727422003284073e-06, + "loss": 1.2491, + "step": 289 + }, + { + "epoch": 0.02, + "learning_rate": 2.380952380952381e-06, + "loss": 1.306, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 2.389162561576355e-06, + "loss": 1.2957, + "step": 291 + }, + { + "epoch": 0.02, + "learning_rate": 2.397372742200329e-06, + "loss": 1.2229, + "step": 292 + }, + { + "epoch": 0.02, + "learning_rate": 2.4055829228243024e-06, + "loss": 1.2346, + "step": 293 + }, + { + "epoch": 0.02, + "learning_rate": 2.4137931034482762e-06, + "loss": 1.3785, + "step": 294 + }, + { + "epoch": 0.02, + "learning_rate": 2.4220032840722497e-06, + "loss": 1.3816, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 2.4302134646962236e-06, + "loss": 1.2245, + "step": 296 + }, + { + "epoch": 0.02, + "learning_rate": 2.4384236453201974e-06, + "loss": 1.3896, + "step": 297 + }, + { + "epoch": 0.02, + "learning_rate": 2.446633825944171e-06, + "loss": 1.3583, + "step": 298 + }, + { + "epoch": 0.02, + "learning_rate": 2.4548440065681448e-06, + "loss": 1.1827, + "step": 299 + }, + { + "epoch": 0.02, + "learning_rate": 2.4630541871921186e-06, + "loss": 1.2428, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 2.471264367816092e-06, + "loss": 1.2054, + "step": 301 + }, + { + "epoch": 0.02, + "learning_rate": 2.479474548440066e-06, + "loss": 1.3668, + "step": 302 + }, + { + "epoch": 0.02, + "learning_rate": 2.48768472906404e-06, + "loss": 1.404, + "step": 303 + }, + { + "epoch": 0.02, + "learning_rate": 2.4958949096880133e-06, + "loss": 1.246, + "step": 304 + }, + { + "epoch": 0.02, + "learning_rate": 2.504105090311987e-06, + "loss": 1.3481, + "step": 305 + }, + { + "epoch": 0.02, + "learning_rate": 2.512315270935961e-06, + "loss": 1.3628, + "step": 306 + }, + { + "epoch": 0.02, + "learning_rate": 2.520525451559935e-06, + "loss": 1.2258, + "step": 307 + }, + { + "epoch": 0.02, + "learning_rate": 2.5287356321839083e-06, + "loss": 1.1824, + "step": 308 + }, + { + "epoch": 0.02, + "learning_rate": 2.536945812807882e-06, + "loss": 1.2794, + "step": 309 + }, + { + "epoch": 0.02, + "learning_rate": 2.545155993431856e-06, + "loss": 1.3101, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 2.5533661740558295e-06, + "loss": 1.2866, + "step": 311 + }, + { + "epoch": 0.02, + "learning_rate": 2.5615763546798034e-06, + "loss": 1.2489, + "step": 312 + }, + { + "epoch": 0.02, + "learning_rate": 2.5697865353037773e-06, + "loss": 1.2447, + "step": 313 + }, + { + "epoch": 0.02, + "learning_rate": 2.5779967159277507e-06, + "loss": 1.1101, + "step": 314 + }, + { + "epoch": 0.02, + "learning_rate": 2.5862068965517246e-06, + "loss": 1.2841, + "step": 315 + }, + { + "epoch": 0.02, + "learning_rate": 2.594417077175698e-06, + "loss": 1.3372, + "step": 316 + }, + { + "epoch": 0.02, + "learning_rate": 2.602627257799672e-06, + "loss": 1.2625, + "step": 317 + }, + { + "epoch": 0.02, + "learning_rate": 2.6108374384236458e-06, + "loss": 1.2411, + "step": 318 + }, + { + "epoch": 0.02, + "learning_rate": 2.6190476190476192e-06, + "loss": 1.4066, + "step": 319 + }, + { + "epoch": 0.02, + "learning_rate": 2.627257799671593e-06, + "loss": 1.2385, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 2.635467980295567e-06, + "loss": 1.2851, + "step": 321 + }, + { + "epoch": 0.02, + "learning_rate": 2.6436781609195404e-06, + "loss": 1.3249, + "step": 322 + }, + { + "epoch": 0.02, + "learning_rate": 2.6518883415435143e-06, + "loss": 1.2782, + "step": 323 + }, + { + "epoch": 0.02, + "learning_rate": 2.660098522167488e-06, + "loss": 1.2341, + "step": 324 + }, + { + "epoch": 0.02, + "learning_rate": 2.6683087027914616e-06, + "loss": 1.3008, + "step": 325 + }, + { + "epoch": 0.02, + "learning_rate": 2.6765188834154355e-06, + "loss": 1.2633, + "step": 326 + }, + { + "epoch": 0.02, + "learning_rate": 2.6847290640394093e-06, + "loss": 1.2854, + "step": 327 + }, + { + "epoch": 0.02, + "learning_rate": 2.692939244663383e-06, + "loss": 1.3673, + "step": 328 + }, + { + "epoch": 0.02, + "learning_rate": 2.7011494252873567e-06, + "loss": 1.2632, + "step": 329 + }, + { + "epoch": 0.02, + "learning_rate": 2.70935960591133e-06, + "loss": 1.1831, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 2.717569786535304e-06, + "loss": 1.2174, + "step": 331 + }, + { + "epoch": 0.02, + "learning_rate": 2.725779967159278e-06, + "loss": 1.2711, + "step": 332 + }, + { + "epoch": 0.02, + "learning_rate": 2.7339901477832513e-06, + "loss": 1.2603, + "step": 333 + }, + { + "epoch": 0.02, + "learning_rate": 2.742200328407225e-06, + "loss": 1.2662, + "step": 334 + }, + { + "epoch": 0.02, + "learning_rate": 2.750410509031199e-06, + "loss": 1.2732, + "step": 335 + }, + { + "epoch": 0.02, + "learning_rate": 2.7586206896551725e-06, + "loss": 1.1307, + "step": 336 + }, + { + "epoch": 0.02, + "learning_rate": 2.7668308702791464e-06, + "loss": 1.2322, + "step": 337 + }, + { + "epoch": 0.02, + "learning_rate": 2.7750410509031202e-06, + "loss": 1.2303, + "step": 338 + }, + { + "epoch": 0.03, + "learning_rate": 2.7832512315270937e-06, + "loss": 1.3112, + "step": 339 + }, + { + "epoch": 0.03, + "learning_rate": 2.7914614121510676e-06, + "loss": 1.2756, + "step": 340 + }, + { + "epoch": 0.03, + "learning_rate": 2.799671592775041e-06, + "loss": 1.3191, + "step": 341 + }, + { + "epoch": 0.03, + "learning_rate": 2.807881773399015e-06, + "loss": 1.228, + "step": 342 + }, + { + "epoch": 0.03, + "learning_rate": 2.8160919540229887e-06, + "loss": 1.3545, + "step": 343 + }, + { + "epoch": 0.03, + "learning_rate": 2.824302134646962e-06, + "loss": 1.3904, + "step": 344 + }, + { + "epoch": 0.03, + "learning_rate": 2.832512315270936e-06, + "loss": 1.2923, + "step": 345 + }, + { + "epoch": 0.03, + "learning_rate": 2.84072249589491e-06, + "loss": 1.3365, + "step": 346 + }, + { + "epoch": 0.03, + "learning_rate": 2.8489326765188834e-06, + "loss": 1.2877, + "step": 347 + }, + { + "epoch": 0.03, + "learning_rate": 2.8571428571428573e-06, + "loss": 1.3579, + "step": 348 + }, + { + "epoch": 0.03, + "learning_rate": 2.865353037766831e-06, + "loss": 1.2426, + "step": 349 + }, + { + "epoch": 0.03, + "learning_rate": 2.8735632183908046e-06, + "loss": 1.242, + "step": 350 + }, + { + "epoch": 0.03, + "learning_rate": 2.8817733990147784e-06, + "loss": 1.2981, + "step": 351 + }, + { + "epoch": 0.03, + "learning_rate": 2.889983579638752e-06, + "loss": 1.3173, + "step": 352 + }, + { + "epoch": 0.03, + "learning_rate": 2.8981937602627258e-06, + "loss": 1.1709, + "step": 353 + }, + { + "epoch": 0.03, + "learning_rate": 2.9064039408866996e-06, + "loss": 1.1909, + "step": 354 + }, + { + "epoch": 0.03, + "learning_rate": 2.914614121510673e-06, + "loss": 1.2903, + "step": 355 + }, + { + "epoch": 0.03, + "learning_rate": 2.922824302134647e-06, + "loss": 1.2581, + "step": 356 + }, + { + "epoch": 0.03, + "learning_rate": 2.931034482758621e-06, + "loss": 1.1986, + "step": 357 + }, + { + "epoch": 0.03, + "learning_rate": 2.9392446633825943e-06, + "loss": 1.1251, + "step": 358 + }, + { + "epoch": 0.03, + "learning_rate": 2.947454844006568e-06, + "loss": 1.2779, + "step": 359 + }, + { + "epoch": 0.03, + "learning_rate": 2.9556650246305424e-06, + "loss": 1.2589, + "step": 360 + }, + { + "epoch": 0.03, + "learning_rate": 2.9638752052545163e-06, + "loss": 1.2414, + "step": 361 + }, + { + "epoch": 0.03, + "learning_rate": 2.9720853858784898e-06, + "loss": 1.2064, + "step": 362 + }, + { + "epoch": 0.03, + "learning_rate": 2.9802955665024636e-06, + "loss": 1.2482, + "step": 363 + }, + { + "epoch": 0.03, + "learning_rate": 2.988505747126437e-06, + "loss": 1.2142, + "step": 364 + }, + { + "epoch": 0.03, + "learning_rate": 2.996715927750411e-06, + "loss": 1.246, + "step": 365 + }, + { + "epoch": 0.03, + "learning_rate": 3.004926108374385e-06, + "loss": 1.1813, + "step": 366 + }, + { + "epoch": 0.03, + "learning_rate": 3.0131362889983583e-06, + "loss": 1.2379, + "step": 367 + }, + { + "epoch": 0.03, + "learning_rate": 3.021346469622332e-06, + "loss": 1.1951, + "step": 368 + }, + { + "epoch": 0.03, + "learning_rate": 3.029556650246306e-06, + "loss": 1.2033, + "step": 369 + }, + { + "epoch": 0.03, + "learning_rate": 3.0377668308702795e-06, + "loss": 1.3003, + "step": 370 + }, + { + "epoch": 0.03, + "learning_rate": 3.0459770114942533e-06, + "loss": 1.2641, + "step": 371 + }, + { + "epoch": 0.03, + "learning_rate": 3.054187192118227e-06, + "loss": 1.2042, + "step": 372 + }, + { + "epoch": 0.03, + "learning_rate": 3.0623973727422007e-06, + "loss": 1.2357, + "step": 373 + }, + { + "epoch": 0.03, + "learning_rate": 3.0706075533661745e-06, + "loss": 1.2226, + "step": 374 + }, + { + "epoch": 0.03, + "learning_rate": 3.078817733990148e-06, + "loss": 1.2014, + "step": 375 + }, + { + "epoch": 0.03, + "learning_rate": 3.087027914614122e-06, + "loss": 1.2833, + "step": 376 + }, + { + "epoch": 0.03, + "learning_rate": 3.0952380952380957e-06, + "loss": 1.2819, + "step": 377 + }, + { + "epoch": 0.03, + "learning_rate": 3.103448275862069e-06, + "loss": 1.1529, + "step": 378 + }, + { + "epoch": 0.03, + "learning_rate": 3.111658456486043e-06, + "loss": 1.2078, + "step": 379 + }, + { + "epoch": 0.03, + "learning_rate": 3.119868637110017e-06, + "loss": 1.265, + "step": 380 + }, + { + "epoch": 0.03, + "learning_rate": 3.1280788177339904e-06, + "loss": 1.2393, + "step": 381 + }, + { + "epoch": 0.03, + "learning_rate": 3.1362889983579642e-06, + "loss": 1.1758, + "step": 382 + }, + { + "epoch": 0.03, + "learning_rate": 3.144499178981938e-06, + "loss": 1.2017, + "step": 383 + }, + { + "epoch": 0.03, + "learning_rate": 3.1527093596059115e-06, + "loss": 1.14, + "step": 384 + }, + { + "epoch": 0.03, + "learning_rate": 3.1609195402298854e-06, + "loss": 1.1271, + "step": 385 + }, + { + "epoch": 0.03, + "learning_rate": 3.169129720853859e-06, + "loss": 1.2382, + "step": 386 + }, + { + "epoch": 0.03, + "learning_rate": 3.1773399014778327e-06, + "loss": 1.2349, + "step": 387 + }, + { + "epoch": 0.03, + "learning_rate": 3.1855500821018066e-06, + "loss": 1.2054, + "step": 388 + }, + { + "epoch": 0.03, + "learning_rate": 3.19376026272578e-06, + "loss": 1.217, + "step": 389 + }, + { + "epoch": 0.03, + "learning_rate": 3.201970443349754e-06, + "loss": 1.2728, + "step": 390 + }, + { + "epoch": 0.03, + "learning_rate": 3.210180623973728e-06, + "loss": 1.2557, + "step": 391 + }, + { + "epoch": 0.03, + "learning_rate": 3.2183908045977012e-06, + "loss": 1.1953, + "step": 392 + }, + { + "epoch": 0.03, + "learning_rate": 3.226600985221675e-06, + "loss": 1.2625, + "step": 393 + }, + { + "epoch": 0.03, + "learning_rate": 3.234811165845649e-06, + "loss": 1.204, + "step": 394 + }, + { + "epoch": 0.03, + "learning_rate": 3.2430213464696224e-06, + "loss": 1.3863, + "step": 395 + }, + { + "epoch": 0.03, + "learning_rate": 3.2512315270935963e-06, + "loss": 1.307, + "step": 396 + }, + { + "epoch": 0.03, + "learning_rate": 3.2594417077175698e-06, + "loss": 1.237, + "step": 397 + }, + { + "epoch": 0.03, + "learning_rate": 3.2676518883415436e-06, + "loss": 1.1927, + "step": 398 + }, + { + "epoch": 0.03, + "learning_rate": 3.2758620689655175e-06, + "loss": 1.1447, + "step": 399 + }, + { + "epoch": 0.03, + "learning_rate": 3.284072249589491e-06, + "loss": 1.2176, + "step": 400 + }, + { + "epoch": 0.03, + "learning_rate": 3.292282430213465e-06, + "loss": 1.2265, + "step": 401 + }, + { + "epoch": 0.03, + "learning_rate": 3.3004926108374387e-06, + "loss": 1.2932, + "step": 402 + }, + { + "epoch": 0.03, + "learning_rate": 3.308702791461412e-06, + "loss": 1.2606, + "step": 403 + }, + { + "epoch": 0.03, + "learning_rate": 3.316912972085386e-06, + "loss": 1.2985, + "step": 404 + }, + { + "epoch": 0.03, + "learning_rate": 3.32512315270936e-06, + "loss": 1.3042, + "step": 405 + }, + { + "epoch": 0.03, + "learning_rate": 3.3333333333333333e-06, + "loss": 1.2885, + "step": 406 + }, + { + "epoch": 0.03, + "learning_rate": 3.341543513957307e-06, + "loss": 1.2304, + "step": 407 + }, + { + "epoch": 0.03, + "learning_rate": 3.349753694581281e-06, + "loss": 1.2168, + "step": 408 + }, + { + "epoch": 0.03, + "learning_rate": 3.3579638752052545e-06, + "loss": 1.2539, + "step": 409 + }, + { + "epoch": 0.03, + "learning_rate": 3.3661740558292284e-06, + "loss": 1.2695, + "step": 410 + }, + { + "epoch": 0.03, + "learning_rate": 3.374384236453202e-06, + "loss": 1.2874, + "step": 411 + }, + { + "epoch": 0.03, + "learning_rate": 3.3825944170771757e-06, + "loss": 1.203, + "step": 412 + }, + { + "epoch": 0.03, + "learning_rate": 3.3908045977011496e-06, + "loss": 1.2815, + "step": 413 + }, + { + "epoch": 0.03, + "learning_rate": 3.399014778325123e-06, + "loss": 1.1912, + "step": 414 + }, + { + "epoch": 0.03, + "learning_rate": 3.407224958949097e-06, + "loss": 1.2655, + "step": 415 + }, + { + "epoch": 0.03, + "learning_rate": 3.415435139573071e-06, + "loss": 1.1781, + "step": 416 + }, + { + "epoch": 0.03, + "learning_rate": 3.423645320197045e-06, + "loss": 1.249, + "step": 417 + }, + { + "epoch": 0.03, + "learning_rate": 3.4318555008210185e-06, + "loss": 1.2945, + "step": 418 + }, + { + "epoch": 0.03, + "learning_rate": 3.4400656814449924e-06, + "loss": 1.0713, + "step": 419 + }, + { + "epoch": 0.03, + "learning_rate": 3.448275862068966e-06, + "loss": 1.2918, + "step": 420 + }, + { + "epoch": 0.03, + "learning_rate": 3.4564860426929397e-06, + "loss": 1.2521, + "step": 421 + }, + { + "epoch": 0.03, + "learning_rate": 3.4646962233169136e-06, + "loss": 1.2802, + "step": 422 + }, + { + "epoch": 0.03, + "learning_rate": 3.472906403940887e-06, + "loss": 1.3591, + "step": 423 + }, + { + "epoch": 0.03, + "learning_rate": 3.481116584564861e-06, + "loss": 1.1951, + "step": 424 + }, + { + "epoch": 0.03, + "learning_rate": 3.4893267651888348e-06, + "loss": 1.2077, + "step": 425 + }, + { + "epoch": 0.03, + "learning_rate": 3.497536945812808e-06, + "loss": 1.1853, + "step": 426 + }, + { + "epoch": 0.03, + "learning_rate": 3.505747126436782e-06, + "loss": 1.2973, + "step": 427 + }, + { + "epoch": 0.03, + "learning_rate": 3.513957307060756e-06, + "loss": 1.3778, + "step": 428 + }, + { + "epoch": 0.03, + "learning_rate": 3.5221674876847294e-06, + "loss": 1.2883, + "step": 429 + }, + { + "epoch": 0.03, + "learning_rate": 3.5303776683087033e-06, + "loss": 1.2062, + "step": 430 + }, + { + "epoch": 0.03, + "learning_rate": 3.5385878489326767e-06, + "loss": 1.3154, + "step": 431 + }, + { + "epoch": 0.03, + "learning_rate": 3.5467980295566506e-06, + "loss": 1.3021, + "step": 432 + }, + { + "epoch": 0.03, + "learning_rate": 3.5550082101806245e-06, + "loss": 1.2791, + "step": 433 + }, + { + "epoch": 0.03, + "learning_rate": 3.563218390804598e-06, + "loss": 1.1884, + "step": 434 + }, + { + "epoch": 0.03, + "learning_rate": 3.5714285714285718e-06, + "loss": 1.1701, + "step": 435 + }, + { + "epoch": 0.03, + "learning_rate": 3.5796387520525457e-06, + "loss": 1.2855, + "step": 436 + }, + { + "epoch": 0.03, + "learning_rate": 3.587848932676519e-06, + "loss": 1.3727, + "step": 437 + }, + { + "epoch": 0.03, + "learning_rate": 3.596059113300493e-06, + "loss": 1.2655, + "step": 438 + }, + { + "epoch": 0.03, + "learning_rate": 3.604269293924467e-06, + "loss": 1.2005, + "step": 439 + }, + { + "epoch": 0.03, + "learning_rate": 3.6124794745484403e-06, + "loss": 1.0957, + "step": 440 + }, + { + "epoch": 0.03, + "learning_rate": 3.620689655172414e-06, + "loss": 1.1557, + "step": 441 + }, + { + "epoch": 0.03, + "learning_rate": 3.628899835796388e-06, + "loss": 1.2331, + "step": 442 + }, + { + "epoch": 0.03, + "learning_rate": 3.6371100164203615e-06, + "loss": 1.1073, + "step": 443 + }, + { + "epoch": 0.03, + "learning_rate": 3.6453201970443354e-06, + "loss": 1.2782, + "step": 444 + }, + { + "epoch": 0.03, + "learning_rate": 3.653530377668309e-06, + "loss": 1.1832, + "step": 445 + }, + { + "epoch": 0.03, + "learning_rate": 3.6617405582922827e-06, + "loss": 1.2152, + "step": 446 + }, + { + "epoch": 0.03, + "learning_rate": 3.6699507389162565e-06, + "loss": 1.2696, + "step": 447 + }, + { + "epoch": 0.03, + "learning_rate": 3.67816091954023e-06, + "loss": 1.2309, + "step": 448 + }, + { + "epoch": 0.03, + "learning_rate": 3.686371100164204e-06, + "loss": 1.2315, + "step": 449 + }, + { + "epoch": 0.03, + "learning_rate": 3.6945812807881777e-06, + "loss": 1.2795, + "step": 450 + }, + { + "epoch": 0.03, + "learning_rate": 3.702791461412151e-06, + "loss": 1.2727, + "step": 451 + }, + { + "epoch": 0.03, + "learning_rate": 3.711001642036125e-06, + "loss": 1.2914, + "step": 452 + }, + { + "epoch": 0.03, + "learning_rate": 3.719211822660099e-06, + "loss": 1.2741, + "step": 453 + }, + { + "epoch": 0.03, + "learning_rate": 3.7274220032840724e-06, + "loss": 1.245, + "step": 454 + }, + { + "epoch": 0.03, + "learning_rate": 3.7356321839080462e-06, + "loss": 1.2044, + "step": 455 + }, + { + "epoch": 0.03, + "learning_rate": 3.7438423645320197e-06, + "loss": 1.2289, + "step": 456 + }, + { + "epoch": 0.03, + "learning_rate": 3.7520525451559936e-06, + "loss": 1.2344, + "step": 457 + }, + { + "epoch": 0.03, + "learning_rate": 3.7602627257799674e-06, + "loss": 1.296, + "step": 458 + }, + { + "epoch": 0.03, + "learning_rate": 3.768472906403941e-06, + "loss": 1.1916, + "step": 459 + }, + { + "epoch": 0.03, + "learning_rate": 3.7766830870279148e-06, + "loss": 1.2825, + "step": 460 + }, + { + "epoch": 0.03, + "learning_rate": 3.7848932676518886e-06, + "loss": 1.2068, + "step": 461 + }, + { + "epoch": 0.03, + "learning_rate": 3.793103448275862e-06, + "loss": 1.1406, + "step": 462 + }, + { + "epoch": 0.03, + "learning_rate": 3.801313628899836e-06, + "loss": 1.2264, + "step": 463 + }, + { + "epoch": 0.03, + "learning_rate": 3.80952380952381e-06, + "loss": 1.3136, + "step": 464 + }, + { + "epoch": 0.03, + "learning_rate": 3.817733990147783e-06, + "loss": 1.1595, + "step": 465 + }, + { + "epoch": 0.03, + "learning_rate": 3.825944170771757e-06, + "loss": 1.1269, + "step": 466 + }, + { + "epoch": 0.03, + "learning_rate": 3.834154351395731e-06, + "loss": 1.2741, + "step": 467 + }, + { + "epoch": 0.03, + "learning_rate": 3.842364532019705e-06, + "loss": 1.1542, + "step": 468 + }, + { + "epoch": 0.03, + "learning_rate": 3.850574712643678e-06, + "loss": 1.305, + "step": 469 + }, + { + "epoch": 0.03, + "learning_rate": 3.858784893267652e-06, + "loss": 1.1878, + "step": 470 + }, + { + "epoch": 0.03, + "learning_rate": 3.8669950738916265e-06, + "loss": 1.2284, + "step": 471 + }, + { + "epoch": 0.03, + "learning_rate": 3.8752052545155995e-06, + "loss": 1.2325, + "step": 472 + }, + { + "epoch": 0.03, + "learning_rate": 3.883415435139573e-06, + "loss": 1.3406, + "step": 473 + }, + { + "epoch": 0.04, + "learning_rate": 3.891625615763547e-06, + "loss": 1.2311, + "step": 474 + }, + { + "epoch": 0.04, + "learning_rate": 3.899835796387521e-06, + "loss": 1.2478, + "step": 475 + }, + { + "epoch": 0.04, + "learning_rate": 3.908045977011495e-06, + "loss": 1.2525, + "step": 476 + }, + { + "epoch": 0.04, + "learning_rate": 3.916256157635469e-06, + "loss": 1.2208, + "step": 477 + }, + { + "epoch": 0.04, + "learning_rate": 3.924466338259442e-06, + "loss": 1.2962, + "step": 478 + }, + { + "epoch": 0.04, + "learning_rate": 3.932676518883416e-06, + "loss": 1.1932, + "step": 479 + }, + { + "epoch": 0.04, + "learning_rate": 3.94088669950739e-06, + "loss": 1.2417, + "step": 480 + }, + { + "epoch": 0.04, + "learning_rate": 3.9490968801313635e-06, + "loss": 1.2575, + "step": 481 + }, + { + "epoch": 0.04, + "learning_rate": 3.957307060755337e-06, + "loss": 1.2642, + "step": 482 + }, + { + "epoch": 0.04, + "learning_rate": 3.96551724137931e-06, + "loss": 1.1981, + "step": 483 + }, + { + "epoch": 0.04, + "learning_rate": 3.973727422003284e-06, + "loss": 1.2485, + "step": 484 + }, + { + "epoch": 0.04, + "learning_rate": 3.981937602627258e-06, + "loss": 1.2226, + "step": 485 + }, + { + "epoch": 0.04, + "learning_rate": 3.990147783251232e-06, + "loss": 1.1902, + "step": 486 + }, + { + "epoch": 0.04, + "learning_rate": 3.998357963875206e-06, + "loss": 1.2042, + "step": 487 + }, + { + "epoch": 0.04, + "learning_rate": 4.00656814449918e-06, + "loss": 1.2186, + "step": 488 + }, + { + "epoch": 0.04, + "learning_rate": 4.014778325123153e-06, + "loss": 1.3268, + "step": 489 + }, + { + "epoch": 0.04, + "learning_rate": 4.022988505747127e-06, + "loss": 1.179, + "step": 490 + }, + { + "epoch": 0.04, + "learning_rate": 4.0311986863711005e-06, + "loss": 1.1437, + "step": 491 + }, + { + "epoch": 0.04, + "learning_rate": 4.039408866995074e-06, + "loss": 1.3347, + "step": 492 + }, + { + "epoch": 0.04, + "learning_rate": 4.047619047619048e-06, + "loss": 1.1571, + "step": 493 + }, + { + "epoch": 0.04, + "learning_rate": 4.055829228243022e-06, + "loss": 1.2293, + "step": 494 + }, + { + "epoch": 0.04, + "learning_rate": 4.064039408866995e-06, + "loss": 1.1575, + "step": 495 + }, + { + "epoch": 0.04, + "learning_rate": 4.072249589490969e-06, + "loss": 1.2581, + "step": 496 + }, + { + "epoch": 0.04, + "learning_rate": 4.080459770114943e-06, + "loss": 1.4065, + "step": 497 + }, + { + "epoch": 0.04, + "learning_rate": 4.088669950738917e-06, + "loss": 1.283, + "step": 498 + }, + { + "epoch": 0.04, + "learning_rate": 4.096880131362891e-06, + "loss": 1.2343, + "step": 499 + }, + { + "epoch": 0.04, + "learning_rate": 4.105090311986864e-06, + "loss": 1.2019, + "step": 500 + }, + { + "epoch": 0.04, + "learning_rate": 4.1133004926108375e-06, + "loss": 1.2715, + "step": 501 + }, + { + "epoch": 0.04, + "learning_rate": 4.121510673234811e-06, + "loss": 1.1947, + "step": 502 + }, + { + "epoch": 0.04, + "learning_rate": 4.129720853858785e-06, + "loss": 1.1493, + "step": 503 + }, + { + "epoch": 0.04, + "learning_rate": 4.137931034482759e-06, + "loss": 1.2206, + "step": 504 + }, + { + "epoch": 0.04, + "learning_rate": 4.146141215106733e-06, + "loss": 1.2394, + "step": 505 + }, + { + "epoch": 0.04, + "learning_rate": 4.154351395730706e-06, + "loss": 1.2452, + "step": 506 + }, + { + "epoch": 0.04, + "learning_rate": 4.16256157635468e-06, + "loss": 1.2828, + "step": 507 + }, + { + "epoch": 0.04, + "learning_rate": 4.170771756978654e-06, + "loss": 1.2641, + "step": 508 + }, + { + "epoch": 0.04, + "learning_rate": 4.178981937602628e-06, + "loss": 1.2013, + "step": 509 + }, + { + "epoch": 0.04, + "learning_rate": 4.1871921182266015e-06, + "loss": 1.1758, + "step": 510 + }, + { + "epoch": 0.04, + "learning_rate": 4.1954022988505746e-06, + "loss": 1.1663, + "step": 511 + }, + { + "epoch": 0.04, + "learning_rate": 4.2036124794745484e-06, + "loss": 1.2322, + "step": 512 + }, + { + "epoch": 0.04, + "learning_rate": 4.211822660098522e-06, + "loss": 1.1998, + "step": 513 + }, + { + "epoch": 0.04, + "learning_rate": 4.220032840722496e-06, + "loss": 1.2359, + "step": 514 + }, + { + "epoch": 0.04, + "learning_rate": 4.22824302134647e-06, + "loss": 1.2541, + "step": 515 + }, + { + "epoch": 0.04, + "learning_rate": 4.236453201970444e-06, + "loss": 1.2722, + "step": 516 + }, + { + "epoch": 0.04, + "learning_rate": 4.244663382594417e-06, + "loss": 1.1021, + "step": 517 + }, + { + "epoch": 0.04, + "learning_rate": 4.252873563218391e-06, + "loss": 1.1214, + "step": 518 + }, + { + "epoch": 0.04, + "learning_rate": 4.261083743842365e-06, + "loss": 1.1802, + "step": 519 + }, + { + "epoch": 0.04, + "learning_rate": 4.2692939244663386e-06, + "loss": 1.2748, + "step": 520 + }, + { + "epoch": 0.04, + "learning_rate": 4.2775041050903124e-06, + "loss": 1.2503, + "step": 521 + }, + { + "epoch": 0.04, + "learning_rate": 4.2857142857142855e-06, + "loss": 1.313, + "step": 522 + }, + { + "epoch": 0.04, + "learning_rate": 4.293924466338259e-06, + "loss": 1.2407, + "step": 523 + }, + { + "epoch": 0.04, + "learning_rate": 4.302134646962233e-06, + "loss": 1.1959, + "step": 524 + }, + { + "epoch": 0.04, + "learning_rate": 4.310344827586207e-06, + "loss": 1.3324, + "step": 525 + }, + { + "epoch": 0.04, + "learning_rate": 4.318555008210181e-06, + "loss": 1.3513, + "step": 526 + }, + { + "epoch": 0.04, + "learning_rate": 4.326765188834155e-06, + "loss": 1.2196, + "step": 527 + }, + { + "epoch": 0.04, + "learning_rate": 4.334975369458129e-06, + "loss": 1.201, + "step": 528 + }, + { + "epoch": 0.04, + "learning_rate": 4.3431855500821026e-06, + "loss": 1.351, + "step": 529 + }, + { + "epoch": 0.04, + "learning_rate": 4.3513957307060764e-06, + "loss": 1.3582, + "step": 530 + }, + { + "epoch": 0.04, + "learning_rate": 4.3596059113300495e-06, + "loss": 1.2047, + "step": 531 + }, + { + "epoch": 0.04, + "learning_rate": 4.367816091954023e-06, + "loss": 1.3357, + "step": 532 + }, + { + "epoch": 0.04, + "learning_rate": 4.376026272577997e-06, + "loss": 1.2225, + "step": 533 + }, + { + "epoch": 0.04, + "learning_rate": 4.384236453201971e-06, + "loss": 1.2174, + "step": 534 + }, + { + "epoch": 0.04, + "learning_rate": 4.392446633825945e-06, + "loss": 1.1979, + "step": 535 + }, + { + "epoch": 0.04, + "learning_rate": 4.400656814449919e-06, + "loss": 1.224, + "step": 536 + }, + { + "epoch": 0.04, + "learning_rate": 4.408866995073892e-06, + "loss": 1.266, + "step": 537 + }, + { + "epoch": 0.04, + "learning_rate": 4.417077175697866e-06, + "loss": 1.2788, + "step": 538 + }, + { + "epoch": 0.04, + "learning_rate": 4.42528735632184e-06, + "loss": 1.333, + "step": 539 + }, + { + "epoch": 0.04, + "learning_rate": 4.4334975369458135e-06, + "loss": 1.1261, + "step": 540 + }, + { + "epoch": 0.04, + "learning_rate": 4.441707717569787e-06, + "loss": 1.2026, + "step": 541 + }, + { + "epoch": 0.04, + "learning_rate": 4.44991789819376e-06, + "loss": 1.291, + "step": 542 + }, + { + "epoch": 0.04, + "learning_rate": 4.458128078817734e-06, + "loss": 1.283, + "step": 543 + }, + { + "epoch": 0.04, + "learning_rate": 4.466338259441708e-06, + "loss": 1.2388, + "step": 544 + }, + { + "epoch": 0.04, + "learning_rate": 4.474548440065682e-06, + "loss": 1.144, + "step": 545 + }, + { + "epoch": 0.04, + "learning_rate": 4.482758620689656e-06, + "loss": 1.2015, + "step": 546 + }, + { + "epoch": 0.04, + "learning_rate": 4.49096880131363e-06, + "loss": 1.2115, + "step": 547 + }, + { + "epoch": 0.04, + "learning_rate": 4.499178981937603e-06, + "loss": 1.2382, + "step": 548 + }, + { + "epoch": 0.04, + "learning_rate": 4.507389162561577e-06, + "loss": 1.1387, + "step": 549 + }, + { + "epoch": 0.04, + "learning_rate": 4.5155993431855505e-06, + "loss": 1.2785, + "step": 550 + }, + { + "epoch": 0.04, + "learning_rate": 4.523809523809524e-06, + "loss": 1.1947, + "step": 551 + }, + { + "epoch": 0.04, + "learning_rate": 4.532019704433498e-06, + "loss": 1.1583, + "step": 552 + }, + { + "epoch": 0.04, + "learning_rate": 4.540229885057471e-06, + "loss": 1.2255, + "step": 553 + }, + { + "epoch": 0.04, + "learning_rate": 4.548440065681445e-06, + "loss": 1.2438, + "step": 554 + }, + { + "epoch": 0.04, + "learning_rate": 4.556650246305419e-06, + "loss": 1.2489, + "step": 555 + }, + { + "epoch": 0.04, + "learning_rate": 4.564860426929393e-06, + "loss": 1.2477, + "step": 556 + }, + { + "epoch": 0.04, + "learning_rate": 4.573070607553367e-06, + "loss": 1.288, + "step": 557 + }, + { + "epoch": 0.04, + "learning_rate": 4.581280788177341e-06, + "loss": 1.066, + "step": 558 + }, + { + "epoch": 0.04, + "learning_rate": 4.589490968801314e-06, + "loss": 1.1549, + "step": 559 + }, + { + "epoch": 0.04, + "learning_rate": 4.5977011494252875e-06, + "loss": 1.1379, + "step": 560 + }, + { + "epoch": 0.04, + "learning_rate": 4.605911330049261e-06, + "loss": 1.2212, + "step": 561 + }, + { + "epoch": 0.04, + "learning_rate": 4.614121510673235e-06, + "loss": 1.2749, + "step": 562 + }, + { + "epoch": 0.04, + "learning_rate": 4.622331691297209e-06, + "loss": 1.1651, + "step": 563 + }, + { + "epoch": 0.04, + "learning_rate": 4.630541871921182e-06, + "loss": 1.237, + "step": 564 + }, + { + "epoch": 0.04, + "learning_rate": 4.638752052545156e-06, + "loss": 1.2979, + "step": 565 + }, + { + "epoch": 0.04, + "learning_rate": 4.64696223316913e-06, + "loss": 1.1799, + "step": 566 + }, + { + "epoch": 0.04, + "learning_rate": 4.655172413793104e-06, + "loss": 1.3012, + "step": 567 + }, + { + "epoch": 0.04, + "learning_rate": 4.663382594417078e-06, + "loss": 1.1587, + "step": 568 + }, + { + "epoch": 0.04, + "learning_rate": 4.6715927750410515e-06, + "loss": 1.0556, + "step": 569 + }, + { + "epoch": 0.04, + "learning_rate": 4.6798029556650245e-06, + "loss": 1.2079, + "step": 570 + }, + { + "epoch": 0.04, + "learning_rate": 4.688013136288998e-06, + "loss": 1.0891, + "step": 571 + }, + { + "epoch": 0.04, + "learning_rate": 4.696223316912972e-06, + "loss": 1.1709, + "step": 572 + }, + { + "epoch": 0.04, + "learning_rate": 4.704433497536946e-06, + "loss": 1.1986, + "step": 573 + }, + { + "epoch": 0.04, + "learning_rate": 4.71264367816092e-06, + "loss": 1.2294, + "step": 574 + }, + { + "epoch": 0.04, + "learning_rate": 4.720853858784893e-06, + "loss": 1.1928, + "step": 575 + }, + { + "epoch": 0.04, + "learning_rate": 4.729064039408867e-06, + "loss": 1.2888, + "step": 576 + }, + { + "epoch": 0.04, + "learning_rate": 4.737274220032841e-06, + "loss": 1.1919, + "step": 577 + }, + { + "epoch": 0.04, + "learning_rate": 4.745484400656815e-06, + "loss": 1.1556, + "step": 578 + }, + { + "epoch": 0.04, + "learning_rate": 4.7536945812807885e-06, + "loss": 1.1491, + "step": 579 + }, + { + "epoch": 0.04, + "learning_rate": 4.761904761904762e-06, + "loss": 1.1009, + "step": 580 + }, + { + "epoch": 0.04, + "learning_rate": 4.770114942528735e-06, + "loss": 1.2992, + "step": 581 + }, + { + "epoch": 0.04, + "learning_rate": 4.77832512315271e-06, + "loss": 1.2172, + "step": 582 + }, + { + "epoch": 0.04, + "learning_rate": 4.786535303776684e-06, + "loss": 1.2101, + "step": 583 + }, + { + "epoch": 0.04, + "learning_rate": 4.794745484400658e-06, + "loss": 1.1762, + "step": 584 + }, + { + "epoch": 0.04, + "learning_rate": 4.802955665024631e-06, + "loss": 1.2478, + "step": 585 + }, + { + "epoch": 0.04, + "learning_rate": 4.811165845648605e-06, + "loss": 1.284, + "step": 586 + }, + { + "epoch": 0.04, + "learning_rate": 4.819376026272579e-06, + "loss": 1.1263, + "step": 587 + }, + { + "epoch": 0.04, + "learning_rate": 4.8275862068965525e-06, + "loss": 1.2529, + "step": 588 + }, + { + "epoch": 0.04, + "learning_rate": 4.835796387520526e-06, + "loss": 1.2569, + "step": 589 + }, + { + "epoch": 0.04, + "learning_rate": 4.844006568144499e-06, + "loss": 1.1844, + "step": 590 + }, + { + "epoch": 0.04, + "learning_rate": 4.852216748768473e-06, + "loss": 1.1362, + "step": 591 + }, + { + "epoch": 0.04, + "learning_rate": 4.860426929392447e-06, + "loss": 1.2063, + "step": 592 + }, + { + "epoch": 0.04, + "learning_rate": 4.868637110016421e-06, + "loss": 1.2072, + "step": 593 + }, + { + "epoch": 0.04, + "learning_rate": 4.876847290640395e-06, + "loss": 1.1802, + "step": 594 + }, + { + "epoch": 0.04, + "learning_rate": 4.885057471264369e-06, + "loss": 1.2418, + "step": 595 + }, + { + "epoch": 0.04, + "learning_rate": 4.893267651888342e-06, + "loss": 1.2161, + "step": 596 + }, + { + "epoch": 0.04, + "learning_rate": 4.901477832512316e-06, + "loss": 1.2966, + "step": 597 + }, + { + "epoch": 0.04, + "learning_rate": 4.9096880131362895e-06, + "loss": 1.1262, + "step": 598 + }, + { + "epoch": 0.04, + "learning_rate": 4.917898193760263e-06, + "loss": 1.2735, + "step": 599 + }, + { + "epoch": 0.04, + "learning_rate": 4.926108374384237e-06, + "loss": 1.2228, + "step": 600 + }, + { + "epoch": 0.04, + "learning_rate": 4.93431855500821e-06, + "loss": 1.2338, + "step": 601 + }, + { + "epoch": 0.04, + "learning_rate": 4.942528735632184e-06, + "loss": 1.2836, + "step": 602 + }, + { + "epoch": 0.04, + "learning_rate": 4.950738916256158e-06, + "loss": 1.2499, + "step": 603 + }, + { + "epoch": 0.04, + "learning_rate": 4.958949096880132e-06, + "loss": 1.1875, + "step": 604 + }, + { + "epoch": 0.04, + "learning_rate": 4.967159277504106e-06, + "loss": 1.3188, + "step": 605 + }, + { + "epoch": 0.04, + "learning_rate": 4.97536945812808e-06, + "loss": 1.2634, + "step": 606 + }, + { + "epoch": 0.04, + "learning_rate": 4.983579638752053e-06, + "loss": 1.2029, + "step": 607 + }, + { + "epoch": 0.04, + "learning_rate": 4.9917898193760265e-06, + "loss": 1.1843, + "step": 608 + }, + { + "epoch": 0.04, + "learning_rate": 5e-06, + "loss": 1.2608, + "step": 609 + }, + { + "epoch": 0.05, + "learning_rate": 5.008210180623974e-06, + "loss": 1.2796, + "step": 610 + }, + { + "epoch": 0.05, + "learning_rate": 5.016420361247948e-06, + "loss": 1.3027, + "step": 611 + }, + { + "epoch": 0.05, + "learning_rate": 5.024630541871922e-06, + "loss": 1.1894, + "step": 612 + }, + { + "epoch": 0.05, + "learning_rate": 5.032840722495895e-06, + "loss": 1.1012, + "step": 613 + }, + { + "epoch": 0.05, + "learning_rate": 5.04105090311987e-06, + "loss": 1.273, + "step": 614 + }, + { + "epoch": 0.05, + "learning_rate": 5.049261083743843e-06, + "loss": 1.1562, + "step": 615 + }, + { + "epoch": 0.05, + "learning_rate": 5.057471264367817e-06, + "loss": 1.1796, + "step": 616 + }, + { + "epoch": 0.05, + "learning_rate": 5.0656814449917905e-06, + "loss": 1.167, + "step": 617 + }, + { + "epoch": 0.05, + "learning_rate": 5.073891625615764e-06, + "loss": 1.2226, + "step": 618 + }, + { + "epoch": 0.05, + "learning_rate": 5.0821018062397374e-06, + "loss": 1.2288, + "step": 619 + }, + { + "epoch": 0.05, + "learning_rate": 5.090311986863712e-06, + "loss": 1.1719, + "step": 620 + }, + { + "epoch": 0.05, + "learning_rate": 5.098522167487685e-06, + "loss": 1.2188, + "step": 621 + }, + { + "epoch": 0.05, + "learning_rate": 5.106732348111659e-06, + "loss": 1.309, + "step": 622 + }, + { + "epoch": 0.05, + "learning_rate": 5.114942528735632e-06, + "loss": 1.1416, + "step": 623 + }, + { + "epoch": 0.05, + "learning_rate": 5.123152709359607e-06, + "loss": 1.256, + "step": 624 + }, + { + "epoch": 0.05, + "learning_rate": 5.13136288998358e-06, + "loss": 1.2191, + "step": 625 + }, + { + "epoch": 0.05, + "learning_rate": 5.1395730706075545e-06, + "loss": 1.1812, + "step": 626 + }, + { + "epoch": 0.05, + "learning_rate": 5.1477832512315276e-06, + "loss": 1.1835, + "step": 627 + }, + { + "epoch": 0.05, + "learning_rate": 5.1559934318555014e-06, + "loss": 1.2866, + "step": 628 + }, + { + "epoch": 0.05, + "learning_rate": 5.1642036124794744e-06, + "loss": 1.1116, + "step": 629 + }, + { + "epoch": 0.05, + "learning_rate": 5.172413793103449e-06, + "loss": 1.2136, + "step": 630 + }, + { + "epoch": 0.05, + "learning_rate": 5.180623973727422e-06, + "loss": 1.2756, + "step": 631 + }, + { + "epoch": 0.05, + "learning_rate": 5.188834154351396e-06, + "loss": 1.2291, + "step": 632 + }, + { + "epoch": 0.05, + "learning_rate": 5.19704433497537e-06, + "loss": 1.1801, + "step": 633 + }, + { + "epoch": 0.05, + "learning_rate": 5.205254515599344e-06, + "loss": 1.3258, + "step": 634 + }, + { + "epoch": 0.05, + "learning_rate": 5.213464696223317e-06, + "loss": 1.1592, + "step": 635 + }, + { + "epoch": 0.05, + "learning_rate": 5.2216748768472915e-06, + "loss": 1.2948, + "step": 636 + }, + { + "epoch": 0.05, + "learning_rate": 5.2298850574712646e-06, + "loss": 1.1694, + "step": 637 + }, + { + "epoch": 0.05, + "learning_rate": 5.2380952380952384e-06, + "loss": 1.1979, + "step": 638 + }, + { + "epoch": 0.05, + "learning_rate": 5.246305418719212e-06, + "loss": 1.2159, + "step": 639 + }, + { + "epoch": 0.05, + "learning_rate": 5.254515599343186e-06, + "loss": 1.1301, + "step": 640 + }, + { + "epoch": 0.05, + "learning_rate": 5.262725779967159e-06, + "loss": 1.223, + "step": 641 + }, + { + "epoch": 0.05, + "learning_rate": 5.270935960591134e-06, + "loss": 1.1968, + "step": 642 + }, + { + "epoch": 0.05, + "learning_rate": 5.279146141215107e-06, + "loss": 1.2299, + "step": 643 + }, + { + "epoch": 0.05, + "learning_rate": 5.287356321839081e-06, + "loss": 1.2526, + "step": 644 + }, + { + "epoch": 0.05, + "learning_rate": 5.295566502463054e-06, + "loss": 1.2, + "step": 645 + }, + { + "epoch": 0.05, + "learning_rate": 5.3037766830870286e-06, + "loss": 1.2237, + "step": 646 + }, + { + "epoch": 0.05, + "learning_rate": 5.311986863711002e-06, + "loss": 1.2796, + "step": 647 + }, + { + "epoch": 0.05, + "learning_rate": 5.320197044334976e-06, + "loss": 1.3156, + "step": 648 + }, + { + "epoch": 0.05, + "learning_rate": 5.328407224958949e-06, + "loss": 1.1686, + "step": 649 + }, + { + "epoch": 0.05, + "learning_rate": 5.336617405582923e-06, + "loss": 1.2083, + "step": 650 + }, + { + "epoch": 0.05, + "learning_rate": 5.344827586206896e-06, + "loss": 1.2084, + "step": 651 + }, + { + "epoch": 0.05, + "learning_rate": 5.353037766830871e-06, + "loss": 1.1444, + "step": 652 + }, + { + "epoch": 0.05, + "learning_rate": 5.361247947454844e-06, + "loss": 1.175, + "step": 653 + }, + { + "epoch": 0.05, + "learning_rate": 5.369458128078819e-06, + "loss": 1.1301, + "step": 654 + }, + { + "epoch": 0.05, + "learning_rate": 5.377668308702792e-06, + "loss": 1.1383, + "step": 655 + }, + { + "epoch": 0.05, + "learning_rate": 5.385878489326766e-06, + "loss": 1.2049, + "step": 656 + }, + { + "epoch": 0.05, + "learning_rate": 5.394088669950739e-06, + "loss": 1.0959, + "step": 657 + }, + { + "epoch": 0.05, + "learning_rate": 5.402298850574713e-06, + "loss": 1.2778, + "step": 658 + }, + { + "epoch": 0.05, + "learning_rate": 5.410509031198686e-06, + "loss": 1.1984, + "step": 659 + }, + { + "epoch": 0.05, + "learning_rate": 5.41871921182266e-06, + "loss": 1.2051, + "step": 660 + }, + { + "epoch": 0.05, + "learning_rate": 5.426929392446634e-06, + "loss": 1.2427, + "step": 661 + }, + { + "epoch": 0.05, + "learning_rate": 5.435139573070608e-06, + "loss": 1.2162, + "step": 662 + }, + { + "epoch": 0.05, + "learning_rate": 5.443349753694581e-06, + "loss": 1.2546, + "step": 663 + }, + { + "epoch": 0.05, + "learning_rate": 5.451559934318556e-06, + "loss": 1.1801, + "step": 664 + }, + { + "epoch": 0.05, + "learning_rate": 5.45977011494253e-06, + "loss": 1.2142, + "step": 665 + }, + { + "epoch": 0.05, + "learning_rate": 5.467980295566503e-06, + "loss": 1.2264, + "step": 666 + }, + { + "epoch": 0.05, + "learning_rate": 5.476190476190477e-06, + "loss": 1.2188, + "step": 667 + }, + { + "epoch": 0.05, + "learning_rate": 5.48440065681445e-06, + "loss": 1.2596, + "step": 668 + }, + { + "epoch": 0.05, + "learning_rate": 5.492610837438424e-06, + "loss": 1.1474, + "step": 669 + }, + { + "epoch": 0.05, + "learning_rate": 5.500821018062398e-06, + "loss": 1.1565, + "step": 670 + }, + { + "epoch": 0.05, + "learning_rate": 5.509031198686372e-06, + "loss": 1.2287, + "step": 671 + }, + { + "epoch": 0.05, + "learning_rate": 5.517241379310345e-06, + "loss": 1.2501, + "step": 672 + }, + { + "epoch": 0.05, + "learning_rate": 5.52545155993432e-06, + "loss": 1.1325, + "step": 673 + }, + { + "epoch": 0.05, + "learning_rate": 5.533661740558293e-06, + "loss": 1.2336, + "step": 674 + }, + { + "epoch": 0.05, + "learning_rate": 5.541871921182267e-06, + "loss": 1.2448, + "step": 675 + }, + { + "epoch": 0.05, + "learning_rate": 5.5500821018062405e-06, + "loss": 1.1498, + "step": 676 + }, + { + "epoch": 0.05, + "learning_rate": 5.558292282430214e-06, + "loss": 1.2021, + "step": 677 + }, + { + "epoch": 0.05, + "learning_rate": 5.566502463054187e-06, + "loss": 1.1608, + "step": 678 + }, + { + "epoch": 0.05, + "learning_rate": 5.574712643678162e-06, + "loss": 1.1915, + "step": 679 + }, + { + "epoch": 0.05, + "learning_rate": 5.582922824302135e-06, + "loss": 1.2343, + "step": 680 + }, + { + "epoch": 0.05, + "learning_rate": 5.591133004926109e-06, + "loss": 1.2192, + "step": 681 + }, + { + "epoch": 0.05, + "learning_rate": 5.599343185550082e-06, + "loss": 1.1549, + "step": 682 + }, + { + "epoch": 0.05, + "learning_rate": 5.607553366174057e-06, + "loss": 1.111, + "step": 683 + }, + { + "epoch": 0.05, + "learning_rate": 5.61576354679803e-06, + "loss": 1.2282, + "step": 684 + }, + { + "epoch": 0.05, + "learning_rate": 5.6239737274220045e-06, + "loss": 1.1567, + "step": 685 + }, + { + "epoch": 0.05, + "learning_rate": 5.6321839080459775e-06, + "loss": 1.2061, + "step": 686 + }, + { + "epoch": 0.05, + "learning_rate": 5.640394088669951e-06, + "loss": 1.2558, + "step": 687 + }, + { + "epoch": 0.05, + "learning_rate": 5.648604269293924e-06, + "loss": 1.1047, + "step": 688 + }, + { + "epoch": 0.05, + "learning_rate": 5.656814449917899e-06, + "loss": 1.1798, + "step": 689 + }, + { + "epoch": 0.05, + "learning_rate": 5.665024630541872e-06, + "loss": 1.1843, + "step": 690 + }, + { + "epoch": 0.05, + "learning_rate": 5.673234811165846e-06, + "loss": 1.2696, + "step": 691 + }, + { + "epoch": 0.05, + "learning_rate": 5.68144499178982e-06, + "loss": 1.1428, + "step": 692 + }, + { + "epoch": 0.05, + "learning_rate": 5.689655172413794e-06, + "loss": 1.1643, + "step": 693 + }, + { + "epoch": 0.05, + "learning_rate": 5.697865353037767e-06, + "loss": 1.1494, + "step": 694 + }, + { + "epoch": 0.05, + "learning_rate": 5.7060755336617415e-06, + "loss": 1.0995, + "step": 695 + }, + { + "epoch": 0.05, + "learning_rate": 5.7142857142857145e-06, + "loss": 1.1629, + "step": 696 + }, + { + "epoch": 0.05, + "learning_rate": 5.722495894909688e-06, + "loss": 1.1785, + "step": 697 + }, + { + "epoch": 0.05, + "learning_rate": 5.730706075533662e-06, + "loss": 1.2128, + "step": 698 + }, + { + "epoch": 0.05, + "learning_rate": 5.738916256157636e-06, + "loss": 1.1375, + "step": 699 + }, + { + "epoch": 0.05, + "learning_rate": 5.747126436781609e-06, + "loss": 1.2157, + "step": 700 + }, + { + "epoch": 0.05, + "learning_rate": 5.755336617405584e-06, + "loss": 1.2736, + "step": 701 + }, + { + "epoch": 0.05, + "learning_rate": 5.763546798029557e-06, + "loss": 1.2105, + "step": 702 + }, + { + "epoch": 0.05, + "learning_rate": 5.771756978653531e-06, + "loss": 1.2662, + "step": 703 + }, + { + "epoch": 0.05, + "learning_rate": 5.779967159277504e-06, + "loss": 1.3063, + "step": 704 + }, + { + "epoch": 0.05, + "learning_rate": 5.7881773399014785e-06, + "loss": 1.21, + "step": 705 + }, + { + "epoch": 0.05, + "learning_rate": 5.7963875205254515e-06, + "loss": 1.2665, + "step": 706 + }, + { + "epoch": 0.05, + "learning_rate": 5.804597701149426e-06, + "loss": 1.3576, + "step": 707 + }, + { + "epoch": 0.05, + "learning_rate": 5.812807881773399e-06, + "loss": 1.24, + "step": 708 + }, + { + "epoch": 0.05, + "learning_rate": 5.821018062397373e-06, + "loss": 1.2144, + "step": 709 + }, + { + "epoch": 0.05, + "learning_rate": 5.829228243021346e-06, + "loss": 1.2207, + "step": 710 + }, + { + "epoch": 0.05, + "learning_rate": 5.837438423645321e-06, + "loss": 1.2681, + "step": 711 + }, + { + "epoch": 0.05, + "learning_rate": 5.845648604269294e-06, + "loss": 1.1276, + "step": 712 + }, + { + "epoch": 0.05, + "learning_rate": 5.853858784893268e-06, + "loss": 1.2957, + "step": 713 + }, + { + "epoch": 0.05, + "learning_rate": 5.862068965517242e-06, + "loss": 1.1956, + "step": 714 + }, + { + "epoch": 0.05, + "learning_rate": 5.8702791461412155e-06, + "loss": 1.2233, + "step": 715 + }, + { + "epoch": 0.05, + "learning_rate": 5.8784893267651886e-06, + "loss": 1.2762, + "step": 716 + }, + { + "epoch": 0.05, + "learning_rate": 5.886699507389163e-06, + "loss": 1.2518, + "step": 717 + }, + { + "epoch": 0.05, + "learning_rate": 5.894909688013136e-06, + "loss": 1.2974, + "step": 718 + }, + { + "epoch": 0.05, + "learning_rate": 5.90311986863711e-06, + "loss": 1.1871, + "step": 719 + }, + { + "epoch": 0.05, + "learning_rate": 5.911330049261085e-06, + "loss": 1.1387, + "step": 720 + }, + { + "epoch": 0.05, + "learning_rate": 5.919540229885058e-06, + "loss": 1.1205, + "step": 721 + }, + { + "epoch": 0.05, + "learning_rate": 5.927750410509033e-06, + "loss": 1.1932, + "step": 722 + }, + { + "epoch": 0.05, + "learning_rate": 5.935960591133006e-06, + "loss": 1.1956, + "step": 723 + }, + { + "epoch": 0.05, + "learning_rate": 5.9441707717569795e-06, + "loss": 1.0961, + "step": 724 + }, + { + "epoch": 0.05, + "learning_rate": 5.9523809523809525e-06, + "loss": 1.1693, + "step": 725 + }, + { + "epoch": 0.05, + "learning_rate": 5.960591133004927e-06, + "loss": 1.0992, + "step": 726 + }, + { + "epoch": 0.05, + "learning_rate": 5.9688013136289e-06, + "loss": 1.2636, + "step": 727 + }, + { + "epoch": 0.05, + "learning_rate": 5.977011494252874e-06, + "loss": 1.203, + "step": 728 + }, + { + "epoch": 0.05, + "learning_rate": 5.985221674876848e-06, + "loss": 1.2331, + "step": 729 + }, + { + "epoch": 0.05, + "learning_rate": 5.993431855500822e-06, + "loss": 1.192, + "step": 730 + }, + { + "epoch": 0.05, + "learning_rate": 6.001642036124795e-06, + "loss": 1.0948, + "step": 731 + }, + { + "epoch": 0.05, + "learning_rate": 6.00985221674877e-06, + "loss": 1.2456, + "step": 732 + }, + { + "epoch": 0.05, + "learning_rate": 6.018062397372743e-06, + "loss": 1.2905, + "step": 733 + }, + { + "epoch": 0.05, + "learning_rate": 6.0262725779967165e-06, + "loss": 1.2767, + "step": 734 + }, + { + "epoch": 0.05, + "learning_rate": 6.03448275862069e-06, + "loss": 1.2177, + "step": 735 + }, + { + "epoch": 0.05, + "learning_rate": 6.042692939244664e-06, + "loss": 1.3, + "step": 736 + }, + { + "epoch": 0.05, + "learning_rate": 6.050903119868637e-06, + "loss": 1.1074, + "step": 737 + }, + { + "epoch": 0.05, + "learning_rate": 6.059113300492612e-06, + "loss": 1.1259, + "step": 738 + }, + { + "epoch": 0.05, + "learning_rate": 6.067323481116585e-06, + "loss": 1.1722, + "step": 739 + }, + { + "epoch": 0.05, + "learning_rate": 6.075533661740559e-06, + "loss": 1.2298, + "step": 740 + }, + { + "epoch": 0.05, + "learning_rate": 6.083743842364532e-06, + "loss": 1.1042, + "step": 741 + }, + { + "epoch": 0.05, + "learning_rate": 6.091954022988507e-06, + "loss": 1.1858, + "step": 742 + }, + { + "epoch": 0.05, + "learning_rate": 6.10016420361248e-06, + "loss": 1.2415, + "step": 743 + }, + { + "epoch": 0.05, + "learning_rate": 6.108374384236454e-06, + "loss": 1.2493, + "step": 744 + }, + { + "epoch": 0.06, + "learning_rate": 6.1165845648604274e-06, + "loss": 1.0902, + "step": 745 + }, + { + "epoch": 0.06, + "learning_rate": 6.124794745484401e-06, + "loss": 1.2423, + "step": 746 + }, + { + "epoch": 0.06, + "learning_rate": 6.133004926108374e-06, + "loss": 1.1422, + "step": 747 + }, + { + "epoch": 0.06, + "learning_rate": 6.141215106732349e-06, + "loss": 1.1787, + "step": 748 + }, + { + "epoch": 0.06, + "learning_rate": 6.149425287356322e-06, + "loss": 1.2675, + "step": 749 + }, + { + "epoch": 0.06, + "learning_rate": 6.157635467980296e-06, + "loss": 1.3218, + "step": 750 + }, + { + "epoch": 0.06, + "learning_rate": 6.16584564860427e-06, + "loss": 1.104, + "step": 751 + }, + { + "epoch": 0.06, + "learning_rate": 6.174055829228244e-06, + "loss": 1.182, + "step": 752 + }, + { + "epoch": 0.06, + "learning_rate": 6.182266009852217e-06, + "loss": 1.2729, + "step": 753 + }, + { + "epoch": 0.06, + "learning_rate": 6.1904761904761914e-06, + "loss": 1.2365, + "step": 754 + }, + { + "epoch": 0.06, + "learning_rate": 6.1986863711001645e-06, + "loss": 1.1792, + "step": 755 + }, + { + "epoch": 0.06, + "learning_rate": 6.206896551724138e-06, + "loss": 1.1543, + "step": 756 + }, + { + "epoch": 0.06, + "learning_rate": 6.215106732348112e-06, + "loss": 1.1425, + "step": 757 + }, + { + "epoch": 0.06, + "learning_rate": 6.223316912972086e-06, + "loss": 1.0677, + "step": 758 + }, + { + "epoch": 0.06, + "learning_rate": 6.231527093596059e-06, + "loss": 1.1925, + "step": 759 + }, + { + "epoch": 0.06, + "learning_rate": 6.239737274220034e-06, + "loss": 1.2119, + "step": 760 + }, + { + "epoch": 0.06, + "learning_rate": 6.247947454844007e-06, + "loss": 1.0626, + "step": 761 + }, + { + "epoch": 0.06, + "learning_rate": 6.256157635467981e-06, + "loss": 1.2338, + "step": 762 + }, + { + "epoch": 0.06, + "learning_rate": 6.264367816091954e-06, + "loss": 1.216, + "step": 763 + }, + { + "epoch": 0.06, + "learning_rate": 6.2725779967159284e-06, + "loss": 1.0962, + "step": 764 + }, + { + "epoch": 0.06, + "learning_rate": 6.2807881773399015e-06, + "loss": 1.109, + "step": 765 + }, + { + "epoch": 0.06, + "learning_rate": 6.288998357963876e-06, + "loss": 1.1183, + "step": 766 + }, + { + "epoch": 0.06, + "learning_rate": 6.297208538587849e-06, + "loss": 1.1258, + "step": 767 + }, + { + "epoch": 0.06, + "learning_rate": 6.305418719211823e-06, + "loss": 1.171, + "step": 768 + }, + { + "epoch": 0.06, + "learning_rate": 6.313628899835796e-06, + "loss": 1.186, + "step": 769 + }, + { + "epoch": 0.06, + "learning_rate": 6.321839080459771e-06, + "loss": 1.2022, + "step": 770 + }, + { + "epoch": 0.06, + "learning_rate": 6.330049261083744e-06, + "loss": 1.2316, + "step": 771 + }, + { + "epoch": 0.06, + "learning_rate": 6.338259441707718e-06, + "loss": 1.2025, + "step": 772 + }, + { + "epoch": 0.06, + "learning_rate": 6.346469622331692e-06, + "loss": 1.2621, + "step": 773 + }, + { + "epoch": 0.06, + "learning_rate": 6.3546798029556655e-06, + "loss": 1.2478, + "step": 774 + }, + { + "epoch": 0.06, + "learning_rate": 6.3628899835796385e-06, + "loss": 1.0892, + "step": 775 + }, + { + "epoch": 0.06, + "learning_rate": 6.371100164203613e-06, + "loss": 1.2165, + "step": 776 + }, + { + "epoch": 0.06, + "learning_rate": 6.379310344827587e-06, + "loss": 1.1364, + "step": 777 + }, + { + "epoch": 0.06, + "learning_rate": 6.38752052545156e-06, + "loss": 1.1142, + "step": 778 + }, + { + "epoch": 0.06, + "learning_rate": 6.395730706075535e-06, + "loss": 1.2727, + "step": 779 + }, + { + "epoch": 0.06, + "learning_rate": 6.403940886699508e-06, + "loss": 1.2978, + "step": 780 + }, + { + "epoch": 0.06, + "learning_rate": 6.412151067323482e-06, + "loss": 1.1176, + "step": 781 + }, + { + "epoch": 0.06, + "learning_rate": 6.420361247947456e-06, + "loss": 1.1569, + "step": 782 + }, + { + "epoch": 0.06, + "learning_rate": 6.4285714285714295e-06, + "loss": 1.174, + "step": 783 + }, + { + "epoch": 0.06, + "learning_rate": 6.4367816091954025e-06, + "loss": 1.1272, + "step": 784 + }, + { + "epoch": 0.06, + "learning_rate": 6.444991789819377e-06, + "loss": 1.2391, + "step": 785 + }, + { + "epoch": 0.06, + "learning_rate": 6.45320197044335e-06, + "loss": 1.2327, + "step": 786 + }, + { + "epoch": 0.06, + "learning_rate": 6.461412151067324e-06, + "loss": 1.1781, + "step": 787 + }, + { + "epoch": 0.06, + "learning_rate": 6.469622331691298e-06, + "loss": 1.2202, + "step": 788 + }, + { + "epoch": 0.06, + "learning_rate": 6.477832512315272e-06, + "loss": 1.0392, + "step": 789 + }, + { + "epoch": 0.06, + "learning_rate": 6.486042692939245e-06, + "loss": 1.1744, + "step": 790 + }, + { + "epoch": 0.06, + "learning_rate": 6.49425287356322e-06, + "loss": 1.2294, + "step": 791 + }, + { + "epoch": 0.06, + "learning_rate": 6.502463054187193e-06, + "loss": 1.1616, + "step": 792 + }, + { + "epoch": 0.06, + "learning_rate": 6.5106732348111665e-06, + "loss": 1.1491, + "step": 793 + }, + { + "epoch": 0.06, + "learning_rate": 6.5188834154351395e-06, + "loss": 1.1792, + "step": 794 + }, + { + "epoch": 0.06, + "learning_rate": 6.527093596059114e-06, + "loss": 1.1802, + "step": 795 + }, + { + "epoch": 0.06, + "learning_rate": 6.535303776683087e-06, + "loss": 1.185, + "step": 796 + }, + { + "epoch": 0.06, + "learning_rate": 6.543513957307062e-06, + "loss": 1.1575, + "step": 797 + }, + { + "epoch": 0.06, + "learning_rate": 6.551724137931035e-06, + "loss": 1.1362, + "step": 798 + }, + { + "epoch": 0.06, + "learning_rate": 6.559934318555009e-06, + "loss": 1.1938, + "step": 799 + }, + { + "epoch": 0.06, + "learning_rate": 6.568144499178982e-06, + "loss": 1.0974, + "step": 800 + }, + { + "epoch": 0.06, + "learning_rate": 6.576354679802957e-06, + "loss": 1.0728, + "step": 801 + }, + { + "epoch": 0.06, + "learning_rate": 6.58456486042693e-06, + "loss": 1.1867, + "step": 802 + }, + { + "epoch": 0.06, + "learning_rate": 6.592775041050904e-06, + "loss": 1.119, + "step": 803 + }, + { + "epoch": 0.06, + "learning_rate": 6.600985221674877e-06, + "loss": 1.1955, + "step": 804 + }, + { + "epoch": 0.06, + "learning_rate": 6.609195402298851e-06, + "loss": 1.2216, + "step": 805 + }, + { + "epoch": 0.06, + "learning_rate": 6.617405582922824e-06, + "loss": 1.1317, + "step": 806 + }, + { + "epoch": 0.06, + "learning_rate": 6.625615763546799e-06, + "loss": 1.2008, + "step": 807 + }, + { + "epoch": 0.06, + "learning_rate": 6.633825944170772e-06, + "loss": 1.1573, + "step": 808 + }, + { + "epoch": 0.06, + "learning_rate": 6.642036124794746e-06, + "loss": 1.221, + "step": 809 + }, + { + "epoch": 0.06, + "learning_rate": 6.65024630541872e-06, + "loss": 1.1674, + "step": 810 + }, + { + "epoch": 0.06, + "learning_rate": 6.658456486042694e-06, + "loss": 1.2038, + "step": 811 + }, + { + "epoch": 0.06, + "learning_rate": 6.666666666666667e-06, + "loss": 1.2109, + "step": 812 + }, + { + "epoch": 0.06, + "learning_rate": 6.674876847290641e-06, + "loss": 1.1706, + "step": 813 + }, + { + "epoch": 0.06, + "learning_rate": 6.683087027914614e-06, + "loss": 1.201, + "step": 814 + }, + { + "epoch": 0.06, + "learning_rate": 6.691297208538588e-06, + "loss": 1.1698, + "step": 815 + }, + { + "epoch": 0.06, + "learning_rate": 6.699507389162562e-06, + "loss": 1.1068, + "step": 816 + }, + { + "epoch": 0.06, + "learning_rate": 6.707717569786536e-06, + "loss": 1.2151, + "step": 817 + }, + { + "epoch": 0.06, + "learning_rate": 6.715927750410509e-06, + "loss": 1.1768, + "step": 818 + }, + { + "epoch": 0.06, + "learning_rate": 6.724137931034484e-06, + "loss": 1.1955, + "step": 819 + }, + { + "epoch": 0.06, + "learning_rate": 6.732348111658457e-06, + "loss": 1.122, + "step": 820 + }, + { + "epoch": 0.06, + "learning_rate": 6.740558292282431e-06, + "loss": 1.0862, + "step": 821 + }, + { + "epoch": 0.06, + "learning_rate": 6.748768472906404e-06, + "loss": 1.1752, + "step": 822 + }, + { + "epoch": 0.06, + "learning_rate": 6.756978653530378e-06, + "loss": 1.1794, + "step": 823 + }, + { + "epoch": 0.06, + "learning_rate": 6.765188834154351e-06, + "loss": 1.2384, + "step": 824 + }, + { + "epoch": 0.06, + "learning_rate": 6.773399014778326e-06, + "loss": 1.1294, + "step": 825 + }, + { + "epoch": 0.06, + "learning_rate": 6.781609195402299e-06, + "loss": 1.1898, + "step": 826 + }, + { + "epoch": 0.06, + "learning_rate": 6.789819376026273e-06, + "loss": 1.1581, + "step": 827 + }, + { + "epoch": 0.06, + "learning_rate": 6.798029556650246e-06, + "loss": 1.1832, + "step": 828 + }, + { + "epoch": 0.06, + "learning_rate": 6.806239737274221e-06, + "loss": 1.1467, + "step": 829 + }, + { + "epoch": 0.06, + "learning_rate": 6.814449917898194e-06, + "loss": 1.1228, + "step": 830 + }, + { + "epoch": 0.06, + "learning_rate": 6.822660098522168e-06, + "loss": 1.0501, + "step": 831 + }, + { + "epoch": 0.06, + "learning_rate": 6.830870279146142e-06, + "loss": 1.1577, + "step": 832 + }, + { + "epoch": 0.06, + "learning_rate": 6.839080459770115e-06, + "loss": 1.1277, + "step": 833 + }, + { + "epoch": 0.06, + "learning_rate": 6.84729064039409e-06, + "loss": 1.2518, + "step": 834 + }, + { + "epoch": 0.06, + "learning_rate": 6.855500821018063e-06, + "loss": 1.1963, + "step": 835 + }, + { + "epoch": 0.06, + "learning_rate": 6.863711001642037e-06, + "loss": 1.1685, + "step": 836 + }, + { + "epoch": 0.06, + "learning_rate": 6.87192118226601e-06, + "loss": 1.1038, + "step": 837 + }, + { + "epoch": 0.06, + "learning_rate": 6.880131362889985e-06, + "loss": 1.1518, + "step": 838 + }, + { + "epoch": 0.06, + "learning_rate": 6.888341543513958e-06, + "loss": 1.2202, + "step": 839 + }, + { + "epoch": 0.06, + "learning_rate": 6.896551724137932e-06, + "loss": 1.2448, + "step": 840 + }, + { + "epoch": 0.06, + "learning_rate": 6.9047619047619055e-06, + "loss": 1.1169, + "step": 841 + }, + { + "epoch": 0.06, + "learning_rate": 6.912972085385879e-06, + "loss": 1.1098, + "step": 842 + }, + { + "epoch": 0.06, + "learning_rate": 6.9211822660098524e-06, + "loss": 1.3329, + "step": 843 + }, + { + "epoch": 0.06, + "learning_rate": 6.929392446633827e-06, + "loss": 1.1551, + "step": 844 + }, + { + "epoch": 0.06, + "learning_rate": 6.9376026272578e-06, + "loss": 1.0271, + "step": 845 + }, + { + "epoch": 0.06, + "learning_rate": 6.945812807881774e-06, + "loss": 1.1354, + "step": 846 + }, + { + "epoch": 0.06, + "learning_rate": 6.954022988505748e-06, + "loss": 1.1598, + "step": 847 + }, + { + "epoch": 0.06, + "learning_rate": 6.962233169129722e-06, + "loss": 1.0606, + "step": 848 + }, + { + "epoch": 0.06, + "learning_rate": 6.970443349753695e-06, + "loss": 1.2562, + "step": 849 + }, + { + "epoch": 0.06, + "learning_rate": 6.9786535303776695e-06, + "loss": 1.1317, + "step": 850 + }, + { + "epoch": 0.06, + "learning_rate": 6.9868637110016425e-06, + "loss": 1.2184, + "step": 851 + }, + { + "epoch": 0.06, + "learning_rate": 6.995073891625616e-06, + "loss": 1.2478, + "step": 852 + }, + { + "epoch": 0.06, + "learning_rate": 7.0032840722495894e-06, + "loss": 1.1617, + "step": 853 + }, + { + "epoch": 0.06, + "learning_rate": 7.011494252873564e-06, + "loss": 1.2032, + "step": 854 + }, + { + "epoch": 0.06, + "learning_rate": 7.019704433497537e-06, + "loss": 1.2675, + "step": 855 + }, + { + "epoch": 0.06, + "learning_rate": 7.027914614121512e-06, + "loss": 1.236, + "step": 856 + }, + { + "epoch": 0.06, + "learning_rate": 7.036124794745485e-06, + "loss": 1.219, + "step": 857 + }, + { + "epoch": 0.06, + "learning_rate": 7.044334975369459e-06, + "loss": 1.1406, + "step": 858 + }, + { + "epoch": 0.06, + "learning_rate": 7.052545155993432e-06, + "loss": 1.1421, + "step": 859 + }, + { + "epoch": 0.06, + "learning_rate": 7.0607553366174065e-06, + "loss": 1.2088, + "step": 860 + }, + { + "epoch": 0.06, + "learning_rate": 7.0689655172413796e-06, + "loss": 1.1657, + "step": 861 + }, + { + "epoch": 0.06, + "learning_rate": 7.0771756978653534e-06, + "loss": 1.0764, + "step": 862 + }, + { + "epoch": 0.06, + "learning_rate": 7.085385878489327e-06, + "loss": 1.2337, + "step": 863 + }, + { + "epoch": 0.06, + "learning_rate": 7.093596059113301e-06, + "loss": 1.1746, + "step": 864 + }, + { + "epoch": 0.06, + "learning_rate": 7.101806239737274e-06, + "loss": 1.1695, + "step": 865 + }, + { + "epoch": 0.06, + "learning_rate": 7.110016420361249e-06, + "loss": 1.232, + "step": 866 + }, + { + "epoch": 0.06, + "learning_rate": 7.118226600985222e-06, + "loss": 1.2484, + "step": 867 + }, + { + "epoch": 0.06, + "learning_rate": 7.126436781609196e-06, + "loss": 1.1889, + "step": 868 + }, + { + "epoch": 0.06, + "learning_rate": 7.13464696223317e-06, + "loss": 1.104, + "step": 869 + }, + { + "epoch": 0.06, + "learning_rate": 7.1428571428571436e-06, + "loss": 1.1823, + "step": 870 + }, + { + "epoch": 0.06, + "learning_rate": 7.151067323481117e-06, + "loss": 1.2747, + "step": 871 + }, + { + "epoch": 0.06, + "learning_rate": 7.159277504105091e-06, + "loss": 1.1959, + "step": 872 + }, + { + "epoch": 0.06, + "learning_rate": 7.167487684729064e-06, + "loss": 1.1781, + "step": 873 + }, + { + "epoch": 0.06, + "learning_rate": 7.175697865353038e-06, + "loss": 1.0766, + "step": 874 + }, + { + "epoch": 0.06, + "learning_rate": 7.183908045977011e-06, + "loss": 1.1862, + "step": 875 + }, + { + "epoch": 0.06, + "learning_rate": 7.192118226600986e-06, + "loss": 1.2252, + "step": 876 + }, + { + "epoch": 0.06, + "learning_rate": 7.200328407224959e-06, + "loss": 1.1792, + "step": 877 + }, + { + "epoch": 0.06, + "learning_rate": 7.208538587848934e-06, + "loss": 1.1764, + "step": 878 + }, + { + "epoch": 0.06, + "learning_rate": 7.216748768472907e-06, + "loss": 1.1431, + "step": 879 + }, + { + "epoch": 0.07, + "learning_rate": 7.224958949096881e-06, + "loss": 1.3068, + "step": 880 + }, + { + "epoch": 0.07, + "learning_rate": 7.233169129720854e-06, + "loss": 1.2255, + "step": 881 + }, + { + "epoch": 0.07, + "learning_rate": 7.241379310344828e-06, + "loss": 1.1548, + "step": 882 + }, + { + "epoch": 0.07, + "learning_rate": 7.249589490968801e-06, + "loss": 1.1817, + "step": 883 + }, + { + "epoch": 0.07, + "learning_rate": 7.257799671592776e-06, + "loss": 1.2626, + "step": 884 + }, + { + "epoch": 0.07, + "learning_rate": 7.266009852216749e-06, + "loss": 1.1658, + "step": 885 + }, + { + "epoch": 0.07, + "learning_rate": 7.274220032840723e-06, + "loss": 1.1151, + "step": 886 + }, + { + "epoch": 0.07, + "learning_rate": 7.282430213464698e-06, + "loss": 1.1878, + "step": 887 + }, + { + "epoch": 0.07, + "learning_rate": 7.290640394088671e-06, + "loss": 1.103, + "step": 888 + }, + { + "epoch": 0.07, + "learning_rate": 7.298850574712645e-06, + "loss": 1.1469, + "step": 889 + }, + { + "epoch": 0.07, + "learning_rate": 7.307060755336618e-06, + "loss": 1.1327, + "step": 890 + }, + { + "epoch": 0.07, + "learning_rate": 7.315270935960592e-06, + "loss": 1.1246, + "step": 891 + }, + { + "epoch": 0.07, + "learning_rate": 7.323481116584565e-06, + "loss": 1.1883, + "step": 892 + }, + { + "epoch": 0.07, + "learning_rate": 7.33169129720854e-06, + "loss": 1.1986, + "step": 893 + }, + { + "epoch": 0.07, + "learning_rate": 7.339901477832513e-06, + "loss": 1.1809, + "step": 894 + }, + { + "epoch": 0.07, + "learning_rate": 7.348111658456487e-06, + "loss": 1.1567, + "step": 895 + }, + { + "epoch": 0.07, + "learning_rate": 7.35632183908046e-06, + "loss": 1.2232, + "step": 896 + }, + { + "epoch": 0.07, + "learning_rate": 7.364532019704435e-06, + "loss": 1.2883, + "step": 897 + }, + { + "epoch": 0.07, + "learning_rate": 7.372742200328408e-06, + "loss": 1.1245, + "step": 898 + }, + { + "epoch": 0.07, + "learning_rate": 7.380952380952382e-06, + "loss": 1.1765, + "step": 899 + }, + { + "epoch": 0.07, + "learning_rate": 7.3891625615763555e-06, + "loss": 1.1802, + "step": 900 + }, + { + "epoch": 0.07, + "learning_rate": 7.397372742200329e-06, + "loss": 1.1794, + "step": 901 + }, + { + "epoch": 0.07, + "learning_rate": 7.405582922824302e-06, + "loss": 1.1125, + "step": 902 + }, + { + "epoch": 0.07, + "learning_rate": 7.413793103448277e-06, + "loss": 1.1249, + "step": 903 + }, + { + "epoch": 0.07, + "learning_rate": 7.42200328407225e-06, + "loss": 1.109, + "step": 904 + }, + { + "epoch": 0.07, + "learning_rate": 7.430213464696224e-06, + "loss": 1.1772, + "step": 905 + }, + { + "epoch": 0.07, + "learning_rate": 7.438423645320198e-06, + "loss": 1.1445, + "step": 906 + }, + { + "epoch": 0.07, + "learning_rate": 7.446633825944172e-06, + "loss": 1.2266, + "step": 907 + }, + { + "epoch": 0.07, + "learning_rate": 7.454844006568145e-06, + "loss": 1.2612, + "step": 908 + }, + { + "epoch": 0.07, + "learning_rate": 7.4630541871921195e-06, + "loss": 1.1673, + "step": 909 + }, + { + "epoch": 0.07, + "learning_rate": 7.4712643678160925e-06, + "loss": 1.1407, + "step": 910 + }, + { + "epoch": 0.07, + "learning_rate": 7.479474548440066e-06, + "loss": 1.2201, + "step": 911 + }, + { + "epoch": 0.07, + "learning_rate": 7.487684729064039e-06, + "loss": 1.2353, + "step": 912 + }, + { + "epoch": 0.07, + "learning_rate": 7.495894909688014e-06, + "loss": 1.0915, + "step": 913 + }, + { + "epoch": 0.07, + "learning_rate": 7.504105090311987e-06, + "loss": 1.1709, + "step": 914 + }, + { + "epoch": 0.07, + "learning_rate": 7.512315270935962e-06, + "loss": 1.1926, + "step": 915 + }, + { + "epoch": 0.07, + "learning_rate": 7.520525451559935e-06, + "loss": 1.2188, + "step": 916 + }, + { + "epoch": 0.07, + "learning_rate": 7.528735632183909e-06, + "loss": 1.143, + "step": 917 + }, + { + "epoch": 0.07, + "learning_rate": 7.536945812807882e-06, + "loss": 1.1471, + "step": 918 + }, + { + "epoch": 0.07, + "learning_rate": 7.5451559934318565e-06, + "loss": 1.0755, + "step": 919 + }, + { + "epoch": 0.07, + "learning_rate": 7.5533661740558295e-06, + "loss": 1.1445, + "step": 920 + }, + { + "epoch": 0.07, + "learning_rate": 7.561576354679803e-06, + "loss": 1.143, + "step": 921 + }, + { + "epoch": 0.07, + "learning_rate": 7.569786535303777e-06, + "loss": 1.2289, + "step": 922 + }, + { + "epoch": 0.07, + "learning_rate": 7.577996715927751e-06, + "loss": 1.142, + "step": 923 + }, + { + "epoch": 0.07, + "learning_rate": 7.586206896551724e-06, + "loss": 1.1257, + "step": 924 + }, + { + "epoch": 0.07, + "learning_rate": 7.594417077175699e-06, + "loss": 1.2097, + "step": 925 + }, + { + "epoch": 0.07, + "learning_rate": 7.602627257799672e-06, + "loss": 1.1522, + "step": 926 + }, + { + "epoch": 0.07, + "learning_rate": 7.610837438423646e-06, + "loss": 1.1378, + "step": 927 + }, + { + "epoch": 0.07, + "learning_rate": 7.61904761904762e-06, + "loss": 1.0805, + "step": 928 + }, + { + "epoch": 0.07, + "learning_rate": 7.6272577996715935e-06, + "loss": 1.0936, + "step": 929 + }, + { + "epoch": 0.07, + "learning_rate": 7.635467980295567e-06, + "loss": 1.2311, + "step": 930 + }, + { + "epoch": 0.07, + "learning_rate": 7.64367816091954e-06, + "loss": 1.1434, + "step": 931 + }, + { + "epoch": 0.07, + "learning_rate": 7.651888341543514e-06, + "loss": 1.2398, + "step": 932 + }, + { + "epoch": 0.07, + "learning_rate": 7.660098522167488e-06, + "loss": 1.1129, + "step": 933 + }, + { + "epoch": 0.07, + "learning_rate": 7.668308702791462e-06, + "loss": 1.1736, + "step": 934 + }, + { + "epoch": 0.07, + "learning_rate": 7.676518883415436e-06, + "loss": 1.145, + "step": 935 + }, + { + "epoch": 0.07, + "learning_rate": 7.68472906403941e-06, + "loss": 1.0913, + "step": 936 + }, + { + "epoch": 0.07, + "learning_rate": 7.692939244663384e-06, + "loss": 1.1705, + "step": 937 + }, + { + "epoch": 0.07, + "learning_rate": 7.701149425287356e-06, + "loss": 1.1492, + "step": 938 + }, + { + "epoch": 0.07, + "learning_rate": 7.709359605911331e-06, + "loss": 1.1592, + "step": 939 + }, + { + "epoch": 0.07, + "learning_rate": 7.717569786535304e-06, + "loss": 1.1978, + "step": 940 + }, + { + "epoch": 0.07, + "learning_rate": 7.725779967159277e-06, + "loss": 1.2626, + "step": 941 + }, + { + "epoch": 0.07, + "learning_rate": 7.733990147783253e-06, + "loss": 1.2509, + "step": 942 + }, + { + "epoch": 0.07, + "learning_rate": 7.742200328407225e-06, + "loss": 1.1592, + "step": 943 + }, + { + "epoch": 0.07, + "learning_rate": 7.750410509031199e-06, + "loss": 1.133, + "step": 944 + }, + { + "epoch": 0.07, + "learning_rate": 7.758620689655173e-06, + "loss": 1.2338, + "step": 945 + }, + { + "epoch": 0.07, + "learning_rate": 7.766830870279147e-06, + "loss": 1.2469, + "step": 946 + }, + { + "epoch": 0.07, + "learning_rate": 7.77504105090312e-06, + "loss": 1.1565, + "step": 947 + }, + { + "epoch": 0.07, + "learning_rate": 7.783251231527095e-06, + "loss": 1.1771, + "step": 948 + }, + { + "epoch": 0.07, + "learning_rate": 7.791461412151068e-06, + "loss": 1.1432, + "step": 949 + }, + { + "epoch": 0.07, + "learning_rate": 7.799671592775042e-06, + "loss": 1.1238, + "step": 950 + }, + { + "epoch": 0.07, + "learning_rate": 7.807881773399016e-06, + "loss": 1.1687, + "step": 951 + }, + { + "epoch": 0.07, + "learning_rate": 7.81609195402299e-06, + "loss": 1.1063, + "step": 952 + }, + { + "epoch": 0.07, + "learning_rate": 7.824302134646962e-06, + "loss": 1.1377, + "step": 953 + }, + { + "epoch": 0.07, + "learning_rate": 7.832512315270938e-06, + "loss": 1.1575, + "step": 954 + }, + { + "epoch": 0.07, + "learning_rate": 7.84072249589491e-06, + "loss": 1.0923, + "step": 955 + }, + { + "epoch": 0.07, + "learning_rate": 7.848932676518884e-06, + "loss": 1.2845, + "step": 956 + }, + { + "epoch": 0.07, + "learning_rate": 7.857142857142858e-06, + "loss": 1.1438, + "step": 957 + }, + { + "epoch": 0.07, + "learning_rate": 7.865353037766832e-06, + "loss": 1.0268, + "step": 958 + }, + { + "epoch": 0.07, + "learning_rate": 7.873563218390805e-06, + "loss": 1.0944, + "step": 959 + }, + { + "epoch": 0.07, + "learning_rate": 7.88177339901478e-06, + "loss": 1.0933, + "step": 960 + }, + { + "epoch": 0.07, + "learning_rate": 7.889983579638753e-06, + "loss": 1.1413, + "step": 961 + }, + { + "epoch": 0.07, + "learning_rate": 7.898193760262727e-06, + "loss": 1.175, + "step": 962 + }, + { + "epoch": 0.07, + "learning_rate": 7.9064039408867e-06, + "loss": 1.151, + "step": 963 + }, + { + "epoch": 0.07, + "learning_rate": 7.914614121510675e-06, + "loss": 1.1593, + "step": 964 + }, + { + "epoch": 0.07, + "learning_rate": 7.922824302134647e-06, + "loss": 1.1991, + "step": 965 + }, + { + "epoch": 0.07, + "learning_rate": 7.93103448275862e-06, + "loss": 1.1202, + "step": 966 + }, + { + "epoch": 0.07, + "learning_rate": 7.939244663382595e-06, + "loss": 1.1595, + "step": 967 + }, + { + "epoch": 0.07, + "learning_rate": 7.947454844006569e-06, + "loss": 1.219, + "step": 968 + }, + { + "epoch": 0.07, + "learning_rate": 7.955665024630542e-06, + "loss": 1.1661, + "step": 969 + }, + { + "epoch": 0.07, + "learning_rate": 7.963875205254516e-06, + "loss": 1.1777, + "step": 970 + }, + { + "epoch": 0.07, + "learning_rate": 7.97208538587849e-06, + "loss": 1.1527, + "step": 971 + }, + { + "epoch": 0.07, + "learning_rate": 7.980295566502464e-06, + "loss": 1.1989, + "step": 972 + }, + { + "epoch": 0.07, + "learning_rate": 7.988505747126438e-06, + "loss": 1.0837, + "step": 973 + }, + { + "epoch": 0.07, + "learning_rate": 7.996715927750412e-06, + "loss": 1.1196, + "step": 974 + }, + { + "epoch": 0.07, + "learning_rate": 8.004926108374384e-06, + "loss": 1.2262, + "step": 975 + }, + { + "epoch": 0.07, + "learning_rate": 8.01313628899836e-06, + "loss": 1.2083, + "step": 976 + }, + { + "epoch": 0.07, + "learning_rate": 8.021346469622332e-06, + "loss": 1.1646, + "step": 977 + }, + { + "epoch": 0.07, + "learning_rate": 8.029556650246306e-06, + "loss": 1.1822, + "step": 978 + }, + { + "epoch": 0.07, + "learning_rate": 8.03776683087028e-06, + "loss": 1.1963, + "step": 979 + }, + { + "epoch": 0.07, + "learning_rate": 8.045977011494253e-06, + "loss": 1.1397, + "step": 980 + }, + { + "epoch": 0.07, + "learning_rate": 8.054187192118227e-06, + "loss": 1.186, + "step": 981 + }, + { + "epoch": 0.07, + "learning_rate": 8.062397372742201e-06, + "loss": 1.26, + "step": 982 + }, + { + "epoch": 0.07, + "learning_rate": 8.070607553366175e-06, + "loss": 1.1484, + "step": 983 + }, + { + "epoch": 0.07, + "learning_rate": 8.078817733990149e-06, + "loss": 1.2994, + "step": 984 + }, + { + "epoch": 0.07, + "learning_rate": 8.087027914614121e-06, + "loss": 1.1738, + "step": 985 + }, + { + "epoch": 0.07, + "learning_rate": 8.095238095238097e-06, + "loss": 1.1347, + "step": 986 + }, + { + "epoch": 0.07, + "learning_rate": 8.103448275862069e-06, + "loss": 1.2361, + "step": 987 + }, + { + "epoch": 0.07, + "learning_rate": 8.111658456486044e-06, + "loss": 1.1611, + "step": 988 + }, + { + "epoch": 0.07, + "learning_rate": 8.119868637110016e-06, + "loss": 1.1617, + "step": 989 + }, + { + "epoch": 0.07, + "learning_rate": 8.12807881773399e-06, + "loss": 1.193, + "step": 990 + }, + { + "epoch": 0.07, + "learning_rate": 8.136288998357964e-06, + "loss": 1.1783, + "step": 991 + }, + { + "epoch": 0.07, + "learning_rate": 8.144499178981938e-06, + "loss": 1.2445, + "step": 992 + }, + { + "epoch": 0.07, + "learning_rate": 8.152709359605912e-06, + "loss": 1.1424, + "step": 993 + }, + { + "epoch": 0.07, + "learning_rate": 8.160919540229886e-06, + "loss": 1.1488, + "step": 994 + }, + { + "epoch": 0.07, + "learning_rate": 8.16912972085386e-06, + "loss": 1.1646, + "step": 995 + }, + { + "epoch": 0.07, + "learning_rate": 8.177339901477834e-06, + "loss": 1.1557, + "step": 996 + }, + { + "epoch": 0.07, + "learning_rate": 8.185550082101807e-06, + "loss": 1.2662, + "step": 997 + }, + { + "epoch": 0.07, + "learning_rate": 8.193760262725781e-06, + "loss": 1.0307, + "step": 998 + }, + { + "epoch": 0.07, + "learning_rate": 8.201970443349755e-06, + "loss": 1.0857, + "step": 999 + }, + { + "epoch": 0.07, + "learning_rate": 8.210180623973727e-06, + "loss": 1.1953, + "step": 1000 + }, + { + "epoch": 0.07, + "learning_rate": 8.218390804597703e-06, + "loss": 1.2529, + "step": 1001 + }, + { + "epoch": 0.07, + "learning_rate": 8.226600985221675e-06, + "loss": 1.2425, + "step": 1002 + }, + { + "epoch": 0.07, + "learning_rate": 8.234811165845649e-06, + "loss": 1.1313, + "step": 1003 + }, + { + "epoch": 0.07, + "learning_rate": 8.243021346469623e-06, + "loss": 1.1301, + "step": 1004 + }, + { + "epoch": 0.07, + "learning_rate": 8.251231527093597e-06, + "loss": 1.1299, + "step": 1005 + }, + { + "epoch": 0.07, + "learning_rate": 8.25944170771757e-06, + "loss": 1.164, + "step": 1006 + }, + { + "epoch": 0.07, + "learning_rate": 8.267651888341544e-06, + "loss": 1.1923, + "step": 1007 + }, + { + "epoch": 0.07, + "learning_rate": 8.275862068965518e-06, + "loss": 1.2426, + "step": 1008 + }, + { + "epoch": 0.07, + "learning_rate": 8.284072249589492e-06, + "loss": 1.1094, + "step": 1009 + }, + { + "epoch": 0.07, + "learning_rate": 8.292282430213466e-06, + "loss": 1.1084, + "step": 1010 + }, + { + "epoch": 0.07, + "learning_rate": 8.30049261083744e-06, + "loss": 1.118, + "step": 1011 + }, + { + "epoch": 0.07, + "learning_rate": 8.308702791461412e-06, + "loss": 1.0555, + "step": 1012 + }, + { + "epoch": 0.07, + "learning_rate": 8.316912972085388e-06, + "loss": 1.0223, + "step": 1013 + }, + { + "epoch": 0.07, + "learning_rate": 8.32512315270936e-06, + "loss": 1.2333, + "step": 1014 + }, + { + "epoch": 0.07, + "learning_rate": 8.333333333333334e-06, + "loss": 1.2027, + "step": 1015 + }, + { + "epoch": 0.08, + "learning_rate": 8.341543513957308e-06, + "loss": 1.1857, + "step": 1016 + }, + { + "epoch": 0.08, + "learning_rate": 8.349753694581281e-06, + "loss": 0.9858, + "step": 1017 + }, + { + "epoch": 0.08, + "learning_rate": 8.357963875205255e-06, + "loss": 1.183, + "step": 1018 + }, + { + "epoch": 0.08, + "learning_rate": 8.36617405582923e-06, + "loss": 1.2974, + "step": 1019 + }, + { + "epoch": 0.08, + "learning_rate": 8.374384236453203e-06, + "loss": 1.0723, + "step": 1020 + }, + { + "epoch": 0.08, + "learning_rate": 8.382594417077177e-06, + "loss": 1.068, + "step": 1021 + }, + { + "epoch": 0.08, + "learning_rate": 8.390804597701149e-06, + "loss": 1.0924, + "step": 1022 + }, + { + "epoch": 0.08, + "learning_rate": 8.399014778325125e-06, + "loss": 1.2131, + "step": 1023 + }, + { + "epoch": 0.08, + "learning_rate": 8.407224958949097e-06, + "loss": 1.2313, + "step": 1024 + }, + { + "epoch": 0.08, + "learning_rate": 8.41543513957307e-06, + "loss": 1.1353, + "step": 1025 + }, + { + "epoch": 0.08, + "learning_rate": 8.423645320197045e-06, + "loss": 1.1336, + "step": 1026 + }, + { + "epoch": 0.08, + "learning_rate": 8.431855500821018e-06, + "loss": 0.99, + "step": 1027 + }, + { + "epoch": 0.08, + "learning_rate": 8.440065681444992e-06, + "loss": 1.2241, + "step": 1028 + }, + { + "epoch": 0.08, + "learning_rate": 8.448275862068966e-06, + "loss": 1.2625, + "step": 1029 + }, + { + "epoch": 0.08, + "learning_rate": 8.45648604269294e-06, + "loss": 1.219, + "step": 1030 + }, + { + "epoch": 0.08, + "learning_rate": 8.464696223316914e-06, + "loss": 1.158, + "step": 1031 + }, + { + "epoch": 0.08, + "learning_rate": 8.472906403940888e-06, + "loss": 1.1049, + "step": 1032 + }, + { + "epoch": 0.08, + "learning_rate": 8.481116584564862e-06, + "loss": 1.2292, + "step": 1033 + }, + { + "epoch": 0.08, + "learning_rate": 8.489326765188834e-06, + "loss": 1.1444, + "step": 1034 + }, + { + "epoch": 0.08, + "learning_rate": 8.49753694581281e-06, + "loss": 1.1204, + "step": 1035 + }, + { + "epoch": 0.08, + "learning_rate": 8.505747126436782e-06, + "loss": 1.2024, + "step": 1036 + }, + { + "epoch": 0.08, + "learning_rate": 8.513957307060756e-06, + "loss": 1.0664, + "step": 1037 + }, + { + "epoch": 0.08, + "learning_rate": 8.52216748768473e-06, + "loss": 1.0697, + "step": 1038 + }, + { + "epoch": 0.08, + "learning_rate": 8.530377668308703e-06, + "loss": 1.1015, + "step": 1039 + }, + { + "epoch": 0.08, + "learning_rate": 8.538587848932677e-06, + "loss": 1.0757, + "step": 1040 + }, + { + "epoch": 0.08, + "learning_rate": 8.546798029556651e-06, + "loss": 1.0602, + "step": 1041 + }, + { + "epoch": 0.08, + "learning_rate": 8.555008210180625e-06, + "loss": 1.0957, + "step": 1042 + }, + { + "epoch": 0.08, + "learning_rate": 8.563218390804599e-06, + "loss": 1.1972, + "step": 1043 + }, + { + "epoch": 0.08, + "learning_rate": 8.571428571428571e-06, + "loss": 1.1091, + "step": 1044 + }, + { + "epoch": 0.08, + "learning_rate": 8.579638752052546e-06, + "loss": 1.0935, + "step": 1045 + }, + { + "epoch": 0.08, + "learning_rate": 8.587848932676519e-06, + "loss": 1.1659, + "step": 1046 + }, + { + "epoch": 0.08, + "learning_rate": 8.596059113300493e-06, + "loss": 1.0915, + "step": 1047 + }, + { + "epoch": 0.08, + "learning_rate": 8.604269293924466e-06, + "loss": 1.1448, + "step": 1048 + }, + { + "epoch": 0.08, + "learning_rate": 8.61247947454844e-06, + "loss": 1.1288, + "step": 1049 + }, + { + "epoch": 0.08, + "learning_rate": 8.620689655172414e-06, + "loss": 1.0097, + "step": 1050 + }, + { + "epoch": 0.08, + "learning_rate": 8.628899835796388e-06, + "loss": 1.2039, + "step": 1051 + }, + { + "epoch": 0.08, + "learning_rate": 8.637110016420362e-06, + "loss": 1.0653, + "step": 1052 + }, + { + "epoch": 0.08, + "learning_rate": 8.645320197044336e-06, + "loss": 1.205, + "step": 1053 + }, + { + "epoch": 0.08, + "learning_rate": 8.65353037766831e-06, + "loss": 1.0748, + "step": 1054 + }, + { + "epoch": 0.08, + "learning_rate": 8.661740558292284e-06, + "loss": 1.0358, + "step": 1055 + }, + { + "epoch": 0.08, + "learning_rate": 8.669950738916257e-06, + "loss": 1.2174, + "step": 1056 + }, + { + "epoch": 0.08, + "learning_rate": 8.678160919540231e-06, + "loss": 1.2247, + "step": 1057 + }, + { + "epoch": 0.08, + "learning_rate": 8.686371100164205e-06, + "loss": 1.1153, + "step": 1058 + }, + { + "epoch": 0.08, + "learning_rate": 8.694581280788177e-06, + "loss": 1.1114, + "step": 1059 + }, + { + "epoch": 0.08, + "learning_rate": 8.702791461412153e-06, + "loss": 1.1194, + "step": 1060 + }, + { + "epoch": 0.08, + "learning_rate": 8.711001642036125e-06, + "loss": 1.1501, + "step": 1061 + }, + { + "epoch": 0.08, + "learning_rate": 8.719211822660099e-06, + "loss": 1.1129, + "step": 1062 + }, + { + "epoch": 0.08, + "learning_rate": 8.727422003284073e-06, + "loss": 1.1329, + "step": 1063 + }, + { + "epoch": 0.08, + "learning_rate": 8.735632183908047e-06, + "loss": 1.1827, + "step": 1064 + }, + { + "epoch": 0.08, + "learning_rate": 8.74384236453202e-06, + "loss": 1.1998, + "step": 1065 + }, + { + "epoch": 0.08, + "learning_rate": 8.752052545155994e-06, + "loss": 1.0167, + "step": 1066 + }, + { + "epoch": 0.08, + "learning_rate": 8.760262725779968e-06, + "loss": 1.1131, + "step": 1067 + }, + { + "epoch": 0.08, + "learning_rate": 8.768472906403942e-06, + "loss": 1.1726, + "step": 1068 + }, + { + "epoch": 0.08, + "learning_rate": 8.776683087027914e-06, + "loss": 1.0656, + "step": 1069 + }, + { + "epoch": 0.08, + "learning_rate": 8.78489326765189e-06, + "loss": 1.188, + "step": 1070 + }, + { + "epoch": 0.08, + "learning_rate": 8.793103448275862e-06, + "loss": 1.045, + "step": 1071 + }, + { + "epoch": 0.08, + "learning_rate": 8.801313628899838e-06, + "loss": 1.0717, + "step": 1072 + }, + { + "epoch": 0.08, + "learning_rate": 8.80952380952381e-06, + "loss": 1.0158, + "step": 1073 + }, + { + "epoch": 0.08, + "learning_rate": 8.817733990147784e-06, + "loss": 1.1019, + "step": 1074 + }, + { + "epoch": 0.08, + "learning_rate": 8.825944170771758e-06, + "loss": 1.0755, + "step": 1075 + }, + { + "epoch": 0.08, + "learning_rate": 8.834154351395731e-06, + "loss": 1.1197, + "step": 1076 + }, + { + "epoch": 0.08, + "learning_rate": 8.842364532019705e-06, + "loss": 1.1085, + "step": 1077 + }, + { + "epoch": 0.08, + "learning_rate": 8.85057471264368e-06, + "loss": 1.0931, + "step": 1078 + }, + { + "epoch": 0.08, + "learning_rate": 8.858784893267653e-06, + "loss": 1.0892, + "step": 1079 + }, + { + "epoch": 0.08, + "learning_rate": 8.866995073891627e-06, + "loss": 1.1705, + "step": 1080 + }, + { + "epoch": 0.08, + "learning_rate": 8.875205254515599e-06, + "loss": 1.2253, + "step": 1081 + }, + { + "epoch": 0.08, + "learning_rate": 8.883415435139575e-06, + "loss": 1.2162, + "step": 1082 + }, + { + "epoch": 0.08, + "learning_rate": 8.891625615763547e-06, + "loss": 1.1818, + "step": 1083 + }, + { + "epoch": 0.08, + "learning_rate": 8.89983579638752e-06, + "loss": 0.9915, + "step": 1084 + }, + { + "epoch": 0.08, + "learning_rate": 8.908045977011495e-06, + "loss": 1.1964, + "step": 1085 + }, + { + "epoch": 0.08, + "learning_rate": 8.916256157635468e-06, + "loss": 1.1845, + "step": 1086 + }, + { + "epoch": 0.08, + "learning_rate": 8.924466338259442e-06, + "loss": 1.1295, + "step": 1087 + }, + { + "epoch": 0.08, + "learning_rate": 8.932676518883416e-06, + "loss": 1.2086, + "step": 1088 + }, + { + "epoch": 0.08, + "learning_rate": 8.94088669950739e-06, + "loss": 1.1454, + "step": 1089 + }, + { + "epoch": 0.08, + "learning_rate": 8.949096880131364e-06, + "loss": 1.0755, + "step": 1090 + }, + { + "epoch": 0.08, + "learning_rate": 8.957307060755338e-06, + "loss": 1.0902, + "step": 1091 + }, + { + "epoch": 0.08, + "learning_rate": 8.965517241379312e-06, + "loss": 1.2115, + "step": 1092 + }, + { + "epoch": 0.08, + "learning_rate": 8.973727422003284e-06, + "loss": 1.1242, + "step": 1093 + }, + { + "epoch": 0.08, + "learning_rate": 8.98193760262726e-06, + "loss": 1.1449, + "step": 1094 + }, + { + "epoch": 0.08, + "learning_rate": 8.990147783251232e-06, + "loss": 1.1905, + "step": 1095 + }, + { + "epoch": 0.08, + "learning_rate": 8.998357963875205e-06, + "loss": 1.1839, + "step": 1096 + }, + { + "epoch": 0.08, + "learning_rate": 9.00656814449918e-06, + "loss": 1.0979, + "step": 1097 + }, + { + "epoch": 0.08, + "learning_rate": 9.014778325123153e-06, + "loss": 1.1086, + "step": 1098 + }, + { + "epoch": 0.08, + "learning_rate": 9.022988505747127e-06, + "loss": 1.1827, + "step": 1099 + }, + { + "epoch": 0.08, + "learning_rate": 9.031198686371101e-06, + "loss": 1.0985, + "step": 1100 + }, + { + "epoch": 0.08, + "learning_rate": 9.039408866995075e-06, + "loss": 1.1076, + "step": 1101 + }, + { + "epoch": 0.08, + "learning_rate": 9.047619047619049e-06, + "loss": 1.1127, + "step": 1102 + }, + { + "epoch": 0.08, + "learning_rate": 9.055829228243021e-06, + "loss": 1.1703, + "step": 1103 + }, + { + "epoch": 0.08, + "learning_rate": 9.064039408866996e-06, + "loss": 1.1193, + "step": 1104 + }, + { + "epoch": 0.08, + "learning_rate": 9.072249589490969e-06, + "loss": 1.1301, + "step": 1105 + }, + { + "epoch": 0.08, + "learning_rate": 9.080459770114942e-06, + "loss": 1.1691, + "step": 1106 + }, + { + "epoch": 0.08, + "learning_rate": 9.088669950738916e-06, + "loss": 1.2609, + "step": 1107 + }, + { + "epoch": 0.08, + "learning_rate": 9.09688013136289e-06, + "loss": 1.1419, + "step": 1108 + }, + { + "epoch": 0.08, + "learning_rate": 9.105090311986866e-06, + "loss": 1.087, + "step": 1109 + }, + { + "epoch": 0.08, + "learning_rate": 9.113300492610838e-06, + "loss": 1.159, + "step": 1110 + }, + { + "epoch": 0.08, + "learning_rate": 9.121510673234812e-06, + "loss": 1.2696, + "step": 1111 + }, + { + "epoch": 0.08, + "learning_rate": 9.129720853858786e-06, + "loss": 1.1012, + "step": 1112 + }, + { + "epoch": 0.08, + "learning_rate": 9.13793103448276e-06, + "loss": 1.2311, + "step": 1113 + }, + { + "epoch": 0.08, + "learning_rate": 9.146141215106733e-06, + "loss": 1.1786, + "step": 1114 + }, + { + "epoch": 0.08, + "learning_rate": 9.154351395730707e-06, + "loss": 1.1383, + "step": 1115 + }, + { + "epoch": 0.08, + "learning_rate": 9.162561576354681e-06, + "loss": 1.1276, + "step": 1116 + }, + { + "epoch": 0.08, + "learning_rate": 9.170771756978655e-06, + "loss": 1.176, + "step": 1117 + }, + { + "epoch": 0.08, + "learning_rate": 9.178981937602627e-06, + "loss": 1.21, + "step": 1118 + }, + { + "epoch": 0.08, + "learning_rate": 9.187192118226603e-06, + "loss": 1.1757, + "step": 1119 + }, + { + "epoch": 0.08, + "learning_rate": 9.195402298850575e-06, + "loss": 1.0026, + "step": 1120 + }, + { + "epoch": 0.08, + "learning_rate": 9.203612479474549e-06, + "loss": 1.023, + "step": 1121 + }, + { + "epoch": 0.08, + "learning_rate": 9.211822660098523e-06, + "loss": 1.0096, + "step": 1122 + }, + { + "epoch": 0.08, + "learning_rate": 9.220032840722497e-06, + "loss": 1.1831, + "step": 1123 + }, + { + "epoch": 0.08, + "learning_rate": 9.22824302134647e-06, + "loss": 1.2996, + "step": 1124 + }, + { + "epoch": 0.08, + "learning_rate": 9.236453201970444e-06, + "loss": 1.0701, + "step": 1125 + }, + { + "epoch": 0.08, + "learning_rate": 9.244663382594418e-06, + "loss": 1.1929, + "step": 1126 + }, + { + "epoch": 0.08, + "learning_rate": 9.252873563218392e-06, + "loss": 1.1551, + "step": 1127 + }, + { + "epoch": 0.08, + "learning_rate": 9.261083743842364e-06, + "loss": 1.1361, + "step": 1128 + }, + { + "epoch": 0.08, + "learning_rate": 9.26929392446634e-06, + "loss": 1.1346, + "step": 1129 + }, + { + "epoch": 0.08, + "learning_rate": 9.277504105090312e-06, + "loss": 1.2347, + "step": 1130 + }, + { + "epoch": 0.08, + "learning_rate": 9.285714285714288e-06, + "loss": 1.1278, + "step": 1131 + }, + { + "epoch": 0.08, + "learning_rate": 9.29392446633826e-06, + "loss": 1.1136, + "step": 1132 + }, + { + "epoch": 0.08, + "learning_rate": 9.302134646962234e-06, + "loss": 1.1552, + "step": 1133 + }, + { + "epoch": 0.08, + "learning_rate": 9.310344827586207e-06, + "loss": 1.1719, + "step": 1134 + }, + { + "epoch": 0.08, + "learning_rate": 9.318555008210181e-06, + "loss": 1.0154, + "step": 1135 + }, + { + "epoch": 0.08, + "learning_rate": 9.326765188834155e-06, + "loss": 1.123, + "step": 1136 + }, + { + "epoch": 0.08, + "learning_rate": 9.334975369458129e-06, + "loss": 1.1401, + "step": 1137 + }, + { + "epoch": 0.08, + "learning_rate": 9.343185550082103e-06, + "loss": 1.1675, + "step": 1138 + }, + { + "epoch": 0.08, + "learning_rate": 9.351395730706077e-06, + "loss": 1.1599, + "step": 1139 + }, + { + "epoch": 0.08, + "learning_rate": 9.359605911330049e-06, + "loss": 1.0652, + "step": 1140 + }, + { + "epoch": 0.08, + "learning_rate": 9.367816091954025e-06, + "loss": 1.2254, + "step": 1141 + }, + { + "epoch": 0.08, + "learning_rate": 9.376026272577997e-06, + "loss": 1.1616, + "step": 1142 + }, + { + "epoch": 0.08, + "learning_rate": 9.38423645320197e-06, + "loss": 1.1628, + "step": 1143 + }, + { + "epoch": 0.08, + "learning_rate": 9.392446633825945e-06, + "loss": 1.2001, + "step": 1144 + }, + { + "epoch": 0.08, + "learning_rate": 9.400656814449918e-06, + "loss": 1.0627, + "step": 1145 + }, + { + "epoch": 0.08, + "learning_rate": 9.408866995073892e-06, + "loss": 1.2042, + "step": 1146 + }, + { + "epoch": 0.08, + "learning_rate": 9.417077175697866e-06, + "loss": 1.0942, + "step": 1147 + }, + { + "epoch": 0.08, + "learning_rate": 9.42528735632184e-06, + "loss": 1.0228, + "step": 1148 + }, + { + "epoch": 0.08, + "learning_rate": 9.433497536945814e-06, + "loss": 1.1872, + "step": 1149 + }, + { + "epoch": 0.08, + "learning_rate": 9.441707717569786e-06, + "loss": 1.1051, + "step": 1150 + }, + { + "epoch": 0.09, + "learning_rate": 9.449917898193762e-06, + "loss": 1.1229, + "step": 1151 + }, + { + "epoch": 0.09, + "learning_rate": 9.458128078817734e-06, + "loss": 1.0681, + "step": 1152 + }, + { + "epoch": 0.09, + "learning_rate": 9.46633825944171e-06, + "loss": 1.1262, + "step": 1153 + }, + { + "epoch": 0.09, + "learning_rate": 9.474548440065682e-06, + "loss": 1.1337, + "step": 1154 + }, + { + "epoch": 0.09, + "learning_rate": 9.482758620689655e-06, + "loss": 1.166, + "step": 1155 + }, + { + "epoch": 0.09, + "learning_rate": 9.49096880131363e-06, + "loss": 1.0853, + "step": 1156 + }, + { + "epoch": 0.09, + "learning_rate": 9.499178981937603e-06, + "loss": 1.1658, + "step": 1157 + }, + { + "epoch": 0.09, + "learning_rate": 9.507389162561577e-06, + "loss": 1.2017, + "step": 1158 + }, + { + "epoch": 0.09, + "learning_rate": 9.515599343185551e-06, + "loss": 1.068, + "step": 1159 + }, + { + "epoch": 0.09, + "learning_rate": 9.523809523809525e-06, + "loss": 1.1877, + "step": 1160 + }, + { + "epoch": 0.09, + "learning_rate": 9.532019704433499e-06, + "loss": 1.2378, + "step": 1161 + }, + { + "epoch": 0.09, + "learning_rate": 9.54022988505747e-06, + "loss": 1.0918, + "step": 1162 + }, + { + "epoch": 0.09, + "learning_rate": 9.548440065681446e-06, + "loss": 1.0987, + "step": 1163 + }, + { + "epoch": 0.09, + "learning_rate": 9.55665024630542e-06, + "loss": 1.1604, + "step": 1164 + }, + { + "epoch": 0.09, + "learning_rate": 9.564860426929392e-06, + "loss": 1.225, + "step": 1165 + }, + { + "epoch": 0.09, + "learning_rate": 9.573070607553368e-06, + "loss": 1.1617, + "step": 1166 + }, + { + "epoch": 0.09, + "learning_rate": 9.58128078817734e-06, + "loss": 1.0908, + "step": 1167 + }, + { + "epoch": 0.09, + "learning_rate": 9.589490968801316e-06, + "loss": 1.2207, + "step": 1168 + }, + { + "epoch": 0.09, + "learning_rate": 9.597701149425288e-06, + "loss": 1.0716, + "step": 1169 + }, + { + "epoch": 0.09, + "learning_rate": 9.605911330049262e-06, + "loss": 1.1811, + "step": 1170 + }, + { + "epoch": 0.09, + "learning_rate": 9.614121510673236e-06, + "loss": 1.1038, + "step": 1171 + }, + { + "epoch": 0.09, + "learning_rate": 9.62233169129721e-06, + "loss": 1.1273, + "step": 1172 + }, + { + "epoch": 0.09, + "learning_rate": 9.630541871921183e-06, + "loss": 1.2651, + "step": 1173 + }, + { + "epoch": 0.09, + "learning_rate": 9.638752052545157e-06, + "loss": 1.1686, + "step": 1174 + }, + { + "epoch": 0.09, + "learning_rate": 9.646962233169131e-06, + "loss": 1.1727, + "step": 1175 + }, + { + "epoch": 0.09, + "learning_rate": 9.655172413793105e-06, + "loss": 1.1474, + "step": 1176 + }, + { + "epoch": 0.09, + "learning_rate": 9.663382594417077e-06, + "loss": 1.1605, + "step": 1177 + }, + { + "epoch": 0.09, + "learning_rate": 9.671592775041053e-06, + "loss": 1.2537, + "step": 1178 + }, + { + "epoch": 0.09, + "learning_rate": 9.679802955665025e-06, + "loss": 1.1129, + "step": 1179 + }, + { + "epoch": 0.09, + "learning_rate": 9.688013136288999e-06, + "loss": 1.1865, + "step": 1180 + }, + { + "epoch": 0.09, + "learning_rate": 9.696223316912973e-06, + "loss": 1.0846, + "step": 1181 + }, + { + "epoch": 0.09, + "learning_rate": 9.704433497536947e-06, + "loss": 1.2985, + "step": 1182 + }, + { + "epoch": 0.09, + "learning_rate": 9.71264367816092e-06, + "loss": 0.9723, + "step": 1183 + }, + { + "epoch": 0.09, + "learning_rate": 9.720853858784894e-06, + "loss": 1.2013, + "step": 1184 + }, + { + "epoch": 0.09, + "learning_rate": 9.729064039408868e-06, + "loss": 1.1594, + "step": 1185 + }, + { + "epoch": 0.09, + "learning_rate": 9.737274220032842e-06, + "loss": 1.1491, + "step": 1186 + }, + { + "epoch": 0.09, + "learning_rate": 9.745484400656814e-06, + "loss": 1.1503, + "step": 1187 + }, + { + "epoch": 0.09, + "learning_rate": 9.75369458128079e-06, + "loss": 1.182, + "step": 1188 + }, + { + "epoch": 0.09, + "learning_rate": 9.761904761904762e-06, + "loss": 1.1216, + "step": 1189 + }, + { + "epoch": 0.09, + "learning_rate": 9.770114942528738e-06, + "loss": 1.0784, + "step": 1190 + }, + { + "epoch": 0.09, + "learning_rate": 9.77832512315271e-06, + "loss": 1.0924, + "step": 1191 + }, + { + "epoch": 0.09, + "learning_rate": 9.786535303776684e-06, + "loss": 1.1591, + "step": 1192 + }, + { + "epoch": 0.09, + "learning_rate": 9.794745484400657e-06, + "loss": 1.2109, + "step": 1193 + }, + { + "epoch": 0.09, + "learning_rate": 9.802955665024631e-06, + "loss": 1.1087, + "step": 1194 + }, + { + "epoch": 0.09, + "learning_rate": 9.811165845648605e-06, + "loss": 1.1109, + "step": 1195 + }, + { + "epoch": 0.09, + "learning_rate": 9.819376026272579e-06, + "loss": 1.1895, + "step": 1196 + }, + { + "epoch": 0.09, + "learning_rate": 9.827586206896553e-06, + "loss": 1.1067, + "step": 1197 + }, + { + "epoch": 0.09, + "learning_rate": 9.835796387520527e-06, + "loss": 1.1187, + "step": 1198 + }, + { + "epoch": 0.09, + "learning_rate": 9.844006568144499e-06, + "loss": 1.2043, + "step": 1199 + }, + { + "epoch": 0.09, + "learning_rate": 9.852216748768475e-06, + "loss": 1.2779, + "step": 1200 + }, + { + "epoch": 0.09, + "learning_rate": 9.860426929392447e-06, + "loss": 1.1898, + "step": 1201 + }, + { + "epoch": 0.09, + "learning_rate": 9.86863711001642e-06, + "loss": 1.092, + "step": 1202 + }, + { + "epoch": 0.09, + "learning_rate": 9.876847290640394e-06, + "loss": 1.219, + "step": 1203 + }, + { + "epoch": 0.09, + "learning_rate": 9.885057471264368e-06, + "loss": 1.1867, + "step": 1204 + }, + { + "epoch": 0.09, + "learning_rate": 9.893267651888342e-06, + "loss": 1.0955, + "step": 1205 + }, + { + "epoch": 0.09, + "learning_rate": 9.901477832512316e-06, + "loss": 1.1398, + "step": 1206 + }, + { + "epoch": 0.09, + "learning_rate": 9.90968801313629e-06, + "loss": 1.0293, + "step": 1207 + }, + { + "epoch": 0.09, + "learning_rate": 9.917898193760264e-06, + "loss": 1.1001, + "step": 1208 + }, + { + "epoch": 0.09, + "learning_rate": 9.926108374384236e-06, + "loss": 1.1607, + "step": 1209 + }, + { + "epoch": 0.09, + "learning_rate": 9.934318555008212e-06, + "loss": 1.1388, + "step": 1210 + }, + { + "epoch": 0.09, + "learning_rate": 9.942528735632184e-06, + "loss": 1.1099, + "step": 1211 + }, + { + "epoch": 0.09, + "learning_rate": 9.95073891625616e-06, + "loss": 1.1468, + "step": 1212 + }, + { + "epoch": 0.09, + "learning_rate": 9.958949096880131e-06, + "loss": 1.0997, + "step": 1213 + }, + { + "epoch": 0.09, + "learning_rate": 9.967159277504105e-06, + "loss": 1.1557, + "step": 1214 + }, + { + "epoch": 0.09, + "learning_rate": 9.97536945812808e-06, + "loss": 1.1463, + "step": 1215 + }, + { + "epoch": 0.09, + "learning_rate": 9.983579638752053e-06, + "loss": 1.1269, + "step": 1216 + }, + { + "epoch": 0.09, + "learning_rate": 9.991789819376027e-06, + "loss": 1.1426, + "step": 1217 + }, + { + "epoch": 0.09, + "learning_rate": 1e-05, + "loss": 1.1863, + "step": 1218 + }, + { + "epoch": 0.09, + "learning_rate": 1.0008210180623975e-05, + "loss": 1.1279, + "step": 1219 + }, + { + "epoch": 0.09, + "learning_rate": 1.0016420361247949e-05, + "loss": 1.0453, + "step": 1220 + }, + { + "epoch": 0.09, + "learning_rate": 1.0024630541871922e-05, + "loss": 0.9988, + "step": 1221 + }, + { + "epoch": 0.09, + "learning_rate": 1.0032840722495896e-05, + "loss": 1.1, + "step": 1222 + }, + { + "epoch": 0.09, + "learning_rate": 1.0041050903119868e-05, + "loss": 1.0857, + "step": 1223 + }, + { + "epoch": 0.09, + "learning_rate": 1.0049261083743844e-05, + "loss": 1.184, + "step": 1224 + }, + { + "epoch": 0.09, + "learning_rate": 1.0057471264367818e-05, + "loss": 0.987, + "step": 1225 + }, + { + "epoch": 0.09, + "learning_rate": 1.006568144499179e-05, + "loss": 1.1339, + "step": 1226 + }, + { + "epoch": 0.09, + "learning_rate": 1.0073891625615764e-05, + "loss": 1.0874, + "step": 1227 + }, + { + "epoch": 0.09, + "learning_rate": 1.008210180623974e-05, + "loss": 1.1019, + "step": 1228 + }, + { + "epoch": 0.09, + "learning_rate": 1.0090311986863712e-05, + "loss": 0.9902, + "step": 1229 + }, + { + "epoch": 0.09, + "learning_rate": 1.0098522167487686e-05, + "loss": 1.1709, + "step": 1230 + }, + { + "epoch": 0.09, + "learning_rate": 1.0106732348111658e-05, + "loss": 1.1254, + "step": 1231 + }, + { + "epoch": 0.09, + "learning_rate": 1.0114942528735633e-05, + "loss": 1.2288, + "step": 1232 + }, + { + "epoch": 0.09, + "learning_rate": 1.0123152709359607e-05, + "loss": 1.0652, + "step": 1233 + }, + { + "epoch": 0.09, + "learning_rate": 1.0131362889983581e-05, + "loss": 1.1061, + "step": 1234 + }, + { + "epoch": 0.09, + "learning_rate": 1.0139573070607553e-05, + "loss": 1.1124, + "step": 1235 + }, + { + "epoch": 0.09, + "learning_rate": 1.0147783251231529e-05, + "loss": 1.223, + "step": 1236 + }, + { + "epoch": 0.09, + "learning_rate": 1.0155993431855503e-05, + "loss": 1.1398, + "step": 1237 + }, + { + "epoch": 0.09, + "learning_rate": 1.0164203612479475e-05, + "loss": 1.2002, + "step": 1238 + }, + { + "epoch": 0.09, + "learning_rate": 1.0172413793103449e-05, + "loss": 1.0276, + "step": 1239 + }, + { + "epoch": 0.09, + "learning_rate": 1.0180623973727424e-05, + "loss": 1.1301, + "step": 1240 + }, + { + "epoch": 0.09, + "learning_rate": 1.0188834154351396e-05, + "loss": 1.0887, + "step": 1241 + }, + { + "epoch": 0.09, + "learning_rate": 1.019704433497537e-05, + "loss": 1.0719, + "step": 1242 + }, + { + "epoch": 0.09, + "learning_rate": 1.0205254515599343e-05, + "loss": 1.2043, + "step": 1243 + }, + { + "epoch": 0.09, + "learning_rate": 1.0213464696223318e-05, + "loss": 1.0813, + "step": 1244 + }, + { + "epoch": 0.09, + "learning_rate": 1.0221674876847292e-05, + "loss": 1.1224, + "step": 1245 + }, + { + "epoch": 0.09, + "learning_rate": 1.0229885057471264e-05, + "loss": 1.0918, + "step": 1246 + }, + { + "epoch": 0.09, + "learning_rate": 1.0238095238095238e-05, + "loss": 1.2287, + "step": 1247 + }, + { + "epoch": 0.09, + "learning_rate": 1.0246305418719214e-05, + "loss": 1.1653, + "step": 1248 + }, + { + "epoch": 0.09, + "learning_rate": 1.0254515599343187e-05, + "loss": 1.1018, + "step": 1249 + }, + { + "epoch": 0.09, + "learning_rate": 1.026272577996716e-05, + "loss": 1.1512, + "step": 1250 + }, + { + "epoch": 0.09, + "learning_rate": 1.0270935960591133e-05, + "loss": 1.1139, + "step": 1251 + }, + { + "epoch": 0.09, + "learning_rate": 1.0279146141215109e-05, + "loss": 1.1196, + "step": 1252 + }, + { + "epoch": 0.09, + "learning_rate": 1.0287356321839081e-05, + "loss": 1.1334, + "step": 1253 + }, + { + "epoch": 0.09, + "learning_rate": 1.0295566502463055e-05, + "loss": 1.1246, + "step": 1254 + }, + { + "epoch": 0.09, + "learning_rate": 1.0303776683087027e-05, + "loss": 1.0626, + "step": 1255 + }, + { + "epoch": 0.09, + "learning_rate": 1.0311986863711003e-05, + "loss": 1.1359, + "step": 1256 + }, + { + "epoch": 0.09, + "learning_rate": 1.0320197044334977e-05, + "loss": 1.1569, + "step": 1257 + }, + { + "epoch": 0.09, + "learning_rate": 1.0328407224958949e-05, + "loss": 1.1427, + "step": 1258 + }, + { + "epoch": 0.09, + "learning_rate": 1.0336617405582923e-05, + "loss": 1.0741, + "step": 1259 + }, + { + "epoch": 0.09, + "learning_rate": 1.0344827586206898e-05, + "loss": 1.1872, + "step": 1260 + }, + { + "epoch": 0.09, + "learning_rate": 1.035303776683087e-05, + "loss": 1.0531, + "step": 1261 + }, + { + "epoch": 0.09, + "learning_rate": 1.0361247947454844e-05, + "loss": 1.0647, + "step": 1262 + }, + { + "epoch": 0.09, + "learning_rate": 1.0369458128078818e-05, + "loss": 1.079, + "step": 1263 + }, + { + "epoch": 0.09, + "learning_rate": 1.0377668308702792e-05, + "loss": 1.1126, + "step": 1264 + }, + { + "epoch": 0.09, + "learning_rate": 1.0385878489326766e-05, + "loss": 1.0773, + "step": 1265 + }, + { + "epoch": 0.09, + "learning_rate": 1.039408866995074e-05, + "loss": 1.1247, + "step": 1266 + }, + { + "epoch": 0.09, + "learning_rate": 1.0402298850574712e-05, + "loss": 1.1435, + "step": 1267 + }, + { + "epoch": 0.09, + "learning_rate": 1.0410509031198688e-05, + "loss": 1.2098, + "step": 1268 + }, + { + "epoch": 0.09, + "learning_rate": 1.0418719211822661e-05, + "loss": 1.2249, + "step": 1269 + }, + { + "epoch": 0.09, + "learning_rate": 1.0426929392446634e-05, + "loss": 1.2085, + "step": 1270 + }, + { + "epoch": 0.09, + "learning_rate": 1.0435139573070608e-05, + "loss": 1.0982, + "step": 1271 + }, + { + "epoch": 0.09, + "learning_rate": 1.0443349753694583e-05, + "loss": 1.1284, + "step": 1272 + }, + { + "epoch": 0.09, + "learning_rate": 1.0451559934318555e-05, + "loss": 0.9674, + "step": 1273 + }, + { + "epoch": 0.09, + "learning_rate": 1.0459770114942529e-05, + "loss": 1.2653, + "step": 1274 + }, + { + "epoch": 0.09, + "learning_rate": 1.0467980295566505e-05, + "loss": 1.1617, + "step": 1275 + }, + { + "epoch": 0.09, + "learning_rate": 1.0476190476190477e-05, + "loss": 1.1914, + "step": 1276 + }, + { + "epoch": 0.09, + "learning_rate": 1.048440065681445e-05, + "loss": 1.2225, + "step": 1277 + }, + { + "epoch": 0.09, + "learning_rate": 1.0492610837438425e-05, + "loss": 1.1476, + "step": 1278 + }, + { + "epoch": 0.09, + "learning_rate": 1.0500821018062399e-05, + "loss": 1.1531, + "step": 1279 + }, + { + "epoch": 0.09, + "learning_rate": 1.0509031198686372e-05, + "loss": 1.1128, + "step": 1280 + }, + { + "epoch": 0.09, + "learning_rate": 1.0517241379310346e-05, + "loss": 1.1875, + "step": 1281 + }, + { + "epoch": 0.09, + "learning_rate": 1.0525451559934318e-05, + "loss": 1.1474, + "step": 1282 + }, + { + "epoch": 0.09, + "learning_rate": 1.0533661740558294e-05, + "loss": 1.0462, + "step": 1283 + }, + { + "epoch": 0.09, + "learning_rate": 1.0541871921182268e-05, + "loss": 1.1008, + "step": 1284 + }, + { + "epoch": 0.09, + "learning_rate": 1.055008210180624e-05, + "loss": 1.148, + "step": 1285 + }, + { + "epoch": 0.1, + "learning_rate": 1.0558292282430214e-05, + "loss": 1.1413, + "step": 1286 + }, + { + "epoch": 0.1, + "learning_rate": 1.056650246305419e-05, + "loss": 1.1202, + "step": 1287 + }, + { + "epoch": 0.1, + "learning_rate": 1.0574712643678162e-05, + "loss": 1.1602, + "step": 1288 + }, + { + "epoch": 0.1, + "learning_rate": 1.0582922824302136e-05, + "loss": 1.1029, + "step": 1289 + }, + { + "epoch": 0.1, + "learning_rate": 1.0591133004926108e-05, + "loss": 1.1554, + "step": 1290 + }, + { + "epoch": 0.1, + "learning_rate": 1.0599343185550083e-05, + "loss": 1.1638, + "step": 1291 + }, + { + "epoch": 0.1, + "learning_rate": 1.0607553366174057e-05, + "loss": 1.2071, + "step": 1292 + }, + { + "epoch": 0.1, + "learning_rate": 1.0615763546798031e-05, + "loss": 1.1104, + "step": 1293 + }, + { + "epoch": 0.1, + "learning_rate": 1.0623973727422003e-05, + "loss": 1.0697, + "step": 1294 + }, + { + "epoch": 0.1, + "learning_rate": 1.0632183908045979e-05, + "loss": 1.193, + "step": 1295 + }, + { + "epoch": 0.1, + "learning_rate": 1.0640394088669953e-05, + "loss": 1.1709, + "step": 1296 + }, + { + "epoch": 0.1, + "learning_rate": 1.0648604269293925e-05, + "loss": 1.1915, + "step": 1297 + }, + { + "epoch": 0.1, + "learning_rate": 1.0656814449917899e-05, + "loss": 1.0579, + "step": 1298 + }, + { + "epoch": 0.1, + "learning_rate": 1.0665024630541874e-05, + "loss": 1.1092, + "step": 1299 + }, + { + "epoch": 0.1, + "learning_rate": 1.0673234811165846e-05, + "loss": 1.1521, + "step": 1300 + }, + { + "epoch": 0.1, + "learning_rate": 1.068144499178982e-05, + "loss": 1.1342, + "step": 1301 + }, + { + "epoch": 0.1, + "learning_rate": 1.0689655172413792e-05, + "loss": 1.1147, + "step": 1302 + }, + { + "epoch": 0.1, + "learning_rate": 1.0697865353037768e-05, + "loss": 0.9733, + "step": 1303 + }, + { + "epoch": 0.1, + "learning_rate": 1.0706075533661742e-05, + "loss": 1.1588, + "step": 1304 + }, + { + "epoch": 0.1, + "learning_rate": 1.0714285714285714e-05, + "loss": 1.1829, + "step": 1305 + }, + { + "epoch": 0.1, + "learning_rate": 1.0722495894909688e-05, + "loss": 1.2291, + "step": 1306 + }, + { + "epoch": 0.1, + "learning_rate": 1.0730706075533664e-05, + "loss": 1.2536, + "step": 1307 + }, + { + "epoch": 0.1, + "learning_rate": 1.0738916256157637e-05, + "loss": 1.0566, + "step": 1308 + }, + { + "epoch": 0.1, + "learning_rate": 1.074712643678161e-05, + "loss": 1.207, + "step": 1309 + }, + { + "epoch": 0.1, + "learning_rate": 1.0755336617405583e-05, + "loss": 1.0036, + "step": 1310 + }, + { + "epoch": 0.1, + "learning_rate": 1.0763546798029559e-05, + "loss": 1.1649, + "step": 1311 + }, + { + "epoch": 0.1, + "learning_rate": 1.0771756978653531e-05, + "loss": 1.0363, + "step": 1312 + }, + { + "epoch": 0.1, + "learning_rate": 1.0779967159277505e-05, + "loss": 1.0958, + "step": 1313 + }, + { + "epoch": 0.1, + "learning_rate": 1.0788177339901477e-05, + "loss": 1.1362, + "step": 1314 + }, + { + "epoch": 0.1, + "learning_rate": 1.0796387520525453e-05, + "loss": 1.1253, + "step": 1315 + }, + { + "epoch": 0.1, + "learning_rate": 1.0804597701149427e-05, + "loss": 1.0649, + "step": 1316 + }, + { + "epoch": 0.1, + "learning_rate": 1.0812807881773399e-05, + "loss": 1.134, + "step": 1317 + }, + { + "epoch": 0.1, + "learning_rate": 1.0821018062397373e-05, + "loss": 1.0752, + "step": 1318 + }, + { + "epoch": 0.1, + "learning_rate": 1.0829228243021348e-05, + "loss": 1.1313, + "step": 1319 + }, + { + "epoch": 0.1, + "learning_rate": 1.083743842364532e-05, + "loss": 1.0658, + "step": 1320 + }, + { + "epoch": 0.1, + "learning_rate": 1.0845648604269294e-05, + "loss": 1.185, + "step": 1321 + }, + { + "epoch": 0.1, + "learning_rate": 1.0853858784893268e-05, + "loss": 1.0199, + "step": 1322 + }, + { + "epoch": 0.1, + "learning_rate": 1.0862068965517242e-05, + "loss": 1.1725, + "step": 1323 + }, + { + "epoch": 0.1, + "learning_rate": 1.0870279146141216e-05, + "loss": 1.1283, + "step": 1324 + }, + { + "epoch": 0.1, + "learning_rate": 1.087848932676519e-05, + "loss": 1.1726, + "step": 1325 + }, + { + "epoch": 0.1, + "learning_rate": 1.0886699507389162e-05, + "loss": 1.152, + "step": 1326 + }, + { + "epoch": 0.1, + "learning_rate": 1.0894909688013138e-05, + "loss": 1.3263, + "step": 1327 + }, + { + "epoch": 0.1, + "learning_rate": 1.0903119868637111e-05, + "loss": 1.1399, + "step": 1328 + }, + { + "epoch": 0.1, + "learning_rate": 1.0911330049261084e-05, + "loss": 1.209, + "step": 1329 + }, + { + "epoch": 0.1, + "learning_rate": 1.091954022988506e-05, + "loss": 1.2682, + "step": 1330 + }, + { + "epoch": 0.1, + "learning_rate": 1.0927750410509033e-05, + "loss": 1.0411, + "step": 1331 + }, + { + "epoch": 0.1, + "learning_rate": 1.0935960591133005e-05, + "loss": 1.0698, + "step": 1332 + }, + { + "epoch": 0.1, + "learning_rate": 1.0944170771756979e-05, + "loss": 1.0379, + "step": 1333 + }, + { + "epoch": 0.1, + "learning_rate": 1.0952380952380955e-05, + "loss": 1.0815, + "step": 1334 + }, + { + "epoch": 0.1, + "learning_rate": 1.0960591133004927e-05, + "loss": 1.1186, + "step": 1335 + }, + { + "epoch": 0.1, + "learning_rate": 1.09688013136289e-05, + "loss": 1.0774, + "step": 1336 + }, + { + "epoch": 0.1, + "learning_rate": 1.0977011494252875e-05, + "loss": 1.2344, + "step": 1337 + }, + { + "epoch": 0.1, + "learning_rate": 1.0985221674876848e-05, + "loss": 1.1412, + "step": 1338 + }, + { + "epoch": 0.1, + "learning_rate": 1.0993431855500822e-05, + "loss": 1.0734, + "step": 1339 + }, + { + "epoch": 0.1, + "learning_rate": 1.1001642036124796e-05, + "loss": 1.1005, + "step": 1340 + }, + { + "epoch": 0.1, + "learning_rate": 1.1009852216748768e-05, + "loss": 1.1744, + "step": 1341 + }, + { + "epoch": 0.1, + "learning_rate": 1.1018062397372744e-05, + "loss": 1.1229, + "step": 1342 + }, + { + "epoch": 0.1, + "learning_rate": 1.1026272577996718e-05, + "loss": 1.181, + "step": 1343 + }, + { + "epoch": 0.1, + "learning_rate": 1.103448275862069e-05, + "loss": 1.0716, + "step": 1344 + }, + { + "epoch": 0.1, + "learning_rate": 1.1042692939244664e-05, + "loss": 1.2162, + "step": 1345 + }, + { + "epoch": 0.1, + "learning_rate": 1.105090311986864e-05, + "loss": 1.1926, + "step": 1346 + }, + { + "epoch": 0.1, + "learning_rate": 1.1059113300492612e-05, + "loss": 1.0868, + "step": 1347 + }, + { + "epoch": 0.1, + "learning_rate": 1.1067323481116585e-05, + "loss": 0.9937, + "step": 1348 + }, + { + "epoch": 0.1, + "learning_rate": 1.1075533661740558e-05, + "loss": 1.1743, + "step": 1349 + }, + { + "epoch": 0.1, + "learning_rate": 1.1083743842364533e-05, + "loss": 1.1057, + "step": 1350 + }, + { + "epoch": 0.1, + "learning_rate": 1.1091954022988507e-05, + "loss": 1.096, + "step": 1351 + }, + { + "epoch": 0.1, + "learning_rate": 1.1100164203612481e-05, + "loss": 1.1884, + "step": 1352 + }, + { + "epoch": 0.1, + "learning_rate": 1.1108374384236453e-05, + "loss": 1.1071, + "step": 1353 + }, + { + "epoch": 0.1, + "learning_rate": 1.1116584564860429e-05, + "loss": 1.189, + "step": 1354 + }, + { + "epoch": 0.1, + "learning_rate": 1.1124794745484403e-05, + "loss": 1.1796, + "step": 1355 + }, + { + "epoch": 0.1, + "learning_rate": 1.1133004926108375e-05, + "loss": 1.1107, + "step": 1356 + }, + { + "epoch": 0.1, + "learning_rate": 1.1141215106732349e-05, + "loss": 1.1456, + "step": 1357 + }, + { + "epoch": 0.1, + "learning_rate": 1.1149425287356324e-05, + "loss": 1.137, + "step": 1358 + }, + { + "epoch": 0.1, + "learning_rate": 1.1157635467980296e-05, + "loss": 1.1666, + "step": 1359 + }, + { + "epoch": 0.1, + "learning_rate": 1.116584564860427e-05, + "loss": 1.0558, + "step": 1360 + }, + { + "epoch": 0.1, + "learning_rate": 1.1174055829228242e-05, + "loss": 1.1548, + "step": 1361 + }, + { + "epoch": 0.1, + "learning_rate": 1.1182266009852218e-05, + "loss": 1.0357, + "step": 1362 + }, + { + "epoch": 0.1, + "learning_rate": 1.1190476190476192e-05, + "loss": 1.084, + "step": 1363 + }, + { + "epoch": 0.1, + "learning_rate": 1.1198686371100164e-05, + "loss": 1.0963, + "step": 1364 + }, + { + "epoch": 0.1, + "learning_rate": 1.1206896551724138e-05, + "loss": 1.0304, + "step": 1365 + }, + { + "epoch": 0.1, + "learning_rate": 1.1215106732348113e-05, + "loss": 1.1982, + "step": 1366 + }, + { + "epoch": 0.1, + "learning_rate": 1.1223316912972087e-05, + "loss": 1.126, + "step": 1367 + }, + { + "epoch": 0.1, + "learning_rate": 1.123152709359606e-05, + "loss": 1.1293, + "step": 1368 + }, + { + "epoch": 0.1, + "learning_rate": 1.1239737274220033e-05, + "loss": 1.0886, + "step": 1369 + }, + { + "epoch": 0.1, + "learning_rate": 1.1247947454844009e-05, + "loss": 1.141, + "step": 1370 + }, + { + "epoch": 0.1, + "learning_rate": 1.1256157635467981e-05, + "loss": 1.0972, + "step": 1371 + }, + { + "epoch": 0.1, + "learning_rate": 1.1264367816091955e-05, + "loss": 1.1435, + "step": 1372 + }, + { + "epoch": 0.1, + "learning_rate": 1.1272577996715927e-05, + "loss": 1.0957, + "step": 1373 + }, + { + "epoch": 0.1, + "learning_rate": 1.1280788177339903e-05, + "loss": 1.0247, + "step": 1374 + }, + { + "epoch": 0.1, + "learning_rate": 1.1288998357963877e-05, + "loss": 1.1672, + "step": 1375 + }, + { + "epoch": 0.1, + "learning_rate": 1.1297208538587849e-05, + "loss": 1.154, + "step": 1376 + }, + { + "epoch": 0.1, + "learning_rate": 1.1305418719211823e-05, + "loss": 1.3261, + "step": 1377 + }, + { + "epoch": 0.1, + "learning_rate": 1.1313628899835798e-05, + "loss": 1.1186, + "step": 1378 + }, + { + "epoch": 0.1, + "learning_rate": 1.132183908045977e-05, + "loss": 1.0863, + "step": 1379 + }, + { + "epoch": 0.1, + "learning_rate": 1.1330049261083744e-05, + "loss": 1.1855, + "step": 1380 + }, + { + "epoch": 0.1, + "learning_rate": 1.1338259441707718e-05, + "loss": 1.1302, + "step": 1381 + }, + { + "epoch": 0.1, + "learning_rate": 1.1346469622331692e-05, + "loss": 1.1434, + "step": 1382 + }, + { + "epoch": 0.1, + "learning_rate": 1.1354679802955666e-05, + "loss": 1.1167, + "step": 1383 + }, + { + "epoch": 0.1, + "learning_rate": 1.136288998357964e-05, + "loss": 1.0306, + "step": 1384 + }, + { + "epoch": 0.1, + "learning_rate": 1.1371100164203615e-05, + "loss": 1.0668, + "step": 1385 + }, + { + "epoch": 0.1, + "learning_rate": 1.1379310344827587e-05, + "loss": 1.12, + "step": 1386 + }, + { + "epoch": 0.1, + "learning_rate": 1.1387520525451561e-05, + "loss": 1.0979, + "step": 1387 + }, + { + "epoch": 0.1, + "learning_rate": 1.1395730706075534e-05, + "loss": 1.1505, + "step": 1388 + }, + { + "epoch": 0.1, + "learning_rate": 1.1403940886699509e-05, + "loss": 1.2124, + "step": 1389 + }, + { + "epoch": 0.1, + "learning_rate": 1.1412151067323483e-05, + "loss": 1.1224, + "step": 1390 + }, + { + "epoch": 0.1, + "learning_rate": 1.1420361247947455e-05, + "loss": 1.1405, + "step": 1391 + }, + { + "epoch": 0.1, + "learning_rate": 1.1428571428571429e-05, + "loss": 1.2091, + "step": 1392 + }, + { + "epoch": 0.1, + "learning_rate": 1.1436781609195405e-05, + "loss": 1.1991, + "step": 1393 + }, + { + "epoch": 0.1, + "learning_rate": 1.1444991789819377e-05, + "loss": 1.0758, + "step": 1394 + }, + { + "epoch": 0.1, + "learning_rate": 1.145320197044335e-05, + "loss": 1.0822, + "step": 1395 + }, + { + "epoch": 0.1, + "learning_rate": 1.1461412151067325e-05, + "loss": 1.1424, + "step": 1396 + }, + { + "epoch": 0.1, + "learning_rate": 1.1469622331691298e-05, + "loss": 1.0855, + "step": 1397 + }, + { + "epoch": 0.1, + "learning_rate": 1.1477832512315272e-05, + "loss": 1.1497, + "step": 1398 + }, + { + "epoch": 0.1, + "learning_rate": 1.1486042692939246e-05, + "loss": 1.1261, + "step": 1399 + }, + { + "epoch": 0.1, + "learning_rate": 1.1494252873563218e-05, + "loss": 1.0366, + "step": 1400 + }, + { + "epoch": 0.1, + "learning_rate": 1.1502463054187194e-05, + "loss": 1.093, + "step": 1401 + }, + { + "epoch": 0.1, + "learning_rate": 1.1510673234811168e-05, + "loss": 1.1573, + "step": 1402 + }, + { + "epoch": 0.1, + "learning_rate": 1.151888341543514e-05, + "loss": 1.1018, + "step": 1403 + }, + { + "epoch": 0.1, + "learning_rate": 1.1527093596059114e-05, + "loss": 1.2228, + "step": 1404 + }, + { + "epoch": 0.1, + "learning_rate": 1.153530377668309e-05, + "loss": 1.026, + "step": 1405 + }, + { + "epoch": 0.1, + "learning_rate": 1.1543513957307062e-05, + "loss": 1.1432, + "step": 1406 + }, + { + "epoch": 0.1, + "learning_rate": 1.1551724137931035e-05, + "loss": 1.0921, + "step": 1407 + }, + { + "epoch": 0.1, + "learning_rate": 1.1559934318555008e-05, + "loss": 1.052, + "step": 1408 + }, + { + "epoch": 0.1, + "learning_rate": 1.1568144499178983e-05, + "loss": 1.0209, + "step": 1409 + }, + { + "epoch": 0.1, + "learning_rate": 1.1576354679802957e-05, + "loss": 1.0898, + "step": 1410 + }, + { + "epoch": 0.1, + "learning_rate": 1.1584564860426931e-05, + "loss": 1.1602, + "step": 1411 + }, + { + "epoch": 0.1, + "learning_rate": 1.1592775041050903e-05, + "loss": 1.153, + "step": 1412 + }, + { + "epoch": 0.1, + "learning_rate": 1.1600985221674879e-05, + "loss": 1.1421, + "step": 1413 + }, + { + "epoch": 0.1, + "learning_rate": 1.1609195402298852e-05, + "loss": 1.1229, + "step": 1414 + }, + { + "epoch": 0.1, + "learning_rate": 1.1617405582922825e-05, + "loss": 1.1243, + "step": 1415 + }, + { + "epoch": 0.1, + "learning_rate": 1.1625615763546799e-05, + "loss": 1.0238, + "step": 1416 + }, + { + "epoch": 0.1, + "learning_rate": 1.1633825944170774e-05, + "loss": 1.1784, + "step": 1417 + }, + { + "epoch": 0.1, + "learning_rate": 1.1642036124794746e-05, + "loss": 1.1362, + "step": 1418 + }, + { + "epoch": 0.1, + "learning_rate": 1.165024630541872e-05, + "loss": 1.167, + "step": 1419 + }, + { + "epoch": 0.1, + "learning_rate": 1.1658456486042692e-05, + "loss": 1.0888, + "step": 1420 + }, + { + "epoch": 0.1, + "learning_rate": 1.1666666666666668e-05, + "loss": 1.1074, + "step": 1421 + }, + { + "epoch": 0.11, + "learning_rate": 1.1674876847290642e-05, + "loss": 1.1013, + "step": 1422 + }, + { + "epoch": 0.11, + "learning_rate": 1.1683087027914614e-05, + "loss": 0.9803, + "step": 1423 + }, + { + "epoch": 0.11, + "learning_rate": 1.1691297208538588e-05, + "loss": 0.9811, + "step": 1424 + }, + { + "epoch": 0.11, + "learning_rate": 1.1699507389162563e-05, + "loss": 1.0489, + "step": 1425 + }, + { + "epoch": 0.11, + "learning_rate": 1.1707717569786536e-05, + "loss": 1.1189, + "step": 1426 + }, + { + "epoch": 0.11, + "learning_rate": 1.171592775041051e-05, + "loss": 1.1876, + "step": 1427 + }, + { + "epoch": 0.11, + "learning_rate": 1.1724137931034483e-05, + "loss": 1.1422, + "step": 1428 + }, + { + "epoch": 0.11, + "learning_rate": 1.1732348111658459e-05, + "loss": 1.0304, + "step": 1429 + }, + { + "epoch": 0.11, + "learning_rate": 1.1740558292282431e-05, + "loss": 1.226, + "step": 1430 + }, + { + "epoch": 0.11, + "learning_rate": 1.1748768472906405e-05, + "loss": 1.1356, + "step": 1431 + }, + { + "epoch": 0.11, + "learning_rate": 1.1756978653530377e-05, + "loss": 1.098, + "step": 1432 + }, + { + "epoch": 0.11, + "learning_rate": 1.1765188834154353e-05, + "loss": 0.9674, + "step": 1433 + }, + { + "epoch": 0.11, + "learning_rate": 1.1773399014778327e-05, + "loss": 1.1659, + "step": 1434 + }, + { + "epoch": 0.11, + "learning_rate": 1.1781609195402299e-05, + "loss": 1.154, + "step": 1435 + }, + { + "epoch": 0.11, + "learning_rate": 1.1789819376026273e-05, + "loss": 1.0531, + "step": 1436 + }, + { + "epoch": 0.11, + "learning_rate": 1.1798029556650248e-05, + "loss": 1.1427, + "step": 1437 + }, + { + "epoch": 0.11, + "learning_rate": 1.180623973727422e-05, + "loss": 1.0966, + "step": 1438 + }, + { + "epoch": 0.11, + "learning_rate": 1.1814449917898194e-05, + "loss": 1.1434, + "step": 1439 + }, + { + "epoch": 0.11, + "learning_rate": 1.182266009852217e-05, + "loss": 1.0721, + "step": 1440 + }, + { + "epoch": 0.11, + "learning_rate": 1.1830870279146142e-05, + "loss": 1.0744, + "step": 1441 + }, + { + "epoch": 0.11, + "learning_rate": 1.1839080459770116e-05, + "loss": 1.159, + "step": 1442 + }, + { + "epoch": 0.11, + "learning_rate": 1.184729064039409e-05, + "loss": 1.0455, + "step": 1443 + }, + { + "epoch": 0.11, + "learning_rate": 1.1855500821018065e-05, + "loss": 1.2254, + "step": 1444 + }, + { + "epoch": 0.11, + "learning_rate": 1.1863711001642037e-05, + "loss": 1.0111, + "step": 1445 + }, + { + "epoch": 0.11, + "learning_rate": 1.1871921182266011e-05, + "loss": 1.0863, + "step": 1446 + }, + { + "epoch": 0.11, + "learning_rate": 1.1880131362889983e-05, + "loss": 1.2063, + "step": 1447 + }, + { + "epoch": 0.11, + "learning_rate": 1.1888341543513959e-05, + "loss": 1.089, + "step": 1448 + }, + { + "epoch": 0.11, + "learning_rate": 1.1896551724137933e-05, + "loss": 1.0609, + "step": 1449 + }, + { + "epoch": 0.11, + "learning_rate": 1.1904761904761905e-05, + "loss": 1.2124, + "step": 1450 + }, + { + "epoch": 0.11, + "learning_rate": 1.1912972085385879e-05, + "loss": 1.1527, + "step": 1451 + }, + { + "epoch": 0.11, + "learning_rate": 1.1921182266009855e-05, + "loss": 1.1741, + "step": 1452 + }, + { + "epoch": 0.11, + "learning_rate": 1.1929392446633827e-05, + "loss": 1.066, + "step": 1453 + }, + { + "epoch": 0.11, + "learning_rate": 1.19376026272578e-05, + "loss": 1.0154, + "step": 1454 + }, + { + "epoch": 0.11, + "learning_rate": 1.1945812807881774e-05, + "loss": 1.1442, + "step": 1455 + }, + { + "epoch": 0.11, + "learning_rate": 1.1954022988505748e-05, + "loss": 1.1043, + "step": 1456 + }, + { + "epoch": 0.11, + "learning_rate": 1.1962233169129722e-05, + "loss": 1.133, + "step": 1457 + }, + { + "epoch": 0.11, + "learning_rate": 1.1970443349753696e-05, + "loss": 1.0705, + "step": 1458 + }, + { + "epoch": 0.11, + "learning_rate": 1.1978653530377668e-05, + "loss": 1.122, + "step": 1459 + }, + { + "epoch": 0.11, + "learning_rate": 1.1986863711001644e-05, + "loss": 0.9535, + "step": 1460 + }, + { + "epoch": 0.11, + "learning_rate": 1.1995073891625618e-05, + "loss": 1.0864, + "step": 1461 + }, + { + "epoch": 0.11, + "learning_rate": 1.200328407224959e-05, + "loss": 1.1569, + "step": 1462 + }, + { + "epoch": 0.11, + "learning_rate": 1.2011494252873564e-05, + "loss": 1.1266, + "step": 1463 + }, + { + "epoch": 0.11, + "learning_rate": 1.201970443349754e-05, + "loss": 1.1511, + "step": 1464 + }, + { + "epoch": 0.11, + "learning_rate": 1.2027914614121511e-05, + "loss": 1.1552, + "step": 1465 + }, + { + "epoch": 0.11, + "learning_rate": 1.2036124794745485e-05, + "loss": 1.032, + "step": 1466 + }, + { + "epoch": 0.11, + "learning_rate": 1.2044334975369458e-05, + "loss": 1.0651, + "step": 1467 + }, + { + "epoch": 0.11, + "learning_rate": 1.2052545155993433e-05, + "loss": 1.0238, + "step": 1468 + }, + { + "epoch": 0.11, + "learning_rate": 1.2060755336617407e-05, + "loss": 1.1028, + "step": 1469 + }, + { + "epoch": 0.11, + "learning_rate": 1.206896551724138e-05, + "loss": 1.2102, + "step": 1470 + }, + { + "epoch": 0.11, + "learning_rate": 1.2077175697865353e-05, + "loss": 1.1256, + "step": 1471 + }, + { + "epoch": 0.11, + "learning_rate": 1.2085385878489329e-05, + "loss": 1.2433, + "step": 1472 + }, + { + "epoch": 0.11, + "learning_rate": 1.2093596059113302e-05, + "loss": 1.1932, + "step": 1473 + }, + { + "epoch": 0.11, + "learning_rate": 1.2101806239737275e-05, + "loss": 1.0521, + "step": 1474 + }, + { + "epoch": 0.11, + "learning_rate": 1.2110016420361248e-05, + "loss": 1.1697, + "step": 1475 + }, + { + "epoch": 0.11, + "learning_rate": 1.2118226600985224e-05, + "loss": 1.1946, + "step": 1476 + }, + { + "epoch": 0.11, + "learning_rate": 1.2126436781609196e-05, + "loss": 1.1323, + "step": 1477 + }, + { + "epoch": 0.11, + "learning_rate": 1.213464696223317e-05, + "loss": 1.0398, + "step": 1478 + }, + { + "epoch": 0.11, + "learning_rate": 1.2142857142857142e-05, + "loss": 1.0217, + "step": 1479 + }, + { + "epoch": 0.11, + "learning_rate": 1.2151067323481118e-05, + "loss": 1.1555, + "step": 1480 + }, + { + "epoch": 0.11, + "learning_rate": 1.2159277504105092e-05, + "loss": 1.0738, + "step": 1481 + }, + { + "epoch": 0.11, + "learning_rate": 1.2167487684729064e-05, + "loss": 1.1415, + "step": 1482 + }, + { + "epoch": 0.11, + "learning_rate": 1.2175697865353038e-05, + "loss": 1.1221, + "step": 1483 + }, + { + "epoch": 0.11, + "learning_rate": 1.2183908045977013e-05, + "loss": 1.1047, + "step": 1484 + }, + { + "epoch": 0.11, + "learning_rate": 1.2192118226600986e-05, + "loss": 1.0992, + "step": 1485 + }, + { + "epoch": 0.11, + "learning_rate": 1.220032840722496e-05, + "loss": 1.1857, + "step": 1486 + }, + { + "epoch": 0.11, + "learning_rate": 1.2208538587848933e-05, + "loss": 1.024, + "step": 1487 + }, + { + "epoch": 0.11, + "learning_rate": 1.2216748768472909e-05, + "loss": 1.1716, + "step": 1488 + }, + { + "epoch": 0.11, + "learning_rate": 1.2224958949096881e-05, + "loss": 1.1248, + "step": 1489 + }, + { + "epoch": 0.11, + "learning_rate": 1.2233169129720855e-05, + "loss": 1.2066, + "step": 1490 + }, + { + "epoch": 0.11, + "learning_rate": 1.2241379310344827e-05, + "loss": 0.9779, + "step": 1491 + }, + { + "epoch": 0.11, + "learning_rate": 1.2249589490968803e-05, + "loss": 1.1786, + "step": 1492 + }, + { + "epoch": 0.11, + "learning_rate": 1.2257799671592776e-05, + "loss": 1.22, + "step": 1493 + }, + { + "epoch": 0.11, + "learning_rate": 1.2266009852216749e-05, + "loss": 1.0909, + "step": 1494 + }, + { + "epoch": 0.11, + "learning_rate": 1.2274220032840724e-05, + "loss": 1.2692, + "step": 1495 + }, + { + "epoch": 0.11, + "learning_rate": 1.2282430213464698e-05, + "loss": 1.2059, + "step": 1496 + }, + { + "epoch": 0.11, + "learning_rate": 1.229064039408867e-05, + "loss": 1.114, + "step": 1497 + }, + { + "epoch": 0.11, + "learning_rate": 1.2298850574712644e-05, + "loss": 1.0984, + "step": 1498 + }, + { + "epoch": 0.11, + "learning_rate": 1.230706075533662e-05, + "loss": 1.1143, + "step": 1499 + }, + { + "epoch": 0.11, + "learning_rate": 1.2315270935960592e-05, + "loss": 1.1355, + "step": 1500 + }, + { + "epoch": 0.11, + "learning_rate": 1.2323481116584566e-05, + "loss": 1.0719, + "step": 1501 + }, + { + "epoch": 0.11, + "learning_rate": 1.233169129720854e-05, + "loss": 1.0228, + "step": 1502 + }, + { + "epoch": 0.11, + "learning_rate": 1.2339901477832515e-05, + "loss": 1.0586, + "step": 1503 + }, + { + "epoch": 0.11, + "learning_rate": 1.2348111658456487e-05, + "loss": 1.1724, + "step": 1504 + }, + { + "epoch": 0.11, + "learning_rate": 1.2356321839080461e-05, + "loss": 1.1443, + "step": 1505 + }, + { + "epoch": 0.11, + "learning_rate": 1.2364532019704433e-05, + "loss": 1.0345, + "step": 1506 + }, + { + "epoch": 0.11, + "learning_rate": 1.2372742200328409e-05, + "loss": 1.0777, + "step": 1507 + }, + { + "epoch": 0.11, + "learning_rate": 1.2380952380952383e-05, + "loss": 1.147, + "step": 1508 + }, + { + "epoch": 0.11, + "learning_rate": 1.2389162561576355e-05, + "loss": 1.0555, + "step": 1509 + }, + { + "epoch": 0.11, + "learning_rate": 1.2397372742200329e-05, + "loss": 1.0569, + "step": 1510 + }, + { + "epoch": 0.11, + "learning_rate": 1.2405582922824304e-05, + "loss": 1.0677, + "step": 1511 + }, + { + "epoch": 0.11, + "learning_rate": 1.2413793103448277e-05, + "loss": 1.179, + "step": 1512 + }, + { + "epoch": 0.11, + "learning_rate": 1.242200328407225e-05, + "loss": 1.0921, + "step": 1513 + }, + { + "epoch": 0.11, + "learning_rate": 1.2430213464696224e-05, + "loss": 1.1344, + "step": 1514 + }, + { + "epoch": 0.11, + "learning_rate": 1.2438423645320198e-05, + "loss": 1.0547, + "step": 1515 + }, + { + "epoch": 0.11, + "learning_rate": 1.2446633825944172e-05, + "loss": 1.1582, + "step": 1516 + }, + { + "epoch": 0.11, + "learning_rate": 1.2454844006568146e-05, + "loss": 1.1925, + "step": 1517 + }, + { + "epoch": 0.11, + "learning_rate": 1.2463054187192118e-05, + "loss": 1.1906, + "step": 1518 + }, + { + "epoch": 0.11, + "learning_rate": 1.2471264367816094e-05, + "loss": 1.0312, + "step": 1519 + }, + { + "epoch": 0.11, + "learning_rate": 1.2479474548440068e-05, + "loss": 1.1551, + "step": 1520 + }, + { + "epoch": 0.11, + "learning_rate": 1.248768472906404e-05, + "loss": 1.0938, + "step": 1521 + }, + { + "epoch": 0.11, + "learning_rate": 1.2495894909688014e-05, + "loss": 1.0133, + "step": 1522 + }, + { + "epoch": 0.11, + "learning_rate": 1.250410509031199e-05, + "loss": 1.0543, + "step": 1523 + }, + { + "epoch": 0.11, + "learning_rate": 1.2512315270935961e-05, + "loss": 0.9239, + "step": 1524 + }, + { + "epoch": 0.11, + "learning_rate": 1.2520525451559935e-05, + "loss": 1.057, + "step": 1525 + }, + { + "epoch": 0.11, + "learning_rate": 1.2528735632183907e-05, + "loss": 1.1246, + "step": 1526 + }, + { + "epoch": 0.11, + "learning_rate": 1.2536945812807883e-05, + "loss": 1.1767, + "step": 1527 + }, + { + "epoch": 0.11, + "learning_rate": 1.2545155993431857e-05, + "loss": 1.147, + "step": 1528 + }, + { + "epoch": 0.11, + "learning_rate": 1.2553366174055829e-05, + "loss": 1.0331, + "step": 1529 + }, + { + "epoch": 0.11, + "learning_rate": 1.2561576354679803e-05, + "loss": 1.1052, + "step": 1530 + }, + { + "epoch": 0.11, + "learning_rate": 1.2569786535303779e-05, + "loss": 1.186, + "step": 1531 + }, + { + "epoch": 0.11, + "learning_rate": 1.2577996715927752e-05, + "loss": 1.0913, + "step": 1532 + }, + { + "epoch": 0.11, + "learning_rate": 1.2586206896551725e-05, + "loss": 1.0488, + "step": 1533 + }, + { + "epoch": 0.11, + "learning_rate": 1.2594417077175698e-05, + "loss": 1.197, + "step": 1534 + }, + { + "epoch": 0.11, + "learning_rate": 1.2602627257799674e-05, + "loss": 1.0078, + "step": 1535 + }, + { + "epoch": 0.11, + "learning_rate": 1.2610837438423646e-05, + "loss": 1.0695, + "step": 1536 + }, + { + "epoch": 0.11, + "learning_rate": 1.261904761904762e-05, + "loss": 1.1704, + "step": 1537 + }, + { + "epoch": 0.11, + "learning_rate": 1.2627257799671592e-05, + "loss": 1.1034, + "step": 1538 + }, + { + "epoch": 0.11, + "learning_rate": 1.2635467980295568e-05, + "loss": 1.1461, + "step": 1539 + }, + { + "epoch": 0.11, + "learning_rate": 1.2643678160919542e-05, + "loss": 1.1492, + "step": 1540 + }, + { + "epoch": 0.11, + "learning_rate": 1.2651888341543514e-05, + "loss": 1.0318, + "step": 1541 + }, + { + "epoch": 0.11, + "learning_rate": 1.2660098522167488e-05, + "loss": 1.0719, + "step": 1542 + }, + { + "epoch": 0.11, + "learning_rate": 1.2668308702791463e-05, + "loss": 0.988, + "step": 1543 + }, + { + "epoch": 0.11, + "learning_rate": 1.2676518883415435e-05, + "loss": 1.0734, + "step": 1544 + }, + { + "epoch": 0.11, + "learning_rate": 1.268472906403941e-05, + "loss": 1.1264, + "step": 1545 + }, + { + "epoch": 0.11, + "learning_rate": 1.2692939244663383e-05, + "loss": 1.0901, + "step": 1546 + }, + { + "epoch": 0.11, + "learning_rate": 1.2701149425287359e-05, + "loss": 1.0295, + "step": 1547 + }, + { + "epoch": 0.11, + "learning_rate": 1.2709359605911331e-05, + "loss": 1.0362, + "step": 1548 + }, + { + "epoch": 0.11, + "learning_rate": 1.2717569786535305e-05, + "loss": 0.9767, + "step": 1549 + }, + { + "epoch": 0.11, + "learning_rate": 1.2725779967159277e-05, + "loss": 1.0837, + "step": 1550 + }, + { + "epoch": 0.11, + "learning_rate": 1.2733990147783253e-05, + "loss": 1.2072, + "step": 1551 + }, + { + "epoch": 0.11, + "learning_rate": 1.2742200328407226e-05, + "loss": 1.0618, + "step": 1552 + }, + { + "epoch": 0.11, + "learning_rate": 1.2750410509031199e-05, + "loss": 1.1878, + "step": 1553 + }, + { + "epoch": 0.11, + "learning_rate": 1.2758620689655174e-05, + "loss": 1.0202, + "step": 1554 + }, + { + "epoch": 0.11, + "learning_rate": 1.2766830870279148e-05, + "loss": 1.0331, + "step": 1555 + }, + { + "epoch": 0.11, + "learning_rate": 1.277504105090312e-05, + "loss": 1.1604, + "step": 1556 + }, + { + "epoch": 0.12, + "learning_rate": 1.2783251231527094e-05, + "loss": 1.1438, + "step": 1557 + }, + { + "epoch": 0.12, + "learning_rate": 1.279146141215107e-05, + "loss": 1.1822, + "step": 1558 + }, + { + "epoch": 0.12, + "learning_rate": 1.2799671592775042e-05, + "loss": 1.057, + "step": 1559 + }, + { + "epoch": 0.12, + "learning_rate": 1.2807881773399016e-05, + "loss": 1.0697, + "step": 1560 + }, + { + "epoch": 0.12, + "learning_rate": 1.281609195402299e-05, + "loss": 1.0721, + "step": 1561 + }, + { + "epoch": 0.12, + "learning_rate": 1.2824302134646963e-05, + "loss": 1.0813, + "step": 1562 + }, + { + "epoch": 0.12, + "learning_rate": 1.2832512315270937e-05, + "loss": 1.0717, + "step": 1563 + }, + { + "epoch": 0.12, + "learning_rate": 1.2840722495894911e-05, + "loss": 0.9739, + "step": 1564 + }, + { + "epoch": 0.12, + "learning_rate": 1.2848932676518883e-05, + "loss": 1.132, + "step": 1565 + }, + { + "epoch": 0.12, + "learning_rate": 1.2857142857142859e-05, + "loss": 1.081, + "step": 1566 + }, + { + "epoch": 0.12, + "learning_rate": 1.2865353037766833e-05, + "loss": 1.15, + "step": 1567 + }, + { + "epoch": 0.12, + "learning_rate": 1.2873563218390805e-05, + "loss": 1.2082, + "step": 1568 + }, + { + "epoch": 0.12, + "learning_rate": 1.2881773399014779e-05, + "loss": 1.1785, + "step": 1569 + }, + { + "epoch": 0.12, + "learning_rate": 1.2889983579638754e-05, + "loss": 0.9331, + "step": 1570 + }, + { + "epoch": 0.12, + "learning_rate": 1.2898193760262727e-05, + "loss": 1.0076, + "step": 1571 + }, + { + "epoch": 0.12, + "learning_rate": 1.29064039408867e-05, + "loss": 1.199, + "step": 1572 + }, + { + "epoch": 0.12, + "learning_rate": 1.2914614121510674e-05, + "loss": 1.0806, + "step": 1573 + }, + { + "epoch": 0.12, + "learning_rate": 1.2922824302134648e-05, + "loss": 1.0459, + "step": 1574 + }, + { + "epoch": 0.12, + "learning_rate": 1.2931034482758622e-05, + "loss": 1.1023, + "step": 1575 + }, + { + "epoch": 0.12, + "learning_rate": 1.2939244663382596e-05, + "loss": 1.1984, + "step": 1576 + }, + { + "epoch": 0.12, + "learning_rate": 1.2947454844006568e-05, + "loss": 1.0417, + "step": 1577 + }, + { + "epoch": 0.12, + "learning_rate": 1.2955665024630544e-05, + "loss": 1.1754, + "step": 1578 + }, + { + "epoch": 0.12, + "learning_rate": 1.2963875205254518e-05, + "loss": 1.1235, + "step": 1579 + }, + { + "epoch": 0.12, + "learning_rate": 1.297208538587849e-05, + "loss": 1.1448, + "step": 1580 + }, + { + "epoch": 0.12, + "learning_rate": 1.2980295566502464e-05, + "loss": 1.0171, + "step": 1581 + }, + { + "epoch": 0.12, + "learning_rate": 1.298850574712644e-05, + "loss": 1.094, + "step": 1582 + }, + { + "epoch": 0.12, + "learning_rate": 1.2996715927750411e-05, + "loss": 1.0845, + "step": 1583 + }, + { + "epoch": 0.12, + "learning_rate": 1.3004926108374385e-05, + "loss": 1.2025, + "step": 1584 + }, + { + "epoch": 0.12, + "learning_rate": 1.3013136288998357e-05, + "loss": 1.0923, + "step": 1585 + }, + { + "epoch": 0.12, + "learning_rate": 1.3021346469622333e-05, + "loss": 1.0712, + "step": 1586 + }, + { + "epoch": 0.12, + "learning_rate": 1.3029556650246307e-05, + "loss": 1.1124, + "step": 1587 + }, + { + "epoch": 0.12, + "learning_rate": 1.3037766830870279e-05, + "loss": 1.1524, + "step": 1588 + }, + { + "epoch": 0.12, + "learning_rate": 1.3045977011494253e-05, + "loss": 1.0488, + "step": 1589 + }, + { + "epoch": 0.12, + "learning_rate": 1.3054187192118228e-05, + "loss": 1.1511, + "step": 1590 + }, + { + "epoch": 0.12, + "learning_rate": 1.3062397372742202e-05, + "loss": 1.2206, + "step": 1591 + }, + { + "epoch": 0.12, + "learning_rate": 1.3070607553366174e-05, + "loss": 1.1296, + "step": 1592 + }, + { + "epoch": 0.12, + "learning_rate": 1.3078817733990148e-05, + "loss": 1.0747, + "step": 1593 + }, + { + "epoch": 0.12, + "learning_rate": 1.3087027914614124e-05, + "loss": 1.295, + "step": 1594 + }, + { + "epoch": 0.12, + "learning_rate": 1.3095238095238096e-05, + "loss": 1.0444, + "step": 1595 + }, + { + "epoch": 0.12, + "learning_rate": 1.310344827586207e-05, + "loss": 1.0447, + "step": 1596 + }, + { + "epoch": 0.12, + "learning_rate": 1.3111658456486042e-05, + "loss": 1.0508, + "step": 1597 + }, + { + "epoch": 0.12, + "learning_rate": 1.3119868637110018e-05, + "loss": 1.18, + "step": 1598 + }, + { + "epoch": 0.12, + "learning_rate": 1.3128078817733992e-05, + "loss": 1.1575, + "step": 1599 + }, + { + "epoch": 0.12, + "learning_rate": 1.3136288998357964e-05, + "loss": 1.0887, + "step": 1600 + }, + { + "epoch": 0.12, + "learning_rate": 1.3144499178981938e-05, + "loss": 1.0599, + "step": 1601 + }, + { + "epoch": 0.12, + "learning_rate": 1.3152709359605913e-05, + "loss": 1.1333, + "step": 1602 + }, + { + "epoch": 0.12, + "learning_rate": 1.3160919540229885e-05, + "loss": 0.9903, + "step": 1603 + }, + { + "epoch": 0.12, + "learning_rate": 1.316912972085386e-05, + "loss": 1.0183, + "step": 1604 + }, + { + "epoch": 0.12, + "learning_rate": 1.3177339901477833e-05, + "loss": 1.0853, + "step": 1605 + }, + { + "epoch": 0.12, + "learning_rate": 1.3185550082101809e-05, + "loss": 1.1925, + "step": 1606 + }, + { + "epoch": 0.12, + "learning_rate": 1.3193760262725781e-05, + "loss": 1.0781, + "step": 1607 + }, + { + "epoch": 0.12, + "learning_rate": 1.3201970443349755e-05, + "loss": 1.2121, + "step": 1608 + }, + { + "epoch": 0.12, + "learning_rate": 1.321018062397373e-05, + "loss": 1.0636, + "step": 1609 + }, + { + "epoch": 0.12, + "learning_rate": 1.3218390804597702e-05, + "loss": 1.0323, + "step": 1610 + }, + { + "epoch": 0.12, + "learning_rate": 1.3226600985221676e-05, + "loss": 1.1516, + "step": 1611 + }, + { + "epoch": 0.12, + "learning_rate": 1.3234811165845649e-05, + "loss": 1.1118, + "step": 1612 + }, + { + "epoch": 0.12, + "learning_rate": 1.3243021346469624e-05, + "loss": 1.1233, + "step": 1613 + }, + { + "epoch": 0.12, + "learning_rate": 1.3251231527093598e-05, + "loss": 1.115, + "step": 1614 + }, + { + "epoch": 0.12, + "learning_rate": 1.325944170771757e-05, + "loss": 1.1168, + "step": 1615 + }, + { + "epoch": 0.12, + "learning_rate": 1.3267651888341544e-05, + "loss": 1.092, + "step": 1616 + }, + { + "epoch": 0.12, + "learning_rate": 1.327586206896552e-05, + "loss": 0.9674, + "step": 1617 + }, + { + "epoch": 0.12, + "learning_rate": 1.3284072249589492e-05, + "loss": 1.1122, + "step": 1618 + }, + { + "epoch": 0.12, + "learning_rate": 1.3292282430213466e-05, + "loss": 0.9966, + "step": 1619 + }, + { + "epoch": 0.12, + "learning_rate": 1.330049261083744e-05, + "loss": 1.0715, + "step": 1620 + }, + { + "epoch": 0.12, + "learning_rate": 1.3308702791461413e-05, + "loss": 1.1316, + "step": 1621 + }, + { + "epoch": 0.12, + "learning_rate": 1.3316912972085387e-05, + "loss": 1.1502, + "step": 1622 + }, + { + "epoch": 0.12, + "learning_rate": 1.3325123152709361e-05, + "loss": 1.0886, + "step": 1623 + }, + { + "epoch": 0.12, + "learning_rate": 1.3333333333333333e-05, + "loss": 1.1525, + "step": 1624 + }, + { + "epoch": 0.12, + "learning_rate": 1.3341543513957309e-05, + "loss": 1.1332, + "step": 1625 + }, + { + "epoch": 0.12, + "learning_rate": 1.3349753694581283e-05, + "loss": 1.0394, + "step": 1626 + }, + { + "epoch": 0.12, + "learning_rate": 1.3357963875205255e-05, + "loss": 1.1389, + "step": 1627 + }, + { + "epoch": 0.12, + "learning_rate": 1.3366174055829229e-05, + "loss": 1.0831, + "step": 1628 + }, + { + "epoch": 0.12, + "learning_rate": 1.3374384236453204e-05, + "loss": 1.1158, + "step": 1629 + }, + { + "epoch": 0.12, + "learning_rate": 1.3382594417077177e-05, + "loss": 1.1191, + "step": 1630 + }, + { + "epoch": 0.12, + "learning_rate": 1.339080459770115e-05, + "loss": 1.219, + "step": 1631 + }, + { + "epoch": 0.12, + "learning_rate": 1.3399014778325124e-05, + "loss": 1.125, + "step": 1632 + }, + { + "epoch": 0.12, + "learning_rate": 1.3407224958949098e-05, + "loss": 1.161, + "step": 1633 + }, + { + "epoch": 0.12, + "learning_rate": 1.3415435139573072e-05, + "loss": 1.1084, + "step": 1634 + }, + { + "epoch": 0.12, + "learning_rate": 1.3423645320197046e-05, + "loss": 1.1262, + "step": 1635 + }, + { + "epoch": 0.12, + "learning_rate": 1.3431855500821018e-05, + "loss": 1.1685, + "step": 1636 + }, + { + "epoch": 0.12, + "learning_rate": 1.3440065681444994e-05, + "loss": 0.9933, + "step": 1637 + }, + { + "epoch": 0.12, + "learning_rate": 1.3448275862068967e-05, + "loss": 0.9815, + "step": 1638 + }, + { + "epoch": 0.12, + "learning_rate": 1.345648604269294e-05, + "loss": 1.1143, + "step": 1639 + }, + { + "epoch": 0.12, + "learning_rate": 1.3464696223316914e-05, + "loss": 1.16, + "step": 1640 + }, + { + "epoch": 0.12, + "learning_rate": 1.3472906403940889e-05, + "loss": 1.1018, + "step": 1641 + }, + { + "epoch": 0.12, + "learning_rate": 1.3481116584564861e-05, + "loss": 1.1379, + "step": 1642 + }, + { + "epoch": 0.12, + "learning_rate": 1.3489326765188835e-05, + "loss": 1.1456, + "step": 1643 + }, + { + "epoch": 0.12, + "learning_rate": 1.3497536945812807e-05, + "loss": 1.0978, + "step": 1644 + }, + { + "epoch": 0.12, + "learning_rate": 1.3505747126436783e-05, + "loss": 1.0507, + "step": 1645 + }, + { + "epoch": 0.12, + "learning_rate": 1.3513957307060757e-05, + "loss": 1.043, + "step": 1646 + }, + { + "epoch": 0.12, + "learning_rate": 1.3522167487684729e-05, + "loss": 1.1173, + "step": 1647 + }, + { + "epoch": 0.12, + "learning_rate": 1.3530377668308703e-05, + "loss": 1.1578, + "step": 1648 + }, + { + "epoch": 0.12, + "learning_rate": 1.3538587848932678e-05, + "loss": 1.0558, + "step": 1649 + }, + { + "epoch": 0.12, + "learning_rate": 1.3546798029556652e-05, + "loss": 1.1995, + "step": 1650 + }, + { + "epoch": 0.12, + "learning_rate": 1.3555008210180624e-05, + "loss": 1.1216, + "step": 1651 + }, + { + "epoch": 0.12, + "learning_rate": 1.3563218390804598e-05, + "loss": 1.1409, + "step": 1652 + }, + { + "epoch": 0.12, + "learning_rate": 1.3571428571428574e-05, + "loss": 1.033, + "step": 1653 + }, + { + "epoch": 0.12, + "learning_rate": 1.3579638752052546e-05, + "loss": 1.1733, + "step": 1654 + }, + { + "epoch": 0.12, + "learning_rate": 1.358784893267652e-05, + "loss": 1.094, + "step": 1655 + }, + { + "epoch": 0.12, + "learning_rate": 1.3596059113300492e-05, + "loss": 1.1187, + "step": 1656 + }, + { + "epoch": 0.12, + "learning_rate": 1.3604269293924468e-05, + "loss": 1.1556, + "step": 1657 + }, + { + "epoch": 0.12, + "learning_rate": 1.3612479474548442e-05, + "loss": 1.1029, + "step": 1658 + }, + { + "epoch": 0.12, + "learning_rate": 1.3620689655172414e-05, + "loss": 1.1416, + "step": 1659 + }, + { + "epoch": 0.12, + "learning_rate": 1.3628899835796388e-05, + "loss": 1.0578, + "step": 1660 + }, + { + "epoch": 0.12, + "learning_rate": 1.3637110016420363e-05, + "loss": 1.1265, + "step": 1661 + }, + { + "epoch": 0.12, + "learning_rate": 1.3645320197044335e-05, + "loss": 1.1358, + "step": 1662 + }, + { + "epoch": 0.12, + "learning_rate": 1.365353037766831e-05, + "loss": 1.0248, + "step": 1663 + }, + { + "epoch": 0.12, + "learning_rate": 1.3661740558292285e-05, + "loss": 1.1808, + "step": 1664 + }, + { + "epoch": 0.12, + "learning_rate": 1.3669950738916259e-05, + "loss": 1.1251, + "step": 1665 + }, + { + "epoch": 0.12, + "learning_rate": 1.367816091954023e-05, + "loss": 1.0064, + "step": 1666 + }, + { + "epoch": 0.12, + "learning_rate": 1.3686371100164205e-05, + "loss": 1.0325, + "step": 1667 + }, + { + "epoch": 0.12, + "learning_rate": 1.369458128078818e-05, + "loss": 1.1337, + "step": 1668 + }, + { + "epoch": 0.12, + "learning_rate": 1.3702791461412152e-05, + "loss": 1.1716, + "step": 1669 + }, + { + "epoch": 0.12, + "learning_rate": 1.3711001642036126e-05, + "loss": 1.0402, + "step": 1670 + }, + { + "epoch": 0.12, + "learning_rate": 1.3719211822660098e-05, + "loss": 1.2322, + "step": 1671 + }, + { + "epoch": 0.12, + "learning_rate": 1.3727422003284074e-05, + "loss": 1.1289, + "step": 1672 + }, + { + "epoch": 0.12, + "learning_rate": 1.3735632183908048e-05, + "loss": 1.1044, + "step": 1673 + }, + { + "epoch": 0.12, + "learning_rate": 1.374384236453202e-05, + "loss": 1.0846, + "step": 1674 + }, + { + "epoch": 0.12, + "learning_rate": 1.3752052545155994e-05, + "loss": 1.1408, + "step": 1675 + }, + { + "epoch": 0.12, + "learning_rate": 1.376026272577997e-05, + "loss": 1.1297, + "step": 1676 + }, + { + "epoch": 0.12, + "learning_rate": 1.3768472906403942e-05, + "loss": 1.1675, + "step": 1677 + }, + { + "epoch": 0.12, + "learning_rate": 1.3776683087027916e-05, + "loss": 1.114, + "step": 1678 + }, + { + "epoch": 0.12, + "learning_rate": 1.378489326765189e-05, + "loss": 1.1161, + "step": 1679 + }, + { + "epoch": 0.12, + "learning_rate": 1.3793103448275863e-05, + "loss": 1.1487, + "step": 1680 + }, + { + "epoch": 0.12, + "learning_rate": 1.3801313628899837e-05, + "loss": 1.1301, + "step": 1681 + }, + { + "epoch": 0.12, + "learning_rate": 1.3809523809523811e-05, + "loss": 1.0787, + "step": 1682 + }, + { + "epoch": 0.12, + "learning_rate": 1.3817733990147783e-05, + "loss": 1.1312, + "step": 1683 + }, + { + "epoch": 0.12, + "learning_rate": 1.3825944170771759e-05, + "loss": 0.9852, + "step": 1684 + }, + { + "epoch": 0.12, + "learning_rate": 1.3834154351395733e-05, + "loss": 1.1682, + "step": 1685 + }, + { + "epoch": 0.12, + "learning_rate": 1.3842364532019705e-05, + "loss": 1.1378, + "step": 1686 + }, + { + "epoch": 0.12, + "learning_rate": 1.3850574712643679e-05, + "loss": 1.236, + "step": 1687 + }, + { + "epoch": 0.12, + "learning_rate": 1.3858784893267654e-05, + "loss": 1.0371, + "step": 1688 + }, + { + "epoch": 0.12, + "learning_rate": 1.3866995073891626e-05, + "loss": 1.0246, + "step": 1689 + }, + { + "epoch": 0.12, + "learning_rate": 1.38752052545156e-05, + "loss": 1.1395, + "step": 1690 + }, + { + "epoch": 0.12, + "learning_rate": 1.3883415435139573e-05, + "loss": 1.0549, + "step": 1691 + }, + { + "epoch": 0.13, + "learning_rate": 1.3891625615763548e-05, + "loss": 1.0669, + "step": 1692 + }, + { + "epoch": 0.13, + "learning_rate": 1.3899835796387522e-05, + "loss": 1.0093, + "step": 1693 + }, + { + "epoch": 0.13, + "learning_rate": 1.3908045977011496e-05, + "loss": 1.0259, + "step": 1694 + }, + { + "epoch": 0.13, + "learning_rate": 1.3916256157635468e-05, + "loss": 1.1754, + "step": 1695 + }, + { + "epoch": 0.13, + "learning_rate": 1.3924466338259444e-05, + "loss": 1.0685, + "step": 1696 + }, + { + "epoch": 0.13, + "learning_rate": 1.3932676518883417e-05, + "loss": 0.977, + "step": 1697 + }, + { + "epoch": 0.13, + "learning_rate": 1.394088669950739e-05, + "loss": 0.9682, + "step": 1698 + }, + { + "epoch": 0.13, + "learning_rate": 1.3949096880131363e-05, + "loss": 1.1479, + "step": 1699 + }, + { + "epoch": 0.13, + "learning_rate": 1.3957307060755339e-05, + "loss": 1.1477, + "step": 1700 + }, + { + "epoch": 0.13, + "learning_rate": 1.3965517241379311e-05, + "loss": 1.065, + "step": 1701 + }, + { + "epoch": 0.13, + "learning_rate": 1.3973727422003285e-05, + "loss": 1.1884, + "step": 1702 + }, + { + "epoch": 0.13, + "learning_rate": 1.3981937602627257e-05, + "loss": 1.1042, + "step": 1703 + }, + { + "epoch": 0.13, + "learning_rate": 1.3990147783251233e-05, + "loss": 1.0715, + "step": 1704 + }, + { + "epoch": 0.13, + "learning_rate": 1.3998357963875207e-05, + "loss": 0.9563, + "step": 1705 + }, + { + "epoch": 0.13, + "learning_rate": 1.4006568144499179e-05, + "loss": 0.9814, + "step": 1706 + }, + { + "epoch": 0.13, + "learning_rate": 1.4014778325123153e-05, + "loss": 1.1, + "step": 1707 + }, + { + "epoch": 0.13, + "learning_rate": 1.4022988505747128e-05, + "loss": 1.1501, + "step": 1708 + }, + { + "epoch": 0.13, + "learning_rate": 1.4031198686371102e-05, + "loss": 1.1412, + "step": 1709 + }, + { + "epoch": 0.13, + "learning_rate": 1.4039408866995074e-05, + "loss": 1.042, + "step": 1710 + }, + { + "epoch": 0.13, + "learning_rate": 1.4047619047619048e-05, + "loss": 1.0704, + "step": 1711 + }, + { + "epoch": 0.13, + "learning_rate": 1.4055829228243024e-05, + "loss": 1.0383, + "step": 1712 + }, + { + "epoch": 0.13, + "learning_rate": 1.4064039408866996e-05, + "loss": 1.1377, + "step": 1713 + }, + { + "epoch": 0.13, + "learning_rate": 1.407224958949097e-05, + "loss": 0.9968, + "step": 1714 + }, + { + "epoch": 0.13, + "learning_rate": 1.4080459770114942e-05, + "loss": 1.0109, + "step": 1715 + }, + { + "epoch": 0.13, + "learning_rate": 1.4088669950738918e-05, + "loss": 1.1428, + "step": 1716 + }, + { + "epoch": 0.13, + "learning_rate": 1.4096880131362891e-05, + "loss": 1.1, + "step": 1717 + }, + { + "epoch": 0.13, + "learning_rate": 1.4105090311986864e-05, + "loss": 0.9821, + "step": 1718 + }, + { + "epoch": 0.13, + "learning_rate": 1.411330049261084e-05, + "loss": 1.0179, + "step": 1719 + }, + { + "epoch": 0.13, + "learning_rate": 1.4121510673234813e-05, + "loss": 1.1352, + "step": 1720 + }, + { + "epoch": 0.13, + "learning_rate": 1.4129720853858785e-05, + "loss": 1.029, + "step": 1721 + }, + { + "epoch": 0.13, + "learning_rate": 1.4137931034482759e-05, + "loss": 1.0311, + "step": 1722 + }, + { + "epoch": 0.13, + "learning_rate": 1.4146141215106735e-05, + "loss": 0.9772, + "step": 1723 + }, + { + "epoch": 0.13, + "learning_rate": 1.4154351395730707e-05, + "loss": 1.1384, + "step": 1724 + }, + { + "epoch": 0.13, + "learning_rate": 1.416256157635468e-05, + "loss": 1.0947, + "step": 1725 + }, + { + "epoch": 0.13, + "learning_rate": 1.4170771756978655e-05, + "loss": 1.076, + "step": 1726 + }, + { + "epoch": 0.13, + "learning_rate": 1.417898193760263e-05, + "loss": 1.1225, + "step": 1727 + }, + { + "epoch": 0.13, + "learning_rate": 1.4187192118226602e-05, + "loss": 1.1123, + "step": 1728 + }, + { + "epoch": 0.13, + "learning_rate": 1.4195402298850576e-05, + "loss": 1.1208, + "step": 1729 + }, + { + "epoch": 0.13, + "learning_rate": 1.4203612479474548e-05, + "loss": 1.1165, + "step": 1730 + }, + { + "epoch": 0.13, + "learning_rate": 1.4211822660098524e-05, + "loss": 1.1232, + "step": 1731 + }, + { + "epoch": 0.13, + "learning_rate": 1.4220032840722498e-05, + "loss": 1.095, + "step": 1732 + }, + { + "epoch": 0.13, + "learning_rate": 1.422824302134647e-05, + "loss": 1.0878, + "step": 1733 + }, + { + "epoch": 0.13, + "learning_rate": 1.4236453201970444e-05, + "loss": 1.0744, + "step": 1734 + }, + { + "epoch": 0.13, + "learning_rate": 1.424466338259442e-05, + "loss": 1.1986, + "step": 1735 + }, + { + "epoch": 0.13, + "learning_rate": 1.4252873563218392e-05, + "loss": 1.01, + "step": 1736 + }, + { + "epoch": 0.13, + "learning_rate": 1.4261083743842366e-05, + "loss": 1.1369, + "step": 1737 + }, + { + "epoch": 0.13, + "learning_rate": 1.426929392446634e-05, + "loss": 1.1369, + "step": 1738 + }, + { + "epoch": 0.13, + "learning_rate": 1.4277504105090313e-05, + "loss": 1.0653, + "step": 1739 + }, + { + "epoch": 0.13, + "learning_rate": 1.4285714285714287e-05, + "loss": 1.1111, + "step": 1740 + }, + { + "epoch": 0.13, + "learning_rate": 1.4293924466338261e-05, + "loss": 1.1063, + "step": 1741 + }, + { + "epoch": 0.13, + "learning_rate": 1.4302134646962233e-05, + "loss": 1.015, + "step": 1742 + }, + { + "epoch": 0.13, + "learning_rate": 1.4310344827586209e-05, + "loss": 1.1691, + "step": 1743 + }, + { + "epoch": 0.13, + "learning_rate": 1.4318555008210183e-05, + "loss": 0.98, + "step": 1744 + }, + { + "epoch": 0.13, + "learning_rate": 1.4326765188834155e-05, + "loss": 1.1304, + "step": 1745 + }, + { + "epoch": 0.13, + "learning_rate": 1.4334975369458129e-05, + "loss": 1.0677, + "step": 1746 + }, + { + "epoch": 0.13, + "learning_rate": 1.4343185550082104e-05, + "loss": 1.1698, + "step": 1747 + }, + { + "epoch": 0.13, + "learning_rate": 1.4351395730706076e-05, + "loss": 1.1481, + "step": 1748 + }, + { + "epoch": 0.13, + "learning_rate": 1.435960591133005e-05, + "loss": 1.048, + "step": 1749 + }, + { + "epoch": 0.13, + "learning_rate": 1.4367816091954022e-05, + "loss": 1.1031, + "step": 1750 + }, + { + "epoch": 0.13, + "learning_rate": 1.4376026272577998e-05, + "loss": 1.0858, + "step": 1751 + }, + { + "epoch": 0.13, + "learning_rate": 1.4384236453201972e-05, + "loss": 1.1736, + "step": 1752 + }, + { + "epoch": 0.13, + "learning_rate": 1.4392446633825946e-05, + "loss": 1.1342, + "step": 1753 + }, + { + "epoch": 0.13, + "learning_rate": 1.4400656814449918e-05, + "loss": 1.0139, + "step": 1754 + }, + { + "epoch": 0.13, + "learning_rate": 1.4408866995073894e-05, + "loss": 0.9892, + "step": 1755 + }, + { + "epoch": 0.13, + "learning_rate": 1.4417077175697867e-05, + "loss": 1.2192, + "step": 1756 + }, + { + "epoch": 0.13, + "learning_rate": 1.442528735632184e-05, + "loss": 1.1144, + "step": 1757 + }, + { + "epoch": 0.13, + "learning_rate": 1.4433497536945813e-05, + "loss": 1.1334, + "step": 1758 + }, + { + "epoch": 0.13, + "learning_rate": 1.4441707717569789e-05, + "loss": 1.0616, + "step": 1759 + }, + { + "epoch": 0.13, + "learning_rate": 1.4449917898193761e-05, + "loss": 1.1119, + "step": 1760 + }, + { + "epoch": 0.13, + "learning_rate": 1.4458128078817735e-05, + "loss": 1.1102, + "step": 1761 + }, + { + "epoch": 0.13, + "learning_rate": 1.4466338259441707e-05, + "loss": 1.078, + "step": 1762 + }, + { + "epoch": 0.13, + "learning_rate": 1.4474548440065683e-05, + "loss": 1.0494, + "step": 1763 + }, + { + "epoch": 0.13, + "learning_rate": 1.4482758620689657e-05, + "loss": 1.1112, + "step": 1764 + }, + { + "epoch": 0.13, + "learning_rate": 1.4490968801313629e-05, + "loss": 0.9979, + "step": 1765 + }, + { + "epoch": 0.13, + "learning_rate": 1.4499178981937603e-05, + "loss": 1.1067, + "step": 1766 + }, + { + "epoch": 0.13, + "learning_rate": 1.4507389162561578e-05, + "loss": 1.0054, + "step": 1767 + }, + { + "epoch": 0.13, + "learning_rate": 1.4515599343185552e-05, + "loss": 1.1458, + "step": 1768 + }, + { + "epoch": 0.13, + "learning_rate": 1.4523809523809524e-05, + "loss": 1.1744, + "step": 1769 + }, + { + "epoch": 0.13, + "learning_rate": 1.4532019704433498e-05, + "loss": 1.0184, + "step": 1770 + }, + { + "epoch": 0.13, + "learning_rate": 1.4540229885057474e-05, + "loss": 1.0513, + "step": 1771 + }, + { + "epoch": 0.13, + "learning_rate": 1.4548440065681446e-05, + "loss": 1.0187, + "step": 1772 + }, + { + "epoch": 0.13, + "learning_rate": 1.455665024630542e-05, + "loss": 1.1512, + "step": 1773 + }, + { + "epoch": 0.13, + "learning_rate": 1.4564860426929395e-05, + "loss": 1.1238, + "step": 1774 + }, + { + "epoch": 0.13, + "learning_rate": 1.4573070607553368e-05, + "loss": 0.9919, + "step": 1775 + }, + { + "epoch": 0.13, + "learning_rate": 1.4581280788177341e-05, + "loss": 1.0161, + "step": 1776 + }, + { + "epoch": 0.13, + "learning_rate": 1.4589490968801314e-05, + "loss": 1.0474, + "step": 1777 + }, + { + "epoch": 0.13, + "learning_rate": 1.459770114942529e-05, + "loss": 1.1005, + "step": 1778 + }, + { + "epoch": 0.13, + "learning_rate": 1.4605911330049263e-05, + "loss": 1.0855, + "step": 1779 + }, + { + "epoch": 0.13, + "learning_rate": 1.4614121510673235e-05, + "loss": 1.1725, + "step": 1780 + }, + { + "epoch": 0.13, + "learning_rate": 1.4622331691297209e-05, + "loss": 1.111, + "step": 1781 + }, + { + "epoch": 0.13, + "learning_rate": 1.4630541871921185e-05, + "loss": 1.1262, + "step": 1782 + }, + { + "epoch": 0.13, + "learning_rate": 1.4638752052545157e-05, + "loss": 1.0814, + "step": 1783 + }, + { + "epoch": 0.13, + "learning_rate": 1.464696223316913e-05, + "loss": 1.0409, + "step": 1784 + }, + { + "epoch": 0.13, + "learning_rate": 1.4655172413793105e-05, + "loss": 1.1672, + "step": 1785 + }, + { + "epoch": 0.13, + "learning_rate": 1.466338259441708e-05, + "loss": 1.0529, + "step": 1786 + }, + { + "epoch": 0.13, + "learning_rate": 1.4671592775041052e-05, + "loss": 1.1125, + "step": 1787 + }, + { + "epoch": 0.13, + "learning_rate": 1.4679802955665026e-05, + "loss": 1.0978, + "step": 1788 + }, + { + "epoch": 0.13, + "learning_rate": 1.4688013136288998e-05, + "loss": 1.0449, + "step": 1789 + }, + { + "epoch": 0.13, + "learning_rate": 1.4696223316912974e-05, + "loss": 0.9422, + "step": 1790 + }, + { + "epoch": 0.13, + "learning_rate": 1.4704433497536948e-05, + "loss": 1.043, + "step": 1791 + }, + { + "epoch": 0.13, + "learning_rate": 1.471264367816092e-05, + "loss": 1.073, + "step": 1792 + }, + { + "epoch": 0.13, + "learning_rate": 1.4720853858784894e-05, + "loss": 1.0772, + "step": 1793 + }, + { + "epoch": 0.13, + "learning_rate": 1.472906403940887e-05, + "loss": 0.9955, + "step": 1794 + }, + { + "epoch": 0.13, + "learning_rate": 1.4737274220032842e-05, + "loss": 1.0249, + "step": 1795 + }, + { + "epoch": 0.13, + "learning_rate": 1.4745484400656815e-05, + "loss": 1.0783, + "step": 1796 + }, + { + "epoch": 0.13, + "learning_rate": 1.475369458128079e-05, + "loss": 1.044, + "step": 1797 + }, + { + "epoch": 0.13, + "learning_rate": 1.4761904761904763e-05, + "loss": 1.0682, + "step": 1798 + }, + { + "epoch": 0.13, + "learning_rate": 1.4770114942528737e-05, + "loss": 1.067, + "step": 1799 + }, + { + "epoch": 0.13, + "learning_rate": 1.4778325123152711e-05, + "loss": 1.0569, + "step": 1800 + }, + { + "epoch": 0.13, + "learning_rate": 1.4786535303776683e-05, + "loss": 0.9819, + "step": 1801 + }, + { + "epoch": 0.13, + "learning_rate": 1.4794745484400659e-05, + "loss": 1.0502, + "step": 1802 + }, + { + "epoch": 0.13, + "learning_rate": 1.4802955665024633e-05, + "loss": 1.0739, + "step": 1803 + }, + { + "epoch": 0.13, + "learning_rate": 1.4811165845648605e-05, + "loss": 1.103, + "step": 1804 + }, + { + "epoch": 0.13, + "learning_rate": 1.4819376026272579e-05, + "loss": 1.0647, + "step": 1805 + }, + { + "epoch": 0.13, + "learning_rate": 1.4827586206896554e-05, + "loss": 1.2282, + "step": 1806 + }, + { + "epoch": 0.13, + "learning_rate": 1.4835796387520526e-05, + "loss": 1.1708, + "step": 1807 + }, + { + "epoch": 0.13, + "learning_rate": 1.48440065681445e-05, + "loss": 1.1398, + "step": 1808 + }, + { + "epoch": 0.13, + "learning_rate": 1.4852216748768472e-05, + "loss": 1.1461, + "step": 1809 + }, + { + "epoch": 0.13, + "learning_rate": 1.4860426929392448e-05, + "loss": 1.0795, + "step": 1810 + }, + { + "epoch": 0.13, + "learning_rate": 1.4868637110016422e-05, + "loss": 1.0146, + "step": 1811 + }, + { + "epoch": 0.13, + "learning_rate": 1.4876847290640396e-05, + "loss": 1.0036, + "step": 1812 + }, + { + "epoch": 0.13, + "learning_rate": 1.4885057471264368e-05, + "loss": 1.0535, + "step": 1813 + }, + { + "epoch": 0.13, + "learning_rate": 1.4893267651888343e-05, + "loss": 1.1098, + "step": 1814 + }, + { + "epoch": 0.13, + "learning_rate": 1.4901477832512317e-05, + "loss": 1.0387, + "step": 1815 + }, + { + "epoch": 0.13, + "learning_rate": 1.490968801313629e-05, + "loss": 1.0586, + "step": 1816 + }, + { + "epoch": 0.13, + "learning_rate": 1.4917898193760263e-05, + "loss": 0.9943, + "step": 1817 + }, + { + "epoch": 0.13, + "learning_rate": 1.4926108374384239e-05, + "loss": 1.1403, + "step": 1818 + }, + { + "epoch": 0.13, + "learning_rate": 1.4934318555008211e-05, + "loss": 1.1014, + "step": 1819 + }, + { + "epoch": 0.13, + "learning_rate": 1.4942528735632185e-05, + "loss": 1.1415, + "step": 1820 + }, + { + "epoch": 0.13, + "learning_rate": 1.4950738916256157e-05, + "loss": 1.029, + "step": 1821 + }, + { + "epoch": 0.13, + "learning_rate": 1.4958949096880133e-05, + "loss": 1.0579, + "step": 1822 + }, + { + "epoch": 0.13, + "learning_rate": 1.4967159277504107e-05, + "loss": 1.0608, + "step": 1823 + }, + { + "epoch": 0.13, + "learning_rate": 1.4975369458128079e-05, + "loss": 1.1331, + "step": 1824 + }, + { + "epoch": 0.13, + "learning_rate": 1.4983579638752053e-05, + "loss": 1.0916, + "step": 1825 + }, + { + "epoch": 0.13, + "learning_rate": 1.4991789819376028e-05, + "loss": 1.0663, + "step": 1826 + }, + { + "epoch": 0.13, + "learning_rate": 1.5000000000000002e-05, + "loss": 1.0869, + "step": 1827 + }, + { + "epoch": 0.14, + "learning_rate": 1.5008210180623974e-05, + "loss": 1.0808, + "step": 1828 + }, + { + "epoch": 0.14, + "learning_rate": 1.501642036124795e-05, + "loss": 1.1595, + "step": 1829 + }, + { + "epoch": 0.14, + "learning_rate": 1.5024630541871924e-05, + "loss": 0.9836, + "step": 1830 + }, + { + "epoch": 0.14, + "learning_rate": 1.5032840722495896e-05, + "loss": 1.0387, + "step": 1831 + }, + { + "epoch": 0.14, + "learning_rate": 1.504105090311987e-05, + "loss": 1.1119, + "step": 1832 + }, + { + "epoch": 0.14, + "learning_rate": 1.5049261083743845e-05, + "loss": 1.1657, + "step": 1833 + }, + { + "epoch": 0.14, + "learning_rate": 1.5057471264367817e-05, + "loss": 0.9578, + "step": 1834 + }, + { + "epoch": 0.14, + "learning_rate": 1.5065681444991791e-05, + "loss": 1.1145, + "step": 1835 + }, + { + "epoch": 0.14, + "learning_rate": 1.5073891625615764e-05, + "loss": 1.0757, + "step": 1836 + }, + { + "epoch": 0.14, + "learning_rate": 1.5082101806239739e-05, + "loss": 1.0977, + "step": 1837 + }, + { + "epoch": 0.14, + "learning_rate": 1.5090311986863713e-05, + "loss": 1.02, + "step": 1838 + }, + { + "epoch": 0.14, + "learning_rate": 1.5098522167487685e-05, + "loss": 1.1789, + "step": 1839 + }, + { + "epoch": 0.14, + "learning_rate": 1.5106732348111659e-05, + "loss": 1.1652, + "step": 1840 + }, + { + "epoch": 0.14, + "learning_rate": 1.5114942528735635e-05, + "loss": 1.0875, + "step": 1841 + }, + { + "epoch": 0.14, + "learning_rate": 1.5123152709359607e-05, + "loss": 1.09, + "step": 1842 + }, + { + "epoch": 0.14, + "learning_rate": 1.513136288998358e-05, + "loss": 1.137, + "step": 1843 + }, + { + "epoch": 0.14, + "learning_rate": 1.5139573070607554e-05, + "loss": 1.0192, + "step": 1844 + }, + { + "epoch": 0.14, + "learning_rate": 1.514778325123153e-05, + "loss": 1.1664, + "step": 1845 + }, + { + "epoch": 0.14, + "learning_rate": 1.5155993431855502e-05, + "loss": 1.0414, + "step": 1846 + }, + { + "epoch": 0.14, + "learning_rate": 1.5164203612479476e-05, + "loss": 1.0361, + "step": 1847 + }, + { + "epoch": 0.14, + "learning_rate": 1.5172413793103448e-05, + "loss": 1.0117, + "step": 1848 + }, + { + "epoch": 0.14, + "learning_rate": 1.5180623973727424e-05, + "loss": 1.0691, + "step": 1849 + }, + { + "epoch": 0.14, + "learning_rate": 1.5188834154351398e-05, + "loss": 1.0496, + "step": 1850 + }, + { + "epoch": 0.14, + "learning_rate": 1.519704433497537e-05, + "loss": 1.1147, + "step": 1851 + }, + { + "epoch": 0.14, + "learning_rate": 1.5205254515599344e-05, + "loss": 1.0238, + "step": 1852 + }, + { + "epoch": 0.14, + "learning_rate": 1.521346469622332e-05, + "loss": 1.1858, + "step": 1853 + }, + { + "epoch": 0.14, + "learning_rate": 1.5221674876847292e-05, + "loss": 1.0376, + "step": 1854 + }, + { + "epoch": 0.14, + "learning_rate": 1.5229885057471265e-05, + "loss": 1.1599, + "step": 1855 + }, + { + "epoch": 0.14, + "learning_rate": 1.523809523809524e-05, + "loss": 1.114, + "step": 1856 + }, + { + "epoch": 0.14, + "learning_rate": 1.5246305418719213e-05, + "loss": 1.1893, + "step": 1857 + }, + { + "epoch": 0.14, + "learning_rate": 1.5254515599343187e-05, + "loss": 1.1601, + "step": 1858 + }, + { + "epoch": 0.14, + "learning_rate": 1.526272577996716e-05, + "loss": 1.1135, + "step": 1859 + }, + { + "epoch": 0.14, + "learning_rate": 1.5270935960591133e-05, + "loss": 0.9857, + "step": 1860 + }, + { + "epoch": 0.14, + "learning_rate": 1.527914614121511e-05, + "loss": 1.0344, + "step": 1861 + }, + { + "epoch": 0.14, + "learning_rate": 1.528735632183908e-05, + "loss": 1.0545, + "step": 1862 + }, + { + "epoch": 0.14, + "learning_rate": 1.5295566502463055e-05, + "loss": 1.0079, + "step": 1863 + }, + { + "epoch": 0.14, + "learning_rate": 1.530377668308703e-05, + "loss": 1.1695, + "step": 1864 + }, + { + "epoch": 0.14, + "learning_rate": 1.5311986863711002e-05, + "loss": 1.1582, + "step": 1865 + }, + { + "epoch": 0.14, + "learning_rate": 1.5320197044334976e-05, + "loss": 0.9872, + "step": 1866 + }, + { + "epoch": 0.14, + "learning_rate": 1.532840722495895e-05, + "loss": 0.9737, + "step": 1867 + }, + { + "epoch": 0.14, + "learning_rate": 1.5336617405582924e-05, + "loss": 1.118, + "step": 1868 + }, + { + "epoch": 0.14, + "learning_rate": 1.5344827586206898e-05, + "loss": 1.057, + "step": 1869 + }, + { + "epoch": 0.14, + "learning_rate": 1.5353037766830872e-05, + "loss": 1.1304, + "step": 1870 + }, + { + "epoch": 0.14, + "learning_rate": 1.5361247947454846e-05, + "loss": 1.1324, + "step": 1871 + }, + { + "epoch": 0.14, + "learning_rate": 1.536945812807882e-05, + "loss": 1.0211, + "step": 1872 + }, + { + "epoch": 0.14, + "learning_rate": 1.5377668308702793e-05, + "loss": 1.0306, + "step": 1873 + }, + { + "epoch": 0.14, + "learning_rate": 1.5385878489326767e-05, + "loss": 1.0336, + "step": 1874 + }, + { + "epoch": 0.14, + "learning_rate": 1.539408866995074e-05, + "loss": 1.1887, + "step": 1875 + }, + { + "epoch": 0.14, + "learning_rate": 1.540229885057471e-05, + "loss": 1.0818, + "step": 1876 + }, + { + "epoch": 0.14, + "learning_rate": 1.541050903119869e-05, + "loss": 1.1349, + "step": 1877 + }, + { + "epoch": 0.14, + "learning_rate": 1.5418719211822663e-05, + "loss": 1.053, + "step": 1878 + }, + { + "epoch": 0.14, + "learning_rate": 1.5426929392446633e-05, + "loss": 1.0375, + "step": 1879 + }, + { + "epoch": 0.14, + "learning_rate": 1.5435139573070607e-05, + "loss": 1.0928, + "step": 1880 + }, + { + "epoch": 0.14, + "learning_rate": 1.5443349753694584e-05, + "loss": 1.0885, + "step": 1881 + }, + { + "epoch": 0.14, + "learning_rate": 1.5451559934318555e-05, + "loss": 1.0055, + "step": 1882 + }, + { + "epoch": 0.14, + "learning_rate": 1.545977011494253e-05, + "loss": 1.0474, + "step": 1883 + }, + { + "epoch": 0.14, + "learning_rate": 1.5467980295566506e-05, + "loss": 1.1637, + "step": 1884 + }, + { + "epoch": 0.14, + "learning_rate": 1.5476190476190476e-05, + "loss": 1.1186, + "step": 1885 + }, + { + "epoch": 0.14, + "learning_rate": 1.548440065681445e-05, + "loss": 1.1808, + "step": 1886 + }, + { + "epoch": 0.14, + "learning_rate": 1.5492610837438424e-05, + "loss": 1.0556, + "step": 1887 + }, + { + "epoch": 0.14, + "learning_rate": 1.5500821018062398e-05, + "loss": 1.0415, + "step": 1888 + }, + { + "epoch": 0.14, + "learning_rate": 1.5509031198686372e-05, + "loss": 1.0719, + "step": 1889 + }, + { + "epoch": 0.14, + "learning_rate": 1.5517241379310346e-05, + "loss": 1.039, + "step": 1890 + }, + { + "epoch": 0.14, + "learning_rate": 1.552545155993432e-05, + "loss": 0.9883, + "step": 1891 + }, + { + "epoch": 0.14, + "learning_rate": 1.5533661740558294e-05, + "loss": 1.1852, + "step": 1892 + }, + { + "epoch": 0.14, + "learning_rate": 1.5541871921182267e-05, + "loss": 1.0747, + "step": 1893 + }, + { + "epoch": 0.14, + "learning_rate": 1.555008210180624e-05, + "loss": 1.0653, + "step": 1894 + }, + { + "epoch": 0.14, + "learning_rate": 1.5558292282430215e-05, + "loss": 1.0828, + "step": 1895 + }, + { + "epoch": 0.14, + "learning_rate": 1.556650246305419e-05, + "loss": 1.0994, + "step": 1896 + }, + { + "epoch": 0.14, + "learning_rate": 1.5574712643678163e-05, + "loss": 1.161, + "step": 1897 + }, + { + "epoch": 0.14, + "learning_rate": 1.5582922824302137e-05, + "loss": 1.0717, + "step": 1898 + }, + { + "epoch": 0.14, + "learning_rate": 1.5591133004926107e-05, + "loss": 0.9513, + "step": 1899 + }, + { + "epoch": 0.14, + "learning_rate": 1.5599343185550085e-05, + "loss": 1.1173, + "step": 1900 + }, + { + "epoch": 0.14, + "learning_rate": 1.560755336617406e-05, + "loss": 1.0663, + "step": 1901 + }, + { + "epoch": 0.14, + "learning_rate": 1.5615763546798032e-05, + "loss": 1.0321, + "step": 1902 + }, + { + "epoch": 0.14, + "learning_rate": 1.5623973727422003e-05, + "loss": 1.0225, + "step": 1903 + }, + { + "epoch": 0.14, + "learning_rate": 1.563218390804598e-05, + "loss": 1.0924, + "step": 1904 + }, + { + "epoch": 0.14, + "learning_rate": 1.5640394088669954e-05, + "loss": 1.0292, + "step": 1905 + }, + { + "epoch": 0.14, + "learning_rate": 1.5648604269293924e-05, + "loss": 1.104, + "step": 1906 + }, + { + "epoch": 0.14, + "learning_rate": 1.5656814449917898e-05, + "loss": 1.2246, + "step": 1907 + }, + { + "epoch": 0.14, + "learning_rate": 1.5665024630541875e-05, + "loss": 0.9208, + "step": 1908 + }, + { + "epoch": 0.14, + "learning_rate": 1.5673234811165846e-05, + "loss": 1.1146, + "step": 1909 + }, + { + "epoch": 0.14, + "learning_rate": 1.568144499178982e-05, + "loss": 0.9894, + "step": 1910 + }, + { + "epoch": 0.14, + "learning_rate": 1.5689655172413794e-05, + "loss": 0.9782, + "step": 1911 + }, + { + "epoch": 0.14, + "learning_rate": 1.5697865353037768e-05, + "loss": 0.9423, + "step": 1912 + }, + { + "epoch": 0.14, + "learning_rate": 1.570607553366174e-05, + "loss": 1.0873, + "step": 1913 + }, + { + "epoch": 0.14, + "learning_rate": 1.5714285714285715e-05, + "loss": 1.1398, + "step": 1914 + }, + { + "epoch": 0.14, + "learning_rate": 1.572249589490969e-05, + "loss": 1.2034, + "step": 1915 + }, + { + "epoch": 0.14, + "learning_rate": 1.5730706075533663e-05, + "loss": 1.0922, + "step": 1916 + }, + { + "epoch": 0.14, + "learning_rate": 1.5738916256157637e-05, + "loss": 1.2049, + "step": 1917 + }, + { + "epoch": 0.14, + "learning_rate": 1.574712643678161e-05, + "loss": 0.9275, + "step": 1918 + }, + { + "epoch": 0.14, + "learning_rate": 1.5755336617405585e-05, + "loss": 1.1583, + "step": 1919 + }, + { + "epoch": 0.14, + "learning_rate": 1.576354679802956e-05, + "loss": 1.1696, + "step": 1920 + }, + { + "epoch": 0.14, + "learning_rate": 1.5771756978653532e-05, + "loss": 1.004, + "step": 1921 + }, + { + "epoch": 0.14, + "learning_rate": 1.5779967159277506e-05, + "loss": 1.0729, + "step": 1922 + }, + { + "epoch": 0.14, + "learning_rate": 1.5788177339901477e-05, + "loss": 1.19, + "step": 1923 + }, + { + "epoch": 0.14, + "learning_rate": 1.5796387520525454e-05, + "loss": 1.0616, + "step": 1924 + }, + { + "epoch": 0.14, + "learning_rate": 1.5804597701149428e-05, + "loss": 1.0681, + "step": 1925 + }, + { + "epoch": 0.14, + "learning_rate": 1.58128078817734e-05, + "loss": 1.0641, + "step": 1926 + }, + { + "epoch": 0.14, + "learning_rate": 1.5821018062397372e-05, + "loss": 1.1639, + "step": 1927 + }, + { + "epoch": 0.14, + "learning_rate": 1.582922824302135e-05, + "loss": 1.1162, + "step": 1928 + }, + { + "epoch": 0.14, + "learning_rate": 1.583743842364532e-05, + "loss": 1.0037, + "step": 1929 + }, + { + "epoch": 0.14, + "learning_rate": 1.5845648604269294e-05, + "loss": 1.1177, + "step": 1930 + }, + { + "epoch": 0.14, + "learning_rate": 1.5853858784893268e-05, + "loss": 1.1143, + "step": 1931 + }, + { + "epoch": 0.14, + "learning_rate": 1.586206896551724e-05, + "loss": 1.0578, + "step": 1932 + }, + { + "epoch": 0.14, + "learning_rate": 1.5870279146141215e-05, + "loss": 1.1414, + "step": 1933 + }, + { + "epoch": 0.14, + "learning_rate": 1.587848932676519e-05, + "loss": 1.0903, + "step": 1934 + }, + { + "epoch": 0.14, + "learning_rate": 1.5886699507389163e-05, + "loss": 0.9767, + "step": 1935 + }, + { + "epoch": 0.14, + "learning_rate": 1.5894909688013137e-05, + "loss": 1.0533, + "step": 1936 + }, + { + "epoch": 0.14, + "learning_rate": 1.590311986863711e-05, + "loss": 1.0307, + "step": 1937 + }, + { + "epoch": 0.14, + "learning_rate": 1.5911330049261085e-05, + "loss": 1.0669, + "step": 1938 + }, + { + "epoch": 0.14, + "learning_rate": 1.591954022988506e-05, + "loss": 1.09, + "step": 1939 + }, + { + "epoch": 0.14, + "learning_rate": 1.5927750410509033e-05, + "loss": 1.1382, + "step": 1940 + }, + { + "epoch": 0.14, + "learning_rate": 1.5935960591133006e-05, + "loss": 1.0804, + "step": 1941 + }, + { + "epoch": 0.14, + "learning_rate": 1.594417077175698e-05, + "loss": 1.021, + "step": 1942 + }, + { + "epoch": 0.14, + "learning_rate": 1.5952380952380954e-05, + "loss": 1.0464, + "step": 1943 + }, + { + "epoch": 0.14, + "learning_rate": 1.5960591133004928e-05, + "loss": 1.1132, + "step": 1944 + }, + { + "epoch": 0.14, + "learning_rate": 1.5968801313628902e-05, + "loss": 1.008, + "step": 1945 + }, + { + "epoch": 0.14, + "learning_rate": 1.5977011494252876e-05, + "loss": 1.084, + "step": 1946 + }, + { + "epoch": 0.14, + "learning_rate": 1.598522167487685e-05, + "loss": 1.2417, + "step": 1947 + }, + { + "epoch": 0.14, + "learning_rate": 1.5993431855500824e-05, + "loss": 1.0145, + "step": 1948 + }, + { + "epoch": 0.14, + "learning_rate": 1.6001642036124797e-05, + "loss": 1.0922, + "step": 1949 + }, + { + "epoch": 0.14, + "learning_rate": 1.6009852216748768e-05, + "loss": 1.0024, + "step": 1950 + }, + { + "epoch": 0.14, + "learning_rate": 1.6018062397372745e-05, + "loss": 1.1313, + "step": 1951 + }, + { + "epoch": 0.14, + "learning_rate": 1.602627257799672e-05, + "loss": 1.0764, + "step": 1952 + }, + { + "epoch": 0.14, + "learning_rate": 1.603448275862069e-05, + "loss": 1.0761, + "step": 1953 + }, + { + "epoch": 0.14, + "learning_rate": 1.6042692939244663e-05, + "loss": 1.1088, + "step": 1954 + }, + { + "epoch": 0.14, + "learning_rate": 1.605090311986864e-05, + "loss": 1.0611, + "step": 1955 + }, + { + "epoch": 0.14, + "learning_rate": 1.605911330049261e-05, + "loss": 1.0136, + "step": 1956 + }, + { + "epoch": 0.14, + "learning_rate": 1.6067323481116585e-05, + "loss": 1.0223, + "step": 1957 + }, + { + "epoch": 0.14, + "learning_rate": 1.607553366174056e-05, + "loss": 1.1704, + "step": 1958 + }, + { + "epoch": 0.14, + "learning_rate": 1.6083743842364533e-05, + "loss": 1.1644, + "step": 1959 + }, + { + "epoch": 0.14, + "learning_rate": 1.6091954022988507e-05, + "loss": 1.079, + "step": 1960 + }, + { + "epoch": 0.14, + "learning_rate": 1.610016420361248e-05, + "loss": 1.1052, + "step": 1961 + }, + { + "epoch": 0.14, + "learning_rate": 1.6108374384236454e-05, + "loss": 1.0875, + "step": 1962 + }, + { + "epoch": 0.15, + "learning_rate": 1.6116584564860428e-05, + "loss": 1.0065, + "step": 1963 + }, + { + "epoch": 0.15, + "learning_rate": 1.6124794745484402e-05, + "loss": 1.1197, + "step": 1964 + }, + { + "epoch": 0.15, + "learning_rate": 1.6133004926108376e-05, + "loss": 1.0148, + "step": 1965 + }, + { + "epoch": 0.15, + "learning_rate": 1.614121510673235e-05, + "loss": 1.0562, + "step": 1966 + }, + { + "epoch": 0.15, + "learning_rate": 1.6149425287356324e-05, + "loss": 0.9798, + "step": 1967 + }, + { + "epoch": 0.15, + "learning_rate": 1.6157635467980298e-05, + "loss": 1.1738, + "step": 1968 + }, + { + "epoch": 0.15, + "learning_rate": 1.616584564860427e-05, + "loss": 1.0411, + "step": 1969 + }, + { + "epoch": 0.15, + "learning_rate": 1.6174055829228242e-05, + "loss": 1.0874, + "step": 1970 + }, + { + "epoch": 0.15, + "learning_rate": 1.618226600985222e-05, + "loss": 1.1769, + "step": 1971 + }, + { + "epoch": 0.15, + "learning_rate": 1.6190476190476193e-05, + "loss": 1.0973, + "step": 1972 + }, + { + "epoch": 0.15, + "learning_rate": 1.6198686371100164e-05, + "loss": 1.1135, + "step": 1973 + }, + { + "epoch": 0.15, + "learning_rate": 1.6206896551724137e-05, + "loss": 1.1161, + "step": 1974 + }, + { + "epoch": 0.15, + "learning_rate": 1.6215106732348115e-05, + "loss": 1.1312, + "step": 1975 + }, + { + "epoch": 0.15, + "learning_rate": 1.622331691297209e-05, + "loss": 0.9905, + "step": 1976 + }, + { + "epoch": 0.15, + "learning_rate": 1.623152709359606e-05, + "loss": 1.0856, + "step": 1977 + }, + { + "epoch": 0.15, + "learning_rate": 1.6239737274220033e-05, + "loss": 1.0598, + "step": 1978 + }, + { + "epoch": 0.15, + "learning_rate": 1.624794745484401e-05, + "loss": 1.0976, + "step": 1979 + }, + { + "epoch": 0.15, + "learning_rate": 1.625615763546798e-05, + "loss": 1.0834, + "step": 1980 + }, + { + "epoch": 0.15, + "learning_rate": 1.6264367816091955e-05, + "loss": 0.9643, + "step": 1981 + }, + { + "epoch": 0.15, + "learning_rate": 1.627257799671593e-05, + "loss": 1.1095, + "step": 1982 + }, + { + "epoch": 0.15, + "learning_rate": 1.6280788177339902e-05, + "loss": 0.9695, + "step": 1983 + }, + { + "epoch": 0.15, + "learning_rate": 1.6288998357963876e-05, + "loss": 1.0426, + "step": 1984 + }, + { + "epoch": 0.15, + "learning_rate": 1.629720853858785e-05, + "loss": 1.032, + "step": 1985 + }, + { + "epoch": 0.15, + "learning_rate": 1.6305418719211824e-05, + "loss": 1.0379, + "step": 1986 + }, + { + "epoch": 0.15, + "learning_rate": 1.6313628899835798e-05, + "loss": 1.2381, + "step": 1987 + }, + { + "epoch": 0.15, + "learning_rate": 1.632183908045977e-05, + "loss": 1.0791, + "step": 1988 + }, + { + "epoch": 0.15, + "learning_rate": 1.6330049261083746e-05, + "loss": 1.0873, + "step": 1989 + }, + { + "epoch": 0.15, + "learning_rate": 1.633825944170772e-05, + "loss": 1.0717, + "step": 1990 + }, + { + "epoch": 0.15, + "learning_rate": 1.6346469622331693e-05, + "loss": 1.1298, + "step": 1991 + }, + { + "epoch": 0.15, + "learning_rate": 1.6354679802955667e-05, + "loss": 1.1435, + "step": 1992 + }, + { + "epoch": 0.15, + "learning_rate": 1.636288998357964e-05, + "loss": 1.1029, + "step": 1993 + }, + { + "epoch": 0.15, + "learning_rate": 1.6371100164203615e-05, + "loss": 1.124, + "step": 1994 + }, + { + "epoch": 0.15, + "learning_rate": 1.637931034482759e-05, + "loss": 1.014, + "step": 1995 + }, + { + "epoch": 0.15, + "learning_rate": 1.6387520525451563e-05, + "loss": 1.2234, + "step": 1996 + }, + { + "epoch": 0.15, + "learning_rate": 1.6395730706075533e-05, + "loss": 1.059, + "step": 1997 + }, + { + "epoch": 0.15, + "learning_rate": 1.640394088669951e-05, + "loss": 1.0756, + "step": 1998 + }, + { + "epoch": 0.15, + "learning_rate": 1.6412151067323484e-05, + "loss": 1.0318, + "step": 1999 + }, + { + "epoch": 0.15, + "learning_rate": 1.6420361247947455e-05, + "loss": 1.111, + "step": 2000 + }, + { + "epoch": 0.15, + "learning_rate": 1.642857142857143e-05, + "loss": 0.92, + "step": 2001 + }, + { + "epoch": 0.15, + "learning_rate": 1.6436781609195406e-05, + "loss": 1.0673, + "step": 2002 + }, + { + "epoch": 0.15, + "learning_rate": 1.6444991789819376e-05, + "loss": 1.0214, + "step": 2003 + }, + { + "epoch": 0.15, + "learning_rate": 1.645320197044335e-05, + "loss": 1.1546, + "step": 2004 + }, + { + "epoch": 0.15, + "learning_rate": 1.6461412151067324e-05, + "loss": 1.0615, + "step": 2005 + }, + { + "epoch": 0.15, + "learning_rate": 1.6469622331691298e-05, + "loss": 1.1439, + "step": 2006 + }, + { + "epoch": 0.15, + "learning_rate": 1.6477832512315272e-05, + "loss": 1.056, + "step": 2007 + }, + { + "epoch": 0.15, + "learning_rate": 1.6486042692939246e-05, + "loss": 1.2432, + "step": 2008 + }, + { + "epoch": 0.15, + "learning_rate": 1.649425287356322e-05, + "loss": 1.0686, + "step": 2009 + }, + { + "epoch": 0.15, + "learning_rate": 1.6502463054187193e-05, + "loss": 1.0633, + "step": 2010 + }, + { + "epoch": 0.15, + "learning_rate": 1.6510673234811167e-05, + "loss": 1.1928, + "step": 2011 + }, + { + "epoch": 0.15, + "learning_rate": 1.651888341543514e-05, + "loss": 1.0458, + "step": 2012 + }, + { + "epoch": 0.15, + "learning_rate": 1.6527093596059115e-05, + "loss": 1.1261, + "step": 2013 + }, + { + "epoch": 0.15, + "learning_rate": 1.653530377668309e-05, + "loss": 1.0034, + "step": 2014 + }, + { + "epoch": 0.15, + "learning_rate": 1.6543513957307063e-05, + "loss": 1.1074, + "step": 2015 + }, + { + "epoch": 0.15, + "learning_rate": 1.6551724137931037e-05, + "loss": 1.1103, + "step": 2016 + }, + { + "epoch": 0.15, + "learning_rate": 1.6559934318555007e-05, + "loss": 1.0082, + "step": 2017 + }, + { + "epoch": 0.15, + "learning_rate": 1.6568144499178984e-05, + "loss": 1.1037, + "step": 2018 + }, + { + "epoch": 0.15, + "learning_rate": 1.6576354679802958e-05, + "loss": 1.1602, + "step": 2019 + }, + { + "epoch": 0.15, + "learning_rate": 1.6584564860426932e-05, + "loss": 1.0353, + "step": 2020 + }, + { + "epoch": 0.15, + "learning_rate": 1.6592775041050903e-05, + "loss": 1.1117, + "step": 2021 + }, + { + "epoch": 0.15, + "learning_rate": 1.660098522167488e-05, + "loss": 1.0714, + "step": 2022 + }, + { + "epoch": 0.15, + "learning_rate": 1.6609195402298854e-05, + "loss": 1.0057, + "step": 2023 + }, + { + "epoch": 0.15, + "learning_rate": 1.6617405582922824e-05, + "loss": 1.0372, + "step": 2024 + }, + { + "epoch": 0.15, + "learning_rate": 1.6625615763546798e-05, + "loss": 1.0347, + "step": 2025 + }, + { + "epoch": 0.15, + "learning_rate": 1.6633825944170775e-05, + "loss": 1.0699, + "step": 2026 + }, + { + "epoch": 0.15, + "learning_rate": 1.6642036124794746e-05, + "loss": 1.023, + "step": 2027 + }, + { + "epoch": 0.15, + "learning_rate": 1.665024630541872e-05, + "loss": 1.0494, + "step": 2028 + }, + { + "epoch": 0.15, + "learning_rate": 1.6658456486042694e-05, + "loss": 1.1878, + "step": 2029 + }, + { + "epoch": 0.15, + "learning_rate": 1.6666666666666667e-05, + "loss": 1.1209, + "step": 2030 + }, + { + "epoch": 0.15, + "learning_rate": 1.667487684729064e-05, + "loss": 1.0935, + "step": 2031 + }, + { + "epoch": 0.15, + "learning_rate": 1.6683087027914615e-05, + "loss": 1.0146, + "step": 2032 + }, + { + "epoch": 0.15, + "learning_rate": 1.669129720853859e-05, + "loss": 1.0764, + "step": 2033 + }, + { + "epoch": 0.15, + "learning_rate": 1.6699507389162563e-05, + "loss": 1.1452, + "step": 2034 + }, + { + "epoch": 0.15, + "learning_rate": 1.6707717569786537e-05, + "loss": 1.086, + "step": 2035 + }, + { + "epoch": 0.15, + "learning_rate": 1.671592775041051e-05, + "loss": 1.0912, + "step": 2036 + }, + { + "epoch": 0.15, + "learning_rate": 1.6724137931034485e-05, + "loss": 0.9474, + "step": 2037 + }, + { + "epoch": 0.15, + "learning_rate": 1.673234811165846e-05, + "loss": 1.0364, + "step": 2038 + }, + { + "epoch": 0.15, + "learning_rate": 1.6740558292282432e-05, + "loss": 1.1013, + "step": 2039 + }, + { + "epoch": 0.15, + "learning_rate": 1.6748768472906406e-05, + "loss": 1.1128, + "step": 2040 + }, + { + "epoch": 0.15, + "learning_rate": 1.6756978653530377e-05, + "loss": 0.9616, + "step": 2041 + }, + { + "epoch": 0.15, + "learning_rate": 1.6765188834154354e-05, + "loss": 1.0009, + "step": 2042 + }, + { + "epoch": 0.15, + "learning_rate": 1.6773399014778328e-05, + "loss": 1.0302, + "step": 2043 + }, + { + "epoch": 0.15, + "learning_rate": 1.6781609195402298e-05, + "loss": 1.0781, + "step": 2044 + }, + { + "epoch": 0.15, + "learning_rate": 1.6789819376026272e-05, + "loss": 1.0917, + "step": 2045 + }, + { + "epoch": 0.15, + "learning_rate": 1.679802955665025e-05, + "loss": 1.1512, + "step": 2046 + }, + { + "epoch": 0.15, + "learning_rate": 1.680623973727422e-05, + "loss": 1.0321, + "step": 2047 + }, + { + "epoch": 0.15, + "learning_rate": 1.6814449917898194e-05, + "loss": 1.1821, + "step": 2048 + }, + { + "epoch": 0.15, + "learning_rate": 1.682266009852217e-05, + "loss": 1.1683, + "step": 2049 + }, + { + "epoch": 0.15, + "learning_rate": 1.683087027914614e-05, + "loss": 1.0991, + "step": 2050 + }, + { + "epoch": 0.15, + "learning_rate": 1.6839080459770115e-05, + "loss": 1.1845, + "step": 2051 + }, + { + "epoch": 0.15, + "learning_rate": 1.684729064039409e-05, + "loss": 1.0316, + "step": 2052 + }, + { + "epoch": 0.15, + "learning_rate": 1.6855500821018067e-05, + "loss": 1.0882, + "step": 2053 + }, + { + "epoch": 0.15, + "learning_rate": 1.6863711001642037e-05, + "loss": 1.051, + "step": 2054 + }, + { + "epoch": 0.15, + "learning_rate": 1.687192118226601e-05, + "loss": 1.0857, + "step": 2055 + }, + { + "epoch": 0.15, + "learning_rate": 1.6880131362889985e-05, + "loss": 1.0176, + "step": 2056 + }, + { + "epoch": 0.15, + "learning_rate": 1.688834154351396e-05, + "loss": 1.2063, + "step": 2057 + }, + { + "epoch": 0.15, + "learning_rate": 1.6896551724137932e-05, + "loss": 1.0761, + "step": 2058 + }, + { + "epoch": 0.15, + "learning_rate": 1.6904761904761906e-05, + "loss": 1.0384, + "step": 2059 + }, + { + "epoch": 0.15, + "learning_rate": 1.691297208538588e-05, + "loss": 1.0661, + "step": 2060 + }, + { + "epoch": 0.15, + "learning_rate": 1.6921182266009854e-05, + "loss": 1.1412, + "step": 2061 + }, + { + "epoch": 0.15, + "learning_rate": 1.6929392446633828e-05, + "loss": 1.1628, + "step": 2062 + }, + { + "epoch": 0.15, + "learning_rate": 1.6937602627257802e-05, + "loss": 1.1398, + "step": 2063 + }, + { + "epoch": 0.15, + "learning_rate": 1.6945812807881776e-05, + "loss": 1.0421, + "step": 2064 + }, + { + "epoch": 0.15, + "learning_rate": 1.695402298850575e-05, + "loss": 1.0991, + "step": 2065 + }, + { + "epoch": 0.15, + "learning_rate": 1.6962233169129723e-05, + "loss": 1.1978, + "step": 2066 + }, + { + "epoch": 0.15, + "learning_rate": 1.6970443349753697e-05, + "loss": 1.1979, + "step": 2067 + }, + { + "epoch": 0.15, + "learning_rate": 1.6978653530377668e-05, + "loss": 1.1139, + "step": 2068 + }, + { + "epoch": 0.15, + "learning_rate": 1.6986863711001645e-05, + "loss": 1.1282, + "step": 2069 + }, + { + "epoch": 0.15, + "learning_rate": 1.699507389162562e-05, + "loss": 1.1075, + "step": 2070 + }, + { + "epoch": 0.15, + "learning_rate": 1.700328407224959e-05, + "loss": 1.0728, + "step": 2071 + }, + { + "epoch": 0.15, + "learning_rate": 1.7011494252873563e-05, + "loss": 1.0928, + "step": 2072 + }, + { + "epoch": 0.15, + "learning_rate": 1.701970443349754e-05, + "loss": 1.2167, + "step": 2073 + }, + { + "epoch": 0.15, + "learning_rate": 1.702791461412151e-05, + "loss": 1.1803, + "step": 2074 + }, + { + "epoch": 0.15, + "learning_rate": 1.7036124794745485e-05, + "loss": 1.1357, + "step": 2075 + }, + { + "epoch": 0.15, + "learning_rate": 1.704433497536946e-05, + "loss": 1.0284, + "step": 2076 + }, + { + "epoch": 0.15, + "learning_rate": 1.7052545155993433e-05, + "loss": 1.054, + "step": 2077 + }, + { + "epoch": 0.15, + "learning_rate": 1.7060755336617407e-05, + "loss": 1.1129, + "step": 2078 + }, + { + "epoch": 0.15, + "learning_rate": 1.706896551724138e-05, + "loss": 1.038, + "step": 2079 + }, + { + "epoch": 0.15, + "learning_rate": 1.7077175697865354e-05, + "loss": 1.0956, + "step": 2080 + }, + { + "epoch": 0.15, + "learning_rate": 1.7085385878489328e-05, + "loss": 1.0656, + "step": 2081 + }, + { + "epoch": 0.15, + "learning_rate": 1.7093596059113302e-05, + "loss": 1.0393, + "step": 2082 + }, + { + "epoch": 0.15, + "learning_rate": 1.7101806239737276e-05, + "loss": 1.0993, + "step": 2083 + }, + { + "epoch": 0.15, + "learning_rate": 1.711001642036125e-05, + "loss": 0.9985, + "step": 2084 + }, + { + "epoch": 0.15, + "learning_rate": 1.7118226600985224e-05, + "loss": 1.1086, + "step": 2085 + }, + { + "epoch": 0.15, + "learning_rate": 1.7126436781609197e-05, + "loss": 1.0574, + "step": 2086 + }, + { + "epoch": 0.15, + "learning_rate": 1.713464696223317e-05, + "loss": 1.1644, + "step": 2087 + }, + { + "epoch": 0.15, + "learning_rate": 1.7142857142857142e-05, + "loss": 1.1001, + "step": 2088 + }, + { + "epoch": 0.15, + "learning_rate": 1.715106732348112e-05, + "loss": 0.9994, + "step": 2089 + }, + { + "epoch": 0.15, + "learning_rate": 1.7159277504105093e-05, + "loss": 1.0825, + "step": 2090 + }, + { + "epoch": 0.15, + "learning_rate": 1.7167487684729063e-05, + "loss": 1.1856, + "step": 2091 + }, + { + "epoch": 0.15, + "learning_rate": 1.7175697865353037e-05, + "loss": 1.151, + "step": 2092 + }, + { + "epoch": 0.15, + "learning_rate": 1.7183908045977015e-05, + "loss": 1.1593, + "step": 2093 + }, + { + "epoch": 0.15, + "learning_rate": 1.7192118226600985e-05, + "loss": 1.0577, + "step": 2094 + }, + { + "epoch": 0.15, + "learning_rate": 1.720032840722496e-05, + "loss": 1.1372, + "step": 2095 + }, + { + "epoch": 0.15, + "learning_rate": 1.7208538587848933e-05, + "loss": 1.0488, + "step": 2096 + }, + { + "epoch": 0.15, + "learning_rate": 1.721674876847291e-05, + "loss": 0.9921, + "step": 2097 + }, + { + "epoch": 0.16, + "learning_rate": 1.722495894909688e-05, + "loss": 1.2176, + "step": 2098 + }, + { + "epoch": 0.16, + "learning_rate": 1.7233169129720854e-05, + "loss": 1.1425, + "step": 2099 + }, + { + "epoch": 0.16, + "learning_rate": 1.7241379310344828e-05, + "loss": 1.1292, + "step": 2100 + }, + { + "epoch": 0.16, + "learning_rate": 1.7249589490968802e-05, + "loss": 1.125, + "step": 2101 + }, + { + "epoch": 0.16, + "learning_rate": 1.7257799671592776e-05, + "loss": 1.2174, + "step": 2102 + }, + { + "epoch": 0.16, + "learning_rate": 1.726600985221675e-05, + "loss": 1.068, + "step": 2103 + }, + { + "epoch": 0.16, + "learning_rate": 1.7274220032840724e-05, + "loss": 0.9103, + "step": 2104 + }, + { + "epoch": 0.16, + "learning_rate": 1.7282430213464698e-05, + "loss": 1.1584, + "step": 2105 + }, + { + "epoch": 0.16, + "learning_rate": 1.729064039408867e-05, + "loss": 1.066, + "step": 2106 + }, + { + "epoch": 0.16, + "learning_rate": 1.7298850574712645e-05, + "loss": 1.1552, + "step": 2107 + }, + { + "epoch": 0.16, + "learning_rate": 1.730706075533662e-05, + "loss": 1.1322, + "step": 2108 + }, + { + "epoch": 0.16, + "learning_rate": 1.7315270935960593e-05, + "loss": 1.1206, + "step": 2109 + }, + { + "epoch": 0.16, + "learning_rate": 1.7323481116584567e-05, + "loss": 1.1498, + "step": 2110 + }, + { + "epoch": 0.16, + "learning_rate": 1.733169129720854e-05, + "loss": 1.0236, + "step": 2111 + }, + { + "epoch": 0.16, + "learning_rate": 1.7339901477832515e-05, + "loss": 1.0857, + "step": 2112 + }, + { + "epoch": 0.16, + "learning_rate": 1.734811165845649e-05, + "loss": 1.0223, + "step": 2113 + }, + { + "epoch": 0.16, + "learning_rate": 1.7356321839080462e-05, + "loss": 0.9986, + "step": 2114 + }, + { + "epoch": 0.16, + "learning_rate": 1.7364532019704433e-05, + "loss": 1.0651, + "step": 2115 + }, + { + "epoch": 0.16, + "learning_rate": 1.737274220032841e-05, + "loss": 1.1177, + "step": 2116 + }, + { + "epoch": 0.16, + "learning_rate": 1.7380952380952384e-05, + "loss": 1.0597, + "step": 2117 + }, + { + "epoch": 0.16, + "learning_rate": 1.7389162561576355e-05, + "loss": 1.2164, + "step": 2118 + }, + { + "epoch": 0.16, + "learning_rate": 1.739737274220033e-05, + "loss": 1.0649, + "step": 2119 + }, + { + "epoch": 0.16, + "learning_rate": 1.7405582922824306e-05, + "loss": 0.972, + "step": 2120 + }, + { + "epoch": 0.16, + "learning_rate": 1.7413793103448276e-05, + "loss": 1.0541, + "step": 2121 + }, + { + "epoch": 0.16, + "learning_rate": 1.742200328407225e-05, + "loss": 1.0702, + "step": 2122 + }, + { + "epoch": 0.16, + "learning_rate": 1.7430213464696224e-05, + "loss": 0.9647, + "step": 2123 + }, + { + "epoch": 0.16, + "learning_rate": 1.7438423645320198e-05, + "loss": 1.0189, + "step": 2124 + }, + { + "epoch": 0.16, + "learning_rate": 1.744663382594417e-05, + "loss": 1.1115, + "step": 2125 + }, + { + "epoch": 0.16, + "learning_rate": 1.7454844006568146e-05, + "loss": 1.0931, + "step": 2126 + }, + { + "epoch": 0.16, + "learning_rate": 1.746305418719212e-05, + "loss": 1.0518, + "step": 2127 + }, + { + "epoch": 0.16, + "learning_rate": 1.7471264367816093e-05, + "loss": 1.1728, + "step": 2128 + }, + { + "epoch": 0.16, + "learning_rate": 1.7479474548440067e-05, + "loss": 1.1035, + "step": 2129 + }, + { + "epoch": 0.16, + "learning_rate": 1.748768472906404e-05, + "loss": 1.2509, + "step": 2130 + }, + { + "epoch": 0.16, + "learning_rate": 1.7495894909688015e-05, + "loss": 1.0682, + "step": 2131 + }, + { + "epoch": 0.16, + "learning_rate": 1.750410509031199e-05, + "loss": 1.0267, + "step": 2132 + }, + { + "epoch": 0.16, + "learning_rate": 1.7512315270935963e-05, + "loss": 1.0621, + "step": 2133 + }, + { + "epoch": 0.16, + "learning_rate": 1.7520525451559937e-05, + "loss": 1.1408, + "step": 2134 + }, + { + "epoch": 0.16, + "learning_rate": 1.7528735632183907e-05, + "loss": 1.127, + "step": 2135 + }, + { + "epoch": 0.16, + "learning_rate": 1.7536945812807884e-05, + "loss": 1.2045, + "step": 2136 + }, + { + "epoch": 0.16, + "learning_rate": 1.7545155993431858e-05, + "loss": 1.0195, + "step": 2137 + }, + { + "epoch": 0.16, + "learning_rate": 1.755336617405583e-05, + "loss": 1.0628, + "step": 2138 + }, + { + "epoch": 0.16, + "learning_rate": 1.7561576354679803e-05, + "loss": 1.1634, + "step": 2139 + }, + { + "epoch": 0.16, + "learning_rate": 1.756978653530378e-05, + "loss": 1.051, + "step": 2140 + }, + { + "epoch": 0.16, + "learning_rate": 1.7577996715927754e-05, + "loss": 1.2, + "step": 2141 + }, + { + "epoch": 0.16, + "learning_rate": 1.7586206896551724e-05, + "loss": 0.9747, + "step": 2142 + }, + { + "epoch": 0.16, + "learning_rate": 1.7594417077175698e-05, + "loss": 1.1762, + "step": 2143 + }, + { + "epoch": 0.16, + "learning_rate": 1.7602627257799675e-05, + "loss": 1.1334, + "step": 2144 + }, + { + "epoch": 0.16, + "learning_rate": 1.7610837438423646e-05, + "loss": 1.0235, + "step": 2145 + }, + { + "epoch": 0.16, + "learning_rate": 1.761904761904762e-05, + "loss": 0.9473, + "step": 2146 + }, + { + "epoch": 0.16, + "learning_rate": 1.7627257799671593e-05, + "loss": 1.1287, + "step": 2147 + }, + { + "epoch": 0.16, + "learning_rate": 1.7635467980295567e-05, + "loss": 1.1286, + "step": 2148 + }, + { + "epoch": 0.16, + "learning_rate": 1.764367816091954e-05, + "loss": 1.1118, + "step": 2149 + }, + { + "epoch": 0.16, + "learning_rate": 1.7651888341543515e-05, + "loss": 1.1693, + "step": 2150 + }, + { + "epoch": 0.16, + "learning_rate": 1.766009852216749e-05, + "loss": 1.136, + "step": 2151 + }, + { + "epoch": 0.16, + "learning_rate": 1.7668308702791463e-05, + "loss": 1.0096, + "step": 2152 + }, + { + "epoch": 0.16, + "learning_rate": 1.7676518883415437e-05, + "loss": 1.0619, + "step": 2153 + }, + { + "epoch": 0.16, + "learning_rate": 1.768472906403941e-05, + "loss": 0.9982, + "step": 2154 + }, + { + "epoch": 0.16, + "learning_rate": 1.7692939244663384e-05, + "loss": 0.9598, + "step": 2155 + }, + { + "epoch": 0.16, + "learning_rate": 1.770114942528736e-05, + "loss": 1.0911, + "step": 2156 + }, + { + "epoch": 0.16, + "learning_rate": 1.7709359605911332e-05, + "loss": 1.057, + "step": 2157 + }, + { + "epoch": 0.16, + "learning_rate": 1.7717569786535306e-05, + "loss": 1.0383, + "step": 2158 + }, + { + "epoch": 0.16, + "learning_rate": 1.7725779967159277e-05, + "loss": 1.053, + "step": 2159 + }, + { + "epoch": 0.16, + "learning_rate": 1.7733990147783254e-05, + "loss": 1.1008, + "step": 2160 + }, + { + "epoch": 0.16, + "learning_rate": 1.7742200328407228e-05, + "loss": 1.0706, + "step": 2161 + }, + { + "epoch": 0.16, + "learning_rate": 1.7750410509031198e-05, + "loss": 1.1481, + "step": 2162 + }, + { + "epoch": 0.16, + "learning_rate": 1.7758620689655175e-05, + "loss": 1.063, + "step": 2163 + }, + { + "epoch": 0.16, + "learning_rate": 1.776683087027915e-05, + "loss": 1.1386, + "step": 2164 + }, + { + "epoch": 0.16, + "learning_rate": 1.777504105090312e-05, + "loss": 1.009, + "step": 2165 + }, + { + "epoch": 0.16, + "learning_rate": 1.7783251231527094e-05, + "loss": 1.0706, + "step": 2166 + }, + { + "epoch": 0.16, + "learning_rate": 1.779146141215107e-05, + "loss": 1.0932, + "step": 2167 + }, + { + "epoch": 0.16, + "learning_rate": 1.779967159277504e-05, + "loss": 1.0947, + "step": 2168 + }, + { + "epoch": 0.16, + "learning_rate": 1.7807881773399015e-05, + "loss": 1.0613, + "step": 2169 + }, + { + "epoch": 0.16, + "learning_rate": 1.781609195402299e-05, + "loss": 0.9789, + "step": 2170 + }, + { + "epoch": 0.16, + "learning_rate": 1.7824302134646963e-05, + "loss": 1.1742, + "step": 2171 + }, + { + "epoch": 0.16, + "learning_rate": 1.7832512315270937e-05, + "loss": 1.0699, + "step": 2172 + }, + { + "epoch": 0.16, + "learning_rate": 1.784072249589491e-05, + "loss": 1.0038, + "step": 2173 + }, + { + "epoch": 0.16, + "learning_rate": 1.7848932676518885e-05, + "loss": 1.0326, + "step": 2174 + }, + { + "epoch": 0.16, + "learning_rate": 1.785714285714286e-05, + "loss": 1.0622, + "step": 2175 + }, + { + "epoch": 0.16, + "learning_rate": 1.7865353037766832e-05, + "loss": 1.0528, + "step": 2176 + }, + { + "epoch": 0.16, + "learning_rate": 1.7873563218390806e-05, + "loss": 1.1403, + "step": 2177 + }, + { + "epoch": 0.16, + "learning_rate": 1.788177339901478e-05, + "loss": 1.0855, + "step": 2178 + }, + { + "epoch": 0.16, + "learning_rate": 1.7889983579638754e-05, + "loss": 1.0019, + "step": 2179 + }, + { + "epoch": 0.16, + "learning_rate": 1.7898193760262728e-05, + "loss": 1.0845, + "step": 2180 + }, + { + "epoch": 0.16, + "learning_rate": 1.7906403940886702e-05, + "loss": 1.1298, + "step": 2181 + }, + { + "epoch": 0.16, + "learning_rate": 1.7914614121510676e-05, + "loss": 1.1194, + "step": 2182 + }, + { + "epoch": 0.16, + "learning_rate": 1.792282430213465e-05, + "loss": 0.9237, + "step": 2183 + }, + { + "epoch": 0.16, + "learning_rate": 1.7931034482758623e-05, + "loss": 1.1162, + "step": 2184 + }, + { + "epoch": 0.16, + "learning_rate": 1.7939244663382597e-05, + "loss": 1.0617, + "step": 2185 + }, + { + "epoch": 0.16, + "learning_rate": 1.7947454844006568e-05, + "loss": 1.0646, + "step": 2186 + }, + { + "epoch": 0.16, + "learning_rate": 1.7955665024630545e-05, + "loss": 1.0121, + "step": 2187 + }, + { + "epoch": 0.16, + "learning_rate": 1.796387520525452e-05, + "loss": 1.0329, + "step": 2188 + }, + { + "epoch": 0.16, + "learning_rate": 1.797208538587849e-05, + "loss": 1.1256, + "step": 2189 + }, + { + "epoch": 0.16, + "learning_rate": 1.7980295566502463e-05, + "loss": 1.0341, + "step": 2190 + }, + { + "epoch": 0.16, + "learning_rate": 1.798850574712644e-05, + "loss": 1.0796, + "step": 2191 + }, + { + "epoch": 0.16, + "learning_rate": 1.799671592775041e-05, + "loss": 1.1001, + "step": 2192 + }, + { + "epoch": 0.16, + "learning_rate": 1.8004926108374385e-05, + "loss": 1.1204, + "step": 2193 + }, + { + "epoch": 0.16, + "learning_rate": 1.801313628899836e-05, + "loss": 1.1607, + "step": 2194 + }, + { + "epoch": 0.16, + "learning_rate": 1.8021346469622333e-05, + "loss": 1.1445, + "step": 2195 + }, + { + "epoch": 0.16, + "learning_rate": 1.8029556650246306e-05, + "loss": 1.0738, + "step": 2196 + }, + { + "epoch": 0.16, + "learning_rate": 1.803776683087028e-05, + "loss": 1.0889, + "step": 2197 + }, + { + "epoch": 0.16, + "learning_rate": 1.8045977011494254e-05, + "loss": 1.0124, + "step": 2198 + }, + { + "epoch": 0.16, + "learning_rate": 1.8054187192118228e-05, + "loss": 1.0488, + "step": 2199 + }, + { + "epoch": 0.16, + "learning_rate": 1.8062397372742202e-05, + "loss": 1.0055, + "step": 2200 + }, + { + "epoch": 0.16, + "learning_rate": 1.8070607553366176e-05, + "loss": 1.0041, + "step": 2201 + }, + { + "epoch": 0.16, + "learning_rate": 1.807881773399015e-05, + "loss": 1.0416, + "step": 2202 + }, + { + "epoch": 0.16, + "learning_rate": 1.8087027914614123e-05, + "loss": 1.0977, + "step": 2203 + }, + { + "epoch": 0.16, + "learning_rate": 1.8095238095238097e-05, + "loss": 0.9407, + "step": 2204 + }, + { + "epoch": 0.16, + "learning_rate": 1.810344827586207e-05, + "loss": 0.9606, + "step": 2205 + }, + { + "epoch": 0.16, + "learning_rate": 1.8111658456486042e-05, + "loss": 0.9815, + "step": 2206 + }, + { + "epoch": 0.16, + "learning_rate": 1.811986863711002e-05, + "loss": 1.008, + "step": 2207 + }, + { + "epoch": 0.16, + "learning_rate": 1.8128078817733993e-05, + "loss": 1.1014, + "step": 2208 + }, + { + "epoch": 0.16, + "learning_rate": 1.8136288998357963e-05, + "loss": 1.0161, + "step": 2209 + }, + { + "epoch": 0.16, + "learning_rate": 1.8144499178981937e-05, + "loss": 1.0815, + "step": 2210 + }, + { + "epoch": 0.16, + "learning_rate": 1.8152709359605914e-05, + "loss": 0.999, + "step": 2211 + }, + { + "epoch": 0.16, + "learning_rate": 1.8160919540229885e-05, + "loss": 1.1371, + "step": 2212 + }, + { + "epoch": 0.16, + "learning_rate": 1.816912972085386e-05, + "loss": 1.1028, + "step": 2213 + }, + { + "epoch": 0.16, + "learning_rate": 1.8177339901477833e-05, + "loss": 1.0947, + "step": 2214 + }, + { + "epoch": 0.16, + "learning_rate": 1.818555008210181e-05, + "loss": 1.0811, + "step": 2215 + }, + { + "epoch": 0.16, + "learning_rate": 1.819376026272578e-05, + "loss": 1.0741, + "step": 2216 + }, + { + "epoch": 0.16, + "learning_rate": 1.8201970443349754e-05, + "loss": 1.0141, + "step": 2217 + }, + { + "epoch": 0.16, + "learning_rate": 1.821018062397373e-05, + "loss": 1.0677, + "step": 2218 + }, + { + "epoch": 0.16, + "learning_rate": 1.8218390804597702e-05, + "loss": 0.9855, + "step": 2219 + }, + { + "epoch": 0.16, + "learning_rate": 1.8226600985221676e-05, + "loss": 1.0324, + "step": 2220 + }, + { + "epoch": 0.16, + "learning_rate": 1.823481116584565e-05, + "loss": 1.0708, + "step": 2221 + }, + { + "epoch": 0.16, + "learning_rate": 1.8243021346469624e-05, + "loss": 1.0739, + "step": 2222 + }, + { + "epoch": 0.16, + "learning_rate": 1.8251231527093598e-05, + "loss": 0.9927, + "step": 2223 + }, + { + "epoch": 0.16, + "learning_rate": 1.825944170771757e-05, + "loss": 1.1747, + "step": 2224 + }, + { + "epoch": 0.16, + "learning_rate": 1.8267651888341545e-05, + "loss": 1.1117, + "step": 2225 + }, + { + "epoch": 0.16, + "learning_rate": 1.827586206896552e-05, + "loss": 1.0743, + "step": 2226 + }, + { + "epoch": 0.16, + "learning_rate": 1.8284072249589493e-05, + "loss": 1.1582, + "step": 2227 + }, + { + "epoch": 0.16, + "learning_rate": 1.8292282430213467e-05, + "loss": 1.1701, + "step": 2228 + }, + { + "epoch": 0.16, + "learning_rate": 1.830049261083744e-05, + "loss": 1.0267, + "step": 2229 + }, + { + "epoch": 0.16, + "learning_rate": 1.8308702791461415e-05, + "loss": 1.1306, + "step": 2230 + }, + { + "epoch": 0.16, + "learning_rate": 1.831691297208539e-05, + "loss": 1.158, + "step": 2231 + }, + { + "epoch": 0.16, + "learning_rate": 1.8325123152709362e-05, + "loss": 1.1139, + "step": 2232 + }, + { + "epoch": 0.16, + "learning_rate": 1.8333333333333333e-05, + "loss": 1.1327, + "step": 2233 + }, + { + "epoch": 0.17, + "learning_rate": 1.834154351395731e-05, + "loss": 1.0979, + "step": 2234 + }, + { + "epoch": 0.17, + "learning_rate": 1.8349753694581284e-05, + "loss": 1.0795, + "step": 2235 + }, + { + "epoch": 0.17, + "learning_rate": 1.8357963875205254e-05, + "loss": 1.1076, + "step": 2236 + }, + { + "epoch": 0.17, + "learning_rate": 1.836617405582923e-05, + "loss": 1.1406, + "step": 2237 + }, + { + "epoch": 0.17, + "learning_rate": 1.8374384236453206e-05, + "loss": 1.1234, + "step": 2238 + }, + { + "epoch": 0.17, + "learning_rate": 1.8382594417077176e-05, + "loss": 0.9902, + "step": 2239 + }, + { + "epoch": 0.17, + "learning_rate": 1.839080459770115e-05, + "loss": 1.1332, + "step": 2240 + }, + { + "epoch": 0.17, + "learning_rate": 1.8399014778325124e-05, + "loss": 1.1414, + "step": 2241 + }, + { + "epoch": 0.17, + "learning_rate": 1.8407224958949098e-05, + "loss": 1.1508, + "step": 2242 + }, + { + "epoch": 0.17, + "learning_rate": 1.841543513957307e-05, + "loss": 1.0407, + "step": 2243 + }, + { + "epoch": 0.17, + "learning_rate": 1.8423645320197045e-05, + "loss": 1.0599, + "step": 2244 + }, + { + "epoch": 0.17, + "learning_rate": 1.843185550082102e-05, + "loss": 0.9697, + "step": 2245 + }, + { + "epoch": 0.17, + "learning_rate": 1.8440065681444993e-05, + "loss": 1.0873, + "step": 2246 + }, + { + "epoch": 0.17, + "learning_rate": 1.8448275862068967e-05, + "loss": 0.9998, + "step": 2247 + }, + { + "epoch": 0.17, + "learning_rate": 1.845648604269294e-05, + "loss": 1.0277, + "step": 2248 + }, + { + "epoch": 0.17, + "learning_rate": 1.8464696223316915e-05, + "loss": 1.0345, + "step": 2249 + }, + { + "epoch": 0.17, + "learning_rate": 1.847290640394089e-05, + "loss": 1.0805, + "step": 2250 + }, + { + "epoch": 0.17, + "learning_rate": 1.8481116584564863e-05, + "loss": 1.1058, + "step": 2251 + }, + { + "epoch": 0.17, + "learning_rate": 1.8489326765188836e-05, + "loss": 1.1209, + "step": 2252 + }, + { + "epoch": 0.17, + "learning_rate": 1.8497536945812807e-05, + "loss": 0.9511, + "step": 2253 + }, + { + "epoch": 0.17, + "learning_rate": 1.8505747126436784e-05, + "loss": 1.0664, + "step": 2254 + }, + { + "epoch": 0.17, + "learning_rate": 1.8513957307060758e-05, + "loss": 1.0474, + "step": 2255 + }, + { + "epoch": 0.17, + "learning_rate": 1.852216748768473e-05, + "loss": 1.0836, + "step": 2256 + }, + { + "epoch": 0.17, + "learning_rate": 1.8530377668308702e-05, + "loss": 1.0149, + "step": 2257 + }, + { + "epoch": 0.17, + "learning_rate": 1.853858784893268e-05, + "loss": 1.0842, + "step": 2258 + }, + { + "epoch": 0.17, + "learning_rate": 1.8546798029556654e-05, + "loss": 0.9892, + "step": 2259 + }, + { + "epoch": 0.17, + "learning_rate": 1.8555008210180624e-05, + "loss": 1.0057, + "step": 2260 + }, + { + "epoch": 0.17, + "learning_rate": 1.8563218390804598e-05, + "loss": 1.075, + "step": 2261 + }, + { + "epoch": 0.17, + "learning_rate": 1.8571428571428575e-05, + "loss": 1.018, + "step": 2262 + }, + { + "epoch": 0.17, + "learning_rate": 1.8579638752052546e-05, + "loss": 1.0826, + "step": 2263 + }, + { + "epoch": 0.17, + "learning_rate": 1.858784893267652e-05, + "loss": 1.0553, + "step": 2264 + }, + { + "epoch": 0.17, + "learning_rate": 1.8596059113300493e-05, + "loss": 1.1647, + "step": 2265 + }, + { + "epoch": 0.17, + "learning_rate": 1.8604269293924467e-05, + "loss": 1.0748, + "step": 2266 + }, + { + "epoch": 0.17, + "learning_rate": 1.861247947454844e-05, + "loss": 0.9957, + "step": 2267 + }, + { + "epoch": 0.17, + "learning_rate": 1.8620689655172415e-05, + "loss": 1.0748, + "step": 2268 + }, + { + "epoch": 0.17, + "learning_rate": 1.862889983579639e-05, + "loss": 1.0807, + "step": 2269 + }, + { + "epoch": 0.17, + "learning_rate": 1.8637110016420363e-05, + "loss": 1.1178, + "step": 2270 + }, + { + "epoch": 0.17, + "learning_rate": 1.8645320197044337e-05, + "loss": 1.1072, + "step": 2271 + }, + { + "epoch": 0.17, + "learning_rate": 1.865353037766831e-05, + "loss": 1.1371, + "step": 2272 + }, + { + "epoch": 0.17, + "learning_rate": 1.8661740558292284e-05, + "loss": 1.1362, + "step": 2273 + }, + { + "epoch": 0.17, + "learning_rate": 1.8669950738916258e-05, + "loss": 1.0853, + "step": 2274 + }, + { + "epoch": 0.17, + "learning_rate": 1.8678160919540232e-05, + "loss": 1.0344, + "step": 2275 + }, + { + "epoch": 0.17, + "learning_rate": 1.8686371100164206e-05, + "loss": 1.0203, + "step": 2276 + }, + { + "epoch": 0.17, + "learning_rate": 1.869458128078818e-05, + "loss": 0.9758, + "step": 2277 + }, + { + "epoch": 0.17, + "learning_rate": 1.8702791461412154e-05, + "loss": 1.001, + "step": 2278 + }, + { + "epoch": 0.17, + "learning_rate": 1.8711001642036128e-05, + "loss": 1.0431, + "step": 2279 + }, + { + "epoch": 0.17, + "learning_rate": 1.8719211822660098e-05, + "loss": 1.002, + "step": 2280 + }, + { + "epoch": 0.17, + "learning_rate": 1.8727422003284075e-05, + "loss": 1.1988, + "step": 2281 + }, + { + "epoch": 0.17, + "learning_rate": 1.873563218390805e-05, + "loss": 1.1223, + "step": 2282 + }, + { + "epoch": 0.17, + "learning_rate": 1.874384236453202e-05, + "loss": 1.0619, + "step": 2283 + }, + { + "epoch": 0.17, + "learning_rate": 1.8752052545155994e-05, + "loss": 1.0539, + "step": 2284 + }, + { + "epoch": 0.17, + "learning_rate": 1.876026272577997e-05, + "loss": 1.0418, + "step": 2285 + }, + { + "epoch": 0.17, + "learning_rate": 1.876847290640394e-05, + "loss": 1.1544, + "step": 2286 + }, + { + "epoch": 0.17, + "learning_rate": 1.8776683087027915e-05, + "loss": 1.0683, + "step": 2287 + }, + { + "epoch": 0.17, + "learning_rate": 1.878489326765189e-05, + "loss": 1.1013, + "step": 2288 + }, + { + "epoch": 0.17, + "learning_rate": 1.8793103448275863e-05, + "loss": 1.0463, + "step": 2289 + }, + { + "epoch": 0.17, + "learning_rate": 1.8801313628899837e-05, + "loss": 1.0544, + "step": 2290 + }, + { + "epoch": 0.17, + "learning_rate": 1.880952380952381e-05, + "loss": 1.144, + "step": 2291 + }, + { + "epoch": 0.17, + "learning_rate": 1.8817733990147784e-05, + "loss": 1.0986, + "step": 2292 + }, + { + "epoch": 0.17, + "learning_rate": 1.882594417077176e-05, + "loss": 1.0591, + "step": 2293 + }, + { + "epoch": 0.17, + "learning_rate": 1.8834154351395732e-05, + "loss": 1.1199, + "step": 2294 + }, + { + "epoch": 0.17, + "learning_rate": 1.8842364532019706e-05, + "loss": 1.1726, + "step": 2295 + }, + { + "epoch": 0.17, + "learning_rate": 1.885057471264368e-05, + "loss": 1.0749, + "step": 2296 + }, + { + "epoch": 0.17, + "learning_rate": 1.8858784893267654e-05, + "loss": 1.1665, + "step": 2297 + }, + { + "epoch": 0.17, + "learning_rate": 1.8866995073891628e-05, + "loss": 1.0722, + "step": 2298 + }, + { + "epoch": 0.17, + "learning_rate": 1.88752052545156e-05, + "loss": 1.0417, + "step": 2299 + }, + { + "epoch": 0.17, + "learning_rate": 1.8883415435139572e-05, + "loss": 1.1235, + "step": 2300 + }, + { + "epoch": 0.17, + "learning_rate": 1.889162561576355e-05, + "loss": 1.0897, + "step": 2301 + }, + { + "epoch": 0.17, + "learning_rate": 1.8899835796387523e-05, + "loss": 1.0316, + "step": 2302 + }, + { + "epoch": 0.17, + "learning_rate": 1.8908045977011497e-05, + "loss": 1.0472, + "step": 2303 + }, + { + "epoch": 0.17, + "learning_rate": 1.8916256157635468e-05, + "loss": 1.0874, + "step": 2304 + }, + { + "epoch": 0.17, + "learning_rate": 1.8924466338259445e-05, + "loss": 1.0315, + "step": 2305 + }, + { + "epoch": 0.17, + "learning_rate": 1.893267651888342e-05, + "loss": 1.1362, + "step": 2306 + }, + { + "epoch": 0.17, + "learning_rate": 1.894088669950739e-05, + "loss": 1.1082, + "step": 2307 + }, + { + "epoch": 0.17, + "learning_rate": 1.8949096880131363e-05, + "loss": 1.1071, + "step": 2308 + }, + { + "epoch": 0.17, + "learning_rate": 1.895730706075534e-05, + "loss": 1.0544, + "step": 2309 + }, + { + "epoch": 0.17, + "learning_rate": 1.896551724137931e-05, + "loss": 1.0556, + "step": 2310 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973727422003285e-05, + "loss": 0.9923, + "step": 2311 + }, + { + "epoch": 0.17, + "learning_rate": 1.898193760262726e-05, + "loss": 1.1906, + "step": 2312 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990147783251232e-05, + "loss": 1.1281, + "step": 2313 + }, + { + "epoch": 0.17, + "learning_rate": 1.8998357963875206e-05, + "loss": 1.0795, + "step": 2314 + }, + { + "epoch": 0.17, + "learning_rate": 1.900656814449918e-05, + "loss": 1.1566, + "step": 2315 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014778325123154e-05, + "loss": 1.0058, + "step": 2316 + }, + { + "epoch": 0.17, + "learning_rate": 1.9022988505747128e-05, + "loss": 0.9866, + "step": 2317 + }, + { + "epoch": 0.17, + "learning_rate": 1.9031198686371102e-05, + "loss": 1.0657, + "step": 2318 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039408866995076e-05, + "loss": 1.0084, + "step": 2319 + }, + { + "epoch": 0.17, + "learning_rate": 1.904761904761905e-05, + "loss": 1.097, + "step": 2320 + }, + { + "epoch": 0.17, + "learning_rate": 1.9055829228243023e-05, + "loss": 0.8812, + "step": 2321 + }, + { + "epoch": 0.17, + "learning_rate": 1.9064039408866997e-05, + "loss": 1.0517, + "step": 2322 + }, + { + "epoch": 0.17, + "learning_rate": 1.907224958949097e-05, + "loss": 1.1103, + "step": 2323 + }, + { + "epoch": 0.17, + "learning_rate": 1.908045977011494e-05, + "loss": 1.1305, + "step": 2324 + }, + { + "epoch": 0.17, + "learning_rate": 1.908866995073892e-05, + "loss": 1.1848, + "step": 2325 + }, + { + "epoch": 0.17, + "learning_rate": 1.9096880131362893e-05, + "loss": 1.0066, + "step": 2326 + }, + { + "epoch": 0.17, + "learning_rate": 1.9105090311986863e-05, + "loss": 1.0624, + "step": 2327 + }, + { + "epoch": 0.17, + "learning_rate": 1.911330049261084e-05, + "loss": 1.0793, + "step": 2328 + }, + { + "epoch": 0.17, + "learning_rate": 1.9121510673234814e-05, + "loss": 1.0634, + "step": 2329 + }, + { + "epoch": 0.17, + "learning_rate": 1.9129720853858785e-05, + "loss": 1.0834, + "step": 2330 + }, + { + "epoch": 0.17, + "learning_rate": 1.913793103448276e-05, + "loss": 1.0142, + "step": 2331 + }, + { + "epoch": 0.17, + "learning_rate": 1.9146141215106736e-05, + "loss": 1.0425, + "step": 2332 + }, + { + "epoch": 0.17, + "learning_rate": 1.9154351395730706e-05, + "loss": 1.067, + "step": 2333 + }, + { + "epoch": 0.17, + "learning_rate": 1.916256157635468e-05, + "loss": 1.0326, + "step": 2334 + }, + { + "epoch": 0.17, + "learning_rate": 1.9170771756978654e-05, + "loss": 0.9615, + "step": 2335 + }, + { + "epoch": 0.17, + "learning_rate": 1.917898193760263e-05, + "loss": 0.9446, + "step": 2336 + }, + { + "epoch": 0.17, + "learning_rate": 1.9187192118226602e-05, + "loss": 1.1613, + "step": 2337 + }, + { + "epoch": 0.17, + "learning_rate": 1.9195402298850576e-05, + "loss": 1.1351, + "step": 2338 + }, + { + "epoch": 0.17, + "learning_rate": 1.920361247947455e-05, + "loss": 1.1416, + "step": 2339 + }, + { + "epoch": 0.17, + "learning_rate": 1.9211822660098524e-05, + "loss": 1.2009, + "step": 2340 + }, + { + "epoch": 0.17, + "learning_rate": 1.9220032840722497e-05, + "loss": 1.1307, + "step": 2341 + }, + { + "epoch": 0.17, + "learning_rate": 1.922824302134647e-05, + "loss": 0.9066, + "step": 2342 + }, + { + "epoch": 0.17, + "learning_rate": 1.9236453201970445e-05, + "loss": 1.0408, + "step": 2343 + }, + { + "epoch": 0.17, + "learning_rate": 1.924466338259442e-05, + "loss": 0.9908, + "step": 2344 + }, + { + "epoch": 0.17, + "learning_rate": 1.9252873563218393e-05, + "loss": 1.0074, + "step": 2345 + }, + { + "epoch": 0.17, + "learning_rate": 1.9261083743842367e-05, + "loss": 1.0155, + "step": 2346 + }, + { + "epoch": 0.17, + "learning_rate": 1.926929392446634e-05, + "loss": 1.0604, + "step": 2347 + }, + { + "epoch": 0.17, + "learning_rate": 1.9277504105090315e-05, + "loss": 1.1278, + "step": 2348 + }, + { + "epoch": 0.17, + "learning_rate": 1.928571428571429e-05, + "loss": 1.0582, + "step": 2349 + }, + { + "epoch": 0.17, + "learning_rate": 1.9293924466338262e-05, + "loss": 1.0599, + "step": 2350 + }, + { + "epoch": 0.17, + "learning_rate": 1.9302134646962233e-05, + "loss": 0.9744, + "step": 2351 + }, + { + "epoch": 0.17, + "learning_rate": 1.931034482758621e-05, + "loss": 1.056, + "step": 2352 + }, + { + "epoch": 0.17, + "learning_rate": 1.9318555008210184e-05, + "loss": 1.1369, + "step": 2353 + }, + { + "epoch": 0.17, + "learning_rate": 1.9326765188834154e-05, + "loss": 1.0791, + "step": 2354 + }, + { + "epoch": 0.17, + "learning_rate": 1.9334975369458128e-05, + "loss": 1.0985, + "step": 2355 + }, + { + "epoch": 0.17, + "learning_rate": 1.9343185550082105e-05, + "loss": 1.068, + "step": 2356 + }, + { + "epoch": 0.17, + "learning_rate": 1.9351395730706076e-05, + "loss": 0.9749, + "step": 2357 + }, + { + "epoch": 0.17, + "learning_rate": 1.935960591133005e-05, + "loss": 1.0958, + "step": 2358 + }, + { + "epoch": 0.17, + "learning_rate": 1.9367816091954024e-05, + "loss": 1.1304, + "step": 2359 + }, + { + "epoch": 0.17, + "learning_rate": 1.9376026272577998e-05, + "loss": 1.1152, + "step": 2360 + }, + { + "epoch": 0.17, + "learning_rate": 1.938423645320197e-05, + "loss": 1.037, + "step": 2361 + }, + { + "epoch": 0.17, + "learning_rate": 1.9392446633825945e-05, + "loss": 1.0291, + "step": 2362 + }, + { + "epoch": 0.17, + "learning_rate": 1.940065681444992e-05, + "loss": 1.0481, + "step": 2363 + }, + { + "epoch": 0.17, + "learning_rate": 1.9408866995073893e-05, + "loss": 1.0816, + "step": 2364 + }, + { + "epoch": 0.17, + "learning_rate": 1.9417077175697867e-05, + "loss": 1.1897, + "step": 2365 + }, + { + "epoch": 0.17, + "learning_rate": 1.942528735632184e-05, + "loss": 1.024, + "step": 2366 + }, + { + "epoch": 0.17, + "learning_rate": 1.9433497536945815e-05, + "loss": 0.9512, + "step": 2367 + }, + { + "epoch": 0.17, + "learning_rate": 1.944170771756979e-05, + "loss": 1.0701, + "step": 2368 + }, + { + "epoch": 0.18, + "learning_rate": 1.9449917898193762e-05, + "loss": 1.2173, + "step": 2369 + }, + { + "epoch": 0.18, + "learning_rate": 1.9458128078817736e-05, + "loss": 1.1445, + "step": 2370 + }, + { + "epoch": 0.18, + "learning_rate": 1.9466338259441707e-05, + "loss": 1.0938, + "step": 2371 + }, + { + "epoch": 0.18, + "learning_rate": 1.9474548440065684e-05, + "loss": 1.0867, + "step": 2372 + }, + { + "epoch": 0.18, + "learning_rate": 1.9482758620689658e-05, + "loss": 1.0602, + "step": 2373 + }, + { + "epoch": 0.18, + "learning_rate": 1.949096880131363e-05, + "loss": 1.0549, + "step": 2374 + }, + { + "epoch": 0.18, + "learning_rate": 1.9499178981937602e-05, + "loss": 1.0823, + "step": 2375 + }, + { + "epoch": 0.18, + "learning_rate": 1.950738916256158e-05, + "loss": 1.0084, + "step": 2376 + }, + { + "epoch": 0.18, + "learning_rate": 1.9515599343185553e-05, + "loss": 0.9874, + "step": 2377 + }, + { + "epoch": 0.18, + "learning_rate": 1.9523809523809524e-05, + "loss": 1.1351, + "step": 2378 + }, + { + "epoch": 0.18, + "learning_rate": 1.9532019704433498e-05, + "loss": 1.0412, + "step": 2379 + }, + { + "epoch": 0.18, + "learning_rate": 1.9540229885057475e-05, + "loss": 1.1274, + "step": 2380 + }, + { + "epoch": 0.18, + "learning_rate": 1.9548440065681445e-05, + "loss": 0.9122, + "step": 2381 + }, + { + "epoch": 0.18, + "learning_rate": 1.955665024630542e-05, + "loss": 1.1103, + "step": 2382 + }, + { + "epoch": 0.18, + "learning_rate": 1.9564860426929397e-05, + "loss": 1.0784, + "step": 2383 + }, + { + "epoch": 0.18, + "learning_rate": 1.9573070607553367e-05, + "loss": 1.0667, + "step": 2384 + }, + { + "epoch": 0.18, + "learning_rate": 1.958128078817734e-05, + "loss": 1.0482, + "step": 2385 + }, + { + "epoch": 0.18, + "learning_rate": 1.9589490968801315e-05, + "loss": 1.0176, + "step": 2386 + }, + { + "epoch": 0.18, + "learning_rate": 1.959770114942529e-05, + "loss": 1.0418, + "step": 2387 + }, + { + "epoch": 0.18, + "learning_rate": 1.9605911330049263e-05, + "loss": 1.0572, + "step": 2388 + }, + { + "epoch": 0.18, + "learning_rate": 1.9614121510673236e-05, + "loss": 1.1019, + "step": 2389 + }, + { + "epoch": 0.18, + "learning_rate": 1.962233169129721e-05, + "loss": 1.1043, + "step": 2390 + }, + { + "epoch": 0.18, + "learning_rate": 1.9630541871921184e-05, + "loss": 1.1146, + "step": 2391 + }, + { + "epoch": 0.18, + "learning_rate": 1.9638752052545158e-05, + "loss": 1.0941, + "step": 2392 + }, + { + "epoch": 0.18, + "learning_rate": 1.9646962233169132e-05, + "loss": 1.0131, + "step": 2393 + }, + { + "epoch": 0.18, + "learning_rate": 1.9655172413793106e-05, + "loss": 1.0416, + "step": 2394 + }, + { + "epoch": 0.18, + "learning_rate": 1.966338259441708e-05, + "loss": 1.0612, + "step": 2395 + }, + { + "epoch": 0.18, + "learning_rate": 1.9671592775041054e-05, + "loss": 0.945, + "step": 2396 + }, + { + "epoch": 0.18, + "learning_rate": 1.9679802955665027e-05, + "loss": 0.966, + "step": 2397 + }, + { + "epoch": 0.18, + "learning_rate": 1.9688013136288998e-05, + "loss": 1.1056, + "step": 2398 + }, + { + "epoch": 0.18, + "learning_rate": 1.9696223316912975e-05, + "loss": 1.1099, + "step": 2399 + }, + { + "epoch": 0.18, + "learning_rate": 1.970443349753695e-05, + "loss": 1.206, + "step": 2400 + }, + { + "epoch": 0.18, + "learning_rate": 1.971264367816092e-05, + "loss": 1.0575, + "step": 2401 + }, + { + "epoch": 0.18, + "learning_rate": 1.9720853858784893e-05, + "loss": 1.0608, + "step": 2402 + }, + { + "epoch": 0.18, + "learning_rate": 1.972906403940887e-05, + "loss": 1.1169, + "step": 2403 + }, + { + "epoch": 0.18, + "learning_rate": 1.973727422003284e-05, + "loss": 1.0329, + "step": 2404 + }, + { + "epoch": 0.18, + "learning_rate": 1.9745484400656815e-05, + "loss": 1.1363, + "step": 2405 + }, + { + "epoch": 0.18, + "learning_rate": 1.975369458128079e-05, + "loss": 1.0893, + "step": 2406 + }, + { + "epoch": 0.18, + "learning_rate": 1.9761904761904763e-05, + "loss": 1.0376, + "step": 2407 + }, + { + "epoch": 0.18, + "learning_rate": 1.9770114942528737e-05, + "loss": 1.0154, + "step": 2408 + }, + { + "epoch": 0.18, + "learning_rate": 1.977832512315271e-05, + "loss": 1.0083, + "step": 2409 + }, + { + "epoch": 0.18, + "learning_rate": 1.9786535303776684e-05, + "loss": 1.0925, + "step": 2410 + }, + { + "epoch": 0.18, + "learning_rate": 1.9794745484400658e-05, + "loss": 1.0709, + "step": 2411 + }, + { + "epoch": 0.18, + "learning_rate": 1.9802955665024632e-05, + "loss": 1.1445, + "step": 2412 + }, + { + "epoch": 0.18, + "learning_rate": 1.9811165845648606e-05, + "loss": 0.9647, + "step": 2413 + }, + { + "epoch": 0.18, + "learning_rate": 1.981937602627258e-05, + "loss": 1.0227, + "step": 2414 + }, + { + "epoch": 0.18, + "learning_rate": 1.9827586206896554e-05, + "loss": 1.1262, + "step": 2415 + }, + { + "epoch": 0.18, + "learning_rate": 1.9835796387520528e-05, + "loss": 1.1621, + "step": 2416 + }, + { + "epoch": 0.18, + "learning_rate": 1.98440065681445e-05, + "loss": 1.1045, + "step": 2417 + }, + { + "epoch": 0.18, + "learning_rate": 1.9852216748768472e-05, + "loss": 1.1528, + "step": 2418 + }, + { + "epoch": 0.18, + "learning_rate": 1.986042692939245e-05, + "loss": 1.2439, + "step": 2419 + }, + { + "epoch": 0.18, + "learning_rate": 1.9868637110016423e-05, + "loss": 0.9034, + "step": 2420 + }, + { + "epoch": 0.18, + "learning_rate": 1.9876847290640397e-05, + "loss": 1.0067, + "step": 2421 + }, + { + "epoch": 0.18, + "learning_rate": 1.9885057471264367e-05, + "loss": 1.0931, + "step": 2422 + }, + { + "epoch": 0.18, + "learning_rate": 1.9893267651888345e-05, + "loss": 0.9636, + "step": 2423 + }, + { + "epoch": 0.18, + "learning_rate": 1.990147783251232e-05, + "loss": 0.9703, + "step": 2424 + }, + { + "epoch": 0.18, + "learning_rate": 1.990968801313629e-05, + "loss": 0.9634, + "step": 2425 + }, + { + "epoch": 0.18, + "learning_rate": 1.9917898193760263e-05, + "loss": 1.0939, + "step": 2426 + }, + { + "epoch": 0.18, + "learning_rate": 1.992610837438424e-05, + "loss": 1.1518, + "step": 2427 + }, + { + "epoch": 0.18, + "learning_rate": 1.993431855500821e-05, + "loss": 1.0339, + "step": 2428 + }, + { + "epoch": 0.18, + "learning_rate": 1.9942528735632185e-05, + "loss": 1.1103, + "step": 2429 + }, + { + "epoch": 0.18, + "learning_rate": 1.995073891625616e-05, + "loss": 1.0745, + "step": 2430 + }, + { + "epoch": 0.18, + "learning_rate": 1.9958949096880132e-05, + "loss": 1.104, + "step": 2431 + }, + { + "epoch": 0.18, + "learning_rate": 1.9967159277504106e-05, + "loss": 1.0877, + "step": 2432 + }, + { + "epoch": 0.18, + "learning_rate": 1.997536945812808e-05, + "loss": 1.0215, + "step": 2433 + }, + { + "epoch": 0.18, + "learning_rate": 1.9983579638752054e-05, + "loss": 1.1941, + "step": 2434 + }, + { + "epoch": 0.18, + "learning_rate": 1.9991789819376028e-05, + "loss": 0.9983, + "step": 2435 + }, + { + "epoch": 0.18, + "learning_rate": 2e-05, + "loss": 1.136, + "step": 2436 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999999204507e-05, + "loss": 1.1279, + "step": 2437 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999968180287e-05, + "loss": 1.1271, + "step": 2438 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999928405643e-05, + "loss": 1.0052, + "step": 2439 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999872721148e-05, + "loss": 1.0389, + "step": 2440 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999980112679e-05, + "loss": 1.046, + "step": 2441 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999713622577e-05, + "loss": 1.1698, + "step": 2442 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999961020851e-05, + "loss": 1.1773, + "step": 2443 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999490884584e-05, + "loss": 1.1488, + "step": 2444 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999355650804e-05, + "loss": 1.1022, + "step": 2445 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999204507166e-05, + "loss": 1.0885, + "step": 2446 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999999037453675e-05, + "loss": 1.0526, + "step": 2447 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999998854490323e-05, + "loss": 0.9739, + "step": 2448 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999865561712e-05, + "loss": 1.2555, + "step": 2449 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999998440834062e-05, + "loss": 1.1132, + "step": 2450 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999821014115e-05, + "loss": 1.0908, + "step": 2451 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999997963538385e-05, + "loss": 1.1538, + "step": 2452 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999997701025765e-05, + "loss": 1.0834, + "step": 2453 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999742260329e-05, + "loss": 1.0491, + "step": 2454 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999997128270966e-05, + "loss": 1.1442, + "step": 2455 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999996818028785e-05, + "loss": 1.2124, + "step": 2456 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999996491876753e-05, + "loss": 1.1694, + "step": 2457 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999996149814874e-05, + "loss": 1.0889, + "step": 2458 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999995791843138e-05, + "loss": 1.0671, + "step": 2459 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999541796156e-05, + "loss": 1.0529, + "step": 2460 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999995028170125e-05, + "loss": 1.1018, + "step": 2461 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999994622468845e-05, + "loss": 0.9944, + "step": 2462 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999994200857714e-05, + "loss": 1.0525, + "step": 2463 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999993763336737e-05, + "loss": 0.9347, + "step": 2464 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999330990591e-05, + "loss": 0.9944, + "step": 2465 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999284056524e-05, + "loss": 1.1376, + "step": 2466 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999992355314722e-05, + "loss": 1.0567, + "step": 2467 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999185415436e-05, + "loss": 1.1664, + "step": 2468 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999991337084155e-05, + "loss": 1.0734, + "step": 2469 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999080410411e-05, + "loss": 1.1467, + "step": 2470 + }, + { + "epoch": 0.18, + "learning_rate": 1.999999025521422e-05, + "loss": 1.0599, + "step": 2471 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999989690414487e-05, + "loss": 1.035, + "step": 2472 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999989109704914e-05, + "loss": 1.0091, + "step": 2473 + }, + { + "epoch": 0.18, + "learning_rate": 1.99999885130855e-05, + "loss": 1.0446, + "step": 2474 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999987900556247e-05, + "loss": 1.1453, + "step": 2475 + }, + { + "epoch": 0.18, + "learning_rate": 1.999998727211716e-05, + "loss": 1.1045, + "step": 2476 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999986627768237e-05, + "loss": 1.1079, + "step": 2477 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999985967509476e-05, + "loss": 1.1516, + "step": 2478 + }, + { + "epoch": 0.18, + "learning_rate": 1.999998529134088e-05, + "loss": 1.1577, + "step": 2479 + }, + { + "epoch": 0.18, + "learning_rate": 1.999998459926245e-05, + "loss": 1.0488, + "step": 2480 + }, + { + "epoch": 0.18, + "learning_rate": 1.999998389127419e-05, + "loss": 1.1065, + "step": 2481 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999983167376097e-05, + "loss": 1.0657, + "step": 2482 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999982427568173e-05, + "loss": 0.9938, + "step": 2483 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999981671850423e-05, + "loss": 1.1151, + "step": 2484 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999980900222842e-05, + "loss": 0.9488, + "step": 2485 + }, + { + "epoch": 0.18, + "learning_rate": 1.999998011268544e-05, + "loss": 0.9686, + "step": 2486 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999979309238204e-05, + "loss": 1.1343, + "step": 2487 + }, + { + "epoch": 0.18, + "learning_rate": 1.999997848988115e-05, + "loss": 1.0954, + "step": 2488 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999977654614274e-05, + "loss": 1.2078, + "step": 2489 + }, + { + "epoch": 0.18, + "learning_rate": 1.999997680343757e-05, + "loss": 1.0626, + "step": 2490 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999975936351054e-05, + "loss": 1.0461, + "step": 2491 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999975053354714e-05, + "loss": 1.1027, + "step": 2492 + }, + { + "epoch": 0.18, + "learning_rate": 1.999997415444856e-05, + "loss": 1.0395, + "step": 2493 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999973239632588e-05, + "loss": 1.0402, + "step": 2494 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999972308906802e-05, + "loss": 1.1279, + "step": 2495 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999971362271203e-05, + "loss": 1.0599, + "step": 2496 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999970399725794e-05, + "loss": 1.2062, + "step": 2497 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999969421270576e-05, + "loss": 0.9875, + "step": 2498 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999968426905548e-05, + "loss": 0.9636, + "step": 2499 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999967416630714e-05, + "loss": 0.8806, + "step": 2500 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999966390446073e-05, + "loss": 1.0928, + "step": 2501 + }, + { + "epoch": 0.18, + "learning_rate": 1.999996534835163e-05, + "loss": 0.9816, + "step": 2502 + }, + { + "epoch": 0.18, + "learning_rate": 1.9999964290347383e-05, + "loss": 1.1471, + "step": 2503 + }, + { + "epoch": 0.19, + "learning_rate": 1.999996321643334e-05, + "loss": 1.1138, + "step": 2504 + }, + { + "epoch": 0.19, + "learning_rate": 1.99999621266095e-05, + "loss": 1.1683, + "step": 2505 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999961020875857e-05, + "loss": 1.057, + "step": 2506 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999959899232423e-05, + "loss": 0.9985, + "step": 2507 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999958761679196e-05, + "loss": 1.0646, + "step": 2508 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999957608216172e-05, + "loss": 1.1277, + "step": 2509 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999956438843363e-05, + "loss": 1.1394, + "step": 2510 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999955253560768e-05, + "loss": 1.0521, + "step": 2511 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999954052368387e-05, + "loss": 1.0843, + "step": 2512 + }, + { + "epoch": 0.19, + "learning_rate": 1.999995283526622e-05, + "loss": 1.01, + "step": 2513 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999951602254274e-05, + "loss": 1.1476, + "step": 2514 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999950353332545e-05, + "loss": 1.0643, + "step": 2515 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999949088501037e-05, + "loss": 1.087, + "step": 2516 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999947807759756e-05, + "loss": 1.0559, + "step": 2517 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999946511108704e-05, + "loss": 1.0369, + "step": 2518 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999945198547875e-05, + "loss": 1.0207, + "step": 2519 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999943870077277e-05, + "loss": 1.048, + "step": 2520 + }, + { + "epoch": 0.19, + "learning_rate": 1.999994252569691e-05, + "loss": 1.0954, + "step": 2521 + }, + { + "epoch": 0.19, + "learning_rate": 1.999994116540678e-05, + "loss": 1.0396, + "step": 2522 + }, + { + "epoch": 0.19, + "learning_rate": 1.999993978920689e-05, + "loss": 1.1381, + "step": 2523 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999938397097233e-05, + "loss": 1.0903, + "step": 2524 + }, + { + "epoch": 0.19, + "learning_rate": 1.999993698907782e-05, + "loss": 1.0858, + "step": 2525 + }, + { + "epoch": 0.19, + "learning_rate": 1.999993556514865e-05, + "loss": 1.0824, + "step": 2526 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999934125309727e-05, + "loss": 1.1391, + "step": 2527 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999932669561048e-05, + "loss": 1.1444, + "step": 2528 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999931197902624e-05, + "loss": 1.0918, + "step": 2529 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999929710334448e-05, + "loss": 1.0544, + "step": 2530 + }, + { + "epoch": 0.19, + "learning_rate": 1.999992820685653e-05, + "loss": 1.1347, + "step": 2531 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999926687468866e-05, + "loss": 1.0762, + "step": 2532 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999925152171468e-05, + "loss": 1.0275, + "step": 2533 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999923600964327e-05, + "loss": 1.0486, + "step": 2534 + }, + { + "epoch": 0.19, + "learning_rate": 1.999992203384745e-05, + "loss": 1.0189, + "step": 2535 + }, + { + "epoch": 0.19, + "learning_rate": 1.999992045082084e-05, + "loss": 1.0792, + "step": 2536 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999918851884502e-05, + "loss": 1.08, + "step": 2537 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999917237038438e-05, + "loss": 1.064, + "step": 2538 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999915606282645e-05, + "loss": 1.0363, + "step": 2539 + }, + { + "epoch": 0.19, + "learning_rate": 1.999991395961713e-05, + "loss": 1.1032, + "step": 2540 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999912297041896e-05, + "loss": 1.057, + "step": 2541 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999910618556945e-05, + "loss": 0.9646, + "step": 2542 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999908924162277e-05, + "loss": 1.0842, + "step": 2543 + }, + { + "epoch": 0.19, + "learning_rate": 1.99999072138579e-05, + "loss": 1.064, + "step": 2544 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999905487643812e-05, + "loss": 1.1744, + "step": 2545 + }, + { + "epoch": 0.19, + "learning_rate": 1.999990374552002e-05, + "loss": 1.1373, + "step": 2546 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999901987486518e-05, + "loss": 1.0815, + "step": 2547 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999900213543323e-05, + "loss": 1.0124, + "step": 2548 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999898423690423e-05, + "loss": 1.0768, + "step": 2549 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999896617927832e-05, + "loss": 1.0679, + "step": 2550 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999894796255546e-05, + "loss": 1.123, + "step": 2551 + }, + { + "epoch": 0.19, + "learning_rate": 1.999989295867357e-05, + "loss": 1.1378, + "step": 2552 + }, + { + "epoch": 0.19, + "learning_rate": 1.999989110518191e-05, + "loss": 1.0733, + "step": 2553 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999889235780567e-05, + "loss": 1.0314, + "step": 2554 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999887350469537e-05, + "loss": 1.1668, + "step": 2555 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999885449248837e-05, + "loss": 0.9754, + "step": 2556 + }, + { + "epoch": 0.19, + "learning_rate": 1.999988353211846e-05, + "loss": 1.1272, + "step": 2557 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999881599078407e-05, + "loss": 1.0574, + "step": 2558 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999879650128687e-05, + "loss": 1.2202, + "step": 2559 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999877685269307e-05, + "loss": 1.0277, + "step": 2560 + }, + { + "epoch": 0.19, + "learning_rate": 1.999987570450026e-05, + "loss": 1.0572, + "step": 2561 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999873707821555e-05, + "loss": 1.106, + "step": 2562 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999871695233197e-05, + "loss": 1.0389, + "step": 2563 + }, + { + "epoch": 0.19, + "learning_rate": 1.999986966673518e-05, + "loss": 1.0786, + "step": 2564 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999867622327515e-05, + "loss": 1.0583, + "step": 2565 + }, + { + "epoch": 0.19, + "learning_rate": 1.999986556201021e-05, + "loss": 1.1511, + "step": 2566 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999863485783255e-05, + "loss": 1.0649, + "step": 2567 + }, + { + "epoch": 0.19, + "learning_rate": 1.999986139364666e-05, + "loss": 1.0225, + "step": 2568 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999859285600434e-05, + "loss": 1.0921, + "step": 2569 + }, + { + "epoch": 0.19, + "learning_rate": 1.999985716164457e-05, + "loss": 1.065, + "step": 2570 + }, + { + "epoch": 0.19, + "learning_rate": 1.999985502177908e-05, + "loss": 1.1003, + "step": 2571 + }, + { + "epoch": 0.19, + "learning_rate": 1.999985286600396e-05, + "loss": 1.0707, + "step": 2572 + }, + { + "epoch": 0.19, + "learning_rate": 1.999985069431922e-05, + "loss": 1.1555, + "step": 2573 + }, + { + "epoch": 0.19, + "learning_rate": 1.999984850672486e-05, + "loss": 1.019, + "step": 2574 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999846303220885e-05, + "loss": 0.9629, + "step": 2575 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999844083807295e-05, + "loss": 1.0831, + "step": 2576 + }, + { + "epoch": 0.19, + "learning_rate": 1.99998418484841e-05, + "loss": 1.1491, + "step": 2577 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999839597251298e-05, + "loss": 1.1567, + "step": 2578 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999837330108895e-05, + "loss": 1.0501, + "step": 2579 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999835047056892e-05, + "loss": 1.0271, + "step": 2580 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999832748095293e-05, + "loss": 0.9669, + "step": 2581 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999830433224105e-05, + "loss": 1.002, + "step": 2582 + }, + { + "epoch": 0.19, + "learning_rate": 1.999982810244333e-05, + "loss": 1.1209, + "step": 2583 + }, + { + "epoch": 0.19, + "learning_rate": 1.999982575575297e-05, + "loss": 1.1056, + "step": 2584 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999823393153033e-05, + "loss": 1.1155, + "step": 2585 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999821014643518e-05, + "loss": 1.0174, + "step": 2586 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999818620224433e-05, + "loss": 1.1225, + "step": 2587 + }, + { + "epoch": 0.19, + "learning_rate": 1.999981620989578e-05, + "loss": 1.0421, + "step": 2588 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999813783657563e-05, + "loss": 1.1475, + "step": 2589 + }, + { + "epoch": 0.19, + "learning_rate": 1.999981134150978e-05, + "loss": 1.1121, + "step": 2590 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999808883452447e-05, + "loss": 1.0981, + "step": 2591 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999806409485557e-05, + "loss": 1.0315, + "step": 2592 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999803919609118e-05, + "loss": 0.9498, + "step": 2593 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999801413823134e-05, + "loss": 1.0535, + "step": 2594 + }, + { + "epoch": 0.19, + "learning_rate": 1.999979889212761e-05, + "loss": 1.0939, + "step": 2595 + }, + { + "epoch": 0.19, + "learning_rate": 1.999979635452255e-05, + "loss": 1.0895, + "step": 2596 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999793801007957e-05, + "loss": 1.0664, + "step": 2597 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999791231583832e-05, + "loss": 1.0356, + "step": 2598 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999788646250186e-05, + "loss": 1.0513, + "step": 2599 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999786045007017e-05, + "loss": 1.0729, + "step": 2600 + }, + { + "epoch": 0.19, + "learning_rate": 1.999978342785433e-05, + "loss": 1.1001, + "step": 2601 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999780794792132e-05, + "loss": 1.1088, + "step": 2602 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999778145820426e-05, + "loss": 1.07, + "step": 2603 + }, + { + "epoch": 0.19, + "learning_rate": 1.999977548093922e-05, + "loss": 1.1454, + "step": 2604 + }, + { + "epoch": 0.19, + "learning_rate": 1.999977280014851e-05, + "loss": 1.0682, + "step": 2605 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999770103448306e-05, + "loss": 1.0587, + "step": 2606 + }, + { + "epoch": 0.19, + "learning_rate": 1.999976739083861e-05, + "loss": 0.9639, + "step": 2607 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999764662319428e-05, + "loss": 1.2569, + "step": 2608 + }, + { + "epoch": 0.19, + "learning_rate": 1.999976191789076e-05, + "loss": 1.1117, + "step": 2609 + }, + { + "epoch": 0.19, + "learning_rate": 1.999975915755262e-05, + "loss": 1.1646, + "step": 2610 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999756381305e-05, + "loss": 1.0582, + "step": 2611 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999753589147918e-05, + "loss": 1.1698, + "step": 2612 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999750781081367e-05, + "loss": 1.0815, + "step": 2613 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999747957105353e-05, + "loss": 1.1568, + "step": 2614 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999745117219887e-05, + "loss": 1.122, + "step": 2615 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999742261424968e-05, + "loss": 1.1619, + "step": 2616 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999739389720602e-05, + "loss": 1.0705, + "step": 2617 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999736502106795e-05, + "loss": 1.1023, + "step": 2618 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999733598583545e-05, + "loss": 1.0104, + "step": 2619 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999730679150868e-05, + "loss": 1.0924, + "step": 2620 + }, + { + "epoch": 0.19, + "learning_rate": 1.999972774380876e-05, + "loss": 1.0631, + "step": 2621 + }, + { + "epoch": 0.19, + "learning_rate": 1.999972479255723e-05, + "loss": 1.1646, + "step": 2622 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999721825396282e-05, + "loss": 0.9647, + "step": 2623 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999718842325918e-05, + "loss": 1.0919, + "step": 2624 + }, + { + "epoch": 0.19, + "learning_rate": 1.999971584334614e-05, + "loss": 1.0183, + "step": 2625 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999712828456963e-05, + "loss": 1.1861, + "step": 2626 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999709797658383e-05, + "loss": 1.0309, + "step": 2627 + }, + { + "epoch": 0.19, + "learning_rate": 1.999970675095041e-05, + "loss": 1.1448, + "step": 2628 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999703688333047e-05, + "loss": 1.059, + "step": 2629 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999700609806294e-05, + "loss": 1.1172, + "step": 2630 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999697515370166e-05, + "loss": 1.0501, + "step": 2631 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999694405024658e-05, + "loss": 1.0774, + "step": 2632 + }, + { + "epoch": 0.19, + "learning_rate": 1.999969127876978e-05, + "loss": 1.1283, + "step": 2633 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999688136605538e-05, + "loss": 1.2153, + "step": 2634 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999684978531936e-05, + "loss": 0.9613, + "step": 2635 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999681804548975e-05, + "loss": 1.0398, + "step": 2636 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999678614656666e-05, + "loss": 1.1978, + "step": 2637 + }, + { + "epoch": 0.19, + "learning_rate": 1.999967540885501e-05, + "loss": 1.1011, + "step": 2638 + }, + { + "epoch": 0.19, + "learning_rate": 1.9999672187144016e-05, + "loss": 1.194, + "step": 2639 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999668949523683e-05, + "loss": 1.1002, + "step": 2640 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999665695994022e-05, + "loss": 1.0952, + "step": 2641 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999662426555035e-05, + "loss": 1.0061, + "step": 2642 + }, + { + "epoch": 0.2, + "learning_rate": 1.999965914120673e-05, + "loss": 1.0763, + "step": 2643 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999655839949107e-05, + "loss": 1.1263, + "step": 2644 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999652522782175e-05, + "loss": 1.1335, + "step": 2645 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999649189705942e-05, + "loss": 1.0639, + "step": 2646 + }, + { + "epoch": 0.2, + "learning_rate": 1.999964584072041e-05, + "loss": 0.9958, + "step": 2647 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999642475825582e-05, + "loss": 0.9812, + "step": 2648 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999639095021468e-05, + "loss": 1.154, + "step": 2649 + }, + { + "epoch": 0.2, + "learning_rate": 1.999963569830807e-05, + "loss": 1.1976, + "step": 2650 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999632285685394e-05, + "loss": 1.0669, + "step": 2651 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999628857153447e-05, + "loss": 1.0683, + "step": 2652 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999625412712232e-05, + "loss": 1.0508, + "step": 2653 + }, + { + "epoch": 0.2, + "learning_rate": 1.999962195236176e-05, + "loss": 1.2107, + "step": 2654 + }, + { + "epoch": 0.2, + "learning_rate": 1.999961847610203e-05, + "loss": 1.0814, + "step": 2655 + }, + { + "epoch": 0.2, + "learning_rate": 1.999961498393305e-05, + "loss": 1.2136, + "step": 2656 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999611475854827e-05, + "loss": 0.9805, + "step": 2657 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999607951867362e-05, + "loss": 1.1358, + "step": 2658 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999604411970667e-05, + "loss": 1.0515, + "step": 2659 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999600856164742e-05, + "loss": 1.0991, + "step": 2660 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999597284449597e-05, + "loss": 1.1709, + "step": 2661 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999593696825235e-05, + "loss": 1.0032, + "step": 2662 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999590093291667e-05, + "loss": 1.0411, + "step": 2663 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999586473848885e-05, + "loss": 1.0562, + "step": 2664 + }, + { + "epoch": 0.2, + "learning_rate": 1.999958283849691e-05, + "loss": 0.9996, + "step": 2665 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999579187235742e-05, + "loss": 1.1201, + "step": 2666 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999575520065388e-05, + "loss": 1.011, + "step": 2667 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999571836985847e-05, + "loss": 0.9913, + "step": 2668 + }, + { + "epoch": 0.2, + "learning_rate": 1.999956813799713e-05, + "loss": 1.0209, + "step": 2669 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999564423099248e-05, + "loss": 1.0979, + "step": 2670 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999560692292203e-05, + "loss": 1.1454, + "step": 2671 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999556945575995e-05, + "loss": 1.1313, + "step": 2672 + }, + { + "epoch": 0.2, + "learning_rate": 1.999955318295064e-05, + "loss": 1.1666, + "step": 2673 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999549404416133e-05, + "loss": 1.0969, + "step": 2674 + }, + { + "epoch": 0.2, + "learning_rate": 1.999954560997249e-05, + "loss": 1.0956, + "step": 2675 + }, + { + "epoch": 0.2, + "learning_rate": 1.999954179961971e-05, + "loss": 1.0253, + "step": 2676 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999537973357806e-05, + "loss": 1.1485, + "step": 2677 + }, + { + "epoch": 0.2, + "learning_rate": 1.999953413118678e-05, + "loss": 1.0005, + "step": 2678 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999530273106637e-05, + "loss": 1.0674, + "step": 2679 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999526399117383e-05, + "loss": 1.0546, + "step": 2680 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999522509219027e-05, + "loss": 1.1272, + "step": 2681 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999518603411574e-05, + "loss": 1.0459, + "step": 2682 + }, + { + "epoch": 0.2, + "learning_rate": 1.999951468169503e-05, + "loss": 1.0622, + "step": 2683 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999510744069402e-05, + "loss": 1.1202, + "step": 2684 + }, + { + "epoch": 0.2, + "learning_rate": 1.999950679053469e-05, + "loss": 1.0832, + "step": 2685 + }, + { + "epoch": 0.2, + "learning_rate": 1.999950282109091e-05, + "loss": 0.9456, + "step": 2686 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999498835738066e-05, + "loss": 1.1028, + "step": 2687 + }, + { + "epoch": 0.2, + "learning_rate": 1.999949483447616e-05, + "loss": 1.1244, + "step": 2688 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999490817305204e-05, + "loss": 1.0339, + "step": 2689 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999486784225196e-05, + "loss": 1.0411, + "step": 2690 + }, + { + "epoch": 0.2, + "learning_rate": 1.999948273523615e-05, + "loss": 1.1244, + "step": 2691 + }, + { + "epoch": 0.2, + "learning_rate": 1.999947867033807e-05, + "loss": 1.1752, + "step": 2692 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999474589530963e-05, + "loss": 1.035, + "step": 2693 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999470492814833e-05, + "loss": 1.1653, + "step": 2694 + }, + { + "epoch": 0.2, + "learning_rate": 1.999946638018969e-05, + "loss": 1.0466, + "step": 2695 + }, + { + "epoch": 0.2, + "learning_rate": 1.999946225165554e-05, + "loss": 1.0602, + "step": 2696 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999458107212384e-05, + "loss": 1.0933, + "step": 2697 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999453946860237e-05, + "loss": 0.9987, + "step": 2698 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999449770599107e-05, + "loss": 1.0962, + "step": 2699 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999445578428988e-05, + "loss": 1.0121, + "step": 2700 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999441370349895e-05, + "loss": 1.0823, + "step": 2701 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999437146361837e-05, + "loss": 1.0246, + "step": 2702 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999432906464815e-05, + "loss": 1.0035, + "step": 2703 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999428650658838e-05, + "loss": 0.9906, + "step": 2704 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999424378943915e-05, + "loss": 1.0655, + "step": 2705 + }, + { + "epoch": 0.2, + "learning_rate": 1.999942009132005e-05, + "loss": 1.06, + "step": 2706 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999415787787253e-05, + "loss": 1.1259, + "step": 2707 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999411468345523e-05, + "loss": 0.9515, + "step": 2708 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999407132994875e-05, + "loss": 0.9815, + "step": 2709 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999402781735315e-05, + "loss": 1.1375, + "step": 2710 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999398414566847e-05, + "loss": 1.1298, + "step": 2711 + }, + { + "epoch": 0.2, + "learning_rate": 1.999939403148948e-05, + "loss": 1.0583, + "step": 2712 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999389632503222e-05, + "loss": 1.1758, + "step": 2713 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999385217608073e-05, + "loss": 0.9771, + "step": 2714 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999380786804052e-05, + "loss": 1.1019, + "step": 2715 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999376340091153e-05, + "loss": 1.039, + "step": 2716 + }, + { + "epoch": 0.2, + "learning_rate": 1.999937187746939e-05, + "loss": 1.0669, + "step": 2717 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999367398938773e-05, + "loss": 1.1122, + "step": 2718 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999362904499304e-05, + "loss": 1.0682, + "step": 2719 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999358394150992e-05, + "loss": 1.1531, + "step": 2720 + }, + { + "epoch": 0.2, + "learning_rate": 1.999935386789384e-05, + "loss": 1.031, + "step": 2721 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999349325727862e-05, + "loss": 1.0826, + "step": 2722 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999344767653065e-05, + "loss": 1.1078, + "step": 2723 + }, + { + "epoch": 0.2, + "learning_rate": 1.999934019366945e-05, + "loss": 1.1069, + "step": 2724 + }, + { + "epoch": 0.2, + "learning_rate": 1.999933560377703e-05, + "loss": 1.0041, + "step": 2725 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999330997975806e-05, + "loss": 1.0676, + "step": 2726 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999326376265796e-05, + "loss": 1.0463, + "step": 2727 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999321738646995e-05, + "loss": 1.1407, + "step": 2728 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999317085119418e-05, + "loss": 1.0399, + "step": 2729 + }, + { + "epoch": 0.2, + "learning_rate": 1.999931241568307e-05, + "loss": 0.993, + "step": 2730 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999307730337963e-05, + "loss": 1.0538, + "step": 2731 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999303029084098e-05, + "loss": 1.1508, + "step": 2732 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999298311921483e-05, + "loss": 0.9898, + "step": 2733 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999293578850127e-05, + "loss": 0.9636, + "step": 2734 + }, + { + "epoch": 0.2, + "learning_rate": 1.999928882987004e-05, + "loss": 1.1943, + "step": 2735 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999284064981227e-05, + "loss": 1.0633, + "step": 2736 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999279284183697e-05, + "loss": 1.0274, + "step": 2737 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999274487477455e-05, + "loss": 1.0985, + "step": 2738 + }, + { + "epoch": 0.2, + "learning_rate": 1.999926967486251e-05, + "loss": 1.0556, + "step": 2739 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999264846338873e-05, + "loss": 0.995, + "step": 2740 + }, + { + "epoch": 0.2, + "learning_rate": 1.999926000190655e-05, + "loss": 1.0765, + "step": 2741 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999255141565542e-05, + "loss": 1.1526, + "step": 2742 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999250265315866e-05, + "loss": 1.0369, + "step": 2743 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999245373157525e-05, + "loss": 1.0882, + "step": 2744 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999240465090527e-05, + "loss": 1.0001, + "step": 2745 + }, + { + "epoch": 0.2, + "learning_rate": 1.999923554111488e-05, + "loss": 1.1597, + "step": 2746 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999230601230595e-05, + "loss": 1.123, + "step": 2747 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999225645437674e-05, + "loss": 1.0742, + "step": 2748 + }, + { + "epoch": 0.2, + "learning_rate": 1.999922067373613e-05, + "loss": 1.0806, + "step": 2749 + }, + { + "epoch": 0.2, + "learning_rate": 1.999921568612597e-05, + "loss": 0.9853, + "step": 2750 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999210682607197e-05, + "loss": 0.9592, + "step": 2751 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999205663179825e-05, + "loss": 1.0049, + "step": 2752 + }, + { + "epoch": 0.2, + "learning_rate": 1.999920062784386e-05, + "loss": 1.1078, + "step": 2753 + }, + { + "epoch": 0.2, + "learning_rate": 1.999919557659931e-05, + "loss": 0.981, + "step": 2754 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999190509446183e-05, + "loss": 1.0301, + "step": 2755 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999185426384485e-05, + "loss": 1.0033, + "step": 2756 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999180327414228e-05, + "loss": 1.1066, + "step": 2757 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999175212535418e-05, + "loss": 1.0045, + "step": 2758 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999170081748062e-05, + "loss": 1.0234, + "step": 2759 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999164935052174e-05, + "loss": 1.0993, + "step": 2760 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999159772447754e-05, + "loss": 1.0613, + "step": 2761 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999154593934814e-05, + "loss": 1.1132, + "step": 2762 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999149399513363e-05, + "loss": 1.1239, + "step": 2763 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999144189183406e-05, + "loss": 1.1391, + "step": 2764 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999138962944954e-05, + "loss": 1.1896, + "step": 2765 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999133720798018e-05, + "loss": 1.1047, + "step": 2766 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999128462742603e-05, + "loss": 1.1481, + "step": 2767 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999123188778717e-05, + "loss": 1.0402, + "step": 2768 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999117898906366e-05, + "loss": 1.1582, + "step": 2769 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999112593125564e-05, + "loss": 1.0938, + "step": 2770 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999107271436318e-05, + "loss": 1.1001, + "step": 2771 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999101933838634e-05, + "loss": 1.1492, + "step": 2772 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999096580332523e-05, + "loss": 1.0652, + "step": 2773 + }, + { + "epoch": 0.2, + "learning_rate": 1.9999091210917992e-05, + "loss": 1.0174, + "step": 2774 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999085825595047e-05, + "loss": 1.0253, + "step": 2775 + }, + { + "epoch": 0.21, + "learning_rate": 1.99990804243637e-05, + "loss": 1.1221, + "step": 2776 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999075007223966e-05, + "loss": 1.0769, + "step": 2777 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999069574175837e-05, + "loss": 1.0477, + "step": 2778 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999064125219338e-05, + "loss": 1.0711, + "step": 2779 + }, + { + "epoch": 0.21, + "learning_rate": 1.999905866035447e-05, + "loss": 1.048, + "step": 2780 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999053179581242e-05, + "loss": 1.0609, + "step": 2781 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999047682899664e-05, + "loss": 1.0684, + "step": 2782 + }, + { + "epoch": 0.21, + "learning_rate": 1.999904217030974e-05, + "loss": 1.1205, + "step": 2783 + }, + { + "epoch": 0.21, + "learning_rate": 1.999903664181149e-05, + "loss": 1.101, + "step": 2784 + }, + { + "epoch": 0.21, + "learning_rate": 1.999903109740491e-05, + "loss": 1.0222, + "step": 2785 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999025537090017e-05, + "loss": 1.0347, + "step": 2786 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999019960866818e-05, + "loss": 1.1106, + "step": 2787 + }, + { + "epoch": 0.21, + "learning_rate": 1.999901436873532e-05, + "loss": 0.9998, + "step": 2788 + }, + { + "epoch": 0.21, + "learning_rate": 1.9999008760695534e-05, + "loss": 0.99, + "step": 2789 + }, + { + "epoch": 0.21, + "learning_rate": 1.999900313674747e-05, + "loss": 1.0689, + "step": 2790 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998997496891134e-05, + "loss": 1.003, + "step": 2791 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998991841126534e-05, + "loss": 1.0745, + "step": 2792 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998986169453682e-05, + "loss": 1.0085, + "step": 2793 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998980481872586e-05, + "loss": 1.0978, + "step": 2794 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998974778383253e-05, + "loss": 1.0752, + "step": 2795 + }, + { + "epoch": 0.21, + "learning_rate": 1.99989690589857e-05, + "loss": 0.9649, + "step": 2796 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998963323679924e-05, + "loss": 1.0195, + "step": 2797 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998957572465945e-05, + "loss": 1.0022, + "step": 2798 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998951805343766e-05, + "loss": 1.2016, + "step": 2799 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998946022313397e-05, + "loss": 1.1444, + "step": 2800 + }, + { + "epoch": 0.21, + "learning_rate": 1.999894022337485e-05, + "loss": 1.0922, + "step": 2801 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998934408528126e-05, + "loss": 1.1079, + "step": 2802 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998928577773248e-05, + "loss": 1.0809, + "step": 2803 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998922731110216e-05, + "loss": 1.1648, + "step": 2804 + }, + { + "epoch": 0.21, + "learning_rate": 1.999891686853904e-05, + "loss": 1.1587, + "step": 2805 + }, + { + "epoch": 0.21, + "learning_rate": 1.999891099005973e-05, + "loss": 1.1089, + "step": 2806 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998905095672293e-05, + "loss": 1.1255, + "step": 2807 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998899185376745e-05, + "loss": 1.1863, + "step": 2808 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998893259173088e-05, + "loss": 1.034, + "step": 2809 + }, + { + "epoch": 0.21, + "learning_rate": 1.999888731706134e-05, + "loss": 0.9982, + "step": 2810 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998881359041502e-05, + "loss": 1.1419, + "step": 2811 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998875385113586e-05, + "loss": 1.1772, + "step": 2812 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998869395277602e-05, + "loss": 1.0904, + "step": 2813 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998863389533564e-05, + "loss": 1.0068, + "step": 2814 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998857367881474e-05, + "loss": 1.0102, + "step": 2815 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998851330321347e-05, + "loss": 1.109, + "step": 2816 + }, + { + "epoch": 0.21, + "learning_rate": 1.999884527685319e-05, + "loss": 1.019, + "step": 2817 + }, + { + "epoch": 0.21, + "learning_rate": 1.999883920747701e-05, + "loss": 1.1254, + "step": 2818 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998833122192823e-05, + "loss": 1.1581, + "step": 2819 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998827021000636e-05, + "loss": 1.0194, + "step": 2820 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998820903900456e-05, + "loss": 1.024, + "step": 2821 + }, + { + "epoch": 0.21, + "learning_rate": 1.99988147708923e-05, + "loss": 1.1057, + "step": 2822 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998808621976165e-05, + "loss": 1.0861, + "step": 2823 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998802457152075e-05, + "loss": 1.1694, + "step": 2824 + }, + { + "epoch": 0.21, + "learning_rate": 1.999879627642003e-05, + "loss": 1.0249, + "step": 2825 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998790079780046e-05, + "loss": 1.0742, + "step": 2826 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998783867232124e-05, + "loss": 1.1655, + "step": 2827 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998777638776286e-05, + "loss": 1.0672, + "step": 2828 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998771394412536e-05, + "loss": 1.0281, + "step": 2829 + }, + { + "epoch": 0.21, + "learning_rate": 1.999876513414088e-05, + "loss": 1.0474, + "step": 2830 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998758857961333e-05, + "loss": 1.0764, + "step": 2831 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998752565873904e-05, + "loss": 1.0916, + "step": 2832 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998746257878606e-05, + "loss": 0.984, + "step": 2833 + }, + { + "epoch": 0.21, + "learning_rate": 1.999873993397544e-05, + "loss": 1.02, + "step": 2834 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998733594164426e-05, + "loss": 1.11, + "step": 2835 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998727238445568e-05, + "loss": 1.0521, + "step": 2836 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998720866818878e-05, + "loss": 1.0717, + "step": 2837 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998714479284368e-05, + "loss": 1.0694, + "step": 2838 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998708075842043e-05, + "loss": 1.04, + "step": 2839 + }, + { + "epoch": 0.21, + "learning_rate": 1.999870165649192e-05, + "loss": 1.058, + "step": 2840 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998695221234003e-05, + "loss": 1.0921, + "step": 2841 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998688770068307e-05, + "loss": 1.0679, + "step": 2842 + }, + { + "epoch": 0.21, + "learning_rate": 1.999868230299484e-05, + "loss": 1.0109, + "step": 2843 + }, + { + "epoch": 0.21, + "learning_rate": 1.999867582001361e-05, + "loss": 1.132, + "step": 2844 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998669321124635e-05, + "loss": 1.0372, + "step": 2845 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998662806327913e-05, + "loss": 1.0734, + "step": 2846 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998656275623466e-05, + "loss": 1.0911, + "step": 2847 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998649729011303e-05, + "loss": 1.0161, + "step": 2848 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998643166491423e-05, + "loss": 1.1504, + "step": 2849 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998636588063852e-05, + "loss": 0.9939, + "step": 2850 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998629993728592e-05, + "loss": 1.0954, + "step": 2851 + }, + { + "epoch": 0.21, + "learning_rate": 1.999862338348565e-05, + "loss": 1.1163, + "step": 2852 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998616757335044e-05, + "loss": 0.9217, + "step": 2853 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998610115276786e-05, + "loss": 1.1025, + "step": 2854 + }, + { + "epoch": 0.21, + "learning_rate": 1.999860345731088e-05, + "loss": 1.0711, + "step": 2855 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998596783437335e-05, + "loss": 1.1603, + "step": 2856 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998590093656167e-05, + "loss": 0.9154, + "step": 2857 + }, + { + "epoch": 0.21, + "learning_rate": 1.999858338796739e-05, + "loss": 1.0825, + "step": 2858 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998576666371004e-05, + "loss": 1.0684, + "step": 2859 + }, + { + "epoch": 0.21, + "learning_rate": 1.999856992886703e-05, + "loss": 1.1225, + "step": 2860 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998563175455473e-05, + "loss": 1.0887, + "step": 2861 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998556406136342e-05, + "loss": 1.1126, + "step": 2862 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998549620909654e-05, + "loss": 1.0346, + "step": 2863 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998542819775413e-05, + "loss": 1.0547, + "step": 2864 + }, + { + "epoch": 0.21, + "learning_rate": 1.999853600273364e-05, + "loss": 1.1128, + "step": 2865 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998529169784337e-05, + "loss": 0.9794, + "step": 2866 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998522320927514e-05, + "loss": 1.0489, + "step": 2867 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998515456163186e-05, + "loss": 1.0575, + "step": 2868 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998508575491362e-05, + "loss": 0.9937, + "step": 2869 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998501678912057e-05, + "loss": 1.0762, + "step": 2870 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998494766425276e-05, + "loss": 1.0744, + "step": 2871 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998487838031036e-05, + "loss": 1.0535, + "step": 2872 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998480893729342e-05, + "loss": 1.0691, + "step": 2873 + }, + { + "epoch": 0.21, + "learning_rate": 1.999847393352021e-05, + "loss": 1.0602, + "step": 2874 + }, + { + "epoch": 0.21, + "learning_rate": 1.999846695740365e-05, + "loss": 1.0764, + "step": 2875 + }, + { + "epoch": 0.21, + "learning_rate": 1.999845996537967e-05, + "loss": 1.1305, + "step": 2876 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998452957448283e-05, + "loss": 1.0995, + "step": 2877 + }, + { + "epoch": 0.21, + "learning_rate": 1.99984459336095e-05, + "loss": 0.9631, + "step": 2878 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998438893863334e-05, + "loss": 1.1372, + "step": 2879 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998431838209794e-05, + "loss": 1.0746, + "step": 2880 + }, + { + "epoch": 0.21, + "learning_rate": 1.999842476664889e-05, + "loss": 1.2113, + "step": 2881 + }, + { + "epoch": 0.21, + "learning_rate": 1.999841767918064e-05, + "loss": 1.1129, + "step": 2882 + }, + { + "epoch": 0.21, + "learning_rate": 1.999841057580505e-05, + "loss": 1.1247, + "step": 2883 + }, + { + "epoch": 0.21, + "learning_rate": 1.999840345652213e-05, + "loss": 1.143, + "step": 2884 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998396321331893e-05, + "loss": 1.0418, + "step": 2885 + }, + { + "epoch": 0.21, + "learning_rate": 1.999838917023435e-05, + "loss": 1.0132, + "step": 2886 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998382003229512e-05, + "loss": 1.1616, + "step": 2887 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998374820317393e-05, + "loss": 1.0195, + "step": 2888 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998367621498002e-05, + "loss": 1.0263, + "step": 2889 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998360406771353e-05, + "loss": 1.0791, + "step": 2890 + }, + { + "epoch": 0.21, + "learning_rate": 1.999835317613745e-05, + "loss": 1.0185, + "step": 2891 + }, + { + "epoch": 0.21, + "learning_rate": 1.999834592959632e-05, + "loss": 1.0717, + "step": 2892 + }, + { + "epoch": 0.21, + "learning_rate": 1.999833866714796e-05, + "loss": 0.964, + "step": 2893 + }, + { + "epoch": 0.21, + "learning_rate": 1.999833138879238e-05, + "loss": 1.0298, + "step": 2894 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998324094529606e-05, + "loss": 1.0464, + "step": 2895 + }, + { + "epoch": 0.21, + "learning_rate": 1.999831678435964e-05, + "loss": 1.099, + "step": 2896 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998309458282493e-05, + "loss": 0.9669, + "step": 2897 + }, + { + "epoch": 0.21, + "learning_rate": 1.999830211629818e-05, + "loss": 1.1341, + "step": 2898 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998294758406712e-05, + "loss": 1.0497, + "step": 2899 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998287384608096e-05, + "loss": 1.0937, + "step": 2900 + }, + { + "epoch": 0.21, + "learning_rate": 1.999827999490235e-05, + "loss": 0.9933, + "step": 2901 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998272589289486e-05, + "loss": 1.1459, + "step": 2902 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998265167769514e-05, + "loss": 1.1323, + "step": 2903 + }, + { + "epoch": 0.21, + "learning_rate": 1.999825773034244e-05, + "loss": 1.0212, + "step": 2904 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998250277008283e-05, + "loss": 1.0988, + "step": 2905 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998242807767054e-05, + "loss": 0.9456, + "step": 2906 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998235322618763e-05, + "loss": 1.0604, + "step": 2907 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998227821563424e-05, + "loss": 1.0665, + "step": 2908 + }, + { + "epoch": 0.21, + "learning_rate": 1.9998220304601046e-05, + "loss": 1.0175, + "step": 2909 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998212771731644e-05, + "loss": 1.091, + "step": 2910 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998205222955228e-05, + "loss": 1.0091, + "step": 2911 + }, + { + "epoch": 0.22, + "learning_rate": 1.999819765827181e-05, + "loss": 0.9996, + "step": 2912 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998190077681402e-05, + "loss": 1.0706, + "step": 2913 + }, + { + "epoch": 0.22, + "learning_rate": 1.999818248118402e-05, + "loss": 1.1167, + "step": 2914 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998174868779668e-05, + "loss": 1.0774, + "step": 2915 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998167240468365e-05, + "loss": 1.0197, + "step": 2916 + }, + { + "epoch": 0.22, + "learning_rate": 1.999815959625012e-05, + "loss": 1.1704, + "step": 2917 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998151936124947e-05, + "loss": 1.1604, + "step": 2918 + }, + { + "epoch": 0.22, + "learning_rate": 1.999814426009286e-05, + "loss": 1.0252, + "step": 2919 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998136568153864e-05, + "loss": 1.0628, + "step": 2920 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998128860307977e-05, + "loss": 1.0976, + "step": 2921 + }, + { + "epoch": 0.22, + "learning_rate": 1.999812113655521e-05, + "loss": 1.088, + "step": 2922 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998113396895576e-05, + "loss": 1.0982, + "step": 2923 + }, + { + "epoch": 0.22, + "learning_rate": 1.999810564132909e-05, + "loss": 1.1666, + "step": 2924 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998097869855755e-05, + "loss": 1.0217, + "step": 2925 + }, + { + "epoch": 0.22, + "learning_rate": 1.999809008247559e-05, + "loss": 1.117, + "step": 2926 + }, + { + "epoch": 0.22, + "learning_rate": 1.999808227918861e-05, + "loss": 1.0298, + "step": 2927 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998074459994825e-05, + "loss": 1.1556, + "step": 2928 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998066624894245e-05, + "loss": 1.0968, + "step": 2929 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998058773886884e-05, + "loss": 0.9764, + "step": 2930 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998050906972753e-05, + "loss": 1.0147, + "step": 2931 + }, + { + "epoch": 0.22, + "learning_rate": 1.999804302415187e-05, + "loss": 0.9498, + "step": 2932 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998035125424237e-05, + "loss": 1.091, + "step": 2933 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998027210789876e-05, + "loss": 0.9182, + "step": 2934 + }, + { + "epoch": 0.22, + "learning_rate": 1.99980192802488e-05, + "loss": 1.0987, + "step": 2935 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998011333801015e-05, + "loss": 1.0213, + "step": 2936 + }, + { + "epoch": 0.22, + "learning_rate": 1.9998003371446538e-05, + "loss": 1.0333, + "step": 2937 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997995393185383e-05, + "loss": 1.1178, + "step": 2938 + }, + { + "epoch": 0.22, + "learning_rate": 1.999798739901756e-05, + "loss": 1.0886, + "step": 2939 + }, + { + "epoch": 0.22, + "learning_rate": 1.999797938894308e-05, + "loss": 1.0346, + "step": 2940 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997971362961954e-05, + "loss": 0.9894, + "step": 2941 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997963321074204e-05, + "loss": 1.1119, + "step": 2942 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997955263279837e-05, + "loss": 1.0836, + "step": 2943 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997947189578864e-05, + "loss": 1.1316, + "step": 2944 + }, + { + "epoch": 0.22, + "learning_rate": 1.99979390999713e-05, + "loss": 1.0813, + "step": 2945 + }, + { + "epoch": 0.22, + "learning_rate": 1.999793099445716e-05, + "loss": 1.1147, + "step": 2946 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997922873036456e-05, + "loss": 1.0781, + "step": 2947 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997914735709195e-05, + "loss": 1.0444, + "step": 2948 + }, + { + "epoch": 0.22, + "learning_rate": 1.99979065824754e-05, + "loss": 1.121, + "step": 2949 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997898413335075e-05, + "loss": 1.041, + "step": 2950 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997890228288237e-05, + "loss": 1.0086, + "step": 2951 + }, + { + "epoch": 0.22, + "learning_rate": 1.99978820273349e-05, + "loss": 1.0676, + "step": 2952 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997873810475076e-05, + "loss": 1.0896, + "step": 2953 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997865577708777e-05, + "loss": 1.0111, + "step": 2954 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997857329036016e-05, + "loss": 1.174, + "step": 2955 + }, + { + "epoch": 0.22, + "learning_rate": 1.999784906445681e-05, + "loss": 0.9653, + "step": 2956 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997840783971164e-05, + "loss": 1.129, + "step": 2957 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997832487579105e-05, + "loss": 1.0333, + "step": 2958 + }, + { + "epoch": 0.22, + "learning_rate": 1.999782417528063e-05, + "loss": 1.0505, + "step": 2959 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997815847075764e-05, + "loss": 1.1465, + "step": 2960 + }, + { + "epoch": 0.22, + "learning_rate": 1.999780750296451e-05, + "loss": 1.0728, + "step": 2961 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997799142946894e-05, + "loss": 1.0446, + "step": 2962 + }, + { + "epoch": 0.22, + "learning_rate": 1.999779076702292e-05, + "loss": 1.0444, + "step": 2963 + }, + { + "epoch": 0.22, + "learning_rate": 1.99977823751926e-05, + "loss": 1.0794, + "step": 2964 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997773967455956e-05, + "loss": 1.0622, + "step": 2965 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997765543813e-05, + "loss": 1.0825, + "step": 2966 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997757104263734e-05, + "loss": 1.0763, + "step": 2967 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997748648808187e-05, + "loss": 1.0337, + "step": 2968 + }, + { + "epoch": 0.22, + "learning_rate": 1.999774017744636e-05, + "loss": 0.9852, + "step": 2969 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997731690178273e-05, + "loss": 1.0086, + "step": 2970 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997723187003936e-05, + "loss": 1.0623, + "step": 2971 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997714667923367e-05, + "loss": 1.0327, + "step": 2972 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997706132936576e-05, + "loss": 1.207, + "step": 2973 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997697582043577e-05, + "loss": 1.0766, + "step": 2974 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997689015244387e-05, + "loss": 1.1287, + "step": 2975 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997680432539016e-05, + "loss": 1.0504, + "step": 2976 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997671833927475e-05, + "loss": 1.0329, + "step": 2977 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997663219409782e-05, + "loss": 1.0232, + "step": 2978 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997654588985953e-05, + "loss": 1.0781, + "step": 2979 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997645942655997e-05, + "loss": 1.0094, + "step": 2980 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997637280419928e-05, + "loss": 1.0207, + "step": 2981 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997628602277762e-05, + "loss": 1.122, + "step": 2982 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997619908229514e-05, + "loss": 1.1635, + "step": 2983 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997611198275193e-05, + "loss": 1.0989, + "step": 2984 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997602472414817e-05, + "loss": 1.0008, + "step": 2985 + }, + { + "epoch": 0.22, + "learning_rate": 1.99975937306484e-05, + "loss": 1.0037, + "step": 2986 + }, + { + "epoch": 0.22, + "learning_rate": 1.999758497297595e-05, + "loss": 1.1159, + "step": 2987 + }, + { + "epoch": 0.22, + "learning_rate": 1.999757619939749e-05, + "loss": 1.1591, + "step": 2988 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997567409913026e-05, + "loss": 1.1133, + "step": 2989 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997558604522574e-05, + "loss": 0.9697, + "step": 2990 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997549783226152e-05, + "loss": 0.9932, + "step": 2991 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997540946023772e-05, + "loss": 1.0603, + "step": 2992 + }, + { + "epoch": 0.22, + "learning_rate": 1.999753209291545e-05, + "loss": 1.0253, + "step": 2993 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997523223901192e-05, + "loss": 1.0543, + "step": 2994 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997514338981018e-05, + "loss": 1.1144, + "step": 2995 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997505438154944e-05, + "loss": 1.1313, + "step": 2996 + }, + { + "epoch": 0.22, + "learning_rate": 1.999749652142298e-05, + "loss": 1.0867, + "step": 2997 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997487588785145e-05, + "loss": 1.1684, + "step": 2998 + }, + { + "epoch": 0.22, + "learning_rate": 1.999747864024145e-05, + "loss": 1.0781, + "step": 2999 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997469675791906e-05, + "loss": 1.0193, + "step": 3000 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997460695436533e-05, + "loss": 1.0035, + "step": 3001 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997451699175342e-05, + "loss": 1.1145, + "step": 3002 + }, + { + "epoch": 0.22, + "learning_rate": 1.999744268700835e-05, + "loss": 1.0496, + "step": 3003 + }, + { + "epoch": 0.22, + "learning_rate": 1.999743365893557e-05, + "loss": 0.9657, + "step": 3004 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997424614957014e-05, + "loss": 1.0801, + "step": 3005 + }, + { + "epoch": 0.22, + "learning_rate": 1.99974155550727e-05, + "loss": 0.8738, + "step": 3006 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997406479282642e-05, + "loss": 1.0742, + "step": 3007 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997397387586848e-05, + "loss": 1.0001, + "step": 3008 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997388279985345e-05, + "loss": 1.0722, + "step": 3009 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997379156478138e-05, + "loss": 1.0036, + "step": 3010 + }, + { + "epoch": 0.22, + "learning_rate": 1.999737001706524e-05, + "loss": 1.0508, + "step": 3011 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997360861746672e-05, + "loss": 1.0521, + "step": 3012 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997351690522447e-05, + "loss": 1.03, + "step": 3013 + }, + { + "epoch": 0.22, + "learning_rate": 1.999734250339258e-05, + "loss": 1.1455, + "step": 3014 + }, + { + "epoch": 0.22, + "learning_rate": 1.999733330035708e-05, + "loss": 1.1627, + "step": 3015 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997324081415968e-05, + "loss": 1.1221, + "step": 3016 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997314846569255e-05, + "loss": 0.9678, + "step": 3017 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997305595816957e-05, + "loss": 1.0251, + "step": 3018 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997296329159092e-05, + "loss": 1.0736, + "step": 3019 + }, + { + "epoch": 0.22, + "learning_rate": 1.999728704659567e-05, + "loss": 1.0079, + "step": 3020 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997277748126705e-05, + "loss": 1.1092, + "step": 3021 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997268433752214e-05, + "loss": 1.0243, + "step": 3022 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997259103472214e-05, + "loss": 1.0083, + "step": 3023 + }, + { + "epoch": 0.22, + "learning_rate": 1.999724975728672e-05, + "loss": 1.0736, + "step": 3024 + }, + { + "epoch": 0.22, + "learning_rate": 1.999724039519574e-05, + "loss": 1.1707, + "step": 3025 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997231017199296e-05, + "loss": 0.9797, + "step": 3026 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997221623297403e-05, + "loss": 1.092, + "step": 3027 + }, + { + "epoch": 0.22, + "learning_rate": 1.999721221349007e-05, + "loss": 1.0522, + "step": 3028 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997202787777315e-05, + "loss": 0.9877, + "step": 3029 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997193346159153e-05, + "loss": 1.0571, + "step": 3030 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997183888635603e-05, + "loss": 1.0432, + "step": 3031 + }, + { + "epoch": 0.22, + "learning_rate": 1.999717441520667e-05, + "loss": 1.027, + "step": 3032 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997164925872386e-05, + "loss": 1.0901, + "step": 3033 + }, + { + "epoch": 0.22, + "learning_rate": 1.999715542063275e-05, + "loss": 1.0079, + "step": 3034 + }, + { + "epoch": 0.22, + "learning_rate": 1.999714589948778e-05, + "loss": 0.9416, + "step": 3035 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997136362437498e-05, + "loss": 0.9929, + "step": 3036 + }, + { + "epoch": 0.22, + "learning_rate": 1.999712680948191e-05, + "loss": 1.0709, + "step": 3037 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997117240621042e-05, + "loss": 0.9624, + "step": 3038 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997107655854902e-05, + "loss": 0.9697, + "step": 3039 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997098055183505e-05, + "loss": 0.9881, + "step": 3040 + }, + { + "epoch": 0.22, + "learning_rate": 1.999708843860687e-05, + "loss": 1.1462, + "step": 3041 + }, + { + "epoch": 0.22, + "learning_rate": 1.999707880612501e-05, + "loss": 1.0844, + "step": 3042 + }, + { + "epoch": 0.22, + "learning_rate": 1.9997069157737935e-05, + "loss": 1.0295, + "step": 3043 + }, + { + "epoch": 0.22, + "learning_rate": 1.999705949344567e-05, + "loss": 1.1646, + "step": 3044 + }, + { + "epoch": 0.22, + "learning_rate": 1.999704981324823e-05, + "loss": 1.0591, + "step": 3045 + }, + { + "epoch": 0.23, + "learning_rate": 1.9997040117145627e-05, + "loss": 0.9262, + "step": 3046 + }, + { + "epoch": 0.23, + "learning_rate": 1.9997030405137872e-05, + "loss": 0.9395, + "step": 3047 + }, + { + "epoch": 0.23, + "learning_rate": 1.9997020677224986e-05, + "loss": 1.1042, + "step": 3048 + }, + { + "epoch": 0.23, + "learning_rate": 1.9997010933406983e-05, + "loss": 1.0506, + "step": 3049 + }, + { + "epoch": 0.23, + "learning_rate": 1.999700117368388e-05, + "loss": 1.0873, + "step": 3050 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996991398055687e-05, + "loss": 1.1301, + "step": 3051 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996981606522428e-05, + "loss": 1.1617, + "step": 3052 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996971799084112e-05, + "loss": 1.1594, + "step": 3053 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996961975740757e-05, + "loss": 1.144, + "step": 3054 + }, + { + "epoch": 0.23, + "learning_rate": 1.999695213649238e-05, + "loss": 1.0626, + "step": 3055 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996942281338995e-05, + "loss": 1.043, + "step": 3056 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996932410280617e-05, + "loss": 1.0834, + "step": 3057 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996922523317265e-05, + "loss": 1.0672, + "step": 3058 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996912620448948e-05, + "loss": 1.0373, + "step": 3059 + }, + { + "epoch": 0.23, + "learning_rate": 1.999690270167569e-05, + "loss": 1.1285, + "step": 3060 + }, + { + "epoch": 0.23, + "learning_rate": 1.99968927669975e-05, + "loss": 1.0125, + "step": 3061 + }, + { + "epoch": 0.23, + "learning_rate": 1.99968828164144e-05, + "loss": 1.0172, + "step": 3062 + }, + { + "epoch": 0.23, + "learning_rate": 1.99968728499264e-05, + "loss": 1.0578, + "step": 3063 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996862867533514e-05, + "loss": 1.0937, + "step": 3064 + }, + { + "epoch": 0.23, + "learning_rate": 1.999685286923577e-05, + "loss": 1.0421, + "step": 3065 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996842855033173e-05, + "loss": 1.1368, + "step": 3066 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996832824925742e-05, + "loss": 1.0843, + "step": 3067 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996822778913492e-05, + "loss": 1.0348, + "step": 3068 + }, + { + "epoch": 0.23, + "learning_rate": 1.999681271699644e-05, + "loss": 1.0534, + "step": 3069 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996802639174603e-05, + "loss": 1.1173, + "step": 3070 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996792545447998e-05, + "loss": 1.0823, + "step": 3071 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996782435816638e-05, + "loss": 1.0245, + "step": 3072 + }, + { + "epoch": 0.23, + "learning_rate": 1.999677231028054e-05, + "loss": 1.0601, + "step": 3073 + }, + { + "epoch": 0.23, + "learning_rate": 1.999676216883972e-05, + "loss": 1.0958, + "step": 3074 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996752011494193e-05, + "loss": 0.9345, + "step": 3075 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996741838243978e-05, + "loss": 1.0083, + "step": 3076 + }, + { + "epoch": 0.23, + "learning_rate": 1.999673164908909e-05, + "loss": 1.1022, + "step": 3077 + }, + { + "epoch": 0.23, + "learning_rate": 1.999672144402954e-05, + "loss": 1.0154, + "step": 3078 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996711223065355e-05, + "loss": 1.1036, + "step": 3079 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996700986196545e-05, + "loss": 1.0164, + "step": 3080 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996690733423122e-05, + "loss": 1.1718, + "step": 3081 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996680464745114e-05, + "loss": 1.0484, + "step": 3082 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996670180162528e-05, + "loss": 1.0573, + "step": 3083 + }, + { + "epoch": 0.23, + "learning_rate": 1.999665987967538e-05, + "loss": 1.077, + "step": 3084 + }, + { + "epoch": 0.23, + "learning_rate": 1.999664956328369e-05, + "loss": 1.1728, + "step": 3085 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996639230987478e-05, + "loss": 1.1075, + "step": 3086 + }, + { + "epoch": 0.23, + "learning_rate": 1.999662888278675e-05, + "loss": 1.0335, + "step": 3087 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996618518681533e-05, + "loss": 1.055, + "step": 3088 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996608138671836e-05, + "loss": 1.0247, + "step": 3089 + }, + { + "epoch": 0.23, + "learning_rate": 1.999659774275768e-05, + "loss": 1.0515, + "step": 3090 + }, + { + "epoch": 0.23, + "learning_rate": 1.999658733093908e-05, + "loss": 1.0183, + "step": 3091 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996576903216052e-05, + "loss": 1.1715, + "step": 3092 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996566459588613e-05, + "loss": 1.0918, + "step": 3093 + }, + { + "epoch": 0.23, + "learning_rate": 1.999655600005678e-05, + "loss": 1.0, + "step": 3094 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996545524620573e-05, + "loss": 1.0602, + "step": 3095 + }, + { + "epoch": 0.23, + "learning_rate": 1.999653503328e-05, + "loss": 1.0746, + "step": 3096 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996524526035083e-05, + "loss": 1.1087, + "step": 3097 + }, + { + "epoch": 0.23, + "learning_rate": 1.999651400288584e-05, + "loss": 1.0483, + "step": 3098 + }, + { + "epoch": 0.23, + "learning_rate": 1.999650346383229e-05, + "loss": 0.9585, + "step": 3099 + }, + { + "epoch": 0.23, + "learning_rate": 1.999649290887444e-05, + "loss": 1.0444, + "step": 3100 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996482338012314e-05, + "loss": 1.1207, + "step": 3101 + }, + { + "epoch": 0.23, + "learning_rate": 1.999647175124593e-05, + "loss": 1.068, + "step": 3102 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996461148575302e-05, + "loss": 0.9702, + "step": 3103 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996450530000445e-05, + "loss": 1.0558, + "step": 3104 + }, + { + "epoch": 0.23, + "learning_rate": 1.999643989552138e-05, + "loss": 1.061, + "step": 3105 + }, + { + "epoch": 0.23, + "learning_rate": 1.999642924513812e-05, + "loss": 0.9564, + "step": 3106 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996418578850688e-05, + "loss": 1.1644, + "step": 3107 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996407896659096e-05, + "loss": 1.0862, + "step": 3108 + }, + { + "epoch": 0.23, + "learning_rate": 1.999639719856336e-05, + "loss": 1.0514, + "step": 3109 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996386484563502e-05, + "loss": 1.059, + "step": 3110 + }, + { + "epoch": 0.23, + "learning_rate": 1.999637575465953e-05, + "loss": 1.0606, + "step": 3111 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996365008851475e-05, + "loss": 1.0402, + "step": 3112 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996354247139343e-05, + "loss": 1.1205, + "step": 3113 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996343469523154e-05, + "loss": 1.016, + "step": 3114 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996332676002925e-05, + "loss": 1.0085, + "step": 3115 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996321866578674e-05, + "loss": 1.1294, + "step": 3116 + }, + { + "epoch": 0.23, + "learning_rate": 1.999631104125042e-05, + "loss": 1.0374, + "step": 3117 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996300200018174e-05, + "loss": 1.0499, + "step": 3118 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996289342881963e-05, + "loss": 1.0212, + "step": 3119 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996278469841793e-05, + "loss": 1.0784, + "step": 3120 + }, + { + "epoch": 0.23, + "learning_rate": 1.999626758089769e-05, + "loss": 1.0039, + "step": 3121 + }, + { + "epoch": 0.23, + "learning_rate": 1.999625667604967e-05, + "loss": 0.947, + "step": 3122 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996245755297744e-05, + "loss": 1.1426, + "step": 3123 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996234818641937e-05, + "loss": 0.959, + "step": 3124 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996223866082264e-05, + "loss": 1.0621, + "step": 3125 + }, + { + "epoch": 0.23, + "learning_rate": 1.999621289761874e-05, + "loss": 1.0424, + "step": 3126 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996201913251387e-05, + "loss": 1.1111, + "step": 3127 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996190912980218e-05, + "loss": 1.1196, + "step": 3128 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996179896805253e-05, + "loss": 1.0012, + "step": 3129 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996168864726506e-05, + "loss": 1.1074, + "step": 3130 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996157816744e-05, + "loss": 1.0919, + "step": 3131 + }, + { + "epoch": 0.23, + "learning_rate": 1.999614675285775e-05, + "loss": 1.1772, + "step": 3132 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996135673067774e-05, + "loss": 1.0868, + "step": 3133 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996124577374088e-05, + "loss": 1.0458, + "step": 3134 + }, + { + "epoch": 0.23, + "learning_rate": 1.999611346577671e-05, + "loss": 1.0605, + "step": 3135 + }, + { + "epoch": 0.23, + "learning_rate": 1.999610233827566e-05, + "loss": 1.0696, + "step": 3136 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996091194870953e-05, + "loss": 1.0963, + "step": 3137 + }, + { + "epoch": 0.23, + "learning_rate": 1.999608003556261e-05, + "loss": 1.0774, + "step": 3138 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996068860350645e-05, + "loss": 1.1138, + "step": 3139 + }, + { + "epoch": 0.23, + "learning_rate": 1.999605766923508e-05, + "loss": 1.3153, + "step": 3140 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996046462215927e-05, + "loss": 1.0645, + "step": 3141 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996035239293206e-05, + "loss": 1.1578, + "step": 3142 + }, + { + "epoch": 0.23, + "learning_rate": 1.9996024000466937e-05, + "loss": 1.0416, + "step": 3143 + }, + { + "epoch": 0.23, + "learning_rate": 1.999601274573714e-05, + "loss": 1.1175, + "step": 3144 + }, + { + "epoch": 0.23, + "learning_rate": 1.999600147510383e-05, + "loss": 1.0873, + "step": 3145 + }, + { + "epoch": 0.23, + "learning_rate": 1.999599018856702e-05, + "loss": 1.0324, + "step": 3146 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995978886126736e-05, + "loss": 1.1105, + "step": 3147 + }, + { + "epoch": 0.23, + "learning_rate": 1.999596756778299e-05, + "loss": 1.1363, + "step": 3148 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995956233535802e-05, + "loss": 1.0917, + "step": 3149 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995944883385196e-05, + "loss": 1.1467, + "step": 3150 + }, + { + "epoch": 0.23, + "learning_rate": 1.999593351733118e-05, + "loss": 1.0101, + "step": 3151 + }, + { + "epoch": 0.23, + "learning_rate": 1.999592213537378e-05, + "loss": 1.0054, + "step": 3152 + }, + { + "epoch": 0.23, + "learning_rate": 1.999591073751301e-05, + "loss": 1.0395, + "step": 3153 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995899323748888e-05, + "loss": 1.0426, + "step": 3154 + }, + { + "epoch": 0.23, + "learning_rate": 1.999588789408143e-05, + "loss": 1.034, + "step": 3155 + }, + { + "epoch": 0.23, + "learning_rate": 1.999587644851066e-05, + "loss": 1.0278, + "step": 3156 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995864987036598e-05, + "loss": 1.0194, + "step": 3157 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995853509659254e-05, + "loss": 1.0914, + "step": 3158 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995842016378647e-05, + "loss": 1.1447, + "step": 3159 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995830507194806e-05, + "loss": 1.0173, + "step": 3160 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995818982107735e-05, + "loss": 1.1189, + "step": 3161 + }, + { + "epoch": 0.23, + "learning_rate": 1.999580744111746e-05, + "loss": 0.9566, + "step": 3162 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995795884224e-05, + "loss": 1.0846, + "step": 3163 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995784311427373e-05, + "loss": 1.0826, + "step": 3164 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995772722727595e-05, + "loss": 1.1791, + "step": 3165 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995761118124686e-05, + "loss": 1.075, + "step": 3166 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995749497618663e-05, + "loss": 0.9804, + "step": 3167 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995737861209545e-05, + "loss": 1.1173, + "step": 3168 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995726208897353e-05, + "loss": 1.1348, + "step": 3169 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995714540682103e-05, + "loss": 1.0407, + "step": 3170 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995702856563814e-05, + "loss": 0.9753, + "step": 3171 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995691156542505e-05, + "loss": 1.0056, + "step": 3172 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995679440618196e-05, + "loss": 1.0872, + "step": 3173 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995667708790898e-05, + "loss": 1.0202, + "step": 3174 + }, + { + "epoch": 0.23, + "learning_rate": 1.999565596106064e-05, + "loss": 1.0166, + "step": 3175 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995644197427438e-05, + "loss": 0.9972, + "step": 3176 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995632417891304e-05, + "loss": 1.0455, + "step": 3177 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995620622452265e-05, + "loss": 1.0777, + "step": 3178 + }, + { + "epoch": 0.23, + "learning_rate": 1.9995608811110338e-05, + "loss": 1.0638, + "step": 3179 + }, + { + "epoch": 0.23, + "learning_rate": 1.999559698386554e-05, + "loss": 1.1623, + "step": 3180 + }, + { + "epoch": 0.24, + "learning_rate": 1.999558514071789e-05, + "loss": 0.9727, + "step": 3181 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995573281667405e-05, + "loss": 0.97, + "step": 3182 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995561406714107e-05, + "loss": 1.0197, + "step": 3183 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995549515858013e-05, + "loss": 1.1415, + "step": 3184 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995537609099146e-05, + "loss": 1.0054, + "step": 3185 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995525686437517e-05, + "loss": 0.9544, + "step": 3186 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995513747873156e-05, + "loss": 1.1053, + "step": 3187 + }, + { + "epoch": 0.24, + "learning_rate": 1.999550179340607e-05, + "loss": 1.0417, + "step": 3188 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995489823036285e-05, + "loss": 0.9894, + "step": 3189 + }, + { + "epoch": 0.24, + "learning_rate": 1.999547783676382e-05, + "loss": 1.0873, + "step": 3190 + }, + { + "epoch": 0.24, + "learning_rate": 1.999546583458869e-05, + "loss": 0.9892, + "step": 3191 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995453816510918e-05, + "loss": 1.0783, + "step": 3192 + }, + { + "epoch": 0.24, + "learning_rate": 1.999544178253052e-05, + "loss": 1.0154, + "step": 3193 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995429732647525e-05, + "loss": 1.0082, + "step": 3194 + }, + { + "epoch": 0.24, + "learning_rate": 1.999541766686194e-05, + "loss": 1.1822, + "step": 3195 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995405585173788e-05, + "loss": 1.0728, + "step": 3196 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995393487583087e-05, + "loss": 0.9995, + "step": 3197 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995381374089858e-05, + "loss": 1.0052, + "step": 3198 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995369244694124e-05, + "loss": 1.0841, + "step": 3199 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995357099395895e-05, + "loss": 1.0363, + "step": 3200 + }, + { + "epoch": 0.24, + "learning_rate": 1.99953449381952e-05, + "loss": 1.1496, + "step": 3201 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995332761092055e-05, + "loss": 1.0685, + "step": 3202 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995320568086474e-05, + "loss": 1.0925, + "step": 3203 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995308359178484e-05, + "loss": 1.1834, + "step": 3204 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995296134368103e-05, + "loss": 1.0247, + "step": 3205 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995283893655347e-05, + "loss": 1.1087, + "step": 3206 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995271637040238e-05, + "loss": 1.0971, + "step": 3207 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995259364522793e-05, + "loss": 1.0434, + "step": 3208 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995247076103034e-05, + "loss": 1.0406, + "step": 3209 + }, + { + "epoch": 0.24, + "learning_rate": 1.999523477178098e-05, + "loss": 1.1049, + "step": 3210 + }, + { + "epoch": 0.24, + "learning_rate": 1.999522245155665e-05, + "loss": 1.1095, + "step": 3211 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995210115430067e-05, + "loss": 1.0749, + "step": 3212 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995197763401247e-05, + "loss": 1.083, + "step": 3213 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995185395470207e-05, + "loss": 1.1269, + "step": 3214 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995173011636973e-05, + "loss": 1.0964, + "step": 3215 + }, + { + "epoch": 0.24, + "learning_rate": 1.999516061190156e-05, + "loss": 0.9782, + "step": 3216 + }, + { + "epoch": 0.24, + "learning_rate": 1.999514819626399e-05, + "loss": 1.0815, + "step": 3217 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995135764724286e-05, + "loss": 0.9725, + "step": 3218 + }, + { + "epoch": 0.24, + "learning_rate": 1.999512331728246e-05, + "loss": 1.0747, + "step": 3219 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995110853938534e-05, + "loss": 1.051, + "step": 3220 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995098374692535e-05, + "loss": 1.0372, + "step": 3221 + }, + { + "epoch": 0.24, + "learning_rate": 1.999508587954447e-05, + "loss": 1.0964, + "step": 3222 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995073368494373e-05, + "loss": 1.0282, + "step": 3223 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995060841542258e-05, + "loss": 1.1402, + "step": 3224 + }, + { + "epoch": 0.24, + "learning_rate": 1.999504829868814e-05, + "loss": 1.0791, + "step": 3225 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995035739932045e-05, + "loss": 1.0756, + "step": 3226 + }, + { + "epoch": 0.24, + "learning_rate": 1.999502316527399e-05, + "loss": 1.0469, + "step": 3227 + }, + { + "epoch": 0.24, + "learning_rate": 1.9995010574713998e-05, + "loss": 1.1049, + "step": 3228 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994997968252082e-05, + "loss": 1.0706, + "step": 3229 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994985345888272e-05, + "loss": 1.1779, + "step": 3230 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994972707622584e-05, + "loss": 1.1001, + "step": 3231 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994960053455038e-05, + "loss": 1.123, + "step": 3232 + }, + { + "epoch": 0.24, + "learning_rate": 1.999494738338565e-05, + "loss": 0.9703, + "step": 3233 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994934697414447e-05, + "loss": 1.0524, + "step": 3234 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994921995541443e-05, + "loss": 1.0103, + "step": 3235 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994909277766663e-05, + "loss": 1.0361, + "step": 3236 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994896544090126e-05, + "loss": 1.1054, + "step": 3237 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994883794511847e-05, + "loss": 0.9837, + "step": 3238 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994871029031856e-05, + "loss": 1.0578, + "step": 3239 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994858247650166e-05, + "loss": 0.9247, + "step": 3240 + }, + { + "epoch": 0.24, + "learning_rate": 1.99948454503668e-05, + "loss": 1.0593, + "step": 3241 + }, + { + "epoch": 0.24, + "learning_rate": 1.999483263718178e-05, + "loss": 1.1501, + "step": 3242 + }, + { + "epoch": 0.24, + "learning_rate": 1.999481980809512e-05, + "loss": 1.0977, + "step": 3243 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994806963106846e-05, + "loss": 1.1004, + "step": 3244 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994794102216976e-05, + "loss": 1.1595, + "step": 3245 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994781225425535e-05, + "loss": 1.0515, + "step": 3246 + }, + { + "epoch": 0.24, + "learning_rate": 1.999476833273254e-05, + "loss": 1.0174, + "step": 3247 + }, + { + "epoch": 0.24, + "learning_rate": 1.999475542413801e-05, + "loss": 1.064, + "step": 3248 + }, + { + "epoch": 0.24, + "learning_rate": 1.999474249964197e-05, + "loss": 1.0914, + "step": 3249 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994729559244433e-05, + "loss": 1.1017, + "step": 3250 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994716602945425e-05, + "loss": 0.9194, + "step": 3251 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994703630744966e-05, + "loss": 1.0235, + "step": 3252 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994690642643077e-05, + "loss": 1.1033, + "step": 3253 + }, + { + "epoch": 0.24, + "learning_rate": 1.999467763863978e-05, + "loss": 1.2116, + "step": 3254 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994664618735093e-05, + "loss": 1.0038, + "step": 3255 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994651582929036e-05, + "loss": 1.0212, + "step": 3256 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994638531221633e-05, + "loss": 1.0584, + "step": 3257 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994625463612904e-05, + "loss": 1.141, + "step": 3258 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994612380102867e-05, + "loss": 1.0568, + "step": 3259 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994599280691544e-05, + "loss": 0.8981, + "step": 3260 + }, + { + "epoch": 0.24, + "learning_rate": 1.999458616537896e-05, + "loss": 1.0829, + "step": 3261 + }, + { + "epoch": 0.24, + "learning_rate": 1.999457303416513e-05, + "loss": 1.0833, + "step": 3262 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994559887050075e-05, + "loss": 1.0672, + "step": 3263 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994546724033825e-05, + "loss": 0.9626, + "step": 3264 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994533545116388e-05, + "loss": 1.1299, + "step": 3265 + }, + { + "epoch": 0.24, + "learning_rate": 1.999452035029779e-05, + "loss": 1.0478, + "step": 3266 + }, + { + "epoch": 0.24, + "learning_rate": 1.999450713957806e-05, + "loss": 1.1183, + "step": 3267 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994493912957207e-05, + "loss": 1.0983, + "step": 3268 + }, + { + "epoch": 0.24, + "learning_rate": 1.999448067043526e-05, + "loss": 1.0393, + "step": 3269 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994467412012234e-05, + "loss": 1.1331, + "step": 3270 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994454137688154e-05, + "loss": 1.0689, + "step": 3271 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994440847463042e-05, + "loss": 0.9954, + "step": 3272 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994427541336917e-05, + "loss": 0.9818, + "step": 3273 + }, + { + "epoch": 0.24, + "learning_rate": 1.99944142193098e-05, + "loss": 1.0058, + "step": 3274 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994400881381715e-05, + "loss": 1.1855, + "step": 3275 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994387527552683e-05, + "loss": 0.9673, + "step": 3276 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994374157822718e-05, + "loss": 1.0008, + "step": 3277 + }, + { + "epoch": 0.24, + "learning_rate": 1.999436077219185e-05, + "loss": 1.0601, + "step": 3278 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994347370660098e-05, + "loss": 1.0736, + "step": 3279 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994333953227477e-05, + "loss": 1.0449, + "step": 3280 + }, + { + "epoch": 0.24, + "learning_rate": 1.999432051989402e-05, + "loss": 1.0705, + "step": 3281 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994307070659738e-05, + "loss": 1.1617, + "step": 3282 + }, + { + "epoch": 0.24, + "learning_rate": 1.999429360552466e-05, + "loss": 1.1035, + "step": 3283 + }, + { + "epoch": 0.24, + "learning_rate": 1.99942801244888e-05, + "loss": 0.9848, + "step": 3284 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994266627552185e-05, + "loss": 1.0939, + "step": 3285 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994253114714835e-05, + "loss": 1.0387, + "step": 3286 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994239585976772e-05, + "loss": 1.0667, + "step": 3287 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994226041338013e-05, + "loss": 1.0959, + "step": 3288 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994212480798586e-05, + "loss": 1.0077, + "step": 3289 + }, + { + "epoch": 0.24, + "learning_rate": 1.999419890435851e-05, + "loss": 1.1365, + "step": 3290 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994185312017808e-05, + "loss": 1.1031, + "step": 3291 + }, + { + "epoch": 0.24, + "learning_rate": 1.99941717037765e-05, + "loss": 0.93, + "step": 3292 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994158079634606e-05, + "loss": 1.1885, + "step": 3293 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994144439592152e-05, + "loss": 1.1106, + "step": 3294 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994130783649154e-05, + "loss": 1.0188, + "step": 3295 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994117111805637e-05, + "loss": 1.107, + "step": 3296 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994103424061624e-05, + "loss": 0.9899, + "step": 3297 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994089720417136e-05, + "loss": 1.0729, + "step": 3298 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994076000872193e-05, + "loss": 1.1123, + "step": 3299 + }, + { + "epoch": 0.24, + "learning_rate": 1.999406226542682e-05, + "loss": 1.0321, + "step": 3300 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994048514081036e-05, + "loss": 1.0369, + "step": 3301 + }, + { + "epoch": 0.24, + "learning_rate": 1.999403474683486e-05, + "loss": 1.1183, + "step": 3302 + }, + { + "epoch": 0.24, + "learning_rate": 1.9994020963688324e-05, + "loss": 1.0398, + "step": 3303 + }, + { + "epoch": 0.24, + "learning_rate": 1.999400716464144e-05, + "loss": 0.9497, + "step": 3304 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993993349694235e-05, + "loss": 1.1233, + "step": 3305 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993979518846726e-05, + "loss": 0.975, + "step": 3306 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993965672098942e-05, + "loss": 1.0133, + "step": 3307 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993951809450903e-05, + "loss": 1.1421, + "step": 3308 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993937930902626e-05, + "loss": 1.059, + "step": 3309 + }, + { + "epoch": 0.24, + "learning_rate": 1.999392403645414e-05, + "loss": 1.104, + "step": 3310 + }, + { + "epoch": 0.24, + "learning_rate": 1.999391012610546e-05, + "loss": 1.0258, + "step": 3311 + }, + { + "epoch": 0.24, + "learning_rate": 1.999389619985662e-05, + "loss": 1.1536, + "step": 3312 + }, + { + "epoch": 0.24, + "learning_rate": 1.9993882257707625e-05, + "loss": 1.1091, + "step": 3313 + }, + { + "epoch": 0.24, + "learning_rate": 1.999386829965851e-05, + "loss": 1.0283, + "step": 3314 + }, + { + "epoch": 0.24, + "learning_rate": 1.999385432570929e-05, + "loss": 1.1044, + "step": 3315 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993840335859996e-05, + "loss": 1.1134, + "step": 3316 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993826330110643e-05, + "loss": 1.0344, + "step": 3317 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993812308461255e-05, + "loss": 1.0713, + "step": 3318 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993798270911856e-05, + "loss": 1.0775, + "step": 3319 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993784217462466e-05, + "loss": 1.0082, + "step": 3320 + }, + { + "epoch": 0.25, + "learning_rate": 1.999377014811311e-05, + "loss": 1.0315, + "step": 3321 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993756062863805e-05, + "loss": 1.054, + "step": 3322 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993741961714578e-05, + "loss": 1.0937, + "step": 3323 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993727844665455e-05, + "loss": 1.1158, + "step": 3324 + }, + { + "epoch": 0.25, + "learning_rate": 1.999371371171645e-05, + "loss": 0.9677, + "step": 3325 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993699562867592e-05, + "loss": 0.9842, + "step": 3326 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993685398118896e-05, + "loss": 1.036, + "step": 3327 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993671217470392e-05, + "loss": 0.9968, + "step": 3328 + }, + { + "epoch": 0.25, + "learning_rate": 1.99936570209221e-05, + "loss": 1.0952, + "step": 3329 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993642808474045e-05, + "loss": 0.9733, + "step": 3330 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993628580126243e-05, + "loss": 1.0435, + "step": 3331 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993614335878724e-05, + "loss": 1.0894, + "step": 3332 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993600075731506e-05, + "loss": 1.1644, + "step": 3333 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993585799684615e-05, + "loss": 1.0094, + "step": 3334 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993571507738073e-05, + "loss": 0.9634, + "step": 3335 + }, + { + "epoch": 0.25, + "learning_rate": 1.99935571998919e-05, + "loss": 1.0179, + "step": 3336 + }, + { + "epoch": 0.25, + "learning_rate": 1.999354287614612e-05, + "loss": 1.033, + "step": 3337 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993528536500756e-05, + "loss": 0.9781, + "step": 3338 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993514180955832e-05, + "loss": 1.0479, + "step": 3339 + }, + { + "epoch": 0.25, + "learning_rate": 1.999349980951137e-05, + "loss": 0.9245, + "step": 3340 + }, + { + "epoch": 0.25, + "learning_rate": 1.999348542216739e-05, + "loss": 1.0252, + "step": 3341 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993471018923924e-05, + "loss": 1.1112, + "step": 3342 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993456599780985e-05, + "loss": 1.0335, + "step": 3343 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993442164738597e-05, + "loss": 1.1781, + "step": 3344 + }, + { + "epoch": 0.25, + "learning_rate": 1.999342771379679e-05, + "loss": 1.0804, + "step": 3345 + }, + { + "epoch": 0.25, + "learning_rate": 1.999341324695558e-05, + "loss": 1.0306, + "step": 3346 + }, + { + "epoch": 0.25, + "learning_rate": 1.999339876421499e-05, + "loss": 1.0362, + "step": 3347 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993384265575048e-05, + "loss": 1.1241, + "step": 3348 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993369751035775e-05, + "loss": 1.0172, + "step": 3349 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993355220597196e-05, + "loss": 1.1547, + "step": 3350 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993340674259327e-05, + "loss": 1.1381, + "step": 3351 + }, + { + "epoch": 0.25, + "learning_rate": 1.99933261120222e-05, + "loss": 1.181, + "step": 3352 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993311533885833e-05, + "loss": 1.1288, + "step": 3353 + }, + { + "epoch": 0.25, + "learning_rate": 1.999329693985025e-05, + "loss": 1.0786, + "step": 3354 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993282329915472e-05, + "loss": 1.0236, + "step": 3355 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993267704081528e-05, + "loss": 1.0096, + "step": 3356 + }, + { + "epoch": 0.25, + "learning_rate": 1.999325306234844e-05, + "loss": 1.1063, + "step": 3357 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993238404716226e-05, + "loss": 1.1012, + "step": 3358 + }, + { + "epoch": 0.25, + "learning_rate": 1.999322373118491e-05, + "loss": 1.1356, + "step": 3359 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993209041754525e-05, + "loss": 1.0681, + "step": 3360 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993194336425082e-05, + "loss": 1.0573, + "step": 3361 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993179615196614e-05, + "loss": 1.1054, + "step": 3362 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993164878069134e-05, + "loss": 1.0296, + "step": 3363 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993150125042677e-05, + "loss": 1.009, + "step": 3364 + }, + { + "epoch": 0.25, + "learning_rate": 1.999313535611726e-05, + "loss": 1.0785, + "step": 3365 + }, + { + "epoch": 0.25, + "learning_rate": 1.999312057129291e-05, + "loss": 1.0155, + "step": 3366 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993105770569644e-05, + "loss": 1.1471, + "step": 3367 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993090953947488e-05, + "loss": 1.0165, + "step": 3368 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993076121426474e-05, + "loss": 1.0242, + "step": 3369 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993061273006613e-05, + "loss": 1.0139, + "step": 3370 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993046408687938e-05, + "loss": 1.1283, + "step": 3371 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993031528470467e-05, + "loss": 1.0041, + "step": 3372 + }, + { + "epoch": 0.25, + "learning_rate": 1.9993016632354226e-05, + "loss": 1.0197, + "step": 3373 + }, + { + "epoch": 0.25, + "learning_rate": 1.999300172033924e-05, + "loss": 0.976, + "step": 3374 + }, + { + "epoch": 0.25, + "learning_rate": 1.999298679242553e-05, + "loss": 1.071, + "step": 3375 + }, + { + "epoch": 0.25, + "learning_rate": 1.999297184861312e-05, + "loss": 1.0104, + "step": 3376 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992956888902036e-05, + "loss": 1.1666, + "step": 3377 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992941913292305e-05, + "loss": 1.1526, + "step": 3378 + }, + { + "epoch": 0.25, + "learning_rate": 1.999292692178394e-05, + "loss": 1.09, + "step": 3379 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992911914376972e-05, + "loss": 1.1087, + "step": 3380 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992896891071426e-05, + "loss": 1.0328, + "step": 3381 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992881851867324e-05, + "loss": 1.1084, + "step": 3382 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992866796764686e-05, + "loss": 1.1236, + "step": 3383 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992851725763547e-05, + "loss": 1.1286, + "step": 3384 + }, + { + "epoch": 0.25, + "learning_rate": 1.999283663886392e-05, + "loss": 1.1568, + "step": 3385 + }, + { + "epoch": 0.25, + "learning_rate": 1.999282153606583e-05, + "loss": 1.0752, + "step": 3386 + }, + { + "epoch": 0.25, + "learning_rate": 1.999280641736931e-05, + "loss": 1.1026, + "step": 3387 + }, + { + "epoch": 0.25, + "learning_rate": 1.999279128277437e-05, + "loss": 1.0175, + "step": 3388 + }, + { + "epoch": 0.25, + "learning_rate": 1.999277613228105e-05, + "loss": 0.9677, + "step": 3389 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992760965889363e-05, + "loss": 1.1454, + "step": 3390 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992745783599337e-05, + "loss": 1.0896, + "step": 3391 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992730585410997e-05, + "loss": 1.1781, + "step": 3392 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992715371324364e-05, + "loss": 1.0949, + "step": 3393 + }, + { + "epoch": 0.25, + "learning_rate": 1.999270014133946e-05, + "loss": 1.0172, + "step": 3394 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992684895456315e-05, + "loss": 1.1274, + "step": 3395 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992669633674954e-05, + "loss": 0.9783, + "step": 3396 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992654355995398e-05, + "loss": 0.9927, + "step": 3397 + }, + { + "epoch": 0.25, + "learning_rate": 1.999263906241767e-05, + "loss": 1.0618, + "step": 3398 + }, + { + "epoch": 0.25, + "learning_rate": 1.99926237529418e-05, + "loss": 1.0073, + "step": 3399 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992608427567806e-05, + "loss": 1.1951, + "step": 3400 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992593086295713e-05, + "loss": 1.0629, + "step": 3401 + }, + { + "epoch": 0.25, + "learning_rate": 1.999257772912555e-05, + "loss": 1.028, + "step": 3402 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992562356057338e-05, + "loss": 1.0428, + "step": 3403 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992546967091104e-05, + "loss": 0.9873, + "step": 3404 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992531562226867e-05, + "loss": 1.0303, + "step": 3405 + }, + { + "epoch": 0.25, + "learning_rate": 1.999251614146466e-05, + "loss": 1.0237, + "step": 3406 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992500704804496e-05, + "loss": 1.0929, + "step": 3407 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992485252246413e-05, + "loss": 0.9287, + "step": 3408 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992469783790426e-05, + "loss": 1.048, + "step": 3409 + }, + { + "epoch": 0.25, + "learning_rate": 1.999245429943656e-05, + "loss": 1.0418, + "step": 3410 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992438799184848e-05, + "loss": 1.0661, + "step": 3411 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992423283035305e-05, + "loss": 1.1795, + "step": 3412 + }, + { + "epoch": 0.25, + "learning_rate": 1.999240775098796e-05, + "loss": 1.0793, + "step": 3413 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992392203042837e-05, + "loss": 1.0709, + "step": 3414 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992376639199966e-05, + "loss": 1.1002, + "step": 3415 + }, + { + "epoch": 0.25, + "learning_rate": 1.999236105945936e-05, + "loss": 1.1077, + "step": 3416 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992345463821054e-05, + "loss": 0.92, + "step": 3417 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992329852285068e-05, + "loss": 1.1295, + "step": 3418 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992314224851425e-05, + "loss": 1.0597, + "step": 3419 + }, + { + "epoch": 0.25, + "learning_rate": 1.999229858152016e-05, + "loss": 1.073, + "step": 3420 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992282922291285e-05, + "loss": 1.0223, + "step": 3421 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992267247164835e-05, + "loss": 0.9427, + "step": 3422 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992251556140828e-05, + "loss": 1.1467, + "step": 3423 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992235849219292e-05, + "loss": 1.1362, + "step": 3424 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992220126400253e-05, + "loss": 1.0327, + "step": 3425 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992204387683737e-05, + "loss": 1.0066, + "step": 3426 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992188633069764e-05, + "loss": 1.0143, + "step": 3427 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992172862558358e-05, + "loss": 1.0439, + "step": 3428 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992157076149553e-05, + "loss": 1.012, + "step": 3429 + }, + { + "epoch": 0.25, + "learning_rate": 1.999214127384337e-05, + "loss": 1.1074, + "step": 3430 + }, + { + "epoch": 0.25, + "learning_rate": 1.999212545563983e-05, + "loss": 1.1529, + "step": 3431 + }, + { + "epoch": 0.25, + "learning_rate": 1.999210962153896e-05, + "loss": 0.925, + "step": 3432 + }, + { + "epoch": 0.25, + "learning_rate": 1.999209377154079e-05, + "loss": 1.0749, + "step": 3433 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992077905645342e-05, + "loss": 0.9355, + "step": 3434 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992062023852638e-05, + "loss": 0.9722, + "step": 3435 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992046126162708e-05, + "loss": 1.0227, + "step": 3436 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992030212575576e-05, + "loss": 1.0758, + "step": 3437 + }, + { + "epoch": 0.25, + "learning_rate": 1.9992014283091268e-05, + "loss": 0.9452, + "step": 3438 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991998337709802e-05, + "loss": 1.1087, + "step": 3439 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991982376431215e-05, + "loss": 0.9229, + "step": 3440 + }, + { + "epoch": 0.25, + "learning_rate": 1.999196639925553e-05, + "loss": 1.0556, + "step": 3441 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991950406182762e-05, + "loss": 1.1112, + "step": 3442 + }, + { + "epoch": 0.25, + "learning_rate": 1.999193439721295e-05, + "loss": 1.0974, + "step": 3443 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991918372346112e-05, + "loss": 1.0744, + "step": 3444 + }, + { + "epoch": 0.25, + "learning_rate": 1.999190233158227e-05, + "loss": 1.053, + "step": 3445 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991886274921463e-05, + "loss": 1.0097, + "step": 3446 + }, + { + "epoch": 0.25, + "learning_rate": 1.99918702023637e-05, + "loss": 0.9405, + "step": 3447 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991854113909016e-05, + "loss": 0.9696, + "step": 3448 + }, + { + "epoch": 0.25, + "learning_rate": 1.999183800955744e-05, + "loss": 1.0549, + "step": 3449 + }, + { + "epoch": 0.25, + "learning_rate": 1.999182188930899e-05, + "loss": 1.012, + "step": 3450 + }, + { + "epoch": 0.25, + "learning_rate": 1.9991805753163693e-05, + "loss": 1.1385, + "step": 3451 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991789601121577e-05, + "loss": 1.0943, + "step": 3452 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991773433182665e-05, + "loss": 1.1201, + "step": 3453 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991757249346987e-05, + "loss": 1.1118, + "step": 3454 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991741049614564e-05, + "loss": 1.0248, + "step": 3455 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991724833985424e-05, + "loss": 1.0698, + "step": 3456 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991708602459592e-05, + "loss": 1.0666, + "step": 3457 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991692355037097e-05, + "loss": 0.9922, + "step": 3458 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991676091717962e-05, + "loss": 0.9685, + "step": 3459 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991659812502214e-05, + "loss": 1.0325, + "step": 3460 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991643517389877e-05, + "loss": 1.0491, + "step": 3461 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991627206380978e-05, + "loss": 0.9835, + "step": 3462 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991610879475543e-05, + "loss": 1.0039, + "step": 3463 + }, + { + "epoch": 0.26, + "learning_rate": 1.99915945366736e-05, + "loss": 1.0838, + "step": 3464 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991578177975168e-05, + "loss": 0.9964, + "step": 3465 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991561803380285e-05, + "loss": 1.0326, + "step": 3466 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991545412888964e-05, + "loss": 1.1521, + "step": 3467 + }, + { + "epoch": 0.26, + "learning_rate": 1.999152900650124e-05, + "loss": 1.0121, + "step": 3468 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991512584217136e-05, + "loss": 1.0308, + "step": 3469 + }, + { + "epoch": 0.26, + "learning_rate": 1.999149614603668e-05, + "loss": 1.0161, + "step": 3470 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991479691959893e-05, + "loss": 1.1421, + "step": 3471 + }, + { + "epoch": 0.26, + "learning_rate": 1.999146322198681e-05, + "loss": 1.1238, + "step": 3472 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991446736117444e-05, + "loss": 0.8787, + "step": 3473 + }, + { + "epoch": 0.26, + "learning_rate": 1.999143023435184e-05, + "loss": 1.0304, + "step": 3474 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991413716690005e-05, + "loss": 1.0216, + "step": 3475 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991397183131974e-05, + "loss": 1.0752, + "step": 3476 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991380633677777e-05, + "loss": 0.9998, + "step": 3477 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991364068327435e-05, + "loss": 0.9994, + "step": 3478 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991347487080975e-05, + "loss": 1.0227, + "step": 3479 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991330889938424e-05, + "loss": 1.1417, + "step": 3480 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991314276899808e-05, + "loss": 1.0973, + "step": 3481 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991297647965156e-05, + "loss": 1.0254, + "step": 3482 + }, + { + "epoch": 0.26, + "learning_rate": 1.999128100313449e-05, + "loss": 1.1437, + "step": 3483 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991264342407842e-05, + "loss": 1.0772, + "step": 3484 + }, + { + "epoch": 0.26, + "learning_rate": 1.999124766578523e-05, + "loss": 1.0351, + "step": 3485 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991230973266692e-05, + "loss": 0.9995, + "step": 3486 + }, + { + "epoch": 0.26, + "learning_rate": 1.999121426485225e-05, + "loss": 1.0261, + "step": 3487 + }, + { + "epoch": 0.26, + "learning_rate": 1.999119754054192e-05, + "loss": 0.9943, + "step": 3488 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991180800335744e-05, + "loss": 1.0802, + "step": 3489 + }, + { + "epoch": 0.26, + "learning_rate": 1.999116404423374e-05, + "loss": 1.0521, + "step": 3490 + }, + { + "epoch": 0.26, + "learning_rate": 1.999114727223594e-05, + "loss": 0.9871, + "step": 3491 + }, + { + "epoch": 0.26, + "learning_rate": 1.999113048434236e-05, + "loss": 1.1935, + "step": 3492 + }, + { + "epoch": 0.26, + "learning_rate": 1.999111368055304e-05, + "loss": 1.0653, + "step": 3493 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991096860868005e-05, + "loss": 1.0798, + "step": 3494 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991080025287272e-05, + "loss": 1.0266, + "step": 3495 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991063173810877e-05, + "loss": 1.0848, + "step": 3496 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991046306438842e-05, + "loss": 1.0812, + "step": 3497 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991029423171194e-05, + "loss": 1.2248, + "step": 3498 + }, + { + "epoch": 0.26, + "learning_rate": 1.9991012524007963e-05, + "loss": 0.9757, + "step": 3499 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990995608949176e-05, + "loss": 0.982, + "step": 3500 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990978677994853e-05, + "loss": 1.0752, + "step": 3501 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990961731145025e-05, + "loss": 1.1254, + "step": 3502 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990944768399723e-05, + "loss": 1.012, + "step": 3503 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990927789758974e-05, + "loss": 1.0093, + "step": 3504 + }, + { + "epoch": 0.26, + "learning_rate": 1.99909107952228e-05, + "loss": 1.057, + "step": 3505 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990893784791224e-05, + "loss": 1.0151, + "step": 3506 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990876758464285e-05, + "loss": 1.1254, + "step": 3507 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990859716242004e-05, + "loss": 1.0359, + "step": 3508 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990842658124407e-05, + "loss": 1.0584, + "step": 3509 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990825584111522e-05, + "loss": 1.0952, + "step": 3510 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990808494203376e-05, + "loss": 1.0551, + "step": 3511 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990791388399996e-05, + "loss": 0.9986, + "step": 3512 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990774266701412e-05, + "loss": 1.2226, + "step": 3513 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990757129107646e-05, + "loss": 1.1863, + "step": 3514 + }, + { + "epoch": 0.26, + "learning_rate": 1.999073997561873e-05, + "loss": 0.987, + "step": 3515 + }, + { + "epoch": 0.26, + "learning_rate": 1.999072280623469e-05, + "loss": 1.0396, + "step": 3516 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990705620955553e-05, + "loss": 1.0192, + "step": 3517 + }, + { + "epoch": 0.26, + "learning_rate": 1.999068841978135e-05, + "loss": 1.0254, + "step": 3518 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990671202712097e-05, + "loss": 1.0001, + "step": 3519 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990653969747834e-05, + "loss": 1.0076, + "step": 3520 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990636720888582e-05, + "loss": 1.1714, + "step": 3521 + }, + { + "epoch": 0.26, + "learning_rate": 1.999061945613437e-05, + "loss": 1.17, + "step": 3522 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990602175485227e-05, + "loss": 0.9581, + "step": 3523 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990584878941174e-05, + "loss": 1.0588, + "step": 3524 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990567566502246e-05, + "loss": 1.0617, + "step": 3525 + }, + { + "epoch": 0.26, + "learning_rate": 1.999055023816847e-05, + "loss": 1.0082, + "step": 3526 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990532893939872e-05, + "loss": 1.0887, + "step": 3527 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990515533816476e-05, + "loss": 1.0415, + "step": 3528 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990498157798317e-05, + "loss": 1.2394, + "step": 3529 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990480765885415e-05, + "loss": 1.0955, + "step": 3530 + }, + { + "epoch": 0.26, + "learning_rate": 1.99904633580778e-05, + "loss": 1.144, + "step": 3531 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990445934375506e-05, + "loss": 1.2086, + "step": 3532 + }, + { + "epoch": 0.26, + "learning_rate": 1.999042849477855e-05, + "loss": 1.0464, + "step": 3533 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990411039286965e-05, + "loss": 0.9437, + "step": 3534 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990393567900783e-05, + "loss": 1.0567, + "step": 3535 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990376080620025e-05, + "loss": 0.9591, + "step": 3536 + }, + { + "epoch": 0.26, + "learning_rate": 1.999035857744472e-05, + "loss": 1.1212, + "step": 3537 + }, + { + "epoch": 0.26, + "learning_rate": 1.99903410583749e-05, + "loss": 1.0257, + "step": 3538 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990323523410593e-05, + "loss": 0.9783, + "step": 3539 + }, + { + "epoch": 0.26, + "learning_rate": 1.999030597255182e-05, + "loss": 1.0482, + "step": 3540 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990288405798616e-05, + "loss": 1.0974, + "step": 3541 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990270823151e-05, + "loss": 1.0901, + "step": 3542 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990253224609013e-05, + "loss": 1.1157, + "step": 3543 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990235610172672e-05, + "loss": 0.9662, + "step": 3544 + }, + { + "epoch": 0.26, + "learning_rate": 1.999021797984201e-05, + "loss": 1.0803, + "step": 3545 + }, + { + "epoch": 0.26, + "learning_rate": 1.999020033361706e-05, + "loss": 1.0706, + "step": 3546 + }, + { + "epoch": 0.26, + "learning_rate": 1.999018267149784e-05, + "loss": 1.0706, + "step": 3547 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990164993484377e-05, + "loss": 1.0341, + "step": 3548 + }, + { + "epoch": 0.26, + "learning_rate": 1.999014729957671e-05, + "loss": 1.0298, + "step": 3549 + }, + { + "epoch": 0.26, + "learning_rate": 1.999012958977486e-05, + "loss": 1.0833, + "step": 3550 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990111864078855e-05, + "loss": 0.9721, + "step": 3551 + }, + { + "epoch": 0.26, + "learning_rate": 1.999009412248873e-05, + "loss": 1.0673, + "step": 3552 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990076365004506e-05, + "loss": 1.0523, + "step": 3553 + }, + { + "epoch": 0.26, + "learning_rate": 1.999005859162621e-05, + "loss": 1.1796, + "step": 3554 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990040802353876e-05, + "loss": 1.2093, + "step": 3555 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990022997187532e-05, + "loss": 1.0295, + "step": 3556 + }, + { + "epoch": 0.26, + "learning_rate": 1.9990005176127203e-05, + "loss": 1.0283, + "step": 3557 + }, + { + "epoch": 0.26, + "learning_rate": 1.998998733917292e-05, + "loss": 1.1378, + "step": 3558 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989969486324707e-05, + "loss": 1.105, + "step": 3559 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989951617582597e-05, + "loss": 1.0892, + "step": 3560 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989933732946615e-05, + "loss": 1.1322, + "step": 3561 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989915832416795e-05, + "loss": 1.0886, + "step": 3562 + }, + { + "epoch": 0.26, + "learning_rate": 1.998989791599316e-05, + "loss": 1.002, + "step": 3563 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989879983675742e-05, + "loss": 1.0468, + "step": 3564 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989862035464567e-05, + "loss": 0.996, + "step": 3565 + }, + { + "epoch": 0.26, + "learning_rate": 1.998984407135966e-05, + "loss": 1.1914, + "step": 3566 + }, + { + "epoch": 0.26, + "learning_rate": 1.998982609136106e-05, + "loss": 1.1419, + "step": 3567 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989808095468787e-05, + "loss": 1.0768, + "step": 3568 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989790083682873e-05, + "loss": 1.0021, + "step": 3569 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989772056003348e-05, + "loss": 1.0536, + "step": 3570 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989754012430233e-05, + "loss": 1.0803, + "step": 3571 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989735952963567e-05, + "loss": 1.1888, + "step": 3572 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989717877603372e-05, + "loss": 1.1043, + "step": 3573 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989699786349677e-05, + "loss": 1.0664, + "step": 3574 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989681679202517e-05, + "loss": 1.0733, + "step": 3575 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989663556161916e-05, + "loss": 1.1311, + "step": 3576 + }, + { + "epoch": 0.26, + "learning_rate": 1.99896454172279e-05, + "loss": 1.0211, + "step": 3577 + }, + { + "epoch": 0.26, + "learning_rate": 1.99896272624005e-05, + "loss": 1.0281, + "step": 3578 + }, + { + "epoch": 0.26, + "learning_rate": 1.998960909167975e-05, + "loss": 1.0352, + "step": 3579 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989590905065673e-05, + "loss": 1.0372, + "step": 3580 + }, + { + "epoch": 0.26, + "learning_rate": 1.99895727025583e-05, + "loss": 0.9982, + "step": 3581 + }, + { + "epoch": 0.26, + "learning_rate": 1.998955448415766e-05, + "loss": 1.0819, + "step": 3582 + }, + { + "epoch": 0.26, + "learning_rate": 1.998953624986378e-05, + "loss": 1.0604, + "step": 3583 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989517999676696e-05, + "loss": 0.9622, + "step": 3584 + }, + { + "epoch": 0.26, + "learning_rate": 1.9989499733596428e-05, + "loss": 1.19, + "step": 3585 + }, + { + "epoch": 0.26, + "learning_rate": 1.998948145162301e-05, + "loss": 1.0544, + "step": 3586 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989463153756468e-05, + "loss": 1.0043, + "step": 3587 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989444839996833e-05, + "loss": 1.1529, + "step": 3588 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989426510344137e-05, + "loss": 1.1006, + "step": 3589 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989408164798402e-05, + "loss": 1.2297, + "step": 3590 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989389803359666e-05, + "loss": 1.0615, + "step": 3591 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989371426027953e-05, + "loss": 1.0564, + "step": 3592 + }, + { + "epoch": 0.27, + "learning_rate": 1.998935303280329e-05, + "loss": 1.0556, + "step": 3593 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989334623685712e-05, + "loss": 1.0644, + "step": 3594 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989316198675243e-05, + "loss": 0.9932, + "step": 3595 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989297757771918e-05, + "loss": 1.0098, + "step": 3596 + }, + { + "epoch": 0.27, + "learning_rate": 1.998927930097576e-05, + "loss": 1.0871, + "step": 3597 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989260828286806e-05, + "loss": 0.9884, + "step": 3598 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989242339705077e-05, + "loss": 1.0607, + "step": 3599 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989223835230608e-05, + "loss": 0.9518, + "step": 3600 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989205314863426e-05, + "loss": 1.185, + "step": 3601 + }, + { + "epoch": 0.27, + "learning_rate": 1.998918677860356e-05, + "loss": 1.1317, + "step": 3602 + }, + { + "epoch": 0.27, + "learning_rate": 1.998916822645104e-05, + "loss": 1.1262, + "step": 3603 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989149658405898e-05, + "loss": 1.0663, + "step": 3604 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989131074468165e-05, + "loss": 1.1433, + "step": 3605 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989112474637867e-05, + "loss": 1.1524, + "step": 3606 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989093858915033e-05, + "loss": 1.2387, + "step": 3607 + }, + { + "epoch": 0.27, + "learning_rate": 1.998907522729969e-05, + "loss": 1.0561, + "step": 3608 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989056579791875e-05, + "loss": 0.9234, + "step": 3609 + }, + { + "epoch": 0.27, + "learning_rate": 1.998903791639161e-05, + "loss": 1.0305, + "step": 3610 + }, + { + "epoch": 0.27, + "learning_rate": 1.998901923709893e-05, + "loss": 1.1805, + "step": 3611 + }, + { + "epoch": 0.27, + "learning_rate": 1.9989000541913866e-05, + "loss": 1.0113, + "step": 3612 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988981830836443e-05, + "loss": 1.0489, + "step": 3613 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988963103866693e-05, + "loss": 1.0858, + "step": 3614 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988944361004647e-05, + "loss": 0.9887, + "step": 3615 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988925602250332e-05, + "loss": 1.021, + "step": 3616 + }, + { + "epoch": 0.27, + "learning_rate": 1.998890682760378e-05, + "loss": 1.0128, + "step": 3617 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988888037065018e-05, + "loss": 1.0065, + "step": 3618 + }, + { + "epoch": 0.27, + "learning_rate": 1.998886923063408e-05, + "loss": 1.0845, + "step": 3619 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988850408310996e-05, + "loss": 1.011, + "step": 3620 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988831570095792e-05, + "loss": 1.0946, + "step": 3621 + }, + { + "epoch": 0.27, + "learning_rate": 1.99888127159885e-05, + "loss": 1.0447, + "step": 3622 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988793845989147e-05, + "loss": 0.9925, + "step": 3623 + }, + { + "epoch": 0.27, + "learning_rate": 1.998877496009777e-05, + "loss": 1.1252, + "step": 3624 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988756058314394e-05, + "loss": 1.0757, + "step": 3625 + }, + { + "epoch": 0.27, + "learning_rate": 1.998873714063905e-05, + "loss": 1.0232, + "step": 3626 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988718207071766e-05, + "loss": 0.925, + "step": 3627 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988699257612577e-05, + "loss": 1.0808, + "step": 3628 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988680292261508e-05, + "loss": 1.0842, + "step": 3629 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988661311018593e-05, + "loss": 1.0542, + "step": 3630 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988642313883863e-05, + "loss": 0.9681, + "step": 3631 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988623300857338e-05, + "loss": 0.9609, + "step": 3632 + }, + { + "epoch": 0.27, + "learning_rate": 1.998860427193906e-05, + "loss": 1.0837, + "step": 3633 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988585227129062e-05, + "loss": 0.9408, + "step": 3634 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988566166427362e-05, + "loss": 1.0583, + "step": 3635 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988547089834e-05, + "loss": 1.0236, + "step": 3636 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988527997348997e-05, + "loss": 0.9807, + "step": 3637 + }, + { + "epoch": 0.27, + "learning_rate": 1.998850888897239e-05, + "loss": 1.0349, + "step": 3638 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988489764704215e-05, + "loss": 1.1008, + "step": 3639 + }, + { + "epoch": 0.27, + "learning_rate": 1.998847062454449e-05, + "loss": 1.0638, + "step": 3640 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988451468493252e-05, + "loss": 1.0835, + "step": 3641 + }, + { + "epoch": 0.27, + "learning_rate": 1.998843229655053e-05, + "loss": 1.06, + "step": 3642 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988413108716352e-05, + "loss": 1.2043, + "step": 3643 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988393904990757e-05, + "loss": 0.9816, + "step": 3644 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988374685373767e-05, + "loss": 1.1688, + "step": 3645 + }, + { + "epoch": 0.27, + "learning_rate": 1.998835544986542e-05, + "loss": 1.0884, + "step": 3646 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988336198465737e-05, + "loss": 1.0082, + "step": 3647 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988316931174758e-05, + "loss": 1.0619, + "step": 3648 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988297647992505e-05, + "loss": 0.9629, + "step": 3649 + }, + { + "epoch": 0.27, + "learning_rate": 1.998827834891902e-05, + "loss": 1.0766, + "step": 3650 + }, + { + "epoch": 0.27, + "learning_rate": 1.998825903395432e-05, + "loss": 1.0393, + "step": 3651 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988239703098447e-05, + "loss": 1.047, + "step": 3652 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988220356351427e-05, + "loss": 1.1249, + "step": 3653 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988200993713293e-05, + "loss": 1.0737, + "step": 3654 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988181615184072e-05, + "loss": 1.0962, + "step": 3655 + }, + { + "epoch": 0.27, + "learning_rate": 1.99881622207638e-05, + "loss": 1.0302, + "step": 3656 + }, + { + "epoch": 0.27, + "learning_rate": 1.99881428104525e-05, + "loss": 1.0264, + "step": 3657 + }, + { + "epoch": 0.27, + "learning_rate": 1.998812338425021e-05, + "loss": 0.9979, + "step": 3658 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988103942156957e-05, + "loss": 1.0385, + "step": 3659 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988084484172776e-05, + "loss": 0.9613, + "step": 3660 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988065010297695e-05, + "loss": 1.0078, + "step": 3661 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988045520531743e-05, + "loss": 1.0623, + "step": 3662 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988026014874956e-05, + "loss": 1.0271, + "step": 3663 + }, + { + "epoch": 0.27, + "learning_rate": 1.9988006493327363e-05, + "loss": 0.9301, + "step": 3664 + }, + { + "epoch": 0.27, + "learning_rate": 1.998798695588899e-05, + "loss": 1.1648, + "step": 3665 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987967402559878e-05, + "loss": 1.0301, + "step": 3666 + }, + { + "epoch": 0.27, + "learning_rate": 1.998794783334005e-05, + "loss": 1.081, + "step": 3667 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987928248229543e-05, + "loss": 0.9422, + "step": 3668 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987908647228382e-05, + "loss": 1.0467, + "step": 3669 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987889030336602e-05, + "loss": 1.0587, + "step": 3670 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987869397554233e-05, + "loss": 0.9926, + "step": 3671 + }, + { + "epoch": 0.27, + "learning_rate": 1.998784974888131e-05, + "loss": 1.1036, + "step": 3672 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987830084317855e-05, + "loss": 1.0597, + "step": 3673 + }, + { + "epoch": 0.27, + "learning_rate": 1.998781040386391e-05, + "loss": 1.0116, + "step": 3674 + }, + { + "epoch": 0.27, + "learning_rate": 1.99877907075195e-05, + "loss": 0.9563, + "step": 3675 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987770995284655e-05, + "loss": 0.9994, + "step": 3676 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987751267159414e-05, + "loss": 1.1095, + "step": 3677 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987731523143804e-05, + "loss": 0.9327, + "step": 3678 + }, + { + "epoch": 0.27, + "learning_rate": 1.998771176323785e-05, + "loss": 1.1739, + "step": 3679 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987691987441595e-05, + "loss": 1.0097, + "step": 3680 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987672195755062e-05, + "loss": 1.1714, + "step": 3681 + }, + { + "epoch": 0.27, + "learning_rate": 1.998765238817829e-05, + "loss": 0.9778, + "step": 3682 + }, + { + "epoch": 0.27, + "learning_rate": 1.99876325647113e-05, + "loss": 1.025, + "step": 3683 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987612725354134e-05, + "loss": 1.1063, + "step": 3684 + }, + { + "epoch": 0.27, + "learning_rate": 1.998759287010682e-05, + "loss": 1.1034, + "step": 3685 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987572998969382e-05, + "loss": 1.1259, + "step": 3686 + }, + { + "epoch": 0.27, + "learning_rate": 1.998755311194186e-05, + "loss": 1.0193, + "step": 3687 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987533209024287e-05, + "loss": 1.1191, + "step": 3688 + }, + { + "epoch": 0.27, + "learning_rate": 1.998751329021669e-05, + "loss": 1.0907, + "step": 3689 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987493355519107e-05, + "loss": 0.9721, + "step": 3690 + }, + { + "epoch": 0.27, + "learning_rate": 1.998747340493156e-05, + "loss": 0.9483, + "step": 3691 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987453438454085e-05, + "loss": 1.1066, + "step": 3692 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987433456086715e-05, + "loss": 1.0999, + "step": 3693 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987413457829484e-05, + "loss": 0.8734, + "step": 3694 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987393443682416e-05, + "loss": 1.0561, + "step": 3695 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987373413645552e-05, + "loss": 1.166, + "step": 3696 + }, + { + "epoch": 0.27, + "learning_rate": 1.998735336771892e-05, + "loss": 1.067, + "step": 3697 + }, + { + "epoch": 0.27, + "learning_rate": 1.998733330590255e-05, + "loss": 0.9599, + "step": 3698 + }, + { + "epoch": 0.27, + "learning_rate": 1.998731322819648e-05, + "loss": 0.9577, + "step": 3699 + }, + { + "epoch": 0.27, + "learning_rate": 1.998729313460073e-05, + "loss": 1.0924, + "step": 3700 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987273025115344e-05, + "loss": 1.1498, + "step": 3701 + }, + { + "epoch": 0.27, + "learning_rate": 1.998725289974035e-05, + "loss": 1.0817, + "step": 3702 + }, + { + "epoch": 0.27, + "learning_rate": 1.998723275847578e-05, + "loss": 1.0807, + "step": 3703 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987212601321663e-05, + "loss": 0.9878, + "step": 3704 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987192428278034e-05, + "loss": 1.0298, + "step": 3705 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987172239344925e-05, + "loss": 1.0489, + "step": 3706 + }, + { + "epoch": 0.27, + "learning_rate": 1.998715203452237e-05, + "loss": 1.0833, + "step": 3707 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987131813810396e-05, + "loss": 1.0231, + "step": 3708 + }, + { + "epoch": 0.27, + "learning_rate": 1.998711157720904e-05, + "loss": 1.1237, + "step": 3709 + }, + { + "epoch": 0.27, + "learning_rate": 1.998709132471833e-05, + "loss": 1.1735, + "step": 3710 + }, + { + "epoch": 0.27, + "learning_rate": 1.9987071056338304e-05, + "loss": 0.9844, + "step": 3711 + }, + { + "epoch": 0.27, + "learning_rate": 1.998705077206899e-05, + "loss": 1.0043, + "step": 3712 + }, + { + "epoch": 0.27, + "learning_rate": 1.998703047191042e-05, + "loss": 0.9204, + "step": 3713 + }, + { + "epoch": 0.27, + "learning_rate": 1.998701015586263e-05, + "loss": 0.9597, + "step": 3714 + }, + { + "epoch": 0.27, + "learning_rate": 1.9986989823925645e-05, + "loss": 1.1109, + "step": 3715 + }, + { + "epoch": 0.27, + "learning_rate": 1.9986969476099505e-05, + "loss": 1.0622, + "step": 3716 + }, + { + "epoch": 0.27, + "learning_rate": 1.9986949112384237e-05, + "loss": 0.9364, + "step": 3717 + }, + { + "epoch": 0.27, + "learning_rate": 1.998692873277988e-05, + "loss": 1.0755, + "step": 3718 + }, + { + "epoch": 0.27, + "learning_rate": 1.9986908337286458e-05, + "loss": 1.1958, + "step": 3719 + }, + { + "epoch": 0.27, + "learning_rate": 1.998688792590401e-05, + "loss": 0.9823, + "step": 3720 + }, + { + "epoch": 0.27, + "learning_rate": 1.9986867498632568e-05, + "loss": 1.0717, + "step": 3721 + }, + { + "epoch": 0.28, + "learning_rate": 1.998684705547216e-05, + "loss": 1.0527, + "step": 3722 + }, + { + "epoch": 0.28, + "learning_rate": 1.998682659642282e-05, + "loss": 1.1061, + "step": 3723 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986806121484585e-05, + "loss": 1.0137, + "step": 3724 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986785630657485e-05, + "loss": 1.1379, + "step": 3725 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986765123941548e-05, + "loss": 1.1062, + "step": 3726 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986744601336814e-05, + "loss": 1.1421, + "step": 3727 + }, + { + "epoch": 0.28, + "learning_rate": 1.998672406284331e-05, + "loss": 1.0831, + "step": 3728 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986703508461072e-05, + "loss": 1.0583, + "step": 3729 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986682938190132e-05, + "loss": 1.0305, + "step": 3730 + }, + { + "epoch": 0.28, + "learning_rate": 1.998666235203052e-05, + "loss": 1.1017, + "step": 3731 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986641749982275e-05, + "loss": 0.9613, + "step": 3732 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986621132045423e-05, + "loss": 1.0124, + "step": 3733 + }, + { + "epoch": 0.28, + "learning_rate": 1.998660049822e-05, + "loss": 1.0309, + "step": 3734 + }, + { + "epoch": 0.28, + "learning_rate": 1.998657984850604e-05, + "loss": 1.1029, + "step": 3735 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986559182903572e-05, + "loss": 1.0598, + "step": 3736 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986538501412632e-05, + "loss": 1.0355, + "step": 3737 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986517804033254e-05, + "loss": 1.0794, + "step": 3738 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986497090765466e-05, + "loss": 0.9782, + "step": 3739 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986476361609306e-05, + "loss": 1.0209, + "step": 3740 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986455616564802e-05, + "loss": 1.0207, + "step": 3741 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986434855631994e-05, + "loss": 1.0783, + "step": 3742 + }, + { + "epoch": 0.28, + "learning_rate": 1.998641407881091e-05, + "loss": 0.9751, + "step": 3743 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986393286101583e-05, + "loss": 1.0113, + "step": 3744 + }, + { + "epoch": 0.28, + "learning_rate": 1.998637247750405e-05, + "loss": 1.0744, + "step": 3745 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986351653018338e-05, + "loss": 1.0286, + "step": 3746 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986330812644484e-05, + "loss": 1.1922, + "step": 3747 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986309956382522e-05, + "loss": 0.9916, + "step": 3748 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986289084232485e-05, + "loss": 1.0002, + "step": 3749 + }, + { + "epoch": 0.28, + "learning_rate": 1.99862681961944e-05, + "loss": 1.0838, + "step": 3750 + }, + { + "epoch": 0.28, + "learning_rate": 1.998624729226831e-05, + "loss": 1.1602, + "step": 3751 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986226372454237e-05, + "loss": 1.0297, + "step": 3752 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986205436752226e-05, + "loss": 1.21, + "step": 3753 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986184485162306e-05, + "loss": 1.0978, + "step": 3754 + }, + { + "epoch": 0.28, + "learning_rate": 1.998616351768451e-05, + "loss": 1.0184, + "step": 3755 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986142534318866e-05, + "loss": 1.022, + "step": 3756 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986121535065416e-05, + "loss": 1.0162, + "step": 3757 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986100519924186e-05, + "loss": 1.0697, + "step": 3758 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986079488895217e-05, + "loss": 1.1045, + "step": 3759 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986058441978536e-05, + "loss": 0.9558, + "step": 3760 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986037379174176e-05, + "loss": 1.0567, + "step": 3761 + }, + { + "epoch": 0.28, + "learning_rate": 1.9986016300482175e-05, + "loss": 1.0546, + "step": 3762 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985995205902567e-05, + "loss": 1.0875, + "step": 3763 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985974095435385e-05, + "loss": 1.1283, + "step": 3764 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985952969080658e-05, + "loss": 1.0162, + "step": 3765 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985931826838424e-05, + "loss": 0.9453, + "step": 3766 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985910668708716e-05, + "loss": 1.1013, + "step": 3767 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985889494691566e-05, + "loss": 1.171, + "step": 3768 + }, + { + "epoch": 0.28, + "learning_rate": 1.998586830478701e-05, + "loss": 1.1081, + "step": 3769 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985847098995078e-05, + "loss": 1.0687, + "step": 3770 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985825877315808e-05, + "loss": 1.0369, + "step": 3771 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985804639749232e-05, + "loss": 1.0756, + "step": 3772 + }, + { + "epoch": 0.28, + "learning_rate": 1.998578338629538e-05, + "loss": 1.1107, + "step": 3773 + }, + { + "epoch": 0.28, + "learning_rate": 1.998576211695429e-05, + "loss": 1.0385, + "step": 3774 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985740831726e-05, + "loss": 1.1482, + "step": 3775 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985719530610538e-05, + "loss": 1.051, + "step": 3776 + }, + { + "epoch": 0.28, + "learning_rate": 1.998569821360794e-05, + "loss": 1.0534, + "step": 3777 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985676880718236e-05, + "loss": 1.1359, + "step": 3778 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985655531941462e-05, + "loss": 1.1194, + "step": 3779 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985634167277657e-05, + "loss": 1.0669, + "step": 3780 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985612786726848e-05, + "loss": 1.0434, + "step": 3781 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985591390289075e-05, + "loss": 1.0066, + "step": 3782 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985569977964365e-05, + "loss": 1.1343, + "step": 3783 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985548549752757e-05, + "loss": 1.0534, + "step": 3784 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985527105654286e-05, + "loss": 0.9592, + "step": 3785 + }, + { + "epoch": 0.28, + "learning_rate": 1.998550564566898e-05, + "loss": 1.023, + "step": 3786 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985484169796882e-05, + "loss": 0.9148, + "step": 3787 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985462678038022e-05, + "loss": 1.1055, + "step": 3788 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985441170392433e-05, + "loss": 1.1081, + "step": 3789 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985419646860148e-05, + "loss": 1.0068, + "step": 3790 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985398107441204e-05, + "loss": 1.0339, + "step": 3791 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985376552135633e-05, + "loss": 1.0927, + "step": 3792 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985354980943473e-05, + "loss": 0.9563, + "step": 3793 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985333393864754e-05, + "loss": 1.0406, + "step": 3794 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985311790899515e-05, + "loss": 1.0877, + "step": 3795 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985290172047786e-05, + "loss": 1.0709, + "step": 3796 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985268537309603e-05, + "loss": 1.0468, + "step": 3797 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985246886685005e-05, + "loss": 1.1139, + "step": 3798 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985225220174015e-05, + "loss": 1.0162, + "step": 3799 + }, + { + "epoch": 0.28, + "learning_rate": 1.998520353777668e-05, + "loss": 1.0199, + "step": 3800 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985181839493027e-05, + "loss": 0.9959, + "step": 3801 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985160125323092e-05, + "loss": 1.1135, + "step": 3802 + }, + { + "epoch": 0.28, + "learning_rate": 1.998513839526691e-05, + "loss": 0.9675, + "step": 3803 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985116649324517e-05, + "loss": 1.0543, + "step": 3804 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985094887495945e-05, + "loss": 0.9905, + "step": 3805 + }, + { + "epoch": 0.28, + "learning_rate": 1.998507310978123e-05, + "loss": 1.0303, + "step": 3806 + }, + { + "epoch": 0.28, + "learning_rate": 1.998505131618041e-05, + "loss": 1.0403, + "step": 3807 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985029506693512e-05, + "loss": 1.1596, + "step": 3808 + }, + { + "epoch": 0.28, + "learning_rate": 1.9985007681320575e-05, + "loss": 1.0946, + "step": 3809 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984985840061637e-05, + "loss": 1.0312, + "step": 3810 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984963982916725e-05, + "loss": 1.1054, + "step": 3811 + }, + { + "epoch": 0.28, + "learning_rate": 1.998494210988588e-05, + "loss": 1.0523, + "step": 3812 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984920220969138e-05, + "loss": 1.0379, + "step": 3813 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984898316166527e-05, + "loss": 1.0701, + "step": 3814 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984876395478084e-05, + "loss": 1.035, + "step": 3815 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984854458903853e-05, + "loss": 1.0505, + "step": 3816 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984832506443855e-05, + "loss": 1.0154, + "step": 3817 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984810538098133e-05, + "loss": 1.0978, + "step": 3818 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984788553866722e-05, + "loss": 1.0854, + "step": 3819 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984766553749652e-05, + "loss": 1.0901, + "step": 3820 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984744537746967e-05, + "loss": 1.0233, + "step": 3821 + }, + { + "epoch": 0.28, + "learning_rate": 1.998472250585869e-05, + "loss": 0.9932, + "step": 3822 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984700458084863e-05, + "loss": 1.045, + "step": 3823 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984678394425522e-05, + "loss": 1.0432, + "step": 3824 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984656314880698e-05, + "loss": 1.1393, + "step": 3825 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984634219450432e-05, + "loss": 1.0412, + "step": 3826 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984612108134758e-05, + "loss": 1.1067, + "step": 3827 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984589980933706e-05, + "loss": 1.088, + "step": 3828 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984567837847313e-05, + "loss": 1.0741, + "step": 3829 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984545678875617e-05, + "loss": 1.1164, + "step": 3830 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984523504018652e-05, + "loss": 1.1001, + "step": 3831 + }, + { + "epoch": 0.28, + "learning_rate": 1.998450131327645e-05, + "loss": 1.0606, + "step": 3832 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984479106649052e-05, + "loss": 1.0096, + "step": 3833 + }, + { + "epoch": 0.28, + "learning_rate": 1.998445688413649e-05, + "loss": 0.9814, + "step": 3834 + }, + { + "epoch": 0.28, + "learning_rate": 1.99844346457388e-05, + "loss": 1.0476, + "step": 3835 + }, + { + "epoch": 0.28, + "learning_rate": 1.998441239145602e-05, + "loss": 1.0191, + "step": 3836 + }, + { + "epoch": 0.28, + "learning_rate": 1.998439012128818e-05, + "loss": 1.0635, + "step": 3837 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984367835235318e-05, + "loss": 1.0165, + "step": 3838 + }, + { + "epoch": 0.28, + "learning_rate": 1.998434553329747e-05, + "loss": 1.041, + "step": 3839 + }, + { + "epoch": 0.28, + "learning_rate": 1.998432321547467e-05, + "loss": 1.0022, + "step": 3840 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984300881766957e-05, + "loss": 1.0015, + "step": 3841 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984278532174362e-05, + "loss": 1.0535, + "step": 3842 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984256166696925e-05, + "loss": 1.0975, + "step": 3843 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984233785334676e-05, + "loss": 1.0301, + "step": 3844 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984211388087655e-05, + "loss": 0.9626, + "step": 3845 + }, + { + "epoch": 0.28, + "learning_rate": 1.99841889749559e-05, + "loss": 1.1836, + "step": 3846 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984166545939438e-05, + "loss": 1.0949, + "step": 3847 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984144101038313e-05, + "loss": 1.0698, + "step": 3848 + }, + { + "epoch": 0.28, + "learning_rate": 1.998412164025256e-05, + "loss": 1.1347, + "step": 3849 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984099163582206e-05, + "loss": 1.0613, + "step": 3850 + }, + { + "epoch": 0.28, + "learning_rate": 1.99840766710273e-05, + "loss": 1.082, + "step": 3851 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984054162587864e-05, + "loss": 1.0788, + "step": 3852 + }, + { + "epoch": 0.28, + "learning_rate": 1.9984031638263946e-05, + "loss": 1.0411, + "step": 3853 + }, + { + "epoch": 0.28, + "learning_rate": 1.998400909805557e-05, + "loss": 0.9996, + "step": 3854 + }, + { + "epoch": 0.28, + "learning_rate": 1.9983986541962785e-05, + "loss": 0.9774, + "step": 3855 + }, + { + "epoch": 0.28, + "learning_rate": 1.9983963969985617e-05, + "loss": 1.0591, + "step": 3856 + }, + { + "epoch": 0.28, + "learning_rate": 1.9983941382124107e-05, + "loss": 1.0704, + "step": 3857 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983918778378287e-05, + "loss": 1.075, + "step": 3858 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983896158748197e-05, + "loss": 1.0559, + "step": 3859 + }, + { + "epoch": 0.29, + "learning_rate": 1.998387352323387e-05, + "loss": 1.0406, + "step": 3860 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983850871835346e-05, + "loss": 1.0609, + "step": 3861 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983828204552656e-05, + "loss": 0.9411, + "step": 3862 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983805521385835e-05, + "loss": 1.072, + "step": 3863 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983782822334924e-05, + "loss": 1.0846, + "step": 3864 + }, + { + "epoch": 0.29, + "learning_rate": 1.998376010739996e-05, + "loss": 1.0776, + "step": 3865 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983737376580976e-05, + "loss": 1.0346, + "step": 3866 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983714629878007e-05, + "loss": 0.9697, + "step": 3867 + }, + { + "epoch": 0.29, + "learning_rate": 1.998369186729109e-05, + "loss": 1.1129, + "step": 3868 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983669088820266e-05, + "loss": 1.0912, + "step": 3869 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983646294465565e-05, + "loss": 1.12, + "step": 3870 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983623484227027e-05, + "loss": 1.1554, + "step": 3871 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983600658104684e-05, + "loss": 1.072, + "step": 3872 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983577816098576e-05, + "loss": 1.1013, + "step": 3873 + }, + { + "epoch": 0.29, + "learning_rate": 1.998355495820874e-05, + "loss": 0.9869, + "step": 3874 + }, + { + "epoch": 0.29, + "learning_rate": 1.998353208443521e-05, + "loss": 1.0976, + "step": 3875 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983509194778022e-05, + "loss": 1.0814, + "step": 3876 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983486289237216e-05, + "loss": 1.0136, + "step": 3877 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983463367812824e-05, + "loss": 1.0095, + "step": 3878 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983440430504887e-05, + "loss": 0.9318, + "step": 3879 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983417477313436e-05, + "loss": 1.08, + "step": 3880 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983394508238514e-05, + "loss": 1.0769, + "step": 3881 + }, + { + "epoch": 0.29, + "learning_rate": 1.998337152328015e-05, + "loss": 1.1051, + "step": 3882 + }, + { + "epoch": 0.29, + "learning_rate": 1.998334852243839e-05, + "loss": 1.0698, + "step": 3883 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983325505713263e-05, + "loss": 1.0597, + "step": 3884 + }, + { + "epoch": 0.29, + "learning_rate": 1.998330247310481e-05, + "loss": 1.179, + "step": 3885 + }, + { + "epoch": 0.29, + "learning_rate": 1.998327942461306e-05, + "loss": 1.1343, + "step": 3886 + }, + { + "epoch": 0.29, + "learning_rate": 1.998325636023806e-05, + "loss": 1.0586, + "step": 3887 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983233279979838e-05, + "loss": 1.0354, + "step": 3888 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983210183838438e-05, + "loss": 1.1534, + "step": 3889 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983187071813896e-05, + "loss": 1.0001, + "step": 3890 + }, + { + "epoch": 0.29, + "learning_rate": 1.998316394390624e-05, + "loss": 1.1491, + "step": 3891 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983140800115516e-05, + "loss": 1.0688, + "step": 3892 + }, + { + "epoch": 0.29, + "learning_rate": 1.998311764044176e-05, + "loss": 1.1663, + "step": 3893 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983094464885002e-05, + "loss": 1.0134, + "step": 3894 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983071273445284e-05, + "loss": 0.9894, + "step": 3895 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983048066122647e-05, + "loss": 1.047, + "step": 3896 + }, + { + "epoch": 0.29, + "learning_rate": 1.998302484291712e-05, + "loss": 1.0134, + "step": 3897 + }, + { + "epoch": 0.29, + "learning_rate": 1.9983001603828745e-05, + "loss": 1.0674, + "step": 3898 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982978348857555e-05, + "loss": 1.0701, + "step": 3899 + }, + { + "epoch": 0.29, + "learning_rate": 1.998295507800359e-05, + "loss": 1.1602, + "step": 3900 + }, + { + "epoch": 0.29, + "learning_rate": 1.998293179126689e-05, + "loss": 1.0479, + "step": 3901 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982908488647483e-05, + "loss": 1.1088, + "step": 3902 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982885170145413e-05, + "loss": 1.0478, + "step": 3903 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982861835760715e-05, + "loss": 1.0508, + "step": 3904 + }, + { + "epoch": 0.29, + "learning_rate": 1.998283848549343e-05, + "loss": 1.0567, + "step": 3905 + }, + { + "epoch": 0.29, + "learning_rate": 1.998281511934359e-05, + "loss": 1.0527, + "step": 3906 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982791737311233e-05, + "loss": 1.0615, + "step": 3907 + }, + { + "epoch": 0.29, + "learning_rate": 1.99827683393964e-05, + "loss": 0.9713, + "step": 3908 + }, + { + "epoch": 0.29, + "learning_rate": 1.998274492559912e-05, + "loss": 0.9921, + "step": 3909 + }, + { + "epoch": 0.29, + "learning_rate": 1.998272149591944e-05, + "loss": 1.1799, + "step": 3910 + }, + { + "epoch": 0.29, + "learning_rate": 1.998269805035739e-05, + "loss": 1.0036, + "step": 3911 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982674588913012e-05, + "loss": 1.0796, + "step": 3912 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982651111586345e-05, + "loss": 1.1279, + "step": 3913 + }, + { + "epoch": 0.29, + "learning_rate": 1.998262761837742e-05, + "loss": 0.9533, + "step": 3914 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982604109286273e-05, + "loss": 0.975, + "step": 3915 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982580584312952e-05, + "loss": 1.1037, + "step": 3916 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982557043457486e-05, + "loss": 1.04, + "step": 3917 + }, + { + "epoch": 0.29, + "learning_rate": 1.998253348671991e-05, + "loss": 1.0994, + "step": 3918 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982509914100272e-05, + "loss": 1.0195, + "step": 3919 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982486325598603e-05, + "loss": 0.96, + "step": 3920 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982462721214937e-05, + "loss": 1.0696, + "step": 3921 + }, + { + "epoch": 0.29, + "learning_rate": 1.998243910094932e-05, + "loss": 1.0039, + "step": 3922 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982415464801786e-05, + "loss": 0.9851, + "step": 3923 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982391812772368e-05, + "loss": 1.1801, + "step": 3924 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982368144861112e-05, + "loss": 1.0695, + "step": 3925 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982344461068047e-05, + "loss": 1.1671, + "step": 3926 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982320761393215e-05, + "loss": 1.108, + "step": 3927 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982297045836655e-05, + "loss": 1.1152, + "step": 3928 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982273314398402e-05, + "loss": 1.1141, + "step": 3929 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982249567078496e-05, + "loss": 1.1079, + "step": 3930 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982225803876973e-05, + "loss": 1.1255, + "step": 3931 + }, + { + "epoch": 0.29, + "learning_rate": 1.998220202479387e-05, + "loss": 0.9812, + "step": 3932 + }, + { + "epoch": 0.29, + "learning_rate": 1.998217822982923e-05, + "loss": 1.055, + "step": 3933 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982154418983083e-05, + "loss": 1.0966, + "step": 3934 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982130592255474e-05, + "loss": 1.0399, + "step": 3935 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982106749646438e-05, + "loss": 0.9162, + "step": 3936 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982082891156015e-05, + "loss": 1.0579, + "step": 3937 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982059016784237e-05, + "loss": 1.1217, + "step": 3938 + }, + { + "epoch": 0.29, + "learning_rate": 1.998203512653115e-05, + "loss": 1.0481, + "step": 3939 + }, + { + "epoch": 0.29, + "learning_rate": 1.9982011220396783e-05, + "loss": 1.0515, + "step": 3940 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981987298381183e-05, + "loss": 1.1395, + "step": 3941 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981963360484383e-05, + "loss": 1.0151, + "step": 3942 + }, + { + "epoch": 0.29, + "learning_rate": 1.998193940670642e-05, + "loss": 1.1058, + "step": 3943 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981915437047337e-05, + "loss": 1.1126, + "step": 3944 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981891451507167e-05, + "loss": 0.9586, + "step": 3945 + }, + { + "epoch": 0.29, + "learning_rate": 1.998186745008595e-05, + "loss": 1.0428, + "step": 3946 + }, + { + "epoch": 0.29, + "learning_rate": 1.998184343278373e-05, + "loss": 1.1259, + "step": 3947 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981819399600537e-05, + "loss": 1.1804, + "step": 3948 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981795350536412e-05, + "loss": 1.1258, + "step": 3949 + }, + { + "epoch": 0.29, + "learning_rate": 1.998177128559139e-05, + "loss": 1.1926, + "step": 3950 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981747204765516e-05, + "loss": 1.0039, + "step": 3951 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981723108058824e-05, + "loss": 1.0358, + "step": 3952 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981698995471355e-05, + "loss": 1.0619, + "step": 3953 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981674867003146e-05, + "loss": 1.1597, + "step": 3954 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981650722654235e-05, + "loss": 1.0672, + "step": 3955 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981626562424656e-05, + "loss": 1.1524, + "step": 3956 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981602386314455e-05, + "loss": 1.0623, + "step": 3957 + }, + { + "epoch": 0.29, + "learning_rate": 1.998157819432367e-05, + "loss": 1.104, + "step": 3958 + }, + { + "epoch": 0.29, + "learning_rate": 1.998155398645233e-05, + "loss": 1.1014, + "step": 3959 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981529762700488e-05, + "loss": 1.1268, + "step": 3960 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981505523068173e-05, + "loss": 0.9619, + "step": 3961 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981481267555424e-05, + "loss": 1.1303, + "step": 3962 + }, + { + "epoch": 0.29, + "learning_rate": 1.998145699616228e-05, + "loss": 1.1077, + "step": 3963 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981432708888784e-05, + "loss": 0.9435, + "step": 3964 + }, + { + "epoch": 0.29, + "learning_rate": 1.998140840573497e-05, + "loss": 1.2017, + "step": 3965 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981384086700877e-05, + "loss": 0.9167, + "step": 3966 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981359751786545e-05, + "loss": 1.0874, + "step": 3967 + }, + { + "epoch": 0.29, + "learning_rate": 1.998133540099201e-05, + "loss": 1.07, + "step": 3968 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981311034317318e-05, + "loss": 1.0448, + "step": 3969 + }, + { + "epoch": 0.29, + "learning_rate": 1.99812866517625e-05, + "loss": 1.0562, + "step": 3970 + }, + { + "epoch": 0.29, + "learning_rate": 1.99812622533276e-05, + "loss": 1.0697, + "step": 3971 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981237839012655e-05, + "loss": 1.1279, + "step": 3972 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981213408817697e-05, + "loss": 0.9898, + "step": 3973 + }, + { + "epoch": 0.29, + "learning_rate": 1.998118896274278e-05, + "loss": 1.0989, + "step": 3974 + }, + { + "epoch": 0.29, + "learning_rate": 1.998116450078793e-05, + "loss": 1.1848, + "step": 3975 + }, + { + "epoch": 0.29, + "learning_rate": 1.998114002295319e-05, + "loss": 1.1861, + "step": 3976 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981115529238598e-05, + "loss": 1.1325, + "step": 3977 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981091019644198e-05, + "loss": 1.0364, + "step": 3978 + }, + { + "epoch": 0.29, + "learning_rate": 1.998106649417002e-05, + "loss": 1.0603, + "step": 3979 + }, + { + "epoch": 0.29, + "learning_rate": 1.998104195281611e-05, + "loss": 1.0144, + "step": 3980 + }, + { + "epoch": 0.29, + "learning_rate": 1.9981017395582507e-05, + "loss": 1.0936, + "step": 3981 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980992822469245e-05, + "loss": 1.1525, + "step": 3982 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980968233476368e-05, + "loss": 1.0618, + "step": 3983 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980943628603914e-05, + "loss": 0.9211, + "step": 3984 + }, + { + "epoch": 0.29, + "learning_rate": 1.998091900785192e-05, + "loss": 1.015, + "step": 3985 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980894371220426e-05, + "loss": 0.9082, + "step": 3986 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980869718709474e-05, + "loss": 1.1791, + "step": 3987 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980845050319103e-05, + "loss": 1.0312, + "step": 3988 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980820366049345e-05, + "loss": 1.0381, + "step": 3989 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980795665900247e-05, + "loss": 1.086, + "step": 3990 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980770949871846e-05, + "loss": 1.0357, + "step": 3991 + }, + { + "epoch": 0.29, + "learning_rate": 1.9980746217964182e-05, + "loss": 1.0992, + "step": 3992 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980721470177293e-05, + "loss": 1.0999, + "step": 3993 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980696706511217e-05, + "loss": 1.1025, + "step": 3994 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980671926966e-05, + "loss": 0.9411, + "step": 3995 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980647131541674e-05, + "loss": 1.0, + "step": 3996 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980622320238278e-05, + "loss": 0.9325, + "step": 3997 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980597493055857e-05, + "loss": 1.1058, + "step": 3998 + }, + { + "epoch": 0.3, + "learning_rate": 1.998057264999445e-05, + "loss": 1.1382, + "step": 3999 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980547791054093e-05, + "loss": 1.0554, + "step": 4000 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980522916234827e-05, + "loss": 1.0448, + "step": 4001 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980498025536696e-05, + "loss": 1.0523, + "step": 4002 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980473118959732e-05, + "loss": 1.1203, + "step": 4003 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980448196503977e-05, + "loss": 1.0198, + "step": 4004 + }, + { + "epoch": 0.3, + "learning_rate": 1.998042325816947e-05, + "loss": 1.0612, + "step": 4005 + }, + { + "epoch": 0.3, + "learning_rate": 1.998039830395626e-05, + "loss": 1.0367, + "step": 4006 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980373333864374e-05, + "loss": 1.1328, + "step": 4007 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980348347893857e-05, + "loss": 1.1628, + "step": 4008 + }, + { + "epoch": 0.3, + "learning_rate": 1.998032334604475e-05, + "loss": 1.0538, + "step": 4009 + }, + { + "epoch": 0.3, + "learning_rate": 1.998029832831709e-05, + "loss": 1.0454, + "step": 4010 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980273294710916e-05, + "loss": 1.014, + "step": 4011 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980248245226274e-05, + "loss": 1.0783, + "step": 4012 + }, + { + "epoch": 0.3, + "learning_rate": 1.99802231798632e-05, + "loss": 1.0305, + "step": 4013 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980198098621727e-05, + "loss": 1.1344, + "step": 4014 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980173001501906e-05, + "loss": 1.1332, + "step": 4015 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980147888503776e-05, + "loss": 1.0536, + "step": 4016 + }, + { + "epoch": 0.3, + "learning_rate": 1.998012275962737e-05, + "loss": 1.0177, + "step": 4017 + }, + { + "epoch": 0.3, + "learning_rate": 1.998009761487273e-05, + "loss": 1.052, + "step": 4018 + }, + { + "epoch": 0.3, + "learning_rate": 1.99800724542399e-05, + "loss": 1.1724, + "step": 4019 + }, + { + "epoch": 0.3, + "learning_rate": 1.9980047277728917e-05, + "loss": 1.0405, + "step": 4020 + }, + { + "epoch": 0.3, + "learning_rate": 1.998002208533982e-05, + "loss": 1.0945, + "step": 4021 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979996877072657e-05, + "loss": 1.0553, + "step": 4022 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979971652927456e-05, + "loss": 1.0184, + "step": 4023 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979946412904263e-05, + "loss": 1.0871, + "step": 4024 + }, + { + "epoch": 0.3, + "learning_rate": 1.997992115700312e-05, + "loss": 1.0756, + "step": 4025 + }, + { + "epoch": 0.3, + "learning_rate": 1.997989588522406e-05, + "loss": 0.9329, + "step": 4026 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979870597567138e-05, + "loss": 1.0308, + "step": 4027 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979845294032378e-05, + "loss": 1.0293, + "step": 4028 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979819974619826e-05, + "loss": 1.0835, + "step": 4029 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979794639329527e-05, + "loss": 1.0718, + "step": 4030 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979769288161517e-05, + "loss": 1.0081, + "step": 4031 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979743921115837e-05, + "loss": 1.161, + "step": 4032 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979718538192524e-05, + "loss": 1.0558, + "step": 4033 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979693139391623e-05, + "loss": 1.0652, + "step": 4034 + }, + { + "epoch": 0.3, + "learning_rate": 1.997966772471317e-05, + "loss": 1.1855, + "step": 4035 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979642294157215e-05, + "loss": 0.9657, + "step": 4036 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979616847723792e-05, + "loss": 1.0906, + "step": 4037 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979591385412937e-05, + "loss": 0.96, + "step": 4038 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979565907224696e-05, + "loss": 1.0468, + "step": 4039 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979540413159107e-05, + "loss": 1.1651, + "step": 4040 + }, + { + "epoch": 0.3, + "learning_rate": 1.997951490321621e-05, + "loss": 1.0573, + "step": 4041 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979489377396055e-05, + "loss": 1.0515, + "step": 4042 + }, + { + "epoch": 0.3, + "learning_rate": 1.997946383569867e-05, + "loss": 0.9961, + "step": 4043 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979438278124103e-05, + "loss": 1.0582, + "step": 4044 + }, + { + "epoch": 0.3, + "learning_rate": 1.997941270467239e-05, + "loss": 1.1737, + "step": 4045 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979387115343578e-05, + "loss": 1.0333, + "step": 4046 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979361510137702e-05, + "loss": 1.0088, + "step": 4047 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979335889054803e-05, + "loss": 1.003, + "step": 4048 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979310252094926e-05, + "loss": 1.1778, + "step": 4049 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979284599258107e-05, + "loss": 1.0997, + "step": 4050 + }, + { + "epoch": 0.3, + "learning_rate": 1.997925893054439e-05, + "loss": 1.0685, + "step": 4051 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979233245953814e-05, + "loss": 1.1305, + "step": 4052 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979207545486422e-05, + "loss": 1.0819, + "step": 4053 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979181829142254e-05, + "loss": 1.1196, + "step": 4054 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979156096921348e-05, + "loss": 1.0981, + "step": 4055 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979130348823748e-05, + "loss": 1.0199, + "step": 4056 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979104584849498e-05, + "loss": 1.0126, + "step": 4057 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979078804998632e-05, + "loss": 0.9814, + "step": 4058 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979053009271194e-05, + "loss": 0.9846, + "step": 4059 + }, + { + "epoch": 0.3, + "learning_rate": 1.9979027197667227e-05, + "loss": 0.9968, + "step": 4060 + }, + { + "epoch": 0.3, + "learning_rate": 1.997900137018677e-05, + "loss": 1.0246, + "step": 4061 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978975526829866e-05, + "loss": 1.0792, + "step": 4062 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978949667596553e-05, + "loss": 1.0972, + "step": 4063 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978923792486874e-05, + "loss": 0.9295, + "step": 4064 + }, + { + "epoch": 0.3, + "learning_rate": 1.997889790150087e-05, + "loss": 0.9837, + "step": 4065 + }, + { + "epoch": 0.3, + "learning_rate": 1.997887199463858e-05, + "loss": 1.0864, + "step": 4066 + }, + { + "epoch": 0.3, + "learning_rate": 1.997884607190005e-05, + "loss": 1.0369, + "step": 4067 + }, + { + "epoch": 0.3, + "learning_rate": 1.997882013328532e-05, + "loss": 1.0166, + "step": 4068 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978794178794428e-05, + "loss": 1.111, + "step": 4069 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978768208427417e-05, + "loss": 1.054, + "step": 4070 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978742222184332e-05, + "loss": 0.9939, + "step": 4071 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978716220065207e-05, + "loss": 1.0336, + "step": 4072 + }, + { + "epoch": 0.3, + "learning_rate": 1.997869020207009e-05, + "loss": 1.013, + "step": 4073 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978664168199016e-05, + "loss": 1.0773, + "step": 4074 + }, + { + "epoch": 0.3, + "learning_rate": 1.997863811845203e-05, + "loss": 1.0298, + "step": 4075 + }, + { + "epoch": 0.3, + "learning_rate": 1.997861205282918e-05, + "loss": 1.0238, + "step": 4076 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978585971330493e-05, + "loss": 1.0247, + "step": 4077 + }, + { + "epoch": 0.3, + "learning_rate": 1.997855987395602e-05, + "loss": 1.1343, + "step": 4078 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978533760705804e-05, + "loss": 1.0301, + "step": 4079 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978507631579883e-05, + "loss": 1.0296, + "step": 4080 + }, + { + "epoch": 0.3, + "learning_rate": 1.99784814865783e-05, + "loss": 1.1702, + "step": 4081 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978455325701094e-05, + "loss": 1.0038, + "step": 4082 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978429148948312e-05, + "loss": 1.135, + "step": 4083 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978402956319988e-05, + "loss": 1.1117, + "step": 4084 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978376747816166e-05, + "loss": 1.1398, + "step": 4085 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978350523436893e-05, + "loss": 1.0988, + "step": 4086 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978324283182203e-05, + "loss": 1.1219, + "step": 4087 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978298027052148e-05, + "loss": 1.0591, + "step": 4088 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978271755046757e-05, + "loss": 1.0045, + "step": 4089 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978245467166085e-05, + "loss": 1.0993, + "step": 4090 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978219163410162e-05, + "loss": 1.0873, + "step": 4091 + }, + { + "epoch": 0.3, + "learning_rate": 1.997819284377904e-05, + "loss": 0.9669, + "step": 4092 + }, + { + "epoch": 0.3, + "learning_rate": 1.997816650827275e-05, + "loss": 1.1824, + "step": 4093 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978140156891346e-05, + "loss": 1.0308, + "step": 4094 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978113789634857e-05, + "loss": 1.1667, + "step": 4095 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978087406503338e-05, + "loss": 1.0065, + "step": 4096 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978061007496822e-05, + "loss": 1.0601, + "step": 4097 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978034592615353e-05, + "loss": 1.0943, + "step": 4098 + }, + { + "epoch": 0.3, + "learning_rate": 1.9978008161858976e-05, + "loss": 0.9957, + "step": 4099 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977981715227728e-05, + "loss": 1.0159, + "step": 4100 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977955252721656e-05, + "loss": 1.151, + "step": 4101 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977928774340797e-05, + "loss": 0.9591, + "step": 4102 + }, + { + "epoch": 0.3, + "learning_rate": 1.99779022800852e-05, + "loss": 1.1496, + "step": 4103 + }, + { + "epoch": 0.3, + "learning_rate": 1.99778757699549e-05, + "loss": 1.049, + "step": 4104 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977849243949945e-05, + "loss": 1.0928, + "step": 4105 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977822702070374e-05, + "loss": 1.0193, + "step": 4106 + }, + { + "epoch": 0.3, + "learning_rate": 1.997779614431623e-05, + "loss": 1.0361, + "step": 4107 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977769570687553e-05, + "loss": 1.0132, + "step": 4108 + }, + { + "epoch": 0.3, + "learning_rate": 1.997774298118439e-05, + "loss": 1.0322, + "step": 4109 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977716375806778e-05, + "loss": 0.9508, + "step": 4110 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977689754554764e-05, + "loss": 1.1665, + "step": 4111 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977663117428388e-05, + "loss": 1.0671, + "step": 4112 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977636464427692e-05, + "loss": 1.1345, + "step": 4113 + }, + { + "epoch": 0.3, + "learning_rate": 1.997760979555272e-05, + "loss": 0.979, + "step": 4114 + }, + { + "epoch": 0.3, + "learning_rate": 1.997758311080351e-05, + "loss": 1.1134, + "step": 4115 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977556410180116e-05, + "loss": 1.1408, + "step": 4116 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977529693682568e-05, + "loss": 1.099, + "step": 4117 + }, + { + "epoch": 0.3, + "learning_rate": 1.997750296131091e-05, + "loss": 1.1122, + "step": 4118 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977476213065192e-05, + "loss": 1.0828, + "step": 4119 + }, + { + "epoch": 0.3, + "learning_rate": 1.997744944894545e-05, + "loss": 1.0905, + "step": 4120 + }, + { + "epoch": 0.3, + "learning_rate": 1.997742266895173e-05, + "loss": 1.0931, + "step": 4121 + }, + { + "epoch": 0.3, + "learning_rate": 1.997739587308407e-05, + "loss": 1.1548, + "step": 4122 + }, + { + "epoch": 0.3, + "learning_rate": 1.997736906134252e-05, + "loss": 1.036, + "step": 4123 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977342233727115e-05, + "loss": 1.007, + "step": 4124 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977315390237905e-05, + "loss": 1.0631, + "step": 4125 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977288530874928e-05, + "loss": 1.1631, + "step": 4126 + }, + { + "epoch": 0.3, + "learning_rate": 1.9977261655638225e-05, + "loss": 1.0759, + "step": 4127 + }, + { + "epoch": 0.31, + "learning_rate": 1.997723476452784e-05, + "loss": 1.0678, + "step": 4128 + }, + { + "epoch": 0.31, + "learning_rate": 1.9977207857543825e-05, + "loss": 1.0323, + "step": 4129 + }, + { + "epoch": 0.31, + "learning_rate": 1.9977180934686208e-05, + "loss": 1.0334, + "step": 4130 + }, + { + "epoch": 0.31, + "learning_rate": 1.997715399595504e-05, + "loss": 1.0516, + "step": 4131 + }, + { + "epoch": 0.31, + "learning_rate": 1.9977127041350366e-05, + "loss": 1.0338, + "step": 4132 + }, + { + "epoch": 0.31, + "learning_rate": 1.9977100070872225e-05, + "loss": 1.0376, + "step": 4133 + }, + { + "epoch": 0.31, + "learning_rate": 1.997707308452066e-05, + "loss": 1.034, + "step": 4134 + }, + { + "epoch": 0.31, + "learning_rate": 1.9977046082295713e-05, + "loss": 1.1133, + "step": 4135 + }, + { + "epoch": 0.31, + "learning_rate": 1.997701906419743e-05, + "loss": 1.0345, + "step": 4136 + }, + { + "epoch": 0.31, + "learning_rate": 1.997699203022585e-05, + "loss": 1.0471, + "step": 4137 + }, + { + "epoch": 0.31, + "learning_rate": 1.997696498038102e-05, + "loss": 1.0872, + "step": 4138 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976937914662985e-05, + "loss": 1.0299, + "step": 4139 + }, + { + "epoch": 0.31, + "learning_rate": 1.997691083307178e-05, + "loss": 1.0482, + "step": 4140 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976883735607452e-05, + "loss": 1.0947, + "step": 4141 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976856622270046e-05, + "loss": 1.0958, + "step": 4142 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976829493059604e-05, + "loss": 1.0263, + "step": 4143 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976802347976168e-05, + "loss": 1.0832, + "step": 4144 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976775187019786e-05, + "loss": 1.1491, + "step": 4145 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976748010190493e-05, + "loss": 1.1129, + "step": 4146 + }, + { + "epoch": 0.31, + "learning_rate": 1.997672081748834e-05, + "loss": 0.9872, + "step": 4147 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976693608913367e-05, + "loss": 1.0638, + "step": 4148 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976666384465614e-05, + "loss": 1.0673, + "step": 4149 + }, + { + "epoch": 0.31, + "learning_rate": 1.997663914414513e-05, + "loss": 1.05, + "step": 4150 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976611887951955e-05, + "loss": 1.0834, + "step": 4151 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976584615886134e-05, + "loss": 1.1302, + "step": 4152 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976557327947708e-05, + "loss": 1.0249, + "step": 4153 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976530024136722e-05, + "loss": 1.1584, + "step": 4154 + }, + { + "epoch": 0.31, + "learning_rate": 1.997650270445322e-05, + "loss": 1.1089, + "step": 4155 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976475368897247e-05, + "loss": 1.0127, + "step": 4156 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976448017468844e-05, + "loss": 1.0069, + "step": 4157 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976420650168052e-05, + "loss": 0.9862, + "step": 4158 + }, + { + "epoch": 0.31, + "learning_rate": 1.997639326699492e-05, + "loss": 1.0897, + "step": 4159 + }, + { + "epoch": 0.31, + "learning_rate": 1.997636586794949e-05, + "loss": 1.2129, + "step": 4160 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976338453031802e-05, + "loss": 1.0547, + "step": 4161 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976311022241903e-05, + "loss": 1.1474, + "step": 4162 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976283575579837e-05, + "loss": 1.0451, + "step": 4163 + }, + { + "epoch": 0.31, + "learning_rate": 1.997625611304564e-05, + "loss": 1.0707, + "step": 4164 + }, + { + "epoch": 0.31, + "learning_rate": 1.997622863463937e-05, + "loss": 1.0861, + "step": 4165 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976201140361064e-05, + "loss": 1.0538, + "step": 4166 + }, + { + "epoch": 0.31, + "learning_rate": 1.997617363021076e-05, + "loss": 1.1458, + "step": 4167 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976146104188508e-05, + "loss": 0.9987, + "step": 4168 + }, + { + "epoch": 0.31, + "learning_rate": 1.997611856229435e-05, + "loss": 1.058, + "step": 4169 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976091004528332e-05, + "loss": 0.9903, + "step": 4170 + }, + { + "epoch": 0.31, + "learning_rate": 1.9976063430890494e-05, + "loss": 1.0184, + "step": 4171 + }, + { + "epoch": 0.31, + "learning_rate": 1.997603584138088e-05, + "loss": 0.9945, + "step": 4172 + }, + { + "epoch": 0.31, + "learning_rate": 1.997600823599954e-05, + "loss": 1.1219, + "step": 4173 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975980614746513e-05, + "loss": 1.0292, + "step": 4174 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975952977621838e-05, + "loss": 1.1324, + "step": 4175 + }, + { + "epoch": 0.31, + "learning_rate": 1.997592532462557e-05, + "loss": 1.0947, + "step": 4176 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975897655757747e-05, + "loss": 0.9959, + "step": 4177 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975869971018413e-05, + "loss": 1.0035, + "step": 4178 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975842270407612e-05, + "loss": 1.0301, + "step": 4179 + }, + { + "epoch": 0.31, + "learning_rate": 1.997581455392539e-05, + "loss": 1.0822, + "step": 4180 + }, + { + "epoch": 0.31, + "learning_rate": 1.997578682157179e-05, + "loss": 1.0059, + "step": 4181 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975759073346854e-05, + "loss": 1.1489, + "step": 4182 + }, + { + "epoch": 0.31, + "learning_rate": 1.997573130925063e-05, + "loss": 1.0083, + "step": 4183 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975703529283158e-05, + "loss": 0.9496, + "step": 4184 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975675733444485e-05, + "loss": 0.8621, + "step": 4185 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975647921734654e-05, + "loss": 0.9874, + "step": 4186 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975620094153713e-05, + "loss": 1.0725, + "step": 4187 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975592250701702e-05, + "loss": 1.0839, + "step": 4188 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975564391378662e-05, + "loss": 1.1075, + "step": 4189 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975536516184647e-05, + "loss": 1.0734, + "step": 4190 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975508625119694e-05, + "loss": 1.1267, + "step": 4191 + }, + { + "epoch": 0.31, + "learning_rate": 1.997548071818385e-05, + "loss": 0.9481, + "step": 4192 + }, + { + "epoch": 0.31, + "learning_rate": 1.997545279537716e-05, + "loss": 1.0287, + "step": 4193 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975424856699664e-05, + "loss": 1.0901, + "step": 4194 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975396902151415e-05, + "loss": 0.9538, + "step": 4195 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975368931732448e-05, + "loss": 1.2056, + "step": 4196 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975340945442815e-05, + "loss": 1.1027, + "step": 4197 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975312943282552e-05, + "loss": 1.1025, + "step": 4198 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975284925251715e-05, + "loss": 0.9066, + "step": 4199 + }, + { + "epoch": 0.31, + "learning_rate": 1.997525689135034e-05, + "loss": 0.9611, + "step": 4200 + }, + { + "epoch": 0.31, + "learning_rate": 1.997522884157847e-05, + "loss": 1.0096, + "step": 4201 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975200775936157e-05, + "loss": 0.928, + "step": 4202 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975172694423443e-05, + "loss": 1.077, + "step": 4203 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975144597040373e-05, + "loss": 1.0098, + "step": 4204 + }, + { + "epoch": 0.31, + "learning_rate": 1.997511648378699e-05, + "loss": 0.9749, + "step": 4205 + }, + { + "epoch": 0.31, + "learning_rate": 1.997508835466334e-05, + "loss": 1.0619, + "step": 4206 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975060209669462e-05, + "loss": 1.1032, + "step": 4207 + }, + { + "epoch": 0.31, + "learning_rate": 1.997503204880541e-05, + "loss": 1.0312, + "step": 4208 + }, + { + "epoch": 0.31, + "learning_rate": 1.9975003872071227e-05, + "loss": 1.1018, + "step": 4209 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974975679466953e-05, + "loss": 1.0984, + "step": 4210 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974947470992632e-05, + "loss": 1.1615, + "step": 4211 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974919246648316e-05, + "loss": 1.0011, + "step": 4212 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974891006434048e-05, + "loss": 1.0701, + "step": 4213 + }, + { + "epoch": 0.31, + "learning_rate": 1.997486275034987e-05, + "loss": 1.0882, + "step": 4214 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974834478395823e-05, + "loss": 1.1221, + "step": 4215 + }, + { + "epoch": 0.31, + "learning_rate": 1.997480619057196e-05, + "loss": 1.087, + "step": 4216 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974777886878323e-05, + "loss": 1.0988, + "step": 4217 + }, + { + "epoch": 0.31, + "learning_rate": 1.997474956731496e-05, + "loss": 1.0447, + "step": 4218 + }, + { + "epoch": 0.31, + "learning_rate": 1.997472123188191e-05, + "loss": 1.1883, + "step": 4219 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974692880579225e-05, + "loss": 0.9465, + "step": 4220 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974664513406946e-05, + "loss": 1.0965, + "step": 4221 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974636130365115e-05, + "loss": 1.1327, + "step": 4222 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974607731453783e-05, + "loss": 0.9239, + "step": 4223 + }, + { + "epoch": 0.31, + "learning_rate": 1.997457931667299e-05, + "loss": 0.9078, + "step": 4224 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974550886022785e-05, + "loss": 1.0385, + "step": 4225 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974522439503214e-05, + "loss": 1.0049, + "step": 4226 + }, + { + "epoch": 0.31, + "learning_rate": 1.997449397711432e-05, + "loss": 0.9914, + "step": 4227 + }, + { + "epoch": 0.31, + "learning_rate": 1.997446549885615e-05, + "loss": 1.0035, + "step": 4228 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974437004728745e-05, + "loss": 1.0685, + "step": 4229 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974408494732156e-05, + "loss": 1.1461, + "step": 4230 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974379968866425e-05, + "loss": 1.0522, + "step": 4231 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974351427131596e-05, + "loss": 0.9962, + "step": 4232 + }, + { + "epoch": 0.31, + "learning_rate": 1.997432286952772e-05, + "loss": 1.0687, + "step": 4233 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974294296054837e-05, + "loss": 1.06, + "step": 4234 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974265706712995e-05, + "loss": 1.127, + "step": 4235 + }, + { + "epoch": 0.31, + "learning_rate": 1.997423710150224e-05, + "loss": 0.9923, + "step": 4236 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974208480422615e-05, + "loss": 1.023, + "step": 4237 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974179843474167e-05, + "loss": 1.044, + "step": 4238 + }, + { + "epoch": 0.31, + "learning_rate": 1.997415119065694e-05, + "loss": 1.0053, + "step": 4239 + }, + { + "epoch": 0.31, + "learning_rate": 1.997412252197099e-05, + "loss": 1.0539, + "step": 4240 + }, + { + "epoch": 0.31, + "learning_rate": 1.997409383741634e-05, + "loss": 1.1003, + "step": 4241 + }, + { + "epoch": 0.31, + "learning_rate": 1.9974065136993058e-05, + "loss": 1.009, + "step": 4242 + }, + { + "epoch": 0.31, + "learning_rate": 1.997403642070118e-05, + "loss": 1.0917, + "step": 4243 + }, + { + "epoch": 0.31, + "learning_rate": 1.997400768854075e-05, + "loss": 1.0447, + "step": 4244 + }, + { + "epoch": 0.31, + "learning_rate": 1.997397894051182e-05, + "loss": 1.0374, + "step": 4245 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973950176614432e-05, + "loss": 1.0507, + "step": 4246 + }, + { + "epoch": 0.31, + "learning_rate": 1.997392139684863e-05, + "loss": 1.0394, + "step": 4247 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973892601214465e-05, + "loss": 1.0844, + "step": 4248 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973863789711975e-05, + "loss": 0.9189, + "step": 4249 + }, + { + "epoch": 0.31, + "learning_rate": 1.997383496234121e-05, + "loss": 1.0624, + "step": 4250 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973806119102222e-05, + "loss": 1.0617, + "step": 4251 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973777259995048e-05, + "loss": 1.0896, + "step": 4252 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973748385019737e-05, + "loss": 1.0842, + "step": 4253 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973719494176335e-05, + "loss": 1.1403, + "step": 4254 + }, + { + "epoch": 0.31, + "learning_rate": 1.997369058746489e-05, + "loss": 1.1036, + "step": 4255 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973661664885446e-05, + "loss": 1.0949, + "step": 4256 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973632726438043e-05, + "loss": 0.9837, + "step": 4257 + }, + { + "epoch": 0.31, + "learning_rate": 1.997360377212274e-05, + "loss": 1.1426, + "step": 4258 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973574801939573e-05, + "loss": 1.0205, + "step": 4259 + }, + { + "epoch": 0.31, + "learning_rate": 1.997354581588859e-05, + "loss": 1.0493, + "step": 4260 + }, + { + "epoch": 0.31, + "learning_rate": 1.997351681396984e-05, + "loss": 1.0368, + "step": 4261 + }, + { + "epoch": 0.31, + "learning_rate": 1.997348779618337e-05, + "loss": 1.0463, + "step": 4262 + }, + { + "epoch": 0.31, + "learning_rate": 1.9973458762529216e-05, + "loss": 1.152, + "step": 4263 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973429713007436e-05, + "loss": 0.9724, + "step": 4264 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973400647618075e-05, + "loss": 0.9477, + "step": 4265 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973371566361175e-05, + "loss": 1.0167, + "step": 4266 + }, + { + "epoch": 0.32, + "learning_rate": 1.997334246923678e-05, + "loss": 0.9446, + "step": 4267 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973313356244944e-05, + "loss": 1.0309, + "step": 4268 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973284227385707e-05, + "loss": 1.1039, + "step": 4269 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973255082659117e-05, + "loss": 1.0816, + "step": 4270 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973225922065222e-05, + "loss": 1.0277, + "step": 4271 + }, + { + "epoch": 0.32, + "learning_rate": 1.997319674560407e-05, + "loss": 1.1397, + "step": 4272 + }, + { + "epoch": 0.32, + "learning_rate": 1.99731675532757e-05, + "loss": 0.9567, + "step": 4273 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973138345080163e-05, + "loss": 0.9913, + "step": 4274 + }, + { + "epoch": 0.32, + "learning_rate": 1.997310912101751e-05, + "loss": 1.1165, + "step": 4275 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973079881087778e-05, + "loss": 1.118, + "step": 4276 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973050625291025e-05, + "loss": 0.9844, + "step": 4277 + }, + { + "epoch": 0.32, + "learning_rate": 1.9973021353627285e-05, + "loss": 1.154, + "step": 4278 + }, + { + "epoch": 0.32, + "learning_rate": 1.997299206609661e-05, + "loss": 1.0884, + "step": 4279 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972962762699053e-05, + "loss": 0.9962, + "step": 4280 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972933443434653e-05, + "loss": 1.0274, + "step": 4281 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972904108303456e-05, + "loss": 1.0506, + "step": 4282 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972874757305513e-05, + "loss": 1.0507, + "step": 4283 + }, + { + "epoch": 0.32, + "learning_rate": 1.997284539044087e-05, + "loss": 1.0088, + "step": 4284 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972816007709572e-05, + "loss": 1.1158, + "step": 4285 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972786609111667e-05, + "loss": 1.1109, + "step": 4286 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972757194647198e-05, + "loss": 0.9757, + "step": 4287 + }, + { + "epoch": 0.32, + "learning_rate": 1.997272776431622e-05, + "loss": 1.066, + "step": 4288 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972698318118772e-05, + "loss": 1.007, + "step": 4289 + }, + { + "epoch": 0.32, + "learning_rate": 1.99726688560549e-05, + "loss": 1.0263, + "step": 4290 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972639378124658e-05, + "loss": 1.0154, + "step": 4291 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972609884328093e-05, + "loss": 1.0174, + "step": 4292 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972580374665242e-05, + "loss": 1.0665, + "step": 4293 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972550849136163e-05, + "loss": 0.9219, + "step": 4294 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972521307740896e-05, + "loss": 1.0978, + "step": 4295 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972491750479493e-05, + "loss": 1.0, + "step": 4296 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972462177351994e-05, + "loss": 0.9085, + "step": 4297 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972432588358456e-05, + "loss": 1.0393, + "step": 4298 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972402983498914e-05, + "loss": 0.9649, + "step": 4299 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972373362773425e-05, + "loss": 0.9592, + "step": 4300 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972343726182033e-05, + "loss": 1.1129, + "step": 4301 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972314073724782e-05, + "loss": 1.0016, + "step": 4302 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972284405401723e-05, + "loss": 0.9335, + "step": 4303 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972254721212904e-05, + "loss": 1.1513, + "step": 4304 + }, + { + "epoch": 0.32, + "learning_rate": 1.997222502115837e-05, + "loss": 1.115, + "step": 4305 + }, + { + "epoch": 0.32, + "learning_rate": 1.997219530523817e-05, + "loss": 1.0789, + "step": 4306 + }, + { + "epoch": 0.32, + "learning_rate": 1.997216557345235e-05, + "loss": 1.0367, + "step": 4307 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972135825800952e-05, + "loss": 1.1418, + "step": 4308 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972106062284035e-05, + "loss": 1.0786, + "step": 4309 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972076282901634e-05, + "loss": 1.0898, + "step": 4310 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972046487653805e-05, + "loss": 1.0288, + "step": 4311 + }, + { + "epoch": 0.32, + "learning_rate": 1.9972016676540592e-05, + "loss": 1.0302, + "step": 4312 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971986849562045e-05, + "loss": 0.8877, + "step": 4313 + }, + { + "epoch": 0.32, + "learning_rate": 1.997195700671821e-05, + "loss": 1.1065, + "step": 4314 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971927148009133e-05, + "loss": 1.108, + "step": 4315 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971897273434862e-05, + "loss": 0.9619, + "step": 4316 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971867382995445e-05, + "loss": 1.0747, + "step": 4317 + }, + { + "epoch": 0.32, + "learning_rate": 1.997183747669093e-05, + "loss": 0.9756, + "step": 4318 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971807554521364e-05, + "loss": 1.1359, + "step": 4319 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971777616486796e-05, + "loss": 1.1716, + "step": 4320 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971747662587273e-05, + "loss": 1.0121, + "step": 4321 + }, + { + "epoch": 0.32, + "learning_rate": 1.997171769282284e-05, + "loss": 1.1294, + "step": 4322 + }, + { + "epoch": 0.32, + "learning_rate": 1.997168770719355e-05, + "loss": 1.0281, + "step": 4323 + }, + { + "epoch": 0.32, + "learning_rate": 1.997165770569944e-05, + "loss": 0.9706, + "step": 4324 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971627688340573e-05, + "loss": 0.9143, + "step": 4325 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971597655116988e-05, + "loss": 1.007, + "step": 4326 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971567606028733e-05, + "loss": 1.1411, + "step": 4327 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971537541075852e-05, + "loss": 1.0532, + "step": 4328 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971507460258404e-05, + "loss": 0.9881, + "step": 4329 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971477363576428e-05, + "loss": 1.1127, + "step": 4330 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971447251029973e-05, + "loss": 1.004, + "step": 4331 + }, + { + "epoch": 0.32, + "learning_rate": 1.997141712261909e-05, + "loss": 1.0329, + "step": 4332 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971386978343825e-05, + "loss": 1.1313, + "step": 4333 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971356818204224e-05, + "loss": 1.0884, + "step": 4334 + }, + { + "epoch": 0.32, + "learning_rate": 1.997132664220034e-05, + "loss": 0.9552, + "step": 4335 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971296450332214e-05, + "loss": 1.064, + "step": 4336 + }, + { + "epoch": 0.32, + "learning_rate": 1.99712662425999e-05, + "loss": 1.0709, + "step": 4337 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971236019003443e-05, + "loss": 1.0812, + "step": 4338 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971205779542893e-05, + "loss": 1.1313, + "step": 4339 + }, + { + "epoch": 0.32, + "learning_rate": 1.99711755242183e-05, + "loss": 0.9937, + "step": 4340 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971145253029706e-05, + "loss": 1.0384, + "step": 4341 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971114965977164e-05, + "loss": 1.112, + "step": 4342 + }, + { + "epoch": 0.32, + "learning_rate": 1.997108466306072e-05, + "loss": 1.0348, + "step": 4343 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971054344280424e-05, + "loss": 1.1348, + "step": 4344 + }, + { + "epoch": 0.32, + "learning_rate": 1.9971024009636323e-05, + "loss": 1.019, + "step": 4345 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970993659128463e-05, + "loss": 1.0422, + "step": 4346 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970963292756895e-05, + "loss": 0.9561, + "step": 4347 + }, + { + "epoch": 0.32, + "learning_rate": 1.997093291052167e-05, + "loss": 1.0875, + "step": 4348 + }, + { + "epoch": 0.32, + "learning_rate": 1.997090251242283e-05, + "loss": 0.9761, + "step": 4349 + }, + { + "epoch": 0.32, + "learning_rate": 1.997087209846043e-05, + "loss": 1.0561, + "step": 4350 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970841668634514e-05, + "loss": 1.0762, + "step": 4351 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970811222945132e-05, + "loss": 1.1037, + "step": 4352 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970780761392333e-05, + "loss": 1.065, + "step": 4353 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970750283976163e-05, + "loss": 1.0785, + "step": 4354 + }, + { + "epoch": 0.32, + "learning_rate": 1.997071979069667e-05, + "loss": 1.0411, + "step": 4355 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970689281553906e-05, + "loss": 1.0036, + "step": 4356 + }, + { + "epoch": 0.32, + "learning_rate": 1.997065875654792e-05, + "loss": 1.0478, + "step": 4357 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970628215678757e-05, + "loss": 1.1139, + "step": 4358 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970597658946466e-05, + "loss": 1.0951, + "step": 4359 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970567086351097e-05, + "loss": 1.0291, + "step": 4360 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970536497892702e-05, + "loss": 1.0136, + "step": 4361 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970505893571324e-05, + "loss": 1.0676, + "step": 4362 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970475273387012e-05, + "loss": 1.0388, + "step": 4363 + }, + { + "epoch": 0.32, + "learning_rate": 1.997044463733982e-05, + "loss": 1.0796, + "step": 4364 + }, + { + "epoch": 0.32, + "learning_rate": 1.997041398542979e-05, + "loss": 1.0964, + "step": 4365 + }, + { + "epoch": 0.32, + "learning_rate": 1.997038331765697e-05, + "loss": 1.0869, + "step": 4366 + }, + { + "epoch": 0.32, + "learning_rate": 1.997035263402142e-05, + "loss": 1.0183, + "step": 4367 + }, + { + "epoch": 0.32, + "learning_rate": 1.997032193452318e-05, + "loss": 1.1023, + "step": 4368 + }, + { + "epoch": 0.32, + "learning_rate": 1.99702912191623e-05, + "loss": 0.9892, + "step": 4369 + }, + { + "epoch": 0.32, + "learning_rate": 1.997026048793883e-05, + "loss": 1.0726, + "step": 4370 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970229740852817e-05, + "loss": 1.114, + "step": 4371 + }, + { + "epoch": 0.32, + "learning_rate": 1.997019897790431e-05, + "loss": 1.0493, + "step": 4372 + }, + { + "epoch": 0.32, + "learning_rate": 1.997016819909336e-05, + "loss": 1.0446, + "step": 4373 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970137404420016e-05, + "loss": 1.0919, + "step": 4374 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970106593884327e-05, + "loss": 1.0388, + "step": 4375 + }, + { + "epoch": 0.32, + "learning_rate": 1.997007576748634e-05, + "loss": 0.9822, + "step": 4376 + }, + { + "epoch": 0.32, + "learning_rate": 1.9970044925226103e-05, + "loss": 1.0838, + "step": 4377 + }, + { + "epoch": 0.32, + "learning_rate": 1.997001406710367e-05, + "loss": 1.0677, + "step": 4378 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969983193119087e-05, + "loss": 1.1416, + "step": 4379 + }, + { + "epoch": 0.32, + "learning_rate": 1.99699523032724e-05, + "loss": 1.0182, + "step": 4380 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969921397563666e-05, + "loss": 1.121, + "step": 4381 + }, + { + "epoch": 0.32, + "learning_rate": 1.996989047599293e-05, + "loss": 1.069, + "step": 4382 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969859538560236e-05, + "loss": 1.0412, + "step": 4383 + }, + { + "epoch": 0.32, + "learning_rate": 1.996982858526564e-05, + "loss": 1.0529, + "step": 4384 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969797616109192e-05, + "loss": 1.0286, + "step": 4385 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969766631090937e-05, + "loss": 0.9796, + "step": 4386 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969735630210926e-05, + "loss": 0.973, + "step": 4387 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969704613469206e-05, + "loss": 1.0358, + "step": 4388 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969673580865833e-05, + "loss": 1.061, + "step": 4389 + }, + { + "epoch": 0.32, + "learning_rate": 1.996964253240085e-05, + "loss": 1.0896, + "step": 4390 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969611468074306e-05, + "loss": 1.0584, + "step": 4391 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969580387886258e-05, + "loss": 1.0439, + "step": 4392 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969549291836745e-05, + "loss": 1.0404, + "step": 4393 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969518179925826e-05, + "loss": 1.1969, + "step": 4394 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969487052153543e-05, + "loss": 1.1359, + "step": 4395 + }, + { + "epoch": 0.32, + "learning_rate": 1.996945590851995e-05, + "loss": 1.0467, + "step": 4396 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969424749025094e-05, + "loss": 0.987, + "step": 4397 + }, + { + "epoch": 0.32, + "learning_rate": 1.9969393573669028e-05, + "loss": 1.1113, + "step": 4398 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969362382451796e-05, + "loss": 1.0682, + "step": 4399 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969331175373456e-05, + "loss": 1.0699, + "step": 4400 + }, + { + "epoch": 0.33, + "learning_rate": 1.996929995243405e-05, + "loss": 1.1339, + "step": 4401 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969268713633628e-05, + "loss": 1.0399, + "step": 4402 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969237458972246e-05, + "loss": 0.9755, + "step": 4403 + }, + { + "epoch": 0.33, + "learning_rate": 1.996920618844995e-05, + "loss": 1.1399, + "step": 4404 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969174902066786e-05, + "loss": 1.0449, + "step": 4405 + }, + { + "epoch": 0.33, + "learning_rate": 1.996914359982281e-05, + "loss": 0.9188, + "step": 4406 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969112281718067e-05, + "loss": 1.0071, + "step": 4407 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969080947752612e-05, + "loss": 1.0865, + "step": 4408 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969049597926492e-05, + "loss": 0.9799, + "step": 4409 + }, + { + "epoch": 0.33, + "learning_rate": 1.9969018232239755e-05, + "loss": 1.1286, + "step": 4410 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968986850692455e-05, + "loss": 1.2012, + "step": 4411 + }, + { + "epoch": 0.33, + "learning_rate": 1.996895545328463e-05, + "loss": 1.0984, + "step": 4412 + }, + { + "epoch": 0.33, + "learning_rate": 1.996892404001635e-05, + "loss": 1.1342, + "step": 4413 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968892610887652e-05, + "loss": 1.0493, + "step": 4414 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968861165898587e-05, + "loss": 1.0826, + "step": 4415 + }, + { + "epoch": 0.33, + "learning_rate": 1.996882970504921e-05, + "loss": 0.9882, + "step": 4416 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968798228339563e-05, + "loss": 1.0736, + "step": 4417 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968766735769703e-05, + "loss": 1.1076, + "step": 4418 + }, + { + "epoch": 0.33, + "learning_rate": 1.996873522733968e-05, + "loss": 1.1227, + "step": 4419 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968703703049536e-05, + "loss": 1.0593, + "step": 4420 + }, + { + "epoch": 0.33, + "learning_rate": 1.996867216289933e-05, + "loss": 0.9648, + "step": 4421 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968640606889108e-05, + "loss": 1.1283, + "step": 4422 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968609035018926e-05, + "loss": 1.0851, + "step": 4423 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968577447288827e-05, + "loss": 1.0691, + "step": 4424 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968545843698863e-05, + "loss": 1.0624, + "step": 4425 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968514224249085e-05, + "loss": 1.0403, + "step": 4426 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968482588939543e-05, + "loss": 1.0836, + "step": 4427 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968450937770288e-05, + "loss": 1.2031, + "step": 4428 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968419270741375e-05, + "loss": 0.9483, + "step": 4429 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968387587852844e-05, + "loss": 0.9942, + "step": 4430 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968355889104752e-05, + "loss": 1.0144, + "step": 4431 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968324174497148e-05, + "loss": 1.0755, + "step": 4432 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968292444030085e-05, + "loss": 1.0414, + "step": 4433 + }, + { + "epoch": 0.33, + "learning_rate": 1.996826069770361e-05, + "loss": 0.8397, + "step": 4434 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968228935517776e-05, + "loss": 1.0515, + "step": 4435 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968197157472632e-05, + "loss": 0.9923, + "step": 4436 + }, + { + "epoch": 0.33, + "learning_rate": 1.996816536356823e-05, + "loss": 1.0515, + "step": 4437 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968133553804615e-05, + "loss": 1.0323, + "step": 4438 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968101728181847e-05, + "loss": 1.085, + "step": 4439 + }, + { + "epoch": 0.33, + "learning_rate": 1.996806988669997e-05, + "loss": 1.0274, + "step": 4440 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968038029359036e-05, + "loss": 1.0195, + "step": 4441 + }, + { + "epoch": 0.33, + "learning_rate": 1.9968006156159098e-05, + "loss": 1.0359, + "step": 4442 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967974267100203e-05, + "loss": 1.0877, + "step": 4443 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967942362182402e-05, + "loss": 1.089, + "step": 4444 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967910441405748e-05, + "loss": 1.1266, + "step": 4445 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967878504770297e-05, + "loss": 1.0456, + "step": 4446 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967846552276088e-05, + "loss": 0.9121, + "step": 4447 + }, + { + "epoch": 0.33, + "learning_rate": 1.996781458392318e-05, + "loss": 1.0992, + "step": 4448 + }, + { + "epoch": 0.33, + "learning_rate": 1.996778259971162e-05, + "loss": 1.0416, + "step": 4449 + }, + { + "epoch": 0.33, + "learning_rate": 1.996775059964146e-05, + "loss": 1.0459, + "step": 4450 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967718583712753e-05, + "loss": 0.9453, + "step": 4451 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967686551925547e-05, + "loss": 1.0441, + "step": 4452 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967654504279896e-05, + "loss": 1.0836, + "step": 4453 + }, + { + "epoch": 0.33, + "learning_rate": 1.996762244077585e-05, + "loss": 1.1328, + "step": 4454 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967590361413456e-05, + "loss": 1.1508, + "step": 4455 + }, + { + "epoch": 0.33, + "learning_rate": 1.996755826619277e-05, + "loss": 0.9742, + "step": 4456 + }, + { + "epoch": 0.33, + "learning_rate": 1.996752615511384e-05, + "loss": 1.0813, + "step": 4457 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967494028176722e-05, + "loss": 1.0078, + "step": 4458 + }, + { + "epoch": 0.33, + "learning_rate": 1.996746188538146e-05, + "loss": 1.1633, + "step": 4459 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967429726728113e-05, + "loss": 1.0982, + "step": 4460 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967397552216723e-05, + "loss": 1.0984, + "step": 4461 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967365361847348e-05, + "loss": 0.9973, + "step": 4462 + }, + { + "epoch": 0.33, + "learning_rate": 1.996733315562004e-05, + "loss": 1.0472, + "step": 4463 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967300933534843e-05, + "loss": 1.0769, + "step": 4464 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967268695591815e-05, + "loss": 1.0752, + "step": 4465 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967236441791005e-05, + "loss": 1.0734, + "step": 4466 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967204172132465e-05, + "loss": 1.0969, + "step": 4467 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967171886616245e-05, + "loss": 0.9745, + "step": 4468 + }, + { + "epoch": 0.33, + "learning_rate": 1.99671395852424e-05, + "loss": 0.9942, + "step": 4469 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967107268010975e-05, + "loss": 1.0646, + "step": 4470 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967074934922027e-05, + "loss": 1.0368, + "step": 4471 + }, + { + "epoch": 0.33, + "learning_rate": 1.9967042585975603e-05, + "loss": 0.9521, + "step": 4472 + }, + { + "epoch": 0.33, + "learning_rate": 1.996701022117176e-05, + "loss": 1.1759, + "step": 4473 + }, + { + "epoch": 0.33, + "learning_rate": 1.996697784051054e-05, + "loss": 1.0895, + "step": 4474 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966945443992008e-05, + "loss": 1.0903, + "step": 4475 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966913031616205e-05, + "loss": 1.0785, + "step": 4476 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966880603383187e-05, + "loss": 0.9944, + "step": 4477 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966848159293004e-05, + "loss": 1.1602, + "step": 4478 + }, + { + "epoch": 0.33, + "learning_rate": 1.996681569934571e-05, + "loss": 1.0969, + "step": 4479 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966783223541353e-05, + "loss": 1.1239, + "step": 4480 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966750731879987e-05, + "loss": 1.0443, + "step": 4481 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966718224361663e-05, + "loss": 1.1005, + "step": 4482 + }, + { + "epoch": 0.33, + "learning_rate": 1.996668570098643e-05, + "loss": 1.0527, + "step": 4483 + }, + { + "epoch": 0.33, + "learning_rate": 1.996665316175435e-05, + "loss": 0.9922, + "step": 4484 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966620606665462e-05, + "loss": 1.0977, + "step": 4485 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966588035719823e-05, + "loss": 1.0695, + "step": 4486 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966555448917487e-05, + "loss": 1.0898, + "step": 4487 + }, + { + "epoch": 0.33, + "learning_rate": 1.99665228462585e-05, + "loss": 1.0008, + "step": 4488 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966490227742926e-05, + "loss": 0.9153, + "step": 4489 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966457593370803e-05, + "loss": 1.0531, + "step": 4490 + }, + { + "epoch": 0.33, + "learning_rate": 1.996642494314219e-05, + "loss": 1.1193, + "step": 4491 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966392277057136e-05, + "loss": 0.9816, + "step": 4492 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966359595115693e-05, + "loss": 1.1002, + "step": 4493 + }, + { + "epoch": 0.33, + "learning_rate": 1.996632689731792e-05, + "loss": 1.0192, + "step": 4494 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966294183663857e-05, + "loss": 1.0341, + "step": 4495 + }, + { + "epoch": 0.33, + "learning_rate": 1.996626145415357e-05, + "loss": 1.0982, + "step": 4496 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966228708787096e-05, + "loss": 1.0638, + "step": 4497 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966195947564498e-05, + "loss": 1.0972, + "step": 4498 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966163170485825e-05, + "loss": 0.9659, + "step": 4499 + }, + { + "epoch": 0.33, + "learning_rate": 1.996613037755113e-05, + "loss": 1.1797, + "step": 4500 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966097568760463e-05, + "loss": 1.1271, + "step": 4501 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966064744113877e-05, + "loss": 1.0298, + "step": 4502 + }, + { + "epoch": 0.33, + "learning_rate": 1.9966031903611424e-05, + "loss": 1.0364, + "step": 4503 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965999047253158e-05, + "loss": 0.9831, + "step": 4504 + }, + { + "epoch": 0.33, + "learning_rate": 1.996596617503913e-05, + "loss": 1.053, + "step": 4505 + }, + { + "epoch": 0.33, + "learning_rate": 1.996593328696939e-05, + "loss": 1.1023, + "step": 4506 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965900383043997e-05, + "loss": 1.0557, + "step": 4507 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965867463262998e-05, + "loss": 1.023, + "step": 4508 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965834527626443e-05, + "loss": 1.0737, + "step": 4509 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965801576134393e-05, + "loss": 1.0541, + "step": 4510 + }, + { + "epoch": 0.33, + "learning_rate": 1.996576860878689e-05, + "loss": 1.0631, + "step": 4511 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965735625583996e-05, + "loss": 0.9893, + "step": 4512 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965702626525754e-05, + "loss": 1.0002, + "step": 4513 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965669611612228e-05, + "loss": 1.201, + "step": 4514 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965636580843464e-05, + "loss": 1.1502, + "step": 4515 + }, + { + "epoch": 0.33, + "learning_rate": 1.996560353421951e-05, + "loss": 1.0579, + "step": 4516 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965570471740426e-05, + "loss": 1.0231, + "step": 4517 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965537393406263e-05, + "loss": 1.0299, + "step": 4518 + }, + { + "epoch": 0.33, + "learning_rate": 1.996550429921707e-05, + "loss": 1.114, + "step": 4519 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965471189172904e-05, + "loss": 1.0257, + "step": 4520 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965438063273813e-05, + "loss": 1.1594, + "step": 4521 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965404921519857e-05, + "loss": 1.1367, + "step": 4522 + }, + { + "epoch": 0.33, + "learning_rate": 1.996537176391108e-05, + "loss": 1.0868, + "step": 4523 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965338590447546e-05, + "loss": 1.128, + "step": 4524 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965305401129295e-05, + "loss": 1.1041, + "step": 4525 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965272195956387e-05, + "loss": 1.0172, + "step": 4526 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965238974928875e-05, + "loss": 0.9854, + "step": 4527 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965205738046807e-05, + "loss": 1.1227, + "step": 4528 + }, + { + "epoch": 0.33, + "learning_rate": 1.996517248531024e-05, + "loss": 1.1509, + "step": 4529 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965139216719227e-05, + "loss": 0.8982, + "step": 4530 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965105932273823e-05, + "loss": 1.0355, + "step": 4531 + }, + { + "epoch": 0.33, + "learning_rate": 1.9965072631974073e-05, + "loss": 1.0626, + "step": 4532 + }, + { + "epoch": 0.33, + "learning_rate": 1.996503931582004e-05, + "loss": 1.119, + "step": 4533 + }, + { + "epoch": 0.34, + "learning_rate": 1.996500598381177e-05, + "loss": 1.0945, + "step": 4534 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964972635949314e-05, + "loss": 1.0413, + "step": 4535 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964939272232736e-05, + "loss": 1.05, + "step": 4536 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964905892662076e-05, + "loss": 1.0764, + "step": 4537 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964872497237396e-05, + "loss": 1.0804, + "step": 4538 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964839085958746e-05, + "loss": 1.1286, + "step": 4539 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964805658826183e-05, + "loss": 1.0327, + "step": 4540 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964772215839752e-05, + "loss": 1.205, + "step": 4541 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964738756999514e-05, + "loss": 1.0411, + "step": 4542 + }, + { + "epoch": 0.34, + "learning_rate": 1.996470528230552e-05, + "loss": 1.0718, + "step": 4543 + }, + { + "epoch": 0.34, + "learning_rate": 1.996467179175782e-05, + "loss": 1.0887, + "step": 4544 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964638285356468e-05, + "loss": 1.1359, + "step": 4545 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964604763101523e-05, + "loss": 1.1456, + "step": 4546 + }, + { + "epoch": 0.34, + "learning_rate": 1.996457122499303e-05, + "loss": 0.9884, + "step": 4547 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964537671031054e-05, + "loss": 1.1559, + "step": 4548 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964504101215632e-05, + "loss": 1.0785, + "step": 4549 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964470515546833e-05, + "loss": 1.0484, + "step": 4550 + }, + { + "epoch": 0.34, + "learning_rate": 1.99644369140247e-05, + "loss": 1.0014, + "step": 4551 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964403296649295e-05, + "loss": 1.0719, + "step": 4552 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964369663420667e-05, + "loss": 1.1361, + "step": 4553 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964336014338865e-05, + "loss": 1.1151, + "step": 4554 + }, + { + "epoch": 0.34, + "learning_rate": 1.996430234940395e-05, + "loss": 1.1395, + "step": 4555 + }, + { + "epoch": 0.34, + "learning_rate": 1.996426866861597e-05, + "loss": 1.0656, + "step": 4556 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964234971974987e-05, + "loss": 0.9584, + "step": 4557 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964201259481043e-05, + "loss": 1.1137, + "step": 4558 + }, + { + "epoch": 0.34, + "learning_rate": 1.99641675311342e-05, + "loss": 1.077, + "step": 4559 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964133786934504e-05, + "loss": 1.0916, + "step": 4560 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964100026882018e-05, + "loss": 1.1113, + "step": 4561 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964066250976792e-05, + "loss": 1.1038, + "step": 4562 + }, + { + "epoch": 0.34, + "learning_rate": 1.9964032459218877e-05, + "loss": 1.0544, + "step": 4563 + }, + { + "epoch": 0.34, + "learning_rate": 1.996399865160833e-05, + "loss": 0.9952, + "step": 4564 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963964828145202e-05, + "loss": 1.0843, + "step": 4565 + }, + { + "epoch": 0.34, + "learning_rate": 1.996393098882955e-05, + "loss": 1.0829, + "step": 4566 + }, + { + "epoch": 0.34, + "learning_rate": 1.996389713366143e-05, + "loss": 1.1955, + "step": 4567 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963863262640884e-05, + "loss": 1.0034, + "step": 4568 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963829375767982e-05, + "loss": 1.0446, + "step": 4569 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963795473042768e-05, + "loss": 1.101, + "step": 4570 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963761554465294e-05, + "loss": 1.0551, + "step": 4571 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963727620035623e-05, + "loss": 1.1155, + "step": 4572 + }, + { + "epoch": 0.34, + "learning_rate": 1.99636936697538e-05, + "loss": 1.0212, + "step": 4573 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963659703619886e-05, + "loss": 1.1014, + "step": 4574 + }, + { + "epoch": 0.34, + "learning_rate": 1.996362572163393e-05, + "loss": 1.0321, + "step": 4575 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963591723795988e-05, + "loss": 1.0848, + "step": 4576 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963557710106113e-05, + "loss": 1.006, + "step": 4577 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963523680564365e-05, + "loss": 1.1382, + "step": 4578 + }, + { + "epoch": 0.34, + "learning_rate": 1.996348963517079e-05, + "loss": 1.0963, + "step": 4579 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963455573925445e-05, + "loss": 1.0293, + "step": 4580 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963421496828385e-05, + "loss": 1.0639, + "step": 4581 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963387403879665e-05, + "loss": 1.1376, + "step": 4582 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963353295079337e-05, + "loss": 0.9924, + "step": 4583 + }, + { + "epoch": 0.34, + "learning_rate": 1.996331917042746e-05, + "loss": 1.0573, + "step": 4584 + }, + { + "epoch": 0.34, + "learning_rate": 1.996328502992408e-05, + "loss": 1.1776, + "step": 4585 + }, + { + "epoch": 0.34, + "learning_rate": 1.996325087356926e-05, + "loss": 0.9401, + "step": 4586 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963216701363046e-05, + "loss": 1.0825, + "step": 4587 + }, + { + "epoch": 0.34, + "learning_rate": 1.99631825133055e-05, + "loss": 1.0541, + "step": 4588 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963148309396673e-05, + "loss": 1.0099, + "step": 4589 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963114089636618e-05, + "loss": 0.9786, + "step": 4590 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963079854025392e-05, + "loss": 1.0668, + "step": 4591 + }, + { + "epoch": 0.34, + "learning_rate": 1.996304560256305e-05, + "loss": 1.09, + "step": 4592 + }, + { + "epoch": 0.34, + "learning_rate": 1.9963011335249642e-05, + "loss": 1.1121, + "step": 4593 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962977052085227e-05, + "loss": 1.0673, + "step": 4594 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962942753069858e-05, + "loss": 1.0399, + "step": 4595 + }, + { + "epoch": 0.34, + "learning_rate": 1.996290843820359e-05, + "loss": 1.0264, + "step": 4596 + }, + { + "epoch": 0.34, + "learning_rate": 1.996287410748648e-05, + "loss": 1.1136, + "step": 4597 + }, + { + "epoch": 0.34, + "learning_rate": 1.996283976091857e-05, + "loss": 1.1168, + "step": 4598 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962805398499933e-05, + "loss": 1.0775, + "step": 4599 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962771020230615e-05, + "loss": 1.0896, + "step": 4600 + }, + { + "epoch": 0.34, + "learning_rate": 1.996273662611067e-05, + "loss": 1.032, + "step": 4601 + }, + { + "epoch": 0.34, + "learning_rate": 1.996270221614015e-05, + "loss": 1.1047, + "step": 4602 + }, + { + "epoch": 0.34, + "learning_rate": 1.996266779031912e-05, + "loss": 1.0887, + "step": 4603 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962633348647622e-05, + "loss": 1.062, + "step": 4604 + }, + { + "epoch": 0.34, + "learning_rate": 1.996259889112572e-05, + "loss": 1.1618, + "step": 4605 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962564417753464e-05, + "loss": 1.0634, + "step": 4606 + }, + { + "epoch": 0.34, + "learning_rate": 1.996252992853091e-05, + "loss": 1.1277, + "step": 4607 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962495423458117e-05, + "loss": 0.9614, + "step": 4608 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962460902535135e-05, + "loss": 1.0038, + "step": 4609 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962426365762018e-05, + "loss": 1.0327, + "step": 4610 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962391813138825e-05, + "loss": 1.0666, + "step": 4611 + }, + { + "epoch": 0.34, + "learning_rate": 1.996235724466561e-05, + "loss": 1.0594, + "step": 4612 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962322660342424e-05, + "loss": 0.9659, + "step": 4613 + }, + { + "epoch": 0.34, + "learning_rate": 1.996228806016933e-05, + "loss": 1.047, + "step": 4614 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962253444146373e-05, + "loss": 1.0918, + "step": 4615 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962218812273618e-05, + "loss": 1.0712, + "step": 4616 + }, + { + "epoch": 0.34, + "learning_rate": 1.996218416455111e-05, + "loss": 1.1381, + "step": 4617 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962149500978915e-05, + "loss": 1.1, + "step": 4618 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962114821557078e-05, + "loss": 1.0221, + "step": 4619 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962080126285662e-05, + "loss": 1.2261, + "step": 4620 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962045415164718e-05, + "loss": 1.1363, + "step": 4621 + }, + { + "epoch": 0.34, + "learning_rate": 1.9962010688194304e-05, + "loss": 1.1504, + "step": 4622 + }, + { + "epoch": 0.34, + "learning_rate": 1.996197594537447e-05, + "loss": 1.0283, + "step": 4623 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961941186705277e-05, + "loss": 1.028, + "step": 4624 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961906412186777e-05, + "loss": 1.0682, + "step": 4625 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961871621819026e-05, + "loss": 1.0898, + "step": 4626 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961836815602083e-05, + "loss": 1.0063, + "step": 4627 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961801993535998e-05, + "loss": 0.9674, + "step": 4628 + }, + { + "epoch": 0.34, + "learning_rate": 1.996176715562083e-05, + "loss": 1.0177, + "step": 4629 + }, + { + "epoch": 0.34, + "learning_rate": 1.996173230185663e-05, + "loss": 1.1715, + "step": 4630 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961697432243458e-05, + "loss": 1.0355, + "step": 4631 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961662546781372e-05, + "loss": 0.9601, + "step": 4632 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961627645470418e-05, + "loss": 1.0651, + "step": 4633 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961592728310658e-05, + "loss": 1.0408, + "step": 4634 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961557795302146e-05, + "loss": 1.0539, + "step": 4635 + }, + { + "epoch": 0.34, + "learning_rate": 1.996152284644494e-05, + "loss": 1.0413, + "step": 4636 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961487881739092e-05, + "loss": 1.0038, + "step": 4637 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961452901184662e-05, + "loss": 1.0663, + "step": 4638 + }, + { + "epoch": 0.34, + "learning_rate": 1.99614179047817e-05, + "loss": 1.0776, + "step": 4639 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961382892530264e-05, + "loss": 1.0198, + "step": 4640 + }, + { + "epoch": 0.34, + "learning_rate": 1.996134786443041e-05, + "loss": 1.1924, + "step": 4641 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961312820482196e-05, + "loss": 1.0449, + "step": 4642 + }, + { + "epoch": 0.34, + "learning_rate": 1.996127776068568e-05, + "loss": 1.2099, + "step": 4643 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961242685040908e-05, + "loss": 0.9711, + "step": 4644 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961207593547943e-05, + "loss": 1.0697, + "step": 4645 + }, + { + "epoch": 0.34, + "learning_rate": 1.996117248620684e-05, + "loss": 1.0892, + "step": 4646 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961137363017652e-05, + "loss": 0.959, + "step": 4647 + }, + { + "epoch": 0.34, + "learning_rate": 1.996110222398044e-05, + "loss": 1.0126, + "step": 4648 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961067069095253e-05, + "loss": 0.9298, + "step": 4649 + }, + { + "epoch": 0.34, + "learning_rate": 1.9961031898362153e-05, + "loss": 1.1749, + "step": 4650 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960996711781196e-05, + "loss": 1.2023, + "step": 4651 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960961509352432e-05, + "loss": 1.0027, + "step": 4652 + }, + { + "epoch": 0.34, + "learning_rate": 1.996092629107592e-05, + "loss": 1.026, + "step": 4653 + }, + { + "epoch": 0.34, + "learning_rate": 1.996089105695172e-05, + "loss": 1.1004, + "step": 4654 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960855806979884e-05, + "loss": 1.0229, + "step": 4655 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960820541160468e-05, + "loss": 1.0501, + "step": 4656 + }, + { + "epoch": 0.34, + "learning_rate": 1.996078525949353e-05, + "loss": 1.0009, + "step": 4657 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960749961979124e-05, + "loss": 0.9756, + "step": 4658 + }, + { + "epoch": 0.34, + "learning_rate": 1.996071464861731e-05, + "loss": 1.1533, + "step": 4659 + }, + { + "epoch": 0.34, + "learning_rate": 1.996067931940814e-05, + "loss": 1.0345, + "step": 4660 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960643974351673e-05, + "loss": 0.9824, + "step": 4661 + }, + { + "epoch": 0.34, + "learning_rate": 1.996060861344796e-05, + "loss": 1.0977, + "step": 4662 + }, + { + "epoch": 0.34, + "learning_rate": 1.996057323669706e-05, + "loss": 1.1493, + "step": 4663 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960537844099036e-05, + "loss": 0.9816, + "step": 4664 + }, + { + "epoch": 0.34, + "learning_rate": 1.996050243565394e-05, + "loss": 1.1178, + "step": 4665 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960467011361825e-05, + "loss": 1.045, + "step": 4666 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960431571222747e-05, + "loss": 1.0553, + "step": 4667 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960396115236768e-05, + "loss": 1.0503, + "step": 4668 + }, + { + "epoch": 0.34, + "learning_rate": 1.9960360643403936e-05, + "loss": 1.0791, + "step": 4669 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960325155724318e-05, + "loss": 1.171, + "step": 4670 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960289652197964e-05, + "loss": 1.1345, + "step": 4671 + }, + { + "epoch": 0.35, + "learning_rate": 1.996025413282493e-05, + "loss": 1.0579, + "step": 4672 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960218597605277e-05, + "loss": 1.0483, + "step": 4673 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960183046539056e-05, + "loss": 1.0519, + "step": 4674 + }, + { + "epoch": 0.35, + "learning_rate": 1.996014747962633e-05, + "loss": 1.0746, + "step": 4675 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960111896867147e-05, + "loss": 1.0879, + "step": 4676 + }, + { + "epoch": 0.35, + "learning_rate": 1.996007629826157e-05, + "loss": 1.0427, + "step": 4677 + }, + { + "epoch": 0.35, + "learning_rate": 1.9960040683809656e-05, + "loss": 1.1348, + "step": 4678 + }, + { + "epoch": 0.35, + "learning_rate": 1.996000505351146e-05, + "loss": 1.0275, + "step": 4679 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959969407367035e-05, + "loss": 1.0443, + "step": 4680 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959933745376446e-05, + "loss": 1.1118, + "step": 4681 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959898067539744e-05, + "loss": 1.0192, + "step": 4682 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959862373856983e-05, + "loss": 1.0619, + "step": 4683 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959826664328224e-05, + "loss": 1.0285, + "step": 4684 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959790938953526e-05, + "loss": 0.9792, + "step": 4685 + }, + { + "epoch": 0.35, + "learning_rate": 1.995975519773294e-05, + "loss": 1.082, + "step": 4686 + }, + { + "epoch": 0.35, + "learning_rate": 1.995971944066653e-05, + "loss": 0.9276, + "step": 4687 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959683667754346e-05, + "loss": 0.9832, + "step": 4688 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959647878996447e-05, + "loss": 1.0749, + "step": 4689 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959612074392892e-05, + "loss": 1.0669, + "step": 4690 + }, + { + "epoch": 0.35, + "learning_rate": 1.995957625394374e-05, + "loss": 1.0438, + "step": 4691 + }, + { + "epoch": 0.35, + "learning_rate": 1.995954041764904e-05, + "loss": 1.0465, + "step": 4692 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959504565508854e-05, + "loss": 1.0901, + "step": 4693 + }, + { + "epoch": 0.35, + "learning_rate": 1.995946869752324e-05, + "loss": 1.047, + "step": 4694 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959432813692255e-05, + "loss": 0.9347, + "step": 4695 + }, + { + "epoch": 0.35, + "learning_rate": 1.995939691401595e-05, + "loss": 1.0554, + "step": 4696 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959360998494396e-05, + "loss": 1.0303, + "step": 4697 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959325067127634e-05, + "loss": 1.0223, + "step": 4698 + }, + { + "epoch": 0.35, + "learning_rate": 1.995928911991573e-05, + "loss": 1.1246, + "step": 4699 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959253156858737e-05, + "loss": 0.9823, + "step": 4700 + }, + { + "epoch": 0.35, + "learning_rate": 1.995921717795672e-05, + "loss": 1.0971, + "step": 4701 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959181183209728e-05, + "loss": 1.0422, + "step": 4702 + }, + { + "epoch": 0.35, + "learning_rate": 1.995914517261782e-05, + "loss": 1.1926, + "step": 4703 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959109146181056e-05, + "loss": 1.0172, + "step": 4704 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959073103899494e-05, + "loss": 0.9604, + "step": 4705 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959037045773187e-05, + "loss": 1.0552, + "step": 4706 + }, + { + "epoch": 0.35, + "learning_rate": 1.9959000971802194e-05, + "loss": 1.0505, + "step": 4707 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958964881986573e-05, + "loss": 1.0651, + "step": 4708 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958928776326384e-05, + "loss": 1.0026, + "step": 4709 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958892654821677e-05, + "loss": 1.1036, + "step": 4710 + }, + { + "epoch": 0.35, + "learning_rate": 1.995885651747252e-05, + "loss": 1.1751, + "step": 4711 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958820364278958e-05, + "loss": 1.0263, + "step": 4712 + }, + { + "epoch": 0.35, + "learning_rate": 1.995878419524106e-05, + "loss": 1.0362, + "step": 4713 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958748010358878e-05, + "loss": 1.0289, + "step": 4714 + }, + { + "epoch": 0.35, + "learning_rate": 1.995871180963247e-05, + "loss": 1.0269, + "step": 4715 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958675593061896e-05, + "loss": 1.1343, + "step": 4716 + }, + { + "epoch": 0.35, + "learning_rate": 1.995863936064721e-05, + "loss": 0.9488, + "step": 4717 + }, + { + "epoch": 0.35, + "learning_rate": 1.995860311238847e-05, + "loss": 1.0639, + "step": 4718 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958566848285735e-05, + "loss": 1.0386, + "step": 4719 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958530568339064e-05, + "loss": 1.0127, + "step": 4720 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958494272548513e-05, + "loss": 1.1726, + "step": 4721 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958457960914143e-05, + "loss": 1.0858, + "step": 4722 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958421633436005e-05, + "loss": 1.1065, + "step": 4723 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958385290114162e-05, + "loss": 1.0913, + "step": 4724 + }, + { + "epoch": 0.35, + "learning_rate": 1.995834893094867e-05, + "loss": 1.161, + "step": 4725 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958312555939586e-05, + "loss": 1.1061, + "step": 4726 + }, + { + "epoch": 0.35, + "learning_rate": 1.995827616508697e-05, + "loss": 1.036, + "step": 4727 + }, + { + "epoch": 0.35, + "learning_rate": 1.995823975839088e-05, + "loss": 1.0033, + "step": 4728 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958203335851375e-05, + "loss": 1.0879, + "step": 4729 + }, + { + "epoch": 0.35, + "learning_rate": 1.995816689746851e-05, + "loss": 0.9639, + "step": 4730 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958130443242343e-05, + "loss": 1.0889, + "step": 4731 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958093973172933e-05, + "loss": 1.0637, + "step": 4732 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958057487260335e-05, + "loss": 0.9591, + "step": 4733 + }, + { + "epoch": 0.35, + "learning_rate": 1.9958020985504615e-05, + "loss": 0.9447, + "step": 4734 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957984467905822e-05, + "loss": 0.9761, + "step": 4735 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957947934464023e-05, + "loss": 1.0424, + "step": 4736 + }, + { + "epoch": 0.35, + "learning_rate": 1.995791138517927e-05, + "loss": 0.9893, + "step": 4737 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957874820051617e-05, + "loss": 1.0212, + "step": 4738 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957838239081132e-05, + "loss": 0.9912, + "step": 4739 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957801642267865e-05, + "loss": 1.0828, + "step": 4740 + }, + { + "epoch": 0.35, + "learning_rate": 1.995776502961188e-05, + "loss": 1.0475, + "step": 4741 + }, + { + "epoch": 0.35, + "learning_rate": 1.995772840111324e-05, + "loss": 0.9924, + "step": 4742 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957691756771987e-05, + "loss": 1.0382, + "step": 4743 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957655096588194e-05, + "loss": 1.102, + "step": 4744 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957618420561914e-05, + "loss": 0.9686, + "step": 4745 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957581728693207e-05, + "loss": 1.0544, + "step": 4746 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957545020982127e-05, + "loss": 1.0514, + "step": 4747 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957508297428733e-05, + "loss": 1.054, + "step": 4748 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957471558033092e-05, + "loss": 1.0312, + "step": 4749 + }, + { + "epoch": 0.35, + "learning_rate": 1.995743480279525e-05, + "loss": 1.0986, + "step": 4750 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957398031715276e-05, + "loss": 1.2201, + "step": 4751 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957361244793226e-05, + "loss": 1.0905, + "step": 4752 + }, + { + "epoch": 0.35, + "learning_rate": 1.995732444202915e-05, + "loss": 1.0079, + "step": 4753 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957287623423118e-05, + "loss": 1.0296, + "step": 4754 + }, + { + "epoch": 0.35, + "learning_rate": 1.995725078897518e-05, + "loss": 1.0392, + "step": 4755 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957213938685403e-05, + "loss": 1.0173, + "step": 4756 + }, + { + "epoch": 0.35, + "learning_rate": 1.995717707255384e-05, + "loss": 1.1282, + "step": 4757 + }, + { + "epoch": 0.35, + "learning_rate": 1.995714019058055e-05, + "loss": 1.099, + "step": 4758 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957103292765595e-05, + "loss": 1.0229, + "step": 4759 + }, + { + "epoch": 0.35, + "learning_rate": 1.9957066379109026e-05, + "loss": 1.0357, + "step": 4760 + }, + { + "epoch": 0.35, + "learning_rate": 1.995702944961091e-05, + "loss": 1.0927, + "step": 4761 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956992504271304e-05, + "loss": 1.018, + "step": 4762 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956955543090264e-05, + "loss": 1.0575, + "step": 4763 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956918566067847e-05, + "loss": 1.1586, + "step": 4764 + }, + { + "epoch": 0.35, + "learning_rate": 1.995688157320412e-05, + "loss": 1.0852, + "step": 4765 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956844564499133e-05, + "loss": 1.0209, + "step": 4766 + }, + { + "epoch": 0.35, + "learning_rate": 1.995680753995295e-05, + "loss": 1.115, + "step": 4767 + }, + { + "epoch": 0.35, + "learning_rate": 1.995677049956563e-05, + "loss": 1.1212, + "step": 4768 + }, + { + "epoch": 0.35, + "learning_rate": 1.995673344333723e-05, + "loss": 1.0765, + "step": 4769 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956696371267806e-05, + "loss": 1.1386, + "step": 4770 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956659283357426e-05, + "loss": 1.0439, + "step": 4771 + }, + { + "epoch": 0.35, + "learning_rate": 1.995662217960614e-05, + "loss": 1.0179, + "step": 4772 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956585060014016e-05, + "loss": 1.0403, + "step": 4773 + }, + { + "epoch": 0.35, + "learning_rate": 1.99565479245811e-05, + "loss": 1.0577, + "step": 4774 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956510773307466e-05, + "loss": 1.0135, + "step": 4775 + }, + { + "epoch": 0.35, + "learning_rate": 1.995647360619316e-05, + "loss": 1.01, + "step": 4776 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956436423238246e-05, + "loss": 1.0491, + "step": 4777 + }, + { + "epoch": 0.35, + "learning_rate": 1.995639922444279e-05, + "loss": 0.978, + "step": 4778 + }, + { + "epoch": 0.35, + "learning_rate": 1.995636200980684e-05, + "loss": 1.0616, + "step": 4779 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956324779330464e-05, + "loss": 0.9619, + "step": 4780 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956287533013715e-05, + "loss": 1.0096, + "step": 4781 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956250270856658e-05, + "loss": 0.9071, + "step": 4782 + }, + { + "epoch": 0.35, + "learning_rate": 1.995621299285935e-05, + "loss": 1.115, + "step": 4783 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956175699021846e-05, + "loss": 1.0207, + "step": 4784 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956138389344208e-05, + "loss": 1.0413, + "step": 4785 + }, + { + "epoch": 0.35, + "learning_rate": 1.99561010638265e-05, + "loss": 1.1589, + "step": 4786 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956063722468776e-05, + "loss": 0.9263, + "step": 4787 + }, + { + "epoch": 0.35, + "learning_rate": 1.9956026365271098e-05, + "loss": 1.1227, + "step": 4788 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955988992233524e-05, + "loss": 0.9643, + "step": 4789 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955951603356113e-05, + "loss": 1.035, + "step": 4790 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955914198638928e-05, + "loss": 1.0049, + "step": 4791 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955876778082023e-05, + "loss": 1.1132, + "step": 4792 + }, + { + "epoch": 0.35, + "learning_rate": 1.995583934168546e-05, + "loss": 0.9228, + "step": 4793 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955801889449304e-05, + "loss": 1.0509, + "step": 4794 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955764421373603e-05, + "loss": 1.1525, + "step": 4795 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955726937458428e-05, + "loss": 1.0444, + "step": 4796 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955689437703834e-05, + "loss": 1.1388, + "step": 4797 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955651922109877e-05, + "loss": 1.0806, + "step": 4798 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955614390676624e-05, + "loss": 1.0841, + "step": 4799 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955576843404126e-05, + "loss": 0.9373, + "step": 4800 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955539280292454e-05, + "loss": 0.9977, + "step": 4801 + }, + { + "epoch": 0.35, + "learning_rate": 1.995550170134166e-05, + "loss": 1.0767, + "step": 4802 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955464106551802e-05, + "loss": 1.0687, + "step": 4803 + }, + { + "epoch": 0.35, + "learning_rate": 1.9955426495922944e-05, + "loss": 1.0966, + "step": 4804 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955388869455142e-05, + "loss": 1.007, + "step": 4805 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955351227148464e-05, + "loss": 1.0478, + "step": 4806 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955313569002962e-05, + "loss": 0.93, + "step": 4807 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955275895018702e-05, + "loss": 1.0851, + "step": 4808 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955238205195736e-05, + "loss": 1.0452, + "step": 4809 + }, + { + "epoch": 0.36, + "learning_rate": 1.995520049953413e-05, + "loss": 1.0049, + "step": 4810 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955162778033946e-05, + "loss": 1.0185, + "step": 4811 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955125040695234e-05, + "loss": 1.127, + "step": 4812 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955087287518066e-05, + "loss": 0.9736, + "step": 4813 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955049518502493e-05, + "loss": 1.041, + "step": 4814 + }, + { + "epoch": 0.36, + "learning_rate": 1.9955011733648583e-05, + "loss": 1.0497, + "step": 4815 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954973932956387e-05, + "loss": 1.0768, + "step": 4816 + }, + { + "epoch": 0.36, + "learning_rate": 1.995493611642597e-05, + "loss": 0.9651, + "step": 4817 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954898284057394e-05, + "loss": 0.9691, + "step": 4818 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954860435850718e-05, + "loss": 1.1381, + "step": 4819 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954822571806e-05, + "loss": 0.9764, + "step": 4820 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954784691923302e-05, + "loss": 1.1114, + "step": 4821 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954746796202683e-05, + "loss": 1.1425, + "step": 4822 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954708884644204e-05, + "loss": 0.9887, + "step": 4823 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954670957247928e-05, + "loss": 0.947, + "step": 4824 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954633014013913e-05, + "loss": 1.114, + "step": 4825 + }, + { + "epoch": 0.36, + "learning_rate": 1.995459505494222e-05, + "loss": 1.032, + "step": 4826 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954557080032907e-05, + "loss": 1.0339, + "step": 4827 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954519089286033e-05, + "loss": 0.9405, + "step": 4828 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954481082701666e-05, + "loss": 1.0433, + "step": 4829 + }, + { + "epoch": 0.36, + "learning_rate": 1.995444306027986e-05, + "loss": 1.0983, + "step": 4830 + }, + { + "epoch": 0.36, + "learning_rate": 1.995440502202068e-05, + "loss": 1.026, + "step": 4831 + }, + { + "epoch": 0.36, + "learning_rate": 1.995436696792418e-05, + "loss": 1.0374, + "step": 4832 + }, + { + "epoch": 0.36, + "learning_rate": 1.995432889799043e-05, + "loss": 1.0052, + "step": 4833 + }, + { + "epoch": 0.36, + "learning_rate": 1.995429081221948e-05, + "loss": 1.1242, + "step": 4834 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954252710611398e-05, + "loss": 1.032, + "step": 4835 + }, + { + "epoch": 0.36, + "learning_rate": 1.995421459316624e-05, + "loss": 1.0511, + "step": 4836 + }, + { + "epoch": 0.36, + "learning_rate": 1.995417645988407e-05, + "loss": 1.0776, + "step": 4837 + }, + { + "epoch": 0.36, + "learning_rate": 1.995413831076495e-05, + "loss": 1.1232, + "step": 4838 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954100145808935e-05, + "loss": 1.0483, + "step": 4839 + }, + { + "epoch": 0.36, + "learning_rate": 1.9954061965016095e-05, + "loss": 0.9688, + "step": 4840 + }, + { + "epoch": 0.36, + "learning_rate": 1.995402376838648e-05, + "loss": 1.0633, + "step": 4841 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953985555920156e-05, + "loss": 0.9808, + "step": 4842 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953947327617183e-05, + "loss": 0.9788, + "step": 4843 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953909083477625e-05, + "loss": 1.1299, + "step": 4844 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953870823501535e-05, + "loss": 1.1279, + "step": 4845 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953832547688986e-05, + "loss": 1.0694, + "step": 4846 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953794256040027e-05, + "loss": 1.0387, + "step": 4847 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953755948554724e-05, + "loss": 1.1956, + "step": 4848 + }, + { + "epoch": 0.36, + "learning_rate": 1.995371762523314e-05, + "loss": 0.9273, + "step": 4849 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953679286075333e-05, + "loss": 1.0327, + "step": 4850 + }, + { + "epoch": 0.36, + "learning_rate": 1.995364093108136e-05, + "loss": 1.0309, + "step": 4851 + }, + { + "epoch": 0.36, + "learning_rate": 1.995360256025129e-05, + "loss": 0.9854, + "step": 4852 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953564173585182e-05, + "loss": 1.0265, + "step": 4853 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953525771083097e-05, + "loss": 0.9299, + "step": 4854 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953487352745095e-05, + "loss": 1.0466, + "step": 4855 + }, + { + "epoch": 0.36, + "learning_rate": 1.995344891857123e-05, + "loss": 1.04, + "step": 4856 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953410468561576e-05, + "loss": 1.0648, + "step": 4857 + }, + { + "epoch": 0.36, + "learning_rate": 1.995337200271619e-05, + "loss": 1.0307, + "step": 4858 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953333521035127e-05, + "loss": 1.0056, + "step": 4859 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953295023518456e-05, + "loss": 0.9914, + "step": 4860 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953256510166234e-05, + "loss": 1.0493, + "step": 4861 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953217980978523e-05, + "loss": 1.1074, + "step": 4862 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953179435955383e-05, + "loss": 1.1813, + "step": 4863 + }, + { + "epoch": 0.36, + "learning_rate": 1.995314087509688e-05, + "loss": 1.0081, + "step": 4864 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953102298403073e-05, + "loss": 1.1003, + "step": 4865 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953063705874018e-05, + "loss": 1.104, + "step": 4866 + }, + { + "epoch": 0.36, + "learning_rate": 1.9953025097509784e-05, + "loss": 0.976, + "step": 4867 + }, + { + "epoch": 0.36, + "learning_rate": 1.995298647331043e-05, + "loss": 0.9496, + "step": 4868 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952947833276016e-05, + "loss": 1.022, + "step": 4869 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952909177406603e-05, + "loss": 0.9931, + "step": 4870 + }, + { + "epoch": 0.36, + "learning_rate": 1.995287050570226e-05, + "loss": 0.9493, + "step": 4871 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952831818163033e-05, + "loss": 1.1059, + "step": 4872 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952793114788998e-05, + "loss": 1.1343, + "step": 4873 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952754395580213e-05, + "loss": 1.0205, + "step": 4874 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952715660536737e-05, + "loss": 1.0425, + "step": 4875 + }, + { + "epoch": 0.36, + "learning_rate": 1.995267690965863e-05, + "loss": 0.9812, + "step": 4876 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952638142945958e-05, + "loss": 1.0219, + "step": 4877 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952599360398784e-05, + "loss": 0.9903, + "step": 4878 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952560562017165e-05, + "loss": 1.0776, + "step": 4879 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952521747801164e-05, + "loss": 1.161, + "step": 4880 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952482917750844e-05, + "loss": 1.0076, + "step": 4881 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952444071866262e-05, + "loss": 1.0619, + "step": 4882 + }, + { + "epoch": 0.36, + "learning_rate": 1.995240521014749e-05, + "loss": 1.0251, + "step": 4883 + }, + { + "epoch": 0.36, + "learning_rate": 1.995236633259458e-05, + "loss": 1.0785, + "step": 4884 + }, + { + "epoch": 0.36, + "learning_rate": 1.99523274392076e-05, + "loss": 0.9068, + "step": 4885 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952288529986607e-05, + "loss": 1.0137, + "step": 4886 + }, + { + "epoch": 0.36, + "learning_rate": 1.995224960493167e-05, + "loss": 1.0396, + "step": 4887 + }, + { + "epoch": 0.36, + "learning_rate": 1.995221066404284e-05, + "loss": 1.0836, + "step": 4888 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952171707320188e-05, + "loss": 1.0392, + "step": 4889 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952132734763774e-05, + "loss": 1.0781, + "step": 4890 + }, + { + "epoch": 0.36, + "learning_rate": 1.995209374637366e-05, + "loss": 0.9632, + "step": 4891 + }, + { + "epoch": 0.36, + "learning_rate": 1.9952054742149905e-05, + "loss": 1.0217, + "step": 4892 + }, + { + "epoch": 0.36, + "learning_rate": 1.995201572209257e-05, + "loss": 1.0574, + "step": 4893 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951976686201725e-05, + "loss": 0.9941, + "step": 4894 + }, + { + "epoch": 0.36, + "learning_rate": 1.995193763447743e-05, + "loss": 1.0823, + "step": 4895 + }, + { + "epoch": 0.36, + "learning_rate": 1.995189856691974e-05, + "loss": 0.9954, + "step": 4896 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951859483528723e-05, + "loss": 1.1328, + "step": 4897 + }, + { + "epoch": 0.36, + "learning_rate": 1.995182038430444e-05, + "loss": 1.1393, + "step": 4898 + }, + { + "epoch": 0.36, + "learning_rate": 1.995178126924695e-05, + "loss": 1.0339, + "step": 4899 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951742138356326e-05, + "loss": 1.0329, + "step": 4900 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951702991632618e-05, + "loss": 1.1293, + "step": 4901 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951663829075892e-05, + "loss": 0.97, + "step": 4902 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951624650686214e-05, + "loss": 0.985, + "step": 4903 + }, + { + "epoch": 0.36, + "learning_rate": 1.995158545646364e-05, + "loss": 0.9999, + "step": 4904 + }, + { + "epoch": 0.36, + "learning_rate": 1.995154624640824e-05, + "loss": 1.0599, + "step": 4905 + }, + { + "epoch": 0.36, + "learning_rate": 1.995150702052007e-05, + "loss": 1.107, + "step": 4906 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951467778799195e-05, + "loss": 1.0097, + "step": 4907 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951428521245675e-05, + "loss": 0.9211, + "step": 4908 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951389247859577e-05, + "loss": 1.1057, + "step": 4909 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951349958640963e-05, + "loss": 1.0839, + "step": 4910 + }, + { + "epoch": 0.36, + "learning_rate": 1.995131065358989e-05, + "loss": 1.0127, + "step": 4911 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951271332706424e-05, + "loss": 1.0427, + "step": 4912 + }, + { + "epoch": 0.36, + "learning_rate": 1.995123199599063e-05, + "loss": 1.0908, + "step": 4913 + }, + { + "epoch": 0.36, + "learning_rate": 1.995119264344257e-05, + "loss": 1.0183, + "step": 4914 + }, + { + "epoch": 0.36, + "learning_rate": 1.99511532750623e-05, + "loss": 1.1176, + "step": 4915 + }, + { + "epoch": 0.36, + "learning_rate": 1.995111389084989e-05, + "loss": 1.1745, + "step": 4916 + }, + { + "epoch": 0.36, + "learning_rate": 1.99510744908054e-05, + "loss": 1.0062, + "step": 4917 + }, + { + "epoch": 0.36, + "learning_rate": 1.9951035074928895e-05, + "loss": 1.1399, + "step": 4918 + }, + { + "epoch": 0.36, + "learning_rate": 1.995099564322043e-05, + "loss": 1.0878, + "step": 4919 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950956195680076e-05, + "loss": 1.0351, + "step": 4920 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950916732307897e-05, + "loss": 0.9648, + "step": 4921 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950877253103947e-05, + "loss": 1.0349, + "step": 4922 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950837758068298e-05, + "loss": 1.0824, + "step": 4923 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950798247201006e-05, + "loss": 1.1375, + "step": 4924 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950758720502136e-05, + "loss": 0.9877, + "step": 4925 + }, + { + "epoch": 0.36, + "learning_rate": 1.995071917797175e-05, + "loss": 0.9553, + "step": 4926 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950679619609915e-05, + "loss": 1.0181, + "step": 4927 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950640045416688e-05, + "loss": 1.072, + "step": 4928 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950600455392138e-05, + "loss": 0.9957, + "step": 4929 + }, + { + "epoch": 0.36, + "learning_rate": 1.995056084953632e-05, + "loss": 1.052, + "step": 4930 + }, + { + "epoch": 0.36, + "learning_rate": 1.995052122784931e-05, + "loss": 0.9915, + "step": 4931 + }, + { + "epoch": 0.36, + "learning_rate": 1.995048159033116e-05, + "loss": 1.0718, + "step": 4932 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950441936981934e-05, + "loss": 1.0274, + "step": 4933 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950402267801698e-05, + "loss": 1.1418, + "step": 4934 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950362582790513e-05, + "loss": 1.0912, + "step": 4935 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950322881948446e-05, + "loss": 0.9491, + "step": 4936 + }, + { + "epoch": 0.36, + "learning_rate": 1.995028316527556e-05, + "loss": 1.0728, + "step": 4937 + }, + { + "epoch": 0.36, + "learning_rate": 1.9950243432771913e-05, + "loss": 1.0028, + "step": 4938 + }, + { + "epoch": 0.36, + "learning_rate": 1.995020368443757e-05, + "loss": 1.0535, + "step": 4939 + }, + { + "epoch": 0.37, + "learning_rate": 1.9950163920272593e-05, + "loss": 1.002, + "step": 4940 + }, + { + "epoch": 0.37, + "learning_rate": 1.995012414027705e-05, + "loss": 1.0567, + "step": 4941 + }, + { + "epoch": 0.37, + "learning_rate": 1.9950084344451003e-05, + "loss": 1.0843, + "step": 4942 + }, + { + "epoch": 0.37, + "learning_rate": 1.9950044532794516e-05, + "loss": 1.0011, + "step": 4943 + }, + { + "epoch": 0.37, + "learning_rate": 1.995000470530765e-05, + "loss": 0.9207, + "step": 4944 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949964861990463e-05, + "loss": 1.0453, + "step": 4945 + }, + { + "epoch": 0.37, + "learning_rate": 1.994992500284303e-05, + "loss": 0.9702, + "step": 4946 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949885127865406e-05, + "loss": 1.2228, + "step": 4947 + }, + { + "epoch": 0.37, + "learning_rate": 1.994984523705766e-05, + "loss": 1.0921, + "step": 4948 + }, + { + "epoch": 0.37, + "learning_rate": 1.994980533041985e-05, + "loss": 1.086, + "step": 4949 + }, + { + "epoch": 0.37, + "learning_rate": 1.994976540795204e-05, + "loss": 1.0084, + "step": 4950 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949725469654302e-05, + "loss": 1.0563, + "step": 4951 + }, + { + "epoch": 0.37, + "learning_rate": 1.994968551552669e-05, + "loss": 1.1244, + "step": 4952 + }, + { + "epoch": 0.37, + "learning_rate": 1.994964554556927e-05, + "loss": 0.9959, + "step": 4953 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949605559782106e-05, + "loss": 1.0445, + "step": 4954 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949565558165263e-05, + "loss": 0.9679, + "step": 4955 + }, + { + "epoch": 0.37, + "learning_rate": 1.99495255407188e-05, + "loss": 1.1353, + "step": 4956 + }, + { + "epoch": 0.37, + "learning_rate": 1.994948550744279e-05, + "loss": 1.0412, + "step": 4957 + }, + { + "epoch": 0.37, + "learning_rate": 1.994944545833729e-05, + "loss": 1.0119, + "step": 4958 + }, + { + "epoch": 0.37, + "learning_rate": 1.994940539340236e-05, + "loss": 1.131, + "step": 4959 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949365312638075e-05, + "loss": 1.0545, + "step": 4960 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949325216044485e-05, + "loss": 1.0162, + "step": 4961 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949285103621665e-05, + "loss": 1.0657, + "step": 4962 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949244975369676e-05, + "loss": 0.9982, + "step": 4963 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949204831288576e-05, + "loss": 1.0664, + "step": 4964 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949164671378436e-05, + "loss": 1.0556, + "step": 4965 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949124495639317e-05, + "loss": 1.0554, + "step": 4966 + }, + { + "epoch": 0.37, + "learning_rate": 1.9949084304071284e-05, + "loss": 1.0646, + "step": 4967 + }, + { + "epoch": 0.37, + "learning_rate": 1.99490440966744e-05, + "loss": 0.954, + "step": 4968 + }, + { + "epoch": 0.37, + "learning_rate": 1.994900387344873e-05, + "loss": 1.0519, + "step": 4969 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948963634394335e-05, + "loss": 1.0676, + "step": 4970 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948923379511283e-05, + "loss": 0.923, + "step": 4971 + }, + { + "epoch": 0.37, + "learning_rate": 1.994888310879964e-05, + "loss": 1.0274, + "step": 4972 + }, + { + "epoch": 0.37, + "learning_rate": 1.994884282225946e-05, + "loss": 1.0296, + "step": 4973 + }, + { + "epoch": 0.37, + "learning_rate": 1.994880251989081e-05, + "loss": 1.1299, + "step": 4974 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948762201693768e-05, + "loss": 1.0418, + "step": 4975 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948721867668383e-05, + "loss": 1.051, + "step": 4976 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948681517814723e-05, + "loss": 1.0396, + "step": 4977 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948641152132853e-05, + "loss": 1.0951, + "step": 4978 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948600770622838e-05, + "loss": 1.0568, + "step": 4979 + }, + { + "epoch": 0.37, + "learning_rate": 1.994856037328474e-05, + "loss": 1.0069, + "step": 4980 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948519960118628e-05, + "loss": 1.0782, + "step": 4981 + }, + { + "epoch": 0.37, + "learning_rate": 1.994847953112456e-05, + "loss": 1.124, + "step": 4982 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948439086302605e-05, + "loss": 1.016, + "step": 4983 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948398625652824e-05, + "loss": 1.127, + "step": 4984 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948358149175284e-05, + "loss": 1.0569, + "step": 4985 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948317656870053e-05, + "loss": 1.068, + "step": 4986 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948277148737184e-05, + "loss": 1.0251, + "step": 4987 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948236624776752e-05, + "loss": 1.1618, + "step": 4988 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948196084988818e-05, + "loss": 0.9802, + "step": 4989 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948155529373444e-05, + "loss": 1.0108, + "step": 4990 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948114957930697e-05, + "loss": 1.0049, + "step": 4991 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948074370660647e-05, + "loss": 0.9486, + "step": 4992 + }, + { + "epoch": 0.37, + "learning_rate": 1.9948033767563347e-05, + "loss": 0.9409, + "step": 4993 + }, + { + "epoch": 0.37, + "learning_rate": 1.994799314863887e-05, + "loss": 1.026, + "step": 4994 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947952513887277e-05, + "loss": 1.1278, + "step": 4995 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947911863308634e-05, + "loss": 1.1327, + "step": 4996 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947871196903007e-05, + "loss": 1.0055, + "step": 4997 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947830514670457e-05, + "loss": 0.9157, + "step": 4998 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947789816611053e-05, + "loss": 1.0828, + "step": 4999 + }, + { + "epoch": 0.37, + "learning_rate": 1.994774910272486e-05, + "loss": 1.0827, + "step": 5000 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947708373011935e-05, + "loss": 1.1201, + "step": 5001 + }, + { + "epoch": 0.37, + "learning_rate": 1.994766762747235e-05, + "loss": 1.0657, + "step": 5002 + }, + { + "epoch": 0.37, + "learning_rate": 1.994762686610617e-05, + "loss": 1.1544, + "step": 5003 + }, + { + "epoch": 0.37, + "learning_rate": 1.994758608891346e-05, + "loss": 0.961, + "step": 5004 + }, + { + "epoch": 0.37, + "learning_rate": 1.994754529589428e-05, + "loss": 1.0313, + "step": 5005 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947504487048698e-05, + "loss": 1.0524, + "step": 5006 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947463662376777e-05, + "loss": 1.1152, + "step": 5007 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947422821878584e-05, + "loss": 1.0244, + "step": 5008 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947381965554184e-05, + "loss": 1.1126, + "step": 5009 + }, + { + "epoch": 0.37, + "learning_rate": 1.994734109340364e-05, + "loss": 1.0649, + "step": 5010 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947300205427026e-05, + "loss": 1.0533, + "step": 5011 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947259301624393e-05, + "loss": 1.0003, + "step": 5012 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947218381995813e-05, + "loss": 0.9159, + "step": 5013 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947177446541355e-05, + "loss": 1.1838, + "step": 5014 + }, + { + "epoch": 0.37, + "learning_rate": 1.994713649526108e-05, + "loss": 1.0227, + "step": 5015 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947095528155045e-05, + "loss": 1.1662, + "step": 5016 + }, + { + "epoch": 0.37, + "learning_rate": 1.994705454522333e-05, + "loss": 1.0935, + "step": 5017 + }, + { + "epoch": 0.37, + "learning_rate": 1.9947013546465992e-05, + "loss": 1.0885, + "step": 5018 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946972531883098e-05, + "loss": 1.07, + "step": 5019 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946931501474715e-05, + "loss": 1.0265, + "step": 5020 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946890455240903e-05, + "loss": 1.0788, + "step": 5021 + }, + { + "epoch": 0.37, + "learning_rate": 1.994684939318173e-05, + "loss": 0.9516, + "step": 5022 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946808315297266e-05, + "loss": 1.1042, + "step": 5023 + }, + { + "epoch": 0.37, + "learning_rate": 1.994676722158757e-05, + "loss": 0.9804, + "step": 5024 + }, + { + "epoch": 0.37, + "learning_rate": 1.994672611205271e-05, + "loss": 0.9938, + "step": 5025 + }, + { + "epoch": 0.37, + "learning_rate": 1.994668498669275e-05, + "loss": 1.0116, + "step": 5026 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946643845507757e-05, + "loss": 0.9873, + "step": 5027 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946602688497797e-05, + "loss": 1.0112, + "step": 5028 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946561515662935e-05, + "loss": 1.0694, + "step": 5029 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946520327003233e-05, + "loss": 1.0041, + "step": 5030 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946479122518764e-05, + "loss": 1.1474, + "step": 5031 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946437902209585e-05, + "loss": 1.0286, + "step": 5032 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946396666075768e-05, + "loss": 0.9276, + "step": 5033 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946355414117375e-05, + "loss": 1.0663, + "step": 5034 + }, + { + "epoch": 0.37, + "learning_rate": 1.994631414633447e-05, + "loss": 1.0765, + "step": 5035 + }, + { + "epoch": 0.37, + "learning_rate": 1.994627286272713e-05, + "loss": 1.0766, + "step": 5036 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946231563295403e-05, + "loss": 1.0159, + "step": 5037 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946190248039367e-05, + "loss": 1.1178, + "step": 5038 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946148916959085e-05, + "loss": 1.0876, + "step": 5039 + }, + { + "epoch": 0.37, + "learning_rate": 1.994610757005462e-05, + "loss": 1.0335, + "step": 5040 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946066207326048e-05, + "loss": 1.0193, + "step": 5041 + }, + { + "epoch": 0.37, + "learning_rate": 1.9946024828773418e-05, + "loss": 1.0221, + "step": 5042 + }, + { + "epoch": 0.37, + "learning_rate": 1.994598343439681e-05, + "loss": 1.0265, + "step": 5043 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945942024196282e-05, + "loss": 1.0491, + "step": 5044 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945900598171904e-05, + "loss": 1.0322, + "step": 5045 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945859156323738e-05, + "loss": 0.9722, + "step": 5046 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945817698651858e-05, + "loss": 1.0642, + "step": 5047 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945776225156322e-05, + "loss": 0.9863, + "step": 5048 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945734735837194e-05, + "loss": 1.0861, + "step": 5049 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945693230694548e-05, + "loss": 0.9839, + "step": 5050 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945651709728446e-05, + "loss": 1.1014, + "step": 5051 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945610172938956e-05, + "loss": 1.0587, + "step": 5052 + }, + { + "epoch": 0.37, + "learning_rate": 1.994556862032614e-05, + "loss": 1.0026, + "step": 5053 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945527051890066e-05, + "loss": 1.0584, + "step": 5054 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945485467630803e-05, + "loss": 1.0252, + "step": 5055 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945443867548415e-05, + "loss": 1.072, + "step": 5056 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945402251642968e-05, + "loss": 0.984, + "step": 5057 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945360619914528e-05, + "loss": 1.1122, + "step": 5058 + }, + { + "epoch": 0.37, + "learning_rate": 1.994531897236316e-05, + "loss": 0.9593, + "step": 5059 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945277308988933e-05, + "loss": 0.9836, + "step": 5060 + }, + { + "epoch": 0.37, + "learning_rate": 1.994523562979191e-05, + "loss": 1.1485, + "step": 5061 + }, + { + "epoch": 0.37, + "learning_rate": 1.994519393477216e-05, + "loss": 0.9838, + "step": 5062 + }, + { + "epoch": 0.37, + "learning_rate": 1.994515222392975e-05, + "loss": 0.9989, + "step": 5063 + }, + { + "epoch": 0.37, + "learning_rate": 1.994511049726475e-05, + "loss": 1.0337, + "step": 5064 + }, + { + "epoch": 0.37, + "learning_rate": 1.9945068754777214e-05, + "loss": 0.9814, + "step": 5065 + }, + { + "epoch": 0.37, + "learning_rate": 1.994502699646722e-05, + "loss": 1.0892, + "step": 5066 + }, + { + "epoch": 0.37, + "learning_rate": 1.994498522233483e-05, + "loss": 1.001, + "step": 5067 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944943432380105e-05, + "loss": 0.9526, + "step": 5068 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944901626603125e-05, + "loss": 1.1509, + "step": 5069 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944859805003946e-05, + "loss": 1.0559, + "step": 5070 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944817967582638e-05, + "loss": 1.1303, + "step": 5071 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944776114339267e-05, + "loss": 1.0726, + "step": 5072 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944734245273896e-05, + "loss": 1.0738, + "step": 5073 + }, + { + "epoch": 0.37, + "learning_rate": 1.99446923603866e-05, + "loss": 1.101, + "step": 5074 + }, + { + "epoch": 0.37, + "learning_rate": 1.9944650459677434e-05, + "loss": 0.9572, + "step": 5075 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944608543146478e-05, + "loss": 1.0821, + "step": 5076 + }, + { + "epoch": 0.38, + "learning_rate": 1.994456661079379e-05, + "loss": 0.9994, + "step": 5077 + }, + { + "epoch": 0.38, + "learning_rate": 1.994452466261944e-05, + "loss": 1.0648, + "step": 5078 + }, + { + "epoch": 0.38, + "learning_rate": 1.994448269862349e-05, + "loss": 1.1572, + "step": 5079 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944440718806014e-05, + "loss": 1.1047, + "step": 5080 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944398723167074e-05, + "loss": 1.0471, + "step": 5081 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944356711706737e-05, + "loss": 1.2647, + "step": 5082 + }, + { + "epoch": 0.38, + "learning_rate": 1.994431468442507e-05, + "loss": 1.0995, + "step": 5083 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944272641322144e-05, + "loss": 1.1135, + "step": 5084 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944230582398018e-05, + "loss": 1.0634, + "step": 5085 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944188507652766e-05, + "loss": 0.9298, + "step": 5086 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944146417086454e-05, + "loss": 1.0199, + "step": 5087 + }, + { + "epoch": 0.38, + "learning_rate": 1.9944104310699144e-05, + "loss": 1.0509, + "step": 5088 + }, + { + "epoch": 0.38, + "learning_rate": 1.994406218849091e-05, + "loss": 1.0084, + "step": 5089 + }, + { + "epoch": 0.38, + "learning_rate": 1.994402005046181e-05, + "loss": 1.0512, + "step": 5090 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943977896611925e-05, + "loss": 0.9186, + "step": 5091 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943935726941307e-05, + "loss": 1.0556, + "step": 5092 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943893541450028e-05, + "loss": 0.9617, + "step": 5093 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943851340138162e-05, + "loss": 1.0359, + "step": 5094 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943809123005767e-05, + "loss": 1.0233, + "step": 5095 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943766890052914e-05, + "loss": 1.0892, + "step": 5096 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943724641279672e-05, + "loss": 1.0313, + "step": 5097 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943682376686103e-05, + "loss": 1.1932, + "step": 5098 + }, + { + "epoch": 0.38, + "learning_rate": 1.994364009627228e-05, + "loss": 1.0569, + "step": 5099 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943597800038267e-05, + "loss": 0.9398, + "step": 5100 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943555487984138e-05, + "loss": 0.9901, + "step": 5101 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943513160109947e-05, + "loss": 1.0189, + "step": 5102 + }, + { + "epoch": 0.38, + "learning_rate": 1.994347081641577e-05, + "loss": 1.0426, + "step": 5103 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943428456901674e-05, + "loss": 1.0857, + "step": 5104 + }, + { + "epoch": 0.38, + "learning_rate": 1.994338608156773e-05, + "loss": 1.0072, + "step": 5105 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943343690413998e-05, + "loss": 1.0142, + "step": 5106 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943301283440543e-05, + "loss": 1.0032, + "step": 5107 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943258860647444e-05, + "loss": 0.9509, + "step": 5108 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943216422034762e-05, + "loss": 1.0278, + "step": 5109 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943173967602563e-05, + "loss": 1.0591, + "step": 5110 + }, + { + "epoch": 0.38, + "learning_rate": 1.994313149735092e-05, + "loss": 1.0695, + "step": 5111 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943089011279892e-05, + "loss": 1.097, + "step": 5112 + }, + { + "epoch": 0.38, + "learning_rate": 1.9943046509389556e-05, + "loss": 1.066, + "step": 5113 + }, + { + "epoch": 0.38, + "learning_rate": 1.994300399167997e-05, + "loss": 1.0689, + "step": 5114 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942961458151212e-05, + "loss": 1.0442, + "step": 5115 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942918908803344e-05, + "loss": 0.9656, + "step": 5116 + }, + { + "epoch": 0.38, + "learning_rate": 1.994287634363643e-05, + "loss": 0.9831, + "step": 5117 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942833762650548e-05, + "loss": 0.999, + "step": 5118 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942791165845755e-05, + "loss": 1.1449, + "step": 5119 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942748553222127e-05, + "loss": 1.2071, + "step": 5120 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942705924779728e-05, + "loss": 0.9668, + "step": 5121 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942663280518623e-05, + "loss": 1.0495, + "step": 5122 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942620620438887e-05, + "loss": 1.0219, + "step": 5123 + }, + { + "epoch": 0.38, + "learning_rate": 1.994257794454058e-05, + "loss": 1.1236, + "step": 5124 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942535252823776e-05, + "loss": 1.0241, + "step": 5125 + }, + { + "epoch": 0.38, + "learning_rate": 1.994249254528854e-05, + "loss": 0.9161, + "step": 5126 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942449821934942e-05, + "loss": 1.1183, + "step": 5127 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942407082763047e-05, + "loss": 1.0648, + "step": 5128 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942364327772923e-05, + "loss": 1.0124, + "step": 5129 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942321556964642e-05, + "loss": 0.9983, + "step": 5130 + }, + { + "epoch": 0.38, + "learning_rate": 1.994227877033827e-05, + "loss": 1.094, + "step": 5131 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942235967893872e-05, + "loss": 1.0445, + "step": 5132 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942193149631522e-05, + "loss": 1.0226, + "step": 5133 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942150315551286e-05, + "loss": 0.9929, + "step": 5134 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942107465653228e-05, + "loss": 0.9576, + "step": 5135 + }, + { + "epoch": 0.38, + "learning_rate": 1.994206459993742e-05, + "loss": 0.9877, + "step": 5136 + }, + { + "epoch": 0.38, + "learning_rate": 1.9942021718403927e-05, + "loss": 1.0192, + "step": 5137 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941978821052827e-05, + "loss": 1.1372, + "step": 5138 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941935907884172e-05, + "loss": 1.0224, + "step": 5139 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941892978898045e-05, + "loss": 1.113, + "step": 5140 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941850034094505e-05, + "loss": 1.0311, + "step": 5141 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941807073473625e-05, + "loss": 1.0633, + "step": 5142 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941764097035475e-05, + "loss": 1.069, + "step": 5143 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941721104780117e-05, + "loss": 1.0402, + "step": 5144 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941678096707628e-05, + "loss": 0.961, + "step": 5145 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941635072818065e-05, + "loss": 1.0485, + "step": 5146 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941592033111504e-05, + "loss": 1.1126, + "step": 5147 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941548977588016e-05, + "loss": 1.0879, + "step": 5148 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941505906247664e-05, + "loss": 1.084, + "step": 5149 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941462819090518e-05, + "loss": 1.1148, + "step": 5150 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941419716116647e-05, + "loss": 1.0761, + "step": 5151 + }, + { + "epoch": 0.38, + "learning_rate": 1.994137659732612e-05, + "loss": 1.0234, + "step": 5152 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941333462719004e-05, + "loss": 0.8869, + "step": 5153 + }, + { + "epoch": 0.38, + "learning_rate": 1.994129031229537e-05, + "loss": 1.0094, + "step": 5154 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941247146055284e-05, + "loss": 1.0625, + "step": 5155 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941203963998815e-05, + "loss": 1.0513, + "step": 5156 + }, + { + "epoch": 0.38, + "learning_rate": 1.9941160766126036e-05, + "loss": 1.0174, + "step": 5157 + }, + { + "epoch": 0.38, + "learning_rate": 1.994111755243701e-05, + "loss": 1.0163, + "step": 5158 + }, + { + "epoch": 0.38, + "learning_rate": 1.994107432293181e-05, + "loss": 1.0468, + "step": 5159 + }, + { + "epoch": 0.38, + "learning_rate": 1.99410310776105e-05, + "loss": 0.9631, + "step": 5160 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940987816473156e-05, + "loss": 1.078, + "step": 5161 + }, + { + "epoch": 0.38, + "learning_rate": 1.994094453951984e-05, + "loss": 1.0034, + "step": 5162 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940901246750625e-05, + "loss": 1.0699, + "step": 5163 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940857938165578e-05, + "loss": 1.051, + "step": 5164 + }, + { + "epoch": 0.38, + "learning_rate": 1.994081461376477e-05, + "loss": 1.0534, + "step": 5165 + }, + { + "epoch": 0.38, + "learning_rate": 1.994077127354826e-05, + "loss": 1.0369, + "step": 5166 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940727917516135e-05, + "loss": 1.1053, + "step": 5167 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940684545668447e-05, + "loss": 0.9202, + "step": 5168 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940641158005274e-05, + "loss": 1.1043, + "step": 5169 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940597754526685e-05, + "loss": 1.1291, + "step": 5170 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940554335232748e-05, + "loss": 0.9446, + "step": 5171 + }, + { + "epoch": 0.38, + "learning_rate": 1.994051090012353e-05, + "loss": 0.967, + "step": 5172 + }, + { + "epoch": 0.38, + "learning_rate": 1.99404674491991e-05, + "loss": 0.9606, + "step": 5173 + }, + { + "epoch": 0.38, + "learning_rate": 1.994042398245953e-05, + "loss": 1.0783, + "step": 5174 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940380499904885e-05, + "loss": 0.993, + "step": 5175 + }, + { + "epoch": 0.38, + "learning_rate": 1.994033700153524e-05, + "loss": 0.9405, + "step": 5176 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940293487350662e-05, + "loss": 0.9803, + "step": 5177 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940249957351216e-05, + "loss": 1.0652, + "step": 5178 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940206411536974e-05, + "loss": 0.9738, + "step": 5179 + }, + { + "epoch": 0.38, + "learning_rate": 1.994016284990801e-05, + "loss": 1.0927, + "step": 5180 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940119272464388e-05, + "loss": 0.928, + "step": 5181 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940075679206177e-05, + "loss": 1.0371, + "step": 5182 + }, + { + "epoch": 0.38, + "learning_rate": 1.9940032070133447e-05, + "loss": 1.1745, + "step": 5183 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939988445246266e-05, + "loss": 0.9212, + "step": 5184 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939944804544713e-05, + "loss": 1.0495, + "step": 5185 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939901148028845e-05, + "loss": 1.1013, + "step": 5186 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939857475698736e-05, + "loss": 0.9197, + "step": 5187 + }, + { + "epoch": 0.38, + "learning_rate": 1.993981378755446e-05, + "loss": 1.041, + "step": 5188 + }, + { + "epoch": 0.38, + "learning_rate": 1.993977008359608e-05, + "loss": 1.1013, + "step": 5189 + }, + { + "epoch": 0.38, + "learning_rate": 1.993972636382367e-05, + "loss": 1.0313, + "step": 5190 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939682628237292e-05, + "loss": 1.0737, + "step": 5191 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939638876837027e-05, + "loss": 0.8876, + "step": 5192 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939595109622934e-05, + "loss": 1.052, + "step": 5193 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939551326595092e-05, + "loss": 1.1021, + "step": 5194 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939507527753565e-05, + "loss": 1.1194, + "step": 5195 + }, + { + "epoch": 0.38, + "learning_rate": 1.993946371309842e-05, + "loss": 1.0243, + "step": 5196 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939419882629733e-05, + "loss": 0.9793, + "step": 5197 + }, + { + "epoch": 0.38, + "learning_rate": 1.993937603634757e-05, + "loss": 1.0629, + "step": 5198 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939332174252006e-05, + "loss": 1.0274, + "step": 5199 + }, + { + "epoch": 0.38, + "learning_rate": 1.99392882963431e-05, + "loss": 1.1176, + "step": 5200 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939244402620934e-05, + "loss": 0.9688, + "step": 5201 + }, + { + "epoch": 0.38, + "learning_rate": 1.993920049308557e-05, + "loss": 0.9887, + "step": 5202 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939156567737084e-05, + "loss": 1.0772, + "step": 5203 + }, + { + "epoch": 0.38, + "learning_rate": 1.993911262657554e-05, + "loss": 0.9292, + "step": 5204 + }, + { + "epoch": 0.38, + "learning_rate": 1.9939068669601008e-05, + "loss": 1.017, + "step": 5205 + }, + { + "epoch": 0.38, + "learning_rate": 1.993902469681356e-05, + "loss": 1.0102, + "step": 5206 + }, + { + "epoch": 0.38, + "learning_rate": 1.993898070821327e-05, + "loss": 1.164, + "step": 5207 + }, + { + "epoch": 0.38, + "learning_rate": 1.99389367038002e-05, + "loss": 1.1549, + "step": 5208 + }, + { + "epoch": 0.38, + "learning_rate": 1.993889268357442e-05, + "loss": 1.1243, + "step": 5209 + }, + { + "epoch": 0.38, + "learning_rate": 1.993884864753601e-05, + "loss": 1.0624, + "step": 5210 + }, + { + "epoch": 0.39, + "learning_rate": 1.993880459568503e-05, + "loss": 1.0867, + "step": 5211 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938760528021562e-05, + "loss": 1.0864, + "step": 5212 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938716444545662e-05, + "loss": 1.0504, + "step": 5213 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938672345257405e-05, + "loss": 1.0221, + "step": 5214 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938628230156866e-05, + "loss": 1.146, + "step": 5215 + }, + { + "epoch": 0.39, + "learning_rate": 1.993858409924411e-05, + "loss": 1.0526, + "step": 5216 + }, + { + "epoch": 0.39, + "learning_rate": 1.993853995251921e-05, + "loss": 1.0107, + "step": 5217 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938495789982234e-05, + "loss": 1.0937, + "step": 5218 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938451611633253e-05, + "loss": 1.0623, + "step": 5219 + }, + { + "epoch": 0.39, + "learning_rate": 1.993840741747234e-05, + "loss": 1.1037, + "step": 5220 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938363207499563e-05, + "loss": 1.0185, + "step": 5221 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938318981714992e-05, + "loss": 1.0671, + "step": 5222 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938274740118694e-05, + "loss": 1.0851, + "step": 5223 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938230482710748e-05, + "loss": 1.0365, + "step": 5224 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938186209491217e-05, + "loss": 0.9628, + "step": 5225 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938141920460174e-05, + "loss": 1.0389, + "step": 5226 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938097615617688e-05, + "loss": 1.1338, + "step": 5227 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938053294963834e-05, + "loss": 1.0196, + "step": 5228 + }, + { + "epoch": 0.39, + "learning_rate": 1.9938008958498675e-05, + "loss": 0.9756, + "step": 5229 + }, + { + "epoch": 0.39, + "learning_rate": 1.993796460622229e-05, + "loss": 1.0395, + "step": 5230 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937920238134744e-05, + "loss": 1.1141, + "step": 5231 + }, + { + "epoch": 0.39, + "learning_rate": 1.993787585423611e-05, + "loss": 0.9648, + "step": 5232 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937831454526457e-05, + "loss": 0.9977, + "step": 5233 + }, + { + "epoch": 0.39, + "learning_rate": 1.993778703900586e-05, + "loss": 0.972, + "step": 5234 + }, + { + "epoch": 0.39, + "learning_rate": 1.993774260767438e-05, + "loss": 0.9864, + "step": 5235 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937698160532095e-05, + "loss": 0.995, + "step": 5236 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937653697579078e-05, + "loss": 1.0407, + "step": 5237 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937609218815392e-05, + "loss": 1.116, + "step": 5238 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937564724241114e-05, + "loss": 1.032, + "step": 5239 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937520213856312e-05, + "loss": 1.0822, + "step": 5240 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937475687661057e-05, + "loss": 1.1235, + "step": 5241 + }, + { + "epoch": 0.39, + "learning_rate": 1.993743114565542e-05, + "loss": 1.0817, + "step": 5242 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937386587839475e-05, + "loss": 1.217, + "step": 5243 + }, + { + "epoch": 0.39, + "learning_rate": 1.993734201421329e-05, + "loss": 1.0833, + "step": 5244 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937297424776935e-05, + "loss": 1.0628, + "step": 5245 + }, + { + "epoch": 0.39, + "learning_rate": 1.993725281953048e-05, + "loss": 1.0851, + "step": 5246 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937208198474e-05, + "loss": 0.9815, + "step": 5247 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937163561607562e-05, + "loss": 0.9147, + "step": 5248 + }, + { + "epoch": 0.39, + "learning_rate": 1.993711890893124e-05, + "loss": 1.0988, + "step": 5249 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937074240445103e-05, + "loss": 1.0009, + "step": 5250 + }, + { + "epoch": 0.39, + "learning_rate": 1.9937029556149228e-05, + "loss": 0.9591, + "step": 5251 + }, + { + "epoch": 0.39, + "learning_rate": 1.993698485604368e-05, + "loss": 1.0034, + "step": 5252 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936940140128525e-05, + "loss": 1.095, + "step": 5253 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936895408403846e-05, + "loss": 0.9243, + "step": 5254 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936850660869704e-05, + "loss": 1.0914, + "step": 5255 + }, + { + "epoch": 0.39, + "learning_rate": 1.993680589752618e-05, + "loss": 1.1265, + "step": 5256 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936761118373334e-05, + "loss": 1.0433, + "step": 5257 + }, + { + "epoch": 0.39, + "learning_rate": 1.993671632341125e-05, + "loss": 0.8803, + "step": 5258 + }, + { + "epoch": 0.39, + "learning_rate": 1.993667151263999e-05, + "loss": 1.0907, + "step": 5259 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936626686059627e-05, + "loss": 1.0781, + "step": 5260 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936581843670233e-05, + "loss": 1.0261, + "step": 5261 + }, + { + "epoch": 0.39, + "learning_rate": 1.993653698547188e-05, + "loss": 1.0991, + "step": 5262 + }, + { + "epoch": 0.39, + "learning_rate": 1.993649211146464e-05, + "loss": 1.1377, + "step": 5263 + }, + { + "epoch": 0.39, + "learning_rate": 1.993644722164858e-05, + "loss": 0.9695, + "step": 5264 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936402316023774e-05, + "loss": 1.0072, + "step": 5265 + }, + { + "epoch": 0.39, + "learning_rate": 1.99363573945903e-05, + "loss": 1.0439, + "step": 5266 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936312457348217e-05, + "loss": 1.0281, + "step": 5267 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936267504297607e-05, + "loss": 1.0928, + "step": 5268 + }, + { + "epoch": 0.39, + "learning_rate": 1.993622253543854e-05, + "loss": 1.1517, + "step": 5269 + }, + { + "epoch": 0.39, + "learning_rate": 1.993617755077108e-05, + "loss": 1.1535, + "step": 5270 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936132550295305e-05, + "loss": 1.0278, + "step": 5271 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936087534011283e-05, + "loss": 1.0628, + "step": 5272 + }, + { + "epoch": 0.39, + "learning_rate": 1.9936042501919092e-05, + "loss": 1.0473, + "step": 5273 + }, + { + "epoch": 0.39, + "learning_rate": 1.99359974540188e-05, + "loss": 0.992, + "step": 5274 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935952390310477e-05, + "loss": 1.1824, + "step": 5275 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935907310794192e-05, + "loss": 1.0315, + "step": 5276 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935862215470026e-05, + "loss": 1.0766, + "step": 5277 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935817104338043e-05, + "loss": 1.0737, + "step": 5278 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935771977398318e-05, + "loss": 1.0288, + "step": 5279 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935726834650922e-05, + "loss": 1.0231, + "step": 5280 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935681676095923e-05, + "loss": 0.9741, + "step": 5281 + }, + { + "epoch": 0.39, + "learning_rate": 1.99356365017334e-05, + "loss": 1.071, + "step": 5282 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935591311563422e-05, + "loss": 0.9571, + "step": 5283 + }, + { + "epoch": 0.39, + "learning_rate": 1.993554610558606e-05, + "loss": 1.0623, + "step": 5284 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935500883801383e-05, + "loss": 1.0838, + "step": 5285 + }, + { + "epoch": 0.39, + "learning_rate": 1.993545564620947e-05, + "loss": 0.9861, + "step": 5286 + }, + { + "epoch": 0.39, + "learning_rate": 1.993541039281039e-05, + "loss": 1.0034, + "step": 5287 + }, + { + "epoch": 0.39, + "learning_rate": 1.993536512360421e-05, + "loss": 1.0403, + "step": 5288 + }, + { + "epoch": 0.39, + "learning_rate": 1.993531983859101e-05, + "loss": 0.9997, + "step": 5289 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935274537770858e-05, + "loss": 1.0317, + "step": 5290 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935229221143827e-05, + "loss": 1.1218, + "step": 5291 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935183888709986e-05, + "loss": 1.0837, + "step": 5292 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935138540469408e-05, + "loss": 1.0749, + "step": 5293 + }, + { + "epoch": 0.39, + "learning_rate": 1.993509317642217e-05, + "loss": 1.1755, + "step": 5294 + }, + { + "epoch": 0.39, + "learning_rate": 1.993504779656834e-05, + "loss": 1.0418, + "step": 5295 + }, + { + "epoch": 0.39, + "learning_rate": 1.9935002400907996e-05, + "loss": 1.1244, + "step": 5296 + }, + { + "epoch": 0.39, + "learning_rate": 1.99349569894412e-05, + "loss": 1.0627, + "step": 5297 + }, + { + "epoch": 0.39, + "learning_rate": 1.993491156216803e-05, + "loss": 1.0451, + "step": 5298 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934866119088555e-05, + "loss": 1.1049, + "step": 5299 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934820660202855e-05, + "loss": 1.0704, + "step": 5300 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934775185510996e-05, + "loss": 1.106, + "step": 5301 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934729695013047e-05, + "loss": 0.9963, + "step": 5302 + }, + { + "epoch": 0.39, + "learning_rate": 1.993468418870909e-05, + "loss": 0.9248, + "step": 5303 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934638666599193e-05, + "loss": 0.9323, + "step": 5304 + }, + { + "epoch": 0.39, + "learning_rate": 1.993459312868343e-05, + "loss": 1.0894, + "step": 5305 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934547574961867e-05, + "loss": 1.0433, + "step": 5306 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934502005434583e-05, + "loss": 0.9724, + "step": 5307 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934456420101647e-05, + "loss": 1.1507, + "step": 5308 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934410818963137e-05, + "loss": 0.9737, + "step": 5309 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934365202019117e-05, + "loss": 1.0603, + "step": 5310 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934319569269663e-05, + "loss": 1.0458, + "step": 5311 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934273920714852e-05, + "loss": 0.9906, + "step": 5312 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934228256354755e-05, + "loss": 1.1409, + "step": 5313 + }, + { + "epoch": 0.39, + "learning_rate": 1.993418257618944e-05, + "loss": 1.0976, + "step": 5314 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934136880218985e-05, + "loss": 0.9805, + "step": 5315 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934091168443457e-05, + "loss": 1.0596, + "step": 5316 + }, + { + "epoch": 0.39, + "learning_rate": 1.9934045440862935e-05, + "loss": 0.9452, + "step": 5317 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933999697477486e-05, + "loss": 1.1338, + "step": 5318 + }, + { + "epoch": 0.39, + "learning_rate": 1.993395393828719e-05, + "loss": 1.0805, + "step": 5319 + }, + { + "epoch": 0.39, + "learning_rate": 1.993390816329211e-05, + "loss": 0.9691, + "step": 5320 + }, + { + "epoch": 0.39, + "learning_rate": 1.993386237249233e-05, + "loss": 1.0583, + "step": 5321 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933816565887912e-05, + "loss": 1.1472, + "step": 5322 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933770743478934e-05, + "loss": 1.0989, + "step": 5323 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933724905265474e-05, + "loss": 1.0814, + "step": 5324 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933679051247595e-05, + "loss": 1.048, + "step": 5325 + }, + { + "epoch": 0.39, + "learning_rate": 1.993363318142538e-05, + "loss": 1.1957, + "step": 5326 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933587295798887e-05, + "loss": 1.0309, + "step": 5327 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933541394368207e-05, + "loss": 0.9961, + "step": 5328 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933495477133403e-05, + "loss": 1.1013, + "step": 5329 + }, + { + "epoch": 0.39, + "learning_rate": 1.993344954409455e-05, + "loss": 1.0768, + "step": 5330 + }, + { + "epoch": 0.39, + "learning_rate": 1.993340359525172e-05, + "loss": 1.0695, + "step": 5331 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933357630604988e-05, + "loss": 1.0697, + "step": 5332 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933311650154424e-05, + "loss": 0.9764, + "step": 5333 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933265653900106e-05, + "loss": 1.0362, + "step": 5334 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933219641842102e-05, + "loss": 0.915, + "step": 5335 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933173613980492e-05, + "loss": 0.984, + "step": 5336 + }, + { + "epoch": 0.39, + "learning_rate": 1.993312757031534e-05, + "loss": 1.1086, + "step": 5337 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933081510846728e-05, + "loss": 0.919, + "step": 5338 + }, + { + "epoch": 0.39, + "learning_rate": 1.9933035435574722e-05, + "loss": 1.0713, + "step": 5339 + }, + { + "epoch": 0.39, + "learning_rate": 1.99329893444994e-05, + "loss": 1.0677, + "step": 5340 + }, + { + "epoch": 0.39, + "learning_rate": 1.9932943237620834e-05, + "loss": 0.9997, + "step": 5341 + }, + { + "epoch": 0.39, + "learning_rate": 1.9932897114939097e-05, + "loss": 1.0381, + "step": 5342 + }, + { + "epoch": 0.39, + "learning_rate": 1.9932850976454262e-05, + "loss": 1.0702, + "step": 5343 + }, + { + "epoch": 0.39, + "learning_rate": 1.993280482216641e-05, + "loss": 1.0946, + "step": 5344 + }, + { + "epoch": 0.39, + "learning_rate": 1.99327586520756e-05, + "loss": 1.0053, + "step": 5345 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932712466181915e-05, + "loss": 1.1372, + "step": 5346 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932666264485428e-05, + "loss": 1.1194, + "step": 5347 + }, + { + "epoch": 0.4, + "learning_rate": 1.993262004698621e-05, + "loss": 1.064, + "step": 5348 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932573813684334e-05, + "loss": 1.0225, + "step": 5349 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932527564579877e-05, + "loss": 1.0452, + "step": 5350 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932481299672912e-05, + "loss": 1.0642, + "step": 5351 + }, + { + "epoch": 0.4, + "learning_rate": 1.993243501896351e-05, + "loss": 1.0569, + "step": 5352 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932388722451744e-05, + "loss": 1.0959, + "step": 5353 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932342410137694e-05, + "loss": 1.1427, + "step": 5354 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932296082021424e-05, + "loss": 0.9909, + "step": 5355 + }, + { + "epoch": 0.4, + "learning_rate": 1.993224973810302e-05, + "loss": 0.9742, + "step": 5356 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932203378382543e-05, + "loss": 1.0303, + "step": 5357 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932157002860072e-05, + "loss": 0.9233, + "step": 5358 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932110611535686e-05, + "loss": 1.0196, + "step": 5359 + }, + { + "epoch": 0.4, + "learning_rate": 1.993206420440945e-05, + "loss": 1.0517, + "step": 5360 + }, + { + "epoch": 0.4, + "learning_rate": 1.9932017781481445e-05, + "loss": 0.9831, + "step": 5361 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931971342751742e-05, + "loss": 0.9931, + "step": 5362 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931924888220413e-05, + "loss": 1.0299, + "step": 5363 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931878417887534e-05, + "loss": 0.9924, + "step": 5364 + }, + { + "epoch": 0.4, + "learning_rate": 1.993183193175318e-05, + "loss": 1.018, + "step": 5365 + }, + { + "epoch": 0.4, + "learning_rate": 1.993178542981742e-05, + "loss": 1.0995, + "step": 5366 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931738912080332e-05, + "loss": 1.0535, + "step": 5367 + }, + { + "epoch": 0.4, + "learning_rate": 1.993169237854199e-05, + "loss": 1.0936, + "step": 5368 + }, + { + "epoch": 0.4, + "learning_rate": 1.993164582920247e-05, + "loss": 1.0902, + "step": 5369 + }, + { + "epoch": 0.4, + "learning_rate": 1.993159926406184e-05, + "loss": 0.9896, + "step": 5370 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931552683120182e-05, + "loss": 1.0282, + "step": 5371 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931506086377564e-05, + "loss": 1.0368, + "step": 5372 + }, + { + "epoch": 0.4, + "learning_rate": 1.993145947383406e-05, + "loss": 0.9814, + "step": 5373 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931412845489748e-05, + "loss": 1.0171, + "step": 5374 + }, + { + "epoch": 0.4, + "learning_rate": 1.99313662013447e-05, + "loss": 0.9745, + "step": 5375 + }, + { + "epoch": 0.4, + "learning_rate": 1.993131954139899e-05, + "loss": 0.971, + "step": 5376 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931272865652693e-05, + "loss": 1.0328, + "step": 5377 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931226174105882e-05, + "loss": 0.8645, + "step": 5378 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931179466758633e-05, + "loss": 1.0833, + "step": 5379 + }, + { + "epoch": 0.4, + "learning_rate": 1.993113274361102e-05, + "loss": 1.0909, + "step": 5380 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931086004663118e-05, + "loss": 1.0285, + "step": 5381 + }, + { + "epoch": 0.4, + "learning_rate": 1.9931039249914997e-05, + "loss": 1.1248, + "step": 5382 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930992479366734e-05, + "loss": 1.0509, + "step": 5383 + }, + { + "epoch": 0.4, + "learning_rate": 1.993094569301841e-05, + "loss": 1.0025, + "step": 5384 + }, + { + "epoch": 0.4, + "learning_rate": 1.993089889087009e-05, + "loss": 1.1557, + "step": 5385 + }, + { + "epoch": 0.4, + "learning_rate": 1.993085207292185e-05, + "loss": 0.9499, + "step": 5386 + }, + { + "epoch": 0.4, + "learning_rate": 1.993080523917377e-05, + "loss": 1.0853, + "step": 5387 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930758389625917e-05, + "loss": 0.9399, + "step": 5388 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930711524278373e-05, + "loss": 1.0301, + "step": 5389 + }, + { + "epoch": 0.4, + "learning_rate": 1.993066464313121e-05, + "loss": 1.0576, + "step": 5390 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930617746184498e-05, + "loss": 1.0851, + "step": 5391 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930570833438316e-05, + "loss": 0.9573, + "step": 5392 + }, + { + "epoch": 0.4, + "learning_rate": 1.993052390489274e-05, + "loss": 1.0435, + "step": 5393 + }, + { + "epoch": 0.4, + "learning_rate": 1.993047696054784e-05, + "loss": 0.998, + "step": 5394 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930430000403693e-05, + "loss": 1.086, + "step": 5395 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930383024460375e-05, + "loss": 1.0103, + "step": 5396 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930336032717962e-05, + "loss": 0.9907, + "step": 5397 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930289025176524e-05, + "loss": 1.0234, + "step": 5398 + }, + { + "epoch": 0.4, + "learning_rate": 1.993024200183614e-05, + "loss": 1.0452, + "step": 5399 + }, + { + "epoch": 0.4, + "learning_rate": 1.993019496269688e-05, + "loss": 0.9757, + "step": 5400 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930147907758822e-05, + "loss": 1.1003, + "step": 5401 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930100837022045e-05, + "loss": 1.0647, + "step": 5402 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930053750486616e-05, + "loss": 1.1139, + "step": 5403 + }, + { + "epoch": 0.4, + "learning_rate": 1.9930006648152617e-05, + "loss": 1.01, + "step": 5404 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929959530020115e-05, + "loss": 1.0816, + "step": 5405 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929912396089194e-05, + "loss": 0.9716, + "step": 5406 + }, + { + "epoch": 0.4, + "learning_rate": 1.992986524635992e-05, + "loss": 1.0138, + "step": 5407 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929818080832374e-05, + "loss": 1.0519, + "step": 5408 + }, + { + "epoch": 0.4, + "learning_rate": 1.992977089950663e-05, + "loss": 1.0026, + "step": 5409 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929723702382765e-05, + "loss": 1.131, + "step": 5410 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929676489460846e-05, + "loss": 1.0961, + "step": 5411 + }, + { + "epoch": 0.4, + "learning_rate": 1.992962926074096e-05, + "loss": 1.0267, + "step": 5412 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929582016223172e-05, + "loss": 0.8864, + "step": 5413 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929534755907563e-05, + "loss": 1.0813, + "step": 5414 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929487479794204e-05, + "loss": 1.0663, + "step": 5415 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929440187883172e-05, + "loss": 0.9785, + "step": 5416 + }, + { + "epoch": 0.4, + "learning_rate": 1.992939288017455e-05, + "loss": 1.0297, + "step": 5417 + }, + { + "epoch": 0.4, + "learning_rate": 1.99293455566684e-05, + "loss": 1.1374, + "step": 5418 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929298217364802e-05, + "loss": 1.1797, + "step": 5419 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929250862263832e-05, + "loss": 0.9915, + "step": 5420 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929203491365568e-05, + "loss": 1.0354, + "step": 5421 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929156104670086e-05, + "loss": 1.0579, + "step": 5422 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929108702177456e-05, + "loss": 0.8261, + "step": 5423 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929061283887753e-05, + "loss": 0.9601, + "step": 5424 + }, + { + "epoch": 0.4, + "learning_rate": 1.9929013849801062e-05, + "loss": 0.9949, + "step": 5425 + }, + { + "epoch": 0.4, + "learning_rate": 1.992896639991745e-05, + "loss": 1.1002, + "step": 5426 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928918934236992e-05, + "loss": 1.0111, + "step": 5427 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928871452759767e-05, + "loss": 1.0886, + "step": 5428 + }, + { + "epoch": 0.4, + "learning_rate": 1.992882395548585e-05, + "loss": 1.0243, + "step": 5429 + }, + { + "epoch": 0.4, + "learning_rate": 1.992877644241532e-05, + "loss": 1.0519, + "step": 5430 + }, + { + "epoch": 0.4, + "learning_rate": 1.992872891354824e-05, + "loss": 1.0635, + "step": 5431 + }, + { + "epoch": 0.4, + "learning_rate": 1.99286813688847e-05, + "loss": 1.027, + "step": 5432 + }, + { + "epoch": 0.4, + "learning_rate": 1.992863380842477e-05, + "loss": 1.0745, + "step": 5433 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928586232168525e-05, + "loss": 0.9565, + "step": 5434 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928538640116044e-05, + "loss": 1.1275, + "step": 5435 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928491032267398e-05, + "loss": 1.1352, + "step": 5436 + }, + { + "epoch": 0.4, + "learning_rate": 1.992844340862266e-05, + "loss": 1.0164, + "step": 5437 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928395769181915e-05, + "loss": 1.0381, + "step": 5438 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928348113945235e-05, + "loss": 1.1005, + "step": 5439 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928300442912695e-05, + "loss": 1.0153, + "step": 5440 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928252756084373e-05, + "loss": 1.0473, + "step": 5441 + }, + { + "epoch": 0.4, + "learning_rate": 1.992820505346034e-05, + "loss": 1.059, + "step": 5442 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928157335040673e-05, + "loss": 1.0573, + "step": 5443 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928109600825455e-05, + "loss": 1.126, + "step": 5444 + }, + { + "epoch": 0.4, + "learning_rate": 1.992806185081475e-05, + "loss": 1.0739, + "step": 5445 + }, + { + "epoch": 0.4, + "learning_rate": 1.9928014085008648e-05, + "loss": 1.0369, + "step": 5446 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927966303407212e-05, + "loss": 1.0413, + "step": 5447 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927918506010525e-05, + "loss": 1.0105, + "step": 5448 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927870692818664e-05, + "loss": 1.0543, + "step": 5449 + }, + { + "epoch": 0.4, + "learning_rate": 1.99278228638317e-05, + "loss": 1.0724, + "step": 5450 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927775019049714e-05, + "loss": 1.0354, + "step": 5451 + }, + { + "epoch": 0.4, + "learning_rate": 1.992772715847278e-05, + "loss": 0.9795, + "step": 5452 + }, + { + "epoch": 0.4, + "learning_rate": 1.992767928210097e-05, + "loss": 1.1161, + "step": 5453 + }, + { + "epoch": 0.4, + "learning_rate": 1.992763138993437e-05, + "loss": 1.0113, + "step": 5454 + }, + { + "epoch": 0.4, + "learning_rate": 1.992758348197305e-05, + "loss": 1.0082, + "step": 5455 + }, + { + "epoch": 0.4, + "learning_rate": 1.992753555821708e-05, + "loss": 1.0927, + "step": 5456 + }, + { + "epoch": 0.4, + "learning_rate": 1.992748761866655e-05, + "loss": 1.1257, + "step": 5457 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927439663321525e-05, + "loss": 1.0804, + "step": 5458 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927391692182087e-05, + "loss": 1.0456, + "step": 5459 + }, + { + "epoch": 0.4, + "learning_rate": 1.992734370524831e-05, + "loss": 0.9194, + "step": 5460 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927295702520274e-05, + "loss": 1.0063, + "step": 5461 + }, + { + "epoch": 0.4, + "learning_rate": 1.992724768399805e-05, + "loss": 1.0803, + "step": 5462 + }, + { + "epoch": 0.4, + "learning_rate": 1.992719964968172e-05, + "loss": 1.0692, + "step": 5463 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927151599571354e-05, + "loss": 0.9911, + "step": 5464 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927103533667033e-05, + "loss": 1.1546, + "step": 5465 + }, + { + "epoch": 0.4, + "learning_rate": 1.9927055451968833e-05, + "loss": 0.9835, + "step": 5466 + }, + { + "epoch": 0.4, + "learning_rate": 1.992700735447683e-05, + "loss": 0.9532, + "step": 5467 + }, + { + "epoch": 0.4, + "learning_rate": 1.99269592411911e-05, + "loss": 1.0357, + "step": 5468 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926911112111716e-05, + "loss": 1.0351, + "step": 5469 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926862967238764e-05, + "loss": 1.1077, + "step": 5470 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926814806572313e-05, + "loss": 1.0306, + "step": 5471 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926766630112442e-05, + "loss": 0.9689, + "step": 5472 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926718437859228e-05, + "loss": 1.0575, + "step": 5473 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926670229812746e-05, + "loss": 1.1325, + "step": 5474 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926622005973073e-05, + "loss": 1.0645, + "step": 5475 + }, + { + "epoch": 0.4, + "learning_rate": 1.992657376634029e-05, + "loss": 1.1072, + "step": 5476 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926525510914466e-05, + "loss": 1.0234, + "step": 5477 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926477239695684e-05, + "loss": 1.1279, + "step": 5478 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926428952684013e-05, + "loss": 1.0613, + "step": 5479 + }, + { + "epoch": 0.4, + "learning_rate": 1.9926380649879542e-05, + "loss": 1.0698, + "step": 5480 + }, + { + "epoch": 0.4, + "learning_rate": 1.992633233128234e-05, + "loss": 1.1289, + "step": 5481 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926283996892486e-05, + "loss": 1.1051, + "step": 5482 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926235646710053e-05, + "loss": 0.928, + "step": 5483 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926187280735125e-05, + "loss": 1.1216, + "step": 5484 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926138898967772e-05, + "loss": 0.9759, + "step": 5485 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926090501408077e-05, + "loss": 0.9608, + "step": 5486 + }, + { + "epoch": 0.41, + "learning_rate": 1.9926042088056112e-05, + "loss": 1.1368, + "step": 5487 + }, + { + "epoch": 0.41, + "learning_rate": 1.992599365891196e-05, + "loss": 1.0393, + "step": 5488 + }, + { + "epoch": 0.41, + "learning_rate": 1.992594521397569e-05, + "loss": 1.0568, + "step": 5489 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925896753247383e-05, + "loss": 1.1152, + "step": 5490 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925848276727117e-05, + "loss": 0.9753, + "step": 5491 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925799784414968e-05, + "loss": 1.1035, + "step": 5492 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925751276311015e-05, + "loss": 1.0625, + "step": 5493 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925702752415332e-05, + "loss": 0.9423, + "step": 5494 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925654212727997e-05, + "loss": 1.0168, + "step": 5495 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925605657249095e-05, + "loss": 1.0073, + "step": 5496 + }, + { + "epoch": 0.41, + "learning_rate": 1.992555708597869e-05, + "loss": 0.9915, + "step": 5497 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925508498916866e-05, + "loss": 1.0307, + "step": 5498 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925459896063702e-05, + "loss": 1.0217, + "step": 5499 + }, + { + "epoch": 0.41, + "learning_rate": 1.992541127741927e-05, + "loss": 0.9589, + "step": 5500 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925362642983656e-05, + "loss": 0.9344, + "step": 5501 + }, + { + "epoch": 0.41, + "learning_rate": 1.992531399275693e-05, + "loss": 0.9983, + "step": 5502 + }, + { + "epoch": 0.41, + "learning_rate": 1.992526532673917e-05, + "loss": 1.0425, + "step": 5503 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925216644930453e-05, + "loss": 1.0131, + "step": 5504 + }, + { + "epoch": 0.41, + "learning_rate": 1.992516794733086e-05, + "loss": 0.9548, + "step": 5505 + }, + { + "epoch": 0.41, + "learning_rate": 1.992511923394047e-05, + "loss": 0.9963, + "step": 5506 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925070504759355e-05, + "loss": 1.0793, + "step": 5507 + }, + { + "epoch": 0.41, + "learning_rate": 1.9925021759787592e-05, + "loss": 1.0812, + "step": 5508 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924972999025268e-05, + "loss": 1.0838, + "step": 5509 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924924222472447e-05, + "loss": 1.0157, + "step": 5510 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924875430129217e-05, + "loss": 0.96, + "step": 5511 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924826621995653e-05, + "loss": 1.0597, + "step": 5512 + }, + { + "epoch": 0.41, + "learning_rate": 1.992477779807183e-05, + "loss": 0.9252, + "step": 5513 + }, + { + "epoch": 0.41, + "learning_rate": 1.992472895835783e-05, + "loss": 0.9923, + "step": 5514 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924680102853725e-05, + "loss": 0.9646, + "step": 5515 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924631231559597e-05, + "loss": 1.1453, + "step": 5516 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924582344475523e-05, + "loss": 1.0002, + "step": 5517 + }, + { + "epoch": 0.41, + "learning_rate": 1.992453344160158e-05, + "loss": 1.1389, + "step": 5518 + }, + { + "epoch": 0.41, + "learning_rate": 1.992448452293785e-05, + "loss": 1.0309, + "step": 5519 + }, + { + "epoch": 0.41, + "learning_rate": 1.99244355884844e-05, + "loss": 0.9763, + "step": 5520 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924386638241318e-05, + "loss": 1.1649, + "step": 5521 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924337672208684e-05, + "loss": 1.0505, + "step": 5522 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924288690386566e-05, + "loss": 1.019, + "step": 5523 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924239692775045e-05, + "loss": 1.0725, + "step": 5524 + }, + { + "epoch": 0.41, + "learning_rate": 1.99241906793742e-05, + "loss": 1.0574, + "step": 5525 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924141650184117e-05, + "loss": 1.0397, + "step": 5526 + }, + { + "epoch": 0.41, + "learning_rate": 1.992409260520486e-05, + "loss": 1.0748, + "step": 5527 + }, + { + "epoch": 0.41, + "learning_rate": 1.9924043544436514e-05, + "loss": 0.9626, + "step": 5528 + }, + { + "epoch": 0.41, + "learning_rate": 1.992399446787916e-05, + "loss": 1.0809, + "step": 5529 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923945375532874e-05, + "loss": 1.1079, + "step": 5530 + }, + { + "epoch": 0.41, + "learning_rate": 1.992389626739773e-05, + "loss": 0.9901, + "step": 5531 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923847143473808e-05, + "loss": 1.0236, + "step": 5532 + }, + { + "epoch": 0.41, + "learning_rate": 1.992379800376119e-05, + "loss": 1.0759, + "step": 5533 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923748848259947e-05, + "loss": 1.0549, + "step": 5534 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923699676970166e-05, + "loss": 1.1221, + "step": 5535 + }, + { + "epoch": 0.41, + "learning_rate": 1.992365048989192e-05, + "loss": 1.0686, + "step": 5536 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923601287025286e-05, + "loss": 1.0122, + "step": 5537 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923552068370347e-05, + "loss": 1.0604, + "step": 5538 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923502833927174e-05, + "loss": 1.0044, + "step": 5539 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923453583695857e-05, + "loss": 1.1065, + "step": 5540 + }, + { + "epoch": 0.41, + "learning_rate": 1.992340431767646e-05, + "loss": 1.0402, + "step": 5541 + }, + { + "epoch": 0.41, + "learning_rate": 1.992335503586907e-05, + "loss": 1.0329, + "step": 5542 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923305738273768e-05, + "loss": 1.1645, + "step": 5543 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923256424890628e-05, + "loss": 1.0269, + "step": 5544 + }, + { + "epoch": 0.41, + "learning_rate": 1.992320709571973e-05, + "loss": 1.0789, + "step": 5545 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923157750761146e-05, + "loss": 1.0398, + "step": 5546 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923108390014966e-05, + "loss": 1.0688, + "step": 5547 + }, + { + "epoch": 0.41, + "learning_rate": 1.9923059013481258e-05, + "loss": 0.9582, + "step": 5548 + }, + { + "epoch": 0.41, + "learning_rate": 1.992300962116011e-05, + "loss": 0.9544, + "step": 5549 + }, + { + "epoch": 0.41, + "learning_rate": 1.992296021305159e-05, + "loss": 1.0191, + "step": 5550 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922910789155785e-05, + "loss": 1.0728, + "step": 5551 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922861349472772e-05, + "loss": 1.0362, + "step": 5552 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922811894002623e-05, + "loss": 1.1263, + "step": 5553 + }, + { + "epoch": 0.41, + "learning_rate": 1.992276242274543e-05, + "loss": 1.039, + "step": 5554 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922712935701262e-05, + "loss": 1.0066, + "step": 5555 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922663432870198e-05, + "loss": 1.0501, + "step": 5556 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922613914252317e-05, + "loss": 1.0245, + "step": 5557 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922564379847703e-05, + "loss": 1.0232, + "step": 5558 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922514829656428e-05, + "loss": 1.0304, + "step": 5559 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922465263678573e-05, + "loss": 1.2169, + "step": 5560 + }, + { + "epoch": 0.41, + "learning_rate": 1.992241568191422e-05, + "loss": 0.9335, + "step": 5561 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922366084363446e-05, + "loss": 1.1381, + "step": 5562 + }, + { + "epoch": 0.41, + "learning_rate": 1.992231647102633e-05, + "loss": 1.0673, + "step": 5563 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922266841902947e-05, + "loss": 1.0636, + "step": 5564 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922217196993382e-05, + "loss": 1.037, + "step": 5565 + }, + { + "epoch": 0.41, + "learning_rate": 1.992216753629771e-05, + "loss": 1.0074, + "step": 5566 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922117859816016e-05, + "loss": 0.9796, + "step": 5567 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922068167548368e-05, + "loss": 1.0338, + "step": 5568 + }, + { + "epoch": 0.41, + "learning_rate": 1.9922018459494855e-05, + "loss": 1.0674, + "step": 5569 + }, + { + "epoch": 0.41, + "learning_rate": 1.992196873565555e-05, + "loss": 1.0862, + "step": 5570 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921918996030538e-05, + "loss": 1.0604, + "step": 5571 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921869240619897e-05, + "loss": 1.1257, + "step": 5572 + }, + { + "epoch": 0.41, + "learning_rate": 1.99218194694237e-05, + "loss": 1.1104, + "step": 5573 + }, + { + "epoch": 0.41, + "learning_rate": 1.992176968244203e-05, + "loss": 1.0806, + "step": 5574 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921719879674966e-05, + "loss": 1.0189, + "step": 5575 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921670061122586e-05, + "loss": 1.0132, + "step": 5576 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921620226784975e-05, + "loss": 1.0783, + "step": 5577 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921570376662206e-05, + "loss": 1.0492, + "step": 5578 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921520510754362e-05, + "loss": 0.9886, + "step": 5579 + }, + { + "epoch": 0.41, + "learning_rate": 1.992147062906152e-05, + "loss": 1.0405, + "step": 5580 + }, + { + "epoch": 0.41, + "learning_rate": 1.992142073158376e-05, + "loss": 0.8943, + "step": 5581 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921370818321163e-05, + "loss": 0.9736, + "step": 5582 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921320889273806e-05, + "loss": 1.1123, + "step": 5583 + }, + { + "epoch": 0.41, + "learning_rate": 1.992127094444177e-05, + "loss": 1.0695, + "step": 5584 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921220983825132e-05, + "loss": 1.0395, + "step": 5585 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921171007423975e-05, + "loss": 0.9966, + "step": 5586 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921121015238376e-05, + "loss": 1.0659, + "step": 5587 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921071007268416e-05, + "loss": 1.1389, + "step": 5588 + }, + { + "epoch": 0.41, + "learning_rate": 1.9921020983514176e-05, + "loss": 0.9703, + "step": 5589 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920970943975728e-05, + "loss": 0.8183, + "step": 5590 + }, + { + "epoch": 0.41, + "learning_rate": 1.992092088865316e-05, + "loss": 1.0619, + "step": 5591 + }, + { + "epoch": 0.41, + "learning_rate": 1.992087081754655e-05, + "loss": 1.1198, + "step": 5592 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920820730655976e-05, + "loss": 1.0433, + "step": 5593 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920770627981518e-05, + "loss": 1.0383, + "step": 5594 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920720509523256e-05, + "loss": 1.0475, + "step": 5595 + }, + { + "epoch": 0.41, + "learning_rate": 1.992067037528127e-05, + "loss": 1.0621, + "step": 5596 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920620225255637e-05, + "loss": 1.0276, + "step": 5597 + }, + { + "epoch": 0.41, + "learning_rate": 1.992057005944644e-05, + "loss": 1.0405, + "step": 5598 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920519877853758e-05, + "loss": 1.069, + "step": 5599 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920469680477673e-05, + "loss": 1.0078, + "step": 5600 + }, + { + "epoch": 0.41, + "learning_rate": 1.992041946731826e-05, + "loss": 0.9875, + "step": 5601 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920369238375605e-05, + "loss": 0.9981, + "step": 5602 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920318993649782e-05, + "loss": 1.0321, + "step": 5603 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920268733140876e-05, + "loss": 1.0971, + "step": 5604 + }, + { + "epoch": 0.41, + "learning_rate": 1.992021845684896e-05, + "loss": 1.0028, + "step": 5605 + }, + { + "epoch": 0.41, + "learning_rate": 1.992016816477412e-05, + "loss": 1.0905, + "step": 5606 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920117856916435e-05, + "loss": 1.0803, + "step": 5607 + }, + { + "epoch": 0.41, + "learning_rate": 1.9920067533275988e-05, + "loss": 1.2053, + "step": 5608 + }, + { + "epoch": 0.41, + "learning_rate": 1.992001719385285e-05, + "loss": 1.0996, + "step": 5609 + }, + { + "epoch": 0.41, + "learning_rate": 1.9919966838647107e-05, + "loss": 0.9627, + "step": 5610 + }, + { + "epoch": 0.41, + "learning_rate": 1.9919916467658843e-05, + "loss": 1.0581, + "step": 5611 + }, + { + "epoch": 0.41, + "learning_rate": 1.9919866080888132e-05, + "loss": 1.019, + "step": 5612 + }, + { + "epoch": 0.41, + "learning_rate": 1.9919815678335053e-05, + "loss": 1.0078, + "step": 5613 + }, + { + "epoch": 0.41, + "learning_rate": 1.991976525999969e-05, + "loss": 0.9725, + "step": 5614 + }, + { + "epoch": 0.41, + "learning_rate": 1.9919714825882125e-05, + "loss": 1.0252, + "step": 5615 + }, + { + "epoch": 0.41, + "learning_rate": 1.991966437598243e-05, + "loss": 0.9744, + "step": 5616 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919613910300698e-05, + "loss": 1.1251, + "step": 5617 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919563428837e-05, + "loss": 1.0783, + "step": 5618 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919512931591417e-05, + "loss": 1.0419, + "step": 5619 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919462418564032e-05, + "loss": 0.9687, + "step": 5620 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919411889754925e-05, + "loss": 1.1595, + "step": 5621 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919361345164172e-05, + "loss": 1.0237, + "step": 5622 + }, + { + "epoch": 0.42, + "learning_rate": 1.991931078479186e-05, + "loss": 1.0686, + "step": 5623 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919260208638067e-05, + "loss": 1.0692, + "step": 5624 + }, + { + "epoch": 0.42, + "learning_rate": 1.991920961670287e-05, + "loss": 1.0465, + "step": 5625 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919159008986354e-05, + "loss": 1.0481, + "step": 5626 + }, + { + "epoch": 0.42, + "learning_rate": 1.99191083854886e-05, + "loss": 0.892, + "step": 5627 + }, + { + "epoch": 0.42, + "learning_rate": 1.9919057746209683e-05, + "loss": 1.0589, + "step": 5628 + }, + { + "epoch": 0.42, + "learning_rate": 1.991900709114969e-05, + "loss": 1.0582, + "step": 5629 + }, + { + "epoch": 0.42, + "learning_rate": 1.99189564203087e-05, + "loss": 1.0125, + "step": 5630 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918905733686787e-05, + "loss": 1.0389, + "step": 5631 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918855031284037e-05, + "loss": 1.0897, + "step": 5632 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918804313100536e-05, + "loss": 0.9768, + "step": 5633 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918753579136356e-05, + "loss": 1.0137, + "step": 5634 + }, + { + "epoch": 0.42, + "learning_rate": 1.991870282939158e-05, + "loss": 1.0165, + "step": 5635 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918652063866293e-05, + "loss": 1.0776, + "step": 5636 + }, + { + "epoch": 0.42, + "learning_rate": 1.991860128256057e-05, + "loss": 1.0747, + "step": 5637 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918550485474493e-05, + "loss": 1.0334, + "step": 5638 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918499672608148e-05, + "loss": 1.038, + "step": 5639 + }, + { + "epoch": 0.42, + "learning_rate": 1.991844884396161e-05, + "loss": 1.0143, + "step": 5640 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918397999534962e-05, + "loss": 0.983, + "step": 5641 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918347139328284e-05, + "loss": 1.0313, + "step": 5642 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918296263341657e-05, + "loss": 1.0812, + "step": 5643 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918245371575165e-05, + "loss": 1.0219, + "step": 5644 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918194464028886e-05, + "loss": 0.9855, + "step": 5645 + }, + { + "epoch": 0.42, + "learning_rate": 1.99181435407029e-05, + "loss": 0.9811, + "step": 5646 + }, + { + "epoch": 0.42, + "learning_rate": 1.991809260159729e-05, + "loss": 0.9774, + "step": 5647 + }, + { + "epoch": 0.42, + "learning_rate": 1.9918041646712136e-05, + "loss": 0.9896, + "step": 5648 + }, + { + "epoch": 0.42, + "learning_rate": 1.991799067604752e-05, + "loss": 1.0511, + "step": 5649 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917939689603523e-05, + "loss": 1.0255, + "step": 5650 + }, + { + "epoch": 0.42, + "learning_rate": 1.991788868738023e-05, + "loss": 1.0963, + "step": 5651 + }, + { + "epoch": 0.42, + "learning_rate": 1.991783766937771e-05, + "loss": 1.1161, + "step": 5652 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917786635596057e-05, + "loss": 1.0065, + "step": 5653 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917735586035343e-05, + "loss": 1.0801, + "step": 5654 + }, + { + "epoch": 0.42, + "learning_rate": 1.991768452069566e-05, + "loss": 0.9681, + "step": 5655 + }, + { + "epoch": 0.42, + "learning_rate": 1.991763343957708e-05, + "loss": 1.0544, + "step": 5656 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917582342679686e-05, + "loss": 1.0569, + "step": 5657 + }, + { + "epoch": 0.42, + "learning_rate": 1.991753123000356e-05, + "loss": 0.9901, + "step": 5658 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917480101548787e-05, + "loss": 1.0266, + "step": 5659 + }, + { + "epoch": 0.42, + "learning_rate": 1.991742895731544e-05, + "loss": 1.0864, + "step": 5660 + }, + { + "epoch": 0.42, + "learning_rate": 1.991737779730361e-05, + "loss": 1.031, + "step": 5661 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917326621513374e-05, + "loss": 1.0493, + "step": 5662 + }, + { + "epoch": 0.42, + "learning_rate": 1.991727542994481e-05, + "loss": 0.9584, + "step": 5663 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917224222598004e-05, + "loss": 0.9743, + "step": 5664 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917172999473036e-05, + "loss": 1.0992, + "step": 5665 + }, + { + "epoch": 0.42, + "learning_rate": 1.991712176056999e-05, + "loss": 1.0316, + "step": 5666 + }, + { + "epoch": 0.42, + "learning_rate": 1.991707050588894e-05, + "loss": 1.0791, + "step": 5667 + }, + { + "epoch": 0.42, + "learning_rate": 1.9917019235429976e-05, + "loss": 1.0989, + "step": 5668 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916967949193177e-05, + "loss": 1.0133, + "step": 5669 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916916647178626e-05, + "loss": 1.1615, + "step": 5670 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916865329386397e-05, + "loss": 0.999, + "step": 5671 + }, + { + "epoch": 0.42, + "learning_rate": 1.991681399581658e-05, + "loss": 0.9793, + "step": 5672 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916762646469255e-05, + "loss": 1.0073, + "step": 5673 + }, + { + "epoch": 0.42, + "learning_rate": 1.99167112813445e-05, + "loss": 1.0498, + "step": 5674 + }, + { + "epoch": 0.42, + "learning_rate": 1.99166599004424e-05, + "loss": 1.0577, + "step": 5675 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916608503763037e-05, + "loss": 1.2026, + "step": 5676 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916557091306493e-05, + "loss": 1.0585, + "step": 5677 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916505663072846e-05, + "loss": 1.078, + "step": 5678 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916454219062177e-05, + "loss": 1.1843, + "step": 5679 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916402759274578e-05, + "loss": 1.0588, + "step": 5680 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916351283710122e-05, + "loss": 1.0748, + "step": 5681 + }, + { + "epoch": 0.42, + "learning_rate": 1.991629979236889e-05, + "loss": 1.0939, + "step": 5682 + }, + { + "epoch": 0.42, + "learning_rate": 1.991624828525097e-05, + "loss": 1.0577, + "step": 5683 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916196762356442e-05, + "loss": 0.9993, + "step": 5684 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916145223685383e-05, + "loss": 1.0525, + "step": 5685 + }, + { + "epoch": 0.42, + "learning_rate": 1.991609366923788e-05, + "loss": 1.0818, + "step": 5686 + }, + { + "epoch": 0.42, + "learning_rate": 1.9916042099014013e-05, + "loss": 1.0206, + "step": 5687 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915990513013866e-05, + "loss": 1.0645, + "step": 5688 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915938911237524e-05, + "loss": 1.1288, + "step": 5689 + }, + { + "epoch": 0.42, + "learning_rate": 1.991588729368506e-05, + "loss": 1.0104, + "step": 5690 + }, + { + "epoch": 0.42, + "learning_rate": 1.991583566035656e-05, + "loss": 1.0203, + "step": 5691 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915784011252114e-05, + "loss": 1.0592, + "step": 5692 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915732346371793e-05, + "loss": 1.0643, + "step": 5693 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915680665715684e-05, + "loss": 1.0216, + "step": 5694 + }, + { + "epoch": 0.42, + "learning_rate": 1.991562896928387e-05, + "loss": 1.099, + "step": 5695 + }, + { + "epoch": 0.42, + "learning_rate": 1.991557725707643e-05, + "loss": 0.9788, + "step": 5696 + }, + { + "epoch": 0.42, + "learning_rate": 1.991552552909345e-05, + "loss": 1.027, + "step": 5697 + }, + { + "epoch": 0.42, + "learning_rate": 1.991547378533501e-05, + "loss": 0.9772, + "step": 5698 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915422025801195e-05, + "loss": 1.0362, + "step": 5699 + }, + { + "epoch": 0.42, + "learning_rate": 1.991537025049208e-05, + "loss": 0.9302, + "step": 5700 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915318459407758e-05, + "loss": 0.9928, + "step": 5701 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915266652548307e-05, + "loss": 0.9936, + "step": 5702 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915214829913805e-05, + "loss": 0.9989, + "step": 5703 + }, + { + "epoch": 0.42, + "learning_rate": 1.991516299150434e-05, + "loss": 1.0538, + "step": 5704 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915111137319994e-05, + "loss": 0.9212, + "step": 5705 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915059267360847e-05, + "loss": 0.9905, + "step": 5706 + }, + { + "epoch": 0.42, + "learning_rate": 1.9915007381626982e-05, + "loss": 1.071, + "step": 5707 + }, + { + "epoch": 0.42, + "learning_rate": 1.991495548011848e-05, + "loss": 0.9896, + "step": 5708 + }, + { + "epoch": 0.42, + "learning_rate": 1.991490356283543e-05, + "loss": 0.991, + "step": 5709 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914851629777904e-05, + "loss": 1.1311, + "step": 5710 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914799680945997e-05, + "loss": 1.0696, + "step": 5711 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914747716339785e-05, + "loss": 1.0266, + "step": 5712 + }, + { + "epoch": 0.42, + "learning_rate": 1.991469573595935e-05, + "loss": 1.0987, + "step": 5713 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914643739804775e-05, + "loss": 0.9731, + "step": 5714 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914591727876143e-05, + "loss": 0.9859, + "step": 5715 + }, + { + "epoch": 0.42, + "learning_rate": 1.991453970017354e-05, + "loss": 1.0219, + "step": 5716 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914487656697043e-05, + "loss": 0.9996, + "step": 5717 + }, + { + "epoch": 0.42, + "learning_rate": 1.991443559744674e-05, + "loss": 0.9369, + "step": 5718 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914383522422712e-05, + "loss": 1.0605, + "step": 5719 + }, + { + "epoch": 0.42, + "learning_rate": 1.991433143162504e-05, + "loss": 0.9574, + "step": 5720 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914279325053812e-05, + "loss": 1.1123, + "step": 5721 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914227202709103e-05, + "loss": 0.9769, + "step": 5722 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914175064591004e-05, + "loss": 0.962, + "step": 5723 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914122910699592e-05, + "loss": 1.0496, + "step": 5724 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914070741034954e-05, + "loss": 1.0585, + "step": 5725 + }, + { + "epoch": 0.42, + "learning_rate": 1.9914018555597173e-05, + "loss": 1.1217, + "step": 5726 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913966354386323e-05, + "loss": 1.0079, + "step": 5727 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913914137402504e-05, + "loss": 1.0209, + "step": 5728 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913861904645783e-05, + "loss": 1.0808, + "step": 5729 + }, + { + "epoch": 0.42, + "learning_rate": 1.991380965611625e-05, + "loss": 1.0431, + "step": 5730 + }, + { + "epoch": 0.42, + "learning_rate": 1.991375739181399e-05, + "loss": 0.9912, + "step": 5731 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913705111739085e-05, + "loss": 1.0559, + "step": 5732 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913652815891612e-05, + "loss": 1.001, + "step": 5733 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913600504271665e-05, + "loss": 1.0471, + "step": 5734 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913548176879317e-05, + "loss": 0.8692, + "step": 5735 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913495833714657e-05, + "loss": 1.1045, + "step": 5736 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913443474777768e-05, + "loss": 1.0776, + "step": 5737 + }, + { + "epoch": 0.42, + "learning_rate": 1.991339110006873e-05, + "loss": 1.0668, + "step": 5738 + }, + { + "epoch": 0.42, + "learning_rate": 1.991333870958763e-05, + "loss": 1.0298, + "step": 5739 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913286303334553e-05, + "loss": 1.0608, + "step": 5740 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913233881309573e-05, + "loss": 1.0131, + "step": 5741 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913181443512783e-05, + "loss": 0.9864, + "step": 5742 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913128989944262e-05, + "loss": 0.9538, + "step": 5743 + }, + { + "epoch": 0.42, + "learning_rate": 1.9913076520604097e-05, + "loss": 1.0645, + "step": 5744 + }, + { + "epoch": 0.42, + "learning_rate": 1.991302403549237e-05, + "loss": 1.007, + "step": 5745 + }, + { + "epoch": 0.42, + "learning_rate": 1.991297153460916e-05, + "loss": 0.9767, + "step": 5746 + }, + { + "epoch": 0.42, + "learning_rate": 1.9912919017954556e-05, + "loss": 1.0026, + "step": 5747 + }, + { + "epoch": 0.42, + "learning_rate": 1.9912866485528637e-05, + "loss": 0.993, + "step": 5748 + }, + { + "epoch": 0.42, + "learning_rate": 1.9912813937331493e-05, + "loss": 1.0652, + "step": 5749 + }, + { + "epoch": 0.42, + "learning_rate": 1.9912761373363202e-05, + "loss": 1.0085, + "step": 5750 + }, + { + "epoch": 0.42, + "learning_rate": 1.991270879362385e-05, + "loss": 1.0112, + "step": 5751 + }, + { + "epoch": 0.43, + "learning_rate": 1.991265619811352e-05, + "loss": 1.1143, + "step": 5752 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912603586832294e-05, + "loss": 1.099, + "step": 5753 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912550959780262e-05, + "loss": 1.0979, + "step": 5754 + }, + { + "epoch": 0.43, + "learning_rate": 1.99124983169575e-05, + "loss": 1.0239, + "step": 5755 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912445658364096e-05, + "loss": 1.082, + "step": 5756 + }, + { + "epoch": 0.43, + "learning_rate": 1.991239298400013e-05, + "loss": 1.0349, + "step": 5757 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912340293865693e-05, + "loss": 1.12, + "step": 5758 + }, + { + "epoch": 0.43, + "learning_rate": 1.991228758796086e-05, + "loss": 1.0103, + "step": 5759 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912234866285723e-05, + "loss": 1.0904, + "step": 5760 + }, + { + "epoch": 0.43, + "learning_rate": 1.9912182128840362e-05, + "loss": 0.9286, + "step": 5761 + }, + { + "epoch": 0.43, + "learning_rate": 1.991212937562486e-05, + "loss": 1.07, + "step": 5762 + }, + { + "epoch": 0.43, + "learning_rate": 1.99120766066393e-05, + "loss": 1.0928, + "step": 5763 + }, + { + "epoch": 0.43, + "learning_rate": 1.991202382188377e-05, + "loss": 1.057, + "step": 5764 + }, + { + "epoch": 0.43, + "learning_rate": 1.991197102135835e-05, + "loss": 1.0692, + "step": 5765 + }, + { + "epoch": 0.43, + "learning_rate": 1.991191820506313e-05, + "loss": 1.0197, + "step": 5766 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911865372998188e-05, + "loss": 1.1364, + "step": 5767 + }, + { + "epoch": 0.43, + "learning_rate": 1.991181252516361e-05, + "loss": 1.0325, + "step": 5768 + }, + { + "epoch": 0.43, + "learning_rate": 1.991175966155948e-05, + "loss": 1.0075, + "step": 5769 + }, + { + "epoch": 0.43, + "learning_rate": 1.991170678218588e-05, + "loss": 0.9859, + "step": 5770 + }, + { + "epoch": 0.43, + "learning_rate": 1.99116538870429e-05, + "loss": 1.0144, + "step": 5771 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911600976130615e-05, + "loss": 1.1063, + "step": 5772 + }, + { + "epoch": 0.43, + "learning_rate": 1.991154804944912e-05, + "loss": 1.0989, + "step": 5773 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911495106998493e-05, + "loss": 1.0767, + "step": 5774 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911442148778815e-05, + "loss": 0.9699, + "step": 5775 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911389174790182e-05, + "loss": 1.0979, + "step": 5776 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911336185032666e-05, + "loss": 1.085, + "step": 5777 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911283179506357e-05, + "loss": 0.9742, + "step": 5778 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911230158211333e-05, + "loss": 0.9751, + "step": 5779 + }, + { + "epoch": 0.43, + "learning_rate": 1.991117712114769e-05, + "loss": 1.0681, + "step": 5780 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911124068315507e-05, + "loss": 1.0646, + "step": 5781 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911070999714862e-05, + "loss": 1.0366, + "step": 5782 + }, + { + "epoch": 0.43, + "learning_rate": 1.9911017915345846e-05, + "loss": 1.1213, + "step": 5783 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910964815208545e-05, + "loss": 1.0433, + "step": 5784 + }, + { + "epoch": 0.43, + "learning_rate": 1.991091169930304e-05, + "loss": 0.9512, + "step": 5785 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910858567629417e-05, + "loss": 1.0587, + "step": 5786 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910805420187756e-05, + "loss": 0.9407, + "step": 5787 + }, + { + "epoch": 0.43, + "learning_rate": 1.991075225697815e-05, + "loss": 0.9713, + "step": 5788 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910699078000675e-05, + "loss": 1.0871, + "step": 5789 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910645883255422e-05, + "loss": 1.0584, + "step": 5790 + }, + { + "epoch": 0.43, + "learning_rate": 1.991059267274247e-05, + "loss": 1.0908, + "step": 5791 + }, + { + "epoch": 0.43, + "learning_rate": 1.991053944646191e-05, + "loss": 1.0439, + "step": 5792 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910486204413824e-05, + "loss": 0.995, + "step": 5793 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910432946598295e-05, + "loss": 1.0874, + "step": 5794 + }, + { + "epoch": 0.43, + "learning_rate": 1.991037967301541e-05, + "loss": 1.0868, + "step": 5795 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910326383665253e-05, + "loss": 0.9702, + "step": 5796 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910273078547907e-05, + "loss": 1.1376, + "step": 5797 + }, + { + "epoch": 0.43, + "learning_rate": 1.991021975766346e-05, + "loss": 1.1214, + "step": 5798 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910166421011992e-05, + "loss": 1.0068, + "step": 5799 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910113068593592e-05, + "loss": 1.1082, + "step": 5800 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910059700408348e-05, + "loss": 1.0218, + "step": 5801 + }, + { + "epoch": 0.43, + "learning_rate": 1.9910006316456338e-05, + "loss": 1.0182, + "step": 5802 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909952916737646e-05, + "loss": 0.9892, + "step": 5803 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909899501252365e-05, + "loss": 1.0296, + "step": 5804 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909846070000574e-05, + "loss": 0.921, + "step": 5805 + }, + { + "epoch": 0.43, + "learning_rate": 1.990979262298236e-05, + "loss": 1.0721, + "step": 5806 + }, + { + "epoch": 0.43, + "learning_rate": 1.990973916019781e-05, + "loss": 0.9298, + "step": 5807 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909685681647005e-05, + "loss": 1.0314, + "step": 5808 + }, + { + "epoch": 0.43, + "learning_rate": 1.990963218733003e-05, + "loss": 1.1242, + "step": 5809 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909578677246975e-05, + "loss": 1.007, + "step": 5810 + }, + { + "epoch": 0.43, + "learning_rate": 1.990952515139792e-05, + "loss": 1.0896, + "step": 5811 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909471609782956e-05, + "loss": 1.1646, + "step": 5812 + }, + { + "epoch": 0.43, + "learning_rate": 1.990941805240216e-05, + "loss": 1.0058, + "step": 5813 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909364479255623e-05, + "loss": 0.9745, + "step": 5814 + }, + { + "epoch": 0.43, + "learning_rate": 1.990931089034343e-05, + "loss": 1.081, + "step": 5815 + }, + { + "epoch": 0.43, + "learning_rate": 1.9909257285665665e-05, + "loss": 0.9205, + "step": 5816 + }, + { + "epoch": 0.43, + "learning_rate": 1.990920366522241e-05, + "loss": 1.1158, + "step": 5817 + }, + { + "epoch": 0.43, + "learning_rate": 1.990915002901376e-05, + "loss": 1.1212, + "step": 5818 + }, + { + "epoch": 0.43, + "learning_rate": 1.990909637703979e-05, + "loss": 1.0705, + "step": 5819 + }, + { + "epoch": 0.43, + "learning_rate": 1.990904270930059e-05, + "loss": 1.0773, + "step": 5820 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908989025796245e-05, + "loss": 0.9951, + "step": 5821 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908935326526838e-05, + "loss": 1.0911, + "step": 5822 + }, + { + "epoch": 0.43, + "learning_rate": 1.990888161149246e-05, + "loss": 1.0805, + "step": 5823 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908827880693195e-05, + "loss": 1.0948, + "step": 5824 + }, + { + "epoch": 0.43, + "learning_rate": 1.990877413412912e-05, + "loss": 1.0452, + "step": 5825 + }, + { + "epoch": 0.43, + "learning_rate": 1.990872037180033e-05, + "loss": 1.0344, + "step": 5826 + }, + { + "epoch": 0.43, + "learning_rate": 1.990866659370691e-05, + "loss": 1.042, + "step": 5827 + }, + { + "epoch": 0.43, + "learning_rate": 1.990861279984894e-05, + "loss": 0.9527, + "step": 5828 + }, + { + "epoch": 0.43, + "learning_rate": 1.990855899022651e-05, + "loss": 1.0555, + "step": 5829 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908505164839704e-05, + "loss": 1.0967, + "step": 5830 + }, + { + "epoch": 0.43, + "learning_rate": 1.990845132368861e-05, + "loss": 0.9926, + "step": 5831 + }, + { + "epoch": 0.43, + "learning_rate": 1.990839746677331e-05, + "loss": 0.9903, + "step": 5832 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908343594093894e-05, + "loss": 1.0289, + "step": 5833 + }, + { + "epoch": 0.43, + "learning_rate": 1.990828970565044e-05, + "loss": 0.9996, + "step": 5834 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908235801443045e-05, + "loss": 0.9983, + "step": 5835 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908181881471784e-05, + "loss": 1.0171, + "step": 5836 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908127945736752e-05, + "loss": 0.9679, + "step": 5837 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908073994238023e-05, + "loss": 1.0426, + "step": 5838 + }, + { + "epoch": 0.43, + "learning_rate": 1.9908020026975693e-05, + "loss": 1.1916, + "step": 5839 + }, + { + "epoch": 0.43, + "learning_rate": 1.990796604394985e-05, + "loss": 0.9723, + "step": 5840 + }, + { + "epoch": 0.43, + "learning_rate": 1.990791204516057e-05, + "loss": 1.0654, + "step": 5841 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907858030607944e-05, + "loss": 1.2125, + "step": 5842 + }, + { + "epoch": 0.43, + "learning_rate": 1.990780400029206e-05, + "loss": 1.1169, + "step": 5843 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907749954213e-05, + "loss": 1.0829, + "step": 5844 + }, + { + "epoch": 0.43, + "learning_rate": 1.990769589237085e-05, + "loss": 1.0418, + "step": 5845 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907641814765696e-05, + "loss": 1.156, + "step": 5846 + }, + { + "epoch": 0.43, + "learning_rate": 1.990758772139763e-05, + "loss": 1.0317, + "step": 5847 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907533612266733e-05, + "loss": 1.1281, + "step": 5848 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907479487373093e-05, + "loss": 1.0615, + "step": 5849 + }, + { + "epoch": 0.43, + "learning_rate": 1.990742534671679e-05, + "loss": 1.1087, + "step": 5850 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907371190297918e-05, + "loss": 0.9262, + "step": 5851 + }, + { + "epoch": 0.43, + "learning_rate": 1.990731701811656e-05, + "loss": 1.0147, + "step": 5852 + }, + { + "epoch": 0.43, + "learning_rate": 1.99072628301728e-05, + "loss": 1.0649, + "step": 5853 + }, + { + "epoch": 0.43, + "learning_rate": 1.990720862646673e-05, + "loss": 1.1336, + "step": 5854 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907154406998432e-05, + "loss": 1.0402, + "step": 5855 + }, + { + "epoch": 0.43, + "learning_rate": 1.9907100171767993e-05, + "loss": 0.9763, + "step": 5856 + }, + { + "epoch": 0.43, + "learning_rate": 1.99070459207755e-05, + "loss": 1.1739, + "step": 5857 + }, + { + "epoch": 0.43, + "learning_rate": 1.990699165402104e-05, + "loss": 0.9379, + "step": 5858 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906937371504695e-05, + "loss": 0.9634, + "step": 5859 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906883073226555e-05, + "loss": 1.018, + "step": 5860 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906828759186707e-05, + "loss": 1.1146, + "step": 5861 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906774429385232e-05, + "loss": 1.0905, + "step": 5862 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906720083822226e-05, + "loss": 1.0086, + "step": 5863 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906665722497767e-05, + "loss": 1.108, + "step": 5864 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906611345411945e-05, + "loss": 1.0071, + "step": 5865 + }, + { + "epoch": 0.43, + "learning_rate": 1.990655695256485e-05, + "loss": 1.0825, + "step": 5866 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906502543956558e-05, + "loss": 1.0074, + "step": 5867 + }, + { + "epoch": 0.43, + "learning_rate": 1.990644811958717e-05, + "loss": 0.9724, + "step": 5868 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906393679456756e-05, + "loss": 1.0831, + "step": 5869 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906339223565418e-05, + "loss": 1.0655, + "step": 5870 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906284751913233e-05, + "loss": 1.1418, + "step": 5871 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906230264500293e-05, + "loss": 1.0147, + "step": 5872 + }, + { + "epoch": 0.43, + "learning_rate": 1.990617576132668e-05, + "loss": 1.052, + "step": 5873 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906121242392484e-05, + "loss": 1.0167, + "step": 5874 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906066707697792e-05, + "loss": 0.9861, + "step": 5875 + }, + { + "epoch": 0.43, + "learning_rate": 1.9906012157242686e-05, + "loss": 1.0459, + "step": 5876 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905957591027262e-05, + "loss": 1.0004, + "step": 5877 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905903009051594e-05, + "loss": 0.989, + "step": 5878 + }, + { + "epoch": 0.43, + "learning_rate": 1.990584841131578e-05, + "loss": 1.0642, + "step": 5879 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905793797819902e-05, + "loss": 0.9697, + "step": 5880 + }, + { + "epoch": 0.43, + "learning_rate": 1.990573916856405e-05, + "loss": 0.9468, + "step": 5881 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905684523548308e-05, + "loss": 1.0115, + "step": 5882 + }, + { + "epoch": 0.43, + "learning_rate": 1.990562986277276e-05, + "loss": 1.0167, + "step": 5883 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905575186237503e-05, + "loss": 0.9944, + "step": 5884 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905520493942614e-05, + "loss": 1.0747, + "step": 5885 + }, + { + "epoch": 0.43, + "learning_rate": 1.9905465785888185e-05, + "loss": 1.1447, + "step": 5886 + }, + { + "epoch": 0.43, + "learning_rate": 1.99054110620743e-05, + "loss": 1.0051, + "step": 5887 + }, + { + "epoch": 0.44, + "learning_rate": 1.990535632250105e-05, + "loss": 0.9442, + "step": 5888 + }, + { + "epoch": 0.44, + "learning_rate": 1.9905301567168515e-05, + "loss": 1.045, + "step": 5889 + }, + { + "epoch": 0.44, + "learning_rate": 1.990524679607679e-05, + "loss": 0.9244, + "step": 5890 + }, + { + "epoch": 0.44, + "learning_rate": 1.990519200922596e-05, + "loss": 1.2119, + "step": 5891 + }, + { + "epoch": 0.44, + "learning_rate": 1.990513720661611e-05, + "loss": 1.0377, + "step": 5892 + }, + { + "epoch": 0.44, + "learning_rate": 1.990508238824733e-05, + "loss": 1.0171, + "step": 5893 + }, + { + "epoch": 0.44, + "learning_rate": 1.99050275541197e-05, + "loss": 0.9399, + "step": 5894 + }, + { + "epoch": 0.44, + "learning_rate": 1.990497270423332e-05, + "loss": 0.9883, + "step": 5895 + }, + { + "epoch": 0.44, + "learning_rate": 1.990491783858827e-05, + "loss": 0.911, + "step": 5896 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904862957184635e-05, + "loss": 1.1209, + "step": 5897 + }, + { + "epoch": 0.44, + "learning_rate": 1.99048080600225e-05, + "loss": 1.0814, + "step": 5898 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904753147101967e-05, + "loss": 0.8598, + "step": 5899 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904698218423106e-05, + "loss": 1.0254, + "step": 5900 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904643273986014e-05, + "loss": 0.9053, + "step": 5901 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904588313790775e-05, + "loss": 1.0766, + "step": 5902 + }, + { + "epoch": 0.44, + "learning_rate": 1.990453333783748e-05, + "loss": 0.9251, + "step": 5903 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904478346126216e-05, + "loss": 1.0409, + "step": 5904 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904423338657062e-05, + "loss": 1.0738, + "step": 5905 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904368315430117e-05, + "loss": 1.1197, + "step": 5906 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904313276445463e-05, + "loss": 1.1057, + "step": 5907 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904258221703187e-05, + "loss": 1.1671, + "step": 5908 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904203151203377e-05, + "loss": 1.0145, + "step": 5909 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904148064946125e-05, + "loss": 0.9792, + "step": 5910 + }, + { + "epoch": 0.44, + "learning_rate": 1.9904092962931516e-05, + "loss": 1.0199, + "step": 5911 + }, + { + "epoch": 0.44, + "learning_rate": 1.990403784515963e-05, + "loss": 1.0711, + "step": 5912 + }, + { + "epoch": 0.44, + "learning_rate": 1.990398271163057e-05, + "loss": 0.9221, + "step": 5913 + }, + { + "epoch": 0.44, + "learning_rate": 1.990392756234441e-05, + "loss": 1.0058, + "step": 5914 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903872397301245e-05, + "loss": 1.0544, + "step": 5915 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903817216501158e-05, + "loss": 1.0995, + "step": 5916 + }, + { + "epoch": 0.44, + "learning_rate": 1.990376201994424e-05, + "loss": 1.0654, + "step": 5917 + }, + { + "epoch": 0.44, + "learning_rate": 1.990370680763058e-05, + "loss": 1.1265, + "step": 5918 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903651579560264e-05, + "loss": 1.0068, + "step": 5919 + }, + { + "epoch": 0.44, + "learning_rate": 1.990359633573338e-05, + "loss": 1.0749, + "step": 5920 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903541076150016e-05, + "loss": 1.0776, + "step": 5921 + }, + { + "epoch": 0.44, + "learning_rate": 1.990348580081026e-05, + "loss": 1.1242, + "step": 5922 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903430509714197e-05, + "loss": 1.0501, + "step": 5923 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903375202861922e-05, + "loss": 1.1519, + "step": 5924 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903319880253517e-05, + "loss": 1.0093, + "step": 5925 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903264541889072e-05, + "loss": 0.9824, + "step": 5926 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903209187768672e-05, + "loss": 1.176, + "step": 5927 + }, + { + "epoch": 0.44, + "learning_rate": 1.9903153817892413e-05, + "loss": 1.0396, + "step": 5928 + }, + { + "epoch": 0.44, + "learning_rate": 1.990309843226037e-05, + "loss": 1.0087, + "step": 5929 + }, + { + "epoch": 0.44, + "learning_rate": 1.990304303087265e-05, + "loss": 1.0539, + "step": 5930 + }, + { + "epoch": 0.44, + "learning_rate": 1.990298761372932e-05, + "loss": 1.1214, + "step": 5931 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902932180830485e-05, + "loss": 1.0731, + "step": 5932 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902876732176225e-05, + "loss": 1.022, + "step": 5933 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902821267766626e-05, + "loss": 1.1339, + "step": 5934 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902765787601785e-05, + "loss": 1.0604, + "step": 5935 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902710291681785e-05, + "loss": 1.2036, + "step": 5936 + }, + { + "epoch": 0.44, + "learning_rate": 1.990265478000671e-05, + "loss": 1.0648, + "step": 5937 + }, + { + "epoch": 0.44, + "learning_rate": 1.990259925257666e-05, + "loss": 1.1204, + "step": 5938 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902543709391712e-05, + "loss": 0.9655, + "step": 5939 + }, + { + "epoch": 0.44, + "learning_rate": 1.990248815045196e-05, + "loss": 1.1519, + "step": 5940 + }, + { + "epoch": 0.44, + "learning_rate": 1.990243257575749e-05, + "loss": 1.0554, + "step": 5941 + }, + { + "epoch": 0.44, + "learning_rate": 1.990237698530839e-05, + "loss": 1.0719, + "step": 5942 + }, + { + "epoch": 0.44, + "learning_rate": 1.990232137910475e-05, + "loss": 1.1177, + "step": 5943 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902265757146663e-05, + "loss": 1.0667, + "step": 5944 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902210119434205e-05, + "loss": 1.0708, + "step": 5945 + }, + { + "epoch": 0.44, + "learning_rate": 1.990215446596748e-05, + "loss": 1.0144, + "step": 5946 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902098796746565e-05, + "loss": 1.0931, + "step": 5947 + }, + { + "epoch": 0.44, + "learning_rate": 1.9902043111771555e-05, + "loss": 0.9556, + "step": 5948 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901987411042532e-05, + "loss": 1.0821, + "step": 5949 + }, + { + "epoch": 0.44, + "learning_rate": 1.990193169455959e-05, + "loss": 0.9586, + "step": 5950 + }, + { + "epoch": 0.44, + "learning_rate": 1.990187596232282e-05, + "loss": 1.0392, + "step": 5951 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901820214332307e-05, + "loss": 1.105, + "step": 5952 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901764450588136e-05, + "loss": 1.0791, + "step": 5953 + }, + { + "epoch": 0.44, + "learning_rate": 1.99017086710904e-05, + "loss": 1.0816, + "step": 5954 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901652875839188e-05, + "loss": 1.0807, + "step": 5955 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901597064834586e-05, + "loss": 1.0021, + "step": 5956 + }, + { + "epoch": 0.44, + "learning_rate": 1.990154123807669e-05, + "loss": 1.0657, + "step": 5957 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901485395565577e-05, + "loss": 1.1299, + "step": 5958 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901429537301345e-05, + "loss": 1.0539, + "step": 5959 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901373663284086e-05, + "loss": 1.0883, + "step": 5960 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901317773513874e-05, + "loss": 1.1388, + "step": 5961 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901261867990814e-05, + "loss": 0.9901, + "step": 5962 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901205946714985e-05, + "loss": 1.0476, + "step": 5963 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901150009686477e-05, + "loss": 1.103, + "step": 5964 + }, + { + "epoch": 0.44, + "learning_rate": 1.9901094056905386e-05, + "loss": 1.0249, + "step": 5965 + }, + { + "epoch": 0.44, + "learning_rate": 1.990103808837179e-05, + "loss": 0.9784, + "step": 5966 + }, + { + "epoch": 0.44, + "learning_rate": 1.990098210408579e-05, + "loss": 1.014, + "step": 5967 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900926104047463e-05, + "loss": 0.9503, + "step": 5968 + }, + { + "epoch": 0.44, + "learning_rate": 1.990087008825691e-05, + "loss": 0.9953, + "step": 5969 + }, + { + "epoch": 0.44, + "learning_rate": 1.990081405671421e-05, + "loss": 0.9824, + "step": 5970 + }, + { + "epoch": 0.44, + "learning_rate": 1.990075800941946e-05, + "loss": 1.1026, + "step": 5971 + }, + { + "epoch": 0.44, + "learning_rate": 1.990070194637274e-05, + "loss": 1.0499, + "step": 5972 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900645867574154e-05, + "loss": 1.0767, + "step": 5973 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900589773023776e-05, + "loss": 1.0338, + "step": 5974 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900533662721702e-05, + "loss": 1.1347, + "step": 5975 + }, + { + "epoch": 0.44, + "learning_rate": 1.990047753666802e-05, + "loss": 1.0602, + "step": 5976 + }, + { + "epoch": 0.44, + "learning_rate": 1.990042139486282e-05, + "loss": 1.147, + "step": 5977 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900365237306195e-05, + "loss": 1.1674, + "step": 5978 + }, + { + "epoch": 0.44, + "learning_rate": 1.9900309063998226e-05, + "loss": 0.9879, + "step": 5979 + }, + { + "epoch": 0.44, + "learning_rate": 1.990025287493901e-05, + "loss": 1.0594, + "step": 5980 + }, + { + "epoch": 0.44, + "learning_rate": 1.990019667012863e-05, + "loss": 1.085, + "step": 5981 + }, + { + "epoch": 0.44, + "learning_rate": 1.990014044956718e-05, + "loss": 1.1455, + "step": 5982 + }, + { + "epoch": 0.44, + "learning_rate": 1.990008421325475e-05, + "loss": 0.9673, + "step": 5983 + }, + { + "epoch": 0.44, + "learning_rate": 1.990002796119143e-05, + "loss": 1.0251, + "step": 5984 + }, + { + "epoch": 0.44, + "learning_rate": 1.98999716933773e-05, + "loss": 0.9972, + "step": 5985 + }, + { + "epoch": 0.44, + "learning_rate": 1.989991540981246e-05, + "loss": 0.8727, + "step": 5986 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899859110496997e-05, + "loss": 1.0178, + "step": 5987 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899802795431e-05, + "loss": 1.0422, + "step": 5988 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899746464614557e-05, + "loss": 1.0354, + "step": 5989 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899690118047758e-05, + "loss": 1.1199, + "step": 5990 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899633755730695e-05, + "loss": 1.1097, + "step": 5991 + }, + { + "epoch": 0.44, + "learning_rate": 1.989957737766346e-05, + "loss": 0.9628, + "step": 5992 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899520983846135e-05, + "loss": 0.9759, + "step": 5993 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899464574278814e-05, + "loss": 0.9751, + "step": 5994 + }, + { + "epoch": 0.44, + "learning_rate": 1.989940814896159e-05, + "loss": 1.1746, + "step": 5995 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899351707894542e-05, + "loss": 1.0983, + "step": 5996 + }, + { + "epoch": 0.44, + "learning_rate": 1.989929525107777e-05, + "loss": 1.007, + "step": 5997 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899238778511365e-05, + "loss": 1.0726, + "step": 5998 + }, + { + "epoch": 0.44, + "learning_rate": 1.989918229019541e-05, + "loss": 1.054, + "step": 5999 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899125786129997e-05, + "loss": 1.1878, + "step": 6000 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899069266315218e-05, + "loss": 1.0903, + "step": 6001 + }, + { + "epoch": 0.44, + "learning_rate": 1.9899012730751163e-05, + "loss": 0.9921, + "step": 6002 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898956179437916e-05, + "loss": 1.0014, + "step": 6003 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898899612375572e-05, + "loss": 0.9937, + "step": 6004 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898843029564222e-05, + "loss": 1.0138, + "step": 6005 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898786431003955e-05, + "loss": 1.1197, + "step": 6006 + }, + { + "epoch": 0.44, + "learning_rate": 1.989872981669486e-05, + "loss": 1.1295, + "step": 6007 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898673186637024e-05, + "loss": 1.0407, + "step": 6008 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898616540830544e-05, + "loss": 1.0168, + "step": 6009 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898559879275508e-05, + "loss": 1.0878, + "step": 6010 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898503201972002e-05, + "loss": 1.0821, + "step": 6011 + }, + { + "epoch": 0.44, + "learning_rate": 1.989844650892012e-05, + "loss": 1.0957, + "step": 6012 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898389800119952e-05, + "loss": 1.0473, + "step": 6013 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898333075571584e-05, + "loss": 1.0234, + "step": 6014 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898276335275115e-05, + "loss": 0.9128, + "step": 6015 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898219579230627e-05, + "loss": 1.0565, + "step": 6016 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898162807438214e-05, + "loss": 0.9507, + "step": 6017 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898106019897964e-05, + "loss": 1.0597, + "step": 6018 + }, + { + "epoch": 0.44, + "learning_rate": 1.9898049216609973e-05, + "loss": 1.0539, + "step": 6019 + }, + { + "epoch": 0.44, + "learning_rate": 1.9897992397574324e-05, + "loss": 1.0637, + "step": 6020 + }, + { + "epoch": 0.44, + "learning_rate": 1.989793556279111e-05, + "loss": 1.1544, + "step": 6021 + }, + { + "epoch": 0.44, + "learning_rate": 1.9897878712260425e-05, + "loss": 1.0639, + "step": 6022 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897821845982354e-05, + "loss": 1.0197, + "step": 6023 + }, + { + "epoch": 0.45, + "learning_rate": 1.989776496395699e-05, + "loss": 0.9494, + "step": 6024 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897708066184423e-05, + "loss": 1.0146, + "step": 6025 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897651152664746e-05, + "loss": 0.9898, + "step": 6026 + }, + { + "epoch": 0.45, + "learning_rate": 1.989759422339805e-05, + "loss": 1.0127, + "step": 6027 + }, + { + "epoch": 0.45, + "learning_rate": 1.989753727838442e-05, + "loss": 1.0344, + "step": 6028 + }, + { + "epoch": 0.45, + "learning_rate": 1.989748031762395e-05, + "loss": 0.9794, + "step": 6029 + }, + { + "epoch": 0.45, + "learning_rate": 1.989742334111673e-05, + "loss": 1.0876, + "step": 6030 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897366348862853e-05, + "loss": 1.0019, + "step": 6031 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897309340862407e-05, + "loss": 1.0154, + "step": 6032 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897252317115483e-05, + "loss": 1.0277, + "step": 6033 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897195277622173e-05, + "loss": 1.0831, + "step": 6034 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897138222382565e-05, + "loss": 0.9795, + "step": 6035 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897081151396752e-05, + "loss": 1.1394, + "step": 6036 + }, + { + "epoch": 0.45, + "learning_rate": 1.9897024064664825e-05, + "loss": 1.0845, + "step": 6037 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896966962186873e-05, + "loss": 1.0892, + "step": 6038 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896909843962992e-05, + "loss": 1.0291, + "step": 6039 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896852709993267e-05, + "loss": 0.9734, + "step": 6040 + }, + { + "epoch": 0.45, + "learning_rate": 1.989679556027779e-05, + "loss": 1.0653, + "step": 6041 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896738394816656e-05, + "loss": 1.0541, + "step": 6042 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896681213609946e-05, + "loss": 0.9239, + "step": 6043 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896624016657764e-05, + "loss": 1.1175, + "step": 6044 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896566803960192e-05, + "loss": 1.0601, + "step": 6045 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896509575517325e-05, + "loss": 1.1039, + "step": 6046 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896452331329254e-05, + "loss": 1.046, + "step": 6047 + }, + { + "epoch": 0.45, + "learning_rate": 1.989639507139607e-05, + "loss": 0.9131, + "step": 6048 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896337795717857e-05, + "loss": 1.0925, + "step": 6049 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896280504294716e-05, + "loss": 1.0311, + "step": 6050 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896223197126736e-05, + "loss": 0.9941, + "step": 6051 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896165874214003e-05, + "loss": 1.0419, + "step": 6052 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896108535556614e-05, + "loss": 0.9587, + "step": 6053 + }, + { + "epoch": 0.45, + "learning_rate": 1.9896051181154655e-05, + "loss": 1.086, + "step": 6054 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895993811008224e-05, + "loss": 1.1225, + "step": 6055 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895936425117405e-05, + "loss": 1.1075, + "step": 6056 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895879023482297e-05, + "loss": 1.16, + "step": 6057 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895821606102985e-05, + "loss": 1.0991, + "step": 6058 + }, + { + "epoch": 0.45, + "learning_rate": 1.989576417297956e-05, + "loss": 1.0569, + "step": 6059 + }, + { + "epoch": 0.45, + "learning_rate": 1.989570672411212e-05, + "loss": 1.1859, + "step": 6060 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895649259500748e-05, + "loss": 1.2308, + "step": 6061 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895591779145542e-05, + "loss": 1.0761, + "step": 6062 + }, + { + "epoch": 0.45, + "learning_rate": 1.989553428304659e-05, + "loss": 1.1224, + "step": 6063 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895476771203987e-05, + "loss": 1.1027, + "step": 6064 + }, + { + "epoch": 0.45, + "learning_rate": 1.989541924361782e-05, + "loss": 1.0533, + "step": 6065 + }, + { + "epoch": 0.45, + "learning_rate": 1.989536170028818e-05, + "loss": 1.0132, + "step": 6066 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895304141215165e-05, + "loss": 1.0437, + "step": 6067 + }, + { + "epoch": 0.45, + "learning_rate": 1.989524656639886e-05, + "loss": 1.1327, + "step": 6068 + }, + { + "epoch": 0.45, + "learning_rate": 1.989518897583936e-05, + "loss": 0.958, + "step": 6069 + }, + { + "epoch": 0.45, + "learning_rate": 1.989513136953676e-05, + "loss": 1.0275, + "step": 6070 + }, + { + "epoch": 0.45, + "learning_rate": 1.989507374749114e-05, + "loss": 1.0494, + "step": 6071 + }, + { + "epoch": 0.45, + "learning_rate": 1.9895016109702607e-05, + "loss": 1.1344, + "step": 6072 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894958456171244e-05, + "loss": 1.1325, + "step": 6073 + }, + { + "epoch": 0.45, + "learning_rate": 1.989490078689714e-05, + "loss": 1.1037, + "step": 6074 + }, + { + "epoch": 0.45, + "learning_rate": 1.989484310188039e-05, + "loss": 1.0239, + "step": 6075 + }, + { + "epoch": 0.45, + "learning_rate": 1.989478540112109e-05, + "loss": 1.0294, + "step": 6076 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894727684619325e-05, + "loss": 1.0668, + "step": 6077 + }, + { + "epoch": 0.45, + "learning_rate": 1.989466995237519e-05, + "loss": 1.0644, + "step": 6078 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894612204388776e-05, + "loss": 1.0097, + "step": 6079 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894554440660178e-05, + "loss": 0.9409, + "step": 6080 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894496661189483e-05, + "loss": 1.0249, + "step": 6081 + }, + { + "epoch": 0.45, + "learning_rate": 1.989443886597679e-05, + "loss": 0.95, + "step": 6082 + }, + { + "epoch": 0.45, + "learning_rate": 1.989438105502218e-05, + "loss": 1.0506, + "step": 6083 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894323228325757e-05, + "loss": 1.0668, + "step": 6084 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894265385887602e-05, + "loss": 0.9783, + "step": 6085 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894207527707817e-05, + "loss": 1.0435, + "step": 6086 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894149653786487e-05, + "loss": 1.0045, + "step": 6087 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894091764123707e-05, + "loss": 1.1017, + "step": 6088 + }, + { + "epoch": 0.45, + "learning_rate": 1.9894033858719566e-05, + "loss": 0.9583, + "step": 6089 + }, + { + "epoch": 0.45, + "learning_rate": 1.989397593757416e-05, + "loss": 0.9906, + "step": 6090 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893918000687583e-05, + "loss": 1.0631, + "step": 6091 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893860048059924e-05, + "loss": 0.9811, + "step": 6092 + }, + { + "epoch": 0.45, + "learning_rate": 1.989380207969127e-05, + "loss": 0.9295, + "step": 6093 + }, + { + "epoch": 0.45, + "learning_rate": 1.989374409558172e-05, + "loss": 1.0056, + "step": 6094 + }, + { + "epoch": 0.45, + "learning_rate": 1.989368609573137e-05, + "loss": 1.0107, + "step": 6095 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893628080140303e-05, + "loss": 0.93, + "step": 6096 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893570048808613e-05, + "loss": 0.99, + "step": 6097 + }, + { + "epoch": 0.45, + "learning_rate": 1.98935120017364e-05, + "loss": 0.8957, + "step": 6098 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893453938923747e-05, + "loss": 1.0746, + "step": 6099 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893395860370753e-05, + "loss": 1.098, + "step": 6100 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893337766077505e-05, + "loss": 0.9311, + "step": 6101 + }, + { + "epoch": 0.45, + "learning_rate": 1.98932796560441e-05, + "loss": 1.0161, + "step": 6102 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893221530270627e-05, + "loss": 0.9778, + "step": 6103 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893163388757182e-05, + "loss": 1.0555, + "step": 6104 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893105231503857e-05, + "loss": 1.0484, + "step": 6105 + }, + { + "epoch": 0.45, + "learning_rate": 1.9893047058510738e-05, + "loss": 1.0012, + "step": 6106 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892988869777925e-05, + "loss": 0.9448, + "step": 6107 + }, + { + "epoch": 0.45, + "learning_rate": 1.989293066530551e-05, + "loss": 0.9458, + "step": 6108 + }, + { + "epoch": 0.45, + "learning_rate": 1.989287244509358e-05, + "loss": 1.0281, + "step": 6109 + }, + { + "epoch": 0.45, + "learning_rate": 1.989281420914224e-05, + "loss": 0.9585, + "step": 6110 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892755957451563e-05, + "loss": 1.0521, + "step": 6111 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892697690021657e-05, + "loss": 0.9436, + "step": 6112 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892639406852615e-05, + "loss": 0.9531, + "step": 6113 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892581107944518e-05, + "loss": 1.1954, + "step": 6114 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892522793297472e-05, + "loss": 1.1093, + "step": 6115 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892464462911564e-05, + "loss": 0.889, + "step": 6116 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892406116786883e-05, + "loss": 1.0232, + "step": 6117 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892347754923523e-05, + "loss": 1.1171, + "step": 6118 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892289377321586e-05, + "loss": 1.039, + "step": 6119 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892230983981154e-05, + "loss": 1.105, + "step": 6120 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892172574902324e-05, + "loss": 1.0131, + "step": 6121 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892114150085192e-05, + "loss": 1.0175, + "step": 6122 + }, + { + "epoch": 0.45, + "learning_rate": 1.9892055709529846e-05, + "loss": 1.0882, + "step": 6123 + }, + { + "epoch": 0.45, + "learning_rate": 1.989199725323638e-05, + "loss": 1.076, + "step": 6124 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891938781204886e-05, + "loss": 1.0898, + "step": 6125 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891880293435463e-05, + "loss": 0.9932, + "step": 6126 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891821789928197e-05, + "loss": 1.02, + "step": 6127 + }, + { + "epoch": 0.45, + "learning_rate": 1.989176327068319e-05, + "loss": 1.0841, + "step": 6128 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891704735700525e-05, + "loss": 1.0379, + "step": 6129 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891646184980297e-05, + "loss": 1.0423, + "step": 6130 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891587618522603e-05, + "loss": 1.0799, + "step": 6131 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891529036327536e-05, + "loss": 1.0582, + "step": 6132 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891470438395186e-05, + "loss": 1.0099, + "step": 6133 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891411824725648e-05, + "loss": 1.108, + "step": 6134 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891353195319017e-05, + "loss": 0.972, + "step": 6135 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891294550175388e-05, + "loss": 1.1365, + "step": 6136 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891235889294842e-05, + "loss": 0.8948, + "step": 6137 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891177212677485e-05, + "loss": 1.0563, + "step": 6138 + }, + { + "epoch": 0.45, + "learning_rate": 1.989111852032341e-05, + "loss": 1.1018, + "step": 6139 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891059812232705e-05, + "loss": 1.0472, + "step": 6140 + }, + { + "epoch": 0.45, + "learning_rate": 1.9891001088405465e-05, + "loss": 1.0732, + "step": 6141 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890942348841782e-05, + "loss": 1.0096, + "step": 6142 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890883593541756e-05, + "loss": 1.0231, + "step": 6143 + }, + { + "epoch": 0.45, + "learning_rate": 1.989082482250547e-05, + "loss": 1.0528, + "step": 6144 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890766035733026e-05, + "loss": 1.0157, + "step": 6145 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890707233224517e-05, + "loss": 1.075, + "step": 6146 + }, + { + "epoch": 0.45, + "learning_rate": 1.989064841498003e-05, + "loss": 0.9945, + "step": 6147 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890589580999667e-05, + "loss": 0.8665, + "step": 6148 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890530731283514e-05, + "loss": 1.0159, + "step": 6149 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890471865831667e-05, + "loss": 1.1028, + "step": 6150 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890412984644225e-05, + "loss": 1.0148, + "step": 6151 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890354087721276e-05, + "loss": 1.0154, + "step": 6152 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890295175062914e-05, + "loss": 0.9928, + "step": 6153 + }, + { + "epoch": 0.45, + "learning_rate": 1.989023624666924e-05, + "loss": 1.0043, + "step": 6154 + }, + { + "epoch": 0.45, + "learning_rate": 1.989017730254033e-05, + "loss": 1.0002, + "step": 6155 + }, + { + "epoch": 0.45, + "learning_rate": 1.98901183426763e-05, + "loss": 1.0805, + "step": 6156 + }, + { + "epoch": 0.45, + "learning_rate": 1.9890059367077227e-05, + "loss": 1.0499, + "step": 6157 + }, + { + "epoch": 0.46, + "learning_rate": 1.9890000375743213e-05, + "loss": 1.0697, + "step": 6158 + }, + { + "epoch": 0.46, + "learning_rate": 1.988994136867435e-05, + "loss": 1.0645, + "step": 6159 + }, + { + "epoch": 0.46, + "learning_rate": 1.988988234587073e-05, + "loss": 1.0923, + "step": 6160 + }, + { + "epoch": 0.46, + "learning_rate": 1.988982330733245e-05, + "loss": 1.0427, + "step": 6161 + }, + { + "epoch": 0.46, + "learning_rate": 1.9889764253059606e-05, + "loss": 1.1326, + "step": 6162 + }, + { + "epoch": 0.46, + "learning_rate": 1.9889705183052284e-05, + "loss": 1.0399, + "step": 6163 + }, + { + "epoch": 0.46, + "learning_rate": 1.9889646097310583e-05, + "loss": 0.9728, + "step": 6164 + }, + { + "epoch": 0.46, + "learning_rate": 1.98895869958346e-05, + "loss": 1.044, + "step": 6165 + }, + { + "epoch": 0.46, + "learning_rate": 1.988952787862442e-05, + "loss": 1.1432, + "step": 6166 + }, + { + "epoch": 0.46, + "learning_rate": 1.988946874568015e-05, + "loss": 1.0267, + "step": 6167 + }, + { + "epoch": 0.46, + "learning_rate": 1.988940959700187e-05, + "loss": 0.9392, + "step": 6168 + }, + { + "epoch": 0.46, + "learning_rate": 1.9889350432589683e-05, + "loss": 1.1325, + "step": 6169 + }, + { + "epoch": 0.46, + "learning_rate": 1.988929125244368e-05, + "loss": 1.0951, + "step": 6170 + }, + { + "epoch": 0.46, + "learning_rate": 1.988923205656396e-05, + "loss": 1.1145, + "step": 6171 + }, + { + "epoch": 0.46, + "learning_rate": 1.988917284495061e-05, + "loss": 1.0421, + "step": 6172 + }, + { + "epoch": 0.46, + "learning_rate": 1.9889113617603727e-05, + "loss": 1.0514, + "step": 6173 + }, + { + "epoch": 0.46, + "learning_rate": 1.988905437452341e-05, + "loss": 0.966, + "step": 6174 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888995115709747e-05, + "loss": 0.9624, + "step": 6175 + }, + { + "epoch": 0.46, + "learning_rate": 1.988893584116283e-05, + "loss": 1.05, + "step": 6176 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888876550882764e-05, + "loss": 1.1395, + "step": 6177 + }, + { + "epoch": 0.46, + "learning_rate": 1.988881724486963e-05, + "loss": 0.9097, + "step": 6178 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888757923123534e-05, + "loss": 1.0492, + "step": 6179 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888698585644566e-05, + "loss": 1.0373, + "step": 6180 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888639232432818e-05, + "loss": 0.9706, + "step": 6181 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888579863488386e-05, + "loss": 0.9816, + "step": 6182 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888520478811366e-05, + "loss": 0.9286, + "step": 6183 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888461078401852e-05, + "loss": 1.0285, + "step": 6184 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888401662259937e-05, + "loss": 1.0278, + "step": 6185 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888342230385717e-05, + "loss": 1.0873, + "step": 6186 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888282782779286e-05, + "loss": 1.0642, + "step": 6187 + }, + { + "epoch": 0.46, + "learning_rate": 1.988822331944074e-05, + "loss": 0.9856, + "step": 6188 + }, + { + "epoch": 0.46, + "learning_rate": 1.988816384037017e-05, + "loss": 0.9914, + "step": 6189 + }, + { + "epoch": 0.46, + "learning_rate": 1.988810434556767e-05, + "loss": 1.1134, + "step": 6190 + }, + { + "epoch": 0.46, + "learning_rate": 1.9888044835033344e-05, + "loss": 1.055, + "step": 6191 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887985308767276e-05, + "loss": 1.0194, + "step": 6192 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887925766769567e-05, + "loss": 1.1396, + "step": 6193 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887866209040308e-05, + "loss": 1.0734, + "step": 6194 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887806635579598e-05, + "loss": 1.0284, + "step": 6195 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887747046387523e-05, + "loss": 1.074, + "step": 6196 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887687441464187e-05, + "loss": 0.9885, + "step": 6197 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887627820809684e-05, + "loss": 1.1039, + "step": 6198 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887568184424106e-05, + "loss": 1.0902, + "step": 6199 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887508532307546e-05, + "loss": 0.9559, + "step": 6200 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887448864460105e-05, + "loss": 0.9626, + "step": 6201 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887389180881872e-05, + "loss": 1.1246, + "step": 6202 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887329481572943e-05, + "loss": 1.0216, + "step": 6203 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887269766533416e-05, + "loss": 1.0531, + "step": 6204 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887210035763383e-05, + "loss": 1.1148, + "step": 6205 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887150289262946e-05, + "loss": 1.1611, + "step": 6206 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887090527032188e-05, + "loss": 1.0667, + "step": 6207 + }, + { + "epoch": 0.46, + "learning_rate": 1.9887030749071212e-05, + "loss": 1.0096, + "step": 6208 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886970955380112e-05, + "loss": 1.1049, + "step": 6209 + }, + { + "epoch": 0.46, + "learning_rate": 1.988691114595898e-05, + "loss": 0.9765, + "step": 6210 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886851320807915e-05, + "loss": 1.0437, + "step": 6211 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886791479927014e-05, + "loss": 1.0593, + "step": 6212 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886731623316366e-05, + "loss": 0.9883, + "step": 6213 + }, + { + "epoch": 0.46, + "learning_rate": 1.988667175097607e-05, + "loss": 1.1156, + "step": 6214 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886611862906223e-05, + "loss": 1.0576, + "step": 6215 + }, + { + "epoch": 0.46, + "learning_rate": 1.988655195910692e-05, + "loss": 1.0152, + "step": 6216 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886492039578246e-05, + "loss": 1.0317, + "step": 6217 + }, + { + "epoch": 0.46, + "learning_rate": 1.988643210432031e-05, + "loss": 1.0401, + "step": 6218 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886372153333204e-05, + "loss": 0.9813, + "step": 6219 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886312186617017e-05, + "loss": 1.0912, + "step": 6220 + }, + { + "epoch": 0.46, + "learning_rate": 1.988625220417185e-05, + "loss": 1.1309, + "step": 6221 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886192205997797e-05, + "loss": 1.0436, + "step": 6222 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886132192094957e-05, + "loss": 1.0794, + "step": 6223 + }, + { + "epoch": 0.46, + "learning_rate": 1.988607216246342e-05, + "loss": 1.1118, + "step": 6224 + }, + { + "epoch": 0.46, + "learning_rate": 1.9886012117103283e-05, + "loss": 0.9566, + "step": 6225 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885952056014642e-05, + "loss": 1.1041, + "step": 6226 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885891979197593e-05, + "loss": 1.0183, + "step": 6227 + }, + { + "epoch": 0.46, + "learning_rate": 1.988583188665223e-05, + "loss": 1.111, + "step": 6228 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885771778378653e-05, + "loss": 1.083, + "step": 6229 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885711654376952e-05, + "loss": 0.9872, + "step": 6230 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885651514647225e-05, + "loss": 1.0392, + "step": 6231 + }, + { + "epoch": 0.46, + "learning_rate": 1.988559135918957e-05, + "loss": 0.9813, + "step": 6232 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885531188004078e-05, + "loss": 1.0077, + "step": 6233 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885471001090848e-05, + "loss": 1.0082, + "step": 6234 + }, + { + "epoch": 0.46, + "learning_rate": 1.988541079844998e-05, + "loss": 0.9977, + "step": 6235 + }, + { + "epoch": 0.46, + "learning_rate": 1.988535058008156e-05, + "loss": 0.9673, + "step": 6236 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885290345985688e-05, + "loss": 1.1342, + "step": 6237 + }, + { + "epoch": 0.46, + "learning_rate": 1.988523009616246e-05, + "loss": 1.0524, + "step": 6238 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885169830611976e-05, + "loss": 1.0428, + "step": 6239 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885109549334326e-05, + "loss": 1.0395, + "step": 6240 + }, + { + "epoch": 0.46, + "learning_rate": 1.9885049252329608e-05, + "loss": 1.1603, + "step": 6241 + }, + { + "epoch": 0.46, + "learning_rate": 1.988498893959792e-05, + "loss": 1.069, + "step": 6242 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884928611139354e-05, + "loss": 1.0046, + "step": 6243 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884868266954006e-05, + "loss": 1.0524, + "step": 6244 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884807907041978e-05, + "loss": 0.9175, + "step": 6245 + }, + { + "epoch": 0.46, + "learning_rate": 1.988474753140336e-05, + "loss": 1.1009, + "step": 6246 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884687140038254e-05, + "loss": 1.0598, + "step": 6247 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884626732946748e-05, + "loss": 0.9561, + "step": 6248 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884566310128943e-05, + "loss": 1.0666, + "step": 6249 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884505871584935e-05, + "loss": 0.9502, + "step": 6250 + }, + { + "epoch": 0.46, + "learning_rate": 1.988444541731482e-05, + "loss": 1.0827, + "step": 6251 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884384947318692e-05, + "loss": 1.0464, + "step": 6252 + }, + { + "epoch": 0.46, + "learning_rate": 1.988432446159665e-05, + "loss": 1.0016, + "step": 6253 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884263960148787e-05, + "loss": 1.1655, + "step": 6254 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884203442975206e-05, + "loss": 1.0881, + "step": 6255 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884142910075997e-05, + "loss": 1.0159, + "step": 6256 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884082361451258e-05, + "loss": 1.0278, + "step": 6257 + }, + { + "epoch": 0.46, + "learning_rate": 1.9884021797101086e-05, + "loss": 1.0919, + "step": 6258 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883961217025574e-05, + "loss": 1.0791, + "step": 6259 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883900621224824e-05, + "loss": 1.066, + "step": 6260 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883840009698927e-05, + "loss": 1.0505, + "step": 6261 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883779382447983e-05, + "loss": 0.9676, + "step": 6262 + }, + { + "epoch": 0.46, + "learning_rate": 1.988371873947209e-05, + "loss": 1.0805, + "step": 6263 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883658080771336e-05, + "loss": 1.0724, + "step": 6264 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883597406345827e-05, + "loss": 1.062, + "step": 6265 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883536716195656e-05, + "loss": 1.0025, + "step": 6266 + }, + { + "epoch": 0.46, + "learning_rate": 1.988347601032092e-05, + "loss": 0.9475, + "step": 6267 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883415288721718e-05, + "loss": 0.9333, + "step": 6268 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883354551398138e-05, + "loss": 0.8978, + "step": 6269 + }, + { + "epoch": 0.46, + "learning_rate": 1.988329379835028e-05, + "loss": 1.0421, + "step": 6270 + }, + { + "epoch": 0.46, + "learning_rate": 1.988323302957825e-05, + "loss": 1.0573, + "step": 6271 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883172245082136e-05, + "loss": 0.9474, + "step": 6272 + }, + { + "epoch": 0.46, + "learning_rate": 1.9883111444862035e-05, + "loss": 1.0931, + "step": 6273 + }, + { + "epoch": 0.46, + "learning_rate": 1.988305062891805e-05, + "loss": 1.0704, + "step": 6274 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882989797250265e-05, + "loss": 0.979, + "step": 6275 + }, + { + "epoch": 0.46, + "learning_rate": 1.988292894985879e-05, + "loss": 1.0615, + "step": 6276 + }, + { + "epoch": 0.46, + "learning_rate": 1.988286808674371e-05, + "loss": 0.9954, + "step": 6277 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882807207905136e-05, + "loss": 0.9437, + "step": 6278 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882746313343153e-05, + "loss": 0.9807, + "step": 6279 + }, + { + "epoch": 0.46, + "learning_rate": 1.988268540305786e-05, + "loss": 1.0232, + "step": 6280 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882624477049363e-05, + "loss": 0.9726, + "step": 6281 + }, + { + "epoch": 0.46, + "learning_rate": 1.988256353531775e-05, + "loss": 1.1858, + "step": 6282 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882502577863116e-05, + "loss": 1.0725, + "step": 6283 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882441604685562e-05, + "loss": 1.1254, + "step": 6284 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882380615785187e-05, + "loss": 0.95, + "step": 6285 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882319611162084e-05, + "loss": 1.0914, + "step": 6286 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882258590816353e-05, + "loss": 1.0672, + "step": 6287 + }, + { + "epoch": 0.46, + "learning_rate": 1.988219755474809e-05, + "loss": 1.0325, + "step": 6288 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882136502957392e-05, + "loss": 1.082, + "step": 6289 + }, + { + "epoch": 0.46, + "learning_rate": 1.9882075435444353e-05, + "loss": 1.152, + "step": 6290 + }, + { + "epoch": 0.46, + "learning_rate": 1.988201435220908e-05, + "loss": 1.0262, + "step": 6291 + }, + { + "epoch": 0.46, + "learning_rate": 1.988195325325166e-05, + "loss": 0.9896, + "step": 6292 + }, + { + "epoch": 0.46, + "learning_rate": 1.9881892138572193e-05, + "loss": 1.0016, + "step": 6293 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881831008170776e-05, + "loss": 1.0937, + "step": 6294 + }, + { + "epoch": 0.47, + "learning_rate": 1.988176986204751e-05, + "loss": 0.9869, + "step": 6295 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881708700202488e-05, + "loss": 1.1762, + "step": 6296 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881647522635813e-05, + "loss": 1.0614, + "step": 6297 + }, + { + "epoch": 0.47, + "learning_rate": 1.988158632934757e-05, + "loss": 1.1677, + "step": 6298 + }, + { + "epoch": 0.47, + "learning_rate": 1.988152512033787e-05, + "loss": 0.9646, + "step": 6299 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881463895606805e-05, + "loss": 1.1126, + "step": 6300 + }, + { + "epoch": 0.47, + "learning_rate": 1.988140265515447e-05, + "loss": 1.0329, + "step": 6301 + }, + { + "epoch": 0.47, + "learning_rate": 1.988134139898097e-05, + "loss": 1.0744, + "step": 6302 + }, + { + "epoch": 0.47, + "learning_rate": 1.988128012708639e-05, + "loss": 1.056, + "step": 6303 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881218839470838e-05, + "loss": 0.9917, + "step": 6304 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881157536134408e-05, + "loss": 0.9958, + "step": 6305 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881096217077198e-05, + "loss": 1.0469, + "step": 6306 + }, + { + "epoch": 0.47, + "learning_rate": 1.9881034882299308e-05, + "loss": 1.0613, + "step": 6307 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880973531800828e-05, + "loss": 0.9736, + "step": 6308 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880912165581864e-05, + "loss": 0.9361, + "step": 6309 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880850783642507e-05, + "loss": 1.011, + "step": 6310 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880789385982863e-05, + "loss": 1.0668, + "step": 6311 + }, + { + "epoch": 0.47, + "learning_rate": 1.988072797260302e-05, + "loss": 1.0159, + "step": 6312 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880666543503076e-05, + "loss": 0.9261, + "step": 6313 + }, + { + "epoch": 0.47, + "learning_rate": 1.988060509868314e-05, + "loss": 1.0133, + "step": 6314 + }, + { + "epoch": 0.47, + "learning_rate": 1.98805436381433e-05, + "loss": 0.9782, + "step": 6315 + }, + { + "epoch": 0.47, + "learning_rate": 1.988048216188366e-05, + "loss": 1.022, + "step": 6316 + }, + { + "epoch": 0.47, + "learning_rate": 1.988042066990431e-05, + "loss": 1.0263, + "step": 6317 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880359162205354e-05, + "loss": 1.0126, + "step": 6318 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880297638786887e-05, + "loss": 0.9951, + "step": 6319 + }, + { + "epoch": 0.47, + "learning_rate": 1.988023609964901e-05, + "loss": 1.0622, + "step": 6320 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880174544791817e-05, + "loss": 1.068, + "step": 6321 + }, + { + "epoch": 0.47, + "learning_rate": 1.9880112974215408e-05, + "loss": 1.147, + "step": 6322 + }, + { + "epoch": 0.47, + "learning_rate": 1.988005138791988e-05, + "loss": 1.0514, + "step": 6323 + }, + { + "epoch": 0.47, + "learning_rate": 1.987998978590533e-05, + "loss": 1.0485, + "step": 6324 + }, + { + "epoch": 0.47, + "learning_rate": 1.987992816817186e-05, + "loss": 0.9617, + "step": 6325 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879866534719566e-05, + "loss": 1.0278, + "step": 6326 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879804885548545e-05, + "loss": 1.0078, + "step": 6327 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879743220658893e-05, + "loss": 1.0363, + "step": 6328 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879681540050714e-05, + "loss": 1.0481, + "step": 6329 + }, + { + "epoch": 0.47, + "learning_rate": 1.98796198437241e-05, + "loss": 1.0634, + "step": 6330 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879558131679156e-05, + "loss": 1.1028, + "step": 6331 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879496403915976e-05, + "loss": 1.0291, + "step": 6332 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879434660434656e-05, + "loss": 1.0712, + "step": 6333 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879372901235302e-05, + "loss": 1.0216, + "step": 6334 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879311126318e-05, + "loss": 1.1854, + "step": 6335 + }, + { + "epoch": 0.47, + "learning_rate": 1.987924933568286e-05, + "loss": 1.0122, + "step": 6336 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879187529329976e-05, + "loss": 1.1451, + "step": 6337 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879125707259442e-05, + "loss": 1.0461, + "step": 6338 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879063869471365e-05, + "loss": 0.9992, + "step": 6339 + }, + { + "epoch": 0.47, + "learning_rate": 1.9879002015965835e-05, + "loss": 1.0573, + "step": 6340 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878940146742956e-05, + "loss": 1.0399, + "step": 6341 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878878261802824e-05, + "loss": 1.1033, + "step": 6342 + }, + { + "epoch": 0.47, + "learning_rate": 1.987881636114554e-05, + "loss": 1.003, + "step": 6343 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878754444771198e-05, + "loss": 1.072, + "step": 6344 + }, + { + "epoch": 0.47, + "learning_rate": 1.98786925126799e-05, + "loss": 1.1507, + "step": 6345 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878630564871743e-05, + "loss": 1.0692, + "step": 6346 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878568601346825e-05, + "loss": 1.0799, + "step": 6347 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878506622105248e-05, + "loss": 1.1286, + "step": 6348 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878444627147107e-05, + "loss": 1.0974, + "step": 6349 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878382616472504e-05, + "loss": 1.0434, + "step": 6350 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878320590081532e-05, + "loss": 0.9969, + "step": 6351 + }, + { + "epoch": 0.47, + "learning_rate": 1.98782585479743e-05, + "loss": 0.9989, + "step": 6352 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878196490150893e-05, + "loss": 0.9899, + "step": 6353 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878134416611415e-05, + "loss": 1.1525, + "step": 6354 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878072327355973e-05, + "loss": 0.9896, + "step": 6355 + }, + { + "epoch": 0.47, + "learning_rate": 1.9878010222384657e-05, + "loss": 1.0511, + "step": 6356 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877948101697567e-05, + "loss": 0.9624, + "step": 6357 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877885965294804e-05, + "loss": 0.9711, + "step": 6358 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877823813176465e-05, + "loss": 1.0977, + "step": 6359 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877761645342654e-05, + "loss": 1.0194, + "step": 6360 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877699461793457e-05, + "loss": 1.0672, + "step": 6361 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877637262528987e-05, + "loss": 0.9488, + "step": 6362 + }, + { + "epoch": 0.47, + "learning_rate": 1.987757504754934e-05, + "loss": 1.0004, + "step": 6363 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877512816854607e-05, + "loss": 1.0573, + "step": 6364 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877450570444893e-05, + "loss": 1.0242, + "step": 6365 + }, + { + "epoch": 0.47, + "learning_rate": 1.98773883083203e-05, + "loss": 1.012, + "step": 6366 + }, + { + "epoch": 0.47, + "learning_rate": 1.987732603048092e-05, + "loss": 1.0925, + "step": 6367 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877263736926856e-05, + "loss": 1.0364, + "step": 6368 + }, + { + "epoch": 0.47, + "learning_rate": 1.987720142765821e-05, + "loss": 1.0639, + "step": 6369 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877139102675077e-05, + "loss": 0.9713, + "step": 6370 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877076761977552e-05, + "loss": 0.9116, + "step": 6371 + }, + { + "epoch": 0.47, + "learning_rate": 1.9877014405565743e-05, + "loss": 1.0236, + "step": 6372 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876952033439748e-05, + "loss": 1.1281, + "step": 6373 + }, + { + "epoch": 0.47, + "learning_rate": 1.987688964559966e-05, + "loss": 1.1232, + "step": 6374 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876827242045583e-05, + "loss": 1.1682, + "step": 6375 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876764822777614e-05, + "loss": 1.084, + "step": 6376 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876702387795852e-05, + "loss": 1.0575, + "step": 6377 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876639937100402e-05, + "loss": 0.9734, + "step": 6378 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876577470691356e-05, + "loss": 1.0334, + "step": 6379 + }, + { + "epoch": 0.47, + "learning_rate": 1.987651498856882e-05, + "loss": 1.0767, + "step": 6380 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876452490732886e-05, + "loss": 1.0211, + "step": 6381 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876389977183657e-05, + "loss": 0.9924, + "step": 6382 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876327447921236e-05, + "loss": 1.0486, + "step": 6383 + }, + { + "epoch": 0.47, + "learning_rate": 1.987626490294572e-05, + "loss": 1.029, + "step": 6384 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876202342257208e-05, + "loss": 0.9887, + "step": 6385 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876139765855798e-05, + "loss": 1.085, + "step": 6386 + }, + { + "epoch": 0.47, + "learning_rate": 1.987607717374159e-05, + "loss": 1.1102, + "step": 6387 + }, + { + "epoch": 0.47, + "learning_rate": 1.9876014565914685e-05, + "loss": 0.9874, + "step": 6388 + }, + { + "epoch": 0.47, + "learning_rate": 1.987595194237518e-05, + "loss": 1.0087, + "step": 6389 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875889303123182e-05, + "loss": 1.0384, + "step": 6390 + }, + { + "epoch": 0.47, + "learning_rate": 1.987582664815878e-05, + "loss": 1.0951, + "step": 6391 + }, + { + "epoch": 0.47, + "learning_rate": 1.987576397748208e-05, + "loss": 1.0314, + "step": 6392 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875701291093182e-05, + "loss": 1.1092, + "step": 6393 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875638588992187e-05, + "loss": 0.9799, + "step": 6394 + }, + { + "epoch": 0.47, + "learning_rate": 1.987557587117919e-05, + "loss": 1.0389, + "step": 6395 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875513137654294e-05, + "loss": 1.0554, + "step": 6396 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875450388417596e-05, + "loss": 1.0733, + "step": 6397 + }, + { + "epoch": 0.47, + "learning_rate": 1.98753876234692e-05, + "loss": 1.0862, + "step": 6398 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875324842809202e-05, + "loss": 1.1691, + "step": 6399 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875262046437706e-05, + "loss": 1.0932, + "step": 6400 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875199234354805e-05, + "loss": 1.0494, + "step": 6401 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875136406560606e-05, + "loss": 0.9962, + "step": 6402 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875073563055206e-05, + "loss": 1.0769, + "step": 6403 + }, + { + "epoch": 0.47, + "learning_rate": 1.9875010703838703e-05, + "loss": 1.1131, + "step": 6404 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874947828911203e-05, + "loss": 1.1229, + "step": 6405 + }, + { + "epoch": 0.47, + "learning_rate": 1.98748849382728e-05, + "loss": 1.1526, + "step": 6406 + }, + { + "epoch": 0.47, + "learning_rate": 1.98748220319236e-05, + "loss": 1.0736, + "step": 6407 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874759109863696e-05, + "loss": 1.0273, + "step": 6408 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874696172093192e-05, + "loss": 1.1626, + "step": 6409 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874633218612188e-05, + "loss": 0.9791, + "step": 6410 + }, + { + "epoch": 0.47, + "learning_rate": 1.987457024942078e-05, + "loss": 0.9855, + "step": 6411 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874507264519075e-05, + "loss": 1.0878, + "step": 6412 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874444263907173e-05, + "loss": 1.0227, + "step": 6413 + }, + { + "epoch": 0.47, + "learning_rate": 1.987438124758517e-05, + "loss": 1.1513, + "step": 6414 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874318215553168e-05, + "loss": 1.0107, + "step": 6415 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874255167811265e-05, + "loss": 1.0909, + "step": 6416 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874192104359563e-05, + "loss": 1.0457, + "step": 6417 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874129025198162e-05, + "loss": 1.0495, + "step": 6418 + }, + { + "epoch": 0.47, + "learning_rate": 1.987406593032717e-05, + "loss": 1.0044, + "step": 6419 + }, + { + "epoch": 0.47, + "learning_rate": 1.9874002819746674e-05, + "loss": 1.0019, + "step": 6420 + }, + { + "epoch": 0.47, + "learning_rate": 1.987393969345678e-05, + "loss": 1.0184, + "step": 6421 + }, + { + "epoch": 0.47, + "learning_rate": 1.9873876551457592e-05, + "loss": 1.0328, + "step": 6422 + }, + { + "epoch": 0.47, + "learning_rate": 1.9873813393749207e-05, + "loss": 1.1013, + "step": 6423 + }, + { + "epoch": 0.47, + "learning_rate": 1.9873750220331724e-05, + "loss": 1.1351, + "step": 6424 + }, + { + "epoch": 0.47, + "learning_rate": 1.987368703120525e-05, + "loss": 1.0259, + "step": 6425 + }, + { + "epoch": 0.47, + "learning_rate": 1.987362382636988e-05, + "loss": 1.0499, + "step": 6426 + }, + { + "epoch": 0.47, + "learning_rate": 1.9873560605825712e-05, + "loss": 0.9634, + "step": 6427 + }, + { + "epoch": 0.47, + "learning_rate": 1.9873497369572852e-05, + "loss": 1.0009, + "step": 6428 + }, + { + "epoch": 0.48, + "learning_rate": 1.98734341176114e-05, + "loss": 1.0612, + "step": 6429 + }, + { + "epoch": 0.48, + "learning_rate": 1.9873370849941458e-05, + "loss": 1.0293, + "step": 6430 + }, + { + "epoch": 0.48, + "learning_rate": 1.987330756656312e-05, + "loss": 1.1496, + "step": 6431 + }, + { + "epoch": 0.48, + "learning_rate": 1.9873244267476496e-05, + "loss": 1.0652, + "step": 6432 + }, + { + "epoch": 0.48, + "learning_rate": 1.9873180952681676e-05, + "loss": 1.0416, + "step": 6433 + }, + { + "epoch": 0.48, + "learning_rate": 1.987311762217877e-05, + "loss": 1.1588, + "step": 6434 + }, + { + "epoch": 0.48, + "learning_rate": 1.9873054275967873e-05, + "loss": 1.0081, + "step": 6435 + }, + { + "epoch": 0.48, + "learning_rate": 1.987299091404909e-05, + "loss": 1.1282, + "step": 6436 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872927536422518e-05, + "loss": 1.0258, + "step": 6437 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872864143088258e-05, + "loss": 1.1283, + "step": 6438 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872800734046417e-05, + "loss": 1.0413, + "step": 6439 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872737309297088e-05, + "loss": 1.0111, + "step": 6440 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872673868840376e-05, + "loss": 1.0705, + "step": 6441 + }, + { + "epoch": 0.48, + "learning_rate": 1.987261041267638e-05, + "loss": 0.9993, + "step": 6442 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872546940805203e-05, + "loss": 1.0335, + "step": 6443 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872483453226947e-05, + "loss": 1.1109, + "step": 6444 + }, + { + "epoch": 0.48, + "learning_rate": 1.987241994994171e-05, + "loss": 1.0575, + "step": 6445 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872356430949594e-05, + "loss": 0.979, + "step": 6446 + }, + { + "epoch": 0.48, + "learning_rate": 1.98722928962507e-05, + "loss": 1.0969, + "step": 6447 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872229345845126e-05, + "loss": 0.9382, + "step": 6448 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872165779732978e-05, + "loss": 1.0595, + "step": 6449 + }, + { + "epoch": 0.48, + "learning_rate": 1.987210219791436e-05, + "loss": 1.1034, + "step": 6450 + }, + { + "epoch": 0.48, + "learning_rate": 1.9872038600389363e-05, + "loss": 0.9879, + "step": 6451 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871974987158097e-05, + "loss": 1.0798, + "step": 6452 + }, + { + "epoch": 0.48, + "learning_rate": 1.987191135822066e-05, + "loss": 1.0785, + "step": 6453 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871847713577154e-05, + "loss": 0.9726, + "step": 6454 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871784053227677e-05, + "loss": 1.0208, + "step": 6455 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871720377172333e-05, + "loss": 0.9867, + "step": 6456 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871656685411222e-05, + "loss": 1.0133, + "step": 6457 + }, + { + "epoch": 0.48, + "learning_rate": 1.987159297794445e-05, + "loss": 1.0459, + "step": 6458 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871529254772112e-05, + "loss": 1.0105, + "step": 6459 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871465515894316e-05, + "loss": 1.091, + "step": 6460 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871401761311157e-05, + "loss": 1.0395, + "step": 6461 + }, + { + "epoch": 0.48, + "learning_rate": 1.987133799102274e-05, + "loss": 1.0521, + "step": 6462 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871274205029163e-05, + "loss": 1.0637, + "step": 6463 + }, + { + "epoch": 0.48, + "learning_rate": 1.987121040333053e-05, + "loss": 0.9821, + "step": 6464 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871146585926943e-05, + "loss": 1.0434, + "step": 6465 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871082752818504e-05, + "loss": 0.9766, + "step": 6466 + }, + { + "epoch": 0.48, + "learning_rate": 1.9871018904005318e-05, + "loss": 1.0256, + "step": 6467 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870955039487473e-05, + "loss": 1.0128, + "step": 6468 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870891159265086e-05, + "loss": 1.0525, + "step": 6469 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870827263338254e-05, + "loss": 1.0659, + "step": 6470 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870763351707072e-05, + "loss": 1.0544, + "step": 6471 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870699424371648e-05, + "loss": 1.0969, + "step": 6472 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870635481332085e-05, + "loss": 1.1796, + "step": 6473 + }, + { + "epoch": 0.48, + "learning_rate": 1.987057152258848e-05, + "loss": 0.973, + "step": 6474 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870507548140937e-05, + "loss": 1.0779, + "step": 6475 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870443557989557e-05, + "loss": 0.9829, + "step": 6476 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870379552134443e-05, + "loss": 0.937, + "step": 6477 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870315530575698e-05, + "loss": 0.9465, + "step": 6478 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870251493313423e-05, + "loss": 1.0557, + "step": 6479 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870187440347716e-05, + "loss": 1.0805, + "step": 6480 + }, + { + "epoch": 0.48, + "learning_rate": 1.987012337167868e-05, + "loss": 1.0267, + "step": 6481 + }, + { + "epoch": 0.48, + "learning_rate": 1.9870059287306423e-05, + "loss": 0.9891, + "step": 6482 + }, + { + "epoch": 0.48, + "learning_rate": 1.986999518723104e-05, + "loss": 0.9509, + "step": 6483 + }, + { + "epoch": 0.48, + "learning_rate": 1.986993107145264e-05, + "loss": 1.0729, + "step": 6484 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869866939971318e-05, + "loss": 0.9997, + "step": 6485 + }, + { + "epoch": 0.48, + "learning_rate": 1.986980279278718e-05, + "loss": 1.1322, + "step": 6486 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869738629900328e-05, + "loss": 0.9723, + "step": 6487 + }, + { + "epoch": 0.48, + "learning_rate": 1.986967445131086e-05, + "loss": 1.0678, + "step": 6488 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869610257018884e-05, + "loss": 0.9226, + "step": 6489 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869546047024497e-05, + "loss": 1.2025, + "step": 6490 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869481821327806e-05, + "loss": 1.0563, + "step": 6491 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869417579928907e-05, + "loss": 1.1065, + "step": 6492 + }, + { + "epoch": 0.48, + "learning_rate": 1.986935332282791e-05, + "loss": 1.1128, + "step": 6493 + }, + { + "epoch": 0.48, + "learning_rate": 1.986928905002491e-05, + "loss": 1.0593, + "step": 6494 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869224761520013e-05, + "loss": 1.0516, + "step": 6495 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869160457313322e-05, + "loss": 1.0696, + "step": 6496 + }, + { + "epoch": 0.48, + "learning_rate": 1.9869096137404938e-05, + "loss": 1.0636, + "step": 6497 + }, + { + "epoch": 0.48, + "learning_rate": 1.986903180179496e-05, + "loss": 0.9745, + "step": 6498 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868967450483498e-05, + "loss": 1.0645, + "step": 6499 + }, + { + "epoch": 0.48, + "learning_rate": 1.986890308347065e-05, + "loss": 1.0304, + "step": 6500 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868838700756513e-05, + "loss": 1.0903, + "step": 6501 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868774302341198e-05, + "loss": 1.1034, + "step": 6502 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868709888224803e-05, + "loss": 1.0712, + "step": 6503 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868645458407432e-05, + "loss": 1.1158, + "step": 6504 + }, + { + "epoch": 0.48, + "learning_rate": 1.986858101288919e-05, + "loss": 1.0195, + "step": 6505 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868516551670175e-05, + "loss": 0.9796, + "step": 6506 + }, + { + "epoch": 0.48, + "learning_rate": 1.986845207475049e-05, + "loss": 1.0533, + "step": 6507 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868387582130243e-05, + "loss": 1.1064, + "step": 6508 + }, + { + "epoch": 0.48, + "learning_rate": 1.986832307380953e-05, + "loss": 1.0797, + "step": 6509 + }, + { + "epoch": 0.48, + "learning_rate": 1.986825854978846e-05, + "loss": 0.9765, + "step": 6510 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868194010067123e-05, + "loss": 1.1078, + "step": 6511 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868129454645636e-05, + "loss": 1.0289, + "step": 6512 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868064883524098e-05, + "loss": 1.1029, + "step": 6513 + }, + { + "epoch": 0.48, + "learning_rate": 1.9868000296702604e-05, + "loss": 0.9598, + "step": 6514 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867935694181267e-05, + "loss": 0.9227, + "step": 6515 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867871075960184e-05, + "loss": 1.0449, + "step": 6516 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867806442039464e-05, + "loss": 0.9568, + "step": 6517 + }, + { + "epoch": 0.48, + "learning_rate": 1.98677417924192e-05, + "loss": 1.0744, + "step": 6518 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867677127099503e-05, + "loss": 0.9749, + "step": 6519 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867612446080468e-05, + "loss": 1.0644, + "step": 6520 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867547749362206e-05, + "loss": 1.0872, + "step": 6521 + }, + { + "epoch": 0.48, + "learning_rate": 1.986748303694482e-05, + "loss": 0.977, + "step": 6522 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867418308828404e-05, + "loss": 1.032, + "step": 6523 + }, + { + "epoch": 0.48, + "learning_rate": 1.986735356501307e-05, + "loss": 1.101, + "step": 6524 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867288805498916e-05, + "loss": 1.0766, + "step": 6525 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867224030286047e-05, + "loss": 1.0548, + "step": 6526 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867159239374567e-05, + "loss": 1.0561, + "step": 6527 + }, + { + "epoch": 0.48, + "learning_rate": 1.9867094432764578e-05, + "loss": 1.1273, + "step": 6528 + }, + { + "epoch": 0.48, + "learning_rate": 1.986702961045618e-05, + "loss": 0.9762, + "step": 6529 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866964772449483e-05, + "loss": 0.9839, + "step": 6530 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866899918744585e-05, + "loss": 1.1379, + "step": 6531 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866835049341587e-05, + "loss": 1.0097, + "step": 6532 + }, + { + "epoch": 0.48, + "learning_rate": 1.98667701642406e-05, + "loss": 1.0189, + "step": 6533 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866705263441722e-05, + "loss": 1.1297, + "step": 6534 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866640346945054e-05, + "loss": 1.0396, + "step": 6535 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866575414750705e-05, + "loss": 1.0904, + "step": 6536 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866510466858774e-05, + "loss": 1.1026, + "step": 6537 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866445503269365e-05, + "loss": 0.9497, + "step": 6538 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866380523982586e-05, + "loss": 1.0228, + "step": 6539 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866315528998538e-05, + "loss": 0.925, + "step": 6540 + }, + { + "epoch": 0.48, + "learning_rate": 1.986625051831732e-05, + "loss": 0.9596, + "step": 6541 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866185491939038e-05, + "loss": 1.0047, + "step": 6542 + }, + { + "epoch": 0.48, + "learning_rate": 1.9866120449863797e-05, + "loss": 0.9734, + "step": 6543 + }, + { + "epoch": 0.48, + "learning_rate": 1.98660553920917e-05, + "loss": 1.0191, + "step": 6544 + }, + { + "epoch": 0.48, + "learning_rate": 1.986599031862285e-05, + "loss": 1.1057, + "step": 6545 + }, + { + "epoch": 0.48, + "learning_rate": 1.986592522945735e-05, + "loss": 0.9372, + "step": 6546 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865860124595304e-05, + "loss": 1.0707, + "step": 6547 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865795004036815e-05, + "loss": 1.1031, + "step": 6548 + }, + { + "epoch": 0.48, + "learning_rate": 1.986572986778199e-05, + "loss": 1.0637, + "step": 6549 + }, + { + "epoch": 0.48, + "learning_rate": 1.986566471583093e-05, + "loss": 0.9908, + "step": 6550 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865599548183737e-05, + "loss": 1.0257, + "step": 6551 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865534364840514e-05, + "loss": 0.9786, + "step": 6552 + }, + { + "epoch": 0.48, + "learning_rate": 1.986546916580137e-05, + "loss": 1.0396, + "step": 6553 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865403951066404e-05, + "loss": 1.0511, + "step": 6554 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865338720635725e-05, + "loss": 1.0469, + "step": 6555 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865273474509426e-05, + "loss": 1.0245, + "step": 6556 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865208212687624e-05, + "loss": 0.9208, + "step": 6557 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865142935170417e-05, + "loss": 1.0697, + "step": 6558 + }, + { + "epoch": 0.48, + "learning_rate": 1.9865077641957906e-05, + "loss": 0.9435, + "step": 6559 + }, + { + "epoch": 0.48, + "learning_rate": 1.98650123330502e-05, + "loss": 1.033, + "step": 6560 + }, + { + "epoch": 0.48, + "learning_rate": 1.9864947008447397e-05, + "loss": 1.0584, + "step": 6561 + }, + { + "epoch": 0.48, + "learning_rate": 1.9864881668149606e-05, + "loss": 1.0294, + "step": 6562 + }, + { + "epoch": 0.48, + "learning_rate": 1.986481631215693e-05, + "loss": 1.0876, + "step": 6563 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864750940469473e-05, + "loss": 1.0327, + "step": 6564 + }, + { + "epoch": 0.49, + "learning_rate": 1.986468555308734e-05, + "loss": 1.0267, + "step": 6565 + }, + { + "epoch": 0.49, + "learning_rate": 1.986462015001063e-05, + "loss": 1.0125, + "step": 6566 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864554731239454e-05, + "loss": 0.9559, + "step": 6567 + }, + { + "epoch": 0.49, + "learning_rate": 1.986448929677391e-05, + "loss": 1.0639, + "step": 6568 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864423846614105e-05, + "loss": 1.1007, + "step": 6569 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864358380760142e-05, + "loss": 1.0291, + "step": 6570 + }, + { + "epoch": 0.49, + "learning_rate": 1.986429289921213e-05, + "loss": 1.093, + "step": 6571 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864227401970164e-05, + "loss": 0.9504, + "step": 6572 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864161889034355e-05, + "loss": 0.9921, + "step": 6573 + }, + { + "epoch": 0.49, + "learning_rate": 1.9864096360404805e-05, + "loss": 1.0706, + "step": 6574 + }, + { + "epoch": 0.49, + "learning_rate": 1.986403081608162e-05, + "loss": 1.1165, + "step": 6575 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863965256064903e-05, + "loss": 1.1522, + "step": 6576 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863899680354758e-05, + "loss": 0.9904, + "step": 6577 + }, + { + "epoch": 0.49, + "learning_rate": 1.986383408895129e-05, + "loss": 1.1101, + "step": 6578 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863768481854604e-05, + "loss": 0.895, + "step": 6579 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863702859064803e-05, + "loss": 1.0547, + "step": 6580 + }, + { + "epoch": 0.49, + "learning_rate": 1.986363722058199e-05, + "loss": 1.1028, + "step": 6581 + }, + { + "epoch": 0.49, + "learning_rate": 1.986357156640627e-05, + "loss": 1.0785, + "step": 6582 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863505896537755e-05, + "loss": 0.9836, + "step": 6583 + }, + { + "epoch": 0.49, + "learning_rate": 1.986344021097654e-05, + "loss": 1.0578, + "step": 6584 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863374509722733e-05, + "loss": 1.0347, + "step": 6585 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863308792776437e-05, + "loss": 0.9823, + "step": 6586 + }, + { + "epoch": 0.49, + "learning_rate": 1.986324306013776e-05, + "loss": 1.0512, + "step": 6587 + }, + { + "epoch": 0.49, + "learning_rate": 1.9863177311806804e-05, + "loss": 1.0118, + "step": 6588 + }, + { + "epoch": 0.49, + "learning_rate": 1.986311154778367e-05, + "loss": 0.9778, + "step": 6589 + }, + { + "epoch": 0.49, + "learning_rate": 1.986304576806847e-05, + "loss": 0.9981, + "step": 6590 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862979972661308e-05, + "loss": 1.176, + "step": 6591 + }, + { + "epoch": 0.49, + "learning_rate": 1.986291416156228e-05, + "loss": 1.1022, + "step": 6592 + }, + { + "epoch": 0.49, + "learning_rate": 1.98628483347715e-05, + "loss": 1.0842, + "step": 6593 + }, + { + "epoch": 0.49, + "learning_rate": 1.986278249228907e-05, + "loss": 1.1688, + "step": 6594 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862716634115094e-05, + "loss": 1.1341, + "step": 6595 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862650760249674e-05, + "loss": 1.094, + "step": 6596 + }, + { + "epoch": 0.49, + "learning_rate": 1.986258487069292e-05, + "loss": 1.0547, + "step": 6597 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862518965444935e-05, + "loss": 1.0621, + "step": 6598 + }, + { + "epoch": 0.49, + "learning_rate": 1.986245304450582e-05, + "loss": 0.9903, + "step": 6599 + }, + { + "epoch": 0.49, + "learning_rate": 1.986238710787569e-05, + "loss": 0.9668, + "step": 6600 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862321155554638e-05, + "loss": 1.0117, + "step": 6601 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862255187542776e-05, + "loss": 1.1619, + "step": 6602 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862189203840208e-05, + "loss": 1.0945, + "step": 6603 + }, + { + "epoch": 0.49, + "learning_rate": 1.9862123204447035e-05, + "loss": 1.0922, + "step": 6604 + }, + { + "epoch": 0.49, + "learning_rate": 1.986205718936337e-05, + "loss": 1.0667, + "step": 6605 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861991158589307e-05, + "loss": 1.0694, + "step": 6606 + }, + { + "epoch": 0.49, + "learning_rate": 1.986192511212496e-05, + "loss": 1.0374, + "step": 6607 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861859049970432e-05, + "loss": 1.1278, + "step": 6608 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861792972125827e-05, + "loss": 1.1001, + "step": 6609 + }, + { + "epoch": 0.49, + "learning_rate": 1.986172687859125e-05, + "loss": 1.0522, + "step": 6610 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861660769366806e-05, + "loss": 0.9232, + "step": 6611 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861594644452604e-05, + "loss": 0.9768, + "step": 6612 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861528503848742e-05, + "loss": 1.1329, + "step": 6613 + }, + { + "epoch": 0.49, + "learning_rate": 1.986146234755533e-05, + "loss": 1.0906, + "step": 6614 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861396175572473e-05, + "loss": 0.972, + "step": 6615 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861329987900278e-05, + "loss": 1.0685, + "step": 6616 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861263784538847e-05, + "loss": 0.9912, + "step": 6617 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861197565488287e-05, + "loss": 1.0525, + "step": 6618 + }, + { + "epoch": 0.49, + "learning_rate": 1.98611313307487e-05, + "loss": 1.1563, + "step": 6619 + }, + { + "epoch": 0.49, + "learning_rate": 1.9861065080320197e-05, + "loss": 0.9875, + "step": 6620 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860998814202878e-05, + "loss": 0.9684, + "step": 6621 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860932532396852e-05, + "loss": 0.95, + "step": 6622 + }, + { + "epoch": 0.49, + "learning_rate": 1.986086623490222e-05, + "loss": 1.1002, + "step": 6623 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860799921719097e-05, + "loss": 0.9873, + "step": 6624 + }, + { + "epoch": 0.49, + "learning_rate": 1.986073359284758e-05, + "loss": 0.9853, + "step": 6625 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860667248287777e-05, + "loss": 1.1367, + "step": 6626 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860600888039795e-05, + "loss": 1.0683, + "step": 6627 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860534512103737e-05, + "loss": 1.0318, + "step": 6628 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860468120479706e-05, + "loss": 1.0414, + "step": 6629 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860401713167816e-05, + "loss": 1.0703, + "step": 6630 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860335290168168e-05, + "loss": 1.0281, + "step": 6631 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860268851480862e-05, + "loss": 0.9408, + "step": 6632 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860202397106016e-05, + "loss": 1.0096, + "step": 6633 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860135927043726e-05, + "loss": 1.064, + "step": 6634 + }, + { + "epoch": 0.49, + "learning_rate": 1.98600694412941e-05, + "loss": 0.9777, + "step": 6635 + }, + { + "epoch": 0.49, + "learning_rate": 1.9860002939857247e-05, + "loss": 0.9247, + "step": 6636 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859936422733268e-05, + "loss": 1.0783, + "step": 6637 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859869889922275e-05, + "loss": 1.0615, + "step": 6638 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859803341424365e-05, + "loss": 1.0575, + "step": 6639 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859736777239652e-05, + "loss": 0.9939, + "step": 6640 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859670197368238e-05, + "loss": 1.1791, + "step": 6641 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859603601810228e-05, + "loss": 1.0465, + "step": 6642 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859536990565737e-05, + "loss": 1.0518, + "step": 6643 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859470363634856e-05, + "loss": 0.9188, + "step": 6644 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859403721017704e-05, + "loss": 1.0067, + "step": 6645 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859337062714376e-05, + "loss": 1.0162, + "step": 6646 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859270388724988e-05, + "loss": 1.1604, + "step": 6647 + }, + { + "epoch": 0.49, + "learning_rate": 1.985920369904964e-05, + "loss": 1.0295, + "step": 6648 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859136993688438e-05, + "loss": 0.9557, + "step": 6649 + }, + { + "epoch": 0.49, + "learning_rate": 1.9859070272641494e-05, + "loss": 0.9913, + "step": 6650 + }, + { + "epoch": 0.49, + "learning_rate": 1.985900353590891e-05, + "loss": 0.9226, + "step": 6651 + }, + { + "epoch": 0.49, + "learning_rate": 1.985893678349079e-05, + "loss": 1.0082, + "step": 6652 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858870015387243e-05, + "loss": 1.0283, + "step": 6653 + }, + { + "epoch": 0.49, + "learning_rate": 1.985880323159838e-05, + "loss": 1.1799, + "step": 6654 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858736432124295e-05, + "loss": 0.9931, + "step": 6655 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858669616965106e-05, + "loss": 0.9261, + "step": 6656 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858602786120913e-05, + "loss": 0.9649, + "step": 6657 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858535939591824e-05, + "loss": 0.991, + "step": 6658 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858469077377945e-05, + "loss": 0.9981, + "step": 6659 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858402199479385e-05, + "loss": 1.1079, + "step": 6660 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858335305896244e-05, + "loss": 0.9884, + "step": 6661 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858268396628638e-05, + "loss": 1.0599, + "step": 6662 + }, + { + "epoch": 0.49, + "learning_rate": 1.985820147167666e-05, + "loss": 0.9784, + "step": 6663 + }, + { + "epoch": 0.49, + "learning_rate": 1.985813453104043e-05, + "loss": 0.9695, + "step": 6664 + }, + { + "epoch": 0.49, + "learning_rate": 1.985806757472005e-05, + "loss": 0.9607, + "step": 6665 + }, + { + "epoch": 0.49, + "learning_rate": 1.9858000602715623e-05, + "loss": 1.0688, + "step": 6666 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857933615027264e-05, + "loss": 0.9959, + "step": 6667 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857866611655067e-05, + "loss": 0.9043, + "step": 6668 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857799592599146e-05, + "loss": 1.0452, + "step": 6669 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857732557859608e-05, + "loss": 1.063, + "step": 6670 + }, + { + "epoch": 0.49, + "learning_rate": 1.985766550743656e-05, + "loss": 0.9789, + "step": 6671 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857598441330107e-05, + "loss": 1.1175, + "step": 6672 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857531359540352e-05, + "loss": 1.0491, + "step": 6673 + }, + { + "epoch": 0.49, + "learning_rate": 1.985746426206741e-05, + "loss": 0.9987, + "step": 6674 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857397148911383e-05, + "loss": 1.0156, + "step": 6675 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857330020072373e-05, + "loss": 1.1044, + "step": 6676 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857262875550497e-05, + "loss": 1.0089, + "step": 6677 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857195715345856e-05, + "loss": 1.0269, + "step": 6678 + }, + { + "epoch": 0.49, + "learning_rate": 1.9857128539458557e-05, + "loss": 0.9868, + "step": 6679 + }, + { + "epoch": 0.49, + "learning_rate": 1.985706134788871e-05, + "loss": 0.9384, + "step": 6680 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856994140636414e-05, + "loss": 0.9237, + "step": 6681 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856926917701784e-05, + "loss": 1.0824, + "step": 6682 + }, + { + "epoch": 0.49, + "learning_rate": 1.985685967908493e-05, + "loss": 0.9776, + "step": 6683 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856792424785943e-05, + "loss": 1.0802, + "step": 6684 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856725154804945e-05, + "loss": 0.9643, + "step": 6685 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856657869142044e-05, + "loss": 0.9874, + "step": 6686 + }, + { + "epoch": 0.49, + "learning_rate": 1.985659056779733e-05, + "loss": 1.064, + "step": 6687 + }, + { + "epoch": 0.49, + "learning_rate": 1.985652325077093e-05, + "loss": 1.0866, + "step": 6688 + }, + { + "epoch": 0.49, + "learning_rate": 1.985645591806294e-05, + "loss": 1.0452, + "step": 6689 + }, + { + "epoch": 0.49, + "learning_rate": 1.985638856967347e-05, + "loss": 1.0321, + "step": 6690 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856321205602627e-05, + "loss": 1.0438, + "step": 6691 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856253825850516e-05, + "loss": 1.1301, + "step": 6692 + }, + { + "epoch": 0.49, + "learning_rate": 1.985618643041725e-05, + "loss": 1.0312, + "step": 6693 + }, + { + "epoch": 0.49, + "learning_rate": 1.985611901930293e-05, + "loss": 1.0779, + "step": 6694 + }, + { + "epoch": 0.49, + "learning_rate": 1.9856051592507665e-05, + "loss": 1.0234, + "step": 6695 + }, + { + "epoch": 0.49, + "learning_rate": 1.9855984150031565e-05, + "loss": 1.0522, + "step": 6696 + }, + { + "epoch": 0.49, + "learning_rate": 1.9855916691874734e-05, + "loss": 1.093, + "step": 6697 + }, + { + "epoch": 0.49, + "learning_rate": 1.985584921803728e-05, + "loss": 1.0685, + "step": 6698 + }, + { + "epoch": 0.49, + "learning_rate": 1.985578172851931e-05, + "loss": 1.0994, + "step": 6699 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855714223320933e-05, + "loss": 1.0856, + "step": 6700 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855646702442257e-05, + "loss": 0.9627, + "step": 6701 + }, + { + "epoch": 0.5, + "learning_rate": 1.985557916588339e-05, + "loss": 1.0332, + "step": 6702 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855511613644433e-05, + "loss": 1.0426, + "step": 6703 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855444045725498e-05, + "loss": 0.9902, + "step": 6704 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855376462126696e-05, + "loss": 1.1206, + "step": 6705 + }, + { + "epoch": 0.5, + "learning_rate": 1.985530886284813e-05, + "loss": 1.0163, + "step": 6706 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855241247889906e-05, + "loss": 1.1414, + "step": 6707 + }, + { + "epoch": 0.5, + "learning_rate": 1.985517361725214e-05, + "loss": 1.0393, + "step": 6708 + }, + { + "epoch": 0.5, + "learning_rate": 1.985510597093493e-05, + "loss": 0.9939, + "step": 6709 + }, + { + "epoch": 0.5, + "learning_rate": 1.9855038308938387e-05, + "loss": 1.1015, + "step": 6710 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854970631262617e-05, + "loss": 1.0051, + "step": 6711 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854902937907735e-05, + "loss": 0.9927, + "step": 6712 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854835228873843e-05, + "loss": 1.0212, + "step": 6713 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854767504161042e-05, + "loss": 1.0384, + "step": 6714 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854699763769455e-05, + "loss": 0.9961, + "step": 6715 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854632007699176e-05, + "loss": 0.9924, + "step": 6716 + }, + { + "epoch": 0.5, + "learning_rate": 1.985456423595032e-05, + "loss": 1.086, + "step": 6717 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854496448522997e-05, + "loss": 1.0848, + "step": 6718 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854428645417307e-05, + "loss": 1.1073, + "step": 6719 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854360826633366e-05, + "loss": 1.0766, + "step": 6720 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854292992171276e-05, + "loss": 1.0303, + "step": 6721 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854225142031144e-05, + "loss": 1.0081, + "step": 6722 + }, + { + "epoch": 0.5, + "learning_rate": 1.9854157276213083e-05, + "loss": 1.0082, + "step": 6723 + }, + { + "epoch": 0.5, + "learning_rate": 1.98540893947172e-05, + "loss": 1.0197, + "step": 6724 + }, + { + "epoch": 0.5, + "learning_rate": 1.98540214975436e-05, + "loss": 0.9617, + "step": 6725 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853953584692395e-05, + "loss": 0.9526, + "step": 6726 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853885656163687e-05, + "loss": 1.0683, + "step": 6727 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853817711957592e-05, + "loss": 1.0365, + "step": 6728 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853749752074208e-05, + "loss": 1.0036, + "step": 6729 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853681776513657e-05, + "loss": 0.9969, + "step": 6730 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853613785276034e-05, + "loss": 1.0751, + "step": 6731 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853545778361454e-05, + "loss": 1.0174, + "step": 6732 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853477755770022e-05, + "loss": 1.0298, + "step": 6733 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853409717501853e-05, + "loss": 1.0621, + "step": 6734 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853341663557043e-05, + "loss": 1.1111, + "step": 6735 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853273593935713e-05, + "loss": 0.9852, + "step": 6736 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853205508637964e-05, + "loss": 1.0467, + "step": 6737 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853137407663906e-05, + "loss": 1.0462, + "step": 6738 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853069291013646e-05, + "loss": 1.0006, + "step": 6739 + }, + { + "epoch": 0.5, + "learning_rate": 1.9853001158687297e-05, + "loss": 1.0703, + "step": 6740 + }, + { + "epoch": 0.5, + "learning_rate": 1.985293301068496e-05, + "loss": 1.0466, + "step": 6741 + }, + { + "epoch": 0.5, + "learning_rate": 1.985286484700675e-05, + "loss": 0.9735, + "step": 6742 + }, + { + "epoch": 0.5, + "learning_rate": 1.985279666765277e-05, + "loss": 1.0846, + "step": 6743 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852728472623135e-05, + "loss": 1.0758, + "step": 6744 + }, + { + "epoch": 0.5, + "learning_rate": 1.985266026191795e-05, + "loss": 0.9382, + "step": 6745 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852592035537323e-05, + "loss": 1.0734, + "step": 6746 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852523793481362e-05, + "loss": 0.9523, + "step": 6747 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852455535750176e-05, + "loss": 1.001, + "step": 6748 + }, + { + "epoch": 0.5, + "learning_rate": 1.985238726234388e-05, + "loss": 1.1242, + "step": 6749 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852318973262566e-05, + "loss": 1.0712, + "step": 6750 + }, + { + "epoch": 0.5, + "learning_rate": 1.985225066850636e-05, + "loss": 1.0415, + "step": 6751 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852182348075362e-05, + "loss": 1.0249, + "step": 6752 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852114011969688e-05, + "loss": 1.074, + "step": 6753 + }, + { + "epoch": 0.5, + "learning_rate": 1.9852045660189438e-05, + "loss": 1.0308, + "step": 6754 + }, + { + "epoch": 0.5, + "learning_rate": 1.985197729273472e-05, + "loss": 1.1422, + "step": 6755 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851908909605657e-05, + "loss": 1.0536, + "step": 6756 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851840510802343e-05, + "loss": 0.911, + "step": 6757 + }, + { + "epoch": 0.5, + "learning_rate": 1.985177209632489e-05, + "loss": 0.9604, + "step": 6758 + }, + { + "epoch": 0.5, + "learning_rate": 1.985170366617341e-05, + "loss": 1.0381, + "step": 6759 + }, + { + "epoch": 0.5, + "learning_rate": 1.985163522034801e-05, + "loss": 1.0046, + "step": 6760 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851566758848796e-05, + "loss": 1.0093, + "step": 6761 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851498281675886e-05, + "loss": 1.1168, + "step": 6762 + }, + { + "epoch": 0.5, + "learning_rate": 1.985142978882938e-05, + "loss": 0.9743, + "step": 6763 + }, + { + "epoch": 0.5, + "learning_rate": 1.985136128030939e-05, + "loss": 1.0318, + "step": 6764 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851292756116028e-05, + "loss": 1.0592, + "step": 6765 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851224216249398e-05, + "loss": 0.944, + "step": 6766 + }, + { + "epoch": 0.5, + "learning_rate": 1.985115566070961e-05, + "loss": 0.9227, + "step": 6767 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851087089496778e-05, + "loss": 0.9901, + "step": 6768 + }, + { + "epoch": 0.5, + "learning_rate": 1.9851018502611007e-05, + "loss": 1.063, + "step": 6769 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850949900052403e-05, + "loss": 1.0682, + "step": 6770 + }, + { + "epoch": 0.5, + "learning_rate": 1.985088128182108e-05, + "loss": 0.9665, + "step": 6771 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850812647917147e-05, + "loss": 1.035, + "step": 6772 + }, + { + "epoch": 0.5, + "learning_rate": 1.985074399834071e-05, + "loss": 1.0802, + "step": 6773 + }, + { + "epoch": 0.5, + "learning_rate": 1.985067533309188e-05, + "loss": 0.934, + "step": 6774 + }, + { + "epoch": 0.5, + "learning_rate": 1.985060665217077e-05, + "loss": 1.1153, + "step": 6775 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850537955577482e-05, + "loss": 1.0915, + "step": 6776 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850469243312133e-05, + "loss": 1.0768, + "step": 6777 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850400515374827e-05, + "loss": 1.0241, + "step": 6778 + }, + { + "epoch": 0.5, + "learning_rate": 1.985033177176567e-05, + "loss": 1.1082, + "step": 6779 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850263012484782e-05, + "loss": 1.0424, + "step": 6780 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850194237532262e-05, + "loss": 1.176, + "step": 6781 + }, + { + "epoch": 0.5, + "learning_rate": 1.9850125446908227e-05, + "loss": 1.0643, + "step": 6782 + }, + { + "epoch": 0.5, + "learning_rate": 1.985005664061278e-05, + "loss": 0.9874, + "step": 6783 + }, + { + "epoch": 0.5, + "learning_rate": 1.984998781864604e-05, + "loss": 1.0614, + "step": 6784 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849918981008108e-05, + "loss": 1.0684, + "step": 6785 + }, + { + "epoch": 0.5, + "learning_rate": 1.984985012769909e-05, + "loss": 0.9652, + "step": 6786 + }, + { + "epoch": 0.5, + "learning_rate": 1.984978125871911e-05, + "loss": 1.0989, + "step": 6787 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849712374068262e-05, + "loss": 1.0945, + "step": 6788 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849643473746667e-05, + "loss": 1.131, + "step": 6789 + }, + { + "epoch": 0.5, + "learning_rate": 1.984957455775443e-05, + "loss": 1.0067, + "step": 6790 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849505626091658e-05, + "loss": 1.076, + "step": 6791 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849436678758467e-05, + "loss": 1.0487, + "step": 6792 + }, + { + "epoch": 0.5, + "learning_rate": 1.984936771575496e-05, + "loss": 1.0998, + "step": 6793 + }, + { + "epoch": 0.5, + "learning_rate": 1.984929873708125e-05, + "loss": 0.9718, + "step": 6794 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849229742737447e-05, + "loss": 0.9654, + "step": 6795 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849160732723663e-05, + "loss": 1.0454, + "step": 6796 + }, + { + "epoch": 0.5, + "learning_rate": 1.9849091707040002e-05, + "loss": 1.111, + "step": 6797 + }, + { + "epoch": 0.5, + "learning_rate": 1.984902266568658e-05, + "loss": 1.0992, + "step": 6798 + }, + { + "epoch": 0.5, + "learning_rate": 1.98489536086635e-05, + "loss": 1.0338, + "step": 6799 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848884535970877e-05, + "loss": 0.9773, + "step": 6800 + }, + { + "epoch": 0.5, + "learning_rate": 1.984881544760882e-05, + "loss": 1.0023, + "step": 6801 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848746343577442e-05, + "loss": 0.9775, + "step": 6802 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848677223876845e-05, + "loss": 1.069, + "step": 6803 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848608088507144e-05, + "loss": 1.1329, + "step": 6804 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848538937468448e-05, + "loss": 1.0, + "step": 6805 + }, + { + "epoch": 0.5, + "learning_rate": 1.984846977076087e-05, + "loss": 0.9971, + "step": 6806 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848400588384516e-05, + "loss": 1.0369, + "step": 6807 + }, + { + "epoch": 0.5, + "learning_rate": 1.98483313903395e-05, + "loss": 1.0779, + "step": 6808 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848262176625926e-05, + "loss": 1.0206, + "step": 6809 + }, + { + "epoch": 0.5, + "learning_rate": 1.984819294724391e-05, + "loss": 1.0, + "step": 6810 + }, + { + "epoch": 0.5, + "learning_rate": 1.984812370219356e-05, + "loss": 1.1403, + "step": 6811 + }, + { + "epoch": 0.5, + "learning_rate": 1.9848054441474983e-05, + "loss": 0.9776, + "step": 6812 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847985165088296e-05, + "loss": 1.123, + "step": 6813 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847915873033604e-05, + "loss": 1.0788, + "step": 6814 + }, + { + "epoch": 0.5, + "learning_rate": 1.984784656531102e-05, + "loss": 1.1417, + "step": 6815 + }, + { + "epoch": 0.5, + "learning_rate": 1.984777724192065e-05, + "loss": 1.134, + "step": 6816 + }, + { + "epoch": 0.5, + "learning_rate": 1.984770790286261e-05, + "loss": 1.0475, + "step": 6817 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847638548137006e-05, + "loss": 1.064, + "step": 6818 + }, + { + "epoch": 0.5, + "learning_rate": 1.984756917774395e-05, + "loss": 1.0298, + "step": 6819 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847499791683552e-05, + "loss": 1.1602, + "step": 6820 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847430389955925e-05, + "loss": 0.9424, + "step": 6821 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847360972561176e-05, + "loss": 0.9872, + "step": 6822 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847291539499415e-05, + "loss": 1.0459, + "step": 6823 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847222090770757e-05, + "loss": 0.9911, + "step": 6824 + }, + { + "epoch": 0.5, + "learning_rate": 1.984715262637531e-05, + "loss": 0.9818, + "step": 6825 + }, + { + "epoch": 0.5, + "learning_rate": 1.984708314631318e-05, + "loss": 1.0188, + "step": 6826 + }, + { + "epoch": 0.5, + "learning_rate": 1.9847013650584482e-05, + "loss": 1.0038, + "step": 6827 + }, + { + "epoch": 0.5, + "learning_rate": 1.984694413918933e-05, + "loss": 1.0506, + "step": 6828 + }, + { + "epoch": 0.5, + "learning_rate": 1.9846874612127828e-05, + "loss": 1.0848, + "step": 6829 + }, + { + "epoch": 0.5, + "learning_rate": 1.984680506940009e-05, + "loss": 1.0138, + "step": 6830 + }, + { + "epoch": 0.5, + "learning_rate": 1.9846735511006226e-05, + "loss": 1.0576, + "step": 6831 + }, + { + "epoch": 0.5, + "learning_rate": 1.9846665936946347e-05, + "loss": 1.0626, + "step": 6832 + }, + { + "epoch": 0.5, + "learning_rate": 1.9846596347220563e-05, + "loss": 1.0156, + "step": 6833 + }, + { + "epoch": 0.5, + "learning_rate": 1.9846526741828987e-05, + "loss": 1.0593, + "step": 6834 + }, + { + "epoch": 0.51, + "learning_rate": 1.9846457120771726e-05, + "loss": 1.0574, + "step": 6835 + }, + { + "epoch": 0.51, + "learning_rate": 1.9846387484048892e-05, + "loss": 0.9249, + "step": 6836 + }, + { + "epoch": 0.51, + "learning_rate": 1.9846317831660597e-05, + "loss": 1.0152, + "step": 6837 + }, + { + "epoch": 0.51, + "learning_rate": 1.9846248163606953e-05, + "loss": 1.0191, + "step": 6838 + }, + { + "epoch": 0.51, + "learning_rate": 1.9846178479888064e-05, + "loss": 1.035, + "step": 6839 + }, + { + "epoch": 0.51, + "learning_rate": 1.984610878050405e-05, + "loss": 1.0367, + "step": 6840 + }, + { + "epoch": 0.51, + "learning_rate": 1.984603906545502e-05, + "loss": 1.0222, + "step": 6841 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845969334741076e-05, + "loss": 1.0179, + "step": 6842 + }, + { + "epoch": 0.51, + "learning_rate": 1.984589958836234e-05, + "loss": 1.1114, + "step": 6843 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845829826318917e-05, + "loss": 1.1192, + "step": 6844 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845760048610922e-05, + "loss": 0.969, + "step": 6845 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845690255238462e-05, + "loss": 0.9702, + "step": 6846 + }, + { + "epoch": 0.51, + "learning_rate": 1.984562044620165e-05, + "loss": 1.0558, + "step": 6847 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845550621500598e-05, + "loss": 1.104, + "step": 6848 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845480781135414e-05, + "loss": 1.0686, + "step": 6849 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845410925106212e-05, + "loss": 1.1217, + "step": 6850 + }, + { + "epoch": 0.51, + "learning_rate": 1.98453410534131e-05, + "loss": 0.9766, + "step": 6851 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845271166056193e-05, + "loss": 1.1043, + "step": 6852 + }, + { + "epoch": 0.51, + "learning_rate": 1.98452012630356e-05, + "loss": 1.1075, + "step": 6853 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845131344351435e-05, + "loss": 0.9997, + "step": 6854 + }, + { + "epoch": 0.51, + "learning_rate": 1.9845061410003805e-05, + "loss": 1.1086, + "step": 6855 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844991459992822e-05, + "loss": 1.0668, + "step": 6856 + }, + { + "epoch": 0.51, + "learning_rate": 1.98449214943186e-05, + "loss": 0.9899, + "step": 6857 + }, + { + "epoch": 0.51, + "learning_rate": 1.984485151298125e-05, + "loss": 1.0075, + "step": 6858 + }, + { + "epoch": 0.51, + "learning_rate": 1.984478151598088e-05, + "loss": 1.0457, + "step": 6859 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844711503317602e-05, + "loss": 1.151, + "step": 6860 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844641474991532e-05, + "loss": 1.0884, + "step": 6861 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844571431002775e-05, + "loss": 1.0473, + "step": 6862 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844501371351447e-05, + "loss": 1.0792, + "step": 6863 + }, + { + "epoch": 0.51, + "learning_rate": 1.984443129603766e-05, + "loss": 0.9495, + "step": 6864 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844361205061522e-05, + "loss": 1.1001, + "step": 6865 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844291098423146e-05, + "loss": 1.0623, + "step": 6866 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844220976122646e-05, + "loss": 1.0717, + "step": 6867 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844150838160127e-05, + "loss": 1.0242, + "step": 6868 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844080684535708e-05, + "loss": 1.0592, + "step": 6869 + }, + { + "epoch": 0.51, + "learning_rate": 1.9844010515249497e-05, + "loss": 1.024, + "step": 6870 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843940330301607e-05, + "loss": 1.0963, + "step": 6871 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843870129692148e-05, + "loss": 0.931, + "step": 6872 + }, + { + "epoch": 0.51, + "learning_rate": 1.984379991342123e-05, + "loss": 1.0656, + "step": 6873 + }, + { + "epoch": 0.51, + "learning_rate": 1.984372968148897e-05, + "loss": 1.0122, + "step": 6874 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843659433895476e-05, + "loss": 1.0915, + "step": 6875 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843589170640863e-05, + "loss": 0.9808, + "step": 6876 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843518891725238e-05, + "loss": 0.9847, + "step": 6877 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843448597148716e-05, + "loss": 0.9959, + "step": 6878 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843378286911408e-05, + "loss": 1.1196, + "step": 6879 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843307961013424e-05, + "loss": 1.0275, + "step": 6880 + }, + { + "epoch": 0.51, + "learning_rate": 1.984323761945488e-05, + "loss": 1.0176, + "step": 6881 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843167262235888e-05, + "loss": 1.068, + "step": 6882 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843096889356553e-05, + "loss": 1.0381, + "step": 6883 + }, + { + "epoch": 0.51, + "learning_rate": 1.9843026500816992e-05, + "loss": 1.0699, + "step": 6884 + }, + { + "epoch": 0.51, + "learning_rate": 1.984295609661732e-05, + "loss": 1.0826, + "step": 6885 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842885676757646e-05, + "loss": 1.0272, + "step": 6886 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842815241238077e-05, + "loss": 1.0639, + "step": 6887 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842744790058736e-05, + "loss": 1.0433, + "step": 6888 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842674323219724e-05, + "loss": 0.9851, + "step": 6889 + }, + { + "epoch": 0.51, + "learning_rate": 1.984260384072116e-05, + "loss": 1.0212, + "step": 6890 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842533342563155e-05, + "loss": 0.9963, + "step": 6891 + }, + { + "epoch": 0.51, + "learning_rate": 1.984246282874582e-05, + "loss": 1.0837, + "step": 6892 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842392299269263e-05, + "loss": 1.011, + "step": 6893 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842321754133605e-05, + "loss": 1.0015, + "step": 6894 + }, + { + "epoch": 0.51, + "learning_rate": 1.9842251193338954e-05, + "loss": 1.0602, + "step": 6895 + }, + { + "epoch": 0.51, + "learning_rate": 1.984218061688542e-05, + "loss": 0.9874, + "step": 6896 + }, + { + "epoch": 0.51, + "learning_rate": 1.984211002477312e-05, + "loss": 1.0525, + "step": 6897 + }, + { + "epoch": 0.51, + "learning_rate": 1.984203941700216e-05, + "loss": 0.9334, + "step": 6898 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841968793572663e-05, + "loss": 1.1382, + "step": 6899 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841898154484727e-05, + "loss": 1.0268, + "step": 6900 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841827499738476e-05, + "loss": 0.953, + "step": 6901 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841756829334014e-05, + "loss": 0.9694, + "step": 6902 + }, + { + "epoch": 0.51, + "learning_rate": 1.984168614327146e-05, + "loss": 1.036, + "step": 6903 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841615441550927e-05, + "loss": 0.9938, + "step": 6904 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841544724172523e-05, + "loss": 1.0585, + "step": 6905 + }, + { + "epoch": 0.51, + "learning_rate": 1.984147399113636e-05, + "loss": 1.0338, + "step": 6906 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841403242442555e-05, + "loss": 1.0405, + "step": 6907 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841332478091217e-05, + "loss": 1.1219, + "step": 6908 + }, + { + "epoch": 0.51, + "learning_rate": 1.984126169808246e-05, + "loss": 0.9902, + "step": 6909 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841190902416398e-05, + "loss": 1.0831, + "step": 6910 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841120091093142e-05, + "loss": 0.9674, + "step": 6911 + }, + { + "epoch": 0.51, + "learning_rate": 1.9841049264112804e-05, + "loss": 0.9997, + "step": 6912 + }, + { + "epoch": 0.51, + "learning_rate": 1.98409784214755e-05, + "loss": 1.0736, + "step": 6913 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840907563181335e-05, + "loss": 1.0423, + "step": 6914 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840836689230428e-05, + "loss": 1.0734, + "step": 6915 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840765799622893e-05, + "loss": 1.0705, + "step": 6916 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840694894358838e-05, + "loss": 1.0834, + "step": 6917 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840623973438382e-05, + "loss": 1.0518, + "step": 6918 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840553036861633e-05, + "loss": 0.9848, + "step": 6919 + }, + { + "epoch": 0.51, + "learning_rate": 1.98404820846287e-05, + "loss": 1.1034, + "step": 6920 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840411116739708e-05, + "loss": 1.0647, + "step": 6921 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840340133194755e-05, + "loss": 1.0275, + "step": 6922 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840269133993967e-05, + "loss": 1.0566, + "step": 6923 + }, + { + "epoch": 0.51, + "learning_rate": 1.984019811913745e-05, + "loss": 1.0391, + "step": 6924 + }, + { + "epoch": 0.51, + "learning_rate": 1.984012708862532e-05, + "loss": 1.0559, + "step": 6925 + }, + { + "epoch": 0.51, + "learning_rate": 1.9840056042457685e-05, + "loss": 1.0073, + "step": 6926 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839984980634664e-05, + "loss": 1.2208, + "step": 6927 + }, + { + "epoch": 0.51, + "learning_rate": 1.983991390315637e-05, + "loss": 1.0398, + "step": 6928 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839842810022907e-05, + "loss": 1.0843, + "step": 6929 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839771701234403e-05, + "loss": 1.0423, + "step": 6930 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839700576790957e-05, + "loss": 1.0012, + "step": 6931 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839629436692693e-05, + "loss": 1.0533, + "step": 6932 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839558280939718e-05, + "loss": 0.9457, + "step": 6933 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839487109532144e-05, + "loss": 1.033, + "step": 6934 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839415922470087e-05, + "loss": 1.0851, + "step": 6935 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839344719753662e-05, + "loss": 0.9849, + "step": 6936 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839273501382977e-05, + "loss": 1.1106, + "step": 6937 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839202267358154e-05, + "loss": 0.9933, + "step": 6938 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839131017679295e-05, + "loss": 1.0848, + "step": 6939 + }, + { + "epoch": 0.51, + "learning_rate": 1.9839059752346525e-05, + "loss": 1.0453, + "step": 6940 + }, + { + "epoch": 0.51, + "learning_rate": 1.983898847135995e-05, + "loss": 1.0687, + "step": 6941 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838917174719682e-05, + "loss": 0.9924, + "step": 6942 + }, + { + "epoch": 0.51, + "learning_rate": 1.983884586242584e-05, + "loss": 1.0103, + "step": 6943 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838774534478535e-05, + "loss": 1.0293, + "step": 6944 + }, + { + "epoch": 0.51, + "learning_rate": 1.983870319087788e-05, + "loss": 0.9429, + "step": 6945 + }, + { + "epoch": 0.51, + "learning_rate": 1.983863183162399e-05, + "loss": 1.0757, + "step": 6946 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838560456716978e-05, + "loss": 1.0898, + "step": 6947 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838489066156956e-05, + "loss": 1.0025, + "step": 6948 + }, + { + "epoch": 0.51, + "learning_rate": 1.983841765994404e-05, + "loss": 0.9982, + "step": 6949 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838346238078338e-05, + "loss": 1.0636, + "step": 6950 + }, + { + "epoch": 0.51, + "learning_rate": 1.983827480055997e-05, + "loss": 0.986, + "step": 6951 + }, + { + "epoch": 0.51, + "learning_rate": 1.983820334738905e-05, + "loss": 0.9428, + "step": 6952 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838131878565687e-05, + "loss": 0.986, + "step": 6953 + }, + { + "epoch": 0.51, + "learning_rate": 1.9838060394089996e-05, + "loss": 1.0021, + "step": 6954 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837988893962095e-05, + "loss": 1.0309, + "step": 6955 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837917378182093e-05, + "loss": 1.0005, + "step": 6956 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837845846750104e-05, + "loss": 0.9489, + "step": 6957 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837774299666244e-05, + "loss": 1.0505, + "step": 6958 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837702736930625e-05, + "loss": 1.0809, + "step": 6959 + }, + { + "epoch": 0.51, + "learning_rate": 1.983763115854336e-05, + "loss": 1.1431, + "step": 6960 + }, + { + "epoch": 0.51, + "learning_rate": 1.983755956450457e-05, + "loss": 1.0177, + "step": 6961 + }, + { + "epoch": 0.51, + "learning_rate": 1.983748795481436e-05, + "loss": 1.0248, + "step": 6962 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837416329472846e-05, + "loss": 1.0835, + "step": 6963 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837344688480148e-05, + "loss": 1.0451, + "step": 6964 + }, + { + "epoch": 0.51, + "learning_rate": 1.983727303183637e-05, + "loss": 1.0451, + "step": 6965 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837201359541637e-05, + "loss": 1.0866, + "step": 6966 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837129671596052e-05, + "loss": 0.9707, + "step": 6967 + }, + { + "epoch": 0.51, + "learning_rate": 1.9837057967999738e-05, + "loss": 1.0087, + "step": 6968 + }, + { + "epoch": 0.51, + "learning_rate": 1.9836986248752806e-05, + "loss": 1.0438, + "step": 6969 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836914513855365e-05, + "loss": 1.0006, + "step": 6970 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836842763307536e-05, + "loss": 1.119, + "step": 6971 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836770997109433e-05, + "loss": 1.0046, + "step": 6972 + }, + { + "epoch": 0.52, + "learning_rate": 1.983669921526117e-05, + "loss": 0.8977, + "step": 6973 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836627417762853e-05, + "loss": 1.0783, + "step": 6974 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836555604614604e-05, + "loss": 1.0341, + "step": 6975 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836483775816537e-05, + "loss": 0.9992, + "step": 6976 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836411931368767e-05, + "loss": 0.978, + "step": 6977 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836340071271403e-05, + "loss": 1.0435, + "step": 6978 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836268195524563e-05, + "loss": 1.0709, + "step": 6979 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836196304128363e-05, + "loss": 1.0763, + "step": 6980 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836124397082917e-05, + "loss": 1.0275, + "step": 6981 + }, + { + "epoch": 0.52, + "learning_rate": 1.9836052474388334e-05, + "loss": 0.9481, + "step": 6982 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835980536044736e-05, + "loss": 0.9852, + "step": 6983 + }, + { + "epoch": 0.52, + "learning_rate": 1.983590858205223e-05, + "loss": 1.054, + "step": 6984 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835836612410936e-05, + "loss": 1.1147, + "step": 6985 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835764627120967e-05, + "loss": 1.0268, + "step": 6986 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835692626182434e-05, + "loss": 1.0007, + "step": 6987 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835620609595458e-05, + "loss": 0.9774, + "step": 6988 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835548577360152e-05, + "loss": 1.0383, + "step": 6989 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835476529476624e-05, + "loss": 1.1441, + "step": 6990 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835404465945e-05, + "loss": 1.0789, + "step": 6991 + }, + { + "epoch": 0.52, + "learning_rate": 1.983533238676538e-05, + "loss": 0.9547, + "step": 6992 + }, + { + "epoch": 0.52, + "learning_rate": 1.983526029193789e-05, + "loss": 1.0875, + "step": 6993 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835188181462646e-05, + "loss": 1.0547, + "step": 6994 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835116055339755e-05, + "loss": 1.0549, + "step": 6995 + }, + { + "epoch": 0.52, + "learning_rate": 1.9835043913569337e-05, + "loss": 1.1072, + "step": 6996 + }, + { + "epoch": 0.52, + "learning_rate": 1.98349717561515e-05, + "loss": 1.0661, + "step": 6997 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834899583086367e-05, + "loss": 1.1207, + "step": 6998 + }, + { + "epoch": 0.52, + "learning_rate": 1.983482739437405e-05, + "loss": 1.0352, + "step": 6999 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834755190014663e-05, + "loss": 0.9575, + "step": 7000 + }, + { + "epoch": 0.52, + "learning_rate": 1.983468297000832e-05, + "loss": 1.0034, + "step": 7001 + }, + { + "epoch": 0.52, + "learning_rate": 1.983461073435514e-05, + "loss": 1.1069, + "step": 7002 + }, + { + "epoch": 0.52, + "learning_rate": 1.983453848305523e-05, + "loss": 1.0418, + "step": 7003 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834466216108715e-05, + "loss": 1.0952, + "step": 7004 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834393933515704e-05, + "loss": 1.0519, + "step": 7005 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834321635276308e-05, + "loss": 1.0173, + "step": 7006 + }, + { + "epoch": 0.52, + "learning_rate": 1.983424932139065e-05, + "loss": 1.0009, + "step": 7007 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834176991858846e-05, + "loss": 1.0366, + "step": 7008 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834104646681003e-05, + "loss": 1.0956, + "step": 7009 + }, + { + "epoch": 0.52, + "learning_rate": 1.9834032285857243e-05, + "loss": 1.146, + "step": 7010 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833959909387675e-05, + "loss": 0.9819, + "step": 7011 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833887517272417e-05, + "loss": 1.0553, + "step": 7012 + }, + { + "epoch": 0.52, + "learning_rate": 1.983381510951159e-05, + "loss": 1.0736, + "step": 7013 + }, + { + "epoch": 0.52, + "learning_rate": 1.98337426861053e-05, + "loss": 1.1023, + "step": 7014 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833670247053663e-05, + "loss": 0.8824, + "step": 7015 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833597792356802e-05, + "loss": 1.0021, + "step": 7016 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833525322014826e-05, + "loss": 1.0288, + "step": 7017 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833452836027853e-05, + "loss": 1.0706, + "step": 7018 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833380334395998e-05, + "loss": 1.0287, + "step": 7019 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833307817119373e-05, + "loss": 1.0997, + "step": 7020 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833235284198098e-05, + "loss": 0.9875, + "step": 7021 + }, + { + "epoch": 0.52, + "learning_rate": 1.9833162735632287e-05, + "loss": 1.083, + "step": 7022 + }, + { + "epoch": 0.52, + "learning_rate": 1.983309017142205e-05, + "loss": 1.0992, + "step": 7023 + }, + { + "epoch": 0.52, + "learning_rate": 1.983301759156751e-05, + "loss": 0.8942, + "step": 7024 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832944996068782e-05, + "loss": 1.0616, + "step": 7025 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832872384925976e-05, + "loss": 0.961, + "step": 7026 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832799758139213e-05, + "loss": 0.9479, + "step": 7027 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832727115708605e-05, + "loss": 1.0155, + "step": 7028 + }, + { + "epoch": 0.52, + "learning_rate": 1.983265445763427e-05, + "loss": 0.991, + "step": 7029 + }, + { + "epoch": 0.52, + "learning_rate": 1.983258178391632e-05, + "loss": 0.9888, + "step": 7030 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832509094554876e-05, + "loss": 1.0164, + "step": 7031 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832436389550048e-05, + "loss": 1.1174, + "step": 7032 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832363668901958e-05, + "loss": 0.9701, + "step": 7033 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832290932610715e-05, + "loss": 0.9958, + "step": 7034 + }, + { + "epoch": 0.52, + "learning_rate": 1.983221818067644e-05, + "loss": 1.0451, + "step": 7035 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832145413099245e-05, + "loss": 1.0401, + "step": 7036 + }, + { + "epoch": 0.52, + "learning_rate": 1.9832072629879246e-05, + "loss": 0.9844, + "step": 7037 + }, + { + "epoch": 0.52, + "learning_rate": 1.983199983101656e-05, + "loss": 1.0332, + "step": 7038 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831927016511308e-05, + "loss": 1.0695, + "step": 7039 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831854186363596e-05, + "loss": 1.0829, + "step": 7040 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831781340573547e-05, + "loss": 1.1549, + "step": 7041 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831708479141273e-05, + "loss": 0.9773, + "step": 7042 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831635602066893e-05, + "loss": 1.0267, + "step": 7043 + }, + { + "epoch": 0.52, + "learning_rate": 1.983156270935052e-05, + "loss": 1.0293, + "step": 7044 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831489800992273e-05, + "loss": 1.0269, + "step": 7045 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831416876992264e-05, + "loss": 1.0782, + "step": 7046 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831343937350612e-05, + "loss": 1.0788, + "step": 7047 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831270982067433e-05, + "loss": 1.0812, + "step": 7048 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831198011142846e-05, + "loss": 1.0542, + "step": 7049 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831125024576958e-05, + "loss": 1.0149, + "step": 7050 + }, + { + "epoch": 0.52, + "learning_rate": 1.9831052022369895e-05, + "loss": 1.0364, + "step": 7051 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830979004521766e-05, + "loss": 0.9491, + "step": 7052 + }, + { + "epoch": 0.52, + "learning_rate": 1.983090597103269e-05, + "loss": 1.0658, + "step": 7053 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830832921902787e-05, + "loss": 1.0304, + "step": 7054 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830759857132166e-05, + "loss": 1.0743, + "step": 7055 + }, + { + "epoch": 0.52, + "learning_rate": 1.983068677672095e-05, + "loss": 1.0553, + "step": 7056 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830613680669248e-05, + "loss": 1.0242, + "step": 7057 + }, + { + "epoch": 0.52, + "learning_rate": 1.983054056897718e-05, + "loss": 1.0399, + "step": 7058 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830467441644865e-05, + "loss": 0.9573, + "step": 7059 + }, + { + "epoch": 0.52, + "learning_rate": 1.983039429867242e-05, + "loss": 1.0107, + "step": 7060 + }, + { + "epoch": 0.52, + "learning_rate": 1.983032114005995e-05, + "loss": 1.075, + "step": 7061 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830247965807584e-05, + "loss": 1.0235, + "step": 7062 + }, + { + "epoch": 0.52, + "learning_rate": 1.9830174775915435e-05, + "loss": 1.0698, + "step": 7063 + }, + { + "epoch": 0.52, + "learning_rate": 1.983010157038362e-05, + "loss": 1.0563, + "step": 7064 + }, + { + "epoch": 0.52, + "learning_rate": 1.983002834921225e-05, + "loss": 1.1742, + "step": 7065 + }, + { + "epoch": 0.52, + "learning_rate": 1.982995511240145e-05, + "loss": 1.1, + "step": 7066 + }, + { + "epoch": 0.52, + "learning_rate": 1.982988185995133e-05, + "loss": 1.0745, + "step": 7067 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829808591862008e-05, + "loss": 1.0229, + "step": 7068 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829735308133598e-05, + "loss": 0.9956, + "step": 7069 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829662008766225e-05, + "loss": 1.1396, + "step": 7070 + }, + { + "epoch": 0.52, + "learning_rate": 1.982958869376e-05, + "loss": 1.1734, + "step": 7071 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829515363115037e-05, + "loss": 1.0151, + "step": 7072 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829442016831458e-05, + "loss": 1.0056, + "step": 7073 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829368654909377e-05, + "loss": 0.9876, + "step": 7074 + }, + { + "epoch": 0.52, + "learning_rate": 1.982929527734891e-05, + "loss": 1.0416, + "step": 7075 + }, + { + "epoch": 0.52, + "learning_rate": 1.982922188415018e-05, + "loss": 1.1671, + "step": 7076 + }, + { + "epoch": 0.52, + "learning_rate": 1.982914847531329e-05, + "loss": 0.9928, + "step": 7077 + }, + { + "epoch": 0.52, + "learning_rate": 1.982907505083837e-05, + "loss": 1.0842, + "step": 7078 + }, + { + "epoch": 0.52, + "learning_rate": 1.9829001610725533e-05, + "loss": 0.9596, + "step": 7079 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828928154974892e-05, + "loss": 0.9441, + "step": 7080 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828854683586573e-05, + "loss": 1.0685, + "step": 7081 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828781196560684e-05, + "loss": 1.0777, + "step": 7082 + }, + { + "epoch": 0.52, + "learning_rate": 1.982870769389734e-05, + "loss": 1.0794, + "step": 7083 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828634175596667e-05, + "loss": 1.1307, + "step": 7084 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828560641658777e-05, + "loss": 0.9136, + "step": 7085 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828487092083787e-05, + "loss": 1.1465, + "step": 7086 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828413526871814e-05, + "loss": 1.0245, + "step": 7087 + }, + { + "epoch": 0.52, + "learning_rate": 1.982833994602298e-05, + "loss": 0.9528, + "step": 7088 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828266349537393e-05, + "loss": 1.0181, + "step": 7089 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828192737415177e-05, + "loss": 1.0694, + "step": 7090 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828119109656448e-05, + "loss": 1.1188, + "step": 7091 + }, + { + "epoch": 0.52, + "learning_rate": 1.9828045466261317e-05, + "loss": 1.1224, + "step": 7092 + }, + { + "epoch": 0.52, + "learning_rate": 1.982797180722991e-05, + "loss": 0.9849, + "step": 7093 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827898132562342e-05, + "loss": 1.0425, + "step": 7094 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827824442258727e-05, + "loss": 1.0899, + "step": 7095 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827750736319185e-05, + "loss": 1.0772, + "step": 7096 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827677014743832e-05, + "loss": 1.0573, + "step": 7097 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827603277532782e-05, + "loss": 1.0899, + "step": 7098 + }, + { + "epoch": 0.52, + "learning_rate": 1.982752952468616e-05, + "loss": 1.0357, + "step": 7099 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827455756204076e-05, + "loss": 1.1273, + "step": 7100 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827381972086656e-05, + "loss": 1.0194, + "step": 7101 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827308172334004e-05, + "loss": 1.0741, + "step": 7102 + }, + { + "epoch": 0.52, + "learning_rate": 1.982723435694625e-05, + "loss": 1.0171, + "step": 7103 + }, + { + "epoch": 0.52, + "learning_rate": 1.9827160525923505e-05, + "loss": 1.0022, + "step": 7104 + }, + { + "epoch": 0.52, + "learning_rate": 1.982708667926589e-05, + "loss": 0.9874, + "step": 7105 + }, + { + "epoch": 0.53, + "learning_rate": 1.982701281697352e-05, + "loss": 1.1904, + "step": 7106 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826938939046515e-05, + "loss": 1.1512, + "step": 7107 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826865045484987e-05, + "loss": 0.9433, + "step": 7108 + }, + { + "epoch": 0.53, + "learning_rate": 1.982679113628906e-05, + "loss": 0.9983, + "step": 7109 + }, + { + "epoch": 0.53, + "learning_rate": 1.982671721145885e-05, + "loss": 1.0137, + "step": 7110 + }, + { + "epoch": 0.53, + "learning_rate": 1.982664327099447e-05, + "loss": 0.9132, + "step": 7111 + }, + { + "epoch": 0.53, + "learning_rate": 1.982656931489604e-05, + "loss": 1.1468, + "step": 7112 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826495343163686e-05, + "loss": 1.0121, + "step": 7113 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826421355797512e-05, + "loss": 1.0945, + "step": 7114 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826347352797644e-05, + "loss": 0.9811, + "step": 7115 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826273334164198e-05, + "loss": 0.967, + "step": 7116 + }, + { + "epoch": 0.53, + "learning_rate": 1.982619929989729e-05, + "loss": 0.9688, + "step": 7117 + }, + { + "epoch": 0.53, + "learning_rate": 1.9826125249997043e-05, + "loss": 1.012, + "step": 7118 + }, + { + "epoch": 0.53, + "learning_rate": 1.982605118446357e-05, + "loss": 1.1084, + "step": 7119 + }, + { + "epoch": 0.53, + "learning_rate": 1.982597710329699e-05, + "loss": 0.9674, + "step": 7120 + }, + { + "epoch": 0.53, + "learning_rate": 1.982590300649742e-05, + "loss": 0.9837, + "step": 7121 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825828894064983e-05, + "loss": 1.0613, + "step": 7122 + }, + { + "epoch": 0.53, + "learning_rate": 1.982575476599979e-05, + "loss": 1.1148, + "step": 7123 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825680622301958e-05, + "loss": 0.9985, + "step": 7124 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825606462971612e-05, + "loss": 1.0553, + "step": 7125 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825532288008867e-05, + "loss": 0.9628, + "step": 7126 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825458097413844e-05, + "loss": 1.0766, + "step": 7127 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825383891186652e-05, + "loss": 1.1899, + "step": 7128 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825309669327416e-05, + "loss": 0.9822, + "step": 7129 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825235431836255e-05, + "loss": 1.0979, + "step": 7130 + }, + { + "epoch": 0.53, + "learning_rate": 1.9825161178713284e-05, + "loss": 0.998, + "step": 7131 + }, + { + "epoch": 0.53, + "learning_rate": 1.982508690995862e-05, + "loss": 1.1417, + "step": 7132 + }, + { + "epoch": 0.53, + "learning_rate": 1.982501262557239e-05, + "loss": 1.0745, + "step": 7133 + }, + { + "epoch": 0.53, + "learning_rate": 1.98249383255547e-05, + "loss": 1.1043, + "step": 7134 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824864009905673e-05, + "loss": 0.9679, + "step": 7135 + }, + { + "epoch": 0.53, + "learning_rate": 1.982478967862543e-05, + "loss": 1.0133, + "step": 7136 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824715331714085e-05, + "loss": 1.0194, + "step": 7137 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824640969171764e-05, + "loss": 1.061, + "step": 7138 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824566590998574e-05, + "loss": 1.0118, + "step": 7139 + }, + { + "epoch": 0.53, + "learning_rate": 1.982449219719464e-05, + "loss": 1.1467, + "step": 7140 + }, + { + "epoch": 0.53, + "learning_rate": 1.982441778776008e-05, + "loss": 0.9649, + "step": 7141 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824343362695012e-05, + "loss": 1.0206, + "step": 7142 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824268921999556e-05, + "loss": 1.0953, + "step": 7143 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824194465673828e-05, + "loss": 0.9544, + "step": 7144 + }, + { + "epoch": 0.53, + "learning_rate": 1.9824119993717946e-05, + "loss": 1.0634, + "step": 7145 + }, + { + "epoch": 0.53, + "learning_rate": 1.982404550613203e-05, + "loss": 1.1122, + "step": 7146 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823971002916198e-05, + "loss": 1.0297, + "step": 7147 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823896484070572e-05, + "loss": 1.1221, + "step": 7148 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823821949595263e-05, + "loss": 0.9713, + "step": 7149 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823747399490397e-05, + "loss": 0.9867, + "step": 7150 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823672833756083e-05, + "loss": 1.0246, + "step": 7151 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823598252392453e-05, + "loss": 1.0944, + "step": 7152 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823523655399617e-05, + "loss": 1.0906, + "step": 7153 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823449042777696e-05, + "loss": 1.0096, + "step": 7154 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823374414526807e-05, + "loss": 1.0567, + "step": 7155 + }, + { + "epoch": 0.53, + "learning_rate": 1.982329977064707e-05, + "loss": 1.114, + "step": 7156 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823225111138602e-05, + "loss": 1.0779, + "step": 7157 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823150436001523e-05, + "loss": 1.1001, + "step": 7158 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823075745235957e-05, + "loss": 1.0964, + "step": 7159 + }, + { + "epoch": 0.53, + "learning_rate": 1.9823001038842013e-05, + "loss": 1.0775, + "step": 7160 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822926316819817e-05, + "loss": 0.9982, + "step": 7161 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822851579169486e-05, + "loss": 1.0095, + "step": 7162 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822776825891137e-05, + "loss": 1.1062, + "step": 7163 + }, + { + "epoch": 0.53, + "learning_rate": 1.982270205698489e-05, + "loss": 1.143, + "step": 7164 + }, + { + "epoch": 0.53, + "learning_rate": 1.982262727245087e-05, + "loss": 1.1433, + "step": 7165 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822552472289187e-05, + "loss": 1.0564, + "step": 7166 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822477656499966e-05, + "loss": 0.9974, + "step": 7167 + }, + { + "epoch": 0.53, + "learning_rate": 1.982240282508332e-05, + "loss": 1.075, + "step": 7168 + }, + { + "epoch": 0.53, + "learning_rate": 1.982232797803937e-05, + "loss": 1.0286, + "step": 7169 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822253115368242e-05, + "loss": 1.0067, + "step": 7170 + }, + { + "epoch": 0.53, + "learning_rate": 1.982217823707005e-05, + "loss": 1.0559, + "step": 7171 + }, + { + "epoch": 0.53, + "learning_rate": 1.9822103343144908e-05, + "loss": 1.0473, + "step": 7172 + }, + { + "epoch": 0.53, + "learning_rate": 1.982202843359294e-05, + "loss": 1.1657, + "step": 7173 + }, + { + "epoch": 0.53, + "learning_rate": 1.982195350841427e-05, + "loss": 0.989, + "step": 7174 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821878567609013e-05, + "loss": 0.9263, + "step": 7175 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821803611177282e-05, + "loss": 1.0275, + "step": 7176 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821728639119207e-05, + "loss": 1.1683, + "step": 7177 + }, + { + "epoch": 0.53, + "learning_rate": 1.98216536514349e-05, + "loss": 1.0911, + "step": 7178 + }, + { + "epoch": 0.53, + "learning_rate": 1.982157864812448e-05, + "loss": 1.0483, + "step": 7179 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821503629188075e-05, + "loss": 1.0129, + "step": 7180 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821428594625795e-05, + "loss": 1.0933, + "step": 7181 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821353544437763e-05, + "loss": 1.0647, + "step": 7182 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821278478624094e-05, + "loss": 1.0983, + "step": 7183 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821203397184917e-05, + "loss": 1.0894, + "step": 7184 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821128300120343e-05, + "loss": 0.9981, + "step": 7185 + }, + { + "epoch": 0.53, + "learning_rate": 1.9821053187430496e-05, + "loss": 1.0325, + "step": 7186 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820978059115495e-05, + "loss": 1.1062, + "step": 7187 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820902915175458e-05, + "loss": 1.0739, + "step": 7188 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820827755610505e-05, + "loss": 1.1191, + "step": 7189 + }, + { + "epoch": 0.53, + "learning_rate": 1.982075258042075e-05, + "loss": 1.0463, + "step": 7190 + }, + { + "epoch": 0.53, + "learning_rate": 1.982067738960632e-05, + "loss": 1.0716, + "step": 7191 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820602183167338e-05, + "loss": 1.0743, + "step": 7192 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820526961103917e-05, + "loss": 1.0418, + "step": 7193 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820451723416177e-05, + "loss": 0.9955, + "step": 7194 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820376470104236e-05, + "loss": 1.1494, + "step": 7195 + }, + { + "epoch": 0.53, + "learning_rate": 1.982030120116822e-05, + "loss": 1.1932, + "step": 7196 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820225916608246e-05, + "loss": 1.0893, + "step": 7197 + }, + { + "epoch": 0.53, + "learning_rate": 1.982015061642443e-05, + "loss": 1.048, + "step": 7198 + }, + { + "epoch": 0.53, + "learning_rate": 1.9820075300616897e-05, + "loss": 1.1303, + "step": 7199 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819999969185765e-05, + "loss": 1.0106, + "step": 7200 + }, + { + "epoch": 0.53, + "learning_rate": 1.981992462213115e-05, + "loss": 1.0508, + "step": 7201 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819849259453182e-05, + "loss": 0.9661, + "step": 7202 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819773881151965e-05, + "loss": 1.0597, + "step": 7203 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819698487227636e-05, + "loss": 1.0065, + "step": 7204 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819623077680306e-05, + "loss": 1.0708, + "step": 7205 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819547652510094e-05, + "loss": 0.8785, + "step": 7206 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819472211717124e-05, + "loss": 1.0689, + "step": 7207 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819396755301513e-05, + "loss": 1.0437, + "step": 7208 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819321283263382e-05, + "loss": 1.0201, + "step": 7209 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819245795602853e-05, + "loss": 1.0069, + "step": 7210 + }, + { + "epoch": 0.53, + "learning_rate": 1.981917029232004e-05, + "loss": 1.0932, + "step": 7211 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819094773415073e-05, + "loss": 1.0034, + "step": 7212 + }, + { + "epoch": 0.53, + "learning_rate": 1.9819019238888065e-05, + "loss": 1.0653, + "step": 7213 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818943688739137e-05, + "loss": 0.9771, + "step": 7214 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818868122968415e-05, + "loss": 0.966, + "step": 7215 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818792541576006e-05, + "loss": 1.0225, + "step": 7216 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818716944562043e-05, + "loss": 1.0724, + "step": 7217 + }, + { + "epoch": 0.53, + "learning_rate": 1.981864133192664e-05, + "loss": 0.9443, + "step": 7218 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818565703669922e-05, + "loss": 1.0357, + "step": 7219 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818490059792004e-05, + "loss": 1.0959, + "step": 7220 + }, + { + "epoch": 0.53, + "learning_rate": 1.981841440029301e-05, + "loss": 1.031, + "step": 7221 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818338725173062e-05, + "loss": 1.0916, + "step": 7222 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818263034432274e-05, + "loss": 1.0777, + "step": 7223 + }, + { + "epoch": 0.53, + "learning_rate": 1.981818732807077e-05, + "loss": 1.0613, + "step": 7224 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818111606088668e-05, + "loss": 1.0121, + "step": 7225 + }, + { + "epoch": 0.53, + "learning_rate": 1.9818035868486098e-05, + "loss": 1.1262, + "step": 7226 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817960115263166e-05, + "loss": 1.1108, + "step": 7227 + }, + { + "epoch": 0.53, + "learning_rate": 1.981788434642e-05, + "loss": 1.1325, + "step": 7228 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817808561956725e-05, + "loss": 1.0484, + "step": 7229 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817732761873457e-05, + "loss": 1.1099, + "step": 7230 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817656946170315e-05, + "loss": 0.998, + "step": 7231 + }, + { + "epoch": 0.53, + "learning_rate": 1.981758111484742e-05, + "loss": 1.0544, + "step": 7232 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817505267904898e-05, + "loss": 1.0704, + "step": 7233 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817429405342864e-05, + "loss": 1.0503, + "step": 7234 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817353527161437e-05, + "loss": 1.1195, + "step": 7235 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817277633360742e-05, + "loss": 1.0008, + "step": 7236 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817201723940903e-05, + "loss": 0.9442, + "step": 7237 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817125798902033e-05, + "loss": 1.1085, + "step": 7238 + }, + { + "epoch": 0.53, + "learning_rate": 1.9817049858244256e-05, + "loss": 1.0579, + "step": 7239 + }, + { + "epoch": 0.53, + "learning_rate": 1.9816973901967692e-05, + "loss": 1.0263, + "step": 7240 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816897930072468e-05, + "loss": 1.0092, + "step": 7241 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816821942558694e-05, + "loss": 1.0429, + "step": 7242 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816745939426504e-05, + "loss": 1.0993, + "step": 7243 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816669920676002e-05, + "loss": 1.0058, + "step": 7244 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816593886307327e-05, + "loss": 0.976, + "step": 7245 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816517836320588e-05, + "loss": 1.0727, + "step": 7246 + }, + { + "epoch": 0.54, + "learning_rate": 1.981644177071591e-05, + "loss": 1.0451, + "step": 7247 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816365689493415e-05, + "loss": 1.0569, + "step": 7248 + }, + { + "epoch": 0.54, + "learning_rate": 1.981628959265322e-05, + "loss": 1.1043, + "step": 7249 + }, + { + "epoch": 0.54, + "learning_rate": 1.981621348019545e-05, + "loss": 1.0346, + "step": 7250 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816137352120225e-05, + "loss": 1.003, + "step": 7251 + }, + { + "epoch": 0.54, + "learning_rate": 1.9816061208427664e-05, + "loss": 1.1066, + "step": 7252 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815985049117893e-05, + "loss": 0.9783, + "step": 7253 + }, + { + "epoch": 0.54, + "learning_rate": 1.981590887419103e-05, + "loss": 0.9127, + "step": 7254 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815832683647195e-05, + "loss": 1.0127, + "step": 7255 + }, + { + "epoch": 0.54, + "learning_rate": 1.981575647748651e-05, + "loss": 1.0189, + "step": 7256 + }, + { + "epoch": 0.54, + "learning_rate": 1.98156802557091e-05, + "loss": 1.1112, + "step": 7257 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815604018315076e-05, + "loss": 1.0667, + "step": 7258 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815527765304575e-05, + "loss": 1.0428, + "step": 7259 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815451496677704e-05, + "loss": 0.9988, + "step": 7260 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815375212434594e-05, + "loss": 1.0672, + "step": 7261 + }, + { + "epoch": 0.54, + "learning_rate": 1.981529891257536e-05, + "loss": 1.0781, + "step": 7262 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815222597100128e-05, + "loss": 1.1561, + "step": 7263 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815146266009015e-05, + "loss": 0.966, + "step": 7264 + }, + { + "epoch": 0.54, + "learning_rate": 1.9815069919302145e-05, + "loss": 0.9904, + "step": 7265 + }, + { + "epoch": 0.54, + "learning_rate": 1.981499355697964e-05, + "loss": 1.0909, + "step": 7266 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814917179041622e-05, + "loss": 1.0312, + "step": 7267 + }, + { + "epoch": 0.54, + "learning_rate": 1.981484078548821e-05, + "loss": 1.156, + "step": 7268 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814764376319527e-05, + "loss": 1.0355, + "step": 7269 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814687951535694e-05, + "loss": 1.009, + "step": 7270 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814611511136833e-05, + "loss": 1.0705, + "step": 7271 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814535055123065e-05, + "loss": 1.0512, + "step": 7272 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814458583494512e-05, + "loss": 1.0388, + "step": 7273 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814382096251297e-05, + "loss": 0.984, + "step": 7274 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814305593393537e-05, + "loss": 1.1438, + "step": 7275 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814229074921362e-05, + "loss": 1.1146, + "step": 7276 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814152540834887e-05, + "loss": 1.1081, + "step": 7277 + }, + { + "epoch": 0.54, + "learning_rate": 1.9814075991134233e-05, + "loss": 1.0281, + "step": 7278 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813999425819527e-05, + "loss": 1.0143, + "step": 7279 + }, + { + "epoch": 0.54, + "learning_rate": 1.981392284489089e-05, + "loss": 1.0066, + "step": 7280 + }, + { + "epoch": 0.54, + "learning_rate": 1.981384624834844e-05, + "loss": 0.9852, + "step": 7281 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813769636192303e-05, + "loss": 0.9628, + "step": 7282 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813693008422598e-05, + "loss": 1.1461, + "step": 7283 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813616365039447e-05, + "loss": 1.0349, + "step": 7284 + }, + { + "epoch": 0.54, + "learning_rate": 1.981353970604297e-05, + "loss": 1.0408, + "step": 7285 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813463031433296e-05, + "loss": 0.9512, + "step": 7286 + }, + { + "epoch": 0.54, + "learning_rate": 1.981338634121054e-05, + "loss": 0.9958, + "step": 7287 + }, + { + "epoch": 0.54, + "learning_rate": 1.981330963537483e-05, + "loss": 1.0121, + "step": 7288 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813232913926283e-05, + "loss": 1.0597, + "step": 7289 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813156176865025e-05, + "loss": 1.1564, + "step": 7290 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813079424191174e-05, + "loss": 1.0256, + "step": 7291 + }, + { + "epoch": 0.54, + "learning_rate": 1.9813002655904856e-05, + "loss": 1.1851, + "step": 7292 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812925872006185e-05, + "loss": 1.046, + "step": 7293 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812849072495295e-05, + "loss": 0.9514, + "step": 7294 + }, + { + "epoch": 0.54, + "learning_rate": 1.98127722573723e-05, + "loss": 1.0327, + "step": 7295 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812695426637326e-05, + "loss": 1.0661, + "step": 7296 + }, + { + "epoch": 0.54, + "learning_rate": 1.981261858029049e-05, + "loss": 1.076, + "step": 7297 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812541718331925e-05, + "loss": 1.0129, + "step": 7298 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812464840761744e-05, + "loss": 0.9992, + "step": 7299 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812387947580073e-05, + "loss": 1.0574, + "step": 7300 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812311038787028e-05, + "loss": 1.0636, + "step": 7301 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812234114382743e-05, + "loss": 0.9827, + "step": 7302 + }, + { + "epoch": 0.54, + "learning_rate": 1.981215717436733e-05, + "loss": 1.084, + "step": 7303 + }, + { + "epoch": 0.54, + "learning_rate": 1.9812080218740917e-05, + "loss": 1.2032, + "step": 7304 + }, + { + "epoch": 0.54, + "learning_rate": 1.981200324750362e-05, + "loss": 1.0836, + "step": 7305 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811926260655576e-05, + "loss": 1.0377, + "step": 7306 + }, + { + "epoch": 0.54, + "learning_rate": 1.981184925819689e-05, + "loss": 1.1518, + "step": 7307 + }, + { + "epoch": 0.54, + "learning_rate": 1.981177224012769e-05, + "loss": 0.9554, + "step": 7308 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811695206448108e-05, + "loss": 0.9122, + "step": 7309 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811618157158254e-05, + "loss": 1.1001, + "step": 7310 + }, + { + "epoch": 0.54, + "learning_rate": 1.981154109225826e-05, + "loss": 1.1664, + "step": 7311 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811464011748242e-05, + "loss": 1.1692, + "step": 7312 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811386915628325e-05, + "loss": 1.049, + "step": 7313 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811309803898634e-05, + "loss": 0.9756, + "step": 7314 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811232676559288e-05, + "loss": 1.0806, + "step": 7315 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811155533610412e-05, + "loss": 0.9929, + "step": 7316 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811078375052126e-05, + "loss": 1.1248, + "step": 7317 + }, + { + "epoch": 0.54, + "learning_rate": 1.9811001200884553e-05, + "loss": 0.9924, + "step": 7318 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810924011107823e-05, + "loss": 1.015, + "step": 7319 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810846805722052e-05, + "loss": 1.1295, + "step": 7320 + }, + { + "epoch": 0.54, + "learning_rate": 1.981076958472736e-05, + "loss": 1.1004, + "step": 7321 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810692348123878e-05, + "loss": 1.0291, + "step": 7322 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810615095911723e-05, + "loss": 1.1075, + "step": 7323 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810537828091018e-05, + "loss": 1.0091, + "step": 7324 + }, + { + "epoch": 0.54, + "learning_rate": 1.981046054466189e-05, + "loss": 1.0627, + "step": 7325 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810383245624463e-05, + "loss": 0.9167, + "step": 7326 + }, + { + "epoch": 0.54, + "learning_rate": 1.981030593097885e-05, + "loss": 1.1481, + "step": 7327 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810228600725184e-05, + "loss": 0.9647, + "step": 7328 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810151254863586e-05, + "loss": 1.0394, + "step": 7329 + }, + { + "epoch": 0.54, + "learning_rate": 1.9810073893394176e-05, + "loss": 1.0205, + "step": 7330 + }, + { + "epoch": 0.54, + "learning_rate": 1.980999651631708e-05, + "loss": 0.9797, + "step": 7331 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809919123632418e-05, + "loss": 1.0341, + "step": 7332 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809841715340314e-05, + "loss": 1.0432, + "step": 7333 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809764291440892e-05, + "loss": 1.0267, + "step": 7334 + }, + { + "epoch": 0.54, + "learning_rate": 1.980968685193428e-05, + "loss": 0.9927, + "step": 7335 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809609396820596e-05, + "loss": 1.1301, + "step": 7336 + }, + { + "epoch": 0.54, + "learning_rate": 1.980953192609996e-05, + "loss": 1.0943, + "step": 7337 + }, + { + "epoch": 0.54, + "learning_rate": 1.98094544397725e-05, + "loss": 1.0606, + "step": 7338 + }, + { + "epoch": 0.54, + "learning_rate": 1.980937693783834e-05, + "loss": 1.0834, + "step": 7339 + }, + { + "epoch": 0.54, + "learning_rate": 1.98092994202976e-05, + "loss": 1.0348, + "step": 7340 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809221887150407e-05, + "loss": 0.9922, + "step": 7341 + }, + { + "epoch": 0.54, + "learning_rate": 1.980914433839688e-05, + "loss": 1.0899, + "step": 7342 + }, + { + "epoch": 0.54, + "learning_rate": 1.9809066774037146e-05, + "loss": 1.0181, + "step": 7343 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808989194071328e-05, + "loss": 1.0581, + "step": 7344 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808911598499545e-05, + "loss": 1.0158, + "step": 7345 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808833987321928e-05, + "loss": 1.0703, + "step": 7346 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808756360538594e-05, + "loss": 1.0003, + "step": 7347 + }, + { + "epoch": 0.54, + "learning_rate": 1.980867871814967e-05, + "loss": 1.0893, + "step": 7348 + }, + { + "epoch": 0.54, + "learning_rate": 1.980860106015528e-05, + "loss": 1.1262, + "step": 7349 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808523386555542e-05, + "loss": 1.053, + "step": 7350 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808445697350585e-05, + "loss": 0.9489, + "step": 7351 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808367992540534e-05, + "loss": 1.1486, + "step": 7352 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808290272125507e-05, + "loss": 1.0555, + "step": 7353 + }, + { + "epoch": 0.54, + "learning_rate": 1.9808212536105632e-05, + "loss": 1.0733, + "step": 7354 + }, + { + "epoch": 0.54, + "learning_rate": 1.980813478448103e-05, + "loss": 0.9306, + "step": 7355 + }, + { + "epoch": 0.54, + "learning_rate": 1.980805701725183e-05, + "loss": 1.0679, + "step": 7356 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807979234418147e-05, + "loss": 1.0502, + "step": 7357 + }, + { + "epoch": 0.54, + "learning_rate": 1.980790143598011e-05, + "loss": 0.9663, + "step": 7358 + }, + { + "epoch": 0.54, + "learning_rate": 1.980782362193784e-05, + "loss": 1.0483, + "step": 7359 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807745792291468e-05, + "loss": 1.0661, + "step": 7360 + }, + { + "epoch": 0.54, + "learning_rate": 1.980766794704111e-05, + "loss": 1.093, + "step": 7361 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807590086186896e-05, + "loss": 0.9712, + "step": 7362 + }, + { + "epoch": 0.54, + "learning_rate": 1.980751220972894e-05, + "loss": 1.0218, + "step": 7363 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807434317667377e-05, + "loss": 1.031, + "step": 7364 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807356410002326e-05, + "loss": 1.1789, + "step": 7365 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807278486733913e-05, + "loss": 1.0547, + "step": 7366 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807200547862258e-05, + "loss": 0.9547, + "step": 7367 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807122593387483e-05, + "loss": 1.1455, + "step": 7368 + }, + { + "epoch": 0.54, + "learning_rate": 1.9807044623309722e-05, + "loss": 0.9925, + "step": 7369 + }, + { + "epoch": 0.54, + "learning_rate": 1.980696663762909e-05, + "loss": 0.9949, + "step": 7370 + }, + { + "epoch": 0.54, + "learning_rate": 1.9806888636345718e-05, + "loss": 1.0773, + "step": 7371 + }, + { + "epoch": 0.54, + "learning_rate": 1.9806810619459727e-05, + "loss": 0.9792, + "step": 7372 + }, + { + "epoch": 0.54, + "learning_rate": 1.9806732586971236e-05, + "loss": 1.0496, + "step": 7373 + }, + { + "epoch": 0.54, + "learning_rate": 1.9806654538880378e-05, + "loss": 1.1067, + "step": 7374 + }, + { + "epoch": 0.54, + "learning_rate": 1.9806576475187268e-05, + "loss": 1.003, + "step": 7375 + }, + { + "epoch": 0.55, + "learning_rate": 1.9806498395892038e-05, + "loss": 0.9907, + "step": 7376 + }, + { + "epoch": 0.55, + "learning_rate": 1.980642030099481e-05, + "loss": 1.007, + "step": 7377 + }, + { + "epoch": 0.55, + "learning_rate": 1.980634219049571e-05, + "loss": 1.0428, + "step": 7378 + }, + { + "epoch": 0.55, + "learning_rate": 1.9806264064394852e-05, + "loss": 1.0589, + "step": 7379 + }, + { + "epoch": 0.55, + "learning_rate": 1.9806185922692375e-05, + "loss": 1.1043, + "step": 7380 + }, + { + "epoch": 0.55, + "learning_rate": 1.9806107765388396e-05, + "loss": 1.0597, + "step": 7381 + }, + { + "epoch": 0.55, + "learning_rate": 1.980602959248304e-05, + "loss": 1.0134, + "step": 7382 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805951403976425e-05, + "loss": 0.956, + "step": 7383 + }, + { + "epoch": 0.55, + "learning_rate": 1.980587319986869e-05, + "loss": 0.9885, + "step": 7384 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805794980159946e-05, + "loss": 1.0201, + "step": 7385 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805716744850324e-05, + "loss": 1.0045, + "step": 7386 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805638493939944e-05, + "loss": 0.9112, + "step": 7387 + }, + { + "epoch": 0.55, + "learning_rate": 1.980556022742894e-05, + "loss": 1.0734, + "step": 7388 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805481945317423e-05, + "loss": 0.9775, + "step": 7389 + }, + { + "epoch": 0.55, + "learning_rate": 1.980540364760553e-05, + "loss": 1.0182, + "step": 7390 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805325334293377e-05, + "loss": 1.0658, + "step": 7391 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805247005381095e-05, + "loss": 0.893, + "step": 7392 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805168660868797e-05, + "loss": 1.0528, + "step": 7393 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805090300756626e-05, + "loss": 1.0098, + "step": 7394 + }, + { + "epoch": 0.55, + "learning_rate": 1.9805011925044693e-05, + "loss": 1.0314, + "step": 7395 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804933533733123e-05, + "loss": 0.9727, + "step": 7396 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804855126822046e-05, + "loss": 1.0064, + "step": 7397 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804776704311587e-05, + "loss": 1.0049, + "step": 7398 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804698266201867e-05, + "loss": 1.0085, + "step": 7399 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804619812493013e-05, + "loss": 0.9085, + "step": 7400 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804541343185148e-05, + "loss": 0.9741, + "step": 7401 + }, + { + "epoch": 0.55, + "learning_rate": 1.98044628582784e-05, + "loss": 1.0003, + "step": 7402 + }, + { + "epoch": 0.55, + "learning_rate": 1.980438435777289e-05, + "loss": 1.024, + "step": 7403 + }, + { + "epoch": 0.55, + "learning_rate": 1.980430584166874e-05, + "loss": 1.0605, + "step": 7404 + }, + { + "epoch": 0.55, + "learning_rate": 1.980422730996609e-05, + "loss": 1.2175, + "step": 7405 + }, + { + "epoch": 0.55, + "learning_rate": 1.980414876266505e-05, + "loss": 1.1286, + "step": 7406 + }, + { + "epoch": 0.55, + "learning_rate": 1.9804070199765748e-05, + "loss": 0.9805, + "step": 7407 + }, + { + "epoch": 0.55, + "learning_rate": 1.980399162126831e-05, + "loss": 1.0475, + "step": 7408 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803913027172865e-05, + "loss": 1.0094, + "step": 7409 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803834417479532e-05, + "loss": 0.9605, + "step": 7410 + }, + { + "epoch": 0.55, + "learning_rate": 1.980375579218844e-05, + "loss": 1.1038, + "step": 7411 + }, + { + "epoch": 0.55, + "learning_rate": 1.980367715129971e-05, + "loss": 1.0498, + "step": 7412 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803598494813475e-05, + "loss": 1.0161, + "step": 7413 + }, + { + "epoch": 0.55, + "learning_rate": 1.980351982272985e-05, + "loss": 1.0995, + "step": 7414 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803441135048968e-05, + "loss": 1.0326, + "step": 7415 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803362431770952e-05, + "loss": 1.0695, + "step": 7416 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803283712895926e-05, + "loss": 1.1317, + "step": 7417 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803204978424015e-05, + "loss": 1.054, + "step": 7418 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803126228355344e-05, + "loss": 0.8603, + "step": 7419 + }, + { + "epoch": 0.55, + "learning_rate": 1.9803047462690045e-05, + "loss": 1.0894, + "step": 7420 + }, + { + "epoch": 0.55, + "learning_rate": 1.980296868142823e-05, + "loss": 1.0263, + "step": 7421 + }, + { + "epoch": 0.55, + "learning_rate": 1.980288988457004e-05, + "loss": 1.0259, + "step": 7422 + }, + { + "epoch": 0.55, + "learning_rate": 1.9802811072115585e-05, + "loss": 0.9307, + "step": 7423 + }, + { + "epoch": 0.55, + "learning_rate": 1.9802732244065002e-05, + "loss": 0.9175, + "step": 7424 + }, + { + "epoch": 0.55, + "learning_rate": 1.980265340041841e-05, + "loss": 1.0203, + "step": 7425 + }, + { + "epoch": 0.55, + "learning_rate": 1.980257454117594e-05, + "loss": 1.0615, + "step": 7426 + }, + { + "epoch": 0.55, + "learning_rate": 1.9802495666337714e-05, + "loss": 1.0068, + "step": 7427 + }, + { + "epoch": 0.55, + "learning_rate": 1.9802416775903856e-05, + "loss": 0.9936, + "step": 7428 + }, + { + "epoch": 0.55, + "learning_rate": 1.980233786987449e-05, + "loss": 1.0593, + "step": 7429 + }, + { + "epoch": 0.55, + "learning_rate": 1.980225894824975e-05, + "loss": 1.0213, + "step": 7430 + }, + { + "epoch": 0.55, + "learning_rate": 1.9802180011029755e-05, + "loss": 1.0849, + "step": 7431 + }, + { + "epoch": 0.55, + "learning_rate": 1.980210105821463e-05, + "loss": 1.0192, + "step": 7432 + }, + { + "epoch": 0.55, + "learning_rate": 1.98020220898045e-05, + "loss": 1.1518, + "step": 7433 + }, + { + "epoch": 0.55, + "learning_rate": 1.98019431057995e-05, + "loss": 1.0706, + "step": 7434 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801864106199743e-05, + "loss": 0.7643, + "step": 7435 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801785091005365e-05, + "loss": 1.0114, + "step": 7436 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801706060216484e-05, + "loss": 0.9986, + "step": 7437 + }, + { + "epoch": 0.55, + "learning_rate": 1.980162701383323e-05, + "loss": 0.983, + "step": 7438 + }, + { + "epoch": 0.55, + "learning_rate": 1.980154795185573e-05, + "loss": 1.0199, + "step": 7439 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801468874284104e-05, + "loss": 1.0785, + "step": 7440 + }, + { + "epoch": 0.55, + "learning_rate": 1.980138978111848e-05, + "loss": 1.0658, + "step": 7441 + }, + { + "epoch": 0.55, + "learning_rate": 1.980131067235899e-05, + "loss": 1.0118, + "step": 7442 + }, + { + "epoch": 0.55, + "learning_rate": 1.980123154800575e-05, + "loss": 1.1551, + "step": 7443 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801152408058893e-05, + "loss": 1.1364, + "step": 7444 + }, + { + "epoch": 0.55, + "learning_rate": 1.9801073252518545e-05, + "loss": 1.0666, + "step": 7445 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800994081384826e-05, + "loss": 1.0666, + "step": 7446 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800914894657866e-05, + "loss": 1.0572, + "step": 7447 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800835692337793e-05, + "loss": 0.9307, + "step": 7448 + }, + { + "epoch": 0.55, + "learning_rate": 1.980075647442473e-05, + "loss": 1.0655, + "step": 7449 + }, + { + "epoch": 0.55, + "learning_rate": 1.98006772409188e-05, + "loss": 0.8923, + "step": 7450 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800597991820138e-05, + "loss": 1.0977, + "step": 7451 + }, + { + "epoch": 0.55, + "learning_rate": 1.980051872712886e-05, + "loss": 1.0041, + "step": 7452 + }, + { + "epoch": 0.55, + "learning_rate": 1.98004394468451e-05, + "loss": 1.0131, + "step": 7453 + }, + { + "epoch": 0.55, + "learning_rate": 1.980036015096898e-05, + "loss": 1.02, + "step": 7454 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800280839500628e-05, + "loss": 1.1253, + "step": 7455 + }, + { + "epoch": 0.55, + "learning_rate": 1.980020151244017e-05, + "loss": 0.982, + "step": 7456 + }, + { + "epoch": 0.55, + "learning_rate": 1.980012216978773e-05, + "loss": 1.0363, + "step": 7457 + }, + { + "epoch": 0.55, + "learning_rate": 1.9800042811543437e-05, + "loss": 1.0305, + "step": 7458 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799963437707415e-05, + "loss": 0.9938, + "step": 7459 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799884048279793e-05, + "loss": 1.144, + "step": 7460 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799804643260697e-05, + "loss": 1.0564, + "step": 7461 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799725222650248e-05, + "loss": 1.0255, + "step": 7462 + }, + { + "epoch": 0.55, + "learning_rate": 1.979964578644858e-05, + "loss": 1.1494, + "step": 7463 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799566334655812e-05, + "loss": 0.9932, + "step": 7464 + }, + { + "epoch": 0.55, + "learning_rate": 1.979948686727208e-05, + "loss": 1.025, + "step": 7465 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799407384297498e-05, + "loss": 1.0801, + "step": 7466 + }, + { + "epoch": 0.55, + "learning_rate": 1.97993278857322e-05, + "loss": 0.969, + "step": 7467 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799248371576315e-05, + "loss": 1.1303, + "step": 7468 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799168841829967e-05, + "loss": 1.0402, + "step": 7469 + }, + { + "epoch": 0.55, + "learning_rate": 1.9799089296493276e-05, + "loss": 0.947, + "step": 7470 + }, + { + "epoch": 0.55, + "learning_rate": 1.979900973556638e-05, + "loss": 1.0797, + "step": 7471 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798930159049396e-05, + "loss": 1.1206, + "step": 7472 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798850566942453e-05, + "loss": 1.0627, + "step": 7473 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798770959245682e-05, + "loss": 1.0801, + "step": 7474 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798691335959206e-05, + "loss": 1.0624, + "step": 7475 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798611697083152e-05, + "loss": 0.9546, + "step": 7476 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798532042617647e-05, + "loss": 1.0121, + "step": 7477 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798452372562816e-05, + "loss": 1.0575, + "step": 7478 + }, + { + "epoch": 0.55, + "learning_rate": 1.979837268691879e-05, + "loss": 1.0964, + "step": 7479 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798292985685693e-05, + "loss": 1.1068, + "step": 7480 + }, + { + "epoch": 0.55, + "learning_rate": 1.979821326886365e-05, + "loss": 1.1378, + "step": 7481 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798133536452792e-05, + "loss": 1.1395, + "step": 7482 + }, + { + "epoch": 0.55, + "learning_rate": 1.9798053788453244e-05, + "loss": 1.1228, + "step": 7483 + }, + { + "epoch": 0.55, + "learning_rate": 1.979797402486513e-05, + "loss": 1.0495, + "step": 7484 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797894245688584e-05, + "loss": 1.0229, + "step": 7485 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797814450923726e-05, + "loss": 1.1039, + "step": 7486 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797734640570684e-05, + "loss": 0.9898, + "step": 7487 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797654814629586e-05, + "loss": 1.0355, + "step": 7488 + }, + { + "epoch": 0.55, + "learning_rate": 1.979757497310056e-05, + "loss": 1.0586, + "step": 7489 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797495115983735e-05, + "loss": 1.0576, + "step": 7490 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797415243279232e-05, + "loss": 1.0609, + "step": 7491 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797335354987185e-05, + "loss": 1.0248, + "step": 7492 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797255451107716e-05, + "loss": 0.9561, + "step": 7493 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797175531640953e-05, + "loss": 1.0525, + "step": 7494 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797095596587023e-05, + "loss": 0.9709, + "step": 7495 + }, + { + "epoch": 0.55, + "learning_rate": 1.9797015645946056e-05, + "loss": 0.9739, + "step": 7496 + }, + { + "epoch": 0.55, + "learning_rate": 1.979693567971818e-05, + "loss": 1.1055, + "step": 7497 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796855697903514e-05, + "loss": 0.9762, + "step": 7498 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796775700502192e-05, + "loss": 1.0602, + "step": 7499 + }, + { + "epoch": 0.55, + "learning_rate": 1.979669568751434e-05, + "loss": 1.0697, + "step": 7500 + }, + { + "epoch": 0.55, + "learning_rate": 1.979661565894009e-05, + "loss": 1.0922, + "step": 7501 + }, + { + "epoch": 0.55, + "learning_rate": 1.979653561477956e-05, + "loss": 0.9441, + "step": 7502 + }, + { + "epoch": 0.55, + "learning_rate": 1.979645555503288e-05, + "loss": 1.0194, + "step": 7503 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796375479700187e-05, + "loss": 1.0449, + "step": 7504 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796295388781592e-05, + "loss": 0.9934, + "step": 7505 + }, + { + "epoch": 0.55, + "learning_rate": 1.979621528227724e-05, + "loss": 1.0701, + "step": 7506 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796135160187243e-05, + "loss": 0.9898, + "step": 7507 + }, + { + "epoch": 0.55, + "learning_rate": 1.9796055022511736e-05, + "loss": 0.9846, + "step": 7508 + }, + { + "epoch": 0.55, + "learning_rate": 1.9795974869250845e-05, + "loss": 1.0399, + "step": 7509 + }, + { + "epoch": 0.55, + "learning_rate": 1.97958947004047e-05, + "loss": 1.0148, + "step": 7510 + }, + { + "epoch": 0.55, + "learning_rate": 1.979581451597343e-05, + "loss": 1.1094, + "step": 7511 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795734315957154e-05, + "loss": 1.1348, + "step": 7512 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795654100356008e-05, + "loss": 1.0686, + "step": 7513 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795573869170116e-05, + "loss": 1.0926, + "step": 7514 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795493622399604e-05, + "loss": 0.988, + "step": 7515 + }, + { + "epoch": 0.56, + "learning_rate": 1.97954133600446e-05, + "loss": 1.0517, + "step": 7516 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795333082105238e-05, + "loss": 1.0808, + "step": 7517 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795252788581642e-05, + "loss": 0.99, + "step": 7518 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795172479473937e-05, + "loss": 0.9707, + "step": 7519 + }, + { + "epoch": 0.56, + "learning_rate": 1.979509215478225e-05, + "loss": 0.999, + "step": 7520 + }, + { + "epoch": 0.56, + "learning_rate": 1.9795011814506716e-05, + "loss": 1.0243, + "step": 7521 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794931458647454e-05, + "loss": 1.0058, + "step": 7522 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794851087204602e-05, + "loss": 1.039, + "step": 7523 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794770700178278e-05, + "loss": 1.1542, + "step": 7524 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794690297568615e-05, + "loss": 0.9973, + "step": 7525 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794609879375738e-05, + "loss": 1.0702, + "step": 7526 + }, + { + "epoch": 0.56, + "learning_rate": 1.979452944559978e-05, + "loss": 0.9978, + "step": 7527 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794448996240863e-05, + "loss": 0.9793, + "step": 7528 + }, + { + "epoch": 0.56, + "learning_rate": 1.979436853129912e-05, + "loss": 1.0846, + "step": 7529 + }, + { + "epoch": 0.56, + "learning_rate": 1.9794288050774673e-05, + "loss": 0.9851, + "step": 7530 + }, + { + "epoch": 0.56, + "learning_rate": 1.979420755466766e-05, + "loss": 1.0917, + "step": 7531 + }, + { + "epoch": 0.56, + "learning_rate": 1.97941270429782e-05, + "loss": 1.0295, + "step": 7532 + }, + { + "epoch": 0.56, + "learning_rate": 1.979404651570642e-05, + "loss": 1.0644, + "step": 7533 + }, + { + "epoch": 0.56, + "learning_rate": 1.979396597285246e-05, + "loss": 0.9334, + "step": 7534 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793885414416435e-05, + "loss": 0.9527, + "step": 7535 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793804840398475e-05, + "loss": 1.1094, + "step": 7536 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793724250798716e-05, + "loss": 1.0805, + "step": 7537 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793643645617285e-05, + "loss": 1.0522, + "step": 7538 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793563024854303e-05, + "loss": 1.0213, + "step": 7539 + }, + { + "epoch": 0.56, + "learning_rate": 1.97934823885099e-05, + "loss": 1.1479, + "step": 7540 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793401736584213e-05, + "loss": 1.068, + "step": 7541 + }, + { + "epoch": 0.56, + "learning_rate": 1.979332106907736e-05, + "loss": 0.9911, + "step": 7542 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793240385989473e-05, + "loss": 1.0195, + "step": 7543 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793159687320684e-05, + "loss": 1.0154, + "step": 7544 + }, + { + "epoch": 0.56, + "learning_rate": 1.9793078973071113e-05, + "loss": 1.0598, + "step": 7545 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792998243240896e-05, + "loss": 0.9841, + "step": 7546 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792917497830162e-05, + "loss": 1.0699, + "step": 7547 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792836736839033e-05, + "loss": 0.9808, + "step": 7548 + }, + { + "epoch": 0.56, + "learning_rate": 1.979275596026764e-05, + "loss": 1.0974, + "step": 7549 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792675168116117e-05, + "loss": 1.0768, + "step": 7550 + }, + { + "epoch": 0.56, + "learning_rate": 1.979259436038458e-05, + "loss": 1.0332, + "step": 7551 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792513537073173e-05, + "loss": 1.0353, + "step": 7552 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792432698182012e-05, + "loss": 0.892, + "step": 7553 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792351843711236e-05, + "loss": 0.9841, + "step": 7554 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792270973660966e-05, + "loss": 0.9953, + "step": 7555 + }, + { + "epoch": 0.56, + "learning_rate": 1.979219008803133e-05, + "loss": 1.1079, + "step": 7556 + }, + { + "epoch": 0.56, + "learning_rate": 1.9792109186822462e-05, + "loss": 1.1952, + "step": 7557 + }, + { + "epoch": 0.56, + "learning_rate": 1.979202827003449e-05, + "loss": 1.1095, + "step": 7558 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791947337667538e-05, + "loss": 1.1021, + "step": 7559 + }, + { + "epoch": 0.56, + "learning_rate": 1.979186638972174e-05, + "loss": 0.9096, + "step": 7560 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791785426197223e-05, + "loss": 0.9451, + "step": 7561 + }, + { + "epoch": 0.56, + "learning_rate": 1.979170444709411e-05, + "loss": 0.9817, + "step": 7562 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791623452412543e-05, + "loss": 1.0095, + "step": 7563 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791542442152642e-05, + "loss": 0.9615, + "step": 7564 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791461416314534e-05, + "loss": 1.0901, + "step": 7565 + }, + { + "epoch": 0.56, + "learning_rate": 1.979138037489835e-05, + "loss": 1.0975, + "step": 7566 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791299317904223e-05, + "loss": 1.0904, + "step": 7567 + }, + { + "epoch": 0.56, + "learning_rate": 1.979121824533228e-05, + "loss": 1.1194, + "step": 7568 + }, + { + "epoch": 0.56, + "learning_rate": 1.9791137157182642e-05, + "loss": 1.0617, + "step": 7569 + }, + { + "epoch": 0.56, + "learning_rate": 1.979105605345545e-05, + "loss": 1.052, + "step": 7570 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790974934150828e-05, + "loss": 1.1138, + "step": 7571 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790893799268906e-05, + "loss": 1.0562, + "step": 7572 + }, + { + "epoch": 0.56, + "learning_rate": 1.979081264880981e-05, + "loss": 1.0304, + "step": 7573 + }, + { + "epoch": 0.56, + "learning_rate": 1.979073148277367e-05, + "loss": 1.0343, + "step": 7574 + }, + { + "epoch": 0.56, + "learning_rate": 1.979065030116062e-05, + "loss": 0.9948, + "step": 7575 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790569103970782e-05, + "loss": 1.0084, + "step": 7576 + }, + { + "epoch": 0.56, + "learning_rate": 1.979048789120429e-05, + "loss": 0.9779, + "step": 7577 + }, + { + "epoch": 0.56, + "learning_rate": 1.979040666286127e-05, + "loss": 0.946, + "step": 7578 + }, + { + "epoch": 0.56, + "learning_rate": 1.979032541894186e-05, + "loss": 0.9674, + "step": 7579 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790244159446176e-05, + "loss": 1.0636, + "step": 7580 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790162884374352e-05, + "loss": 1.0734, + "step": 7581 + }, + { + "epoch": 0.56, + "learning_rate": 1.9790081593726525e-05, + "loss": 1.1676, + "step": 7582 + }, + { + "epoch": 0.56, + "learning_rate": 1.979000028750281e-05, + "loss": 0.9701, + "step": 7583 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789918965703352e-05, + "loss": 1.0144, + "step": 7584 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789837628328274e-05, + "loss": 1.0762, + "step": 7585 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789756275377697e-05, + "loss": 0.9963, + "step": 7586 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789674906851763e-05, + "loss": 1.2059, + "step": 7587 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789593522750597e-05, + "loss": 0.9208, + "step": 7588 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789512123074325e-05, + "loss": 0.9964, + "step": 7589 + }, + { + "epoch": 0.56, + "learning_rate": 1.978943070782308e-05, + "loss": 1.0331, + "step": 7590 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789349276996993e-05, + "loss": 1.1012, + "step": 7591 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789267830596185e-05, + "loss": 0.9451, + "step": 7592 + }, + { + "epoch": 0.56, + "learning_rate": 1.97891863686208e-05, + "loss": 0.8631, + "step": 7593 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789104891070953e-05, + "loss": 1.1469, + "step": 7594 + }, + { + "epoch": 0.56, + "learning_rate": 1.9789023397946784e-05, + "loss": 1.0267, + "step": 7595 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788941889248417e-05, + "loss": 1.0518, + "step": 7596 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788860364975987e-05, + "loss": 1.0321, + "step": 7597 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788778825129616e-05, + "loss": 1.0038, + "step": 7598 + }, + { + "epoch": 0.56, + "learning_rate": 1.978869726970944e-05, + "loss": 1.0334, + "step": 7599 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788615698715588e-05, + "loss": 1.0511, + "step": 7600 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788534112148187e-05, + "loss": 0.9554, + "step": 7601 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788452510007364e-05, + "loss": 0.9938, + "step": 7602 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788370892293258e-05, + "loss": 0.9621, + "step": 7603 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788289259005994e-05, + "loss": 0.9699, + "step": 7604 + }, + { + "epoch": 0.56, + "learning_rate": 1.97882076101457e-05, + "loss": 1.0423, + "step": 7605 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788125945712512e-05, + "loss": 1.0666, + "step": 7606 + }, + { + "epoch": 0.56, + "learning_rate": 1.9788044265706553e-05, + "loss": 0.9712, + "step": 7607 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787962570127953e-05, + "loss": 1.0916, + "step": 7608 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787880858976843e-05, + "loss": 1.0089, + "step": 7609 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787799132253364e-05, + "loss": 0.9963, + "step": 7610 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787717389957628e-05, + "loss": 1.0877, + "step": 7611 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787635632089778e-05, + "loss": 0.9838, + "step": 7612 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787553858649935e-05, + "loss": 0.971, + "step": 7613 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787472069638238e-05, + "loss": 1.0106, + "step": 7614 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787390265054813e-05, + "loss": 1.1276, + "step": 7615 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787308444899788e-05, + "loss": 1.0214, + "step": 7616 + }, + { + "epoch": 0.56, + "learning_rate": 1.97872266091733e-05, + "loss": 0.9282, + "step": 7617 + }, + { + "epoch": 0.56, + "learning_rate": 1.9787144757875468e-05, + "loss": 1.1317, + "step": 7618 + }, + { + "epoch": 0.56, + "learning_rate": 1.978706289100643e-05, + "loss": 1.0286, + "step": 7619 + }, + { + "epoch": 0.56, + "learning_rate": 1.978698100856632e-05, + "loss": 1.0784, + "step": 7620 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786899110555258e-05, + "loss": 1.0372, + "step": 7621 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786817196973383e-05, + "loss": 1.0668, + "step": 7622 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786735267820817e-05, + "loss": 1.0161, + "step": 7623 + }, + { + "epoch": 0.56, + "learning_rate": 1.97866533230977e-05, + "loss": 1.092, + "step": 7624 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786571362804156e-05, + "loss": 0.9315, + "step": 7625 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786489386940318e-05, + "loss": 1.1094, + "step": 7626 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786407395506314e-05, + "loss": 1.0109, + "step": 7627 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786325388502275e-05, + "loss": 1.0807, + "step": 7628 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786243365928335e-05, + "loss": 1.1073, + "step": 7629 + }, + { + "epoch": 0.56, + "learning_rate": 1.978616132778462e-05, + "loss": 1.0971, + "step": 7630 + }, + { + "epoch": 0.56, + "learning_rate": 1.9786079274071265e-05, + "loss": 1.0438, + "step": 7631 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785997204788394e-05, + "loss": 1.098, + "step": 7632 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785915119936143e-05, + "loss": 1.0176, + "step": 7633 + }, + { + "epoch": 0.56, + "learning_rate": 1.978583301951464e-05, + "loss": 0.9531, + "step": 7634 + }, + { + "epoch": 0.56, + "learning_rate": 1.978575090352402e-05, + "loss": 0.9779, + "step": 7635 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785668771964407e-05, + "loss": 0.9787, + "step": 7636 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785586624835935e-05, + "loss": 0.9716, + "step": 7637 + }, + { + "epoch": 0.56, + "learning_rate": 1.978550446213874e-05, + "loss": 0.9743, + "step": 7638 + }, + { + "epoch": 0.56, + "learning_rate": 1.978542228387294e-05, + "loss": 1.0488, + "step": 7639 + }, + { + "epoch": 0.56, + "learning_rate": 1.978534009003868e-05, + "loss": 1.1321, + "step": 7640 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785257880636077e-05, + "loss": 1.0141, + "step": 7641 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785175655665273e-05, + "loss": 1.0488, + "step": 7642 + }, + { + "epoch": 0.56, + "learning_rate": 1.9785093415126394e-05, + "loss": 1.1157, + "step": 7643 + }, + { + "epoch": 0.56, + "learning_rate": 1.978501115901957e-05, + "loss": 1.087, + "step": 7644 + }, + { + "epoch": 0.56, + "learning_rate": 1.9784928887344933e-05, + "loss": 1.0174, + "step": 7645 + }, + { + "epoch": 0.56, + "learning_rate": 1.9784846600102616e-05, + "loss": 1.045, + "step": 7646 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784764297292748e-05, + "loss": 1.0042, + "step": 7647 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784681978915457e-05, + "loss": 1.0275, + "step": 7648 + }, + { + "epoch": 0.57, + "learning_rate": 1.978459964497088e-05, + "loss": 1.0338, + "step": 7649 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784517295459146e-05, + "loss": 1.0037, + "step": 7650 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784434930380386e-05, + "loss": 1.0422, + "step": 7651 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784352549734725e-05, + "loss": 0.9934, + "step": 7652 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784270153522303e-05, + "loss": 0.9922, + "step": 7653 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784187741743242e-05, + "loss": 1.0612, + "step": 7654 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784105314397684e-05, + "loss": 0.9356, + "step": 7655 + }, + { + "epoch": 0.57, + "learning_rate": 1.9784022871485754e-05, + "loss": 0.9907, + "step": 7656 + }, + { + "epoch": 0.57, + "learning_rate": 1.978394041300758e-05, + "loss": 1.0891, + "step": 7657 + }, + { + "epoch": 0.57, + "learning_rate": 1.97838579389633e-05, + "loss": 0.975, + "step": 7658 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783775449353042e-05, + "loss": 1.1884, + "step": 7659 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783692944176933e-05, + "loss": 1.0951, + "step": 7660 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783610423435113e-05, + "loss": 1.12, + "step": 7661 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783527887127708e-05, + "loss": 1.0091, + "step": 7662 + }, + { + "epoch": 0.57, + "learning_rate": 1.978344533525485e-05, + "loss": 0.8953, + "step": 7663 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783362767816668e-05, + "loss": 0.9873, + "step": 7664 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783280184813298e-05, + "loss": 1.0264, + "step": 7665 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783197586244868e-05, + "loss": 0.981, + "step": 7666 + }, + { + "epoch": 0.57, + "learning_rate": 1.9783114972111513e-05, + "loss": 1.0067, + "step": 7667 + }, + { + "epoch": 0.57, + "learning_rate": 1.978303234241336e-05, + "loss": 0.9834, + "step": 7668 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782949697150542e-05, + "loss": 1.0914, + "step": 7669 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782867036323195e-05, + "loss": 0.9306, + "step": 7670 + }, + { + "epoch": 0.57, + "learning_rate": 1.978278435993144e-05, + "loss": 1.1182, + "step": 7671 + }, + { + "epoch": 0.57, + "learning_rate": 1.978270166797542e-05, + "loss": 1.0946, + "step": 7672 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782618960455262e-05, + "loss": 1.0965, + "step": 7673 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782536237371094e-05, + "loss": 1.0033, + "step": 7674 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782453498723052e-05, + "loss": 0.9618, + "step": 7675 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782370744511268e-05, + "loss": 1.0537, + "step": 7676 + }, + { + "epoch": 0.57, + "learning_rate": 1.978228797473587e-05, + "loss": 1.017, + "step": 7677 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782205189396993e-05, + "loss": 1.0774, + "step": 7678 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782122388494767e-05, + "loss": 1.0158, + "step": 7679 + }, + { + "epoch": 0.57, + "learning_rate": 1.9782039572029323e-05, + "loss": 1.1224, + "step": 7680 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781956740000798e-05, + "loss": 0.9986, + "step": 7681 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781873892409316e-05, + "loss": 1.0292, + "step": 7682 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781791029255013e-05, + "loss": 0.9867, + "step": 7683 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781708150538022e-05, + "loss": 0.9855, + "step": 7684 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781625256258474e-05, + "loss": 0.9855, + "step": 7685 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781542346416497e-05, + "loss": 1.1064, + "step": 7686 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781459421012226e-05, + "loss": 0.9979, + "step": 7687 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781376480045797e-05, + "loss": 1.0239, + "step": 7688 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781293523517333e-05, + "loss": 0.9896, + "step": 7689 + }, + { + "epoch": 0.57, + "learning_rate": 1.9781210551426973e-05, + "loss": 1.1033, + "step": 7690 + }, + { + "epoch": 0.57, + "learning_rate": 1.978112756377485e-05, + "loss": 1.053, + "step": 7691 + }, + { + "epoch": 0.57, + "learning_rate": 1.978104456056109e-05, + "loss": 0.9735, + "step": 7692 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780961541785826e-05, + "loss": 1.03, + "step": 7693 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780878507449198e-05, + "loss": 0.9628, + "step": 7694 + }, + { + "epoch": 0.57, + "learning_rate": 1.978079545755133e-05, + "loss": 1.0487, + "step": 7695 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780712392092352e-05, + "loss": 1.1068, + "step": 7696 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780629311072407e-05, + "loss": 1.0658, + "step": 7697 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780546214491616e-05, + "loss": 1.0352, + "step": 7698 + }, + { + "epoch": 0.57, + "learning_rate": 1.978046310235012e-05, + "loss": 0.937, + "step": 7699 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780379974648043e-05, + "loss": 1.0295, + "step": 7700 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780296831385525e-05, + "loss": 1.0468, + "step": 7701 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780213672562693e-05, + "loss": 1.1415, + "step": 7702 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780130498179676e-05, + "loss": 1.0457, + "step": 7703 + }, + { + "epoch": 0.57, + "learning_rate": 1.9780047308236617e-05, + "loss": 1.0858, + "step": 7704 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779964102733643e-05, + "loss": 0.9864, + "step": 7705 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779880881670885e-05, + "loss": 1.0737, + "step": 7706 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779797645048476e-05, + "loss": 1.0654, + "step": 7707 + }, + { + "epoch": 0.57, + "learning_rate": 1.977971439286655e-05, + "loss": 1.0634, + "step": 7708 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779631125125233e-05, + "loss": 0.9901, + "step": 7709 + }, + { + "epoch": 0.57, + "learning_rate": 1.977954784182467e-05, + "loss": 1.1055, + "step": 7710 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779464542964983e-05, + "loss": 1.0177, + "step": 7711 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779381228546308e-05, + "loss": 0.961, + "step": 7712 + }, + { + "epoch": 0.57, + "learning_rate": 1.977929789856878e-05, + "loss": 1.1567, + "step": 7713 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779214553032524e-05, + "loss": 1.0669, + "step": 7714 + }, + { + "epoch": 0.57, + "learning_rate": 1.977913119193768e-05, + "loss": 1.109, + "step": 7715 + }, + { + "epoch": 0.57, + "learning_rate": 1.9779047815284377e-05, + "loss": 1.0087, + "step": 7716 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778964423072753e-05, + "loss": 0.9584, + "step": 7717 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778881015302934e-05, + "loss": 0.9706, + "step": 7718 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778797591975053e-05, + "loss": 0.9799, + "step": 7719 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778714153089248e-05, + "loss": 1.1904, + "step": 7720 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778630698645646e-05, + "loss": 1.0296, + "step": 7721 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778547228644383e-05, + "loss": 0.9293, + "step": 7722 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778463743085595e-05, + "loss": 1.096, + "step": 7723 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778380241969407e-05, + "loss": 1.0762, + "step": 7724 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778296725295958e-05, + "loss": 1.0582, + "step": 7725 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778213193065377e-05, + "loss": 1.0574, + "step": 7726 + }, + { + "epoch": 0.57, + "learning_rate": 1.9778129645277803e-05, + "loss": 1.065, + "step": 7727 + }, + { + "epoch": 0.57, + "learning_rate": 1.977804608193336e-05, + "loss": 0.9981, + "step": 7728 + }, + { + "epoch": 0.57, + "learning_rate": 1.977796250303219e-05, + "loss": 1.0678, + "step": 7729 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777878908574416e-05, + "loss": 1.0021, + "step": 7730 + }, + { + "epoch": 0.57, + "learning_rate": 1.977779529856018e-05, + "loss": 1.0844, + "step": 7731 + }, + { + "epoch": 0.57, + "learning_rate": 1.977771167298961e-05, + "loss": 1.0977, + "step": 7732 + }, + { + "epoch": 0.57, + "learning_rate": 1.977762803186284e-05, + "loss": 1.1538, + "step": 7733 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777544375180005e-05, + "loss": 1.0295, + "step": 7734 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777460702941237e-05, + "loss": 1.015, + "step": 7735 + }, + { + "epoch": 0.57, + "learning_rate": 1.977737701514667e-05, + "loss": 1.0748, + "step": 7736 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777293311796435e-05, + "loss": 1.0396, + "step": 7737 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777209592890668e-05, + "loss": 1.0641, + "step": 7738 + }, + { + "epoch": 0.57, + "learning_rate": 1.9777125858429497e-05, + "loss": 0.9732, + "step": 7739 + }, + { + "epoch": 0.57, + "learning_rate": 1.977704210841306e-05, + "loss": 0.9883, + "step": 7740 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776958342841487e-05, + "loss": 1.0064, + "step": 7741 + }, + { + "epoch": 0.57, + "learning_rate": 1.977687456171492e-05, + "loss": 1.0774, + "step": 7742 + }, + { + "epoch": 0.57, + "learning_rate": 1.977679076503348e-05, + "loss": 1.0214, + "step": 7743 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776706952797306e-05, + "loss": 0.9512, + "step": 7744 + }, + { + "epoch": 0.57, + "learning_rate": 1.977662312500653e-05, + "loss": 0.9425, + "step": 7745 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776539281661287e-05, + "loss": 1.0278, + "step": 7746 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776455422761713e-05, + "loss": 1.0508, + "step": 7747 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776371548307934e-05, + "loss": 1.0324, + "step": 7748 + }, + { + "epoch": 0.57, + "learning_rate": 1.977628765830009e-05, + "loss": 1.0778, + "step": 7749 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776203752738313e-05, + "loss": 0.9842, + "step": 7750 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776119831622735e-05, + "loss": 0.9195, + "step": 7751 + }, + { + "epoch": 0.57, + "learning_rate": 1.9776035894953488e-05, + "loss": 1.0492, + "step": 7752 + }, + { + "epoch": 0.57, + "learning_rate": 1.977595194273071e-05, + "loss": 1.0002, + "step": 7753 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775867974954534e-05, + "loss": 0.9433, + "step": 7754 + }, + { + "epoch": 0.57, + "learning_rate": 1.977578399162509e-05, + "loss": 1.092, + "step": 7755 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775699992742514e-05, + "loss": 0.9408, + "step": 7756 + }, + { + "epoch": 0.57, + "learning_rate": 1.977561597830694e-05, + "loss": 1.0078, + "step": 7757 + }, + { + "epoch": 0.57, + "learning_rate": 1.97755319483185e-05, + "loss": 1.0504, + "step": 7758 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775447902777326e-05, + "loss": 1.0064, + "step": 7759 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775363841683557e-05, + "loss": 1.0587, + "step": 7760 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775279765037325e-05, + "loss": 1.1087, + "step": 7761 + }, + { + "epoch": 0.57, + "learning_rate": 1.977519567283876e-05, + "loss": 1.0142, + "step": 7762 + }, + { + "epoch": 0.57, + "learning_rate": 1.9775111565088002e-05, + "loss": 1.1104, + "step": 7763 + }, + { + "epoch": 0.57, + "learning_rate": 1.977502744178518e-05, + "loss": 0.9987, + "step": 7764 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774943302930424e-05, + "loss": 0.9631, + "step": 7765 + }, + { + "epoch": 0.57, + "learning_rate": 1.977485914852388e-05, + "loss": 1.1067, + "step": 7766 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774774978565673e-05, + "loss": 1.0589, + "step": 7767 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774690793055934e-05, + "loss": 1.1155, + "step": 7768 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774606591994808e-05, + "loss": 1.0659, + "step": 7769 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774522375382422e-05, + "loss": 1.0474, + "step": 7770 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774438143218908e-05, + "loss": 1.0516, + "step": 7771 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774353895504404e-05, + "loss": 0.9921, + "step": 7772 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774269632239043e-05, + "loss": 1.0669, + "step": 7773 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774185353422963e-05, + "loss": 1.0018, + "step": 7774 + }, + { + "epoch": 0.57, + "learning_rate": 1.9774101059056286e-05, + "loss": 1.0688, + "step": 7775 + }, + { + "epoch": 0.57, + "learning_rate": 1.977401674913916e-05, + "loss": 1.0255, + "step": 7776 + }, + { + "epoch": 0.57, + "learning_rate": 1.9773932423671712e-05, + "loss": 1.1001, + "step": 7777 + }, + { + "epoch": 0.57, + "learning_rate": 1.9773848082654073e-05, + "loss": 1.078, + "step": 7778 + }, + { + "epoch": 0.57, + "learning_rate": 1.977376372608639e-05, + "loss": 1.1026, + "step": 7779 + }, + { + "epoch": 0.57, + "learning_rate": 1.977367935396878e-05, + "loss": 0.9609, + "step": 7780 + }, + { + "epoch": 0.57, + "learning_rate": 1.977359496630139e-05, + "loss": 1.0218, + "step": 7781 + }, + { + "epoch": 0.58, + "learning_rate": 1.977351056308435e-05, + "loss": 1.0184, + "step": 7782 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773426144317796e-05, + "loss": 1.0666, + "step": 7783 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773341710001856e-05, + "loss": 1.0565, + "step": 7784 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773257260136674e-05, + "loss": 1.0621, + "step": 7785 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773172794722377e-05, + "loss": 1.0765, + "step": 7786 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773088313759102e-05, + "loss": 0.9519, + "step": 7787 + }, + { + "epoch": 0.58, + "learning_rate": 1.9773003817246986e-05, + "loss": 1.0668, + "step": 7788 + }, + { + "epoch": 0.58, + "learning_rate": 1.977291930518616e-05, + "loss": 1.0135, + "step": 7789 + }, + { + "epoch": 0.58, + "learning_rate": 1.977283477757676e-05, + "loss": 1.0774, + "step": 7790 + }, + { + "epoch": 0.58, + "learning_rate": 1.9772750234418917e-05, + "loss": 0.9661, + "step": 7791 + }, + { + "epoch": 0.58, + "learning_rate": 1.977266567571277e-05, + "loss": 1.0365, + "step": 7792 + }, + { + "epoch": 0.58, + "learning_rate": 1.977258110145845e-05, + "loss": 0.9371, + "step": 7793 + }, + { + "epoch": 0.58, + "learning_rate": 1.9772496511656093e-05, + "loss": 1.0976, + "step": 7794 + }, + { + "epoch": 0.58, + "learning_rate": 1.9772411906305837e-05, + "loss": 1.0352, + "step": 7795 + }, + { + "epoch": 0.58, + "learning_rate": 1.977232728540781e-05, + "loss": 1.0099, + "step": 7796 + }, + { + "epoch": 0.58, + "learning_rate": 1.977224264896215e-05, + "loss": 1.1013, + "step": 7797 + }, + { + "epoch": 0.58, + "learning_rate": 1.9772157996968996e-05, + "loss": 0.9549, + "step": 7798 + }, + { + "epoch": 0.58, + "learning_rate": 1.9772073329428477e-05, + "loss": 0.9404, + "step": 7799 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771988646340725e-05, + "loss": 1.1153, + "step": 7800 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771903947705884e-05, + "loss": 0.9909, + "step": 7801 + }, + { + "epoch": 0.58, + "learning_rate": 1.977181923352408e-05, + "loss": 1.0906, + "step": 7802 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771734503795453e-05, + "loss": 0.9929, + "step": 7803 + }, + { + "epoch": 0.58, + "learning_rate": 1.977164975852014e-05, + "loss": 1.0387, + "step": 7804 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771564997698266e-05, + "loss": 1.0656, + "step": 7805 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771480221329974e-05, + "loss": 1.0276, + "step": 7806 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771395429415396e-05, + "loss": 1.0719, + "step": 7807 + }, + { + "epoch": 0.58, + "learning_rate": 1.977131062195467e-05, + "loss": 1.0581, + "step": 7808 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771225798947927e-05, + "loss": 0.9766, + "step": 7809 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771140960395304e-05, + "loss": 0.9734, + "step": 7810 + }, + { + "epoch": 0.58, + "learning_rate": 1.9771056106296937e-05, + "loss": 1.0684, + "step": 7811 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770971236652957e-05, + "loss": 1.0419, + "step": 7812 + }, + { + "epoch": 0.58, + "learning_rate": 1.97708863514635e-05, + "loss": 1.089, + "step": 7813 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770801450728708e-05, + "loss": 1.0256, + "step": 7814 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770716534448703e-05, + "loss": 1.1106, + "step": 7815 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770631602623635e-05, + "loss": 1.0437, + "step": 7816 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770546655253628e-05, + "loss": 1.0368, + "step": 7817 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770461692338822e-05, + "loss": 1.0988, + "step": 7818 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770376713879355e-05, + "loss": 1.1243, + "step": 7819 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770291719875353e-05, + "loss": 1.094, + "step": 7820 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770206710326958e-05, + "loss": 1.0522, + "step": 7821 + }, + { + "epoch": 0.58, + "learning_rate": 1.9770121685234306e-05, + "loss": 1.0915, + "step": 7822 + }, + { + "epoch": 0.58, + "learning_rate": 1.977003664459753e-05, + "loss": 1.0436, + "step": 7823 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769951588416765e-05, + "loss": 1.0008, + "step": 7824 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769866516692143e-05, + "loss": 1.0334, + "step": 7825 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769781429423806e-05, + "loss": 1.0988, + "step": 7826 + }, + { + "epoch": 0.58, + "learning_rate": 1.976969632661189e-05, + "loss": 1.0265, + "step": 7827 + }, + { + "epoch": 0.58, + "learning_rate": 1.976961120825652e-05, + "loss": 0.991, + "step": 7828 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769526074357843e-05, + "loss": 1.072, + "step": 7829 + }, + { + "epoch": 0.58, + "learning_rate": 1.976944092491599e-05, + "loss": 1.02, + "step": 7830 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769355759931093e-05, + "loss": 1.0672, + "step": 7831 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769270579403294e-05, + "loss": 0.9369, + "step": 7832 + }, + { + "epoch": 0.58, + "learning_rate": 1.9769185383332725e-05, + "loss": 1.0837, + "step": 7833 + }, + { + "epoch": 0.58, + "learning_rate": 1.976910017171952e-05, + "loss": 0.9714, + "step": 7834 + }, + { + "epoch": 0.58, + "learning_rate": 1.976901494456382e-05, + "loss": 0.9958, + "step": 7835 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768929701865753e-05, + "loss": 1.0496, + "step": 7836 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768844443625463e-05, + "loss": 0.9597, + "step": 7837 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768759169843076e-05, + "loss": 0.9596, + "step": 7838 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768673880518733e-05, + "loss": 1.064, + "step": 7839 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768588575652572e-05, + "loss": 0.9737, + "step": 7840 + }, + { + "epoch": 0.58, + "learning_rate": 1.976850325524473e-05, + "loss": 1.1343, + "step": 7841 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768417919295336e-05, + "loss": 1.0914, + "step": 7842 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768332567804527e-05, + "loss": 0.9503, + "step": 7843 + }, + { + "epoch": 0.58, + "learning_rate": 1.976824720077244e-05, + "loss": 1.0153, + "step": 7844 + }, + { + "epoch": 0.58, + "learning_rate": 1.9768161818199215e-05, + "loss": 1.1215, + "step": 7845 + }, + { + "epoch": 0.58, + "learning_rate": 1.976807642008498e-05, + "loss": 1.0231, + "step": 7846 + }, + { + "epoch": 0.58, + "learning_rate": 1.976799100642988e-05, + "loss": 1.1308, + "step": 7847 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767905577234042e-05, + "loss": 1.0399, + "step": 7848 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767820132497612e-05, + "loss": 1.0475, + "step": 7849 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767734672220712e-05, + "loss": 0.9785, + "step": 7850 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767649196403492e-05, + "loss": 1.0838, + "step": 7851 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767563705046077e-05, + "loss": 1.0561, + "step": 7852 + }, + { + "epoch": 0.58, + "learning_rate": 1.976747819814861e-05, + "loss": 0.9418, + "step": 7853 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767392675711224e-05, + "loss": 1.078, + "step": 7854 + }, + { + "epoch": 0.58, + "learning_rate": 1.976730713773406e-05, + "loss": 1.0173, + "step": 7855 + }, + { + "epoch": 0.58, + "learning_rate": 1.976722158421725e-05, + "loss": 1.0645, + "step": 7856 + }, + { + "epoch": 0.58, + "learning_rate": 1.9767136015160925e-05, + "loss": 1.0856, + "step": 7857 + }, + { + "epoch": 0.58, + "learning_rate": 1.976705043056523e-05, + "loss": 0.9875, + "step": 7858 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766964830430296e-05, + "loss": 1.0113, + "step": 7859 + }, + { + "epoch": 0.58, + "learning_rate": 1.976687921475626e-05, + "loss": 1.0687, + "step": 7860 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766793583543264e-05, + "loss": 1.0387, + "step": 7861 + }, + { + "epoch": 0.58, + "learning_rate": 1.976670793679143e-05, + "loss": 1.0805, + "step": 7862 + }, + { + "epoch": 0.58, + "learning_rate": 1.976662227450091e-05, + "loss": 1.0325, + "step": 7863 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766536596671834e-05, + "loss": 0.9557, + "step": 7864 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766450903304335e-05, + "loss": 0.9331, + "step": 7865 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766365194398553e-05, + "loss": 0.9566, + "step": 7866 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766279469954625e-05, + "loss": 1.0164, + "step": 7867 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766193729972687e-05, + "loss": 1.0378, + "step": 7868 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766107974452872e-05, + "loss": 1.012, + "step": 7869 + }, + { + "epoch": 0.58, + "learning_rate": 1.9766022203395318e-05, + "loss": 0.9997, + "step": 7870 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765936416800168e-05, + "loss": 1.0618, + "step": 7871 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765850614667547e-05, + "loss": 1.1375, + "step": 7872 + }, + { + "epoch": 0.58, + "learning_rate": 1.97657647969976e-05, + "loss": 1.0315, + "step": 7873 + }, + { + "epoch": 0.58, + "learning_rate": 1.976567896379046e-05, + "loss": 1.0589, + "step": 7874 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765593115046268e-05, + "loss": 1.07, + "step": 7875 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765507250765153e-05, + "loss": 1.1009, + "step": 7876 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765421370947257e-05, + "loss": 1.0833, + "step": 7877 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765335475592714e-05, + "loss": 1.1071, + "step": 7878 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765249564701662e-05, + "loss": 1.0269, + "step": 7879 + }, + { + "epoch": 0.58, + "learning_rate": 1.976516363827424e-05, + "loss": 0.9706, + "step": 7880 + }, + { + "epoch": 0.58, + "learning_rate": 1.9765077696310582e-05, + "loss": 1.0436, + "step": 7881 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764991738810824e-05, + "loss": 1.0927, + "step": 7882 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764905765775103e-05, + "loss": 1.06, + "step": 7883 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764819777203557e-05, + "loss": 1.1612, + "step": 7884 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764733773096327e-05, + "loss": 1.075, + "step": 7885 + }, + { + "epoch": 0.58, + "learning_rate": 1.976464775345354e-05, + "loss": 1.1011, + "step": 7886 + }, + { + "epoch": 0.58, + "learning_rate": 1.976456171827534e-05, + "loss": 1.088, + "step": 7887 + }, + { + "epoch": 0.58, + "learning_rate": 1.976447566756186e-05, + "loss": 0.9827, + "step": 7888 + }, + { + "epoch": 0.58, + "learning_rate": 1.976438960131324e-05, + "loss": 1.0562, + "step": 7889 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764303519529618e-05, + "loss": 1.0797, + "step": 7890 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764217422211128e-05, + "loss": 0.9673, + "step": 7891 + }, + { + "epoch": 0.58, + "learning_rate": 1.9764131309357907e-05, + "loss": 0.8983, + "step": 7892 + }, + { + "epoch": 0.58, + "learning_rate": 1.976404518097009e-05, + "loss": 1.0106, + "step": 7893 + }, + { + "epoch": 0.58, + "learning_rate": 1.976395903704782e-05, + "loss": 1.1219, + "step": 7894 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763872877591232e-05, + "loss": 1.0491, + "step": 7895 + }, + { + "epoch": 0.58, + "learning_rate": 1.976378670260046e-05, + "loss": 1.0036, + "step": 7896 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763700512075644e-05, + "loss": 0.9589, + "step": 7897 + }, + { + "epoch": 0.58, + "learning_rate": 1.976361430601692e-05, + "loss": 0.9777, + "step": 7898 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763528084424422e-05, + "loss": 0.9886, + "step": 7899 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763441847298296e-05, + "loss": 1.0233, + "step": 7900 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763355594638672e-05, + "loss": 1.0111, + "step": 7901 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763269326445688e-05, + "loss": 1.0378, + "step": 7902 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763183042719484e-05, + "loss": 0.9872, + "step": 7903 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763096743460194e-05, + "loss": 1.0048, + "step": 7904 + }, + { + "epoch": 0.58, + "learning_rate": 1.9763010428667956e-05, + "loss": 1.0309, + "step": 7905 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762924098342906e-05, + "loss": 0.9181, + "step": 7906 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762837752485186e-05, + "loss": 1.0564, + "step": 7907 + }, + { + "epoch": 0.58, + "learning_rate": 1.976275139109493e-05, + "loss": 1.1125, + "step": 7908 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762665014172278e-05, + "loss": 1.0588, + "step": 7909 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762578621717365e-05, + "loss": 1.0627, + "step": 7910 + }, + { + "epoch": 0.58, + "learning_rate": 1.976249221373033e-05, + "loss": 0.9304, + "step": 7911 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762405790211307e-05, + "loss": 0.989, + "step": 7912 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762319351160437e-05, + "loss": 1.0246, + "step": 7913 + }, + { + "epoch": 0.58, + "learning_rate": 1.976223289657786e-05, + "loss": 1.0936, + "step": 7914 + }, + { + "epoch": 0.58, + "learning_rate": 1.9762146426463705e-05, + "loss": 1.0802, + "step": 7915 + }, + { + "epoch": 0.58, + "learning_rate": 1.976205994081812e-05, + "loss": 0.9669, + "step": 7916 + }, + { + "epoch": 0.58, + "learning_rate": 1.976197343964123e-05, + "loss": 0.9791, + "step": 7917 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761886922933187e-05, + "loss": 1.0782, + "step": 7918 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761800390694117e-05, + "loss": 1.0503, + "step": 7919 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761713842924166e-05, + "loss": 1.06, + "step": 7920 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761627279623466e-05, + "loss": 1.0315, + "step": 7921 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761540700792157e-05, + "loss": 1.1109, + "step": 7922 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761454106430372e-05, + "loss": 0.9914, + "step": 7923 + }, + { + "epoch": 0.59, + "learning_rate": 1.976136749653826e-05, + "loss": 1.0197, + "step": 7924 + }, + { + "epoch": 0.59, + "learning_rate": 1.976128087111595e-05, + "loss": 0.975, + "step": 7925 + }, + { + "epoch": 0.59, + "learning_rate": 1.976119423016358e-05, + "loss": 1.0862, + "step": 7926 + }, + { + "epoch": 0.59, + "learning_rate": 1.9761107573681292e-05, + "loss": 1.0554, + "step": 7927 + }, + { + "epoch": 0.59, + "learning_rate": 1.976102090166922e-05, + "loss": 1.0087, + "step": 7928 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760934214127502e-05, + "loss": 0.9904, + "step": 7929 + }, + { + "epoch": 0.59, + "learning_rate": 1.976084751105628e-05, + "loss": 1.0281, + "step": 7930 + }, + { + "epoch": 0.59, + "learning_rate": 1.976076079245569e-05, + "loss": 0.8967, + "step": 7931 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760674058325863e-05, + "loss": 1.0441, + "step": 7932 + }, + { + "epoch": 0.59, + "learning_rate": 1.976058730866695e-05, + "loss": 0.9581, + "step": 7933 + }, + { + "epoch": 0.59, + "learning_rate": 1.976050054347908e-05, + "loss": 0.9541, + "step": 7934 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760413762762392e-05, + "loss": 1.0326, + "step": 7935 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760326966517027e-05, + "loss": 1.0708, + "step": 7936 + }, + { + "epoch": 0.59, + "learning_rate": 1.976024015474312e-05, + "loss": 0.9501, + "step": 7937 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760153327440812e-05, + "loss": 1.0356, + "step": 7938 + }, + { + "epoch": 0.59, + "learning_rate": 1.9760066484610237e-05, + "loss": 1.0836, + "step": 7939 + }, + { + "epoch": 0.59, + "learning_rate": 1.975997962625154e-05, + "loss": 0.9721, + "step": 7940 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759892752364854e-05, + "loss": 0.9006, + "step": 7941 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759805862950316e-05, + "loss": 0.9964, + "step": 7942 + }, + { + "epoch": 0.59, + "learning_rate": 1.975971895800807e-05, + "loss": 1.0128, + "step": 7943 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759632037538248e-05, + "loss": 1.063, + "step": 7944 + }, + { + "epoch": 0.59, + "learning_rate": 1.975954510154099e-05, + "loss": 1.0272, + "step": 7945 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759458150016438e-05, + "loss": 0.9846, + "step": 7946 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759371182964727e-05, + "loss": 1.0808, + "step": 7947 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759284200385995e-05, + "loss": 1.0817, + "step": 7948 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759197202280383e-05, + "loss": 0.879, + "step": 7949 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759110188648026e-05, + "loss": 1.024, + "step": 7950 + }, + { + "epoch": 0.59, + "learning_rate": 1.9759023159489068e-05, + "loss": 0.9528, + "step": 7951 + }, + { + "epoch": 0.59, + "learning_rate": 1.975893611480364e-05, + "loss": 1.0456, + "step": 7952 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758849054591885e-05, + "loss": 1.1062, + "step": 7953 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758761978853942e-05, + "loss": 1.0875, + "step": 7954 + }, + { + "epoch": 0.59, + "learning_rate": 1.975867488758995e-05, + "loss": 1.101, + "step": 7955 + }, + { + "epoch": 0.59, + "learning_rate": 1.975858778080004e-05, + "loss": 1.0519, + "step": 7956 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758500658484363e-05, + "loss": 1.0177, + "step": 7957 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758413520643044e-05, + "loss": 1.117, + "step": 7958 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758326367276235e-05, + "loss": 0.9364, + "step": 7959 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758239198384068e-05, + "loss": 1.1152, + "step": 7960 + }, + { + "epoch": 0.59, + "learning_rate": 1.9758152013966677e-05, + "loss": 1.0218, + "step": 7961 + }, + { + "epoch": 0.59, + "learning_rate": 1.975806481402421e-05, + "loss": 0.9928, + "step": 7962 + }, + { + "epoch": 0.59, + "learning_rate": 1.97579775985568e-05, + "loss": 1.0989, + "step": 7963 + }, + { + "epoch": 0.59, + "learning_rate": 1.975789036756459e-05, + "loss": 1.0771, + "step": 7964 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757803121047714e-05, + "loss": 0.924, + "step": 7965 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757715859006314e-05, + "loss": 1.043, + "step": 7966 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757628581440524e-05, + "loss": 1.0134, + "step": 7967 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757541288350487e-05, + "loss": 0.8919, + "step": 7968 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757453979736345e-05, + "loss": 1.0279, + "step": 7969 + }, + { + "epoch": 0.59, + "learning_rate": 1.975736665559823e-05, + "loss": 1.1106, + "step": 7970 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757279315936287e-05, + "loss": 0.9851, + "step": 7971 + }, + { + "epoch": 0.59, + "learning_rate": 1.975719196075065e-05, + "loss": 1.0765, + "step": 7972 + }, + { + "epoch": 0.59, + "learning_rate": 1.9757104590041462e-05, + "loss": 1.0447, + "step": 7973 + }, + { + "epoch": 0.59, + "learning_rate": 1.975701720380886e-05, + "loss": 0.9416, + "step": 7974 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756929802052984e-05, + "loss": 1.031, + "step": 7975 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756842384773968e-05, + "loss": 0.9566, + "step": 7976 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756754951971962e-05, + "loss": 0.935, + "step": 7977 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756667503647095e-05, + "loss": 0.9324, + "step": 7978 + }, + { + "epoch": 0.59, + "learning_rate": 1.975658003979951e-05, + "loss": 0.8903, + "step": 7979 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756492560429342e-05, + "loss": 1.1859, + "step": 7980 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756405065536738e-05, + "loss": 1.1213, + "step": 7981 + }, + { + "epoch": 0.59, + "learning_rate": 1.975631755512183e-05, + "loss": 1.1079, + "step": 7982 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756230029184765e-05, + "loss": 1.1098, + "step": 7983 + }, + { + "epoch": 0.59, + "learning_rate": 1.9756142487725674e-05, + "loss": 1.1005, + "step": 7984 + }, + { + "epoch": 0.59, + "learning_rate": 1.97560549307447e-05, + "loss": 1.1318, + "step": 7985 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755967358241985e-05, + "loss": 1.0503, + "step": 7986 + }, + { + "epoch": 0.59, + "learning_rate": 1.975587977021766e-05, + "loss": 0.9854, + "step": 7987 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755792166671875e-05, + "loss": 1.0661, + "step": 7988 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755704547604766e-05, + "loss": 1.0016, + "step": 7989 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755616913016466e-05, + "loss": 0.9509, + "step": 7990 + }, + { + "epoch": 0.59, + "learning_rate": 1.975552926290712e-05, + "loss": 0.9877, + "step": 7991 + }, + { + "epoch": 0.59, + "learning_rate": 1.975544159727687e-05, + "loss": 1.0804, + "step": 7992 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755353916125847e-05, + "loss": 1.0709, + "step": 7993 + }, + { + "epoch": 0.59, + "learning_rate": 1.97552662194542e-05, + "loss": 1.0532, + "step": 7994 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755178507262062e-05, + "loss": 0.9681, + "step": 7995 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755090779549574e-05, + "loss": 0.9972, + "step": 7996 + }, + { + "epoch": 0.59, + "learning_rate": 1.9755003036316874e-05, + "loss": 1.0073, + "step": 7997 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754915277564105e-05, + "loss": 1.1374, + "step": 7998 + }, + { + "epoch": 0.59, + "learning_rate": 1.975482750329141e-05, + "loss": 1.1344, + "step": 7999 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754739713498917e-05, + "loss": 1.0736, + "step": 8000 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754651908186776e-05, + "loss": 0.9682, + "step": 8001 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754564087355125e-05, + "loss": 0.9647, + "step": 8002 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754476251004102e-05, + "loss": 1.1398, + "step": 8003 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754388399133843e-05, + "loss": 1.1123, + "step": 8004 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754300531744497e-05, + "loss": 1.0895, + "step": 8005 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754212648836197e-05, + "loss": 1.0201, + "step": 8006 + }, + { + "epoch": 0.59, + "learning_rate": 1.975412475040908e-05, + "loss": 1.1265, + "step": 8007 + }, + { + "epoch": 0.59, + "learning_rate": 1.9754036836463293e-05, + "loss": 0.9783, + "step": 8008 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753948906998972e-05, + "loss": 1.0777, + "step": 8009 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753860962016258e-05, + "loss": 1.0743, + "step": 8010 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753773001515294e-05, + "loss": 1.0204, + "step": 8011 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753685025496213e-05, + "loss": 1.0723, + "step": 8012 + }, + { + "epoch": 0.59, + "learning_rate": 1.975359703395916e-05, + "loss": 1.06, + "step": 8013 + }, + { + "epoch": 0.59, + "learning_rate": 1.975350902690427e-05, + "loss": 0.9938, + "step": 8014 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753421004331693e-05, + "loss": 1.0038, + "step": 8015 + }, + { + "epoch": 0.59, + "learning_rate": 1.975333296624156e-05, + "loss": 0.9966, + "step": 8016 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753244912634012e-05, + "loss": 1.0407, + "step": 8017 + }, + { + "epoch": 0.59, + "learning_rate": 1.9753156843509194e-05, + "loss": 1.096, + "step": 8018 + }, + { + "epoch": 0.59, + "learning_rate": 1.975306875886724e-05, + "loss": 1.1127, + "step": 8019 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752980658708298e-05, + "loss": 1.0059, + "step": 8020 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752892543032497e-05, + "loss": 1.0743, + "step": 8021 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752804411839988e-05, + "loss": 1.1659, + "step": 8022 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752716265130906e-05, + "loss": 0.9919, + "step": 8023 + }, + { + "epoch": 0.59, + "learning_rate": 1.975262810290539e-05, + "loss": 1.055, + "step": 8024 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752539925163586e-05, + "loss": 1.0849, + "step": 8025 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752451731905625e-05, + "loss": 0.9152, + "step": 8026 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752363523131658e-05, + "loss": 1.0634, + "step": 8027 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752275298841815e-05, + "loss": 1.0632, + "step": 8028 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752187059036246e-05, + "loss": 0.9826, + "step": 8029 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752098803715087e-05, + "loss": 1.106, + "step": 8030 + }, + { + "epoch": 0.59, + "learning_rate": 1.9752010532878474e-05, + "loss": 1.0474, + "step": 8031 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751922246526557e-05, + "loss": 1.1276, + "step": 8032 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751833944659467e-05, + "loss": 1.0133, + "step": 8033 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751745627277353e-05, + "loss": 1.1264, + "step": 8034 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751657294380346e-05, + "loss": 1.015, + "step": 8035 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751568945968596e-05, + "loss": 1.0067, + "step": 8036 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751480582042238e-05, + "loss": 1.0057, + "step": 8037 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751392202601414e-05, + "loss": 1.0894, + "step": 8038 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751303807646264e-05, + "loss": 0.9425, + "step": 8039 + }, + { + "epoch": 0.59, + "learning_rate": 1.975121539717693e-05, + "loss": 1.1795, + "step": 8040 + }, + { + "epoch": 0.59, + "learning_rate": 1.975112697119355e-05, + "loss": 1.0023, + "step": 8041 + }, + { + "epoch": 0.59, + "learning_rate": 1.9751038529696267e-05, + "loss": 1.1153, + "step": 8042 + }, + { + "epoch": 0.59, + "learning_rate": 1.9750950072685223e-05, + "loss": 1.1368, + "step": 8043 + }, + { + "epoch": 0.59, + "learning_rate": 1.9750861600160557e-05, + "loss": 1.0442, + "step": 8044 + }, + { + "epoch": 0.59, + "learning_rate": 1.975077311212241e-05, + "loss": 0.9659, + "step": 8045 + }, + { + "epoch": 0.59, + "learning_rate": 1.975068460857092e-05, + "loss": 1.1047, + "step": 8046 + }, + { + "epoch": 0.59, + "learning_rate": 1.975059608950623e-05, + "loss": 1.0916, + "step": 8047 + }, + { + "epoch": 0.59, + "learning_rate": 1.975050755492848e-05, + "loss": 1.0666, + "step": 8048 + }, + { + "epoch": 0.59, + "learning_rate": 1.9750419004837817e-05, + "loss": 0.8683, + "step": 8049 + }, + { + "epoch": 0.59, + "learning_rate": 1.9750330439234374e-05, + "loss": 1.0908, + "step": 8050 + }, + { + "epoch": 0.59, + "learning_rate": 1.9750241858118296e-05, + "loss": 1.0877, + "step": 8051 + }, + { + "epoch": 0.59, + "learning_rate": 1.975015326148972e-05, + "loss": 0.9627, + "step": 8052 + }, + { + "epoch": 0.6, + "learning_rate": 1.9750064649348794e-05, + "loss": 1.0781, + "step": 8053 + }, + { + "epoch": 0.6, + "learning_rate": 1.974997602169565e-05, + "loss": 0.9866, + "step": 8054 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749887378530435e-05, + "loss": 1.0984, + "step": 8055 + }, + { + "epoch": 0.6, + "learning_rate": 1.974979871985329e-05, + "loss": 1.0614, + "step": 8056 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749710045664355e-05, + "loss": 1.1035, + "step": 8057 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749621355963772e-05, + "loss": 1.0939, + "step": 8058 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749532650751678e-05, + "loss": 0.9961, + "step": 8059 + }, + { + "epoch": 0.6, + "learning_rate": 1.974944393002822e-05, + "loss": 1.0899, + "step": 8060 + }, + { + "epoch": 0.6, + "learning_rate": 1.974935519379353e-05, + "loss": 1.0434, + "step": 8061 + }, + { + "epoch": 0.6, + "learning_rate": 1.974926644204776e-05, + "loss": 0.8668, + "step": 8062 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749177674791048e-05, + "loss": 1.0867, + "step": 8063 + }, + { + "epoch": 0.6, + "learning_rate": 1.974908889202353e-05, + "loss": 0.8648, + "step": 8064 + }, + { + "epoch": 0.6, + "learning_rate": 1.9749000093745355e-05, + "loss": 0.9289, + "step": 8065 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748911279956658e-05, + "loss": 1.0541, + "step": 8066 + }, + { + "epoch": 0.6, + "learning_rate": 1.974882245065758e-05, + "loss": 0.9849, + "step": 8067 + }, + { + "epoch": 0.6, + "learning_rate": 1.974873360584827e-05, + "loss": 0.9845, + "step": 8068 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748644745528862e-05, + "loss": 1.0294, + "step": 8069 + }, + { + "epoch": 0.6, + "learning_rate": 1.97485558696995e-05, + "loss": 1.1012, + "step": 8070 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748466978360326e-05, + "loss": 1.0981, + "step": 8071 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748378071511477e-05, + "loss": 1.0818, + "step": 8072 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748289149153105e-05, + "loss": 1.0435, + "step": 8073 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748200211285337e-05, + "loss": 1.0036, + "step": 8074 + }, + { + "epoch": 0.6, + "learning_rate": 1.9748111257908324e-05, + "loss": 1.0619, + "step": 8075 + }, + { + "epoch": 0.6, + "learning_rate": 1.974802228902221e-05, + "loss": 0.993, + "step": 8076 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747933304627124e-05, + "loss": 1.1433, + "step": 8077 + }, + { + "epoch": 0.6, + "learning_rate": 1.974784430472322e-05, + "loss": 1.0359, + "step": 8078 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747755289310637e-05, + "loss": 0.9713, + "step": 8079 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747666258389512e-05, + "loss": 1.0762, + "step": 8080 + }, + { + "epoch": 0.6, + "learning_rate": 1.974757721195999e-05, + "loss": 0.9412, + "step": 8081 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747488150022214e-05, + "loss": 1.0844, + "step": 8082 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747399072576325e-05, + "loss": 1.0335, + "step": 8083 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747309979622458e-05, + "loss": 1.1152, + "step": 8084 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747220871160763e-05, + "loss": 0.9176, + "step": 8085 + }, + { + "epoch": 0.6, + "learning_rate": 1.974713174719138e-05, + "loss": 0.9845, + "step": 8086 + }, + { + "epoch": 0.6, + "learning_rate": 1.9747042607714448e-05, + "loss": 1.0363, + "step": 8087 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746953452730112e-05, + "loss": 0.9091, + "step": 8088 + }, + { + "epoch": 0.6, + "learning_rate": 1.974686428223851e-05, + "loss": 1.0708, + "step": 8089 + }, + { + "epoch": 0.6, + "learning_rate": 1.974677509623979e-05, + "loss": 0.9616, + "step": 8090 + }, + { + "epoch": 0.6, + "learning_rate": 1.974668589473409e-05, + "loss": 0.9979, + "step": 8091 + }, + { + "epoch": 0.6, + "learning_rate": 1.974659667772155e-05, + "loss": 1.0268, + "step": 8092 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746507445202313e-05, + "loss": 1.0733, + "step": 8093 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746418197176528e-05, + "loss": 0.9626, + "step": 8094 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746328933644327e-05, + "loss": 0.9608, + "step": 8095 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746239654605856e-05, + "loss": 1.099, + "step": 8096 + }, + { + "epoch": 0.6, + "learning_rate": 1.974615036006126e-05, + "loss": 1.0053, + "step": 8097 + }, + { + "epoch": 0.6, + "learning_rate": 1.9746061050010672e-05, + "loss": 1.0871, + "step": 8098 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745971724454243e-05, + "loss": 1.0727, + "step": 8099 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745882383392117e-05, + "loss": 1.0566, + "step": 8100 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745793026824426e-05, + "loss": 1.003, + "step": 8101 + }, + { + "epoch": 0.6, + "learning_rate": 1.974570365475132e-05, + "loss": 1.0064, + "step": 8102 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745614267172943e-05, + "loss": 0.8744, + "step": 8103 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745524864089428e-05, + "loss": 1.0393, + "step": 8104 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745435445500924e-05, + "loss": 0.9708, + "step": 8105 + }, + { + "epoch": 0.6, + "learning_rate": 1.974534601140757e-05, + "loss": 1.0781, + "step": 8106 + }, + { + "epoch": 0.6, + "learning_rate": 1.974525656180951e-05, + "loss": 0.983, + "step": 8107 + }, + { + "epoch": 0.6, + "learning_rate": 1.9745167096706885e-05, + "loss": 1.0313, + "step": 8108 + }, + { + "epoch": 0.6, + "learning_rate": 1.974507761609984e-05, + "loss": 1.1438, + "step": 8109 + }, + { + "epoch": 0.6, + "learning_rate": 1.974498811998852e-05, + "loss": 0.9437, + "step": 8110 + }, + { + "epoch": 0.6, + "learning_rate": 1.974489860837306e-05, + "loss": 1.0344, + "step": 8111 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744809081253606e-05, + "loss": 0.9587, + "step": 8112 + }, + { + "epoch": 0.6, + "learning_rate": 1.97447195386303e-05, + "loss": 1.0108, + "step": 8113 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744629980503283e-05, + "loss": 0.9715, + "step": 8114 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744540406872703e-05, + "loss": 1.1094, + "step": 8115 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744450817738693e-05, + "loss": 1.0866, + "step": 8116 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744361213101408e-05, + "loss": 1.0346, + "step": 8117 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744271592960977e-05, + "loss": 1.0858, + "step": 8118 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744181957317553e-05, + "loss": 0.9301, + "step": 8119 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744092306171275e-05, + "loss": 1.1476, + "step": 8120 + }, + { + "epoch": 0.6, + "learning_rate": 1.9744002639522287e-05, + "loss": 1.0908, + "step": 8121 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743912957370727e-05, + "loss": 1.1678, + "step": 8122 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743823259716743e-05, + "loss": 0.9792, + "step": 8123 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743733546560473e-05, + "loss": 1.1365, + "step": 8124 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743643817902064e-05, + "loss": 1.0278, + "step": 8125 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743554073741658e-05, + "loss": 0.9592, + "step": 8126 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743464314079393e-05, + "loss": 1.0613, + "step": 8127 + }, + { + "epoch": 0.6, + "learning_rate": 1.974337453891542e-05, + "loss": 1.025, + "step": 8128 + }, + { + "epoch": 0.6, + "learning_rate": 1.974328474824988e-05, + "loss": 1.1325, + "step": 8129 + }, + { + "epoch": 0.6, + "learning_rate": 1.974319494208291e-05, + "loss": 0.9677, + "step": 8130 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743105120414656e-05, + "loss": 1.1441, + "step": 8131 + }, + { + "epoch": 0.6, + "learning_rate": 1.9743015283245258e-05, + "loss": 1.1134, + "step": 8132 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742925430574866e-05, + "loss": 0.9719, + "step": 8133 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742835562403618e-05, + "loss": 0.9872, + "step": 8134 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742745678731658e-05, + "loss": 0.9994, + "step": 8135 + }, + { + "epoch": 0.6, + "learning_rate": 1.974265577955913e-05, + "loss": 0.9488, + "step": 8136 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742565864886173e-05, + "loss": 0.9918, + "step": 8137 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742475934712937e-05, + "loss": 1.054, + "step": 8138 + }, + { + "epoch": 0.6, + "learning_rate": 1.974238598903956e-05, + "loss": 1.1019, + "step": 8139 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742296027866186e-05, + "loss": 1.0011, + "step": 8140 + }, + { + "epoch": 0.6, + "learning_rate": 1.974220605119296e-05, + "loss": 1.0878, + "step": 8141 + }, + { + "epoch": 0.6, + "learning_rate": 1.974211605902002e-05, + "loss": 0.9782, + "step": 8142 + }, + { + "epoch": 0.6, + "learning_rate": 1.9742026051347513e-05, + "loss": 1.0851, + "step": 8143 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741936028175582e-05, + "loss": 1.0387, + "step": 8144 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741845989504375e-05, + "loss": 1.1397, + "step": 8145 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741755935334026e-05, + "loss": 1.0266, + "step": 8146 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741665865664682e-05, + "loss": 1.2145, + "step": 8147 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741575780496492e-05, + "loss": 1.0444, + "step": 8148 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741485679829592e-05, + "loss": 1.0801, + "step": 8149 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741395563664123e-05, + "loss": 0.9913, + "step": 8150 + }, + { + "epoch": 0.6, + "learning_rate": 1.974130543200024e-05, + "loss": 1.1013, + "step": 8151 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741215284838072e-05, + "loss": 0.9163, + "step": 8152 + }, + { + "epoch": 0.6, + "learning_rate": 1.9741125122177775e-05, + "loss": 1.0288, + "step": 8153 + }, + { + "epoch": 0.6, + "learning_rate": 1.974103494401949e-05, + "loss": 1.0379, + "step": 8154 + }, + { + "epoch": 0.6, + "learning_rate": 1.974094475036335e-05, + "loss": 1.0642, + "step": 8155 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740854541209513e-05, + "loss": 1.0453, + "step": 8156 + }, + { + "epoch": 0.6, + "learning_rate": 1.974076431655811e-05, + "loss": 0.997, + "step": 8157 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740674076409295e-05, + "loss": 0.9843, + "step": 8158 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740583820763205e-05, + "loss": 1.0409, + "step": 8159 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740493549619987e-05, + "loss": 1.0586, + "step": 8160 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740403262979783e-05, + "loss": 1.1607, + "step": 8161 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740312960842734e-05, + "loss": 1.0657, + "step": 8162 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740222643208988e-05, + "loss": 0.96, + "step": 8163 + }, + { + "epoch": 0.6, + "learning_rate": 1.974013231007869e-05, + "loss": 1.0189, + "step": 8164 + }, + { + "epoch": 0.6, + "learning_rate": 1.9740041961451977e-05, + "loss": 0.9543, + "step": 8165 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739951597328996e-05, + "loss": 1.0077, + "step": 8166 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739861217709894e-05, + "loss": 1.0677, + "step": 8167 + }, + { + "epoch": 0.6, + "learning_rate": 1.973977082259481e-05, + "loss": 1.1089, + "step": 8168 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739680411983895e-05, + "loss": 0.9541, + "step": 8169 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739589985877283e-05, + "loss": 0.9222, + "step": 8170 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739499544275125e-05, + "loss": 1.1058, + "step": 8171 + }, + { + "epoch": 0.6, + "learning_rate": 1.973940908717756e-05, + "loss": 1.0322, + "step": 8172 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739318614584737e-05, + "loss": 1.0548, + "step": 8173 + }, + { + "epoch": 0.6, + "learning_rate": 1.9739228126496794e-05, + "loss": 1.0326, + "step": 8174 + }, + { + "epoch": 0.6, + "learning_rate": 1.973913762291388e-05, + "loss": 1.1111, + "step": 8175 + }, + { + "epoch": 0.6, + "learning_rate": 1.973904710383614e-05, + "loss": 1.0088, + "step": 8176 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738956569263713e-05, + "loss": 1.0551, + "step": 8177 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738866019196747e-05, + "loss": 1.0696, + "step": 8178 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738775453635383e-05, + "loss": 1.065, + "step": 8179 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738684872579767e-05, + "loss": 0.9523, + "step": 8180 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738594276030044e-05, + "loss": 0.9314, + "step": 8181 + }, + { + "epoch": 0.6, + "learning_rate": 1.973850366398635e-05, + "loss": 1.0714, + "step": 8182 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738413036448847e-05, + "loss": 0.9816, + "step": 8183 + }, + { + "epoch": 0.6, + "learning_rate": 1.973832239341766e-05, + "loss": 1.0039, + "step": 8184 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738231734892946e-05, + "loss": 1.0171, + "step": 8185 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738141060874844e-05, + "loss": 1.028, + "step": 8186 + }, + { + "epoch": 0.6, + "learning_rate": 1.9738050371363496e-05, + "loss": 1.1212, + "step": 8187 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737959666359052e-05, + "loss": 1.0557, + "step": 8188 + }, + { + "epoch": 0.61, + "learning_rate": 1.973786894586165e-05, + "loss": 1.0141, + "step": 8189 + }, + { + "epoch": 0.61, + "learning_rate": 1.973777820987144e-05, + "loss": 1.0381, + "step": 8190 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737687458388566e-05, + "loss": 1.0513, + "step": 8191 + }, + { + "epoch": 0.61, + "learning_rate": 1.973759669141317e-05, + "loss": 1.0495, + "step": 8192 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737505908945397e-05, + "loss": 1.0888, + "step": 8193 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737415110985388e-05, + "loss": 1.0838, + "step": 8194 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737324297533293e-05, + "loss": 1.1, + "step": 8195 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737233468589253e-05, + "loss": 0.9541, + "step": 8196 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737142624153418e-05, + "loss": 1.0613, + "step": 8197 + }, + { + "epoch": 0.61, + "learning_rate": 1.9737051764225926e-05, + "loss": 1.0059, + "step": 8198 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736960888806923e-05, + "loss": 1.04, + "step": 8199 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736869997896555e-05, + "loss": 1.0655, + "step": 8200 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736779091494967e-05, + "loss": 1.0999, + "step": 8201 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736688169602302e-05, + "loss": 1.0878, + "step": 8202 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736597232218706e-05, + "loss": 0.9361, + "step": 8203 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736506279344323e-05, + "loss": 1.0569, + "step": 8204 + }, + { + "epoch": 0.61, + "learning_rate": 1.97364153109793e-05, + "loss": 0.9208, + "step": 8205 + }, + { + "epoch": 0.61, + "learning_rate": 1.9736324327123774e-05, + "loss": 0.9469, + "step": 8206 + }, + { + "epoch": 0.61, + "learning_rate": 1.97362333277779e-05, + "loss": 1.11, + "step": 8207 + }, + { + "epoch": 0.61, + "learning_rate": 1.973614231294182e-05, + "loss": 1.0361, + "step": 8208 + }, + { + "epoch": 0.61, + "learning_rate": 1.973605128261567e-05, + "loss": 1.0513, + "step": 8209 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735960236799606e-05, + "loss": 1.0269, + "step": 8210 + }, + { + "epoch": 0.61, + "learning_rate": 1.973586917549377e-05, + "loss": 1.0095, + "step": 8211 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735778098698303e-05, + "loss": 0.9594, + "step": 8212 + }, + { + "epoch": 0.61, + "learning_rate": 1.973568700641335e-05, + "loss": 0.9363, + "step": 8213 + }, + { + "epoch": 0.61, + "learning_rate": 1.973559589863906e-05, + "loss": 1.1375, + "step": 8214 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735504775375578e-05, + "loss": 1.0194, + "step": 8215 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735413636623048e-05, + "loss": 1.0908, + "step": 8216 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735322482381612e-05, + "loss": 1.0471, + "step": 8217 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735231312651416e-05, + "loss": 1.0987, + "step": 8218 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735140127432607e-05, + "loss": 0.9516, + "step": 8219 + }, + { + "epoch": 0.61, + "learning_rate": 1.9735048926725332e-05, + "loss": 1.0435, + "step": 8220 + }, + { + "epoch": 0.61, + "learning_rate": 1.973495771052973e-05, + "loss": 1.0123, + "step": 8221 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734866478845953e-05, + "loss": 0.9038, + "step": 8222 + }, + { + "epoch": 0.61, + "learning_rate": 1.973477523167414e-05, + "loss": 0.9416, + "step": 8223 + }, + { + "epoch": 0.61, + "learning_rate": 1.973468396901444e-05, + "loss": 0.9437, + "step": 8224 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734592690866998e-05, + "loss": 1.073, + "step": 8225 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734501397231955e-05, + "loss": 1.0242, + "step": 8226 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734410088109464e-05, + "loss": 1.0502, + "step": 8227 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734318763499664e-05, + "loss": 0.9803, + "step": 8228 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734227423402702e-05, + "loss": 0.981, + "step": 8229 + }, + { + "epoch": 0.61, + "learning_rate": 1.9734136067818723e-05, + "loss": 1.0939, + "step": 8230 + }, + { + "epoch": 0.61, + "learning_rate": 1.973404469674787e-05, + "loss": 0.9621, + "step": 8231 + }, + { + "epoch": 0.61, + "learning_rate": 1.97339533101903e-05, + "loss": 1.1159, + "step": 8232 + }, + { + "epoch": 0.61, + "learning_rate": 1.973386190814614e-05, + "loss": 1.0446, + "step": 8233 + }, + { + "epoch": 0.61, + "learning_rate": 1.973377049061555e-05, + "loss": 0.9607, + "step": 8234 + }, + { + "epoch": 0.61, + "learning_rate": 1.973367905759867e-05, + "loss": 1.0304, + "step": 8235 + }, + { + "epoch": 0.61, + "learning_rate": 1.9733587609095644e-05, + "loss": 1.0215, + "step": 8236 + }, + { + "epoch": 0.61, + "learning_rate": 1.973349614510662e-05, + "loss": 0.9625, + "step": 8237 + }, + { + "epoch": 0.61, + "learning_rate": 1.9733404665631742e-05, + "loss": 1.098, + "step": 8238 + }, + { + "epoch": 0.61, + "learning_rate": 1.9733313170671157e-05, + "loss": 1.1006, + "step": 8239 + }, + { + "epoch": 0.61, + "learning_rate": 1.973322166022501e-05, + "loss": 1.0, + "step": 8240 + }, + { + "epoch": 0.61, + "learning_rate": 1.973313013429345e-05, + "loss": 0.9551, + "step": 8241 + }, + { + "epoch": 0.61, + "learning_rate": 1.9733038592876618e-05, + "loss": 0.9523, + "step": 8242 + }, + { + "epoch": 0.61, + "learning_rate": 1.973294703597466e-05, + "loss": 1.0059, + "step": 8243 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732855463587718e-05, + "loss": 0.9422, + "step": 8244 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732763875715947e-05, + "loss": 1.0359, + "step": 8245 + }, + { + "epoch": 0.61, + "learning_rate": 1.973267227235949e-05, + "loss": 1.0293, + "step": 8246 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732580653518486e-05, + "loss": 1.1103, + "step": 8247 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732489019193087e-05, + "loss": 1.0079, + "step": 8248 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732397369383438e-05, + "loss": 1.0629, + "step": 8249 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732305704089684e-05, + "loss": 0.9301, + "step": 8250 + }, + { + "epoch": 0.61, + "learning_rate": 1.973221402331197e-05, + "loss": 0.9196, + "step": 8251 + }, + { + "epoch": 0.61, + "learning_rate": 1.973212232705044e-05, + "loss": 0.982, + "step": 8252 + }, + { + "epoch": 0.61, + "learning_rate": 1.9732030615305247e-05, + "loss": 1.1176, + "step": 8253 + }, + { + "epoch": 0.61, + "learning_rate": 1.9731938888076532e-05, + "loss": 1.0592, + "step": 8254 + }, + { + "epoch": 0.61, + "learning_rate": 1.973184714536444e-05, + "loss": 1.0191, + "step": 8255 + }, + { + "epoch": 0.61, + "learning_rate": 1.973175538716912e-05, + "loss": 1.0287, + "step": 8256 + }, + { + "epoch": 0.61, + "learning_rate": 1.9731663613490715e-05, + "loss": 1.0003, + "step": 8257 + }, + { + "epoch": 0.61, + "learning_rate": 1.973157182432937e-05, + "loss": 1.0355, + "step": 8258 + }, + { + "epoch": 0.61, + "learning_rate": 1.973148001968524e-05, + "loss": 1.0673, + "step": 8259 + }, + { + "epoch": 0.61, + "learning_rate": 1.973138819955846e-05, + "loss": 1.053, + "step": 8260 + }, + { + "epoch": 0.61, + "learning_rate": 1.973129636394918e-05, + "loss": 1.0745, + "step": 8261 + }, + { + "epoch": 0.61, + "learning_rate": 1.9731204512857548e-05, + "loss": 1.0692, + "step": 8262 + }, + { + "epoch": 0.61, + "learning_rate": 1.973111264628371e-05, + "loss": 1.0336, + "step": 8263 + }, + { + "epoch": 0.61, + "learning_rate": 1.9731020764227812e-05, + "loss": 0.965, + "step": 8264 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730928866689994e-05, + "loss": 1.1424, + "step": 8265 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730836953670414e-05, + "loss": 1.091, + "step": 8266 + }, + { + "epoch": 0.61, + "learning_rate": 1.973074502516921e-05, + "loss": 1.0471, + "step": 8267 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730653081186524e-05, + "loss": 0.9737, + "step": 8268 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730561121722514e-05, + "loss": 1.1235, + "step": 8269 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730469146777317e-05, + "loss": 1.0322, + "step": 8270 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730377156351085e-05, + "loss": 1.0528, + "step": 8271 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730285150443963e-05, + "loss": 1.0788, + "step": 8272 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730193129056095e-05, + "loss": 1.0524, + "step": 8273 + }, + { + "epoch": 0.61, + "learning_rate": 1.973010109218763e-05, + "loss": 0.9733, + "step": 8274 + }, + { + "epoch": 0.61, + "learning_rate": 1.9730009039838715e-05, + "loss": 1.0823, + "step": 8275 + }, + { + "epoch": 0.61, + "learning_rate": 1.972991697200949e-05, + "loss": 1.0139, + "step": 8276 + }, + { + "epoch": 0.61, + "learning_rate": 1.972982488870011e-05, + "loss": 1.0598, + "step": 8277 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729732789910715e-05, + "loss": 1.145, + "step": 8278 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729640675641456e-05, + "loss": 1.0437, + "step": 8279 + }, + { + "epoch": 0.61, + "learning_rate": 1.972954854589248e-05, + "loss": 1.0618, + "step": 8280 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729456400663926e-05, + "loss": 0.9636, + "step": 8281 + }, + { + "epoch": 0.61, + "learning_rate": 1.972936423995595e-05, + "loss": 0.9124, + "step": 8282 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729272063768694e-05, + "loss": 0.9849, + "step": 8283 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729179872102307e-05, + "loss": 1.1128, + "step": 8284 + }, + { + "epoch": 0.61, + "learning_rate": 1.9729087664956932e-05, + "loss": 0.9821, + "step": 8285 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728995442332718e-05, + "loss": 1.1701, + "step": 8286 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728903204229813e-05, + "loss": 1.074, + "step": 8287 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728810950648357e-05, + "loss": 0.9954, + "step": 8288 + }, + { + "epoch": 0.61, + "learning_rate": 1.972871868158851e-05, + "loss": 1.0596, + "step": 8289 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728626397050405e-05, + "loss": 1.0432, + "step": 8290 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728534097034196e-05, + "loss": 1.051, + "step": 8291 + }, + { + "epoch": 0.61, + "learning_rate": 1.972844178154003e-05, + "loss": 1.0912, + "step": 8292 + }, + { + "epoch": 0.61, + "learning_rate": 1.972834945056805e-05, + "loss": 0.981, + "step": 8293 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728257104118406e-05, + "loss": 1.0639, + "step": 8294 + }, + { + "epoch": 0.61, + "learning_rate": 1.9728164742191244e-05, + "loss": 1.0901, + "step": 8295 + }, + { + "epoch": 0.61, + "learning_rate": 1.972807236478671e-05, + "loss": 1.1113, + "step": 8296 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727979971904957e-05, + "loss": 1.0424, + "step": 8297 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727887563546123e-05, + "loss": 1.1019, + "step": 8298 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727795139710357e-05, + "loss": 1.0596, + "step": 8299 + }, + { + "epoch": 0.61, + "learning_rate": 1.972770270039781e-05, + "loss": 1.0635, + "step": 8300 + }, + { + "epoch": 0.61, + "learning_rate": 1.972761024560863e-05, + "loss": 0.965, + "step": 8301 + }, + { + "epoch": 0.61, + "learning_rate": 1.972751777534296e-05, + "loss": 1.1123, + "step": 8302 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727425289600946e-05, + "loss": 1.0127, + "step": 8303 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727332788382737e-05, + "loss": 1.1243, + "step": 8304 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727240271688487e-05, + "loss": 1.1261, + "step": 8305 + }, + { + "epoch": 0.61, + "learning_rate": 1.972714773951833e-05, + "loss": 1.0689, + "step": 8306 + }, + { + "epoch": 0.61, + "learning_rate": 1.9727055191872424e-05, + "loss": 1.017, + "step": 8307 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726962628750912e-05, + "loss": 1.0881, + "step": 8308 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726870050153943e-05, + "loss": 1.0948, + "step": 8309 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726777456081657e-05, + "loss": 1.0039, + "step": 8310 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726684846534212e-05, + "loss": 1.0926, + "step": 8311 + }, + { + "epoch": 0.61, + "learning_rate": 1.972659222151175e-05, + "loss": 1.0508, + "step": 8312 + }, + { + "epoch": 0.61, + "learning_rate": 1.972649958101442e-05, + "loss": 1.0127, + "step": 8313 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726406925042365e-05, + "loss": 1.0513, + "step": 8314 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726314253595738e-05, + "loss": 1.0749, + "step": 8315 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726221566674683e-05, + "loss": 1.0551, + "step": 8316 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726128864279352e-05, + "loss": 1.0409, + "step": 8317 + }, + { + "epoch": 0.61, + "learning_rate": 1.9726036146409884e-05, + "loss": 1.0686, + "step": 8318 + }, + { + "epoch": 0.61, + "learning_rate": 1.9725943413066436e-05, + "loss": 1.0783, + "step": 8319 + }, + { + "epoch": 0.61, + "learning_rate": 1.9725850664249146e-05, + "loss": 1.0504, + "step": 8320 + }, + { + "epoch": 0.61, + "learning_rate": 1.972575789995817e-05, + "loss": 1.1724, + "step": 8321 + }, + { + "epoch": 0.61, + "learning_rate": 1.9725665120193654e-05, + "loss": 1.0843, + "step": 8322 + }, + { + "epoch": 0.61, + "learning_rate": 1.9725572324955743e-05, + "loss": 1.0237, + "step": 8323 + }, + { + "epoch": 0.62, + "learning_rate": 1.9725479514244587e-05, + "loss": 1.0493, + "step": 8324 + }, + { + "epoch": 0.62, + "learning_rate": 1.9725386688060327e-05, + "loss": 1.0108, + "step": 8325 + }, + { + "epoch": 0.62, + "learning_rate": 1.972529384640312e-05, + "loss": 1.0889, + "step": 8326 + }, + { + "epoch": 0.62, + "learning_rate": 1.972520098927311e-05, + "loss": 0.9752, + "step": 8327 + }, + { + "epoch": 0.62, + "learning_rate": 1.972510811667044e-05, + "loss": 1.1278, + "step": 8328 + }, + { + "epoch": 0.62, + "learning_rate": 1.972501522859527e-05, + "loss": 1.0822, + "step": 8329 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724922325047733e-05, + "loss": 1.0025, + "step": 8330 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724829406027985e-05, + "loss": 1.0602, + "step": 8331 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724736471536174e-05, + "loss": 1.0638, + "step": 8332 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724643521572445e-05, + "loss": 1.0474, + "step": 8333 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724550556136948e-05, + "loss": 1.0562, + "step": 8334 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724457575229832e-05, + "loss": 1.0281, + "step": 8335 + }, + { + "epoch": 0.62, + "learning_rate": 1.972436457885124e-05, + "loss": 1.0252, + "step": 8336 + }, + { + "epoch": 0.62, + "learning_rate": 1.972427156700132e-05, + "loss": 1.0847, + "step": 8337 + }, + { + "epoch": 0.62, + "learning_rate": 1.9724178539680232e-05, + "loss": 1.1987, + "step": 8338 + }, + { + "epoch": 0.62, + "learning_rate": 1.972408549688811e-05, + "loss": 1.0055, + "step": 8339 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723992438625106e-05, + "loss": 1.039, + "step": 8340 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723899364891372e-05, + "loss": 1.0196, + "step": 8341 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723806275687055e-05, + "loss": 1.0073, + "step": 8342 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723713171012297e-05, + "loss": 1.0281, + "step": 8343 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723620050867252e-05, + "loss": 1.0522, + "step": 8344 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723526915252066e-05, + "loss": 0.9129, + "step": 8345 + }, + { + "epoch": 0.62, + "learning_rate": 1.972343376416689e-05, + "loss": 1.0387, + "step": 8346 + }, + { + "epoch": 0.62, + "learning_rate": 1.972334059761187e-05, + "loss": 1.0134, + "step": 8347 + }, + { + "epoch": 0.62, + "learning_rate": 1.972324741558715e-05, + "loss": 1.0732, + "step": 8348 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723154218092887e-05, + "loss": 1.134, + "step": 8349 + }, + { + "epoch": 0.62, + "learning_rate": 1.9723061005129222e-05, + "loss": 0.9519, + "step": 8350 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722967776696308e-05, + "loss": 1.0488, + "step": 8351 + }, + { + "epoch": 0.62, + "learning_rate": 1.972287453279429e-05, + "loss": 0.8735, + "step": 8352 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722781273423318e-05, + "loss": 0.849, + "step": 8353 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722687998583538e-05, + "loss": 0.9107, + "step": 8354 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722594708275103e-05, + "loss": 0.9956, + "step": 8355 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722501402498156e-05, + "loss": 1.1013, + "step": 8356 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722408081252853e-05, + "loss": 1.0106, + "step": 8357 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722314744539336e-05, + "loss": 1.1099, + "step": 8358 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722221392357757e-05, + "loss": 1.0359, + "step": 8359 + }, + { + "epoch": 0.62, + "learning_rate": 1.972212802470826e-05, + "loss": 1.0402, + "step": 8360 + }, + { + "epoch": 0.62, + "learning_rate": 1.9722034641590996e-05, + "loss": 1.0667, + "step": 8361 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721941243006116e-05, + "loss": 1.0262, + "step": 8362 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721847828953768e-05, + "loss": 1.1733, + "step": 8363 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721754399434096e-05, + "loss": 1.1065, + "step": 8364 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721660954447254e-05, + "loss": 1.0456, + "step": 8365 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721567493993387e-05, + "loss": 1.0255, + "step": 8366 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721474018072647e-05, + "loss": 1.0905, + "step": 8367 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721380526685177e-05, + "loss": 1.0821, + "step": 8368 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721287019831133e-05, + "loss": 0.9785, + "step": 8369 + }, + { + "epoch": 0.62, + "learning_rate": 1.972119349751066e-05, + "loss": 1.0022, + "step": 8370 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721099959723905e-05, + "loss": 1.1128, + "step": 8371 + }, + { + "epoch": 0.62, + "learning_rate": 1.9721006406471023e-05, + "loss": 0.9779, + "step": 8372 + }, + { + "epoch": 0.62, + "learning_rate": 1.972091283775216e-05, + "loss": 0.8907, + "step": 8373 + }, + { + "epoch": 0.62, + "learning_rate": 1.972081925356746e-05, + "loss": 1.0986, + "step": 8374 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720725653917076e-05, + "loss": 1.0295, + "step": 8375 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720632038801156e-05, + "loss": 1.0531, + "step": 8376 + }, + { + "epoch": 0.62, + "learning_rate": 1.972053840821985e-05, + "loss": 1.1066, + "step": 8377 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720444762173306e-05, + "loss": 1.0286, + "step": 8378 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720351100661674e-05, + "loss": 0.9294, + "step": 8379 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720257423685105e-05, + "loss": 1.1201, + "step": 8380 + }, + { + "epoch": 0.62, + "learning_rate": 1.9720163731243743e-05, + "loss": 0.9885, + "step": 8381 + }, + { + "epoch": 0.62, + "learning_rate": 1.972007002333774e-05, + "loss": 0.9917, + "step": 8382 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719976299967244e-05, + "loss": 0.9956, + "step": 8383 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719882561132407e-05, + "loss": 1.0957, + "step": 8384 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719788806833374e-05, + "loss": 1.1024, + "step": 8385 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719695037070295e-05, + "loss": 1.1578, + "step": 8386 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719601251843322e-05, + "loss": 1.0882, + "step": 8387 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719507451152603e-05, + "loss": 0.9693, + "step": 8388 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719413634998285e-05, + "loss": 1.0438, + "step": 8389 + }, + { + "epoch": 0.62, + "learning_rate": 1.971931980338052e-05, + "loss": 0.9664, + "step": 8390 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719225956299456e-05, + "loss": 1.0698, + "step": 8391 + }, + { + "epoch": 0.62, + "learning_rate": 1.9719132093755244e-05, + "loss": 1.0275, + "step": 8392 + }, + { + "epoch": 0.62, + "learning_rate": 1.971903821574803e-05, + "loss": 1.0636, + "step": 8393 + }, + { + "epoch": 0.62, + "learning_rate": 1.9718944322277966e-05, + "loss": 1.0475, + "step": 8394 + }, + { + "epoch": 0.62, + "learning_rate": 1.97188504133452e-05, + "loss": 1.0977, + "step": 8395 + }, + { + "epoch": 0.62, + "learning_rate": 1.9718756488949885e-05, + "loss": 1.0, + "step": 8396 + }, + { + "epoch": 0.62, + "learning_rate": 1.971866254909216e-05, + "loss": 1.0336, + "step": 8397 + }, + { + "epoch": 0.62, + "learning_rate": 1.971856859377219e-05, + "loss": 0.9821, + "step": 8398 + }, + { + "epoch": 0.62, + "learning_rate": 1.971847462299011e-05, + "loss": 1.1547, + "step": 8399 + }, + { + "epoch": 0.62, + "learning_rate": 1.971838063674608e-05, + "loss": 0.9343, + "step": 8400 + }, + { + "epoch": 0.62, + "learning_rate": 1.9718286635040247e-05, + "loss": 1.1112, + "step": 8401 + }, + { + "epoch": 0.62, + "learning_rate": 1.9718192617872756e-05, + "loss": 1.0337, + "step": 8402 + }, + { + "epoch": 0.62, + "learning_rate": 1.9718098585243762e-05, + "loss": 0.9553, + "step": 8403 + }, + { + "epoch": 0.62, + "learning_rate": 1.971800453715341e-05, + "loss": 1.0268, + "step": 8404 + }, + { + "epoch": 0.62, + "learning_rate": 1.971791047360185e-05, + "loss": 1.0222, + "step": 8405 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717816394589238e-05, + "loss": 0.954, + "step": 8406 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717722300115716e-05, + "loss": 1.0708, + "step": 8407 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717628190181437e-05, + "loss": 0.9956, + "step": 8408 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717534064786554e-05, + "loss": 1.0759, + "step": 8409 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717439923931208e-05, + "loss": 0.9558, + "step": 8410 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717345767615556e-05, + "loss": 0.9997, + "step": 8411 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717251595839748e-05, + "loss": 1.0757, + "step": 8412 + }, + { + "epoch": 0.62, + "learning_rate": 1.971715740860393e-05, + "loss": 1.0896, + "step": 8413 + }, + { + "epoch": 0.62, + "learning_rate": 1.9717063205908255e-05, + "loss": 0.9626, + "step": 8414 + }, + { + "epoch": 0.62, + "learning_rate": 1.971696898775287e-05, + "loss": 1.0131, + "step": 8415 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716874754137926e-05, + "loss": 0.9567, + "step": 8416 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716780505063574e-05, + "loss": 1.0161, + "step": 8417 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716686240529966e-05, + "loss": 1.0604, + "step": 8418 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716591960537245e-05, + "loss": 1.0447, + "step": 8419 + }, + { + "epoch": 0.62, + "learning_rate": 1.971649766508557e-05, + "loss": 1.0863, + "step": 8420 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716403354175078e-05, + "loss": 0.9523, + "step": 8421 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716309027805934e-05, + "loss": 1.0545, + "step": 8422 + }, + { + "epoch": 0.62, + "learning_rate": 1.971621468597828e-05, + "loss": 1.0245, + "step": 8423 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716120328692267e-05, + "loss": 1.0635, + "step": 8424 + }, + { + "epoch": 0.62, + "learning_rate": 1.9716025955948046e-05, + "loss": 1.0944, + "step": 8425 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715931567745768e-05, + "loss": 0.9492, + "step": 8426 + }, + { + "epoch": 0.62, + "learning_rate": 1.971583716408558e-05, + "loss": 1.0084, + "step": 8427 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715742744967635e-05, + "loss": 1.0871, + "step": 8428 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715648310392082e-05, + "loss": 1.1372, + "step": 8429 + }, + { + "epoch": 0.62, + "learning_rate": 1.971555386035907e-05, + "loss": 0.9832, + "step": 8430 + }, + { + "epoch": 0.62, + "learning_rate": 1.971545939486875e-05, + "loss": 1.0824, + "step": 8431 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715364913921278e-05, + "loss": 1.1085, + "step": 8432 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715270417516796e-05, + "loss": 1.0404, + "step": 8433 + }, + { + "epoch": 0.62, + "learning_rate": 1.971517590565546e-05, + "loss": 1.1073, + "step": 8434 + }, + { + "epoch": 0.62, + "learning_rate": 1.9715081378337415e-05, + "loss": 0.9731, + "step": 8435 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714986835562815e-05, + "loss": 1.0588, + "step": 8436 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714892277331812e-05, + "loss": 1.1373, + "step": 8437 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714797703644555e-05, + "loss": 1.0549, + "step": 8438 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714703114501193e-05, + "loss": 0.9145, + "step": 8439 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714608509901872e-05, + "loss": 1.1122, + "step": 8440 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714513889846754e-05, + "loss": 0.9494, + "step": 8441 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714419254335985e-05, + "loss": 1.0632, + "step": 8442 + }, + { + "epoch": 0.62, + "learning_rate": 1.971432460336971e-05, + "loss": 0.9343, + "step": 8443 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714229936948084e-05, + "loss": 1.0427, + "step": 8444 + }, + { + "epoch": 0.62, + "learning_rate": 1.9714135255071257e-05, + "loss": 1.0427, + "step": 8445 + }, + { + "epoch": 0.62, + "learning_rate": 1.971404055773938e-05, + "loss": 1.0637, + "step": 8446 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713945844952605e-05, + "loss": 1.0409, + "step": 8447 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713851116711078e-05, + "loss": 1.1742, + "step": 8448 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713756373014954e-05, + "loss": 1.0047, + "step": 8449 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713661613864385e-05, + "loss": 1.1453, + "step": 8450 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713566839259517e-05, + "loss": 1.0593, + "step": 8451 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713472049200505e-05, + "loss": 0.9033, + "step": 8452 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713377243687497e-05, + "loss": 1.0377, + "step": 8453 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713282422720643e-05, + "loss": 1.1127, + "step": 8454 + }, + { + "epoch": 0.62, + "learning_rate": 1.9713187586300096e-05, + "loss": 0.996, + "step": 8455 + }, + { + "epoch": 0.62, + "learning_rate": 1.971309273442601e-05, + "loss": 1.1243, + "step": 8456 + }, + { + "epoch": 0.62, + "learning_rate": 1.971299786709853e-05, + "loss": 0.9897, + "step": 8457 + }, + { + "epoch": 0.62, + "learning_rate": 1.971290298431781e-05, + "loss": 0.9706, + "step": 8458 + }, + { + "epoch": 0.63, + "learning_rate": 1.9712808086084e-05, + "loss": 0.9651, + "step": 8459 + }, + { + "epoch": 0.63, + "learning_rate": 1.971271317239725e-05, + "loss": 0.9079, + "step": 8460 + }, + { + "epoch": 0.63, + "learning_rate": 1.9712618243257712e-05, + "loss": 1.0738, + "step": 8461 + }, + { + "epoch": 0.63, + "learning_rate": 1.971252329866554e-05, + "loss": 1.1028, + "step": 8462 + }, + { + "epoch": 0.63, + "learning_rate": 1.971242833862088e-05, + "loss": 1.1361, + "step": 8463 + }, + { + "epoch": 0.63, + "learning_rate": 1.9712333363123886e-05, + "loss": 1.0508, + "step": 8464 + }, + { + "epoch": 0.63, + "learning_rate": 1.971223837217471e-05, + "loss": 0.966, + "step": 8465 + }, + { + "epoch": 0.63, + "learning_rate": 1.97121433657735e-05, + "loss": 0.9498, + "step": 8466 + }, + { + "epoch": 0.63, + "learning_rate": 1.971204834392041e-05, + "loss": 1.0692, + "step": 8467 + }, + { + "epoch": 0.63, + "learning_rate": 1.971195330661559e-05, + "loss": 1.166, + "step": 8468 + }, + { + "epoch": 0.63, + "learning_rate": 1.971185825385919e-05, + "loss": 1.0908, + "step": 8469 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711763185651365e-05, + "loss": 1.014, + "step": 8470 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711668101992263e-05, + "loss": 0.9095, + "step": 8471 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711573002882035e-05, + "loss": 0.975, + "step": 8472 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711477888320836e-05, + "loss": 1.0971, + "step": 8473 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711382758308814e-05, + "loss": 1.1142, + "step": 8474 + }, + { + "epoch": 0.63, + "learning_rate": 1.9711287612846118e-05, + "loss": 1.1125, + "step": 8475 + }, + { + "epoch": 0.63, + "learning_rate": 1.97111924519329e-05, + "loss": 1.0927, + "step": 8476 + }, + { + "epoch": 0.63, + "learning_rate": 1.971109727556932e-05, + "loss": 1.0371, + "step": 8477 + }, + { + "epoch": 0.63, + "learning_rate": 1.971100208375552e-05, + "loss": 1.1204, + "step": 8478 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710906876491656e-05, + "loss": 1.1308, + "step": 8479 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710811653777875e-05, + "loss": 1.0161, + "step": 8480 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710716415614335e-05, + "loss": 1.0151, + "step": 8481 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710621162001182e-05, + "loss": 1.1038, + "step": 8482 + }, + { + "epoch": 0.63, + "learning_rate": 1.971052589293857e-05, + "loss": 1.0642, + "step": 8483 + }, + { + "epoch": 0.63, + "learning_rate": 1.971043060842665e-05, + "loss": 1.1228, + "step": 8484 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710335308465576e-05, + "loss": 1.0874, + "step": 8485 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710239993055498e-05, + "loss": 1.0521, + "step": 8486 + }, + { + "epoch": 0.63, + "learning_rate": 1.9710144662196565e-05, + "loss": 1.0087, + "step": 8487 + }, + { + "epoch": 0.63, + "learning_rate": 1.971004931588893e-05, + "loss": 1.1625, + "step": 8488 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709953954132746e-05, + "loss": 1.0129, + "step": 8489 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709858576928168e-05, + "loss": 0.9539, + "step": 8490 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709763184275336e-05, + "loss": 1.085, + "step": 8491 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709667776174415e-05, + "loss": 1.0827, + "step": 8492 + }, + { + "epoch": 0.63, + "learning_rate": 1.970957235262555e-05, + "loss": 0.9331, + "step": 8493 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709476913628896e-05, + "loss": 1.0768, + "step": 8494 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709381459184604e-05, + "loss": 1.0648, + "step": 8495 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709285989292822e-05, + "loss": 0.9104, + "step": 8496 + }, + { + "epoch": 0.63, + "learning_rate": 1.9709190503953704e-05, + "loss": 1.0678, + "step": 8497 + }, + { + "epoch": 0.63, + "learning_rate": 1.970909500316741e-05, + "loss": 0.8763, + "step": 8498 + }, + { + "epoch": 0.63, + "learning_rate": 1.970899948693408e-05, + "loss": 0.9803, + "step": 8499 + }, + { + "epoch": 0.63, + "learning_rate": 1.9708903955253866e-05, + "loss": 1.0222, + "step": 8500 + }, + { + "epoch": 0.63, + "learning_rate": 1.970880840812693e-05, + "loss": 1.0151, + "step": 8501 + }, + { + "epoch": 0.63, + "learning_rate": 1.970871284555342e-05, + "loss": 1.1212, + "step": 8502 + }, + { + "epoch": 0.63, + "learning_rate": 1.9708617267533486e-05, + "loss": 1.0358, + "step": 8503 + }, + { + "epoch": 0.63, + "learning_rate": 1.9708521674067275e-05, + "loss": 0.9619, + "step": 8504 + }, + { + "epoch": 0.63, + "learning_rate": 1.970842606515495e-05, + "loss": 0.9946, + "step": 8505 + }, + { + "epoch": 0.63, + "learning_rate": 1.970833044079666e-05, + "loss": 1.0896, + "step": 8506 + }, + { + "epoch": 0.63, + "learning_rate": 1.970823480099255e-05, + "loss": 1.0017, + "step": 8507 + }, + { + "epoch": 0.63, + "learning_rate": 1.970813914574278e-05, + "loss": 1.0604, + "step": 8508 + }, + { + "epoch": 0.63, + "learning_rate": 1.97080434750475e-05, + "loss": 0.9832, + "step": 8509 + }, + { + "epoch": 0.63, + "learning_rate": 1.9707947788906863e-05, + "loss": 1.0398, + "step": 8510 + }, + { + "epoch": 0.63, + "learning_rate": 1.9707852087321016e-05, + "loss": 1.0326, + "step": 8511 + }, + { + "epoch": 0.63, + "learning_rate": 1.970775637029012e-05, + "loss": 1.0346, + "step": 8512 + }, + { + "epoch": 0.63, + "learning_rate": 1.970766063781432e-05, + "loss": 1.044, + "step": 8513 + }, + { + "epoch": 0.63, + "learning_rate": 1.970756488989377e-05, + "loss": 1.0065, + "step": 8514 + }, + { + "epoch": 0.63, + "learning_rate": 1.9707469126528625e-05, + "loss": 1.1337, + "step": 8515 + }, + { + "epoch": 0.63, + "learning_rate": 1.9707373347719036e-05, + "loss": 1.0514, + "step": 8516 + }, + { + "epoch": 0.63, + "learning_rate": 1.970727755346515e-05, + "loss": 1.0423, + "step": 8517 + }, + { + "epoch": 0.63, + "learning_rate": 1.970718174376713e-05, + "loss": 1.0489, + "step": 8518 + }, + { + "epoch": 0.63, + "learning_rate": 1.9707085918625116e-05, + "loss": 1.0274, + "step": 8519 + }, + { + "epoch": 0.63, + "learning_rate": 1.9706990078039275e-05, + "loss": 1.0281, + "step": 8520 + }, + { + "epoch": 0.63, + "learning_rate": 1.970689422200975e-05, + "loss": 1.0602, + "step": 8521 + }, + { + "epoch": 0.63, + "learning_rate": 1.970679835053669e-05, + "loss": 1.0591, + "step": 8522 + }, + { + "epoch": 0.63, + "learning_rate": 1.9706702463620256e-05, + "loss": 1.1223, + "step": 8523 + }, + { + "epoch": 0.63, + "learning_rate": 1.97066065612606e-05, + "loss": 1.1399, + "step": 8524 + }, + { + "epoch": 0.63, + "learning_rate": 1.9706510643457868e-05, + "loss": 0.9649, + "step": 8525 + }, + { + "epoch": 0.63, + "learning_rate": 1.970641471021222e-05, + "loss": 0.9586, + "step": 8526 + }, + { + "epoch": 0.63, + "learning_rate": 1.9706318761523803e-05, + "loss": 1.0064, + "step": 8527 + }, + { + "epoch": 0.63, + "learning_rate": 1.970622279739277e-05, + "loss": 1.0575, + "step": 8528 + }, + { + "epoch": 0.63, + "learning_rate": 1.9706126817819274e-05, + "loss": 1.1158, + "step": 8529 + }, + { + "epoch": 0.63, + "learning_rate": 1.970603082280347e-05, + "loss": 1.0284, + "step": 8530 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705934812345515e-05, + "loss": 1.0852, + "step": 8531 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705838786445556e-05, + "loss": 1.0918, + "step": 8532 + }, + { + "epoch": 0.63, + "learning_rate": 1.970574274510374e-05, + "loss": 1.0604, + "step": 8533 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705646688320232e-05, + "loss": 1.0118, + "step": 8534 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705550616095177e-05, + "loss": 1.041, + "step": 8535 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705454528428728e-05, + "loss": 1.2076, + "step": 8536 + }, + { + "epoch": 0.63, + "learning_rate": 1.970535842532104e-05, + "loss": 1.0479, + "step": 8537 + }, + { + "epoch": 0.63, + "learning_rate": 1.970526230677227e-05, + "loss": 0.9508, + "step": 8538 + }, + { + "epoch": 0.63, + "learning_rate": 1.970516617278256e-05, + "loss": 1.0445, + "step": 8539 + }, + { + "epoch": 0.63, + "learning_rate": 1.9705070023352075e-05, + "loss": 1.1024, + "step": 8540 + }, + { + "epoch": 0.63, + "learning_rate": 1.970497385848096e-05, + "loss": 1.0367, + "step": 8541 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704877678169372e-05, + "loss": 0.9097, + "step": 8542 + }, + { + "epoch": 0.63, + "learning_rate": 1.970478148241746e-05, + "loss": 1.0741, + "step": 8543 + }, + { + "epoch": 0.63, + "learning_rate": 1.970468527122538e-05, + "loss": 0.9911, + "step": 8544 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704589044593287e-05, + "loss": 0.989, + "step": 8545 + }, + { + "epoch": 0.63, + "learning_rate": 1.970449280252133e-05, + "loss": 0.9583, + "step": 8546 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704396545009665e-05, + "loss": 1.0148, + "step": 8547 + }, + { + "epoch": 0.63, + "learning_rate": 1.970430027205844e-05, + "loss": 1.1208, + "step": 8548 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704203983667817e-05, + "loss": 1.0461, + "step": 8549 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704107679837943e-05, + "loss": 1.0117, + "step": 8550 + }, + { + "epoch": 0.63, + "learning_rate": 1.9704011360568973e-05, + "loss": 0.9901, + "step": 8551 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703915025861058e-05, + "loss": 1.0705, + "step": 8552 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703818675714356e-05, + "loss": 0.992, + "step": 8553 + }, + { + "epoch": 0.63, + "learning_rate": 1.970372231012901e-05, + "loss": 1.1321, + "step": 8554 + }, + { + "epoch": 0.63, + "learning_rate": 1.970362592910519e-05, + "loss": 1.0617, + "step": 8555 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703529532643033e-05, + "loss": 0.9368, + "step": 8556 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703433120742702e-05, + "loss": 1.0845, + "step": 8557 + }, + { + "epoch": 0.63, + "learning_rate": 1.970333669340435e-05, + "loss": 0.9731, + "step": 8558 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703240250628126e-05, + "loss": 1.0492, + "step": 8559 + }, + { + "epoch": 0.63, + "learning_rate": 1.9703143792414185e-05, + "loss": 1.1707, + "step": 8560 + }, + { + "epoch": 0.63, + "learning_rate": 1.970304731876268e-05, + "loss": 0.9863, + "step": 8561 + }, + { + "epoch": 0.63, + "learning_rate": 1.9702950829673767e-05, + "loss": 0.9525, + "step": 8562 + }, + { + "epoch": 0.63, + "learning_rate": 1.9702854325147598e-05, + "loss": 0.9685, + "step": 8563 + }, + { + "epoch": 0.63, + "learning_rate": 1.970275780518433e-05, + "loss": 1.1026, + "step": 8564 + }, + { + "epoch": 0.63, + "learning_rate": 1.9702661269784107e-05, + "loss": 1.0254, + "step": 8565 + }, + { + "epoch": 0.63, + "learning_rate": 1.9702564718947093e-05, + "loss": 0.9337, + "step": 8566 + }, + { + "epoch": 0.63, + "learning_rate": 1.9702468152673436e-05, + "loss": 1.065, + "step": 8567 + }, + { + "epoch": 0.63, + "learning_rate": 1.970237157096329e-05, + "loss": 1.1239, + "step": 8568 + }, + { + "epoch": 0.63, + "learning_rate": 1.970227497381681e-05, + "loss": 1.0941, + "step": 8569 + }, + { + "epoch": 0.63, + "learning_rate": 1.970217836123415e-05, + "loss": 1.0087, + "step": 8570 + }, + { + "epoch": 0.63, + "learning_rate": 1.970208173321546e-05, + "loss": 1.0886, + "step": 8571 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701985089760902e-05, + "loss": 1.1189, + "step": 8572 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701888430870623e-05, + "loss": 1.0504, + "step": 8573 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701791756544777e-05, + "loss": 1.1047, + "step": 8574 + }, + { + "epoch": 0.63, + "learning_rate": 1.970169506678352e-05, + "loss": 1.0441, + "step": 8575 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701598361587004e-05, + "loss": 1.05, + "step": 8576 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701501640955385e-05, + "loss": 1.0489, + "step": 8577 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701404904888816e-05, + "loss": 1.0586, + "step": 8578 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701308153387447e-05, + "loss": 1.03, + "step": 8579 + }, + { + "epoch": 0.63, + "learning_rate": 1.970121138645144e-05, + "loss": 1.1205, + "step": 8580 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701114604080942e-05, + "loss": 1.0062, + "step": 8581 + }, + { + "epoch": 0.63, + "learning_rate": 1.9701017806276114e-05, + "loss": 1.0901, + "step": 8582 + }, + { + "epoch": 0.63, + "learning_rate": 1.97009209930371e-05, + "loss": 1.0585, + "step": 8583 + }, + { + "epoch": 0.63, + "learning_rate": 1.9700824164364062e-05, + "loss": 1.0065, + "step": 8584 + }, + { + "epoch": 0.63, + "learning_rate": 1.970072732025715e-05, + "loss": 0.9658, + "step": 8585 + }, + { + "epoch": 0.63, + "learning_rate": 1.970063046071652e-05, + "loss": 1.0283, + "step": 8586 + }, + { + "epoch": 0.63, + "learning_rate": 1.9700533585742327e-05, + "loss": 0.9801, + "step": 8587 + }, + { + "epoch": 0.63, + "learning_rate": 1.970043669533472e-05, + "loss": 1.0222, + "step": 8588 + }, + { + "epoch": 0.63, + "learning_rate": 1.9700339789493862e-05, + "loss": 1.0805, + "step": 8589 + }, + { + "epoch": 0.63, + "learning_rate": 1.9700242868219897e-05, + "loss": 1.1162, + "step": 8590 + }, + { + "epoch": 0.63, + "learning_rate": 1.970014593151299e-05, + "loss": 0.9708, + "step": 8591 + }, + { + "epoch": 0.63, + "learning_rate": 1.9700048979373286e-05, + "loss": 0.9864, + "step": 8592 + }, + { + "epoch": 0.63, + "learning_rate": 1.9699952011800944e-05, + "loss": 1.0237, + "step": 8593 + }, + { + "epoch": 0.64, + "learning_rate": 1.9699855028796116e-05, + "loss": 1.0155, + "step": 8594 + }, + { + "epoch": 0.64, + "learning_rate": 1.969975803035896e-05, + "loss": 1.0589, + "step": 8595 + }, + { + "epoch": 0.64, + "learning_rate": 1.9699661016489628e-05, + "loss": 1.0647, + "step": 8596 + }, + { + "epoch": 0.64, + "learning_rate": 1.9699563987188273e-05, + "loss": 1.0748, + "step": 8597 + }, + { + "epoch": 0.64, + "learning_rate": 1.969946694245505e-05, + "loss": 1.0403, + "step": 8598 + }, + { + "epoch": 0.64, + "learning_rate": 1.9699369882290113e-05, + "loss": 0.9977, + "step": 8599 + }, + { + "epoch": 0.64, + "learning_rate": 1.969927280669362e-05, + "loss": 1.0566, + "step": 8600 + }, + { + "epoch": 0.64, + "learning_rate": 1.9699175715665717e-05, + "loss": 1.0285, + "step": 8601 + }, + { + "epoch": 0.64, + "learning_rate": 1.969907860920657e-05, + "loss": 0.9853, + "step": 8602 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698981487316327e-05, + "loss": 1.0759, + "step": 8603 + }, + { + "epoch": 0.64, + "learning_rate": 1.969888434999514e-05, + "loss": 1.0229, + "step": 8604 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698787197243168e-05, + "loss": 1.0544, + "step": 8605 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698690029060566e-05, + "loss": 1.1266, + "step": 8606 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698592845447487e-05, + "loss": 1.0009, + "step": 8607 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698495646404084e-05, + "loss": 1.0706, + "step": 8608 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698398431930515e-05, + "loss": 1.1337, + "step": 8609 + }, + { + "epoch": 0.64, + "learning_rate": 1.969830120202693e-05, + "loss": 0.9905, + "step": 8610 + }, + { + "epoch": 0.64, + "learning_rate": 1.969820395669349e-05, + "loss": 1.0246, + "step": 8611 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698106695930343e-05, + "loss": 1.0052, + "step": 8612 + }, + { + "epoch": 0.64, + "learning_rate": 1.9698009419737648e-05, + "loss": 1.0618, + "step": 8613 + }, + { + "epoch": 0.64, + "learning_rate": 1.969791212811556e-05, + "loss": 0.9049, + "step": 8614 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697814821064233e-05, + "loss": 0.9726, + "step": 8615 + }, + { + "epoch": 0.64, + "learning_rate": 1.969771749858382e-05, + "loss": 0.9225, + "step": 8616 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697620160674477e-05, + "loss": 1.0531, + "step": 8617 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697522807336358e-05, + "loss": 1.0425, + "step": 8618 + }, + { + "epoch": 0.64, + "learning_rate": 1.969742543856962e-05, + "loss": 1.0945, + "step": 8619 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697328054374418e-05, + "loss": 1.036, + "step": 8620 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697230654750905e-05, + "loss": 1.0769, + "step": 8621 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697133239699237e-05, + "loss": 1.0239, + "step": 8622 + }, + { + "epoch": 0.64, + "learning_rate": 1.9697035809219567e-05, + "loss": 1.1147, + "step": 8623 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696938363312053e-05, + "loss": 1.0098, + "step": 8624 + }, + { + "epoch": 0.64, + "learning_rate": 1.969684090197685e-05, + "loss": 1.0042, + "step": 8625 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696743425214112e-05, + "loss": 1.0611, + "step": 8626 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696645933023993e-05, + "loss": 1.0123, + "step": 8627 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696548425406646e-05, + "loss": 1.0002, + "step": 8628 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696450902362234e-05, + "loss": 1.0091, + "step": 8629 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696353363890905e-05, + "loss": 1.0875, + "step": 8630 + }, + { + "epoch": 0.64, + "learning_rate": 1.969625580999282e-05, + "loss": 1.0444, + "step": 8631 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696158240668123e-05, + "loss": 0.9167, + "step": 8632 + }, + { + "epoch": 0.64, + "learning_rate": 1.9696060655916985e-05, + "loss": 1.1152, + "step": 8633 + }, + { + "epoch": 0.64, + "learning_rate": 1.969596305573955e-05, + "loss": 1.0865, + "step": 8634 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695865440135975e-05, + "loss": 1.0396, + "step": 8635 + }, + { + "epoch": 0.64, + "learning_rate": 1.969576780910642e-05, + "loss": 1.1241, + "step": 8636 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695670162651034e-05, + "loss": 1.0816, + "step": 8637 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695572500769976e-05, + "loss": 0.9176, + "step": 8638 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695474823463402e-05, + "loss": 1.1393, + "step": 8639 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695377130731467e-05, + "loss": 1.1495, + "step": 8640 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695279422574326e-05, + "loss": 1.1267, + "step": 8641 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695181698992132e-05, + "loss": 1.0043, + "step": 8642 + }, + { + "epoch": 0.64, + "learning_rate": 1.9695083959985046e-05, + "loss": 1.1174, + "step": 8643 + }, + { + "epoch": 0.64, + "learning_rate": 1.9694986205553218e-05, + "loss": 1.0661, + "step": 8644 + }, + { + "epoch": 0.64, + "learning_rate": 1.96948884356968e-05, + "loss": 1.0928, + "step": 8645 + }, + { + "epoch": 0.64, + "learning_rate": 1.969479065041596e-05, + "loss": 1.0848, + "step": 8646 + }, + { + "epoch": 0.64, + "learning_rate": 1.9694692849710846e-05, + "loss": 0.9797, + "step": 8647 + }, + { + "epoch": 0.64, + "learning_rate": 1.9694595033581616e-05, + "loss": 1.0775, + "step": 8648 + }, + { + "epoch": 0.64, + "learning_rate": 1.969449720202842e-05, + "loss": 1.1086, + "step": 8649 + }, + { + "epoch": 0.64, + "learning_rate": 1.969439935505142e-05, + "loss": 0.9844, + "step": 8650 + }, + { + "epoch": 0.64, + "learning_rate": 1.969430149265077e-05, + "loss": 1.1253, + "step": 8651 + }, + { + "epoch": 0.64, + "learning_rate": 1.969420361482662e-05, + "loss": 1.1185, + "step": 8652 + }, + { + "epoch": 0.64, + "learning_rate": 1.9694105721579137e-05, + "loss": 1.0105, + "step": 8653 + }, + { + "epoch": 0.64, + "learning_rate": 1.9694007812908467e-05, + "loss": 0.9535, + "step": 8654 + }, + { + "epoch": 0.64, + "learning_rate": 1.969390988881477e-05, + "loss": 1.1418, + "step": 8655 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693811949298204e-05, + "loss": 1.0307, + "step": 8656 + }, + { + "epoch": 0.64, + "learning_rate": 1.969371399435892e-05, + "loss": 1.0218, + "step": 8657 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693616023997075e-05, + "loss": 1.1257, + "step": 8658 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693518038212824e-05, + "loss": 1.067, + "step": 8659 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693420037006326e-05, + "loss": 0.9483, + "step": 8660 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693322020377736e-05, + "loss": 1.0307, + "step": 8661 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693223988327207e-05, + "loss": 1.1454, + "step": 8662 + }, + { + "epoch": 0.64, + "learning_rate": 1.9693125940854902e-05, + "loss": 0.9984, + "step": 8663 + }, + { + "epoch": 0.64, + "learning_rate": 1.969302787796097e-05, + "loss": 1.0151, + "step": 8664 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692929799645565e-05, + "loss": 1.0098, + "step": 8665 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692831705908853e-05, + "loss": 1.1066, + "step": 8666 + }, + { + "epoch": 0.64, + "learning_rate": 1.969273359675098e-05, + "loss": 1.0106, + "step": 8667 + }, + { + "epoch": 0.64, + "learning_rate": 1.969263547217211e-05, + "loss": 1.1474, + "step": 8668 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692537332172396e-05, + "loss": 1.0208, + "step": 8669 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692439176751993e-05, + "loss": 0.9568, + "step": 8670 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692341005911057e-05, + "loss": 1.0386, + "step": 8671 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692242819649743e-05, + "loss": 0.9702, + "step": 8672 + }, + { + "epoch": 0.64, + "learning_rate": 1.9692144617968216e-05, + "loss": 1.0546, + "step": 8673 + }, + { + "epoch": 0.64, + "learning_rate": 1.969204640086662e-05, + "loss": 1.0445, + "step": 8674 + }, + { + "epoch": 0.64, + "learning_rate": 1.969194816834512e-05, + "loss": 1.1408, + "step": 8675 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691849920403867e-05, + "loss": 1.1612, + "step": 8676 + }, + { + "epoch": 0.64, + "learning_rate": 1.969175165704302e-05, + "loss": 1.0226, + "step": 8677 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691653378262733e-05, + "loss": 0.929, + "step": 8678 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691555084063165e-05, + "loss": 0.9742, + "step": 8679 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691456774444476e-05, + "loss": 1.086, + "step": 8680 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691358449406813e-05, + "loss": 1.1203, + "step": 8681 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691260108950337e-05, + "loss": 1.0262, + "step": 8682 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691161753075206e-05, + "loss": 1.0757, + "step": 8683 + }, + { + "epoch": 0.64, + "learning_rate": 1.9691063381781575e-05, + "loss": 1.0352, + "step": 8684 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690964995069603e-05, + "loss": 1.0126, + "step": 8685 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690866592939443e-05, + "loss": 0.989, + "step": 8686 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690768175391254e-05, + "loss": 1.0507, + "step": 8687 + }, + { + "epoch": 0.64, + "learning_rate": 1.969066974242519e-05, + "loss": 0.9474, + "step": 8688 + }, + { + "epoch": 0.64, + "learning_rate": 1.969057129404141e-05, + "loss": 1.0137, + "step": 8689 + }, + { + "epoch": 0.64, + "learning_rate": 1.969047283024007e-05, + "loss": 0.9655, + "step": 8690 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690374351021325e-05, + "loss": 1.074, + "step": 8691 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690275856385335e-05, + "loss": 1.0076, + "step": 8692 + }, + { + "epoch": 0.64, + "learning_rate": 1.969017734633225e-05, + "loss": 0.8897, + "step": 8693 + }, + { + "epoch": 0.64, + "learning_rate": 1.9690078820862238e-05, + "loss": 1.1014, + "step": 8694 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689980279975446e-05, + "loss": 1.0983, + "step": 8695 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689881723672032e-05, + "loss": 0.9416, + "step": 8696 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689783151952157e-05, + "loss": 1.066, + "step": 8697 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689684564815977e-05, + "loss": 1.0227, + "step": 8698 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689585962263646e-05, + "loss": 1.0485, + "step": 8699 + }, + { + "epoch": 0.64, + "learning_rate": 1.968948734429532e-05, + "loss": 1.0503, + "step": 8700 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689388710911163e-05, + "loss": 0.9596, + "step": 8701 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689290062111327e-05, + "loss": 0.9872, + "step": 8702 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689191397895966e-05, + "loss": 0.9913, + "step": 8703 + }, + { + "epoch": 0.64, + "learning_rate": 1.9689092718265243e-05, + "loss": 0.9732, + "step": 8704 + }, + { + "epoch": 0.64, + "learning_rate": 1.968899402321931e-05, + "loss": 1.0101, + "step": 8705 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688895312758324e-05, + "loss": 1.1068, + "step": 8706 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688796586882447e-05, + "loss": 1.0197, + "step": 8707 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688697845591834e-05, + "loss": 1.1005, + "step": 8708 + }, + { + "epoch": 0.64, + "learning_rate": 1.968859908888664e-05, + "loss": 0.9813, + "step": 8709 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688500316767022e-05, + "loss": 1.009, + "step": 8710 + }, + { + "epoch": 0.64, + "learning_rate": 1.968840152923314e-05, + "loss": 1.1902, + "step": 8711 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688302726285153e-05, + "loss": 0.9974, + "step": 8712 + }, + { + "epoch": 0.64, + "learning_rate": 1.968820390792321e-05, + "loss": 1.047, + "step": 8713 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688105074147478e-05, + "loss": 1.0188, + "step": 8714 + }, + { + "epoch": 0.64, + "learning_rate": 1.9688006224958105e-05, + "loss": 1.0168, + "step": 8715 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687907360355254e-05, + "loss": 1.1086, + "step": 8716 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687808480339083e-05, + "loss": 0.9706, + "step": 8717 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687709584909744e-05, + "loss": 1.0469, + "step": 8718 + }, + { + "epoch": 0.64, + "learning_rate": 1.96876106740674e-05, + "loss": 1.0047, + "step": 8719 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687511747812203e-05, + "loss": 0.9759, + "step": 8720 + }, + { + "epoch": 0.64, + "learning_rate": 1.968741280614432e-05, + "loss": 1.0653, + "step": 8721 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687313849063894e-05, + "loss": 1.103, + "step": 8722 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687214876571095e-05, + "loss": 0.9879, + "step": 8723 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687115888666074e-05, + "loss": 1.0776, + "step": 8724 + }, + { + "epoch": 0.64, + "learning_rate": 1.9687016885348994e-05, + "loss": 1.0864, + "step": 8725 + }, + { + "epoch": 0.64, + "learning_rate": 1.9686917866620004e-05, + "loss": 1.0841, + "step": 8726 + }, + { + "epoch": 0.64, + "learning_rate": 1.9686818832479266e-05, + "loss": 1.1448, + "step": 8727 + }, + { + "epoch": 0.64, + "learning_rate": 1.968671978292694e-05, + "loss": 1.0289, + "step": 8728 + }, + { + "epoch": 0.64, + "learning_rate": 1.968662071796318e-05, + "loss": 1.0702, + "step": 8729 + }, + { + "epoch": 0.65, + "learning_rate": 1.9686521637588147e-05, + "loss": 1.0709, + "step": 8730 + }, + { + "epoch": 0.65, + "learning_rate": 1.9686422541801997e-05, + "loss": 1.1361, + "step": 8731 + }, + { + "epoch": 0.65, + "learning_rate": 1.9686323430604887e-05, + "loss": 1.0022, + "step": 8732 + }, + { + "epoch": 0.65, + "learning_rate": 1.9686224303996974e-05, + "loss": 1.112, + "step": 8733 + }, + { + "epoch": 0.65, + "learning_rate": 1.968612516197842e-05, + "loss": 0.9886, + "step": 8734 + }, + { + "epoch": 0.65, + "learning_rate": 1.9686026004549372e-05, + "loss": 1.069, + "step": 8735 + }, + { + "epoch": 0.65, + "learning_rate": 1.968592683171e-05, + "loss": 0.8849, + "step": 8736 + }, + { + "epoch": 0.65, + "learning_rate": 1.9685827643460458e-05, + "loss": 0.9238, + "step": 8737 + }, + { + "epoch": 0.65, + "learning_rate": 1.96857284398009e-05, + "loss": 1.0851, + "step": 8738 + }, + { + "epoch": 0.65, + "learning_rate": 1.968562922073149e-05, + "loss": 1.0148, + "step": 8739 + }, + { + "epoch": 0.65, + "learning_rate": 1.968552998625238e-05, + "loss": 1.0205, + "step": 8740 + }, + { + "epoch": 0.65, + "learning_rate": 1.968543073636373e-05, + "loss": 1.0463, + "step": 8741 + }, + { + "epoch": 0.65, + "learning_rate": 1.96853314710657e-05, + "loss": 1.0455, + "step": 8742 + }, + { + "epoch": 0.65, + "learning_rate": 1.9685232190358444e-05, + "loss": 1.0881, + "step": 8743 + }, + { + "epoch": 0.65, + "learning_rate": 1.9685132894242125e-05, + "loss": 1.0406, + "step": 8744 + }, + { + "epoch": 0.65, + "learning_rate": 1.9685033582716894e-05, + "loss": 1.0473, + "step": 8745 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684934255782918e-05, + "loss": 1.0004, + "step": 8746 + }, + { + "epoch": 0.65, + "learning_rate": 1.968483491344035e-05, + "loss": 1.0886, + "step": 8747 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684735555689345e-05, + "loss": 1.0972, + "step": 8748 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684636182530067e-05, + "loss": 1.1023, + "step": 8749 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684536793962667e-05, + "loss": 1.0865, + "step": 8750 + }, + { + "epoch": 0.65, + "learning_rate": 1.968443738998731e-05, + "loss": 1.0999, + "step": 8751 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684337970604153e-05, + "loss": 1.1404, + "step": 8752 + }, + { + "epoch": 0.65, + "learning_rate": 1.968423853581335e-05, + "loss": 1.0587, + "step": 8753 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684139085615065e-05, + "loss": 0.95, + "step": 8754 + }, + { + "epoch": 0.65, + "learning_rate": 1.9684039620009452e-05, + "loss": 0.9251, + "step": 8755 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683940138996673e-05, + "loss": 0.999, + "step": 8756 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683840642576882e-05, + "loss": 1.0299, + "step": 8757 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683741130750237e-05, + "loss": 0.9948, + "step": 8758 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683641603516902e-05, + "loss": 1.0228, + "step": 8759 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683542060877027e-05, + "loss": 1.0771, + "step": 8760 + }, + { + "epoch": 0.65, + "learning_rate": 1.968344250283078e-05, + "loss": 1.0082, + "step": 8761 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683342929378312e-05, + "loss": 1.0677, + "step": 8762 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683243340519783e-05, + "loss": 1.0114, + "step": 8763 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683143736255355e-05, + "loss": 1.0891, + "step": 8764 + }, + { + "epoch": 0.65, + "learning_rate": 1.9683044116585183e-05, + "loss": 1.0159, + "step": 8765 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682944481509426e-05, + "loss": 1.1042, + "step": 8766 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682844831028244e-05, + "loss": 1.0218, + "step": 8767 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682745165141792e-05, + "loss": 1.0891, + "step": 8768 + }, + { + "epoch": 0.65, + "learning_rate": 1.968264548385023e-05, + "loss": 1.0296, + "step": 8769 + }, + { + "epoch": 0.65, + "learning_rate": 1.968254578715372e-05, + "loss": 1.0308, + "step": 8770 + }, + { + "epoch": 0.65, + "learning_rate": 1.968244607505242e-05, + "loss": 1.0018, + "step": 8771 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682346347546484e-05, + "loss": 1.0632, + "step": 8772 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682246604636074e-05, + "loss": 1.1535, + "step": 8773 + }, + { + "epoch": 0.65, + "learning_rate": 1.968214684632135e-05, + "loss": 0.9463, + "step": 8774 + }, + { + "epoch": 0.65, + "learning_rate": 1.9682047072602464e-05, + "loss": 1.0613, + "step": 8775 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681947283479587e-05, + "loss": 1.1255, + "step": 8776 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681847478952866e-05, + "loss": 1.0893, + "step": 8777 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681747659022465e-05, + "loss": 1.0545, + "step": 8778 + }, + { + "epoch": 0.65, + "learning_rate": 1.968164782368854e-05, + "loss": 1.0241, + "step": 8779 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681547972951252e-05, + "loss": 1.0114, + "step": 8780 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681448106810762e-05, + "loss": 1.0276, + "step": 8781 + }, + { + "epoch": 0.65, + "learning_rate": 1.968134822526723e-05, + "loss": 1.0304, + "step": 8782 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681248328320804e-05, + "loss": 0.9861, + "step": 8783 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681148415971655e-05, + "loss": 0.9135, + "step": 8784 + }, + { + "epoch": 0.65, + "learning_rate": 1.9681048488219936e-05, + "loss": 1.2168, + "step": 8785 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680948545065804e-05, + "loss": 1.0251, + "step": 8786 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680848586509425e-05, + "loss": 1.2081, + "step": 8787 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680748612550955e-05, + "loss": 1.0799, + "step": 8788 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680648623190553e-05, + "loss": 0.9409, + "step": 8789 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680548618428377e-05, + "loss": 0.9355, + "step": 8790 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680448598264585e-05, + "loss": 1.0887, + "step": 8791 + }, + { + "epoch": 0.65, + "learning_rate": 1.968034856269934e-05, + "loss": 1.0494, + "step": 8792 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680248511732797e-05, + "loss": 1.0343, + "step": 8793 + }, + { + "epoch": 0.65, + "learning_rate": 1.9680148445365115e-05, + "loss": 1.0429, + "step": 8794 + }, + { + "epoch": 0.65, + "learning_rate": 1.968004836359646e-05, + "loss": 1.0308, + "step": 8795 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679948266426986e-05, + "loss": 0.97, + "step": 8796 + }, + { + "epoch": 0.65, + "learning_rate": 1.967984815385685e-05, + "loss": 1.1596, + "step": 8797 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679748025886213e-05, + "loss": 1.1056, + "step": 8798 + }, + { + "epoch": 0.65, + "learning_rate": 1.967964788251524e-05, + "loss": 1.0219, + "step": 8799 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679547723744082e-05, + "loss": 1.0763, + "step": 8800 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679447549572905e-05, + "loss": 0.9899, + "step": 8801 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679347360001862e-05, + "loss": 0.9679, + "step": 8802 + }, + { + "epoch": 0.65, + "learning_rate": 1.967924715503112e-05, + "loss": 1.0942, + "step": 8803 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679146934660832e-05, + "loss": 1.0357, + "step": 8804 + }, + { + "epoch": 0.65, + "learning_rate": 1.9679046698891162e-05, + "loss": 1.1446, + "step": 8805 + }, + { + "epoch": 0.65, + "learning_rate": 1.967894644772226e-05, + "loss": 1.1157, + "step": 8806 + }, + { + "epoch": 0.65, + "learning_rate": 1.9678846181154304e-05, + "loss": 1.0916, + "step": 8807 + }, + { + "epoch": 0.65, + "learning_rate": 1.9678745899187434e-05, + "loss": 1.1859, + "step": 8808 + }, + { + "epoch": 0.65, + "learning_rate": 1.967864560182182e-05, + "loss": 0.943, + "step": 8809 + }, + { + "epoch": 0.65, + "learning_rate": 1.967854528905762e-05, + "loss": 1.0448, + "step": 8810 + }, + { + "epoch": 0.65, + "learning_rate": 1.967844496089499e-05, + "loss": 0.9776, + "step": 8811 + }, + { + "epoch": 0.65, + "learning_rate": 1.9678344617334097e-05, + "loss": 1.0743, + "step": 8812 + }, + { + "epoch": 0.65, + "learning_rate": 1.9678244258375093e-05, + "loss": 1.0505, + "step": 8813 + }, + { + "epoch": 0.65, + "learning_rate": 1.967814388401814e-05, + "loss": 0.9983, + "step": 8814 + }, + { + "epoch": 0.65, + "learning_rate": 1.96780434942634e-05, + "loss": 0.9754, + "step": 8815 + }, + { + "epoch": 0.65, + "learning_rate": 1.967794308911103e-05, + "loss": 1.0974, + "step": 8816 + }, + { + "epoch": 0.65, + "learning_rate": 1.9677842668561196e-05, + "loss": 1.0455, + "step": 8817 + }, + { + "epoch": 0.65, + "learning_rate": 1.967774223261405e-05, + "loss": 1.1278, + "step": 8818 + }, + { + "epoch": 0.65, + "learning_rate": 1.9677641781269748e-05, + "loss": 1.1441, + "step": 8819 + }, + { + "epoch": 0.65, + "learning_rate": 1.9677541314528464e-05, + "loss": 1.0405, + "step": 8820 + }, + { + "epoch": 0.65, + "learning_rate": 1.967744083239035e-05, + "loss": 1.0353, + "step": 8821 + }, + { + "epoch": 0.65, + "learning_rate": 1.9677340334855565e-05, + "loss": 1.0251, + "step": 8822 + }, + { + "epoch": 0.65, + "learning_rate": 1.967723982192427e-05, + "loss": 1.1607, + "step": 8823 + }, + { + "epoch": 0.65, + "learning_rate": 1.9677139293596624e-05, + "loss": 1.0755, + "step": 8824 + }, + { + "epoch": 0.65, + "learning_rate": 1.967703874987279e-05, + "loss": 1.0841, + "step": 8825 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676938190752923e-05, + "loss": 1.0163, + "step": 8826 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676837616237186e-05, + "loss": 1.0992, + "step": 8827 + }, + { + "epoch": 0.65, + "learning_rate": 1.967673702632574e-05, + "loss": 0.975, + "step": 8828 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676636421018742e-05, + "loss": 1.0679, + "step": 8829 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676535800316356e-05, + "loss": 0.9759, + "step": 8830 + }, + { + "epoch": 0.65, + "learning_rate": 1.967643516421874e-05, + "loss": 1.1846, + "step": 8831 + }, + { + "epoch": 0.65, + "learning_rate": 1.967633451272605e-05, + "loss": 1.0684, + "step": 8832 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676233845838455e-05, + "loss": 0.8993, + "step": 8833 + }, + { + "epoch": 0.65, + "learning_rate": 1.967613316355611e-05, + "loss": 1.0689, + "step": 8834 + }, + { + "epoch": 0.65, + "learning_rate": 1.9676032465879173e-05, + "loss": 1.0504, + "step": 8835 + }, + { + "epoch": 0.65, + "learning_rate": 1.967593175280781e-05, + "loss": 0.9703, + "step": 8836 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675831024342176e-05, + "loss": 1.0817, + "step": 8837 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675730280482435e-05, + "loss": 1.0451, + "step": 8838 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675629521228745e-05, + "loss": 1.1119, + "step": 8839 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675528746581265e-05, + "loss": 0.9166, + "step": 8840 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675427956540157e-05, + "loss": 1.1187, + "step": 8841 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675327151105584e-05, + "loss": 1.1414, + "step": 8842 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675226330277705e-05, + "loss": 1.0161, + "step": 8843 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675125494056675e-05, + "loss": 0.9497, + "step": 8844 + }, + { + "epoch": 0.65, + "learning_rate": 1.9675024642442662e-05, + "loss": 0.8981, + "step": 8845 + }, + { + "epoch": 0.65, + "learning_rate": 1.967492377543582e-05, + "loss": 1.0872, + "step": 8846 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674822893036317e-05, + "loss": 1.0232, + "step": 8847 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674721995244306e-05, + "loss": 1.114, + "step": 8848 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674621082059953e-05, + "loss": 1.0519, + "step": 8849 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674520153483414e-05, + "loss": 0.9661, + "step": 8850 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674419209514854e-05, + "loss": 1.0576, + "step": 8851 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674318250154428e-05, + "loss": 1.0273, + "step": 8852 + }, + { + "epoch": 0.65, + "learning_rate": 1.9674217275402302e-05, + "loss": 0.9305, + "step": 8853 + }, + { + "epoch": 0.65, + "learning_rate": 1.967411628525864e-05, + "loss": 1.0227, + "step": 8854 + }, + { + "epoch": 0.65, + "learning_rate": 1.967401527972359e-05, + "loss": 1.0606, + "step": 8855 + }, + { + "epoch": 0.65, + "learning_rate": 1.9673914258797324e-05, + "loss": 1.0937, + "step": 8856 + }, + { + "epoch": 0.65, + "learning_rate": 1.9673813222479997e-05, + "loss": 0.8799, + "step": 8857 + }, + { + "epoch": 0.65, + "learning_rate": 1.9673712170771773e-05, + "loss": 1.024, + "step": 8858 + }, + { + "epoch": 0.65, + "learning_rate": 1.967361110367281e-05, + "loss": 0.9757, + "step": 8859 + }, + { + "epoch": 0.65, + "learning_rate": 1.967351002118327e-05, + "loss": 0.9349, + "step": 8860 + }, + { + "epoch": 0.65, + "learning_rate": 1.9673408923303313e-05, + "loss": 0.9923, + "step": 8861 + }, + { + "epoch": 0.65, + "learning_rate": 1.96733078100331e-05, + "loss": 1.02, + "step": 8862 + }, + { + "epoch": 0.65, + "learning_rate": 1.9673206681372798e-05, + "loss": 1.0726, + "step": 8863 + }, + { + "epoch": 0.65, + "learning_rate": 1.967310553732256e-05, + "loss": 0.9503, + "step": 8864 + }, + { + "epoch": 0.66, + "learning_rate": 1.967300437788255e-05, + "loss": 1.026, + "step": 8865 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672903203052928e-05, + "loss": 1.0162, + "step": 8866 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672802012833855e-05, + "loss": 1.0409, + "step": 8867 + }, + { + "epoch": 0.66, + "learning_rate": 1.967270080722549e-05, + "loss": 0.9999, + "step": 8868 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672599586228e-05, + "loss": 1.1574, + "step": 8869 + }, + { + "epoch": 0.66, + "learning_rate": 1.967249834984154e-05, + "loss": 0.9202, + "step": 8870 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672397098066276e-05, + "loss": 0.9739, + "step": 8871 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672295830902365e-05, + "loss": 1.0459, + "step": 8872 + }, + { + "epoch": 0.66, + "learning_rate": 1.967219454834997e-05, + "loss": 1.0209, + "step": 8873 + }, + { + "epoch": 0.66, + "learning_rate": 1.9672093250409253e-05, + "loss": 1.0186, + "step": 8874 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671991937080372e-05, + "loss": 1.0465, + "step": 8875 + }, + { + "epoch": 0.66, + "learning_rate": 1.967189060836349e-05, + "loss": 1.0815, + "step": 8876 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671789264258772e-05, + "loss": 1.0777, + "step": 8877 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671687904766372e-05, + "loss": 1.0888, + "step": 8878 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671586529886456e-05, + "loss": 1.1003, + "step": 8879 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671485139619184e-05, + "loss": 1.0333, + "step": 8880 + }, + { + "epoch": 0.66, + "learning_rate": 1.967138373396472e-05, + "loss": 0.9573, + "step": 8881 + }, + { + "epoch": 0.66, + "learning_rate": 1.967128231292322e-05, + "loss": 1.0157, + "step": 8882 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671180876494848e-05, + "loss": 1.0323, + "step": 8883 + }, + { + "epoch": 0.66, + "learning_rate": 1.9671079424679766e-05, + "loss": 1.0166, + "step": 8884 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670977957478135e-05, + "loss": 1.0547, + "step": 8885 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670876474890116e-05, + "loss": 1.0334, + "step": 8886 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670774976915875e-05, + "loss": 1.0231, + "step": 8887 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670673463555566e-05, + "loss": 0.9716, + "step": 8888 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670571934809354e-05, + "loss": 1.0641, + "step": 8889 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670470390677398e-05, + "loss": 1.0478, + "step": 8890 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670368831159865e-05, + "loss": 1.0768, + "step": 8891 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670267256256913e-05, + "loss": 1.0982, + "step": 8892 + }, + { + "epoch": 0.66, + "learning_rate": 1.9670165665968705e-05, + "loss": 1.0535, + "step": 8893 + }, + { + "epoch": 0.66, + "learning_rate": 1.96700640602954e-05, + "loss": 1.0033, + "step": 8894 + }, + { + "epoch": 0.66, + "learning_rate": 1.9669962439237164e-05, + "loss": 1.0194, + "step": 8895 + }, + { + "epoch": 0.66, + "learning_rate": 1.9669860802794155e-05, + "loss": 1.0688, + "step": 8896 + }, + { + "epoch": 0.66, + "learning_rate": 1.9669759150966534e-05, + "loss": 1.0251, + "step": 8897 + }, + { + "epoch": 0.66, + "learning_rate": 1.9669657483754467e-05, + "loss": 1.0721, + "step": 8898 + }, + { + "epoch": 0.66, + "learning_rate": 1.966955580115811e-05, + "loss": 1.0961, + "step": 8899 + }, + { + "epoch": 0.66, + "learning_rate": 1.966945410317763e-05, + "loss": 1.0564, + "step": 8900 + }, + { + "epoch": 0.66, + "learning_rate": 1.966935238981319e-05, + "loss": 0.9319, + "step": 8901 + }, + { + "epoch": 0.66, + "learning_rate": 1.9669250661064944e-05, + "loss": 1.0032, + "step": 8902 + }, + { + "epoch": 0.66, + "learning_rate": 1.966914891693306e-05, + "loss": 0.9384, + "step": 8903 + }, + { + "epoch": 0.66, + "learning_rate": 1.96690471574177e-05, + "loss": 1.0323, + "step": 8904 + }, + { + "epoch": 0.66, + "learning_rate": 1.9668945382519022e-05, + "loss": 0.922, + "step": 8905 + }, + { + "epoch": 0.66, + "learning_rate": 1.966884359223719e-05, + "loss": 0.9325, + "step": 8906 + }, + { + "epoch": 0.66, + "learning_rate": 1.9668741786572368e-05, + "loss": 1.0589, + "step": 8907 + }, + { + "epoch": 0.66, + "learning_rate": 1.966863996552472e-05, + "loss": 1.0758, + "step": 8908 + }, + { + "epoch": 0.66, + "learning_rate": 1.9668538129094398e-05, + "loss": 0.9384, + "step": 8909 + }, + { + "epoch": 0.66, + "learning_rate": 1.966843627728157e-05, + "loss": 1.0137, + "step": 8910 + }, + { + "epoch": 0.66, + "learning_rate": 1.9668334410086403e-05, + "loss": 1.0355, + "step": 8911 + }, + { + "epoch": 0.66, + "learning_rate": 1.966823252750905e-05, + "loss": 1.076, + "step": 8912 + }, + { + "epoch": 0.66, + "learning_rate": 1.966813062954968e-05, + "loss": 1.0231, + "step": 8913 + }, + { + "epoch": 0.66, + "learning_rate": 1.9668028716208453e-05, + "loss": 1.151, + "step": 8914 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667926787485528e-05, + "loss": 1.0314, + "step": 8915 + }, + { + "epoch": 0.66, + "learning_rate": 1.966782484338107e-05, + "loss": 1.0528, + "step": 8916 + }, + { + "epoch": 0.66, + "learning_rate": 1.966772288389524e-05, + "loss": 1.0973, + "step": 8917 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667620909028206e-05, + "loss": 0.9903, + "step": 8918 + }, + { + "epoch": 0.66, + "learning_rate": 1.966751891878012e-05, + "loss": 1.1028, + "step": 8919 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667416913151155e-05, + "loss": 1.0261, + "step": 8920 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667314892141465e-05, + "loss": 1.004, + "step": 8921 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667212855751213e-05, + "loss": 1.0729, + "step": 8922 + }, + { + "epoch": 0.66, + "learning_rate": 1.966711080398057e-05, + "loss": 0.9147, + "step": 8923 + }, + { + "epoch": 0.66, + "learning_rate": 1.9667008736829687e-05, + "loss": 1.0456, + "step": 8924 + }, + { + "epoch": 0.66, + "learning_rate": 1.966690665429873e-05, + "loss": 0.9811, + "step": 8925 + }, + { + "epoch": 0.66, + "learning_rate": 1.9666804556387866e-05, + "loss": 0.9717, + "step": 8926 + }, + { + "epoch": 0.66, + "learning_rate": 1.9666702443097258e-05, + "loss": 0.9637, + "step": 8927 + }, + { + "epoch": 0.66, + "learning_rate": 1.966660031442706e-05, + "loss": 1.0524, + "step": 8928 + }, + { + "epoch": 0.66, + "learning_rate": 1.966649817037744e-05, + "loss": 0.9986, + "step": 8929 + }, + { + "epoch": 0.66, + "learning_rate": 1.9666396010948556e-05, + "loss": 1.0451, + "step": 8930 + }, + { + "epoch": 0.66, + "learning_rate": 1.9666293836140577e-05, + "loss": 1.0562, + "step": 8931 + }, + { + "epoch": 0.66, + "learning_rate": 1.9666191645953665e-05, + "loss": 1.0131, + "step": 8932 + }, + { + "epoch": 0.66, + "learning_rate": 1.966608944038798e-05, + "loss": 0.9646, + "step": 8933 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665987219443684e-05, + "loss": 1.0973, + "step": 8934 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665884983120937e-05, + "loss": 1.0266, + "step": 8935 + }, + { + "epoch": 0.66, + "learning_rate": 1.966578273141991e-05, + "loss": 1.0559, + "step": 8936 + }, + { + "epoch": 0.66, + "learning_rate": 1.966568046434076e-05, + "loss": 0.9757, + "step": 8937 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665578181883652e-05, + "loss": 0.8776, + "step": 8938 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665475884048743e-05, + "loss": 1.0081, + "step": 8939 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665373570836204e-05, + "loss": 1.0924, + "step": 8940 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665271242246192e-05, + "loss": 1.0644, + "step": 8941 + }, + { + "epoch": 0.66, + "learning_rate": 1.9665168898278876e-05, + "loss": 1.0745, + "step": 8942 + }, + { + "epoch": 0.66, + "learning_rate": 1.966506653893441e-05, + "loss": 0.9996, + "step": 8943 + }, + { + "epoch": 0.66, + "learning_rate": 1.966496416421296e-05, + "loss": 1.0826, + "step": 8944 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664861774114695e-05, + "loss": 1.0173, + "step": 8945 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664759368639773e-05, + "loss": 1.0617, + "step": 8946 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664656947788354e-05, + "loss": 1.1007, + "step": 8947 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664554511560604e-05, + "loss": 1.0206, + "step": 8948 + }, + { + "epoch": 0.66, + "learning_rate": 1.966445205995669e-05, + "loss": 1.0846, + "step": 8949 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664349592976766e-05, + "loss": 1.0358, + "step": 8950 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664247110621e-05, + "loss": 0.9959, + "step": 8951 + }, + { + "epoch": 0.66, + "learning_rate": 1.9664144612889558e-05, + "loss": 1.1171, + "step": 8952 + }, + { + "epoch": 0.66, + "learning_rate": 1.96640420997826e-05, + "loss": 0.9885, + "step": 8953 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663939571300294e-05, + "loss": 0.9585, + "step": 8954 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663837027442788e-05, + "loss": 0.9656, + "step": 8955 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663734468210264e-05, + "loss": 1.0886, + "step": 8956 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663631893602874e-05, + "loss": 1.1232, + "step": 8957 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663529303620784e-05, + "loss": 1.0334, + "step": 8958 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663426698264153e-05, + "loss": 1.1793, + "step": 8959 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663324077533155e-05, + "loss": 0.9911, + "step": 8960 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663221441427943e-05, + "loss": 1.0178, + "step": 8961 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663118789948684e-05, + "loss": 1.1435, + "step": 8962 + }, + { + "epoch": 0.66, + "learning_rate": 1.9663016123095543e-05, + "loss": 1.0116, + "step": 8963 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662913440868682e-05, + "loss": 1.045, + "step": 8964 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662810743268262e-05, + "loss": 0.9805, + "step": 8965 + }, + { + "epoch": 0.66, + "learning_rate": 1.966270803029445e-05, + "loss": 1.0415, + "step": 8966 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662605301947406e-05, + "loss": 1.157, + "step": 8967 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662502558227295e-05, + "loss": 1.062, + "step": 8968 + }, + { + "epoch": 0.66, + "learning_rate": 1.966239979913428e-05, + "loss": 1.0964, + "step": 8969 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662297024668532e-05, + "loss": 1.0695, + "step": 8970 + }, + { + "epoch": 0.66, + "learning_rate": 1.96621942348302e-05, + "loss": 0.9527, + "step": 8971 + }, + { + "epoch": 0.66, + "learning_rate": 1.9662091429619456e-05, + "loss": 1.2111, + "step": 8972 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661988609036465e-05, + "loss": 0.9598, + "step": 8973 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661885773081392e-05, + "loss": 1.0437, + "step": 8974 + }, + { + "epoch": 0.66, + "learning_rate": 1.966178292175439e-05, + "loss": 1.022, + "step": 8975 + }, + { + "epoch": 0.66, + "learning_rate": 1.966168005505563e-05, + "loss": 1.0307, + "step": 8976 + }, + { + "epoch": 0.66, + "learning_rate": 1.966157717298528e-05, + "loss": 1.0204, + "step": 8977 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661474275543498e-05, + "loss": 1.1047, + "step": 8978 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661371362730447e-05, + "loss": 1.0032, + "step": 8979 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661268434546293e-05, + "loss": 1.0264, + "step": 8980 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661165490991198e-05, + "loss": 1.1105, + "step": 8981 + }, + { + "epoch": 0.66, + "learning_rate": 1.9661062532065328e-05, + "loss": 1.0796, + "step": 8982 + }, + { + "epoch": 0.66, + "learning_rate": 1.9660959557768846e-05, + "loss": 0.994, + "step": 8983 + }, + { + "epoch": 0.66, + "learning_rate": 1.9660856568101914e-05, + "loss": 1.0456, + "step": 8984 + }, + { + "epoch": 0.66, + "learning_rate": 1.96607535630647e-05, + "loss": 0.9742, + "step": 8985 + }, + { + "epoch": 0.66, + "learning_rate": 1.9660650542657363e-05, + "loss": 1.0571, + "step": 8986 + }, + { + "epoch": 0.66, + "learning_rate": 1.966054750688007e-05, + "loss": 1.0099, + "step": 8987 + }, + { + "epoch": 0.66, + "learning_rate": 1.9660444455732985e-05, + "loss": 0.993, + "step": 8988 + }, + { + "epoch": 0.66, + "learning_rate": 1.966034138921627e-05, + "loss": 1.059, + "step": 8989 + }, + { + "epoch": 0.66, + "learning_rate": 1.966023830733009e-05, + "loss": 0.9086, + "step": 8990 + }, + { + "epoch": 0.66, + "learning_rate": 1.966013521007461e-05, + "loss": 1.0006, + "step": 8991 + }, + { + "epoch": 0.66, + "learning_rate": 1.966003209744999e-05, + "loss": 0.9929, + "step": 8992 + }, + { + "epoch": 0.66, + "learning_rate": 1.9659928969456402e-05, + "loss": 1.0895, + "step": 8993 + }, + { + "epoch": 0.66, + "learning_rate": 1.9659825826094004e-05, + "loss": 0.9566, + "step": 8994 + }, + { + "epoch": 0.66, + "learning_rate": 1.9659722667362955e-05, + "loss": 1.0497, + "step": 8995 + }, + { + "epoch": 0.66, + "learning_rate": 1.9659619493263436e-05, + "loss": 1.0358, + "step": 8996 + }, + { + "epoch": 0.66, + "learning_rate": 1.9659516303795594e-05, + "loss": 1.058, + "step": 8997 + }, + { + "epoch": 0.66, + "learning_rate": 1.96594130989596e-05, + "loss": 0.9961, + "step": 8998 + }, + { + "epoch": 0.66, + "learning_rate": 1.965930987875562e-05, + "loss": 1.071, + "step": 8999 + }, + { + "epoch": 0.67, + "learning_rate": 1.9659206643183816e-05, + "loss": 0.9568, + "step": 9000 + }, + { + "epoch": 0.67, + "learning_rate": 1.965910339224435e-05, + "loss": 0.9946, + "step": 9001 + }, + { + "epoch": 0.67, + "learning_rate": 1.965900012593739e-05, + "loss": 0.9739, + "step": 9002 + }, + { + "epoch": 0.67, + "learning_rate": 1.96588968442631e-05, + "loss": 1.0596, + "step": 9003 + }, + { + "epoch": 0.67, + "learning_rate": 1.965879354722164e-05, + "loss": 1.0053, + "step": 9004 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658690234813184e-05, + "loss": 1.0371, + "step": 9005 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658586907037883e-05, + "loss": 1.0039, + "step": 9006 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658483563895913e-05, + "loss": 1.0209, + "step": 9007 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658380205387434e-05, + "loss": 1.0413, + "step": 9008 + }, + { + "epoch": 0.67, + "learning_rate": 1.965827683151261e-05, + "loss": 1.0999, + "step": 9009 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658173442271603e-05, + "loss": 1.0877, + "step": 9010 + }, + { + "epoch": 0.67, + "learning_rate": 1.9658070037664582e-05, + "loss": 1.0467, + "step": 9011 + }, + { + "epoch": 0.67, + "learning_rate": 1.965796661769171e-05, + "loss": 1.0157, + "step": 9012 + }, + { + "epoch": 0.67, + "learning_rate": 1.9657863182353148e-05, + "loss": 1.0271, + "step": 9013 + }, + { + "epoch": 0.67, + "learning_rate": 1.965775973164907e-05, + "loss": 1.0244, + "step": 9014 + }, + { + "epoch": 0.67, + "learning_rate": 1.965765626557963e-05, + "loss": 1.0235, + "step": 9015 + }, + { + "epoch": 0.67, + "learning_rate": 1.9657552784144998e-05, + "loss": 1.0597, + "step": 9016 + }, + { + "epoch": 0.67, + "learning_rate": 1.965744928734534e-05, + "loss": 1.053, + "step": 9017 + }, + { + "epoch": 0.67, + "learning_rate": 1.9657345775180815e-05, + "loss": 0.9568, + "step": 9018 + }, + { + "epoch": 0.67, + "learning_rate": 1.965724224765159e-05, + "loss": 1.0476, + "step": 9019 + }, + { + "epoch": 0.67, + "learning_rate": 1.9657138704757835e-05, + "loss": 0.9803, + "step": 9020 + }, + { + "epoch": 0.67, + "learning_rate": 1.965703514649971e-05, + "loss": 1.0324, + "step": 9021 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656931572877374e-05, + "loss": 1.0002, + "step": 9022 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656827983891003e-05, + "loss": 1.0921, + "step": 9023 + }, + { + "epoch": 0.67, + "learning_rate": 1.965672437954076e-05, + "loss": 1.0645, + "step": 9024 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656620759826805e-05, + "loss": 0.9935, + "step": 9025 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656517124749302e-05, + "loss": 1.1256, + "step": 9026 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656413474308417e-05, + "loss": 1.0107, + "step": 9027 + }, + { + "epoch": 0.67, + "learning_rate": 1.965630980850432e-05, + "loss": 1.0589, + "step": 9028 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656206127337173e-05, + "loss": 1.119, + "step": 9029 + }, + { + "epoch": 0.67, + "learning_rate": 1.9656102430807135e-05, + "loss": 1.0766, + "step": 9030 + }, + { + "epoch": 0.67, + "learning_rate": 1.965599871891438e-05, + "loss": 0.9722, + "step": 9031 + }, + { + "epoch": 0.67, + "learning_rate": 1.965589499165907e-05, + "loss": 0.9899, + "step": 9032 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655791249041366e-05, + "loss": 1.0246, + "step": 9033 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655687491061435e-05, + "loss": 0.9886, + "step": 9034 + }, + { + "epoch": 0.67, + "learning_rate": 1.965558371771945e-05, + "loss": 1.0157, + "step": 9035 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655479929015566e-05, + "loss": 1.0934, + "step": 9036 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655376124949953e-05, + "loss": 1.0398, + "step": 9037 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655272305522773e-05, + "loss": 0.9925, + "step": 9038 + }, + { + "epoch": 0.67, + "learning_rate": 1.965516847073419e-05, + "loss": 1.0698, + "step": 9039 + }, + { + "epoch": 0.67, + "learning_rate": 1.9655064620584377e-05, + "loss": 1.098, + "step": 9040 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654960755073492e-05, + "loss": 0.9997, + "step": 9041 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654856874201705e-05, + "loss": 1.0155, + "step": 9042 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654752977969176e-05, + "loss": 1.1221, + "step": 9043 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654649066376074e-05, + "loss": 0.9583, + "step": 9044 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654545139422564e-05, + "loss": 1.088, + "step": 9045 + }, + { + "epoch": 0.67, + "learning_rate": 1.965444119710881e-05, + "loss": 0.9405, + "step": 9046 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654337239434978e-05, + "loss": 1.0465, + "step": 9047 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654233266401236e-05, + "loss": 0.9902, + "step": 9048 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654129278007745e-05, + "loss": 1.0089, + "step": 9049 + }, + { + "epoch": 0.67, + "learning_rate": 1.9654025274254673e-05, + "loss": 1.0397, + "step": 9050 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653921255142183e-05, + "loss": 1.0535, + "step": 9051 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653817220670444e-05, + "loss": 1.0049, + "step": 9052 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653713170839617e-05, + "loss": 1.0376, + "step": 9053 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653609105649876e-05, + "loss": 1.0268, + "step": 9054 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653505025101375e-05, + "loss": 0.9341, + "step": 9055 + }, + { + "epoch": 0.67, + "learning_rate": 1.965340092919429e-05, + "loss": 0.9378, + "step": 9056 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653296817928783e-05, + "loss": 1.0457, + "step": 9057 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653192691305017e-05, + "loss": 0.9534, + "step": 9058 + }, + { + "epoch": 0.67, + "learning_rate": 1.9653088549323157e-05, + "loss": 1.106, + "step": 9059 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652984391983378e-05, + "loss": 0.9226, + "step": 9060 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652880219285834e-05, + "loss": 1.0759, + "step": 9061 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652776031230695e-05, + "loss": 1.0708, + "step": 9062 + }, + { + "epoch": 0.67, + "learning_rate": 1.965267182781813e-05, + "loss": 0.9789, + "step": 9063 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652567609048302e-05, + "loss": 1.099, + "step": 9064 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652463374921374e-05, + "loss": 0.9629, + "step": 9065 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652359125437515e-05, + "loss": 1.1196, + "step": 9066 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652254860596895e-05, + "loss": 0.9676, + "step": 9067 + }, + { + "epoch": 0.67, + "learning_rate": 1.9652150580399675e-05, + "loss": 1.0205, + "step": 9068 + }, + { + "epoch": 0.67, + "learning_rate": 1.965204628484602e-05, + "loss": 0.9259, + "step": 9069 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651941973936094e-05, + "loss": 0.9793, + "step": 9070 + }, + { + "epoch": 0.67, + "learning_rate": 1.965183764767007e-05, + "loss": 1.0489, + "step": 9071 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651733306048112e-05, + "loss": 1.1096, + "step": 9072 + }, + { + "epoch": 0.67, + "learning_rate": 1.965162894907038e-05, + "loss": 0.9718, + "step": 9073 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651524576737048e-05, + "loss": 1.0565, + "step": 9074 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651420189048278e-05, + "loss": 1.0497, + "step": 9075 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651315786004237e-05, + "loss": 0.9845, + "step": 9076 + }, + { + "epoch": 0.67, + "learning_rate": 1.965121136760509e-05, + "loss": 1.0661, + "step": 9077 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651106933851004e-05, + "loss": 1.1157, + "step": 9078 + }, + { + "epoch": 0.67, + "learning_rate": 1.9651002484742144e-05, + "loss": 1.1084, + "step": 9079 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650898020278677e-05, + "loss": 1.0269, + "step": 9080 + }, + { + "epoch": 0.67, + "learning_rate": 1.965079354046077e-05, + "loss": 1.0161, + "step": 9081 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650689045288586e-05, + "loss": 0.967, + "step": 9082 + }, + { + "epoch": 0.67, + "learning_rate": 1.96505845347623e-05, + "loss": 1.0781, + "step": 9083 + }, + { + "epoch": 0.67, + "learning_rate": 1.965048000888207e-05, + "loss": 0.971, + "step": 9084 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650375467648058e-05, + "loss": 1.0534, + "step": 9085 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650270911060444e-05, + "loss": 1.0625, + "step": 9086 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650166339119385e-05, + "loss": 1.0042, + "step": 9087 + }, + { + "epoch": 0.67, + "learning_rate": 1.9650061751825048e-05, + "loss": 0.9186, + "step": 9088 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649957149177602e-05, + "loss": 1.1052, + "step": 9089 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649852531177212e-05, + "loss": 1.0801, + "step": 9090 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649747897824042e-05, + "loss": 1.0284, + "step": 9091 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649643249118265e-05, + "loss": 0.9288, + "step": 9092 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649538585060045e-05, + "loss": 0.9778, + "step": 9093 + }, + { + "epoch": 0.67, + "learning_rate": 1.964943390564954e-05, + "loss": 0.9152, + "step": 9094 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649329210886932e-05, + "loss": 0.9328, + "step": 9095 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649224500772374e-05, + "loss": 0.9373, + "step": 9096 + }, + { + "epoch": 0.67, + "learning_rate": 1.964911977530604e-05, + "loss": 1.0686, + "step": 9097 + }, + { + "epoch": 0.67, + "learning_rate": 1.9649015034488097e-05, + "loss": 1.0356, + "step": 9098 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648910278318704e-05, + "loss": 1.1012, + "step": 9099 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648805506798038e-05, + "loss": 0.9869, + "step": 9100 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648700719926255e-05, + "loss": 0.9114, + "step": 9101 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648595917703532e-05, + "loss": 0.9854, + "step": 9102 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648491100130027e-05, + "loss": 1.0805, + "step": 9103 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648386267205913e-05, + "loss": 1.0291, + "step": 9104 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648281418931353e-05, + "loss": 0.9775, + "step": 9105 + }, + { + "epoch": 0.67, + "learning_rate": 1.9648176555306517e-05, + "loss": 0.8896, + "step": 9106 + }, + { + "epoch": 0.67, + "learning_rate": 1.964807167633157e-05, + "loss": 0.9477, + "step": 9107 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647966782006674e-05, + "loss": 1.1265, + "step": 9108 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647861872332007e-05, + "loss": 1.0427, + "step": 9109 + }, + { + "epoch": 0.67, + "learning_rate": 1.964775694730773e-05, + "loss": 1.0003, + "step": 9110 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647652006934005e-05, + "loss": 1.0227, + "step": 9111 + }, + { + "epoch": 0.67, + "learning_rate": 1.964754705121101e-05, + "loss": 1.1715, + "step": 9112 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647442080138898e-05, + "loss": 1.1286, + "step": 9113 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647337093717848e-05, + "loss": 1.0831, + "step": 9114 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647232091948023e-05, + "loss": 1.0718, + "step": 9115 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647127074829592e-05, + "loss": 1.0388, + "step": 9116 + }, + { + "epoch": 0.67, + "learning_rate": 1.9647022042362717e-05, + "loss": 0.9706, + "step": 9117 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646916994547567e-05, + "loss": 1.1031, + "step": 9118 + }, + { + "epoch": 0.67, + "learning_rate": 1.964681193138431e-05, + "loss": 1.0592, + "step": 9119 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646706852873113e-05, + "loss": 1.0584, + "step": 9120 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646601759014147e-05, + "loss": 0.953, + "step": 9121 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646496649807575e-05, + "loss": 1.0284, + "step": 9122 + }, + { + "epoch": 0.67, + "learning_rate": 1.964639152525356e-05, + "loss": 1.0903, + "step": 9123 + }, + { + "epoch": 0.67, + "learning_rate": 1.964628638535228e-05, + "loss": 1.0216, + "step": 9124 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646181230103893e-05, + "loss": 1.1036, + "step": 9125 + }, + { + "epoch": 0.67, + "learning_rate": 1.9646076059508573e-05, + "loss": 0.9561, + "step": 9126 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645970873566477e-05, + "loss": 0.9608, + "step": 9127 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645865672277787e-05, + "loss": 1.062, + "step": 9128 + }, + { + "epoch": 0.67, + "learning_rate": 1.964576045564266e-05, + "loss": 1.0798, + "step": 9129 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645655223661263e-05, + "loss": 1.0165, + "step": 9130 + }, + { + "epoch": 0.67, + "learning_rate": 1.964554997633377e-05, + "loss": 1.0023, + "step": 9131 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645444713660346e-05, + "loss": 1.0056, + "step": 9132 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645339435641154e-05, + "loss": 0.8967, + "step": 9133 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645234142276367e-05, + "loss": 0.9854, + "step": 9134 + }, + { + "epoch": 0.67, + "learning_rate": 1.9645128833566153e-05, + "loss": 1.0294, + "step": 9135 + }, + { + "epoch": 0.68, + "learning_rate": 1.964502350951067e-05, + "loss": 1.1304, + "step": 9136 + }, + { + "epoch": 0.68, + "learning_rate": 1.96449181701101e-05, + "loss": 0.9448, + "step": 9137 + }, + { + "epoch": 0.68, + "learning_rate": 1.96448128153646e-05, + "loss": 1.0903, + "step": 9138 + }, + { + "epoch": 0.68, + "learning_rate": 1.964470744527434e-05, + "loss": 1.0753, + "step": 9139 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644602059839487e-05, + "loss": 1.0485, + "step": 9140 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644496659060213e-05, + "loss": 1.0262, + "step": 9141 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644391242936683e-05, + "loss": 1.1484, + "step": 9142 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644285811469063e-05, + "loss": 1.0633, + "step": 9143 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644180364657522e-05, + "loss": 0.8969, + "step": 9144 + }, + { + "epoch": 0.68, + "learning_rate": 1.9644074902502227e-05, + "loss": 0.9708, + "step": 9145 + }, + { + "epoch": 0.68, + "learning_rate": 1.964396942500335e-05, + "loss": 0.958, + "step": 9146 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643863932161055e-05, + "loss": 0.9655, + "step": 9147 + }, + { + "epoch": 0.68, + "learning_rate": 1.964375842397551e-05, + "loss": 1.1124, + "step": 9148 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643652900446884e-05, + "loss": 1.132, + "step": 9149 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643547361575344e-05, + "loss": 0.9933, + "step": 9150 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643441807361056e-05, + "loss": 0.9063, + "step": 9151 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643336237804192e-05, + "loss": 0.9635, + "step": 9152 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643230652904916e-05, + "loss": 1.0065, + "step": 9153 + }, + { + "epoch": 0.68, + "learning_rate": 1.96431250526634e-05, + "loss": 1.0202, + "step": 9154 + }, + { + "epoch": 0.68, + "learning_rate": 1.9643019437079812e-05, + "loss": 1.0532, + "step": 9155 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642913806154314e-05, + "loss": 1.0068, + "step": 9156 + }, + { + "epoch": 0.68, + "learning_rate": 1.964280815988708e-05, + "loss": 0.9869, + "step": 9157 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642702498278274e-05, + "loss": 1.0261, + "step": 9158 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642596821328072e-05, + "loss": 1.0202, + "step": 9159 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642491129036632e-05, + "loss": 1.0818, + "step": 9160 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642385421404125e-05, + "loss": 1.0112, + "step": 9161 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642279698430723e-05, + "loss": 1.0329, + "step": 9162 + }, + { + "epoch": 0.68, + "learning_rate": 1.9642173960116592e-05, + "loss": 1.1099, + "step": 9163 + }, + { + "epoch": 0.68, + "learning_rate": 1.96420682064619e-05, + "loss": 0.9932, + "step": 9164 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641962437466814e-05, + "loss": 1.1386, + "step": 9165 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641856653131505e-05, + "loss": 1.0963, + "step": 9166 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641750853456143e-05, + "loss": 0.9836, + "step": 9167 + }, + { + "epoch": 0.68, + "learning_rate": 1.964164503844089e-05, + "loss": 1.1794, + "step": 9168 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641539208085917e-05, + "loss": 1.0314, + "step": 9169 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641433362391396e-05, + "loss": 1.0049, + "step": 9170 + }, + { + "epoch": 0.68, + "learning_rate": 1.964132750135749e-05, + "loss": 1.0204, + "step": 9171 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641221624984373e-05, + "loss": 0.99, + "step": 9172 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641115733272206e-05, + "loss": 1.1326, + "step": 9173 + }, + { + "epoch": 0.68, + "learning_rate": 1.9641009826221163e-05, + "loss": 1.0087, + "step": 9174 + }, + { + "epoch": 0.68, + "learning_rate": 1.964090390383141e-05, + "loss": 1.0586, + "step": 9175 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640797966103122e-05, + "loss": 0.9876, + "step": 9176 + }, + { + "epoch": 0.68, + "learning_rate": 1.964069201303646e-05, + "loss": 1.0743, + "step": 9177 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640586044631594e-05, + "loss": 1.1103, + "step": 9178 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640480060888694e-05, + "loss": 0.9545, + "step": 9179 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640374061807928e-05, + "loss": 1.0479, + "step": 9180 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640268047389462e-05, + "loss": 0.9652, + "step": 9181 + }, + { + "epoch": 0.68, + "learning_rate": 1.9640162017633472e-05, + "loss": 0.985, + "step": 9182 + }, + { + "epoch": 0.68, + "learning_rate": 1.964005597254012e-05, + "loss": 1.0563, + "step": 9183 + }, + { + "epoch": 0.68, + "learning_rate": 1.963994991210958e-05, + "loss": 1.0793, + "step": 9184 + }, + { + "epoch": 0.68, + "learning_rate": 1.9639843836342015e-05, + "loss": 1.0351, + "step": 9185 + }, + { + "epoch": 0.68, + "learning_rate": 1.9639737745237594e-05, + "loss": 1.0413, + "step": 9186 + }, + { + "epoch": 0.68, + "learning_rate": 1.9639631638796493e-05, + "loss": 1.0813, + "step": 9187 + }, + { + "epoch": 0.68, + "learning_rate": 1.963952551701887e-05, + "loss": 0.9188, + "step": 9188 + }, + { + "epoch": 0.68, + "learning_rate": 1.9639419379904904e-05, + "loss": 0.9618, + "step": 9189 + }, + { + "epoch": 0.68, + "learning_rate": 1.963931322745476e-05, + "loss": 1.0796, + "step": 9190 + }, + { + "epoch": 0.68, + "learning_rate": 1.9639207059668605e-05, + "loss": 1.0458, + "step": 9191 + }, + { + "epoch": 0.68, + "learning_rate": 1.963910087654661e-05, + "loss": 1.1532, + "step": 9192 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638994678088946e-05, + "loss": 1.0423, + "step": 9193 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638888464295777e-05, + "loss": 0.9863, + "step": 9194 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638782235167273e-05, + "loss": 1.063, + "step": 9195 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638675990703603e-05, + "loss": 1.1337, + "step": 9196 + }, + { + "epoch": 0.68, + "learning_rate": 1.963856973090494e-05, + "loss": 1.0527, + "step": 9197 + }, + { + "epoch": 0.68, + "learning_rate": 1.963846345577145e-05, + "loss": 1.0571, + "step": 9198 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638357165303305e-05, + "loss": 0.9626, + "step": 9199 + }, + { + "epoch": 0.68, + "learning_rate": 1.963825085950067e-05, + "loss": 1.1517, + "step": 9200 + }, + { + "epoch": 0.68, + "learning_rate": 1.9638144538363714e-05, + "loss": 1.1279, + "step": 9201 + }, + { + "epoch": 0.68, + "learning_rate": 1.963803820189261e-05, + "loss": 0.9749, + "step": 9202 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637931850087523e-05, + "loss": 0.983, + "step": 9203 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637825482948627e-05, + "loss": 1.0392, + "step": 9204 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637719100476087e-05, + "loss": 1.044, + "step": 9205 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637612702670075e-05, + "loss": 1.0483, + "step": 9206 + }, + { + "epoch": 0.68, + "learning_rate": 1.963750628953076e-05, + "loss": 1.0285, + "step": 9207 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637399861058305e-05, + "loss": 1.0515, + "step": 9208 + }, + { + "epoch": 0.68, + "learning_rate": 1.963729341725289e-05, + "loss": 1.0363, + "step": 9209 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637186958114678e-05, + "loss": 0.9813, + "step": 9210 + }, + { + "epoch": 0.68, + "learning_rate": 1.9637080483643837e-05, + "loss": 1.1248, + "step": 9211 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636973993840543e-05, + "loss": 0.9515, + "step": 9212 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636867488704957e-05, + "loss": 1.0999, + "step": 9213 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636760968237255e-05, + "loss": 1.1687, + "step": 9214 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636654432437603e-05, + "loss": 1.1145, + "step": 9215 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636547881306173e-05, + "loss": 0.9808, + "step": 9216 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636441314843132e-05, + "loss": 1.0593, + "step": 9217 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636334733048652e-05, + "loss": 0.9836, + "step": 9218 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636228135922903e-05, + "loss": 0.9944, + "step": 9219 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636121523466047e-05, + "loss": 1.0504, + "step": 9220 + }, + { + "epoch": 0.68, + "learning_rate": 1.9636014895678264e-05, + "loss": 0.9519, + "step": 9221 + }, + { + "epoch": 0.68, + "learning_rate": 1.963590825255972e-05, + "loss": 1.0568, + "step": 9222 + }, + { + "epoch": 0.68, + "learning_rate": 1.963580159411058e-05, + "loss": 1.0395, + "step": 9223 + }, + { + "epoch": 0.68, + "learning_rate": 1.963569492033102e-05, + "loss": 1.0634, + "step": 9224 + }, + { + "epoch": 0.68, + "learning_rate": 1.9635588231221207e-05, + "loss": 1.0432, + "step": 9225 + }, + { + "epoch": 0.68, + "learning_rate": 1.963548152678131e-05, + "loss": 1.1345, + "step": 9226 + }, + { + "epoch": 0.68, + "learning_rate": 1.9635374807011502e-05, + "loss": 0.9181, + "step": 9227 + }, + { + "epoch": 0.68, + "learning_rate": 1.963526807191195e-05, + "loss": 1.0174, + "step": 9228 + }, + { + "epoch": 0.68, + "learning_rate": 1.9635161321482822e-05, + "loss": 1.1022, + "step": 9229 + }, + { + "epoch": 0.68, + "learning_rate": 1.9635054555724292e-05, + "loss": 1.0601, + "step": 9230 + }, + { + "epoch": 0.68, + "learning_rate": 1.963494777463653e-05, + "loss": 1.0524, + "step": 9231 + }, + { + "epoch": 0.68, + "learning_rate": 1.96348409782197e-05, + "loss": 0.9734, + "step": 9232 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634734166473975e-05, + "loss": 0.9869, + "step": 9233 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634627339399533e-05, + "loss": 1.0279, + "step": 9234 + }, + { + "epoch": 0.68, + "learning_rate": 1.963452049699653e-05, + "loss": 0.9906, + "step": 9235 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634413639265145e-05, + "loss": 1.1084, + "step": 9236 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634306766205546e-05, + "loss": 1.0013, + "step": 9237 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634199877817902e-05, + "loss": 0.9735, + "step": 9238 + }, + { + "epoch": 0.68, + "learning_rate": 1.9634092974102384e-05, + "loss": 1.1086, + "step": 9239 + }, + { + "epoch": 0.68, + "learning_rate": 1.963398605505916e-05, + "loss": 1.0251, + "step": 9240 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633879120688406e-05, + "loss": 0.9526, + "step": 9241 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633772170990285e-05, + "loss": 0.9308, + "step": 9242 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633665205964974e-05, + "loss": 1.115, + "step": 9243 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633558225612636e-05, + "loss": 1.0708, + "step": 9244 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633451229933447e-05, + "loss": 0.9674, + "step": 9245 + }, + { + "epoch": 0.68, + "learning_rate": 1.9633344218927575e-05, + "loss": 1.066, + "step": 9246 + }, + { + "epoch": 0.68, + "learning_rate": 1.963323719259519e-05, + "loss": 0.981, + "step": 9247 + }, + { + "epoch": 0.68, + "learning_rate": 1.963313015093646e-05, + "loss": 1.111, + "step": 9248 + }, + { + "epoch": 0.68, + "learning_rate": 1.963302309395156e-05, + "loss": 1.1725, + "step": 9249 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632916021640652e-05, + "loss": 1.1259, + "step": 9250 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632808934003923e-05, + "loss": 1.0087, + "step": 9251 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632701831041526e-05, + "loss": 1.0604, + "step": 9252 + }, + { + "epoch": 0.68, + "learning_rate": 1.963259471275364e-05, + "loss": 1.0893, + "step": 9253 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632487579140434e-05, + "loss": 0.9721, + "step": 9254 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632380430202075e-05, + "loss": 1.1627, + "step": 9255 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632273265938738e-05, + "loss": 0.9886, + "step": 9256 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632166086350596e-05, + "loss": 0.9798, + "step": 9257 + }, + { + "epoch": 0.68, + "learning_rate": 1.9632058891437812e-05, + "loss": 1.1274, + "step": 9258 + }, + { + "epoch": 0.68, + "learning_rate": 1.963195168120056e-05, + "loss": 0.958, + "step": 9259 + }, + { + "epoch": 0.68, + "learning_rate": 1.963184445563901e-05, + "loss": 0.8619, + "step": 9260 + }, + { + "epoch": 0.68, + "learning_rate": 1.9631737214753337e-05, + "loss": 1.0458, + "step": 9261 + }, + { + "epoch": 0.68, + "learning_rate": 1.9631629958543703e-05, + "loss": 1.0455, + "step": 9262 + }, + { + "epoch": 0.68, + "learning_rate": 1.9631522687010287e-05, + "loss": 1.1002, + "step": 9263 + }, + { + "epoch": 0.68, + "learning_rate": 1.9631415400153256e-05, + "loss": 1.0129, + "step": 9264 + }, + { + "epoch": 0.68, + "learning_rate": 1.963130809797278e-05, + "loss": 1.0426, + "step": 9265 + }, + { + "epoch": 0.68, + "learning_rate": 1.9631200780469032e-05, + "loss": 1.0648, + "step": 9266 + }, + { + "epoch": 0.68, + "learning_rate": 1.963109344764218e-05, + "loss": 1.0414, + "step": 9267 + }, + { + "epoch": 0.68, + "learning_rate": 1.9630986099492396e-05, + "loss": 1.0444, + "step": 9268 + }, + { + "epoch": 0.68, + "learning_rate": 1.963087873601985e-05, + "loss": 1.0717, + "step": 9269 + }, + { + "epoch": 0.68, + "learning_rate": 1.9630771357224713e-05, + "loss": 0.9851, + "step": 9270 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630663963107158e-05, + "loss": 0.9877, + "step": 9271 + }, + { + "epoch": 0.69, + "learning_rate": 1.963055655366736e-05, + "loss": 1.0558, + "step": 9272 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630449128905477e-05, + "loss": 1.0032, + "step": 9273 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630341688821687e-05, + "loss": 1.0589, + "step": 9274 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630234233416168e-05, + "loss": 1.0239, + "step": 9275 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630126762689077e-05, + "loss": 0.9741, + "step": 9276 + }, + { + "epoch": 0.69, + "learning_rate": 1.9630019276640597e-05, + "loss": 1.0682, + "step": 9277 + }, + { + "epoch": 0.69, + "learning_rate": 1.962991177527089e-05, + "loss": 1.0496, + "step": 9278 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629804258580135e-05, + "loss": 1.1201, + "step": 9279 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629696726568496e-05, + "loss": 1.0687, + "step": 9280 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629589179236152e-05, + "loss": 1.111, + "step": 9281 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629481616583262e-05, + "loss": 0.9636, + "step": 9282 + }, + { + "epoch": 0.69, + "learning_rate": 1.962937403861001e-05, + "loss": 1.0209, + "step": 9283 + }, + { + "epoch": 0.69, + "learning_rate": 1.962926644531656e-05, + "loss": 0.9753, + "step": 9284 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629158836703082e-05, + "loss": 0.9527, + "step": 9285 + }, + { + "epoch": 0.69, + "learning_rate": 1.9629051212769758e-05, + "loss": 1.0061, + "step": 9286 + }, + { + "epoch": 0.69, + "learning_rate": 1.9628943573516744e-05, + "loss": 0.9722, + "step": 9287 + }, + { + "epoch": 0.69, + "learning_rate": 1.962883591894422e-05, + "loss": 1.0361, + "step": 9288 + }, + { + "epoch": 0.69, + "learning_rate": 1.962872824905236e-05, + "loss": 1.0078, + "step": 9289 + }, + { + "epoch": 0.69, + "learning_rate": 1.9628620563841324e-05, + "loss": 1.0764, + "step": 9290 + }, + { + "epoch": 0.69, + "learning_rate": 1.9628512863311297e-05, + "loss": 1.082, + "step": 9291 + }, + { + "epoch": 0.69, + "learning_rate": 1.962840514746244e-05, + "loss": 1.0525, + "step": 9292 + }, + { + "epoch": 0.69, + "learning_rate": 1.9628297416294926e-05, + "loss": 0.9695, + "step": 9293 + }, + { + "epoch": 0.69, + "learning_rate": 1.9628189669808938e-05, + "loss": 1.0794, + "step": 9294 + }, + { + "epoch": 0.69, + "learning_rate": 1.962808190800463e-05, + "loss": 1.0664, + "step": 9295 + }, + { + "epoch": 0.69, + "learning_rate": 1.962797413088218e-05, + "loss": 0.9976, + "step": 9296 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627866338441768e-05, + "loss": 1.0829, + "step": 9297 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627758530683554e-05, + "loss": 0.9686, + "step": 9298 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627650707607714e-05, + "loss": 1.0719, + "step": 9299 + }, + { + "epoch": 0.69, + "learning_rate": 1.962754286921442e-05, + "loss": 1.0693, + "step": 9300 + }, + { + "epoch": 0.69, + "learning_rate": 1.962743501550384e-05, + "loss": 1.0801, + "step": 9301 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627327146476155e-05, + "loss": 1.1124, + "step": 9302 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627219262131526e-05, + "loss": 0.8666, + "step": 9303 + }, + { + "epoch": 0.69, + "learning_rate": 1.962711136247013e-05, + "loss": 1.0886, + "step": 9304 + }, + { + "epoch": 0.69, + "learning_rate": 1.9627003447492136e-05, + "loss": 1.0625, + "step": 9305 + }, + { + "epoch": 0.69, + "learning_rate": 1.962689551719772e-05, + "loss": 0.995, + "step": 9306 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626787571587053e-05, + "loss": 1.043, + "step": 9307 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626679610660298e-05, + "loss": 1.1045, + "step": 9308 + }, + { + "epoch": 0.69, + "learning_rate": 1.962657163441764e-05, + "loss": 0.9596, + "step": 9309 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626463642859238e-05, + "loss": 0.9902, + "step": 9310 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626355635985272e-05, + "loss": 1.076, + "step": 9311 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626247613795914e-05, + "loss": 1.0061, + "step": 9312 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626139576291336e-05, + "loss": 1.061, + "step": 9313 + }, + { + "epoch": 0.69, + "learning_rate": 1.9626031523471704e-05, + "loss": 1.0027, + "step": 9314 + }, + { + "epoch": 0.69, + "learning_rate": 1.962592345533719e-05, + "loss": 1.1484, + "step": 9315 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625815371887978e-05, + "loss": 1.1198, + "step": 9316 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625707273124225e-05, + "loss": 0.9539, + "step": 9317 + }, + { + "epoch": 0.69, + "learning_rate": 1.962559915904611e-05, + "loss": 1.1428, + "step": 9318 + }, + { + "epoch": 0.69, + "learning_rate": 1.962549102965381e-05, + "loss": 1.1054, + "step": 9319 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625382884947485e-05, + "loss": 0.9586, + "step": 9320 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625274724927316e-05, + "loss": 0.9406, + "step": 9321 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625166549593472e-05, + "loss": 1.0259, + "step": 9322 + }, + { + "epoch": 0.69, + "learning_rate": 1.9625058358946127e-05, + "loss": 0.9446, + "step": 9323 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624950152985452e-05, + "loss": 0.9721, + "step": 9324 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624841931711618e-05, + "loss": 0.9236, + "step": 9325 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624733695124797e-05, + "loss": 0.9974, + "step": 9326 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624625443225165e-05, + "loss": 1.0362, + "step": 9327 + }, + { + "epoch": 0.69, + "learning_rate": 1.962451717601289e-05, + "loss": 0.9302, + "step": 9328 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624408893488147e-05, + "loss": 0.9982, + "step": 9329 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624300595651107e-05, + "loss": 0.9939, + "step": 9330 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624192282501943e-05, + "loss": 1.0714, + "step": 9331 + }, + { + "epoch": 0.69, + "learning_rate": 1.9624083954040826e-05, + "loss": 1.1103, + "step": 9332 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623975610267928e-05, + "loss": 1.0609, + "step": 9333 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623867251183423e-05, + "loss": 0.9691, + "step": 9334 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623758876787482e-05, + "loss": 1.007, + "step": 9335 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623650487080282e-05, + "loss": 1.0616, + "step": 9336 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623542082061986e-05, + "loss": 1.057, + "step": 9337 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623433661732777e-05, + "loss": 0.9967, + "step": 9338 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623325226092823e-05, + "loss": 1.0628, + "step": 9339 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623216775142292e-05, + "loss": 0.9425, + "step": 9340 + }, + { + "epoch": 0.69, + "learning_rate": 1.9623108308881366e-05, + "loss": 1.048, + "step": 9341 + }, + { + "epoch": 0.69, + "learning_rate": 1.962299982731021e-05, + "loss": 0.9475, + "step": 9342 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622891330428996e-05, + "loss": 1.0733, + "step": 9343 + }, + { + "epoch": 0.69, + "learning_rate": 1.96227828182379e-05, + "loss": 1.0017, + "step": 9344 + }, + { + "epoch": 0.69, + "learning_rate": 1.96226742907371e-05, + "loss": 0.9898, + "step": 9345 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622565747926757e-05, + "loss": 1.0135, + "step": 9346 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622457189807052e-05, + "loss": 1.0357, + "step": 9347 + }, + { + "epoch": 0.69, + "learning_rate": 1.962234861637816e-05, + "loss": 1.0022, + "step": 9348 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622240027640243e-05, + "loss": 1.0457, + "step": 9349 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622131423593477e-05, + "loss": 1.0404, + "step": 9350 + }, + { + "epoch": 0.69, + "learning_rate": 1.9622022804238044e-05, + "loss": 0.9299, + "step": 9351 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621914169574106e-05, + "loss": 1.0138, + "step": 9352 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621805519601844e-05, + "loss": 1.1018, + "step": 9353 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621696854321426e-05, + "loss": 1.0922, + "step": 9354 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621588173733026e-05, + "loss": 0.9949, + "step": 9355 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621479477836813e-05, + "loss": 1.0068, + "step": 9356 + }, + { + "epoch": 0.69, + "learning_rate": 1.962137076663297e-05, + "loss": 1.0399, + "step": 9357 + }, + { + "epoch": 0.69, + "learning_rate": 1.962126204012166e-05, + "loss": 1.0412, + "step": 9358 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621153298303058e-05, + "loss": 1.0577, + "step": 9359 + }, + { + "epoch": 0.69, + "learning_rate": 1.9621044541177344e-05, + "loss": 0.9321, + "step": 9360 + }, + { + "epoch": 0.69, + "learning_rate": 1.9620935768744683e-05, + "loss": 1.1312, + "step": 9361 + }, + { + "epoch": 0.69, + "learning_rate": 1.962082698100525e-05, + "loss": 1.0044, + "step": 9362 + }, + { + "epoch": 0.69, + "learning_rate": 1.962071817795922e-05, + "loss": 1.0543, + "step": 9363 + }, + { + "epoch": 0.69, + "learning_rate": 1.9620609359606765e-05, + "loss": 1.0202, + "step": 9364 + }, + { + "epoch": 0.69, + "learning_rate": 1.9620500525948056e-05, + "loss": 0.894, + "step": 9365 + }, + { + "epoch": 0.69, + "learning_rate": 1.962039167698327e-05, + "loss": 1.1615, + "step": 9366 + }, + { + "epoch": 0.69, + "learning_rate": 1.962028281271258e-05, + "loss": 1.0554, + "step": 9367 + }, + { + "epoch": 0.69, + "learning_rate": 1.962017393313616e-05, + "loss": 1.0884, + "step": 9368 + }, + { + "epoch": 0.69, + "learning_rate": 1.9620065038254176e-05, + "loss": 1.1447, + "step": 9369 + }, + { + "epoch": 0.69, + "learning_rate": 1.9619956128066806e-05, + "loss": 1.0179, + "step": 9370 + }, + { + "epoch": 0.69, + "learning_rate": 1.9619847202574224e-05, + "loss": 1.0951, + "step": 9371 + }, + { + "epoch": 0.69, + "learning_rate": 1.9619738261776605e-05, + "loss": 1.0916, + "step": 9372 + }, + { + "epoch": 0.69, + "learning_rate": 1.961962930567412e-05, + "loss": 1.0659, + "step": 9373 + }, + { + "epoch": 0.69, + "learning_rate": 1.961952033426694e-05, + "loss": 0.9949, + "step": 9374 + }, + { + "epoch": 0.69, + "learning_rate": 1.961941134755524e-05, + "loss": 1.037, + "step": 9375 + }, + { + "epoch": 0.69, + "learning_rate": 1.96193023455392e-05, + "loss": 0.993, + "step": 9376 + }, + { + "epoch": 0.69, + "learning_rate": 1.9619193328218983e-05, + "loss": 1.011, + "step": 9377 + }, + { + "epoch": 0.69, + "learning_rate": 1.961908429559477e-05, + "loss": 1.0176, + "step": 9378 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618975247666726e-05, + "loss": 0.9648, + "step": 9379 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618866184435034e-05, + "loss": 1.0923, + "step": 9380 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618757105899862e-05, + "loss": 1.0505, + "step": 9381 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618648012061387e-05, + "loss": 1.2383, + "step": 9382 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618538902919777e-05, + "loss": 1.0212, + "step": 9383 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618429778475213e-05, + "loss": 1.0174, + "step": 9384 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618320638727863e-05, + "loss": 1.0076, + "step": 9385 + }, + { + "epoch": 0.69, + "learning_rate": 1.96182114836779e-05, + "loss": 0.9915, + "step": 9386 + }, + { + "epoch": 0.69, + "learning_rate": 1.9618102313325507e-05, + "loss": 1.0853, + "step": 9387 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617993127670845e-05, + "loss": 0.9807, + "step": 9388 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617883926714096e-05, + "loss": 1.0161, + "step": 9389 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617774710455433e-05, + "loss": 0.9666, + "step": 9390 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617665478895025e-05, + "loss": 0.9614, + "step": 9391 + }, + { + "epoch": 0.69, + "learning_rate": 1.961755623203305e-05, + "loss": 1.0994, + "step": 9392 + }, + { + "epoch": 0.69, + "learning_rate": 1.961744696986968e-05, + "loss": 0.9911, + "step": 9393 + }, + { + "epoch": 0.69, + "learning_rate": 1.961733769240509e-05, + "loss": 1.0846, + "step": 9394 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617228399639454e-05, + "loss": 1.0307, + "step": 9395 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617119091572943e-05, + "loss": 1.0524, + "step": 9396 + }, + { + "epoch": 0.69, + "learning_rate": 1.9617009768205734e-05, + "loss": 1.0064, + "step": 9397 + }, + { + "epoch": 0.69, + "learning_rate": 1.9616900429538002e-05, + "loss": 1.0254, + "step": 9398 + }, + { + "epoch": 0.69, + "learning_rate": 1.9616791075569915e-05, + "loss": 1.0693, + "step": 9399 + }, + { + "epoch": 0.69, + "learning_rate": 1.9616681706301658e-05, + "loss": 0.9921, + "step": 9400 + }, + { + "epoch": 0.69, + "learning_rate": 1.961657232173339e-05, + "loss": 1.0647, + "step": 9401 + }, + { + "epoch": 0.69, + "learning_rate": 1.9616462921865296e-05, + "loss": 1.0455, + "step": 9402 + }, + { + "epoch": 0.69, + "learning_rate": 1.961635350669755e-05, + "loss": 0.988, + "step": 9403 + }, + { + "epoch": 0.69, + "learning_rate": 1.961624407623032e-05, + "loss": 0.9839, + "step": 9404 + }, + { + "epoch": 0.69, + "learning_rate": 1.9616134630463787e-05, + "loss": 1.0629, + "step": 9405 + }, + { + "epoch": 0.7, + "learning_rate": 1.9616025169398117e-05, + "loss": 1.0627, + "step": 9406 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615915693033488e-05, + "loss": 1.0211, + "step": 9407 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615806201370078e-05, + "loss": 0.9717, + "step": 9408 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615696694408057e-05, + "loss": 0.9282, + "step": 9409 + }, + { + "epoch": 0.7, + "learning_rate": 1.96155871721476e-05, + "loss": 1.1247, + "step": 9410 + }, + { + "epoch": 0.7, + "learning_rate": 1.961547763458888e-05, + "loss": 1.0434, + "step": 9411 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615368081732077e-05, + "loss": 0.9503, + "step": 9412 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615258513577357e-05, + "loss": 1.0399, + "step": 9413 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615148930124897e-05, + "loss": 0.9842, + "step": 9414 + }, + { + "epoch": 0.7, + "learning_rate": 1.9615039331374875e-05, + "loss": 1.0529, + "step": 9415 + }, + { + "epoch": 0.7, + "learning_rate": 1.9614929717327463e-05, + "loss": 0.9971, + "step": 9416 + }, + { + "epoch": 0.7, + "learning_rate": 1.9614820087982836e-05, + "loss": 0.997, + "step": 9417 + }, + { + "epoch": 0.7, + "learning_rate": 1.9614710443341166e-05, + "loss": 1.0078, + "step": 9418 + }, + { + "epoch": 0.7, + "learning_rate": 1.961460078340263e-05, + "loss": 0.9981, + "step": 9419 + }, + { + "epoch": 0.7, + "learning_rate": 1.9614491108167404e-05, + "loss": 1.0514, + "step": 9420 + }, + { + "epoch": 0.7, + "learning_rate": 1.9614381417635657e-05, + "loss": 1.0754, + "step": 9421 + }, + { + "epoch": 0.7, + "learning_rate": 1.961427171180757e-05, + "loss": 0.9607, + "step": 9422 + }, + { + "epoch": 0.7, + "learning_rate": 1.961416199068331e-05, + "loss": 1.0651, + "step": 9423 + }, + { + "epoch": 0.7, + "learning_rate": 1.961405225426306e-05, + "loss": 1.0932, + "step": 9424 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613942502546986e-05, + "loss": 0.9421, + "step": 9425 + }, + { + "epoch": 0.7, + "learning_rate": 1.961383273553527e-05, + "loss": 1.0351, + "step": 9426 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613722953228082e-05, + "loss": 1.0422, + "step": 9427 + }, + { + "epoch": 0.7, + "learning_rate": 1.96136131556256e-05, + "loss": 0.9396, + "step": 9428 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613503342727995e-05, + "loss": 0.9978, + "step": 9429 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613393514535444e-05, + "loss": 0.9417, + "step": 9430 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613283671048125e-05, + "loss": 0.9515, + "step": 9431 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613173812266205e-05, + "loss": 0.9811, + "step": 9432 + }, + { + "epoch": 0.7, + "learning_rate": 1.9613063938189863e-05, + "loss": 1.1406, + "step": 9433 + }, + { + "epoch": 0.7, + "learning_rate": 1.961295404881928e-05, + "loss": 1.0427, + "step": 9434 + }, + { + "epoch": 0.7, + "learning_rate": 1.961284414415462e-05, + "loss": 0.9537, + "step": 9435 + }, + { + "epoch": 0.7, + "learning_rate": 1.961273422419606e-05, + "loss": 1.1446, + "step": 9436 + }, + { + "epoch": 0.7, + "learning_rate": 1.961262428894378e-05, + "loss": 0.9673, + "step": 9437 + }, + { + "epoch": 0.7, + "learning_rate": 1.9612514338397954e-05, + "loss": 1.0673, + "step": 9438 + }, + { + "epoch": 0.7, + "learning_rate": 1.9612404372558754e-05, + "loss": 1.0871, + "step": 9439 + }, + { + "epoch": 0.7, + "learning_rate": 1.9612294391426355e-05, + "loss": 1.1104, + "step": 9440 + }, + { + "epoch": 0.7, + "learning_rate": 1.9612184395000935e-05, + "loss": 1.0575, + "step": 9441 + }, + { + "epoch": 0.7, + "learning_rate": 1.9612074383282666e-05, + "loss": 0.9597, + "step": 9442 + }, + { + "epoch": 0.7, + "learning_rate": 1.9611964356271723e-05, + "loss": 1.1344, + "step": 9443 + }, + { + "epoch": 0.7, + "learning_rate": 1.9611854313968285e-05, + "loss": 1.0852, + "step": 9444 + }, + { + "epoch": 0.7, + "learning_rate": 1.9611744256372524e-05, + "loss": 0.9774, + "step": 9445 + }, + { + "epoch": 0.7, + "learning_rate": 1.9611634183484613e-05, + "loss": 1.0945, + "step": 9446 + }, + { + "epoch": 0.7, + "learning_rate": 1.961152409530473e-05, + "loss": 1.1094, + "step": 9447 + }, + { + "epoch": 0.7, + "learning_rate": 1.961141399183305e-05, + "loss": 1.0809, + "step": 9448 + }, + { + "epoch": 0.7, + "learning_rate": 1.961130387306975e-05, + "loss": 1.0151, + "step": 9449 + }, + { + "epoch": 0.7, + "learning_rate": 1.9611193739015e-05, + "loss": 1.0936, + "step": 9450 + }, + { + "epoch": 0.7, + "learning_rate": 1.961108358966898e-05, + "loss": 1.0365, + "step": 9451 + }, + { + "epoch": 0.7, + "learning_rate": 1.961097342503186e-05, + "loss": 1.0798, + "step": 9452 + }, + { + "epoch": 0.7, + "learning_rate": 1.9610863245103828e-05, + "loss": 0.9349, + "step": 9453 + }, + { + "epoch": 0.7, + "learning_rate": 1.961075304988504e-05, + "loss": 1.0771, + "step": 9454 + }, + { + "epoch": 0.7, + "learning_rate": 1.9610642839375688e-05, + "loss": 1.0778, + "step": 9455 + }, + { + "epoch": 0.7, + "learning_rate": 1.9610532613575938e-05, + "loss": 0.995, + "step": 9456 + }, + { + "epoch": 0.7, + "learning_rate": 1.9610422372485965e-05, + "loss": 1.0007, + "step": 9457 + }, + { + "epoch": 0.7, + "learning_rate": 1.961031211610595e-05, + "loss": 0.9951, + "step": 9458 + }, + { + "epoch": 0.7, + "learning_rate": 1.961020184443607e-05, + "loss": 1.0423, + "step": 9459 + }, + { + "epoch": 0.7, + "learning_rate": 1.9610091557476495e-05, + "loss": 1.049, + "step": 9460 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609981255227397e-05, + "loss": 1.038, + "step": 9461 + }, + { + "epoch": 0.7, + "learning_rate": 1.960987093768896e-05, + "loss": 1.0687, + "step": 9462 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609760604861353e-05, + "loss": 1.0665, + "step": 9463 + }, + { + "epoch": 0.7, + "learning_rate": 1.960965025674476e-05, + "loss": 0.9501, + "step": 9464 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609539893339343e-05, + "loss": 0.9264, + "step": 9465 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609429514645294e-05, + "loss": 1.0505, + "step": 9466 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609319120662775e-05, + "loss": 1.0561, + "step": 9467 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609208711391968e-05, + "loss": 1.0945, + "step": 9468 + }, + { + "epoch": 0.7, + "learning_rate": 1.9609098286833047e-05, + "loss": 0.9987, + "step": 9469 + }, + { + "epoch": 0.7, + "learning_rate": 1.960898784698619e-05, + "loss": 0.8958, + "step": 9470 + }, + { + "epoch": 0.7, + "learning_rate": 1.9608877391851568e-05, + "loss": 1.0349, + "step": 9471 + }, + { + "epoch": 0.7, + "learning_rate": 1.960876692142936e-05, + "loss": 1.2051, + "step": 9472 + }, + { + "epoch": 0.7, + "learning_rate": 1.9608656435719742e-05, + "loss": 1.1048, + "step": 9473 + }, + { + "epoch": 0.7, + "learning_rate": 1.960854593472289e-05, + "loss": 1.0125, + "step": 9474 + }, + { + "epoch": 0.7, + "learning_rate": 1.9608435418438975e-05, + "loss": 1.1257, + "step": 9475 + }, + { + "epoch": 0.7, + "learning_rate": 1.960832488686818e-05, + "loss": 0.9573, + "step": 9476 + }, + { + "epoch": 0.7, + "learning_rate": 1.9608214340010678e-05, + "loss": 1.1043, + "step": 9477 + }, + { + "epoch": 0.7, + "learning_rate": 1.9608103777866643e-05, + "loss": 1.1084, + "step": 9478 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607993200436256e-05, + "loss": 1.103, + "step": 9479 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607882607719685e-05, + "loss": 1.0044, + "step": 9480 + }, + { + "epoch": 0.7, + "learning_rate": 1.960777199971711e-05, + "loss": 1.1049, + "step": 9481 + }, + { + "epoch": 0.7, + "learning_rate": 1.960766137642871e-05, + "loss": 1.0175, + "step": 9482 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607550737854658e-05, + "loss": 1.0547, + "step": 9483 + }, + { + "epoch": 0.7, + "learning_rate": 1.960744008399513e-05, + "loss": 0.9465, + "step": 9484 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607329414850304e-05, + "loss": 1.0506, + "step": 9485 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607218730420346e-05, + "loss": 0.9876, + "step": 9486 + }, + { + "epoch": 0.7, + "learning_rate": 1.9607108030705452e-05, + "loss": 1.0459, + "step": 9487 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606997315705777e-05, + "loss": 0.9926, + "step": 9488 + }, + { + "epoch": 0.7, + "learning_rate": 1.960688658542151e-05, + "loss": 0.936, + "step": 9489 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606775839852827e-05, + "loss": 0.9494, + "step": 9490 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606665078999897e-05, + "loss": 1.0742, + "step": 9491 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606554302862902e-05, + "loss": 1.1473, + "step": 9492 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606443511442017e-05, + "loss": 0.9953, + "step": 9493 + }, + { + "epoch": 0.7, + "learning_rate": 1.960633270473742e-05, + "loss": 1.0972, + "step": 9494 + }, + { + "epoch": 0.7, + "learning_rate": 1.960622188274928e-05, + "loss": 1.0203, + "step": 9495 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606111045477784e-05, + "loss": 1.0642, + "step": 9496 + }, + { + "epoch": 0.7, + "learning_rate": 1.9606000192923095e-05, + "loss": 1.0516, + "step": 9497 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605889325085402e-05, + "loss": 1.0217, + "step": 9498 + }, + { + "epoch": 0.7, + "learning_rate": 1.960577844196488e-05, + "loss": 1.0479, + "step": 9499 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605667543561697e-05, + "loss": 0.9411, + "step": 9500 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605556629876035e-05, + "loss": 1.0802, + "step": 9501 + }, + { + "epoch": 0.7, + "learning_rate": 1.960544570090807e-05, + "loss": 0.972, + "step": 9502 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605334756657977e-05, + "loss": 1.0656, + "step": 9503 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605223797125938e-05, + "loss": 1.1404, + "step": 9504 + }, + { + "epoch": 0.7, + "learning_rate": 1.9605112822312124e-05, + "loss": 1.0099, + "step": 9505 + }, + { + "epoch": 0.7, + "learning_rate": 1.960500183221671e-05, + "loss": 1.0987, + "step": 9506 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604890826839877e-05, + "loss": 1.0072, + "step": 9507 + }, + { + "epoch": 0.7, + "learning_rate": 1.96047798061818e-05, + "loss": 0.9641, + "step": 9508 + }, + { + "epoch": 0.7, + "learning_rate": 1.960466877024266e-05, + "loss": 1.0733, + "step": 9509 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604557719022624e-05, + "loss": 1.0398, + "step": 9510 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604446652521874e-05, + "loss": 1.124, + "step": 9511 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604335570740592e-05, + "loss": 0.9514, + "step": 9512 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604224473678944e-05, + "loss": 1.0728, + "step": 9513 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604113361337113e-05, + "loss": 1.0291, + "step": 9514 + }, + { + "epoch": 0.7, + "learning_rate": 1.9604002233715282e-05, + "loss": 1.2, + "step": 9515 + }, + { + "epoch": 0.7, + "learning_rate": 1.960389109081361e-05, + "loss": 1.0771, + "step": 9516 + }, + { + "epoch": 0.7, + "learning_rate": 1.960377993263229e-05, + "loss": 1.0427, + "step": 9517 + }, + { + "epoch": 0.7, + "learning_rate": 1.9603668759171496e-05, + "loss": 1.091, + "step": 9518 + }, + { + "epoch": 0.7, + "learning_rate": 1.96035575704314e-05, + "loss": 1.0861, + "step": 9519 + }, + { + "epoch": 0.7, + "learning_rate": 1.960344636641218e-05, + "loss": 1.0922, + "step": 9520 + }, + { + "epoch": 0.7, + "learning_rate": 1.9603335147114015e-05, + "loss": 0.9856, + "step": 9521 + }, + { + "epoch": 0.7, + "learning_rate": 1.9603223912537083e-05, + "loss": 1.0617, + "step": 9522 + }, + { + "epoch": 0.7, + "learning_rate": 1.9603112662681557e-05, + "loss": 1.0425, + "step": 9523 + }, + { + "epoch": 0.7, + "learning_rate": 1.9603001397547617e-05, + "loss": 1.0358, + "step": 9524 + }, + { + "epoch": 0.7, + "learning_rate": 1.9602890117135442e-05, + "loss": 1.0969, + "step": 9525 + }, + { + "epoch": 0.7, + "learning_rate": 1.96027788214452e-05, + "loss": 0.9605, + "step": 9526 + }, + { + "epoch": 0.7, + "learning_rate": 1.9602667510477078e-05, + "loss": 1.0809, + "step": 9527 + }, + { + "epoch": 0.7, + "learning_rate": 1.960255618423125e-05, + "loss": 0.9883, + "step": 9528 + }, + { + "epoch": 0.7, + "learning_rate": 1.960244484270789e-05, + "loss": 0.9329, + "step": 9529 + }, + { + "epoch": 0.7, + "learning_rate": 1.960233348590718e-05, + "loss": 1.0196, + "step": 9530 + }, + { + "epoch": 0.7, + "learning_rate": 1.9602222113829295e-05, + "loss": 1.0757, + "step": 9531 + }, + { + "epoch": 0.7, + "learning_rate": 1.960211072647441e-05, + "loss": 0.9911, + "step": 9532 + }, + { + "epoch": 0.7, + "learning_rate": 1.9601999323842705e-05, + "loss": 1.0937, + "step": 9533 + }, + { + "epoch": 0.7, + "learning_rate": 1.960188790593436e-05, + "loss": 1.1498, + "step": 9534 + }, + { + "epoch": 0.7, + "learning_rate": 1.9601776472749542e-05, + "loss": 1.0169, + "step": 9535 + }, + { + "epoch": 0.7, + "learning_rate": 1.960166502428844e-05, + "loss": 1.1435, + "step": 9536 + }, + { + "epoch": 0.7, + "learning_rate": 1.9601553560551223e-05, + "loss": 1.0473, + "step": 9537 + }, + { + "epoch": 0.7, + "learning_rate": 1.9601442081538075e-05, + "loss": 1.0813, + "step": 9538 + }, + { + "epoch": 0.7, + "learning_rate": 1.960133058724917e-05, + "loss": 0.9529, + "step": 9539 + }, + { + "epoch": 0.7, + "learning_rate": 1.960121907768468e-05, + "loss": 1.058, + "step": 9540 + }, + { + "epoch": 0.7, + "learning_rate": 1.9601107552844795e-05, + "loss": 1.0244, + "step": 9541 + }, + { + "epoch": 0.71, + "learning_rate": 1.9600996012729682e-05, + "loss": 1.1087, + "step": 9542 + }, + { + "epoch": 0.71, + "learning_rate": 1.960088445733952e-05, + "loss": 0.9717, + "step": 9543 + }, + { + "epoch": 0.71, + "learning_rate": 1.960077288667449e-05, + "loss": 1.0129, + "step": 9544 + }, + { + "epoch": 0.71, + "learning_rate": 1.9600661300734766e-05, + "loss": 1.0583, + "step": 9545 + }, + { + "epoch": 0.71, + "learning_rate": 1.960054969952053e-05, + "loss": 1.0786, + "step": 9546 + }, + { + "epoch": 0.71, + "learning_rate": 1.9600438083031953e-05, + "loss": 0.9856, + "step": 9547 + }, + { + "epoch": 0.71, + "learning_rate": 1.960032645126922e-05, + "loss": 1.027, + "step": 9548 + }, + { + "epoch": 0.71, + "learning_rate": 1.9600214804232505e-05, + "loss": 1.0545, + "step": 9549 + }, + { + "epoch": 0.71, + "learning_rate": 1.9600103141921984e-05, + "loss": 1.01, + "step": 9550 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599991464337838e-05, + "loss": 1.1656, + "step": 9551 + }, + { + "epoch": 0.71, + "learning_rate": 1.959987977148024e-05, + "loss": 1.0321, + "step": 9552 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599768063349375e-05, + "loss": 1.0331, + "step": 9553 + }, + { + "epoch": 0.71, + "learning_rate": 1.959965633994541e-05, + "loss": 1.0436, + "step": 9554 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599544601268534e-05, + "loss": 0.9953, + "step": 9555 + }, + { + "epoch": 0.71, + "learning_rate": 1.959943284731892e-05, + "loss": 0.9889, + "step": 9556 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599321078096744e-05, + "loss": 0.9445, + "step": 9557 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599209293602187e-05, + "loss": 1.0837, + "step": 9558 + }, + { + "epoch": 0.71, + "learning_rate": 1.9599097493835423e-05, + "loss": 1.0654, + "step": 9559 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598985678796634e-05, + "loss": 1.0341, + "step": 9560 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598873848485994e-05, + "loss": 1.1075, + "step": 9561 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598762002903685e-05, + "loss": 1.0237, + "step": 9562 + }, + { + "epoch": 0.71, + "learning_rate": 1.959865014204988e-05, + "loss": 0.9967, + "step": 9563 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598538265924763e-05, + "loss": 1.002, + "step": 9564 + }, + { + "epoch": 0.71, + "learning_rate": 1.959842637452851e-05, + "loss": 1.0841, + "step": 9565 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598314467861298e-05, + "loss": 1.0571, + "step": 9566 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598202545923302e-05, + "loss": 1.005, + "step": 9567 + }, + { + "epoch": 0.71, + "learning_rate": 1.9598090608714705e-05, + "loss": 0.9364, + "step": 9568 + }, + { + "epoch": 0.71, + "learning_rate": 1.959797865623568e-05, + "loss": 1.0225, + "step": 9569 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597866688486412e-05, + "loss": 0.985, + "step": 9570 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597754705467075e-05, + "loss": 0.9999, + "step": 9571 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597642707177845e-05, + "loss": 1.0038, + "step": 9572 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597530693618908e-05, + "loss": 1.032, + "step": 9573 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597418664790433e-05, + "loss": 1.0395, + "step": 9574 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597306620692604e-05, + "loss": 0.9215, + "step": 9575 + }, + { + "epoch": 0.71, + "learning_rate": 1.9597194561325596e-05, + "loss": 0.9601, + "step": 9576 + }, + { + "epoch": 0.71, + "learning_rate": 1.959708248668959e-05, + "loss": 1.0256, + "step": 9577 + }, + { + "epoch": 0.71, + "learning_rate": 1.959697039678476e-05, + "loss": 0.9903, + "step": 9578 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596858291611293e-05, + "loss": 1.1016, + "step": 9579 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596746171169357e-05, + "loss": 1.1535, + "step": 9580 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596634035459138e-05, + "loss": 1.0567, + "step": 9581 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596521884480808e-05, + "loss": 0.9334, + "step": 9582 + }, + { + "epoch": 0.71, + "learning_rate": 1.959640971823455e-05, + "loss": 1.0392, + "step": 9583 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596297536720542e-05, + "loss": 1.1008, + "step": 9584 + }, + { + "epoch": 0.71, + "learning_rate": 1.9596185339938963e-05, + "loss": 1.0463, + "step": 9585 + }, + { + "epoch": 0.71, + "learning_rate": 1.959607312788999e-05, + "loss": 1.0656, + "step": 9586 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595960900573802e-05, + "loss": 1.11, + "step": 9587 + }, + { + "epoch": 0.71, + "learning_rate": 1.959584865799058e-05, + "loss": 1.0819, + "step": 9588 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595736400140497e-05, + "loss": 0.9861, + "step": 9589 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595624127023732e-05, + "loss": 0.9484, + "step": 9590 + }, + { + "epoch": 0.71, + "learning_rate": 1.959551183864047e-05, + "loss": 0.9617, + "step": 9591 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595399534990888e-05, + "loss": 1.0249, + "step": 9592 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595287216075157e-05, + "loss": 1.0785, + "step": 9593 + }, + { + "epoch": 0.71, + "learning_rate": 1.9595174881893466e-05, + "loss": 1.0073, + "step": 9594 + }, + { + "epoch": 0.71, + "learning_rate": 1.959506253244599e-05, + "loss": 0.8973, + "step": 9595 + }, + { + "epoch": 0.71, + "learning_rate": 1.95949501677329e-05, + "loss": 0.9595, + "step": 9596 + }, + { + "epoch": 0.71, + "learning_rate": 1.9594837787754385e-05, + "loss": 1.0371, + "step": 9597 + }, + { + "epoch": 0.71, + "learning_rate": 1.9594725392510625e-05, + "loss": 1.1292, + "step": 9598 + }, + { + "epoch": 0.71, + "learning_rate": 1.959461298200179e-05, + "loss": 0.9319, + "step": 9599 + }, + { + "epoch": 0.71, + "learning_rate": 1.959450055622806e-05, + "loss": 1.0117, + "step": 9600 + }, + { + "epoch": 0.71, + "learning_rate": 1.9594388115189624e-05, + "loss": 1.0117, + "step": 9601 + }, + { + "epoch": 0.71, + "learning_rate": 1.9594275658886646e-05, + "loss": 0.9731, + "step": 9602 + }, + { + "epoch": 0.71, + "learning_rate": 1.959416318731932e-05, + "loss": 0.9913, + "step": 9603 + }, + { + "epoch": 0.71, + "learning_rate": 1.9594050700487814e-05, + "loss": 0.9563, + "step": 9604 + }, + { + "epoch": 0.71, + "learning_rate": 1.959393819839231e-05, + "loss": 1.0352, + "step": 9605 + }, + { + "epoch": 0.71, + "learning_rate": 1.959382568103299e-05, + "loss": 1.0566, + "step": 9606 + }, + { + "epoch": 0.71, + "learning_rate": 1.9593713148410026e-05, + "loss": 1.098, + "step": 9607 + }, + { + "epoch": 0.71, + "learning_rate": 1.9593600600523607e-05, + "loss": 1.0357, + "step": 9608 + }, + { + "epoch": 0.71, + "learning_rate": 1.9593488037373904e-05, + "loss": 1.0646, + "step": 9609 + }, + { + "epoch": 0.71, + "learning_rate": 1.95933754589611e-05, + "loss": 1.0441, + "step": 9610 + }, + { + "epoch": 0.71, + "learning_rate": 1.9593262865285373e-05, + "loss": 1.1844, + "step": 9611 + }, + { + "epoch": 0.71, + "learning_rate": 1.95931502563469e-05, + "loss": 1.0435, + "step": 9612 + }, + { + "epoch": 0.71, + "learning_rate": 1.9593037632145864e-05, + "loss": 1.0177, + "step": 9613 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592924992682444e-05, + "loss": 1.0566, + "step": 9614 + }, + { + "epoch": 0.71, + "learning_rate": 1.959281233795681e-05, + "loss": 1.063, + "step": 9615 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592699667969156e-05, + "loss": 1.0908, + "step": 9616 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592586982719654e-05, + "loss": 1.0422, + "step": 9617 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592474282208482e-05, + "loss": 0.9882, + "step": 9618 + }, + { + "epoch": 0.71, + "learning_rate": 1.959236156643582e-05, + "loss": 1.0808, + "step": 9619 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592248835401846e-05, + "loss": 1.1353, + "step": 9620 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592136089106745e-05, + "loss": 0.964, + "step": 9621 + }, + { + "epoch": 0.71, + "learning_rate": 1.9592023327550694e-05, + "loss": 1.0194, + "step": 9622 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591910550733865e-05, + "loss": 0.9839, + "step": 9623 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591797758656446e-05, + "loss": 1.0477, + "step": 9624 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591684951318617e-05, + "loss": 1.0842, + "step": 9625 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591572128720552e-05, + "loss": 1.009, + "step": 9626 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591459290862435e-05, + "loss": 1.131, + "step": 9627 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591346437744444e-05, + "loss": 1.0803, + "step": 9628 + }, + { + "epoch": 0.71, + "learning_rate": 1.9591233569366755e-05, + "loss": 0.9787, + "step": 9629 + }, + { + "epoch": 0.71, + "learning_rate": 1.959112068572955e-05, + "loss": 1.0417, + "step": 9630 + }, + { + "epoch": 0.71, + "learning_rate": 1.959100778683301e-05, + "loss": 0.9332, + "step": 9631 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590894872677316e-05, + "loss": 0.9702, + "step": 9632 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590781943262643e-05, + "loss": 1.0071, + "step": 9633 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590668998589176e-05, + "loss": 1.0479, + "step": 9634 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590556038657088e-05, + "loss": 1.027, + "step": 9635 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590443063466565e-05, + "loss": 1.1313, + "step": 9636 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590330073017784e-05, + "loss": 1.0531, + "step": 9637 + }, + { + "epoch": 0.71, + "learning_rate": 1.9590217067310927e-05, + "loss": 0.991, + "step": 9638 + }, + { + "epoch": 0.71, + "learning_rate": 1.959010404634617e-05, + "loss": 0.9973, + "step": 9639 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589991010123693e-05, + "loss": 0.966, + "step": 9640 + }, + { + "epoch": 0.71, + "learning_rate": 1.958987795864368e-05, + "loss": 0.9794, + "step": 9641 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589764891906308e-05, + "loss": 1.0389, + "step": 9642 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589651809911755e-05, + "loss": 1.1299, + "step": 9643 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589538712660206e-05, + "loss": 1.079, + "step": 9644 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589425600151836e-05, + "loss": 1.0597, + "step": 9645 + }, + { + "epoch": 0.71, + "learning_rate": 1.958931247238683e-05, + "loss": 1.0741, + "step": 9646 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589199329365363e-05, + "loss": 1.0938, + "step": 9647 + }, + { + "epoch": 0.71, + "learning_rate": 1.9589086171087618e-05, + "loss": 1.0577, + "step": 9648 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588972997553773e-05, + "loss": 1.0429, + "step": 9649 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588859808764008e-05, + "loss": 1.096, + "step": 9650 + }, + { + "epoch": 0.71, + "learning_rate": 1.958874660471851e-05, + "loss": 0.936, + "step": 9651 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588633385417444e-05, + "loss": 0.9937, + "step": 9652 + }, + { + "epoch": 0.71, + "learning_rate": 1.958852015086101e-05, + "loss": 0.9098, + "step": 9653 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588406901049367e-05, + "loss": 0.9802, + "step": 9654 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588293635982712e-05, + "loss": 1.0508, + "step": 9655 + }, + { + "epoch": 0.71, + "learning_rate": 1.958818035566122e-05, + "loss": 1.0137, + "step": 9656 + }, + { + "epoch": 0.71, + "learning_rate": 1.9588067060085068e-05, + "loss": 0.9152, + "step": 9657 + }, + { + "epoch": 0.71, + "learning_rate": 1.958795374925444e-05, + "loss": 1.0715, + "step": 9658 + }, + { + "epoch": 0.71, + "learning_rate": 1.958784042316951e-05, + "loss": 1.0766, + "step": 9659 + }, + { + "epoch": 0.71, + "learning_rate": 1.9587727081830467e-05, + "loss": 1.1005, + "step": 9660 + }, + { + "epoch": 0.71, + "learning_rate": 1.9587613725237488e-05, + "loss": 1.0744, + "step": 9661 + }, + { + "epoch": 0.71, + "learning_rate": 1.958750035339075e-05, + "loss": 0.9712, + "step": 9662 + }, + { + "epoch": 0.71, + "learning_rate": 1.958738696629044e-05, + "loss": 1.0173, + "step": 9663 + }, + { + "epoch": 0.71, + "learning_rate": 1.9587273563936732e-05, + "loss": 1.0159, + "step": 9664 + }, + { + "epoch": 0.71, + "learning_rate": 1.9587160146329812e-05, + "loss": 1.1182, + "step": 9665 + }, + { + "epoch": 0.71, + "learning_rate": 1.9587046713469854e-05, + "loss": 0.9778, + "step": 9666 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586933265357042e-05, + "loss": 0.9152, + "step": 9667 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586819801991556e-05, + "loss": 1.0062, + "step": 9668 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586706323373577e-05, + "loss": 1.0567, + "step": 9669 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586592829503287e-05, + "loss": 1.0739, + "step": 9670 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586479320380863e-05, + "loss": 1.1272, + "step": 9671 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586365796006487e-05, + "loss": 0.8961, + "step": 9672 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586252256380344e-05, + "loss": 1.0242, + "step": 9673 + }, + { + "epoch": 0.71, + "learning_rate": 1.958613870150261e-05, + "loss": 0.9859, + "step": 9674 + }, + { + "epoch": 0.71, + "learning_rate": 1.9586025131373465e-05, + "loss": 1.0202, + "step": 9675 + }, + { + "epoch": 0.71, + "learning_rate": 1.958591154599309e-05, + "loss": 0.9936, + "step": 9676 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585797945361668e-05, + "loss": 1.1352, + "step": 9677 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585684329479378e-05, + "loss": 1.0125, + "step": 9678 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585570698346402e-05, + "loss": 0.9799, + "step": 9679 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585457051962923e-05, + "loss": 1.0588, + "step": 9680 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585343390329114e-05, + "loss": 1.0146, + "step": 9681 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585229713445165e-05, + "loss": 0.9226, + "step": 9682 + }, + { + "epoch": 0.72, + "learning_rate": 1.958511602131125e-05, + "loss": 1.011, + "step": 9683 + }, + { + "epoch": 0.72, + "learning_rate": 1.9585002313927553e-05, + "loss": 1.03, + "step": 9684 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584888591294256e-05, + "loss": 1.1568, + "step": 9685 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584774853411537e-05, + "loss": 1.0949, + "step": 9686 + }, + { + "epoch": 0.72, + "learning_rate": 1.958466110027958e-05, + "loss": 0.8733, + "step": 9687 + }, + { + "epoch": 0.72, + "learning_rate": 1.958454733189856e-05, + "loss": 1.1621, + "step": 9688 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584433548268667e-05, + "loss": 1.0078, + "step": 9689 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584319749390076e-05, + "loss": 1.0153, + "step": 9690 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584205935262967e-05, + "loss": 1.0519, + "step": 9691 + }, + { + "epoch": 0.72, + "learning_rate": 1.9584092105887527e-05, + "loss": 1.0354, + "step": 9692 + }, + { + "epoch": 0.72, + "learning_rate": 1.958397826126393e-05, + "loss": 1.1387, + "step": 9693 + }, + { + "epoch": 0.72, + "learning_rate": 1.958386440139236e-05, + "loss": 0.9352, + "step": 9694 + }, + { + "epoch": 0.72, + "learning_rate": 1.9583750526273005e-05, + "loss": 1.0318, + "step": 9695 + }, + { + "epoch": 0.72, + "learning_rate": 1.958363663590604e-05, + "loss": 0.99, + "step": 9696 + }, + { + "epoch": 0.72, + "learning_rate": 1.958352273029164e-05, + "loss": 1.0921, + "step": 9697 + }, + { + "epoch": 0.72, + "learning_rate": 1.9583408809429995e-05, + "loss": 1.0325, + "step": 9698 + }, + { + "epoch": 0.72, + "learning_rate": 1.9583294873321285e-05, + "loss": 1.0054, + "step": 9699 + }, + { + "epoch": 0.72, + "learning_rate": 1.958318092196569e-05, + "loss": 1.0728, + "step": 9700 + }, + { + "epoch": 0.72, + "learning_rate": 1.958306695536339e-05, + "loss": 1.0989, + "step": 9701 + }, + { + "epoch": 0.72, + "learning_rate": 1.9582952973514567e-05, + "loss": 0.972, + "step": 9702 + }, + { + "epoch": 0.72, + "learning_rate": 1.9582838976419404e-05, + "loss": 1.108, + "step": 9703 + }, + { + "epoch": 0.72, + "learning_rate": 1.9582724964078078e-05, + "loss": 1.0563, + "step": 9704 + }, + { + "epoch": 0.72, + "learning_rate": 1.958261093649078e-05, + "loss": 1.1112, + "step": 9705 + }, + { + "epoch": 0.72, + "learning_rate": 1.958249689365768e-05, + "loss": 0.9638, + "step": 9706 + }, + { + "epoch": 0.72, + "learning_rate": 1.9582382835578966e-05, + "loss": 1.1351, + "step": 9707 + }, + { + "epoch": 0.72, + "learning_rate": 1.958226876225482e-05, + "loss": 1.0658, + "step": 9708 + }, + { + "epoch": 0.72, + "learning_rate": 1.958215467368542e-05, + "loss": 1.0206, + "step": 9709 + }, + { + "epoch": 0.72, + "learning_rate": 1.958204056987095e-05, + "loss": 1.1067, + "step": 9710 + }, + { + "epoch": 0.72, + "learning_rate": 1.958192645081159e-05, + "loss": 1.0558, + "step": 9711 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581812316507517e-05, + "loss": 0.9721, + "step": 9712 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581698166958926e-05, + "loss": 1.0158, + "step": 9713 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581584002165984e-05, + "loss": 1.0813, + "step": 9714 + }, + { + "epoch": 0.72, + "learning_rate": 1.958146982212888e-05, + "loss": 0.9653, + "step": 9715 + }, + { + "epoch": 0.72, + "learning_rate": 1.95813556268478e-05, + "loss": 1.0192, + "step": 9716 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581241416322915e-05, + "loss": 1.0799, + "step": 9717 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581127190554415e-05, + "loss": 1.0211, + "step": 9718 + }, + { + "epoch": 0.72, + "learning_rate": 1.9581012949542474e-05, + "loss": 1.002, + "step": 9719 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580898693287284e-05, + "loss": 1.0672, + "step": 9720 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580784421789018e-05, + "loss": 1.0607, + "step": 9721 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580670135047862e-05, + "loss": 1.0882, + "step": 9722 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580555833063995e-05, + "loss": 1.0684, + "step": 9723 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580441515837605e-05, + "loss": 0.9669, + "step": 9724 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580327183368867e-05, + "loss": 1.0347, + "step": 9725 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580212835657964e-05, + "loss": 1.0028, + "step": 9726 + }, + { + "epoch": 0.72, + "learning_rate": 1.9580098472705083e-05, + "loss": 0.9705, + "step": 9727 + }, + { + "epoch": 0.72, + "learning_rate": 1.95799840945104e-05, + "loss": 1.0897, + "step": 9728 + }, + { + "epoch": 0.72, + "learning_rate": 1.95798697010741e-05, + "loss": 1.0247, + "step": 9729 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579755292396365e-05, + "loss": 1.0506, + "step": 9730 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579640868477376e-05, + "loss": 1.0662, + "step": 9731 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579526429317313e-05, + "loss": 1.1056, + "step": 9732 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579411974916363e-05, + "loss": 0.9854, + "step": 9733 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579297505274705e-05, + "loss": 1.0015, + "step": 9734 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579183020392522e-05, + "loss": 1.1658, + "step": 9735 + }, + { + "epoch": 0.72, + "learning_rate": 1.9579068520269997e-05, + "loss": 1.0782, + "step": 9736 + }, + { + "epoch": 0.72, + "learning_rate": 1.957895400490731e-05, + "loss": 1.0962, + "step": 9737 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578839474304647e-05, + "loss": 1.0767, + "step": 9738 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578724928462184e-05, + "loss": 1.1565, + "step": 9739 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578610367380104e-05, + "loss": 0.9536, + "step": 9740 + }, + { + "epoch": 0.72, + "learning_rate": 1.95784957910586e-05, + "loss": 1.0312, + "step": 9741 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578381199497838e-05, + "loss": 1.1212, + "step": 9742 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578266592698012e-05, + "loss": 0.9143, + "step": 9743 + }, + { + "epoch": 0.72, + "learning_rate": 1.95781519706593e-05, + "loss": 0.9852, + "step": 9744 + }, + { + "epoch": 0.72, + "learning_rate": 1.9578037333381888e-05, + "loss": 0.9542, + "step": 9745 + }, + { + "epoch": 0.72, + "learning_rate": 1.9577922680865952e-05, + "loss": 1.123, + "step": 9746 + }, + { + "epoch": 0.72, + "learning_rate": 1.957780801311168e-05, + "loss": 0.9907, + "step": 9747 + }, + { + "epoch": 0.72, + "learning_rate": 1.957769333011925e-05, + "loss": 1.0339, + "step": 9748 + }, + { + "epoch": 0.72, + "learning_rate": 1.9577578631888852e-05, + "loss": 1.0574, + "step": 9749 + }, + { + "epoch": 0.72, + "learning_rate": 1.9577463918420663e-05, + "loss": 0.9204, + "step": 9750 + }, + { + "epoch": 0.72, + "learning_rate": 1.9577349189714862e-05, + "loss": 1.0783, + "step": 9751 + }, + { + "epoch": 0.72, + "learning_rate": 1.9577234445771637e-05, + "loss": 1.0056, + "step": 9752 + }, + { + "epoch": 0.72, + "learning_rate": 1.957711968659117e-05, + "loss": 0.875, + "step": 9753 + }, + { + "epoch": 0.72, + "learning_rate": 1.957700491217364e-05, + "loss": 1.0361, + "step": 9754 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576890122519235e-05, + "loss": 0.9513, + "step": 9755 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576775317628133e-05, + "loss": 1.0228, + "step": 9756 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576660497500518e-05, + "loss": 1.0262, + "step": 9757 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576545662136576e-05, + "loss": 1.0562, + "step": 9758 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576430811536483e-05, + "loss": 1.0268, + "step": 9759 + }, + { + "epoch": 0.72, + "learning_rate": 1.957631594570043e-05, + "loss": 1.0253, + "step": 9760 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576201064628594e-05, + "loss": 1.1191, + "step": 9761 + }, + { + "epoch": 0.72, + "learning_rate": 1.9576086168321156e-05, + "loss": 0.9163, + "step": 9762 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575971256778307e-05, + "loss": 1.0226, + "step": 9763 + }, + { + "epoch": 0.72, + "learning_rate": 1.957585633000022e-05, + "loss": 0.9737, + "step": 9764 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575741387987086e-05, + "loss": 1.1724, + "step": 9765 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575626430739083e-05, + "loss": 0.9832, + "step": 9766 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575511458256394e-05, + "loss": 1.0815, + "step": 9767 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575396470539204e-05, + "loss": 1.0711, + "step": 9768 + }, + { + "epoch": 0.72, + "learning_rate": 1.95752814675877e-05, + "loss": 1.0072, + "step": 9769 + }, + { + "epoch": 0.72, + "learning_rate": 1.9575166449402053e-05, + "loss": 1.0369, + "step": 9770 + }, + { + "epoch": 0.72, + "learning_rate": 1.957505141598246e-05, + "loss": 1.0083, + "step": 9771 + }, + { + "epoch": 0.72, + "learning_rate": 1.957493636732909e-05, + "loss": 0.9342, + "step": 9772 + }, + { + "epoch": 0.72, + "learning_rate": 1.957482130344214e-05, + "loss": 1.0272, + "step": 9773 + }, + { + "epoch": 0.72, + "learning_rate": 1.9574706224321782e-05, + "loss": 0.9847, + "step": 9774 + }, + { + "epoch": 0.72, + "learning_rate": 1.9574591129968203e-05, + "loss": 0.9686, + "step": 9775 + }, + { + "epoch": 0.72, + "learning_rate": 1.9574476020381586e-05, + "loss": 0.9954, + "step": 9776 + }, + { + "epoch": 0.72, + "learning_rate": 1.957436089556212e-05, + "loss": 1.0239, + "step": 9777 + }, + { + "epoch": 0.72, + "learning_rate": 1.9574245755509976e-05, + "loss": 0.9521, + "step": 9778 + }, + { + "epoch": 0.72, + "learning_rate": 1.957413060022535e-05, + "loss": 0.9962, + "step": 9779 + }, + { + "epoch": 0.72, + "learning_rate": 1.9574015429708417e-05, + "loss": 1.1359, + "step": 9780 + }, + { + "epoch": 0.72, + "learning_rate": 1.9573900243959362e-05, + "loss": 1.0072, + "step": 9781 + }, + { + "epoch": 0.72, + "learning_rate": 1.9573785042978368e-05, + "loss": 1.1442, + "step": 9782 + }, + { + "epoch": 0.72, + "learning_rate": 1.957366982676562e-05, + "loss": 0.9987, + "step": 9783 + }, + { + "epoch": 0.72, + "learning_rate": 1.95735545953213e-05, + "loss": 1.0802, + "step": 9784 + }, + { + "epoch": 0.72, + "learning_rate": 1.957343934864559e-05, + "loss": 0.9884, + "step": 9785 + }, + { + "epoch": 0.72, + "learning_rate": 1.9573324086738678e-05, + "loss": 1.0638, + "step": 9786 + }, + { + "epoch": 0.72, + "learning_rate": 1.957320880960074e-05, + "loss": 1.1308, + "step": 9787 + }, + { + "epoch": 0.72, + "learning_rate": 1.957309351723197e-05, + "loss": 1.0054, + "step": 9788 + }, + { + "epoch": 0.72, + "learning_rate": 1.957297820963254e-05, + "loss": 1.0445, + "step": 9789 + }, + { + "epoch": 0.72, + "learning_rate": 1.957286288680264e-05, + "loss": 1.0745, + "step": 9790 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572747548742453e-05, + "loss": 0.9901, + "step": 9791 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572632195452162e-05, + "loss": 1.0256, + "step": 9792 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572516826931947e-05, + "loss": 0.9109, + "step": 9793 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572401443182e-05, + "loss": 0.9681, + "step": 9794 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572286044202496e-05, + "loss": 0.9449, + "step": 9795 + }, + { + "epoch": 0.72, + "learning_rate": 1.9572170629993623e-05, + "loss": 1.0694, + "step": 9796 + }, + { + "epoch": 0.72, + "learning_rate": 1.957205520055556e-05, + "loss": 1.0616, + "step": 9797 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571939755888497e-05, + "loss": 1.0103, + "step": 9798 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571824295992618e-05, + "loss": 1.0641, + "step": 9799 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571708820868096e-05, + "loss": 0.9489, + "step": 9800 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571593330515127e-05, + "loss": 1.0387, + "step": 9801 + }, + { + "epoch": 0.72, + "learning_rate": 1.957147782493389e-05, + "loss": 1.0867, + "step": 9802 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571362304124567e-05, + "loss": 1.0564, + "step": 9803 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571246768087345e-05, + "loss": 1.0491, + "step": 9804 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571131216822404e-05, + "loss": 1.0866, + "step": 9805 + }, + { + "epoch": 0.72, + "learning_rate": 1.9571015650329936e-05, + "loss": 1.0633, + "step": 9806 + }, + { + "epoch": 0.72, + "learning_rate": 1.9570900068610112e-05, + "loss": 1.1603, + "step": 9807 + }, + { + "epoch": 0.72, + "learning_rate": 1.9570784471663126e-05, + "loss": 1.1213, + "step": 9808 + }, + { + "epoch": 0.72, + "learning_rate": 1.9570668859489158e-05, + "loss": 1.0585, + "step": 9809 + }, + { + "epoch": 0.72, + "learning_rate": 1.9570553232088393e-05, + "loss": 1.1698, + "step": 9810 + }, + { + "epoch": 0.72, + "learning_rate": 1.957043758946102e-05, + "loss": 1.0245, + "step": 9811 + }, + { + "epoch": 0.73, + "learning_rate": 1.957032193160721e-05, + "loss": 0.992, + "step": 9812 + }, + { + "epoch": 0.73, + "learning_rate": 1.957020625852716e-05, + "loss": 1.0405, + "step": 9813 + }, + { + "epoch": 0.73, + "learning_rate": 1.9570090570221042e-05, + "loss": 1.058, + "step": 9814 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569974866689054e-05, + "loss": 1.0328, + "step": 9815 + }, + { + "epoch": 0.73, + "learning_rate": 1.956985914793137e-05, + "loss": 0.9639, + "step": 9816 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569743413948177e-05, + "loss": 1.1141, + "step": 9817 + }, + { + "epoch": 0.73, + "learning_rate": 1.956962766473966e-05, + "loss": 1.0005, + "step": 9818 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569511900306003e-05, + "loss": 0.8843, + "step": 9819 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569396120647387e-05, + "loss": 1.0199, + "step": 9820 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569280325764e-05, + "loss": 0.9645, + "step": 9821 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569164515656024e-05, + "loss": 1.0707, + "step": 9822 + }, + { + "epoch": 0.73, + "learning_rate": 1.9569048690323646e-05, + "loss": 1.0223, + "step": 9823 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568932849767047e-05, + "loss": 1.0368, + "step": 9824 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568816993986413e-05, + "loss": 0.942, + "step": 9825 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568701122981927e-05, + "loss": 1.056, + "step": 9826 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568585236753776e-05, + "loss": 1.0748, + "step": 9827 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568469335302142e-05, + "loss": 0.9898, + "step": 9828 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568353418627212e-05, + "loss": 0.9196, + "step": 9829 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568237486729168e-05, + "loss": 1.0088, + "step": 9830 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568121539608194e-05, + "loss": 0.9848, + "step": 9831 + }, + { + "epoch": 0.73, + "learning_rate": 1.9568005577264477e-05, + "loss": 0.9639, + "step": 9832 + }, + { + "epoch": 0.73, + "learning_rate": 1.9567889599698198e-05, + "loss": 1.0523, + "step": 9833 + }, + { + "epoch": 0.73, + "learning_rate": 1.9567773606909544e-05, + "loss": 1.0728, + "step": 9834 + }, + { + "epoch": 0.73, + "learning_rate": 1.95676575988987e-05, + "loss": 1.0136, + "step": 9835 + }, + { + "epoch": 0.73, + "learning_rate": 1.956754157566585e-05, + "loss": 1.0175, + "step": 9836 + }, + { + "epoch": 0.73, + "learning_rate": 1.9567425537211175e-05, + "loss": 1.0613, + "step": 9837 + }, + { + "epoch": 0.73, + "learning_rate": 1.9567309483534863e-05, + "loss": 1.0069, + "step": 9838 + }, + { + "epoch": 0.73, + "learning_rate": 1.9567193414637104e-05, + "loss": 1.1017, + "step": 9839 + }, + { + "epoch": 0.73, + "learning_rate": 1.956707733051807e-05, + "loss": 0.9602, + "step": 9840 + }, + { + "epoch": 0.73, + "learning_rate": 1.9566961231177962e-05, + "loss": 0.9782, + "step": 9841 + }, + { + "epoch": 0.73, + "learning_rate": 1.956684511661695e-05, + "loss": 1.0361, + "step": 9842 + }, + { + "epoch": 0.73, + "learning_rate": 1.956672898683522e-05, + "loss": 1.0006, + "step": 9843 + }, + { + "epoch": 0.73, + "learning_rate": 1.9566612841832967e-05, + "loss": 1.0526, + "step": 9844 + }, + { + "epoch": 0.73, + "learning_rate": 1.956649668161037e-05, + "loss": 1.0204, + "step": 9845 + }, + { + "epoch": 0.73, + "learning_rate": 1.956638050616761e-05, + "loss": 1.0311, + "step": 9846 + }, + { + "epoch": 0.73, + "learning_rate": 1.956626431550488e-05, + "loss": 0.8477, + "step": 9847 + }, + { + "epoch": 0.73, + "learning_rate": 1.9566148109622357e-05, + "loss": 1.0813, + "step": 9848 + }, + { + "epoch": 0.73, + "learning_rate": 1.9566031888520228e-05, + "loss": 1.0156, + "step": 9849 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565915652198682e-05, + "loss": 0.9385, + "step": 9850 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565799400657903e-05, + "loss": 1.0533, + "step": 9851 + }, + { + "epoch": 0.73, + "learning_rate": 1.956568313389807e-05, + "loss": 0.9829, + "step": 9852 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565566851919377e-05, + "loss": 1.0821, + "step": 9853 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565450554721997e-05, + "loss": 0.9431, + "step": 9854 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565334242306128e-05, + "loss": 0.9977, + "step": 9855 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565217914671947e-05, + "loss": 1.0055, + "step": 9856 + }, + { + "epoch": 0.73, + "learning_rate": 1.9565101571819636e-05, + "loss": 1.0678, + "step": 9857 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564985213749393e-05, + "loss": 1.0887, + "step": 9858 + }, + { + "epoch": 0.73, + "learning_rate": 1.956486884046139e-05, + "loss": 1.1012, + "step": 9859 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564752451955823e-05, + "loss": 0.9372, + "step": 9860 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564636048232866e-05, + "loss": 1.1048, + "step": 9861 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564519629292714e-05, + "loss": 0.9408, + "step": 9862 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564403195135547e-05, + "loss": 1.0545, + "step": 9863 + }, + { + "epoch": 0.73, + "learning_rate": 1.956428674576155e-05, + "loss": 1.0748, + "step": 9864 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564170281170913e-05, + "loss": 1.0786, + "step": 9865 + }, + { + "epoch": 0.73, + "learning_rate": 1.9564053801363813e-05, + "loss": 1.0634, + "step": 9866 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563937306340445e-05, + "loss": 1.0886, + "step": 9867 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563820796100984e-05, + "loss": 1.1266, + "step": 9868 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563704270645625e-05, + "loss": 0.9936, + "step": 9869 + }, + { + "epoch": 0.73, + "learning_rate": 1.956358772997455e-05, + "loss": 1.0816, + "step": 9870 + }, + { + "epoch": 0.73, + "learning_rate": 1.956347117408794e-05, + "loss": 1.1058, + "step": 9871 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563354602985988e-05, + "loss": 0.9967, + "step": 9872 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563238016668877e-05, + "loss": 0.9466, + "step": 9873 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563121415136783e-05, + "loss": 1.0738, + "step": 9874 + }, + { + "epoch": 0.73, + "learning_rate": 1.9563004798389905e-05, + "loss": 1.0411, + "step": 9875 + }, + { + "epoch": 0.73, + "learning_rate": 1.9562888166428428e-05, + "loss": 1.1739, + "step": 9876 + }, + { + "epoch": 0.73, + "learning_rate": 1.9562771519252527e-05, + "loss": 0.9552, + "step": 9877 + }, + { + "epoch": 0.73, + "learning_rate": 1.956265485686239e-05, + "loss": 1.0744, + "step": 9878 + }, + { + "epoch": 0.73, + "learning_rate": 1.956253817925821e-05, + "loss": 1.0079, + "step": 9879 + }, + { + "epoch": 0.73, + "learning_rate": 1.956242148644017e-05, + "loss": 1.005, + "step": 9880 + }, + { + "epoch": 0.73, + "learning_rate": 1.9562304778408454e-05, + "loss": 1.043, + "step": 9881 + }, + { + "epoch": 0.73, + "learning_rate": 1.956218805516325e-05, + "loss": 0.9521, + "step": 9882 + }, + { + "epoch": 0.73, + "learning_rate": 1.9562071316704736e-05, + "loss": 0.8187, + "step": 9883 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561954563033104e-05, + "loss": 1.0432, + "step": 9884 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561837794148544e-05, + "loss": 0.9572, + "step": 9885 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561721010051232e-05, + "loss": 1.0692, + "step": 9886 + }, + { + "epoch": 0.73, + "learning_rate": 1.956160421074136e-05, + "loss": 1.0094, + "step": 9887 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561487396219112e-05, + "loss": 1.0191, + "step": 9888 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561370566484678e-05, + "loss": 1.019, + "step": 9889 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561253721538236e-05, + "loss": 0.9057, + "step": 9890 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561136861379978e-05, + "loss": 0.862, + "step": 9891 + }, + { + "epoch": 0.73, + "learning_rate": 1.9561019986010088e-05, + "loss": 0.9938, + "step": 9892 + }, + { + "epoch": 0.73, + "learning_rate": 1.956090309542875e-05, + "loss": 0.9466, + "step": 9893 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560786189636155e-05, + "loss": 0.9979, + "step": 9894 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560669268632483e-05, + "loss": 0.9742, + "step": 9895 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560552332417927e-05, + "loss": 1.0614, + "step": 9896 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560435380992668e-05, + "loss": 1.0492, + "step": 9897 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560318414356892e-05, + "loss": 1.0534, + "step": 9898 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560201432510783e-05, + "loss": 0.9853, + "step": 9899 + }, + { + "epoch": 0.73, + "learning_rate": 1.9560084435454535e-05, + "loss": 1.0579, + "step": 9900 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559967423188325e-05, + "loss": 1.049, + "step": 9901 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559850395712348e-05, + "loss": 0.9193, + "step": 9902 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559733353026783e-05, + "loss": 0.9425, + "step": 9903 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559616295131824e-05, + "loss": 1.0999, + "step": 9904 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559499222027646e-05, + "loss": 0.975, + "step": 9905 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559382133714442e-05, + "loss": 0.9711, + "step": 9906 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559265030192403e-05, + "loss": 1.0443, + "step": 9907 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559147911461708e-05, + "loss": 0.9806, + "step": 9908 + }, + { + "epoch": 0.73, + "learning_rate": 1.9559030777522543e-05, + "loss": 1.1043, + "step": 9909 + }, + { + "epoch": 0.73, + "learning_rate": 1.9558913628375097e-05, + "loss": 0.9449, + "step": 9910 + }, + { + "epoch": 0.73, + "learning_rate": 1.955879646401956e-05, + "loss": 1.001, + "step": 9911 + }, + { + "epoch": 0.73, + "learning_rate": 1.955867928445611e-05, + "loss": 0.9492, + "step": 9912 + }, + { + "epoch": 0.73, + "learning_rate": 1.955856208968494e-05, + "loss": 1.0584, + "step": 9913 + }, + { + "epoch": 0.73, + "learning_rate": 1.9558444879706236e-05, + "loss": 1.0707, + "step": 9914 + }, + { + "epoch": 0.73, + "learning_rate": 1.955832765452018e-05, + "loss": 0.989, + "step": 9915 + }, + { + "epoch": 0.73, + "learning_rate": 1.9558210414126964e-05, + "loss": 1.0687, + "step": 9916 + }, + { + "epoch": 0.73, + "learning_rate": 1.955809315852677e-05, + "loss": 1.0935, + "step": 9917 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557975887719785e-05, + "loss": 1.0897, + "step": 9918 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557858601706198e-05, + "loss": 0.9659, + "step": 9919 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557741300486198e-05, + "loss": 1.0781, + "step": 9920 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557623984059968e-05, + "loss": 1.0478, + "step": 9921 + }, + { + "epoch": 0.73, + "learning_rate": 1.955750665242769e-05, + "loss": 1.0476, + "step": 9922 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557389305589562e-05, + "loss": 1.0041, + "step": 9923 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557271943545762e-05, + "loss": 1.135, + "step": 9924 + }, + { + "epoch": 0.73, + "learning_rate": 1.9557154566296478e-05, + "loss": 0.9292, + "step": 9925 + }, + { + "epoch": 0.73, + "learning_rate": 1.95570371738419e-05, + "loss": 1.0713, + "step": 9926 + }, + { + "epoch": 0.73, + "learning_rate": 1.955691976618221e-05, + "loss": 0.9718, + "step": 9927 + }, + { + "epoch": 0.73, + "learning_rate": 1.95568023433176e-05, + "loss": 0.9438, + "step": 9928 + }, + { + "epoch": 0.73, + "learning_rate": 1.955668490524825e-05, + "loss": 1.0852, + "step": 9929 + }, + { + "epoch": 0.73, + "learning_rate": 1.9556567451974353e-05, + "loss": 0.9777, + "step": 9930 + }, + { + "epoch": 0.73, + "learning_rate": 1.9556449983496096e-05, + "loss": 1.0549, + "step": 9931 + }, + { + "epoch": 0.73, + "learning_rate": 1.9556332499813663e-05, + "loss": 1.0132, + "step": 9932 + }, + { + "epoch": 0.73, + "learning_rate": 1.9556215000927242e-05, + "loss": 1.0748, + "step": 9933 + }, + { + "epoch": 0.73, + "learning_rate": 1.9556097486837017e-05, + "loss": 1.099, + "step": 9934 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555979957543177e-05, + "loss": 1.0147, + "step": 9935 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555862413045913e-05, + "loss": 0.9852, + "step": 9936 + }, + { + "epoch": 0.73, + "learning_rate": 1.955574485334541e-05, + "loss": 1.0418, + "step": 9937 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555627278441848e-05, + "loss": 1.0828, + "step": 9938 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555509688335427e-05, + "loss": 1.0865, + "step": 9939 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555392083026323e-05, + "loss": 0.9465, + "step": 9940 + }, + { + "epoch": 0.73, + "learning_rate": 1.9555274462514726e-05, + "loss": 0.9594, + "step": 9941 + }, + { + "epoch": 0.73, + "learning_rate": 1.955515682680082e-05, + "loss": 0.915, + "step": 9942 + }, + { + "epoch": 0.73, + "learning_rate": 1.95550391758848e-05, + "loss": 1.098, + "step": 9943 + }, + { + "epoch": 0.73, + "learning_rate": 1.9554921509766856e-05, + "loss": 0.9869, + "step": 9944 + }, + { + "epoch": 0.73, + "learning_rate": 1.9554803828447163e-05, + "loss": 0.8874, + "step": 9945 + }, + { + "epoch": 0.73, + "learning_rate": 1.9554686131925913e-05, + "loss": 1.0649, + "step": 9946 + }, + { + "epoch": 0.73, + "learning_rate": 1.9554568420203294e-05, + "loss": 0.9727, + "step": 9947 + }, + { + "epoch": 0.74, + "learning_rate": 1.9554450693279493e-05, + "loss": 0.9542, + "step": 9948 + }, + { + "epoch": 0.74, + "learning_rate": 1.9554332951154697e-05, + "loss": 1.0668, + "step": 9949 + }, + { + "epoch": 0.74, + "learning_rate": 1.9554215193829098e-05, + "loss": 1.0037, + "step": 9950 + }, + { + "epoch": 0.74, + "learning_rate": 1.9554097421302877e-05, + "loss": 0.7921, + "step": 9951 + }, + { + "epoch": 0.74, + "learning_rate": 1.9553979633576222e-05, + "loss": 1.0123, + "step": 9952 + }, + { + "epoch": 0.74, + "learning_rate": 1.9553861830649327e-05, + "loss": 0.9822, + "step": 9953 + }, + { + "epoch": 0.74, + "learning_rate": 1.9553744012522368e-05, + "loss": 0.9848, + "step": 9954 + }, + { + "epoch": 0.74, + "learning_rate": 1.9553626179195542e-05, + "loss": 1.0997, + "step": 9955 + }, + { + "epoch": 0.74, + "learning_rate": 1.955350833066904e-05, + "loss": 1.1567, + "step": 9956 + }, + { + "epoch": 0.74, + "learning_rate": 1.955339046694303e-05, + "loss": 0.9505, + "step": 9957 + }, + { + "epoch": 0.74, + "learning_rate": 1.955327258801772e-05, + "loss": 0.9877, + "step": 9958 + }, + { + "epoch": 0.74, + "learning_rate": 1.955315469389329e-05, + "loss": 1.0141, + "step": 9959 + }, + { + "epoch": 0.74, + "learning_rate": 1.9553036784569928e-05, + "loss": 1.0893, + "step": 9960 + }, + { + "epoch": 0.74, + "learning_rate": 1.955291886004782e-05, + "loss": 0.9672, + "step": 9961 + }, + { + "epoch": 0.74, + "learning_rate": 1.9552800920327156e-05, + "loss": 1.1485, + "step": 9962 + }, + { + "epoch": 0.74, + "learning_rate": 1.955268296540812e-05, + "loss": 1.0206, + "step": 9963 + }, + { + "epoch": 0.74, + "learning_rate": 1.95525649952909e-05, + "loss": 1.0535, + "step": 9964 + }, + { + "epoch": 0.74, + "learning_rate": 1.955244700997569e-05, + "loss": 1.0278, + "step": 9965 + }, + { + "epoch": 0.74, + "learning_rate": 1.9552329009462674e-05, + "loss": 0.9389, + "step": 9966 + }, + { + "epoch": 0.74, + "learning_rate": 1.9552210993752037e-05, + "loss": 1.0203, + "step": 9967 + }, + { + "epoch": 0.74, + "learning_rate": 1.955209296284397e-05, + "loss": 1.0638, + "step": 9968 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551974916738657e-05, + "loss": 0.9556, + "step": 9969 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551856855436295e-05, + "loss": 1.1067, + "step": 9970 + }, + { + "epoch": 0.74, + "learning_rate": 1.955173877893706e-05, + "loss": 0.9546, + "step": 9971 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551620687241146e-05, + "loss": 1.0605, + "step": 9972 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551502580348743e-05, + "loss": 1.0433, + "step": 9973 + }, + { + "epoch": 0.74, + "learning_rate": 1.955138445826003e-05, + "loss": 1.041, + "step": 9974 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551266320975208e-05, + "loss": 0.8861, + "step": 9975 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551148168494455e-05, + "loss": 0.9816, + "step": 9976 + }, + { + "epoch": 0.74, + "learning_rate": 1.9551030000817958e-05, + "loss": 0.9924, + "step": 9977 + }, + { + "epoch": 0.74, + "learning_rate": 1.9550911817945913e-05, + "loss": 0.9767, + "step": 9978 + }, + { + "epoch": 0.74, + "learning_rate": 1.9550793619878504e-05, + "loss": 0.995, + "step": 9979 + }, + { + "epoch": 0.74, + "learning_rate": 1.955067540661592e-05, + "loss": 1.0481, + "step": 9980 + }, + { + "epoch": 0.74, + "learning_rate": 1.955055717815835e-05, + "loss": 0.9731, + "step": 9981 + }, + { + "epoch": 0.74, + "learning_rate": 1.9550438934505974e-05, + "loss": 0.9746, + "step": 9982 + }, + { + "epoch": 0.74, + "learning_rate": 1.9550320675658988e-05, + "loss": 1.0523, + "step": 9983 + }, + { + "epoch": 0.74, + "learning_rate": 1.955020240161758e-05, + "loss": 1.0159, + "step": 9984 + }, + { + "epoch": 0.74, + "learning_rate": 1.9550084112381937e-05, + "loss": 0.9739, + "step": 9985 + }, + { + "epoch": 0.74, + "learning_rate": 1.9549965807952245e-05, + "loss": 1.032, + "step": 9986 + }, + { + "epoch": 0.74, + "learning_rate": 1.9549847488328696e-05, + "loss": 1.1013, + "step": 9987 + }, + { + "epoch": 0.74, + "learning_rate": 1.9549729153511474e-05, + "loss": 0.966, + "step": 9988 + }, + { + "epoch": 0.74, + "learning_rate": 1.954961080350077e-05, + "loss": 0.9313, + "step": 9989 + }, + { + "epoch": 0.74, + "learning_rate": 1.9549492438296776e-05, + "loss": 0.9466, + "step": 9990 + }, + { + "epoch": 0.74, + "learning_rate": 1.954937405789967e-05, + "loss": 0.9619, + "step": 9991 + }, + { + "epoch": 0.74, + "learning_rate": 1.954925566230965e-05, + "loss": 1.0536, + "step": 9992 + }, + { + "epoch": 0.74, + "learning_rate": 1.9549137251526898e-05, + "loss": 0.9465, + "step": 9993 + }, + { + "epoch": 0.74, + "learning_rate": 1.954901882555161e-05, + "loss": 1.0637, + "step": 9994 + }, + { + "epoch": 0.74, + "learning_rate": 1.954890038438397e-05, + "loss": 1.0861, + "step": 9995 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548781928024157e-05, + "loss": 0.9507, + "step": 9996 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548663456472378e-05, + "loss": 1.0415, + "step": 9997 + }, + { + "epoch": 0.74, + "learning_rate": 1.954854496972881e-05, + "loss": 1.0882, + "step": 9998 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548426467793642e-05, + "loss": 0.965, + "step": 9999 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548307950667062e-05, + "loss": 1.1499, + "step": 10000 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548189418349265e-05, + "loss": 0.9635, + "step": 10001 + }, + { + "epoch": 0.74, + "learning_rate": 1.9548070870840436e-05, + "loss": 1.1406, + "step": 10002 + }, + { + "epoch": 0.74, + "learning_rate": 1.954795230814076e-05, + "loss": 1.0178, + "step": 10003 + }, + { + "epoch": 0.74, + "learning_rate": 1.9547833730250427e-05, + "loss": 0.9991, + "step": 10004 + }, + { + "epoch": 0.74, + "learning_rate": 1.954771513716963e-05, + "loss": 1.0502, + "step": 10005 + }, + { + "epoch": 0.74, + "learning_rate": 1.9547596528898555e-05, + "loss": 0.9214, + "step": 10006 + }, + { + "epoch": 0.74, + "learning_rate": 1.954747790543739e-05, + "loss": 1.033, + "step": 10007 + }, + { + "epoch": 0.74, + "learning_rate": 1.954735926678632e-05, + "loss": 1.0223, + "step": 10008 + }, + { + "epoch": 0.74, + "learning_rate": 1.954724061294554e-05, + "loss": 0.9108, + "step": 10009 + }, + { + "epoch": 0.74, + "learning_rate": 1.9547121943915243e-05, + "loss": 1.0334, + "step": 10010 + }, + { + "epoch": 0.74, + "learning_rate": 1.95470032596956e-05, + "loss": 1.0627, + "step": 10011 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546884560286823e-05, + "loss": 1.0604, + "step": 10012 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546765845689084e-05, + "loss": 1.0316, + "step": 10013 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546647115902576e-05, + "loss": 1.0549, + "step": 10014 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546528370927494e-05, + "loss": 0.9633, + "step": 10015 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546409610764017e-05, + "loss": 1.0618, + "step": 10016 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546290835412342e-05, + "loss": 1.0825, + "step": 10017 + }, + { + "epoch": 0.74, + "learning_rate": 1.9546172044872654e-05, + "loss": 1.0837, + "step": 10018 + }, + { + "epoch": 0.74, + "learning_rate": 1.954605323914514e-05, + "loss": 1.0337, + "step": 10019 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545934418229994e-05, + "loss": 0.9658, + "step": 10020 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545815582127404e-05, + "loss": 1.1083, + "step": 10021 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545696730837558e-05, + "loss": 1.0105, + "step": 10022 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545577864360644e-05, + "loss": 1.1869, + "step": 10023 + }, + { + "epoch": 0.74, + "learning_rate": 1.954545898269685e-05, + "loss": 1.0439, + "step": 10024 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545340085846373e-05, + "loss": 1.1244, + "step": 10025 + }, + { + "epoch": 0.74, + "learning_rate": 1.9545221173809393e-05, + "loss": 1.0207, + "step": 10026 + }, + { + "epoch": 0.74, + "learning_rate": 1.95451022465861e-05, + "loss": 0.9029, + "step": 10027 + }, + { + "epoch": 0.74, + "learning_rate": 1.954498330417669e-05, + "loss": 1.0058, + "step": 10028 + }, + { + "epoch": 0.74, + "learning_rate": 1.9544864346581347e-05, + "loss": 0.9661, + "step": 10029 + }, + { + "epoch": 0.74, + "learning_rate": 1.9544745373800264e-05, + "loss": 1.1171, + "step": 10030 + }, + { + "epoch": 0.74, + "learning_rate": 1.954462638583362e-05, + "loss": 1.0669, + "step": 10031 + }, + { + "epoch": 0.74, + "learning_rate": 1.954450738268162e-05, + "loss": 1.1161, + "step": 10032 + }, + { + "epoch": 0.74, + "learning_rate": 1.9544388364344442e-05, + "loss": 1.0603, + "step": 10033 + }, + { + "epoch": 0.74, + "learning_rate": 1.9544269330822277e-05, + "loss": 1.0497, + "step": 10034 + }, + { + "epoch": 0.74, + "learning_rate": 1.954415028211532e-05, + "loss": 1.008, + "step": 10035 + }, + { + "epoch": 0.74, + "learning_rate": 1.9544031218223757e-05, + "loss": 0.9574, + "step": 10036 + }, + { + "epoch": 0.74, + "learning_rate": 1.954391213914777e-05, + "loss": 1.0184, + "step": 10037 + }, + { + "epoch": 0.74, + "learning_rate": 1.954379304488756e-05, + "loss": 0.9767, + "step": 10038 + }, + { + "epoch": 0.74, + "learning_rate": 1.9543673935443312e-05, + "loss": 1.0175, + "step": 10039 + }, + { + "epoch": 0.74, + "learning_rate": 1.9543554810815216e-05, + "loss": 1.043, + "step": 10040 + }, + { + "epoch": 0.74, + "learning_rate": 1.954343567100346e-05, + "loss": 1.1295, + "step": 10041 + }, + { + "epoch": 0.74, + "learning_rate": 1.9543316516008236e-05, + "loss": 1.0858, + "step": 10042 + }, + { + "epoch": 0.74, + "learning_rate": 1.9543197345829727e-05, + "loss": 1.1157, + "step": 10043 + }, + { + "epoch": 0.74, + "learning_rate": 1.954307816046813e-05, + "loss": 1.0209, + "step": 10044 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542958959923632e-05, + "loss": 0.955, + "step": 10045 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542839744196427e-05, + "loss": 1.027, + "step": 10046 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542720513286696e-05, + "loss": 0.9941, + "step": 10047 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542601267194633e-05, + "loss": 1.0291, + "step": 10048 + }, + { + "epoch": 0.74, + "learning_rate": 1.954248200592043e-05, + "loss": 1.0147, + "step": 10049 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542362729464273e-05, + "loss": 0.9495, + "step": 10050 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542243437826355e-05, + "loss": 1.0392, + "step": 10051 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542124131006862e-05, + "loss": 1.0845, + "step": 10052 + }, + { + "epoch": 0.74, + "learning_rate": 1.9542004809005987e-05, + "loss": 0.9996, + "step": 10053 + }, + { + "epoch": 0.74, + "learning_rate": 1.9541885471823924e-05, + "loss": 1.0133, + "step": 10054 + }, + { + "epoch": 0.74, + "learning_rate": 1.954176611946085e-05, + "loss": 1.1384, + "step": 10055 + }, + { + "epoch": 0.74, + "learning_rate": 1.9541646751916967e-05, + "loss": 1.0653, + "step": 10056 + }, + { + "epoch": 0.74, + "learning_rate": 1.954152736919246e-05, + "loss": 0.9774, + "step": 10057 + }, + { + "epoch": 0.74, + "learning_rate": 1.954140797128752e-05, + "loss": 1.0086, + "step": 10058 + }, + { + "epoch": 0.74, + "learning_rate": 1.9541288558202335e-05, + "loss": 0.969, + "step": 10059 + }, + { + "epoch": 0.74, + "learning_rate": 1.9541169129937094e-05, + "loss": 1.0877, + "step": 10060 + }, + { + "epoch": 0.74, + "learning_rate": 1.9541049686491993e-05, + "loss": 0.8561, + "step": 10061 + }, + { + "epoch": 0.74, + "learning_rate": 1.9540930227867218e-05, + "loss": 1.0375, + "step": 10062 + }, + { + "epoch": 0.74, + "learning_rate": 1.9540810754062958e-05, + "loss": 0.9552, + "step": 10063 + }, + { + "epoch": 0.74, + "learning_rate": 1.9540691265079404e-05, + "loss": 1.1067, + "step": 10064 + }, + { + "epoch": 0.74, + "learning_rate": 1.954057176091675e-05, + "loss": 1.0186, + "step": 10065 + }, + { + "epoch": 0.74, + "learning_rate": 1.954045224157518e-05, + "loss": 0.9761, + "step": 10066 + }, + { + "epoch": 0.74, + "learning_rate": 1.954033270705489e-05, + "loss": 1.164, + "step": 10067 + }, + { + "epoch": 0.74, + "learning_rate": 1.9540213157356064e-05, + "loss": 1.0757, + "step": 10068 + }, + { + "epoch": 0.74, + "learning_rate": 1.9540093592478896e-05, + "loss": 1.0998, + "step": 10069 + }, + { + "epoch": 0.74, + "learning_rate": 1.953997401242358e-05, + "loss": 1.0351, + "step": 10070 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539854417190296e-05, + "loss": 1.1073, + "step": 10071 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539734806779243e-05, + "loss": 1.0413, + "step": 10072 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539615181190606e-05, + "loss": 1.0779, + "step": 10073 + }, + { + "epoch": 0.74, + "learning_rate": 1.953949554042458e-05, + "loss": 1.0258, + "step": 10074 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539375884481354e-05, + "loss": 1.052, + "step": 10075 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539256213361115e-05, + "loss": 1.0962, + "step": 10076 + }, + { + "epoch": 0.74, + "learning_rate": 1.9539136527064055e-05, + "loss": 1.0315, + "step": 10077 + }, + { + "epoch": 0.74, + "learning_rate": 1.953901682559037e-05, + "loss": 1.0417, + "step": 10078 + }, + { + "epoch": 0.74, + "learning_rate": 1.953889710894024e-05, + "loss": 1.07, + "step": 10079 + }, + { + "epoch": 0.74, + "learning_rate": 1.9538777377113868e-05, + "loss": 0.9325, + "step": 10080 + }, + { + "epoch": 0.74, + "learning_rate": 1.953865763011143e-05, + "loss": 1.0092, + "step": 10081 + }, + { + "epoch": 0.74, + "learning_rate": 1.953853786793313e-05, + "loss": 0.9605, + "step": 10082 + }, + { + "epoch": 0.75, + "learning_rate": 1.9538418090579147e-05, + "loss": 1.0464, + "step": 10083 + }, + { + "epoch": 0.75, + "learning_rate": 1.953829829804968e-05, + "loss": 0.9737, + "step": 10084 + }, + { + "epoch": 0.75, + "learning_rate": 1.9538178490344922e-05, + "loss": 1.0708, + "step": 10085 + }, + { + "epoch": 0.75, + "learning_rate": 1.9538058667465053e-05, + "loss": 1.157, + "step": 10086 + }, + { + "epoch": 0.75, + "learning_rate": 1.953793882941027e-05, + "loss": 1.0138, + "step": 10087 + }, + { + "epoch": 0.75, + "learning_rate": 1.9537818976180763e-05, + "loss": 0.9612, + "step": 10088 + }, + { + "epoch": 0.75, + "learning_rate": 1.953769910777672e-05, + "loss": 1.0858, + "step": 10089 + }, + { + "epoch": 0.75, + "learning_rate": 1.953757922419834e-05, + "loss": 1.0994, + "step": 10090 + }, + { + "epoch": 0.75, + "learning_rate": 1.9537459325445805e-05, + "loss": 0.98, + "step": 10091 + }, + { + "epoch": 0.75, + "learning_rate": 1.9537339411519306e-05, + "loss": 1.1397, + "step": 10092 + }, + { + "epoch": 0.75, + "learning_rate": 1.953721948241904e-05, + "loss": 1.0258, + "step": 10093 + }, + { + "epoch": 0.75, + "learning_rate": 1.9537099538145195e-05, + "loss": 0.928, + "step": 10094 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536979578697957e-05, + "loss": 1.0788, + "step": 10095 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536859604077522e-05, + "loss": 1.0043, + "step": 10096 + }, + { + "epoch": 0.75, + "learning_rate": 1.953673961428408e-05, + "loss": 1.02, + "step": 10097 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536619609317827e-05, + "loss": 1.0602, + "step": 10098 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536499589178943e-05, + "loss": 1.0364, + "step": 10099 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536379553867625e-05, + "loss": 1.0384, + "step": 10100 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536259503384067e-05, + "loss": 1.1244, + "step": 10101 + }, + { + "epoch": 0.75, + "learning_rate": 1.953613943772845e-05, + "loss": 1.106, + "step": 10102 + }, + { + "epoch": 0.75, + "learning_rate": 1.9536019356900977e-05, + "loss": 1.024, + "step": 10103 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535899260901832e-05, + "loss": 0.9922, + "step": 10104 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535779149731206e-05, + "loss": 1.0438, + "step": 10105 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535659023389298e-05, + "loss": 1.0146, + "step": 10106 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535538881876288e-05, + "loss": 1.076, + "step": 10107 + }, + { + "epoch": 0.75, + "learning_rate": 1.953541872519237e-05, + "loss": 1.0556, + "step": 10108 + }, + { + "epoch": 0.75, + "learning_rate": 1.953529855333774e-05, + "loss": 1.0676, + "step": 10109 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535178366312587e-05, + "loss": 1.085, + "step": 10110 + }, + { + "epoch": 0.75, + "learning_rate": 1.9535058164117103e-05, + "loss": 1.0344, + "step": 10111 + }, + { + "epoch": 0.75, + "learning_rate": 1.9534937946751476e-05, + "loss": 1.1049, + "step": 10112 + }, + { + "epoch": 0.75, + "learning_rate": 1.95348177142159e-05, + "loss": 0.9477, + "step": 10113 + }, + { + "epoch": 0.75, + "learning_rate": 1.9534697466510565e-05, + "loss": 0.959, + "step": 10114 + }, + { + "epoch": 0.75, + "learning_rate": 1.953457720363566e-05, + "loss": 1.014, + "step": 10115 + }, + { + "epoch": 0.75, + "learning_rate": 1.953445692559138e-05, + "loss": 0.9901, + "step": 10116 + }, + { + "epoch": 0.75, + "learning_rate": 1.9534336632377918e-05, + "loss": 1.073, + "step": 10117 + }, + { + "epoch": 0.75, + "learning_rate": 1.953421632399546e-05, + "loss": 0.8943, + "step": 10118 + }, + { + "epoch": 0.75, + "learning_rate": 1.9534096000444203e-05, + "loss": 1.096, + "step": 10119 + }, + { + "epoch": 0.75, + "learning_rate": 1.9533975661724335e-05, + "loss": 1.0269, + "step": 10120 + }, + { + "epoch": 0.75, + "learning_rate": 1.953385530783605e-05, + "loss": 1.1127, + "step": 10121 + }, + { + "epoch": 0.75, + "learning_rate": 1.9533734938779533e-05, + "loss": 1.0581, + "step": 10122 + }, + { + "epoch": 0.75, + "learning_rate": 1.953361455455498e-05, + "loss": 1.0764, + "step": 10123 + }, + { + "epoch": 0.75, + "learning_rate": 1.9533494155162587e-05, + "loss": 1.0862, + "step": 10124 + }, + { + "epoch": 0.75, + "learning_rate": 1.953337374060254e-05, + "loss": 0.9391, + "step": 10125 + }, + { + "epoch": 0.75, + "learning_rate": 1.953325331087503e-05, + "loss": 1.117, + "step": 10126 + }, + { + "epoch": 0.75, + "learning_rate": 1.953313286598025e-05, + "loss": 1.0444, + "step": 10127 + }, + { + "epoch": 0.75, + "learning_rate": 1.9533012405918395e-05, + "loss": 1.0365, + "step": 10128 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532891930689654e-05, + "loss": 1.0236, + "step": 10129 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532771440294215e-05, + "loss": 0.976, + "step": 10130 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532650934732276e-05, + "loss": 1.1268, + "step": 10131 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532530414004024e-05, + "loss": 1.08, + "step": 10132 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532409878109652e-05, + "loss": 1.0584, + "step": 10133 + }, + { + "epoch": 0.75, + "learning_rate": 1.953228932704935e-05, + "loss": 0.9739, + "step": 10134 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532168760823314e-05, + "loss": 1.1069, + "step": 10135 + }, + { + "epoch": 0.75, + "learning_rate": 1.9532048179431736e-05, + "loss": 1.1322, + "step": 10136 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531927582874805e-05, + "loss": 1.0197, + "step": 10137 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531806971152715e-05, + "loss": 1.0488, + "step": 10138 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531686344265655e-05, + "loss": 1.0178, + "step": 10139 + }, + { + "epoch": 0.75, + "learning_rate": 1.953156570221382e-05, + "loss": 0.9862, + "step": 10140 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531445044997396e-05, + "loss": 0.9786, + "step": 10141 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531324372616583e-05, + "loss": 1.0168, + "step": 10142 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531203685071566e-05, + "loss": 1.066, + "step": 10143 + }, + { + "epoch": 0.75, + "learning_rate": 1.9531082982362545e-05, + "loss": 0.9547, + "step": 10144 + }, + { + "epoch": 0.75, + "learning_rate": 1.9530962264489703e-05, + "loss": 1.026, + "step": 10145 + }, + { + "epoch": 0.75, + "learning_rate": 1.9530841531453237e-05, + "loss": 1.1037, + "step": 10146 + }, + { + "epoch": 0.75, + "learning_rate": 1.9530720783253342e-05, + "loss": 0.9829, + "step": 10147 + }, + { + "epoch": 0.75, + "learning_rate": 1.9530600019890206e-05, + "loss": 1.0364, + "step": 10148 + }, + { + "epoch": 0.75, + "learning_rate": 1.953047924136402e-05, + "loss": 1.0763, + "step": 10149 + }, + { + "epoch": 0.75, + "learning_rate": 1.9530358447674975e-05, + "loss": 1.1, + "step": 10150 + }, + { + "epoch": 0.75, + "learning_rate": 1.953023763882327e-05, + "loss": 1.032, + "step": 10151 + }, + { + "epoch": 0.75, + "learning_rate": 1.953011681480909e-05, + "loss": 0.9915, + "step": 10152 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529995975632634e-05, + "loss": 0.9914, + "step": 10153 + }, + { + "epoch": 0.75, + "learning_rate": 1.952987512129409e-05, + "loss": 1.0647, + "step": 10154 + }, + { + "epoch": 0.75, + "learning_rate": 1.952975425179365e-05, + "loss": 0.9918, + "step": 10155 + }, + { + "epoch": 0.75, + "learning_rate": 1.952963336713151e-05, + "loss": 1.0639, + "step": 10156 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529512467307853e-05, + "loss": 1.0647, + "step": 10157 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529391552322885e-05, + "loss": 1.0455, + "step": 10158 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529270622176787e-05, + "loss": 0.9661, + "step": 10159 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529149676869757e-05, + "loss": 1.0083, + "step": 10160 + }, + { + "epoch": 0.75, + "learning_rate": 1.9529028716401987e-05, + "loss": 0.997, + "step": 10161 + }, + { + "epoch": 0.75, + "learning_rate": 1.952890774077367e-05, + "loss": 0.9593, + "step": 10162 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528786749984995e-05, + "loss": 0.9553, + "step": 10163 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528665744036156e-05, + "loss": 1.0237, + "step": 10164 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528544722927348e-05, + "loss": 0.9565, + "step": 10165 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528423686658762e-05, + "loss": 0.9057, + "step": 10166 + }, + { + "epoch": 0.75, + "learning_rate": 1.952830263523059e-05, + "loss": 0.9976, + "step": 10167 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528181568643025e-05, + "loss": 0.9641, + "step": 10168 + }, + { + "epoch": 0.75, + "learning_rate": 1.9528060486896257e-05, + "loss": 1.0099, + "step": 10169 + }, + { + "epoch": 0.75, + "learning_rate": 1.9527939389990483e-05, + "loss": 1.0732, + "step": 10170 + }, + { + "epoch": 0.75, + "learning_rate": 1.952781827792589e-05, + "loss": 0.9942, + "step": 10171 + }, + { + "epoch": 0.75, + "learning_rate": 1.952769715070268e-05, + "loss": 1.051, + "step": 10172 + }, + { + "epoch": 0.75, + "learning_rate": 1.952757600832104e-05, + "loss": 1.0763, + "step": 10173 + }, + { + "epoch": 0.75, + "learning_rate": 1.952745485078116e-05, + "loss": 1.1427, + "step": 10174 + }, + { + "epoch": 0.75, + "learning_rate": 1.952733367808324e-05, + "loss": 1.0277, + "step": 10175 + }, + { + "epoch": 0.75, + "learning_rate": 1.952721249022746e-05, + "loss": 1.0665, + "step": 10176 + }, + { + "epoch": 0.75, + "learning_rate": 1.952709128721403e-05, + "loss": 0.9665, + "step": 10177 + }, + { + "epoch": 0.75, + "learning_rate": 1.952697006904313e-05, + "loss": 0.983, + "step": 10178 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526848835714956e-05, + "loss": 0.9093, + "step": 10179 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526727587229702e-05, + "loss": 1.1268, + "step": 10180 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526606323587562e-05, + "loss": 1.0299, + "step": 10181 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526485044788726e-05, + "loss": 1.1284, + "step": 10182 + }, + { + "epoch": 0.75, + "learning_rate": 1.952636375083339e-05, + "loss": 1.0594, + "step": 10183 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526242441721747e-05, + "loss": 1.1299, + "step": 10184 + }, + { + "epoch": 0.75, + "learning_rate": 1.9526121117453986e-05, + "loss": 1.1515, + "step": 10185 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525999778030306e-05, + "loss": 0.9571, + "step": 10186 + }, + { + "epoch": 0.75, + "learning_rate": 1.952587842345089e-05, + "loss": 1.1118, + "step": 10187 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525757053715947e-05, + "loss": 1.0553, + "step": 10188 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525635668825656e-05, + "loss": 0.9962, + "step": 10189 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525514268780213e-05, + "loss": 0.9864, + "step": 10190 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525392853579813e-05, + "loss": 1.0126, + "step": 10191 + }, + { + "epoch": 0.75, + "learning_rate": 1.9525271423224655e-05, + "loss": 1.0462, + "step": 10192 + }, + { + "epoch": 0.75, + "learning_rate": 1.952514997771492e-05, + "loss": 1.0909, + "step": 10193 + }, + { + "epoch": 0.75, + "learning_rate": 1.952502851705081e-05, + "loss": 0.9943, + "step": 10194 + }, + { + "epoch": 0.75, + "learning_rate": 1.9524907041232516e-05, + "loss": 1.0446, + "step": 10195 + }, + { + "epoch": 0.75, + "learning_rate": 1.952478555026023e-05, + "loss": 1.0101, + "step": 10196 + }, + { + "epoch": 0.75, + "learning_rate": 1.952466404413415e-05, + "loss": 0.9159, + "step": 10197 + }, + { + "epoch": 0.75, + "learning_rate": 1.9524542522854462e-05, + "loss": 1.0848, + "step": 10198 + }, + { + "epoch": 0.75, + "learning_rate": 1.9524420986421367e-05, + "loss": 0.9578, + "step": 10199 + }, + { + "epoch": 0.75, + "learning_rate": 1.9524299434835052e-05, + "loss": 1.0611, + "step": 10200 + }, + { + "epoch": 0.75, + "learning_rate": 1.952417786809571e-05, + "loss": 0.9719, + "step": 10201 + }, + { + "epoch": 0.75, + "learning_rate": 1.9524056286203544e-05, + "loss": 1.0284, + "step": 10202 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523934689158736e-05, + "loss": 0.9832, + "step": 10203 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523813076961488e-05, + "loss": 0.9926, + "step": 10204 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523691449611983e-05, + "loss": 1.0409, + "step": 10205 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523569807110427e-05, + "loss": 1.047, + "step": 10206 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523448149457003e-05, + "loss": 1.0299, + "step": 10207 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523326476651915e-05, + "loss": 1.1164, + "step": 10208 + }, + { + "epoch": 0.75, + "learning_rate": 1.9523204788695346e-05, + "loss": 1.1074, + "step": 10209 + }, + { + "epoch": 0.75, + "learning_rate": 1.95230830855875e-05, + "loss": 1.0418, + "step": 10210 + }, + { + "epoch": 0.75, + "learning_rate": 1.952296136732856e-05, + "loss": 1.1403, + "step": 10211 + }, + { + "epoch": 0.75, + "learning_rate": 1.9522839633918728e-05, + "loss": 1.0014, + "step": 10212 + }, + { + "epoch": 0.75, + "learning_rate": 1.952271788535819e-05, + "loss": 0.8931, + "step": 10213 + }, + { + "epoch": 0.75, + "learning_rate": 1.9522596121647146e-05, + "loss": 1.0679, + "step": 10214 + }, + { + "epoch": 0.75, + "learning_rate": 1.9522474342785788e-05, + "loss": 1.0106, + "step": 10215 + }, + { + "epoch": 0.75, + "learning_rate": 1.952235254877431e-05, + "loss": 0.984, + "step": 10216 + }, + { + "epoch": 0.75, + "learning_rate": 1.9522230739612906e-05, + "loss": 0.9911, + "step": 10217 + }, + { + "epoch": 0.76, + "learning_rate": 1.952210891530177e-05, + "loss": 1.0818, + "step": 10218 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521987075841092e-05, + "loss": 1.0662, + "step": 10219 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521865221231068e-05, + "loss": 1.0455, + "step": 10220 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521743351471897e-05, + "loss": 1.0953, + "step": 10221 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521621466563766e-05, + "loss": 1.039, + "step": 10222 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521499566506874e-05, + "loss": 1.0079, + "step": 10223 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521377651301407e-05, + "loss": 1.0578, + "step": 10224 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521255720947566e-05, + "loss": 0.9824, + "step": 10225 + }, + { + "epoch": 0.76, + "learning_rate": 1.9521133775445547e-05, + "loss": 1.0226, + "step": 10226 + }, + { + "epoch": 0.76, + "learning_rate": 1.952101181479554e-05, + "loss": 1.1039, + "step": 10227 + }, + { + "epoch": 0.76, + "learning_rate": 1.9520889838997737e-05, + "loss": 1.0217, + "step": 10228 + }, + { + "epoch": 0.76, + "learning_rate": 1.9520767848052335e-05, + "loss": 0.9002, + "step": 10229 + }, + { + "epoch": 0.76, + "learning_rate": 1.952064584195953e-05, + "loss": 1.1156, + "step": 10230 + }, + { + "epoch": 0.76, + "learning_rate": 1.952052382071951e-05, + "loss": 0.9939, + "step": 10231 + }, + { + "epoch": 0.76, + "learning_rate": 1.9520401784332475e-05, + "loss": 1.024, + "step": 10232 + }, + { + "epoch": 0.76, + "learning_rate": 1.9520279732798614e-05, + "loss": 1.122, + "step": 10233 + }, + { + "epoch": 0.76, + "learning_rate": 1.952015766611813e-05, + "loss": 1.0125, + "step": 10234 + }, + { + "epoch": 0.76, + "learning_rate": 1.952003558429121e-05, + "loss": 1.0267, + "step": 10235 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519913487318046e-05, + "loss": 0.9826, + "step": 10236 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519791375198836e-05, + "loss": 1.0417, + "step": 10237 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519669247933775e-05, + "loss": 1.0236, + "step": 10238 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519547105523057e-05, + "loss": 0.9655, + "step": 10239 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519424947966878e-05, + "loss": 1.1375, + "step": 10240 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519302775265425e-05, + "loss": 1.0394, + "step": 10241 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519180587418904e-05, + "loss": 1.0203, + "step": 10242 + }, + { + "epoch": 0.76, + "learning_rate": 1.9519058384427495e-05, + "loss": 1.0364, + "step": 10243 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518936166291404e-05, + "loss": 0.9611, + "step": 10244 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518813933010822e-05, + "loss": 1.0484, + "step": 10245 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518691684585945e-05, + "loss": 1.0089, + "step": 10246 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518569421016962e-05, + "loss": 1.012, + "step": 10247 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518447142304074e-05, + "loss": 0.9966, + "step": 10248 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518324848447473e-05, + "loss": 1.1315, + "step": 10249 + }, + { + "epoch": 0.76, + "learning_rate": 1.9518202539447353e-05, + "loss": 1.084, + "step": 10250 + }, + { + "epoch": 0.76, + "learning_rate": 1.951808021530391e-05, + "loss": 0.9571, + "step": 10251 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517957876017335e-05, + "loss": 1.1707, + "step": 10252 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517835521587826e-05, + "loss": 1.0223, + "step": 10253 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517713152015577e-05, + "loss": 1.0716, + "step": 10254 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517590767300784e-05, + "loss": 0.9936, + "step": 10255 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517468367443638e-05, + "loss": 1.0827, + "step": 10256 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517345952444338e-05, + "loss": 0.926, + "step": 10257 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517223522303077e-05, + "loss": 1.0976, + "step": 10258 + }, + { + "epoch": 0.76, + "learning_rate": 1.9517101077020048e-05, + "loss": 1.0233, + "step": 10259 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516978616595445e-05, + "loss": 0.958, + "step": 10260 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516856141029467e-05, + "loss": 1.0427, + "step": 10261 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516733650322308e-05, + "loss": 1.028, + "step": 10262 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516611144474163e-05, + "loss": 1.0269, + "step": 10263 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516488623485227e-05, + "loss": 1.0477, + "step": 10264 + }, + { + "epoch": 0.76, + "learning_rate": 1.951636608735569e-05, + "loss": 0.9453, + "step": 10265 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516243536085752e-05, + "loss": 1.0591, + "step": 10266 + }, + { + "epoch": 0.76, + "learning_rate": 1.9516120969675606e-05, + "loss": 1.0528, + "step": 10267 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515998388125447e-05, + "loss": 1.0885, + "step": 10268 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515875791435475e-05, + "loss": 1.0398, + "step": 10269 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515753179605878e-05, + "loss": 1.0319, + "step": 10270 + }, + { + "epoch": 0.76, + "learning_rate": 1.951563055263685e-05, + "loss": 1.0618, + "step": 10271 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515507910528596e-05, + "loss": 0.9804, + "step": 10272 + }, + { + "epoch": 0.76, + "learning_rate": 1.95153852532813e-05, + "loss": 0.9994, + "step": 10273 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515262580895167e-05, + "loss": 1.0343, + "step": 10274 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515139893370385e-05, + "loss": 1.0589, + "step": 10275 + }, + { + "epoch": 0.76, + "learning_rate": 1.9515017190707152e-05, + "loss": 1.0234, + "step": 10276 + }, + { + "epoch": 0.76, + "learning_rate": 1.9514894472905662e-05, + "loss": 1.0471, + "step": 10277 + }, + { + "epoch": 0.76, + "learning_rate": 1.951477173996611e-05, + "loss": 0.9823, + "step": 10278 + }, + { + "epoch": 0.76, + "learning_rate": 1.95146489918887e-05, + "loss": 1.0225, + "step": 10279 + }, + { + "epoch": 0.76, + "learning_rate": 1.9514526228673612e-05, + "loss": 1.0997, + "step": 10280 + }, + { + "epoch": 0.76, + "learning_rate": 1.951440345032105e-05, + "loss": 0.9616, + "step": 10281 + }, + { + "epoch": 0.76, + "learning_rate": 1.9514280656831207e-05, + "loss": 1.0088, + "step": 10282 + }, + { + "epoch": 0.76, + "learning_rate": 1.951415784820428e-05, + "loss": 1.0381, + "step": 10283 + }, + { + "epoch": 0.76, + "learning_rate": 1.9514035024440467e-05, + "loss": 1.1154, + "step": 10284 + }, + { + "epoch": 0.76, + "learning_rate": 1.951391218553996e-05, + "loss": 1.1125, + "step": 10285 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513789331502952e-05, + "loss": 0.9502, + "step": 10286 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513666462329644e-05, + "loss": 1.0691, + "step": 10287 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513543578020226e-05, + "loss": 1.0334, + "step": 10288 + }, + { + "epoch": 0.76, + "learning_rate": 1.95134206785749e-05, + "loss": 1.1149, + "step": 10289 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513297763993856e-05, + "loss": 0.9235, + "step": 10290 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513174834277292e-05, + "loss": 1.1105, + "step": 10291 + }, + { + "epoch": 0.76, + "learning_rate": 1.9513051889425402e-05, + "loss": 0.9881, + "step": 10292 + }, + { + "epoch": 0.76, + "learning_rate": 1.951292892943838e-05, + "loss": 0.948, + "step": 10293 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512805954316427e-05, + "loss": 1.0894, + "step": 10294 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512682964059735e-05, + "loss": 1.055, + "step": 10295 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512559958668503e-05, + "loss": 0.9478, + "step": 10296 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512436938142923e-05, + "loss": 1.0879, + "step": 10297 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512313902483192e-05, + "loss": 1.0376, + "step": 10298 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512190851689503e-05, + "loss": 0.9981, + "step": 10299 + }, + { + "epoch": 0.76, + "learning_rate": 1.9512067785762055e-05, + "loss": 1.0749, + "step": 10300 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511944704701043e-05, + "loss": 1.0079, + "step": 10301 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511821608506666e-05, + "loss": 1.0266, + "step": 10302 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511698497179117e-05, + "loss": 1.0821, + "step": 10303 + }, + { + "epoch": 0.76, + "learning_rate": 1.951157537071859e-05, + "loss": 0.9957, + "step": 10304 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511452229125282e-05, + "loss": 0.9426, + "step": 10305 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511329072399392e-05, + "loss": 0.924, + "step": 10306 + }, + { + "epoch": 0.76, + "learning_rate": 1.9511205900541114e-05, + "loss": 1.0109, + "step": 10307 + }, + { + "epoch": 0.76, + "learning_rate": 1.951108271355064e-05, + "loss": 1.0739, + "step": 10308 + }, + { + "epoch": 0.76, + "learning_rate": 1.951095951142817e-05, + "loss": 1.0297, + "step": 10309 + }, + { + "epoch": 0.76, + "learning_rate": 1.95108362941739e-05, + "loss": 1.0467, + "step": 10310 + }, + { + "epoch": 0.76, + "learning_rate": 1.9510713061788026e-05, + "loss": 1.1138, + "step": 10311 + }, + { + "epoch": 0.76, + "learning_rate": 1.9510589814270745e-05, + "loss": 1.0042, + "step": 10312 + }, + { + "epoch": 0.76, + "learning_rate": 1.951046655162225e-05, + "loss": 1.0676, + "step": 10313 + }, + { + "epoch": 0.76, + "learning_rate": 1.951034327384274e-05, + "loss": 0.9611, + "step": 10314 + }, + { + "epoch": 0.76, + "learning_rate": 1.951021998093241e-05, + "loss": 1.0757, + "step": 10315 + }, + { + "epoch": 0.76, + "learning_rate": 1.9510096672891454e-05, + "loss": 1.0113, + "step": 10316 + }, + { + "epoch": 0.76, + "learning_rate": 1.9509973349720074e-05, + "loss": 1.0199, + "step": 10317 + }, + { + "epoch": 0.76, + "learning_rate": 1.950985001141846e-05, + "loss": 1.0389, + "step": 10318 + }, + { + "epoch": 0.76, + "learning_rate": 1.950972665798681e-05, + "loss": 1.0195, + "step": 10319 + }, + { + "epoch": 0.76, + "learning_rate": 1.9509603289425322e-05, + "loss": 1.0232, + "step": 10320 + }, + { + "epoch": 0.76, + "learning_rate": 1.950947990573419e-05, + "loss": 1.0603, + "step": 10321 + }, + { + "epoch": 0.76, + "learning_rate": 1.9509356506913613e-05, + "loss": 1.0791, + "step": 10322 + }, + { + "epoch": 0.76, + "learning_rate": 1.9509233092963783e-05, + "loss": 1.0366, + "step": 10323 + }, + { + "epoch": 0.76, + "learning_rate": 1.9509109663884902e-05, + "loss": 1.0338, + "step": 10324 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508986219677166e-05, + "loss": 1.0106, + "step": 10325 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508862760340763e-05, + "loss": 0.9092, + "step": 10326 + }, + { + "epoch": 0.76, + "learning_rate": 1.95087392858759e-05, + "loss": 0.9816, + "step": 10327 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508615796282768e-05, + "loss": 1.0017, + "step": 10328 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508492291561564e-05, + "loss": 1.0554, + "step": 10329 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508368771712484e-05, + "loss": 1.1379, + "step": 10330 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508245236735724e-05, + "loss": 0.9972, + "step": 10331 + }, + { + "epoch": 0.76, + "learning_rate": 1.9508121686631485e-05, + "loss": 1.0166, + "step": 10332 + }, + { + "epoch": 0.76, + "learning_rate": 1.950799812139996e-05, + "loss": 0.9476, + "step": 10333 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507874541041342e-05, + "loss": 1.1142, + "step": 10334 + }, + { + "epoch": 0.76, + "learning_rate": 1.950775094555584e-05, + "loss": 1.002, + "step": 10335 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507627334943634e-05, + "loss": 0.954, + "step": 10336 + }, + { + "epoch": 0.76, + "learning_rate": 1.950750370920493e-05, + "loss": 1.1123, + "step": 10337 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507380068339924e-05, + "loss": 1.0381, + "step": 10338 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507256412348817e-05, + "loss": 0.9828, + "step": 10339 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507132741231796e-05, + "loss": 1.1612, + "step": 10340 + }, + { + "epoch": 0.76, + "learning_rate": 1.9507009054989067e-05, + "loss": 1.0236, + "step": 10341 + }, + { + "epoch": 0.76, + "learning_rate": 1.950688535362082e-05, + "loss": 1.0019, + "step": 10342 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506761637127254e-05, + "loss": 1.0311, + "step": 10343 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506637905508567e-05, + "loss": 1.1569, + "step": 10344 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506514158764954e-05, + "loss": 1.0392, + "step": 10345 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506390396896616e-05, + "loss": 1.0481, + "step": 10346 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506266619903743e-05, + "loss": 0.9779, + "step": 10347 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506142827786537e-05, + "loss": 1.101, + "step": 10348 + }, + { + "epoch": 0.76, + "learning_rate": 1.9506019020545195e-05, + "loss": 1.1173, + "step": 10349 + }, + { + "epoch": 0.76, + "learning_rate": 1.9505895198179914e-05, + "loss": 0.9849, + "step": 10350 + }, + { + "epoch": 0.76, + "learning_rate": 1.9505771360690887e-05, + "loss": 0.9731, + "step": 10351 + }, + { + "epoch": 0.76, + "learning_rate": 1.9505647508078318e-05, + "loss": 1.0923, + "step": 10352 + }, + { + "epoch": 0.76, + "learning_rate": 1.9505523640342394e-05, + "loss": 1.1231, + "step": 10353 + }, + { + "epoch": 0.77, + "learning_rate": 1.950539975748332e-05, + "loss": 0.9633, + "step": 10354 + }, + { + "epoch": 0.77, + "learning_rate": 1.9505275859501292e-05, + "loss": 1.0136, + "step": 10355 + }, + { + "epoch": 0.77, + "learning_rate": 1.9505151946396505e-05, + "loss": 1.0781, + "step": 10356 + }, + { + "epoch": 0.77, + "learning_rate": 1.950502801816916e-05, + "loss": 1.0361, + "step": 10357 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504904074819448e-05, + "loss": 0.9661, + "step": 10358 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504780116347573e-05, + "loss": 1.037, + "step": 10359 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504656142753725e-05, + "loss": 1.1052, + "step": 10360 + }, + { + "epoch": 0.77, + "learning_rate": 1.950453215403811e-05, + "loss": 0.9991, + "step": 10361 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504408150200915e-05, + "loss": 1.0817, + "step": 10362 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504284131242346e-05, + "loss": 1.0256, + "step": 10363 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504160097162593e-05, + "loss": 1.0071, + "step": 10364 + }, + { + "epoch": 0.77, + "learning_rate": 1.9504036047961862e-05, + "loss": 1.0647, + "step": 10365 + }, + { + "epoch": 0.77, + "learning_rate": 1.9503911983640342e-05, + "loss": 1.0144, + "step": 10366 + }, + { + "epoch": 0.77, + "learning_rate": 1.9503787904198238e-05, + "loss": 1.0122, + "step": 10367 + }, + { + "epoch": 0.77, + "learning_rate": 1.950366380963574e-05, + "loss": 0.8997, + "step": 10368 + }, + { + "epoch": 0.77, + "learning_rate": 1.950353969995305e-05, + "loss": 1.0983, + "step": 10369 + }, + { + "epoch": 0.77, + "learning_rate": 1.9503415575150365e-05, + "loss": 1.0461, + "step": 10370 + }, + { + "epoch": 0.77, + "learning_rate": 1.950329143522788e-05, + "loss": 0.9532, + "step": 10371 + }, + { + "epoch": 0.77, + "learning_rate": 1.9503167280185796e-05, + "loss": 1.0706, + "step": 10372 + }, + { + "epoch": 0.77, + "learning_rate": 1.950304311002431e-05, + "loss": 1.0171, + "step": 10373 + }, + { + "epoch": 0.77, + "learning_rate": 1.950291892474361e-05, + "loss": 0.966, + "step": 10374 + }, + { + "epoch": 0.77, + "learning_rate": 1.950279472434391e-05, + "loss": 0.947, + "step": 10375 + }, + { + "epoch": 0.77, + "learning_rate": 1.95026705088254e-05, + "loss": 1.0916, + "step": 10376 + }, + { + "epoch": 0.77, + "learning_rate": 1.9502546278188274e-05, + "loss": 1.0394, + "step": 10377 + }, + { + "epoch": 0.77, + "learning_rate": 1.9502422032432737e-05, + "loss": 0.9372, + "step": 10378 + }, + { + "epoch": 0.77, + "learning_rate": 1.9502297771558977e-05, + "loss": 0.9158, + "step": 10379 + }, + { + "epoch": 0.77, + "learning_rate": 1.95021734955672e-05, + "loss": 1.0104, + "step": 10380 + }, + { + "epoch": 0.77, + "learning_rate": 1.95020492044576e-05, + "loss": 1.1272, + "step": 10381 + }, + { + "epoch": 0.77, + "learning_rate": 1.950192489823038e-05, + "loss": 1.0405, + "step": 10382 + }, + { + "epoch": 0.77, + "learning_rate": 1.9501800576885728e-05, + "loss": 0.9933, + "step": 10383 + }, + { + "epoch": 0.77, + "learning_rate": 1.950167624042385e-05, + "loss": 1.0042, + "step": 10384 + }, + { + "epoch": 0.77, + "learning_rate": 1.950155188884494e-05, + "loss": 1.0635, + "step": 10385 + }, + { + "epoch": 0.77, + "learning_rate": 1.9501427522149195e-05, + "loss": 1.124, + "step": 10386 + }, + { + "epoch": 0.77, + "learning_rate": 1.9501303140336816e-05, + "loss": 1.103, + "step": 10387 + }, + { + "epoch": 0.77, + "learning_rate": 1.9501178743408002e-05, + "loss": 1.1125, + "step": 10388 + }, + { + "epoch": 0.77, + "learning_rate": 1.9501054331362948e-05, + "loss": 1.113, + "step": 10389 + }, + { + "epoch": 0.77, + "learning_rate": 1.950092990420185e-05, + "loss": 1.0463, + "step": 10390 + }, + { + "epoch": 0.77, + "learning_rate": 1.950080546192491e-05, + "loss": 1.0305, + "step": 10391 + }, + { + "epoch": 0.77, + "learning_rate": 1.9500681004532323e-05, + "loss": 1.0774, + "step": 10392 + }, + { + "epoch": 0.77, + "learning_rate": 1.950055653202429e-05, + "loss": 1.0912, + "step": 10393 + }, + { + "epoch": 0.77, + "learning_rate": 1.950043204440101e-05, + "loss": 1.0069, + "step": 10394 + }, + { + "epoch": 0.77, + "learning_rate": 1.950030754166267e-05, + "loss": 1.0549, + "step": 10395 + }, + { + "epoch": 0.77, + "learning_rate": 1.9500183023809487e-05, + "loss": 0.9999, + "step": 10396 + }, + { + "epoch": 0.77, + "learning_rate": 1.9500058490841646e-05, + "loss": 1.0487, + "step": 10397 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499933942759342e-05, + "loss": 0.927, + "step": 10398 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499809379562786e-05, + "loss": 1.0686, + "step": 10399 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499684801252167e-05, + "loss": 1.1547, + "step": 10400 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499560207827685e-05, + "loss": 0.9852, + "step": 10401 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499435599289536e-05, + "loss": 0.9569, + "step": 10402 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499310975637925e-05, + "loss": 1.0301, + "step": 10403 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499186336873043e-05, + "loss": 1.0424, + "step": 10404 + }, + { + "epoch": 0.77, + "learning_rate": 1.9499061682995094e-05, + "loss": 1.0759, + "step": 10405 + }, + { + "epoch": 0.77, + "learning_rate": 1.9498937014004272e-05, + "loss": 1.0782, + "step": 10406 + }, + { + "epoch": 0.77, + "learning_rate": 1.9498812329900776e-05, + "loss": 1.106, + "step": 10407 + }, + { + "epoch": 0.77, + "learning_rate": 1.949868763068481e-05, + "loss": 1.1044, + "step": 10408 + }, + { + "epoch": 0.77, + "learning_rate": 1.9498562916356562e-05, + "loss": 1.0063, + "step": 10409 + }, + { + "epoch": 0.77, + "learning_rate": 1.949843818691624e-05, + "loss": 1.0285, + "step": 10410 + }, + { + "epoch": 0.77, + "learning_rate": 1.949831344236404e-05, + "loss": 1.0195, + "step": 10411 + }, + { + "epoch": 0.77, + "learning_rate": 1.9498188682700156e-05, + "loss": 1.0237, + "step": 10412 + }, + { + "epoch": 0.77, + "learning_rate": 1.949806390792479e-05, + "loss": 1.0668, + "step": 10413 + }, + { + "epoch": 0.77, + "learning_rate": 1.949793911803814e-05, + "loss": 0.9747, + "step": 10414 + }, + { + "epoch": 0.77, + "learning_rate": 1.949781431304041e-05, + "loss": 1.0339, + "step": 10415 + }, + { + "epoch": 0.77, + "learning_rate": 1.9497689492931787e-05, + "loss": 1.0425, + "step": 10416 + }, + { + "epoch": 0.77, + "learning_rate": 1.9497564657712477e-05, + "loss": 1.1376, + "step": 10417 + }, + { + "epoch": 0.77, + "learning_rate": 1.949743980738268e-05, + "loss": 1.0662, + "step": 10418 + }, + { + "epoch": 0.77, + "learning_rate": 1.949731494194259e-05, + "loss": 1.0169, + "step": 10419 + }, + { + "epoch": 0.77, + "learning_rate": 1.949719006139241e-05, + "loss": 1.0452, + "step": 10420 + }, + { + "epoch": 0.77, + "learning_rate": 1.9497065165732334e-05, + "loss": 0.8748, + "step": 10421 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496940254962564e-05, + "loss": 1.0434, + "step": 10422 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496815329083302e-05, + "loss": 0.9522, + "step": 10423 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496690388094737e-05, + "loss": 1.0884, + "step": 10424 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496565431997076e-05, + "loss": 1.0459, + "step": 10425 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496440460790512e-05, + "loss": 1.0346, + "step": 10426 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496315474475252e-05, + "loss": 1.0706, + "step": 10427 + }, + { + "epoch": 0.77, + "learning_rate": 1.9496190473051486e-05, + "loss": 0.9442, + "step": 10428 + }, + { + "epoch": 0.77, + "learning_rate": 1.949606545651942e-05, + "loss": 1.092, + "step": 10429 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495940424879248e-05, + "loss": 1.0307, + "step": 10430 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495815378131173e-05, + "loss": 1.0221, + "step": 10431 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495690316275387e-05, + "loss": 1.1106, + "step": 10432 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495565239312094e-05, + "loss": 1.1075, + "step": 10433 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495440147241495e-05, + "loss": 1.0985, + "step": 10434 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495315040063788e-05, + "loss": 1.0785, + "step": 10435 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495189917779168e-05, + "loss": 0.9989, + "step": 10436 + }, + { + "epoch": 0.77, + "learning_rate": 1.9495064780387837e-05, + "loss": 1.0123, + "step": 10437 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494939627889998e-05, + "loss": 1.0297, + "step": 10438 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494814460285842e-05, + "loss": 1.0091, + "step": 10439 + }, + { + "epoch": 0.77, + "learning_rate": 1.949468927757557e-05, + "loss": 0.9889, + "step": 10440 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494564079759387e-05, + "loss": 1.1039, + "step": 10441 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494438866837487e-05, + "loss": 1.0403, + "step": 10442 + }, + { + "epoch": 0.77, + "learning_rate": 1.949431363881007e-05, + "loss": 0.8789, + "step": 10443 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494188395677338e-05, + "loss": 1.0408, + "step": 10444 + }, + { + "epoch": 0.77, + "learning_rate": 1.9494063137439484e-05, + "loss": 1.1189, + "step": 10445 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493937864096715e-05, + "loss": 1.1035, + "step": 10446 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493812575649225e-05, + "loss": 1.042, + "step": 10447 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493687272097215e-05, + "loss": 1.1072, + "step": 10448 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493561953440884e-05, + "loss": 0.9898, + "step": 10449 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493436619680435e-05, + "loss": 0.9724, + "step": 10450 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493311270816063e-05, + "loss": 1.0528, + "step": 10451 + }, + { + "epoch": 0.77, + "learning_rate": 1.949318590684797e-05, + "loss": 1.1076, + "step": 10452 + }, + { + "epoch": 0.77, + "learning_rate": 1.9493060527776348e-05, + "loss": 0.931, + "step": 10453 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492935133601406e-05, + "loss": 1.0222, + "step": 10454 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492809724323337e-05, + "loss": 1.0167, + "step": 10455 + }, + { + "epoch": 0.77, + "learning_rate": 1.949268429994235e-05, + "loss": 1.07, + "step": 10456 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492558860458635e-05, + "loss": 1.1001, + "step": 10457 + }, + { + "epoch": 0.77, + "learning_rate": 1.949243340587239e-05, + "loss": 1.0554, + "step": 10458 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492307936183828e-05, + "loss": 1.0736, + "step": 10459 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492182451393135e-05, + "loss": 0.9487, + "step": 10460 + }, + { + "epoch": 0.77, + "learning_rate": 1.9492056951500515e-05, + "loss": 1.1233, + "step": 10461 + }, + { + "epoch": 0.77, + "learning_rate": 1.949193143650617e-05, + "loss": 0.9901, + "step": 10462 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491805906410298e-05, + "loss": 0.9452, + "step": 10463 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491680361213098e-05, + "loss": 0.9496, + "step": 10464 + }, + { + "epoch": 0.77, + "learning_rate": 1.949155480091477e-05, + "loss": 1.0769, + "step": 10465 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491429225515516e-05, + "loss": 1.0417, + "step": 10466 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491303635015528e-05, + "loss": 1.0827, + "step": 10467 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491178029415015e-05, + "loss": 1.0568, + "step": 10468 + }, + { + "epoch": 0.77, + "learning_rate": 1.9491052408714176e-05, + "loss": 1.0218, + "step": 10469 + }, + { + "epoch": 0.77, + "learning_rate": 1.949092677291321e-05, + "loss": 1.0743, + "step": 10470 + }, + { + "epoch": 0.77, + "learning_rate": 1.949080112201231e-05, + "loss": 1.0996, + "step": 10471 + }, + { + "epoch": 0.77, + "learning_rate": 1.9490675456011684e-05, + "loss": 1.0746, + "step": 10472 + }, + { + "epoch": 0.77, + "learning_rate": 1.9490549774911527e-05, + "loss": 0.9241, + "step": 10473 + }, + { + "epoch": 0.77, + "learning_rate": 1.9490424078712042e-05, + "loss": 1.0347, + "step": 10474 + }, + { + "epoch": 0.77, + "learning_rate": 1.949029836741343e-05, + "loss": 1.0722, + "step": 10475 + }, + { + "epoch": 0.77, + "learning_rate": 1.9490172641015888e-05, + "loss": 1.0822, + "step": 10476 + }, + { + "epoch": 0.77, + "learning_rate": 1.9490046899519615e-05, + "loss": 1.0304, + "step": 10477 + }, + { + "epoch": 0.77, + "learning_rate": 1.9489921142924814e-05, + "loss": 1.094, + "step": 10478 + }, + { + "epoch": 0.77, + "learning_rate": 1.948979537123169e-05, + "loss": 1.0259, + "step": 10479 + }, + { + "epoch": 0.77, + "learning_rate": 1.948966958444043e-05, + "loss": 0.9753, + "step": 10480 + }, + { + "epoch": 0.77, + "learning_rate": 1.9489543782551244e-05, + "loss": 0.9574, + "step": 10481 + }, + { + "epoch": 0.77, + "learning_rate": 1.948941796556433e-05, + "loss": 0.9826, + "step": 10482 + }, + { + "epoch": 0.77, + "learning_rate": 1.9489292133479888e-05, + "loss": 1.0964, + "step": 10483 + }, + { + "epoch": 0.77, + "learning_rate": 1.9489166286298115e-05, + "loss": 0.9455, + "step": 10484 + }, + { + "epoch": 0.77, + "learning_rate": 1.948904042401922e-05, + "loss": 1.0512, + "step": 10485 + }, + { + "epoch": 0.77, + "learning_rate": 1.9488914546643393e-05, + "loss": 1.0377, + "step": 10486 + }, + { + "epoch": 0.77, + "learning_rate": 1.948878865417084e-05, + "loss": 1.0884, + "step": 10487 + }, + { + "epoch": 0.77, + "learning_rate": 1.948866274660176e-05, + "loss": 1.038, + "step": 10488 + }, + { + "epoch": 0.78, + "learning_rate": 1.948853682393635e-05, + "loss": 1.0072, + "step": 10489 + }, + { + "epoch": 0.78, + "learning_rate": 1.948841088617482e-05, + "loss": 1.11, + "step": 10490 + }, + { + "epoch": 0.78, + "learning_rate": 1.948828493331736e-05, + "loss": 0.9554, + "step": 10491 + }, + { + "epoch": 0.78, + "learning_rate": 1.9488158965364174e-05, + "loss": 1.054, + "step": 10492 + }, + { + "epoch": 0.78, + "learning_rate": 1.9488032982315467e-05, + "loss": 1.0303, + "step": 10493 + }, + { + "epoch": 0.78, + "learning_rate": 1.948790698417143e-05, + "loss": 1.0234, + "step": 10494 + }, + { + "epoch": 0.78, + "learning_rate": 1.9487780970932273e-05, + "loss": 1.0194, + "step": 10495 + }, + { + "epoch": 0.78, + "learning_rate": 1.948765494259819e-05, + "loss": 1.0352, + "step": 10496 + }, + { + "epoch": 0.78, + "learning_rate": 1.9487528899169383e-05, + "loss": 1.048, + "step": 10497 + }, + { + "epoch": 0.78, + "learning_rate": 1.9487402840646056e-05, + "loss": 1.0796, + "step": 10498 + }, + { + "epoch": 0.78, + "learning_rate": 1.9487276767028405e-05, + "loss": 1.0602, + "step": 10499 + }, + { + "epoch": 0.78, + "learning_rate": 1.948715067831663e-05, + "loss": 1.027, + "step": 10500 + }, + { + "epoch": 0.78, + "learning_rate": 1.948702457451094e-05, + "loss": 0.9796, + "step": 10501 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486898455611524e-05, + "loss": 0.9509, + "step": 10502 + }, + { + "epoch": 0.78, + "learning_rate": 1.948677232161859e-05, + "loss": 1.0966, + "step": 10503 + }, + { + "epoch": 0.78, + "learning_rate": 1.948664617253234e-05, + "loss": 1.043, + "step": 10504 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486520008352967e-05, + "loss": 1.0809, + "step": 10505 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486393829080678e-05, + "loss": 1.0409, + "step": 10506 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486267634715675e-05, + "loss": 1.0132, + "step": 10507 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486141425258155e-05, + "loss": 0.9951, + "step": 10508 + }, + { + "epoch": 0.78, + "learning_rate": 1.9486015200708316e-05, + "loss": 0.8663, + "step": 10509 + }, + { + "epoch": 0.78, + "learning_rate": 1.9485888961066367e-05, + "loss": 1.0004, + "step": 10510 + }, + { + "epoch": 0.78, + "learning_rate": 1.94857627063325e-05, + "loss": 1.0726, + "step": 10511 + }, + { + "epoch": 0.78, + "learning_rate": 1.948563643650692e-05, + "loss": 1.0286, + "step": 10512 + }, + { + "epoch": 0.78, + "learning_rate": 1.9485510151589833e-05, + "loss": 1.0152, + "step": 10513 + }, + { + "epoch": 0.78, + "learning_rate": 1.9485383851581435e-05, + "loss": 1.0999, + "step": 10514 + }, + { + "epoch": 0.78, + "learning_rate": 1.948525753648192e-05, + "loss": 1.0022, + "step": 10515 + }, + { + "epoch": 0.78, + "learning_rate": 1.9485131206291502e-05, + "loss": 1.0631, + "step": 10516 + }, + { + "epoch": 0.78, + "learning_rate": 1.9485004861010375e-05, + "loss": 1.1117, + "step": 10517 + }, + { + "epoch": 0.78, + "learning_rate": 1.9484878500638742e-05, + "loss": 1.0447, + "step": 10518 + }, + { + "epoch": 0.78, + "learning_rate": 1.94847521251768e-05, + "loss": 0.9299, + "step": 10519 + }, + { + "epoch": 0.78, + "learning_rate": 1.9484625734624755e-05, + "loss": 0.9977, + "step": 10520 + }, + { + "epoch": 0.78, + "learning_rate": 1.9484499328982806e-05, + "loss": 1.1419, + "step": 10521 + }, + { + "epoch": 0.78, + "learning_rate": 1.9484372908251155e-05, + "loss": 1.0772, + "step": 10522 + }, + { + "epoch": 0.78, + "learning_rate": 1.9484246472430002e-05, + "loss": 1.0612, + "step": 10523 + }, + { + "epoch": 0.78, + "learning_rate": 1.948412002151955e-05, + "loss": 0.9956, + "step": 10524 + }, + { + "epoch": 0.78, + "learning_rate": 1.9483993555519995e-05, + "loss": 1.1522, + "step": 10525 + }, + { + "epoch": 0.78, + "learning_rate": 1.948386707443154e-05, + "loss": 1.0206, + "step": 10526 + }, + { + "epoch": 0.78, + "learning_rate": 1.9483740578254394e-05, + "loss": 1.0544, + "step": 10527 + }, + { + "epoch": 0.78, + "learning_rate": 1.9483614066988752e-05, + "loss": 1.09, + "step": 10528 + }, + { + "epoch": 0.78, + "learning_rate": 1.9483487540634813e-05, + "loss": 1.0611, + "step": 10529 + }, + { + "epoch": 0.78, + "learning_rate": 1.948336099919278e-05, + "loss": 0.9558, + "step": 10530 + }, + { + "epoch": 0.78, + "learning_rate": 1.948323444266286e-05, + "loss": 1.1027, + "step": 10531 + }, + { + "epoch": 0.78, + "learning_rate": 1.9483107871045246e-05, + "loss": 0.9325, + "step": 10532 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482981284340145e-05, + "loss": 1.127, + "step": 10533 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482854682547755e-05, + "loss": 0.9706, + "step": 10534 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482728065668278e-05, + "loss": 1.089, + "step": 10535 + }, + { + "epoch": 0.78, + "learning_rate": 1.948260143370192e-05, + "loss": 1.0827, + "step": 10536 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482474786648877e-05, + "loss": 0.9546, + "step": 10537 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482348124509353e-05, + "loss": 1.1184, + "step": 10538 + }, + { + "epoch": 0.78, + "learning_rate": 1.9482221447283546e-05, + "loss": 0.9821, + "step": 10539 + }, + { + "epoch": 0.78, + "learning_rate": 1.948209475497166e-05, + "loss": 1.023, + "step": 10540 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481968047573904e-05, + "loss": 1.1524, + "step": 10541 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481841325090465e-05, + "loss": 0.9534, + "step": 10542 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481714587521553e-05, + "loss": 0.9801, + "step": 10543 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481587834867373e-05, + "loss": 1.0151, + "step": 10544 + }, + { + "epoch": 0.78, + "learning_rate": 1.948146106712812e-05, + "loss": 1.0492, + "step": 10545 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481334284303995e-05, + "loss": 0.9834, + "step": 10546 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481207486395207e-05, + "loss": 1.071, + "step": 10547 + }, + { + "epoch": 0.78, + "learning_rate": 1.9481080673401955e-05, + "loss": 0.9568, + "step": 10548 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480953845324434e-05, + "loss": 0.9706, + "step": 10549 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480827002162853e-05, + "loss": 1.0966, + "step": 10550 + }, + { + "epoch": 0.78, + "learning_rate": 1.948070014391741e-05, + "loss": 1.0622, + "step": 10551 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480573270588318e-05, + "loss": 0.9948, + "step": 10552 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480446382175763e-05, + "loss": 1.0177, + "step": 10553 + }, + { + "epoch": 0.78, + "learning_rate": 1.948031947867995e-05, + "loss": 1.1655, + "step": 10554 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480192560101087e-05, + "loss": 1.0451, + "step": 10555 + }, + { + "epoch": 0.78, + "learning_rate": 1.9480065626439375e-05, + "loss": 1.1374, + "step": 10556 + }, + { + "epoch": 0.78, + "learning_rate": 1.9479938677695012e-05, + "loss": 1.2389, + "step": 10557 + }, + { + "epoch": 0.78, + "learning_rate": 1.9479811713868204e-05, + "loss": 1.1126, + "step": 10558 + }, + { + "epoch": 0.78, + "learning_rate": 1.947968473495915e-05, + "loss": 1.1112, + "step": 10559 + }, + { + "epoch": 0.78, + "learning_rate": 1.9479557740968052e-05, + "loss": 1.1207, + "step": 10560 + }, + { + "epoch": 0.78, + "learning_rate": 1.9479430731895116e-05, + "loss": 0.9819, + "step": 10561 + }, + { + "epoch": 0.78, + "learning_rate": 1.947930370774054e-05, + "loss": 1.0009, + "step": 10562 + }, + { + "epoch": 0.78, + "learning_rate": 1.947917666850453e-05, + "loss": 1.0542, + "step": 10563 + }, + { + "epoch": 0.78, + "learning_rate": 1.947904961418728e-05, + "loss": 0.986, + "step": 10564 + }, + { + "epoch": 0.78, + "learning_rate": 1.9478922544789004e-05, + "loss": 0.98, + "step": 10565 + }, + { + "epoch": 0.78, + "learning_rate": 1.9478795460309894e-05, + "loss": 1.0208, + "step": 10566 + }, + { + "epoch": 0.78, + "learning_rate": 1.9478668360750156e-05, + "loss": 1.0263, + "step": 10567 + }, + { + "epoch": 0.78, + "learning_rate": 1.9478541246109996e-05, + "loss": 1.0482, + "step": 10568 + }, + { + "epoch": 0.78, + "learning_rate": 1.947841411638961e-05, + "loss": 0.8627, + "step": 10569 + }, + { + "epoch": 0.78, + "learning_rate": 1.94782869715892e-05, + "loss": 1.0825, + "step": 10570 + }, + { + "epoch": 0.78, + "learning_rate": 1.9478159811708973e-05, + "loss": 1.1003, + "step": 10571 + }, + { + "epoch": 0.78, + "learning_rate": 1.947803263674913e-05, + "loss": 1.0345, + "step": 10572 + }, + { + "epoch": 0.78, + "learning_rate": 1.947790544670987e-05, + "loss": 0.9896, + "step": 10573 + }, + { + "epoch": 0.78, + "learning_rate": 1.9477778241591406e-05, + "loss": 1.1718, + "step": 10574 + }, + { + "epoch": 0.78, + "learning_rate": 1.9477651021393925e-05, + "loss": 0.9996, + "step": 10575 + }, + { + "epoch": 0.78, + "learning_rate": 1.947752378611764e-05, + "loss": 1.1015, + "step": 10576 + }, + { + "epoch": 0.78, + "learning_rate": 1.9477396535762745e-05, + "loss": 1.0578, + "step": 10577 + }, + { + "epoch": 0.78, + "learning_rate": 1.9477269270329456e-05, + "loss": 1.0877, + "step": 10578 + }, + { + "epoch": 0.78, + "learning_rate": 1.947714198981796e-05, + "loss": 1.0544, + "step": 10579 + }, + { + "epoch": 0.78, + "learning_rate": 1.947701469422847e-05, + "loss": 1.0506, + "step": 10580 + }, + { + "epoch": 0.78, + "learning_rate": 1.9476887383561185e-05, + "loss": 1.0692, + "step": 10581 + }, + { + "epoch": 0.78, + "learning_rate": 1.9476760057816308e-05, + "loss": 0.9629, + "step": 10582 + }, + { + "epoch": 0.78, + "learning_rate": 1.947663271699404e-05, + "loss": 1.0105, + "step": 10583 + }, + { + "epoch": 0.78, + "learning_rate": 1.9476505361094587e-05, + "loss": 0.9775, + "step": 10584 + }, + { + "epoch": 0.78, + "learning_rate": 1.947637799011815e-05, + "loss": 1.0344, + "step": 10585 + }, + { + "epoch": 0.78, + "learning_rate": 1.947625060406493e-05, + "loss": 1.0781, + "step": 10586 + }, + { + "epoch": 0.78, + "learning_rate": 1.9476123202935126e-05, + "loss": 1.0685, + "step": 10587 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475995786728952e-05, + "loss": 1.1203, + "step": 10588 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475868355446602e-05, + "loss": 1.0577, + "step": 10589 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475740909088282e-05, + "loss": 1.0189, + "step": 10590 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475613447654195e-05, + "loss": 0.9671, + "step": 10591 + }, + { + "epoch": 0.78, + "learning_rate": 1.947548597114454e-05, + "loss": 1.0253, + "step": 10592 + }, + { + "epoch": 0.78, + "learning_rate": 1.947535847955952e-05, + "loss": 1.0237, + "step": 10593 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475230972899344e-05, + "loss": 1.0103, + "step": 10594 + }, + { + "epoch": 0.78, + "learning_rate": 1.9475103451164213e-05, + "loss": 1.0128, + "step": 10595 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474975914354324e-05, + "loss": 0.9524, + "step": 10596 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474848362469884e-05, + "loss": 0.9675, + "step": 10597 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474720795511097e-05, + "loss": 1.1394, + "step": 10598 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474593213478167e-05, + "loss": 1.0271, + "step": 10599 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474465616371292e-05, + "loss": 0.9295, + "step": 10600 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474338004190677e-05, + "loss": 1.0406, + "step": 10601 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474210376936527e-05, + "loss": 0.9245, + "step": 10602 + }, + { + "epoch": 0.78, + "learning_rate": 1.9474082734609047e-05, + "loss": 1.0025, + "step": 10603 + }, + { + "epoch": 0.78, + "learning_rate": 1.947395507720843e-05, + "loss": 1.0982, + "step": 10604 + }, + { + "epoch": 0.78, + "learning_rate": 1.947382740473489e-05, + "loss": 1.0506, + "step": 10605 + }, + { + "epoch": 0.78, + "learning_rate": 1.947369971718863e-05, + "loss": 1.0255, + "step": 10606 + }, + { + "epoch": 0.78, + "learning_rate": 1.9473572014569844e-05, + "loss": 1.2057, + "step": 10607 + }, + { + "epoch": 0.78, + "learning_rate": 1.9473444296878742e-05, + "loss": 1.0252, + "step": 10608 + }, + { + "epoch": 0.78, + "learning_rate": 1.9473316564115526e-05, + "loss": 0.9936, + "step": 10609 + }, + { + "epoch": 0.78, + "learning_rate": 1.94731888162804e-05, + "loss": 0.9197, + "step": 10610 + }, + { + "epoch": 0.78, + "learning_rate": 1.947306105337356e-05, + "loss": 1.1049, + "step": 10611 + }, + { + "epoch": 0.78, + "learning_rate": 1.9472933275395224e-05, + "loss": 1.0332, + "step": 10612 + }, + { + "epoch": 0.78, + "learning_rate": 1.947280548234558e-05, + "loss": 1.1074, + "step": 10613 + }, + { + "epoch": 0.78, + "learning_rate": 1.947267767422484e-05, + "loss": 1.0963, + "step": 10614 + }, + { + "epoch": 0.78, + "learning_rate": 1.947254985103321e-05, + "loss": 1.0399, + "step": 10615 + }, + { + "epoch": 0.78, + "learning_rate": 1.9472422012770882e-05, + "loss": 1.0509, + "step": 10616 + }, + { + "epoch": 0.78, + "learning_rate": 1.947229415943807e-05, + "loss": 1.0494, + "step": 10617 + }, + { + "epoch": 0.78, + "learning_rate": 1.947216629103497e-05, + "loss": 1.0481, + "step": 10618 + }, + { + "epoch": 0.78, + "learning_rate": 1.947203840756179e-05, + "loss": 1.145, + "step": 10619 + }, + { + "epoch": 0.78, + "learning_rate": 1.9471910509018737e-05, + "loss": 1.0425, + "step": 10620 + }, + { + "epoch": 0.78, + "learning_rate": 1.9471782595406004e-05, + "loss": 0.9709, + "step": 10621 + }, + { + "epoch": 0.78, + "learning_rate": 1.9471654666723807e-05, + "loss": 0.9376, + "step": 10622 + }, + { + "epoch": 0.78, + "learning_rate": 1.9471526722972337e-05, + "loss": 1.0099, + "step": 10623 + }, + { + "epoch": 0.79, + "learning_rate": 1.9471398764151806e-05, + "loss": 0.9094, + "step": 10624 + }, + { + "epoch": 0.79, + "learning_rate": 1.9471270790262413e-05, + "loss": 0.9903, + "step": 10625 + }, + { + "epoch": 0.79, + "learning_rate": 1.9471142801304368e-05, + "loss": 1.098, + "step": 10626 + }, + { + "epoch": 0.79, + "learning_rate": 1.9471014797277865e-05, + "loss": 1.0529, + "step": 10627 + }, + { + "epoch": 0.79, + "learning_rate": 1.947088677818312e-05, + "loss": 0.9623, + "step": 10628 + }, + { + "epoch": 0.79, + "learning_rate": 1.9470758744020322e-05, + "loss": 1.0817, + "step": 10629 + }, + { + "epoch": 0.79, + "learning_rate": 1.9470630694789686e-05, + "loss": 1.0993, + "step": 10630 + }, + { + "epoch": 0.79, + "learning_rate": 1.9470502630491412e-05, + "loss": 1.0425, + "step": 10631 + }, + { + "epoch": 0.79, + "learning_rate": 1.9470374551125705e-05, + "loss": 1.0865, + "step": 10632 + }, + { + "epoch": 0.79, + "learning_rate": 1.9470246456692763e-05, + "loss": 0.9641, + "step": 10633 + }, + { + "epoch": 0.79, + "learning_rate": 1.94701183471928e-05, + "loss": 0.9584, + "step": 10634 + }, + { + "epoch": 0.79, + "learning_rate": 1.946999022262601e-05, + "loss": 1.0542, + "step": 10635 + }, + { + "epoch": 0.79, + "learning_rate": 1.94698620829926e-05, + "loss": 1.1112, + "step": 10636 + }, + { + "epoch": 0.79, + "learning_rate": 1.946973392829278e-05, + "loss": 0.9633, + "step": 10637 + }, + { + "epoch": 0.79, + "learning_rate": 1.9469605758526747e-05, + "loss": 1.132, + "step": 10638 + }, + { + "epoch": 0.79, + "learning_rate": 1.9469477573694705e-05, + "loss": 1.0075, + "step": 10639 + }, + { + "epoch": 0.79, + "learning_rate": 1.9469349373796864e-05, + "loss": 1.0582, + "step": 10640 + }, + { + "epoch": 0.79, + "learning_rate": 1.9469221158833417e-05, + "loss": 0.9932, + "step": 10641 + }, + { + "epoch": 0.79, + "learning_rate": 1.9469092928804577e-05, + "loss": 1.0664, + "step": 10642 + }, + { + "epoch": 0.79, + "learning_rate": 1.9468964683710546e-05, + "loss": 0.9971, + "step": 10643 + }, + { + "epoch": 0.79, + "learning_rate": 1.9468836423551528e-05, + "loss": 1.0448, + "step": 10644 + }, + { + "epoch": 0.79, + "learning_rate": 1.9468708148327728e-05, + "loss": 1.0086, + "step": 10645 + }, + { + "epoch": 0.79, + "learning_rate": 1.946857985803935e-05, + "loss": 0.9614, + "step": 10646 + }, + { + "epoch": 0.79, + "learning_rate": 1.9468451552686594e-05, + "loss": 0.9948, + "step": 10647 + }, + { + "epoch": 0.79, + "learning_rate": 1.946832323226967e-05, + "loss": 0.9345, + "step": 10648 + }, + { + "epoch": 0.79, + "learning_rate": 1.9468194896788775e-05, + "loss": 1.0628, + "step": 10649 + }, + { + "epoch": 0.79, + "learning_rate": 1.946806654624412e-05, + "loss": 1.0535, + "step": 10650 + }, + { + "epoch": 0.79, + "learning_rate": 1.9467938180635903e-05, + "loss": 1.0464, + "step": 10651 + }, + { + "epoch": 0.79, + "learning_rate": 1.9467809799964338e-05, + "loss": 0.9343, + "step": 10652 + }, + { + "epoch": 0.79, + "learning_rate": 1.946768140422962e-05, + "loss": 1.0094, + "step": 10653 + }, + { + "epoch": 0.79, + "learning_rate": 1.9467552993431955e-05, + "loss": 1.0984, + "step": 10654 + }, + { + "epoch": 0.79, + "learning_rate": 1.9467424567571548e-05, + "loss": 0.9517, + "step": 10655 + }, + { + "epoch": 0.79, + "learning_rate": 1.946729612664861e-05, + "loss": 1.074, + "step": 10656 + }, + { + "epoch": 0.79, + "learning_rate": 1.9467167670663332e-05, + "loss": 1.0015, + "step": 10657 + }, + { + "epoch": 0.79, + "learning_rate": 1.946703919961593e-05, + "loss": 1.1, + "step": 10658 + }, + { + "epoch": 0.79, + "learning_rate": 1.94669107135066e-05, + "loss": 1.0071, + "step": 10659 + }, + { + "epoch": 0.79, + "learning_rate": 1.9466782212335556e-05, + "loss": 0.9938, + "step": 10660 + }, + { + "epoch": 0.79, + "learning_rate": 1.946665369610299e-05, + "loss": 1.0191, + "step": 10661 + }, + { + "epoch": 0.79, + "learning_rate": 1.9466525164809124e-05, + "loss": 1.0597, + "step": 10662 + }, + { + "epoch": 0.79, + "learning_rate": 1.9466396618454143e-05, + "loss": 1.0741, + "step": 10663 + }, + { + "epoch": 0.79, + "learning_rate": 1.9466268057038263e-05, + "loss": 1.0454, + "step": 10664 + }, + { + "epoch": 0.79, + "learning_rate": 1.9466139480561686e-05, + "loss": 0.9617, + "step": 10665 + }, + { + "epoch": 0.79, + "learning_rate": 1.946601088902462e-05, + "loss": 1.0403, + "step": 10666 + }, + { + "epoch": 0.79, + "learning_rate": 1.946588228242726e-05, + "loss": 1.1152, + "step": 10667 + }, + { + "epoch": 0.79, + "learning_rate": 1.946575366076982e-05, + "loss": 1.0513, + "step": 10668 + }, + { + "epoch": 0.79, + "learning_rate": 1.94656250240525e-05, + "loss": 1.1441, + "step": 10669 + }, + { + "epoch": 0.79, + "learning_rate": 1.9465496372275508e-05, + "loss": 1.03, + "step": 10670 + }, + { + "epoch": 0.79, + "learning_rate": 1.9465367705439045e-05, + "loss": 0.9317, + "step": 10671 + }, + { + "epoch": 0.79, + "learning_rate": 1.946523902354332e-05, + "loss": 0.9915, + "step": 10672 + }, + { + "epoch": 0.79, + "learning_rate": 1.9465110326588533e-05, + "loss": 1.0384, + "step": 10673 + }, + { + "epoch": 0.79, + "learning_rate": 1.946498161457489e-05, + "loss": 1.0061, + "step": 10674 + }, + { + "epoch": 0.79, + "learning_rate": 1.9464852887502595e-05, + "loss": 0.956, + "step": 10675 + }, + { + "epoch": 0.79, + "learning_rate": 1.946472414537186e-05, + "loss": 0.99, + "step": 10676 + }, + { + "epoch": 0.79, + "learning_rate": 1.946459538818288e-05, + "loss": 1.0885, + "step": 10677 + }, + { + "epoch": 0.79, + "learning_rate": 1.9464466615935868e-05, + "loss": 1.1221, + "step": 10678 + }, + { + "epoch": 0.79, + "learning_rate": 1.9464337828631018e-05, + "loss": 1.0186, + "step": 10679 + }, + { + "epoch": 0.79, + "learning_rate": 1.946420902626855e-05, + "loss": 0.9871, + "step": 10680 + }, + { + "epoch": 0.79, + "learning_rate": 1.9464080208848656e-05, + "loss": 1.0836, + "step": 10681 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463951376371546e-05, + "loss": 1.0454, + "step": 10682 + }, + { + "epoch": 0.79, + "learning_rate": 1.946382252883743e-05, + "loss": 1.0051, + "step": 10683 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463693666246503e-05, + "loss": 0.9957, + "step": 10684 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463564788598973e-05, + "loss": 1.0417, + "step": 10685 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463435895895052e-05, + "loss": 1.0126, + "step": 10686 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463306988134935e-05, + "loss": 1.0829, + "step": 10687 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463178065318834e-05, + "loss": 1.0275, + "step": 10688 + }, + { + "epoch": 0.79, + "learning_rate": 1.9463049127446953e-05, + "loss": 1.0571, + "step": 10689 + }, + { + "epoch": 0.79, + "learning_rate": 1.9462920174519495e-05, + "loss": 1.0533, + "step": 10690 + }, + { + "epoch": 0.79, + "learning_rate": 1.9462791206536666e-05, + "loss": 1.02, + "step": 10691 + }, + { + "epoch": 0.79, + "learning_rate": 1.946266222349867e-05, + "loss": 1.0716, + "step": 10692 + }, + { + "epoch": 0.79, + "learning_rate": 1.946253322540572e-05, + "loss": 1.0029, + "step": 10693 + }, + { + "epoch": 0.79, + "learning_rate": 1.946240421225801e-05, + "loss": 0.9729, + "step": 10694 + }, + { + "epoch": 0.79, + "learning_rate": 1.9462275184055753e-05, + "loss": 1.0381, + "step": 10695 + }, + { + "epoch": 0.79, + "learning_rate": 1.946214614079915e-05, + "loss": 1.0646, + "step": 10696 + }, + { + "epoch": 0.79, + "learning_rate": 1.9462017082488406e-05, + "loss": 1.0695, + "step": 10697 + }, + { + "epoch": 0.79, + "learning_rate": 1.9461888009123732e-05, + "loss": 1.0466, + "step": 10698 + }, + { + "epoch": 0.79, + "learning_rate": 1.9461758920705328e-05, + "loss": 0.9319, + "step": 10699 + }, + { + "epoch": 0.79, + "learning_rate": 1.94616298172334e-05, + "loss": 1.017, + "step": 10700 + }, + { + "epoch": 0.79, + "learning_rate": 1.9461500698708154e-05, + "loss": 1.0878, + "step": 10701 + }, + { + "epoch": 0.79, + "learning_rate": 1.9461371565129798e-05, + "loss": 0.9446, + "step": 10702 + }, + { + "epoch": 0.79, + "learning_rate": 1.9461242416498532e-05, + "loss": 1.0721, + "step": 10703 + }, + { + "epoch": 0.79, + "learning_rate": 1.946111325281457e-05, + "loss": 0.9143, + "step": 10704 + }, + { + "epoch": 0.79, + "learning_rate": 1.946098407407811e-05, + "loss": 1.0195, + "step": 10705 + }, + { + "epoch": 0.79, + "learning_rate": 1.9460854880289358e-05, + "loss": 1.0148, + "step": 10706 + }, + { + "epoch": 0.79, + "learning_rate": 1.946072567144852e-05, + "loss": 1.0352, + "step": 10707 + }, + { + "epoch": 0.79, + "learning_rate": 1.9460596447555804e-05, + "loss": 0.893, + "step": 10708 + }, + { + "epoch": 0.79, + "learning_rate": 1.9460467208611415e-05, + "loss": 1.0352, + "step": 10709 + }, + { + "epoch": 0.79, + "learning_rate": 1.946033795461556e-05, + "loss": 1.133, + "step": 10710 + }, + { + "epoch": 0.79, + "learning_rate": 1.946020868556844e-05, + "loss": 1.0873, + "step": 10711 + }, + { + "epoch": 0.79, + "learning_rate": 1.9460079401470265e-05, + "loss": 1.0722, + "step": 10712 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459950102321238e-05, + "loss": 0.8998, + "step": 10713 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459820788121563e-05, + "loss": 0.9203, + "step": 10714 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459691458871454e-05, + "loss": 0.9791, + "step": 10715 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459562114571108e-05, + "loss": 1.1274, + "step": 10716 + }, + { + "epoch": 0.79, + "learning_rate": 1.945943275522073e-05, + "loss": 1.1281, + "step": 10717 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459303380820538e-05, + "loss": 1.0803, + "step": 10718 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459173991370727e-05, + "loss": 1.0274, + "step": 10719 + }, + { + "epoch": 0.79, + "learning_rate": 1.9459044586871502e-05, + "loss": 1.0976, + "step": 10720 + }, + { + "epoch": 0.79, + "learning_rate": 1.9458915167323077e-05, + "loss": 1.0088, + "step": 10721 + }, + { + "epoch": 0.79, + "learning_rate": 1.945878573272565e-05, + "loss": 1.0939, + "step": 10722 + }, + { + "epoch": 0.79, + "learning_rate": 1.945865628307943e-05, + "loss": 1.1465, + "step": 10723 + }, + { + "epoch": 0.79, + "learning_rate": 1.9458526818384627e-05, + "loss": 1.0444, + "step": 10724 + }, + { + "epoch": 0.79, + "learning_rate": 1.9458397338641443e-05, + "loss": 1.0861, + "step": 10725 + }, + { + "epoch": 0.79, + "learning_rate": 1.945826784385008e-05, + "loss": 0.9428, + "step": 10726 + }, + { + "epoch": 0.79, + "learning_rate": 1.945813833401075e-05, + "loss": 1.1189, + "step": 10727 + }, + { + "epoch": 0.79, + "learning_rate": 1.9458008809123657e-05, + "loss": 1.0593, + "step": 10728 + }, + { + "epoch": 0.79, + "learning_rate": 1.945787926918901e-05, + "loss": 1.0478, + "step": 10729 + }, + { + "epoch": 0.79, + "learning_rate": 1.945774971420701e-05, + "loss": 1.1087, + "step": 10730 + }, + { + "epoch": 0.79, + "learning_rate": 1.9457620144177865e-05, + "loss": 1.1303, + "step": 10731 + }, + { + "epoch": 0.79, + "learning_rate": 1.945749055910178e-05, + "loss": 1.1409, + "step": 10732 + }, + { + "epoch": 0.79, + "learning_rate": 1.9457360958978965e-05, + "loss": 1.0197, + "step": 10733 + }, + { + "epoch": 0.79, + "learning_rate": 1.9457231343809625e-05, + "loss": 1.156, + "step": 10734 + }, + { + "epoch": 0.79, + "learning_rate": 1.9457101713593965e-05, + "loss": 1.1219, + "step": 10735 + }, + { + "epoch": 0.79, + "learning_rate": 1.945697206833219e-05, + "loss": 1.025, + "step": 10736 + }, + { + "epoch": 0.79, + "learning_rate": 1.945684240802451e-05, + "loss": 1.1143, + "step": 10737 + }, + { + "epoch": 0.79, + "learning_rate": 1.9456712732671125e-05, + "loss": 0.9379, + "step": 10738 + }, + { + "epoch": 0.79, + "learning_rate": 1.945658304227225e-05, + "loss": 1.1004, + "step": 10739 + }, + { + "epoch": 0.79, + "learning_rate": 1.9456453336828087e-05, + "loss": 0.9687, + "step": 10740 + }, + { + "epoch": 0.79, + "learning_rate": 1.9456323616338837e-05, + "loss": 1.0722, + "step": 10741 + }, + { + "epoch": 0.79, + "learning_rate": 1.9456193880804716e-05, + "loss": 0.9372, + "step": 10742 + }, + { + "epoch": 0.79, + "learning_rate": 1.9456064130225925e-05, + "loss": 1.0086, + "step": 10743 + }, + { + "epoch": 0.79, + "learning_rate": 1.945593436460267e-05, + "loss": 0.9695, + "step": 10744 + }, + { + "epoch": 0.79, + "learning_rate": 1.9455804583935158e-05, + "loss": 1.0063, + "step": 10745 + }, + { + "epoch": 0.79, + "learning_rate": 1.94556747882236e-05, + "loss": 1.1367, + "step": 10746 + }, + { + "epoch": 0.79, + "learning_rate": 1.9455544977468192e-05, + "loss": 1.0442, + "step": 10747 + }, + { + "epoch": 0.79, + "learning_rate": 1.945541515166915e-05, + "loss": 1.0246, + "step": 10748 + }, + { + "epoch": 0.79, + "learning_rate": 1.945528531082668e-05, + "loss": 1.0339, + "step": 10749 + }, + { + "epoch": 0.79, + "learning_rate": 1.9455155454940986e-05, + "loss": 1.2056, + "step": 10750 + }, + { + "epoch": 0.79, + "learning_rate": 1.9455025584012274e-05, + "loss": 1.0859, + "step": 10751 + }, + { + "epoch": 0.79, + "learning_rate": 1.9454895698040752e-05, + "loss": 0.9714, + "step": 10752 + }, + { + "epoch": 0.79, + "learning_rate": 1.9454765797026626e-05, + "loss": 0.967, + "step": 10753 + }, + { + "epoch": 0.79, + "learning_rate": 1.9454635880970102e-05, + "loss": 1.1065, + "step": 10754 + }, + { + "epoch": 0.79, + "learning_rate": 1.945450594987139e-05, + "loss": 1.1137, + "step": 10755 + }, + { + "epoch": 0.79, + "learning_rate": 1.945437600373069e-05, + "loss": 1.1216, + "step": 10756 + }, + { + "epoch": 0.79, + "learning_rate": 1.945424604254822e-05, + "loss": 1.0067, + "step": 10757 + }, + { + "epoch": 0.79, + "learning_rate": 1.9454116066324175e-05, + "loss": 1.101, + "step": 10758 + }, + { + "epoch": 0.79, + "learning_rate": 1.9453986075058766e-05, + "loss": 1.0928, + "step": 10759 + }, + { + "epoch": 0.8, + "learning_rate": 1.94538560687522e-05, + "loss": 1.0448, + "step": 10760 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453726047404688e-05, + "loss": 0.9917, + "step": 10761 + }, + { + "epoch": 0.8, + "learning_rate": 1.945359601101643e-05, + "loss": 0.987, + "step": 10762 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453465959587638e-05, + "loss": 1.0113, + "step": 10763 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453335893118517e-05, + "loss": 0.9154, + "step": 10764 + }, + { + "epoch": 0.8, + "learning_rate": 1.9453205811609273e-05, + "loss": 1.0105, + "step": 10765 + }, + { + "epoch": 0.8, + "learning_rate": 1.945307571506011e-05, + "loss": 1.0453, + "step": 10766 + }, + { + "epoch": 0.8, + "learning_rate": 1.9452945603471243e-05, + "loss": 1.0372, + "step": 10767 + }, + { + "epoch": 0.8, + "learning_rate": 1.9452815476842875e-05, + "loss": 1.0357, + "step": 10768 + }, + { + "epoch": 0.8, + "learning_rate": 1.9452685335175213e-05, + "loss": 1.042, + "step": 10769 + }, + { + "epoch": 0.8, + "learning_rate": 1.945255517846846e-05, + "loss": 1.0572, + "step": 10770 + }, + { + "epoch": 0.8, + "learning_rate": 1.945242500672283e-05, + "loss": 1.0747, + "step": 10771 + }, + { + "epoch": 0.8, + "learning_rate": 1.945229481993853e-05, + "loss": 1.041, + "step": 10772 + }, + { + "epoch": 0.8, + "learning_rate": 1.9452164618115762e-05, + "loss": 1.04, + "step": 10773 + }, + { + "epoch": 0.8, + "learning_rate": 1.9452034401254732e-05, + "loss": 1.0606, + "step": 10774 + }, + { + "epoch": 0.8, + "learning_rate": 1.9451904169355652e-05, + "loss": 1.066, + "step": 10775 + }, + { + "epoch": 0.8, + "learning_rate": 1.9451773922418732e-05, + "loss": 1.0014, + "step": 10776 + }, + { + "epoch": 0.8, + "learning_rate": 1.945164366044417e-05, + "loss": 0.9871, + "step": 10777 + }, + { + "epoch": 0.8, + "learning_rate": 1.945151338343218e-05, + "loss": 1.0142, + "step": 10778 + }, + { + "epoch": 0.8, + "learning_rate": 1.9451383091382968e-05, + "loss": 0.9982, + "step": 10779 + }, + { + "epoch": 0.8, + "learning_rate": 1.945125278429674e-05, + "loss": 1.0178, + "step": 10780 + }, + { + "epoch": 0.8, + "learning_rate": 1.9451122462173703e-05, + "loss": 0.9983, + "step": 10781 + }, + { + "epoch": 0.8, + "learning_rate": 1.945099212501407e-05, + "loss": 0.9966, + "step": 10782 + }, + { + "epoch": 0.8, + "learning_rate": 1.9450861772818035e-05, + "loss": 1.0559, + "step": 10783 + }, + { + "epoch": 0.8, + "learning_rate": 1.945073140558582e-05, + "loss": 1.0019, + "step": 10784 + }, + { + "epoch": 0.8, + "learning_rate": 1.9450601023317627e-05, + "loss": 0.9776, + "step": 10785 + }, + { + "epoch": 0.8, + "learning_rate": 1.9450470626013663e-05, + "loss": 0.9565, + "step": 10786 + }, + { + "epoch": 0.8, + "learning_rate": 1.9450340213674132e-05, + "loss": 1.0289, + "step": 10787 + }, + { + "epoch": 0.8, + "learning_rate": 1.945020978629925e-05, + "loss": 0.9502, + "step": 10788 + }, + { + "epoch": 0.8, + "learning_rate": 1.9450079343889215e-05, + "loss": 1.0715, + "step": 10789 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449948886444238e-05, + "loss": 0.9724, + "step": 10790 + }, + { + "epoch": 0.8, + "learning_rate": 1.944981841396453e-05, + "loss": 1.0201, + "step": 10791 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449687926450298e-05, + "loss": 1.1699, + "step": 10792 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449557423901745e-05, + "loss": 0.9655, + "step": 10793 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449426906319085e-05, + "loss": 0.9364, + "step": 10794 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449296373702516e-05, + "loss": 1.0462, + "step": 10795 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449165826052257e-05, + "loss": 1.104, + "step": 10796 + }, + { + "epoch": 0.8, + "learning_rate": 1.944903526336851e-05, + "loss": 0.9382, + "step": 10797 + }, + { + "epoch": 0.8, + "learning_rate": 1.9448904685651477e-05, + "loss": 1.0601, + "step": 10798 + }, + { + "epoch": 0.8, + "learning_rate": 1.944877409290138e-05, + "loss": 1.019, + "step": 10799 + }, + { + "epoch": 0.8, + "learning_rate": 1.944864348511841e-05, + "loss": 1.0989, + "step": 10800 + }, + { + "epoch": 0.8, + "learning_rate": 1.944851286230279e-05, + "loss": 1.1164, + "step": 10801 + }, + { + "epoch": 0.8, + "learning_rate": 1.944838222445472e-05, + "loss": 1.0415, + "step": 10802 + }, + { + "epoch": 0.8, + "learning_rate": 1.9448251571574408e-05, + "loss": 1.0275, + "step": 10803 + }, + { + "epoch": 0.8, + "learning_rate": 1.9448120903662066e-05, + "loss": 1.0202, + "step": 10804 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447990220717894e-05, + "loss": 1.0577, + "step": 10805 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447859522742106e-05, + "loss": 1.0608, + "step": 10806 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447728809734912e-05, + "loss": 0.9463, + "step": 10807 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447598081696512e-05, + "loss": 1.0769, + "step": 10808 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447467338627122e-05, + "loss": 1.0298, + "step": 10809 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447336580526947e-05, + "loss": 1.0643, + "step": 10810 + }, + { + "epoch": 0.8, + "learning_rate": 1.9447205807396192e-05, + "loss": 0.9939, + "step": 10811 + }, + { + "epoch": 0.8, + "learning_rate": 1.944707501923507e-05, + "loss": 1.0951, + "step": 10812 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446944216043785e-05, + "loss": 1.124, + "step": 10813 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446813397822546e-05, + "loss": 1.0386, + "step": 10814 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446682564571565e-05, + "loss": 1.02, + "step": 10815 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446551716291044e-05, + "loss": 1.0199, + "step": 10816 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446420852981198e-05, + "loss": 1.0359, + "step": 10817 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446289974642232e-05, + "loss": 1.0678, + "step": 10818 + }, + { + "epoch": 0.8, + "learning_rate": 1.944615908127435e-05, + "loss": 1.0768, + "step": 10819 + }, + { + "epoch": 0.8, + "learning_rate": 1.9446028172877764e-05, + "loss": 1.043, + "step": 10820 + }, + { + "epoch": 0.8, + "learning_rate": 1.9445897249452685e-05, + "loss": 1.1115, + "step": 10821 + }, + { + "epoch": 0.8, + "learning_rate": 1.944576631099932e-05, + "loss": 1.0153, + "step": 10822 + }, + { + "epoch": 0.8, + "learning_rate": 1.944563535751787e-05, + "loss": 1.0185, + "step": 10823 + }, + { + "epoch": 0.8, + "learning_rate": 1.9445504389008554e-05, + "loss": 1.004, + "step": 10824 + }, + { + "epoch": 0.8, + "learning_rate": 1.9445373405471576e-05, + "loss": 1.0663, + "step": 10825 + }, + { + "epoch": 0.8, + "learning_rate": 1.944524240690714e-05, + "loss": 1.0133, + "step": 10826 + }, + { + "epoch": 0.8, + "learning_rate": 1.9445111393315464e-05, + "loss": 1.0249, + "step": 10827 + }, + { + "epoch": 0.8, + "learning_rate": 1.9444980364696747e-05, + "loss": 1.0064, + "step": 10828 + }, + { + "epoch": 0.8, + "learning_rate": 1.94448493210512e-05, + "loss": 1.0879, + "step": 10829 + }, + { + "epoch": 0.8, + "learning_rate": 1.9444718262379036e-05, + "loss": 0.9679, + "step": 10830 + }, + { + "epoch": 0.8, + "learning_rate": 1.9444587188680463e-05, + "loss": 0.9972, + "step": 10831 + }, + { + "epoch": 0.8, + "learning_rate": 1.944445609995568e-05, + "loss": 0.9704, + "step": 10832 + }, + { + "epoch": 0.8, + "learning_rate": 1.9444324996204907e-05, + "loss": 0.9469, + "step": 10833 + }, + { + "epoch": 0.8, + "learning_rate": 1.944419387742835e-05, + "loss": 1.0311, + "step": 10834 + }, + { + "epoch": 0.8, + "learning_rate": 1.944406274362621e-05, + "loss": 1.0381, + "step": 10835 + }, + { + "epoch": 0.8, + "learning_rate": 1.9443931594798706e-05, + "loss": 1.001, + "step": 10836 + }, + { + "epoch": 0.8, + "learning_rate": 1.944380043094604e-05, + "loss": 1.1357, + "step": 10837 + }, + { + "epoch": 0.8, + "learning_rate": 1.944366925206842e-05, + "loss": 1.0435, + "step": 10838 + }, + { + "epoch": 0.8, + "learning_rate": 1.944353805816606e-05, + "loss": 1.0827, + "step": 10839 + }, + { + "epoch": 0.8, + "learning_rate": 1.9443406849239167e-05, + "loss": 1.0612, + "step": 10840 + }, + { + "epoch": 0.8, + "learning_rate": 1.9443275625287947e-05, + "loss": 1.0191, + "step": 10841 + }, + { + "epoch": 0.8, + "learning_rate": 1.9443144386312612e-05, + "loss": 1.0382, + "step": 10842 + }, + { + "epoch": 0.8, + "learning_rate": 1.944301313231337e-05, + "loss": 1.0709, + "step": 10843 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442881863290428e-05, + "loss": 1.0489, + "step": 10844 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442750579243995e-05, + "loss": 1.0474, + "step": 10845 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442619280174285e-05, + "loss": 1.02, + "step": 10846 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442487966081497e-05, + "loss": 0.9583, + "step": 10847 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442356636965847e-05, + "loss": 1.0054, + "step": 10848 + }, + { + "epoch": 0.8, + "learning_rate": 1.9442225292827547e-05, + "loss": 0.9335, + "step": 10849 + }, + { + "epoch": 0.8, + "learning_rate": 1.94420939336668e-05, + "loss": 1.0406, + "step": 10850 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441962559483816e-05, + "loss": 1.1038, + "step": 10851 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441831170278802e-05, + "loss": 1.0781, + "step": 10852 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441699766051974e-05, + "loss": 0.9843, + "step": 10853 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441568346803535e-05, + "loss": 1.0477, + "step": 10854 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441436912533692e-05, + "loss": 1.0955, + "step": 10855 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441305463242664e-05, + "loss": 1.0182, + "step": 10856 + }, + { + "epoch": 0.8, + "learning_rate": 1.944117399893065e-05, + "loss": 1.0431, + "step": 10857 + }, + { + "epoch": 0.8, + "learning_rate": 1.9441042519597864e-05, + "loss": 1.0556, + "step": 10858 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440911025244515e-05, + "loss": 1.1634, + "step": 10859 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440779515870815e-05, + "loss": 1.0037, + "step": 10860 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440647991476968e-05, + "loss": 1.0857, + "step": 10861 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440516452063183e-05, + "loss": 1.1158, + "step": 10862 + }, + { + "epoch": 0.8, + "learning_rate": 1.944038489762967e-05, + "loss": 0.9539, + "step": 10863 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440253328176643e-05, + "loss": 0.9836, + "step": 10864 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440121743704304e-05, + "loss": 0.9588, + "step": 10865 + }, + { + "epoch": 0.8, + "learning_rate": 1.943999014421287e-05, + "loss": 1.0999, + "step": 10866 + }, + { + "epoch": 0.8, + "learning_rate": 1.943985852970255e-05, + "loss": 1.0742, + "step": 10867 + }, + { + "epoch": 0.8, + "learning_rate": 1.9439726900173545e-05, + "loss": 1.0978, + "step": 10868 + }, + { + "epoch": 0.8, + "learning_rate": 1.943959525562607e-05, + "loss": 1.0758, + "step": 10869 + }, + { + "epoch": 0.8, + "learning_rate": 1.9439463596060335e-05, + "loss": 0.9497, + "step": 10870 + }, + { + "epoch": 0.8, + "learning_rate": 1.943933192147655e-05, + "loss": 0.9906, + "step": 10871 + }, + { + "epoch": 0.8, + "learning_rate": 1.943920023187492e-05, + "loss": 0.9818, + "step": 10872 + }, + { + "epoch": 0.8, + "learning_rate": 1.943906852725566e-05, + "loss": 0.9625, + "step": 10873 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438936807618972e-05, + "loss": 0.9917, + "step": 10874 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438805072965076e-05, + "loss": 1.0459, + "step": 10875 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438673323294174e-05, + "loss": 1.0838, + "step": 10876 + }, + { + "epoch": 0.8, + "learning_rate": 1.943854155860648e-05, + "loss": 1.0034, + "step": 10877 + }, + { + "epoch": 0.8, + "learning_rate": 1.94384097789022e-05, + "loss": 1.0056, + "step": 10878 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438277984181548e-05, + "loss": 1.0111, + "step": 10879 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438146174444723e-05, + "loss": 1.0182, + "step": 10880 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438014349691946e-05, + "loss": 0.987, + "step": 10881 + }, + { + "epoch": 0.8, + "learning_rate": 1.9437882509923426e-05, + "loss": 0.986, + "step": 10882 + }, + { + "epoch": 0.8, + "learning_rate": 1.943775065513937e-05, + "loss": 1.061, + "step": 10883 + }, + { + "epoch": 0.8, + "learning_rate": 1.9437618785339985e-05, + "loss": 1.0795, + "step": 10884 + }, + { + "epoch": 0.8, + "learning_rate": 1.9437486900525483e-05, + "loss": 1.0561, + "step": 10885 + }, + { + "epoch": 0.8, + "learning_rate": 1.943735500069608e-05, + "loss": 1.0262, + "step": 10886 + }, + { + "epoch": 0.8, + "learning_rate": 1.9437223085851973e-05, + "loss": 1.0259, + "step": 10887 + }, + { + "epoch": 0.8, + "learning_rate": 1.9437091155993385e-05, + "loss": 1.0156, + "step": 10888 + }, + { + "epoch": 0.8, + "learning_rate": 1.9436959211120517e-05, + "loss": 1.0049, + "step": 10889 + }, + { + "epoch": 0.8, + "learning_rate": 1.9436827251233578e-05, + "loss": 1.031, + "step": 10890 + }, + { + "epoch": 0.8, + "learning_rate": 1.943669527633279e-05, + "loss": 1.1357, + "step": 10891 + }, + { + "epoch": 0.8, + "learning_rate": 1.9436563286418348e-05, + "loss": 0.9546, + "step": 10892 + }, + { + "epoch": 0.8, + "learning_rate": 1.9436431281490472e-05, + "loss": 0.9778, + "step": 10893 + }, + { + "epoch": 0.8, + "learning_rate": 1.943629926154937e-05, + "loss": 1.0407, + "step": 10894 + }, + { + "epoch": 0.81, + "learning_rate": 1.9436167226595247e-05, + "loss": 1.1806, + "step": 10895 + }, + { + "epoch": 0.81, + "learning_rate": 1.9436035176628322e-05, + "loss": 1.0123, + "step": 10896 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435903111648798e-05, + "loss": 0.9172, + "step": 10897 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435771031656886e-05, + "loss": 1.0828, + "step": 10898 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435638936652795e-05, + "loss": 1.0397, + "step": 10899 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435506826636743e-05, + "loss": 1.0179, + "step": 10900 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435374701608933e-05, + "loss": 0.9645, + "step": 10901 + }, + { + "epoch": 0.81, + "learning_rate": 1.9435242561569574e-05, + "loss": 1.0597, + "step": 10902 + }, + { + "epoch": 0.81, + "learning_rate": 1.943511040651888e-05, + "loss": 1.0504, + "step": 10903 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434978236457063e-05, + "loss": 1.098, + "step": 10904 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434846051384327e-05, + "loss": 1.0758, + "step": 10905 + }, + { + "epoch": 0.81, + "learning_rate": 1.943471385130089e-05, + "loss": 1.0122, + "step": 10906 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434581636206955e-05, + "loss": 1.1015, + "step": 10907 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434449406102735e-05, + "loss": 0.9154, + "step": 10908 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434317160988442e-05, + "loss": 0.9948, + "step": 10909 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434184900864287e-05, + "loss": 0.8947, + "step": 10910 + }, + { + "epoch": 0.81, + "learning_rate": 1.9434052625730478e-05, + "loss": 1.0469, + "step": 10911 + }, + { + "epoch": 0.81, + "learning_rate": 1.9433920335587225e-05, + "loss": 1.0223, + "step": 10912 + }, + { + "epoch": 0.81, + "learning_rate": 1.943378803043474e-05, + "loss": 0.9509, + "step": 10913 + }, + { + "epoch": 0.81, + "learning_rate": 1.9433655710273237e-05, + "loss": 0.8779, + "step": 10914 + }, + { + "epoch": 0.81, + "learning_rate": 1.943352337510292e-05, + "loss": 0.9275, + "step": 10915 + }, + { + "epoch": 0.81, + "learning_rate": 1.9433391024924e-05, + "loss": 1.0699, + "step": 10916 + }, + { + "epoch": 0.81, + "learning_rate": 1.943325865973669e-05, + "loss": 0.9998, + "step": 10917 + }, + { + "epoch": 0.81, + "learning_rate": 1.9433126279541205e-05, + "loss": 0.8926, + "step": 10918 + }, + { + "epoch": 0.81, + "learning_rate": 1.9432993884337748e-05, + "loss": 1.0313, + "step": 10919 + }, + { + "epoch": 0.81, + "learning_rate": 1.9432861474126535e-05, + "loss": 1.0602, + "step": 10920 + }, + { + "epoch": 0.81, + "learning_rate": 1.943272904890777e-05, + "loss": 1.0477, + "step": 10921 + }, + { + "epoch": 0.81, + "learning_rate": 1.943259660868167e-05, + "loss": 0.9644, + "step": 10922 + }, + { + "epoch": 0.81, + "learning_rate": 1.9432464153448444e-05, + "loss": 0.9891, + "step": 10923 + }, + { + "epoch": 0.81, + "learning_rate": 1.9432331683208303e-05, + "loss": 0.995, + "step": 10924 + }, + { + "epoch": 0.81, + "learning_rate": 1.9432199197961457e-05, + "loss": 1.0554, + "step": 10925 + }, + { + "epoch": 0.81, + "learning_rate": 1.943206669770812e-05, + "loss": 1.1272, + "step": 10926 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431934182448494e-05, + "loss": 1.0574, + "step": 10927 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431801652182798e-05, + "loss": 0.9732, + "step": 10928 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431669106911242e-05, + "loss": 1.0157, + "step": 10929 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431536546634034e-05, + "loss": 1.0667, + "step": 10930 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431403971351387e-05, + "loss": 0.918, + "step": 10931 + }, + { + "epoch": 0.81, + "learning_rate": 1.943127138106351e-05, + "loss": 1.0326, + "step": 10932 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431138775770617e-05, + "loss": 0.9475, + "step": 10933 + }, + { + "epoch": 0.81, + "learning_rate": 1.9431006155472915e-05, + "loss": 1.0324, + "step": 10934 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430873520170617e-05, + "loss": 0.985, + "step": 10935 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430740869863934e-05, + "loss": 1.0058, + "step": 10936 + }, + { + "epoch": 0.81, + "learning_rate": 1.943060820455308e-05, + "loss": 0.9902, + "step": 10937 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430475524238264e-05, + "loss": 1.0552, + "step": 10938 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430342828919694e-05, + "loss": 1.0791, + "step": 10939 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430210118597582e-05, + "loss": 0.9876, + "step": 10940 + }, + { + "epoch": 0.81, + "learning_rate": 1.9430077393272145e-05, + "loss": 0.9798, + "step": 10941 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429944652943586e-05, + "loss": 1.0412, + "step": 10942 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429811897612122e-05, + "loss": 1.0815, + "step": 10943 + }, + { + "epoch": 0.81, + "learning_rate": 1.942967912727796e-05, + "loss": 1.0053, + "step": 10944 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429546341941316e-05, + "loss": 0.9194, + "step": 10945 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429413541602397e-05, + "loss": 0.9732, + "step": 10946 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429280726261417e-05, + "loss": 1.0833, + "step": 10947 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429147895918585e-05, + "loss": 1.0278, + "step": 10948 + }, + { + "epoch": 0.81, + "learning_rate": 1.9429015050574116e-05, + "loss": 1.0367, + "step": 10949 + }, + { + "epoch": 0.81, + "learning_rate": 1.9428882190228218e-05, + "loss": 1.1091, + "step": 10950 + }, + { + "epoch": 0.81, + "learning_rate": 1.94287493148811e-05, + "loss": 1.1159, + "step": 10951 + }, + { + "epoch": 0.81, + "learning_rate": 1.942861642453298e-05, + "loss": 1.1121, + "step": 10952 + }, + { + "epoch": 0.81, + "learning_rate": 1.9428483519184067e-05, + "loss": 0.9791, + "step": 10953 + }, + { + "epoch": 0.81, + "learning_rate": 1.942835059883457e-05, + "loss": 0.9851, + "step": 10954 + }, + { + "epoch": 0.81, + "learning_rate": 1.94282176634847e-05, + "loss": 1.0588, + "step": 10955 + }, + { + "epoch": 0.81, + "learning_rate": 1.9428084713134674e-05, + "loss": 0.9961, + "step": 10956 + }, + { + "epoch": 0.81, + "learning_rate": 1.9427951747784698e-05, + "loss": 0.9413, + "step": 10957 + }, + { + "epoch": 0.81, + "learning_rate": 1.942781876743499e-05, + "loss": 1.0561, + "step": 10958 + }, + { + "epoch": 0.81, + "learning_rate": 1.942768577208575e-05, + "loss": 1.0233, + "step": 10959 + }, + { + "epoch": 0.81, + "learning_rate": 1.9427552761737203e-05, + "loss": 1.0015, + "step": 10960 + }, + { + "epoch": 0.81, + "learning_rate": 1.942741973638955e-05, + "loss": 0.915, + "step": 10961 + }, + { + "epoch": 0.81, + "learning_rate": 1.942728669604301e-05, + "loss": 0.9789, + "step": 10962 + }, + { + "epoch": 0.81, + "learning_rate": 1.9427153640697792e-05, + "loss": 1.1249, + "step": 10963 + }, + { + "epoch": 0.81, + "learning_rate": 1.9427020570354104e-05, + "loss": 0.9983, + "step": 10964 + }, + { + "epoch": 0.81, + "learning_rate": 1.9426887485012166e-05, + "loss": 0.9586, + "step": 10965 + }, + { + "epoch": 0.81, + "learning_rate": 1.9426754384672182e-05, + "loss": 1.0012, + "step": 10966 + }, + { + "epoch": 0.81, + "learning_rate": 1.942662126933437e-05, + "loss": 1.1017, + "step": 10967 + }, + { + "epoch": 0.81, + "learning_rate": 1.9426488138998935e-05, + "loss": 1.0176, + "step": 10968 + }, + { + "epoch": 0.81, + "learning_rate": 1.9426354993666092e-05, + "loss": 1.0132, + "step": 10969 + }, + { + "epoch": 0.81, + "learning_rate": 1.9426221833336053e-05, + "loss": 1.0915, + "step": 10970 + }, + { + "epoch": 0.81, + "learning_rate": 1.942608865800903e-05, + "loss": 1.0843, + "step": 10971 + }, + { + "epoch": 0.81, + "learning_rate": 1.9425955467685237e-05, + "loss": 1.1286, + "step": 10972 + }, + { + "epoch": 0.81, + "learning_rate": 1.942582226236488e-05, + "loss": 0.9434, + "step": 10973 + }, + { + "epoch": 0.81, + "learning_rate": 1.942568904204818e-05, + "loss": 1.0297, + "step": 10974 + }, + { + "epoch": 0.81, + "learning_rate": 1.9425555806735337e-05, + "loss": 1.0583, + "step": 10975 + }, + { + "epoch": 0.81, + "learning_rate": 1.9425422556426574e-05, + "loss": 0.9822, + "step": 10976 + }, + { + "epoch": 0.81, + "learning_rate": 1.94252892911221e-05, + "loss": 0.9988, + "step": 10977 + }, + { + "epoch": 0.81, + "learning_rate": 1.942515601082212e-05, + "loss": 0.9214, + "step": 10978 + }, + { + "epoch": 0.81, + "learning_rate": 1.9425022715526855e-05, + "loss": 1.0543, + "step": 10979 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424889405236514e-05, + "loss": 1.1149, + "step": 10980 + }, + { + "epoch": 0.81, + "learning_rate": 1.942475607995131e-05, + "loss": 1.0878, + "step": 10981 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424622739671454e-05, + "loss": 0.9278, + "step": 10982 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424489384397155e-05, + "loss": 0.9972, + "step": 10983 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424356014128628e-05, + "loss": 1.0096, + "step": 10984 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424222628866086e-05, + "loss": 1.0533, + "step": 10985 + }, + { + "epoch": 0.81, + "learning_rate": 1.9424089228609743e-05, + "loss": 1.0073, + "step": 10986 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423955813359806e-05, + "loss": 1.1635, + "step": 10987 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423822383116494e-05, + "loss": 1.0616, + "step": 10988 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423688937880015e-05, + "loss": 1.2087, + "step": 10989 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423555477650578e-05, + "loss": 1.0281, + "step": 10990 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423422002428404e-05, + "loss": 0.9647, + "step": 10991 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423288512213693e-05, + "loss": 1.0266, + "step": 10992 + }, + { + "epoch": 0.81, + "learning_rate": 1.942315500700667e-05, + "loss": 1.0113, + "step": 10993 + }, + { + "epoch": 0.81, + "learning_rate": 1.9423021486807542e-05, + "loss": 0.9922, + "step": 10994 + }, + { + "epoch": 0.81, + "learning_rate": 1.9422887951616523e-05, + "loss": 0.982, + "step": 10995 + }, + { + "epoch": 0.81, + "learning_rate": 1.942275440143382e-05, + "loss": 0.998, + "step": 10996 + }, + { + "epoch": 0.81, + "learning_rate": 1.9422620836259655e-05, + "loss": 1.0271, + "step": 10997 + }, + { + "epoch": 0.81, + "learning_rate": 1.942248725609423e-05, + "loss": 0.9704, + "step": 10998 + }, + { + "epoch": 0.81, + "learning_rate": 1.9422353660937763e-05, + "loss": 0.937, + "step": 10999 + }, + { + "epoch": 0.81, + "learning_rate": 1.9422220050790467e-05, + "loss": 0.9572, + "step": 11000 + }, + { + "epoch": 0.81, + "learning_rate": 1.9422086425652552e-05, + "loss": 1.0958, + "step": 11001 + }, + { + "epoch": 0.81, + "learning_rate": 1.9421952785524237e-05, + "loss": 0.8939, + "step": 11002 + }, + { + "epoch": 0.81, + "learning_rate": 1.942181913040573e-05, + "loss": 0.916, + "step": 11003 + }, + { + "epoch": 0.81, + "learning_rate": 1.942168546029724e-05, + "loss": 1.0534, + "step": 11004 + }, + { + "epoch": 0.81, + "learning_rate": 1.942155177519898e-05, + "loss": 1.1752, + "step": 11005 + }, + { + "epoch": 0.81, + "learning_rate": 1.942141807511117e-05, + "loss": 0.9799, + "step": 11006 + }, + { + "epoch": 0.81, + "learning_rate": 1.9421284360034024e-05, + "loss": 1.0145, + "step": 11007 + }, + { + "epoch": 0.81, + "learning_rate": 1.9421150629967744e-05, + "loss": 1.0377, + "step": 11008 + }, + { + "epoch": 0.81, + "learning_rate": 1.9421016884912545e-05, + "loss": 1.0597, + "step": 11009 + }, + { + "epoch": 0.81, + "learning_rate": 1.9420883124868644e-05, + "loss": 0.9834, + "step": 11010 + }, + { + "epoch": 0.81, + "learning_rate": 1.9420749349836255e-05, + "loss": 0.9877, + "step": 11011 + }, + { + "epoch": 0.81, + "learning_rate": 1.942061555981559e-05, + "loss": 1.0462, + "step": 11012 + }, + { + "epoch": 0.81, + "learning_rate": 1.9420481754806858e-05, + "loss": 1.0337, + "step": 11013 + }, + { + "epoch": 0.81, + "learning_rate": 1.9420347934810277e-05, + "loss": 0.9969, + "step": 11014 + }, + { + "epoch": 0.81, + "learning_rate": 1.9420214099826057e-05, + "loss": 1.0203, + "step": 11015 + }, + { + "epoch": 0.81, + "learning_rate": 1.942008024985441e-05, + "loss": 1.0415, + "step": 11016 + }, + { + "epoch": 0.81, + "learning_rate": 1.941994638489555e-05, + "loss": 1.0495, + "step": 11017 + }, + { + "epoch": 0.81, + "learning_rate": 1.941981250494969e-05, + "loss": 0.911, + "step": 11018 + }, + { + "epoch": 0.81, + "learning_rate": 1.9419678610017045e-05, + "loss": 1.1048, + "step": 11019 + }, + { + "epoch": 0.81, + "learning_rate": 1.941954470009782e-05, + "loss": 0.9217, + "step": 11020 + }, + { + "epoch": 0.81, + "learning_rate": 1.941941077519224e-05, + "loss": 1.0399, + "step": 11021 + }, + { + "epoch": 0.81, + "learning_rate": 1.9419276835300507e-05, + "loss": 1.1131, + "step": 11022 + }, + { + "epoch": 0.81, + "learning_rate": 1.9419142880422847e-05, + "loss": 1.0974, + "step": 11023 + }, + { + "epoch": 0.81, + "learning_rate": 1.941900891055946e-05, + "loss": 1.0544, + "step": 11024 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418874925710565e-05, + "loss": 0.9874, + "step": 11025 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418740925876377e-05, + "loss": 1.0687, + "step": 11026 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418606911057102e-05, + "loss": 0.9526, + "step": 11027 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418472881252964e-05, + "loss": 1.045, + "step": 11028 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418338836464166e-05, + "loss": 1.0334, + "step": 11029 + }, + { + "epoch": 0.81, + "learning_rate": 1.9418204776690928e-05, + "loss": 1.1865, + "step": 11030 + }, + { + "epoch": 0.82, + "learning_rate": 1.9418070701933463e-05, + "loss": 0.888, + "step": 11031 + }, + { + "epoch": 0.82, + "learning_rate": 1.941793661219198e-05, + "loss": 1.1192, + "step": 11032 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417802507466693e-05, + "loss": 0.906, + "step": 11033 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417668387757817e-05, + "loss": 1.0974, + "step": 11034 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417534253065565e-05, + "loss": 0.9566, + "step": 11035 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417400103390154e-05, + "loss": 0.9536, + "step": 11036 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417265938731794e-05, + "loss": 0.9596, + "step": 11037 + }, + { + "epoch": 0.82, + "learning_rate": 1.9417131759090696e-05, + "loss": 0.9755, + "step": 11038 + }, + { + "epoch": 0.82, + "learning_rate": 1.941699756446708e-05, + "loss": 1.105, + "step": 11039 + }, + { + "epoch": 0.82, + "learning_rate": 1.9416863354861152e-05, + "loss": 1.1679, + "step": 11040 + }, + { + "epoch": 0.82, + "learning_rate": 1.941672913027313e-05, + "loss": 0.999, + "step": 11041 + }, + { + "epoch": 0.82, + "learning_rate": 1.9416594890703225e-05, + "loss": 0.9871, + "step": 11042 + }, + { + "epoch": 0.82, + "learning_rate": 1.9416460636151657e-05, + "loss": 1.1084, + "step": 11043 + }, + { + "epoch": 0.82, + "learning_rate": 1.941632636661863e-05, + "loss": 0.9073, + "step": 11044 + }, + { + "epoch": 0.82, + "learning_rate": 1.9416192082104368e-05, + "loss": 1.0503, + "step": 11045 + }, + { + "epoch": 0.82, + "learning_rate": 1.9416057782609076e-05, + "loss": 1.0432, + "step": 11046 + }, + { + "epoch": 0.82, + "learning_rate": 1.941592346813297e-05, + "loss": 0.8987, + "step": 11047 + }, + { + "epoch": 0.82, + "learning_rate": 1.9415789138676263e-05, + "loss": 0.9865, + "step": 11048 + }, + { + "epoch": 0.82, + "learning_rate": 1.941565479423917e-05, + "loss": 1.1431, + "step": 11049 + }, + { + "epoch": 0.82, + "learning_rate": 1.941552043482191e-05, + "loss": 1.0329, + "step": 11050 + }, + { + "epoch": 0.82, + "learning_rate": 1.9415386060424687e-05, + "loss": 1.1303, + "step": 11051 + }, + { + "epoch": 0.82, + "learning_rate": 1.941525167104772e-05, + "loss": 1.0177, + "step": 11052 + }, + { + "epoch": 0.82, + "learning_rate": 1.9415117266691227e-05, + "loss": 1.1919, + "step": 11053 + }, + { + "epoch": 0.82, + "learning_rate": 1.9414982847355413e-05, + "loss": 1.0413, + "step": 11054 + }, + { + "epoch": 0.82, + "learning_rate": 1.9414848413040496e-05, + "loss": 1.0001, + "step": 11055 + }, + { + "epoch": 0.82, + "learning_rate": 1.941471396374669e-05, + "loss": 0.9517, + "step": 11056 + }, + { + "epoch": 0.82, + "learning_rate": 1.941457949947421e-05, + "loss": 1.074, + "step": 11057 + }, + { + "epoch": 0.82, + "learning_rate": 1.9414445020223265e-05, + "loss": 0.9856, + "step": 11058 + }, + { + "epoch": 0.82, + "learning_rate": 1.941431052599408e-05, + "loss": 1.0454, + "step": 11059 + }, + { + "epoch": 0.82, + "learning_rate": 1.9414176016786854e-05, + "loss": 1.096, + "step": 11060 + }, + { + "epoch": 0.82, + "learning_rate": 1.9414041492601812e-05, + "loss": 1.0352, + "step": 11061 + }, + { + "epoch": 0.82, + "learning_rate": 1.9413906953439164e-05, + "loss": 1.1004, + "step": 11062 + }, + { + "epoch": 0.82, + "learning_rate": 1.9413772399299128e-05, + "loss": 0.9448, + "step": 11063 + }, + { + "epoch": 0.82, + "learning_rate": 1.941363783018191e-05, + "loss": 1.1817, + "step": 11064 + }, + { + "epoch": 0.82, + "learning_rate": 1.941350324608773e-05, + "loss": 0.9649, + "step": 11065 + }, + { + "epoch": 0.82, + "learning_rate": 1.9413368647016805e-05, + "loss": 1.0232, + "step": 11066 + }, + { + "epoch": 0.82, + "learning_rate": 1.941323403296934e-05, + "loss": 1.1142, + "step": 11067 + }, + { + "epoch": 0.82, + "learning_rate": 1.9413099403945557e-05, + "loss": 1.062, + "step": 11068 + }, + { + "epoch": 0.82, + "learning_rate": 1.9412964759945668e-05, + "loss": 1.0898, + "step": 11069 + }, + { + "epoch": 0.82, + "learning_rate": 1.9412830100969885e-05, + "loss": 0.9834, + "step": 11070 + }, + { + "epoch": 0.82, + "learning_rate": 1.9412695427018424e-05, + "loss": 0.9976, + "step": 11071 + }, + { + "epoch": 0.82, + "learning_rate": 1.94125607380915e-05, + "loss": 1.0743, + "step": 11072 + }, + { + "epoch": 0.82, + "learning_rate": 1.9412426034189328e-05, + "loss": 1.0683, + "step": 11073 + }, + { + "epoch": 0.82, + "learning_rate": 1.941229131531212e-05, + "loss": 0.9495, + "step": 11074 + }, + { + "epoch": 0.82, + "learning_rate": 1.941215658146009e-05, + "loss": 0.956, + "step": 11075 + }, + { + "epoch": 0.82, + "learning_rate": 1.9412021832633456e-05, + "loss": 1.1327, + "step": 11076 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411887068832426e-05, + "loss": 0.988, + "step": 11077 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411752290057222e-05, + "loss": 0.9753, + "step": 11078 + }, + { + "epoch": 0.82, + "learning_rate": 1.941161749630805e-05, + "loss": 1.1219, + "step": 11079 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411482687585134e-05, + "loss": 1.1664, + "step": 11080 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411347863888683e-05, + "loss": 1.09, + "step": 11081 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411213025218912e-05, + "loss": 0.9385, + "step": 11082 + }, + { + "epoch": 0.82, + "learning_rate": 1.9411078171576035e-05, + "loss": 1.0521, + "step": 11083 + }, + { + "epoch": 0.82, + "learning_rate": 1.9410943302960268e-05, + "loss": 1.0515, + "step": 11084 + }, + { + "epoch": 0.82, + "learning_rate": 1.9410808419371822e-05, + "loss": 0.963, + "step": 11085 + }, + { + "epoch": 0.82, + "learning_rate": 1.941067352081092e-05, + "loss": 0.935, + "step": 11086 + }, + { + "epoch": 0.82, + "learning_rate": 1.9410538607277765e-05, + "loss": 1.0408, + "step": 11087 + }, + { + "epoch": 0.82, + "learning_rate": 1.941040367877258e-05, + "loss": 1.0141, + "step": 11088 + }, + { + "epoch": 0.82, + "learning_rate": 1.9410268735295578e-05, + "loss": 1.0105, + "step": 11089 + }, + { + "epoch": 0.82, + "learning_rate": 1.9410133776846973e-05, + "loss": 1.0469, + "step": 11090 + }, + { + "epoch": 0.82, + "learning_rate": 1.9409998803426983e-05, + "loss": 1.0044, + "step": 11091 + }, + { + "epoch": 0.82, + "learning_rate": 1.9409863815035814e-05, + "loss": 0.96, + "step": 11092 + }, + { + "epoch": 0.82, + "learning_rate": 1.940972881167369e-05, + "loss": 1.0493, + "step": 11093 + }, + { + "epoch": 0.82, + "learning_rate": 1.940959379334082e-05, + "loss": 1.0165, + "step": 11094 + }, + { + "epoch": 0.82, + "learning_rate": 1.940945876003742e-05, + "loss": 1.1199, + "step": 11095 + }, + { + "epoch": 0.82, + "learning_rate": 1.9409323711763707e-05, + "loss": 1.0507, + "step": 11096 + }, + { + "epoch": 0.82, + "learning_rate": 1.9409188648519898e-05, + "loss": 1.1246, + "step": 11097 + }, + { + "epoch": 0.82, + "learning_rate": 1.94090535703062e-05, + "loss": 1.0475, + "step": 11098 + }, + { + "epoch": 0.82, + "learning_rate": 1.9408918477122833e-05, + "loss": 1.0719, + "step": 11099 + }, + { + "epoch": 0.82, + "learning_rate": 1.940878336897001e-05, + "loss": 0.9786, + "step": 11100 + }, + { + "epoch": 0.82, + "learning_rate": 1.9408648245847952e-05, + "loss": 1.1183, + "step": 11101 + }, + { + "epoch": 0.82, + "learning_rate": 1.9408513107756866e-05, + "loss": 1.0576, + "step": 11102 + }, + { + "epoch": 0.82, + "learning_rate": 1.940837795469697e-05, + "loss": 1.0097, + "step": 11103 + }, + { + "epoch": 0.82, + "learning_rate": 1.940824278666848e-05, + "loss": 0.8837, + "step": 11104 + }, + { + "epoch": 0.82, + "learning_rate": 1.9408107603671614e-05, + "loss": 1.0396, + "step": 11105 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407972405706577e-05, + "loss": 0.9767, + "step": 11106 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407837192773594e-05, + "loss": 1.0646, + "step": 11107 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407701964872876e-05, + "loss": 0.9426, + "step": 11108 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407566722004637e-05, + "loss": 1.0432, + "step": 11109 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407431464169096e-05, + "loss": 0.9848, + "step": 11110 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407296191366466e-05, + "loss": 1.019, + "step": 11111 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407160903596964e-05, + "loss": 1.0155, + "step": 11112 + }, + { + "epoch": 0.82, + "learning_rate": 1.9407025600860803e-05, + "loss": 1.1258, + "step": 11113 + }, + { + "epoch": 0.82, + "learning_rate": 1.9406890283158196e-05, + "loss": 1.021, + "step": 11114 + }, + { + "epoch": 0.82, + "learning_rate": 1.9406754950489365e-05, + "loss": 0.9981, + "step": 11115 + }, + { + "epoch": 0.82, + "learning_rate": 1.9406619602854518e-05, + "loss": 0.9749, + "step": 11116 + }, + { + "epoch": 0.82, + "learning_rate": 1.9406484240253876e-05, + "loss": 1.0323, + "step": 11117 + }, + { + "epoch": 0.82, + "learning_rate": 1.940634886268765e-05, + "loss": 1.084, + "step": 11118 + }, + { + "epoch": 0.82, + "learning_rate": 1.940621347015606e-05, + "loss": 1.0961, + "step": 11119 + }, + { + "epoch": 0.82, + "learning_rate": 1.9406078062659317e-05, + "loss": 1.0183, + "step": 11120 + }, + { + "epoch": 0.82, + "learning_rate": 1.940594264019764e-05, + "loss": 1.0738, + "step": 11121 + }, + { + "epoch": 0.82, + "learning_rate": 1.940580720277124e-05, + "loss": 1.009, + "step": 11122 + }, + { + "epoch": 0.82, + "learning_rate": 1.940567175038034e-05, + "loss": 1.0053, + "step": 11123 + }, + { + "epoch": 0.82, + "learning_rate": 1.9405536283025147e-05, + "loss": 1.1122, + "step": 11124 + }, + { + "epoch": 0.82, + "learning_rate": 1.9405400800705882e-05, + "loss": 1.0378, + "step": 11125 + }, + { + "epoch": 0.82, + "learning_rate": 1.9405265303422756e-05, + "loss": 1.0297, + "step": 11126 + }, + { + "epoch": 0.82, + "learning_rate": 1.940512979117599e-05, + "loss": 0.8981, + "step": 11127 + }, + { + "epoch": 0.82, + "learning_rate": 1.94049942639658e-05, + "loss": 0.963, + "step": 11128 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404858721792393e-05, + "loss": 0.9985, + "step": 11129 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404723164655995e-05, + "loss": 1.0703, + "step": 11130 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404587592556814e-05, + "loss": 1.0134, + "step": 11131 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404452005495068e-05, + "loss": 1.0698, + "step": 11132 + }, + { + "epoch": 0.82, + "learning_rate": 1.940431640347098e-05, + "loss": 1.0462, + "step": 11133 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404180786484754e-05, + "loss": 0.9418, + "step": 11134 + }, + { + "epoch": 0.82, + "learning_rate": 1.9404045154536607e-05, + "loss": 1.0188, + "step": 11135 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403909507626764e-05, + "loss": 1.0387, + "step": 11136 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403773845755433e-05, + "loss": 1.09, + "step": 11137 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403638168922834e-05, + "loss": 1.0795, + "step": 11138 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403502477129184e-05, + "loss": 1.0884, + "step": 11139 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403366770374692e-05, + "loss": 1.0667, + "step": 11140 + }, + { + "epoch": 0.82, + "learning_rate": 1.940323104865958e-05, + "loss": 0.9436, + "step": 11141 + }, + { + "epoch": 0.82, + "learning_rate": 1.9403095311984058e-05, + "loss": 1.1039, + "step": 11142 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402959560348353e-05, + "loss": 1.0452, + "step": 11143 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402823793752667e-05, + "loss": 1.0048, + "step": 11144 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402688012197227e-05, + "loss": 1.0561, + "step": 11145 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402552215682244e-05, + "loss": 1.1809, + "step": 11146 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402416404207934e-05, + "loss": 1.0829, + "step": 11147 + }, + { + "epoch": 0.82, + "learning_rate": 1.9402280577774513e-05, + "loss": 0.9842, + "step": 11148 + }, + { + "epoch": 0.82, + "learning_rate": 1.94021447363822e-05, + "loss": 1.0493, + "step": 11149 + }, + { + "epoch": 0.82, + "learning_rate": 1.940200888003121e-05, + "loss": 1.0722, + "step": 11150 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401873008721757e-05, + "loss": 1.0324, + "step": 11151 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401737122454058e-05, + "loss": 0.9874, + "step": 11152 + }, + { + "epoch": 0.82, + "learning_rate": 1.940160122122833e-05, + "loss": 1.0707, + "step": 11153 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401465305044792e-05, + "loss": 0.9314, + "step": 11154 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401329373903653e-05, + "loss": 0.9833, + "step": 11155 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401193427805138e-05, + "loss": 1.0024, + "step": 11156 + }, + { + "epoch": 0.82, + "learning_rate": 1.9401057466749454e-05, + "loss": 0.9943, + "step": 11157 + }, + { + "epoch": 0.82, + "learning_rate": 1.9400921490736825e-05, + "loss": 1.0507, + "step": 11158 + }, + { + "epoch": 0.82, + "learning_rate": 1.940078549976746e-05, + "loss": 1.0428, + "step": 11159 + }, + { + "epoch": 0.82, + "learning_rate": 1.9400649493841587e-05, + "loss": 0.9858, + "step": 11160 + }, + { + "epoch": 0.82, + "learning_rate": 1.940051347295941e-05, + "loss": 1.0265, + "step": 11161 + }, + { + "epoch": 0.82, + "learning_rate": 1.9400377437121153e-05, + "loss": 1.0418, + "step": 11162 + }, + { + "epoch": 0.82, + "learning_rate": 1.940024138632703e-05, + "loss": 1.0239, + "step": 11163 + }, + { + "epoch": 0.82, + "learning_rate": 1.9400105320577255e-05, + "loss": 1.0022, + "step": 11164 + }, + { + "epoch": 0.82, + "learning_rate": 1.939996923987205e-05, + "loss": 1.1459, + "step": 11165 + }, + { + "epoch": 0.83, + "learning_rate": 1.9399833144211625e-05, + "loss": 1.1401, + "step": 11166 + }, + { + "epoch": 0.83, + "learning_rate": 1.93996970335962e-05, + "loss": 1.1959, + "step": 11167 + }, + { + "epoch": 0.83, + "learning_rate": 1.9399560908025994e-05, + "loss": 1.1716, + "step": 11168 + }, + { + "epoch": 0.83, + "learning_rate": 1.939942476750122e-05, + "loss": 0.9562, + "step": 11169 + }, + { + "epoch": 0.83, + "learning_rate": 1.9399288612022097e-05, + "loss": 0.9126, + "step": 11170 + }, + { + "epoch": 0.83, + "learning_rate": 1.939915244158884e-05, + "loss": 1.0053, + "step": 11171 + }, + { + "epoch": 0.83, + "learning_rate": 1.9399016256201665e-05, + "loss": 0.9001, + "step": 11172 + }, + { + "epoch": 0.83, + "learning_rate": 1.9398880055860793e-05, + "loss": 1.1036, + "step": 11173 + }, + { + "epoch": 0.83, + "learning_rate": 1.9398743840566432e-05, + "loss": 1.0001, + "step": 11174 + }, + { + "epoch": 0.83, + "learning_rate": 1.9398607610318806e-05, + "loss": 0.9837, + "step": 11175 + }, + { + "epoch": 0.83, + "learning_rate": 1.939847136511813e-05, + "loss": 1.071, + "step": 11176 + }, + { + "epoch": 0.83, + "learning_rate": 1.939833510496462e-05, + "loss": 0.9107, + "step": 11177 + }, + { + "epoch": 0.83, + "learning_rate": 1.9398198829858495e-05, + "loss": 1.0287, + "step": 11178 + }, + { + "epoch": 0.83, + "learning_rate": 1.9398062539799968e-05, + "loss": 1.071, + "step": 11179 + }, + { + "epoch": 0.83, + "learning_rate": 1.939792623478926e-05, + "loss": 1.0197, + "step": 11180 + }, + { + "epoch": 0.83, + "learning_rate": 1.9397789914826584e-05, + "loss": 1.0278, + "step": 11181 + }, + { + "epoch": 0.83, + "learning_rate": 1.939765357991216e-05, + "loss": 1.054, + "step": 11182 + }, + { + "epoch": 0.83, + "learning_rate": 1.93975172300462e-05, + "loss": 0.9799, + "step": 11183 + }, + { + "epoch": 0.83, + "learning_rate": 1.9397380865228933e-05, + "loss": 0.8569, + "step": 11184 + }, + { + "epoch": 0.83, + "learning_rate": 1.939724448546056e-05, + "loss": 1.0572, + "step": 11185 + }, + { + "epoch": 0.83, + "learning_rate": 1.9397108090741308e-05, + "loss": 1.019, + "step": 11186 + }, + { + "epoch": 0.83, + "learning_rate": 1.939697168107139e-05, + "loss": 0.9411, + "step": 11187 + }, + { + "epoch": 0.83, + "learning_rate": 1.9396835256451025e-05, + "loss": 1.0637, + "step": 11188 + }, + { + "epoch": 0.83, + "learning_rate": 1.939669881688043e-05, + "loss": 1.0232, + "step": 11189 + }, + { + "epoch": 0.83, + "learning_rate": 1.9396562362359823e-05, + "loss": 1.217, + "step": 11190 + }, + { + "epoch": 0.83, + "learning_rate": 1.9396425892889418e-05, + "loss": 1.0698, + "step": 11191 + }, + { + "epoch": 0.83, + "learning_rate": 1.9396289408469434e-05, + "loss": 1.0439, + "step": 11192 + }, + { + "epoch": 0.83, + "learning_rate": 1.939615290910009e-05, + "loss": 0.9755, + "step": 11193 + }, + { + "epoch": 0.83, + "learning_rate": 1.9396016394781598e-05, + "loss": 1.0066, + "step": 11194 + }, + { + "epoch": 0.83, + "learning_rate": 1.939587986551418e-05, + "loss": 1.1567, + "step": 11195 + }, + { + "epoch": 0.83, + "learning_rate": 1.939574332129805e-05, + "loss": 1.0098, + "step": 11196 + }, + { + "epoch": 0.83, + "learning_rate": 1.939560676213343e-05, + "loss": 1.0476, + "step": 11197 + }, + { + "epoch": 0.83, + "learning_rate": 1.939547018802053e-05, + "loss": 0.921, + "step": 11198 + }, + { + "epoch": 0.83, + "learning_rate": 1.9395333598959576e-05, + "loss": 0.9978, + "step": 11199 + }, + { + "epoch": 0.83, + "learning_rate": 1.9395196994950778e-05, + "loss": 1.0226, + "step": 11200 + }, + { + "epoch": 0.83, + "learning_rate": 1.9395060375994358e-05, + "loss": 1.0208, + "step": 11201 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394923742090528e-05, + "loss": 1.05, + "step": 11202 + }, + { + "epoch": 0.83, + "learning_rate": 1.939478709323951e-05, + "loss": 1.1484, + "step": 11203 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394650429441522e-05, + "loss": 0.9832, + "step": 11204 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394513750696777e-05, + "loss": 0.9523, + "step": 11205 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394377057005496e-05, + "loss": 1.0159, + "step": 11206 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394240348367898e-05, + "loss": 1.0429, + "step": 11207 + }, + { + "epoch": 0.83, + "learning_rate": 1.9394103624784197e-05, + "loss": 1.0608, + "step": 11208 + }, + { + "epoch": 0.83, + "learning_rate": 1.9393966886254606e-05, + "loss": 1.0771, + "step": 11209 + }, + { + "epoch": 0.83, + "learning_rate": 1.9393830132779353e-05, + "loss": 1.0936, + "step": 11210 + }, + { + "epoch": 0.83, + "learning_rate": 1.939369336435865e-05, + "loss": 0.8964, + "step": 11211 + }, + { + "epoch": 0.83, + "learning_rate": 1.9393556580992716e-05, + "loss": 0.9321, + "step": 11212 + }, + { + "epoch": 0.83, + "learning_rate": 1.9393419782681765e-05, + "loss": 0.993, + "step": 11213 + }, + { + "epoch": 0.83, + "learning_rate": 1.939328296942602e-05, + "loss": 0.9744, + "step": 11214 + }, + { + "epoch": 0.83, + "learning_rate": 1.93931461412257e-05, + "loss": 1.0249, + "step": 11215 + }, + { + "epoch": 0.83, + "learning_rate": 1.939300929808101e-05, + "loss": 1.0906, + "step": 11216 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392872439992185e-05, + "loss": 1.0069, + "step": 11217 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392735566959427e-05, + "loss": 1.1384, + "step": 11218 + }, + { + "epoch": 0.83, + "learning_rate": 1.939259867898297e-05, + "loss": 0.9522, + "step": 11219 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392461776063016e-05, + "loss": 0.9621, + "step": 11220 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392324858199792e-05, + "loss": 1.0442, + "step": 11221 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392187925393512e-05, + "loss": 1.0532, + "step": 11222 + }, + { + "epoch": 0.83, + "learning_rate": 1.9392050977644398e-05, + "loss": 1.071, + "step": 11223 + }, + { + "epoch": 0.83, + "learning_rate": 1.939191401495266e-05, + "loss": 1.0377, + "step": 11224 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391777037318523e-05, + "loss": 1.0421, + "step": 11225 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391640044742207e-05, + "loss": 1.0233, + "step": 11226 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391503037223924e-05, + "loss": 1.0144, + "step": 11227 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391366014763895e-05, + "loss": 1.031, + "step": 11228 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391228977362334e-05, + "loss": 1.0344, + "step": 11229 + }, + { + "epoch": 0.83, + "learning_rate": 1.9391091925019464e-05, + "loss": 1.0624, + "step": 11230 + }, + { + "epoch": 0.83, + "learning_rate": 1.9390954857735502e-05, + "loss": 0.9605, + "step": 11231 + }, + { + "epoch": 0.83, + "learning_rate": 1.9390817775510665e-05, + "loss": 0.9557, + "step": 11232 + }, + { + "epoch": 0.83, + "learning_rate": 1.939068067834517e-05, + "loss": 1.018, + "step": 11233 + }, + { + "epoch": 0.83, + "learning_rate": 1.9390543566239236e-05, + "loss": 1.1265, + "step": 11234 + }, + { + "epoch": 0.83, + "learning_rate": 1.9390406439193085e-05, + "loss": 0.9159, + "step": 11235 + }, + { + "epoch": 0.83, + "learning_rate": 1.939026929720693e-05, + "loss": 0.9136, + "step": 11236 + }, + { + "epoch": 0.83, + "learning_rate": 1.939013214028099e-05, + "loss": 1.0453, + "step": 11237 + }, + { + "epoch": 0.83, + "learning_rate": 1.9389994968415485e-05, + "loss": 1.0089, + "step": 11238 + }, + { + "epoch": 0.83, + "learning_rate": 1.938985778161063e-05, + "loss": 0.989, + "step": 11239 + }, + { + "epoch": 0.83, + "learning_rate": 1.9389720579866646e-05, + "loss": 1.0674, + "step": 11240 + }, + { + "epoch": 0.83, + "learning_rate": 1.938958336318375e-05, + "loss": 1.0673, + "step": 11241 + }, + { + "epoch": 0.83, + "learning_rate": 1.9389446131562167e-05, + "loss": 1.0424, + "step": 11242 + }, + { + "epoch": 0.83, + "learning_rate": 1.9389308885002107e-05, + "loss": 1.071, + "step": 11243 + }, + { + "epoch": 0.83, + "learning_rate": 1.938917162350379e-05, + "loss": 1.0241, + "step": 11244 + }, + { + "epoch": 0.83, + "learning_rate": 1.9389034347067434e-05, + "loss": 0.9259, + "step": 11245 + }, + { + "epoch": 0.83, + "learning_rate": 1.938889705569326e-05, + "loss": 1.2029, + "step": 11246 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388759749381484e-05, + "loss": 1.0289, + "step": 11247 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388622428132325e-05, + "loss": 1.0834, + "step": 11248 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388485091946004e-05, + "loss": 1.0583, + "step": 11249 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388347740822738e-05, + "loss": 1.1489, + "step": 11250 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388210374762747e-05, + "loss": 1.0383, + "step": 11251 + }, + { + "epoch": 0.83, + "learning_rate": 1.9388072993766242e-05, + "loss": 1.0276, + "step": 11252 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387935597833452e-05, + "loss": 1.068, + "step": 11253 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387798186964585e-05, + "loss": 1.064, + "step": 11254 + }, + { + "epoch": 0.83, + "learning_rate": 1.938766076115987e-05, + "loss": 1.0039, + "step": 11255 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387523320419523e-05, + "loss": 1.0897, + "step": 11256 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387385864743758e-05, + "loss": 1.0151, + "step": 11257 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387248394132797e-05, + "loss": 0.9613, + "step": 11258 + }, + { + "epoch": 0.83, + "learning_rate": 1.9387110908586855e-05, + "loss": 1.1591, + "step": 11259 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386973408106158e-05, + "loss": 1.0298, + "step": 11260 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386835892690918e-05, + "loss": 1.0022, + "step": 11261 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386698362341355e-05, + "loss": 0.9213, + "step": 11262 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386560817057694e-05, + "loss": 1.0663, + "step": 11263 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386423256840147e-05, + "loss": 0.898, + "step": 11264 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386285681688935e-05, + "loss": 0.9711, + "step": 11265 + }, + { + "epoch": 0.83, + "learning_rate": 1.9386148091604274e-05, + "loss": 0.9941, + "step": 11266 + }, + { + "epoch": 0.83, + "learning_rate": 1.938601048658639e-05, + "loss": 1.0218, + "step": 11267 + }, + { + "epoch": 0.83, + "learning_rate": 1.9385872866635492e-05, + "loss": 0.9949, + "step": 11268 + }, + { + "epoch": 0.83, + "learning_rate": 1.938573523175181e-05, + "loss": 1.1025, + "step": 11269 + }, + { + "epoch": 0.83, + "learning_rate": 1.9385597581935553e-05, + "loss": 1.005, + "step": 11270 + }, + { + "epoch": 0.83, + "learning_rate": 1.9385459917186947e-05, + "loss": 1.0748, + "step": 11271 + }, + { + "epoch": 0.83, + "learning_rate": 1.938532223750621e-05, + "loss": 1.0154, + "step": 11272 + }, + { + "epoch": 0.83, + "learning_rate": 1.938518454289356e-05, + "loss": 1.0364, + "step": 11273 + }, + { + "epoch": 0.83, + "learning_rate": 1.9385046833349212e-05, + "loss": 1.0546, + "step": 11274 + }, + { + "epoch": 0.83, + "learning_rate": 1.9384909108873393e-05, + "loss": 1.0026, + "step": 11275 + }, + { + "epoch": 0.83, + "learning_rate": 1.9384771369466314e-05, + "loss": 1.1428, + "step": 11276 + }, + { + "epoch": 0.83, + "learning_rate": 1.93846336151282e-05, + "loss": 1.0629, + "step": 11277 + }, + { + "epoch": 0.83, + "learning_rate": 1.9384495845859268e-05, + "loss": 0.944, + "step": 11278 + }, + { + "epoch": 0.83, + "learning_rate": 1.938435806165974e-05, + "loss": 1.1072, + "step": 11279 + }, + { + "epoch": 0.83, + "learning_rate": 1.9384220262529827e-05, + "loss": 1.0362, + "step": 11280 + }, + { + "epoch": 0.83, + "learning_rate": 1.9384082448469756e-05, + "loss": 1.019, + "step": 11281 + }, + { + "epoch": 0.83, + "learning_rate": 1.9383944619479744e-05, + "loss": 1.1214, + "step": 11282 + }, + { + "epoch": 0.83, + "learning_rate": 1.938380677556001e-05, + "loss": 0.9667, + "step": 11283 + }, + { + "epoch": 0.83, + "learning_rate": 1.938366891671078e-05, + "loss": 0.9878, + "step": 11284 + }, + { + "epoch": 0.83, + "learning_rate": 1.938353104293226e-05, + "loss": 0.9845, + "step": 11285 + }, + { + "epoch": 0.83, + "learning_rate": 1.9383393154224678e-05, + "loss": 1.1, + "step": 11286 + }, + { + "epoch": 0.83, + "learning_rate": 1.938325525058825e-05, + "loss": 0.9848, + "step": 11287 + }, + { + "epoch": 0.83, + "learning_rate": 1.93831173320232e-05, + "loss": 1.118, + "step": 11288 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382979398529743e-05, + "loss": 1.0106, + "step": 11289 + }, + { + "epoch": 0.83, + "learning_rate": 1.93828414501081e-05, + "loss": 0.9679, + "step": 11290 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382703486758493e-05, + "loss": 1.0149, + "step": 11291 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382565508481136e-05, + "loss": 0.9786, + "step": 11292 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382427515276254e-05, + "loss": 1.0621, + "step": 11293 + }, + { + "epoch": 0.83, + "learning_rate": 1.938228950714406e-05, + "loss": 1.0682, + "step": 11294 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382151484084782e-05, + "loss": 1.0844, + "step": 11295 + }, + { + "epoch": 0.83, + "learning_rate": 1.9382013446098634e-05, + "loss": 1.0126, + "step": 11296 + }, + { + "epoch": 0.83, + "learning_rate": 1.9381875393185836e-05, + "loss": 1.0136, + "step": 11297 + }, + { + "epoch": 0.83, + "learning_rate": 1.938173732534661e-05, + "loss": 1.1301, + "step": 11298 + }, + { + "epoch": 0.83, + "learning_rate": 1.938159924258117e-05, + "loss": 0.9668, + "step": 11299 + }, + { + "epoch": 0.83, + "learning_rate": 1.9381461144889747e-05, + "loss": 0.9417, + "step": 11300 + }, + { + "epoch": 0.84, + "learning_rate": 1.9381323032272548e-05, + "loss": 1.0393, + "step": 11301 + }, + { + "epoch": 0.84, + "learning_rate": 1.93811849047298e-05, + "loss": 1.1216, + "step": 11302 + }, + { + "epoch": 0.84, + "learning_rate": 1.938104676226172e-05, + "loss": 1.0554, + "step": 11303 + }, + { + "epoch": 0.84, + "learning_rate": 1.938090860486853e-05, + "loss": 1.0918, + "step": 11304 + }, + { + "epoch": 0.84, + "learning_rate": 1.938077043255045e-05, + "loss": 0.9823, + "step": 11305 + }, + { + "epoch": 0.84, + "learning_rate": 1.9380632245307694e-05, + "loss": 1.074, + "step": 11306 + }, + { + "epoch": 0.84, + "learning_rate": 1.938049404314049e-05, + "loss": 1.0991, + "step": 11307 + }, + { + "epoch": 0.84, + "learning_rate": 1.938035582604905e-05, + "loss": 1.0472, + "step": 11308 + }, + { + "epoch": 0.84, + "learning_rate": 1.9380217594033602e-05, + "loss": 1.0827, + "step": 11309 + }, + { + "epoch": 0.84, + "learning_rate": 1.9380079347094363e-05, + "loss": 1.0893, + "step": 11310 + }, + { + "epoch": 0.84, + "learning_rate": 1.9379941085231548e-05, + "loss": 0.8766, + "step": 11311 + }, + { + "epoch": 0.84, + "learning_rate": 1.9379802808445385e-05, + "loss": 0.9615, + "step": 11312 + }, + { + "epoch": 0.84, + "learning_rate": 1.9379664516736087e-05, + "loss": 0.9861, + "step": 11313 + }, + { + "epoch": 0.84, + "learning_rate": 1.9379526210103877e-05, + "loss": 0.9283, + "step": 11314 + }, + { + "epoch": 0.84, + "learning_rate": 1.937938788854898e-05, + "loss": 1.0027, + "step": 11315 + }, + { + "epoch": 0.84, + "learning_rate": 1.9379249552071602e-05, + "loss": 1.0622, + "step": 11316 + }, + { + "epoch": 0.84, + "learning_rate": 1.937911120067198e-05, + "loss": 0.9521, + "step": 11317 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378972834350325e-05, + "loss": 1.0348, + "step": 11318 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378834453106856e-05, + "loss": 1.0744, + "step": 11319 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378696056941797e-05, + "loss": 1.0351, + "step": 11320 + }, + { + "epoch": 0.84, + "learning_rate": 1.937855764585537e-05, + "loss": 1.0108, + "step": 11321 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378419219847784e-05, + "loss": 1.0475, + "step": 11322 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378280778919273e-05, + "loss": 1.1109, + "step": 11323 + }, + { + "epoch": 0.84, + "learning_rate": 1.937814232307005e-05, + "loss": 0.9706, + "step": 11324 + }, + { + "epoch": 0.84, + "learning_rate": 1.9378003852300335e-05, + "loss": 1.0884, + "step": 11325 + }, + { + "epoch": 0.84, + "learning_rate": 1.9377865366610355e-05, + "loss": 1.1067, + "step": 11326 + }, + { + "epoch": 0.84, + "learning_rate": 1.937772686600032e-05, + "loss": 1.0121, + "step": 11327 + }, + { + "epoch": 0.84, + "learning_rate": 1.9377588350470456e-05, + "loss": 1.0695, + "step": 11328 + }, + { + "epoch": 0.84, + "learning_rate": 1.9377449820020984e-05, + "loss": 0.9965, + "step": 11329 + }, + { + "epoch": 0.84, + "learning_rate": 1.9377311274652126e-05, + "loss": 1.1167, + "step": 11330 + }, + { + "epoch": 0.84, + "learning_rate": 1.93771727143641e-05, + "loss": 0.9617, + "step": 11331 + }, + { + "epoch": 0.84, + "learning_rate": 1.9377034139157122e-05, + "loss": 1.0561, + "step": 11332 + }, + { + "epoch": 0.84, + "learning_rate": 1.937689554903142e-05, + "loss": 1.0246, + "step": 11333 + }, + { + "epoch": 0.84, + "learning_rate": 1.9376756943987212e-05, + "loss": 0.9817, + "step": 11334 + }, + { + "epoch": 0.84, + "learning_rate": 1.9376618324024715e-05, + "loss": 1.0307, + "step": 11335 + }, + { + "epoch": 0.84, + "learning_rate": 1.9376479689144155e-05, + "loss": 1.0891, + "step": 11336 + }, + { + "epoch": 0.84, + "learning_rate": 1.937634103934575e-05, + "loss": 1.0255, + "step": 11337 + }, + { + "epoch": 0.84, + "learning_rate": 1.937620237462972e-05, + "loss": 0.8761, + "step": 11338 + }, + { + "epoch": 0.84, + "learning_rate": 1.9376063694996284e-05, + "loss": 1.038, + "step": 11339 + }, + { + "epoch": 0.84, + "learning_rate": 1.937592500044567e-05, + "loss": 0.9105, + "step": 11340 + }, + { + "epoch": 0.84, + "learning_rate": 1.9375786290978088e-05, + "loss": 0.9868, + "step": 11341 + }, + { + "epoch": 0.84, + "learning_rate": 1.9375647566593765e-05, + "loss": 1.0501, + "step": 11342 + }, + { + "epoch": 0.84, + "learning_rate": 1.9375508827292925e-05, + "loss": 1.0619, + "step": 11343 + }, + { + "epoch": 0.84, + "learning_rate": 1.937537007307578e-05, + "loss": 1.0411, + "step": 11344 + }, + { + "epoch": 0.84, + "learning_rate": 1.9375231303942556e-05, + "loss": 1.0509, + "step": 11345 + }, + { + "epoch": 0.84, + "learning_rate": 1.9375092519893476e-05, + "loss": 1.0957, + "step": 11346 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374953720928754e-05, + "loss": 0.94, + "step": 11347 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374814907048618e-05, + "loss": 0.944, + "step": 11348 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374676078253288e-05, + "loss": 1.0285, + "step": 11349 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374537234542976e-05, + "loss": 0.9722, + "step": 11350 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374398375917914e-05, + "loss": 0.9741, + "step": 11351 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374259502378317e-05, + "loss": 1.1357, + "step": 11352 + }, + { + "epoch": 0.84, + "learning_rate": 1.9374120613924408e-05, + "loss": 1.0952, + "step": 11353 + }, + { + "epoch": 0.84, + "learning_rate": 1.9373981710556403e-05, + "loss": 1.1091, + "step": 11354 + }, + { + "epoch": 0.84, + "learning_rate": 1.937384279227453e-05, + "loss": 1.0366, + "step": 11355 + }, + { + "epoch": 0.84, + "learning_rate": 1.9373703859079008e-05, + "loss": 1.0309, + "step": 11356 + }, + { + "epoch": 0.84, + "learning_rate": 1.937356491097006e-05, + "loss": 1.0054, + "step": 11357 + }, + { + "epoch": 0.84, + "learning_rate": 1.9373425947947902e-05, + "loss": 1.1457, + "step": 11358 + }, + { + "epoch": 0.84, + "learning_rate": 1.9373286970012754e-05, + "loss": 1.0811, + "step": 11359 + }, + { + "epoch": 0.84, + "learning_rate": 1.9373147977164843e-05, + "loss": 0.928, + "step": 11360 + }, + { + "epoch": 0.84, + "learning_rate": 1.937300896940439e-05, + "loss": 1.0702, + "step": 11361 + }, + { + "epoch": 0.84, + "learning_rate": 1.9372869946731613e-05, + "loss": 1.0314, + "step": 11362 + }, + { + "epoch": 0.84, + "learning_rate": 1.937273090914673e-05, + "loss": 1.0027, + "step": 11363 + }, + { + "epoch": 0.84, + "learning_rate": 1.937259185664997e-05, + "loss": 1.003, + "step": 11364 + }, + { + "epoch": 0.84, + "learning_rate": 1.937245278924155e-05, + "loss": 1.0359, + "step": 11365 + }, + { + "epoch": 0.84, + "learning_rate": 1.937231370692169e-05, + "loss": 0.9711, + "step": 11366 + }, + { + "epoch": 0.84, + "learning_rate": 1.9372174609690615e-05, + "loss": 1.0223, + "step": 11367 + }, + { + "epoch": 0.84, + "learning_rate": 1.9372035497548545e-05, + "loss": 1.0238, + "step": 11368 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371896370495697e-05, + "loss": 1.0553, + "step": 11369 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371757228532297e-05, + "loss": 1.0602, + "step": 11370 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371618071658566e-05, + "loss": 1.1278, + "step": 11371 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371478899874727e-05, + "loss": 0.9835, + "step": 11372 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371339713180998e-05, + "loss": 1.0475, + "step": 11373 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371200511577602e-05, + "loss": 1.0286, + "step": 11374 + }, + { + "epoch": 0.84, + "learning_rate": 1.9371061295064755e-05, + "loss": 1.0351, + "step": 11375 + }, + { + "epoch": 0.84, + "learning_rate": 1.937092206364269e-05, + "loss": 0.9927, + "step": 11376 + }, + { + "epoch": 0.84, + "learning_rate": 1.937078281731162e-05, + "loss": 1.0121, + "step": 11377 + }, + { + "epoch": 0.84, + "learning_rate": 1.9370643556071768e-05, + "loss": 1.0293, + "step": 11378 + }, + { + "epoch": 0.84, + "learning_rate": 1.9370504279923357e-05, + "loss": 1.0533, + "step": 11379 + }, + { + "epoch": 0.84, + "learning_rate": 1.9370364988866607e-05, + "loss": 1.0529, + "step": 11380 + }, + { + "epoch": 0.84, + "learning_rate": 1.937022568290174e-05, + "loss": 1.0836, + "step": 11381 + }, + { + "epoch": 0.84, + "learning_rate": 1.9370086362028978e-05, + "loss": 0.955, + "step": 11382 + }, + { + "epoch": 0.84, + "learning_rate": 1.9369947026248542e-05, + "loss": 1.0481, + "step": 11383 + }, + { + "epoch": 0.84, + "learning_rate": 1.9369807675560656e-05, + "loss": 0.9504, + "step": 11384 + }, + { + "epoch": 0.84, + "learning_rate": 1.9369668309965538e-05, + "loss": 0.9926, + "step": 11385 + }, + { + "epoch": 0.84, + "learning_rate": 1.9369528929463417e-05, + "loss": 1.0376, + "step": 11386 + }, + { + "epoch": 0.84, + "learning_rate": 1.9369389534054505e-05, + "loss": 0.996, + "step": 11387 + }, + { + "epoch": 0.84, + "learning_rate": 1.936925012373903e-05, + "loss": 0.9746, + "step": 11388 + }, + { + "epoch": 0.84, + "learning_rate": 1.936911069851721e-05, + "loss": 0.9896, + "step": 11389 + }, + { + "epoch": 0.84, + "learning_rate": 1.9368971258389272e-05, + "loss": 1.0568, + "step": 11390 + }, + { + "epoch": 0.84, + "learning_rate": 1.936883180335543e-05, + "loss": 1.0486, + "step": 11391 + }, + { + "epoch": 0.84, + "learning_rate": 1.9368692333415913e-05, + "loss": 1.0749, + "step": 11392 + }, + { + "epoch": 0.84, + "learning_rate": 1.9368552848570944e-05, + "loss": 0.95, + "step": 11393 + }, + { + "epoch": 0.84, + "learning_rate": 1.936841334882074e-05, + "loss": 1.0415, + "step": 11394 + }, + { + "epoch": 0.84, + "learning_rate": 1.9368273834165522e-05, + "loss": 1.0217, + "step": 11395 + }, + { + "epoch": 0.84, + "learning_rate": 1.9368134304605514e-05, + "loss": 0.9721, + "step": 11396 + }, + { + "epoch": 0.84, + "learning_rate": 1.936799476014094e-05, + "loss": 1.0475, + "step": 11397 + }, + { + "epoch": 0.84, + "learning_rate": 1.9367855200772018e-05, + "loss": 1.027, + "step": 11398 + }, + { + "epoch": 0.84, + "learning_rate": 1.9367715626498977e-05, + "loss": 0.953, + "step": 11399 + }, + { + "epoch": 0.84, + "learning_rate": 1.9367576037322032e-05, + "loss": 1.0458, + "step": 11400 + }, + { + "epoch": 0.84, + "learning_rate": 1.9367436433241407e-05, + "loss": 1.072, + "step": 11401 + }, + { + "epoch": 0.84, + "learning_rate": 1.9367296814257325e-05, + "loss": 1.1289, + "step": 11402 + }, + { + "epoch": 0.84, + "learning_rate": 1.936715718037001e-05, + "loss": 0.9249, + "step": 11403 + }, + { + "epoch": 0.84, + "learning_rate": 1.936701753157968e-05, + "loss": 0.9047, + "step": 11404 + }, + { + "epoch": 0.84, + "learning_rate": 1.936687786788656e-05, + "loss": 0.9596, + "step": 11405 + }, + { + "epoch": 0.84, + "learning_rate": 1.9366738189290867e-05, + "loss": 1.0321, + "step": 11406 + }, + { + "epoch": 0.84, + "learning_rate": 1.9366598495792833e-05, + "loss": 1.0149, + "step": 11407 + }, + { + "epoch": 0.84, + "learning_rate": 1.9366458787392673e-05, + "loss": 0.9577, + "step": 11408 + }, + { + "epoch": 0.84, + "learning_rate": 1.9366319064090612e-05, + "loss": 1.0533, + "step": 11409 + }, + { + "epoch": 0.84, + "learning_rate": 1.936617932588687e-05, + "loss": 0.986, + "step": 11410 + }, + { + "epoch": 0.84, + "learning_rate": 1.936603957278167e-05, + "loss": 1.0056, + "step": 11411 + }, + { + "epoch": 0.84, + "learning_rate": 1.9365899804775233e-05, + "loss": 0.9569, + "step": 11412 + }, + { + "epoch": 0.84, + "learning_rate": 1.936576002186779e-05, + "loss": 0.9951, + "step": 11413 + }, + { + "epoch": 0.84, + "learning_rate": 1.9365620224059552e-05, + "loss": 1.0763, + "step": 11414 + }, + { + "epoch": 0.84, + "learning_rate": 1.9365480411350748e-05, + "loss": 0.9764, + "step": 11415 + }, + { + "epoch": 0.84, + "learning_rate": 1.9365340583741597e-05, + "loss": 0.9513, + "step": 11416 + }, + { + "epoch": 0.84, + "learning_rate": 1.9365200741232324e-05, + "loss": 0.9977, + "step": 11417 + }, + { + "epoch": 0.84, + "learning_rate": 1.936506088382315e-05, + "loss": 1.1299, + "step": 11418 + }, + { + "epoch": 0.84, + "learning_rate": 1.93649210115143e-05, + "loss": 1.0416, + "step": 11419 + }, + { + "epoch": 0.84, + "learning_rate": 1.9364781124305994e-05, + "loss": 0.973, + "step": 11420 + }, + { + "epoch": 0.84, + "learning_rate": 1.9364641222198456e-05, + "loss": 1.1227, + "step": 11421 + }, + { + "epoch": 0.84, + "learning_rate": 1.9364501305191904e-05, + "loss": 1.0953, + "step": 11422 + }, + { + "epoch": 0.84, + "learning_rate": 1.936436137328657e-05, + "loss": 1.1266, + "step": 11423 + }, + { + "epoch": 0.84, + "learning_rate": 1.9364221426482665e-05, + "loss": 1.1868, + "step": 11424 + }, + { + "epoch": 0.84, + "learning_rate": 1.9364081464780423e-05, + "loss": 1.0189, + "step": 11425 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363941488180058e-05, + "loss": 1.0713, + "step": 11426 + }, + { + "epoch": 0.84, + "learning_rate": 1.93638014966818e-05, + "loss": 1.0335, + "step": 11427 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363661490285864e-05, + "loss": 1.0602, + "step": 11428 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363521468992476e-05, + "loss": 0.9369, + "step": 11429 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363381432801862e-05, + "loss": 1.061, + "step": 11430 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363241381714242e-05, + "loss": 0.9835, + "step": 11431 + }, + { + "epoch": 0.84, + "learning_rate": 1.9363101315729834e-05, + "loss": 1.0468, + "step": 11432 + }, + { + "epoch": 0.84, + "learning_rate": 1.936296123484887e-05, + "loss": 1.0939, + "step": 11433 + }, + { + "epoch": 0.84, + "learning_rate": 1.936282113907157e-05, + "loss": 0.9949, + "step": 11434 + }, + { + "epoch": 0.84, + "learning_rate": 1.9362681028398153e-05, + "loss": 1.1387, + "step": 11435 + }, + { + "epoch": 0.84, + "learning_rate": 1.9362540902828846e-05, + "loss": 0.9943, + "step": 11436 + }, + { + "epoch": 0.85, + "learning_rate": 1.936240076236387e-05, + "loss": 1.0115, + "step": 11437 + }, + { + "epoch": 0.85, + "learning_rate": 1.9362260607003447e-05, + "loss": 1.0386, + "step": 11438 + }, + { + "epoch": 0.85, + "learning_rate": 1.9362120436747803e-05, + "loss": 0.9229, + "step": 11439 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361980251597158e-05, + "loss": 1.0593, + "step": 11440 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361840051551737e-05, + "loss": 1.0835, + "step": 11441 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361699836611764e-05, + "loss": 1.0867, + "step": 11442 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361559606777458e-05, + "loss": 1.0372, + "step": 11443 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361419362049047e-05, + "loss": 1.0296, + "step": 11444 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361279102426747e-05, + "loss": 1.0044, + "step": 11445 + }, + { + "epoch": 0.85, + "learning_rate": 1.9361138827910793e-05, + "loss": 0.9357, + "step": 11446 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360998538501397e-05, + "loss": 1.0333, + "step": 11447 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360858234198784e-05, + "loss": 1.0527, + "step": 11448 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360717915003183e-05, + "loss": 1.078, + "step": 11449 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360577580914812e-05, + "loss": 1.0287, + "step": 11450 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360437231933897e-05, + "loss": 1.0988, + "step": 11451 + }, + { + "epoch": 0.85, + "learning_rate": 1.936029686806066e-05, + "loss": 1.0943, + "step": 11452 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360156489295325e-05, + "loss": 0.9694, + "step": 11453 + }, + { + "epoch": 0.85, + "learning_rate": 1.9360016095638115e-05, + "loss": 1.0709, + "step": 11454 + }, + { + "epoch": 0.85, + "learning_rate": 1.9359875687089253e-05, + "loss": 1.053, + "step": 11455 + }, + { + "epoch": 0.85, + "learning_rate": 1.9359735263648962e-05, + "loss": 1.1177, + "step": 11456 + }, + { + "epoch": 0.85, + "learning_rate": 1.9359594825317466e-05, + "loss": 0.9386, + "step": 11457 + }, + { + "epoch": 0.85, + "learning_rate": 1.935945437209499e-05, + "loss": 1.0306, + "step": 11458 + }, + { + "epoch": 0.85, + "learning_rate": 1.9359313903981756e-05, + "loss": 0.9757, + "step": 11459 + }, + { + "epoch": 0.85, + "learning_rate": 1.9359173420977986e-05, + "loss": 0.9901, + "step": 11460 + }, + { + "epoch": 0.85, + "learning_rate": 1.935903292308391e-05, + "loss": 1.0493, + "step": 11461 + }, + { + "epoch": 0.85, + "learning_rate": 1.9358892410299737e-05, + "loss": 1.0535, + "step": 11462 + }, + { + "epoch": 0.85, + "learning_rate": 1.935875188262571e-05, + "loss": 1.1048, + "step": 11463 + }, + { + "epoch": 0.85, + "learning_rate": 1.9358611340062036e-05, + "loss": 1.0076, + "step": 11464 + }, + { + "epoch": 0.85, + "learning_rate": 1.935847078260895e-05, + "loss": 1.069, + "step": 11465 + }, + { + "epoch": 0.85, + "learning_rate": 1.9358330210266665e-05, + "loss": 1.0678, + "step": 11466 + }, + { + "epoch": 0.85, + "learning_rate": 1.9358189623035413e-05, + "loss": 0.9634, + "step": 11467 + }, + { + "epoch": 0.85, + "learning_rate": 1.9358049020915416e-05, + "loss": 0.9948, + "step": 11468 + }, + { + "epoch": 0.85, + "learning_rate": 1.9357908403906897e-05, + "loss": 0.9963, + "step": 11469 + }, + { + "epoch": 0.85, + "learning_rate": 1.935776777201008e-05, + "loss": 1.0778, + "step": 11470 + }, + { + "epoch": 0.85, + "learning_rate": 1.9357627125225185e-05, + "loss": 0.9974, + "step": 11471 + }, + { + "epoch": 0.85, + "learning_rate": 1.9357486463552445e-05, + "loss": 0.9955, + "step": 11472 + }, + { + "epoch": 0.85, + "learning_rate": 1.9357345786992075e-05, + "loss": 1.0948, + "step": 11473 + }, + { + "epoch": 0.85, + "learning_rate": 1.93572050955443e-05, + "loss": 1.0468, + "step": 11474 + }, + { + "epoch": 0.85, + "learning_rate": 1.9357064389209345e-05, + "loss": 0.9656, + "step": 11475 + }, + { + "epoch": 0.85, + "learning_rate": 1.935692366798744e-05, + "loss": 1.0367, + "step": 11476 + }, + { + "epoch": 0.85, + "learning_rate": 1.9356782931878796e-05, + "loss": 1.068, + "step": 11477 + }, + { + "epoch": 0.85, + "learning_rate": 1.935664218088365e-05, + "loss": 1.0022, + "step": 11478 + }, + { + "epoch": 0.85, + "learning_rate": 1.9356501415002217e-05, + "loss": 0.9728, + "step": 11479 + }, + { + "epoch": 0.85, + "learning_rate": 1.9356360634234724e-05, + "loss": 1.0179, + "step": 11480 + }, + { + "epoch": 0.85, + "learning_rate": 1.93562198385814e-05, + "loss": 1.0815, + "step": 11481 + }, + { + "epoch": 0.85, + "learning_rate": 1.9356079028042457e-05, + "loss": 1.0898, + "step": 11482 + }, + { + "epoch": 0.85, + "learning_rate": 1.935593820261813e-05, + "loss": 1.0868, + "step": 11483 + }, + { + "epoch": 0.85, + "learning_rate": 1.935579736230864e-05, + "loss": 1.0376, + "step": 11484 + }, + { + "epoch": 0.85, + "learning_rate": 1.935565650711421e-05, + "loss": 1.032, + "step": 11485 + }, + { + "epoch": 0.85, + "learning_rate": 1.9355515637035064e-05, + "loss": 0.8945, + "step": 11486 + }, + { + "epoch": 0.85, + "learning_rate": 1.9355374752071426e-05, + "loss": 0.9991, + "step": 11487 + }, + { + "epoch": 0.85, + "learning_rate": 1.9355233852223522e-05, + "loss": 1.1033, + "step": 11488 + }, + { + "epoch": 0.85, + "learning_rate": 1.9355092937491572e-05, + "loss": 1.1149, + "step": 11489 + }, + { + "epoch": 0.85, + "learning_rate": 1.9354952007875806e-05, + "loss": 1.1122, + "step": 11490 + }, + { + "epoch": 0.85, + "learning_rate": 1.935481106337644e-05, + "loss": 0.9753, + "step": 11491 + }, + { + "epoch": 0.85, + "learning_rate": 1.935467010399371e-05, + "loss": 0.9304, + "step": 11492 + }, + { + "epoch": 0.85, + "learning_rate": 1.935452912972783e-05, + "loss": 1.0856, + "step": 11493 + }, + { + "epoch": 0.85, + "learning_rate": 1.935438814057903e-05, + "loss": 1.014, + "step": 11494 + }, + { + "epoch": 0.85, + "learning_rate": 1.935424713654753e-05, + "loss": 0.9821, + "step": 11495 + }, + { + "epoch": 0.85, + "learning_rate": 1.935410611763356e-05, + "loss": 0.9597, + "step": 11496 + }, + { + "epoch": 0.85, + "learning_rate": 1.9353965083837335e-05, + "loss": 1.1027, + "step": 11497 + }, + { + "epoch": 0.85, + "learning_rate": 1.935382403515909e-05, + "loss": 0.9758, + "step": 11498 + }, + { + "epoch": 0.85, + "learning_rate": 1.9353682971599044e-05, + "loss": 0.9677, + "step": 11499 + }, + { + "epoch": 0.85, + "learning_rate": 1.935354189315742e-05, + "loss": 1.0321, + "step": 11500 + }, + { + "epoch": 0.85, + "learning_rate": 1.935340079983445e-05, + "loss": 0.9296, + "step": 11501 + }, + { + "epoch": 0.85, + "learning_rate": 1.9353259691630348e-05, + "loss": 1.1011, + "step": 11502 + }, + { + "epoch": 0.85, + "learning_rate": 1.9353118568545345e-05, + "loss": 0.9955, + "step": 11503 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352977430579664e-05, + "loss": 0.9739, + "step": 11504 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352836277733532e-05, + "loss": 1.1478, + "step": 11505 + }, + { + "epoch": 0.85, + "learning_rate": 1.935269511000717e-05, + "loss": 0.9949, + "step": 11506 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352553927400802e-05, + "loss": 1.0597, + "step": 11507 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352412729914654e-05, + "loss": 1.0229, + "step": 11508 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352271517548957e-05, + "loss": 0.8605, + "step": 11509 + }, + { + "epoch": 0.85, + "learning_rate": 1.9352130290303927e-05, + "loss": 1.0881, + "step": 11510 + }, + { + "epoch": 0.85, + "learning_rate": 1.935198904817979e-05, + "loss": 0.9213, + "step": 11511 + }, + { + "epoch": 0.85, + "learning_rate": 1.935184779117677e-05, + "loss": 1.0714, + "step": 11512 + }, + { + "epoch": 0.85, + "learning_rate": 1.93517065192951e-05, + "loss": 1.0069, + "step": 11513 + }, + { + "epoch": 0.85, + "learning_rate": 1.9351565232534993e-05, + "loss": 1.0239, + "step": 11514 + }, + { + "epoch": 0.85, + "learning_rate": 1.9351423930896685e-05, + "loss": 0.9966, + "step": 11515 + }, + { + "epoch": 0.85, + "learning_rate": 1.9351282614380393e-05, + "loss": 0.9567, + "step": 11516 + }, + { + "epoch": 0.85, + "learning_rate": 1.9351141282986342e-05, + "loss": 1.0119, + "step": 11517 + }, + { + "epoch": 0.85, + "learning_rate": 1.935099993671476e-05, + "loss": 1.0708, + "step": 11518 + }, + { + "epoch": 0.85, + "learning_rate": 1.935085857556587e-05, + "loss": 0.9954, + "step": 11519 + }, + { + "epoch": 0.85, + "learning_rate": 1.9350717199539903e-05, + "loss": 1.0689, + "step": 11520 + }, + { + "epoch": 0.85, + "learning_rate": 1.9350575808637075e-05, + "loss": 0.948, + "step": 11521 + }, + { + "epoch": 0.85, + "learning_rate": 1.935043440285761e-05, + "loss": 0.9601, + "step": 11522 + }, + { + "epoch": 0.85, + "learning_rate": 1.9350292982201745e-05, + "loss": 1.1008, + "step": 11523 + }, + { + "epoch": 0.85, + "learning_rate": 1.9350151546669696e-05, + "loss": 1.0188, + "step": 11524 + }, + { + "epoch": 0.85, + "learning_rate": 1.9350010096261686e-05, + "loss": 1.092, + "step": 11525 + }, + { + "epoch": 0.85, + "learning_rate": 1.9349868630977948e-05, + "loss": 1.0261, + "step": 11526 + }, + { + "epoch": 0.85, + "learning_rate": 1.93497271508187e-05, + "loss": 1.0104, + "step": 11527 + }, + { + "epoch": 0.85, + "learning_rate": 1.934958565578417e-05, + "loss": 1.0928, + "step": 11528 + }, + { + "epoch": 0.85, + "learning_rate": 1.9349444145874586e-05, + "loss": 1.0681, + "step": 11529 + }, + { + "epoch": 0.85, + "learning_rate": 1.9349302621090168e-05, + "loss": 1.0102, + "step": 11530 + }, + { + "epoch": 0.85, + "learning_rate": 1.9349161081431146e-05, + "loss": 0.9427, + "step": 11531 + }, + { + "epoch": 0.85, + "learning_rate": 1.934901952689774e-05, + "loss": 1.0117, + "step": 11532 + }, + { + "epoch": 0.85, + "learning_rate": 1.9348877957490174e-05, + "loss": 0.9111, + "step": 11533 + }, + { + "epoch": 0.85, + "learning_rate": 1.934873637320868e-05, + "loss": 0.9746, + "step": 11534 + }, + { + "epoch": 0.85, + "learning_rate": 1.9348594774053483e-05, + "loss": 1.0004, + "step": 11535 + }, + { + "epoch": 0.85, + "learning_rate": 1.9348453160024803e-05, + "loss": 1.0391, + "step": 11536 + }, + { + "epoch": 0.85, + "learning_rate": 1.9348311531122868e-05, + "loss": 1.1318, + "step": 11537 + }, + { + "epoch": 0.85, + "learning_rate": 1.93481698873479e-05, + "loss": 1.0411, + "step": 11538 + }, + { + "epoch": 0.85, + "learning_rate": 1.9348028228700134e-05, + "loss": 1.0643, + "step": 11539 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347886555179783e-05, + "loss": 1.0317, + "step": 11540 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347744866787082e-05, + "loss": 1.0595, + "step": 11541 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347603163522254e-05, + "loss": 1.0869, + "step": 11542 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347461445385517e-05, + "loss": 1.0664, + "step": 11543 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347319712377106e-05, + "loss": 0.983, + "step": 11544 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347177964497244e-05, + "loss": 0.9924, + "step": 11545 + }, + { + "epoch": 0.85, + "learning_rate": 1.9347036201746153e-05, + "loss": 0.8833, + "step": 11546 + }, + { + "epoch": 0.85, + "learning_rate": 1.9346894424124067e-05, + "loss": 1.0084, + "step": 11547 + }, + { + "epoch": 0.85, + "learning_rate": 1.93467526316312e-05, + "loss": 1.1047, + "step": 11548 + }, + { + "epoch": 0.85, + "learning_rate": 1.934661082426778e-05, + "loss": 1.0523, + "step": 11549 + }, + { + "epoch": 0.85, + "learning_rate": 1.934646900203404e-05, + "loss": 1.1734, + "step": 11550 + }, + { + "epoch": 0.85, + "learning_rate": 1.934632716493021e-05, + "loss": 1.106, + "step": 11551 + }, + { + "epoch": 0.85, + "learning_rate": 1.9346185312956495e-05, + "loss": 0.9697, + "step": 11552 + }, + { + "epoch": 0.85, + "learning_rate": 1.9346043446113136e-05, + "loss": 1.0698, + "step": 11553 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345901564400357e-05, + "loss": 1.0525, + "step": 11554 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345759667818385e-05, + "loss": 1.0528, + "step": 11555 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345617756367438e-05, + "loss": 0.9732, + "step": 11556 + }, + { + "epoch": 0.85, + "learning_rate": 1.934547583004775e-05, + "loss": 0.9034, + "step": 11557 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345333888859542e-05, + "loss": 1.0553, + "step": 11558 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345191932803044e-05, + "loss": 1.112, + "step": 11559 + }, + { + "epoch": 0.85, + "learning_rate": 1.9345049961878478e-05, + "loss": 1.0111, + "step": 11560 + }, + { + "epoch": 0.85, + "learning_rate": 1.934490797608607e-05, + "loss": 1.0104, + "step": 11561 + }, + { + "epoch": 0.85, + "learning_rate": 1.934476597542605e-05, + "loss": 1.0428, + "step": 11562 + }, + { + "epoch": 0.85, + "learning_rate": 1.9344623959898635e-05, + "loss": 1.079, + "step": 11563 + }, + { + "epoch": 0.85, + "learning_rate": 1.9344481929504062e-05, + "loss": 0.972, + "step": 11564 + }, + { + "epoch": 0.85, + "learning_rate": 1.9344339884242553e-05, + "loss": 0.9791, + "step": 11565 + }, + { + "epoch": 0.85, + "learning_rate": 1.934419782411433e-05, + "loss": 1.0536, + "step": 11566 + }, + { + "epoch": 0.85, + "learning_rate": 1.9344055749119625e-05, + "loss": 1.0019, + "step": 11567 + }, + { + "epoch": 0.85, + "learning_rate": 1.9343913659258657e-05, + "loss": 0.9508, + "step": 11568 + }, + { + "epoch": 0.85, + "learning_rate": 1.9343771554531657e-05, + "loss": 0.9893, + "step": 11569 + }, + { + "epoch": 0.85, + "learning_rate": 1.934362943493885e-05, + "loss": 1.0175, + "step": 11570 + }, + { + "epoch": 0.85, + "learning_rate": 1.9343487300480466e-05, + "loss": 1.1, + "step": 11571 + }, + { + "epoch": 0.86, + "learning_rate": 1.9343345151156724e-05, + "loss": 1.0644, + "step": 11572 + }, + { + "epoch": 0.86, + "learning_rate": 1.9343202986967852e-05, + "loss": 1.1033, + "step": 11573 + }, + { + "epoch": 0.86, + "learning_rate": 1.934306080791408e-05, + "loss": 0.9749, + "step": 11574 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342918613995633e-05, + "loss": 0.9228, + "step": 11575 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342776405212737e-05, + "loss": 1.0037, + "step": 11576 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342634181565615e-05, + "loss": 1.0296, + "step": 11577 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342491943054495e-05, + "loss": 1.0802, + "step": 11578 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342349689679606e-05, + "loss": 1.0937, + "step": 11579 + }, + { + "epoch": 0.86, + "learning_rate": 1.9342207421441174e-05, + "loss": 1.0144, + "step": 11580 + }, + { + "epoch": 0.86, + "learning_rate": 1.934206513833942e-05, + "loss": 0.9937, + "step": 11581 + }, + { + "epoch": 0.86, + "learning_rate": 1.9341922840374572e-05, + "loss": 1.0538, + "step": 11582 + }, + { + "epoch": 0.86, + "learning_rate": 1.934178052754686e-05, + "loss": 0.9562, + "step": 11583 + }, + { + "epoch": 0.86, + "learning_rate": 1.9341638199856514e-05, + "loss": 0.9675, + "step": 11584 + }, + { + "epoch": 0.86, + "learning_rate": 1.934149585730375e-05, + "loss": 1.0307, + "step": 11585 + }, + { + "epoch": 0.86, + "learning_rate": 1.9341353499888802e-05, + "loss": 1.021, + "step": 11586 + }, + { + "epoch": 0.86, + "learning_rate": 1.9341211127611894e-05, + "loss": 1.0326, + "step": 11587 + }, + { + "epoch": 0.86, + "learning_rate": 1.934106874047325e-05, + "loss": 1.0731, + "step": 11588 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340926338473106e-05, + "loss": 1.0327, + "step": 11589 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340783921611677e-05, + "loss": 1.1289, + "step": 11590 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340641489889194e-05, + "loss": 1.0474, + "step": 11591 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340499043305884e-05, + "loss": 1.0364, + "step": 11592 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340356581861973e-05, + "loss": 1.0273, + "step": 11593 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340214105557693e-05, + "loss": 0.9975, + "step": 11594 + }, + { + "epoch": 0.86, + "learning_rate": 1.9340071614393263e-05, + "loss": 1.086, + "step": 11595 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339929108368914e-05, + "loss": 1.0344, + "step": 11596 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339786587484867e-05, + "loss": 1.0303, + "step": 11597 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339644051741357e-05, + "loss": 0.9014, + "step": 11598 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339501501138606e-05, + "loss": 1.1633, + "step": 11599 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339358935676844e-05, + "loss": 1.0226, + "step": 11600 + }, + { + "epoch": 0.86, + "learning_rate": 1.9339216355356295e-05, + "loss": 1.037, + "step": 11601 + }, + { + "epoch": 0.86, + "learning_rate": 1.933907376017718e-05, + "loss": 0.9703, + "step": 11602 + }, + { + "epoch": 0.86, + "learning_rate": 1.9338931150139738e-05, + "loss": 0.9542, + "step": 11603 + }, + { + "epoch": 0.86, + "learning_rate": 1.933878852524419e-05, + "loss": 0.9937, + "step": 11604 + }, + { + "epoch": 0.86, + "learning_rate": 1.9338645885490762e-05, + "loss": 1.06, + "step": 11605 + }, + { + "epoch": 0.86, + "learning_rate": 1.9338503230879683e-05, + "loss": 1.0002, + "step": 11606 + }, + { + "epoch": 0.86, + "learning_rate": 1.933836056141118e-05, + "loss": 1.1191, + "step": 11607 + }, + { + "epoch": 0.86, + "learning_rate": 1.9338217877085477e-05, + "loss": 1.0041, + "step": 11608 + }, + { + "epoch": 0.86, + "learning_rate": 1.9338075177902803e-05, + "loss": 0.9937, + "step": 11609 + }, + { + "epoch": 0.86, + "learning_rate": 1.9337932463863385e-05, + "loss": 1.0809, + "step": 11610 + }, + { + "epoch": 0.86, + "learning_rate": 1.933778973496745e-05, + "loss": 0.9672, + "step": 11611 + }, + { + "epoch": 0.86, + "learning_rate": 1.9337646991215226e-05, + "loss": 1.0145, + "step": 11612 + }, + { + "epoch": 0.86, + "learning_rate": 1.933750423260694e-05, + "loss": 0.8833, + "step": 11613 + }, + { + "epoch": 0.86, + "learning_rate": 1.9337361459142817e-05, + "loss": 0.9656, + "step": 11614 + }, + { + "epoch": 0.86, + "learning_rate": 1.9337218670823085e-05, + "loss": 0.8754, + "step": 11615 + }, + { + "epoch": 0.86, + "learning_rate": 1.933707586764797e-05, + "loss": 1.0575, + "step": 11616 + }, + { + "epoch": 0.86, + "learning_rate": 1.9336933049617704e-05, + "loss": 0.974, + "step": 11617 + }, + { + "epoch": 0.86, + "learning_rate": 1.933679021673251e-05, + "loss": 0.9974, + "step": 11618 + }, + { + "epoch": 0.86, + "learning_rate": 1.933664736899262e-05, + "loss": 0.9859, + "step": 11619 + }, + { + "epoch": 0.86, + "learning_rate": 1.9336504506398252e-05, + "loss": 0.9844, + "step": 11620 + }, + { + "epoch": 0.86, + "learning_rate": 1.933636162894964e-05, + "loss": 1.1108, + "step": 11621 + }, + { + "epoch": 0.86, + "learning_rate": 1.9336218736647013e-05, + "loss": 1.0356, + "step": 11622 + }, + { + "epoch": 0.86, + "learning_rate": 1.933607582949059e-05, + "loss": 1.0999, + "step": 11623 + }, + { + "epoch": 0.86, + "learning_rate": 1.933593290748061e-05, + "loss": 1.0211, + "step": 11624 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335789970617292e-05, + "loss": 0.9968, + "step": 11625 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335647018900866e-05, + "loss": 1.0286, + "step": 11626 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335504052331558e-05, + "loss": 1.0643, + "step": 11627 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335361070909598e-05, + "loss": 1.0637, + "step": 11628 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335218074635214e-05, + "loss": 1.1084, + "step": 11629 + }, + { + "epoch": 0.86, + "learning_rate": 1.9335075063508632e-05, + "loss": 1.1148, + "step": 11630 + }, + { + "epoch": 0.86, + "learning_rate": 1.9334932037530073e-05, + "loss": 0.9519, + "step": 11631 + }, + { + "epoch": 0.86, + "learning_rate": 1.9334788996699777e-05, + "loss": 0.9654, + "step": 11632 + }, + { + "epoch": 0.86, + "learning_rate": 1.933464594101796e-05, + "loss": 1.0654, + "step": 11633 + }, + { + "epoch": 0.86, + "learning_rate": 1.933450287048486e-05, + "loss": 1.0506, + "step": 11634 + }, + { + "epoch": 0.86, + "learning_rate": 1.9334359785100697e-05, + "loss": 0.9031, + "step": 11635 + }, + { + "epoch": 0.86, + "learning_rate": 1.93342166848657e-05, + "loss": 0.9581, + "step": 11636 + }, + { + "epoch": 0.86, + "learning_rate": 1.9334073569780098e-05, + "loss": 1.0276, + "step": 11637 + }, + { + "epoch": 0.86, + "learning_rate": 1.933393043984412e-05, + "loss": 0.9805, + "step": 11638 + }, + { + "epoch": 0.86, + "learning_rate": 1.933378729505799e-05, + "loss": 0.8639, + "step": 11639 + }, + { + "epoch": 0.86, + "learning_rate": 1.933364413542194e-05, + "loss": 1.0989, + "step": 11640 + }, + { + "epoch": 0.86, + "learning_rate": 1.9333500960936196e-05, + "loss": 1.0434, + "step": 11641 + }, + { + "epoch": 0.86, + "learning_rate": 1.9333357771600983e-05, + "loss": 1.0257, + "step": 11642 + }, + { + "epoch": 0.86, + "learning_rate": 1.9333214567416534e-05, + "loss": 1.0669, + "step": 11643 + }, + { + "epoch": 0.86, + "learning_rate": 1.9333071348383076e-05, + "loss": 1.0937, + "step": 11644 + }, + { + "epoch": 0.86, + "learning_rate": 1.9332928114500832e-05, + "loss": 1.0433, + "step": 11645 + }, + { + "epoch": 0.86, + "learning_rate": 1.9332784865770033e-05, + "loss": 1.1538, + "step": 11646 + }, + { + "epoch": 0.86, + "learning_rate": 1.9332641602190905e-05, + "loss": 1.0529, + "step": 11647 + }, + { + "epoch": 0.86, + "learning_rate": 1.933249832376368e-05, + "loss": 1.0083, + "step": 11648 + }, + { + "epoch": 0.86, + "learning_rate": 1.9332355030488583e-05, + "loss": 1.0099, + "step": 11649 + }, + { + "epoch": 0.86, + "learning_rate": 1.9332211722365843e-05, + "loss": 1.1017, + "step": 11650 + }, + { + "epoch": 0.86, + "learning_rate": 1.933206839939569e-05, + "loss": 1.0419, + "step": 11651 + }, + { + "epoch": 0.86, + "learning_rate": 1.9331925061578345e-05, + "loss": 1.0663, + "step": 11652 + }, + { + "epoch": 0.86, + "learning_rate": 1.9331781708914042e-05, + "loss": 1.1214, + "step": 11653 + }, + { + "epoch": 0.86, + "learning_rate": 1.933163834140301e-05, + "loss": 1.0923, + "step": 11654 + }, + { + "epoch": 0.86, + "learning_rate": 1.9331494959045475e-05, + "loss": 0.9298, + "step": 11655 + }, + { + "epoch": 0.86, + "learning_rate": 1.9331351561841662e-05, + "loss": 1.0803, + "step": 11656 + }, + { + "epoch": 0.86, + "learning_rate": 1.9331208149791804e-05, + "loss": 1.0101, + "step": 11657 + }, + { + "epoch": 0.86, + "learning_rate": 1.933106472289613e-05, + "loss": 1.0312, + "step": 11658 + }, + { + "epoch": 0.86, + "learning_rate": 1.9330921281154862e-05, + "loss": 0.9978, + "step": 11659 + }, + { + "epoch": 0.86, + "learning_rate": 1.9330777824568236e-05, + "loss": 1.0352, + "step": 11660 + }, + { + "epoch": 0.86, + "learning_rate": 1.933063435313647e-05, + "loss": 0.9645, + "step": 11661 + }, + { + "epoch": 0.86, + "learning_rate": 1.93304908668598e-05, + "loss": 1.1132, + "step": 11662 + }, + { + "epoch": 0.86, + "learning_rate": 1.9330347365738454e-05, + "loss": 1.0331, + "step": 11663 + }, + { + "epoch": 0.86, + "learning_rate": 1.9330203849772657e-05, + "loss": 1.0057, + "step": 11664 + }, + { + "epoch": 0.86, + "learning_rate": 1.9330060318962644e-05, + "loss": 1.0624, + "step": 11665 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329916773308636e-05, + "loss": 1.0147, + "step": 11666 + }, + { + "epoch": 0.86, + "learning_rate": 1.932977321281086e-05, + "loss": 1.025, + "step": 11667 + }, + { + "epoch": 0.86, + "learning_rate": 1.932962963746955e-05, + "loss": 1.1627, + "step": 11668 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329486047284937e-05, + "loss": 1.1077, + "step": 11669 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329342442257243e-05, + "loss": 1.0707, + "step": 11670 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329198822386693e-05, + "loss": 1.0007, + "step": 11671 + }, + { + "epoch": 0.86, + "learning_rate": 1.9329055187673527e-05, + "loss": 1.0133, + "step": 11672 + }, + { + "epoch": 0.86, + "learning_rate": 1.932891153811797e-05, + "loss": 0.9452, + "step": 11673 + }, + { + "epoch": 0.86, + "learning_rate": 1.932876787372024e-05, + "loss": 1.0182, + "step": 11674 + }, + { + "epoch": 0.86, + "learning_rate": 1.932862419448058e-05, + "loss": 0.8908, + "step": 11675 + }, + { + "epoch": 0.86, + "learning_rate": 1.9328480500399213e-05, + "loss": 1.0027, + "step": 11676 + }, + { + "epoch": 0.86, + "learning_rate": 1.9328336791476366e-05, + "loss": 1.0, + "step": 11677 + }, + { + "epoch": 0.86, + "learning_rate": 1.9328193067712268e-05, + "loss": 1.028, + "step": 11678 + }, + { + "epoch": 0.86, + "learning_rate": 1.9328049329107146e-05, + "loss": 0.9164, + "step": 11679 + }, + { + "epoch": 0.86, + "learning_rate": 1.9327905575661235e-05, + "loss": 0.9421, + "step": 11680 + }, + { + "epoch": 0.86, + "learning_rate": 1.9327761807374754e-05, + "loss": 0.9985, + "step": 11681 + }, + { + "epoch": 0.86, + "learning_rate": 1.932761802424794e-05, + "loss": 0.9688, + "step": 11682 + }, + { + "epoch": 0.86, + "learning_rate": 1.9327474226281023e-05, + "loss": 1.1565, + "step": 11683 + }, + { + "epoch": 0.86, + "learning_rate": 1.9327330413474223e-05, + "loss": 0.9837, + "step": 11684 + }, + { + "epoch": 0.86, + "learning_rate": 1.932718658582778e-05, + "loss": 1.0702, + "step": 11685 + }, + { + "epoch": 0.86, + "learning_rate": 1.932704274334191e-05, + "loss": 0.9159, + "step": 11686 + }, + { + "epoch": 0.86, + "learning_rate": 1.932689888601685e-05, + "loss": 1.0666, + "step": 11687 + }, + { + "epoch": 0.86, + "learning_rate": 1.932675501385283e-05, + "loss": 1.1282, + "step": 11688 + }, + { + "epoch": 0.86, + "learning_rate": 1.932661112685007e-05, + "loss": 1.054, + "step": 11689 + }, + { + "epoch": 0.86, + "learning_rate": 1.9326467225008812e-05, + "loss": 1.0295, + "step": 11690 + }, + { + "epoch": 0.86, + "learning_rate": 1.9326323308329277e-05, + "loss": 0.9739, + "step": 11691 + }, + { + "epoch": 0.86, + "learning_rate": 1.9326179376811694e-05, + "loss": 1.0423, + "step": 11692 + }, + { + "epoch": 0.86, + "learning_rate": 1.9326035430456293e-05, + "loss": 1.0198, + "step": 11693 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325891469263303e-05, + "loss": 1.1135, + "step": 11694 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325747493232952e-05, + "loss": 1.0612, + "step": 11695 + }, + { + "epoch": 0.86, + "learning_rate": 1.932560350236547e-05, + "loss": 1.0582, + "step": 11696 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325459496661087e-05, + "loss": 1.0493, + "step": 11697 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325315476120036e-05, + "loss": 1.0449, + "step": 11698 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325171440742535e-05, + "loss": 0.9639, + "step": 11699 + }, + { + "epoch": 0.86, + "learning_rate": 1.9325027390528822e-05, + "loss": 1.091, + "step": 11700 + }, + { + "epoch": 0.86, + "learning_rate": 1.9324883325479124e-05, + "loss": 1.0806, + "step": 11701 + }, + { + "epoch": 0.86, + "learning_rate": 1.9324739245593667e-05, + "loss": 1.0484, + "step": 11702 + }, + { + "epoch": 0.86, + "learning_rate": 1.932459515087269e-05, + "loss": 0.9689, + "step": 11703 + }, + { + "epoch": 0.86, + "learning_rate": 1.932445104131641e-05, + "loss": 1.0822, + "step": 11704 + }, + { + "epoch": 0.86, + "learning_rate": 1.9324306916925065e-05, + "loss": 0.9894, + "step": 11705 + }, + { + "epoch": 0.86, + "learning_rate": 1.932416277769888e-05, + "loss": 0.9258, + "step": 11706 + }, + { + "epoch": 0.87, + "learning_rate": 1.9324018623638083e-05, + "loss": 1.0336, + "step": 11707 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323874454742908e-05, + "loss": 1.0757, + "step": 11708 + }, + { + "epoch": 0.87, + "learning_rate": 1.932373027101358e-05, + "loss": 0.8867, + "step": 11709 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323586072450334e-05, + "loss": 1.0074, + "step": 11710 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323441859053393e-05, + "loss": 1.0917, + "step": 11711 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323297630822994e-05, + "loss": 1.0315, + "step": 11712 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323153387759357e-05, + "loss": 1.1307, + "step": 11713 + }, + { + "epoch": 0.87, + "learning_rate": 1.9323009129862716e-05, + "loss": 0.9421, + "step": 11714 + }, + { + "epoch": 0.87, + "learning_rate": 1.9322864857133306e-05, + "loss": 0.9185, + "step": 11715 + }, + { + "epoch": 0.87, + "learning_rate": 1.9322720569571346e-05, + "loss": 1.0848, + "step": 11716 + }, + { + "epoch": 0.87, + "learning_rate": 1.9322576267177076e-05, + "loss": 0.9808, + "step": 11717 + }, + { + "epoch": 0.87, + "learning_rate": 1.9322431949950714e-05, + "loss": 1.0612, + "step": 11718 + }, + { + "epoch": 0.87, + "learning_rate": 1.93222876178925e-05, + "loss": 1.0531, + "step": 11719 + }, + { + "epoch": 0.87, + "learning_rate": 1.9322143271002662e-05, + "loss": 1.1357, + "step": 11720 + }, + { + "epoch": 0.87, + "learning_rate": 1.9321998909281422e-05, + "loss": 0.947, + "step": 11721 + }, + { + "epoch": 0.87, + "learning_rate": 1.9321854532729022e-05, + "loss": 0.9524, + "step": 11722 + }, + { + "epoch": 0.87, + "learning_rate": 1.932171014134568e-05, + "loss": 1.0538, + "step": 11723 + }, + { + "epoch": 0.87, + "learning_rate": 1.932156573513163e-05, + "loss": 1.0452, + "step": 11724 + }, + { + "epoch": 0.87, + "learning_rate": 1.9321421314087107e-05, + "loss": 1.003, + "step": 11725 + }, + { + "epoch": 0.87, + "learning_rate": 1.932127687821233e-05, + "loss": 1.0, + "step": 11726 + }, + { + "epoch": 0.87, + "learning_rate": 1.932113242750754e-05, + "loss": 1.1137, + "step": 11727 + }, + { + "epoch": 0.87, + "learning_rate": 1.9320987961972958e-05, + "loss": 1.026, + "step": 11728 + }, + { + "epoch": 0.87, + "learning_rate": 1.932084348160882e-05, + "loss": 1.033, + "step": 11729 + }, + { + "epoch": 0.87, + "learning_rate": 1.932069898641535e-05, + "loss": 1.0027, + "step": 11730 + }, + { + "epoch": 0.87, + "learning_rate": 1.9320554476392787e-05, + "loss": 0.9733, + "step": 11731 + }, + { + "epoch": 0.87, + "learning_rate": 1.932040995154135e-05, + "loss": 1.027, + "step": 11732 + }, + { + "epoch": 0.87, + "learning_rate": 1.9320265411861276e-05, + "loss": 1.0594, + "step": 11733 + }, + { + "epoch": 0.87, + "learning_rate": 1.9320120857352793e-05, + "loss": 0.9759, + "step": 11734 + }, + { + "epoch": 0.87, + "learning_rate": 1.931997628801613e-05, + "loss": 1.1036, + "step": 11735 + }, + { + "epoch": 0.87, + "learning_rate": 1.931983170385152e-05, + "loss": 0.9706, + "step": 11736 + }, + { + "epoch": 0.87, + "learning_rate": 1.931968710485919e-05, + "loss": 0.9976, + "step": 11737 + }, + { + "epoch": 0.87, + "learning_rate": 1.931954249103937e-05, + "loss": 0.8996, + "step": 11738 + }, + { + "epoch": 0.87, + "learning_rate": 1.9319397862392292e-05, + "loss": 1.0155, + "step": 11739 + }, + { + "epoch": 0.87, + "learning_rate": 1.9319253218918187e-05, + "loss": 1.023, + "step": 11740 + }, + { + "epoch": 0.87, + "learning_rate": 1.9319108560617282e-05, + "loss": 1.041, + "step": 11741 + }, + { + "epoch": 0.87, + "learning_rate": 1.9318963887489814e-05, + "loss": 1.0165, + "step": 11742 + }, + { + "epoch": 0.87, + "learning_rate": 1.9318819199536e-05, + "loss": 1.1228, + "step": 11743 + }, + { + "epoch": 0.87, + "learning_rate": 1.931867449675608e-05, + "loss": 1.0281, + "step": 11744 + }, + { + "epoch": 0.87, + "learning_rate": 1.9318529779150284e-05, + "loss": 1.0901, + "step": 11745 + }, + { + "epoch": 0.87, + "learning_rate": 1.931838504671884e-05, + "loss": 1.0541, + "step": 11746 + }, + { + "epoch": 0.87, + "learning_rate": 1.9318240299461977e-05, + "loss": 1.0129, + "step": 11747 + }, + { + "epoch": 0.87, + "learning_rate": 1.9318095537379928e-05, + "loss": 1.0299, + "step": 11748 + }, + { + "epoch": 0.87, + "learning_rate": 1.9317950760472923e-05, + "loss": 1.0456, + "step": 11749 + }, + { + "epoch": 0.87, + "learning_rate": 1.931780596874119e-05, + "loss": 1.0128, + "step": 11750 + }, + { + "epoch": 0.87, + "learning_rate": 1.9317661162184966e-05, + "loss": 1.1259, + "step": 11751 + }, + { + "epoch": 0.87, + "learning_rate": 1.931751634080447e-05, + "loss": 1.0614, + "step": 11752 + }, + { + "epoch": 0.87, + "learning_rate": 1.931737150459994e-05, + "loss": 0.9976, + "step": 11753 + }, + { + "epoch": 0.87, + "learning_rate": 1.9317226653571608e-05, + "loss": 1.0281, + "step": 11754 + }, + { + "epoch": 0.87, + "learning_rate": 1.93170817877197e-05, + "loss": 1.0652, + "step": 11755 + }, + { + "epoch": 0.87, + "learning_rate": 1.931693690704445e-05, + "loss": 1.1277, + "step": 11756 + }, + { + "epoch": 0.87, + "learning_rate": 1.9316792011546085e-05, + "loss": 1.0754, + "step": 11757 + }, + { + "epoch": 0.87, + "learning_rate": 1.9316647101224838e-05, + "loss": 0.9829, + "step": 11758 + }, + { + "epoch": 0.87, + "learning_rate": 1.9316502176080937e-05, + "loss": 0.9818, + "step": 11759 + }, + { + "epoch": 0.87, + "learning_rate": 1.9316357236114614e-05, + "loss": 1.1125, + "step": 11760 + }, + { + "epoch": 0.87, + "learning_rate": 1.9316212281326102e-05, + "loss": 0.9839, + "step": 11761 + }, + { + "epoch": 0.87, + "learning_rate": 1.931606731171563e-05, + "loss": 1.0705, + "step": 11762 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315922327283427e-05, + "loss": 0.9851, + "step": 11763 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315777328029727e-05, + "loss": 0.9585, + "step": 11764 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315632313954757e-05, + "loss": 0.9523, + "step": 11765 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315487285058746e-05, + "loss": 1.0363, + "step": 11766 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315342241341933e-05, + "loss": 0.9237, + "step": 11767 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315197182804543e-05, + "loss": 1.0042, + "step": 11768 + }, + { + "epoch": 0.87, + "learning_rate": 1.9315052109446804e-05, + "loss": 1.0207, + "step": 11769 + }, + { + "epoch": 0.87, + "learning_rate": 1.9314907021268953e-05, + "loss": 1.0029, + "step": 11770 + }, + { + "epoch": 0.87, + "learning_rate": 1.9314761918271217e-05, + "loss": 1.1161, + "step": 11771 + }, + { + "epoch": 0.87, + "learning_rate": 1.931461680045383e-05, + "loss": 1.1591, + "step": 11772 + }, + { + "epoch": 0.87, + "learning_rate": 1.931447166781702e-05, + "loss": 1.1005, + "step": 11773 + }, + { + "epoch": 0.87, + "learning_rate": 1.9314326520361017e-05, + "loss": 1.156, + "step": 11774 + }, + { + "epoch": 0.87, + "learning_rate": 1.931418135808606e-05, + "loss": 1.1224, + "step": 11775 + }, + { + "epoch": 0.87, + "learning_rate": 1.9314036180992364e-05, + "loss": 1.0095, + "step": 11776 + }, + { + "epoch": 0.87, + "learning_rate": 1.9313890989080177e-05, + "loss": 0.9488, + "step": 11777 + }, + { + "epoch": 0.87, + "learning_rate": 1.931374578234972e-05, + "loss": 1.1183, + "step": 11778 + }, + { + "epoch": 0.87, + "learning_rate": 1.9313600560801227e-05, + "loss": 1.1073, + "step": 11779 + }, + { + "epoch": 0.87, + "learning_rate": 1.9313455324434928e-05, + "loss": 1.0391, + "step": 11780 + }, + { + "epoch": 0.87, + "learning_rate": 1.9313310073251058e-05, + "loss": 0.9509, + "step": 11781 + }, + { + "epoch": 0.87, + "learning_rate": 1.931316480724984e-05, + "loss": 1.0818, + "step": 11782 + }, + { + "epoch": 0.87, + "learning_rate": 1.9313019526431517e-05, + "loss": 1.1014, + "step": 11783 + }, + { + "epoch": 0.87, + "learning_rate": 1.9312874230796307e-05, + "loss": 1.1096, + "step": 11784 + }, + { + "epoch": 0.87, + "learning_rate": 1.931272892034445e-05, + "loss": 1.0332, + "step": 11785 + }, + { + "epoch": 0.87, + "learning_rate": 1.9312583595076172e-05, + "loss": 1.0569, + "step": 11786 + }, + { + "epoch": 0.87, + "learning_rate": 1.931243825499171e-05, + "loss": 1.0111, + "step": 11787 + }, + { + "epoch": 0.87, + "learning_rate": 1.931229290009129e-05, + "loss": 0.9055, + "step": 11788 + }, + { + "epoch": 0.87, + "learning_rate": 1.9312147530375145e-05, + "loss": 1.0099, + "step": 11789 + }, + { + "epoch": 0.87, + "learning_rate": 1.9312002145843508e-05, + "loss": 0.9906, + "step": 11790 + }, + { + "epoch": 0.87, + "learning_rate": 1.9311856746496608e-05, + "loss": 0.9777, + "step": 11791 + }, + { + "epoch": 0.87, + "learning_rate": 1.9311711332334678e-05, + "loss": 1.0781, + "step": 11792 + }, + { + "epoch": 0.87, + "learning_rate": 1.931156590335795e-05, + "loss": 0.9789, + "step": 11793 + }, + { + "epoch": 0.87, + "learning_rate": 1.9311420459566647e-05, + "loss": 0.938, + "step": 11794 + }, + { + "epoch": 0.87, + "learning_rate": 1.9311275000961014e-05, + "loss": 1.0551, + "step": 11795 + }, + { + "epoch": 0.87, + "learning_rate": 1.9311129527541276e-05, + "loss": 0.9309, + "step": 11796 + }, + { + "epoch": 0.87, + "learning_rate": 1.931098403930766e-05, + "loss": 1.134, + "step": 11797 + }, + { + "epoch": 0.87, + "learning_rate": 1.9310838536260403e-05, + "loss": 1.0056, + "step": 11798 + }, + { + "epoch": 0.87, + "learning_rate": 1.9310693018399736e-05, + "loss": 0.9836, + "step": 11799 + }, + { + "epoch": 0.87, + "learning_rate": 1.931054748572589e-05, + "loss": 1.0115, + "step": 11800 + }, + { + "epoch": 0.87, + "learning_rate": 1.9310401938239094e-05, + "loss": 1.0475, + "step": 11801 + }, + { + "epoch": 0.87, + "learning_rate": 1.9310256375939582e-05, + "loss": 1.0128, + "step": 11802 + }, + { + "epoch": 0.87, + "learning_rate": 1.9310110798827585e-05, + "loss": 1.0666, + "step": 11803 + }, + { + "epoch": 0.87, + "learning_rate": 1.930996520690334e-05, + "loss": 1.0164, + "step": 11804 + }, + { + "epoch": 0.87, + "learning_rate": 1.9309819600167066e-05, + "loss": 0.9966, + "step": 11805 + }, + { + "epoch": 0.87, + "learning_rate": 1.930967397861901e-05, + "loss": 1.0821, + "step": 11806 + }, + { + "epoch": 0.87, + "learning_rate": 1.9309528342259392e-05, + "loss": 0.9291, + "step": 11807 + }, + { + "epoch": 0.87, + "learning_rate": 1.9309382691088445e-05, + "loss": 0.9194, + "step": 11808 + }, + { + "epoch": 0.87, + "learning_rate": 1.9309237025106408e-05, + "loss": 1.1196, + "step": 11809 + }, + { + "epoch": 0.87, + "learning_rate": 1.9309091344313506e-05, + "loss": 0.9825, + "step": 11810 + }, + { + "epoch": 0.87, + "learning_rate": 1.930894564870997e-05, + "loss": 0.9886, + "step": 11811 + }, + { + "epoch": 0.87, + "learning_rate": 1.930879993829604e-05, + "loss": 1.0543, + "step": 11812 + }, + { + "epoch": 0.87, + "learning_rate": 1.930865421307194e-05, + "loss": 0.9801, + "step": 11813 + }, + { + "epoch": 0.87, + "learning_rate": 1.9308508473037904e-05, + "loss": 1.0482, + "step": 11814 + }, + { + "epoch": 0.87, + "learning_rate": 1.9308362718194163e-05, + "loss": 0.9542, + "step": 11815 + }, + { + "epoch": 0.87, + "learning_rate": 1.9308216948540952e-05, + "loss": 1.0343, + "step": 11816 + }, + { + "epoch": 0.87, + "learning_rate": 1.93080711640785e-05, + "loss": 1.1632, + "step": 11817 + }, + { + "epoch": 0.87, + "learning_rate": 1.930792536480704e-05, + "loss": 0.9889, + "step": 11818 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307779550726807e-05, + "loss": 0.9858, + "step": 11819 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307633721838027e-05, + "loss": 1.0946, + "step": 11820 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307487878140936e-05, + "loss": 0.9569, + "step": 11821 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307342019635764e-05, + "loss": 1.1053, + "step": 11822 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307196146322746e-05, + "loss": 1.0324, + "step": 11823 + }, + { + "epoch": 0.87, + "learning_rate": 1.9307050258202107e-05, + "loss": 1.0703, + "step": 11824 + }, + { + "epoch": 0.87, + "learning_rate": 1.930690435527409e-05, + "loss": 1.0444, + "step": 11825 + }, + { + "epoch": 0.87, + "learning_rate": 1.9306758437538916e-05, + "loss": 0.9862, + "step": 11826 + }, + { + "epoch": 0.87, + "learning_rate": 1.9306612504996827e-05, + "loss": 0.885, + "step": 11827 + }, + { + "epoch": 0.87, + "learning_rate": 1.930646655764805e-05, + "loss": 1.014, + "step": 11828 + }, + { + "epoch": 0.87, + "learning_rate": 1.9306320595492815e-05, + "loss": 0.9154, + "step": 11829 + }, + { + "epoch": 0.87, + "learning_rate": 1.930617461853136e-05, + "loss": 0.961, + "step": 11830 + }, + { + "epoch": 0.87, + "learning_rate": 1.9306028626763914e-05, + "loss": 1.0261, + "step": 11831 + }, + { + "epoch": 0.87, + "learning_rate": 1.930588262019071e-05, + "loss": 0.8992, + "step": 11832 + }, + { + "epoch": 0.87, + "learning_rate": 1.9305736598811977e-05, + "loss": 1.026, + "step": 11833 + }, + { + "epoch": 0.87, + "learning_rate": 1.930559056262795e-05, + "loss": 1.0479, + "step": 11834 + }, + { + "epoch": 0.87, + "learning_rate": 1.9305444511638868e-05, + "loss": 1.0248, + "step": 11835 + }, + { + "epoch": 0.87, + "learning_rate": 1.9305298445844947e-05, + "loss": 0.9503, + "step": 11836 + }, + { + "epoch": 0.87, + "learning_rate": 1.930515236524644e-05, + "loss": 1.0725, + "step": 11837 + }, + { + "epoch": 0.87, + "learning_rate": 1.930500626984356e-05, + "loss": 0.9997, + "step": 11838 + }, + { + "epoch": 0.87, + "learning_rate": 1.9304860159636555e-05, + "loss": 0.9312, + "step": 11839 + }, + { + "epoch": 0.87, + "learning_rate": 1.9304714034625646e-05, + "loss": 1.0643, + "step": 11840 + }, + { + "epoch": 0.87, + "learning_rate": 1.9304567894811072e-05, + "loss": 1.0473, + "step": 11841 + }, + { + "epoch": 0.87, + "learning_rate": 1.930442174019306e-05, + "loss": 1.0803, + "step": 11842 + }, + { + "epoch": 0.88, + "learning_rate": 1.9304275570771853e-05, + "loss": 1.0009, + "step": 11843 + }, + { + "epoch": 0.88, + "learning_rate": 1.9304129386547673e-05, + "loss": 0.981, + "step": 11844 + }, + { + "epoch": 0.88, + "learning_rate": 1.9303983187520755e-05, + "loss": 0.9776, + "step": 11845 + }, + { + "epoch": 0.88, + "learning_rate": 1.9303836973691334e-05, + "loss": 0.9777, + "step": 11846 + }, + { + "epoch": 0.88, + "learning_rate": 1.930369074505964e-05, + "loss": 1.0286, + "step": 11847 + }, + { + "epoch": 0.88, + "learning_rate": 1.930354450162591e-05, + "loss": 0.9703, + "step": 11848 + }, + { + "epoch": 0.88, + "learning_rate": 1.930339824339037e-05, + "loss": 1.0102, + "step": 11849 + }, + { + "epoch": 0.88, + "learning_rate": 1.930325197035326e-05, + "loss": 1.0896, + "step": 11850 + }, + { + "epoch": 0.88, + "learning_rate": 1.930310568251481e-05, + "loss": 0.9697, + "step": 11851 + }, + { + "epoch": 0.88, + "learning_rate": 1.930295937987525e-05, + "loss": 0.9955, + "step": 11852 + }, + { + "epoch": 0.88, + "learning_rate": 1.9302813062434812e-05, + "loss": 1.0303, + "step": 11853 + }, + { + "epoch": 0.88, + "learning_rate": 1.930266673019374e-05, + "loss": 0.9229, + "step": 11854 + }, + { + "epoch": 0.88, + "learning_rate": 1.930252038315225e-05, + "loss": 0.9985, + "step": 11855 + }, + { + "epoch": 0.88, + "learning_rate": 1.9302374021310586e-05, + "loss": 1.058, + "step": 11856 + }, + { + "epoch": 0.88, + "learning_rate": 1.9302227644668977e-05, + "loss": 1.0498, + "step": 11857 + }, + { + "epoch": 0.88, + "learning_rate": 1.930208125322766e-05, + "loss": 1.0304, + "step": 11858 + }, + { + "epoch": 0.88, + "learning_rate": 1.930193484698686e-05, + "loss": 1.0098, + "step": 11859 + }, + { + "epoch": 0.88, + "learning_rate": 1.930178842594682e-05, + "loss": 0.998, + "step": 11860 + }, + { + "epoch": 0.88, + "learning_rate": 1.9301641990107763e-05, + "loss": 1.056, + "step": 11861 + }, + { + "epoch": 0.88, + "learning_rate": 1.930149553946993e-05, + "loss": 1.0055, + "step": 11862 + }, + { + "epoch": 0.88, + "learning_rate": 1.9301349074033548e-05, + "loss": 1.1178, + "step": 11863 + }, + { + "epoch": 0.88, + "learning_rate": 1.9301202593798854e-05, + "loss": 1.0959, + "step": 11864 + }, + { + "epoch": 0.88, + "learning_rate": 1.930105609876608e-05, + "loss": 1.1054, + "step": 11865 + }, + { + "epoch": 0.88, + "learning_rate": 1.930090958893546e-05, + "loss": 1.0069, + "step": 11866 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300763064307225e-05, + "loss": 0.99, + "step": 11867 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300616524881608e-05, + "loss": 1.0952, + "step": 11868 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300469970658844e-05, + "loss": 1.0866, + "step": 11869 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300323401639166e-05, + "loss": 1.0355, + "step": 11870 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300176817822806e-05, + "loss": 0.9997, + "step": 11871 + }, + { + "epoch": 0.88, + "learning_rate": 1.9300030219209996e-05, + "loss": 0.9594, + "step": 11872 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299883605800974e-05, + "loss": 1.0168, + "step": 11873 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299736977595967e-05, + "loss": 0.9609, + "step": 11874 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299590334595212e-05, + "loss": 0.9842, + "step": 11875 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299443676798947e-05, + "loss": 1.021, + "step": 11876 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299297004207394e-05, + "loss": 1.0755, + "step": 11877 + }, + { + "epoch": 0.88, + "learning_rate": 1.9299150316820795e-05, + "loss": 0.978, + "step": 11878 + }, + { + "epoch": 0.88, + "learning_rate": 1.929900361463938e-05, + "loss": 1.071, + "step": 11879 + }, + { + "epoch": 0.88, + "learning_rate": 1.9298856897663385e-05, + "loss": 0.871, + "step": 11880 + }, + { + "epoch": 0.88, + "learning_rate": 1.929871016589304e-05, + "loss": 1.0273, + "step": 11881 + }, + { + "epoch": 0.88, + "learning_rate": 1.929856341932858e-05, + "loss": 0.9659, + "step": 11882 + }, + { + "epoch": 0.88, + "learning_rate": 1.9298416657970242e-05, + "loss": 0.9581, + "step": 11883 + }, + { + "epoch": 0.88, + "learning_rate": 1.9298269881818252e-05, + "loss": 1.085, + "step": 11884 + }, + { + "epoch": 0.88, + "learning_rate": 1.929812309087285e-05, + "loss": 0.9534, + "step": 11885 + }, + { + "epoch": 0.88, + "learning_rate": 1.9297976285134267e-05, + "loss": 1.0992, + "step": 11886 + }, + { + "epoch": 0.88, + "learning_rate": 1.9297829464602735e-05, + "loss": 1.0408, + "step": 11887 + }, + { + "epoch": 0.88, + "learning_rate": 1.929768262927849e-05, + "loss": 1.0065, + "step": 11888 + }, + { + "epoch": 0.88, + "learning_rate": 1.9297535779161768e-05, + "loss": 0.9753, + "step": 11889 + }, + { + "epoch": 0.88, + "learning_rate": 1.9297388914252794e-05, + "loss": 0.9151, + "step": 11890 + }, + { + "epoch": 0.88, + "learning_rate": 1.929724203455181e-05, + "loss": 1.0439, + "step": 11891 + }, + { + "epoch": 0.88, + "learning_rate": 1.929709514005905e-05, + "loss": 1.0319, + "step": 11892 + }, + { + "epoch": 0.88, + "learning_rate": 1.929694823077474e-05, + "loss": 1.1262, + "step": 11893 + }, + { + "epoch": 0.88, + "learning_rate": 1.929680130669912e-05, + "loss": 0.952, + "step": 11894 + }, + { + "epoch": 0.88, + "learning_rate": 1.9296654367832425e-05, + "loss": 1.0178, + "step": 11895 + }, + { + "epoch": 0.88, + "learning_rate": 1.9296507414174882e-05, + "loss": 1.0491, + "step": 11896 + }, + { + "epoch": 0.88, + "learning_rate": 1.9296360445726732e-05, + "loss": 1.0528, + "step": 11897 + }, + { + "epoch": 0.88, + "learning_rate": 1.92962134624882e-05, + "loss": 1.1295, + "step": 11898 + }, + { + "epoch": 0.88, + "learning_rate": 1.9296066464459532e-05, + "loss": 1.0626, + "step": 11899 + }, + { + "epoch": 0.88, + "learning_rate": 1.929591945164095e-05, + "loss": 1.0132, + "step": 11900 + }, + { + "epoch": 0.88, + "learning_rate": 1.92957724240327e-05, + "loss": 1.0399, + "step": 11901 + }, + { + "epoch": 0.88, + "learning_rate": 1.9295625381635e-05, + "loss": 0.9717, + "step": 11902 + }, + { + "epoch": 0.88, + "learning_rate": 1.92954783244481e-05, + "loss": 1.0543, + "step": 11903 + }, + { + "epoch": 0.88, + "learning_rate": 1.9295331252472223e-05, + "loss": 1.022, + "step": 11904 + }, + { + "epoch": 0.88, + "learning_rate": 1.929518416570761e-05, + "loss": 1.0147, + "step": 11905 + }, + { + "epoch": 0.88, + "learning_rate": 1.929503706415449e-05, + "loss": 1.0254, + "step": 11906 + }, + { + "epoch": 0.88, + "learning_rate": 1.92948899478131e-05, + "loss": 0.975, + "step": 11907 + }, + { + "epoch": 0.88, + "learning_rate": 1.929474281668367e-05, + "loss": 1.1303, + "step": 11908 + }, + { + "epoch": 0.88, + "learning_rate": 1.929459567076644e-05, + "loss": 1.0524, + "step": 11909 + }, + { + "epoch": 0.88, + "learning_rate": 1.9294448510061642e-05, + "loss": 1.0698, + "step": 11910 + }, + { + "epoch": 0.88, + "learning_rate": 1.929430133456951e-05, + "loss": 1.0151, + "step": 11911 + }, + { + "epoch": 0.88, + "learning_rate": 1.9294154144290273e-05, + "loss": 1.0445, + "step": 11912 + }, + { + "epoch": 0.88, + "learning_rate": 1.9294006939224174e-05, + "loss": 1.0166, + "step": 11913 + }, + { + "epoch": 0.88, + "learning_rate": 1.929385971937144e-05, + "loss": 1.1406, + "step": 11914 + }, + { + "epoch": 0.88, + "learning_rate": 1.929371248473231e-05, + "loss": 1.0013, + "step": 11915 + }, + { + "epoch": 0.88, + "learning_rate": 1.929356523530702e-05, + "loss": 1.0547, + "step": 11916 + }, + { + "epoch": 0.88, + "learning_rate": 1.9293417971095797e-05, + "loss": 1.1113, + "step": 11917 + }, + { + "epoch": 0.88, + "learning_rate": 1.9293270692098876e-05, + "loss": 1.1131, + "step": 11918 + }, + { + "epoch": 0.88, + "learning_rate": 1.92931233983165e-05, + "loss": 0.965, + "step": 11919 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292976089748895e-05, + "loss": 1.1237, + "step": 11920 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292828766396298e-05, + "loss": 1.0874, + "step": 11921 + }, + { + "epoch": 0.88, + "learning_rate": 1.929268142825894e-05, + "loss": 1.0635, + "step": 11922 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292534075337064e-05, + "loss": 0.9813, + "step": 11923 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292386707630898e-05, + "loss": 1.0538, + "step": 11924 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292239325140676e-05, + "loss": 0.9739, + "step": 11925 + }, + { + "epoch": 0.88, + "learning_rate": 1.9292091927866633e-05, + "loss": 1.0175, + "step": 11926 + }, + { + "epoch": 0.88, + "learning_rate": 1.929194451580901e-05, + "loss": 1.1065, + "step": 11927 + }, + { + "epoch": 0.88, + "learning_rate": 1.929179708896803e-05, + "loss": 1.0218, + "step": 11928 + }, + { + "epoch": 0.88, + "learning_rate": 1.9291649647343938e-05, + "loss": 1.0878, + "step": 11929 + }, + { + "epoch": 0.88, + "learning_rate": 1.9291502190936963e-05, + "loss": 1.0234, + "step": 11930 + }, + { + "epoch": 0.88, + "learning_rate": 1.929135471974734e-05, + "loss": 1.0248, + "step": 11931 + }, + { + "epoch": 0.88, + "learning_rate": 1.9291207233775305e-05, + "loss": 1.0741, + "step": 11932 + }, + { + "epoch": 0.88, + "learning_rate": 1.9291059733021096e-05, + "loss": 1.0582, + "step": 11933 + }, + { + "epoch": 0.88, + "learning_rate": 1.929091221748494e-05, + "loss": 1.12, + "step": 11934 + }, + { + "epoch": 0.88, + "learning_rate": 1.9290764687167076e-05, + "loss": 1.0218, + "step": 11935 + }, + { + "epoch": 0.88, + "learning_rate": 1.929061714206774e-05, + "loss": 1.0917, + "step": 11936 + }, + { + "epoch": 0.88, + "learning_rate": 1.9290469582187164e-05, + "loss": 1.0563, + "step": 11937 + }, + { + "epoch": 0.88, + "learning_rate": 1.9290322007525584e-05, + "loss": 1.0954, + "step": 11938 + }, + { + "epoch": 0.88, + "learning_rate": 1.9290174418083234e-05, + "loss": 1.058, + "step": 11939 + }, + { + "epoch": 0.88, + "learning_rate": 1.929002681386035e-05, + "loss": 0.9758, + "step": 11940 + }, + { + "epoch": 0.88, + "learning_rate": 1.9289879194857164e-05, + "loss": 0.9606, + "step": 11941 + }, + { + "epoch": 0.88, + "learning_rate": 1.9289731561073917e-05, + "loss": 1.0111, + "step": 11942 + }, + { + "epoch": 0.88, + "learning_rate": 1.928958391251084e-05, + "loss": 1.0184, + "step": 11943 + }, + { + "epoch": 0.88, + "learning_rate": 1.9289436249168162e-05, + "loss": 1.0305, + "step": 11944 + }, + { + "epoch": 0.88, + "learning_rate": 1.928928857104613e-05, + "loss": 0.9588, + "step": 11945 + }, + { + "epoch": 0.88, + "learning_rate": 1.928914087814497e-05, + "loss": 1.101, + "step": 11946 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288993170464922e-05, + "loss": 1.0148, + "step": 11947 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288845448006216e-05, + "loss": 0.9444, + "step": 11948 + }, + { + "epoch": 0.88, + "learning_rate": 1.928869771076909e-05, + "loss": 1.115, + "step": 11949 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288549958753778e-05, + "loss": 1.0506, + "step": 11950 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288402191960517e-05, + "loss": 1.0571, + "step": 11951 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288254410389544e-05, + "loss": 1.029, + "step": 11952 + }, + { + "epoch": 0.88, + "learning_rate": 1.9288106614041088e-05, + "loss": 1.0892, + "step": 11953 + }, + { + "epoch": 0.88, + "learning_rate": 1.928795880291539e-05, + "loss": 1.0357, + "step": 11954 + }, + { + "epoch": 0.88, + "learning_rate": 1.928781097701268e-05, + "loss": 1.0532, + "step": 11955 + }, + { + "epoch": 0.88, + "learning_rate": 1.9287663136333196e-05, + "loss": 1.0355, + "step": 11956 + }, + { + "epoch": 0.88, + "learning_rate": 1.9287515280877175e-05, + "loss": 0.9478, + "step": 11957 + }, + { + "epoch": 0.88, + "learning_rate": 1.9287367410644843e-05, + "loss": 1.0803, + "step": 11958 + }, + { + "epoch": 0.88, + "learning_rate": 1.928721952563645e-05, + "loss": 1.1099, + "step": 11959 + }, + { + "epoch": 0.88, + "learning_rate": 1.9287071625852222e-05, + "loss": 1.0836, + "step": 11960 + }, + { + "epoch": 0.88, + "learning_rate": 1.9286923711292393e-05, + "loss": 1.0436, + "step": 11961 + }, + { + "epoch": 0.88, + "learning_rate": 1.9286775781957203e-05, + "loss": 0.9951, + "step": 11962 + }, + { + "epoch": 0.88, + "learning_rate": 1.9286627837846885e-05, + "loss": 0.9779, + "step": 11963 + }, + { + "epoch": 0.88, + "learning_rate": 1.9286479878961676e-05, + "loss": 1.0259, + "step": 11964 + }, + { + "epoch": 0.88, + "learning_rate": 1.928633190530181e-05, + "loss": 0.9949, + "step": 11965 + }, + { + "epoch": 0.88, + "learning_rate": 1.9286183916867526e-05, + "loss": 1.0426, + "step": 11966 + }, + { + "epoch": 0.88, + "learning_rate": 1.928603591365905e-05, + "loss": 1.0917, + "step": 11967 + }, + { + "epoch": 0.88, + "learning_rate": 1.928588789567663e-05, + "loss": 0.9152, + "step": 11968 + }, + { + "epoch": 0.88, + "learning_rate": 1.928573986292049e-05, + "loss": 0.9515, + "step": 11969 + }, + { + "epoch": 0.88, + "learning_rate": 1.9285591815390875e-05, + "loss": 1.0218, + "step": 11970 + }, + { + "epoch": 0.88, + "learning_rate": 1.9285443753088012e-05, + "loss": 1.0336, + "step": 11971 + }, + { + "epoch": 0.88, + "learning_rate": 1.9285295676012143e-05, + "loss": 1.0464, + "step": 11972 + }, + { + "epoch": 0.88, + "learning_rate": 1.92851475841635e-05, + "loss": 1.0559, + "step": 11973 + }, + { + "epoch": 0.88, + "learning_rate": 1.9284999477542322e-05, + "loss": 0.9721, + "step": 11974 + }, + { + "epoch": 0.88, + "learning_rate": 1.928485135614884e-05, + "loss": 1.0229, + "step": 11975 + }, + { + "epoch": 0.88, + "learning_rate": 1.92847032199833e-05, + "loss": 0.9871, + "step": 11976 + }, + { + "epoch": 0.88, + "learning_rate": 1.9284555069045923e-05, + "loss": 0.9308, + "step": 11977 + }, + { + "epoch": 0.89, + "learning_rate": 1.9284406903336953e-05, + "loss": 1.0668, + "step": 11978 + }, + { + "epoch": 0.89, + "learning_rate": 1.9284258722856622e-05, + "loss": 1.0536, + "step": 11979 + }, + { + "epoch": 0.89, + "learning_rate": 1.9284110527605172e-05, + "loss": 1.0719, + "step": 11980 + }, + { + "epoch": 0.89, + "learning_rate": 1.9283962317582837e-05, + "loss": 0.9323, + "step": 11981 + }, + { + "epoch": 0.89, + "learning_rate": 1.9283814092789845e-05, + "loss": 1.0388, + "step": 11982 + }, + { + "epoch": 0.89, + "learning_rate": 1.928366585322644e-05, + "loss": 1.1051, + "step": 11983 + }, + { + "epoch": 0.89, + "learning_rate": 1.9283517598892856e-05, + "loss": 0.8997, + "step": 11984 + }, + { + "epoch": 0.89, + "learning_rate": 1.9283369329789328e-05, + "loss": 0.999, + "step": 11985 + }, + { + "epoch": 0.89, + "learning_rate": 1.928322104591609e-05, + "loss": 1.0731, + "step": 11986 + }, + { + "epoch": 0.89, + "learning_rate": 1.9283072747273386e-05, + "loss": 1.0641, + "step": 11987 + }, + { + "epoch": 0.89, + "learning_rate": 1.928292443386144e-05, + "loss": 1.093, + "step": 11988 + }, + { + "epoch": 0.89, + "learning_rate": 1.9282776105680495e-05, + "loss": 1.061, + "step": 11989 + }, + { + "epoch": 0.89, + "learning_rate": 1.928262776273079e-05, + "loss": 1.0701, + "step": 11990 + }, + { + "epoch": 0.89, + "learning_rate": 1.9282479405012554e-05, + "loss": 1.0215, + "step": 11991 + }, + { + "epoch": 0.89, + "learning_rate": 1.9282331032526026e-05, + "loss": 1.0235, + "step": 11992 + }, + { + "epoch": 0.89, + "learning_rate": 1.9282182645271445e-05, + "loss": 1.076, + "step": 11993 + }, + { + "epoch": 0.89, + "learning_rate": 1.928203424324904e-05, + "loss": 1.0328, + "step": 11994 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281885826459054e-05, + "loss": 1.0264, + "step": 11995 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281737394901722e-05, + "loss": 1.0053, + "step": 11996 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281588948577275e-05, + "loss": 1.0014, + "step": 11997 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281440487485954e-05, + "loss": 0.9458, + "step": 11998 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281292011627993e-05, + "loss": 0.9238, + "step": 11999 + }, + { + "epoch": 0.89, + "learning_rate": 1.9281143521003632e-05, + "loss": 1.0903, + "step": 12000 + }, + { + "epoch": 0.89, + "learning_rate": 1.92809950156131e-05, + "loss": 1.037, + "step": 12001 + }, + { + "epoch": 0.89, + "learning_rate": 1.928084649545664e-05, + "loss": 0.9237, + "step": 12002 + }, + { + "epoch": 0.89, + "learning_rate": 1.928069796053449e-05, + "loss": 1.0386, + "step": 12003 + }, + { + "epoch": 0.89, + "learning_rate": 1.9280549410846877e-05, + "loss": 0.9457, + "step": 12004 + }, + { + "epoch": 0.89, + "learning_rate": 1.9280400846394046e-05, + "loss": 0.9386, + "step": 12005 + }, + { + "epoch": 0.89, + "learning_rate": 1.9280252267176226e-05, + "loss": 1.0436, + "step": 12006 + }, + { + "epoch": 0.89, + "learning_rate": 1.9280103673193662e-05, + "loss": 1.0205, + "step": 12007 + }, + { + "epoch": 0.89, + "learning_rate": 1.9279955064446582e-05, + "loss": 1.0214, + "step": 12008 + }, + { + "epoch": 0.89, + "learning_rate": 1.927980644093523e-05, + "loss": 1.0724, + "step": 12009 + }, + { + "epoch": 0.89, + "learning_rate": 1.9279657802659836e-05, + "loss": 1.0007, + "step": 12010 + }, + { + "epoch": 0.89, + "learning_rate": 1.927950914962064e-05, + "loss": 0.984, + "step": 12011 + }, + { + "epoch": 0.89, + "learning_rate": 1.9279360481817878e-05, + "loss": 0.9525, + "step": 12012 + }, + { + "epoch": 0.89, + "learning_rate": 1.9279211799251785e-05, + "loss": 0.9699, + "step": 12013 + }, + { + "epoch": 0.89, + "learning_rate": 1.9279063101922602e-05, + "loss": 1.1111, + "step": 12014 + }, + { + "epoch": 0.89, + "learning_rate": 1.9278914389830557e-05, + "loss": 1.0302, + "step": 12015 + }, + { + "epoch": 0.89, + "learning_rate": 1.9278765662975896e-05, + "loss": 1.0193, + "step": 12016 + }, + { + "epoch": 0.89, + "learning_rate": 1.9278616921358854e-05, + "loss": 0.8685, + "step": 12017 + }, + { + "epoch": 0.89, + "learning_rate": 1.927846816497966e-05, + "loss": 0.9688, + "step": 12018 + }, + { + "epoch": 0.89, + "learning_rate": 1.9278319393838557e-05, + "loss": 0.9487, + "step": 12019 + }, + { + "epoch": 0.89, + "learning_rate": 1.927817060793578e-05, + "loss": 1.0584, + "step": 12020 + }, + { + "epoch": 0.89, + "learning_rate": 1.927802180727157e-05, + "loss": 1.1234, + "step": 12021 + }, + { + "epoch": 0.89, + "learning_rate": 1.927787299184616e-05, + "loss": 1.0507, + "step": 12022 + }, + { + "epoch": 0.89, + "learning_rate": 1.927772416165978e-05, + "loss": 1.1931, + "step": 12023 + }, + { + "epoch": 0.89, + "learning_rate": 1.927757531671268e-05, + "loss": 1.0468, + "step": 12024 + }, + { + "epoch": 0.89, + "learning_rate": 1.9277426457005087e-05, + "loss": 1.1031, + "step": 12025 + }, + { + "epoch": 0.89, + "learning_rate": 1.9277277582537243e-05, + "loss": 1.0677, + "step": 12026 + }, + { + "epoch": 0.89, + "learning_rate": 1.9277128693309385e-05, + "loss": 1.0472, + "step": 12027 + }, + { + "epoch": 0.89, + "learning_rate": 1.9276979789321744e-05, + "loss": 0.9012, + "step": 12028 + }, + { + "epoch": 0.89, + "learning_rate": 1.9276830870574565e-05, + "loss": 1.0683, + "step": 12029 + }, + { + "epoch": 0.89, + "learning_rate": 1.9276681937068077e-05, + "loss": 1.0567, + "step": 12030 + }, + { + "epoch": 0.89, + "learning_rate": 1.927653298880252e-05, + "loss": 1.0526, + "step": 12031 + }, + { + "epoch": 0.89, + "learning_rate": 1.9276384025778137e-05, + "loss": 1.0818, + "step": 12032 + }, + { + "epoch": 0.89, + "learning_rate": 1.9276235047995156e-05, + "loss": 1.0443, + "step": 12033 + }, + { + "epoch": 0.89, + "learning_rate": 1.927608605545382e-05, + "loss": 0.933, + "step": 12034 + }, + { + "epoch": 0.89, + "learning_rate": 1.9275937048154363e-05, + "loss": 1.0298, + "step": 12035 + }, + { + "epoch": 0.89, + "learning_rate": 1.927578802609702e-05, + "loss": 1.0144, + "step": 12036 + }, + { + "epoch": 0.89, + "learning_rate": 1.9275638989282034e-05, + "loss": 1.067, + "step": 12037 + }, + { + "epoch": 0.89, + "learning_rate": 1.9275489937709635e-05, + "loss": 1.0055, + "step": 12038 + }, + { + "epoch": 0.89, + "learning_rate": 1.927534087138007e-05, + "loss": 0.9469, + "step": 12039 + }, + { + "epoch": 0.89, + "learning_rate": 1.927519179029357e-05, + "loss": 1.0722, + "step": 12040 + }, + { + "epoch": 0.89, + "learning_rate": 1.927504269445037e-05, + "loss": 0.9927, + "step": 12041 + }, + { + "epoch": 0.89, + "learning_rate": 1.927489358385071e-05, + "loss": 1.03, + "step": 12042 + }, + { + "epoch": 0.89, + "learning_rate": 1.927474445849483e-05, + "loss": 0.9411, + "step": 12043 + }, + { + "epoch": 0.89, + "learning_rate": 1.9274595318382957e-05, + "loss": 1.0109, + "step": 12044 + }, + { + "epoch": 0.89, + "learning_rate": 1.9274446163515342e-05, + "loss": 1.0635, + "step": 12045 + }, + { + "epoch": 0.89, + "learning_rate": 1.9274296993892213e-05, + "loss": 1.0223, + "step": 12046 + }, + { + "epoch": 0.89, + "learning_rate": 1.9274147809513814e-05, + "loss": 1.1312, + "step": 12047 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273998610380375e-05, + "loss": 0.9318, + "step": 12048 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273849396492137e-05, + "loss": 1.1009, + "step": 12049 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273700167849338e-05, + "loss": 1.1316, + "step": 12050 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273550924452214e-05, + "loss": 1.0338, + "step": 12051 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273401666301002e-05, + "loss": 1.0074, + "step": 12052 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273252393395944e-05, + "loss": 1.0759, + "step": 12053 + }, + { + "epoch": 0.89, + "learning_rate": 1.9273103105737273e-05, + "loss": 1.0464, + "step": 12054 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272953803325225e-05, + "loss": 1.0508, + "step": 12055 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272804486160043e-05, + "loss": 0.9777, + "step": 12056 + }, + { + "epoch": 0.89, + "learning_rate": 1.927265515424196e-05, + "loss": 0.9983, + "step": 12057 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272505807571216e-05, + "loss": 0.9859, + "step": 12058 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272356446148047e-05, + "loss": 1.0967, + "step": 12059 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272207069972695e-05, + "loss": 0.9762, + "step": 12060 + }, + { + "epoch": 0.89, + "learning_rate": 1.9272057679045392e-05, + "loss": 1.1598, + "step": 12061 + }, + { + "epoch": 0.89, + "learning_rate": 1.927190827336638e-05, + "loss": 0.9899, + "step": 12062 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271758852935892e-05, + "loss": 0.9942, + "step": 12063 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271609417754168e-05, + "loss": 1.0214, + "step": 12064 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271459967821446e-05, + "loss": 0.8937, + "step": 12065 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271310503137964e-05, + "loss": 0.9666, + "step": 12066 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271161023703962e-05, + "loss": 1.0169, + "step": 12067 + }, + { + "epoch": 0.89, + "learning_rate": 1.9271011529519675e-05, + "loss": 1.0331, + "step": 12068 + }, + { + "epoch": 0.89, + "learning_rate": 1.9270862020585338e-05, + "loss": 1.0362, + "step": 12069 + }, + { + "epoch": 0.89, + "learning_rate": 1.9270712496901197e-05, + "loss": 1.0089, + "step": 12070 + }, + { + "epoch": 0.89, + "learning_rate": 1.927056295846748e-05, + "loss": 1.146, + "step": 12071 + }, + { + "epoch": 0.89, + "learning_rate": 1.9270413405284435e-05, + "loss": 1.0536, + "step": 12072 + }, + { + "epoch": 0.89, + "learning_rate": 1.9270263837352292e-05, + "loss": 0.9524, + "step": 12073 + }, + { + "epoch": 0.89, + "learning_rate": 1.927011425467129e-05, + "loss": 1.0492, + "step": 12074 + }, + { + "epoch": 0.89, + "learning_rate": 1.9269964657241675e-05, + "loss": 1.0341, + "step": 12075 + }, + { + "epoch": 0.89, + "learning_rate": 1.926981504506367e-05, + "loss": 1.1347, + "step": 12076 + }, + { + "epoch": 0.89, + "learning_rate": 1.926966541813753e-05, + "loss": 0.9313, + "step": 12077 + }, + { + "epoch": 0.89, + "learning_rate": 1.926951577646348e-05, + "loss": 0.9545, + "step": 12078 + }, + { + "epoch": 0.89, + "learning_rate": 1.9269366120041763e-05, + "loss": 1.1086, + "step": 12079 + }, + { + "epoch": 0.89, + "learning_rate": 1.926921644887262e-05, + "loss": 0.9353, + "step": 12080 + }, + { + "epoch": 0.89, + "learning_rate": 1.9269066762956286e-05, + "loss": 1.0771, + "step": 12081 + }, + { + "epoch": 0.89, + "learning_rate": 1.9268917062293e-05, + "loss": 1.096, + "step": 12082 + }, + { + "epoch": 0.89, + "learning_rate": 1.9268767346882998e-05, + "loss": 1.1383, + "step": 12083 + }, + { + "epoch": 0.89, + "learning_rate": 1.926861761672652e-05, + "loss": 1.0296, + "step": 12084 + }, + { + "epoch": 0.89, + "learning_rate": 1.9268467871823803e-05, + "loss": 1.1426, + "step": 12085 + }, + { + "epoch": 0.89, + "learning_rate": 1.9268318112175086e-05, + "loss": 1.0724, + "step": 12086 + }, + { + "epoch": 0.89, + "learning_rate": 1.9268168337780607e-05, + "loss": 1.0986, + "step": 12087 + }, + { + "epoch": 0.89, + "learning_rate": 1.926801854864061e-05, + "loss": 1.0298, + "step": 12088 + }, + { + "epoch": 0.89, + "learning_rate": 1.9267868744755323e-05, + "loss": 1.0541, + "step": 12089 + }, + { + "epoch": 0.89, + "learning_rate": 1.9267718926124992e-05, + "loss": 1.0078, + "step": 12090 + }, + { + "epoch": 0.89, + "learning_rate": 1.926756909274985e-05, + "loss": 1.1042, + "step": 12091 + }, + { + "epoch": 0.89, + "learning_rate": 1.9267419244630142e-05, + "loss": 1.0758, + "step": 12092 + }, + { + "epoch": 0.89, + "learning_rate": 1.9267269381766097e-05, + "loss": 1.0097, + "step": 12093 + }, + { + "epoch": 0.89, + "learning_rate": 1.9267119504157967e-05, + "loss": 1.0295, + "step": 12094 + }, + { + "epoch": 0.89, + "learning_rate": 1.9266969611805974e-05, + "loss": 0.9624, + "step": 12095 + }, + { + "epoch": 0.89, + "learning_rate": 1.926681970471037e-05, + "loss": 1.117, + "step": 12096 + }, + { + "epoch": 0.89, + "learning_rate": 1.9266669782871386e-05, + "loss": 0.9957, + "step": 12097 + }, + { + "epoch": 0.89, + "learning_rate": 1.9266519846289264e-05, + "loss": 1.0435, + "step": 12098 + }, + { + "epoch": 0.89, + "learning_rate": 1.9266369894964242e-05, + "loss": 1.0338, + "step": 12099 + }, + { + "epoch": 0.89, + "learning_rate": 1.9266219928896558e-05, + "loss": 1.0801, + "step": 12100 + }, + { + "epoch": 0.89, + "learning_rate": 1.926606994808645e-05, + "loss": 0.9414, + "step": 12101 + }, + { + "epoch": 0.89, + "learning_rate": 1.9265919952534157e-05, + "loss": 1.163, + "step": 12102 + }, + { + "epoch": 0.89, + "learning_rate": 1.926576994223992e-05, + "loss": 1.0884, + "step": 12103 + }, + { + "epoch": 0.89, + "learning_rate": 1.9265619917203975e-05, + "loss": 0.9888, + "step": 12104 + }, + { + "epoch": 0.89, + "learning_rate": 1.926546987742656e-05, + "loss": 1.1384, + "step": 12105 + }, + { + "epoch": 0.89, + "learning_rate": 1.9265319822907916e-05, + "loss": 1.041, + "step": 12106 + }, + { + "epoch": 0.89, + "learning_rate": 1.926516975364828e-05, + "loss": 1.0914, + "step": 12107 + }, + { + "epoch": 0.89, + "learning_rate": 1.926501966964789e-05, + "loss": 0.9549, + "step": 12108 + }, + { + "epoch": 0.89, + "learning_rate": 1.926486957090699e-05, + "loss": 1.0414, + "step": 12109 + }, + { + "epoch": 0.89, + "learning_rate": 1.9264719457425812e-05, + "loss": 0.9715, + "step": 12110 + }, + { + "epoch": 0.89, + "learning_rate": 1.92645693292046e-05, + "loss": 1.0868, + "step": 12111 + }, + { + "epoch": 0.89, + "learning_rate": 1.9264419186243592e-05, + "loss": 1.0462, + "step": 12112 + }, + { + "epoch": 0.9, + "learning_rate": 1.9264269028543018e-05, + "loss": 0.9644, + "step": 12113 + }, + { + "epoch": 0.9, + "learning_rate": 1.9264118856103132e-05, + "loss": 1.113, + "step": 12114 + }, + { + "epoch": 0.9, + "learning_rate": 1.926396866892416e-05, + "loss": 0.9327, + "step": 12115 + }, + { + "epoch": 0.9, + "learning_rate": 1.9263818467006353e-05, + "loss": 1.0325, + "step": 12116 + }, + { + "epoch": 0.9, + "learning_rate": 1.926366825034994e-05, + "loss": 0.9216, + "step": 12117 + }, + { + "epoch": 0.9, + "learning_rate": 1.9263518018955163e-05, + "loss": 1.0128, + "step": 12118 + }, + { + "epoch": 0.9, + "learning_rate": 1.9263367772822262e-05, + "loss": 1.0118, + "step": 12119 + }, + { + "epoch": 0.9, + "learning_rate": 1.9263217511951476e-05, + "loss": 0.9771, + "step": 12120 + }, + { + "epoch": 0.9, + "learning_rate": 1.926306723634304e-05, + "loss": 1.054, + "step": 12121 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262916945997202e-05, + "loss": 1.0596, + "step": 12122 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262766640914193e-05, + "loss": 1.052, + "step": 12123 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262616321094254e-05, + "loss": 1.0636, + "step": 12124 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262465986537625e-05, + "loss": 1.0229, + "step": 12125 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262315637244548e-05, + "loss": 0.9678, + "step": 12126 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262165273215255e-05, + "loss": 1.0261, + "step": 12127 + }, + { + "epoch": 0.9, + "learning_rate": 1.9262014894449992e-05, + "loss": 1.0565, + "step": 12128 + }, + { + "epoch": 0.9, + "learning_rate": 1.9261864500948998e-05, + "loss": 1.0428, + "step": 12129 + }, + { + "epoch": 0.9, + "learning_rate": 1.9261714092712503e-05, + "loss": 1.0852, + "step": 12130 + }, + { + "epoch": 0.9, + "learning_rate": 1.926156366974076e-05, + "loss": 0.9844, + "step": 12131 + }, + { + "epoch": 0.9, + "learning_rate": 1.9261413232034e-05, + "loss": 0.9501, + "step": 12132 + }, + { + "epoch": 0.9, + "learning_rate": 1.9261262779592466e-05, + "loss": 1.1389, + "step": 12133 + }, + { + "epoch": 0.9, + "learning_rate": 1.926111231241639e-05, + "loss": 1.1165, + "step": 12134 + }, + { + "epoch": 0.9, + "learning_rate": 1.926096183050602e-05, + "loss": 1.0321, + "step": 12135 + }, + { + "epoch": 0.9, + "learning_rate": 1.9260811333861594e-05, + "loss": 1.1242, + "step": 12136 + }, + { + "epoch": 0.9, + "learning_rate": 1.9260660822483348e-05, + "loss": 0.9486, + "step": 12137 + }, + { + "epoch": 0.9, + "learning_rate": 1.9260510296371523e-05, + "loss": 1.0035, + "step": 12138 + }, + { + "epoch": 0.9, + "learning_rate": 1.926035975552636e-05, + "loss": 1.0446, + "step": 12139 + }, + { + "epoch": 0.9, + "learning_rate": 1.9260209199948094e-05, + "loss": 0.9029, + "step": 12140 + }, + { + "epoch": 0.9, + "learning_rate": 1.926005862963697e-05, + "loss": 1.0436, + "step": 12141 + }, + { + "epoch": 0.9, + "learning_rate": 1.9259908044593225e-05, + "loss": 1.0324, + "step": 12142 + }, + { + "epoch": 0.9, + "learning_rate": 1.92597574448171e-05, + "loss": 0.9797, + "step": 12143 + }, + { + "epoch": 0.9, + "learning_rate": 1.925960683030883e-05, + "loss": 1.0702, + "step": 12144 + }, + { + "epoch": 0.9, + "learning_rate": 1.925945620106866e-05, + "loss": 1.1368, + "step": 12145 + }, + { + "epoch": 0.9, + "learning_rate": 1.925930555709683e-05, + "loss": 1.0625, + "step": 12146 + }, + { + "epoch": 0.9, + "learning_rate": 1.9259154898393574e-05, + "loss": 1.0805, + "step": 12147 + }, + { + "epoch": 0.9, + "learning_rate": 1.9259004224959134e-05, + "loss": 1.0495, + "step": 12148 + }, + { + "epoch": 0.9, + "learning_rate": 1.9258853536793755e-05, + "loss": 1.0096, + "step": 12149 + }, + { + "epoch": 0.9, + "learning_rate": 1.9258702833897666e-05, + "loss": 1.0142, + "step": 12150 + }, + { + "epoch": 0.9, + "learning_rate": 1.925855211627112e-05, + "loss": 0.8924, + "step": 12151 + }, + { + "epoch": 0.9, + "learning_rate": 1.9258401383914347e-05, + "loss": 1.0529, + "step": 12152 + }, + { + "epoch": 0.9, + "learning_rate": 1.9258250636827588e-05, + "loss": 1.1209, + "step": 12153 + }, + { + "epoch": 0.9, + "learning_rate": 1.9258099875011088e-05, + "loss": 1.0767, + "step": 12154 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257949098465085e-05, + "loss": 1.0611, + "step": 12155 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257798307189815e-05, + "loss": 1.1031, + "step": 12156 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257647501185517e-05, + "loss": 1.0558, + "step": 12157 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257496680452438e-05, + "loss": 0.9676, + "step": 12158 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257345844990814e-05, + "loss": 0.9374, + "step": 12159 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257194994800887e-05, + "loss": 0.9565, + "step": 12160 + }, + { + "epoch": 0.9, + "learning_rate": 1.9257044129882893e-05, + "loss": 0.8742, + "step": 12161 + }, + { + "epoch": 0.9, + "learning_rate": 1.9256893250237073e-05, + "loss": 1.0557, + "step": 12162 + }, + { + "epoch": 0.9, + "learning_rate": 1.925674235586367e-05, + "loss": 0.9761, + "step": 12163 + }, + { + "epoch": 0.9, + "learning_rate": 1.9256591446762924e-05, + "loss": 0.9101, + "step": 12164 + }, + { + "epoch": 0.9, + "learning_rate": 1.9256440522935072e-05, + "loss": 1.026, + "step": 12165 + }, + { + "epoch": 0.9, + "learning_rate": 1.9256289584380353e-05, + "loss": 1.0485, + "step": 12166 + }, + { + "epoch": 0.9, + "learning_rate": 1.925613863109901e-05, + "loss": 1.0328, + "step": 12167 + }, + { + "epoch": 0.9, + "learning_rate": 1.925598766309129e-05, + "loss": 1.0755, + "step": 12168 + }, + { + "epoch": 0.9, + "learning_rate": 1.925583668035742e-05, + "loss": 1.064, + "step": 12169 + }, + { + "epoch": 0.9, + "learning_rate": 1.9255685682897645e-05, + "loss": 1.0839, + "step": 12170 + }, + { + "epoch": 0.9, + "learning_rate": 1.925553467071221e-05, + "loss": 1.0791, + "step": 12171 + }, + { + "epoch": 0.9, + "learning_rate": 1.925538364380135e-05, + "loss": 1.1413, + "step": 12172 + }, + { + "epoch": 0.9, + "learning_rate": 1.9255232602165306e-05, + "loss": 0.9894, + "step": 12173 + }, + { + "epoch": 0.9, + "learning_rate": 1.9255081545804324e-05, + "loss": 0.9524, + "step": 12174 + }, + { + "epoch": 0.9, + "learning_rate": 1.9254930474718632e-05, + "loss": 0.992, + "step": 12175 + }, + { + "epoch": 0.9, + "learning_rate": 1.9254779388908486e-05, + "loss": 1.0497, + "step": 12176 + }, + { + "epoch": 0.9, + "learning_rate": 1.9254628288374112e-05, + "loss": 1.0464, + "step": 12177 + }, + { + "epoch": 0.9, + "learning_rate": 1.925447717311576e-05, + "loss": 1.0457, + "step": 12178 + }, + { + "epoch": 0.9, + "learning_rate": 1.925432604313367e-05, + "loss": 1.0764, + "step": 12179 + }, + { + "epoch": 0.9, + "learning_rate": 1.925417489842808e-05, + "loss": 0.9776, + "step": 12180 + }, + { + "epoch": 0.9, + "learning_rate": 1.9254023738999223e-05, + "loss": 1.1376, + "step": 12181 + }, + { + "epoch": 0.9, + "learning_rate": 1.9253872564847352e-05, + "loss": 1.0962, + "step": 12182 + }, + { + "epoch": 0.9, + "learning_rate": 1.92537213759727e-05, + "loss": 1.0814, + "step": 12183 + }, + { + "epoch": 0.9, + "learning_rate": 1.9253570172375507e-05, + "loss": 0.9878, + "step": 12184 + }, + { + "epoch": 0.9, + "learning_rate": 1.925341895405602e-05, + "loss": 0.977, + "step": 12185 + }, + { + "epoch": 0.9, + "learning_rate": 1.9253267721014476e-05, + "loss": 1.089, + "step": 12186 + }, + { + "epoch": 0.9, + "learning_rate": 1.9253116473251115e-05, + "loss": 1.0138, + "step": 12187 + }, + { + "epoch": 0.9, + "learning_rate": 1.925296521076618e-05, + "loss": 1.063, + "step": 12188 + }, + { + "epoch": 0.9, + "learning_rate": 1.9252813933559906e-05, + "loss": 1.0644, + "step": 12189 + }, + { + "epoch": 0.9, + "learning_rate": 1.925266264163254e-05, + "loss": 0.9672, + "step": 12190 + }, + { + "epoch": 0.9, + "learning_rate": 1.9252511334984318e-05, + "loss": 0.9741, + "step": 12191 + }, + { + "epoch": 0.9, + "learning_rate": 1.925236001361548e-05, + "loss": 1.0858, + "step": 12192 + }, + { + "epoch": 0.9, + "learning_rate": 1.9252208677526276e-05, + "loss": 0.9228, + "step": 12193 + }, + { + "epoch": 0.9, + "learning_rate": 1.925205732671694e-05, + "loss": 0.9217, + "step": 12194 + }, + { + "epoch": 0.9, + "learning_rate": 1.925190596118771e-05, + "loss": 0.9437, + "step": 12195 + }, + { + "epoch": 0.9, + "learning_rate": 1.9251754580938833e-05, + "loss": 1.0174, + "step": 12196 + }, + { + "epoch": 0.9, + "learning_rate": 1.925160318597054e-05, + "loss": 1.0041, + "step": 12197 + }, + { + "epoch": 0.9, + "learning_rate": 1.9251451776283087e-05, + "loss": 1.0084, + "step": 12198 + }, + { + "epoch": 0.9, + "learning_rate": 1.9251300351876704e-05, + "loss": 0.9807, + "step": 12199 + }, + { + "epoch": 0.9, + "learning_rate": 1.9251148912751632e-05, + "loss": 0.9922, + "step": 12200 + }, + { + "epoch": 0.9, + "learning_rate": 1.925099745890812e-05, + "loss": 1.0396, + "step": 12201 + }, + { + "epoch": 0.9, + "learning_rate": 1.9250845990346397e-05, + "loss": 1.1993, + "step": 12202 + }, + { + "epoch": 0.9, + "learning_rate": 1.9250694507066715e-05, + "loss": 1.0517, + "step": 12203 + }, + { + "epoch": 0.9, + "learning_rate": 1.9250543009069306e-05, + "loss": 1.0268, + "step": 12204 + }, + { + "epoch": 0.9, + "learning_rate": 1.925039149635442e-05, + "loss": 1.0173, + "step": 12205 + }, + { + "epoch": 0.9, + "learning_rate": 1.925023996892229e-05, + "loss": 1.0356, + "step": 12206 + }, + { + "epoch": 0.9, + "learning_rate": 1.925008842677316e-05, + "loss": 1.0439, + "step": 12207 + }, + { + "epoch": 0.9, + "learning_rate": 1.9249936869907274e-05, + "loss": 0.951, + "step": 12208 + }, + { + "epoch": 0.9, + "learning_rate": 1.924978529832487e-05, + "loss": 0.9236, + "step": 12209 + }, + { + "epoch": 0.9, + "learning_rate": 1.924963371202619e-05, + "loss": 1.0225, + "step": 12210 + }, + { + "epoch": 0.9, + "learning_rate": 1.9249482111011474e-05, + "loss": 1.065, + "step": 12211 + }, + { + "epoch": 0.9, + "learning_rate": 1.9249330495280968e-05, + "loss": 1.1108, + "step": 12212 + }, + { + "epoch": 0.9, + "learning_rate": 1.924917886483491e-05, + "loss": 1.0122, + "step": 12213 + }, + { + "epoch": 0.9, + "learning_rate": 1.9249027219673535e-05, + "loss": 1.0688, + "step": 12214 + }, + { + "epoch": 0.9, + "learning_rate": 1.9248875559797095e-05, + "loss": 1.0843, + "step": 12215 + }, + { + "epoch": 0.9, + "learning_rate": 1.9248723885205822e-05, + "loss": 1.0015, + "step": 12216 + }, + { + "epoch": 0.9, + "learning_rate": 1.9248572195899965e-05, + "loss": 1.0484, + "step": 12217 + }, + { + "epoch": 0.9, + "learning_rate": 1.924842049187976e-05, + "loss": 0.981, + "step": 12218 + }, + { + "epoch": 0.9, + "learning_rate": 1.9248268773145453e-05, + "loss": 1.1114, + "step": 12219 + }, + { + "epoch": 0.9, + "learning_rate": 1.924811703969728e-05, + "loss": 0.9909, + "step": 12220 + }, + { + "epoch": 0.9, + "learning_rate": 1.9247965291535483e-05, + "loss": 1.0173, + "step": 12221 + }, + { + "epoch": 0.9, + "learning_rate": 1.924781352866031e-05, + "loss": 0.8906, + "step": 12222 + }, + { + "epoch": 0.9, + "learning_rate": 1.9247661751071994e-05, + "loss": 1.0856, + "step": 12223 + }, + { + "epoch": 0.9, + "learning_rate": 1.9247509958770785e-05, + "loss": 1.0195, + "step": 12224 + }, + { + "epoch": 0.9, + "learning_rate": 1.924735815175692e-05, + "loss": 1.0965, + "step": 12225 + }, + { + "epoch": 0.9, + "learning_rate": 1.9247206330030637e-05, + "loss": 0.9609, + "step": 12226 + }, + { + "epoch": 0.9, + "learning_rate": 1.9247054493592182e-05, + "loss": 1.0179, + "step": 12227 + }, + { + "epoch": 0.9, + "learning_rate": 1.9246902642441798e-05, + "loss": 1.0244, + "step": 12228 + }, + { + "epoch": 0.9, + "learning_rate": 1.924675077657972e-05, + "loss": 1.0749, + "step": 12229 + }, + { + "epoch": 0.9, + "learning_rate": 1.9246598896006198e-05, + "loss": 1.129, + "step": 12230 + }, + { + "epoch": 0.9, + "learning_rate": 1.9246447000721467e-05, + "loss": 1.0616, + "step": 12231 + }, + { + "epoch": 0.9, + "learning_rate": 1.9246295090725772e-05, + "loss": 0.9109, + "step": 12232 + }, + { + "epoch": 0.9, + "learning_rate": 1.924614316601935e-05, + "loss": 1.0129, + "step": 12233 + }, + { + "epoch": 0.9, + "learning_rate": 1.9245991226602452e-05, + "loss": 1.1514, + "step": 12234 + }, + { + "epoch": 0.9, + "learning_rate": 1.924583927247531e-05, + "loss": 0.9858, + "step": 12235 + }, + { + "epoch": 0.9, + "learning_rate": 1.9245687303638173e-05, + "loss": 1.0275, + "step": 12236 + }, + { + "epoch": 0.9, + "learning_rate": 1.924553532009128e-05, + "loss": 0.9805, + "step": 12237 + }, + { + "epoch": 0.9, + "learning_rate": 1.924538332183487e-05, + "loss": 0.985, + "step": 12238 + }, + { + "epoch": 0.9, + "learning_rate": 1.924523130886919e-05, + "loss": 1.0747, + "step": 12239 + }, + { + "epoch": 0.9, + "learning_rate": 1.924507928119448e-05, + "loss": 1.21, + "step": 12240 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244927238810978e-05, + "loss": 1.071, + "step": 12241 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244775181718932e-05, + "loss": 1.029, + "step": 12242 + }, + { + "epoch": 0.9, + "learning_rate": 1.924462310991858e-05, + "loss": 1.0712, + "step": 12243 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244471023410164e-05, + "loss": 1.0865, + "step": 12244 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244318922193925e-05, + "loss": 0.9897, + "step": 12245 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244166806270108e-05, + "loss": 0.9742, + "step": 12246 + }, + { + "epoch": 0.9, + "learning_rate": 1.9244014675638957e-05, + "loss": 1.0411, + "step": 12247 + }, + { + "epoch": 0.9, + "learning_rate": 1.924386253030071e-05, + "loss": 0.9763, + "step": 12248 + }, + { + "epoch": 0.91, + "learning_rate": 1.924371037025561e-05, + "loss": 1.0568, + "step": 12249 + }, + { + "epoch": 0.91, + "learning_rate": 1.9243558195503896e-05, + "loss": 1.0399, + "step": 12250 + }, + { + "epoch": 0.91, + "learning_rate": 1.9243406006045812e-05, + "loss": 1.1389, + "step": 12251 + }, + { + "epoch": 0.91, + "learning_rate": 1.9243253801881606e-05, + "loss": 1.0806, + "step": 12252 + }, + { + "epoch": 0.91, + "learning_rate": 1.9243101583011515e-05, + "loss": 1.0392, + "step": 12253 + }, + { + "epoch": 0.91, + "learning_rate": 1.924294934943578e-05, + "loss": 1.0661, + "step": 12254 + }, + { + "epoch": 0.91, + "learning_rate": 1.9242797101154643e-05, + "loss": 1.0317, + "step": 12255 + }, + { + "epoch": 0.91, + "learning_rate": 1.9242644838168352e-05, + "loss": 1.0443, + "step": 12256 + }, + { + "epoch": 0.91, + "learning_rate": 1.9242492560477143e-05, + "loss": 1.1241, + "step": 12257 + }, + { + "epoch": 0.91, + "learning_rate": 1.9242340268081262e-05, + "loss": 1.0932, + "step": 12258 + }, + { + "epoch": 0.91, + "learning_rate": 1.924218796098095e-05, + "loss": 1.0329, + "step": 12259 + }, + { + "epoch": 0.91, + "learning_rate": 1.9242035639176445e-05, + "loss": 1.0304, + "step": 12260 + }, + { + "epoch": 0.91, + "learning_rate": 1.9241883302668e-05, + "loss": 1.0371, + "step": 12261 + }, + { + "epoch": 0.91, + "learning_rate": 1.9241730951455844e-05, + "loss": 1.0558, + "step": 12262 + }, + { + "epoch": 0.91, + "learning_rate": 1.924157858554023e-05, + "loss": 0.9345, + "step": 12263 + }, + { + "epoch": 0.91, + "learning_rate": 1.92414262049214e-05, + "loss": 0.9954, + "step": 12264 + }, + { + "epoch": 0.91, + "learning_rate": 1.9241273809599588e-05, + "loss": 0.9787, + "step": 12265 + }, + { + "epoch": 0.91, + "learning_rate": 1.9241121399575045e-05, + "loss": 1.0225, + "step": 12266 + }, + { + "epoch": 0.91, + "learning_rate": 1.9240968974848007e-05, + "loss": 1.1687, + "step": 12267 + }, + { + "epoch": 0.91, + "learning_rate": 1.924081653541872e-05, + "loss": 1.0417, + "step": 12268 + }, + { + "epoch": 0.91, + "learning_rate": 1.924066408128743e-05, + "loss": 1.168, + "step": 12269 + }, + { + "epoch": 0.91, + "learning_rate": 1.9240511612454372e-05, + "loss": 0.9599, + "step": 12270 + }, + { + "epoch": 0.91, + "learning_rate": 1.924035912891979e-05, + "loss": 1.0257, + "step": 12271 + }, + { + "epoch": 0.91, + "learning_rate": 1.9240206630683933e-05, + "loss": 1.0066, + "step": 12272 + }, + { + "epoch": 0.91, + "learning_rate": 1.924005411774704e-05, + "loss": 1.077, + "step": 12273 + }, + { + "epoch": 0.91, + "learning_rate": 1.923990159010935e-05, + "loss": 0.9316, + "step": 12274 + }, + { + "epoch": 0.91, + "learning_rate": 1.923974904777111e-05, + "loss": 1.0228, + "step": 12275 + }, + { + "epoch": 0.91, + "learning_rate": 1.923959649073256e-05, + "loss": 1.0272, + "step": 12276 + }, + { + "epoch": 0.91, + "learning_rate": 1.9239443918993944e-05, + "loss": 1.1085, + "step": 12277 + }, + { + "epoch": 0.91, + "learning_rate": 1.923929133255551e-05, + "loss": 1.0181, + "step": 12278 + }, + { + "epoch": 0.91, + "learning_rate": 1.923913873141749e-05, + "loss": 0.9948, + "step": 12279 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238986115580138e-05, + "loss": 0.9503, + "step": 12280 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238833485043686e-05, + "loss": 1.0136, + "step": 12281 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238680839808384e-05, + "loss": 1.123, + "step": 12282 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238528179874475e-05, + "loss": 1.0844, + "step": 12283 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238375505242197e-05, + "loss": 0.9298, + "step": 12284 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238222815911796e-05, + "loss": 1.0293, + "step": 12285 + }, + { + "epoch": 0.91, + "learning_rate": 1.9238070111883517e-05, + "loss": 1.1091, + "step": 12286 + }, + { + "epoch": 0.91, + "learning_rate": 1.92379173931576e-05, + "loss": 0.944, + "step": 12287 + }, + { + "epoch": 0.91, + "learning_rate": 1.9237764659734287e-05, + "loss": 1.0945, + "step": 12288 + }, + { + "epoch": 0.91, + "learning_rate": 1.9237611911613823e-05, + "loss": 1.0789, + "step": 12289 + }, + { + "epoch": 0.91, + "learning_rate": 1.923745914879645e-05, + "loss": 0.967, + "step": 12290 + }, + { + "epoch": 0.91, + "learning_rate": 1.9237306371282414e-05, + "loss": 0.9455, + "step": 12291 + }, + { + "epoch": 0.91, + "learning_rate": 1.923715357907195e-05, + "loss": 1.0139, + "step": 12292 + }, + { + "epoch": 0.91, + "learning_rate": 1.9237000772165313e-05, + "loss": 1.0371, + "step": 12293 + }, + { + "epoch": 0.91, + "learning_rate": 1.923684795056274e-05, + "loss": 0.972, + "step": 12294 + }, + { + "epoch": 0.91, + "learning_rate": 1.923669511426447e-05, + "loss": 1.0114, + "step": 12295 + }, + { + "epoch": 0.91, + "learning_rate": 1.9236542263270755e-05, + "loss": 1.0875, + "step": 12296 + }, + { + "epoch": 0.91, + "learning_rate": 1.923638939758183e-05, + "loss": 0.9747, + "step": 12297 + }, + { + "epoch": 0.91, + "learning_rate": 1.923623651719794e-05, + "loss": 1.0109, + "step": 12298 + }, + { + "epoch": 0.91, + "learning_rate": 1.9236083622119334e-05, + "loss": 1.0056, + "step": 12299 + }, + { + "epoch": 0.91, + "learning_rate": 1.923593071234625e-05, + "loss": 0.9641, + "step": 12300 + }, + { + "epoch": 0.91, + "learning_rate": 1.923577778787893e-05, + "loss": 0.962, + "step": 12301 + }, + { + "epoch": 0.91, + "learning_rate": 1.9235624848717623e-05, + "loss": 1.1264, + "step": 12302 + }, + { + "epoch": 0.91, + "learning_rate": 1.9235471894862566e-05, + "loss": 1.0494, + "step": 12303 + }, + { + "epoch": 0.91, + "learning_rate": 1.9235318926314006e-05, + "loss": 0.9095, + "step": 12304 + }, + { + "epoch": 0.91, + "learning_rate": 1.9235165943072188e-05, + "loss": 1.0233, + "step": 12305 + }, + { + "epoch": 0.91, + "learning_rate": 1.923501294513735e-05, + "loss": 0.9831, + "step": 12306 + }, + { + "epoch": 0.91, + "learning_rate": 1.923485993250974e-05, + "loss": 1.0747, + "step": 12307 + }, + { + "epoch": 0.91, + "learning_rate": 1.9234706905189603e-05, + "loss": 1.0701, + "step": 12308 + }, + { + "epoch": 0.91, + "learning_rate": 1.9234553863177177e-05, + "loss": 1.1445, + "step": 12309 + }, + { + "epoch": 0.91, + "learning_rate": 1.9234400806472707e-05, + "loss": 1.0006, + "step": 12310 + }, + { + "epoch": 0.91, + "learning_rate": 1.9234247735076437e-05, + "loss": 1.0038, + "step": 12311 + }, + { + "epoch": 0.91, + "learning_rate": 1.9234094648988614e-05, + "loss": 1.0284, + "step": 12312 + }, + { + "epoch": 0.91, + "learning_rate": 1.9233941548209472e-05, + "loss": 1.051, + "step": 12313 + }, + { + "epoch": 0.91, + "learning_rate": 1.923378843273927e-05, + "loss": 0.9921, + "step": 12314 + }, + { + "epoch": 0.91, + "learning_rate": 1.9233635302578235e-05, + "loss": 0.9364, + "step": 12315 + }, + { + "epoch": 0.91, + "learning_rate": 1.923348215772662e-05, + "loss": 0.974, + "step": 12316 + }, + { + "epoch": 0.91, + "learning_rate": 1.923332899818467e-05, + "loss": 1.0066, + "step": 12317 + }, + { + "epoch": 0.91, + "learning_rate": 1.9233175823952623e-05, + "loss": 1.0393, + "step": 12318 + }, + { + "epoch": 0.91, + "learning_rate": 1.9233022635030728e-05, + "loss": 1.0148, + "step": 12319 + }, + { + "epoch": 0.91, + "learning_rate": 1.9232869431419225e-05, + "loss": 1.0204, + "step": 12320 + }, + { + "epoch": 0.91, + "learning_rate": 1.923271621311836e-05, + "loss": 0.9741, + "step": 12321 + }, + { + "epoch": 0.91, + "learning_rate": 1.923256298012837e-05, + "loss": 1.0155, + "step": 12322 + }, + { + "epoch": 0.91, + "learning_rate": 1.9232409732449512e-05, + "loss": 1.0236, + "step": 12323 + }, + { + "epoch": 0.91, + "learning_rate": 1.9232256470082017e-05, + "loss": 1.0632, + "step": 12324 + }, + { + "epoch": 0.91, + "learning_rate": 1.9232103193026137e-05, + "loss": 1.1758, + "step": 12325 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231949901282113e-05, + "loss": 0.8883, + "step": 12326 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231796594850185e-05, + "loss": 0.9631, + "step": 12327 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231643273730604e-05, + "loss": 0.9852, + "step": 12328 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231489937923608e-05, + "loss": 1.0426, + "step": 12329 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231336587429446e-05, + "loss": 0.9915, + "step": 12330 + }, + { + "epoch": 0.91, + "learning_rate": 1.923118322224836e-05, + "loss": 0.9787, + "step": 12331 + }, + { + "epoch": 0.91, + "learning_rate": 1.9231029842380594e-05, + "loss": 0.9762, + "step": 12332 + }, + { + "epoch": 0.91, + "learning_rate": 1.923087644782639e-05, + "loss": 1.0482, + "step": 12333 + }, + { + "epoch": 0.91, + "learning_rate": 1.923072303858599e-05, + "loss": 1.0504, + "step": 12334 + }, + { + "epoch": 0.91, + "learning_rate": 1.923056961465965e-05, + "loss": 0.9872, + "step": 12335 + }, + { + "epoch": 0.91, + "learning_rate": 1.92304161760476e-05, + "loss": 1.0312, + "step": 12336 + }, + { + "epoch": 0.91, + "learning_rate": 1.923026272275009e-05, + "loss": 0.9778, + "step": 12337 + }, + { + "epoch": 0.91, + "learning_rate": 1.9230109254767367e-05, + "loss": 1.0249, + "step": 12338 + }, + { + "epoch": 0.91, + "learning_rate": 1.922995577209967e-05, + "loss": 1.0534, + "step": 12339 + }, + { + "epoch": 0.91, + "learning_rate": 1.9229802274747245e-05, + "loss": 0.9915, + "step": 12340 + }, + { + "epoch": 0.91, + "learning_rate": 1.922964876271034e-05, + "loss": 1.0368, + "step": 12341 + }, + { + "epoch": 0.91, + "learning_rate": 1.9229495235989192e-05, + "loss": 0.9431, + "step": 12342 + }, + { + "epoch": 0.91, + "learning_rate": 1.922934169458405e-05, + "loss": 1.0684, + "step": 12343 + }, + { + "epoch": 0.91, + "learning_rate": 1.922918813849516e-05, + "loss": 0.9936, + "step": 12344 + }, + { + "epoch": 0.91, + "learning_rate": 1.922903456772276e-05, + "loss": 1.0002, + "step": 12345 + }, + { + "epoch": 0.91, + "learning_rate": 1.92288809822671e-05, + "loss": 1.0585, + "step": 12346 + }, + { + "epoch": 0.91, + "learning_rate": 1.9228727382128423e-05, + "loss": 1.0056, + "step": 12347 + }, + { + "epoch": 0.91, + "learning_rate": 1.922857376730697e-05, + "loss": 1.1092, + "step": 12348 + }, + { + "epoch": 0.91, + "learning_rate": 1.9228420137802992e-05, + "loss": 1.0527, + "step": 12349 + }, + { + "epoch": 0.91, + "learning_rate": 1.9228266493616728e-05, + "loss": 0.9965, + "step": 12350 + }, + { + "epoch": 0.91, + "learning_rate": 1.922811283474842e-05, + "loss": 1.1368, + "step": 12351 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227959161198323e-05, + "loss": 1.0538, + "step": 12352 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227805472966673e-05, + "loss": 1.0002, + "step": 12353 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227651770053714e-05, + "loss": 1.0918, + "step": 12354 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227498052459698e-05, + "loss": 0.9804, + "step": 12355 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227344320184863e-05, + "loss": 1.0007, + "step": 12356 + }, + { + "epoch": 0.91, + "learning_rate": 1.922719057322945e-05, + "loss": 1.0928, + "step": 12357 + }, + { + "epoch": 0.91, + "learning_rate": 1.9227036811593715e-05, + "loss": 0.9754, + "step": 12358 + }, + { + "epoch": 0.91, + "learning_rate": 1.9226883035277893e-05, + "loss": 0.9745, + "step": 12359 + }, + { + "epoch": 0.91, + "learning_rate": 1.9226729244282235e-05, + "loss": 1.0823, + "step": 12360 + }, + { + "epoch": 0.91, + "learning_rate": 1.922657543860698e-05, + "loss": 0.9771, + "step": 12361 + }, + { + "epoch": 0.91, + "learning_rate": 1.9226421618252374e-05, + "loss": 1.0076, + "step": 12362 + }, + { + "epoch": 0.91, + "learning_rate": 1.922626778321867e-05, + "loss": 1.0278, + "step": 12363 + }, + { + "epoch": 0.91, + "learning_rate": 1.9226113933506098e-05, + "loss": 1.0208, + "step": 12364 + }, + { + "epoch": 0.91, + "learning_rate": 1.9225960069114916e-05, + "loss": 0.9767, + "step": 12365 + }, + { + "epoch": 0.91, + "learning_rate": 1.922580619004536e-05, + "loss": 1.1322, + "step": 12366 + }, + { + "epoch": 0.91, + "learning_rate": 1.922565229629768e-05, + "loss": 1.0257, + "step": 12367 + }, + { + "epoch": 0.91, + "learning_rate": 1.922549838787212e-05, + "loss": 0.9751, + "step": 12368 + }, + { + "epoch": 0.91, + "learning_rate": 1.9225344464768924e-05, + "loss": 0.9507, + "step": 12369 + }, + { + "epoch": 0.91, + "learning_rate": 1.9225190526988334e-05, + "loss": 1.0117, + "step": 12370 + }, + { + "epoch": 0.91, + "learning_rate": 1.92250365745306e-05, + "loss": 0.9254, + "step": 12371 + }, + { + "epoch": 0.91, + "learning_rate": 1.9224882607395966e-05, + "loss": 1.024, + "step": 12372 + }, + { + "epoch": 0.91, + "learning_rate": 1.9224728625584673e-05, + "loss": 1.0607, + "step": 12373 + }, + { + "epoch": 0.91, + "learning_rate": 1.9224574629096975e-05, + "loss": 1.0587, + "step": 12374 + }, + { + "epoch": 0.91, + "learning_rate": 1.9224420617933103e-05, + "loss": 1.0052, + "step": 12375 + }, + { + "epoch": 0.91, + "learning_rate": 1.9224266592093315e-05, + "loss": 1.0026, + "step": 12376 + }, + { + "epoch": 0.91, + "learning_rate": 1.922411255157785e-05, + "loss": 1.084, + "step": 12377 + }, + { + "epoch": 0.91, + "learning_rate": 1.922395849638695e-05, + "loss": 1.0329, + "step": 12378 + }, + { + "epoch": 0.91, + "learning_rate": 1.9223804426520867e-05, + "loss": 0.8774, + "step": 12379 + }, + { + "epoch": 0.91, + "learning_rate": 1.9223650341979846e-05, + "loss": 1.058, + "step": 12380 + }, + { + "epoch": 0.91, + "learning_rate": 1.9223496242764127e-05, + "loss": 1.1082, + "step": 12381 + }, + { + "epoch": 0.91, + "learning_rate": 1.9223342128873953e-05, + "loss": 0.9844, + "step": 12382 + }, + { + "epoch": 0.91, + "learning_rate": 1.9223188000309577e-05, + "loss": 1.038, + "step": 12383 + }, + { + "epoch": 0.92, + "learning_rate": 1.9223033857071242e-05, + "loss": 1.0683, + "step": 12384 + }, + { + "epoch": 0.92, + "learning_rate": 1.922287969915919e-05, + "loss": 0.911, + "step": 12385 + }, + { + "epoch": 0.92, + "learning_rate": 1.922272552657367e-05, + "loss": 1.0816, + "step": 12386 + }, + { + "epoch": 0.92, + "learning_rate": 1.9222571339314926e-05, + "loss": 0.988, + "step": 12387 + }, + { + "epoch": 0.92, + "learning_rate": 1.9222417137383202e-05, + "loss": 0.9829, + "step": 12388 + }, + { + "epoch": 0.92, + "learning_rate": 1.9222262920778746e-05, + "loss": 0.9473, + "step": 12389 + }, + { + "epoch": 0.92, + "learning_rate": 1.92221086895018e-05, + "loss": 1.1084, + "step": 12390 + }, + { + "epoch": 0.92, + "learning_rate": 1.9221954443552612e-05, + "loss": 1.0141, + "step": 12391 + }, + { + "epoch": 0.92, + "learning_rate": 1.922180018293142e-05, + "loss": 1.0973, + "step": 12392 + }, + { + "epoch": 0.92, + "learning_rate": 1.9221645907638485e-05, + "loss": 1.1416, + "step": 12393 + }, + { + "epoch": 0.92, + "learning_rate": 1.922149161767404e-05, + "loss": 1.0472, + "step": 12394 + }, + { + "epoch": 0.92, + "learning_rate": 1.9221337313038332e-05, + "loss": 1.0387, + "step": 12395 + }, + { + "epoch": 0.92, + "learning_rate": 1.9221182993731612e-05, + "loss": 1.0847, + "step": 12396 + }, + { + "epoch": 0.92, + "learning_rate": 1.922102865975412e-05, + "loss": 1.0469, + "step": 12397 + }, + { + "epoch": 0.92, + "learning_rate": 1.92208743111061e-05, + "loss": 0.986, + "step": 12398 + }, + { + "epoch": 0.92, + "learning_rate": 1.9220719947787807e-05, + "loss": 1.0948, + "step": 12399 + }, + { + "epoch": 0.92, + "learning_rate": 1.9220565569799475e-05, + "loss": 0.9886, + "step": 12400 + }, + { + "epoch": 0.92, + "learning_rate": 1.922041117714136e-05, + "loss": 1.0994, + "step": 12401 + }, + { + "epoch": 0.92, + "learning_rate": 1.92202567698137e-05, + "loss": 1.0206, + "step": 12402 + }, + { + "epoch": 0.92, + "learning_rate": 1.9220102347816744e-05, + "loss": 1.0936, + "step": 12403 + }, + { + "epoch": 0.92, + "learning_rate": 1.9219947911150733e-05, + "loss": 1.1583, + "step": 12404 + }, + { + "epoch": 0.92, + "learning_rate": 1.9219793459815924e-05, + "loss": 1.0553, + "step": 12405 + }, + { + "epoch": 0.92, + "learning_rate": 1.921963899381255e-05, + "loss": 1.0674, + "step": 12406 + }, + { + "epoch": 0.92, + "learning_rate": 1.921948451314086e-05, + "loss": 0.9984, + "step": 12407 + }, + { + "epoch": 0.92, + "learning_rate": 1.921933001780111e-05, + "loss": 1.0624, + "step": 12408 + }, + { + "epoch": 0.92, + "learning_rate": 1.9219175507793533e-05, + "loss": 0.9897, + "step": 12409 + }, + { + "epoch": 0.92, + "learning_rate": 1.921902098311838e-05, + "loss": 1.0547, + "step": 12410 + }, + { + "epoch": 0.92, + "learning_rate": 1.92188664437759e-05, + "loss": 1.0338, + "step": 12411 + }, + { + "epoch": 0.92, + "learning_rate": 1.921871188976633e-05, + "loss": 1.0414, + "step": 12412 + }, + { + "epoch": 0.92, + "learning_rate": 1.9218557321089924e-05, + "loss": 0.9918, + "step": 12413 + }, + { + "epoch": 0.92, + "learning_rate": 1.9218402737746928e-05, + "loss": 1.0639, + "step": 12414 + }, + { + "epoch": 0.92, + "learning_rate": 1.9218248139737583e-05, + "loss": 1.0442, + "step": 12415 + }, + { + "epoch": 0.92, + "learning_rate": 1.9218093527062134e-05, + "loss": 1.1061, + "step": 12416 + }, + { + "epoch": 0.92, + "learning_rate": 1.9217938899720834e-05, + "loss": 0.993, + "step": 12417 + }, + { + "epoch": 0.92, + "learning_rate": 1.9217784257713922e-05, + "loss": 1.1028, + "step": 12418 + }, + { + "epoch": 0.92, + "learning_rate": 1.921762960104165e-05, + "loss": 1.012, + "step": 12419 + }, + { + "epoch": 0.92, + "learning_rate": 1.921747492970426e-05, + "loss": 0.9795, + "step": 12420 + }, + { + "epoch": 0.92, + "learning_rate": 1.9217320243702e-05, + "loss": 1.0193, + "step": 12421 + }, + { + "epoch": 0.92, + "learning_rate": 1.9217165543035113e-05, + "loss": 1.0879, + "step": 12422 + }, + { + "epoch": 0.92, + "learning_rate": 1.9217010827703852e-05, + "loss": 1.1501, + "step": 12423 + }, + { + "epoch": 0.92, + "learning_rate": 1.921685609770845e-05, + "loss": 1.0589, + "step": 12424 + }, + { + "epoch": 0.92, + "learning_rate": 1.9216701353049168e-05, + "loss": 1.1109, + "step": 12425 + }, + { + "epoch": 0.92, + "learning_rate": 1.9216546593726247e-05, + "loss": 1.0092, + "step": 12426 + }, + { + "epoch": 0.92, + "learning_rate": 1.9216391819739933e-05, + "loss": 1.0251, + "step": 12427 + }, + { + "epoch": 0.92, + "learning_rate": 1.921623703109047e-05, + "loss": 0.9882, + "step": 12428 + }, + { + "epoch": 0.92, + "learning_rate": 1.92160822277781e-05, + "loss": 1.0792, + "step": 12429 + }, + { + "epoch": 0.92, + "learning_rate": 1.9215927409803082e-05, + "loss": 1.0874, + "step": 12430 + }, + { + "epoch": 0.92, + "learning_rate": 1.9215772577165655e-05, + "loss": 0.9778, + "step": 12431 + }, + { + "epoch": 0.92, + "learning_rate": 1.9215617729866064e-05, + "loss": 0.9995, + "step": 12432 + }, + { + "epoch": 0.92, + "learning_rate": 1.9215462867904556e-05, + "loss": 1.082, + "step": 12433 + }, + { + "epoch": 0.92, + "learning_rate": 1.921530799128138e-05, + "loss": 0.9668, + "step": 12434 + }, + { + "epoch": 0.92, + "learning_rate": 1.921515309999678e-05, + "loss": 1.0407, + "step": 12435 + }, + { + "epoch": 0.92, + "learning_rate": 1.9214998194051002e-05, + "loss": 1.0734, + "step": 12436 + }, + { + "epoch": 0.92, + "learning_rate": 1.9214843273444298e-05, + "loss": 0.9448, + "step": 12437 + }, + { + "epoch": 0.92, + "learning_rate": 1.9214688338176906e-05, + "loss": 0.9786, + "step": 12438 + }, + { + "epoch": 0.92, + "learning_rate": 1.921453338824908e-05, + "loss": 0.9914, + "step": 12439 + }, + { + "epoch": 0.92, + "learning_rate": 1.921437842366106e-05, + "loss": 1.038, + "step": 12440 + }, + { + "epoch": 0.92, + "learning_rate": 1.9214223444413094e-05, + "loss": 1.0247, + "step": 12441 + }, + { + "epoch": 0.92, + "learning_rate": 1.9214068450505436e-05, + "loss": 0.9673, + "step": 12442 + }, + { + "epoch": 0.92, + "learning_rate": 1.9213913441938327e-05, + "loss": 0.9902, + "step": 12443 + }, + { + "epoch": 0.92, + "learning_rate": 1.921375841871201e-05, + "loss": 1.027, + "step": 12444 + }, + { + "epoch": 0.92, + "learning_rate": 1.9213603380826735e-05, + "loss": 1.0325, + "step": 12445 + }, + { + "epoch": 0.92, + "learning_rate": 1.921344832828275e-05, + "loss": 1.0847, + "step": 12446 + }, + { + "epoch": 0.92, + "learning_rate": 1.92132932610803e-05, + "loss": 0.9149, + "step": 12447 + }, + { + "epoch": 0.92, + "learning_rate": 1.9213138179219634e-05, + "loss": 1.0229, + "step": 12448 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212983082700994e-05, + "loss": 1.0566, + "step": 12449 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212827971524636e-05, + "loss": 1.1229, + "step": 12450 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212672845690793e-05, + "loss": 1.0546, + "step": 12451 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212517705199722e-05, + "loss": 1.0287, + "step": 12452 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212362550051668e-05, + "loss": 0.89, + "step": 12453 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212207380246878e-05, + "loss": 1.1782, + "step": 12454 + }, + { + "epoch": 0.92, + "learning_rate": 1.9212052195785595e-05, + "loss": 1.0365, + "step": 12455 + }, + { + "epoch": 0.92, + "learning_rate": 1.921189699666807e-05, + "loss": 1.0547, + "step": 12456 + }, + { + "epoch": 0.92, + "learning_rate": 1.921174178289455e-05, + "loss": 1.0582, + "step": 12457 + }, + { + "epoch": 0.92, + "learning_rate": 1.921158655446528e-05, + "loss": 0.9069, + "step": 12458 + }, + { + "epoch": 0.92, + "learning_rate": 1.9211431311380508e-05, + "loss": 1.0487, + "step": 12459 + }, + { + "epoch": 0.92, + "learning_rate": 1.9211276053640477e-05, + "loss": 1.0094, + "step": 12460 + }, + { + "epoch": 0.92, + "learning_rate": 1.9211120781245438e-05, + "loss": 1.0103, + "step": 12461 + }, + { + "epoch": 0.92, + "learning_rate": 1.921096549419564e-05, + "loss": 0.9501, + "step": 12462 + }, + { + "epoch": 0.92, + "learning_rate": 1.9210810192491327e-05, + "loss": 1.0293, + "step": 12463 + }, + { + "epoch": 0.92, + "learning_rate": 1.9210654876132747e-05, + "loss": 1.1353, + "step": 12464 + }, + { + "epoch": 0.92, + "learning_rate": 1.9210499545120148e-05, + "loss": 1.051, + "step": 12465 + }, + { + "epoch": 0.92, + "learning_rate": 1.9210344199453777e-05, + "loss": 1.0978, + "step": 12466 + }, + { + "epoch": 0.92, + "learning_rate": 1.9210188839133875e-05, + "loss": 1.0551, + "step": 12467 + }, + { + "epoch": 0.92, + "learning_rate": 1.92100334641607e-05, + "loss": 0.9552, + "step": 12468 + }, + { + "epoch": 0.92, + "learning_rate": 1.9209878074534492e-05, + "loss": 0.9609, + "step": 12469 + }, + { + "epoch": 0.92, + "learning_rate": 1.9209722670255498e-05, + "loss": 1.0936, + "step": 12470 + }, + { + "epoch": 0.92, + "learning_rate": 1.920956725132397e-05, + "loss": 0.9736, + "step": 12471 + }, + { + "epoch": 0.92, + "learning_rate": 1.9209411817740145e-05, + "loss": 0.9617, + "step": 12472 + }, + { + "epoch": 0.92, + "learning_rate": 1.9209256369504285e-05, + "loss": 1.0531, + "step": 12473 + }, + { + "epoch": 0.92, + "learning_rate": 1.9209100906616628e-05, + "loss": 1.0983, + "step": 12474 + }, + { + "epoch": 0.92, + "learning_rate": 1.9208945429077423e-05, + "loss": 0.9715, + "step": 12475 + }, + { + "epoch": 0.92, + "learning_rate": 1.920878993688692e-05, + "loss": 1.0586, + "step": 12476 + }, + { + "epoch": 0.92, + "learning_rate": 1.920863443004536e-05, + "loss": 1.0254, + "step": 12477 + }, + { + "epoch": 0.92, + "learning_rate": 1.9208478908552995e-05, + "loss": 0.9822, + "step": 12478 + }, + { + "epoch": 0.92, + "learning_rate": 1.920832337241007e-05, + "loss": 1.0529, + "step": 12479 + }, + { + "epoch": 0.92, + "learning_rate": 1.920816782161684e-05, + "loss": 0.9686, + "step": 12480 + }, + { + "epoch": 0.92, + "learning_rate": 1.9208012256173544e-05, + "loss": 1.0108, + "step": 12481 + }, + { + "epoch": 0.92, + "learning_rate": 1.920785667608043e-05, + "loss": 0.9591, + "step": 12482 + }, + { + "epoch": 0.92, + "learning_rate": 1.920770108133775e-05, + "loss": 1.0297, + "step": 12483 + }, + { + "epoch": 0.92, + "learning_rate": 1.9207545471945747e-05, + "loss": 1.0751, + "step": 12484 + }, + { + "epoch": 0.92, + "learning_rate": 1.9207389847904675e-05, + "loss": 1.0214, + "step": 12485 + }, + { + "epoch": 0.92, + "learning_rate": 1.9207234209214774e-05, + "loss": 1.0098, + "step": 12486 + }, + { + "epoch": 0.92, + "learning_rate": 1.9207078555876296e-05, + "loss": 1.0597, + "step": 12487 + }, + { + "epoch": 0.92, + "learning_rate": 1.920692288788949e-05, + "loss": 1.0474, + "step": 12488 + }, + { + "epoch": 0.92, + "learning_rate": 1.92067672052546e-05, + "loss": 1.1018, + "step": 12489 + }, + { + "epoch": 0.92, + "learning_rate": 1.9206611507971872e-05, + "loss": 1.028, + "step": 12490 + }, + { + "epoch": 0.92, + "learning_rate": 1.920645579604156e-05, + "loss": 1.046, + "step": 12491 + }, + { + "epoch": 0.92, + "learning_rate": 1.9206300069463908e-05, + "loss": 0.9821, + "step": 12492 + }, + { + "epoch": 0.92, + "learning_rate": 1.9206144328239164e-05, + "loss": 0.9577, + "step": 12493 + }, + { + "epoch": 0.92, + "learning_rate": 1.920598857236758e-05, + "loss": 0.9579, + "step": 12494 + }, + { + "epoch": 0.92, + "learning_rate": 1.9205832801849397e-05, + "loss": 1.0508, + "step": 12495 + }, + { + "epoch": 0.92, + "learning_rate": 1.9205677016684864e-05, + "loss": 1.0032, + "step": 12496 + }, + { + "epoch": 0.92, + "learning_rate": 1.920552121687423e-05, + "loss": 0.9276, + "step": 12497 + }, + { + "epoch": 0.92, + "learning_rate": 1.9205365402417745e-05, + "loss": 0.9782, + "step": 12498 + }, + { + "epoch": 0.92, + "learning_rate": 1.9205209573315655e-05, + "loss": 0.9384, + "step": 12499 + }, + { + "epoch": 0.92, + "learning_rate": 1.9205053729568212e-05, + "loss": 1.0043, + "step": 12500 + }, + { + "epoch": 0.92, + "learning_rate": 1.9204897871175654e-05, + "loss": 0.9644, + "step": 12501 + }, + { + "epoch": 0.92, + "learning_rate": 1.920474199813824e-05, + "loss": 1.0231, + "step": 12502 + }, + { + "epoch": 0.92, + "learning_rate": 1.920458611045621e-05, + "loss": 1.0534, + "step": 12503 + }, + { + "epoch": 0.92, + "learning_rate": 1.9204430208129816e-05, + "loss": 1.0709, + "step": 12504 + }, + { + "epoch": 0.92, + "learning_rate": 1.9204274291159306e-05, + "loss": 1.2265, + "step": 12505 + }, + { + "epoch": 0.92, + "learning_rate": 1.9204118359544924e-05, + "loss": 1.0979, + "step": 12506 + }, + { + "epoch": 0.92, + "learning_rate": 1.920396241328693e-05, + "loss": 0.9646, + "step": 12507 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203806452385552e-05, + "loss": 1.0214, + "step": 12508 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203650476841057e-05, + "loss": 1.039, + "step": 12509 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203494486653683e-05, + "loss": 0.9739, + "step": 12510 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203338481823684e-05, + "loss": 1.0374, + "step": 12511 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203182462351305e-05, + "loss": 1.1118, + "step": 12512 + }, + { + "epoch": 0.92, + "learning_rate": 1.9203026428236792e-05, + "loss": 1.1453, + "step": 12513 + }, + { + "epoch": 0.92, + "learning_rate": 1.9202870379480396e-05, + "loss": 0.9149, + "step": 12514 + }, + { + "epoch": 0.92, + "learning_rate": 1.9202714316082365e-05, + "loss": 1.0866, + "step": 12515 + }, + { + "epoch": 0.92, + "learning_rate": 1.9202558238042946e-05, + "loss": 0.9713, + "step": 12516 + }, + { + "epoch": 0.92, + "learning_rate": 1.9202402145362394e-05, + "loss": 1.057, + "step": 12517 + }, + { + "epoch": 0.92, + "learning_rate": 1.9202246038040947e-05, + "loss": 1.0628, + "step": 12518 + }, + { + "epoch": 0.93, + "learning_rate": 1.920208991607886e-05, + "loss": 0.9693, + "step": 12519 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201933779476374e-05, + "loss": 1.0833, + "step": 12520 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201777628233753e-05, + "loss": 1.0599, + "step": 12521 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201621462351227e-05, + "loss": 0.963, + "step": 12522 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201465281829054e-05, + "loss": 1.0405, + "step": 12523 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201309086667486e-05, + "loss": 1.0449, + "step": 12524 + }, + { + "epoch": 0.93, + "learning_rate": 1.9201152876866763e-05, + "loss": 1.0405, + "step": 12525 + }, + { + "epoch": 0.93, + "learning_rate": 1.9200996652427138e-05, + "loss": 0.9685, + "step": 12526 + }, + { + "epoch": 0.93, + "learning_rate": 1.920084041334886e-05, + "loss": 1.0203, + "step": 12527 + }, + { + "epoch": 0.93, + "learning_rate": 1.920068415963218e-05, + "loss": 1.017, + "step": 12528 + }, + { + "epoch": 0.93, + "learning_rate": 1.9200527891277338e-05, + "loss": 0.9924, + "step": 12529 + }, + { + "epoch": 0.93, + "learning_rate": 1.920037160828459e-05, + "loss": 1.1467, + "step": 12530 + }, + { + "epoch": 0.93, + "learning_rate": 1.920021531065418e-05, + "loss": 0.9748, + "step": 12531 + }, + { + "epoch": 0.93, + "learning_rate": 1.920005899838636e-05, + "loss": 0.9891, + "step": 12532 + }, + { + "epoch": 0.93, + "learning_rate": 1.919990267148138e-05, + "loss": 1.0483, + "step": 12533 + }, + { + "epoch": 0.93, + "learning_rate": 1.9199746329939483e-05, + "loss": 0.9833, + "step": 12534 + }, + { + "epoch": 0.93, + "learning_rate": 1.9199589973760923e-05, + "loss": 1.1516, + "step": 12535 + }, + { + "epoch": 0.93, + "learning_rate": 1.9199433602945948e-05, + "loss": 0.9586, + "step": 12536 + }, + { + "epoch": 0.93, + "learning_rate": 1.9199277217494803e-05, + "loss": 0.7738, + "step": 12537 + }, + { + "epoch": 0.93, + "learning_rate": 1.9199120817407745e-05, + "loss": 0.948, + "step": 12538 + }, + { + "epoch": 0.93, + "learning_rate": 1.919896440268501e-05, + "loss": 0.9912, + "step": 12539 + }, + { + "epoch": 0.93, + "learning_rate": 1.919880797332686e-05, + "loss": 1.0059, + "step": 12540 + }, + { + "epoch": 0.93, + "learning_rate": 1.9198651529333537e-05, + "loss": 1.0512, + "step": 12541 + }, + { + "epoch": 0.93, + "learning_rate": 1.9198495070705288e-05, + "loss": 0.9713, + "step": 12542 + }, + { + "epoch": 0.93, + "learning_rate": 1.919833859744237e-05, + "loss": 1.0442, + "step": 12543 + }, + { + "epoch": 0.93, + "learning_rate": 1.9198182109545022e-05, + "loss": 1.0266, + "step": 12544 + }, + { + "epoch": 0.93, + "learning_rate": 1.91980256070135e-05, + "loss": 0.9696, + "step": 12545 + }, + { + "epoch": 0.93, + "learning_rate": 1.919786908984805e-05, + "loss": 1.014, + "step": 12546 + }, + { + "epoch": 0.93, + "learning_rate": 1.9197712558048925e-05, + "loss": 1.0782, + "step": 12547 + }, + { + "epoch": 0.93, + "learning_rate": 1.9197556011616366e-05, + "loss": 0.9554, + "step": 12548 + }, + { + "epoch": 0.93, + "learning_rate": 1.9197399450550632e-05, + "loss": 1.1116, + "step": 12549 + }, + { + "epoch": 0.93, + "learning_rate": 1.9197242874851964e-05, + "loss": 1.0592, + "step": 12550 + }, + { + "epoch": 0.93, + "learning_rate": 1.9197086284520615e-05, + "loss": 0.9343, + "step": 12551 + }, + { + "epoch": 0.93, + "learning_rate": 1.9196929679556832e-05, + "loss": 1.1134, + "step": 12552 + }, + { + "epoch": 0.93, + "learning_rate": 1.9196773059960868e-05, + "loss": 0.9356, + "step": 12553 + }, + { + "epoch": 0.93, + "learning_rate": 1.9196616425732965e-05, + "loss": 1.0354, + "step": 12554 + }, + { + "epoch": 0.93, + "learning_rate": 1.919645977687338e-05, + "loss": 1.0861, + "step": 12555 + }, + { + "epoch": 0.93, + "learning_rate": 1.919630311338236e-05, + "loss": 1.082, + "step": 12556 + }, + { + "epoch": 0.93, + "learning_rate": 1.9196146435260155e-05, + "loss": 1.0612, + "step": 12557 + }, + { + "epoch": 0.93, + "learning_rate": 1.919598974250701e-05, + "loss": 1.0371, + "step": 12558 + }, + { + "epoch": 0.93, + "learning_rate": 1.9195833035123174e-05, + "loss": 0.9559, + "step": 12559 + }, + { + "epoch": 0.93, + "learning_rate": 1.91956763131089e-05, + "loss": 1.087, + "step": 12560 + }, + { + "epoch": 0.93, + "learning_rate": 1.9195519576464442e-05, + "loss": 1.0154, + "step": 12561 + }, + { + "epoch": 0.93, + "learning_rate": 1.9195362825190043e-05, + "loss": 1.0045, + "step": 12562 + }, + { + "epoch": 0.93, + "learning_rate": 1.919520605928595e-05, + "loss": 0.9571, + "step": 12563 + }, + { + "epoch": 0.93, + "learning_rate": 1.9195049278752416e-05, + "loss": 0.9198, + "step": 12564 + }, + { + "epoch": 0.93, + "learning_rate": 1.919489248358969e-05, + "loss": 0.9296, + "step": 12565 + }, + { + "epoch": 0.93, + "learning_rate": 1.9194735673798023e-05, + "loss": 0.9575, + "step": 12566 + }, + { + "epoch": 0.93, + "learning_rate": 1.9194578849377664e-05, + "loss": 1.0898, + "step": 12567 + }, + { + "epoch": 0.93, + "learning_rate": 1.9194422010328856e-05, + "loss": 1.0716, + "step": 12568 + }, + { + "epoch": 0.93, + "learning_rate": 1.919426515665186e-05, + "loss": 0.9988, + "step": 12569 + }, + { + "epoch": 0.93, + "learning_rate": 1.9194108288346917e-05, + "loss": 1.0324, + "step": 12570 + }, + { + "epoch": 0.93, + "learning_rate": 1.919395140541428e-05, + "loss": 0.832, + "step": 12571 + }, + { + "epoch": 0.93, + "learning_rate": 1.91937945078542e-05, + "loss": 1.0277, + "step": 12572 + }, + { + "epoch": 0.93, + "learning_rate": 1.9193637595666925e-05, + "loss": 1.0669, + "step": 12573 + }, + { + "epoch": 0.93, + "learning_rate": 1.9193480668852696e-05, + "loss": 1.0834, + "step": 12574 + }, + { + "epoch": 0.93, + "learning_rate": 1.9193323727411777e-05, + "loss": 0.8953, + "step": 12575 + }, + { + "epoch": 0.93, + "learning_rate": 1.9193166771344412e-05, + "loss": 0.9395, + "step": 12576 + }, + { + "epoch": 0.93, + "learning_rate": 1.919300980065085e-05, + "loss": 1.0621, + "step": 12577 + }, + { + "epoch": 0.93, + "learning_rate": 1.9192852815331342e-05, + "loss": 0.9279, + "step": 12578 + }, + { + "epoch": 0.93, + "learning_rate": 1.9192695815386132e-05, + "loss": 1.0962, + "step": 12579 + }, + { + "epoch": 0.93, + "learning_rate": 1.919253880081548e-05, + "loss": 1.0826, + "step": 12580 + }, + { + "epoch": 0.93, + "learning_rate": 1.9192381771619626e-05, + "loss": 1.0658, + "step": 12581 + }, + { + "epoch": 0.93, + "learning_rate": 1.9192224727798824e-05, + "loss": 0.9255, + "step": 12582 + }, + { + "epoch": 0.93, + "learning_rate": 1.9192067669353328e-05, + "loss": 1.0098, + "step": 12583 + }, + { + "epoch": 0.93, + "learning_rate": 1.919191059628338e-05, + "loss": 1.0481, + "step": 12584 + }, + { + "epoch": 0.93, + "learning_rate": 1.9191753508589234e-05, + "loss": 0.9684, + "step": 12585 + }, + { + "epoch": 0.93, + "learning_rate": 1.9191596406271144e-05, + "loss": 0.9536, + "step": 12586 + }, + { + "epoch": 0.93, + "learning_rate": 1.9191439289329352e-05, + "loss": 1.1251, + "step": 12587 + }, + { + "epoch": 0.93, + "learning_rate": 1.9191282157764113e-05, + "loss": 0.9924, + "step": 12588 + }, + { + "epoch": 0.93, + "learning_rate": 1.9191125011575673e-05, + "loss": 1.0321, + "step": 12589 + }, + { + "epoch": 0.93, + "learning_rate": 1.9190967850764286e-05, + "loss": 1.0161, + "step": 12590 + }, + { + "epoch": 0.93, + "learning_rate": 1.9190810675330204e-05, + "loss": 1.0961, + "step": 12591 + }, + { + "epoch": 0.93, + "learning_rate": 1.919065348527367e-05, + "loss": 1.0624, + "step": 12592 + }, + { + "epoch": 0.93, + "learning_rate": 1.919049628059494e-05, + "loss": 1.0157, + "step": 12593 + }, + { + "epoch": 0.93, + "learning_rate": 1.9190339061294257e-05, + "loss": 1.0838, + "step": 12594 + }, + { + "epoch": 0.93, + "learning_rate": 1.9190181827371882e-05, + "loss": 1.0463, + "step": 12595 + }, + { + "epoch": 0.93, + "learning_rate": 1.9190024578828055e-05, + "loss": 1.1143, + "step": 12596 + }, + { + "epoch": 0.93, + "learning_rate": 1.918986731566303e-05, + "loss": 1.0662, + "step": 12597 + }, + { + "epoch": 0.93, + "learning_rate": 1.918971003787706e-05, + "loss": 1.0184, + "step": 12598 + }, + { + "epoch": 0.93, + "learning_rate": 1.9189552745470396e-05, + "loss": 1.0438, + "step": 12599 + }, + { + "epoch": 0.93, + "learning_rate": 1.918939543844328e-05, + "loss": 1.058, + "step": 12600 + }, + { + "epoch": 0.93, + "learning_rate": 1.918923811679597e-05, + "loss": 1.0138, + "step": 12601 + }, + { + "epoch": 0.93, + "learning_rate": 1.9189080780528712e-05, + "loss": 1.0927, + "step": 12602 + }, + { + "epoch": 0.93, + "learning_rate": 1.918892342964176e-05, + "loss": 0.9942, + "step": 12603 + }, + { + "epoch": 0.93, + "learning_rate": 1.918876606413536e-05, + "loss": 1.0875, + "step": 12604 + }, + { + "epoch": 0.93, + "learning_rate": 1.9188608684009764e-05, + "loss": 1.0546, + "step": 12605 + }, + { + "epoch": 0.93, + "learning_rate": 1.9188451289265225e-05, + "loss": 0.9923, + "step": 12606 + }, + { + "epoch": 0.93, + "learning_rate": 1.9188293879901988e-05, + "loss": 0.9943, + "step": 12607 + }, + { + "epoch": 0.93, + "learning_rate": 1.918813645592031e-05, + "loss": 1.0456, + "step": 12608 + }, + { + "epoch": 0.93, + "learning_rate": 1.918797901732044e-05, + "loss": 1.082, + "step": 12609 + }, + { + "epoch": 0.93, + "learning_rate": 1.9187821564102625e-05, + "loss": 1.0071, + "step": 12610 + }, + { + "epoch": 0.93, + "learning_rate": 1.9187664096267118e-05, + "loss": 1.151, + "step": 12611 + }, + { + "epoch": 0.93, + "learning_rate": 1.9187506613814167e-05, + "loss": 1.0547, + "step": 12612 + }, + { + "epoch": 0.93, + "learning_rate": 1.9187349116744026e-05, + "loss": 1.0441, + "step": 12613 + }, + { + "epoch": 0.93, + "learning_rate": 1.9187191605056946e-05, + "loss": 0.9753, + "step": 12614 + }, + { + "epoch": 0.93, + "learning_rate": 1.918703407875317e-05, + "loss": 0.9252, + "step": 12615 + }, + { + "epoch": 0.93, + "learning_rate": 1.918687653783296e-05, + "loss": 1.0321, + "step": 12616 + }, + { + "epoch": 0.93, + "learning_rate": 1.918671898229656e-05, + "loss": 0.8928, + "step": 12617 + }, + { + "epoch": 0.93, + "learning_rate": 1.9186561412144216e-05, + "loss": 0.9295, + "step": 12618 + }, + { + "epoch": 0.93, + "learning_rate": 1.918640382737619e-05, + "loss": 1.1283, + "step": 12619 + }, + { + "epoch": 0.93, + "learning_rate": 1.9186246227992726e-05, + "loss": 0.985, + "step": 12620 + }, + { + "epoch": 0.93, + "learning_rate": 1.9186088613994075e-05, + "loss": 0.9946, + "step": 12621 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185930985380486e-05, + "loss": 1.054, + "step": 12622 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185773342152218e-05, + "loss": 0.973, + "step": 12623 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185615684309508e-05, + "loss": 1.1198, + "step": 12624 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185458011852623e-05, + "loss": 0.915, + "step": 12625 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185300324781798e-05, + "loss": 0.9968, + "step": 12626 + }, + { + "epoch": 0.93, + "learning_rate": 1.9185142623097296e-05, + "loss": 1.0522, + "step": 12627 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184984906799365e-05, + "loss": 1.1294, + "step": 12628 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184827175888255e-05, + "loss": 1.0441, + "step": 12629 + }, + { + "epoch": 0.93, + "learning_rate": 1.918466943036421e-05, + "loss": 0.9502, + "step": 12630 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184511670227494e-05, + "loss": 1.0583, + "step": 12631 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184353895478346e-05, + "loss": 0.9645, + "step": 12632 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184196106117027e-05, + "loss": 0.9696, + "step": 12633 + }, + { + "epoch": 0.93, + "learning_rate": 1.9184038302143783e-05, + "loss": 0.9064, + "step": 12634 + }, + { + "epoch": 0.93, + "learning_rate": 1.9183880483558865e-05, + "loss": 0.9538, + "step": 12635 + }, + { + "epoch": 0.93, + "learning_rate": 1.918372265036252e-05, + "loss": 1.0122, + "step": 12636 + }, + { + "epoch": 0.93, + "learning_rate": 1.918356480255501e-05, + "loss": 1.0887, + "step": 12637 + }, + { + "epoch": 0.93, + "learning_rate": 1.9183406940136576e-05, + "loss": 1.062, + "step": 12638 + }, + { + "epoch": 0.93, + "learning_rate": 1.9183249063107472e-05, + "loss": 1.0994, + "step": 12639 + }, + { + "epoch": 0.93, + "learning_rate": 1.9183091171467954e-05, + "loss": 1.0992, + "step": 12640 + }, + { + "epoch": 0.93, + "learning_rate": 1.9182933265218266e-05, + "loss": 1.0133, + "step": 12641 + }, + { + "epoch": 0.93, + "learning_rate": 1.9182775344358667e-05, + "loss": 0.9903, + "step": 12642 + }, + { + "epoch": 0.93, + "learning_rate": 1.91826174088894e-05, + "loss": 0.9355, + "step": 12643 + }, + { + "epoch": 0.93, + "learning_rate": 1.918245945881072e-05, + "loss": 1.0244, + "step": 12644 + }, + { + "epoch": 0.93, + "learning_rate": 1.918230149412288e-05, + "loss": 1.0104, + "step": 12645 + }, + { + "epoch": 0.93, + "learning_rate": 1.918214351482613e-05, + "loss": 1.0052, + "step": 12646 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181985520920715e-05, + "loss": 0.9951, + "step": 12647 + }, + { + "epoch": 0.93, + "learning_rate": 1.91818275124069e-05, + "loss": 1.0059, + "step": 12648 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181669489284928e-05, + "loss": 1.0913, + "step": 12649 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181511451555046e-05, + "loss": 0.9334, + "step": 12650 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181353399217515e-05, + "loss": 0.8953, + "step": 12651 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181195332272582e-05, + "loss": 0.9745, + "step": 12652 + }, + { + "epoch": 0.93, + "learning_rate": 1.9181037250720495e-05, + "loss": 1.0179, + "step": 12653 + }, + { + "epoch": 0.93, + "learning_rate": 1.9180879154561514e-05, + "loss": 1.0611, + "step": 12654 + }, + { + "epoch": 0.94, + "learning_rate": 1.918072104379588e-05, + "loss": 1.0788, + "step": 12655 + }, + { + "epoch": 0.94, + "learning_rate": 1.918056291842386e-05, + "loss": 0.9571, + "step": 12656 + }, + { + "epoch": 0.94, + "learning_rate": 1.9180404778445685e-05, + "loss": 1.0059, + "step": 12657 + }, + { + "epoch": 0.94, + "learning_rate": 1.918024662386162e-05, + "loss": 0.9864, + "step": 12658 + }, + { + "epoch": 0.94, + "learning_rate": 1.9180088454671918e-05, + "loss": 1.0199, + "step": 12659 + }, + { + "epoch": 0.94, + "learning_rate": 1.9179930270876825e-05, + "loss": 0.9173, + "step": 12660 + }, + { + "epoch": 0.94, + "learning_rate": 1.9179772072476593e-05, + "loss": 0.9515, + "step": 12661 + }, + { + "epoch": 0.94, + "learning_rate": 1.9179613859471474e-05, + "loss": 1.0004, + "step": 12662 + }, + { + "epoch": 0.94, + "learning_rate": 1.9179455631861725e-05, + "loss": 1.0394, + "step": 12663 + }, + { + "epoch": 0.94, + "learning_rate": 1.917929738964759e-05, + "loss": 1.0207, + "step": 12664 + }, + { + "epoch": 0.94, + "learning_rate": 1.9179139132829326e-05, + "loss": 1.0699, + "step": 12665 + }, + { + "epoch": 0.94, + "learning_rate": 1.917898086140718e-05, + "loss": 1.0295, + "step": 12666 + }, + { + "epoch": 0.94, + "learning_rate": 1.9178822575381407e-05, + "loss": 1.0487, + "step": 12667 + }, + { + "epoch": 0.94, + "learning_rate": 1.917866427475226e-05, + "loss": 1.0583, + "step": 12668 + }, + { + "epoch": 0.94, + "learning_rate": 1.917850595951999e-05, + "loss": 0.9719, + "step": 12669 + }, + { + "epoch": 0.94, + "learning_rate": 1.917834762968485e-05, + "loss": 1.0985, + "step": 12670 + }, + { + "epoch": 0.94, + "learning_rate": 1.917818928524709e-05, + "loss": 0.9447, + "step": 12671 + }, + { + "epoch": 0.94, + "learning_rate": 1.9178030926206957e-05, + "loss": 1.1358, + "step": 12672 + }, + { + "epoch": 0.94, + "learning_rate": 1.9177872552564713e-05, + "loss": 1.1222, + "step": 12673 + }, + { + "epoch": 0.94, + "learning_rate": 1.9177714164320603e-05, + "loss": 1.0797, + "step": 12674 + }, + { + "epoch": 0.94, + "learning_rate": 1.9177555761474882e-05, + "loss": 1.0629, + "step": 12675 + }, + { + "epoch": 0.94, + "learning_rate": 1.91773973440278e-05, + "loss": 0.9249, + "step": 12676 + }, + { + "epoch": 0.94, + "learning_rate": 1.9177238911979612e-05, + "loss": 0.9515, + "step": 12677 + }, + { + "epoch": 0.94, + "learning_rate": 1.9177080465330565e-05, + "loss": 1.0409, + "step": 12678 + }, + { + "epoch": 0.94, + "learning_rate": 1.917692200408092e-05, + "loss": 1.0835, + "step": 12679 + }, + { + "epoch": 0.94, + "learning_rate": 1.917676352823092e-05, + "loss": 0.9926, + "step": 12680 + }, + { + "epoch": 0.94, + "learning_rate": 1.917660503778082e-05, + "loss": 0.9749, + "step": 12681 + }, + { + "epoch": 0.94, + "learning_rate": 1.917644653273087e-05, + "loss": 0.9908, + "step": 12682 + }, + { + "epoch": 0.94, + "learning_rate": 1.917628801308133e-05, + "loss": 0.9399, + "step": 12683 + }, + { + "epoch": 0.94, + "learning_rate": 1.9176129478832445e-05, + "loss": 0.9478, + "step": 12684 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175970929984468e-05, + "loss": 0.9604, + "step": 12685 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175812366537655e-05, + "loss": 0.9788, + "step": 12686 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175653788492252e-05, + "loss": 0.995, + "step": 12687 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175495195848518e-05, + "loss": 1.0949, + "step": 12688 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175336588606703e-05, + "loss": 0.9508, + "step": 12689 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175177966767058e-05, + "loss": 1.0994, + "step": 12690 + }, + { + "epoch": 0.94, + "learning_rate": 1.9175019330329836e-05, + "loss": 1.0869, + "step": 12691 + }, + { + "epoch": 0.94, + "learning_rate": 1.917486067929529e-05, + "loss": 0.9869, + "step": 12692 + }, + { + "epoch": 0.94, + "learning_rate": 1.917470201366367e-05, + "loss": 0.9126, + "step": 12693 + }, + { + "epoch": 0.94, + "learning_rate": 1.9174543333435233e-05, + "loss": 1.0109, + "step": 12694 + }, + { + "epoch": 0.94, + "learning_rate": 1.9174384638610228e-05, + "loss": 1.0619, + "step": 12695 + }, + { + "epoch": 0.94, + "learning_rate": 1.9174225929188907e-05, + "loss": 0.9279, + "step": 12696 + }, + { + "epoch": 0.94, + "learning_rate": 1.9174067205171525e-05, + "loss": 1.0737, + "step": 12697 + }, + { + "epoch": 0.94, + "learning_rate": 1.9173908466558332e-05, + "loss": 1.0308, + "step": 12698 + }, + { + "epoch": 0.94, + "learning_rate": 1.9173749713349582e-05, + "loss": 0.8989, + "step": 12699 + }, + { + "epoch": 0.94, + "learning_rate": 1.917359094554553e-05, + "loss": 0.9144, + "step": 12700 + }, + { + "epoch": 0.94, + "learning_rate": 1.9173432163146426e-05, + "loss": 0.9855, + "step": 12701 + }, + { + "epoch": 0.94, + "learning_rate": 1.917327336615252e-05, + "loss": 1.1023, + "step": 12702 + }, + { + "epoch": 0.94, + "learning_rate": 1.9173114554564068e-05, + "loss": 1.0554, + "step": 12703 + }, + { + "epoch": 0.94, + "learning_rate": 1.9172955728381325e-05, + "loss": 1.0758, + "step": 12704 + }, + { + "epoch": 0.94, + "learning_rate": 1.9172796887604534e-05, + "loss": 1.0248, + "step": 12705 + }, + { + "epoch": 0.94, + "learning_rate": 1.917263803223396e-05, + "loss": 1.0317, + "step": 12706 + }, + { + "epoch": 0.94, + "learning_rate": 1.9172479162269848e-05, + "loss": 0.9368, + "step": 12707 + }, + { + "epoch": 0.94, + "learning_rate": 1.917232027771245e-05, + "loss": 0.9921, + "step": 12708 + }, + { + "epoch": 0.94, + "learning_rate": 1.917216137856203e-05, + "loss": 1.1246, + "step": 12709 + }, + { + "epoch": 0.94, + "learning_rate": 1.9172002464818828e-05, + "loss": 0.9835, + "step": 12710 + }, + { + "epoch": 0.94, + "learning_rate": 1.9171843536483102e-05, + "loss": 1.0804, + "step": 12711 + }, + { + "epoch": 0.94, + "learning_rate": 1.91716845935551e-05, + "loss": 0.9489, + "step": 12712 + }, + { + "epoch": 0.94, + "learning_rate": 1.917152563603508e-05, + "loss": 1.03, + "step": 12713 + }, + { + "epoch": 0.94, + "learning_rate": 1.91713666639233e-05, + "loss": 1.0408, + "step": 12714 + }, + { + "epoch": 0.94, + "learning_rate": 1.9171207677220002e-05, + "loss": 1.0095, + "step": 12715 + }, + { + "epoch": 0.94, + "learning_rate": 1.9171048675925446e-05, + "loss": 1.0352, + "step": 12716 + }, + { + "epoch": 0.94, + "learning_rate": 1.9170889660039883e-05, + "loss": 1.1198, + "step": 12717 + }, + { + "epoch": 0.94, + "learning_rate": 1.917073062956356e-05, + "loss": 0.9818, + "step": 12718 + }, + { + "epoch": 0.94, + "learning_rate": 1.9170571584496742e-05, + "loss": 1.0824, + "step": 12719 + }, + { + "epoch": 0.94, + "learning_rate": 1.9170412524839674e-05, + "loss": 0.9433, + "step": 12720 + }, + { + "epoch": 0.94, + "learning_rate": 1.917025345059261e-05, + "loss": 1.0494, + "step": 12721 + }, + { + "epoch": 0.94, + "learning_rate": 1.9170094361755803e-05, + "loss": 1.0889, + "step": 12722 + }, + { + "epoch": 0.94, + "learning_rate": 1.9169935258329514e-05, + "loss": 0.9821, + "step": 12723 + }, + { + "epoch": 0.94, + "learning_rate": 1.916977614031398e-05, + "loss": 1.0003, + "step": 12724 + }, + { + "epoch": 0.94, + "learning_rate": 1.916961700770947e-05, + "loss": 1.0465, + "step": 12725 + }, + { + "epoch": 0.94, + "learning_rate": 1.9169457860516227e-05, + "loss": 1.0611, + "step": 12726 + }, + { + "epoch": 0.94, + "learning_rate": 1.9169298698734508e-05, + "loss": 1.1087, + "step": 12727 + }, + { + "epoch": 0.94, + "learning_rate": 1.9169139522364567e-05, + "loss": 1.0273, + "step": 12728 + }, + { + "epoch": 0.94, + "learning_rate": 1.9168980331406656e-05, + "loss": 1.0587, + "step": 12729 + }, + { + "epoch": 0.94, + "learning_rate": 1.9168821125861028e-05, + "loss": 0.959, + "step": 12730 + }, + { + "epoch": 0.94, + "learning_rate": 1.9168661905727938e-05, + "loss": 1.0249, + "step": 12731 + }, + { + "epoch": 0.94, + "learning_rate": 1.9168502671007636e-05, + "loss": 1.031, + "step": 12732 + }, + { + "epoch": 0.94, + "learning_rate": 1.916834342170038e-05, + "loss": 1.0957, + "step": 12733 + }, + { + "epoch": 0.94, + "learning_rate": 1.916818415780642e-05, + "loss": 1.0602, + "step": 12734 + }, + { + "epoch": 0.94, + "learning_rate": 1.916802487932601e-05, + "loss": 0.9988, + "step": 12735 + }, + { + "epoch": 0.94, + "learning_rate": 1.91678655862594e-05, + "loss": 0.9702, + "step": 12736 + }, + { + "epoch": 0.94, + "learning_rate": 1.916770627860685e-05, + "loss": 0.9305, + "step": 12737 + }, + { + "epoch": 0.94, + "learning_rate": 1.9167546956368612e-05, + "loss": 0.9341, + "step": 12738 + }, + { + "epoch": 0.94, + "learning_rate": 1.9167387619544936e-05, + "loss": 0.9406, + "step": 12739 + }, + { + "epoch": 0.94, + "learning_rate": 1.916722826813608e-05, + "loss": 1.0548, + "step": 12740 + }, + { + "epoch": 0.94, + "learning_rate": 1.916706890214229e-05, + "loss": 1.1147, + "step": 12741 + }, + { + "epoch": 0.94, + "learning_rate": 1.916690952156383e-05, + "loss": 0.9801, + "step": 12742 + }, + { + "epoch": 0.94, + "learning_rate": 1.9166750126400945e-05, + "loss": 0.944, + "step": 12743 + }, + { + "epoch": 0.94, + "learning_rate": 1.9166590716653895e-05, + "loss": 1.0436, + "step": 12744 + }, + { + "epoch": 0.94, + "learning_rate": 1.9166431292322927e-05, + "loss": 0.956, + "step": 12745 + }, + { + "epoch": 0.94, + "learning_rate": 1.9166271853408297e-05, + "loss": 1.0133, + "step": 12746 + }, + { + "epoch": 0.94, + "learning_rate": 1.916611239991026e-05, + "loss": 1.0494, + "step": 12747 + }, + { + "epoch": 0.94, + "learning_rate": 1.916595293182907e-05, + "loss": 1.0146, + "step": 12748 + }, + { + "epoch": 0.94, + "learning_rate": 1.9165793449164983e-05, + "loss": 1.0311, + "step": 12749 + }, + { + "epoch": 0.94, + "learning_rate": 1.9165633951918247e-05, + "loss": 1.1334, + "step": 12750 + }, + { + "epoch": 0.94, + "learning_rate": 1.916547444008912e-05, + "loss": 0.9584, + "step": 12751 + }, + { + "epoch": 0.94, + "learning_rate": 1.916531491367785e-05, + "loss": 1.1637, + "step": 12752 + }, + { + "epoch": 0.94, + "learning_rate": 1.91651553726847e-05, + "loss": 1.1714, + "step": 12753 + }, + { + "epoch": 0.94, + "learning_rate": 1.916499581710992e-05, + "loss": 1.1392, + "step": 12754 + }, + { + "epoch": 0.94, + "learning_rate": 1.916483624695376e-05, + "loss": 1.0945, + "step": 12755 + }, + { + "epoch": 0.94, + "learning_rate": 1.9164676662216474e-05, + "loss": 1.1435, + "step": 12756 + }, + { + "epoch": 0.94, + "learning_rate": 1.916451706289832e-05, + "loss": 0.9949, + "step": 12757 + }, + { + "epoch": 0.94, + "learning_rate": 1.9164357448999556e-05, + "loss": 1.0751, + "step": 12758 + }, + { + "epoch": 0.94, + "learning_rate": 1.9164197820520423e-05, + "loss": 1.0565, + "step": 12759 + }, + { + "epoch": 0.94, + "learning_rate": 1.9164038177461187e-05, + "loss": 0.9221, + "step": 12760 + }, + { + "epoch": 0.94, + "learning_rate": 1.9163878519822098e-05, + "loss": 1.0313, + "step": 12761 + }, + { + "epoch": 0.94, + "learning_rate": 1.9163718847603404e-05, + "loss": 1.0728, + "step": 12762 + }, + { + "epoch": 0.94, + "learning_rate": 1.9163559160805368e-05, + "loss": 1.0601, + "step": 12763 + }, + { + "epoch": 0.94, + "learning_rate": 1.916339945942824e-05, + "loss": 1.1617, + "step": 12764 + }, + { + "epoch": 0.94, + "learning_rate": 1.9163239743472276e-05, + "loss": 1.1424, + "step": 12765 + }, + { + "epoch": 0.94, + "learning_rate": 1.916308001293773e-05, + "loss": 0.9789, + "step": 12766 + }, + { + "epoch": 0.94, + "learning_rate": 1.916292026782485e-05, + "loss": 0.9031, + "step": 12767 + }, + { + "epoch": 0.94, + "learning_rate": 1.91627605081339e-05, + "loss": 1.0226, + "step": 12768 + }, + { + "epoch": 0.94, + "learning_rate": 1.9162600733865126e-05, + "loss": 1.0913, + "step": 12769 + }, + { + "epoch": 0.94, + "learning_rate": 1.9162440945018783e-05, + "loss": 1.0863, + "step": 12770 + }, + { + "epoch": 0.94, + "learning_rate": 1.916228114159513e-05, + "loss": 1.0496, + "step": 12771 + }, + { + "epoch": 0.94, + "learning_rate": 1.916212132359442e-05, + "loss": 1.0334, + "step": 12772 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161961491016906e-05, + "loss": 1.0664, + "step": 12773 + }, + { + "epoch": 0.94, + "learning_rate": 1.916180164386284e-05, + "loss": 1.0293, + "step": 12774 + }, + { + "epoch": 0.94, + "learning_rate": 1.916164178213248e-05, + "loss": 0.9752, + "step": 12775 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161481905826078e-05, + "loss": 1.0197, + "step": 12776 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161322014943893e-05, + "loss": 0.9649, + "step": 12777 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161162109486172e-05, + "loss": 1.0013, + "step": 12778 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161002189453176e-05, + "loss": 0.9851, + "step": 12779 + }, + { + "epoch": 0.94, + "learning_rate": 1.9160842254845152e-05, + "loss": 1.0151, + "step": 12780 + }, + { + "epoch": 0.94, + "learning_rate": 1.9160682305662364e-05, + "loss": 1.0571, + "step": 12781 + }, + { + "epoch": 0.94, + "learning_rate": 1.916052234190506e-05, + "loss": 0.9795, + "step": 12782 + }, + { + "epoch": 0.94, + "learning_rate": 1.9160362363573493e-05, + "loss": 1.1546, + "step": 12783 + }, + { + "epoch": 0.94, + "learning_rate": 1.9160202370667925e-05, + "loss": 1.0071, + "step": 12784 + }, + { + "epoch": 0.94, + "learning_rate": 1.9160042363188603e-05, + "loss": 0.9182, + "step": 12785 + }, + { + "epoch": 0.94, + "learning_rate": 1.9159882341135787e-05, + "loss": 1.0468, + "step": 12786 + }, + { + "epoch": 0.94, + "learning_rate": 1.915972230450973e-05, + "loss": 1.1058, + "step": 12787 + }, + { + "epoch": 0.94, + "learning_rate": 1.9159562253310683e-05, + "loss": 0.9893, + "step": 12788 + }, + { + "epoch": 0.94, + "learning_rate": 1.9159402187538902e-05, + "loss": 0.9724, + "step": 12789 + }, + { + "epoch": 0.95, + "learning_rate": 1.9159242107194648e-05, + "loss": 1.0578, + "step": 12790 + }, + { + "epoch": 0.95, + "learning_rate": 1.9159082012278168e-05, + "loss": 0.9828, + "step": 12791 + }, + { + "epoch": 0.95, + "learning_rate": 1.9158921902789722e-05, + "loss": 1.0415, + "step": 12792 + }, + { + "epoch": 0.95, + "learning_rate": 1.915876177872956e-05, + "loss": 1.1429, + "step": 12793 + }, + { + "epoch": 0.95, + "learning_rate": 1.915860164009794e-05, + "loss": 1.0221, + "step": 12794 + }, + { + "epoch": 0.95, + "learning_rate": 1.9158441486895112e-05, + "loss": 0.8681, + "step": 12795 + }, + { + "epoch": 0.95, + "learning_rate": 1.915828131912134e-05, + "loss": 1.0629, + "step": 12796 + }, + { + "epoch": 0.95, + "learning_rate": 1.915812113677687e-05, + "loss": 1.093, + "step": 12797 + }, + { + "epoch": 0.95, + "learning_rate": 1.915796093986196e-05, + "loss": 0.9687, + "step": 12798 + }, + { + "epoch": 0.95, + "learning_rate": 1.915780072837687e-05, + "loss": 1.0745, + "step": 12799 + }, + { + "epoch": 0.95, + "learning_rate": 1.9157640502321844e-05, + "loss": 1.0708, + "step": 12800 + }, + { + "epoch": 0.95, + "learning_rate": 1.9157480261697147e-05, + "loss": 0.9693, + "step": 12801 + }, + { + "epoch": 0.95, + "learning_rate": 1.915732000650303e-05, + "loss": 1.0695, + "step": 12802 + }, + { + "epoch": 0.95, + "learning_rate": 1.9157159736739745e-05, + "loss": 1.0121, + "step": 12803 + }, + { + "epoch": 0.95, + "learning_rate": 1.915699945240755e-05, + "loss": 1.02, + "step": 12804 + }, + { + "epoch": 0.95, + "learning_rate": 1.9156839153506706e-05, + "loss": 1.0436, + "step": 12805 + }, + { + "epoch": 0.95, + "learning_rate": 1.9156678840037453e-05, + "loss": 0.9217, + "step": 12806 + }, + { + "epoch": 0.95, + "learning_rate": 1.915651851200006e-05, + "loss": 1.0313, + "step": 12807 + }, + { + "epoch": 0.95, + "learning_rate": 1.915635816939478e-05, + "loss": 1.1463, + "step": 12808 + }, + { + "epoch": 0.95, + "learning_rate": 1.915619781222186e-05, + "loss": 1.1104, + "step": 12809 + }, + { + "epoch": 0.95, + "learning_rate": 1.9156037440481563e-05, + "loss": 1.0905, + "step": 12810 + }, + { + "epoch": 0.95, + "learning_rate": 1.9155877054174138e-05, + "loss": 0.9902, + "step": 12811 + }, + { + "epoch": 0.95, + "learning_rate": 1.9155716653299847e-05, + "loss": 0.9659, + "step": 12812 + }, + { + "epoch": 0.95, + "learning_rate": 1.915555623785894e-05, + "loss": 1.1513, + "step": 12813 + }, + { + "epoch": 0.95, + "learning_rate": 1.9155395807851676e-05, + "loss": 0.9418, + "step": 12814 + }, + { + "epoch": 0.95, + "learning_rate": 1.9155235363278308e-05, + "loss": 1.1522, + "step": 12815 + }, + { + "epoch": 0.95, + "learning_rate": 1.915507490413909e-05, + "loss": 1.0216, + "step": 12816 + }, + { + "epoch": 0.95, + "learning_rate": 1.915491443043428e-05, + "loss": 0.9525, + "step": 12817 + }, + { + "epoch": 0.95, + "learning_rate": 1.9154753942164132e-05, + "loss": 0.9673, + "step": 12818 + }, + { + "epoch": 0.95, + "learning_rate": 1.91545934393289e-05, + "loss": 1.0606, + "step": 12819 + }, + { + "epoch": 0.95, + "learning_rate": 1.9154432921928842e-05, + "loss": 0.9578, + "step": 12820 + }, + { + "epoch": 0.95, + "learning_rate": 1.9154272389964212e-05, + "loss": 0.9737, + "step": 12821 + }, + { + "epoch": 0.95, + "learning_rate": 1.9154111843435267e-05, + "loss": 1.1001, + "step": 12822 + }, + { + "epoch": 0.95, + "learning_rate": 1.9153951282342257e-05, + "loss": 1.1539, + "step": 12823 + }, + { + "epoch": 0.95, + "learning_rate": 1.9153790706685445e-05, + "loss": 1.0739, + "step": 12824 + }, + { + "epoch": 0.95, + "learning_rate": 1.915363011646508e-05, + "loss": 0.9997, + "step": 12825 + }, + { + "epoch": 0.95, + "learning_rate": 1.9153469511681423e-05, + "loss": 1.0195, + "step": 12826 + }, + { + "epoch": 0.95, + "learning_rate": 1.915330889233473e-05, + "loss": 0.9337, + "step": 12827 + }, + { + "epoch": 0.95, + "learning_rate": 1.915314825842525e-05, + "loss": 0.9982, + "step": 12828 + }, + { + "epoch": 0.95, + "learning_rate": 1.915298760995324e-05, + "loss": 1.025, + "step": 12829 + }, + { + "epoch": 0.95, + "learning_rate": 1.915282694691896e-05, + "loss": 1.0227, + "step": 12830 + }, + { + "epoch": 0.95, + "learning_rate": 1.915266626932266e-05, + "loss": 0.9585, + "step": 12831 + }, + { + "epoch": 0.95, + "learning_rate": 1.9152505577164605e-05, + "loss": 1.011, + "step": 12832 + }, + { + "epoch": 0.95, + "learning_rate": 1.915234487044504e-05, + "loss": 0.8935, + "step": 12833 + }, + { + "epoch": 0.95, + "learning_rate": 1.9152184149164225e-05, + "loss": 0.9933, + "step": 12834 + }, + { + "epoch": 0.95, + "learning_rate": 1.915202341332242e-05, + "loss": 1.0826, + "step": 12835 + }, + { + "epoch": 0.95, + "learning_rate": 1.9151862662919875e-05, + "loss": 1.0597, + "step": 12836 + }, + { + "epoch": 0.95, + "learning_rate": 1.9151701897956844e-05, + "loss": 1.0256, + "step": 12837 + }, + { + "epoch": 0.95, + "learning_rate": 1.915154111843359e-05, + "loss": 1.0255, + "step": 12838 + }, + { + "epoch": 0.95, + "learning_rate": 1.9151380324350365e-05, + "loss": 1.0754, + "step": 12839 + }, + { + "epoch": 0.95, + "learning_rate": 1.9151219515707424e-05, + "loss": 0.8734, + "step": 12840 + }, + { + "epoch": 0.95, + "learning_rate": 1.9151058692505026e-05, + "loss": 1.0441, + "step": 12841 + }, + { + "epoch": 0.95, + "learning_rate": 1.915089785474342e-05, + "loss": 0.9723, + "step": 12842 + }, + { + "epoch": 0.95, + "learning_rate": 1.9150737002422867e-05, + "loss": 1.0224, + "step": 12843 + }, + { + "epoch": 0.95, + "learning_rate": 1.9150576135543625e-05, + "loss": 1.0015, + "step": 12844 + }, + { + "epoch": 0.95, + "learning_rate": 1.9150415254105947e-05, + "loss": 1.0205, + "step": 12845 + }, + { + "epoch": 0.95, + "learning_rate": 1.9150254358110086e-05, + "loss": 1.0165, + "step": 12846 + }, + { + "epoch": 0.95, + "learning_rate": 1.9150093447556304e-05, + "loss": 0.9984, + "step": 12847 + }, + { + "epoch": 0.95, + "learning_rate": 1.9149932522444855e-05, + "loss": 0.9514, + "step": 12848 + }, + { + "epoch": 0.95, + "learning_rate": 1.9149771582775993e-05, + "loss": 0.9714, + "step": 12849 + }, + { + "epoch": 0.95, + "learning_rate": 1.9149610628549978e-05, + "loss": 1.0178, + "step": 12850 + }, + { + "epoch": 0.95, + "learning_rate": 1.914944965976706e-05, + "loss": 1.0469, + "step": 12851 + }, + { + "epoch": 0.95, + "learning_rate": 1.91492886764275e-05, + "loss": 1.1075, + "step": 12852 + }, + { + "epoch": 0.95, + "learning_rate": 1.914912767853155e-05, + "loss": 1.0763, + "step": 12853 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148966666079472e-05, + "loss": 0.9786, + "step": 12854 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148805639071518e-05, + "loss": 0.9711, + "step": 12855 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148644597507943e-05, + "loss": 0.9963, + "step": 12856 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148483541389008e-05, + "loss": 0.9572, + "step": 12857 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148322470714967e-05, + "loss": 1.0035, + "step": 12858 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148161385486074e-05, + "loss": 1.0578, + "step": 12859 + }, + { + "epoch": 0.95, + "learning_rate": 1.9148000285702587e-05, + "loss": 1.1467, + "step": 12860 + }, + { + "epoch": 0.95, + "learning_rate": 1.9147839171364763e-05, + "loss": 1.0481, + "step": 12861 + }, + { + "epoch": 0.95, + "learning_rate": 1.914767804247286e-05, + "loss": 1.0095, + "step": 12862 + }, + { + "epoch": 0.95, + "learning_rate": 1.914751689902713e-05, + "loss": 1.1793, + "step": 12863 + }, + { + "epoch": 0.95, + "learning_rate": 1.914735574102783e-05, + "loss": 1.0178, + "step": 12864 + }, + { + "epoch": 0.95, + "learning_rate": 1.914719456847522e-05, + "loss": 1.0109, + "step": 12865 + }, + { + "epoch": 0.95, + "learning_rate": 1.9147033381369553e-05, + "loss": 0.9981, + "step": 12866 + }, + { + "epoch": 0.95, + "learning_rate": 1.914687217971109e-05, + "loss": 1.0493, + "step": 12867 + }, + { + "epoch": 0.95, + "learning_rate": 1.914671096350008e-05, + "loss": 1.0292, + "step": 12868 + }, + { + "epoch": 0.95, + "learning_rate": 1.9146549732736786e-05, + "loss": 1.0124, + "step": 12869 + }, + { + "epoch": 0.95, + "learning_rate": 1.914638848742146e-05, + "loss": 1.0268, + "step": 12870 + }, + { + "epoch": 0.95, + "learning_rate": 1.9146227227554365e-05, + "loss": 0.9006, + "step": 12871 + }, + { + "epoch": 0.95, + "learning_rate": 1.914606595313575e-05, + "loss": 1.1326, + "step": 12872 + }, + { + "epoch": 0.95, + "learning_rate": 1.9145904664165876e-05, + "loss": 0.9764, + "step": 12873 + }, + { + "epoch": 0.95, + "learning_rate": 1.9145743360645e-05, + "loss": 1.0473, + "step": 12874 + }, + { + "epoch": 0.95, + "learning_rate": 1.9145582042573373e-05, + "loss": 1.065, + "step": 12875 + }, + { + "epoch": 0.95, + "learning_rate": 1.914542070995126e-05, + "loss": 0.9726, + "step": 12876 + }, + { + "epoch": 0.95, + "learning_rate": 1.914525936277891e-05, + "loss": 0.9741, + "step": 12877 + }, + { + "epoch": 0.95, + "learning_rate": 1.9145098001056584e-05, + "loss": 1.1413, + "step": 12878 + }, + { + "epoch": 0.95, + "learning_rate": 1.914493662478454e-05, + "loss": 1.0374, + "step": 12879 + }, + { + "epoch": 0.95, + "learning_rate": 1.914477523396303e-05, + "loss": 1.0806, + "step": 12880 + }, + { + "epoch": 0.95, + "learning_rate": 1.9144613828592317e-05, + "loss": 0.9262, + "step": 12881 + }, + { + "epoch": 0.95, + "learning_rate": 1.9144452408672653e-05, + "loss": 1.0368, + "step": 12882 + }, + { + "epoch": 0.95, + "learning_rate": 1.9144290974204294e-05, + "loss": 0.9858, + "step": 12883 + }, + { + "epoch": 0.95, + "learning_rate": 1.9144129525187503e-05, + "loss": 1.0767, + "step": 12884 + }, + { + "epoch": 0.95, + "learning_rate": 1.914396806162253e-05, + "loss": 0.9995, + "step": 12885 + }, + { + "epoch": 0.95, + "learning_rate": 1.914380658350963e-05, + "loss": 1.055, + "step": 12886 + }, + { + "epoch": 0.95, + "learning_rate": 1.914364509084907e-05, + "loss": 1.0306, + "step": 12887 + }, + { + "epoch": 0.95, + "learning_rate": 1.9143483583641104e-05, + "loss": 1.0588, + "step": 12888 + }, + { + "epoch": 0.95, + "learning_rate": 1.9143322061885983e-05, + "loss": 0.954, + "step": 12889 + }, + { + "epoch": 0.95, + "learning_rate": 1.9143160525583968e-05, + "loss": 1.0217, + "step": 12890 + }, + { + "epoch": 0.95, + "learning_rate": 1.9142998974735315e-05, + "loss": 1.0254, + "step": 12891 + }, + { + "epoch": 0.95, + "learning_rate": 1.914283740934028e-05, + "loss": 0.9925, + "step": 12892 + }, + { + "epoch": 0.95, + "learning_rate": 1.914267582939912e-05, + "loss": 1.1748, + "step": 12893 + }, + { + "epoch": 0.95, + "learning_rate": 1.91425142349121e-05, + "loss": 1.0308, + "step": 12894 + }, + { + "epoch": 0.95, + "learning_rate": 1.914235262587947e-05, + "loss": 0.9787, + "step": 12895 + }, + { + "epoch": 0.95, + "learning_rate": 1.914219100230148e-05, + "loss": 0.9751, + "step": 12896 + }, + { + "epoch": 0.95, + "learning_rate": 1.9142029364178404e-05, + "loss": 1.0009, + "step": 12897 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141867711510485e-05, + "loss": 0.9706, + "step": 12898 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141706044297987e-05, + "loss": 1.0257, + "step": 12899 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141544362541164e-05, + "loss": 1.1683, + "step": 12900 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141382666240273e-05, + "loss": 1.1243, + "step": 12901 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141220955395577e-05, + "loss": 0.9339, + "step": 12902 + }, + { + "epoch": 0.95, + "learning_rate": 1.9141059230007328e-05, + "loss": 1.0157, + "step": 12903 + }, + { + "epoch": 0.95, + "learning_rate": 1.9140897490075784e-05, + "loss": 1.1142, + "step": 12904 + }, + { + "epoch": 0.95, + "learning_rate": 1.9140735735601203e-05, + "loss": 1.0469, + "step": 12905 + }, + { + "epoch": 0.95, + "learning_rate": 1.9140573966583842e-05, + "loss": 1.0513, + "step": 12906 + }, + { + "epoch": 0.95, + "learning_rate": 1.914041218302396e-05, + "loss": 1.0403, + "step": 12907 + }, + { + "epoch": 0.95, + "learning_rate": 1.914025038492181e-05, + "loss": 0.9947, + "step": 12908 + }, + { + "epoch": 0.95, + "learning_rate": 1.9140088572277652e-05, + "loss": 0.8685, + "step": 12909 + }, + { + "epoch": 0.95, + "learning_rate": 1.9139926745091747e-05, + "loss": 1.0066, + "step": 12910 + }, + { + "epoch": 0.95, + "learning_rate": 1.913976490336435e-05, + "loss": 1.0044, + "step": 12911 + }, + { + "epoch": 0.95, + "learning_rate": 1.9139603047095713e-05, + "loss": 1.0667, + "step": 12912 + }, + { + "epoch": 0.95, + "learning_rate": 1.9139441176286104e-05, + "loss": 1.0737, + "step": 12913 + }, + { + "epoch": 0.95, + "learning_rate": 1.9139279290935774e-05, + "loss": 1.0235, + "step": 12914 + }, + { + "epoch": 0.95, + "learning_rate": 1.913911739104498e-05, + "loss": 0.9986, + "step": 12915 + }, + { + "epoch": 0.95, + "learning_rate": 1.913895547661398e-05, + "loss": 1.0062, + "step": 12916 + }, + { + "epoch": 0.95, + "learning_rate": 1.9138793547643034e-05, + "loss": 1.0627, + "step": 12917 + }, + { + "epoch": 0.95, + "learning_rate": 1.9138631604132397e-05, + "loss": 1.0058, + "step": 12918 + }, + { + "epoch": 0.95, + "learning_rate": 1.913846964608233e-05, + "loss": 0.9867, + "step": 12919 + }, + { + "epoch": 0.95, + "learning_rate": 1.9138307673493086e-05, + "loss": 1.0192, + "step": 12920 + }, + { + "epoch": 0.95, + "learning_rate": 1.9138145686364928e-05, + "loss": 1.0094, + "step": 12921 + }, + { + "epoch": 0.95, + "learning_rate": 1.9137983684698106e-05, + "loss": 1.0961, + "step": 12922 + }, + { + "epoch": 0.95, + "learning_rate": 1.9137821668492885e-05, + "loss": 1.052, + "step": 12923 + }, + { + "epoch": 0.95, + "learning_rate": 1.9137659637749523e-05, + "loss": 1.0423, + "step": 12924 + }, + { + "epoch": 0.96, + "learning_rate": 1.9137497592468274e-05, + "loss": 0.9826, + "step": 12925 + }, + { + "epoch": 0.96, + "learning_rate": 1.9137335532649394e-05, + "loss": 1.085, + "step": 12926 + }, + { + "epoch": 0.96, + "learning_rate": 1.913717345829315e-05, + "loss": 1.0902, + "step": 12927 + }, + { + "epoch": 0.96, + "learning_rate": 1.913701136939979e-05, + "loss": 1.0484, + "step": 12928 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136849265969576e-05, + "loss": 1.023, + "step": 12929 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136687148002765e-05, + "loss": 0.9919, + "step": 12930 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136525015499616e-05, + "loss": 1.0734, + "step": 12931 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136362868460385e-05, + "loss": 1.0908, + "step": 12932 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136200706885334e-05, + "loss": 1.0318, + "step": 12933 + }, + { + "epoch": 0.96, + "learning_rate": 1.9136038530774716e-05, + "loss": 0.9434, + "step": 12934 + }, + { + "epoch": 0.96, + "learning_rate": 1.9135876340128792e-05, + "loss": 1.1365, + "step": 12935 + }, + { + "epoch": 0.96, + "learning_rate": 1.9135714134947824e-05, + "loss": 1.1957, + "step": 12936 + }, + { + "epoch": 0.96, + "learning_rate": 1.913555191523206e-05, + "loss": 1.0224, + "step": 12937 + }, + { + "epoch": 0.96, + "learning_rate": 1.9135389680981763e-05, + "loss": 1.0098, + "step": 12938 + }, + { + "epoch": 0.96, + "learning_rate": 1.9135227432197196e-05, + "loss": 0.93, + "step": 12939 + }, + { + "epoch": 0.96, + "learning_rate": 1.913506516887861e-05, + "loss": 0.9852, + "step": 12940 + }, + { + "epoch": 0.96, + "learning_rate": 1.9134902891026267e-05, + "loss": 1.0749, + "step": 12941 + }, + { + "epoch": 0.96, + "learning_rate": 1.9134740598640428e-05, + "loss": 1.0688, + "step": 12942 + }, + { + "epoch": 0.96, + "learning_rate": 1.913457829172134e-05, + "loss": 0.9805, + "step": 12943 + }, + { + "epoch": 0.96, + "learning_rate": 1.9134415970269274e-05, + "loss": 0.9872, + "step": 12944 + }, + { + "epoch": 0.96, + "learning_rate": 1.913425363428448e-05, + "loss": 1.0581, + "step": 12945 + }, + { + "epoch": 0.96, + "learning_rate": 1.913409128376722e-05, + "loss": 0.9714, + "step": 12946 + }, + { + "epoch": 0.96, + "learning_rate": 1.9133928918717752e-05, + "loss": 1.0849, + "step": 12947 + }, + { + "epoch": 0.96, + "learning_rate": 1.9133766539136332e-05, + "loss": 1.0101, + "step": 12948 + }, + { + "epoch": 0.96, + "learning_rate": 1.9133604145023225e-05, + "loss": 1.1192, + "step": 12949 + }, + { + "epoch": 0.96, + "learning_rate": 1.9133441736378678e-05, + "loss": 1.0275, + "step": 12950 + }, + { + "epoch": 0.96, + "learning_rate": 1.913327931320296e-05, + "loss": 1.106, + "step": 12951 + }, + { + "epoch": 0.96, + "learning_rate": 1.9133116875496322e-05, + "loss": 1.0334, + "step": 12952 + }, + { + "epoch": 0.96, + "learning_rate": 1.913295442325903e-05, + "loss": 1.109, + "step": 12953 + }, + { + "epoch": 0.96, + "learning_rate": 1.9132791956491335e-05, + "loss": 1.0406, + "step": 12954 + }, + { + "epoch": 0.96, + "learning_rate": 1.91326294751935e-05, + "loss": 1.0345, + "step": 12955 + }, + { + "epoch": 0.96, + "learning_rate": 1.913246697936578e-05, + "loss": 1.0324, + "step": 12956 + }, + { + "epoch": 0.96, + "learning_rate": 1.913230446900844e-05, + "loss": 1.026, + "step": 12957 + }, + { + "epoch": 0.96, + "learning_rate": 1.9132141944121732e-05, + "loss": 0.9837, + "step": 12958 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131979404705917e-05, + "loss": 0.923, + "step": 12959 + }, + { + "epoch": 0.96, + "learning_rate": 1.913181685076125e-05, + "loss": 0.9652, + "step": 12960 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131654282288e-05, + "loss": 0.9476, + "step": 12961 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131491699286413e-05, + "loss": 1.0186, + "step": 12962 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131329101756755e-05, + "loss": 1.0702, + "step": 12963 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131166489699282e-05, + "loss": 1.1155, + "step": 12964 + }, + { + "epoch": 0.96, + "learning_rate": 1.9131003863114256e-05, + "loss": 1.075, + "step": 12965 + }, + { + "epoch": 0.96, + "learning_rate": 1.9130841222001933e-05, + "loss": 1.0355, + "step": 12966 + }, + { + "epoch": 0.96, + "learning_rate": 1.913067856636257e-05, + "loss": 1.0835, + "step": 12967 + }, + { + "epoch": 0.96, + "learning_rate": 1.913051589619643e-05, + "loss": 1.0878, + "step": 12968 + }, + { + "epoch": 0.96, + "learning_rate": 1.9130353211503768e-05, + "loss": 1.0549, + "step": 12969 + }, + { + "epoch": 0.96, + "learning_rate": 1.913019051228485e-05, + "loss": 1.1199, + "step": 12970 + }, + { + "epoch": 0.96, + "learning_rate": 1.9130027798539927e-05, + "loss": 0.8996, + "step": 12971 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129865070269256e-05, + "loss": 1.0387, + "step": 12972 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129702327473105e-05, + "loss": 0.992, + "step": 12973 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129539570151725e-05, + "loss": 0.918, + "step": 12974 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129376798305383e-05, + "loss": 1.0849, + "step": 12975 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129214011934326e-05, + "loss": 1.0518, + "step": 12976 + }, + { + "epoch": 0.96, + "learning_rate": 1.9129051211038823e-05, + "loss": 0.9588, + "step": 12977 + }, + { + "epoch": 0.96, + "learning_rate": 1.9128888395619135e-05, + "loss": 0.9574, + "step": 12978 + }, + { + "epoch": 0.96, + "learning_rate": 1.912872556567551e-05, + "loss": 1.04, + "step": 12979 + }, + { + "epoch": 0.96, + "learning_rate": 1.9128562721208217e-05, + "loss": 0.9852, + "step": 12980 + }, + { + "epoch": 0.96, + "learning_rate": 1.912839986221751e-05, + "loss": 1.0662, + "step": 12981 + }, + { + "epoch": 0.96, + "learning_rate": 1.912823698870365e-05, + "loss": 1.0928, + "step": 12982 + }, + { + "epoch": 0.96, + "learning_rate": 1.9128074100666897e-05, + "loss": 0.9614, + "step": 12983 + }, + { + "epoch": 0.96, + "learning_rate": 1.9127911198107505e-05, + "loss": 1.0477, + "step": 12984 + }, + { + "epoch": 0.96, + "learning_rate": 1.912774828102574e-05, + "loss": 0.8667, + "step": 12985 + }, + { + "epoch": 0.96, + "learning_rate": 1.9127585349421855e-05, + "loss": 1.027, + "step": 12986 + }, + { + "epoch": 0.96, + "learning_rate": 1.9127422403296116e-05, + "loss": 1.0802, + "step": 12987 + }, + { + "epoch": 0.96, + "learning_rate": 1.9127259442648776e-05, + "loss": 1.0444, + "step": 12988 + }, + { + "epoch": 0.96, + "learning_rate": 1.9127096467480097e-05, + "loss": 0.9951, + "step": 12989 + }, + { + "epoch": 0.96, + "learning_rate": 1.912693347779034e-05, + "loss": 0.9956, + "step": 12990 + }, + { + "epoch": 0.96, + "learning_rate": 1.912677047357976e-05, + "loss": 0.994, + "step": 12991 + }, + { + "epoch": 0.96, + "learning_rate": 1.912660745484862e-05, + "loss": 1.0106, + "step": 12992 + }, + { + "epoch": 0.96, + "learning_rate": 1.9126444421597176e-05, + "loss": 1.1252, + "step": 12993 + }, + { + "epoch": 0.96, + "learning_rate": 1.9126281373825694e-05, + "loss": 0.9729, + "step": 12994 + }, + { + "epoch": 0.96, + "learning_rate": 1.9126118311534424e-05, + "loss": 1.0249, + "step": 12995 + }, + { + "epoch": 0.96, + "learning_rate": 1.912595523472363e-05, + "loss": 1.0573, + "step": 12996 + }, + { + "epoch": 0.96, + "learning_rate": 1.9125792143393575e-05, + "loss": 0.9639, + "step": 12997 + }, + { + "epoch": 0.96, + "learning_rate": 1.9125629037544516e-05, + "loss": 0.983, + "step": 12998 + }, + { + "epoch": 0.96, + "learning_rate": 1.912546591717671e-05, + "loss": 1.0569, + "step": 12999 + }, + { + "epoch": 0.96, + "learning_rate": 1.912530278229042e-05, + "loss": 1.1539, + "step": 13000 + }, + { + "epoch": 0.96, + "learning_rate": 1.9125139632885902e-05, + "loss": 1.0083, + "step": 13001 + }, + { + "epoch": 0.96, + "learning_rate": 1.9124976468963415e-05, + "loss": 1.0144, + "step": 13002 + }, + { + "epoch": 0.96, + "learning_rate": 1.9124813290523227e-05, + "loss": 0.8453, + "step": 13003 + }, + { + "epoch": 0.96, + "learning_rate": 1.9124650097565588e-05, + "loss": 1.1539, + "step": 13004 + }, + { + "epoch": 0.96, + "learning_rate": 1.912448689009076e-05, + "loss": 1.0415, + "step": 13005 + }, + { + "epoch": 0.96, + "learning_rate": 1.9124323668099008e-05, + "loss": 1.0616, + "step": 13006 + }, + { + "epoch": 0.96, + "learning_rate": 1.9124160431590584e-05, + "loss": 1.0315, + "step": 13007 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123997180565754e-05, + "loss": 0.9412, + "step": 13008 + }, + { + "epoch": 0.96, + "learning_rate": 1.912383391502477e-05, + "loss": 1.0639, + "step": 13009 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123670634967904e-05, + "loss": 1.0491, + "step": 13010 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123507340395402e-05, + "loss": 1.1155, + "step": 13011 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123344031307534e-05, + "loss": 1.0836, + "step": 13012 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123180707704557e-05, + "loss": 1.1108, + "step": 13013 + }, + { + "epoch": 0.96, + "learning_rate": 1.9123017369586726e-05, + "loss": 1.0038, + "step": 13014 + }, + { + "epoch": 0.96, + "learning_rate": 1.912285401695431e-05, + "loss": 1.1148, + "step": 13015 + }, + { + "epoch": 0.96, + "learning_rate": 1.912269064980756e-05, + "loss": 1.0548, + "step": 13016 + }, + { + "epoch": 0.96, + "learning_rate": 1.912252726814674e-05, + "loss": 0.9904, + "step": 13017 + }, + { + "epoch": 0.96, + "learning_rate": 1.9122363871972112e-05, + "loss": 1.0758, + "step": 13018 + }, + { + "epoch": 0.96, + "learning_rate": 1.9122200461283933e-05, + "loss": 1.0482, + "step": 13019 + }, + { + "epoch": 0.96, + "learning_rate": 1.912203703608246e-05, + "loss": 1.0466, + "step": 13020 + }, + { + "epoch": 0.96, + "learning_rate": 1.912187359636796e-05, + "loss": 0.9938, + "step": 13021 + }, + { + "epoch": 0.96, + "learning_rate": 1.912171014214069e-05, + "loss": 1.1343, + "step": 13022 + }, + { + "epoch": 0.96, + "learning_rate": 1.912154667340091e-05, + "loss": 1.0006, + "step": 13023 + }, + { + "epoch": 0.96, + "learning_rate": 1.9121383190148878e-05, + "loss": 1.1295, + "step": 13024 + }, + { + "epoch": 0.96, + "learning_rate": 1.9121219692384854e-05, + "loss": 0.8999, + "step": 13025 + }, + { + "epoch": 0.96, + "learning_rate": 1.9121056180109103e-05, + "loss": 1.066, + "step": 13026 + }, + { + "epoch": 0.96, + "learning_rate": 1.9120892653321882e-05, + "loss": 0.9163, + "step": 13027 + }, + { + "epoch": 0.96, + "learning_rate": 1.9120729112023452e-05, + "loss": 1.0336, + "step": 13028 + }, + { + "epoch": 0.96, + "learning_rate": 1.9120565556214067e-05, + "loss": 1.0217, + "step": 13029 + }, + { + "epoch": 0.96, + "learning_rate": 1.9120401985894e-05, + "loss": 1.1066, + "step": 13030 + }, + { + "epoch": 0.96, + "learning_rate": 1.91202384010635e-05, + "loss": 0.93, + "step": 13031 + }, + { + "epoch": 0.96, + "learning_rate": 1.912007480172283e-05, + "loss": 1.1102, + "step": 13032 + }, + { + "epoch": 0.96, + "learning_rate": 1.9119911187872256e-05, + "loss": 0.9773, + "step": 13033 + }, + { + "epoch": 0.96, + "learning_rate": 1.911974755951203e-05, + "loss": 1.0911, + "step": 13034 + }, + { + "epoch": 0.96, + "learning_rate": 1.9119583916642416e-05, + "loss": 0.9992, + "step": 13035 + }, + { + "epoch": 0.96, + "learning_rate": 1.9119420259263676e-05, + "loss": 0.9679, + "step": 13036 + }, + { + "epoch": 0.96, + "learning_rate": 1.911925658737607e-05, + "loss": 0.9663, + "step": 13037 + }, + { + "epoch": 0.96, + "learning_rate": 1.9119092900979855e-05, + "loss": 1.0507, + "step": 13038 + }, + { + "epoch": 0.96, + "learning_rate": 1.9118929200075295e-05, + "loss": 1.0778, + "step": 13039 + }, + { + "epoch": 0.96, + "learning_rate": 1.9118765484662645e-05, + "loss": 0.9821, + "step": 13040 + }, + { + "epoch": 0.96, + "learning_rate": 1.9118601754742175e-05, + "loss": 0.9929, + "step": 13041 + }, + { + "epoch": 0.96, + "learning_rate": 1.911843801031414e-05, + "loss": 1.0425, + "step": 13042 + }, + { + "epoch": 0.96, + "learning_rate": 1.9118274251378798e-05, + "loss": 1.021, + "step": 13043 + }, + { + "epoch": 0.96, + "learning_rate": 1.9118110477936412e-05, + "loss": 0.9423, + "step": 13044 + }, + { + "epoch": 0.96, + "learning_rate": 1.9117946689987243e-05, + "loss": 0.9439, + "step": 13045 + }, + { + "epoch": 0.96, + "learning_rate": 1.9117782887531553e-05, + "loss": 1.0585, + "step": 13046 + }, + { + "epoch": 0.96, + "learning_rate": 1.9117619070569598e-05, + "loss": 1.1457, + "step": 13047 + }, + { + "epoch": 0.96, + "learning_rate": 1.9117455239101644e-05, + "loss": 0.9821, + "step": 13048 + }, + { + "epoch": 0.96, + "learning_rate": 1.911729139312795e-05, + "loss": 1.1058, + "step": 13049 + }, + { + "epoch": 0.96, + "learning_rate": 1.9117127532648775e-05, + "loss": 0.95, + "step": 13050 + }, + { + "epoch": 0.96, + "learning_rate": 1.911696365766438e-05, + "loss": 0.9542, + "step": 13051 + }, + { + "epoch": 0.96, + "learning_rate": 1.9116799768175024e-05, + "loss": 1.0003, + "step": 13052 + }, + { + "epoch": 0.96, + "learning_rate": 1.9116635864180972e-05, + "loss": 0.9386, + "step": 13053 + }, + { + "epoch": 0.96, + "learning_rate": 1.9116471945682483e-05, + "loss": 0.9553, + "step": 13054 + }, + { + "epoch": 0.96, + "learning_rate": 1.9116308012679816e-05, + "loss": 1.0012, + "step": 13055 + }, + { + "epoch": 0.96, + "learning_rate": 1.9116144065173238e-05, + "loss": 0.9943, + "step": 13056 + }, + { + "epoch": 0.96, + "learning_rate": 1.9115980103163002e-05, + "loss": 0.9571, + "step": 13057 + }, + { + "epoch": 0.96, + "learning_rate": 1.911581612664937e-05, + "loss": 1.0876, + "step": 13058 + }, + { + "epoch": 0.96, + "learning_rate": 1.9115652135632607e-05, + "loss": 1.072, + "step": 13059 + }, + { + "epoch": 0.96, + "learning_rate": 1.9115488130112972e-05, + "loss": 1.1717, + "step": 13060 + }, + { + "epoch": 0.97, + "learning_rate": 1.9115324110090724e-05, + "loss": 0.8468, + "step": 13061 + }, + { + "epoch": 0.97, + "learning_rate": 1.911516007556613e-05, + "loss": 0.9897, + "step": 13062 + }, + { + "epoch": 0.97, + "learning_rate": 1.9114996026539445e-05, + "loss": 1.152, + "step": 13063 + }, + { + "epoch": 0.97, + "learning_rate": 1.9114831963010928e-05, + "loss": 1.0634, + "step": 13064 + }, + { + "epoch": 0.97, + "learning_rate": 1.9114667884980848e-05, + "loss": 1.0842, + "step": 13065 + }, + { + "epoch": 0.97, + "learning_rate": 1.9114503792449463e-05, + "loss": 0.975, + "step": 13066 + }, + { + "epoch": 0.97, + "learning_rate": 1.911433968541703e-05, + "loss": 1.0594, + "step": 13067 + }, + { + "epoch": 0.97, + "learning_rate": 1.911417556388381e-05, + "loss": 1.0854, + "step": 13068 + }, + { + "epoch": 0.97, + "learning_rate": 1.911401142785007e-05, + "loss": 1.1293, + "step": 13069 + }, + { + "epoch": 0.97, + "learning_rate": 1.911384727731607e-05, + "loss": 0.9842, + "step": 13070 + }, + { + "epoch": 0.97, + "learning_rate": 1.9113683112282067e-05, + "loss": 1.0614, + "step": 13071 + }, + { + "epoch": 0.97, + "learning_rate": 1.9113518932748326e-05, + "loss": 1.0072, + "step": 13072 + }, + { + "epoch": 0.97, + "learning_rate": 1.911335473871511e-05, + "loss": 1.0686, + "step": 13073 + }, + { + "epoch": 0.97, + "learning_rate": 1.911319053018267e-05, + "loss": 0.9727, + "step": 13074 + }, + { + "epoch": 0.97, + "learning_rate": 1.911302630715128e-05, + "loss": 1.065, + "step": 13075 + }, + { + "epoch": 0.97, + "learning_rate": 1.9112862069621193e-05, + "loss": 0.9733, + "step": 13076 + }, + { + "epoch": 0.97, + "learning_rate": 1.911269781759267e-05, + "loss": 0.8726, + "step": 13077 + }, + { + "epoch": 0.97, + "learning_rate": 1.9112533551065982e-05, + "loss": 0.9447, + "step": 13078 + }, + { + "epoch": 0.97, + "learning_rate": 1.9112369270041377e-05, + "loss": 0.9925, + "step": 13079 + }, + { + "epoch": 0.97, + "learning_rate": 1.9112204974519128e-05, + "loss": 1.0001, + "step": 13080 + }, + { + "epoch": 0.97, + "learning_rate": 1.911204066449949e-05, + "loss": 1.1375, + "step": 13081 + }, + { + "epoch": 0.97, + "learning_rate": 1.911187633998272e-05, + "loss": 0.9781, + "step": 13082 + }, + { + "epoch": 0.97, + "learning_rate": 1.9111712000969092e-05, + "loss": 0.9857, + "step": 13083 + }, + { + "epoch": 0.97, + "learning_rate": 1.9111547647458856e-05, + "loss": 0.9976, + "step": 13084 + }, + { + "epoch": 0.97, + "learning_rate": 1.911138327945228e-05, + "loss": 1.0882, + "step": 13085 + }, + { + "epoch": 0.97, + "learning_rate": 1.9111218896949626e-05, + "loss": 1.0093, + "step": 13086 + }, + { + "epoch": 0.97, + "learning_rate": 1.9111054499951153e-05, + "loss": 0.9919, + "step": 13087 + }, + { + "epoch": 0.97, + "learning_rate": 1.911089008845712e-05, + "loss": 0.9447, + "step": 13088 + }, + { + "epoch": 0.97, + "learning_rate": 1.911072566246779e-05, + "loss": 1.1462, + "step": 13089 + }, + { + "epoch": 0.97, + "learning_rate": 1.911056122198343e-05, + "loss": 1.1093, + "step": 13090 + }, + { + "epoch": 0.97, + "learning_rate": 1.9110396767004293e-05, + "loss": 1.1095, + "step": 13091 + }, + { + "epoch": 0.97, + "learning_rate": 1.911023229753065e-05, + "loss": 1.0269, + "step": 13092 + }, + { + "epoch": 0.97, + "learning_rate": 1.911006781356275e-05, + "loss": 0.9662, + "step": 13093 + }, + { + "epoch": 0.97, + "learning_rate": 1.910990331510087e-05, + "loss": 1.0239, + "step": 13094 + }, + { + "epoch": 0.97, + "learning_rate": 1.910973880214526e-05, + "loss": 1.0008, + "step": 13095 + }, + { + "epoch": 0.97, + "learning_rate": 1.9109574274696186e-05, + "loss": 0.9476, + "step": 13096 + }, + { + "epoch": 0.97, + "learning_rate": 1.9109409732753913e-05, + "loss": 1.0296, + "step": 13097 + }, + { + "epoch": 0.97, + "learning_rate": 1.9109245176318696e-05, + "loss": 1.0086, + "step": 13098 + }, + { + "epoch": 0.97, + "learning_rate": 1.9109080605390804e-05, + "loss": 0.9444, + "step": 13099 + }, + { + "epoch": 0.97, + "learning_rate": 1.9108916019970492e-05, + "loss": 1.0213, + "step": 13100 + }, + { + "epoch": 0.97, + "learning_rate": 1.9108751420058026e-05, + "loss": 0.9222, + "step": 13101 + }, + { + "epoch": 0.97, + "learning_rate": 1.9108586805653667e-05, + "loss": 1.0651, + "step": 13102 + }, + { + "epoch": 0.97, + "learning_rate": 1.9108422176757675e-05, + "loss": 1.0548, + "step": 13103 + }, + { + "epoch": 0.97, + "learning_rate": 1.9108257533370315e-05, + "loss": 0.9796, + "step": 13104 + }, + { + "epoch": 0.97, + "learning_rate": 1.910809287549185e-05, + "loss": 1.0938, + "step": 13105 + }, + { + "epoch": 0.97, + "learning_rate": 1.9107928203122536e-05, + "loss": 1.0306, + "step": 13106 + }, + { + "epoch": 0.97, + "learning_rate": 1.910776351626264e-05, + "loss": 1.0995, + "step": 13107 + }, + { + "epoch": 0.97, + "learning_rate": 1.910759881491242e-05, + "loss": 1.0813, + "step": 13108 + }, + { + "epoch": 0.97, + "learning_rate": 1.9107434099072147e-05, + "loss": 0.9729, + "step": 13109 + }, + { + "epoch": 0.97, + "learning_rate": 1.910726936874207e-05, + "loss": 1.0951, + "step": 13110 + }, + { + "epoch": 0.97, + "learning_rate": 1.910710462392246e-05, + "loss": 1.0792, + "step": 13111 + }, + { + "epoch": 0.97, + "learning_rate": 1.910693986461358e-05, + "loss": 1.1539, + "step": 13112 + }, + { + "epoch": 0.97, + "learning_rate": 1.9106775090815684e-05, + "loss": 1.0931, + "step": 13113 + }, + { + "epoch": 0.97, + "learning_rate": 1.910661030252904e-05, + "loss": 0.9599, + "step": 13114 + }, + { + "epoch": 0.97, + "learning_rate": 1.910644549975391e-05, + "loss": 1.0675, + "step": 13115 + }, + { + "epoch": 0.97, + "learning_rate": 1.9106280682490555e-05, + "loss": 1.1361, + "step": 13116 + }, + { + "epoch": 0.97, + "learning_rate": 1.910611585073924e-05, + "loss": 1.0398, + "step": 13117 + }, + { + "epoch": 0.97, + "learning_rate": 1.9105951004500224e-05, + "loss": 1.076, + "step": 13118 + }, + { + "epoch": 0.97, + "learning_rate": 1.910578614377377e-05, + "loss": 0.9054, + "step": 13119 + }, + { + "epoch": 0.97, + "learning_rate": 1.910562126856014e-05, + "loss": 0.9085, + "step": 13120 + }, + { + "epoch": 0.97, + "learning_rate": 1.9105456378859597e-05, + "loss": 1.0697, + "step": 13121 + }, + { + "epoch": 0.97, + "learning_rate": 1.9105291474672402e-05, + "loss": 0.9588, + "step": 13122 + }, + { + "epoch": 0.97, + "learning_rate": 1.9105126555998822e-05, + "loss": 0.959, + "step": 13123 + }, + { + "epoch": 0.97, + "learning_rate": 1.9104961622839116e-05, + "loss": 0.9892, + "step": 13124 + }, + { + "epoch": 0.97, + "learning_rate": 1.9104796675193543e-05, + "loss": 1.1085, + "step": 13125 + }, + { + "epoch": 0.97, + "learning_rate": 1.910463171306237e-05, + "loss": 1.0654, + "step": 13126 + }, + { + "epoch": 0.97, + "learning_rate": 1.9104466736445857e-05, + "loss": 0.943, + "step": 13127 + }, + { + "epoch": 0.97, + "learning_rate": 1.910430174534427e-05, + "loss": 1.0362, + "step": 13128 + }, + { + "epoch": 0.97, + "learning_rate": 1.9104136739757865e-05, + "loss": 1.0625, + "step": 13129 + }, + { + "epoch": 0.97, + "learning_rate": 1.9103971719686913e-05, + "loss": 0.9709, + "step": 13130 + }, + { + "epoch": 0.97, + "learning_rate": 1.910380668513167e-05, + "loss": 1.0695, + "step": 13131 + }, + { + "epoch": 0.97, + "learning_rate": 1.9103641636092406e-05, + "loss": 1.057, + "step": 13132 + }, + { + "epoch": 0.97, + "learning_rate": 1.910347657256937e-05, + "loss": 1.1179, + "step": 13133 + }, + { + "epoch": 0.97, + "learning_rate": 1.910331149456284e-05, + "loss": 1.1172, + "step": 13134 + }, + { + "epoch": 0.97, + "learning_rate": 1.9103146402073068e-05, + "loss": 1.114, + "step": 13135 + }, + { + "epoch": 0.97, + "learning_rate": 1.9102981295100324e-05, + "loss": 1.126, + "step": 13136 + }, + { + "epoch": 0.97, + "learning_rate": 1.910281617364486e-05, + "loss": 0.9638, + "step": 13137 + }, + { + "epoch": 0.97, + "learning_rate": 1.910265103770695e-05, + "loss": 0.9107, + "step": 13138 + }, + { + "epoch": 0.97, + "learning_rate": 1.910248588728685e-05, + "loss": 0.9534, + "step": 13139 + }, + { + "epoch": 0.97, + "learning_rate": 1.9102320722384832e-05, + "loss": 0.967, + "step": 13140 + }, + { + "epoch": 0.97, + "learning_rate": 1.9102155543001145e-05, + "loss": 0.9639, + "step": 13141 + }, + { + "epoch": 0.97, + "learning_rate": 1.9101990349136065e-05, + "loss": 1.1256, + "step": 13142 + }, + { + "epoch": 0.97, + "learning_rate": 1.910182514078984e-05, + "loss": 1.0974, + "step": 13143 + }, + { + "epoch": 0.97, + "learning_rate": 1.910165991796275e-05, + "loss": 1.008, + "step": 13144 + }, + { + "epoch": 0.97, + "learning_rate": 1.9101494680655043e-05, + "loss": 1.0314, + "step": 13145 + }, + { + "epoch": 0.97, + "learning_rate": 1.9101329428866993e-05, + "loss": 1.0665, + "step": 13146 + }, + { + "epoch": 0.97, + "learning_rate": 1.9101164162598854e-05, + "loss": 0.9927, + "step": 13147 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100998881850897e-05, + "loss": 1.0186, + "step": 13148 + }, + { + "epoch": 0.97, + "learning_rate": 1.910083358662338e-05, + "loss": 1.0649, + "step": 13149 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100668276916562e-05, + "loss": 0.9898, + "step": 13150 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100502952730716e-05, + "loss": 0.9631, + "step": 13151 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100337614066102e-05, + "loss": 0.8526, + "step": 13152 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100172260922974e-05, + "loss": 1.0787, + "step": 13153 + }, + { + "epoch": 0.97, + "learning_rate": 1.9100006893301607e-05, + "loss": 1.0289, + "step": 13154 + }, + { + "epoch": 0.97, + "learning_rate": 1.9099841511202258e-05, + "loss": 1.0843, + "step": 13155 + }, + { + "epoch": 0.97, + "learning_rate": 1.9099676114625195e-05, + "loss": 1.1195, + "step": 13156 + }, + { + "epoch": 0.97, + "learning_rate": 1.9099510703570672e-05, + "loss": 0.9622, + "step": 13157 + }, + { + "epoch": 0.97, + "learning_rate": 1.909934527803896e-05, + "loss": 0.986, + "step": 13158 + }, + { + "epoch": 0.97, + "learning_rate": 1.909917983803032e-05, + "loss": 0.993, + "step": 13159 + }, + { + "epoch": 0.97, + "learning_rate": 1.9099014383545012e-05, + "loss": 0.9349, + "step": 13160 + }, + { + "epoch": 0.97, + "learning_rate": 1.909884891458331e-05, + "loss": 1.0863, + "step": 13161 + }, + { + "epoch": 0.97, + "learning_rate": 1.9098683431145463e-05, + "loss": 0.9958, + "step": 13162 + }, + { + "epoch": 0.97, + "learning_rate": 1.9098517933231743e-05, + "loss": 1.0025, + "step": 13163 + }, + { + "epoch": 0.97, + "learning_rate": 1.909835242084241e-05, + "loss": 1.1068, + "step": 13164 + }, + { + "epoch": 0.97, + "learning_rate": 1.9098186893977727e-05, + "loss": 1.0176, + "step": 13165 + }, + { + "epoch": 0.97, + "learning_rate": 1.9098021352637963e-05, + "loss": 1.1115, + "step": 13166 + }, + { + "epoch": 0.97, + "learning_rate": 1.909785579682337e-05, + "loss": 1.0053, + "step": 13167 + }, + { + "epoch": 0.97, + "learning_rate": 1.9097690226534226e-05, + "loss": 1.0164, + "step": 13168 + }, + { + "epoch": 0.97, + "learning_rate": 1.9097524641770786e-05, + "loss": 0.9556, + "step": 13169 + }, + { + "epoch": 0.97, + "learning_rate": 1.9097359042533315e-05, + "loss": 1.0776, + "step": 13170 + }, + { + "epoch": 0.97, + "learning_rate": 1.9097193428822073e-05, + "loss": 0.9658, + "step": 13171 + }, + { + "epoch": 0.97, + "learning_rate": 1.9097027800637328e-05, + "loss": 0.971, + "step": 13172 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096862157979342e-05, + "loss": 1.0045, + "step": 13173 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096696500848378e-05, + "loss": 1.0236, + "step": 13174 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096530829244698e-05, + "loss": 0.9394, + "step": 13175 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096365143168568e-05, + "loss": 0.9887, + "step": 13176 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096199442620254e-05, + "loss": 0.9256, + "step": 13177 + }, + { + "epoch": 0.97, + "learning_rate": 1.9096033727600016e-05, + "loss": 1.1215, + "step": 13178 + }, + { + "epoch": 0.97, + "learning_rate": 1.909586799810812e-05, + "loss": 0.9548, + "step": 13179 + }, + { + "epoch": 0.97, + "learning_rate": 1.9095702254144822e-05, + "loss": 0.9924, + "step": 13180 + }, + { + "epoch": 0.97, + "learning_rate": 1.90955364957104e-05, + "loss": 0.9534, + "step": 13181 + }, + { + "epoch": 0.97, + "learning_rate": 1.9095370722805104e-05, + "loss": 1.0298, + "step": 13182 + }, + { + "epoch": 0.97, + "learning_rate": 1.9095204935429205e-05, + "loss": 1.0611, + "step": 13183 + }, + { + "epoch": 0.97, + "learning_rate": 1.9095039133582963e-05, + "loss": 1.1077, + "step": 13184 + }, + { + "epoch": 0.97, + "learning_rate": 1.9094873317266648e-05, + "loss": 0.9358, + "step": 13185 + }, + { + "epoch": 0.97, + "learning_rate": 1.9094707486480516e-05, + "loss": 0.9919, + "step": 13186 + }, + { + "epoch": 0.97, + "learning_rate": 1.9094541641224836e-05, + "loss": 0.941, + "step": 13187 + }, + { + "epoch": 0.97, + "learning_rate": 1.909437578149987e-05, + "loss": 0.9316, + "step": 13188 + }, + { + "epoch": 0.97, + "learning_rate": 1.9094209907305883e-05, + "loss": 1.1239, + "step": 13189 + }, + { + "epoch": 0.97, + "learning_rate": 1.909404401864314e-05, + "loss": 1.1117, + "step": 13190 + }, + { + "epoch": 0.97, + "learning_rate": 1.9093878115511902e-05, + "loss": 1.0717, + "step": 13191 + }, + { + "epoch": 0.97, + "learning_rate": 1.9093712197912432e-05, + "loss": 1.0086, + "step": 13192 + }, + { + "epoch": 0.97, + "learning_rate": 1.9093546265844996e-05, + "loss": 1.0432, + "step": 13193 + }, + { + "epoch": 0.97, + "learning_rate": 1.909338031930986e-05, + "loss": 1.0242, + "step": 13194 + }, + { + "epoch": 0.97, + "learning_rate": 1.9093214358307283e-05, + "loss": 1.0033, + "step": 13195 + }, + { + "epoch": 0.98, + "learning_rate": 1.9093048382837533e-05, + "loss": 1.069, + "step": 13196 + }, + { + "epoch": 0.98, + "learning_rate": 1.9092882392900877e-05, + "loss": 1.0624, + "step": 13197 + }, + { + "epoch": 0.98, + "learning_rate": 1.909271638849757e-05, + "loss": 0.9196, + "step": 13198 + }, + { + "epoch": 0.98, + "learning_rate": 1.9092550369627886e-05, + "loss": 0.9318, + "step": 13199 + }, + { + "epoch": 0.98, + "learning_rate": 1.909238433629208e-05, + "loss": 0.9577, + "step": 13200 + }, + { + "epoch": 0.98, + "learning_rate": 1.9092218288490425e-05, + "loss": 0.9725, + "step": 13201 + }, + { + "epoch": 0.98, + "learning_rate": 1.9092052226223176e-05, + "loss": 1.0042, + "step": 13202 + }, + { + "epoch": 0.98, + "learning_rate": 1.9091886149490606e-05, + "loss": 0.9348, + "step": 13203 + }, + { + "epoch": 0.98, + "learning_rate": 1.9091720058292974e-05, + "loss": 1.0502, + "step": 13204 + }, + { + "epoch": 0.98, + "learning_rate": 1.9091553952630546e-05, + "loss": 1.0079, + "step": 13205 + }, + { + "epoch": 0.98, + "learning_rate": 1.9091387832503584e-05, + "loss": 0.9973, + "step": 13206 + }, + { + "epoch": 0.98, + "learning_rate": 1.9091221697912355e-05, + "loss": 0.9525, + "step": 13207 + }, + { + "epoch": 0.98, + "learning_rate": 1.909105554885712e-05, + "loss": 1.0379, + "step": 13208 + }, + { + "epoch": 0.98, + "learning_rate": 1.9090889385338148e-05, + "loss": 1.1346, + "step": 13209 + }, + { + "epoch": 0.98, + "learning_rate": 1.9090723207355705e-05, + "loss": 1.0759, + "step": 13210 + }, + { + "epoch": 0.98, + "learning_rate": 1.9090557014910045e-05, + "loss": 1.0056, + "step": 13211 + }, + { + "epoch": 0.98, + "learning_rate": 1.909039080800144e-05, + "loss": 1.0385, + "step": 13212 + }, + { + "epoch": 0.98, + "learning_rate": 1.9090224586630155e-05, + "loss": 1.058, + "step": 13213 + }, + { + "epoch": 0.98, + "learning_rate": 1.9090058350796453e-05, + "loss": 1.0036, + "step": 13214 + }, + { + "epoch": 0.98, + "learning_rate": 1.9089892100500598e-05, + "loss": 0.9798, + "step": 13215 + }, + { + "epoch": 0.98, + "learning_rate": 1.9089725835742855e-05, + "loss": 1.0558, + "step": 13216 + }, + { + "epoch": 0.98, + "learning_rate": 1.9089559556523488e-05, + "loss": 1.0471, + "step": 13217 + }, + { + "epoch": 0.98, + "learning_rate": 1.9089393262842758e-05, + "loss": 1.1377, + "step": 13218 + }, + { + "epoch": 0.98, + "learning_rate": 1.908922695470094e-05, + "loss": 1.0511, + "step": 13219 + }, + { + "epoch": 0.98, + "learning_rate": 1.908906063209829e-05, + "loss": 1.0039, + "step": 13220 + }, + { + "epoch": 0.98, + "learning_rate": 1.9088894295035074e-05, + "loss": 0.9907, + "step": 13221 + }, + { + "epoch": 0.98, + "learning_rate": 1.9088727943511555e-05, + "loss": 1.0392, + "step": 13222 + }, + { + "epoch": 0.98, + "learning_rate": 1.9088561577528e-05, + "loss": 1.031, + "step": 13223 + }, + { + "epoch": 0.98, + "learning_rate": 1.9088395197084673e-05, + "loss": 0.9191, + "step": 13224 + }, + { + "epoch": 0.98, + "learning_rate": 1.9088228802181842e-05, + "loss": 1.0546, + "step": 13225 + }, + { + "epoch": 0.98, + "learning_rate": 1.908806239281977e-05, + "loss": 1.0087, + "step": 13226 + }, + { + "epoch": 0.98, + "learning_rate": 1.908789596899872e-05, + "loss": 1.043, + "step": 13227 + }, + { + "epoch": 0.98, + "learning_rate": 1.908772953071896e-05, + "loss": 1.0848, + "step": 13228 + }, + { + "epoch": 0.98, + "learning_rate": 1.9087563077980745e-05, + "loss": 0.8991, + "step": 13229 + }, + { + "epoch": 0.98, + "learning_rate": 1.908739661078435e-05, + "loss": 0.9498, + "step": 13230 + }, + { + "epoch": 0.98, + "learning_rate": 1.908723012913004e-05, + "loss": 1.1269, + "step": 13231 + }, + { + "epoch": 0.98, + "learning_rate": 1.9087063633018073e-05, + "loss": 1.0086, + "step": 13232 + }, + { + "epoch": 0.98, + "learning_rate": 1.908689712244872e-05, + "loss": 1.1173, + "step": 13233 + }, + { + "epoch": 0.98, + "learning_rate": 1.908673059742224e-05, + "loss": 1.085, + "step": 13234 + }, + { + "epoch": 0.98, + "learning_rate": 1.9086564057938906e-05, + "loss": 1.0055, + "step": 13235 + }, + { + "epoch": 0.98, + "learning_rate": 1.908639750399898e-05, + "loss": 1.1764, + "step": 13236 + }, + { + "epoch": 0.98, + "learning_rate": 1.9086230935602723e-05, + "loss": 1.0241, + "step": 13237 + }, + { + "epoch": 0.98, + "learning_rate": 1.9086064352750403e-05, + "loss": 1.0063, + "step": 13238 + }, + { + "epoch": 0.98, + "learning_rate": 1.9085897755442283e-05, + "loss": 1.0261, + "step": 13239 + }, + { + "epoch": 0.98, + "learning_rate": 1.9085731143678635e-05, + "loss": 0.9046, + "step": 13240 + }, + { + "epoch": 0.98, + "learning_rate": 1.9085564517459713e-05, + "loss": 0.9956, + "step": 13241 + }, + { + "epoch": 0.98, + "learning_rate": 1.908539787678579e-05, + "loss": 1.0856, + "step": 13242 + }, + { + "epoch": 0.98, + "learning_rate": 1.9085231221657128e-05, + "loss": 1.0088, + "step": 13243 + }, + { + "epoch": 0.98, + "learning_rate": 1.9085064552073994e-05, + "loss": 1.0084, + "step": 13244 + }, + { + "epoch": 0.98, + "learning_rate": 1.9084897868036653e-05, + "loss": 1.0142, + "step": 13245 + }, + { + "epoch": 0.98, + "learning_rate": 1.9084731169545366e-05, + "loss": 1.0911, + "step": 13246 + }, + { + "epoch": 0.98, + "learning_rate": 1.9084564456600407e-05, + "loss": 0.9893, + "step": 13247 + }, + { + "epoch": 0.98, + "learning_rate": 1.908439772920203e-05, + "loss": 0.991, + "step": 13248 + }, + { + "epoch": 0.98, + "learning_rate": 1.908423098735051e-05, + "loss": 0.8987, + "step": 13249 + }, + { + "epoch": 0.98, + "learning_rate": 1.908406423104611e-05, + "loss": 1.0142, + "step": 13250 + }, + { + "epoch": 0.98, + "learning_rate": 1.908389746028909e-05, + "loss": 1.0179, + "step": 13251 + }, + { + "epoch": 0.98, + "learning_rate": 1.908373067507972e-05, + "loss": 1.0892, + "step": 13252 + }, + { + "epoch": 0.98, + "learning_rate": 1.9083563875418265e-05, + "loss": 1.0072, + "step": 13253 + }, + { + "epoch": 0.98, + "learning_rate": 1.9083397061304988e-05, + "loss": 0.8875, + "step": 13254 + }, + { + "epoch": 0.98, + "learning_rate": 1.908323023274016e-05, + "loss": 1.123, + "step": 13255 + }, + { + "epoch": 0.98, + "learning_rate": 1.9083063389724042e-05, + "loss": 1.0941, + "step": 13256 + }, + { + "epoch": 0.98, + "learning_rate": 1.90828965322569e-05, + "loss": 0.9507, + "step": 13257 + }, + { + "epoch": 0.98, + "learning_rate": 1.9082729660338997e-05, + "loss": 0.9768, + "step": 13258 + }, + { + "epoch": 0.98, + "learning_rate": 1.9082562773970604e-05, + "loss": 1.0601, + "step": 13259 + }, + { + "epoch": 0.98, + "learning_rate": 1.908239587315198e-05, + "loss": 1.0614, + "step": 13260 + }, + { + "epoch": 0.98, + "learning_rate": 1.90822289578834e-05, + "loss": 0.9871, + "step": 13261 + }, + { + "epoch": 0.98, + "learning_rate": 1.9082062028165117e-05, + "loss": 0.9561, + "step": 13262 + }, + { + "epoch": 0.98, + "learning_rate": 1.908189508399741e-05, + "loss": 0.9691, + "step": 13263 + }, + { + "epoch": 0.98, + "learning_rate": 1.908172812538053e-05, + "loss": 1.0754, + "step": 13264 + }, + { + "epoch": 0.98, + "learning_rate": 1.9081561152314756e-05, + "loss": 1.0344, + "step": 13265 + }, + { + "epoch": 0.98, + "learning_rate": 1.9081394164800343e-05, + "loss": 1.0353, + "step": 13266 + }, + { + "epoch": 0.98, + "learning_rate": 1.9081227162837568e-05, + "loss": 1.0132, + "step": 13267 + }, + { + "epoch": 0.98, + "learning_rate": 1.9081060146426687e-05, + "loss": 0.9017, + "step": 13268 + }, + { + "epoch": 0.98, + "learning_rate": 1.9080893115567972e-05, + "loss": 0.9291, + "step": 13269 + }, + { + "epoch": 0.98, + "learning_rate": 1.9080726070261684e-05, + "loss": 0.9933, + "step": 13270 + }, + { + "epoch": 0.98, + "learning_rate": 1.908055901050809e-05, + "loss": 1.0722, + "step": 13271 + }, + { + "epoch": 0.98, + "learning_rate": 1.908039193630746e-05, + "loss": 1.0574, + "step": 13272 + }, + { + "epoch": 0.98, + "learning_rate": 1.9080224847660053e-05, + "loss": 1.0805, + "step": 13273 + }, + { + "epoch": 0.98, + "learning_rate": 1.9080057744566138e-05, + "loss": 1.0047, + "step": 13274 + }, + { + "epoch": 0.98, + "learning_rate": 1.9079890627025984e-05, + "loss": 1.0932, + "step": 13275 + }, + { + "epoch": 0.98, + "learning_rate": 1.907972349503985e-05, + "loss": 0.9616, + "step": 13276 + }, + { + "epoch": 0.98, + "learning_rate": 1.907955634860801e-05, + "loss": 1.0765, + "step": 13277 + }, + { + "epoch": 0.98, + "learning_rate": 1.907938918773072e-05, + "loss": 1.101, + "step": 13278 + }, + { + "epoch": 0.98, + "learning_rate": 1.907922201240826e-05, + "loss": 1.0219, + "step": 13279 + }, + { + "epoch": 0.98, + "learning_rate": 1.907905482264088e-05, + "loss": 1.0156, + "step": 13280 + }, + { + "epoch": 0.98, + "learning_rate": 1.9078887618428857e-05, + "loss": 1.016, + "step": 13281 + }, + { + "epoch": 0.98, + "learning_rate": 1.9078720399772455e-05, + "loss": 1.1457, + "step": 13282 + }, + { + "epoch": 0.98, + "learning_rate": 1.9078553166671935e-05, + "loss": 1.0398, + "step": 13283 + }, + { + "epoch": 0.98, + "learning_rate": 1.907838591912757e-05, + "loss": 1.0188, + "step": 13284 + }, + { + "epoch": 0.98, + "learning_rate": 1.907821865713962e-05, + "loss": 0.9536, + "step": 13285 + }, + { + "epoch": 0.98, + "learning_rate": 1.907805138070836e-05, + "loss": 0.9954, + "step": 13286 + }, + { + "epoch": 0.98, + "learning_rate": 1.9077884089834043e-05, + "loss": 1.0387, + "step": 13287 + }, + { + "epoch": 0.98, + "learning_rate": 1.9077716784516948e-05, + "loss": 0.9831, + "step": 13288 + }, + { + "epoch": 0.98, + "learning_rate": 1.907754946475733e-05, + "loss": 1.0692, + "step": 13289 + }, + { + "epoch": 0.98, + "learning_rate": 1.9077382130555464e-05, + "loss": 1.0463, + "step": 13290 + }, + { + "epoch": 0.98, + "learning_rate": 1.9077214781911613e-05, + "loss": 1.0679, + "step": 13291 + }, + { + "epoch": 0.98, + "learning_rate": 1.907704741882604e-05, + "loss": 0.9808, + "step": 13292 + }, + { + "epoch": 0.98, + "learning_rate": 1.9076880041299018e-05, + "loss": 1.0836, + "step": 13293 + }, + { + "epoch": 0.98, + "learning_rate": 1.907671264933081e-05, + "loss": 1.012, + "step": 13294 + }, + { + "epoch": 0.98, + "learning_rate": 1.907654524292168e-05, + "loss": 0.9533, + "step": 13295 + }, + { + "epoch": 0.98, + "learning_rate": 1.9076377822071896e-05, + "loss": 0.9802, + "step": 13296 + }, + { + "epoch": 0.98, + "learning_rate": 1.9076210386781725e-05, + "loss": 0.9454, + "step": 13297 + }, + { + "epoch": 0.98, + "learning_rate": 1.9076042937051435e-05, + "loss": 0.9027, + "step": 13298 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075875472881293e-05, + "loss": 0.9643, + "step": 13299 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075707994271556e-05, + "loss": 0.857, + "step": 13300 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075540501222504e-05, + "loss": 1.1512, + "step": 13301 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075372993734392e-05, + "loss": 1.0469, + "step": 13302 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075205471807493e-05, + "loss": 0.9738, + "step": 13303 + }, + { + "epoch": 0.98, + "learning_rate": 1.9075037935442073e-05, + "loss": 1.0403, + "step": 13304 + }, + { + "epoch": 0.98, + "learning_rate": 1.9074870384638394e-05, + "loss": 1.0015, + "step": 13305 + }, + { + "epoch": 0.98, + "learning_rate": 1.907470281939673e-05, + "loss": 0.9538, + "step": 13306 + }, + { + "epoch": 0.98, + "learning_rate": 1.9074535239717345e-05, + "loss": 1.0987, + "step": 13307 + }, + { + "epoch": 0.98, + "learning_rate": 1.9074367645600497e-05, + "loss": 1.0349, + "step": 13308 + }, + { + "epoch": 0.98, + "learning_rate": 1.9074200037046468e-05, + "loss": 1.0682, + "step": 13309 + }, + { + "epoch": 0.98, + "learning_rate": 1.9074032414055514e-05, + "loss": 1.0102, + "step": 13310 + }, + { + "epoch": 0.98, + "learning_rate": 1.9073864776627903e-05, + "loss": 1.0382, + "step": 13311 + }, + { + "epoch": 0.98, + "learning_rate": 1.9073697124763903e-05, + "loss": 1.0249, + "step": 13312 + }, + { + "epoch": 0.98, + "learning_rate": 1.907352945846378e-05, + "loss": 1.0076, + "step": 13313 + }, + { + "epoch": 0.98, + "learning_rate": 1.9073361777727803e-05, + "loss": 1.0525, + "step": 13314 + }, + { + "epoch": 0.98, + "learning_rate": 1.9073194082556238e-05, + "loss": 0.9266, + "step": 13315 + }, + { + "epoch": 0.98, + "learning_rate": 1.9073026372949348e-05, + "loss": 1.1105, + "step": 13316 + }, + { + "epoch": 0.98, + "learning_rate": 1.9072858648907404e-05, + "loss": 1.0734, + "step": 13317 + }, + { + "epoch": 0.98, + "learning_rate": 1.9072690910430674e-05, + "loss": 1.0077, + "step": 13318 + }, + { + "epoch": 0.98, + "learning_rate": 1.907252315751942e-05, + "loss": 1.0705, + "step": 13319 + }, + { + "epoch": 0.98, + "learning_rate": 1.9072355390173914e-05, + "loss": 1.0498, + "step": 13320 + }, + { + "epoch": 0.98, + "learning_rate": 1.9072187608394416e-05, + "loss": 1.0071, + "step": 13321 + }, + { + "epoch": 0.98, + "learning_rate": 1.90720198121812e-05, + "loss": 1.0945, + "step": 13322 + }, + { + "epoch": 0.98, + "learning_rate": 1.907185200153453e-05, + "loss": 1.017, + "step": 13323 + }, + { + "epoch": 0.98, + "learning_rate": 1.9071684176454675e-05, + "loss": 1.0488, + "step": 13324 + }, + { + "epoch": 0.98, + "learning_rate": 1.90715163369419e-05, + "loss": 1.0426, + "step": 13325 + }, + { + "epoch": 0.98, + "learning_rate": 1.9071348482996472e-05, + "loss": 1.0304, + "step": 13326 + }, + { + "epoch": 0.98, + "learning_rate": 1.9071180614618657e-05, + "loss": 0.9457, + "step": 13327 + }, + { + "epoch": 0.98, + "learning_rate": 1.907101273180872e-05, + "loss": 1.0462, + "step": 13328 + }, + { + "epoch": 0.98, + "learning_rate": 1.9070844834566937e-05, + "loss": 1.1212, + "step": 13329 + }, + { + "epoch": 0.98, + "learning_rate": 1.9070676922893568e-05, + "loss": 1.0376, + "step": 13330 + }, + { + "epoch": 0.99, + "learning_rate": 1.9070508996788882e-05, + "loss": 1.0208, + "step": 13331 + }, + { + "epoch": 0.99, + "learning_rate": 1.9070341056253147e-05, + "loss": 1.0522, + "step": 13332 + }, + { + "epoch": 0.99, + "learning_rate": 1.9070173101286626e-05, + "loss": 1.051, + "step": 13333 + }, + { + "epoch": 0.99, + "learning_rate": 1.9070005131889592e-05, + "loss": 0.9137, + "step": 13334 + }, + { + "epoch": 0.99, + "learning_rate": 1.906983714806231e-05, + "loss": 1.1231, + "step": 13335 + }, + { + "epoch": 0.99, + "learning_rate": 1.9069669149805044e-05, + "loss": 1.0078, + "step": 13336 + }, + { + "epoch": 0.99, + "learning_rate": 1.9069501137118067e-05, + "loss": 1.116, + "step": 13337 + }, + { + "epoch": 0.99, + "learning_rate": 1.9069333110001646e-05, + "loss": 1.1013, + "step": 13338 + }, + { + "epoch": 0.99, + "learning_rate": 1.906916506845604e-05, + "loss": 0.9641, + "step": 13339 + }, + { + "epoch": 0.99, + "learning_rate": 1.9068997012481523e-05, + "loss": 1.0689, + "step": 13340 + }, + { + "epoch": 0.99, + "learning_rate": 1.9068828942078365e-05, + "loss": 0.9617, + "step": 13341 + }, + { + "epoch": 0.99, + "learning_rate": 1.9068660857246826e-05, + "loss": 0.8718, + "step": 13342 + }, + { + "epoch": 0.99, + "learning_rate": 1.906849275798718e-05, + "loss": 0.951, + "step": 13343 + }, + { + "epoch": 0.99, + "learning_rate": 1.9068324644299688e-05, + "loss": 0.9454, + "step": 13344 + }, + { + "epoch": 0.99, + "learning_rate": 1.9068156516184627e-05, + "loss": 1.0541, + "step": 13345 + }, + { + "epoch": 0.99, + "learning_rate": 1.9067988373642254e-05, + "loss": 0.9922, + "step": 13346 + }, + { + "epoch": 0.99, + "learning_rate": 1.9067820216672846e-05, + "loss": 0.9689, + "step": 13347 + }, + { + "epoch": 0.99, + "learning_rate": 1.906765204527666e-05, + "loss": 1.1336, + "step": 13348 + }, + { + "epoch": 0.99, + "learning_rate": 1.9067483859453974e-05, + "loss": 1.0062, + "step": 13349 + }, + { + "epoch": 0.99, + "learning_rate": 1.906731565920505e-05, + "loss": 1.1586, + "step": 13350 + }, + { + "epoch": 0.99, + "learning_rate": 1.9067147444530153e-05, + "loss": 1.0191, + "step": 13351 + }, + { + "epoch": 0.99, + "learning_rate": 1.9066979215429557e-05, + "loss": 0.8526, + "step": 13352 + }, + { + "epoch": 0.99, + "learning_rate": 1.9066810971903526e-05, + "loss": 1.1063, + "step": 13353 + }, + { + "epoch": 0.99, + "learning_rate": 1.906664271395233e-05, + "loss": 0.9116, + "step": 13354 + }, + { + "epoch": 0.99, + "learning_rate": 1.9066474441576236e-05, + "loss": 1.0474, + "step": 13355 + }, + { + "epoch": 0.99, + "learning_rate": 1.9066306154775505e-05, + "loss": 1.1265, + "step": 13356 + }, + { + "epoch": 0.99, + "learning_rate": 1.9066137853550417e-05, + "loss": 1.0377, + "step": 13357 + }, + { + "epoch": 0.99, + "learning_rate": 1.906596953790123e-05, + "loss": 1.039, + "step": 13358 + }, + { + "epoch": 0.99, + "learning_rate": 1.906580120782822e-05, + "loss": 0.9576, + "step": 13359 + }, + { + "epoch": 0.99, + "learning_rate": 1.9065632863331643e-05, + "loss": 1.1227, + "step": 13360 + }, + { + "epoch": 0.99, + "learning_rate": 1.906546450441178e-05, + "loss": 0.994, + "step": 13361 + }, + { + "epoch": 0.99, + "learning_rate": 1.906529613106889e-05, + "loss": 0.9364, + "step": 13362 + }, + { + "epoch": 0.99, + "learning_rate": 1.906512774330324e-05, + "loss": 1.0827, + "step": 13363 + }, + { + "epoch": 0.99, + "learning_rate": 1.9064959341115105e-05, + "loss": 1.1044, + "step": 13364 + }, + { + "epoch": 0.99, + "learning_rate": 1.906479092450475e-05, + "loss": 0.9811, + "step": 13365 + }, + { + "epoch": 0.99, + "learning_rate": 1.906462249347244e-05, + "loss": 0.9662, + "step": 13366 + }, + { + "epoch": 0.99, + "learning_rate": 1.9064454048018448e-05, + "loss": 1.0692, + "step": 13367 + }, + { + "epoch": 0.99, + "learning_rate": 1.9064285588143038e-05, + "loss": 0.9667, + "step": 13368 + }, + { + "epoch": 0.99, + "learning_rate": 1.9064117113846482e-05, + "loss": 1.1008, + "step": 13369 + }, + { + "epoch": 0.99, + "learning_rate": 1.906394862512904e-05, + "loss": 1.0596, + "step": 13370 + }, + { + "epoch": 0.99, + "learning_rate": 1.906378012199099e-05, + "loss": 1.0898, + "step": 13371 + }, + { + "epoch": 0.99, + "learning_rate": 1.9063611604432592e-05, + "loss": 0.957, + "step": 13372 + }, + { + "epoch": 0.99, + "learning_rate": 1.9063443072454123e-05, + "loss": 0.9246, + "step": 13373 + }, + { + "epoch": 0.99, + "learning_rate": 1.906327452605584e-05, + "loss": 1.1361, + "step": 13374 + }, + { + "epoch": 0.99, + "learning_rate": 1.906310596523802e-05, + "loss": 1.031, + "step": 13375 + }, + { + "epoch": 0.99, + "learning_rate": 1.9062937390000926e-05, + "loss": 1.1465, + "step": 13376 + }, + { + "epoch": 0.99, + "learning_rate": 1.906276880034483e-05, + "loss": 1.0261, + "step": 13377 + }, + { + "epoch": 0.99, + "learning_rate": 1.9062600196269998e-05, + "loss": 1.0667, + "step": 13378 + }, + { + "epoch": 0.99, + "learning_rate": 1.9062431577776698e-05, + "loss": 1.0147, + "step": 13379 + }, + { + "epoch": 0.99, + "learning_rate": 1.90622629448652e-05, + "loss": 1.047, + "step": 13380 + }, + { + "epoch": 0.99, + "learning_rate": 1.9062094297535768e-05, + "loss": 1.0837, + "step": 13381 + }, + { + "epoch": 0.99, + "learning_rate": 1.906192563578868e-05, + "loss": 0.9688, + "step": 13382 + }, + { + "epoch": 0.99, + "learning_rate": 1.9061756959624193e-05, + "loss": 0.9692, + "step": 13383 + }, + { + "epoch": 0.99, + "learning_rate": 1.906158826904258e-05, + "loss": 1.004, + "step": 13384 + }, + { + "epoch": 0.99, + "learning_rate": 1.9061419564044114e-05, + "loss": 0.9608, + "step": 13385 + }, + { + "epoch": 0.99, + "learning_rate": 1.9061250844629054e-05, + "loss": 1.0434, + "step": 13386 + }, + { + "epoch": 0.99, + "learning_rate": 1.906108211079768e-05, + "loss": 0.987, + "step": 13387 + }, + { + "epoch": 0.99, + "learning_rate": 1.906091336255025e-05, + "loss": 0.9758, + "step": 13388 + }, + { + "epoch": 0.99, + "learning_rate": 1.9060744599887035e-05, + "loss": 1.0824, + "step": 13389 + }, + { + "epoch": 0.99, + "learning_rate": 1.906057582280831e-05, + "loss": 1.033, + "step": 13390 + }, + { + "epoch": 0.99, + "learning_rate": 1.906040703131433e-05, + "loss": 1.0617, + "step": 13391 + }, + { + "epoch": 0.99, + "learning_rate": 1.9060238225405382e-05, + "loss": 0.9436, + "step": 13392 + }, + { + "epoch": 0.99, + "learning_rate": 1.906006940508172e-05, + "loss": 0.9626, + "step": 13393 + }, + { + "epoch": 0.99, + "learning_rate": 1.9059900570343615e-05, + "loss": 1.1831, + "step": 13394 + }, + { + "epoch": 0.99, + "learning_rate": 1.905973172119134e-05, + "loss": 1.1427, + "step": 13395 + }, + { + "epoch": 0.99, + "learning_rate": 1.905956285762516e-05, + "loss": 1.0911, + "step": 13396 + }, + { + "epoch": 0.99, + "learning_rate": 1.9059393979645347e-05, + "loss": 0.9656, + "step": 13397 + }, + { + "epoch": 0.99, + "learning_rate": 1.905922508725217e-05, + "loss": 0.9158, + "step": 13398 + }, + { + "epoch": 0.99, + "learning_rate": 1.9059056180445893e-05, + "loss": 1.0517, + "step": 13399 + }, + { + "epoch": 0.99, + "learning_rate": 1.9058887259226786e-05, + "loss": 1.0545, + "step": 13400 + }, + { + "epoch": 0.99, + "learning_rate": 1.905871832359512e-05, + "loss": 0.9793, + "step": 13401 + }, + { + "epoch": 0.99, + "learning_rate": 1.905854937355116e-05, + "loss": 1.0571, + "step": 13402 + }, + { + "epoch": 0.99, + "learning_rate": 1.9058380409095184e-05, + "loss": 1.0715, + "step": 13403 + }, + { + "epoch": 0.99, + "learning_rate": 1.9058211430227448e-05, + "loss": 1.0906, + "step": 13404 + }, + { + "epoch": 0.99, + "learning_rate": 1.905804243694823e-05, + "loss": 1.093, + "step": 13405 + }, + { + "epoch": 0.99, + "learning_rate": 1.9057873429257797e-05, + "loss": 0.9671, + "step": 13406 + }, + { + "epoch": 0.99, + "learning_rate": 1.9057704407156415e-05, + "loss": 1.0329, + "step": 13407 + }, + { + "epoch": 0.99, + "learning_rate": 1.9057535370644356e-05, + "loss": 0.9723, + "step": 13408 + }, + { + "epoch": 0.99, + "learning_rate": 1.9057366319721888e-05, + "loss": 1.0368, + "step": 13409 + }, + { + "epoch": 0.99, + "learning_rate": 1.9057197254389283e-05, + "loss": 1.0598, + "step": 13410 + }, + { + "epoch": 0.99, + "learning_rate": 1.90570281746468e-05, + "loss": 1.0216, + "step": 13411 + }, + { + "epoch": 0.99, + "learning_rate": 1.905685908049472e-05, + "loss": 1.0698, + "step": 13412 + }, + { + "epoch": 0.99, + "learning_rate": 1.9056689971933302e-05, + "loss": 0.9459, + "step": 13413 + }, + { + "epoch": 0.99, + "learning_rate": 1.905652084896282e-05, + "loss": 0.9672, + "step": 13414 + }, + { + "epoch": 0.99, + "learning_rate": 1.9056351711583548e-05, + "loss": 1.0126, + "step": 13415 + }, + { + "epoch": 0.99, + "learning_rate": 1.905618255979575e-05, + "loss": 1.0338, + "step": 13416 + }, + { + "epoch": 0.99, + "learning_rate": 1.905601339359969e-05, + "loss": 0.9516, + "step": 13417 + }, + { + "epoch": 0.99, + "learning_rate": 1.9055844212995645e-05, + "loss": 0.9848, + "step": 13418 + }, + { + "epoch": 0.99, + "learning_rate": 1.9055675017983885e-05, + "loss": 1.009, + "step": 13419 + }, + { + "epoch": 0.99, + "learning_rate": 1.905550580856467e-05, + "loss": 0.9714, + "step": 13420 + }, + { + "epoch": 0.99, + "learning_rate": 1.9055336584738274e-05, + "loss": 1.0297, + "step": 13421 + }, + { + "epoch": 0.99, + "learning_rate": 1.905516734650497e-05, + "loss": 0.9558, + "step": 13422 + }, + { + "epoch": 0.99, + "learning_rate": 1.9054998093865028e-05, + "loss": 1.0859, + "step": 13423 + }, + { + "epoch": 0.99, + "learning_rate": 1.905482882681871e-05, + "loss": 1.0446, + "step": 13424 + }, + { + "epoch": 0.99, + "learning_rate": 1.9054659545366287e-05, + "loss": 1.0114, + "step": 13425 + }, + { + "epoch": 0.99, + "learning_rate": 1.905449024950803e-05, + "loss": 0.9948, + "step": 13426 + }, + { + "epoch": 0.99, + "learning_rate": 1.905432093924421e-05, + "loss": 1.0723, + "step": 13427 + }, + { + "epoch": 0.99, + "learning_rate": 1.9054151614575096e-05, + "loss": 0.9966, + "step": 13428 + }, + { + "epoch": 0.99, + "learning_rate": 1.9053982275500954e-05, + "loss": 1.0294, + "step": 13429 + }, + { + "epoch": 0.99, + "learning_rate": 1.9053812922022058e-05, + "loss": 0.9611, + "step": 13430 + }, + { + "epoch": 0.99, + "learning_rate": 1.9053643554138676e-05, + "loss": 0.9628, + "step": 13431 + }, + { + "epoch": 0.99, + "learning_rate": 1.9053474171851074e-05, + "loss": 1.2066, + "step": 13432 + }, + { + "epoch": 0.99, + "learning_rate": 1.9053304775159525e-05, + "loss": 1.005, + "step": 13433 + }, + { + "epoch": 0.99, + "learning_rate": 1.90531353640643e-05, + "loss": 1.1674, + "step": 13434 + }, + { + "epoch": 0.99, + "learning_rate": 1.905296593856566e-05, + "loss": 1.0105, + "step": 13435 + }, + { + "epoch": 0.99, + "learning_rate": 1.905279649866389e-05, + "loss": 1.0144, + "step": 13436 + }, + { + "epoch": 0.99, + "learning_rate": 1.9052627044359246e-05, + "loss": 0.9518, + "step": 13437 + }, + { + "epoch": 0.99, + "learning_rate": 1.9052457575652e-05, + "loss": 0.9686, + "step": 13438 + }, + { + "epoch": 0.99, + "learning_rate": 1.9052288092542423e-05, + "loss": 0.8966, + "step": 13439 + }, + { + "epoch": 0.99, + "learning_rate": 1.9052118595030785e-05, + "loss": 0.9898, + "step": 13440 + }, + { + "epoch": 0.99, + "learning_rate": 1.9051949083117358e-05, + "loss": 1.0725, + "step": 13441 + }, + { + "epoch": 0.99, + "learning_rate": 1.905177955680241e-05, + "loss": 0.9584, + "step": 13442 + }, + { + "epoch": 0.99, + "learning_rate": 1.905161001608621e-05, + "loss": 0.9695, + "step": 13443 + }, + { + "epoch": 0.99, + "learning_rate": 1.9051440460969023e-05, + "loss": 1.0286, + "step": 13444 + }, + { + "epoch": 0.99, + "learning_rate": 1.905127089145113e-05, + "loss": 1.018, + "step": 13445 + }, + { + "epoch": 0.99, + "learning_rate": 1.9051101307532795e-05, + "loss": 0.914, + "step": 13446 + }, + { + "epoch": 0.99, + "learning_rate": 1.9050931709214282e-05, + "loss": 0.9812, + "step": 13447 + }, + { + "epoch": 0.99, + "learning_rate": 1.905076209649587e-05, + "loss": 0.9433, + "step": 13448 + }, + { + "epoch": 0.99, + "learning_rate": 1.9050592469377824e-05, + "loss": 1.0376, + "step": 13449 + }, + { + "epoch": 0.99, + "learning_rate": 1.9050422827860414e-05, + "loss": 0.9817, + "step": 13450 + }, + { + "epoch": 0.99, + "learning_rate": 1.905025317194391e-05, + "loss": 1.0465, + "step": 13451 + }, + { + "epoch": 0.99, + "learning_rate": 1.9050083501628588e-05, + "loss": 1.0534, + "step": 13452 + }, + { + "epoch": 0.99, + "learning_rate": 1.9049913816914706e-05, + "loss": 1.0756, + "step": 13453 + }, + { + "epoch": 0.99, + "learning_rate": 1.9049744117802546e-05, + "loss": 1.1019, + "step": 13454 + }, + { + "epoch": 0.99, + "learning_rate": 1.9049574404292367e-05, + "loss": 1.0293, + "step": 13455 + }, + { + "epoch": 0.99, + "learning_rate": 1.904940467638445e-05, + "loss": 1.1392, + "step": 13456 + }, + { + "epoch": 0.99, + "learning_rate": 1.9049234934079057e-05, + "loss": 1.0435, + "step": 13457 + }, + { + "epoch": 0.99, + "learning_rate": 1.9049065177376463e-05, + "loss": 0.9255, + "step": 13458 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048895406276934e-05, + "loss": 1.0279, + "step": 13459 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048725620780738e-05, + "loss": 1.1648, + "step": 13460 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048555820888153e-05, + "loss": 0.89, + "step": 13461 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048386006599448e-05, + "loss": 0.9083, + "step": 13462 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048216177914885e-05, + "loss": 1.029, + "step": 13463 + }, + { + "epoch": 0.99, + "learning_rate": 1.9048046334834743e-05, + "loss": 1.0611, + "step": 13464 + }, + { + "epoch": 0.99, + "learning_rate": 1.9047876477359287e-05, + "loss": 0.9705, + "step": 13465 + }, + { + "epoch": 0.99, + "learning_rate": 1.904770660548879e-05, + "loss": 0.9719, + "step": 13466 + }, + { + "epoch": 1.0, + "learning_rate": 1.904753671922352e-05, + "loss": 1.0026, + "step": 13467 + }, + { + "epoch": 1.0, + "learning_rate": 1.904736681856375e-05, + "loss": 1.0688, + "step": 13468 + }, + { + "epoch": 1.0, + "learning_rate": 1.904719690350975e-05, + "loss": 0.9842, + "step": 13469 + }, + { + "epoch": 1.0, + "learning_rate": 1.9047026974061784e-05, + "loss": 0.9373, + "step": 13470 + }, + { + "epoch": 1.0, + "learning_rate": 1.904685703022013e-05, + "loss": 1.057, + "step": 13471 + }, + { + "epoch": 1.0, + "learning_rate": 1.9046687071985057e-05, + "loss": 0.9881, + "step": 13472 + }, + { + "epoch": 1.0, + "learning_rate": 1.9046517099356834e-05, + "loss": 1.0645, + "step": 13473 + }, + { + "epoch": 1.0, + "learning_rate": 1.904634711233573e-05, + "loss": 1.0708, + "step": 13474 + }, + { + "epoch": 1.0, + "learning_rate": 1.904617711092202e-05, + "loss": 0.9937, + "step": 13475 + }, + { + "epoch": 1.0, + "learning_rate": 1.9046007095115972e-05, + "loss": 1.0033, + "step": 13476 + }, + { + "epoch": 1.0, + "learning_rate": 1.9045837064917853e-05, + "loss": 1.0383, + "step": 13477 + }, + { + "epoch": 1.0, + "learning_rate": 1.904566702032794e-05, + "loss": 1.0231, + "step": 13478 + }, + { + "epoch": 1.0, + "learning_rate": 1.9045496961346496e-05, + "loss": 1.0312, + "step": 13479 + }, + { + "epoch": 1.0, + "learning_rate": 1.90453268879738e-05, + "loss": 1.0381, + "step": 13480 + }, + { + "epoch": 1.0, + "learning_rate": 1.9045156800210116e-05, + "loss": 1.0323, + "step": 13481 + }, + { + "epoch": 1.0, + "learning_rate": 1.9044986698055718e-05, + "loss": 0.997, + "step": 13482 + }, + { + "epoch": 1.0, + "learning_rate": 1.9044816581510874e-05, + "loss": 0.9478, + "step": 13483 + }, + { + "epoch": 1.0, + "learning_rate": 1.9044646450575856e-05, + "loss": 1.1374, + "step": 13484 + }, + { + "epoch": 1.0, + "learning_rate": 1.904447630525094e-05, + "loss": 0.9922, + "step": 13485 + }, + { + "epoch": 1.0, + "learning_rate": 1.9044306145536384e-05, + "loss": 1.1207, + "step": 13486 + }, + { + "epoch": 1.0, + "learning_rate": 1.904413597143247e-05, + "loss": 1.1152, + "step": 13487 + }, + { + "epoch": 1.0, + "learning_rate": 1.9043965782939465e-05, + "loss": 1.0731, + "step": 13488 + }, + { + "epoch": 1.0, + "learning_rate": 1.904379558005764e-05, + "loss": 0.9305, + "step": 13489 + }, + { + "epoch": 1.0, + "learning_rate": 1.9043625362787268e-05, + "loss": 1.1064, + "step": 13490 + }, + { + "epoch": 1.0, + "learning_rate": 1.9043455131128613e-05, + "loss": 0.9614, + "step": 13491 + }, + { + "epoch": 1.0, + "learning_rate": 1.9043284885081953e-05, + "loss": 1.0215, + "step": 13492 + }, + { + "epoch": 1.0, + "learning_rate": 1.9043114624647556e-05, + "loss": 0.9825, + "step": 13493 + }, + { + "epoch": 1.0, + "learning_rate": 1.9042944349825695e-05, + "loss": 0.9821, + "step": 13494 + }, + { + "epoch": 1.0, + "learning_rate": 1.9042774060616634e-05, + "loss": 1.0818, + "step": 13495 + }, + { + "epoch": 1.0, + "learning_rate": 1.904260375702065e-05, + "loss": 0.9071, + "step": 13496 + }, + { + "epoch": 1.0, + "learning_rate": 1.9042433439038016e-05, + "loss": 1.0041, + "step": 13497 + }, + { + "epoch": 1.0, + "learning_rate": 1.9042263106668998e-05, + "loss": 0.9631, + "step": 13498 + }, + { + "epoch": 1.0, + "learning_rate": 1.9042092759913867e-05, + "loss": 0.993, + "step": 13499 + }, + { + "epoch": 1.0, + "learning_rate": 1.9041922398772898e-05, + "loss": 1.0162, + "step": 13500 + }, + { + "epoch": 1.0, + "learning_rate": 1.904175202324636e-05, + "loss": 0.9469, + "step": 13501 + }, + { + "epoch": 1.0, + "learning_rate": 1.9041581633334523e-05, + "loss": 0.981, + "step": 13502 + }, + { + "epoch": 1.0, + "learning_rate": 1.9041411229037662e-05, + "loss": 1.0316, + "step": 13503 + }, + { + "epoch": 1.0, + "learning_rate": 1.904124081035604e-05, + "loss": 1.0814, + "step": 13504 + }, + { + "epoch": 1.0, + "learning_rate": 1.9041070377289936e-05, + "loss": 1.0287, + "step": 13505 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040899929839616e-05, + "loss": 1.0963, + "step": 13506 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040729468005356e-05, + "loss": 0.9233, + "step": 13507 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040558991787422e-05, + "loss": 1.0732, + "step": 13508 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040388501186094e-05, + "loss": 0.898, + "step": 13509 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040217996201634e-05, + "loss": 0.9934, + "step": 13510 + }, + { + "epoch": 1.0, + "learning_rate": 1.9040047476834314e-05, + "loss": 0.9445, + "step": 13511 + }, + { + "epoch": 1.0, + "learning_rate": 1.903987694308441e-05, + "loss": 0.9405, + "step": 13512 + }, + { + "epoch": 1.0, + "learning_rate": 1.903970639495219e-05, + "loss": 0.9434, + "step": 13513 + }, + { + "epoch": 1.0, + "learning_rate": 1.9039535832437928e-05, + "loss": 0.8663, + "step": 13514 + }, + { + "epoch": 1.0, + "learning_rate": 1.9039365255541892e-05, + "loss": 1.0915, + "step": 13515 + }, + { + "epoch": 1.0, + "learning_rate": 1.9039194664264358e-05, + "loss": 1.1408, + "step": 13516 + }, + { + "epoch": 1.0, + "learning_rate": 1.9039024058605594e-05, + "loss": 1.1344, + "step": 13517 + }, + { + "epoch": 1.0, + "learning_rate": 1.903885343856587e-05, + "loss": 1.011, + "step": 13518 + }, + { + "epoch": 1.0, + "learning_rate": 1.9038682804145462e-05, + "loss": 1.0, + "step": 13519 + }, + { + "epoch": 1.0, + "learning_rate": 1.9038512155344637e-05, + "loss": 0.9706, + "step": 13520 + }, + { + "epoch": 1.0, + "learning_rate": 1.903834149216367e-05, + "loss": 1.0185, + "step": 13521 + }, + { + "epoch": 1.0, + "learning_rate": 1.903817081460283e-05, + "loss": 0.8309, + "step": 13522 + }, + { + "epoch": 1.0, + "learning_rate": 1.9038000122662388e-05, + "loss": 1.0988, + "step": 13523 + }, + { + "epoch": 1.0, + "learning_rate": 1.9037829416342618e-05, + "loss": 1.0368, + "step": 13524 + }, + { + "epoch": 1.0, + "learning_rate": 1.9037658695643793e-05, + "loss": 1.1044, + "step": 13525 + }, + { + "epoch": 1.0, + "learning_rate": 1.903748796056618e-05, + "loss": 1.0755, + "step": 13526 + }, + { + "epoch": 1.0, + "learning_rate": 1.9037317211110055e-05, + "loss": 1.0806, + "step": 13527 + }, + { + "epoch": 1.0, + "learning_rate": 1.9037146447275684e-05, + "loss": 1.0713, + "step": 13528 + }, + { + "epoch": 1.0, + "learning_rate": 1.9036975669063345e-05, + "loss": 1.0294, + "step": 13529 + }, + { + "epoch": 1.0, + "learning_rate": 1.9036804876473307e-05, + "loss": 1.1013, + "step": 13530 + }, + { + "epoch": 1.0, + "learning_rate": 1.903663406950584e-05, + "loss": 1.0748, + "step": 13531 + }, + { + "epoch": 1.0, + "learning_rate": 1.9036463248161218e-05, + "loss": 0.9942, + "step": 13532 + }, + { + "epoch": 1.0, + "learning_rate": 1.9036292412439713e-05, + "loss": 0.878, + "step": 13533 + }, + { + "epoch": 1.0, + "learning_rate": 1.90361215623416e-05, + "loss": 0.99, + "step": 13534 + }, + { + "epoch": 1.0, + "learning_rate": 1.903595069786714e-05, + "loss": 0.9569, + "step": 13535 + }, + { + "epoch": 1.0, + "learning_rate": 1.9035779819016615e-05, + "loss": 1.0274, + "step": 13536 + }, + { + "epoch": 1.0, + "learning_rate": 1.903560892579029e-05, + "loss": 1.0374, + "step": 13537 + }, + { + "epoch": 1.0, + "learning_rate": 1.9035438018188448e-05, + "loss": 1.048, + "step": 13538 + }, + { + "epoch": 1.0, + "learning_rate": 1.903526709621135e-05, + "loss": 1.005, + "step": 13539 + }, + { + "epoch": 1.0, + "learning_rate": 1.903509615985927e-05, + "loss": 1.0149, + "step": 13540 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034925209132484e-05, + "loss": 0.9248, + "step": 13541 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034754244031256e-05, + "loss": 1.0045, + "step": 13542 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034583264555866e-05, + "loss": 1.1102, + "step": 13543 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034412270706585e-05, + "loss": 1.0215, + "step": 13544 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034241262483683e-05, + "loss": 1.0563, + "step": 13545 + }, + { + "epoch": 1.0, + "learning_rate": 1.9034070239887428e-05, + "loss": 0.9652, + "step": 13546 + }, + { + "epoch": 1.0, + "learning_rate": 1.90338992029181e-05, + "loss": 1.0262, + "step": 13547 + }, + { + "epoch": 1.0, + "learning_rate": 1.9033728151575967e-05, + "loss": 1.0175, + "step": 13548 + }, + { + "epoch": 1.0, + "learning_rate": 1.9033557085861303e-05, + "loss": 0.9585, + "step": 13549 + }, + { + "epoch": 1.0, + "learning_rate": 1.9033386005774378e-05, + "loss": 0.8477, + "step": 13550 + }, + { + "epoch": 1.0, + "learning_rate": 1.9033214911315468e-05, + "loss": 1.0495, + "step": 13551 + }, + { + "epoch": 1.0, + "learning_rate": 1.9033043802484838e-05, + "loss": 0.981, + "step": 13552 + }, + { + "epoch": 1.0, + "learning_rate": 1.9032872679282767e-05, + "loss": 1.0231, + "step": 13553 + }, + { + "epoch": 1.0, + "learning_rate": 1.903270154170952e-05, + "loss": 1.1256, + "step": 13554 + }, + { + "epoch": 1.0, + "learning_rate": 1.903253038976538e-05, + "loss": 1.0114, + "step": 13555 + }, + { + "epoch": 1.0, + "learning_rate": 1.9032359223450614e-05, + "loss": 1.0965, + "step": 13556 + }, + { + "epoch": 1.0, + "learning_rate": 1.903218804276549e-05, + "loss": 0.9347, + "step": 13557 + }, + { + "epoch": 1.0, + "learning_rate": 1.9032016847710286e-05, + "loss": 0.9992, + "step": 13558 + }, + { + "epoch": 1.0, + "learning_rate": 1.903184563828527e-05, + "loss": 0.9508, + "step": 13559 + }, + { + "epoch": 1.0, + "learning_rate": 1.903167441449072e-05, + "loss": 1.0351, + "step": 13560 + }, + { + "epoch": 1.0, + "learning_rate": 1.9031503176326903e-05, + "loss": 1.0446, + "step": 13561 + }, + { + "epoch": 1.0, + "learning_rate": 1.9031331923794095e-05, + "loss": 1.0531, + "step": 13562 + }, + { + "epoch": 1.0, + "learning_rate": 1.9031160656892566e-05, + "loss": 0.9497, + "step": 13563 + }, + { + "epoch": 1.0, + "learning_rate": 1.903098937562259e-05, + "loss": 0.8621, + "step": 13564 + }, + { + "epoch": 1.0, + "learning_rate": 1.903081807998444e-05, + "loss": 1.0335, + "step": 13565 + }, + { + "epoch": 1.0, + "learning_rate": 1.9030646769978383e-05, + "loss": 1.0448, + "step": 13566 + }, + { + "epoch": 1.0, + "learning_rate": 1.9030475445604703e-05, + "loss": 1.0308, + "step": 13567 + }, + { + "epoch": 1.0, + "learning_rate": 1.9030304106863663e-05, + "loss": 0.9777, + "step": 13568 + }, + { + "epoch": 1.0, + "learning_rate": 1.9030132753755538e-05, + "loss": 1.2217, + "step": 13569 + }, + { + "epoch": 1.0, + "learning_rate": 1.90299613862806e-05, + "loss": 0.9232, + "step": 13570 + }, + { + "epoch": 1.0, + "learning_rate": 1.9029790004439125e-05, + "loss": 0.9737, + "step": 13571 + }, + { + "epoch": 1.0, + "learning_rate": 1.9029618608231382e-05, + "loss": 0.9907, + "step": 13572 + }, + { + "epoch": 1.0, + "learning_rate": 1.9029447197657645e-05, + "loss": 1.0515, + "step": 13573 + }, + { + "epoch": 1.0, + "learning_rate": 1.9029275772718186e-05, + "loss": 1.1124, + "step": 13574 + }, + { + "epoch": 1.0, + "learning_rate": 1.902910433341328e-05, + "loss": 1.0293, + "step": 13575 + }, + { + "epoch": 1.0, + "learning_rate": 1.90289328797432e-05, + "loss": 1.014, + "step": 13576 + }, + { + "epoch": 1.0, + "learning_rate": 1.902876141170821e-05, + "loss": 0.9961, + "step": 13577 + }, + { + "epoch": 1.0, + "learning_rate": 1.9028589929308597e-05, + "loss": 1.0005, + "step": 13578 + }, + { + "epoch": 1.0, + "learning_rate": 1.9028418432544623e-05, + "loss": 1.0374, + "step": 13579 + }, + { + "epoch": 1.0, + "learning_rate": 1.9028246921416563e-05, + "loss": 1.1636, + "step": 13580 + }, + { + "epoch": 1.0, + "learning_rate": 1.9028075395924694e-05, + "loss": 0.9916, + "step": 13581 + }, + { + "epoch": 1.0, + "learning_rate": 1.9027903856069287e-05, + "loss": 1.0608, + "step": 13582 + }, + { + "epoch": 1.0, + "learning_rate": 1.902773230185061e-05, + "loss": 0.9965, + "step": 13583 + }, + { + "epoch": 1.0, + "learning_rate": 1.9027560733268946e-05, + "loss": 1.1383, + "step": 13584 + }, + { + "epoch": 1.0, + "learning_rate": 1.9027389150324557e-05, + "loss": 1.034, + "step": 13585 + }, + { + "epoch": 1.0, + "learning_rate": 1.9027217553017724e-05, + "loss": 0.9376, + "step": 13586 + }, + { + "epoch": 1.0, + "learning_rate": 1.9027045941348716e-05, + "loss": 1.0545, + "step": 13587 + }, + { + "epoch": 1.0, + "learning_rate": 1.9026874315317808e-05, + "loss": 1.1917, + "step": 13588 + }, + { + "epoch": 1.0, + "learning_rate": 1.902670267492527e-05, + "loss": 0.8817, + "step": 13589 + }, + { + "epoch": 1.0, + "learning_rate": 1.9026531020171378e-05, + "loss": 0.9865, + "step": 13590 + }, + { + "epoch": 1.0, + "learning_rate": 1.9026359351056406e-05, + "loss": 0.9379, + "step": 13591 + }, + { + "epoch": 1.0, + "learning_rate": 1.9026187667580624e-05, + "loss": 0.9664, + "step": 13592 + }, + { + "epoch": 1.0, + "learning_rate": 1.9026015969744306e-05, + "loss": 1.076, + "step": 13593 + }, + { + "epoch": 1.0, + "learning_rate": 1.9025844257547727e-05, + "loss": 1.0626, + "step": 13594 + }, + { + "epoch": 1.0, + "learning_rate": 1.9025672530991158e-05, + "loss": 0.9769, + "step": 13595 + }, + { + "epoch": 1.0, + "learning_rate": 1.9025500790074873e-05, + "loss": 1.0694, + "step": 13596 + }, + { + "epoch": 1.0, + "learning_rate": 1.9025329034799144e-05, + "loss": 1.0946, + "step": 13597 + }, + { + "epoch": 1.0, + "learning_rate": 1.902515726516425e-05, + "loss": 0.937, + "step": 13598 + }, + { + "epoch": 1.0, + "learning_rate": 1.9024985481170456e-05, + "loss": 0.9846, + "step": 13599 + }, + { + "epoch": 1.0, + "learning_rate": 1.9024813682818043e-05, + "loss": 1.0076, + "step": 13600 + }, + { + "epoch": 1.0, + "learning_rate": 1.9024641870107276e-05, + "loss": 0.8395, + "step": 13601 + }, + { + "epoch": 1.01, + "learning_rate": 1.9024470043038437e-05, + "loss": 1.0326, + "step": 13602 + }, + { + "epoch": 1.01, + "learning_rate": 1.9024298201611793e-05, + "loss": 0.8814, + "step": 13603 + }, + { + "epoch": 1.01, + "learning_rate": 1.902412634582762e-05, + "loss": 1.0092, + "step": 13604 + }, + { + "epoch": 1.01, + "learning_rate": 1.9023954475686192e-05, + "loss": 1.0308, + "step": 13605 + }, + { + "epoch": 1.01, + "learning_rate": 1.9023782591187777e-05, + "loss": 1.0447, + "step": 13606 + }, + { + "epoch": 1.01, + "learning_rate": 1.902361069233266e-05, + "loss": 1.0133, + "step": 13607 + }, + { + "epoch": 1.01, + "learning_rate": 1.9023438779121104e-05, + "loss": 1.0255, + "step": 13608 + }, + { + "epoch": 1.01, + "learning_rate": 1.9023266851553384e-05, + "loss": 1.0011, + "step": 13609 + }, + { + "epoch": 1.01, + "learning_rate": 1.9023094909629778e-05, + "loss": 0.9844, + "step": 13610 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022922953350558e-05, + "loss": 1.0703, + "step": 13611 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022750982715994e-05, + "loss": 1.0183, + "step": 13612 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022578997726366e-05, + "loss": 1.0293, + "step": 13613 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022406998381942e-05, + "loss": 1.0674, + "step": 13614 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022234984682996e-05, + "loss": 1.0275, + "step": 13615 + }, + { + "epoch": 1.01, + "learning_rate": 1.9022062956629807e-05, + "loss": 1.0263, + "step": 13616 + }, + { + "epoch": 1.01, + "learning_rate": 1.902189091422264e-05, + "loss": 0.9528, + "step": 13617 + }, + { + "epoch": 1.01, + "learning_rate": 1.9021718857461777e-05, + "loss": 1.0663, + "step": 13618 + }, + { + "epoch": 1.01, + "learning_rate": 1.9021546786347485e-05, + "loss": 0.9625, + "step": 13619 + }, + { + "epoch": 1.01, + "learning_rate": 1.9021374700880044e-05, + "loss": 0.8836, + "step": 13620 + }, + { + "epoch": 1.01, + "learning_rate": 1.9021202601059725e-05, + "loss": 1.0221, + "step": 13621 + }, + { + "epoch": 1.01, + "learning_rate": 1.90210304868868e-05, + "loss": 1.0756, + "step": 13622 + }, + { + "epoch": 1.01, + "learning_rate": 1.902085835836155e-05, + "loss": 1.0626, + "step": 13623 + }, + { + "epoch": 1.01, + "learning_rate": 1.9020686215484237e-05, + "loss": 1.168, + "step": 13624 + }, + { + "epoch": 1.01, + "learning_rate": 1.9020514058255145e-05, + "loss": 1.0154, + "step": 13625 + }, + { + "epoch": 1.01, + "learning_rate": 1.9020341886674544e-05, + "loss": 1.0535, + "step": 13626 + }, + { + "epoch": 1.01, + "learning_rate": 1.9020169700742703e-05, + "loss": 1.0507, + "step": 13627 + }, + { + "epoch": 1.01, + "learning_rate": 1.9019997500459906e-05, + "loss": 0.9443, + "step": 13628 + }, + { + "epoch": 1.01, + "learning_rate": 1.9019825285826422e-05, + "loss": 1.0067, + "step": 13629 + }, + { + "epoch": 1.01, + "learning_rate": 1.9019653056842523e-05, + "loss": 1.0298, + "step": 13630 + }, + { + "epoch": 1.01, + "learning_rate": 1.9019480813508483e-05, + "loss": 0.99, + "step": 13631 + }, + { + "epoch": 1.01, + "learning_rate": 1.9019308555824587e-05, + "loss": 0.9679, + "step": 13632 + }, + { + "epoch": 1.01, + "learning_rate": 1.901913628379109e-05, + "loss": 1.0258, + "step": 13633 + }, + { + "epoch": 1.01, + "learning_rate": 1.901896399740828e-05, + "loss": 0.9645, + "step": 13634 + }, + { + "epoch": 1.01, + "learning_rate": 1.9018791696676427e-05, + "loss": 1.0506, + "step": 13635 + }, + { + "epoch": 1.01, + "learning_rate": 1.9018619381595807e-05, + "loss": 1.1215, + "step": 13636 + }, + { + "epoch": 1.01, + "learning_rate": 1.9018447052166692e-05, + "loss": 1.029, + "step": 13637 + }, + { + "epoch": 1.01, + "learning_rate": 1.9018274708389357e-05, + "loss": 0.9967, + "step": 13638 + }, + { + "epoch": 1.01, + "learning_rate": 1.9018102350264077e-05, + "loss": 1.101, + "step": 13639 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017929977791123e-05, + "loss": 1.1113, + "step": 13640 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017757590970773e-05, + "loss": 1.0041, + "step": 13641 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017585189803297e-05, + "loss": 1.0505, + "step": 13642 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017412774288973e-05, + "loss": 1.1006, + "step": 13643 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017240344428077e-05, + "loss": 0.9108, + "step": 13644 + }, + { + "epoch": 1.01, + "learning_rate": 1.9017067900220878e-05, + "loss": 1.0237, + "step": 13645 + }, + { + "epoch": 1.01, + "learning_rate": 1.901689544166765e-05, + "loss": 1.0871, + "step": 13646 + }, + { + "epoch": 1.01, + "learning_rate": 1.901672296876868e-05, + "loss": 1.1009, + "step": 13647 + }, + { + "epoch": 1.01, + "learning_rate": 1.9016550481524224e-05, + "loss": 1.0402, + "step": 13648 + }, + { + "epoch": 1.01, + "learning_rate": 1.9016377979934568e-05, + "loss": 1.0226, + "step": 13649 + }, + { + "epoch": 1.01, + "learning_rate": 1.9016205463999983e-05, + "loss": 1.0231, + "step": 13650 + }, + { + "epoch": 1.01, + "learning_rate": 1.9016032933720745e-05, + "loss": 1.0316, + "step": 13651 + }, + { + "epoch": 1.01, + "learning_rate": 1.901586038909713e-05, + "loss": 0.9725, + "step": 13652 + }, + { + "epoch": 1.01, + "learning_rate": 1.9015687830129405e-05, + "loss": 0.9927, + "step": 13653 + }, + { + "epoch": 1.01, + "learning_rate": 1.9015515256817854e-05, + "loss": 0.8875, + "step": 13654 + }, + { + "epoch": 1.01, + "learning_rate": 1.9015342669162743e-05, + "loss": 0.9404, + "step": 13655 + }, + { + "epoch": 1.01, + "learning_rate": 1.9015170067164354e-05, + "loss": 0.9978, + "step": 13656 + }, + { + "epoch": 1.01, + "learning_rate": 1.9014997450822958e-05, + "loss": 1.0867, + "step": 13657 + }, + { + "epoch": 1.01, + "learning_rate": 1.901482482013883e-05, + "loss": 0.9457, + "step": 13658 + }, + { + "epoch": 1.01, + "learning_rate": 1.9014652175112242e-05, + "loss": 0.9171, + "step": 13659 + }, + { + "epoch": 1.01, + "learning_rate": 1.9014479515743476e-05, + "loss": 1.1, + "step": 13660 + }, + { + "epoch": 1.01, + "learning_rate": 1.9014306842032798e-05, + "loss": 1.0798, + "step": 13661 + }, + { + "epoch": 1.01, + "learning_rate": 1.901413415398049e-05, + "loss": 0.9204, + "step": 13662 + }, + { + "epoch": 1.01, + "learning_rate": 1.901396145158682e-05, + "loss": 1.0876, + "step": 13663 + }, + { + "epoch": 1.01, + "learning_rate": 1.901378873485207e-05, + "loss": 1.122, + "step": 13664 + }, + { + "epoch": 1.01, + "learning_rate": 1.901361600377651e-05, + "loss": 1.0721, + "step": 13665 + }, + { + "epoch": 1.01, + "learning_rate": 1.9013443258360415e-05, + "loss": 1.0011, + "step": 13666 + }, + { + "epoch": 1.01, + "learning_rate": 1.9013270498604058e-05, + "loss": 1.068, + "step": 13667 + }, + { + "epoch": 1.01, + "learning_rate": 1.901309772450772e-05, + "loss": 1.0054, + "step": 13668 + }, + { + "epoch": 1.01, + "learning_rate": 1.9012924936071676e-05, + "loss": 1.0332, + "step": 13669 + }, + { + "epoch": 1.01, + "learning_rate": 1.9012752133296193e-05, + "loss": 1.0184, + "step": 13670 + }, + { + "epoch": 1.01, + "learning_rate": 1.901257931618155e-05, + "loss": 0.9613, + "step": 13671 + }, + { + "epoch": 1.01, + "learning_rate": 1.9012406484728022e-05, + "loss": 0.9985, + "step": 13672 + }, + { + "epoch": 1.01, + "learning_rate": 1.9012233638935886e-05, + "loss": 1.0811, + "step": 13673 + }, + { + "epoch": 1.01, + "learning_rate": 1.9012060778805413e-05, + "loss": 1.0241, + "step": 13674 + }, + { + "epoch": 1.01, + "learning_rate": 1.9011887904336882e-05, + "loss": 1.0119, + "step": 13675 + }, + { + "epoch": 1.01, + "learning_rate": 1.9011715015530567e-05, + "loss": 1.0745, + "step": 13676 + }, + { + "epoch": 1.01, + "learning_rate": 1.901154211238674e-05, + "loss": 1.0906, + "step": 13677 + }, + { + "epoch": 1.01, + "learning_rate": 1.9011369194905684e-05, + "loss": 1.0797, + "step": 13678 + }, + { + "epoch": 1.01, + "learning_rate": 1.9011196263087662e-05, + "loss": 1.0641, + "step": 13679 + }, + { + "epoch": 1.01, + "learning_rate": 1.901102331693296e-05, + "loss": 1.0614, + "step": 13680 + }, + { + "epoch": 1.01, + "learning_rate": 1.9010850356441847e-05, + "loss": 1.1003, + "step": 13681 + }, + { + "epoch": 1.01, + "learning_rate": 1.90106773816146e-05, + "loss": 1.0689, + "step": 13682 + }, + { + "epoch": 1.01, + "learning_rate": 1.9010504392451494e-05, + "loss": 1.0337, + "step": 13683 + }, + { + "epoch": 1.01, + "learning_rate": 1.9010331388952804e-05, + "loss": 0.9772, + "step": 13684 + }, + { + "epoch": 1.01, + "learning_rate": 1.9010158371118807e-05, + "loss": 0.901, + "step": 13685 + }, + { + "epoch": 1.01, + "learning_rate": 1.9009985338949773e-05, + "loss": 1.016, + "step": 13686 + }, + { + "epoch": 1.01, + "learning_rate": 1.9009812292445987e-05, + "loss": 1.0323, + "step": 13687 + }, + { + "epoch": 1.01, + "learning_rate": 1.9009639231607714e-05, + "loss": 1.0303, + "step": 13688 + }, + { + "epoch": 1.01, + "learning_rate": 1.9009466156435237e-05, + "loss": 1.0494, + "step": 13689 + }, + { + "epoch": 1.01, + "learning_rate": 1.9009293066928825e-05, + "loss": 0.9909, + "step": 13690 + }, + { + "epoch": 1.01, + "learning_rate": 1.900911996308876e-05, + "loss": 0.9397, + "step": 13691 + }, + { + "epoch": 1.01, + "learning_rate": 1.9008946844915313e-05, + "loss": 1.0569, + "step": 13692 + }, + { + "epoch": 1.01, + "learning_rate": 1.9008773712408762e-05, + "loss": 0.9815, + "step": 13693 + }, + { + "epoch": 1.01, + "learning_rate": 1.9008600565569377e-05, + "loss": 1.1519, + "step": 13694 + }, + { + "epoch": 1.01, + "learning_rate": 1.900842740439744e-05, + "loss": 0.9817, + "step": 13695 + }, + { + "epoch": 1.01, + "learning_rate": 1.9008254228893225e-05, + "loss": 1.0328, + "step": 13696 + }, + { + "epoch": 1.01, + "learning_rate": 1.9008081039057004e-05, + "loss": 1.0327, + "step": 13697 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007907834889057e-05, + "loss": 0.9797, + "step": 13698 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007734616389654e-05, + "loss": 1.0996, + "step": 13699 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007561383559077e-05, + "loss": 1.0077, + "step": 13700 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007388136397598e-05, + "loss": 1.0433, + "step": 13701 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007214874905493e-05, + "loss": 1.0486, + "step": 13702 + }, + { + "epoch": 1.01, + "learning_rate": 1.9007041599083037e-05, + "loss": 0.9931, + "step": 13703 + }, + { + "epoch": 1.01, + "learning_rate": 1.9006868308930512e-05, + "loss": 1.0073, + "step": 13704 + }, + { + "epoch": 1.01, + "learning_rate": 1.9006695004448182e-05, + "loss": 1.0556, + "step": 13705 + }, + { + "epoch": 1.01, + "learning_rate": 1.900652168563633e-05, + "loss": 1.0236, + "step": 13706 + }, + { + "epoch": 1.01, + "learning_rate": 1.9006348352495237e-05, + "loss": 0.9937, + "step": 13707 + }, + { + "epoch": 1.01, + "learning_rate": 1.9006175005025165e-05, + "loss": 1.1183, + "step": 13708 + }, + { + "epoch": 1.01, + "learning_rate": 1.90060016432264e-05, + "loss": 1.08, + "step": 13709 + }, + { + "epoch": 1.01, + "learning_rate": 1.900582826709922e-05, + "loss": 0.9687, + "step": 13710 + }, + { + "epoch": 1.01, + "learning_rate": 1.900565487664389e-05, + "loss": 1.0148, + "step": 13711 + }, + { + "epoch": 1.01, + "learning_rate": 1.9005481471860694e-05, + "loss": 0.9715, + "step": 13712 + }, + { + "epoch": 1.01, + "learning_rate": 1.9005308052749906e-05, + "loss": 1.0417, + "step": 13713 + }, + { + "epoch": 1.01, + "learning_rate": 1.90051346193118e-05, + "loss": 0.9864, + "step": 13714 + }, + { + "epoch": 1.01, + "learning_rate": 1.9004961171546657e-05, + "loss": 1.1659, + "step": 13715 + }, + { + "epoch": 1.01, + "learning_rate": 1.9004787709454745e-05, + "loss": 0.983, + "step": 13716 + }, + { + "epoch": 1.01, + "learning_rate": 1.900461423303635e-05, + "loss": 1.0824, + "step": 13717 + }, + { + "epoch": 1.01, + "learning_rate": 1.900444074229174e-05, + "loss": 1.1147, + "step": 13718 + }, + { + "epoch": 1.01, + "learning_rate": 1.9004267237221192e-05, + "loss": 0.9941, + "step": 13719 + }, + { + "epoch": 1.01, + "learning_rate": 1.9004093717824984e-05, + "loss": 0.9568, + "step": 13720 + }, + { + "epoch": 1.01, + "learning_rate": 1.900392018410339e-05, + "loss": 1.0293, + "step": 13721 + }, + { + "epoch": 1.01, + "learning_rate": 1.9003746636056696e-05, + "loss": 0.9837, + "step": 13722 + }, + { + "epoch": 1.01, + "learning_rate": 1.9003573073685163e-05, + "loss": 1.071, + "step": 13723 + }, + { + "epoch": 1.01, + "learning_rate": 1.9003399496989076e-05, + "loss": 1.0541, + "step": 13724 + }, + { + "epoch": 1.01, + "learning_rate": 1.9003225905968707e-05, + "loss": 1.1361, + "step": 13725 + }, + { + "epoch": 1.01, + "learning_rate": 1.9003052300624336e-05, + "loss": 1.1996, + "step": 13726 + }, + { + "epoch": 1.01, + "learning_rate": 1.900287868095624e-05, + "loss": 1.0177, + "step": 13727 + }, + { + "epoch": 1.01, + "learning_rate": 1.900270504696469e-05, + "loss": 1.1249, + "step": 13728 + }, + { + "epoch": 1.01, + "learning_rate": 1.9002531398649966e-05, + "loss": 1.025, + "step": 13729 + }, + { + "epoch": 1.01, + "learning_rate": 1.900235773601234e-05, + "loss": 0.9208, + "step": 13730 + }, + { + "epoch": 1.01, + "learning_rate": 1.9002184059052097e-05, + "loss": 0.9996, + "step": 13731 + }, + { + "epoch": 1.01, + "learning_rate": 1.9002010367769504e-05, + "loss": 1.0086, + "step": 13732 + }, + { + "epoch": 1.01, + "learning_rate": 1.9001836662164845e-05, + "loss": 0.9237, + "step": 13733 + }, + { + "epoch": 1.01, + "learning_rate": 1.900166294223839e-05, + "loss": 1.0452, + "step": 13734 + }, + { + "epoch": 1.01, + "learning_rate": 1.900148920799042e-05, + "loss": 1.073, + "step": 13735 + }, + { + "epoch": 1.01, + "learning_rate": 1.900131545942121e-05, + "loss": 0.9934, + "step": 13736 + }, + { + "epoch": 1.02, + "learning_rate": 1.900114169653103e-05, + "loss": 1.0401, + "step": 13737 + }, + { + "epoch": 1.02, + "learning_rate": 1.9000967919320172e-05, + "loss": 0.9855, + "step": 13738 + }, + { + "epoch": 1.02, + "learning_rate": 1.9000794127788892e-05, + "loss": 1.0404, + "step": 13739 + }, + { + "epoch": 1.02, + "learning_rate": 1.9000620321937484e-05, + "loss": 1.0347, + "step": 13740 + }, + { + "epoch": 1.02, + "learning_rate": 1.900044650176622e-05, + "loss": 1.0683, + "step": 13741 + }, + { + "epoch": 1.02, + "learning_rate": 1.9000272667275366e-05, + "loss": 1.0201, + "step": 13742 + }, + { + "epoch": 1.02, + "learning_rate": 1.9000098818465215e-05, + "loss": 0.9892, + "step": 13743 + }, + { + "epoch": 1.02, + "learning_rate": 1.899992495533603e-05, + "loss": 1.0072, + "step": 13744 + }, + { + "epoch": 1.02, + "learning_rate": 1.8999751077888095e-05, + "loss": 0.9439, + "step": 13745 + }, + { + "epoch": 1.02, + "learning_rate": 1.8999577186121688e-05, + "loss": 0.9352, + "step": 13746 + }, + { + "epoch": 1.02, + "learning_rate": 1.899940328003708e-05, + "loss": 1.1659, + "step": 13747 + }, + { + "epoch": 1.02, + "learning_rate": 1.899922935963455e-05, + "loss": 1.024, + "step": 13748 + }, + { + "epoch": 1.02, + "learning_rate": 1.8999055424914372e-05, + "loss": 1.0052, + "step": 13749 + }, + { + "epoch": 1.02, + "learning_rate": 1.899888147587683e-05, + "loss": 0.9896, + "step": 13750 + }, + { + "epoch": 1.02, + "learning_rate": 1.8998707512522195e-05, + "loss": 0.8888, + "step": 13751 + }, + { + "epoch": 1.02, + "learning_rate": 1.8998533534850747e-05, + "loss": 1.0522, + "step": 13752 + }, + { + "epoch": 1.02, + "learning_rate": 1.8998359542862754e-05, + "loss": 0.8609, + "step": 13753 + }, + { + "epoch": 1.02, + "learning_rate": 1.8998185536558505e-05, + "loss": 1.1209, + "step": 13754 + }, + { + "epoch": 1.02, + "learning_rate": 1.899801151593827e-05, + "loss": 1.0357, + "step": 13755 + }, + { + "epoch": 1.02, + "learning_rate": 1.899783748100233e-05, + "loss": 1.0724, + "step": 13756 + }, + { + "epoch": 1.02, + "learning_rate": 1.8997663431750957e-05, + "loss": 0.9545, + "step": 13757 + }, + { + "epoch": 1.02, + "learning_rate": 1.899748936818443e-05, + "loss": 1.0315, + "step": 13758 + }, + { + "epoch": 1.02, + "learning_rate": 1.899731529030303e-05, + "loss": 0.9539, + "step": 13759 + }, + { + "epoch": 1.02, + "learning_rate": 1.8997141198107027e-05, + "loss": 1.0445, + "step": 13760 + }, + { + "epoch": 1.02, + "learning_rate": 1.8996967091596697e-05, + "loss": 1.0769, + "step": 13761 + }, + { + "epoch": 1.02, + "learning_rate": 1.899679297077233e-05, + "loss": 1.0041, + "step": 13762 + }, + { + "epoch": 1.02, + "learning_rate": 1.8996618835634187e-05, + "loss": 1.097, + "step": 13763 + }, + { + "epoch": 1.02, + "learning_rate": 1.8996444686182556e-05, + "loss": 1.0593, + "step": 13764 + }, + { + "epoch": 1.02, + "learning_rate": 1.899627052241771e-05, + "loss": 1.1551, + "step": 13765 + }, + { + "epoch": 1.02, + "learning_rate": 1.899609634433992e-05, + "loss": 1.1213, + "step": 13766 + }, + { + "epoch": 1.02, + "learning_rate": 1.8995922151949477e-05, + "loss": 1.0084, + "step": 13767 + }, + { + "epoch": 1.02, + "learning_rate": 1.8995747945246648e-05, + "loss": 1.17, + "step": 13768 + }, + { + "epoch": 1.02, + "learning_rate": 1.8995573724231714e-05, + "loss": 1.0408, + "step": 13769 + }, + { + "epoch": 1.02, + "learning_rate": 1.899539948890495e-05, + "loss": 1.1052, + "step": 13770 + }, + { + "epoch": 1.02, + "learning_rate": 1.8995225239266634e-05, + "loss": 1.0353, + "step": 13771 + }, + { + "epoch": 1.02, + "learning_rate": 1.8995050975317043e-05, + "loss": 1.0341, + "step": 13772 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994876697056454e-05, + "loss": 0.9954, + "step": 13773 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994702404485146e-05, + "loss": 0.9945, + "step": 13774 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994528097603396e-05, + "loss": 0.9309, + "step": 13775 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994353776411477e-05, + "loss": 1.0328, + "step": 13776 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994179440909673e-05, + "loss": 1.0027, + "step": 13777 + }, + { + "epoch": 1.02, + "learning_rate": 1.8994005091098256e-05, + "loss": 1.0275, + "step": 13778 + }, + { + "epoch": 1.02, + "learning_rate": 1.899383072697751e-05, + "loss": 1.0811, + "step": 13779 + }, + { + "epoch": 1.02, + "learning_rate": 1.8993656348547702e-05, + "loss": 0.9796, + "step": 13780 + }, + { + "epoch": 1.02, + "learning_rate": 1.899348195580912e-05, + "loss": 1.0243, + "step": 13781 + }, + { + "epoch": 1.02, + "learning_rate": 1.8993307548762034e-05, + "loss": 1.0173, + "step": 13782 + }, + { + "epoch": 1.02, + "learning_rate": 1.8993133127406725e-05, + "loss": 1.1057, + "step": 13783 + }, + { + "epoch": 1.02, + "learning_rate": 1.899295869174347e-05, + "loss": 0.935, + "step": 13784 + }, + { + "epoch": 1.02, + "learning_rate": 1.8992784241772546e-05, + "loss": 1.0386, + "step": 13785 + }, + { + "epoch": 1.02, + "learning_rate": 1.8992609777494232e-05, + "loss": 1.032, + "step": 13786 + }, + { + "epoch": 1.02, + "learning_rate": 1.8992435298908803e-05, + "loss": 1.0436, + "step": 13787 + }, + { + "epoch": 1.02, + "learning_rate": 1.8992260806016544e-05, + "loss": 0.9464, + "step": 13788 + }, + { + "epoch": 1.02, + "learning_rate": 1.899208629881772e-05, + "loss": 0.9813, + "step": 13789 + }, + { + "epoch": 1.02, + "learning_rate": 1.899191177731262e-05, + "loss": 1.0133, + "step": 13790 + }, + { + "epoch": 1.02, + "learning_rate": 1.8991737241501512e-05, + "loss": 1.0562, + "step": 13791 + }, + { + "epoch": 1.02, + "learning_rate": 1.8991562691384682e-05, + "loss": 1.0244, + "step": 13792 + }, + { + "epoch": 1.02, + "learning_rate": 1.89913881269624e-05, + "loss": 0.9683, + "step": 13793 + }, + { + "epoch": 1.02, + "learning_rate": 1.8991213548234954e-05, + "loss": 1.0322, + "step": 13794 + }, + { + "epoch": 1.02, + "learning_rate": 1.8991038955202614e-05, + "loss": 1.0234, + "step": 13795 + }, + { + "epoch": 1.02, + "learning_rate": 1.899086434786566e-05, + "loss": 1.1592, + "step": 13796 + }, + { + "epoch": 1.02, + "learning_rate": 1.8990689726224366e-05, + "loss": 1.0681, + "step": 13797 + }, + { + "epoch": 1.02, + "learning_rate": 1.899051509027902e-05, + "loss": 0.946, + "step": 13798 + }, + { + "epoch": 1.02, + "learning_rate": 1.899034044002989e-05, + "loss": 1.025, + "step": 13799 + }, + { + "epoch": 1.02, + "learning_rate": 1.8990165775477255e-05, + "loss": 1.082, + "step": 13800 + }, + { + "epoch": 1.02, + "learning_rate": 1.8989991096621395e-05, + "loss": 1.0679, + "step": 13801 + }, + { + "epoch": 1.02, + "learning_rate": 1.898981640346259e-05, + "loss": 0.9456, + "step": 13802 + }, + { + "epoch": 1.02, + "learning_rate": 1.8989641696001112e-05, + "loss": 1.0182, + "step": 13803 + }, + { + "epoch": 1.02, + "learning_rate": 1.8989466974237244e-05, + "loss": 1.0058, + "step": 13804 + }, + { + "epoch": 1.02, + "learning_rate": 1.8989292238171268e-05, + "loss": 0.9951, + "step": 13805 + }, + { + "epoch": 1.02, + "learning_rate": 1.898911748780345e-05, + "loss": 1.0418, + "step": 13806 + }, + { + "epoch": 1.02, + "learning_rate": 1.898894272313408e-05, + "loss": 0.9777, + "step": 13807 + }, + { + "epoch": 1.02, + "learning_rate": 1.8988767944163425e-05, + "loss": 0.9613, + "step": 13808 + }, + { + "epoch": 1.02, + "learning_rate": 1.898859315089177e-05, + "loss": 1.0099, + "step": 13809 + }, + { + "epoch": 1.02, + "learning_rate": 1.8988418343319396e-05, + "loss": 1.0924, + "step": 13810 + }, + { + "epoch": 1.02, + "learning_rate": 1.8988243521446574e-05, + "loss": 1.1365, + "step": 13811 + }, + { + "epoch": 1.02, + "learning_rate": 1.8988068685273586e-05, + "loss": 0.9123, + "step": 13812 + }, + { + "epoch": 1.02, + "learning_rate": 1.8987893834800706e-05, + "loss": 0.9245, + "step": 13813 + }, + { + "epoch": 1.02, + "learning_rate": 1.898771897002822e-05, + "loss": 1.0223, + "step": 13814 + }, + { + "epoch": 1.02, + "learning_rate": 1.89875440909564e-05, + "loss": 0.9479, + "step": 13815 + }, + { + "epoch": 1.02, + "learning_rate": 1.8987369197585528e-05, + "loss": 1.0579, + "step": 13816 + }, + { + "epoch": 1.02, + "learning_rate": 1.898719428991588e-05, + "loss": 1.0062, + "step": 13817 + }, + { + "epoch": 1.02, + "learning_rate": 1.898701936794773e-05, + "loss": 1.0209, + "step": 13818 + }, + { + "epoch": 1.02, + "learning_rate": 1.8986844431681363e-05, + "loss": 1.0467, + "step": 13819 + }, + { + "epoch": 1.02, + "learning_rate": 1.8986669481117056e-05, + "loss": 0.9658, + "step": 13820 + }, + { + "epoch": 1.02, + "learning_rate": 1.898649451625509e-05, + "loss": 0.9212, + "step": 13821 + }, + { + "epoch": 1.02, + "learning_rate": 1.8986319537095735e-05, + "loss": 1.0453, + "step": 13822 + }, + { + "epoch": 1.02, + "learning_rate": 1.8986144543639272e-05, + "loss": 1.0349, + "step": 13823 + }, + { + "epoch": 1.02, + "learning_rate": 1.8985969535885988e-05, + "loss": 0.9572, + "step": 13824 + }, + { + "epoch": 1.02, + "learning_rate": 1.898579451383615e-05, + "loss": 0.9848, + "step": 13825 + }, + { + "epoch": 1.02, + "learning_rate": 1.8985619477490045e-05, + "loss": 1.0443, + "step": 13826 + }, + { + "epoch": 1.02, + "learning_rate": 1.8985444426847946e-05, + "loss": 0.9932, + "step": 13827 + }, + { + "epoch": 1.02, + "learning_rate": 1.8985269361910132e-05, + "loss": 1.0707, + "step": 13828 + }, + { + "epoch": 1.02, + "learning_rate": 1.898509428267689e-05, + "loss": 1.0313, + "step": 13829 + }, + { + "epoch": 1.02, + "learning_rate": 1.898491918914848e-05, + "loss": 1.0342, + "step": 13830 + }, + { + "epoch": 1.02, + "learning_rate": 1.89847440813252e-05, + "loss": 0.9891, + "step": 13831 + }, + { + "epoch": 1.02, + "learning_rate": 1.898456895920732e-05, + "loss": 1.0084, + "step": 13832 + }, + { + "epoch": 1.02, + "learning_rate": 1.8984393822795118e-05, + "loss": 1.0715, + "step": 13833 + }, + { + "epoch": 1.02, + "learning_rate": 1.8984218672088876e-05, + "loss": 1.0254, + "step": 13834 + }, + { + "epoch": 1.02, + "learning_rate": 1.898404350708887e-05, + "loss": 1.0244, + "step": 13835 + }, + { + "epoch": 1.02, + "learning_rate": 1.898386832779538e-05, + "loss": 1.0599, + "step": 13836 + }, + { + "epoch": 1.02, + "learning_rate": 1.8983693134208677e-05, + "loss": 0.9843, + "step": 13837 + }, + { + "epoch": 1.02, + "learning_rate": 1.8983517926329057e-05, + "loss": 1.0032, + "step": 13838 + }, + { + "epoch": 1.02, + "learning_rate": 1.898334270415678e-05, + "loss": 1.0643, + "step": 13839 + }, + { + "epoch": 1.02, + "learning_rate": 1.898316746769214e-05, + "loss": 1.0281, + "step": 13840 + }, + { + "epoch": 1.02, + "learning_rate": 1.8982992216935406e-05, + "loss": 0.9294, + "step": 13841 + }, + { + "epoch": 1.02, + "learning_rate": 1.898281695188686e-05, + "loss": 0.9711, + "step": 13842 + }, + { + "epoch": 1.02, + "learning_rate": 1.898264167254678e-05, + "loss": 0.9773, + "step": 13843 + }, + { + "epoch": 1.02, + "learning_rate": 1.8982466378915446e-05, + "loss": 0.9386, + "step": 13844 + }, + { + "epoch": 1.02, + "learning_rate": 1.8982291070993136e-05, + "loss": 1.1024, + "step": 13845 + }, + { + "epoch": 1.02, + "learning_rate": 1.898211574878013e-05, + "loss": 1.023, + "step": 13846 + }, + { + "epoch": 1.02, + "learning_rate": 1.898194041227671e-05, + "loss": 0.9455, + "step": 13847 + }, + { + "epoch": 1.02, + "learning_rate": 1.8981765061483146e-05, + "loss": 1.002, + "step": 13848 + }, + { + "epoch": 1.02, + "learning_rate": 1.8981589696399726e-05, + "loss": 1.017, + "step": 13849 + }, + { + "epoch": 1.02, + "learning_rate": 1.898141431702672e-05, + "loss": 1.0305, + "step": 13850 + }, + { + "epoch": 1.02, + "learning_rate": 1.8981238923364417e-05, + "loss": 1.045, + "step": 13851 + }, + { + "epoch": 1.02, + "learning_rate": 1.8981063515413093e-05, + "loss": 0.9545, + "step": 13852 + }, + { + "epoch": 1.02, + "learning_rate": 1.898088809317302e-05, + "loss": 1.0562, + "step": 13853 + }, + { + "epoch": 1.02, + "learning_rate": 1.8980712656644487e-05, + "loss": 1.0779, + "step": 13854 + }, + { + "epoch": 1.02, + "learning_rate": 1.8980537205827766e-05, + "loss": 0.9611, + "step": 13855 + }, + { + "epoch": 1.02, + "learning_rate": 1.898036174072314e-05, + "loss": 1.0773, + "step": 13856 + }, + { + "epoch": 1.02, + "learning_rate": 1.8980186261330888e-05, + "loss": 0.9734, + "step": 13857 + }, + { + "epoch": 1.02, + "learning_rate": 1.8980010767651285e-05, + "loss": 0.8951, + "step": 13858 + }, + { + "epoch": 1.02, + "learning_rate": 1.8979835259684614e-05, + "loss": 1.0334, + "step": 13859 + }, + { + "epoch": 1.02, + "learning_rate": 1.8979659737431158e-05, + "loss": 1.0017, + "step": 13860 + }, + { + "epoch": 1.02, + "learning_rate": 1.8979484200891187e-05, + "loss": 0.9913, + "step": 13861 + }, + { + "epoch": 1.02, + "learning_rate": 1.8979308650064988e-05, + "loss": 0.9719, + "step": 13862 + }, + { + "epoch": 1.02, + "learning_rate": 1.8979133084952837e-05, + "loss": 1.0165, + "step": 13863 + }, + { + "epoch": 1.02, + "learning_rate": 1.897895750555501e-05, + "loss": 1.1727, + "step": 13864 + }, + { + "epoch": 1.02, + "learning_rate": 1.8978781911871795e-05, + "loss": 1.0072, + "step": 13865 + }, + { + "epoch": 1.02, + "learning_rate": 1.8978606303903463e-05, + "loss": 1.0541, + "step": 13866 + }, + { + "epoch": 1.02, + "learning_rate": 1.89784306816503e-05, + "loss": 1.1547, + "step": 13867 + }, + { + "epoch": 1.02, + "learning_rate": 1.897825504511258e-05, + "loss": 1.068, + "step": 13868 + }, + { + "epoch": 1.02, + "learning_rate": 1.8978079394290584e-05, + "loss": 1.0378, + "step": 13869 + }, + { + "epoch": 1.02, + "learning_rate": 1.8977903729184594e-05, + "loss": 1.0182, + "step": 13870 + }, + { + "epoch": 1.02, + "learning_rate": 1.8977728049794888e-05, + "loss": 0.9598, + "step": 13871 + }, + { + "epoch": 1.02, + "learning_rate": 1.8977552356121743e-05, + "loss": 0.9333, + "step": 13872 + }, + { + "epoch": 1.03, + "learning_rate": 1.897737664816544e-05, + "loss": 1.1285, + "step": 13873 + }, + { + "epoch": 1.03, + "learning_rate": 1.8977200925926262e-05, + "loss": 1.1103, + "step": 13874 + }, + { + "epoch": 1.03, + "learning_rate": 1.8977025189404485e-05, + "loss": 1.0553, + "step": 13875 + }, + { + "epoch": 1.03, + "learning_rate": 1.8976849438600387e-05, + "loss": 1.002, + "step": 13876 + }, + { + "epoch": 1.03, + "learning_rate": 1.8976673673514254e-05, + "loss": 0.9971, + "step": 13877 + }, + { + "epoch": 1.03, + "learning_rate": 1.897649789414636e-05, + "loss": 1.0689, + "step": 13878 + }, + { + "epoch": 1.03, + "learning_rate": 1.8976322100496983e-05, + "loss": 0.9771, + "step": 13879 + }, + { + "epoch": 1.03, + "learning_rate": 1.8976146292566412e-05, + "loss": 0.9548, + "step": 13880 + }, + { + "epoch": 1.03, + "learning_rate": 1.897597047035492e-05, + "loss": 0.9566, + "step": 13881 + }, + { + "epoch": 1.03, + "learning_rate": 1.8975794633862784e-05, + "loss": 0.9628, + "step": 13882 + }, + { + "epoch": 1.03, + "learning_rate": 1.8975618783090285e-05, + "loss": 1.031, + "step": 13883 + }, + { + "epoch": 1.03, + "learning_rate": 1.897544291803771e-05, + "loss": 1.0536, + "step": 13884 + }, + { + "epoch": 1.03, + "learning_rate": 1.897526703870533e-05, + "loss": 0.9559, + "step": 13885 + }, + { + "epoch": 1.03, + "learning_rate": 1.8975091145093433e-05, + "loss": 0.9574, + "step": 13886 + }, + { + "epoch": 1.03, + "learning_rate": 1.897491523720229e-05, + "loss": 0.9059, + "step": 13887 + }, + { + "epoch": 1.03, + "learning_rate": 1.8974739315032187e-05, + "loss": 1.0206, + "step": 13888 + }, + { + "epoch": 1.03, + "learning_rate": 1.89745633785834e-05, + "loss": 1.0238, + "step": 13889 + }, + { + "epoch": 1.03, + "learning_rate": 1.8974387427856217e-05, + "loss": 1.0444, + "step": 13890 + }, + { + "epoch": 1.03, + "learning_rate": 1.8974211462850907e-05, + "loss": 1.0217, + "step": 13891 + }, + { + "epoch": 1.03, + "learning_rate": 1.8974035483567756e-05, + "loss": 1.0058, + "step": 13892 + }, + { + "epoch": 1.03, + "learning_rate": 1.897385949000704e-05, + "loss": 0.9112, + "step": 13893 + }, + { + "epoch": 1.03, + "learning_rate": 1.897368348216905e-05, + "loss": 1.0264, + "step": 13894 + }, + { + "epoch": 1.03, + "learning_rate": 1.897350746005405e-05, + "loss": 1.0511, + "step": 13895 + }, + { + "epoch": 1.03, + "learning_rate": 1.8973331423662334e-05, + "loss": 1.0383, + "step": 13896 + }, + { + "epoch": 1.03, + "learning_rate": 1.897315537299417e-05, + "loss": 1.0289, + "step": 13897 + }, + { + "epoch": 1.03, + "learning_rate": 1.897297930804985e-05, + "loss": 0.973, + "step": 13898 + }, + { + "epoch": 1.03, + "learning_rate": 1.8972803228829644e-05, + "loss": 1.065, + "step": 13899 + }, + { + "epoch": 1.03, + "learning_rate": 1.8972627135333838e-05, + "loss": 1.0198, + "step": 13900 + }, + { + "epoch": 1.03, + "learning_rate": 1.897245102756271e-05, + "loss": 1.0739, + "step": 13901 + }, + { + "epoch": 1.03, + "learning_rate": 1.8972274905516544e-05, + "loss": 1.1061, + "step": 13902 + }, + { + "epoch": 1.03, + "learning_rate": 1.8972098769195612e-05, + "loss": 1.1071, + "step": 13903 + }, + { + "epoch": 1.03, + "learning_rate": 1.8971922618600203e-05, + "loss": 1.0811, + "step": 13904 + }, + { + "epoch": 1.03, + "learning_rate": 1.8971746453730592e-05, + "loss": 1.0235, + "step": 13905 + }, + { + "epoch": 1.03, + "learning_rate": 1.8971570274587063e-05, + "loss": 0.9934, + "step": 13906 + }, + { + "epoch": 1.03, + "learning_rate": 1.897139408116989e-05, + "loss": 1.0314, + "step": 13907 + }, + { + "epoch": 1.03, + "learning_rate": 1.897121787347936e-05, + "loss": 1.0207, + "step": 13908 + }, + { + "epoch": 1.03, + "learning_rate": 1.8971041651515748e-05, + "loss": 0.9206, + "step": 13909 + }, + { + "epoch": 1.03, + "learning_rate": 1.897086541527934e-05, + "loss": 0.9756, + "step": 13910 + }, + { + "epoch": 1.03, + "learning_rate": 1.8970689164770417e-05, + "loss": 1.1297, + "step": 13911 + }, + { + "epoch": 1.03, + "learning_rate": 1.8970512899989247e-05, + "loss": 1.0334, + "step": 13912 + }, + { + "epoch": 1.03, + "learning_rate": 1.8970336620936125e-05, + "loss": 1.0451, + "step": 13913 + }, + { + "epoch": 1.03, + "learning_rate": 1.8970160327611325e-05, + "loss": 1.0434, + "step": 13914 + }, + { + "epoch": 1.03, + "learning_rate": 1.8969984020015127e-05, + "loss": 1.0176, + "step": 13915 + }, + { + "epoch": 1.03, + "learning_rate": 1.8969807698147816e-05, + "loss": 0.8941, + "step": 13916 + }, + { + "epoch": 1.03, + "learning_rate": 1.8969631362009668e-05, + "loss": 1.0734, + "step": 13917 + }, + { + "epoch": 1.03, + "learning_rate": 1.8969455011600962e-05, + "loss": 0.8707, + "step": 13918 + }, + { + "epoch": 1.03, + "learning_rate": 1.8969278646921984e-05, + "loss": 1.0681, + "step": 13919 + }, + { + "epoch": 1.03, + "learning_rate": 1.896910226797301e-05, + "loss": 0.9172, + "step": 13920 + }, + { + "epoch": 1.03, + "learning_rate": 1.8968925874754322e-05, + "loss": 0.9412, + "step": 13921 + }, + { + "epoch": 1.03, + "learning_rate": 1.8968749467266204e-05, + "loss": 1.0415, + "step": 13922 + }, + { + "epoch": 1.03, + "learning_rate": 1.8968573045508935e-05, + "loss": 0.9875, + "step": 13923 + }, + { + "epoch": 1.03, + "learning_rate": 1.8968396609482793e-05, + "loss": 1.0952, + "step": 13924 + }, + { + "epoch": 1.03, + "learning_rate": 1.896822015918806e-05, + "loss": 1.0041, + "step": 13925 + }, + { + "epoch": 1.03, + "learning_rate": 1.8968043694625014e-05, + "loss": 1.1322, + "step": 13926 + }, + { + "epoch": 1.03, + "learning_rate": 1.8967867215793944e-05, + "loss": 0.9749, + "step": 13927 + }, + { + "epoch": 1.03, + "learning_rate": 1.896769072269512e-05, + "loss": 0.9727, + "step": 13928 + }, + { + "epoch": 1.03, + "learning_rate": 1.8967514215328836e-05, + "loss": 1.0349, + "step": 13929 + }, + { + "epoch": 1.03, + "learning_rate": 1.896733769369536e-05, + "loss": 0.9298, + "step": 13930 + }, + { + "epoch": 1.03, + "learning_rate": 1.8967161157794978e-05, + "loss": 1.0186, + "step": 13931 + }, + { + "epoch": 1.03, + "learning_rate": 1.896698460762797e-05, + "loss": 0.9169, + "step": 13932 + }, + { + "epoch": 1.03, + "learning_rate": 1.8966808043194623e-05, + "loss": 0.9505, + "step": 13933 + }, + { + "epoch": 1.03, + "learning_rate": 1.896663146449521e-05, + "loss": 0.8813, + "step": 13934 + }, + { + "epoch": 1.03, + "learning_rate": 1.896645487153001e-05, + "loss": 0.9989, + "step": 13935 + }, + { + "epoch": 1.03, + "learning_rate": 1.8966278264299313e-05, + "loss": 1.0226, + "step": 13936 + }, + { + "epoch": 1.03, + "learning_rate": 1.8966101642803397e-05, + "loss": 1.056, + "step": 13937 + }, + { + "epoch": 1.03, + "learning_rate": 1.896592500704254e-05, + "loss": 0.9346, + "step": 13938 + }, + { + "epoch": 1.03, + "learning_rate": 1.8965748357017023e-05, + "loss": 1.0055, + "step": 13939 + }, + { + "epoch": 1.03, + "learning_rate": 1.896557169272713e-05, + "loss": 1.0724, + "step": 13940 + }, + { + "epoch": 1.03, + "learning_rate": 1.896539501417314e-05, + "loss": 1.0429, + "step": 13941 + }, + { + "epoch": 1.03, + "learning_rate": 1.8965218321355337e-05, + "loss": 1.0999, + "step": 13942 + }, + { + "epoch": 1.03, + "learning_rate": 1.8965041614273997e-05, + "loss": 0.9972, + "step": 13943 + }, + { + "epoch": 1.03, + "learning_rate": 1.8964864892929407e-05, + "loss": 1.0125, + "step": 13944 + }, + { + "epoch": 1.03, + "learning_rate": 1.8964688157321844e-05, + "loss": 1.0929, + "step": 13945 + }, + { + "epoch": 1.03, + "learning_rate": 1.896451140745159e-05, + "loss": 1.0159, + "step": 13946 + }, + { + "epoch": 1.03, + "learning_rate": 1.896433464331893e-05, + "loss": 1.038, + "step": 13947 + }, + { + "epoch": 1.03, + "learning_rate": 1.8964157864924133e-05, + "loss": 0.9635, + "step": 13948 + }, + { + "epoch": 1.03, + "learning_rate": 1.8963981072267494e-05, + "loss": 0.9672, + "step": 13949 + }, + { + "epoch": 1.03, + "learning_rate": 1.896380426534929e-05, + "loss": 1.0261, + "step": 13950 + }, + { + "epoch": 1.03, + "learning_rate": 1.8963627444169803e-05, + "loss": 1.0399, + "step": 13951 + }, + { + "epoch": 1.03, + "learning_rate": 1.8963450608729313e-05, + "loss": 0.9573, + "step": 13952 + }, + { + "epoch": 1.03, + "learning_rate": 1.89632737590281e-05, + "loss": 1.0422, + "step": 13953 + }, + { + "epoch": 1.03, + "learning_rate": 1.8963096895066445e-05, + "loss": 0.857, + "step": 13954 + }, + { + "epoch": 1.03, + "learning_rate": 1.896292001684463e-05, + "loss": 0.9945, + "step": 13955 + }, + { + "epoch": 1.03, + "learning_rate": 1.8962743124362944e-05, + "loss": 1.021, + "step": 13956 + }, + { + "epoch": 1.03, + "learning_rate": 1.8962566217621657e-05, + "loss": 0.9951, + "step": 13957 + }, + { + "epoch": 1.03, + "learning_rate": 1.8962389296621055e-05, + "loss": 0.9275, + "step": 13958 + }, + { + "epoch": 1.03, + "learning_rate": 1.8962212361361422e-05, + "loss": 1.0985, + "step": 13959 + }, + { + "epoch": 1.03, + "learning_rate": 1.896203541184304e-05, + "loss": 1.1011, + "step": 13960 + }, + { + "epoch": 1.03, + "learning_rate": 1.8961858448066183e-05, + "loss": 0.9128, + "step": 13961 + }, + { + "epoch": 1.03, + "learning_rate": 1.896168147003114e-05, + "loss": 1.0341, + "step": 13962 + }, + { + "epoch": 1.03, + "learning_rate": 1.8961504477738186e-05, + "loss": 1.0903, + "step": 13963 + }, + { + "epoch": 1.03, + "learning_rate": 1.896132747118761e-05, + "loss": 1.0398, + "step": 13964 + }, + { + "epoch": 1.03, + "learning_rate": 1.896115045037969e-05, + "loss": 0.9603, + "step": 13965 + }, + { + "epoch": 1.03, + "learning_rate": 1.896097341531471e-05, + "loss": 1.011, + "step": 13966 + }, + { + "epoch": 1.03, + "learning_rate": 1.8960796365992947e-05, + "loss": 0.9801, + "step": 13967 + }, + { + "epoch": 1.03, + "learning_rate": 1.8960619302414684e-05, + "loss": 1.0915, + "step": 13968 + }, + { + "epoch": 1.03, + "learning_rate": 1.8960442224580205e-05, + "loss": 0.9877, + "step": 13969 + }, + { + "epoch": 1.03, + "learning_rate": 1.8960265132489793e-05, + "loss": 1.0391, + "step": 13970 + }, + { + "epoch": 1.03, + "learning_rate": 1.8960088026143725e-05, + "loss": 1.0405, + "step": 13971 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959910905542285e-05, + "loss": 0.9615, + "step": 13972 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959733770685756e-05, + "loss": 1.0733, + "step": 13973 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959556621574417e-05, + "loss": 0.9113, + "step": 13974 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959379458208554e-05, + "loss": 0.963, + "step": 13975 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959202280588444e-05, + "loss": 0.934, + "step": 13976 + }, + { + "epoch": 1.03, + "learning_rate": 1.8959025088714373e-05, + "loss": 0.8936, + "step": 13977 + }, + { + "epoch": 1.03, + "learning_rate": 1.895884788258662e-05, + "loss": 0.8738, + "step": 13978 + }, + { + "epoch": 1.03, + "learning_rate": 1.895867066220547e-05, + "loss": 1.0522, + "step": 13979 + }, + { + "epoch": 1.03, + "learning_rate": 1.8958493427571202e-05, + "loss": 1.0189, + "step": 13980 + }, + { + "epoch": 1.03, + "learning_rate": 1.89583161786841e-05, + "loss": 1.0383, + "step": 13981 + }, + { + "epoch": 1.03, + "learning_rate": 1.8958138915544443e-05, + "loss": 1.0381, + "step": 13982 + }, + { + "epoch": 1.03, + "learning_rate": 1.895796163815252e-05, + "loss": 1.0349, + "step": 13983 + }, + { + "epoch": 1.03, + "learning_rate": 1.89577843465086e-05, + "loss": 1.0669, + "step": 13984 + }, + { + "epoch": 1.03, + "learning_rate": 1.895760704061298e-05, + "loss": 0.9597, + "step": 13985 + }, + { + "epoch": 1.03, + "learning_rate": 1.8957429720465935e-05, + "loss": 0.899, + "step": 13986 + }, + { + "epoch": 1.03, + "learning_rate": 1.8957252386067745e-05, + "loss": 0.9682, + "step": 13987 + }, + { + "epoch": 1.03, + "learning_rate": 1.8957075037418694e-05, + "loss": 0.9699, + "step": 13988 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956897674519065e-05, + "loss": 0.8984, + "step": 13989 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956720297369143e-05, + "loss": 1.0037, + "step": 13990 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956542905969207e-05, + "loss": 1.0342, + "step": 13991 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956365500319537e-05, + "loss": 1.028, + "step": 13992 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956188080420415e-05, + "loss": 0.9947, + "step": 13993 + }, + { + "epoch": 1.03, + "learning_rate": 1.8956010646272132e-05, + "loss": 1.0924, + "step": 13994 + }, + { + "epoch": 1.03, + "learning_rate": 1.8955833197874963e-05, + "loss": 0.8938, + "step": 13995 + }, + { + "epoch": 1.03, + "learning_rate": 1.895565573522919e-05, + "loss": 0.8801, + "step": 13996 + }, + { + "epoch": 1.03, + "learning_rate": 1.8955478258335096e-05, + "loss": 0.988, + "step": 13997 + }, + { + "epoch": 1.03, + "learning_rate": 1.8955300767192965e-05, + "loss": 1.0536, + "step": 13998 + }, + { + "epoch": 1.03, + "learning_rate": 1.8955123261803078e-05, + "loss": 1.0767, + "step": 13999 + }, + { + "epoch": 1.03, + "learning_rate": 1.895494574216572e-05, + "loss": 1.0816, + "step": 14000 + }, + { + "epoch": 1.03, + "learning_rate": 1.8954768208281173e-05, + "loss": 1.0703, + "step": 14001 + }, + { + "epoch": 1.03, + "learning_rate": 1.8954590660149712e-05, + "loss": 1.0467, + "step": 14002 + }, + { + "epoch": 1.03, + "learning_rate": 1.8954413097771626e-05, + "loss": 1.0752, + "step": 14003 + }, + { + "epoch": 1.03, + "learning_rate": 1.8954235521147203e-05, + "loss": 1.0285, + "step": 14004 + }, + { + "epoch": 1.03, + "learning_rate": 1.8954057930276714e-05, + "loss": 0.9986, + "step": 14005 + }, + { + "epoch": 1.03, + "learning_rate": 1.895388032516045e-05, + "loss": 0.9906, + "step": 14006 + }, + { + "epoch": 1.03, + "learning_rate": 1.895370270579869e-05, + "loss": 1.0364, + "step": 14007 + }, + { + "epoch": 1.04, + "learning_rate": 1.8953525072191716e-05, + "loss": 0.9614, + "step": 14008 + }, + { + "epoch": 1.04, + "learning_rate": 1.8953347424339812e-05, + "loss": 1.0278, + "step": 14009 + }, + { + "epoch": 1.04, + "learning_rate": 1.8953169762243257e-05, + "loss": 1.004, + "step": 14010 + }, + { + "epoch": 1.04, + "learning_rate": 1.8952992085902342e-05, + "loss": 0.9109, + "step": 14011 + }, + { + "epoch": 1.04, + "learning_rate": 1.8952814395317345e-05, + "loss": 1.0192, + "step": 14012 + }, + { + "epoch": 1.04, + "learning_rate": 1.8952636690488547e-05, + "loss": 1.0154, + "step": 14013 + }, + { + "epoch": 1.04, + "learning_rate": 1.895245897141623e-05, + "loss": 1.0677, + "step": 14014 + }, + { + "epoch": 1.04, + "learning_rate": 1.895228123810068e-05, + "loss": 1.0562, + "step": 14015 + }, + { + "epoch": 1.04, + "learning_rate": 1.895210349054218e-05, + "loss": 1.0818, + "step": 14016 + }, + { + "epoch": 1.04, + "learning_rate": 1.895192572874101e-05, + "loss": 1.0371, + "step": 14017 + }, + { + "epoch": 1.04, + "learning_rate": 1.8951747952697453e-05, + "loss": 1.0776, + "step": 14018 + }, + { + "epoch": 1.04, + "learning_rate": 1.8951570162411796e-05, + "loss": 1.0049, + "step": 14019 + }, + { + "epoch": 1.04, + "learning_rate": 1.895139235788432e-05, + "loss": 1.1039, + "step": 14020 + }, + { + "epoch": 1.04, + "learning_rate": 1.8951214539115303e-05, + "loss": 1.0017, + "step": 14021 + }, + { + "epoch": 1.04, + "learning_rate": 1.8951036706105036e-05, + "loss": 1.1504, + "step": 14022 + }, + { + "epoch": 1.04, + "learning_rate": 1.8950858858853794e-05, + "loss": 1.0856, + "step": 14023 + }, + { + "epoch": 1.04, + "learning_rate": 1.8950680997361864e-05, + "loss": 0.9339, + "step": 14024 + }, + { + "epoch": 1.04, + "learning_rate": 1.895050312162953e-05, + "loss": 1.0248, + "step": 14025 + }, + { + "epoch": 1.04, + "learning_rate": 1.8950325231657072e-05, + "loss": 1.1162, + "step": 14026 + }, + { + "epoch": 1.04, + "learning_rate": 1.8950147327444776e-05, + "loss": 1.0028, + "step": 14027 + }, + { + "epoch": 1.04, + "learning_rate": 1.8949969408992923e-05, + "loss": 1.0669, + "step": 14028 + }, + { + "epoch": 1.04, + "learning_rate": 1.8949791476301796e-05, + "loss": 0.9717, + "step": 14029 + }, + { + "epoch": 1.04, + "learning_rate": 1.894961352937168e-05, + "loss": 0.9896, + "step": 14030 + }, + { + "epoch": 1.04, + "learning_rate": 1.8949435568202855e-05, + "loss": 0.9868, + "step": 14031 + }, + { + "epoch": 1.04, + "learning_rate": 1.894925759279561e-05, + "loss": 1.118, + "step": 14032 + }, + { + "epoch": 1.04, + "learning_rate": 1.894907960315022e-05, + "loss": 0.9756, + "step": 14033 + }, + { + "epoch": 1.04, + "learning_rate": 1.8948901599266974e-05, + "loss": 1.0087, + "step": 14034 + }, + { + "epoch": 1.04, + "learning_rate": 1.8948723581146154e-05, + "loss": 1.025, + "step": 14035 + }, + { + "epoch": 1.04, + "learning_rate": 1.8948545548788045e-05, + "loss": 1.1458, + "step": 14036 + }, + { + "epoch": 1.04, + "learning_rate": 1.8948367502192925e-05, + "loss": 1.1017, + "step": 14037 + }, + { + "epoch": 1.04, + "learning_rate": 1.894818944136108e-05, + "loss": 1.0683, + "step": 14038 + }, + { + "epoch": 1.04, + "learning_rate": 1.8948011366292793e-05, + "loss": 1.0242, + "step": 14039 + }, + { + "epoch": 1.04, + "learning_rate": 1.894783327698835e-05, + "loss": 1.0827, + "step": 14040 + }, + { + "epoch": 1.04, + "learning_rate": 1.8947655173448033e-05, + "loss": 1.1364, + "step": 14041 + }, + { + "epoch": 1.04, + "learning_rate": 1.8947477055672123e-05, + "loss": 1.0172, + "step": 14042 + }, + { + "epoch": 1.04, + "learning_rate": 1.8947298923660906e-05, + "loss": 0.9812, + "step": 14043 + }, + { + "epoch": 1.04, + "learning_rate": 1.8947120777414664e-05, + "loss": 1.0847, + "step": 14044 + }, + { + "epoch": 1.04, + "learning_rate": 1.894694261693368e-05, + "loss": 0.9526, + "step": 14045 + }, + { + "epoch": 1.04, + "learning_rate": 1.894676444221824e-05, + "loss": 0.9948, + "step": 14046 + }, + { + "epoch": 1.04, + "learning_rate": 1.8946586253268625e-05, + "loss": 1.0723, + "step": 14047 + }, + { + "epoch": 1.04, + "learning_rate": 1.894640805008512e-05, + "loss": 1.0153, + "step": 14048 + }, + { + "epoch": 1.04, + "learning_rate": 1.8946229832668005e-05, + "loss": 0.9721, + "step": 14049 + }, + { + "epoch": 1.04, + "learning_rate": 1.894605160101757e-05, + "loss": 0.9501, + "step": 14050 + }, + { + "epoch": 1.04, + "learning_rate": 1.894587335513409e-05, + "loss": 1.1431, + "step": 14051 + }, + { + "epoch": 1.04, + "learning_rate": 1.894569509501786e-05, + "loss": 0.8823, + "step": 14052 + }, + { + "epoch": 1.04, + "learning_rate": 1.8945516820669156e-05, + "loss": 1.0139, + "step": 14053 + }, + { + "epoch": 1.04, + "learning_rate": 1.894533853208826e-05, + "loss": 0.8725, + "step": 14054 + }, + { + "epoch": 1.04, + "learning_rate": 1.8945160229275458e-05, + "loss": 0.9216, + "step": 14055 + }, + { + "epoch": 1.04, + "learning_rate": 1.8944981912231032e-05, + "loss": 1.0156, + "step": 14056 + }, + { + "epoch": 1.04, + "learning_rate": 1.8944803580955274e-05, + "loss": 1.0934, + "step": 14057 + }, + { + "epoch": 1.04, + "learning_rate": 1.8944625235448457e-05, + "loss": 0.9993, + "step": 14058 + }, + { + "epoch": 1.04, + "learning_rate": 1.8944446875710867e-05, + "loss": 1.1399, + "step": 14059 + }, + { + "epoch": 1.04, + "learning_rate": 1.89442685017428e-05, + "loss": 1.0609, + "step": 14060 + }, + { + "epoch": 1.04, + "learning_rate": 1.894409011354452e-05, + "loss": 1.1157, + "step": 14061 + }, + { + "epoch": 1.04, + "learning_rate": 1.8943911711116323e-05, + "loss": 1.0283, + "step": 14062 + }, + { + "epoch": 1.04, + "learning_rate": 1.894373329445849e-05, + "loss": 0.9874, + "step": 14063 + }, + { + "epoch": 1.04, + "learning_rate": 1.8943554863571306e-05, + "loss": 1.0153, + "step": 14064 + }, + { + "epoch": 1.04, + "learning_rate": 1.894337641845506e-05, + "loss": 1.0703, + "step": 14065 + }, + { + "epoch": 1.04, + "learning_rate": 1.8943197959110023e-05, + "loss": 0.9919, + "step": 14066 + }, + { + "epoch": 1.04, + "learning_rate": 1.8943019485536487e-05, + "loss": 1.0944, + "step": 14067 + }, + { + "epoch": 1.04, + "learning_rate": 1.8942840997734736e-05, + "loss": 1.0993, + "step": 14068 + }, + { + "epoch": 1.04, + "learning_rate": 1.8942662495705053e-05, + "loss": 0.9803, + "step": 14069 + }, + { + "epoch": 1.04, + "learning_rate": 1.894248397944772e-05, + "loss": 0.9225, + "step": 14070 + }, + { + "epoch": 1.04, + "learning_rate": 1.8942305448963024e-05, + "loss": 0.9947, + "step": 14071 + }, + { + "epoch": 1.04, + "learning_rate": 1.894212690425125e-05, + "loss": 0.9103, + "step": 14072 + }, + { + "epoch": 1.04, + "learning_rate": 1.894194834531268e-05, + "loss": 1.0096, + "step": 14073 + }, + { + "epoch": 1.04, + "learning_rate": 1.8941769772147595e-05, + "loss": 1.0976, + "step": 14074 + }, + { + "epoch": 1.04, + "learning_rate": 1.8941591184756284e-05, + "loss": 0.9593, + "step": 14075 + }, + { + "epoch": 1.04, + "learning_rate": 1.894141258313903e-05, + "loss": 0.9811, + "step": 14076 + }, + { + "epoch": 1.04, + "learning_rate": 1.8941233967296116e-05, + "loss": 0.9818, + "step": 14077 + }, + { + "epoch": 1.04, + "learning_rate": 1.8941055337227825e-05, + "loss": 1.096, + "step": 14078 + }, + { + "epoch": 1.04, + "learning_rate": 1.8940876692934443e-05, + "loss": 0.9881, + "step": 14079 + }, + { + "epoch": 1.04, + "learning_rate": 1.8940698034416257e-05, + "loss": 1.0434, + "step": 14080 + }, + { + "epoch": 1.04, + "learning_rate": 1.8940519361673546e-05, + "loss": 1.0153, + "step": 14081 + }, + { + "epoch": 1.04, + "learning_rate": 1.89403406747066e-05, + "loss": 0.979, + "step": 14082 + }, + { + "epoch": 1.04, + "learning_rate": 1.8940161973515697e-05, + "loss": 0.9874, + "step": 14083 + }, + { + "epoch": 1.04, + "learning_rate": 1.8939983258101122e-05, + "loss": 1.1241, + "step": 14084 + }, + { + "epoch": 1.04, + "learning_rate": 1.8939804528463167e-05, + "loss": 0.9439, + "step": 14085 + }, + { + "epoch": 1.04, + "learning_rate": 1.893962578460211e-05, + "loss": 0.9335, + "step": 14086 + }, + { + "epoch": 1.04, + "learning_rate": 1.893944702651823e-05, + "loss": 1.0093, + "step": 14087 + }, + { + "epoch": 1.04, + "learning_rate": 1.8939268254211822e-05, + "loss": 0.925, + "step": 14088 + }, + { + "epoch": 1.04, + "learning_rate": 1.8939089467683166e-05, + "loss": 1.036, + "step": 14089 + }, + { + "epoch": 1.04, + "learning_rate": 1.8938910666932546e-05, + "loss": 0.9794, + "step": 14090 + }, + { + "epoch": 1.04, + "learning_rate": 1.893873185196025e-05, + "loss": 1.0551, + "step": 14091 + }, + { + "epoch": 1.04, + "learning_rate": 1.893855302276656e-05, + "loss": 0.9894, + "step": 14092 + }, + { + "epoch": 1.04, + "learning_rate": 1.8938374179351753e-05, + "loss": 1.134, + "step": 14093 + }, + { + "epoch": 1.04, + "learning_rate": 1.8938195321716125e-05, + "loss": 0.9377, + "step": 14094 + }, + { + "epoch": 1.04, + "learning_rate": 1.8938016449859958e-05, + "loss": 1.0135, + "step": 14095 + }, + { + "epoch": 1.04, + "learning_rate": 1.893783756378353e-05, + "loss": 0.9732, + "step": 14096 + }, + { + "epoch": 1.04, + "learning_rate": 1.8937658663487133e-05, + "loss": 0.9299, + "step": 14097 + }, + { + "epoch": 1.04, + "learning_rate": 1.8937479748971047e-05, + "loss": 1.16, + "step": 14098 + }, + { + "epoch": 1.04, + "learning_rate": 1.893730082023556e-05, + "loss": 1.1154, + "step": 14099 + }, + { + "epoch": 1.04, + "learning_rate": 1.8937121877280957e-05, + "loss": 1.1206, + "step": 14100 + }, + { + "epoch": 1.04, + "learning_rate": 1.8936942920107523e-05, + "loss": 0.9416, + "step": 14101 + }, + { + "epoch": 1.04, + "learning_rate": 1.8936763948715536e-05, + "loss": 1.113, + "step": 14102 + }, + { + "epoch": 1.04, + "learning_rate": 1.8936584963105288e-05, + "loss": 1.005, + "step": 14103 + }, + { + "epoch": 1.04, + "learning_rate": 1.893640596327706e-05, + "loss": 1.175, + "step": 14104 + }, + { + "epoch": 1.04, + "learning_rate": 1.8936226949231136e-05, + "loss": 1.0928, + "step": 14105 + }, + { + "epoch": 1.04, + "learning_rate": 1.8936047920967805e-05, + "loss": 0.954, + "step": 14106 + }, + { + "epoch": 1.04, + "learning_rate": 1.8935868878487348e-05, + "loss": 1.0383, + "step": 14107 + }, + { + "epoch": 1.04, + "learning_rate": 1.8935689821790053e-05, + "loss": 1.032, + "step": 14108 + }, + { + "epoch": 1.04, + "learning_rate": 1.89355107508762e-05, + "loss": 1.0548, + "step": 14109 + }, + { + "epoch": 1.04, + "learning_rate": 1.8935331665746083e-05, + "loss": 1.0115, + "step": 14110 + }, + { + "epoch": 1.04, + "learning_rate": 1.8935152566399978e-05, + "loss": 1.027, + "step": 14111 + }, + { + "epoch": 1.04, + "learning_rate": 1.8934973452838176e-05, + "loss": 1.0444, + "step": 14112 + }, + { + "epoch": 1.04, + "learning_rate": 1.8934794325060956e-05, + "loss": 1.0531, + "step": 14113 + }, + { + "epoch": 1.04, + "learning_rate": 1.8934615183068605e-05, + "loss": 0.9304, + "step": 14114 + }, + { + "epoch": 1.04, + "learning_rate": 1.8934436026861412e-05, + "loss": 1.0667, + "step": 14115 + }, + { + "epoch": 1.04, + "learning_rate": 1.893425685643966e-05, + "loss": 1.0206, + "step": 14116 + }, + { + "epoch": 1.04, + "learning_rate": 1.893407767180363e-05, + "loss": 0.9867, + "step": 14117 + }, + { + "epoch": 1.04, + "learning_rate": 1.8933898472953612e-05, + "loss": 1.0599, + "step": 14118 + }, + { + "epoch": 1.04, + "learning_rate": 1.8933719259889888e-05, + "loss": 0.9664, + "step": 14119 + }, + { + "epoch": 1.04, + "learning_rate": 1.8933540032612744e-05, + "loss": 1.1116, + "step": 14120 + }, + { + "epoch": 1.04, + "learning_rate": 1.893336079112247e-05, + "loss": 1.0333, + "step": 14121 + }, + { + "epoch": 1.04, + "learning_rate": 1.8933181535419343e-05, + "loss": 1.088, + "step": 14122 + }, + { + "epoch": 1.04, + "learning_rate": 1.893300226550365e-05, + "loss": 0.9611, + "step": 14123 + }, + { + "epoch": 1.04, + "learning_rate": 1.8932822981375682e-05, + "loss": 1.0283, + "step": 14124 + }, + { + "epoch": 1.04, + "learning_rate": 1.893264368303572e-05, + "loss": 0.9867, + "step": 14125 + }, + { + "epoch": 1.04, + "learning_rate": 1.8932464370484046e-05, + "loss": 0.9796, + "step": 14126 + }, + { + "epoch": 1.04, + "learning_rate": 1.8932285043720955e-05, + "loss": 1.0863, + "step": 14127 + }, + { + "epoch": 1.04, + "learning_rate": 1.8932105702746724e-05, + "loss": 0.9954, + "step": 14128 + }, + { + "epoch": 1.04, + "learning_rate": 1.893192634756164e-05, + "loss": 1.0209, + "step": 14129 + }, + { + "epoch": 1.04, + "learning_rate": 1.8931746978165987e-05, + "loss": 0.9843, + "step": 14130 + }, + { + "epoch": 1.04, + "learning_rate": 1.8931567594560056e-05, + "loss": 0.8968, + "step": 14131 + }, + { + "epoch": 1.04, + "learning_rate": 1.8931388196744124e-05, + "loss": 1.0402, + "step": 14132 + }, + { + "epoch": 1.04, + "learning_rate": 1.8931208784718483e-05, + "loss": 1.045, + "step": 14133 + }, + { + "epoch": 1.04, + "learning_rate": 1.893102935848342e-05, + "loss": 1.0078, + "step": 14134 + }, + { + "epoch": 1.04, + "learning_rate": 1.8930849918039213e-05, + "loss": 1.0303, + "step": 14135 + }, + { + "epoch": 1.04, + "learning_rate": 1.8930670463386155e-05, + "loss": 0.9887, + "step": 14136 + }, + { + "epoch": 1.04, + "learning_rate": 1.8930490994524525e-05, + "loss": 0.8861, + "step": 14137 + }, + { + "epoch": 1.04, + "learning_rate": 1.893031151145461e-05, + "loss": 0.9689, + "step": 14138 + }, + { + "epoch": 1.04, + "learning_rate": 1.8930132014176707e-05, + "loss": 1.0095, + "step": 14139 + }, + { + "epoch": 1.04, + "learning_rate": 1.892995250269108e-05, + "loss": 1.0219, + "step": 14140 + }, + { + "epoch": 1.04, + "learning_rate": 1.8929772976998033e-05, + "loss": 1.0623, + "step": 14141 + }, + { + "epoch": 1.04, + "learning_rate": 1.8929593437097844e-05, + "loss": 0.974, + "step": 14142 + }, + { + "epoch": 1.05, + "learning_rate": 1.8929413882990798e-05, + "loss": 1.0762, + "step": 14143 + }, + { + "epoch": 1.05, + "learning_rate": 1.8929234314677183e-05, + "loss": 0.9467, + "step": 14144 + }, + { + "epoch": 1.05, + "learning_rate": 1.8929054732157284e-05, + "loss": 0.9279, + "step": 14145 + }, + { + "epoch": 1.05, + "learning_rate": 1.892887513543139e-05, + "loss": 1.095, + "step": 14146 + }, + { + "epoch": 1.05, + "learning_rate": 1.892869552449978e-05, + "loss": 0.9765, + "step": 14147 + }, + { + "epoch": 1.05, + "learning_rate": 1.8928515899362742e-05, + "loss": 0.8983, + "step": 14148 + }, + { + "epoch": 1.05, + "learning_rate": 1.8928336260020563e-05, + "loss": 1.0959, + "step": 14149 + }, + { + "epoch": 1.05, + "learning_rate": 1.8928156606473533e-05, + "loss": 1.0694, + "step": 14150 + }, + { + "epoch": 1.05, + "learning_rate": 1.892797693872193e-05, + "loss": 0.9696, + "step": 14151 + }, + { + "epoch": 1.05, + "learning_rate": 1.8927797256766044e-05, + "loss": 0.995, + "step": 14152 + }, + { + "epoch": 1.05, + "learning_rate": 1.892761756060616e-05, + "loss": 1.003, + "step": 14153 + }, + { + "epoch": 1.05, + "learning_rate": 1.8927437850242564e-05, + "loss": 1.0217, + "step": 14154 + }, + { + "epoch": 1.05, + "learning_rate": 1.892725812567554e-05, + "loss": 0.8812, + "step": 14155 + }, + { + "epoch": 1.05, + "learning_rate": 1.892707838690538e-05, + "loss": 0.9857, + "step": 14156 + }, + { + "epoch": 1.05, + "learning_rate": 1.8926898633932364e-05, + "loss": 1.0444, + "step": 14157 + }, + { + "epoch": 1.05, + "learning_rate": 1.8926718866756782e-05, + "loss": 1.0458, + "step": 14158 + }, + { + "epoch": 1.05, + "learning_rate": 1.8926539085378914e-05, + "loss": 0.9853, + "step": 14159 + }, + { + "epoch": 1.05, + "learning_rate": 1.8926359289799055e-05, + "loss": 1.0488, + "step": 14160 + }, + { + "epoch": 1.05, + "learning_rate": 1.8926179480017482e-05, + "loss": 0.9522, + "step": 14161 + }, + { + "epoch": 1.05, + "learning_rate": 1.8925999656034487e-05, + "loss": 1.0299, + "step": 14162 + }, + { + "epoch": 1.05, + "learning_rate": 1.8925819817850354e-05, + "loss": 1.0042, + "step": 14163 + }, + { + "epoch": 1.05, + "learning_rate": 1.892563996546537e-05, + "loss": 1.1441, + "step": 14164 + }, + { + "epoch": 1.05, + "learning_rate": 1.8925460098879822e-05, + "loss": 0.9047, + "step": 14165 + }, + { + "epoch": 1.05, + "learning_rate": 1.8925280218093994e-05, + "loss": 0.9034, + "step": 14166 + }, + { + "epoch": 1.05, + "learning_rate": 1.892510032310817e-05, + "loss": 0.9907, + "step": 14167 + }, + { + "epoch": 1.05, + "learning_rate": 1.892492041392264e-05, + "loss": 0.9974, + "step": 14168 + }, + { + "epoch": 1.05, + "learning_rate": 1.8924740490537693e-05, + "loss": 1.0496, + "step": 14169 + }, + { + "epoch": 1.05, + "learning_rate": 1.8924560552953607e-05, + "loss": 1.1379, + "step": 14170 + }, + { + "epoch": 1.05, + "learning_rate": 1.8924380601170674e-05, + "loss": 0.9633, + "step": 14171 + }, + { + "epoch": 1.05, + "learning_rate": 1.8924200635189178e-05, + "loss": 1.0937, + "step": 14172 + }, + { + "epoch": 1.05, + "learning_rate": 1.892402065500941e-05, + "loss": 1.0237, + "step": 14173 + }, + { + "epoch": 1.05, + "learning_rate": 1.892384066063165e-05, + "loss": 1.0139, + "step": 14174 + }, + { + "epoch": 1.05, + "learning_rate": 1.892366065205619e-05, + "loss": 0.8539, + "step": 14175 + }, + { + "epoch": 1.05, + "learning_rate": 1.8923480629283312e-05, + "loss": 0.9956, + "step": 14176 + }, + { + "epoch": 1.05, + "learning_rate": 1.8923300592313303e-05, + "loss": 0.9877, + "step": 14177 + }, + { + "epoch": 1.05, + "learning_rate": 1.892312054114645e-05, + "loss": 1.0401, + "step": 14178 + }, + { + "epoch": 1.05, + "learning_rate": 1.8922940475783042e-05, + "loss": 1.1099, + "step": 14179 + }, + { + "epoch": 1.05, + "learning_rate": 1.892276039622336e-05, + "loss": 1.1166, + "step": 14180 + }, + { + "epoch": 1.05, + "learning_rate": 1.8922580302467696e-05, + "loss": 0.9147, + "step": 14181 + }, + { + "epoch": 1.05, + "learning_rate": 1.8922400194516336e-05, + "loss": 1.0029, + "step": 14182 + }, + { + "epoch": 1.05, + "learning_rate": 1.8922220072369565e-05, + "loss": 1.1342, + "step": 14183 + }, + { + "epoch": 1.05, + "learning_rate": 1.8922039936027666e-05, + "loss": 1.1105, + "step": 14184 + }, + { + "epoch": 1.05, + "learning_rate": 1.8921859785490932e-05, + "loss": 0.983, + "step": 14185 + }, + { + "epoch": 1.05, + "learning_rate": 1.8921679620759647e-05, + "loss": 1.0755, + "step": 14186 + }, + { + "epoch": 1.05, + "learning_rate": 1.8921499441834095e-05, + "loss": 0.9626, + "step": 14187 + }, + { + "epoch": 1.05, + "learning_rate": 1.8921319248714565e-05, + "loss": 1.0458, + "step": 14188 + }, + { + "epoch": 1.05, + "learning_rate": 1.8921139041401345e-05, + "loss": 1.0718, + "step": 14189 + }, + { + "epoch": 1.05, + "learning_rate": 1.8920958819894722e-05, + "loss": 1.0605, + "step": 14190 + }, + { + "epoch": 1.05, + "learning_rate": 1.892077858419498e-05, + "loss": 1.1717, + "step": 14191 + }, + { + "epoch": 1.05, + "learning_rate": 1.8920598334302408e-05, + "loss": 1.1553, + "step": 14192 + }, + { + "epoch": 1.05, + "learning_rate": 1.8920418070217292e-05, + "loss": 1.0782, + "step": 14193 + }, + { + "epoch": 1.05, + "learning_rate": 1.8920237791939915e-05, + "loss": 0.9334, + "step": 14194 + }, + { + "epoch": 1.05, + "learning_rate": 1.8920057499470566e-05, + "loss": 1.1206, + "step": 14195 + }, + { + "epoch": 1.05, + "learning_rate": 1.891987719280954e-05, + "loss": 1.0178, + "step": 14196 + }, + { + "epoch": 1.05, + "learning_rate": 1.8919696871957112e-05, + "loss": 1.151, + "step": 14197 + }, + { + "epoch": 1.05, + "learning_rate": 1.8919516536913573e-05, + "loss": 1.0883, + "step": 14198 + }, + { + "epoch": 1.05, + "learning_rate": 1.8919336187679213e-05, + "loss": 1.0127, + "step": 14199 + }, + { + "epoch": 1.05, + "learning_rate": 1.8919155824254317e-05, + "loss": 0.9321, + "step": 14200 + }, + { + "epoch": 1.05, + "learning_rate": 1.8918975446639173e-05, + "loss": 1.0407, + "step": 14201 + }, + { + "epoch": 1.05, + "learning_rate": 1.8918795054834066e-05, + "loss": 1.023, + "step": 14202 + }, + { + "epoch": 1.05, + "learning_rate": 1.8918614648839283e-05, + "loss": 1.0636, + "step": 14203 + }, + { + "epoch": 1.05, + "learning_rate": 1.891843422865511e-05, + "loss": 1.0409, + "step": 14204 + }, + { + "epoch": 1.05, + "learning_rate": 1.8918253794281835e-05, + "loss": 1.0013, + "step": 14205 + }, + { + "epoch": 1.05, + "learning_rate": 1.891807334571975e-05, + "loss": 1.1261, + "step": 14206 + }, + { + "epoch": 1.05, + "learning_rate": 1.8917892882969135e-05, + "loss": 1.0427, + "step": 14207 + }, + { + "epoch": 1.05, + "learning_rate": 1.891771240603028e-05, + "loss": 1.0289, + "step": 14208 + }, + { + "epoch": 1.05, + "learning_rate": 1.8917531914903473e-05, + "loss": 0.9869, + "step": 14209 + }, + { + "epoch": 1.05, + "learning_rate": 1.8917351409589e-05, + "loss": 1.0318, + "step": 14210 + }, + { + "epoch": 1.05, + "learning_rate": 1.891717089008715e-05, + "loss": 0.9232, + "step": 14211 + }, + { + "epoch": 1.05, + "learning_rate": 1.8916990356398208e-05, + "loss": 1.0044, + "step": 14212 + }, + { + "epoch": 1.05, + "learning_rate": 1.891680980852246e-05, + "loss": 0.9778, + "step": 14213 + }, + { + "epoch": 1.05, + "learning_rate": 1.8916629246460198e-05, + "loss": 0.9471, + "step": 14214 + }, + { + "epoch": 1.05, + "learning_rate": 1.8916448670211704e-05, + "loss": 0.9946, + "step": 14215 + }, + { + "epoch": 1.05, + "learning_rate": 1.8916268079777273e-05, + "loss": 0.9644, + "step": 14216 + }, + { + "epoch": 1.05, + "learning_rate": 1.891608747515718e-05, + "loss": 0.9323, + "step": 14217 + }, + { + "epoch": 1.05, + "learning_rate": 1.8915906856351722e-05, + "loss": 1.0826, + "step": 14218 + }, + { + "epoch": 1.05, + "learning_rate": 1.8915726223361182e-05, + "loss": 1.0507, + "step": 14219 + }, + { + "epoch": 1.05, + "learning_rate": 1.8915545576185853e-05, + "loss": 1.0087, + "step": 14220 + }, + { + "epoch": 1.05, + "learning_rate": 1.891536491482602e-05, + "loss": 0.996, + "step": 14221 + }, + { + "epoch": 1.05, + "learning_rate": 1.8915184239281964e-05, + "loss": 1.0244, + "step": 14222 + }, + { + "epoch": 1.05, + "learning_rate": 1.891500354955398e-05, + "loss": 0.9502, + "step": 14223 + }, + { + "epoch": 1.05, + "learning_rate": 1.891482284564235e-05, + "loss": 1.0413, + "step": 14224 + }, + { + "epoch": 1.05, + "learning_rate": 1.8914642127547368e-05, + "loss": 1.054, + "step": 14225 + }, + { + "epoch": 1.05, + "learning_rate": 1.8914461395269314e-05, + "loss": 0.985, + "step": 14226 + }, + { + "epoch": 1.05, + "learning_rate": 1.8914280648808482e-05, + "loss": 1.119, + "step": 14227 + }, + { + "epoch": 1.05, + "learning_rate": 1.8914099888165158e-05, + "loss": 0.9701, + "step": 14228 + }, + { + "epoch": 1.05, + "learning_rate": 1.8913919113339626e-05, + "loss": 0.9861, + "step": 14229 + }, + { + "epoch": 1.05, + "learning_rate": 1.8913738324332177e-05, + "loss": 1.0241, + "step": 14230 + }, + { + "epoch": 1.05, + "learning_rate": 1.89135575211431e-05, + "loss": 1.015, + "step": 14231 + }, + { + "epoch": 1.05, + "learning_rate": 1.8913376703772678e-05, + "loss": 0.9707, + "step": 14232 + }, + { + "epoch": 1.05, + "learning_rate": 1.8913195872221203e-05, + "loss": 0.9259, + "step": 14233 + }, + { + "epoch": 1.05, + "learning_rate": 1.8913015026488957e-05, + "loss": 1.0587, + "step": 14234 + }, + { + "epoch": 1.05, + "learning_rate": 1.8912834166576238e-05, + "loss": 1.0132, + "step": 14235 + }, + { + "epoch": 1.05, + "learning_rate": 1.8912653292483324e-05, + "loss": 1.0486, + "step": 14236 + }, + { + "epoch": 1.05, + "learning_rate": 1.8912472404210508e-05, + "loss": 1.0571, + "step": 14237 + }, + { + "epoch": 1.05, + "learning_rate": 1.8912291501758073e-05, + "loss": 0.9865, + "step": 14238 + }, + { + "epoch": 1.05, + "learning_rate": 1.891211058512631e-05, + "loss": 0.891, + "step": 14239 + }, + { + "epoch": 1.05, + "learning_rate": 1.891192965431551e-05, + "loss": 1.0307, + "step": 14240 + }, + { + "epoch": 1.05, + "learning_rate": 1.8911748709325954e-05, + "loss": 1.2067, + "step": 14241 + }, + { + "epoch": 1.05, + "learning_rate": 1.8911567750157935e-05, + "loss": 1.0621, + "step": 14242 + }, + { + "epoch": 1.05, + "learning_rate": 1.891138677681174e-05, + "loss": 0.9686, + "step": 14243 + }, + { + "epoch": 1.05, + "learning_rate": 1.891120578928765e-05, + "loss": 1.051, + "step": 14244 + }, + { + "epoch": 1.05, + "learning_rate": 1.8911024787585964e-05, + "loss": 0.9938, + "step": 14245 + }, + { + "epoch": 1.05, + "learning_rate": 1.8910843771706967e-05, + "loss": 1.0554, + "step": 14246 + }, + { + "epoch": 1.05, + "learning_rate": 1.8910662741650943e-05, + "loss": 1.0307, + "step": 14247 + }, + { + "epoch": 1.05, + "learning_rate": 1.8910481697418183e-05, + "loss": 1.0241, + "step": 14248 + }, + { + "epoch": 1.05, + "learning_rate": 1.8910300639008968e-05, + "loss": 1.0788, + "step": 14249 + }, + { + "epoch": 1.05, + "learning_rate": 1.89101195664236e-05, + "loss": 1.0612, + "step": 14250 + }, + { + "epoch": 1.05, + "learning_rate": 1.8909938479662355e-05, + "loss": 1.0723, + "step": 14251 + }, + { + "epoch": 1.05, + "learning_rate": 1.890975737872553e-05, + "loss": 0.9878, + "step": 14252 + }, + { + "epoch": 1.05, + "learning_rate": 1.8909576263613404e-05, + "loss": 0.9139, + "step": 14253 + }, + { + "epoch": 1.05, + "learning_rate": 1.890939513432627e-05, + "loss": 1.0791, + "step": 14254 + }, + { + "epoch": 1.05, + "learning_rate": 1.8909213990864418e-05, + "loss": 1.0844, + "step": 14255 + }, + { + "epoch": 1.05, + "learning_rate": 1.8909032833228132e-05, + "loss": 1.0892, + "step": 14256 + }, + { + "epoch": 1.05, + "learning_rate": 1.8908851661417702e-05, + "loss": 1.1068, + "step": 14257 + }, + { + "epoch": 1.05, + "learning_rate": 1.890867047543342e-05, + "loss": 1.0644, + "step": 14258 + }, + { + "epoch": 1.05, + "learning_rate": 1.8908489275275564e-05, + "loss": 1.0028, + "step": 14259 + }, + { + "epoch": 1.05, + "learning_rate": 1.8908308060944436e-05, + "loss": 1.0153, + "step": 14260 + }, + { + "epoch": 1.05, + "learning_rate": 1.890812683244031e-05, + "loss": 0.9694, + "step": 14261 + }, + { + "epoch": 1.05, + "learning_rate": 1.890794558976349e-05, + "loss": 0.9755, + "step": 14262 + }, + { + "epoch": 1.05, + "learning_rate": 1.8907764332914252e-05, + "loss": 1.0073, + "step": 14263 + }, + { + "epoch": 1.05, + "learning_rate": 1.8907583061892886e-05, + "loss": 1.1339, + "step": 14264 + }, + { + "epoch": 1.05, + "learning_rate": 1.890740177669969e-05, + "loss": 1.047, + "step": 14265 + }, + { + "epoch": 1.05, + "learning_rate": 1.8907220477334937e-05, + "loss": 1.0697, + "step": 14266 + }, + { + "epoch": 1.05, + "learning_rate": 1.8907039163798927e-05, + "loss": 1.0067, + "step": 14267 + }, + { + "epoch": 1.05, + "learning_rate": 1.8906857836091947e-05, + "loss": 0.9873, + "step": 14268 + }, + { + "epoch": 1.05, + "learning_rate": 1.8906676494214283e-05, + "loss": 1.0064, + "step": 14269 + }, + { + "epoch": 1.05, + "learning_rate": 1.890649513816622e-05, + "loss": 1.0096, + "step": 14270 + }, + { + "epoch": 1.05, + "learning_rate": 1.8906313767948057e-05, + "loss": 1.0678, + "step": 14271 + }, + { + "epoch": 1.05, + "learning_rate": 1.8906132383560072e-05, + "loss": 1.1923, + "step": 14272 + }, + { + "epoch": 1.05, + "learning_rate": 1.8905950985002557e-05, + "loss": 1.0194, + "step": 14273 + }, + { + "epoch": 1.05, + "learning_rate": 1.8905769572275805e-05, + "loss": 1.0694, + "step": 14274 + }, + { + "epoch": 1.05, + "learning_rate": 1.89055881453801e-05, + "loss": 1.0982, + "step": 14275 + }, + { + "epoch": 1.05, + "learning_rate": 1.890540670431573e-05, + "loss": 0.9728, + "step": 14276 + }, + { + "epoch": 1.05, + "learning_rate": 1.8905225249082986e-05, + "loss": 1.0561, + "step": 14277 + }, + { + "epoch": 1.05, + "learning_rate": 1.8905043779682156e-05, + "loss": 1.1394, + "step": 14278 + }, + { + "epoch": 1.06, + "learning_rate": 1.890486229611353e-05, + "loss": 1.0393, + "step": 14279 + }, + { + "epoch": 1.06, + "learning_rate": 1.8904680798377393e-05, + "loss": 1.1592, + "step": 14280 + }, + { + "epoch": 1.06, + "learning_rate": 1.890449928647404e-05, + "loss": 1.0434, + "step": 14281 + }, + { + "epoch": 1.06, + "learning_rate": 1.890431776040375e-05, + "loss": 1.1613, + "step": 14282 + }, + { + "epoch": 1.06, + "learning_rate": 1.8904136220166824e-05, + "loss": 1.0348, + "step": 14283 + }, + { + "epoch": 1.06, + "learning_rate": 1.8903954665763544e-05, + "loss": 1.0507, + "step": 14284 + }, + { + "epoch": 1.06, + "learning_rate": 1.8903773097194194e-05, + "loss": 0.9823, + "step": 14285 + }, + { + "epoch": 1.06, + "learning_rate": 1.8903591514459076e-05, + "loss": 0.9535, + "step": 14286 + }, + { + "epoch": 1.06, + "learning_rate": 1.8903409917558466e-05, + "loss": 1.0249, + "step": 14287 + }, + { + "epoch": 1.06, + "learning_rate": 1.890322830649266e-05, + "loss": 1.0518, + "step": 14288 + }, + { + "epoch": 1.06, + "learning_rate": 1.8903046681261946e-05, + "loss": 1.0935, + "step": 14289 + }, + { + "epoch": 1.06, + "learning_rate": 1.8902865041866612e-05, + "loss": 0.9628, + "step": 14290 + }, + { + "epoch": 1.06, + "learning_rate": 1.8902683388306945e-05, + "loss": 1.0777, + "step": 14291 + }, + { + "epoch": 1.06, + "learning_rate": 1.8902501720583237e-05, + "loss": 1.0981, + "step": 14292 + }, + { + "epoch": 1.06, + "learning_rate": 1.8902320038695777e-05, + "loss": 0.9553, + "step": 14293 + }, + { + "epoch": 1.06, + "learning_rate": 1.890213834264485e-05, + "loss": 0.8919, + "step": 14294 + }, + { + "epoch": 1.06, + "learning_rate": 1.8901956632430753e-05, + "loss": 1.0745, + "step": 14295 + }, + { + "epoch": 1.06, + "learning_rate": 1.8901774908053768e-05, + "loss": 0.993, + "step": 14296 + }, + { + "epoch": 1.06, + "learning_rate": 1.8901593169514186e-05, + "loss": 1.1486, + "step": 14297 + }, + { + "epoch": 1.06, + "learning_rate": 1.8901411416812298e-05, + "loss": 1.0046, + "step": 14298 + }, + { + "epoch": 1.06, + "learning_rate": 1.890122964994839e-05, + "loss": 0.9602, + "step": 14299 + }, + { + "epoch": 1.06, + "learning_rate": 1.8901047868922755e-05, + "loss": 0.9485, + "step": 14300 + }, + { + "epoch": 1.06, + "learning_rate": 1.890086607373568e-05, + "loss": 1.0402, + "step": 14301 + }, + { + "epoch": 1.06, + "learning_rate": 1.8900684264387453e-05, + "loss": 1.0392, + "step": 14302 + }, + { + "epoch": 1.06, + "learning_rate": 1.8900502440878364e-05, + "loss": 1.0684, + "step": 14303 + }, + { + "epoch": 1.06, + "learning_rate": 1.8900320603208707e-05, + "loss": 1.0552, + "step": 14304 + }, + { + "epoch": 1.06, + "learning_rate": 1.8900138751378763e-05, + "loss": 1.1042, + "step": 14305 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899956885388827e-05, + "loss": 1.0063, + "step": 14306 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899775005239187e-05, + "loss": 1.0052, + "step": 14307 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899593110930134e-05, + "loss": 0.9897, + "step": 14308 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899411202461953e-05, + "loss": 1.0408, + "step": 14309 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899229279834935e-05, + "loss": 1.0738, + "step": 14310 + }, + { + "epoch": 1.06, + "learning_rate": 1.8899047343049375e-05, + "loss": 0.9743, + "step": 14311 + }, + { + "epoch": 1.06, + "learning_rate": 1.8898865392105555e-05, + "loss": 1.0145, + "step": 14312 + }, + { + "epoch": 1.06, + "learning_rate": 1.889868342700377e-05, + "loss": 1.0781, + "step": 14313 + }, + { + "epoch": 1.06, + "learning_rate": 1.8898501447744304e-05, + "loss": 0.9789, + "step": 14314 + }, + { + "epoch": 1.06, + "learning_rate": 1.8898319454327453e-05, + "loss": 1.0084, + "step": 14315 + }, + { + "epoch": 1.06, + "learning_rate": 1.88981374467535e-05, + "loss": 0.9803, + "step": 14316 + }, + { + "epoch": 1.06, + "learning_rate": 1.8897955425022737e-05, + "loss": 1.0376, + "step": 14317 + }, + { + "epoch": 1.06, + "learning_rate": 1.8897773389135457e-05, + "loss": 1.0138, + "step": 14318 + }, + { + "epoch": 1.06, + "learning_rate": 1.8897591339091943e-05, + "loss": 1.0344, + "step": 14319 + }, + { + "epoch": 1.06, + "learning_rate": 1.8897409274892492e-05, + "loss": 0.9984, + "step": 14320 + }, + { + "epoch": 1.06, + "learning_rate": 1.889722719653739e-05, + "loss": 1.0126, + "step": 14321 + }, + { + "epoch": 1.06, + "learning_rate": 1.8897045104026927e-05, + "loss": 1.0101, + "step": 14322 + }, + { + "epoch": 1.06, + "learning_rate": 1.8896862997361388e-05, + "loss": 1.0237, + "step": 14323 + }, + { + "epoch": 1.06, + "learning_rate": 1.8896680876541075e-05, + "loss": 1.0552, + "step": 14324 + }, + { + "epoch": 1.06, + "learning_rate": 1.889649874156626e-05, + "loss": 0.9229, + "step": 14325 + }, + { + "epoch": 1.06, + "learning_rate": 1.889631659243725e-05, + "loss": 1.0306, + "step": 14326 + }, + { + "epoch": 1.06, + "learning_rate": 1.8896134429154328e-05, + "loss": 1.0223, + "step": 14327 + }, + { + "epoch": 1.06, + "learning_rate": 1.889595225171778e-05, + "loss": 0.9445, + "step": 14328 + }, + { + "epoch": 1.06, + "learning_rate": 1.8895770060127905e-05, + "loss": 1.0527, + "step": 14329 + }, + { + "epoch": 1.06, + "learning_rate": 1.889558785438498e-05, + "loss": 1.0316, + "step": 14330 + }, + { + "epoch": 1.06, + "learning_rate": 1.8895405634489307e-05, + "loss": 0.9819, + "step": 14331 + }, + { + "epoch": 1.06, + "learning_rate": 1.889522340044117e-05, + "loss": 1.0178, + "step": 14332 + }, + { + "epoch": 1.06, + "learning_rate": 1.8895041152240857e-05, + "loss": 1.0683, + "step": 14333 + }, + { + "epoch": 1.06, + "learning_rate": 1.8894858889888663e-05, + "loss": 0.935, + "step": 14334 + }, + { + "epoch": 1.06, + "learning_rate": 1.8894676613384873e-05, + "loss": 0.9577, + "step": 14335 + }, + { + "epoch": 1.06, + "learning_rate": 1.8894494322729785e-05, + "loss": 0.9974, + "step": 14336 + }, + { + "epoch": 1.06, + "learning_rate": 1.889431201792368e-05, + "loss": 1.0436, + "step": 14337 + }, + { + "epoch": 1.06, + "learning_rate": 1.8894129698966853e-05, + "loss": 0.9731, + "step": 14338 + }, + { + "epoch": 1.06, + "learning_rate": 1.8893947365859592e-05, + "loss": 1.0287, + "step": 14339 + }, + { + "epoch": 1.06, + "learning_rate": 1.8893765018602188e-05, + "loss": 1.0474, + "step": 14340 + }, + { + "epoch": 1.06, + "learning_rate": 1.8893582657194933e-05, + "loss": 1.1152, + "step": 14341 + }, + { + "epoch": 1.06, + "learning_rate": 1.8893400281638114e-05, + "loss": 1.0309, + "step": 14342 + }, + { + "epoch": 1.06, + "learning_rate": 1.889321789193202e-05, + "loss": 1.0271, + "step": 14343 + }, + { + "epoch": 1.06, + "learning_rate": 1.8893035488076948e-05, + "loss": 0.9714, + "step": 14344 + }, + { + "epoch": 1.06, + "learning_rate": 1.889285307007318e-05, + "loss": 0.8873, + "step": 14345 + }, + { + "epoch": 1.06, + "learning_rate": 1.8892670637921015e-05, + "loss": 0.9701, + "step": 14346 + }, + { + "epoch": 1.06, + "learning_rate": 1.8892488191620733e-05, + "loss": 0.8714, + "step": 14347 + }, + { + "epoch": 1.06, + "learning_rate": 1.8892305731172632e-05, + "loss": 1.0292, + "step": 14348 + }, + { + "epoch": 1.06, + "learning_rate": 1.8892123256576998e-05, + "loss": 1.0743, + "step": 14349 + }, + { + "epoch": 1.06, + "learning_rate": 1.8891940767834124e-05, + "loss": 1.0011, + "step": 14350 + }, + { + "epoch": 1.06, + "learning_rate": 1.88917582649443e-05, + "loss": 1.0841, + "step": 14351 + }, + { + "epoch": 1.06, + "learning_rate": 1.8891575747907814e-05, + "loss": 0.9669, + "step": 14352 + }, + { + "epoch": 1.06, + "learning_rate": 1.889139321672496e-05, + "loss": 1.0806, + "step": 14353 + }, + { + "epoch": 1.06, + "learning_rate": 1.8891210671396028e-05, + "loss": 0.9525, + "step": 14354 + }, + { + "epoch": 1.06, + "learning_rate": 1.8891028111921305e-05, + "loss": 1.0184, + "step": 14355 + }, + { + "epoch": 1.06, + "learning_rate": 1.8890845538301085e-05, + "loss": 1.1092, + "step": 14356 + }, + { + "epoch": 1.06, + "learning_rate": 1.8890662950535655e-05, + "loss": 1.0468, + "step": 14357 + }, + { + "epoch": 1.06, + "learning_rate": 1.8890480348625307e-05, + "loss": 0.9838, + "step": 14358 + }, + { + "epoch": 1.06, + "learning_rate": 1.8890297732570335e-05, + "loss": 1.0141, + "step": 14359 + }, + { + "epoch": 1.06, + "learning_rate": 1.8890115102371025e-05, + "loss": 1.0685, + "step": 14360 + }, + { + "epoch": 1.06, + "learning_rate": 1.888993245802767e-05, + "loss": 1.0169, + "step": 14361 + }, + { + "epoch": 1.06, + "learning_rate": 1.888974979954056e-05, + "loss": 1.0563, + "step": 14362 + }, + { + "epoch": 1.06, + "learning_rate": 1.8889567126909984e-05, + "loss": 1.036, + "step": 14363 + }, + { + "epoch": 1.06, + "learning_rate": 1.8889384440136228e-05, + "loss": 1.0157, + "step": 14364 + }, + { + "epoch": 1.06, + "learning_rate": 1.8889201739219597e-05, + "loss": 0.9797, + "step": 14365 + }, + { + "epoch": 1.06, + "learning_rate": 1.8889019024160368e-05, + "loss": 1.0032, + "step": 14366 + }, + { + "epoch": 1.06, + "learning_rate": 1.8888836294958844e-05, + "loss": 0.8861, + "step": 14367 + }, + { + "epoch": 1.06, + "learning_rate": 1.88886535516153e-05, + "loss": 1.0154, + "step": 14368 + }, + { + "epoch": 1.06, + "learning_rate": 1.888847079413004e-05, + "loss": 1.0027, + "step": 14369 + }, + { + "epoch": 1.06, + "learning_rate": 1.888828802250335e-05, + "loss": 1.0456, + "step": 14370 + }, + { + "epoch": 1.06, + "learning_rate": 1.888810523673552e-05, + "loss": 1.0304, + "step": 14371 + }, + { + "epoch": 1.06, + "learning_rate": 1.8887922436826844e-05, + "loss": 1.0805, + "step": 14372 + }, + { + "epoch": 1.06, + "learning_rate": 1.888773962277761e-05, + "loss": 1.0832, + "step": 14373 + }, + { + "epoch": 1.06, + "learning_rate": 1.8887556794588104e-05, + "loss": 1.1065, + "step": 14374 + }, + { + "epoch": 1.06, + "learning_rate": 1.8887373952258628e-05, + "loss": 1.0045, + "step": 14375 + }, + { + "epoch": 1.06, + "learning_rate": 1.8887191095789465e-05, + "loss": 1.0047, + "step": 14376 + }, + { + "epoch": 1.06, + "learning_rate": 1.888700822518091e-05, + "loss": 0.9977, + "step": 14377 + }, + { + "epoch": 1.06, + "learning_rate": 1.888682534043325e-05, + "loss": 1.0355, + "step": 14378 + }, + { + "epoch": 1.06, + "learning_rate": 1.8886642441546777e-05, + "loss": 0.9936, + "step": 14379 + }, + { + "epoch": 1.06, + "learning_rate": 1.8886459528521787e-05, + "loss": 0.9346, + "step": 14380 + }, + { + "epoch": 1.06, + "learning_rate": 1.8886276601358563e-05, + "loss": 1.0129, + "step": 14381 + }, + { + "epoch": 1.06, + "learning_rate": 1.8886093660057407e-05, + "loss": 1.0404, + "step": 14382 + }, + { + "epoch": 1.06, + "learning_rate": 1.8885910704618596e-05, + "loss": 0.9486, + "step": 14383 + }, + { + "epoch": 1.06, + "learning_rate": 1.888572773504243e-05, + "loss": 0.9935, + "step": 14384 + }, + { + "epoch": 1.06, + "learning_rate": 1.88855447513292e-05, + "loss": 1.0486, + "step": 14385 + }, + { + "epoch": 1.06, + "learning_rate": 1.8885361753479192e-05, + "loss": 0.9993, + "step": 14386 + }, + { + "epoch": 1.06, + "learning_rate": 1.8885178741492703e-05, + "loss": 0.9999, + "step": 14387 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884995715370023e-05, + "loss": 0.9824, + "step": 14388 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884812675111438e-05, + "loss": 0.9966, + "step": 14389 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884629620717248e-05, + "loss": 1.0363, + "step": 14390 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884446552187735e-05, + "loss": 0.9204, + "step": 14391 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884263469523197e-05, + "loss": 0.9943, + "step": 14392 + }, + { + "epoch": 1.06, + "learning_rate": 1.8884080372723922e-05, + "loss": 1.0402, + "step": 14393 + }, + { + "epoch": 1.06, + "learning_rate": 1.8883897261790202e-05, + "loss": 1.0873, + "step": 14394 + }, + { + "epoch": 1.06, + "learning_rate": 1.888371413672233e-05, + "loss": 1.0506, + "step": 14395 + }, + { + "epoch": 1.06, + "learning_rate": 1.8883530997520597e-05, + "loss": 1.0648, + "step": 14396 + }, + { + "epoch": 1.06, + "learning_rate": 1.8883347844185288e-05, + "loss": 1.0012, + "step": 14397 + }, + { + "epoch": 1.06, + "learning_rate": 1.8883164676716706e-05, + "loss": 1.0069, + "step": 14398 + }, + { + "epoch": 1.06, + "learning_rate": 1.8882981495115132e-05, + "loss": 0.9107, + "step": 14399 + }, + { + "epoch": 1.06, + "learning_rate": 1.8882798299380864e-05, + "loss": 1.1593, + "step": 14400 + }, + { + "epoch": 1.06, + "learning_rate": 1.888261508951419e-05, + "loss": 1.0499, + "step": 14401 + }, + { + "epoch": 1.06, + "learning_rate": 1.88824318655154e-05, + "loss": 0.9111, + "step": 14402 + }, + { + "epoch": 1.06, + "learning_rate": 1.8882248627384793e-05, + "loss": 1.0843, + "step": 14403 + }, + { + "epoch": 1.06, + "learning_rate": 1.8882065375122653e-05, + "loss": 0.9175, + "step": 14404 + }, + { + "epoch": 1.06, + "learning_rate": 1.8881882108729273e-05, + "loss": 1.1086, + "step": 14405 + }, + { + "epoch": 1.06, + "learning_rate": 1.8881698828204947e-05, + "loss": 1.0292, + "step": 14406 + }, + { + "epoch": 1.06, + "learning_rate": 1.8881515533549963e-05, + "loss": 1.0276, + "step": 14407 + }, + { + "epoch": 1.06, + "learning_rate": 1.8881332224764617e-05, + "loss": 0.988, + "step": 14408 + }, + { + "epoch": 1.06, + "learning_rate": 1.88811489018492e-05, + "loss": 1.0134, + "step": 14409 + }, + { + "epoch": 1.06, + "learning_rate": 1.8880965564804e-05, + "loss": 0.9851, + "step": 14410 + }, + { + "epoch": 1.06, + "learning_rate": 1.888078221362931e-05, + "loss": 1.1358, + "step": 14411 + }, + { + "epoch": 1.06, + "learning_rate": 1.8880598848325426e-05, + "loss": 0.9658, + "step": 14412 + }, + { + "epoch": 1.06, + "learning_rate": 1.8880415468892632e-05, + "loss": 1.0503, + "step": 14413 + }, + { + "epoch": 1.07, + "learning_rate": 1.8880232075331228e-05, + "loss": 0.9732, + "step": 14414 + }, + { + "epoch": 1.07, + "learning_rate": 1.8880048667641496e-05, + "loss": 1.0523, + "step": 14415 + }, + { + "epoch": 1.07, + "learning_rate": 1.887986524582374e-05, + "loss": 1.1231, + "step": 14416 + }, + { + "epoch": 1.07, + "learning_rate": 1.8879681809878243e-05, + "loss": 1.0956, + "step": 14417 + }, + { + "epoch": 1.07, + "learning_rate": 1.8879498359805297e-05, + "loss": 0.9534, + "step": 14418 + }, + { + "epoch": 1.07, + "learning_rate": 1.88793148956052e-05, + "loss": 1.0821, + "step": 14419 + }, + { + "epoch": 1.07, + "learning_rate": 1.8879131417278237e-05, + "loss": 1.0442, + "step": 14420 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878947924824706e-05, + "loss": 1.1259, + "step": 14421 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878764418244894e-05, + "loss": 1.0248, + "step": 14422 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878580897539094e-05, + "loss": 0.9862, + "step": 14423 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878397362707603e-05, + "loss": 0.9766, + "step": 14424 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878213813750702e-05, + "loss": 1.0476, + "step": 14425 + }, + { + "epoch": 1.07, + "learning_rate": 1.8878030250668694e-05, + "loss": 1.124, + "step": 14426 + }, + { + "epoch": 1.07, + "learning_rate": 1.887784667346187e-05, + "loss": 1.0454, + "step": 14427 + }, + { + "epoch": 1.07, + "learning_rate": 1.8877663082130513e-05, + "loss": 1.027, + "step": 14428 + }, + { + "epoch": 1.07, + "learning_rate": 1.8877479476674922e-05, + "loss": 1.0043, + "step": 14429 + }, + { + "epoch": 1.07, + "learning_rate": 1.887729585709539e-05, + "loss": 1.0831, + "step": 14430 + }, + { + "epoch": 1.07, + "learning_rate": 1.887711222339221e-05, + "loss": 1.0876, + "step": 14431 + }, + { + "epoch": 1.07, + "learning_rate": 1.8876928575565666e-05, + "loss": 1.0141, + "step": 14432 + }, + { + "epoch": 1.07, + "learning_rate": 1.8876744913616058e-05, + "loss": 1.0409, + "step": 14433 + }, + { + "epoch": 1.07, + "learning_rate": 1.8876561237543676e-05, + "loss": 0.9879, + "step": 14434 + }, + { + "epoch": 1.07, + "learning_rate": 1.8876377547348813e-05, + "loss": 0.9313, + "step": 14435 + }, + { + "epoch": 1.07, + "learning_rate": 1.887619384303176e-05, + "loss": 1.058, + "step": 14436 + }, + { + "epoch": 1.07, + "learning_rate": 1.8876010124592808e-05, + "loss": 0.9645, + "step": 14437 + }, + { + "epoch": 1.07, + "learning_rate": 1.8875826392032253e-05, + "loss": 1.0494, + "step": 14438 + }, + { + "epoch": 1.07, + "learning_rate": 1.8875642645350382e-05, + "loss": 0.9785, + "step": 14439 + }, + { + "epoch": 1.07, + "learning_rate": 1.8875458884547492e-05, + "loss": 0.8746, + "step": 14440 + }, + { + "epoch": 1.07, + "learning_rate": 1.8875275109623875e-05, + "loss": 1.001, + "step": 14441 + }, + { + "epoch": 1.07, + "learning_rate": 1.8875091320579823e-05, + "loss": 1.0267, + "step": 14442 + }, + { + "epoch": 1.07, + "learning_rate": 1.8874907517415626e-05, + "loss": 1.0043, + "step": 14443 + }, + { + "epoch": 1.07, + "learning_rate": 1.887472370013158e-05, + "loss": 0.9196, + "step": 14444 + }, + { + "epoch": 1.07, + "learning_rate": 1.8874539868727972e-05, + "loss": 0.9568, + "step": 14445 + }, + { + "epoch": 1.07, + "learning_rate": 1.88743560232051e-05, + "loss": 1.0563, + "step": 14446 + }, + { + "epoch": 1.07, + "learning_rate": 1.8874172163563254e-05, + "loss": 0.9641, + "step": 14447 + }, + { + "epoch": 1.07, + "learning_rate": 1.887398828980273e-05, + "loss": 1.0576, + "step": 14448 + }, + { + "epoch": 1.07, + "learning_rate": 1.8873804401923815e-05, + "loss": 0.992, + "step": 14449 + }, + { + "epoch": 1.07, + "learning_rate": 1.8873620499926804e-05, + "loss": 0.8619, + "step": 14450 + }, + { + "epoch": 1.07, + "learning_rate": 1.887343658381199e-05, + "loss": 0.9208, + "step": 14451 + }, + { + "epoch": 1.07, + "learning_rate": 1.8873252653579665e-05, + "loss": 0.9994, + "step": 14452 + }, + { + "epoch": 1.07, + "learning_rate": 1.8873068709230122e-05, + "loss": 1.0247, + "step": 14453 + }, + { + "epoch": 1.07, + "learning_rate": 1.8872884750763655e-05, + "loss": 0.9789, + "step": 14454 + }, + { + "epoch": 1.07, + "learning_rate": 1.8872700778180556e-05, + "loss": 1.0265, + "step": 14455 + }, + { + "epoch": 1.07, + "learning_rate": 1.8872516791481114e-05, + "loss": 0.9953, + "step": 14456 + }, + { + "epoch": 1.07, + "learning_rate": 1.8872332790665625e-05, + "loss": 1.0343, + "step": 14457 + }, + { + "epoch": 1.07, + "learning_rate": 1.8872148775734386e-05, + "loss": 1.0162, + "step": 14458 + }, + { + "epoch": 1.07, + "learning_rate": 1.887196474668768e-05, + "loss": 1.0224, + "step": 14459 + }, + { + "epoch": 1.07, + "learning_rate": 1.887178070352581e-05, + "loss": 1.1175, + "step": 14460 + }, + { + "epoch": 1.07, + "learning_rate": 1.887159664624906e-05, + "loss": 0.9898, + "step": 14461 + }, + { + "epoch": 1.07, + "learning_rate": 1.8871412574857727e-05, + "loss": 1.0344, + "step": 14462 + }, + { + "epoch": 1.07, + "learning_rate": 1.8871228489352102e-05, + "loss": 1.0393, + "step": 14463 + }, + { + "epoch": 1.07, + "learning_rate": 1.8871044389732483e-05, + "loss": 0.9717, + "step": 14464 + }, + { + "epoch": 1.07, + "learning_rate": 1.887086027599916e-05, + "loss": 0.9709, + "step": 14465 + }, + { + "epoch": 1.07, + "learning_rate": 1.8870676148152422e-05, + "loss": 1.0283, + "step": 14466 + }, + { + "epoch": 1.07, + "learning_rate": 1.8870492006192566e-05, + "loss": 1.0258, + "step": 14467 + }, + { + "epoch": 1.07, + "learning_rate": 1.887030785011988e-05, + "loss": 0.9582, + "step": 14468 + }, + { + "epoch": 1.07, + "learning_rate": 1.887012367993467e-05, + "loss": 1.0832, + "step": 14469 + }, + { + "epoch": 1.07, + "learning_rate": 1.8869939495637216e-05, + "loss": 1.0031, + "step": 14470 + }, + { + "epoch": 1.07, + "learning_rate": 1.8869755297227812e-05, + "loss": 0.9884, + "step": 14471 + }, + { + "epoch": 1.07, + "learning_rate": 1.8869571084706758e-05, + "loss": 1.0594, + "step": 14472 + }, + { + "epoch": 1.07, + "learning_rate": 1.886938685807434e-05, + "loss": 0.9997, + "step": 14473 + }, + { + "epoch": 1.07, + "learning_rate": 1.886920261733086e-05, + "loss": 1.0874, + "step": 14474 + }, + { + "epoch": 1.07, + "learning_rate": 1.88690183624766e-05, + "loss": 0.9619, + "step": 14475 + }, + { + "epoch": 1.07, + "learning_rate": 1.8868834093511862e-05, + "loss": 1.0261, + "step": 14476 + }, + { + "epoch": 1.07, + "learning_rate": 1.8868649810436934e-05, + "loss": 1.0283, + "step": 14477 + }, + { + "epoch": 1.07, + "learning_rate": 1.886846551325211e-05, + "loss": 1.0689, + "step": 14478 + }, + { + "epoch": 1.07, + "learning_rate": 1.8868281201957688e-05, + "loss": 1.0545, + "step": 14479 + }, + { + "epoch": 1.07, + "learning_rate": 1.8868096876553955e-05, + "loss": 1.0815, + "step": 14480 + }, + { + "epoch": 1.07, + "learning_rate": 1.8867912537041205e-05, + "loss": 0.9862, + "step": 14481 + }, + { + "epoch": 1.07, + "learning_rate": 1.8867728183419735e-05, + "loss": 1.1025, + "step": 14482 + }, + { + "epoch": 1.07, + "learning_rate": 1.886754381568984e-05, + "loss": 0.9741, + "step": 14483 + }, + { + "epoch": 1.07, + "learning_rate": 1.8867359433851807e-05, + "loss": 1.0668, + "step": 14484 + }, + { + "epoch": 1.07, + "learning_rate": 1.886717503790593e-05, + "loss": 1.0731, + "step": 14485 + }, + { + "epoch": 1.07, + "learning_rate": 1.8866990627852506e-05, + "loss": 0.9688, + "step": 14486 + }, + { + "epoch": 1.07, + "learning_rate": 1.8866806203691827e-05, + "loss": 0.9568, + "step": 14487 + }, + { + "epoch": 1.07, + "learning_rate": 1.8866621765424183e-05, + "loss": 0.9238, + "step": 14488 + }, + { + "epoch": 1.07, + "learning_rate": 1.8866437313049873e-05, + "loss": 1.0332, + "step": 14489 + }, + { + "epoch": 1.07, + "learning_rate": 1.886625284656919e-05, + "loss": 0.9856, + "step": 14490 + }, + { + "epoch": 1.07, + "learning_rate": 1.8866068365982423e-05, + "loss": 1.0274, + "step": 14491 + }, + { + "epoch": 1.07, + "learning_rate": 1.886588387128987e-05, + "loss": 1.0258, + "step": 14492 + }, + { + "epoch": 1.07, + "learning_rate": 1.886569936249182e-05, + "loss": 0.9629, + "step": 14493 + }, + { + "epoch": 1.07, + "learning_rate": 1.8865514839588574e-05, + "loss": 0.8931, + "step": 14494 + }, + { + "epoch": 1.07, + "learning_rate": 1.8865330302580415e-05, + "loss": 0.9945, + "step": 14495 + }, + { + "epoch": 1.07, + "learning_rate": 1.8865145751467647e-05, + "loss": 0.9515, + "step": 14496 + }, + { + "epoch": 1.07, + "learning_rate": 1.8864961186250554e-05, + "loss": 0.9271, + "step": 14497 + }, + { + "epoch": 1.07, + "learning_rate": 1.8864776606929438e-05, + "loss": 0.9994, + "step": 14498 + }, + { + "epoch": 1.07, + "learning_rate": 1.886459201350459e-05, + "loss": 1.0204, + "step": 14499 + }, + { + "epoch": 1.07, + "learning_rate": 1.8864407405976303e-05, + "loss": 0.9988, + "step": 14500 + }, + { + "epoch": 1.07, + "learning_rate": 1.8864222784344867e-05, + "loss": 0.9867, + "step": 14501 + }, + { + "epoch": 1.07, + "learning_rate": 1.886403814861058e-05, + "loss": 1.0617, + "step": 14502 + }, + { + "epoch": 1.07, + "learning_rate": 1.8863853498773737e-05, + "loss": 0.8933, + "step": 14503 + }, + { + "epoch": 1.07, + "learning_rate": 1.8863668834834632e-05, + "loss": 1.044, + "step": 14504 + }, + { + "epoch": 1.07, + "learning_rate": 1.8863484156793553e-05, + "loss": 0.9771, + "step": 14505 + }, + { + "epoch": 1.07, + "learning_rate": 1.88632994646508e-05, + "loss": 0.998, + "step": 14506 + }, + { + "epoch": 1.07, + "learning_rate": 1.886311475840666e-05, + "loss": 1.1077, + "step": 14507 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862930038061436e-05, + "loss": 1.0275, + "step": 14508 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862745303615417e-05, + "loss": 1.0389, + "step": 14509 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862560555068895e-05, + "loss": 1.0303, + "step": 14510 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862375792422167e-05, + "loss": 1.0285, + "step": 14511 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862191015675523e-05, + "loss": 0.8201, + "step": 14512 + }, + { + "epoch": 1.07, + "learning_rate": 1.8862006224829263e-05, + "loss": 1.0075, + "step": 14513 + }, + { + "epoch": 1.07, + "learning_rate": 1.886182141988368e-05, + "loss": 0.9761, + "step": 14514 + }, + { + "epoch": 1.07, + "learning_rate": 1.886163660083906e-05, + "loss": 0.961, + "step": 14515 + }, + { + "epoch": 1.07, + "learning_rate": 1.886145176769571e-05, + "loss": 0.9448, + "step": 14516 + }, + { + "epoch": 1.07, + "learning_rate": 1.886126692045391e-05, + "loss": 0.9905, + "step": 14517 + }, + { + "epoch": 1.07, + "learning_rate": 1.8861082059113967e-05, + "loss": 0.97, + "step": 14518 + }, + { + "epoch": 1.07, + "learning_rate": 1.8860897183676163e-05, + "loss": 1.0745, + "step": 14519 + }, + { + "epoch": 1.07, + "learning_rate": 1.8860712294140804e-05, + "loss": 0.9503, + "step": 14520 + }, + { + "epoch": 1.07, + "learning_rate": 1.8860527390508176e-05, + "loss": 1.0998, + "step": 14521 + }, + { + "epoch": 1.07, + "learning_rate": 1.8860342472778576e-05, + "loss": 1.0275, + "step": 14522 + }, + { + "epoch": 1.07, + "learning_rate": 1.88601575409523e-05, + "loss": 1.0174, + "step": 14523 + }, + { + "epoch": 1.07, + "learning_rate": 1.8859972595029636e-05, + "loss": 1.0618, + "step": 14524 + }, + { + "epoch": 1.07, + "learning_rate": 1.8859787635010883e-05, + "loss": 1.0589, + "step": 14525 + }, + { + "epoch": 1.07, + "learning_rate": 1.885960266089634e-05, + "loss": 1.0927, + "step": 14526 + }, + { + "epoch": 1.07, + "learning_rate": 1.885941767268629e-05, + "loss": 1.0933, + "step": 14527 + }, + { + "epoch": 1.07, + "learning_rate": 1.8859232670381036e-05, + "loss": 1.0972, + "step": 14528 + }, + { + "epoch": 1.07, + "learning_rate": 1.885904765398087e-05, + "loss": 0.9306, + "step": 14529 + }, + { + "epoch": 1.07, + "learning_rate": 1.885886262348608e-05, + "loss": 1.0704, + "step": 14530 + }, + { + "epoch": 1.07, + "learning_rate": 1.8858677578896973e-05, + "loss": 1.0876, + "step": 14531 + }, + { + "epoch": 1.07, + "learning_rate": 1.8858492520213837e-05, + "loss": 0.9124, + "step": 14532 + }, + { + "epoch": 1.07, + "learning_rate": 1.885830744743696e-05, + "loss": 0.9101, + "step": 14533 + }, + { + "epoch": 1.07, + "learning_rate": 1.885812236056665e-05, + "loss": 1.0317, + "step": 14534 + }, + { + "epoch": 1.07, + "learning_rate": 1.885793725960319e-05, + "loss": 1.0521, + "step": 14535 + }, + { + "epoch": 1.07, + "learning_rate": 1.885775214454688e-05, + "loss": 0.962, + "step": 14536 + }, + { + "epoch": 1.07, + "learning_rate": 1.8857567015398012e-05, + "loss": 1.0016, + "step": 14537 + }, + { + "epoch": 1.07, + "learning_rate": 1.8857381872156878e-05, + "loss": 1.0585, + "step": 14538 + }, + { + "epoch": 1.07, + "learning_rate": 1.885719671482378e-05, + "loss": 1.0727, + "step": 14539 + }, + { + "epoch": 1.07, + "learning_rate": 1.885701154339901e-05, + "loss": 0.9537, + "step": 14540 + }, + { + "epoch": 1.07, + "learning_rate": 1.8856826357882862e-05, + "loss": 0.9993, + "step": 14541 + }, + { + "epoch": 1.07, + "learning_rate": 1.885664115827563e-05, + "loss": 1.0404, + "step": 14542 + }, + { + "epoch": 1.07, + "learning_rate": 1.8856455944577602e-05, + "loss": 1.0325, + "step": 14543 + }, + { + "epoch": 1.07, + "learning_rate": 1.8856270716789088e-05, + "loss": 1.0316, + "step": 14544 + }, + { + "epoch": 1.07, + "learning_rate": 1.8856085474910367e-05, + "loss": 0.8965, + "step": 14545 + }, + { + "epoch": 1.07, + "learning_rate": 1.8855900218941744e-05, + "loss": 1.0541, + "step": 14546 + }, + { + "epoch": 1.07, + "learning_rate": 1.885571494888351e-05, + "loss": 1.1815, + "step": 14547 + }, + { + "epoch": 1.07, + "learning_rate": 1.885552966473596e-05, + "loss": 0.9555, + "step": 14548 + }, + { + "epoch": 1.08, + "learning_rate": 1.8855344366499392e-05, + "loss": 1.0505, + "step": 14549 + }, + { + "epoch": 1.08, + "learning_rate": 1.8855159054174096e-05, + "loss": 1.0001, + "step": 14550 + }, + { + "epoch": 1.08, + "learning_rate": 1.8854973727760365e-05, + "loss": 1.0325, + "step": 14551 + }, + { + "epoch": 1.08, + "learning_rate": 1.88547883872585e-05, + "loss": 0.9321, + "step": 14552 + }, + { + "epoch": 1.08, + "learning_rate": 1.8854603032668796e-05, + "loss": 0.9457, + "step": 14553 + }, + { + "epoch": 1.08, + "learning_rate": 1.8854417663991543e-05, + "loss": 0.9457, + "step": 14554 + }, + { + "epoch": 1.08, + "learning_rate": 1.8854232281227038e-05, + "loss": 0.9684, + "step": 14555 + }, + { + "epoch": 1.08, + "learning_rate": 1.8854046884375576e-05, + "loss": 1.0208, + "step": 14556 + }, + { + "epoch": 1.08, + "learning_rate": 1.8853861473437454e-05, + "loss": 1.1359, + "step": 14557 + }, + { + "epoch": 1.08, + "learning_rate": 1.8853676048412964e-05, + "loss": 1.083, + "step": 14558 + }, + { + "epoch": 1.08, + "learning_rate": 1.8853490609302398e-05, + "loss": 1.0026, + "step": 14559 + }, + { + "epoch": 1.08, + "learning_rate": 1.885330515610606e-05, + "loss": 0.9259, + "step": 14560 + }, + { + "epoch": 1.08, + "learning_rate": 1.885311968882424e-05, + "loss": 0.9981, + "step": 14561 + }, + { + "epoch": 1.08, + "learning_rate": 1.885293420745723e-05, + "loss": 1.094, + "step": 14562 + }, + { + "epoch": 1.08, + "learning_rate": 1.8852748712005333e-05, + "loss": 1.0316, + "step": 14563 + }, + { + "epoch": 1.08, + "learning_rate": 1.8852563202468835e-05, + "loss": 1.0518, + "step": 14564 + }, + { + "epoch": 1.08, + "learning_rate": 1.885237767884804e-05, + "loss": 1.0209, + "step": 14565 + }, + { + "epoch": 1.08, + "learning_rate": 1.8852192141143235e-05, + "loss": 0.9822, + "step": 14566 + }, + { + "epoch": 1.08, + "learning_rate": 1.8852006589354718e-05, + "loss": 0.9762, + "step": 14567 + }, + { + "epoch": 1.08, + "learning_rate": 1.8851821023482788e-05, + "loss": 1.0781, + "step": 14568 + }, + { + "epoch": 1.08, + "learning_rate": 1.885163544352774e-05, + "loss": 1.1331, + "step": 14569 + }, + { + "epoch": 1.08, + "learning_rate": 1.885144984948986e-05, + "loss": 1.1739, + "step": 14570 + }, + { + "epoch": 1.08, + "learning_rate": 1.8851264241369453e-05, + "loss": 1.0439, + "step": 14571 + }, + { + "epoch": 1.08, + "learning_rate": 1.8851078619166813e-05, + "loss": 1.0275, + "step": 14572 + }, + { + "epoch": 1.08, + "learning_rate": 1.8850892982882233e-05, + "loss": 1.0356, + "step": 14573 + }, + { + "epoch": 1.08, + "learning_rate": 1.885070733251601e-05, + "loss": 0.9073, + "step": 14574 + }, + { + "epoch": 1.08, + "learning_rate": 1.8850521668068434e-05, + "loss": 1.0299, + "step": 14575 + }, + { + "epoch": 1.08, + "learning_rate": 1.8850335989539807e-05, + "loss": 1.0538, + "step": 14576 + }, + { + "epoch": 1.08, + "learning_rate": 1.885015029693042e-05, + "loss": 0.9403, + "step": 14577 + }, + { + "epoch": 1.08, + "learning_rate": 1.8849964590240576e-05, + "loss": 1.0584, + "step": 14578 + }, + { + "epoch": 1.08, + "learning_rate": 1.8849778869470562e-05, + "loss": 0.9425, + "step": 14579 + }, + { + "epoch": 1.08, + "learning_rate": 1.8849593134620675e-05, + "loss": 1.0726, + "step": 14580 + }, + { + "epoch": 1.08, + "learning_rate": 1.8849407385691213e-05, + "loss": 0.973, + "step": 14581 + }, + { + "epoch": 1.08, + "learning_rate": 1.884922162268247e-05, + "loss": 0.9885, + "step": 14582 + }, + { + "epoch": 1.08, + "learning_rate": 1.8849035845594744e-05, + "loss": 1.0233, + "step": 14583 + }, + { + "epoch": 1.08, + "learning_rate": 1.8848850054428326e-05, + "loss": 0.9712, + "step": 14584 + }, + { + "epoch": 1.08, + "learning_rate": 1.8848664249183516e-05, + "loss": 1.0024, + "step": 14585 + }, + { + "epoch": 1.08, + "learning_rate": 1.8848478429860605e-05, + "loss": 1.1537, + "step": 14586 + }, + { + "epoch": 1.08, + "learning_rate": 1.8848292596459892e-05, + "loss": 1.0704, + "step": 14587 + }, + { + "epoch": 1.08, + "learning_rate": 1.8848106748981673e-05, + "loss": 0.8711, + "step": 14588 + }, + { + "epoch": 1.08, + "learning_rate": 1.8847920887426245e-05, + "loss": 0.9601, + "step": 14589 + }, + { + "epoch": 1.08, + "learning_rate": 1.88477350117939e-05, + "loss": 1.0824, + "step": 14590 + }, + { + "epoch": 1.08, + "learning_rate": 1.8847549122084935e-05, + "loss": 0.9414, + "step": 14591 + }, + { + "epoch": 1.08, + "learning_rate": 1.8847363218299643e-05, + "loss": 0.9409, + "step": 14592 + }, + { + "epoch": 1.08, + "learning_rate": 1.884717730043833e-05, + "loss": 1.09, + "step": 14593 + }, + { + "epoch": 1.08, + "learning_rate": 1.8846991368501277e-05, + "loss": 1.0906, + "step": 14594 + }, + { + "epoch": 1.08, + "learning_rate": 1.8846805422488793e-05, + "loss": 1.0089, + "step": 14595 + }, + { + "epoch": 1.08, + "learning_rate": 1.884661946240116e-05, + "loss": 0.9829, + "step": 14596 + }, + { + "epoch": 1.08, + "learning_rate": 1.884643348823869e-05, + "loss": 1.0379, + "step": 14597 + }, + { + "epoch": 1.08, + "learning_rate": 1.884624750000167e-05, + "loss": 0.9392, + "step": 14598 + }, + { + "epoch": 1.08, + "learning_rate": 1.8846061497690392e-05, + "loss": 1.0281, + "step": 14599 + }, + { + "epoch": 1.08, + "learning_rate": 1.884587548130516e-05, + "loss": 0.9772, + "step": 14600 + }, + { + "epoch": 1.08, + "learning_rate": 1.8845689450846268e-05, + "loss": 1.0584, + "step": 14601 + }, + { + "epoch": 1.08, + "learning_rate": 1.884550340631401e-05, + "loss": 1.0571, + "step": 14602 + }, + { + "epoch": 1.08, + "learning_rate": 1.884531734770868e-05, + "loss": 1.0555, + "step": 14603 + }, + { + "epoch": 1.08, + "learning_rate": 1.884513127503058e-05, + "loss": 1.0665, + "step": 14604 + }, + { + "epoch": 1.08, + "learning_rate": 1.884494518828e-05, + "loss": 1.0778, + "step": 14605 + }, + { + "epoch": 1.08, + "learning_rate": 1.8844759087457237e-05, + "loss": 0.9609, + "step": 14606 + }, + { + "epoch": 1.08, + "learning_rate": 1.8844572972562596e-05, + "loss": 1.023, + "step": 14607 + }, + { + "epoch": 1.08, + "learning_rate": 1.8844386843596362e-05, + "loss": 0.9872, + "step": 14608 + }, + { + "epoch": 1.08, + "learning_rate": 1.8844200700558835e-05, + "loss": 1.1429, + "step": 14609 + }, + { + "epoch": 1.08, + "learning_rate": 1.884401454345031e-05, + "loss": 0.9534, + "step": 14610 + }, + { + "epoch": 1.08, + "learning_rate": 1.8843828372271086e-05, + "loss": 1.0436, + "step": 14611 + }, + { + "epoch": 1.08, + "learning_rate": 1.8843642187021455e-05, + "loss": 1.0094, + "step": 14612 + }, + { + "epoch": 1.08, + "learning_rate": 1.8843455987701718e-05, + "loss": 1.0318, + "step": 14613 + }, + { + "epoch": 1.08, + "learning_rate": 1.884326977431217e-05, + "loss": 0.9424, + "step": 14614 + }, + { + "epoch": 1.08, + "learning_rate": 1.8843083546853106e-05, + "loss": 1.088, + "step": 14615 + }, + { + "epoch": 1.08, + "learning_rate": 1.884289730532482e-05, + "loss": 1.1065, + "step": 14616 + }, + { + "epoch": 1.08, + "learning_rate": 1.8842711049727612e-05, + "loss": 1.0009, + "step": 14617 + }, + { + "epoch": 1.08, + "learning_rate": 1.884252478006178e-05, + "loss": 1.0612, + "step": 14618 + }, + { + "epoch": 1.08, + "learning_rate": 1.8842338496327614e-05, + "loss": 1.133, + "step": 14619 + }, + { + "epoch": 1.08, + "learning_rate": 1.8842152198525413e-05, + "loss": 0.9683, + "step": 14620 + }, + { + "epoch": 1.08, + "learning_rate": 1.884196588665548e-05, + "loss": 1.0935, + "step": 14621 + }, + { + "epoch": 1.08, + "learning_rate": 1.88417795607181e-05, + "loss": 1.017, + "step": 14622 + }, + { + "epoch": 1.08, + "learning_rate": 1.884159322071358e-05, + "loss": 1.0349, + "step": 14623 + }, + { + "epoch": 1.08, + "learning_rate": 1.884140686664221e-05, + "loss": 0.9666, + "step": 14624 + }, + { + "epoch": 1.08, + "learning_rate": 1.8841220498504287e-05, + "loss": 1.0465, + "step": 14625 + }, + { + "epoch": 1.08, + "learning_rate": 1.8841034116300108e-05, + "loss": 1.0686, + "step": 14626 + }, + { + "epoch": 1.08, + "learning_rate": 1.8840847720029972e-05, + "loss": 0.9283, + "step": 14627 + }, + { + "epoch": 1.08, + "learning_rate": 1.8840661309694176e-05, + "loss": 1.0714, + "step": 14628 + }, + { + "epoch": 1.08, + "learning_rate": 1.884047488529301e-05, + "loss": 1.0126, + "step": 14629 + }, + { + "epoch": 1.08, + "learning_rate": 1.8840288446826777e-05, + "loss": 1.0589, + "step": 14630 + }, + { + "epoch": 1.08, + "learning_rate": 1.884010199429577e-05, + "loss": 1.0343, + "step": 14631 + }, + { + "epoch": 1.08, + "learning_rate": 1.883991552770029e-05, + "loss": 1.0578, + "step": 14632 + }, + { + "epoch": 1.08, + "learning_rate": 1.883972904704063e-05, + "loss": 0.9949, + "step": 14633 + }, + { + "epoch": 1.08, + "learning_rate": 1.8839542552317086e-05, + "loss": 0.9592, + "step": 14634 + }, + { + "epoch": 1.08, + "learning_rate": 1.8839356043529956e-05, + "loss": 0.9462, + "step": 14635 + }, + { + "epoch": 1.08, + "learning_rate": 1.8839169520679538e-05, + "loss": 0.9244, + "step": 14636 + }, + { + "epoch": 1.08, + "learning_rate": 1.8838982983766126e-05, + "loss": 1.0504, + "step": 14637 + }, + { + "epoch": 1.08, + "learning_rate": 1.8838796432790023e-05, + "loss": 0.9167, + "step": 14638 + }, + { + "epoch": 1.08, + "learning_rate": 1.883860986775152e-05, + "loss": 1.0455, + "step": 14639 + }, + { + "epoch": 1.08, + "learning_rate": 1.8838423288650914e-05, + "loss": 1.0501, + "step": 14640 + }, + { + "epoch": 1.08, + "learning_rate": 1.88382366954885e-05, + "loss": 0.987, + "step": 14641 + }, + { + "epoch": 1.08, + "learning_rate": 1.8838050088264582e-05, + "loss": 1.0769, + "step": 14642 + }, + { + "epoch": 1.08, + "learning_rate": 1.8837863466979452e-05, + "loss": 1.0563, + "step": 14643 + }, + { + "epoch": 1.08, + "learning_rate": 1.8837676831633406e-05, + "loss": 1.0794, + "step": 14644 + }, + { + "epoch": 1.08, + "learning_rate": 1.8837490182226745e-05, + "loss": 1.1556, + "step": 14645 + }, + { + "epoch": 1.08, + "learning_rate": 1.8837303518759763e-05, + "loss": 0.9557, + "step": 14646 + }, + { + "epoch": 1.08, + "learning_rate": 1.8837116841232757e-05, + "loss": 1.0178, + "step": 14647 + }, + { + "epoch": 1.08, + "learning_rate": 1.8836930149646024e-05, + "loss": 0.9965, + "step": 14648 + }, + { + "epoch": 1.08, + "learning_rate": 1.883674344399986e-05, + "loss": 0.9537, + "step": 14649 + }, + { + "epoch": 1.08, + "learning_rate": 1.8836556724294568e-05, + "loss": 0.9573, + "step": 14650 + }, + { + "epoch": 1.08, + "learning_rate": 1.8836369990530435e-05, + "loss": 0.9135, + "step": 14651 + }, + { + "epoch": 1.08, + "learning_rate": 1.8836183242707768e-05, + "loss": 0.9078, + "step": 14652 + }, + { + "epoch": 1.08, + "learning_rate": 1.883599648082686e-05, + "loss": 1.1052, + "step": 14653 + }, + { + "epoch": 1.08, + "learning_rate": 1.883580970488801e-05, + "loss": 1.0208, + "step": 14654 + }, + { + "epoch": 1.08, + "learning_rate": 1.8835622914891508e-05, + "loss": 0.9787, + "step": 14655 + }, + { + "epoch": 1.08, + "learning_rate": 1.883543611083766e-05, + "loss": 1.0269, + "step": 14656 + }, + { + "epoch": 1.08, + "learning_rate": 1.883524929272676e-05, + "loss": 1.0374, + "step": 14657 + }, + { + "epoch": 1.08, + "learning_rate": 1.88350624605591e-05, + "loss": 1.0903, + "step": 14658 + }, + { + "epoch": 1.08, + "learning_rate": 1.8834875614334985e-05, + "loss": 1.0653, + "step": 14659 + }, + { + "epoch": 1.08, + "learning_rate": 1.8834688754054708e-05, + "loss": 0.9445, + "step": 14660 + }, + { + "epoch": 1.08, + "learning_rate": 1.883450187971857e-05, + "loss": 0.9888, + "step": 14661 + }, + { + "epoch": 1.08, + "learning_rate": 1.883431499132686e-05, + "loss": 1.1221, + "step": 14662 + }, + { + "epoch": 1.08, + "learning_rate": 1.8834128088879892e-05, + "loss": 0.9628, + "step": 14663 + }, + { + "epoch": 1.08, + "learning_rate": 1.8833941172377945e-05, + "loss": 1.0522, + "step": 14664 + }, + { + "epoch": 1.08, + "learning_rate": 1.8833754241821327e-05, + "loss": 1.0043, + "step": 14665 + }, + { + "epoch": 1.08, + "learning_rate": 1.883356729721033e-05, + "loss": 1.1299, + "step": 14666 + }, + { + "epoch": 1.08, + "learning_rate": 1.8833380338545252e-05, + "loss": 1.0417, + "step": 14667 + }, + { + "epoch": 1.08, + "learning_rate": 1.88331933658264e-05, + "loss": 1.0213, + "step": 14668 + }, + { + "epoch": 1.08, + "learning_rate": 1.8833006379054058e-05, + "loss": 0.9409, + "step": 14669 + }, + { + "epoch": 1.08, + "learning_rate": 1.8832819378228528e-05, + "loss": 1.1011, + "step": 14670 + }, + { + "epoch": 1.08, + "learning_rate": 1.883263236335011e-05, + "loss": 1.082, + "step": 14671 + }, + { + "epoch": 1.08, + "learning_rate": 1.8832445334419103e-05, + "loss": 1.1868, + "step": 14672 + }, + { + "epoch": 1.08, + "learning_rate": 1.8832258291435798e-05, + "loss": 0.9972, + "step": 14673 + }, + { + "epoch": 1.08, + "learning_rate": 1.88320712344005e-05, + "loss": 1.0977, + "step": 14674 + }, + { + "epoch": 1.08, + "learning_rate": 1.88318841633135e-05, + "loss": 1.0525, + "step": 14675 + }, + { + "epoch": 1.08, + "learning_rate": 1.8831697078175103e-05, + "loss": 0.9602, + "step": 14676 + }, + { + "epoch": 1.08, + "learning_rate": 1.8831509978985595e-05, + "loss": 0.9152, + "step": 14677 + }, + { + "epoch": 1.08, + "learning_rate": 1.8831322865745288e-05, + "loss": 1.0148, + "step": 14678 + }, + { + "epoch": 1.08, + "learning_rate": 1.8831135738454473e-05, + "loss": 1.0877, + "step": 14679 + }, + { + "epoch": 1.08, + "learning_rate": 1.883094859711344e-05, + "loss": 1.0882, + "step": 14680 + }, + { + "epoch": 1.08, + "learning_rate": 1.8830761441722497e-05, + "loss": 1.0034, + "step": 14681 + }, + { + "epoch": 1.08, + "learning_rate": 1.8830574272281944e-05, + "loss": 1.0217, + "step": 14682 + }, + { + "epoch": 1.08, + "learning_rate": 1.883038708879207e-05, + "loss": 1.0085, + "step": 14683 + }, + { + "epoch": 1.08, + "learning_rate": 1.8830199891253175e-05, + "loss": 1.0938, + "step": 14684 + }, + { + "epoch": 1.09, + "learning_rate": 1.883001267966556e-05, + "loss": 1.0016, + "step": 14685 + }, + { + "epoch": 1.09, + "learning_rate": 1.882982545402952e-05, + "loss": 1.0294, + "step": 14686 + }, + { + "epoch": 1.09, + "learning_rate": 1.8829638214345357e-05, + "loss": 0.9867, + "step": 14687 + }, + { + "epoch": 1.09, + "learning_rate": 1.8829450960613364e-05, + "loss": 1.0221, + "step": 14688 + }, + { + "epoch": 1.09, + "learning_rate": 1.882926369283384e-05, + "loss": 1.0616, + "step": 14689 + }, + { + "epoch": 1.09, + "learning_rate": 1.8829076411007085e-05, + "loss": 1.0223, + "step": 14690 + }, + { + "epoch": 1.09, + "learning_rate": 1.8828889115133396e-05, + "loss": 1.0313, + "step": 14691 + }, + { + "epoch": 1.09, + "learning_rate": 1.882870180521307e-05, + "loss": 1.0169, + "step": 14692 + }, + { + "epoch": 1.09, + "learning_rate": 1.8828514481246407e-05, + "loss": 1.1228, + "step": 14693 + }, + { + "epoch": 1.09, + "learning_rate": 1.88283271432337e-05, + "loss": 0.9044, + "step": 14694 + }, + { + "epoch": 1.09, + "learning_rate": 1.8828139791175255e-05, + "loss": 1.0469, + "step": 14695 + }, + { + "epoch": 1.09, + "learning_rate": 1.8827952425071364e-05, + "loss": 1.0731, + "step": 14696 + }, + { + "epoch": 1.09, + "learning_rate": 1.882776504492233e-05, + "loss": 1.0218, + "step": 14697 + }, + { + "epoch": 1.09, + "learning_rate": 1.8827577650728443e-05, + "loss": 1.0103, + "step": 14698 + }, + { + "epoch": 1.09, + "learning_rate": 1.882739024249001e-05, + "loss": 1.0492, + "step": 14699 + }, + { + "epoch": 1.09, + "learning_rate": 1.8827202820207323e-05, + "loss": 1.1137, + "step": 14700 + }, + { + "epoch": 1.09, + "learning_rate": 1.8827015383880684e-05, + "loss": 1.0439, + "step": 14701 + }, + { + "epoch": 1.09, + "learning_rate": 1.882682793351039e-05, + "loss": 1.1366, + "step": 14702 + }, + { + "epoch": 1.09, + "learning_rate": 1.8826640469096736e-05, + "loss": 1.0431, + "step": 14703 + }, + { + "epoch": 1.09, + "learning_rate": 1.8826452990640027e-05, + "loss": 1.048, + "step": 14704 + }, + { + "epoch": 1.09, + "learning_rate": 1.8826265498140558e-05, + "loss": 1.0281, + "step": 14705 + }, + { + "epoch": 1.09, + "learning_rate": 1.8826077991598625e-05, + "loss": 0.958, + "step": 14706 + }, + { + "epoch": 1.09, + "learning_rate": 1.882589047101453e-05, + "loss": 0.9756, + "step": 14707 + }, + { + "epoch": 1.09, + "learning_rate": 1.8825702936388568e-05, + "loss": 1.0533, + "step": 14708 + }, + { + "epoch": 1.09, + "learning_rate": 1.882551538772104e-05, + "loss": 1.0717, + "step": 14709 + }, + { + "epoch": 1.09, + "learning_rate": 1.882532782501224e-05, + "loss": 0.9981, + "step": 14710 + }, + { + "epoch": 1.09, + "learning_rate": 1.8825140248262473e-05, + "loss": 1.0317, + "step": 14711 + }, + { + "epoch": 1.09, + "learning_rate": 1.8824952657472034e-05, + "loss": 1.0599, + "step": 14712 + }, + { + "epoch": 1.09, + "learning_rate": 1.8824765052641218e-05, + "loss": 1.0941, + "step": 14713 + }, + { + "epoch": 1.09, + "learning_rate": 1.882457743377033e-05, + "loss": 0.9916, + "step": 14714 + }, + { + "epoch": 1.09, + "learning_rate": 1.882438980085967e-05, + "loss": 0.9565, + "step": 14715 + }, + { + "epoch": 1.09, + "learning_rate": 1.8824202153909526e-05, + "loss": 1.0274, + "step": 14716 + }, + { + "epoch": 1.09, + "learning_rate": 1.8824014492920205e-05, + "loss": 1.0358, + "step": 14717 + }, + { + "epoch": 1.09, + "learning_rate": 1.8823826817892e-05, + "loss": 0.967, + "step": 14718 + }, + { + "epoch": 1.09, + "learning_rate": 1.8823639128825214e-05, + "loss": 1.0804, + "step": 14719 + }, + { + "epoch": 1.09, + "learning_rate": 1.8823451425720148e-05, + "loss": 1.0269, + "step": 14720 + }, + { + "epoch": 1.09, + "learning_rate": 1.8823263708577093e-05, + "loss": 1.0995, + "step": 14721 + }, + { + "epoch": 1.09, + "learning_rate": 1.8823075977396354e-05, + "loss": 1.0482, + "step": 14722 + }, + { + "epoch": 1.09, + "learning_rate": 1.8822888232178225e-05, + "loss": 0.9776, + "step": 14723 + }, + { + "epoch": 1.09, + "learning_rate": 1.882270047292301e-05, + "loss": 1.0015, + "step": 14724 + }, + { + "epoch": 1.09, + "learning_rate": 1.8822512699631e-05, + "loss": 1.0812, + "step": 14725 + }, + { + "epoch": 1.09, + "learning_rate": 1.8822324912302504e-05, + "loss": 1.0321, + "step": 14726 + }, + { + "epoch": 1.09, + "learning_rate": 1.8822137110937814e-05, + "loss": 1.0106, + "step": 14727 + }, + { + "epoch": 1.09, + "learning_rate": 1.882194929553723e-05, + "loss": 1.0944, + "step": 14728 + }, + { + "epoch": 1.09, + "learning_rate": 1.8821761466101048e-05, + "loss": 1.0153, + "step": 14729 + }, + { + "epoch": 1.09, + "learning_rate": 1.8821573622629575e-05, + "loss": 1.0273, + "step": 14730 + }, + { + "epoch": 1.09, + "learning_rate": 1.8821385765123097e-05, + "loss": 0.9911, + "step": 14731 + }, + { + "epoch": 1.09, + "learning_rate": 1.8821197893581928e-05, + "loss": 1.0523, + "step": 14732 + }, + { + "epoch": 1.09, + "learning_rate": 1.8821010008006358e-05, + "loss": 1.0719, + "step": 14733 + }, + { + "epoch": 1.09, + "learning_rate": 1.8820822108396685e-05, + "loss": 0.9081, + "step": 14734 + }, + { + "epoch": 1.09, + "learning_rate": 1.882063419475321e-05, + "loss": 1.0278, + "step": 14735 + }, + { + "epoch": 1.09, + "learning_rate": 1.8820446267076233e-05, + "loss": 1.0451, + "step": 14736 + }, + { + "epoch": 1.09, + "learning_rate": 1.882025832536605e-05, + "loss": 1.0784, + "step": 14737 + }, + { + "epoch": 1.09, + "learning_rate": 1.8820070369622966e-05, + "loss": 0.9324, + "step": 14738 + }, + { + "epoch": 1.09, + "learning_rate": 1.8819882399847276e-05, + "loss": 1.0124, + "step": 14739 + }, + { + "epoch": 1.09, + "learning_rate": 1.8819694416039276e-05, + "loss": 0.9785, + "step": 14740 + }, + { + "epoch": 1.09, + "learning_rate": 1.8819506418199273e-05, + "loss": 1.0784, + "step": 14741 + }, + { + "epoch": 1.09, + "learning_rate": 1.8819318406327557e-05, + "loss": 0.982, + "step": 14742 + }, + { + "epoch": 1.09, + "learning_rate": 1.8819130380424434e-05, + "loss": 0.9642, + "step": 14743 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818942340490198e-05, + "loss": 1.0273, + "step": 14744 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818754286525155e-05, + "loss": 1.0451, + "step": 14745 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818566218529598e-05, + "loss": 1.0929, + "step": 14746 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818378136503827e-05, + "loss": 1.0867, + "step": 14747 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818190040448146e-05, + "loss": 0.9711, + "step": 14748 + }, + { + "epoch": 1.09, + "learning_rate": 1.8818001930362847e-05, + "loss": 0.97, + "step": 14749 + }, + { + "epoch": 1.09, + "learning_rate": 1.8817813806248237e-05, + "loss": 1.0155, + "step": 14750 + }, + { + "epoch": 1.09, + "learning_rate": 1.881762566810461e-05, + "loss": 0.993, + "step": 14751 + }, + { + "epoch": 1.09, + "learning_rate": 1.8817437515932263e-05, + "loss": 0.9638, + "step": 14752 + }, + { + "epoch": 1.09, + "learning_rate": 1.8817249349731502e-05, + "loss": 0.9429, + "step": 14753 + }, + { + "epoch": 1.09, + "learning_rate": 1.8817061169502622e-05, + "loss": 1.0959, + "step": 14754 + }, + { + "epoch": 1.09, + "learning_rate": 1.881687297524593e-05, + "loss": 0.9526, + "step": 14755 + }, + { + "epoch": 1.09, + "learning_rate": 1.8816684766961715e-05, + "loss": 1.0456, + "step": 14756 + }, + { + "epoch": 1.09, + "learning_rate": 1.8816496544650277e-05, + "loss": 0.9855, + "step": 14757 + }, + { + "epoch": 1.09, + "learning_rate": 1.8816308308311922e-05, + "loss": 0.88, + "step": 14758 + }, + { + "epoch": 1.09, + "learning_rate": 1.881612005794695e-05, + "loss": 1.0604, + "step": 14759 + }, + { + "epoch": 1.09, + "learning_rate": 1.8815931793555653e-05, + "loss": 0.9809, + "step": 14760 + }, + { + "epoch": 1.09, + "learning_rate": 1.8815743515138335e-05, + "loss": 1.0299, + "step": 14761 + }, + { + "epoch": 1.09, + "learning_rate": 1.8815555222695296e-05, + "loss": 1.0481, + "step": 14762 + }, + { + "epoch": 1.09, + "learning_rate": 1.8815366916226835e-05, + "loss": 1.0355, + "step": 14763 + }, + { + "epoch": 1.09, + "learning_rate": 1.8815178595733253e-05, + "loss": 1.0681, + "step": 14764 + }, + { + "epoch": 1.09, + "learning_rate": 1.8814990261214845e-05, + "loss": 1.0782, + "step": 14765 + }, + { + "epoch": 1.09, + "learning_rate": 1.8814801912671913e-05, + "loss": 0.9824, + "step": 14766 + }, + { + "epoch": 1.09, + "learning_rate": 1.881461355010476e-05, + "loss": 1.013, + "step": 14767 + }, + { + "epoch": 1.09, + "learning_rate": 1.8814425173513685e-05, + "loss": 1.0564, + "step": 14768 + }, + { + "epoch": 1.09, + "learning_rate": 1.881423678289898e-05, + "loss": 1.0601, + "step": 14769 + }, + { + "epoch": 1.09, + "learning_rate": 1.8814048378260954e-05, + "loss": 1.0674, + "step": 14770 + }, + { + "epoch": 1.09, + "learning_rate": 1.8813859959599903e-05, + "loss": 1.0256, + "step": 14771 + }, + { + "epoch": 1.09, + "learning_rate": 1.8813671526916123e-05, + "loss": 1.1099, + "step": 14772 + }, + { + "epoch": 1.09, + "learning_rate": 1.8813483080209926e-05, + "loss": 1.0015, + "step": 14773 + }, + { + "epoch": 1.09, + "learning_rate": 1.8813294619481593e-05, + "loss": 1.1299, + "step": 14774 + }, + { + "epoch": 1.09, + "learning_rate": 1.881310614473144e-05, + "loss": 0.9416, + "step": 14775 + }, + { + "epoch": 1.09, + "learning_rate": 1.8812917655959764e-05, + "loss": 0.9884, + "step": 14776 + }, + { + "epoch": 1.09, + "learning_rate": 1.8812729153166857e-05, + "loss": 0.9915, + "step": 14777 + }, + { + "epoch": 1.09, + "learning_rate": 1.8812540636353027e-05, + "loss": 0.9215, + "step": 14778 + }, + { + "epoch": 1.09, + "learning_rate": 1.881235210551857e-05, + "loss": 1.0109, + "step": 14779 + }, + { + "epoch": 1.09, + "learning_rate": 1.8812163560663788e-05, + "loss": 0.9282, + "step": 14780 + }, + { + "epoch": 1.09, + "learning_rate": 1.8811975001788978e-05, + "loss": 0.8943, + "step": 14781 + }, + { + "epoch": 1.09, + "learning_rate": 1.8811786428894442e-05, + "loss": 1.1085, + "step": 14782 + }, + { + "epoch": 1.09, + "learning_rate": 1.8811597841980484e-05, + "loss": 0.9917, + "step": 14783 + }, + { + "epoch": 1.09, + "learning_rate": 1.8811409241047395e-05, + "loss": 0.9519, + "step": 14784 + }, + { + "epoch": 1.09, + "learning_rate": 1.8811220626095477e-05, + "loss": 1.0396, + "step": 14785 + }, + { + "epoch": 1.09, + "learning_rate": 1.881103199712504e-05, + "loss": 0.9872, + "step": 14786 + }, + { + "epoch": 1.09, + "learning_rate": 1.8810843354136375e-05, + "loss": 0.9596, + "step": 14787 + }, + { + "epoch": 1.09, + "learning_rate": 1.8810654697129783e-05, + "loss": 0.9628, + "step": 14788 + }, + { + "epoch": 1.09, + "learning_rate": 1.8810466026105568e-05, + "loss": 1.0763, + "step": 14789 + }, + { + "epoch": 1.09, + "learning_rate": 1.8810277341064026e-05, + "loss": 0.9821, + "step": 14790 + }, + { + "epoch": 1.09, + "learning_rate": 1.8810088642005456e-05, + "loss": 0.9727, + "step": 14791 + }, + { + "epoch": 1.09, + "learning_rate": 1.8809899928930166e-05, + "loss": 0.9934, + "step": 14792 + }, + { + "epoch": 1.09, + "learning_rate": 1.8809711201838448e-05, + "loss": 0.99, + "step": 14793 + }, + { + "epoch": 1.09, + "learning_rate": 1.8809522460730606e-05, + "loss": 1.111, + "step": 14794 + }, + { + "epoch": 1.09, + "learning_rate": 1.880933370560694e-05, + "loss": 0.9172, + "step": 14795 + }, + { + "epoch": 1.09, + "learning_rate": 1.880914493646775e-05, + "loss": 1.0434, + "step": 14796 + }, + { + "epoch": 1.09, + "learning_rate": 1.8808956153313338e-05, + "loss": 0.9814, + "step": 14797 + }, + { + "epoch": 1.09, + "learning_rate": 1.8808767356144e-05, + "loss": 1.0451, + "step": 14798 + }, + { + "epoch": 1.09, + "learning_rate": 1.880857854496004e-05, + "loss": 1.0249, + "step": 14799 + }, + { + "epoch": 1.09, + "learning_rate": 1.880838971976176e-05, + "loss": 0.8848, + "step": 14800 + }, + { + "epoch": 1.09, + "learning_rate": 1.8808200880549455e-05, + "loss": 1.0807, + "step": 14801 + }, + { + "epoch": 1.09, + "learning_rate": 1.880801202732343e-05, + "loss": 0.9794, + "step": 14802 + }, + { + "epoch": 1.09, + "learning_rate": 1.8807823160083985e-05, + "loss": 1.085, + "step": 14803 + }, + { + "epoch": 1.09, + "learning_rate": 1.880763427883142e-05, + "loss": 1.0723, + "step": 14804 + }, + { + "epoch": 1.09, + "learning_rate": 1.880744538356603e-05, + "loss": 1.0299, + "step": 14805 + }, + { + "epoch": 1.09, + "learning_rate": 1.8807256474288125e-05, + "loss": 0.9709, + "step": 14806 + }, + { + "epoch": 1.09, + "learning_rate": 1.8807067550997996e-05, + "loss": 0.9876, + "step": 14807 + }, + { + "epoch": 1.09, + "learning_rate": 1.8806878613695954e-05, + "loss": 1.0223, + "step": 14808 + }, + { + "epoch": 1.09, + "learning_rate": 1.8806689662382293e-05, + "loss": 1.0119, + "step": 14809 + }, + { + "epoch": 1.09, + "learning_rate": 1.8806500697057314e-05, + "loss": 1.0581, + "step": 14810 + }, + { + "epoch": 1.09, + "learning_rate": 1.8806311717721316e-05, + "loss": 0.9885, + "step": 14811 + }, + { + "epoch": 1.09, + "learning_rate": 1.8806122724374604e-05, + "loss": 1.0201, + "step": 14812 + }, + { + "epoch": 1.09, + "learning_rate": 1.880593371701748e-05, + "loss": 1.0707, + "step": 14813 + }, + { + "epoch": 1.09, + "learning_rate": 1.8805744695650235e-05, + "loss": 0.9894, + "step": 14814 + }, + { + "epoch": 1.09, + "learning_rate": 1.8805555660273182e-05, + "loss": 1.0593, + "step": 14815 + }, + { + "epoch": 1.09, + "learning_rate": 1.8805366610886614e-05, + "loss": 1.0224, + "step": 14816 + }, + { + "epoch": 1.09, + "learning_rate": 1.8805177547490833e-05, + "loss": 0.9713, + "step": 14817 + }, + { + "epoch": 1.09, + "learning_rate": 1.880498847008614e-05, + "loss": 0.9542, + "step": 14818 + }, + { + "epoch": 1.09, + "learning_rate": 1.8804799378672836e-05, + "loss": 1.0644, + "step": 14819 + }, + { + "epoch": 1.1, + "learning_rate": 1.8804610273251223e-05, + "loss": 1.0762, + "step": 14820 + }, + { + "epoch": 1.1, + "learning_rate": 1.88044211538216e-05, + "loss": 1.0389, + "step": 14821 + }, + { + "epoch": 1.1, + "learning_rate": 1.880423202038427e-05, + "loss": 1.0323, + "step": 14822 + }, + { + "epoch": 1.1, + "learning_rate": 1.8804042872939534e-05, + "loss": 0.9897, + "step": 14823 + }, + { + "epoch": 1.1, + "learning_rate": 1.8803853711487692e-05, + "loss": 1.0684, + "step": 14824 + }, + { + "epoch": 1.1, + "learning_rate": 1.880366453602904e-05, + "loss": 0.849, + "step": 14825 + }, + { + "epoch": 1.1, + "learning_rate": 1.880347534656389e-05, + "loss": 0.9706, + "step": 14826 + }, + { + "epoch": 1.1, + "learning_rate": 1.8803286143092534e-05, + "loss": 0.9907, + "step": 14827 + }, + { + "epoch": 1.1, + "learning_rate": 1.8803096925615276e-05, + "loss": 1.009, + "step": 14828 + }, + { + "epoch": 1.1, + "learning_rate": 1.8802907694132416e-05, + "loss": 0.9615, + "step": 14829 + }, + { + "epoch": 1.1, + "learning_rate": 1.8802718448644256e-05, + "loss": 1.0704, + "step": 14830 + }, + { + "epoch": 1.1, + "learning_rate": 1.8802529189151094e-05, + "loss": 1.0564, + "step": 14831 + }, + { + "epoch": 1.1, + "learning_rate": 1.880233991565324e-05, + "loss": 1.0355, + "step": 14832 + }, + { + "epoch": 1.1, + "learning_rate": 1.8802150628150985e-05, + "loss": 1.0925, + "step": 14833 + }, + { + "epoch": 1.1, + "learning_rate": 1.8801961326644633e-05, + "loss": 0.9373, + "step": 14834 + }, + { + "epoch": 1.1, + "learning_rate": 1.880177201113449e-05, + "loss": 0.9845, + "step": 14835 + }, + { + "epoch": 1.1, + "learning_rate": 1.8801582681620854e-05, + "loss": 1.0337, + "step": 14836 + }, + { + "epoch": 1.1, + "learning_rate": 1.8801393338104023e-05, + "loss": 0.9744, + "step": 14837 + }, + { + "epoch": 1.1, + "learning_rate": 1.88012039805843e-05, + "loss": 1.0722, + "step": 14838 + }, + { + "epoch": 1.1, + "learning_rate": 1.8801014609061995e-05, + "loss": 1.0023, + "step": 14839 + }, + { + "epoch": 1.1, + "learning_rate": 1.8800825223537392e-05, + "loss": 1.0729, + "step": 14840 + }, + { + "epoch": 1.1, + "learning_rate": 1.8800635824010805e-05, + "loss": 1.0593, + "step": 14841 + }, + { + "epoch": 1.1, + "learning_rate": 1.8800446410482537e-05, + "loss": 1.0433, + "step": 14842 + }, + { + "epoch": 1.1, + "learning_rate": 1.8800256982952882e-05, + "loss": 0.9274, + "step": 14843 + }, + { + "epoch": 1.1, + "learning_rate": 1.8800067541422145e-05, + "loss": 0.9567, + "step": 14844 + }, + { + "epoch": 1.1, + "learning_rate": 1.8799878085890624e-05, + "loss": 1.1391, + "step": 14845 + }, + { + "epoch": 1.1, + "learning_rate": 1.879968861635862e-05, + "loss": 0.926, + "step": 14846 + }, + { + "epoch": 1.1, + "learning_rate": 1.8799499132826444e-05, + "loss": 1.0622, + "step": 14847 + }, + { + "epoch": 1.1, + "learning_rate": 1.8799309635294388e-05, + "loss": 1.0409, + "step": 14848 + }, + { + "epoch": 1.1, + "learning_rate": 1.8799120123762758e-05, + "loss": 0.9421, + "step": 14849 + }, + { + "epoch": 1.1, + "learning_rate": 1.879893059823185e-05, + "loss": 0.9776, + "step": 14850 + }, + { + "epoch": 1.1, + "learning_rate": 1.8798741058701974e-05, + "loss": 1.0238, + "step": 14851 + }, + { + "epoch": 1.1, + "learning_rate": 1.8798551505173424e-05, + "loss": 1.0528, + "step": 14852 + }, + { + "epoch": 1.1, + "learning_rate": 1.8798361937646502e-05, + "loss": 1.0047, + "step": 14853 + }, + { + "epoch": 1.1, + "learning_rate": 1.8798172356121514e-05, + "loss": 1.0143, + "step": 14854 + }, + { + "epoch": 1.1, + "learning_rate": 1.879798276059876e-05, + "loss": 1.0756, + "step": 14855 + }, + { + "epoch": 1.1, + "learning_rate": 1.8797793151078544e-05, + "loss": 1.0339, + "step": 14856 + }, + { + "epoch": 1.1, + "learning_rate": 1.8797603527561162e-05, + "loss": 1.0116, + "step": 14857 + }, + { + "epoch": 1.1, + "learning_rate": 1.879741389004692e-05, + "loss": 0.9994, + "step": 14858 + }, + { + "epoch": 1.1, + "learning_rate": 1.8797224238536115e-05, + "loss": 1.0853, + "step": 14859 + }, + { + "epoch": 1.1, + "learning_rate": 1.8797034573029057e-05, + "loss": 0.9368, + "step": 14860 + }, + { + "epoch": 1.1, + "learning_rate": 1.879684489352604e-05, + "loss": 1.171, + "step": 14861 + }, + { + "epoch": 1.1, + "learning_rate": 1.879665520002737e-05, + "loss": 1.0591, + "step": 14862 + }, + { + "epoch": 1.1, + "learning_rate": 1.8796465492533348e-05, + "loss": 1.084, + "step": 14863 + }, + { + "epoch": 1.1, + "learning_rate": 1.8796275771044273e-05, + "loss": 0.9459, + "step": 14864 + }, + { + "epoch": 1.1, + "learning_rate": 1.8796086035560452e-05, + "loss": 0.8651, + "step": 14865 + }, + { + "epoch": 1.1, + "learning_rate": 1.879589628608218e-05, + "loss": 0.9254, + "step": 14866 + }, + { + "epoch": 1.1, + "learning_rate": 1.8795706522609764e-05, + "loss": 1.0115, + "step": 14867 + }, + { + "epoch": 1.1, + "learning_rate": 1.8795516745143503e-05, + "loss": 1.0157, + "step": 14868 + }, + { + "epoch": 1.1, + "learning_rate": 1.8795326953683706e-05, + "loss": 0.9584, + "step": 14869 + }, + { + "epoch": 1.1, + "learning_rate": 1.8795137148230668e-05, + "loss": 0.95, + "step": 14870 + }, + { + "epoch": 1.1, + "learning_rate": 1.879494732878469e-05, + "loss": 1.0581, + "step": 14871 + }, + { + "epoch": 1.1, + "learning_rate": 1.8794757495346078e-05, + "loss": 0.9601, + "step": 14872 + }, + { + "epoch": 1.1, + "learning_rate": 1.8794567647915133e-05, + "loss": 0.9514, + "step": 14873 + }, + { + "epoch": 1.1, + "learning_rate": 1.8794377786492156e-05, + "loss": 0.9349, + "step": 14874 + }, + { + "epoch": 1.1, + "learning_rate": 1.8794187911077452e-05, + "loss": 1.0679, + "step": 14875 + }, + { + "epoch": 1.1, + "learning_rate": 1.879399802167132e-05, + "loss": 1.0576, + "step": 14876 + }, + { + "epoch": 1.1, + "learning_rate": 1.879380811827406e-05, + "loss": 1.0888, + "step": 14877 + }, + { + "epoch": 1.1, + "learning_rate": 1.8793618200885978e-05, + "loss": 0.978, + "step": 14878 + }, + { + "epoch": 1.1, + "learning_rate": 1.879342826950738e-05, + "loss": 1.079, + "step": 14879 + }, + { + "epoch": 1.1, + "learning_rate": 1.879323832413856e-05, + "loss": 1.0182, + "step": 14880 + }, + { + "epoch": 1.1, + "learning_rate": 1.8793048364779823e-05, + "loss": 1.0402, + "step": 14881 + }, + { + "epoch": 1.1, + "learning_rate": 1.879285839143147e-05, + "loss": 0.9404, + "step": 14882 + }, + { + "epoch": 1.1, + "learning_rate": 1.8792668404093808e-05, + "loss": 0.9596, + "step": 14883 + }, + { + "epoch": 1.1, + "learning_rate": 1.8792478402767134e-05, + "loss": 1.0859, + "step": 14884 + }, + { + "epoch": 1.1, + "learning_rate": 1.879228838745176e-05, + "loss": 0.9582, + "step": 14885 + }, + { + "epoch": 1.1, + "learning_rate": 1.8792098358147973e-05, + "loss": 1.0452, + "step": 14886 + }, + { + "epoch": 1.1, + "learning_rate": 1.8791908314856086e-05, + "loss": 1.0611, + "step": 14887 + }, + { + "epoch": 1.1, + "learning_rate": 1.8791718257576398e-05, + "loss": 1.0533, + "step": 14888 + }, + { + "epoch": 1.1, + "learning_rate": 1.8791528186309214e-05, + "loss": 0.99, + "step": 14889 + }, + { + "epoch": 1.1, + "learning_rate": 1.8791338101054834e-05, + "loss": 1.1187, + "step": 14890 + }, + { + "epoch": 1.1, + "learning_rate": 1.879114800181356e-05, + "loss": 0.9918, + "step": 14891 + }, + { + "epoch": 1.1, + "learning_rate": 1.8790957888585693e-05, + "loss": 0.9841, + "step": 14892 + }, + { + "epoch": 1.1, + "learning_rate": 1.8790767761371543e-05, + "loss": 0.99, + "step": 14893 + }, + { + "epoch": 1.1, + "learning_rate": 1.87905776201714e-05, + "loss": 0.987, + "step": 14894 + }, + { + "epoch": 1.1, + "learning_rate": 1.879038746498558e-05, + "loss": 0.996, + "step": 14895 + }, + { + "epoch": 1.1, + "learning_rate": 1.8790197295814375e-05, + "loss": 1.089, + "step": 14896 + }, + { + "epoch": 1.1, + "learning_rate": 1.8790007112658096e-05, + "loss": 1.0552, + "step": 14897 + }, + { + "epoch": 1.1, + "learning_rate": 1.878981691551704e-05, + "loss": 0.9782, + "step": 14898 + }, + { + "epoch": 1.1, + "learning_rate": 1.878962670439151e-05, + "loss": 0.9156, + "step": 14899 + }, + { + "epoch": 1.1, + "learning_rate": 1.878943647928181e-05, + "loss": 0.961, + "step": 14900 + }, + { + "epoch": 1.1, + "learning_rate": 1.8789246240188243e-05, + "loss": 1.0417, + "step": 14901 + }, + { + "epoch": 1.1, + "learning_rate": 1.8789055987111112e-05, + "loss": 0.9568, + "step": 14902 + }, + { + "epoch": 1.1, + "learning_rate": 1.8788865720050718e-05, + "loss": 1.1289, + "step": 14903 + }, + { + "epoch": 1.1, + "learning_rate": 1.878867543900736e-05, + "loss": 1.0374, + "step": 14904 + }, + { + "epoch": 1.1, + "learning_rate": 1.878848514398135e-05, + "loss": 0.9449, + "step": 14905 + }, + { + "epoch": 1.1, + "learning_rate": 1.8788294834972985e-05, + "loss": 0.9769, + "step": 14906 + }, + { + "epoch": 1.1, + "learning_rate": 1.878810451198257e-05, + "loss": 1.0153, + "step": 14907 + }, + { + "epoch": 1.1, + "learning_rate": 1.8787914175010405e-05, + "loss": 1.0982, + "step": 14908 + }, + { + "epoch": 1.1, + "learning_rate": 1.8787723824056795e-05, + "loss": 1.0099, + "step": 14909 + }, + { + "epoch": 1.1, + "learning_rate": 1.8787533459122042e-05, + "loss": 0.9696, + "step": 14910 + }, + { + "epoch": 1.1, + "learning_rate": 1.8787343080206447e-05, + "loss": 1.0208, + "step": 14911 + }, + { + "epoch": 1.1, + "learning_rate": 1.8787152687310313e-05, + "loss": 1.1117, + "step": 14912 + }, + { + "epoch": 1.1, + "learning_rate": 1.878696228043395e-05, + "loss": 1.025, + "step": 14913 + }, + { + "epoch": 1.1, + "learning_rate": 1.8786771859577656e-05, + "loss": 0.9584, + "step": 14914 + }, + { + "epoch": 1.1, + "learning_rate": 1.878658142474173e-05, + "loss": 0.9515, + "step": 14915 + }, + { + "epoch": 1.1, + "learning_rate": 1.8786390975926482e-05, + "loss": 0.8902, + "step": 14916 + }, + { + "epoch": 1.1, + "learning_rate": 1.8786200513132207e-05, + "loss": 1.0005, + "step": 14917 + }, + { + "epoch": 1.1, + "learning_rate": 1.8786010036359218e-05, + "loss": 0.9914, + "step": 14918 + }, + { + "epoch": 1.1, + "learning_rate": 1.8785819545607807e-05, + "loss": 0.8801, + "step": 14919 + }, + { + "epoch": 1.1, + "learning_rate": 1.878562904087829e-05, + "loss": 1.0709, + "step": 14920 + }, + { + "epoch": 1.1, + "learning_rate": 1.878543852217096e-05, + "loss": 0.9572, + "step": 14921 + }, + { + "epoch": 1.1, + "learning_rate": 1.878524798948612e-05, + "loss": 0.9036, + "step": 14922 + }, + { + "epoch": 1.1, + "learning_rate": 1.8785057442824074e-05, + "loss": 1.1142, + "step": 14923 + }, + { + "epoch": 1.1, + "learning_rate": 1.8784866882185134e-05, + "loss": 1.0493, + "step": 14924 + }, + { + "epoch": 1.1, + "learning_rate": 1.8784676307569594e-05, + "loss": 0.9922, + "step": 14925 + }, + { + "epoch": 1.1, + "learning_rate": 1.878448571897776e-05, + "loss": 1.0561, + "step": 14926 + }, + { + "epoch": 1.1, + "learning_rate": 1.8784295116409936e-05, + "loss": 0.9853, + "step": 14927 + }, + { + "epoch": 1.1, + "learning_rate": 1.878410449986642e-05, + "loss": 1.001, + "step": 14928 + }, + { + "epoch": 1.1, + "learning_rate": 1.8783913869347522e-05, + "loss": 1.0303, + "step": 14929 + }, + { + "epoch": 1.1, + "learning_rate": 1.8783723224853543e-05, + "loss": 0.9414, + "step": 14930 + }, + { + "epoch": 1.1, + "learning_rate": 1.8783532566384787e-05, + "loss": 1.0523, + "step": 14931 + }, + { + "epoch": 1.1, + "learning_rate": 1.8783341893941553e-05, + "loss": 0.9122, + "step": 14932 + }, + { + "epoch": 1.1, + "learning_rate": 1.8783151207524152e-05, + "loss": 1.0174, + "step": 14933 + }, + { + "epoch": 1.1, + "learning_rate": 1.878296050713288e-05, + "loss": 1.0096, + "step": 14934 + }, + { + "epoch": 1.1, + "learning_rate": 1.8782769792768043e-05, + "loss": 0.9523, + "step": 14935 + }, + { + "epoch": 1.1, + "learning_rate": 1.8782579064429945e-05, + "loss": 0.9462, + "step": 14936 + }, + { + "epoch": 1.1, + "learning_rate": 1.8782388322118894e-05, + "loss": 1.0171, + "step": 14937 + }, + { + "epoch": 1.1, + "learning_rate": 1.8782197565835183e-05, + "loss": 0.9992, + "step": 14938 + }, + { + "epoch": 1.1, + "learning_rate": 1.8782006795579125e-05, + "loss": 0.9417, + "step": 14939 + }, + { + "epoch": 1.1, + "learning_rate": 1.8781816011351018e-05, + "loss": 1.0843, + "step": 14940 + }, + { + "epoch": 1.1, + "learning_rate": 1.878162521315117e-05, + "loss": 1.1929, + "step": 14941 + }, + { + "epoch": 1.1, + "learning_rate": 1.878143440097988e-05, + "loss": 1.0812, + "step": 14942 + }, + { + "epoch": 1.1, + "learning_rate": 1.8781243574837453e-05, + "loss": 1.0391, + "step": 14943 + }, + { + "epoch": 1.1, + "learning_rate": 1.8781052734724192e-05, + "loss": 1.0538, + "step": 14944 + }, + { + "epoch": 1.1, + "learning_rate": 1.87808618806404e-05, + "loss": 1.0354, + "step": 14945 + }, + { + "epoch": 1.1, + "learning_rate": 1.878067101258639e-05, + "loss": 1.0564, + "step": 14946 + }, + { + "epoch": 1.1, + "learning_rate": 1.8780480130562454e-05, + "loss": 1.1325, + "step": 14947 + }, + { + "epoch": 1.1, + "learning_rate": 1.87802892345689e-05, + "loss": 1.0695, + "step": 14948 + }, + { + "epoch": 1.1, + "learning_rate": 1.878009832460603e-05, + "loss": 1.1082, + "step": 14949 + }, + { + "epoch": 1.1, + "learning_rate": 1.8779907400674152e-05, + "loss": 0.9583, + "step": 14950 + }, + { + "epoch": 1.1, + "learning_rate": 1.8779716462773564e-05, + "loss": 1.0212, + "step": 14951 + }, + { + "epoch": 1.1, + "learning_rate": 1.8779525510904572e-05, + "loss": 0.9253, + "step": 14952 + }, + { + "epoch": 1.1, + "learning_rate": 1.877933454506748e-05, + "loss": 0.96, + "step": 14953 + }, + { + "epoch": 1.1, + "learning_rate": 1.8779143565262598e-05, + "loss": 0.9157, + "step": 14954 + }, + { + "epoch": 1.11, + "learning_rate": 1.8778952571490222e-05, + "loss": 1.0234, + "step": 14955 + }, + { + "epoch": 1.11, + "learning_rate": 1.8778761563750656e-05, + "loss": 0.9719, + "step": 14956 + }, + { + "epoch": 1.11, + "learning_rate": 1.8778570542044208e-05, + "loss": 1.1099, + "step": 14957 + }, + { + "epoch": 1.11, + "learning_rate": 1.8778379506371175e-05, + "loss": 1.0008, + "step": 14958 + }, + { + "epoch": 1.11, + "learning_rate": 1.877818845673187e-05, + "loss": 1.0922, + "step": 14959 + }, + { + "epoch": 1.11, + "learning_rate": 1.8777997393126592e-05, + "loss": 1.1279, + "step": 14960 + }, + { + "epoch": 1.11, + "learning_rate": 1.8777806315555648e-05, + "loss": 0.988, + "step": 14961 + }, + { + "epoch": 1.11, + "learning_rate": 1.877761522401934e-05, + "loss": 1.0545, + "step": 14962 + }, + { + "epoch": 1.11, + "learning_rate": 1.8777424118517968e-05, + "loss": 0.9901, + "step": 14963 + }, + { + "epoch": 1.11, + "learning_rate": 1.877723299905184e-05, + "loss": 0.9733, + "step": 14964 + }, + { + "epoch": 1.11, + "learning_rate": 1.877704186562126e-05, + "loss": 1.1232, + "step": 14965 + }, + { + "epoch": 1.11, + "learning_rate": 1.8776850718226537e-05, + "loss": 0.9725, + "step": 14966 + }, + { + "epoch": 1.11, + "learning_rate": 1.8776659556867964e-05, + "loss": 0.9695, + "step": 14967 + }, + { + "epoch": 1.11, + "learning_rate": 1.8776468381545854e-05, + "loss": 0.978, + "step": 14968 + }, + { + "epoch": 1.11, + "learning_rate": 1.877627719226051e-05, + "loss": 1.0159, + "step": 14969 + }, + { + "epoch": 1.11, + "learning_rate": 1.877608598901223e-05, + "loss": 1.004, + "step": 14970 + }, + { + "epoch": 1.11, + "learning_rate": 1.8775894771801325e-05, + "loss": 1.0794, + "step": 14971 + }, + { + "epoch": 1.11, + "learning_rate": 1.8775703540628098e-05, + "loss": 1.1134, + "step": 14972 + }, + { + "epoch": 1.11, + "learning_rate": 1.877551229549285e-05, + "loss": 1.0794, + "step": 14973 + }, + { + "epoch": 1.11, + "learning_rate": 1.877532103639589e-05, + "loss": 1.0812, + "step": 14974 + }, + { + "epoch": 1.11, + "learning_rate": 1.8775129763337522e-05, + "loss": 1.0236, + "step": 14975 + }, + { + "epoch": 1.11, + "learning_rate": 1.8774938476318044e-05, + "loss": 1.045, + "step": 14976 + }, + { + "epoch": 1.11, + "learning_rate": 1.8774747175337765e-05, + "loss": 1.0404, + "step": 14977 + }, + { + "epoch": 1.11, + "learning_rate": 1.8774555860396984e-05, + "loss": 1.0951, + "step": 14978 + }, + { + "epoch": 1.11, + "learning_rate": 1.877436453149602e-05, + "loss": 0.9833, + "step": 14979 + }, + { + "epoch": 1.11, + "learning_rate": 1.8774173188635162e-05, + "loss": 0.9762, + "step": 14980 + }, + { + "epoch": 1.11, + "learning_rate": 1.877398183181472e-05, + "loss": 1.1011, + "step": 14981 + }, + { + "epoch": 1.11, + "learning_rate": 1.8773790461034997e-05, + "loss": 1.0838, + "step": 14982 + }, + { + "epoch": 1.11, + "learning_rate": 1.8773599076296306e-05, + "loss": 1.0428, + "step": 14983 + }, + { + "epoch": 1.11, + "learning_rate": 1.877340767759894e-05, + "loss": 1.1173, + "step": 14984 + }, + { + "epoch": 1.11, + "learning_rate": 1.8773216264943204e-05, + "loss": 1.1299, + "step": 14985 + }, + { + "epoch": 1.11, + "learning_rate": 1.877302483832941e-05, + "loss": 1.0749, + "step": 14986 + }, + { + "epoch": 1.11, + "learning_rate": 1.877283339775786e-05, + "loss": 1.0439, + "step": 14987 + }, + { + "epoch": 1.11, + "learning_rate": 1.8772641943228855e-05, + "loss": 1.0451, + "step": 14988 + }, + { + "epoch": 1.11, + "learning_rate": 1.8772450474742702e-05, + "loss": 1.04, + "step": 14989 + }, + { + "epoch": 1.11, + "learning_rate": 1.877225899229971e-05, + "loss": 1.1026, + "step": 14990 + }, + { + "epoch": 1.11, + "learning_rate": 1.8772067495900178e-05, + "loss": 1.0612, + "step": 14991 + }, + { + "epoch": 1.11, + "learning_rate": 1.877187598554441e-05, + "loss": 1.0393, + "step": 14992 + }, + { + "epoch": 1.11, + "learning_rate": 1.8771684461232714e-05, + "loss": 0.9186, + "step": 14993 + }, + { + "epoch": 1.11, + "learning_rate": 1.877149292296539e-05, + "loss": 0.9637, + "step": 14994 + }, + { + "epoch": 1.11, + "learning_rate": 1.877130137074275e-05, + "loss": 1.0884, + "step": 14995 + }, + { + "epoch": 1.11, + "learning_rate": 1.8771109804565096e-05, + "loss": 1.0773, + "step": 14996 + }, + { + "epoch": 1.11, + "learning_rate": 1.8770918224432732e-05, + "loss": 1.0012, + "step": 14997 + }, + { + "epoch": 1.11, + "learning_rate": 1.877072663034596e-05, + "loss": 0.9598, + "step": 14998 + }, + { + "epoch": 1.11, + "learning_rate": 1.877053502230509e-05, + "loss": 1.0804, + "step": 14999 + }, + { + "epoch": 1.11, + "learning_rate": 1.8770343400310424e-05, + "loss": 1.0244, + "step": 15000 + }, + { + "epoch": 1.11, + "learning_rate": 1.8770151764362262e-05, + "loss": 1.0294, + "step": 15001 + }, + { + "epoch": 1.11, + "learning_rate": 1.876996011446092e-05, + "loss": 1.1224, + "step": 15002 + }, + { + "epoch": 1.11, + "learning_rate": 1.8769768450606695e-05, + "loss": 1.0922, + "step": 15003 + }, + { + "epoch": 1.11, + "learning_rate": 1.8769576772799893e-05, + "loss": 0.9819, + "step": 15004 + }, + { + "epoch": 1.11, + "learning_rate": 1.8769385081040822e-05, + "loss": 1.1042, + "step": 15005 + }, + { + "epoch": 1.11, + "learning_rate": 1.8769193375329783e-05, + "loss": 1.023, + "step": 15006 + }, + { + "epoch": 1.11, + "learning_rate": 1.876900165566708e-05, + "loss": 0.9955, + "step": 15007 + }, + { + "epoch": 1.11, + "learning_rate": 1.8768809922053023e-05, + "loss": 0.9957, + "step": 15008 + }, + { + "epoch": 1.11, + "learning_rate": 1.8768618174487917e-05, + "loss": 1.0153, + "step": 15009 + }, + { + "epoch": 1.11, + "learning_rate": 1.8768426412972065e-05, + "loss": 0.9876, + "step": 15010 + }, + { + "epoch": 1.11, + "learning_rate": 1.8768234637505772e-05, + "loss": 1.0615, + "step": 15011 + }, + { + "epoch": 1.11, + "learning_rate": 1.8768042848089342e-05, + "loss": 0.9497, + "step": 15012 + }, + { + "epoch": 1.11, + "learning_rate": 1.876785104472308e-05, + "loss": 1.1378, + "step": 15013 + }, + { + "epoch": 1.11, + "learning_rate": 1.8767659227407294e-05, + "loss": 1.0012, + "step": 15014 + }, + { + "epoch": 1.11, + "learning_rate": 1.8767467396142286e-05, + "loss": 1.0936, + "step": 15015 + }, + { + "epoch": 1.11, + "learning_rate": 1.8767275550928366e-05, + "loss": 0.9765, + "step": 15016 + }, + { + "epoch": 1.11, + "learning_rate": 1.8767083691765833e-05, + "loss": 0.9948, + "step": 15017 + }, + { + "epoch": 1.11, + "learning_rate": 1.8766891818654996e-05, + "loss": 0.9211, + "step": 15018 + }, + { + "epoch": 1.11, + "learning_rate": 1.876669993159616e-05, + "loss": 1.0202, + "step": 15019 + }, + { + "epoch": 1.11, + "learning_rate": 1.8766508030589632e-05, + "loss": 0.9809, + "step": 15020 + }, + { + "epoch": 1.11, + "learning_rate": 1.8766316115635712e-05, + "loss": 1.0931, + "step": 15021 + }, + { + "epoch": 1.11, + "learning_rate": 1.876612418673471e-05, + "loss": 1.0745, + "step": 15022 + }, + { + "epoch": 1.11, + "learning_rate": 1.876593224388693e-05, + "loss": 0.9935, + "step": 15023 + }, + { + "epoch": 1.11, + "learning_rate": 1.8765740287092675e-05, + "loss": 0.9957, + "step": 15024 + }, + { + "epoch": 1.11, + "learning_rate": 1.8765548316352254e-05, + "loss": 0.965, + "step": 15025 + }, + { + "epoch": 1.11, + "learning_rate": 1.876535633166597e-05, + "loss": 1.0152, + "step": 15026 + }, + { + "epoch": 1.11, + "learning_rate": 1.876516433303413e-05, + "loss": 1.0621, + "step": 15027 + }, + { + "epoch": 1.11, + "learning_rate": 1.876497232045704e-05, + "loss": 1.0915, + "step": 15028 + }, + { + "epoch": 1.11, + "learning_rate": 1.8764780293935004e-05, + "loss": 0.9704, + "step": 15029 + }, + { + "epoch": 1.11, + "learning_rate": 1.8764588253468328e-05, + "loss": 1.072, + "step": 15030 + }, + { + "epoch": 1.11, + "learning_rate": 1.876439619905732e-05, + "loss": 0.955, + "step": 15031 + }, + { + "epoch": 1.11, + "learning_rate": 1.876420413070228e-05, + "loss": 0.9255, + "step": 15032 + }, + { + "epoch": 1.11, + "learning_rate": 1.8764012048403516e-05, + "loss": 1.0114, + "step": 15033 + }, + { + "epoch": 1.11, + "learning_rate": 1.8763819952161334e-05, + "loss": 1.1387, + "step": 15034 + }, + { + "epoch": 1.11, + "learning_rate": 1.8763627841976044e-05, + "loss": 1.1035, + "step": 15035 + }, + { + "epoch": 1.11, + "learning_rate": 1.8763435717847944e-05, + "loss": 1.0304, + "step": 15036 + }, + { + "epoch": 1.11, + "learning_rate": 1.8763243579777343e-05, + "loss": 1.0802, + "step": 15037 + }, + { + "epoch": 1.11, + "learning_rate": 1.8763051427764546e-05, + "loss": 1.0094, + "step": 15038 + }, + { + "epoch": 1.11, + "learning_rate": 1.8762859261809865e-05, + "loss": 1.0438, + "step": 15039 + }, + { + "epoch": 1.11, + "learning_rate": 1.8762667081913594e-05, + "loss": 0.9576, + "step": 15040 + }, + { + "epoch": 1.11, + "learning_rate": 1.8762474888076048e-05, + "loss": 1.0832, + "step": 15041 + }, + { + "epoch": 1.11, + "learning_rate": 1.876228268029753e-05, + "loss": 1.0395, + "step": 15042 + }, + { + "epoch": 1.11, + "learning_rate": 1.8762090458578346e-05, + "loss": 1.0869, + "step": 15043 + }, + { + "epoch": 1.11, + "learning_rate": 1.87618982229188e-05, + "loss": 0.9804, + "step": 15044 + }, + { + "epoch": 1.11, + "learning_rate": 1.87617059733192e-05, + "loss": 1.0204, + "step": 15045 + }, + { + "epoch": 1.11, + "learning_rate": 1.876151370977985e-05, + "loss": 0.9939, + "step": 15046 + }, + { + "epoch": 1.11, + "learning_rate": 1.876132143230106e-05, + "loss": 1.0274, + "step": 15047 + }, + { + "epoch": 1.11, + "learning_rate": 1.8761129140883126e-05, + "loss": 0.9806, + "step": 15048 + }, + { + "epoch": 1.11, + "learning_rate": 1.8760936835526368e-05, + "loss": 0.9272, + "step": 15049 + }, + { + "epoch": 1.11, + "learning_rate": 1.8760744516231086e-05, + "loss": 1.1115, + "step": 15050 + }, + { + "epoch": 1.11, + "learning_rate": 1.876055218299758e-05, + "loss": 1.0052, + "step": 15051 + }, + { + "epoch": 1.11, + "learning_rate": 1.876035983582616e-05, + "loss": 1.034, + "step": 15052 + }, + { + "epoch": 1.11, + "learning_rate": 1.8760167474717135e-05, + "loss": 0.9551, + "step": 15053 + }, + { + "epoch": 1.11, + "learning_rate": 1.8759975099670812e-05, + "loss": 1.0535, + "step": 15054 + }, + { + "epoch": 1.11, + "learning_rate": 1.8759782710687487e-05, + "loss": 1.0538, + "step": 15055 + }, + { + "epoch": 1.11, + "learning_rate": 1.875959030776748e-05, + "loss": 1.0507, + "step": 15056 + }, + { + "epoch": 1.11, + "learning_rate": 1.8759397890911087e-05, + "loss": 1.0427, + "step": 15057 + }, + { + "epoch": 1.11, + "learning_rate": 1.875920546011862e-05, + "loss": 1.0405, + "step": 15058 + }, + { + "epoch": 1.11, + "learning_rate": 1.875901301539038e-05, + "loss": 1.1255, + "step": 15059 + }, + { + "epoch": 1.11, + "learning_rate": 1.8758820556726674e-05, + "loss": 0.9155, + "step": 15060 + }, + { + "epoch": 1.11, + "learning_rate": 1.875862808412781e-05, + "loss": 0.9742, + "step": 15061 + }, + { + "epoch": 1.11, + "learning_rate": 1.8758435597594096e-05, + "loss": 1.0309, + "step": 15062 + }, + { + "epoch": 1.11, + "learning_rate": 1.8758243097125833e-05, + "loss": 0.9359, + "step": 15063 + }, + { + "epoch": 1.11, + "learning_rate": 1.8758050582723335e-05, + "loss": 0.9412, + "step": 15064 + }, + { + "epoch": 1.11, + "learning_rate": 1.8757858054386902e-05, + "loss": 0.9772, + "step": 15065 + }, + { + "epoch": 1.11, + "learning_rate": 1.8757665512116844e-05, + "loss": 0.9466, + "step": 15066 + }, + { + "epoch": 1.11, + "learning_rate": 1.8757472955913464e-05, + "loss": 0.9988, + "step": 15067 + }, + { + "epoch": 1.11, + "learning_rate": 1.875728038577707e-05, + "loss": 1.0811, + "step": 15068 + }, + { + "epoch": 1.11, + "learning_rate": 1.8757087801707965e-05, + "loss": 0.928, + "step": 15069 + }, + { + "epoch": 1.11, + "learning_rate": 1.8756895203706463e-05, + "loss": 1.031, + "step": 15070 + }, + { + "epoch": 1.11, + "learning_rate": 1.8756702591772865e-05, + "loss": 1.0608, + "step": 15071 + }, + { + "epoch": 1.11, + "learning_rate": 1.8756509965907476e-05, + "loss": 1.0761, + "step": 15072 + }, + { + "epoch": 1.11, + "learning_rate": 1.8756317326110607e-05, + "loss": 1.0716, + "step": 15073 + }, + { + "epoch": 1.11, + "learning_rate": 1.875612467238256e-05, + "loss": 1.0188, + "step": 15074 + }, + { + "epoch": 1.11, + "learning_rate": 1.8755932004723647e-05, + "loss": 1.0579, + "step": 15075 + }, + { + "epoch": 1.11, + "learning_rate": 1.875573932313417e-05, + "loss": 1.0755, + "step": 15076 + }, + { + "epoch": 1.11, + "learning_rate": 1.875554662761444e-05, + "loss": 1.0401, + "step": 15077 + }, + { + "epoch": 1.11, + "learning_rate": 1.8755353918164756e-05, + "loss": 1.0291, + "step": 15078 + }, + { + "epoch": 1.11, + "learning_rate": 1.875516119478543e-05, + "loss": 1.0391, + "step": 15079 + }, + { + "epoch": 1.11, + "learning_rate": 1.875496845747677e-05, + "loss": 1.0647, + "step": 15080 + }, + { + "epoch": 1.11, + "learning_rate": 1.875477570623908e-05, + "loss": 1.0595, + "step": 15081 + }, + { + "epoch": 1.11, + "learning_rate": 1.875458294107267e-05, + "loss": 1.0955, + "step": 15082 + }, + { + "epoch": 1.11, + "learning_rate": 1.8754390161977837e-05, + "loss": 1.0034, + "step": 15083 + }, + { + "epoch": 1.11, + "learning_rate": 1.87541973689549e-05, + "loss": 1.0614, + "step": 15084 + }, + { + "epoch": 1.11, + "learning_rate": 1.8754004562004156e-05, + "loss": 0.9393, + "step": 15085 + }, + { + "epoch": 1.11, + "learning_rate": 1.875381174112592e-05, + "loss": 0.9858, + "step": 15086 + }, + { + "epoch": 1.11, + "learning_rate": 1.875361890632049e-05, + "loss": 1.0843, + "step": 15087 + }, + { + "epoch": 1.11, + "learning_rate": 1.875342605758818e-05, + "loss": 0.9426, + "step": 15088 + }, + { + "epoch": 1.11, + "learning_rate": 1.8753233194929297e-05, + "loss": 1.0991, + "step": 15089 + }, + { + "epoch": 1.11, + "learning_rate": 1.875304031834414e-05, + "loss": 0.9599, + "step": 15090 + }, + { + "epoch": 1.12, + "learning_rate": 1.8752847427833028e-05, + "loss": 0.9895, + "step": 15091 + }, + { + "epoch": 1.12, + "learning_rate": 1.8752654523396256e-05, + "loss": 1.0509, + "step": 15092 + }, + { + "epoch": 1.12, + "learning_rate": 1.8752461605034137e-05, + "loss": 1.1094, + "step": 15093 + }, + { + "epoch": 1.12, + "learning_rate": 1.8752268672746976e-05, + "loss": 0.9927, + "step": 15094 + }, + { + "epoch": 1.12, + "learning_rate": 1.8752075726535085e-05, + "loss": 0.9256, + "step": 15095 + }, + { + "epoch": 1.12, + "learning_rate": 1.875188276639876e-05, + "loss": 0.9769, + "step": 15096 + }, + { + "epoch": 1.12, + "learning_rate": 1.8751689792338317e-05, + "loss": 0.87, + "step": 15097 + }, + { + "epoch": 1.12, + "learning_rate": 1.8751496804354065e-05, + "loss": 0.9638, + "step": 15098 + }, + { + "epoch": 1.12, + "learning_rate": 1.87513038024463e-05, + "loss": 0.9884, + "step": 15099 + }, + { + "epoch": 1.12, + "learning_rate": 1.8751110786615342e-05, + "loss": 1.0131, + "step": 15100 + }, + { + "epoch": 1.12, + "learning_rate": 1.875091775686149e-05, + "loss": 0.952, + "step": 15101 + }, + { + "epoch": 1.12, + "learning_rate": 1.875072471318505e-05, + "loss": 1.0516, + "step": 15102 + }, + { + "epoch": 1.12, + "learning_rate": 1.8750531655586338e-05, + "loss": 1.0658, + "step": 15103 + }, + { + "epoch": 1.12, + "learning_rate": 1.875033858406565e-05, + "loss": 1.0067, + "step": 15104 + }, + { + "epoch": 1.12, + "learning_rate": 1.8750145498623304e-05, + "loss": 1.0389, + "step": 15105 + }, + { + "epoch": 1.12, + "learning_rate": 1.8749952399259595e-05, + "loss": 0.9869, + "step": 15106 + }, + { + "epoch": 1.12, + "learning_rate": 1.874975928597484e-05, + "loss": 0.9759, + "step": 15107 + }, + { + "epoch": 1.12, + "learning_rate": 1.8749566158769344e-05, + "loss": 1.0241, + "step": 15108 + }, + { + "epoch": 1.12, + "learning_rate": 1.874937301764341e-05, + "loss": 1.0764, + "step": 15109 + }, + { + "epoch": 1.12, + "learning_rate": 1.8749179862597353e-05, + "loss": 1.0384, + "step": 15110 + }, + { + "epoch": 1.12, + "learning_rate": 1.8748986693631477e-05, + "loss": 1.0597, + "step": 15111 + }, + { + "epoch": 1.12, + "learning_rate": 1.8748793510746085e-05, + "loss": 1.0042, + "step": 15112 + }, + { + "epoch": 1.12, + "learning_rate": 1.8748600313941488e-05, + "loss": 1.0054, + "step": 15113 + }, + { + "epoch": 1.12, + "learning_rate": 1.8748407103217992e-05, + "loss": 1.1113, + "step": 15114 + }, + { + "epoch": 1.12, + "learning_rate": 1.874821387857591e-05, + "loss": 0.9528, + "step": 15115 + }, + { + "epoch": 1.12, + "learning_rate": 1.874802064001554e-05, + "loss": 0.9268, + "step": 15116 + }, + { + "epoch": 1.12, + "learning_rate": 1.8747827387537198e-05, + "loss": 1.0699, + "step": 15117 + }, + { + "epoch": 1.12, + "learning_rate": 1.8747634121141185e-05, + "loss": 1.0509, + "step": 15118 + }, + { + "epoch": 1.12, + "learning_rate": 1.8747440840827813e-05, + "loss": 1.026, + "step": 15119 + }, + { + "epoch": 1.12, + "learning_rate": 1.8747247546597385e-05, + "loss": 1.122, + "step": 15120 + }, + { + "epoch": 1.12, + "learning_rate": 1.8747054238450216e-05, + "loss": 0.9418, + "step": 15121 + }, + { + "epoch": 1.12, + "learning_rate": 1.8746860916386605e-05, + "loss": 1.0325, + "step": 15122 + }, + { + "epoch": 1.12, + "learning_rate": 1.8746667580406865e-05, + "loss": 0.9816, + "step": 15123 + }, + { + "epoch": 1.12, + "learning_rate": 1.8746474230511302e-05, + "loss": 1.0222, + "step": 15124 + }, + { + "epoch": 1.12, + "learning_rate": 1.8746280866700225e-05, + "loss": 1.0147, + "step": 15125 + }, + { + "epoch": 1.12, + "learning_rate": 1.8746087488973937e-05, + "loss": 1.0128, + "step": 15126 + }, + { + "epoch": 1.12, + "learning_rate": 1.8745894097332752e-05, + "loss": 1.0739, + "step": 15127 + }, + { + "epoch": 1.12, + "learning_rate": 1.8745700691776974e-05, + "loss": 0.9911, + "step": 15128 + }, + { + "epoch": 1.12, + "learning_rate": 1.8745507272306908e-05, + "loss": 1.0629, + "step": 15129 + }, + { + "epoch": 1.12, + "learning_rate": 1.8745313838922868e-05, + "loss": 1.0661, + "step": 15130 + }, + { + "epoch": 1.12, + "learning_rate": 1.8745120391625157e-05, + "loss": 1.0854, + "step": 15131 + }, + { + "epoch": 1.12, + "learning_rate": 1.874492693041409e-05, + "loss": 0.9831, + "step": 15132 + }, + { + "epoch": 1.12, + "learning_rate": 1.8744733455289964e-05, + "loss": 1.0081, + "step": 15133 + }, + { + "epoch": 1.12, + "learning_rate": 1.8744539966253092e-05, + "loss": 1.0361, + "step": 15134 + }, + { + "epoch": 1.12, + "learning_rate": 1.8744346463303785e-05, + "loss": 1.0521, + "step": 15135 + }, + { + "epoch": 1.12, + "learning_rate": 1.8744152946442345e-05, + "loss": 1.025, + "step": 15136 + }, + { + "epoch": 1.12, + "learning_rate": 1.8743959415669083e-05, + "loss": 1.0702, + "step": 15137 + }, + { + "epoch": 1.12, + "learning_rate": 1.874376587098431e-05, + "loss": 0.9757, + "step": 15138 + }, + { + "epoch": 1.12, + "learning_rate": 1.874357231238833e-05, + "loss": 1.0543, + "step": 15139 + }, + { + "epoch": 1.12, + "learning_rate": 1.8743378739881447e-05, + "loss": 1.0252, + "step": 15140 + }, + { + "epoch": 1.12, + "learning_rate": 1.874318515346398e-05, + "loss": 0.9744, + "step": 15141 + }, + { + "epoch": 1.12, + "learning_rate": 1.8742991553136223e-05, + "loss": 1.0023, + "step": 15142 + }, + { + "epoch": 1.12, + "learning_rate": 1.8742797938898496e-05, + "loss": 0.946, + "step": 15143 + }, + { + "epoch": 1.12, + "learning_rate": 1.8742604310751106e-05, + "loss": 0.9058, + "step": 15144 + }, + { + "epoch": 1.12, + "learning_rate": 1.874241066869435e-05, + "loss": 1.0729, + "step": 15145 + }, + { + "epoch": 1.12, + "learning_rate": 1.874221701272855e-05, + "loss": 0.8957, + "step": 15146 + }, + { + "epoch": 1.12, + "learning_rate": 1.8742023342854005e-05, + "loss": 0.9923, + "step": 15147 + }, + { + "epoch": 1.12, + "learning_rate": 1.8741829659071027e-05, + "loss": 1.0916, + "step": 15148 + }, + { + "epoch": 1.12, + "learning_rate": 1.874163596137992e-05, + "loss": 0.9758, + "step": 15149 + }, + { + "epoch": 1.12, + "learning_rate": 1.8741442249781e-05, + "loss": 1.0202, + "step": 15150 + }, + { + "epoch": 1.12, + "learning_rate": 1.8741248524274567e-05, + "loss": 0.9981, + "step": 15151 + }, + { + "epoch": 1.12, + "learning_rate": 1.8741054784860932e-05, + "loss": 1.1162, + "step": 15152 + }, + { + "epoch": 1.12, + "learning_rate": 1.8740861031540406e-05, + "loss": 0.9834, + "step": 15153 + }, + { + "epoch": 1.12, + "learning_rate": 1.8740667264313295e-05, + "loss": 0.9844, + "step": 15154 + }, + { + "epoch": 1.12, + "learning_rate": 1.874047348317991e-05, + "loss": 0.9476, + "step": 15155 + }, + { + "epoch": 1.12, + "learning_rate": 1.8740279688140554e-05, + "loss": 1.0056, + "step": 15156 + }, + { + "epoch": 1.12, + "learning_rate": 1.8740085879195536e-05, + "loss": 1.0567, + "step": 15157 + }, + { + "epoch": 1.12, + "learning_rate": 1.8739892056345172e-05, + "loss": 0.9246, + "step": 15158 + }, + { + "epoch": 1.12, + "learning_rate": 1.873969821958976e-05, + "loss": 1.1012, + "step": 15159 + }, + { + "epoch": 1.12, + "learning_rate": 1.873950436892961e-05, + "loss": 1.0946, + "step": 15160 + }, + { + "epoch": 1.12, + "learning_rate": 1.8739310504365043e-05, + "loss": 0.9603, + "step": 15161 + }, + { + "epoch": 1.12, + "learning_rate": 1.873911662589635e-05, + "loss": 1.0853, + "step": 15162 + }, + { + "epoch": 1.12, + "learning_rate": 1.873892273352385e-05, + "loss": 0.9501, + "step": 15163 + }, + { + "epoch": 1.12, + "learning_rate": 1.873872882724785e-05, + "loss": 1.1125, + "step": 15164 + }, + { + "epoch": 1.12, + "learning_rate": 1.873853490706866e-05, + "loss": 1.0612, + "step": 15165 + }, + { + "epoch": 1.12, + "learning_rate": 1.8738340972986583e-05, + "loss": 1.0725, + "step": 15166 + }, + { + "epoch": 1.12, + "learning_rate": 1.8738147025001927e-05, + "loss": 1.0086, + "step": 15167 + }, + { + "epoch": 1.12, + "learning_rate": 1.8737953063115007e-05, + "loss": 1.0127, + "step": 15168 + }, + { + "epoch": 1.12, + "learning_rate": 1.8737759087326132e-05, + "loss": 0.9882, + "step": 15169 + }, + { + "epoch": 1.12, + "learning_rate": 1.8737565097635607e-05, + "loss": 0.983, + "step": 15170 + }, + { + "epoch": 1.12, + "learning_rate": 1.8737371094043736e-05, + "loss": 1.008, + "step": 15171 + }, + { + "epoch": 1.12, + "learning_rate": 1.8737177076550838e-05, + "loss": 1.0503, + "step": 15172 + }, + { + "epoch": 1.12, + "learning_rate": 1.8736983045157207e-05, + "loss": 0.966, + "step": 15173 + }, + { + "epoch": 1.12, + "learning_rate": 1.873678899986317e-05, + "loss": 1.0542, + "step": 15174 + }, + { + "epoch": 1.12, + "learning_rate": 1.8736594940669027e-05, + "loss": 1.0781, + "step": 15175 + }, + { + "epoch": 1.12, + "learning_rate": 1.873640086757508e-05, + "loss": 1.1024, + "step": 15176 + }, + { + "epoch": 1.12, + "learning_rate": 1.873620678058165e-05, + "loss": 0.9895, + "step": 15177 + }, + { + "epoch": 1.12, + "learning_rate": 1.8736012679689037e-05, + "loss": 1.0217, + "step": 15178 + }, + { + "epoch": 1.12, + "learning_rate": 1.8735818564897555e-05, + "loss": 1.0018, + "step": 15179 + }, + { + "epoch": 1.12, + "learning_rate": 1.8735624436207507e-05, + "loss": 0.9948, + "step": 15180 + }, + { + "epoch": 1.12, + "learning_rate": 1.8735430293619206e-05, + "loss": 1.0276, + "step": 15181 + }, + { + "epoch": 1.12, + "learning_rate": 1.8735236137132965e-05, + "loss": 1.0443, + "step": 15182 + }, + { + "epoch": 1.12, + "learning_rate": 1.8735041966749086e-05, + "loss": 1.06, + "step": 15183 + }, + { + "epoch": 1.12, + "learning_rate": 1.873484778246788e-05, + "loss": 1.1775, + "step": 15184 + }, + { + "epoch": 1.12, + "learning_rate": 1.8734653584289656e-05, + "loss": 1.1089, + "step": 15185 + }, + { + "epoch": 1.12, + "learning_rate": 1.8734459372214723e-05, + "loss": 1.1459, + "step": 15186 + }, + { + "epoch": 1.12, + "learning_rate": 1.8734265146243388e-05, + "loss": 1.0218, + "step": 15187 + }, + { + "epoch": 1.12, + "learning_rate": 1.8734070906375966e-05, + "loss": 0.995, + "step": 15188 + }, + { + "epoch": 1.12, + "learning_rate": 1.8733876652612758e-05, + "loss": 0.8793, + "step": 15189 + }, + { + "epoch": 1.12, + "learning_rate": 1.8733682384954083e-05, + "loss": 1.0061, + "step": 15190 + }, + { + "epoch": 1.12, + "learning_rate": 1.8733488103400236e-05, + "loss": 0.9965, + "step": 15191 + }, + { + "epoch": 1.12, + "learning_rate": 1.873329380795154e-05, + "loss": 1.0294, + "step": 15192 + }, + { + "epoch": 1.12, + "learning_rate": 1.8733099498608302e-05, + "loss": 1.0078, + "step": 15193 + }, + { + "epoch": 1.12, + "learning_rate": 1.873290517537082e-05, + "loss": 0.9921, + "step": 15194 + }, + { + "epoch": 1.12, + "learning_rate": 1.8732710838239416e-05, + "loss": 1.1117, + "step": 15195 + }, + { + "epoch": 1.12, + "learning_rate": 1.8732516487214393e-05, + "loss": 0.9976, + "step": 15196 + }, + { + "epoch": 1.12, + "learning_rate": 1.8732322122296057e-05, + "loss": 1.1049, + "step": 15197 + }, + { + "epoch": 1.12, + "learning_rate": 1.8732127743484728e-05, + "loss": 0.9768, + "step": 15198 + }, + { + "epoch": 1.12, + "learning_rate": 1.8731933350780704e-05, + "loss": 1.0539, + "step": 15199 + }, + { + "epoch": 1.12, + "learning_rate": 1.8731738944184302e-05, + "loss": 1.0188, + "step": 15200 + }, + { + "epoch": 1.12, + "learning_rate": 1.8731544523695825e-05, + "loss": 1.1282, + "step": 15201 + }, + { + "epoch": 1.12, + "learning_rate": 1.8731350089315587e-05, + "loss": 1.1164, + "step": 15202 + }, + { + "epoch": 1.12, + "learning_rate": 1.8731155641043897e-05, + "loss": 0.953, + "step": 15203 + }, + { + "epoch": 1.12, + "learning_rate": 1.873096117888106e-05, + "loss": 1.0012, + "step": 15204 + }, + { + "epoch": 1.12, + "learning_rate": 1.8730766702827394e-05, + "loss": 1.0059, + "step": 15205 + }, + { + "epoch": 1.12, + "learning_rate": 1.87305722128832e-05, + "loss": 1.1153, + "step": 15206 + }, + { + "epoch": 1.12, + "learning_rate": 1.873037770904879e-05, + "loss": 1.0661, + "step": 15207 + }, + { + "epoch": 1.12, + "learning_rate": 1.8730183191324474e-05, + "loss": 1.08, + "step": 15208 + }, + { + "epoch": 1.12, + "learning_rate": 1.8729988659710563e-05, + "loss": 1.0086, + "step": 15209 + }, + { + "epoch": 1.12, + "learning_rate": 1.8729794114207365e-05, + "loss": 0.9325, + "step": 15210 + }, + { + "epoch": 1.12, + "learning_rate": 1.8729599554815188e-05, + "loss": 0.9266, + "step": 15211 + }, + { + "epoch": 1.12, + "learning_rate": 1.8729404981534344e-05, + "loss": 1.0442, + "step": 15212 + }, + { + "epoch": 1.12, + "learning_rate": 1.872921039436514e-05, + "loss": 0.9879, + "step": 15213 + }, + { + "epoch": 1.12, + "learning_rate": 1.872901579330789e-05, + "loss": 0.8616, + "step": 15214 + }, + { + "epoch": 1.12, + "learning_rate": 1.87288211783629e-05, + "loss": 1.0071, + "step": 15215 + }, + { + "epoch": 1.12, + "learning_rate": 1.872862654953048e-05, + "loss": 1.019, + "step": 15216 + }, + { + "epoch": 1.12, + "learning_rate": 1.872843190681094e-05, + "loss": 1.0394, + "step": 15217 + }, + { + "epoch": 1.12, + "learning_rate": 1.8728237250204587e-05, + "loss": 0.9605, + "step": 15218 + }, + { + "epoch": 1.12, + "learning_rate": 1.8728042579711737e-05, + "loss": 1.1024, + "step": 15219 + }, + { + "epoch": 1.12, + "learning_rate": 1.8727847895332693e-05, + "loss": 0.9988, + "step": 15220 + }, + { + "epoch": 1.12, + "learning_rate": 1.8727653197067773e-05, + "loss": 1.0106, + "step": 15221 + }, + { + "epoch": 1.12, + "learning_rate": 1.8727458484917277e-05, + "loss": 0.9175, + "step": 15222 + }, + { + "epoch": 1.12, + "learning_rate": 1.872726375888152e-05, + "loss": 1.0459, + "step": 15223 + }, + { + "epoch": 1.12, + "learning_rate": 1.8727069018960813e-05, + "loss": 0.9507, + "step": 15224 + }, + { + "epoch": 1.12, + "learning_rate": 1.8726874265155462e-05, + "loss": 1.0083, + "step": 15225 + }, + { + "epoch": 1.13, + "learning_rate": 1.8726679497465785e-05, + "loss": 0.9511, + "step": 15226 + }, + { + "epoch": 1.13, + "learning_rate": 1.8726484715892078e-05, + "loss": 1.0376, + "step": 15227 + }, + { + "epoch": 1.13, + "learning_rate": 1.8726289920434658e-05, + "loss": 0.9684, + "step": 15228 + }, + { + "epoch": 1.13, + "learning_rate": 1.872609511109384e-05, + "loss": 1.0112, + "step": 15229 + }, + { + "epoch": 1.13, + "learning_rate": 1.872590028786993e-05, + "loss": 1.0061, + "step": 15230 + }, + { + "epoch": 1.13, + "learning_rate": 1.872570545076324e-05, + "loss": 0.9155, + "step": 15231 + }, + { + "epoch": 1.13, + "learning_rate": 1.8725510599774072e-05, + "loss": 1.0563, + "step": 15232 + }, + { + "epoch": 1.13, + "learning_rate": 1.872531573490274e-05, + "loss": 1.0822, + "step": 15233 + }, + { + "epoch": 1.13, + "learning_rate": 1.8725120856149562e-05, + "loss": 1.0473, + "step": 15234 + }, + { + "epoch": 1.13, + "learning_rate": 1.8724925963514837e-05, + "loss": 1.0163, + "step": 15235 + }, + { + "epoch": 1.13, + "learning_rate": 1.872473105699888e-05, + "loss": 1.0535, + "step": 15236 + }, + { + "epoch": 1.13, + "learning_rate": 1.8724536136602e-05, + "loss": 1.048, + "step": 15237 + }, + { + "epoch": 1.13, + "learning_rate": 1.872434120232451e-05, + "loss": 1.0211, + "step": 15238 + }, + { + "epoch": 1.13, + "learning_rate": 1.872414625416672e-05, + "loss": 0.9447, + "step": 15239 + }, + { + "epoch": 1.13, + "learning_rate": 1.8723951292128936e-05, + "loss": 0.9953, + "step": 15240 + }, + { + "epoch": 1.13, + "learning_rate": 1.8723756316211465e-05, + "loss": 1.0485, + "step": 15241 + }, + { + "epoch": 1.13, + "learning_rate": 1.8723561326414627e-05, + "loss": 0.9196, + "step": 15242 + }, + { + "epoch": 1.13, + "learning_rate": 1.872336632273873e-05, + "loss": 1.082, + "step": 15243 + }, + { + "epoch": 1.13, + "learning_rate": 1.872317130518408e-05, + "loss": 0.9717, + "step": 15244 + }, + { + "epoch": 1.13, + "learning_rate": 1.8722976273750987e-05, + "loss": 1.0892, + "step": 15245 + }, + { + "epoch": 1.13, + "learning_rate": 1.872278122843977e-05, + "loss": 0.9771, + "step": 15246 + }, + { + "epoch": 1.13, + "learning_rate": 1.8722586169250726e-05, + "loss": 1.0214, + "step": 15247 + }, + { + "epoch": 1.13, + "learning_rate": 1.8722391096184177e-05, + "loss": 0.9624, + "step": 15248 + }, + { + "epoch": 1.13, + "learning_rate": 1.8722196009240426e-05, + "loss": 1.0563, + "step": 15249 + }, + { + "epoch": 1.13, + "learning_rate": 1.8722000908419784e-05, + "loss": 0.8544, + "step": 15250 + }, + { + "epoch": 1.13, + "learning_rate": 1.872180579372257e-05, + "loss": 1.0751, + "step": 15251 + }, + { + "epoch": 1.13, + "learning_rate": 1.8721610665149083e-05, + "loss": 1.0837, + "step": 15252 + }, + { + "epoch": 1.13, + "learning_rate": 1.8721415522699635e-05, + "loss": 1.1234, + "step": 15253 + }, + { + "epoch": 1.13, + "learning_rate": 1.8721220366374544e-05, + "loss": 1.0651, + "step": 15254 + }, + { + "epoch": 1.13, + "learning_rate": 1.8721025196174116e-05, + "loss": 0.9962, + "step": 15255 + }, + { + "epoch": 1.13, + "learning_rate": 1.872083001209866e-05, + "loss": 0.9201, + "step": 15256 + }, + { + "epoch": 1.13, + "learning_rate": 1.872063481414849e-05, + "loss": 1.0992, + "step": 15257 + }, + { + "epoch": 1.13, + "learning_rate": 1.8720439602323914e-05, + "loss": 1.0088, + "step": 15258 + }, + { + "epoch": 1.13, + "learning_rate": 1.8720244376625242e-05, + "loss": 0.983, + "step": 15259 + }, + { + "epoch": 1.13, + "learning_rate": 1.872004913705279e-05, + "loss": 1.0264, + "step": 15260 + }, + { + "epoch": 1.13, + "learning_rate": 1.8719853883606862e-05, + "loss": 0.9559, + "step": 15261 + }, + { + "epoch": 1.13, + "learning_rate": 1.8719658616287774e-05, + "loss": 1.0463, + "step": 15262 + }, + { + "epoch": 1.13, + "learning_rate": 1.8719463335095827e-05, + "loss": 0.9647, + "step": 15263 + }, + { + "epoch": 1.13, + "learning_rate": 1.8719268040031348e-05, + "loss": 1.0401, + "step": 15264 + }, + { + "epoch": 1.13, + "learning_rate": 1.871907273109463e-05, + "loss": 0.9914, + "step": 15265 + }, + { + "epoch": 1.13, + "learning_rate": 1.8718877408285996e-05, + "loss": 1.0316, + "step": 15266 + }, + { + "epoch": 1.13, + "learning_rate": 1.871868207160575e-05, + "loss": 0.9913, + "step": 15267 + }, + { + "epoch": 1.13, + "learning_rate": 1.8718486721054207e-05, + "loss": 1.1285, + "step": 15268 + }, + { + "epoch": 1.13, + "learning_rate": 1.8718291356631678e-05, + "loss": 0.9771, + "step": 15269 + }, + { + "epoch": 1.13, + "learning_rate": 1.8718095978338473e-05, + "loss": 0.9035, + "step": 15270 + }, + { + "epoch": 1.13, + "learning_rate": 1.87179005861749e-05, + "loss": 1.0075, + "step": 15271 + }, + { + "epoch": 1.13, + "learning_rate": 1.871770518014127e-05, + "loss": 1.0732, + "step": 15272 + }, + { + "epoch": 1.13, + "learning_rate": 1.87175097602379e-05, + "loss": 1.0427, + "step": 15273 + }, + { + "epoch": 1.13, + "learning_rate": 1.8717314326465095e-05, + "loss": 0.9465, + "step": 15274 + }, + { + "epoch": 1.13, + "learning_rate": 1.8717118878823168e-05, + "loss": 1.0113, + "step": 15275 + }, + { + "epoch": 1.13, + "learning_rate": 1.8716923417312425e-05, + "loss": 1.0394, + "step": 15276 + }, + { + "epoch": 1.13, + "learning_rate": 1.8716727941933188e-05, + "loss": 1.0382, + "step": 15277 + }, + { + "epoch": 1.13, + "learning_rate": 1.8716532452685758e-05, + "loss": 1.0279, + "step": 15278 + }, + { + "epoch": 1.13, + "learning_rate": 1.871633694957045e-05, + "loss": 0.89, + "step": 15279 + }, + { + "epoch": 1.13, + "learning_rate": 1.8716141432587577e-05, + "loss": 0.9701, + "step": 15280 + }, + { + "epoch": 1.13, + "learning_rate": 1.8715945901737443e-05, + "loss": 1.1194, + "step": 15281 + }, + { + "epoch": 1.13, + "learning_rate": 1.8715750357020367e-05, + "loss": 0.997, + "step": 15282 + }, + { + "epoch": 1.13, + "learning_rate": 1.8715554798436656e-05, + "loss": 1.0837, + "step": 15283 + }, + { + "epoch": 1.13, + "learning_rate": 1.8715359225986623e-05, + "loss": 0.9616, + "step": 15284 + }, + { + "epoch": 1.13, + "learning_rate": 1.8715163639670576e-05, + "loss": 1.1245, + "step": 15285 + }, + { + "epoch": 1.13, + "learning_rate": 1.8714968039488834e-05, + "loss": 0.9395, + "step": 15286 + }, + { + "epoch": 1.13, + "learning_rate": 1.8714772425441697e-05, + "loss": 1.0699, + "step": 15287 + }, + { + "epoch": 1.13, + "learning_rate": 1.8714576797529482e-05, + "loss": 0.9044, + "step": 15288 + }, + { + "epoch": 1.13, + "learning_rate": 1.8714381155752503e-05, + "loss": 1.0998, + "step": 15289 + }, + { + "epoch": 1.13, + "learning_rate": 1.8714185500111064e-05, + "loss": 0.9, + "step": 15290 + }, + { + "epoch": 1.13, + "learning_rate": 1.8713989830605482e-05, + "loss": 1.0011, + "step": 15291 + }, + { + "epoch": 1.13, + "learning_rate": 1.8713794147236072e-05, + "loss": 0.9517, + "step": 15292 + }, + { + "epoch": 1.13, + "learning_rate": 1.8713598450003134e-05, + "loss": 1.0439, + "step": 15293 + }, + { + "epoch": 1.13, + "learning_rate": 1.8713402738906987e-05, + "loss": 0.972, + "step": 15294 + }, + { + "epoch": 1.13, + "learning_rate": 1.8713207013947944e-05, + "loss": 0.9962, + "step": 15295 + }, + { + "epoch": 1.13, + "learning_rate": 1.871301127512631e-05, + "loss": 1.0033, + "step": 15296 + }, + { + "epoch": 1.13, + "learning_rate": 1.87128155224424e-05, + "loss": 1.1101, + "step": 15297 + }, + { + "epoch": 1.13, + "learning_rate": 1.8712619755896525e-05, + "loss": 1.023, + "step": 15298 + }, + { + "epoch": 1.13, + "learning_rate": 1.8712423975488997e-05, + "loss": 1.0535, + "step": 15299 + }, + { + "epoch": 1.13, + "learning_rate": 1.8712228181220128e-05, + "loss": 1.0852, + "step": 15300 + }, + { + "epoch": 1.13, + "learning_rate": 1.8712032373090228e-05, + "loss": 1.0074, + "step": 15301 + }, + { + "epoch": 1.13, + "learning_rate": 1.871183655109961e-05, + "loss": 1.0546, + "step": 15302 + }, + { + "epoch": 1.13, + "learning_rate": 1.8711640715248585e-05, + "loss": 1.0387, + "step": 15303 + }, + { + "epoch": 1.13, + "learning_rate": 1.871144486553746e-05, + "loss": 0.9723, + "step": 15304 + }, + { + "epoch": 1.13, + "learning_rate": 1.8711249001966555e-05, + "loss": 1.0735, + "step": 15305 + }, + { + "epoch": 1.13, + "learning_rate": 1.8711053124536177e-05, + "loss": 1.0396, + "step": 15306 + }, + { + "epoch": 1.13, + "learning_rate": 1.871085723324664e-05, + "loss": 0.967, + "step": 15307 + }, + { + "epoch": 1.13, + "learning_rate": 1.871066132809825e-05, + "loss": 1.0559, + "step": 15308 + }, + { + "epoch": 1.13, + "learning_rate": 1.8710465409091324e-05, + "loss": 1.1607, + "step": 15309 + }, + { + "epoch": 1.13, + "learning_rate": 1.871026947622617e-05, + "loss": 0.9394, + "step": 15310 + }, + { + "epoch": 1.13, + "learning_rate": 1.8710073529503106e-05, + "loss": 1.0656, + "step": 15311 + }, + { + "epoch": 1.13, + "learning_rate": 1.8709877568922438e-05, + "loss": 1.0861, + "step": 15312 + }, + { + "epoch": 1.13, + "learning_rate": 1.870968159448448e-05, + "loss": 1.1135, + "step": 15313 + }, + { + "epoch": 1.13, + "learning_rate": 1.870948560618954e-05, + "loss": 1.0761, + "step": 15314 + }, + { + "epoch": 1.13, + "learning_rate": 1.870928960403794e-05, + "loss": 1.0374, + "step": 15315 + }, + { + "epoch": 1.13, + "learning_rate": 1.8709093588029975e-05, + "loss": 1.1462, + "step": 15316 + }, + { + "epoch": 1.13, + "learning_rate": 1.8708897558165972e-05, + "loss": 0.9307, + "step": 15317 + }, + { + "epoch": 1.13, + "learning_rate": 1.870870151444624e-05, + "loss": 1.0359, + "step": 15318 + }, + { + "epoch": 1.13, + "learning_rate": 1.8708505456871085e-05, + "loss": 1.0065, + "step": 15319 + }, + { + "epoch": 1.13, + "learning_rate": 1.8708309385440824e-05, + "loss": 0.995, + "step": 15320 + }, + { + "epoch": 1.13, + "learning_rate": 1.8708113300155764e-05, + "loss": 0.96, + "step": 15321 + }, + { + "epoch": 1.13, + "learning_rate": 1.8707917201016223e-05, + "loss": 0.9802, + "step": 15322 + }, + { + "epoch": 1.13, + "learning_rate": 1.870772108802251e-05, + "loss": 0.9518, + "step": 15323 + }, + { + "epoch": 1.13, + "learning_rate": 1.870752496117494e-05, + "loss": 0.9511, + "step": 15324 + }, + { + "epoch": 1.13, + "learning_rate": 1.870732882047382e-05, + "loss": 1.035, + "step": 15325 + }, + { + "epoch": 1.13, + "learning_rate": 1.8707132665919465e-05, + "loss": 0.986, + "step": 15326 + }, + { + "epoch": 1.13, + "learning_rate": 1.8706936497512185e-05, + "loss": 1.0532, + "step": 15327 + }, + { + "epoch": 1.13, + "learning_rate": 1.8706740315252294e-05, + "loss": 1.0165, + "step": 15328 + }, + { + "epoch": 1.13, + "learning_rate": 1.8706544119140102e-05, + "loss": 1.0902, + "step": 15329 + }, + { + "epoch": 1.13, + "learning_rate": 1.870634790917593e-05, + "loss": 1.0055, + "step": 15330 + }, + { + "epoch": 1.13, + "learning_rate": 1.8706151685360073e-05, + "loss": 1.0127, + "step": 15331 + }, + { + "epoch": 1.13, + "learning_rate": 1.870595544769286e-05, + "loss": 1.0938, + "step": 15332 + }, + { + "epoch": 1.13, + "learning_rate": 1.8705759196174593e-05, + "loss": 0.9464, + "step": 15333 + }, + { + "epoch": 1.13, + "learning_rate": 1.8705562930805592e-05, + "loss": 1.0702, + "step": 15334 + }, + { + "epoch": 1.13, + "learning_rate": 1.8705366651586162e-05, + "loss": 1.0882, + "step": 15335 + }, + { + "epoch": 1.13, + "learning_rate": 1.870517035851662e-05, + "loss": 0.9661, + "step": 15336 + }, + { + "epoch": 1.13, + "learning_rate": 1.8704974051597273e-05, + "loss": 0.9294, + "step": 15337 + }, + { + "epoch": 1.13, + "learning_rate": 1.8704777730828438e-05, + "loss": 1.0413, + "step": 15338 + }, + { + "epoch": 1.13, + "learning_rate": 1.8704581396210427e-05, + "loss": 1.0612, + "step": 15339 + }, + { + "epoch": 1.13, + "learning_rate": 1.8704385047743553e-05, + "loss": 0.9628, + "step": 15340 + }, + { + "epoch": 1.13, + "learning_rate": 1.8704188685428128e-05, + "loss": 1.0399, + "step": 15341 + }, + { + "epoch": 1.13, + "learning_rate": 1.870399230926446e-05, + "loss": 0.9836, + "step": 15342 + }, + { + "epoch": 1.13, + "learning_rate": 1.8703795919252865e-05, + "loss": 1.1095, + "step": 15343 + }, + { + "epoch": 1.13, + "learning_rate": 1.8703599515393655e-05, + "loss": 0.9693, + "step": 15344 + }, + { + "epoch": 1.13, + "learning_rate": 1.8703403097687144e-05, + "loss": 0.9338, + "step": 15345 + }, + { + "epoch": 1.13, + "learning_rate": 1.8703206666133642e-05, + "loss": 1.0517, + "step": 15346 + }, + { + "epoch": 1.13, + "learning_rate": 1.8703010220733463e-05, + "loss": 1.0644, + "step": 15347 + }, + { + "epoch": 1.13, + "learning_rate": 1.870281376148692e-05, + "loss": 1.0135, + "step": 15348 + }, + { + "epoch": 1.13, + "learning_rate": 1.8702617288394328e-05, + "loss": 1.0388, + "step": 15349 + }, + { + "epoch": 1.13, + "learning_rate": 1.8702420801455992e-05, + "loss": 0.9686, + "step": 15350 + }, + { + "epoch": 1.13, + "learning_rate": 1.8702224300672228e-05, + "loss": 1.1025, + "step": 15351 + }, + { + "epoch": 1.13, + "learning_rate": 1.870202778604335e-05, + "loss": 1.0239, + "step": 15352 + }, + { + "epoch": 1.13, + "learning_rate": 1.8701831257569678e-05, + "loss": 0.9956, + "step": 15353 + }, + { + "epoch": 1.13, + "learning_rate": 1.8701634715251507e-05, + "loss": 1.0362, + "step": 15354 + }, + { + "epoch": 1.13, + "learning_rate": 1.8701438159089163e-05, + "loss": 1.0565, + "step": 15355 + }, + { + "epoch": 1.13, + "learning_rate": 1.8701241589082956e-05, + "loss": 1.0381, + "step": 15356 + }, + { + "epoch": 1.13, + "learning_rate": 1.87010450052332e-05, + "loss": 1.0436, + "step": 15357 + }, + { + "epoch": 1.13, + "learning_rate": 1.8700848407540203e-05, + "loss": 1.0096, + "step": 15358 + }, + { + "epoch": 1.13, + "learning_rate": 1.8700651796004282e-05, + "loss": 1.0357, + "step": 15359 + }, + { + "epoch": 1.13, + "learning_rate": 1.870045517062575e-05, + "loss": 1.0309, + "step": 15360 + }, + { + "epoch": 1.14, + "learning_rate": 1.8700258531404916e-05, + "loss": 1.0904, + "step": 15361 + }, + { + "epoch": 1.14, + "learning_rate": 1.8700061878342098e-05, + "loss": 1.0011, + "step": 15362 + }, + { + "epoch": 1.14, + "learning_rate": 1.8699865211437602e-05, + "loss": 0.969, + "step": 15363 + }, + { + "epoch": 1.14, + "learning_rate": 1.8699668530691745e-05, + "loss": 0.9983, + "step": 15364 + }, + { + "epoch": 1.14, + "learning_rate": 1.869947183610484e-05, + "loss": 1.0669, + "step": 15365 + }, + { + "epoch": 1.14, + "learning_rate": 1.8699275127677207e-05, + "loss": 1.0888, + "step": 15366 + }, + { + "epoch": 1.14, + "learning_rate": 1.8699078405409144e-05, + "loss": 1.0358, + "step": 15367 + }, + { + "epoch": 1.14, + "learning_rate": 1.8698881669300972e-05, + "loss": 1.1151, + "step": 15368 + }, + { + "epoch": 1.14, + "learning_rate": 1.8698684919353004e-05, + "loss": 1.0479, + "step": 15369 + }, + { + "epoch": 1.14, + "learning_rate": 1.8698488155565554e-05, + "loss": 1.0126, + "step": 15370 + }, + { + "epoch": 1.14, + "learning_rate": 1.8698291377938934e-05, + "loss": 1.031, + "step": 15371 + }, + { + "epoch": 1.14, + "learning_rate": 1.869809458647346e-05, + "loss": 1.0227, + "step": 15372 + }, + { + "epoch": 1.14, + "learning_rate": 1.869789778116944e-05, + "loss": 1.1353, + "step": 15373 + }, + { + "epoch": 1.14, + "learning_rate": 1.8697700962027186e-05, + "loss": 1.0483, + "step": 15374 + }, + { + "epoch": 1.14, + "learning_rate": 1.8697504129047014e-05, + "loss": 0.9501, + "step": 15375 + }, + { + "epoch": 1.14, + "learning_rate": 1.869730728222924e-05, + "loss": 1.0958, + "step": 15376 + }, + { + "epoch": 1.14, + "learning_rate": 1.8697110421574175e-05, + "loss": 1.0046, + "step": 15377 + }, + { + "epoch": 1.14, + "learning_rate": 1.869691354708213e-05, + "loss": 1.0086, + "step": 15378 + }, + { + "epoch": 1.14, + "learning_rate": 1.869671665875342e-05, + "loss": 0.9642, + "step": 15379 + }, + { + "epoch": 1.14, + "learning_rate": 1.8696519756588363e-05, + "loss": 0.9663, + "step": 15380 + }, + { + "epoch": 1.14, + "learning_rate": 1.8696322840587263e-05, + "loss": 1.0417, + "step": 15381 + }, + { + "epoch": 1.14, + "learning_rate": 1.869612591075044e-05, + "loss": 1.1355, + "step": 15382 + }, + { + "epoch": 1.14, + "learning_rate": 1.8695928967078202e-05, + "loss": 1.051, + "step": 15383 + }, + { + "epoch": 1.14, + "learning_rate": 1.8695732009570865e-05, + "loss": 1.0661, + "step": 15384 + }, + { + "epoch": 1.14, + "learning_rate": 1.8695535038228745e-05, + "loss": 0.9329, + "step": 15385 + }, + { + "epoch": 1.14, + "learning_rate": 1.869533805305215e-05, + "loss": 1.0687, + "step": 15386 + }, + { + "epoch": 1.14, + "learning_rate": 1.8695141054041402e-05, + "loss": 0.9875, + "step": 15387 + }, + { + "epoch": 1.14, + "learning_rate": 1.8694944041196806e-05, + "loss": 1.1186, + "step": 15388 + }, + { + "epoch": 1.14, + "learning_rate": 1.8694747014518678e-05, + "loss": 1.098, + "step": 15389 + }, + { + "epoch": 1.14, + "learning_rate": 1.8694549974007336e-05, + "loss": 0.9855, + "step": 15390 + }, + { + "epoch": 1.14, + "learning_rate": 1.8694352919663082e-05, + "loss": 1.0775, + "step": 15391 + }, + { + "epoch": 1.14, + "learning_rate": 1.869415585148624e-05, + "loss": 1.1222, + "step": 15392 + }, + { + "epoch": 1.14, + "learning_rate": 1.8693958769477123e-05, + "loss": 0.9756, + "step": 15393 + }, + { + "epoch": 1.14, + "learning_rate": 1.869376167363604e-05, + "loss": 1.054, + "step": 15394 + }, + { + "epoch": 1.14, + "learning_rate": 1.869356456396331e-05, + "loss": 1.0527, + "step": 15395 + }, + { + "epoch": 1.14, + "learning_rate": 1.869336744045924e-05, + "loss": 0.973, + "step": 15396 + }, + { + "epoch": 1.14, + "learning_rate": 1.8693170303124147e-05, + "loss": 1.1266, + "step": 15397 + }, + { + "epoch": 1.14, + "learning_rate": 1.8692973151958343e-05, + "loss": 1.1361, + "step": 15398 + }, + { + "epoch": 1.14, + "learning_rate": 1.8692775986962143e-05, + "loss": 0.99, + "step": 15399 + }, + { + "epoch": 1.14, + "learning_rate": 1.8692578808135864e-05, + "loss": 1.0232, + "step": 15400 + }, + { + "epoch": 1.14, + "learning_rate": 1.8692381615479814e-05, + "loss": 1.0571, + "step": 15401 + }, + { + "epoch": 1.14, + "learning_rate": 1.869218440899431e-05, + "loss": 1.0758, + "step": 15402 + }, + { + "epoch": 1.14, + "learning_rate": 1.8691987188679663e-05, + "loss": 1.0148, + "step": 15403 + }, + { + "epoch": 1.14, + "learning_rate": 1.869178995453619e-05, + "loss": 0.9527, + "step": 15404 + }, + { + "epoch": 1.14, + "learning_rate": 1.8691592706564206e-05, + "loss": 1.0681, + "step": 15405 + }, + { + "epoch": 1.14, + "learning_rate": 1.869139544476402e-05, + "loss": 1.0334, + "step": 15406 + }, + { + "epoch": 1.14, + "learning_rate": 1.8691198169135948e-05, + "loss": 1.1578, + "step": 15407 + }, + { + "epoch": 1.14, + "learning_rate": 1.869100087968031e-05, + "loss": 1.0232, + "step": 15408 + }, + { + "epoch": 1.14, + "learning_rate": 1.8690803576397405e-05, + "loss": 0.9437, + "step": 15409 + }, + { + "epoch": 1.14, + "learning_rate": 1.869060625928756e-05, + "loss": 1.0741, + "step": 15410 + }, + { + "epoch": 1.14, + "learning_rate": 1.8690408928351083e-05, + "loss": 1.0351, + "step": 15411 + }, + { + "epoch": 1.14, + "learning_rate": 1.8690211583588292e-05, + "loss": 1.059, + "step": 15412 + }, + { + "epoch": 1.14, + "learning_rate": 1.8690014224999497e-05, + "loss": 1.0002, + "step": 15413 + }, + { + "epoch": 1.14, + "learning_rate": 1.8689816852585013e-05, + "loss": 1.0262, + "step": 15414 + }, + { + "epoch": 1.14, + "learning_rate": 1.868961946634516e-05, + "loss": 0.9777, + "step": 15415 + }, + { + "epoch": 1.14, + "learning_rate": 1.868942206628024e-05, + "loss": 0.9795, + "step": 15416 + }, + { + "epoch": 1.14, + "learning_rate": 1.8689224652390576e-05, + "loss": 0.9853, + "step": 15417 + }, + { + "epoch": 1.14, + "learning_rate": 1.8689027224676483e-05, + "loss": 0.9501, + "step": 15418 + }, + { + "epoch": 1.14, + "learning_rate": 1.868882978313827e-05, + "loss": 0.9687, + "step": 15419 + }, + { + "epoch": 1.14, + "learning_rate": 1.868863232777625e-05, + "loss": 1.0353, + "step": 15420 + }, + { + "epoch": 1.14, + "learning_rate": 1.8688434858590744e-05, + "loss": 0.9487, + "step": 15421 + }, + { + "epoch": 1.14, + "learning_rate": 1.8688237375582062e-05, + "loss": 0.982, + "step": 15422 + }, + { + "epoch": 1.14, + "learning_rate": 1.8688039878750517e-05, + "loss": 1.0765, + "step": 15423 + }, + { + "epoch": 1.14, + "learning_rate": 1.868784236809643e-05, + "loss": 0.9501, + "step": 15424 + }, + { + "epoch": 1.14, + "learning_rate": 1.8687644843620103e-05, + "loss": 1.096, + "step": 15425 + }, + { + "epoch": 1.14, + "learning_rate": 1.868744730532186e-05, + "loss": 0.9731, + "step": 15426 + }, + { + "epoch": 1.14, + "learning_rate": 1.8687249753202016e-05, + "loss": 1.1695, + "step": 15427 + }, + { + "epoch": 1.14, + "learning_rate": 1.868705218726088e-05, + "loss": 1.0291, + "step": 15428 + }, + { + "epoch": 1.14, + "learning_rate": 1.8686854607498765e-05, + "loss": 0.9631, + "step": 15429 + }, + { + "epoch": 1.14, + "learning_rate": 1.8686657013915993e-05, + "loss": 0.9977, + "step": 15430 + }, + { + "epoch": 1.14, + "learning_rate": 1.8686459406512873e-05, + "loss": 0.9817, + "step": 15431 + }, + { + "epoch": 1.14, + "learning_rate": 1.8686261785289718e-05, + "loss": 0.9558, + "step": 15432 + }, + { + "epoch": 1.14, + "learning_rate": 1.8686064150246842e-05, + "loss": 0.9499, + "step": 15433 + }, + { + "epoch": 1.14, + "learning_rate": 1.868586650138457e-05, + "loss": 0.9866, + "step": 15434 + }, + { + "epoch": 1.14, + "learning_rate": 1.8685668838703204e-05, + "loss": 1.01, + "step": 15435 + }, + { + "epoch": 1.14, + "learning_rate": 1.8685471162203065e-05, + "loss": 1.0864, + "step": 15436 + }, + { + "epoch": 1.14, + "learning_rate": 1.8685273471884466e-05, + "loss": 1.0238, + "step": 15437 + }, + { + "epoch": 1.14, + "learning_rate": 1.868507576774772e-05, + "loss": 0.9937, + "step": 15438 + }, + { + "epoch": 1.14, + "learning_rate": 1.868487804979314e-05, + "loss": 1.0889, + "step": 15439 + }, + { + "epoch": 1.14, + "learning_rate": 1.8684680318021047e-05, + "loss": 1.0982, + "step": 15440 + }, + { + "epoch": 1.14, + "learning_rate": 1.8684482572431748e-05, + "loss": 1.0448, + "step": 15441 + }, + { + "epoch": 1.14, + "learning_rate": 1.8684284813025567e-05, + "loss": 1.1065, + "step": 15442 + }, + { + "epoch": 1.14, + "learning_rate": 1.868408703980281e-05, + "loss": 1.0397, + "step": 15443 + }, + { + "epoch": 1.14, + "learning_rate": 1.8683889252763795e-05, + "loss": 0.944, + "step": 15444 + }, + { + "epoch": 1.14, + "learning_rate": 1.8683691451908836e-05, + "loss": 1.0062, + "step": 15445 + }, + { + "epoch": 1.14, + "learning_rate": 1.8683493637238247e-05, + "loss": 1.02, + "step": 15446 + }, + { + "epoch": 1.14, + "learning_rate": 1.8683295808752348e-05, + "loss": 1.0107, + "step": 15447 + }, + { + "epoch": 1.14, + "learning_rate": 1.8683097966451446e-05, + "loss": 1.0214, + "step": 15448 + }, + { + "epoch": 1.14, + "learning_rate": 1.868290011033586e-05, + "loss": 1.0567, + "step": 15449 + }, + { + "epoch": 1.14, + "learning_rate": 1.8682702240405906e-05, + "loss": 1.0047, + "step": 15450 + }, + { + "epoch": 1.14, + "learning_rate": 1.8682504356661898e-05, + "loss": 0.9514, + "step": 15451 + }, + { + "epoch": 1.14, + "learning_rate": 1.8682306459104145e-05, + "loss": 1.0708, + "step": 15452 + }, + { + "epoch": 1.14, + "learning_rate": 1.8682108547732968e-05, + "loss": 1.0294, + "step": 15453 + }, + { + "epoch": 1.14, + "learning_rate": 1.8681910622548683e-05, + "loss": 0.9538, + "step": 15454 + }, + { + "epoch": 1.14, + "learning_rate": 1.86817126835516e-05, + "loss": 0.926, + "step": 15455 + }, + { + "epoch": 1.14, + "learning_rate": 1.8681514730742038e-05, + "loss": 1.0473, + "step": 15456 + }, + { + "epoch": 1.14, + "learning_rate": 1.868131676412031e-05, + "loss": 0.9517, + "step": 15457 + }, + { + "epoch": 1.14, + "learning_rate": 1.868111878368673e-05, + "loss": 1.0095, + "step": 15458 + }, + { + "epoch": 1.14, + "learning_rate": 1.8680920789441616e-05, + "loss": 0.9404, + "step": 15459 + }, + { + "epoch": 1.14, + "learning_rate": 1.868072278138528e-05, + "loss": 1.023, + "step": 15460 + }, + { + "epoch": 1.14, + "learning_rate": 1.868052475951804e-05, + "loss": 0.9582, + "step": 15461 + }, + { + "epoch": 1.14, + "learning_rate": 1.8680326723840207e-05, + "loss": 1.0507, + "step": 15462 + }, + { + "epoch": 1.14, + "learning_rate": 1.86801286743521e-05, + "loss": 1.1644, + "step": 15463 + }, + { + "epoch": 1.14, + "learning_rate": 1.8679930611054033e-05, + "loss": 0.9412, + "step": 15464 + }, + { + "epoch": 1.14, + "learning_rate": 1.8679732533946318e-05, + "loss": 1.0171, + "step": 15465 + }, + { + "epoch": 1.14, + "learning_rate": 1.8679534443029276e-05, + "loss": 1.0537, + "step": 15466 + }, + { + "epoch": 1.14, + "learning_rate": 1.8679336338303215e-05, + "loss": 0.9955, + "step": 15467 + }, + { + "epoch": 1.14, + "learning_rate": 1.8679138219768454e-05, + "loss": 0.9913, + "step": 15468 + }, + { + "epoch": 1.14, + "learning_rate": 1.867894008742531e-05, + "loss": 0.9768, + "step": 15469 + }, + { + "epoch": 1.14, + "learning_rate": 1.8678741941274097e-05, + "loss": 0.901, + "step": 15470 + }, + { + "epoch": 1.14, + "learning_rate": 1.8678543781315132e-05, + "loss": 1.0132, + "step": 15471 + }, + { + "epoch": 1.14, + "learning_rate": 1.8678345607548724e-05, + "loss": 1.0078, + "step": 15472 + }, + { + "epoch": 1.14, + "learning_rate": 1.8678147419975193e-05, + "loss": 1.0187, + "step": 15473 + }, + { + "epoch": 1.14, + "learning_rate": 1.8677949218594857e-05, + "loss": 1.1153, + "step": 15474 + }, + { + "epoch": 1.14, + "learning_rate": 1.8677751003408025e-05, + "loss": 1.0892, + "step": 15475 + }, + { + "epoch": 1.14, + "learning_rate": 1.8677552774415014e-05, + "loss": 1.0397, + "step": 15476 + }, + { + "epoch": 1.14, + "learning_rate": 1.8677354531616142e-05, + "loss": 0.9901, + "step": 15477 + }, + { + "epoch": 1.14, + "learning_rate": 1.8677156275011724e-05, + "loss": 1.0246, + "step": 15478 + }, + { + "epoch": 1.14, + "learning_rate": 1.8676958004602075e-05, + "loss": 1.0833, + "step": 15479 + }, + { + "epoch": 1.14, + "learning_rate": 1.8676759720387513e-05, + "loss": 0.8886, + "step": 15480 + }, + { + "epoch": 1.14, + "learning_rate": 1.867656142236834e-05, + "loss": 0.9942, + "step": 15481 + }, + { + "epoch": 1.14, + "learning_rate": 1.867636311054489e-05, + "loss": 1.0566, + "step": 15482 + }, + { + "epoch": 1.14, + "learning_rate": 1.867616478491747e-05, + "loss": 0.9502, + "step": 15483 + }, + { + "epoch": 1.14, + "learning_rate": 1.8675966445486397e-05, + "loss": 0.9944, + "step": 15484 + }, + { + "epoch": 1.14, + "learning_rate": 1.8675768092251982e-05, + "loss": 1.1104, + "step": 15485 + }, + { + "epoch": 1.14, + "learning_rate": 1.8675569725214546e-05, + "loss": 1.0611, + "step": 15486 + }, + { + "epoch": 1.14, + "learning_rate": 1.86753713443744e-05, + "loss": 0.9843, + "step": 15487 + }, + { + "epoch": 1.14, + "learning_rate": 1.8675172949731867e-05, + "loss": 1.0392, + "step": 15488 + }, + { + "epoch": 1.14, + "learning_rate": 1.8674974541287252e-05, + "loss": 0.9878, + "step": 15489 + }, + { + "epoch": 1.14, + "learning_rate": 1.867477611904088e-05, + "loss": 1.0136, + "step": 15490 + }, + { + "epoch": 1.14, + "learning_rate": 1.8674577682993063e-05, + "loss": 1.0386, + "step": 15491 + }, + { + "epoch": 1.14, + "learning_rate": 1.867437923314412e-05, + "loss": 0.9744, + "step": 15492 + }, + { + "epoch": 1.14, + "learning_rate": 1.8674180769494362e-05, + "loss": 1.0705, + "step": 15493 + }, + { + "epoch": 1.14, + "learning_rate": 1.8673982292044104e-05, + "loss": 1.0693, + "step": 15494 + }, + { + "epoch": 1.14, + "learning_rate": 1.8673783800793667e-05, + "loss": 1.0837, + "step": 15495 + }, + { + "epoch": 1.14, + "learning_rate": 1.8673585295743362e-05, + "loss": 1.0138, + "step": 15496 + }, + { + "epoch": 1.15, + "learning_rate": 1.8673386776893508e-05, + "loss": 1.0638, + "step": 15497 + }, + { + "epoch": 1.15, + "learning_rate": 1.8673188244244417e-05, + "loss": 1.0072, + "step": 15498 + }, + { + "epoch": 1.15, + "learning_rate": 1.867298969779641e-05, + "loss": 1.0602, + "step": 15499 + }, + { + "epoch": 1.15, + "learning_rate": 1.8672791137549803e-05, + "loss": 1.0344, + "step": 15500 + }, + { + "epoch": 1.15, + "learning_rate": 1.8672592563504907e-05, + "loss": 0.9923, + "step": 15501 + }, + { + "epoch": 1.15, + "learning_rate": 1.867239397566204e-05, + "loss": 0.9697, + "step": 15502 + }, + { + "epoch": 1.15, + "learning_rate": 1.8672195374021516e-05, + "loss": 0.9838, + "step": 15503 + }, + { + "epoch": 1.15, + "learning_rate": 1.867199675858366e-05, + "loss": 0.9868, + "step": 15504 + }, + { + "epoch": 1.15, + "learning_rate": 1.8671798129348776e-05, + "loss": 1.0344, + "step": 15505 + }, + { + "epoch": 1.15, + "learning_rate": 1.8671599486317187e-05, + "loss": 0.9692, + "step": 15506 + }, + { + "epoch": 1.15, + "learning_rate": 1.8671400829489206e-05, + "loss": 1.0532, + "step": 15507 + }, + { + "epoch": 1.15, + "learning_rate": 1.867120215886515e-05, + "loss": 0.9697, + "step": 15508 + }, + { + "epoch": 1.15, + "learning_rate": 1.8671003474445335e-05, + "loss": 1.0419, + "step": 15509 + }, + { + "epoch": 1.15, + "learning_rate": 1.867080477623008e-05, + "loss": 1.1137, + "step": 15510 + }, + { + "epoch": 1.15, + "learning_rate": 1.8670606064219697e-05, + "loss": 1.0946, + "step": 15511 + }, + { + "epoch": 1.15, + "learning_rate": 1.8670407338414507e-05, + "loss": 0.9639, + "step": 15512 + }, + { + "epoch": 1.15, + "learning_rate": 1.867020859881482e-05, + "loss": 1.1121, + "step": 15513 + }, + { + "epoch": 1.15, + "learning_rate": 1.8670009845420956e-05, + "loss": 1.0289, + "step": 15514 + }, + { + "epoch": 1.15, + "learning_rate": 1.866981107823323e-05, + "loss": 1.0769, + "step": 15515 + }, + { + "epoch": 1.15, + "learning_rate": 1.8669612297251957e-05, + "loss": 0.9331, + "step": 15516 + }, + { + "epoch": 1.15, + "learning_rate": 1.8669413502477454e-05, + "loss": 1.0887, + "step": 15517 + }, + { + "epoch": 1.15, + "learning_rate": 1.8669214693910043e-05, + "loss": 0.999, + "step": 15518 + }, + { + "epoch": 1.15, + "learning_rate": 1.8669015871550033e-05, + "loss": 0.9165, + "step": 15519 + }, + { + "epoch": 1.15, + "learning_rate": 1.8668817035397744e-05, + "loss": 0.9685, + "step": 15520 + }, + { + "epoch": 1.15, + "learning_rate": 1.866861818545349e-05, + "loss": 0.9301, + "step": 15521 + }, + { + "epoch": 1.15, + "learning_rate": 1.8668419321717585e-05, + "loss": 0.9715, + "step": 15522 + }, + { + "epoch": 1.15, + "learning_rate": 1.8668220444190356e-05, + "loss": 1.1206, + "step": 15523 + }, + { + "epoch": 1.15, + "learning_rate": 1.8668021552872105e-05, + "loss": 0.9604, + "step": 15524 + }, + { + "epoch": 1.15, + "learning_rate": 1.8667822647763162e-05, + "loss": 1.0297, + "step": 15525 + }, + { + "epoch": 1.15, + "learning_rate": 1.866762372886383e-05, + "loss": 1.001, + "step": 15526 + }, + { + "epoch": 1.15, + "learning_rate": 1.866742479617444e-05, + "loss": 1.1508, + "step": 15527 + }, + { + "epoch": 1.15, + "learning_rate": 1.8667225849695302e-05, + "loss": 0.9109, + "step": 15528 + }, + { + "epoch": 1.15, + "learning_rate": 1.8667026889426726e-05, + "loss": 1.0437, + "step": 15529 + }, + { + "epoch": 1.15, + "learning_rate": 1.8666827915369037e-05, + "loss": 1.1032, + "step": 15530 + }, + { + "epoch": 1.15, + "learning_rate": 1.8666628927522546e-05, + "loss": 1.07, + "step": 15531 + }, + { + "epoch": 1.15, + "learning_rate": 1.8666429925887575e-05, + "loss": 0.9793, + "step": 15532 + }, + { + "epoch": 1.15, + "learning_rate": 1.8666230910464436e-05, + "loss": 1.1513, + "step": 15533 + }, + { + "epoch": 1.15, + "learning_rate": 1.866603188125345e-05, + "loss": 0.9868, + "step": 15534 + }, + { + "epoch": 1.15, + "learning_rate": 1.866583283825493e-05, + "loss": 1.0275, + "step": 15535 + }, + { + "epoch": 1.15, + "learning_rate": 1.8665633781469195e-05, + "loss": 1.0696, + "step": 15536 + }, + { + "epoch": 1.15, + "learning_rate": 1.866543471089656e-05, + "loss": 0.9311, + "step": 15537 + }, + { + "epoch": 1.15, + "learning_rate": 1.866523562653734e-05, + "loss": 1.0777, + "step": 15538 + }, + { + "epoch": 1.15, + "learning_rate": 1.8665036528391858e-05, + "loss": 1.0707, + "step": 15539 + }, + { + "epoch": 1.15, + "learning_rate": 1.8664837416460427e-05, + "loss": 1.0834, + "step": 15540 + }, + { + "epoch": 1.15, + "learning_rate": 1.866463829074336e-05, + "loss": 1.0429, + "step": 15541 + }, + { + "epoch": 1.15, + "learning_rate": 1.866443915124098e-05, + "loss": 0.9738, + "step": 15542 + }, + { + "epoch": 1.15, + "learning_rate": 1.8664239997953598e-05, + "loss": 1.0163, + "step": 15543 + }, + { + "epoch": 1.15, + "learning_rate": 1.866404083088154e-05, + "loss": 0.943, + "step": 15544 + }, + { + "epoch": 1.15, + "learning_rate": 1.8663841650025113e-05, + "loss": 0.9946, + "step": 15545 + }, + { + "epoch": 1.15, + "learning_rate": 1.866364245538464e-05, + "loss": 0.9585, + "step": 15546 + }, + { + "epoch": 1.15, + "learning_rate": 1.8663443246960434e-05, + "loss": 0.989, + "step": 15547 + }, + { + "epoch": 1.15, + "learning_rate": 1.866324402475281e-05, + "loss": 1.0196, + "step": 15548 + }, + { + "epoch": 1.15, + "learning_rate": 1.8663044788762096e-05, + "loss": 1.0695, + "step": 15549 + }, + { + "epoch": 1.15, + "learning_rate": 1.8662845538988596e-05, + "loss": 0.9947, + "step": 15550 + }, + { + "epoch": 1.15, + "learning_rate": 1.8662646275432637e-05, + "loss": 1.0398, + "step": 15551 + }, + { + "epoch": 1.15, + "learning_rate": 1.8662446998094526e-05, + "loss": 0.9439, + "step": 15552 + }, + { + "epoch": 1.15, + "learning_rate": 1.8662247706974593e-05, + "loss": 0.9405, + "step": 15553 + }, + { + "epoch": 1.15, + "learning_rate": 1.8662048402073147e-05, + "loss": 0.9059, + "step": 15554 + }, + { + "epoch": 1.15, + "learning_rate": 1.86618490833905e-05, + "loss": 1.0239, + "step": 15555 + }, + { + "epoch": 1.15, + "learning_rate": 1.8661649750926978e-05, + "loss": 1.1279, + "step": 15556 + }, + { + "epoch": 1.15, + "learning_rate": 1.8661450404682897e-05, + "loss": 1.0189, + "step": 15557 + }, + { + "epoch": 1.15, + "learning_rate": 1.8661251044658572e-05, + "loss": 1.049, + "step": 15558 + }, + { + "epoch": 1.15, + "learning_rate": 1.866105167085432e-05, + "loss": 0.9861, + "step": 15559 + }, + { + "epoch": 1.15, + "learning_rate": 1.8660852283270455e-05, + "loss": 0.9356, + "step": 15560 + }, + { + "epoch": 1.15, + "learning_rate": 1.8660652881907304e-05, + "loss": 1.0006, + "step": 15561 + }, + { + "epoch": 1.15, + "learning_rate": 1.8660453466765173e-05, + "loss": 0.9472, + "step": 15562 + }, + { + "epoch": 1.15, + "learning_rate": 1.866025403784439e-05, + "loss": 1.0073, + "step": 15563 + }, + { + "epoch": 1.15, + "learning_rate": 1.866005459514526e-05, + "loss": 0.9845, + "step": 15564 + }, + { + "epoch": 1.15, + "learning_rate": 1.8659855138668114e-05, + "loss": 1.035, + "step": 15565 + }, + { + "epoch": 1.15, + "learning_rate": 1.8659655668413255e-05, + "loss": 1.0491, + "step": 15566 + }, + { + "epoch": 1.15, + "learning_rate": 1.8659456184381013e-05, + "loss": 1.0606, + "step": 15567 + }, + { + "epoch": 1.15, + "learning_rate": 1.86592566865717e-05, + "loss": 0.9519, + "step": 15568 + }, + { + "epoch": 1.15, + "learning_rate": 1.865905717498563e-05, + "loss": 0.98, + "step": 15569 + }, + { + "epoch": 1.15, + "learning_rate": 1.8658857649623128e-05, + "loss": 1.0038, + "step": 15570 + }, + { + "epoch": 1.15, + "learning_rate": 1.8658658110484508e-05, + "loss": 0.978, + "step": 15571 + }, + { + "epoch": 1.15, + "learning_rate": 1.8658458557570082e-05, + "loss": 1.0499, + "step": 15572 + }, + { + "epoch": 1.15, + "learning_rate": 1.8658258990880174e-05, + "loss": 1.0688, + "step": 15573 + }, + { + "epoch": 1.15, + "learning_rate": 1.86580594104151e-05, + "loss": 0.9874, + "step": 15574 + }, + { + "epoch": 1.15, + "learning_rate": 1.865785981617518e-05, + "loss": 0.9414, + "step": 15575 + }, + { + "epoch": 1.15, + "learning_rate": 1.8657660208160725e-05, + "loss": 1.0194, + "step": 15576 + }, + { + "epoch": 1.15, + "learning_rate": 1.865746058637206e-05, + "loss": 0.9263, + "step": 15577 + }, + { + "epoch": 1.15, + "learning_rate": 1.86572609508095e-05, + "loss": 0.9844, + "step": 15578 + }, + { + "epoch": 1.15, + "learning_rate": 1.8657061301473356e-05, + "loss": 1.0107, + "step": 15579 + }, + { + "epoch": 1.15, + "learning_rate": 1.8656861638363956e-05, + "loss": 0.9392, + "step": 15580 + }, + { + "epoch": 1.15, + "learning_rate": 1.865666196148161e-05, + "loss": 0.9089, + "step": 15581 + }, + { + "epoch": 1.15, + "learning_rate": 1.8656462270826638e-05, + "loss": 0.9385, + "step": 15582 + }, + { + "epoch": 1.15, + "learning_rate": 1.865626256639936e-05, + "loss": 1.0106, + "step": 15583 + }, + { + "epoch": 1.15, + "learning_rate": 1.8656062848200093e-05, + "loss": 1.0466, + "step": 15584 + }, + { + "epoch": 1.15, + "learning_rate": 1.8655863116229157e-05, + "loss": 0.9648, + "step": 15585 + }, + { + "epoch": 1.15, + "learning_rate": 1.865566337048686e-05, + "loss": 0.9177, + "step": 15586 + }, + { + "epoch": 1.15, + "learning_rate": 1.8655463610973528e-05, + "loss": 1.116, + "step": 15587 + }, + { + "epoch": 1.15, + "learning_rate": 1.8655263837689482e-05, + "loss": 0.9225, + "step": 15588 + }, + { + "epoch": 1.15, + "learning_rate": 1.865506405063503e-05, + "loss": 1.0387, + "step": 15589 + }, + { + "epoch": 1.15, + "learning_rate": 1.86548642498105e-05, + "loss": 1.0135, + "step": 15590 + }, + { + "epoch": 1.15, + "learning_rate": 1.86546644352162e-05, + "loss": 1.1148, + "step": 15591 + }, + { + "epoch": 1.15, + "learning_rate": 1.8654464606852453e-05, + "loss": 0.9979, + "step": 15592 + }, + { + "epoch": 1.15, + "learning_rate": 1.865426476471958e-05, + "loss": 1.0318, + "step": 15593 + }, + { + "epoch": 1.15, + "learning_rate": 1.865406490881789e-05, + "loss": 1.0177, + "step": 15594 + }, + { + "epoch": 1.15, + "learning_rate": 1.865386503914771e-05, + "loss": 1.0786, + "step": 15595 + }, + { + "epoch": 1.15, + "learning_rate": 1.8653665155709357e-05, + "loss": 0.985, + "step": 15596 + }, + { + "epoch": 1.15, + "learning_rate": 1.8653465258503144e-05, + "loss": 1.0198, + "step": 15597 + }, + { + "epoch": 1.15, + "learning_rate": 1.865326534752939e-05, + "loss": 0.9755, + "step": 15598 + }, + { + "epoch": 1.15, + "learning_rate": 1.8653065422788415e-05, + "loss": 1.0501, + "step": 15599 + }, + { + "epoch": 1.15, + "learning_rate": 1.865286548428054e-05, + "loss": 0.9779, + "step": 15600 + }, + { + "epoch": 1.15, + "learning_rate": 1.865266553200608e-05, + "loss": 0.9813, + "step": 15601 + }, + { + "epoch": 1.15, + "learning_rate": 1.8652465565965347e-05, + "loss": 1.0168, + "step": 15602 + }, + { + "epoch": 1.15, + "learning_rate": 1.865226558615867e-05, + "loss": 1.0748, + "step": 15603 + }, + { + "epoch": 1.15, + "learning_rate": 1.8652065592586362e-05, + "loss": 0.9896, + "step": 15604 + }, + { + "epoch": 1.15, + "learning_rate": 1.8651865585248742e-05, + "loss": 1.0464, + "step": 15605 + }, + { + "epoch": 1.15, + "learning_rate": 1.8651665564146122e-05, + "loss": 1.1329, + "step": 15606 + }, + { + "epoch": 1.15, + "learning_rate": 1.865146552927883e-05, + "loss": 0.9751, + "step": 15607 + }, + { + "epoch": 1.15, + "learning_rate": 1.8651265480647182e-05, + "loss": 0.9555, + "step": 15608 + }, + { + "epoch": 1.15, + "learning_rate": 1.8651065418251492e-05, + "loss": 1.0571, + "step": 15609 + }, + { + "epoch": 1.15, + "learning_rate": 1.8650865342092083e-05, + "loss": 0.9888, + "step": 15610 + }, + { + "epoch": 1.15, + "learning_rate": 1.8650665252169266e-05, + "loss": 1.0642, + "step": 15611 + }, + { + "epoch": 1.15, + "learning_rate": 1.865046514848337e-05, + "loss": 1.0018, + "step": 15612 + }, + { + "epoch": 1.15, + "learning_rate": 1.8650265031034705e-05, + "loss": 1.0624, + "step": 15613 + }, + { + "epoch": 1.15, + "learning_rate": 1.8650064899823592e-05, + "loss": 1.1091, + "step": 15614 + }, + { + "epoch": 1.15, + "learning_rate": 1.8649864754850352e-05, + "loss": 0.9905, + "step": 15615 + }, + { + "epoch": 1.15, + "learning_rate": 1.8649664596115297e-05, + "loss": 1.1659, + "step": 15616 + }, + { + "epoch": 1.15, + "learning_rate": 1.864946442361875e-05, + "loss": 0.9973, + "step": 15617 + }, + { + "epoch": 1.15, + "learning_rate": 1.864926423736103e-05, + "loss": 1.0696, + "step": 15618 + }, + { + "epoch": 1.15, + "learning_rate": 1.8649064037342456e-05, + "loss": 0.9864, + "step": 15619 + }, + { + "epoch": 1.15, + "learning_rate": 1.864886382356334e-05, + "loss": 1.0686, + "step": 15620 + }, + { + "epoch": 1.15, + "learning_rate": 1.864866359602401e-05, + "loss": 1.0294, + "step": 15621 + }, + { + "epoch": 1.15, + "learning_rate": 1.864846335472478e-05, + "loss": 0.9238, + "step": 15622 + }, + { + "epoch": 1.15, + "learning_rate": 1.864826309966597e-05, + "loss": 0.9932, + "step": 15623 + }, + { + "epoch": 1.15, + "learning_rate": 1.8648062830847894e-05, + "loss": 1.0691, + "step": 15624 + }, + { + "epoch": 1.15, + "learning_rate": 1.8647862548270872e-05, + "loss": 1.0021, + "step": 15625 + }, + { + "epoch": 1.15, + "learning_rate": 1.8647662251935226e-05, + "loss": 1.0159, + "step": 15626 + }, + { + "epoch": 1.15, + "learning_rate": 1.8647461941841274e-05, + "loss": 1.0834, + "step": 15627 + }, + { + "epoch": 1.15, + "learning_rate": 1.8647261617989333e-05, + "loss": 1.0395, + "step": 15628 + }, + { + "epoch": 1.15, + "learning_rate": 1.8647061280379723e-05, + "loss": 1.0235, + "step": 15629 + }, + { + "epoch": 1.15, + "learning_rate": 1.8646860929012763e-05, + "loss": 0.9592, + "step": 15630 + }, + { + "epoch": 1.15, + "learning_rate": 1.864666056388877e-05, + "loss": 1.0467, + "step": 15631 + }, + { + "epoch": 1.16, + "learning_rate": 1.8646460185008065e-05, + "loss": 1.0644, + "step": 15632 + }, + { + "epoch": 1.16, + "learning_rate": 1.8646259792370963e-05, + "loss": 1.0327, + "step": 15633 + }, + { + "epoch": 1.16, + "learning_rate": 1.8646059385977785e-05, + "loss": 0.9145, + "step": 15634 + }, + { + "epoch": 1.16, + "learning_rate": 1.8645858965828853e-05, + "loss": 1.0784, + "step": 15635 + }, + { + "epoch": 1.16, + "learning_rate": 1.8645658531924482e-05, + "loss": 1.1104, + "step": 15636 + }, + { + "epoch": 1.16, + "learning_rate": 1.8645458084264994e-05, + "loss": 0.9796, + "step": 15637 + }, + { + "epoch": 1.16, + "learning_rate": 1.8645257622850704e-05, + "loss": 1.1017, + "step": 15638 + }, + { + "epoch": 1.16, + "learning_rate": 1.864505714768193e-05, + "loss": 1.0657, + "step": 15639 + }, + { + "epoch": 1.16, + "learning_rate": 1.8644856658758995e-05, + "loss": 1.0284, + "step": 15640 + }, + { + "epoch": 1.16, + "learning_rate": 1.864465615608222e-05, + "loss": 1.0251, + "step": 15641 + }, + { + "epoch": 1.16, + "learning_rate": 1.864445563965192e-05, + "loss": 0.9326, + "step": 15642 + }, + { + "epoch": 1.16, + "learning_rate": 1.864425510946841e-05, + "loss": 1.039, + "step": 15643 + }, + { + "epoch": 1.16, + "learning_rate": 1.864405456553202e-05, + "loss": 1.1226, + "step": 15644 + }, + { + "epoch": 1.16, + "learning_rate": 1.864385400784306e-05, + "loss": 0.9457, + "step": 15645 + }, + { + "epoch": 1.16, + "learning_rate": 1.8643653436401854e-05, + "loss": 1.0455, + "step": 15646 + }, + { + "epoch": 1.16, + "learning_rate": 1.8643452851208715e-05, + "loss": 0.9515, + "step": 15647 + }, + { + "epoch": 1.16, + "learning_rate": 1.8643252252263973e-05, + "loss": 0.9484, + "step": 15648 + }, + { + "epoch": 1.16, + "learning_rate": 1.8643051639567935e-05, + "loss": 1.0821, + "step": 15649 + }, + { + "epoch": 1.16, + "learning_rate": 1.8642851013120923e-05, + "loss": 1.0071, + "step": 15650 + }, + { + "epoch": 1.16, + "learning_rate": 1.8642650372923264e-05, + "loss": 0.9937, + "step": 15651 + }, + { + "epoch": 1.16, + "learning_rate": 1.864244971897527e-05, + "loss": 1.0549, + "step": 15652 + }, + { + "epoch": 1.16, + "learning_rate": 1.8642249051277264e-05, + "loss": 1.1048, + "step": 15653 + }, + { + "epoch": 1.16, + "learning_rate": 1.864204836982956e-05, + "loss": 1.0136, + "step": 15654 + }, + { + "epoch": 1.16, + "learning_rate": 1.8641847674632483e-05, + "loss": 1.0612, + "step": 15655 + }, + { + "epoch": 1.16, + "learning_rate": 1.864164696568635e-05, + "loss": 0.9961, + "step": 15656 + }, + { + "epoch": 1.16, + "learning_rate": 1.864144624299148e-05, + "loss": 1.1081, + "step": 15657 + }, + { + "epoch": 1.16, + "learning_rate": 1.864124550654819e-05, + "loss": 1.0771, + "step": 15658 + }, + { + "epoch": 1.16, + "learning_rate": 1.8641044756356807e-05, + "loss": 1.0238, + "step": 15659 + }, + { + "epoch": 1.16, + "learning_rate": 1.8640843992417642e-05, + "loss": 1.0679, + "step": 15660 + }, + { + "epoch": 1.16, + "learning_rate": 1.8640643214731018e-05, + "loss": 1.0545, + "step": 15661 + }, + { + "epoch": 1.16, + "learning_rate": 1.8640442423297255e-05, + "loss": 1.1795, + "step": 15662 + }, + { + "epoch": 1.16, + "learning_rate": 1.864024161811667e-05, + "loss": 0.9656, + "step": 15663 + }, + { + "epoch": 1.16, + "learning_rate": 1.8640040799189586e-05, + "loss": 1.0584, + "step": 15664 + }, + { + "epoch": 1.16, + "learning_rate": 1.8639839966516323e-05, + "loss": 1.0213, + "step": 15665 + }, + { + "epoch": 1.16, + "learning_rate": 1.8639639120097193e-05, + "loss": 1.0442, + "step": 15666 + }, + { + "epoch": 1.16, + "learning_rate": 1.8639438259932525e-05, + "loss": 1.0628, + "step": 15667 + }, + { + "epoch": 1.16, + "learning_rate": 1.8639237386022632e-05, + "loss": 1.0082, + "step": 15668 + }, + { + "epoch": 1.16, + "learning_rate": 1.8639036498367837e-05, + "loss": 0.987, + "step": 15669 + }, + { + "epoch": 1.16, + "learning_rate": 1.863883559696846e-05, + "loss": 1.0189, + "step": 15670 + }, + { + "epoch": 1.16, + "learning_rate": 1.8638634681824817e-05, + "loss": 0.9088, + "step": 15671 + }, + { + "epoch": 1.16, + "learning_rate": 1.863843375293723e-05, + "loss": 1.1302, + "step": 15672 + }, + { + "epoch": 1.16, + "learning_rate": 1.8638232810306017e-05, + "loss": 1.116, + "step": 15673 + }, + { + "epoch": 1.16, + "learning_rate": 1.86380318539315e-05, + "loss": 1.0498, + "step": 15674 + }, + { + "epoch": 1.16, + "learning_rate": 1.8637830883814e-05, + "loss": 1.1223, + "step": 15675 + }, + { + "epoch": 1.16, + "learning_rate": 1.8637629899953836e-05, + "loss": 0.9795, + "step": 15676 + }, + { + "epoch": 1.16, + "learning_rate": 1.8637428902351323e-05, + "loss": 1.0868, + "step": 15677 + }, + { + "epoch": 1.16, + "learning_rate": 1.8637227891006784e-05, + "loss": 1.0551, + "step": 15678 + }, + { + "epoch": 1.16, + "learning_rate": 1.863702686592054e-05, + "loss": 1.0697, + "step": 15679 + }, + { + "epoch": 1.16, + "learning_rate": 1.863682582709291e-05, + "loss": 1.0197, + "step": 15680 + }, + { + "epoch": 1.16, + "learning_rate": 1.8636624774524216e-05, + "loss": 1.0313, + "step": 15681 + }, + { + "epoch": 1.16, + "learning_rate": 1.8636423708214774e-05, + "loss": 1.039, + "step": 15682 + }, + { + "epoch": 1.16, + "learning_rate": 1.8636222628164905e-05, + "loss": 0.9813, + "step": 15683 + }, + { + "epoch": 1.16, + "learning_rate": 1.8636021534374926e-05, + "loss": 1.0138, + "step": 15684 + }, + { + "epoch": 1.16, + "learning_rate": 1.8635820426845164e-05, + "loss": 1.1062, + "step": 15685 + }, + { + "epoch": 1.16, + "learning_rate": 1.8635619305575935e-05, + "loss": 1.0187, + "step": 15686 + }, + { + "epoch": 1.16, + "learning_rate": 1.8635418170567557e-05, + "loss": 0.8968, + "step": 15687 + }, + { + "epoch": 1.16, + "learning_rate": 1.8635217021820355e-05, + "loss": 0.9545, + "step": 15688 + }, + { + "epoch": 1.16, + "learning_rate": 1.8635015859334645e-05, + "loss": 0.9494, + "step": 15689 + }, + { + "epoch": 1.16, + "learning_rate": 1.863481468311075e-05, + "loss": 0.9072, + "step": 15690 + }, + { + "epoch": 1.16, + "learning_rate": 1.8634613493148984e-05, + "loss": 1.0757, + "step": 15691 + }, + { + "epoch": 1.16, + "learning_rate": 1.8634412289449676e-05, + "loss": 1.0316, + "step": 15692 + }, + { + "epoch": 1.16, + "learning_rate": 1.863421107201314e-05, + "loss": 1.0559, + "step": 15693 + }, + { + "epoch": 1.16, + "learning_rate": 1.86340098408397e-05, + "loss": 0.9869, + "step": 15694 + }, + { + "epoch": 1.16, + "learning_rate": 1.863380859592967e-05, + "loss": 1.0387, + "step": 15695 + }, + { + "epoch": 1.16, + "learning_rate": 1.8633607337283376e-05, + "loss": 1.0795, + "step": 15696 + }, + { + "epoch": 1.16, + "learning_rate": 1.8633406064901135e-05, + "loss": 0.9832, + "step": 15697 + }, + { + "epoch": 1.16, + "learning_rate": 1.863320477878327e-05, + "loss": 0.9528, + "step": 15698 + }, + { + "epoch": 1.16, + "learning_rate": 1.86330034789301e-05, + "loss": 1.0296, + "step": 15699 + }, + { + "epoch": 1.16, + "learning_rate": 1.8632802165341944e-05, + "loss": 1.0348, + "step": 15700 + }, + { + "epoch": 1.16, + "learning_rate": 1.8632600838019123e-05, + "loss": 1.0732, + "step": 15701 + }, + { + "epoch": 1.16, + "learning_rate": 1.8632399496961957e-05, + "loss": 1.1012, + "step": 15702 + }, + { + "epoch": 1.16, + "learning_rate": 1.863219814217077e-05, + "loss": 1.0315, + "step": 15703 + }, + { + "epoch": 1.16, + "learning_rate": 1.8631996773645877e-05, + "loss": 1.0255, + "step": 15704 + }, + { + "epoch": 1.16, + "learning_rate": 1.8631795391387602e-05, + "loss": 1.0034, + "step": 15705 + }, + { + "epoch": 1.16, + "learning_rate": 1.863159399539626e-05, + "loss": 1.004, + "step": 15706 + }, + { + "epoch": 1.16, + "learning_rate": 1.863139258567218e-05, + "loss": 0.9622, + "step": 15707 + }, + { + "epoch": 1.16, + "learning_rate": 1.8631191162215677e-05, + "loss": 0.947, + "step": 15708 + }, + { + "epoch": 1.16, + "learning_rate": 1.863098972502707e-05, + "loss": 0.9665, + "step": 15709 + }, + { + "epoch": 1.16, + "learning_rate": 1.8630788274106684e-05, + "loss": 0.9262, + "step": 15710 + }, + { + "epoch": 1.16, + "learning_rate": 1.8630586809454835e-05, + "loss": 0.9832, + "step": 15711 + }, + { + "epoch": 1.16, + "learning_rate": 1.863038533107185e-05, + "loss": 0.9547, + "step": 15712 + }, + { + "epoch": 1.16, + "learning_rate": 1.8630183838958043e-05, + "loss": 0.9843, + "step": 15713 + }, + { + "epoch": 1.16, + "learning_rate": 1.8629982333113735e-05, + "loss": 0.9764, + "step": 15714 + }, + { + "epoch": 1.16, + "learning_rate": 1.862978081353925e-05, + "loss": 0.9885, + "step": 15715 + }, + { + "epoch": 1.16, + "learning_rate": 1.862957928023491e-05, + "loss": 0.9965, + "step": 15716 + }, + { + "epoch": 1.16, + "learning_rate": 1.8629377733201032e-05, + "loss": 0.9612, + "step": 15717 + }, + { + "epoch": 1.16, + "learning_rate": 1.8629176172437936e-05, + "loss": 1.083, + "step": 15718 + }, + { + "epoch": 1.16, + "learning_rate": 1.8628974597945944e-05, + "loss": 1.0411, + "step": 15719 + }, + { + "epoch": 1.16, + "learning_rate": 1.8628773009725375e-05, + "loss": 0.9808, + "step": 15720 + }, + { + "epoch": 1.16, + "learning_rate": 1.8628571407776556e-05, + "loss": 1.1122, + "step": 15721 + }, + { + "epoch": 1.16, + "learning_rate": 1.8628369792099802e-05, + "loss": 0.8634, + "step": 15722 + }, + { + "epoch": 1.16, + "learning_rate": 1.8628168162695434e-05, + "loss": 0.974, + "step": 15723 + }, + { + "epoch": 1.16, + "learning_rate": 1.8627966519563775e-05, + "loss": 1.0521, + "step": 15724 + }, + { + "epoch": 1.16, + "learning_rate": 1.862776486270514e-05, + "loss": 1.1148, + "step": 15725 + }, + { + "epoch": 1.16, + "learning_rate": 1.862756319211986e-05, + "loss": 1.0463, + "step": 15726 + }, + { + "epoch": 1.16, + "learning_rate": 1.862736150780825e-05, + "loss": 0.9355, + "step": 15727 + }, + { + "epoch": 1.16, + "learning_rate": 1.862715980977063e-05, + "loss": 1.1032, + "step": 15728 + }, + { + "epoch": 1.16, + "learning_rate": 1.862695809800732e-05, + "loss": 0.999, + "step": 15729 + }, + { + "epoch": 1.16, + "learning_rate": 1.8626756372518645e-05, + "loss": 0.9772, + "step": 15730 + }, + { + "epoch": 1.16, + "learning_rate": 1.8626554633304925e-05, + "loss": 0.9178, + "step": 15731 + }, + { + "epoch": 1.16, + "learning_rate": 1.862635288036648e-05, + "loss": 0.9653, + "step": 15732 + }, + { + "epoch": 1.16, + "learning_rate": 1.862615111370363e-05, + "loss": 1.0894, + "step": 15733 + }, + { + "epoch": 1.16, + "learning_rate": 1.86259493333167e-05, + "loss": 1.0433, + "step": 15734 + }, + { + "epoch": 1.16, + "learning_rate": 1.8625747539206004e-05, + "loss": 0.9821, + "step": 15735 + }, + { + "epoch": 1.16, + "learning_rate": 1.8625545731371865e-05, + "loss": 1.0879, + "step": 15736 + }, + { + "epoch": 1.16, + "learning_rate": 1.8625343909814613e-05, + "loss": 0.9454, + "step": 15737 + }, + { + "epoch": 1.16, + "learning_rate": 1.862514207453456e-05, + "loss": 1.0486, + "step": 15738 + }, + { + "epoch": 1.16, + "learning_rate": 1.8624940225532027e-05, + "loss": 1.1585, + "step": 15739 + }, + { + "epoch": 1.16, + "learning_rate": 1.8624738362807337e-05, + "loss": 0.9553, + "step": 15740 + }, + { + "epoch": 1.16, + "learning_rate": 1.8624536486360812e-05, + "loss": 1.0353, + "step": 15741 + }, + { + "epoch": 1.16, + "learning_rate": 1.8624334596192777e-05, + "loss": 1.0896, + "step": 15742 + }, + { + "epoch": 1.16, + "learning_rate": 1.8624132692303546e-05, + "loss": 1.0396, + "step": 15743 + }, + { + "epoch": 1.16, + "learning_rate": 1.8623930774693445e-05, + "loss": 1.0891, + "step": 15744 + }, + { + "epoch": 1.16, + "learning_rate": 1.8623728843362793e-05, + "loss": 0.9502, + "step": 15745 + }, + { + "epoch": 1.16, + "learning_rate": 1.862352689831191e-05, + "loss": 0.9943, + "step": 15746 + }, + { + "epoch": 1.16, + "learning_rate": 1.862332493954112e-05, + "loss": 1.0528, + "step": 15747 + }, + { + "epoch": 1.16, + "learning_rate": 1.862312296705074e-05, + "loss": 1.0493, + "step": 15748 + }, + { + "epoch": 1.16, + "learning_rate": 1.86229209808411e-05, + "loss": 1.0548, + "step": 15749 + }, + { + "epoch": 1.16, + "learning_rate": 1.8622718980912515e-05, + "loss": 0.9666, + "step": 15750 + }, + { + "epoch": 1.16, + "learning_rate": 1.8622516967265304e-05, + "loss": 1.1024, + "step": 15751 + }, + { + "epoch": 1.16, + "learning_rate": 1.8622314939899798e-05, + "loss": 0.9428, + "step": 15752 + }, + { + "epoch": 1.16, + "learning_rate": 1.862211289881631e-05, + "loss": 1.1773, + "step": 15753 + }, + { + "epoch": 1.16, + "learning_rate": 1.8621910844015158e-05, + "loss": 0.9786, + "step": 15754 + }, + { + "epoch": 1.16, + "learning_rate": 1.8621708775496675e-05, + "loss": 0.9557, + "step": 15755 + }, + { + "epoch": 1.16, + "learning_rate": 1.8621506693261174e-05, + "loss": 1.021, + "step": 15756 + }, + { + "epoch": 1.16, + "learning_rate": 1.8621304597308984e-05, + "loss": 1.065, + "step": 15757 + }, + { + "epoch": 1.16, + "learning_rate": 1.8621102487640418e-05, + "loss": 1.1262, + "step": 15758 + }, + { + "epoch": 1.16, + "learning_rate": 1.8620900364255802e-05, + "loss": 1.0338, + "step": 15759 + }, + { + "epoch": 1.16, + "learning_rate": 1.8620698227155457e-05, + "loss": 1.0193, + "step": 15760 + }, + { + "epoch": 1.16, + "learning_rate": 1.8620496076339703e-05, + "loss": 1.1185, + "step": 15761 + }, + { + "epoch": 1.16, + "learning_rate": 1.8620293911808863e-05, + "loss": 1.019, + "step": 15762 + }, + { + "epoch": 1.16, + "learning_rate": 1.862009173356326e-05, + "loss": 1.0386, + "step": 15763 + }, + { + "epoch": 1.16, + "learning_rate": 1.8619889541603215e-05, + "loss": 1.0613, + "step": 15764 + }, + { + "epoch": 1.16, + "learning_rate": 1.8619687335929047e-05, + "loss": 1.0911, + "step": 15765 + }, + { + "epoch": 1.16, + "learning_rate": 1.8619485116541084e-05, + "loss": 1.0545, + "step": 15766 + }, + { + "epoch": 1.17, + "learning_rate": 1.8619282883439637e-05, + "loss": 0.9078, + "step": 15767 + }, + { + "epoch": 1.17, + "learning_rate": 1.861908063662504e-05, + "loss": 1.0714, + "step": 15768 + }, + { + "epoch": 1.17, + "learning_rate": 1.8618878376097606e-05, + "loss": 1.0331, + "step": 15769 + }, + { + "epoch": 1.17, + "learning_rate": 1.8618676101857663e-05, + "loss": 1.0742, + "step": 15770 + }, + { + "epoch": 1.17, + "learning_rate": 1.8618473813905527e-05, + "loss": 1.0065, + "step": 15771 + }, + { + "epoch": 1.17, + "learning_rate": 1.8618271512241523e-05, + "loss": 1.0511, + "step": 15772 + }, + { + "epoch": 1.17, + "learning_rate": 1.861806919686597e-05, + "loss": 1.0326, + "step": 15773 + }, + { + "epoch": 1.17, + "learning_rate": 1.8617866867779194e-05, + "loss": 1.0123, + "step": 15774 + }, + { + "epoch": 1.17, + "learning_rate": 1.8617664524981518e-05, + "loss": 0.9268, + "step": 15775 + }, + { + "epoch": 1.17, + "learning_rate": 1.8617462168473256e-05, + "loss": 1.0809, + "step": 15776 + }, + { + "epoch": 1.17, + "learning_rate": 1.861725979825474e-05, + "loss": 0.9966, + "step": 15777 + }, + { + "epoch": 1.17, + "learning_rate": 1.8617057414326285e-05, + "loss": 1.0676, + "step": 15778 + }, + { + "epoch": 1.17, + "learning_rate": 1.8616855016688214e-05, + "loss": 1.0108, + "step": 15779 + }, + { + "epoch": 1.17, + "learning_rate": 1.861665260534085e-05, + "loss": 1.0603, + "step": 15780 + }, + { + "epoch": 1.17, + "learning_rate": 1.8616450180284515e-05, + "loss": 1.0377, + "step": 15781 + }, + { + "epoch": 1.17, + "learning_rate": 1.8616247741519532e-05, + "loss": 0.9168, + "step": 15782 + }, + { + "epoch": 1.17, + "learning_rate": 1.861604528904622e-05, + "loss": 0.9056, + "step": 15783 + }, + { + "epoch": 1.17, + "learning_rate": 1.8615842822864906e-05, + "loss": 1.0497, + "step": 15784 + }, + { + "epoch": 1.17, + "learning_rate": 1.8615640342975907e-05, + "loss": 0.9354, + "step": 15785 + }, + { + "epoch": 1.17, + "learning_rate": 1.861543784937955e-05, + "loss": 0.9821, + "step": 15786 + }, + { + "epoch": 1.17, + "learning_rate": 1.8615235342076153e-05, + "loss": 0.9163, + "step": 15787 + }, + { + "epoch": 1.17, + "learning_rate": 1.861503282106604e-05, + "loss": 0.9283, + "step": 15788 + }, + { + "epoch": 1.17, + "learning_rate": 1.8614830286349532e-05, + "loss": 1.1099, + "step": 15789 + }, + { + "epoch": 1.17, + "learning_rate": 1.8614627737926952e-05, + "loss": 1.0128, + "step": 15790 + }, + { + "epoch": 1.17, + "learning_rate": 1.8614425175798626e-05, + "loss": 0.9664, + "step": 15791 + }, + { + "epoch": 1.17, + "learning_rate": 1.8614222599964867e-05, + "loss": 0.9964, + "step": 15792 + }, + { + "epoch": 1.17, + "learning_rate": 1.861402001042601e-05, + "loss": 1.0234, + "step": 15793 + }, + { + "epoch": 1.17, + "learning_rate": 1.8613817407182366e-05, + "loss": 0.9791, + "step": 15794 + }, + { + "epoch": 1.17, + "learning_rate": 1.8613614790234264e-05, + "loss": 1.0265, + "step": 15795 + }, + { + "epoch": 1.17, + "learning_rate": 1.8613412159582024e-05, + "loss": 0.9843, + "step": 15796 + }, + { + "epoch": 1.17, + "learning_rate": 1.8613209515225965e-05, + "loss": 0.9983, + "step": 15797 + }, + { + "epoch": 1.17, + "learning_rate": 1.861300685716642e-05, + "loss": 1.0261, + "step": 15798 + }, + { + "epoch": 1.17, + "learning_rate": 1.8612804185403698e-05, + "loss": 1.0561, + "step": 15799 + }, + { + "epoch": 1.17, + "learning_rate": 1.861260149993813e-05, + "loss": 0.9777, + "step": 15800 + }, + { + "epoch": 1.17, + "learning_rate": 1.8612398800770033e-05, + "loss": 1.0396, + "step": 15801 + }, + { + "epoch": 1.17, + "learning_rate": 1.8612196087899736e-05, + "loss": 1.0438, + "step": 15802 + }, + { + "epoch": 1.17, + "learning_rate": 1.8611993361327557e-05, + "loss": 0.9503, + "step": 15803 + }, + { + "epoch": 1.17, + "learning_rate": 1.8611790621053823e-05, + "loss": 1.0896, + "step": 15804 + }, + { + "epoch": 1.17, + "learning_rate": 1.861158786707885e-05, + "loss": 1.0643, + "step": 15805 + }, + { + "epoch": 1.17, + "learning_rate": 1.8611385099402964e-05, + "loss": 1.1424, + "step": 15806 + }, + { + "epoch": 1.17, + "learning_rate": 1.861118231802649e-05, + "loss": 1.0639, + "step": 15807 + }, + { + "epoch": 1.17, + "learning_rate": 1.8610979522949747e-05, + "loss": 0.9644, + "step": 15808 + }, + { + "epoch": 1.17, + "learning_rate": 1.8610776714173057e-05, + "loss": 1.0435, + "step": 15809 + }, + { + "epoch": 1.17, + "learning_rate": 1.8610573891696744e-05, + "loss": 1.1239, + "step": 15810 + }, + { + "epoch": 1.17, + "learning_rate": 1.8610371055521134e-05, + "loss": 0.9702, + "step": 15811 + }, + { + "epoch": 1.17, + "learning_rate": 1.8610168205646544e-05, + "loss": 0.9951, + "step": 15812 + }, + { + "epoch": 1.17, + "learning_rate": 1.86099653420733e-05, + "loss": 0.9353, + "step": 15813 + }, + { + "epoch": 1.17, + "learning_rate": 1.860976246480173e-05, + "loss": 1.0638, + "step": 15814 + }, + { + "epoch": 1.17, + "learning_rate": 1.8609559573832147e-05, + "loss": 1.0033, + "step": 15815 + }, + { + "epoch": 1.17, + "learning_rate": 1.8609356669164876e-05, + "loss": 0.9944, + "step": 15816 + }, + { + "epoch": 1.17, + "learning_rate": 1.8609153750800243e-05, + "loss": 1.0433, + "step": 15817 + }, + { + "epoch": 1.17, + "learning_rate": 1.8608950818738568e-05, + "loss": 0.9948, + "step": 15818 + }, + { + "epoch": 1.17, + "learning_rate": 1.860874787298018e-05, + "loss": 0.953, + "step": 15819 + }, + { + "epoch": 1.17, + "learning_rate": 1.8608544913525396e-05, + "loss": 0.9636, + "step": 15820 + }, + { + "epoch": 1.17, + "learning_rate": 1.8608341940374536e-05, + "loss": 0.9385, + "step": 15821 + }, + { + "epoch": 1.17, + "learning_rate": 1.860813895352793e-05, + "loss": 0.907, + "step": 15822 + }, + { + "epoch": 1.17, + "learning_rate": 1.8607935952985897e-05, + "loss": 0.9904, + "step": 15823 + }, + { + "epoch": 1.17, + "learning_rate": 1.8607732938748762e-05, + "loss": 1.0571, + "step": 15824 + }, + { + "epoch": 1.17, + "learning_rate": 1.860752991081685e-05, + "loss": 0.9953, + "step": 15825 + }, + { + "epoch": 1.17, + "learning_rate": 1.8607326869190477e-05, + "loss": 1.0921, + "step": 15826 + }, + { + "epoch": 1.17, + "learning_rate": 1.860712381386997e-05, + "loss": 1.1223, + "step": 15827 + }, + { + "epoch": 1.17, + "learning_rate": 1.8606920744855652e-05, + "loss": 1.0654, + "step": 15828 + }, + { + "epoch": 1.17, + "learning_rate": 1.860671766214785e-05, + "loss": 0.9416, + "step": 15829 + }, + { + "epoch": 1.17, + "learning_rate": 1.860651456574688e-05, + "loss": 1.0584, + "step": 15830 + }, + { + "epoch": 1.17, + "learning_rate": 1.860631145565307e-05, + "loss": 0.97, + "step": 15831 + }, + { + "epoch": 1.17, + "learning_rate": 1.8606108331866743e-05, + "loss": 1.1764, + "step": 15832 + }, + { + "epoch": 1.17, + "learning_rate": 1.8605905194388215e-05, + "loss": 1.0512, + "step": 15833 + }, + { + "epoch": 1.17, + "learning_rate": 1.860570204321782e-05, + "loss": 1.1283, + "step": 15834 + }, + { + "epoch": 1.17, + "learning_rate": 1.8605498878355878e-05, + "loss": 0.9189, + "step": 15835 + }, + { + "epoch": 1.17, + "learning_rate": 1.8605295699802703e-05, + "loss": 1.0521, + "step": 15836 + }, + { + "epoch": 1.17, + "learning_rate": 1.8605092507558633e-05, + "loss": 1.0659, + "step": 15837 + }, + { + "epoch": 1.17, + "learning_rate": 1.860488930162398e-05, + "loss": 1.101, + "step": 15838 + }, + { + "epoch": 1.17, + "learning_rate": 1.860468608199907e-05, + "loss": 1.0019, + "step": 15839 + }, + { + "epoch": 1.17, + "learning_rate": 1.860448284868423e-05, + "loss": 1.0014, + "step": 15840 + }, + { + "epoch": 1.17, + "learning_rate": 1.860427960167978e-05, + "loss": 1.0162, + "step": 15841 + }, + { + "epoch": 1.17, + "learning_rate": 1.8604076340986047e-05, + "loss": 0.9624, + "step": 15842 + }, + { + "epoch": 1.17, + "learning_rate": 1.8603873066603352e-05, + "loss": 0.9921, + "step": 15843 + }, + { + "epoch": 1.17, + "learning_rate": 1.8603669778532015e-05, + "loss": 1.0215, + "step": 15844 + }, + { + "epoch": 1.17, + "learning_rate": 1.8603466476772362e-05, + "loss": 1.0893, + "step": 15845 + }, + { + "epoch": 1.17, + "learning_rate": 1.8603263161324717e-05, + "loss": 0.9136, + "step": 15846 + }, + { + "epoch": 1.17, + "learning_rate": 1.8603059832189406e-05, + "loss": 1.1085, + "step": 15847 + }, + { + "epoch": 1.17, + "learning_rate": 1.8602856489366747e-05, + "loss": 0.9561, + "step": 15848 + }, + { + "epoch": 1.17, + "learning_rate": 1.860265313285707e-05, + "loss": 0.9367, + "step": 15849 + }, + { + "epoch": 1.17, + "learning_rate": 1.8602449762660694e-05, + "loss": 1.0452, + "step": 15850 + }, + { + "epoch": 1.17, + "learning_rate": 1.860224637877794e-05, + "loss": 0.9592, + "step": 15851 + }, + { + "epoch": 1.17, + "learning_rate": 1.860204298120914e-05, + "loss": 1.0681, + "step": 15852 + }, + { + "epoch": 1.17, + "learning_rate": 1.8601839569954608e-05, + "loss": 0.8756, + "step": 15853 + }, + { + "epoch": 1.17, + "learning_rate": 1.8601636145014673e-05, + "loss": 1.1341, + "step": 15854 + }, + { + "epoch": 1.17, + "learning_rate": 1.860143270638966e-05, + "loss": 0.9946, + "step": 15855 + }, + { + "epoch": 1.17, + "learning_rate": 1.860122925407989e-05, + "loss": 1.1077, + "step": 15856 + }, + { + "epoch": 1.17, + "learning_rate": 1.8601025788085687e-05, + "loss": 1.0692, + "step": 15857 + }, + { + "epoch": 1.17, + "learning_rate": 1.8600822308407376e-05, + "loss": 1.1252, + "step": 15858 + }, + { + "epoch": 1.17, + "learning_rate": 1.860061881504528e-05, + "loss": 1.081, + "step": 15859 + }, + { + "epoch": 1.17, + "learning_rate": 1.860041530799972e-05, + "loss": 1.033, + "step": 15860 + }, + { + "epoch": 1.17, + "learning_rate": 1.8600211787271024e-05, + "loss": 0.9679, + "step": 15861 + }, + { + "epoch": 1.17, + "learning_rate": 1.8600008252859514e-05, + "loss": 0.9692, + "step": 15862 + }, + { + "epoch": 1.17, + "learning_rate": 1.8599804704765512e-05, + "loss": 0.9375, + "step": 15863 + }, + { + "epoch": 1.17, + "learning_rate": 1.8599601142989346e-05, + "loss": 1.0368, + "step": 15864 + }, + { + "epoch": 1.17, + "learning_rate": 1.8599397567531338e-05, + "loss": 1.0235, + "step": 15865 + }, + { + "epoch": 1.17, + "learning_rate": 1.859919397839181e-05, + "loss": 0.9606, + "step": 15866 + }, + { + "epoch": 1.17, + "learning_rate": 1.859899037557109e-05, + "loss": 1.0161, + "step": 15867 + }, + { + "epoch": 1.17, + "learning_rate": 1.8598786759069498e-05, + "loss": 1.0158, + "step": 15868 + }, + { + "epoch": 1.17, + "learning_rate": 1.8598583128887358e-05, + "loss": 0.962, + "step": 15869 + }, + { + "epoch": 1.17, + "learning_rate": 1.8598379485024995e-05, + "loss": 1.1522, + "step": 15870 + }, + { + "epoch": 1.17, + "learning_rate": 1.8598175827482735e-05, + "loss": 1.0029, + "step": 15871 + }, + { + "epoch": 1.17, + "learning_rate": 1.85979721562609e-05, + "loss": 1.025, + "step": 15872 + }, + { + "epoch": 1.17, + "learning_rate": 1.8597768471359814e-05, + "loss": 1.0038, + "step": 15873 + }, + { + "epoch": 1.17, + "learning_rate": 1.85975647727798e-05, + "loss": 0.9517, + "step": 15874 + }, + { + "epoch": 1.17, + "learning_rate": 1.8597361060521188e-05, + "loss": 1.1324, + "step": 15875 + }, + { + "epoch": 1.17, + "learning_rate": 1.8597157334584294e-05, + "loss": 0.9634, + "step": 15876 + }, + { + "epoch": 1.17, + "learning_rate": 1.8596953594969444e-05, + "loss": 1.1269, + "step": 15877 + }, + { + "epoch": 1.17, + "learning_rate": 1.8596749841676963e-05, + "loss": 0.9626, + "step": 15878 + }, + { + "epoch": 1.17, + "learning_rate": 1.859654607470718e-05, + "loss": 0.9961, + "step": 15879 + }, + { + "epoch": 1.17, + "learning_rate": 1.8596342294060415e-05, + "loss": 1.0424, + "step": 15880 + }, + { + "epoch": 1.17, + "learning_rate": 1.859613849973699e-05, + "loss": 1.0315, + "step": 15881 + }, + { + "epoch": 1.17, + "learning_rate": 1.8595934691737234e-05, + "loss": 1.0119, + "step": 15882 + }, + { + "epoch": 1.17, + "learning_rate": 1.859573087006147e-05, + "loss": 0.9766, + "step": 15883 + }, + { + "epoch": 1.17, + "learning_rate": 1.859552703471002e-05, + "loss": 0.9693, + "step": 15884 + }, + { + "epoch": 1.17, + "learning_rate": 1.8595323185683206e-05, + "loss": 0.9768, + "step": 15885 + }, + { + "epoch": 1.17, + "learning_rate": 1.859511932298136e-05, + "loss": 1.0139, + "step": 15886 + }, + { + "epoch": 1.17, + "learning_rate": 1.85949154466048e-05, + "loss": 1.1051, + "step": 15887 + }, + { + "epoch": 1.17, + "learning_rate": 1.8594711556553854e-05, + "loss": 0.9692, + "step": 15888 + }, + { + "epoch": 1.17, + "learning_rate": 1.859450765282884e-05, + "loss": 1.0616, + "step": 15889 + }, + { + "epoch": 1.17, + "learning_rate": 1.8594303735430093e-05, + "loss": 0.9947, + "step": 15890 + }, + { + "epoch": 1.17, + "learning_rate": 1.859409980435793e-05, + "loss": 1.0154, + "step": 15891 + }, + { + "epoch": 1.17, + "learning_rate": 1.859389585961268e-05, + "loss": 1.0143, + "step": 15892 + }, + { + "epoch": 1.17, + "learning_rate": 1.859369190119466e-05, + "loss": 0.9702, + "step": 15893 + }, + { + "epoch": 1.17, + "learning_rate": 1.85934879291042e-05, + "loss": 1.0965, + "step": 15894 + }, + { + "epoch": 1.17, + "learning_rate": 1.8593283943341624e-05, + "loss": 1.033, + "step": 15895 + }, + { + "epoch": 1.17, + "learning_rate": 1.8593079943907258e-05, + "loss": 0.9377, + "step": 15896 + }, + { + "epoch": 1.17, + "learning_rate": 1.859287593080142e-05, + "loss": 0.9607, + "step": 15897 + }, + { + "epoch": 1.17, + "learning_rate": 1.8592671904024448e-05, + "loss": 1.0418, + "step": 15898 + }, + { + "epoch": 1.17, + "learning_rate": 1.859246786357665e-05, + "loss": 1.002, + "step": 15899 + }, + { + "epoch": 1.17, + "learning_rate": 1.8592263809458363e-05, + "loss": 0.9953, + "step": 15900 + }, + { + "epoch": 1.17, + "learning_rate": 1.8592059741669905e-05, + "loss": 1.0932, + "step": 15901 + }, + { + "epoch": 1.17, + "learning_rate": 1.8591855660211605e-05, + "loss": 1.0505, + "step": 15902 + }, + { + "epoch": 1.18, + "learning_rate": 1.859165156508378e-05, + "loss": 1.0218, + "step": 15903 + }, + { + "epoch": 1.18, + "learning_rate": 1.8591447456286767e-05, + "loss": 0.9633, + "step": 15904 + }, + { + "epoch": 1.18, + "learning_rate": 1.8591243333820882e-05, + "loss": 1.0191, + "step": 15905 + }, + { + "epoch": 1.18, + "learning_rate": 1.859103919768645e-05, + "loss": 0.9222, + "step": 15906 + }, + { + "epoch": 1.18, + "learning_rate": 1.8590835047883796e-05, + "loss": 1.087, + "step": 15907 + }, + { + "epoch": 1.18, + "learning_rate": 1.859063088441325e-05, + "loss": 1.0454, + "step": 15908 + }, + { + "epoch": 1.18, + "learning_rate": 1.8590426707275132e-05, + "loss": 0.9438, + "step": 15909 + }, + { + "epoch": 1.18, + "learning_rate": 1.859022251646977e-05, + "loss": 0.9215, + "step": 15910 + }, + { + "epoch": 1.18, + "learning_rate": 1.8590018311997484e-05, + "loss": 1.0249, + "step": 15911 + }, + { + "epoch": 1.18, + "learning_rate": 1.8589814093858604e-05, + "loss": 1.0478, + "step": 15912 + }, + { + "epoch": 1.18, + "learning_rate": 1.8589609862053454e-05, + "loss": 0.9481, + "step": 15913 + }, + { + "epoch": 1.18, + "learning_rate": 1.8589405616582352e-05, + "loss": 1.0181, + "step": 15914 + }, + { + "epoch": 1.18, + "learning_rate": 1.858920135744563e-05, + "loss": 1.0146, + "step": 15915 + }, + { + "epoch": 1.18, + "learning_rate": 1.8588997084643616e-05, + "loss": 1.0951, + "step": 15916 + }, + { + "epoch": 1.18, + "learning_rate": 1.8588792798176628e-05, + "loss": 0.9737, + "step": 15917 + }, + { + "epoch": 1.18, + "learning_rate": 1.8588588498044992e-05, + "loss": 0.9753, + "step": 15918 + }, + { + "epoch": 1.18, + "learning_rate": 1.858838418424904e-05, + "loss": 1.0422, + "step": 15919 + }, + { + "epoch": 1.18, + "learning_rate": 1.8588179856789085e-05, + "loss": 0.9662, + "step": 15920 + }, + { + "epoch": 1.18, + "learning_rate": 1.858797551566546e-05, + "loss": 1.0827, + "step": 15921 + }, + { + "epoch": 1.18, + "learning_rate": 1.8587771160878488e-05, + "loss": 0.9086, + "step": 15922 + }, + { + "epoch": 1.18, + "learning_rate": 1.85875667924285e-05, + "loss": 1.0592, + "step": 15923 + }, + { + "epoch": 1.18, + "learning_rate": 1.858736241031581e-05, + "loss": 1.0458, + "step": 15924 + }, + { + "epoch": 1.18, + "learning_rate": 1.8587158014540753e-05, + "loss": 0.9841, + "step": 15925 + }, + { + "epoch": 1.18, + "learning_rate": 1.8586953605103653e-05, + "loss": 1.0119, + "step": 15926 + }, + { + "epoch": 1.18, + "learning_rate": 1.858674918200483e-05, + "loss": 0.9599, + "step": 15927 + }, + { + "epoch": 1.18, + "learning_rate": 1.858654474524461e-05, + "loss": 0.9496, + "step": 15928 + }, + { + "epoch": 1.18, + "learning_rate": 1.858634029482332e-05, + "loss": 1.0535, + "step": 15929 + }, + { + "epoch": 1.18, + "learning_rate": 1.858613583074129e-05, + "loss": 1.062, + "step": 15930 + }, + { + "epoch": 1.18, + "learning_rate": 1.8585931352998835e-05, + "loss": 1.106, + "step": 15931 + }, + { + "epoch": 1.18, + "learning_rate": 1.8585726861596288e-05, + "loss": 1.1197, + "step": 15932 + }, + { + "epoch": 1.18, + "learning_rate": 1.8585522356533973e-05, + "loss": 1.0819, + "step": 15933 + }, + { + "epoch": 1.18, + "learning_rate": 1.8585317837812217e-05, + "loss": 1.0215, + "step": 15934 + }, + { + "epoch": 1.18, + "learning_rate": 1.858511330543134e-05, + "loss": 1.0032, + "step": 15935 + }, + { + "epoch": 1.18, + "learning_rate": 1.8584908759391672e-05, + "loss": 1.1385, + "step": 15936 + }, + { + "epoch": 1.18, + "learning_rate": 1.8584704199693536e-05, + "loss": 1.0501, + "step": 15937 + }, + { + "epoch": 1.18, + "learning_rate": 1.858449962633726e-05, + "loss": 0.9914, + "step": 15938 + }, + { + "epoch": 1.18, + "learning_rate": 1.8584295039323166e-05, + "loss": 0.8982, + "step": 15939 + }, + { + "epoch": 1.18, + "learning_rate": 1.8584090438651583e-05, + "loss": 1.085, + "step": 15940 + }, + { + "epoch": 1.18, + "learning_rate": 1.8583885824322838e-05, + "loss": 1.1249, + "step": 15941 + }, + { + "epoch": 1.18, + "learning_rate": 1.8583681196337247e-05, + "loss": 1.0423, + "step": 15942 + }, + { + "epoch": 1.18, + "learning_rate": 1.8583476554695145e-05, + "loss": 1.0984, + "step": 15943 + }, + { + "epoch": 1.18, + "learning_rate": 1.8583271899396856e-05, + "loss": 0.9763, + "step": 15944 + }, + { + "epoch": 1.18, + "learning_rate": 1.85830672304427e-05, + "loss": 0.9663, + "step": 15945 + }, + { + "epoch": 1.18, + "learning_rate": 1.8582862547833013e-05, + "loss": 1.0158, + "step": 15946 + }, + { + "epoch": 1.18, + "learning_rate": 1.858265785156811e-05, + "loss": 0.971, + "step": 15947 + }, + { + "epoch": 1.18, + "learning_rate": 1.8582453141648322e-05, + "loss": 1.0159, + "step": 15948 + }, + { + "epoch": 1.18, + "learning_rate": 1.8582248418073973e-05, + "loss": 0.9524, + "step": 15949 + }, + { + "epoch": 1.18, + "learning_rate": 1.858204368084539e-05, + "loss": 1.0623, + "step": 15950 + }, + { + "epoch": 1.18, + "learning_rate": 1.85818389299629e-05, + "loss": 1.0391, + "step": 15951 + }, + { + "epoch": 1.18, + "learning_rate": 1.8581634165426824e-05, + "loss": 0.9494, + "step": 15952 + }, + { + "epoch": 1.18, + "learning_rate": 1.8581429387237494e-05, + "loss": 0.9706, + "step": 15953 + }, + { + "epoch": 1.18, + "learning_rate": 1.8581224595395234e-05, + "loss": 0.9686, + "step": 15954 + }, + { + "epoch": 1.18, + "learning_rate": 1.8581019789900364e-05, + "loss": 1.1372, + "step": 15955 + }, + { + "epoch": 1.18, + "learning_rate": 1.8580814970753217e-05, + "loss": 1.0144, + "step": 15956 + }, + { + "epoch": 1.18, + "learning_rate": 1.8580610137954115e-05, + "loss": 0.92, + "step": 15957 + }, + { + "epoch": 1.18, + "learning_rate": 1.8580405291503386e-05, + "loss": 1.0434, + "step": 15958 + }, + { + "epoch": 1.18, + "learning_rate": 1.8580200431401353e-05, + "loss": 1.0034, + "step": 15959 + }, + { + "epoch": 1.18, + "learning_rate": 1.8579995557648347e-05, + "loss": 1.0048, + "step": 15960 + }, + { + "epoch": 1.18, + "learning_rate": 1.857979067024469e-05, + "loss": 1.017, + "step": 15961 + }, + { + "epoch": 1.18, + "learning_rate": 1.8579585769190704e-05, + "loss": 1.0692, + "step": 15962 + }, + { + "epoch": 1.18, + "learning_rate": 1.8579380854486724e-05, + "loss": 1.1144, + "step": 15963 + }, + { + "epoch": 1.18, + "learning_rate": 1.8579175926133073e-05, + "loss": 0.9247, + "step": 15964 + }, + { + "epoch": 1.18, + "learning_rate": 1.8578970984130076e-05, + "loss": 1.0224, + "step": 15965 + }, + { + "epoch": 1.18, + "learning_rate": 1.8578766028478057e-05, + "loss": 0.9703, + "step": 15966 + }, + { + "epoch": 1.18, + "learning_rate": 1.8578561059177342e-05, + "loss": 1.1084, + "step": 15967 + }, + { + "epoch": 1.18, + "learning_rate": 1.857835607622826e-05, + "loss": 0.9927, + "step": 15968 + }, + { + "epoch": 1.18, + "learning_rate": 1.857815107963114e-05, + "loss": 0.9023, + "step": 15969 + }, + { + "epoch": 1.18, + "learning_rate": 1.8577946069386303e-05, + "loss": 0.9972, + "step": 15970 + }, + { + "epoch": 1.18, + "learning_rate": 1.8577741045494074e-05, + "loss": 1.0477, + "step": 15971 + }, + { + "epoch": 1.18, + "learning_rate": 1.8577536007954783e-05, + "loss": 1.0468, + "step": 15972 + }, + { + "epoch": 1.18, + "learning_rate": 1.8577330956768757e-05, + "loss": 1.036, + "step": 15973 + }, + { + "epoch": 1.18, + "learning_rate": 1.8577125891936316e-05, + "loss": 0.9326, + "step": 15974 + }, + { + "epoch": 1.18, + "learning_rate": 1.8576920813457794e-05, + "loss": 1.0385, + "step": 15975 + }, + { + "epoch": 1.18, + "learning_rate": 1.857671572133351e-05, + "loss": 1.0472, + "step": 15976 + }, + { + "epoch": 1.18, + "learning_rate": 1.8576510615563798e-05, + "loss": 1.071, + "step": 15977 + }, + { + "epoch": 1.18, + "learning_rate": 1.8576305496148977e-05, + "loss": 1.0526, + "step": 15978 + }, + { + "epoch": 1.18, + "learning_rate": 1.857610036308938e-05, + "loss": 0.9843, + "step": 15979 + }, + { + "epoch": 1.18, + "learning_rate": 1.857589521638533e-05, + "loss": 0.8904, + "step": 15980 + }, + { + "epoch": 1.18, + "learning_rate": 1.8575690056037147e-05, + "loss": 1.0069, + "step": 15981 + }, + { + "epoch": 1.18, + "learning_rate": 1.857548488204517e-05, + "loss": 1.0019, + "step": 15982 + }, + { + "epoch": 1.18, + "learning_rate": 1.8575279694409717e-05, + "loss": 0.9835, + "step": 15983 + }, + { + "epoch": 1.18, + "learning_rate": 1.8575074493131116e-05, + "loss": 1.045, + "step": 15984 + }, + { + "epoch": 1.18, + "learning_rate": 1.85748692782097e-05, + "loss": 1.0831, + "step": 15985 + }, + { + "epoch": 1.18, + "learning_rate": 1.8574664049645785e-05, + "loss": 0.9759, + "step": 15986 + }, + { + "epoch": 1.18, + "learning_rate": 1.85744588074397e-05, + "loss": 0.9643, + "step": 15987 + }, + { + "epoch": 1.18, + "learning_rate": 1.857425355159178e-05, + "loss": 1.0655, + "step": 15988 + }, + { + "epoch": 1.18, + "learning_rate": 1.857404828210234e-05, + "loss": 1.0341, + "step": 15989 + }, + { + "epoch": 1.18, + "learning_rate": 1.8573842998971715e-05, + "loss": 0.8995, + "step": 15990 + }, + { + "epoch": 1.18, + "learning_rate": 1.8573637702200226e-05, + "loss": 1.0241, + "step": 15991 + }, + { + "epoch": 1.18, + "learning_rate": 1.8573432391788203e-05, + "loss": 1.0178, + "step": 15992 + }, + { + "epoch": 1.18, + "learning_rate": 1.8573227067735972e-05, + "loss": 1.1213, + "step": 15993 + }, + { + "epoch": 1.18, + "learning_rate": 1.8573021730043857e-05, + "loss": 0.9664, + "step": 15994 + }, + { + "epoch": 1.18, + "learning_rate": 1.857281637871219e-05, + "loss": 1.0264, + "step": 15995 + }, + { + "epoch": 1.18, + "learning_rate": 1.8572611013741293e-05, + "loss": 1.1022, + "step": 15996 + }, + { + "epoch": 1.18, + "learning_rate": 1.8572405635131496e-05, + "loss": 1.0278, + "step": 15997 + }, + { + "epoch": 1.18, + "learning_rate": 1.857220024288312e-05, + "loss": 0.9916, + "step": 15998 + }, + { + "epoch": 1.18, + "learning_rate": 1.8571994836996503e-05, + "loss": 0.9768, + "step": 15999 + }, + { + "epoch": 1.18, + "learning_rate": 1.8571789417471962e-05, + "loss": 1.0251, + "step": 16000 + }, + { + "epoch": 1.18, + "learning_rate": 1.8571583984309826e-05, + "loss": 0.988, + "step": 16001 + }, + { + "epoch": 1.18, + "learning_rate": 1.857137853751042e-05, + "loss": 0.9422, + "step": 16002 + }, + { + "epoch": 1.18, + "learning_rate": 1.8571173077074078e-05, + "loss": 1.0325, + "step": 16003 + }, + { + "epoch": 1.18, + "learning_rate": 1.857096760300112e-05, + "loss": 0.9839, + "step": 16004 + }, + { + "epoch": 1.18, + "learning_rate": 1.8570762115291878e-05, + "loss": 1.0635, + "step": 16005 + }, + { + "epoch": 1.18, + "learning_rate": 1.8570556613946673e-05, + "loss": 1.1169, + "step": 16006 + }, + { + "epoch": 1.18, + "learning_rate": 1.857035109896584e-05, + "loss": 0.8586, + "step": 16007 + }, + { + "epoch": 1.18, + "learning_rate": 1.8570145570349693e-05, + "loss": 1.0227, + "step": 16008 + }, + { + "epoch": 1.18, + "learning_rate": 1.8569940028098572e-05, + "loss": 0.9557, + "step": 16009 + }, + { + "epoch": 1.18, + "learning_rate": 1.8569734472212797e-05, + "loss": 1.0407, + "step": 16010 + }, + { + "epoch": 1.18, + "learning_rate": 1.8569528902692697e-05, + "loss": 0.9716, + "step": 16011 + }, + { + "epoch": 1.18, + "learning_rate": 1.85693233195386e-05, + "loss": 0.9776, + "step": 16012 + }, + { + "epoch": 1.18, + "learning_rate": 1.8569117722750833e-05, + "loss": 1.07, + "step": 16013 + }, + { + "epoch": 1.18, + "learning_rate": 1.8568912112329722e-05, + "loss": 1.0817, + "step": 16014 + }, + { + "epoch": 1.18, + "learning_rate": 1.8568706488275595e-05, + "loss": 1.0497, + "step": 16015 + }, + { + "epoch": 1.18, + "learning_rate": 1.8568500850588778e-05, + "loss": 1.0075, + "step": 16016 + }, + { + "epoch": 1.18, + "learning_rate": 1.8568295199269596e-05, + "loss": 1.0839, + "step": 16017 + }, + { + "epoch": 1.18, + "learning_rate": 1.8568089534318384e-05, + "loss": 1.0435, + "step": 16018 + }, + { + "epoch": 1.18, + "learning_rate": 1.856788385573546e-05, + "loss": 0.9551, + "step": 16019 + }, + { + "epoch": 1.18, + "learning_rate": 1.8567678163521155e-05, + "loss": 1.0509, + "step": 16020 + }, + { + "epoch": 1.18, + "learning_rate": 1.8567472457675803e-05, + "loss": 1.0132, + "step": 16021 + }, + { + "epoch": 1.18, + "learning_rate": 1.8567266738199718e-05, + "loss": 0.9128, + "step": 16022 + }, + { + "epoch": 1.18, + "learning_rate": 1.8567061005093237e-05, + "loss": 1.0879, + "step": 16023 + }, + { + "epoch": 1.18, + "learning_rate": 1.856685525835668e-05, + "loss": 0.8663, + "step": 16024 + }, + { + "epoch": 1.18, + "learning_rate": 1.8566649497990385e-05, + "loss": 0.9725, + "step": 16025 + }, + { + "epoch": 1.18, + "learning_rate": 1.8566443723994673e-05, + "loss": 1.0133, + "step": 16026 + }, + { + "epoch": 1.18, + "learning_rate": 1.8566237936369868e-05, + "loss": 1.0507, + "step": 16027 + }, + { + "epoch": 1.18, + "learning_rate": 1.8566032135116306e-05, + "loss": 1.1074, + "step": 16028 + }, + { + "epoch": 1.18, + "learning_rate": 1.8565826320234306e-05, + "loss": 0.9072, + "step": 16029 + }, + { + "epoch": 1.18, + "learning_rate": 1.85656204917242e-05, + "loss": 1.0155, + "step": 16030 + }, + { + "epoch": 1.18, + "learning_rate": 1.856541464958631e-05, + "loss": 0.8546, + "step": 16031 + }, + { + "epoch": 1.18, + "learning_rate": 1.8565208793820974e-05, + "loss": 0.9895, + "step": 16032 + }, + { + "epoch": 1.18, + "learning_rate": 1.856500292442851e-05, + "loss": 0.9579, + "step": 16033 + }, + { + "epoch": 1.18, + "learning_rate": 1.856479704140925e-05, + "loss": 0.9685, + "step": 16034 + }, + { + "epoch": 1.18, + "learning_rate": 1.8564591144763518e-05, + "loss": 1.0311, + "step": 16035 + }, + { + "epoch": 1.18, + "learning_rate": 1.8564385234491643e-05, + "loss": 1.0614, + "step": 16036 + }, + { + "epoch": 1.18, + "learning_rate": 1.8564179310593958e-05, + "loss": 1.0111, + "step": 16037 + }, + { + "epoch": 1.19, + "learning_rate": 1.8563973373070786e-05, + "loss": 0.9285, + "step": 16038 + }, + { + "epoch": 1.19, + "learning_rate": 1.8563767421922456e-05, + "loss": 0.9354, + "step": 16039 + }, + { + "epoch": 1.19, + "learning_rate": 1.8563561457149292e-05, + "loss": 1.0161, + "step": 16040 + }, + { + "epoch": 1.19, + "learning_rate": 1.8563355478751624e-05, + "loss": 1.1148, + "step": 16041 + }, + { + "epoch": 1.19, + "learning_rate": 1.856314948672978e-05, + "loss": 0.9412, + "step": 16042 + }, + { + "epoch": 1.19, + "learning_rate": 1.8562943481084084e-05, + "loss": 0.9493, + "step": 16043 + }, + { + "epoch": 1.19, + "learning_rate": 1.8562737461814877e-05, + "loss": 1.0226, + "step": 16044 + }, + { + "epoch": 1.19, + "learning_rate": 1.856253142892247e-05, + "loss": 0.9139, + "step": 16045 + }, + { + "epoch": 1.19, + "learning_rate": 1.8562325382407202e-05, + "loss": 1.0445, + "step": 16046 + }, + { + "epoch": 1.19, + "learning_rate": 1.8562119322269394e-05, + "loss": 0.9518, + "step": 16047 + }, + { + "epoch": 1.19, + "learning_rate": 1.8561913248509375e-05, + "loss": 0.923, + "step": 16048 + }, + { + "epoch": 1.19, + "learning_rate": 1.8561707161127477e-05, + "loss": 1.0718, + "step": 16049 + }, + { + "epoch": 1.19, + "learning_rate": 1.8561501060124026e-05, + "loss": 1.051, + "step": 16050 + }, + { + "epoch": 1.19, + "learning_rate": 1.856129494549935e-05, + "loss": 1.0669, + "step": 16051 + }, + { + "epoch": 1.19, + "learning_rate": 1.856108881725377e-05, + "loss": 0.9821, + "step": 16052 + }, + { + "epoch": 1.19, + "learning_rate": 1.856088267538763e-05, + "loss": 1.08, + "step": 16053 + }, + { + "epoch": 1.19, + "learning_rate": 1.856067651990124e-05, + "loss": 0.9813, + "step": 16054 + }, + { + "epoch": 1.19, + "learning_rate": 1.8560470350794942e-05, + "loss": 1.0164, + "step": 16055 + }, + { + "epoch": 1.19, + "learning_rate": 1.8560264168069054e-05, + "loss": 1.0582, + "step": 16056 + }, + { + "epoch": 1.19, + "learning_rate": 1.856005797172391e-05, + "loss": 1.0259, + "step": 16057 + }, + { + "epoch": 1.19, + "learning_rate": 1.8559851761759837e-05, + "loss": 1.0198, + "step": 16058 + }, + { + "epoch": 1.19, + "learning_rate": 1.855964553817716e-05, + "loss": 0.9824, + "step": 16059 + }, + { + "epoch": 1.19, + "learning_rate": 1.855943930097621e-05, + "loss": 1.0885, + "step": 16060 + }, + { + "epoch": 1.19, + "learning_rate": 1.8559233050157317e-05, + "loss": 1.0801, + "step": 16061 + }, + { + "epoch": 1.19, + "learning_rate": 1.8559026785720804e-05, + "loss": 1.0795, + "step": 16062 + }, + { + "epoch": 1.19, + "learning_rate": 1.8558820507667002e-05, + "loss": 1.0736, + "step": 16063 + }, + { + "epoch": 1.19, + "learning_rate": 1.855861421599624e-05, + "loss": 0.9552, + "step": 16064 + }, + { + "epoch": 1.19, + "learning_rate": 1.8558407910708843e-05, + "loss": 1.0479, + "step": 16065 + }, + { + "epoch": 1.19, + "learning_rate": 1.8558201591805146e-05, + "loss": 1.0941, + "step": 16066 + }, + { + "epoch": 1.19, + "learning_rate": 1.855799525928547e-05, + "loss": 0.8472, + "step": 16067 + }, + { + "epoch": 1.19, + "learning_rate": 1.8557788913150145e-05, + "loss": 0.9783, + "step": 16068 + }, + { + "epoch": 1.19, + "learning_rate": 1.8557582553399504e-05, + "loss": 0.999, + "step": 16069 + }, + { + "epoch": 1.19, + "learning_rate": 1.855737618003387e-05, + "loss": 0.9524, + "step": 16070 + }, + { + "epoch": 1.19, + "learning_rate": 1.8557169793053568e-05, + "loss": 0.9859, + "step": 16071 + }, + { + "epoch": 1.19, + "learning_rate": 1.8556963392458937e-05, + "loss": 0.9219, + "step": 16072 + }, + { + "epoch": 1.19, + "learning_rate": 1.8556756978250295e-05, + "loss": 1.1112, + "step": 16073 + }, + { + "epoch": 1.19, + "learning_rate": 1.855655055042798e-05, + "loss": 1.0034, + "step": 16074 + }, + { + "epoch": 1.19, + "learning_rate": 1.8556344108992314e-05, + "loss": 0.9421, + "step": 16075 + }, + { + "epoch": 1.19, + "learning_rate": 1.8556137653943626e-05, + "loss": 1.0361, + "step": 16076 + }, + { + "epoch": 1.19, + "learning_rate": 1.8555931185282245e-05, + "loss": 1.0458, + "step": 16077 + }, + { + "epoch": 1.19, + "learning_rate": 1.85557247030085e-05, + "loss": 0.9283, + "step": 16078 + }, + { + "epoch": 1.19, + "learning_rate": 1.855551820712272e-05, + "loss": 1.1103, + "step": 16079 + }, + { + "epoch": 1.19, + "learning_rate": 1.8555311697625234e-05, + "loss": 1.0337, + "step": 16080 + }, + { + "epoch": 1.19, + "learning_rate": 1.855510517451637e-05, + "loss": 1.0773, + "step": 16081 + }, + { + "epoch": 1.19, + "learning_rate": 1.855489863779645e-05, + "loss": 0.9803, + "step": 16082 + }, + { + "epoch": 1.19, + "learning_rate": 1.8554692087465817e-05, + "loss": 0.9967, + "step": 16083 + }, + { + "epoch": 1.19, + "learning_rate": 1.8554485523524785e-05, + "loss": 0.9941, + "step": 16084 + }, + { + "epoch": 1.19, + "learning_rate": 1.8554278945973693e-05, + "loss": 1.1362, + "step": 16085 + }, + { + "epoch": 1.19, + "learning_rate": 1.8554072354812865e-05, + "loss": 1.0337, + "step": 16086 + }, + { + "epoch": 1.19, + "learning_rate": 1.855386575004263e-05, + "loss": 0.9301, + "step": 16087 + }, + { + "epoch": 1.19, + "learning_rate": 1.8553659131663316e-05, + "loss": 1.0199, + "step": 16088 + }, + { + "epoch": 1.19, + "learning_rate": 1.8553452499675252e-05, + "loss": 0.9635, + "step": 16089 + }, + { + "epoch": 1.19, + "learning_rate": 1.855324585407877e-05, + "loss": 1.0531, + "step": 16090 + }, + { + "epoch": 1.19, + "learning_rate": 1.8553039194874194e-05, + "loss": 0.977, + "step": 16091 + }, + { + "epoch": 1.19, + "learning_rate": 1.8552832522061856e-05, + "loss": 0.9776, + "step": 16092 + }, + { + "epoch": 1.19, + "learning_rate": 1.8552625835642083e-05, + "loss": 1.2417, + "step": 16093 + }, + { + "epoch": 1.19, + "learning_rate": 1.8552419135615204e-05, + "loss": 1.1, + "step": 16094 + }, + { + "epoch": 1.19, + "learning_rate": 1.855221242198155e-05, + "loss": 0.9474, + "step": 16095 + }, + { + "epoch": 1.19, + "learning_rate": 1.855200569474145e-05, + "loss": 1.0373, + "step": 16096 + }, + { + "epoch": 1.19, + "learning_rate": 1.855179895389523e-05, + "loss": 1.0453, + "step": 16097 + }, + { + "epoch": 1.19, + "learning_rate": 1.8551592199443216e-05, + "loss": 1.1034, + "step": 16098 + }, + { + "epoch": 1.19, + "learning_rate": 1.8551385431385747e-05, + "loss": 1.0683, + "step": 16099 + }, + { + "epoch": 1.19, + "learning_rate": 1.8551178649723144e-05, + "loss": 1.1434, + "step": 16100 + }, + { + "epoch": 1.19, + "learning_rate": 1.855097185445574e-05, + "loss": 1.0662, + "step": 16101 + }, + { + "epoch": 1.19, + "learning_rate": 1.8550765045583857e-05, + "loss": 0.9531, + "step": 16102 + }, + { + "epoch": 1.19, + "learning_rate": 1.8550558223107835e-05, + "loss": 1.0111, + "step": 16103 + }, + { + "epoch": 1.19, + "learning_rate": 1.8550351387027992e-05, + "loss": 1.0506, + "step": 16104 + }, + { + "epoch": 1.19, + "learning_rate": 1.8550144537344665e-05, + "loss": 0.9249, + "step": 16105 + }, + { + "epoch": 1.19, + "learning_rate": 1.8549937674058182e-05, + "loss": 1.0345, + "step": 16106 + }, + { + "epoch": 1.19, + "learning_rate": 1.8549730797168868e-05, + "loss": 1.0079, + "step": 16107 + }, + { + "epoch": 1.19, + "learning_rate": 1.8549523906677052e-05, + "loss": 1.0294, + "step": 16108 + }, + { + "epoch": 1.19, + "learning_rate": 1.8549317002583073e-05, + "loss": 0.8807, + "step": 16109 + }, + { + "epoch": 1.19, + "learning_rate": 1.854911008488725e-05, + "loss": 1.0392, + "step": 16110 + }, + { + "epoch": 1.19, + "learning_rate": 1.8548903153589912e-05, + "loss": 1.0688, + "step": 16111 + }, + { + "epoch": 1.19, + "learning_rate": 1.854869620869139e-05, + "loss": 0.9902, + "step": 16112 + }, + { + "epoch": 1.19, + "learning_rate": 1.854848925019202e-05, + "loss": 0.9312, + "step": 16113 + }, + { + "epoch": 1.19, + "learning_rate": 1.8548282278092125e-05, + "loss": 0.9636, + "step": 16114 + }, + { + "epoch": 1.19, + "learning_rate": 1.8548075292392037e-05, + "loss": 1.0274, + "step": 16115 + }, + { + "epoch": 1.19, + "learning_rate": 1.854786829309208e-05, + "loss": 1.0728, + "step": 16116 + }, + { + "epoch": 1.19, + "learning_rate": 1.8547661280192585e-05, + "loss": 1.0783, + "step": 16117 + }, + { + "epoch": 1.19, + "learning_rate": 1.854745425369389e-05, + "loss": 1.0232, + "step": 16118 + }, + { + "epoch": 1.19, + "learning_rate": 1.8547247213596315e-05, + "loss": 1.0722, + "step": 16119 + }, + { + "epoch": 1.19, + "learning_rate": 1.8547040159900186e-05, + "loss": 1.0324, + "step": 16120 + }, + { + "epoch": 1.19, + "learning_rate": 1.8546833092605845e-05, + "loss": 0.9068, + "step": 16121 + }, + { + "epoch": 1.19, + "learning_rate": 1.8546626011713613e-05, + "loss": 1.0235, + "step": 16122 + }, + { + "epoch": 1.19, + "learning_rate": 1.8546418917223818e-05, + "loss": 1.0204, + "step": 16123 + }, + { + "epoch": 1.19, + "learning_rate": 1.8546211809136797e-05, + "loss": 0.9885, + "step": 16124 + }, + { + "epoch": 1.19, + "learning_rate": 1.8546004687452878e-05, + "loss": 1.044, + "step": 16125 + }, + { + "epoch": 1.19, + "learning_rate": 1.854579755217238e-05, + "loss": 1.0566, + "step": 16126 + }, + { + "epoch": 1.19, + "learning_rate": 1.8545590403295642e-05, + "loss": 1.1209, + "step": 16127 + }, + { + "epoch": 1.19, + "learning_rate": 1.8545383240822997e-05, + "loss": 1.0922, + "step": 16128 + }, + { + "epoch": 1.19, + "learning_rate": 1.8545176064754765e-05, + "loss": 1.0016, + "step": 16129 + }, + { + "epoch": 1.19, + "learning_rate": 1.854496887509128e-05, + "loss": 0.984, + "step": 16130 + }, + { + "epoch": 1.19, + "learning_rate": 1.854476167183288e-05, + "loss": 1.026, + "step": 16131 + }, + { + "epoch": 1.19, + "learning_rate": 1.8544554454979878e-05, + "loss": 0.9803, + "step": 16132 + }, + { + "epoch": 1.19, + "learning_rate": 1.854434722453261e-05, + "loss": 1.0406, + "step": 16133 + }, + { + "epoch": 1.19, + "learning_rate": 1.854413998049142e-05, + "loss": 0.9696, + "step": 16134 + }, + { + "epoch": 1.19, + "learning_rate": 1.854393272285661e-05, + "loss": 0.9351, + "step": 16135 + }, + { + "epoch": 1.19, + "learning_rate": 1.8543725451628537e-05, + "loss": 1.0337, + "step": 16136 + }, + { + "epoch": 1.19, + "learning_rate": 1.8543518166807515e-05, + "loss": 1.0171, + "step": 16137 + }, + { + "epoch": 1.19, + "learning_rate": 1.8543310868393875e-05, + "loss": 0.9092, + "step": 16138 + }, + { + "epoch": 1.19, + "learning_rate": 1.8543103556387954e-05, + "loss": 1.0658, + "step": 16139 + }, + { + "epoch": 1.19, + "learning_rate": 1.854289623079008e-05, + "loss": 1.0192, + "step": 16140 + }, + { + "epoch": 1.19, + "learning_rate": 1.854268889160057e-05, + "loss": 1.0461, + "step": 16141 + }, + { + "epoch": 1.19, + "learning_rate": 1.854248153881977e-05, + "loss": 1.07, + "step": 16142 + }, + { + "epoch": 1.19, + "learning_rate": 1.8542274172448006e-05, + "loss": 0.973, + "step": 16143 + }, + { + "epoch": 1.19, + "learning_rate": 1.8542066792485608e-05, + "loss": 0.9991, + "step": 16144 + }, + { + "epoch": 1.19, + "learning_rate": 1.8541859398932898e-05, + "loss": 1.0682, + "step": 16145 + }, + { + "epoch": 1.19, + "learning_rate": 1.8541651991790216e-05, + "loss": 0.9003, + "step": 16146 + }, + { + "epoch": 1.19, + "learning_rate": 1.8541444571057885e-05, + "loss": 1.0312, + "step": 16147 + }, + { + "epoch": 1.19, + "learning_rate": 1.854123713673624e-05, + "loss": 0.9712, + "step": 16148 + }, + { + "epoch": 1.19, + "learning_rate": 1.8541029688825607e-05, + "loss": 0.9447, + "step": 16149 + }, + { + "epoch": 1.19, + "learning_rate": 1.8540822227326317e-05, + "loss": 0.9735, + "step": 16150 + }, + { + "epoch": 1.19, + "learning_rate": 1.8540614752238703e-05, + "loss": 1.1455, + "step": 16151 + }, + { + "epoch": 1.19, + "learning_rate": 1.8540407263563092e-05, + "loss": 1.083, + "step": 16152 + }, + { + "epoch": 1.19, + "learning_rate": 1.854019976129982e-05, + "loss": 0.9398, + "step": 16153 + }, + { + "epoch": 1.19, + "learning_rate": 1.853999224544921e-05, + "loss": 0.9641, + "step": 16154 + }, + { + "epoch": 1.19, + "learning_rate": 1.853978471601159e-05, + "loss": 1.0648, + "step": 16155 + }, + { + "epoch": 1.19, + "learning_rate": 1.85395771729873e-05, + "loss": 0.9466, + "step": 16156 + }, + { + "epoch": 1.19, + "learning_rate": 1.8539369616376663e-05, + "loss": 1.0923, + "step": 16157 + }, + { + "epoch": 1.19, + "learning_rate": 1.853916204618001e-05, + "loss": 0.9142, + "step": 16158 + }, + { + "epoch": 1.19, + "learning_rate": 1.8538954462397675e-05, + "loss": 0.9777, + "step": 16159 + }, + { + "epoch": 1.19, + "learning_rate": 1.8538746865029983e-05, + "loss": 1.0081, + "step": 16160 + }, + { + "epoch": 1.19, + "learning_rate": 1.853853925407727e-05, + "loss": 1.0141, + "step": 16161 + }, + { + "epoch": 1.19, + "learning_rate": 1.853833162953986e-05, + "loss": 1.0438, + "step": 16162 + }, + { + "epoch": 1.19, + "learning_rate": 1.8538123991418086e-05, + "loss": 0.9704, + "step": 16163 + }, + { + "epoch": 1.19, + "learning_rate": 1.8537916339712282e-05, + "loss": 1.0425, + "step": 16164 + }, + { + "epoch": 1.19, + "learning_rate": 1.8537708674422773e-05, + "loss": 1.1303, + "step": 16165 + }, + { + "epoch": 1.19, + "learning_rate": 1.853750099554989e-05, + "loss": 1.0239, + "step": 16166 + }, + { + "epoch": 1.19, + "learning_rate": 1.853729330309397e-05, + "loss": 1.0746, + "step": 16167 + }, + { + "epoch": 1.19, + "learning_rate": 1.853708559705534e-05, + "loss": 1.0058, + "step": 16168 + }, + { + "epoch": 1.19, + "learning_rate": 1.853687787743432e-05, + "loss": 0.9831, + "step": 16169 + }, + { + "epoch": 1.19, + "learning_rate": 1.8536670144231255e-05, + "loss": 1.039, + "step": 16170 + }, + { + "epoch": 1.19, + "learning_rate": 1.853646239744647e-05, + "loss": 0.9913, + "step": 16171 + }, + { + "epoch": 1.19, + "learning_rate": 1.8536254637080295e-05, + "loss": 1.0432, + "step": 16172 + }, + { + "epoch": 1.2, + "learning_rate": 1.853604686313306e-05, + "loss": 1.1254, + "step": 16173 + }, + { + "epoch": 1.2, + "learning_rate": 1.8535839075605097e-05, + "loss": 1.1041, + "step": 16174 + }, + { + "epoch": 1.2, + "learning_rate": 1.8535631274496738e-05, + "loss": 0.9899, + "step": 16175 + }, + { + "epoch": 1.2, + "learning_rate": 1.853542345980831e-05, + "loss": 0.9522, + "step": 16176 + }, + { + "epoch": 1.2, + "learning_rate": 1.8535215631540145e-05, + "loss": 1.0748, + "step": 16177 + }, + { + "epoch": 1.2, + "learning_rate": 1.8535007789692575e-05, + "loss": 1.0052, + "step": 16178 + }, + { + "epoch": 1.2, + "learning_rate": 1.853479993426593e-05, + "loss": 0.9659, + "step": 16179 + }, + { + "epoch": 1.2, + "learning_rate": 1.8534592065260543e-05, + "loss": 0.9039, + "step": 16180 + }, + { + "epoch": 1.2, + "learning_rate": 1.8534384182676736e-05, + "loss": 0.9871, + "step": 16181 + }, + { + "epoch": 1.2, + "learning_rate": 1.8534176286514848e-05, + "loss": 1.0035, + "step": 16182 + }, + { + "epoch": 1.2, + "learning_rate": 1.853396837677521e-05, + "loss": 0.9945, + "step": 16183 + }, + { + "epoch": 1.2, + "learning_rate": 1.8533760453458146e-05, + "loss": 1.0641, + "step": 16184 + }, + { + "epoch": 1.2, + "learning_rate": 1.8533552516563995e-05, + "loss": 0.9635, + "step": 16185 + }, + { + "epoch": 1.2, + "learning_rate": 1.8533344566093086e-05, + "loss": 1.0144, + "step": 16186 + }, + { + "epoch": 1.2, + "learning_rate": 1.8533136602045746e-05, + "loss": 1.0501, + "step": 16187 + }, + { + "epoch": 1.2, + "learning_rate": 1.8532928624422304e-05, + "loss": 1.1231, + "step": 16188 + }, + { + "epoch": 1.2, + "learning_rate": 1.85327206332231e-05, + "loss": 1.0676, + "step": 16189 + }, + { + "epoch": 1.2, + "learning_rate": 1.8532512628448457e-05, + "loss": 1.0232, + "step": 16190 + }, + { + "epoch": 1.2, + "learning_rate": 1.853230461009871e-05, + "loss": 1.0271, + "step": 16191 + }, + { + "epoch": 1.2, + "learning_rate": 1.8532096578174184e-05, + "loss": 0.9865, + "step": 16192 + }, + { + "epoch": 1.2, + "learning_rate": 1.8531888532675217e-05, + "loss": 1.048, + "step": 16193 + }, + { + "epoch": 1.2, + "learning_rate": 1.8531680473602138e-05, + "loss": 1.0801, + "step": 16194 + }, + { + "epoch": 1.2, + "learning_rate": 1.853147240095528e-05, + "loss": 0.9828, + "step": 16195 + }, + { + "epoch": 1.2, + "learning_rate": 1.8531264314734968e-05, + "loss": 1.0027, + "step": 16196 + }, + { + "epoch": 1.2, + "learning_rate": 1.8531056214941535e-05, + "loss": 1.0234, + "step": 16197 + }, + { + "epoch": 1.2, + "learning_rate": 1.8530848101575317e-05, + "loss": 1.0557, + "step": 16198 + }, + { + "epoch": 1.2, + "learning_rate": 1.853063997463664e-05, + "loss": 1.0027, + "step": 16199 + }, + { + "epoch": 1.2, + "learning_rate": 1.853043183412584e-05, + "loss": 1.0785, + "step": 16200 + }, + { + "epoch": 1.2, + "learning_rate": 1.853022368004324e-05, + "loss": 0.9904, + "step": 16201 + }, + { + "epoch": 1.2, + "learning_rate": 1.8530015512389178e-05, + "loss": 1.0898, + "step": 16202 + }, + { + "epoch": 1.2, + "learning_rate": 1.8529807331163986e-05, + "loss": 1.0047, + "step": 16203 + }, + { + "epoch": 1.2, + "learning_rate": 1.852959913636799e-05, + "loss": 1.108, + "step": 16204 + }, + { + "epoch": 1.2, + "learning_rate": 1.8529390928001523e-05, + "loss": 1.0154, + "step": 16205 + }, + { + "epoch": 1.2, + "learning_rate": 1.852918270606492e-05, + "loss": 1.0381, + "step": 16206 + }, + { + "epoch": 1.2, + "learning_rate": 1.852897447055851e-05, + "loss": 1.0081, + "step": 16207 + }, + { + "epoch": 1.2, + "learning_rate": 1.852876622148262e-05, + "loss": 0.967, + "step": 16208 + }, + { + "epoch": 1.2, + "learning_rate": 1.8528557958837586e-05, + "loss": 1.0238, + "step": 16209 + }, + { + "epoch": 1.2, + "learning_rate": 1.8528349682623738e-05, + "loss": 1.1056, + "step": 16210 + }, + { + "epoch": 1.2, + "learning_rate": 1.8528141392841408e-05, + "loss": 1.0258, + "step": 16211 + }, + { + "epoch": 1.2, + "learning_rate": 1.8527933089490928e-05, + "loss": 0.9577, + "step": 16212 + }, + { + "epoch": 1.2, + "learning_rate": 1.852772477257263e-05, + "loss": 0.9578, + "step": 16213 + }, + { + "epoch": 1.2, + "learning_rate": 1.852751644208684e-05, + "loss": 0.9727, + "step": 16214 + }, + { + "epoch": 1.2, + "learning_rate": 1.8527308098033896e-05, + "loss": 0.9139, + "step": 16215 + }, + { + "epoch": 1.2, + "learning_rate": 1.8527099740414127e-05, + "loss": 1.0823, + "step": 16216 + }, + { + "epoch": 1.2, + "learning_rate": 1.852689136922786e-05, + "loss": 1.0946, + "step": 16217 + }, + { + "epoch": 1.2, + "learning_rate": 1.8526682984475438e-05, + "loss": 1.0599, + "step": 16218 + }, + { + "epoch": 1.2, + "learning_rate": 1.852647458615718e-05, + "loss": 0.9414, + "step": 16219 + }, + { + "epoch": 1.2, + "learning_rate": 1.8526266174273423e-05, + "loss": 0.9931, + "step": 16220 + }, + { + "epoch": 1.2, + "learning_rate": 1.85260577488245e-05, + "loss": 0.965, + "step": 16221 + }, + { + "epoch": 1.2, + "learning_rate": 1.8525849309810742e-05, + "loss": 0.8691, + "step": 16222 + }, + { + "epoch": 1.2, + "learning_rate": 1.8525640857232478e-05, + "loss": 0.9996, + "step": 16223 + }, + { + "epoch": 1.2, + "learning_rate": 1.8525432391090045e-05, + "loss": 1.0566, + "step": 16224 + }, + { + "epoch": 1.2, + "learning_rate": 1.8525223911383764e-05, + "loss": 1.0234, + "step": 16225 + }, + { + "epoch": 1.2, + "learning_rate": 1.852501541811398e-05, + "loss": 1.0171, + "step": 16226 + }, + { + "epoch": 1.2, + "learning_rate": 1.8524806911281015e-05, + "loss": 1.0051, + "step": 16227 + }, + { + "epoch": 1.2, + "learning_rate": 1.8524598390885207e-05, + "loss": 0.9495, + "step": 16228 + }, + { + "epoch": 1.2, + "learning_rate": 1.8524389856926883e-05, + "loss": 1.0756, + "step": 16229 + }, + { + "epoch": 1.2, + "learning_rate": 1.8524181309406376e-05, + "loss": 1.0076, + "step": 16230 + }, + { + "epoch": 1.2, + "learning_rate": 1.852397274832402e-05, + "loss": 0.9862, + "step": 16231 + }, + { + "epoch": 1.2, + "learning_rate": 1.8523764173680147e-05, + "loss": 0.9576, + "step": 16232 + }, + { + "epoch": 1.2, + "learning_rate": 1.8523555585475086e-05, + "loss": 1.0181, + "step": 16233 + }, + { + "epoch": 1.2, + "learning_rate": 1.8523346983709166e-05, + "loss": 1.0116, + "step": 16234 + }, + { + "epoch": 1.2, + "learning_rate": 1.852313836838273e-05, + "loss": 0.9177, + "step": 16235 + }, + { + "epoch": 1.2, + "learning_rate": 1.85229297394961e-05, + "loss": 1.0453, + "step": 16236 + }, + { + "epoch": 1.2, + "learning_rate": 1.852272109704961e-05, + "loss": 1.0114, + "step": 16237 + }, + { + "epoch": 1.2, + "learning_rate": 1.8522512441043588e-05, + "loss": 1.0035, + "step": 16238 + }, + { + "epoch": 1.2, + "learning_rate": 1.8522303771478376e-05, + "loss": 1.0985, + "step": 16239 + }, + { + "epoch": 1.2, + "learning_rate": 1.85220950883543e-05, + "loss": 1.0403, + "step": 16240 + }, + { + "epoch": 1.2, + "learning_rate": 1.8521886391671692e-05, + "loss": 1.1226, + "step": 16241 + }, + { + "epoch": 1.2, + "learning_rate": 1.8521677681430884e-05, + "loss": 1.0524, + "step": 16242 + }, + { + "epoch": 1.2, + "learning_rate": 1.852146895763221e-05, + "loss": 1.0644, + "step": 16243 + }, + { + "epoch": 1.2, + "learning_rate": 1.8521260220276e-05, + "loss": 1.0011, + "step": 16244 + }, + { + "epoch": 1.2, + "learning_rate": 1.8521051469362586e-05, + "loss": 1.1204, + "step": 16245 + }, + { + "epoch": 1.2, + "learning_rate": 1.8520842704892307e-05, + "loss": 0.9789, + "step": 16246 + }, + { + "epoch": 1.2, + "learning_rate": 1.8520633926865485e-05, + "loss": 1.0008, + "step": 16247 + }, + { + "epoch": 1.2, + "learning_rate": 1.8520425135282452e-05, + "loss": 0.9991, + "step": 16248 + }, + { + "epoch": 1.2, + "learning_rate": 1.852021633014355e-05, + "loss": 1.0683, + "step": 16249 + }, + { + "epoch": 1.2, + "learning_rate": 1.8520007511449107e-05, + "loss": 0.9817, + "step": 16250 + }, + { + "epoch": 1.2, + "learning_rate": 1.851979867919945e-05, + "loss": 1.1106, + "step": 16251 + }, + { + "epoch": 1.2, + "learning_rate": 1.8519589833394915e-05, + "loss": 0.9875, + "step": 16252 + }, + { + "epoch": 1.2, + "learning_rate": 1.8519380974035837e-05, + "loss": 1.0728, + "step": 16253 + }, + { + "epoch": 1.2, + "learning_rate": 1.851917210112254e-05, + "loss": 1.0392, + "step": 16254 + }, + { + "epoch": 1.2, + "learning_rate": 1.851896321465537e-05, + "loss": 0.9924, + "step": 16255 + }, + { + "epoch": 1.2, + "learning_rate": 1.8518754314634648e-05, + "loss": 1.0546, + "step": 16256 + }, + { + "epoch": 1.2, + "learning_rate": 1.851854540106071e-05, + "loss": 0.9665, + "step": 16257 + }, + { + "epoch": 1.2, + "learning_rate": 1.8518336473933888e-05, + "loss": 1.0855, + "step": 16258 + }, + { + "epoch": 1.2, + "learning_rate": 1.8518127533254514e-05, + "loss": 0.9839, + "step": 16259 + }, + { + "epoch": 1.2, + "learning_rate": 1.851791857902292e-05, + "loss": 1.0283, + "step": 16260 + }, + { + "epoch": 1.2, + "learning_rate": 1.8517709611239444e-05, + "loss": 0.9635, + "step": 16261 + }, + { + "epoch": 1.2, + "learning_rate": 1.851750062990441e-05, + "loss": 1.0484, + "step": 16262 + }, + { + "epoch": 1.2, + "learning_rate": 1.8517291635018156e-05, + "loss": 1.0729, + "step": 16263 + }, + { + "epoch": 1.2, + "learning_rate": 1.851708262658101e-05, + "loss": 0.9698, + "step": 16264 + }, + { + "epoch": 1.2, + "learning_rate": 1.8516873604593312e-05, + "loss": 1.1359, + "step": 16265 + }, + { + "epoch": 1.2, + "learning_rate": 1.8516664569055386e-05, + "loss": 1.0628, + "step": 16266 + }, + { + "epoch": 1.2, + "learning_rate": 1.8516455519967568e-05, + "loss": 1.1518, + "step": 16267 + }, + { + "epoch": 1.2, + "learning_rate": 1.8516246457330195e-05, + "loss": 1.0043, + "step": 16268 + }, + { + "epoch": 1.2, + "learning_rate": 1.8516037381143594e-05, + "loss": 0.9241, + "step": 16269 + }, + { + "epoch": 1.2, + "learning_rate": 1.85158282914081e-05, + "loss": 1.0083, + "step": 16270 + }, + { + "epoch": 1.2, + "learning_rate": 1.8515619188124044e-05, + "loss": 0.9878, + "step": 16271 + }, + { + "epoch": 1.2, + "learning_rate": 1.8515410071291758e-05, + "loss": 1.0559, + "step": 16272 + }, + { + "epoch": 1.2, + "learning_rate": 1.851520094091158e-05, + "loss": 0.9998, + "step": 16273 + }, + { + "epoch": 1.2, + "learning_rate": 1.8514991796983834e-05, + "loss": 0.9856, + "step": 16274 + }, + { + "epoch": 1.2, + "learning_rate": 1.8514782639508864e-05, + "loss": 1.0277, + "step": 16275 + }, + { + "epoch": 1.2, + "learning_rate": 1.851457346848699e-05, + "loss": 1.0531, + "step": 16276 + }, + { + "epoch": 1.2, + "learning_rate": 1.8514364283918558e-05, + "loss": 1.0057, + "step": 16277 + }, + { + "epoch": 1.2, + "learning_rate": 1.8514155085803892e-05, + "loss": 1.1136, + "step": 16278 + }, + { + "epoch": 1.2, + "learning_rate": 1.8513945874143327e-05, + "loss": 1.0209, + "step": 16279 + }, + { + "epoch": 1.2, + "learning_rate": 1.8513736648937196e-05, + "loss": 1.0372, + "step": 16280 + }, + { + "epoch": 1.2, + "learning_rate": 1.851352741018583e-05, + "loss": 0.943, + "step": 16281 + }, + { + "epoch": 1.2, + "learning_rate": 1.8513318157889568e-05, + "loss": 1.0159, + "step": 16282 + }, + { + "epoch": 1.2, + "learning_rate": 1.8513108892048733e-05, + "loss": 0.9636, + "step": 16283 + }, + { + "epoch": 1.2, + "learning_rate": 1.8512899612663665e-05, + "loss": 0.8921, + "step": 16284 + }, + { + "epoch": 1.2, + "learning_rate": 1.85126903197347e-05, + "loss": 0.9942, + "step": 16285 + }, + { + "epoch": 1.2, + "learning_rate": 1.851248101326216e-05, + "loss": 0.9446, + "step": 16286 + }, + { + "epoch": 1.2, + "learning_rate": 1.8512271693246387e-05, + "loss": 1.0255, + "step": 16287 + }, + { + "epoch": 1.2, + "learning_rate": 1.851206235968771e-05, + "loss": 1.0038, + "step": 16288 + }, + { + "epoch": 1.2, + "learning_rate": 1.851185301258647e-05, + "loss": 1.0318, + "step": 16289 + }, + { + "epoch": 1.2, + "learning_rate": 1.851164365194299e-05, + "loss": 1.0563, + "step": 16290 + }, + { + "epoch": 1.2, + "learning_rate": 1.8511434277757608e-05, + "loss": 1.0564, + "step": 16291 + }, + { + "epoch": 1.2, + "learning_rate": 1.8511224890030653e-05, + "loss": 1.0432, + "step": 16292 + }, + { + "epoch": 1.2, + "learning_rate": 1.851101548876246e-05, + "loss": 1.0274, + "step": 16293 + }, + { + "epoch": 1.2, + "learning_rate": 1.8510806073953365e-05, + "loss": 0.9041, + "step": 16294 + }, + { + "epoch": 1.2, + "learning_rate": 1.85105966456037e-05, + "loss": 0.9854, + "step": 16295 + }, + { + "epoch": 1.2, + "learning_rate": 1.8510387203713798e-05, + "loss": 1.0185, + "step": 16296 + }, + { + "epoch": 1.2, + "learning_rate": 1.851017774828399e-05, + "loss": 1.0789, + "step": 16297 + }, + { + "epoch": 1.2, + "learning_rate": 1.8509968279314613e-05, + "loss": 1.0337, + "step": 16298 + }, + { + "epoch": 1.2, + "learning_rate": 1.8509758796805996e-05, + "loss": 1.0884, + "step": 16299 + }, + { + "epoch": 1.2, + "learning_rate": 1.8509549300758472e-05, + "loss": 0.9779, + "step": 16300 + }, + { + "epoch": 1.2, + "learning_rate": 1.8509339791172384e-05, + "loss": 1.0869, + "step": 16301 + }, + { + "epoch": 1.2, + "learning_rate": 1.8509130268048056e-05, + "loss": 1.006, + "step": 16302 + }, + { + "epoch": 1.2, + "learning_rate": 1.850892073138582e-05, + "loss": 0.8896, + "step": 16303 + }, + { + "epoch": 1.2, + "learning_rate": 1.8508711181186014e-05, + "loss": 0.9998, + "step": 16304 + }, + { + "epoch": 1.2, + "learning_rate": 1.850850161744897e-05, + "loss": 1.0078, + "step": 16305 + }, + { + "epoch": 1.2, + "learning_rate": 1.8508292040175025e-05, + "loss": 0.9536, + "step": 16306 + }, + { + "epoch": 1.2, + "learning_rate": 1.8508082449364505e-05, + "loss": 0.928, + "step": 16307 + }, + { + "epoch": 1.2, + "learning_rate": 1.850787284501775e-05, + "loss": 0.9944, + "step": 16308 + }, + { + "epoch": 1.21, + "learning_rate": 1.850766322713509e-05, + "loss": 1.0719, + "step": 16309 + }, + { + "epoch": 1.21, + "learning_rate": 1.850745359571686e-05, + "loss": 0.9889, + "step": 16310 + }, + { + "epoch": 1.21, + "learning_rate": 1.8507243950763393e-05, + "loss": 1.004, + "step": 16311 + }, + { + "epoch": 1.21, + "learning_rate": 1.8507034292275024e-05, + "loss": 1.0675, + "step": 16312 + }, + { + "epoch": 1.21, + "learning_rate": 1.8506824620252082e-05, + "loss": 1.0971, + "step": 16313 + }, + { + "epoch": 1.21, + "learning_rate": 1.8506614934694903e-05, + "loss": 1.0159, + "step": 16314 + }, + { + "epoch": 1.21, + "learning_rate": 1.8506405235603827e-05, + "loss": 1.0882, + "step": 16315 + }, + { + "epoch": 1.21, + "learning_rate": 1.8506195522979177e-05, + "loss": 1.0067, + "step": 16316 + }, + { + "epoch": 1.21, + "learning_rate": 1.8505985796821296e-05, + "loss": 0.9892, + "step": 16317 + }, + { + "epoch": 1.21, + "learning_rate": 1.8505776057130508e-05, + "loss": 1.0519, + "step": 16318 + }, + { + "epoch": 1.21, + "learning_rate": 1.8505566303907156e-05, + "loss": 1.1219, + "step": 16319 + }, + { + "epoch": 1.21, + "learning_rate": 1.8505356537151568e-05, + "loss": 1.166, + "step": 16320 + }, + { + "epoch": 1.21, + "learning_rate": 1.850514675686408e-05, + "loss": 0.9745, + "step": 16321 + }, + { + "epoch": 1.21, + "learning_rate": 1.8504936963045024e-05, + "loss": 1.0889, + "step": 16322 + }, + { + "epoch": 1.21, + "learning_rate": 1.850472715569474e-05, + "loss": 0.9944, + "step": 16323 + }, + { + "epoch": 1.21, + "learning_rate": 1.850451733481355e-05, + "loss": 0.9241, + "step": 16324 + }, + { + "epoch": 1.21, + "learning_rate": 1.8504307500401797e-05, + "loss": 1.0003, + "step": 16325 + }, + { + "epoch": 1.21, + "learning_rate": 1.8504097652459813e-05, + "loss": 1.0696, + "step": 16326 + }, + { + "epoch": 1.21, + "learning_rate": 1.850388779098793e-05, + "loss": 0.9796, + "step": 16327 + }, + { + "epoch": 1.21, + "learning_rate": 1.8503677915986485e-05, + "loss": 1.0594, + "step": 16328 + }, + { + "epoch": 1.21, + "learning_rate": 1.850346802745581e-05, + "loss": 0.9283, + "step": 16329 + }, + { + "epoch": 1.21, + "learning_rate": 1.850325812539624e-05, + "loss": 0.9391, + "step": 16330 + }, + { + "epoch": 1.21, + "learning_rate": 1.8503048209808106e-05, + "loss": 0.9603, + "step": 16331 + }, + { + "epoch": 1.21, + "learning_rate": 1.8502838280691746e-05, + "loss": 0.9604, + "step": 16332 + }, + { + "epoch": 1.21, + "learning_rate": 1.850262833804749e-05, + "loss": 1.0366, + "step": 16333 + }, + { + "epoch": 1.21, + "learning_rate": 1.850241838187567e-05, + "loss": 1.0955, + "step": 16334 + }, + { + "epoch": 1.21, + "learning_rate": 1.850220841217663e-05, + "loss": 1.002, + "step": 16335 + }, + { + "epoch": 1.21, + "learning_rate": 1.8501998428950696e-05, + "loss": 0.9739, + "step": 16336 + }, + { + "epoch": 1.21, + "learning_rate": 1.8501788432198206e-05, + "loss": 0.9928, + "step": 16337 + }, + { + "epoch": 1.21, + "learning_rate": 1.850157842191949e-05, + "loss": 0.9734, + "step": 16338 + }, + { + "epoch": 1.21, + "learning_rate": 1.8501368398114885e-05, + "loss": 0.9318, + "step": 16339 + }, + { + "epoch": 1.21, + "learning_rate": 1.8501158360784722e-05, + "loss": 1.0041, + "step": 16340 + }, + { + "epoch": 1.21, + "learning_rate": 1.850094830992934e-05, + "loss": 1.0613, + "step": 16341 + }, + { + "epoch": 1.21, + "learning_rate": 1.8500738245549072e-05, + "loss": 0.9832, + "step": 16342 + }, + { + "epoch": 1.21, + "learning_rate": 1.8500528167644253e-05, + "loss": 0.9807, + "step": 16343 + }, + { + "epoch": 1.21, + "learning_rate": 1.850031807621521e-05, + "loss": 0.9862, + "step": 16344 + }, + { + "epoch": 1.21, + "learning_rate": 1.8500107971262287e-05, + "loss": 1.0342, + "step": 16345 + }, + { + "epoch": 1.21, + "learning_rate": 1.849989785278581e-05, + "loss": 1.0386, + "step": 16346 + }, + { + "epoch": 1.21, + "learning_rate": 1.8499687720786118e-05, + "loss": 1.048, + "step": 16347 + }, + { + "epoch": 1.21, + "learning_rate": 1.849947757526355e-05, + "loss": 0.9775, + "step": 16348 + }, + { + "epoch": 1.21, + "learning_rate": 1.8499267416218428e-05, + "loss": 0.9468, + "step": 16349 + }, + { + "epoch": 1.21, + "learning_rate": 1.8499057243651094e-05, + "loss": 1.0374, + "step": 16350 + }, + { + "epoch": 1.21, + "learning_rate": 1.8498847057561884e-05, + "loss": 1.0654, + "step": 16351 + }, + { + "epoch": 1.21, + "learning_rate": 1.849863685795113e-05, + "loss": 1.0702, + "step": 16352 + }, + { + "epoch": 1.21, + "learning_rate": 1.8498426644819166e-05, + "loss": 1.0575, + "step": 16353 + }, + { + "epoch": 1.21, + "learning_rate": 1.8498216418166324e-05, + "loss": 1.0458, + "step": 16354 + }, + { + "epoch": 1.21, + "learning_rate": 1.8498006177992943e-05, + "loss": 1.0243, + "step": 16355 + }, + { + "epoch": 1.21, + "learning_rate": 1.8497795924299358e-05, + "loss": 1.0186, + "step": 16356 + }, + { + "epoch": 1.21, + "learning_rate": 1.84975856570859e-05, + "loss": 1.0938, + "step": 16357 + }, + { + "epoch": 1.21, + "learning_rate": 1.8497375376352905e-05, + "loss": 0.9872, + "step": 16358 + }, + { + "epoch": 1.21, + "learning_rate": 1.8497165082100705e-05, + "loss": 1.0702, + "step": 16359 + }, + { + "epoch": 1.21, + "learning_rate": 1.8496954774329636e-05, + "loss": 1.0675, + "step": 16360 + }, + { + "epoch": 1.21, + "learning_rate": 1.849674445304004e-05, + "loss": 0.9033, + "step": 16361 + }, + { + "epoch": 1.21, + "learning_rate": 1.8496534118232238e-05, + "loss": 0.9458, + "step": 16362 + }, + { + "epoch": 1.21, + "learning_rate": 1.8496323769906576e-05, + "loss": 1.0164, + "step": 16363 + }, + { + "epoch": 1.21, + "learning_rate": 1.8496113408063384e-05, + "loss": 0.984, + "step": 16364 + }, + { + "epoch": 1.21, + "learning_rate": 1.8495903032702995e-05, + "loss": 0.9762, + "step": 16365 + }, + { + "epoch": 1.21, + "learning_rate": 1.849569264382575e-05, + "loss": 1.0607, + "step": 16366 + }, + { + "epoch": 1.21, + "learning_rate": 1.8495482241431976e-05, + "loss": 1.0864, + "step": 16367 + }, + { + "epoch": 1.21, + "learning_rate": 1.8495271825522013e-05, + "loss": 1.0393, + "step": 16368 + }, + { + "epoch": 1.21, + "learning_rate": 1.849506139609619e-05, + "loss": 0.9498, + "step": 16369 + }, + { + "epoch": 1.21, + "learning_rate": 1.849485095315485e-05, + "loss": 0.9747, + "step": 16370 + }, + { + "epoch": 1.21, + "learning_rate": 1.849464049669832e-05, + "loss": 1.011, + "step": 16371 + }, + { + "epoch": 1.21, + "learning_rate": 1.8494430026726945e-05, + "loss": 1.0988, + "step": 16372 + }, + { + "epoch": 1.21, + "learning_rate": 1.849421954324105e-05, + "loss": 1.0298, + "step": 16373 + }, + { + "epoch": 1.21, + "learning_rate": 1.8494009046240973e-05, + "loss": 0.9355, + "step": 16374 + }, + { + "epoch": 1.21, + "learning_rate": 1.849379853572705e-05, + "loss": 1.1255, + "step": 16375 + }, + { + "epoch": 1.21, + "learning_rate": 1.849358801169961e-05, + "loss": 1.0107, + "step": 16376 + }, + { + "epoch": 1.21, + "learning_rate": 1.8493377474158997e-05, + "loss": 0.9737, + "step": 16377 + }, + { + "epoch": 1.21, + "learning_rate": 1.8493166923105543e-05, + "loss": 0.997, + "step": 16378 + }, + { + "epoch": 1.21, + "learning_rate": 1.8492956358539584e-05, + "loss": 1.0401, + "step": 16379 + }, + { + "epoch": 1.21, + "learning_rate": 1.8492745780461446e-05, + "loss": 1.0315, + "step": 16380 + }, + { + "epoch": 1.21, + "learning_rate": 1.8492535188871476e-05, + "loss": 0.9932, + "step": 16381 + }, + { + "epoch": 1.21, + "learning_rate": 1.8492324583770003e-05, + "loss": 1.0397, + "step": 16382 + }, + { + "epoch": 1.21, + "learning_rate": 1.8492113965157364e-05, + "loss": 0.8539, + "step": 16383 + }, + { + "epoch": 1.21, + "learning_rate": 1.849190333303389e-05, + "loss": 0.9831, + "step": 16384 + }, + { + "epoch": 1.21, + "learning_rate": 1.8491692687399925e-05, + "loss": 1.076, + "step": 16385 + }, + { + "epoch": 1.21, + "learning_rate": 1.849148202825579e-05, + "loss": 0.962, + "step": 16386 + }, + { + "epoch": 1.21, + "learning_rate": 1.8491271355601837e-05, + "loss": 1.0353, + "step": 16387 + }, + { + "epoch": 1.21, + "learning_rate": 1.849106066943839e-05, + "loss": 1.0091, + "step": 16388 + }, + { + "epoch": 1.21, + "learning_rate": 1.8490849969765785e-05, + "loss": 1.0233, + "step": 16389 + }, + { + "epoch": 1.21, + "learning_rate": 1.849063925658436e-05, + "loss": 1.0599, + "step": 16390 + }, + { + "epoch": 1.21, + "learning_rate": 1.849042852989445e-05, + "loss": 0.919, + "step": 16391 + }, + { + "epoch": 1.21, + "learning_rate": 1.849021778969639e-05, + "loss": 1.0687, + "step": 16392 + }, + { + "epoch": 1.21, + "learning_rate": 1.8490007035990515e-05, + "loss": 1.0448, + "step": 16393 + }, + { + "epoch": 1.21, + "learning_rate": 1.8489796268777157e-05, + "loss": 1.1002, + "step": 16394 + }, + { + "epoch": 1.21, + "learning_rate": 1.848958548805666e-05, + "loss": 0.9676, + "step": 16395 + }, + { + "epoch": 1.21, + "learning_rate": 1.848937469382935e-05, + "loss": 0.9621, + "step": 16396 + }, + { + "epoch": 1.21, + "learning_rate": 1.8489163886095567e-05, + "loss": 0.9578, + "step": 16397 + }, + { + "epoch": 1.21, + "learning_rate": 1.848895306485565e-05, + "loss": 1.0821, + "step": 16398 + }, + { + "epoch": 1.21, + "learning_rate": 1.8488742230109923e-05, + "loss": 1.0408, + "step": 16399 + }, + { + "epoch": 1.21, + "learning_rate": 1.848853138185873e-05, + "loss": 0.9878, + "step": 16400 + }, + { + "epoch": 1.21, + "learning_rate": 1.848832052010241e-05, + "loss": 1.0053, + "step": 16401 + }, + { + "epoch": 1.21, + "learning_rate": 1.848810964484129e-05, + "loss": 0.9946, + "step": 16402 + }, + { + "epoch": 1.21, + "learning_rate": 1.848789875607571e-05, + "loss": 1.1417, + "step": 16403 + }, + { + "epoch": 1.21, + "learning_rate": 1.8487687853806006e-05, + "loss": 1.0527, + "step": 16404 + }, + { + "epoch": 1.21, + "learning_rate": 1.848747693803251e-05, + "loss": 0.9752, + "step": 16405 + }, + { + "epoch": 1.21, + "learning_rate": 1.8487266008755558e-05, + "loss": 1.0604, + "step": 16406 + }, + { + "epoch": 1.21, + "learning_rate": 1.848705506597549e-05, + "loss": 1.0022, + "step": 16407 + }, + { + "epoch": 1.21, + "learning_rate": 1.8486844109692638e-05, + "loss": 1.0388, + "step": 16408 + }, + { + "epoch": 1.21, + "learning_rate": 1.8486633139907336e-05, + "loss": 0.9972, + "step": 16409 + }, + { + "epoch": 1.21, + "learning_rate": 1.848642215661993e-05, + "loss": 0.9518, + "step": 16410 + }, + { + "epoch": 1.21, + "learning_rate": 1.848621115983074e-05, + "loss": 1.0134, + "step": 16411 + }, + { + "epoch": 1.21, + "learning_rate": 1.8486000149540108e-05, + "loss": 1.0137, + "step": 16412 + }, + { + "epoch": 1.21, + "learning_rate": 1.8485789125748375e-05, + "loss": 1.0678, + "step": 16413 + }, + { + "epoch": 1.21, + "learning_rate": 1.8485578088455875e-05, + "loss": 1.1248, + "step": 16414 + }, + { + "epoch": 1.21, + "learning_rate": 1.848536703766294e-05, + "loss": 1.1108, + "step": 16415 + }, + { + "epoch": 1.21, + "learning_rate": 1.8485155973369904e-05, + "loss": 1.0545, + "step": 16416 + }, + { + "epoch": 1.21, + "learning_rate": 1.848494489557711e-05, + "loss": 1.0741, + "step": 16417 + }, + { + "epoch": 1.21, + "learning_rate": 1.848473380428489e-05, + "loss": 1.0065, + "step": 16418 + }, + { + "epoch": 1.21, + "learning_rate": 1.848452269949358e-05, + "loss": 1.0504, + "step": 16419 + }, + { + "epoch": 1.21, + "learning_rate": 1.8484311581203514e-05, + "loss": 1.0077, + "step": 16420 + }, + { + "epoch": 1.21, + "learning_rate": 1.848410044941503e-05, + "loss": 1.0221, + "step": 16421 + }, + { + "epoch": 1.21, + "learning_rate": 1.8483889304128465e-05, + "loss": 0.9642, + "step": 16422 + }, + { + "epoch": 1.21, + "learning_rate": 1.8483678145344152e-05, + "loss": 1.099, + "step": 16423 + }, + { + "epoch": 1.21, + "learning_rate": 1.848346697306243e-05, + "loss": 1.0324, + "step": 16424 + }, + { + "epoch": 1.21, + "learning_rate": 1.8483255787283633e-05, + "loss": 0.9538, + "step": 16425 + }, + { + "epoch": 1.21, + "learning_rate": 1.8483044588008096e-05, + "loss": 1.0834, + "step": 16426 + }, + { + "epoch": 1.21, + "learning_rate": 1.8482833375236157e-05, + "loss": 1.021, + "step": 16427 + }, + { + "epoch": 1.21, + "learning_rate": 1.8482622148968154e-05, + "loss": 1.0969, + "step": 16428 + }, + { + "epoch": 1.21, + "learning_rate": 1.8482410909204415e-05, + "loss": 1.0304, + "step": 16429 + }, + { + "epoch": 1.21, + "learning_rate": 1.8482199655945287e-05, + "loss": 1.0159, + "step": 16430 + }, + { + "epoch": 1.21, + "learning_rate": 1.8481988389191098e-05, + "loss": 1.0461, + "step": 16431 + }, + { + "epoch": 1.21, + "learning_rate": 1.8481777108942186e-05, + "loss": 0.9728, + "step": 16432 + }, + { + "epoch": 1.21, + "learning_rate": 1.848156581519889e-05, + "loss": 0.9647, + "step": 16433 + }, + { + "epoch": 1.21, + "learning_rate": 1.8481354507961545e-05, + "loss": 1.0508, + "step": 16434 + }, + { + "epoch": 1.21, + "learning_rate": 1.8481143187230483e-05, + "loss": 1.0148, + "step": 16435 + }, + { + "epoch": 1.21, + "learning_rate": 1.8480931853006047e-05, + "loss": 1.0501, + "step": 16436 + }, + { + "epoch": 1.21, + "learning_rate": 1.848072050528857e-05, + "loss": 1.176, + "step": 16437 + }, + { + "epoch": 1.21, + "learning_rate": 1.848050914407838e-05, + "loss": 1.0678, + "step": 16438 + }, + { + "epoch": 1.21, + "learning_rate": 1.848029776937583e-05, + "loss": 0.9564, + "step": 16439 + }, + { + "epoch": 1.21, + "learning_rate": 1.8480086381181246e-05, + "loss": 1.1406, + "step": 16440 + }, + { + "epoch": 1.21, + "learning_rate": 1.8479874979494966e-05, + "loss": 1.0808, + "step": 16441 + }, + { + "epoch": 1.21, + "learning_rate": 1.847966356431732e-05, + "loss": 1.0796, + "step": 16442 + }, + { + "epoch": 1.21, + "learning_rate": 1.847945213564866e-05, + "loss": 1.0235, + "step": 16443 + }, + { + "epoch": 1.22, + "learning_rate": 1.847924069348931e-05, + "loss": 1.0413, + "step": 16444 + }, + { + "epoch": 1.22, + "learning_rate": 1.8479029237839604e-05, + "loss": 1.0739, + "step": 16445 + }, + { + "epoch": 1.22, + "learning_rate": 1.847881776869989e-05, + "loss": 1.092, + "step": 16446 + }, + { + "epoch": 1.22, + "learning_rate": 1.8478606286070497e-05, + "loss": 0.9478, + "step": 16447 + }, + { + "epoch": 1.22, + "learning_rate": 1.8478394789951762e-05, + "loss": 1.1044, + "step": 16448 + }, + { + "epoch": 1.22, + "learning_rate": 1.847818328034402e-05, + "loss": 1.0398, + "step": 16449 + }, + { + "epoch": 1.22, + "learning_rate": 1.8477971757247616e-05, + "loss": 1.0918, + "step": 16450 + }, + { + "epoch": 1.22, + "learning_rate": 1.8477760220662872e-05, + "loss": 1.0167, + "step": 16451 + }, + { + "epoch": 1.22, + "learning_rate": 1.8477548670590138e-05, + "loss": 1.0363, + "step": 16452 + }, + { + "epoch": 1.22, + "learning_rate": 1.8477337107029746e-05, + "loss": 1.02, + "step": 16453 + }, + { + "epoch": 1.22, + "learning_rate": 1.8477125529982028e-05, + "loss": 1.0119, + "step": 16454 + }, + { + "epoch": 1.22, + "learning_rate": 1.8476913939447326e-05, + "loss": 0.9988, + "step": 16455 + }, + { + "epoch": 1.22, + "learning_rate": 1.8476702335425976e-05, + "loss": 1.0462, + "step": 16456 + }, + { + "epoch": 1.22, + "learning_rate": 1.8476490717918313e-05, + "loss": 1.0141, + "step": 16457 + }, + { + "epoch": 1.22, + "learning_rate": 1.8476279086924676e-05, + "loss": 1.0291, + "step": 16458 + }, + { + "epoch": 1.22, + "learning_rate": 1.84760674424454e-05, + "loss": 0.94, + "step": 16459 + }, + { + "epoch": 1.22, + "learning_rate": 1.8475855784480822e-05, + "loss": 0.9267, + "step": 16460 + }, + { + "epoch": 1.22, + "learning_rate": 1.847564411303128e-05, + "loss": 1.1622, + "step": 16461 + }, + { + "epoch": 1.22, + "learning_rate": 1.8475432428097106e-05, + "loss": 1.1108, + "step": 16462 + }, + { + "epoch": 1.22, + "learning_rate": 1.8475220729678644e-05, + "loss": 0.942, + "step": 16463 + }, + { + "epoch": 1.22, + "learning_rate": 1.8475009017776223e-05, + "loss": 0.9473, + "step": 16464 + }, + { + "epoch": 1.22, + "learning_rate": 1.8474797292390186e-05, + "loss": 0.9314, + "step": 16465 + }, + { + "epoch": 1.22, + "learning_rate": 1.8474585553520866e-05, + "loss": 1.0433, + "step": 16466 + }, + { + "epoch": 1.22, + "learning_rate": 1.8474373801168607e-05, + "loss": 1.0627, + "step": 16467 + }, + { + "epoch": 1.22, + "learning_rate": 1.8474162035333735e-05, + "loss": 1.0933, + "step": 16468 + }, + { + "epoch": 1.22, + "learning_rate": 1.8473950256016596e-05, + "loss": 0.8941, + "step": 16469 + }, + { + "epoch": 1.22, + "learning_rate": 1.847373846321752e-05, + "loss": 1.0402, + "step": 16470 + }, + { + "epoch": 1.22, + "learning_rate": 1.847352665693685e-05, + "loss": 1.0115, + "step": 16471 + }, + { + "epoch": 1.22, + "learning_rate": 1.847331483717492e-05, + "loss": 0.9527, + "step": 16472 + }, + { + "epoch": 1.22, + "learning_rate": 1.847310300393207e-05, + "loss": 1.0358, + "step": 16473 + }, + { + "epoch": 1.22, + "learning_rate": 1.8472891157208627e-05, + "loss": 1.0114, + "step": 16474 + }, + { + "epoch": 1.22, + "learning_rate": 1.847267929700494e-05, + "loss": 1.026, + "step": 16475 + }, + { + "epoch": 1.22, + "learning_rate": 1.847246742332134e-05, + "loss": 0.9979, + "step": 16476 + }, + { + "epoch": 1.22, + "learning_rate": 1.8472255536158167e-05, + "loss": 1.0194, + "step": 16477 + }, + { + "epoch": 1.22, + "learning_rate": 1.847204363551576e-05, + "loss": 0.9879, + "step": 16478 + }, + { + "epoch": 1.22, + "learning_rate": 1.8471831721394444e-05, + "loss": 1.0397, + "step": 16479 + }, + { + "epoch": 1.22, + "learning_rate": 1.847161979379457e-05, + "loss": 1.0304, + "step": 16480 + }, + { + "epoch": 1.22, + "learning_rate": 1.847140785271647e-05, + "loss": 0.9464, + "step": 16481 + }, + { + "epoch": 1.22, + "learning_rate": 1.8471195898160483e-05, + "loss": 1.0115, + "step": 16482 + }, + { + "epoch": 1.22, + "learning_rate": 1.8470983930126944e-05, + "loss": 1.081, + "step": 16483 + }, + { + "epoch": 1.22, + "learning_rate": 1.8470771948616186e-05, + "loss": 0.9964, + "step": 16484 + }, + { + "epoch": 1.22, + "learning_rate": 1.8470559953628554e-05, + "loss": 1.0592, + "step": 16485 + }, + { + "epoch": 1.22, + "learning_rate": 1.8470347945164384e-05, + "loss": 1.1076, + "step": 16486 + }, + { + "epoch": 1.22, + "learning_rate": 1.8470135923224012e-05, + "loss": 1.1848, + "step": 16487 + }, + { + "epoch": 1.22, + "learning_rate": 1.8469923887807772e-05, + "loss": 1.0639, + "step": 16488 + }, + { + "epoch": 1.22, + "learning_rate": 1.8469711838916007e-05, + "loss": 1.0313, + "step": 16489 + }, + { + "epoch": 1.22, + "learning_rate": 1.8469499776549053e-05, + "loss": 1.0052, + "step": 16490 + }, + { + "epoch": 1.22, + "learning_rate": 1.846928770070724e-05, + "loss": 1.0324, + "step": 16491 + }, + { + "epoch": 1.22, + "learning_rate": 1.8469075611390916e-05, + "loss": 0.9544, + "step": 16492 + }, + { + "epoch": 1.22, + "learning_rate": 1.8468863508600412e-05, + "loss": 0.9738, + "step": 16493 + }, + { + "epoch": 1.22, + "learning_rate": 1.8468651392336067e-05, + "loss": 1.0258, + "step": 16494 + }, + { + "epoch": 1.22, + "learning_rate": 1.8468439262598223e-05, + "loss": 1.0295, + "step": 16495 + }, + { + "epoch": 1.22, + "learning_rate": 1.846822711938721e-05, + "loss": 1.0445, + "step": 16496 + }, + { + "epoch": 1.22, + "learning_rate": 1.8468014962703365e-05, + "loss": 1.1498, + "step": 16497 + }, + { + "epoch": 1.22, + "learning_rate": 1.8467802792547036e-05, + "loss": 0.9567, + "step": 16498 + }, + { + "epoch": 1.22, + "learning_rate": 1.846759060891855e-05, + "loss": 1.0324, + "step": 16499 + }, + { + "epoch": 1.22, + "learning_rate": 1.846737841181825e-05, + "loss": 0.9706, + "step": 16500 + }, + { + "epoch": 1.22, + "learning_rate": 1.8467166201246472e-05, + "loss": 0.9422, + "step": 16501 + }, + { + "epoch": 1.22, + "learning_rate": 1.8466953977203554e-05, + "loss": 1.0204, + "step": 16502 + }, + { + "epoch": 1.22, + "learning_rate": 1.8466741739689837e-05, + "loss": 1.074, + "step": 16503 + }, + { + "epoch": 1.22, + "learning_rate": 1.8466529488705647e-05, + "loss": 0.9991, + "step": 16504 + }, + { + "epoch": 1.22, + "learning_rate": 1.8466317224251335e-05, + "loss": 1.0308, + "step": 16505 + }, + { + "epoch": 1.22, + "learning_rate": 1.8466104946327232e-05, + "loss": 0.973, + "step": 16506 + }, + { + "epoch": 1.22, + "learning_rate": 1.8465892654933678e-05, + "loss": 0.9854, + "step": 16507 + }, + { + "epoch": 1.22, + "learning_rate": 1.846568035007101e-05, + "loss": 1.0114, + "step": 16508 + }, + { + "epoch": 1.22, + "learning_rate": 1.8465468031739565e-05, + "loss": 1.0502, + "step": 16509 + }, + { + "epoch": 1.22, + "learning_rate": 1.846525569993968e-05, + "loss": 0.9939, + "step": 16510 + }, + { + "epoch": 1.22, + "learning_rate": 1.8465043354671697e-05, + "loss": 1.0072, + "step": 16511 + }, + { + "epoch": 1.22, + "learning_rate": 1.846483099593595e-05, + "loss": 1.0895, + "step": 16512 + }, + { + "epoch": 1.22, + "learning_rate": 1.8464618623732774e-05, + "loss": 1.0204, + "step": 16513 + }, + { + "epoch": 1.22, + "learning_rate": 1.8464406238062515e-05, + "loss": 0.9947, + "step": 16514 + }, + { + "epoch": 1.22, + "learning_rate": 1.846419383892551e-05, + "loss": 1.0971, + "step": 16515 + }, + { + "epoch": 1.22, + "learning_rate": 1.846398142632209e-05, + "loss": 0.9419, + "step": 16516 + }, + { + "epoch": 1.22, + "learning_rate": 1.8463769000252595e-05, + "loss": 1.0314, + "step": 16517 + }, + { + "epoch": 1.22, + "learning_rate": 1.8463556560717366e-05, + "loss": 1.104, + "step": 16518 + }, + { + "epoch": 1.22, + "learning_rate": 1.846334410771674e-05, + "loss": 1.0338, + "step": 16519 + }, + { + "epoch": 1.22, + "learning_rate": 1.8463131641251054e-05, + "loss": 1.1138, + "step": 16520 + }, + { + "epoch": 1.22, + "learning_rate": 1.8462919161320648e-05, + "loss": 1.0502, + "step": 16521 + }, + { + "epoch": 1.22, + "learning_rate": 1.8462706667925857e-05, + "loss": 1.0267, + "step": 16522 + }, + { + "epoch": 1.22, + "learning_rate": 1.8462494161067023e-05, + "loss": 1.0907, + "step": 16523 + }, + { + "epoch": 1.22, + "learning_rate": 1.846228164074448e-05, + "loss": 1.0231, + "step": 16524 + }, + { + "epoch": 1.22, + "learning_rate": 1.846206910695857e-05, + "loss": 0.9683, + "step": 16525 + }, + { + "epoch": 1.22, + "learning_rate": 1.8461856559709627e-05, + "loss": 0.9948, + "step": 16526 + }, + { + "epoch": 1.22, + "learning_rate": 1.846164399899799e-05, + "loss": 1.0061, + "step": 16527 + }, + { + "epoch": 1.22, + "learning_rate": 1.8461431424824002e-05, + "loss": 1.0723, + "step": 16528 + }, + { + "epoch": 1.22, + "learning_rate": 1.8461218837187996e-05, + "loss": 1.028, + "step": 16529 + }, + { + "epoch": 1.22, + "learning_rate": 1.8461006236090313e-05, + "loss": 0.9674, + "step": 16530 + }, + { + "epoch": 1.22, + "learning_rate": 1.846079362153129e-05, + "loss": 1.0237, + "step": 16531 + }, + { + "epoch": 1.22, + "learning_rate": 1.8460580993511262e-05, + "loss": 0.9745, + "step": 16532 + }, + { + "epoch": 1.22, + "learning_rate": 1.8460368352030572e-05, + "loss": 1.0078, + "step": 16533 + }, + { + "epoch": 1.22, + "learning_rate": 1.8460155697089562e-05, + "loss": 1.0284, + "step": 16534 + }, + { + "epoch": 1.22, + "learning_rate": 1.8459943028688562e-05, + "loss": 0.9903, + "step": 16535 + }, + { + "epoch": 1.22, + "learning_rate": 1.8459730346827914e-05, + "loss": 0.8964, + "step": 16536 + }, + { + "epoch": 1.22, + "learning_rate": 1.8459517651507953e-05, + "loss": 0.9458, + "step": 16537 + }, + { + "epoch": 1.22, + "learning_rate": 1.8459304942729024e-05, + "loss": 0.9289, + "step": 16538 + }, + { + "epoch": 1.22, + "learning_rate": 1.845909222049146e-05, + "loss": 1.0283, + "step": 16539 + }, + { + "epoch": 1.22, + "learning_rate": 1.8458879484795604e-05, + "loss": 1.0168, + "step": 16540 + }, + { + "epoch": 1.22, + "learning_rate": 1.845866673564179e-05, + "loss": 0.9948, + "step": 16541 + }, + { + "epoch": 1.22, + "learning_rate": 1.8458453973030356e-05, + "loss": 0.9963, + "step": 16542 + }, + { + "epoch": 1.22, + "learning_rate": 1.8458241196961643e-05, + "loss": 1.0859, + "step": 16543 + }, + { + "epoch": 1.22, + "learning_rate": 1.8458028407435992e-05, + "loss": 0.9829, + "step": 16544 + }, + { + "epoch": 1.22, + "learning_rate": 1.845781560445374e-05, + "loss": 0.995, + "step": 16545 + }, + { + "epoch": 1.22, + "learning_rate": 1.8457602788015218e-05, + "loss": 1.0471, + "step": 16546 + }, + { + "epoch": 1.22, + "learning_rate": 1.8457389958120777e-05, + "loss": 1.037, + "step": 16547 + }, + { + "epoch": 1.22, + "learning_rate": 1.845717711477075e-05, + "loss": 1.1155, + "step": 16548 + }, + { + "epoch": 1.22, + "learning_rate": 1.8456964257965474e-05, + "loss": 0.9732, + "step": 16549 + }, + { + "epoch": 1.22, + "learning_rate": 1.8456751387705285e-05, + "loss": 1.0647, + "step": 16550 + }, + { + "epoch": 1.22, + "learning_rate": 1.845653850399053e-05, + "loss": 1.1116, + "step": 16551 + }, + { + "epoch": 1.22, + "learning_rate": 1.845632560682154e-05, + "loss": 1.067, + "step": 16552 + }, + { + "epoch": 1.22, + "learning_rate": 1.8456112696198657e-05, + "loss": 0.9679, + "step": 16553 + }, + { + "epoch": 1.22, + "learning_rate": 1.8455899772122222e-05, + "loss": 0.8854, + "step": 16554 + }, + { + "epoch": 1.22, + "learning_rate": 1.8455686834592573e-05, + "loss": 0.9547, + "step": 16555 + }, + { + "epoch": 1.22, + "learning_rate": 1.845547388361004e-05, + "loss": 1.0339, + "step": 16556 + }, + { + "epoch": 1.22, + "learning_rate": 1.8455260919174977e-05, + "loss": 0.9899, + "step": 16557 + }, + { + "epoch": 1.22, + "learning_rate": 1.8455047941287715e-05, + "loss": 1.0044, + "step": 16558 + }, + { + "epoch": 1.22, + "learning_rate": 1.8454834949948587e-05, + "loss": 0.9871, + "step": 16559 + }, + { + "epoch": 1.22, + "learning_rate": 1.845462194515794e-05, + "loss": 1.0211, + "step": 16560 + }, + { + "epoch": 1.22, + "learning_rate": 1.845440892691611e-05, + "loss": 0.8705, + "step": 16561 + }, + { + "epoch": 1.22, + "learning_rate": 1.8454195895223437e-05, + "loss": 1.1316, + "step": 16562 + }, + { + "epoch": 1.22, + "learning_rate": 1.845398285008026e-05, + "loss": 1.0664, + "step": 16563 + }, + { + "epoch": 1.22, + "learning_rate": 1.8453769791486917e-05, + "loss": 1.0483, + "step": 16564 + }, + { + "epoch": 1.22, + "learning_rate": 1.8453556719443746e-05, + "loss": 1.0407, + "step": 16565 + }, + { + "epoch": 1.22, + "learning_rate": 1.8453343633951085e-05, + "loss": 1.0036, + "step": 16566 + }, + { + "epoch": 1.22, + "learning_rate": 1.845313053500928e-05, + "loss": 1.0864, + "step": 16567 + }, + { + "epoch": 1.22, + "learning_rate": 1.8452917422618663e-05, + "loss": 1.0878, + "step": 16568 + }, + { + "epoch": 1.22, + "learning_rate": 1.8452704296779577e-05, + "loss": 1.1041, + "step": 16569 + }, + { + "epoch": 1.22, + "learning_rate": 1.8452491157492358e-05, + "loss": 1.0423, + "step": 16570 + }, + { + "epoch": 1.22, + "learning_rate": 1.8452278004757346e-05, + "loss": 0.9461, + "step": 16571 + }, + { + "epoch": 1.22, + "learning_rate": 1.845206483857488e-05, + "loss": 1.0057, + "step": 16572 + }, + { + "epoch": 1.22, + "learning_rate": 1.84518516589453e-05, + "loss": 0.9553, + "step": 16573 + }, + { + "epoch": 1.22, + "learning_rate": 1.845163846586895e-05, + "loss": 1.0361, + "step": 16574 + }, + { + "epoch": 1.22, + "learning_rate": 1.845142525934616e-05, + "loss": 0.9698, + "step": 16575 + }, + { + "epoch": 1.22, + "learning_rate": 1.845121203937727e-05, + "loss": 1.0357, + "step": 16576 + }, + { + "epoch": 1.22, + "learning_rate": 1.8450998805962628e-05, + "loss": 1.0984, + "step": 16577 + }, + { + "epoch": 1.22, + "learning_rate": 1.8450785559102567e-05, + "loss": 1.041, + "step": 16578 + }, + { + "epoch": 1.23, + "learning_rate": 1.8450572298797424e-05, + "loss": 1.0609, + "step": 16579 + }, + { + "epoch": 1.23, + "learning_rate": 1.8450359025047543e-05, + "loss": 1.1134, + "step": 16580 + }, + { + "epoch": 1.23, + "learning_rate": 1.845014573785326e-05, + "loss": 1.1429, + "step": 16581 + }, + { + "epoch": 1.23, + "learning_rate": 1.8449932437214923e-05, + "loss": 1.04, + "step": 16582 + }, + { + "epoch": 1.23, + "learning_rate": 1.8449719123132857e-05, + "loss": 1.0694, + "step": 16583 + }, + { + "epoch": 1.23, + "learning_rate": 1.8449505795607412e-05, + "loss": 1.1481, + "step": 16584 + }, + { + "epoch": 1.23, + "learning_rate": 1.844929245463892e-05, + "loss": 0.9107, + "step": 16585 + }, + { + "epoch": 1.23, + "learning_rate": 1.844907910022773e-05, + "loss": 1.0558, + "step": 16586 + }, + { + "epoch": 1.23, + "learning_rate": 1.8448865732374176e-05, + "loss": 1.0379, + "step": 16587 + }, + { + "epoch": 1.23, + "learning_rate": 1.8448652351078593e-05, + "loss": 1.0113, + "step": 16588 + }, + { + "epoch": 1.23, + "learning_rate": 1.8448438956341328e-05, + "loss": 0.8981, + "step": 16589 + }, + { + "epoch": 1.23, + "learning_rate": 1.8448225548162714e-05, + "loss": 0.9948, + "step": 16590 + }, + { + "epoch": 1.23, + "learning_rate": 1.8448012126543097e-05, + "loss": 1.0171, + "step": 16591 + }, + { + "epoch": 1.23, + "learning_rate": 1.8447798691482815e-05, + "loss": 1.0119, + "step": 16592 + }, + { + "epoch": 1.23, + "learning_rate": 1.8447585242982207e-05, + "loss": 1.0411, + "step": 16593 + }, + { + "epoch": 1.23, + "learning_rate": 1.8447371781041606e-05, + "loss": 1.1586, + "step": 16594 + }, + { + "epoch": 1.23, + "learning_rate": 1.8447158305661363e-05, + "loss": 1.0751, + "step": 16595 + }, + { + "epoch": 1.23, + "learning_rate": 1.844694481684181e-05, + "loss": 0.9585, + "step": 16596 + }, + { + "epoch": 1.23, + "learning_rate": 1.8446731314583285e-05, + "loss": 1.0471, + "step": 16597 + }, + { + "epoch": 1.23, + "learning_rate": 1.8446517798886138e-05, + "loss": 1.026, + "step": 16598 + }, + { + "epoch": 1.23, + "learning_rate": 1.8446304269750697e-05, + "loss": 1.1492, + "step": 16599 + }, + { + "epoch": 1.23, + "learning_rate": 1.844609072717731e-05, + "loss": 1.0601, + "step": 16600 + }, + { + "epoch": 1.23, + "learning_rate": 1.8445877171166313e-05, + "loss": 1.0886, + "step": 16601 + }, + { + "epoch": 1.23, + "learning_rate": 1.8445663601718044e-05, + "loss": 0.9578, + "step": 16602 + }, + { + "epoch": 1.23, + "learning_rate": 1.8445450018832844e-05, + "loss": 1.024, + "step": 16603 + }, + { + "epoch": 1.23, + "learning_rate": 1.844523642251106e-05, + "loss": 1.0628, + "step": 16604 + }, + { + "epoch": 1.23, + "learning_rate": 1.8445022812753023e-05, + "loss": 1.0659, + "step": 16605 + }, + { + "epoch": 1.23, + "learning_rate": 1.8444809189559073e-05, + "loss": 1.012, + "step": 16606 + }, + { + "epoch": 1.23, + "learning_rate": 1.844459555292955e-05, + "loss": 1.0865, + "step": 16607 + }, + { + "epoch": 1.23, + "learning_rate": 1.8444381902864803e-05, + "loss": 1.0822, + "step": 16608 + }, + { + "epoch": 1.23, + "learning_rate": 1.8444168239365163e-05, + "loss": 0.9691, + "step": 16609 + }, + { + "epoch": 1.23, + "learning_rate": 1.844395456243097e-05, + "loss": 1.0661, + "step": 16610 + }, + { + "epoch": 1.23, + "learning_rate": 1.8443740872062568e-05, + "loss": 1.052, + "step": 16611 + }, + { + "epoch": 1.23, + "learning_rate": 1.8443527168260294e-05, + "loss": 1.0405, + "step": 16612 + }, + { + "epoch": 1.23, + "learning_rate": 1.844331345102449e-05, + "loss": 1.0244, + "step": 16613 + }, + { + "epoch": 1.23, + "learning_rate": 1.8443099720355494e-05, + "loss": 1.0382, + "step": 16614 + }, + { + "epoch": 1.23, + "learning_rate": 1.844288597625365e-05, + "loss": 1.1347, + "step": 16615 + }, + { + "epoch": 1.23, + "learning_rate": 1.844267221871929e-05, + "loss": 1.0875, + "step": 16616 + }, + { + "epoch": 1.23, + "learning_rate": 1.8442458447752763e-05, + "loss": 0.9409, + "step": 16617 + }, + { + "epoch": 1.23, + "learning_rate": 1.8442244663354406e-05, + "loss": 1.09, + "step": 16618 + }, + { + "epoch": 1.23, + "learning_rate": 1.8442030865524557e-05, + "loss": 1.0766, + "step": 16619 + }, + { + "epoch": 1.23, + "learning_rate": 1.844181705426356e-05, + "loss": 1.0286, + "step": 16620 + }, + { + "epoch": 1.23, + "learning_rate": 1.844160322957175e-05, + "loss": 1.0285, + "step": 16621 + }, + { + "epoch": 1.23, + "learning_rate": 1.844138939144947e-05, + "loss": 1.0787, + "step": 16622 + }, + { + "epoch": 1.23, + "learning_rate": 1.844117553989706e-05, + "loss": 0.9532, + "step": 16623 + }, + { + "epoch": 1.23, + "learning_rate": 1.8440961674914865e-05, + "loss": 1.0942, + "step": 16624 + }, + { + "epoch": 1.23, + "learning_rate": 1.8440747796503217e-05, + "loss": 1.0303, + "step": 16625 + }, + { + "epoch": 1.23, + "learning_rate": 1.8440533904662458e-05, + "loss": 0.9539, + "step": 16626 + }, + { + "epoch": 1.23, + "learning_rate": 1.8440319999392937e-05, + "loss": 1.0411, + "step": 16627 + }, + { + "epoch": 1.23, + "learning_rate": 1.8440106080694982e-05, + "loss": 0.9724, + "step": 16628 + }, + { + "epoch": 1.23, + "learning_rate": 1.843989214856894e-05, + "loss": 0.9896, + "step": 16629 + }, + { + "epoch": 1.23, + "learning_rate": 1.843967820301515e-05, + "loss": 0.9851, + "step": 16630 + }, + { + "epoch": 1.23, + "learning_rate": 1.8439464244033953e-05, + "loss": 1.0698, + "step": 16631 + }, + { + "epoch": 1.23, + "learning_rate": 1.843925027162569e-05, + "loss": 0.9353, + "step": 16632 + }, + { + "epoch": 1.23, + "learning_rate": 1.8439036285790698e-05, + "loss": 0.9394, + "step": 16633 + }, + { + "epoch": 1.23, + "learning_rate": 1.8438822286529325e-05, + "loss": 1.1103, + "step": 16634 + }, + { + "epoch": 1.23, + "learning_rate": 1.8438608273841905e-05, + "loss": 1.0838, + "step": 16635 + }, + { + "epoch": 1.23, + "learning_rate": 1.8438394247728778e-05, + "loss": 1.0703, + "step": 16636 + }, + { + "epoch": 1.23, + "learning_rate": 1.8438180208190288e-05, + "loss": 1.0511, + "step": 16637 + }, + { + "epoch": 1.23, + "learning_rate": 1.8437966155226772e-05, + "loss": 1.1302, + "step": 16638 + }, + { + "epoch": 1.23, + "learning_rate": 1.8437752088838574e-05, + "loss": 0.9693, + "step": 16639 + }, + { + "epoch": 1.23, + "learning_rate": 1.843753800902603e-05, + "loss": 0.9603, + "step": 16640 + }, + { + "epoch": 1.23, + "learning_rate": 1.8437323915789488e-05, + "loss": 1.0932, + "step": 16641 + }, + { + "epoch": 1.23, + "learning_rate": 1.8437109809129285e-05, + "loss": 0.9454, + "step": 16642 + }, + { + "epoch": 1.23, + "learning_rate": 1.8436895689045757e-05, + "loss": 1.0576, + "step": 16643 + }, + { + "epoch": 1.23, + "learning_rate": 1.843668155553925e-05, + "loss": 0.9893, + "step": 16644 + }, + { + "epoch": 1.23, + "learning_rate": 1.8436467408610105e-05, + "loss": 0.9513, + "step": 16645 + }, + { + "epoch": 1.23, + "learning_rate": 1.8436253248258658e-05, + "loss": 1.0476, + "step": 16646 + }, + { + "epoch": 1.23, + "learning_rate": 1.8436039074485256e-05, + "loss": 1.0003, + "step": 16647 + }, + { + "epoch": 1.23, + "learning_rate": 1.8435824887290237e-05, + "loss": 1.062, + "step": 16648 + }, + { + "epoch": 1.23, + "learning_rate": 1.8435610686673938e-05, + "loss": 0.9485, + "step": 16649 + }, + { + "epoch": 1.23, + "learning_rate": 1.8435396472636706e-05, + "loss": 0.9923, + "step": 16650 + }, + { + "epoch": 1.23, + "learning_rate": 1.8435182245178877e-05, + "loss": 1.0249, + "step": 16651 + }, + { + "epoch": 1.23, + "learning_rate": 1.8434968004300798e-05, + "loss": 0.9492, + "step": 16652 + }, + { + "epoch": 1.23, + "learning_rate": 1.8434753750002802e-05, + "loss": 1.0952, + "step": 16653 + }, + { + "epoch": 1.23, + "learning_rate": 1.8434539482285233e-05, + "loss": 0.9718, + "step": 16654 + }, + { + "epoch": 1.23, + "learning_rate": 1.843432520114843e-05, + "loss": 0.9845, + "step": 16655 + }, + { + "epoch": 1.23, + "learning_rate": 1.843411090659274e-05, + "loss": 1.1261, + "step": 16656 + }, + { + "epoch": 1.23, + "learning_rate": 1.8433896598618498e-05, + "loss": 0.9989, + "step": 16657 + }, + { + "epoch": 1.23, + "learning_rate": 1.843368227722605e-05, + "loss": 0.9154, + "step": 16658 + }, + { + "epoch": 1.23, + "learning_rate": 1.8433467942415734e-05, + "loss": 1.031, + "step": 16659 + }, + { + "epoch": 1.23, + "learning_rate": 1.8433253594187887e-05, + "loss": 0.9253, + "step": 16660 + }, + { + "epoch": 1.23, + "learning_rate": 1.8433039232542858e-05, + "loss": 1.0672, + "step": 16661 + }, + { + "epoch": 1.23, + "learning_rate": 1.8432824857480983e-05, + "loss": 1.0104, + "step": 16662 + }, + { + "epoch": 1.23, + "learning_rate": 1.8432610469002605e-05, + "loss": 1.0283, + "step": 16663 + }, + { + "epoch": 1.23, + "learning_rate": 1.8432396067108065e-05, + "loss": 1.0596, + "step": 16664 + }, + { + "epoch": 1.23, + "learning_rate": 1.84321816517977e-05, + "loss": 1.0671, + "step": 16665 + }, + { + "epoch": 1.23, + "learning_rate": 1.843196722307186e-05, + "loss": 0.9479, + "step": 16666 + }, + { + "epoch": 1.23, + "learning_rate": 1.843175278093088e-05, + "loss": 0.9672, + "step": 16667 + }, + { + "epoch": 1.23, + "learning_rate": 1.8431538325375098e-05, + "loss": 1.0454, + "step": 16668 + }, + { + "epoch": 1.23, + "learning_rate": 1.8431323856404862e-05, + "loss": 1.0859, + "step": 16669 + }, + { + "epoch": 1.23, + "learning_rate": 1.8431109374020508e-05, + "loss": 0.9869, + "step": 16670 + }, + { + "epoch": 1.23, + "learning_rate": 1.843089487822238e-05, + "loss": 1.0823, + "step": 16671 + }, + { + "epoch": 1.23, + "learning_rate": 1.843068036901082e-05, + "loss": 0.9992, + "step": 16672 + }, + { + "epoch": 1.23, + "learning_rate": 1.843046584638617e-05, + "loss": 0.9714, + "step": 16673 + }, + { + "epoch": 1.23, + "learning_rate": 1.8430251310348767e-05, + "loss": 1.0105, + "step": 16674 + }, + { + "epoch": 1.23, + "learning_rate": 1.8430036760898956e-05, + "loss": 1.0013, + "step": 16675 + }, + { + "epoch": 1.23, + "learning_rate": 1.8429822198037077e-05, + "loss": 0.9667, + "step": 16676 + }, + { + "epoch": 1.23, + "learning_rate": 1.8429607621763467e-05, + "loss": 1.0536, + "step": 16677 + }, + { + "epoch": 1.23, + "learning_rate": 1.8429393032078477e-05, + "loss": 1.0033, + "step": 16678 + }, + { + "epoch": 1.23, + "learning_rate": 1.842917842898244e-05, + "loss": 1.073, + "step": 16679 + }, + { + "epoch": 1.23, + "learning_rate": 1.8428963812475705e-05, + "loss": 1.0463, + "step": 16680 + }, + { + "epoch": 1.23, + "learning_rate": 1.8428749182558608e-05, + "loss": 1.0875, + "step": 16681 + }, + { + "epoch": 1.23, + "learning_rate": 1.8428534539231492e-05, + "loss": 1.0049, + "step": 16682 + }, + { + "epoch": 1.23, + "learning_rate": 1.8428319882494693e-05, + "loss": 1.0298, + "step": 16683 + }, + { + "epoch": 1.23, + "learning_rate": 1.8428105212348562e-05, + "loss": 1.0164, + "step": 16684 + }, + { + "epoch": 1.23, + "learning_rate": 1.8427890528793436e-05, + "loss": 1.0037, + "step": 16685 + }, + { + "epoch": 1.23, + "learning_rate": 1.842767583182966e-05, + "loss": 0.9955, + "step": 16686 + }, + { + "epoch": 1.23, + "learning_rate": 1.8427461121457568e-05, + "loss": 1.0235, + "step": 16687 + }, + { + "epoch": 1.23, + "learning_rate": 1.8427246397677504e-05, + "loss": 0.9588, + "step": 16688 + }, + { + "epoch": 1.23, + "learning_rate": 1.8427031660489817e-05, + "loss": 0.949, + "step": 16689 + }, + { + "epoch": 1.23, + "learning_rate": 1.842681690989484e-05, + "loss": 1.0361, + "step": 16690 + }, + { + "epoch": 1.23, + "learning_rate": 1.8426602145892917e-05, + "loss": 1.0607, + "step": 16691 + }, + { + "epoch": 1.23, + "learning_rate": 1.8426387368484392e-05, + "loss": 0.9605, + "step": 16692 + }, + { + "epoch": 1.23, + "learning_rate": 1.8426172577669606e-05, + "loss": 1.0142, + "step": 16693 + }, + { + "epoch": 1.23, + "learning_rate": 1.8425957773448896e-05, + "loss": 1.0393, + "step": 16694 + }, + { + "epoch": 1.23, + "learning_rate": 1.8425742955822614e-05, + "loss": 0.995, + "step": 16695 + }, + { + "epoch": 1.23, + "learning_rate": 1.842552812479109e-05, + "loss": 1.039, + "step": 16696 + }, + { + "epoch": 1.23, + "learning_rate": 1.8425313280354675e-05, + "loss": 0.998, + "step": 16697 + }, + { + "epoch": 1.23, + "learning_rate": 1.8425098422513707e-05, + "loss": 1.1649, + "step": 16698 + }, + { + "epoch": 1.23, + "learning_rate": 1.8424883551268524e-05, + "loss": 1.0626, + "step": 16699 + }, + { + "epoch": 1.23, + "learning_rate": 1.8424668666619475e-05, + "loss": 1.0269, + "step": 16700 + }, + { + "epoch": 1.23, + "learning_rate": 1.84244537685669e-05, + "loss": 0.9719, + "step": 16701 + }, + { + "epoch": 1.23, + "learning_rate": 1.8424238857111138e-05, + "loss": 0.9253, + "step": 16702 + }, + { + "epoch": 1.23, + "learning_rate": 1.8424023932252534e-05, + "loss": 1.0008, + "step": 16703 + }, + { + "epoch": 1.23, + "learning_rate": 1.8423808993991422e-05, + "loss": 1.0883, + "step": 16704 + }, + { + "epoch": 1.23, + "learning_rate": 1.842359404232816e-05, + "loss": 1.0644, + "step": 16705 + }, + { + "epoch": 1.23, + "learning_rate": 1.8423379077263072e-05, + "loss": 1.0724, + "step": 16706 + }, + { + "epoch": 1.23, + "learning_rate": 1.8423164098796512e-05, + "loss": 1.0343, + "step": 16707 + }, + { + "epoch": 1.23, + "learning_rate": 1.842294910692882e-05, + "loss": 0.986, + "step": 16708 + }, + { + "epoch": 1.23, + "learning_rate": 1.8422734101660334e-05, + "loss": 1.0484, + "step": 16709 + }, + { + "epoch": 1.23, + "learning_rate": 1.84225190829914e-05, + "loss": 1.1367, + "step": 16710 + }, + { + "epoch": 1.23, + "learning_rate": 1.8422304050922356e-05, + "loss": 0.957, + "step": 16711 + }, + { + "epoch": 1.23, + "learning_rate": 1.842208900545355e-05, + "loss": 0.9195, + "step": 16712 + }, + { + "epoch": 1.23, + "learning_rate": 1.8421873946585323e-05, + "loss": 1.0314, + "step": 16713 + }, + { + "epoch": 1.23, + "learning_rate": 1.8421658874318008e-05, + "loss": 1.0325, + "step": 16714 + }, + { + "epoch": 1.24, + "learning_rate": 1.842144378865196e-05, + "loss": 0.8835, + "step": 16715 + }, + { + "epoch": 1.24, + "learning_rate": 1.8421228689587512e-05, + "loss": 0.993, + "step": 16716 + }, + { + "epoch": 1.24, + "learning_rate": 1.8421013577125012e-05, + "loss": 0.9931, + "step": 16717 + }, + { + "epoch": 1.24, + "learning_rate": 1.84207984512648e-05, + "loss": 1.0135, + "step": 16718 + }, + { + "epoch": 1.24, + "learning_rate": 1.8420583312007217e-05, + "loss": 1.0782, + "step": 16719 + }, + { + "epoch": 1.24, + "learning_rate": 1.8420368159352606e-05, + "loss": 1.0199, + "step": 16720 + }, + { + "epoch": 1.24, + "learning_rate": 1.842015299330131e-05, + "loss": 1.0225, + "step": 16721 + }, + { + "epoch": 1.24, + "learning_rate": 1.841993781385367e-05, + "loss": 1.0421, + "step": 16722 + }, + { + "epoch": 1.24, + "learning_rate": 1.841972262101003e-05, + "loss": 1.0634, + "step": 16723 + }, + { + "epoch": 1.24, + "learning_rate": 1.8419507414770734e-05, + "loss": 1.115, + "step": 16724 + }, + { + "epoch": 1.24, + "learning_rate": 1.8419292195136123e-05, + "loss": 0.9425, + "step": 16725 + }, + { + "epoch": 1.24, + "learning_rate": 1.8419076962106536e-05, + "loss": 1.0261, + "step": 16726 + }, + { + "epoch": 1.24, + "learning_rate": 1.8418861715682316e-05, + "loss": 0.9127, + "step": 16727 + }, + { + "epoch": 1.24, + "learning_rate": 1.841864645586381e-05, + "loss": 1.0348, + "step": 16728 + }, + { + "epoch": 1.24, + "learning_rate": 1.841843118265136e-05, + "loss": 0.9825, + "step": 16729 + }, + { + "epoch": 1.24, + "learning_rate": 1.8418215896045303e-05, + "loss": 1.0916, + "step": 16730 + }, + { + "epoch": 1.24, + "learning_rate": 1.8418000596045988e-05, + "loss": 1.0179, + "step": 16731 + }, + { + "epoch": 1.24, + "learning_rate": 1.841778528265375e-05, + "loss": 0.8646, + "step": 16732 + }, + { + "epoch": 1.24, + "learning_rate": 1.841756995586894e-05, + "loss": 1.0682, + "step": 16733 + }, + { + "epoch": 1.24, + "learning_rate": 1.8417354615691893e-05, + "loss": 1.0896, + "step": 16734 + }, + { + "epoch": 1.24, + "learning_rate": 1.841713926212296e-05, + "loss": 1.0285, + "step": 16735 + }, + { + "epoch": 1.24, + "learning_rate": 1.8416923895162477e-05, + "loss": 1.046, + "step": 16736 + }, + { + "epoch": 1.24, + "learning_rate": 1.8416708514810787e-05, + "loss": 0.99, + "step": 16737 + }, + { + "epoch": 1.24, + "learning_rate": 1.841649312106823e-05, + "loss": 0.9623, + "step": 16738 + }, + { + "epoch": 1.24, + "learning_rate": 1.8416277713935163e-05, + "loss": 1.1147, + "step": 16739 + }, + { + "epoch": 1.24, + "learning_rate": 1.8416062293411912e-05, + "loss": 0.9726, + "step": 16740 + }, + { + "epoch": 1.24, + "learning_rate": 1.841584685949883e-05, + "loss": 0.9824, + "step": 16741 + }, + { + "epoch": 1.24, + "learning_rate": 1.8415631412196253e-05, + "loss": 1.0349, + "step": 16742 + }, + { + "epoch": 1.24, + "learning_rate": 1.8415415951504525e-05, + "loss": 1.0227, + "step": 16743 + }, + { + "epoch": 1.24, + "learning_rate": 1.8415200477423994e-05, + "loss": 1.0135, + "step": 16744 + }, + { + "epoch": 1.24, + "learning_rate": 1.8414984989954997e-05, + "loss": 1.0612, + "step": 16745 + }, + { + "epoch": 1.24, + "learning_rate": 1.8414769489097885e-05, + "loss": 1.0216, + "step": 16746 + }, + { + "epoch": 1.24, + "learning_rate": 1.841455397485299e-05, + "loss": 0.9362, + "step": 16747 + }, + { + "epoch": 1.24, + "learning_rate": 1.841433844722066e-05, + "loss": 0.982, + "step": 16748 + }, + { + "epoch": 1.24, + "learning_rate": 1.841412290620124e-05, + "loss": 0.9799, + "step": 16749 + }, + { + "epoch": 1.24, + "learning_rate": 1.841390735179507e-05, + "loss": 1.0462, + "step": 16750 + }, + { + "epoch": 1.24, + "learning_rate": 1.8413691784002493e-05, + "loss": 1.0476, + "step": 16751 + }, + { + "epoch": 1.24, + "learning_rate": 1.8413476202823855e-05, + "loss": 1.0567, + "step": 16752 + }, + { + "epoch": 1.24, + "learning_rate": 1.8413260608259492e-05, + "loss": 0.9252, + "step": 16753 + }, + { + "epoch": 1.24, + "learning_rate": 1.8413045000309758e-05, + "loss": 0.9908, + "step": 16754 + }, + { + "epoch": 1.24, + "learning_rate": 1.8412829378974986e-05, + "loss": 0.9805, + "step": 16755 + }, + { + "epoch": 1.24, + "learning_rate": 1.8412613744255524e-05, + "loss": 1.0435, + "step": 16756 + }, + { + "epoch": 1.24, + "learning_rate": 1.8412398096151712e-05, + "loss": 0.9022, + "step": 16757 + }, + { + "epoch": 1.24, + "learning_rate": 1.84121824346639e-05, + "loss": 1.1086, + "step": 16758 + }, + { + "epoch": 1.24, + "learning_rate": 1.8411966759792422e-05, + "loss": 1.1165, + "step": 16759 + }, + { + "epoch": 1.24, + "learning_rate": 1.8411751071537626e-05, + "loss": 0.9274, + "step": 16760 + }, + { + "epoch": 1.24, + "learning_rate": 1.8411535369899854e-05, + "loss": 1.0231, + "step": 16761 + }, + { + "epoch": 1.24, + "learning_rate": 1.8411319654879447e-05, + "loss": 0.9632, + "step": 16762 + }, + { + "epoch": 1.24, + "learning_rate": 1.8411103926476756e-05, + "loss": 1.0324, + "step": 16763 + }, + { + "epoch": 1.24, + "learning_rate": 1.8410888184692116e-05, + "loss": 1.0502, + "step": 16764 + }, + { + "epoch": 1.24, + "learning_rate": 1.8410672429525874e-05, + "loss": 0.9713, + "step": 16765 + }, + { + "epoch": 1.24, + "learning_rate": 1.841045666097837e-05, + "loss": 0.9919, + "step": 16766 + }, + { + "epoch": 1.24, + "learning_rate": 1.8410240879049954e-05, + "loss": 0.9813, + "step": 16767 + }, + { + "epoch": 1.24, + "learning_rate": 1.8410025083740964e-05, + "loss": 1.1124, + "step": 16768 + }, + { + "epoch": 1.24, + "learning_rate": 1.8409809275051743e-05, + "loss": 1.0081, + "step": 16769 + }, + { + "epoch": 1.24, + "learning_rate": 1.8409593452982637e-05, + "loss": 0.9955, + "step": 16770 + }, + { + "epoch": 1.24, + "learning_rate": 1.8409377617533986e-05, + "loss": 0.9769, + "step": 16771 + }, + { + "epoch": 1.24, + "learning_rate": 1.840916176870614e-05, + "loss": 0.9741, + "step": 16772 + }, + { + "epoch": 1.24, + "learning_rate": 1.8408945906499433e-05, + "loss": 0.9842, + "step": 16773 + }, + { + "epoch": 1.24, + "learning_rate": 1.8408730030914213e-05, + "loss": 0.9845, + "step": 16774 + }, + { + "epoch": 1.24, + "learning_rate": 1.8408514141950826e-05, + "loss": 1.0415, + "step": 16775 + }, + { + "epoch": 1.24, + "learning_rate": 1.8408298239609618e-05, + "loss": 1.0269, + "step": 16776 + }, + { + "epoch": 1.24, + "learning_rate": 1.840808232389092e-05, + "loss": 0.9452, + "step": 16777 + }, + { + "epoch": 1.24, + "learning_rate": 1.8407866394795088e-05, + "loss": 1.0264, + "step": 16778 + }, + { + "epoch": 1.24, + "learning_rate": 1.8407650452322458e-05, + "loss": 0.9874, + "step": 16779 + }, + { + "epoch": 1.24, + "learning_rate": 1.840743449647338e-05, + "loss": 1.0187, + "step": 16780 + }, + { + "epoch": 1.24, + "learning_rate": 1.8407218527248192e-05, + "loss": 0.9264, + "step": 16781 + }, + { + "epoch": 1.24, + "learning_rate": 1.8407002544647237e-05, + "loss": 0.987, + "step": 16782 + }, + { + "epoch": 1.24, + "learning_rate": 1.8406786548670867e-05, + "loss": 1.0785, + "step": 16783 + }, + { + "epoch": 1.24, + "learning_rate": 1.8406570539319415e-05, + "loss": 1.0151, + "step": 16784 + }, + { + "epoch": 1.24, + "learning_rate": 1.840635451659323e-05, + "loss": 1.036, + "step": 16785 + }, + { + "epoch": 1.24, + "learning_rate": 1.8406138480492657e-05, + "loss": 1.078, + "step": 16786 + }, + { + "epoch": 1.24, + "learning_rate": 1.8405922431018038e-05, + "loss": 0.86, + "step": 16787 + }, + { + "epoch": 1.24, + "learning_rate": 1.8405706368169713e-05, + "loss": 1.0439, + "step": 16788 + }, + { + "epoch": 1.24, + "learning_rate": 1.8405490291948033e-05, + "loss": 0.9961, + "step": 16789 + }, + { + "epoch": 1.24, + "learning_rate": 1.840527420235334e-05, + "loss": 1.0107, + "step": 16790 + }, + { + "epoch": 1.24, + "learning_rate": 1.8405058099385973e-05, + "loss": 1.0004, + "step": 16791 + }, + { + "epoch": 1.24, + "learning_rate": 1.840484198304628e-05, + "loss": 0.9221, + "step": 16792 + }, + { + "epoch": 1.24, + "learning_rate": 1.84046258533346e-05, + "loss": 1.0237, + "step": 16793 + }, + { + "epoch": 1.24, + "learning_rate": 1.8404409710251286e-05, + "loss": 0.991, + "step": 16794 + }, + { + "epoch": 1.24, + "learning_rate": 1.8404193553796673e-05, + "loss": 0.961, + "step": 16795 + }, + { + "epoch": 1.24, + "learning_rate": 1.840397738397111e-05, + "loss": 0.9642, + "step": 16796 + }, + { + "epoch": 1.24, + "learning_rate": 1.840376120077494e-05, + "loss": 1.0804, + "step": 16797 + }, + { + "epoch": 1.24, + "learning_rate": 1.8403545004208504e-05, + "loss": 0.9443, + "step": 16798 + }, + { + "epoch": 1.24, + "learning_rate": 1.840332879427215e-05, + "loss": 1.0313, + "step": 16799 + }, + { + "epoch": 1.24, + "learning_rate": 1.8403112570966217e-05, + "loss": 1.0425, + "step": 16800 + }, + { + "epoch": 1.24, + "learning_rate": 1.8402896334291055e-05, + "loss": 1.0091, + "step": 16801 + }, + { + "epoch": 1.24, + "learning_rate": 1.8402680084247006e-05, + "loss": 0.9646, + "step": 16802 + }, + { + "epoch": 1.24, + "learning_rate": 1.8402463820834412e-05, + "loss": 1.0611, + "step": 16803 + }, + { + "epoch": 1.24, + "learning_rate": 1.8402247544053616e-05, + "loss": 0.9926, + "step": 16804 + }, + { + "epoch": 1.24, + "learning_rate": 1.8402031253904966e-05, + "loss": 1.1029, + "step": 16805 + }, + { + "epoch": 1.24, + "learning_rate": 1.8401814950388804e-05, + "loss": 0.94, + "step": 16806 + }, + { + "epoch": 1.24, + "learning_rate": 1.8401598633505477e-05, + "loss": 1.0099, + "step": 16807 + }, + { + "epoch": 1.24, + "learning_rate": 1.8401382303255322e-05, + "loss": 0.9453, + "step": 16808 + }, + { + "epoch": 1.24, + "learning_rate": 1.840116595963869e-05, + "loss": 0.9612, + "step": 16809 + }, + { + "epoch": 1.24, + "learning_rate": 1.8400949602655925e-05, + "loss": 1.0926, + "step": 16810 + }, + { + "epoch": 1.24, + "learning_rate": 1.8400733232307373e-05, + "loss": 1.0181, + "step": 16811 + }, + { + "epoch": 1.24, + "learning_rate": 1.840051684859337e-05, + "loss": 1.0305, + "step": 16812 + }, + { + "epoch": 1.24, + "learning_rate": 1.8400300451514264e-05, + "loss": 1.0748, + "step": 16813 + }, + { + "epoch": 1.24, + "learning_rate": 1.84000840410704e-05, + "loss": 0.9398, + "step": 16814 + }, + { + "epoch": 1.24, + "learning_rate": 1.8399867617262127e-05, + "loss": 0.9988, + "step": 16815 + }, + { + "epoch": 1.24, + "learning_rate": 1.8399651180089778e-05, + "loss": 1.0755, + "step": 16816 + }, + { + "epoch": 1.24, + "learning_rate": 1.8399434729553708e-05, + "loss": 1.0224, + "step": 16817 + }, + { + "epoch": 1.24, + "learning_rate": 1.8399218265654256e-05, + "loss": 1.0252, + "step": 16818 + }, + { + "epoch": 1.24, + "learning_rate": 1.839900178839177e-05, + "loss": 0.9821, + "step": 16819 + }, + { + "epoch": 1.24, + "learning_rate": 1.8398785297766592e-05, + "loss": 1.0086, + "step": 16820 + }, + { + "epoch": 1.24, + "learning_rate": 1.8398568793779067e-05, + "loss": 0.9298, + "step": 16821 + }, + { + "epoch": 1.24, + "learning_rate": 1.839835227642954e-05, + "loss": 1.1148, + "step": 16822 + }, + { + "epoch": 1.24, + "learning_rate": 1.8398135745718352e-05, + "loss": 1.069, + "step": 16823 + }, + { + "epoch": 1.24, + "learning_rate": 1.839791920164585e-05, + "loss": 1.0272, + "step": 16824 + }, + { + "epoch": 1.24, + "learning_rate": 1.8397702644212382e-05, + "loss": 0.9923, + "step": 16825 + }, + { + "epoch": 1.24, + "learning_rate": 1.8397486073418288e-05, + "loss": 1.1116, + "step": 16826 + }, + { + "epoch": 1.24, + "learning_rate": 1.8397269489263912e-05, + "loss": 0.87, + "step": 16827 + }, + { + "epoch": 1.24, + "learning_rate": 1.83970528917496e-05, + "loss": 0.9237, + "step": 16828 + }, + { + "epoch": 1.24, + "learning_rate": 1.83968362808757e-05, + "loss": 0.9925, + "step": 16829 + }, + { + "epoch": 1.24, + "learning_rate": 1.8396619656642548e-05, + "loss": 0.9514, + "step": 16830 + }, + { + "epoch": 1.24, + "learning_rate": 1.8396403019050503e-05, + "loss": 1.0192, + "step": 16831 + }, + { + "epoch": 1.24, + "learning_rate": 1.8396186368099896e-05, + "loss": 0.9641, + "step": 16832 + }, + { + "epoch": 1.24, + "learning_rate": 1.8395969703791073e-05, + "loss": 0.9969, + "step": 16833 + }, + { + "epoch": 1.24, + "learning_rate": 1.839575302612439e-05, + "loss": 1.0002, + "step": 16834 + }, + { + "epoch": 1.24, + "learning_rate": 1.8395536335100178e-05, + "loss": 1.02, + "step": 16835 + }, + { + "epoch": 1.24, + "learning_rate": 1.839531963071879e-05, + "loss": 1.0262, + "step": 16836 + }, + { + "epoch": 1.24, + "learning_rate": 1.839510291298057e-05, + "loss": 1.0862, + "step": 16837 + }, + { + "epoch": 1.24, + "learning_rate": 1.839488618188586e-05, + "loss": 0.9687, + "step": 16838 + }, + { + "epoch": 1.24, + "learning_rate": 1.8394669437435008e-05, + "loss": 1.0951, + "step": 16839 + }, + { + "epoch": 1.24, + "learning_rate": 1.839445267962835e-05, + "loss": 1.073, + "step": 16840 + }, + { + "epoch": 1.24, + "learning_rate": 1.8394235908466244e-05, + "loss": 0.9717, + "step": 16841 + }, + { + "epoch": 1.24, + "learning_rate": 1.839401912394903e-05, + "loss": 0.9664, + "step": 16842 + }, + { + "epoch": 1.24, + "learning_rate": 1.839380232607705e-05, + "loss": 1.0527, + "step": 16843 + }, + { + "epoch": 1.24, + "learning_rate": 1.839358551485065e-05, + "loss": 1.0325, + "step": 16844 + }, + { + "epoch": 1.24, + "learning_rate": 1.8393368690270176e-05, + "loss": 0.927, + "step": 16845 + }, + { + "epoch": 1.24, + "learning_rate": 1.839315185233597e-05, + "loss": 1.0157, + "step": 16846 + }, + { + "epoch": 1.24, + "learning_rate": 1.8392935001048383e-05, + "loss": 1.0874, + "step": 16847 + }, + { + "epoch": 1.24, + "learning_rate": 1.8392718136407755e-05, + "loss": 1.0778, + "step": 16848 + }, + { + "epoch": 1.24, + "learning_rate": 1.839250125841443e-05, + "loss": 0.9874, + "step": 16849 + }, + { + "epoch": 1.25, + "learning_rate": 1.839228436706876e-05, + "loss": 0.9403, + "step": 16850 + }, + { + "epoch": 1.25, + "learning_rate": 1.8392067462371087e-05, + "loss": 1.0032, + "step": 16851 + }, + { + "epoch": 1.25, + "learning_rate": 1.8391850544321748e-05, + "loss": 1.0741, + "step": 16852 + }, + { + "epoch": 1.25, + "learning_rate": 1.83916336129211e-05, + "loss": 1.0541, + "step": 16853 + }, + { + "epoch": 1.25, + "learning_rate": 1.839141666816948e-05, + "loss": 1.0323, + "step": 16854 + }, + { + "epoch": 1.25, + "learning_rate": 1.8391199710067237e-05, + "loss": 0.9818, + "step": 16855 + }, + { + "epoch": 1.25, + "learning_rate": 1.8390982738614717e-05, + "loss": 0.951, + "step": 16856 + }, + { + "epoch": 1.25, + "learning_rate": 1.8390765753812263e-05, + "loss": 0.907, + "step": 16857 + }, + { + "epoch": 1.25, + "learning_rate": 1.839054875566022e-05, + "loss": 1.0134, + "step": 16858 + }, + { + "epoch": 1.25, + "learning_rate": 1.8390331744158936e-05, + "loss": 0.9993, + "step": 16859 + }, + { + "epoch": 1.25, + "learning_rate": 1.839011471930875e-05, + "loss": 1.1039, + "step": 16860 + }, + { + "epoch": 1.25, + "learning_rate": 1.8389897681110015e-05, + "loss": 1.068, + "step": 16861 + }, + { + "epoch": 1.25, + "learning_rate": 1.8389680629563072e-05, + "loss": 1.1195, + "step": 16862 + }, + { + "epoch": 1.25, + "learning_rate": 1.838946356466827e-05, + "loss": 1.1294, + "step": 16863 + }, + { + "epoch": 1.25, + "learning_rate": 1.8389246486425945e-05, + "loss": 1.0174, + "step": 16864 + }, + { + "epoch": 1.25, + "learning_rate": 1.8389029394836452e-05, + "loss": 1.0454, + "step": 16865 + }, + { + "epoch": 1.25, + "learning_rate": 1.8388812289900135e-05, + "loss": 1.0489, + "step": 16866 + }, + { + "epoch": 1.25, + "learning_rate": 1.8388595171617337e-05, + "loss": 0.9884, + "step": 16867 + }, + { + "epoch": 1.25, + "learning_rate": 1.8388378039988403e-05, + "loss": 0.9903, + "step": 16868 + }, + { + "epoch": 1.25, + "learning_rate": 1.838816089501368e-05, + "loss": 1.0653, + "step": 16869 + }, + { + "epoch": 1.25, + "learning_rate": 1.838794373669351e-05, + "loss": 1.1201, + "step": 16870 + }, + { + "epoch": 1.25, + "learning_rate": 1.8387726565028247e-05, + "loss": 0.9781, + "step": 16871 + }, + { + "epoch": 1.25, + "learning_rate": 1.838750938001823e-05, + "loss": 1.1121, + "step": 16872 + }, + { + "epoch": 1.25, + "learning_rate": 1.8387292181663804e-05, + "loss": 0.9814, + "step": 16873 + }, + { + "epoch": 1.25, + "learning_rate": 1.8387074969965317e-05, + "loss": 1.0914, + "step": 16874 + }, + { + "epoch": 1.25, + "learning_rate": 1.8386857744923113e-05, + "loss": 1.0356, + "step": 16875 + }, + { + "epoch": 1.25, + "learning_rate": 1.8386640506537538e-05, + "loss": 1.0545, + "step": 16876 + }, + { + "epoch": 1.25, + "learning_rate": 1.8386423254808937e-05, + "loss": 1.0113, + "step": 16877 + }, + { + "epoch": 1.25, + "learning_rate": 1.8386205989737657e-05, + "loss": 0.9568, + "step": 16878 + }, + { + "epoch": 1.25, + "learning_rate": 1.8385988711324043e-05, + "loss": 1.0129, + "step": 16879 + }, + { + "epoch": 1.25, + "learning_rate": 1.8385771419568443e-05, + "loss": 1.0695, + "step": 16880 + }, + { + "epoch": 1.25, + "learning_rate": 1.83855541144712e-05, + "loss": 0.9751, + "step": 16881 + }, + { + "epoch": 1.25, + "learning_rate": 1.838533679603266e-05, + "loss": 1.1168, + "step": 16882 + }, + { + "epoch": 1.25, + "learning_rate": 1.8385119464253168e-05, + "loss": 1.0268, + "step": 16883 + }, + { + "epoch": 1.25, + "learning_rate": 1.8384902119133073e-05, + "loss": 1.07, + "step": 16884 + }, + { + "epoch": 1.25, + "learning_rate": 1.8384684760672717e-05, + "loss": 1.0734, + "step": 16885 + }, + { + "epoch": 1.25, + "learning_rate": 1.838446738887245e-05, + "loss": 1.087, + "step": 16886 + }, + { + "epoch": 1.25, + "learning_rate": 1.8384250003732613e-05, + "loss": 0.9597, + "step": 16887 + }, + { + "epoch": 1.25, + "learning_rate": 1.8384032605253555e-05, + "loss": 1.0837, + "step": 16888 + }, + { + "epoch": 1.25, + "learning_rate": 1.8383815193435623e-05, + "loss": 1.0792, + "step": 16889 + }, + { + "epoch": 1.25, + "learning_rate": 1.8383597768279154e-05, + "loss": 1.0836, + "step": 16890 + }, + { + "epoch": 1.25, + "learning_rate": 1.838338032978451e-05, + "loss": 0.9504, + "step": 16891 + }, + { + "epoch": 1.25, + "learning_rate": 1.8383162877952025e-05, + "loss": 1.0743, + "step": 16892 + }, + { + "epoch": 1.25, + "learning_rate": 1.8382945412782045e-05, + "loss": 0.996, + "step": 16893 + }, + { + "epoch": 1.25, + "learning_rate": 1.8382727934274918e-05, + "loss": 1.0146, + "step": 16894 + }, + { + "epoch": 1.25, + "learning_rate": 1.8382510442430993e-05, + "loss": 1.0434, + "step": 16895 + }, + { + "epoch": 1.25, + "learning_rate": 1.8382292937250613e-05, + "loss": 0.9871, + "step": 16896 + }, + { + "epoch": 1.25, + "learning_rate": 1.8382075418734125e-05, + "loss": 0.9727, + "step": 16897 + }, + { + "epoch": 1.25, + "learning_rate": 1.838185788688188e-05, + "loss": 1.0595, + "step": 16898 + }, + { + "epoch": 1.25, + "learning_rate": 1.8381640341694212e-05, + "loss": 0.8975, + "step": 16899 + }, + { + "epoch": 1.25, + "learning_rate": 1.8381422783171478e-05, + "loss": 0.9365, + "step": 16900 + }, + { + "epoch": 1.25, + "learning_rate": 1.838120521131402e-05, + "loss": 0.9077, + "step": 16901 + }, + { + "epoch": 1.25, + "learning_rate": 1.8380987626122184e-05, + "loss": 1.0513, + "step": 16902 + }, + { + "epoch": 1.25, + "learning_rate": 1.8380770027596316e-05, + "loss": 0.9914, + "step": 16903 + }, + { + "epoch": 1.25, + "learning_rate": 1.838055241573676e-05, + "loss": 1.0113, + "step": 16904 + }, + { + "epoch": 1.25, + "learning_rate": 1.838033479054387e-05, + "loss": 1.0266, + "step": 16905 + }, + { + "epoch": 1.25, + "learning_rate": 1.8380117152017986e-05, + "loss": 1.027, + "step": 16906 + }, + { + "epoch": 1.25, + "learning_rate": 1.8379899500159454e-05, + "loss": 1.0173, + "step": 16907 + }, + { + "epoch": 1.25, + "learning_rate": 1.8379681834968624e-05, + "loss": 1.0135, + "step": 16908 + }, + { + "epoch": 1.25, + "learning_rate": 1.8379464156445837e-05, + "loss": 0.9787, + "step": 16909 + }, + { + "epoch": 1.25, + "learning_rate": 1.8379246464591443e-05, + "loss": 1.0338, + "step": 16910 + }, + { + "epoch": 1.25, + "learning_rate": 1.837902875940579e-05, + "loss": 1.0581, + "step": 16911 + }, + { + "epoch": 1.25, + "learning_rate": 1.837881104088922e-05, + "loss": 1.0771, + "step": 16912 + }, + { + "epoch": 1.25, + "learning_rate": 1.8378593309042083e-05, + "loss": 0.9879, + "step": 16913 + }, + { + "epoch": 1.25, + "learning_rate": 1.8378375563864722e-05, + "loss": 0.989, + "step": 16914 + }, + { + "epoch": 1.25, + "learning_rate": 1.837815780535749e-05, + "loss": 0.9908, + "step": 16915 + }, + { + "epoch": 1.25, + "learning_rate": 1.8377940033520724e-05, + "loss": 1.0378, + "step": 16916 + }, + { + "epoch": 1.25, + "learning_rate": 1.8377722248354775e-05, + "loss": 1.0929, + "step": 16917 + }, + { + "epoch": 1.25, + "learning_rate": 1.8377504449859994e-05, + "loss": 1.0352, + "step": 16918 + }, + { + "epoch": 1.25, + "learning_rate": 1.8377286638036718e-05, + "loss": 0.9998, + "step": 16919 + }, + { + "epoch": 1.25, + "learning_rate": 1.8377068812885305e-05, + "loss": 0.9657, + "step": 16920 + }, + { + "epoch": 1.25, + "learning_rate": 1.837685097440609e-05, + "loss": 0.9401, + "step": 16921 + }, + { + "epoch": 1.25, + "learning_rate": 1.8376633122599427e-05, + "loss": 1.0234, + "step": 16922 + }, + { + "epoch": 1.25, + "learning_rate": 1.837641525746566e-05, + "loss": 1.028, + "step": 16923 + }, + { + "epoch": 1.25, + "learning_rate": 1.837619737900514e-05, + "loss": 1.0668, + "step": 16924 + }, + { + "epoch": 1.25, + "learning_rate": 1.8375979487218204e-05, + "loss": 1.0003, + "step": 16925 + }, + { + "epoch": 1.25, + "learning_rate": 1.8375761582105207e-05, + "loss": 0.9466, + "step": 16926 + }, + { + "epoch": 1.25, + "learning_rate": 1.8375543663666494e-05, + "loss": 1.0986, + "step": 16927 + }, + { + "epoch": 1.25, + "learning_rate": 1.8375325731902408e-05, + "loss": 0.8758, + "step": 16928 + }, + { + "epoch": 1.25, + "learning_rate": 1.83751077868133e-05, + "loss": 0.9324, + "step": 16929 + }, + { + "epoch": 1.25, + "learning_rate": 1.837488982839952e-05, + "loss": 1.0663, + "step": 16930 + }, + { + "epoch": 1.25, + "learning_rate": 1.8374671856661404e-05, + "loss": 0.9749, + "step": 16931 + }, + { + "epoch": 1.25, + "learning_rate": 1.8374453871599308e-05, + "loss": 1.0733, + "step": 16932 + }, + { + "epoch": 1.25, + "learning_rate": 1.8374235873213574e-05, + "loss": 1.0633, + "step": 16933 + }, + { + "epoch": 1.25, + "learning_rate": 1.837401786150455e-05, + "loss": 1.0461, + "step": 16934 + }, + { + "epoch": 1.25, + "learning_rate": 1.8373799836472585e-05, + "loss": 1.0018, + "step": 16935 + }, + { + "epoch": 1.25, + "learning_rate": 1.8373581798118026e-05, + "loss": 0.9628, + "step": 16936 + }, + { + "epoch": 1.25, + "learning_rate": 1.8373363746441213e-05, + "loss": 1.0169, + "step": 16937 + }, + { + "epoch": 1.25, + "learning_rate": 1.83731456814425e-05, + "loss": 0.9971, + "step": 16938 + }, + { + "epoch": 1.25, + "learning_rate": 1.8372927603122236e-05, + "loss": 0.9665, + "step": 16939 + }, + { + "epoch": 1.25, + "learning_rate": 1.837270951148076e-05, + "loss": 1.0384, + "step": 16940 + }, + { + "epoch": 1.25, + "learning_rate": 1.8372491406518426e-05, + "loss": 1.0842, + "step": 16941 + }, + { + "epoch": 1.25, + "learning_rate": 1.8372273288235574e-05, + "loss": 0.952, + "step": 16942 + }, + { + "epoch": 1.25, + "learning_rate": 1.8372055156632557e-05, + "loss": 1.1088, + "step": 16943 + }, + { + "epoch": 1.25, + "learning_rate": 1.8371837011709723e-05, + "loss": 1.032, + "step": 16944 + }, + { + "epoch": 1.25, + "learning_rate": 1.837161885346741e-05, + "loss": 1.046, + "step": 16945 + }, + { + "epoch": 1.25, + "learning_rate": 1.8371400681905977e-05, + "loss": 1.0637, + "step": 16946 + }, + { + "epoch": 1.25, + "learning_rate": 1.837118249702576e-05, + "loss": 1.1453, + "step": 16947 + }, + { + "epoch": 1.25, + "learning_rate": 1.8370964298827114e-05, + "loss": 1.0315, + "step": 16948 + }, + { + "epoch": 1.25, + "learning_rate": 1.8370746087310384e-05, + "loss": 1.017, + "step": 16949 + }, + { + "epoch": 1.25, + "learning_rate": 1.8370527862475917e-05, + "loss": 0.9836, + "step": 16950 + }, + { + "epoch": 1.25, + "learning_rate": 1.837030962432406e-05, + "loss": 0.8917, + "step": 16951 + }, + { + "epoch": 1.25, + "learning_rate": 1.8370091372855156e-05, + "loss": 1.0831, + "step": 16952 + }, + { + "epoch": 1.25, + "learning_rate": 1.836987310806956e-05, + "loss": 1.0212, + "step": 16953 + }, + { + "epoch": 1.25, + "learning_rate": 1.8369654829967617e-05, + "loss": 1.0938, + "step": 16954 + }, + { + "epoch": 1.25, + "learning_rate": 1.8369436538549673e-05, + "loss": 0.9585, + "step": 16955 + }, + { + "epoch": 1.25, + "learning_rate": 1.8369218233816072e-05, + "loss": 0.9927, + "step": 16956 + }, + { + "epoch": 1.25, + "learning_rate": 1.8368999915767168e-05, + "loss": 1.1167, + "step": 16957 + }, + { + "epoch": 1.25, + "learning_rate": 1.83687815844033e-05, + "loss": 1.0282, + "step": 16958 + }, + { + "epoch": 1.25, + "learning_rate": 1.8368563239724823e-05, + "loss": 0.9788, + "step": 16959 + }, + { + "epoch": 1.25, + "learning_rate": 1.8368344881732084e-05, + "loss": 0.9887, + "step": 16960 + }, + { + "epoch": 1.25, + "learning_rate": 1.8368126510425424e-05, + "loss": 1.0025, + "step": 16961 + }, + { + "epoch": 1.25, + "learning_rate": 1.83679081258052e-05, + "loss": 1.1565, + "step": 16962 + }, + { + "epoch": 1.25, + "learning_rate": 1.8367689727871746e-05, + "loss": 1.137, + "step": 16963 + }, + { + "epoch": 1.25, + "learning_rate": 1.8367471316625422e-05, + "loss": 1.0391, + "step": 16964 + }, + { + "epoch": 1.25, + "learning_rate": 1.8367252892066574e-05, + "loss": 1.0944, + "step": 16965 + }, + { + "epoch": 1.25, + "learning_rate": 1.836703445419554e-05, + "loss": 1.0331, + "step": 16966 + }, + { + "epoch": 1.25, + "learning_rate": 1.8366816003012678e-05, + "loss": 0.9879, + "step": 16967 + }, + { + "epoch": 1.25, + "learning_rate": 1.836659753851833e-05, + "loss": 1.0888, + "step": 16968 + }, + { + "epoch": 1.25, + "learning_rate": 1.8366379060712845e-05, + "loss": 0.9025, + "step": 16969 + }, + { + "epoch": 1.25, + "learning_rate": 1.836616056959657e-05, + "loss": 1.0713, + "step": 16970 + }, + { + "epoch": 1.25, + "learning_rate": 1.8365942065169856e-05, + "loss": 1.0665, + "step": 16971 + }, + { + "epoch": 1.25, + "learning_rate": 1.8365723547433046e-05, + "loss": 1.0252, + "step": 16972 + }, + { + "epoch": 1.25, + "learning_rate": 1.836550501638649e-05, + "loss": 0.9905, + "step": 16973 + }, + { + "epoch": 1.25, + "learning_rate": 1.8365286472030533e-05, + "loss": 1.1304, + "step": 16974 + }, + { + "epoch": 1.25, + "learning_rate": 1.8365067914365525e-05, + "loss": 1.063, + "step": 16975 + }, + { + "epoch": 1.25, + "learning_rate": 1.8364849343391817e-05, + "loss": 1.066, + "step": 16976 + }, + { + "epoch": 1.25, + "learning_rate": 1.836463075910975e-05, + "loss": 1.1266, + "step": 16977 + }, + { + "epoch": 1.25, + "learning_rate": 1.836441216151968e-05, + "loss": 1.0467, + "step": 16978 + }, + { + "epoch": 1.25, + "learning_rate": 1.8364193550621943e-05, + "loss": 0.9757, + "step": 16979 + }, + { + "epoch": 1.25, + "learning_rate": 1.83639749264169e-05, + "loss": 0.9962, + "step": 16980 + }, + { + "epoch": 1.25, + "learning_rate": 1.8363756288904888e-05, + "loss": 0.9404, + "step": 16981 + }, + { + "epoch": 1.25, + "learning_rate": 1.8363537638086263e-05, + "loss": 0.9691, + "step": 16982 + }, + { + "epoch": 1.25, + "learning_rate": 1.8363318973961366e-05, + "loss": 0.9342, + "step": 16983 + }, + { + "epoch": 1.25, + "learning_rate": 1.8363100296530548e-05, + "loss": 1.0507, + "step": 16984 + }, + { + "epoch": 1.26, + "learning_rate": 1.836288160579416e-05, + "loss": 1.047, + "step": 16985 + }, + { + "epoch": 1.26, + "learning_rate": 1.836266290175255e-05, + "loss": 0.9423, + "step": 16986 + }, + { + "epoch": 1.26, + "learning_rate": 1.8362444184406055e-05, + "loss": 0.999, + "step": 16987 + }, + { + "epoch": 1.26, + "learning_rate": 1.8362225453755035e-05, + "loss": 1.0558, + "step": 16988 + }, + { + "epoch": 1.26, + "learning_rate": 1.8362006709799835e-05, + "loss": 0.9876, + "step": 16989 + }, + { + "epoch": 1.26, + "learning_rate": 1.83617879525408e-05, + "loss": 0.934, + "step": 16990 + }, + { + "epoch": 1.26, + "learning_rate": 1.8361569181978282e-05, + "loss": 1.054, + "step": 16991 + }, + { + "epoch": 1.26, + "learning_rate": 1.8361350398112624e-05, + "loss": 1.0317, + "step": 16992 + }, + { + "epoch": 1.26, + "learning_rate": 1.836113160094418e-05, + "loss": 1.0559, + "step": 16993 + }, + { + "epoch": 1.26, + "learning_rate": 1.8360912790473293e-05, + "loss": 0.9662, + "step": 16994 + }, + { + "epoch": 1.26, + "learning_rate": 1.8360693966700315e-05, + "loss": 1.006, + "step": 16995 + }, + { + "epoch": 1.26, + "learning_rate": 1.836047512962559e-05, + "loss": 0.9199, + "step": 16996 + }, + { + "epoch": 1.26, + "learning_rate": 1.836025627924947e-05, + "loss": 1.0706, + "step": 16997 + }, + { + "epoch": 1.26, + "learning_rate": 1.8360037415572306e-05, + "loss": 1.051, + "step": 16998 + }, + { + "epoch": 1.26, + "learning_rate": 1.8359818538594437e-05, + "loss": 0.968, + "step": 16999 + }, + { + "epoch": 1.26, + "learning_rate": 1.8359599648316216e-05, + "loss": 0.8902, + "step": 17000 + }, + { + "epoch": 1.26, + "learning_rate": 1.8359380744738e-05, + "loss": 1.065, + "step": 17001 + }, + { + "epoch": 1.26, + "learning_rate": 1.8359161827860118e-05, + "loss": 1.0996, + "step": 17002 + }, + { + "epoch": 1.26, + "learning_rate": 1.8358942897682935e-05, + "loss": 0.9666, + "step": 17003 + }, + { + "epoch": 1.26, + "learning_rate": 1.8358723954206793e-05, + "loss": 1.0803, + "step": 17004 + }, + { + "epoch": 1.26, + "learning_rate": 1.8358504997432038e-05, + "loss": 1.0264, + "step": 17005 + }, + { + "epoch": 1.26, + "learning_rate": 1.8358286027359024e-05, + "loss": 0.9412, + "step": 17006 + }, + { + "epoch": 1.26, + "learning_rate": 1.8358067043988094e-05, + "loss": 1.042, + "step": 17007 + }, + { + "epoch": 1.26, + "learning_rate": 1.83578480473196e-05, + "loss": 1.1302, + "step": 17008 + }, + { + "epoch": 1.26, + "learning_rate": 1.8357629037353893e-05, + "loss": 1.0381, + "step": 17009 + }, + { + "epoch": 1.26, + "learning_rate": 1.8357410014091313e-05, + "loss": 0.9651, + "step": 17010 + }, + { + "epoch": 1.26, + "learning_rate": 1.8357190977532214e-05, + "loss": 0.9295, + "step": 17011 + }, + { + "epoch": 1.26, + "learning_rate": 1.8356971927676946e-05, + "loss": 0.9294, + "step": 17012 + }, + { + "epoch": 1.26, + "learning_rate": 1.8356752864525854e-05, + "loss": 1.098, + "step": 17013 + }, + { + "epoch": 1.26, + "learning_rate": 1.8356533788079284e-05, + "loss": 0.9274, + "step": 17014 + }, + { + "epoch": 1.26, + "learning_rate": 1.8356314698337595e-05, + "loss": 0.9867, + "step": 17015 + }, + { + "epoch": 1.26, + "learning_rate": 1.8356095595301123e-05, + "loss": 0.9868, + "step": 17016 + }, + { + "epoch": 1.26, + "learning_rate": 1.8355876478970226e-05, + "loss": 1.057, + "step": 17017 + }, + { + "epoch": 1.26, + "learning_rate": 1.8355657349345248e-05, + "loss": 1.1705, + "step": 17018 + }, + { + "epoch": 1.26, + "learning_rate": 1.8355438206426538e-05, + "loss": 1.1209, + "step": 17019 + }, + { + "epoch": 1.26, + "learning_rate": 1.8355219050214446e-05, + "loss": 1.0258, + "step": 17020 + }, + { + "epoch": 1.26, + "learning_rate": 1.8354999880709324e-05, + "loss": 0.9624, + "step": 17021 + }, + { + "epoch": 1.26, + "learning_rate": 1.835478069791151e-05, + "loss": 1.0075, + "step": 17022 + }, + { + "epoch": 1.26, + "learning_rate": 1.8354561501821365e-05, + "loss": 1.0126, + "step": 17023 + }, + { + "epoch": 1.26, + "learning_rate": 1.835434229243923e-05, + "loss": 1.0199, + "step": 17024 + }, + { + "epoch": 1.26, + "learning_rate": 1.8354123069765456e-05, + "loss": 0.9443, + "step": 17025 + }, + { + "epoch": 1.26, + "learning_rate": 1.8353903833800387e-05, + "loss": 1.0819, + "step": 17026 + }, + { + "epoch": 1.26, + "learning_rate": 1.8353684584544386e-05, + "loss": 0.9531, + "step": 17027 + }, + { + "epoch": 1.26, + "learning_rate": 1.8353465321997787e-05, + "loss": 0.9462, + "step": 17028 + }, + { + "epoch": 1.26, + "learning_rate": 1.8353246046160942e-05, + "loss": 0.9983, + "step": 17029 + }, + { + "epoch": 1.26, + "learning_rate": 1.8353026757034206e-05, + "loss": 1.1125, + "step": 17030 + }, + { + "epoch": 1.26, + "learning_rate": 1.835280745461792e-05, + "loss": 1.0551, + "step": 17031 + }, + { + "epoch": 1.26, + "learning_rate": 1.835258813891244e-05, + "loss": 0.9451, + "step": 17032 + }, + { + "epoch": 1.26, + "learning_rate": 1.835236880991811e-05, + "loss": 1.069, + "step": 17033 + }, + { + "epoch": 1.26, + "learning_rate": 1.8352149467635283e-05, + "loss": 1.0516, + "step": 17034 + }, + { + "epoch": 1.26, + "learning_rate": 1.8351930112064304e-05, + "loss": 0.9433, + "step": 17035 + }, + { + "epoch": 1.26, + "learning_rate": 1.8351710743205525e-05, + "loss": 1.1365, + "step": 17036 + }, + { + "epoch": 1.26, + "learning_rate": 1.8351491361059294e-05, + "loss": 0.9374, + "step": 17037 + }, + { + "epoch": 1.26, + "learning_rate": 1.8351271965625958e-05, + "loss": 0.9806, + "step": 17038 + }, + { + "epoch": 1.26, + "learning_rate": 1.8351052556905867e-05, + "loss": 1.0177, + "step": 17039 + }, + { + "epoch": 1.26, + "learning_rate": 1.8350833134899373e-05, + "loss": 0.9126, + "step": 17040 + }, + { + "epoch": 1.26, + "learning_rate": 1.835061369960682e-05, + "loss": 1.0056, + "step": 17041 + }, + { + "epoch": 1.26, + "learning_rate": 1.835039425102856e-05, + "loss": 0.9735, + "step": 17042 + }, + { + "epoch": 1.26, + "learning_rate": 1.835017478916495e-05, + "loss": 1.0322, + "step": 17043 + }, + { + "epoch": 1.26, + "learning_rate": 1.8349955314016322e-05, + "loss": 0.9993, + "step": 17044 + }, + { + "epoch": 1.26, + "learning_rate": 1.8349735825583038e-05, + "loss": 1.0617, + "step": 17045 + }, + { + "epoch": 1.26, + "learning_rate": 1.8349516323865444e-05, + "loss": 1.0144, + "step": 17046 + }, + { + "epoch": 1.26, + "learning_rate": 1.834929680886389e-05, + "loss": 1.0201, + "step": 17047 + }, + { + "epoch": 1.26, + "learning_rate": 1.8349077280578726e-05, + "loss": 0.9087, + "step": 17048 + }, + { + "epoch": 1.26, + "learning_rate": 1.8348857739010296e-05, + "loss": 1.0211, + "step": 17049 + }, + { + "epoch": 1.26, + "learning_rate": 1.8348638184158952e-05, + "loss": 1.0038, + "step": 17050 + }, + { + "epoch": 1.26, + "learning_rate": 1.8348418616025044e-05, + "loss": 0.9958, + "step": 17051 + }, + { + "epoch": 1.26, + "learning_rate": 1.8348199034608925e-05, + "loss": 0.9531, + "step": 17052 + }, + { + "epoch": 1.26, + "learning_rate": 1.834797943991094e-05, + "loss": 1.0575, + "step": 17053 + }, + { + "epoch": 1.26, + "learning_rate": 1.8347759831931436e-05, + "loss": 0.904, + "step": 17054 + }, + { + "epoch": 1.26, + "learning_rate": 1.8347540210670768e-05, + "loss": 1.0307, + "step": 17055 + }, + { + "epoch": 1.26, + "learning_rate": 1.8347320576129284e-05, + "loss": 1.0496, + "step": 17056 + }, + { + "epoch": 1.26, + "learning_rate": 1.834710092830733e-05, + "loss": 0.9985, + "step": 17057 + }, + { + "epoch": 1.26, + "learning_rate": 1.834688126720526e-05, + "loss": 1.1238, + "step": 17058 + }, + { + "epoch": 1.26, + "learning_rate": 1.834666159282342e-05, + "loss": 1.0409, + "step": 17059 + }, + { + "epoch": 1.26, + "learning_rate": 1.8346441905162162e-05, + "loss": 0.9262, + "step": 17060 + }, + { + "epoch": 1.26, + "learning_rate": 1.8346222204221833e-05, + "loss": 1.0543, + "step": 17061 + }, + { + "epoch": 1.26, + "learning_rate": 1.8346002490002782e-05, + "loss": 1.0912, + "step": 17062 + }, + { + "epoch": 1.26, + "learning_rate": 1.8345782762505365e-05, + "loss": 0.9486, + "step": 17063 + }, + { + "epoch": 1.26, + "learning_rate": 1.8345563021729927e-05, + "loss": 0.9031, + "step": 17064 + }, + { + "epoch": 1.26, + "learning_rate": 1.834534326767681e-05, + "loss": 1.0975, + "step": 17065 + }, + { + "epoch": 1.26, + "learning_rate": 1.834512350034638e-05, + "loss": 1.0581, + "step": 17066 + }, + { + "epoch": 1.26, + "learning_rate": 1.8344903719738972e-05, + "loss": 1.0686, + "step": 17067 + }, + { + "epoch": 1.26, + "learning_rate": 1.8344683925854948e-05, + "loss": 0.9977, + "step": 17068 + }, + { + "epoch": 1.26, + "learning_rate": 1.8344464118694647e-05, + "loss": 1.0302, + "step": 17069 + }, + { + "epoch": 1.26, + "learning_rate": 1.8344244298258424e-05, + "loss": 0.9492, + "step": 17070 + }, + { + "epoch": 1.26, + "learning_rate": 1.8344024464546626e-05, + "loss": 1.0599, + "step": 17071 + }, + { + "epoch": 1.26, + "learning_rate": 1.8343804617559606e-05, + "loss": 1.0915, + "step": 17072 + }, + { + "epoch": 1.26, + "learning_rate": 1.8343584757297712e-05, + "loss": 1.0368, + "step": 17073 + }, + { + "epoch": 1.26, + "learning_rate": 1.8343364883761295e-05, + "loss": 1.0631, + "step": 17074 + }, + { + "epoch": 1.26, + "learning_rate": 1.83431449969507e-05, + "loss": 0.9745, + "step": 17075 + }, + { + "epoch": 1.26, + "learning_rate": 1.8342925096866283e-05, + "loss": 0.9544, + "step": 17076 + }, + { + "epoch": 1.26, + "learning_rate": 1.8342705183508394e-05, + "loss": 1.0924, + "step": 17077 + }, + { + "epoch": 1.26, + "learning_rate": 1.8342485256877376e-05, + "loss": 1.0462, + "step": 17078 + }, + { + "epoch": 1.26, + "learning_rate": 1.8342265316973586e-05, + "loss": 1.0334, + "step": 17079 + }, + { + "epoch": 1.26, + "learning_rate": 1.8342045363797372e-05, + "loss": 1.0891, + "step": 17080 + }, + { + "epoch": 1.26, + "learning_rate": 1.834182539734908e-05, + "loss": 0.9764, + "step": 17081 + }, + { + "epoch": 1.26, + "learning_rate": 1.8341605417629064e-05, + "loss": 0.9244, + "step": 17082 + }, + { + "epoch": 1.26, + "learning_rate": 1.8341385424637677e-05, + "loss": 1.0042, + "step": 17083 + }, + { + "epoch": 1.26, + "learning_rate": 1.8341165418375266e-05, + "loss": 1.1186, + "step": 17084 + }, + { + "epoch": 1.26, + "learning_rate": 1.8340945398842172e-05, + "loss": 1.0859, + "step": 17085 + }, + { + "epoch": 1.26, + "learning_rate": 1.834072536603876e-05, + "loss": 1.136, + "step": 17086 + }, + { + "epoch": 1.26, + "learning_rate": 1.834050531996537e-05, + "loss": 1.1347, + "step": 17087 + }, + { + "epoch": 1.26, + "learning_rate": 1.8340285260622358e-05, + "loss": 0.9685, + "step": 17088 + }, + { + "epoch": 1.26, + "learning_rate": 1.8340065188010067e-05, + "loss": 1.0179, + "step": 17089 + }, + { + "epoch": 1.26, + "learning_rate": 1.8339845102128856e-05, + "loss": 1.004, + "step": 17090 + }, + { + "epoch": 1.26, + "learning_rate": 1.833962500297907e-05, + "loss": 1.0701, + "step": 17091 + }, + { + "epoch": 1.26, + "learning_rate": 1.8339404890561056e-05, + "loss": 0.9659, + "step": 17092 + }, + { + "epoch": 1.26, + "learning_rate": 1.8339184764875175e-05, + "loss": 1.0196, + "step": 17093 + }, + { + "epoch": 1.26, + "learning_rate": 1.8338964625921766e-05, + "loss": 1.0591, + "step": 17094 + }, + { + "epoch": 1.26, + "learning_rate": 1.833874447370118e-05, + "loss": 0.9713, + "step": 17095 + }, + { + "epoch": 1.26, + "learning_rate": 1.8338524308213778e-05, + "loss": 0.9425, + "step": 17096 + }, + { + "epoch": 1.26, + "learning_rate": 1.8338304129459904e-05, + "loss": 0.9916, + "step": 17097 + }, + { + "epoch": 1.26, + "learning_rate": 1.83380839374399e-05, + "loss": 0.9889, + "step": 17098 + }, + { + "epoch": 1.26, + "learning_rate": 1.833786373215413e-05, + "loss": 1.04, + "step": 17099 + }, + { + "epoch": 1.26, + "learning_rate": 1.8337643513602933e-05, + "loss": 0.9812, + "step": 17100 + }, + { + "epoch": 1.26, + "learning_rate": 1.8337423281786667e-05, + "loss": 1.0612, + "step": 17101 + }, + { + "epoch": 1.26, + "learning_rate": 1.833720303670568e-05, + "loss": 0.9661, + "step": 17102 + }, + { + "epoch": 1.26, + "learning_rate": 1.8336982778360325e-05, + "loss": 0.9477, + "step": 17103 + }, + { + "epoch": 1.26, + "learning_rate": 1.8336762506750945e-05, + "loss": 1.1187, + "step": 17104 + }, + { + "epoch": 1.26, + "learning_rate": 1.8336542221877897e-05, + "loss": 1.0593, + "step": 17105 + }, + { + "epoch": 1.26, + "learning_rate": 1.833632192374153e-05, + "loss": 1.095, + "step": 17106 + }, + { + "epoch": 1.26, + "learning_rate": 1.8336101612342193e-05, + "loss": 1.0224, + "step": 17107 + }, + { + "epoch": 1.26, + "learning_rate": 1.8335881287680236e-05, + "loss": 0.9394, + "step": 17108 + }, + { + "epoch": 1.26, + "learning_rate": 1.8335660949756016e-05, + "loss": 1.0153, + "step": 17109 + }, + { + "epoch": 1.26, + "learning_rate": 1.833544059856988e-05, + "loss": 0.994, + "step": 17110 + }, + { + "epoch": 1.26, + "learning_rate": 1.833522023412217e-05, + "loss": 1.0011, + "step": 17111 + }, + { + "epoch": 1.26, + "learning_rate": 1.8334999856413248e-05, + "loss": 1.0259, + "step": 17112 + }, + { + "epoch": 1.26, + "learning_rate": 1.833477946544346e-05, + "loss": 0.9829, + "step": 17113 + }, + { + "epoch": 1.26, + "learning_rate": 1.8334559061213157e-05, + "loss": 1.04, + "step": 17114 + }, + { + "epoch": 1.26, + "learning_rate": 1.8334338643722687e-05, + "loss": 1.0864, + "step": 17115 + }, + { + "epoch": 1.26, + "learning_rate": 1.8334118212972408e-05, + "loss": 1.0909, + "step": 17116 + }, + { + "epoch": 1.26, + "learning_rate": 1.8333897768962665e-05, + "loss": 1.1016, + "step": 17117 + }, + { + "epoch": 1.26, + "learning_rate": 1.8333677311693807e-05, + "loss": 0.9948, + "step": 17118 + }, + { + "epoch": 1.26, + "learning_rate": 1.833345684116619e-05, + "loss": 0.8913, + "step": 17119 + }, + { + "epoch": 1.26, + "learning_rate": 1.8333236357380163e-05, + "loss": 0.9765, + "step": 17120 + }, + { + "epoch": 1.27, + "learning_rate": 1.8333015860336077e-05, + "loss": 0.9794, + "step": 17121 + }, + { + "epoch": 1.27, + "learning_rate": 1.833279535003428e-05, + "loss": 0.9539, + "step": 17122 + }, + { + "epoch": 1.27, + "learning_rate": 1.8332574826475126e-05, + "loss": 1.0676, + "step": 17123 + }, + { + "epoch": 1.27, + "learning_rate": 1.833235428965896e-05, + "loss": 1.0093, + "step": 17124 + }, + { + "epoch": 1.27, + "learning_rate": 1.833213373958614e-05, + "loss": 0.9403, + "step": 17125 + }, + { + "epoch": 1.27, + "learning_rate": 1.8331913176257018e-05, + "loss": 0.9753, + "step": 17126 + }, + { + "epoch": 1.27, + "learning_rate": 1.833169259967194e-05, + "loss": 0.9596, + "step": 17127 + }, + { + "epoch": 1.27, + "learning_rate": 1.8331472009831256e-05, + "loss": 0.9858, + "step": 17128 + }, + { + "epoch": 1.27, + "learning_rate": 1.8331251406735326e-05, + "loss": 1.0165, + "step": 17129 + }, + { + "epoch": 1.27, + "learning_rate": 1.8331030790384485e-05, + "loss": 0.9912, + "step": 17130 + }, + { + "epoch": 1.27, + "learning_rate": 1.8330810160779096e-05, + "loss": 0.9768, + "step": 17131 + }, + { + "epoch": 1.27, + "learning_rate": 1.833058951791951e-05, + "loss": 0.7893, + "step": 17132 + }, + { + "epoch": 1.27, + "learning_rate": 1.833036886180607e-05, + "loss": 1.0466, + "step": 17133 + }, + { + "epoch": 1.27, + "learning_rate": 1.833014819243914e-05, + "loss": 1.0826, + "step": 17134 + }, + { + "epoch": 1.27, + "learning_rate": 1.8329927509819057e-05, + "loss": 0.9716, + "step": 17135 + }, + { + "epoch": 1.27, + "learning_rate": 1.832970681394618e-05, + "loss": 1.0646, + "step": 17136 + }, + { + "epoch": 1.27, + "learning_rate": 1.8329486104820862e-05, + "loss": 1.0885, + "step": 17137 + }, + { + "epoch": 1.27, + "learning_rate": 1.8329265382443448e-05, + "loss": 1.0442, + "step": 17138 + }, + { + "epoch": 1.27, + "learning_rate": 1.832904464681429e-05, + "loss": 0.9466, + "step": 17139 + }, + { + "epoch": 1.27, + "learning_rate": 1.8328823897933745e-05, + "loss": 1.0452, + "step": 17140 + }, + { + "epoch": 1.27, + "learning_rate": 1.8328603135802157e-05, + "loss": 0.9322, + "step": 17141 + }, + { + "epoch": 1.27, + "learning_rate": 1.8328382360419884e-05, + "loss": 1.0114, + "step": 17142 + }, + { + "epoch": 1.27, + "learning_rate": 1.8328161571787272e-05, + "loss": 0.8983, + "step": 17143 + }, + { + "epoch": 1.27, + "learning_rate": 1.8327940769904676e-05, + "loss": 0.9344, + "step": 17144 + }, + { + "epoch": 1.27, + "learning_rate": 1.8327719954772444e-05, + "loss": 0.9869, + "step": 17145 + }, + { + "epoch": 1.27, + "learning_rate": 1.8327499126390927e-05, + "loss": 0.9444, + "step": 17146 + }, + { + "epoch": 1.27, + "learning_rate": 1.832727828476048e-05, + "loss": 1.0687, + "step": 17147 + }, + { + "epoch": 1.27, + "learning_rate": 1.832705742988145e-05, + "loss": 1.0074, + "step": 17148 + }, + { + "epoch": 1.27, + "learning_rate": 1.8326836561754193e-05, + "loss": 0.9836, + "step": 17149 + }, + { + "epoch": 1.27, + "learning_rate": 1.832661568037906e-05, + "loss": 0.9762, + "step": 17150 + }, + { + "epoch": 1.27, + "learning_rate": 1.8326394785756397e-05, + "loss": 1.1354, + "step": 17151 + }, + { + "epoch": 1.27, + "learning_rate": 1.832617387788656e-05, + "loss": 0.9747, + "step": 17152 + }, + { + "epoch": 1.27, + "learning_rate": 1.83259529567699e-05, + "loss": 0.9838, + "step": 17153 + }, + { + "epoch": 1.27, + "learning_rate": 1.8325732022406766e-05, + "loss": 0.931, + "step": 17154 + }, + { + "epoch": 1.27, + "learning_rate": 1.8325511074797515e-05, + "loss": 1.0627, + "step": 17155 + }, + { + "epoch": 1.27, + "learning_rate": 1.8325290113942494e-05, + "loss": 1.0506, + "step": 17156 + }, + { + "epoch": 1.27, + "learning_rate": 1.8325069139842053e-05, + "loss": 1.054, + "step": 17157 + }, + { + "epoch": 1.27, + "learning_rate": 1.8324848152496548e-05, + "loss": 1.0074, + "step": 17158 + }, + { + "epoch": 1.27, + "learning_rate": 1.8324627151906328e-05, + "loss": 1.017, + "step": 17159 + }, + { + "epoch": 1.27, + "learning_rate": 1.8324406138071747e-05, + "loss": 1.1206, + "step": 17160 + }, + { + "epoch": 1.27, + "learning_rate": 1.8324185110993153e-05, + "loss": 1.0418, + "step": 17161 + }, + { + "epoch": 1.27, + "learning_rate": 1.83239640706709e-05, + "loss": 0.9785, + "step": 17162 + }, + { + "epoch": 1.27, + "learning_rate": 1.8323743017105342e-05, + "loss": 1.101, + "step": 17163 + }, + { + "epoch": 1.27, + "learning_rate": 1.8323521950296827e-05, + "loss": 0.9789, + "step": 17164 + }, + { + "epoch": 1.27, + "learning_rate": 1.8323300870245703e-05, + "loss": 0.9914, + "step": 17165 + }, + { + "epoch": 1.27, + "learning_rate": 1.832307977695233e-05, + "loss": 1.0653, + "step": 17166 + }, + { + "epoch": 1.27, + "learning_rate": 1.8322858670417056e-05, + "loss": 0.8686, + "step": 17167 + }, + { + "epoch": 1.27, + "learning_rate": 1.8322637550640233e-05, + "loss": 1.0069, + "step": 17168 + }, + { + "epoch": 1.27, + "learning_rate": 1.832241641762221e-05, + "loss": 1.0425, + "step": 17169 + }, + { + "epoch": 1.27, + "learning_rate": 1.8322195271363344e-05, + "loss": 0.9918, + "step": 17170 + }, + { + "epoch": 1.27, + "learning_rate": 1.8321974111863986e-05, + "loss": 0.932, + "step": 17171 + }, + { + "epoch": 1.27, + "learning_rate": 1.8321752939124487e-05, + "loss": 1.0811, + "step": 17172 + }, + { + "epoch": 1.27, + "learning_rate": 1.8321531753145195e-05, + "loss": 0.9275, + "step": 17173 + }, + { + "epoch": 1.27, + "learning_rate": 1.8321310553926467e-05, + "loss": 1.0411, + "step": 17174 + }, + { + "epoch": 1.27, + "learning_rate": 1.832108934146865e-05, + "loss": 1.0752, + "step": 17175 + }, + { + "epoch": 1.27, + "learning_rate": 1.8320868115772103e-05, + "loss": 1.0416, + "step": 17176 + }, + { + "epoch": 1.27, + "learning_rate": 1.832064687683717e-05, + "loss": 1.053, + "step": 17177 + }, + { + "epoch": 1.27, + "learning_rate": 1.832042562466421e-05, + "loss": 1.0421, + "step": 17178 + }, + { + "epoch": 1.27, + "learning_rate": 1.8320204359253567e-05, + "loss": 1.0119, + "step": 17179 + }, + { + "epoch": 1.27, + "learning_rate": 1.8319983080605604e-05, + "loss": 1.1892, + "step": 17180 + }, + { + "epoch": 1.27, + "learning_rate": 1.8319761788720662e-05, + "loss": 1.0446, + "step": 17181 + }, + { + "epoch": 1.27, + "learning_rate": 1.83195404835991e-05, + "loss": 1.0656, + "step": 17182 + }, + { + "epoch": 1.27, + "learning_rate": 1.831931916524127e-05, + "loss": 1.0918, + "step": 17183 + }, + { + "epoch": 1.27, + "learning_rate": 1.831909783364752e-05, + "loss": 0.9645, + "step": 17184 + }, + { + "epoch": 1.27, + "learning_rate": 1.8318876488818204e-05, + "loss": 1.031, + "step": 17185 + }, + { + "epoch": 1.27, + "learning_rate": 1.831865513075368e-05, + "loss": 0.9746, + "step": 17186 + }, + { + "epoch": 1.27, + "learning_rate": 1.831843375945429e-05, + "loss": 1.0756, + "step": 17187 + }, + { + "epoch": 1.27, + "learning_rate": 1.8318212374920387e-05, + "loss": 1.0697, + "step": 17188 + }, + { + "epoch": 1.27, + "learning_rate": 1.831799097715233e-05, + "loss": 1.0136, + "step": 17189 + }, + { + "epoch": 1.27, + "learning_rate": 1.8317769566150474e-05, + "loss": 1.1039, + "step": 17190 + }, + { + "epoch": 1.27, + "learning_rate": 1.831754814191516e-05, + "loss": 1.1422, + "step": 17191 + }, + { + "epoch": 1.27, + "learning_rate": 1.8317326704446747e-05, + "loss": 1.1396, + "step": 17192 + }, + { + "epoch": 1.27, + "learning_rate": 1.8317105253745587e-05, + "loss": 1.0214, + "step": 17193 + }, + { + "epoch": 1.27, + "learning_rate": 1.831688378981203e-05, + "loss": 0.9171, + "step": 17194 + }, + { + "epoch": 1.27, + "learning_rate": 1.831666231264643e-05, + "loss": 0.9388, + "step": 17195 + }, + { + "epoch": 1.27, + "learning_rate": 1.831644082224914e-05, + "loss": 1.0894, + "step": 17196 + }, + { + "epoch": 1.27, + "learning_rate": 1.8316219318620514e-05, + "loss": 1.0879, + "step": 17197 + }, + { + "epoch": 1.27, + "learning_rate": 1.8315997801760897e-05, + "loss": 1.077, + "step": 17198 + }, + { + "epoch": 1.27, + "learning_rate": 1.8315776271670652e-05, + "loss": 1.0213, + "step": 17199 + }, + { + "epoch": 1.27, + "learning_rate": 1.831555472835012e-05, + "loss": 0.921, + "step": 17200 + }, + { + "epoch": 1.27, + "learning_rate": 1.8315333171799665e-05, + "loss": 0.9309, + "step": 17201 + }, + { + "epoch": 1.27, + "learning_rate": 1.831511160201963e-05, + "loss": 1.0569, + "step": 17202 + }, + { + "epoch": 1.27, + "learning_rate": 1.8314890019010373e-05, + "loss": 0.9137, + "step": 17203 + }, + { + "epoch": 1.27, + "learning_rate": 1.8314668422772245e-05, + "loss": 1.061, + "step": 17204 + }, + { + "epoch": 1.27, + "learning_rate": 1.83144468133056e-05, + "loss": 0.9947, + "step": 17205 + }, + { + "epoch": 1.27, + "learning_rate": 1.8314225190610785e-05, + "loss": 1.1012, + "step": 17206 + }, + { + "epoch": 1.27, + "learning_rate": 1.831400355468816e-05, + "loss": 1.064, + "step": 17207 + }, + { + "epoch": 1.27, + "learning_rate": 1.8313781905538074e-05, + "loss": 1.0532, + "step": 17208 + }, + { + "epoch": 1.27, + "learning_rate": 1.831356024316088e-05, + "loss": 1.0326, + "step": 17209 + }, + { + "epoch": 1.27, + "learning_rate": 1.831333856755693e-05, + "loss": 1.0139, + "step": 17210 + }, + { + "epoch": 1.27, + "learning_rate": 1.8313116878726575e-05, + "loss": 1.0834, + "step": 17211 + }, + { + "epoch": 1.27, + "learning_rate": 1.8312895176670173e-05, + "loss": 1.0796, + "step": 17212 + }, + { + "epoch": 1.27, + "learning_rate": 1.8312673461388077e-05, + "loss": 1.0454, + "step": 17213 + }, + { + "epoch": 1.27, + "learning_rate": 1.8312451732880628e-05, + "loss": 0.9671, + "step": 17214 + }, + { + "epoch": 1.27, + "learning_rate": 1.8312229991148193e-05, + "loss": 1.0472, + "step": 17215 + }, + { + "epoch": 1.27, + "learning_rate": 1.8312008236191118e-05, + "loss": 1.0109, + "step": 17216 + }, + { + "epoch": 1.27, + "learning_rate": 1.8311786468009757e-05, + "loss": 1.0701, + "step": 17217 + }, + { + "epoch": 1.27, + "learning_rate": 1.831156468660446e-05, + "loss": 1.0168, + "step": 17218 + }, + { + "epoch": 1.27, + "learning_rate": 1.8311342891975584e-05, + "loss": 1.0307, + "step": 17219 + }, + { + "epoch": 1.27, + "learning_rate": 1.8311121084123483e-05, + "loss": 1.0506, + "step": 17220 + }, + { + "epoch": 1.27, + "learning_rate": 1.8310899263048506e-05, + "loss": 0.9669, + "step": 17221 + }, + { + "epoch": 1.27, + "learning_rate": 1.8310677428751006e-05, + "loss": 1.0409, + "step": 17222 + }, + { + "epoch": 1.27, + "learning_rate": 1.8310455581231336e-05, + "loss": 1.0591, + "step": 17223 + }, + { + "epoch": 1.27, + "learning_rate": 1.831023372048985e-05, + "loss": 0.9299, + "step": 17224 + }, + { + "epoch": 1.27, + "learning_rate": 1.8310011846526905e-05, + "loss": 1.0128, + "step": 17225 + }, + { + "epoch": 1.27, + "learning_rate": 1.830978995934285e-05, + "loss": 0.9797, + "step": 17226 + }, + { + "epoch": 1.27, + "learning_rate": 1.8309568058938034e-05, + "loss": 1.0704, + "step": 17227 + }, + { + "epoch": 1.27, + "learning_rate": 1.8309346145312815e-05, + "loss": 1.0171, + "step": 17228 + }, + { + "epoch": 1.27, + "learning_rate": 1.8309124218467547e-05, + "loss": 1.1206, + "step": 17229 + }, + { + "epoch": 1.27, + "learning_rate": 1.830890227840258e-05, + "loss": 0.9964, + "step": 17230 + }, + { + "epoch": 1.27, + "learning_rate": 1.8308680325118267e-05, + "loss": 1.0196, + "step": 17231 + }, + { + "epoch": 1.27, + "learning_rate": 1.8308458358614967e-05, + "loss": 0.9504, + "step": 17232 + }, + { + "epoch": 1.27, + "learning_rate": 1.8308236378893027e-05, + "loss": 1.0318, + "step": 17233 + }, + { + "epoch": 1.27, + "learning_rate": 1.83080143859528e-05, + "loss": 1.0364, + "step": 17234 + }, + { + "epoch": 1.27, + "learning_rate": 1.830779237979464e-05, + "loss": 1.0365, + "step": 17235 + }, + { + "epoch": 1.27, + "learning_rate": 1.83075703604189e-05, + "loss": 1.1146, + "step": 17236 + }, + { + "epoch": 1.27, + "learning_rate": 1.830734832782594e-05, + "loss": 0.9521, + "step": 17237 + }, + { + "epoch": 1.27, + "learning_rate": 1.8307126282016107e-05, + "loss": 0.9378, + "step": 17238 + }, + { + "epoch": 1.27, + "learning_rate": 1.8306904222989748e-05, + "loss": 1.0675, + "step": 17239 + }, + { + "epoch": 1.27, + "learning_rate": 1.830668215074723e-05, + "loss": 1.0131, + "step": 17240 + }, + { + "epoch": 1.27, + "learning_rate": 1.83064600652889e-05, + "loss": 1.0424, + "step": 17241 + }, + { + "epoch": 1.27, + "learning_rate": 1.8306237966615106e-05, + "loss": 1.0064, + "step": 17242 + }, + { + "epoch": 1.27, + "learning_rate": 1.830601585472621e-05, + "loss": 1.1687, + "step": 17243 + }, + { + "epoch": 1.27, + "learning_rate": 1.8305793729622557e-05, + "loss": 1.0802, + "step": 17244 + }, + { + "epoch": 1.27, + "learning_rate": 1.830557159130451e-05, + "loss": 0.9851, + "step": 17245 + }, + { + "epoch": 1.27, + "learning_rate": 1.8305349439772415e-05, + "loss": 0.9297, + "step": 17246 + }, + { + "epoch": 1.27, + "learning_rate": 1.830512727502663e-05, + "loss": 1.1231, + "step": 17247 + }, + { + "epoch": 1.27, + "learning_rate": 1.8304905097067504e-05, + "loss": 0.9387, + "step": 17248 + }, + { + "epoch": 1.27, + "learning_rate": 1.8304682905895393e-05, + "loss": 1.0779, + "step": 17249 + }, + { + "epoch": 1.27, + "learning_rate": 1.8304460701510652e-05, + "loss": 1.1218, + "step": 17250 + }, + { + "epoch": 1.27, + "learning_rate": 1.8304238483913633e-05, + "loss": 1.0623, + "step": 17251 + }, + { + "epoch": 1.27, + "learning_rate": 1.830401625310469e-05, + "loss": 1.0027, + "step": 17252 + }, + { + "epoch": 1.27, + "learning_rate": 1.830379400908417e-05, + "loss": 1.0359, + "step": 17253 + }, + { + "epoch": 1.27, + "learning_rate": 1.830357175185244e-05, + "loss": 0.9929, + "step": 17254 + }, + { + "epoch": 1.27, + "learning_rate": 1.8303349481409843e-05, + "loss": 1.0133, + "step": 17255 + }, + { + "epoch": 1.28, + "learning_rate": 1.8303127197756736e-05, + "loss": 1.0956, + "step": 17256 + }, + { + "epoch": 1.28, + "learning_rate": 1.8302904900893473e-05, + "loss": 1.1055, + "step": 17257 + }, + { + "epoch": 1.28, + "learning_rate": 1.8302682590820403e-05, + "loss": 0.9652, + "step": 17258 + }, + { + "epoch": 1.28, + "learning_rate": 1.830246026753789e-05, + "loss": 1.0725, + "step": 17259 + }, + { + "epoch": 1.28, + "learning_rate": 1.8302237931046276e-05, + "loss": 1.0004, + "step": 17260 + }, + { + "epoch": 1.28, + "learning_rate": 1.830201558134592e-05, + "loss": 0.9676, + "step": 17261 + }, + { + "epoch": 1.28, + "learning_rate": 1.8301793218437182e-05, + "loss": 0.9147, + "step": 17262 + }, + { + "epoch": 1.28, + "learning_rate": 1.8301570842320407e-05, + "loss": 1.0844, + "step": 17263 + }, + { + "epoch": 1.28, + "learning_rate": 1.830134845299595e-05, + "loss": 1.0045, + "step": 17264 + }, + { + "epoch": 1.28, + "learning_rate": 1.8301126050464168e-05, + "loss": 0.9448, + "step": 17265 + }, + { + "epoch": 1.28, + "learning_rate": 1.8300903634725414e-05, + "loss": 0.9507, + "step": 17266 + }, + { + "epoch": 1.28, + "learning_rate": 1.830068120578004e-05, + "loss": 0.987, + "step": 17267 + }, + { + "epoch": 1.28, + "learning_rate": 1.83004587636284e-05, + "loss": 0.9602, + "step": 17268 + }, + { + "epoch": 1.28, + "learning_rate": 1.8300236308270848e-05, + "loss": 1.0858, + "step": 17269 + }, + { + "epoch": 1.28, + "learning_rate": 1.8300013839707742e-05, + "loss": 1.0543, + "step": 17270 + }, + { + "epoch": 1.28, + "learning_rate": 1.8299791357939433e-05, + "loss": 1.098, + "step": 17271 + }, + { + "epoch": 1.28, + "learning_rate": 1.829956886296627e-05, + "loss": 0.9369, + "step": 17272 + }, + { + "epoch": 1.28, + "learning_rate": 1.8299346354788615e-05, + "loss": 1.0181, + "step": 17273 + }, + { + "epoch": 1.28, + "learning_rate": 1.8299123833406818e-05, + "loss": 1.0467, + "step": 17274 + }, + { + "epoch": 1.28, + "learning_rate": 1.8298901298821235e-05, + "loss": 0.9871, + "step": 17275 + }, + { + "epoch": 1.28, + "learning_rate": 1.8298678751032216e-05, + "loss": 0.9851, + "step": 17276 + }, + { + "epoch": 1.28, + "learning_rate": 1.829845619004012e-05, + "loss": 0.8956, + "step": 17277 + }, + { + "epoch": 1.28, + "learning_rate": 1.8298233615845297e-05, + "loss": 1.0809, + "step": 17278 + }, + { + "epoch": 1.28, + "learning_rate": 1.829801102844811e-05, + "loss": 0.8986, + "step": 17279 + }, + { + "epoch": 1.28, + "learning_rate": 1.8297788427848898e-05, + "loss": 1.0066, + "step": 17280 + }, + { + "epoch": 1.28, + "learning_rate": 1.8297565814048024e-05, + "loss": 1.0223, + "step": 17281 + }, + { + "epoch": 1.28, + "learning_rate": 1.8297343187045843e-05, + "loss": 1.0714, + "step": 17282 + }, + { + "epoch": 1.28, + "learning_rate": 1.829712054684271e-05, + "loss": 1.0472, + "step": 17283 + }, + { + "epoch": 1.28, + "learning_rate": 1.8296897893438975e-05, + "loss": 1.0229, + "step": 17284 + }, + { + "epoch": 1.28, + "learning_rate": 1.8296675226834995e-05, + "loss": 1.0661, + "step": 17285 + }, + { + "epoch": 1.28, + "learning_rate": 1.829645254703112e-05, + "loss": 1.1121, + "step": 17286 + }, + { + "epoch": 1.28, + "learning_rate": 1.829622985402771e-05, + "loss": 1.0861, + "step": 17287 + }, + { + "epoch": 1.28, + "learning_rate": 1.829600714782512e-05, + "loss": 1.0544, + "step": 17288 + }, + { + "epoch": 1.28, + "learning_rate": 1.8295784428423696e-05, + "loss": 1.0971, + "step": 17289 + }, + { + "epoch": 1.28, + "learning_rate": 1.82955616958238e-05, + "loss": 1.0559, + "step": 17290 + }, + { + "epoch": 1.28, + "learning_rate": 1.829533895002579e-05, + "loss": 1.0573, + "step": 17291 + }, + { + "epoch": 1.28, + "learning_rate": 1.8295116191030007e-05, + "loss": 1.0466, + "step": 17292 + }, + { + "epoch": 1.28, + "learning_rate": 1.8294893418836814e-05, + "loss": 0.9921, + "step": 17293 + }, + { + "epoch": 1.28, + "learning_rate": 1.8294670633446567e-05, + "loss": 1.0395, + "step": 17294 + }, + { + "epoch": 1.28, + "learning_rate": 1.829444783485961e-05, + "loss": 1.0641, + "step": 17295 + }, + { + "epoch": 1.28, + "learning_rate": 1.829422502307631e-05, + "loss": 1.0341, + "step": 17296 + }, + { + "epoch": 1.28, + "learning_rate": 1.8294002198097022e-05, + "loss": 1.0528, + "step": 17297 + }, + { + "epoch": 1.28, + "learning_rate": 1.8293779359922094e-05, + "loss": 1.1213, + "step": 17298 + }, + { + "epoch": 1.28, + "learning_rate": 1.8293556508551877e-05, + "loss": 0.9952, + "step": 17299 + }, + { + "epoch": 1.28, + "learning_rate": 1.829333364398673e-05, + "loss": 0.9679, + "step": 17300 + }, + { + "epoch": 1.28, + "learning_rate": 1.829311076622701e-05, + "loss": 1.0062, + "step": 17301 + }, + { + "epoch": 1.28, + "learning_rate": 1.829288787527307e-05, + "loss": 1.1515, + "step": 17302 + }, + { + "epoch": 1.28, + "learning_rate": 1.829266497112526e-05, + "loss": 1.0311, + "step": 17303 + }, + { + "epoch": 1.28, + "learning_rate": 1.8292442053783942e-05, + "loss": 0.9334, + "step": 17304 + }, + { + "epoch": 1.28, + "learning_rate": 1.829221912324947e-05, + "loss": 1.0771, + "step": 17305 + }, + { + "epoch": 1.28, + "learning_rate": 1.8291996179522192e-05, + "loss": 0.9324, + "step": 17306 + }, + { + "epoch": 1.28, + "learning_rate": 1.8291773222602467e-05, + "loss": 0.9867, + "step": 17307 + }, + { + "epoch": 1.28, + "learning_rate": 1.8291550252490648e-05, + "loss": 0.9557, + "step": 17308 + }, + { + "epoch": 1.28, + "learning_rate": 1.8291327269187094e-05, + "loss": 1.0055, + "step": 17309 + }, + { + "epoch": 1.28, + "learning_rate": 1.8291104272692157e-05, + "loss": 1.1197, + "step": 17310 + }, + { + "epoch": 1.28, + "learning_rate": 1.829088126300619e-05, + "loss": 1.0488, + "step": 17311 + }, + { + "epoch": 1.28, + "learning_rate": 1.8290658240129554e-05, + "loss": 0.9305, + "step": 17312 + }, + { + "epoch": 1.28, + "learning_rate": 1.8290435204062593e-05, + "loss": 1.0838, + "step": 17313 + }, + { + "epoch": 1.28, + "learning_rate": 1.8290212154805672e-05, + "loss": 1.0479, + "step": 17314 + }, + { + "epoch": 1.28, + "learning_rate": 1.828998909235914e-05, + "loss": 1.0446, + "step": 17315 + }, + { + "epoch": 1.28, + "learning_rate": 1.8289766016723357e-05, + "loss": 1.1413, + "step": 17316 + }, + { + "epoch": 1.28, + "learning_rate": 1.828954292789867e-05, + "loss": 1.0258, + "step": 17317 + }, + { + "epoch": 1.28, + "learning_rate": 1.8289319825885445e-05, + "loss": 1.1366, + "step": 17318 + }, + { + "epoch": 1.28, + "learning_rate": 1.8289096710684027e-05, + "loss": 1.0799, + "step": 17319 + }, + { + "epoch": 1.28, + "learning_rate": 1.8288873582294775e-05, + "loss": 0.9633, + "step": 17320 + }, + { + "epoch": 1.28, + "learning_rate": 1.8288650440718046e-05, + "loss": 1.0676, + "step": 17321 + }, + { + "epoch": 1.28, + "learning_rate": 1.828842728595419e-05, + "loss": 1.0346, + "step": 17322 + }, + { + "epoch": 1.28, + "learning_rate": 1.8288204118003565e-05, + "loss": 0.9763, + "step": 17323 + }, + { + "epoch": 1.28, + "learning_rate": 1.8287980936866524e-05, + "loss": 1.0202, + "step": 17324 + }, + { + "epoch": 1.28, + "learning_rate": 1.8287757742543427e-05, + "loss": 0.9463, + "step": 17325 + }, + { + "epoch": 1.28, + "learning_rate": 1.8287534535034626e-05, + "loss": 1.0441, + "step": 17326 + }, + { + "epoch": 1.28, + "learning_rate": 1.8287311314340476e-05, + "loss": 1.0789, + "step": 17327 + }, + { + "epoch": 1.28, + "learning_rate": 1.828708808046133e-05, + "loss": 1.0003, + "step": 17328 + }, + { + "epoch": 1.28, + "learning_rate": 1.8286864833397547e-05, + "loss": 0.9627, + "step": 17329 + }, + { + "epoch": 1.28, + "learning_rate": 1.8286641573149483e-05, + "loss": 1.0763, + "step": 17330 + }, + { + "epoch": 1.28, + "learning_rate": 1.8286418299717486e-05, + "loss": 1.0761, + "step": 17331 + }, + { + "epoch": 1.28, + "learning_rate": 1.828619501310192e-05, + "loss": 1.1635, + "step": 17332 + }, + { + "epoch": 1.28, + "learning_rate": 1.8285971713303133e-05, + "loss": 1.1108, + "step": 17333 + }, + { + "epoch": 1.28, + "learning_rate": 1.8285748400321487e-05, + "loss": 0.9669, + "step": 17334 + }, + { + "epoch": 1.28, + "learning_rate": 1.8285525074157333e-05, + "loss": 0.9978, + "step": 17335 + }, + { + "epoch": 1.28, + "learning_rate": 1.8285301734811025e-05, + "loss": 0.974, + "step": 17336 + }, + { + "epoch": 1.28, + "learning_rate": 1.8285078382282925e-05, + "loss": 0.941, + "step": 17337 + }, + { + "epoch": 1.28, + "learning_rate": 1.8284855016573377e-05, + "loss": 1.0455, + "step": 17338 + }, + { + "epoch": 1.28, + "learning_rate": 1.828463163768275e-05, + "loss": 1.0162, + "step": 17339 + }, + { + "epoch": 1.28, + "learning_rate": 1.828440824561139e-05, + "loss": 1.0707, + "step": 17340 + }, + { + "epoch": 1.28, + "learning_rate": 1.8284184840359652e-05, + "loss": 1.0805, + "step": 17341 + }, + { + "epoch": 1.28, + "learning_rate": 1.8283961421927897e-05, + "loss": 1.0971, + "step": 17342 + }, + { + "epoch": 1.28, + "learning_rate": 1.828373799031648e-05, + "loss": 0.9946, + "step": 17343 + }, + { + "epoch": 1.28, + "learning_rate": 1.8283514545525753e-05, + "loss": 1.0094, + "step": 17344 + }, + { + "epoch": 1.28, + "learning_rate": 1.828329108755607e-05, + "loss": 1.1091, + "step": 17345 + }, + { + "epoch": 1.28, + "learning_rate": 1.828306761640779e-05, + "loss": 1.0801, + "step": 17346 + }, + { + "epoch": 1.28, + "learning_rate": 1.828284413208127e-05, + "loss": 0.9204, + "step": 17347 + }, + { + "epoch": 1.28, + "learning_rate": 1.8282620634576865e-05, + "loss": 1.061, + "step": 17348 + }, + { + "epoch": 1.28, + "learning_rate": 1.8282397123894925e-05, + "loss": 1.0447, + "step": 17349 + }, + { + "epoch": 1.28, + "learning_rate": 1.8282173600035814e-05, + "loss": 1.0172, + "step": 17350 + }, + { + "epoch": 1.28, + "learning_rate": 1.828195006299988e-05, + "loss": 0.9634, + "step": 17351 + }, + { + "epoch": 1.28, + "learning_rate": 1.8281726512787484e-05, + "loss": 1.1066, + "step": 17352 + }, + { + "epoch": 1.28, + "learning_rate": 1.828150294939898e-05, + "loss": 1.0973, + "step": 17353 + }, + { + "epoch": 1.28, + "learning_rate": 1.8281279372834718e-05, + "loss": 1.0894, + "step": 17354 + }, + { + "epoch": 1.28, + "learning_rate": 1.8281055783095065e-05, + "loss": 1.0598, + "step": 17355 + }, + { + "epoch": 1.28, + "learning_rate": 1.8280832180180368e-05, + "loss": 0.9604, + "step": 17356 + }, + { + "epoch": 1.28, + "learning_rate": 1.8280608564090988e-05, + "loss": 0.9895, + "step": 17357 + }, + { + "epoch": 1.28, + "learning_rate": 1.8280384934827277e-05, + "loss": 1.0514, + "step": 17358 + }, + { + "epoch": 1.28, + "learning_rate": 1.8280161292389588e-05, + "loss": 0.9554, + "step": 17359 + }, + { + "epoch": 1.28, + "learning_rate": 1.8279937636778284e-05, + "loss": 1.023, + "step": 17360 + }, + { + "epoch": 1.28, + "learning_rate": 1.827971396799372e-05, + "loss": 1.0897, + "step": 17361 + }, + { + "epoch": 1.28, + "learning_rate": 1.827949028603625e-05, + "loss": 1.0722, + "step": 17362 + }, + { + "epoch": 1.28, + "learning_rate": 1.8279266590906223e-05, + "loss": 1.154, + "step": 17363 + }, + { + "epoch": 1.28, + "learning_rate": 1.8279042882604005e-05, + "loss": 1.032, + "step": 17364 + }, + { + "epoch": 1.28, + "learning_rate": 1.827881916112995e-05, + "loss": 1.0524, + "step": 17365 + }, + { + "epoch": 1.28, + "learning_rate": 1.827859542648441e-05, + "loss": 1.1088, + "step": 17366 + }, + { + "epoch": 1.28, + "learning_rate": 1.8278371678667742e-05, + "loss": 0.948, + "step": 17367 + }, + { + "epoch": 1.28, + "learning_rate": 1.8278147917680305e-05, + "loss": 1.1043, + "step": 17368 + }, + { + "epoch": 1.28, + "learning_rate": 1.8277924143522454e-05, + "loss": 0.9474, + "step": 17369 + }, + { + "epoch": 1.28, + "learning_rate": 1.8277700356194542e-05, + "loss": 1.0548, + "step": 17370 + }, + { + "epoch": 1.28, + "learning_rate": 1.827747655569693e-05, + "loss": 1.0656, + "step": 17371 + }, + { + "epoch": 1.28, + "learning_rate": 1.827725274202997e-05, + "loss": 0.9824, + "step": 17372 + }, + { + "epoch": 1.28, + "learning_rate": 1.827702891519402e-05, + "loss": 1.0623, + "step": 17373 + }, + { + "epoch": 1.28, + "learning_rate": 1.8276805075189432e-05, + "loss": 1.0925, + "step": 17374 + }, + { + "epoch": 1.28, + "learning_rate": 1.8276581222016567e-05, + "loss": 1.0148, + "step": 17375 + }, + { + "epoch": 1.28, + "learning_rate": 1.827635735567578e-05, + "loss": 0.9781, + "step": 17376 + }, + { + "epoch": 1.28, + "learning_rate": 1.827613347616743e-05, + "loss": 0.9708, + "step": 17377 + }, + { + "epoch": 1.28, + "learning_rate": 1.8275909583491866e-05, + "loss": 1.0413, + "step": 17378 + }, + { + "epoch": 1.28, + "learning_rate": 1.827568567764945e-05, + "loss": 0.9656, + "step": 17379 + }, + { + "epoch": 1.28, + "learning_rate": 1.8275461758640537e-05, + "loss": 1.0583, + "step": 17380 + }, + { + "epoch": 1.28, + "learning_rate": 1.8275237826465478e-05, + "loss": 1.0959, + "step": 17381 + }, + { + "epoch": 1.28, + "learning_rate": 1.827501388112464e-05, + "loss": 1.0742, + "step": 17382 + }, + { + "epoch": 1.28, + "learning_rate": 1.827478992261837e-05, + "loss": 1.0714, + "step": 17383 + }, + { + "epoch": 1.28, + "learning_rate": 1.827456595094703e-05, + "loss": 1.1477, + "step": 17384 + }, + { + "epoch": 1.28, + "learning_rate": 1.8274341966110975e-05, + "loss": 1.0657, + "step": 17385 + }, + { + "epoch": 1.28, + "learning_rate": 1.8274117968110556e-05, + "loss": 1.0137, + "step": 17386 + }, + { + "epoch": 1.28, + "learning_rate": 1.8273893956946137e-05, + "loss": 1.0294, + "step": 17387 + }, + { + "epoch": 1.28, + "learning_rate": 1.8273669932618067e-05, + "loss": 1.0121, + "step": 17388 + }, + { + "epoch": 1.28, + "learning_rate": 1.827344589512671e-05, + "loss": 0.9787, + "step": 17389 + }, + { + "epoch": 1.28, + "learning_rate": 1.8273221844472418e-05, + "loss": 1.1063, + "step": 17390 + }, + { + "epoch": 1.29, + "learning_rate": 1.827299778065555e-05, + "loss": 1.0425, + "step": 17391 + }, + { + "epoch": 1.29, + "learning_rate": 1.827277370367646e-05, + "loss": 1.0002, + "step": 17392 + }, + { + "epoch": 1.29, + "learning_rate": 1.8272549613535504e-05, + "loss": 1.107, + "step": 17393 + }, + { + "epoch": 1.29, + "learning_rate": 1.827232551023304e-05, + "loss": 0.989, + "step": 17394 + }, + { + "epoch": 1.29, + "learning_rate": 1.8272101393769425e-05, + "loss": 0.9897, + "step": 17395 + }, + { + "epoch": 1.29, + "learning_rate": 1.827187726414502e-05, + "loss": 0.9682, + "step": 17396 + }, + { + "epoch": 1.29, + "learning_rate": 1.827165312136017e-05, + "loss": 0.9068, + "step": 17397 + }, + { + "epoch": 1.29, + "learning_rate": 1.827142896541524e-05, + "loss": 0.9858, + "step": 17398 + }, + { + "epoch": 1.29, + "learning_rate": 1.8271204796310585e-05, + "loss": 0.9196, + "step": 17399 + }, + { + "epoch": 1.29, + "learning_rate": 1.827098061404656e-05, + "loss": 0.9558, + "step": 17400 + }, + { + "epoch": 1.29, + "learning_rate": 1.8270756418623528e-05, + "loss": 1.0653, + "step": 17401 + }, + { + "epoch": 1.29, + "learning_rate": 1.8270532210041838e-05, + "loss": 1.0118, + "step": 17402 + }, + { + "epoch": 1.29, + "learning_rate": 1.8270307988301847e-05, + "loss": 1.0895, + "step": 17403 + }, + { + "epoch": 1.29, + "learning_rate": 1.8270083753403917e-05, + "loss": 1.0072, + "step": 17404 + }, + { + "epoch": 1.29, + "learning_rate": 1.8269859505348405e-05, + "loss": 0.9895, + "step": 17405 + }, + { + "epoch": 1.29, + "learning_rate": 1.826963524413566e-05, + "loss": 1.0431, + "step": 17406 + }, + { + "epoch": 1.29, + "learning_rate": 1.8269410969766045e-05, + "loss": 0.967, + "step": 17407 + }, + { + "epoch": 1.29, + "learning_rate": 1.8269186682239915e-05, + "loss": 1.013, + "step": 17408 + }, + { + "epoch": 1.29, + "learning_rate": 1.826896238155763e-05, + "loss": 0.9323, + "step": 17409 + }, + { + "epoch": 1.29, + "learning_rate": 1.826873806771954e-05, + "loss": 1.0104, + "step": 17410 + }, + { + "epoch": 1.29, + "learning_rate": 1.8268513740726008e-05, + "loss": 0.9587, + "step": 17411 + }, + { + "epoch": 1.29, + "learning_rate": 1.8268289400577387e-05, + "loss": 0.9801, + "step": 17412 + }, + { + "epoch": 1.29, + "learning_rate": 1.826806504727404e-05, + "loss": 1.0255, + "step": 17413 + }, + { + "epoch": 1.29, + "learning_rate": 1.826784068081632e-05, + "loss": 0.8889, + "step": 17414 + }, + { + "epoch": 1.29, + "learning_rate": 1.826761630120458e-05, + "loss": 1.0255, + "step": 17415 + }, + { + "epoch": 1.29, + "learning_rate": 1.8267391908439183e-05, + "loss": 0.9983, + "step": 17416 + }, + { + "epoch": 1.29, + "learning_rate": 1.8267167502520483e-05, + "loss": 1.0191, + "step": 17417 + }, + { + "epoch": 1.29, + "learning_rate": 1.826694308344884e-05, + "loss": 1.1302, + "step": 17418 + }, + { + "epoch": 1.29, + "learning_rate": 1.8266718651224603e-05, + "loss": 1.0246, + "step": 17419 + }, + { + "epoch": 1.29, + "learning_rate": 1.826649420584814e-05, + "loss": 1.0899, + "step": 17420 + }, + { + "epoch": 1.29, + "learning_rate": 1.8266269747319802e-05, + "loss": 1.033, + "step": 17421 + }, + { + "epoch": 1.29, + "learning_rate": 1.8266045275639945e-05, + "loss": 0.9973, + "step": 17422 + }, + { + "epoch": 1.29, + "learning_rate": 1.826582079080893e-05, + "loss": 1.0563, + "step": 17423 + }, + { + "epoch": 1.29, + "learning_rate": 1.8265596292827116e-05, + "loss": 1.0377, + "step": 17424 + }, + { + "epoch": 1.29, + "learning_rate": 1.8265371781694852e-05, + "loss": 1.0314, + "step": 17425 + }, + { + "epoch": 1.29, + "learning_rate": 1.82651472574125e-05, + "loss": 1.0087, + "step": 17426 + }, + { + "epoch": 1.29, + "learning_rate": 1.8264922719980416e-05, + "loss": 0.9872, + "step": 17427 + }, + { + "epoch": 1.29, + "learning_rate": 1.8264698169398962e-05, + "loss": 0.996, + "step": 17428 + }, + { + "epoch": 1.29, + "learning_rate": 1.826447360566849e-05, + "loss": 1.0406, + "step": 17429 + }, + { + "epoch": 1.29, + "learning_rate": 1.8264249028789357e-05, + "loss": 1.0576, + "step": 17430 + }, + { + "epoch": 1.29, + "learning_rate": 1.8264024438761924e-05, + "loss": 0.9516, + "step": 17431 + }, + { + "epoch": 1.29, + "learning_rate": 1.8263799835586544e-05, + "loss": 0.9001, + "step": 17432 + }, + { + "epoch": 1.29, + "learning_rate": 1.8263575219263576e-05, + "loss": 1.0016, + "step": 17433 + }, + { + "epoch": 1.29, + "learning_rate": 1.8263350589793383e-05, + "loss": 1.1295, + "step": 17434 + }, + { + "epoch": 1.29, + "learning_rate": 1.8263125947176314e-05, + "loss": 1.115, + "step": 17435 + }, + { + "epoch": 1.29, + "learning_rate": 1.826290129141273e-05, + "loss": 1.0174, + "step": 17436 + }, + { + "epoch": 1.29, + "learning_rate": 1.826267662250299e-05, + "loss": 1.0383, + "step": 17437 + }, + { + "epoch": 1.29, + "learning_rate": 1.8262451940447447e-05, + "loss": 1.1803, + "step": 17438 + }, + { + "epoch": 1.29, + "learning_rate": 1.8262227245246464e-05, + "loss": 1.1152, + "step": 17439 + }, + { + "epoch": 1.29, + "learning_rate": 1.8262002536900393e-05, + "loss": 1.1069, + "step": 17440 + }, + { + "epoch": 1.29, + "learning_rate": 1.8261777815409595e-05, + "loss": 0.9564, + "step": 17441 + }, + { + "epoch": 1.29, + "learning_rate": 1.8261553080774427e-05, + "loss": 0.9109, + "step": 17442 + }, + { + "epoch": 1.29, + "learning_rate": 1.8261328332995246e-05, + "loss": 1.0263, + "step": 17443 + }, + { + "epoch": 1.29, + "learning_rate": 1.826110357207241e-05, + "loss": 1.0651, + "step": 17444 + }, + { + "epoch": 1.29, + "learning_rate": 1.8260878798006277e-05, + "loss": 1.0654, + "step": 17445 + }, + { + "epoch": 1.29, + "learning_rate": 1.8260654010797204e-05, + "loss": 1.1084, + "step": 17446 + }, + { + "epoch": 1.29, + "learning_rate": 1.8260429210445547e-05, + "loss": 1.0811, + "step": 17447 + }, + { + "epoch": 1.29, + "learning_rate": 1.826020439695167e-05, + "loss": 0.9766, + "step": 17448 + }, + { + "epoch": 1.29, + "learning_rate": 1.825997957031592e-05, + "loss": 1.1433, + "step": 17449 + }, + { + "epoch": 1.29, + "learning_rate": 1.8259754730538665e-05, + "loss": 1.0361, + "step": 17450 + }, + { + "epoch": 1.29, + "learning_rate": 1.8259529877620255e-05, + "loss": 1.0485, + "step": 17451 + }, + { + "epoch": 1.29, + "learning_rate": 1.825930501156105e-05, + "loss": 0.9952, + "step": 17452 + }, + { + "epoch": 1.29, + "learning_rate": 1.8259080132361415e-05, + "loss": 1.0339, + "step": 17453 + }, + { + "epoch": 1.29, + "learning_rate": 1.82588552400217e-05, + "loss": 1.0224, + "step": 17454 + }, + { + "epoch": 1.29, + "learning_rate": 1.8258630334542263e-05, + "loss": 1.0418, + "step": 17455 + }, + { + "epoch": 1.29, + "learning_rate": 1.8258405415923462e-05, + "loss": 1.0498, + "step": 17456 + }, + { + "epoch": 1.29, + "learning_rate": 1.825818048416566e-05, + "loss": 0.9709, + "step": 17457 + }, + { + "epoch": 1.29, + "learning_rate": 1.825795553926921e-05, + "loss": 0.9647, + "step": 17458 + }, + { + "epoch": 1.29, + "learning_rate": 1.8257730581234464e-05, + "loss": 1.0512, + "step": 17459 + }, + { + "epoch": 1.29, + "learning_rate": 1.8257505610061795e-05, + "loss": 1.0177, + "step": 17460 + }, + { + "epoch": 1.29, + "learning_rate": 1.8257280625751547e-05, + "loss": 1.1299, + "step": 17461 + }, + { + "epoch": 1.29, + "learning_rate": 1.825705562830409e-05, + "loss": 0.9626, + "step": 17462 + }, + { + "epoch": 1.29, + "learning_rate": 1.825683061771977e-05, + "loss": 1.0017, + "step": 17463 + }, + { + "epoch": 1.29, + "learning_rate": 1.8256605593998952e-05, + "loss": 0.8788, + "step": 17464 + }, + { + "epoch": 1.29, + "learning_rate": 1.825638055714199e-05, + "loss": 1.1575, + "step": 17465 + }, + { + "epoch": 1.29, + "learning_rate": 1.8256155507149254e-05, + "loss": 1.093, + "step": 17466 + }, + { + "epoch": 1.29, + "learning_rate": 1.8255930444021083e-05, + "loss": 1.0388, + "step": 17467 + }, + { + "epoch": 1.29, + "learning_rate": 1.825570536775785e-05, + "loss": 0.9841, + "step": 17468 + }, + { + "epoch": 1.29, + "learning_rate": 1.8255480278359906e-05, + "loss": 1.0656, + "step": 17469 + }, + { + "epoch": 1.29, + "learning_rate": 1.8255255175827612e-05, + "loss": 0.9623, + "step": 17470 + }, + { + "epoch": 1.29, + "learning_rate": 1.8255030060161324e-05, + "loss": 0.9487, + "step": 17471 + }, + { + "epoch": 1.29, + "learning_rate": 1.8254804931361404e-05, + "loss": 1.1019, + "step": 17472 + }, + { + "epoch": 1.29, + "learning_rate": 1.8254579789428204e-05, + "loss": 0.9415, + "step": 17473 + }, + { + "epoch": 1.29, + "learning_rate": 1.8254354634362087e-05, + "loss": 0.9742, + "step": 17474 + }, + { + "epoch": 1.29, + "learning_rate": 1.825412946616341e-05, + "loss": 1.1286, + "step": 17475 + }, + { + "epoch": 1.29, + "learning_rate": 1.825390428483253e-05, + "loss": 1.0128, + "step": 17476 + }, + { + "epoch": 1.29, + "learning_rate": 1.8253679090369807e-05, + "loss": 0.9874, + "step": 17477 + }, + { + "epoch": 1.29, + "learning_rate": 1.8253453882775602e-05, + "loss": 1.0684, + "step": 17478 + }, + { + "epoch": 1.29, + "learning_rate": 1.8253228662050265e-05, + "loss": 0.9271, + "step": 17479 + }, + { + "epoch": 1.29, + "learning_rate": 1.825300342819416e-05, + "loss": 1.0009, + "step": 17480 + }, + { + "epoch": 1.29, + "learning_rate": 1.8252778181207648e-05, + "loss": 0.9879, + "step": 17481 + }, + { + "epoch": 1.29, + "learning_rate": 1.825255292109108e-05, + "loss": 1.0173, + "step": 17482 + }, + { + "epoch": 1.29, + "learning_rate": 1.825232764784482e-05, + "loss": 1.1217, + "step": 17483 + }, + { + "epoch": 1.29, + "learning_rate": 1.8252102361469223e-05, + "loss": 0.9216, + "step": 17484 + }, + { + "epoch": 1.29, + "learning_rate": 1.8251877061964652e-05, + "loss": 1.0395, + "step": 17485 + }, + { + "epoch": 1.29, + "learning_rate": 1.8251651749331464e-05, + "loss": 0.9723, + "step": 17486 + }, + { + "epoch": 1.29, + "learning_rate": 1.825142642357001e-05, + "loss": 1.0525, + "step": 17487 + }, + { + "epoch": 1.29, + "learning_rate": 1.825120108468066e-05, + "loss": 1.0605, + "step": 17488 + }, + { + "epoch": 1.29, + "learning_rate": 1.8250975732663767e-05, + "loss": 1.0318, + "step": 17489 + }, + { + "epoch": 1.29, + "learning_rate": 1.8250750367519686e-05, + "loss": 1.0345, + "step": 17490 + }, + { + "epoch": 1.29, + "learning_rate": 1.8250524989248783e-05, + "loss": 1.116, + "step": 17491 + }, + { + "epoch": 1.29, + "learning_rate": 1.825029959785141e-05, + "loss": 0.9817, + "step": 17492 + }, + { + "epoch": 1.29, + "learning_rate": 1.8250074193327934e-05, + "loss": 1.0889, + "step": 17493 + }, + { + "epoch": 1.29, + "learning_rate": 1.8249848775678703e-05, + "loss": 1.1277, + "step": 17494 + }, + { + "epoch": 1.29, + "learning_rate": 1.824962334490408e-05, + "loss": 1.0243, + "step": 17495 + }, + { + "epoch": 1.29, + "learning_rate": 1.8249397901004427e-05, + "loss": 0.9485, + "step": 17496 + }, + { + "epoch": 1.29, + "learning_rate": 1.8249172443980097e-05, + "loss": 1.1034, + "step": 17497 + }, + { + "epoch": 1.29, + "learning_rate": 1.8248946973831455e-05, + "loss": 1.0316, + "step": 17498 + }, + { + "epoch": 1.29, + "learning_rate": 1.824872149055885e-05, + "loss": 1.0557, + "step": 17499 + }, + { + "epoch": 1.29, + "learning_rate": 1.8248495994162654e-05, + "loss": 1.0302, + "step": 17500 + }, + { + "epoch": 1.29, + "learning_rate": 1.8248270484643217e-05, + "loss": 0.978, + "step": 17501 + }, + { + "epoch": 1.29, + "learning_rate": 1.8248044962000898e-05, + "loss": 1.0958, + "step": 17502 + }, + { + "epoch": 1.29, + "learning_rate": 1.8247819426236063e-05, + "loss": 1.0715, + "step": 17503 + }, + { + "epoch": 1.29, + "learning_rate": 1.8247593877349058e-05, + "loss": 1.1801, + "step": 17504 + }, + { + "epoch": 1.29, + "learning_rate": 1.8247368315340256e-05, + "loss": 1.095, + "step": 17505 + }, + { + "epoch": 1.29, + "learning_rate": 1.8247142740210002e-05, + "loss": 0.9599, + "step": 17506 + }, + { + "epoch": 1.29, + "learning_rate": 1.8246917151958666e-05, + "loss": 1.047, + "step": 17507 + }, + { + "epoch": 1.29, + "learning_rate": 1.8246691550586603e-05, + "loss": 0.9149, + "step": 17508 + }, + { + "epoch": 1.29, + "learning_rate": 1.824646593609417e-05, + "loss": 0.9578, + "step": 17509 + }, + { + "epoch": 1.29, + "learning_rate": 1.824624030848173e-05, + "loss": 0.911, + "step": 17510 + }, + { + "epoch": 1.29, + "learning_rate": 1.8246014667749638e-05, + "loss": 1.0565, + "step": 17511 + }, + { + "epoch": 1.29, + "learning_rate": 1.8245789013898253e-05, + "loss": 1.0293, + "step": 17512 + }, + { + "epoch": 1.29, + "learning_rate": 1.8245563346927937e-05, + "loss": 1.0642, + "step": 17513 + }, + { + "epoch": 1.29, + "learning_rate": 1.824533766683905e-05, + "loss": 1.0025, + "step": 17514 + }, + { + "epoch": 1.29, + "learning_rate": 1.8245111973631944e-05, + "loss": 0.9283, + "step": 17515 + }, + { + "epoch": 1.29, + "learning_rate": 1.8244886267306987e-05, + "loss": 1.0888, + "step": 17516 + }, + { + "epoch": 1.29, + "learning_rate": 1.824466054786453e-05, + "loss": 1.1747, + "step": 17517 + }, + { + "epoch": 1.29, + "learning_rate": 1.824443481530494e-05, + "loss": 1.0234, + "step": 17518 + }, + { + "epoch": 1.29, + "learning_rate": 1.8244209069628573e-05, + "loss": 0.9269, + "step": 17519 + }, + { + "epoch": 1.29, + "learning_rate": 1.8243983310835786e-05, + "loss": 1.0702, + "step": 17520 + }, + { + "epoch": 1.29, + "learning_rate": 1.8243757538926937e-05, + "loss": 1.043, + "step": 17521 + }, + { + "epoch": 1.29, + "learning_rate": 1.8243531753902387e-05, + "loss": 0.9442, + "step": 17522 + }, + { + "epoch": 1.29, + "learning_rate": 1.82433059557625e-05, + "loss": 0.9769, + "step": 17523 + }, + { + "epoch": 1.29, + "learning_rate": 1.824308014450763e-05, + "loss": 1.0286, + "step": 17524 + }, + { + "epoch": 1.29, + "learning_rate": 1.8242854320138134e-05, + "loss": 1.0423, + "step": 17525 + }, + { + "epoch": 1.29, + "learning_rate": 1.824262848265438e-05, + "loss": 1.0048, + "step": 17526 + }, + { + "epoch": 1.3, + "learning_rate": 1.824240263205672e-05, + "loss": 1.047, + "step": 17527 + }, + { + "epoch": 1.3, + "learning_rate": 1.8242176768345515e-05, + "loss": 1.0017, + "step": 17528 + }, + { + "epoch": 1.3, + "learning_rate": 1.8241950891521126e-05, + "loss": 1.1166, + "step": 17529 + }, + { + "epoch": 1.3, + "learning_rate": 1.8241725001583908e-05, + "loss": 1.017, + "step": 17530 + }, + { + "epoch": 1.3, + "learning_rate": 1.8241499098534223e-05, + "loss": 0.9962, + "step": 17531 + }, + { + "epoch": 1.3, + "learning_rate": 1.8241273182372435e-05, + "loss": 1.0421, + "step": 17532 + }, + { + "epoch": 1.3, + "learning_rate": 1.82410472530989e-05, + "loss": 1.0443, + "step": 17533 + }, + { + "epoch": 1.3, + "learning_rate": 1.824082131071397e-05, + "loss": 0.9027, + "step": 17534 + }, + { + "epoch": 1.3, + "learning_rate": 1.8240595355218014e-05, + "loss": 0.9001, + "step": 17535 + }, + { + "epoch": 1.3, + "learning_rate": 1.824036938661139e-05, + "loss": 0.966, + "step": 17536 + }, + { + "epoch": 1.3, + "learning_rate": 1.8240143404894457e-05, + "loss": 1.0033, + "step": 17537 + }, + { + "epoch": 1.3, + "learning_rate": 1.8239917410067574e-05, + "loss": 0.9713, + "step": 17538 + }, + { + "epoch": 1.3, + "learning_rate": 1.8239691402131097e-05, + "loss": 1.1199, + "step": 17539 + }, + { + "epoch": 1.3, + "learning_rate": 1.823946538108539e-05, + "loss": 1.0725, + "step": 17540 + }, + { + "epoch": 1.3, + "learning_rate": 1.8239239346930814e-05, + "loss": 0.8964, + "step": 17541 + }, + { + "epoch": 1.3, + "learning_rate": 1.8239013299667726e-05, + "loss": 1.0094, + "step": 17542 + }, + { + "epoch": 1.3, + "learning_rate": 1.8238787239296484e-05, + "loss": 0.9931, + "step": 17543 + }, + { + "epoch": 1.3, + "learning_rate": 1.823856116581745e-05, + "loss": 0.9815, + "step": 17544 + }, + { + "epoch": 1.3, + "learning_rate": 1.823833507923098e-05, + "loss": 0.9504, + "step": 17545 + }, + { + "epoch": 1.3, + "learning_rate": 1.8238108979537438e-05, + "loss": 1.0816, + "step": 17546 + }, + { + "epoch": 1.3, + "learning_rate": 1.8237882866737188e-05, + "loss": 0.9672, + "step": 17547 + }, + { + "epoch": 1.3, + "learning_rate": 1.8237656740830577e-05, + "loss": 0.8832, + "step": 17548 + }, + { + "epoch": 1.3, + "learning_rate": 1.8237430601817978e-05, + "loss": 1.0359, + "step": 17549 + }, + { + "epoch": 1.3, + "learning_rate": 1.8237204449699742e-05, + "loss": 1.0645, + "step": 17550 + }, + { + "epoch": 1.3, + "learning_rate": 1.823697828447623e-05, + "loss": 1.0058, + "step": 17551 + }, + { + "epoch": 1.3, + "learning_rate": 1.8236752106147807e-05, + "loss": 0.9865, + "step": 17552 + }, + { + "epoch": 1.3, + "learning_rate": 1.8236525914714826e-05, + "loss": 1.1103, + "step": 17553 + }, + { + "epoch": 1.3, + "learning_rate": 1.823629971017765e-05, + "loss": 0.9235, + "step": 17554 + }, + { + "epoch": 1.3, + "learning_rate": 1.8236073492536637e-05, + "loss": 1.0774, + "step": 17555 + }, + { + "epoch": 1.3, + "learning_rate": 1.8235847261792156e-05, + "loss": 1.0822, + "step": 17556 + }, + { + "epoch": 1.3, + "learning_rate": 1.8235621017944555e-05, + "loss": 1.0189, + "step": 17557 + }, + { + "epoch": 1.3, + "learning_rate": 1.8235394760994202e-05, + "loss": 1.1044, + "step": 17558 + }, + { + "epoch": 1.3, + "learning_rate": 1.8235168490941452e-05, + "loss": 1.022, + "step": 17559 + }, + { + "epoch": 1.3, + "learning_rate": 1.8234942207786664e-05, + "loss": 1.0266, + "step": 17560 + }, + { + "epoch": 1.3, + "learning_rate": 1.8234715911530204e-05, + "loss": 0.9983, + "step": 17561 + }, + { + "epoch": 1.3, + "learning_rate": 1.823448960217243e-05, + "loss": 1.012, + "step": 17562 + }, + { + "epoch": 1.3, + "learning_rate": 1.8234263279713696e-05, + "loss": 1.0985, + "step": 17563 + }, + { + "epoch": 1.3, + "learning_rate": 1.823403694415437e-05, + "loss": 1.0198, + "step": 17564 + }, + { + "epoch": 1.3, + "learning_rate": 1.8233810595494806e-05, + "loss": 0.8937, + "step": 17565 + }, + { + "epoch": 1.3, + "learning_rate": 1.823358423373537e-05, + "loss": 0.9782, + "step": 17566 + }, + { + "epoch": 1.3, + "learning_rate": 1.823335785887642e-05, + "loss": 1.1348, + "step": 17567 + }, + { + "epoch": 1.3, + "learning_rate": 1.8233131470918316e-05, + "loss": 0.9842, + "step": 17568 + }, + { + "epoch": 1.3, + "learning_rate": 1.8232905069861414e-05, + "loss": 1.0932, + "step": 17569 + }, + { + "epoch": 1.3, + "learning_rate": 1.8232678655706083e-05, + "loss": 1.0578, + "step": 17570 + }, + { + "epoch": 1.3, + "learning_rate": 1.8232452228452674e-05, + "loss": 0.9235, + "step": 17571 + }, + { + "epoch": 1.3, + "learning_rate": 1.823222578810155e-05, + "loss": 0.9716, + "step": 17572 + }, + { + "epoch": 1.3, + "learning_rate": 1.8231999334653076e-05, + "loss": 1.0545, + "step": 17573 + }, + { + "epoch": 1.3, + "learning_rate": 1.823177286810761e-05, + "loss": 1.0847, + "step": 17574 + }, + { + "epoch": 1.3, + "learning_rate": 1.8231546388465507e-05, + "loss": 1.133, + "step": 17575 + }, + { + "epoch": 1.3, + "learning_rate": 1.8231319895727135e-05, + "loss": 0.9724, + "step": 17576 + }, + { + "epoch": 1.3, + "learning_rate": 1.823109338989285e-05, + "loss": 1.0097, + "step": 17577 + }, + { + "epoch": 1.3, + "learning_rate": 1.8230866870963013e-05, + "loss": 1.0181, + "step": 17578 + }, + { + "epoch": 1.3, + "learning_rate": 1.823064033893798e-05, + "loss": 1.0324, + "step": 17579 + }, + { + "epoch": 1.3, + "learning_rate": 1.8230413793818122e-05, + "loss": 1.0483, + "step": 17580 + }, + { + "epoch": 1.3, + "learning_rate": 1.823018723560379e-05, + "loss": 1.05, + "step": 17581 + }, + { + "epoch": 1.3, + "learning_rate": 1.822996066429535e-05, + "loss": 1.1334, + "step": 17582 + }, + { + "epoch": 1.3, + "learning_rate": 1.822973407989316e-05, + "loss": 0.9048, + "step": 17583 + }, + { + "epoch": 1.3, + "learning_rate": 1.822950748239758e-05, + "loss": 1.0115, + "step": 17584 + }, + { + "epoch": 1.3, + "learning_rate": 1.822928087180897e-05, + "loss": 1.0858, + "step": 17585 + }, + { + "epoch": 1.3, + "learning_rate": 1.8229054248127696e-05, + "loss": 0.9715, + "step": 17586 + }, + { + "epoch": 1.3, + "learning_rate": 1.822882761135411e-05, + "loss": 0.9927, + "step": 17587 + }, + { + "epoch": 1.3, + "learning_rate": 1.822860096148858e-05, + "loss": 0.9403, + "step": 17588 + }, + { + "epoch": 1.3, + "learning_rate": 1.8228374298531462e-05, + "loss": 0.97, + "step": 17589 + }, + { + "epoch": 1.3, + "learning_rate": 1.8228147622483118e-05, + "loss": 1.001, + "step": 17590 + }, + { + "epoch": 1.3, + "learning_rate": 1.822792093334391e-05, + "loss": 1.0158, + "step": 17591 + }, + { + "epoch": 1.3, + "learning_rate": 1.82276942311142e-05, + "loss": 1.0062, + "step": 17592 + }, + { + "epoch": 1.3, + "learning_rate": 1.822746751579434e-05, + "loss": 0.9672, + "step": 17593 + }, + { + "epoch": 1.3, + "learning_rate": 1.82272407873847e-05, + "loss": 0.9686, + "step": 17594 + }, + { + "epoch": 1.3, + "learning_rate": 1.822701404588564e-05, + "loss": 1.0514, + "step": 17595 + }, + { + "epoch": 1.3, + "learning_rate": 1.8226787291297515e-05, + "loss": 0.9905, + "step": 17596 + }, + { + "epoch": 1.3, + "learning_rate": 1.822656052362069e-05, + "loss": 0.921, + "step": 17597 + }, + { + "epoch": 1.3, + "learning_rate": 1.8226333742855527e-05, + "loss": 1.1408, + "step": 17598 + }, + { + "epoch": 1.3, + "learning_rate": 1.8226106949002383e-05, + "loss": 1.03, + "step": 17599 + }, + { + "epoch": 1.3, + "learning_rate": 1.822588014206162e-05, + "loss": 1.0763, + "step": 17600 + }, + { + "epoch": 1.3, + "learning_rate": 1.82256533220336e-05, + "loss": 1.0476, + "step": 17601 + }, + { + "epoch": 1.3, + "learning_rate": 1.8225426488918684e-05, + "loss": 0.9949, + "step": 17602 + }, + { + "epoch": 1.3, + "learning_rate": 1.822519964271723e-05, + "loss": 1.0677, + "step": 17603 + }, + { + "epoch": 1.3, + "learning_rate": 1.82249727834296e-05, + "loss": 1.0499, + "step": 17604 + }, + { + "epoch": 1.3, + "learning_rate": 1.822474591105616e-05, + "loss": 1.0106, + "step": 17605 + }, + { + "epoch": 1.3, + "learning_rate": 1.8224519025597268e-05, + "loss": 1.0031, + "step": 17606 + }, + { + "epoch": 1.3, + "learning_rate": 1.822429212705328e-05, + "loss": 0.9012, + "step": 17607 + }, + { + "epoch": 1.3, + "learning_rate": 1.822406521542456e-05, + "loss": 0.9879, + "step": 17608 + }, + { + "epoch": 1.3, + "learning_rate": 1.8223838290711473e-05, + "loss": 1.0015, + "step": 17609 + }, + { + "epoch": 1.3, + "learning_rate": 1.8223611352914376e-05, + "loss": 1.0233, + "step": 17610 + }, + { + "epoch": 1.3, + "learning_rate": 1.822338440203363e-05, + "loss": 1.0911, + "step": 17611 + }, + { + "epoch": 1.3, + "learning_rate": 1.82231574380696e-05, + "loss": 0.93, + "step": 17612 + }, + { + "epoch": 1.3, + "learning_rate": 1.8222930461022642e-05, + "loss": 1.0363, + "step": 17613 + }, + { + "epoch": 1.3, + "learning_rate": 1.822270347089312e-05, + "loss": 1.0956, + "step": 17614 + }, + { + "epoch": 1.3, + "learning_rate": 1.8222476467681393e-05, + "loss": 1.0406, + "step": 17615 + }, + { + "epoch": 1.3, + "learning_rate": 1.8222249451387822e-05, + "loss": 0.986, + "step": 17616 + }, + { + "epoch": 1.3, + "learning_rate": 1.8222022422012776e-05, + "loss": 1.1317, + "step": 17617 + }, + { + "epoch": 1.3, + "learning_rate": 1.8221795379556604e-05, + "loss": 0.9463, + "step": 17618 + }, + { + "epoch": 1.3, + "learning_rate": 1.8221568324019675e-05, + "loss": 1.0143, + "step": 17619 + }, + { + "epoch": 1.3, + "learning_rate": 1.8221341255402348e-05, + "loss": 0.9986, + "step": 17620 + }, + { + "epoch": 1.3, + "learning_rate": 1.822111417370499e-05, + "loss": 1.122, + "step": 17621 + }, + { + "epoch": 1.3, + "learning_rate": 1.822088707892795e-05, + "loss": 1.0513, + "step": 17622 + }, + { + "epoch": 1.3, + "learning_rate": 1.82206599710716e-05, + "loss": 1.0266, + "step": 17623 + }, + { + "epoch": 1.3, + "learning_rate": 1.8220432850136297e-05, + "loss": 1.0007, + "step": 17624 + }, + { + "epoch": 1.3, + "learning_rate": 1.8220205716122402e-05, + "loss": 0.9878, + "step": 17625 + }, + { + "epoch": 1.3, + "learning_rate": 1.8219978569030278e-05, + "loss": 0.951, + "step": 17626 + }, + { + "epoch": 1.3, + "learning_rate": 1.8219751408860283e-05, + "loss": 0.9817, + "step": 17627 + }, + { + "epoch": 1.3, + "learning_rate": 1.8219524235612785e-05, + "loss": 1.1164, + "step": 17628 + }, + { + "epoch": 1.3, + "learning_rate": 1.821929704928814e-05, + "loss": 0.9123, + "step": 17629 + }, + { + "epoch": 1.3, + "learning_rate": 1.821906984988671e-05, + "loss": 1.0484, + "step": 17630 + }, + { + "epoch": 1.3, + "learning_rate": 1.821884263740886e-05, + "loss": 0.9679, + "step": 17631 + }, + { + "epoch": 1.3, + "learning_rate": 1.8218615411854944e-05, + "loss": 0.9357, + "step": 17632 + }, + { + "epoch": 1.3, + "learning_rate": 1.8218388173225335e-05, + "loss": 1.0368, + "step": 17633 + }, + { + "epoch": 1.3, + "learning_rate": 1.8218160921520386e-05, + "loss": 1.1339, + "step": 17634 + }, + { + "epoch": 1.3, + "learning_rate": 1.8217933656740457e-05, + "loss": 1.0249, + "step": 17635 + }, + { + "epoch": 1.3, + "learning_rate": 1.8217706378885918e-05, + "loss": 0.9641, + "step": 17636 + }, + { + "epoch": 1.3, + "learning_rate": 1.8217479087957123e-05, + "loss": 0.9045, + "step": 17637 + }, + { + "epoch": 1.3, + "learning_rate": 1.8217251783954436e-05, + "loss": 1.0817, + "step": 17638 + }, + { + "epoch": 1.3, + "learning_rate": 1.821702446687822e-05, + "loss": 0.9599, + "step": 17639 + }, + { + "epoch": 1.3, + "learning_rate": 1.8216797136728834e-05, + "loss": 1.0115, + "step": 17640 + }, + { + "epoch": 1.3, + "learning_rate": 1.821656979350664e-05, + "loss": 1.0321, + "step": 17641 + }, + { + "epoch": 1.3, + "learning_rate": 1.8216342437212006e-05, + "loss": 0.9828, + "step": 17642 + }, + { + "epoch": 1.3, + "learning_rate": 1.8216115067845287e-05, + "loss": 1.0124, + "step": 17643 + }, + { + "epoch": 1.3, + "learning_rate": 1.8215887685406844e-05, + "loss": 1.0272, + "step": 17644 + }, + { + "epoch": 1.3, + "learning_rate": 1.8215660289897043e-05, + "loss": 0.9242, + "step": 17645 + }, + { + "epoch": 1.3, + "learning_rate": 1.8215432881316246e-05, + "loss": 1.0024, + "step": 17646 + }, + { + "epoch": 1.3, + "learning_rate": 1.8215205459664813e-05, + "loss": 0.9699, + "step": 17647 + }, + { + "epoch": 1.3, + "learning_rate": 1.8214978024943102e-05, + "loss": 0.956, + "step": 17648 + }, + { + "epoch": 1.3, + "learning_rate": 1.821475057715148e-05, + "loss": 1.0012, + "step": 17649 + }, + { + "epoch": 1.3, + "learning_rate": 1.8214523116290305e-05, + "loss": 0.9499, + "step": 17650 + }, + { + "epoch": 1.3, + "learning_rate": 1.8214295642359947e-05, + "loss": 1.0507, + "step": 17651 + }, + { + "epoch": 1.3, + "learning_rate": 1.821406815536076e-05, + "loss": 1.0887, + "step": 17652 + }, + { + "epoch": 1.3, + "learning_rate": 1.8213840655293105e-05, + "loss": 0.9959, + "step": 17653 + }, + { + "epoch": 1.3, + "learning_rate": 1.821361314215735e-05, + "loss": 0.9768, + "step": 17654 + }, + { + "epoch": 1.3, + "learning_rate": 1.821338561595385e-05, + "loss": 1.1706, + "step": 17655 + }, + { + "epoch": 1.3, + "learning_rate": 1.8213158076682976e-05, + "loss": 1.1112, + "step": 17656 + }, + { + "epoch": 1.3, + "learning_rate": 1.8212930524345082e-05, + "loss": 1.057, + "step": 17657 + }, + { + "epoch": 1.3, + "learning_rate": 1.8212702958940536e-05, + "loss": 1.0454, + "step": 17658 + }, + { + "epoch": 1.3, + "learning_rate": 1.8212475380469694e-05, + "loss": 1.1122, + "step": 17659 + }, + { + "epoch": 1.3, + "learning_rate": 1.821224778893292e-05, + "loss": 1.0307, + "step": 17660 + }, + { + "epoch": 1.3, + "learning_rate": 1.8212020184330584e-05, + "loss": 0.9532, + "step": 17661 + }, + { + "epoch": 1.31, + "learning_rate": 1.8211792566663035e-05, + "loss": 1.0521, + "step": 17662 + }, + { + "epoch": 1.31, + "learning_rate": 1.8211564935930643e-05, + "loss": 1.0386, + "step": 17663 + }, + { + "epoch": 1.31, + "learning_rate": 1.821133729213377e-05, + "loss": 1.0065, + "step": 17664 + }, + { + "epoch": 1.31, + "learning_rate": 1.8211109635272776e-05, + "loss": 1.0132, + "step": 17665 + }, + { + "epoch": 1.31, + "learning_rate": 1.8210881965348024e-05, + "loss": 1.0479, + "step": 17666 + }, + { + "epoch": 1.31, + "learning_rate": 1.8210654282359875e-05, + "loss": 1.0169, + "step": 17667 + }, + { + "epoch": 1.31, + "learning_rate": 1.8210426586308693e-05, + "loss": 1.0482, + "step": 17668 + }, + { + "epoch": 1.31, + "learning_rate": 1.8210198877194844e-05, + "loss": 1.0494, + "step": 17669 + }, + { + "epoch": 1.31, + "learning_rate": 1.820997115501868e-05, + "loss": 1.0517, + "step": 17670 + }, + { + "epoch": 1.31, + "learning_rate": 1.820974341978057e-05, + "loss": 1.0099, + "step": 17671 + }, + { + "epoch": 1.31, + "learning_rate": 1.8209515671480883e-05, + "loss": 0.9961, + "step": 17672 + }, + { + "epoch": 1.31, + "learning_rate": 1.8209287910119968e-05, + "loss": 0.9269, + "step": 17673 + }, + { + "epoch": 1.31, + "learning_rate": 1.820906013569819e-05, + "loss": 1.0385, + "step": 17674 + }, + { + "epoch": 1.31, + "learning_rate": 1.8208832348215924e-05, + "loss": 1.0664, + "step": 17675 + }, + { + "epoch": 1.31, + "learning_rate": 1.8208604547673518e-05, + "loss": 0.9694, + "step": 17676 + }, + { + "epoch": 1.31, + "learning_rate": 1.8208376734071337e-05, + "loss": 0.9076, + "step": 17677 + }, + { + "epoch": 1.31, + "learning_rate": 1.8208148907409748e-05, + "loss": 1.0677, + "step": 17678 + }, + { + "epoch": 1.31, + "learning_rate": 1.8207921067689113e-05, + "loss": 1.0829, + "step": 17679 + }, + { + "epoch": 1.31, + "learning_rate": 1.820769321490979e-05, + "loss": 1.0221, + "step": 17680 + }, + { + "epoch": 1.31, + "learning_rate": 1.820746534907215e-05, + "loss": 0.9658, + "step": 17681 + }, + { + "epoch": 1.31, + "learning_rate": 1.8207237470176546e-05, + "loss": 1.0496, + "step": 17682 + }, + { + "epoch": 1.31, + "learning_rate": 1.8207009578223343e-05, + "loss": 1.0576, + "step": 17683 + }, + { + "epoch": 1.31, + "learning_rate": 1.820678167321291e-05, + "loss": 1.0458, + "step": 17684 + }, + { + "epoch": 1.31, + "learning_rate": 1.8206553755145604e-05, + "loss": 1.0952, + "step": 17685 + }, + { + "epoch": 1.31, + "learning_rate": 1.8206325824021783e-05, + "loss": 1.0764, + "step": 17686 + }, + { + "epoch": 1.31, + "learning_rate": 1.8206097879841823e-05, + "loss": 1.0015, + "step": 17687 + }, + { + "epoch": 1.31, + "learning_rate": 1.8205869922606074e-05, + "loss": 1.026, + "step": 17688 + }, + { + "epoch": 1.31, + "learning_rate": 1.8205641952314904e-05, + "loss": 1.0469, + "step": 17689 + }, + { + "epoch": 1.31, + "learning_rate": 1.8205413968968675e-05, + "loss": 1.0022, + "step": 17690 + }, + { + "epoch": 1.31, + "learning_rate": 1.820518597256775e-05, + "loss": 1.0283, + "step": 17691 + }, + { + "epoch": 1.31, + "learning_rate": 1.8204957963112492e-05, + "loss": 0.997, + "step": 17692 + }, + { + "epoch": 1.31, + "learning_rate": 1.8204729940603262e-05, + "loss": 0.993, + "step": 17693 + }, + { + "epoch": 1.31, + "learning_rate": 1.8204501905040424e-05, + "loss": 0.9073, + "step": 17694 + }, + { + "epoch": 1.31, + "learning_rate": 1.820427385642434e-05, + "loss": 1.0618, + "step": 17695 + }, + { + "epoch": 1.31, + "learning_rate": 1.8204045794755375e-05, + "loss": 1.0678, + "step": 17696 + }, + { + "epoch": 1.31, + "learning_rate": 1.8203817720033892e-05, + "loss": 1.1081, + "step": 17697 + }, + { + "epoch": 1.31, + "learning_rate": 1.820358963226025e-05, + "loss": 0.9715, + "step": 17698 + }, + { + "epoch": 1.31, + "learning_rate": 1.8203361531434816e-05, + "loss": 0.9751, + "step": 17699 + }, + { + "epoch": 1.31, + "learning_rate": 1.820313341755795e-05, + "loss": 1.0638, + "step": 17700 + }, + { + "epoch": 1.31, + "learning_rate": 1.8202905290630016e-05, + "loss": 0.9212, + "step": 17701 + }, + { + "epoch": 1.31, + "learning_rate": 1.8202677150651377e-05, + "loss": 0.9673, + "step": 17702 + }, + { + "epoch": 1.31, + "learning_rate": 1.82024489976224e-05, + "loss": 0.9071, + "step": 17703 + }, + { + "epoch": 1.31, + "learning_rate": 1.8202220831543437e-05, + "loss": 1.0318, + "step": 17704 + }, + { + "epoch": 1.31, + "learning_rate": 1.8201992652414864e-05, + "loss": 1.0449, + "step": 17705 + }, + { + "epoch": 1.31, + "learning_rate": 1.8201764460237033e-05, + "loss": 1.0443, + "step": 17706 + }, + { + "epoch": 1.31, + "learning_rate": 1.8201536255010318e-05, + "loss": 1.0225, + "step": 17707 + }, + { + "epoch": 1.31, + "learning_rate": 1.8201308036735073e-05, + "loss": 1.0346, + "step": 17708 + }, + { + "epoch": 1.31, + "learning_rate": 1.8201079805411663e-05, + "loss": 1.034, + "step": 17709 + }, + { + "epoch": 1.31, + "learning_rate": 1.8200851561040455e-05, + "loss": 0.9861, + "step": 17710 + }, + { + "epoch": 1.31, + "learning_rate": 1.820062330362181e-05, + "loss": 1.005, + "step": 17711 + }, + { + "epoch": 1.31, + "learning_rate": 1.8200395033156087e-05, + "loss": 0.983, + "step": 17712 + }, + { + "epoch": 1.31, + "learning_rate": 1.8200166749643653e-05, + "loss": 0.9254, + "step": 17713 + }, + { + "epoch": 1.31, + "learning_rate": 1.8199938453084875e-05, + "loss": 1.0643, + "step": 17714 + }, + { + "epoch": 1.31, + "learning_rate": 1.8199710143480106e-05, + "loss": 1.1589, + "step": 17715 + }, + { + "epoch": 1.31, + "learning_rate": 1.819948182082972e-05, + "loss": 0.9323, + "step": 17716 + }, + { + "epoch": 1.31, + "learning_rate": 1.8199253485134077e-05, + "loss": 1.0131, + "step": 17717 + }, + { + "epoch": 1.31, + "learning_rate": 1.819902513639354e-05, + "loss": 0.9133, + "step": 17718 + }, + { + "epoch": 1.31, + "learning_rate": 1.8198796774608464e-05, + "loss": 0.9613, + "step": 17719 + }, + { + "epoch": 1.31, + "learning_rate": 1.8198568399779227e-05, + "loss": 1.0162, + "step": 17720 + }, + { + "epoch": 1.31, + "learning_rate": 1.819834001190618e-05, + "loss": 1.058, + "step": 17721 + }, + { + "epoch": 1.31, + "learning_rate": 1.8198111610989694e-05, + "loss": 1.0964, + "step": 17722 + }, + { + "epoch": 1.31, + "learning_rate": 1.8197883197030127e-05, + "loss": 0.9731, + "step": 17723 + }, + { + "epoch": 1.31, + "learning_rate": 1.8197654770027848e-05, + "loss": 0.9651, + "step": 17724 + }, + { + "epoch": 1.31, + "learning_rate": 1.8197426329983217e-05, + "loss": 1.0749, + "step": 17725 + }, + { + "epoch": 1.31, + "learning_rate": 1.81971978768966e-05, + "loss": 0.9977, + "step": 17726 + }, + { + "epoch": 1.31, + "learning_rate": 1.8196969410768354e-05, + "loss": 1.0398, + "step": 17727 + }, + { + "epoch": 1.31, + "learning_rate": 1.8196740931598847e-05, + "loss": 1.0735, + "step": 17728 + }, + { + "epoch": 1.31, + "learning_rate": 1.819651243938845e-05, + "loss": 1.0512, + "step": 17729 + }, + { + "epoch": 1.31, + "learning_rate": 1.819628393413751e-05, + "loss": 0.9554, + "step": 17730 + }, + { + "epoch": 1.31, + "learning_rate": 1.8196055415846404e-05, + "loss": 1.0632, + "step": 17731 + }, + { + "epoch": 1.31, + "learning_rate": 1.8195826884515487e-05, + "loss": 0.9416, + "step": 17732 + }, + { + "epoch": 1.31, + "learning_rate": 1.8195598340145132e-05, + "loss": 0.9404, + "step": 17733 + }, + { + "epoch": 1.31, + "learning_rate": 1.8195369782735697e-05, + "loss": 0.9812, + "step": 17734 + }, + { + "epoch": 1.31, + "learning_rate": 1.8195141212287544e-05, + "loss": 0.9092, + "step": 17735 + }, + { + "epoch": 1.31, + "learning_rate": 1.8194912628801038e-05, + "loss": 1.1588, + "step": 17736 + }, + { + "epoch": 1.31, + "learning_rate": 1.8194684032276543e-05, + "loss": 0.9382, + "step": 17737 + }, + { + "epoch": 1.31, + "learning_rate": 1.8194455422714425e-05, + "loss": 1.0776, + "step": 17738 + }, + { + "epoch": 1.31, + "learning_rate": 1.8194226800115044e-05, + "loss": 0.9556, + "step": 17739 + }, + { + "epoch": 1.31, + "learning_rate": 1.819399816447877e-05, + "loss": 1.0772, + "step": 17740 + }, + { + "epoch": 1.31, + "learning_rate": 1.8193769515805955e-05, + "loss": 0.9968, + "step": 17741 + }, + { + "epoch": 1.31, + "learning_rate": 1.8193540854096972e-05, + "loss": 1.0239, + "step": 17742 + }, + { + "epoch": 1.31, + "learning_rate": 1.8193312179352186e-05, + "loss": 1.0377, + "step": 17743 + }, + { + "epoch": 1.31, + "learning_rate": 1.8193083491571953e-05, + "loss": 1.0258, + "step": 17744 + }, + { + "epoch": 1.31, + "learning_rate": 1.8192854790756645e-05, + "loss": 1.1148, + "step": 17745 + }, + { + "epoch": 1.31, + "learning_rate": 1.819262607690662e-05, + "loss": 1.1884, + "step": 17746 + }, + { + "epoch": 1.31, + "learning_rate": 1.8192397350022248e-05, + "loss": 0.9707, + "step": 17747 + }, + { + "epoch": 1.31, + "learning_rate": 1.8192168610103884e-05, + "loss": 1.0072, + "step": 17748 + }, + { + "epoch": 1.31, + "learning_rate": 1.81919398571519e-05, + "loss": 1.0307, + "step": 17749 + }, + { + "epoch": 1.31, + "learning_rate": 1.8191711091166655e-05, + "loss": 0.8983, + "step": 17750 + }, + { + "epoch": 1.31, + "learning_rate": 1.8191482312148515e-05, + "loss": 1.0967, + "step": 17751 + }, + { + "epoch": 1.31, + "learning_rate": 1.8191253520097848e-05, + "loss": 1.0424, + "step": 17752 + }, + { + "epoch": 1.31, + "learning_rate": 1.819102471501501e-05, + "loss": 1.0111, + "step": 17753 + }, + { + "epoch": 1.31, + "learning_rate": 1.8190795896900367e-05, + "loss": 1.1256, + "step": 17754 + }, + { + "epoch": 1.31, + "learning_rate": 1.8190567065754288e-05, + "loss": 1.1828, + "step": 17755 + }, + { + "epoch": 1.31, + "learning_rate": 1.8190338221577132e-05, + "loss": 1.0399, + "step": 17756 + }, + { + "epoch": 1.31, + "learning_rate": 1.819010936436927e-05, + "loss": 0.9576, + "step": 17757 + }, + { + "epoch": 1.31, + "learning_rate": 1.8189880494131055e-05, + "loss": 0.9799, + "step": 17758 + }, + { + "epoch": 1.31, + "learning_rate": 1.818965161086286e-05, + "loss": 1.0656, + "step": 17759 + }, + { + "epoch": 1.31, + "learning_rate": 1.8189422714565047e-05, + "loss": 1.0378, + "step": 17760 + }, + { + "epoch": 1.31, + "learning_rate": 1.818919380523798e-05, + "loss": 1.0046, + "step": 17761 + }, + { + "epoch": 1.31, + "learning_rate": 1.818896488288202e-05, + "loss": 0.9853, + "step": 17762 + }, + { + "epoch": 1.31, + "learning_rate": 1.8188735947497536e-05, + "loss": 1.0842, + "step": 17763 + }, + { + "epoch": 1.31, + "learning_rate": 1.8188506999084892e-05, + "loss": 0.9872, + "step": 17764 + }, + { + "epoch": 1.31, + "learning_rate": 1.8188278037644448e-05, + "loss": 1.066, + "step": 17765 + }, + { + "epoch": 1.31, + "learning_rate": 1.8188049063176572e-05, + "loss": 1.0948, + "step": 17766 + }, + { + "epoch": 1.31, + "learning_rate": 1.8187820075681625e-05, + "loss": 1.0094, + "step": 17767 + }, + { + "epoch": 1.31, + "learning_rate": 1.8187591075159975e-05, + "loss": 0.91, + "step": 17768 + }, + { + "epoch": 1.31, + "learning_rate": 1.8187362061611985e-05, + "loss": 0.9764, + "step": 17769 + }, + { + "epoch": 1.31, + "learning_rate": 1.818713303503802e-05, + "loss": 1.0586, + "step": 17770 + }, + { + "epoch": 1.31, + "learning_rate": 1.818690399543844e-05, + "loss": 0.9894, + "step": 17771 + }, + { + "epoch": 1.31, + "learning_rate": 1.818667494281362e-05, + "loss": 1.1161, + "step": 17772 + }, + { + "epoch": 1.31, + "learning_rate": 1.818644587716391e-05, + "loss": 1.0463, + "step": 17773 + }, + { + "epoch": 1.31, + "learning_rate": 1.818621679848968e-05, + "loss": 0.9904, + "step": 17774 + }, + { + "epoch": 1.31, + "learning_rate": 1.8185987706791304e-05, + "loss": 1.0467, + "step": 17775 + }, + { + "epoch": 1.31, + "learning_rate": 1.8185758602069135e-05, + "loss": 0.9457, + "step": 17776 + }, + { + "epoch": 1.31, + "learning_rate": 1.8185529484323542e-05, + "loss": 1.0812, + "step": 17777 + }, + { + "epoch": 1.31, + "learning_rate": 1.818530035355489e-05, + "loss": 1.0142, + "step": 17778 + }, + { + "epoch": 1.31, + "learning_rate": 1.818507120976354e-05, + "loss": 1.1043, + "step": 17779 + }, + { + "epoch": 1.31, + "learning_rate": 1.8184842052949856e-05, + "loss": 1.1089, + "step": 17780 + }, + { + "epoch": 1.31, + "learning_rate": 1.8184612883114208e-05, + "loss": 0.9689, + "step": 17781 + }, + { + "epoch": 1.31, + "learning_rate": 1.818438370025696e-05, + "loss": 0.9976, + "step": 17782 + }, + { + "epoch": 1.31, + "learning_rate": 1.818415450437847e-05, + "loss": 0.9587, + "step": 17783 + }, + { + "epoch": 1.31, + "learning_rate": 1.818392529547911e-05, + "loss": 1.0468, + "step": 17784 + }, + { + "epoch": 1.31, + "learning_rate": 1.8183696073559244e-05, + "loss": 1.0359, + "step": 17785 + }, + { + "epoch": 1.31, + "learning_rate": 1.8183466838619233e-05, + "loss": 1.0324, + "step": 17786 + }, + { + "epoch": 1.31, + "learning_rate": 1.8183237590659442e-05, + "loss": 1.0286, + "step": 17787 + }, + { + "epoch": 1.31, + "learning_rate": 1.818300832968024e-05, + "loss": 0.9302, + "step": 17788 + }, + { + "epoch": 1.31, + "learning_rate": 1.8182779055681988e-05, + "loss": 1.0966, + "step": 17789 + }, + { + "epoch": 1.31, + "learning_rate": 1.818254976866505e-05, + "loss": 1.0303, + "step": 17790 + }, + { + "epoch": 1.31, + "learning_rate": 1.8182320468629792e-05, + "loss": 1.1003, + "step": 17791 + }, + { + "epoch": 1.31, + "learning_rate": 1.8182091155576583e-05, + "loss": 1.118, + "step": 17792 + }, + { + "epoch": 1.31, + "learning_rate": 1.8181861829505782e-05, + "loss": 0.9068, + "step": 17793 + }, + { + "epoch": 1.31, + "learning_rate": 1.8181632490417755e-05, + "loss": 0.949, + "step": 17794 + }, + { + "epoch": 1.31, + "learning_rate": 1.8181403138312864e-05, + "loss": 0.9598, + "step": 17795 + }, + { + "epoch": 1.31, + "learning_rate": 1.8181173773191485e-05, + "loss": 1.0129, + "step": 17796 + }, + { + "epoch": 1.32, + "learning_rate": 1.818094439505397e-05, + "loss": 1.096, + "step": 17797 + }, + { + "epoch": 1.32, + "learning_rate": 1.8180715003900693e-05, + "loss": 0.9591, + "step": 17798 + }, + { + "epoch": 1.32, + "learning_rate": 1.8180485599732014e-05, + "loss": 1.0169, + "step": 17799 + }, + { + "epoch": 1.32, + "learning_rate": 1.81802561825483e-05, + "loss": 0.9444, + "step": 17800 + }, + { + "epoch": 1.32, + "learning_rate": 1.8180026752349914e-05, + "loss": 0.9845, + "step": 17801 + }, + { + "epoch": 1.32, + "learning_rate": 1.8179797309137227e-05, + "loss": 0.8899, + "step": 17802 + }, + { + "epoch": 1.32, + "learning_rate": 1.8179567852910595e-05, + "loss": 1.0335, + "step": 17803 + }, + { + "epoch": 1.32, + "learning_rate": 1.8179338383670388e-05, + "loss": 1.1182, + "step": 17804 + }, + { + "epoch": 1.32, + "learning_rate": 1.8179108901416974e-05, + "loss": 1.1313, + "step": 17805 + }, + { + "epoch": 1.32, + "learning_rate": 1.817887940615071e-05, + "loss": 1.0021, + "step": 17806 + }, + { + "epoch": 1.32, + "learning_rate": 1.817864989787197e-05, + "loss": 1.0629, + "step": 17807 + }, + { + "epoch": 1.32, + "learning_rate": 1.817842037658111e-05, + "loss": 0.9576, + "step": 17808 + }, + { + "epoch": 1.32, + "learning_rate": 1.8178190842278505e-05, + "loss": 1.0079, + "step": 17809 + }, + { + "epoch": 1.32, + "learning_rate": 1.8177961294964514e-05, + "loss": 0.9812, + "step": 17810 + }, + { + "epoch": 1.32, + "learning_rate": 1.8177731734639504e-05, + "loss": 0.9969, + "step": 17811 + }, + { + "epoch": 1.32, + "learning_rate": 1.8177502161303837e-05, + "loss": 1.0529, + "step": 17812 + }, + { + "epoch": 1.32, + "learning_rate": 1.8177272574957884e-05, + "loss": 0.9411, + "step": 17813 + }, + { + "epoch": 1.32, + "learning_rate": 1.8177042975602005e-05, + "loss": 1.0244, + "step": 17814 + }, + { + "epoch": 1.32, + "learning_rate": 1.817681336323657e-05, + "loss": 1.0798, + "step": 17815 + }, + { + "epoch": 1.32, + "learning_rate": 1.8176583737861942e-05, + "loss": 1.157, + "step": 17816 + }, + { + "epoch": 1.32, + "learning_rate": 1.8176354099478485e-05, + "loss": 1.0646, + "step": 17817 + }, + { + "epoch": 1.32, + "learning_rate": 1.8176124448086564e-05, + "loss": 0.9957, + "step": 17818 + }, + { + "epoch": 1.32, + "learning_rate": 1.8175894783686548e-05, + "loss": 0.9277, + "step": 17819 + }, + { + "epoch": 1.32, + "learning_rate": 1.8175665106278803e-05, + "loss": 1.0149, + "step": 17820 + }, + { + "epoch": 1.32, + "learning_rate": 1.8175435415863688e-05, + "loss": 1.0365, + "step": 17821 + }, + { + "epoch": 1.32, + "learning_rate": 1.8175205712441576e-05, + "loss": 1.0183, + "step": 17822 + }, + { + "epoch": 1.32, + "learning_rate": 1.8174975996012825e-05, + "loss": 0.9794, + "step": 17823 + }, + { + "epoch": 1.32, + "learning_rate": 1.8174746266577802e-05, + "loss": 0.9434, + "step": 17824 + }, + { + "epoch": 1.32, + "learning_rate": 1.8174516524136877e-05, + "loss": 1.0388, + "step": 17825 + }, + { + "epoch": 1.32, + "learning_rate": 1.8174286768690414e-05, + "loss": 0.9889, + "step": 17826 + }, + { + "epoch": 1.32, + "learning_rate": 1.817405700023878e-05, + "loss": 1.079, + "step": 17827 + }, + { + "epoch": 1.32, + "learning_rate": 1.8173827218782335e-05, + "loss": 1.0978, + "step": 17828 + }, + { + "epoch": 1.32, + "learning_rate": 1.817359742432145e-05, + "loss": 1.008, + "step": 17829 + }, + { + "epoch": 1.32, + "learning_rate": 1.8173367616856485e-05, + "loss": 0.9957, + "step": 17830 + }, + { + "epoch": 1.32, + "learning_rate": 1.8173137796387812e-05, + "loss": 0.9652, + "step": 17831 + }, + { + "epoch": 1.32, + "learning_rate": 1.8172907962915795e-05, + "loss": 1.084, + "step": 17832 + }, + { + "epoch": 1.32, + "learning_rate": 1.8172678116440795e-05, + "loss": 1.0959, + "step": 17833 + }, + { + "epoch": 1.32, + "learning_rate": 1.8172448256963183e-05, + "loss": 0.9326, + "step": 17834 + }, + { + "epoch": 1.32, + "learning_rate": 1.8172218384483323e-05, + "loss": 1.0135, + "step": 17835 + }, + { + "epoch": 1.32, + "learning_rate": 1.817198849900158e-05, + "loss": 0.9297, + "step": 17836 + }, + { + "epoch": 1.32, + "learning_rate": 1.817175860051832e-05, + "loss": 1.007, + "step": 17837 + }, + { + "epoch": 1.32, + "learning_rate": 1.817152868903391e-05, + "loss": 0.9906, + "step": 17838 + }, + { + "epoch": 1.32, + "learning_rate": 1.817129876454871e-05, + "loss": 0.9997, + "step": 17839 + }, + { + "epoch": 1.32, + "learning_rate": 1.8171068827063094e-05, + "loss": 1.031, + "step": 17840 + }, + { + "epoch": 1.32, + "learning_rate": 1.8170838876577422e-05, + "loss": 1.0787, + "step": 17841 + }, + { + "epoch": 1.32, + "learning_rate": 1.8170608913092068e-05, + "loss": 0.9987, + "step": 17842 + }, + { + "epoch": 1.32, + "learning_rate": 1.8170378936607387e-05, + "loss": 1.0465, + "step": 17843 + }, + { + "epoch": 1.32, + "learning_rate": 1.8170148947123753e-05, + "loss": 1.0207, + "step": 17844 + }, + { + "epoch": 1.32, + "learning_rate": 1.816991894464153e-05, + "loss": 1.0314, + "step": 17845 + }, + { + "epoch": 1.32, + "learning_rate": 1.816968892916108e-05, + "loss": 0.9913, + "step": 17846 + }, + { + "epoch": 1.32, + "learning_rate": 1.8169458900682772e-05, + "loss": 0.9515, + "step": 17847 + }, + { + "epoch": 1.32, + "learning_rate": 1.8169228859206974e-05, + "loss": 1.005, + "step": 17848 + }, + { + "epoch": 1.32, + "learning_rate": 1.8168998804734046e-05, + "loss": 0.9987, + "step": 17849 + }, + { + "epoch": 1.32, + "learning_rate": 1.816876873726436e-05, + "loss": 0.9639, + "step": 17850 + }, + { + "epoch": 1.32, + "learning_rate": 1.816853865679828e-05, + "loss": 1.0437, + "step": 17851 + }, + { + "epoch": 1.32, + "learning_rate": 1.816830856333617e-05, + "loss": 1.0434, + "step": 17852 + }, + { + "epoch": 1.32, + "learning_rate": 1.81680784568784e-05, + "loss": 0.9325, + "step": 17853 + }, + { + "epoch": 1.32, + "learning_rate": 1.8167848337425336e-05, + "loss": 0.981, + "step": 17854 + }, + { + "epoch": 1.32, + "learning_rate": 1.8167618204977336e-05, + "loss": 0.9167, + "step": 17855 + }, + { + "epoch": 1.32, + "learning_rate": 1.816738805953478e-05, + "loss": 1.0113, + "step": 17856 + }, + { + "epoch": 1.32, + "learning_rate": 1.816715790109802e-05, + "loss": 1.1086, + "step": 17857 + }, + { + "epoch": 1.32, + "learning_rate": 1.8166927729667433e-05, + "loss": 1.004, + "step": 17858 + }, + { + "epoch": 1.32, + "learning_rate": 1.8166697545243377e-05, + "loss": 0.8974, + "step": 17859 + }, + { + "epoch": 1.32, + "learning_rate": 1.8166467347826225e-05, + "loss": 1.0944, + "step": 17860 + }, + { + "epoch": 1.32, + "learning_rate": 1.8166237137416337e-05, + "loss": 1.0515, + "step": 17861 + }, + { + "epoch": 1.32, + "learning_rate": 1.816600691401409e-05, + "loss": 1.0474, + "step": 17862 + }, + { + "epoch": 1.32, + "learning_rate": 1.816577667761983e-05, + "loss": 1.0032, + "step": 17863 + }, + { + "epoch": 1.32, + "learning_rate": 1.8165546428233946e-05, + "loss": 1.0164, + "step": 17864 + }, + { + "epoch": 1.32, + "learning_rate": 1.8165316165856794e-05, + "loss": 1.0636, + "step": 17865 + }, + { + "epoch": 1.32, + "learning_rate": 1.8165085890488738e-05, + "loss": 1.0743, + "step": 17866 + }, + { + "epoch": 1.32, + "learning_rate": 1.816485560213015e-05, + "loss": 1.0368, + "step": 17867 + }, + { + "epoch": 1.32, + "learning_rate": 1.816462530078139e-05, + "loss": 0.9865, + "step": 17868 + }, + { + "epoch": 1.32, + "learning_rate": 1.816439498644283e-05, + "loss": 1.0726, + "step": 17869 + }, + { + "epoch": 1.32, + "learning_rate": 1.8164164659114835e-05, + "loss": 0.9786, + "step": 17870 + }, + { + "epoch": 1.32, + "learning_rate": 1.816393431879777e-05, + "loss": 1.0943, + "step": 17871 + }, + { + "epoch": 1.32, + "learning_rate": 1.8163703965492004e-05, + "loss": 1.0476, + "step": 17872 + }, + { + "epoch": 1.32, + "learning_rate": 1.81634735991979e-05, + "loss": 1.1355, + "step": 17873 + }, + { + "epoch": 1.32, + "learning_rate": 1.8163243219915825e-05, + "loss": 1.0032, + "step": 17874 + }, + { + "epoch": 1.32, + "learning_rate": 1.8163012827646147e-05, + "loss": 0.9774, + "step": 17875 + }, + { + "epoch": 1.32, + "learning_rate": 1.8162782422389234e-05, + "loss": 0.9602, + "step": 17876 + }, + { + "epoch": 1.32, + "learning_rate": 1.816255200414545e-05, + "loss": 1.0396, + "step": 17877 + }, + { + "epoch": 1.32, + "learning_rate": 1.8162321572915164e-05, + "loss": 0.8459, + "step": 17878 + }, + { + "epoch": 1.32, + "learning_rate": 1.816209112869874e-05, + "loss": 1.0131, + "step": 17879 + }, + { + "epoch": 1.32, + "learning_rate": 1.8161860671496547e-05, + "loss": 1.0983, + "step": 17880 + }, + { + "epoch": 1.32, + "learning_rate": 1.8161630201308948e-05, + "loss": 1.0998, + "step": 17881 + }, + { + "epoch": 1.32, + "learning_rate": 1.8161399718136314e-05, + "loss": 1.0661, + "step": 17882 + }, + { + "epoch": 1.32, + "learning_rate": 1.816116922197901e-05, + "loss": 1.0282, + "step": 17883 + }, + { + "epoch": 1.32, + "learning_rate": 1.81609387128374e-05, + "loss": 0.9264, + "step": 17884 + }, + { + "epoch": 1.32, + "learning_rate": 1.8160708190711856e-05, + "loss": 1.0144, + "step": 17885 + }, + { + "epoch": 1.32, + "learning_rate": 1.8160477655602743e-05, + "loss": 0.9611, + "step": 17886 + }, + { + "epoch": 1.32, + "learning_rate": 1.8160247107510424e-05, + "loss": 0.973, + "step": 17887 + }, + { + "epoch": 1.32, + "learning_rate": 1.8160016546435272e-05, + "loss": 0.976, + "step": 17888 + }, + { + "epoch": 1.32, + "learning_rate": 1.8159785972377646e-05, + "loss": 0.9898, + "step": 17889 + }, + { + "epoch": 1.32, + "learning_rate": 1.815955538533792e-05, + "loss": 1.0103, + "step": 17890 + }, + { + "epoch": 1.32, + "learning_rate": 1.8159324785316456e-05, + "loss": 0.9674, + "step": 17891 + }, + { + "epoch": 1.32, + "learning_rate": 1.8159094172313624e-05, + "loss": 0.9526, + "step": 17892 + }, + { + "epoch": 1.32, + "learning_rate": 1.8158863546329792e-05, + "loss": 0.9324, + "step": 17893 + }, + { + "epoch": 1.32, + "learning_rate": 1.8158632907365322e-05, + "loss": 1.0971, + "step": 17894 + }, + { + "epoch": 1.32, + "learning_rate": 1.8158402255420587e-05, + "loss": 0.9381, + "step": 17895 + }, + { + "epoch": 1.32, + "learning_rate": 1.8158171590495947e-05, + "loss": 1.0325, + "step": 17896 + }, + { + "epoch": 1.32, + "learning_rate": 1.8157940912591777e-05, + "loss": 0.8717, + "step": 17897 + }, + { + "epoch": 1.32, + "learning_rate": 1.8157710221708432e-05, + "loss": 1.0974, + "step": 17898 + }, + { + "epoch": 1.32, + "learning_rate": 1.8157479517846294e-05, + "loss": 1.1745, + "step": 17899 + }, + { + "epoch": 1.32, + "learning_rate": 1.8157248801005723e-05, + "loss": 1.061, + "step": 17900 + }, + { + "epoch": 1.32, + "learning_rate": 1.8157018071187083e-05, + "loss": 1.052, + "step": 17901 + }, + { + "epoch": 1.32, + "learning_rate": 1.8156787328390743e-05, + "loss": 1.0196, + "step": 17902 + }, + { + "epoch": 1.32, + "learning_rate": 1.8156556572617076e-05, + "loss": 1.0633, + "step": 17903 + }, + { + "epoch": 1.32, + "learning_rate": 1.815632580386644e-05, + "loss": 0.9492, + "step": 17904 + }, + { + "epoch": 1.32, + "learning_rate": 1.8156095022139205e-05, + "loss": 1.0079, + "step": 17905 + }, + { + "epoch": 1.32, + "learning_rate": 1.815586422743574e-05, + "loss": 1.0332, + "step": 17906 + }, + { + "epoch": 1.32, + "learning_rate": 1.8155633419756416e-05, + "loss": 0.9614, + "step": 17907 + }, + { + "epoch": 1.32, + "learning_rate": 1.8155402599101594e-05, + "loss": 0.9912, + "step": 17908 + }, + { + "epoch": 1.32, + "learning_rate": 1.8155171765471643e-05, + "loss": 1.0838, + "step": 17909 + }, + { + "epoch": 1.32, + "learning_rate": 1.815494091886693e-05, + "loss": 0.9436, + "step": 17910 + }, + { + "epoch": 1.32, + "learning_rate": 1.8154710059287822e-05, + "loss": 1.056, + "step": 17911 + }, + { + "epoch": 1.32, + "learning_rate": 1.8154479186734688e-05, + "loss": 1.0103, + "step": 17912 + }, + { + "epoch": 1.32, + "learning_rate": 1.8154248301207892e-05, + "loss": 0.8998, + "step": 17913 + }, + { + "epoch": 1.32, + "learning_rate": 1.8154017402707807e-05, + "loss": 1.0199, + "step": 17914 + }, + { + "epoch": 1.32, + "learning_rate": 1.8153786491234797e-05, + "loss": 1.0432, + "step": 17915 + }, + { + "epoch": 1.32, + "learning_rate": 1.8153555566789226e-05, + "loss": 1.0601, + "step": 17916 + }, + { + "epoch": 1.32, + "learning_rate": 1.8153324629371467e-05, + "loss": 0.939, + "step": 17917 + }, + { + "epoch": 1.32, + "learning_rate": 1.8153093678981883e-05, + "loss": 1.0579, + "step": 17918 + }, + { + "epoch": 1.32, + "learning_rate": 1.8152862715620844e-05, + "loss": 1.0384, + "step": 17919 + }, + { + "epoch": 1.32, + "learning_rate": 1.815263173928872e-05, + "loss": 1.1214, + "step": 17920 + }, + { + "epoch": 1.32, + "learning_rate": 1.815240074998587e-05, + "loss": 1.0176, + "step": 17921 + }, + { + "epoch": 1.32, + "learning_rate": 1.8152169747712673e-05, + "loss": 1.049, + "step": 17922 + }, + { + "epoch": 1.32, + "learning_rate": 1.815193873246949e-05, + "loss": 1.0393, + "step": 17923 + }, + { + "epoch": 1.32, + "learning_rate": 1.8151707704256688e-05, + "loss": 1.0905, + "step": 17924 + }, + { + "epoch": 1.32, + "learning_rate": 1.8151476663074635e-05, + "loss": 0.9775, + "step": 17925 + }, + { + "epoch": 1.32, + "learning_rate": 1.81512456089237e-05, + "loss": 0.9518, + "step": 17926 + }, + { + "epoch": 1.32, + "learning_rate": 1.815101454180425e-05, + "loss": 0.9977, + "step": 17927 + }, + { + "epoch": 1.32, + "learning_rate": 1.815078346171665e-05, + "loss": 0.961, + "step": 17928 + }, + { + "epoch": 1.32, + "learning_rate": 1.8150552368661276e-05, + "loss": 1.035, + "step": 17929 + }, + { + "epoch": 1.32, + "learning_rate": 1.8150321262638485e-05, + "loss": 1.0974, + "step": 17930 + }, + { + "epoch": 1.32, + "learning_rate": 1.815009014364865e-05, + "loss": 1.0322, + "step": 17931 + }, + { + "epoch": 1.32, + "learning_rate": 1.8149859011692136e-05, + "loss": 0.9107, + "step": 17932 + }, + { + "epoch": 1.33, + "learning_rate": 1.8149627866769317e-05, + "loss": 0.9912, + "step": 17933 + }, + { + "epoch": 1.33, + "learning_rate": 1.8149396708880554e-05, + "loss": 1.0626, + "step": 17934 + }, + { + "epoch": 1.33, + "learning_rate": 1.8149165538026217e-05, + "loss": 1.0219, + "step": 17935 + }, + { + "epoch": 1.33, + "learning_rate": 1.8148934354206676e-05, + "loss": 1.0945, + "step": 17936 + }, + { + "epoch": 1.33, + "learning_rate": 1.81487031574223e-05, + "loss": 1.0894, + "step": 17937 + }, + { + "epoch": 1.33, + "learning_rate": 1.814847194767345e-05, + "loss": 0.9475, + "step": 17938 + }, + { + "epoch": 1.33, + "learning_rate": 1.8148240724960496e-05, + "loss": 1.0242, + "step": 17939 + }, + { + "epoch": 1.33, + "learning_rate": 1.8148009489283813e-05, + "loss": 1.0365, + "step": 17940 + }, + { + "epoch": 1.33, + "learning_rate": 1.814777824064376e-05, + "loss": 1.1178, + "step": 17941 + }, + { + "epoch": 1.33, + "learning_rate": 1.814754697904071e-05, + "loss": 0.9662, + "step": 17942 + }, + { + "epoch": 1.33, + "learning_rate": 1.8147315704475025e-05, + "loss": 0.9444, + "step": 17943 + }, + { + "epoch": 1.33, + "learning_rate": 1.814708441694708e-05, + "loss": 1.0839, + "step": 17944 + }, + { + "epoch": 1.33, + "learning_rate": 1.814685311645724e-05, + "loss": 1.0314, + "step": 17945 + }, + { + "epoch": 1.33, + "learning_rate": 1.8146621803005876e-05, + "loss": 1.1675, + "step": 17946 + }, + { + "epoch": 1.33, + "learning_rate": 1.8146390476593353e-05, + "loss": 0.9702, + "step": 17947 + }, + { + "epoch": 1.33, + "learning_rate": 1.8146159137220036e-05, + "loss": 1.0525, + "step": 17948 + }, + { + "epoch": 1.33, + "learning_rate": 1.8145927784886298e-05, + "loss": 1.115, + "step": 17949 + }, + { + "epoch": 1.33, + "learning_rate": 1.8145696419592506e-05, + "loss": 1.0865, + "step": 17950 + }, + { + "epoch": 1.33, + "learning_rate": 1.8145465041339028e-05, + "loss": 0.9818, + "step": 17951 + }, + { + "epoch": 1.33, + "learning_rate": 1.8145233650126228e-05, + "loss": 1.0078, + "step": 17952 + }, + { + "epoch": 1.33, + "learning_rate": 1.814500224595448e-05, + "loss": 0.9996, + "step": 17953 + }, + { + "epoch": 1.33, + "learning_rate": 1.814477082882415e-05, + "loss": 1.1205, + "step": 17954 + }, + { + "epoch": 1.33, + "learning_rate": 1.814453939873561e-05, + "loss": 1.0871, + "step": 17955 + }, + { + "epoch": 1.33, + "learning_rate": 1.8144307955689217e-05, + "loss": 1.1373, + "step": 17956 + }, + { + "epoch": 1.33, + "learning_rate": 1.814407649968535e-05, + "loss": 0.9534, + "step": 17957 + }, + { + "epoch": 1.33, + "learning_rate": 1.8143845030724376e-05, + "loss": 1.0666, + "step": 17958 + }, + { + "epoch": 1.33, + "learning_rate": 1.814361354880666e-05, + "loss": 1.0257, + "step": 17959 + }, + { + "epoch": 1.33, + "learning_rate": 1.814338205393257e-05, + "loss": 0.9764, + "step": 17960 + }, + { + "epoch": 1.33, + "learning_rate": 1.8143150546102476e-05, + "loss": 1.0411, + "step": 17961 + }, + { + "epoch": 1.33, + "learning_rate": 1.8142919025316748e-05, + "loss": 1.0152, + "step": 17962 + }, + { + "epoch": 1.33, + "learning_rate": 1.814268749157575e-05, + "loss": 1.0465, + "step": 17963 + }, + { + "epoch": 1.33, + "learning_rate": 1.8142455944879854e-05, + "loss": 0.9878, + "step": 17964 + }, + { + "epoch": 1.33, + "learning_rate": 1.8142224385229425e-05, + "loss": 0.9786, + "step": 17965 + }, + { + "epoch": 1.33, + "learning_rate": 1.8141992812624834e-05, + "loss": 0.9146, + "step": 17966 + }, + { + "epoch": 1.33, + "learning_rate": 1.8141761227066453e-05, + "loss": 1.0672, + "step": 17967 + }, + { + "epoch": 1.33, + "learning_rate": 1.814152962855464e-05, + "loss": 1.0069, + "step": 17968 + }, + { + "epoch": 1.33, + "learning_rate": 1.8141298017089774e-05, + "loss": 1.0936, + "step": 17969 + }, + { + "epoch": 1.33, + "learning_rate": 1.8141066392672218e-05, + "loss": 0.9312, + "step": 17970 + }, + { + "epoch": 1.33, + "learning_rate": 1.8140834755302342e-05, + "loss": 0.9928, + "step": 17971 + }, + { + "epoch": 1.33, + "learning_rate": 1.8140603104980513e-05, + "loss": 1.019, + "step": 17972 + }, + { + "epoch": 1.33, + "learning_rate": 1.8140371441707107e-05, + "loss": 1.012, + "step": 17973 + }, + { + "epoch": 1.33, + "learning_rate": 1.814013976548248e-05, + "loss": 1.0341, + "step": 17974 + }, + { + "epoch": 1.33, + "learning_rate": 1.813990807630701e-05, + "loss": 0.9953, + "step": 17975 + }, + { + "epoch": 1.33, + "learning_rate": 1.813967637418106e-05, + "loss": 0.9737, + "step": 17976 + }, + { + "epoch": 1.33, + "learning_rate": 1.8139444659105005e-05, + "loss": 0.8965, + "step": 17977 + }, + { + "epoch": 1.33, + "learning_rate": 1.813921293107921e-05, + "loss": 0.9466, + "step": 17978 + }, + { + "epoch": 1.33, + "learning_rate": 1.8138981190104038e-05, + "loss": 1.0725, + "step": 17979 + }, + { + "epoch": 1.33, + "learning_rate": 1.8138749436179867e-05, + "loss": 1.0266, + "step": 17980 + }, + { + "epoch": 1.33, + "learning_rate": 1.8138517669307064e-05, + "loss": 0.9814, + "step": 17981 + }, + { + "epoch": 1.33, + "learning_rate": 1.8138285889485992e-05, + "loss": 0.9304, + "step": 17982 + }, + { + "epoch": 1.33, + "learning_rate": 1.813805409671703e-05, + "loss": 1.104, + "step": 17983 + }, + { + "epoch": 1.33, + "learning_rate": 1.8137822291000535e-05, + "loss": 0.9439, + "step": 17984 + }, + { + "epoch": 1.33, + "learning_rate": 1.813759047233688e-05, + "loss": 1.0392, + "step": 17985 + }, + { + "epoch": 1.33, + "learning_rate": 1.813735864072644e-05, + "loss": 0.8699, + "step": 17986 + }, + { + "epoch": 1.33, + "learning_rate": 1.8137126796169575e-05, + "loss": 0.9762, + "step": 17987 + }, + { + "epoch": 1.33, + "learning_rate": 1.813689493866666e-05, + "loss": 0.964, + "step": 17988 + }, + { + "epoch": 1.33, + "learning_rate": 1.813666306821806e-05, + "loss": 1.0654, + "step": 17989 + }, + { + "epoch": 1.33, + "learning_rate": 1.8136431184824147e-05, + "loss": 1.0691, + "step": 17990 + }, + { + "epoch": 1.33, + "learning_rate": 1.8136199288485285e-05, + "loss": 0.9711, + "step": 17991 + }, + { + "epoch": 1.33, + "learning_rate": 1.8135967379201848e-05, + "loss": 1.0134, + "step": 17992 + }, + { + "epoch": 1.33, + "learning_rate": 1.8135735456974206e-05, + "loss": 1.0892, + "step": 17993 + }, + { + "epoch": 1.33, + "learning_rate": 1.8135503521802722e-05, + "loss": 1.1321, + "step": 17994 + }, + { + "epoch": 1.33, + "learning_rate": 1.8135271573687768e-05, + "loss": 1.0587, + "step": 17995 + }, + { + "epoch": 1.33, + "learning_rate": 1.8135039612629717e-05, + "loss": 0.9927, + "step": 17996 + }, + { + "epoch": 1.33, + "learning_rate": 1.8134807638628934e-05, + "loss": 0.9472, + "step": 17997 + }, + { + "epoch": 1.33, + "learning_rate": 1.8134575651685783e-05, + "loss": 1.1113, + "step": 17998 + }, + { + "epoch": 1.33, + "learning_rate": 1.8134343651800643e-05, + "loss": 1.0505, + "step": 17999 + }, + { + "epoch": 1.33, + "learning_rate": 1.8134111638973876e-05, + "loss": 1.0682, + "step": 18000 + }, + { + "epoch": 1.33, + "learning_rate": 1.813387961320586e-05, + "loss": 1.0956, + "step": 18001 + }, + { + "epoch": 1.33, + "learning_rate": 1.813364757449695e-05, + "loss": 0.9308, + "step": 18002 + }, + { + "epoch": 1.33, + "learning_rate": 1.813341552284753e-05, + "loss": 1.0232, + "step": 18003 + }, + { + "epoch": 1.33, + "learning_rate": 1.8133183458257958e-05, + "loss": 1.1745, + "step": 18004 + }, + { + "epoch": 1.33, + "learning_rate": 1.8132951380728608e-05, + "loss": 1.0657, + "step": 18005 + }, + { + "epoch": 1.33, + "learning_rate": 1.813271929025985e-05, + "loss": 0.9682, + "step": 18006 + }, + { + "epoch": 1.33, + "learning_rate": 1.813248718685205e-05, + "loss": 1.0401, + "step": 18007 + }, + { + "epoch": 1.33, + "learning_rate": 1.813225507050558e-05, + "loss": 0.996, + "step": 18008 + }, + { + "epoch": 1.33, + "learning_rate": 1.8132022941220807e-05, + "loss": 0.9767, + "step": 18009 + }, + { + "epoch": 1.33, + "learning_rate": 1.8131790798998103e-05, + "loss": 1.05, + "step": 18010 + }, + { + "epoch": 1.33, + "learning_rate": 1.813155864383784e-05, + "loss": 0.9652, + "step": 18011 + }, + { + "epoch": 1.33, + "learning_rate": 1.813132647574038e-05, + "loss": 1.0163, + "step": 18012 + }, + { + "epoch": 1.33, + "learning_rate": 1.8131094294706097e-05, + "loss": 1.0708, + "step": 18013 + }, + { + "epoch": 1.33, + "learning_rate": 1.8130862100735357e-05, + "loss": 0.9539, + "step": 18014 + }, + { + "epoch": 1.33, + "learning_rate": 1.8130629893828534e-05, + "loss": 1.0728, + "step": 18015 + }, + { + "epoch": 1.33, + "learning_rate": 1.8130397673985993e-05, + "loss": 0.9625, + "step": 18016 + }, + { + "epoch": 1.33, + "learning_rate": 1.813016544120811e-05, + "loss": 1.0532, + "step": 18017 + }, + { + "epoch": 1.33, + "learning_rate": 1.8129933195495246e-05, + "loss": 1.0721, + "step": 18018 + }, + { + "epoch": 1.33, + "learning_rate": 1.8129700936847774e-05, + "loss": 1.0249, + "step": 18019 + }, + { + "epoch": 1.33, + "learning_rate": 1.812946866526607e-05, + "loss": 1.0193, + "step": 18020 + }, + { + "epoch": 1.33, + "learning_rate": 1.812923638075049e-05, + "loss": 1.0759, + "step": 18021 + }, + { + "epoch": 1.33, + "learning_rate": 1.8129004083301414e-05, + "loss": 1.1195, + "step": 18022 + }, + { + "epoch": 1.33, + "learning_rate": 1.812877177291921e-05, + "loss": 0.973, + "step": 18023 + }, + { + "epoch": 1.33, + "learning_rate": 1.812853944960425e-05, + "loss": 1.1277, + "step": 18024 + }, + { + "epoch": 1.33, + "learning_rate": 1.8128307113356894e-05, + "loss": 1.0249, + "step": 18025 + }, + { + "epoch": 1.33, + "learning_rate": 1.8128074764177518e-05, + "loss": 0.957, + "step": 18026 + }, + { + "epoch": 1.33, + "learning_rate": 1.8127842402066497e-05, + "loss": 1.0787, + "step": 18027 + }, + { + "epoch": 1.33, + "learning_rate": 1.812761002702419e-05, + "loss": 0.9237, + "step": 18028 + }, + { + "epoch": 1.33, + "learning_rate": 1.8127377639050972e-05, + "loss": 1.055, + "step": 18029 + }, + { + "epoch": 1.33, + "learning_rate": 1.8127145238147214e-05, + "loss": 1.0578, + "step": 18030 + }, + { + "epoch": 1.33, + "learning_rate": 1.8126912824313284e-05, + "loss": 1.0101, + "step": 18031 + }, + { + "epoch": 1.33, + "learning_rate": 1.812668039754955e-05, + "loss": 1.0777, + "step": 18032 + }, + { + "epoch": 1.33, + "learning_rate": 1.8126447957856385e-05, + "loss": 1.0746, + "step": 18033 + }, + { + "epoch": 1.33, + "learning_rate": 1.812621550523416e-05, + "loss": 1.1109, + "step": 18034 + }, + { + "epoch": 1.33, + "learning_rate": 1.812598303968324e-05, + "loss": 1.0453, + "step": 18035 + }, + { + "epoch": 1.33, + "learning_rate": 1.8125750561203995e-05, + "loss": 1.0501, + "step": 18036 + }, + { + "epoch": 1.33, + "learning_rate": 1.8125518069796803e-05, + "loss": 0.9833, + "step": 18037 + }, + { + "epoch": 1.33, + "learning_rate": 1.812528556546202e-05, + "loss": 1.0082, + "step": 18038 + }, + { + "epoch": 1.33, + "learning_rate": 1.812505304820003e-05, + "loss": 0.9412, + "step": 18039 + }, + { + "epoch": 1.33, + "learning_rate": 1.8124820518011196e-05, + "loss": 1.0534, + "step": 18040 + }, + { + "epoch": 1.33, + "learning_rate": 1.8124587974895888e-05, + "loss": 0.9789, + "step": 18041 + }, + { + "epoch": 1.33, + "learning_rate": 1.8124355418854478e-05, + "loss": 1.1905, + "step": 18042 + }, + { + "epoch": 1.33, + "learning_rate": 1.8124122849887335e-05, + "loss": 1.0464, + "step": 18043 + }, + { + "epoch": 1.33, + "learning_rate": 1.8123890267994825e-05, + "loss": 1.099, + "step": 18044 + }, + { + "epoch": 1.33, + "learning_rate": 1.8123657673177327e-05, + "loss": 0.9292, + "step": 18045 + }, + { + "epoch": 1.33, + "learning_rate": 1.8123425065435204e-05, + "loss": 1.0145, + "step": 18046 + }, + { + "epoch": 1.33, + "learning_rate": 1.8123192444768826e-05, + "loss": 0.979, + "step": 18047 + }, + { + "epoch": 1.33, + "learning_rate": 1.8122959811178568e-05, + "loss": 1.0471, + "step": 18048 + }, + { + "epoch": 1.33, + "learning_rate": 1.8122727164664793e-05, + "loss": 0.9216, + "step": 18049 + }, + { + "epoch": 1.33, + "learning_rate": 1.8122494505227877e-05, + "loss": 0.9286, + "step": 18050 + }, + { + "epoch": 1.33, + "learning_rate": 1.8122261832868192e-05, + "loss": 0.9695, + "step": 18051 + }, + { + "epoch": 1.33, + "learning_rate": 1.81220291475861e-05, + "loss": 1.0747, + "step": 18052 + }, + { + "epoch": 1.33, + "learning_rate": 1.8121796449381975e-05, + "loss": 0.9624, + "step": 18053 + }, + { + "epoch": 1.33, + "learning_rate": 1.8121563738256195e-05, + "loss": 1.0096, + "step": 18054 + }, + { + "epoch": 1.33, + "learning_rate": 1.8121331014209117e-05, + "loss": 1.0234, + "step": 18055 + }, + { + "epoch": 1.33, + "learning_rate": 1.812109827724112e-05, + "loss": 0.9799, + "step": 18056 + }, + { + "epoch": 1.33, + "learning_rate": 1.812086552735257e-05, + "loss": 1.0384, + "step": 18057 + }, + { + "epoch": 1.33, + "learning_rate": 1.8120632764543842e-05, + "loss": 1.0463, + "step": 18058 + }, + { + "epoch": 1.33, + "learning_rate": 1.81203999888153e-05, + "loss": 1.0544, + "step": 18059 + }, + { + "epoch": 1.33, + "learning_rate": 1.812016720016732e-05, + "loss": 0.8945, + "step": 18060 + }, + { + "epoch": 1.33, + "learning_rate": 1.811993439860027e-05, + "loss": 1.1301, + "step": 18061 + }, + { + "epoch": 1.33, + "learning_rate": 1.811970158411452e-05, + "loss": 0.9905, + "step": 18062 + }, + { + "epoch": 1.33, + "learning_rate": 1.8119468756710443e-05, + "loss": 0.9598, + "step": 18063 + }, + { + "epoch": 1.33, + "learning_rate": 1.8119235916388405e-05, + "loss": 0.9404, + "step": 18064 + }, + { + "epoch": 1.33, + "learning_rate": 1.811900306314878e-05, + "loss": 0.97, + "step": 18065 + }, + { + "epoch": 1.33, + "learning_rate": 1.8118770196991937e-05, + "loss": 1.0617, + "step": 18066 + }, + { + "epoch": 1.33, + "learning_rate": 1.811853731791825e-05, + "loss": 1.0128, + "step": 18067 + }, + { + "epoch": 1.34, + "learning_rate": 1.8118304425928078e-05, + "loss": 1.1052, + "step": 18068 + }, + { + "epoch": 1.34, + "learning_rate": 1.8118071521021807e-05, + "loss": 1.0193, + "step": 18069 + }, + { + "epoch": 1.34, + "learning_rate": 1.8117838603199797e-05, + "loss": 1.046, + "step": 18070 + }, + { + "epoch": 1.34, + "learning_rate": 1.8117605672462423e-05, + "loss": 1.1527, + "step": 18071 + }, + { + "epoch": 1.34, + "learning_rate": 1.8117372728810055e-05, + "loss": 1.0564, + "step": 18072 + }, + { + "epoch": 1.34, + "learning_rate": 1.811713977224306e-05, + "loss": 1.0815, + "step": 18073 + }, + { + "epoch": 1.34, + "learning_rate": 1.8116906802761817e-05, + "loss": 1.0873, + "step": 18074 + }, + { + "epoch": 1.34, + "learning_rate": 1.811667382036669e-05, + "loss": 0.997, + "step": 18075 + }, + { + "epoch": 1.34, + "learning_rate": 1.8116440825058047e-05, + "loss": 1.0517, + "step": 18076 + }, + { + "epoch": 1.34, + "learning_rate": 1.811620781683627e-05, + "loss": 0.9511, + "step": 18077 + }, + { + "epoch": 1.34, + "learning_rate": 1.8115974795701717e-05, + "loss": 1.093, + "step": 18078 + }, + { + "epoch": 1.34, + "learning_rate": 1.8115741761654763e-05, + "loss": 1.0275, + "step": 18079 + }, + { + "epoch": 1.34, + "learning_rate": 1.8115508714695784e-05, + "loss": 1.0949, + "step": 18080 + }, + { + "epoch": 1.34, + "learning_rate": 1.8115275654825146e-05, + "loss": 1.0093, + "step": 18081 + }, + { + "epoch": 1.34, + "learning_rate": 1.811504258204322e-05, + "loss": 1.0639, + "step": 18082 + }, + { + "epoch": 1.34, + "learning_rate": 1.8114809496350376e-05, + "loss": 1.1619, + "step": 18083 + }, + { + "epoch": 1.34, + "learning_rate": 1.811457639774699e-05, + "loss": 1.0206, + "step": 18084 + }, + { + "epoch": 1.34, + "learning_rate": 1.8114343286233424e-05, + "loss": 1.0459, + "step": 18085 + }, + { + "epoch": 1.34, + "learning_rate": 1.8114110161810056e-05, + "loss": 0.9399, + "step": 18086 + }, + { + "epoch": 1.34, + "learning_rate": 1.8113877024477253e-05, + "loss": 1.0499, + "step": 18087 + }, + { + "epoch": 1.34, + "learning_rate": 1.811364387423539e-05, + "loss": 1.0295, + "step": 18088 + }, + { + "epoch": 1.34, + "learning_rate": 1.811341071108484e-05, + "loss": 1.0054, + "step": 18089 + }, + { + "epoch": 1.34, + "learning_rate": 1.8113177535025964e-05, + "loss": 0.9788, + "step": 18090 + }, + { + "epoch": 1.34, + "learning_rate": 1.8112944346059137e-05, + "loss": 0.9609, + "step": 18091 + }, + { + "epoch": 1.34, + "learning_rate": 1.8112711144184737e-05, + "loss": 1.0577, + "step": 18092 + }, + { + "epoch": 1.34, + "learning_rate": 1.8112477929403124e-05, + "loss": 1.0139, + "step": 18093 + }, + { + "epoch": 1.34, + "learning_rate": 1.811224470171468e-05, + "loss": 0.9922, + "step": 18094 + }, + { + "epoch": 1.34, + "learning_rate": 1.8112011461119768e-05, + "loss": 0.9968, + "step": 18095 + }, + { + "epoch": 1.34, + "learning_rate": 1.811177820761876e-05, + "loss": 0.9727, + "step": 18096 + }, + { + "epoch": 1.34, + "learning_rate": 1.8111544941212035e-05, + "loss": 1.0432, + "step": 18097 + }, + { + "epoch": 1.34, + "learning_rate": 1.8111311661899952e-05, + "loss": 1.0956, + "step": 18098 + }, + { + "epoch": 1.34, + "learning_rate": 1.811107836968289e-05, + "loss": 0.9634, + "step": 18099 + }, + { + "epoch": 1.34, + "learning_rate": 1.811084506456122e-05, + "loss": 1.0194, + "step": 18100 + }, + { + "epoch": 1.34, + "learning_rate": 1.811061174653531e-05, + "loss": 1.035, + "step": 18101 + }, + { + "epoch": 1.34, + "learning_rate": 1.8110378415605533e-05, + "loss": 1.0249, + "step": 18102 + }, + { + "epoch": 1.34, + "learning_rate": 1.811014507177226e-05, + "loss": 0.9856, + "step": 18103 + }, + { + "epoch": 1.34, + "learning_rate": 1.810991171503586e-05, + "loss": 0.9866, + "step": 18104 + }, + { + "epoch": 1.34, + "learning_rate": 1.8109678345396712e-05, + "loss": 1.0825, + "step": 18105 + }, + { + "epoch": 1.34, + "learning_rate": 1.810944496285518e-05, + "loss": 1.0167, + "step": 18106 + }, + { + "epoch": 1.34, + "learning_rate": 1.8109211567411633e-05, + "loss": 0.9908, + "step": 18107 + }, + { + "epoch": 1.34, + "learning_rate": 1.810897815906645e-05, + "loss": 0.9752, + "step": 18108 + }, + { + "epoch": 1.34, + "learning_rate": 1.8108744737819997e-05, + "loss": 1.0394, + "step": 18109 + }, + { + "epoch": 1.34, + "learning_rate": 1.810851130367265e-05, + "loss": 0.9998, + "step": 18110 + }, + { + "epoch": 1.34, + "learning_rate": 1.8108277856624777e-05, + "loss": 1.0741, + "step": 18111 + }, + { + "epoch": 1.34, + "learning_rate": 1.8108044396676748e-05, + "loss": 1.0069, + "step": 18112 + }, + { + "epoch": 1.34, + "learning_rate": 1.8107810923828936e-05, + "loss": 1.1441, + "step": 18113 + }, + { + "epoch": 1.34, + "learning_rate": 1.8107577438081713e-05, + "loss": 1.2432, + "step": 18114 + }, + { + "epoch": 1.34, + "learning_rate": 1.8107343939435453e-05, + "loss": 1.0374, + "step": 18115 + }, + { + "epoch": 1.34, + "learning_rate": 1.8107110427890523e-05, + "loss": 1.0639, + "step": 18116 + }, + { + "epoch": 1.34, + "learning_rate": 1.8106876903447295e-05, + "loss": 0.8749, + "step": 18117 + }, + { + "epoch": 1.34, + "learning_rate": 1.8106643366106144e-05, + "loss": 0.9437, + "step": 18118 + }, + { + "epoch": 1.34, + "learning_rate": 1.810640981586744e-05, + "loss": 1.0271, + "step": 18119 + }, + { + "epoch": 1.34, + "learning_rate": 1.810617625273155e-05, + "loss": 1.0234, + "step": 18120 + }, + { + "epoch": 1.34, + "learning_rate": 1.810594267669885e-05, + "loss": 0.9968, + "step": 18121 + }, + { + "epoch": 1.34, + "learning_rate": 1.8105709087769713e-05, + "loss": 0.99, + "step": 18122 + }, + { + "epoch": 1.34, + "learning_rate": 1.810547548594451e-05, + "loss": 1.0315, + "step": 18123 + }, + { + "epoch": 1.34, + "learning_rate": 1.810524187122361e-05, + "loss": 1.0186, + "step": 18124 + }, + { + "epoch": 1.34, + "learning_rate": 1.8105008243607386e-05, + "loss": 1.0169, + "step": 18125 + }, + { + "epoch": 1.34, + "learning_rate": 1.8104774603096212e-05, + "loss": 1.0709, + "step": 18126 + }, + { + "epoch": 1.34, + "learning_rate": 1.8104540949690456e-05, + "loss": 0.9795, + "step": 18127 + }, + { + "epoch": 1.34, + "learning_rate": 1.810430728339049e-05, + "loss": 1.042, + "step": 18128 + }, + { + "epoch": 1.34, + "learning_rate": 1.810407360419669e-05, + "loss": 0.9729, + "step": 18129 + }, + { + "epoch": 1.34, + "learning_rate": 1.8103839912109422e-05, + "loss": 0.9955, + "step": 18130 + }, + { + "epoch": 1.34, + "learning_rate": 1.8103606207129063e-05, + "loss": 1.0927, + "step": 18131 + }, + { + "epoch": 1.34, + "learning_rate": 1.8103372489255978e-05, + "loss": 0.9935, + "step": 18132 + }, + { + "epoch": 1.34, + "learning_rate": 1.8103138758490547e-05, + "loss": 0.9511, + "step": 18133 + }, + { + "epoch": 1.34, + "learning_rate": 1.810290501483314e-05, + "loss": 1.0329, + "step": 18134 + }, + { + "epoch": 1.34, + "learning_rate": 1.8102671258284123e-05, + "loss": 1.1078, + "step": 18135 + }, + { + "epoch": 1.34, + "learning_rate": 1.8102437488843872e-05, + "loss": 1.0103, + "step": 18136 + }, + { + "epoch": 1.34, + "learning_rate": 1.8102203706512762e-05, + "loss": 1.0754, + "step": 18137 + }, + { + "epoch": 1.34, + "learning_rate": 1.8101969911291163e-05, + "loss": 0.8937, + "step": 18138 + }, + { + "epoch": 1.34, + "learning_rate": 1.810173610317944e-05, + "loss": 0.9782, + "step": 18139 + }, + { + "epoch": 1.34, + "learning_rate": 1.8101502282177972e-05, + "loss": 0.9296, + "step": 18140 + }, + { + "epoch": 1.34, + "learning_rate": 1.8101268448287134e-05, + "loss": 1.202, + "step": 18141 + }, + { + "epoch": 1.34, + "learning_rate": 1.8101034601507293e-05, + "loss": 1.0611, + "step": 18142 + }, + { + "epoch": 1.34, + "learning_rate": 1.810080074183882e-05, + "loss": 0.9718, + "step": 18143 + }, + { + "epoch": 1.34, + "learning_rate": 1.810056686928209e-05, + "loss": 1.0132, + "step": 18144 + }, + { + "epoch": 1.34, + "learning_rate": 1.8100332983837473e-05, + "loss": 0.9836, + "step": 18145 + }, + { + "epoch": 1.34, + "learning_rate": 1.8100099085505344e-05, + "loss": 0.9946, + "step": 18146 + }, + { + "epoch": 1.34, + "learning_rate": 1.8099865174286072e-05, + "loss": 1.1021, + "step": 18147 + }, + { + "epoch": 1.34, + "learning_rate": 1.809963125018003e-05, + "loss": 0.8669, + "step": 18148 + }, + { + "epoch": 1.34, + "learning_rate": 1.809939731318759e-05, + "loss": 1.0267, + "step": 18149 + }, + { + "epoch": 1.34, + "learning_rate": 1.8099163363309124e-05, + "loss": 1.0725, + "step": 18150 + }, + { + "epoch": 1.34, + "learning_rate": 1.8098929400545008e-05, + "loss": 1.0085, + "step": 18151 + }, + { + "epoch": 1.34, + "learning_rate": 1.809869542489561e-05, + "loss": 0.9344, + "step": 18152 + }, + { + "epoch": 1.34, + "learning_rate": 1.80984614363613e-05, + "loss": 1.0831, + "step": 18153 + }, + { + "epoch": 1.34, + "learning_rate": 1.809822743494246e-05, + "loss": 1.0017, + "step": 18154 + }, + { + "epoch": 1.34, + "learning_rate": 1.809799342063945e-05, + "loss": 1.0771, + "step": 18155 + }, + { + "epoch": 1.34, + "learning_rate": 1.8097759393452655e-05, + "loss": 1.0147, + "step": 18156 + }, + { + "epoch": 1.34, + "learning_rate": 1.8097525353382436e-05, + "loss": 1.0168, + "step": 18157 + }, + { + "epoch": 1.34, + "learning_rate": 1.809729130042917e-05, + "loss": 0.9741, + "step": 18158 + }, + { + "epoch": 1.34, + "learning_rate": 1.809705723459323e-05, + "loss": 1.0774, + "step": 18159 + }, + { + "epoch": 1.34, + "learning_rate": 1.809682315587499e-05, + "loss": 1.0496, + "step": 18160 + }, + { + "epoch": 1.34, + "learning_rate": 1.8096589064274817e-05, + "loss": 1.0267, + "step": 18161 + }, + { + "epoch": 1.34, + "learning_rate": 1.8096354959793086e-05, + "loss": 1.0568, + "step": 18162 + }, + { + "epoch": 1.34, + "learning_rate": 1.8096120842430173e-05, + "loss": 0.9172, + "step": 18163 + }, + { + "epoch": 1.34, + "learning_rate": 1.8095886712186446e-05, + "loss": 1.0531, + "step": 18164 + }, + { + "epoch": 1.34, + "learning_rate": 1.809565256906228e-05, + "loss": 0.908, + "step": 18165 + }, + { + "epoch": 1.34, + "learning_rate": 1.8095418413058045e-05, + "loss": 0.9903, + "step": 18166 + }, + { + "epoch": 1.34, + "learning_rate": 1.809518424417412e-05, + "loss": 0.9619, + "step": 18167 + }, + { + "epoch": 1.34, + "learning_rate": 1.8094950062410867e-05, + "loss": 1.0754, + "step": 18168 + }, + { + "epoch": 1.34, + "learning_rate": 1.8094715867768665e-05, + "loss": 0.9384, + "step": 18169 + }, + { + "epoch": 1.34, + "learning_rate": 1.809448166024789e-05, + "loss": 0.9929, + "step": 18170 + }, + { + "epoch": 1.34, + "learning_rate": 1.8094247439848908e-05, + "loss": 0.9878, + "step": 18171 + }, + { + "epoch": 1.34, + "learning_rate": 1.8094013206572093e-05, + "loss": 0.974, + "step": 18172 + }, + { + "epoch": 1.34, + "learning_rate": 1.8093778960417817e-05, + "loss": 1.0922, + "step": 18173 + }, + { + "epoch": 1.34, + "learning_rate": 1.8093544701386458e-05, + "loss": 1.0314, + "step": 18174 + }, + { + "epoch": 1.34, + "learning_rate": 1.8093310429478385e-05, + "loss": 1.1375, + "step": 18175 + }, + { + "epoch": 1.34, + "learning_rate": 1.8093076144693972e-05, + "loss": 0.9585, + "step": 18176 + }, + { + "epoch": 1.34, + "learning_rate": 1.8092841847033587e-05, + "loss": 1.0499, + "step": 18177 + }, + { + "epoch": 1.34, + "learning_rate": 1.8092607536497607e-05, + "loss": 1.0303, + "step": 18178 + }, + { + "epoch": 1.34, + "learning_rate": 1.8092373213086407e-05, + "loss": 0.9798, + "step": 18179 + }, + { + "epoch": 1.34, + "learning_rate": 1.8092138876800355e-05, + "loss": 1.0168, + "step": 18180 + }, + { + "epoch": 1.34, + "learning_rate": 1.8091904527639828e-05, + "loss": 1.0382, + "step": 18181 + }, + { + "epoch": 1.34, + "learning_rate": 1.8091670165605193e-05, + "loss": 1.1396, + "step": 18182 + }, + { + "epoch": 1.34, + "learning_rate": 1.809143579069683e-05, + "loss": 0.9063, + "step": 18183 + }, + { + "epoch": 1.34, + "learning_rate": 1.8091201402915108e-05, + "loss": 0.9705, + "step": 18184 + }, + { + "epoch": 1.34, + "learning_rate": 1.80909670022604e-05, + "loss": 1.0987, + "step": 18185 + }, + { + "epoch": 1.34, + "learning_rate": 1.8090732588733076e-05, + "loss": 1.0519, + "step": 18186 + }, + { + "epoch": 1.34, + "learning_rate": 1.8090498162333516e-05, + "loss": 1.0724, + "step": 18187 + }, + { + "epoch": 1.34, + "learning_rate": 1.8090263723062083e-05, + "loss": 1.0479, + "step": 18188 + }, + { + "epoch": 1.34, + "learning_rate": 1.8090029270919162e-05, + "loss": 1.0277, + "step": 18189 + }, + { + "epoch": 1.34, + "learning_rate": 1.808979480590512e-05, + "loss": 0.9992, + "step": 18190 + }, + { + "epoch": 1.34, + "learning_rate": 1.808956032802033e-05, + "loss": 1.0353, + "step": 18191 + }, + { + "epoch": 1.34, + "learning_rate": 1.8089325837265165e-05, + "loss": 1.0261, + "step": 18192 + }, + { + "epoch": 1.34, + "learning_rate": 1.8089091333639998e-05, + "loss": 1.0177, + "step": 18193 + }, + { + "epoch": 1.34, + "learning_rate": 1.8088856817145202e-05, + "loss": 0.9502, + "step": 18194 + }, + { + "epoch": 1.34, + "learning_rate": 1.8088622287781153e-05, + "loss": 1.1223, + "step": 18195 + }, + { + "epoch": 1.34, + "learning_rate": 1.8088387745548218e-05, + "loss": 1.0546, + "step": 18196 + }, + { + "epoch": 1.34, + "learning_rate": 1.8088153190446775e-05, + "loss": 1.0159, + "step": 18197 + }, + { + "epoch": 1.34, + "learning_rate": 1.8087918622477197e-05, + "loss": 1.116, + "step": 18198 + }, + { + "epoch": 1.34, + "learning_rate": 1.8087684041639858e-05, + "loss": 0.9815, + "step": 18199 + }, + { + "epoch": 1.34, + "learning_rate": 1.8087449447935125e-05, + "loss": 1.0923, + "step": 18200 + }, + { + "epoch": 1.34, + "learning_rate": 1.808721484136338e-05, + "loss": 1.0094, + "step": 18201 + }, + { + "epoch": 1.34, + "learning_rate": 1.8086980221924987e-05, + "loss": 0.9982, + "step": 18202 + }, + { + "epoch": 1.35, + "learning_rate": 1.8086745589620326e-05, + "loss": 0.8408, + "step": 18203 + }, + { + "epoch": 1.35, + "learning_rate": 1.808651094444977e-05, + "loss": 0.9866, + "step": 18204 + }, + { + "epoch": 1.35, + "learning_rate": 1.8086276286413692e-05, + "loss": 1.0045, + "step": 18205 + }, + { + "epoch": 1.35, + "learning_rate": 1.808604161551246e-05, + "loss": 0.9675, + "step": 18206 + }, + { + "epoch": 1.35, + "learning_rate": 1.8085806931746456e-05, + "loss": 1.0083, + "step": 18207 + }, + { + "epoch": 1.35, + "learning_rate": 1.8085572235116043e-05, + "loss": 1.0647, + "step": 18208 + }, + { + "epoch": 1.35, + "learning_rate": 1.8085337525621605e-05, + "loss": 1.0602, + "step": 18209 + }, + { + "epoch": 1.35, + "learning_rate": 1.8085102803263512e-05, + "loss": 0.9783, + "step": 18210 + }, + { + "epoch": 1.35, + "learning_rate": 1.8084868068042132e-05, + "loss": 1.0198, + "step": 18211 + }, + { + "epoch": 1.35, + "learning_rate": 1.8084633319957846e-05, + "loss": 0.9981, + "step": 18212 + }, + { + "epoch": 1.35, + "learning_rate": 1.8084398559011022e-05, + "loss": 1.0613, + "step": 18213 + }, + { + "epoch": 1.35, + "learning_rate": 1.8084163785202037e-05, + "loss": 1.0698, + "step": 18214 + }, + { + "epoch": 1.35, + "learning_rate": 1.808392899853126e-05, + "loss": 1.1182, + "step": 18215 + }, + { + "epoch": 1.35, + "learning_rate": 1.808369419899907e-05, + "loss": 1.0568, + "step": 18216 + }, + { + "epoch": 1.35, + "learning_rate": 1.8083459386605836e-05, + "loss": 0.8915, + "step": 18217 + }, + { + "epoch": 1.35, + "learning_rate": 1.808322456135194e-05, + "loss": 1.0003, + "step": 18218 + }, + { + "epoch": 1.35, + "learning_rate": 1.808298972323774e-05, + "loss": 0.9705, + "step": 18219 + }, + { + "epoch": 1.35, + "learning_rate": 1.8082754872263626e-05, + "loss": 1.0426, + "step": 18220 + }, + { + "epoch": 1.35, + "learning_rate": 1.808252000842996e-05, + "loss": 1.0568, + "step": 18221 + }, + { + "epoch": 1.35, + "learning_rate": 1.8082285131737124e-05, + "loss": 0.9312, + "step": 18222 + }, + { + "epoch": 1.35, + "learning_rate": 1.8082050242185487e-05, + "loss": 1.0047, + "step": 18223 + }, + { + "epoch": 1.35, + "learning_rate": 1.808181533977542e-05, + "loss": 0.9915, + "step": 18224 + }, + { + "epoch": 1.35, + "learning_rate": 1.8081580424507307e-05, + "loss": 1.0098, + "step": 18225 + }, + { + "epoch": 1.35, + "learning_rate": 1.808134549638151e-05, + "loss": 1.0798, + "step": 18226 + }, + { + "epoch": 1.35, + "learning_rate": 1.8081110555398412e-05, + "loss": 0.9843, + "step": 18227 + }, + { + "epoch": 1.35, + "learning_rate": 1.808087560155838e-05, + "loss": 1.1857, + "step": 18228 + }, + { + "epoch": 1.35, + "learning_rate": 1.8080640634861792e-05, + "loss": 0.9936, + "step": 18229 + }, + { + "epoch": 1.35, + "learning_rate": 1.808040565530902e-05, + "loss": 1.0487, + "step": 18230 + }, + { + "epoch": 1.35, + "learning_rate": 1.8080170662900437e-05, + "loss": 1.0575, + "step": 18231 + }, + { + "epoch": 1.35, + "learning_rate": 1.807993565763642e-05, + "loss": 0.954, + "step": 18232 + }, + { + "epoch": 1.35, + "learning_rate": 1.8079700639517338e-05, + "loss": 1.037, + "step": 18233 + }, + { + "epoch": 1.35, + "learning_rate": 1.807946560854357e-05, + "loss": 0.9774, + "step": 18234 + }, + { + "epoch": 1.35, + "learning_rate": 1.8079230564715487e-05, + "loss": 0.9704, + "step": 18235 + }, + { + "epoch": 1.35, + "learning_rate": 1.8078995508033466e-05, + "loss": 1.0886, + "step": 18236 + }, + { + "epoch": 1.35, + "learning_rate": 1.807876043849788e-05, + "loss": 0.9661, + "step": 18237 + }, + { + "epoch": 1.35, + "learning_rate": 1.8078525356109095e-05, + "loss": 1.0104, + "step": 18238 + }, + { + "epoch": 1.35, + "learning_rate": 1.8078290260867498e-05, + "loss": 1.138, + "step": 18239 + }, + { + "epoch": 1.35, + "learning_rate": 1.8078055152773453e-05, + "loss": 1.1083, + "step": 18240 + }, + { + "epoch": 1.35, + "learning_rate": 1.8077820031827337e-05, + "loss": 0.9262, + "step": 18241 + }, + { + "epoch": 1.35, + "learning_rate": 1.807758489802953e-05, + "loss": 0.9162, + "step": 18242 + }, + { + "epoch": 1.35, + "learning_rate": 1.8077349751380398e-05, + "loss": 1.0493, + "step": 18243 + }, + { + "epoch": 1.35, + "learning_rate": 1.8077114591880316e-05, + "loss": 1.0921, + "step": 18244 + }, + { + "epoch": 1.35, + "learning_rate": 1.8076879419529665e-05, + "loss": 0.9983, + "step": 18245 + }, + { + "epoch": 1.35, + "learning_rate": 1.807664423432881e-05, + "loss": 0.8497, + "step": 18246 + }, + { + "epoch": 1.35, + "learning_rate": 1.8076409036278133e-05, + "loss": 0.9895, + "step": 18247 + }, + { + "epoch": 1.35, + "learning_rate": 1.8076173825378002e-05, + "loss": 1.0756, + "step": 18248 + }, + { + "epoch": 1.35, + "learning_rate": 1.8075938601628795e-05, + "loss": 1.0097, + "step": 18249 + }, + { + "epoch": 1.35, + "learning_rate": 1.8075703365030887e-05, + "loss": 1.0033, + "step": 18250 + }, + { + "epoch": 1.35, + "learning_rate": 1.8075468115584648e-05, + "loss": 1.1639, + "step": 18251 + }, + { + "epoch": 1.35, + "learning_rate": 1.8075232853290458e-05, + "loss": 1.0979, + "step": 18252 + }, + { + "epoch": 1.35, + "learning_rate": 1.807499757814868e-05, + "loss": 1.1183, + "step": 18253 + }, + { + "epoch": 1.35, + "learning_rate": 1.8074762290159706e-05, + "loss": 1.0922, + "step": 18254 + }, + { + "epoch": 1.35, + "learning_rate": 1.8074526989323893e-05, + "loss": 0.9927, + "step": 18255 + }, + { + "epoch": 1.35, + "learning_rate": 1.807429167564163e-05, + "loss": 1.1209, + "step": 18256 + }, + { + "epoch": 1.35, + "learning_rate": 1.8074056349113282e-05, + "loss": 0.9516, + "step": 18257 + }, + { + "epoch": 1.35, + "learning_rate": 1.8073821009739226e-05, + "loss": 1.0296, + "step": 18258 + }, + { + "epoch": 1.35, + "learning_rate": 1.8073585657519833e-05, + "loss": 1.002, + "step": 18259 + }, + { + "epoch": 1.35, + "learning_rate": 1.8073350292455484e-05, + "loss": 1.0505, + "step": 18260 + }, + { + "epoch": 1.35, + "learning_rate": 1.8073114914546552e-05, + "loss": 1.0805, + "step": 18261 + }, + { + "epoch": 1.35, + "learning_rate": 1.8072879523793404e-05, + "loss": 0.9569, + "step": 18262 + }, + { + "epoch": 1.35, + "learning_rate": 1.8072644120196426e-05, + "loss": 0.9892, + "step": 18263 + }, + { + "epoch": 1.35, + "learning_rate": 1.8072408703755984e-05, + "loss": 1.0637, + "step": 18264 + }, + { + "epoch": 1.35, + "learning_rate": 1.8072173274472457e-05, + "loss": 1.1236, + "step": 18265 + }, + { + "epoch": 1.35, + "learning_rate": 1.8071937832346213e-05, + "loss": 1.0315, + "step": 18266 + }, + { + "epoch": 1.35, + "learning_rate": 1.8071702377377634e-05, + "loss": 1.049, + "step": 18267 + }, + { + "epoch": 1.35, + "learning_rate": 1.807146690956709e-05, + "loss": 1.0586, + "step": 18268 + }, + { + "epoch": 1.35, + "learning_rate": 1.8071231428914962e-05, + "loss": 0.9972, + "step": 18269 + }, + { + "epoch": 1.35, + "learning_rate": 1.807099593542162e-05, + "loss": 1.0233, + "step": 18270 + }, + { + "epoch": 1.35, + "learning_rate": 1.8070760429087437e-05, + "loss": 1.0462, + "step": 18271 + }, + { + "epoch": 1.35, + "learning_rate": 1.807052490991279e-05, + "loss": 0.9781, + "step": 18272 + }, + { + "epoch": 1.35, + "learning_rate": 1.8070289377898053e-05, + "loss": 1.0226, + "step": 18273 + }, + { + "epoch": 1.35, + "learning_rate": 1.80700538330436e-05, + "loss": 1.0444, + "step": 18274 + }, + { + "epoch": 1.35, + "learning_rate": 1.806981827534981e-05, + "loss": 0.8982, + "step": 18275 + }, + { + "epoch": 1.35, + "learning_rate": 1.8069582704817055e-05, + "loss": 1.0178, + "step": 18276 + }, + { + "epoch": 1.35, + "learning_rate": 1.8069347121445705e-05, + "loss": 0.9035, + "step": 18277 + }, + { + "epoch": 1.35, + "learning_rate": 1.806911152523614e-05, + "loss": 1.001, + "step": 18278 + }, + { + "epoch": 1.35, + "learning_rate": 1.8068875916188736e-05, + "loss": 1.0085, + "step": 18279 + }, + { + "epoch": 1.35, + "learning_rate": 1.806864029430387e-05, + "loss": 1.0338, + "step": 18280 + }, + { + "epoch": 1.35, + "learning_rate": 1.8068404659581905e-05, + "loss": 1.0374, + "step": 18281 + }, + { + "epoch": 1.35, + "learning_rate": 1.8068169012023226e-05, + "loss": 0.9164, + "step": 18282 + }, + { + "epoch": 1.35, + "learning_rate": 1.8067933351628205e-05, + "loss": 0.968, + "step": 18283 + }, + { + "epoch": 1.35, + "learning_rate": 1.806769767839722e-05, + "loss": 0.8065, + "step": 18284 + }, + { + "epoch": 1.35, + "learning_rate": 1.8067461992330644e-05, + "loss": 1.0098, + "step": 18285 + }, + { + "epoch": 1.35, + "learning_rate": 1.806722629342885e-05, + "loss": 0.9357, + "step": 18286 + }, + { + "epoch": 1.35, + "learning_rate": 1.8066990581692214e-05, + "loss": 0.9834, + "step": 18287 + }, + { + "epoch": 1.35, + "learning_rate": 1.806675485712111e-05, + "loss": 1.0311, + "step": 18288 + }, + { + "epoch": 1.35, + "learning_rate": 1.8066519119715916e-05, + "loss": 1.0777, + "step": 18289 + }, + { + "epoch": 1.35, + "learning_rate": 1.8066283369477005e-05, + "loss": 1.0519, + "step": 18290 + }, + { + "epoch": 1.35, + "learning_rate": 1.8066047606404753e-05, + "loss": 0.9425, + "step": 18291 + }, + { + "epoch": 1.35, + "learning_rate": 1.8065811830499534e-05, + "loss": 0.9654, + "step": 18292 + }, + { + "epoch": 1.35, + "learning_rate": 1.8065576041761722e-05, + "loss": 1.0151, + "step": 18293 + }, + { + "epoch": 1.35, + "learning_rate": 1.8065340240191697e-05, + "loss": 0.9376, + "step": 18294 + }, + { + "epoch": 1.35, + "learning_rate": 1.806510442578983e-05, + "loss": 1.0412, + "step": 18295 + }, + { + "epoch": 1.35, + "learning_rate": 1.80648685985565e-05, + "loss": 1.072, + "step": 18296 + }, + { + "epoch": 1.35, + "learning_rate": 1.8064632758492076e-05, + "loss": 1.1135, + "step": 18297 + }, + { + "epoch": 1.35, + "learning_rate": 1.8064396905596936e-05, + "loss": 1.0867, + "step": 18298 + }, + { + "epoch": 1.35, + "learning_rate": 1.8064161039871456e-05, + "loss": 1.0849, + "step": 18299 + }, + { + "epoch": 1.35, + "learning_rate": 1.8063925161316015e-05, + "loss": 0.9755, + "step": 18300 + }, + { + "epoch": 1.35, + "learning_rate": 1.806368926993098e-05, + "loss": 1.1187, + "step": 18301 + }, + { + "epoch": 1.35, + "learning_rate": 1.8063453365716733e-05, + "loss": 1.0751, + "step": 18302 + }, + { + "epoch": 1.35, + "learning_rate": 1.806321744867365e-05, + "loss": 1.0646, + "step": 18303 + }, + { + "epoch": 1.35, + "learning_rate": 1.8062981518802098e-05, + "loss": 0.9866, + "step": 18304 + }, + { + "epoch": 1.35, + "learning_rate": 1.806274557610246e-05, + "loss": 1.0079, + "step": 18305 + }, + { + "epoch": 1.35, + "learning_rate": 1.806250962057511e-05, + "loss": 1.1487, + "step": 18306 + }, + { + "epoch": 1.35, + "learning_rate": 1.8062273652220422e-05, + "loss": 0.8758, + "step": 18307 + }, + { + "epoch": 1.35, + "learning_rate": 1.806203767103877e-05, + "loss": 0.9131, + "step": 18308 + }, + { + "epoch": 1.35, + "learning_rate": 1.8061801677030534e-05, + "loss": 0.9945, + "step": 18309 + }, + { + "epoch": 1.35, + "learning_rate": 1.8061565670196087e-05, + "loss": 1.0366, + "step": 18310 + }, + { + "epoch": 1.35, + "learning_rate": 1.8061329650535803e-05, + "loss": 1.0308, + "step": 18311 + }, + { + "epoch": 1.35, + "learning_rate": 1.8061093618050058e-05, + "loss": 1.1188, + "step": 18312 + }, + { + "epoch": 1.35, + "learning_rate": 1.806085757273923e-05, + "loss": 0.8886, + "step": 18313 + }, + { + "epoch": 1.35, + "learning_rate": 1.8060621514603695e-05, + "loss": 1.0227, + "step": 18314 + }, + { + "epoch": 1.35, + "learning_rate": 1.8060385443643823e-05, + "loss": 0.9141, + "step": 18315 + }, + { + "epoch": 1.35, + "learning_rate": 1.8060149359859994e-05, + "loss": 0.9019, + "step": 18316 + }, + { + "epoch": 1.35, + "learning_rate": 1.8059913263252584e-05, + "loss": 1.0212, + "step": 18317 + }, + { + "epoch": 1.35, + "learning_rate": 1.8059677153821966e-05, + "loss": 0.9269, + "step": 18318 + }, + { + "epoch": 1.35, + "learning_rate": 1.8059441031568516e-05, + "loss": 1.0344, + "step": 18319 + }, + { + "epoch": 1.35, + "learning_rate": 1.8059204896492613e-05, + "loss": 0.9742, + "step": 18320 + }, + { + "epoch": 1.35, + "learning_rate": 1.805896874859463e-05, + "loss": 1.1036, + "step": 18321 + }, + { + "epoch": 1.35, + "learning_rate": 1.8058732587874938e-05, + "loss": 1.0261, + "step": 18322 + }, + { + "epoch": 1.35, + "learning_rate": 1.8058496414333924e-05, + "loss": 0.9974, + "step": 18323 + }, + { + "epoch": 1.35, + "learning_rate": 1.8058260227971955e-05, + "loss": 0.9042, + "step": 18324 + }, + { + "epoch": 1.35, + "learning_rate": 1.8058024028789408e-05, + "loss": 0.8561, + "step": 18325 + }, + { + "epoch": 1.35, + "learning_rate": 1.8057787816786664e-05, + "loss": 0.9467, + "step": 18326 + }, + { + "epoch": 1.35, + "learning_rate": 1.805755159196409e-05, + "loss": 1.0168, + "step": 18327 + }, + { + "epoch": 1.35, + "learning_rate": 1.8057315354322073e-05, + "loss": 0.9041, + "step": 18328 + }, + { + "epoch": 1.35, + "learning_rate": 1.8057079103860977e-05, + "loss": 1.1131, + "step": 18329 + }, + { + "epoch": 1.35, + "learning_rate": 1.8056842840581183e-05, + "loss": 0.926, + "step": 18330 + }, + { + "epoch": 1.35, + "learning_rate": 1.8056606564483074e-05, + "loss": 1.1587, + "step": 18331 + }, + { + "epoch": 1.35, + "learning_rate": 1.8056370275567014e-05, + "loss": 1.0145, + "step": 18332 + }, + { + "epoch": 1.35, + "learning_rate": 1.8056133973833385e-05, + "loss": 1.0209, + "step": 18333 + }, + { + "epoch": 1.35, + "learning_rate": 1.805589765928256e-05, + "loss": 0.9584, + "step": 18334 + }, + { + "epoch": 1.35, + "learning_rate": 1.805566133191492e-05, + "loss": 1.1332, + "step": 18335 + }, + { + "epoch": 1.35, + "learning_rate": 1.8055424991730836e-05, + "loss": 1.0972, + "step": 18336 + }, + { + "epoch": 1.35, + "learning_rate": 1.8055188638730686e-05, + "loss": 0.986, + "step": 18337 + }, + { + "epoch": 1.35, + "learning_rate": 1.805495227291485e-05, + "loss": 0.9665, + "step": 18338 + }, + { + "epoch": 1.36, + "learning_rate": 1.8054715894283697e-05, + "loss": 0.9898, + "step": 18339 + }, + { + "epoch": 1.36, + "learning_rate": 1.8054479502837606e-05, + "loss": 1.0105, + "step": 18340 + }, + { + "epoch": 1.36, + "learning_rate": 1.8054243098576954e-05, + "loss": 0.9873, + "step": 18341 + }, + { + "epoch": 1.36, + "learning_rate": 1.805400668150212e-05, + "loss": 0.9829, + "step": 18342 + }, + { + "epoch": 1.36, + "learning_rate": 1.805377025161347e-05, + "loss": 1.1131, + "step": 18343 + }, + { + "epoch": 1.36, + "learning_rate": 1.805353380891139e-05, + "loss": 1.0344, + "step": 18344 + }, + { + "epoch": 1.36, + "learning_rate": 1.8053297353396254e-05, + "loss": 0.9918, + "step": 18345 + }, + { + "epoch": 1.36, + "learning_rate": 1.8053060885068435e-05, + "loss": 1.0934, + "step": 18346 + }, + { + "epoch": 1.36, + "learning_rate": 1.805282440392831e-05, + "loss": 1.092, + "step": 18347 + }, + { + "epoch": 1.36, + "learning_rate": 1.8052587909976265e-05, + "loss": 1.0341, + "step": 18348 + }, + { + "epoch": 1.36, + "learning_rate": 1.805235140321266e-05, + "loss": 1.0426, + "step": 18349 + }, + { + "epoch": 1.36, + "learning_rate": 1.805211488363788e-05, + "loss": 1.069, + "step": 18350 + }, + { + "epoch": 1.36, + "learning_rate": 1.80518783512523e-05, + "loss": 1.0113, + "step": 18351 + }, + { + "epoch": 1.36, + "learning_rate": 1.8051641806056303e-05, + "loss": 1.1871, + "step": 18352 + }, + { + "epoch": 1.36, + "learning_rate": 1.805140524805025e-05, + "loss": 1.1057, + "step": 18353 + }, + { + "epoch": 1.36, + "learning_rate": 1.805116867723453e-05, + "loss": 1.0486, + "step": 18354 + }, + { + "epoch": 1.36, + "learning_rate": 1.805093209360952e-05, + "loss": 0.99, + "step": 18355 + }, + { + "epoch": 1.36, + "learning_rate": 1.8050695497175586e-05, + "loss": 0.9879, + "step": 18356 + }, + { + "epoch": 1.36, + "learning_rate": 1.8050458887933117e-05, + "loss": 1.0707, + "step": 18357 + }, + { + "epoch": 1.36, + "learning_rate": 1.805022226588248e-05, + "loss": 1.0823, + "step": 18358 + }, + { + "epoch": 1.36, + "learning_rate": 1.8049985631024052e-05, + "loss": 1.0033, + "step": 18359 + }, + { + "epoch": 1.36, + "learning_rate": 1.8049748983358217e-05, + "loss": 0.906, + "step": 18360 + }, + { + "epoch": 1.36, + "learning_rate": 1.804951232288534e-05, + "loss": 1.0457, + "step": 18361 + }, + { + "epoch": 1.36, + "learning_rate": 1.804927564960581e-05, + "loss": 0.8952, + "step": 18362 + }, + { + "epoch": 1.36, + "learning_rate": 1.8049038963519995e-05, + "loss": 0.9377, + "step": 18363 + }, + { + "epoch": 1.36, + "learning_rate": 1.8048802264628273e-05, + "loss": 1.0075, + "step": 18364 + }, + { + "epoch": 1.36, + "learning_rate": 1.8048565552931022e-05, + "loss": 1.0054, + "step": 18365 + }, + { + "epoch": 1.36, + "learning_rate": 1.8048328828428622e-05, + "loss": 0.8821, + "step": 18366 + }, + { + "epoch": 1.36, + "learning_rate": 1.8048092091121442e-05, + "loss": 1.1629, + "step": 18367 + }, + { + "epoch": 1.36, + "learning_rate": 1.8047855341009865e-05, + "loss": 0.982, + "step": 18368 + }, + { + "epoch": 1.36, + "learning_rate": 1.8047618578094263e-05, + "loss": 1.0892, + "step": 18369 + }, + { + "epoch": 1.36, + "learning_rate": 1.804738180237502e-05, + "loss": 1.1034, + "step": 18370 + }, + { + "epoch": 1.36, + "learning_rate": 1.8047145013852504e-05, + "loss": 1.0609, + "step": 18371 + }, + { + "epoch": 1.36, + "learning_rate": 1.8046908212527096e-05, + "loss": 0.9719, + "step": 18372 + }, + { + "epoch": 1.36, + "learning_rate": 1.804667139839917e-05, + "loss": 1.0708, + "step": 18373 + }, + { + "epoch": 1.36, + "learning_rate": 1.8046434571469108e-05, + "loss": 1.0664, + "step": 18374 + }, + { + "epoch": 1.36, + "learning_rate": 1.804619773173728e-05, + "loss": 0.9229, + "step": 18375 + }, + { + "epoch": 1.36, + "learning_rate": 1.8045960879204072e-05, + "loss": 1.0408, + "step": 18376 + }, + { + "epoch": 1.36, + "learning_rate": 1.804572401386985e-05, + "loss": 1.0834, + "step": 18377 + }, + { + "epoch": 1.36, + "learning_rate": 1.8045487135735e-05, + "loss": 1.0525, + "step": 18378 + }, + { + "epoch": 1.36, + "learning_rate": 1.8045250244799892e-05, + "loss": 0.9689, + "step": 18379 + }, + { + "epoch": 1.36, + "learning_rate": 1.8045013341064907e-05, + "loss": 0.9922, + "step": 18380 + }, + { + "epoch": 1.36, + "learning_rate": 1.8044776424530424e-05, + "loss": 0.9819, + "step": 18381 + }, + { + "epoch": 1.36, + "learning_rate": 1.8044539495196813e-05, + "loss": 0.964, + "step": 18382 + }, + { + "epoch": 1.36, + "learning_rate": 1.8044302553064455e-05, + "loss": 1.1254, + "step": 18383 + }, + { + "epoch": 1.36, + "learning_rate": 1.804406559813373e-05, + "loss": 1.1219, + "step": 18384 + }, + { + "epoch": 1.36, + "learning_rate": 1.8043828630405008e-05, + "loss": 1.072, + "step": 18385 + }, + { + "epoch": 1.36, + "learning_rate": 1.8043591649878673e-05, + "loss": 0.9024, + "step": 18386 + }, + { + "epoch": 1.36, + "learning_rate": 1.8043354656555097e-05, + "loss": 1.036, + "step": 18387 + }, + { + "epoch": 1.36, + "learning_rate": 1.804311765043466e-05, + "loss": 0.8336, + "step": 18388 + }, + { + "epoch": 1.36, + "learning_rate": 1.8042880631517735e-05, + "loss": 0.978, + "step": 18389 + }, + { + "epoch": 1.36, + "learning_rate": 1.8042643599804704e-05, + "loss": 0.9843, + "step": 18390 + }, + { + "epoch": 1.36, + "learning_rate": 1.804240655529594e-05, + "loss": 1.0774, + "step": 18391 + }, + { + "epoch": 1.36, + "learning_rate": 1.8042169497991825e-05, + "loss": 1.03, + "step": 18392 + }, + { + "epoch": 1.36, + "learning_rate": 1.8041932427892735e-05, + "loss": 1.0039, + "step": 18393 + }, + { + "epoch": 1.36, + "learning_rate": 1.804169534499904e-05, + "loss": 1.0753, + "step": 18394 + }, + { + "epoch": 1.36, + "learning_rate": 1.8041458249311126e-05, + "loss": 1.0714, + "step": 18395 + }, + { + "epoch": 1.36, + "learning_rate": 1.8041221140829368e-05, + "loss": 0.9813, + "step": 18396 + }, + { + "epoch": 1.36, + "learning_rate": 1.804098401955414e-05, + "loss": 1.0252, + "step": 18397 + }, + { + "epoch": 1.36, + "learning_rate": 1.804074688548582e-05, + "loss": 1.0384, + "step": 18398 + }, + { + "epoch": 1.36, + "learning_rate": 1.804050973862479e-05, + "loss": 0.9668, + "step": 18399 + }, + { + "epoch": 1.36, + "learning_rate": 1.8040272578971425e-05, + "loss": 0.9755, + "step": 18400 + }, + { + "epoch": 1.36, + "learning_rate": 1.8040035406526103e-05, + "loss": 0.8435, + "step": 18401 + }, + { + "epoch": 1.36, + "learning_rate": 1.8039798221289195e-05, + "loss": 1.0114, + "step": 18402 + }, + { + "epoch": 1.36, + "learning_rate": 1.8039561023261082e-05, + "loss": 1.0161, + "step": 18403 + }, + { + "epoch": 1.36, + "learning_rate": 1.8039323812442146e-05, + "loss": 0.9227, + "step": 18404 + }, + { + "epoch": 1.36, + "learning_rate": 1.8039086588832758e-05, + "loss": 0.954, + "step": 18405 + }, + { + "epoch": 1.36, + "learning_rate": 1.80388493524333e-05, + "loss": 0.8899, + "step": 18406 + }, + { + "epoch": 1.36, + "learning_rate": 1.8038612103244148e-05, + "loss": 1.112, + "step": 18407 + }, + { + "epoch": 1.36, + "learning_rate": 1.803837484126568e-05, + "loss": 1.1275, + "step": 18408 + }, + { + "epoch": 1.36, + "learning_rate": 1.803813756649827e-05, + "loss": 1.0815, + "step": 18409 + }, + { + "epoch": 1.36, + "learning_rate": 1.80379002789423e-05, + "loss": 1.0761, + "step": 18410 + }, + { + "epoch": 1.36, + "learning_rate": 1.8037662978598148e-05, + "loss": 1.0359, + "step": 18411 + }, + { + "epoch": 1.36, + "learning_rate": 1.8037425665466183e-05, + "loss": 1.0502, + "step": 18412 + }, + { + "epoch": 1.36, + "learning_rate": 1.8037188339546795e-05, + "loss": 0.9868, + "step": 18413 + }, + { + "epoch": 1.36, + "learning_rate": 1.803695100084035e-05, + "loss": 0.9926, + "step": 18414 + }, + { + "epoch": 1.36, + "learning_rate": 1.8036713649347236e-05, + "loss": 0.9459, + "step": 18415 + }, + { + "epoch": 1.36, + "learning_rate": 1.803647628506782e-05, + "loss": 0.9793, + "step": 18416 + }, + { + "epoch": 1.36, + "learning_rate": 1.803623890800249e-05, + "loss": 0.8916, + "step": 18417 + }, + { + "epoch": 1.36, + "learning_rate": 1.8036001518151617e-05, + "loss": 1.0809, + "step": 18418 + }, + { + "epoch": 1.36, + "learning_rate": 1.803576411551558e-05, + "loss": 1.0402, + "step": 18419 + }, + { + "epoch": 1.36, + "learning_rate": 1.803552670009476e-05, + "loss": 1.0358, + "step": 18420 + }, + { + "epoch": 1.36, + "learning_rate": 1.803528927188953e-05, + "loss": 1.0511, + "step": 18421 + }, + { + "epoch": 1.36, + "learning_rate": 1.803505183090027e-05, + "loss": 0.9601, + "step": 18422 + }, + { + "epoch": 1.36, + "learning_rate": 1.8034814377127356e-05, + "loss": 1.0004, + "step": 18423 + }, + { + "epoch": 1.36, + "learning_rate": 1.8034576910571168e-05, + "loss": 0.9812, + "step": 18424 + }, + { + "epoch": 1.36, + "learning_rate": 1.8034339431232088e-05, + "loss": 1.045, + "step": 18425 + }, + { + "epoch": 1.36, + "learning_rate": 1.8034101939110484e-05, + "loss": 0.9351, + "step": 18426 + }, + { + "epoch": 1.36, + "learning_rate": 1.8033864434206736e-05, + "loss": 1.0678, + "step": 18427 + }, + { + "epoch": 1.36, + "learning_rate": 1.8033626916521227e-05, + "loss": 0.9836, + "step": 18428 + }, + { + "epoch": 1.36, + "learning_rate": 1.8033389386054337e-05, + "loss": 1.0034, + "step": 18429 + }, + { + "epoch": 1.36, + "learning_rate": 1.8033151842806434e-05, + "loss": 0.9528, + "step": 18430 + }, + { + "epoch": 1.36, + "learning_rate": 1.80329142867779e-05, + "loss": 0.9818, + "step": 18431 + }, + { + "epoch": 1.36, + "learning_rate": 1.803267671796912e-05, + "loss": 1.0166, + "step": 18432 + }, + { + "epoch": 1.36, + "learning_rate": 1.8032439136380463e-05, + "loss": 0.9798, + "step": 18433 + }, + { + "epoch": 1.36, + "learning_rate": 1.803220154201231e-05, + "loss": 1.0273, + "step": 18434 + }, + { + "epoch": 1.36, + "learning_rate": 1.803196393486504e-05, + "loss": 0.9855, + "step": 18435 + }, + { + "epoch": 1.36, + "learning_rate": 1.8031726314939033e-05, + "loss": 1.0826, + "step": 18436 + }, + { + "epoch": 1.36, + "learning_rate": 1.803148868223466e-05, + "loss": 0.9398, + "step": 18437 + }, + { + "epoch": 1.36, + "learning_rate": 1.8031251036752306e-05, + "loss": 0.9266, + "step": 18438 + }, + { + "epoch": 1.36, + "learning_rate": 1.803101337849235e-05, + "loss": 0.9944, + "step": 18439 + }, + { + "epoch": 1.36, + "learning_rate": 1.803077570745516e-05, + "loss": 1.0477, + "step": 18440 + }, + { + "epoch": 1.36, + "learning_rate": 1.8030538023641123e-05, + "loss": 0.9372, + "step": 18441 + }, + { + "epoch": 1.36, + "learning_rate": 1.8030300327050614e-05, + "loss": 1.0882, + "step": 18442 + }, + { + "epoch": 1.36, + "learning_rate": 1.8030062617684013e-05, + "loss": 1.0495, + "step": 18443 + }, + { + "epoch": 1.36, + "learning_rate": 1.8029824895541698e-05, + "loss": 1.0687, + "step": 18444 + }, + { + "epoch": 1.36, + "learning_rate": 1.8029587160624046e-05, + "loss": 0.9349, + "step": 18445 + }, + { + "epoch": 1.36, + "learning_rate": 1.8029349412931436e-05, + "loss": 0.9647, + "step": 18446 + }, + { + "epoch": 1.36, + "learning_rate": 1.8029111652464245e-05, + "loss": 0.9552, + "step": 18447 + }, + { + "epoch": 1.36, + "learning_rate": 1.802887387922285e-05, + "loss": 0.9462, + "step": 18448 + }, + { + "epoch": 1.36, + "learning_rate": 1.8028636093207635e-05, + "loss": 0.977, + "step": 18449 + }, + { + "epoch": 1.36, + "learning_rate": 1.8028398294418976e-05, + "loss": 0.9344, + "step": 18450 + }, + { + "epoch": 1.36, + "learning_rate": 1.802816048285725e-05, + "loss": 0.9278, + "step": 18451 + }, + { + "epoch": 1.36, + "learning_rate": 1.8027922658522835e-05, + "loss": 0.9534, + "step": 18452 + }, + { + "epoch": 1.36, + "learning_rate": 1.8027684821416106e-05, + "loss": 0.8648, + "step": 18453 + }, + { + "epoch": 1.36, + "learning_rate": 1.802744697153745e-05, + "loss": 0.8833, + "step": 18454 + }, + { + "epoch": 1.36, + "learning_rate": 1.8027209108887238e-05, + "loss": 0.9283, + "step": 18455 + }, + { + "epoch": 1.36, + "learning_rate": 1.8026971233465854e-05, + "loss": 1.0557, + "step": 18456 + }, + { + "epoch": 1.36, + "learning_rate": 1.802673334527367e-05, + "loss": 1.0056, + "step": 18457 + }, + { + "epoch": 1.36, + "learning_rate": 1.802649544431107e-05, + "loss": 1.1026, + "step": 18458 + }, + { + "epoch": 1.36, + "learning_rate": 1.802625753057843e-05, + "loss": 1.0773, + "step": 18459 + }, + { + "epoch": 1.36, + "learning_rate": 1.802601960407613e-05, + "loss": 0.9477, + "step": 18460 + }, + { + "epoch": 1.36, + "learning_rate": 1.8025781664804546e-05, + "loss": 1.0652, + "step": 18461 + }, + { + "epoch": 1.36, + "learning_rate": 1.802554371276406e-05, + "loss": 1.0566, + "step": 18462 + }, + { + "epoch": 1.36, + "learning_rate": 1.8025305747955048e-05, + "loss": 0.8645, + "step": 18463 + }, + { + "epoch": 1.36, + "learning_rate": 1.802506777037789e-05, + "loss": 1.0917, + "step": 18464 + }, + { + "epoch": 1.36, + "learning_rate": 1.8024829780032963e-05, + "loss": 1.0244, + "step": 18465 + }, + { + "epoch": 1.36, + "learning_rate": 1.8024591776920648e-05, + "loss": 1.0262, + "step": 18466 + }, + { + "epoch": 1.36, + "learning_rate": 1.802435376104132e-05, + "loss": 1.0376, + "step": 18467 + }, + { + "epoch": 1.36, + "learning_rate": 1.8024115732395363e-05, + "loss": 1.0594, + "step": 18468 + }, + { + "epoch": 1.36, + "learning_rate": 1.8023877690983148e-05, + "loss": 0.9545, + "step": 18469 + }, + { + "epoch": 1.36, + "learning_rate": 1.8023639636805063e-05, + "loss": 1.0582, + "step": 18470 + }, + { + "epoch": 1.36, + "learning_rate": 1.8023401569861484e-05, + "loss": 1.1072, + "step": 18471 + }, + { + "epoch": 1.36, + "learning_rate": 1.8023163490152782e-05, + "loss": 0.9533, + "step": 18472 + }, + { + "epoch": 1.36, + "learning_rate": 1.8022925397679345e-05, + "loss": 0.9796, + "step": 18473 + }, + { + "epoch": 1.37, + "learning_rate": 1.8022687292441548e-05, + "loss": 0.9914, + "step": 18474 + }, + { + "epoch": 1.37, + "learning_rate": 1.802244917443977e-05, + "loss": 0.9432, + "step": 18475 + }, + { + "epoch": 1.37, + "learning_rate": 1.802221104367439e-05, + "loss": 1.0047, + "step": 18476 + }, + { + "epoch": 1.37, + "learning_rate": 1.8021972900145787e-05, + "loss": 0.912, + "step": 18477 + }, + { + "epoch": 1.37, + "learning_rate": 1.802173474385434e-05, + "loss": 0.9508, + "step": 18478 + }, + { + "epoch": 1.37, + "learning_rate": 1.8021496574800426e-05, + "loss": 0.9979, + "step": 18479 + }, + { + "epoch": 1.37, + "learning_rate": 1.8021258392984427e-05, + "loss": 1.122, + "step": 18480 + }, + { + "epoch": 1.37, + "learning_rate": 1.802102019840672e-05, + "loss": 1.0513, + "step": 18481 + }, + { + "epoch": 1.37, + "learning_rate": 1.8020781991067688e-05, + "loss": 1.0296, + "step": 18482 + }, + { + "epoch": 1.37, + "learning_rate": 1.8020543770967704e-05, + "loss": 1.103, + "step": 18483 + }, + { + "epoch": 1.37, + "learning_rate": 1.802030553810715e-05, + "loss": 0.8831, + "step": 18484 + }, + { + "epoch": 1.37, + "learning_rate": 1.8020067292486406e-05, + "loss": 1.0726, + "step": 18485 + }, + { + "epoch": 1.37, + "learning_rate": 1.8019829034105845e-05, + "loss": 0.9976, + "step": 18486 + }, + { + "epoch": 1.37, + "learning_rate": 1.8019590762965858e-05, + "loss": 1.0135, + "step": 18487 + }, + { + "epoch": 1.37, + "learning_rate": 1.8019352479066813e-05, + "loss": 0.9401, + "step": 18488 + }, + { + "epoch": 1.37, + "learning_rate": 1.801911418240909e-05, + "loss": 1.0876, + "step": 18489 + }, + { + "epoch": 1.37, + "learning_rate": 1.8018875872993077e-05, + "loss": 0.9514, + "step": 18490 + }, + { + "epoch": 1.37, + "learning_rate": 1.801863755081914e-05, + "loss": 0.9469, + "step": 18491 + }, + { + "epoch": 1.37, + "learning_rate": 1.8018399215887672e-05, + "loss": 1.0292, + "step": 18492 + }, + { + "epoch": 1.37, + "learning_rate": 1.8018160868199044e-05, + "loss": 1.0389, + "step": 18493 + }, + { + "epoch": 1.37, + "learning_rate": 1.8017922507753635e-05, + "loss": 1.0333, + "step": 18494 + }, + { + "epoch": 1.37, + "learning_rate": 1.8017684134551826e-05, + "loss": 1.0474, + "step": 18495 + }, + { + "epoch": 1.37, + "learning_rate": 1.8017445748593998e-05, + "loss": 0.9205, + "step": 18496 + }, + { + "epoch": 1.37, + "learning_rate": 1.801720734988053e-05, + "loss": 1.0186, + "step": 18497 + }, + { + "epoch": 1.37, + "learning_rate": 1.8016968938411796e-05, + "loss": 0.9585, + "step": 18498 + }, + { + "epoch": 1.37, + "learning_rate": 1.8016730514188178e-05, + "loss": 0.9309, + "step": 18499 + }, + { + "epoch": 1.37, + "learning_rate": 1.8016492077210063e-05, + "loss": 1.0582, + "step": 18500 + }, + { + "epoch": 1.37, + "learning_rate": 1.8016253627477817e-05, + "loss": 1.0728, + "step": 18501 + }, + { + "epoch": 1.37, + "learning_rate": 1.801601516499183e-05, + "loss": 1.0507, + "step": 18502 + }, + { + "epoch": 1.37, + "learning_rate": 1.8015776689752477e-05, + "loss": 1.0542, + "step": 18503 + }, + { + "epoch": 1.37, + "learning_rate": 1.8015538201760136e-05, + "loss": 1.1318, + "step": 18504 + }, + { + "epoch": 1.37, + "learning_rate": 1.801529970101519e-05, + "loss": 0.9603, + "step": 18505 + }, + { + "epoch": 1.37, + "learning_rate": 1.8015061187518017e-05, + "loss": 0.936, + "step": 18506 + }, + { + "epoch": 1.37, + "learning_rate": 1.8014822661269e-05, + "loss": 0.9731, + "step": 18507 + }, + { + "epoch": 1.37, + "learning_rate": 1.801458412226851e-05, + "loss": 0.9941, + "step": 18508 + }, + { + "epoch": 1.37, + "learning_rate": 1.8014345570516932e-05, + "loss": 1.028, + "step": 18509 + }, + { + "epoch": 1.37, + "learning_rate": 1.8014107006014642e-05, + "loss": 1.0164, + "step": 18510 + }, + { + "epoch": 1.37, + "learning_rate": 1.8013868428762026e-05, + "loss": 0.9999, + "step": 18511 + }, + { + "epoch": 1.37, + "learning_rate": 1.801362983875946e-05, + "loss": 0.8765, + "step": 18512 + }, + { + "epoch": 1.37, + "learning_rate": 1.8013391236007324e-05, + "loss": 1.0102, + "step": 18513 + }, + { + "epoch": 1.37, + "learning_rate": 1.8013152620505997e-05, + "loss": 1.0361, + "step": 18514 + }, + { + "epoch": 1.37, + "learning_rate": 1.801291399225586e-05, + "loss": 0.935, + "step": 18515 + }, + { + "epoch": 1.37, + "learning_rate": 1.801267535125729e-05, + "loss": 0.9588, + "step": 18516 + }, + { + "epoch": 1.37, + "learning_rate": 1.8012436697510665e-05, + "loss": 0.9128, + "step": 18517 + }, + { + "epoch": 1.37, + "learning_rate": 1.801219803101637e-05, + "loss": 0.9912, + "step": 18518 + }, + { + "epoch": 1.37, + "learning_rate": 1.8011959351774786e-05, + "loss": 1.0948, + "step": 18519 + }, + { + "epoch": 1.37, + "learning_rate": 1.8011720659786286e-05, + "loss": 1.1363, + "step": 18520 + }, + { + "epoch": 1.37, + "learning_rate": 1.8011481955051254e-05, + "loss": 0.9808, + "step": 18521 + }, + { + "epoch": 1.37, + "learning_rate": 1.8011243237570067e-05, + "loss": 0.9844, + "step": 18522 + }, + { + "epoch": 1.37, + "learning_rate": 1.8011004507343108e-05, + "loss": 0.9586, + "step": 18523 + }, + { + "epoch": 1.37, + "learning_rate": 1.8010765764370756e-05, + "loss": 1.0014, + "step": 18524 + }, + { + "epoch": 1.37, + "learning_rate": 1.801052700865339e-05, + "loss": 1.0881, + "step": 18525 + }, + { + "epoch": 1.37, + "learning_rate": 1.8010288240191388e-05, + "loss": 0.9983, + "step": 18526 + }, + { + "epoch": 1.37, + "learning_rate": 1.8010049458985137e-05, + "loss": 1.0931, + "step": 18527 + }, + { + "epoch": 1.37, + "learning_rate": 1.800981066503501e-05, + "loss": 1.1008, + "step": 18528 + }, + { + "epoch": 1.37, + "learning_rate": 1.8009571858341384e-05, + "loss": 1.0158, + "step": 18529 + }, + { + "epoch": 1.37, + "learning_rate": 1.8009333038904648e-05, + "loss": 0.9548, + "step": 18530 + }, + { + "epoch": 1.37, + "learning_rate": 1.8009094206725178e-05, + "loss": 1.1008, + "step": 18531 + }, + { + "epoch": 1.37, + "learning_rate": 1.8008855361803354e-05, + "loss": 1.0315, + "step": 18532 + }, + { + "epoch": 1.37, + "learning_rate": 1.8008616504139556e-05, + "loss": 0.9462, + "step": 18533 + }, + { + "epoch": 1.37, + "learning_rate": 1.800837763373416e-05, + "loss": 1.0577, + "step": 18534 + }, + { + "epoch": 1.37, + "learning_rate": 1.8008138750587552e-05, + "loss": 1.0348, + "step": 18535 + }, + { + "epoch": 1.37, + "learning_rate": 1.800789985470011e-05, + "loss": 0.9719, + "step": 18536 + }, + { + "epoch": 1.37, + "learning_rate": 1.8007660946072215e-05, + "loss": 1.0439, + "step": 18537 + }, + { + "epoch": 1.37, + "learning_rate": 1.8007422024704244e-05, + "loss": 0.9075, + "step": 18538 + }, + { + "epoch": 1.37, + "learning_rate": 1.8007183090596577e-05, + "loss": 1.0273, + "step": 18539 + }, + { + "epoch": 1.37, + "learning_rate": 1.80069441437496e-05, + "loss": 1.0255, + "step": 18540 + }, + { + "epoch": 1.37, + "learning_rate": 1.800670518416369e-05, + "loss": 0.991, + "step": 18541 + }, + { + "epoch": 1.37, + "learning_rate": 1.8006466211839225e-05, + "loss": 1.139, + "step": 18542 + }, + { + "epoch": 1.37, + "learning_rate": 1.800622722677659e-05, + "loss": 1.0386, + "step": 18543 + }, + { + "epoch": 1.37, + "learning_rate": 1.8005988228976157e-05, + "loss": 1.0421, + "step": 18544 + }, + { + "epoch": 1.37, + "learning_rate": 1.800574921843831e-05, + "loss": 1.0303, + "step": 18545 + }, + { + "epoch": 1.37, + "learning_rate": 1.8005510195163438e-05, + "loss": 1.1034, + "step": 18546 + }, + { + "epoch": 1.37, + "learning_rate": 1.800527115915191e-05, + "loss": 0.9782, + "step": 18547 + }, + { + "epoch": 1.37, + "learning_rate": 1.800503211040411e-05, + "loss": 1.0224, + "step": 18548 + }, + { + "epoch": 1.37, + "learning_rate": 1.8004793048920417e-05, + "loss": 1.0173, + "step": 18549 + }, + { + "epoch": 1.37, + "learning_rate": 1.8004553974701214e-05, + "loss": 0.9479, + "step": 18550 + }, + { + "epoch": 1.37, + "learning_rate": 1.800431488774688e-05, + "loss": 1.0141, + "step": 18551 + }, + { + "epoch": 1.37, + "learning_rate": 1.8004075788057798e-05, + "loss": 1.0477, + "step": 18552 + }, + { + "epoch": 1.37, + "learning_rate": 1.8003836675634343e-05, + "loss": 0.9309, + "step": 18553 + }, + { + "epoch": 1.37, + "learning_rate": 1.80035975504769e-05, + "loss": 1.007, + "step": 18554 + }, + { + "epoch": 1.37, + "learning_rate": 1.800335841258585e-05, + "loss": 0.9034, + "step": 18555 + }, + { + "epoch": 1.37, + "learning_rate": 1.8003119261961567e-05, + "loss": 1.1165, + "step": 18556 + }, + { + "epoch": 1.37, + "learning_rate": 1.8002880098604432e-05, + "loss": 1.0129, + "step": 18557 + }, + { + "epoch": 1.37, + "learning_rate": 1.8002640922514837e-05, + "loss": 1.0239, + "step": 18558 + }, + { + "epoch": 1.37, + "learning_rate": 1.800240173369315e-05, + "loss": 1.075, + "step": 18559 + }, + { + "epoch": 1.37, + "learning_rate": 1.800216253213976e-05, + "loss": 1.0031, + "step": 18560 + }, + { + "epoch": 1.37, + "learning_rate": 1.800192331785504e-05, + "loss": 1.06, + "step": 18561 + }, + { + "epoch": 1.37, + "learning_rate": 1.8001684090839378e-05, + "loss": 1.0119, + "step": 18562 + }, + { + "epoch": 1.37, + "learning_rate": 1.800144485109315e-05, + "loss": 0.99, + "step": 18563 + }, + { + "epoch": 1.37, + "learning_rate": 1.8001205598616737e-05, + "loss": 1.1162, + "step": 18564 + }, + { + "epoch": 1.37, + "learning_rate": 1.8000966333410525e-05, + "loss": 0.9553, + "step": 18565 + }, + { + "epoch": 1.37, + "learning_rate": 1.8000727055474884e-05, + "loss": 0.9521, + "step": 18566 + }, + { + "epoch": 1.37, + "learning_rate": 1.8000487764810205e-05, + "loss": 1.0697, + "step": 18567 + }, + { + "epoch": 1.37, + "learning_rate": 1.800024846141686e-05, + "loss": 0.9578, + "step": 18568 + }, + { + "epoch": 1.37, + "learning_rate": 1.8000009145295236e-05, + "loss": 0.9628, + "step": 18569 + }, + { + "epoch": 1.37, + "learning_rate": 1.7999769816445712e-05, + "loss": 1.098, + "step": 18570 + }, + { + "epoch": 1.37, + "learning_rate": 1.799953047486867e-05, + "loss": 1.0181, + "step": 18571 + }, + { + "epoch": 1.37, + "learning_rate": 1.799929112056449e-05, + "loss": 1.0238, + "step": 18572 + }, + { + "epoch": 1.37, + "learning_rate": 1.799905175353355e-05, + "loss": 1.0445, + "step": 18573 + }, + { + "epoch": 1.37, + "learning_rate": 1.799881237377623e-05, + "loss": 1.0514, + "step": 18574 + }, + { + "epoch": 1.37, + "learning_rate": 1.799857298129292e-05, + "loss": 1.0459, + "step": 18575 + }, + { + "epoch": 1.37, + "learning_rate": 1.7998333576083994e-05, + "loss": 1.0425, + "step": 18576 + }, + { + "epoch": 1.37, + "learning_rate": 1.7998094158149833e-05, + "loss": 1.0337, + "step": 18577 + }, + { + "epoch": 1.37, + "learning_rate": 1.7997854727490815e-05, + "loss": 0.9943, + "step": 18578 + }, + { + "epoch": 1.37, + "learning_rate": 1.799761528410733e-05, + "loss": 1.0524, + "step": 18579 + }, + { + "epoch": 1.37, + "learning_rate": 1.799737582799975e-05, + "loss": 1.0138, + "step": 18580 + }, + { + "epoch": 1.37, + "learning_rate": 1.799713635916846e-05, + "loss": 0.9754, + "step": 18581 + }, + { + "epoch": 1.37, + "learning_rate": 1.7996896877613838e-05, + "loss": 1.0352, + "step": 18582 + }, + { + "epoch": 1.37, + "learning_rate": 1.7996657383336272e-05, + "loss": 1.0621, + "step": 18583 + }, + { + "epoch": 1.37, + "learning_rate": 1.799641787633614e-05, + "loss": 1.027, + "step": 18584 + }, + { + "epoch": 1.37, + "learning_rate": 1.7996178356613816e-05, + "loss": 1.0728, + "step": 18585 + }, + { + "epoch": 1.37, + "learning_rate": 1.799593882416969e-05, + "loss": 1.0642, + "step": 18586 + }, + { + "epoch": 1.37, + "learning_rate": 1.799569927900414e-05, + "loss": 0.9199, + "step": 18587 + }, + { + "epoch": 1.37, + "learning_rate": 1.7995459721117544e-05, + "loss": 1.0552, + "step": 18588 + }, + { + "epoch": 1.37, + "learning_rate": 1.7995220150510285e-05, + "loss": 1.0036, + "step": 18589 + }, + { + "epoch": 1.37, + "learning_rate": 1.799498056718275e-05, + "loss": 1.0335, + "step": 18590 + }, + { + "epoch": 1.37, + "learning_rate": 1.7994740971135314e-05, + "loss": 0.9726, + "step": 18591 + }, + { + "epoch": 1.37, + "learning_rate": 1.799450136236836e-05, + "loss": 1.0272, + "step": 18592 + }, + { + "epoch": 1.37, + "learning_rate": 1.7994261740882265e-05, + "loss": 1.1042, + "step": 18593 + }, + { + "epoch": 1.37, + "learning_rate": 1.7994022106677418e-05, + "loss": 1.1446, + "step": 18594 + }, + { + "epoch": 1.37, + "learning_rate": 1.799378245975419e-05, + "loss": 0.9475, + "step": 18595 + }, + { + "epoch": 1.37, + "learning_rate": 1.7993542800112975e-05, + "loss": 0.9985, + "step": 18596 + }, + { + "epoch": 1.37, + "learning_rate": 1.799330312775415e-05, + "loss": 1.0543, + "step": 18597 + }, + { + "epoch": 1.37, + "learning_rate": 1.7993063442678088e-05, + "loss": 1.0113, + "step": 18598 + }, + { + "epoch": 1.37, + "learning_rate": 1.7992823744885176e-05, + "loss": 0.9741, + "step": 18599 + }, + { + "epoch": 1.37, + "learning_rate": 1.79925840343758e-05, + "loss": 1.0834, + "step": 18600 + }, + { + "epoch": 1.37, + "learning_rate": 1.7992344311150334e-05, + "loss": 1.0082, + "step": 18601 + }, + { + "epoch": 1.37, + "learning_rate": 1.7992104575209162e-05, + "loss": 1.104, + "step": 18602 + }, + { + "epoch": 1.37, + "learning_rate": 1.799186482655267e-05, + "loss": 1.028, + "step": 18603 + }, + { + "epoch": 1.37, + "learning_rate": 1.7991625065181232e-05, + "loss": 1.0629, + "step": 18604 + }, + { + "epoch": 1.37, + "learning_rate": 1.7991385291095236e-05, + "loss": 0.8578, + "step": 18605 + }, + { + "epoch": 1.37, + "learning_rate": 1.799114550429506e-05, + "loss": 1.05, + "step": 18606 + }, + { + "epoch": 1.37, + "learning_rate": 1.799090570478108e-05, + "loss": 1.0059, + "step": 18607 + }, + { + "epoch": 1.37, + "learning_rate": 1.799066589255369e-05, + "loss": 1.0019, + "step": 18608 + }, + { + "epoch": 1.38, + "learning_rate": 1.7990426067613264e-05, + "loss": 0.916, + "step": 18609 + }, + { + "epoch": 1.38, + "learning_rate": 1.7990186229960183e-05, + "loss": 1.0363, + "step": 18610 + }, + { + "epoch": 1.38, + "learning_rate": 1.7989946379594828e-05, + "loss": 0.9974, + "step": 18611 + }, + { + "epoch": 1.38, + "learning_rate": 1.7989706516517586e-05, + "loss": 0.984, + "step": 18612 + }, + { + "epoch": 1.38, + "learning_rate": 1.798946664072883e-05, + "loss": 0.891, + "step": 18613 + }, + { + "epoch": 1.38, + "learning_rate": 1.7989226752228952e-05, + "loss": 0.9786, + "step": 18614 + }, + { + "epoch": 1.38, + "learning_rate": 1.7988986851018327e-05, + "loss": 1.0538, + "step": 18615 + }, + { + "epoch": 1.38, + "learning_rate": 1.798874693709734e-05, + "loss": 1.0385, + "step": 18616 + }, + { + "epoch": 1.38, + "learning_rate": 1.7988507010466367e-05, + "loss": 1.0646, + "step": 18617 + }, + { + "epoch": 1.38, + "learning_rate": 1.79882670711258e-05, + "loss": 0.9994, + "step": 18618 + }, + { + "epoch": 1.38, + "learning_rate": 1.798802711907601e-05, + "loss": 0.9694, + "step": 18619 + }, + { + "epoch": 1.38, + "learning_rate": 1.798778715431738e-05, + "loss": 1.0729, + "step": 18620 + }, + { + "epoch": 1.38, + "learning_rate": 1.79875471768503e-05, + "loss": 1.0141, + "step": 18621 + }, + { + "epoch": 1.38, + "learning_rate": 1.7987307186675144e-05, + "loss": 0.9884, + "step": 18622 + }, + { + "epoch": 1.38, + "learning_rate": 1.7987067183792298e-05, + "loss": 0.9159, + "step": 18623 + }, + { + "epoch": 1.38, + "learning_rate": 1.7986827168202145e-05, + "loss": 1.0943, + "step": 18624 + }, + { + "epoch": 1.38, + "learning_rate": 1.798658713990506e-05, + "loss": 1.0164, + "step": 18625 + }, + { + "epoch": 1.38, + "learning_rate": 1.798634709890143e-05, + "loss": 0.9946, + "step": 18626 + }, + { + "epoch": 1.38, + "learning_rate": 1.7986107045191637e-05, + "loss": 0.9987, + "step": 18627 + }, + { + "epoch": 1.38, + "learning_rate": 1.798586697877606e-05, + "loss": 0.9683, + "step": 18628 + }, + { + "epoch": 1.38, + "learning_rate": 1.7985626899655083e-05, + "loss": 1.0264, + "step": 18629 + }, + { + "epoch": 1.38, + "learning_rate": 1.7985386807829092e-05, + "loss": 1.0096, + "step": 18630 + }, + { + "epoch": 1.38, + "learning_rate": 1.798514670329846e-05, + "loss": 1.1294, + "step": 18631 + }, + { + "epoch": 1.38, + "learning_rate": 1.7984906586063578e-05, + "loss": 0.98, + "step": 18632 + }, + { + "epoch": 1.38, + "learning_rate": 1.798466645612482e-05, + "loss": 1.0499, + "step": 18633 + }, + { + "epoch": 1.38, + "learning_rate": 1.7984426313482575e-05, + "loss": 1.0256, + "step": 18634 + }, + { + "epoch": 1.38, + "learning_rate": 1.798418615813722e-05, + "loss": 1.0269, + "step": 18635 + }, + { + "epoch": 1.38, + "learning_rate": 1.798394599008914e-05, + "loss": 1.0518, + "step": 18636 + }, + { + "epoch": 1.38, + "learning_rate": 1.7983705809338716e-05, + "loss": 0.9754, + "step": 18637 + }, + { + "epoch": 1.38, + "learning_rate": 1.7983465615886332e-05, + "loss": 0.9587, + "step": 18638 + }, + { + "epoch": 1.38, + "learning_rate": 1.7983225409732364e-05, + "loss": 1.1059, + "step": 18639 + }, + { + "epoch": 1.38, + "learning_rate": 1.7982985190877204e-05, + "loss": 1.0759, + "step": 18640 + }, + { + "epoch": 1.38, + "learning_rate": 1.7982744959321222e-05, + "loss": 1.099, + "step": 18641 + }, + { + "epoch": 1.38, + "learning_rate": 1.798250471506481e-05, + "loss": 1.0951, + "step": 18642 + }, + { + "epoch": 1.38, + "learning_rate": 1.798226445810835e-05, + "loss": 1.0583, + "step": 18643 + }, + { + "epoch": 1.38, + "learning_rate": 1.798202418845222e-05, + "loss": 0.9725, + "step": 18644 + }, + { + "epoch": 1.38, + "learning_rate": 1.7981783906096803e-05, + "loss": 0.9296, + "step": 18645 + }, + { + "epoch": 1.38, + "learning_rate": 1.7981543611042482e-05, + "loss": 1.0656, + "step": 18646 + }, + { + "epoch": 1.38, + "learning_rate": 1.7981303303289642e-05, + "loss": 0.9723, + "step": 18647 + }, + { + "epoch": 1.38, + "learning_rate": 1.798106298283866e-05, + "loss": 1.0135, + "step": 18648 + }, + { + "epoch": 1.38, + "learning_rate": 1.7980822649689923e-05, + "loss": 1.1012, + "step": 18649 + }, + { + "epoch": 1.38, + "learning_rate": 1.798058230384381e-05, + "loss": 1.0034, + "step": 18650 + }, + { + "epoch": 1.38, + "learning_rate": 1.7980341945300702e-05, + "loss": 1.0855, + "step": 18651 + }, + { + "epoch": 1.38, + "learning_rate": 1.7980101574060985e-05, + "loss": 0.8944, + "step": 18652 + }, + { + "epoch": 1.38, + "learning_rate": 1.7979861190125043e-05, + "loss": 0.9878, + "step": 18653 + }, + { + "epoch": 1.38, + "learning_rate": 1.7979620793493256e-05, + "loss": 1.0123, + "step": 18654 + }, + { + "epoch": 1.38, + "learning_rate": 1.7979380384166004e-05, + "loss": 1.0296, + "step": 18655 + }, + { + "epoch": 1.38, + "learning_rate": 1.7979139962143673e-05, + "loss": 1.1095, + "step": 18656 + }, + { + "epoch": 1.38, + "learning_rate": 1.7978899527426646e-05, + "loss": 0.9921, + "step": 18657 + }, + { + "epoch": 1.38, + "learning_rate": 1.7978659080015305e-05, + "loss": 0.9857, + "step": 18658 + }, + { + "epoch": 1.38, + "learning_rate": 1.7978418619910027e-05, + "loss": 1.067, + "step": 18659 + }, + { + "epoch": 1.38, + "learning_rate": 1.79781781471112e-05, + "loss": 0.9768, + "step": 18660 + }, + { + "epoch": 1.38, + "learning_rate": 1.797793766161921e-05, + "loss": 1.0998, + "step": 18661 + }, + { + "epoch": 1.38, + "learning_rate": 1.7977697163434432e-05, + "loss": 1.0863, + "step": 18662 + }, + { + "epoch": 1.38, + "learning_rate": 1.7977456652557254e-05, + "loss": 1.1163, + "step": 18663 + }, + { + "epoch": 1.38, + "learning_rate": 1.7977216128988056e-05, + "loss": 0.9345, + "step": 18664 + }, + { + "epoch": 1.38, + "learning_rate": 1.7976975592727218e-05, + "loss": 1.0929, + "step": 18665 + }, + { + "epoch": 1.38, + "learning_rate": 1.797673504377513e-05, + "loss": 0.9985, + "step": 18666 + }, + { + "epoch": 1.38, + "learning_rate": 1.7976494482132173e-05, + "loss": 0.914, + "step": 18667 + }, + { + "epoch": 1.38, + "learning_rate": 1.7976253907798722e-05, + "loss": 0.9318, + "step": 18668 + }, + { + "epoch": 1.38, + "learning_rate": 1.7976013320775166e-05, + "loss": 0.9444, + "step": 18669 + }, + { + "epoch": 1.38, + "learning_rate": 1.7975772721061887e-05, + "loss": 0.9577, + "step": 18670 + }, + { + "epoch": 1.38, + "learning_rate": 1.797553210865927e-05, + "loss": 1.0257, + "step": 18671 + }, + { + "epoch": 1.38, + "learning_rate": 1.797529148356769e-05, + "loss": 0.9696, + "step": 18672 + }, + { + "epoch": 1.38, + "learning_rate": 1.797505084578754e-05, + "loss": 0.9967, + "step": 18673 + }, + { + "epoch": 1.38, + "learning_rate": 1.79748101953192e-05, + "loss": 0.9534, + "step": 18674 + }, + { + "epoch": 1.38, + "learning_rate": 1.797456953216305e-05, + "loss": 1.0024, + "step": 18675 + }, + { + "epoch": 1.38, + "learning_rate": 1.7974328856319473e-05, + "loss": 1.072, + "step": 18676 + }, + { + "epoch": 1.38, + "learning_rate": 1.797408816778885e-05, + "loss": 0.9646, + "step": 18677 + }, + { + "epoch": 1.38, + "learning_rate": 1.797384746657157e-05, + "loss": 1.0061, + "step": 18678 + }, + { + "epoch": 1.38, + "learning_rate": 1.7973606752668012e-05, + "loss": 1.1064, + "step": 18679 + }, + { + "epoch": 1.38, + "learning_rate": 1.797336602607856e-05, + "loss": 1.0144, + "step": 18680 + }, + { + "epoch": 1.38, + "learning_rate": 1.7973125286803595e-05, + "loss": 1.1059, + "step": 18681 + }, + { + "epoch": 1.38, + "learning_rate": 1.7972884534843503e-05, + "loss": 0.9825, + "step": 18682 + }, + { + "epoch": 1.38, + "learning_rate": 1.797264377019867e-05, + "loss": 1.1739, + "step": 18683 + }, + { + "epoch": 1.38, + "learning_rate": 1.797240299286947e-05, + "loss": 0.9444, + "step": 18684 + }, + { + "epoch": 1.38, + "learning_rate": 1.7972162202856293e-05, + "loss": 1.1396, + "step": 18685 + }, + { + "epoch": 1.38, + "learning_rate": 1.7971921400159517e-05, + "loss": 1.0255, + "step": 18686 + }, + { + "epoch": 1.38, + "learning_rate": 1.797168058477953e-05, + "loss": 1.1206, + "step": 18687 + }, + { + "epoch": 1.38, + "learning_rate": 1.7971439756716713e-05, + "loss": 1.0767, + "step": 18688 + }, + { + "epoch": 1.38, + "learning_rate": 1.7971198915971447e-05, + "loss": 0.92, + "step": 18689 + }, + { + "epoch": 1.38, + "learning_rate": 1.7970958062544123e-05, + "loss": 1.0026, + "step": 18690 + }, + { + "epoch": 1.38, + "learning_rate": 1.7970717196435113e-05, + "loss": 1.0194, + "step": 18691 + }, + { + "epoch": 1.38, + "learning_rate": 1.797047631764481e-05, + "loss": 0.9879, + "step": 18692 + }, + { + "epoch": 1.38, + "learning_rate": 1.797023542617359e-05, + "loss": 1.0862, + "step": 18693 + }, + { + "epoch": 1.38, + "learning_rate": 1.7969994522021844e-05, + "loss": 0.9902, + "step": 18694 + }, + { + "epoch": 1.38, + "learning_rate": 1.7969753605189947e-05, + "loss": 0.9588, + "step": 18695 + }, + { + "epoch": 1.38, + "learning_rate": 1.7969512675678288e-05, + "loss": 0.9283, + "step": 18696 + }, + { + "epoch": 1.38, + "learning_rate": 1.7969271733487247e-05, + "loss": 1.0665, + "step": 18697 + }, + { + "epoch": 1.38, + "learning_rate": 1.796903077861721e-05, + "loss": 1.0971, + "step": 18698 + }, + { + "epoch": 1.38, + "learning_rate": 1.7968789811068558e-05, + "loss": 1.0624, + "step": 18699 + }, + { + "epoch": 1.38, + "learning_rate": 1.796854883084168e-05, + "loss": 1.0215, + "step": 18700 + }, + { + "epoch": 1.38, + "learning_rate": 1.796830783793695e-05, + "loss": 0.9183, + "step": 18701 + }, + { + "epoch": 1.38, + "learning_rate": 1.7968066832354756e-05, + "loss": 0.9878, + "step": 18702 + }, + { + "epoch": 1.38, + "learning_rate": 1.796782581409548e-05, + "loss": 0.9513, + "step": 18703 + }, + { + "epoch": 1.38, + "learning_rate": 1.796758478315951e-05, + "loss": 1.033, + "step": 18704 + }, + { + "epoch": 1.38, + "learning_rate": 1.796734373954723e-05, + "loss": 1.0941, + "step": 18705 + }, + { + "epoch": 1.38, + "learning_rate": 1.7967102683259013e-05, + "loss": 1.0674, + "step": 18706 + }, + { + "epoch": 1.38, + "learning_rate": 1.7966861614295255e-05, + "loss": 0.9671, + "step": 18707 + }, + { + "epoch": 1.38, + "learning_rate": 1.796662053265633e-05, + "loss": 0.9668, + "step": 18708 + }, + { + "epoch": 1.38, + "learning_rate": 1.796637943834263e-05, + "loss": 0.9605, + "step": 18709 + }, + { + "epoch": 1.38, + "learning_rate": 1.796613833135453e-05, + "loss": 0.8725, + "step": 18710 + }, + { + "epoch": 1.38, + "learning_rate": 1.796589721169242e-05, + "loss": 1.0777, + "step": 18711 + }, + { + "epoch": 1.38, + "learning_rate": 1.7965656079356683e-05, + "loss": 0.9811, + "step": 18712 + }, + { + "epoch": 1.38, + "learning_rate": 1.79654149343477e-05, + "loss": 0.9659, + "step": 18713 + }, + { + "epoch": 1.38, + "learning_rate": 1.7965173776665856e-05, + "loss": 0.999, + "step": 18714 + }, + { + "epoch": 1.38, + "learning_rate": 1.7964932606311532e-05, + "loss": 0.958, + "step": 18715 + }, + { + "epoch": 1.38, + "learning_rate": 1.7964691423285118e-05, + "loss": 1.1232, + "step": 18716 + }, + { + "epoch": 1.38, + "learning_rate": 1.7964450227586993e-05, + "loss": 1.0787, + "step": 18717 + }, + { + "epoch": 1.38, + "learning_rate": 1.7964209019217542e-05, + "loss": 1.0087, + "step": 18718 + }, + { + "epoch": 1.38, + "learning_rate": 1.7963967798177147e-05, + "loss": 1.0224, + "step": 18719 + }, + { + "epoch": 1.38, + "learning_rate": 1.796372656446619e-05, + "loss": 1.0018, + "step": 18720 + }, + { + "epoch": 1.38, + "learning_rate": 1.7963485318085064e-05, + "loss": 0.8867, + "step": 18721 + }, + { + "epoch": 1.38, + "learning_rate": 1.7963244059034142e-05, + "loss": 0.9047, + "step": 18722 + }, + { + "epoch": 1.38, + "learning_rate": 1.7963002787313817e-05, + "loss": 1.0859, + "step": 18723 + }, + { + "epoch": 1.38, + "learning_rate": 1.7962761502924468e-05, + "loss": 1.0064, + "step": 18724 + }, + { + "epoch": 1.38, + "learning_rate": 1.7962520205866476e-05, + "loss": 1.0768, + "step": 18725 + }, + { + "epoch": 1.38, + "learning_rate": 1.796227889614023e-05, + "loss": 0.984, + "step": 18726 + }, + { + "epoch": 1.38, + "learning_rate": 1.796203757374611e-05, + "loss": 1.0163, + "step": 18727 + }, + { + "epoch": 1.38, + "learning_rate": 1.7961796238684507e-05, + "loss": 1.0869, + "step": 18728 + }, + { + "epoch": 1.38, + "learning_rate": 1.79615548909558e-05, + "loss": 0.9239, + "step": 18729 + }, + { + "epoch": 1.38, + "learning_rate": 1.7961313530560367e-05, + "loss": 1.0995, + "step": 18730 + }, + { + "epoch": 1.38, + "learning_rate": 1.79610721574986e-05, + "loss": 1.0567, + "step": 18731 + }, + { + "epoch": 1.38, + "learning_rate": 1.7960830771770883e-05, + "loss": 1.0341, + "step": 18732 + }, + { + "epoch": 1.38, + "learning_rate": 1.79605893733776e-05, + "loss": 1.0876, + "step": 18733 + }, + { + "epoch": 1.38, + "learning_rate": 1.796034796231913e-05, + "loss": 0.9222, + "step": 18734 + }, + { + "epoch": 1.38, + "learning_rate": 1.796010653859586e-05, + "loss": 0.9506, + "step": 18735 + }, + { + "epoch": 1.38, + "learning_rate": 1.7959865102208177e-05, + "loss": 0.9438, + "step": 18736 + }, + { + "epoch": 1.38, + "learning_rate": 1.7959623653156463e-05, + "loss": 1.0025, + "step": 18737 + }, + { + "epoch": 1.38, + "learning_rate": 1.7959382191441098e-05, + "loss": 0.9677, + "step": 18738 + }, + { + "epoch": 1.38, + "learning_rate": 1.7959140717062472e-05, + "loss": 1.0242, + "step": 18739 + }, + { + "epoch": 1.38, + "learning_rate": 1.7958899230020964e-05, + "loss": 1.0164, + "step": 18740 + }, + { + "epoch": 1.38, + "learning_rate": 1.7958657730316968e-05, + "loss": 1.0236, + "step": 18741 + }, + { + "epoch": 1.38, + "learning_rate": 1.7958416217950855e-05, + "loss": 0.9326, + "step": 18742 + }, + { + "epoch": 1.38, + "learning_rate": 1.7958174692923016e-05, + "loss": 1.1691, + "step": 18743 + }, + { + "epoch": 1.38, + "learning_rate": 1.7957933155233838e-05, + "loss": 0.9613, + "step": 18744 + }, + { + "epoch": 1.39, + "learning_rate": 1.79576916048837e-05, + "loss": 1.0703, + "step": 18745 + }, + { + "epoch": 1.39, + "learning_rate": 1.795745004187299e-05, + "loss": 1.0129, + "step": 18746 + }, + { + "epoch": 1.39, + "learning_rate": 1.7957208466202093e-05, + "loss": 1.0127, + "step": 18747 + }, + { + "epoch": 1.39, + "learning_rate": 1.7956966877871385e-05, + "loss": 0.7995, + "step": 18748 + }, + { + "epoch": 1.39, + "learning_rate": 1.7956725276881263e-05, + "loss": 0.9778, + "step": 18749 + }, + { + "epoch": 1.39, + "learning_rate": 1.79564836632321e-05, + "loss": 1.0784, + "step": 18750 + }, + { + "epoch": 1.39, + "learning_rate": 1.7956242036924288e-05, + "loss": 1.0154, + "step": 18751 + }, + { + "epoch": 1.39, + "learning_rate": 1.795600039795821e-05, + "loss": 0.9073, + "step": 18752 + }, + { + "epoch": 1.39, + "learning_rate": 1.7955758746334245e-05, + "loss": 0.888, + "step": 18753 + }, + { + "epoch": 1.39, + "learning_rate": 1.7955517082052787e-05, + "loss": 1.0074, + "step": 18754 + }, + { + "epoch": 1.39, + "learning_rate": 1.795527540511421e-05, + "loss": 0.8943, + "step": 18755 + }, + { + "epoch": 1.39, + "learning_rate": 1.795503371551891e-05, + "loss": 0.8689, + "step": 18756 + }, + { + "epoch": 1.39, + "learning_rate": 1.7954792013267258e-05, + "loss": 1.0666, + "step": 18757 + }, + { + "epoch": 1.39, + "learning_rate": 1.7954550298359647e-05, + "loss": 1.0196, + "step": 18758 + }, + { + "epoch": 1.39, + "learning_rate": 1.7954308570796462e-05, + "loss": 1.0506, + "step": 18759 + }, + { + "epoch": 1.39, + "learning_rate": 1.7954066830578087e-05, + "loss": 1.068, + "step": 18760 + }, + { + "epoch": 1.39, + "learning_rate": 1.7953825077704904e-05, + "loss": 1.0821, + "step": 18761 + }, + { + "epoch": 1.39, + "learning_rate": 1.79535833121773e-05, + "loss": 1.0426, + "step": 18762 + }, + { + "epoch": 1.39, + "learning_rate": 1.7953341533995657e-05, + "loss": 1.1178, + "step": 18763 + }, + { + "epoch": 1.39, + "learning_rate": 1.7953099743160365e-05, + "loss": 0.998, + "step": 18764 + }, + { + "epoch": 1.39, + "learning_rate": 1.79528579396718e-05, + "loss": 0.9354, + "step": 18765 + }, + { + "epoch": 1.39, + "learning_rate": 1.7952616123530355e-05, + "loss": 0.9853, + "step": 18766 + }, + { + "epoch": 1.39, + "learning_rate": 1.7952374294736413e-05, + "loss": 0.974, + "step": 18767 + }, + { + "epoch": 1.39, + "learning_rate": 1.7952132453290354e-05, + "loss": 0.9647, + "step": 18768 + }, + { + "epoch": 1.39, + "learning_rate": 1.795189059919257e-05, + "loss": 1.0337, + "step": 18769 + }, + { + "epoch": 1.39, + "learning_rate": 1.7951648732443437e-05, + "loss": 0.9597, + "step": 18770 + }, + { + "epoch": 1.39, + "learning_rate": 1.7951406853043346e-05, + "loss": 1.0394, + "step": 18771 + }, + { + "epoch": 1.39, + "learning_rate": 1.7951164960992682e-05, + "loss": 1.0599, + "step": 18772 + }, + { + "epoch": 1.39, + "learning_rate": 1.795092305629183e-05, + "loss": 1.0021, + "step": 18773 + }, + { + "epoch": 1.39, + "learning_rate": 1.7950681138941168e-05, + "loss": 1.038, + "step": 18774 + }, + { + "epoch": 1.39, + "learning_rate": 1.795043920894109e-05, + "loss": 1.0198, + "step": 18775 + }, + { + "epoch": 1.39, + "learning_rate": 1.7950197266291975e-05, + "loss": 0.9962, + "step": 18776 + }, + { + "epoch": 1.39, + "learning_rate": 1.794995531099421e-05, + "loss": 1.0192, + "step": 18777 + }, + { + "epoch": 1.39, + "learning_rate": 1.7949713343048182e-05, + "loss": 1.0465, + "step": 18778 + }, + { + "epoch": 1.39, + "learning_rate": 1.794947136245427e-05, + "loss": 1.0112, + "step": 18779 + }, + { + "epoch": 1.39, + "learning_rate": 1.7949229369212866e-05, + "loss": 1.0485, + "step": 18780 + }, + { + "epoch": 1.39, + "learning_rate": 1.7948987363324352e-05, + "loss": 1.0332, + "step": 18781 + }, + { + "epoch": 1.39, + "learning_rate": 1.794874534478911e-05, + "loss": 1.0829, + "step": 18782 + }, + { + "epoch": 1.39, + "learning_rate": 1.794850331360753e-05, + "loss": 1.0292, + "step": 18783 + }, + { + "epoch": 1.39, + "learning_rate": 1.7948261269779993e-05, + "loss": 1.0539, + "step": 18784 + }, + { + "epoch": 1.39, + "learning_rate": 1.7948019213306888e-05, + "loss": 0.9689, + "step": 18785 + }, + { + "epoch": 1.39, + "learning_rate": 1.79477771441886e-05, + "loss": 1.0689, + "step": 18786 + }, + { + "epoch": 1.39, + "learning_rate": 1.794753506242551e-05, + "loss": 1.0609, + "step": 18787 + }, + { + "epoch": 1.39, + "learning_rate": 1.7947292968018005e-05, + "loss": 1.1028, + "step": 18788 + }, + { + "epoch": 1.39, + "learning_rate": 1.7947050860966473e-05, + "loss": 1.0277, + "step": 18789 + }, + { + "epoch": 1.39, + "learning_rate": 1.7946808741271293e-05, + "loss": 1.022, + "step": 18790 + }, + { + "epoch": 1.39, + "learning_rate": 1.7946566608932855e-05, + "loss": 0.9516, + "step": 18791 + }, + { + "epoch": 1.39, + "learning_rate": 1.7946324463951547e-05, + "loss": 0.9689, + "step": 18792 + }, + { + "epoch": 1.39, + "learning_rate": 1.7946082306327746e-05, + "loss": 0.987, + "step": 18793 + }, + { + "epoch": 1.39, + "learning_rate": 1.7945840136061843e-05, + "loss": 0.9446, + "step": 18794 + }, + { + "epoch": 1.39, + "learning_rate": 1.7945597953154224e-05, + "loss": 1.0161, + "step": 18795 + }, + { + "epoch": 1.39, + "learning_rate": 1.7945355757605268e-05, + "loss": 1.0305, + "step": 18796 + }, + { + "epoch": 1.39, + "learning_rate": 1.794511354941537e-05, + "loss": 1.0061, + "step": 18797 + }, + { + "epoch": 1.39, + "learning_rate": 1.7944871328584906e-05, + "loss": 1.0572, + "step": 18798 + }, + { + "epoch": 1.39, + "learning_rate": 1.794462909511427e-05, + "loss": 1.0908, + "step": 18799 + }, + { + "epoch": 1.39, + "learning_rate": 1.7944386849003836e-05, + "loss": 0.9793, + "step": 18800 + }, + { + "epoch": 1.39, + "learning_rate": 1.7944144590254e-05, + "loss": 1.0381, + "step": 18801 + }, + { + "epoch": 1.39, + "learning_rate": 1.7943902318865145e-05, + "loss": 1.0242, + "step": 18802 + }, + { + "epoch": 1.39, + "learning_rate": 1.7943660034837655e-05, + "loss": 1.0596, + "step": 18803 + }, + { + "epoch": 1.39, + "learning_rate": 1.7943417738171912e-05, + "loss": 0.9809, + "step": 18804 + }, + { + "epoch": 1.39, + "learning_rate": 1.794317542886831e-05, + "loss": 1.1525, + "step": 18805 + }, + { + "epoch": 1.39, + "learning_rate": 1.7942933106927224e-05, + "loss": 1.0126, + "step": 18806 + }, + { + "epoch": 1.39, + "learning_rate": 1.794269077234905e-05, + "loss": 1.0444, + "step": 18807 + }, + { + "epoch": 1.39, + "learning_rate": 1.7942448425134165e-05, + "loss": 0.915, + "step": 18808 + }, + { + "epoch": 1.39, + "learning_rate": 1.794220606528296e-05, + "loss": 1.1952, + "step": 18809 + }, + { + "epoch": 1.39, + "learning_rate": 1.7941963692795817e-05, + "loss": 1.0358, + "step": 18810 + }, + { + "epoch": 1.39, + "learning_rate": 1.7941721307673123e-05, + "loss": 1.0785, + "step": 18811 + }, + { + "epoch": 1.39, + "learning_rate": 1.7941478909915268e-05, + "loss": 1.0443, + "step": 18812 + }, + { + "epoch": 1.39, + "learning_rate": 1.794123649952263e-05, + "loss": 0.9517, + "step": 18813 + }, + { + "epoch": 1.39, + "learning_rate": 1.79409940764956e-05, + "loss": 0.9963, + "step": 18814 + }, + { + "epoch": 1.39, + "learning_rate": 1.7940751640834562e-05, + "loss": 1.0532, + "step": 18815 + }, + { + "epoch": 1.39, + "learning_rate": 1.79405091925399e-05, + "loss": 1.0661, + "step": 18816 + }, + { + "epoch": 1.39, + "learning_rate": 1.7940266731612005e-05, + "loss": 1.0588, + "step": 18817 + }, + { + "epoch": 1.39, + "learning_rate": 1.7940024258051254e-05, + "loss": 1.063, + "step": 18818 + }, + { + "epoch": 1.39, + "learning_rate": 1.7939781771858043e-05, + "loss": 0.9975, + "step": 18819 + }, + { + "epoch": 1.39, + "learning_rate": 1.7939539273032753e-05, + "loss": 0.9864, + "step": 18820 + }, + { + "epoch": 1.39, + "learning_rate": 1.7939296761575765e-05, + "loss": 0.8701, + "step": 18821 + }, + { + "epoch": 1.39, + "learning_rate": 1.7939054237487472e-05, + "loss": 1.1053, + "step": 18822 + }, + { + "epoch": 1.39, + "learning_rate": 1.793881170076826e-05, + "loss": 1.0591, + "step": 18823 + }, + { + "epoch": 1.39, + "learning_rate": 1.7938569151418506e-05, + "loss": 0.9941, + "step": 18824 + }, + { + "epoch": 1.39, + "learning_rate": 1.7938326589438605e-05, + "loss": 0.9664, + "step": 18825 + }, + { + "epoch": 1.39, + "learning_rate": 1.793808401482894e-05, + "loss": 1.0634, + "step": 18826 + }, + { + "epoch": 1.39, + "learning_rate": 1.79378414275899e-05, + "loss": 1.0243, + "step": 18827 + }, + { + "epoch": 1.39, + "learning_rate": 1.7937598827721864e-05, + "loss": 1.0026, + "step": 18828 + }, + { + "epoch": 1.39, + "learning_rate": 1.7937356215225222e-05, + "loss": 1.1192, + "step": 18829 + }, + { + "epoch": 1.39, + "learning_rate": 1.7937113590100363e-05, + "loss": 0.9846, + "step": 18830 + }, + { + "epoch": 1.39, + "learning_rate": 1.7936870952347666e-05, + "loss": 1.0663, + "step": 18831 + }, + { + "epoch": 1.39, + "learning_rate": 1.7936628301967524e-05, + "loss": 0.9896, + "step": 18832 + }, + { + "epoch": 1.39, + "learning_rate": 1.793638563896032e-05, + "loss": 1.0574, + "step": 18833 + }, + { + "epoch": 1.39, + "learning_rate": 1.7936142963326438e-05, + "loss": 1.0479, + "step": 18834 + }, + { + "epoch": 1.39, + "learning_rate": 1.7935900275066268e-05, + "loss": 0.9243, + "step": 18835 + }, + { + "epoch": 1.39, + "learning_rate": 1.7935657574180193e-05, + "loss": 1.0316, + "step": 18836 + }, + { + "epoch": 1.39, + "learning_rate": 1.7935414860668598e-05, + "loss": 0.9863, + "step": 18837 + }, + { + "epoch": 1.39, + "learning_rate": 1.7935172134531872e-05, + "loss": 1.0171, + "step": 18838 + }, + { + "epoch": 1.39, + "learning_rate": 1.7934929395770404e-05, + "loss": 1.0465, + "step": 18839 + }, + { + "epoch": 1.39, + "learning_rate": 1.7934686644384575e-05, + "loss": 0.9878, + "step": 18840 + }, + { + "epoch": 1.39, + "learning_rate": 1.7934443880374773e-05, + "loss": 0.9954, + "step": 18841 + }, + { + "epoch": 1.39, + "learning_rate": 1.7934201103741388e-05, + "loss": 1.0083, + "step": 18842 + }, + { + "epoch": 1.39, + "learning_rate": 1.79339583144848e-05, + "loss": 1.0995, + "step": 18843 + }, + { + "epoch": 1.39, + "learning_rate": 1.7933715512605395e-05, + "loss": 1.0556, + "step": 18844 + }, + { + "epoch": 1.39, + "learning_rate": 1.7933472698103566e-05, + "loss": 0.8575, + "step": 18845 + }, + { + "epoch": 1.39, + "learning_rate": 1.7933229870979693e-05, + "loss": 1.0007, + "step": 18846 + }, + { + "epoch": 1.39, + "learning_rate": 1.7932987031234164e-05, + "loss": 1.0334, + "step": 18847 + }, + { + "epoch": 1.39, + "learning_rate": 1.7932744178867367e-05, + "loss": 1.1031, + "step": 18848 + }, + { + "epoch": 1.39, + "learning_rate": 1.7932501313879688e-05, + "loss": 0.9832, + "step": 18849 + }, + { + "epoch": 1.39, + "learning_rate": 1.7932258436271515e-05, + "loss": 0.9565, + "step": 18850 + }, + { + "epoch": 1.39, + "learning_rate": 1.793201554604323e-05, + "loss": 0.9623, + "step": 18851 + }, + { + "epoch": 1.39, + "learning_rate": 1.793177264319522e-05, + "loss": 0.99, + "step": 18852 + }, + { + "epoch": 1.39, + "learning_rate": 1.7931529727727878e-05, + "loss": 1.0653, + "step": 18853 + }, + { + "epoch": 1.39, + "learning_rate": 1.7931286799641583e-05, + "loss": 1.057, + "step": 18854 + }, + { + "epoch": 1.39, + "learning_rate": 1.7931043858936724e-05, + "loss": 0.9877, + "step": 18855 + }, + { + "epoch": 1.39, + "learning_rate": 1.7930800905613686e-05, + "loss": 1.0544, + "step": 18856 + }, + { + "epoch": 1.39, + "learning_rate": 1.793055793967286e-05, + "loss": 1.0356, + "step": 18857 + }, + { + "epoch": 1.39, + "learning_rate": 1.7930314961114626e-05, + "loss": 1.1509, + "step": 18858 + }, + { + "epoch": 1.39, + "learning_rate": 1.793007196993938e-05, + "loss": 1.0046, + "step": 18859 + }, + { + "epoch": 1.39, + "learning_rate": 1.79298289661475e-05, + "loss": 1.1839, + "step": 18860 + }, + { + "epoch": 1.39, + "learning_rate": 1.7929585949739376e-05, + "loss": 0.994, + "step": 18861 + }, + { + "epoch": 1.39, + "learning_rate": 1.7929342920715394e-05, + "loss": 1.0148, + "step": 18862 + }, + { + "epoch": 1.39, + "learning_rate": 1.792909987907594e-05, + "loss": 1.0553, + "step": 18863 + }, + { + "epoch": 1.39, + "learning_rate": 1.7928856824821402e-05, + "loss": 0.9143, + "step": 18864 + }, + { + "epoch": 1.39, + "learning_rate": 1.7928613757952168e-05, + "loss": 1.0347, + "step": 18865 + }, + { + "epoch": 1.39, + "learning_rate": 1.792837067846862e-05, + "loss": 0.9753, + "step": 18866 + }, + { + "epoch": 1.39, + "learning_rate": 1.792812758637115e-05, + "loss": 1.0034, + "step": 18867 + }, + { + "epoch": 1.39, + "learning_rate": 1.792788448166014e-05, + "loss": 1.083, + "step": 18868 + }, + { + "epoch": 1.39, + "learning_rate": 1.7927641364335982e-05, + "loss": 0.985, + "step": 18869 + }, + { + "epoch": 1.39, + "learning_rate": 1.792739823439906e-05, + "loss": 1.0747, + "step": 18870 + }, + { + "epoch": 1.39, + "learning_rate": 1.792715509184976e-05, + "loss": 1.0703, + "step": 18871 + }, + { + "epoch": 1.39, + "learning_rate": 1.792691193668847e-05, + "loss": 1.0265, + "step": 18872 + }, + { + "epoch": 1.39, + "learning_rate": 1.7926668768915572e-05, + "loss": 1.0433, + "step": 18873 + }, + { + "epoch": 1.39, + "learning_rate": 1.7926425588531465e-05, + "loss": 1.0272, + "step": 18874 + }, + { + "epoch": 1.39, + "learning_rate": 1.792618239553652e-05, + "loss": 1.036, + "step": 18875 + }, + { + "epoch": 1.39, + "learning_rate": 1.792593918993114e-05, + "loss": 0.9359, + "step": 18876 + }, + { + "epoch": 1.39, + "learning_rate": 1.7925695971715702e-05, + "loss": 1.0526, + "step": 18877 + }, + { + "epoch": 1.39, + "learning_rate": 1.7925452740890593e-05, + "loss": 1.0897, + "step": 18878 + }, + { + "epoch": 1.39, + "learning_rate": 1.7925209497456203e-05, + "loss": 0.9741, + "step": 18879 + }, + { + "epoch": 1.4, + "learning_rate": 1.792496624141292e-05, + "loss": 0.9514, + "step": 18880 + }, + { + "epoch": 1.4, + "learning_rate": 1.7924722972761126e-05, + "loss": 0.9692, + "step": 18881 + }, + { + "epoch": 1.4, + "learning_rate": 1.792447969150121e-05, + "loss": 1.0941, + "step": 18882 + }, + { + "epoch": 1.4, + "learning_rate": 1.7924236397633565e-05, + "loss": 1.021, + "step": 18883 + }, + { + "epoch": 1.4, + "learning_rate": 1.792399309115857e-05, + "loss": 0.934, + "step": 18884 + }, + { + "epoch": 1.4, + "learning_rate": 1.7923749772076616e-05, + "loss": 1.0297, + "step": 18885 + }, + { + "epoch": 1.4, + "learning_rate": 1.792350644038809e-05, + "loss": 1.0574, + "step": 18886 + }, + { + "epoch": 1.4, + "learning_rate": 1.792326309609338e-05, + "loss": 0.9933, + "step": 18887 + }, + { + "epoch": 1.4, + "learning_rate": 1.792301973919287e-05, + "loss": 1.0396, + "step": 18888 + }, + { + "epoch": 1.4, + "learning_rate": 1.792277636968695e-05, + "loss": 1.043, + "step": 18889 + }, + { + "epoch": 1.4, + "learning_rate": 1.7922532987576006e-05, + "loss": 1.014, + "step": 18890 + }, + { + "epoch": 1.4, + "learning_rate": 1.7922289592860426e-05, + "loss": 0.9895, + "step": 18891 + }, + { + "epoch": 1.4, + "learning_rate": 1.792204618554059e-05, + "loss": 1.0783, + "step": 18892 + }, + { + "epoch": 1.4, + "learning_rate": 1.7921802765616896e-05, + "loss": 1.0028, + "step": 18893 + }, + { + "epoch": 1.4, + "learning_rate": 1.7921559333089732e-05, + "loss": 0.9867, + "step": 18894 + }, + { + "epoch": 1.4, + "learning_rate": 1.7921315887959477e-05, + "loss": 0.8525, + "step": 18895 + }, + { + "epoch": 1.4, + "learning_rate": 1.792107243022652e-05, + "loss": 0.9627, + "step": 18896 + }, + { + "epoch": 1.4, + "learning_rate": 1.792082895989125e-05, + "loss": 1.0525, + "step": 18897 + }, + { + "epoch": 1.4, + "learning_rate": 1.792058547695406e-05, + "loss": 1.1273, + "step": 18898 + }, + { + "epoch": 1.4, + "learning_rate": 1.7920341981415325e-05, + "loss": 1.0208, + "step": 18899 + }, + { + "epoch": 1.4, + "learning_rate": 1.7920098473275444e-05, + "loss": 0.8967, + "step": 18900 + }, + { + "epoch": 1.4, + "learning_rate": 1.7919854952534796e-05, + "loss": 1.0657, + "step": 18901 + }, + { + "epoch": 1.4, + "learning_rate": 1.7919611419193774e-05, + "loss": 1.0555, + "step": 18902 + }, + { + "epoch": 1.4, + "learning_rate": 1.7919367873252764e-05, + "loss": 0.9788, + "step": 18903 + }, + { + "epoch": 1.4, + "learning_rate": 1.7919124314712152e-05, + "loss": 1.0253, + "step": 18904 + }, + { + "epoch": 1.4, + "learning_rate": 1.7918880743572326e-05, + "loss": 0.9897, + "step": 18905 + }, + { + "epoch": 1.4, + "learning_rate": 1.7918637159833677e-05, + "loss": 0.8959, + "step": 18906 + }, + { + "epoch": 1.4, + "learning_rate": 1.7918393563496588e-05, + "loss": 0.9857, + "step": 18907 + }, + { + "epoch": 1.4, + "learning_rate": 1.7918149954561447e-05, + "loss": 1.0943, + "step": 18908 + }, + { + "epoch": 1.4, + "learning_rate": 1.791790633302864e-05, + "loss": 1.089, + "step": 18909 + }, + { + "epoch": 1.4, + "learning_rate": 1.7917662698898563e-05, + "loss": 0.9474, + "step": 18910 + }, + { + "epoch": 1.4, + "learning_rate": 1.7917419052171596e-05, + "loss": 1.1017, + "step": 18911 + }, + { + "epoch": 1.4, + "learning_rate": 1.7917175392848128e-05, + "loss": 1.0764, + "step": 18912 + }, + { + "epoch": 1.4, + "learning_rate": 1.7916931720928548e-05, + "loss": 1.1176, + "step": 18913 + }, + { + "epoch": 1.4, + "learning_rate": 1.7916688036413243e-05, + "loss": 0.9482, + "step": 18914 + }, + { + "epoch": 1.4, + "learning_rate": 1.7916444339302602e-05, + "loss": 0.9702, + "step": 18915 + }, + { + "epoch": 1.4, + "learning_rate": 1.7916200629597008e-05, + "loss": 0.9781, + "step": 18916 + }, + { + "epoch": 1.4, + "learning_rate": 1.7915956907296853e-05, + "loss": 0.9584, + "step": 18917 + }, + { + "epoch": 1.4, + "learning_rate": 1.7915713172402522e-05, + "loss": 1.1241, + "step": 18918 + }, + { + "epoch": 1.4, + "learning_rate": 1.791546942491441e-05, + "loss": 1.1051, + "step": 18919 + }, + { + "epoch": 1.4, + "learning_rate": 1.7915225664832893e-05, + "loss": 1.0215, + "step": 18920 + }, + { + "epoch": 1.4, + "learning_rate": 1.791498189215837e-05, + "loss": 0.993, + "step": 18921 + }, + { + "epoch": 1.4, + "learning_rate": 1.791473810689122e-05, + "loss": 1.0174, + "step": 18922 + }, + { + "epoch": 1.4, + "learning_rate": 1.791449430903184e-05, + "loss": 1.0054, + "step": 18923 + }, + { + "epoch": 1.4, + "learning_rate": 1.7914250498580607e-05, + "loss": 1.0093, + "step": 18924 + }, + { + "epoch": 1.4, + "learning_rate": 1.7914006675537918e-05, + "loss": 1.0338, + "step": 18925 + }, + { + "epoch": 1.4, + "learning_rate": 1.791376283990416e-05, + "loss": 0.9595, + "step": 18926 + }, + { + "epoch": 1.4, + "learning_rate": 1.7913518991679713e-05, + "loss": 1.0064, + "step": 18927 + }, + { + "epoch": 1.4, + "learning_rate": 1.7913275130864976e-05, + "loss": 0.9654, + "step": 18928 + }, + { + "epoch": 1.4, + "learning_rate": 1.791303125746033e-05, + "loss": 1.0855, + "step": 18929 + }, + { + "epoch": 1.4, + "learning_rate": 1.791278737146616e-05, + "loss": 0.9722, + "step": 18930 + }, + { + "epoch": 1.4, + "learning_rate": 1.7912543472882864e-05, + "loss": 1.0116, + "step": 18931 + }, + { + "epoch": 1.4, + "learning_rate": 1.7912299561710822e-05, + "loss": 0.98, + "step": 18932 + }, + { + "epoch": 1.4, + "learning_rate": 1.7912055637950424e-05, + "loss": 1.0036, + "step": 18933 + }, + { + "epoch": 1.4, + "learning_rate": 1.791181170160206e-05, + "loss": 1.0026, + "step": 18934 + }, + { + "epoch": 1.4, + "learning_rate": 1.7911567752666117e-05, + "loss": 1.013, + "step": 18935 + }, + { + "epoch": 1.4, + "learning_rate": 1.791132379114298e-05, + "loss": 0.9554, + "step": 18936 + }, + { + "epoch": 1.4, + "learning_rate": 1.7911079817033047e-05, + "loss": 1.1166, + "step": 18937 + }, + { + "epoch": 1.4, + "learning_rate": 1.791083583033669e-05, + "loss": 0.993, + "step": 18938 + }, + { + "epoch": 1.4, + "learning_rate": 1.7910591831054314e-05, + "loss": 1.0171, + "step": 18939 + }, + { + "epoch": 1.4, + "learning_rate": 1.7910347819186297e-05, + "loss": 1.0319, + "step": 18940 + }, + { + "epoch": 1.4, + "learning_rate": 1.7910103794733028e-05, + "loss": 0.971, + "step": 18941 + }, + { + "epoch": 1.4, + "learning_rate": 1.79098597576949e-05, + "loss": 0.8603, + "step": 18942 + }, + { + "epoch": 1.4, + "learning_rate": 1.7909615708072296e-05, + "loss": 1.0268, + "step": 18943 + }, + { + "epoch": 1.4, + "learning_rate": 1.7909371645865606e-05, + "loss": 0.9859, + "step": 18944 + }, + { + "epoch": 1.4, + "learning_rate": 1.7909127571075222e-05, + "loss": 0.9643, + "step": 18945 + }, + { + "epoch": 1.4, + "learning_rate": 1.7908883483701526e-05, + "loss": 1.1587, + "step": 18946 + }, + { + "epoch": 1.4, + "learning_rate": 1.7908639383744912e-05, + "loss": 0.9836, + "step": 18947 + }, + { + "epoch": 1.4, + "learning_rate": 1.7908395271205766e-05, + "loss": 0.925, + "step": 18948 + }, + { + "epoch": 1.4, + "learning_rate": 1.7908151146084474e-05, + "loss": 1.0458, + "step": 18949 + }, + { + "epoch": 1.4, + "learning_rate": 1.7907907008381426e-05, + "loss": 1.0369, + "step": 18950 + }, + { + "epoch": 1.4, + "learning_rate": 1.7907662858097014e-05, + "loss": 1.0362, + "step": 18951 + }, + { + "epoch": 1.4, + "learning_rate": 1.7907418695231622e-05, + "loss": 1.0653, + "step": 18952 + }, + { + "epoch": 1.4, + "learning_rate": 1.7907174519785643e-05, + "loss": 0.9992, + "step": 18953 + }, + { + "epoch": 1.4, + "learning_rate": 1.7906930331759456e-05, + "loss": 0.9824, + "step": 18954 + }, + { + "epoch": 1.4, + "learning_rate": 1.7906686131153464e-05, + "loss": 0.9115, + "step": 18955 + }, + { + "epoch": 1.4, + "learning_rate": 1.7906441917968044e-05, + "loss": 0.9727, + "step": 18956 + }, + { + "epoch": 1.4, + "learning_rate": 1.7906197692203585e-05, + "loss": 1.0628, + "step": 18957 + }, + { + "epoch": 1.4, + "learning_rate": 1.790595345386048e-05, + "loss": 1.0161, + "step": 18958 + }, + { + "epoch": 1.4, + "learning_rate": 1.7905709202939118e-05, + "loss": 0.9509, + "step": 18959 + }, + { + "epoch": 1.4, + "learning_rate": 1.7905464939439886e-05, + "loss": 1.0772, + "step": 18960 + }, + { + "epoch": 1.4, + "learning_rate": 1.790522066336317e-05, + "loss": 0.9802, + "step": 18961 + }, + { + "epoch": 1.4, + "learning_rate": 1.7904976374709364e-05, + "loss": 1.0165, + "step": 18962 + }, + { + "epoch": 1.4, + "learning_rate": 1.7904732073478854e-05, + "loss": 0.9232, + "step": 18963 + }, + { + "epoch": 1.4, + "learning_rate": 1.790448775967202e-05, + "loss": 1.0122, + "step": 18964 + }, + { + "epoch": 1.4, + "learning_rate": 1.790424343328927e-05, + "loss": 0.9713, + "step": 18965 + }, + { + "epoch": 1.4, + "learning_rate": 1.7903999094330977e-05, + "loss": 1.0945, + "step": 18966 + }, + { + "epoch": 1.4, + "learning_rate": 1.7903754742797534e-05, + "loss": 1.0449, + "step": 18967 + }, + { + "epoch": 1.4, + "learning_rate": 1.790351037868933e-05, + "loss": 1.0162, + "step": 18968 + }, + { + "epoch": 1.4, + "learning_rate": 1.7903266002006755e-05, + "loss": 0.957, + "step": 18969 + }, + { + "epoch": 1.4, + "learning_rate": 1.79030216127502e-05, + "loss": 1.0698, + "step": 18970 + }, + { + "epoch": 1.4, + "learning_rate": 1.7902777210920045e-05, + "loss": 1.0856, + "step": 18971 + }, + { + "epoch": 1.4, + "learning_rate": 1.7902532796516686e-05, + "loss": 0.9902, + "step": 18972 + }, + { + "epoch": 1.4, + "learning_rate": 1.7902288369540516e-05, + "loss": 0.9723, + "step": 18973 + }, + { + "epoch": 1.4, + "learning_rate": 1.7902043929991913e-05, + "loss": 0.9739, + "step": 18974 + }, + { + "epoch": 1.4, + "learning_rate": 1.7901799477871268e-05, + "loss": 1.1433, + "step": 18975 + }, + { + "epoch": 1.4, + "learning_rate": 1.790155501317898e-05, + "loss": 1.071, + "step": 18976 + }, + { + "epoch": 1.4, + "learning_rate": 1.7901310535915426e-05, + "loss": 1.0168, + "step": 18977 + }, + { + "epoch": 1.4, + "learning_rate": 1.7901066046081002e-05, + "loss": 1.0505, + "step": 18978 + }, + { + "epoch": 1.4, + "learning_rate": 1.7900821543676097e-05, + "loss": 1.0255, + "step": 18979 + }, + { + "epoch": 1.4, + "learning_rate": 1.7900577028701095e-05, + "loss": 1.0107, + "step": 18980 + }, + { + "epoch": 1.4, + "learning_rate": 1.790033250115639e-05, + "loss": 0.947, + "step": 18981 + }, + { + "epoch": 1.4, + "learning_rate": 1.790008796104237e-05, + "loss": 1.1241, + "step": 18982 + }, + { + "epoch": 1.4, + "learning_rate": 1.7899843408359417e-05, + "loss": 1.018, + "step": 18983 + }, + { + "epoch": 1.4, + "learning_rate": 1.7899598843107934e-05, + "loss": 0.9757, + "step": 18984 + }, + { + "epoch": 1.4, + "learning_rate": 1.7899354265288294e-05, + "loss": 1.0006, + "step": 18985 + }, + { + "epoch": 1.4, + "learning_rate": 1.7899109674900905e-05, + "loss": 0.9173, + "step": 18986 + }, + { + "epoch": 1.4, + "learning_rate": 1.7898865071946138e-05, + "loss": 1.0039, + "step": 18987 + }, + { + "epoch": 1.4, + "learning_rate": 1.7898620456424388e-05, + "loss": 1.0433, + "step": 18988 + }, + { + "epoch": 1.4, + "learning_rate": 1.789837582833605e-05, + "loss": 1.0708, + "step": 18989 + }, + { + "epoch": 1.4, + "learning_rate": 1.789813118768151e-05, + "loss": 0.9945, + "step": 18990 + }, + { + "epoch": 1.4, + "learning_rate": 1.7897886534461153e-05, + "loss": 1.1056, + "step": 18991 + }, + { + "epoch": 1.4, + "learning_rate": 1.7897641868675373e-05, + "loss": 1.0002, + "step": 18992 + }, + { + "epoch": 1.4, + "learning_rate": 1.789739719032456e-05, + "loss": 0.945, + "step": 18993 + }, + { + "epoch": 1.4, + "learning_rate": 1.7897152499409095e-05, + "loss": 1.0715, + "step": 18994 + }, + { + "epoch": 1.4, + "learning_rate": 1.7896907795929377e-05, + "loss": 1.0254, + "step": 18995 + }, + { + "epoch": 1.4, + "learning_rate": 1.789666307988579e-05, + "loss": 0.9327, + "step": 18996 + }, + { + "epoch": 1.4, + "learning_rate": 1.7896418351278732e-05, + "loss": 1.0063, + "step": 18997 + }, + { + "epoch": 1.4, + "learning_rate": 1.789617361010858e-05, + "loss": 1.0279, + "step": 18998 + }, + { + "epoch": 1.4, + "learning_rate": 1.7895928856375727e-05, + "loss": 1.0818, + "step": 18999 + }, + { + "epoch": 1.4, + "learning_rate": 1.7895684090080567e-05, + "loss": 0.9422, + "step": 19000 + }, + { + "epoch": 1.4, + "learning_rate": 1.7895439311223484e-05, + "loss": 1.0992, + "step": 19001 + }, + { + "epoch": 1.4, + "learning_rate": 1.7895194519804873e-05, + "loss": 1.0932, + "step": 19002 + }, + { + "epoch": 1.4, + "learning_rate": 1.789494971582512e-05, + "loss": 1.0609, + "step": 19003 + }, + { + "epoch": 1.4, + "learning_rate": 1.7894704899284615e-05, + "loss": 0.9964, + "step": 19004 + }, + { + "epoch": 1.4, + "learning_rate": 1.7894460070183744e-05, + "loss": 1.0392, + "step": 19005 + }, + { + "epoch": 1.4, + "learning_rate": 1.7894215228522906e-05, + "loss": 1.0565, + "step": 19006 + }, + { + "epoch": 1.4, + "learning_rate": 1.789397037430248e-05, + "loss": 1.0363, + "step": 19007 + }, + { + "epoch": 1.4, + "learning_rate": 1.789372550752286e-05, + "loss": 1.1598, + "step": 19008 + }, + { + "epoch": 1.4, + "learning_rate": 1.7893480628184438e-05, + "loss": 1.1132, + "step": 19009 + }, + { + "epoch": 1.4, + "learning_rate": 1.78932357362876e-05, + "loss": 0.885, + "step": 19010 + }, + { + "epoch": 1.4, + "learning_rate": 1.7892990831832737e-05, + "loss": 0.9304, + "step": 19011 + }, + { + "epoch": 1.4, + "learning_rate": 1.7892745914820237e-05, + "loss": 0.9868, + "step": 19012 + }, + { + "epoch": 1.4, + "learning_rate": 1.7892500985250495e-05, + "loss": 1.0936, + "step": 19013 + }, + { + "epoch": 1.4, + "learning_rate": 1.7892256043123892e-05, + "loss": 1.0247, + "step": 19014 + }, + { + "epoch": 1.41, + "learning_rate": 1.7892011088440825e-05, + "loss": 1.0822, + "step": 19015 + }, + { + "epoch": 1.41, + "learning_rate": 1.7891766121201684e-05, + "loss": 0.9109, + "step": 19016 + }, + { + "epoch": 1.41, + "learning_rate": 1.7891521141406852e-05, + "loss": 0.9416, + "step": 19017 + }, + { + "epoch": 1.41, + "learning_rate": 1.7891276149056725e-05, + "loss": 0.931, + "step": 19018 + }, + { + "epoch": 1.41, + "learning_rate": 1.789103114415169e-05, + "loss": 0.9316, + "step": 19019 + }, + { + "epoch": 1.41, + "learning_rate": 1.7890786126692137e-05, + "loss": 1.0184, + "step": 19020 + }, + { + "epoch": 1.41, + "learning_rate": 1.7890541096678457e-05, + "loss": 1.0075, + "step": 19021 + }, + { + "epoch": 1.41, + "learning_rate": 1.789029605411104e-05, + "loss": 1.0578, + "step": 19022 + }, + { + "epoch": 1.41, + "learning_rate": 1.7890050998990268e-05, + "loss": 1.0725, + "step": 19023 + }, + { + "epoch": 1.41, + "learning_rate": 1.7889805931316547e-05, + "loss": 1.0118, + "step": 19024 + }, + { + "epoch": 1.41, + "learning_rate": 1.788956085109025e-05, + "loss": 1.0462, + "step": 19025 + }, + { + "epoch": 1.41, + "learning_rate": 1.7889315758311778e-05, + "loss": 0.962, + "step": 19026 + }, + { + "epoch": 1.41, + "learning_rate": 1.788907065298152e-05, + "loss": 1.123, + "step": 19027 + }, + { + "epoch": 1.41, + "learning_rate": 1.788882553509986e-05, + "loss": 1.0621, + "step": 19028 + }, + { + "epoch": 1.41, + "learning_rate": 1.7888580404667194e-05, + "loss": 1.0775, + "step": 19029 + }, + { + "epoch": 1.41, + "learning_rate": 1.7888335261683906e-05, + "loss": 1.0923, + "step": 19030 + }, + { + "epoch": 1.41, + "learning_rate": 1.7888090106150394e-05, + "loss": 0.994, + "step": 19031 + }, + { + "epoch": 1.41, + "learning_rate": 1.788784493806704e-05, + "loss": 1.1375, + "step": 19032 + }, + { + "epoch": 1.41, + "learning_rate": 1.7887599757434238e-05, + "loss": 1.0601, + "step": 19033 + }, + { + "epoch": 1.41, + "learning_rate": 1.788735456425238e-05, + "loss": 0.9204, + "step": 19034 + }, + { + "epoch": 1.41, + "learning_rate": 1.7887109358521848e-05, + "loss": 1.0187, + "step": 19035 + }, + { + "epoch": 1.41, + "learning_rate": 1.7886864140243043e-05, + "loss": 0.9539, + "step": 19036 + }, + { + "epoch": 1.41, + "learning_rate": 1.7886618909416346e-05, + "loss": 1.0127, + "step": 19037 + }, + { + "epoch": 1.41, + "learning_rate": 1.7886373666042157e-05, + "loss": 0.9895, + "step": 19038 + }, + { + "epoch": 1.41, + "learning_rate": 1.7886128410120856e-05, + "loss": 1.0395, + "step": 19039 + }, + { + "epoch": 1.41, + "learning_rate": 1.788588314165284e-05, + "loss": 1.0325, + "step": 19040 + }, + { + "epoch": 1.41, + "learning_rate": 1.7885637860638497e-05, + "loss": 0.9376, + "step": 19041 + }, + { + "epoch": 1.41, + "learning_rate": 1.7885392567078216e-05, + "loss": 0.9861, + "step": 19042 + }, + { + "epoch": 1.41, + "learning_rate": 1.7885147260972384e-05, + "loss": 1.037, + "step": 19043 + }, + { + "epoch": 1.41, + "learning_rate": 1.78849019423214e-05, + "loss": 0.9781, + "step": 19044 + }, + { + "epoch": 1.41, + "learning_rate": 1.7884656611125647e-05, + "loss": 1.1363, + "step": 19045 + }, + { + "epoch": 1.41, + "learning_rate": 1.7884411267385524e-05, + "loss": 0.9239, + "step": 19046 + }, + { + "epoch": 1.41, + "learning_rate": 1.788416591110141e-05, + "loss": 0.8748, + "step": 19047 + }, + { + "epoch": 1.41, + "learning_rate": 1.7883920542273703e-05, + "loss": 0.9841, + "step": 19048 + }, + { + "epoch": 1.41, + "learning_rate": 1.788367516090279e-05, + "loss": 0.9153, + "step": 19049 + }, + { + "epoch": 1.41, + "learning_rate": 1.7883429766989064e-05, + "loss": 0.9815, + "step": 19050 + }, + { + "epoch": 1.41, + "learning_rate": 1.788318436053291e-05, + "loss": 0.9476, + "step": 19051 + }, + { + "epoch": 1.41, + "learning_rate": 1.7882938941534728e-05, + "loss": 1.0888, + "step": 19052 + }, + { + "epoch": 1.41, + "learning_rate": 1.78826935099949e-05, + "loss": 1.0697, + "step": 19053 + }, + { + "epoch": 1.41, + "learning_rate": 1.7882448065913822e-05, + "loss": 1.0271, + "step": 19054 + }, + { + "epoch": 1.41, + "learning_rate": 1.7882202609291882e-05, + "loss": 0.9857, + "step": 19055 + }, + { + "epoch": 1.41, + "learning_rate": 1.7881957140129466e-05, + "loss": 0.9682, + "step": 19056 + }, + { + "epoch": 1.41, + "learning_rate": 1.7881711658426972e-05, + "loss": 0.9965, + "step": 19057 + }, + { + "epoch": 1.41, + "learning_rate": 1.788146616418479e-05, + "loss": 1.1084, + "step": 19058 + }, + { + "epoch": 1.41, + "learning_rate": 1.7881220657403303e-05, + "loss": 0.9479, + "step": 19059 + }, + { + "epoch": 1.41, + "learning_rate": 1.7880975138082907e-05, + "loss": 1.0618, + "step": 19060 + }, + { + "epoch": 1.41, + "learning_rate": 1.7880729606223997e-05, + "loss": 0.9321, + "step": 19061 + }, + { + "epoch": 1.41, + "learning_rate": 1.7880484061826956e-05, + "loss": 1.0263, + "step": 19062 + }, + { + "epoch": 1.41, + "learning_rate": 1.788023850489218e-05, + "loss": 1.0732, + "step": 19063 + }, + { + "epoch": 1.41, + "learning_rate": 1.7879992935420052e-05, + "loss": 0.9739, + "step": 19064 + }, + { + "epoch": 1.41, + "learning_rate": 1.787974735341097e-05, + "loss": 0.909, + "step": 19065 + }, + { + "epoch": 1.41, + "learning_rate": 1.7879501758865325e-05, + "loss": 1.0826, + "step": 19066 + }, + { + "epoch": 1.41, + "learning_rate": 1.7879256151783503e-05, + "loss": 0.894, + "step": 19067 + }, + { + "epoch": 1.41, + "learning_rate": 1.7879010532165898e-05, + "loss": 0.9702, + "step": 19068 + }, + { + "epoch": 1.41, + "learning_rate": 1.78787649000129e-05, + "loss": 1.1145, + "step": 19069 + }, + { + "epoch": 1.41, + "learning_rate": 1.7878519255324902e-05, + "loss": 1.0797, + "step": 19070 + }, + { + "epoch": 1.41, + "learning_rate": 1.787827359810229e-05, + "loss": 1.0105, + "step": 19071 + }, + { + "epoch": 1.41, + "learning_rate": 1.787802792834546e-05, + "loss": 1.0187, + "step": 19072 + }, + { + "epoch": 1.41, + "learning_rate": 1.7877782246054795e-05, + "loss": 1.0414, + "step": 19073 + }, + { + "epoch": 1.41, + "learning_rate": 1.7877536551230697e-05, + "loss": 0.9727, + "step": 19074 + }, + { + "epoch": 1.41, + "learning_rate": 1.7877290843873547e-05, + "loss": 1.0943, + "step": 19075 + }, + { + "epoch": 1.41, + "learning_rate": 1.7877045123983744e-05, + "loss": 0.9339, + "step": 19076 + }, + { + "epoch": 1.41, + "learning_rate": 1.787679939156167e-05, + "loss": 0.9925, + "step": 19077 + }, + { + "epoch": 1.41, + "learning_rate": 1.7876553646607722e-05, + "loss": 1.0717, + "step": 19078 + }, + { + "epoch": 1.41, + "learning_rate": 1.787630788912229e-05, + "loss": 1.1072, + "step": 19079 + }, + { + "epoch": 1.41, + "learning_rate": 1.7876062119105767e-05, + "loss": 1.0262, + "step": 19080 + }, + { + "epoch": 1.41, + "learning_rate": 1.7875816336558543e-05, + "loss": 1.0331, + "step": 19081 + }, + { + "epoch": 1.41, + "learning_rate": 1.7875570541481004e-05, + "loss": 1.066, + "step": 19082 + }, + { + "epoch": 1.41, + "learning_rate": 1.7875324733873547e-05, + "loss": 0.9437, + "step": 19083 + }, + { + "epoch": 1.41, + "learning_rate": 1.7875078913736562e-05, + "loss": 0.9884, + "step": 19084 + }, + { + "epoch": 1.41, + "learning_rate": 1.7874833081070435e-05, + "loss": 0.9216, + "step": 19085 + }, + { + "epoch": 1.41, + "learning_rate": 1.7874587235875566e-05, + "loss": 1.0232, + "step": 19086 + }, + { + "epoch": 1.41, + "learning_rate": 1.7874341378152338e-05, + "loss": 1.0779, + "step": 19087 + }, + { + "epoch": 1.41, + "learning_rate": 1.7874095507901147e-05, + "loss": 1.0802, + "step": 19088 + }, + { + "epoch": 1.41, + "learning_rate": 1.7873849625122382e-05, + "loss": 0.9955, + "step": 19089 + }, + { + "epoch": 1.41, + "learning_rate": 1.7873603729816434e-05, + "loss": 0.9708, + "step": 19090 + }, + { + "epoch": 1.41, + "learning_rate": 1.78733578219837e-05, + "loss": 1.0486, + "step": 19091 + }, + { + "epoch": 1.41, + "learning_rate": 1.7873111901624563e-05, + "loss": 0.9805, + "step": 19092 + }, + { + "epoch": 1.41, + "learning_rate": 1.787286596873942e-05, + "loss": 0.9656, + "step": 19093 + }, + { + "epoch": 1.41, + "learning_rate": 1.7872620023328657e-05, + "loss": 1.0546, + "step": 19094 + }, + { + "epoch": 1.41, + "learning_rate": 1.7872374065392665e-05, + "loss": 0.8744, + "step": 19095 + }, + { + "epoch": 1.41, + "learning_rate": 1.7872128094931846e-05, + "loss": 1.0428, + "step": 19096 + }, + { + "epoch": 1.41, + "learning_rate": 1.787188211194658e-05, + "loss": 0.9835, + "step": 19097 + }, + { + "epoch": 1.41, + "learning_rate": 1.787163611643726e-05, + "loss": 1.0221, + "step": 19098 + }, + { + "epoch": 1.41, + "learning_rate": 1.7871390108404282e-05, + "loss": 1.0022, + "step": 19099 + }, + { + "epoch": 1.41, + "learning_rate": 1.7871144087848033e-05, + "loss": 1.0335, + "step": 19100 + }, + { + "epoch": 1.41, + "learning_rate": 1.787089805476891e-05, + "loss": 0.8311, + "step": 19101 + }, + { + "epoch": 1.41, + "learning_rate": 1.78706520091673e-05, + "loss": 1.0339, + "step": 19102 + }, + { + "epoch": 1.41, + "learning_rate": 1.7870405951043594e-05, + "loss": 0.966, + "step": 19103 + }, + { + "epoch": 1.41, + "learning_rate": 1.7870159880398183e-05, + "loss": 1.0978, + "step": 19104 + }, + { + "epoch": 1.41, + "learning_rate": 1.7869913797231464e-05, + "loss": 1.0542, + "step": 19105 + }, + { + "epoch": 1.41, + "learning_rate": 1.7869667701543823e-05, + "loss": 1.0595, + "step": 19106 + }, + { + "epoch": 1.41, + "learning_rate": 1.7869421593335656e-05, + "loss": 0.9152, + "step": 19107 + }, + { + "epoch": 1.41, + "learning_rate": 1.7869175472607345e-05, + "loss": 1.008, + "step": 19108 + }, + { + "epoch": 1.41, + "learning_rate": 1.7868929339359292e-05, + "loss": 0.9083, + "step": 19109 + }, + { + "epoch": 1.41, + "learning_rate": 1.7868683193591888e-05, + "loss": 1.0627, + "step": 19110 + }, + { + "epoch": 1.41, + "learning_rate": 1.786843703530552e-05, + "loss": 1.0277, + "step": 19111 + }, + { + "epoch": 1.41, + "learning_rate": 1.7868190864500578e-05, + "loss": 1.0445, + "step": 19112 + }, + { + "epoch": 1.41, + "learning_rate": 1.786794468117746e-05, + "loss": 1.0182, + "step": 19113 + }, + { + "epoch": 1.41, + "learning_rate": 1.7867698485336553e-05, + "loss": 0.9927, + "step": 19114 + }, + { + "epoch": 1.41, + "learning_rate": 1.7867452276978252e-05, + "loss": 1.0153, + "step": 19115 + }, + { + "epoch": 1.41, + "learning_rate": 1.7867206056102946e-05, + "loss": 0.9211, + "step": 19116 + }, + { + "epoch": 1.41, + "learning_rate": 1.7866959822711028e-05, + "loss": 0.9281, + "step": 19117 + }, + { + "epoch": 1.41, + "learning_rate": 1.7866713576802886e-05, + "loss": 1.063, + "step": 19118 + }, + { + "epoch": 1.41, + "learning_rate": 1.786646731837892e-05, + "loss": 0.9612, + "step": 19119 + }, + { + "epoch": 1.41, + "learning_rate": 1.786622104743952e-05, + "loss": 0.992, + "step": 19120 + }, + { + "epoch": 1.41, + "learning_rate": 1.7865974763985067e-05, + "loss": 0.9515, + "step": 19121 + }, + { + "epoch": 1.41, + "learning_rate": 1.7865728468015967e-05, + "loss": 1.0175, + "step": 19122 + }, + { + "epoch": 1.41, + "learning_rate": 1.7865482159532602e-05, + "loss": 1.0152, + "step": 19123 + }, + { + "epoch": 1.41, + "learning_rate": 1.7865235838535366e-05, + "loss": 0.9866, + "step": 19124 + }, + { + "epoch": 1.41, + "learning_rate": 1.7864989505024658e-05, + "loss": 1.0485, + "step": 19125 + }, + { + "epoch": 1.41, + "learning_rate": 1.786474315900086e-05, + "loss": 0.9973, + "step": 19126 + }, + { + "epoch": 1.41, + "learning_rate": 1.7864496800464367e-05, + "loss": 0.9592, + "step": 19127 + }, + { + "epoch": 1.41, + "learning_rate": 1.7864250429415575e-05, + "loss": 1.0535, + "step": 19128 + }, + { + "epoch": 1.41, + "learning_rate": 1.786400404585487e-05, + "loss": 1.0607, + "step": 19129 + }, + { + "epoch": 1.41, + "learning_rate": 1.786375764978265e-05, + "loss": 1.0687, + "step": 19130 + }, + { + "epoch": 1.41, + "learning_rate": 1.7863511241199302e-05, + "loss": 1.0288, + "step": 19131 + }, + { + "epoch": 1.41, + "learning_rate": 1.786326482010522e-05, + "loss": 1.0, + "step": 19132 + }, + { + "epoch": 1.41, + "learning_rate": 1.7863018386500803e-05, + "loss": 0.8836, + "step": 19133 + }, + { + "epoch": 1.41, + "learning_rate": 1.786277194038643e-05, + "loss": 0.9638, + "step": 19134 + }, + { + "epoch": 1.41, + "learning_rate": 1.7862525481762497e-05, + "loss": 1.1441, + "step": 19135 + }, + { + "epoch": 1.41, + "learning_rate": 1.7862279010629402e-05, + "loss": 1.082, + "step": 19136 + }, + { + "epoch": 1.41, + "learning_rate": 1.7862032526987532e-05, + "loss": 0.9983, + "step": 19137 + }, + { + "epoch": 1.41, + "learning_rate": 1.7861786030837285e-05, + "loss": 1.0058, + "step": 19138 + }, + { + "epoch": 1.41, + "learning_rate": 1.7861539522179042e-05, + "loss": 0.9845, + "step": 19139 + }, + { + "epoch": 1.41, + "learning_rate": 1.7861293001013206e-05, + "loss": 1.0, + "step": 19140 + }, + { + "epoch": 1.41, + "learning_rate": 1.7861046467340166e-05, + "loss": 1.0325, + "step": 19141 + }, + { + "epoch": 1.41, + "learning_rate": 1.786079992116031e-05, + "loss": 1.0213, + "step": 19142 + }, + { + "epoch": 1.41, + "learning_rate": 1.786055336247404e-05, + "loss": 1.0112, + "step": 19143 + }, + { + "epoch": 1.41, + "learning_rate": 1.7860306791281742e-05, + "loss": 1.0308, + "step": 19144 + }, + { + "epoch": 1.41, + "learning_rate": 1.7860060207583804e-05, + "loss": 0.9516, + "step": 19145 + }, + { + "epoch": 1.41, + "learning_rate": 1.7859813611380623e-05, + "loss": 0.9566, + "step": 19146 + }, + { + "epoch": 1.41, + "learning_rate": 1.785956700267259e-05, + "loss": 0.9849, + "step": 19147 + }, + { + "epoch": 1.41, + "learning_rate": 1.78593203814601e-05, + "loss": 0.9097, + "step": 19148 + }, + { + "epoch": 1.41, + "learning_rate": 1.785907374774354e-05, + "loss": 1.0241, + "step": 19149 + }, + { + "epoch": 1.41, + "learning_rate": 1.7858827101523312e-05, + "loss": 1.0308, + "step": 19150 + }, + { + "epoch": 1.42, + "learning_rate": 1.7858580442799803e-05, + "loss": 1.084, + "step": 19151 + }, + { + "epoch": 1.42, + "learning_rate": 1.7858333771573403e-05, + "loss": 0.9934, + "step": 19152 + }, + { + "epoch": 1.42, + "learning_rate": 1.7858087087844505e-05, + "loss": 1.0217, + "step": 19153 + }, + { + "epoch": 1.42, + "learning_rate": 1.7857840391613503e-05, + "loss": 0.9841, + "step": 19154 + }, + { + "epoch": 1.42, + "learning_rate": 1.785759368288079e-05, + "loss": 0.9387, + "step": 19155 + }, + { + "epoch": 1.42, + "learning_rate": 1.785734696164676e-05, + "loss": 1.0828, + "step": 19156 + }, + { + "epoch": 1.42, + "learning_rate": 1.7857100227911798e-05, + "loss": 1.0167, + "step": 19157 + }, + { + "epoch": 1.42, + "learning_rate": 1.7856853481676307e-05, + "loss": 1.0108, + "step": 19158 + }, + { + "epoch": 1.42, + "learning_rate": 1.7856606722940673e-05, + "loss": 0.9805, + "step": 19159 + }, + { + "epoch": 1.42, + "learning_rate": 1.785635995170529e-05, + "loss": 0.9557, + "step": 19160 + }, + { + "epoch": 1.42, + "learning_rate": 1.785611316797055e-05, + "loss": 1.1152, + "step": 19161 + }, + { + "epoch": 1.42, + "learning_rate": 1.7855866371736845e-05, + "loss": 1.023, + "step": 19162 + }, + { + "epoch": 1.42, + "learning_rate": 1.7855619563004573e-05, + "loss": 0.9681, + "step": 19163 + }, + { + "epoch": 1.42, + "learning_rate": 1.785537274177412e-05, + "loss": 0.9211, + "step": 19164 + }, + { + "epoch": 1.42, + "learning_rate": 1.7855125908045882e-05, + "loss": 1.0985, + "step": 19165 + }, + { + "epoch": 1.42, + "learning_rate": 1.785487906182025e-05, + "loss": 1.0598, + "step": 19166 + }, + { + "epoch": 1.42, + "learning_rate": 1.785463220309762e-05, + "loss": 1.138, + "step": 19167 + }, + { + "epoch": 1.42, + "learning_rate": 1.785438533187838e-05, + "loss": 1.0577, + "step": 19168 + }, + { + "epoch": 1.42, + "learning_rate": 1.7854138448162926e-05, + "loss": 1.0737, + "step": 19169 + }, + { + "epoch": 1.42, + "learning_rate": 1.7853891551951653e-05, + "loss": 1.0561, + "step": 19170 + }, + { + "epoch": 1.42, + "learning_rate": 1.785364464324495e-05, + "loss": 1.0026, + "step": 19171 + }, + { + "epoch": 1.42, + "learning_rate": 1.7853397722043206e-05, + "loss": 1.0167, + "step": 19172 + }, + { + "epoch": 1.42, + "learning_rate": 1.7853150788346822e-05, + "loss": 0.9794, + "step": 19173 + }, + { + "epoch": 1.42, + "learning_rate": 1.7852903842156187e-05, + "loss": 1.0221, + "step": 19174 + }, + { + "epoch": 1.42, + "learning_rate": 1.7852656883471694e-05, + "loss": 0.9765, + "step": 19175 + }, + { + "epoch": 1.42, + "learning_rate": 1.785240991229374e-05, + "loss": 0.9691, + "step": 19176 + }, + { + "epoch": 1.42, + "learning_rate": 1.7852162928622713e-05, + "loss": 0.9729, + "step": 19177 + }, + { + "epoch": 1.42, + "learning_rate": 1.7851915932459004e-05, + "loss": 1.0891, + "step": 19178 + }, + { + "epoch": 1.42, + "learning_rate": 1.785166892380301e-05, + "loss": 0.9673, + "step": 19179 + }, + { + "epoch": 1.42, + "learning_rate": 1.7851421902655125e-05, + "loss": 1.0219, + "step": 19180 + }, + { + "epoch": 1.42, + "learning_rate": 1.7851174869015734e-05, + "loss": 1.0477, + "step": 19181 + }, + { + "epoch": 1.42, + "learning_rate": 1.7850927822885243e-05, + "loss": 0.9799, + "step": 19182 + }, + { + "epoch": 1.42, + "learning_rate": 1.7850680764264036e-05, + "loss": 1.0415, + "step": 19183 + }, + { + "epoch": 1.42, + "learning_rate": 1.785043369315251e-05, + "loss": 1.0914, + "step": 19184 + }, + { + "epoch": 1.42, + "learning_rate": 1.785018660955105e-05, + "loss": 1.0936, + "step": 19185 + }, + { + "epoch": 1.42, + "learning_rate": 1.7849939513460063e-05, + "loss": 1.0217, + "step": 19186 + }, + { + "epoch": 1.42, + "learning_rate": 1.784969240487993e-05, + "loss": 0.983, + "step": 19187 + }, + { + "epoch": 1.42, + "learning_rate": 1.7849445283811052e-05, + "loss": 1.0222, + "step": 19188 + }, + { + "epoch": 1.42, + "learning_rate": 1.7849198150253816e-05, + "loss": 1.0838, + "step": 19189 + }, + { + "epoch": 1.42, + "learning_rate": 1.784895100420862e-05, + "loss": 0.9935, + "step": 19190 + }, + { + "epoch": 1.42, + "learning_rate": 1.7848703845675854e-05, + "loss": 1.0132, + "step": 19191 + }, + { + "epoch": 1.42, + "learning_rate": 1.784845667465591e-05, + "loss": 0.9551, + "step": 19192 + }, + { + "epoch": 1.42, + "learning_rate": 1.7848209491149186e-05, + "loss": 1.0531, + "step": 19193 + }, + { + "epoch": 1.42, + "learning_rate": 1.7847962295156076e-05, + "loss": 0.9978, + "step": 19194 + }, + { + "epoch": 1.42, + "learning_rate": 1.7847715086676966e-05, + "loss": 0.9958, + "step": 19195 + }, + { + "epoch": 1.42, + "learning_rate": 1.7847467865712258e-05, + "loss": 1.0232, + "step": 19196 + }, + { + "epoch": 1.42, + "learning_rate": 1.7847220632262335e-05, + "loss": 0.8904, + "step": 19197 + }, + { + "epoch": 1.42, + "learning_rate": 1.78469733863276e-05, + "loss": 0.8993, + "step": 19198 + }, + { + "epoch": 1.42, + "learning_rate": 1.7846726127908444e-05, + "loss": 1.1251, + "step": 19199 + }, + { + "epoch": 1.42, + "learning_rate": 1.7846478857005255e-05, + "loss": 0.9651, + "step": 19200 + }, + { + "epoch": 1.42, + "learning_rate": 1.7846231573618434e-05, + "loss": 0.9428, + "step": 19201 + }, + { + "epoch": 1.42, + "learning_rate": 1.784598427774837e-05, + "loss": 1.1128, + "step": 19202 + }, + { + "epoch": 1.42, + "learning_rate": 1.7845736969395455e-05, + "loss": 1.0083, + "step": 19203 + }, + { + "epoch": 1.42, + "learning_rate": 1.7845489648560087e-05, + "loss": 0.95, + "step": 19204 + }, + { + "epoch": 1.42, + "learning_rate": 1.7845242315242656e-05, + "loss": 0.9358, + "step": 19205 + }, + { + "epoch": 1.42, + "learning_rate": 1.784499496944356e-05, + "loss": 1.0672, + "step": 19206 + }, + { + "epoch": 1.42, + "learning_rate": 1.7844747611163183e-05, + "loss": 1.1117, + "step": 19207 + }, + { + "epoch": 1.42, + "learning_rate": 1.7844500240401928e-05, + "loss": 0.9887, + "step": 19208 + }, + { + "epoch": 1.42, + "learning_rate": 1.7844252857160188e-05, + "loss": 1.0422, + "step": 19209 + }, + { + "epoch": 1.42, + "learning_rate": 1.784400546143835e-05, + "loss": 0.9934, + "step": 19210 + }, + { + "epoch": 1.42, + "learning_rate": 1.7843758053236816e-05, + "loss": 1.0978, + "step": 19211 + }, + { + "epoch": 1.42, + "learning_rate": 1.784351063255597e-05, + "loss": 0.9886, + "step": 19212 + }, + { + "epoch": 1.42, + "learning_rate": 1.7843263199396212e-05, + "loss": 0.9612, + "step": 19213 + }, + { + "epoch": 1.42, + "learning_rate": 1.7843015753757936e-05, + "loss": 0.9967, + "step": 19214 + }, + { + "epoch": 1.42, + "learning_rate": 1.7842768295641536e-05, + "loss": 1.0049, + "step": 19215 + }, + { + "epoch": 1.42, + "learning_rate": 1.78425208250474e-05, + "loss": 1.0116, + "step": 19216 + }, + { + "epoch": 1.42, + "learning_rate": 1.7842273341975925e-05, + "loss": 0.8901, + "step": 19217 + }, + { + "epoch": 1.42, + "learning_rate": 1.784202584642751e-05, + "loss": 1.0639, + "step": 19218 + }, + { + "epoch": 1.42, + "learning_rate": 1.784177833840254e-05, + "loss": 0.9694, + "step": 19219 + }, + { + "epoch": 1.42, + "learning_rate": 1.7841530817901415e-05, + "loss": 1.0411, + "step": 19220 + }, + { + "epoch": 1.42, + "learning_rate": 1.7841283284924528e-05, + "loss": 0.9718, + "step": 19221 + }, + { + "epoch": 1.42, + "learning_rate": 1.784103573947227e-05, + "loss": 0.9786, + "step": 19222 + }, + { + "epoch": 1.42, + "learning_rate": 1.7840788181545033e-05, + "loss": 1.0227, + "step": 19223 + }, + { + "epoch": 1.42, + "learning_rate": 1.784054061114322e-05, + "loss": 0.9712, + "step": 19224 + }, + { + "epoch": 1.42, + "learning_rate": 1.7840293028267214e-05, + "loss": 1.1118, + "step": 19225 + }, + { + "epoch": 1.42, + "learning_rate": 1.7840045432917416e-05, + "loss": 0.9559, + "step": 19226 + }, + { + "epoch": 1.42, + "learning_rate": 1.7839797825094218e-05, + "loss": 0.9979, + "step": 19227 + }, + { + "epoch": 1.42, + "learning_rate": 1.7839550204798012e-05, + "loss": 1.0352, + "step": 19228 + }, + { + "epoch": 1.42, + "learning_rate": 1.78393025720292e-05, + "loss": 1.0294, + "step": 19229 + }, + { + "epoch": 1.42, + "learning_rate": 1.7839054926788164e-05, + "loss": 0.8525, + "step": 19230 + }, + { + "epoch": 1.42, + "learning_rate": 1.7838807269075305e-05, + "loss": 1.0292, + "step": 19231 + }, + { + "epoch": 1.42, + "learning_rate": 1.7838559598891013e-05, + "loss": 0.9996, + "step": 19232 + }, + { + "epoch": 1.42, + "learning_rate": 1.783831191623569e-05, + "loss": 0.9319, + "step": 19233 + }, + { + "epoch": 1.42, + "learning_rate": 1.7838064221109724e-05, + "loss": 1.0827, + "step": 19234 + }, + { + "epoch": 1.42, + "learning_rate": 1.7837816513513505e-05, + "loss": 1.1, + "step": 19235 + }, + { + "epoch": 1.42, + "learning_rate": 1.7837568793447437e-05, + "loss": 1.1003, + "step": 19236 + }, + { + "epoch": 1.42, + "learning_rate": 1.7837321060911908e-05, + "loss": 1.0426, + "step": 19237 + }, + { + "epoch": 1.42, + "learning_rate": 1.7837073315907313e-05, + "loss": 0.8908, + "step": 19238 + }, + { + "epoch": 1.42, + "learning_rate": 1.7836825558434047e-05, + "loss": 1.0395, + "step": 19239 + }, + { + "epoch": 1.42, + "learning_rate": 1.7836577788492503e-05, + "loss": 0.9931, + "step": 19240 + }, + { + "epoch": 1.42, + "learning_rate": 1.7836330006083078e-05, + "loss": 1.0864, + "step": 19241 + }, + { + "epoch": 1.42, + "learning_rate": 1.7836082211206162e-05, + "loss": 1.1016, + "step": 19242 + }, + { + "epoch": 1.42, + "learning_rate": 1.7835834403862147e-05, + "loss": 0.9658, + "step": 19243 + }, + { + "epoch": 1.42, + "learning_rate": 1.783558658405144e-05, + "loss": 1.0824, + "step": 19244 + }, + { + "epoch": 1.42, + "learning_rate": 1.783533875177442e-05, + "loss": 1.1153, + "step": 19245 + }, + { + "epoch": 1.42, + "learning_rate": 1.7835090907031493e-05, + "loss": 1.0808, + "step": 19246 + }, + { + "epoch": 1.42, + "learning_rate": 1.783484304982305e-05, + "loss": 0.9393, + "step": 19247 + }, + { + "epoch": 1.42, + "learning_rate": 1.7834595180149478e-05, + "loss": 1.0607, + "step": 19248 + }, + { + "epoch": 1.42, + "learning_rate": 1.783434729801118e-05, + "loss": 1.0228, + "step": 19249 + }, + { + "epoch": 1.42, + "learning_rate": 1.7834099403408548e-05, + "loss": 0.925, + "step": 19250 + }, + { + "epoch": 1.42, + "learning_rate": 1.7833851496341972e-05, + "loss": 1.1043, + "step": 19251 + }, + { + "epoch": 1.42, + "learning_rate": 1.7833603576811856e-05, + "loss": 0.9552, + "step": 19252 + }, + { + "epoch": 1.42, + "learning_rate": 1.7833355644818584e-05, + "loss": 1.0777, + "step": 19253 + }, + { + "epoch": 1.42, + "learning_rate": 1.7833107700362558e-05, + "loss": 1.118, + "step": 19254 + }, + { + "epoch": 1.42, + "learning_rate": 1.783285974344417e-05, + "loss": 1.0368, + "step": 19255 + }, + { + "epoch": 1.42, + "learning_rate": 1.7832611774063813e-05, + "loss": 0.9598, + "step": 19256 + }, + { + "epoch": 1.42, + "learning_rate": 1.7832363792221887e-05, + "loss": 0.9258, + "step": 19257 + }, + { + "epoch": 1.42, + "learning_rate": 1.7832115797918774e-05, + "loss": 1.0441, + "step": 19258 + }, + { + "epoch": 1.42, + "learning_rate": 1.7831867791154883e-05, + "loss": 1.137, + "step": 19259 + }, + { + "epoch": 1.42, + "learning_rate": 1.78316197719306e-05, + "loss": 0.9963, + "step": 19260 + }, + { + "epoch": 1.42, + "learning_rate": 1.7831371740246322e-05, + "loss": 1.0011, + "step": 19261 + }, + { + "epoch": 1.42, + "learning_rate": 1.7831123696102443e-05, + "loss": 0.9527, + "step": 19262 + }, + { + "epoch": 1.42, + "learning_rate": 1.7830875639499362e-05, + "loss": 1.0456, + "step": 19263 + }, + { + "epoch": 1.42, + "learning_rate": 1.7830627570437468e-05, + "loss": 0.9143, + "step": 19264 + }, + { + "epoch": 1.42, + "learning_rate": 1.783037948891716e-05, + "loss": 1.0467, + "step": 19265 + }, + { + "epoch": 1.42, + "learning_rate": 1.7830131394938824e-05, + "loss": 0.9994, + "step": 19266 + }, + { + "epoch": 1.42, + "learning_rate": 1.782988328850286e-05, + "loss": 0.9833, + "step": 19267 + }, + { + "epoch": 1.42, + "learning_rate": 1.7829635169609673e-05, + "loss": 1.0445, + "step": 19268 + }, + { + "epoch": 1.42, + "learning_rate": 1.782938703825964e-05, + "loss": 1.1291, + "step": 19269 + }, + { + "epoch": 1.42, + "learning_rate": 1.7829138894453168e-05, + "loss": 1.2252, + "step": 19270 + }, + { + "epoch": 1.42, + "learning_rate": 1.782889073819065e-05, + "loss": 1.1453, + "step": 19271 + }, + { + "epoch": 1.42, + "learning_rate": 1.7828642569472477e-05, + "loss": 1.0652, + "step": 19272 + }, + { + "epoch": 1.42, + "learning_rate": 1.7828394388299043e-05, + "loss": 1.0677, + "step": 19273 + }, + { + "epoch": 1.42, + "learning_rate": 1.782814619467075e-05, + "loss": 0.9623, + "step": 19274 + }, + { + "epoch": 1.42, + "learning_rate": 1.7827897988587986e-05, + "loss": 1.0474, + "step": 19275 + }, + { + "epoch": 1.42, + "learning_rate": 1.7827649770051147e-05, + "loss": 0.984, + "step": 19276 + }, + { + "epoch": 1.42, + "learning_rate": 1.7827401539060632e-05, + "loss": 0.9515, + "step": 19277 + }, + { + "epoch": 1.42, + "learning_rate": 1.782715329561683e-05, + "loss": 0.9343, + "step": 19278 + }, + { + "epoch": 1.42, + "learning_rate": 1.782690503972014e-05, + "loss": 1.0662, + "step": 19279 + }, + { + "epoch": 1.42, + "learning_rate": 1.782665677137096e-05, + "loss": 1.1496, + "step": 19280 + }, + { + "epoch": 1.42, + "learning_rate": 1.7826408490569676e-05, + "loss": 1.0411, + "step": 19281 + }, + { + "epoch": 1.42, + "learning_rate": 1.782616019731669e-05, + "loss": 1.0735, + "step": 19282 + }, + { + "epoch": 1.42, + "learning_rate": 1.7825911891612395e-05, + "loss": 1.0237, + "step": 19283 + }, + { + "epoch": 1.42, + "learning_rate": 1.7825663573457185e-05, + "loss": 1.0385, + "step": 19284 + }, + { + "epoch": 1.42, + "learning_rate": 1.782541524285146e-05, + "loss": 0.9407, + "step": 19285 + }, + { + "epoch": 1.43, + "learning_rate": 1.782516689979561e-05, + "loss": 1.0737, + "step": 19286 + }, + { + "epoch": 1.43, + "learning_rate": 1.7824918544290028e-05, + "loss": 0.9762, + "step": 19287 + }, + { + "epoch": 1.43, + "learning_rate": 1.7824670176335116e-05, + "loss": 0.9476, + "step": 19288 + }, + { + "epoch": 1.43, + "learning_rate": 1.7824421795931262e-05, + "loss": 0.9543, + "step": 19289 + }, + { + "epoch": 1.43, + "learning_rate": 1.782417340307887e-05, + "loss": 0.8965, + "step": 19290 + }, + { + "epoch": 1.43, + "learning_rate": 1.7823924997778326e-05, + "loss": 1.0708, + "step": 19291 + }, + { + "epoch": 1.43, + "learning_rate": 1.782367658003003e-05, + "loss": 0.9987, + "step": 19292 + }, + { + "epoch": 1.43, + "learning_rate": 1.7823428149834378e-05, + "loss": 1.0419, + "step": 19293 + }, + { + "epoch": 1.43, + "learning_rate": 1.7823179707191763e-05, + "loss": 0.981, + "step": 19294 + }, + { + "epoch": 1.43, + "learning_rate": 1.7822931252102582e-05, + "loss": 0.9337, + "step": 19295 + }, + { + "epoch": 1.43, + "learning_rate": 1.7822682784567227e-05, + "loss": 1.0202, + "step": 19296 + }, + { + "epoch": 1.43, + "learning_rate": 1.78224343045861e-05, + "loss": 1.0908, + "step": 19297 + }, + { + "epoch": 1.43, + "learning_rate": 1.7822185812159587e-05, + "loss": 0.9904, + "step": 19298 + }, + { + "epoch": 1.43, + "learning_rate": 1.782193730728809e-05, + "loss": 1.1344, + "step": 19299 + }, + { + "epoch": 1.43, + "learning_rate": 1.7821688789972003e-05, + "loss": 0.9573, + "step": 19300 + }, + { + "epoch": 1.43, + "learning_rate": 1.782144026021172e-05, + "loss": 0.9349, + "step": 19301 + }, + { + "epoch": 1.43, + "learning_rate": 1.782119171800764e-05, + "loss": 1.0639, + "step": 19302 + }, + { + "epoch": 1.43, + "learning_rate": 1.7820943163360155e-05, + "loss": 1.1418, + "step": 19303 + }, + { + "epoch": 1.43, + "learning_rate": 1.782069459626966e-05, + "loss": 0.9055, + "step": 19304 + }, + { + "epoch": 1.43, + "learning_rate": 1.782044601673655e-05, + "loss": 1.0297, + "step": 19305 + }, + { + "epoch": 1.43, + "learning_rate": 1.7820197424761227e-05, + "loss": 1.0125, + "step": 19306 + }, + { + "epoch": 1.43, + "learning_rate": 1.781994882034408e-05, + "loss": 0.9594, + "step": 19307 + }, + { + "epoch": 1.43, + "learning_rate": 1.7819700203485505e-05, + "loss": 0.9921, + "step": 19308 + }, + { + "epoch": 1.43, + "learning_rate": 1.78194515741859e-05, + "loss": 1.0154, + "step": 19309 + }, + { + "epoch": 1.43, + "learning_rate": 1.7819202932445655e-05, + "loss": 1.1257, + "step": 19310 + }, + { + "epoch": 1.43, + "learning_rate": 1.7818954278265178e-05, + "loss": 0.9708, + "step": 19311 + }, + { + "epoch": 1.43, + "learning_rate": 1.7818705611644852e-05, + "loss": 0.9979, + "step": 19312 + }, + { + "epoch": 1.43, + "learning_rate": 1.7818456932585076e-05, + "loss": 1.0885, + "step": 19313 + }, + { + "epoch": 1.43, + "learning_rate": 1.781820824108625e-05, + "loss": 1.0385, + "step": 19314 + }, + { + "epoch": 1.43, + "learning_rate": 1.7817959537148765e-05, + "loss": 1.0424, + "step": 19315 + }, + { + "epoch": 1.43, + "learning_rate": 1.781771082077302e-05, + "loss": 1.08, + "step": 19316 + }, + { + "epoch": 1.43, + "learning_rate": 1.781746209195941e-05, + "loss": 0.9244, + "step": 19317 + }, + { + "epoch": 1.43, + "learning_rate": 1.7817213350708322e-05, + "loss": 1.0613, + "step": 19318 + }, + { + "epoch": 1.43, + "learning_rate": 1.7816964597020165e-05, + "loss": 1.0467, + "step": 19319 + }, + { + "epoch": 1.43, + "learning_rate": 1.781671583089533e-05, + "loss": 0.9063, + "step": 19320 + }, + { + "epoch": 1.43, + "learning_rate": 1.781646705233421e-05, + "loss": 1.0325, + "step": 19321 + }, + { + "epoch": 1.43, + "learning_rate": 1.7816218261337203e-05, + "loss": 1.0296, + "step": 19322 + }, + { + "epoch": 1.43, + "learning_rate": 1.7815969457904706e-05, + "loss": 1.0272, + "step": 19323 + }, + { + "epoch": 1.43, + "learning_rate": 1.781572064203711e-05, + "loss": 1.0045, + "step": 19324 + }, + { + "epoch": 1.43, + "learning_rate": 1.7815471813734818e-05, + "loss": 1.003, + "step": 19325 + }, + { + "epoch": 1.43, + "learning_rate": 1.781522297299822e-05, + "loss": 1.0588, + "step": 19326 + }, + { + "epoch": 1.43, + "learning_rate": 1.7814974119827715e-05, + "loss": 0.9876, + "step": 19327 + }, + { + "epoch": 1.43, + "learning_rate": 1.78147252542237e-05, + "loss": 1.1245, + "step": 19328 + }, + { + "epoch": 1.43, + "learning_rate": 1.7814476376186568e-05, + "loss": 0.9953, + "step": 19329 + }, + { + "epoch": 1.43, + "learning_rate": 1.7814227485716712e-05, + "loss": 0.8875, + "step": 19330 + }, + { + "epoch": 1.43, + "learning_rate": 1.7813978582814534e-05, + "loss": 1.0203, + "step": 19331 + }, + { + "epoch": 1.43, + "learning_rate": 1.781372966748043e-05, + "loss": 0.9625, + "step": 19332 + }, + { + "epoch": 1.43, + "learning_rate": 1.781348073971479e-05, + "loss": 1.0611, + "step": 19333 + }, + { + "epoch": 1.43, + "learning_rate": 1.781323179951802e-05, + "loss": 1.0652, + "step": 19334 + }, + { + "epoch": 1.43, + "learning_rate": 1.7812982846890505e-05, + "loss": 0.9737, + "step": 19335 + }, + { + "epoch": 1.43, + "learning_rate": 1.7812733881832645e-05, + "loss": 1.0181, + "step": 19336 + }, + { + "epoch": 1.43, + "learning_rate": 1.781248490434484e-05, + "loss": 1.032, + "step": 19337 + }, + { + "epoch": 1.43, + "learning_rate": 1.7812235914427484e-05, + "loss": 0.9763, + "step": 19338 + }, + { + "epoch": 1.43, + "learning_rate": 1.781198691208097e-05, + "loss": 1.0632, + "step": 19339 + }, + { + "epoch": 1.43, + "learning_rate": 1.78117378973057e-05, + "loss": 1.1111, + "step": 19340 + }, + { + "epoch": 1.43, + "learning_rate": 1.7811488870102063e-05, + "loss": 1.0746, + "step": 19341 + }, + { + "epoch": 1.43, + "learning_rate": 1.781123983047046e-05, + "loss": 1.0497, + "step": 19342 + }, + { + "epoch": 1.43, + "learning_rate": 1.7810990778411287e-05, + "loss": 1.01, + "step": 19343 + }, + { + "epoch": 1.43, + "learning_rate": 1.781074171392494e-05, + "loss": 0.9283, + "step": 19344 + }, + { + "epoch": 1.43, + "learning_rate": 1.7810492637011812e-05, + "loss": 1.0836, + "step": 19345 + }, + { + "epoch": 1.43, + "learning_rate": 1.7810243547672305e-05, + "loss": 0.9219, + "step": 19346 + }, + { + "epoch": 1.43, + "learning_rate": 1.7809994445906815e-05, + "loss": 1.0325, + "step": 19347 + }, + { + "epoch": 1.43, + "learning_rate": 1.780974533171573e-05, + "loss": 0.89, + "step": 19348 + }, + { + "epoch": 1.43, + "learning_rate": 1.7809496205099455e-05, + "loss": 0.9584, + "step": 19349 + }, + { + "epoch": 1.43, + "learning_rate": 1.780924706605838e-05, + "loss": 1.0499, + "step": 19350 + }, + { + "epoch": 1.43, + "learning_rate": 1.7808997914592905e-05, + "loss": 1.0967, + "step": 19351 + }, + { + "epoch": 1.43, + "learning_rate": 1.780874875070343e-05, + "loss": 0.9632, + "step": 19352 + }, + { + "epoch": 1.43, + "learning_rate": 1.7808499574390343e-05, + "loss": 1.0422, + "step": 19353 + }, + { + "epoch": 1.43, + "learning_rate": 1.780825038565405e-05, + "loss": 1.0395, + "step": 19354 + }, + { + "epoch": 1.43, + "learning_rate": 1.7808001184494935e-05, + "loss": 1.1087, + "step": 19355 + }, + { + "epoch": 1.43, + "learning_rate": 1.7807751970913408e-05, + "loss": 1.0899, + "step": 19356 + }, + { + "epoch": 1.43, + "learning_rate": 1.7807502744909858e-05, + "loss": 0.9728, + "step": 19357 + }, + { + "epoch": 1.43, + "learning_rate": 1.7807253506484682e-05, + "loss": 0.9282, + "step": 19358 + }, + { + "epoch": 1.43, + "learning_rate": 1.780700425563828e-05, + "loss": 1.0482, + "step": 19359 + }, + { + "epoch": 1.43, + "learning_rate": 1.7806754992371042e-05, + "loss": 0.9576, + "step": 19360 + }, + { + "epoch": 1.43, + "learning_rate": 1.780650571668337e-05, + "loss": 1.0187, + "step": 19361 + }, + { + "epoch": 1.43, + "learning_rate": 1.7806256428575658e-05, + "loss": 1.0376, + "step": 19362 + }, + { + "epoch": 1.43, + "learning_rate": 1.7806007128048306e-05, + "loss": 1.0319, + "step": 19363 + }, + { + "epoch": 1.43, + "learning_rate": 1.7805757815101706e-05, + "loss": 0.9793, + "step": 19364 + }, + { + "epoch": 1.43, + "learning_rate": 1.7805508489736262e-05, + "loss": 1.0192, + "step": 19365 + }, + { + "epoch": 1.43, + "learning_rate": 1.780525915195236e-05, + "loss": 1.0867, + "step": 19366 + }, + { + "epoch": 1.43, + "learning_rate": 1.7805009801750404e-05, + "loss": 1.0119, + "step": 19367 + }, + { + "epoch": 1.43, + "learning_rate": 1.7804760439130787e-05, + "loss": 1.0032, + "step": 19368 + }, + { + "epoch": 1.43, + "learning_rate": 1.780451106409391e-05, + "loss": 0.9704, + "step": 19369 + }, + { + "epoch": 1.43, + "learning_rate": 1.780426167664017e-05, + "loss": 1.0653, + "step": 19370 + }, + { + "epoch": 1.43, + "learning_rate": 1.780401227676996e-05, + "loss": 1.1458, + "step": 19371 + }, + { + "epoch": 1.43, + "learning_rate": 1.780376286448368e-05, + "loss": 1.133, + "step": 19372 + }, + { + "epoch": 1.43, + "learning_rate": 1.7803513439781717e-05, + "loss": 0.9932, + "step": 19373 + }, + { + "epoch": 1.43, + "learning_rate": 1.7803264002664482e-05, + "loss": 1.0148, + "step": 19374 + }, + { + "epoch": 1.43, + "learning_rate": 1.7803014553132366e-05, + "loss": 1.0467, + "step": 19375 + }, + { + "epoch": 1.43, + "learning_rate": 1.7802765091185763e-05, + "loss": 1.0373, + "step": 19376 + }, + { + "epoch": 1.43, + "learning_rate": 1.7802515616825075e-05, + "loss": 0.9849, + "step": 19377 + }, + { + "epoch": 1.43, + "learning_rate": 1.7802266130050695e-05, + "loss": 1.0794, + "step": 19378 + }, + { + "epoch": 1.43, + "learning_rate": 1.7802016630863018e-05, + "loss": 0.9783, + "step": 19379 + }, + { + "epoch": 1.43, + "learning_rate": 1.7801767119262448e-05, + "loss": 1.1377, + "step": 19380 + }, + { + "epoch": 1.43, + "learning_rate": 1.7801517595249377e-05, + "loss": 0.9554, + "step": 19381 + }, + { + "epoch": 1.43, + "learning_rate": 1.7801268058824203e-05, + "loss": 1.0909, + "step": 19382 + }, + { + "epoch": 1.43, + "learning_rate": 1.7801018509987323e-05, + "loss": 0.9667, + "step": 19383 + }, + { + "epoch": 1.43, + "learning_rate": 1.7800768948739134e-05, + "loss": 1.0164, + "step": 19384 + }, + { + "epoch": 1.43, + "learning_rate": 1.7800519375080035e-05, + "loss": 1.0388, + "step": 19385 + }, + { + "epoch": 1.43, + "learning_rate": 1.780026978901042e-05, + "loss": 0.9683, + "step": 19386 + }, + { + "epoch": 1.43, + "learning_rate": 1.7800020190530687e-05, + "loss": 1.1032, + "step": 19387 + }, + { + "epoch": 1.43, + "learning_rate": 1.7799770579641237e-05, + "loss": 1.0232, + "step": 19388 + }, + { + "epoch": 1.43, + "learning_rate": 1.7799520956342457e-05, + "loss": 1.0279, + "step": 19389 + }, + { + "epoch": 1.43, + "learning_rate": 1.7799271320634756e-05, + "loss": 1.031, + "step": 19390 + }, + { + "epoch": 1.43, + "learning_rate": 1.7799021672518523e-05, + "loss": 0.9908, + "step": 19391 + }, + { + "epoch": 1.43, + "learning_rate": 1.7798772011994163e-05, + "loss": 0.9806, + "step": 19392 + }, + { + "epoch": 1.43, + "learning_rate": 1.779852233906206e-05, + "loss": 1.0646, + "step": 19393 + }, + { + "epoch": 1.43, + "learning_rate": 1.7798272653722626e-05, + "loss": 0.9966, + "step": 19394 + }, + { + "epoch": 1.43, + "learning_rate": 1.779802295597625e-05, + "loss": 1.0369, + "step": 19395 + }, + { + "epoch": 1.43, + "learning_rate": 1.7797773245823334e-05, + "loss": 1.043, + "step": 19396 + }, + { + "epoch": 1.43, + "learning_rate": 1.7797523523264266e-05, + "loss": 1.0351, + "step": 19397 + }, + { + "epoch": 1.43, + "learning_rate": 1.7797273788299455e-05, + "loss": 1.0123, + "step": 19398 + }, + { + "epoch": 1.43, + "learning_rate": 1.779702404092929e-05, + "loss": 1.0281, + "step": 19399 + }, + { + "epoch": 1.43, + "learning_rate": 1.7796774281154173e-05, + "loss": 0.952, + "step": 19400 + }, + { + "epoch": 1.43, + "learning_rate": 1.77965245089745e-05, + "loss": 0.9507, + "step": 19401 + }, + { + "epoch": 1.43, + "learning_rate": 1.7796274724390664e-05, + "loss": 0.946, + "step": 19402 + }, + { + "epoch": 1.43, + "learning_rate": 1.779602492740307e-05, + "loss": 0.9995, + "step": 19403 + }, + { + "epoch": 1.43, + "learning_rate": 1.779577511801211e-05, + "loss": 1.0063, + "step": 19404 + }, + { + "epoch": 1.43, + "learning_rate": 1.7795525296218187e-05, + "loss": 1.0699, + "step": 19405 + }, + { + "epoch": 1.43, + "learning_rate": 1.7795275462021694e-05, + "loss": 0.9787, + "step": 19406 + }, + { + "epoch": 1.43, + "learning_rate": 1.7795025615423026e-05, + "loss": 0.86, + "step": 19407 + }, + { + "epoch": 1.43, + "learning_rate": 1.7794775756422583e-05, + "loss": 1.1516, + "step": 19408 + }, + { + "epoch": 1.43, + "learning_rate": 1.779452588502077e-05, + "loss": 1.0155, + "step": 19409 + }, + { + "epoch": 1.43, + "learning_rate": 1.7794276001217973e-05, + "loss": 0.9802, + "step": 19410 + }, + { + "epoch": 1.43, + "learning_rate": 1.7794026105014593e-05, + "loss": 1.0296, + "step": 19411 + }, + { + "epoch": 1.43, + "learning_rate": 1.7793776196411033e-05, + "loss": 0.9555, + "step": 19412 + }, + { + "epoch": 1.43, + "learning_rate": 1.7793526275407683e-05, + "loss": 0.939, + "step": 19413 + }, + { + "epoch": 1.43, + "learning_rate": 1.7793276342004947e-05, + "loss": 1.009, + "step": 19414 + }, + { + "epoch": 1.43, + "learning_rate": 1.779302639620322e-05, + "loss": 1.1372, + "step": 19415 + }, + { + "epoch": 1.43, + "learning_rate": 1.7792776438002896e-05, + "loss": 1.0008, + "step": 19416 + }, + { + "epoch": 1.43, + "learning_rate": 1.779252646740438e-05, + "loss": 1.0287, + "step": 19417 + }, + { + "epoch": 1.43, + "learning_rate": 1.7792276484408063e-05, + "loss": 1.0438, + "step": 19418 + }, + { + "epoch": 1.43, + "learning_rate": 1.7792026489014346e-05, + "loss": 1.1094, + "step": 19419 + }, + { + "epoch": 1.43, + "learning_rate": 1.7791776481223628e-05, + "loss": 1.0446, + "step": 19420 + }, + { + "epoch": 1.44, + "learning_rate": 1.7791526461036305e-05, + "loss": 1.034, + "step": 19421 + }, + { + "epoch": 1.44, + "learning_rate": 1.7791276428452776e-05, + "loss": 1.1091, + "step": 19422 + }, + { + "epoch": 1.44, + "learning_rate": 1.7791026383473436e-05, + "loss": 1.0372, + "step": 19423 + }, + { + "epoch": 1.44, + "learning_rate": 1.7790776326098683e-05, + "loss": 0.9947, + "step": 19424 + }, + { + "epoch": 1.44, + "learning_rate": 1.7790526256328917e-05, + "loss": 0.9813, + "step": 19425 + }, + { + "epoch": 1.44, + "learning_rate": 1.7790276174164535e-05, + "loss": 0.9825, + "step": 19426 + }, + { + "epoch": 1.44, + "learning_rate": 1.7790026079605937e-05, + "loss": 1.001, + "step": 19427 + }, + { + "epoch": 1.44, + "learning_rate": 1.778977597265352e-05, + "loss": 1.1223, + "step": 19428 + }, + { + "epoch": 1.44, + "learning_rate": 1.7789525853307677e-05, + "loss": 1.0738, + "step": 19429 + }, + { + "epoch": 1.44, + "learning_rate": 1.7789275721568812e-05, + "loss": 1.0293, + "step": 19430 + }, + { + "epoch": 1.44, + "learning_rate": 1.778902557743732e-05, + "loss": 1.1109, + "step": 19431 + }, + { + "epoch": 1.44, + "learning_rate": 1.77887754209136e-05, + "loss": 1.0358, + "step": 19432 + }, + { + "epoch": 1.44, + "learning_rate": 1.7788525251998047e-05, + "loss": 1.0278, + "step": 19433 + }, + { + "epoch": 1.44, + "learning_rate": 1.7788275070691065e-05, + "loss": 1.1361, + "step": 19434 + }, + { + "epoch": 1.44, + "learning_rate": 1.778802487699305e-05, + "loss": 1.0552, + "step": 19435 + }, + { + "epoch": 1.44, + "learning_rate": 1.7787774670904397e-05, + "loss": 1.0725, + "step": 19436 + }, + { + "epoch": 1.44, + "learning_rate": 1.7787524452425506e-05, + "loss": 1.0296, + "step": 19437 + }, + { + "epoch": 1.44, + "learning_rate": 1.7787274221556775e-05, + "loss": 1.0491, + "step": 19438 + }, + { + "epoch": 1.44, + "learning_rate": 1.7787023978298602e-05, + "loss": 0.9159, + "step": 19439 + }, + { + "epoch": 1.44, + "learning_rate": 1.7786773722651383e-05, + "loss": 1.0513, + "step": 19440 + }, + { + "epoch": 1.44, + "learning_rate": 1.7786523454615525e-05, + "loss": 1.0393, + "step": 19441 + }, + { + "epoch": 1.44, + "learning_rate": 1.7786273174191416e-05, + "loss": 0.9474, + "step": 19442 + }, + { + "epoch": 1.44, + "learning_rate": 1.7786022881379458e-05, + "loss": 0.9984, + "step": 19443 + }, + { + "epoch": 1.44, + "learning_rate": 1.7785772576180046e-05, + "loss": 1.0259, + "step": 19444 + }, + { + "epoch": 1.44, + "learning_rate": 1.7785522258593583e-05, + "loss": 1.0639, + "step": 19445 + }, + { + "epoch": 1.44, + "learning_rate": 1.7785271928620466e-05, + "loss": 1.0641, + "step": 19446 + }, + { + "epoch": 1.44, + "learning_rate": 1.7785021586261092e-05, + "loss": 0.9443, + "step": 19447 + }, + { + "epoch": 1.44, + "learning_rate": 1.7784771231515863e-05, + "loss": 1.0169, + "step": 19448 + }, + { + "epoch": 1.44, + "learning_rate": 1.778452086438517e-05, + "loss": 1.001, + "step": 19449 + }, + { + "epoch": 1.44, + "learning_rate": 1.7784270484869418e-05, + "loss": 1.0935, + "step": 19450 + }, + { + "epoch": 1.44, + "learning_rate": 1.7784020092969004e-05, + "loss": 1.1174, + "step": 19451 + }, + { + "epoch": 1.44, + "learning_rate": 1.7783769688684325e-05, + "loss": 0.9849, + "step": 19452 + }, + { + "epoch": 1.44, + "learning_rate": 1.778351927201578e-05, + "loss": 0.9656, + "step": 19453 + }, + { + "epoch": 1.44, + "learning_rate": 1.7783268842963766e-05, + "loss": 0.9004, + "step": 19454 + }, + { + "epoch": 1.44, + "learning_rate": 1.7783018401528682e-05, + "loss": 1.0329, + "step": 19455 + }, + { + "epoch": 1.44, + "learning_rate": 1.778276794771093e-05, + "loss": 0.9957, + "step": 19456 + }, + { + "epoch": 1.44, + "learning_rate": 1.7782517481510903e-05, + "loss": 1.0172, + "step": 19457 + }, + { + "epoch": 1.44, + "learning_rate": 1.7782267002929003e-05, + "loss": 0.9887, + "step": 19458 + }, + { + "epoch": 1.44, + "learning_rate": 1.7782016511965626e-05, + "loss": 1.0395, + "step": 19459 + }, + { + "epoch": 1.44, + "learning_rate": 1.7781766008621173e-05, + "loss": 0.841, + "step": 19460 + }, + { + "epoch": 1.44, + "learning_rate": 1.7781515492896045e-05, + "loss": 0.9296, + "step": 19461 + }, + { + "epoch": 1.44, + "learning_rate": 1.7781264964790632e-05, + "loss": 0.9985, + "step": 19462 + }, + { + "epoch": 1.44, + "learning_rate": 1.7781014424305342e-05, + "loss": 1.0074, + "step": 19463 + }, + { + "epoch": 1.44, + "learning_rate": 1.7780763871440564e-05, + "loss": 0.9651, + "step": 19464 + }, + { + "epoch": 1.44, + "learning_rate": 1.7780513306196704e-05, + "loss": 0.9857, + "step": 19465 + }, + { + "epoch": 1.44, + "learning_rate": 1.7780262728574162e-05, + "loss": 1.0007, + "step": 19466 + }, + { + "epoch": 1.44, + "learning_rate": 1.7780012138573332e-05, + "loss": 1.0715, + "step": 19467 + }, + { + "epoch": 1.44, + "learning_rate": 1.7779761536194612e-05, + "loss": 0.9648, + "step": 19468 + }, + { + "epoch": 1.44, + "learning_rate": 1.7779510921438404e-05, + "loss": 0.9765, + "step": 19469 + }, + { + "epoch": 1.44, + "learning_rate": 1.7779260294305106e-05, + "loss": 0.982, + "step": 19470 + }, + { + "epoch": 1.44, + "learning_rate": 1.7779009654795115e-05, + "loss": 1.0517, + "step": 19471 + }, + { + "epoch": 1.44, + "learning_rate": 1.777875900290883e-05, + "loss": 1.0092, + "step": 19472 + }, + { + "epoch": 1.44, + "learning_rate": 1.7778508338646653e-05, + "loss": 0.9598, + "step": 19473 + }, + { + "epoch": 1.44, + "learning_rate": 1.777825766200898e-05, + "loss": 0.9978, + "step": 19474 + }, + { + "epoch": 1.44, + "learning_rate": 1.777800697299621e-05, + "loss": 1.0328, + "step": 19475 + }, + { + "epoch": 1.44, + "learning_rate": 1.7777756271608742e-05, + "loss": 0.9936, + "step": 19476 + }, + { + "epoch": 1.44, + "learning_rate": 1.7777505557846976e-05, + "loss": 1.0409, + "step": 19477 + }, + { + "epoch": 1.44, + "learning_rate": 1.777725483171131e-05, + "loss": 1.035, + "step": 19478 + }, + { + "epoch": 1.44, + "learning_rate": 1.7777004093202142e-05, + "loss": 0.9612, + "step": 19479 + }, + { + "epoch": 1.44, + "learning_rate": 1.777675334231987e-05, + "loss": 0.9765, + "step": 19480 + }, + { + "epoch": 1.44, + "learning_rate": 1.7776502579064896e-05, + "loss": 0.9862, + "step": 19481 + }, + { + "epoch": 1.44, + "learning_rate": 1.777625180343762e-05, + "loss": 0.9427, + "step": 19482 + }, + { + "epoch": 1.44, + "learning_rate": 1.777600101543844e-05, + "loss": 1.0022, + "step": 19483 + }, + { + "epoch": 1.44, + "learning_rate": 1.7775750215067747e-05, + "loss": 1.0615, + "step": 19484 + }, + { + "epoch": 1.44, + "learning_rate": 1.7775499402325952e-05, + "loss": 0.8542, + "step": 19485 + }, + { + "epoch": 1.44, + "learning_rate": 1.7775248577213445e-05, + "loss": 0.9624, + "step": 19486 + }, + { + "epoch": 1.44, + "learning_rate": 1.7774997739730627e-05, + "loss": 1.0367, + "step": 19487 + }, + { + "epoch": 1.44, + "learning_rate": 1.7774746889877907e-05, + "loss": 0.9798, + "step": 19488 + }, + { + "epoch": 1.44, + "learning_rate": 1.777449602765567e-05, + "loss": 1.0147, + "step": 19489 + }, + { + "epoch": 1.44, + "learning_rate": 1.7774245153064322e-05, + "loss": 1.1083, + "step": 19490 + }, + { + "epoch": 1.44, + "learning_rate": 1.7773994266104264e-05, + "loss": 0.9331, + "step": 19491 + }, + { + "epoch": 1.44, + "learning_rate": 1.7773743366775888e-05, + "loss": 0.966, + "step": 19492 + }, + { + "epoch": 1.44, + "learning_rate": 1.77734924550796e-05, + "loss": 1.0058, + "step": 19493 + }, + { + "epoch": 1.44, + "learning_rate": 1.7773241531015797e-05, + "loss": 1.0511, + "step": 19494 + }, + { + "epoch": 1.44, + "learning_rate": 1.7772990594584877e-05, + "loss": 1.1302, + "step": 19495 + }, + { + "epoch": 1.44, + "learning_rate": 1.7772739645787236e-05, + "loss": 1.0883, + "step": 19496 + }, + { + "epoch": 1.44, + "learning_rate": 1.7772488684623287e-05, + "loss": 0.8781, + "step": 19497 + }, + { + "epoch": 1.44, + "learning_rate": 1.7772237711093412e-05, + "loss": 1.1329, + "step": 19498 + }, + { + "epoch": 1.44, + "learning_rate": 1.777198672519802e-05, + "loss": 1.0824, + "step": 19499 + }, + { + "epoch": 1.44, + "learning_rate": 1.777173572693751e-05, + "loss": 1.0851, + "step": 19500 + }, + { + "epoch": 1.44, + "learning_rate": 1.7771484716312277e-05, + "loss": 0.9385, + "step": 19501 + }, + { + "epoch": 1.44, + "learning_rate": 1.7771233693322726e-05, + "loss": 1.0469, + "step": 19502 + }, + { + "epoch": 1.44, + "learning_rate": 1.777098265796925e-05, + "loss": 1.0587, + "step": 19503 + }, + { + "epoch": 1.44, + "learning_rate": 1.7770731610252253e-05, + "loss": 1.0077, + "step": 19504 + }, + { + "epoch": 1.44, + "learning_rate": 1.7770480550172135e-05, + "loss": 1.0626, + "step": 19505 + }, + { + "epoch": 1.44, + "learning_rate": 1.7770229477729293e-05, + "loss": 1.1164, + "step": 19506 + }, + { + "epoch": 1.44, + "learning_rate": 1.7769978392924128e-05, + "loss": 1.069, + "step": 19507 + }, + { + "epoch": 1.44, + "learning_rate": 1.7769727295757035e-05, + "loss": 0.952, + "step": 19508 + }, + { + "epoch": 1.44, + "learning_rate": 1.7769476186228417e-05, + "loss": 1.1265, + "step": 19509 + }, + { + "epoch": 1.44, + "learning_rate": 1.776922506433868e-05, + "loss": 1.0584, + "step": 19510 + }, + { + "epoch": 1.44, + "learning_rate": 1.7768973930088212e-05, + "loss": 0.9912, + "step": 19511 + }, + { + "epoch": 1.44, + "learning_rate": 1.7768722783477417e-05, + "loss": 0.9426, + "step": 19512 + }, + { + "epoch": 1.44, + "learning_rate": 1.7768471624506694e-05, + "loss": 1.0096, + "step": 19513 + }, + { + "epoch": 1.44, + "learning_rate": 1.776822045317645e-05, + "loss": 1.0413, + "step": 19514 + }, + { + "epoch": 1.44, + "learning_rate": 1.7767969269487075e-05, + "loss": 0.9961, + "step": 19515 + }, + { + "epoch": 1.44, + "learning_rate": 1.7767718073438968e-05, + "loss": 1.0655, + "step": 19516 + }, + { + "epoch": 1.44, + "learning_rate": 1.776746686503254e-05, + "loss": 1.0128, + "step": 19517 + }, + { + "epoch": 1.44, + "learning_rate": 1.776721564426818e-05, + "loss": 0.9378, + "step": 19518 + }, + { + "epoch": 1.44, + "learning_rate": 1.776696441114629e-05, + "loss": 0.9705, + "step": 19519 + }, + { + "epoch": 1.44, + "learning_rate": 1.776671316566727e-05, + "loss": 1.0162, + "step": 19520 + }, + { + "epoch": 1.44, + "learning_rate": 1.7766461907831526e-05, + "loss": 0.9967, + "step": 19521 + }, + { + "epoch": 1.44, + "learning_rate": 1.7766210637639444e-05, + "loss": 0.9505, + "step": 19522 + }, + { + "epoch": 1.44, + "learning_rate": 1.7765959355091438e-05, + "loss": 1.001, + "step": 19523 + }, + { + "epoch": 1.44, + "learning_rate": 1.77657080601879e-05, + "loss": 0.9776, + "step": 19524 + }, + { + "epoch": 1.44, + "learning_rate": 1.776545675292923e-05, + "loss": 1.0605, + "step": 19525 + }, + { + "epoch": 1.44, + "learning_rate": 1.7765205433315833e-05, + "loss": 1.0364, + "step": 19526 + }, + { + "epoch": 1.44, + "learning_rate": 1.7764954101348103e-05, + "loss": 0.8929, + "step": 19527 + }, + { + "epoch": 1.44, + "learning_rate": 1.7764702757026442e-05, + "loss": 1.0265, + "step": 19528 + }, + { + "epoch": 1.44, + "learning_rate": 1.7764451400351253e-05, + "loss": 0.9828, + "step": 19529 + }, + { + "epoch": 1.44, + "learning_rate": 1.7764200031322928e-05, + "loss": 0.9608, + "step": 19530 + }, + { + "epoch": 1.44, + "learning_rate": 1.7763948649941874e-05, + "loss": 1.0932, + "step": 19531 + }, + { + "epoch": 1.44, + "learning_rate": 1.776369725620849e-05, + "loss": 0.9701, + "step": 19532 + }, + { + "epoch": 1.44, + "learning_rate": 1.7763445850123174e-05, + "loss": 0.9207, + "step": 19533 + }, + { + "epoch": 1.44, + "learning_rate": 1.7763194431686326e-05, + "loss": 1.0144, + "step": 19534 + }, + { + "epoch": 1.44, + "learning_rate": 1.7762943000898346e-05, + "loss": 1.0733, + "step": 19535 + }, + { + "epoch": 1.44, + "learning_rate": 1.7762691557759636e-05, + "loss": 0.9088, + "step": 19536 + }, + { + "epoch": 1.44, + "learning_rate": 1.7762440102270596e-05, + "loss": 1.109, + "step": 19537 + }, + { + "epoch": 1.44, + "learning_rate": 1.7762188634431623e-05, + "loss": 0.9598, + "step": 19538 + }, + { + "epoch": 1.44, + "learning_rate": 1.776193715424312e-05, + "loss": 0.9603, + "step": 19539 + }, + { + "epoch": 1.44, + "learning_rate": 1.7761685661705485e-05, + "loss": 1.0924, + "step": 19540 + }, + { + "epoch": 1.44, + "learning_rate": 1.776143415681912e-05, + "loss": 0.9962, + "step": 19541 + }, + { + "epoch": 1.44, + "learning_rate": 1.7761182639584424e-05, + "loss": 1.0063, + "step": 19542 + }, + { + "epoch": 1.44, + "learning_rate": 1.7760931110001797e-05, + "loss": 1.1095, + "step": 19543 + }, + { + "epoch": 1.44, + "learning_rate": 1.776067956807164e-05, + "loss": 0.9915, + "step": 19544 + }, + { + "epoch": 1.44, + "learning_rate": 1.7760428013794352e-05, + "loss": 0.9918, + "step": 19545 + }, + { + "epoch": 1.44, + "learning_rate": 1.7760176447170337e-05, + "loss": 1.0379, + "step": 19546 + }, + { + "epoch": 1.44, + "learning_rate": 1.775992486819999e-05, + "loss": 0.9615, + "step": 19547 + }, + { + "epoch": 1.44, + "learning_rate": 1.775967327688371e-05, + "loss": 0.9373, + "step": 19548 + }, + { + "epoch": 1.44, + "learning_rate": 1.775942167322191e-05, + "loss": 0.8922, + "step": 19549 + }, + { + "epoch": 1.44, + "learning_rate": 1.7759170057214974e-05, + "loss": 0.9604, + "step": 19550 + }, + { + "epoch": 1.44, + "learning_rate": 1.775891842886331e-05, + "loss": 1.0007, + "step": 19551 + }, + { + "epoch": 1.44, + "learning_rate": 1.775866678816732e-05, + "loss": 1.0417, + "step": 19552 + }, + { + "epoch": 1.44, + "learning_rate": 1.7758415135127404e-05, + "loss": 1.0542, + "step": 19553 + }, + { + "epoch": 1.44, + "learning_rate": 1.7758163469743955e-05, + "loss": 1.0356, + "step": 19554 + }, + { + "epoch": 1.44, + "learning_rate": 1.7757911792017383e-05, + "loss": 1.1152, + "step": 19555 + }, + { + "epoch": 1.44, + "learning_rate": 1.775766010194808e-05, + "loss": 1.1101, + "step": 19556 + }, + { + "epoch": 1.45, + "learning_rate": 1.7757408399536458e-05, + "loss": 0.9976, + "step": 19557 + }, + { + "epoch": 1.45, + "learning_rate": 1.7757156684782905e-05, + "loss": 0.9458, + "step": 19558 + }, + { + "epoch": 1.45, + "learning_rate": 1.7756904957687827e-05, + "loss": 1.0281, + "step": 19559 + }, + { + "epoch": 1.45, + "learning_rate": 1.7756653218251626e-05, + "loss": 1.0207, + "step": 19560 + }, + { + "epoch": 1.45, + "learning_rate": 1.77564014664747e-05, + "loss": 1.0213, + "step": 19561 + }, + { + "epoch": 1.45, + "learning_rate": 1.7756149702357448e-05, + "loss": 1.0076, + "step": 19562 + }, + { + "epoch": 1.45, + "learning_rate": 1.7755897925900275e-05, + "loss": 1.0612, + "step": 19563 + }, + { + "epoch": 1.45, + "learning_rate": 1.775564613710358e-05, + "loss": 0.9582, + "step": 19564 + }, + { + "epoch": 1.45, + "learning_rate": 1.775539433596776e-05, + "loss": 1.0306, + "step": 19565 + }, + { + "epoch": 1.45, + "learning_rate": 1.7755142522493223e-05, + "loss": 1.1093, + "step": 19566 + }, + { + "epoch": 1.45, + "learning_rate": 1.775489069668036e-05, + "loss": 1.0759, + "step": 19567 + }, + { + "epoch": 1.45, + "learning_rate": 1.775463885852958e-05, + "loss": 1.0247, + "step": 19568 + }, + { + "epoch": 1.45, + "learning_rate": 1.7754387008041282e-05, + "loss": 0.9677, + "step": 19569 + }, + { + "epoch": 1.45, + "learning_rate": 1.7754135145215863e-05, + "loss": 1.0941, + "step": 19570 + }, + { + "epoch": 1.45, + "learning_rate": 1.7753883270053727e-05, + "loss": 1.0533, + "step": 19571 + }, + { + "epoch": 1.45, + "learning_rate": 1.7753631382555272e-05, + "loss": 1.071, + "step": 19572 + }, + { + "epoch": 1.45, + "learning_rate": 1.7753379482720903e-05, + "loss": 0.9991, + "step": 19573 + }, + { + "epoch": 1.45, + "learning_rate": 1.7753127570551017e-05, + "loss": 1.0007, + "step": 19574 + }, + { + "epoch": 1.45, + "learning_rate": 1.7752875646046014e-05, + "loss": 1.1078, + "step": 19575 + }, + { + "epoch": 1.45, + "learning_rate": 1.77526237092063e-05, + "loss": 1.1301, + "step": 19576 + }, + { + "epoch": 1.45, + "learning_rate": 1.775237176003227e-05, + "loss": 0.9971, + "step": 19577 + }, + { + "epoch": 1.45, + "learning_rate": 1.7752119798524327e-05, + "loss": 0.8493, + "step": 19578 + }, + { + "epoch": 1.45, + "learning_rate": 1.7751867824682877e-05, + "loss": 0.9465, + "step": 19579 + }, + { + "epoch": 1.45, + "learning_rate": 1.7751615838508312e-05, + "loss": 1.0654, + "step": 19580 + }, + { + "epoch": 1.45, + "learning_rate": 1.7751363840001035e-05, + "loss": 0.9458, + "step": 19581 + }, + { + "epoch": 1.45, + "learning_rate": 1.7751111829161454e-05, + "loss": 0.955, + "step": 19582 + }, + { + "epoch": 1.45, + "learning_rate": 1.7750859805989963e-05, + "loss": 0.9844, + "step": 19583 + }, + { + "epoch": 1.45, + "learning_rate": 1.7750607770486964e-05, + "loss": 0.8897, + "step": 19584 + }, + { + "epoch": 1.45, + "learning_rate": 1.775035572265286e-05, + "loss": 1.0607, + "step": 19585 + }, + { + "epoch": 1.45, + "learning_rate": 1.7750103662488052e-05, + "loss": 0.9986, + "step": 19586 + }, + { + "epoch": 1.45, + "learning_rate": 1.774985158999294e-05, + "loss": 1.0678, + "step": 19587 + }, + { + "epoch": 1.45, + "learning_rate": 1.7749599505167923e-05, + "loss": 1.071, + "step": 19588 + }, + { + "epoch": 1.45, + "learning_rate": 1.7749347408013404e-05, + "loss": 0.9998, + "step": 19589 + }, + { + "epoch": 1.45, + "learning_rate": 1.7749095298529785e-05, + "loss": 1.0493, + "step": 19590 + }, + { + "epoch": 1.45, + "learning_rate": 1.7748843176717466e-05, + "loss": 1.0037, + "step": 19591 + }, + { + "epoch": 1.45, + "learning_rate": 1.774859104257685e-05, + "loss": 1.0358, + "step": 19592 + }, + { + "epoch": 1.45, + "learning_rate": 1.7748338896108333e-05, + "loss": 1.0604, + "step": 19593 + }, + { + "epoch": 1.45, + "learning_rate": 1.7748086737312323e-05, + "loss": 0.9385, + "step": 19594 + }, + { + "epoch": 1.45, + "learning_rate": 1.774783456618922e-05, + "loss": 1.1067, + "step": 19595 + }, + { + "epoch": 1.45, + "learning_rate": 1.7747582382739417e-05, + "loss": 0.9712, + "step": 19596 + }, + { + "epoch": 1.45, + "learning_rate": 1.7747330186963327e-05, + "loss": 1.0917, + "step": 19597 + }, + { + "epoch": 1.45, + "learning_rate": 1.774707797886134e-05, + "loss": 0.998, + "step": 19598 + }, + { + "epoch": 1.45, + "learning_rate": 1.7746825758433863e-05, + "loss": 1.0892, + "step": 19599 + }, + { + "epoch": 1.45, + "learning_rate": 1.77465735256813e-05, + "loss": 1.0039, + "step": 19600 + }, + { + "epoch": 1.45, + "learning_rate": 1.7746321280604052e-05, + "loss": 1.0716, + "step": 19601 + }, + { + "epoch": 1.45, + "learning_rate": 1.7746069023202514e-05, + "loss": 1.1123, + "step": 19602 + }, + { + "epoch": 1.45, + "learning_rate": 1.774581675347709e-05, + "loss": 1.1952, + "step": 19603 + }, + { + "epoch": 1.45, + "learning_rate": 1.7745564471428184e-05, + "loss": 1.1161, + "step": 19604 + }, + { + "epoch": 1.45, + "learning_rate": 1.7745312177056195e-05, + "loss": 1.0136, + "step": 19605 + }, + { + "epoch": 1.45, + "learning_rate": 1.7745059870361524e-05, + "loss": 0.9844, + "step": 19606 + }, + { + "epoch": 1.45, + "learning_rate": 1.774480755134458e-05, + "loss": 1.0606, + "step": 19607 + }, + { + "epoch": 1.45, + "learning_rate": 1.774455522000575e-05, + "loss": 0.9954, + "step": 19608 + }, + { + "epoch": 1.45, + "learning_rate": 1.7744302876345445e-05, + "loss": 1.0924, + "step": 19609 + }, + { + "epoch": 1.45, + "learning_rate": 1.7744050520364068e-05, + "loss": 0.9942, + "step": 19610 + }, + { + "epoch": 1.45, + "learning_rate": 1.7743798152062017e-05, + "loss": 0.9151, + "step": 19611 + }, + { + "epoch": 1.45, + "learning_rate": 1.7743545771439695e-05, + "loss": 1.0703, + "step": 19612 + }, + { + "epoch": 1.45, + "learning_rate": 1.77432933784975e-05, + "loss": 0.9364, + "step": 19613 + }, + { + "epoch": 1.45, + "learning_rate": 1.7743040973235834e-05, + "loss": 0.9227, + "step": 19614 + }, + { + "epoch": 1.45, + "learning_rate": 1.7742788555655104e-05, + "loss": 1.0348, + "step": 19615 + }, + { + "epoch": 1.45, + "learning_rate": 1.774253612575571e-05, + "loss": 1.0369, + "step": 19616 + }, + { + "epoch": 1.45, + "learning_rate": 1.7742283683538046e-05, + "loss": 0.824, + "step": 19617 + }, + { + "epoch": 1.45, + "learning_rate": 1.7742031229002522e-05, + "loss": 0.9782, + "step": 19618 + }, + { + "epoch": 1.45, + "learning_rate": 1.7741778762149536e-05, + "loss": 1.003, + "step": 19619 + }, + { + "epoch": 1.45, + "learning_rate": 1.774152628297949e-05, + "loss": 0.9652, + "step": 19620 + }, + { + "epoch": 1.45, + "learning_rate": 1.7741273791492792e-05, + "loss": 1.1145, + "step": 19621 + }, + { + "epoch": 1.45, + "learning_rate": 1.7741021287689837e-05, + "loss": 0.959, + "step": 19622 + }, + { + "epoch": 1.45, + "learning_rate": 1.7740768771571023e-05, + "loss": 1.0396, + "step": 19623 + }, + { + "epoch": 1.45, + "learning_rate": 1.774051624313676e-05, + "loss": 0.8963, + "step": 19624 + }, + { + "epoch": 1.45, + "learning_rate": 1.7740263702387447e-05, + "loss": 0.8407, + "step": 19625 + }, + { + "epoch": 1.45, + "learning_rate": 1.7740011149323485e-05, + "loss": 1.0296, + "step": 19626 + }, + { + "epoch": 1.45, + "learning_rate": 1.7739758583945274e-05, + "loss": 0.8579, + "step": 19627 + }, + { + "epoch": 1.45, + "learning_rate": 1.773950600625322e-05, + "loss": 0.9264, + "step": 19628 + }, + { + "epoch": 1.45, + "learning_rate": 1.773925341624772e-05, + "loss": 1.0602, + "step": 19629 + }, + { + "epoch": 1.45, + "learning_rate": 1.7739000813929183e-05, + "loss": 1.0527, + "step": 19630 + }, + { + "epoch": 1.45, + "learning_rate": 1.7738748199298e-05, + "loss": 1.0244, + "step": 19631 + }, + { + "epoch": 1.45, + "learning_rate": 1.7738495572354585e-05, + "loss": 1.0165, + "step": 19632 + }, + { + "epoch": 1.45, + "learning_rate": 1.7738242933099332e-05, + "loss": 1.0989, + "step": 19633 + }, + { + "epoch": 1.45, + "learning_rate": 1.7737990281532645e-05, + "loss": 1.0794, + "step": 19634 + }, + { + "epoch": 1.45, + "learning_rate": 1.7737737617654923e-05, + "loss": 1.0707, + "step": 19635 + }, + { + "epoch": 1.45, + "learning_rate": 1.7737484941466577e-05, + "loss": 0.92, + "step": 19636 + }, + { + "epoch": 1.45, + "learning_rate": 1.7737232252968e-05, + "loss": 1.1389, + "step": 19637 + }, + { + "epoch": 1.45, + "learning_rate": 1.77369795521596e-05, + "loss": 1.0074, + "step": 19638 + }, + { + "epoch": 1.45, + "learning_rate": 1.7736726839041774e-05, + "loss": 1.0265, + "step": 19639 + }, + { + "epoch": 1.45, + "learning_rate": 1.7736474113614926e-05, + "loss": 0.9717, + "step": 19640 + }, + { + "epoch": 1.45, + "learning_rate": 1.7736221375879458e-05, + "loss": 0.993, + "step": 19641 + }, + { + "epoch": 1.45, + "learning_rate": 1.7735968625835775e-05, + "loss": 1.0739, + "step": 19642 + }, + { + "epoch": 1.45, + "learning_rate": 1.7735715863484277e-05, + "loss": 0.9847, + "step": 19643 + }, + { + "epoch": 1.45, + "learning_rate": 1.7735463088825362e-05, + "loss": 1.0463, + "step": 19644 + }, + { + "epoch": 1.45, + "learning_rate": 1.773521030185944e-05, + "loss": 1.1154, + "step": 19645 + }, + { + "epoch": 1.45, + "learning_rate": 1.7734957502586906e-05, + "loss": 0.9947, + "step": 19646 + }, + { + "epoch": 1.45, + "learning_rate": 1.7734704691008163e-05, + "loss": 0.9847, + "step": 19647 + }, + { + "epoch": 1.45, + "learning_rate": 1.773445186712362e-05, + "loss": 1.0999, + "step": 19648 + }, + { + "epoch": 1.45, + "learning_rate": 1.7734199030933675e-05, + "loss": 0.9741, + "step": 19649 + }, + { + "epoch": 1.45, + "learning_rate": 1.7733946182438726e-05, + "loss": 1.0445, + "step": 19650 + }, + { + "epoch": 1.45, + "learning_rate": 1.773369332163918e-05, + "loss": 0.9391, + "step": 19651 + }, + { + "epoch": 1.45, + "learning_rate": 1.7733440448535443e-05, + "loss": 0.9702, + "step": 19652 + }, + { + "epoch": 1.45, + "learning_rate": 1.773318756312791e-05, + "loss": 1.0118, + "step": 19653 + }, + { + "epoch": 1.45, + "learning_rate": 1.7732934665416985e-05, + "loss": 0.9455, + "step": 19654 + }, + { + "epoch": 1.45, + "learning_rate": 1.7732681755403073e-05, + "loss": 1.0696, + "step": 19655 + }, + { + "epoch": 1.45, + "learning_rate": 1.7732428833086573e-05, + "loss": 1.0142, + "step": 19656 + }, + { + "epoch": 1.45, + "learning_rate": 1.7732175898467892e-05, + "loss": 1.0588, + "step": 19657 + }, + { + "epoch": 1.45, + "learning_rate": 1.773192295154743e-05, + "loss": 1.0059, + "step": 19658 + }, + { + "epoch": 1.45, + "learning_rate": 1.773166999232559e-05, + "loss": 1.0348, + "step": 19659 + }, + { + "epoch": 1.45, + "learning_rate": 1.773141702080277e-05, + "loss": 1.0596, + "step": 19660 + }, + { + "epoch": 1.45, + "learning_rate": 1.7731164036979377e-05, + "loss": 1.0867, + "step": 19661 + }, + { + "epoch": 1.45, + "learning_rate": 1.7730911040855814e-05, + "loss": 1.1323, + "step": 19662 + }, + { + "epoch": 1.45, + "learning_rate": 1.7730658032432483e-05, + "loss": 1.0365, + "step": 19663 + }, + { + "epoch": 1.45, + "learning_rate": 1.7730405011709782e-05, + "loss": 1.0183, + "step": 19664 + }, + { + "epoch": 1.45, + "learning_rate": 1.7730151978688123e-05, + "loss": 1.0504, + "step": 19665 + }, + { + "epoch": 1.45, + "learning_rate": 1.7729898933367897e-05, + "loss": 1.0261, + "step": 19666 + }, + { + "epoch": 1.45, + "learning_rate": 1.7729645875749516e-05, + "loss": 1.1298, + "step": 19667 + }, + { + "epoch": 1.45, + "learning_rate": 1.7729392805833378e-05, + "loss": 1.0486, + "step": 19668 + }, + { + "epoch": 1.45, + "learning_rate": 1.7729139723619887e-05, + "loss": 0.986, + "step": 19669 + }, + { + "epoch": 1.45, + "learning_rate": 1.7728886629109443e-05, + "loss": 0.973, + "step": 19670 + }, + { + "epoch": 1.45, + "learning_rate": 1.7728633522302452e-05, + "loss": 0.8774, + "step": 19671 + }, + { + "epoch": 1.45, + "learning_rate": 1.7728380403199318e-05, + "loss": 1.048, + "step": 19672 + }, + { + "epoch": 1.45, + "learning_rate": 1.772812727180044e-05, + "loss": 0.9869, + "step": 19673 + }, + { + "epoch": 1.45, + "learning_rate": 1.772787412810622e-05, + "loss": 0.9634, + "step": 19674 + }, + { + "epoch": 1.45, + "learning_rate": 1.7727620972117065e-05, + "loss": 1.0372, + "step": 19675 + }, + { + "epoch": 1.45, + "learning_rate": 1.7727367803833378e-05, + "loss": 1.0401, + "step": 19676 + }, + { + "epoch": 1.45, + "learning_rate": 1.7727114623255554e-05, + "loss": 1.0588, + "step": 19677 + }, + { + "epoch": 1.45, + "learning_rate": 1.7726861430384005e-05, + "loss": 0.8461, + "step": 19678 + }, + { + "epoch": 1.45, + "learning_rate": 1.7726608225219128e-05, + "loss": 1.0694, + "step": 19679 + }, + { + "epoch": 1.45, + "learning_rate": 1.772635500776133e-05, + "loss": 0.8109, + "step": 19680 + }, + { + "epoch": 1.45, + "learning_rate": 1.772610177801101e-05, + "loss": 1.0594, + "step": 19681 + }, + { + "epoch": 1.45, + "learning_rate": 1.772584853596857e-05, + "loss": 1.0174, + "step": 19682 + }, + { + "epoch": 1.45, + "learning_rate": 1.772559528163442e-05, + "loss": 1.0543, + "step": 19683 + }, + { + "epoch": 1.45, + "learning_rate": 1.7725342015008956e-05, + "loss": 1.0032, + "step": 19684 + }, + { + "epoch": 1.45, + "learning_rate": 1.7725088736092585e-05, + "loss": 1.076, + "step": 19685 + }, + { + "epoch": 1.45, + "learning_rate": 1.7724835444885708e-05, + "loss": 0.9107, + "step": 19686 + }, + { + "epoch": 1.45, + "learning_rate": 1.772458214138873e-05, + "loss": 0.9585, + "step": 19687 + }, + { + "epoch": 1.45, + "learning_rate": 1.7724328825602047e-05, + "loss": 0.9819, + "step": 19688 + }, + { + "epoch": 1.45, + "learning_rate": 1.7724075497526072e-05, + "loss": 1.0785, + "step": 19689 + }, + { + "epoch": 1.45, + "learning_rate": 1.77238221571612e-05, + "loss": 1.0759, + "step": 19690 + }, + { + "epoch": 1.45, + "learning_rate": 1.772356880450784e-05, + "loss": 1.0109, + "step": 19691 + }, + { + "epoch": 1.46, + "learning_rate": 1.7723315439566394e-05, + "loss": 1.1141, + "step": 19692 + }, + { + "epoch": 1.46, + "learning_rate": 1.772306206233726e-05, + "loss": 0.9518, + "step": 19693 + }, + { + "epoch": 1.46, + "learning_rate": 1.772280867282085e-05, + "loss": 1.0752, + "step": 19694 + }, + { + "epoch": 1.46, + "learning_rate": 1.7722555271017555e-05, + "loss": 1.005, + "step": 19695 + }, + { + "epoch": 1.46, + "learning_rate": 1.772230185692779e-05, + "loss": 1.0282, + "step": 19696 + }, + { + "epoch": 1.46, + "learning_rate": 1.772204843055195e-05, + "loss": 1.0296, + "step": 19697 + }, + { + "epoch": 1.46, + "learning_rate": 1.7721794991890446e-05, + "loss": 1.0658, + "step": 19698 + }, + { + "epoch": 1.46, + "learning_rate": 1.7721541540943672e-05, + "loss": 1.0154, + "step": 19699 + }, + { + "epoch": 1.46, + "learning_rate": 1.7721288077712037e-05, + "loss": 0.9912, + "step": 19700 + }, + { + "epoch": 1.46, + "learning_rate": 1.7721034602195944e-05, + "loss": 1.003, + "step": 19701 + }, + { + "epoch": 1.46, + "learning_rate": 1.7720781114395794e-05, + "loss": 0.9473, + "step": 19702 + }, + { + "epoch": 1.46, + "learning_rate": 1.7720527614311995e-05, + "loss": 1.0372, + "step": 19703 + }, + { + "epoch": 1.46, + "learning_rate": 1.7720274101944944e-05, + "loss": 0.9026, + "step": 19704 + }, + { + "epoch": 1.46, + "learning_rate": 1.772002057729505e-05, + "loss": 1.0089, + "step": 19705 + }, + { + "epoch": 1.46, + "learning_rate": 1.771976704036271e-05, + "loss": 1.0098, + "step": 19706 + }, + { + "epoch": 1.46, + "learning_rate": 1.7719513491148335e-05, + "loss": 1.0474, + "step": 19707 + }, + { + "epoch": 1.46, + "learning_rate": 1.7719259929652324e-05, + "loss": 0.9652, + "step": 19708 + }, + { + "epoch": 1.46, + "learning_rate": 1.7719006355875076e-05, + "loss": 1.059, + "step": 19709 + }, + { + "epoch": 1.46, + "learning_rate": 1.7718752769817006e-05, + "loss": 0.9463, + "step": 19710 + }, + { + "epoch": 1.46, + "learning_rate": 1.7718499171478507e-05, + "loss": 0.995, + "step": 19711 + }, + { + "epoch": 1.46, + "learning_rate": 1.771824556085999e-05, + "loss": 0.9567, + "step": 19712 + }, + { + "epoch": 1.46, + "learning_rate": 1.7717991937961847e-05, + "loss": 0.9715, + "step": 19713 + }, + { + "epoch": 1.46, + "learning_rate": 1.7717738302784496e-05, + "loss": 0.9376, + "step": 19714 + }, + { + "epoch": 1.46, + "learning_rate": 1.771748465532833e-05, + "loss": 1.0756, + "step": 19715 + }, + { + "epoch": 1.46, + "learning_rate": 1.7717230995593757e-05, + "loss": 0.9358, + "step": 19716 + }, + { + "epoch": 1.46, + "learning_rate": 1.7716977323581182e-05, + "loss": 1.0601, + "step": 19717 + }, + { + "epoch": 1.46, + "learning_rate": 1.7716723639291008e-05, + "loss": 1.0573, + "step": 19718 + }, + { + "epoch": 1.46, + "learning_rate": 1.7716469942723632e-05, + "loss": 0.9179, + "step": 19719 + }, + { + "epoch": 1.46, + "learning_rate": 1.7716216233879466e-05, + "loss": 1.0333, + "step": 19720 + }, + { + "epoch": 1.46, + "learning_rate": 1.771596251275891e-05, + "loss": 1.0368, + "step": 19721 + }, + { + "epoch": 1.46, + "learning_rate": 1.771570877936237e-05, + "loss": 1.0665, + "step": 19722 + }, + { + "epoch": 1.46, + "learning_rate": 1.7715455033690244e-05, + "loss": 1.006, + "step": 19723 + }, + { + "epoch": 1.46, + "learning_rate": 1.771520127574294e-05, + "loss": 0.974, + "step": 19724 + }, + { + "epoch": 1.46, + "learning_rate": 1.771494750552086e-05, + "loss": 1.0199, + "step": 19725 + }, + { + "epoch": 1.46, + "learning_rate": 1.7714693723024413e-05, + "loss": 1.0462, + "step": 19726 + }, + { + "epoch": 1.46, + "learning_rate": 1.7714439928253996e-05, + "loss": 1.0146, + "step": 19727 + }, + { + "epoch": 1.46, + "learning_rate": 1.7714186121210014e-05, + "loss": 0.9646, + "step": 19728 + }, + { + "epoch": 1.46, + "learning_rate": 1.7713932301892877e-05, + "loss": 0.9992, + "step": 19729 + }, + { + "epoch": 1.46, + "learning_rate": 1.7713678470302978e-05, + "loss": 1.0034, + "step": 19730 + }, + { + "epoch": 1.46, + "learning_rate": 1.771342462644073e-05, + "loss": 1.1418, + "step": 19731 + }, + { + "epoch": 1.46, + "learning_rate": 1.7713170770306538e-05, + "loss": 0.9681, + "step": 19732 + }, + { + "epoch": 1.46, + "learning_rate": 1.7712916901900795e-05, + "loss": 0.9357, + "step": 19733 + }, + { + "epoch": 1.46, + "learning_rate": 1.7712663021223914e-05, + "loss": 0.9871, + "step": 19734 + }, + { + "epoch": 1.46, + "learning_rate": 1.7712409128276297e-05, + "loss": 0.9763, + "step": 19735 + }, + { + "epoch": 1.46, + "learning_rate": 1.7712155223058347e-05, + "loss": 1.0326, + "step": 19736 + }, + { + "epoch": 1.46, + "learning_rate": 1.7711901305570468e-05, + "loss": 0.9703, + "step": 19737 + }, + { + "epoch": 1.46, + "learning_rate": 1.7711647375813066e-05, + "loss": 0.9905, + "step": 19738 + }, + { + "epoch": 1.46, + "learning_rate": 1.771139343378654e-05, + "loss": 1.0216, + "step": 19739 + }, + { + "epoch": 1.46, + "learning_rate": 1.77111394794913e-05, + "loss": 1.0416, + "step": 19740 + }, + { + "epoch": 1.46, + "learning_rate": 1.771088551292775e-05, + "loss": 0.9669, + "step": 19741 + }, + { + "epoch": 1.46, + "learning_rate": 1.7710631534096287e-05, + "loss": 0.9781, + "step": 19742 + }, + { + "epoch": 1.46, + "learning_rate": 1.771037754299732e-05, + "loss": 1.0178, + "step": 19743 + }, + { + "epoch": 1.46, + "learning_rate": 1.7710123539631255e-05, + "loss": 1.0662, + "step": 19744 + }, + { + "epoch": 1.46, + "learning_rate": 1.770986952399849e-05, + "loss": 1.0607, + "step": 19745 + }, + { + "epoch": 1.46, + "learning_rate": 1.7709615496099437e-05, + "loss": 0.9498, + "step": 19746 + }, + { + "epoch": 1.46, + "learning_rate": 1.7709361455934495e-05, + "loss": 0.9984, + "step": 19747 + }, + { + "epoch": 1.46, + "learning_rate": 1.770910740350407e-05, + "loss": 0.9732, + "step": 19748 + }, + { + "epoch": 1.46, + "learning_rate": 1.770885333880856e-05, + "loss": 1.0425, + "step": 19749 + }, + { + "epoch": 1.46, + "learning_rate": 1.770859926184838e-05, + "loss": 0.8957, + "step": 19750 + }, + { + "epoch": 1.46, + "learning_rate": 1.770834517262393e-05, + "loss": 0.9439, + "step": 19751 + }, + { + "epoch": 1.46, + "learning_rate": 1.770809107113561e-05, + "loss": 1.0669, + "step": 19752 + }, + { + "epoch": 1.46, + "learning_rate": 1.770783695738383e-05, + "loss": 0.9604, + "step": 19753 + }, + { + "epoch": 1.46, + "learning_rate": 1.7707582831368988e-05, + "loss": 1.1303, + "step": 19754 + }, + { + "epoch": 1.46, + "learning_rate": 1.7707328693091494e-05, + "loss": 0.9159, + "step": 19755 + }, + { + "epoch": 1.46, + "learning_rate": 1.770707454255175e-05, + "loss": 0.9876, + "step": 19756 + }, + { + "epoch": 1.46, + "learning_rate": 1.7706820379750162e-05, + "loss": 0.9939, + "step": 19757 + }, + { + "epoch": 1.46, + "learning_rate": 1.770656620468713e-05, + "loss": 1.008, + "step": 19758 + }, + { + "epoch": 1.46, + "learning_rate": 1.7706312017363064e-05, + "loss": 0.9198, + "step": 19759 + }, + { + "epoch": 1.46, + "learning_rate": 1.7706057817778363e-05, + "loss": 1.1633, + "step": 19760 + }, + { + "epoch": 1.46, + "learning_rate": 1.7705803605933436e-05, + "loss": 1.0264, + "step": 19761 + }, + { + "epoch": 1.46, + "learning_rate": 1.7705549381828686e-05, + "loss": 1.0485, + "step": 19762 + }, + { + "epoch": 1.46, + "learning_rate": 1.770529514546452e-05, + "loss": 0.9156, + "step": 19763 + }, + { + "epoch": 1.46, + "learning_rate": 1.7705040896841334e-05, + "loss": 1.1022, + "step": 19764 + }, + { + "epoch": 1.46, + "learning_rate": 1.7704786635959542e-05, + "loss": 0.934, + "step": 19765 + }, + { + "epoch": 1.46, + "learning_rate": 1.7704532362819543e-05, + "loss": 1.0403, + "step": 19766 + }, + { + "epoch": 1.46, + "learning_rate": 1.7704278077421745e-05, + "loss": 0.9796, + "step": 19767 + }, + { + "epoch": 1.46, + "learning_rate": 1.770402377976655e-05, + "loss": 1.0054, + "step": 19768 + }, + { + "epoch": 1.46, + "learning_rate": 1.7703769469854363e-05, + "loss": 1.0003, + "step": 19769 + }, + { + "epoch": 1.46, + "learning_rate": 1.7703515147685587e-05, + "loss": 1.0911, + "step": 19770 + }, + { + "epoch": 1.46, + "learning_rate": 1.7703260813260633e-05, + "loss": 0.9522, + "step": 19771 + }, + { + "epoch": 1.46, + "learning_rate": 1.7703006466579896e-05, + "loss": 1.0284, + "step": 19772 + }, + { + "epoch": 1.46, + "learning_rate": 1.770275210764379e-05, + "loss": 1.0348, + "step": 19773 + }, + { + "epoch": 1.46, + "learning_rate": 1.770249773645272e-05, + "loss": 0.9286, + "step": 19774 + }, + { + "epoch": 1.46, + "learning_rate": 1.770224335300708e-05, + "loss": 1.0944, + "step": 19775 + }, + { + "epoch": 1.46, + "learning_rate": 1.770198895730728e-05, + "loss": 0.9702, + "step": 19776 + }, + { + "epoch": 1.46, + "learning_rate": 1.7701734549353728e-05, + "loss": 0.9348, + "step": 19777 + }, + { + "epoch": 1.46, + "learning_rate": 1.7701480129146827e-05, + "loss": 1.0643, + "step": 19778 + }, + { + "epoch": 1.46, + "learning_rate": 1.770122569668698e-05, + "loss": 1.0809, + "step": 19779 + }, + { + "epoch": 1.46, + "learning_rate": 1.770097125197459e-05, + "loss": 0.982, + "step": 19780 + }, + { + "epoch": 1.46, + "learning_rate": 1.7700716795010073e-05, + "loss": 0.9494, + "step": 19781 + }, + { + "epoch": 1.46, + "learning_rate": 1.770046232579382e-05, + "loss": 1.021, + "step": 19782 + }, + { + "epoch": 1.46, + "learning_rate": 1.770020784432624e-05, + "loss": 1.0331, + "step": 19783 + }, + { + "epoch": 1.46, + "learning_rate": 1.7699953350607745e-05, + "loss": 1.0486, + "step": 19784 + }, + { + "epoch": 1.46, + "learning_rate": 1.769969884463873e-05, + "loss": 1.0488, + "step": 19785 + }, + { + "epoch": 1.46, + "learning_rate": 1.769944432641961e-05, + "loss": 1.0202, + "step": 19786 + }, + { + "epoch": 1.46, + "learning_rate": 1.7699189795950777e-05, + "loss": 1.0778, + "step": 19787 + }, + { + "epoch": 1.46, + "learning_rate": 1.7698935253232646e-05, + "loss": 0.8814, + "step": 19788 + }, + { + "epoch": 1.46, + "learning_rate": 1.769868069826562e-05, + "loss": 0.9773, + "step": 19789 + }, + { + "epoch": 1.46, + "learning_rate": 1.7698426131050103e-05, + "loss": 0.8262, + "step": 19790 + }, + { + "epoch": 1.46, + "learning_rate": 1.76981715515865e-05, + "loss": 0.99, + "step": 19791 + }, + { + "epoch": 1.46, + "learning_rate": 1.7697916959875218e-05, + "loss": 1.0001, + "step": 19792 + }, + { + "epoch": 1.46, + "learning_rate": 1.7697662355916656e-05, + "loss": 1.0994, + "step": 19793 + }, + { + "epoch": 1.46, + "learning_rate": 1.7697407739711227e-05, + "loss": 1.0736, + "step": 19794 + }, + { + "epoch": 1.46, + "learning_rate": 1.769715311125933e-05, + "loss": 0.983, + "step": 19795 + }, + { + "epoch": 1.46, + "learning_rate": 1.7696898470561376e-05, + "loss": 1.0372, + "step": 19796 + }, + { + "epoch": 1.46, + "learning_rate": 1.7696643817617763e-05, + "loss": 0.9366, + "step": 19797 + }, + { + "epoch": 1.46, + "learning_rate": 1.76963891524289e-05, + "loss": 0.999, + "step": 19798 + }, + { + "epoch": 1.46, + "learning_rate": 1.7696134474995194e-05, + "loss": 1.0003, + "step": 19799 + }, + { + "epoch": 1.46, + "learning_rate": 1.7695879785317046e-05, + "loss": 1.1103, + "step": 19800 + }, + { + "epoch": 1.46, + "learning_rate": 1.7695625083394868e-05, + "loss": 0.8996, + "step": 19801 + }, + { + "epoch": 1.46, + "learning_rate": 1.7695370369229056e-05, + "loss": 1.0707, + "step": 19802 + }, + { + "epoch": 1.46, + "learning_rate": 1.7695115642820023e-05, + "loss": 0.9786, + "step": 19803 + }, + { + "epoch": 1.46, + "learning_rate": 1.769486090416817e-05, + "loss": 0.8925, + "step": 19804 + }, + { + "epoch": 1.46, + "learning_rate": 1.76946061532739e-05, + "loss": 0.975, + "step": 19805 + }, + { + "epoch": 1.46, + "learning_rate": 1.7694351390137626e-05, + "loss": 0.8842, + "step": 19806 + }, + { + "epoch": 1.46, + "learning_rate": 1.7694096614759745e-05, + "loss": 1.0079, + "step": 19807 + }, + { + "epoch": 1.46, + "learning_rate": 1.769384182714067e-05, + "loss": 0.95, + "step": 19808 + }, + { + "epoch": 1.46, + "learning_rate": 1.7693587027280804e-05, + "loss": 1.0172, + "step": 19809 + }, + { + "epoch": 1.46, + "learning_rate": 1.7693332215180548e-05, + "loss": 0.9608, + "step": 19810 + }, + { + "epoch": 1.46, + "learning_rate": 1.769307739084031e-05, + "loss": 0.9662, + "step": 19811 + }, + { + "epoch": 1.46, + "learning_rate": 1.76928225542605e-05, + "loss": 1.0463, + "step": 19812 + }, + { + "epoch": 1.46, + "learning_rate": 1.769256770544151e-05, + "loss": 0.9141, + "step": 19813 + }, + { + "epoch": 1.46, + "learning_rate": 1.7692312844383766e-05, + "loss": 0.9917, + "step": 19814 + }, + { + "epoch": 1.46, + "learning_rate": 1.7692057971087657e-05, + "loss": 1.0449, + "step": 19815 + }, + { + "epoch": 1.46, + "learning_rate": 1.7691803085553594e-05, + "loss": 0.9261, + "step": 19816 + }, + { + "epoch": 1.46, + "learning_rate": 1.7691548187781983e-05, + "loss": 1.0316, + "step": 19817 + }, + { + "epoch": 1.46, + "learning_rate": 1.7691293277773225e-05, + "loss": 1.0084, + "step": 19818 + }, + { + "epoch": 1.46, + "learning_rate": 1.7691038355527735e-05, + "loss": 1.0275, + "step": 19819 + }, + { + "epoch": 1.46, + "learning_rate": 1.7690783421045913e-05, + "loss": 0.9125, + "step": 19820 + }, + { + "epoch": 1.46, + "learning_rate": 1.7690528474328163e-05, + "loss": 1.0516, + "step": 19821 + }, + { + "epoch": 1.46, + "learning_rate": 1.7690273515374888e-05, + "loss": 0.993, + "step": 19822 + }, + { + "epoch": 1.46, + "learning_rate": 1.7690018544186506e-05, + "loss": 0.9351, + "step": 19823 + }, + { + "epoch": 1.46, + "learning_rate": 1.7689763560763407e-05, + "loss": 1.0229, + "step": 19824 + }, + { + "epoch": 1.46, + "learning_rate": 1.768950856510601e-05, + "loss": 1.0099, + "step": 19825 + }, + { + "epoch": 1.46, + "learning_rate": 1.7689253557214714e-05, + "loss": 1.0454, + "step": 19826 + }, + { + "epoch": 1.47, + "learning_rate": 1.768899853708992e-05, + "loss": 0.9713, + "step": 19827 + }, + { + "epoch": 1.47, + "learning_rate": 1.7688743504732046e-05, + "loss": 0.9484, + "step": 19828 + }, + { + "epoch": 1.47, + "learning_rate": 1.768848846014149e-05, + "loss": 1.0261, + "step": 19829 + }, + { + "epoch": 1.47, + "learning_rate": 1.7688233403318656e-05, + "loss": 0.9655, + "step": 19830 + }, + { + "epoch": 1.47, + "learning_rate": 1.7687978334263952e-05, + "loss": 1.0085, + "step": 19831 + }, + { + "epoch": 1.47, + "learning_rate": 1.7687723252977787e-05, + "loss": 0.9602, + "step": 19832 + }, + { + "epoch": 1.47, + "learning_rate": 1.7687468159460565e-05, + "loss": 0.9417, + "step": 19833 + }, + { + "epoch": 1.47, + "learning_rate": 1.768721305371269e-05, + "loss": 1.0317, + "step": 19834 + }, + { + "epoch": 1.47, + "learning_rate": 1.768695793573457e-05, + "loss": 0.956, + "step": 19835 + }, + { + "epoch": 1.47, + "learning_rate": 1.768670280552661e-05, + "loss": 0.991, + "step": 19836 + }, + { + "epoch": 1.47, + "learning_rate": 1.768644766308921e-05, + "loss": 0.9934, + "step": 19837 + }, + { + "epoch": 1.47, + "learning_rate": 1.7686192508422787e-05, + "loss": 1.1242, + "step": 19838 + }, + { + "epoch": 1.47, + "learning_rate": 1.7685937341527744e-05, + "loss": 0.9033, + "step": 19839 + }, + { + "epoch": 1.47, + "learning_rate": 1.768568216240448e-05, + "loss": 0.927, + "step": 19840 + }, + { + "epoch": 1.47, + "learning_rate": 1.7685426971053407e-05, + "loss": 1.0204, + "step": 19841 + }, + { + "epoch": 1.47, + "learning_rate": 1.768517176747493e-05, + "loss": 1.0397, + "step": 19842 + }, + { + "epoch": 1.47, + "learning_rate": 1.7684916551669456e-05, + "loss": 0.9118, + "step": 19843 + }, + { + "epoch": 1.47, + "learning_rate": 1.7684661323637388e-05, + "loss": 1.0946, + "step": 19844 + }, + { + "epoch": 1.47, + "learning_rate": 1.7684406083379134e-05, + "loss": 1.0248, + "step": 19845 + }, + { + "epoch": 1.47, + "learning_rate": 1.7684150830895102e-05, + "loss": 0.9976, + "step": 19846 + }, + { + "epoch": 1.47, + "learning_rate": 1.768389556618569e-05, + "loss": 0.917, + "step": 19847 + }, + { + "epoch": 1.47, + "learning_rate": 1.7683640289251317e-05, + "loss": 0.952, + "step": 19848 + }, + { + "epoch": 1.47, + "learning_rate": 1.768338500009238e-05, + "loss": 1.0659, + "step": 19849 + }, + { + "epoch": 1.47, + "learning_rate": 1.7683129698709285e-05, + "loss": 1.0317, + "step": 19850 + }, + { + "epoch": 1.47, + "learning_rate": 1.768287438510244e-05, + "loss": 1.0612, + "step": 19851 + }, + { + "epoch": 1.47, + "learning_rate": 1.7682619059272256e-05, + "loss": 1.0269, + "step": 19852 + }, + { + "epoch": 1.47, + "learning_rate": 1.7682363721219133e-05, + "loss": 1.0767, + "step": 19853 + }, + { + "epoch": 1.47, + "learning_rate": 1.768210837094348e-05, + "loss": 1.0421, + "step": 19854 + }, + { + "epoch": 1.47, + "learning_rate": 1.7681853008445704e-05, + "loss": 1.0967, + "step": 19855 + }, + { + "epoch": 1.47, + "learning_rate": 1.7681597633726208e-05, + "loss": 0.9629, + "step": 19856 + }, + { + "epoch": 1.47, + "learning_rate": 1.76813422467854e-05, + "loss": 0.9535, + "step": 19857 + }, + { + "epoch": 1.47, + "learning_rate": 1.7681086847623686e-05, + "loss": 0.8991, + "step": 19858 + }, + { + "epoch": 1.47, + "learning_rate": 1.7680831436241473e-05, + "loss": 1.1172, + "step": 19859 + }, + { + "epoch": 1.47, + "learning_rate": 1.7680576012639164e-05, + "loss": 1.0181, + "step": 19860 + }, + { + "epoch": 1.47, + "learning_rate": 1.768032057681717e-05, + "loss": 1.0616, + "step": 19861 + }, + { + "epoch": 1.47, + "learning_rate": 1.76800651287759e-05, + "loss": 1.1208, + "step": 19862 + }, + { + "epoch": 1.47, + "learning_rate": 1.7679809668515754e-05, + "loss": 0.9715, + "step": 19863 + }, + { + "epoch": 1.47, + "learning_rate": 1.767955419603714e-05, + "loss": 1.0967, + "step": 19864 + }, + { + "epoch": 1.47, + "learning_rate": 1.7679298711340464e-05, + "loss": 1.0011, + "step": 19865 + }, + { + "epoch": 1.47, + "learning_rate": 1.7679043214426135e-05, + "loss": 1.0624, + "step": 19866 + }, + { + "epoch": 1.47, + "learning_rate": 1.767878770529456e-05, + "loss": 0.9767, + "step": 19867 + }, + { + "epoch": 1.47, + "learning_rate": 1.767853218394614e-05, + "loss": 1.0411, + "step": 19868 + }, + { + "epoch": 1.47, + "learning_rate": 1.7678276650381286e-05, + "loss": 1.0709, + "step": 19869 + }, + { + "epoch": 1.47, + "learning_rate": 1.7678021104600404e-05, + "loss": 1.1265, + "step": 19870 + }, + { + "epoch": 1.47, + "learning_rate": 1.7677765546603903e-05, + "loss": 1.0804, + "step": 19871 + }, + { + "epoch": 1.47, + "learning_rate": 1.7677509976392184e-05, + "loss": 0.9452, + "step": 19872 + }, + { + "epoch": 1.47, + "learning_rate": 1.767725439396566e-05, + "loss": 0.9409, + "step": 19873 + }, + { + "epoch": 1.47, + "learning_rate": 1.767699879932473e-05, + "loss": 0.9771, + "step": 19874 + }, + { + "epoch": 1.47, + "learning_rate": 1.7676743192469806e-05, + "loss": 1.0635, + "step": 19875 + }, + { + "epoch": 1.47, + "learning_rate": 1.7676487573401292e-05, + "loss": 1.0042, + "step": 19876 + }, + { + "epoch": 1.47, + "learning_rate": 1.76762319421196e-05, + "loss": 1.0971, + "step": 19877 + }, + { + "epoch": 1.47, + "learning_rate": 1.7675976298625127e-05, + "loss": 1.0798, + "step": 19878 + }, + { + "epoch": 1.47, + "learning_rate": 1.767572064291829e-05, + "loss": 0.8876, + "step": 19879 + }, + { + "epoch": 1.47, + "learning_rate": 1.767546497499949e-05, + "loss": 1.0814, + "step": 19880 + }, + { + "epoch": 1.47, + "learning_rate": 1.7675209294869134e-05, + "loss": 0.961, + "step": 19881 + }, + { + "epoch": 1.47, + "learning_rate": 1.767495360252763e-05, + "loss": 0.9755, + "step": 19882 + }, + { + "epoch": 1.47, + "learning_rate": 1.767469789797539e-05, + "loss": 1.0245, + "step": 19883 + }, + { + "epoch": 1.47, + "learning_rate": 1.767444218121281e-05, + "loss": 1.017, + "step": 19884 + }, + { + "epoch": 1.47, + "learning_rate": 1.7674186452240302e-05, + "loss": 1.0636, + "step": 19885 + }, + { + "epoch": 1.47, + "learning_rate": 1.7673930711058273e-05, + "loss": 1.0687, + "step": 19886 + }, + { + "epoch": 1.47, + "learning_rate": 1.7673674957667135e-05, + "loss": 1.1025, + "step": 19887 + }, + { + "epoch": 1.47, + "learning_rate": 1.7673419192067283e-05, + "loss": 0.9029, + "step": 19888 + }, + { + "epoch": 1.47, + "learning_rate": 1.7673163414259136e-05, + "loss": 0.9959, + "step": 19889 + }, + { + "epoch": 1.47, + "learning_rate": 1.7672907624243093e-05, + "loss": 1.0211, + "step": 19890 + }, + { + "epoch": 1.47, + "learning_rate": 1.7672651822019567e-05, + "loss": 0.9909, + "step": 19891 + }, + { + "epoch": 1.47, + "learning_rate": 1.7672396007588957e-05, + "loss": 1.1559, + "step": 19892 + }, + { + "epoch": 1.47, + "learning_rate": 1.7672140180951678e-05, + "loss": 1.0329, + "step": 19893 + }, + { + "epoch": 1.47, + "learning_rate": 1.7671884342108136e-05, + "loss": 1.0702, + "step": 19894 + }, + { + "epoch": 1.47, + "learning_rate": 1.7671628491058734e-05, + "loss": 0.9911, + "step": 19895 + }, + { + "epoch": 1.47, + "learning_rate": 1.7671372627803878e-05, + "loss": 0.9292, + "step": 19896 + }, + { + "epoch": 1.47, + "learning_rate": 1.767111675234398e-05, + "loss": 1.0693, + "step": 19897 + }, + { + "epoch": 1.47, + "learning_rate": 1.7670860864679443e-05, + "loss": 1.1118, + "step": 19898 + }, + { + "epoch": 1.47, + "learning_rate": 1.7670604964810677e-05, + "loss": 1.0154, + "step": 19899 + }, + { + "epoch": 1.47, + "learning_rate": 1.767034905273809e-05, + "loss": 1.0338, + "step": 19900 + }, + { + "epoch": 1.47, + "learning_rate": 1.7670093128462087e-05, + "loss": 1.0322, + "step": 19901 + }, + { + "epoch": 1.47, + "learning_rate": 1.7669837191983076e-05, + "loss": 1.1082, + "step": 19902 + }, + { + "epoch": 1.47, + "learning_rate": 1.766958124330146e-05, + "loss": 1.1598, + "step": 19903 + }, + { + "epoch": 1.47, + "learning_rate": 1.7669325282417653e-05, + "loss": 0.9692, + "step": 19904 + }, + { + "epoch": 1.47, + "learning_rate": 1.766906930933206e-05, + "loss": 1.0243, + "step": 19905 + }, + { + "epoch": 1.47, + "learning_rate": 1.7668813324045086e-05, + "loss": 0.9757, + "step": 19906 + }, + { + "epoch": 1.47, + "learning_rate": 1.7668557326557142e-05, + "loss": 1.0772, + "step": 19907 + }, + { + "epoch": 1.47, + "learning_rate": 1.766830131686863e-05, + "loss": 0.9607, + "step": 19908 + }, + { + "epoch": 1.47, + "learning_rate": 1.7668045294979963e-05, + "loss": 0.9448, + "step": 19909 + }, + { + "epoch": 1.47, + "learning_rate": 1.7667789260891542e-05, + "loss": 1.0503, + "step": 19910 + }, + { + "epoch": 1.47, + "learning_rate": 1.766753321460378e-05, + "loss": 0.9964, + "step": 19911 + }, + { + "epoch": 1.47, + "learning_rate": 1.7667277156117083e-05, + "loss": 1.014, + "step": 19912 + }, + { + "epoch": 1.47, + "learning_rate": 1.7667021085431858e-05, + "loss": 1.0412, + "step": 19913 + }, + { + "epoch": 1.47, + "learning_rate": 1.766676500254851e-05, + "loss": 0.9793, + "step": 19914 + }, + { + "epoch": 1.47, + "learning_rate": 1.7666508907467453e-05, + "loss": 1.0319, + "step": 19915 + }, + { + "epoch": 1.47, + "learning_rate": 1.7666252800189087e-05, + "loss": 1.0019, + "step": 19916 + }, + { + "epoch": 1.47, + "learning_rate": 1.7665996680713822e-05, + "loss": 0.9784, + "step": 19917 + }, + { + "epoch": 1.47, + "learning_rate": 1.7665740549042066e-05, + "loss": 1.0355, + "step": 19918 + }, + { + "epoch": 1.47, + "learning_rate": 1.7665484405174224e-05, + "loss": 0.9778, + "step": 19919 + }, + { + "epoch": 1.47, + "learning_rate": 1.766522824911071e-05, + "loss": 0.9776, + "step": 19920 + }, + { + "epoch": 1.47, + "learning_rate": 1.7664972080851925e-05, + "loss": 0.9266, + "step": 19921 + }, + { + "epoch": 1.47, + "learning_rate": 1.766471590039828e-05, + "loss": 1.0083, + "step": 19922 + }, + { + "epoch": 1.47, + "learning_rate": 1.7664459707750185e-05, + "loss": 1.0482, + "step": 19923 + }, + { + "epoch": 1.47, + "learning_rate": 1.7664203502908043e-05, + "loss": 0.9753, + "step": 19924 + }, + { + "epoch": 1.47, + "learning_rate": 1.766394728587226e-05, + "loss": 0.9602, + "step": 19925 + }, + { + "epoch": 1.47, + "learning_rate": 1.7663691056643245e-05, + "loss": 1.1159, + "step": 19926 + }, + { + "epoch": 1.47, + "learning_rate": 1.7663434815221413e-05, + "loss": 1.029, + "step": 19927 + }, + { + "epoch": 1.47, + "learning_rate": 1.766317856160716e-05, + "loss": 0.9789, + "step": 19928 + }, + { + "epoch": 1.47, + "learning_rate": 1.7662922295800904e-05, + "loss": 0.9199, + "step": 19929 + }, + { + "epoch": 1.47, + "learning_rate": 1.7662666017803047e-05, + "loss": 1.0314, + "step": 19930 + }, + { + "epoch": 1.47, + "learning_rate": 1.7662409727613997e-05, + "loss": 1.0046, + "step": 19931 + }, + { + "epoch": 1.47, + "learning_rate": 1.7662153425234165e-05, + "loss": 1.0388, + "step": 19932 + }, + { + "epoch": 1.47, + "learning_rate": 1.7661897110663956e-05, + "loss": 1.024, + "step": 19933 + }, + { + "epoch": 1.47, + "learning_rate": 1.766164078390377e-05, + "loss": 1.1148, + "step": 19934 + }, + { + "epoch": 1.47, + "learning_rate": 1.7661384444954033e-05, + "loss": 0.9161, + "step": 19935 + }, + { + "epoch": 1.47, + "learning_rate": 1.7661128093815143e-05, + "loss": 0.9266, + "step": 19936 + }, + { + "epoch": 1.47, + "learning_rate": 1.76608717304875e-05, + "loss": 0.989, + "step": 19937 + }, + { + "epoch": 1.47, + "learning_rate": 1.7660615354971528e-05, + "loss": 1.1002, + "step": 19938 + }, + { + "epoch": 1.47, + "learning_rate": 1.7660358967267624e-05, + "loss": 1.0258, + "step": 19939 + }, + { + "epoch": 1.47, + "learning_rate": 1.76601025673762e-05, + "loss": 0.9776, + "step": 19940 + }, + { + "epoch": 1.47, + "learning_rate": 1.765984615529766e-05, + "loss": 1.0155, + "step": 19941 + }, + { + "epoch": 1.47, + "learning_rate": 1.7659589731032415e-05, + "loss": 0.9431, + "step": 19942 + }, + { + "epoch": 1.47, + "learning_rate": 1.765933329458087e-05, + "loss": 0.9991, + "step": 19943 + }, + { + "epoch": 1.47, + "learning_rate": 1.765907684594344e-05, + "loss": 1.0442, + "step": 19944 + }, + { + "epoch": 1.47, + "learning_rate": 1.7658820385120523e-05, + "loss": 0.9731, + "step": 19945 + }, + { + "epoch": 1.47, + "learning_rate": 1.7658563912112538e-05, + "loss": 1.1517, + "step": 19946 + }, + { + "epoch": 1.47, + "learning_rate": 1.765830742691988e-05, + "loss": 1.0716, + "step": 19947 + }, + { + "epoch": 1.47, + "learning_rate": 1.765805092954297e-05, + "loss": 0.9906, + "step": 19948 + }, + { + "epoch": 1.47, + "learning_rate": 1.765779441998221e-05, + "loss": 1.0552, + "step": 19949 + }, + { + "epoch": 1.47, + "learning_rate": 1.765753789823801e-05, + "loss": 0.9736, + "step": 19950 + }, + { + "epoch": 1.47, + "learning_rate": 1.7657281364310774e-05, + "loss": 0.9851, + "step": 19951 + }, + { + "epoch": 1.47, + "learning_rate": 1.765702481820091e-05, + "loss": 0.9969, + "step": 19952 + }, + { + "epoch": 1.47, + "learning_rate": 1.7656768259908834e-05, + "loss": 0.9797, + "step": 19953 + }, + { + "epoch": 1.47, + "learning_rate": 1.765651168943495e-05, + "loss": 0.9739, + "step": 19954 + }, + { + "epoch": 1.47, + "learning_rate": 1.7656255106779663e-05, + "loss": 0.9826, + "step": 19955 + }, + { + "epoch": 1.47, + "learning_rate": 1.7655998511943386e-05, + "loss": 0.9575, + "step": 19956 + }, + { + "epoch": 1.47, + "learning_rate": 1.765574190492652e-05, + "loss": 0.9322, + "step": 19957 + }, + { + "epoch": 1.47, + "learning_rate": 1.765548528572948e-05, + "loss": 1.05, + "step": 19958 + }, + { + "epoch": 1.47, + "learning_rate": 1.7655228654352674e-05, + "loss": 1.0625, + "step": 19959 + }, + { + "epoch": 1.47, + "learning_rate": 1.765497201079651e-05, + "loss": 0.9811, + "step": 19960 + }, + { + "epoch": 1.47, + "learning_rate": 1.7654715355061395e-05, + "loss": 0.9458, + "step": 19961 + }, + { + "epoch": 1.47, + "learning_rate": 1.7654458687147733e-05, + "loss": 0.9698, + "step": 19962 + }, + { + "epoch": 1.48, + "learning_rate": 1.765420200705594e-05, + "loss": 0.969, + "step": 19963 + }, + { + "epoch": 1.48, + "learning_rate": 1.7653945314786422e-05, + "loss": 1.0247, + "step": 19964 + }, + { + "epoch": 1.48, + "learning_rate": 1.7653688610339586e-05, + "loss": 1.0664, + "step": 19965 + }, + { + "epoch": 1.48, + "learning_rate": 1.765343189371584e-05, + "loss": 1.1151, + "step": 19966 + }, + { + "epoch": 1.48, + "learning_rate": 1.765317516491559e-05, + "loss": 1.0803, + "step": 19967 + }, + { + "epoch": 1.48, + "learning_rate": 1.7652918423939252e-05, + "loss": 0.956, + "step": 19968 + }, + { + "epoch": 1.48, + "learning_rate": 1.765266167078723e-05, + "loss": 1.0125, + "step": 19969 + }, + { + "epoch": 1.48, + "learning_rate": 1.7652404905459933e-05, + "loss": 0.9577, + "step": 19970 + }, + { + "epoch": 1.48, + "learning_rate": 1.765214812795777e-05, + "loss": 0.9385, + "step": 19971 + }, + { + "epoch": 1.48, + "learning_rate": 1.765189133828115e-05, + "loss": 0.8499, + "step": 19972 + }, + { + "epoch": 1.48, + "learning_rate": 1.7651634536430477e-05, + "loss": 1.052, + "step": 19973 + }, + { + "epoch": 1.48, + "learning_rate": 1.7651377722406163e-05, + "loss": 1.0369, + "step": 19974 + }, + { + "epoch": 1.48, + "learning_rate": 1.7651120896208616e-05, + "loss": 1.0144, + "step": 19975 + }, + { + "epoch": 1.48, + "learning_rate": 1.7650864057838248e-05, + "loss": 1.0038, + "step": 19976 + }, + { + "epoch": 1.48, + "learning_rate": 1.765060720729546e-05, + "loss": 0.9305, + "step": 19977 + }, + { + "epoch": 1.48, + "learning_rate": 1.765035034458067e-05, + "loss": 0.9923, + "step": 19978 + }, + { + "epoch": 1.48, + "learning_rate": 1.765009346969428e-05, + "loss": 0.9807, + "step": 19979 + }, + { + "epoch": 1.48, + "learning_rate": 1.7649836582636702e-05, + "loss": 1.003, + "step": 19980 + }, + { + "epoch": 1.48, + "learning_rate": 1.764957968340834e-05, + "loss": 1.2089, + "step": 19981 + }, + { + "epoch": 1.48, + "learning_rate": 1.764932277200961e-05, + "loss": 0.9822, + "step": 19982 + }, + { + "epoch": 1.48, + "learning_rate": 1.7649065848440917e-05, + "loss": 0.9215, + "step": 19983 + }, + { + "epoch": 1.48, + "learning_rate": 1.7648808912702668e-05, + "loss": 1.0201, + "step": 19984 + }, + { + "epoch": 1.48, + "learning_rate": 1.7648551964795272e-05, + "loss": 0.9665, + "step": 19985 + }, + { + "epoch": 1.48, + "learning_rate": 1.7648295004719144e-05, + "loss": 0.9717, + "step": 19986 + }, + { + "epoch": 1.48, + "learning_rate": 1.7648038032474683e-05, + "loss": 1.0187, + "step": 19987 + }, + { + "epoch": 1.48, + "learning_rate": 1.7647781048062308e-05, + "loss": 0.9174, + "step": 19988 + }, + { + "epoch": 1.48, + "learning_rate": 1.7647524051482418e-05, + "loss": 0.9185, + "step": 19989 + }, + { + "epoch": 1.48, + "learning_rate": 1.764726704273543e-05, + "loss": 1.0257, + "step": 19990 + }, + { + "epoch": 1.48, + "learning_rate": 1.7647010021821747e-05, + "loss": 0.9595, + "step": 19991 + }, + { + "epoch": 1.48, + "learning_rate": 1.7646752988741783e-05, + "loss": 1.0313, + "step": 19992 + }, + { + "epoch": 1.48, + "learning_rate": 1.7646495943495943e-05, + "loss": 1.0516, + "step": 19993 + }, + { + "epoch": 1.48, + "learning_rate": 1.7646238886084637e-05, + "loss": 1.0511, + "step": 19994 + }, + { + "epoch": 1.48, + "learning_rate": 1.7645981816508277e-05, + "loss": 0.9993, + "step": 19995 + }, + { + "epoch": 1.48, + "learning_rate": 1.7645724734767267e-05, + "loss": 0.9438, + "step": 19996 + }, + { + "epoch": 1.48, + "learning_rate": 1.7645467640862016e-05, + "loss": 1.018, + "step": 19997 + }, + { + "epoch": 1.48, + "learning_rate": 1.764521053479294e-05, + "loss": 1.0135, + "step": 19998 + }, + { + "epoch": 1.48, + "learning_rate": 1.764495341656044e-05, + "loss": 1.0506, + "step": 19999 + }, + { + "epoch": 1.48, + "learning_rate": 1.7644696286164928e-05, + "loss": 1.021, + "step": 20000 + }, + { + "epoch": 1.48, + "learning_rate": 1.7644439143606816e-05, + "loss": 1.0069, + "step": 20001 + }, + { + "epoch": 1.48, + "learning_rate": 1.764418198888651e-05, + "loss": 1.0667, + "step": 20002 + }, + { + "epoch": 1.48, + "learning_rate": 1.7643924822004423e-05, + "loss": 0.9731, + "step": 20003 + }, + { + "epoch": 1.48, + "learning_rate": 1.7643667642960957e-05, + "loss": 0.9713, + "step": 20004 + }, + { + "epoch": 1.48, + "learning_rate": 1.7643410451756527e-05, + "loss": 0.9965, + "step": 20005 + }, + { + "epoch": 1.48, + "learning_rate": 1.7643153248391536e-05, + "loss": 1.0205, + "step": 20006 + }, + { + "epoch": 1.48, + "learning_rate": 1.7642896032866402e-05, + "loss": 0.9563, + "step": 20007 + }, + { + "epoch": 1.48, + "learning_rate": 1.764263880518153e-05, + "loss": 1.0016, + "step": 20008 + }, + { + "epoch": 1.48, + "learning_rate": 1.7642381565337325e-05, + "loss": 0.8923, + "step": 20009 + }, + { + "epoch": 1.48, + "learning_rate": 1.7642124313334205e-05, + "loss": 1.0158, + "step": 20010 + }, + { + "epoch": 1.48, + "learning_rate": 1.764186704917257e-05, + "loss": 0.9983, + "step": 20011 + }, + { + "epoch": 1.48, + "learning_rate": 1.764160977285284e-05, + "loss": 0.9563, + "step": 20012 + }, + { + "epoch": 1.48, + "learning_rate": 1.7641352484375414e-05, + "loss": 1.0543, + "step": 20013 + }, + { + "epoch": 1.48, + "learning_rate": 1.7641095183740704e-05, + "loss": 0.9612, + "step": 20014 + }, + { + "epoch": 1.48, + "learning_rate": 1.7640837870949123e-05, + "loss": 0.9899, + "step": 20015 + }, + { + "epoch": 1.48, + "learning_rate": 1.7640580546001077e-05, + "loss": 0.7792, + "step": 20016 + }, + { + "epoch": 1.48, + "learning_rate": 1.764032320889698e-05, + "loss": 1.0729, + "step": 20017 + }, + { + "epoch": 1.48, + "learning_rate": 1.7640065859637236e-05, + "loss": 1.1176, + "step": 20018 + }, + { + "epoch": 1.48, + "learning_rate": 1.7639808498222257e-05, + "loss": 0.9922, + "step": 20019 + }, + { + "epoch": 1.48, + "learning_rate": 1.763955112465245e-05, + "loss": 1.0373, + "step": 20020 + }, + { + "epoch": 1.48, + "learning_rate": 1.7639293738928226e-05, + "loss": 0.9678, + "step": 20021 + }, + { + "epoch": 1.48, + "learning_rate": 1.7639036341049997e-05, + "loss": 0.9388, + "step": 20022 + }, + { + "epoch": 1.48, + "learning_rate": 1.7638778931018173e-05, + "loss": 1.0484, + "step": 20023 + }, + { + "epoch": 1.48, + "learning_rate": 1.7638521508833157e-05, + "loss": 1.152, + "step": 20024 + }, + { + "epoch": 1.48, + "learning_rate": 1.7638264074495364e-05, + "loss": 1.0152, + "step": 20025 + }, + { + "epoch": 1.48, + "learning_rate": 1.7638006628005202e-05, + "loss": 0.9603, + "step": 20026 + }, + { + "epoch": 1.48, + "learning_rate": 1.7637749169363082e-05, + "loss": 1.1387, + "step": 20027 + }, + { + "epoch": 1.48, + "learning_rate": 1.763749169856941e-05, + "loss": 1.0536, + "step": 20028 + }, + { + "epoch": 1.48, + "learning_rate": 1.7637234215624597e-05, + "loss": 0.9774, + "step": 20029 + }, + { + "epoch": 1.48, + "learning_rate": 1.763697672052906e-05, + "loss": 1.0642, + "step": 20030 + }, + { + "epoch": 1.48, + "learning_rate": 1.7636719213283196e-05, + "loss": 1.0189, + "step": 20031 + }, + { + "epoch": 1.48, + "learning_rate": 1.7636461693887425e-05, + "loss": 1.0244, + "step": 20032 + }, + { + "epoch": 1.48, + "learning_rate": 1.7636204162342148e-05, + "loss": 1.0334, + "step": 20033 + }, + { + "epoch": 1.48, + "learning_rate": 1.7635946618647783e-05, + "loss": 0.9907, + "step": 20034 + }, + { + "epoch": 1.48, + "learning_rate": 1.7635689062804736e-05, + "loss": 0.9843, + "step": 20035 + }, + { + "epoch": 1.48, + "learning_rate": 1.7635431494813414e-05, + "loss": 1.0427, + "step": 20036 + }, + { + "epoch": 1.48, + "learning_rate": 1.7635173914674233e-05, + "loss": 1.0004, + "step": 20037 + }, + { + "epoch": 1.48, + "learning_rate": 1.7634916322387597e-05, + "loss": 1.0495, + "step": 20038 + }, + { + "epoch": 1.48, + "learning_rate": 1.7634658717953923e-05, + "loss": 0.9172, + "step": 20039 + }, + { + "epoch": 1.48, + "learning_rate": 1.7634401101373612e-05, + "loss": 1.0354, + "step": 20040 + }, + { + "epoch": 1.48, + "learning_rate": 1.763414347264708e-05, + "loss": 0.9034, + "step": 20041 + }, + { + "epoch": 1.48, + "learning_rate": 1.7633885831774736e-05, + "loss": 0.9376, + "step": 20042 + }, + { + "epoch": 1.48, + "learning_rate": 1.7633628178756986e-05, + "loss": 0.9274, + "step": 20043 + }, + { + "epoch": 1.48, + "learning_rate": 1.7633370513594244e-05, + "loss": 0.9972, + "step": 20044 + }, + { + "epoch": 1.48, + "learning_rate": 1.763311283628692e-05, + "loss": 1.0123, + "step": 20045 + }, + { + "epoch": 1.48, + "learning_rate": 1.763285514683542e-05, + "loss": 1.0372, + "step": 20046 + }, + { + "epoch": 1.48, + "learning_rate": 1.7632597445240154e-05, + "loss": 1.0364, + "step": 20047 + }, + { + "epoch": 1.48, + "learning_rate": 1.7632339731501543e-05, + "loss": 0.9149, + "step": 20048 + }, + { + "epoch": 1.48, + "learning_rate": 1.7632082005619983e-05, + "loss": 1.1423, + "step": 20049 + }, + { + "epoch": 1.48, + "learning_rate": 1.763182426759589e-05, + "loss": 1.0403, + "step": 20050 + }, + { + "epoch": 1.48, + "learning_rate": 1.7631566517429676e-05, + "loss": 1.0095, + "step": 20051 + }, + { + "epoch": 1.48, + "learning_rate": 1.7631308755121748e-05, + "loss": 1.0093, + "step": 20052 + }, + { + "epoch": 1.48, + "learning_rate": 1.7631050980672518e-05, + "loss": 0.9948, + "step": 20053 + }, + { + "epoch": 1.48, + "learning_rate": 1.7630793194082395e-05, + "loss": 1.0294, + "step": 20054 + }, + { + "epoch": 1.48, + "learning_rate": 1.7630535395351787e-05, + "loss": 0.9976, + "step": 20055 + }, + { + "epoch": 1.48, + "learning_rate": 1.7630277584481106e-05, + "loss": 0.9937, + "step": 20056 + }, + { + "epoch": 1.48, + "learning_rate": 1.7630019761470762e-05, + "loss": 1.0213, + "step": 20057 + }, + { + "epoch": 1.48, + "learning_rate": 1.762976192632117e-05, + "loss": 0.9846, + "step": 20058 + }, + { + "epoch": 1.48, + "learning_rate": 1.7629504079032733e-05, + "loss": 0.9797, + "step": 20059 + }, + { + "epoch": 1.48, + "learning_rate": 1.7629246219605865e-05, + "loss": 1.0413, + "step": 20060 + }, + { + "epoch": 1.48, + "learning_rate": 1.7628988348040973e-05, + "loss": 0.96, + "step": 20061 + }, + { + "epoch": 1.48, + "learning_rate": 1.7628730464338476e-05, + "loss": 0.9368, + "step": 20062 + }, + { + "epoch": 1.48, + "learning_rate": 1.7628472568498772e-05, + "loss": 1.0432, + "step": 20063 + }, + { + "epoch": 1.48, + "learning_rate": 1.762821466052228e-05, + "loss": 1.0985, + "step": 20064 + }, + { + "epoch": 1.48, + "learning_rate": 1.7627956740409406e-05, + "loss": 1.0939, + "step": 20065 + }, + { + "epoch": 1.48, + "learning_rate": 1.7627698808160564e-05, + "loss": 1.1302, + "step": 20066 + }, + { + "epoch": 1.48, + "learning_rate": 1.762744086377616e-05, + "loss": 1.0912, + "step": 20067 + }, + { + "epoch": 1.48, + "learning_rate": 1.7627182907256607e-05, + "loss": 0.9413, + "step": 20068 + }, + { + "epoch": 1.48, + "learning_rate": 1.7626924938602316e-05, + "loss": 1.0058, + "step": 20069 + }, + { + "epoch": 1.48, + "learning_rate": 1.7626666957813695e-05, + "loss": 0.9909, + "step": 20070 + }, + { + "epoch": 1.48, + "learning_rate": 1.7626408964891158e-05, + "loss": 0.9993, + "step": 20071 + }, + { + "epoch": 1.48, + "learning_rate": 1.7626150959835114e-05, + "loss": 1.0323, + "step": 20072 + }, + { + "epoch": 1.48, + "learning_rate": 1.762589294264597e-05, + "loss": 1.088, + "step": 20073 + }, + { + "epoch": 1.48, + "learning_rate": 1.762563491332414e-05, + "loss": 1.0519, + "step": 20074 + }, + { + "epoch": 1.48, + "learning_rate": 1.7625376871870037e-05, + "loss": 0.966, + "step": 20075 + }, + { + "epoch": 1.48, + "learning_rate": 1.7625118818284067e-05, + "loss": 1.0821, + "step": 20076 + }, + { + "epoch": 1.48, + "learning_rate": 1.762486075256664e-05, + "loss": 0.9236, + "step": 20077 + }, + { + "epoch": 1.48, + "learning_rate": 1.762460267471817e-05, + "loss": 1.0231, + "step": 20078 + }, + { + "epoch": 1.48, + "learning_rate": 1.762434458473907e-05, + "loss": 0.9983, + "step": 20079 + }, + { + "epoch": 1.48, + "learning_rate": 1.7624086482629742e-05, + "loss": 0.9765, + "step": 20080 + }, + { + "epoch": 1.48, + "learning_rate": 1.7623828368390603e-05, + "loss": 0.985, + "step": 20081 + }, + { + "epoch": 1.48, + "learning_rate": 1.762357024202206e-05, + "loss": 0.947, + "step": 20082 + }, + { + "epoch": 1.48, + "learning_rate": 1.762331210352453e-05, + "loss": 0.9869, + "step": 20083 + }, + { + "epoch": 1.48, + "learning_rate": 1.7623053952898417e-05, + "loss": 1.004, + "step": 20084 + }, + { + "epoch": 1.48, + "learning_rate": 1.7622795790144134e-05, + "loss": 1.062, + "step": 20085 + }, + { + "epoch": 1.48, + "learning_rate": 1.762253761526209e-05, + "loss": 1.0177, + "step": 20086 + }, + { + "epoch": 1.48, + "learning_rate": 1.76222794282527e-05, + "loss": 0.9436, + "step": 20087 + }, + { + "epoch": 1.48, + "learning_rate": 1.7622021229116377e-05, + "loss": 0.9424, + "step": 20088 + }, + { + "epoch": 1.48, + "learning_rate": 1.762176301785352e-05, + "loss": 1.084, + "step": 20089 + }, + { + "epoch": 1.48, + "learning_rate": 1.7621504794464553e-05, + "loss": 1.0327, + "step": 20090 + }, + { + "epoch": 1.48, + "learning_rate": 1.762124655894988e-05, + "loss": 0.9507, + "step": 20091 + }, + { + "epoch": 1.48, + "learning_rate": 1.7620988311309907e-05, + "loss": 0.9975, + "step": 20092 + }, + { + "epoch": 1.48, + "learning_rate": 1.7620730051545053e-05, + "loss": 1.0962, + "step": 20093 + }, + { + "epoch": 1.48, + "learning_rate": 1.762047177965573e-05, + "loss": 0.9919, + "step": 20094 + }, + { + "epoch": 1.48, + "learning_rate": 1.7620213495642346e-05, + "loss": 1.093, + "step": 20095 + }, + { + "epoch": 1.48, + "learning_rate": 1.7619955199505306e-05, + "loss": 1.1234, + "step": 20096 + }, + { + "epoch": 1.48, + "learning_rate": 1.761969689124503e-05, + "loss": 0.9975, + "step": 20097 + }, + { + "epoch": 1.49, + "learning_rate": 1.7619438570861924e-05, + "loss": 0.9411, + "step": 20098 + }, + { + "epoch": 1.49, + "learning_rate": 1.76191802383564e-05, + "loss": 1.0436, + "step": 20099 + }, + { + "epoch": 1.49, + "learning_rate": 1.761892189372887e-05, + "loss": 1.0657, + "step": 20100 + }, + { + "epoch": 1.49, + "learning_rate": 1.7618663536979745e-05, + "loss": 1.0658, + "step": 20101 + }, + { + "epoch": 1.49, + "learning_rate": 1.7618405168109434e-05, + "loss": 1.0419, + "step": 20102 + }, + { + "epoch": 1.49, + "learning_rate": 1.7618146787118352e-05, + "loss": 0.9362, + "step": 20103 + }, + { + "epoch": 1.49, + "learning_rate": 1.7617888394006903e-05, + "loss": 0.9246, + "step": 20104 + }, + { + "epoch": 1.49, + "learning_rate": 1.7617629988775503e-05, + "loss": 0.9291, + "step": 20105 + }, + { + "epoch": 1.49, + "learning_rate": 1.7617371571424566e-05, + "loss": 1.1979, + "step": 20106 + }, + { + "epoch": 1.49, + "learning_rate": 1.76171131419545e-05, + "loss": 1.0428, + "step": 20107 + }, + { + "epoch": 1.49, + "learning_rate": 1.7616854700365714e-05, + "loss": 1.0911, + "step": 20108 + }, + { + "epoch": 1.49, + "learning_rate": 1.7616596246658623e-05, + "loss": 1.0676, + "step": 20109 + }, + { + "epoch": 1.49, + "learning_rate": 1.7616337780833633e-05, + "loss": 1.0486, + "step": 20110 + }, + { + "epoch": 1.49, + "learning_rate": 1.761607930289116e-05, + "loss": 1.0686, + "step": 20111 + }, + { + "epoch": 1.49, + "learning_rate": 1.7615820812831613e-05, + "loss": 0.9883, + "step": 20112 + }, + { + "epoch": 1.49, + "learning_rate": 1.761556231065541e-05, + "loss": 1.0863, + "step": 20113 + }, + { + "epoch": 1.49, + "learning_rate": 1.7615303796362947e-05, + "loss": 1.0626, + "step": 20114 + }, + { + "epoch": 1.49, + "learning_rate": 1.761504526995465e-05, + "loss": 0.9259, + "step": 20115 + }, + { + "epoch": 1.49, + "learning_rate": 1.7614786731430925e-05, + "loss": 1.0136, + "step": 20116 + }, + { + "epoch": 1.49, + "learning_rate": 1.761452818079218e-05, + "loss": 1.0432, + "step": 20117 + }, + { + "epoch": 1.49, + "learning_rate": 1.7614269618038834e-05, + "loss": 1.0902, + "step": 20118 + }, + { + "epoch": 1.49, + "learning_rate": 1.7614011043171292e-05, + "loss": 1.0593, + "step": 20119 + }, + { + "epoch": 1.49, + "learning_rate": 1.7613752456189966e-05, + "loss": 0.9441, + "step": 20120 + }, + { + "epoch": 1.49, + "learning_rate": 1.7613493857095273e-05, + "loss": 0.9737, + "step": 20121 + }, + { + "epoch": 1.49, + "learning_rate": 1.7613235245887615e-05, + "loss": 1.0122, + "step": 20122 + }, + { + "epoch": 1.49, + "learning_rate": 1.761297662256741e-05, + "loss": 0.9428, + "step": 20123 + }, + { + "epoch": 1.49, + "learning_rate": 1.761271798713507e-05, + "loss": 1.0727, + "step": 20124 + }, + { + "epoch": 1.49, + "learning_rate": 1.7612459339591005e-05, + "loss": 1.0885, + "step": 20125 + }, + { + "epoch": 1.49, + "learning_rate": 1.7612200679935624e-05, + "loss": 1.0582, + "step": 20126 + }, + { + "epoch": 1.49, + "learning_rate": 1.761194200816934e-05, + "loss": 1.1132, + "step": 20127 + }, + { + "epoch": 1.49, + "learning_rate": 1.7611683324292567e-05, + "loss": 1.0304, + "step": 20128 + }, + { + "epoch": 1.49, + "learning_rate": 1.7611424628305714e-05, + "loss": 1.0592, + "step": 20129 + }, + { + "epoch": 1.49, + "learning_rate": 1.7611165920209194e-05, + "loss": 1.112, + "step": 20130 + }, + { + "epoch": 1.49, + "learning_rate": 1.761090720000342e-05, + "loss": 0.9413, + "step": 20131 + }, + { + "epoch": 1.49, + "learning_rate": 1.7610648467688796e-05, + "loss": 0.9653, + "step": 20132 + }, + { + "epoch": 1.49, + "learning_rate": 1.7610389723265743e-05, + "loss": 0.9823, + "step": 20133 + }, + { + "epoch": 1.49, + "learning_rate": 1.761013096673467e-05, + "loss": 1.0705, + "step": 20134 + }, + { + "epoch": 1.49, + "learning_rate": 1.7609872198095984e-05, + "loss": 1.0143, + "step": 20135 + }, + { + "epoch": 1.49, + "learning_rate": 1.76096134173501e-05, + "loss": 0.9762, + "step": 20136 + }, + { + "epoch": 1.49, + "learning_rate": 1.760935462449743e-05, + "loss": 0.9687, + "step": 20137 + }, + { + "epoch": 1.49, + "learning_rate": 1.760909581953839e-05, + "loss": 0.9978, + "step": 20138 + }, + { + "epoch": 1.49, + "learning_rate": 1.7608837002473388e-05, + "loss": 1.025, + "step": 20139 + }, + { + "epoch": 1.49, + "learning_rate": 1.760857817330283e-05, + "loss": 1.0277, + "step": 20140 + }, + { + "epoch": 1.49, + "learning_rate": 1.7608319332027135e-05, + "loss": 0.9412, + "step": 20141 + }, + { + "epoch": 1.49, + "learning_rate": 1.760806047864671e-05, + "loss": 1.0649, + "step": 20142 + }, + { + "epoch": 1.49, + "learning_rate": 1.7607801613161974e-05, + "loss": 0.9259, + "step": 20143 + }, + { + "epoch": 1.49, + "learning_rate": 1.7607542735573334e-05, + "loss": 1.0443, + "step": 20144 + }, + { + "epoch": 1.49, + "learning_rate": 1.76072838458812e-05, + "loss": 1.0149, + "step": 20145 + }, + { + "epoch": 1.49, + "learning_rate": 1.7607024944085987e-05, + "loss": 1.0298, + "step": 20146 + }, + { + "epoch": 1.49, + "learning_rate": 1.7606766030188108e-05, + "loss": 1.0461, + "step": 20147 + }, + { + "epoch": 1.49, + "learning_rate": 1.7606507104187974e-05, + "loss": 0.9958, + "step": 20148 + }, + { + "epoch": 1.49, + "learning_rate": 1.7606248166085992e-05, + "loss": 1.0534, + "step": 20149 + }, + { + "epoch": 1.49, + "learning_rate": 1.760598921588258e-05, + "loss": 1.0647, + "step": 20150 + }, + { + "epoch": 1.49, + "learning_rate": 1.7605730253578146e-05, + "loss": 1.0221, + "step": 20151 + }, + { + "epoch": 1.49, + "learning_rate": 1.7605471279173108e-05, + "loss": 1.0679, + "step": 20152 + }, + { + "epoch": 1.49, + "learning_rate": 1.7605212292667872e-05, + "loss": 1.0295, + "step": 20153 + }, + { + "epoch": 1.49, + "learning_rate": 1.7604953294062853e-05, + "loss": 0.9715, + "step": 20154 + }, + { + "epoch": 1.49, + "learning_rate": 1.7604694283358458e-05, + "loss": 1.0542, + "step": 20155 + }, + { + "epoch": 1.49, + "learning_rate": 1.7604435260555106e-05, + "loss": 1.0683, + "step": 20156 + }, + { + "epoch": 1.49, + "learning_rate": 1.7604176225653207e-05, + "loss": 1.0593, + "step": 20157 + }, + { + "epoch": 1.49, + "learning_rate": 1.7603917178653173e-05, + "loss": 1.0834, + "step": 20158 + }, + { + "epoch": 1.49, + "learning_rate": 1.7603658119555415e-05, + "loss": 1.0292, + "step": 20159 + }, + { + "epoch": 1.49, + "learning_rate": 1.7603399048360342e-05, + "loss": 1.0788, + "step": 20160 + }, + { + "epoch": 1.49, + "learning_rate": 1.7603139965068375e-05, + "loss": 1.0279, + "step": 20161 + }, + { + "epoch": 1.49, + "learning_rate": 1.760288086967992e-05, + "loss": 1.0448, + "step": 20162 + }, + { + "epoch": 1.49, + "learning_rate": 1.7602621762195388e-05, + "loss": 1.0523, + "step": 20163 + }, + { + "epoch": 1.49, + "learning_rate": 1.7602362642615195e-05, + "loss": 0.9271, + "step": 20164 + }, + { + "epoch": 1.49, + "learning_rate": 1.760210351093975e-05, + "loss": 1.0119, + "step": 20165 + }, + { + "epoch": 1.49, + "learning_rate": 1.7601844367169472e-05, + "loss": 0.9877, + "step": 20166 + }, + { + "epoch": 1.49, + "learning_rate": 1.7601585211304762e-05, + "loss": 0.964, + "step": 20167 + }, + { + "epoch": 1.49, + "learning_rate": 1.7601326043346042e-05, + "loss": 0.9007, + "step": 20168 + }, + { + "epoch": 1.49, + "learning_rate": 1.7601066863293724e-05, + "loss": 1.1068, + "step": 20169 + }, + { + "epoch": 1.49, + "learning_rate": 1.7600807671148214e-05, + "loss": 1.0122, + "step": 20170 + }, + { + "epoch": 1.49, + "learning_rate": 1.7600548466909927e-05, + "loss": 0.969, + "step": 20171 + }, + { + "epoch": 1.49, + "learning_rate": 1.760028925057928e-05, + "loss": 1.0017, + "step": 20172 + }, + { + "epoch": 1.49, + "learning_rate": 1.7600030022156674e-05, + "loss": 1.1179, + "step": 20173 + }, + { + "epoch": 1.49, + "learning_rate": 1.7599770781642534e-05, + "loss": 0.9875, + "step": 20174 + }, + { + "epoch": 1.49, + "learning_rate": 1.759951152903727e-05, + "loss": 1.0712, + "step": 20175 + }, + { + "epoch": 1.49, + "learning_rate": 1.7599252264341288e-05, + "loss": 1.0627, + "step": 20176 + }, + { + "epoch": 1.49, + "learning_rate": 1.759899298755501e-05, + "loss": 1.0194, + "step": 20177 + }, + { + "epoch": 1.49, + "learning_rate": 1.7598733698678835e-05, + "loss": 0.96, + "step": 20178 + }, + { + "epoch": 1.49, + "learning_rate": 1.7598474397713187e-05, + "loss": 1.046, + "step": 20179 + }, + { + "epoch": 1.49, + "learning_rate": 1.7598215084658475e-05, + "loss": 0.9196, + "step": 20180 + }, + { + "epoch": 1.49, + "learning_rate": 1.759795575951511e-05, + "loss": 1.1172, + "step": 20181 + }, + { + "epoch": 1.49, + "learning_rate": 1.7597696422283507e-05, + "loss": 1.0581, + "step": 20182 + }, + { + "epoch": 1.49, + "learning_rate": 1.7597437072964075e-05, + "loss": 1.0688, + "step": 20183 + }, + { + "epoch": 1.49, + "learning_rate": 1.7597177711557236e-05, + "loss": 1.0612, + "step": 20184 + }, + { + "epoch": 1.49, + "learning_rate": 1.7596918338063392e-05, + "loss": 0.9278, + "step": 20185 + }, + { + "epoch": 1.49, + "learning_rate": 1.759665895248296e-05, + "loss": 0.9978, + "step": 20186 + }, + { + "epoch": 1.49, + "learning_rate": 1.7596399554816352e-05, + "loss": 1.12, + "step": 20187 + }, + { + "epoch": 1.49, + "learning_rate": 1.759614014506398e-05, + "loss": 0.9067, + "step": 20188 + }, + { + "epoch": 1.49, + "learning_rate": 1.759588072322626e-05, + "loss": 1.0825, + "step": 20189 + }, + { + "epoch": 1.49, + "learning_rate": 1.75956212893036e-05, + "loss": 1.0579, + "step": 20190 + }, + { + "epoch": 1.49, + "learning_rate": 1.7595361843296418e-05, + "loss": 1.1123, + "step": 20191 + }, + { + "epoch": 1.49, + "learning_rate": 1.7595102385205122e-05, + "loss": 1.087, + "step": 20192 + }, + { + "epoch": 1.49, + "learning_rate": 1.7594842915030124e-05, + "loss": 1.0165, + "step": 20193 + }, + { + "epoch": 1.49, + "learning_rate": 1.7594583432771846e-05, + "loss": 1.1213, + "step": 20194 + }, + { + "epoch": 1.49, + "learning_rate": 1.7594323938430694e-05, + "loss": 0.9649, + "step": 20195 + }, + { + "epoch": 1.49, + "learning_rate": 1.7594064432007077e-05, + "loss": 1.0509, + "step": 20196 + }, + { + "epoch": 1.49, + "learning_rate": 1.7593804913501413e-05, + "loss": 0.9493, + "step": 20197 + }, + { + "epoch": 1.49, + "learning_rate": 1.7593545382914115e-05, + "loss": 1.0586, + "step": 20198 + }, + { + "epoch": 1.49, + "learning_rate": 1.7593285840245596e-05, + "loss": 1.037, + "step": 20199 + }, + { + "epoch": 1.49, + "learning_rate": 1.7593026285496267e-05, + "loss": 1.0283, + "step": 20200 + }, + { + "epoch": 1.49, + "learning_rate": 1.759276671866654e-05, + "loss": 1.1347, + "step": 20201 + }, + { + "epoch": 1.49, + "learning_rate": 1.759250713975683e-05, + "loss": 1.026, + "step": 20202 + }, + { + "epoch": 1.49, + "learning_rate": 1.759224754876755e-05, + "loss": 0.9589, + "step": 20203 + }, + { + "epoch": 1.49, + "learning_rate": 1.7591987945699117e-05, + "loss": 0.9144, + "step": 20204 + }, + { + "epoch": 1.49, + "learning_rate": 1.7591728330551934e-05, + "loss": 0.8334, + "step": 20205 + }, + { + "epoch": 1.49, + "learning_rate": 1.7591468703326423e-05, + "loss": 1.0559, + "step": 20206 + }, + { + "epoch": 1.49, + "learning_rate": 1.7591209064022995e-05, + "loss": 0.9666, + "step": 20207 + }, + { + "epoch": 1.49, + "learning_rate": 1.759094941264206e-05, + "loss": 0.9294, + "step": 20208 + }, + { + "epoch": 1.49, + "learning_rate": 1.759068974918403e-05, + "loss": 1.0018, + "step": 20209 + }, + { + "epoch": 1.49, + "learning_rate": 1.7590430073649327e-05, + "loss": 1.0557, + "step": 20210 + }, + { + "epoch": 1.49, + "learning_rate": 1.7590170386038353e-05, + "loss": 0.9707, + "step": 20211 + }, + { + "epoch": 1.49, + "learning_rate": 1.7589910686351532e-05, + "loss": 0.9669, + "step": 20212 + }, + { + "epoch": 1.49, + "learning_rate": 1.758965097458927e-05, + "loss": 0.9854, + "step": 20213 + }, + { + "epoch": 1.49, + "learning_rate": 1.758939125075198e-05, + "loss": 0.991, + "step": 20214 + }, + { + "epoch": 1.49, + "learning_rate": 1.758913151484008e-05, + "loss": 1.0464, + "step": 20215 + }, + { + "epoch": 1.49, + "learning_rate": 1.7588871766853975e-05, + "loss": 1.056, + "step": 20216 + }, + { + "epoch": 1.49, + "learning_rate": 1.7588612006794088e-05, + "loss": 0.9703, + "step": 20217 + }, + { + "epoch": 1.49, + "learning_rate": 1.758835223466083e-05, + "loss": 1.055, + "step": 20218 + }, + { + "epoch": 1.49, + "learning_rate": 1.7588092450454608e-05, + "loss": 1.0308, + "step": 20219 + }, + { + "epoch": 1.49, + "learning_rate": 1.758783265417584e-05, + "loss": 1.0703, + "step": 20220 + }, + { + "epoch": 1.49, + "learning_rate": 1.7587572845824937e-05, + "loss": 1.0181, + "step": 20221 + }, + { + "epoch": 1.49, + "learning_rate": 1.758731302540232e-05, + "loss": 1.0625, + "step": 20222 + }, + { + "epoch": 1.49, + "learning_rate": 1.758705319290839e-05, + "loss": 1.0151, + "step": 20223 + }, + { + "epoch": 1.49, + "learning_rate": 1.758679334834357e-05, + "loss": 1.0749, + "step": 20224 + }, + { + "epoch": 1.49, + "learning_rate": 1.7586533491708273e-05, + "loss": 1.0202, + "step": 20225 + }, + { + "epoch": 1.49, + "learning_rate": 1.7586273623002905e-05, + "loss": 0.9507, + "step": 20226 + }, + { + "epoch": 1.49, + "learning_rate": 1.7586013742227887e-05, + "loss": 0.9542, + "step": 20227 + }, + { + "epoch": 1.49, + "learning_rate": 1.7585753849383628e-05, + "loss": 1.0371, + "step": 20228 + }, + { + "epoch": 1.49, + "learning_rate": 1.7585493944470546e-05, + "loss": 0.9977, + "step": 20229 + }, + { + "epoch": 1.49, + "learning_rate": 1.758523402748905e-05, + "loss": 0.906, + "step": 20230 + }, + { + "epoch": 1.49, + "learning_rate": 1.7584974098439554e-05, + "loss": 1.0271, + "step": 20231 + }, + { + "epoch": 1.49, + "learning_rate": 1.7584714157322475e-05, + "loss": 1.1137, + "step": 20232 + }, + { + "epoch": 1.5, + "learning_rate": 1.758445420413822e-05, + "loss": 0.965, + "step": 20233 + }, + { + "epoch": 1.5, + "learning_rate": 1.7584194238887212e-05, + "loss": 1.0601, + "step": 20234 + }, + { + "epoch": 1.5, + "learning_rate": 1.758393426156986e-05, + "loss": 0.9143, + "step": 20235 + }, + { + "epoch": 1.5, + "learning_rate": 1.758367427218657e-05, + "loss": 1.0325, + "step": 20236 + }, + { + "epoch": 1.5, + "learning_rate": 1.7583414270737768e-05, + "loss": 0.9813, + "step": 20237 + }, + { + "epoch": 1.5, + "learning_rate": 1.758315425722386e-05, + "loss": 0.7811, + "step": 20238 + }, + { + "epoch": 1.5, + "learning_rate": 1.7582894231645267e-05, + "loss": 1.1057, + "step": 20239 + }, + { + "epoch": 1.5, + "learning_rate": 1.7582634194002396e-05, + "loss": 0.9855, + "step": 20240 + }, + { + "epoch": 1.5, + "learning_rate": 1.7582374144295656e-05, + "loss": 1.1068, + "step": 20241 + }, + { + "epoch": 1.5, + "learning_rate": 1.7582114082525475e-05, + "loss": 0.9265, + "step": 20242 + }, + { + "epoch": 1.5, + "learning_rate": 1.7581854008692255e-05, + "loss": 0.9698, + "step": 20243 + }, + { + "epoch": 1.5, + "learning_rate": 1.7581593922796414e-05, + "loss": 0.9307, + "step": 20244 + }, + { + "epoch": 1.5, + "learning_rate": 1.7581333824838367e-05, + "loss": 0.9705, + "step": 20245 + }, + { + "epoch": 1.5, + "learning_rate": 1.7581073714818527e-05, + "loss": 1.0394, + "step": 20246 + }, + { + "epoch": 1.5, + "learning_rate": 1.7580813592737307e-05, + "loss": 1.0288, + "step": 20247 + }, + { + "epoch": 1.5, + "learning_rate": 1.758055345859512e-05, + "loss": 0.9562, + "step": 20248 + }, + { + "epoch": 1.5, + "learning_rate": 1.758029331239238e-05, + "loss": 1.0093, + "step": 20249 + }, + { + "epoch": 1.5, + "learning_rate": 1.7580033154129505e-05, + "loss": 0.9162, + "step": 20250 + }, + { + "epoch": 1.5, + "learning_rate": 1.75797729838069e-05, + "loss": 1.0222, + "step": 20251 + }, + { + "epoch": 1.5, + "learning_rate": 1.757951280142499e-05, + "loss": 0.9395, + "step": 20252 + }, + { + "epoch": 1.5, + "learning_rate": 1.757925260698418e-05, + "loss": 1.0369, + "step": 20253 + }, + { + "epoch": 1.5, + "learning_rate": 1.757899240048489e-05, + "loss": 1.1148, + "step": 20254 + }, + { + "epoch": 1.5, + "learning_rate": 1.757873218192753e-05, + "loss": 0.9797, + "step": 20255 + }, + { + "epoch": 1.5, + "learning_rate": 1.7578471951312517e-05, + "loss": 1.0233, + "step": 20256 + }, + { + "epoch": 1.5, + "learning_rate": 1.7578211708640263e-05, + "loss": 0.9637, + "step": 20257 + }, + { + "epoch": 1.5, + "learning_rate": 1.7577951453911183e-05, + "loss": 1.1383, + "step": 20258 + }, + { + "epoch": 1.5, + "learning_rate": 1.757769118712569e-05, + "loss": 0.9228, + "step": 20259 + }, + { + "epoch": 1.5, + "learning_rate": 1.7577430908284198e-05, + "loss": 1.0471, + "step": 20260 + }, + { + "epoch": 1.5, + "learning_rate": 1.7577170617387122e-05, + "loss": 0.9488, + "step": 20261 + }, + { + "epoch": 1.5, + "learning_rate": 1.7576910314434878e-05, + "loss": 0.9858, + "step": 20262 + }, + { + "epoch": 1.5, + "learning_rate": 1.7576649999427874e-05, + "loss": 1.0818, + "step": 20263 + }, + { + "epoch": 1.5, + "learning_rate": 1.757638967236653e-05, + "loss": 0.9328, + "step": 20264 + }, + { + "epoch": 1.5, + "learning_rate": 1.757612933325126e-05, + "loss": 0.8966, + "step": 20265 + }, + { + "epoch": 1.5, + "learning_rate": 1.757586898208248e-05, + "loss": 0.8672, + "step": 20266 + }, + { + "epoch": 1.5, + "learning_rate": 1.7575608618860594e-05, + "loss": 1.0187, + "step": 20267 + }, + { + "epoch": 1.5, + "learning_rate": 1.7575348243586027e-05, + "loss": 1.0214, + "step": 20268 + }, + { + "epoch": 1.5, + "learning_rate": 1.7575087856259188e-05, + "loss": 1.0103, + "step": 20269 + }, + { + "epoch": 1.5, + "learning_rate": 1.7574827456880492e-05, + "loss": 1.082, + "step": 20270 + }, + { + "epoch": 1.5, + "learning_rate": 1.7574567045450357e-05, + "loss": 1.025, + "step": 20271 + }, + { + "epoch": 1.5, + "learning_rate": 1.757430662196919e-05, + "loss": 1.0197, + "step": 20272 + }, + { + "epoch": 1.5, + "learning_rate": 1.7574046186437414e-05, + "loss": 1.0559, + "step": 20273 + }, + { + "epoch": 1.5, + "learning_rate": 1.7573785738855432e-05, + "loss": 1.0295, + "step": 20274 + }, + { + "epoch": 1.5, + "learning_rate": 1.7573525279223675e-05, + "loss": 1.0644, + "step": 20275 + }, + { + "epoch": 1.5, + "learning_rate": 1.7573264807542538e-05, + "loss": 1.0545, + "step": 20276 + }, + { + "epoch": 1.5, + "learning_rate": 1.757300432381245e-05, + "loss": 1.1361, + "step": 20277 + }, + { + "epoch": 1.5, + "learning_rate": 1.757274382803382e-05, + "loss": 0.8973, + "step": 20278 + }, + { + "epoch": 1.5, + "learning_rate": 1.7572483320207062e-05, + "loss": 0.9223, + "step": 20279 + }, + { + "epoch": 1.5, + "learning_rate": 1.7572222800332593e-05, + "loss": 1.0109, + "step": 20280 + }, + { + "epoch": 1.5, + "learning_rate": 1.7571962268410826e-05, + "loss": 1.1201, + "step": 20281 + }, + { + "epoch": 1.5, + "learning_rate": 1.7571701724442173e-05, + "loss": 1.0505, + "step": 20282 + }, + { + "epoch": 1.5, + "learning_rate": 1.757144116842705e-05, + "loss": 0.9873, + "step": 20283 + }, + { + "epoch": 1.5, + "learning_rate": 1.7571180600365875e-05, + "loss": 0.9061, + "step": 20284 + }, + { + "epoch": 1.5, + "learning_rate": 1.7570920020259062e-05, + "loss": 0.9725, + "step": 20285 + }, + { + "epoch": 1.5, + "learning_rate": 1.757065942810702e-05, + "loss": 0.987, + "step": 20286 + }, + { + "epoch": 1.5, + "learning_rate": 1.757039882391017e-05, + "loss": 1.0477, + "step": 20287 + }, + { + "epoch": 1.5, + "learning_rate": 1.757013820766892e-05, + "loss": 1.0516, + "step": 20288 + }, + { + "epoch": 1.5, + "learning_rate": 1.7569877579383692e-05, + "loss": 0.9011, + "step": 20289 + }, + { + "epoch": 1.5, + "learning_rate": 1.7569616939054897e-05, + "loss": 1.1054, + "step": 20290 + }, + { + "epoch": 1.5, + "learning_rate": 1.7569356286682947e-05, + "loss": 1.0709, + "step": 20291 + }, + { + "epoch": 1.5, + "learning_rate": 1.756909562226826e-05, + "loss": 0.889, + "step": 20292 + }, + { + "epoch": 1.5, + "learning_rate": 1.7568834945811253e-05, + "loss": 0.9554, + "step": 20293 + }, + { + "epoch": 1.5, + "learning_rate": 1.7568574257312335e-05, + "loss": 1.1127, + "step": 20294 + }, + { + "epoch": 1.5, + "learning_rate": 1.7568313556771922e-05, + "loss": 0.9794, + "step": 20295 + }, + { + "epoch": 1.5, + "learning_rate": 1.7568052844190434e-05, + "loss": 1.1058, + "step": 20296 + }, + { + "epoch": 1.5, + "learning_rate": 1.756779211956828e-05, + "loss": 1.0407, + "step": 20297 + }, + { + "epoch": 1.5, + "learning_rate": 1.756753138290588e-05, + "loss": 0.86, + "step": 20298 + }, + { + "epoch": 1.5, + "learning_rate": 1.7567270634203642e-05, + "loss": 1.0312, + "step": 20299 + }, + { + "epoch": 1.5, + "learning_rate": 1.7567009873461992e-05, + "loss": 0.9833, + "step": 20300 + }, + { + "epoch": 1.5, + "learning_rate": 1.7566749100681328e-05, + "loss": 1.008, + "step": 20301 + }, + { + "epoch": 1.5, + "learning_rate": 1.756648831586208e-05, + "loss": 0.9289, + "step": 20302 + }, + { + "epoch": 1.5, + "learning_rate": 1.7566227519004658e-05, + "loss": 0.8772, + "step": 20303 + }, + { + "epoch": 1.5, + "learning_rate": 1.7565966710109473e-05, + "loss": 1.0941, + "step": 20304 + }, + { + "epoch": 1.5, + "learning_rate": 1.7565705889176943e-05, + "loss": 1.0197, + "step": 20305 + }, + { + "epoch": 1.5, + "learning_rate": 1.7565445056207486e-05, + "loss": 0.9679, + "step": 20306 + }, + { + "epoch": 1.5, + "learning_rate": 1.7565184211201515e-05, + "loss": 1.0633, + "step": 20307 + }, + { + "epoch": 1.5, + "learning_rate": 1.756492335415944e-05, + "loss": 1.0194, + "step": 20308 + }, + { + "epoch": 1.5, + "learning_rate": 1.7564662485081682e-05, + "loss": 0.9673, + "step": 20309 + }, + { + "epoch": 1.5, + "learning_rate": 1.7564401603968654e-05, + "loss": 1.0009, + "step": 20310 + }, + { + "epoch": 1.5, + "learning_rate": 1.7564140710820772e-05, + "loss": 0.9196, + "step": 20311 + }, + { + "epoch": 1.5, + "learning_rate": 1.756387980563845e-05, + "loss": 1.0012, + "step": 20312 + }, + { + "epoch": 1.5, + "learning_rate": 1.7563618888422106e-05, + "loss": 1.1038, + "step": 20313 + }, + { + "epoch": 1.5, + "learning_rate": 1.756335795917215e-05, + "loss": 1.0342, + "step": 20314 + }, + { + "epoch": 1.5, + "learning_rate": 1.7563097017889e-05, + "loss": 0.9331, + "step": 20315 + }, + { + "epoch": 1.5, + "learning_rate": 1.756283606457307e-05, + "loss": 0.9648, + "step": 20316 + }, + { + "epoch": 1.5, + "learning_rate": 1.7562575099224775e-05, + "loss": 1.0301, + "step": 20317 + }, + { + "epoch": 1.5, + "learning_rate": 1.7562314121844535e-05, + "loss": 1.0286, + "step": 20318 + }, + { + "epoch": 1.5, + "learning_rate": 1.7562053132432758e-05, + "loss": 1.0077, + "step": 20319 + }, + { + "epoch": 1.5, + "learning_rate": 1.7561792130989864e-05, + "loss": 0.8909, + "step": 20320 + }, + { + "epoch": 1.5, + "learning_rate": 1.7561531117516266e-05, + "loss": 0.9886, + "step": 20321 + }, + { + "epoch": 1.5, + "learning_rate": 1.7561270092012385e-05, + "loss": 0.8708, + "step": 20322 + }, + { + "epoch": 1.5, + "learning_rate": 1.7561009054478626e-05, + "loss": 0.966, + "step": 20323 + }, + { + "epoch": 1.5, + "learning_rate": 1.756074800491541e-05, + "loss": 1.0661, + "step": 20324 + }, + { + "epoch": 1.5, + "learning_rate": 1.7560486943323152e-05, + "loss": 1.0456, + "step": 20325 + }, + { + "epoch": 1.5, + "learning_rate": 1.756022586970227e-05, + "loss": 0.9947, + "step": 20326 + }, + { + "epoch": 1.5, + "learning_rate": 1.7559964784053177e-05, + "loss": 1.0019, + "step": 20327 + }, + { + "epoch": 1.5, + "learning_rate": 1.7559703686376285e-05, + "loss": 1.0228, + "step": 20328 + }, + { + "epoch": 1.5, + "learning_rate": 1.7559442576672016e-05, + "loss": 0.9648, + "step": 20329 + }, + { + "epoch": 1.5, + "learning_rate": 1.755918145494078e-05, + "loss": 1.0673, + "step": 20330 + }, + { + "epoch": 1.5, + "learning_rate": 1.7558920321182995e-05, + "loss": 1.0185, + "step": 20331 + }, + { + "epoch": 1.5, + "learning_rate": 1.7558659175399076e-05, + "loss": 1.067, + "step": 20332 + }, + { + "epoch": 1.5, + "learning_rate": 1.7558398017589438e-05, + "loss": 1.0849, + "step": 20333 + }, + { + "epoch": 1.5, + "learning_rate": 1.7558136847754496e-05, + "loss": 0.9806, + "step": 20334 + }, + { + "epoch": 1.5, + "learning_rate": 1.755787566589467e-05, + "loss": 1.0412, + "step": 20335 + }, + { + "epoch": 1.5, + "learning_rate": 1.7557614472010368e-05, + "loss": 0.9921, + "step": 20336 + }, + { + "epoch": 1.5, + "learning_rate": 1.755735326610201e-05, + "loss": 1.0457, + "step": 20337 + }, + { + "epoch": 1.5, + "learning_rate": 1.755709204817001e-05, + "loss": 0.8461, + "step": 20338 + }, + { + "epoch": 1.5, + "learning_rate": 1.755683081821479e-05, + "loss": 1.0224, + "step": 20339 + }, + { + "epoch": 1.5, + "learning_rate": 1.7556569576236757e-05, + "loss": 1.0923, + "step": 20340 + }, + { + "epoch": 1.5, + "learning_rate": 1.755630832223633e-05, + "loss": 1.0117, + "step": 20341 + }, + { + "epoch": 1.5, + "learning_rate": 1.7556047056213926e-05, + "loss": 0.9728, + "step": 20342 + }, + { + "epoch": 1.5, + "learning_rate": 1.7555785778169956e-05, + "loss": 0.9498, + "step": 20343 + }, + { + "epoch": 1.5, + "learning_rate": 1.7555524488104844e-05, + "loss": 0.9935, + "step": 20344 + }, + { + "epoch": 1.5, + "learning_rate": 1.7555263186018993e-05, + "loss": 1.0224, + "step": 20345 + }, + { + "epoch": 1.5, + "learning_rate": 1.7555001871912832e-05, + "loss": 0.9375, + "step": 20346 + }, + { + "epoch": 1.5, + "learning_rate": 1.755474054578677e-05, + "loss": 1.0864, + "step": 20347 + }, + { + "epoch": 1.5, + "learning_rate": 1.7554479207641225e-05, + "loss": 1.0159, + "step": 20348 + }, + { + "epoch": 1.5, + "learning_rate": 1.755421785747661e-05, + "loss": 0.963, + "step": 20349 + }, + { + "epoch": 1.5, + "learning_rate": 1.7553956495293342e-05, + "loss": 0.9647, + "step": 20350 + }, + { + "epoch": 1.5, + "learning_rate": 1.7553695121091838e-05, + "loss": 0.9893, + "step": 20351 + }, + { + "epoch": 1.5, + "learning_rate": 1.7553433734872514e-05, + "loss": 1.0889, + "step": 20352 + }, + { + "epoch": 1.5, + "learning_rate": 1.7553172336635786e-05, + "loss": 1.0687, + "step": 20353 + }, + { + "epoch": 1.5, + "learning_rate": 1.7552910926382065e-05, + "loss": 0.8664, + "step": 20354 + }, + { + "epoch": 1.5, + "learning_rate": 1.7552649504111773e-05, + "loss": 1.1411, + "step": 20355 + }, + { + "epoch": 1.5, + "learning_rate": 1.7552388069825322e-05, + "loss": 0.9429, + "step": 20356 + }, + { + "epoch": 1.5, + "learning_rate": 1.7552126623523133e-05, + "loss": 0.9136, + "step": 20357 + }, + { + "epoch": 1.5, + "learning_rate": 1.7551865165205616e-05, + "loss": 1.0312, + "step": 20358 + }, + { + "epoch": 1.5, + "learning_rate": 1.755160369487319e-05, + "loss": 1.0018, + "step": 20359 + }, + { + "epoch": 1.5, + "learning_rate": 1.7551342212526273e-05, + "loss": 0.9416, + "step": 20360 + }, + { + "epoch": 1.5, + "learning_rate": 1.7551080718165277e-05, + "loss": 0.8993, + "step": 20361 + }, + { + "epoch": 1.5, + "learning_rate": 1.7550819211790617e-05, + "loss": 0.9043, + "step": 20362 + }, + { + "epoch": 1.5, + "learning_rate": 1.7550557693402716e-05, + "loss": 1.0631, + "step": 20363 + }, + { + "epoch": 1.5, + "learning_rate": 1.7550296163001986e-05, + "loss": 1.0764, + "step": 20364 + }, + { + "epoch": 1.5, + "learning_rate": 1.755003462058884e-05, + "loss": 1.0319, + "step": 20365 + }, + { + "epoch": 1.5, + "learning_rate": 1.7549773066163697e-05, + "loss": 1.1372, + "step": 20366 + }, + { + "epoch": 1.5, + "learning_rate": 1.754951149972697e-05, + "loss": 1.0039, + "step": 20367 + }, + { + "epoch": 1.5, + "learning_rate": 1.7549249921279084e-05, + "loss": 0.965, + "step": 20368 + }, + { + "epoch": 1.51, + "learning_rate": 1.7548988330820445e-05, + "loss": 0.938, + "step": 20369 + }, + { + "epoch": 1.51, + "learning_rate": 1.7548726728351476e-05, + "loss": 0.9758, + "step": 20370 + }, + { + "epoch": 1.51, + "learning_rate": 1.7548465113872593e-05, + "loss": 1.0395, + "step": 20371 + }, + { + "epoch": 1.51, + "learning_rate": 1.7548203487384205e-05, + "loss": 1.109, + "step": 20372 + }, + { + "epoch": 1.51, + "learning_rate": 1.7547941848886737e-05, + "loss": 1.0188, + "step": 20373 + }, + { + "epoch": 1.51, + "learning_rate": 1.75476801983806e-05, + "loss": 0.9985, + "step": 20374 + }, + { + "epoch": 1.51, + "learning_rate": 1.7547418535866212e-05, + "loss": 1.0398, + "step": 20375 + }, + { + "epoch": 1.51, + "learning_rate": 1.754715686134399e-05, + "loss": 1.0842, + "step": 20376 + }, + { + "epoch": 1.51, + "learning_rate": 1.7546895174814345e-05, + "loss": 1.0377, + "step": 20377 + }, + { + "epoch": 1.51, + "learning_rate": 1.7546633476277705e-05, + "loss": 1.1147, + "step": 20378 + }, + { + "epoch": 1.51, + "learning_rate": 1.7546371765734477e-05, + "loss": 1.005, + "step": 20379 + }, + { + "epoch": 1.51, + "learning_rate": 1.7546110043185076e-05, + "loss": 0.9403, + "step": 20380 + }, + { + "epoch": 1.51, + "learning_rate": 1.7545848308629924e-05, + "loss": 1.0019, + "step": 20381 + }, + { + "epoch": 1.51, + "learning_rate": 1.7545586562069436e-05, + "loss": 1.005, + "step": 20382 + }, + { + "epoch": 1.51, + "learning_rate": 1.7545324803504024e-05, + "loss": 1.0033, + "step": 20383 + }, + { + "epoch": 1.51, + "learning_rate": 1.754506303293411e-05, + "loss": 1.0371, + "step": 20384 + }, + { + "epoch": 1.51, + "learning_rate": 1.754480125036011e-05, + "loss": 0.9509, + "step": 20385 + }, + { + "epoch": 1.51, + "learning_rate": 1.754453945578244e-05, + "loss": 0.9248, + "step": 20386 + }, + { + "epoch": 1.51, + "learning_rate": 1.7544277649201516e-05, + "loss": 1.0407, + "step": 20387 + }, + { + "epoch": 1.51, + "learning_rate": 1.754401583061775e-05, + "loss": 1.077, + "step": 20388 + }, + { + "epoch": 1.51, + "learning_rate": 1.7543754000031568e-05, + "loss": 1.0568, + "step": 20389 + }, + { + "epoch": 1.51, + "learning_rate": 1.7543492157443378e-05, + "loss": 1.1098, + "step": 20390 + }, + { + "epoch": 1.51, + "learning_rate": 1.75432303028536e-05, + "loss": 1.0309, + "step": 20391 + }, + { + "epoch": 1.51, + "learning_rate": 1.7542968436262653e-05, + "loss": 0.9517, + "step": 20392 + }, + { + "epoch": 1.51, + "learning_rate": 1.754270655767095e-05, + "loss": 0.9718, + "step": 20393 + }, + { + "epoch": 1.51, + "learning_rate": 1.7542444667078907e-05, + "loss": 1.1155, + "step": 20394 + }, + { + "epoch": 1.51, + "learning_rate": 1.7542182764486944e-05, + "loss": 1.0109, + "step": 20395 + }, + { + "epoch": 1.51, + "learning_rate": 1.7541920849895475e-05, + "loss": 0.8739, + "step": 20396 + }, + { + "epoch": 1.51, + "learning_rate": 1.7541658923304917e-05, + "loss": 1.0179, + "step": 20397 + }, + { + "epoch": 1.51, + "learning_rate": 1.754139698471569e-05, + "loss": 1.0482, + "step": 20398 + }, + { + "epoch": 1.51, + "learning_rate": 1.7541135034128207e-05, + "loss": 0.9543, + "step": 20399 + }, + { + "epoch": 1.51, + "learning_rate": 1.754087307154289e-05, + "loss": 1.0525, + "step": 20400 + }, + { + "epoch": 1.51, + "learning_rate": 1.754061109696015e-05, + "loss": 1.1227, + "step": 20401 + }, + { + "epoch": 1.51, + "learning_rate": 1.75403491103804e-05, + "loss": 1.0204, + "step": 20402 + }, + { + "epoch": 1.51, + "learning_rate": 1.7540087111804066e-05, + "loss": 0.9649, + "step": 20403 + }, + { + "epoch": 1.51, + "learning_rate": 1.7539825101231563e-05, + "loss": 0.9792, + "step": 20404 + }, + { + "epoch": 1.51, + "learning_rate": 1.7539563078663305e-05, + "loss": 1.0968, + "step": 20405 + }, + { + "epoch": 1.51, + "learning_rate": 1.753930104409971e-05, + "loss": 0.8882, + "step": 20406 + }, + { + "epoch": 1.51, + "learning_rate": 1.7539038997541195e-05, + "loss": 1.072, + "step": 20407 + }, + { + "epoch": 1.51, + "learning_rate": 1.7538776938988175e-05, + "loss": 1.107, + "step": 20408 + }, + { + "epoch": 1.51, + "learning_rate": 1.753851486844107e-05, + "loss": 1.0616, + "step": 20409 + }, + { + "epoch": 1.51, + "learning_rate": 1.75382527859003e-05, + "loss": 1.0677, + "step": 20410 + }, + { + "epoch": 1.51, + "learning_rate": 1.7537990691366272e-05, + "loss": 0.9763, + "step": 20411 + }, + { + "epoch": 1.51, + "learning_rate": 1.7537728584839407e-05, + "loss": 0.9503, + "step": 20412 + }, + { + "epoch": 1.51, + "learning_rate": 1.753746646632013e-05, + "loss": 1.0997, + "step": 20413 + }, + { + "epoch": 1.51, + "learning_rate": 1.7537204335808848e-05, + "loss": 0.9803, + "step": 20414 + }, + { + "epoch": 1.51, + "learning_rate": 1.7536942193305983e-05, + "loss": 0.8472, + "step": 20415 + }, + { + "epoch": 1.51, + "learning_rate": 1.7536680038811947e-05, + "loss": 0.995, + "step": 20416 + }, + { + "epoch": 1.51, + "learning_rate": 1.7536417872327164e-05, + "loss": 1.0479, + "step": 20417 + }, + { + "epoch": 1.51, + "learning_rate": 1.753615569385205e-05, + "loss": 1.0686, + "step": 20418 + }, + { + "epoch": 1.51, + "learning_rate": 1.7535893503387022e-05, + "loss": 1.0987, + "step": 20419 + }, + { + "epoch": 1.51, + "learning_rate": 1.753563130093249e-05, + "loss": 0.9189, + "step": 20420 + }, + { + "epoch": 1.51, + "learning_rate": 1.7535369086488878e-05, + "loss": 1.0179, + "step": 20421 + }, + { + "epoch": 1.51, + "learning_rate": 1.7535106860056598e-05, + "loss": 0.964, + "step": 20422 + }, + { + "epoch": 1.51, + "learning_rate": 1.7534844621636078e-05, + "loss": 1.0174, + "step": 20423 + }, + { + "epoch": 1.51, + "learning_rate": 1.7534582371227724e-05, + "loss": 0.9939, + "step": 20424 + }, + { + "epoch": 1.51, + "learning_rate": 1.7534320108831957e-05, + "loss": 1.0191, + "step": 20425 + }, + { + "epoch": 1.51, + "learning_rate": 1.7534057834449193e-05, + "loss": 1.0727, + "step": 20426 + }, + { + "epoch": 1.51, + "learning_rate": 1.7533795548079852e-05, + "loss": 1.0586, + "step": 20427 + }, + { + "epoch": 1.51, + "learning_rate": 1.7533533249724352e-05, + "loss": 0.8724, + "step": 20428 + }, + { + "epoch": 1.51, + "learning_rate": 1.7533270939383108e-05, + "loss": 1.0586, + "step": 20429 + }, + { + "epoch": 1.51, + "learning_rate": 1.7533008617056538e-05, + "loss": 0.9893, + "step": 20430 + }, + { + "epoch": 1.51, + "learning_rate": 1.7532746282745058e-05, + "loss": 0.952, + "step": 20431 + }, + { + "epoch": 1.51, + "learning_rate": 1.7532483936449085e-05, + "loss": 1.0062, + "step": 20432 + }, + { + "epoch": 1.51, + "learning_rate": 1.7532221578169042e-05, + "loss": 0.9653, + "step": 20433 + }, + { + "epoch": 1.51, + "learning_rate": 1.7531959207905337e-05, + "loss": 1.0586, + "step": 20434 + }, + { + "epoch": 1.51, + "learning_rate": 1.75316968256584e-05, + "loss": 1.0798, + "step": 20435 + }, + { + "epoch": 1.51, + "learning_rate": 1.753143443142863e-05, + "loss": 0.9827, + "step": 20436 + }, + { + "epoch": 1.51, + "learning_rate": 1.7531172025216465e-05, + "loss": 1.0109, + "step": 20437 + }, + { + "epoch": 1.51, + "learning_rate": 1.753090960702231e-05, + "loss": 1.0293, + "step": 20438 + }, + { + "epoch": 1.51, + "learning_rate": 1.7530647176846588e-05, + "loss": 0.9962, + "step": 20439 + }, + { + "epoch": 1.51, + "learning_rate": 1.753038473468971e-05, + "loss": 1.121, + "step": 20440 + }, + { + "epoch": 1.51, + "learning_rate": 1.7530122280552098e-05, + "loss": 1.0008, + "step": 20441 + }, + { + "epoch": 1.51, + "learning_rate": 1.752985981443417e-05, + "loss": 1.0645, + "step": 20442 + }, + { + "epoch": 1.51, + "learning_rate": 1.7529597336336347e-05, + "loss": 0.8929, + "step": 20443 + }, + { + "epoch": 1.51, + "learning_rate": 1.7529334846259035e-05, + "loss": 0.9481, + "step": 20444 + }, + { + "epoch": 1.51, + "learning_rate": 1.7529072344202665e-05, + "loss": 1.0385, + "step": 20445 + }, + { + "epoch": 1.51, + "learning_rate": 1.7528809830167642e-05, + "loss": 0.9452, + "step": 20446 + }, + { + "epoch": 1.51, + "learning_rate": 1.75285473041544e-05, + "loss": 1.007, + "step": 20447 + }, + { + "epoch": 1.51, + "learning_rate": 1.752828476616334e-05, + "loss": 1.0517, + "step": 20448 + }, + { + "epoch": 1.51, + "learning_rate": 1.7528022216194886e-05, + "loss": 1.0424, + "step": 20449 + }, + { + "epoch": 1.51, + "learning_rate": 1.752775965424946e-05, + "loss": 1.039, + "step": 20450 + }, + { + "epoch": 1.51, + "learning_rate": 1.7527497080327475e-05, + "loss": 0.9464, + "step": 20451 + }, + { + "epoch": 1.51, + "learning_rate": 1.752723449442935e-05, + "loss": 1.0458, + "step": 20452 + }, + { + "epoch": 1.51, + "learning_rate": 1.7526971896555502e-05, + "loss": 1.0223, + "step": 20453 + }, + { + "epoch": 1.51, + "learning_rate": 1.752670928670635e-05, + "loss": 0.863, + "step": 20454 + }, + { + "epoch": 1.51, + "learning_rate": 1.752644666488231e-05, + "loss": 1.0231, + "step": 20455 + }, + { + "epoch": 1.51, + "learning_rate": 1.7526184031083803e-05, + "loss": 0.9947, + "step": 20456 + }, + { + "epoch": 1.51, + "learning_rate": 1.752592138531124e-05, + "loss": 1.0461, + "step": 20457 + }, + { + "epoch": 1.51, + "learning_rate": 1.7525658727565048e-05, + "loss": 1.0629, + "step": 20458 + }, + { + "epoch": 1.51, + "learning_rate": 1.752539605784564e-05, + "loss": 1.0616, + "step": 20459 + }, + { + "epoch": 1.51, + "learning_rate": 1.7525133376153433e-05, + "loss": 0.9659, + "step": 20460 + }, + { + "epoch": 1.51, + "learning_rate": 1.7524870682488852e-05, + "loss": 1.0195, + "step": 20461 + }, + { + "epoch": 1.51, + "learning_rate": 1.7524607976852307e-05, + "loss": 1.0369, + "step": 20462 + }, + { + "epoch": 1.51, + "learning_rate": 1.7524345259244216e-05, + "loss": 0.9764, + "step": 20463 + }, + { + "epoch": 1.51, + "learning_rate": 1.7524082529664997e-05, + "loss": 1.065, + "step": 20464 + }, + { + "epoch": 1.51, + "learning_rate": 1.7523819788115075e-05, + "loss": 1.1101, + "step": 20465 + }, + { + "epoch": 1.51, + "learning_rate": 1.752355703459486e-05, + "loss": 0.9306, + "step": 20466 + }, + { + "epoch": 1.51, + "learning_rate": 1.752329426910478e-05, + "loss": 0.9693, + "step": 20467 + }, + { + "epoch": 1.51, + "learning_rate": 1.752303149164524e-05, + "loss": 0.9033, + "step": 20468 + }, + { + "epoch": 1.51, + "learning_rate": 1.7522768702216666e-05, + "loss": 1.0127, + "step": 20469 + }, + { + "epoch": 1.51, + "learning_rate": 1.7522505900819476e-05, + "loss": 1.0334, + "step": 20470 + }, + { + "epoch": 1.51, + "learning_rate": 1.7522243087454086e-05, + "loss": 0.933, + "step": 20471 + }, + { + "epoch": 1.51, + "learning_rate": 1.7521980262120913e-05, + "loss": 1.095, + "step": 20472 + }, + { + "epoch": 1.51, + "learning_rate": 1.7521717424820378e-05, + "loss": 1.1346, + "step": 20473 + }, + { + "epoch": 1.51, + "learning_rate": 1.75214545755529e-05, + "loss": 1.0424, + "step": 20474 + }, + { + "epoch": 1.51, + "learning_rate": 1.7521191714318893e-05, + "loss": 1.095, + "step": 20475 + }, + { + "epoch": 1.51, + "learning_rate": 1.7520928841118777e-05, + "loss": 0.9378, + "step": 20476 + }, + { + "epoch": 1.51, + "learning_rate": 1.7520665955952972e-05, + "loss": 1.007, + "step": 20477 + }, + { + "epoch": 1.51, + "learning_rate": 1.7520403058821895e-05, + "loss": 0.8683, + "step": 20478 + }, + { + "epoch": 1.51, + "learning_rate": 1.7520140149725967e-05, + "loss": 1.0218, + "step": 20479 + }, + { + "epoch": 1.51, + "learning_rate": 1.7519877228665602e-05, + "loss": 0.9505, + "step": 20480 + }, + { + "epoch": 1.51, + "learning_rate": 1.751961429564122e-05, + "loss": 0.9705, + "step": 20481 + }, + { + "epoch": 1.51, + "learning_rate": 1.7519351350653235e-05, + "loss": 0.92, + "step": 20482 + }, + { + "epoch": 1.51, + "learning_rate": 1.7519088393702076e-05, + "loss": 0.9734, + "step": 20483 + }, + { + "epoch": 1.51, + "learning_rate": 1.7518825424788152e-05, + "loss": 1.029, + "step": 20484 + }, + { + "epoch": 1.51, + "learning_rate": 1.7518562443911883e-05, + "loss": 0.9686, + "step": 20485 + }, + { + "epoch": 1.51, + "learning_rate": 1.7518299451073693e-05, + "loss": 1.0452, + "step": 20486 + }, + { + "epoch": 1.51, + "learning_rate": 1.751803644627399e-05, + "loss": 1.0396, + "step": 20487 + }, + { + "epoch": 1.51, + "learning_rate": 1.7517773429513206e-05, + "loss": 1.0384, + "step": 20488 + }, + { + "epoch": 1.51, + "learning_rate": 1.7517510400791745e-05, + "loss": 0.9521, + "step": 20489 + }, + { + "epoch": 1.51, + "learning_rate": 1.7517247360110038e-05, + "loss": 1.0661, + "step": 20490 + }, + { + "epoch": 1.51, + "learning_rate": 1.7516984307468496e-05, + "loss": 0.8747, + "step": 20491 + }, + { + "epoch": 1.51, + "learning_rate": 1.7516721242867536e-05, + "loss": 1.109, + "step": 20492 + }, + { + "epoch": 1.51, + "learning_rate": 1.7516458166307583e-05, + "loss": 1.0038, + "step": 20493 + }, + { + "epoch": 1.51, + "learning_rate": 1.7516195077789056e-05, + "loss": 0.9412, + "step": 20494 + }, + { + "epoch": 1.51, + "learning_rate": 1.7515931977312365e-05, + "loss": 1.0032, + "step": 20495 + }, + { + "epoch": 1.51, + "learning_rate": 1.7515668864877934e-05, + "loss": 0.9787, + "step": 20496 + }, + { + "epoch": 1.51, + "learning_rate": 1.7515405740486186e-05, + "loss": 1.029, + "step": 20497 + }, + { + "epoch": 1.51, + "learning_rate": 1.751514260413753e-05, + "loss": 1.0578, + "step": 20498 + }, + { + "epoch": 1.51, + "learning_rate": 1.7514879455832393e-05, + "loss": 1.0468, + "step": 20499 + }, + { + "epoch": 1.51, + "learning_rate": 1.751461629557119e-05, + "loss": 0.9238, + "step": 20500 + }, + { + "epoch": 1.51, + "learning_rate": 1.7514353123354337e-05, + "loss": 1.055, + "step": 20501 + }, + { + "epoch": 1.51, + "learning_rate": 1.7514089939182257e-05, + "loss": 1.1128, + "step": 20502 + }, + { + "epoch": 1.51, + "learning_rate": 1.751382674305537e-05, + "loss": 0.9845, + "step": 20503 + }, + { + "epoch": 1.52, + "learning_rate": 1.751356353497409e-05, + "loss": 1.0484, + "step": 20504 + }, + { + "epoch": 1.52, + "learning_rate": 1.7513300314938838e-05, + "loss": 1.0852, + "step": 20505 + }, + { + "epoch": 1.52, + "learning_rate": 1.7513037082950032e-05, + "loss": 1.0451, + "step": 20506 + }, + { + "epoch": 1.52, + "learning_rate": 1.751277383900809e-05, + "loss": 0.9946, + "step": 20507 + }, + { + "epoch": 1.52, + "learning_rate": 1.751251058311344e-05, + "loss": 0.9414, + "step": 20508 + }, + { + "epoch": 1.52, + "learning_rate": 1.7512247315266483e-05, + "loss": 0.9928, + "step": 20509 + }, + { + "epoch": 1.52, + "learning_rate": 1.7511984035467655e-05, + "loss": 0.9868, + "step": 20510 + }, + { + "epoch": 1.52, + "learning_rate": 1.7511720743717366e-05, + "loss": 0.9537, + "step": 20511 + }, + { + "epoch": 1.52, + "learning_rate": 1.7511457440016037e-05, + "loss": 0.927, + "step": 20512 + }, + { + "epoch": 1.52, + "learning_rate": 1.7511194124364084e-05, + "loss": 1.1333, + "step": 20513 + }, + { + "epoch": 1.52, + "learning_rate": 1.7510930796761932e-05, + "loss": 0.9884, + "step": 20514 + }, + { + "epoch": 1.52, + "learning_rate": 1.7510667457209995e-05, + "loss": 1.0398, + "step": 20515 + }, + { + "epoch": 1.52, + "learning_rate": 1.7510404105708694e-05, + "loss": 0.9615, + "step": 20516 + }, + { + "epoch": 1.52, + "learning_rate": 1.7510140742258448e-05, + "loss": 1.0647, + "step": 20517 + }, + { + "epoch": 1.52, + "learning_rate": 1.7509877366859673e-05, + "loss": 0.957, + "step": 20518 + }, + { + "epoch": 1.52, + "learning_rate": 1.7509613979512795e-05, + "loss": 1.0078, + "step": 20519 + }, + { + "epoch": 1.52, + "learning_rate": 1.7509350580218225e-05, + "loss": 1.06, + "step": 20520 + }, + { + "epoch": 1.52, + "learning_rate": 1.7509087168976387e-05, + "loss": 1.0243, + "step": 20521 + }, + { + "epoch": 1.52, + "learning_rate": 1.75088237457877e-05, + "loss": 0.9171, + "step": 20522 + }, + { + "epoch": 1.52, + "learning_rate": 1.7508560310652577e-05, + "loss": 1.0712, + "step": 20523 + }, + { + "epoch": 1.52, + "learning_rate": 1.7508296863571447e-05, + "loss": 1.0384, + "step": 20524 + }, + { + "epoch": 1.52, + "learning_rate": 1.750803340454472e-05, + "loss": 0.8875, + "step": 20525 + }, + { + "epoch": 1.52, + "learning_rate": 1.7507769933572823e-05, + "loss": 1.1026, + "step": 20526 + }, + { + "epoch": 1.52, + "learning_rate": 1.750750645065617e-05, + "loss": 1.05, + "step": 20527 + }, + { + "epoch": 1.52, + "learning_rate": 1.7507242955795182e-05, + "loss": 1.0818, + "step": 20528 + }, + { + "epoch": 1.52, + "learning_rate": 1.7506979448990277e-05, + "loss": 0.952, + "step": 20529 + }, + { + "epoch": 1.52, + "learning_rate": 1.7506715930241877e-05, + "loss": 1.0181, + "step": 20530 + }, + { + "epoch": 1.52, + "learning_rate": 1.75064523995504e-05, + "loss": 1.1628, + "step": 20531 + }, + { + "epoch": 1.52, + "learning_rate": 1.7506188856916262e-05, + "loss": 1.0179, + "step": 20532 + }, + { + "epoch": 1.52, + "learning_rate": 1.7505925302339886e-05, + "loss": 1.0914, + "step": 20533 + }, + { + "epoch": 1.52, + "learning_rate": 1.750566173582169e-05, + "loss": 0.968, + "step": 20534 + }, + { + "epoch": 1.52, + "learning_rate": 1.7505398157362094e-05, + "loss": 0.8978, + "step": 20535 + }, + { + "epoch": 1.52, + "learning_rate": 1.7505134566961516e-05, + "loss": 1.0171, + "step": 20536 + }, + { + "epoch": 1.52, + "learning_rate": 1.7504870964620375e-05, + "loss": 0.9454, + "step": 20537 + }, + { + "epoch": 1.52, + "learning_rate": 1.7504607350339093e-05, + "loss": 0.9559, + "step": 20538 + }, + { + "epoch": 1.52, + "learning_rate": 1.7504343724118087e-05, + "loss": 0.8915, + "step": 20539 + }, + { + "epoch": 1.52, + "learning_rate": 1.750408008595778e-05, + "loss": 1.0397, + "step": 20540 + }, + { + "epoch": 1.52, + "learning_rate": 1.7503816435858588e-05, + "loss": 1.0537, + "step": 20541 + }, + { + "epoch": 1.52, + "learning_rate": 1.7503552773820933e-05, + "loss": 1.0134, + "step": 20542 + }, + { + "epoch": 1.52, + "learning_rate": 1.750328909984523e-05, + "loss": 0.9662, + "step": 20543 + }, + { + "epoch": 1.52, + "learning_rate": 1.7503025413931903e-05, + "loss": 0.9026, + "step": 20544 + }, + { + "epoch": 1.52, + "learning_rate": 1.750276171608137e-05, + "loss": 1.0467, + "step": 20545 + }, + { + "epoch": 1.52, + "learning_rate": 1.750249800629405e-05, + "loss": 1.0825, + "step": 20546 + }, + { + "epoch": 1.52, + "learning_rate": 1.7502234284570364e-05, + "loss": 0.9914, + "step": 20547 + }, + { + "epoch": 1.52, + "learning_rate": 1.7501970550910728e-05, + "loss": 0.9688, + "step": 20548 + }, + { + "epoch": 1.52, + "learning_rate": 1.7501706805315567e-05, + "loss": 1.0314, + "step": 20549 + }, + { + "epoch": 1.52, + "learning_rate": 1.7501443047785295e-05, + "loss": 1.0724, + "step": 20550 + }, + { + "epoch": 1.52, + "learning_rate": 1.7501179278320336e-05, + "loss": 1.0247, + "step": 20551 + }, + { + "epoch": 1.52, + "learning_rate": 1.750091549692111e-05, + "loss": 0.9651, + "step": 20552 + }, + { + "epoch": 1.52, + "learning_rate": 1.750065170358803e-05, + "loss": 0.9966, + "step": 20553 + }, + { + "epoch": 1.52, + "learning_rate": 1.7500387898321525e-05, + "loss": 1.067, + "step": 20554 + }, + { + "epoch": 1.52, + "learning_rate": 1.750012408112201e-05, + "loss": 1.007, + "step": 20555 + }, + { + "epoch": 1.52, + "learning_rate": 1.7499860251989902e-05, + "loss": 1.1097, + "step": 20556 + }, + { + "epoch": 1.52, + "learning_rate": 1.7499596410925624e-05, + "loss": 1.0387, + "step": 20557 + }, + { + "epoch": 1.52, + "learning_rate": 1.74993325579296e-05, + "loss": 1.136, + "step": 20558 + }, + { + "epoch": 1.52, + "learning_rate": 1.7499068693002238e-05, + "loss": 0.9925, + "step": 20559 + }, + { + "epoch": 1.52, + "learning_rate": 1.749880481614397e-05, + "loss": 1.0143, + "step": 20560 + }, + { + "epoch": 1.52, + "learning_rate": 1.7498540927355206e-05, + "loss": 1.0922, + "step": 20561 + }, + { + "epoch": 1.52, + "learning_rate": 1.7498277026636376e-05, + "loss": 1.0057, + "step": 20562 + }, + { + "epoch": 1.52, + "learning_rate": 1.7498013113987893e-05, + "loss": 0.9598, + "step": 20563 + }, + { + "epoch": 1.52, + "learning_rate": 1.749774918941018e-05, + "loss": 0.9582, + "step": 20564 + }, + { + "epoch": 1.52, + "learning_rate": 1.7497485252903654e-05, + "loss": 0.9222, + "step": 20565 + }, + { + "epoch": 1.52, + "learning_rate": 1.7497221304468734e-05, + "loss": 1.065, + "step": 20566 + }, + { + "epoch": 1.52, + "learning_rate": 1.7496957344105845e-05, + "loss": 1.0162, + "step": 20567 + }, + { + "epoch": 1.52, + "learning_rate": 1.74966933718154e-05, + "loss": 1.0274, + "step": 20568 + }, + { + "epoch": 1.52, + "learning_rate": 1.7496429387597828e-05, + "loss": 1.0588, + "step": 20569 + }, + { + "epoch": 1.52, + "learning_rate": 1.7496165391453544e-05, + "loss": 1.0524, + "step": 20570 + }, + { + "epoch": 1.52, + "learning_rate": 1.7495901383382964e-05, + "loss": 1.042, + "step": 20571 + }, + { + "epoch": 1.52, + "learning_rate": 1.7495637363386516e-05, + "loss": 1.0742, + "step": 20572 + }, + { + "epoch": 1.52, + "learning_rate": 1.7495373331464614e-05, + "loss": 0.8498, + "step": 20573 + }, + { + "epoch": 1.52, + "learning_rate": 1.749510928761768e-05, + "loss": 1.1377, + "step": 20574 + }, + { + "epoch": 1.52, + "learning_rate": 1.7494845231846135e-05, + "loss": 1.1335, + "step": 20575 + }, + { + "epoch": 1.52, + "learning_rate": 1.74945811641504e-05, + "loss": 1.0293, + "step": 20576 + }, + { + "epoch": 1.52, + "learning_rate": 1.7494317084530892e-05, + "loss": 1.0341, + "step": 20577 + }, + { + "epoch": 1.52, + "learning_rate": 1.749405299298803e-05, + "loss": 0.9974, + "step": 20578 + }, + { + "epoch": 1.52, + "learning_rate": 1.749378888952224e-05, + "loss": 1.0366, + "step": 20579 + }, + { + "epoch": 1.52, + "learning_rate": 1.7493524774133944e-05, + "loss": 0.9373, + "step": 20580 + }, + { + "epoch": 1.52, + "learning_rate": 1.7493260646823552e-05, + "loss": 1.1342, + "step": 20581 + }, + { + "epoch": 1.52, + "learning_rate": 1.7492996507591486e-05, + "loss": 0.9782, + "step": 20582 + }, + { + "epoch": 1.52, + "learning_rate": 1.7492732356438176e-05, + "loss": 1.0958, + "step": 20583 + }, + { + "epoch": 1.52, + "learning_rate": 1.749246819336403e-05, + "loss": 1.0184, + "step": 20584 + }, + { + "epoch": 1.52, + "learning_rate": 1.749220401836948e-05, + "loss": 1.0186, + "step": 20585 + }, + { + "epoch": 1.52, + "learning_rate": 1.7491939831454942e-05, + "loss": 1.1325, + "step": 20586 + }, + { + "epoch": 1.52, + "learning_rate": 1.749167563262083e-05, + "loss": 0.9995, + "step": 20587 + }, + { + "epoch": 1.52, + "learning_rate": 1.749141142186757e-05, + "loss": 1.0081, + "step": 20588 + }, + { + "epoch": 1.52, + "learning_rate": 1.7491147199195584e-05, + "loss": 1.0219, + "step": 20589 + }, + { + "epoch": 1.52, + "learning_rate": 1.7490882964605287e-05, + "loss": 0.9086, + "step": 20590 + }, + { + "epoch": 1.52, + "learning_rate": 1.74906187180971e-05, + "loss": 0.9269, + "step": 20591 + }, + { + "epoch": 1.52, + "learning_rate": 1.7490354459671454e-05, + "loss": 1.0101, + "step": 20592 + }, + { + "epoch": 1.52, + "learning_rate": 1.7490090189328753e-05, + "loss": 1.0005, + "step": 20593 + }, + { + "epoch": 1.52, + "learning_rate": 1.7489825907069435e-05, + "loss": 0.9748, + "step": 20594 + }, + { + "epoch": 1.52, + "learning_rate": 1.74895616128939e-05, + "loss": 0.9771, + "step": 20595 + }, + { + "epoch": 1.52, + "learning_rate": 1.7489297306802588e-05, + "loss": 1.1253, + "step": 20596 + }, + { + "epoch": 1.52, + "learning_rate": 1.7489032988795907e-05, + "loss": 0.9137, + "step": 20597 + }, + { + "epoch": 1.52, + "learning_rate": 1.7488768658874285e-05, + "loss": 1.0222, + "step": 20598 + }, + { + "epoch": 1.52, + "learning_rate": 1.748850431703814e-05, + "loss": 0.9786, + "step": 20599 + }, + { + "epoch": 1.52, + "learning_rate": 1.7488239963287884e-05, + "loss": 0.9436, + "step": 20600 + }, + { + "epoch": 1.52, + "learning_rate": 1.748797559762395e-05, + "loss": 0.9764, + "step": 20601 + }, + { + "epoch": 1.52, + "learning_rate": 1.7487711220046758e-05, + "loss": 0.9724, + "step": 20602 + }, + { + "epoch": 1.52, + "learning_rate": 1.748744683055672e-05, + "loss": 0.9592, + "step": 20603 + }, + { + "epoch": 1.52, + "learning_rate": 1.7487182429154262e-05, + "loss": 1.0076, + "step": 20604 + }, + { + "epoch": 1.52, + "learning_rate": 1.7486918015839804e-05, + "loss": 0.9305, + "step": 20605 + }, + { + "epoch": 1.52, + "learning_rate": 1.7486653590613768e-05, + "loss": 1.0336, + "step": 20606 + }, + { + "epoch": 1.52, + "learning_rate": 1.7486389153476572e-05, + "loss": 1.1557, + "step": 20607 + }, + { + "epoch": 1.52, + "learning_rate": 1.7486124704428636e-05, + "loss": 0.9842, + "step": 20608 + }, + { + "epoch": 1.52, + "learning_rate": 1.7485860243470383e-05, + "loss": 0.934, + "step": 20609 + }, + { + "epoch": 1.52, + "learning_rate": 1.7485595770602234e-05, + "loss": 0.9849, + "step": 20610 + }, + { + "epoch": 1.52, + "learning_rate": 1.7485331285824615e-05, + "loss": 0.9878, + "step": 20611 + }, + { + "epoch": 1.52, + "learning_rate": 1.7485066789137934e-05, + "loss": 1.0222, + "step": 20612 + }, + { + "epoch": 1.52, + "learning_rate": 1.7484802280542623e-05, + "loss": 0.9697, + "step": 20613 + }, + { + "epoch": 1.52, + "learning_rate": 1.7484537760039092e-05, + "loss": 0.9383, + "step": 20614 + }, + { + "epoch": 1.52, + "learning_rate": 1.7484273227627776e-05, + "loss": 1.0575, + "step": 20615 + }, + { + "epoch": 1.52, + "learning_rate": 1.7484008683309084e-05, + "loss": 0.9384, + "step": 20616 + }, + { + "epoch": 1.52, + "learning_rate": 1.7483744127083443e-05, + "loss": 0.9765, + "step": 20617 + }, + { + "epoch": 1.52, + "learning_rate": 1.7483479558951274e-05, + "loss": 1.013, + "step": 20618 + }, + { + "epoch": 1.52, + "learning_rate": 1.7483214978912993e-05, + "loss": 0.9375, + "step": 20619 + }, + { + "epoch": 1.52, + "learning_rate": 1.7482950386969023e-05, + "loss": 1.0963, + "step": 20620 + }, + { + "epoch": 1.52, + "learning_rate": 1.7482685783119792e-05, + "loss": 0.9812, + "step": 20621 + }, + { + "epoch": 1.52, + "learning_rate": 1.7482421167365712e-05, + "loss": 1.0798, + "step": 20622 + }, + { + "epoch": 1.52, + "learning_rate": 1.7482156539707205e-05, + "loss": 0.9554, + "step": 20623 + }, + { + "epoch": 1.52, + "learning_rate": 1.7481891900144693e-05, + "loss": 0.9859, + "step": 20624 + }, + { + "epoch": 1.52, + "learning_rate": 1.74816272486786e-05, + "loss": 0.9262, + "step": 20625 + }, + { + "epoch": 1.52, + "learning_rate": 1.7481362585309346e-05, + "loss": 1.053, + "step": 20626 + }, + { + "epoch": 1.52, + "learning_rate": 1.748109791003735e-05, + "loss": 0.976, + "step": 20627 + }, + { + "epoch": 1.52, + "learning_rate": 1.7480833222863033e-05, + "loss": 1.0153, + "step": 20628 + }, + { + "epoch": 1.52, + "learning_rate": 1.748056852378682e-05, + "loss": 1.0668, + "step": 20629 + }, + { + "epoch": 1.52, + "learning_rate": 1.7480303812809127e-05, + "loss": 1.0204, + "step": 20630 + }, + { + "epoch": 1.52, + "learning_rate": 1.748003908993038e-05, + "loss": 0.9734, + "step": 20631 + }, + { + "epoch": 1.52, + "learning_rate": 1.7479774355150998e-05, + "loss": 0.9659, + "step": 20632 + }, + { + "epoch": 1.52, + "learning_rate": 1.7479509608471402e-05, + "loss": 0.9336, + "step": 20633 + }, + { + "epoch": 1.52, + "learning_rate": 1.7479244849892013e-05, + "loss": 0.9045, + "step": 20634 + }, + { + "epoch": 1.52, + "learning_rate": 1.747898007941325e-05, + "loss": 0.954, + "step": 20635 + }, + { + "epoch": 1.52, + "learning_rate": 1.747871529703554e-05, + "loss": 1.0999, + "step": 20636 + }, + { + "epoch": 1.52, + "learning_rate": 1.7478450502759298e-05, + "loss": 1.103, + "step": 20637 + }, + { + "epoch": 1.52, + "learning_rate": 1.7478185696584948e-05, + "loss": 1.0037, + "step": 20638 + }, + { + "epoch": 1.53, + "learning_rate": 1.7477920878512915e-05, + "loss": 1.0289, + "step": 20639 + }, + { + "epoch": 1.53, + "learning_rate": 1.7477656048543617e-05, + "loss": 1.0542, + "step": 20640 + }, + { + "epoch": 1.53, + "learning_rate": 1.747739120667747e-05, + "loss": 1.0246, + "step": 20641 + }, + { + "epoch": 1.53, + "learning_rate": 1.7477126352914905e-05, + "loss": 1.0588, + "step": 20642 + }, + { + "epoch": 1.53, + "learning_rate": 1.747686148725634e-05, + "loss": 1.0414, + "step": 20643 + }, + { + "epoch": 1.53, + "learning_rate": 1.7476596609702193e-05, + "loss": 0.9462, + "step": 20644 + }, + { + "epoch": 1.53, + "learning_rate": 1.7476331720252886e-05, + "loss": 1.029, + "step": 20645 + }, + { + "epoch": 1.53, + "learning_rate": 1.7476066818908842e-05, + "loss": 0.9067, + "step": 20646 + }, + { + "epoch": 1.53, + "learning_rate": 1.7475801905670486e-05, + "loss": 1.1155, + "step": 20647 + }, + { + "epoch": 1.53, + "learning_rate": 1.7475536980538236e-05, + "loss": 1.0225, + "step": 20648 + }, + { + "epoch": 1.53, + "learning_rate": 1.747527204351251e-05, + "loss": 0.9976, + "step": 20649 + }, + { + "epoch": 1.53, + "learning_rate": 1.7475007094593736e-05, + "loss": 1.007, + "step": 20650 + }, + { + "epoch": 1.53, + "learning_rate": 1.7474742133782334e-05, + "loss": 1.0419, + "step": 20651 + }, + { + "epoch": 1.53, + "learning_rate": 1.7474477161078725e-05, + "loss": 1.0214, + "step": 20652 + }, + { + "epoch": 1.53, + "learning_rate": 1.7474212176483325e-05, + "loss": 1.0162, + "step": 20653 + }, + { + "epoch": 1.53, + "learning_rate": 1.7473947179996562e-05, + "loss": 1.0534, + "step": 20654 + }, + { + "epoch": 1.53, + "learning_rate": 1.747368217161886e-05, + "loss": 1.009, + "step": 20655 + }, + { + "epoch": 1.53, + "learning_rate": 1.747341715135063e-05, + "loss": 1.0495, + "step": 20656 + }, + { + "epoch": 1.53, + "learning_rate": 1.7473152119192304e-05, + "loss": 1.0282, + "step": 20657 + }, + { + "epoch": 1.53, + "learning_rate": 1.74728870751443e-05, + "loss": 1.0104, + "step": 20658 + }, + { + "epoch": 1.53, + "learning_rate": 1.747262201920704e-05, + "loss": 1.0031, + "step": 20659 + }, + { + "epoch": 1.53, + "learning_rate": 1.7472356951380944e-05, + "loss": 1.0197, + "step": 20660 + }, + { + "epoch": 1.53, + "learning_rate": 1.7472091871666434e-05, + "loss": 0.9461, + "step": 20661 + }, + { + "epoch": 1.53, + "learning_rate": 1.7471826780063933e-05, + "loss": 0.8655, + "step": 20662 + }, + { + "epoch": 1.53, + "learning_rate": 1.7471561676573866e-05, + "loss": 0.9984, + "step": 20663 + }, + { + "epoch": 1.53, + "learning_rate": 1.7471296561196647e-05, + "loss": 1.0253, + "step": 20664 + }, + { + "epoch": 1.53, + "learning_rate": 1.7471031433932703e-05, + "loss": 1.1898, + "step": 20665 + }, + { + "epoch": 1.53, + "learning_rate": 1.7470766294782457e-05, + "loss": 1.109, + "step": 20666 + }, + { + "epoch": 1.53, + "learning_rate": 1.7470501143746327e-05, + "loss": 0.9819, + "step": 20667 + }, + { + "epoch": 1.53, + "learning_rate": 1.7470235980824735e-05, + "loss": 1.0437, + "step": 20668 + }, + { + "epoch": 1.53, + "learning_rate": 1.746997080601811e-05, + "loss": 0.9937, + "step": 20669 + }, + { + "epoch": 1.53, + "learning_rate": 1.7469705619326866e-05, + "loss": 1.1711, + "step": 20670 + }, + { + "epoch": 1.53, + "learning_rate": 1.746944042075142e-05, + "loss": 0.9673, + "step": 20671 + }, + { + "epoch": 1.53, + "learning_rate": 1.746917521029221e-05, + "loss": 1.0039, + "step": 20672 + }, + { + "epoch": 1.53, + "learning_rate": 1.7468909987949647e-05, + "loss": 1.0297, + "step": 20673 + }, + { + "epoch": 1.53, + "learning_rate": 1.7468644753724155e-05, + "loss": 0.9336, + "step": 20674 + }, + { + "epoch": 1.53, + "learning_rate": 1.7468379507616155e-05, + "loss": 1.022, + "step": 20675 + }, + { + "epoch": 1.53, + "learning_rate": 1.746811424962607e-05, + "loss": 0.9441, + "step": 20676 + }, + { + "epoch": 1.53, + "learning_rate": 1.7467848979754323e-05, + "loss": 0.9857, + "step": 20677 + }, + { + "epoch": 1.53, + "learning_rate": 1.7467583698001334e-05, + "loss": 1.0076, + "step": 20678 + }, + { + "epoch": 1.53, + "learning_rate": 1.746731840436753e-05, + "loss": 1.0577, + "step": 20679 + }, + { + "epoch": 1.53, + "learning_rate": 1.746705309885332e-05, + "loss": 1.0658, + "step": 20680 + }, + { + "epoch": 1.53, + "learning_rate": 1.746678778145914e-05, + "loss": 1.0379, + "step": 20681 + }, + { + "epoch": 1.53, + "learning_rate": 1.746652245218541e-05, + "loss": 1.0592, + "step": 20682 + }, + { + "epoch": 1.53, + "learning_rate": 1.746625711103255e-05, + "loss": 1.0484, + "step": 20683 + }, + { + "epoch": 1.53, + "learning_rate": 1.746599175800098e-05, + "loss": 1.0183, + "step": 20684 + }, + { + "epoch": 1.53, + "learning_rate": 1.7465726393091124e-05, + "loss": 1.0703, + "step": 20685 + }, + { + "epoch": 1.53, + "learning_rate": 1.7465461016303406e-05, + "loss": 1.0766, + "step": 20686 + }, + { + "epoch": 1.53, + "learning_rate": 1.7465195627638244e-05, + "loss": 0.9954, + "step": 20687 + }, + { + "epoch": 1.53, + "learning_rate": 1.7464930227096064e-05, + "loss": 1.0201, + "step": 20688 + }, + { + "epoch": 1.53, + "learning_rate": 1.7464664814677285e-05, + "loss": 0.9378, + "step": 20689 + }, + { + "epoch": 1.53, + "learning_rate": 1.7464399390382333e-05, + "loss": 0.9303, + "step": 20690 + }, + { + "epoch": 1.53, + "learning_rate": 1.7464133954211626e-05, + "loss": 0.9739, + "step": 20691 + }, + { + "epoch": 1.53, + "learning_rate": 1.7463868506165595e-05, + "loss": 1.0503, + "step": 20692 + }, + { + "epoch": 1.53, + "learning_rate": 1.746360304624465e-05, + "loss": 1.0153, + "step": 20693 + }, + { + "epoch": 1.53, + "learning_rate": 1.746333757444922e-05, + "loss": 0.8385, + "step": 20694 + }, + { + "epoch": 1.53, + "learning_rate": 1.746307209077973e-05, + "loss": 1.0147, + "step": 20695 + }, + { + "epoch": 1.53, + "learning_rate": 1.7462806595236593e-05, + "loss": 0.9767, + "step": 20696 + }, + { + "epoch": 1.53, + "learning_rate": 1.746254108782024e-05, + "loss": 1.0692, + "step": 20697 + }, + { + "epoch": 1.53, + "learning_rate": 1.746227556853109e-05, + "loss": 1.0027, + "step": 20698 + }, + { + "epoch": 1.53, + "learning_rate": 1.746201003736957e-05, + "loss": 1.0635, + "step": 20699 + }, + { + "epoch": 1.53, + "learning_rate": 1.7461744494336098e-05, + "loss": 0.9877, + "step": 20700 + }, + { + "epoch": 1.53, + "learning_rate": 1.7461478939431094e-05, + "loss": 0.9584, + "step": 20701 + }, + { + "epoch": 1.53, + "learning_rate": 1.7461213372654985e-05, + "loss": 1.0178, + "step": 20702 + }, + { + "epoch": 1.53, + "learning_rate": 1.7460947794008193e-05, + "loss": 1.0303, + "step": 20703 + }, + { + "epoch": 1.53, + "learning_rate": 1.7460682203491142e-05, + "loss": 1.0156, + "step": 20704 + }, + { + "epoch": 1.53, + "learning_rate": 1.746041660110425e-05, + "loss": 0.9752, + "step": 20705 + }, + { + "epoch": 1.53, + "learning_rate": 1.7460150986847942e-05, + "loss": 1.0852, + "step": 20706 + }, + { + "epoch": 1.53, + "learning_rate": 1.7459885360722643e-05, + "loss": 0.9459, + "step": 20707 + }, + { + "epoch": 1.53, + "learning_rate": 1.745961972272877e-05, + "loss": 1.0465, + "step": 20708 + }, + { + "epoch": 1.53, + "learning_rate": 1.745935407286675e-05, + "loss": 1.0334, + "step": 20709 + }, + { + "epoch": 1.53, + "learning_rate": 1.7459088411137003e-05, + "loss": 1.0823, + "step": 20710 + }, + { + "epoch": 1.53, + "learning_rate": 1.7458822737539952e-05, + "loss": 1.0006, + "step": 20711 + }, + { + "epoch": 1.53, + "learning_rate": 1.7458557052076024e-05, + "loss": 1.0388, + "step": 20712 + }, + { + "epoch": 1.53, + "learning_rate": 1.745829135474564e-05, + "loss": 1.0709, + "step": 20713 + }, + { + "epoch": 1.53, + "learning_rate": 1.7458025645549216e-05, + "loss": 1.0105, + "step": 20714 + }, + { + "epoch": 1.53, + "learning_rate": 1.745775992448718e-05, + "loss": 0.9399, + "step": 20715 + }, + { + "epoch": 1.53, + "learning_rate": 1.7457494191559957e-05, + "loss": 0.9873, + "step": 20716 + }, + { + "epoch": 1.53, + "learning_rate": 1.7457228446767968e-05, + "loss": 1.0014, + "step": 20717 + }, + { + "epoch": 1.53, + "learning_rate": 1.7456962690111635e-05, + "loss": 0.9411, + "step": 20718 + }, + { + "epoch": 1.53, + "learning_rate": 1.745669692159138e-05, + "loss": 1.0351, + "step": 20719 + }, + { + "epoch": 1.53, + "learning_rate": 1.7456431141207632e-05, + "loss": 1.0255, + "step": 20720 + }, + { + "epoch": 1.53, + "learning_rate": 1.74561653489608e-05, + "loss": 1.0588, + "step": 20721 + }, + { + "epoch": 1.53, + "learning_rate": 1.745589954485132e-05, + "loss": 1.0547, + "step": 20722 + }, + { + "epoch": 1.53, + "learning_rate": 1.7455633728879613e-05, + "loss": 0.9198, + "step": 20723 + }, + { + "epoch": 1.53, + "learning_rate": 1.7455367901046096e-05, + "loss": 1.0205, + "step": 20724 + }, + { + "epoch": 1.53, + "learning_rate": 1.7455102061351197e-05, + "loss": 1.0067, + "step": 20725 + }, + { + "epoch": 1.53, + "learning_rate": 1.7454836209795332e-05, + "loss": 1.013, + "step": 20726 + }, + { + "epoch": 1.53, + "learning_rate": 1.7454570346378937e-05, + "loss": 1.0436, + "step": 20727 + }, + { + "epoch": 1.53, + "learning_rate": 1.7454304471102423e-05, + "loss": 0.9975, + "step": 20728 + }, + { + "epoch": 1.53, + "learning_rate": 1.7454038583966218e-05, + "loss": 1.0782, + "step": 20729 + }, + { + "epoch": 1.53, + "learning_rate": 1.7453772684970746e-05, + "loss": 0.9507, + "step": 20730 + }, + { + "epoch": 1.53, + "learning_rate": 1.7453506774116426e-05, + "loss": 1.0266, + "step": 20731 + }, + { + "epoch": 1.53, + "learning_rate": 1.7453240851403685e-05, + "loss": 1.0829, + "step": 20732 + }, + { + "epoch": 1.53, + "learning_rate": 1.7452974916832944e-05, + "loss": 0.9106, + "step": 20733 + }, + { + "epoch": 1.53, + "learning_rate": 1.7452708970404626e-05, + "loss": 1.0442, + "step": 20734 + }, + { + "epoch": 1.53, + "learning_rate": 1.7452443012119154e-05, + "loss": 1.0766, + "step": 20735 + }, + { + "epoch": 1.53, + "learning_rate": 1.7452177041976953e-05, + "loss": 1.0285, + "step": 20736 + }, + { + "epoch": 1.53, + "learning_rate": 1.7451911059978445e-05, + "loss": 1.0741, + "step": 20737 + }, + { + "epoch": 1.53, + "learning_rate": 1.745164506612405e-05, + "loss": 1.1068, + "step": 20738 + }, + { + "epoch": 1.53, + "learning_rate": 1.74513790604142e-05, + "loss": 1.1535, + "step": 20739 + }, + { + "epoch": 1.53, + "learning_rate": 1.7451113042849308e-05, + "loss": 1.1135, + "step": 20740 + }, + { + "epoch": 1.53, + "learning_rate": 1.7450847013429803e-05, + "loss": 0.9297, + "step": 20741 + }, + { + "epoch": 1.53, + "learning_rate": 1.745058097215611e-05, + "loss": 1.0468, + "step": 20742 + }, + { + "epoch": 1.53, + "learning_rate": 1.7450314919028643e-05, + "loss": 1.0585, + "step": 20743 + }, + { + "epoch": 1.53, + "learning_rate": 1.7450048854047836e-05, + "loss": 0.9818, + "step": 20744 + }, + { + "epoch": 1.53, + "learning_rate": 1.7449782777214106e-05, + "loss": 1.0476, + "step": 20745 + }, + { + "epoch": 1.53, + "learning_rate": 1.7449516688527882e-05, + "loss": 0.9924, + "step": 20746 + }, + { + "epoch": 1.53, + "learning_rate": 1.744925058798958e-05, + "loss": 1.0236, + "step": 20747 + }, + { + "epoch": 1.53, + "learning_rate": 1.7448984475599626e-05, + "loss": 1.0475, + "step": 20748 + }, + { + "epoch": 1.53, + "learning_rate": 1.7448718351358445e-05, + "loss": 0.9793, + "step": 20749 + }, + { + "epoch": 1.53, + "learning_rate": 1.7448452215266464e-05, + "loss": 0.919, + "step": 20750 + }, + { + "epoch": 1.53, + "learning_rate": 1.7448186067324098e-05, + "loss": 0.9916, + "step": 20751 + }, + { + "epoch": 1.53, + "learning_rate": 1.7447919907531775e-05, + "loss": 0.9523, + "step": 20752 + }, + { + "epoch": 1.53, + "learning_rate": 1.7447653735889916e-05, + "loss": 0.978, + "step": 20753 + }, + { + "epoch": 1.53, + "learning_rate": 1.744738755239895e-05, + "loss": 1.01, + "step": 20754 + }, + { + "epoch": 1.53, + "learning_rate": 1.7447121357059297e-05, + "loss": 1.0107, + "step": 20755 + }, + { + "epoch": 1.53, + "learning_rate": 1.7446855149871378e-05, + "loss": 1.095, + "step": 20756 + }, + { + "epoch": 1.53, + "learning_rate": 1.744658893083562e-05, + "loss": 0.9617, + "step": 20757 + }, + { + "epoch": 1.53, + "learning_rate": 1.7446322699952445e-05, + "loss": 0.9912, + "step": 20758 + }, + { + "epoch": 1.53, + "learning_rate": 1.7446056457222278e-05, + "loss": 1.0301, + "step": 20759 + }, + { + "epoch": 1.53, + "learning_rate": 1.7445790202645542e-05, + "loss": 1.0099, + "step": 20760 + }, + { + "epoch": 1.53, + "learning_rate": 1.7445523936222658e-05, + "loss": 0.9903, + "step": 20761 + }, + { + "epoch": 1.53, + "learning_rate": 1.7445257657954054e-05, + "loss": 1.0605, + "step": 20762 + }, + { + "epoch": 1.53, + "learning_rate": 1.7444991367840153e-05, + "loss": 1.1159, + "step": 20763 + }, + { + "epoch": 1.53, + "learning_rate": 1.7444725065881373e-05, + "loss": 1.0535, + "step": 20764 + }, + { + "epoch": 1.53, + "learning_rate": 1.7444458752078145e-05, + "loss": 1.0911, + "step": 20765 + }, + { + "epoch": 1.53, + "learning_rate": 1.744419242643089e-05, + "loss": 0.9818, + "step": 20766 + }, + { + "epoch": 1.53, + "learning_rate": 1.744392608894003e-05, + "loss": 0.9504, + "step": 20767 + }, + { + "epoch": 1.53, + "learning_rate": 1.7443659739605993e-05, + "loss": 0.8992, + "step": 20768 + }, + { + "epoch": 1.53, + "learning_rate": 1.7443393378429195e-05, + "loss": 1.0528, + "step": 20769 + }, + { + "epoch": 1.53, + "learning_rate": 1.7443127005410067e-05, + "loss": 1.0204, + "step": 20770 + }, + { + "epoch": 1.53, + "learning_rate": 1.7442860620549032e-05, + "loss": 1.0638, + "step": 20771 + }, + { + "epoch": 1.53, + "learning_rate": 1.744259422384651e-05, + "loss": 0.9342, + "step": 20772 + }, + { + "epoch": 1.53, + "learning_rate": 1.7442327815302928e-05, + "loss": 1.1734, + "step": 20773 + }, + { + "epoch": 1.53, + "learning_rate": 1.7442061394918707e-05, + "loss": 1.0847, + "step": 20774 + }, + { + "epoch": 1.54, + "learning_rate": 1.7441794962694276e-05, + "loss": 0.8625, + "step": 20775 + }, + { + "epoch": 1.54, + "learning_rate": 1.7441528518630053e-05, + "loss": 0.9952, + "step": 20776 + }, + { + "epoch": 1.54, + "learning_rate": 1.7441262062726467e-05, + "loss": 1.0428, + "step": 20777 + }, + { + "epoch": 1.54, + "learning_rate": 1.744099559498394e-05, + "loss": 0.9902, + "step": 20778 + }, + { + "epoch": 1.54, + "learning_rate": 1.7440729115402892e-05, + "loss": 1.0353, + "step": 20779 + }, + { + "epoch": 1.54, + "learning_rate": 1.744046262398375e-05, + "loss": 0.9364, + "step": 20780 + }, + { + "epoch": 1.54, + "learning_rate": 1.7440196120726944e-05, + "loss": 1.1116, + "step": 20781 + }, + { + "epoch": 1.54, + "learning_rate": 1.743992960563289e-05, + "loss": 0.9008, + "step": 20782 + }, + { + "epoch": 1.54, + "learning_rate": 1.7439663078702013e-05, + "loss": 1.0599, + "step": 20783 + }, + { + "epoch": 1.54, + "learning_rate": 1.743939653993474e-05, + "loss": 1.0455, + "step": 20784 + }, + { + "epoch": 1.54, + "learning_rate": 1.743912998933149e-05, + "loss": 1.0695, + "step": 20785 + }, + { + "epoch": 1.54, + "learning_rate": 1.7438863426892695e-05, + "loss": 1.0636, + "step": 20786 + }, + { + "epoch": 1.54, + "learning_rate": 1.7438596852618772e-05, + "loss": 1.0191, + "step": 20787 + }, + { + "epoch": 1.54, + "learning_rate": 1.743833026651015e-05, + "loss": 1.0537, + "step": 20788 + }, + { + "epoch": 1.54, + "learning_rate": 1.743806366856725e-05, + "loss": 1.0804, + "step": 20789 + }, + { + "epoch": 1.54, + "learning_rate": 1.7437797058790496e-05, + "loss": 0.9757, + "step": 20790 + }, + { + "epoch": 1.54, + "learning_rate": 1.7437530437180314e-05, + "loss": 1.0181, + "step": 20791 + }, + { + "epoch": 1.54, + "learning_rate": 1.7437263803737124e-05, + "loss": 1.0382, + "step": 20792 + }, + { + "epoch": 1.54, + "learning_rate": 1.743699715846136e-05, + "loss": 1.099, + "step": 20793 + }, + { + "epoch": 1.54, + "learning_rate": 1.743673050135344e-05, + "loss": 0.9366, + "step": 20794 + }, + { + "epoch": 1.54, + "learning_rate": 1.7436463832413782e-05, + "loss": 1.0647, + "step": 20795 + }, + { + "epoch": 1.54, + "learning_rate": 1.743619715164282e-05, + "loss": 0.9235, + "step": 20796 + }, + { + "epoch": 1.54, + "learning_rate": 1.7435930459040973e-05, + "loss": 0.9771, + "step": 20797 + }, + { + "epoch": 1.54, + "learning_rate": 1.7435663754608668e-05, + "loss": 1.0533, + "step": 20798 + }, + { + "epoch": 1.54, + "learning_rate": 1.743539703834633e-05, + "loss": 0.8255, + "step": 20799 + }, + { + "epoch": 1.54, + "learning_rate": 1.743513031025438e-05, + "loss": 1.0914, + "step": 20800 + }, + { + "epoch": 1.54, + "learning_rate": 1.7434863570333247e-05, + "loss": 0.8843, + "step": 20801 + }, + { + "epoch": 1.54, + "learning_rate": 1.743459681858335e-05, + "loss": 1.0824, + "step": 20802 + }, + { + "epoch": 1.54, + "learning_rate": 1.743433005500512e-05, + "loss": 1.0797, + "step": 20803 + }, + { + "epoch": 1.54, + "learning_rate": 1.743406327959897e-05, + "loss": 0.9855, + "step": 20804 + }, + { + "epoch": 1.54, + "learning_rate": 1.7433796492365332e-05, + "loss": 1.181, + "step": 20805 + }, + { + "epoch": 1.54, + "learning_rate": 1.7433529693304634e-05, + "loss": 1.0097, + "step": 20806 + }, + { + "epoch": 1.54, + "learning_rate": 1.7433262882417295e-05, + "loss": 0.918, + "step": 20807 + }, + { + "epoch": 1.54, + "learning_rate": 1.7432996059703744e-05, + "loss": 0.9959, + "step": 20808 + }, + { + "epoch": 1.54, + "learning_rate": 1.74327292251644e-05, + "loss": 0.9665, + "step": 20809 + }, + { + "epoch": 1.54, + "learning_rate": 1.743246237879969e-05, + "loss": 1.0113, + "step": 20810 + }, + { + "epoch": 1.54, + "learning_rate": 1.743219552061004e-05, + "loss": 1.142, + "step": 20811 + }, + { + "epoch": 1.54, + "learning_rate": 1.7431928650595874e-05, + "loss": 0.9981, + "step": 20812 + }, + { + "epoch": 1.54, + "learning_rate": 1.7431661768757614e-05, + "loss": 1.0808, + "step": 20813 + }, + { + "epoch": 1.54, + "learning_rate": 1.7431394875095685e-05, + "loss": 0.9431, + "step": 20814 + }, + { + "epoch": 1.54, + "learning_rate": 1.7431127969610514e-05, + "loss": 0.8998, + "step": 20815 + }, + { + "epoch": 1.54, + "learning_rate": 1.743086105230253e-05, + "loss": 1.0062, + "step": 20816 + }, + { + "epoch": 1.54, + "learning_rate": 1.7430594123172145e-05, + "loss": 1.035, + "step": 20817 + }, + { + "epoch": 1.54, + "learning_rate": 1.7430327182219794e-05, + "loss": 1.0843, + "step": 20818 + }, + { + "epoch": 1.54, + "learning_rate": 1.74300602294459e-05, + "loss": 1.0139, + "step": 20819 + }, + { + "epoch": 1.54, + "learning_rate": 1.7429793264850887e-05, + "loss": 0.9645, + "step": 20820 + }, + { + "epoch": 1.54, + "learning_rate": 1.7429526288435177e-05, + "loss": 0.9757, + "step": 20821 + }, + { + "epoch": 1.54, + "learning_rate": 1.7429259300199194e-05, + "loss": 0.9328, + "step": 20822 + }, + { + "epoch": 1.54, + "learning_rate": 1.742899230014337e-05, + "loss": 0.9955, + "step": 20823 + }, + { + "epoch": 1.54, + "learning_rate": 1.7428725288268124e-05, + "loss": 0.9907, + "step": 20824 + }, + { + "epoch": 1.54, + "learning_rate": 1.7428458264573885e-05, + "loss": 0.9684, + "step": 20825 + }, + { + "epoch": 1.54, + "learning_rate": 1.7428191229061072e-05, + "loss": 1.0312, + "step": 20826 + }, + { + "epoch": 1.54, + "learning_rate": 1.7427924181730114e-05, + "loss": 1.0178, + "step": 20827 + }, + { + "epoch": 1.54, + "learning_rate": 1.7427657122581437e-05, + "loss": 0.9625, + "step": 20828 + }, + { + "epoch": 1.54, + "learning_rate": 1.742739005161546e-05, + "loss": 1.0776, + "step": 20829 + }, + { + "epoch": 1.54, + "learning_rate": 1.7427122968832614e-05, + "loss": 1.0029, + "step": 20830 + }, + { + "epoch": 1.54, + "learning_rate": 1.742685587423332e-05, + "loss": 0.9715, + "step": 20831 + }, + { + "epoch": 1.54, + "learning_rate": 1.742658876781801e-05, + "loss": 1.0737, + "step": 20832 + }, + { + "epoch": 1.54, + "learning_rate": 1.7426321649587098e-05, + "loss": 1.0091, + "step": 20833 + }, + { + "epoch": 1.54, + "learning_rate": 1.7426054519541014e-05, + "loss": 0.9547, + "step": 20834 + }, + { + "epoch": 1.54, + "learning_rate": 1.7425787377680186e-05, + "loss": 0.9712, + "step": 20835 + }, + { + "epoch": 1.54, + "learning_rate": 1.7425520224005038e-05, + "loss": 1.0054, + "step": 20836 + }, + { + "epoch": 1.54, + "learning_rate": 1.742525305851599e-05, + "loss": 1.078, + "step": 20837 + }, + { + "epoch": 1.54, + "learning_rate": 1.7424985881213473e-05, + "loss": 1.0271, + "step": 20838 + }, + { + "epoch": 1.54, + "learning_rate": 1.742471869209791e-05, + "loss": 0.9952, + "step": 20839 + }, + { + "epoch": 1.54, + "learning_rate": 1.7424451491169722e-05, + "loss": 0.9354, + "step": 20840 + }, + { + "epoch": 1.54, + "learning_rate": 1.7424184278429343e-05, + "loss": 0.9938, + "step": 20841 + }, + { + "epoch": 1.54, + "learning_rate": 1.742391705387719e-05, + "loss": 0.9474, + "step": 20842 + }, + { + "epoch": 1.54, + "learning_rate": 1.742364981751369e-05, + "loss": 0.9936, + "step": 20843 + }, + { + "epoch": 1.54, + "learning_rate": 1.7423382569339274e-05, + "loss": 1.0631, + "step": 20844 + }, + { + "epoch": 1.54, + "learning_rate": 1.742311530935436e-05, + "loss": 1.0616, + "step": 20845 + }, + { + "epoch": 1.54, + "learning_rate": 1.7422848037559375e-05, + "loss": 1.0373, + "step": 20846 + }, + { + "epoch": 1.54, + "learning_rate": 1.7422580753954747e-05, + "loss": 0.9722, + "step": 20847 + }, + { + "epoch": 1.54, + "learning_rate": 1.7422313458540896e-05, + "loss": 1.0546, + "step": 20848 + }, + { + "epoch": 1.54, + "learning_rate": 1.742204615131825e-05, + "loss": 1.0134, + "step": 20849 + }, + { + "epoch": 1.54, + "learning_rate": 1.7421778832287238e-05, + "loss": 1.1008, + "step": 20850 + }, + { + "epoch": 1.54, + "learning_rate": 1.742151150144828e-05, + "loss": 0.9546, + "step": 20851 + }, + { + "epoch": 1.54, + "learning_rate": 1.7421244158801807e-05, + "loss": 1.0054, + "step": 20852 + }, + { + "epoch": 1.54, + "learning_rate": 1.742097680434824e-05, + "loss": 1.04, + "step": 20853 + }, + { + "epoch": 1.54, + "learning_rate": 1.7420709438088e-05, + "loss": 1.0576, + "step": 20854 + }, + { + "epoch": 1.54, + "learning_rate": 1.7420442060021522e-05, + "loss": 0.9196, + "step": 20855 + }, + { + "epoch": 1.54, + "learning_rate": 1.7420174670149227e-05, + "loss": 1.0237, + "step": 20856 + }, + { + "epoch": 1.54, + "learning_rate": 1.7419907268471536e-05, + "loss": 1.0033, + "step": 20857 + }, + { + "epoch": 1.54, + "learning_rate": 1.7419639854988882e-05, + "loss": 1.125, + "step": 20858 + }, + { + "epoch": 1.54, + "learning_rate": 1.7419372429701686e-05, + "loss": 0.9996, + "step": 20859 + }, + { + "epoch": 1.54, + "learning_rate": 1.7419104992610377e-05, + "loss": 1.0086, + "step": 20860 + }, + { + "epoch": 1.54, + "learning_rate": 1.7418837543715373e-05, + "loss": 1.0369, + "step": 20861 + }, + { + "epoch": 1.54, + "learning_rate": 1.741857008301711e-05, + "loss": 0.9637, + "step": 20862 + }, + { + "epoch": 1.54, + "learning_rate": 1.7418302610516006e-05, + "loss": 1.0336, + "step": 20863 + }, + { + "epoch": 1.54, + "learning_rate": 1.741803512621249e-05, + "loss": 0.9761, + "step": 20864 + }, + { + "epoch": 1.54, + "learning_rate": 1.7417767630106984e-05, + "loss": 1.0594, + "step": 20865 + }, + { + "epoch": 1.54, + "learning_rate": 1.7417500122199916e-05, + "loss": 0.9968, + "step": 20866 + }, + { + "epoch": 1.54, + "learning_rate": 1.7417232602491714e-05, + "loss": 1.0346, + "step": 20867 + }, + { + "epoch": 1.54, + "learning_rate": 1.7416965070982798e-05, + "loss": 0.9799, + "step": 20868 + }, + { + "epoch": 1.54, + "learning_rate": 1.74166975276736e-05, + "loss": 1.0514, + "step": 20869 + }, + { + "epoch": 1.54, + "learning_rate": 1.741642997256454e-05, + "loss": 1.0187, + "step": 20870 + }, + { + "epoch": 1.54, + "learning_rate": 1.7416162405656046e-05, + "loss": 0.9557, + "step": 20871 + }, + { + "epoch": 1.54, + "learning_rate": 1.7415894826948543e-05, + "loss": 1.0925, + "step": 20872 + }, + { + "epoch": 1.54, + "learning_rate": 1.741562723644246e-05, + "loss": 1.0133, + "step": 20873 + }, + { + "epoch": 1.54, + "learning_rate": 1.741535963413822e-05, + "loss": 1.0341, + "step": 20874 + }, + { + "epoch": 1.54, + "learning_rate": 1.7415092020036243e-05, + "loss": 0.962, + "step": 20875 + }, + { + "epoch": 1.54, + "learning_rate": 1.7414824394136967e-05, + "loss": 1.0386, + "step": 20876 + }, + { + "epoch": 1.54, + "learning_rate": 1.741455675644081e-05, + "loss": 0.9832, + "step": 20877 + }, + { + "epoch": 1.54, + "learning_rate": 1.74142891069482e-05, + "loss": 1.112, + "step": 20878 + }, + { + "epoch": 1.54, + "learning_rate": 1.7414021445659562e-05, + "loss": 1.1054, + "step": 20879 + }, + { + "epoch": 1.54, + "learning_rate": 1.7413753772575323e-05, + "loss": 0.8978, + "step": 20880 + }, + { + "epoch": 1.54, + "learning_rate": 1.7413486087695903e-05, + "loss": 1.0444, + "step": 20881 + }, + { + "epoch": 1.54, + "learning_rate": 1.7413218391021736e-05, + "loss": 1.0422, + "step": 20882 + }, + { + "epoch": 1.54, + "learning_rate": 1.741295068255324e-05, + "loss": 0.9465, + "step": 20883 + }, + { + "epoch": 1.54, + "learning_rate": 1.7412682962290852e-05, + "loss": 1.002, + "step": 20884 + }, + { + "epoch": 1.54, + "learning_rate": 1.741241523023499e-05, + "loss": 0.95, + "step": 20885 + }, + { + "epoch": 1.54, + "learning_rate": 1.7412147486386082e-05, + "loss": 0.9668, + "step": 20886 + }, + { + "epoch": 1.54, + "learning_rate": 1.741187973074455e-05, + "loss": 1.1301, + "step": 20887 + }, + { + "epoch": 1.54, + "learning_rate": 1.7411611963310826e-05, + "loss": 1.0556, + "step": 20888 + }, + { + "epoch": 1.54, + "learning_rate": 1.7411344184085336e-05, + "loss": 1.0199, + "step": 20889 + }, + { + "epoch": 1.54, + "learning_rate": 1.7411076393068496e-05, + "loss": 0.9581, + "step": 20890 + }, + { + "epoch": 1.54, + "learning_rate": 1.7410808590260745e-05, + "loss": 1.0024, + "step": 20891 + }, + { + "epoch": 1.54, + "learning_rate": 1.7410540775662505e-05, + "loss": 1.1256, + "step": 20892 + }, + { + "epoch": 1.54, + "learning_rate": 1.7410272949274196e-05, + "loss": 1.0044, + "step": 20893 + }, + { + "epoch": 1.54, + "learning_rate": 1.741000511109625e-05, + "loss": 1.1101, + "step": 20894 + }, + { + "epoch": 1.54, + "learning_rate": 1.7409737261129095e-05, + "loss": 1.0396, + "step": 20895 + }, + { + "epoch": 1.54, + "learning_rate": 1.740946939937315e-05, + "loss": 1.0544, + "step": 20896 + }, + { + "epoch": 1.54, + "learning_rate": 1.7409201525828847e-05, + "loss": 0.9319, + "step": 20897 + }, + { + "epoch": 1.54, + "learning_rate": 1.740893364049661e-05, + "loss": 1.0678, + "step": 20898 + }, + { + "epoch": 1.54, + "learning_rate": 1.7408665743376866e-05, + "loss": 0.9057, + "step": 20899 + }, + { + "epoch": 1.54, + "learning_rate": 1.7408397834470042e-05, + "loss": 0.9504, + "step": 20900 + }, + { + "epoch": 1.54, + "learning_rate": 1.740812991377656e-05, + "loss": 0.9301, + "step": 20901 + }, + { + "epoch": 1.54, + "learning_rate": 1.740786198129685e-05, + "loss": 1.0678, + "step": 20902 + }, + { + "epoch": 1.54, + "learning_rate": 1.740759403703134e-05, + "loss": 1.0494, + "step": 20903 + }, + { + "epoch": 1.54, + "learning_rate": 1.7407326080980452e-05, + "loss": 1.0769, + "step": 20904 + }, + { + "epoch": 1.54, + "learning_rate": 1.7407058113144618e-05, + "loss": 1.0469, + "step": 20905 + }, + { + "epoch": 1.54, + "learning_rate": 1.7406790133524255e-05, + "loss": 0.8956, + "step": 20906 + }, + { + "epoch": 1.54, + "learning_rate": 1.7406522142119796e-05, + "loss": 1.0962, + "step": 20907 + }, + { + "epoch": 1.54, + "learning_rate": 1.740625413893167e-05, + "loss": 0.8858, + "step": 20908 + }, + { + "epoch": 1.54, + "learning_rate": 1.7405986123960298e-05, + "loss": 0.9651, + "step": 20909 + }, + { + "epoch": 1.55, + "learning_rate": 1.7405718097206105e-05, + "loss": 0.9793, + "step": 20910 + }, + { + "epoch": 1.55, + "learning_rate": 1.7405450058669525e-05, + "loss": 1.1254, + "step": 20911 + }, + { + "epoch": 1.55, + "learning_rate": 1.7405182008350974e-05, + "loss": 0.9953, + "step": 20912 + }, + { + "epoch": 1.55, + "learning_rate": 1.740491394625089e-05, + "loss": 1.0099, + "step": 20913 + }, + { + "epoch": 1.55, + "learning_rate": 1.740464587236969e-05, + "loss": 0.9787, + "step": 20914 + }, + { + "epoch": 1.55, + "learning_rate": 1.7404377786707806e-05, + "loss": 0.948, + "step": 20915 + }, + { + "epoch": 1.55, + "learning_rate": 1.7404109689265664e-05, + "loss": 1.1142, + "step": 20916 + }, + { + "epoch": 1.55, + "learning_rate": 1.7403841580043686e-05, + "loss": 1.085, + "step": 20917 + }, + { + "epoch": 1.55, + "learning_rate": 1.7403573459042306e-05, + "loss": 0.9262, + "step": 20918 + }, + { + "epoch": 1.55, + "learning_rate": 1.7403305326261947e-05, + "loss": 0.9522, + "step": 20919 + }, + { + "epoch": 1.55, + "learning_rate": 1.740303718170303e-05, + "loss": 0.9366, + "step": 20920 + }, + { + "epoch": 1.55, + "learning_rate": 1.740276902536599e-05, + "loss": 1.0394, + "step": 20921 + }, + { + "epoch": 1.55, + "learning_rate": 1.740250085725125e-05, + "loss": 1.1626, + "step": 20922 + }, + { + "epoch": 1.55, + "learning_rate": 1.7402232677359236e-05, + "loss": 1.0096, + "step": 20923 + }, + { + "epoch": 1.55, + "learning_rate": 1.7401964485690378e-05, + "loss": 0.976, + "step": 20924 + }, + { + "epoch": 1.55, + "learning_rate": 1.74016962822451e-05, + "loss": 0.9978, + "step": 20925 + }, + { + "epoch": 1.55, + "learning_rate": 1.740142806702383e-05, + "loss": 1.0364, + "step": 20926 + }, + { + "epoch": 1.55, + "learning_rate": 1.740115984002699e-05, + "loss": 1.0054, + "step": 20927 + }, + { + "epoch": 1.55, + "learning_rate": 1.740089160125501e-05, + "loss": 1.081, + "step": 20928 + }, + { + "epoch": 1.55, + "learning_rate": 1.740062335070832e-05, + "loss": 0.9053, + "step": 20929 + }, + { + "epoch": 1.55, + "learning_rate": 1.7400355088387346e-05, + "loss": 0.9484, + "step": 20930 + }, + { + "epoch": 1.55, + "learning_rate": 1.7400086814292513e-05, + "loss": 0.9725, + "step": 20931 + }, + { + "epoch": 1.55, + "learning_rate": 1.7399818528424247e-05, + "loss": 1.0698, + "step": 20932 + }, + { + "epoch": 1.55, + "learning_rate": 1.7399550230782975e-05, + "loss": 0.9439, + "step": 20933 + }, + { + "epoch": 1.55, + "learning_rate": 1.7399281921369124e-05, + "loss": 1.1231, + "step": 20934 + }, + { + "epoch": 1.55, + "learning_rate": 1.7399013600183117e-05, + "loss": 1.0258, + "step": 20935 + }, + { + "epoch": 1.55, + "learning_rate": 1.739874526722539e-05, + "loss": 0.9533, + "step": 20936 + }, + { + "epoch": 1.55, + "learning_rate": 1.7398476922496367e-05, + "loss": 1.0217, + "step": 20937 + }, + { + "epoch": 1.55, + "learning_rate": 1.739820856599647e-05, + "loss": 0.9957, + "step": 20938 + }, + { + "epoch": 1.55, + "learning_rate": 1.7397940197726127e-05, + "loss": 1.0333, + "step": 20939 + }, + { + "epoch": 1.55, + "learning_rate": 1.739767181768577e-05, + "loss": 1.0143, + "step": 20940 + }, + { + "epoch": 1.55, + "learning_rate": 1.7397403425875825e-05, + "loss": 1.0866, + "step": 20941 + }, + { + "epoch": 1.55, + "learning_rate": 1.7397135022296715e-05, + "loss": 0.9231, + "step": 20942 + }, + { + "epoch": 1.55, + "learning_rate": 1.7396866606948868e-05, + "loss": 1.0618, + "step": 20943 + }, + { + "epoch": 1.55, + "learning_rate": 1.7396598179832715e-05, + "loss": 0.9667, + "step": 20944 + }, + { + "epoch": 1.55, + "learning_rate": 1.7396329740948678e-05, + "loss": 1.0073, + "step": 20945 + }, + { + "epoch": 1.55, + "learning_rate": 1.7396061290297183e-05, + "loss": 1.0044, + "step": 20946 + }, + { + "epoch": 1.55, + "learning_rate": 1.7395792827878664e-05, + "loss": 1.1336, + "step": 20947 + }, + { + "epoch": 1.55, + "learning_rate": 1.7395524353693547e-05, + "loss": 1.058, + "step": 20948 + }, + { + "epoch": 1.55, + "learning_rate": 1.7395255867742252e-05, + "loss": 1.0011, + "step": 20949 + }, + { + "epoch": 1.55, + "learning_rate": 1.739498737002521e-05, + "loss": 1.0473, + "step": 20950 + }, + { + "epoch": 1.55, + "learning_rate": 1.7394718860542852e-05, + "loss": 0.9992, + "step": 20951 + }, + { + "epoch": 1.55, + "learning_rate": 1.7394450339295602e-05, + "loss": 1.0054, + "step": 20952 + }, + { + "epoch": 1.55, + "learning_rate": 1.7394181806283888e-05, + "loss": 1.0719, + "step": 20953 + }, + { + "epoch": 1.55, + "learning_rate": 1.7393913261508136e-05, + "loss": 0.9828, + "step": 20954 + }, + { + "epoch": 1.55, + "learning_rate": 1.7393644704968774e-05, + "loss": 1.0653, + "step": 20955 + }, + { + "epoch": 1.55, + "learning_rate": 1.7393376136666224e-05, + "loss": 0.9998, + "step": 20956 + }, + { + "epoch": 1.55, + "learning_rate": 1.7393107556600925e-05, + "loss": 0.9966, + "step": 20957 + }, + { + "epoch": 1.55, + "learning_rate": 1.7392838964773293e-05, + "loss": 1.1048, + "step": 20958 + }, + { + "epoch": 1.55, + "learning_rate": 1.7392570361183764e-05, + "loss": 0.9355, + "step": 20959 + }, + { + "epoch": 1.55, + "learning_rate": 1.7392301745832756e-05, + "loss": 1.0384, + "step": 20960 + }, + { + "epoch": 1.55, + "learning_rate": 1.739203311872071e-05, + "loss": 1.0193, + "step": 20961 + }, + { + "epoch": 1.55, + "learning_rate": 1.7391764479848037e-05, + "loss": 1.0334, + "step": 20962 + }, + { + "epoch": 1.55, + "learning_rate": 1.7391495829215176e-05, + "loss": 1.1618, + "step": 20963 + }, + { + "epoch": 1.55, + "learning_rate": 1.739122716682255e-05, + "loss": 0.961, + "step": 20964 + }, + { + "epoch": 1.55, + "learning_rate": 1.7390958492670586e-05, + "loss": 0.9353, + "step": 20965 + }, + { + "epoch": 1.55, + "learning_rate": 1.7390689806759718e-05, + "loss": 0.945, + "step": 20966 + }, + { + "epoch": 1.55, + "learning_rate": 1.7390421109090362e-05, + "loss": 1.0843, + "step": 20967 + }, + { + "epoch": 1.55, + "learning_rate": 1.7390152399662956e-05, + "loss": 1.0902, + "step": 20968 + }, + { + "epoch": 1.55, + "learning_rate": 1.738988367847792e-05, + "loss": 0.9924, + "step": 20969 + }, + { + "epoch": 1.55, + "learning_rate": 1.7389614945535688e-05, + "loss": 0.9403, + "step": 20970 + }, + { + "epoch": 1.55, + "learning_rate": 1.738934620083668e-05, + "loss": 0.9384, + "step": 20971 + }, + { + "epoch": 1.55, + "learning_rate": 1.7389077444381334e-05, + "loss": 0.9283, + "step": 20972 + }, + { + "epoch": 1.55, + "learning_rate": 1.7388808676170068e-05, + "loss": 1.144, + "step": 20973 + }, + { + "epoch": 1.55, + "learning_rate": 1.7388539896203313e-05, + "loss": 0.9706, + "step": 20974 + }, + { + "epoch": 1.55, + "learning_rate": 1.7388271104481495e-05, + "loss": 0.9367, + "step": 20975 + }, + { + "epoch": 1.55, + "learning_rate": 1.738800230100505e-05, + "loss": 1.094, + "step": 20976 + }, + { + "epoch": 1.55, + "learning_rate": 1.738773348577439e-05, + "loss": 0.9807, + "step": 20977 + }, + { + "epoch": 1.55, + "learning_rate": 1.738746465878996e-05, + "loss": 0.921, + "step": 20978 + }, + { + "epoch": 1.55, + "learning_rate": 1.7387195820052175e-05, + "loss": 0.8085, + "step": 20979 + }, + { + "epoch": 1.55, + "learning_rate": 1.7386926969561466e-05, + "loss": 0.9989, + "step": 20980 + }, + { + "epoch": 1.55, + "learning_rate": 1.7386658107318264e-05, + "loss": 1.118, + "step": 20981 + }, + { + "epoch": 1.55, + "learning_rate": 1.7386389233322994e-05, + "loss": 1.0172, + "step": 20982 + }, + { + "epoch": 1.55, + "learning_rate": 1.7386120347576085e-05, + "loss": 0.9187, + "step": 20983 + }, + { + "epoch": 1.55, + "learning_rate": 1.7385851450077965e-05, + "loss": 1.0394, + "step": 20984 + }, + { + "epoch": 1.55, + "learning_rate": 1.738558254082906e-05, + "loss": 0.9304, + "step": 20985 + }, + { + "epoch": 1.55, + "learning_rate": 1.7385313619829796e-05, + "loss": 1.1246, + "step": 20986 + }, + { + "epoch": 1.55, + "learning_rate": 1.738504468708061e-05, + "loss": 1.0349, + "step": 20987 + }, + { + "epoch": 1.55, + "learning_rate": 1.7384775742581918e-05, + "loss": 0.9506, + "step": 20988 + }, + { + "epoch": 1.55, + "learning_rate": 1.7384506786334155e-05, + "loss": 1.1208, + "step": 20989 + }, + { + "epoch": 1.55, + "learning_rate": 1.738423781833775e-05, + "loss": 1.0173, + "step": 20990 + }, + { + "epoch": 1.55, + "learning_rate": 1.7383968838593125e-05, + "loss": 1.0431, + "step": 20991 + }, + { + "epoch": 1.55, + "learning_rate": 1.7383699847100715e-05, + "loss": 1.0463, + "step": 20992 + }, + { + "epoch": 1.55, + "learning_rate": 1.738343084386094e-05, + "loss": 1.0374, + "step": 20993 + }, + { + "epoch": 1.55, + "learning_rate": 1.7383161828874234e-05, + "loss": 1.1322, + "step": 20994 + }, + { + "epoch": 1.55, + "learning_rate": 1.738289280214102e-05, + "loss": 1.067, + "step": 20995 + }, + { + "epoch": 1.55, + "learning_rate": 1.7382623763661735e-05, + "loss": 1.0743, + "step": 20996 + }, + { + "epoch": 1.55, + "learning_rate": 1.73823547134368e-05, + "loss": 0.8631, + "step": 20997 + }, + { + "epoch": 1.55, + "learning_rate": 1.738208565146664e-05, + "loss": 0.9453, + "step": 20998 + }, + { + "epoch": 1.55, + "learning_rate": 1.738181657775169e-05, + "loss": 1.0946, + "step": 20999 + }, + { + "epoch": 1.55, + "learning_rate": 1.7381547492292376e-05, + "loss": 1.0352, + "step": 21000 + }, + { + "epoch": 1.55, + "learning_rate": 1.738127839508913e-05, + "loss": 0.9731, + "step": 21001 + }, + { + "epoch": 1.55, + "learning_rate": 1.7381009286142366e-05, + "loss": 0.9614, + "step": 21002 + }, + { + "epoch": 1.55, + "learning_rate": 1.738074016545253e-05, + "loss": 0.9564, + "step": 21003 + }, + { + "epoch": 1.55, + "learning_rate": 1.7380471033020036e-05, + "loss": 1.0396, + "step": 21004 + }, + { + "epoch": 1.55, + "learning_rate": 1.7380201888845322e-05, + "loss": 0.9599, + "step": 21005 + }, + { + "epoch": 1.55, + "learning_rate": 1.7379932732928807e-05, + "loss": 0.9161, + "step": 21006 + }, + { + "epoch": 1.55, + "learning_rate": 1.737966356527093e-05, + "loss": 1.0193, + "step": 21007 + }, + { + "epoch": 1.55, + "learning_rate": 1.7379394385872116e-05, + "loss": 1.0202, + "step": 21008 + }, + { + "epoch": 1.55, + "learning_rate": 1.7379125194732786e-05, + "loss": 1.0318, + "step": 21009 + }, + { + "epoch": 1.55, + "learning_rate": 1.7378855991853373e-05, + "loss": 1.1282, + "step": 21010 + }, + { + "epoch": 1.55, + "learning_rate": 1.737858677723431e-05, + "loss": 0.8677, + "step": 21011 + }, + { + "epoch": 1.55, + "learning_rate": 1.7378317550876018e-05, + "loss": 1.0727, + "step": 21012 + }, + { + "epoch": 1.55, + "learning_rate": 1.737804831277893e-05, + "loss": 1.0486, + "step": 21013 + }, + { + "epoch": 1.55, + "learning_rate": 1.7377779062943473e-05, + "loss": 0.9565, + "step": 21014 + }, + { + "epoch": 1.55, + "learning_rate": 1.737750980137007e-05, + "loss": 1.033, + "step": 21015 + }, + { + "epoch": 1.55, + "learning_rate": 1.737724052805916e-05, + "loss": 1.0497, + "step": 21016 + }, + { + "epoch": 1.55, + "learning_rate": 1.7376971243011165e-05, + "loss": 1.0361, + "step": 21017 + }, + { + "epoch": 1.55, + "learning_rate": 1.7376701946226514e-05, + "loss": 1.0325, + "step": 21018 + }, + { + "epoch": 1.55, + "learning_rate": 1.7376432637705637e-05, + "loss": 0.9554, + "step": 21019 + }, + { + "epoch": 1.55, + "learning_rate": 1.737616331744896e-05, + "loss": 0.9633, + "step": 21020 + }, + { + "epoch": 1.55, + "learning_rate": 1.737589398545691e-05, + "loss": 1.011, + "step": 21021 + }, + { + "epoch": 1.55, + "learning_rate": 1.7375624641729917e-05, + "loss": 0.9836, + "step": 21022 + }, + { + "epoch": 1.55, + "learning_rate": 1.7375355286268415e-05, + "loss": 1.1601, + "step": 21023 + }, + { + "epoch": 1.55, + "learning_rate": 1.737508591907283e-05, + "loss": 0.9621, + "step": 21024 + }, + { + "epoch": 1.55, + "learning_rate": 1.7374816540143585e-05, + "loss": 0.9627, + "step": 21025 + }, + { + "epoch": 1.55, + "learning_rate": 1.7374547149481115e-05, + "loss": 1.0277, + "step": 21026 + }, + { + "epoch": 1.55, + "learning_rate": 1.7374277747085842e-05, + "loss": 1.0909, + "step": 21027 + }, + { + "epoch": 1.55, + "learning_rate": 1.73740083329582e-05, + "loss": 0.8719, + "step": 21028 + }, + { + "epoch": 1.55, + "learning_rate": 1.737373890709862e-05, + "loss": 0.961, + "step": 21029 + }, + { + "epoch": 1.55, + "learning_rate": 1.737346946950752e-05, + "loss": 0.9577, + "step": 21030 + }, + { + "epoch": 1.55, + "learning_rate": 1.7373200020185342e-05, + "loss": 1.1088, + "step": 21031 + }, + { + "epoch": 1.55, + "learning_rate": 1.737293055913251e-05, + "loss": 1.0018, + "step": 21032 + }, + { + "epoch": 1.55, + "learning_rate": 1.7372661086349446e-05, + "loss": 1.0313, + "step": 21033 + }, + { + "epoch": 1.55, + "learning_rate": 1.7372391601836582e-05, + "loss": 0.9852, + "step": 21034 + }, + { + "epoch": 1.55, + "learning_rate": 1.7372122105594352e-05, + "loss": 1.1214, + "step": 21035 + }, + { + "epoch": 1.55, + "learning_rate": 1.7371852597623182e-05, + "loss": 1.0157, + "step": 21036 + }, + { + "epoch": 1.55, + "learning_rate": 1.7371583077923496e-05, + "loss": 1.0627, + "step": 21037 + }, + { + "epoch": 1.55, + "learning_rate": 1.737131354649573e-05, + "loss": 1.0427, + "step": 21038 + }, + { + "epoch": 1.55, + "learning_rate": 1.737104400334031e-05, + "loss": 0.9937, + "step": 21039 + }, + { + "epoch": 1.55, + "learning_rate": 1.7370774448457662e-05, + "loss": 0.9812, + "step": 21040 + }, + { + "epoch": 1.55, + "learning_rate": 1.737050488184822e-05, + "loss": 0.9726, + "step": 21041 + }, + { + "epoch": 1.55, + "learning_rate": 1.7370235303512407e-05, + "loss": 1.0846, + "step": 21042 + }, + { + "epoch": 1.55, + "learning_rate": 1.736996571345066e-05, + "loss": 0.9717, + "step": 21043 + }, + { + "epoch": 1.55, + "learning_rate": 1.7369696111663398e-05, + "loss": 0.9607, + "step": 21044 + }, + { + "epoch": 1.56, + "learning_rate": 1.736942649815106e-05, + "loss": 1.0069, + "step": 21045 + }, + { + "epoch": 1.56, + "learning_rate": 1.7369156872914062e-05, + "loss": 0.9707, + "step": 21046 + }, + { + "epoch": 1.56, + "learning_rate": 1.736888723595285e-05, + "loss": 0.92, + "step": 21047 + }, + { + "epoch": 1.56, + "learning_rate": 1.7368617587267836e-05, + "loss": 0.9758, + "step": 21048 + }, + { + "epoch": 1.56, + "learning_rate": 1.736834792685946e-05, + "loss": 0.906, + "step": 21049 + }, + { + "epoch": 1.56, + "learning_rate": 1.7368078254728148e-05, + "loss": 1.0532, + "step": 21050 + }, + { + "epoch": 1.56, + "learning_rate": 1.7367808570874333e-05, + "loss": 1.0327, + "step": 21051 + }, + { + "epoch": 1.56, + "learning_rate": 1.7367538875298433e-05, + "loss": 0.9612, + "step": 21052 + }, + { + "epoch": 1.56, + "learning_rate": 1.7367269168000888e-05, + "loss": 1.0361, + "step": 21053 + }, + { + "epoch": 1.56, + "learning_rate": 1.736699944898212e-05, + "loss": 1.035, + "step": 21054 + }, + { + "epoch": 1.56, + "learning_rate": 1.7366729718242567e-05, + "loss": 1.0041, + "step": 21055 + }, + { + "epoch": 1.56, + "learning_rate": 1.7366459975782648e-05, + "loss": 1.04, + "step": 21056 + }, + { + "epoch": 1.56, + "learning_rate": 1.7366190221602795e-05, + "loss": 1.1279, + "step": 21057 + }, + { + "epoch": 1.56, + "learning_rate": 1.7365920455703445e-05, + "loss": 0.9814, + "step": 21058 + }, + { + "epoch": 1.56, + "learning_rate": 1.7365650678085015e-05, + "loss": 1.0801, + "step": 21059 + }, + { + "epoch": 1.56, + "learning_rate": 1.736538088874794e-05, + "loss": 1.0434, + "step": 21060 + }, + { + "epoch": 1.56, + "learning_rate": 1.7365111087692654e-05, + "loss": 1.0767, + "step": 21061 + }, + { + "epoch": 1.56, + "learning_rate": 1.736484127491958e-05, + "loss": 1.0449, + "step": 21062 + }, + { + "epoch": 1.56, + "learning_rate": 1.7364571450429147e-05, + "loss": 1.1212, + "step": 21063 + }, + { + "epoch": 1.56, + "learning_rate": 1.7364301614221786e-05, + "loss": 1.0092, + "step": 21064 + }, + { + "epoch": 1.56, + "learning_rate": 1.736403176629793e-05, + "loss": 1.0803, + "step": 21065 + }, + { + "epoch": 1.56, + "learning_rate": 1.7363761906658003e-05, + "loss": 1.0218, + "step": 21066 + }, + { + "epoch": 1.56, + "learning_rate": 1.7363492035302435e-05, + "loss": 1.0782, + "step": 21067 + }, + { + "epoch": 1.56, + "learning_rate": 1.7363222152231662e-05, + "loss": 1.0079, + "step": 21068 + }, + { + "epoch": 1.56, + "learning_rate": 1.7362952257446098e-05, + "loss": 0.9124, + "step": 21069 + }, + { + "epoch": 1.56, + "learning_rate": 1.736268235094619e-05, + "loss": 0.9812, + "step": 21070 + }, + { + "epoch": 1.56, + "learning_rate": 1.7362412432732356e-05, + "loss": 0.9733, + "step": 21071 + }, + { + "epoch": 1.56, + "learning_rate": 1.736214250280503e-05, + "loss": 1.0729, + "step": 21072 + }, + { + "epoch": 1.56, + "learning_rate": 1.7361872561164644e-05, + "loss": 0.9439, + "step": 21073 + }, + { + "epoch": 1.56, + "learning_rate": 1.736160260781162e-05, + "loss": 1.0225, + "step": 21074 + }, + { + "epoch": 1.56, + "learning_rate": 1.7361332642746395e-05, + "loss": 1.0305, + "step": 21075 + }, + { + "epoch": 1.56, + "learning_rate": 1.7361062665969394e-05, + "loss": 1.1196, + "step": 21076 + }, + { + "epoch": 1.56, + "learning_rate": 1.7360792677481044e-05, + "loss": 1.1181, + "step": 21077 + }, + { + "epoch": 1.56, + "learning_rate": 1.736052267728178e-05, + "loss": 1.1565, + "step": 21078 + }, + { + "epoch": 1.56, + "learning_rate": 1.7360252665372033e-05, + "loss": 1.1162, + "step": 21079 + }, + { + "epoch": 1.56, + "learning_rate": 1.7359982641752223e-05, + "loss": 1.0226, + "step": 21080 + }, + { + "epoch": 1.56, + "learning_rate": 1.735971260642279e-05, + "loss": 1.077, + "step": 21081 + }, + { + "epoch": 1.56, + "learning_rate": 1.735944255938416e-05, + "loss": 0.9658, + "step": 21082 + }, + { + "epoch": 1.56, + "learning_rate": 1.735917250063676e-05, + "loss": 1.0711, + "step": 21083 + }, + { + "epoch": 1.56, + "learning_rate": 1.7358902430181027e-05, + "loss": 1.0091, + "step": 21084 + }, + { + "epoch": 1.56, + "learning_rate": 1.735863234801738e-05, + "loss": 0.9441, + "step": 21085 + }, + { + "epoch": 1.56, + "learning_rate": 1.7358362254146258e-05, + "loss": 1.1092, + "step": 21086 + }, + { + "epoch": 1.56, + "learning_rate": 1.7358092148568085e-05, + "loss": 1.1271, + "step": 21087 + }, + { + "epoch": 1.56, + "learning_rate": 1.735782203128329e-05, + "loss": 1.0878, + "step": 21088 + }, + { + "epoch": 1.56, + "learning_rate": 1.735755190229231e-05, + "loss": 0.9252, + "step": 21089 + }, + { + "epoch": 1.56, + "learning_rate": 1.735728176159557e-05, + "loss": 1.0184, + "step": 21090 + }, + { + "epoch": 1.56, + "learning_rate": 1.7357011609193495e-05, + "loss": 1.0354, + "step": 21091 + }, + { + "epoch": 1.56, + "learning_rate": 1.735674144508652e-05, + "loss": 1.0195, + "step": 21092 + }, + { + "epoch": 1.56, + "learning_rate": 1.735647126927508e-05, + "loss": 0.9936, + "step": 21093 + }, + { + "epoch": 1.56, + "learning_rate": 1.73562010817596e-05, + "loss": 1.0005, + "step": 21094 + }, + { + "epoch": 1.56, + "learning_rate": 1.7355930882540507e-05, + "loss": 1.058, + "step": 21095 + }, + { + "epoch": 1.56, + "learning_rate": 1.7355660671618232e-05, + "loss": 0.9536, + "step": 21096 + }, + { + "epoch": 1.56, + "learning_rate": 1.7355390448993205e-05, + "loss": 1.0007, + "step": 21097 + }, + { + "epoch": 1.56, + "learning_rate": 1.7355120214665858e-05, + "loss": 0.9026, + "step": 21098 + }, + { + "epoch": 1.56, + "learning_rate": 1.7354849968636622e-05, + "loss": 1.0424, + "step": 21099 + }, + { + "epoch": 1.56, + "learning_rate": 1.7354579710905924e-05, + "loss": 1.124, + "step": 21100 + }, + { + "epoch": 1.56, + "learning_rate": 1.7354309441474194e-05, + "loss": 1.0269, + "step": 21101 + }, + { + "epoch": 1.56, + "learning_rate": 1.7354039160341863e-05, + "loss": 0.9127, + "step": 21102 + }, + { + "epoch": 1.56, + "learning_rate": 1.735376886750936e-05, + "loss": 0.9996, + "step": 21103 + }, + { + "epoch": 1.56, + "learning_rate": 1.7353498562977118e-05, + "loss": 1.0631, + "step": 21104 + }, + { + "epoch": 1.56, + "learning_rate": 1.7353228246745565e-05, + "loss": 0.9741, + "step": 21105 + }, + { + "epoch": 1.56, + "learning_rate": 1.735295791881513e-05, + "loss": 1.0743, + "step": 21106 + }, + { + "epoch": 1.56, + "learning_rate": 1.7352687579186244e-05, + "loss": 0.9972, + "step": 21107 + }, + { + "epoch": 1.56, + "learning_rate": 1.735241722785934e-05, + "loss": 0.9661, + "step": 21108 + }, + { + "epoch": 1.56, + "learning_rate": 1.735214686483484e-05, + "loss": 0.9255, + "step": 21109 + }, + { + "epoch": 1.56, + "learning_rate": 1.7351876490113186e-05, + "loss": 1.0045, + "step": 21110 + }, + { + "epoch": 1.56, + "learning_rate": 1.7351606103694798e-05, + "loss": 0.9172, + "step": 21111 + }, + { + "epoch": 1.56, + "learning_rate": 1.735133570558011e-05, + "loss": 1.0738, + "step": 21112 + }, + { + "epoch": 1.56, + "learning_rate": 1.7351065295769555e-05, + "loss": 1.0038, + "step": 21113 + }, + { + "epoch": 1.56, + "learning_rate": 1.7350794874263556e-05, + "loss": 1.0462, + "step": 21114 + }, + { + "epoch": 1.56, + "learning_rate": 1.735052444106255e-05, + "loss": 1.0865, + "step": 21115 + }, + { + "epoch": 1.56, + "learning_rate": 1.7350253996166962e-05, + "loss": 1.1046, + "step": 21116 + }, + { + "epoch": 1.56, + "learning_rate": 1.7349983539577228e-05, + "loss": 1.0313, + "step": 21117 + }, + { + "epoch": 1.56, + "learning_rate": 1.7349713071293774e-05, + "loss": 0.9631, + "step": 21118 + }, + { + "epoch": 1.56, + "learning_rate": 1.7349442591317033e-05, + "loss": 1.0738, + "step": 21119 + }, + { + "epoch": 1.56, + "learning_rate": 1.7349172099647433e-05, + "loss": 0.9984, + "step": 21120 + }, + { + "epoch": 1.56, + "learning_rate": 1.7348901596285406e-05, + "loss": 0.9727, + "step": 21121 + }, + { + "epoch": 1.56, + "learning_rate": 1.734863108123138e-05, + "loss": 0.9825, + "step": 21122 + }, + { + "epoch": 1.56, + "learning_rate": 1.7348360554485787e-05, + "loss": 0.9979, + "step": 21123 + }, + { + "epoch": 1.56, + "learning_rate": 1.734809001604906e-05, + "loss": 1.0731, + "step": 21124 + }, + { + "epoch": 1.56, + "learning_rate": 1.7347819465921625e-05, + "loss": 1.039, + "step": 21125 + }, + { + "epoch": 1.56, + "learning_rate": 1.7347548904103914e-05, + "loss": 1.0112, + "step": 21126 + }, + { + "epoch": 1.56, + "learning_rate": 1.734727833059636e-05, + "loss": 1.0562, + "step": 21127 + }, + { + "epoch": 1.56, + "learning_rate": 1.7347007745399393e-05, + "loss": 0.9657, + "step": 21128 + }, + { + "epoch": 1.56, + "learning_rate": 1.7346737148513437e-05, + "loss": 1.0479, + "step": 21129 + }, + { + "epoch": 1.56, + "learning_rate": 1.7346466539938927e-05, + "loss": 0.9771, + "step": 21130 + }, + { + "epoch": 1.56, + "learning_rate": 1.7346195919676295e-05, + "loss": 1.0161, + "step": 21131 + }, + { + "epoch": 1.56, + "learning_rate": 1.7345925287725973e-05, + "loss": 0.9796, + "step": 21132 + }, + { + "epoch": 1.56, + "learning_rate": 1.7345654644088386e-05, + "loss": 0.9029, + "step": 21133 + }, + { + "epoch": 1.56, + "learning_rate": 1.734538398876397e-05, + "loss": 1.0074, + "step": 21134 + }, + { + "epoch": 1.56, + "learning_rate": 1.7345113321753153e-05, + "loss": 1.0951, + "step": 21135 + }, + { + "epoch": 1.56, + "learning_rate": 1.7344842643056367e-05, + "loss": 1.104, + "step": 21136 + }, + { + "epoch": 1.56, + "learning_rate": 1.7344571952674036e-05, + "loss": 0.9497, + "step": 21137 + }, + { + "epoch": 1.56, + "learning_rate": 1.73443012506066e-05, + "loss": 1.0157, + "step": 21138 + }, + { + "epoch": 1.56, + "learning_rate": 1.7344030536854487e-05, + "loss": 1.0494, + "step": 21139 + }, + { + "epoch": 1.56, + "learning_rate": 1.7343759811418125e-05, + "loss": 0.9289, + "step": 21140 + }, + { + "epoch": 1.56, + "learning_rate": 1.7343489074297943e-05, + "loss": 0.9011, + "step": 21141 + }, + { + "epoch": 1.56, + "learning_rate": 1.7343218325494378e-05, + "loss": 0.9672, + "step": 21142 + }, + { + "epoch": 1.56, + "learning_rate": 1.734294756500786e-05, + "loss": 1.0619, + "step": 21143 + }, + { + "epoch": 1.56, + "learning_rate": 1.7342676792838814e-05, + "loss": 1.1685, + "step": 21144 + }, + { + "epoch": 1.56, + "learning_rate": 1.7342406008987675e-05, + "loss": 0.9491, + "step": 21145 + }, + { + "epoch": 1.56, + "learning_rate": 1.7342135213454875e-05, + "loss": 1.0483, + "step": 21146 + }, + { + "epoch": 1.56, + "learning_rate": 1.734186440624084e-05, + "loss": 1.0182, + "step": 21147 + }, + { + "epoch": 1.56, + "learning_rate": 1.7341593587346006e-05, + "loss": 1.0537, + "step": 21148 + }, + { + "epoch": 1.56, + "learning_rate": 1.7341322756770802e-05, + "loss": 0.8583, + "step": 21149 + }, + { + "epoch": 1.56, + "learning_rate": 1.7341051914515657e-05, + "loss": 0.9845, + "step": 21150 + }, + { + "epoch": 1.56, + "learning_rate": 1.7340781060581006e-05, + "loss": 1.0018, + "step": 21151 + }, + { + "epoch": 1.56, + "learning_rate": 1.734051019496727e-05, + "loss": 0.9939, + "step": 21152 + }, + { + "epoch": 1.56, + "learning_rate": 1.7340239317674896e-05, + "loss": 1.0133, + "step": 21153 + }, + { + "epoch": 1.56, + "learning_rate": 1.7339968428704302e-05, + "loss": 1.0042, + "step": 21154 + }, + { + "epoch": 1.56, + "learning_rate": 1.7339697528055923e-05, + "loss": 1.0574, + "step": 21155 + }, + { + "epoch": 1.56, + "learning_rate": 1.733942661573019e-05, + "loss": 0.8975, + "step": 21156 + }, + { + "epoch": 1.56, + "learning_rate": 1.7339155691727537e-05, + "loss": 0.9658, + "step": 21157 + }, + { + "epoch": 1.56, + "learning_rate": 1.7338884756048392e-05, + "loss": 1.0185, + "step": 21158 + }, + { + "epoch": 1.56, + "learning_rate": 1.7338613808693186e-05, + "loss": 1.0368, + "step": 21159 + }, + { + "epoch": 1.56, + "learning_rate": 1.7338342849662345e-05, + "loss": 0.9151, + "step": 21160 + }, + { + "epoch": 1.56, + "learning_rate": 1.7338071878956315e-05, + "loss": 1.0402, + "step": 21161 + }, + { + "epoch": 1.56, + "learning_rate": 1.7337800896575513e-05, + "loss": 1.0557, + "step": 21162 + }, + { + "epoch": 1.56, + "learning_rate": 1.7337529902520372e-05, + "loss": 0.9917, + "step": 21163 + }, + { + "epoch": 1.56, + "learning_rate": 1.7337258896791328e-05, + "loss": 1.0697, + "step": 21164 + }, + { + "epoch": 1.56, + "learning_rate": 1.733698787938881e-05, + "loss": 0.9908, + "step": 21165 + }, + { + "epoch": 1.56, + "learning_rate": 1.7336716850313252e-05, + "loss": 1.0294, + "step": 21166 + }, + { + "epoch": 1.56, + "learning_rate": 1.733644580956508e-05, + "loss": 1.0712, + "step": 21167 + }, + { + "epoch": 1.56, + "learning_rate": 1.7336174757144727e-05, + "loss": 0.9513, + "step": 21168 + }, + { + "epoch": 1.56, + "learning_rate": 1.7335903693052626e-05, + "loss": 1.0433, + "step": 21169 + }, + { + "epoch": 1.56, + "learning_rate": 1.7335632617289207e-05, + "loss": 0.9659, + "step": 21170 + }, + { + "epoch": 1.56, + "learning_rate": 1.73353615298549e-05, + "loss": 1.1327, + "step": 21171 + }, + { + "epoch": 1.56, + "learning_rate": 1.733509043075014e-05, + "loss": 1.0346, + "step": 21172 + }, + { + "epoch": 1.56, + "learning_rate": 1.7334819319975357e-05, + "loss": 0.9807, + "step": 21173 + }, + { + "epoch": 1.56, + "learning_rate": 1.733454819753098e-05, + "loss": 1.0307, + "step": 21174 + }, + { + "epoch": 1.56, + "learning_rate": 1.733427706341744e-05, + "loss": 0.9643, + "step": 21175 + }, + { + "epoch": 1.56, + "learning_rate": 1.733400591763517e-05, + "loss": 1.0268, + "step": 21176 + }, + { + "epoch": 1.56, + "learning_rate": 1.7333734760184605e-05, + "loss": 0.9986, + "step": 21177 + }, + { + "epoch": 1.56, + "learning_rate": 1.7333463591066172e-05, + "loss": 0.9693, + "step": 21178 + }, + { + "epoch": 1.56, + "learning_rate": 1.73331924102803e-05, + "loss": 1.1046, + "step": 21179 + }, + { + "epoch": 1.56, + "learning_rate": 1.7332921217827427e-05, + "loss": 1.0915, + "step": 21180 + }, + { + "epoch": 1.57, + "learning_rate": 1.7332650013707982e-05, + "loss": 0.9371, + "step": 21181 + }, + { + "epoch": 1.57, + "learning_rate": 1.733237879792239e-05, + "loss": 1.0236, + "step": 21182 + }, + { + "epoch": 1.57, + "learning_rate": 1.733210757047109e-05, + "loss": 1.0463, + "step": 21183 + }, + { + "epoch": 1.57, + "learning_rate": 1.733183633135452e-05, + "loss": 0.9618, + "step": 21184 + }, + { + "epoch": 1.57, + "learning_rate": 1.7331565080573097e-05, + "loss": 1.0772, + "step": 21185 + }, + { + "epoch": 1.57, + "learning_rate": 1.733129381812726e-05, + "loss": 0.8771, + "step": 21186 + }, + { + "epoch": 1.57, + "learning_rate": 1.7331022544017438e-05, + "loss": 1.1266, + "step": 21187 + }, + { + "epoch": 1.57, + "learning_rate": 1.7330751258244065e-05, + "loss": 1.0979, + "step": 21188 + }, + { + "epoch": 1.57, + "learning_rate": 1.733047996080757e-05, + "loss": 1.0015, + "step": 21189 + }, + { + "epoch": 1.57, + "learning_rate": 1.733020865170839e-05, + "loss": 1.046, + "step": 21190 + }, + { + "epoch": 1.57, + "learning_rate": 1.732993733094695e-05, + "loss": 1.04, + "step": 21191 + }, + { + "epoch": 1.57, + "learning_rate": 1.7329665998523684e-05, + "loss": 0.981, + "step": 21192 + }, + { + "epoch": 1.57, + "learning_rate": 1.7329394654439028e-05, + "loss": 0.9356, + "step": 21193 + }, + { + "epoch": 1.57, + "learning_rate": 1.7329123298693405e-05, + "loss": 1.052, + "step": 21194 + }, + { + "epoch": 1.57, + "learning_rate": 1.7328851931287252e-05, + "loss": 1.0568, + "step": 21195 + }, + { + "epoch": 1.57, + "learning_rate": 1.7328580552221006e-05, + "loss": 1.1008, + "step": 21196 + }, + { + "epoch": 1.57, + "learning_rate": 1.7328309161495088e-05, + "loss": 0.9588, + "step": 21197 + }, + { + "epoch": 1.57, + "learning_rate": 1.7328037759109937e-05, + "loss": 0.9596, + "step": 21198 + }, + { + "epoch": 1.57, + "learning_rate": 1.7327766345065984e-05, + "loss": 1.0016, + "step": 21199 + }, + { + "epoch": 1.57, + "learning_rate": 1.7327494919363655e-05, + "loss": 1.036, + "step": 21200 + }, + { + "epoch": 1.57, + "learning_rate": 1.7327223482003393e-05, + "loss": 1.0892, + "step": 21201 + }, + { + "epoch": 1.57, + "learning_rate": 1.732695203298562e-05, + "loss": 0.9331, + "step": 21202 + }, + { + "epoch": 1.57, + "learning_rate": 1.7326680572310767e-05, + "loss": 0.8878, + "step": 21203 + }, + { + "epoch": 1.57, + "learning_rate": 1.7326409099979275e-05, + "loss": 0.9798, + "step": 21204 + }, + { + "epoch": 1.57, + "learning_rate": 1.732613761599157e-05, + "loss": 0.9628, + "step": 21205 + }, + { + "epoch": 1.57, + "learning_rate": 1.7325866120348084e-05, + "loss": 1.0675, + "step": 21206 + }, + { + "epoch": 1.57, + "learning_rate": 1.7325594613049253e-05, + "loss": 0.9219, + "step": 21207 + }, + { + "epoch": 1.57, + "learning_rate": 1.7325323094095502e-05, + "loss": 1.0946, + "step": 21208 + }, + { + "epoch": 1.57, + "learning_rate": 1.7325051563487266e-05, + "loss": 1.0029, + "step": 21209 + }, + { + "epoch": 1.57, + "learning_rate": 1.732478002122498e-05, + "loss": 0.9652, + "step": 21210 + }, + { + "epoch": 1.57, + "learning_rate": 1.7324508467309075e-05, + "loss": 0.9474, + "step": 21211 + }, + { + "epoch": 1.57, + "learning_rate": 1.732423690173998e-05, + "loss": 0.9056, + "step": 21212 + }, + { + "epoch": 1.57, + "learning_rate": 1.732396532451813e-05, + "loss": 1.0365, + "step": 21213 + }, + { + "epoch": 1.57, + "learning_rate": 1.7323693735643956e-05, + "loss": 1.0691, + "step": 21214 + }, + { + "epoch": 1.57, + "learning_rate": 1.732342213511789e-05, + "loss": 1.1007, + "step": 21215 + }, + { + "epoch": 1.57, + "learning_rate": 1.732315052294036e-05, + "loss": 1.0239, + "step": 21216 + }, + { + "epoch": 1.57, + "learning_rate": 1.732287889911181e-05, + "loss": 0.9562, + "step": 21217 + }, + { + "epoch": 1.57, + "learning_rate": 1.7322607263632657e-05, + "loss": 1.0354, + "step": 21218 + }, + { + "epoch": 1.57, + "learning_rate": 1.7322335616503347e-05, + "loss": 0.9865, + "step": 21219 + }, + { + "epoch": 1.57, + "learning_rate": 1.73220639577243e-05, + "loss": 1.0649, + "step": 21220 + }, + { + "epoch": 1.57, + "learning_rate": 1.7321792287295956e-05, + "loss": 1.1316, + "step": 21221 + }, + { + "epoch": 1.57, + "learning_rate": 1.732152060521875e-05, + "loss": 0.9686, + "step": 21222 + }, + { + "epoch": 1.57, + "learning_rate": 1.7321248911493104e-05, + "loss": 1.0571, + "step": 21223 + }, + { + "epoch": 1.57, + "learning_rate": 1.7320977206119457e-05, + "loss": 0.9331, + "step": 21224 + }, + { + "epoch": 1.57, + "learning_rate": 1.7320705489098242e-05, + "loss": 0.976, + "step": 21225 + }, + { + "epoch": 1.57, + "learning_rate": 1.7320433760429884e-05, + "loss": 1.099, + "step": 21226 + }, + { + "epoch": 1.57, + "learning_rate": 1.7320162020114827e-05, + "loss": 1.0801, + "step": 21227 + }, + { + "epoch": 1.57, + "learning_rate": 1.7319890268153493e-05, + "loss": 1.0269, + "step": 21228 + }, + { + "epoch": 1.57, + "learning_rate": 1.7319618504546317e-05, + "loss": 0.9583, + "step": 21229 + }, + { + "epoch": 1.57, + "learning_rate": 1.731934672929374e-05, + "loss": 1.0061, + "step": 21230 + }, + { + "epoch": 1.57, + "learning_rate": 1.731907494239618e-05, + "loss": 0.9585, + "step": 21231 + }, + { + "epoch": 1.57, + "learning_rate": 1.7318803143854078e-05, + "loss": 1.0561, + "step": 21232 + }, + { + "epoch": 1.57, + "learning_rate": 1.7318531333667865e-05, + "loss": 0.9958, + "step": 21233 + }, + { + "epoch": 1.57, + "learning_rate": 1.7318259511837972e-05, + "loss": 0.907, + "step": 21234 + }, + { + "epoch": 1.57, + "learning_rate": 1.731798767836484e-05, + "loss": 0.9325, + "step": 21235 + }, + { + "epoch": 1.57, + "learning_rate": 1.731771583324889e-05, + "loss": 1.044, + "step": 21236 + }, + { + "epoch": 1.57, + "learning_rate": 1.7317443976490555e-05, + "loss": 0.992, + "step": 21237 + }, + { + "epoch": 1.57, + "learning_rate": 1.7317172108090275e-05, + "loss": 0.9362, + "step": 21238 + }, + { + "epoch": 1.57, + "learning_rate": 1.7316900228048473e-05, + "loss": 1.0302, + "step": 21239 + }, + { + "epoch": 1.57, + "learning_rate": 1.7316628336365595e-05, + "loss": 1.038, + "step": 21240 + }, + { + "epoch": 1.57, + "learning_rate": 1.7316356433042063e-05, + "loss": 0.9618, + "step": 21241 + }, + { + "epoch": 1.57, + "learning_rate": 1.731608451807831e-05, + "loss": 0.9761, + "step": 21242 + }, + { + "epoch": 1.57, + "learning_rate": 1.7315812591474776e-05, + "loss": 1.0081, + "step": 21243 + }, + { + "epoch": 1.57, + "learning_rate": 1.7315540653231887e-05, + "loss": 1.0488, + "step": 21244 + }, + { + "epoch": 1.57, + "learning_rate": 1.731526870335008e-05, + "loss": 1.0158, + "step": 21245 + }, + { + "epoch": 1.57, + "learning_rate": 1.7314996741829783e-05, + "loss": 1.0046, + "step": 21246 + }, + { + "epoch": 1.57, + "learning_rate": 1.731472476867143e-05, + "loss": 1.0151, + "step": 21247 + }, + { + "epoch": 1.57, + "learning_rate": 1.7314452783875454e-05, + "loss": 0.9339, + "step": 21248 + }, + { + "epoch": 1.57, + "learning_rate": 1.7314180787442292e-05, + "loss": 1.0452, + "step": 21249 + }, + { + "epoch": 1.57, + "learning_rate": 1.731390877937237e-05, + "loss": 0.9796, + "step": 21250 + }, + { + "epoch": 1.57, + "learning_rate": 1.7313636759666125e-05, + "loss": 1.0563, + "step": 21251 + }, + { + "epoch": 1.57, + "learning_rate": 1.7313364728323984e-05, + "loss": 0.8582, + "step": 21252 + }, + { + "epoch": 1.57, + "learning_rate": 1.7313092685346392e-05, + "loss": 1.024, + "step": 21253 + }, + { + "epoch": 1.57, + "learning_rate": 1.731282063073377e-05, + "loss": 1.0568, + "step": 21254 + }, + { + "epoch": 1.57, + "learning_rate": 1.7312548564486556e-05, + "loss": 1.1015, + "step": 21255 + }, + { + "epoch": 1.57, + "learning_rate": 1.731227648660518e-05, + "loss": 1.0059, + "step": 21256 + }, + { + "epoch": 1.57, + "learning_rate": 1.7312004397090076e-05, + "loss": 1.0011, + "step": 21257 + }, + { + "epoch": 1.57, + "learning_rate": 1.731173229594168e-05, + "loss": 0.963, + "step": 21258 + }, + { + "epoch": 1.57, + "learning_rate": 1.731146018316042e-05, + "loss": 1.0738, + "step": 21259 + }, + { + "epoch": 1.57, + "learning_rate": 1.7311188058746738e-05, + "loss": 1.0414, + "step": 21260 + }, + { + "epoch": 1.57, + "learning_rate": 1.731091592270105e-05, + "loss": 1.0892, + "step": 21261 + }, + { + "epoch": 1.57, + "learning_rate": 1.731064377502381e-05, + "loss": 1.1412, + "step": 21262 + }, + { + "epoch": 1.57, + "learning_rate": 1.7310371615715436e-05, + "loss": 1.0064, + "step": 21263 + }, + { + "epoch": 1.57, + "learning_rate": 1.7310099444776366e-05, + "loss": 1.0496, + "step": 21264 + }, + { + "epoch": 1.57, + "learning_rate": 1.730982726220703e-05, + "loss": 1.0789, + "step": 21265 + }, + { + "epoch": 1.57, + "learning_rate": 1.7309555068007863e-05, + "loss": 1.0763, + "step": 21266 + }, + { + "epoch": 1.57, + "learning_rate": 1.73092828621793e-05, + "loss": 1.0071, + "step": 21267 + }, + { + "epoch": 1.57, + "learning_rate": 1.7309010644721774e-05, + "loss": 0.9939, + "step": 21268 + }, + { + "epoch": 1.57, + "learning_rate": 1.7308738415635716e-05, + "loss": 0.9073, + "step": 21269 + }, + { + "epoch": 1.57, + "learning_rate": 1.730846617492156e-05, + "loss": 1.1189, + "step": 21270 + }, + { + "epoch": 1.57, + "learning_rate": 1.7308193922579737e-05, + "loss": 0.9632, + "step": 21271 + }, + { + "epoch": 1.57, + "learning_rate": 1.7307921658610683e-05, + "loss": 0.9352, + "step": 21272 + }, + { + "epoch": 1.57, + "learning_rate": 1.7307649383014832e-05, + "loss": 0.9655, + "step": 21273 + }, + { + "epoch": 1.57, + "learning_rate": 1.7307377095792613e-05, + "loss": 1.01, + "step": 21274 + }, + { + "epoch": 1.57, + "learning_rate": 1.730710479694446e-05, + "loss": 1.0931, + "step": 21275 + }, + { + "epoch": 1.57, + "learning_rate": 1.730683248647081e-05, + "loss": 1.022, + "step": 21276 + }, + { + "epoch": 1.57, + "learning_rate": 1.7306560164372095e-05, + "loss": 0.8549, + "step": 21277 + }, + { + "epoch": 1.57, + "learning_rate": 1.7306287830648747e-05, + "loss": 0.9381, + "step": 21278 + }, + { + "epoch": 1.57, + "learning_rate": 1.73060154853012e-05, + "loss": 1.0121, + "step": 21279 + }, + { + "epoch": 1.57, + "learning_rate": 1.7305743128329883e-05, + "loss": 1.0903, + "step": 21280 + }, + { + "epoch": 1.57, + "learning_rate": 1.7305470759735238e-05, + "loss": 1.0254, + "step": 21281 + }, + { + "epoch": 1.57, + "learning_rate": 1.7305198379517692e-05, + "loss": 1.0246, + "step": 21282 + }, + { + "epoch": 1.57, + "learning_rate": 1.730492598767768e-05, + "loss": 0.8859, + "step": 21283 + }, + { + "epoch": 1.57, + "learning_rate": 1.7304653584215634e-05, + "loss": 1.0616, + "step": 21284 + }, + { + "epoch": 1.57, + "learning_rate": 1.7304381169131986e-05, + "loss": 1.0205, + "step": 21285 + }, + { + "epoch": 1.57, + "learning_rate": 1.7304108742427178e-05, + "loss": 0.9628, + "step": 21286 + }, + { + "epoch": 1.57, + "learning_rate": 1.7303836304101637e-05, + "loss": 1.0595, + "step": 21287 + }, + { + "epoch": 1.57, + "learning_rate": 1.7303563854155793e-05, + "loss": 0.9966, + "step": 21288 + }, + { + "epoch": 1.57, + "learning_rate": 1.7303291392590087e-05, + "loss": 0.9638, + "step": 21289 + }, + { + "epoch": 1.57, + "learning_rate": 1.7303018919404946e-05, + "loss": 0.9744, + "step": 21290 + }, + { + "epoch": 1.57, + "learning_rate": 1.7302746434600806e-05, + "loss": 1.0039, + "step": 21291 + }, + { + "epoch": 1.57, + "learning_rate": 1.7302473938178102e-05, + "loss": 0.9747, + "step": 21292 + }, + { + "epoch": 1.57, + "learning_rate": 1.7302201430137267e-05, + "loss": 0.9864, + "step": 21293 + }, + { + "epoch": 1.57, + "learning_rate": 1.7301928910478735e-05, + "loss": 0.9386, + "step": 21294 + }, + { + "epoch": 1.57, + "learning_rate": 1.730165637920294e-05, + "loss": 1.0182, + "step": 21295 + }, + { + "epoch": 1.57, + "learning_rate": 1.730138383631031e-05, + "loss": 1.1085, + "step": 21296 + }, + { + "epoch": 1.57, + "learning_rate": 1.7301111281801285e-05, + "loss": 1.0342, + "step": 21297 + }, + { + "epoch": 1.57, + "learning_rate": 1.7300838715676297e-05, + "loss": 0.987, + "step": 21298 + }, + { + "epoch": 1.57, + "learning_rate": 1.7300566137935777e-05, + "loss": 1.041, + "step": 21299 + }, + { + "epoch": 1.57, + "learning_rate": 1.7300293548580162e-05, + "loss": 1.0662, + "step": 21300 + }, + { + "epoch": 1.57, + "learning_rate": 1.7300020947609885e-05, + "loss": 1.0398, + "step": 21301 + }, + { + "epoch": 1.57, + "learning_rate": 1.7299748335025377e-05, + "loss": 1.0338, + "step": 21302 + }, + { + "epoch": 1.57, + "learning_rate": 1.7299475710827075e-05, + "loss": 1.1156, + "step": 21303 + }, + { + "epoch": 1.57, + "learning_rate": 1.7299203075015413e-05, + "loss": 1.1115, + "step": 21304 + }, + { + "epoch": 1.57, + "learning_rate": 1.7298930427590824e-05, + "loss": 1.0043, + "step": 21305 + }, + { + "epoch": 1.57, + "learning_rate": 1.729865776855374e-05, + "loss": 1.0404, + "step": 21306 + }, + { + "epoch": 1.57, + "learning_rate": 1.7298385097904594e-05, + "loss": 1.0813, + "step": 21307 + }, + { + "epoch": 1.57, + "learning_rate": 1.7298112415643823e-05, + "loss": 0.9563, + "step": 21308 + }, + { + "epoch": 1.57, + "learning_rate": 1.7297839721771857e-05, + "loss": 1.0426, + "step": 21309 + }, + { + "epoch": 1.57, + "learning_rate": 1.7297567016289133e-05, + "loss": 0.9463, + "step": 21310 + }, + { + "epoch": 1.57, + "learning_rate": 1.7297294299196087e-05, + "loss": 0.9836, + "step": 21311 + }, + { + "epoch": 1.57, + "learning_rate": 1.729702157049315e-05, + "loss": 1.0347, + "step": 21312 + }, + { + "epoch": 1.57, + "learning_rate": 1.7296748830180757e-05, + "loss": 1.0278, + "step": 21313 + }, + { + "epoch": 1.57, + "learning_rate": 1.729647607825934e-05, + "loss": 1.0318, + "step": 21314 + }, + { + "epoch": 1.57, + "learning_rate": 1.7296203314729332e-05, + "loss": 0.9558, + "step": 21315 + }, + { + "epoch": 1.58, + "learning_rate": 1.729593053959117e-05, + "loss": 1.0442, + "step": 21316 + }, + { + "epoch": 1.58, + "learning_rate": 1.7295657752845286e-05, + "loss": 1.0633, + "step": 21317 + }, + { + "epoch": 1.58, + "learning_rate": 1.7295384954492118e-05, + "loss": 1.0214, + "step": 21318 + }, + { + "epoch": 1.58, + "learning_rate": 1.7295112144532097e-05, + "loss": 1.0801, + "step": 21319 + }, + { + "epoch": 1.58, + "learning_rate": 1.7294839322965652e-05, + "loss": 0.9834, + "step": 21320 + }, + { + "epoch": 1.58, + "learning_rate": 1.7294566489793226e-05, + "loss": 0.9955, + "step": 21321 + }, + { + "epoch": 1.58, + "learning_rate": 1.7294293645015246e-05, + "loss": 1.0247, + "step": 21322 + }, + { + "epoch": 1.58, + "learning_rate": 1.7294020788632152e-05, + "loss": 0.9414, + "step": 21323 + }, + { + "epoch": 1.58, + "learning_rate": 1.7293747920644375e-05, + "loss": 1.0321, + "step": 21324 + }, + { + "epoch": 1.58, + "learning_rate": 1.7293475041052348e-05, + "loss": 0.9663, + "step": 21325 + }, + { + "epoch": 1.58, + "learning_rate": 1.7293202149856508e-05, + "loss": 1.0396, + "step": 21326 + }, + { + "epoch": 1.58, + "learning_rate": 1.729292924705729e-05, + "loss": 1.1017, + "step": 21327 + }, + { + "epoch": 1.58, + "learning_rate": 1.729265633265512e-05, + "loss": 0.9543, + "step": 21328 + }, + { + "epoch": 1.58, + "learning_rate": 1.7292383406650442e-05, + "loss": 1.0472, + "step": 21329 + }, + { + "epoch": 1.58, + "learning_rate": 1.7292110469043685e-05, + "loss": 0.9767, + "step": 21330 + }, + { + "epoch": 1.58, + "learning_rate": 1.7291837519835284e-05, + "loss": 1.0904, + "step": 21331 + }, + { + "epoch": 1.58, + "learning_rate": 1.7291564559025676e-05, + "loss": 1.0283, + "step": 21332 + }, + { + "epoch": 1.58, + "learning_rate": 1.7291291586615294e-05, + "loss": 0.9712, + "step": 21333 + }, + { + "epoch": 1.58, + "learning_rate": 1.729101860260457e-05, + "loss": 0.9802, + "step": 21334 + }, + { + "epoch": 1.58, + "learning_rate": 1.7290745606993937e-05, + "loss": 1.0901, + "step": 21335 + }, + { + "epoch": 1.58, + "learning_rate": 1.7290472599783833e-05, + "loss": 1.0154, + "step": 21336 + }, + { + "epoch": 1.58, + "learning_rate": 1.7290199580974694e-05, + "loss": 0.9999, + "step": 21337 + }, + { + "epoch": 1.58, + "learning_rate": 1.728992655056695e-05, + "loss": 0.9981, + "step": 21338 + }, + { + "epoch": 1.58, + "learning_rate": 1.728965350856104e-05, + "loss": 0.9712, + "step": 21339 + }, + { + "epoch": 1.58, + "learning_rate": 1.728938045495739e-05, + "loss": 1.0593, + "step": 21340 + }, + { + "epoch": 1.58, + "learning_rate": 1.7289107389756443e-05, + "loss": 1.041, + "step": 21341 + }, + { + "epoch": 1.58, + "learning_rate": 1.7288834312958632e-05, + "loss": 1.1154, + "step": 21342 + }, + { + "epoch": 1.58, + "learning_rate": 1.7288561224564388e-05, + "loss": 1.0278, + "step": 21343 + }, + { + "epoch": 1.58, + "learning_rate": 1.7288288124574148e-05, + "loss": 1.0329, + "step": 21344 + }, + { + "epoch": 1.58, + "learning_rate": 1.7288015012988348e-05, + "loss": 1.0789, + "step": 21345 + }, + { + "epoch": 1.58, + "learning_rate": 1.7287741889807416e-05, + "loss": 0.9286, + "step": 21346 + }, + { + "epoch": 1.58, + "learning_rate": 1.7287468755031792e-05, + "loss": 0.9287, + "step": 21347 + }, + { + "epoch": 1.58, + "learning_rate": 1.7287195608661913e-05, + "loss": 1.009, + "step": 21348 + }, + { + "epoch": 1.58, + "learning_rate": 1.7286922450698207e-05, + "loss": 1.0854, + "step": 21349 + }, + { + "epoch": 1.58, + "learning_rate": 1.7286649281141113e-05, + "loss": 0.9046, + "step": 21350 + }, + { + "epoch": 1.58, + "learning_rate": 1.7286376099991062e-05, + "loss": 1.07, + "step": 21351 + }, + { + "epoch": 1.58, + "learning_rate": 1.728610290724849e-05, + "loss": 0.9849, + "step": 21352 + }, + { + "epoch": 1.58, + "learning_rate": 1.7285829702913838e-05, + "loss": 0.9927, + "step": 21353 + }, + { + "epoch": 1.58, + "learning_rate": 1.728555648698753e-05, + "loss": 0.9326, + "step": 21354 + }, + { + "epoch": 1.58, + "learning_rate": 1.728528325947001e-05, + "loss": 1.0967, + "step": 21355 + }, + { + "epoch": 1.58, + "learning_rate": 1.7285010020361705e-05, + "loss": 1.0567, + "step": 21356 + }, + { + "epoch": 1.58, + "learning_rate": 1.7284736769663056e-05, + "loss": 0.9245, + "step": 21357 + }, + { + "epoch": 1.58, + "learning_rate": 1.7284463507374493e-05, + "loss": 0.9835, + "step": 21358 + }, + { + "epoch": 1.58, + "learning_rate": 1.7284190233496455e-05, + "loss": 1.1149, + "step": 21359 + }, + { + "epoch": 1.58, + "learning_rate": 1.7283916948029374e-05, + "loss": 1.0333, + "step": 21360 + }, + { + "epoch": 1.58, + "learning_rate": 1.7283643650973683e-05, + "loss": 1.0004, + "step": 21361 + }, + { + "epoch": 1.58, + "learning_rate": 1.728337034232982e-05, + "loss": 1.0378, + "step": 21362 + }, + { + "epoch": 1.58, + "learning_rate": 1.728309702209822e-05, + "loss": 1.0277, + "step": 21363 + }, + { + "epoch": 1.58, + "learning_rate": 1.7282823690279315e-05, + "loss": 1.1254, + "step": 21364 + }, + { + "epoch": 1.58, + "learning_rate": 1.7282550346873544e-05, + "loss": 0.9969, + "step": 21365 + }, + { + "epoch": 1.58, + "learning_rate": 1.728227699188134e-05, + "loss": 1.0191, + "step": 21366 + }, + { + "epoch": 1.58, + "learning_rate": 1.7282003625303137e-05, + "loss": 0.993, + "step": 21367 + }, + { + "epoch": 1.58, + "learning_rate": 1.728173024713937e-05, + "loss": 1.0133, + "step": 21368 + }, + { + "epoch": 1.58, + "learning_rate": 1.7281456857390473e-05, + "loss": 0.9666, + "step": 21369 + }, + { + "epoch": 1.58, + "learning_rate": 1.7281183456056884e-05, + "loss": 0.9874, + "step": 21370 + }, + { + "epoch": 1.58, + "learning_rate": 1.7280910043139035e-05, + "loss": 1.0224, + "step": 21371 + }, + { + "epoch": 1.58, + "learning_rate": 1.7280636618637364e-05, + "loss": 1.1858, + "step": 21372 + }, + { + "epoch": 1.58, + "learning_rate": 1.7280363182552304e-05, + "loss": 1.0711, + "step": 21373 + }, + { + "epoch": 1.58, + "learning_rate": 1.7280089734884288e-05, + "loss": 0.9654, + "step": 21374 + }, + { + "epoch": 1.58, + "learning_rate": 1.7279816275633754e-05, + "loss": 0.9945, + "step": 21375 + }, + { + "epoch": 1.58, + "learning_rate": 1.727954280480114e-05, + "loss": 1.0977, + "step": 21376 + }, + { + "epoch": 1.58, + "learning_rate": 1.7279269322386873e-05, + "loss": 1.1048, + "step": 21377 + }, + { + "epoch": 1.58, + "learning_rate": 1.7278995828391397e-05, + "loss": 1.0646, + "step": 21378 + }, + { + "epoch": 1.58, + "learning_rate": 1.7278722322815137e-05, + "loss": 0.9808, + "step": 21379 + }, + { + "epoch": 1.58, + "learning_rate": 1.7278448805658537e-05, + "loss": 1.0045, + "step": 21380 + }, + { + "epoch": 1.58, + "learning_rate": 1.727817527692203e-05, + "loss": 1.159, + "step": 21381 + }, + { + "epoch": 1.58, + "learning_rate": 1.727790173660605e-05, + "loss": 0.937, + "step": 21382 + }, + { + "epoch": 1.58, + "learning_rate": 1.7277628184711033e-05, + "loss": 1.0444, + "step": 21383 + }, + { + "epoch": 1.58, + "learning_rate": 1.7277354621237412e-05, + "loss": 1.0524, + "step": 21384 + }, + { + "epoch": 1.58, + "learning_rate": 1.7277081046185626e-05, + "loss": 1.0346, + "step": 21385 + }, + { + "epoch": 1.58, + "learning_rate": 1.7276807459556105e-05, + "loss": 0.9674, + "step": 21386 + }, + { + "epoch": 1.58, + "learning_rate": 1.727653386134929e-05, + "loss": 1.008, + "step": 21387 + }, + { + "epoch": 1.58, + "learning_rate": 1.7276260251565614e-05, + "loss": 1.0074, + "step": 21388 + }, + { + "epoch": 1.58, + "learning_rate": 1.727598663020551e-05, + "loss": 0.9639, + "step": 21389 + }, + { + "epoch": 1.58, + "learning_rate": 1.7275712997269417e-05, + "loss": 1.0764, + "step": 21390 + }, + { + "epoch": 1.58, + "learning_rate": 1.727543935275777e-05, + "loss": 0.9287, + "step": 21391 + }, + { + "epoch": 1.58, + "learning_rate": 1.7275165696671e-05, + "loss": 0.914, + "step": 21392 + }, + { + "epoch": 1.58, + "learning_rate": 1.7274892029009548e-05, + "loss": 1.0908, + "step": 21393 + }, + { + "epoch": 1.58, + "learning_rate": 1.7274618349773848e-05, + "loss": 1.0133, + "step": 21394 + }, + { + "epoch": 1.58, + "learning_rate": 1.7274344658964327e-05, + "loss": 1.0888, + "step": 21395 + }, + { + "epoch": 1.58, + "learning_rate": 1.7274070956581435e-05, + "loss": 1.0831, + "step": 21396 + }, + { + "epoch": 1.58, + "learning_rate": 1.7273797242625597e-05, + "loss": 1.0559, + "step": 21397 + }, + { + "epoch": 1.58, + "learning_rate": 1.7273523517097252e-05, + "loss": 1.0479, + "step": 21398 + }, + { + "epoch": 1.58, + "learning_rate": 1.7273249779996835e-05, + "loss": 1.0325, + "step": 21399 + }, + { + "epoch": 1.58, + "learning_rate": 1.7272976031324783e-05, + "loss": 1.0472, + "step": 21400 + }, + { + "epoch": 1.58, + "learning_rate": 1.727270227108153e-05, + "loss": 1.0455, + "step": 21401 + }, + { + "epoch": 1.58, + "learning_rate": 1.7272428499267512e-05, + "loss": 1.0197, + "step": 21402 + }, + { + "epoch": 1.58, + "learning_rate": 1.7272154715883165e-05, + "loss": 1.0891, + "step": 21403 + }, + { + "epoch": 1.58, + "learning_rate": 1.727188092092892e-05, + "loss": 0.9015, + "step": 21404 + }, + { + "epoch": 1.58, + "learning_rate": 1.727160711440522e-05, + "loss": 1.0985, + "step": 21405 + }, + { + "epoch": 1.58, + "learning_rate": 1.7271333296312496e-05, + "loss": 1.001, + "step": 21406 + }, + { + "epoch": 1.58, + "learning_rate": 1.727105946665119e-05, + "loss": 1.0559, + "step": 21407 + }, + { + "epoch": 1.58, + "learning_rate": 1.727078562542172e-05, + "loss": 0.9785, + "step": 21408 + }, + { + "epoch": 1.58, + "learning_rate": 1.7270511772624545e-05, + "loss": 1.0427, + "step": 21409 + }, + { + "epoch": 1.58, + "learning_rate": 1.7270237908260088e-05, + "loss": 1.0048, + "step": 21410 + }, + { + "epoch": 1.58, + "learning_rate": 1.726996403232878e-05, + "loss": 0.9814, + "step": 21411 + }, + { + "epoch": 1.58, + "learning_rate": 1.726969014483107e-05, + "loss": 1.0011, + "step": 21412 + }, + { + "epoch": 1.58, + "learning_rate": 1.7269416245767384e-05, + "loss": 1.0409, + "step": 21413 + }, + { + "epoch": 1.58, + "learning_rate": 1.7269142335138166e-05, + "loss": 0.9894, + "step": 21414 + }, + { + "epoch": 1.58, + "learning_rate": 1.726886841294384e-05, + "loss": 1.0791, + "step": 21415 + }, + { + "epoch": 1.58, + "learning_rate": 1.7268594479184854e-05, + "loss": 0.9958, + "step": 21416 + }, + { + "epoch": 1.58, + "learning_rate": 1.726832053386163e-05, + "loss": 1.0596, + "step": 21417 + }, + { + "epoch": 1.58, + "learning_rate": 1.726804657697462e-05, + "loss": 1.1691, + "step": 21418 + }, + { + "epoch": 1.58, + "learning_rate": 1.726777260852425e-05, + "loss": 1.0727, + "step": 21419 + }, + { + "epoch": 1.58, + "learning_rate": 1.7267498628510954e-05, + "loss": 1.0371, + "step": 21420 + }, + { + "epoch": 1.58, + "learning_rate": 1.7267224636935175e-05, + "loss": 1.025, + "step": 21421 + }, + { + "epoch": 1.58, + "learning_rate": 1.7266950633797345e-05, + "loss": 1.0501, + "step": 21422 + }, + { + "epoch": 1.58, + "learning_rate": 1.72666766190979e-05, + "loss": 1.0556, + "step": 21423 + }, + { + "epoch": 1.58, + "learning_rate": 1.7266402592837276e-05, + "loss": 1.0557, + "step": 21424 + }, + { + "epoch": 1.58, + "learning_rate": 1.726612855501591e-05, + "loss": 1.052, + "step": 21425 + }, + { + "epoch": 1.58, + "learning_rate": 1.7265854505634235e-05, + "loss": 1.0884, + "step": 21426 + }, + { + "epoch": 1.58, + "learning_rate": 1.7265580444692695e-05, + "loss": 1.0265, + "step": 21427 + }, + { + "epoch": 1.58, + "learning_rate": 1.7265306372191717e-05, + "loss": 1.0122, + "step": 21428 + }, + { + "epoch": 1.58, + "learning_rate": 1.726503228813174e-05, + "loss": 1.0116, + "step": 21429 + }, + { + "epoch": 1.58, + "learning_rate": 1.7264758192513204e-05, + "loss": 1.0907, + "step": 21430 + }, + { + "epoch": 1.58, + "learning_rate": 1.7264484085336537e-05, + "loss": 0.9658, + "step": 21431 + }, + { + "epoch": 1.58, + "learning_rate": 1.7264209966602183e-05, + "loss": 0.924, + "step": 21432 + }, + { + "epoch": 1.58, + "learning_rate": 1.7263935836310572e-05, + "loss": 1.0179, + "step": 21433 + }, + { + "epoch": 1.58, + "learning_rate": 1.7263661694462146e-05, + "loss": 0.9654, + "step": 21434 + }, + { + "epoch": 1.58, + "learning_rate": 1.7263387541057338e-05, + "loss": 1.1223, + "step": 21435 + }, + { + "epoch": 1.58, + "learning_rate": 1.7263113376096584e-05, + "loss": 1.0369, + "step": 21436 + }, + { + "epoch": 1.58, + "learning_rate": 1.7262839199580316e-05, + "loss": 1.0519, + "step": 21437 + }, + { + "epoch": 1.58, + "learning_rate": 1.726256501150898e-05, + "loss": 0.9266, + "step": 21438 + }, + { + "epoch": 1.58, + "learning_rate": 1.7262290811883006e-05, + "loss": 1.1646, + "step": 21439 + }, + { + "epoch": 1.58, + "learning_rate": 1.7262016600702833e-05, + "loss": 0.903, + "step": 21440 + }, + { + "epoch": 1.58, + "learning_rate": 1.7261742377968892e-05, + "loss": 1.042, + "step": 21441 + }, + { + "epoch": 1.58, + "learning_rate": 1.7261468143681623e-05, + "loss": 1.0505, + "step": 21442 + }, + { + "epoch": 1.58, + "learning_rate": 1.7261193897841467e-05, + "loss": 1.0684, + "step": 21443 + }, + { + "epoch": 1.58, + "learning_rate": 1.726091964044885e-05, + "loss": 1.0332, + "step": 21444 + }, + { + "epoch": 1.58, + "learning_rate": 1.7260645371504214e-05, + "loss": 1.0468, + "step": 21445 + }, + { + "epoch": 1.58, + "learning_rate": 1.7260371091007997e-05, + "loss": 1.0692, + "step": 21446 + }, + { + "epoch": 1.58, + "learning_rate": 1.7260096798960636e-05, + "loss": 1.0055, + "step": 21447 + }, + { + "epoch": 1.58, + "learning_rate": 1.725982249536256e-05, + "loss": 0.9694, + "step": 21448 + }, + { + "epoch": 1.58, + "learning_rate": 1.7259548180214212e-05, + "loss": 1.0485, + "step": 21449 + }, + { + "epoch": 1.58, + "learning_rate": 1.7259273853516028e-05, + "loss": 1.0293, + "step": 21450 + }, + { + "epoch": 1.59, + "learning_rate": 1.725899951526844e-05, + "loss": 0.9896, + "step": 21451 + }, + { + "epoch": 1.59, + "learning_rate": 1.725872516547189e-05, + "loss": 1.106, + "step": 21452 + }, + { + "epoch": 1.59, + "learning_rate": 1.7258450804126813e-05, + "loss": 1.0043, + "step": 21453 + }, + { + "epoch": 1.59, + "learning_rate": 1.7258176431233643e-05, + "loss": 1.0628, + "step": 21454 + }, + { + "epoch": 1.59, + "learning_rate": 1.7257902046792816e-05, + "loss": 0.95, + "step": 21455 + }, + { + "epoch": 1.59, + "learning_rate": 1.7257627650804774e-05, + "loss": 0.9636, + "step": 21456 + }, + { + "epoch": 1.59, + "learning_rate": 1.725735324326995e-05, + "loss": 1.0439, + "step": 21457 + }, + { + "epoch": 1.59, + "learning_rate": 1.725707882418878e-05, + "loss": 1.0187, + "step": 21458 + }, + { + "epoch": 1.59, + "learning_rate": 1.72568043935617e-05, + "loss": 1.1425, + "step": 21459 + }, + { + "epoch": 1.59, + "learning_rate": 1.725652995138915e-05, + "loss": 1.0695, + "step": 21460 + }, + { + "epoch": 1.59, + "learning_rate": 1.725625549767157e-05, + "loss": 1.0669, + "step": 21461 + }, + { + "epoch": 1.59, + "learning_rate": 1.7255981032409385e-05, + "loss": 1.061, + "step": 21462 + }, + { + "epoch": 1.59, + "learning_rate": 1.7255706555603037e-05, + "loss": 0.963, + "step": 21463 + }, + { + "epoch": 1.59, + "learning_rate": 1.7255432067252964e-05, + "loss": 0.9229, + "step": 21464 + }, + { + "epoch": 1.59, + "learning_rate": 1.7255157567359605e-05, + "loss": 1.0182, + "step": 21465 + }, + { + "epoch": 1.59, + "learning_rate": 1.7254883055923392e-05, + "loss": 0.9356, + "step": 21466 + }, + { + "epoch": 1.59, + "learning_rate": 1.7254608532944766e-05, + "loss": 0.9951, + "step": 21467 + }, + { + "epoch": 1.59, + "learning_rate": 1.725433399842416e-05, + "loss": 1.1176, + "step": 21468 + }, + { + "epoch": 1.59, + "learning_rate": 1.725405945236201e-05, + "loss": 1.0808, + "step": 21469 + }, + { + "epoch": 1.59, + "learning_rate": 1.725378489475876e-05, + "loss": 1.0948, + "step": 21470 + }, + { + "epoch": 1.59, + "learning_rate": 1.725351032561484e-05, + "loss": 0.9442, + "step": 21471 + }, + { + "epoch": 1.59, + "learning_rate": 1.7253235744930686e-05, + "loss": 0.9807, + "step": 21472 + }, + { + "epoch": 1.59, + "learning_rate": 1.725296115270674e-05, + "loss": 0.9387, + "step": 21473 + }, + { + "epoch": 1.59, + "learning_rate": 1.725268654894344e-05, + "loss": 0.9506, + "step": 21474 + }, + { + "epoch": 1.59, + "learning_rate": 1.7252411933641217e-05, + "loss": 1.0452, + "step": 21475 + }, + { + "epoch": 1.59, + "learning_rate": 1.725213730680051e-05, + "loss": 1.087, + "step": 21476 + }, + { + "epoch": 1.59, + "learning_rate": 1.7251862668421755e-05, + "loss": 1.0572, + "step": 21477 + }, + { + "epoch": 1.59, + "learning_rate": 1.7251588018505393e-05, + "loss": 1.0528, + "step": 21478 + }, + { + "epoch": 1.59, + "learning_rate": 1.7251313357051856e-05, + "loss": 1.0723, + "step": 21479 + }, + { + "epoch": 1.59, + "learning_rate": 1.7251038684061584e-05, + "loss": 1.017, + "step": 21480 + }, + { + "epoch": 1.59, + "learning_rate": 1.7250763999535012e-05, + "loss": 1.0941, + "step": 21481 + }, + { + "epoch": 1.59, + "learning_rate": 1.725048930347258e-05, + "loss": 1.0864, + "step": 21482 + }, + { + "epoch": 1.59, + "learning_rate": 1.725021459587472e-05, + "loss": 1.0006, + "step": 21483 + }, + { + "epoch": 1.59, + "learning_rate": 1.7249939876741878e-05, + "loss": 1.0374, + "step": 21484 + }, + { + "epoch": 1.59, + "learning_rate": 1.724966514607448e-05, + "loss": 0.9889, + "step": 21485 + }, + { + "epoch": 1.59, + "learning_rate": 1.724939040387297e-05, + "loss": 0.9902, + "step": 21486 + }, + { + "epoch": 1.59, + "learning_rate": 1.7249115650137786e-05, + "loss": 0.9366, + "step": 21487 + }, + { + "epoch": 1.59, + "learning_rate": 1.724884088486936e-05, + "loss": 0.9605, + "step": 21488 + }, + { + "epoch": 1.59, + "learning_rate": 1.7248566108068136e-05, + "loss": 1.0436, + "step": 21489 + }, + { + "epoch": 1.59, + "learning_rate": 1.7248291319734542e-05, + "loss": 0.963, + "step": 21490 + }, + { + "epoch": 1.59, + "learning_rate": 1.724801651986902e-05, + "loss": 0.9497, + "step": 21491 + }, + { + "epoch": 1.59, + "learning_rate": 1.724774170847201e-05, + "loss": 0.9558, + "step": 21492 + }, + { + "epoch": 1.59, + "learning_rate": 1.7247466885543948e-05, + "loss": 0.9814, + "step": 21493 + }, + { + "epoch": 1.59, + "learning_rate": 1.724719205108527e-05, + "loss": 0.9966, + "step": 21494 + }, + { + "epoch": 1.59, + "learning_rate": 1.724691720509641e-05, + "loss": 0.9658, + "step": 21495 + }, + { + "epoch": 1.59, + "learning_rate": 1.7246642347577812e-05, + "loss": 0.9543, + "step": 21496 + }, + { + "epoch": 1.59, + "learning_rate": 1.7246367478529906e-05, + "loss": 1.0449, + "step": 21497 + }, + { + "epoch": 1.59, + "learning_rate": 1.7246092597953133e-05, + "loss": 0.902, + "step": 21498 + }, + { + "epoch": 1.59, + "learning_rate": 1.724581770584793e-05, + "loss": 0.9213, + "step": 21499 + }, + { + "epoch": 1.59, + "learning_rate": 1.724554280221474e-05, + "loss": 1.0153, + "step": 21500 + }, + { + "epoch": 1.59, + "learning_rate": 1.724526788705399e-05, + "loss": 0.9291, + "step": 21501 + }, + { + "epoch": 1.59, + "learning_rate": 1.7244992960366125e-05, + "loss": 0.98, + "step": 21502 + }, + { + "epoch": 1.59, + "learning_rate": 1.724471802215158e-05, + "loss": 0.9324, + "step": 21503 + }, + { + "epoch": 1.59, + "learning_rate": 1.724444307241079e-05, + "loss": 0.9883, + "step": 21504 + }, + { + "epoch": 1.59, + "learning_rate": 1.7244168111144196e-05, + "loss": 1.0121, + "step": 21505 + }, + { + "epoch": 1.59, + "learning_rate": 1.724389313835223e-05, + "loss": 0.9832, + "step": 21506 + }, + { + "epoch": 1.59, + "learning_rate": 1.724361815403534e-05, + "loss": 0.9503, + "step": 21507 + }, + { + "epoch": 1.59, + "learning_rate": 1.7243343158193956e-05, + "loss": 1.0399, + "step": 21508 + }, + { + "epoch": 1.59, + "learning_rate": 1.7243068150828514e-05, + "loss": 1.0366, + "step": 21509 + }, + { + "epoch": 1.59, + "learning_rate": 1.7242793131939457e-05, + "loss": 0.9808, + "step": 21510 + }, + { + "epoch": 1.59, + "learning_rate": 1.7242518101527218e-05, + "loss": 0.929, + "step": 21511 + }, + { + "epoch": 1.59, + "learning_rate": 1.7242243059592237e-05, + "loss": 1.0019, + "step": 21512 + }, + { + "epoch": 1.59, + "learning_rate": 1.7241968006134952e-05, + "loss": 0.9615, + "step": 21513 + }, + { + "epoch": 1.59, + "learning_rate": 1.72416929411558e-05, + "loss": 1.0026, + "step": 21514 + }, + { + "epoch": 1.59, + "learning_rate": 1.7241417864655214e-05, + "loss": 1.0052, + "step": 21515 + }, + { + "epoch": 1.59, + "learning_rate": 1.724114277663364e-05, + "loss": 1.0137, + "step": 21516 + }, + { + "epoch": 1.59, + "learning_rate": 1.724086767709151e-05, + "loss": 0.9484, + "step": 21517 + }, + { + "epoch": 1.59, + "learning_rate": 1.7240592566029263e-05, + "loss": 0.951, + "step": 21518 + }, + { + "epoch": 1.59, + "learning_rate": 1.7240317443447333e-05, + "loss": 1.0607, + "step": 21519 + }, + { + "epoch": 1.59, + "learning_rate": 1.724004230934617e-05, + "loss": 1.0912, + "step": 21520 + }, + { + "epoch": 1.59, + "learning_rate": 1.7239767163726197e-05, + "loss": 0.9641, + "step": 21521 + }, + { + "epoch": 1.59, + "learning_rate": 1.7239492006587858e-05, + "loss": 1.1057, + "step": 21522 + }, + { + "epoch": 1.59, + "learning_rate": 1.723921683793159e-05, + "loss": 1.0623, + "step": 21523 + }, + { + "epoch": 1.59, + "learning_rate": 1.7238941657757835e-05, + "loss": 1.0611, + "step": 21524 + }, + { + "epoch": 1.59, + "learning_rate": 1.7238666466067026e-05, + "loss": 0.9778, + "step": 21525 + }, + { + "epoch": 1.59, + "learning_rate": 1.7238391262859605e-05, + "loss": 1.0464, + "step": 21526 + }, + { + "epoch": 1.59, + "learning_rate": 1.7238116048136002e-05, + "loss": 0.9951, + "step": 21527 + }, + { + "epoch": 1.59, + "learning_rate": 1.723784082189666e-05, + "loss": 0.9936, + "step": 21528 + }, + { + "epoch": 1.59, + "learning_rate": 1.723756558414202e-05, + "loss": 0.9837, + "step": 21529 + }, + { + "epoch": 1.59, + "learning_rate": 1.723729033487252e-05, + "loss": 0.8841, + "step": 21530 + }, + { + "epoch": 1.59, + "learning_rate": 1.723701507408859e-05, + "loss": 0.9977, + "step": 21531 + }, + { + "epoch": 1.59, + "learning_rate": 1.7236739801790672e-05, + "loss": 0.9605, + "step": 21532 + }, + { + "epoch": 1.59, + "learning_rate": 1.7236464517979206e-05, + "loss": 0.9543, + "step": 21533 + }, + { + "epoch": 1.59, + "learning_rate": 1.7236189222654627e-05, + "loss": 0.9632, + "step": 21534 + }, + { + "epoch": 1.59, + "learning_rate": 1.7235913915817373e-05, + "loss": 1.0133, + "step": 21535 + }, + { + "epoch": 1.59, + "learning_rate": 1.7235638597467888e-05, + "loss": 1.0351, + "step": 21536 + }, + { + "epoch": 1.59, + "learning_rate": 1.7235363267606605e-05, + "loss": 0.9776, + "step": 21537 + }, + { + "epoch": 1.59, + "learning_rate": 1.723508792623396e-05, + "loss": 1.0652, + "step": 21538 + }, + { + "epoch": 1.59, + "learning_rate": 1.7234812573350394e-05, + "loss": 1.0631, + "step": 21539 + }, + { + "epoch": 1.59, + "learning_rate": 1.7234537208956347e-05, + "loss": 0.9956, + "step": 21540 + }, + { + "epoch": 1.59, + "learning_rate": 1.7234261833052254e-05, + "loss": 0.9342, + "step": 21541 + }, + { + "epoch": 1.59, + "learning_rate": 1.723398644563855e-05, + "loss": 1.0036, + "step": 21542 + }, + { + "epoch": 1.59, + "learning_rate": 1.723371104671568e-05, + "loss": 1.0172, + "step": 21543 + }, + { + "epoch": 1.59, + "learning_rate": 1.723343563628408e-05, + "loss": 1.0126, + "step": 21544 + }, + { + "epoch": 1.59, + "learning_rate": 1.7233160214344188e-05, + "loss": 1.0974, + "step": 21545 + }, + { + "epoch": 1.59, + "learning_rate": 1.723288478089644e-05, + "loss": 0.9618, + "step": 21546 + }, + { + "epoch": 1.59, + "learning_rate": 1.7232609335941276e-05, + "loss": 1.0501, + "step": 21547 + }, + { + "epoch": 1.59, + "learning_rate": 1.7232333879479135e-05, + "loss": 1.041, + "step": 21548 + }, + { + "epoch": 1.59, + "learning_rate": 1.7232058411510455e-05, + "loss": 0.9655, + "step": 21549 + }, + { + "epoch": 1.59, + "learning_rate": 1.7231782932035672e-05, + "loss": 0.987, + "step": 21550 + }, + { + "epoch": 1.59, + "learning_rate": 1.7231507441055228e-05, + "loss": 1.0551, + "step": 21551 + }, + { + "epoch": 1.59, + "learning_rate": 1.7231231938569555e-05, + "loss": 0.9805, + "step": 21552 + }, + { + "epoch": 1.59, + "learning_rate": 1.7230956424579098e-05, + "loss": 1.085, + "step": 21553 + }, + { + "epoch": 1.59, + "learning_rate": 1.7230680899084294e-05, + "loss": 0.8987, + "step": 21554 + }, + { + "epoch": 1.59, + "learning_rate": 1.7230405362085576e-05, + "loss": 1.0239, + "step": 21555 + }, + { + "epoch": 1.59, + "learning_rate": 1.7230129813583388e-05, + "loss": 1.0565, + "step": 21556 + }, + { + "epoch": 1.59, + "learning_rate": 1.7229854253578168e-05, + "loss": 0.9994, + "step": 21557 + }, + { + "epoch": 1.59, + "learning_rate": 1.7229578682070355e-05, + "loss": 1.0566, + "step": 21558 + }, + { + "epoch": 1.59, + "learning_rate": 1.7229303099060384e-05, + "loss": 1.0361, + "step": 21559 + }, + { + "epoch": 1.59, + "learning_rate": 1.7229027504548695e-05, + "loss": 0.9864, + "step": 21560 + }, + { + "epoch": 1.59, + "learning_rate": 1.7228751898535728e-05, + "loss": 0.9581, + "step": 21561 + }, + { + "epoch": 1.59, + "learning_rate": 1.7228476281021917e-05, + "loss": 0.9228, + "step": 21562 + }, + { + "epoch": 1.59, + "learning_rate": 1.7228200652007706e-05, + "loss": 1.0141, + "step": 21563 + }, + { + "epoch": 1.59, + "learning_rate": 1.7227925011493532e-05, + "loss": 0.969, + "step": 21564 + }, + { + "epoch": 1.59, + "learning_rate": 1.722764935947983e-05, + "loss": 1.0209, + "step": 21565 + }, + { + "epoch": 1.59, + "learning_rate": 1.7227373695967044e-05, + "loss": 0.9031, + "step": 21566 + }, + { + "epoch": 1.59, + "learning_rate": 1.722709802095561e-05, + "loss": 1.0046, + "step": 21567 + }, + { + "epoch": 1.59, + "learning_rate": 1.7226822334445964e-05, + "loss": 1.0793, + "step": 21568 + }, + { + "epoch": 1.59, + "learning_rate": 1.7226546636438548e-05, + "loss": 1.0811, + "step": 21569 + }, + { + "epoch": 1.59, + "learning_rate": 1.7226270926933798e-05, + "loss": 0.9451, + "step": 21570 + }, + { + "epoch": 1.59, + "learning_rate": 1.7225995205932158e-05, + "loss": 0.9748, + "step": 21571 + }, + { + "epoch": 1.59, + "learning_rate": 1.722571947343406e-05, + "loss": 1.0183, + "step": 21572 + }, + { + "epoch": 1.59, + "learning_rate": 1.7225443729439946e-05, + "loss": 1.0517, + "step": 21573 + }, + { + "epoch": 1.59, + "learning_rate": 1.722516797395026e-05, + "loss": 1.0307, + "step": 21574 + }, + { + "epoch": 1.59, + "learning_rate": 1.722489220696543e-05, + "loss": 1.097, + "step": 21575 + }, + { + "epoch": 1.59, + "learning_rate": 1.72246164284859e-05, + "loss": 0.9854, + "step": 21576 + }, + { + "epoch": 1.59, + "learning_rate": 1.7224340638512107e-05, + "loss": 1.0358, + "step": 21577 + }, + { + "epoch": 1.59, + "learning_rate": 1.7224064837044496e-05, + "loss": 0.9309, + "step": 21578 + }, + { + "epoch": 1.59, + "learning_rate": 1.7223789024083494e-05, + "loss": 1.0354, + "step": 21579 + }, + { + "epoch": 1.59, + "learning_rate": 1.7223513199629553e-05, + "loss": 0.9889, + "step": 21580 + }, + { + "epoch": 1.59, + "learning_rate": 1.7223237363683102e-05, + "loss": 1.0557, + "step": 21581 + }, + { + "epoch": 1.59, + "learning_rate": 1.722296151624459e-05, + "loss": 0.9971, + "step": 21582 + }, + { + "epoch": 1.59, + "learning_rate": 1.7222685657314446e-05, + "loss": 1.1225, + "step": 21583 + }, + { + "epoch": 1.59, + "learning_rate": 1.722240978689311e-05, + "loss": 0.9961, + "step": 21584 + }, + { + "epoch": 1.59, + "learning_rate": 1.7222133904981025e-05, + "loss": 0.9562, + "step": 21585 + }, + { + "epoch": 1.59, + "learning_rate": 1.722185801157863e-05, + "loss": 1.0245, + "step": 21586 + }, + { + "epoch": 1.6, + "learning_rate": 1.7221582106686363e-05, + "loss": 0.9922, + "step": 21587 + }, + { + "epoch": 1.6, + "learning_rate": 1.7221306190304658e-05, + "loss": 1.0742, + "step": 21588 + }, + { + "epoch": 1.6, + "learning_rate": 1.7221030262433964e-05, + "loss": 0.9751, + "step": 21589 + }, + { + "epoch": 1.6, + "learning_rate": 1.7220754323074707e-05, + "loss": 1.1463, + "step": 21590 + }, + { + "epoch": 1.6, + "learning_rate": 1.722047837222734e-05, + "loss": 0.9998, + "step": 21591 + }, + { + "epoch": 1.6, + "learning_rate": 1.722020240989229e-05, + "loss": 0.9983, + "step": 21592 + }, + { + "epoch": 1.6, + "learning_rate": 1.7219926436070005e-05, + "loss": 1.0771, + "step": 21593 + }, + { + "epoch": 1.6, + "learning_rate": 1.721965045076092e-05, + "loss": 0.9505, + "step": 21594 + }, + { + "epoch": 1.6, + "learning_rate": 1.721937445396547e-05, + "loss": 0.9413, + "step": 21595 + }, + { + "epoch": 1.6, + "learning_rate": 1.7219098445684102e-05, + "loss": 1.0159, + "step": 21596 + }, + { + "epoch": 1.6, + "learning_rate": 1.7218822425917248e-05, + "loss": 1.0487, + "step": 21597 + }, + { + "epoch": 1.6, + "learning_rate": 1.721854639466536e-05, + "loss": 0.89, + "step": 21598 + }, + { + "epoch": 1.6, + "learning_rate": 1.721827035192886e-05, + "loss": 1.0133, + "step": 21599 + }, + { + "epoch": 1.6, + "learning_rate": 1.7217994297708195e-05, + "loss": 1.0207, + "step": 21600 + }, + { + "epoch": 1.6, + "learning_rate": 1.7217718232003807e-05, + "loss": 0.9736, + "step": 21601 + }, + { + "epoch": 1.6, + "learning_rate": 1.7217442154816132e-05, + "loss": 1.0089, + "step": 21602 + }, + { + "epoch": 1.6, + "learning_rate": 1.721716606614561e-05, + "loss": 0.9854, + "step": 21603 + }, + { + "epoch": 1.6, + "learning_rate": 1.721688996599268e-05, + "loss": 1.0205, + "step": 21604 + }, + { + "epoch": 1.6, + "learning_rate": 1.7216613854357783e-05, + "loss": 1.0626, + "step": 21605 + }, + { + "epoch": 1.6, + "learning_rate": 1.7216337731241356e-05, + "loss": 0.9617, + "step": 21606 + }, + { + "epoch": 1.6, + "learning_rate": 1.7216061596643838e-05, + "loss": 1.006, + "step": 21607 + }, + { + "epoch": 1.6, + "learning_rate": 1.721578545056567e-05, + "loss": 0.9713, + "step": 21608 + }, + { + "epoch": 1.6, + "learning_rate": 1.7215509293007286e-05, + "loss": 1.1665, + "step": 21609 + }, + { + "epoch": 1.6, + "learning_rate": 1.7215233123969134e-05, + "loss": 0.8619, + "step": 21610 + }, + { + "epoch": 1.6, + "learning_rate": 1.721495694345165e-05, + "loss": 1.0741, + "step": 21611 + }, + { + "epoch": 1.6, + "learning_rate": 1.721468075145527e-05, + "loss": 1.0143, + "step": 21612 + }, + { + "epoch": 1.6, + "learning_rate": 1.721440454798044e-05, + "loss": 0.9939, + "step": 21613 + }, + { + "epoch": 1.6, + "learning_rate": 1.7214128333027593e-05, + "loss": 1.0076, + "step": 21614 + }, + { + "epoch": 1.6, + "learning_rate": 1.7213852106597172e-05, + "loss": 0.9488, + "step": 21615 + }, + { + "epoch": 1.6, + "learning_rate": 1.7213575868689614e-05, + "loss": 0.9543, + "step": 21616 + }, + { + "epoch": 1.6, + "learning_rate": 1.7213299619305362e-05, + "loss": 0.9572, + "step": 21617 + }, + { + "epoch": 1.6, + "learning_rate": 1.721302335844485e-05, + "loss": 1.004, + "step": 21618 + }, + { + "epoch": 1.6, + "learning_rate": 1.7212747086108526e-05, + "loss": 0.8913, + "step": 21619 + }, + { + "epoch": 1.6, + "learning_rate": 1.721247080229682e-05, + "loss": 0.9756, + "step": 21620 + }, + { + "epoch": 1.6, + "learning_rate": 1.721219450701018e-05, + "loss": 1.0813, + "step": 21621 + }, + { + "epoch": 1.6, + "learning_rate": 1.7211918200249037e-05, + "loss": 1.0215, + "step": 21622 + }, + { + "epoch": 1.6, + "learning_rate": 1.7211641882013838e-05, + "loss": 1.057, + "step": 21623 + }, + { + "epoch": 1.6, + "learning_rate": 1.721136555230502e-05, + "loss": 0.9922, + "step": 21624 + }, + { + "epoch": 1.6, + "learning_rate": 1.7211089211123025e-05, + "loss": 0.9811, + "step": 21625 + }, + { + "epoch": 1.6, + "learning_rate": 1.721081285846829e-05, + "loss": 1.0047, + "step": 21626 + }, + { + "epoch": 1.6, + "learning_rate": 1.721053649434125e-05, + "loss": 0.9973, + "step": 21627 + }, + { + "epoch": 1.6, + "learning_rate": 1.721026011874235e-05, + "loss": 0.9786, + "step": 21628 + }, + { + "epoch": 1.6, + "learning_rate": 1.7209983731672037e-05, + "loss": 1.0093, + "step": 21629 + }, + { + "epoch": 1.6, + "learning_rate": 1.7209707333130738e-05, + "loss": 0.9786, + "step": 21630 + }, + { + "epoch": 1.6, + "learning_rate": 1.7209430923118895e-05, + "loss": 1.1219, + "step": 21631 + }, + { + "epoch": 1.6, + "learning_rate": 1.7209154501636952e-05, + "loss": 1.0002, + "step": 21632 + }, + { + "epoch": 1.6, + "learning_rate": 1.720887806868535e-05, + "loss": 0.8515, + "step": 21633 + }, + { + "epoch": 1.6, + "learning_rate": 1.7208601624264525e-05, + "loss": 1.0574, + "step": 21634 + }, + { + "epoch": 1.6, + "learning_rate": 1.7208325168374918e-05, + "loss": 1.0405, + "step": 21635 + }, + { + "epoch": 1.6, + "learning_rate": 1.720804870101697e-05, + "loss": 0.9486, + "step": 21636 + }, + { + "epoch": 1.6, + "learning_rate": 1.7207772222191116e-05, + "loss": 1.0781, + "step": 21637 + }, + { + "epoch": 1.6, + "learning_rate": 1.7207495731897802e-05, + "loss": 0.97, + "step": 21638 + }, + { + "epoch": 1.6, + "learning_rate": 1.7207219230137465e-05, + "loss": 1.0961, + "step": 21639 + }, + { + "epoch": 1.6, + "learning_rate": 1.7206942716910548e-05, + "loss": 1.004, + "step": 21640 + }, + { + "epoch": 1.6, + "learning_rate": 1.7206666192217484e-05, + "loss": 1.1012, + "step": 21641 + }, + { + "epoch": 1.6, + "learning_rate": 1.720638965605872e-05, + "loss": 0.907, + "step": 21642 + }, + { + "epoch": 1.6, + "learning_rate": 1.720611310843469e-05, + "loss": 1.0111, + "step": 21643 + }, + { + "epoch": 1.6, + "learning_rate": 1.720583654934584e-05, + "loss": 1.0118, + "step": 21644 + }, + { + "epoch": 1.6, + "learning_rate": 1.7205559978792606e-05, + "loss": 0.9613, + "step": 21645 + }, + { + "epoch": 1.6, + "learning_rate": 1.720528339677543e-05, + "loss": 1.12, + "step": 21646 + }, + { + "epoch": 1.6, + "learning_rate": 1.720500680329475e-05, + "loss": 0.9733, + "step": 21647 + }, + { + "epoch": 1.6, + "learning_rate": 1.7204730198351006e-05, + "loss": 0.965, + "step": 21648 + }, + { + "epoch": 1.6, + "learning_rate": 1.720445358194464e-05, + "loss": 0.9481, + "step": 21649 + }, + { + "epoch": 1.6, + "learning_rate": 1.7204176954076094e-05, + "loss": 0.8997, + "step": 21650 + }, + { + "epoch": 1.6, + "learning_rate": 1.7203900314745805e-05, + "loss": 0.9498, + "step": 21651 + }, + { + "epoch": 1.6, + "learning_rate": 1.7203623663954215e-05, + "loss": 1.0332, + "step": 21652 + }, + { + "epoch": 1.6, + "learning_rate": 1.720334700170176e-05, + "loss": 1.0374, + "step": 21653 + }, + { + "epoch": 1.6, + "learning_rate": 1.7203070327988883e-05, + "loss": 0.9544, + "step": 21654 + }, + { + "epoch": 1.6, + "learning_rate": 1.7202793642816026e-05, + "loss": 1.062, + "step": 21655 + }, + { + "epoch": 1.6, + "learning_rate": 1.7202516946183626e-05, + "loss": 0.9711, + "step": 21656 + }, + { + "epoch": 1.6, + "learning_rate": 1.7202240238092124e-05, + "loss": 0.971, + "step": 21657 + }, + { + "epoch": 1.6, + "learning_rate": 1.7201963518541964e-05, + "loss": 1.0599, + "step": 21658 + }, + { + "epoch": 1.6, + "learning_rate": 1.720168678753358e-05, + "loss": 1.0116, + "step": 21659 + }, + { + "epoch": 1.6, + "learning_rate": 1.7201410045067416e-05, + "loss": 1.0207, + "step": 21660 + }, + { + "epoch": 1.6, + "learning_rate": 1.720113329114391e-05, + "loss": 1.0938, + "step": 21661 + }, + { + "epoch": 1.6, + "learning_rate": 1.7200856525763504e-05, + "loss": 0.9787, + "step": 21662 + }, + { + "epoch": 1.6, + "learning_rate": 1.7200579748926642e-05, + "loss": 0.9531, + "step": 21663 + }, + { + "epoch": 1.6, + "learning_rate": 1.7200302960633758e-05, + "loss": 1.0031, + "step": 21664 + }, + { + "epoch": 1.6, + "learning_rate": 1.7200026160885298e-05, + "loss": 1.0, + "step": 21665 + }, + { + "epoch": 1.6, + "learning_rate": 1.7199749349681694e-05, + "loss": 1.0339, + "step": 21666 + }, + { + "epoch": 1.6, + "learning_rate": 1.7199472527023395e-05, + "loss": 0.9685, + "step": 21667 + }, + { + "epoch": 1.6, + "learning_rate": 1.719919569291084e-05, + "loss": 1.0539, + "step": 21668 + }, + { + "epoch": 1.6, + "learning_rate": 1.7198918847344467e-05, + "loss": 0.9142, + "step": 21669 + }, + { + "epoch": 1.6, + "learning_rate": 1.7198641990324715e-05, + "loss": 1.0716, + "step": 21670 + }, + { + "epoch": 1.6, + "learning_rate": 1.7198365121852027e-05, + "loss": 1.0025, + "step": 21671 + }, + { + "epoch": 1.6, + "learning_rate": 1.7198088241926844e-05, + "loss": 0.9885, + "step": 21672 + }, + { + "epoch": 1.6, + "learning_rate": 1.7197811350549603e-05, + "loss": 0.9333, + "step": 21673 + }, + { + "epoch": 1.6, + "learning_rate": 1.719753444772075e-05, + "loss": 1.0244, + "step": 21674 + }, + { + "epoch": 1.6, + "learning_rate": 1.719725753344072e-05, + "loss": 0.9854, + "step": 21675 + }, + { + "epoch": 1.6, + "learning_rate": 1.719698060770996e-05, + "loss": 0.9609, + "step": 21676 + }, + { + "epoch": 1.6, + "learning_rate": 1.7196703670528905e-05, + "loss": 0.9284, + "step": 21677 + }, + { + "epoch": 1.6, + "learning_rate": 1.7196426721898e-05, + "loss": 0.9288, + "step": 21678 + }, + { + "epoch": 1.6, + "learning_rate": 1.719614976181768e-05, + "loss": 1.1027, + "step": 21679 + }, + { + "epoch": 1.6, + "learning_rate": 1.7195872790288387e-05, + "loss": 1.0817, + "step": 21680 + }, + { + "epoch": 1.6, + "learning_rate": 1.7195595807310566e-05, + "loss": 0.9464, + "step": 21681 + }, + { + "epoch": 1.6, + "learning_rate": 1.7195318812884655e-05, + "loss": 1.0807, + "step": 21682 + }, + { + "epoch": 1.6, + "learning_rate": 1.7195041807011094e-05, + "loss": 1.0448, + "step": 21683 + }, + { + "epoch": 1.6, + "learning_rate": 1.7194764789690323e-05, + "loss": 0.9005, + "step": 21684 + }, + { + "epoch": 1.6, + "learning_rate": 1.7194487760922786e-05, + "loss": 0.9844, + "step": 21685 + }, + { + "epoch": 1.6, + "learning_rate": 1.7194210720708923e-05, + "loss": 1.0742, + "step": 21686 + }, + { + "epoch": 1.6, + "learning_rate": 1.7193933669049172e-05, + "loss": 1.0488, + "step": 21687 + }, + { + "epoch": 1.6, + "learning_rate": 1.7193656605943976e-05, + "loss": 0.9723, + "step": 21688 + }, + { + "epoch": 1.6, + "learning_rate": 1.7193379531393776e-05, + "loss": 1.129, + "step": 21689 + }, + { + "epoch": 1.6, + "learning_rate": 1.719310244539901e-05, + "loss": 1.0642, + "step": 21690 + }, + { + "epoch": 1.6, + "learning_rate": 1.7192825347960123e-05, + "loss": 1.0445, + "step": 21691 + }, + { + "epoch": 1.6, + "learning_rate": 1.7192548239077553e-05, + "loss": 1.0118, + "step": 21692 + }, + { + "epoch": 1.6, + "learning_rate": 1.719227111875174e-05, + "loss": 1.1104, + "step": 21693 + }, + { + "epoch": 1.6, + "learning_rate": 1.7191993986983127e-05, + "loss": 1.0906, + "step": 21694 + }, + { + "epoch": 1.6, + "learning_rate": 1.7191716843772156e-05, + "loss": 1.0631, + "step": 21695 + }, + { + "epoch": 1.6, + "learning_rate": 1.7191439689119266e-05, + "loss": 1.0477, + "step": 21696 + }, + { + "epoch": 1.6, + "learning_rate": 1.71911625230249e-05, + "loss": 1.082, + "step": 21697 + }, + { + "epoch": 1.6, + "learning_rate": 1.7190885345489493e-05, + "loss": 0.9756, + "step": 21698 + }, + { + "epoch": 1.6, + "learning_rate": 1.7190608156513494e-05, + "loss": 0.9381, + "step": 21699 + }, + { + "epoch": 1.6, + "learning_rate": 1.7190330956097342e-05, + "loss": 1.05, + "step": 21700 + }, + { + "epoch": 1.6, + "learning_rate": 1.7190053744241472e-05, + "loss": 0.9309, + "step": 21701 + }, + { + "epoch": 1.6, + "learning_rate": 1.718977652094633e-05, + "loss": 1.0994, + "step": 21702 + }, + { + "epoch": 1.6, + "learning_rate": 1.7189499286212358e-05, + "loss": 1.0246, + "step": 21703 + }, + { + "epoch": 1.6, + "learning_rate": 1.7189222040039998e-05, + "loss": 1.0693, + "step": 21704 + }, + { + "epoch": 1.6, + "learning_rate": 1.718894478242968e-05, + "loss": 0.9419, + "step": 21705 + }, + { + "epoch": 1.6, + "learning_rate": 1.718866751338186e-05, + "loss": 0.8978, + "step": 21706 + }, + { + "epoch": 1.6, + "learning_rate": 1.7188390232896973e-05, + "loss": 1.0574, + "step": 21707 + }, + { + "epoch": 1.6, + "learning_rate": 1.7188112940975462e-05, + "loss": 1.071, + "step": 21708 + }, + { + "epoch": 1.6, + "learning_rate": 1.7187835637617764e-05, + "loss": 0.9781, + "step": 21709 + }, + { + "epoch": 1.6, + "learning_rate": 1.7187558322824317e-05, + "loss": 1.0411, + "step": 21710 + }, + { + "epoch": 1.6, + "learning_rate": 1.7187280996595573e-05, + "loss": 1.0616, + "step": 21711 + }, + { + "epoch": 1.6, + "learning_rate": 1.718700365893197e-05, + "loss": 1.0889, + "step": 21712 + }, + { + "epoch": 1.6, + "learning_rate": 1.7186726309833943e-05, + "loss": 0.9797, + "step": 21713 + }, + { + "epoch": 1.6, + "learning_rate": 1.7186448949301937e-05, + "loss": 0.9897, + "step": 21714 + }, + { + "epoch": 1.6, + "learning_rate": 1.7186171577336395e-05, + "loss": 1.1151, + "step": 21715 + }, + { + "epoch": 1.6, + "learning_rate": 1.7185894193937756e-05, + "loss": 1.1113, + "step": 21716 + }, + { + "epoch": 1.6, + "learning_rate": 1.7185616799106465e-05, + "loss": 0.9348, + "step": 21717 + }, + { + "epoch": 1.6, + "learning_rate": 1.7185339392842957e-05, + "loss": 0.95, + "step": 21718 + }, + { + "epoch": 1.6, + "learning_rate": 1.7185061975147676e-05, + "loss": 0.9798, + "step": 21719 + }, + { + "epoch": 1.6, + "learning_rate": 1.7184784546021068e-05, + "loss": 1.0889, + "step": 21720 + }, + { + "epoch": 1.6, + "learning_rate": 1.7184507105463568e-05, + "loss": 1.0401, + "step": 21721 + }, + { + "epoch": 1.61, + "learning_rate": 1.718422965347562e-05, + "loss": 1.0019, + "step": 21722 + }, + { + "epoch": 1.61, + "learning_rate": 1.7183952190057668e-05, + "loss": 1.1284, + "step": 21723 + }, + { + "epoch": 1.61, + "learning_rate": 1.7183674715210147e-05, + "loss": 1.026, + "step": 21724 + }, + { + "epoch": 1.61, + "learning_rate": 1.7183397228933504e-05, + "loss": 0.9537, + "step": 21725 + }, + { + "epoch": 1.61, + "learning_rate": 1.718311973122818e-05, + "loss": 1.0414, + "step": 21726 + }, + { + "epoch": 1.61, + "learning_rate": 1.7182842222094613e-05, + "loss": 0.9354, + "step": 21727 + }, + { + "epoch": 1.61, + "learning_rate": 1.7182564701533248e-05, + "loss": 1.0837, + "step": 21728 + }, + { + "epoch": 1.61, + "learning_rate": 1.7182287169544528e-05, + "loss": 0.9626, + "step": 21729 + }, + { + "epoch": 1.61, + "learning_rate": 1.7182009626128886e-05, + "loss": 0.9766, + "step": 21730 + }, + { + "epoch": 1.61, + "learning_rate": 1.7181732071286777e-05, + "loss": 1.019, + "step": 21731 + }, + { + "epoch": 1.61, + "learning_rate": 1.718145450501863e-05, + "loss": 0.9692, + "step": 21732 + }, + { + "epoch": 1.61, + "learning_rate": 1.7181176927324893e-05, + "loss": 1.0292, + "step": 21733 + }, + { + "epoch": 1.61, + "learning_rate": 1.7180899338206003e-05, + "loss": 1.1465, + "step": 21734 + }, + { + "epoch": 1.61, + "learning_rate": 1.718062173766241e-05, + "loss": 1.0525, + "step": 21735 + }, + { + "epoch": 1.61, + "learning_rate": 1.7180344125694545e-05, + "loss": 0.9048, + "step": 21736 + }, + { + "epoch": 1.61, + "learning_rate": 1.7180066502302862e-05, + "loss": 1.099, + "step": 21737 + }, + { + "epoch": 1.61, + "learning_rate": 1.7179788867487788e-05, + "loss": 0.9569, + "step": 21738 + }, + { + "epoch": 1.61, + "learning_rate": 1.7179511221249777e-05, + "loss": 0.9984, + "step": 21739 + }, + { + "epoch": 1.61, + "learning_rate": 1.717923356358927e-05, + "loss": 1.069, + "step": 21740 + }, + { + "epoch": 1.61, + "learning_rate": 1.71789558945067e-05, + "loss": 1.0026, + "step": 21741 + }, + { + "epoch": 1.61, + "learning_rate": 1.7178678214002517e-05, + "loss": 0.9758, + "step": 21742 + }, + { + "epoch": 1.61, + "learning_rate": 1.7178400522077154e-05, + "loss": 0.9351, + "step": 21743 + }, + { + "epoch": 1.61, + "learning_rate": 1.7178122818731063e-05, + "loss": 1.0827, + "step": 21744 + }, + { + "epoch": 1.61, + "learning_rate": 1.717784510396468e-05, + "loss": 0.9965, + "step": 21745 + }, + { + "epoch": 1.61, + "learning_rate": 1.7177567377778452e-05, + "loss": 1.0329, + "step": 21746 + }, + { + "epoch": 1.61, + "learning_rate": 1.7177289640172813e-05, + "loss": 0.8909, + "step": 21747 + }, + { + "epoch": 1.61, + "learning_rate": 1.717701189114821e-05, + "loss": 0.9535, + "step": 21748 + }, + { + "epoch": 1.61, + "learning_rate": 1.7176734130705082e-05, + "loss": 0.9926, + "step": 21749 + }, + { + "epoch": 1.61, + "learning_rate": 1.7176456358843875e-05, + "loss": 0.9354, + "step": 21750 + }, + { + "epoch": 1.61, + "learning_rate": 1.717617857556503e-05, + "loss": 1.1238, + "step": 21751 + }, + { + "epoch": 1.61, + "learning_rate": 1.7175900780868984e-05, + "loss": 1.016, + "step": 21752 + }, + { + "epoch": 1.61, + "learning_rate": 1.7175622974756184e-05, + "loss": 1.1156, + "step": 21753 + }, + { + "epoch": 1.61, + "learning_rate": 1.717534515722707e-05, + "loss": 1.0469, + "step": 21754 + }, + { + "epoch": 1.61, + "learning_rate": 1.7175067328282087e-05, + "loss": 0.9994, + "step": 21755 + }, + { + "epoch": 1.61, + "learning_rate": 1.717478948792167e-05, + "loss": 0.964, + "step": 21756 + }, + { + "epoch": 1.61, + "learning_rate": 1.717451163614627e-05, + "loss": 1.0505, + "step": 21757 + }, + { + "epoch": 1.61, + "learning_rate": 1.717423377295632e-05, + "loss": 0.9056, + "step": 21758 + }, + { + "epoch": 1.61, + "learning_rate": 1.717395589835227e-05, + "loss": 1.1046, + "step": 21759 + }, + { + "epoch": 1.61, + "learning_rate": 1.717367801233456e-05, + "loss": 0.9411, + "step": 21760 + }, + { + "epoch": 1.61, + "learning_rate": 1.7173400114903628e-05, + "loss": 0.9567, + "step": 21761 + }, + { + "epoch": 1.61, + "learning_rate": 1.717312220605992e-05, + "loss": 0.9638, + "step": 21762 + }, + { + "epoch": 1.61, + "learning_rate": 1.7172844285803878e-05, + "loss": 1.0416, + "step": 21763 + }, + { + "epoch": 1.61, + "learning_rate": 1.7172566354135943e-05, + "loss": 1.025, + "step": 21764 + }, + { + "epoch": 1.61, + "learning_rate": 1.7172288411056557e-05, + "loss": 1.0052, + "step": 21765 + }, + { + "epoch": 1.61, + "learning_rate": 1.7172010456566163e-05, + "loss": 1.0472, + "step": 21766 + }, + { + "epoch": 1.61, + "learning_rate": 1.71717324906652e-05, + "loss": 0.967, + "step": 21767 + }, + { + "epoch": 1.61, + "learning_rate": 1.7171454513354117e-05, + "loss": 1.0089, + "step": 21768 + }, + { + "epoch": 1.61, + "learning_rate": 1.7171176524633356e-05, + "loss": 1.113, + "step": 21769 + }, + { + "epoch": 1.61, + "learning_rate": 1.717089852450335e-05, + "loss": 1.055, + "step": 21770 + }, + { + "epoch": 1.61, + "learning_rate": 1.7170620512964548e-05, + "loss": 1.0686, + "step": 21771 + }, + { + "epoch": 1.61, + "learning_rate": 1.717034249001739e-05, + "loss": 0.9957, + "step": 21772 + }, + { + "epoch": 1.61, + "learning_rate": 1.7170064455662324e-05, + "loss": 0.9604, + "step": 21773 + }, + { + "epoch": 1.61, + "learning_rate": 1.7169786409899783e-05, + "loss": 1.0197, + "step": 21774 + }, + { + "epoch": 1.61, + "learning_rate": 1.716950835273022e-05, + "loss": 0.9796, + "step": 21775 + }, + { + "epoch": 1.61, + "learning_rate": 1.716923028415407e-05, + "loss": 1.0256, + "step": 21776 + }, + { + "epoch": 1.61, + "learning_rate": 1.7168952204171772e-05, + "loss": 0.9966, + "step": 21777 + }, + { + "epoch": 1.61, + "learning_rate": 1.7168674112783778e-05, + "loss": 1.0541, + "step": 21778 + }, + { + "epoch": 1.61, + "learning_rate": 1.7168396009990525e-05, + "loss": 1.0039, + "step": 21779 + }, + { + "epoch": 1.61, + "learning_rate": 1.716811789579246e-05, + "loss": 1.0714, + "step": 21780 + }, + { + "epoch": 1.61, + "learning_rate": 1.716783977019002e-05, + "loss": 0.967, + "step": 21781 + }, + { + "epoch": 1.61, + "learning_rate": 1.716756163318365e-05, + "loss": 0.9786, + "step": 21782 + }, + { + "epoch": 1.61, + "learning_rate": 1.716728348477379e-05, + "loss": 1.0084, + "step": 21783 + }, + { + "epoch": 1.61, + "learning_rate": 1.716700532496088e-05, + "loss": 0.961, + "step": 21784 + }, + { + "epoch": 1.61, + "learning_rate": 1.7166727153745374e-05, + "loss": 1.0133, + "step": 21785 + }, + { + "epoch": 1.61, + "learning_rate": 1.7166448971127706e-05, + "loss": 0.9477, + "step": 21786 + }, + { + "epoch": 1.61, + "learning_rate": 1.716617077710832e-05, + "loss": 1.0707, + "step": 21787 + }, + { + "epoch": 1.61, + "learning_rate": 1.7165892571687658e-05, + "loss": 1.0723, + "step": 21788 + }, + { + "epoch": 1.61, + "learning_rate": 1.7165614354866166e-05, + "loss": 1.0192, + "step": 21789 + }, + { + "epoch": 1.61, + "learning_rate": 1.7165336126644282e-05, + "loss": 0.9347, + "step": 21790 + }, + { + "epoch": 1.61, + "learning_rate": 1.7165057887022452e-05, + "loss": 1.032, + "step": 21791 + }, + { + "epoch": 1.61, + "learning_rate": 1.7164779636001117e-05, + "loss": 1.0242, + "step": 21792 + }, + { + "epoch": 1.61, + "learning_rate": 1.716450137358072e-05, + "loss": 1.1572, + "step": 21793 + }, + { + "epoch": 1.61, + "learning_rate": 1.7164223099761705e-05, + "loss": 1.0945, + "step": 21794 + }, + { + "epoch": 1.61, + "learning_rate": 1.7163944814544508e-05, + "loss": 0.9252, + "step": 21795 + }, + { + "epoch": 1.61, + "learning_rate": 1.716366651792958e-05, + "loss": 0.9778, + "step": 21796 + }, + { + "epoch": 1.61, + "learning_rate": 1.7163388209917363e-05, + "loss": 1.0155, + "step": 21797 + }, + { + "epoch": 1.61, + "learning_rate": 1.7163109890508302e-05, + "loss": 1.1115, + "step": 21798 + }, + { + "epoch": 1.61, + "learning_rate": 1.716283155970283e-05, + "loss": 1.0241, + "step": 21799 + }, + { + "epoch": 1.61, + "learning_rate": 1.7162553217501393e-05, + "loss": 0.9574, + "step": 21800 + }, + { + "epoch": 1.61, + "learning_rate": 1.7162274863904443e-05, + "loss": 0.8905, + "step": 21801 + }, + { + "epoch": 1.61, + "learning_rate": 1.7161996498912413e-05, + "loss": 0.9924, + "step": 21802 + }, + { + "epoch": 1.61, + "learning_rate": 1.7161718122525746e-05, + "loss": 1.0435, + "step": 21803 + }, + { + "epoch": 1.61, + "learning_rate": 1.7161439734744894e-05, + "loss": 0.9779, + "step": 21804 + }, + { + "epoch": 1.61, + "learning_rate": 1.716116133557029e-05, + "loss": 1.0455, + "step": 21805 + }, + { + "epoch": 1.61, + "learning_rate": 1.716088292500238e-05, + "loss": 0.9982, + "step": 21806 + }, + { + "epoch": 1.61, + "learning_rate": 1.7160604503041613e-05, + "loss": 0.8993, + "step": 21807 + }, + { + "epoch": 1.61, + "learning_rate": 1.7160326069688423e-05, + "loss": 1.054, + "step": 21808 + }, + { + "epoch": 1.61, + "learning_rate": 1.7160047624943253e-05, + "loss": 1.0859, + "step": 21809 + }, + { + "epoch": 1.61, + "learning_rate": 1.7159769168806554e-05, + "loss": 1.0313, + "step": 21810 + }, + { + "epoch": 1.61, + "learning_rate": 1.7159490701278763e-05, + "loss": 1.1331, + "step": 21811 + }, + { + "epoch": 1.61, + "learning_rate": 1.7159212222360327e-05, + "loss": 1.0978, + "step": 21812 + }, + { + "epoch": 1.61, + "learning_rate": 1.7158933732051687e-05, + "loss": 1.0344, + "step": 21813 + }, + { + "epoch": 1.61, + "learning_rate": 1.7158655230353282e-05, + "loss": 0.9669, + "step": 21814 + }, + { + "epoch": 1.61, + "learning_rate": 1.7158376717265564e-05, + "loss": 0.9249, + "step": 21815 + }, + { + "epoch": 1.61, + "learning_rate": 1.7158098192788967e-05, + "loss": 0.8635, + "step": 21816 + }, + { + "epoch": 1.61, + "learning_rate": 1.715781965692394e-05, + "loss": 0.9856, + "step": 21817 + }, + { + "epoch": 1.61, + "learning_rate": 1.7157541109670924e-05, + "loss": 0.9635, + "step": 21818 + }, + { + "epoch": 1.61, + "learning_rate": 1.7157262551030364e-05, + "loss": 1.0478, + "step": 21819 + }, + { + "epoch": 1.61, + "learning_rate": 1.7156983981002696e-05, + "loss": 1.092, + "step": 21820 + }, + { + "epoch": 1.61, + "learning_rate": 1.7156705399588377e-05, + "loss": 0.9845, + "step": 21821 + }, + { + "epoch": 1.61, + "learning_rate": 1.7156426806787835e-05, + "loss": 1.0573, + "step": 21822 + }, + { + "epoch": 1.61, + "learning_rate": 1.7156148202601524e-05, + "loss": 1.0407, + "step": 21823 + }, + { + "epoch": 1.61, + "learning_rate": 1.7155869587029884e-05, + "loss": 1.032, + "step": 21824 + }, + { + "epoch": 1.61, + "learning_rate": 1.7155590960073355e-05, + "loss": 0.9479, + "step": 21825 + }, + { + "epoch": 1.61, + "learning_rate": 1.7155312321732385e-05, + "loss": 0.9467, + "step": 21826 + }, + { + "epoch": 1.61, + "learning_rate": 1.7155033672007413e-05, + "loss": 0.9434, + "step": 21827 + }, + { + "epoch": 1.61, + "learning_rate": 1.7154755010898892e-05, + "loss": 0.9444, + "step": 21828 + }, + { + "epoch": 1.61, + "learning_rate": 1.715447633840725e-05, + "loss": 1.0739, + "step": 21829 + }, + { + "epoch": 1.61, + "learning_rate": 1.7154197654532944e-05, + "loss": 1.008, + "step": 21830 + }, + { + "epoch": 1.61, + "learning_rate": 1.715391895927641e-05, + "loss": 1.1073, + "step": 21831 + }, + { + "epoch": 1.61, + "learning_rate": 1.7153640252638094e-05, + "loss": 1.0508, + "step": 21832 + }, + { + "epoch": 1.61, + "learning_rate": 1.7153361534618435e-05, + "loss": 1.0788, + "step": 21833 + }, + { + "epoch": 1.61, + "learning_rate": 1.7153082805217886e-05, + "loss": 0.9313, + "step": 21834 + }, + { + "epoch": 1.61, + "learning_rate": 1.715280406443688e-05, + "loss": 1.0569, + "step": 21835 + }, + { + "epoch": 1.61, + "learning_rate": 1.7152525312275868e-05, + "loss": 1.1184, + "step": 21836 + }, + { + "epoch": 1.61, + "learning_rate": 1.715224654873529e-05, + "loss": 1.0186, + "step": 21837 + }, + { + "epoch": 1.61, + "learning_rate": 1.7151967773815586e-05, + "loss": 0.9732, + "step": 21838 + }, + { + "epoch": 1.61, + "learning_rate": 1.7151688987517208e-05, + "loss": 1.1074, + "step": 21839 + }, + { + "epoch": 1.61, + "learning_rate": 1.7151410189840592e-05, + "loss": 1.0151, + "step": 21840 + }, + { + "epoch": 1.61, + "learning_rate": 1.715113138078619e-05, + "loss": 1.0165, + "step": 21841 + }, + { + "epoch": 1.61, + "learning_rate": 1.7150852560354437e-05, + "loss": 1.1279, + "step": 21842 + }, + { + "epoch": 1.61, + "learning_rate": 1.7150573728545778e-05, + "loss": 1.0332, + "step": 21843 + }, + { + "epoch": 1.61, + "learning_rate": 1.7150294885360664e-05, + "loss": 0.9272, + "step": 21844 + }, + { + "epoch": 1.61, + "learning_rate": 1.7150016030799528e-05, + "loss": 1.0433, + "step": 21845 + }, + { + "epoch": 1.61, + "learning_rate": 1.714973716486282e-05, + "loss": 1.1252, + "step": 21846 + }, + { + "epoch": 1.61, + "learning_rate": 1.7149458287550983e-05, + "loss": 0.9008, + "step": 21847 + }, + { + "epoch": 1.61, + "learning_rate": 1.7149179398864458e-05, + "loss": 0.9153, + "step": 21848 + }, + { + "epoch": 1.61, + "learning_rate": 1.7148900498803694e-05, + "loss": 0.985, + "step": 21849 + }, + { + "epoch": 1.61, + "learning_rate": 1.7148621587369134e-05, + "loss": 1.0279, + "step": 21850 + }, + { + "epoch": 1.61, + "learning_rate": 1.714834266456121e-05, + "loss": 1.0812, + "step": 21851 + }, + { + "epoch": 1.61, + "learning_rate": 1.714806373038038e-05, + "loss": 0.9958, + "step": 21852 + }, + { + "epoch": 1.61, + "learning_rate": 1.7147784784827086e-05, + "loss": 0.9584, + "step": 21853 + }, + { + "epoch": 1.61, + "learning_rate": 1.7147505827901768e-05, + "loss": 0.9754, + "step": 21854 + }, + { + "epoch": 1.61, + "learning_rate": 1.714722685960487e-05, + "loss": 1.0156, + "step": 21855 + }, + { + "epoch": 1.61, + "learning_rate": 1.7146947879936835e-05, + "loss": 0.991, + "step": 21856 + }, + { + "epoch": 1.61, + "learning_rate": 1.7146668888898105e-05, + "loss": 0.9491, + "step": 21857 + }, + { + "epoch": 1.62, + "learning_rate": 1.7146389886489133e-05, + "loss": 1.0222, + "step": 21858 + }, + { + "epoch": 1.62, + "learning_rate": 1.7146110872710354e-05, + "loss": 0.9697, + "step": 21859 + }, + { + "epoch": 1.62, + "learning_rate": 1.7145831847562217e-05, + "loss": 0.8877, + "step": 21860 + }, + { + "epoch": 1.62, + "learning_rate": 1.714555281104516e-05, + "loss": 1.0898, + "step": 21861 + }, + { + "epoch": 1.62, + "learning_rate": 1.7145273763159632e-05, + "loss": 0.8982, + "step": 21862 + }, + { + "epoch": 1.62, + "learning_rate": 1.714499470390608e-05, + "loss": 1.0409, + "step": 21863 + }, + { + "epoch": 1.62, + "learning_rate": 1.7144715633284938e-05, + "loss": 1.0322, + "step": 21864 + }, + { + "epoch": 1.62, + "learning_rate": 1.714443655129666e-05, + "loss": 0.9932, + "step": 21865 + }, + { + "epoch": 1.62, + "learning_rate": 1.714415745794168e-05, + "loss": 1.0472, + "step": 21866 + }, + { + "epoch": 1.62, + "learning_rate": 1.7143878353220454e-05, + "loss": 1.1278, + "step": 21867 + }, + { + "epoch": 1.62, + "learning_rate": 1.7143599237133416e-05, + "loss": 1.0353, + "step": 21868 + }, + { + "epoch": 1.62, + "learning_rate": 1.7143320109681013e-05, + "loss": 1.0544, + "step": 21869 + }, + { + "epoch": 1.62, + "learning_rate": 1.714304097086369e-05, + "loss": 1.0421, + "step": 21870 + }, + { + "epoch": 1.62, + "learning_rate": 1.7142761820681896e-05, + "loss": 1.0886, + "step": 21871 + }, + { + "epoch": 1.62, + "learning_rate": 1.7142482659136067e-05, + "loss": 0.9484, + "step": 21872 + }, + { + "epoch": 1.62, + "learning_rate": 1.7142203486226647e-05, + "loss": 0.9799, + "step": 21873 + }, + { + "epoch": 1.62, + "learning_rate": 1.7141924301954088e-05, + "loss": 0.9596, + "step": 21874 + }, + { + "epoch": 1.62, + "learning_rate": 1.714164510631883e-05, + "loss": 1.0068, + "step": 21875 + }, + { + "epoch": 1.62, + "learning_rate": 1.714136589932131e-05, + "loss": 0.9162, + "step": 21876 + }, + { + "epoch": 1.62, + "learning_rate": 1.7141086680961986e-05, + "loss": 0.9912, + "step": 21877 + }, + { + "epoch": 1.62, + "learning_rate": 1.714080745124129e-05, + "loss": 0.9347, + "step": 21878 + }, + { + "epoch": 1.62, + "learning_rate": 1.714052821015968e-05, + "loss": 0.9991, + "step": 21879 + }, + { + "epoch": 1.62, + "learning_rate": 1.7140248957717585e-05, + "loss": 1.1626, + "step": 21880 + }, + { + "epoch": 1.62, + "learning_rate": 1.7139969693915456e-05, + "loss": 1.0001, + "step": 21881 + }, + { + "epoch": 1.62, + "learning_rate": 1.713969041875374e-05, + "loss": 1.0515, + "step": 21882 + }, + { + "epoch": 1.62, + "learning_rate": 1.713941113223288e-05, + "loss": 1.0116, + "step": 21883 + }, + { + "epoch": 1.62, + "learning_rate": 1.7139131834353313e-05, + "loss": 1.0309, + "step": 21884 + }, + { + "epoch": 1.62, + "learning_rate": 1.7138852525115495e-05, + "loss": 0.9433, + "step": 21885 + }, + { + "epoch": 1.62, + "learning_rate": 1.7138573204519864e-05, + "loss": 1.1524, + "step": 21886 + }, + { + "epoch": 1.62, + "learning_rate": 1.7138293872566863e-05, + "loss": 0.9915, + "step": 21887 + }, + { + "epoch": 1.62, + "learning_rate": 1.713801452925694e-05, + "loss": 1.0133, + "step": 21888 + }, + { + "epoch": 1.62, + "learning_rate": 1.713773517459054e-05, + "loss": 1.0466, + "step": 21889 + }, + { + "epoch": 1.62, + "learning_rate": 1.71374558085681e-05, + "loss": 0.9416, + "step": 21890 + }, + { + "epoch": 1.62, + "learning_rate": 1.7137176431190074e-05, + "loss": 1.0129, + "step": 21891 + }, + { + "epoch": 1.62, + "learning_rate": 1.71368970424569e-05, + "loss": 1.0279, + "step": 21892 + }, + { + "epoch": 1.62, + "learning_rate": 1.7136617642369026e-05, + "loss": 0.9905, + "step": 21893 + }, + { + "epoch": 1.62, + "learning_rate": 1.7136338230926895e-05, + "loss": 1.0087, + "step": 21894 + }, + { + "epoch": 1.62, + "learning_rate": 1.7136058808130952e-05, + "loss": 1.0654, + "step": 21895 + }, + { + "epoch": 1.62, + "learning_rate": 1.7135779373981643e-05, + "loss": 1.0416, + "step": 21896 + }, + { + "epoch": 1.62, + "learning_rate": 1.7135499928479413e-05, + "loss": 1.0043, + "step": 21897 + }, + { + "epoch": 1.62, + "learning_rate": 1.7135220471624702e-05, + "loss": 1.1052, + "step": 21898 + }, + { + "epoch": 1.62, + "learning_rate": 1.7134941003417954e-05, + "loss": 1.0507, + "step": 21899 + }, + { + "epoch": 1.62, + "learning_rate": 1.7134661523859622e-05, + "loss": 1.0339, + "step": 21900 + }, + { + "epoch": 1.62, + "learning_rate": 1.7134382032950145e-05, + "loss": 1.103, + "step": 21901 + }, + { + "epoch": 1.62, + "learning_rate": 1.7134102530689968e-05, + "loss": 1.0541, + "step": 21902 + }, + { + "epoch": 1.62, + "learning_rate": 1.7133823017079533e-05, + "loss": 0.984, + "step": 21903 + }, + { + "epoch": 1.62, + "learning_rate": 1.713354349211929e-05, + "loss": 1.0274, + "step": 21904 + }, + { + "epoch": 1.62, + "learning_rate": 1.7133263955809684e-05, + "loss": 0.8837, + "step": 21905 + }, + { + "epoch": 1.62, + "learning_rate": 1.7132984408151153e-05, + "loss": 0.97, + "step": 21906 + }, + { + "epoch": 1.62, + "learning_rate": 1.7132704849144147e-05, + "loss": 1.071, + "step": 21907 + }, + { + "epoch": 1.62, + "learning_rate": 1.7132425278789108e-05, + "loss": 1.0528, + "step": 21908 + }, + { + "epoch": 1.62, + "learning_rate": 1.7132145697086487e-05, + "loss": 1.0138, + "step": 21909 + }, + { + "epoch": 1.62, + "learning_rate": 1.713186610403672e-05, + "loss": 1.0187, + "step": 21910 + }, + { + "epoch": 1.62, + "learning_rate": 1.713158649964026e-05, + "loss": 1.0636, + "step": 21911 + }, + { + "epoch": 1.62, + "learning_rate": 1.7131306883897542e-05, + "loss": 0.96, + "step": 21912 + }, + { + "epoch": 1.62, + "learning_rate": 1.7131027256809022e-05, + "loss": 1.0941, + "step": 21913 + }, + { + "epoch": 1.62, + "learning_rate": 1.713074761837514e-05, + "loss": 1.0533, + "step": 21914 + }, + { + "epoch": 1.62, + "learning_rate": 1.7130467968596335e-05, + "loss": 0.9537, + "step": 21915 + }, + { + "epoch": 1.62, + "learning_rate": 1.7130188307473063e-05, + "loss": 1.0114, + "step": 21916 + }, + { + "epoch": 1.62, + "learning_rate": 1.712990863500576e-05, + "loss": 0.9161, + "step": 21917 + }, + { + "epoch": 1.62, + "learning_rate": 1.7129628951194875e-05, + "loss": 1.0194, + "step": 21918 + }, + { + "epoch": 1.62, + "learning_rate": 1.712934925604085e-05, + "loss": 0.9636, + "step": 21919 + }, + { + "epoch": 1.62, + "learning_rate": 1.7129069549544138e-05, + "loss": 0.9911, + "step": 21920 + }, + { + "epoch": 1.62, + "learning_rate": 1.7128789831705176e-05, + "loss": 0.9524, + "step": 21921 + }, + { + "epoch": 1.62, + "learning_rate": 1.712851010252441e-05, + "loss": 0.8451, + "step": 21922 + }, + { + "epoch": 1.62, + "learning_rate": 1.712823036200229e-05, + "loss": 1.0315, + "step": 21923 + }, + { + "epoch": 1.62, + "learning_rate": 1.7127950610139254e-05, + "loss": 0.9428, + "step": 21924 + }, + { + "epoch": 1.62, + "learning_rate": 1.7127670846935752e-05, + "loss": 1.0355, + "step": 21925 + }, + { + "epoch": 1.62, + "learning_rate": 1.7127391072392227e-05, + "loss": 1.0949, + "step": 21926 + }, + { + "epoch": 1.62, + "learning_rate": 1.7127111286509122e-05, + "loss": 1.0278, + "step": 21927 + }, + { + "epoch": 1.62, + "learning_rate": 1.712683148928689e-05, + "loss": 1.1054, + "step": 21928 + }, + { + "epoch": 1.62, + "learning_rate": 1.712655168072597e-05, + "loss": 0.9988, + "step": 21929 + }, + { + "epoch": 1.62, + "learning_rate": 1.7126271860826807e-05, + "loss": 0.991, + "step": 21930 + }, + { + "epoch": 1.62, + "learning_rate": 1.712599202958985e-05, + "loss": 1.0393, + "step": 21931 + }, + { + "epoch": 1.62, + "learning_rate": 1.7125712187015534e-05, + "loss": 1.0696, + "step": 21932 + }, + { + "epoch": 1.62, + "learning_rate": 1.712543233310432e-05, + "loss": 0.9103, + "step": 21933 + }, + { + "epoch": 1.62, + "learning_rate": 1.712515246785664e-05, + "loss": 0.9382, + "step": 21934 + }, + { + "epoch": 1.62, + "learning_rate": 1.7124872591272947e-05, + "loss": 1.0395, + "step": 21935 + }, + { + "epoch": 1.62, + "learning_rate": 1.7124592703353683e-05, + "loss": 0.9445, + "step": 21936 + }, + { + "epoch": 1.62, + "learning_rate": 1.7124312804099298e-05, + "loss": 1.0266, + "step": 21937 + }, + { + "epoch": 1.62, + "learning_rate": 1.712403289351023e-05, + "loss": 0.903, + "step": 21938 + }, + { + "epoch": 1.62, + "learning_rate": 1.7123752971586925e-05, + "loss": 0.9952, + "step": 21939 + }, + { + "epoch": 1.62, + "learning_rate": 1.7123473038329832e-05, + "loss": 1.0776, + "step": 21940 + }, + { + "epoch": 1.62, + "learning_rate": 1.7123193093739398e-05, + "loss": 1.0259, + "step": 21941 + }, + { + "epoch": 1.62, + "learning_rate": 1.7122913137816063e-05, + "loss": 0.9448, + "step": 21942 + }, + { + "epoch": 1.62, + "learning_rate": 1.7122633170560277e-05, + "loss": 1.058, + "step": 21943 + }, + { + "epoch": 1.62, + "learning_rate": 1.7122353191972484e-05, + "loss": 1.008, + "step": 21944 + }, + { + "epoch": 1.62, + "learning_rate": 1.7122073202053126e-05, + "loss": 1.0055, + "step": 21945 + }, + { + "epoch": 1.62, + "learning_rate": 1.7121793200802655e-05, + "loss": 1.0186, + "step": 21946 + }, + { + "epoch": 1.62, + "learning_rate": 1.712151318822151e-05, + "loss": 0.9811, + "step": 21947 + }, + { + "epoch": 1.62, + "learning_rate": 1.712123316431014e-05, + "loss": 1.0565, + "step": 21948 + }, + { + "epoch": 1.62, + "learning_rate": 1.712095312906899e-05, + "loss": 1.051, + "step": 21949 + }, + { + "epoch": 1.62, + "learning_rate": 1.7120673082498505e-05, + "loss": 1.0217, + "step": 21950 + }, + { + "epoch": 1.62, + "learning_rate": 1.712039302459913e-05, + "loss": 0.997, + "step": 21951 + }, + { + "epoch": 1.62, + "learning_rate": 1.7120112955371316e-05, + "loss": 0.9783, + "step": 21952 + }, + { + "epoch": 1.62, + "learning_rate": 1.7119832874815502e-05, + "loss": 1.0446, + "step": 21953 + }, + { + "epoch": 1.62, + "learning_rate": 1.7119552782932136e-05, + "loss": 1.0287, + "step": 21954 + }, + { + "epoch": 1.62, + "learning_rate": 1.711927267972166e-05, + "loss": 1.0463, + "step": 21955 + }, + { + "epoch": 1.62, + "learning_rate": 1.7118992565184528e-05, + "loss": 1.0685, + "step": 21956 + }, + { + "epoch": 1.62, + "learning_rate": 1.7118712439321175e-05, + "loss": 0.9648, + "step": 21957 + }, + { + "epoch": 1.62, + "learning_rate": 1.711843230213206e-05, + "loss": 0.9947, + "step": 21958 + }, + { + "epoch": 1.62, + "learning_rate": 1.7118152153617614e-05, + "loss": 1.0701, + "step": 21959 + }, + { + "epoch": 1.62, + "learning_rate": 1.7117871993778293e-05, + "loss": 1.0595, + "step": 21960 + }, + { + "epoch": 1.62, + "learning_rate": 1.711759182261454e-05, + "loss": 1.0732, + "step": 21961 + }, + { + "epoch": 1.62, + "learning_rate": 1.71173116401268e-05, + "loss": 1.0216, + "step": 21962 + }, + { + "epoch": 1.62, + "learning_rate": 1.7117031446315518e-05, + "loss": 1.1016, + "step": 21963 + }, + { + "epoch": 1.62, + "learning_rate": 1.7116751241181144e-05, + "loss": 0.9017, + "step": 21964 + }, + { + "epoch": 1.62, + "learning_rate": 1.7116471024724114e-05, + "loss": 0.957, + "step": 21965 + }, + { + "epoch": 1.62, + "learning_rate": 1.7116190796944887e-05, + "loss": 1.0083, + "step": 21966 + }, + { + "epoch": 1.62, + "learning_rate": 1.71159105578439e-05, + "loss": 0.961, + "step": 21967 + }, + { + "epoch": 1.62, + "learning_rate": 1.7115630307421597e-05, + "loss": 1.0555, + "step": 21968 + }, + { + "epoch": 1.62, + "learning_rate": 1.7115350045678432e-05, + "loss": 0.9686, + "step": 21969 + }, + { + "epoch": 1.62, + "learning_rate": 1.711506977261485e-05, + "loss": 1.0268, + "step": 21970 + }, + { + "epoch": 1.62, + "learning_rate": 1.7114789488231286e-05, + "loss": 0.9813, + "step": 21971 + }, + { + "epoch": 1.62, + "learning_rate": 1.71145091925282e-05, + "loss": 0.9522, + "step": 21972 + }, + { + "epoch": 1.62, + "learning_rate": 1.7114228885506027e-05, + "loss": 1.0305, + "step": 21973 + }, + { + "epoch": 1.62, + "learning_rate": 1.711394856716522e-05, + "loss": 1.053, + "step": 21974 + }, + { + "epoch": 1.62, + "learning_rate": 1.7113668237506223e-05, + "loss": 0.8988, + "step": 21975 + }, + { + "epoch": 1.62, + "learning_rate": 1.7113387896529482e-05, + "loss": 1.0064, + "step": 21976 + }, + { + "epoch": 1.62, + "learning_rate": 1.7113107544235443e-05, + "loss": 0.9945, + "step": 21977 + }, + { + "epoch": 1.62, + "learning_rate": 1.7112827180624552e-05, + "loss": 0.9698, + "step": 21978 + }, + { + "epoch": 1.62, + "learning_rate": 1.7112546805697253e-05, + "loss": 0.9991, + "step": 21979 + }, + { + "epoch": 1.62, + "learning_rate": 1.7112266419453994e-05, + "loss": 0.9882, + "step": 21980 + }, + { + "epoch": 1.62, + "learning_rate": 1.711198602189522e-05, + "loss": 1.0445, + "step": 21981 + }, + { + "epoch": 1.62, + "learning_rate": 1.7111705613021377e-05, + "loss": 1.0671, + "step": 21982 + }, + { + "epoch": 1.62, + "learning_rate": 1.7111425192832917e-05, + "loss": 0.9674, + "step": 21983 + }, + { + "epoch": 1.62, + "learning_rate": 1.711114476133028e-05, + "loss": 1.0145, + "step": 21984 + }, + { + "epoch": 1.62, + "learning_rate": 1.711086431851391e-05, + "loss": 1.0309, + "step": 21985 + }, + { + "epoch": 1.62, + "learning_rate": 1.7110583864384257e-05, + "loss": 1.1076, + "step": 21986 + }, + { + "epoch": 1.62, + "learning_rate": 1.711030339894177e-05, + "loss": 1.0256, + "step": 21987 + }, + { + "epoch": 1.62, + "learning_rate": 1.711002292218689e-05, + "loss": 1.064, + "step": 21988 + }, + { + "epoch": 1.62, + "learning_rate": 1.7109742434120067e-05, + "loss": 1.0318, + "step": 21989 + }, + { + "epoch": 1.62, + "learning_rate": 1.7109461934741745e-05, + "loss": 1.1483, + "step": 21990 + }, + { + "epoch": 1.62, + "learning_rate": 1.710918142405237e-05, + "loss": 0.9932, + "step": 21991 + }, + { + "epoch": 1.62, + "learning_rate": 1.7108900902052387e-05, + "loss": 0.9717, + "step": 21992 + }, + { + "epoch": 1.63, + "learning_rate": 1.7108620368742247e-05, + "loss": 0.9488, + "step": 21993 + }, + { + "epoch": 1.63, + "learning_rate": 1.7108339824122394e-05, + "loss": 1.0495, + "step": 21994 + }, + { + "epoch": 1.63, + "learning_rate": 1.7108059268193275e-05, + "loss": 0.9808, + "step": 21995 + }, + { + "epoch": 1.63, + "learning_rate": 1.710777870095533e-05, + "loss": 1.1118, + "step": 21996 + }, + { + "epoch": 1.63, + "learning_rate": 1.7107498122409017e-05, + "loss": 0.988, + "step": 21997 + }, + { + "epoch": 1.63, + "learning_rate": 1.7107217532554772e-05, + "loss": 0.9822, + "step": 21998 + }, + { + "epoch": 1.63, + "learning_rate": 1.7106936931393047e-05, + "loss": 1.0199, + "step": 21999 + }, + { + "epoch": 1.63, + "learning_rate": 1.710665631892429e-05, + "loss": 0.9692, + "step": 22000 + }, + { + "epoch": 1.63, + "learning_rate": 1.710637569514894e-05, + "loss": 1.0209, + "step": 22001 + }, + { + "epoch": 1.63, + "learning_rate": 1.7106095060067448e-05, + "loss": 0.9403, + "step": 22002 + }, + { + "epoch": 1.63, + "learning_rate": 1.7105814413680263e-05, + "loss": 0.9645, + "step": 22003 + }, + { + "epoch": 1.63, + "learning_rate": 1.710553375598783e-05, + "loss": 0.8615, + "step": 22004 + }, + { + "epoch": 1.63, + "learning_rate": 1.710525308699059e-05, + "loss": 1.0099, + "step": 22005 + }, + { + "epoch": 1.63, + "learning_rate": 1.7104972406689e-05, + "loss": 1.0716, + "step": 22006 + }, + { + "epoch": 1.63, + "learning_rate": 1.710469171508349e-05, + "loss": 1.0838, + "step": 22007 + }, + { + "epoch": 1.63, + "learning_rate": 1.7104411012174525e-05, + "loss": 1.0059, + "step": 22008 + }, + { + "epoch": 1.63, + "learning_rate": 1.7104130297962545e-05, + "loss": 1.0239, + "step": 22009 + }, + { + "epoch": 1.63, + "learning_rate": 1.7103849572447994e-05, + "loss": 1.027, + "step": 22010 + }, + { + "epoch": 1.63, + "learning_rate": 1.710356883563132e-05, + "loss": 1.0672, + "step": 22011 + }, + { + "epoch": 1.63, + "learning_rate": 1.7103288087512967e-05, + "loss": 1.0507, + "step": 22012 + }, + { + "epoch": 1.63, + "learning_rate": 1.710300732809339e-05, + "loss": 1.0738, + "step": 22013 + }, + { + "epoch": 1.63, + "learning_rate": 1.7102726557373022e-05, + "loss": 1.0307, + "step": 22014 + }, + { + "epoch": 1.63, + "learning_rate": 1.710244577535232e-05, + "loss": 0.999, + "step": 22015 + }, + { + "epoch": 1.63, + "learning_rate": 1.710216498203173e-05, + "loss": 1.0548, + "step": 22016 + }, + { + "epoch": 1.63, + "learning_rate": 1.7101884177411697e-05, + "loss": 1.0494, + "step": 22017 + }, + { + "epoch": 1.63, + "learning_rate": 1.710160336149267e-05, + "loss": 0.9627, + "step": 22018 + }, + { + "epoch": 1.63, + "learning_rate": 1.710132253427509e-05, + "loss": 0.9479, + "step": 22019 + }, + { + "epoch": 1.63, + "learning_rate": 1.7101041695759412e-05, + "loss": 0.9969, + "step": 22020 + }, + { + "epoch": 1.63, + "learning_rate": 1.7100760845946073e-05, + "loss": 0.9602, + "step": 22021 + }, + { + "epoch": 1.63, + "learning_rate": 1.710047998483553e-05, + "loss": 1.0177, + "step": 22022 + }, + { + "epoch": 1.63, + "learning_rate": 1.710019911242822e-05, + "loss": 1.0413, + "step": 22023 + }, + { + "epoch": 1.63, + "learning_rate": 1.70999182287246e-05, + "loss": 0.9923, + "step": 22024 + }, + { + "epoch": 1.63, + "learning_rate": 1.709963733372511e-05, + "loss": 0.9775, + "step": 22025 + }, + { + "epoch": 1.63, + "learning_rate": 1.7099356427430197e-05, + "loss": 1.0278, + "step": 22026 + }, + { + "epoch": 1.63, + "learning_rate": 1.7099075509840307e-05, + "loss": 0.9169, + "step": 22027 + }, + { + "epoch": 1.63, + "learning_rate": 1.7098794580955894e-05, + "loss": 0.9798, + "step": 22028 + }, + { + "epoch": 1.63, + "learning_rate": 1.7098513640777397e-05, + "loss": 1.0441, + "step": 22029 + }, + { + "epoch": 1.63, + "learning_rate": 1.709823268930527e-05, + "loss": 1.0643, + "step": 22030 + }, + { + "epoch": 1.63, + "learning_rate": 1.7097951726539953e-05, + "loss": 0.9269, + "step": 22031 + }, + { + "epoch": 1.63, + "learning_rate": 1.7097670752481897e-05, + "loss": 1.0146, + "step": 22032 + }, + { + "epoch": 1.63, + "learning_rate": 1.709738976713155e-05, + "loss": 1.0932, + "step": 22033 + }, + { + "epoch": 1.63, + "learning_rate": 1.7097108770489356e-05, + "loss": 0.9519, + "step": 22034 + }, + { + "epoch": 1.63, + "learning_rate": 1.7096827762555765e-05, + "loss": 0.9812, + "step": 22035 + }, + { + "epoch": 1.63, + "learning_rate": 1.7096546743331222e-05, + "loss": 0.914, + "step": 22036 + }, + { + "epoch": 1.63, + "learning_rate": 1.7096265712816174e-05, + "loss": 0.9312, + "step": 22037 + }, + { + "epoch": 1.63, + "learning_rate": 1.709598467101107e-05, + "loss": 1.0631, + "step": 22038 + }, + { + "epoch": 1.63, + "learning_rate": 1.7095703617916357e-05, + "loss": 0.9689, + "step": 22039 + }, + { + "epoch": 1.63, + "learning_rate": 1.709542255353248e-05, + "loss": 1.0494, + "step": 22040 + }, + { + "epoch": 1.63, + "learning_rate": 1.7095141477859885e-05, + "loss": 0.96, + "step": 22041 + }, + { + "epoch": 1.63, + "learning_rate": 1.7094860390899022e-05, + "loss": 0.9296, + "step": 22042 + }, + { + "epoch": 1.63, + "learning_rate": 1.709457929265034e-05, + "loss": 0.9761, + "step": 22043 + }, + { + "epoch": 1.63, + "learning_rate": 1.709429818311428e-05, + "loss": 0.9715, + "step": 22044 + }, + { + "epoch": 1.63, + "learning_rate": 1.7094017062291297e-05, + "loss": 0.9392, + "step": 22045 + }, + { + "epoch": 1.63, + "learning_rate": 1.7093735930181833e-05, + "loss": 0.997, + "step": 22046 + }, + { + "epoch": 1.63, + "learning_rate": 1.7093454786786335e-05, + "loss": 1.0164, + "step": 22047 + }, + { + "epoch": 1.63, + "learning_rate": 1.7093173632105252e-05, + "loss": 0.9847, + "step": 22048 + }, + { + "epoch": 1.63, + "learning_rate": 1.7092892466139034e-05, + "loss": 1.0096, + "step": 22049 + }, + { + "epoch": 1.63, + "learning_rate": 1.7092611288888125e-05, + "loss": 0.9509, + "step": 22050 + }, + { + "epoch": 1.63, + "learning_rate": 1.709233010035297e-05, + "loss": 1.0489, + "step": 22051 + }, + { + "epoch": 1.63, + "learning_rate": 1.7092048900534025e-05, + "loss": 0.9884, + "step": 22052 + }, + { + "epoch": 1.63, + "learning_rate": 1.7091767689431728e-05, + "loss": 1.1409, + "step": 22053 + }, + { + "epoch": 1.63, + "learning_rate": 1.709148646704653e-05, + "loss": 1.0774, + "step": 22054 + }, + { + "epoch": 1.63, + "learning_rate": 1.709120523337888e-05, + "loss": 1.0184, + "step": 22055 + }, + { + "epoch": 1.63, + "learning_rate": 1.709092398842922e-05, + "loss": 1.0463, + "step": 22056 + }, + { + "epoch": 1.63, + "learning_rate": 1.7090642732198004e-05, + "loss": 0.9798, + "step": 22057 + }, + { + "epoch": 1.63, + "learning_rate": 1.7090361464685675e-05, + "loss": 1.1069, + "step": 22058 + }, + { + "epoch": 1.63, + "learning_rate": 1.7090080185892686e-05, + "loss": 0.9202, + "step": 22059 + }, + { + "epoch": 1.63, + "learning_rate": 1.708979889581948e-05, + "loss": 0.9576, + "step": 22060 + }, + { + "epoch": 1.63, + "learning_rate": 1.7089517594466502e-05, + "loss": 0.9855, + "step": 22061 + }, + { + "epoch": 1.63, + "learning_rate": 1.7089236281834205e-05, + "loss": 0.9951, + "step": 22062 + }, + { + "epoch": 1.63, + "learning_rate": 1.7088954957923037e-05, + "loss": 1.1093, + "step": 22063 + }, + { + "epoch": 1.63, + "learning_rate": 1.708867362273344e-05, + "loss": 1.0295, + "step": 22064 + }, + { + "epoch": 1.63, + "learning_rate": 1.7088392276265866e-05, + "loss": 1.0076, + "step": 22065 + }, + { + "epoch": 1.63, + "learning_rate": 1.708811091852076e-05, + "loss": 1.0436, + "step": 22066 + }, + { + "epoch": 1.63, + "learning_rate": 1.7087829549498573e-05, + "loss": 1.1195, + "step": 22067 + }, + { + "epoch": 1.63, + "learning_rate": 1.7087548169199747e-05, + "loss": 1.0637, + "step": 22068 + }, + { + "epoch": 1.63, + "learning_rate": 1.708726677762474e-05, + "loss": 0.9396, + "step": 22069 + }, + { + "epoch": 1.63, + "learning_rate": 1.7086985374773986e-05, + "loss": 0.9383, + "step": 22070 + }, + { + "epoch": 1.63, + "learning_rate": 1.7086703960647945e-05, + "loss": 1.0497, + "step": 22071 + }, + { + "epoch": 1.63, + "learning_rate": 1.7086422535247055e-05, + "loss": 0.9617, + "step": 22072 + }, + { + "epoch": 1.63, + "learning_rate": 1.708614109857177e-05, + "loss": 0.9539, + "step": 22073 + }, + { + "epoch": 1.63, + "learning_rate": 1.7085859650622537e-05, + "loss": 1.113, + "step": 22074 + }, + { + "epoch": 1.63, + "learning_rate": 1.7085578191399798e-05, + "loss": 1.0748, + "step": 22075 + }, + { + "epoch": 1.63, + "learning_rate": 1.708529672090401e-05, + "loss": 1.038, + "step": 22076 + }, + { + "epoch": 1.63, + "learning_rate": 1.7085015239135617e-05, + "loss": 0.9473, + "step": 22077 + }, + { + "epoch": 1.63, + "learning_rate": 1.7084733746095063e-05, + "loss": 1.0258, + "step": 22078 + }, + { + "epoch": 1.63, + "learning_rate": 1.70844522417828e-05, + "loss": 0.9933, + "step": 22079 + }, + { + "epoch": 1.63, + "learning_rate": 1.7084170726199275e-05, + "loss": 1.0152, + "step": 22080 + }, + { + "epoch": 1.63, + "learning_rate": 1.7083889199344936e-05, + "loss": 1.044, + "step": 22081 + }, + { + "epoch": 1.63, + "learning_rate": 1.708360766122023e-05, + "loss": 1.11, + "step": 22082 + }, + { + "epoch": 1.63, + "learning_rate": 1.7083326111825605e-05, + "loss": 1.0038, + "step": 22083 + }, + { + "epoch": 1.63, + "learning_rate": 1.7083044551161512e-05, + "loss": 1.0912, + "step": 22084 + }, + { + "epoch": 1.63, + "learning_rate": 1.7082762979228395e-05, + "loss": 0.9621, + "step": 22085 + }, + { + "epoch": 1.63, + "learning_rate": 1.7082481396026703e-05, + "loss": 1.0599, + "step": 22086 + }, + { + "epoch": 1.63, + "learning_rate": 1.7082199801556888e-05, + "loss": 0.9521, + "step": 22087 + }, + { + "epoch": 1.63, + "learning_rate": 1.708191819581939e-05, + "loss": 1.0493, + "step": 22088 + }, + { + "epoch": 1.63, + "learning_rate": 1.7081636578814663e-05, + "loss": 0.9968, + "step": 22089 + }, + { + "epoch": 1.63, + "learning_rate": 1.7081354950543155e-05, + "loss": 1.0669, + "step": 22090 + }, + { + "epoch": 1.63, + "learning_rate": 1.708107331100531e-05, + "loss": 0.9836, + "step": 22091 + }, + { + "epoch": 1.63, + "learning_rate": 1.708079166020158e-05, + "loss": 1.0021, + "step": 22092 + }, + { + "epoch": 1.63, + "learning_rate": 1.708050999813241e-05, + "loss": 0.9796, + "step": 22093 + }, + { + "epoch": 1.63, + "learning_rate": 1.7080228324798253e-05, + "loss": 1.0448, + "step": 22094 + }, + { + "epoch": 1.63, + "learning_rate": 1.707994664019955e-05, + "loss": 1.0683, + "step": 22095 + }, + { + "epoch": 1.63, + "learning_rate": 1.7079664944336755e-05, + "loss": 0.9779, + "step": 22096 + }, + { + "epoch": 1.63, + "learning_rate": 1.7079383237210318e-05, + "loss": 1.0937, + "step": 22097 + }, + { + "epoch": 1.63, + "learning_rate": 1.707910151882068e-05, + "loss": 1.1265, + "step": 22098 + }, + { + "epoch": 1.63, + "learning_rate": 1.707881978916829e-05, + "loss": 1.0376, + "step": 22099 + }, + { + "epoch": 1.63, + "learning_rate": 1.7078538048253603e-05, + "loss": 1.0888, + "step": 22100 + }, + { + "epoch": 1.63, + "learning_rate": 1.707825629607706e-05, + "loss": 0.9804, + "step": 22101 + }, + { + "epoch": 1.63, + "learning_rate": 1.707797453263912e-05, + "loss": 0.9759, + "step": 22102 + }, + { + "epoch": 1.63, + "learning_rate": 1.7077692757940218e-05, + "loss": 1.0081, + "step": 22103 + }, + { + "epoch": 1.63, + "learning_rate": 1.7077410971980808e-05, + "loss": 0.9322, + "step": 22104 + }, + { + "epoch": 1.63, + "learning_rate": 1.707712917476134e-05, + "loss": 1.0075, + "step": 22105 + }, + { + "epoch": 1.63, + "learning_rate": 1.707684736628226e-05, + "loss": 1.0463, + "step": 22106 + }, + { + "epoch": 1.63, + "learning_rate": 1.7076565546544017e-05, + "loss": 1.0161, + "step": 22107 + }, + { + "epoch": 1.63, + "learning_rate": 1.707628371554706e-05, + "loss": 0.9085, + "step": 22108 + }, + { + "epoch": 1.63, + "learning_rate": 1.7076001873291836e-05, + "loss": 1.0795, + "step": 22109 + }, + { + "epoch": 1.63, + "learning_rate": 1.7075720019778793e-05, + "loss": 0.9705, + "step": 22110 + }, + { + "epoch": 1.63, + "learning_rate": 1.7075438155008384e-05, + "loss": 0.979, + "step": 22111 + }, + { + "epoch": 1.63, + "learning_rate": 1.707515627898105e-05, + "loss": 1.0328, + "step": 22112 + }, + { + "epoch": 1.63, + "learning_rate": 1.7074874391697248e-05, + "loss": 1.027, + "step": 22113 + }, + { + "epoch": 1.63, + "learning_rate": 1.707459249315742e-05, + "loss": 0.9108, + "step": 22114 + }, + { + "epoch": 1.63, + "learning_rate": 1.707431058336202e-05, + "loss": 0.9313, + "step": 22115 + }, + { + "epoch": 1.63, + "learning_rate": 1.7074028662311487e-05, + "loss": 0.8942, + "step": 22116 + }, + { + "epoch": 1.63, + "learning_rate": 1.707374673000628e-05, + "loss": 1.0772, + "step": 22117 + }, + { + "epoch": 1.63, + "learning_rate": 1.7073464786446842e-05, + "loss": 1.0644, + "step": 22118 + }, + { + "epoch": 1.63, + "learning_rate": 1.7073182831633623e-05, + "loss": 1.0789, + "step": 22119 + }, + { + "epoch": 1.63, + "learning_rate": 1.707290086556707e-05, + "loss": 0.9648, + "step": 22120 + }, + { + "epoch": 1.63, + "learning_rate": 1.707261888824763e-05, + "loss": 1.0521, + "step": 22121 + }, + { + "epoch": 1.63, + "learning_rate": 1.707233689967576e-05, + "loss": 1.0166, + "step": 22122 + }, + { + "epoch": 1.63, + "learning_rate": 1.70720548998519e-05, + "loss": 1.0538, + "step": 22123 + }, + { + "epoch": 1.63, + "learning_rate": 1.7071772888776504e-05, + "loss": 1.0687, + "step": 22124 + }, + { + "epoch": 1.63, + "learning_rate": 1.707149086645002e-05, + "loss": 1.1428, + "step": 22125 + }, + { + "epoch": 1.63, + "learning_rate": 1.707120883287289e-05, + "loss": 0.9533, + "step": 22126 + }, + { + "epoch": 1.63, + "learning_rate": 1.7070926788045573e-05, + "loss": 1.0427, + "step": 22127 + }, + { + "epoch": 1.64, + "learning_rate": 1.7070644731968508e-05, + "loss": 0.898, + "step": 22128 + }, + { + "epoch": 1.64, + "learning_rate": 1.7070362664642154e-05, + "loss": 1.0018, + "step": 22129 + }, + { + "epoch": 1.64, + "learning_rate": 1.7070080586066948e-05, + "loss": 1.0573, + "step": 22130 + }, + { + "epoch": 1.64, + "learning_rate": 1.7069798496243348e-05, + "loss": 1.128, + "step": 22131 + }, + { + "epoch": 1.64, + "learning_rate": 1.70695163951718e-05, + "loss": 1.0007, + "step": 22132 + }, + { + "epoch": 1.64, + "learning_rate": 1.706923428285275e-05, + "loss": 1.0084, + "step": 22133 + }, + { + "epoch": 1.64, + "learning_rate": 1.706895215928665e-05, + "loss": 1.0317, + "step": 22134 + }, + { + "epoch": 1.64, + "learning_rate": 1.7068670024473952e-05, + "loss": 0.9673, + "step": 22135 + }, + { + "epoch": 1.64, + "learning_rate": 1.7068387878415103e-05, + "loss": 1.1055, + "step": 22136 + }, + { + "epoch": 1.64, + "learning_rate": 1.7068105721110543e-05, + "loss": 0.9176, + "step": 22137 + }, + { + "epoch": 1.64, + "learning_rate": 1.7067823552560733e-05, + "loss": 1.0578, + "step": 22138 + }, + { + "epoch": 1.64, + "learning_rate": 1.7067541372766113e-05, + "loss": 1.0671, + "step": 22139 + }, + { + "epoch": 1.64, + "learning_rate": 1.7067259181727137e-05, + "loss": 1.1142, + "step": 22140 + }, + { + "epoch": 1.64, + "learning_rate": 1.7066976979444253e-05, + "loss": 0.9974, + "step": 22141 + }, + { + "epoch": 1.64, + "learning_rate": 1.7066694765917912e-05, + "loss": 0.937, + "step": 22142 + }, + { + "epoch": 1.64, + "learning_rate": 1.706641254114856e-05, + "loss": 0.9093, + "step": 22143 + }, + { + "epoch": 1.64, + "learning_rate": 1.7066130305136644e-05, + "loss": 0.9566, + "step": 22144 + }, + { + "epoch": 1.64, + "learning_rate": 1.7065848057882616e-05, + "loss": 1.0381, + "step": 22145 + }, + { + "epoch": 1.64, + "learning_rate": 1.7065565799386926e-05, + "loss": 1.0935, + "step": 22146 + }, + { + "epoch": 1.64, + "learning_rate": 1.706528352965002e-05, + "loss": 1.1059, + "step": 22147 + }, + { + "epoch": 1.64, + "learning_rate": 1.7065001248672356e-05, + "loss": 1.1223, + "step": 22148 + }, + { + "epoch": 1.64, + "learning_rate": 1.706471895645437e-05, + "loss": 0.988, + "step": 22149 + }, + { + "epoch": 1.64, + "learning_rate": 1.7064436652996517e-05, + "loss": 1.072, + "step": 22150 + }, + { + "epoch": 1.64, + "learning_rate": 1.706415433829925e-05, + "loss": 1.0493, + "step": 22151 + }, + { + "epoch": 1.64, + "learning_rate": 1.706387201236301e-05, + "loss": 0.9996, + "step": 22152 + }, + { + "epoch": 1.64, + "learning_rate": 1.7063589675188255e-05, + "loss": 0.9839, + "step": 22153 + }, + { + "epoch": 1.64, + "learning_rate": 1.7063307326775427e-05, + "loss": 1.0684, + "step": 22154 + }, + { + "epoch": 1.64, + "learning_rate": 1.706302496712498e-05, + "loss": 0.9525, + "step": 22155 + }, + { + "epoch": 1.64, + "learning_rate": 1.706274259623736e-05, + "loss": 1.1125, + "step": 22156 + }, + { + "epoch": 1.64, + "learning_rate": 1.7062460214113017e-05, + "loss": 1.0052, + "step": 22157 + }, + { + "epoch": 1.64, + "learning_rate": 1.7062177820752403e-05, + "loss": 1.0401, + "step": 22158 + }, + { + "epoch": 1.64, + "learning_rate": 1.7061895416155964e-05, + "loss": 0.9323, + "step": 22159 + }, + { + "epoch": 1.64, + "learning_rate": 1.7061613000324148e-05, + "loss": 0.9882, + "step": 22160 + }, + { + "epoch": 1.64, + "learning_rate": 1.7061330573257412e-05, + "loss": 1.0398, + "step": 22161 + }, + { + "epoch": 1.64, + "learning_rate": 1.7061048134956194e-05, + "loss": 0.9967, + "step": 22162 + }, + { + "epoch": 1.64, + "learning_rate": 1.7060765685420953e-05, + "loss": 0.9912, + "step": 22163 + }, + { + "epoch": 1.64, + "learning_rate": 1.7060483224652135e-05, + "loss": 1.0249, + "step": 22164 + }, + { + "epoch": 1.64, + "learning_rate": 1.706020075265019e-05, + "loss": 0.9981, + "step": 22165 + }, + { + "epoch": 1.64, + "learning_rate": 1.7059918269415565e-05, + "loss": 0.8466, + "step": 22166 + }, + { + "epoch": 1.64, + "learning_rate": 1.705963577494871e-05, + "loss": 1.0575, + "step": 22167 + }, + { + "epoch": 1.64, + "learning_rate": 1.705935326925008e-05, + "loss": 1.034, + "step": 22168 + }, + { + "epoch": 1.64, + "learning_rate": 1.7059070752320114e-05, + "loss": 1.0535, + "step": 22169 + }, + { + "epoch": 1.64, + "learning_rate": 1.7058788224159267e-05, + "loss": 0.9638, + "step": 22170 + }, + { + "epoch": 1.64, + "learning_rate": 1.7058505684767994e-05, + "loss": 1.0026, + "step": 22171 + }, + { + "epoch": 1.64, + "learning_rate": 1.7058223134146737e-05, + "loss": 1.0174, + "step": 22172 + }, + { + "epoch": 1.64, + "learning_rate": 1.705794057229595e-05, + "loss": 1.1524, + "step": 22173 + }, + { + "epoch": 1.64, + "learning_rate": 1.7057657999216077e-05, + "loss": 1.0776, + "step": 22174 + }, + { + "epoch": 1.64, + "learning_rate": 1.705737541490757e-05, + "loss": 0.9583, + "step": 22175 + }, + { + "epoch": 1.64, + "learning_rate": 1.7057092819370885e-05, + "loss": 0.9619, + "step": 22176 + }, + { + "epoch": 1.64, + "learning_rate": 1.7056810212606466e-05, + "loss": 0.982, + "step": 22177 + }, + { + "epoch": 1.64, + "learning_rate": 1.7056527594614763e-05, + "loss": 0.9563, + "step": 22178 + }, + { + "epoch": 1.64, + "learning_rate": 1.705624496539622e-05, + "loss": 0.9836, + "step": 22179 + }, + { + "epoch": 1.64, + "learning_rate": 1.70559623249513e-05, + "loss": 0.9925, + "step": 22180 + }, + { + "epoch": 1.64, + "learning_rate": 1.7055679673280438e-05, + "loss": 0.8953, + "step": 22181 + }, + { + "epoch": 1.64, + "learning_rate": 1.7055397010384093e-05, + "loss": 0.9586, + "step": 22182 + }, + { + "epoch": 1.64, + "learning_rate": 1.7055114336262716e-05, + "loss": 0.928, + "step": 22183 + }, + { + "epoch": 1.64, + "learning_rate": 1.705483165091675e-05, + "loss": 1.0187, + "step": 22184 + }, + { + "epoch": 1.64, + "learning_rate": 1.7054548954346643e-05, + "loss": 0.8989, + "step": 22185 + }, + { + "epoch": 1.64, + "learning_rate": 1.7054266246552857e-05, + "loss": 1.0572, + "step": 22186 + }, + { + "epoch": 1.64, + "learning_rate": 1.705398352753583e-05, + "loss": 0.8899, + "step": 22187 + }, + { + "epoch": 1.64, + "learning_rate": 1.7053700797296015e-05, + "loss": 0.953, + "step": 22188 + }, + { + "epoch": 1.64, + "learning_rate": 1.7053418055833865e-05, + "loss": 1.0955, + "step": 22189 + }, + { + "epoch": 1.64, + "learning_rate": 1.705313530314983e-05, + "loss": 1.0745, + "step": 22190 + }, + { + "epoch": 1.64, + "learning_rate": 1.7052852539244355e-05, + "loss": 1.0129, + "step": 22191 + }, + { + "epoch": 1.64, + "learning_rate": 1.7052569764117897e-05, + "loss": 1.094, + "step": 22192 + }, + { + "epoch": 1.64, + "learning_rate": 1.7052286977770895e-05, + "loss": 1.0771, + "step": 22193 + }, + { + "epoch": 1.64, + "learning_rate": 1.7052004180203806e-05, + "loss": 1.0748, + "step": 22194 + }, + { + "epoch": 1.64, + "learning_rate": 1.7051721371417085e-05, + "loss": 0.9194, + "step": 22195 + }, + { + "epoch": 1.64, + "learning_rate": 1.7051438551411172e-05, + "loss": 0.9786, + "step": 22196 + }, + { + "epoch": 1.64, + "learning_rate": 1.705115572018652e-05, + "loss": 0.9597, + "step": 22197 + }, + { + "epoch": 1.64, + "learning_rate": 1.7050872877743583e-05, + "loss": 0.9692, + "step": 22198 + }, + { + "epoch": 1.64, + "learning_rate": 1.705059002408281e-05, + "loss": 1.0963, + "step": 22199 + }, + { + "epoch": 1.64, + "learning_rate": 1.705030715920464e-05, + "loss": 1.0844, + "step": 22200 + }, + { + "epoch": 1.64, + "learning_rate": 1.7050024283109538e-05, + "loss": 1.1373, + "step": 22201 + }, + { + "epoch": 1.64, + "learning_rate": 1.7049741395797953e-05, + "loss": 1.1241, + "step": 22202 + }, + { + "epoch": 1.64, + "learning_rate": 1.7049458497270324e-05, + "loss": 0.9372, + "step": 22203 + }, + { + "epoch": 1.64, + "learning_rate": 1.704917558752711e-05, + "loss": 0.9509, + "step": 22204 + }, + { + "epoch": 1.64, + "learning_rate": 1.7048892666568756e-05, + "loss": 0.9594, + "step": 22205 + }, + { + "epoch": 1.64, + "learning_rate": 1.7048609734395718e-05, + "loss": 1.0108, + "step": 22206 + }, + { + "epoch": 1.64, + "learning_rate": 1.7048326791008442e-05, + "loss": 1.0066, + "step": 22207 + }, + { + "epoch": 1.64, + "learning_rate": 1.7048043836407377e-05, + "loss": 0.9298, + "step": 22208 + }, + { + "epoch": 1.64, + "learning_rate": 1.7047760870592976e-05, + "loss": 0.9547, + "step": 22209 + }, + { + "epoch": 1.64, + "learning_rate": 1.7047477893565688e-05, + "loss": 0.9638, + "step": 22210 + }, + { + "epoch": 1.64, + "learning_rate": 1.7047194905325963e-05, + "loss": 1.1199, + "step": 22211 + }, + { + "epoch": 1.64, + "learning_rate": 1.7046911905874254e-05, + "loss": 0.9763, + "step": 22212 + }, + { + "epoch": 1.64, + "learning_rate": 1.704662889521101e-05, + "loss": 1.0332, + "step": 22213 + }, + { + "epoch": 1.64, + "learning_rate": 1.7046345873336677e-05, + "loss": 1.024, + "step": 22214 + }, + { + "epoch": 1.64, + "learning_rate": 1.704606284025171e-05, + "loss": 0.9325, + "step": 22215 + }, + { + "epoch": 1.64, + "learning_rate": 1.7045779795956553e-05, + "loss": 1.0894, + "step": 22216 + }, + { + "epoch": 1.64, + "learning_rate": 1.7045496740451666e-05, + "loss": 1.0244, + "step": 22217 + }, + { + "epoch": 1.64, + "learning_rate": 1.7045213673737494e-05, + "loss": 0.9578, + "step": 22218 + }, + { + "epoch": 1.64, + "learning_rate": 1.7044930595814487e-05, + "loss": 0.9342, + "step": 22219 + }, + { + "epoch": 1.64, + "learning_rate": 1.70446475066831e-05, + "loss": 1.1057, + "step": 22220 + }, + { + "epoch": 1.64, + "learning_rate": 1.7044364406343776e-05, + "loss": 1.0385, + "step": 22221 + }, + { + "epoch": 1.64, + "learning_rate": 1.704408129479697e-05, + "loss": 1.0172, + "step": 22222 + }, + { + "epoch": 1.64, + "learning_rate": 1.7043798172043127e-05, + "loss": 0.9724, + "step": 22223 + }, + { + "epoch": 1.64, + "learning_rate": 1.7043515038082708e-05, + "loss": 1.0197, + "step": 22224 + }, + { + "epoch": 1.64, + "learning_rate": 1.7043231892916154e-05, + "loss": 1.0573, + "step": 22225 + }, + { + "epoch": 1.64, + "learning_rate": 1.704294873654392e-05, + "loss": 1.0113, + "step": 22226 + }, + { + "epoch": 1.64, + "learning_rate": 1.7042665568966457e-05, + "loss": 0.9229, + "step": 22227 + }, + { + "epoch": 1.64, + "learning_rate": 1.704238239018421e-05, + "loss": 1.0537, + "step": 22228 + }, + { + "epoch": 1.64, + "learning_rate": 1.7042099200197638e-05, + "loss": 0.9836, + "step": 22229 + }, + { + "epoch": 1.64, + "learning_rate": 1.7041815999007183e-05, + "loss": 1.0223, + "step": 22230 + }, + { + "epoch": 1.64, + "learning_rate": 1.70415327866133e-05, + "loss": 1.0637, + "step": 22231 + }, + { + "epoch": 1.64, + "learning_rate": 1.7041249563016443e-05, + "loss": 1.041, + "step": 22232 + }, + { + "epoch": 1.64, + "learning_rate": 1.7040966328217056e-05, + "loss": 1.0002, + "step": 22233 + }, + { + "epoch": 1.64, + "learning_rate": 1.704068308221559e-05, + "loss": 0.9135, + "step": 22234 + }, + { + "epoch": 1.64, + "learning_rate": 1.70403998250125e-05, + "loss": 1.1265, + "step": 22235 + }, + { + "epoch": 1.64, + "learning_rate": 1.7040116556608237e-05, + "loss": 0.9428, + "step": 22236 + }, + { + "epoch": 1.64, + "learning_rate": 1.7039833277003243e-05, + "loss": 1.0451, + "step": 22237 + }, + { + "epoch": 1.64, + "learning_rate": 1.703954998619798e-05, + "loss": 1.0064, + "step": 22238 + }, + { + "epoch": 1.64, + "learning_rate": 1.7039266684192894e-05, + "loss": 1.0316, + "step": 22239 + }, + { + "epoch": 1.64, + "learning_rate": 1.7038983370988434e-05, + "loss": 1.0204, + "step": 22240 + }, + { + "epoch": 1.64, + "learning_rate": 1.7038700046585054e-05, + "loss": 1.0695, + "step": 22241 + }, + { + "epoch": 1.64, + "learning_rate": 1.70384167109832e-05, + "loss": 0.9302, + "step": 22242 + }, + { + "epoch": 1.64, + "learning_rate": 1.7038133364183326e-05, + "loss": 0.9499, + "step": 22243 + }, + { + "epoch": 1.64, + "learning_rate": 1.7037850006185883e-05, + "loss": 0.9783, + "step": 22244 + }, + { + "epoch": 1.64, + "learning_rate": 1.7037566636991322e-05, + "loss": 1.0055, + "step": 22245 + }, + { + "epoch": 1.64, + "learning_rate": 1.7037283256600094e-05, + "loss": 0.9577, + "step": 22246 + }, + { + "epoch": 1.64, + "learning_rate": 1.7036999865012648e-05, + "loss": 0.9678, + "step": 22247 + }, + { + "epoch": 1.64, + "learning_rate": 1.7036716462229433e-05, + "loss": 1.1159, + "step": 22248 + }, + { + "epoch": 1.64, + "learning_rate": 1.7036433048250905e-05, + "loss": 0.9343, + "step": 22249 + }, + { + "epoch": 1.64, + "learning_rate": 1.7036149623077516e-05, + "loss": 0.906, + "step": 22250 + }, + { + "epoch": 1.64, + "learning_rate": 1.7035866186709712e-05, + "loss": 1.0392, + "step": 22251 + }, + { + "epoch": 1.64, + "learning_rate": 1.7035582739147944e-05, + "loss": 0.9834, + "step": 22252 + }, + { + "epoch": 1.64, + "learning_rate": 1.7035299280392666e-05, + "loss": 1.0273, + "step": 22253 + }, + { + "epoch": 1.64, + "learning_rate": 1.7035015810444325e-05, + "loss": 1.1115, + "step": 22254 + }, + { + "epoch": 1.64, + "learning_rate": 1.7034732329303376e-05, + "loss": 1.0073, + "step": 22255 + }, + { + "epoch": 1.64, + "learning_rate": 1.703444883697027e-05, + "loss": 0.9487, + "step": 22256 + }, + { + "epoch": 1.64, + "learning_rate": 1.7034165333445454e-05, + "loss": 1.1365, + "step": 22257 + }, + { + "epoch": 1.64, + "learning_rate": 1.7033881818729383e-05, + "loss": 1.0116, + "step": 22258 + }, + { + "epoch": 1.64, + "learning_rate": 1.7033598292822505e-05, + "loss": 0.928, + "step": 22259 + }, + { + "epoch": 1.64, + "learning_rate": 1.7033314755725276e-05, + "loss": 1.0082, + "step": 22260 + }, + { + "epoch": 1.64, + "learning_rate": 1.703303120743814e-05, + "loss": 1.0629, + "step": 22261 + }, + { + "epoch": 1.64, + "learning_rate": 1.7032747647961555e-05, + "loss": 1.0524, + "step": 22262 + }, + { + "epoch": 1.64, + "learning_rate": 1.7032464077295966e-05, + "loss": 1.0082, + "step": 22263 + }, + { + "epoch": 1.65, + "learning_rate": 1.703218049544183e-05, + "loss": 1.0797, + "step": 22264 + }, + { + "epoch": 1.65, + "learning_rate": 1.7031896902399596e-05, + "loss": 1.0339, + "step": 22265 + }, + { + "epoch": 1.65, + "learning_rate": 1.7031613298169713e-05, + "loss": 1.0665, + "step": 22266 + }, + { + "epoch": 1.65, + "learning_rate": 1.7031329682752632e-05, + "loss": 1.0373, + "step": 22267 + }, + { + "epoch": 1.65, + "learning_rate": 1.7031046056148814e-05, + "loss": 0.9732, + "step": 22268 + }, + { + "epoch": 1.65, + "learning_rate": 1.7030762418358696e-05, + "loss": 0.9183, + "step": 22269 + }, + { + "epoch": 1.65, + "learning_rate": 1.7030478769382735e-05, + "loss": 0.9914, + "step": 22270 + }, + { + "epoch": 1.65, + "learning_rate": 1.7030195109221383e-05, + "loss": 1.0624, + "step": 22271 + }, + { + "epoch": 1.65, + "learning_rate": 1.702991143787509e-05, + "loss": 1.0453, + "step": 22272 + }, + { + "epoch": 1.65, + "learning_rate": 1.7029627755344315e-05, + "loss": 1.0071, + "step": 22273 + }, + { + "epoch": 1.65, + "learning_rate": 1.7029344061629497e-05, + "loss": 1.0289, + "step": 22274 + }, + { + "epoch": 1.65, + "learning_rate": 1.7029060356731094e-05, + "loss": 0.9444, + "step": 22275 + }, + { + "epoch": 1.65, + "learning_rate": 1.7028776640649554e-05, + "loss": 1.0627, + "step": 22276 + }, + { + "epoch": 1.65, + "learning_rate": 1.7028492913385337e-05, + "loss": 1.0586, + "step": 22277 + }, + { + "epoch": 1.65, + "learning_rate": 1.7028209174938886e-05, + "loss": 1.0612, + "step": 22278 + }, + { + "epoch": 1.65, + "learning_rate": 1.7027925425310654e-05, + "loss": 1.0187, + "step": 22279 + }, + { + "epoch": 1.65, + "learning_rate": 1.7027641664501093e-05, + "loss": 1.0977, + "step": 22280 + }, + { + "epoch": 1.65, + "learning_rate": 1.7027357892510656e-05, + "loss": 0.9692, + "step": 22281 + }, + { + "epoch": 1.65, + "learning_rate": 1.702707410933979e-05, + "loss": 1.0173, + "step": 22282 + }, + { + "epoch": 1.65, + "learning_rate": 1.702679031498895e-05, + "loss": 0.9704, + "step": 22283 + }, + { + "epoch": 1.65, + "learning_rate": 1.702650650945859e-05, + "loss": 1.0436, + "step": 22284 + }, + { + "epoch": 1.65, + "learning_rate": 1.7026222692749162e-05, + "loss": 0.9877, + "step": 22285 + }, + { + "epoch": 1.65, + "learning_rate": 1.702593886486111e-05, + "loss": 1.0032, + "step": 22286 + }, + { + "epoch": 1.65, + "learning_rate": 1.7025655025794886e-05, + "loss": 1.0322, + "step": 22287 + }, + { + "epoch": 1.65, + "learning_rate": 1.702537117555095e-05, + "loss": 1.0323, + "step": 22288 + }, + { + "epoch": 1.65, + "learning_rate": 1.702508731412975e-05, + "loss": 1.0077, + "step": 22289 + }, + { + "epoch": 1.65, + "learning_rate": 1.7024803441531734e-05, + "loss": 1.0006, + "step": 22290 + }, + { + "epoch": 1.65, + "learning_rate": 1.7024519557757357e-05, + "loss": 0.9843, + "step": 22291 + }, + { + "epoch": 1.65, + "learning_rate": 1.702423566280707e-05, + "loss": 1.0252, + "step": 22292 + }, + { + "epoch": 1.65, + "learning_rate": 1.7023951756681323e-05, + "loss": 0.9273, + "step": 22293 + }, + { + "epoch": 1.65, + "learning_rate": 1.702366783938057e-05, + "loss": 1.0195, + "step": 22294 + }, + { + "epoch": 1.65, + "learning_rate": 1.7023383910905263e-05, + "loss": 1.1058, + "step": 22295 + }, + { + "epoch": 1.65, + "learning_rate": 1.7023099971255855e-05, + "loss": 0.9546, + "step": 22296 + }, + { + "epoch": 1.65, + "learning_rate": 1.7022816020432792e-05, + "loss": 0.9513, + "step": 22297 + }, + { + "epoch": 1.65, + "learning_rate": 1.7022532058436528e-05, + "loss": 1.0295, + "step": 22298 + }, + { + "epoch": 1.65, + "learning_rate": 1.702224808526752e-05, + "loss": 1.0517, + "step": 22299 + }, + { + "epoch": 1.65, + "learning_rate": 1.702196410092621e-05, + "loss": 1.1252, + "step": 22300 + }, + { + "epoch": 1.65, + "learning_rate": 1.7021680105413063e-05, + "loss": 0.9381, + "step": 22301 + }, + { + "epoch": 1.65, + "learning_rate": 1.702139609872852e-05, + "loss": 0.9949, + "step": 22302 + }, + { + "epoch": 1.65, + "learning_rate": 1.7021112080873034e-05, + "loss": 1.0437, + "step": 22303 + }, + { + "epoch": 1.65, + "learning_rate": 1.702082805184706e-05, + "loss": 0.9782, + "step": 22304 + }, + { + "epoch": 1.65, + "learning_rate": 1.702054401165105e-05, + "loss": 0.7901, + "step": 22305 + }, + { + "epoch": 1.65, + "learning_rate": 1.7020259960285455e-05, + "loss": 0.9653, + "step": 22306 + }, + { + "epoch": 1.65, + "learning_rate": 1.701997589775073e-05, + "loss": 0.981, + "step": 22307 + }, + { + "epoch": 1.65, + "learning_rate": 1.701969182404732e-05, + "loss": 0.9643, + "step": 22308 + }, + { + "epoch": 1.65, + "learning_rate": 1.701940773917568e-05, + "loss": 0.9768, + "step": 22309 + }, + { + "epoch": 1.65, + "learning_rate": 1.7019123643136264e-05, + "loss": 0.9001, + "step": 22310 + }, + { + "epoch": 1.65, + "learning_rate": 1.7018839535929523e-05, + "loss": 1.0458, + "step": 22311 + }, + { + "epoch": 1.65, + "learning_rate": 1.701855541755591e-05, + "loss": 0.9388, + "step": 22312 + }, + { + "epoch": 1.65, + "learning_rate": 1.7018271288015874e-05, + "loss": 1.0322, + "step": 22313 + }, + { + "epoch": 1.65, + "learning_rate": 1.7017987147309868e-05, + "loss": 0.9159, + "step": 22314 + }, + { + "epoch": 1.65, + "learning_rate": 1.7017702995438344e-05, + "loss": 1.0527, + "step": 22315 + }, + { + "epoch": 1.65, + "learning_rate": 1.701741883240176e-05, + "loss": 1.0009, + "step": 22316 + }, + { + "epoch": 1.65, + "learning_rate": 1.7017134658200558e-05, + "loss": 0.9145, + "step": 22317 + }, + { + "epoch": 1.65, + "learning_rate": 1.7016850472835197e-05, + "loss": 1.0679, + "step": 22318 + }, + { + "epoch": 1.65, + "learning_rate": 1.7016566276306128e-05, + "loss": 0.9247, + "step": 22319 + }, + { + "epoch": 1.65, + "learning_rate": 1.70162820686138e-05, + "loss": 1.0584, + "step": 22320 + }, + { + "epoch": 1.65, + "learning_rate": 1.7015997849758668e-05, + "loss": 1.1061, + "step": 22321 + }, + { + "epoch": 1.65, + "learning_rate": 1.7015713619741186e-05, + "loss": 0.9549, + "step": 22322 + }, + { + "epoch": 1.65, + "learning_rate": 1.701542937856181e-05, + "loss": 1.0085, + "step": 22323 + }, + { + "epoch": 1.65, + "learning_rate": 1.7015145126220976e-05, + "loss": 1.0191, + "step": 22324 + }, + { + "epoch": 1.65, + "learning_rate": 1.701486086271915e-05, + "loss": 0.9842, + "step": 22325 + }, + { + "epoch": 1.65, + "learning_rate": 1.701457658805678e-05, + "loss": 1.0789, + "step": 22326 + }, + { + "epoch": 1.65, + "learning_rate": 1.701429230223432e-05, + "loss": 0.9915, + "step": 22327 + }, + { + "epoch": 1.65, + "learning_rate": 1.701400800525222e-05, + "loss": 1.0838, + "step": 22328 + }, + { + "epoch": 1.65, + "learning_rate": 1.7013723697110937e-05, + "loss": 1.0448, + "step": 22329 + }, + { + "epoch": 1.65, + "learning_rate": 1.7013439377810917e-05, + "loss": 1.0612, + "step": 22330 + }, + { + "epoch": 1.65, + "learning_rate": 1.7013155047352613e-05, + "loss": 1.0187, + "step": 22331 + }, + { + "epoch": 1.65, + "learning_rate": 1.7012870705736483e-05, + "loss": 1.0218, + "step": 22332 + }, + { + "epoch": 1.65, + "learning_rate": 1.7012586352962975e-05, + "loss": 0.9949, + "step": 22333 + }, + { + "epoch": 1.65, + "learning_rate": 1.7012301989032545e-05, + "loss": 1.0758, + "step": 22334 + }, + { + "epoch": 1.65, + "learning_rate": 1.701201761394564e-05, + "loss": 1.0141, + "step": 22335 + }, + { + "epoch": 1.65, + "learning_rate": 1.701173322770272e-05, + "loss": 1.0432, + "step": 22336 + }, + { + "epoch": 1.65, + "learning_rate": 1.7011448830304225e-05, + "loss": 1.1002, + "step": 22337 + }, + { + "epoch": 1.65, + "learning_rate": 1.701116442175062e-05, + "loss": 1.1401, + "step": 22338 + }, + { + "epoch": 1.65, + "learning_rate": 1.701088000204235e-05, + "loss": 0.9545, + "step": 22339 + }, + { + "epoch": 1.65, + "learning_rate": 1.7010595571179872e-05, + "loss": 0.9076, + "step": 22340 + }, + { + "epoch": 1.65, + "learning_rate": 1.701031112916364e-05, + "loss": 0.9399, + "step": 22341 + }, + { + "epoch": 1.65, + "learning_rate": 1.70100266759941e-05, + "loss": 0.9669, + "step": 22342 + }, + { + "epoch": 1.65, + "learning_rate": 1.7009742211671706e-05, + "loss": 1.0265, + "step": 22343 + }, + { + "epoch": 1.65, + "learning_rate": 1.7009457736196914e-05, + "loss": 1.0179, + "step": 22344 + }, + { + "epoch": 1.65, + "learning_rate": 1.7009173249570176e-05, + "loss": 1.0701, + "step": 22345 + }, + { + "epoch": 1.65, + "learning_rate": 1.7008888751791942e-05, + "loss": 1.1484, + "step": 22346 + }, + { + "epoch": 1.65, + "learning_rate": 1.700860424286267e-05, + "loss": 1.0113, + "step": 22347 + }, + { + "epoch": 1.65, + "learning_rate": 1.7008319722782805e-05, + "loss": 0.9802, + "step": 22348 + }, + { + "epoch": 1.65, + "learning_rate": 1.7008035191552804e-05, + "loss": 0.9751, + "step": 22349 + }, + { + "epoch": 1.65, + "learning_rate": 1.7007750649173122e-05, + "loss": 0.9497, + "step": 22350 + }, + { + "epoch": 1.65, + "learning_rate": 1.700746609564421e-05, + "loss": 1.0979, + "step": 22351 + }, + { + "epoch": 1.65, + "learning_rate": 1.7007181530966513e-05, + "loss": 1.0375, + "step": 22352 + }, + { + "epoch": 1.65, + "learning_rate": 1.7006896955140495e-05, + "loss": 1.104, + "step": 22353 + }, + { + "epoch": 1.65, + "learning_rate": 1.7006612368166605e-05, + "loss": 1.0486, + "step": 22354 + }, + { + "epoch": 1.65, + "learning_rate": 1.7006327770045293e-05, + "loss": 1.0086, + "step": 22355 + }, + { + "epoch": 1.65, + "learning_rate": 1.700604316077701e-05, + "loss": 0.9826, + "step": 22356 + }, + { + "epoch": 1.65, + "learning_rate": 1.700575854036222e-05, + "loss": 1.0714, + "step": 22357 + }, + { + "epoch": 1.65, + "learning_rate": 1.7005473908801366e-05, + "loss": 1.0704, + "step": 22358 + }, + { + "epoch": 1.65, + "learning_rate": 1.70051892660949e-05, + "loss": 1.0531, + "step": 22359 + }, + { + "epoch": 1.65, + "learning_rate": 1.700490461224328e-05, + "loss": 1.1342, + "step": 22360 + }, + { + "epoch": 1.65, + "learning_rate": 1.700461994724696e-05, + "loss": 1.0485, + "step": 22361 + }, + { + "epoch": 1.65, + "learning_rate": 1.7004335271106388e-05, + "loss": 1.0965, + "step": 22362 + }, + { + "epoch": 1.65, + "learning_rate": 1.7004050583822018e-05, + "loss": 1.0293, + "step": 22363 + }, + { + "epoch": 1.65, + "learning_rate": 1.7003765885394303e-05, + "loss": 1.1364, + "step": 22364 + }, + { + "epoch": 1.65, + "learning_rate": 1.7003481175823696e-05, + "loss": 0.8916, + "step": 22365 + }, + { + "epoch": 1.65, + "learning_rate": 1.7003196455110656e-05, + "loss": 1.012, + "step": 22366 + }, + { + "epoch": 1.65, + "learning_rate": 1.7002911723255622e-05, + "loss": 0.9662, + "step": 22367 + }, + { + "epoch": 1.65, + "learning_rate": 1.7002626980259063e-05, + "loss": 1.1975, + "step": 22368 + }, + { + "epoch": 1.65, + "learning_rate": 1.700234222612142e-05, + "loss": 0.9462, + "step": 22369 + }, + { + "epoch": 1.65, + "learning_rate": 1.7002057460843156e-05, + "loss": 1.0219, + "step": 22370 + }, + { + "epoch": 1.65, + "learning_rate": 1.7001772684424716e-05, + "loss": 1.1166, + "step": 22371 + }, + { + "epoch": 1.65, + "learning_rate": 1.7001487896866553e-05, + "loss": 1.0389, + "step": 22372 + }, + { + "epoch": 1.65, + "learning_rate": 1.7001203098169127e-05, + "loss": 1.0962, + "step": 22373 + }, + { + "epoch": 1.65, + "learning_rate": 1.7000918288332886e-05, + "loss": 0.9577, + "step": 22374 + }, + { + "epoch": 1.65, + "learning_rate": 1.7000633467358284e-05, + "loss": 0.9817, + "step": 22375 + }, + { + "epoch": 1.65, + "learning_rate": 1.7000348635245774e-05, + "loss": 0.9674, + "step": 22376 + }, + { + "epoch": 1.65, + "learning_rate": 1.7000063791995808e-05, + "loss": 1.0166, + "step": 22377 + }, + { + "epoch": 1.65, + "learning_rate": 1.6999778937608843e-05, + "loss": 0.9188, + "step": 22378 + }, + { + "epoch": 1.65, + "learning_rate": 1.699949407208533e-05, + "loss": 0.9288, + "step": 22379 + }, + { + "epoch": 1.65, + "learning_rate": 1.699920919542572e-05, + "loss": 1.0299, + "step": 22380 + }, + { + "epoch": 1.65, + "learning_rate": 1.699892430763047e-05, + "loss": 1.0396, + "step": 22381 + }, + { + "epoch": 1.65, + "learning_rate": 1.699863940870003e-05, + "loss": 0.9676, + "step": 22382 + }, + { + "epoch": 1.65, + "learning_rate": 1.6998354498634856e-05, + "loss": 1.0009, + "step": 22383 + }, + { + "epoch": 1.65, + "learning_rate": 1.69980695774354e-05, + "loss": 1.0311, + "step": 22384 + }, + { + "epoch": 1.65, + "learning_rate": 1.6997784645102117e-05, + "loss": 0.9405, + "step": 22385 + }, + { + "epoch": 1.65, + "learning_rate": 1.6997499701635454e-05, + "loss": 0.9265, + "step": 22386 + }, + { + "epoch": 1.65, + "learning_rate": 1.6997214747035872e-05, + "loss": 1.0426, + "step": 22387 + }, + { + "epoch": 1.65, + "learning_rate": 1.6996929781303823e-05, + "loss": 1.0138, + "step": 22388 + }, + { + "epoch": 1.65, + "learning_rate": 1.6996644804439756e-05, + "loss": 1.0401, + "step": 22389 + }, + { + "epoch": 1.65, + "learning_rate": 1.699635981644413e-05, + "loss": 1.0483, + "step": 22390 + }, + { + "epoch": 1.65, + "learning_rate": 1.6996074817317395e-05, + "loss": 1.0094, + "step": 22391 + }, + { + "epoch": 1.65, + "learning_rate": 1.6995789807060003e-05, + "loss": 0.956, + "step": 22392 + }, + { + "epoch": 1.65, + "learning_rate": 1.699550478567241e-05, + "loss": 1.031, + "step": 22393 + }, + { + "epoch": 1.65, + "learning_rate": 1.699521975315507e-05, + "loss": 1.0489, + "step": 22394 + }, + { + "epoch": 1.65, + "learning_rate": 1.6994934709508436e-05, + "loss": 0.945, + "step": 22395 + }, + { + "epoch": 1.65, + "learning_rate": 1.699464965473296e-05, + "loss": 0.9648, + "step": 22396 + }, + { + "epoch": 1.65, + "learning_rate": 1.6994364588829098e-05, + "loss": 0.9785, + "step": 22397 + }, + { + "epoch": 1.65, + "learning_rate": 1.69940795117973e-05, + "loss": 1.0521, + "step": 22398 + }, + { + "epoch": 1.66, + "learning_rate": 1.6993794423638023e-05, + "loss": 0.988, + "step": 22399 + }, + { + "epoch": 1.66, + "learning_rate": 1.6993509324351718e-05, + "loss": 1.0388, + "step": 22400 + }, + { + "epoch": 1.66, + "learning_rate": 1.6993224213938843e-05, + "loss": 1.0451, + "step": 22401 + }, + { + "epoch": 1.66, + "learning_rate": 1.6992939092399844e-05, + "loss": 0.922, + "step": 22402 + }, + { + "epoch": 1.66, + "learning_rate": 1.6992653959735183e-05, + "loss": 1.0796, + "step": 22403 + }, + { + "epoch": 1.66, + "learning_rate": 1.6992368815945303e-05, + "loss": 0.9401, + "step": 22404 + }, + { + "epoch": 1.66, + "learning_rate": 1.6992083661030668e-05, + "loss": 0.9636, + "step": 22405 + }, + { + "epoch": 1.66, + "learning_rate": 1.699179849499173e-05, + "loss": 1.0501, + "step": 22406 + }, + { + "epoch": 1.66, + "learning_rate": 1.6991513317828938e-05, + "loss": 1.0699, + "step": 22407 + }, + { + "epoch": 1.66, + "learning_rate": 1.6991228129542753e-05, + "loss": 0.9479, + "step": 22408 + }, + { + "epoch": 1.66, + "learning_rate": 1.6990942930133617e-05, + "loss": 0.9606, + "step": 22409 + }, + { + "epoch": 1.66, + "learning_rate": 1.6990657719601996e-05, + "loss": 0.9319, + "step": 22410 + }, + { + "epoch": 1.66, + "learning_rate": 1.6990372497948334e-05, + "loss": 1.0533, + "step": 22411 + }, + { + "epoch": 1.66, + "learning_rate": 1.6990087265173093e-05, + "loss": 1.0179, + "step": 22412 + }, + { + "epoch": 1.66, + "learning_rate": 1.6989802021276722e-05, + "loss": 0.964, + "step": 22413 + }, + { + "epoch": 1.66, + "learning_rate": 1.6989516766259676e-05, + "loss": 0.9216, + "step": 22414 + }, + { + "epoch": 1.66, + "learning_rate": 1.698923150012241e-05, + "loss": 1.0463, + "step": 22415 + }, + { + "epoch": 1.66, + "learning_rate": 1.6988946222865378e-05, + "loss": 1.0639, + "step": 22416 + }, + { + "epoch": 1.66, + "learning_rate": 1.6988660934489027e-05, + "loss": 0.9301, + "step": 22417 + }, + { + "epoch": 1.66, + "learning_rate": 1.698837563499382e-05, + "loss": 0.9797, + "step": 22418 + }, + { + "epoch": 1.66, + "learning_rate": 1.698809032438021e-05, + "loss": 1.0321, + "step": 22419 + }, + { + "epoch": 1.66, + "learning_rate": 1.6987805002648645e-05, + "loss": 0.965, + "step": 22420 + }, + { + "epoch": 1.66, + "learning_rate": 1.6987519669799582e-05, + "loss": 0.8892, + "step": 22421 + }, + { + "epoch": 1.66, + "learning_rate": 1.6987234325833475e-05, + "loss": 0.9867, + "step": 22422 + }, + { + "epoch": 1.66, + "learning_rate": 1.698694897075078e-05, + "loss": 1.0167, + "step": 22423 + }, + { + "epoch": 1.66, + "learning_rate": 1.6986663604551948e-05, + "loss": 0.9716, + "step": 22424 + }, + { + "epoch": 1.66, + "learning_rate": 1.6986378227237432e-05, + "loss": 1.0257, + "step": 22425 + }, + { + "epoch": 1.66, + "learning_rate": 1.698609283880769e-05, + "loss": 0.959, + "step": 22426 + }, + { + "epoch": 1.66, + "learning_rate": 1.6985807439263177e-05, + "loss": 0.9638, + "step": 22427 + }, + { + "epoch": 1.66, + "learning_rate": 1.6985522028604342e-05, + "loss": 0.9626, + "step": 22428 + }, + { + "epoch": 1.66, + "learning_rate": 1.6985236606831642e-05, + "loss": 0.9284, + "step": 22429 + }, + { + "epoch": 1.66, + "learning_rate": 1.6984951173945527e-05, + "loss": 1.0343, + "step": 22430 + }, + { + "epoch": 1.66, + "learning_rate": 1.698466572994646e-05, + "loss": 1.0593, + "step": 22431 + }, + { + "epoch": 1.66, + "learning_rate": 1.6984380274834888e-05, + "loss": 0.8974, + "step": 22432 + }, + { + "epoch": 1.66, + "learning_rate": 1.6984094808611264e-05, + "loss": 0.982, + "step": 22433 + }, + { + "epoch": 1.66, + "learning_rate": 1.698380933127605e-05, + "loss": 0.9538, + "step": 22434 + }, + { + "epoch": 1.66, + "learning_rate": 1.698352384282969e-05, + "loss": 1.0821, + "step": 22435 + }, + { + "epoch": 1.66, + "learning_rate": 1.698323834327265e-05, + "loss": 1.0786, + "step": 22436 + }, + { + "epoch": 1.66, + "learning_rate": 1.6982952832605373e-05, + "loss": 1.0492, + "step": 22437 + }, + { + "epoch": 1.66, + "learning_rate": 1.6982667310828317e-05, + "loss": 0.9221, + "step": 22438 + }, + { + "epoch": 1.66, + "learning_rate": 1.698238177794194e-05, + "loss": 1.1118, + "step": 22439 + }, + { + "epoch": 1.66, + "learning_rate": 1.6982096233946692e-05, + "loss": 0.9403, + "step": 22440 + }, + { + "epoch": 1.66, + "learning_rate": 1.6981810678843027e-05, + "loss": 0.9962, + "step": 22441 + }, + { + "epoch": 1.66, + "learning_rate": 1.6981525112631402e-05, + "loss": 0.9714, + "step": 22442 + }, + { + "epoch": 1.66, + "learning_rate": 1.6981239535312272e-05, + "loss": 0.9622, + "step": 22443 + }, + { + "epoch": 1.66, + "learning_rate": 1.698095394688609e-05, + "loss": 1.059, + "step": 22444 + }, + { + "epoch": 1.66, + "learning_rate": 1.698066834735331e-05, + "loss": 1.0086, + "step": 22445 + }, + { + "epoch": 1.66, + "learning_rate": 1.698038273671439e-05, + "loss": 1.0022, + "step": 22446 + }, + { + "epoch": 1.66, + "learning_rate": 1.6980097114969776e-05, + "loss": 1.0249, + "step": 22447 + }, + { + "epoch": 1.66, + "learning_rate": 1.697981148211993e-05, + "loss": 1.042, + "step": 22448 + }, + { + "epoch": 1.66, + "learning_rate": 1.69795258381653e-05, + "loss": 0.973, + "step": 22449 + }, + { + "epoch": 1.66, + "learning_rate": 1.6979240183106346e-05, + "loss": 1.0262, + "step": 22450 + }, + { + "epoch": 1.66, + "learning_rate": 1.6978954516943522e-05, + "loss": 0.9301, + "step": 22451 + }, + { + "epoch": 1.66, + "learning_rate": 1.697866883967728e-05, + "loss": 0.9888, + "step": 22452 + }, + { + "epoch": 1.66, + "learning_rate": 1.6978383151308077e-05, + "loss": 1.0647, + "step": 22453 + }, + { + "epoch": 1.66, + "learning_rate": 1.6978097451836367e-05, + "loss": 0.9491, + "step": 22454 + }, + { + "epoch": 1.66, + "learning_rate": 1.6977811741262604e-05, + "loss": 1.1132, + "step": 22455 + }, + { + "epoch": 1.66, + "learning_rate": 1.697752601958724e-05, + "loss": 1.0629, + "step": 22456 + }, + { + "epoch": 1.66, + "learning_rate": 1.697724028681073e-05, + "loss": 1.114, + "step": 22457 + }, + { + "epoch": 1.66, + "learning_rate": 1.697695454293353e-05, + "loss": 1.0268, + "step": 22458 + }, + { + "epoch": 1.66, + "learning_rate": 1.69766687879561e-05, + "loss": 1.0674, + "step": 22459 + }, + { + "epoch": 1.66, + "learning_rate": 1.697638302187889e-05, + "loss": 0.9908, + "step": 22460 + }, + { + "epoch": 1.66, + "learning_rate": 1.6976097244702354e-05, + "loss": 1.0681, + "step": 22461 + }, + { + "epoch": 1.66, + "learning_rate": 1.6975811456426944e-05, + "loss": 1.0274, + "step": 22462 + }, + { + "epoch": 1.66, + "learning_rate": 1.697552565705312e-05, + "loss": 0.987, + "step": 22463 + }, + { + "epoch": 1.66, + "learning_rate": 1.6975239846581333e-05, + "loss": 1.0571, + "step": 22464 + }, + { + "epoch": 1.66, + "learning_rate": 1.697495402501204e-05, + "loss": 1.036, + "step": 22465 + }, + { + "epoch": 1.66, + "learning_rate": 1.6974668192345696e-05, + "loss": 1.053, + "step": 22466 + }, + { + "epoch": 1.66, + "learning_rate": 1.697438234858275e-05, + "loss": 1.0731, + "step": 22467 + }, + { + "epoch": 1.66, + "learning_rate": 1.697409649372367e-05, + "loss": 0.9953, + "step": 22468 + }, + { + "epoch": 1.66, + "learning_rate": 1.6973810627768895e-05, + "loss": 1.0384, + "step": 22469 + }, + { + "epoch": 1.66, + "learning_rate": 1.697352475071889e-05, + "loss": 0.9455, + "step": 22470 + }, + { + "epoch": 1.66, + "learning_rate": 1.6973238862574107e-05, + "loss": 0.994, + "step": 22471 + }, + { + "epoch": 1.66, + "learning_rate": 1.6972952963335e-05, + "loss": 1.1487, + "step": 22472 + }, + { + "epoch": 1.66, + "learning_rate": 1.6972667053002024e-05, + "loss": 0.9173, + "step": 22473 + }, + { + "epoch": 1.66, + "learning_rate": 1.6972381131575633e-05, + "loss": 0.9645, + "step": 22474 + }, + { + "epoch": 1.66, + "learning_rate": 1.697209519905629e-05, + "loss": 1.0242, + "step": 22475 + }, + { + "epoch": 1.66, + "learning_rate": 1.697180925544444e-05, + "loss": 0.8909, + "step": 22476 + }, + { + "epoch": 1.66, + "learning_rate": 1.6971523300740537e-05, + "loss": 1.0833, + "step": 22477 + }, + { + "epoch": 1.66, + "learning_rate": 1.6971237334945043e-05, + "loss": 1.0324, + "step": 22478 + }, + { + "epoch": 1.66, + "learning_rate": 1.697095135805841e-05, + "loss": 1.0992, + "step": 22479 + }, + { + "epoch": 1.66, + "learning_rate": 1.6970665370081093e-05, + "loss": 0.9431, + "step": 22480 + }, + { + "epoch": 1.66, + "learning_rate": 1.6970379371013548e-05, + "loss": 0.9454, + "step": 22481 + }, + { + "epoch": 1.66, + "learning_rate": 1.6970093360856232e-05, + "loss": 1.0287, + "step": 22482 + }, + { + "epoch": 1.66, + "learning_rate": 1.696980733960959e-05, + "loss": 0.9909, + "step": 22483 + }, + { + "epoch": 1.66, + "learning_rate": 1.696952130727409e-05, + "loss": 0.9429, + "step": 22484 + }, + { + "epoch": 1.66, + "learning_rate": 1.696923526385018e-05, + "loss": 0.9676, + "step": 22485 + }, + { + "epoch": 1.66, + "learning_rate": 1.6968949209338315e-05, + "loss": 1.0078, + "step": 22486 + }, + { + "epoch": 1.66, + "learning_rate": 1.696866314373895e-05, + "loss": 1.0885, + "step": 22487 + }, + { + "epoch": 1.66, + "learning_rate": 1.696837706705255e-05, + "loss": 1.0292, + "step": 22488 + }, + { + "epoch": 1.66, + "learning_rate": 1.696809097927955e-05, + "loss": 0.9225, + "step": 22489 + }, + { + "epoch": 1.66, + "learning_rate": 1.6967804880420426e-05, + "loss": 1.0208, + "step": 22490 + }, + { + "epoch": 1.66, + "learning_rate": 1.696751877047562e-05, + "loss": 1.0362, + "step": 22491 + }, + { + "epoch": 1.66, + "learning_rate": 1.6967232649445592e-05, + "loss": 0.9791, + "step": 22492 + }, + { + "epoch": 1.66, + "learning_rate": 1.69669465173308e-05, + "loss": 0.9479, + "step": 22493 + }, + { + "epoch": 1.66, + "learning_rate": 1.696666037413169e-05, + "loss": 1.0391, + "step": 22494 + }, + { + "epoch": 1.66, + "learning_rate": 1.6966374219848726e-05, + "loss": 1.0232, + "step": 22495 + }, + { + "epoch": 1.66, + "learning_rate": 1.6966088054482363e-05, + "loss": 1.0387, + "step": 22496 + }, + { + "epoch": 1.66, + "learning_rate": 1.6965801878033047e-05, + "loss": 0.977, + "step": 22497 + }, + { + "epoch": 1.66, + "learning_rate": 1.6965515690501245e-05, + "loss": 1.0208, + "step": 22498 + }, + { + "epoch": 1.66, + "learning_rate": 1.6965229491887405e-05, + "loss": 0.9539, + "step": 22499 + }, + { + "epoch": 1.66, + "learning_rate": 1.6964943282191986e-05, + "loss": 0.9169, + "step": 22500 + }, + { + "epoch": 1.66, + "learning_rate": 1.6964657061415438e-05, + "loss": 1.0447, + "step": 22501 + }, + { + "epoch": 1.66, + "learning_rate": 1.6964370829558228e-05, + "loss": 0.8582, + "step": 22502 + }, + { + "epoch": 1.66, + "learning_rate": 1.6964084586620798e-05, + "loss": 1.001, + "step": 22503 + }, + { + "epoch": 1.66, + "learning_rate": 1.6963798332603607e-05, + "loss": 1.0171, + "step": 22504 + }, + { + "epoch": 1.66, + "learning_rate": 1.6963512067507117e-05, + "loss": 1.0316, + "step": 22505 + }, + { + "epoch": 1.66, + "learning_rate": 1.6963225791331778e-05, + "loss": 1.0882, + "step": 22506 + }, + { + "epoch": 1.66, + "learning_rate": 1.6962939504078045e-05, + "loss": 0.9234, + "step": 22507 + }, + { + "epoch": 1.66, + "learning_rate": 1.6962653205746374e-05, + "loss": 1.0124, + "step": 22508 + }, + { + "epoch": 1.66, + "learning_rate": 1.6962366896337226e-05, + "loss": 1.0057, + "step": 22509 + }, + { + "epoch": 1.66, + "learning_rate": 1.6962080575851046e-05, + "loss": 0.8934, + "step": 22510 + }, + { + "epoch": 1.66, + "learning_rate": 1.69617942442883e-05, + "loss": 0.952, + "step": 22511 + }, + { + "epoch": 1.66, + "learning_rate": 1.6961507901649438e-05, + "loss": 0.9263, + "step": 22512 + }, + { + "epoch": 1.66, + "learning_rate": 1.6961221547934918e-05, + "loss": 1.0103, + "step": 22513 + }, + { + "epoch": 1.66, + "learning_rate": 1.6960935183145187e-05, + "loss": 0.9874, + "step": 22514 + }, + { + "epoch": 1.66, + "learning_rate": 1.6960648807280715e-05, + "loss": 1.0157, + "step": 22515 + }, + { + "epoch": 1.66, + "learning_rate": 1.696036242034195e-05, + "loss": 1.0341, + "step": 22516 + }, + { + "epoch": 1.66, + "learning_rate": 1.6960076022329345e-05, + "loss": 1.0283, + "step": 22517 + }, + { + "epoch": 1.66, + "learning_rate": 1.695978961324336e-05, + "loss": 1.0782, + "step": 22518 + }, + { + "epoch": 1.66, + "learning_rate": 1.6959503193084448e-05, + "loss": 0.9093, + "step": 22519 + }, + { + "epoch": 1.66, + "learning_rate": 1.6959216761853067e-05, + "loss": 1.1017, + "step": 22520 + }, + { + "epoch": 1.66, + "learning_rate": 1.695893031954967e-05, + "loss": 1.0721, + "step": 22521 + }, + { + "epoch": 1.66, + "learning_rate": 1.6958643866174718e-05, + "loss": 1.0534, + "step": 22522 + }, + { + "epoch": 1.66, + "learning_rate": 1.695835740172866e-05, + "loss": 1.0462, + "step": 22523 + }, + { + "epoch": 1.66, + "learning_rate": 1.6958070926211955e-05, + "loss": 1.013, + "step": 22524 + }, + { + "epoch": 1.66, + "learning_rate": 1.6957784439625063e-05, + "loss": 0.8605, + "step": 22525 + }, + { + "epoch": 1.66, + "learning_rate": 1.6957497941968434e-05, + "loss": 1.0243, + "step": 22526 + }, + { + "epoch": 1.66, + "learning_rate": 1.695721143324252e-05, + "loss": 1.0174, + "step": 22527 + }, + { + "epoch": 1.66, + "learning_rate": 1.6956924913447787e-05, + "loss": 1.0832, + "step": 22528 + }, + { + "epoch": 1.66, + "learning_rate": 1.6956638382584684e-05, + "loss": 1.1346, + "step": 22529 + }, + { + "epoch": 1.66, + "learning_rate": 1.695635184065367e-05, + "loss": 1.0345, + "step": 22530 + }, + { + "epoch": 1.66, + "learning_rate": 1.69560652876552e-05, + "loss": 0.994, + "step": 22531 + }, + { + "epoch": 1.66, + "learning_rate": 1.6955778723589728e-05, + "loss": 1.0613, + "step": 22532 + }, + { + "epoch": 1.66, + "learning_rate": 1.695549214845771e-05, + "loss": 0.951, + "step": 22533 + }, + { + "epoch": 1.67, + "learning_rate": 1.695520556225961e-05, + "loss": 0.9777, + "step": 22534 + }, + { + "epoch": 1.67, + "learning_rate": 1.6954918964995873e-05, + "loss": 1.0435, + "step": 22535 + }, + { + "epoch": 1.67, + "learning_rate": 1.695463235666696e-05, + "loss": 0.9996, + "step": 22536 + }, + { + "epoch": 1.67, + "learning_rate": 1.6954345737273325e-05, + "loss": 1.1186, + "step": 22537 + }, + { + "epoch": 1.67, + "learning_rate": 1.6954059106815427e-05, + "loss": 1.0448, + "step": 22538 + }, + { + "epoch": 1.67, + "learning_rate": 1.6953772465293725e-05, + "loss": 0.9289, + "step": 22539 + }, + { + "epoch": 1.67, + "learning_rate": 1.6953485812708664e-05, + "loss": 0.9882, + "step": 22540 + }, + { + "epoch": 1.67, + "learning_rate": 1.695319914906071e-05, + "loss": 0.9519, + "step": 22541 + }, + { + "epoch": 1.67, + "learning_rate": 1.695291247435031e-05, + "loss": 0.953, + "step": 22542 + }, + { + "epoch": 1.67, + "learning_rate": 1.6952625788577932e-05, + "loss": 0.9582, + "step": 22543 + }, + { + "epoch": 1.67, + "learning_rate": 1.6952339091744023e-05, + "loss": 0.9788, + "step": 22544 + }, + { + "epoch": 1.67, + "learning_rate": 1.6952052383849042e-05, + "loss": 0.8689, + "step": 22545 + }, + { + "epoch": 1.67, + "learning_rate": 1.6951765664893444e-05, + "loss": 0.936, + "step": 22546 + }, + { + "epoch": 1.67, + "learning_rate": 1.695147893487769e-05, + "loss": 0.9918, + "step": 22547 + }, + { + "epoch": 1.67, + "learning_rate": 1.6951192193802232e-05, + "loss": 1.0305, + "step": 22548 + }, + { + "epoch": 1.67, + "learning_rate": 1.6950905441667527e-05, + "loss": 0.8958, + "step": 22549 + }, + { + "epoch": 1.67, + "learning_rate": 1.695061867847403e-05, + "loss": 0.9909, + "step": 22550 + }, + { + "epoch": 1.67, + "learning_rate": 1.6950331904222194e-05, + "loss": 1.0431, + "step": 22551 + }, + { + "epoch": 1.67, + "learning_rate": 1.695004511891248e-05, + "loss": 1.0353, + "step": 22552 + }, + { + "epoch": 1.67, + "learning_rate": 1.694975832254535e-05, + "loss": 1.0684, + "step": 22553 + }, + { + "epoch": 1.67, + "learning_rate": 1.694947151512125e-05, + "loss": 0.9799, + "step": 22554 + }, + { + "epoch": 1.67, + "learning_rate": 1.6949184696640638e-05, + "loss": 0.9633, + "step": 22555 + }, + { + "epoch": 1.67, + "learning_rate": 1.6948897867103977e-05, + "loss": 1.0141, + "step": 22556 + }, + { + "epoch": 1.67, + "learning_rate": 1.6948611026511717e-05, + "loss": 1.0428, + "step": 22557 + }, + { + "epoch": 1.67, + "learning_rate": 1.6948324174864318e-05, + "loss": 1.0215, + "step": 22558 + }, + { + "epoch": 1.67, + "learning_rate": 1.694803731216223e-05, + "loss": 1.0897, + "step": 22559 + }, + { + "epoch": 1.67, + "learning_rate": 1.6947750438405917e-05, + "loss": 1.0051, + "step": 22560 + }, + { + "epoch": 1.67, + "learning_rate": 1.6947463553595837e-05, + "loss": 0.9393, + "step": 22561 + }, + { + "epoch": 1.67, + "learning_rate": 1.6947176657732434e-05, + "loss": 1.0468, + "step": 22562 + }, + { + "epoch": 1.67, + "learning_rate": 1.6946889750816176e-05, + "loss": 0.9154, + "step": 22563 + }, + { + "epoch": 1.67, + "learning_rate": 1.6946602832847517e-05, + "loss": 0.9571, + "step": 22564 + }, + { + "epoch": 1.67, + "learning_rate": 1.6946315903826908e-05, + "loss": 1.0086, + "step": 22565 + }, + { + "epoch": 1.67, + "learning_rate": 1.6946028963754812e-05, + "loss": 1.0071, + "step": 22566 + }, + { + "epoch": 1.67, + "learning_rate": 1.6945742012631688e-05, + "loss": 1.0009, + "step": 22567 + }, + { + "epoch": 1.67, + "learning_rate": 1.6945455050457982e-05, + "loss": 1.0374, + "step": 22568 + }, + { + "epoch": 1.67, + "learning_rate": 1.694516807723416e-05, + "loss": 1.0759, + "step": 22569 + }, + { + "epoch": 1.67, + "learning_rate": 1.694488109296067e-05, + "loss": 1.1554, + "step": 22570 + }, + { + "epoch": 1.67, + "learning_rate": 1.6944594097637976e-05, + "loss": 1.0707, + "step": 22571 + }, + { + "epoch": 1.67, + "learning_rate": 1.6944307091266532e-05, + "loss": 1.0381, + "step": 22572 + }, + { + "epoch": 1.67, + "learning_rate": 1.6944020073846796e-05, + "loss": 0.8802, + "step": 22573 + }, + { + "epoch": 1.67, + "learning_rate": 1.6943733045379223e-05, + "loss": 1.0043, + "step": 22574 + }, + { + "epoch": 1.67, + "learning_rate": 1.694344600586427e-05, + "loss": 0.969, + "step": 22575 + }, + { + "epoch": 1.67, + "learning_rate": 1.6943158955302395e-05, + "loss": 0.9688, + "step": 22576 + }, + { + "epoch": 1.67, + "learning_rate": 1.694287189369405e-05, + "loss": 1.0455, + "step": 22577 + }, + { + "epoch": 1.67, + "learning_rate": 1.6942584821039697e-05, + "loss": 0.9582, + "step": 22578 + }, + { + "epoch": 1.67, + "learning_rate": 1.6942297737339792e-05, + "loss": 0.9124, + "step": 22579 + }, + { + "epoch": 1.67, + "learning_rate": 1.6942010642594788e-05, + "loss": 1.0218, + "step": 22580 + }, + { + "epoch": 1.67, + "learning_rate": 1.694172353680515e-05, + "loss": 1.0421, + "step": 22581 + }, + { + "epoch": 1.67, + "learning_rate": 1.6941436419971324e-05, + "loss": 1.0058, + "step": 22582 + }, + { + "epoch": 1.67, + "learning_rate": 1.6941149292093772e-05, + "loss": 0.8453, + "step": 22583 + }, + { + "epoch": 1.67, + "learning_rate": 1.694086215317295e-05, + "loss": 1.0161, + "step": 22584 + }, + { + "epoch": 1.67, + "learning_rate": 1.6940575003209323e-05, + "loss": 1.0268, + "step": 22585 + }, + { + "epoch": 1.67, + "learning_rate": 1.6940287842203334e-05, + "loss": 1.0595, + "step": 22586 + }, + { + "epoch": 1.67, + "learning_rate": 1.6940000670155446e-05, + "loss": 0.9552, + "step": 22587 + }, + { + "epoch": 1.67, + "learning_rate": 1.693971348706612e-05, + "loss": 1.0675, + "step": 22588 + }, + { + "epoch": 1.67, + "learning_rate": 1.6939426292935805e-05, + "loss": 1.0346, + "step": 22589 + }, + { + "epoch": 1.67, + "learning_rate": 1.6939139087764965e-05, + "loss": 1.0305, + "step": 22590 + }, + { + "epoch": 1.67, + "learning_rate": 1.693885187155405e-05, + "loss": 0.9937, + "step": 22591 + }, + { + "epoch": 1.67, + "learning_rate": 1.6938564644303525e-05, + "loss": 1.0473, + "step": 22592 + }, + { + "epoch": 1.67, + "learning_rate": 1.6938277406013843e-05, + "loss": 0.927, + "step": 22593 + }, + { + "epoch": 1.67, + "learning_rate": 1.693799015668546e-05, + "loss": 1.042, + "step": 22594 + }, + { + "epoch": 1.67, + "learning_rate": 1.6937702896318834e-05, + "loss": 1.1238, + "step": 22595 + }, + { + "epoch": 1.67, + "learning_rate": 1.6937415624914423e-05, + "loss": 1.0018, + "step": 22596 + }, + { + "epoch": 1.67, + "learning_rate": 1.693712834247268e-05, + "loss": 0.9742, + "step": 22597 + }, + { + "epoch": 1.67, + "learning_rate": 1.693684104899407e-05, + "loss": 1.0057, + "step": 22598 + }, + { + "epoch": 1.67, + "learning_rate": 1.6936553744479037e-05, + "loss": 0.9972, + "step": 22599 + }, + { + "epoch": 1.67, + "learning_rate": 1.693626642892805e-05, + "loss": 1.0146, + "step": 22600 + }, + { + "epoch": 1.67, + "learning_rate": 1.6935979102341568e-05, + "loss": 0.9629, + "step": 22601 + }, + { + "epoch": 1.67, + "learning_rate": 1.6935691764720034e-05, + "loss": 0.93, + "step": 22602 + }, + { + "epoch": 1.67, + "learning_rate": 1.693540441606392e-05, + "loss": 1.0548, + "step": 22603 + }, + { + "epoch": 1.67, + "learning_rate": 1.6935117056373674e-05, + "loss": 0.9782, + "step": 22604 + }, + { + "epoch": 1.67, + "learning_rate": 1.6934829685649757e-05, + "loss": 1.0857, + "step": 22605 + }, + { + "epoch": 1.67, + "learning_rate": 1.6934542303892626e-05, + "loss": 1.0163, + "step": 22606 + }, + { + "epoch": 1.67, + "learning_rate": 1.6934254911102735e-05, + "loss": 0.9557, + "step": 22607 + }, + { + "epoch": 1.67, + "learning_rate": 1.6933967507280545e-05, + "loss": 0.9431, + "step": 22608 + }, + { + "epoch": 1.67, + "learning_rate": 1.693368009242651e-05, + "loss": 0.9305, + "step": 22609 + }, + { + "epoch": 1.67, + "learning_rate": 1.693339266654109e-05, + "loss": 1.0137, + "step": 22610 + }, + { + "epoch": 1.67, + "learning_rate": 1.6933105229624743e-05, + "loss": 1.0823, + "step": 22611 + }, + { + "epoch": 1.67, + "learning_rate": 1.6932817781677923e-05, + "loss": 1.115, + "step": 22612 + }, + { + "epoch": 1.67, + "learning_rate": 1.6932530322701094e-05, + "loss": 0.9407, + "step": 22613 + }, + { + "epoch": 1.67, + "learning_rate": 1.6932242852694704e-05, + "loss": 0.9782, + "step": 22614 + }, + { + "epoch": 1.67, + "learning_rate": 1.6931955371659216e-05, + "loss": 0.9649, + "step": 22615 + }, + { + "epoch": 1.67, + "learning_rate": 1.693166787959508e-05, + "loss": 1.0586, + "step": 22616 + }, + { + "epoch": 1.67, + "learning_rate": 1.6931380376502768e-05, + "loss": 1.0348, + "step": 22617 + }, + { + "epoch": 1.67, + "learning_rate": 1.6931092862382725e-05, + "loss": 0.9884, + "step": 22618 + }, + { + "epoch": 1.67, + "learning_rate": 1.693080533723541e-05, + "loss": 1.0469, + "step": 22619 + }, + { + "epoch": 1.67, + "learning_rate": 1.6930517801061288e-05, + "loss": 0.9332, + "step": 22620 + }, + { + "epoch": 1.67, + "learning_rate": 1.693023025386081e-05, + "loss": 1.0748, + "step": 22621 + }, + { + "epoch": 1.67, + "learning_rate": 1.6929942695634434e-05, + "loss": 1.0284, + "step": 22622 + }, + { + "epoch": 1.67, + "learning_rate": 1.692965512638262e-05, + "loss": 1.0142, + "step": 22623 + }, + { + "epoch": 1.67, + "learning_rate": 1.6929367546105825e-05, + "loss": 1.0237, + "step": 22624 + }, + { + "epoch": 1.67, + "learning_rate": 1.6929079954804503e-05, + "loss": 1.0307, + "step": 22625 + }, + { + "epoch": 1.67, + "learning_rate": 1.692879235247911e-05, + "loss": 1.0098, + "step": 22626 + }, + { + "epoch": 1.67, + "learning_rate": 1.6928504739130115e-05, + "loss": 1.0033, + "step": 22627 + }, + { + "epoch": 1.67, + "learning_rate": 1.6928217114757963e-05, + "loss": 1.1051, + "step": 22628 + }, + { + "epoch": 1.67, + "learning_rate": 1.6927929479363118e-05, + "loss": 1.0248, + "step": 22629 + }, + { + "epoch": 1.67, + "learning_rate": 1.6927641832946036e-05, + "loss": 1.0545, + "step": 22630 + }, + { + "epoch": 1.67, + "learning_rate": 1.6927354175507176e-05, + "loss": 1.0242, + "step": 22631 + }, + { + "epoch": 1.67, + "learning_rate": 1.6927066507046993e-05, + "loss": 0.9343, + "step": 22632 + }, + { + "epoch": 1.67, + "learning_rate": 1.692677882756595e-05, + "loss": 0.9178, + "step": 22633 + }, + { + "epoch": 1.67, + "learning_rate": 1.6926491137064496e-05, + "loss": 1.016, + "step": 22634 + }, + { + "epoch": 1.67, + "learning_rate": 1.6926203435543097e-05, + "loss": 0.9612, + "step": 22635 + }, + { + "epoch": 1.67, + "learning_rate": 1.692591572300221e-05, + "loss": 1.0641, + "step": 22636 + }, + { + "epoch": 1.67, + "learning_rate": 1.6925627999442286e-05, + "loss": 1.1004, + "step": 22637 + }, + { + "epoch": 1.67, + "learning_rate": 1.6925340264863785e-05, + "loss": 0.9839, + "step": 22638 + }, + { + "epoch": 1.67, + "learning_rate": 1.6925052519267172e-05, + "loss": 1.0263, + "step": 22639 + }, + { + "epoch": 1.67, + "learning_rate": 1.6924764762652893e-05, + "loss": 0.9687, + "step": 22640 + }, + { + "epoch": 1.67, + "learning_rate": 1.692447699502142e-05, + "loss": 1.0214, + "step": 22641 + }, + { + "epoch": 1.67, + "learning_rate": 1.69241892163732e-05, + "loss": 0.9044, + "step": 22642 + }, + { + "epoch": 1.67, + "learning_rate": 1.6923901426708692e-05, + "loss": 0.9906, + "step": 22643 + }, + { + "epoch": 1.67, + "learning_rate": 1.6923613626028357e-05, + "loss": 0.9446, + "step": 22644 + }, + { + "epoch": 1.67, + "learning_rate": 1.692332581433265e-05, + "loss": 1.0747, + "step": 22645 + }, + { + "epoch": 1.67, + "learning_rate": 1.6923037991622035e-05, + "loss": 0.9657, + "step": 22646 + }, + { + "epoch": 1.67, + "learning_rate": 1.6922750157896963e-05, + "loss": 1.0094, + "step": 22647 + }, + { + "epoch": 1.67, + "learning_rate": 1.6922462313157895e-05, + "loss": 1.0761, + "step": 22648 + }, + { + "epoch": 1.67, + "learning_rate": 1.692217445740529e-05, + "loss": 1.1025, + "step": 22649 + }, + { + "epoch": 1.67, + "learning_rate": 1.69218865906396e-05, + "loss": 1.0835, + "step": 22650 + }, + { + "epoch": 1.67, + "learning_rate": 1.6921598712861294e-05, + "loss": 0.9247, + "step": 22651 + }, + { + "epoch": 1.67, + "learning_rate": 1.692131082407082e-05, + "loss": 0.9333, + "step": 22652 + }, + { + "epoch": 1.67, + "learning_rate": 1.692102292426864e-05, + "loss": 1.0455, + "step": 22653 + }, + { + "epoch": 1.67, + "learning_rate": 1.692073501345521e-05, + "loss": 1.149, + "step": 22654 + }, + { + "epoch": 1.67, + "learning_rate": 1.692044709163099e-05, + "loss": 1.074, + "step": 22655 + }, + { + "epoch": 1.67, + "learning_rate": 1.692015915879644e-05, + "loss": 0.9889, + "step": 22656 + }, + { + "epoch": 1.67, + "learning_rate": 1.6919871214952013e-05, + "loss": 1.0338, + "step": 22657 + }, + { + "epoch": 1.67, + "learning_rate": 1.691958326009817e-05, + "loss": 0.963, + "step": 22658 + }, + { + "epoch": 1.67, + "learning_rate": 1.691929529423537e-05, + "loss": 1.0167, + "step": 22659 + }, + { + "epoch": 1.67, + "learning_rate": 1.6919007317364073e-05, + "loss": 1.0473, + "step": 22660 + }, + { + "epoch": 1.67, + "learning_rate": 1.691871932948473e-05, + "loss": 1.0144, + "step": 22661 + }, + { + "epoch": 1.67, + "learning_rate": 1.6918431330597805e-05, + "loss": 1.0198, + "step": 22662 + }, + { + "epoch": 1.67, + "learning_rate": 1.6918143320703754e-05, + "loss": 1.0584, + "step": 22663 + }, + { + "epoch": 1.67, + "learning_rate": 1.6917855299803036e-05, + "loss": 1.0699, + "step": 22664 + }, + { + "epoch": 1.67, + "learning_rate": 1.691756726789611e-05, + "loss": 0.9015, + "step": 22665 + }, + { + "epoch": 1.67, + "learning_rate": 1.6917279224983434e-05, + "loss": 1.0073, + "step": 22666 + }, + { + "epoch": 1.67, + "learning_rate": 1.6916991171065466e-05, + "loss": 1.057, + "step": 22667 + }, + { + "epoch": 1.67, + "learning_rate": 1.6916703106142663e-05, + "loss": 1.0806, + "step": 22668 + }, + { + "epoch": 1.67, + "learning_rate": 1.6916415030215482e-05, + "loss": 0.8879, + "step": 22669 + }, + { + "epoch": 1.68, + "learning_rate": 1.6916126943284388e-05, + "loss": 1.1327, + "step": 22670 + }, + { + "epoch": 1.68, + "learning_rate": 1.6915838845349834e-05, + "loss": 1.0051, + "step": 22671 + }, + { + "epoch": 1.68, + "learning_rate": 1.691555073641228e-05, + "loss": 0.9914, + "step": 22672 + }, + { + "epoch": 1.68, + "learning_rate": 1.6915262616472178e-05, + "loss": 0.8947, + "step": 22673 + }, + { + "epoch": 1.68, + "learning_rate": 1.6914974485529997e-05, + "loss": 1.0837, + "step": 22674 + }, + { + "epoch": 1.68, + "learning_rate": 1.691468634358619e-05, + "loss": 0.9259, + "step": 22675 + }, + { + "epoch": 1.68, + "learning_rate": 1.6914398190641217e-05, + "loss": 0.9494, + "step": 22676 + }, + { + "epoch": 1.68, + "learning_rate": 1.6914110026695533e-05, + "loss": 0.9079, + "step": 22677 + }, + { + "epoch": 1.68, + "learning_rate": 1.6913821851749597e-05, + "loss": 1.046, + "step": 22678 + }, + { + "epoch": 1.68, + "learning_rate": 1.6913533665803875e-05, + "loss": 1.0729, + "step": 22679 + }, + { + "epoch": 1.68, + "learning_rate": 1.6913245468858817e-05, + "loss": 1.0161, + "step": 22680 + }, + { + "epoch": 1.68, + "learning_rate": 1.6912957260914887e-05, + "loss": 1.079, + "step": 22681 + }, + { + "epoch": 1.68, + "learning_rate": 1.691266904197254e-05, + "loss": 0.9885, + "step": 22682 + }, + { + "epoch": 1.68, + "learning_rate": 1.691238081203223e-05, + "loss": 0.9659, + "step": 22683 + }, + { + "epoch": 1.68, + "learning_rate": 1.6912092571094426e-05, + "loss": 1.085, + "step": 22684 + }, + { + "epoch": 1.68, + "learning_rate": 1.6911804319159582e-05, + "loss": 1.0094, + "step": 22685 + }, + { + "epoch": 1.68, + "learning_rate": 1.6911516056228154e-05, + "loss": 1.0278, + "step": 22686 + }, + { + "epoch": 1.68, + "learning_rate": 1.6911227782300605e-05, + "loss": 1.0226, + "step": 22687 + }, + { + "epoch": 1.68, + "learning_rate": 1.691093949737739e-05, + "loss": 1.0384, + "step": 22688 + }, + { + "epoch": 1.68, + "learning_rate": 1.691065120145897e-05, + "loss": 1.0407, + "step": 22689 + }, + { + "epoch": 1.68, + "learning_rate": 1.69103628945458e-05, + "loss": 0.9195, + "step": 22690 + }, + { + "epoch": 1.68, + "learning_rate": 1.6910074576638345e-05, + "loss": 1.0353, + "step": 22691 + }, + { + "epoch": 1.68, + "learning_rate": 1.690978624773706e-05, + "loss": 1.0411, + "step": 22692 + }, + { + "epoch": 1.68, + "learning_rate": 1.69094979078424e-05, + "loss": 1.07, + "step": 22693 + }, + { + "epoch": 1.68, + "learning_rate": 1.690920955695483e-05, + "loss": 0.9918, + "step": 22694 + }, + { + "epoch": 1.68, + "learning_rate": 1.690892119507481e-05, + "loss": 1.1533, + "step": 22695 + }, + { + "epoch": 1.68, + "learning_rate": 1.690863282220279e-05, + "loss": 1.067, + "step": 22696 + }, + { + "epoch": 1.68, + "learning_rate": 1.6908344438339238e-05, + "loss": 1.1291, + "step": 22697 + }, + { + "epoch": 1.68, + "learning_rate": 1.690805604348461e-05, + "loss": 1.1765, + "step": 22698 + }, + { + "epoch": 1.68, + "learning_rate": 1.690776763763936e-05, + "loss": 1.0325, + "step": 22699 + }, + { + "epoch": 1.68, + "learning_rate": 1.690747922080395e-05, + "loss": 1.1328, + "step": 22700 + }, + { + "epoch": 1.68, + "learning_rate": 1.690719079297884e-05, + "loss": 0.9825, + "step": 22701 + }, + { + "epoch": 1.68, + "learning_rate": 1.6906902354164487e-05, + "loss": 1.0011, + "step": 22702 + }, + { + "epoch": 1.68, + "learning_rate": 1.6906613904361354e-05, + "loss": 0.9739, + "step": 22703 + }, + { + "epoch": 1.68, + "learning_rate": 1.6906325443569897e-05, + "loss": 1.0864, + "step": 22704 + }, + { + "epoch": 1.68, + "learning_rate": 1.6906036971790572e-05, + "loss": 0.9414, + "step": 22705 + }, + { + "epoch": 1.68, + "learning_rate": 1.6905748489023845e-05, + "loss": 1.0351, + "step": 22706 + }, + { + "epoch": 1.68, + "learning_rate": 1.690545999527017e-05, + "loss": 1.0318, + "step": 22707 + }, + { + "epoch": 1.68, + "learning_rate": 1.6905171490530006e-05, + "loss": 1.0019, + "step": 22708 + }, + { + "epoch": 1.68, + "learning_rate": 1.690488297480381e-05, + "loss": 0.9812, + "step": 22709 + }, + { + "epoch": 1.68, + "learning_rate": 1.6904594448092048e-05, + "loss": 0.9318, + "step": 22710 + }, + { + "epoch": 1.68, + "learning_rate": 1.6904305910395176e-05, + "loss": 1.0522, + "step": 22711 + }, + { + "epoch": 1.68, + "learning_rate": 1.690401736171365e-05, + "loss": 1.1238, + "step": 22712 + }, + { + "epoch": 1.68, + "learning_rate": 1.690372880204793e-05, + "loss": 1.1149, + "step": 22713 + }, + { + "epoch": 1.68, + "learning_rate": 1.690344023139848e-05, + "loss": 0.9905, + "step": 22714 + }, + { + "epoch": 1.68, + "learning_rate": 1.6903151649765754e-05, + "loss": 1.0174, + "step": 22715 + }, + { + "epoch": 1.68, + "learning_rate": 1.690286305715021e-05, + "loss": 0.9995, + "step": 22716 + }, + { + "epoch": 1.68, + "learning_rate": 1.6902574453552308e-05, + "loss": 0.9593, + "step": 22717 + }, + { + "epoch": 1.68, + "learning_rate": 1.6902285838972515e-05, + "loss": 1.021, + "step": 22718 + }, + { + "epoch": 1.68, + "learning_rate": 1.690199721341128e-05, + "loss": 0.9595, + "step": 22719 + }, + { + "epoch": 1.68, + "learning_rate": 1.6901708576869068e-05, + "loss": 1.0087, + "step": 22720 + }, + { + "epoch": 1.68, + "learning_rate": 1.6901419929346334e-05, + "loss": 0.9454, + "step": 22721 + }, + { + "epoch": 1.68, + "learning_rate": 1.690113127084354e-05, + "loss": 0.9104, + "step": 22722 + }, + { + "epoch": 1.68, + "learning_rate": 1.690084260136115e-05, + "loss": 0.9641, + "step": 22723 + }, + { + "epoch": 1.68, + "learning_rate": 1.690055392089961e-05, + "loss": 0.8827, + "step": 22724 + }, + { + "epoch": 1.68, + "learning_rate": 1.6900265229459395e-05, + "loss": 0.9813, + "step": 22725 + }, + { + "epoch": 1.68, + "learning_rate": 1.6899976527040954e-05, + "loss": 0.9999, + "step": 22726 + }, + { + "epoch": 1.68, + "learning_rate": 1.6899687813644746e-05, + "loss": 1.0041, + "step": 22727 + }, + { + "epoch": 1.68, + "learning_rate": 1.6899399089271235e-05, + "loss": 0.9382, + "step": 22728 + }, + { + "epoch": 1.68, + "learning_rate": 1.689911035392088e-05, + "loss": 1.0121, + "step": 22729 + }, + { + "epoch": 1.68, + "learning_rate": 1.6898821607594137e-05, + "loss": 1.0504, + "step": 22730 + }, + { + "epoch": 1.68, + "learning_rate": 1.6898532850291468e-05, + "loss": 1.0333, + "step": 22731 + }, + { + "epoch": 1.68, + "learning_rate": 1.6898244082013336e-05, + "loss": 1.0331, + "step": 22732 + }, + { + "epoch": 1.68, + "learning_rate": 1.6897955302760194e-05, + "loss": 1.0325, + "step": 22733 + }, + { + "epoch": 1.68, + "learning_rate": 1.6897666512532504e-05, + "loss": 1.0349, + "step": 22734 + }, + { + "epoch": 1.68, + "learning_rate": 1.6897377711330726e-05, + "loss": 1.0138, + "step": 22735 + }, + { + "epoch": 1.68, + "learning_rate": 1.6897088899155315e-05, + "loss": 1.1028, + "step": 22736 + }, + { + "epoch": 1.68, + "learning_rate": 1.6896800076006737e-05, + "loss": 1.0692, + "step": 22737 + }, + { + "epoch": 1.68, + "learning_rate": 1.689651124188545e-05, + "loss": 1.0079, + "step": 22738 + }, + { + "epoch": 1.68, + "learning_rate": 1.689622239679191e-05, + "loss": 1.0237, + "step": 22739 + }, + { + "epoch": 1.68, + "learning_rate": 1.6895933540726578e-05, + "loss": 1.0303, + "step": 22740 + }, + { + "epoch": 1.68, + "learning_rate": 1.6895644673689916e-05, + "loss": 1.1384, + "step": 22741 + }, + { + "epoch": 1.68, + "learning_rate": 1.6895355795682382e-05, + "loss": 1.0185, + "step": 22742 + }, + { + "epoch": 1.68, + "learning_rate": 1.6895066906704434e-05, + "loss": 1.0444, + "step": 22743 + }, + { + "epoch": 1.68, + "learning_rate": 1.6894778006756537e-05, + "loss": 1.0732, + "step": 22744 + }, + { + "epoch": 1.68, + "learning_rate": 1.6894489095839144e-05, + "loss": 0.9991, + "step": 22745 + }, + { + "epoch": 1.68, + "learning_rate": 1.689420017395272e-05, + "loss": 1.0428, + "step": 22746 + }, + { + "epoch": 1.68, + "learning_rate": 1.6893911241097718e-05, + "loss": 1.0377, + "step": 22747 + }, + { + "epoch": 1.68, + "learning_rate": 1.6893622297274602e-05, + "loss": 1.0376, + "step": 22748 + }, + { + "epoch": 1.68, + "learning_rate": 1.6893333342483837e-05, + "loss": 0.944, + "step": 22749 + }, + { + "epoch": 1.68, + "learning_rate": 1.689304437672587e-05, + "loss": 1.0279, + "step": 22750 + }, + { + "epoch": 1.68, + "learning_rate": 1.6892755400001174e-05, + "loss": 1.01, + "step": 22751 + }, + { + "epoch": 1.68, + "learning_rate": 1.68924664123102e-05, + "loss": 0.8682, + "step": 22752 + }, + { + "epoch": 1.68, + "learning_rate": 1.6892177413653413e-05, + "loss": 0.9686, + "step": 22753 + }, + { + "epoch": 1.68, + "learning_rate": 1.689188840403127e-05, + "loss": 1.0384, + "step": 22754 + }, + { + "epoch": 1.68, + "learning_rate": 1.689159938344423e-05, + "loss": 0.9653, + "step": 22755 + }, + { + "epoch": 1.68, + "learning_rate": 1.6891310351892753e-05, + "loss": 1.0228, + "step": 22756 + }, + { + "epoch": 1.68, + "learning_rate": 1.68910213093773e-05, + "loss": 0.9239, + "step": 22757 + }, + { + "epoch": 1.68, + "learning_rate": 1.6890732255898328e-05, + "loss": 0.9453, + "step": 22758 + }, + { + "epoch": 1.68, + "learning_rate": 1.6890443191456305e-05, + "loss": 1.0025, + "step": 22759 + }, + { + "epoch": 1.68, + "learning_rate": 1.6890154116051683e-05, + "loss": 1.1808, + "step": 22760 + }, + { + "epoch": 1.68, + "learning_rate": 1.6889865029684923e-05, + "loss": 1.0009, + "step": 22761 + }, + { + "epoch": 1.68, + "learning_rate": 1.6889575932356487e-05, + "loss": 1.0042, + "step": 22762 + }, + { + "epoch": 1.68, + "learning_rate": 1.6889286824066838e-05, + "loss": 1.0602, + "step": 22763 + }, + { + "epoch": 1.68, + "learning_rate": 1.688899770481643e-05, + "loss": 0.9886, + "step": 22764 + }, + { + "epoch": 1.68, + "learning_rate": 1.6888708574605723e-05, + "loss": 0.9667, + "step": 22765 + }, + { + "epoch": 1.68, + "learning_rate": 1.688841943343518e-05, + "loss": 0.9841, + "step": 22766 + }, + { + "epoch": 1.68, + "learning_rate": 1.6888130281305263e-05, + "loss": 0.979, + "step": 22767 + }, + { + "epoch": 1.68, + "learning_rate": 1.6887841118216424e-05, + "loss": 1.088, + "step": 22768 + }, + { + "epoch": 1.68, + "learning_rate": 1.6887551944169134e-05, + "loss": 1.1225, + "step": 22769 + }, + { + "epoch": 1.68, + "learning_rate": 1.6887262759163846e-05, + "loss": 1.0002, + "step": 22770 + }, + { + "epoch": 1.68, + "learning_rate": 1.6886973563201018e-05, + "loss": 1.1148, + "step": 22771 + }, + { + "epoch": 1.68, + "learning_rate": 1.6886684356281118e-05, + "loss": 1.002, + "step": 22772 + }, + { + "epoch": 1.68, + "learning_rate": 1.68863951384046e-05, + "loss": 0.9864, + "step": 22773 + }, + { + "epoch": 1.68, + "learning_rate": 1.6886105909571925e-05, + "loss": 0.9835, + "step": 22774 + }, + { + "epoch": 1.68, + "learning_rate": 1.6885816669783555e-05, + "loss": 0.9757, + "step": 22775 + }, + { + "epoch": 1.68, + "learning_rate": 1.688552741903995e-05, + "loss": 1.0038, + "step": 22776 + }, + { + "epoch": 1.68, + "learning_rate": 1.6885238157341567e-05, + "loss": 1.0428, + "step": 22777 + }, + { + "epoch": 1.68, + "learning_rate": 1.688494888468887e-05, + "loss": 1.07, + "step": 22778 + }, + { + "epoch": 1.68, + "learning_rate": 1.6884659601082317e-05, + "loss": 0.9875, + "step": 22779 + }, + { + "epoch": 1.68, + "learning_rate": 1.688437030652237e-05, + "loss": 0.983, + "step": 22780 + }, + { + "epoch": 1.68, + "learning_rate": 1.688408100100949e-05, + "loss": 1.028, + "step": 22781 + }, + { + "epoch": 1.68, + "learning_rate": 1.6883791684544135e-05, + "loss": 0.9798, + "step": 22782 + }, + { + "epoch": 1.68, + "learning_rate": 1.6883502357126764e-05, + "loss": 0.9577, + "step": 22783 + }, + { + "epoch": 1.68, + "learning_rate": 1.688321301875784e-05, + "loss": 1.0039, + "step": 22784 + }, + { + "epoch": 1.68, + "learning_rate": 1.6882923669437823e-05, + "loss": 1.0235, + "step": 22785 + }, + { + "epoch": 1.68, + "learning_rate": 1.688263430916717e-05, + "loss": 1.044, + "step": 22786 + }, + { + "epoch": 1.68, + "learning_rate": 1.6882344937946345e-05, + "loss": 1.0224, + "step": 22787 + }, + { + "epoch": 1.68, + "learning_rate": 1.6882055555775812e-05, + "loss": 1.0911, + "step": 22788 + }, + { + "epoch": 1.68, + "learning_rate": 1.6881766162656027e-05, + "loss": 1.0515, + "step": 22789 + }, + { + "epoch": 1.68, + "learning_rate": 1.6881476758587446e-05, + "loss": 1.006, + "step": 22790 + }, + { + "epoch": 1.68, + "learning_rate": 1.6881187343570537e-05, + "loss": 1.0343, + "step": 22791 + }, + { + "epoch": 1.68, + "learning_rate": 1.6880897917605755e-05, + "loss": 1.0184, + "step": 22792 + }, + { + "epoch": 1.68, + "learning_rate": 1.6880608480693563e-05, + "loss": 1.0009, + "step": 22793 + }, + { + "epoch": 1.68, + "learning_rate": 1.6880319032834426e-05, + "loss": 1.1126, + "step": 22794 + }, + { + "epoch": 1.68, + "learning_rate": 1.6880029574028794e-05, + "loss": 1.0031, + "step": 22795 + }, + { + "epoch": 1.68, + "learning_rate": 1.6879740104277135e-05, + "loss": 0.8628, + "step": 22796 + }, + { + "epoch": 1.68, + "learning_rate": 1.687945062357991e-05, + "loss": 0.9074, + "step": 22797 + }, + { + "epoch": 1.68, + "learning_rate": 1.6879161131937575e-05, + "loss": 1.0036, + "step": 22798 + }, + { + "epoch": 1.68, + "learning_rate": 1.6878871629350595e-05, + "loss": 1.0195, + "step": 22799 + }, + { + "epoch": 1.68, + "learning_rate": 1.687858211581943e-05, + "loss": 1.0127, + "step": 22800 + }, + { + "epoch": 1.68, + "learning_rate": 1.6878292591344535e-05, + "loss": 1.0323, + "step": 22801 + }, + { + "epoch": 1.68, + "learning_rate": 1.687800305592638e-05, + "loss": 1.0008, + "step": 22802 + }, + { + "epoch": 1.68, + "learning_rate": 1.6877713509565417e-05, + "loss": 0.966, + "step": 22803 + }, + { + "epoch": 1.68, + "learning_rate": 1.6877423952262112e-05, + "loss": 1.0506, + "step": 22804 + }, + { + "epoch": 1.69, + "learning_rate": 1.687713438401692e-05, + "loss": 1.0496, + "step": 22805 + }, + { + "epoch": 1.69, + "learning_rate": 1.6876844804830308e-05, + "loss": 0.9639, + "step": 22806 + }, + { + "epoch": 1.69, + "learning_rate": 1.6876555214702737e-05, + "loss": 1.0294, + "step": 22807 + }, + { + "epoch": 1.69, + "learning_rate": 1.687626561363466e-05, + "loss": 0.9701, + "step": 22808 + }, + { + "epoch": 1.69, + "learning_rate": 1.6875976001626547e-05, + "loss": 0.9422, + "step": 22809 + }, + { + "epoch": 1.69, + "learning_rate": 1.6875686378678856e-05, + "loss": 1.0206, + "step": 22810 + }, + { + "epoch": 1.69, + "learning_rate": 1.6875396744792042e-05, + "loss": 1.0115, + "step": 22811 + }, + { + "epoch": 1.69, + "learning_rate": 1.687510709996657e-05, + "loss": 0.9187, + "step": 22812 + }, + { + "epoch": 1.69, + "learning_rate": 1.68748174442029e-05, + "loss": 1.0009, + "step": 22813 + }, + { + "epoch": 1.69, + "learning_rate": 1.68745277775015e-05, + "loss": 0.9679, + "step": 22814 + }, + { + "epoch": 1.69, + "learning_rate": 1.6874238099862822e-05, + "loss": 0.9923, + "step": 22815 + }, + { + "epoch": 1.69, + "learning_rate": 1.6873948411287326e-05, + "loss": 0.9818, + "step": 22816 + }, + { + "epoch": 1.69, + "learning_rate": 1.687365871177548e-05, + "loss": 1.065, + "step": 22817 + }, + { + "epoch": 1.69, + "learning_rate": 1.687336900132774e-05, + "loss": 0.906, + "step": 22818 + }, + { + "epoch": 1.69, + "learning_rate": 1.6873079279944567e-05, + "loss": 1.1095, + "step": 22819 + }, + { + "epoch": 1.69, + "learning_rate": 1.6872789547626423e-05, + "loss": 0.9332, + "step": 22820 + }, + { + "epoch": 1.69, + "learning_rate": 1.6872499804373775e-05, + "loss": 1.0093, + "step": 22821 + }, + { + "epoch": 1.69, + "learning_rate": 1.6872210050187073e-05, + "loss": 1.1311, + "step": 22822 + }, + { + "epoch": 1.69, + "learning_rate": 1.6871920285066785e-05, + "loss": 0.9811, + "step": 22823 + }, + { + "epoch": 1.69, + "learning_rate": 1.687163050901337e-05, + "loss": 1.0683, + "step": 22824 + }, + { + "epoch": 1.69, + "learning_rate": 1.6871340722027288e-05, + "loss": 0.9566, + "step": 22825 + }, + { + "epoch": 1.69, + "learning_rate": 1.6871050924109e-05, + "loss": 0.9503, + "step": 22826 + }, + { + "epoch": 1.69, + "learning_rate": 1.687076111525897e-05, + "loss": 1.0277, + "step": 22827 + }, + { + "epoch": 1.69, + "learning_rate": 1.6870471295477657e-05, + "loss": 0.9421, + "step": 22828 + }, + { + "epoch": 1.69, + "learning_rate": 1.6870181464765526e-05, + "loss": 1.0657, + "step": 22829 + }, + { + "epoch": 1.69, + "learning_rate": 1.6869891623123028e-05, + "loss": 1.0118, + "step": 22830 + }, + { + "epoch": 1.69, + "learning_rate": 1.6869601770550637e-05, + "loss": 0.9756, + "step": 22831 + }, + { + "epoch": 1.69, + "learning_rate": 1.6869311907048804e-05, + "loss": 1.0014, + "step": 22832 + }, + { + "epoch": 1.69, + "learning_rate": 1.6869022032617993e-05, + "loss": 1.0829, + "step": 22833 + }, + { + "epoch": 1.69, + "learning_rate": 1.6868732147258668e-05, + "loss": 0.991, + "step": 22834 + }, + { + "epoch": 1.69, + "learning_rate": 1.686844225097129e-05, + "loss": 0.9894, + "step": 22835 + }, + { + "epoch": 1.69, + "learning_rate": 1.6868152343756315e-05, + "loss": 0.9977, + "step": 22836 + }, + { + "epoch": 1.69, + "learning_rate": 1.6867862425614213e-05, + "loss": 0.9948, + "step": 22837 + }, + { + "epoch": 1.69, + "learning_rate": 1.686757249654544e-05, + "loss": 1.0006, + "step": 22838 + }, + { + "epoch": 1.69, + "learning_rate": 1.6867282556550453e-05, + "loss": 1.0941, + "step": 22839 + }, + { + "epoch": 1.69, + "learning_rate": 1.686699260562972e-05, + "loss": 1.0017, + "step": 22840 + }, + { + "epoch": 1.69, + "learning_rate": 1.68667026437837e-05, + "loss": 0.9342, + "step": 22841 + }, + { + "epoch": 1.69, + "learning_rate": 1.6866412671012854e-05, + "loss": 0.9883, + "step": 22842 + }, + { + "epoch": 1.69, + "learning_rate": 1.6866122687317642e-05, + "loss": 1.0309, + "step": 22843 + }, + { + "epoch": 1.69, + "learning_rate": 1.686583269269853e-05, + "loss": 0.9595, + "step": 22844 + }, + { + "epoch": 1.69, + "learning_rate": 1.6865542687155974e-05, + "loss": 1.084, + "step": 22845 + }, + { + "epoch": 1.69, + "learning_rate": 1.686525267069044e-05, + "loss": 0.9902, + "step": 22846 + }, + { + "epoch": 1.69, + "learning_rate": 1.686496264330239e-05, + "loss": 1.0544, + "step": 22847 + }, + { + "epoch": 1.69, + "learning_rate": 1.6864672604992274e-05, + "loss": 0.9979, + "step": 22848 + }, + { + "epoch": 1.69, + "learning_rate": 1.6864382555760568e-05, + "loss": 1.0099, + "step": 22849 + }, + { + "epoch": 1.69, + "learning_rate": 1.6864092495607724e-05, + "loss": 0.9599, + "step": 22850 + }, + { + "epoch": 1.69, + "learning_rate": 1.6863802424534212e-05, + "loss": 0.9995, + "step": 22851 + }, + { + "epoch": 1.69, + "learning_rate": 1.6863512342540485e-05, + "loss": 0.9463, + "step": 22852 + }, + { + "epoch": 1.69, + "learning_rate": 1.6863222249627008e-05, + "loss": 1.0632, + "step": 22853 + }, + { + "epoch": 1.69, + "learning_rate": 1.6862932145794243e-05, + "loss": 1.0614, + "step": 22854 + }, + { + "epoch": 1.69, + "learning_rate": 1.6862642031042655e-05, + "loss": 0.9946, + "step": 22855 + }, + { + "epoch": 1.69, + "learning_rate": 1.68623519053727e-05, + "loss": 1.0693, + "step": 22856 + }, + { + "epoch": 1.69, + "learning_rate": 1.686206176878484e-05, + "loss": 1.0323, + "step": 22857 + }, + { + "epoch": 1.69, + "learning_rate": 1.6861771621279536e-05, + "loss": 1.0461, + "step": 22858 + }, + { + "epoch": 1.69, + "learning_rate": 1.6861481462857253e-05, + "loss": 1.0212, + "step": 22859 + }, + { + "epoch": 1.69, + "learning_rate": 1.686119129351845e-05, + "loss": 1.0056, + "step": 22860 + }, + { + "epoch": 1.69, + "learning_rate": 1.686090111326359e-05, + "loss": 0.9648, + "step": 22861 + }, + { + "epoch": 1.69, + "learning_rate": 1.6860610922093138e-05, + "loss": 0.8685, + "step": 22862 + }, + { + "epoch": 1.69, + "learning_rate": 1.686032072000755e-05, + "loss": 1.0135, + "step": 22863 + }, + { + "epoch": 1.69, + "learning_rate": 1.686003050700729e-05, + "loss": 0.9584, + "step": 22864 + }, + { + "epoch": 1.69, + "learning_rate": 1.685974028309282e-05, + "loss": 0.9865, + "step": 22865 + }, + { + "epoch": 1.69, + "learning_rate": 1.68594500482646e-05, + "loss": 0.995, + "step": 22866 + }, + { + "epoch": 1.69, + "learning_rate": 1.6859159802523094e-05, + "loss": 1.0899, + "step": 22867 + }, + { + "epoch": 1.69, + "learning_rate": 1.6858869545868764e-05, + "loss": 1.0281, + "step": 22868 + }, + { + "epoch": 1.69, + "learning_rate": 1.685857927830207e-05, + "loss": 0.9463, + "step": 22869 + }, + { + "epoch": 1.69, + "learning_rate": 1.6858288999823472e-05, + "loss": 1.0378, + "step": 22870 + }, + { + "epoch": 1.69, + "learning_rate": 1.6857998710433436e-05, + "loss": 1.0555, + "step": 22871 + }, + { + "epoch": 1.69, + "learning_rate": 1.6857708410132423e-05, + "loss": 0.9763, + "step": 22872 + }, + { + "epoch": 1.69, + "learning_rate": 1.6857418098920893e-05, + "loss": 1.1235, + "step": 22873 + }, + { + "epoch": 1.69, + "learning_rate": 1.685712777679931e-05, + "loss": 0.9521, + "step": 22874 + }, + { + "epoch": 1.69, + "learning_rate": 1.6856837443768136e-05, + "loss": 1.0622, + "step": 22875 + }, + { + "epoch": 1.69, + "learning_rate": 1.685654709982783e-05, + "loss": 0.9823, + "step": 22876 + }, + { + "epoch": 1.69, + "learning_rate": 1.6856256744978853e-05, + "loss": 1.0121, + "step": 22877 + }, + { + "epoch": 1.69, + "learning_rate": 1.6855966379221675e-05, + "loss": 1.0137, + "step": 22878 + }, + { + "epoch": 1.69, + "learning_rate": 1.6855676002556748e-05, + "loss": 0.8571, + "step": 22879 + }, + { + "epoch": 1.69, + "learning_rate": 1.6855385614984538e-05, + "loss": 0.9482, + "step": 22880 + }, + { + "epoch": 1.69, + "learning_rate": 1.685509521650551e-05, + "loss": 1.1531, + "step": 22881 + }, + { + "epoch": 1.69, + "learning_rate": 1.6854804807120126e-05, + "loss": 0.9608, + "step": 22882 + }, + { + "epoch": 1.69, + "learning_rate": 1.6854514386828845e-05, + "loss": 1.1642, + "step": 22883 + }, + { + "epoch": 1.69, + "learning_rate": 1.6854223955632124e-05, + "loss": 0.9837, + "step": 22884 + }, + { + "epoch": 1.69, + "learning_rate": 1.6853933513530436e-05, + "loss": 0.9832, + "step": 22885 + }, + { + "epoch": 1.69, + "learning_rate": 1.685364306052424e-05, + "loss": 1.0513, + "step": 22886 + }, + { + "epoch": 1.69, + "learning_rate": 1.685335259661399e-05, + "loss": 1.0974, + "step": 22887 + }, + { + "epoch": 1.69, + "learning_rate": 1.6853062121800155e-05, + "loss": 1.0601, + "step": 22888 + }, + { + "epoch": 1.69, + "learning_rate": 1.68527716360832e-05, + "loss": 1.0103, + "step": 22889 + }, + { + "epoch": 1.69, + "learning_rate": 1.6852481139463578e-05, + "loss": 1.0593, + "step": 22890 + }, + { + "epoch": 1.69, + "learning_rate": 1.685219063194176e-05, + "loss": 1.1076, + "step": 22891 + }, + { + "epoch": 1.69, + "learning_rate": 1.6851900113518204e-05, + "loss": 1.1271, + "step": 22892 + }, + { + "epoch": 1.69, + "learning_rate": 1.6851609584193377e-05, + "loss": 1.0427, + "step": 22893 + }, + { + "epoch": 1.69, + "learning_rate": 1.6851319043967733e-05, + "loss": 1.1141, + "step": 22894 + }, + { + "epoch": 1.69, + "learning_rate": 1.6851028492841738e-05, + "loss": 1.1198, + "step": 22895 + }, + { + "epoch": 1.69, + "learning_rate": 1.6850737930815855e-05, + "loss": 0.9884, + "step": 22896 + }, + { + "epoch": 1.69, + "learning_rate": 1.685044735789055e-05, + "loss": 1.0439, + "step": 22897 + }, + { + "epoch": 1.69, + "learning_rate": 1.6850156774066276e-05, + "loss": 1.0102, + "step": 22898 + }, + { + "epoch": 1.69, + "learning_rate": 1.6849866179343502e-05, + "loss": 0.9279, + "step": 22899 + }, + { + "epoch": 1.69, + "learning_rate": 1.6849575573722687e-05, + "loss": 0.9391, + "step": 22900 + }, + { + "epoch": 1.69, + "learning_rate": 1.6849284957204298e-05, + "loss": 1.1461, + "step": 22901 + }, + { + "epoch": 1.69, + "learning_rate": 1.6848994329788796e-05, + "loss": 1.1011, + "step": 22902 + }, + { + "epoch": 1.69, + "learning_rate": 1.6848703691476642e-05, + "loss": 1.0161, + "step": 22903 + }, + { + "epoch": 1.69, + "learning_rate": 1.6848413042268298e-05, + "loss": 0.8672, + "step": 22904 + }, + { + "epoch": 1.69, + "learning_rate": 1.6848122382164223e-05, + "loss": 0.9961, + "step": 22905 + }, + { + "epoch": 1.69, + "learning_rate": 1.6847831711164885e-05, + "loss": 1.0234, + "step": 22906 + }, + { + "epoch": 1.69, + "learning_rate": 1.6847541029270748e-05, + "loss": 1.0457, + "step": 22907 + }, + { + "epoch": 1.69, + "learning_rate": 1.6847250336482267e-05, + "loss": 1.0873, + "step": 22908 + }, + { + "epoch": 1.69, + "learning_rate": 1.684695963279991e-05, + "loss": 1.0545, + "step": 22909 + }, + { + "epoch": 1.69, + "learning_rate": 1.6846668918224137e-05, + "loss": 1.0171, + "step": 22910 + }, + { + "epoch": 1.69, + "learning_rate": 1.6846378192755413e-05, + "loss": 1.118, + "step": 22911 + }, + { + "epoch": 1.69, + "learning_rate": 1.6846087456394206e-05, + "loss": 0.9251, + "step": 22912 + }, + { + "epoch": 1.69, + "learning_rate": 1.6845796709140964e-05, + "loss": 1.0722, + "step": 22913 + }, + { + "epoch": 1.69, + "learning_rate": 1.684550595099616e-05, + "loss": 1.006, + "step": 22914 + }, + { + "epoch": 1.69, + "learning_rate": 1.684521518196025e-05, + "loss": 1.0428, + "step": 22915 + }, + { + "epoch": 1.69, + "learning_rate": 1.6844924402033703e-05, + "loss": 0.9817, + "step": 22916 + }, + { + "epoch": 1.69, + "learning_rate": 1.684463361121698e-05, + "loss": 1.0141, + "step": 22917 + }, + { + "epoch": 1.69, + "learning_rate": 1.6844342809510542e-05, + "loss": 0.9444, + "step": 22918 + }, + { + "epoch": 1.69, + "learning_rate": 1.684405199691485e-05, + "loss": 1.0597, + "step": 22919 + }, + { + "epoch": 1.69, + "learning_rate": 1.6843761173430375e-05, + "loss": 0.9759, + "step": 22920 + }, + { + "epoch": 1.69, + "learning_rate": 1.684347033905757e-05, + "loss": 1.0954, + "step": 22921 + }, + { + "epoch": 1.69, + "learning_rate": 1.6843179493796902e-05, + "loss": 1.1526, + "step": 22922 + }, + { + "epoch": 1.69, + "learning_rate": 1.6842888637648833e-05, + "loss": 1.0452, + "step": 22923 + }, + { + "epoch": 1.69, + "learning_rate": 1.684259777061383e-05, + "loss": 1.0951, + "step": 22924 + }, + { + "epoch": 1.69, + "learning_rate": 1.6842306892692346e-05, + "loss": 0.9939, + "step": 22925 + }, + { + "epoch": 1.69, + "learning_rate": 1.6842016003884855e-05, + "loss": 0.947, + "step": 22926 + }, + { + "epoch": 1.69, + "learning_rate": 1.6841725104191812e-05, + "loss": 1.1211, + "step": 22927 + }, + { + "epoch": 1.69, + "learning_rate": 1.684143419361368e-05, + "loss": 0.9431, + "step": 22928 + }, + { + "epoch": 1.69, + "learning_rate": 1.6841143272150925e-05, + "loss": 1.0575, + "step": 22929 + }, + { + "epoch": 1.69, + "learning_rate": 1.6840852339804013e-05, + "loss": 1.0582, + "step": 22930 + }, + { + "epoch": 1.69, + "learning_rate": 1.6840561396573402e-05, + "loss": 1.0155, + "step": 22931 + }, + { + "epoch": 1.69, + "learning_rate": 1.6840270442459552e-05, + "loss": 0.9681, + "step": 22932 + }, + { + "epoch": 1.69, + "learning_rate": 1.6839979477462932e-05, + "loss": 1.0164, + "step": 22933 + }, + { + "epoch": 1.69, + "learning_rate": 1.6839688501583998e-05, + "loss": 1.0042, + "step": 22934 + }, + { + "epoch": 1.69, + "learning_rate": 1.6839397514823222e-05, + "loss": 0.9946, + "step": 22935 + }, + { + "epoch": 1.69, + "learning_rate": 1.6839106517181064e-05, + "loss": 0.965, + "step": 22936 + }, + { + "epoch": 1.69, + "learning_rate": 1.6838815508657982e-05, + "loss": 1.0341, + "step": 22937 + }, + { + "epoch": 1.69, + "learning_rate": 1.6838524489254444e-05, + "loss": 1.0346, + "step": 22938 + }, + { + "epoch": 1.69, + "learning_rate": 1.683823345897091e-05, + "loss": 1.1075, + "step": 22939 + }, + { + "epoch": 1.7, + "learning_rate": 1.6837942417807846e-05, + "loss": 0.9594, + "step": 22940 + }, + { + "epoch": 1.7, + "learning_rate": 1.6837651365765712e-05, + "loss": 0.9456, + "step": 22941 + }, + { + "epoch": 1.7, + "learning_rate": 1.6837360302844975e-05, + "loss": 0.9406, + "step": 22942 + }, + { + "epoch": 1.7, + "learning_rate": 1.683706922904609e-05, + "loss": 1.0744, + "step": 22943 + }, + { + "epoch": 1.7, + "learning_rate": 1.6836778144369533e-05, + "loss": 1.0736, + "step": 22944 + }, + { + "epoch": 1.7, + "learning_rate": 1.6836487048815754e-05, + "loss": 0.9377, + "step": 22945 + }, + { + "epoch": 1.7, + "learning_rate": 1.6836195942385224e-05, + "loss": 0.9369, + "step": 22946 + }, + { + "epoch": 1.7, + "learning_rate": 1.6835904825078406e-05, + "loss": 0.9866, + "step": 22947 + }, + { + "epoch": 1.7, + "learning_rate": 1.683561369689576e-05, + "loss": 0.9515, + "step": 22948 + }, + { + "epoch": 1.7, + "learning_rate": 1.6835322557837747e-05, + "loss": 0.967, + "step": 22949 + }, + { + "epoch": 1.7, + "learning_rate": 1.683503140790484e-05, + "loss": 0.9687, + "step": 22950 + }, + { + "epoch": 1.7, + "learning_rate": 1.683474024709749e-05, + "loss": 1.0109, + "step": 22951 + }, + { + "epoch": 1.7, + "learning_rate": 1.6834449075416168e-05, + "loss": 0.9852, + "step": 22952 + }, + { + "epoch": 1.7, + "learning_rate": 1.6834157892861337e-05, + "loss": 1.0011, + "step": 22953 + }, + { + "epoch": 1.7, + "learning_rate": 1.6833866699433457e-05, + "loss": 0.9961, + "step": 22954 + }, + { + "epoch": 1.7, + "learning_rate": 1.6833575495132993e-05, + "loss": 0.9958, + "step": 22955 + }, + { + "epoch": 1.7, + "learning_rate": 1.6833284279960408e-05, + "loss": 1.048, + "step": 22956 + }, + { + "epoch": 1.7, + "learning_rate": 1.6832993053916165e-05, + "loss": 0.9254, + "step": 22957 + }, + { + "epoch": 1.7, + "learning_rate": 1.683270181700073e-05, + "loss": 0.9843, + "step": 22958 + }, + { + "epoch": 1.7, + "learning_rate": 1.6832410569214562e-05, + "loss": 1.0252, + "step": 22959 + }, + { + "epoch": 1.7, + "learning_rate": 1.6832119310558128e-05, + "loss": 1.0191, + "step": 22960 + }, + { + "epoch": 1.7, + "learning_rate": 1.683182804103189e-05, + "loss": 0.922, + "step": 22961 + }, + { + "epoch": 1.7, + "learning_rate": 1.683153676063631e-05, + "loss": 1.0184, + "step": 22962 + }, + { + "epoch": 1.7, + "learning_rate": 1.6831245469371855e-05, + "loss": 0.8984, + "step": 22963 + }, + { + "epoch": 1.7, + "learning_rate": 1.683095416723898e-05, + "loss": 0.8809, + "step": 22964 + }, + { + "epoch": 1.7, + "learning_rate": 1.683066285423816e-05, + "loss": 0.9924, + "step": 22965 + }, + { + "epoch": 1.7, + "learning_rate": 1.683037153036985e-05, + "loss": 1.0425, + "step": 22966 + }, + { + "epoch": 1.7, + "learning_rate": 1.6830080195634523e-05, + "loss": 0.9566, + "step": 22967 + }, + { + "epoch": 1.7, + "learning_rate": 1.682978885003263e-05, + "loss": 1.0068, + "step": 22968 + }, + { + "epoch": 1.7, + "learning_rate": 1.6829497493564645e-05, + "loss": 0.9577, + "step": 22969 + }, + { + "epoch": 1.7, + "learning_rate": 1.6829206126231024e-05, + "loss": 1.0234, + "step": 22970 + }, + { + "epoch": 1.7, + "learning_rate": 1.6828914748032233e-05, + "loss": 0.9982, + "step": 22971 + }, + { + "epoch": 1.7, + "learning_rate": 1.682862335896874e-05, + "loss": 1.0062, + "step": 22972 + }, + { + "epoch": 1.7, + "learning_rate": 1.6828331959041e-05, + "loss": 1.0628, + "step": 22973 + }, + { + "epoch": 1.7, + "learning_rate": 1.6828040548249487e-05, + "loss": 1.1045, + "step": 22974 + }, + { + "epoch": 1.7, + "learning_rate": 1.6827749126594656e-05, + "loss": 1.0215, + "step": 22975 + }, + { + "epoch": 1.7, + "learning_rate": 1.6827457694076974e-05, + "loss": 1.0523, + "step": 22976 + }, + { + "epoch": 1.7, + "learning_rate": 1.6827166250696905e-05, + "loss": 0.9134, + "step": 22977 + }, + { + "epoch": 1.7, + "learning_rate": 1.682687479645491e-05, + "loss": 1.029, + "step": 22978 + }, + { + "epoch": 1.7, + "learning_rate": 1.6826583331351458e-05, + "loss": 1.0724, + "step": 22979 + }, + { + "epoch": 1.7, + "learning_rate": 1.682629185538701e-05, + "loss": 0.8632, + "step": 22980 + }, + { + "epoch": 1.7, + "learning_rate": 1.6826000368562028e-05, + "loss": 1.0137, + "step": 22981 + }, + { + "epoch": 1.7, + "learning_rate": 1.6825708870876978e-05, + "loss": 1.1254, + "step": 22982 + }, + { + "epoch": 1.7, + "learning_rate": 1.682541736233232e-05, + "loss": 1.0644, + "step": 22983 + }, + { + "epoch": 1.7, + "learning_rate": 1.682512584292852e-05, + "loss": 1.0802, + "step": 22984 + }, + { + "epoch": 1.7, + "learning_rate": 1.6824834312666046e-05, + "loss": 1.0074, + "step": 22985 + }, + { + "epoch": 1.7, + "learning_rate": 1.6824542771545357e-05, + "loss": 0.9596, + "step": 22986 + }, + { + "epoch": 1.7, + "learning_rate": 1.6824251219566922e-05, + "loss": 1.0169, + "step": 22987 + }, + { + "epoch": 1.7, + "learning_rate": 1.6823959656731195e-05, + "loss": 1.1304, + "step": 22988 + }, + { + "epoch": 1.7, + "learning_rate": 1.682366808303865e-05, + "loss": 1.0563, + "step": 22989 + }, + { + "epoch": 1.7, + "learning_rate": 1.682337649848974e-05, + "loss": 0.9932, + "step": 22990 + }, + { + "epoch": 1.7, + "learning_rate": 1.6823084903084944e-05, + "loss": 0.9683, + "step": 22991 + }, + { + "epoch": 1.7, + "learning_rate": 1.6822793296824714e-05, + "loss": 1.0029, + "step": 22992 + }, + { + "epoch": 1.7, + "learning_rate": 1.6822501679709516e-05, + "loss": 1.0646, + "step": 22993 + }, + { + "epoch": 1.7, + "learning_rate": 1.6822210051739815e-05, + "loss": 1.0388, + "step": 22994 + }, + { + "epoch": 1.7, + "learning_rate": 1.682191841291608e-05, + "loss": 0.9782, + "step": 22995 + }, + { + "epoch": 1.7, + "learning_rate": 1.682162676323877e-05, + "loss": 1.0447, + "step": 22996 + }, + { + "epoch": 1.7, + "learning_rate": 1.6821335102708346e-05, + "loss": 0.9358, + "step": 22997 + }, + { + "epoch": 1.7, + "learning_rate": 1.682104343132528e-05, + "loss": 1.0789, + "step": 22998 + }, + { + "epoch": 1.7, + "learning_rate": 1.6820751749090022e-05, + "loss": 0.9385, + "step": 22999 + }, + { + "epoch": 1.7, + "learning_rate": 1.6820460056003054e-05, + "loss": 0.9422, + "step": 23000 + }, + { + "epoch": 1.7, + "learning_rate": 1.6820168352064828e-05, + "loss": 1.0356, + "step": 23001 + }, + { + "epoch": 1.7, + "learning_rate": 1.6819876637275815e-05, + "loss": 0.9743, + "step": 23002 + }, + { + "epoch": 1.7, + "learning_rate": 1.6819584911636472e-05, + "loss": 1.0035, + "step": 23003 + }, + { + "epoch": 1.7, + "learning_rate": 1.6819293175147268e-05, + "loss": 1.1056, + "step": 23004 + }, + { + "epoch": 1.7, + "learning_rate": 1.6819001427808666e-05, + "loss": 1.0031, + "step": 23005 + }, + { + "epoch": 1.7, + "learning_rate": 1.6818709669621134e-05, + "loss": 1.1237, + "step": 23006 + }, + { + "epoch": 1.7, + "learning_rate": 1.681841790058513e-05, + "loss": 1.0841, + "step": 23007 + }, + { + "epoch": 1.7, + "learning_rate": 1.681812612070112e-05, + "loss": 1.0379, + "step": 23008 + }, + { + "epoch": 1.7, + "learning_rate": 1.681783432996957e-05, + "loss": 1.0352, + "step": 23009 + }, + { + "epoch": 1.7, + "learning_rate": 1.681754252839094e-05, + "loss": 1.0523, + "step": 23010 + }, + { + "epoch": 1.7, + "learning_rate": 1.6817250715965697e-05, + "loss": 0.9368, + "step": 23011 + }, + { + "epoch": 1.7, + "learning_rate": 1.681695889269431e-05, + "loss": 1.0251, + "step": 23012 + }, + { + "epoch": 1.7, + "learning_rate": 1.681666705857724e-05, + "loss": 1.0436, + "step": 23013 + }, + { + "epoch": 1.7, + "learning_rate": 1.6816375213614945e-05, + "loss": 1.0487, + "step": 23014 + }, + { + "epoch": 1.7, + "learning_rate": 1.68160833578079e-05, + "loss": 1.008, + "step": 23015 + }, + { + "epoch": 1.7, + "learning_rate": 1.6815791491156562e-05, + "loss": 1.0957, + "step": 23016 + }, + { + "epoch": 1.7, + "learning_rate": 1.6815499613661393e-05, + "loss": 0.9755, + "step": 23017 + }, + { + "epoch": 1.7, + "learning_rate": 1.6815207725322865e-05, + "loss": 1.0544, + "step": 23018 + }, + { + "epoch": 1.7, + "learning_rate": 1.681491582614144e-05, + "loss": 0.9917, + "step": 23019 + }, + { + "epoch": 1.7, + "learning_rate": 1.681462391611758e-05, + "loss": 1.0213, + "step": 23020 + }, + { + "epoch": 1.7, + "learning_rate": 1.681433199525175e-05, + "loss": 1.0334, + "step": 23021 + }, + { + "epoch": 1.7, + "learning_rate": 1.6814040063544416e-05, + "loss": 0.964, + "step": 23022 + }, + { + "epoch": 1.7, + "learning_rate": 1.6813748120996044e-05, + "loss": 1.1687, + "step": 23023 + }, + { + "epoch": 1.7, + "learning_rate": 1.6813456167607096e-05, + "loss": 1.0229, + "step": 23024 + }, + { + "epoch": 1.7, + "learning_rate": 1.6813164203378033e-05, + "loss": 1.1099, + "step": 23025 + }, + { + "epoch": 1.7, + "learning_rate": 1.6812872228309328e-05, + "loss": 0.9424, + "step": 23026 + }, + { + "epoch": 1.7, + "learning_rate": 1.6812580242401433e-05, + "loss": 0.9521, + "step": 23027 + }, + { + "epoch": 1.7, + "learning_rate": 1.6812288245654827e-05, + "loss": 1.032, + "step": 23028 + }, + { + "epoch": 1.7, + "learning_rate": 1.6811996238069967e-05, + "loss": 1.0669, + "step": 23029 + }, + { + "epoch": 1.7, + "learning_rate": 1.6811704219647317e-05, + "loss": 0.9779, + "step": 23030 + }, + { + "epoch": 1.7, + "learning_rate": 1.6811412190387347e-05, + "loss": 0.9684, + "step": 23031 + }, + { + "epoch": 1.7, + "learning_rate": 1.6811120150290513e-05, + "loss": 0.9274, + "step": 23032 + }, + { + "epoch": 1.7, + "learning_rate": 1.6810828099357285e-05, + "loss": 0.992, + "step": 23033 + }, + { + "epoch": 1.7, + "learning_rate": 1.6810536037588127e-05, + "loss": 1.0362, + "step": 23034 + }, + { + "epoch": 1.7, + "learning_rate": 1.6810243964983506e-05, + "loss": 0.9941, + "step": 23035 + }, + { + "epoch": 1.7, + "learning_rate": 1.680995188154388e-05, + "loss": 1.0436, + "step": 23036 + }, + { + "epoch": 1.7, + "learning_rate": 1.680965978726972e-05, + "loss": 1.1333, + "step": 23037 + }, + { + "epoch": 1.7, + "learning_rate": 1.6809367682161493e-05, + "loss": 0.9936, + "step": 23038 + }, + { + "epoch": 1.7, + "learning_rate": 1.6809075566219655e-05, + "loss": 0.9775, + "step": 23039 + }, + { + "epoch": 1.7, + "learning_rate": 1.6808783439444675e-05, + "loss": 0.9622, + "step": 23040 + }, + { + "epoch": 1.7, + "learning_rate": 1.6808491301837018e-05, + "loss": 0.9622, + "step": 23041 + }, + { + "epoch": 1.7, + "learning_rate": 1.6808199153397152e-05, + "loss": 0.9701, + "step": 23042 + }, + { + "epoch": 1.7, + "learning_rate": 1.6807906994125536e-05, + "loss": 0.9769, + "step": 23043 + }, + { + "epoch": 1.7, + "learning_rate": 1.680761482402264e-05, + "loss": 1.0501, + "step": 23044 + }, + { + "epoch": 1.7, + "learning_rate": 1.6807322643088922e-05, + "loss": 1.03, + "step": 23045 + }, + { + "epoch": 1.7, + "learning_rate": 1.6807030451324852e-05, + "loss": 1.0754, + "step": 23046 + }, + { + "epoch": 1.7, + "learning_rate": 1.6806738248730896e-05, + "loss": 1.0503, + "step": 23047 + }, + { + "epoch": 1.7, + "learning_rate": 1.6806446035307516e-05, + "loss": 1.0458, + "step": 23048 + }, + { + "epoch": 1.7, + "learning_rate": 1.6806153811055175e-05, + "loss": 1.038, + "step": 23049 + }, + { + "epoch": 1.7, + "learning_rate": 1.6805861575974345e-05, + "loss": 0.9826, + "step": 23050 + }, + { + "epoch": 1.7, + "learning_rate": 1.6805569330065482e-05, + "loss": 0.9603, + "step": 23051 + }, + { + "epoch": 1.7, + "learning_rate": 1.6805277073329062e-05, + "loss": 0.9828, + "step": 23052 + }, + { + "epoch": 1.7, + "learning_rate": 1.6804984805765538e-05, + "loss": 1.0835, + "step": 23053 + }, + { + "epoch": 1.7, + "learning_rate": 1.6804692527375385e-05, + "loss": 0.9154, + "step": 23054 + }, + { + "epoch": 1.7, + "learning_rate": 1.680440023815906e-05, + "loss": 0.9778, + "step": 23055 + }, + { + "epoch": 1.7, + "learning_rate": 1.6804107938117032e-05, + "loss": 0.9862, + "step": 23056 + }, + { + "epoch": 1.7, + "learning_rate": 1.6803815627249765e-05, + "loss": 1.0665, + "step": 23057 + }, + { + "epoch": 1.7, + "learning_rate": 1.6803523305557726e-05, + "loss": 0.9215, + "step": 23058 + }, + { + "epoch": 1.7, + "learning_rate": 1.6803230973041377e-05, + "loss": 1.0923, + "step": 23059 + }, + { + "epoch": 1.7, + "learning_rate": 1.6802938629701188e-05, + "loss": 0.9704, + "step": 23060 + }, + { + "epoch": 1.7, + "learning_rate": 1.680264627553762e-05, + "loss": 0.9242, + "step": 23061 + }, + { + "epoch": 1.7, + "learning_rate": 1.680235391055114e-05, + "loss": 0.9455, + "step": 23062 + }, + { + "epoch": 1.7, + "learning_rate": 1.6802061534742213e-05, + "loss": 1.0087, + "step": 23063 + }, + { + "epoch": 1.7, + "learning_rate": 1.6801769148111298e-05, + "loss": 0.96, + "step": 23064 + }, + { + "epoch": 1.7, + "learning_rate": 1.6801476750658866e-05, + "loss": 0.9116, + "step": 23065 + }, + { + "epoch": 1.7, + "learning_rate": 1.680118434238539e-05, + "loss": 1.0022, + "step": 23066 + }, + { + "epoch": 1.7, + "learning_rate": 1.680089192329132e-05, + "loss": 1.0429, + "step": 23067 + }, + { + "epoch": 1.7, + "learning_rate": 1.680059949337713e-05, + "loss": 1.0372, + "step": 23068 + }, + { + "epoch": 1.7, + "learning_rate": 1.6800307052643282e-05, + "loss": 1.1674, + "step": 23069 + }, + { + "epoch": 1.7, + "learning_rate": 1.6800014601090243e-05, + "loss": 1.0198, + "step": 23070 + }, + { + "epoch": 1.7, + "learning_rate": 1.6799722138718482e-05, + "loss": 0.9775, + "step": 23071 + }, + { + "epoch": 1.7, + "learning_rate": 1.6799429665528454e-05, + "loss": 0.9237, + "step": 23072 + }, + { + "epoch": 1.7, + "learning_rate": 1.6799137181520635e-05, + "loss": 1.0891, + "step": 23073 + }, + { + "epoch": 1.7, + "learning_rate": 1.6798844686695485e-05, + "loss": 0.9974, + "step": 23074 + }, + { + "epoch": 1.7, + "learning_rate": 1.6798552181053473e-05, + "loss": 1.0595, + "step": 23075 + }, + { + "epoch": 1.71, + "learning_rate": 1.6798259664595057e-05, + "loss": 0.9637, + "step": 23076 + }, + { + "epoch": 1.71, + "learning_rate": 1.679796713732071e-05, + "loss": 1.024, + "step": 23077 + }, + { + "epoch": 1.71, + "learning_rate": 1.6797674599230896e-05, + "loss": 1.1613, + "step": 23078 + }, + { + "epoch": 1.71, + "learning_rate": 1.6797382050326073e-05, + "loss": 0.9949, + "step": 23079 + }, + { + "epoch": 1.71, + "learning_rate": 1.679708949060672e-05, + "loss": 0.9987, + "step": 23080 + }, + { + "epoch": 1.71, + "learning_rate": 1.679679692007329e-05, + "loss": 1.0097, + "step": 23081 + }, + { + "epoch": 1.71, + "learning_rate": 1.6796504338726255e-05, + "loss": 1.1063, + "step": 23082 + }, + { + "epoch": 1.71, + "learning_rate": 1.6796211746566082e-05, + "loss": 1.0733, + "step": 23083 + }, + { + "epoch": 1.71, + "learning_rate": 1.6795919143593228e-05, + "loss": 1.0022, + "step": 23084 + }, + { + "epoch": 1.71, + "learning_rate": 1.6795626529808165e-05, + "loss": 1.031, + "step": 23085 + }, + { + "epoch": 1.71, + "learning_rate": 1.6795333905211356e-05, + "loss": 1.0895, + "step": 23086 + }, + { + "epoch": 1.71, + "learning_rate": 1.679504126980327e-05, + "loss": 0.9724, + "step": 23087 + }, + { + "epoch": 1.71, + "learning_rate": 1.679474862358437e-05, + "loss": 1.0147, + "step": 23088 + }, + { + "epoch": 1.71, + "learning_rate": 1.6794455966555124e-05, + "loss": 0.9222, + "step": 23089 + }, + { + "epoch": 1.71, + "learning_rate": 1.6794163298715996e-05, + "loss": 1.0131, + "step": 23090 + }, + { + "epoch": 1.71, + "learning_rate": 1.679387062006745e-05, + "loss": 0.9525, + "step": 23091 + }, + { + "epoch": 1.71, + "learning_rate": 1.679357793060995e-05, + "loss": 1.0177, + "step": 23092 + }, + { + "epoch": 1.71, + "learning_rate": 1.6793285230343972e-05, + "loss": 1.0074, + "step": 23093 + }, + { + "epoch": 1.71, + "learning_rate": 1.6792992519269972e-05, + "loss": 0.862, + "step": 23094 + }, + { + "epoch": 1.71, + "learning_rate": 1.6792699797388414e-05, + "loss": 0.9769, + "step": 23095 + }, + { + "epoch": 1.71, + "learning_rate": 1.679240706469977e-05, + "loss": 1.0367, + "step": 23096 + }, + { + "epoch": 1.71, + "learning_rate": 1.6792114321204502e-05, + "loss": 0.9605, + "step": 23097 + }, + { + "epoch": 1.71, + "learning_rate": 1.679182156690308e-05, + "loss": 1.0096, + "step": 23098 + }, + { + "epoch": 1.71, + "learning_rate": 1.6791528801795967e-05, + "loss": 0.9942, + "step": 23099 + }, + { + "epoch": 1.71, + "learning_rate": 1.6791236025883627e-05, + "loss": 0.996, + "step": 23100 + }, + { + "epoch": 1.71, + "learning_rate": 1.679094323916653e-05, + "loss": 0.9547, + "step": 23101 + }, + { + "epoch": 1.71, + "learning_rate": 1.6790650441645135e-05, + "loss": 0.9479, + "step": 23102 + }, + { + "epoch": 1.71, + "learning_rate": 1.6790357633319918e-05, + "loss": 1.0115, + "step": 23103 + }, + { + "epoch": 1.71, + "learning_rate": 1.6790064814191334e-05, + "loss": 1.1436, + "step": 23104 + }, + { + "epoch": 1.71, + "learning_rate": 1.6789771984259858e-05, + "loss": 0.9148, + "step": 23105 + }, + { + "epoch": 1.71, + "learning_rate": 1.678947914352595e-05, + "loss": 0.9935, + "step": 23106 + }, + { + "epoch": 1.71, + "learning_rate": 1.6789186291990075e-05, + "loss": 1.0743, + "step": 23107 + }, + { + "epoch": 1.71, + "learning_rate": 1.6788893429652708e-05, + "loss": 0.9919, + "step": 23108 + }, + { + "epoch": 1.71, + "learning_rate": 1.6788600556514304e-05, + "loss": 1.0193, + "step": 23109 + }, + { + "epoch": 1.71, + "learning_rate": 1.6788307672575337e-05, + "loss": 1.0702, + "step": 23110 + }, + { + "epoch": 1.71, + "learning_rate": 1.678801477783627e-05, + "loss": 0.996, + "step": 23111 + }, + { + "epoch": 1.71, + "learning_rate": 1.6787721872297564e-05, + "loss": 0.9573, + "step": 23112 + }, + { + "epoch": 1.71, + "learning_rate": 1.678742895595969e-05, + "loss": 0.9388, + "step": 23113 + }, + { + "epoch": 1.71, + "learning_rate": 1.6787136028823117e-05, + "loss": 0.969, + "step": 23114 + }, + { + "epoch": 1.71, + "learning_rate": 1.6786843090888307e-05, + "loss": 1.0726, + "step": 23115 + }, + { + "epoch": 1.71, + "learning_rate": 1.6786550142155726e-05, + "loss": 1.0802, + "step": 23116 + }, + { + "epoch": 1.71, + "learning_rate": 1.678625718262584e-05, + "loss": 1.0344, + "step": 23117 + }, + { + "epoch": 1.71, + "learning_rate": 1.6785964212299118e-05, + "loss": 0.9877, + "step": 23118 + }, + { + "epoch": 1.71, + "learning_rate": 1.6785671231176025e-05, + "loss": 0.9937, + "step": 23119 + }, + { + "epoch": 1.71, + "learning_rate": 1.6785378239257023e-05, + "loss": 0.9709, + "step": 23120 + }, + { + "epoch": 1.71, + "learning_rate": 1.6785085236542582e-05, + "loss": 0.919, + "step": 23121 + }, + { + "epoch": 1.71, + "learning_rate": 1.678479222303317e-05, + "loss": 1.0807, + "step": 23122 + }, + { + "epoch": 1.71, + "learning_rate": 1.6784499198729246e-05, + "loss": 0.9909, + "step": 23123 + }, + { + "epoch": 1.71, + "learning_rate": 1.6784206163631286e-05, + "loss": 0.9953, + "step": 23124 + }, + { + "epoch": 1.71, + "learning_rate": 1.6783913117739747e-05, + "loss": 1.0244, + "step": 23125 + }, + { + "epoch": 1.71, + "learning_rate": 1.6783620061055098e-05, + "loss": 1.0433, + "step": 23126 + }, + { + "epoch": 1.71, + "learning_rate": 1.6783326993577813e-05, + "loss": 1.0648, + "step": 23127 + }, + { + "epoch": 1.71, + "learning_rate": 1.6783033915308345e-05, + "loss": 0.9319, + "step": 23128 + }, + { + "epoch": 1.71, + "learning_rate": 1.6782740826247173e-05, + "loss": 0.9447, + "step": 23129 + }, + { + "epoch": 1.71, + "learning_rate": 1.6782447726394756e-05, + "loss": 1.0629, + "step": 23130 + }, + { + "epoch": 1.71, + "learning_rate": 1.6782154615751557e-05, + "loss": 0.8851, + "step": 23131 + }, + { + "epoch": 1.71, + "learning_rate": 1.678186149431805e-05, + "loss": 1.0704, + "step": 23132 + }, + { + "epoch": 1.71, + "learning_rate": 1.67815683620947e-05, + "loss": 0.9877, + "step": 23133 + }, + { + "epoch": 1.71, + "learning_rate": 1.678127521908197e-05, + "loss": 0.9875, + "step": 23134 + }, + { + "epoch": 1.71, + "learning_rate": 1.678098206528033e-05, + "loss": 0.9559, + "step": 23135 + }, + { + "epoch": 1.71, + "learning_rate": 1.6780688900690242e-05, + "loss": 1.0973, + "step": 23136 + }, + { + "epoch": 1.71, + "learning_rate": 1.6780395725312177e-05, + "loss": 1.0622, + "step": 23137 + }, + { + "epoch": 1.71, + "learning_rate": 1.67801025391466e-05, + "loss": 1.0309, + "step": 23138 + }, + { + "epoch": 1.71, + "learning_rate": 1.6779809342193975e-05, + "loss": 0.9899, + "step": 23139 + }, + { + "epoch": 1.71, + "learning_rate": 1.677951613445477e-05, + "loss": 1.0337, + "step": 23140 + }, + { + "epoch": 1.71, + "learning_rate": 1.6779222915929452e-05, + "loss": 1.0177, + "step": 23141 + }, + { + "epoch": 1.71, + "learning_rate": 1.677892968661849e-05, + "loss": 1.0551, + "step": 23142 + }, + { + "epoch": 1.71, + "learning_rate": 1.6778636446522343e-05, + "loss": 0.9211, + "step": 23143 + }, + { + "epoch": 1.71, + "learning_rate": 1.6778343195641486e-05, + "loss": 1.0266, + "step": 23144 + }, + { + "epoch": 1.71, + "learning_rate": 1.677804993397638e-05, + "loss": 1.0397, + "step": 23145 + }, + { + "epoch": 1.71, + "learning_rate": 1.6777756661527494e-05, + "loss": 1.0742, + "step": 23146 + }, + { + "epoch": 1.71, + "learning_rate": 1.6777463378295298e-05, + "loss": 1.0274, + "step": 23147 + }, + { + "epoch": 1.71, + "learning_rate": 1.677717008428025e-05, + "loss": 1.052, + "step": 23148 + }, + { + "epoch": 1.71, + "learning_rate": 1.677687677948282e-05, + "loss": 0.9368, + "step": 23149 + }, + { + "epoch": 1.71, + "learning_rate": 1.677658346390348e-05, + "loss": 0.9874, + "step": 23150 + }, + { + "epoch": 1.71, + "learning_rate": 1.6776290137542694e-05, + "loss": 1.0176, + "step": 23151 + }, + { + "epoch": 1.71, + "learning_rate": 1.6775996800400923e-05, + "loss": 1.0125, + "step": 23152 + }, + { + "epoch": 1.71, + "learning_rate": 1.677570345247864e-05, + "loss": 1.0003, + "step": 23153 + }, + { + "epoch": 1.71, + "learning_rate": 1.677541009377631e-05, + "loss": 0.9872, + "step": 23154 + }, + { + "epoch": 1.71, + "learning_rate": 1.6775116724294402e-05, + "loss": 0.9394, + "step": 23155 + }, + { + "epoch": 1.71, + "learning_rate": 1.677482334403338e-05, + "loss": 1.0396, + "step": 23156 + }, + { + "epoch": 1.71, + "learning_rate": 1.6774529952993705e-05, + "loss": 1.0899, + "step": 23157 + }, + { + "epoch": 1.71, + "learning_rate": 1.6774236551175853e-05, + "loss": 0.9854, + "step": 23158 + }, + { + "epoch": 1.71, + "learning_rate": 1.6773943138580285e-05, + "loss": 1.0562, + "step": 23159 + }, + { + "epoch": 1.71, + "learning_rate": 1.6773649715207477e-05, + "loss": 0.9959, + "step": 23160 + }, + { + "epoch": 1.71, + "learning_rate": 1.6773356281057886e-05, + "loss": 0.9564, + "step": 23161 + }, + { + "epoch": 1.71, + "learning_rate": 1.677306283613198e-05, + "loss": 1.0469, + "step": 23162 + }, + { + "epoch": 1.71, + "learning_rate": 1.677276938043023e-05, + "loss": 0.9808, + "step": 23163 + }, + { + "epoch": 1.71, + "learning_rate": 1.6772475913953098e-05, + "loss": 1.1645, + "step": 23164 + }, + { + "epoch": 1.71, + "learning_rate": 1.677218243670106e-05, + "loss": 1.0416, + "step": 23165 + }, + { + "epoch": 1.71, + "learning_rate": 1.677188894867457e-05, + "loss": 0.956, + "step": 23166 + }, + { + "epoch": 1.71, + "learning_rate": 1.6771595449874106e-05, + "loss": 0.8946, + "step": 23167 + }, + { + "epoch": 1.71, + "learning_rate": 1.6771301940300127e-05, + "loss": 0.9252, + "step": 23168 + }, + { + "epoch": 1.71, + "learning_rate": 1.6771008419953105e-05, + "loss": 0.9221, + "step": 23169 + }, + { + "epoch": 1.71, + "learning_rate": 1.6770714888833505e-05, + "loss": 0.9882, + "step": 23170 + }, + { + "epoch": 1.71, + "learning_rate": 1.6770421346941797e-05, + "loss": 0.9353, + "step": 23171 + }, + { + "epoch": 1.71, + "learning_rate": 1.6770127794278442e-05, + "loss": 0.9741, + "step": 23172 + }, + { + "epoch": 1.71, + "learning_rate": 1.676983423084391e-05, + "loss": 1.0882, + "step": 23173 + }, + { + "epoch": 1.71, + "learning_rate": 1.6769540656638672e-05, + "loss": 0.8453, + "step": 23174 + }, + { + "epoch": 1.71, + "learning_rate": 1.676924707166319e-05, + "loss": 1.0253, + "step": 23175 + }, + { + "epoch": 1.71, + "learning_rate": 1.6768953475917933e-05, + "loss": 0.9069, + "step": 23176 + }, + { + "epoch": 1.71, + "learning_rate": 1.676865986940337e-05, + "loss": 0.9472, + "step": 23177 + }, + { + "epoch": 1.71, + "learning_rate": 1.6768366252119964e-05, + "loss": 1.0139, + "step": 23178 + }, + { + "epoch": 1.71, + "learning_rate": 1.6768072624068183e-05, + "loss": 1.0471, + "step": 23179 + }, + { + "epoch": 1.71, + "learning_rate": 1.6767778985248496e-05, + "loss": 0.8933, + "step": 23180 + }, + { + "epoch": 1.71, + "learning_rate": 1.676748533566137e-05, + "loss": 1.0492, + "step": 23181 + }, + { + "epoch": 1.71, + "learning_rate": 1.6767191675307273e-05, + "loss": 1.0066, + "step": 23182 + }, + { + "epoch": 1.71, + "learning_rate": 1.6766898004186672e-05, + "loss": 0.9794, + "step": 23183 + }, + { + "epoch": 1.71, + "learning_rate": 1.6766604322300028e-05, + "loss": 0.978, + "step": 23184 + }, + { + "epoch": 1.71, + "learning_rate": 1.6766310629647816e-05, + "loss": 1.0616, + "step": 23185 + }, + { + "epoch": 1.71, + "learning_rate": 1.67660169262305e-05, + "loss": 1.0592, + "step": 23186 + }, + { + "epoch": 1.71, + "learning_rate": 1.676572321204855e-05, + "loss": 1.0238, + "step": 23187 + }, + { + "epoch": 1.71, + "learning_rate": 1.676542948710243e-05, + "loss": 1.0182, + "step": 23188 + }, + { + "epoch": 1.71, + "learning_rate": 1.6765135751392606e-05, + "loss": 0.9659, + "step": 23189 + }, + { + "epoch": 1.71, + "learning_rate": 1.676484200491955e-05, + "loss": 1.0321, + "step": 23190 + }, + { + "epoch": 1.71, + "learning_rate": 1.6764548247683726e-05, + "loss": 1.088, + "step": 23191 + }, + { + "epoch": 1.71, + "learning_rate": 1.6764254479685602e-05, + "loss": 0.9312, + "step": 23192 + }, + { + "epoch": 1.71, + "learning_rate": 1.676396070092565e-05, + "loss": 1.0599, + "step": 23193 + }, + { + "epoch": 1.71, + "learning_rate": 1.676366691140433e-05, + "loss": 0.977, + "step": 23194 + }, + { + "epoch": 1.71, + "learning_rate": 1.6763373111122116e-05, + "loss": 1.0367, + "step": 23195 + }, + { + "epoch": 1.71, + "learning_rate": 1.6763079300079468e-05, + "loss": 0.9514, + "step": 23196 + }, + { + "epoch": 1.71, + "learning_rate": 1.676278547827686e-05, + "loss": 0.9733, + "step": 23197 + }, + { + "epoch": 1.71, + "learning_rate": 1.6762491645714757e-05, + "loss": 0.9814, + "step": 23198 + }, + { + "epoch": 1.71, + "learning_rate": 1.6762197802393622e-05, + "loss": 0.9528, + "step": 23199 + }, + { + "epoch": 1.71, + "learning_rate": 1.676190394831393e-05, + "loss": 0.9527, + "step": 23200 + }, + { + "epoch": 1.71, + "learning_rate": 1.6761610083476147e-05, + "loss": 0.9273, + "step": 23201 + }, + { + "epoch": 1.71, + "learning_rate": 1.6761316207880742e-05, + "loss": 0.9562, + "step": 23202 + }, + { + "epoch": 1.71, + "learning_rate": 1.6761022321528175e-05, + "loss": 1.0901, + "step": 23203 + }, + { + "epoch": 1.71, + "learning_rate": 1.676072842441892e-05, + "loss": 0.9693, + "step": 23204 + }, + { + "epoch": 1.71, + "learning_rate": 1.6760434516553444e-05, + "loss": 1.0176, + "step": 23205 + }, + { + "epoch": 1.71, + "learning_rate": 1.676014059793221e-05, + "loss": 0.9888, + "step": 23206 + }, + { + "epoch": 1.71, + "learning_rate": 1.675984666855569e-05, + "loss": 0.9736, + "step": 23207 + }, + { + "epoch": 1.71, + "learning_rate": 1.675955272842435e-05, + "loss": 1.0653, + "step": 23208 + }, + { + "epoch": 1.71, + "learning_rate": 1.675925877753866e-05, + "loss": 0.9752, + "step": 23209 + }, + { + "epoch": 1.71, + "learning_rate": 1.6758964815899087e-05, + "loss": 0.9997, + "step": 23210 + }, + { + "epoch": 1.72, + "learning_rate": 1.6758670843506096e-05, + "loss": 1.0444, + "step": 23211 + }, + { + "epoch": 1.72, + "learning_rate": 1.6758376860360157e-05, + "loss": 0.9314, + "step": 23212 + }, + { + "epoch": 1.72, + "learning_rate": 1.6758082866461737e-05, + "loss": 0.9979, + "step": 23213 + }, + { + "epoch": 1.72, + "learning_rate": 1.6757788861811302e-05, + "loss": 0.9144, + "step": 23214 + }, + { + "epoch": 1.72, + "learning_rate": 1.6757494846409325e-05, + "loss": 1.0389, + "step": 23215 + }, + { + "epoch": 1.72, + "learning_rate": 1.6757200820256267e-05, + "loss": 1.0587, + "step": 23216 + }, + { + "epoch": 1.72, + "learning_rate": 1.67569067833526e-05, + "loss": 1.0024, + "step": 23217 + }, + { + "epoch": 1.72, + "learning_rate": 1.6756612735698792e-05, + "loss": 1.0027, + "step": 23218 + }, + { + "epoch": 1.72, + "learning_rate": 1.675631867729531e-05, + "loss": 0.9848, + "step": 23219 + }, + { + "epoch": 1.72, + "learning_rate": 1.675602460814262e-05, + "loss": 1.1132, + "step": 23220 + }, + { + "epoch": 1.72, + "learning_rate": 1.6755730528241193e-05, + "loss": 1.051, + "step": 23221 + }, + { + "epoch": 1.72, + "learning_rate": 1.6755436437591495e-05, + "loss": 1.0933, + "step": 23222 + }, + { + "epoch": 1.72, + "learning_rate": 1.6755142336193997e-05, + "loss": 0.9871, + "step": 23223 + }, + { + "epoch": 1.72, + "learning_rate": 1.675484822404916e-05, + "loss": 1.0767, + "step": 23224 + }, + { + "epoch": 1.72, + "learning_rate": 1.675455410115746e-05, + "loss": 1.094, + "step": 23225 + }, + { + "epoch": 1.72, + "learning_rate": 1.6754259967519357e-05, + "loss": 1.0616, + "step": 23226 + }, + { + "epoch": 1.72, + "learning_rate": 1.6753965823135326e-05, + "loss": 0.8775, + "step": 23227 + }, + { + "epoch": 1.72, + "learning_rate": 1.6753671668005828e-05, + "loss": 1.1148, + "step": 23228 + }, + { + "epoch": 1.72, + "learning_rate": 1.6753377502131338e-05, + "loss": 1.0227, + "step": 23229 + }, + { + "epoch": 1.72, + "learning_rate": 1.6753083325512323e-05, + "loss": 0.9861, + "step": 23230 + }, + { + "epoch": 1.72, + "learning_rate": 1.6752789138149247e-05, + "loss": 1.1145, + "step": 23231 + }, + { + "epoch": 1.72, + "learning_rate": 1.6752494940042582e-05, + "loss": 1.0764, + "step": 23232 + }, + { + "epoch": 1.72, + "learning_rate": 1.675220073119279e-05, + "loss": 1.0898, + "step": 23233 + }, + { + "epoch": 1.72, + "learning_rate": 1.6751906511600346e-05, + "loss": 1.0476, + "step": 23234 + }, + { + "epoch": 1.72, + "learning_rate": 1.6751612281265713e-05, + "loss": 1.046, + "step": 23235 + }, + { + "epoch": 1.72, + "learning_rate": 1.6751318040189365e-05, + "loss": 0.9417, + "step": 23236 + }, + { + "epoch": 1.72, + "learning_rate": 1.6751023788371764e-05, + "loss": 0.9254, + "step": 23237 + }, + { + "epoch": 1.72, + "learning_rate": 1.675072952581338e-05, + "loss": 0.8925, + "step": 23238 + }, + { + "epoch": 1.72, + "learning_rate": 1.6750435252514688e-05, + "loss": 0.9791, + "step": 23239 + }, + { + "epoch": 1.72, + "learning_rate": 1.6750140968476144e-05, + "loss": 1.0163, + "step": 23240 + }, + { + "epoch": 1.72, + "learning_rate": 1.6749846673698225e-05, + "loss": 1.0556, + "step": 23241 + }, + { + "epoch": 1.72, + "learning_rate": 1.6749552368181392e-05, + "loss": 1.0027, + "step": 23242 + }, + { + "epoch": 1.72, + "learning_rate": 1.6749258051926122e-05, + "loss": 1.012, + "step": 23243 + }, + { + "epoch": 1.72, + "learning_rate": 1.6748963724932877e-05, + "loss": 0.9766, + "step": 23244 + }, + { + "epoch": 1.72, + "learning_rate": 1.674866938720213e-05, + "loss": 1.031, + "step": 23245 + }, + { + "epoch": 1.72, + "learning_rate": 1.6748375038734344e-05, + "loss": 1.0286, + "step": 23246 + }, + { + "epoch": 1.72, + "learning_rate": 1.674808067952999e-05, + "loss": 1.102, + "step": 23247 + }, + { + "epoch": 1.72, + "learning_rate": 1.674778630958954e-05, + "loss": 1.0405, + "step": 23248 + }, + { + "epoch": 1.72, + "learning_rate": 1.6747491928913454e-05, + "loss": 1.0177, + "step": 23249 + }, + { + "epoch": 1.72, + "learning_rate": 1.6747197537502205e-05, + "loss": 1.0003, + "step": 23250 + }, + { + "epoch": 1.72, + "learning_rate": 1.6746903135356263e-05, + "loss": 1.0407, + "step": 23251 + }, + { + "epoch": 1.72, + "learning_rate": 1.6746608722476097e-05, + "loss": 0.9692, + "step": 23252 + }, + { + "epoch": 1.72, + "learning_rate": 1.674631429886217e-05, + "loss": 0.9507, + "step": 23253 + }, + { + "epoch": 1.72, + "learning_rate": 1.6746019864514952e-05, + "loss": 0.9224, + "step": 23254 + }, + { + "epoch": 1.72, + "learning_rate": 1.6745725419434916e-05, + "loss": 0.9507, + "step": 23255 + }, + { + "epoch": 1.72, + "learning_rate": 1.6745430963622526e-05, + "loss": 0.9958, + "step": 23256 + }, + { + "epoch": 1.72, + "learning_rate": 1.674513649707825e-05, + "loss": 0.9643, + "step": 23257 + }, + { + "epoch": 1.72, + "learning_rate": 1.6744842019802564e-05, + "loss": 0.9854, + "step": 23258 + }, + { + "epoch": 1.72, + "learning_rate": 1.6744547531795928e-05, + "loss": 1.0272, + "step": 23259 + }, + { + "epoch": 1.72, + "learning_rate": 1.674425303305881e-05, + "loss": 0.9027, + "step": 23260 + }, + { + "epoch": 1.72, + "learning_rate": 1.6743958523591684e-05, + "loss": 1.0229, + "step": 23261 + }, + { + "epoch": 1.72, + "learning_rate": 1.674366400339502e-05, + "loss": 0.9417, + "step": 23262 + }, + { + "epoch": 1.72, + "learning_rate": 1.674336947246928e-05, + "loss": 1.091, + "step": 23263 + }, + { + "epoch": 1.72, + "learning_rate": 1.6743074930814932e-05, + "loss": 0.9931, + "step": 23264 + }, + { + "epoch": 1.72, + "learning_rate": 1.6742780378432454e-05, + "loss": 0.9925, + "step": 23265 + }, + { + "epoch": 1.72, + "learning_rate": 1.6742485815322304e-05, + "loss": 1.0026, + "step": 23266 + }, + { + "epoch": 1.72, + "learning_rate": 1.6742191241484962e-05, + "loss": 1.0355, + "step": 23267 + }, + { + "epoch": 1.72, + "learning_rate": 1.6741896656920887e-05, + "loss": 1.1451, + "step": 23268 + }, + { + "epoch": 1.72, + "learning_rate": 1.674160206163055e-05, + "loss": 0.8432, + "step": 23269 + }, + { + "epoch": 1.72, + "learning_rate": 1.674130745561442e-05, + "loss": 0.9743, + "step": 23270 + }, + { + "epoch": 1.72, + "learning_rate": 1.6741012838872968e-05, + "loss": 1.0233, + "step": 23271 + }, + { + "epoch": 1.72, + "learning_rate": 1.674071821140666e-05, + "loss": 0.9981, + "step": 23272 + }, + { + "epoch": 1.72, + "learning_rate": 1.6740423573215963e-05, + "loss": 1.0296, + "step": 23273 + }, + { + "epoch": 1.72, + "learning_rate": 1.6740128924301355e-05, + "loss": 0.9863, + "step": 23274 + }, + { + "epoch": 1.72, + "learning_rate": 1.6739834264663293e-05, + "loss": 0.968, + "step": 23275 + }, + { + "epoch": 1.72, + "learning_rate": 1.6739539594302254e-05, + "loss": 1.0042, + "step": 23276 + }, + { + "epoch": 1.72, + "learning_rate": 1.6739244913218703e-05, + "loss": 1.0221, + "step": 23277 + }, + { + "epoch": 1.72, + "learning_rate": 1.673895022141311e-05, + "loss": 0.9839, + "step": 23278 + }, + { + "epoch": 1.72, + "learning_rate": 1.6738655518885942e-05, + "loss": 1.0432, + "step": 23279 + }, + { + "epoch": 1.72, + "learning_rate": 1.673836080563767e-05, + "loss": 1.049, + "step": 23280 + }, + { + "epoch": 1.72, + "learning_rate": 1.6738066081668764e-05, + "loss": 1.0652, + "step": 23281 + }, + { + "epoch": 1.72, + "learning_rate": 1.6737771346979693e-05, + "loss": 1.0149, + "step": 23282 + }, + { + "epoch": 1.72, + "learning_rate": 1.6737476601570922e-05, + "loss": 1.1152, + "step": 23283 + }, + { + "epoch": 1.72, + "learning_rate": 1.6737181845442922e-05, + "loss": 1.0372, + "step": 23284 + }, + { + "epoch": 1.72, + "learning_rate": 1.673688707859616e-05, + "loss": 1.1011, + "step": 23285 + }, + { + "epoch": 1.72, + "learning_rate": 1.673659230103111e-05, + "loss": 0.9739, + "step": 23286 + }, + { + "epoch": 1.72, + "learning_rate": 1.673629751274824e-05, + "loss": 0.9869, + "step": 23287 + }, + { + "epoch": 1.72, + "learning_rate": 1.673600271374801e-05, + "loss": 1.0488, + "step": 23288 + }, + { + "epoch": 1.72, + "learning_rate": 1.67357079040309e-05, + "loss": 0.9175, + "step": 23289 + }, + { + "epoch": 1.72, + "learning_rate": 1.6735413083597376e-05, + "loss": 0.9786, + "step": 23290 + }, + { + "epoch": 1.72, + "learning_rate": 1.6735118252447905e-05, + "loss": 1.0452, + "step": 23291 + }, + { + "epoch": 1.72, + "learning_rate": 1.673482341058296e-05, + "loss": 0.9763, + "step": 23292 + }, + { + "epoch": 1.72, + "learning_rate": 1.6734528558003002e-05, + "loss": 0.9907, + "step": 23293 + }, + { + "epoch": 1.72, + "learning_rate": 1.673423369470851e-05, + "loss": 0.9915, + "step": 23294 + }, + { + "epoch": 1.72, + "learning_rate": 1.6733938820699946e-05, + "loss": 1.09, + "step": 23295 + }, + { + "epoch": 1.72, + "learning_rate": 1.6733643935977783e-05, + "loss": 0.9893, + "step": 23296 + }, + { + "epoch": 1.72, + "learning_rate": 1.673334904054249e-05, + "loss": 0.9636, + "step": 23297 + }, + { + "epoch": 1.72, + "learning_rate": 1.673305413439453e-05, + "loss": 1.02, + "step": 23298 + }, + { + "epoch": 1.72, + "learning_rate": 1.6732759217534383e-05, + "loss": 1.0302, + "step": 23299 + }, + { + "epoch": 1.72, + "learning_rate": 1.673246428996251e-05, + "loss": 1.057, + "step": 23300 + }, + { + "epoch": 1.72, + "learning_rate": 1.6732169351679383e-05, + "loss": 1.0869, + "step": 23301 + }, + { + "epoch": 1.72, + "learning_rate": 1.6731874402685467e-05, + "loss": 0.9944, + "step": 23302 + }, + { + "epoch": 1.72, + "learning_rate": 1.673157944298124e-05, + "loss": 0.9284, + "step": 23303 + }, + { + "epoch": 1.72, + "learning_rate": 1.6731284472567165e-05, + "loss": 1.061, + "step": 23304 + }, + { + "epoch": 1.72, + "learning_rate": 1.6730989491443714e-05, + "loss": 1.0509, + "step": 23305 + }, + { + "epoch": 1.72, + "learning_rate": 1.6730694499611355e-05, + "loss": 1.0695, + "step": 23306 + }, + { + "epoch": 1.72, + "learning_rate": 1.6730399497070554e-05, + "loss": 1.0927, + "step": 23307 + }, + { + "epoch": 1.72, + "learning_rate": 1.6730104483821786e-05, + "loss": 1.1079, + "step": 23308 + }, + { + "epoch": 1.72, + "learning_rate": 1.672980945986552e-05, + "loss": 0.8876, + "step": 23309 + }, + { + "epoch": 1.72, + "learning_rate": 1.672951442520222e-05, + "loss": 1.0039, + "step": 23310 + }, + { + "epoch": 1.72, + "learning_rate": 1.672921937983236e-05, + "loss": 1.0805, + "step": 23311 + }, + { + "epoch": 1.72, + "learning_rate": 1.6728924323756407e-05, + "loss": 0.933, + "step": 23312 + }, + { + "epoch": 1.72, + "learning_rate": 1.6728629256974835e-05, + "loss": 0.9341, + "step": 23313 + }, + { + "epoch": 1.72, + "learning_rate": 1.672833417948811e-05, + "loss": 0.924, + "step": 23314 + }, + { + "epoch": 1.72, + "learning_rate": 1.67280390912967e-05, + "loss": 1.0175, + "step": 23315 + }, + { + "epoch": 1.72, + "learning_rate": 1.6727743992401074e-05, + "loss": 1.0101, + "step": 23316 + }, + { + "epoch": 1.72, + "learning_rate": 1.672744888280171e-05, + "loss": 0.9015, + "step": 23317 + }, + { + "epoch": 1.72, + "learning_rate": 1.6727153762499062e-05, + "loss": 1.0607, + "step": 23318 + }, + { + "epoch": 1.72, + "learning_rate": 1.6726858631493616e-05, + "loss": 0.9173, + "step": 23319 + }, + { + "epoch": 1.72, + "learning_rate": 1.672656348978583e-05, + "loss": 1.0532, + "step": 23320 + }, + { + "epoch": 1.72, + "learning_rate": 1.6726268337376176e-05, + "loss": 0.9794, + "step": 23321 + }, + { + "epoch": 1.72, + "learning_rate": 1.6725973174265128e-05, + "loss": 0.9953, + "step": 23322 + }, + { + "epoch": 1.72, + "learning_rate": 1.6725678000453154e-05, + "loss": 1.0565, + "step": 23323 + }, + { + "epoch": 1.72, + "learning_rate": 1.6725382815940724e-05, + "loss": 1.0655, + "step": 23324 + }, + { + "epoch": 1.72, + "learning_rate": 1.6725087620728303e-05, + "loss": 1.0213, + "step": 23325 + }, + { + "epoch": 1.72, + "learning_rate": 1.6724792414816365e-05, + "loss": 1.0401, + "step": 23326 + }, + { + "epoch": 1.72, + "learning_rate": 1.6724497198205378e-05, + "loss": 1.0598, + "step": 23327 + }, + { + "epoch": 1.72, + "learning_rate": 1.672420197089581e-05, + "loss": 0.9837, + "step": 23328 + }, + { + "epoch": 1.72, + "learning_rate": 1.6723906732888136e-05, + "loss": 0.9719, + "step": 23329 + }, + { + "epoch": 1.72, + "learning_rate": 1.6723611484182824e-05, + "loss": 0.9686, + "step": 23330 + }, + { + "epoch": 1.72, + "learning_rate": 1.672331622478034e-05, + "loss": 1.0091, + "step": 23331 + }, + { + "epoch": 1.72, + "learning_rate": 1.6723020954681155e-05, + "loss": 1.2003, + "step": 23332 + }, + { + "epoch": 1.72, + "learning_rate": 1.6722725673885742e-05, + "loss": 1.0568, + "step": 23333 + }, + { + "epoch": 1.72, + "learning_rate": 1.672243038239457e-05, + "loss": 0.9458, + "step": 23334 + }, + { + "epoch": 1.72, + "learning_rate": 1.67221350802081e-05, + "loss": 0.9404, + "step": 23335 + }, + { + "epoch": 1.72, + "learning_rate": 1.6721839767326818e-05, + "loss": 1.1172, + "step": 23336 + }, + { + "epoch": 1.72, + "learning_rate": 1.6721544443751182e-05, + "loss": 0.9193, + "step": 23337 + }, + { + "epoch": 1.72, + "learning_rate": 1.6721249109481664e-05, + "loss": 0.9259, + "step": 23338 + }, + { + "epoch": 1.72, + "learning_rate": 1.6720953764518734e-05, + "loss": 1.0732, + "step": 23339 + }, + { + "epoch": 1.72, + "learning_rate": 1.6720658408862864e-05, + "loss": 1.0302, + "step": 23340 + }, + { + "epoch": 1.72, + "learning_rate": 1.6720363042514524e-05, + "loss": 1.0844, + "step": 23341 + }, + { + "epoch": 1.72, + "learning_rate": 1.6720067665474182e-05, + "loss": 0.9628, + "step": 23342 + }, + { + "epoch": 1.72, + "learning_rate": 1.671977227774231e-05, + "loss": 0.9277, + "step": 23343 + }, + { + "epoch": 1.72, + "learning_rate": 1.6719476879319372e-05, + "loss": 1.0185, + "step": 23344 + }, + { + "epoch": 1.72, + "learning_rate": 1.6719181470205847e-05, + "loss": 1.1077, + "step": 23345 + }, + { + "epoch": 1.73, + "learning_rate": 1.6718886050402196e-05, + "loss": 1.0473, + "step": 23346 + }, + { + "epoch": 1.73, + "learning_rate": 1.6718590619908895e-05, + "loss": 1.0415, + "step": 23347 + }, + { + "epoch": 1.73, + "learning_rate": 1.671829517872641e-05, + "loss": 0.9337, + "step": 23348 + }, + { + "epoch": 1.73, + "learning_rate": 1.6717999726855217e-05, + "loss": 1.0295, + "step": 23349 + }, + { + "epoch": 1.73, + "learning_rate": 1.6717704264295786e-05, + "loss": 0.9787, + "step": 23350 + }, + { + "epoch": 1.73, + "learning_rate": 1.6717408791048578e-05, + "loss": 0.97, + "step": 23351 + }, + { + "epoch": 1.73, + "learning_rate": 1.671711330711407e-05, + "loss": 0.9589, + "step": 23352 + }, + { + "epoch": 1.73, + "learning_rate": 1.671681781249273e-05, + "loss": 1.0086, + "step": 23353 + }, + { + "epoch": 1.73, + "learning_rate": 1.671652230718503e-05, + "loss": 1.0236, + "step": 23354 + }, + { + "epoch": 1.73, + "learning_rate": 1.6716226791191442e-05, + "loss": 0.9779, + "step": 23355 + }, + { + "epoch": 1.73, + "learning_rate": 1.671593126451243e-05, + "loss": 1.006, + "step": 23356 + }, + { + "epoch": 1.73, + "learning_rate": 1.6715635727148466e-05, + "loss": 0.9443, + "step": 23357 + }, + { + "epoch": 1.73, + "learning_rate": 1.6715340179100023e-05, + "loss": 1.0362, + "step": 23358 + }, + { + "epoch": 1.73, + "learning_rate": 1.671504462036757e-05, + "loss": 1.0102, + "step": 23359 + }, + { + "epoch": 1.73, + "learning_rate": 1.6714749050951578e-05, + "loss": 0.8615, + "step": 23360 + }, + { + "epoch": 1.73, + "learning_rate": 1.671445347085252e-05, + "loss": 1.0347, + "step": 23361 + }, + { + "epoch": 1.73, + "learning_rate": 1.6714157880070857e-05, + "loss": 1.085, + "step": 23362 + }, + { + "epoch": 1.73, + "learning_rate": 1.671386227860707e-05, + "loss": 1.0455, + "step": 23363 + }, + { + "epoch": 1.73, + "learning_rate": 1.671356666646162e-05, + "loss": 1.1328, + "step": 23364 + }, + { + "epoch": 1.73, + "learning_rate": 1.671327104363498e-05, + "loss": 0.9838, + "step": 23365 + }, + { + "epoch": 1.73, + "learning_rate": 1.6712975410127623e-05, + "loss": 1.1075, + "step": 23366 + }, + { + "epoch": 1.73, + "learning_rate": 1.6712679765940018e-05, + "loss": 1.0577, + "step": 23367 + }, + { + "epoch": 1.73, + "learning_rate": 1.6712384111072636e-05, + "loss": 1.0518, + "step": 23368 + }, + { + "epoch": 1.73, + "learning_rate": 1.6712088445525946e-05, + "loss": 0.9905, + "step": 23369 + }, + { + "epoch": 1.73, + "learning_rate": 1.6711792769300425e-05, + "loss": 0.9215, + "step": 23370 + }, + { + "epoch": 1.73, + "learning_rate": 1.6711497082396532e-05, + "loss": 1.0692, + "step": 23371 + }, + { + "epoch": 1.73, + "learning_rate": 1.6711201384814745e-05, + "loss": 0.9935, + "step": 23372 + }, + { + "epoch": 1.73, + "learning_rate": 1.6710905676555532e-05, + "loss": 1.0896, + "step": 23373 + }, + { + "epoch": 1.73, + "learning_rate": 1.6710609957619363e-05, + "loss": 1.0603, + "step": 23374 + }, + { + "epoch": 1.73, + "learning_rate": 1.671031422800671e-05, + "loss": 1.0686, + "step": 23375 + }, + { + "epoch": 1.73, + "learning_rate": 1.6710018487718044e-05, + "loss": 1.0652, + "step": 23376 + }, + { + "epoch": 1.73, + "learning_rate": 1.670972273675384e-05, + "loss": 1.0215, + "step": 23377 + }, + { + "epoch": 1.73, + "learning_rate": 1.6709426975114555e-05, + "loss": 1.0164, + "step": 23378 + }, + { + "epoch": 1.73, + "learning_rate": 1.670913120280067e-05, + "loss": 0.9839, + "step": 23379 + }, + { + "epoch": 1.73, + "learning_rate": 1.6708835419812653e-05, + "loss": 0.9768, + "step": 23380 + }, + { + "epoch": 1.73, + "learning_rate": 1.6708539626150977e-05, + "loss": 1.1124, + "step": 23381 + }, + { + "epoch": 1.73, + "learning_rate": 1.6708243821816107e-05, + "loss": 0.985, + "step": 23382 + }, + { + "epoch": 1.73, + "learning_rate": 1.670794800680852e-05, + "loss": 1.0265, + "step": 23383 + }, + { + "epoch": 1.73, + "learning_rate": 1.6707652181128683e-05, + "loss": 0.979, + "step": 23384 + }, + { + "epoch": 1.73, + "learning_rate": 1.6707356344777065e-05, + "loss": 1.0718, + "step": 23385 + }, + { + "epoch": 1.73, + "learning_rate": 1.6707060497754143e-05, + "loss": 0.9356, + "step": 23386 + }, + { + "epoch": 1.73, + "learning_rate": 1.670676464006038e-05, + "loss": 1.0896, + "step": 23387 + }, + { + "epoch": 1.73, + "learning_rate": 1.6706468771696254e-05, + "loss": 1.0992, + "step": 23388 + }, + { + "epoch": 1.73, + "learning_rate": 1.6706172892662232e-05, + "loss": 1.1028, + "step": 23389 + }, + { + "epoch": 1.73, + "learning_rate": 1.670587700295878e-05, + "loss": 1.0846, + "step": 23390 + }, + { + "epoch": 1.73, + "learning_rate": 1.670558110258638e-05, + "loss": 0.9721, + "step": 23391 + }, + { + "epoch": 1.73, + "learning_rate": 1.6705285191545493e-05, + "loss": 0.9884, + "step": 23392 + }, + { + "epoch": 1.73, + "learning_rate": 1.6704989269836592e-05, + "loss": 0.8919, + "step": 23393 + }, + { + "epoch": 1.73, + "learning_rate": 1.6704693337460153e-05, + "loss": 0.9918, + "step": 23394 + }, + { + "epoch": 1.73, + "learning_rate": 1.6704397394416637e-05, + "loss": 0.9246, + "step": 23395 + }, + { + "epoch": 1.73, + "learning_rate": 1.6704101440706524e-05, + "loss": 1.0735, + "step": 23396 + }, + { + "epoch": 1.73, + "learning_rate": 1.670380547633028e-05, + "loss": 1.0603, + "step": 23397 + }, + { + "epoch": 1.73, + "learning_rate": 1.6703509501288384e-05, + "loss": 1.0631, + "step": 23398 + }, + { + "epoch": 1.73, + "learning_rate": 1.6703213515581293e-05, + "loss": 1.0825, + "step": 23399 + }, + { + "epoch": 1.73, + "learning_rate": 1.6702917519209488e-05, + "loss": 0.9619, + "step": 23400 + }, + { + "epoch": 1.73, + "learning_rate": 1.6702621512173436e-05, + "loss": 0.9073, + "step": 23401 + }, + { + "epoch": 1.73, + "learning_rate": 1.670232549447361e-05, + "loss": 0.9832, + "step": 23402 + }, + { + "epoch": 1.73, + "learning_rate": 1.6702029466110477e-05, + "loss": 0.9944, + "step": 23403 + }, + { + "epoch": 1.73, + "learning_rate": 1.6701733427084515e-05, + "loss": 1.0371, + "step": 23404 + }, + { + "epoch": 1.73, + "learning_rate": 1.670143737739619e-05, + "loss": 1.1135, + "step": 23405 + }, + { + "epoch": 1.73, + "learning_rate": 1.6701141317045974e-05, + "loss": 1.0584, + "step": 23406 + }, + { + "epoch": 1.73, + "learning_rate": 1.6700845246034337e-05, + "loss": 1.0104, + "step": 23407 + }, + { + "epoch": 1.73, + "learning_rate": 1.6700549164361754e-05, + "loss": 1.0088, + "step": 23408 + }, + { + "epoch": 1.73, + "learning_rate": 1.6700253072028688e-05, + "loss": 1.0637, + "step": 23409 + }, + { + "epoch": 1.73, + "learning_rate": 1.669995696903562e-05, + "loss": 0.9466, + "step": 23410 + }, + { + "epoch": 1.73, + "learning_rate": 1.6699660855383014e-05, + "loss": 1.0143, + "step": 23411 + }, + { + "epoch": 1.73, + "learning_rate": 1.6699364731071345e-05, + "loss": 0.9864, + "step": 23412 + }, + { + "epoch": 1.73, + "learning_rate": 1.669906859610108e-05, + "loss": 1.0562, + "step": 23413 + }, + { + "epoch": 1.73, + "learning_rate": 1.6698772450472694e-05, + "loss": 0.9882, + "step": 23414 + }, + { + "epoch": 1.73, + "learning_rate": 1.6698476294186656e-05, + "loss": 1.1739, + "step": 23415 + }, + { + "epoch": 1.73, + "learning_rate": 1.669818012724344e-05, + "loss": 1.0401, + "step": 23416 + }, + { + "epoch": 1.73, + "learning_rate": 1.6697883949643516e-05, + "loss": 1.0663, + "step": 23417 + }, + { + "epoch": 1.73, + "learning_rate": 1.6697587761387354e-05, + "loss": 1.0985, + "step": 23418 + }, + { + "epoch": 1.73, + "learning_rate": 1.6697291562475423e-05, + "loss": 1.0198, + "step": 23419 + }, + { + "epoch": 1.73, + "learning_rate": 1.66969953529082e-05, + "loss": 1.0067, + "step": 23420 + }, + { + "epoch": 1.73, + "learning_rate": 1.669669913268615e-05, + "loss": 0.9714, + "step": 23421 + }, + { + "epoch": 1.73, + "learning_rate": 1.669640290180975e-05, + "loss": 1.0648, + "step": 23422 + }, + { + "epoch": 1.73, + "learning_rate": 1.6696106660279466e-05, + "loss": 1.031, + "step": 23423 + }, + { + "epoch": 1.73, + "learning_rate": 1.6695810408095776e-05, + "loss": 0.9451, + "step": 23424 + }, + { + "epoch": 1.73, + "learning_rate": 1.6695514145259145e-05, + "loss": 1.0328, + "step": 23425 + }, + { + "epoch": 1.73, + "learning_rate": 1.6695217871770045e-05, + "loss": 0.9551, + "step": 23426 + }, + { + "epoch": 1.73, + "learning_rate": 1.6694921587628954e-05, + "loss": 1.0266, + "step": 23427 + }, + { + "epoch": 1.73, + "learning_rate": 1.669462529283634e-05, + "loss": 1.0096, + "step": 23428 + }, + { + "epoch": 1.73, + "learning_rate": 1.6694328987392666e-05, + "loss": 0.9718, + "step": 23429 + }, + { + "epoch": 1.73, + "learning_rate": 1.6694032671298412e-05, + "loss": 1.0588, + "step": 23430 + }, + { + "epoch": 1.73, + "learning_rate": 1.669373634455405e-05, + "loss": 1.0901, + "step": 23431 + }, + { + "epoch": 1.73, + "learning_rate": 1.669344000716005e-05, + "loss": 1.1051, + "step": 23432 + }, + { + "epoch": 1.73, + "learning_rate": 1.6693143659116877e-05, + "loss": 0.9293, + "step": 23433 + }, + { + "epoch": 1.73, + "learning_rate": 1.6692847300425015e-05, + "loss": 1.1247, + "step": 23434 + }, + { + "epoch": 1.73, + "learning_rate": 1.6692550931084924e-05, + "loss": 1.083, + "step": 23435 + }, + { + "epoch": 1.73, + "learning_rate": 1.669225455109708e-05, + "loss": 0.9185, + "step": 23436 + }, + { + "epoch": 1.73, + "learning_rate": 1.669195816046196e-05, + "loss": 0.979, + "step": 23437 + }, + { + "epoch": 1.73, + "learning_rate": 1.6691661759180026e-05, + "loss": 1.0246, + "step": 23438 + }, + { + "epoch": 1.73, + "learning_rate": 1.6691365347251757e-05, + "loss": 1.0403, + "step": 23439 + }, + { + "epoch": 1.73, + "learning_rate": 1.6691068924677618e-05, + "loss": 1.0444, + "step": 23440 + }, + { + "epoch": 1.73, + "learning_rate": 1.6690772491458087e-05, + "loss": 1.1689, + "step": 23441 + }, + { + "epoch": 1.73, + "learning_rate": 1.6690476047593633e-05, + "loss": 0.944, + "step": 23442 + }, + { + "epoch": 1.73, + "learning_rate": 1.6690179593084725e-05, + "loss": 1.1142, + "step": 23443 + }, + { + "epoch": 1.73, + "learning_rate": 1.6689883127931836e-05, + "loss": 1.0344, + "step": 23444 + }, + { + "epoch": 1.73, + "learning_rate": 1.6689586652135442e-05, + "loss": 0.9501, + "step": 23445 + }, + { + "epoch": 1.73, + "learning_rate": 1.668929016569601e-05, + "loss": 0.9331, + "step": 23446 + }, + { + "epoch": 1.73, + "learning_rate": 1.6688993668614012e-05, + "loss": 1.0615, + "step": 23447 + }, + { + "epoch": 1.73, + "learning_rate": 1.668869716088992e-05, + "loss": 1.1479, + "step": 23448 + }, + { + "epoch": 1.73, + "learning_rate": 1.668840064252421e-05, + "loss": 1.0564, + "step": 23449 + }, + { + "epoch": 1.73, + "learning_rate": 1.6688104113517348e-05, + "loss": 1.04, + "step": 23450 + }, + { + "epoch": 1.73, + "learning_rate": 1.6687807573869808e-05, + "loss": 0.9845, + "step": 23451 + }, + { + "epoch": 1.73, + "learning_rate": 1.6687511023582064e-05, + "loss": 1.0401, + "step": 23452 + }, + { + "epoch": 1.73, + "learning_rate": 1.6687214462654583e-05, + "loss": 0.9603, + "step": 23453 + }, + { + "epoch": 1.73, + "learning_rate": 1.668691789108784e-05, + "loss": 1.065, + "step": 23454 + }, + { + "epoch": 1.73, + "learning_rate": 1.668662130888231e-05, + "loss": 1.0321, + "step": 23455 + }, + { + "epoch": 1.73, + "learning_rate": 1.6686324716038457e-05, + "loss": 0.9858, + "step": 23456 + }, + { + "epoch": 1.73, + "learning_rate": 1.6686028112556756e-05, + "loss": 0.9813, + "step": 23457 + }, + { + "epoch": 1.73, + "learning_rate": 1.668573149843768e-05, + "loss": 1.0254, + "step": 23458 + }, + { + "epoch": 1.73, + "learning_rate": 1.6685434873681706e-05, + "loss": 0.9632, + "step": 23459 + }, + { + "epoch": 1.73, + "learning_rate": 1.6685138238289295e-05, + "loss": 0.9665, + "step": 23460 + }, + { + "epoch": 1.73, + "learning_rate": 1.6684841592260926e-05, + "loss": 1.1101, + "step": 23461 + }, + { + "epoch": 1.73, + "learning_rate": 1.6684544935597072e-05, + "loss": 0.9854, + "step": 23462 + }, + { + "epoch": 1.73, + "learning_rate": 1.66842482682982e-05, + "loss": 0.8481, + "step": 23463 + }, + { + "epoch": 1.73, + "learning_rate": 1.668395159036479e-05, + "loss": 0.9791, + "step": 23464 + }, + { + "epoch": 1.73, + "learning_rate": 1.6683654901797305e-05, + "loss": 0.9893, + "step": 23465 + }, + { + "epoch": 1.73, + "learning_rate": 1.668335820259622e-05, + "loss": 0.9655, + "step": 23466 + }, + { + "epoch": 1.73, + "learning_rate": 1.6683061492762006e-05, + "loss": 0.9366, + "step": 23467 + }, + { + "epoch": 1.73, + "learning_rate": 1.6682764772295138e-05, + "loss": 0.9837, + "step": 23468 + }, + { + "epoch": 1.73, + "learning_rate": 1.668246804119609e-05, + "loss": 1.1256, + "step": 23469 + }, + { + "epoch": 1.73, + "learning_rate": 1.668217129946533e-05, + "loss": 1.0555, + "step": 23470 + }, + { + "epoch": 1.73, + "learning_rate": 1.6681874547103327e-05, + "loss": 1.0429, + "step": 23471 + }, + { + "epoch": 1.73, + "learning_rate": 1.668157778411056e-05, + "loss": 1.0397, + "step": 23472 + }, + { + "epoch": 1.73, + "learning_rate": 1.66812810104875e-05, + "loss": 1.0599, + "step": 23473 + }, + { + "epoch": 1.73, + "learning_rate": 1.6680984226234614e-05, + "loss": 1.0627, + "step": 23474 + }, + { + "epoch": 1.73, + "learning_rate": 1.668068743135238e-05, + "loss": 1.1177, + "step": 23475 + }, + { + "epoch": 1.73, + "learning_rate": 1.6680390625841267e-05, + "loss": 1.0128, + "step": 23476 + }, + { + "epoch": 1.73, + "learning_rate": 1.668009380970175e-05, + "loss": 0.9139, + "step": 23477 + }, + { + "epoch": 1.73, + "learning_rate": 1.6679796982934296e-05, + "loss": 0.955, + "step": 23478 + }, + { + "epoch": 1.73, + "learning_rate": 1.6679500145539385e-05, + "loss": 1.0932, + "step": 23479 + }, + { + "epoch": 1.73, + "learning_rate": 1.6679203297517483e-05, + "loss": 0.9339, + "step": 23480 + }, + { + "epoch": 1.73, + "learning_rate": 1.667890643886906e-05, + "loss": 0.9889, + "step": 23481 + }, + { + "epoch": 1.74, + "learning_rate": 1.6678609569594598e-05, + "loss": 0.8772, + "step": 23482 + }, + { + "epoch": 1.74, + "learning_rate": 1.6678312689694563e-05, + "loss": 1.007, + "step": 23483 + }, + { + "epoch": 1.74, + "learning_rate": 1.6678015799169426e-05, + "loss": 1.0475, + "step": 23484 + }, + { + "epoch": 1.74, + "learning_rate": 1.6677718898019662e-05, + "loss": 1.0184, + "step": 23485 + }, + { + "epoch": 1.74, + "learning_rate": 1.6677421986245744e-05, + "loss": 0.9212, + "step": 23486 + }, + { + "epoch": 1.74, + "learning_rate": 1.667712506384814e-05, + "loss": 1.0884, + "step": 23487 + }, + { + "epoch": 1.74, + "learning_rate": 1.6676828130827328e-05, + "loss": 0.9458, + "step": 23488 + }, + { + "epoch": 1.74, + "learning_rate": 1.6676531187183775e-05, + "loss": 0.9411, + "step": 23489 + }, + { + "epoch": 1.74, + "learning_rate": 1.667623423291796e-05, + "loss": 1.0901, + "step": 23490 + }, + { + "epoch": 1.74, + "learning_rate": 1.667593726803035e-05, + "loss": 1.0745, + "step": 23491 + }, + { + "epoch": 1.74, + "learning_rate": 1.6675640292521422e-05, + "loss": 1.0813, + "step": 23492 + }, + { + "epoch": 1.74, + "learning_rate": 1.6675343306391645e-05, + "loss": 1.0915, + "step": 23493 + }, + { + "epoch": 1.74, + "learning_rate": 1.667504630964149e-05, + "loss": 1.0099, + "step": 23494 + }, + { + "epoch": 1.74, + "learning_rate": 1.6674749302271434e-05, + "loss": 1.0462, + "step": 23495 + }, + { + "epoch": 1.74, + "learning_rate": 1.6674452284281946e-05, + "loss": 1.0243, + "step": 23496 + }, + { + "epoch": 1.74, + "learning_rate": 1.6674155255673497e-05, + "loss": 1.0017, + "step": 23497 + }, + { + "epoch": 1.74, + "learning_rate": 1.667385821644657e-05, + "loss": 1.0, + "step": 23498 + }, + { + "epoch": 1.74, + "learning_rate": 1.6673561166601625e-05, + "loss": 1.0038, + "step": 23499 + }, + { + "epoch": 1.74, + "learning_rate": 1.6673264106139137e-05, + "loss": 1.092, + "step": 23500 + }, + { + "epoch": 1.74, + "learning_rate": 1.667296703505959e-05, + "loss": 0.9279, + "step": 23501 + }, + { + "epoch": 1.74, + "learning_rate": 1.6672669953363442e-05, + "loss": 1.0871, + "step": 23502 + }, + { + "epoch": 1.74, + "learning_rate": 1.6672372861051173e-05, + "loss": 1.0536, + "step": 23503 + }, + { + "epoch": 1.74, + "learning_rate": 1.667207575812325e-05, + "loss": 1.0034, + "step": 23504 + }, + { + "epoch": 1.74, + "learning_rate": 1.6671778644580155e-05, + "loss": 1.0434, + "step": 23505 + }, + { + "epoch": 1.74, + "learning_rate": 1.6671481520422355e-05, + "loss": 1.0938, + "step": 23506 + }, + { + "epoch": 1.74, + "learning_rate": 1.667118438565032e-05, + "loss": 0.9472, + "step": 23507 + }, + { + "epoch": 1.74, + "learning_rate": 1.667088724026453e-05, + "loss": 1.0331, + "step": 23508 + }, + { + "epoch": 1.74, + "learning_rate": 1.667059008426545e-05, + "loss": 1.0267, + "step": 23509 + }, + { + "epoch": 1.74, + "learning_rate": 1.6670292917653564e-05, + "loss": 1.0792, + "step": 23510 + }, + { + "epoch": 1.74, + "learning_rate": 1.666999574042933e-05, + "loss": 1.0153, + "step": 23511 + }, + { + "epoch": 1.74, + "learning_rate": 1.666969855259323e-05, + "loss": 0.9916, + "step": 23512 + }, + { + "epoch": 1.74, + "learning_rate": 1.6669401354145735e-05, + "loss": 0.9829, + "step": 23513 + }, + { + "epoch": 1.74, + "learning_rate": 1.6669104145087322e-05, + "loss": 0.9963, + "step": 23514 + }, + { + "epoch": 1.74, + "learning_rate": 1.6668806925418452e-05, + "loss": 1.0323, + "step": 23515 + }, + { + "epoch": 1.74, + "learning_rate": 1.666850969513961e-05, + "loss": 1.024, + "step": 23516 + }, + { + "epoch": 1.74, + "learning_rate": 1.6668212454251264e-05, + "loss": 1.0375, + "step": 23517 + }, + { + "epoch": 1.74, + "learning_rate": 1.6667915202753888e-05, + "loss": 0.9953, + "step": 23518 + }, + { + "epoch": 1.74, + "learning_rate": 1.6667617940647956e-05, + "loss": 1.0149, + "step": 23519 + }, + { + "epoch": 1.74, + "learning_rate": 1.6667320667933936e-05, + "loss": 0.8728, + "step": 23520 + }, + { + "epoch": 1.74, + "learning_rate": 1.6667023384612306e-05, + "loss": 0.9362, + "step": 23521 + }, + { + "epoch": 1.74, + "learning_rate": 1.6666726090683536e-05, + "loss": 1.0852, + "step": 23522 + }, + { + "epoch": 1.74, + "learning_rate": 1.66664287861481e-05, + "loss": 0.9751, + "step": 23523 + }, + { + "epoch": 1.74, + "learning_rate": 1.666613147100647e-05, + "loss": 0.9721, + "step": 23524 + }, + { + "epoch": 1.74, + "learning_rate": 1.6665834145259125e-05, + "loss": 1.0607, + "step": 23525 + }, + { + "epoch": 1.74, + "learning_rate": 1.666553680890653e-05, + "loss": 0.9225, + "step": 23526 + }, + { + "epoch": 1.74, + "learning_rate": 1.666523946194916e-05, + "loss": 1.0283, + "step": 23527 + }, + { + "epoch": 1.74, + "learning_rate": 1.6664942104387495e-05, + "loss": 0.9338, + "step": 23528 + }, + { + "epoch": 1.74, + "learning_rate": 1.6664644736221996e-05, + "loss": 1.0157, + "step": 23529 + }, + { + "epoch": 1.74, + "learning_rate": 1.6664347357453146e-05, + "loss": 1.1023, + "step": 23530 + }, + { + "epoch": 1.74, + "learning_rate": 1.6664049968081415e-05, + "loss": 0.9713, + "step": 23531 + }, + { + "epoch": 1.74, + "learning_rate": 1.6663752568107275e-05, + "loss": 1.0925, + "step": 23532 + }, + { + "epoch": 1.74, + "learning_rate": 1.66634551575312e-05, + "loss": 0.9313, + "step": 23533 + }, + { + "epoch": 1.74, + "learning_rate": 1.666315773635366e-05, + "loss": 1.0397, + "step": 23534 + }, + { + "epoch": 1.74, + "learning_rate": 1.6662860304575135e-05, + "loss": 1.0559, + "step": 23535 + }, + { + "epoch": 1.74, + "learning_rate": 1.6662562862196095e-05, + "loss": 0.9629, + "step": 23536 + }, + { + "epoch": 1.74, + "learning_rate": 1.666226540921701e-05, + "loss": 1.1067, + "step": 23537 + }, + { + "epoch": 1.74, + "learning_rate": 1.666196794563836e-05, + "loss": 0.9374, + "step": 23538 + }, + { + "epoch": 1.74, + "learning_rate": 1.666167047146061e-05, + "loss": 1.0033, + "step": 23539 + }, + { + "epoch": 1.74, + "learning_rate": 1.6661372986684245e-05, + "loss": 0.9935, + "step": 23540 + }, + { + "epoch": 1.74, + "learning_rate": 1.666107549130972e-05, + "loss": 1.046, + "step": 23541 + }, + { + "epoch": 1.74, + "learning_rate": 1.6660777985337526e-05, + "loss": 0.9547, + "step": 23542 + }, + { + "epoch": 1.74, + "learning_rate": 1.6660480468768127e-05, + "loss": 1.0682, + "step": 23543 + }, + { + "epoch": 1.74, + "learning_rate": 1.6660182941602e-05, + "loss": 0.9828, + "step": 23544 + }, + { + "epoch": 1.74, + "learning_rate": 1.665988540383962e-05, + "loss": 1.0233, + "step": 23545 + }, + { + "epoch": 1.74, + "learning_rate": 1.6659587855481454e-05, + "loss": 1.0052, + "step": 23546 + }, + { + "epoch": 1.74, + "learning_rate": 1.665929029652798e-05, + "loss": 1.0383, + "step": 23547 + }, + { + "epoch": 1.74, + "learning_rate": 1.6658992726979673e-05, + "loss": 0.9617, + "step": 23548 + }, + { + "epoch": 1.74, + "learning_rate": 1.6658695146837e-05, + "loss": 0.959, + "step": 23549 + }, + { + "epoch": 1.74, + "learning_rate": 1.665839755610044e-05, + "loss": 1.0121, + "step": 23550 + }, + { + "epoch": 1.74, + "learning_rate": 1.6658099954770462e-05, + "loss": 0.948, + "step": 23551 + }, + { + "epoch": 1.74, + "learning_rate": 1.6657802342847546e-05, + "loss": 1.0777, + "step": 23552 + }, + { + "epoch": 1.74, + "learning_rate": 1.6657504720332162e-05, + "loss": 0.9318, + "step": 23553 + }, + { + "epoch": 1.74, + "learning_rate": 1.665720708722478e-05, + "loss": 1.0203, + "step": 23554 + }, + { + "epoch": 1.74, + "learning_rate": 1.665690944352588e-05, + "loss": 0.8987, + "step": 23555 + }, + { + "epoch": 1.74, + "learning_rate": 1.6656611789235928e-05, + "loss": 0.9452, + "step": 23556 + }, + { + "epoch": 1.74, + "learning_rate": 1.6656314124355407e-05, + "loss": 1.0011, + "step": 23557 + }, + { + "epoch": 1.74, + "learning_rate": 1.6656016448884783e-05, + "loss": 1.124, + "step": 23558 + }, + { + "epoch": 1.74, + "learning_rate": 1.665571876282453e-05, + "loss": 0.9635, + "step": 23559 + }, + { + "epoch": 1.74, + "learning_rate": 1.6655421066175127e-05, + "loss": 0.9717, + "step": 23560 + }, + { + "epoch": 1.74, + "learning_rate": 1.665512335893704e-05, + "loss": 0.9367, + "step": 23561 + }, + { + "epoch": 1.74, + "learning_rate": 1.665482564111075e-05, + "loss": 0.9814, + "step": 23562 + }, + { + "epoch": 1.74, + "learning_rate": 1.665452791269673e-05, + "loss": 1.1181, + "step": 23563 + }, + { + "epoch": 1.74, + "learning_rate": 1.665423017369545e-05, + "loss": 1.1466, + "step": 23564 + }, + { + "epoch": 1.74, + "learning_rate": 1.665393242410738e-05, + "loss": 0.9013, + "step": 23565 + }, + { + "epoch": 1.74, + "learning_rate": 1.6653634663933005e-05, + "loss": 1.0474, + "step": 23566 + }, + { + "epoch": 1.74, + "learning_rate": 1.665333689317279e-05, + "loss": 1.0126, + "step": 23567 + }, + { + "epoch": 1.74, + "learning_rate": 1.6653039111827212e-05, + "loss": 1.0727, + "step": 23568 + }, + { + "epoch": 1.74, + "learning_rate": 1.6652741319896744e-05, + "loss": 1.0557, + "step": 23569 + }, + { + "epoch": 1.74, + "learning_rate": 1.6652443517381857e-05, + "loss": 1.0223, + "step": 23570 + }, + { + "epoch": 1.74, + "learning_rate": 1.6652145704283033e-05, + "loss": 1.0265, + "step": 23571 + }, + { + "epoch": 1.74, + "learning_rate": 1.6651847880600735e-05, + "loss": 1.0403, + "step": 23572 + }, + { + "epoch": 1.74, + "learning_rate": 1.665155004633544e-05, + "loss": 1.0618, + "step": 23573 + }, + { + "epoch": 1.74, + "learning_rate": 1.665125220148763e-05, + "loss": 1.0114, + "step": 23574 + }, + { + "epoch": 1.74, + "learning_rate": 1.665095434605777e-05, + "loss": 0.8286, + "step": 23575 + }, + { + "epoch": 1.74, + "learning_rate": 1.665065648004634e-05, + "loss": 1.0331, + "step": 23576 + }, + { + "epoch": 1.74, + "learning_rate": 1.6650358603453808e-05, + "loss": 0.9745, + "step": 23577 + }, + { + "epoch": 1.74, + "learning_rate": 1.665006071628065e-05, + "loss": 1.1, + "step": 23578 + }, + { + "epoch": 1.74, + "learning_rate": 1.6649762818527345e-05, + "loss": 0.9935, + "step": 23579 + }, + { + "epoch": 1.74, + "learning_rate": 1.6649464910194356e-05, + "loss": 0.9892, + "step": 23580 + }, + { + "epoch": 1.74, + "learning_rate": 1.6649166991282167e-05, + "loss": 1.0605, + "step": 23581 + }, + { + "epoch": 1.74, + "learning_rate": 1.6648869061791248e-05, + "loss": 0.9038, + "step": 23582 + }, + { + "epoch": 1.74, + "learning_rate": 1.6648571121722073e-05, + "loss": 1.0275, + "step": 23583 + }, + { + "epoch": 1.74, + "learning_rate": 1.6648273171075116e-05, + "loss": 0.9314, + "step": 23584 + }, + { + "epoch": 1.74, + "learning_rate": 1.6647975209850856e-05, + "loss": 1.0339, + "step": 23585 + }, + { + "epoch": 1.74, + "learning_rate": 1.664767723804976e-05, + "loss": 1.0586, + "step": 23586 + }, + { + "epoch": 1.74, + "learning_rate": 1.6647379255672303e-05, + "loss": 0.9726, + "step": 23587 + }, + { + "epoch": 1.74, + "learning_rate": 1.664708126271896e-05, + "loss": 1.0096, + "step": 23588 + }, + { + "epoch": 1.74, + "learning_rate": 1.6646783259190206e-05, + "loss": 0.972, + "step": 23589 + }, + { + "epoch": 1.74, + "learning_rate": 1.664648524508652e-05, + "loss": 0.947, + "step": 23590 + }, + { + "epoch": 1.74, + "learning_rate": 1.6646187220408365e-05, + "loss": 1.0761, + "step": 23591 + }, + { + "epoch": 1.74, + "learning_rate": 1.6645889185156224e-05, + "loss": 0.9243, + "step": 23592 + }, + { + "epoch": 1.74, + "learning_rate": 1.6645591139330564e-05, + "loss": 1.0453, + "step": 23593 + }, + { + "epoch": 1.74, + "learning_rate": 1.664529308293187e-05, + "loss": 0.9313, + "step": 23594 + }, + { + "epoch": 1.74, + "learning_rate": 1.664499501596061e-05, + "loss": 0.9849, + "step": 23595 + }, + { + "epoch": 1.74, + "learning_rate": 1.6644696938417254e-05, + "loss": 0.987, + "step": 23596 + }, + { + "epoch": 1.74, + "learning_rate": 1.664439885030228e-05, + "loss": 1.016, + "step": 23597 + }, + { + "epoch": 1.74, + "learning_rate": 1.6644100751616165e-05, + "loss": 1.017, + "step": 23598 + }, + { + "epoch": 1.74, + "learning_rate": 1.664380264235938e-05, + "loss": 1.0141, + "step": 23599 + }, + { + "epoch": 1.74, + "learning_rate": 1.6643504522532398e-05, + "loss": 0.9942, + "step": 23600 + }, + { + "epoch": 1.74, + "learning_rate": 1.66432063921357e-05, + "loss": 1.2048, + "step": 23601 + }, + { + "epoch": 1.74, + "learning_rate": 1.6642908251169753e-05, + "loss": 1.1334, + "step": 23602 + }, + { + "epoch": 1.74, + "learning_rate": 1.664261009963503e-05, + "loss": 1.0115, + "step": 23603 + }, + { + "epoch": 1.74, + "learning_rate": 1.6642311937532016e-05, + "loss": 0.95, + "step": 23604 + }, + { + "epoch": 1.74, + "learning_rate": 1.6642013764861176e-05, + "loss": 1.0902, + "step": 23605 + }, + { + "epoch": 1.74, + "learning_rate": 1.6641715581622988e-05, + "loss": 0.914, + "step": 23606 + }, + { + "epoch": 1.74, + "learning_rate": 1.664141738781792e-05, + "loss": 0.9878, + "step": 23607 + }, + { + "epoch": 1.74, + "learning_rate": 1.664111918344646e-05, + "loss": 0.9623, + "step": 23608 + }, + { + "epoch": 1.74, + "learning_rate": 1.664082096850907e-05, + "loss": 0.9758, + "step": 23609 + }, + { + "epoch": 1.74, + "learning_rate": 1.664052274300623e-05, + "loss": 0.9842, + "step": 23610 + }, + { + "epoch": 1.74, + "learning_rate": 1.664022450693841e-05, + "loss": 1.101, + "step": 23611 + }, + { + "epoch": 1.74, + "learning_rate": 1.6639926260306094e-05, + "loss": 1.0675, + "step": 23612 + }, + { + "epoch": 1.74, + "learning_rate": 1.6639628003109748e-05, + "loss": 1.0781, + "step": 23613 + }, + { + "epoch": 1.74, + "learning_rate": 1.663932973534985e-05, + "loss": 0.9852, + "step": 23614 + }, + { + "epoch": 1.74, + "learning_rate": 1.6639031457026868e-05, + "loss": 0.9848, + "step": 23615 + }, + { + "epoch": 1.74, + "learning_rate": 1.663873316814129e-05, + "loss": 1.0164, + "step": 23616 + }, + { + "epoch": 1.75, + "learning_rate": 1.6638434868693575e-05, + "loss": 0.9344, + "step": 23617 + }, + { + "epoch": 1.75, + "learning_rate": 1.6638136558684207e-05, + "loss": 1.041, + "step": 23618 + }, + { + "epoch": 1.75, + "learning_rate": 1.663783823811366e-05, + "loss": 1.0239, + "step": 23619 + }, + { + "epoch": 1.75, + "learning_rate": 1.6637539906982407e-05, + "loss": 1.0094, + "step": 23620 + }, + { + "epoch": 1.75, + "learning_rate": 1.6637241565290923e-05, + "loss": 1.0334, + "step": 23621 + }, + { + "epoch": 1.75, + "learning_rate": 1.6636943213039683e-05, + "loss": 1.0215, + "step": 23622 + }, + { + "epoch": 1.75, + "learning_rate": 1.663664485022916e-05, + "loss": 0.9517, + "step": 23623 + }, + { + "epoch": 1.75, + "learning_rate": 1.663634647685983e-05, + "loss": 0.9888, + "step": 23624 + }, + { + "epoch": 1.75, + "learning_rate": 1.663604809293217e-05, + "loss": 1.0206, + "step": 23625 + }, + { + "epoch": 1.75, + "learning_rate": 1.6635749698446653e-05, + "loss": 0.9846, + "step": 23626 + }, + { + "epoch": 1.75, + "learning_rate": 1.663545129340375e-05, + "loss": 1.0498, + "step": 23627 + }, + { + "epoch": 1.75, + "learning_rate": 1.663515287780394e-05, + "loss": 1.0094, + "step": 23628 + }, + { + "epoch": 1.75, + "learning_rate": 1.6634854451647698e-05, + "loss": 0.8819, + "step": 23629 + }, + { + "epoch": 1.75, + "learning_rate": 1.6634556014935496e-05, + "loss": 1.0479, + "step": 23630 + }, + { + "epoch": 1.75, + "learning_rate": 1.663425756766781e-05, + "loss": 1.0359, + "step": 23631 + }, + { + "epoch": 1.75, + "learning_rate": 1.6633959109845118e-05, + "loss": 1.0524, + "step": 23632 + }, + { + "epoch": 1.75, + "learning_rate": 1.6633660641467887e-05, + "loss": 0.9319, + "step": 23633 + }, + { + "epoch": 1.75, + "learning_rate": 1.6633362162536602e-05, + "loss": 0.9702, + "step": 23634 + }, + { + "epoch": 1.75, + "learning_rate": 1.663306367305173e-05, + "loss": 0.9971, + "step": 23635 + }, + { + "epoch": 1.75, + "learning_rate": 1.663276517301375e-05, + "loss": 1.0445, + "step": 23636 + }, + { + "epoch": 1.75, + "learning_rate": 1.6632466662423134e-05, + "loss": 1.0643, + "step": 23637 + }, + { + "epoch": 1.75, + "learning_rate": 1.6632168141280358e-05, + "loss": 1.0352, + "step": 23638 + }, + { + "epoch": 1.75, + "learning_rate": 1.6631869609585896e-05, + "loss": 1.0385, + "step": 23639 + }, + { + "epoch": 1.75, + "learning_rate": 1.663157106734023e-05, + "loss": 0.9047, + "step": 23640 + }, + { + "epoch": 1.75, + "learning_rate": 1.6631272514543826e-05, + "loss": 0.9555, + "step": 23641 + }, + { + "epoch": 1.75, + "learning_rate": 1.6630973951197163e-05, + "loss": 1.0606, + "step": 23642 + }, + { + "epoch": 1.75, + "learning_rate": 1.6630675377300714e-05, + "loss": 1.0427, + "step": 23643 + }, + { + "epoch": 1.75, + "learning_rate": 1.6630376792854955e-05, + "loss": 0.964, + "step": 23644 + }, + { + "epoch": 1.75, + "learning_rate": 1.6630078197860366e-05, + "loss": 1.0151, + "step": 23645 + }, + { + "epoch": 1.75, + "learning_rate": 1.662977959231741e-05, + "loss": 0.9923, + "step": 23646 + }, + { + "epoch": 1.75, + "learning_rate": 1.6629480976226572e-05, + "loss": 0.972, + "step": 23647 + }, + { + "epoch": 1.75, + "learning_rate": 1.662918234958833e-05, + "loss": 0.9257, + "step": 23648 + }, + { + "epoch": 1.75, + "learning_rate": 1.6628883712403148e-05, + "loss": 0.9661, + "step": 23649 + }, + { + "epoch": 1.75, + "learning_rate": 1.662858506467151e-05, + "loss": 1.0437, + "step": 23650 + }, + { + "epoch": 1.75, + "learning_rate": 1.6628286406393885e-05, + "loss": 1.001, + "step": 23651 + }, + { + "epoch": 1.75, + "learning_rate": 1.6627987737570756e-05, + "loss": 0.9813, + "step": 23652 + }, + { + "epoch": 1.75, + "learning_rate": 1.6627689058202588e-05, + "loss": 1.035, + "step": 23653 + }, + { + "epoch": 1.75, + "learning_rate": 1.6627390368289863e-05, + "loss": 0.9175, + "step": 23654 + }, + { + "epoch": 1.75, + "learning_rate": 1.6627091667833055e-05, + "loss": 1.0431, + "step": 23655 + }, + { + "epoch": 1.75, + "learning_rate": 1.6626792956832642e-05, + "loss": 1.0033, + "step": 23656 + }, + { + "epoch": 1.75, + "learning_rate": 1.662649423528909e-05, + "loss": 1.0045, + "step": 23657 + }, + { + "epoch": 1.75, + "learning_rate": 1.6626195503202883e-05, + "loss": 1.0149, + "step": 23658 + }, + { + "epoch": 1.75, + "learning_rate": 1.6625896760574497e-05, + "loss": 1.051, + "step": 23659 + }, + { + "epoch": 1.75, + "learning_rate": 1.6625598007404403e-05, + "loss": 0.9859, + "step": 23660 + }, + { + "epoch": 1.75, + "learning_rate": 1.6625299243693075e-05, + "loss": 1.0297, + "step": 23661 + }, + { + "epoch": 1.75, + "learning_rate": 1.662500046944099e-05, + "loss": 1.0456, + "step": 23662 + }, + { + "epoch": 1.75, + "learning_rate": 1.6624701684648627e-05, + "loss": 0.9874, + "step": 23663 + }, + { + "epoch": 1.75, + "learning_rate": 1.6624402889316458e-05, + "loss": 1.0073, + "step": 23664 + }, + { + "epoch": 1.75, + "learning_rate": 1.6624104083444956e-05, + "loss": 0.899, + "step": 23665 + }, + { + "epoch": 1.75, + "learning_rate": 1.6623805267034603e-05, + "loss": 0.9814, + "step": 23666 + }, + { + "epoch": 1.75, + "learning_rate": 1.662350644008587e-05, + "loss": 0.9312, + "step": 23667 + }, + { + "epoch": 1.75, + "learning_rate": 1.6623207602599228e-05, + "loss": 1.1032, + "step": 23668 + }, + { + "epoch": 1.75, + "learning_rate": 1.6622908754575166e-05, + "loss": 1.0957, + "step": 23669 + }, + { + "epoch": 1.75, + "learning_rate": 1.6622609896014142e-05, + "loss": 0.9618, + "step": 23670 + }, + { + "epoch": 1.75, + "learning_rate": 1.6622311026916646e-05, + "loss": 0.909, + "step": 23671 + }, + { + "epoch": 1.75, + "learning_rate": 1.662201214728315e-05, + "loss": 1.139, + "step": 23672 + }, + { + "epoch": 1.75, + "learning_rate": 1.662171325711412e-05, + "loss": 1.0577, + "step": 23673 + }, + { + "epoch": 1.75, + "learning_rate": 1.6621414356410043e-05, + "loss": 1.0975, + "step": 23674 + }, + { + "epoch": 1.75, + "learning_rate": 1.6621115445171393e-05, + "loss": 1.0858, + "step": 23675 + }, + { + "epoch": 1.75, + "learning_rate": 1.662081652339864e-05, + "loss": 0.9777, + "step": 23676 + }, + { + "epoch": 1.75, + "learning_rate": 1.6620517591092263e-05, + "loss": 1.0643, + "step": 23677 + }, + { + "epoch": 1.75, + "learning_rate": 1.6620218648252736e-05, + "loss": 1.0284, + "step": 23678 + }, + { + "epoch": 1.75, + "learning_rate": 1.661991969488054e-05, + "loss": 1.0638, + "step": 23679 + }, + { + "epoch": 1.75, + "learning_rate": 1.6619620730976142e-05, + "loss": 1.0206, + "step": 23680 + }, + { + "epoch": 1.75, + "learning_rate": 1.6619321756540022e-05, + "loss": 0.966, + "step": 23681 + }, + { + "epoch": 1.75, + "learning_rate": 1.661902277157266e-05, + "loss": 1.0131, + "step": 23682 + }, + { + "epoch": 1.75, + "learning_rate": 1.6618723776074524e-05, + "loss": 1.0055, + "step": 23683 + }, + { + "epoch": 1.75, + "learning_rate": 1.6618424770046095e-05, + "loss": 1.0236, + "step": 23684 + }, + { + "epoch": 1.75, + "learning_rate": 1.6618125753487846e-05, + "loss": 1.0576, + "step": 23685 + }, + { + "epoch": 1.75, + "learning_rate": 1.6617826726400253e-05, + "loss": 1.0918, + "step": 23686 + }, + { + "epoch": 1.75, + "learning_rate": 1.6617527688783793e-05, + "loss": 0.9977, + "step": 23687 + }, + { + "epoch": 1.75, + "learning_rate": 1.6617228640638944e-05, + "loss": 1.0775, + "step": 23688 + }, + { + "epoch": 1.75, + "learning_rate": 1.6616929581966176e-05, + "loss": 0.9372, + "step": 23689 + }, + { + "epoch": 1.75, + "learning_rate": 1.661663051276597e-05, + "loss": 1.062, + "step": 23690 + }, + { + "epoch": 1.75, + "learning_rate": 1.661633143303879e-05, + "loss": 1.021, + "step": 23691 + }, + { + "epoch": 1.75, + "learning_rate": 1.6616032342785132e-05, + "loss": 1.0432, + "step": 23692 + }, + { + "epoch": 1.75, + "learning_rate": 1.661573324200546e-05, + "loss": 0.9481, + "step": 23693 + }, + { + "epoch": 1.75, + "learning_rate": 1.6615434130700246e-05, + "loss": 0.9863, + "step": 23694 + }, + { + "epoch": 1.75, + "learning_rate": 1.6615135008869975e-05, + "loss": 1.0095, + "step": 23695 + }, + { + "epoch": 1.75, + "learning_rate": 1.6614835876515116e-05, + "loss": 1.0943, + "step": 23696 + }, + { + "epoch": 1.75, + "learning_rate": 1.6614536733636154e-05, + "loss": 1.0091, + "step": 23697 + }, + { + "epoch": 1.75, + "learning_rate": 1.6614237580233553e-05, + "loss": 0.9585, + "step": 23698 + }, + { + "epoch": 1.75, + "learning_rate": 1.6613938416307794e-05, + "loss": 1.0266, + "step": 23699 + }, + { + "epoch": 1.75, + "learning_rate": 1.6613639241859355e-05, + "loss": 1.0942, + "step": 23700 + }, + { + "epoch": 1.75, + "learning_rate": 1.6613340056888713e-05, + "loss": 0.9074, + "step": 23701 + }, + { + "epoch": 1.75, + "learning_rate": 1.661304086139634e-05, + "loss": 0.9414, + "step": 23702 + }, + { + "epoch": 1.75, + "learning_rate": 1.6612741655382713e-05, + "loss": 1.0711, + "step": 23703 + }, + { + "epoch": 1.75, + "learning_rate": 1.6612442438848307e-05, + "loss": 1.0526, + "step": 23704 + }, + { + "epoch": 1.75, + "learning_rate": 1.6612143211793603e-05, + "loss": 1.0062, + "step": 23705 + }, + { + "epoch": 1.75, + "learning_rate": 1.661184397421907e-05, + "loss": 0.9834, + "step": 23706 + }, + { + "epoch": 1.75, + "learning_rate": 1.6611544726125193e-05, + "loss": 1.0199, + "step": 23707 + }, + { + "epoch": 1.75, + "learning_rate": 1.661124546751244e-05, + "loss": 0.9494, + "step": 23708 + }, + { + "epoch": 1.75, + "learning_rate": 1.6610946198381292e-05, + "loss": 1.0462, + "step": 23709 + }, + { + "epoch": 1.75, + "learning_rate": 1.6610646918732222e-05, + "loss": 1.0266, + "step": 23710 + }, + { + "epoch": 1.75, + "learning_rate": 1.6610347628565702e-05, + "loss": 0.995, + "step": 23711 + }, + { + "epoch": 1.75, + "learning_rate": 1.6610048327882222e-05, + "loss": 0.8768, + "step": 23712 + }, + { + "epoch": 1.75, + "learning_rate": 1.6609749016682245e-05, + "loss": 0.9675, + "step": 23713 + }, + { + "epoch": 1.75, + "learning_rate": 1.6609449694966252e-05, + "loss": 1.1276, + "step": 23714 + }, + { + "epoch": 1.75, + "learning_rate": 1.660915036273472e-05, + "loss": 1.0006, + "step": 23715 + }, + { + "epoch": 1.75, + "learning_rate": 1.6608851019988126e-05, + "loss": 1.0282, + "step": 23716 + }, + { + "epoch": 1.75, + "learning_rate": 1.6608551666726943e-05, + "loss": 1.0233, + "step": 23717 + }, + { + "epoch": 1.75, + "learning_rate": 1.6608252302951647e-05, + "loss": 1.087, + "step": 23718 + }, + { + "epoch": 1.75, + "learning_rate": 1.660795292866272e-05, + "loss": 1.0262, + "step": 23719 + }, + { + "epoch": 1.75, + "learning_rate": 1.6607653543860633e-05, + "loss": 0.9545, + "step": 23720 + }, + { + "epoch": 1.75, + "learning_rate": 1.6607354148545862e-05, + "loss": 1.1084, + "step": 23721 + }, + { + "epoch": 1.75, + "learning_rate": 1.6607054742718888e-05, + "loss": 1.0499, + "step": 23722 + }, + { + "epoch": 1.75, + "learning_rate": 1.660675532638018e-05, + "loss": 1.0339, + "step": 23723 + }, + { + "epoch": 1.75, + "learning_rate": 1.660645589953022e-05, + "loss": 1.0785, + "step": 23724 + }, + { + "epoch": 1.75, + "learning_rate": 1.6606156462169484e-05, + "loss": 1.0468, + "step": 23725 + }, + { + "epoch": 1.75, + "learning_rate": 1.660585701429845e-05, + "loss": 0.9644, + "step": 23726 + }, + { + "epoch": 1.75, + "learning_rate": 1.660555755591759e-05, + "loss": 1.0033, + "step": 23727 + }, + { + "epoch": 1.75, + "learning_rate": 1.6605258087027384e-05, + "loss": 0.9037, + "step": 23728 + }, + { + "epoch": 1.75, + "learning_rate": 1.6604958607628307e-05, + "loss": 0.986, + "step": 23729 + }, + { + "epoch": 1.75, + "learning_rate": 1.660465911772083e-05, + "loss": 1.0798, + "step": 23730 + }, + { + "epoch": 1.75, + "learning_rate": 1.660435961730544e-05, + "loss": 0.9594, + "step": 23731 + }, + { + "epoch": 1.75, + "learning_rate": 1.6604060106382608e-05, + "loss": 1.0977, + "step": 23732 + }, + { + "epoch": 1.75, + "learning_rate": 1.6603760584952807e-05, + "loss": 1.0327, + "step": 23733 + }, + { + "epoch": 1.75, + "learning_rate": 1.6603461053016522e-05, + "loss": 1.0823, + "step": 23734 + }, + { + "epoch": 1.75, + "learning_rate": 1.6603161510574222e-05, + "loss": 1.0738, + "step": 23735 + }, + { + "epoch": 1.75, + "learning_rate": 1.6602861957626387e-05, + "loss": 1.2292, + "step": 23736 + }, + { + "epoch": 1.75, + "learning_rate": 1.6602562394173496e-05, + "loss": 0.9899, + "step": 23737 + }, + { + "epoch": 1.75, + "learning_rate": 1.660226282021602e-05, + "loss": 1.0333, + "step": 23738 + }, + { + "epoch": 1.75, + "learning_rate": 1.660196323575444e-05, + "loss": 1.0522, + "step": 23739 + }, + { + "epoch": 1.75, + "learning_rate": 1.660166364078923e-05, + "loss": 1.0251, + "step": 23740 + }, + { + "epoch": 1.75, + "learning_rate": 1.6601364035320865e-05, + "loss": 0.9688, + "step": 23741 + }, + { + "epoch": 1.75, + "learning_rate": 1.6601064419349828e-05, + "loss": 1.0519, + "step": 23742 + }, + { + "epoch": 1.75, + "learning_rate": 1.6600764792876588e-05, + "loss": 0.9379, + "step": 23743 + }, + { + "epoch": 1.75, + "learning_rate": 1.660046515590163e-05, + "loss": 1.0444, + "step": 23744 + }, + { + "epoch": 1.75, + "learning_rate": 1.6600165508425426e-05, + "loss": 1.1181, + "step": 23745 + }, + { + "epoch": 1.75, + "learning_rate": 1.6599865850448453e-05, + "loss": 1.04, + "step": 23746 + }, + { + "epoch": 1.75, + "learning_rate": 1.659956618197118e-05, + "loss": 0.9994, + "step": 23747 + }, + { + "epoch": 1.75, + "learning_rate": 1.6599266502994103e-05, + "loss": 1.086, + "step": 23748 + }, + { + "epoch": 1.75, + "learning_rate": 1.659896681351768e-05, + "loss": 0.9672, + "step": 23749 + }, + { + "epoch": 1.75, + "learning_rate": 1.65986671135424e-05, + "loss": 1.0514, + "step": 23750 + }, + { + "epoch": 1.75, + "learning_rate": 1.6598367403068732e-05, + "loss": 0.9844, + "step": 23751 + }, + { + "epoch": 1.76, + "learning_rate": 1.6598067682097156e-05, + "loss": 1.0903, + "step": 23752 + }, + { + "epoch": 1.76, + "learning_rate": 1.6597767950628152e-05, + "loss": 0.9569, + "step": 23753 + }, + { + "epoch": 1.76, + "learning_rate": 1.659746820866219e-05, + "loss": 1.0488, + "step": 23754 + }, + { + "epoch": 1.76, + "learning_rate": 1.659716845619975e-05, + "loss": 1.0487, + "step": 23755 + }, + { + "epoch": 1.76, + "learning_rate": 1.659686869324131e-05, + "loss": 0.9699, + "step": 23756 + }, + { + "epoch": 1.76, + "learning_rate": 1.659656891978735e-05, + "loss": 0.8966, + "step": 23757 + }, + { + "epoch": 1.76, + "learning_rate": 1.6596269135838343e-05, + "loss": 1.0417, + "step": 23758 + }, + { + "epoch": 1.76, + "learning_rate": 1.6595969341394763e-05, + "loss": 1.0833, + "step": 23759 + }, + { + "epoch": 1.76, + "learning_rate": 1.659566953645709e-05, + "loss": 0.9923, + "step": 23760 + }, + { + "epoch": 1.76, + "learning_rate": 1.65953697210258e-05, + "loss": 1.0583, + "step": 23761 + }, + { + "epoch": 1.76, + "learning_rate": 1.6595069895101374e-05, + "loss": 1.0403, + "step": 23762 + }, + { + "epoch": 1.76, + "learning_rate": 1.6594770058684286e-05, + "loss": 1.0188, + "step": 23763 + }, + { + "epoch": 1.76, + "learning_rate": 1.6594470211775015e-05, + "loss": 0.9796, + "step": 23764 + }, + { + "epoch": 1.76, + "learning_rate": 1.659417035437403e-05, + "loss": 0.9673, + "step": 23765 + }, + { + "epoch": 1.76, + "learning_rate": 1.6593870486481818e-05, + "loss": 0.9646, + "step": 23766 + }, + { + "epoch": 1.76, + "learning_rate": 1.6593570608098852e-05, + "loss": 0.944, + "step": 23767 + }, + { + "epoch": 1.76, + "learning_rate": 1.659327071922561e-05, + "loss": 1.127, + "step": 23768 + }, + { + "epoch": 1.76, + "learning_rate": 1.659297081986257e-05, + "loss": 0.9581, + "step": 23769 + }, + { + "epoch": 1.76, + "learning_rate": 1.6592670910010205e-05, + "loss": 1.1196, + "step": 23770 + }, + { + "epoch": 1.76, + "learning_rate": 1.6592370989668995e-05, + "loss": 1.0296, + "step": 23771 + }, + { + "epoch": 1.76, + "learning_rate": 1.6592071058839422e-05, + "loss": 1.0641, + "step": 23772 + }, + { + "epoch": 1.76, + "learning_rate": 1.6591771117521952e-05, + "loss": 0.9772, + "step": 23773 + }, + { + "epoch": 1.76, + "learning_rate": 1.659147116571707e-05, + "loss": 1.0324, + "step": 23774 + }, + { + "epoch": 1.76, + "learning_rate": 1.6591171203425254e-05, + "loss": 0.9773, + "step": 23775 + }, + { + "epoch": 1.76, + "learning_rate": 1.6590871230646978e-05, + "loss": 1.1314, + "step": 23776 + }, + { + "epoch": 1.76, + "learning_rate": 1.659057124738272e-05, + "loss": 1.026, + "step": 23777 + }, + { + "epoch": 1.76, + "learning_rate": 1.6590271253632954e-05, + "loss": 0.9978, + "step": 23778 + }, + { + "epoch": 1.76, + "learning_rate": 1.6589971249398168e-05, + "loss": 1.0112, + "step": 23779 + }, + { + "epoch": 1.76, + "learning_rate": 1.6589671234678825e-05, + "loss": 1.072, + "step": 23780 + }, + { + "epoch": 1.76, + "learning_rate": 1.6589371209475413e-05, + "loss": 0.8831, + "step": 23781 + }, + { + "epoch": 1.76, + "learning_rate": 1.6589071173788407e-05, + "loss": 0.9507, + "step": 23782 + }, + { + "epoch": 1.76, + "learning_rate": 1.658877112761828e-05, + "loss": 0.9087, + "step": 23783 + }, + { + "epoch": 1.76, + "learning_rate": 1.6588471070965514e-05, + "loss": 1.0316, + "step": 23784 + }, + { + "epoch": 1.76, + "learning_rate": 1.6588171003830583e-05, + "loss": 1.0005, + "step": 23785 + }, + { + "epoch": 1.76, + "learning_rate": 1.6587870926213965e-05, + "loss": 0.9829, + "step": 23786 + }, + { + "epoch": 1.76, + "learning_rate": 1.658757083811614e-05, + "loss": 0.9835, + "step": 23787 + }, + { + "epoch": 1.76, + "learning_rate": 1.6587270739537586e-05, + "loss": 1.0746, + "step": 23788 + }, + { + "epoch": 1.76, + "learning_rate": 1.658697063047878e-05, + "loss": 0.9682, + "step": 23789 + }, + { + "epoch": 1.76, + "learning_rate": 1.6586670510940194e-05, + "loss": 1.0422, + "step": 23790 + }, + { + "epoch": 1.76, + "learning_rate": 1.6586370380922313e-05, + "loss": 1.0074, + "step": 23791 + }, + { + "epoch": 1.76, + "learning_rate": 1.6586070240425608e-05, + "loss": 0.9827, + "step": 23792 + }, + { + "epoch": 1.76, + "learning_rate": 1.6585770089450563e-05, + "loss": 0.967, + "step": 23793 + }, + { + "epoch": 1.76, + "learning_rate": 1.6585469927997646e-05, + "loss": 0.9129, + "step": 23794 + }, + { + "epoch": 1.76, + "learning_rate": 1.6585169756067345e-05, + "loss": 1.0359, + "step": 23795 + }, + { + "epoch": 1.76, + "learning_rate": 1.658486957366013e-05, + "loss": 0.9327, + "step": 23796 + }, + { + "epoch": 1.76, + "learning_rate": 1.6584569380776484e-05, + "loss": 0.9447, + "step": 23797 + }, + { + "epoch": 1.76, + "learning_rate": 1.6584269177416883e-05, + "loss": 0.95, + "step": 23798 + }, + { + "epoch": 1.76, + "learning_rate": 1.6583968963581806e-05, + "loss": 0.9774, + "step": 23799 + }, + { + "epoch": 1.76, + "learning_rate": 1.6583668739271726e-05, + "loss": 0.9922, + "step": 23800 + }, + { + "epoch": 1.76, + "learning_rate": 1.658336850448712e-05, + "loss": 0.9627, + "step": 23801 + }, + { + "epoch": 1.76, + "learning_rate": 1.6583068259228473e-05, + "loss": 1.0793, + "step": 23802 + }, + { + "epoch": 1.76, + "learning_rate": 1.6582768003496256e-05, + "loss": 0.9763, + "step": 23803 + }, + { + "epoch": 1.76, + "learning_rate": 1.658246773729095e-05, + "loss": 1.0224, + "step": 23804 + }, + { + "epoch": 1.76, + "learning_rate": 1.6582167460613035e-05, + "loss": 1.0261, + "step": 23805 + }, + { + "epoch": 1.76, + "learning_rate": 1.6581867173462984e-05, + "loss": 1.1489, + "step": 23806 + }, + { + "epoch": 1.76, + "learning_rate": 1.6581566875841274e-05, + "loss": 1.0232, + "step": 23807 + }, + { + "epoch": 1.76, + "learning_rate": 1.6581266567748387e-05, + "loss": 1.0452, + "step": 23808 + }, + { + "epoch": 1.76, + "learning_rate": 1.65809662491848e-05, + "loss": 0.9546, + "step": 23809 + }, + { + "epoch": 1.76, + "learning_rate": 1.6580665920150992e-05, + "loss": 0.9363, + "step": 23810 + }, + { + "epoch": 1.76, + "learning_rate": 1.6580365580647434e-05, + "loss": 1.0112, + "step": 23811 + }, + { + "epoch": 1.76, + "learning_rate": 1.658006523067461e-05, + "loss": 1.0451, + "step": 23812 + }, + { + "epoch": 1.76, + "learning_rate": 1.6579764870233e-05, + "loss": 0.917, + "step": 23813 + }, + { + "epoch": 1.76, + "learning_rate": 1.657946449932307e-05, + "loss": 0.9771, + "step": 23814 + }, + { + "epoch": 1.76, + "learning_rate": 1.6579164117945313e-05, + "loss": 0.9085, + "step": 23815 + }, + { + "epoch": 1.76, + "learning_rate": 1.6578863726100197e-05, + "loss": 0.9459, + "step": 23816 + }, + { + "epoch": 1.76, + "learning_rate": 1.6578563323788203e-05, + "loss": 1.0801, + "step": 23817 + }, + { + "epoch": 1.76, + "learning_rate": 1.6578262911009813e-05, + "loss": 1.1159, + "step": 23818 + }, + { + "epoch": 1.76, + "learning_rate": 1.6577962487765497e-05, + "loss": 0.8847, + "step": 23819 + }, + { + "epoch": 1.76, + "learning_rate": 1.6577662054055737e-05, + "loss": 1.0524, + "step": 23820 + }, + { + "epoch": 1.76, + "learning_rate": 1.657736160988101e-05, + "loss": 0.9712, + "step": 23821 + }, + { + "epoch": 1.76, + "learning_rate": 1.6577061155241797e-05, + "loss": 1.1121, + "step": 23822 + }, + { + "epoch": 1.76, + "learning_rate": 1.6576760690138573e-05, + "loss": 0.9549, + "step": 23823 + }, + { + "epoch": 1.76, + "learning_rate": 1.6576460214571816e-05, + "loss": 1.0755, + "step": 23824 + }, + { + "epoch": 1.76, + "learning_rate": 1.6576159728542006e-05, + "loss": 0.931, + "step": 23825 + }, + { + "epoch": 1.76, + "learning_rate": 1.657585923204962e-05, + "loss": 0.9324, + "step": 23826 + }, + { + "epoch": 1.76, + "learning_rate": 1.6575558725095134e-05, + "loss": 0.9342, + "step": 23827 + }, + { + "epoch": 1.76, + "learning_rate": 1.6575258207679033e-05, + "loss": 1.0252, + "step": 23828 + }, + { + "epoch": 1.76, + "learning_rate": 1.6574957679801788e-05, + "loss": 1.0534, + "step": 23829 + }, + { + "epoch": 1.76, + "learning_rate": 1.6574657141463875e-05, + "loss": 0.955, + "step": 23830 + }, + { + "epoch": 1.76, + "learning_rate": 1.657435659266578e-05, + "loss": 0.9716, + "step": 23831 + }, + { + "epoch": 1.76, + "learning_rate": 1.6574056033407976e-05, + "loss": 1.0459, + "step": 23832 + }, + { + "epoch": 1.76, + "learning_rate": 1.6573755463690945e-05, + "loss": 1.1023, + "step": 23833 + }, + { + "epoch": 1.76, + "learning_rate": 1.6573454883515165e-05, + "loss": 1.0788, + "step": 23834 + }, + { + "epoch": 1.76, + "learning_rate": 1.657315429288111e-05, + "loss": 1.0243, + "step": 23835 + }, + { + "epoch": 1.76, + "learning_rate": 1.657285369178926e-05, + "loss": 1.0593, + "step": 23836 + }, + { + "epoch": 1.76, + "learning_rate": 1.657255308024009e-05, + "loss": 1.0366, + "step": 23837 + }, + { + "epoch": 1.76, + "learning_rate": 1.6572252458234093e-05, + "loss": 0.9841, + "step": 23838 + }, + { + "epoch": 1.76, + "learning_rate": 1.6571951825771728e-05, + "loss": 1.0752, + "step": 23839 + }, + { + "epoch": 1.76, + "learning_rate": 1.6571651182853483e-05, + "loss": 0.9957, + "step": 23840 + }, + { + "epoch": 1.76, + "learning_rate": 1.6571350529479836e-05, + "loss": 0.9544, + "step": 23841 + }, + { + "epoch": 1.76, + "learning_rate": 1.6571049865651265e-05, + "loss": 1.0139, + "step": 23842 + }, + { + "epoch": 1.76, + "learning_rate": 1.6570749191368247e-05, + "loss": 1.0314, + "step": 23843 + }, + { + "epoch": 1.76, + "learning_rate": 1.6570448506631257e-05, + "loss": 1.0275, + "step": 23844 + }, + { + "epoch": 1.76, + "learning_rate": 1.6570147811440783e-05, + "loss": 1.1043, + "step": 23845 + }, + { + "epoch": 1.76, + "learning_rate": 1.6569847105797295e-05, + "loss": 1.1332, + "step": 23846 + }, + { + "epoch": 1.76, + "learning_rate": 1.656954638970128e-05, + "loss": 1.0916, + "step": 23847 + }, + { + "epoch": 1.76, + "learning_rate": 1.65692456631532e-05, + "loss": 1.0308, + "step": 23848 + }, + { + "epoch": 1.76, + "learning_rate": 1.6568944926153553e-05, + "loss": 0.9893, + "step": 23849 + }, + { + "epoch": 1.76, + "learning_rate": 1.6568644178702803e-05, + "loss": 0.9996, + "step": 23850 + }, + { + "epoch": 1.76, + "learning_rate": 1.6568343420801438e-05, + "loss": 1.1065, + "step": 23851 + }, + { + "epoch": 1.76, + "learning_rate": 1.6568042652449933e-05, + "loss": 1.048, + "step": 23852 + }, + { + "epoch": 1.76, + "learning_rate": 1.6567741873648765e-05, + "loss": 0.994, + "step": 23853 + }, + { + "epoch": 1.76, + "learning_rate": 1.656744108439841e-05, + "loss": 0.9416, + "step": 23854 + }, + { + "epoch": 1.76, + "learning_rate": 1.6567140284699353e-05, + "loss": 0.9641, + "step": 23855 + }, + { + "epoch": 1.76, + "learning_rate": 1.6566839474552074e-05, + "loss": 0.9802, + "step": 23856 + }, + { + "epoch": 1.76, + "learning_rate": 1.656653865395704e-05, + "loss": 1.0258, + "step": 23857 + }, + { + "epoch": 1.76, + "learning_rate": 1.656623782291474e-05, + "loss": 0.963, + "step": 23858 + }, + { + "epoch": 1.76, + "learning_rate": 1.6565936981425653e-05, + "loss": 0.9998, + "step": 23859 + }, + { + "epoch": 1.76, + "learning_rate": 1.656563612949025e-05, + "loss": 1.0082, + "step": 23860 + }, + { + "epoch": 1.76, + "learning_rate": 1.6565335267109015e-05, + "loss": 0.9691, + "step": 23861 + }, + { + "epoch": 1.76, + "learning_rate": 1.6565034394282427e-05, + "loss": 0.9705, + "step": 23862 + }, + { + "epoch": 1.76, + "learning_rate": 1.656473351101096e-05, + "loss": 0.965, + "step": 23863 + }, + { + "epoch": 1.76, + "learning_rate": 1.6564432617295095e-05, + "loss": 1.0828, + "step": 23864 + }, + { + "epoch": 1.76, + "learning_rate": 1.6564131713135315e-05, + "loss": 1.0645, + "step": 23865 + }, + { + "epoch": 1.76, + "learning_rate": 1.6563830798532096e-05, + "loss": 0.968, + "step": 23866 + }, + { + "epoch": 1.76, + "learning_rate": 1.6563529873485916e-05, + "loss": 0.9655, + "step": 23867 + }, + { + "epoch": 1.76, + "learning_rate": 1.6563228937997253e-05, + "loss": 0.9714, + "step": 23868 + }, + { + "epoch": 1.76, + "learning_rate": 1.6562927992066585e-05, + "loss": 0.9999, + "step": 23869 + }, + { + "epoch": 1.76, + "learning_rate": 1.6562627035694395e-05, + "loss": 1.0309, + "step": 23870 + }, + { + "epoch": 1.76, + "learning_rate": 1.6562326068881156e-05, + "loss": 1.0185, + "step": 23871 + }, + { + "epoch": 1.76, + "learning_rate": 1.6562025091627352e-05, + "loss": 1.0043, + "step": 23872 + }, + { + "epoch": 1.76, + "learning_rate": 1.656172410393346e-05, + "loss": 0.9622, + "step": 23873 + }, + { + "epoch": 1.76, + "learning_rate": 1.656142310579996e-05, + "loss": 1.1029, + "step": 23874 + }, + { + "epoch": 1.76, + "learning_rate": 1.656112209722733e-05, + "loss": 0.9604, + "step": 23875 + }, + { + "epoch": 1.76, + "learning_rate": 1.656082107821605e-05, + "loss": 0.9211, + "step": 23876 + }, + { + "epoch": 1.76, + "learning_rate": 1.656052004876659e-05, + "loss": 0.9697, + "step": 23877 + }, + { + "epoch": 1.76, + "learning_rate": 1.6560219008879443e-05, + "loss": 1.07, + "step": 23878 + }, + { + "epoch": 1.76, + "learning_rate": 1.6559917958555077e-05, + "loss": 0.9745, + "step": 23879 + }, + { + "epoch": 1.76, + "learning_rate": 1.655961689779398e-05, + "loss": 1.0081, + "step": 23880 + }, + { + "epoch": 1.76, + "learning_rate": 1.6559315826596624e-05, + "loss": 1.0283, + "step": 23881 + }, + { + "epoch": 1.76, + "learning_rate": 1.6559014744963492e-05, + "loss": 1.0406, + "step": 23882 + }, + { + "epoch": 1.76, + "learning_rate": 1.655871365289506e-05, + "loss": 1.0844, + "step": 23883 + }, + { + "epoch": 1.76, + "learning_rate": 1.6558412550391808e-05, + "loss": 1.0605, + "step": 23884 + }, + { + "epoch": 1.76, + "learning_rate": 1.6558111437454214e-05, + "loss": 1.0408, + "step": 23885 + }, + { + "epoch": 1.76, + "learning_rate": 1.6557810314082764e-05, + "loss": 0.9969, + "step": 23886 + }, + { + "epoch": 1.76, + "learning_rate": 1.6557509180277925e-05, + "loss": 1.1093, + "step": 23887 + }, + { + "epoch": 1.77, + "learning_rate": 1.6557208036040188e-05, + "loss": 1.0215, + "step": 23888 + }, + { + "epoch": 1.77, + "learning_rate": 1.6556906881370023e-05, + "loss": 1.0366, + "step": 23889 + }, + { + "epoch": 1.77, + "learning_rate": 1.6556605716267917e-05, + "loss": 1.0759, + "step": 23890 + }, + { + "epoch": 1.77, + "learning_rate": 1.655630454073434e-05, + "loss": 0.9602, + "step": 23891 + }, + { + "epoch": 1.77, + "learning_rate": 1.6556003354769782e-05, + "loss": 1.0552, + "step": 23892 + }, + { + "epoch": 1.77, + "learning_rate": 1.655570215837471e-05, + "loss": 1.0177, + "step": 23893 + }, + { + "epoch": 1.77, + "learning_rate": 1.6555400951549617e-05, + "loss": 0.9079, + "step": 23894 + }, + { + "epoch": 1.77, + "learning_rate": 1.655509973429497e-05, + "loss": 0.9791, + "step": 23895 + }, + { + "epoch": 1.77, + "learning_rate": 1.6554798506611253e-05, + "loss": 0.988, + "step": 23896 + }, + { + "epoch": 1.77, + "learning_rate": 1.6554497268498945e-05, + "loss": 1.0222, + "step": 23897 + }, + { + "epoch": 1.77, + "learning_rate": 1.655419601995853e-05, + "loss": 1.0538, + "step": 23898 + }, + { + "epoch": 1.77, + "learning_rate": 1.6553894760990478e-05, + "loss": 1.0456, + "step": 23899 + }, + { + "epoch": 1.77, + "learning_rate": 1.6553593491595277e-05, + "loss": 0.9995, + "step": 23900 + }, + { + "epoch": 1.77, + "learning_rate": 1.6553292211773403e-05, + "loss": 0.9807, + "step": 23901 + }, + { + "epoch": 1.77, + "learning_rate": 1.655299092152533e-05, + "loss": 1.0496, + "step": 23902 + }, + { + "epoch": 1.77, + "learning_rate": 1.6552689620851547e-05, + "loss": 1.0775, + "step": 23903 + }, + { + "epoch": 1.77, + "learning_rate": 1.6552388309752524e-05, + "loss": 0.9692, + "step": 23904 + }, + { + "epoch": 1.77, + "learning_rate": 1.6552086988228747e-05, + "loss": 0.9299, + "step": 23905 + }, + { + "epoch": 1.77, + "learning_rate": 1.6551785656280696e-05, + "loss": 0.9838, + "step": 23906 + }, + { + "epoch": 1.77, + "learning_rate": 1.6551484313908847e-05, + "loss": 1.0659, + "step": 23907 + }, + { + "epoch": 1.77, + "learning_rate": 1.6551182961113678e-05, + "loss": 0.9987, + "step": 23908 + }, + { + "epoch": 1.77, + "learning_rate": 1.655088159789567e-05, + "loss": 1.0214, + "step": 23909 + }, + { + "epoch": 1.77, + "learning_rate": 1.6550580224255302e-05, + "loss": 1.0284, + "step": 23910 + }, + { + "epoch": 1.77, + "learning_rate": 1.6550278840193058e-05, + "loss": 1.1696, + "step": 23911 + }, + { + "epoch": 1.77, + "learning_rate": 1.6549977445709414e-05, + "loss": 1.0231, + "step": 23912 + }, + { + "epoch": 1.77, + "learning_rate": 1.654967604080485e-05, + "loss": 0.9514, + "step": 23913 + }, + { + "epoch": 1.77, + "learning_rate": 1.6549374625479845e-05, + "loss": 0.9865, + "step": 23914 + }, + { + "epoch": 1.77, + "learning_rate": 1.6549073199734875e-05, + "loss": 1.0237, + "step": 23915 + }, + { + "epoch": 1.77, + "learning_rate": 1.6548771763570428e-05, + "loss": 0.8841, + "step": 23916 + }, + { + "epoch": 1.77, + "learning_rate": 1.6548470316986977e-05, + "loss": 1.0206, + "step": 23917 + }, + { + "epoch": 1.77, + "learning_rate": 1.6548168859985e-05, + "loss": 0.9527, + "step": 23918 + }, + { + "epoch": 1.77, + "learning_rate": 1.6547867392564986e-05, + "loss": 0.907, + "step": 23919 + }, + { + "epoch": 1.77, + "learning_rate": 1.6547565914727405e-05, + "loss": 1.0248, + "step": 23920 + }, + { + "epoch": 1.77, + "learning_rate": 1.654726442647274e-05, + "loss": 0.94, + "step": 23921 + }, + { + "epoch": 1.77, + "learning_rate": 1.6546962927801474e-05, + "loss": 1.0221, + "step": 23922 + }, + { + "epoch": 1.77, + "learning_rate": 1.654666141871408e-05, + "loss": 1.009, + "step": 23923 + }, + { + "epoch": 1.77, + "learning_rate": 1.6546359899211043e-05, + "loss": 0.9976, + "step": 23924 + }, + { + "epoch": 1.77, + "learning_rate": 1.654605836929284e-05, + "loss": 1.1182, + "step": 23925 + }, + { + "epoch": 1.77, + "learning_rate": 1.6545756828959954e-05, + "loss": 1.0344, + "step": 23926 + }, + { + "epoch": 1.77, + "learning_rate": 1.654545527821286e-05, + "loss": 1.0416, + "step": 23927 + }, + { + "epoch": 1.77, + "learning_rate": 1.6545153717052042e-05, + "loss": 1.0463, + "step": 23928 + }, + { + "epoch": 1.77, + "learning_rate": 1.654485214547798e-05, + "loss": 0.9279, + "step": 23929 + }, + { + "epoch": 1.77, + "learning_rate": 1.654455056349115e-05, + "loss": 1.0261, + "step": 23930 + }, + { + "epoch": 1.77, + "learning_rate": 1.6544248971092033e-05, + "loss": 1.1085, + "step": 23931 + }, + { + "epoch": 1.77, + "learning_rate": 1.654394736828111e-05, + "loss": 1.0181, + "step": 23932 + }, + { + "epoch": 1.77, + "learning_rate": 1.654364575505886e-05, + "loss": 1.0216, + "step": 23933 + }, + { + "epoch": 1.77, + "learning_rate": 1.6543344131425763e-05, + "loss": 0.9537, + "step": 23934 + }, + { + "epoch": 1.77, + "learning_rate": 1.65430424973823e-05, + "loss": 1.0151, + "step": 23935 + }, + { + "epoch": 1.77, + "learning_rate": 1.6542740852928947e-05, + "loss": 0.924, + "step": 23936 + }, + { + "epoch": 1.77, + "learning_rate": 1.6542439198066193e-05, + "loss": 1.059, + "step": 23937 + }, + { + "epoch": 1.77, + "learning_rate": 1.6542137532794505e-05, + "loss": 0.932, + "step": 23938 + }, + { + "epoch": 1.77, + "learning_rate": 1.6541835857114375e-05, + "loss": 1.0069, + "step": 23939 + }, + { + "epoch": 1.77, + "learning_rate": 1.6541534171026276e-05, + "loss": 1.0442, + "step": 23940 + }, + { + "epoch": 1.77, + "learning_rate": 1.654123247453069e-05, + "loss": 0.9535, + "step": 23941 + }, + { + "epoch": 1.77, + "learning_rate": 1.6540930767628096e-05, + "loss": 1.0157, + "step": 23942 + }, + { + "epoch": 1.77, + "learning_rate": 1.6540629050318973e-05, + "loss": 0.99, + "step": 23943 + }, + { + "epoch": 1.77, + "learning_rate": 1.6540327322603806e-05, + "loss": 0.9641, + "step": 23944 + }, + { + "epoch": 1.77, + "learning_rate": 1.654002558448307e-05, + "loss": 0.927, + "step": 23945 + }, + { + "epoch": 1.77, + "learning_rate": 1.6539723835957246e-05, + "loss": 0.9999, + "step": 23946 + }, + { + "epoch": 1.77, + "learning_rate": 1.6539422077026818e-05, + "loss": 0.9997, + "step": 23947 + }, + { + "epoch": 1.77, + "learning_rate": 1.653912030769226e-05, + "loss": 0.9935, + "step": 23948 + }, + { + "epoch": 1.77, + "learning_rate": 1.6538818527954058e-05, + "loss": 1.0301, + "step": 23949 + }, + { + "epoch": 1.77, + "learning_rate": 1.653851673781269e-05, + "loss": 0.9873, + "step": 23950 + }, + { + "epoch": 1.77, + "learning_rate": 1.6538214937268633e-05, + "loss": 1.0564, + "step": 23951 + }, + { + "epoch": 1.77, + "learning_rate": 1.653791312632237e-05, + "loss": 1.0144, + "step": 23952 + }, + { + "epoch": 1.77, + "learning_rate": 1.6537611304974378e-05, + "loss": 1.0277, + "step": 23953 + }, + { + "epoch": 1.77, + "learning_rate": 1.6537309473225144e-05, + "loss": 1.071, + "step": 23954 + }, + { + "epoch": 1.77, + "learning_rate": 1.6537007631075145e-05, + "loss": 1.0141, + "step": 23955 + }, + { + "epoch": 1.77, + "learning_rate": 1.6536705778524855e-05, + "loss": 1.0632, + "step": 23956 + }, + { + "epoch": 1.77, + "learning_rate": 1.6536403915574763e-05, + "loss": 0.9229, + "step": 23957 + }, + { + "epoch": 1.77, + "learning_rate": 1.6536102042225346e-05, + "loss": 0.9737, + "step": 23958 + }, + { + "epoch": 1.77, + "learning_rate": 1.6535800158477085e-05, + "loss": 1.0049, + "step": 23959 + }, + { + "epoch": 1.77, + "learning_rate": 1.6535498264330457e-05, + "loss": 0.832, + "step": 23960 + }, + { + "epoch": 1.77, + "learning_rate": 1.6535196359785948e-05, + "loss": 0.9256, + "step": 23961 + }, + { + "epoch": 1.77, + "learning_rate": 1.653489444484403e-05, + "loss": 0.9121, + "step": 23962 + }, + { + "epoch": 1.77, + "learning_rate": 1.6534592519505196e-05, + "loss": 0.986, + "step": 23963 + }, + { + "epoch": 1.77, + "learning_rate": 1.653429058376991e-05, + "loss": 1.0701, + "step": 23964 + }, + { + "epoch": 1.77, + "learning_rate": 1.6533988637638666e-05, + "loss": 1.0188, + "step": 23965 + }, + { + "epoch": 1.77, + "learning_rate": 1.653368668111194e-05, + "loss": 0.8634, + "step": 23966 + }, + { + "epoch": 1.77, + "learning_rate": 1.653338471419021e-05, + "loss": 1.0381, + "step": 23967 + }, + { + "epoch": 1.77, + "learning_rate": 1.653308273687396e-05, + "loss": 1.0722, + "step": 23968 + }, + { + "epoch": 1.77, + "learning_rate": 1.653278074916367e-05, + "loss": 1.1122, + "step": 23969 + }, + { + "epoch": 1.77, + "learning_rate": 1.6532478751059816e-05, + "loss": 1.0571, + "step": 23970 + }, + { + "epoch": 1.77, + "learning_rate": 1.6532176742562883e-05, + "loss": 1.0984, + "step": 23971 + }, + { + "epoch": 1.77, + "learning_rate": 1.653187472367335e-05, + "loss": 0.9611, + "step": 23972 + }, + { + "epoch": 1.77, + "learning_rate": 1.65315726943917e-05, + "loss": 1.022, + "step": 23973 + }, + { + "epoch": 1.77, + "learning_rate": 1.653127065471841e-05, + "loss": 1.0218, + "step": 23974 + }, + { + "epoch": 1.77, + "learning_rate": 1.6530968604653962e-05, + "loss": 0.9461, + "step": 23975 + }, + { + "epoch": 1.77, + "learning_rate": 1.6530666544198836e-05, + "loss": 1.0356, + "step": 23976 + }, + { + "epoch": 1.77, + "learning_rate": 1.6530364473353512e-05, + "loss": 1.0215, + "step": 23977 + }, + { + "epoch": 1.77, + "learning_rate": 1.6530062392118472e-05, + "loss": 0.8645, + "step": 23978 + }, + { + "epoch": 1.77, + "learning_rate": 1.6529760300494197e-05, + "loss": 1.0502, + "step": 23979 + }, + { + "epoch": 1.77, + "learning_rate": 1.6529458198481168e-05, + "loss": 1.036, + "step": 23980 + }, + { + "epoch": 1.77, + "learning_rate": 1.6529156086079863e-05, + "loss": 0.8855, + "step": 23981 + }, + { + "epoch": 1.77, + "learning_rate": 1.6528853963290762e-05, + "loss": 0.9671, + "step": 23982 + }, + { + "epoch": 1.77, + "learning_rate": 1.652855183011435e-05, + "loss": 1.172, + "step": 23983 + }, + { + "epoch": 1.77, + "learning_rate": 1.6528249686551106e-05, + "loss": 0.9524, + "step": 23984 + }, + { + "epoch": 1.77, + "learning_rate": 1.652794753260151e-05, + "loss": 0.9067, + "step": 23985 + }, + { + "epoch": 1.77, + "learning_rate": 1.652764536826604e-05, + "loss": 1.0462, + "step": 23986 + }, + { + "epoch": 1.77, + "learning_rate": 1.6527343193545183e-05, + "loss": 1.0271, + "step": 23987 + }, + { + "epoch": 1.77, + "learning_rate": 1.6527041008439418e-05, + "loss": 1.0702, + "step": 23988 + }, + { + "epoch": 1.77, + "learning_rate": 1.6526738812949222e-05, + "loss": 0.9945, + "step": 23989 + }, + { + "epoch": 1.77, + "learning_rate": 1.6526436607075073e-05, + "loss": 1.0549, + "step": 23990 + }, + { + "epoch": 1.77, + "learning_rate": 1.6526134390817462e-05, + "loss": 1.0519, + "step": 23991 + }, + { + "epoch": 1.77, + "learning_rate": 1.6525832164176863e-05, + "loss": 1.1016, + "step": 23992 + }, + { + "epoch": 1.77, + "learning_rate": 1.652552992715376e-05, + "loss": 0.9702, + "step": 23993 + }, + { + "epoch": 1.77, + "learning_rate": 1.6525227679748632e-05, + "loss": 1.078, + "step": 23994 + }, + { + "epoch": 1.77, + "learning_rate": 1.6524925421961958e-05, + "loss": 1.0094, + "step": 23995 + }, + { + "epoch": 1.77, + "learning_rate": 1.6524623153794224e-05, + "loss": 1.0552, + "step": 23996 + }, + { + "epoch": 1.77, + "learning_rate": 1.6524320875245905e-05, + "loss": 0.9646, + "step": 23997 + }, + { + "epoch": 1.77, + "learning_rate": 1.6524018586317486e-05, + "loss": 1.0114, + "step": 23998 + }, + { + "epoch": 1.77, + "learning_rate": 1.652371628700945e-05, + "loss": 1.0056, + "step": 23999 + }, + { + "epoch": 1.77, + "learning_rate": 1.652341397732227e-05, + "loss": 0.9339, + "step": 24000 + }, + { + "epoch": 1.77, + "learning_rate": 1.6523111657256433e-05, + "loss": 0.9976, + "step": 24001 + }, + { + "epoch": 1.77, + "learning_rate": 1.6522809326812418e-05, + "loss": 0.9883, + "step": 24002 + }, + { + "epoch": 1.77, + "learning_rate": 1.6522506985990708e-05, + "loss": 1.0458, + "step": 24003 + }, + { + "epoch": 1.77, + "learning_rate": 1.652220463479178e-05, + "loss": 0.9904, + "step": 24004 + }, + { + "epoch": 1.77, + "learning_rate": 1.6521902273216122e-05, + "loss": 1.0035, + "step": 24005 + }, + { + "epoch": 1.77, + "learning_rate": 1.652159990126421e-05, + "loss": 1.0549, + "step": 24006 + }, + { + "epoch": 1.77, + "learning_rate": 1.6521297518936523e-05, + "loss": 0.9912, + "step": 24007 + }, + { + "epoch": 1.77, + "learning_rate": 1.6520995126233547e-05, + "loss": 1.0204, + "step": 24008 + }, + { + "epoch": 1.77, + "learning_rate": 1.6520692723155762e-05, + "loss": 1.0462, + "step": 24009 + }, + { + "epoch": 1.77, + "learning_rate": 1.6520390309703646e-05, + "loss": 1.0105, + "step": 24010 + }, + { + "epoch": 1.77, + "learning_rate": 1.6520087885877684e-05, + "loss": 1.0677, + "step": 24011 + }, + { + "epoch": 1.77, + "learning_rate": 1.6519785451678354e-05, + "loss": 1.0178, + "step": 24012 + }, + { + "epoch": 1.77, + "learning_rate": 1.651948300710614e-05, + "loss": 1.0107, + "step": 24013 + }, + { + "epoch": 1.77, + "learning_rate": 1.651918055216152e-05, + "loss": 0.8421, + "step": 24014 + }, + { + "epoch": 1.77, + "learning_rate": 1.6518878086844982e-05, + "loss": 1.0346, + "step": 24015 + }, + { + "epoch": 1.77, + "learning_rate": 1.6518575611156995e-05, + "loss": 0.8905, + "step": 24016 + }, + { + "epoch": 1.77, + "learning_rate": 1.6518273125098052e-05, + "loss": 1.1039, + "step": 24017 + }, + { + "epoch": 1.77, + "learning_rate": 1.6517970628668627e-05, + "loss": 1.0569, + "step": 24018 + }, + { + "epoch": 1.77, + "learning_rate": 1.651766812186921e-05, + "loss": 0.9969, + "step": 24019 + }, + { + "epoch": 1.77, + "learning_rate": 1.651736560470027e-05, + "loss": 1.0313, + "step": 24020 + }, + { + "epoch": 1.77, + "learning_rate": 1.6517063077162293e-05, + "loss": 1.1074, + "step": 24021 + }, + { + "epoch": 1.77, + "learning_rate": 1.651676053925577e-05, + "loss": 0.9804, + "step": 24022 + }, + { + "epoch": 1.78, + "learning_rate": 1.6516457990981164e-05, + "loss": 1.0506, + "step": 24023 + }, + { + "epoch": 1.78, + "learning_rate": 1.6516155432338975e-05, + "loss": 0.9282, + "step": 24024 + }, + { + "epoch": 1.78, + "learning_rate": 1.6515852863329672e-05, + "loss": 0.9989, + "step": 24025 + }, + { + "epoch": 1.78, + "learning_rate": 1.6515550283953743e-05, + "loss": 0.9811, + "step": 24026 + }, + { + "epoch": 1.78, + "learning_rate": 1.6515247694211665e-05, + "loss": 1.0152, + "step": 24027 + }, + { + "epoch": 1.78, + "learning_rate": 1.651494509410392e-05, + "loss": 1.0295, + "step": 24028 + }, + { + "epoch": 1.78, + "learning_rate": 1.6514642483630992e-05, + "loss": 0.9464, + "step": 24029 + }, + { + "epoch": 1.78, + "learning_rate": 1.651433986279336e-05, + "loss": 0.8911, + "step": 24030 + }, + { + "epoch": 1.78, + "learning_rate": 1.6514037231591506e-05, + "loss": 0.9588, + "step": 24031 + }, + { + "epoch": 1.78, + "learning_rate": 1.6513734590025916e-05, + "loss": 1.0324, + "step": 24032 + }, + { + "epoch": 1.78, + "learning_rate": 1.6513431938097065e-05, + "loss": 0.9401, + "step": 24033 + }, + { + "epoch": 1.78, + "learning_rate": 1.6513129275805436e-05, + "loss": 0.9347, + "step": 24034 + }, + { + "epoch": 1.78, + "learning_rate": 1.6512826603151514e-05, + "loss": 1.0557, + "step": 24035 + }, + { + "epoch": 1.78, + "learning_rate": 1.6512523920135775e-05, + "loss": 0.891, + "step": 24036 + }, + { + "epoch": 1.78, + "learning_rate": 1.6512221226758704e-05, + "loss": 1.1026, + "step": 24037 + }, + { + "epoch": 1.78, + "learning_rate": 1.651191852302078e-05, + "loss": 1.0183, + "step": 24038 + }, + { + "epoch": 1.78, + "learning_rate": 1.6511615808922493e-05, + "loss": 0.9267, + "step": 24039 + }, + { + "epoch": 1.78, + "learning_rate": 1.6511313084464314e-05, + "loss": 0.9207, + "step": 24040 + }, + { + "epoch": 1.78, + "learning_rate": 1.6511010349646733e-05, + "loss": 1.0453, + "step": 24041 + }, + { + "epoch": 1.78, + "learning_rate": 1.6510707604470223e-05, + "loss": 0.9868, + "step": 24042 + }, + { + "epoch": 1.78, + "learning_rate": 1.6510404848935273e-05, + "loss": 1.1542, + "step": 24043 + }, + { + "epoch": 1.78, + "learning_rate": 1.651010208304236e-05, + "loss": 0.9132, + "step": 24044 + }, + { + "epoch": 1.78, + "learning_rate": 1.6509799306791972e-05, + "loss": 1.0277, + "step": 24045 + }, + { + "epoch": 1.78, + "learning_rate": 1.6509496520184584e-05, + "loss": 0.8723, + "step": 24046 + }, + { + "epoch": 1.78, + "learning_rate": 1.650919372322068e-05, + "loss": 1.0145, + "step": 24047 + }, + { + "epoch": 1.78, + "learning_rate": 1.650889091590074e-05, + "loss": 1.0674, + "step": 24048 + }, + { + "epoch": 1.78, + "learning_rate": 1.6508588098225248e-05, + "loss": 1.0963, + "step": 24049 + }, + { + "epoch": 1.78, + "learning_rate": 1.650828527019469e-05, + "loss": 0.985, + "step": 24050 + }, + { + "epoch": 1.78, + "learning_rate": 1.650798243180954e-05, + "loss": 1.0045, + "step": 24051 + }, + { + "epoch": 1.78, + "learning_rate": 1.650767958307028e-05, + "loss": 0.9577, + "step": 24052 + }, + { + "epoch": 1.78, + "learning_rate": 1.65073767239774e-05, + "loss": 0.9714, + "step": 24053 + }, + { + "epoch": 1.78, + "learning_rate": 1.6507073854531374e-05, + "loss": 0.9545, + "step": 24054 + }, + { + "epoch": 1.78, + "learning_rate": 1.6506770974732685e-05, + "loss": 1.1305, + "step": 24055 + }, + { + "epoch": 1.78, + "learning_rate": 1.650646808458182e-05, + "loss": 1.1115, + "step": 24056 + }, + { + "epoch": 1.78, + "learning_rate": 1.650616518407926e-05, + "loss": 1.0418, + "step": 24057 + }, + { + "epoch": 1.78, + "learning_rate": 1.6505862273225476e-05, + "loss": 1.0858, + "step": 24058 + }, + { + "epoch": 1.78, + "learning_rate": 1.6505559352020965e-05, + "loss": 0.9021, + "step": 24059 + }, + { + "epoch": 1.78, + "learning_rate": 1.6505256420466202e-05, + "loss": 1.0245, + "step": 24060 + }, + { + "epoch": 1.78, + "learning_rate": 1.6504953478561665e-05, + "loss": 1.0437, + "step": 24061 + }, + { + "epoch": 1.78, + "learning_rate": 1.6504650526307843e-05, + "loss": 1.0128, + "step": 24062 + }, + { + "epoch": 1.78, + "learning_rate": 1.6504347563705213e-05, + "loss": 1.0506, + "step": 24063 + }, + { + "epoch": 1.78, + "learning_rate": 1.6504044590754263e-05, + "loss": 1.0779, + "step": 24064 + }, + { + "epoch": 1.78, + "learning_rate": 1.6503741607455466e-05, + "loss": 0.8749, + "step": 24065 + }, + { + "epoch": 1.78, + "learning_rate": 1.650343861380931e-05, + "loss": 1.1356, + "step": 24066 + }, + { + "epoch": 1.78, + "learning_rate": 1.650313560981628e-05, + "loss": 0.993, + "step": 24067 + }, + { + "epoch": 1.78, + "learning_rate": 1.6502832595476852e-05, + "loss": 0.9074, + "step": 24068 + }, + { + "epoch": 1.78, + "learning_rate": 1.6502529570791508e-05, + "loss": 1.0402, + "step": 24069 + }, + { + "epoch": 1.78, + "learning_rate": 1.6502226535760737e-05, + "loss": 0.8812, + "step": 24070 + }, + { + "epoch": 1.78, + "learning_rate": 1.6501923490385014e-05, + "loss": 0.9971, + "step": 24071 + }, + { + "epoch": 1.78, + "learning_rate": 1.6501620434664826e-05, + "loss": 1.0219, + "step": 24072 + }, + { + "epoch": 1.78, + "learning_rate": 1.6501317368600648e-05, + "loss": 1.0968, + "step": 24073 + }, + { + "epoch": 1.78, + "learning_rate": 1.650101429219297e-05, + "loss": 1.1564, + "step": 24074 + }, + { + "epoch": 1.78, + "learning_rate": 1.650071120544227e-05, + "loss": 0.9647, + "step": 24075 + }, + { + "epoch": 1.78, + "learning_rate": 1.6500408108349037e-05, + "loss": 0.9301, + "step": 24076 + }, + { + "epoch": 1.78, + "learning_rate": 1.650010500091374e-05, + "loss": 0.9491, + "step": 24077 + }, + { + "epoch": 1.78, + "learning_rate": 1.6499801883136873e-05, + "loss": 0.9211, + "step": 24078 + }, + { + "epoch": 1.78, + "learning_rate": 1.6499498755018916e-05, + "loss": 0.9426, + "step": 24079 + }, + { + "epoch": 1.78, + "learning_rate": 1.6499195616560347e-05, + "loss": 1.0237, + "step": 24080 + }, + { + "epoch": 1.78, + "learning_rate": 1.649889246776165e-05, + "loss": 1.114, + "step": 24081 + }, + { + "epoch": 1.78, + "learning_rate": 1.649858930862331e-05, + "loss": 1.0465, + "step": 24082 + }, + { + "epoch": 1.78, + "learning_rate": 1.6498286139145808e-05, + "loss": 0.8398, + "step": 24083 + }, + { + "epoch": 1.78, + "learning_rate": 1.6497982959329622e-05, + "loss": 0.986, + "step": 24084 + }, + { + "epoch": 1.78, + "learning_rate": 1.649767976917524e-05, + "loss": 0.916, + "step": 24085 + }, + { + "epoch": 1.78, + "learning_rate": 1.6497376568683142e-05, + "loss": 1.0733, + "step": 24086 + }, + { + "epoch": 1.78, + "learning_rate": 1.6497073357853813e-05, + "loss": 1.103, + "step": 24087 + }, + { + "epoch": 1.78, + "learning_rate": 1.6496770136687732e-05, + "loss": 1.0069, + "step": 24088 + }, + { + "epoch": 1.78, + "learning_rate": 1.6496466905185382e-05, + "loss": 1.0803, + "step": 24089 + }, + { + "epoch": 1.78, + "learning_rate": 1.6496163663347248e-05, + "loss": 1.0021, + "step": 24090 + }, + { + "epoch": 1.78, + "learning_rate": 1.649586041117381e-05, + "loss": 0.9781, + "step": 24091 + }, + { + "epoch": 1.78, + "learning_rate": 1.649555714866555e-05, + "loss": 0.9309, + "step": 24092 + }, + { + "epoch": 1.78, + "learning_rate": 1.649525387582295e-05, + "loss": 1.082, + "step": 24093 + }, + { + "epoch": 1.78, + "learning_rate": 1.6494950592646497e-05, + "loss": 0.9809, + "step": 24094 + }, + { + "epoch": 1.78, + "learning_rate": 1.649464729913667e-05, + "loss": 0.9313, + "step": 24095 + }, + { + "epoch": 1.78, + "learning_rate": 1.649434399529395e-05, + "loss": 1.1163, + "step": 24096 + }, + { + "epoch": 1.78, + "learning_rate": 1.649404068111882e-05, + "loss": 1.0618, + "step": 24097 + }, + { + "epoch": 1.78, + "learning_rate": 1.649373735661177e-05, + "loss": 1.0844, + "step": 24098 + }, + { + "epoch": 1.78, + "learning_rate": 1.6493434021773277e-05, + "loss": 1.0541, + "step": 24099 + }, + { + "epoch": 1.78, + "learning_rate": 1.6493130676603818e-05, + "loss": 1.0337, + "step": 24100 + }, + { + "epoch": 1.78, + "learning_rate": 1.6492827321103885e-05, + "loss": 1.0894, + "step": 24101 + }, + { + "epoch": 1.78, + "learning_rate": 1.6492523955273954e-05, + "loss": 1.0261, + "step": 24102 + }, + { + "epoch": 1.78, + "learning_rate": 1.649222057911451e-05, + "loss": 0.9101, + "step": 24103 + }, + { + "epoch": 1.78, + "learning_rate": 1.649191719262604e-05, + "loss": 1.0512, + "step": 24104 + }, + { + "epoch": 1.78, + "learning_rate": 1.6491613795809018e-05, + "loss": 1.0886, + "step": 24105 + }, + { + "epoch": 1.78, + "learning_rate": 1.6491310388663934e-05, + "loss": 0.9354, + "step": 24106 + }, + { + "epoch": 1.78, + "learning_rate": 1.6491006971191265e-05, + "loss": 0.9418, + "step": 24107 + }, + { + "epoch": 1.78, + "learning_rate": 1.6490703543391496e-05, + "loss": 1.0767, + "step": 24108 + }, + { + "epoch": 1.78, + "learning_rate": 1.6490400105265115e-05, + "loss": 0.9526, + "step": 24109 + }, + { + "epoch": 1.78, + "learning_rate": 1.6490096656812602e-05, + "loss": 0.9915, + "step": 24110 + }, + { + "epoch": 1.78, + "learning_rate": 1.6489793198034433e-05, + "loss": 0.9476, + "step": 24111 + }, + { + "epoch": 1.78, + "learning_rate": 1.6489489728931098e-05, + "loss": 1.0124, + "step": 24112 + }, + { + "epoch": 1.78, + "learning_rate": 1.6489186249503078e-05, + "loss": 0.9429, + "step": 24113 + }, + { + "epoch": 1.78, + "learning_rate": 1.6488882759750855e-05, + "loss": 0.9732, + "step": 24114 + }, + { + "epoch": 1.78, + "learning_rate": 1.6488579259674914e-05, + "loss": 1.0203, + "step": 24115 + }, + { + "epoch": 1.78, + "learning_rate": 1.6488275749275734e-05, + "loss": 1.0761, + "step": 24116 + }, + { + "epoch": 1.78, + "learning_rate": 1.6487972228553798e-05, + "loss": 1.0348, + "step": 24117 + }, + { + "epoch": 1.78, + "learning_rate": 1.6487668697509596e-05, + "loss": 0.9405, + "step": 24118 + }, + { + "epoch": 1.78, + "learning_rate": 1.6487365156143604e-05, + "loss": 0.965, + "step": 24119 + }, + { + "epoch": 1.78, + "learning_rate": 1.6487061604456304e-05, + "loss": 0.9484, + "step": 24120 + }, + { + "epoch": 1.78, + "learning_rate": 1.6486758042448187e-05, + "loss": 1.0765, + "step": 24121 + }, + { + "epoch": 1.78, + "learning_rate": 1.6486454470119725e-05, + "loss": 1.0466, + "step": 24122 + }, + { + "epoch": 1.78, + "learning_rate": 1.6486150887471413e-05, + "loss": 1.0448, + "step": 24123 + }, + { + "epoch": 1.78, + "learning_rate": 1.6485847294503722e-05, + "loss": 0.9254, + "step": 24124 + }, + { + "epoch": 1.78, + "learning_rate": 1.6485543691217145e-05, + "loss": 1.0533, + "step": 24125 + }, + { + "epoch": 1.78, + "learning_rate": 1.648524007761216e-05, + "loss": 1.063, + "step": 24126 + }, + { + "epoch": 1.78, + "learning_rate": 1.6484936453689248e-05, + "loss": 1.1124, + "step": 24127 + }, + { + "epoch": 1.78, + "learning_rate": 1.6484632819448898e-05, + "loss": 1.0424, + "step": 24128 + }, + { + "epoch": 1.78, + "learning_rate": 1.6484329174891585e-05, + "loss": 1.059, + "step": 24129 + }, + { + "epoch": 1.78, + "learning_rate": 1.6484025520017804e-05, + "loss": 0.9253, + "step": 24130 + }, + { + "epoch": 1.78, + "learning_rate": 1.648372185482803e-05, + "loss": 0.9716, + "step": 24131 + }, + { + "epoch": 1.78, + "learning_rate": 1.6483418179322745e-05, + "loss": 1.0572, + "step": 24132 + }, + { + "epoch": 1.78, + "learning_rate": 1.6483114493502432e-05, + "loss": 0.9627, + "step": 24133 + }, + { + "epoch": 1.78, + "learning_rate": 1.648281079736758e-05, + "loss": 1.0684, + "step": 24134 + }, + { + "epoch": 1.78, + "learning_rate": 1.6482507090918666e-05, + "loss": 0.9395, + "step": 24135 + }, + { + "epoch": 1.78, + "learning_rate": 1.648220337415618e-05, + "loss": 1.008, + "step": 24136 + }, + { + "epoch": 1.78, + "learning_rate": 1.6481899647080598e-05, + "loss": 0.9495, + "step": 24137 + }, + { + "epoch": 1.78, + "learning_rate": 1.648159590969241e-05, + "loss": 1.0358, + "step": 24138 + }, + { + "epoch": 1.78, + "learning_rate": 1.6481292161992092e-05, + "loss": 1.0403, + "step": 24139 + }, + { + "epoch": 1.78, + "learning_rate": 1.648098840398013e-05, + "loss": 1.0867, + "step": 24140 + }, + { + "epoch": 1.78, + "learning_rate": 1.6480684635657013e-05, + "loss": 1.0782, + "step": 24141 + }, + { + "epoch": 1.78, + "learning_rate": 1.6480380857023214e-05, + "loss": 1.0551, + "step": 24142 + }, + { + "epoch": 1.78, + "learning_rate": 1.6480077068079228e-05, + "loss": 0.9439, + "step": 24143 + }, + { + "epoch": 1.78, + "learning_rate": 1.6479773268825527e-05, + "loss": 0.8785, + "step": 24144 + }, + { + "epoch": 1.78, + "learning_rate": 1.64794694592626e-05, + "loss": 1.0612, + "step": 24145 + }, + { + "epoch": 1.78, + "learning_rate": 1.6479165639390932e-05, + "loss": 1.0614, + "step": 24146 + }, + { + "epoch": 1.78, + "learning_rate": 1.6478861809211004e-05, + "loss": 1.0116, + "step": 24147 + }, + { + "epoch": 1.78, + "learning_rate": 1.64785579687233e-05, + "loss": 1.0723, + "step": 24148 + }, + { + "epoch": 1.78, + "learning_rate": 1.64782541179283e-05, + "loss": 1.0054, + "step": 24149 + }, + { + "epoch": 1.78, + "learning_rate": 1.6477950256826492e-05, + "loss": 1.0431, + "step": 24150 + }, + { + "epoch": 1.78, + "learning_rate": 1.6477646385418357e-05, + "loss": 0.9213, + "step": 24151 + }, + { + "epoch": 1.78, + "learning_rate": 1.6477342503704382e-05, + "loss": 1.0789, + "step": 24152 + }, + { + "epoch": 1.78, + "learning_rate": 1.6477038611685045e-05, + "loss": 1.0289, + "step": 24153 + }, + { + "epoch": 1.78, + "learning_rate": 1.6476734709360834e-05, + "loss": 1.032, + "step": 24154 + }, + { + "epoch": 1.78, + "learning_rate": 1.6476430796732227e-05, + "loss": 1.0729, + "step": 24155 + }, + { + "epoch": 1.78, + "learning_rate": 1.6476126873799714e-05, + "loss": 1.0583, + "step": 24156 + }, + { + "epoch": 1.78, + "learning_rate": 1.6475822940563778e-05, + "loss": 1.0092, + "step": 24157 + }, + { + "epoch": 1.79, + "learning_rate": 1.64755189970249e-05, + "loss": 0.9587, + "step": 24158 + }, + { + "epoch": 1.79, + "learning_rate": 1.6475215043183558e-05, + "loss": 0.946, + "step": 24159 + }, + { + "epoch": 1.79, + "learning_rate": 1.6474911079040246e-05, + "loss": 0.9529, + "step": 24160 + }, + { + "epoch": 1.79, + "learning_rate": 1.6474607104595442e-05, + "loss": 1.0397, + "step": 24161 + }, + { + "epoch": 1.79, + "learning_rate": 1.6474303119849633e-05, + "loss": 1.0269, + "step": 24162 + }, + { + "epoch": 1.79, + "learning_rate": 1.64739991248033e-05, + "loss": 1.0756, + "step": 24163 + }, + { + "epoch": 1.79, + "learning_rate": 1.6473695119456925e-05, + "loss": 0.9164, + "step": 24164 + }, + { + "epoch": 1.79, + "learning_rate": 1.6473391103810995e-05, + "loss": 1.0891, + "step": 24165 + }, + { + "epoch": 1.79, + "learning_rate": 1.6473087077865995e-05, + "loss": 1.029, + "step": 24166 + }, + { + "epoch": 1.79, + "learning_rate": 1.64727830416224e-05, + "loss": 1.0104, + "step": 24167 + }, + { + "epoch": 1.79, + "learning_rate": 1.6472478995080705e-05, + "loss": 1.0433, + "step": 24168 + }, + { + "epoch": 1.79, + "learning_rate": 1.6472174938241384e-05, + "loss": 1.004, + "step": 24169 + }, + { + "epoch": 1.79, + "learning_rate": 1.6471870871104928e-05, + "loss": 1.0127, + "step": 24170 + }, + { + "epoch": 1.79, + "learning_rate": 1.6471566793671817e-05, + "loss": 0.9318, + "step": 24171 + }, + { + "epoch": 1.79, + "learning_rate": 1.6471262705942536e-05, + "loss": 1.1063, + "step": 24172 + }, + { + "epoch": 1.79, + "learning_rate": 1.647095860791757e-05, + "loss": 1.0042, + "step": 24173 + }, + { + "epoch": 1.79, + "learning_rate": 1.6470654499597402e-05, + "loss": 0.9821, + "step": 24174 + }, + { + "epoch": 1.79, + "learning_rate": 1.647035038098251e-05, + "loss": 0.9269, + "step": 24175 + }, + { + "epoch": 1.79, + "learning_rate": 1.647004625207339e-05, + "loss": 1.0507, + "step": 24176 + }, + { + "epoch": 1.79, + "learning_rate": 1.6469742112870516e-05, + "loss": 1.0615, + "step": 24177 + }, + { + "epoch": 1.79, + "learning_rate": 1.6469437963374372e-05, + "loss": 0.8929, + "step": 24178 + }, + { + "epoch": 1.79, + "learning_rate": 1.6469133803585447e-05, + "loss": 1.025, + "step": 24179 + }, + { + "epoch": 1.79, + "learning_rate": 1.6468829633504225e-05, + "loss": 1.1225, + "step": 24180 + }, + { + "epoch": 1.79, + "learning_rate": 1.6468525453131187e-05, + "loss": 1.0018, + "step": 24181 + }, + { + "epoch": 1.79, + "learning_rate": 1.6468221262466816e-05, + "loss": 1.0316, + "step": 24182 + }, + { + "epoch": 1.79, + "learning_rate": 1.64679170615116e-05, + "loss": 1.1036, + "step": 24183 + }, + { + "epoch": 1.79, + "learning_rate": 1.646761285026602e-05, + "loss": 1.0146, + "step": 24184 + }, + { + "epoch": 1.79, + "learning_rate": 1.6467308628730557e-05, + "loss": 0.8922, + "step": 24185 + }, + { + "epoch": 1.79, + "learning_rate": 1.6467004396905698e-05, + "loss": 1.0212, + "step": 24186 + }, + { + "epoch": 1.79, + "learning_rate": 1.646670015479193e-05, + "loss": 0.9446, + "step": 24187 + }, + { + "epoch": 1.79, + "learning_rate": 1.6466395902389737e-05, + "loss": 1.0726, + "step": 24188 + }, + { + "epoch": 1.79, + "learning_rate": 1.6466091639699597e-05, + "loss": 0.9955, + "step": 24189 + }, + { + "epoch": 1.79, + "learning_rate": 1.6465787366721998e-05, + "loss": 1.0331, + "step": 24190 + }, + { + "epoch": 1.79, + "learning_rate": 1.6465483083457425e-05, + "loss": 1.0665, + "step": 24191 + }, + { + "epoch": 1.79, + "learning_rate": 1.646517878990636e-05, + "loss": 1.0301, + "step": 24192 + }, + { + "epoch": 1.79, + "learning_rate": 1.6464874486069292e-05, + "loss": 1.1076, + "step": 24193 + }, + { + "epoch": 1.79, + "learning_rate": 1.6464570171946695e-05, + "loss": 1.0533, + "step": 24194 + }, + { + "epoch": 1.79, + "learning_rate": 1.646426584753906e-05, + "loss": 1.0364, + "step": 24195 + }, + { + "epoch": 1.79, + "learning_rate": 1.6463961512846876e-05, + "loss": 1.0189, + "step": 24196 + }, + { + "epoch": 1.79, + "learning_rate": 1.6463657167870616e-05, + "loss": 0.9652, + "step": 24197 + }, + { + "epoch": 1.79, + "learning_rate": 1.6463352812610773e-05, + "loss": 1.0978, + "step": 24198 + }, + { + "epoch": 1.79, + "learning_rate": 1.6463048447067828e-05, + "loss": 0.8815, + "step": 24199 + }, + { + "epoch": 1.79, + "learning_rate": 1.6462744071242265e-05, + "loss": 0.9958, + "step": 24200 + }, + { + "epoch": 1.79, + "learning_rate": 1.6462439685134566e-05, + "loss": 1.0957, + "step": 24201 + }, + { + "epoch": 1.79, + "learning_rate": 1.6462135288745226e-05, + "loss": 1.0499, + "step": 24202 + }, + { + "epoch": 1.79, + "learning_rate": 1.6461830882074715e-05, + "loss": 1.0912, + "step": 24203 + }, + { + "epoch": 1.79, + "learning_rate": 1.6461526465123524e-05, + "loss": 0.9859, + "step": 24204 + }, + { + "epoch": 1.79, + "learning_rate": 1.646122203789214e-05, + "loss": 1.0289, + "step": 24205 + }, + { + "epoch": 1.79, + "learning_rate": 1.646091760038104e-05, + "loss": 0.9639, + "step": 24206 + }, + { + "epoch": 1.79, + "learning_rate": 1.646061315259071e-05, + "loss": 1.0955, + "step": 24207 + }, + { + "epoch": 1.79, + "learning_rate": 1.6460308694521644e-05, + "loss": 0.9774, + "step": 24208 + }, + { + "epoch": 1.79, + "learning_rate": 1.6460004226174313e-05, + "loss": 1.028, + "step": 24209 + }, + { + "epoch": 1.79, + "learning_rate": 1.645969974754921e-05, + "loss": 1.0277, + "step": 24210 + }, + { + "epoch": 1.79, + "learning_rate": 1.645939525864682e-05, + "loss": 1.0157, + "step": 24211 + }, + { + "epoch": 1.79, + "learning_rate": 1.6459090759467625e-05, + "loss": 0.9621, + "step": 24212 + }, + { + "epoch": 1.79, + "learning_rate": 1.6458786250012106e-05, + "loss": 1.0711, + "step": 24213 + }, + { + "epoch": 1.79, + "learning_rate": 1.6458481730280748e-05, + "loss": 1.0095, + "step": 24214 + }, + { + "epoch": 1.79, + "learning_rate": 1.645817720027404e-05, + "loss": 1.0027, + "step": 24215 + }, + { + "epoch": 1.79, + "learning_rate": 1.6457872659992467e-05, + "loss": 0.9672, + "step": 24216 + }, + { + "epoch": 1.79, + "learning_rate": 1.645756810943651e-05, + "loss": 1.0743, + "step": 24217 + }, + { + "epoch": 1.79, + "learning_rate": 1.6457263548606653e-05, + "loss": 0.9597, + "step": 24218 + }, + { + "epoch": 1.79, + "learning_rate": 1.645695897750338e-05, + "loss": 1.0118, + "step": 24219 + }, + { + "epoch": 1.79, + "learning_rate": 1.6456654396127184e-05, + "loss": 1.0036, + "step": 24220 + }, + { + "epoch": 1.79, + "learning_rate": 1.6456349804478538e-05, + "loss": 1.0579, + "step": 24221 + }, + { + "epoch": 1.79, + "learning_rate": 1.6456045202557934e-05, + "loss": 0.9769, + "step": 24222 + }, + { + "epoch": 1.79, + "learning_rate": 1.6455740590365855e-05, + "loss": 0.9669, + "step": 24223 + }, + { + "epoch": 1.79, + "learning_rate": 1.645543596790278e-05, + "loss": 1.0063, + "step": 24224 + }, + { + "epoch": 1.79, + "learning_rate": 1.6455131335169205e-05, + "loss": 1.0956, + "step": 24225 + }, + { + "epoch": 1.79, + "learning_rate": 1.645482669216561e-05, + "loss": 0.9504, + "step": 24226 + }, + { + "epoch": 1.79, + "learning_rate": 1.6454522038892473e-05, + "loss": 1.0434, + "step": 24227 + }, + { + "epoch": 1.79, + "learning_rate": 1.6454217375350286e-05, + "loss": 1.1106, + "step": 24228 + }, + { + "epoch": 1.79, + "learning_rate": 1.645391270153953e-05, + "loss": 1.0653, + "step": 24229 + }, + { + "epoch": 1.79, + "learning_rate": 1.6453608017460693e-05, + "loss": 0.9616, + "step": 24230 + }, + { + "epoch": 1.79, + "learning_rate": 1.6453303323114255e-05, + "loss": 0.9829, + "step": 24231 + }, + { + "epoch": 1.79, + "learning_rate": 1.6452998618500705e-05, + "loss": 0.8863, + "step": 24232 + }, + { + "epoch": 1.79, + "learning_rate": 1.645269390362053e-05, + "loss": 1.0612, + "step": 24233 + }, + { + "epoch": 1.79, + "learning_rate": 1.6452389178474208e-05, + "loss": 1.0145, + "step": 24234 + }, + { + "epoch": 1.79, + "learning_rate": 1.6452084443062227e-05, + "loss": 0.9939, + "step": 24235 + }, + { + "epoch": 1.79, + "learning_rate": 1.645177969738507e-05, + "loss": 1.0002, + "step": 24236 + }, + { + "epoch": 1.79, + "learning_rate": 1.6451474941443228e-05, + "loss": 0.9685, + "step": 24237 + }, + { + "epoch": 1.79, + "learning_rate": 1.6451170175237182e-05, + "loss": 1.0754, + "step": 24238 + }, + { + "epoch": 1.79, + "learning_rate": 1.6450865398767414e-05, + "loss": 1.0061, + "step": 24239 + }, + { + "epoch": 1.79, + "learning_rate": 1.6450560612034412e-05, + "loss": 0.9961, + "step": 24240 + }, + { + "epoch": 1.79, + "learning_rate": 1.645025581503866e-05, + "loss": 1.0053, + "step": 24241 + }, + { + "epoch": 1.79, + "learning_rate": 1.6449951007780644e-05, + "loss": 1.0326, + "step": 24242 + }, + { + "epoch": 1.79, + "learning_rate": 1.6449646190260847e-05, + "loss": 0.9964, + "step": 24243 + }, + { + "epoch": 1.79, + "learning_rate": 1.6449341362479758e-05, + "loss": 1.0472, + "step": 24244 + }, + { + "epoch": 1.79, + "learning_rate": 1.6449036524437857e-05, + "loss": 1.0596, + "step": 24245 + }, + { + "epoch": 1.79, + "learning_rate": 1.644873167613563e-05, + "loss": 1.0219, + "step": 24246 + }, + { + "epoch": 1.79, + "learning_rate": 1.6448426817573566e-05, + "loss": 1.0789, + "step": 24247 + }, + { + "epoch": 1.79, + "learning_rate": 1.6448121948752145e-05, + "loss": 0.8951, + "step": 24248 + }, + { + "epoch": 1.79, + "learning_rate": 1.6447817069671857e-05, + "loss": 1.0287, + "step": 24249 + }, + { + "epoch": 1.79, + "learning_rate": 1.644751218033318e-05, + "loss": 0.9994, + "step": 24250 + }, + { + "epoch": 1.79, + "learning_rate": 1.6447207280736605e-05, + "loss": 1.0825, + "step": 24251 + }, + { + "epoch": 1.79, + "learning_rate": 1.6446902370882618e-05, + "loss": 1.0383, + "step": 24252 + }, + { + "epoch": 1.79, + "learning_rate": 1.64465974507717e-05, + "loss": 0.9587, + "step": 24253 + }, + { + "epoch": 1.79, + "learning_rate": 1.6446292520404337e-05, + "loss": 1.159, + "step": 24254 + }, + { + "epoch": 1.79, + "learning_rate": 1.6445987579781014e-05, + "loss": 1.0506, + "step": 24255 + }, + { + "epoch": 1.79, + "learning_rate": 1.644568262890222e-05, + "loss": 0.8815, + "step": 24256 + }, + { + "epoch": 1.79, + "learning_rate": 1.6445377667768433e-05, + "loss": 0.9761, + "step": 24257 + }, + { + "epoch": 1.79, + "learning_rate": 1.6445072696380148e-05, + "loss": 0.9949, + "step": 24258 + }, + { + "epoch": 1.79, + "learning_rate": 1.644476771473784e-05, + "loss": 1.0462, + "step": 24259 + }, + { + "epoch": 1.79, + "learning_rate": 1.6444462722842e-05, + "loss": 0.9963, + "step": 24260 + }, + { + "epoch": 1.79, + "learning_rate": 1.644415772069311e-05, + "loss": 1.0529, + "step": 24261 + }, + { + "epoch": 1.79, + "learning_rate": 1.644385270829166e-05, + "loss": 1.0395, + "step": 24262 + }, + { + "epoch": 1.79, + "learning_rate": 1.6443547685638135e-05, + "loss": 0.9901, + "step": 24263 + }, + { + "epoch": 1.79, + "learning_rate": 1.6443242652733015e-05, + "loss": 1.0132, + "step": 24264 + }, + { + "epoch": 1.79, + "learning_rate": 1.6442937609576787e-05, + "loss": 1.007, + "step": 24265 + }, + { + "epoch": 1.79, + "learning_rate": 1.6442632556169936e-05, + "loss": 0.9228, + "step": 24266 + }, + { + "epoch": 1.79, + "learning_rate": 1.6442327492512952e-05, + "loss": 0.9717, + "step": 24267 + }, + { + "epoch": 1.79, + "learning_rate": 1.6442022418606318e-05, + "loss": 0.9471, + "step": 24268 + }, + { + "epoch": 1.79, + "learning_rate": 1.6441717334450514e-05, + "loss": 1.1519, + "step": 24269 + }, + { + "epoch": 1.79, + "learning_rate": 1.6441412240046035e-05, + "loss": 1.018, + "step": 24270 + }, + { + "epoch": 1.79, + "learning_rate": 1.6441107135393358e-05, + "loss": 1.0225, + "step": 24271 + }, + { + "epoch": 1.79, + "learning_rate": 1.6440802020492975e-05, + "loss": 1.0119, + "step": 24272 + }, + { + "epoch": 1.79, + "learning_rate": 1.6440496895345367e-05, + "loss": 0.849, + "step": 24273 + }, + { + "epoch": 1.79, + "learning_rate": 1.6440191759951016e-05, + "loss": 1.0074, + "step": 24274 + }, + { + "epoch": 1.79, + "learning_rate": 1.6439886614310418e-05, + "loss": 0.9239, + "step": 24275 + }, + { + "epoch": 1.79, + "learning_rate": 1.643958145842405e-05, + "loss": 1.0546, + "step": 24276 + }, + { + "epoch": 1.79, + "learning_rate": 1.64392762922924e-05, + "loss": 1.0159, + "step": 24277 + }, + { + "epoch": 1.79, + "learning_rate": 1.643897111591595e-05, + "loss": 0.9967, + "step": 24278 + }, + { + "epoch": 1.79, + "learning_rate": 1.6438665929295195e-05, + "loss": 0.9737, + "step": 24279 + }, + { + "epoch": 1.79, + "learning_rate": 1.643836073243061e-05, + "loss": 1.017, + "step": 24280 + }, + { + "epoch": 1.79, + "learning_rate": 1.643805552532269e-05, + "loss": 1.0155, + "step": 24281 + }, + { + "epoch": 1.79, + "learning_rate": 1.643775030797191e-05, + "loss": 1.0759, + "step": 24282 + }, + { + "epoch": 1.79, + "learning_rate": 1.6437445080378767e-05, + "loss": 1.0497, + "step": 24283 + }, + { + "epoch": 1.79, + "learning_rate": 1.6437139842543736e-05, + "loss": 1.1018, + "step": 24284 + }, + { + "epoch": 1.79, + "learning_rate": 1.6436834594467314e-05, + "loss": 1.0161, + "step": 24285 + }, + { + "epoch": 1.79, + "learning_rate": 1.6436529336149975e-05, + "loss": 1.113, + "step": 24286 + }, + { + "epoch": 1.79, + "learning_rate": 1.6436224067592212e-05, + "loss": 1.0143, + "step": 24287 + }, + { + "epoch": 1.79, + "learning_rate": 1.6435918788794508e-05, + "loss": 1.014, + "step": 24288 + }, + { + "epoch": 1.79, + "learning_rate": 1.6435613499757348e-05, + "loss": 1.0235, + "step": 24289 + }, + { + "epoch": 1.79, + "learning_rate": 1.643530820048122e-05, + "loss": 1.0802, + "step": 24290 + }, + { + "epoch": 1.79, + "learning_rate": 1.6435002890966608e-05, + "loss": 0.976, + "step": 24291 + }, + { + "epoch": 1.79, + "learning_rate": 1.6434697571214e-05, + "loss": 1.12, + "step": 24292 + }, + { + "epoch": 1.79, + "learning_rate": 1.643439224122388e-05, + "loss": 1.1027, + "step": 24293 + }, + { + "epoch": 1.8, + "learning_rate": 1.6434086900996736e-05, + "loss": 1.1032, + "step": 24294 + }, + { + "epoch": 1.8, + "learning_rate": 1.6433781550533048e-05, + "loss": 1.07, + "step": 24295 + }, + { + "epoch": 1.8, + "learning_rate": 1.643347618983331e-05, + "loss": 1.0927, + "step": 24296 + }, + { + "epoch": 1.8, + "learning_rate": 1.6433170818897998e-05, + "loss": 1.0246, + "step": 24297 + }, + { + "epoch": 1.8, + "learning_rate": 1.6432865437727607e-05, + "loss": 0.9629, + "step": 24298 + }, + { + "epoch": 1.8, + "learning_rate": 1.6432560046322616e-05, + "loss": 1.0198, + "step": 24299 + }, + { + "epoch": 1.8, + "learning_rate": 1.6432254644683518e-05, + "loss": 1.0868, + "step": 24300 + }, + { + "epoch": 1.8, + "learning_rate": 1.6431949232810792e-05, + "loss": 1.0189, + "step": 24301 + }, + { + "epoch": 1.8, + "learning_rate": 1.643164381070493e-05, + "loss": 0.9968, + "step": 24302 + }, + { + "epoch": 1.8, + "learning_rate": 1.643133837836641e-05, + "loss": 0.9967, + "step": 24303 + }, + { + "epoch": 1.8, + "learning_rate": 1.6431032935795722e-05, + "loss": 1.0521, + "step": 24304 + }, + { + "epoch": 1.8, + "learning_rate": 1.643072748299336e-05, + "loss": 1.1317, + "step": 24305 + }, + { + "epoch": 1.8, + "learning_rate": 1.6430422019959794e-05, + "loss": 1.0139, + "step": 24306 + }, + { + "epoch": 1.8, + "learning_rate": 1.6430116546695526e-05, + "loss": 0.9928, + "step": 24307 + }, + { + "epoch": 1.8, + "learning_rate": 1.642981106320103e-05, + "loss": 0.9805, + "step": 24308 + }, + { + "epoch": 1.8, + "learning_rate": 1.6429505569476798e-05, + "loss": 0.9451, + "step": 24309 + }, + { + "epoch": 1.8, + "learning_rate": 1.6429200065523313e-05, + "loss": 1.0352, + "step": 24310 + }, + { + "epoch": 1.8, + "learning_rate": 1.6428894551341065e-05, + "loss": 1.028, + "step": 24311 + }, + { + "epoch": 1.8, + "learning_rate": 1.6428589026930535e-05, + "loss": 1.041, + "step": 24312 + }, + { + "epoch": 1.8, + "learning_rate": 1.6428283492292215e-05, + "loss": 1.0928, + "step": 24313 + }, + { + "epoch": 1.8, + "learning_rate": 1.642797794742659e-05, + "loss": 1.0333, + "step": 24314 + }, + { + "epoch": 1.8, + "learning_rate": 1.642767239233414e-05, + "loss": 1.0826, + "step": 24315 + }, + { + "epoch": 1.8, + "learning_rate": 1.6427366827015355e-05, + "loss": 0.8637, + "step": 24316 + }, + { + "epoch": 1.8, + "learning_rate": 1.6427061251470723e-05, + "loss": 0.9888, + "step": 24317 + }, + { + "epoch": 1.8, + "learning_rate": 1.642675566570073e-05, + "loss": 1.0449, + "step": 24318 + }, + { + "epoch": 1.8, + "learning_rate": 1.6426450069705854e-05, + "loss": 0.9761, + "step": 24319 + }, + { + "epoch": 1.8, + "learning_rate": 1.6426144463486595e-05, + "loss": 1.1037, + "step": 24320 + }, + { + "epoch": 1.8, + "learning_rate": 1.642583884704343e-05, + "loss": 0.9768, + "step": 24321 + }, + { + "epoch": 1.8, + "learning_rate": 1.6425533220376845e-05, + "loss": 1.011, + "step": 24322 + }, + { + "epoch": 1.8, + "learning_rate": 1.642522758348733e-05, + "loss": 1.0071, + "step": 24323 + }, + { + "epoch": 1.8, + "learning_rate": 1.642492193637537e-05, + "loss": 0.9162, + "step": 24324 + }, + { + "epoch": 1.8, + "learning_rate": 1.642461627904145e-05, + "loss": 0.8997, + "step": 24325 + }, + { + "epoch": 1.8, + "learning_rate": 1.642431061148606e-05, + "loss": 1.0081, + "step": 24326 + }, + { + "epoch": 1.8, + "learning_rate": 1.6424004933709685e-05, + "loss": 1.0868, + "step": 24327 + }, + { + "epoch": 1.8, + "learning_rate": 1.6423699245712807e-05, + "loss": 1.0142, + "step": 24328 + }, + { + "epoch": 1.8, + "learning_rate": 1.6423393547495914e-05, + "loss": 1.0057, + "step": 24329 + }, + { + "epoch": 1.8, + "learning_rate": 1.6423087839059497e-05, + "loss": 0.9903, + "step": 24330 + }, + { + "epoch": 1.8, + "learning_rate": 1.6422782120404038e-05, + "loss": 0.9308, + "step": 24331 + }, + { + "epoch": 1.8, + "learning_rate": 1.6422476391530024e-05, + "loss": 0.9298, + "step": 24332 + }, + { + "epoch": 1.8, + "learning_rate": 1.6422170652437944e-05, + "loss": 0.9756, + "step": 24333 + }, + { + "epoch": 1.8, + "learning_rate": 1.642186490312828e-05, + "loss": 0.9282, + "step": 24334 + }, + { + "epoch": 1.8, + "learning_rate": 1.642155914360152e-05, + "loss": 1.0436, + "step": 24335 + }, + { + "epoch": 1.8, + "learning_rate": 1.6421253373858154e-05, + "loss": 1.0723, + "step": 24336 + }, + { + "epoch": 1.8, + "learning_rate": 1.6420947593898665e-05, + "loss": 1.0171, + "step": 24337 + }, + { + "epoch": 1.8, + "learning_rate": 1.642064180372354e-05, + "loss": 0.9689, + "step": 24338 + }, + { + "epoch": 1.8, + "learning_rate": 1.6420336003333266e-05, + "loss": 1.0145, + "step": 24339 + }, + { + "epoch": 1.8, + "learning_rate": 1.6420030192728328e-05, + "loss": 1.0491, + "step": 24340 + }, + { + "epoch": 1.8, + "learning_rate": 1.6419724371909213e-05, + "loss": 1.0643, + "step": 24341 + }, + { + "epoch": 1.8, + "learning_rate": 1.6419418540876412e-05, + "loss": 0.9978, + "step": 24342 + }, + { + "epoch": 1.8, + "learning_rate": 1.641911269963041e-05, + "loss": 0.9916, + "step": 24343 + }, + { + "epoch": 1.8, + "learning_rate": 1.6418806848171684e-05, + "loss": 0.9376, + "step": 24344 + }, + { + "epoch": 1.8, + "learning_rate": 1.6418500986500732e-05, + "loss": 0.8711, + "step": 24345 + }, + { + "epoch": 1.8, + "learning_rate": 1.6418195114618036e-05, + "loss": 0.9826, + "step": 24346 + }, + { + "epoch": 1.8, + "learning_rate": 1.6417889232524084e-05, + "loss": 1.0235, + "step": 24347 + }, + { + "epoch": 1.8, + "learning_rate": 1.6417583340219362e-05, + "loss": 0.9945, + "step": 24348 + }, + { + "epoch": 1.8, + "learning_rate": 1.6417277437704357e-05, + "loss": 1.0874, + "step": 24349 + }, + { + "epoch": 1.8, + "learning_rate": 1.6416971524979555e-05, + "loss": 0.9027, + "step": 24350 + }, + { + "epoch": 1.8, + "learning_rate": 1.6416665602045444e-05, + "loss": 0.9971, + "step": 24351 + }, + { + "epoch": 1.8, + "learning_rate": 1.641635966890251e-05, + "loss": 0.9923, + "step": 24352 + }, + { + "epoch": 1.8, + "learning_rate": 1.641605372555124e-05, + "loss": 1.0053, + "step": 24353 + }, + { + "epoch": 1.8, + "learning_rate": 1.6415747771992118e-05, + "loss": 1.1006, + "step": 24354 + }, + { + "epoch": 1.8, + "learning_rate": 1.6415441808225632e-05, + "loss": 1.0899, + "step": 24355 + }, + { + "epoch": 1.8, + "learning_rate": 1.641513583425227e-05, + "loss": 1.0352, + "step": 24356 + }, + { + "epoch": 1.8, + "learning_rate": 1.6414829850072522e-05, + "loss": 0.9831, + "step": 24357 + }, + { + "epoch": 1.8, + "learning_rate": 1.641452385568687e-05, + "loss": 0.9538, + "step": 24358 + }, + { + "epoch": 1.8, + "learning_rate": 1.6414217851095804e-05, + "loss": 0.9322, + "step": 24359 + }, + { + "epoch": 1.8, + "learning_rate": 1.641391183629981e-05, + "loss": 0.9875, + "step": 24360 + }, + { + "epoch": 1.8, + "learning_rate": 1.641360581129937e-05, + "loss": 1.0125, + "step": 24361 + }, + { + "epoch": 1.8, + "learning_rate": 1.6413299776094976e-05, + "loss": 1.1158, + "step": 24362 + }, + { + "epoch": 1.8, + "learning_rate": 1.6412993730687115e-05, + "loss": 1.012, + "step": 24363 + }, + { + "epoch": 1.8, + "learning_rate": 1.641268767507627e-05, + "loss": 0.9944, + "step": 24364 + }, + { + "epoch": 1.8, + "learning_rate": 1.6412381609262932e-05, + "loss": 1.0371, + "step": 24365 + }, + { + "epoch": 1.8, + "learning_rate": 1.6412075533247593e-05, + "loss": 0.9312, + "step": 24366 + }, + { + "epoch": 1.8, + "learning_rate": 1.6411769447030724e-05, + "loss": 1.0583, + "step": 24367 + }, + { + "epoch": 1.8, + "learning_rate": 1.6411463350612825e-05, + "loss": 1.1481, + "step": 24368 + }, + { + "epoch": 1.8, + "learning_rate": 1.641115724399438e-05, + "loss": 0.9545, + "step": 24369 + }, + { + "epoch": 1.8, + "learning_rate": 1.6410851127175876e-05, + "loss": 1.0122, + "step": 24370 + }, + { + "epoch": 1.8, + "learning_rate": 1.64105450001578e-05, + "loss": 0.966, + "step": 24371 + }, + { + "epoch": 1.8, + "learning_rate": 1.6410238862940637e-05, + "loss": 1.0689, + "step": 24372 + }, + { + "epoch": 1.8, + "learning_rate": 1.6409932715524876e-05, + "loss": 1.0789, + "step": 24373 + }, + { + "epoch": 1.8, + "learning_rate": 1.6409626557911004e-05, + "loss": 0.9885, + "step": 24374 + }, + { + "epoch": 1.8, + "learning_rate": 1.6409320390099506e-05, + "loss": 1.0949, + "step": 24375 + }, + { + "epoch": 1.8, + "learning_rate": 1.6409014212090877e-05, + "loss": 0.9542, + "step": 24376 + }, + { + "epoch": 1.8, + "learning_rate": 1.6408708023885593e-05, + "loss": 1.0944, + "step": 24377 + }, + { + "epoch": 1.8, + "learning_rate": 1.6408401825484148e-05, + "loss": 0.9988, + "step": 24378 + }, + { + "epoch": 1.8, + "learning_rate": 1.6408095616887027e-05, + "loss": 0.9574, + "step": 24379 + }, + { + "epoch": 1.8, + "learning_rate": 1.6407789398094717e-05, + "loss": 1.0955, + "step": 24380 + }, + { + "epoch": 1.8, + "learning_rate": 1.6407483169107704e-05, + "loss": 0.9399, + "step": 24381 + }, + { + "epoch": 1.8, + "learning_rate": 1.640717692992648e-05, + "loss": 1.1072, + "step": 24382 + }, + { + "epoch": 1.8, + "learning_rate": 1.640687068055153e-05, + "loss": 1.0904, + "step": 24383 + }, + { + "epoch": 1.8, + "learning_rate": 1.6406564420983338e-05, + "loss": 1.0108, + "step": 24384 + }, + { + "epoch": 1.8, + "learning_rate": 1.6406258151222392e-05, + "loss": 1.0571, + "step": 24385 + }, + { + "epoch": 1.8, + "learning_rate": 1.6405951871269183e-05, + "loss": 1.0388, + "step": 24386 + }, + { + "epoch": 1.8, + "learning_rate": 1.64056455811242e-05, + "loss": 1.1659, + "step": 24387 + }, + { + "epoch": 1.8, + "learning_rate": 1.6405339280787922e-05, + "loss": 0.9764, + "step": 24388 + }, + { + "epoch": 1.8, + "learning_rate": 1.6405032970260843e-05, + "loss": 1.0286, + "step": 24389 + }, + { + "epoch": 1.8, + "learning_rate": 1.6404726649543445e-05, + "loss": 1.0467, + "step": 24390 + }, + { + "epoch": 1.8, + "learning_rate": 1.640442031863622e-05, + "loss": 0.8016, + "step": 24391 + }, + { + "epoch": 1.8, + "learning_rate": 1.6404113977539654e-05, + "loss": 1.041, + "step": 24392 + }, + { + "epoch": 1.8, + "learning_rate": 1.6403807626254236e-05, + "loss": 1.1048, + "step": 24393 + }, + { + "epoch": 1.8, + "learning_rate": 1.6403501264780452e-05, + "loss": 1.1139, + "step": 24394 + }, + { + "epoch": 1.8, + "learning_rate": 1.6403194893118786e-05, + "loss": 1.0922, + "step": 24395 + }, + { + "epoch": 1.8, + "learning_rate": 1.640288851126973e-05, + "loss": 0.9571, + "step": 24396 + }, + { + "epoch": 1.8, + "learning_rate": 1.640258211923377e-05, + "loss": 1.0103, + "step": 24397 + }, + { + "epoch": 1.8, + "learning_rate": 1.6402275717011393e-05, + "loss": 1.0467, + "step": 24398 + }, + { + "epoch": 1.8, + "learning_rate": 1.640196930460309e-05, + "loss": 1.0099, + "step": 24399 + }, + { + "epoch": 1.8, + "learning_rate": 1.6401662882009342e-05, + "loss": 1.1371, + "step": 24400 + }, + { + "epoch": 1.8, + "learning_rate": 1.640135644923064e-05, + "loss": 0.9878, + "step": 24401 + }, + { + "epoch": 1.8, + "learning_rate": 1.6401050006267472e-05, + "loss": 0.9719, + "step": 24402 + }, + { + "epoch": 1.8, + "learning_rate": 1.6400743553120326e-05, + "loss": 1.0154, + "step": 24403 + }, + { + "epoch": 1.8, + "learning_rate": 1.6400437089789688e-05, + "loss": 0.9676, + "step": 24404 + }, + { + "epoch": 1.8, + "learning_rate": 1.6400130616276044e-05, + "loss": 0.9162, + "step": 24405 + }, + { + "epoch": 1.8, + "learning_rate": 1.6399824132579888e-05, + "loss": 1.0185, + "step": 24406 + }, + { + "epoch": 1.8, + "learning_rate": 1.6399517638701698e-05, + "loss": 0.9215, + "step": 24407 + }, + { + "epoch": 1.8, + "learning_rate": 1.6399211134641976e-05, + "loss": 1.0943, + "step": 24408 + }, + { + "epoch": 1.8, + "learning_rate": 1.639890462040119e-05, + "loss": 0.9891, + "step": 24409 + }, + { + "epoch": 1.8, + "learning_rate": 1.6398598095979845e-05, + "loss": 0.9941, + "step": 24410 + }, + { + "epoch": 1.8, + "learning_rate": 1.639829156137842e-05, + "loss": 0.931, + "step": 24411 + }, + { + "epoch": 1.8, + "learning_rate": 1.6397985016597404e-05, + "loss": 1.0844, + "step": 24412 + }, + { + "epoch": 1.8, + "learning_rate": 1.6397678461637287e-05, + "loss": 1.0155, + "step": 24413 + }, + { + "epoch": 1.8, + "learning_rate": 1.6397371896498554e-05, + "loss": 1.0371, + "step": 24414 + }, + { + "epoch": 1.8, + "learning_rate": 1.6397065321181694e-05, + "loss": 1.0172, + "step": 24415 + }, + { + "epoch": 1.8, + "learning_rate": 1.6396758735687194e-05, + "loss": 0.94, + "step": 24416 + }, + { + "epoch": 1.8, + "learning_rate": 1.6396452140015546e-05, + "loss": 1.0861, + "step": 24417 + }, + { + "epoch": 1.8, + "learning_rate": 1.639614553416723e-05, + "loss": 0.9267, + "step": 24418 + }, + { + "epoch": 1.8, + "learning_rate": 1.6395838918142736e-05, + "loss": 1.0586, + "step": 24419 + }, + { + "epoch": 1.8, + "learning_rate": 1.639553229194256e-05, + "loss": 0.9903, + "step": 24420 + }, + { + "epoch": 1.8, + "learning_rate": 1.639522565556718e-05, + "loss": 1.0812, + "step": 24421 + }, + { + "epoch": 1.8, + "learning_rate": 1.639491900901709e-05, + "loss": 0.9644, + "step": 24422 + }, + { + "epoch": 1.8, + "learning_rate": 1.639461235229277e-05, + "loss": 1.1072, + "step": 24423 + }, + { + "epoch": 1.8, + "learning_rate": 1.6394305685394717e-05, + "loss": 1.0885, + "step": 24424 + }, + { + "epoch": 1.8, + "learning_rate": 1.6393999008323417e-05, + "loss": 0.992, + "step": 24425 + }, + { + "epoch": 1.8, + "learning_rate": 1.639369232107935e-05, + "loss": 1.0409, + "step": 24426 + }, + { + "epoch": 1.8, + "learning_rate": 1.6393385623663017e-05, + "loss": 1.0946, + "step": 24427 + }, + { + "epoch": 1.8, + "learning_rate": 1.6393078916074895e-05, + "loss": 0.9687, + "step": 24428 + }, + { + "epoch": 1.81, + "learning_rate": 1.6392772198315475e-05, + "loss": 1.0186, + "step": 24429 + }, + { + "epoch": 1.81, + "learning_rate": 1.639246547038525e-05, + "loss": 0.8663, + "step": 24430 + }, + { + "epoch": 1.81, + "learning_rate": 1.6392158732284703e-05, + "loss": 0.8636, + "step": 24431 + }, + { + "epoch": 1.81, + "learning_rate": 1.639185198401432e-05, + "loss": 0.9995, + "step": 24432 + }, + { + "epoch": 1.81, + "learning_rate": 1.639154522557459e-05, + "loss": 0.9629, + "step": 24433 + }, + { + "epoch": 1.81, + "learning_rate": 1.639123845696601e-05, + "loss": 0.9173, + "step": 24434 + }, + { + "epoch": 1.81, + "learning_rate": 1.6390931678189056e-05, + "loss": 1.0572, + "step": 24435 + }, + { + "epoch": 1.81, + "learning_rate": 1.6390624889244225e-05, + "loss": 1.0835, + "step": 24436 + }, + { + "epoch": 1.81, + "learning_rate": 1.6390318090131997e-05, + "loss": 1.0185, + "step": 24437 + }, + { + "epoch": 1.81, + "learning_rate": 1.639001128085287e-05, + "loss": 1.0919, + "step": 24438 + }, + { + "epoch": 1.81, + "learning_rate": 1.638970446140732e-05, + "loss": 0.9553, + "step": 24439 + }, + { + "epoch": 1.81, + "learning_rate": 1.6389397631795842e-05, + "loss": 0.9451, + "step": 24440 + }, + { + "epoch": 1.81, + "learning_rate": 1.6389090792018928e-05, + "loss": 1.0379, + "step": 24441 + }, + { + "epoch": 1.81, + "learning_rate": 1.638878394207706e-05, + "loss": 0.9574, + "step": 24442 + }, + { + "epoch": 1.81, + "learning_rate": 1.6388477081970727e-05, + "loss": 0.9944, + "step": 24443 + }, + { + "epoch": 1.81, + "learning_rate": 1.638817021170042e-05, + "loss": 1.0323, + "step": 24444 + }, + { + "epoch": 1.81, + "learning_rate": 1.638786333126663e-05, + "loss": 0.9385, + "step": 24445 + }, + { + "epoch": 1.81, + "learning_rate": 1.638755644066983e-05, + "loss": 0.9765, + "step": 24446 + }, + { + "epoch": 1.81, + "learning_rate": 1.638724953991053e-05, + "loss": 1.0043, + "step": 24447 + }, + { + "epoch": 1.81, + "learning_rate": 1.63869426289892e-05, + "loss": 0.9848, + "step": 24448 + }, + { + "epoch": 1.81, + "learning_rate": 1.6386635707906338e-05, + "loss": 1.1374, + "step": 24449 + }, + { + "epoch": 1.81, + "learning_rate": 1.6386328776662432e-05, + "loss": 1.048, + "step": 24450 + }, + { + "epoch": 1.81, + "learning_rate": 1.6386021835257966e-05, + "loss": 1.0712, + "step": 24451 + }, + { + "epoch": 1.81, + "learning_rate": 1.6385714883693427e-05, + "loss": 1.0851, + "step": 24452 + }, + { + "epoch": 1.81, + "learning_rate": 1.638540792196931e-05, + "loss": 1.0621, + "step": 24453 + }, + { + "epoch": 1.81, + "learning_rate": 1.6385100950086106e-05, + "loss": 0.963, + "step": 24454 + }, + { + "epoch": 1.81, + "learning_rate": 1.6384793968044293e-05, + "loss": 0.9441, + "step": 24455 + }, + { + "epoch": 1.81, + "learning_rate": 1.6384486975844365e-05, + "loss": 0.9636, + "step": 24456 + }, + { + "epoch": 1.81, + "learning_rate": 1.638417997348681e-05, + "loss": 0.9887, + "step": 24457 + }, + { + "epoch": 1.81, + "learning_rate": 1.638387296097211e-05, + "loss": 1.0166, + "step": 24458 + }, + { + "epoch": 1.81, + "learning_rate": 1.638356593830077e-05, + "loss": 1.0883, + "step": 24459 + }, + { + "epoch": 1.81, + "learning_rate": 1.6383258905473258e-05, + "loss": 1.0752, + "step": 24460 + }, + { + "epoch": 1.81, + "learning_rate": 1.6382951862490077e-05, + "loss": 0.9752, + "step": 24461 + }, + { + "epoch": 1.81, + "learning_rate": 1.638264480935171e-05, + "loss": 1.0281, + "step": 24462 + }, + { + "epoch": 1.81, + "learning_rate": 1.6382337746058646e-05, + "loss": 1.1298, + "step": 24463 + }, + { + "epoch": 1.81, + "learning_rate": 1.6382030672611376e-05, + "loss": 1.0694, + "step": 24464 + }, + { + "epoch": 1.81, + "learning_rate": 1.6381723589010386e-05, + "loss": 1.0933, + "step": 24465 + }, + { + "epoch": 1.81, + "learning_rate": 1.6381416495256165e-05, + "loss": 1.1088, + "step": 24466 + }, + { + "epoch": 1.81, + "learning_rate": 1.63811093913492e-05, + "loss": 1.0023, + "step": 24467 + }, + { + "epoch": 1.81, + "learning_rate": 1.6380802277289984e-05, + "loss": 1.0471, + "step": 24468 + }, + { + "epoch": 1.81, + "learning_rate": 1.6380495153079003e-05, + "loss": 0.9363, + "step": 24469 + }, + { + "epoch": 1.81, + "learning_rate": 1.638018801871674e-05, + "loss": 1.0261, + "step": 24470 + }, + { + "epoch": 1.81, + "learning_rate": 1.637988087420369e-05, + "loss": 1.0075, + "step": 24471 + }, + { + "epoch": 1.81, + "learning_rate": 1.6379573719540348e-05, + "loss": 1.0272, + "step": 24472 + }, + { + "epoch": 1.81, + "learning_rate": 1.637926655472719e-05, + "loss": 1.0249, + "step": 24473 + }, + { + "epoch": 1.81, + "learning_rate": 1.6378959379764713e-05, + "loss": 1.0204, + "step": 24474 + }, + { + "epoch": 1.81, + "learning_rate": 1.63786521946534e-05, + "loss": 0.9352, + "step": 24475 + }, + { + "epoch": 1.81, + "learning_rate": 1.6378344999393746e-05, + "loss": 0.9519, + "step": 24476 + }, + { + "epoch": 1.81, + "learning_rate": 1.6378037793986234e-05, + "loss": 1.1123, + "step": 24477 + }, + { + "epoch": 1.81, + "learning_rate": 1.6377730578431355e-05, + "loss": 0.9003, + "step": 24478 + }, + { + "epoch": 1.81, + "learning_rate": 1.6377423352729598e-05, + "loss": 1.013, + "step": 24479 + }, + { + "epoch": 1.81, + "learning_rate": 1.637711611688145e-05, + "loss": 1.1416, + "step": 24480 + }, + { + "epoch": 1.81, + "learning_rate": 1.6376808870887405e-05, + "loss": 1.0573, + "step": 24481 + }, + { + "epoch": 1.81, + "learning_rate": 1.637650161474795e-05, + "loss": 1.0353, + "step": 24482 + }, + { + "epoch": 1.81, + "learning_rate": 1.637619434846357e-05, + "loss": 0.9935, + "step": 24483 + }, + { + "epoch": 1.81, + "learning_rate": 1.637588707203475e-05, + "loss": 0.9663, + "step": 24484 + }, + { + "epoch": 1.81, + "learning_rate": 1.637557978546199e-05, + "loss": 1.0765, + "step": 24485 + }, + { + "epoch": 1.81, + "learning_rate": 1.6375272488745775e-05, + "loss": 1.0341, + "step": 24486 + }, + { + "epoch": 1.81, + "learning_rate": 1.6374965181886592e-05, + "loss": 1.0395, + "step": 24487 + }, + { + "epoch": 1.81, + "learning_rate": 1.637465786488493e-05, + "loss": 0.9255, + "step": 24488 + }, + { + "epoch": 1.81, + "learning_rate": 1.6374350537741276e-05, + "loss": 1.006, + "step": 24489 + }, + { + "epoch": 1.81, + "learning_rate": 1.6374043200456124e-05, + "loss": 1.127, + "step": 24490 + }, + { + "epoch": 1.81, + "learning_rate": 1.6373735853029962e-05, + "loss": 1.0831, + "step": 24491 + }, + { + "epoch": 1.81, + "learning_rate": 1.6373428495463277e-05, + "loss": 0.9734, + "step": 24492 + }, + { + "epoch": 1.81, + "learning_rate": 1.6373121127756558e-05, + "loss": 0.9351, + "step": 24493 + }, + { + "epoch": 1.81, + "learning_rate": 1.6372813749910292e-05, + "loss": 0.9752, + "step": 24494 + }, + { + "epoch": 1.81, + "learning_rate": 1.6372506361924974e-05, + "loss": 1.0258, + "step": 24495 + }, + { + "epoch": 1.81, + "learning_rate": 1.6372198963801088e-05, + "loss": 1.0399, + "step": 24496 + }, + { + "epoch": 1.81, + "learning_rate": 1.6371891555539122e-05, + "loss": 1.1044, + "step": 24497 + }, + { + "epoch": 1.81, + "learning_rate": 1.637158413713957e-05, + "loss": 1.0943, + "step": 24498 + }, + { + "epoch": 1.81, + "learning_rate": 1.637127670860292e-05, + "loss": 1.0071, + "step": 24499 + }, + { + "epoch": 1.81, + "learning_rate": 1.637096926992966e-05, + "loss": 0.8955, + "step": 24500 + }, + { + "epoch": 1.81, + "learning_rate": 1.637066182112028e-05, + "loss": 0.9924, + "step": 24501 + }, + { + "epoch": 1.81, + "learning_rate": 1.6370354362175265e-05, + "loss": 1.0196, + "step": 24502 + }, + { + "epoch": 1.81, + "learning_rate": 1.6370046893095107e-05, + "loss": 1.0928, + "step": 24503 + }, + { + "epoch": 1.81, + "learning_rate": 1.63697394138803e-05, + "loss": 0.9399, + "step": 24504 + }, + { + "epoch": 1.81, + "learning_rate": 1.6369431924531323e-05, + "loss": 1.0692, + "step": 24505 + }, + { + "epoch": 1.81, + "learning_rate": 1.6369124425048676e-05, + "loss": 0.9749, + "step": 24506 + }, + { + "epoch": 1.81, + "learning_rate": 1.636881691543284e-05, + "loss": 1.0487, + "step": 24507 + }, + { + "epoch": 1.81, + "learning_rate": 1.6368509395684308e-05, + "loss": 0.9943, + "step": 24508 + }, + { + "epoch": 1.81, + "learning_rate": 1.6368201865803566e-05, + "loss": 0.9514, + "step": 24509 + }, + { + "epoch": 1.81, + "learning_rate": 1.636789432579111e-05, + "loss": 1.0129, + "step": 24510 + }, + { + "epoch": 1.81, + "learning_rate": 1.6367586775647427e-05, + "loss": 0.9786, + "step": 24511 + }, + { + "epoch": 1.81, + "learning_rate": 1.6367279215373e-05, + "loss": 1.0031, + "step": 24512 + }, + { + "epoch": 1.81, + "learning_rate": 1.6366971644968327e-05, + "loss": 1.0782, + "step": 24513 + }, + { + "epoch": 1.81, + "learning_rate": 1.636666406443389e-05, + "loss": 1.0526, + "step": 24514 + }, + { + "epoch": 1.81, + "learning_rate": 1.6366356473770182e-05, + "loss": 1.0848, + "step": 24515 + }, + { + "epoch": 1.81, + "learning_rate": 1.6366048872977692e-05, + "loss": 1.0007, + "step": 24516 + }, + { + "epoch": 1.81, + "learning_rate": 1.636574126205691e-05, + "loss": 0.9907, + "step": 24517 + }, + { + "epoch": 1.81, + "learning_rate": 1.6365433641008325e-05, + "loss": 1.0463, + "step": 24518 + }, + { + "epoch": 1.81, + "learning_rate": 1.6365126009832424e-05, + "loss": 1.0453, + "step": 24519 + }, + { + "epoch": 1.81, + "learning_rate": 1.63648183685297e-05, + "loss": 0.8734, + "step": 24520 + }, + { + "epoch": 1.81, + "learning_rate": 1.6364510717100644e-05, + "loss": 0.9606, + "step": 24521 + }, + { + "epoch": 1.81, + "learning_rate": 1.6364203055545737e-05, + "loss": 0.9693, + "step": 24522 + }, + { + "epoch": 1.81, + "learning_rate": 1.6363895383865476e-05, + "loss": 1.0556, + "step": 24523 + }, + { + "epoch": 1.81, + "learning_rate": 1.636358770206035e-05, + "loss": 0.9214, + "step": 24524 + }, + { + "epoch": 1.81, + "learning_rate": 1.6363280010130848e-05, + "loss": 1.0064, + "step": 24525 + }, + { + "epoch": 1.81, + "learning_rate": 1.6362972308077452e-05, + "loss": 1.0059, + "step": 24526 + }, + { + "epoch": 1.81, + "learning_rate": 1.6362664595900665e-05, + "loss": 0.9526, + "step": 24527 + }, + { + "epoch": 1.81, + "learning_rate": 1.636235687360097e-05, + "loss": 1.0891, + "step": 24528 + }, + { + "epoch": 1.81, + "learning_rate": 1.636204914117885e-05, + "loss": 1.0735, + "step": 24529 + }, + { + "epoch": 1.81, + "learning_rate": 1.6361741398634808e-05, + "loss": 1.0726, + "step": 24530 + }, + { + "epoch": 1.81, + "learning_rate": 1.6361433645969324e-05, + "loss": 1.0539, + "step": 24531 + }, + { + "epoch": 1.81, + "learning_rate": 1.6361125883182887e-05, + "loss": 1.0512, + "step": 24532 + }, + { + "epoch": 1.81, + "learning_rate": 1.6360818110275992e-05, + "loss": 1.0029, + "step": 24533 + }, + { + "epoch": 1.81, + "learning_rate": 1.6360510327249127e-05, + "loss": 0.9398, + "step": 24534 + }, + { + "epoch": 1.81, + "learning_rate": 1.636020253410278e-05, + "loss": 1.0731, + "step": 24535 + }, + { + "epoch": 1.81, + "learning_rate": 1.635989473083744e-05, + "loss": 0.9041, + "step": 24536 + }, + { + "epoch": 1.81, + "learning_rate": 1.6359586917453603e-05, + "loss": 1.1416, + "step": 24537 + }, + { + "epoch": 1.81, + "learning_rate": 1.635927909395175e-05, + "loss": 1.0746, + "step": 24538 + }, + { + "epoch": 1.81, + "learning_rate": 1.6358971260332376e-05, + "loss": 1.0207, + "step": 24539 + }, + { + "epoch": 1.81, + "learning_rate": 1.6358663416595972e-05, + "loss": 1.0781, + "step": 24540 + }, + { + "epoch": 1.81, + "learning_rate": 1.6358355562743024e-05, + "loss": 1.0053, + "step": 24541 + }, + { + "epoch": 1.81, + "learning_rate": 1.6358047698774023e-05, + "loss": 0.9522, + "step": 24542 + }, + { + "epoch": 1.81, + "learning_rate": 1.635773982468946e-05, + "loss": 1.0428, + "step": 24543 + }, + { + "epoch": 1.81, + "learning_rate": 1.635743194048982e-05, + "loss": 1.1279, + "step": 24544 + }, + { + "epoch": 1.81, + "learning_rate": 1.6357124046175603e-05, + "loss": 0.9732, + "step": 24545 + }, + { + "epoch": 1.81, + "learning_rate": 1.635681614174729e-05, + "loss": 0.9682, + "step": 24546 + }, + { + "epoch": 1.81, + "learning_rate": 1.6356508227205374e-05, + "loss": 1.0646, + "step": 24547 + }, + { + "epoch": 1.81, + "learning_rate": 1.635620030255034e-05, + "loss": 1.0281, + "step": 24548 + }, + { + "epoch": 1.81, + "learning_rate": 1.6355892367782687e-05, + "loss": 0.9918, + "step": 24549 + }, + { + "epoch": 1.81, + "learning_rate": 1.63555844229029e-05, + "loss": 1.0039, + "step": 24550 + }, + { + "epoch": 1.81, + "learning_rate": 1.6355276467911466e-05, + "loss": 0.9652, + "step": 24551 + }, + { + "epoch": 1.81, + "learning_rate": 1.635496850280888e-05, + "loss": 1.0307, + "step": 24552 + }, + { + "epoch": 1.81, + "learning_rate": 1.6354660527595632e-05, + "loss": 0.9405, + "step": 24553 + }, + { + "epoch": 1.81, + "learning_rate": 1.6354352542272206e-05, + "loss": 0.9655, + "step": 24554 + }, + { + "epoch": 1.81, + "learning_rate": 1.6354044546839097e-05, + "loss": 1.0213, + "step": 24555 + }, + { + "epoch": 1.81, + "learning_rate": 1.6353736541296795e-05, + "loss": 1.0078, + "step": 24556 + }, + { + "epoch": 1.81, + "learning_rate": 1.6353428525645792e-05, + "loss": 0.9617, + "step": 24557 + }, + { + "epoch": 1.81, + "learning_rate": 1.6353120499886573e-05, + "loss": 1.1057, + "step": 24558 + }, + { + "epoch": 1.81, + "learning_rate": 1.635281246401963e-05, + "loss": 1.0244, + "step": 24559 + }, + { + "epoch": 1.81, + "learning_rate": 1.635250441804545e-05, + "loss": 1.0436, + "step": 24560 + }, + { + "epoch": 1.81, + "learning_rate": 1.635219636196453e-05, + "loss": 1.0282, + "step": 24561 + }, + { + "epoch": 1.81, + "learning_rate": 1.6351888295777355e-05, + "loss": 1.0083, + "step": 24562 + }, + { + "epoch": 1.81, + "learning_rate": 1.6351580219484414e-05, + "loss": 1.0922, + "step": 24563 + }, + { + "epoch": 1.82, + "learning_rate": 1.6351272133086203e-05, + "loss": 1.0543, + "step": 24564 + }, + { + "epoch": 1.82, + "learning_rate": 1.635096403658321e-05, + "loss": 0.9815, + "step": 24565 + }, + { + "epoch": 1.82, + "learning_rate": 1.6350655929975924e-05, + "loss": 1.0494, + "step": 24566 + }, + { + "epoch": 1.82, + "learning_rate": 1.6350347813264834e-05, + "loss": 1.0445, + "step": 24567 + }, + { + "epoch": 1.82, + "learning_rate": 1.6350039686450433e-05, + "loss": 0.9662, + "step": 24568 + }, + { + "epoch": 1.82, + "learning_rate": 1.634973154953321e-05, + "loss": 1.0602, + "step": 24569 + }, + { + "epoch": 1.82, + "learning_rate": 1.634942340251365e-05, + "loss": 1.0009, + "step": 24570 + }, + { + "epoch": 1.82, + "learning_rate": 1.634911524539225e-05, + "loss": 0.8601, + "step": 24571 + }, + { + "epoch": 1.82, + "learning_rate": 1.6348807078169504e-05, + "loss": 1.1332, + "step": 24572 + }, + { + "epoch": 1.82, + "learning_rate": 1.634849890084589e-05, + "loss": 1.143, + "step": 24573 + }, + { + "epoch": 1.82, + "learning_rate": 1.634819071342191e-05, + "loss": 1.0066, + "step": 24574 + }, + { + "epoch": 1.82, + "learning_rate": 1.6347882515898045e-05, + "loss": 1.0482, + "step": 24575 + }, + { + "epoch": 1.82, + "learning_rate": 1.6347574308274793e-05, + "loss": 0.9841, + "step": 24576 + }, + { + "epoch": 1.82, + "learning_rate": 1.634726609055264e-05, + "loss": 0.9845, + "step": 24577 + }, + { + "epoch": 1.82, + "learning_rate": 1.6346957862732076e-05, + "loss": 0.9725, + "step": 24578 + }, + { + "epoch": 1.82, + "learning_rate": 1.6346649624813594e-05, + "loss": 1.0612, + "step": 24579 + }, + { + "epoch": 1.82, + "learning_rate": 1.634634137679769e-05, + "loss": 1.0717, + "step": 24580 + }, + { + "epoch": 1.82, + "learning_rate": 1.634603311868484e-05, + "loss": 0.943, + "step": 24581 + }, + { + "epoch": 1.82, + "learning_rate": 1.634572485047554e-05, + "loss": 0.9394, + "step": 24582 + }, + { + "epoch": 1.82, + "learning_rate": 1.6345416572170286e-05, + "loss": 1.0486, + "step": 24583 + }, + { + "epoch": 1.82, + "learning_rate": 1.6345108283769565e-05, + "loss": 1.058, + "step": 24584 + }, + { + "epoch": 1.82, + "learning_rate": 1.6344799985273868e-05, + "loss": 0.9706, + "step": 24585 + }, + { + "epoch": 1.82, + "learning_rate": 1.6344491676683686e-05, + "loss": 1.0424, + "step": 24586 + }, + { + "epoch": 1.82, + "learning_rate": 1.6344183357999507e-05, + "loss": 0.9954, + "step": 24587 + }, + { + "epoch": 1.82, + "learning_rate": 1.6343875029221823e-05, + "loss": 0.9712, + "step": 24588 + }, + { + "epoch": 1.82, + "learning_rate": 1.634356669035112e-05, + "loss": 1.0601, + "step": 24589 + }, + { + "epoch": 1.82, + "learning_rate": 1.6343258341387903e-05, + "loss": 1.0813, + "step": 24590 + }, + { + "epoch": 1.82, + "learning_rate": 1.6342949982332647e-05, + "loss": 0.9822, + "step": 24591 + }, + { + "epoch": 1.82, + "learning_rate": 1.6342641613185847e-05, + "loss": 0.8942, + "step": 24592 + }, + { + "epoch": 1.82, + "learning_rate": 1.6342333233947998e-05, + "loss": 0.9423, + "step": 24593 + }, + { + "epoch": 1.82, + "learning_rate": 1.6342024844619587e-05, + "loss": 0.9241, + "step": 24594 + }, + { + "epoch": 1.82, + "learning_rate": 1.6341716445201107e-05, + "loss": 1.0091, + "step": 24595 + }, + { + "epoch": 1.82, + "learning_rate": 1.634140803569304e-05, + "loss": 1.0407, + "step": 24596 + }, + { + "epoch": 1.82, + "learning_rate": 1.634109961609589e-05, + "loss": 0.905, + "step": 24597 + }, + { + "epoch": 1.82, + "learning_rate": 1.634079118641014e-05, + "loss": 1.0088, + "step": 24598 + }, + { + "epoch": 1.82, + "learning_rate": 1.6340482746636278e-05, + "loss": 1.0783, + "step": 24599 + }, + { + "epoch": 1.82, + "learning_rate": 1.6340174296774804e-05, + "loss": 0.9613, + "step": 24600 + }, + { + "epoch": 1.82, + "learning_rate": 1.63398658368262e-05, + "loss": 0.9787, + "step": 24601 + }, + { + "epoch": 1.82, + "learning_rate": 1.6339557366790963e-05, + "loss": 1.0678, + "step": 24602 + }, + { + "epoch": 1.82, + "learning_rate": 1.633924888666958e-05, + "loss": 0.9746, + "step": 24603 + }, + { + "epoch": 1.82, + "learning_rate": 1.6338940396462543e-05, + "loss": 0.9431, + "step": 24604 + }, + { + "epoch": 1.82, + "learning_rate": 1.633863189617034e-05, + "loss": 1.0129, + "step": 24605 + }, + { + "epoch": 1.82, + "learning_rate": 1.6338323385793467e-05, + "loss": 1.0284, + "step": 24606 + }, + { + "epoch": 1.82, + "learning_rate": 1.633801486533241e-05, + "loss": 1.036, + "step": 24607 + }, + { + "epoch": 1.82, + "learning_rate": 1.6337706334787662e-05, + "loss": 1.0433, + "step": 24608 + }, + { + "epoch": 1.82, + "learning_rate": 1.6337397794159716e-05, + "loss": 0.9622, + "step": 24609 + }, + { + "epoch": 1.82, + "learning_rate": 1.633708924344906e-05, + "loss": 1.1127, + "step": 24610 + }, + { + "epoch": 1.82, + "learning_rate": 1.6336780682656186e-05, + "loss": 0.9919, + "step": 24611 + }, + { + "epoch": 1.82, + "learning_rate": 1.6336472111781585e-05, + "loss": 1.0184, + "step": 24612 + }, + { + "epoch": 1.82, + "learning_rate": 1.6336163530825746e-05, + "loss": 1.088, + "step": 24613 + }, + { + "epoch": 1.82, + "learning_rate": 1.6335854939789162e-05, + "loss": 1.0005, + "step": 24614 + }, + { + "epoch": 1.82, + "learning_rate": 1.6335546338672323e-05, + "loss": 1.006, + "step": 24615 + }, + { + "epoch": 1.82, + "learning_rate": 1.633523772747572e-05, + "loss": 1.1498, + "step": 24616 + }, + { + "epoch": 1.82, + "learning_rate": 1.6334929106199847e-05, + "loss": 1.0329, + "step": 24617 + }, + { + "epoch": 1.82, + "learning_rate": 1.6334620474845192e-05, + "loss": 0.9346, + "step": 24618 + }, + { + "epoch": 1.82, + "learning_rate": 1.6334311833412245e-05, + "loss": 1.0061, + "step": 24619 + }, + { + "epoch": 1.82, + "learning_rate": 1.63340031819015e-05, + "loss": 0.9914, + "step": 24620 + }, + { + "epoch": 1.82, + "learning_rate": 1.6333694520313443e-05, + "loss": 1.0606, + "step": 24621 + }, + { + "epoch": 1.82, + "learning_rate": 1.6333385848648572e-05, + "loss": 1.0135, + "step": 24622 + }, + { + "epoch": 1.82, + "learning_rate": 1.6333077166907373e-05, + "loss": 1.0143, + "step": 24623 + }, + { + "epoch": 1.82, + "learning_rate": 1.633276847509034e-05, + "loss": 1.0062, + "step": 24624 + }, + { + "epoch": 1.82, + "learning_rate": 1.6332459773197964e-05, + "loss": 1.0004, + "step": 24625 + }, + { + "epoch": 1.82, + "learning_rate": 1.6332151061230733e-05, + "loss": 1.0052, + "step": 24626 + }, + { + "epoch": 1.82, + "learning_rate": 1.633184233918914e-05, + "loss": 1.0752, + "step": 24627 + }, + { + "epoch": 1.82, + "learning_rate": 1.6331533607073678e-05, + "loss": 1.0557, + "step": 24628 + }, + { + "epoch": 1.82, + "learning_rate": 1.6331224864884837e-05, + "loss": 1.1113, + "step": 24629 + }, + { + "epoch": 1.82, + "learning_rate": 1.6330916112623107e-05, + "loss": 1.0204, + "step": 24630 + }, + { + "epoch": 1.82, + "learning_rate": 1.633060735028898e-05, + "loss": 1.0386, + "step": 24631 + }, + { + "epoch": 1.82, + "learning_rate": 1.6330298577882946e-05, + "loss": 1.0999, + "step": 24632 + }, + { + "epoch": 1.82, + "learning_rate": 1.6329989795405497e-05, + "loss": 0.9417, + "step": 24633 + }, + { + "epoch": 1.82, + "learning_rate": 1.632968100285713e-05, + "loss": 1.0549, + "step": 24634 + }, + { + "epoch": 1.82, + "learning_rate": 1.6329372200238328e-05, + "loss": 1.1314, + "step": 24635 + }, + { + "epoch": 1.82, + "learning_rate": 1.6329063387549582e-05, + "loss": 0.9676, + "step": 24636 + }, + { + "epoch": 1.82, + "learning_rate": 1.632875456479139e-05, + "loss": 0.9345, + "step": 24637 + }, + { + "epoch": 1.82, + "learning_rate": 1.632844573196424e-05, + "loss": 1.1285, + "step": 24638 + }, + { + "epoch": 1.82, + "learning_rate": 1.6328136889068623e-05, + "loss": 1.0814, + "step": 24639 + }, + { + "epoch": 1.82, + "learning_rate": 1.632782803610503e-05, + "loss": 0.9948, + "step": 24640 + }, + { + "epoch": 1.82, + "learning_rate": 1.6327519173073955e-05, + "loss": 0.991, + "step": 24641 + }, + { + "epoch": 1.82, + "learning_rate": 1.6327210299975887e-05, + "loss": 1.0466, + "step": 24642 + }, + { + "epoch": 1.82, + "learning_rate": 1.6326901416811317e-05, + "loss": 1.0071, + "step": 24643 + }, + { + "epoch": 1.82, + "learning_rate": 1.6326592523580737e-05, + "loss": 1.0667, + "step": 24644 + }, + { + "epoch": 1.82, + "learning_rate": 1.632628362028464e-05, + "loss": 1.0197, + "step": 24645 + }, + { + "epoch": 1.82, + "learning_rate": 1.6325974706923515e-05, + "loss": 0.914, + "step": 24646 + }, + { + "epoch": 1.82, + "learning_rate": 1.6325665783497856e-05, + "loss": 0.9562, + "step": 24647 + }, + { + "epoch": 1.82, + "learning_rate": 1.632535685000815e-05, + "loss": 0.9955, + "step": 24648 + }, + { + "epoch": 1.82, + "learning_rate": 1.6325047906454896e-05, + "loss": 1.0117, + "step": 24649 + }, + { + "epoch": 1.82, + "learning_rate": 1.632473895283858e-05, + "loss": 1.0239, + "step": 24650 + }, + { + "epoch": 1.82, + "learning_rate": 1.6324429989159697e-05, + "loss": 1.0504, + "step": 24651 + }, + { + "epoch": 1.82, + "learning_rate": 1.6324121015418733e-05, + "loss": 1.0159, + "step": 24652 + }, + { + "epoch": 1.82, + "learning_rate": 1.6323812031616183e-05, + "loss": 0.9028, + "step": 24653 + }, + { + "epoch": 1.82, + "learning_rate": 1.6323503037752542e-05, + "loss": 0.9496, + "step": 24654 + }, + { + "epoch": 1.82, + "learning_rate": 1.6323194033828297e-05, + "loss": 1.0869, + "step": 24655 + }, + { + "epoch": 1.82, + "learning_rate": 1.6322885019843937e-05, + "loss": 0.9375, + "step": 24656 + }, + { + "epoch": 1.82, + "learning_rate": 1.632257599579996e-05, + "loss": 0.9451, + "step": 24657 + }, + { + "epoch": 1.82, + "learning_rate": 1.6322266961696856e-05, + "loss": 1.1416, + "step": 24658 + }, + { + "epoch": 1.82, + "learning_rate": 1.6321957917535114e-05, + "loss": 0.9982, + "step": 24659 + }, + { + "epoch": 1.82, + "learning_rate": 1.632164886331523e-05, + "loss": 1.0703, + "step": 24660 + }, + { + "epoch": 1.82, + "learning_rate": 1.632133979903769e-05, + "loss": 1.1683, + "step": 24661 + }, + { + "epoch": 1.82, + "learning_rate": 1.632103072470299e-05, + "loss": 1.0497, + "step": 24662 + }, + { + "epoch": 1.82, + "learning_rate": 1.632072164031162e-05, + "loss": 1.0398, + "step": 24663 + }, + { + "epoch": 1.82, + "learning_rate": 1.6320412545864074e-05, + "loss": 1.0654, + "step": 24664 + }, + { + "epoch": 1.82, + "learning_rate": 1.632010344136084e-05, + "loss": 0.9768, + "step": 24665 + }, + { + "epoch": 1.82, + "learning_rate": 1.6319794326802414e-05, + "loss": 0.9568, + "step": 24666 + }, + { + "epoch": 1.82, + "learning_rate": 1.6319485202189286e-05, + "loss": 0.9283, + "step": 24667 + }, + { + "epoch": 1.82, + "learning_rate": 1.6319176067521942e-05, + "loss": 1.0445, + "step": 24668 + }, + { + "epoch": 1.82, + "learning_rate": 1.6318866922800886e-05, + "loss": 0.9993, + "step": 24669 + }, + { + "epoch": 1.82, + "learning_rate": 1.63185577680266e-05, + "loss": 0.9681, + "step": 24670 + }, + { + "epoch": 1.82, + "learning_rate": 1.631824860319958e-05, + "loss": 1.0667, + "step": 24671 + }, + { + "epoch": 1.82, + "learning_rate": 1.6317939428320314e-05, + "loss": 1.0677, + "step": 24672 + }, + { + "epoch": 1.82, + "learning_rate": 1.63176302433893e-05, + "loss": 1.0368, + "step": 24673 + }, + { + "epoch": 1.82, + "learning_rate": 1.6317321048407023e-05, + "loss": 1.0454, + "step": 24674 + }, + { + "epoch": 1.82, + "learning_rate": 1.6317011843373985e-05, + "loss": 1.0687, + "step": 24675 + }, + { + "epoch": 1.82, + "learning_rate": 1.6316702628290668e-05, + "loss": 1.0742, + "step": 24676 + }, + { + "epoch": 1.82, + "learning_rate": 1.6316393403157565e-05, + "loss": 0.9908, + "step": 24677 + }, + { + "epoch": 1.82, + "learning_rate": 1.6316084167975172e-05, + "loss": 1.089, + "step": 24678 + }, + { + "epoch": 1.82, + "learning_rate": 1.631577492274398e-05, + "loss": 0.9421, + "step": 24679 + }, + { + "epoch": 1.82, + "learning_rate": 1.6315465667464484e-05, + "loss": 1.0546, + "step": 24680 + }, + { + "epoch": 1.82, + "learning_rate": 1.6315156402137168e-05, + "loss": 1.0341, + "step": 24681 + }, + { + "epoch": 1.82, + "learning_rate": 1.6314847126762526e-05, + "loss": 1.0092, + "step": 24682 + }, + { + "epoch": 1.82, + "learning_rate": 1.6314537841341058e-05, + "loss": 0.9923, + "step": 24683 + }, + { + "epoch": 1.82, + "learning_rate": 1.6314228545873246e-05, + "loss": 0.9314, + "step": 24684 + }, + { + "epoch": 1.82, + "learning_rate": 1.6313919240359594e-05, + "loss": 0.9539, + "step": 24685 + }, + { + "epoch": 1.82, + "learning_rate": 1.631360992480058e-05, + "loss": 1.0965, + "step": 24686 + }, + { + "epoch": 1.82, + "learning_rate": 1.6313300599196706e-05, + "loss": 1.0016, + "step": 24687 + }, + { + "epoch": 1.82, + "learning_rate": 1.631299126354846e-05, + "loss": 0.987, + "step": 24688 + }, + { + "epoch": 1.82, + "learning_rate": 1.6312681917856335e-05, + "loss": 0.9129, + "step": 24689 + }, + { + "epoch": 1.82, + "learning_rate": 1.6312372562120825e-05, + "loss": 1.0134, + "step": 24690 + }, + { + "epoch": 1.82, + "learning_rate": 1.631206319634242e-05, + "loss": 0.93, + "step": 24691 + }, + { + "epoch": 1.82, + "learning_rate": 1.6311753820521616e-05, + "loss": 1.0146, + "step": 24692 + }, + { + "epoch": 1.82, + "learning_rate": 1.6311444434658895e-05, + "loss": 1.0268, + "step": 24693 + }, + { + "epoch": 1.82, + "learning_rate": 1.631113503875476e-05, + "loss": 0.8951, + "step": 24694 + }, + { + "epoch": 1.82, + "learning_rate": 1.6310825632809697e-05, + "loss": 0.996, + "step": 24695 + }, + { + "epoch": 1.82, + "learning_rate": 1.6310516216824205e-05, + "loss": 1.0583, + "step": 24696 + }, + { + "epoch": 1.82, + "learning_rate": 1.6310206790798767e-05, + "loss": 0.9678, + "step": 24697 + }, + { + "epoch": 1.82, + "learning_rate": 1.6309897354733884e-05, + "loss": 0.9656, + "step": 24698 + }, + { + "epoch": 1.82, + "learning_rate": 1.6309587908630045e-05, + "loss": 1.0631, + "step": 24699 + }, + { + "epoch": 1.83, + "learning_rate": 1.630927845248774e-05, + "loss": 1.0939, + "step": 24700 + }, + { + "epoch": 1.83, + "learning_rate": 1.6308968986307463e-05, + "loss": 1.1257, + "step": 24701 + }, + { + "epoch": 1.83, + "learning_rate": 1.630865951008971e-05, + "loss": 1.0475, + "step": 24702 + }, + { + "epoch": 1.83, + "learning_rate": 1.6308350023834966e-05, + "loss": 0.9656, + "step": 24703 + }, + { + "epoch": 1.83, + "learning_rate": 1.630804052754373e-05, + "loss": 0.9165, + "step": 24704 + }, + { + "epoch": 1.83, + "learning_rate": 1.630773102121649e-05, + "loss": 0.9733, + "step": 24705 + }, + { + "epoch": 1.83, + "learning_rate": 1.6307421504853737e-05, + "loss": 0.8734, + "step": 24706 + }, + { + "epoch": 1.83, + "learning_rate": 1.6307111978455972e-05, + "loss": 1.0542, + "step": 24707 + }, + { + "epoch": 1.83, + "learning_rate": 1.6306802442023683e-05, + "loss": 1.0159, + "step": 24708 + }, + { + "epoch": 1.83, + "learning_rate": 1.6306492895557356e-05, + "loss": 0.9933, + "step": 24709 + }, + { + "epoch": 1.83, + "learning_rate": 1.6306183339057494e-05, + "loss": 1.0095, + "step": 24710 + }, + { + "epoch": 1.83, + "learning_rate": 1.630587377252458e-05, + "loss": 1.0478, + "step": 24711 + }, + { + "epoch": 1.83, + "learning_rate": 1.6305564195959114e-05, + "loss": 1.0593, + "step": 24712 + }, + { + "epoch": 1.83, + "learning_rate": 1.6305254609361586e-05, + "loss": 1.0572, + "step": 24713 + }, + { + "epoch": 1.83, + "learning_rate": 1.6304945012732488e-05, + "loss": 0.9771, + "step": 24714 + }, + { + "epoch": 1.83, + "learning_rate": 1.630463540607231e-05, + "loss": 0.9293, + "step": 24715 + }, + { + "epoch": 1.83, + "learning_rate": 1.630432578938155e-05, + "loss": 1.0448, + "step": 24716 + }, + { + "epoch": 1.83, + "learning_rate": 1.63040161626607e-05, + "loss": 0.9662, + "step": 24717 + }, + { + "epoch": 1.83, + "learning_rate": 1.6303706525910245e-05, + "loss": 0.9536, + "step": 24718 + }, + { + "epoch": 1.83, + "learning_rate": 1.6303396879130688e-05, + "loss": 0.9862, + "step": 24719 + }, + { + "epoch": 1.83, + "learning_rate": 1.6303087222322514e-05, + "loss": 1.0455, + "step": 24720 + }, + { + "epoch": 1.83, + "learning_rate": 1.630277755548622e-05, + "loss": 1.0691, + "step": 24721 + }, + { + "epoch": 1.83, + "learning_rate": 1.6302467878622294e-05, + "loss": 1.0456, + "step": 24722 + }, + { + "epoch": 1.83, + "learning_rate": 1.6302158191731233e-05, + "loss": 0.9483, + "step": 24723 + }, + { + "epoch": 1.83, + "learning_rate": 1.630184849481353e-05, + "loss": 0.9505, + "step": 24724 + }, + { + "epoch": 1.83, + "learning_rate": 1.630153878786967e-05, + "loss": 1.0808, + "step": 24725 + }, + { + "epoch": 1.83, + "learning_rate": 1.6301229070900163e-05, + "loss": 0.9328, + "step": 24726 + }, + { + "epoch": 1.83, + "learning_rate": 1.6300919343905484e-05, + "loss": 0.9966, + "step": 24727 + }, + { + "epoch": 1.83, + "learning_rate": 1.6300609606886136e-05, + "loss": 0.9638, + "step": 24728 + }, + { + "epoch": 1.83, + "learning_rate": 1.6300299859842604e-05, + "loss": 1.0046, + "step": 24729 + }, + { + "epoch": 1.83, + "learning_rate": 1.6299990102775385e-05, + "loss": 0.9674, + "step": 24730 + }, + { + "epoch": 1.83, + "learning_rate": 1.6299680335684976e-05, + "loss": 1.0022, + "step": 24731 + }, + { + "epoch": 1.83, + "learning_rate": 1.6299370558571862e-05, + "loss": 1.0364, + "step": 24732 + }, + { + "epoch": 1.83, + "learning_rate": 1.6299060771436543e-05, + "loss": 0.893, + "step": 24733 + }, + { + "epoch": 1.83, + "learning_rate": 1.6298750974279506e-05, + "loss": 0.9787, + "step": 24734 + }, + { + "epoch": 1.83, + "learning_rate": 1.6298441167101245e-05, + "loss": 1.1325, + "step": 24735 + }, + { + "epoch": 1.83, + "learning_rate": 1.6298131349902257e-05, + "loss": 1.0077, + "step": 24736 + }, + { + "epoch": 1.83, + "learning_rate": 1.629782152268303e-05, + "loss": 1.0535, + "step": 24737 + }, + { + "epoch": 1.83, + "learning_rate": 1.6297511685444063e-05, + "loss": 0.9904, + "step": 24738 + }, + { + "epoch": 1.83, + "learning_rate": 1.629720183818584e-05, + "loss": 1.0192, + "step": 24739 + }, + { + "epoch": 1.83, + "learning_rate": 1.629689198090886e-05, + "loss": 1.0546, + "step": 24740 + }, + { + "epoch": 1.83, + "learning_rate": 1.629658211361362e-05, + "loss": 0.9493, + "step": 24741 + }, + { + "epoch": 1.83, + "learning_rate": 1.62962722363006e-05, + "loss": 1.0216, + "step": 24742 + }, + { + "epoch": 1.83, + "learning_rate": 1.6295962348970307e-05, + "loss": 1.0788, + "step": 24743 + }, + { + "epoch": 1.83, + "learning_rate": 1.6295652451623228e-05, + "loss": 0.9578, + "step": 24744 + }, + { + "epoch": 1.83, + "learning_rate": 1.6295342544259853e-05, + "loss": 0.9242, + "step": 24745 + }, + { + "epoch": 1.83, + "learning_rate": 1.629503262688068e-05, + "loss": 1.0171, + "step": 24746 + }, + { + "epoch": 1.83, + "learning_rate": 1.62947226994862e-05, + "loss": 1.0314, + "step": 24747 + }, + { + "epoch": 1.83, + "learning_rate": 1.6294412762076902e-05, + "loss": 0.9717, + "step": 24748 + }, + { + "epoch": 1.83, + "learning_rate": 1.629410281465329e-05, + "loss": 1.0116, + "step": 24749 + }, + { + "epoch": 1.83, + "learning_rate": 1.6293792857215843e-05, + "loss": 1.0604, + "step": 24750 + }, + { + "epoch": 1.83, + "learning_rate": 1.6293482889765068e-05, + "loss": 0.8916, + "step": 24751 + }, + { + "epoch": 1.83, + "learning_rate": 1.629317291230145e-05, + "loss": 0.9191, + "step": 24752 + }, + { + "epoch": 1.83, + "learning_rate": 1.6292862924825484e-05, + "loss": 1.0278, + "step": 24753 + }, + { + "epoch": 1.83, + "learning_rate": 1.6292552927337663e-05, + "loss": 1.0414, + "step": 24754 + }, + { + "epoch": 1.83, + "learning_rate": 1.629224291983848e-05, + "loss": 1.0885, + "step": 24755 + }, + { + "epoch": 1.83, + "learning_rate": 1.629193290232843e-05, + "loss": 1.0944, + "step": 24756 + }, + { + "epoch": 1.83, + "learning_rate": 1.6291622874808003e-05, + "loss": 0.9718, + "step": 24757 + }, + { + "epoch": 1.83, + "learning_rate": 1.6291312837277696e-05, + "loss": 0.967, + "step": 24758 + }, + { + "epoch": 1.83, + "learning_rate": 1.6291002789737995e-05, + "loss": 1.0459, + "step": 24759 + }, + { + "epoch": 1.83, + "learning_rate": 1.6290692732189403e-05, + "loss": 1.1053, + "step": 24760 + }, + { + "epoch": 1.83, + "learning_rate": 1.6290382664632408e-05, + "loss": 0.9512, + "step": 24761 + }, + { + "epoch": 1.83, + "learning_rate": 1.6290072587067506e-05, + "loss": 1.1286, + "step": 24762 + }, + { + "epoch": 1.83, + "learning_rate": 1.6289762499495187e-05, + "loss": 1.0693, + "step": 24763 + }, + { + "epoch": 1.83, + "learning_rate": 1.6289452401915947e-05, + "loss": 1.0171, + "step": 24764 + }, + { + "epoch": 1.83, + "learning_rate": 1.6289142294330276e-05, + "loss": 0.9341, + "step": 24765 + }, + { + "epoch": 1.83, + "learning_rate": 1.6288832176738672e-05, + "loss": 0.8855, + "step": 24766 + }, + { + "epoch": 1.83, + "learning_rate": 1.6288522049141625e-05, + "loss": 0.921, + "step": 24767 + }, + { + "epoch": 1.83, + "learning_rate": 1.6288211911539627e-05, + "loss": 0.9423, + "step": 24768 + }, + { + "epoch": 1.83, + "learning_rate": 1.6287901763933174e-05, + "loss": 0.889, + "step": 24769 + }, + { + "epoch": 1.83, + "learning_rate": 1.628759160632276e-05, + "loss": 0.977, + "step": 24770 + }, + { + "epoch": 1.83, + "learning_rate": 1.628728143870888e-05, + "loss": 1.0906, + "step": 24771 + }, + { + "epoch": 1.83, + "learning_rate": 1.6286971261092027e-05, + "loss": 0.9931, + "step": 24772 + }, + { + "epoch": 1.83, + "learning_rate": 1.628666107347269e-05, + "loss": 0.9776, + "step": 24773 + }, + { + "epoch": 1.83, + "learning_rate": 1.6286350875851363e-05, + "loss": 1.0137, + "step": 24774 + }, + { + "epoch": 1.83, + "learning_rate": 1.6286040668228543e-05, + "loss": 1.0541, + "step": 24775 + }, + { + "epoch": 1.83, + "learning_rate": 1.628573045060472e-05, + "loss": 1.0016, + "step": 24776 + }, + { + "epoch": 1.83, + "learning_rate": 1.6285420222980396e-05, + "loss": 1.0299, + "step": 24777 + }, + { + "epoch": 1.83, + "learning_rate": 1.6285109985356052e-05, + "loss": 1.0663, + "step": 24778 + }, + { + "epoch": 1.83, + "learning_rate": 1.628479973773219e-05, + "loss": 0.9986, + "step": 24779 + }, + { + "epoch": 1.83, + "learning_rate": 1.6284489480109306e-05, + "loss": 0.9016, + "step": 24780 + }, + { + "epoch": 1.83, + "learning_rate": 1.6284179212487883e-05, + "loss": 0.9749, + "step": 24781 + }, + { + "epoch": 1.83, + "learning_rate": 1.6283868934868425e-05, + "loss": 1.0769, + "step": 24782 + }, + { + "epoch": 1.83, + "learning_rate": 1.6283558647251417e-05, + "loss": 1.0566, + "step": 24783 + }, + { + "epoch": 1.83, + "learning_rate": 1.628324834963736e-05, + "loss": 0.917, + "step": 24784 + }, + { + "epoch": 1.83, + "learning_rate": 1.6282938042026746e-05, + "loss": 0.9391, + "step": 24785 + }, + { + "epoch": 1.83, + "learning_rate": 1.628262772442007e-05, + "loss": 0.9263, + "step": 24786 + }, + { + "epoch": 1.83, + "learning_rate": 1.6282317396817817e-05, + "loss": 0.9937, + "step": 24787 + }, + { + "epoch": 1.83, + "learning_rate": 1.6282007059220488e-05, + "loss": 0.946, + "step": 24788 + }, + { + "epoch": 1.83, + "learning_rate": 1.6281696711628577e-05, + "loss": 1.0296, + "step": 24789 + }, + { + "epoch": 1.83, + "learning_rate": 1.6281386354042575e-05, + "loss": 1.0883, + "step": 24790 + }, + { + "epoch": 1.83, + "learning_rate": 1.628107598646298e-05, + "loss": 1.0103, + "step": 24791 + }, + { + "epoch": 1.83, + "learning_rate": 1.6280765608890284e-05, + "loss": 1.0065, + "step": 24792 + }, + { + "epoch": 1.83, + "learning_rate": 1.6280455221324977e-05, + "loss": 0.8919, + "step": 24793 + }, + { + "epoch": 1.83, + "learning_rate": 1.6280144823767555e-05, + "loss": 1.1657, + "step": 24794 + }, + { + "epoch": 1.83, + "learning_rate": 1.627983441621851e-05, + "loss": 0.9225, + "step": 24795 + }, + { + "epoch": 1.83, + "learning_rate": 1.6279523998678342e-05, + "loss": 1.0459, + "step": 24796 + }, + { + "epoch": 1.83, + "learning_rate": 1.627921357114754e-05, + "loss": 1.0325, + "step": 24797 + }, + { + "epoch": 1.83, + "learning_rate": 1.62789031336266e-05, + "loss": 1.0993, + "step": 24798 + }, + { + "epoch": 1.83, + "learning_rate": 1.6278592686116015e-05, + "loss": 1.0488, + "step": 24799 + }, + { + "epoch": 1.83, + "learning_rate": 1.627828222861628e-05, + "loss": 0.9249, + "step": 24800 + }, + { + "epoch": 1.83, + "learning_rate": 1.6277971761127885e-05, + "loss": 1.0094, + "step": 24801 + }, + { + "epoch": 1.83, + "learning_rate": 1.627766128365133e-05, + "loss": 0.9839, + "step": 24802 + }, + { + "epoch": 1.83, + "learning_rate": 1.6277350796187102e-05, + "loss": 1.0105, + "step": 24803 + }, + { + "epoch": 1.83, + "learning_rate": 1.62770402987357e-05, + "loss": 0.9333, + "step": 24804 + }, + { + "epoch": 1.83, + "learning_rate": 1.627672979129762e-05, + "loss": 1.0447, + "step": 24805 + }, + { + "epoch": 1.83, + "learning_rate": 1.6276419273873346e-05, + "loss": 1.0079, + "step": 24806 + }, + { + "epoch": 1.83, + "learning_rate": 1.6276108746463384e-05, + "loss": 1.1276, + "step": 24807 + }, + { + "epoch": 1.83, + "learning_rate": 1.627579820906822e-05, + "loss": 0.9903, + "step": 24808 + }, + { + "epoch": 1.83, + "learning_rate": 1.627548766168835e-05, + "loss": 1.0122, + "step": 24809 + }, + { + "epoch": 1.83, + "learning_rate": 1.6275177104324274e-05, + "loss": 1.1772, + "step": 24810 + }, + { + "epoch": 1.83, + "learning_rate": 1.627486653697648e-05, + "loss": 0.9881, + "step": 24811 + }, + { + "epoch": 1.83, + "learning_rate": 1.627455595964546e-05, + "loss": 0.9575, + "step": 24812 + }, + { + "epoch": 1.83, + "learning_rate": 1.6274245372331713e-05, + "loss": 0.955, + "step": 24813 + }, + { + "epoch": 1.83, + "learning_rate": 1.6273934775035733e-05, + "loss": 0.9895, + "step": 24814 + }, + { + "epoch": 1.83, + "learning_rate": 1.6273624167758008e-05, + "loss": 0.9822, + "step": 24815 + }, + { + "epoch": 1.83, + "learning_rate": 1.627331355049904e-05, + "loss": 1.0037, + "step": 24816 + }, + { + "epoch": 1.83, + "learning_rate": 1.6273002923259316e-05, + "loss": 0.9134, + "step": 24817 + }, + { + "epoch": 1.83, + "learning_rate": 1.627269228603934e-05, + "loss": 1.0164, + "step": 24818 + }, + { + "epoch": 1.83, + "learning_rate": 1.6272381638839596e-05, + "loss": 1.0074, + "step": 24819 + }, + { + "epoch": 1.83, + "learning_rate": 1.6272070981660585e-05, + "loss": 0.86, + "step": 24820 + }, + { + "epoch": 1.83, + "learning_rate": 1.6271760314502798e-05, + "loss": 1.0537, + "step": 24821 + }, + { + "epoch": 1.83, + "learning_rate": 1.6271449637366728e-05, + "loss": 1.0169, + "step": 24822 + }, + { + "epoch": 1.83, + "learning_rate": 1.6271138950252875e-05, + "loss": 0.9944, + "step": 24823 + }, + { + "epoch": 1.83, + "learning_rate": 1.627082825316173e-05, + "loss": 0.9509, + "step": 24824 + }, + { + "epoch": 1.83, + "learning_rate": 1.6270517546093782e-05, + "loss": 1.008, + "step": 24825 + }, + { + "epoch": 1.83, + "learning_rate": 1.6270206829049535e-05, + "loss": 1.0007, + "step": 24826 + }, + { + "epoch": 1.83, + "learning_rate": 1.6269896102029473e-05, + "loss": 1.0003, + "step": 24827 + }, + { + "epoch": 1.83, + "learning_rate": 1.6269585365034103e-05, + "loss": 0.9307, + "step": 24828 + }, + { + "epoch": 1.83, + "learning_rate": 1.6269274618063908e-05, + "loss": 1.0632, + "step": 24829 + }, + { + "epoch": 1.83, + "learning_rate": 1.6268963861119388e-05, + "loss": 1.0341, + "step": 24830 + }, + { + "epoch": 1.83, + "learning_rate": 1.6268653094201038e-05, + "loss": 0.9797, + "step": 24831 + }, + { + "epoch": 1.83, + "learning_rate": 1.6268342317309346e-05, + "loss": 0.9329, + "step": 24832 + }, + { + "epoch": 1.83, + "learning_rate": 1.6268031530444814e-05, + "loss": 0.9952, + "step": 24833 + }, + { + "epoch": 1.83, + "learning_rate": 1.6267720733607936e-05, + "loss": 0.9402, + "step": 24834 + }, + { + "epoch": 1.84, + "learning_rate": 1.62674099267992e-05, + "loss": 0.9549, + "step": 24835 + }, + { + "epoch": 1.84, + "learning_rate": 1.6267099110019103e-05, + "loss": 0.9504, + "step": 24836 + }, + { + "epoch": 1.84, + "learning_rate": 1.626678828326815e-05, + "loss": 0.9855, + "step": 24837 + }, + { + "epoch": 1.84, + "learning_rate": 1.6266477446546816e-05, + "loss": 1.0707, + "step": 24838 + }, + { + "epoch": 1.84, + "learning_rate": 1.626616659985561e-05, + "loss": 1.0328, + "step": 24839 + }, + { + "epoch": 1.84, + "learning_rate": 1.6265855743195026e-05, + "loss": 1.0027, + "step": 24840 + }, + { + "epoch": 1.84, + "learning_rate": 1.626554487656555e-05, + "loss": 0.9559, + "step": 24841 + }, + { + "epoch": 1.84, + "learning_rate": 1.6265233999967685e-05, + "loss": 0.9305, + "step": 24842 + }, + { + "epoch": 1.84, + "learning_rate": 1.6264923113401923e-05, + "loss": 1.0661, + "step": 24843 + }, + { + "epoch": 1.84, + "learning_rate": 1.6264612216868752e-05, + "loss": 1.0517, + "step": 24844 + }, + { + "epoch": 1.84, + "learning_rate": 1.6264301310368676e-05, + "loss": 1.1346, + "step": 24845 + }, + { + "epoch": 1.84, + "learning_rate": 1.626399039390219e-05, + "loss": 1.0622, + "step": 24846 + }, + { + "epoch": 1.84, + "learning_rate": 1.6263679467469783e-05, + "loss": 0.9744, + "step": 24847 + }, + { + "epoch": 1.84, + "learning_rate": 1.626336853107195e-05, + "loss": 1.0126, + "step": 24848 + }, + { + "epoch": 1.84, + "learning_rate": 1.6263057584709187e-05, + "loss": 1.04, + "step": 24849 + }, + { + "epoch": 1.84, + "learning_rate": 1.6262746628381992e-05, + "loss": 1.0086, + "step": 24850 + }, + { + "epoch": 1.84, + "learning_rate": 1.6262435662090853e-05, + "loss": 0.9647, + "step": 24851 + }, + { + "epoch": 1.84, + "learning_rate": 1.626212468583627e-05, + "loss": 1.0536, + "step": 24852 + }, + { + "epoch": 1.84, + "learning_rate": 1.6261813699618734e-05, + "loss": 1.0822, + "step": 24853 + }, + { + "epoch": 1.84, + "learning_rate": 1.6261502703438744e-05, + "loss": 0.9895, + "step": 24854 + }, + { + "epoch": 1.84, + "learning_rate": 1.6261191697296795e-05, + "loss": 1.0307, + "step": 24855 + }, + { + "epoch": 1.84, + "learning_rate": 1.6260880681193377e-05, + "loss": 0.9275, + "step": 24856 + }, + { + "epoch": 1.84, + "learning_rate": 1.6260569655128987e-05, + "loss": 0.9767, + "step": 24857 + }, + { + "epoch": 1.84, + "learning_rate": 1.6260258619104122e-05, + "loss": 0.9618, + "step": 24858 + }, + { + "epoch": 1.84, + "learning_rate": 1.6259947573119275e-05, + "loss": 0.9913, + "step": 24859 + }, + { + "epoch": 1.84, + "learning_rate": 1.625963651717494e-05, + "loss": 1.0936, + "step": 24860 + }, + { + "epoch": 1.84, + "learning_rate": 1.6259325451271616e-05, + "loss": 1.0772, + "step": 24861 + }, + { + "epoch": 1.84, + "learning_rate": 1.625901437540979e-05, + "loss": 1.021, + "step": 24862 + }, + { + "epoch": 1.84, + "learning_rate": 1.6258703289589962e-05, + "loss": 0.9998, + "step": 24863 + }, + { + "epoch": 1.84, + "learning_rate": 1.6258392193812632e-05, + "loss": 0.8789, + "step": 24864 + }, + { + "epoch": 1.84, + "learning_rate": 1.6258081088078285e-05, + "loss": 1.103, + "step": 24865 + }, + { + "epoch": 1.84, + "learning_rate": 1.625776997238742e-05, + "loss": 1.1095, + "step": 24866 + }, + { + "epoch": 1.84, + "learning_rate": 1.6257458846740534e-05, + "loss": 0.814, + "step": 24867 + }, + { + "epoch": 1.84, + "learning_rate": 1.6257147711138118e-05, + "loss": 1.034, + "step": 24868 + }, + { + "epoch": 1.84, + "learning_rate": 1.6256836565580675e-05, + "loss": 1.0606, + "step": 24869 + }, + { + "epoch": 1.84, + "learning_rate": 1.625652541006869e-05, + "loss": 1.0554, + "step": 24870 + }, + { + "epoch": 1.84, + "learning_rate": 1.6256214244602665e-05, + "loss": 1.0556, + "step": 24871 + }, + { + "epoch": 1.84, + "learning_rate": 1.625590306918309e-05, + "loss": 1.0294, + "step": 24872 + }, + { + "epoch": 1.84, + "learning_rate": 1.6255591883810463e-05, + "loss": 1.0039, + "step": 24873 + }, + { + "epoch": 1.84, + "learning_rate": 1.625528068848528e-05, + "loss": 0.8631, + "step": 24874 + }, + { + "epoch": 1.84, + "learning_rate": 1.625496948320804e-05, + "loss": 1.0416, + "step": 24875 + }, + { + "epoch": 1.84, + "learning_rate": 1.6254658267979226e-05, + "loss": 1.0362, + "step": 24876 + }, + { + "epoch": 1.84, + "learning_rate": 1.6254347042799345e-05, + "loss": 0.8699, + "step": 24877 + }, + { + "epoch": 1.84, + "learning_rate": 1.6254035807668883e-05, + "loss": 0.9358, + "step": 24878 + }, + { + "epoch": 1.84, + "learning_rate": 1.625372456258834e-05, + "loss": 0.9887, + "step": 24879 + }, + { + "epoch": 1.84, + "learning_rate": 1.6253413307558214e-05, + "loss": 0.8809, + "step": 24880 + }, + { + "epoch": 1.84, + "learning_rate": 1.6253102042578992e-05, + "loss": 1.0273, + "step": 24881 + }, + { + "epoch": 1.84, + "learning_rate": 1.6252790767651175e-05, + "loss": 0.9992, + "step": 24882 + }, + { + "epoch": 1.84, + "learning_rate": 1.625247948277526e-05, + "loss": 1.0312, + "step": 24883 + }, + { + "epoch": 1.84, + "learning_rate": 1.625216818795174e-05, + "loss": 1.1244, + "step": 24884 + }, + { + "epoch": 1.84, + "learning_rate": 1.625185688318111e-05, + "loss": 0.9597, + "step": 24885 + }, + { + "epoch": 1.84, + "learning_rate": 1.6251545568463863e-05, + "loss": 1.1245, + "step": 24886 + }, + { + "epoch": 1.84, + "learning_rate": 1.6251234243800498e-05, + "loss": 1.029, + "step": 24887 + }, + { + "epoch": 1.84, + "learning_rate": 1.6250922909191505e-05, + "loss": 1.1115, + "step": 24888 + }, + { + "epoch": 1.84, + "learning_rate": 1.6250611564637387e-05, + "loss": 0.9471, + "step": 24889 + }, + { + "epoch": 1.84, + "learning_rate": 1.6250300210138638e-05, + "loss": 0.9846, + "step": 24890 + }, + { + "epoch": 1.84, + "learning_rate": 1.6249988845695746e-05, + "loss": 1.0499, + "step": 24891 + }, + { + "epoch": 1.84, + "learning_rate": 1.624967747130921e-05, + "loss": 0.9329, + "step": 24892 + }, + { + "epoch": 1.84, + "learning_rate": 1.624936608697953e-05, + "loss": 1.1152, + "step": 24893 + }, + { + "epoch": 1.84, + "learning_rate": 1.6249054692707198e-05, + "loss": 1.0492, + "step": 24894 + }, + { + "epoch": 1.84, + "learning_rate": 1.624874328849271e-05, + "loss": 1.0978, + "step": 24895 + }, + { + "epoch": 1.84, + "learning_rate": 1.624843187433656e-05, + "loss": 1.0217, + "step": 24896 + }, + { + "epoch": 1.84, + "learning_rate": 1.624812045023924e-05, + "loss": 1.1762, + "step": 24897 + }, + { + "epoch": 1.84, + "learning_rate": 1.624780901620126e-05, + "loss": 1.0471, + "step": 24898 + }, + { + "epoch": 1.84, + "learning_rate": 1.6247497572223094e-05, + "loss": 0.9485, + "step": 24899 + }, + { + "epoch": 1.84, + "learning_rate": 1.6247186118305255e-05, + "loss": 1.0881, + "step": 24900 + }, + { + "epoch": 1.84, + "learning_rate": 1.624687465444823e-05, + "loss": 1.0602, + "step": 24901 + }, + { + "epoch": 1.84, + "learning_rate": 1.6246563180652514e-05, + "loss": 0.9696, + "step": 24902 + }, + { + "epoch": 1.84, + "learning_rate": 1.624625169691861e-05, + "loss": 0.9825, + "step": 24903 + }, + { + "epoch": 1.84, + "learning_rate": 1.6245940203247012e-05, + "loss": 1.1196, + "step": 24904 + }, + { + "epoch": 1.84, + "learning_rate": 1.624562869963821e-05, + "loss": 1.0858, + "step": 24905 + }, + { + "epoch": 1.84, + "learning_rate": 1.6245317186092696e-05, + "loss": 0.9971, + "step": 24906 + }, + { + "epoch": 1.84, + "learning_rate": 1.6245005662610976e-05, + "loss": 1.039, + "step": 24907 + }, + { + "epoch": 1.84, + "learning_rate": 1.6244694129193542e-05, + "loss": 1.0841, + "step": 24908 + }, + { + "epoch": 1.84, + "learning_rate": 1.624438258584089e-05, + "loss": 0.9782, + "step": 24909 + }, + { + "epoch": 1.84, + "learning_rate": 1.624407103255351e-05, + "loss": 1.0535, + "step": 24910 + }, + { + "epoch": 1.84, + "learning_rate": 1.6243759469331906e-05, + "loss": 0.9661, + "step": 24911 + }, + { + "epoch": 1.84, + "learning_rate": 1.624344789617657e-05, + "loss": 1.0073, + "step": 24912 + }, + { + "epoch": 1.84, + "learning_rate": 1.6243136313087998e-05, + "loss": 0.9552, + "step": 24913 + }, + { + "epoch": 1.84, + "learning_rate": 1.6242824720066685e-05, + "loss": 0.9375, + "step": 24914 + }, + { + "epoch": 1.84, + "learning_rate": 1.6242513117113125e-05, + "loss": 0.9847, + "step": 24915 + }, + { + "epoch": 1.84, + "learning_rate": 1.624220150422782e-05, + "loss": 1.0095, + "step": 24916 + }, + { + "epoch": 1.84, + "learning_rate": 1.6241889881411258e-05, + "loss": 1.089, + "step": 24917 + }, + { + "epoch": 1.84, + "learning_rate": 1.6241578248663937e-05, + "loss": 1.0468, + "step": 24918 + }, + { + "epoch": 1.84, + "learning_rate": 1.6241266605986358e-05, + "loss": 1.0505, + "step": 24919 + }, + { + "epoch": 1.84, + "learning_rate": 1.624095495337901e-05, + "loss": 1.0331, + "step": 24920 + }, + { + "epoch": 1.84, + "learning_rate": 1.6240643290842396e-05, + "loss": 0.9474, + "step": 24921 + }, + { + "epoch": 1.84, + "learning_rate": 1.624033161837701e-05, + "loss": 0.9972, + "step": 24922 + }, + { + "epoch": 1.84, + "learning_rate": 1.624001993598334e-05, + "loss": 1.0665, + "step": 24923 + }, + { + "epoch": 1.84, + "learning_rate": 1.6239708243661887e-05, + "loss": 0.9095, + "step": 24924 + }, + { + "epoch": 1.84, + "learning_rate": 1.623939654141315e-05, + "loss": 1.0432, + "step": 24925 + }, + { + "epoch": 1.84, + "learning_rate": 1.623908482923762e-05, + "loss": 1.065, + "step": 24926 + }, + { + "epoch": 1.84, + "learning_rate": 1.62387731071358e-05, + "loss": 1.0697, + "step": 24927 + }, + { + "epoch": 1.84, + "learning_rate": 1.623846137510818e-05, + "loss": 1.0133, + "step": 24928 + }, + { + "epoch": 1.84, + "learning_rate": 1.6238149633155255e-05, + "loss": 1.0401, + "step": 24929 + }, + { + "epoch": 1.84, + "learning_rate": 1.6237837881277524e-05, + "loss": 0.9829, + "step": 24930 + }, + { + "epoch": 1.84, + "learning_rate": 1.6237526119475484e-05, + "loss": 1.051, + "step": 24931 + }, + { + "epoch": 1.84, + "learning_rate": 1.6237214347749626e-05, + "loss": 1.0344, + "step": 24932 + }, + { + "epoch": 1.84, + "learning_rate": 1.6236902566100453e-05, + "loss": 0.9724, + "step": 24933 + }, + { + "epoch": 1.84, + "learning_rate": 1.623659077452845e-05, + "loss": 0.8917, + "step": 24934 + }, + { + "epoch": 1.84, + "learning_rate": 1.623627897303413e-05, + "loss": 0.9314, + "step": 24935 + }, + { + "epoch": 1.84, + "learning_rate": 1.623596716161797e-05, + "loss": 1.0206, + "step": 24936 + }, + { + "epoch": 1.84, + "learning_rate": 1.6235655340280482e-05, + "loss": 0.9983, + "step": 24937 + }, + { + "epoch": 1.84, + "learning_rate": 1.6235343509022154e-05, + "loss": 1.0154, + "step": 24938 + }, + { + "epoch": 1.84, + "learning_rate": 1.623503166784348e-05, + "loss": 1.0655, + "step": 24939 + }, + { + "epoch": 1.84, + "learning_rate": 1.6234719816744968e-05, + "loss": 0.8488, + "step": 24940 + }, + { + "epoch": 1.84, + "learning_rate": 1.62344079557271e-05, + "loss": 1.0071, + "step": 24941 + }, + { + "epoch": 1.84, + "learning_rate": 1.623409608479038e-05, + "loss": 1.0965, + "step": 24942 + }, + { + "epoch": 1.84, + "learning_rate": 1.62337842039353e-05, + "loss": 0.9451, + "step": 24943 + }, + { + "epoch": 1.84, + "learning_rate": 1.623347231316236e-05, + "loss": 0.9542, + "step": 24944 + }, + { + "epoch": 1.84, + "learning_rate": 1.6233160412472055e-05, + "loss": 1.03, + "step": 24945 + }, + { + "epoch": 1.84, + "learning_rate": 1.6232848501864883e-05, + "loss": 1.0331, + "step": 24946 + }, + { + "epoch": 1.84, + "learning_rate": 1.6232536581341337e-05, + "loss": 0.9057, + "step": 24947 + }, + { + "epoch": 1.84, + "learning_rate": 1.6232224650901912e-05, + "loss": 0.8989, + "step": 24948 + }, + { + "epoch": 1.84, + "learning_rate": 1.6231912710547105e-05, + "loss": 0.9529, + "step": 24949 + }, + { + "epoch": 1.84, + "learning_rate": 1.623160076027742e-05, + "loss": 1.1617, + "step": 24950 + }, + { + "epoch": 1.84, + "learning_rate": 1.6231288800093345e-05, + "loss": 0.9972, + "step": 24951 + }, + { + "epoch": 1.84, + "learning_rate": 1.623097682999538e-05, + "loss": 1.0341, + "step": 24952 + }, + { + "epoch": 1.84, + "learning_rate": 1.6230664849984015e-05, + "loss": 1.0787, + "step": 24953 + }, + { + "epoch": 1.84, + "learning_rate": 1.6230352860059757e-05, + "loss": 0.9461, + "step": 24954 + }, + { + "epoch": 1.84, + "learning_rate": 1.623004086022309e-05, + "loss": 1.0087, + "step": 24955 + }, + { + "epoch": 1.84, + "learning_rate": 1.6229728850474525e-05, + "loss": 0.943, + "step": 24956 + }, + { + "epoch": 1.84, + "learning_rate": 1.6229416830814544e-05, + "loss": 1.0915, + "step": 24957 + }, + { + "epoch": 1.84, + "learning_rate": 1.6229104801243655e-05, + "loss": 0.9433, + "step": 24958 + }, + { + "epoch": 1.84, + "learning_rate": 1.622879276176235e-05, + "loss": 1.0505, + "step": 24959 + }, + { + "epoch": 1.84, + "learning_rate": 1.622848071237112e-05, + "loss": 1.0382, + "step": 24960 + }, + { + "epoch": 1.84, + "learning_rate": 1.622816865307047e-05, + "loss": 0.9509, + "step": 24961 + }, + { + "epoch": 1.84, + "learning_rate": 1.6227856583860892e-05, + "loss": 1.0016, + "step": 24962 + }, + { + "epoch": 1.84, + "learning_rate": 1.622754450474288e-05, + "loss": 0.9877, + "step": 24963 + }, + { + "epoch": 1.84, + "learning_rate": 1.622723241571694e-05, + "loss": 0.9432, + "step": 24964 + }, + { + "epoch": 1.84, + "learning_rate": 1.622692031678356e-05, + "loss": 1.008, + "step": 24965 + }, + { + "epoch": 1.84, + "learning_rate": 1.6226608207943234e-05, + "loss": 0.9954, + "step": 24966 + }, + { + "epoch": 1.84, + "learning_rate": 1.6226296089196466e-05, + "loss": 0.9743, + "step": 24967 + }, + { + "epoch": 1.84, + "learning_rate": 1.622598396054375e-05, + "loss": 0.9558, + "step": 24968 + }, + { + "epoch": 1.84, + "learning_rate": 1.6225671821985587e-05, + "loss": 0.9345, + "step": 24969 + }, + { + "epoch": 1.85, + "learning_rate": 1.6225359673522466e-05, + "loss": 1.0338, + "step": 24970 + }, + { + "epoch": 1.85, + "learning_rate": 1.6225047515154884e-05, + "loss": 1.0046, + "step": 24971 + }, + { + "epoch": 1.85, + "learning_rate": 1.6224735346883346e-05, + "loss": 1.0424, + "step": 24972 + }, + { + "epoch": 1.85, + "learning_rate": 1.6224423168708343e-05, + "loss": 1.0082, + "step": 24973 + }, + { + "epoch": 1.85, + "learning_rate": 1.6224110980630365e-05, + "loss": 1.0172, + "step": 24974 + }, + { + "epoch": 1.85, + "learning_rate": 1.622379878264992e-05, + "loss": 1.0219, + "step": 24975 + }, + { + "epoch": 1.85, + "learning_rate": 1.62234865747675e-05, + "loss": 0.9079, + "step": 24976 + }, + { + "epoch": 1.85, + "learning_rate": 1.62231743569836e-05, + "loss": 0.9746, + "step": 24977 + }, + { + "epoch": 1.85, + "learning_rate": 1.6222862129298725e-05, + "loss": 1.0834, + "step": 24978 + }, + { + "epoch": 1.85, + "learning_rate": 1.622254989171336e-05, + "loss": 0.9806, + "step": 24979 + }, + { + "epoch": 1.85, + "learning_rate": 1.6222237644228008e-05, + "loss": 0.9953, + "step": 24980 + }, + { + "epoch": 1.85, + "learning_rate": 1.6221925386843162e-05, + "loss": 0.9618, + "step": 24981 + }, + { + "epoch": 1.85, + "learning_rate": 1.6221613119559323e-05, + "loss": 0.994, + "step": 24982 + }, + { + "epoch": 1.85, + "learning_rate": 1.622130084237699e-05, + "loss": 0.9362, + "step": 24983 + }, + { + "epoch": 1.85, + "learning_rate": 1.6220988555296653e-05, + "loss": 1.0281, + "step": 24984 + }, + { + "epoch": 1.85, + "learning_rate": 1.6220676258318814e-05, + "loss": 1.0109, + "step": 24985 + }, + { + "epoch": 1.85, + "learning_rate": 1.6220363951443968e-05, + "loss": 1.0073, + "step": 24986 + }, + { + "epoch": 1.85, + "learning_rate": 1.6220051634672612e-05, + "loss": 1.0576, + "step": 24987 + }, + { + "epoch": 1.85, + "learning_rate": 1.6219739308005245e-05, + "loss": 0.8764, + "step": 24988 + }, + { + "epoch": 1.85, + "learning_rate": 1.621942697144236e-05, + "loss": 0.9299, + "step": 24989 + }, + { + "epoch": 1.85, + "learning_rate": 1.6219114624984453e-05, + "loss": 0.913, + "step": 24990 + }, + { + "epoch": 1.85, + "learning_rate": 1.6218802268632027e-05, + "loss": 0.9631, + "step": 24991 + }, + { + "epoch": 1.85, + "learning_rate": 1.6218489902385572e-05, + "loss": 0.9954, + "step": 24992 + }, + { + "epoch": 1.85, + "learning_rate": 1.6218177526245592e-05, + "loss": 0.9644, + "step": 24993 + }, + { + "epoch": 1.85, + "learning_rate": 1.621786514021258e-05, + "loss": 1.0923, + "step": 24994 + }, + { + "epoch": 1.85, + "learning_rate": 1.621755274428703e-05, + "loss": 1.0396, + "step": 24995 + }, + { + "epoch": 1.85, + "learning_rate": 1.6217240338469443e-05, + "loss": 0.8835, + "step": 24996 + }, + { + "epoch": 1.85, + "learning_rate": 1.6216927922760323e-05, + "loss": 1.0218, + "step": 24997 + }, + { + "epoch": 1.85, + "learning_rate": 1.621661549716015e-05, + "loss": 0.9402, + "step": 24998 + }, + { + "epoch": 1.85, + "learning_rate": 1.6216303061669437e-05, + "loss": 1.0373, + "step": 24999 + }, + { + "epoch": 1.85, + "learning_rate": 1.621599061628867e-05, + "loss": 1.0219, + "step": 25000 + }, + { + "epoch": 1.85, + "learning_rate": 1.621567816101836e-05, + "loss": 1.123, + "step": 25001 + }, + { + "epoch": 1.85, + "learning_rate": 1.6215365695858988e-05, + "loss": 0.9314, + "step": 25002 + }, + { + "epoch": 1.85, + "learning_rate": 1.6215053220811055e-05, + "loss": 0.9148, + "step": 25003 + }, + { + "epoch": 1.85, + "learning_rate": 1.6214740735875064e-05, + "loss": 0.9765, + "step": 25004 + }, + { + "epoch": 1.85, + "learning_rate": 1.6214428241051514e-05, + "loss": 0.9176, + "step": 25005 + }, + { + "epoch": 1.85, + "learning_rate": 1.6214115736340895e-05, + "loss": 0.941, + "step": 25006 + }, + { + "epoch": 1.85, + "learning_rate": 1.6213803221743707e-05, + "loss": 0.8286, + "step": 25007 + }, + { + "epoch": 1.85, + "learning_rate": 1.6213490697260445e-05, + "loss": 1.0385, + "step": 25008 + }, + { + "epoch": 1.85, + "learning_rate": 1.6213178162891607e-05, + "loss": 0.8732, + "step": 25009 + }, + { + "epoch": 1.85, + "learning_rate": 1.6212865618637695e-05, + "loss": 0.9487, + "step": 25010 + }, + { + "epoch": 1.85, + "learning_rate": 1.62125530644992e-05, + "loss": 1.0978, + "step": 25011 + }, + { + "epoch": 1.85, + "learning_rate": 1.6212240500476623e-05, + "loss": 1.0345, + "step": 25012 + }, + { + "epoch": 1.85, + "learning_rate": 1.621192792657046e-05, + "loss": 0.9942, + "step": 25013 + }, + { + "epoch": 1.85, + "learning_rate": 1.621161534278121e-05, + "loss": 1.0508, + "step": 25014 + }, + { + "epoch": 1.85, + "learning_rate": 1.6211302749109366e-05, + "loss": 0.8807, + "step": 25015 + }, + { + "epoch": 1.85, + "learning_rate": 1.6210990145555433e-05, + "loss": 1.0958, + "step": 25016 + }, + { + "epoch": 1.85, + "learning_rate": 1.6210677532119898e-05, + "loss": 0.9225, + "step": 25017 + }, + { + "epoch": 1.85, + "learning_rate": 1.621036490880327e-05, + "loss": 1.076, + "step": 25018 + }, + { + "epoch": 1.85, + "learning_rate": 1.6210052275606037e-05, + "loss": 0.9344, + "step": 25019 + }, + { + "epoch": 1.85, + "learning_rate": 1.62097396325287e-05, + "loss": 1.0019, + "step": 25020 + }, + { + "epoch": 1.85, + "learning_rate": 1.6209426979571752e-05, + "loss": 1.0449, + "step": 25021 + }, + { + "epoch": 1.85, + "learning_rate": 1.6209114316735697e-05, + "loss": 1.0301, + "step": 25022 + }, + { + "epoch": 1.85, + "learning_rate": 1.6208801644021033e-05, + "loss": 0.9182, + "step": 25023 + }, + { + "epoch": 1.85, + "learning_rate": 1.620848896142825e-05, + "loss": 1.0708, + "step": 25024 + }, + { + "epoch": 1.85, + "learning_rate": 1.6208176268957853e-05, + "loss": 1.1002, + "step": 25025 + }, + { + "epoch": 1.85, + "learning_rate": 1.6207863566610337e-05, + "loss": 1.1042, + "step": 25026 + }, + { + "epoch": 1.85, + "learning_rate": 1.6207550854386196e-05, + "loss": 1.0749, + "step": 25027 + }, + { + "epoch": 1.85, + "learning_rate": 1.6207238132285933e-05, + "loss": 1.0021, + "step": 25028 + }, + { + "epoch": 1.85, + "learning_rate": 1.620692540031004e-05, + "loss": 1.071, + "step": 25029 + }, + { + "epoch": 1.85, + "learning_rate": 1.620661265845902e-05, + "loss": 1.0029, + "step": 25030 + }, + { + "epoch": 1.85, + "learning_rate": 1.620629990673337e-05, + "loss": 0.8687, + "step": 25031 + }, + { + "epoch": 1.85, + "learning_rate": 1.620598714513358e-05, + "loss": 0.9887, + "step": 25032 + }, + { + "epoch": 1.85, + "learning_rate": 1.6205674373660157e-05, + "loss": 0.8251, + "step": 25033 + }, + { + "epoch": 1.85, + "learning_rate": 1.6205361592313598e-05, + "loss": 0.9969, + "step": 25034 + }, + { + "epoch": 1.85, + "learning_rate": 1.6205048801094393e-05, + "loss": 1.0261, + "step": 25035 + }, + { + "epoch": 1.85, + "learning_rate": 1.6204736000003042e-05, + "loss": 0.9997, + "step": 25036 + }, + { + "epoch": 1.85, + "learning_rate": 1.6204423189040047e-05, + "loss": 1.1362, + "step": 25037 + }, + { + "epoch": 1.85, + "learning_rate": 1.6204110368205907e-05, + "loss": 0.9111, + "step": 25038 + }, + { + "epoch": 1.85, + "learning_rate": 1.6203797537501112e-05, + "loss": 1.0988, + "step": 25039 + }, + { + "epoch": 1.85, + "learning_rate": 1.6203484696926164e-05, + "loss": 0.9776, + "step": 25040 + }, + { + "epoch": 1.85, + "learning_rate": 1.6203171846481564e-05, + "loss": 1.0184, + "step": 25041 + }, + { + "epoch": 1.85, + "learning_rate": 1.62028589861678e-05, + "loss": 1.0205, + "step": 25042 + }, + { + "epoch": 1.85, + "learning_rate": 1.6202546115985386e-05, + "loss": 1.134, + "step": 25043 + }, + { + "epoch": 1.85, + "learning_rate": 1.6202233235934803e-05, + "loss": 1.0042, + "step": 25044 + }, + { + "epoch": 1.85, + "learning_rate": 1.6201920346016557e-05, + "loss": 1.0784, + "step": 25045 + }, + { + "epoch": 1.85, + "learning_rate": 1.6201607446231143e-05, + "loss": 1.0435, + "step": 25046 + }, + { + "epoch": 1.85, + "learning_rate": 1.6201294536579062e-05, + "loss": 0.9613, + "step": 25047 + }, + { + "epoch": 1.85, + "learning_rate": 1.620098161706081e-05, + "loss": 1.1239, + "step": 25048 + }, + { + "epoch": 1.85, + "learning_rate": 1.6200668687676883e-05, + "loss": 1.122, + "step": 25049 + }, + { + "epoch": 1.85, + "learning_rate": 1.6200355748427784e-05, + "loss": 0.945, + "step": 25050 + }, + { + "epoch": 1.85, + "learning_rate": 1.6200042799314007e-05, + "loss": 0.9756, + "step": 25051 + }, + { + "epoch": 1.85, + "learning_rate": 1.6199729840336047e-05, + "loss": 1.101, + "step": 25052 + }, + { + "epoch": 1.85, + "learning_rate": 1.6199416871494408e-05, + "loss": 1.0735, + "step": 25053 + }, + { + "epoch": 1.85, + "learning_rate": 1.619910389278959e-05, + "loss": 0.915, + "step": 25054 + }, + { + "epoch": 1.85, + "learning_rate": 1.6198790904222082e-05, + "loss": 1.0609, + "step": 25055 + }, + { + "epoch": 1.85, + "learning_rate": 1.6198477905792383e-05, + "loss": 1.0266, + "step": 25056 + }, + { + "epoch": 1.85, + "learning_rate": 1.6198164897500998e-05, + "loss": 1.0435, + "step": 25057 + }, + { + "epoch": 1.85, + "learning_rate": 1.6197851879348422e-05, + "loss": 1.0494, + "step": 25058 + }, + { + "epoch": 1.85, + "learning_rate": 1.6197538851335152e-05, + "loss": 0.9294, + "step": 25059 + }, + { + "epoch": 1.85, + "learning_rate": 1.6197225813461687e-05, + "loss": 0.9942, + "step": 25060 + }, + { + "epoch": 1.85, + "learning_rate": 1.619691276572852e-05, + "loss": 1.0328, + "step": 25061 + }, + { + "epoch": 1.85, + "learning_rate": 1.6196599708136157e-05, + "loss": 0.9962, + "step": 25062 + }, + { + "epoch": 1.85, + "learning_rate": 1.6196286640685094e-05, + "loss": 0.8703, + "step": 25063 + }, + { + "epoch": 1.85, + "learning_rate": 1.6195973563375828e-05, + "loss": 1.0576, + "step": 25064 + }, + { + "epoch": 1.85, + "learning_rate": 1.6195660476208854e-05, + "loss": 1.0169, + "step": 25065 + }, + { + "epoch": 1.85, + "learning_rate": 1.6195347379184672e-05, + "loss": 0.9993, + "step": 25066 + }, + { + "epoch": 1.85, + "learning_rate": 1.6195034272303782e-05, + "loss": 1.0652, + "step": 25067 + }, + { + "epoch": 1.85, + "learning_rate": 1.619472115556668e-05, + "loss": 1.0274, + "step": 25068 + }, + { + "epoch": 1.85, + "learning_rate": 1.619440802897387e-05, + "loss": 1.001, + "step": 25069 + }, + { + "epoch": 1.85, + "learning_rate": 1.619409489252584e-05, + "loss": 1.041, + "step": 25070 + }, + { + "epoch": 1.85, + "learning_rate": 1.6193781746223096e-05, + "loss": 1.0651, + "step": 25071 + }, + { + "epoch": 1.85, + "learning_rate": 1.6193468590066132e-05, + "loss": 0.8644, + "step": 25072 + }, + { + "epoch": 1.85, + "learning_rate": 1.619315542405545e-05, + "loss": 1.0306, + "step": 25073 + }, + { + "epoch": 1.85, + "learning_rate": 1.6192842248191546e-05, + "loss": 1.0723, + "step": 25074 + }, + { + "epoch": 1.85, + "learning_rate": 1.619252906247492e-05, + "loss": 1.078, + "step": 25075 + }, + { + "epoch": 1.85, + "learning_rate": 1.6192215866906066e-05, + "loss": 0.8965, + "step": 25076 + }, + { + "epoch": 1.85, + "learning_rate": 1.6191902661485486e-05, + "loss": 0.9072, + "step": 25077 + }, + { + "epoch": 1.85, + "learning_rate": 1.619158944621368e-05, + "loss": 0.989, + "step": 25078 + }, + { + "epoch": 1.85, + "learning_rate": 1.6191276221091144e-05, + "loss": 1.1454, + "step": 25079 + }, + { + "epoch": 1.85, + "learning_rate": 1.6190962986118374e-05, + "loss": 1.0052, + "step": 25080 + }, + { + "epoch": 1.85, + "learning_rate": 1.619064974129587e-05, + "loss": 0.9688, + "step": 25081 + }, + { + "epoch": 1.85, + "learning_rate": 1.6190336486624133e-05, + "loss": 0.9374, + "step": 25082 + }, + { + "epoch": 1.85, + "learning_rate": 1.6190023222103655e-05, + "loss": 1.044, + "step": 25083 + }, + { + "epoch": 1.85, + "learning_rate": 1.6189709947734942e-05, + "loss": 0.8719, + "step": 25084 + }, + { + "epoch": 1.85, + "learning_rate": 1.6189396663518486e-05, + "loss": 0.929, + "step": 25085 + }, + { + "epoch": 1.85, + "learning_rate": 1.6189083369454793e-05, + "loss": 1.0504, + "step": 25086 + }, + { + "epoch": 1.85, + "learning_rate": 1.6188770065544353e-05, + "loss": 1.015, + "step": 25087 + }, + { + "epoch": 1.85, + "learning_rate": 1.6188456751787672e-05, + "loss": 1.0574, + "step": 25088 + }, + { + "epoch": 1.85, + "learning_rate": 1.618814342818524e-05, + "loss": 0.946, + "step": 25089 + }, + { + "epoch": 1.85, + "learning_rate": 1.6187830094737565e-05, + "loss": 1.0932, + "step": 25090 + }, + { + "epoch": 1.85, + "learning_rate": 1.6187516751445138e-05, + "loss": 0.9579, + "step": 25091 + }, + { + "epoch": 1.85, + "learning_rate": 1.6187203398308463e-05, + "loss": 0.9877, + "step": 25092 + }, + { + "epoch": 1.85, + "learning_rate": 1.6186890035328035e-05, + "loss": 0.8682, + "step": 25093 + }, + { + "epoch": 1.85, + "learning_rate": 1.6186576662504353e-05, + "loss": 1.0609, + "step": 25094 + }, + { + "epoch": 1.85, + "learning_rate": 1.6186263279837914e-05, + "loss": 1.1507, + "step": 25095 + }, + { + "epoch": 1.85, + "learning_rate": 1.6185949887329223e-05, + "loss": 0.952, + "step": 25096 + }, + { + "epoch": 1.85, + "learning_rate": 1.6185636484978768e-05, + "loss": 0.9652, + "step": 25097 + }, + { + "epoch": 1.85, + "learning_rate": 1.6185323072787058e-05, + "loss": 1.0273, + "step": 25098 + }, + { + "epoch": 1.85, + "learning_rate": 1.618500965075459e-05, + "loss": 1.0647, + "step": 25099 + }, + { + "epoch": 1.85, + "learning_rate": 1.6184696218881855e-05, + "loss": 0.9789, + "step": 25100 + }, + { + "epoch": 1.85, + "learning_rate": 1.618438277716936e-05, + "loss": 0.9251, + "step": 25101 + }, + { + "epoch": 1.85, + "learning_rate": 1.61840693256176e-05, + "loss": 0.8683, + "step": 25102 + }, + { + "epoch": 1.85, + "learning_rate": 1.618375586422707e-05, + "loss": 0.9306, + "step": 25103 + }, + { + "epoch": 1.85, + "learning_rate": 1.618344239299828e-05, + "loss": 0.9094, + "step": 25104 + }, + { + "epoch": 1.85, + "learning_rate": 1.6183128911931718e-05, + "loss": 0.9378, + "step": 25105 + }, + { + "epoch": 1.86, + "learning_rate": 1.6182815421027885e-05, + "loss": 1.001, + "step": 25106 + }, + { + "epoch": 1.86, + "learning_rate": 1.6182501920287283e-05, + "loss": 0.9198, + "step": 25107 + }, + { + "epoch": 1.86, + "learning_rate": 1.618218840971041e-05, + "loss": 1.0745, + "step": 25108 + }, + { + "epoch": 1.86, + "learning_rate": 1.618187488929776e-05, + "loss": 1.0776, + "step": 25109 + }, + { + "epoch": 1.86, + "learning_rate": 1.618156135904984e-05, + "loss": 1.0206, + "step": 25110 + }, + { + "epoch": 1.86, + "learning_rate": 1.618124781896714e-05, + "loss": 1.0705, + "step": 25111 + }, + { + "epoch": 1.86, + "learning_rate": 1.618093426905017e-05, + "loss": 0.9474, + "step": 25112 + }, + { + "epoch": 1.86, + "learning_rate": 1.6180620709299414e-05, + "loss": 0.9665, + "step": 25113 + }, + { + "epoch": 1.86, + "learning_rate": 1.618030713971538e-05, + "loss": 1.0548, + "step": 25114 + }, + { + "epoch": 1.86, + "learning_rate": 1.617999356029857e-05, + "loss": 0.9608, + "step": 25115 + }, + { + "epoch": 1.86, + "learning_rate": 1.6179679971049477e-05, + "loss": 1.1, + "step": 25116 + }, + { + "epoch": 1.86, + "learning_rate": 1.61793663719686e-05, + "loss": 0.9545, + "step": 25117 + }, + { + "epoch": 1.86, + "learning_rate": 1.6179052763056442e-05, + "loss": 1.0282, + "step": 25118 + }, + { + "epoch": 1.86, + "learning_rate": 1.6178739144313497e-05, + "loss": 1.0249, + "step": 25119 + }, + { + "epoch": 1.86, + "learning_rate": 1.6178425515740268e-05, + "loss": 0.9934, + "step": 25120 + }, + { + "epoch": 1.86, + "learning_rate": 1.6178111877337252e-05, + "loss": 1.0385, + "step": 25121 + }, + { + "epoch": 1.86, + "learning_rate": 1.6177798229104947e-05, + "loss": 1.0182, + "step": 25122 + }, + { + "epoch": 1.86, + "learning_rate": 1.6177484571043855e-05, + "loss": 0.9477, + "step": 25123 + }, + { + "epoch": 1.86, + "learning_rate": 1.6177170903154477e-05, + "loss": 0.9134, + "step": 25124 + }, + { + "epoch": 1.86, + "learning_rate": 1.61768572254373e-05, + "loss": 0.9233, + "step": 25125 + }, + { + "epoch": 1.86, + "learning_rate": 1.6176543537892838e-05, + "loss": 0.9117, + "step": 25126 + }, + { + "epoch": 1.86, + "learning_rate": 1.6176229840521583e-05, + "loss": 1.0793, + "step": 25127 + }, + { + "epoch": 1.86, + "learning_rate": 1.6175916133324032e-05, + "loss": 1.0497, + "step": 25128 + }, + { + "epoch": 1.86, + "learning_rate": 1.617560241630069e-05, + "loss": 0.9802, + "step": 25129 + }, + { + "epoch": 1.86, + "learning_rate": 1.617528868945205e-05, + "loss": 0.9678, + "step": 25130 + }, + { + "epoch": 1.86, + "learning_rate": 1.6174974952778616e-05, + "loss": 1.0161, + "step": 25131 + }, + { + "epoch": 1.86, + "learning_rate": 1.6174661206280884e-05, + "loss": 0.9572, + "step": 25132 + }, + { + "epoch": 1.86, + "learning_rate": 1.6174347449959355e-05, + "loss": 0.9553, + "step": 25133 + }, + { + "epoch": 1.86, + "learning_rate": 1.6174033683814525e-05, + "loss": 0.9597, + "step": 25134 + }, + { + "epoch": 1.86, + "learning_rate": 1.61737199078469e-05, + "loss": 0.9903, + "step": 25135 + }, + { + "epoch": 1.86, + "learning_rate": 1.6173406122056973e-05, + "loss": 0.9235, + "step": 25136 + }, + { + "epoch": 1.86, + "learning_rate": 1.617309232644525e-05, + "loss": 0.9139, + "step": 25137 + }, + { + "epoch": 1.86, + "learning_rate": 1.6172778521012217e-05, + "loss": 0.9726, + "step": 25138 + }, + { + "epoch": 1.86, + "learning_rate": 1.6172464705758385e-05, + "loss": 0.9729, + "step": 25139 + }, + { + "epoch": 1.86, + "learning_rate": 1.617215088068425e-05, + "loss": 0.9315, + "step": 25140 + }, + { + "epoch": 1.86, + "learning_rate": 1.617183704579031e-05, + "loss": 0.974, + "step": 25141 + }, + { + "epoch": 1.86, + "learning_rate": 1.617152320107707e-05, + "loss": 1.0687, + "step": 25142 + }, + { + "epoch": 1.86, + "learning_rate": 1.617120934654502e-05, + "loss": 0.9656, + "step": 25143 + }, + { + "epoch": 1.86, + "learning_rate": 1.617089548219467e-05, + "loss": 1.0026, + "step": 25144 + }, + { + "epoch": 1.86, + "learning_rate": 1.6170581608026506e-05, + "loss": 1.0442, + "step": 25145 + }, + { + "epoch": 1.86, + "learning_rate": 1.617026772404104e-05, + "loss": 1.0454, + "step": 25146 + }, + { + "epoch": 1.86, + "learning_rate": 1.6169953830238765e-05, + "loss": 1.0496, + "step": 25147 + }, + { + "epoch": 1.86, + "learning_rate": 1.6169639926620182e-05, + "loss": 0.9448, + "step": 25148 + }, + { + "epoch": 1.86, + "learning_rate": 1.616932601318579e-05, + "loss": 1.173, + "step": 25149 + }, + { + "epoch": 1.86, + "learning_rate": 1.6169012089936088e-05, + "loss": 0.9955, + "step": 25150 + }, + { + "epoch": 1.86, + "learning_rate": 1.6168698156871577e-05, + "loss": 0.9105, + "step": 25151 + }, + { + "epoch": 1.86, + "learning_rate": 1.6168384213992756e-05, + "loss": 0.9299, + "step": 25152 + }, + { + "epoch": 1.86, + "learning_rate": 1.6168070261300123e-05, + "loss": 1.006, + "step": 25153 + }, + { + "epoch": 1.86, + "learning_rate": 1.6167756298794176e-05, + "loss": 0.9108, + "step": 25154 + }, + { + "epoch": 1.86, + "learning_rate": 1.616744232647542e-05, + "loss": 0.9668, + "step": 25155 + }, + { + "epoch": 1.86, + "learning_rate": 1.6167128344344355e-05, + "loss": 1.0149, + "step": 25156 + }, + { + "epoch": 1.86, + "learning_rate": 1.6166814352401472e-05, + "loss": 0.948, + "step": 25157 + }, + { + "epoch": 1.86, + "learning_rate": 1.6166500350647275e-05, + "loss": 1.0648, + "step": 25158 + }, + { + "epoch": 1.86, + "learning_rate": 1.6166186339082267e-05, + "loss": 0.9567, + "step": 25159 + }, + { + "epoch": 1.86, + "learning_rate": 1.6165872317706945e-05, + "loss": 1.0003, + "step": 25160 + }, + { + "epoch": 1.86, + "learning_rate": 1.6165558286521805e-05, + "loss": 1.0217, + "step": 25161 + }, + { + "epoch": 1.86, + "learning_rate": 1.6165244245527353e-05, + "loss": 1.0455, + "step": 25162 + }, + { + "epoch": 1.86, + "learning_rate": 1.6164930194724083e-05, + "loss": 1.0495, + "step": 25163 + }, + { + "epoch": 1.86, + "learning_rate": 1.61646161341125e-05, + "loss": 1.0154, + "step": 25164 + }, + { + "epoch": 1.86, + "learning_rate": 1.61643020636931e-05, + "loss": 1.1386, + "step": 25165 + }, + { + "epoch": 1.86, + "learning_rate": 1.616398798346638e-05, + "loss": 1.0287, + "step": 25166 + }, + { + "epoch": 1.86, + "learning_rate": 1.6163673893432848e-05, + "loss": 1.0492, + "step": 25167 + }, + { + "epoch": 1.86, + "learning_rate": 1.6163359793592998e-05, + "loss": 0.9536, + "step": 25168 + }, + { + "epoch": 1.86, + "learning_rate": 1.6163045683947332e-05, + "loss": 0.9967, + "step": 25169 + }, + { + "epoch": 1.86, + "learning_rate": 1.6162731564496348e-05, + "loss": 0.9958, + "step": 25170 + }, + { + "epoch": 1.86, + "learning_rate": 1.6162417435240544e-05, + "loss": 1.0949, + "step": 25171 + }, + { + "epoch": 1.86, + "learning_rate": 1.6162103296180422e-05, + "loss": 1.1449, + "step": 25172 + }, + { + "epoch": 1.86, + "learning_rate": 1.6161789147316483e-05, + "loss": 1.1159, + "step": 25173 + }, + { + "epoch": 1.86, + "learning_rate": 1.6161474988649226e-05, + "loss": 0.9845, + "step": 25174 + }, + { + "epoch": 1.86, + "learning_rate": 1.616116082017915e-05, + "loss": 1.0457, + "step": 25175 + }, + { + "epoch": 1.86, + "learning_rate": 1.6160846641906757e-05, + "loss": 0.9992, + "step": 25176 + }, + { + "epoch": 1.86, + "learning_rate": 1.616053245383254e-05, + "loss": 0.9754, + "step": 25177 + }, + { + "epoch": 1.86, + "learning_rate": 1.6160218255957013e-05, + "loss": 1.0288, + "step": 25178 + }, + { + "epoch": 1.86, + "learning_rate": 1.615990404828066e-05, + "loss": 0.9542, + "step": 25179 + }, + { + "epoch": 1.86, + "learning_rate": 1.615958983080399e-05, + "loss": 1.0749, + "step": 25180 + }, + { + "epoch": 1.86, + "learning_rate": 1.6159275603527497e-05, + "loss": 0.9863, + "step": 25181 + }, + { + "epoch": 1.86, + "learning_rate": 1.615896136645169e-05, + "loss": 1.075, + "step": 25182 + }, + { + "epoch": 1.86, + "learning_rate": 1.615864711957706e-05, + "loss": 0.9291, + "step": 25183 + }, + { + "epoch": 1.86, + "learning_rate": 1.6158332862904114e-05, + "loss": 0.9417, + "step": 25184 + }, + { + "epoch": 1.86, + "learning_rate": 1.6158018596433343e-05, + "loss": 1.0213, + "step": 25185 + }, + { + "epoch": 1.86, + "learning_rate": 1.6157704320165258e-05, + "loss": 1.0447, + "step": 25186 + }, + { + "epoch": 1.86, + "learning_rate": 1.6157390034100352e-05, + "loss": 0.9813, + "step": 25187 + }, + { + "epoch": 1.86, + "learning_rate": 1.6157075738239124e-05, + "loss": 0.9264, + "step": 25188 + }, + { + "epoch": 1.86, + "learning_rate": 1.615676143258208e-05, + "loss": 0.9538, + "step": 25189 + }, + { + "epoch": 1.86, + "learning_rate": 1.6156447117129714e-05, + "loss": 1.0258, + "step": 25190 + }, + { + "epoch": 1.86, + "learning_rate": 1.6156132791882527e-05, + "loss": 1.0166, + "step": 25191 + }, + { + "epoch": 1.86, + "learning_rate": 1.6155818456841025e-05, + "loss": 1.0917, + "step": 25192 + }, + { + "epoch": 1.86, + "learning_rate": 1.6155504112005703e-05, + "loss": 1.0248, + "step": 25193 + }, + { + "epoch": 1.86, + "learning_rate": 1.6155189757377062e-05, + "loss": 1.0849, + "step": 25194 + }, + { + "epoch": 1.86, + "learning_rate": 1.61548753929556e-05, + "loss": 0.9963, + "step": 25195 + }, + { + "epoch": 1.86, + "learning_rate": 1.6154561018741823e-05, + "loss": 1.05, + "step": 25196 + }, + { + "epoch": 1.86, + "learning_rate": 1.6154246634736226e-05, + "loss": 0.8949, + "step": 25197 + }, + { + "epoch": 1.86, + "learning_rate": 1.615393224093931e-05, + "loss": 0.9998, + "step": 25198 + }, + { + "epoch": 1.86, + "learning_rate": 1.6153617837351575e-05, + "loss": 0.9503, + "step": 25199 + }, + { + "epoch": 1.86, + "learning_rate": 1.6153303423973524e-05, + "loss": 0.9705, + "step": 25200 + }, + { + "epoch": 1.86, + "learning_rate": 1.6152989000805653e-05, + "loss": 1.0064, + "step": 25201 + }, + { + "epoch": 1.86, + "learning_rate": 1.6152674567848466e-05, + "loss": 1.0599, + "step": 25202 + }, + { + "epoch": 1.86, + "learning_rate": 1.6152360125102465e-05, + "loss": 1.0264, + "step": 25203 + }, + { + "epoch": 1.86, + "learning_rate": 1.6152045672568145e-05, + "loss": 0.9688, + "step": 25204 + }, + { + "epoch": 1.86, + "learning_rate": 1.6151731210246003e-05, + "loss": 1.0228, + "step": 25205 + }, + { + "epoch": 1.86, + "learning_rate": 1.615141673813655e-05, + "loss": 0.9619, + "step": 25206 + }, + { + "epoch": 1.86, + "learning_rate": 1.615110225624028e-05, + "loss": 1.0841, + "step": 25207 + }, + { + "epoch": 1.86, + "learning_rate": 1.6150787764557694e-05, + "loss": 1.0424, + "step": 25208 + }, + { + "epoch": 1.86, + "learning_rate": 1.6150473263089294e-05, + "loss": 0.9386, + "step": 25209 + }, + { + "epoch": 1.86, + "learning_rate": 1.6150158751835577e-05, + "loss": 0.8943, + "step": 25210 + }, + { + "epoch": 1.86, + "learning_rate": 1.6149844230797043e-05, + "loss": 1.0344, + "step": 25211 + }, + { + "epoch": 1.86, + "learning_rate": 1.6149529699974202e-05, + "loss": 0.9039, + "step": 25212 + }, + { + "epoch": 1.86, + "learning_rate": 1.614921515936754e-05, + "loss": 1.1327, + "step": 25213 + }, + { + "epoch": 1.86, + "learning_rate": 1.614890060897757e-05, + "loss": 0.9754, + "step": 25214 + }, + { + "epoch": 1.86, + "learning_rate": 1.6148586048804786e-05, + "loss": 1.083, + "step": 25215 + }, + { + "epoch": 1.86, + "learning_rate": 1.614827147884969e-05, + "loss": 0.9362, + "step": 25216 + }, + { + "epoch": 1.86, + "learning_rate": 1.614795689911278e-05, + "loss": 1.0482, + "step": 25217 + }, + { + "epoch": 1.86, + "learning_rate": 1.6147642309594556e-05, + "loss": 1.0025, + "step": 25218 + }, + { + "epoch": 1.86, + "learning_rate": 1.6147327710295525e-05, + "loss": 1.0142, + "step": 25219 + }, + { + "epoch": 1.86, + "learning_rate": 1.6147013101216184e-05, + "loss": 1.0524, + "step": 25220 + }, + { + "epoch": 1.86, + "learning_rate": 1.6146698482357034e-05, + "loss": 1.0587, + "step": 25221 + }, + { + "epoch": 1.86, + "learning_rate": 1.614638385371857e-05, + "loss": 0.9972, + "step": 25222 + }, + { + "epoch": 1.86, + "learning_rate": 1.6146069215301297e-05, + "loss": 0.9223, + "step": 25223 + }, + { + "epoch": 1.86, + "learning_rate": 1.614575456710572e-05, + "loss": 1.0586, + "step": 25224 + }, + { + "epoch": 1.86, + "learning_rate": 1.6145439909132337e-05, + "loss": 1.079, + "step": 25225 + }, + { + "epoch": 1.86, + "learning_rate": 1.6145125241381643e-05, + "loss": 1.0722, + "step": 25226 + }, + { + "epoch": 1.86, + "learning_rate": 1.614481056385414e-05, + "loss": 1.0354, + "step": 25227 + }, + { + "epoch": 1.86, + "learning_rate": 1.6144495876550335e-05, + "loss": 1.0648, + "step": 25228 + }, + { + "epoch": 1.86, + "learning_rate": 1.614418117947073e-05, + "loss": 1.1079, + "step": 25229 + }, + { + "epoch": 1.86, + "learning_rate": 1.6143866472615816e-05, + "loss": 0.9438, + "step": 25230 + }, + { + "epoch": 1.86, + "learning_rate": 1.61435517559861e-05, + "loss": 1.0272, + "step": 25231 + }, + { + "epoch": 1.86, + "learning_rate": 1.6143237029582076e-05, + "loss": 1.0042, + "step": 25232 + }, + { + "epoch": 1.86, + "learning_rate": 1.6142922293404252e-05, + "loss": 1.0054, + "step": 25233 + }, + { + "epoch": 1.86, + "learning_rate": 1.6142607547453128e-05, + "loss": 1.0545, + "step": 25234 + }, + { + "epoch": 1.86, + "learning_rate": 1.6142292791729205e-05, + "loss": 1.0273, + "step": 25235 + }, + { + "epoch": 1.86, + "learning_rate": 1.6141978026232975e-05, + "loss": 1.0746, + "step": 25236 + }, + { + "epoch": 1.86, + "learning_rate": 1.6141663250964953e-05, + "loss": 0.9867, + "step": 25237 + }, + { + "epoch": 1.86, + "learning_rate": 1.6141348465925626e-05, + "loss": 1.1022, + "step": 25238 + }, + { + "epoch": 1.86, + "learning_rate": 1.614103367111551e-05, + "loss": 1.0389, + "step": 25239 + }, + { + "epoch": 1.86, + "learning_rate": 1.6140718866535094e-05, + "loss": 1.0868, + "step": 25240 + }, + { + "epoch": 1.87, + "learning_rate": 1.614040405218488e-05, + "loss": 1.0804, + "step": 25241 + }, + { + "epoch": 1.87, + "learning_rate": 1.614008922806537e-05, + "loss": 1.0358, + "step": 25242 + }, + { + "epoch": 1.87, + "learning_rate": 1.613977439417707e-05, + "loss": 0.9685, + "step": 25243 + }, + { + "epoch": 1.87, + "learning_rate": 1.6139459550520477e-05, + "loss": 1.0399, + "step": 25244 + }, + { + "epoch": 1.87, + "learning_rate": 1.6139144697096087e-05, + "loss": 1.0664, + "step": 25245 + }, + { + "epoch": 1.87, + "learning_rate": 1.613882983390441e-05, + "loss": 1.0442, + "step": 25246 + }, + { + "epoch": 1.87, + "learning_rate": 1.613851496094594e-05, + "loss": 1.0814, + "step": 25247 + }, + { + "epoch": 1.87, + "learning_rate": 1.613820007822118e-05, + "loss": 1.1326, + "step": 25248 + }, + { + "epoch": 1.87, + "learning_rate": 1.6137885185730635e-05, + "loss": 1.0605, + "step": 25249 + }, + { + "epoch": 1.87, + "learning_rate": 1.6137570283474796e-05, + "loss": 0.8241, + "step": 25250 + }, + { + "epoch": 1.87, + "learning_rate": 1.6137255371454175e-05, + "loss": 1.0566, + "step": 25251 + }, + { + "epoch": 1.87, + "learning_rate": 1.613694044966927e-05, + "loss": 1.0091, + "step": 25252 + }, + { + "epoch": 1.87, + "learning_rate": 1.613662551812058e-05, + "loss": 0.9308, + "step": 25253 + }, + { + "epoch": 1.87, + "learning_rate": 1.6136310576808604e-05, + "loss": 1.1186, + "step": 25254 + }, + { + "epoch": 1.87, + "learning_rate": 1.6135995625733845e-05, + "loss": 0.9803, + "step": 25255 + }, + { + "epoch": 1.87, + "learning_rate": 1.613568066489681e-05, + "loss": 1.0372, + "step": 25256 + }, + { + "epoch": 1.87, + "learning_rate": 1.6135365694297988e-05, + "loss": 1.0494, + "step": 25257 + }, + { + "epoch": 1.87, + "learning_rate": 1.6135050713937895e-05, + "loss": 0.9942, + "step": 25258 + }, + { + "epoch": 1.87, + "learning_rate": 1.6134735723817018e-05, + "loss": 1.0397, + "step": 25259 + }, + { + "epoch": 1.87, + "learning_rate": 1.6134420723935865e-05, + "loss": 1.0118, + "step": 25260 + }, + { + "epoch": 1.87, + "learning_rate": 1.6134105714294938e-05, + "loss": 0.9624, + "step": 25261 + }, + { + "epoch": 1.87, + "learning_rate": 1.6133790694894734e-05, + "loss": 1.0279, + "step": 25262 + }, + { + "epoch": 1.87, + "learning_rate": 1.613347566573576e-05, + "loss": 0.9711, + "step": 25263 + }, + { + "epoch": 1.87, + "learning_rate": 1.613316062681851e-05, + "loss": 0.9884, + "step": 25264 + }, + { + "epoch": 1.87, + "learning_rate": 1.6132845578143496e-05, + "loss": 1.0192, + "step": 25265 + }, + { + "epoch": 1.87, + "learning_rate": 1.6132530519711207e-05, + "loss": 1.0139, + "step": 25266 + }, + { + "epoch": 1.87, + "learning_rate": 1.613221545152215e-05, + "loss": 0.9292, + "step": 25267 + }, + { + "epoch": 1.87, + "learning_rate": 1.6131900373576828e-05, + "loss": 0.9437, + "step": 25268 + }, + { + "epoch": 1.87, + "learning_rate": 1.6131585285875736e-05, + "loss": 0.9435, + "step": 25269 + }, + { + "epoch": 1.87, + "learning_rate": 1.6131270188419386e-05, + "loss": 1.0451, + "step": 25270 + }, + { + "epoch": 1.87, + "learning_rate": 1.613095508120827e-05, + "loss": 0.9962, + "step": 25271 + }, + { + "epoch": 1.87, + "learning_rate": 1.6130639964242887e-05, + "loss": 0.9889, + "step": 25272 + }, + { + "epoch": 1.87, + "learning_rate": 1.613032483752375e-05, + "loss": 1.0378, + "step": 25273 + }, + { + "epoch": 1.87, + "learning_rate": 1.6130009701051353e-05, + "loss": 1.108, + "step": 25274 + }, + { + "epoch": 1.87, + "learning_rate": 1.6129694554826197e-05, + "loss": 0.986, + "step": 25275 + }, + { + "epoch": 1.87, + "learning_rate": 1.6129379398848785e-05, + "loss": 0.9517, + "step": 25276 + }, + { + "epoch": 1.87, + "learning_rate": 1.612906423311962e-05, + "loss": 1.0561, + "step": 25277 + }, + { + "epoch": 1.87, + "learning_rate": 1.61287490576392e-05, + "loss": 0.9901, + "step": 25278 + }, + { + "epoch": 1.87, + "learning_rate": 1.6128433872408028e-05, + "loss": 1.0222, + "step": 25279 + }, + { + "epoch": 1.87, + "learning_rate": 1.6128118677426602e-05, + "loss": 1.0333, + "step": 25280 + }, + { + "epoch": 1.87, + "learning_rate": 1.612780347269543e-05, + "loss": 1.0524, + "step": 25281 + }, + { + "epoch": 1.87, + "learning_rate": 1.612748825821501e-05, + "loss": 0.9868, + "step": 25282 + }, + { + "epoch": 1.87, + "learning_rate": 1.6127173033985845e-05, + "loss": 0.9919, + "step": 25283 + }, + { + "epoch": 1.87, + "learning_rate": 1.6126857800008435e-05, + "loss": 0.9775, + "step": 25284 + }, + { + "epoch": 1.87, + "learning_rate": 1.612654255628328e-05, + "loss": 0.9535, + "step": 25285 + }, + { + "epoch": 1.87, + "learning_rate": 1.6126227302810886e-05, + "loss": 1.0631, + "step": 25286 + }, + { + "epoch": 1.87, + "learning_rate": 1.612591203959175e-05, + "loss": 1.1212, + "step": 25287 + }, + { + "epoch": 1.87, + "learning_rate": 1.6125596766626376e-05, + "loss": 0.9981, + "step": 25288 + }, + { + "epoch": 1.87, + "learning_rate": 1.6125281483915263e-05, + "loss": 1.006, + "step": 25289 + }, + { + "epoch": 1.87, + "learning_rate": 1.6124966191458922e-05, + "loss": 1.0736, + "step": 25290 + }, + { + "epoch": 1.87, + "learning_rate": 1.612465088925784e-05, + "loss": 1.0108, + "step": 25291 + }, + { + "epoch": 1.87, + "learning_rate": 1.612433557731253e-05, + "loss": 1.0137, + "step": 25292 + }, + { + "epoch": 1.87, + "learning_rate": 1.6124020255623486e-05, + "loss": 0.9139, + "step": 25293 + }, + { + "epoch": 1.87, + "learning_rate": 1.612370492419122e-05, + "loss": 1.0414, + "step": 25294 + }, + { + "epoch": 1.87, + "learning_rate": 1.612338958301622e-05, + "loss": 1.0954, + "step": 25295 + }, + { + "epoch": 1.87, + "learning_rate": 1.6123074232099e-05, + "loss": 1.0777, + "step": 25296 + }, + { + "epoch": 1.87, + "learning_rate": 1.6122758871440055e-05, + "loss": 0.9902, + "step": 25297 + }, + { + "epoch": 1.87, + "learning_rate": 1.6122443501039888e-05, + "loss": 0.9398, + "step": 25298 + }, + { + "epoch": 1.87, + "learning_rate": 1.6122128120898997e-05, + "loss": 1.0078, + "step": 25299 + }, + { + "epoch": 1.87, + "learning_rate": 1.6121812731017895e-05, + "loss": 1.0683, + "step": 25300 + }, + { + "epoch": 1.87, + "learning_rate": 1.612149733139707e-05, + "loss": 1.0499, + "step": 25301 + }, + { + "epoch": 1.87, + "learning_rate": 1.6121181922037036e-05, + "loss": 0.9262, + "step": 25302 + }, + { + "epoch": 1.87, + "learning_rate": 1.6120866502938285e-05, + "loss": 0.9471, + "step": 25303 + }, + { + "epoch": 1.87, + "learning_rate": 1.6120551074101324e-05, + "loss": 1.0088, + "step": 25304 + }, + { + "epoch": 1.87, + "learning_rate": 1.6120235635526653e-05, + "loss": 0.9896, + "step": 25305 + }, + { + "epoch": 1.87, + "learning_rate": 1.6119920187214775e-05, + "loss": 0.9786, + "step": 25306 + }, + { + "epoch": 1.87, + "learning_rate": 1.6119604729166193e-05, + "loss": 1.0689, + "step": 25307 + }, + { + "epoch": 1.87, + "learning_rate": 1.6119289261381408e-05, + "loss": 1.0426, + "step": 25308 + }, + { + "epoch": 1.87, + "learning_rate": 1.6118973783860917e-05, + "loss": 1.0333, + "step": 25309 + }, + { + "epoch": 1.87, + "learning_rate": 1.6118658296605233e-05, + "loss": 1.0134, + "step": 25310 + }, + { + "epoch": 1.87, + "learning_rate": 1.6118342799614846e-05, + "loss": 1.008, + "step": 25311 + }, + { + "epoch": 1.87, + "learning_rate": 1.6118027292890266e-05, + "loss": 0.9241, + "step": 25312 + }, + { + "epoch": 1.87, + "learning_rate": 1.611771177643199e-05, + "loss": 1.0124, + "step": 25313 + }, + { + "epoch": 1.87, + "learning_rate": 1.611739625024052e-05, + "loss": 0.9857, + "step": 25314 + }, + { + "epoch": 1.87, + "learning_rate": 1.6117080714316363e-05, + "loss": 0.9917, + "step": 25315 + }, + { + "epoch": 1.87, + "learning_rate": 1.611676516866002e-05, + "loss": 1.0233, + "step": 25316 + }, + { + "epoch": 1.87, + "learning_rate": 1.611644961327199e-05, + "loss": 0.8941, + "step": 25317 + }, + { + "epoch": 1.87, + "learning_rate": 1.6116134048152776e-05, + "loss": 1.0598, + "step": 25318 + }, + { + "epoch": 1.87, + "learning_rate": 1.611581847330288e-05, + "loss": 0.9574, + "step": 25319 + }, + { + "epoch": 1.87, + "learning_rate": 1.61155028887228e-05, + "loss": 1.0514, + "step": 25320 + }, + { + "epoch": 1.87, + "learning_rate": 1.611518729441305e-05, + "loss": 1.0193, + "step": 25321 + }, + { + "epoch": 1.87, + "learning_rate": 1.611487169037412e-05, + "loss": 1.1277, + "step": 25322 + }, + { + "epoch": 1.87, + "learning_rate": 1.6114556076606516e-05, + "loss": 1.0195, + "step": 25323 + }, + { + "epoch": 1.87, + "learning_rate": 1.6114240453110743e-05, + "loss": 1.042, + "step": 25324 + }, + { + "epoch": 1.87, + "learning_rate": 1.61139248198873e-05, + "loss": 1.0863, + "step": 25325 + }, + { + "epoch": 1.87, + "learning_rate": 1.611360917693669e-05, + "loss": 1.01, + "step": 25326 + }, + { + "epoch": 1.87, + "learning_rate": 1.6113293524259416e-05, + "loss": 1.1015, + "step": 25327 + }, + { + "epoch": 1.87, + "learning_rate": 1.611297786185598e-05, + "loss": 1.0968, + "step": 25328 + }, + { + "epoch": 1.87, + "learning_rate": 1.611266218972688e-05, + "loss": 0.9671, + "step": 25329 + }, + { + "epoch": 1.87, + "learning_rate": 1.6112346507872624e-05, + "loss": 0.9713, + "step": 25330 + }, + { + "epoch": 1.87, + "learning_rate": 1.6112030816293715e-05, + "loss": 0.9655, + "step": 25331 + }, + { + "epoch": 1.87, + "learning_rate": 1.6111715114990646e-05, + "loss": 0.9751, + "step": 25332 + }, + { + "epoch": 1.87, + "learning_rate": 1.6111399403963935e-05, + "loss": 0.9202, + "step": 25333 + }, + { + "epoch": 1.87, + "learning_rate": 1.6111083683214067e-05, + "loss": 1.0719, + "step": 25334 + }, + { + "epoch": 1.87, + "learning_rate": 1.6110767952741556e-05, + "loss": 1.011, + "step": 25335 + }, + { + "epoch": 1.87, + "learning_rate": 1.6110452212546902e-05, + "loss": 0.8942, + "step": 25336 + }, + { + "epoch": 1.87, + "learning_rate": 1.61101364626306e-05, + "loss": 1.1466, + "step": 25337 + }, + { + "epoch": 1.87, + "learning_rate": 1.6109820702993162e-05, + "loss": 0.9825, + "step": 25338 + }, + { + "epoch": 1.87, + "learning_rate": 1.6109504933635086e-05, + "loss": 1.0097, + "step": 25339 + }, + { + "epoch": 1.87, + "learning_rate": 1.6109189154556877e-05, + "loss": 0.9867, + "step": 25340 + }, + { + "epoch": 1.87, + "learning_rate": 1.6108873365759035e-05, + "loss": 1.0554, + "step": 25341 + }, + { + "epoch": 1.87, + "learning_rate": 1.610855756724206e-05, + "loss": 1.1657, + "step": 25342 + }, + { + "epoch": 1.87, + "learning_rate": 1.6108241759006463e-05, + "loss": 1.0194, + "step": 25343 + }, + { + "epoch": 1.87, + "learning_rate": 1.6107925941052734e-05, + "loss": 0.9609, + "step": 25344 + }, + { + "epoch": 1.87, + "learning_rate": 1.6107610113381386e-05, + "loss": 1.0084, + "step": 25345 + }, + { + "epoch": 1.87, + "learning_rate": 1.6107294275992917e-05, + "loss": 1.0267, + "step": 25346 + }, + { + "epoch": 1.87, + "learning_rate": 1.610697842888783e-05, + "loss": 0.9593, + "step": 25347 + }, + { + "epoch": 1.87, + "learning_rate": 1.610666257206663e-05, + "loss": 1.0492, + "step": 25348 + }, + { + "epoch": 1.87, + "learning_rate": 1.6106346705529813e-05, + "loss": 1.0384, + "step": 25349 + }, + { + "epoch": 1.87, + "learning_rate": 1.610603082927789e-05, + "loss": 1.0856, + "step": 25350 + }, + { + "epoch": 1.87, + "learning_rate": 1.6105714943311356e-05, + "loss": 1.0708, + "step": 25351 + }, + { + "epoch": 1.87, + "learning_rate": 1.610539904763072e-05, + "loss": 1.0001, + "step": 25352 + }, + { + "epoch": 1.87, + "learning_rate": 1.610508314223648e-05, + "loss": 0.9738, + "step": 25353 + }, + { + "epoch": 1.87, + "learning_rate": 1.610476722712914e-05, + "loss": 0.9669, + "step": 25354 + }, + { + "epoch": 1.87, + "learning_rate": 1.61044513023092e-05, + "loss": 0.8699, + "step": 25355 + }, + { + "epoch": 1.87, + "learning_rate": 1.610413536777717e-05, + "loss": 0.945, + "step": 25356 + }, + { + "epoch": 1.87, + "learning_rate": 1.6103819423533545e-05, + "loss": 0.9499, + "step": 25357 + }, + { + "epoch": 1.87, + "learning_rate": 1.610350346957883e-05, + "loss": 1.0575, + "step": 25358 + }, + { + "epoch": 1.87, + "learning_rate": 1.6103187505913528e-05, + "loss": 1.0682, + "step": 25359 + }, + { + "epoch": 1.87, + "learning_rate": 1.6102871532538144e-05, + "loss": 1.0487, + "step": 25360 + }, + { + "epoch": 1.87, + "learning_rate": 1.6102555549453177e-05, + "loss": 0.9493, + "step": 25361 + }, + { + "epoch": 1.87, + "learning_rate": 1.6102239556659136e-05, + "loss": 1.0013, + "step": 25362 + }, + { + "epoch": 1.87, + "learning_rate": 1.6101923554156515e-05, + "loss": 1.0223, + "step": 25363 + }, + { + "epoch": 1.87, + "learning_rate": 1.6101607541945823e-05, + "loss": 0.9327, + "step": 25364 + }, + { + "epoch": 1.87, + "learning_rate": 1.610129152002756e-05, + "loss": 0.8759, + "step": 25365 + }, + { + "epoch": 1.87, + "learning_rate": 1.610097548840223e-05, + "loss": 0.9289, + "step": 25366 + }, + { + "epoch": 1.87, + "learning_rate": 1.6100659447070333e-05, + "loss": 0.956, + "step": 25367 + }, + { + "epoch": 1.87, + "learning_rate": 1.6100343396032375e-05, + "loss": 1.1066, + "step": 25368 + }, + { + "epoch": 1.87, + "learning_rate": 1.6100027335288857e-05, + "loss": 1.0449, + "step": 25369 + }, + { + "epoch": 1.87, + "learning_rate": 1.6099711264840285e-05, + "loss": 0.9244, + "step": 25370 + }, + { + "epoch": 1.87, + "learning_rate": 1.609939518468716e-05, + "loss": 0.9836, + "step": 25371 + }, + { + "epoch": 1.87, + "learning_rate": 1.6099079094829983e-05, + "loss": 0.9299, + "step": 25372 + }, + { + "epoch": 1.87, + "learning_rate": 1.6098762995269258e-05, + "loss": 1.1196, + "step": 25373 + }, + { + "epoch": 1.87, + "learning_rate": 1.6098446886005486e-05, + "loss": 0.9757, + "step": 25374 + }, + { + "epoch": 1.87, + "learning_rate": 1.6098130767039176e-05, + "loss": 1.0782, + "step": 25375 + }, + { + "epoch": 1.88, + "learning_rate": 1.6097814638370828e-05, + "loss": 0.7664, + "step": 25376 + }, + { + "epoch": 1.88, + "learning_rate": 1.6097498500000944e-05, + "loss": 1.0095, + "step": 25377 + }, + { + "epoch": 1.88, + "learning_rate": 1.6097182351930026e-05, + "loss": 0.8877, + "step": 25378 + }, + { + "epoch": 1.88, + "learning_rate": 1.6096866194158575e-05, + "loss": 1.0984, + "step": 25379 + }, + { + "epoch": 1.88, + "learning_rate": 1.6096550026687102e-05, + "loss": 1.0178, + "step": 25380 + }, + { + "epoch": 1.88, + "learning_rate": 1.6096233849516103e-05, + "loss": 0.9736, + "step": 25381 + }, + { + "epoch": 1.88, + "learning_rate": 1.6095917662646085e-05, + "loss": 0.9863, + "step": 25382 + }, + { + "epoch": 1.88, + "learning_rate": 1.6095601466077546e-05, + "loss": 0.8971, + "step": 25383 + }, + { + "epoch": 1.88, + "learning_rate": 1.6095285259810992e-05, + "loss": 1.0279, + "step": 25384 + }, + { + "epoch": 1.88, + "learning_rate": 1.6094969043846928e-05, + "loss": 1.0811, + "step": 25385 + }, + { + "epoch": 1.88, + "learning_rate": 1.6094652818185857e-05, + "loss": 0.9358, + "step": 25386 + }, + { + "epoch": 1.88, + "learning_rate": 1.6094336582828278e-05, + "loss": 0.9367, + "step": 25387 + }, + { + "epoch": 1.88, + "learning_rate": 1.60940203377747e-05, + "loss": 1.0921, + "step": 25388 + }, + { + "epoch": 1.88, + "learning_rate": 1.609370408302562e-05, + "loss": 0.9406, + "step": 25389 + }, + { + "epoch": 1.88, + "learning_rate": 1.6093387818581546e-05, + "loss": 1.1438, + "step": 25390 + }, + { + "epoch": 1.88, + "learning_rate": 1.6093071544442977e-05, + "loss": 0.9304, + "step": 25391 + }, + { + "epoch": 1.88, + "learning_rate": 1.6092755260610418e-05, + "loss": 1.0553, + "step": 25392 + }, + { + "epoch": 1.88, + "learning_rate": 1.6092438967084376e-05, + "loss": 1.0111, + "step": 25393 + }, + { + "epoch": 1.88, + "learning_rate": 1.6092122663865347e-05, + "loss": 1.0538, + "step": 25394 + }, + { + "epoch": 1.88, + "learning_rate": 1.6091806350953838e-05, + "loss": 1.0613, + "step": 25395 + }, + { + "epoch": 1.88, + "learning_rate": 1.6091490028350355e-05, + "loss": 0.9933, + "step": 25396 + }, + { + "epoch": 1.88, + "learning_rate": 1.6091173696055398e-05, + "loss": 0.9578, + "step": 25397 + }, + { + "epoch": 1.88, + "learning_rate": 1.609085735406947e-05, + "loss": 0.9591, + "step": 25398 + }, + { + "epoch": 1.88, + "learning_rate": 1.6090541002393075e-05, + "loss": 1.0219, + "step": 25399 + }, + { + "epoch": 1.88, + "learning_rate": 1.6090224641026718e-05, + "loss": 0.9451, + "step": 25400 + }, + { + "epoch": 1.88, + "learning_rate": 1.60899082699709e-05, + "loss": 0.9744, + "step": 25401 + }, + { + "epoch": 1.88, + "learning_rate": 1.6089591889226123e-05, + "loss": 0.9914, + "step": 25402 + }, + { + "epoch": 1.88, + "learning_rate": 1.6089275498792895e-05, + "loss": 1.0903, + "step": 25403 + }, + { + "epoch": 1.88, + "learning_rate": 1.6088959098671714e-05, + "loss": 1.0224, + "step": 25404 + }, + { + "epoch": 1.88, + "learning_rate": 1.608864268886309e-05, + "loss": 0.9849, + "step": 25405 + }, + { + "epoch": 1.88, + "learning_rate": 1.6088326269367517e-05, + "loss": 0.9644, + "step": 25406 + }, + { + "epoch": 1.88, + "learning_rate": 1.6088009840185506e-05, + "loss": 1.0392, + "step": 25407 + }, + { + "epoch": 1.88, + "learning_rate": 1.6087693401317562e-05, + "loss": 1.1068, + "step": 25408 + }, + { + "epoch": 1.88, + "learning_rate": 1.6087376952764183e-05, + "loss": 1.0569, + "step": 25409 + }, + { + "epoch": 1.88, + "learning_rate": 1.6087060494525876e-05, + "loss": 1.0593, + "step": 25410 + }, + { + "epoch": 1.88, + "learning_rate": 1.608674402660314e-05, + "loss": 1.0016, + "step": 25411 + }, + { + "epoch": 1.88, + "learning_rate": 1.6086427548996483e-05, + "loss": 1.0393, + "step": 25412 + }, + { + "epoch": 1.88, + "learning_rate": 1.6086111061706406e-05, + "loss": 1.089, + "step": 25413 + }, + { + "epoch": 1.88, + "learning_rate": 1.608579456473341e-05, + "loss": 0.9746, + "step": 25414 + }, + { + "epoch": 1.88, + "learning_rate": 1.6085478058078007e-05, + "loss": 0.9605, + "step": 25415 + }, + { + "epoch": 1.88, + "learning_rate": 1.6085161541740694e-05, + "loss": 0.9299, + "step": 25416 + }, + { + "epoch": 1.88, + "learning_rate": 1.608484501572198e-05, + "loss": 0.9639, + "step": 25417 + }, + { + "epoch": 1.88, + "learning_rate": 1.6084528480022358e-05, + "loss": 1.0044, + "step": 25418 + }, + { + "epoch": 1.88, + "learning_rate": 1.6084211934642342e-05, + "loss": 1.0388, + "step": 25419 + }, + { + "epoch": 1.88, + "learning_rate": 1.608389537958243e-05, + "loss": 0.9596, + "step": 25420 + }, + { + "epoch": 1.88, + "learning_rate": 1.608357881484313e-05, + "loss": 0.9201, + "step": 25421 + }, + { + "epoch": 1.88, + "learning_rate": 1.608326224042494e-05, + "loss": 0.9416, + "step": 25422 + }, + { + "epoch": 1.88, + "learning_rate": 1.608294565632837e-05, + "loss": 0.9832, + "step": 25423 + }, + { + "epoch": 1.88, + "learning_rate": 1.6082629062553915e-05, + "loss": 1.0068, + "step": 25424 + }, + { + "epoch": 1.88, + "learning_rate": 1.6082312459102093e-05, + "loss": 1.136, + "step": 25425 + }, + { + "epoch": 1.88, + "learning_rate": 1.608199584597339e-05, + "loss": 0.9541, + "step": 25426 + }, + { + "epoch": 1.88, + "learning_rate": 1.6081679223168324e-05, + "loss": 1.0032, + "step": 25427 + }, + { + "epoch": 1.88, + "learning_rate": 1.6081362590687392e-05, + "loss": 1.105, + "step": 25428 + }, + { + "epoch": 1.88, + "learning_rate": 1.60810459485311e-05, + "loss": 1.0676, + "step": 25429 + }, + { + "epoch": 1.88, + "learning_rate": 1.6080729296699947e-05, + "loss": 0.9048, + "step": 25430 + }, + { + "epoch": 1.88, + "learning_rate": 1.6080412635194444e-05, + "loss": 0.9124, + "step": 25431 + }, + { + "epoch": 1.88, + "learning_rate": 1.6080095964015094e-05, + "loss": 0.9853, + "step": 25432 + }, + { + "epoch": 1.88, + "learning_rate": 1.6079779283162393e-05, + "loss": 1.0724, + "step": 25433 + }, + { + "epoch": 1.88, + "learning_rate": 1.607946259263685e-05, + "loss": 1.0366, + "step": 25434 + }, + { + "epoch": 1.88, + "learning_rate": 1.6079145892438975e-05, + "loss": 0.9435, + "step": 25435 + }, + { + "epoch": 1.88, + "learning_rate": 1.6078829182569263e-05, + "loss": 1.1534, + "step": 25436 + }, + { + "epoch": 1.88, + "learning_rate": 1.607851246302822e-05, + "loss": 0.998, + "step": 25437 + }, + { + "epoch": 1.88, + "learning_rate": 1.607819573381635e-05, + "loss": 0.985, + "step": 25438 + }, + { + "epoch": 1.88, + "learning_rate": 1.607787899493416e-05, + "loss": 0.8514, + "step": 25439 + }, + { + "epoch": 1.88, + "learning_rate": 1.6077562246382148e-05, + "loss": 0.8695, + "step": 25440 + }, + { + "epoch": 1.88, + "learning_rate": 1.607724548816082e-05, + "loss": 0.9795, + "step": 25441 + }, + { + "epoch": 1.88, + "learning_rate": 1.6076928720270687e-05, + "loss": 0.9823, + "step": 25442 + }, + { + "epoch": 1.88, + "learning_rate": 1.6076611942712242e-05, + "loss": 1.0637, + "step": 25443 + }, + { + "epoch": 1.88, + "learning_rate": 1.6076295155486003e-05, + "loss": 0.9454, + "step": 25444 + }, + { + "epoch": 1.88, + "learning_rate": 1.607597835859246e-05, + "loss": 0.9692, + "step": 25445 + }, + { + "epoch": 1.88, + "learning_rate": 1.607566155203212e-05, + "loss": 0.9111, + "step": 25446 + }, + { + "epoch": 1.88, + "learning_rate": 1.6075344735805492e-05, + "loss": 0.7938, + "step": 25447 + }, + { + "epoch": 1.88, + "learning_rate": 1.6075027909913077e-05, + "loss": 0.952, + "step": 25448 + }, + { + "epoch": 1.88, + "learning_rate": 1.607471107435538e-05, + "loss": 1.0684, + "step": 25449 + }, + { + "epoch": 1.88, + "learning_rate": 1.6074394229132902e-05, + "loss": 1.0078, + "step": 25450 + }, + { + "epoch": 1.88, + "learning_rate": 1.6074077374246152e-05, + "loss": 1.016, + "step": 25451 + }, + { + "epoch": 1.88, + "learning_rate": 1.607376050969563e-05, + "loss": 0.9772, + "step": 25452 + }, + { + "epoch": 1.88, + "learning_rate": 1.6073443635481846e-05, + "loss": 1.0574, + "step": 25453 + }, + { + "epoch": 1.88, + "learning_rate": 1.6073126751605295e-05, + "loss": 0.9356, + "step": 25454 + }, + { + "epoch": 1.88, + "learning_rate": 1.607280985806649e-05, + "loss": 0.9899, + "step": 25455 + }, + { + "epoch": 1.88, + "learning_rate": 1.607249295486593e-05, + "loss": 1.0552, + "step": 25456 + }, + { + "epoch": 1.88, + "learning_rate": 1.6072176042004123e-05, + "loss": 0.9669, + "step": 25457 + }, + { + "epoch": 1.88, + "learning_rate": 1.6071859119481565e-05, + "loss": 0.8704, + "step": 25458 + }, + { + "epoch": 1.88, + "learning_rate": 1.607154218729877e-05, + "loss": 0.9625, + "step": 25459 + }, + { + "epoch": 1.88, + "learning_rate": 1.6071225245456238e-05, + "loss": 0.9969, + "step": 25460 + }, + { + "epoch": 1.88, + "learning_rate": 1.6070908293954474e-05, + "loss": 1.0445, + "step": 25461 + }, + { + "epoch": 1.88, + "learning_rate": 1.607059133279398e-05, + "loss": 0.9662, + "step": 25462 + }, + { + "epoch": 1.88, + "learning_rate": 1.607027436197526e-05, + "loss": 0.9738, + "step": 25463 + }, + { + "epoch": 1.88, + "learning_rate": 1.6069957381498828e-05, + "loss": 0.98, + "step": 25464 + }, + { + "epoch": 1.88, + "learning_rate": 1.6069640391365175e-05, + "loss": 1.0429, + "step": 25465 + }, + { + "epoch": 1.88, + "learning_rate": 1.6069323391574812e-05, + "loss": 1.0624, + "step": 25466 + }, + { + "epoch": 1.88, + "learning_rate": 1.6069006382128246e-05, + "loss": 0.9899, + "step": 25467 + }, + { + "epoch": 1.88, + "learning_rate": 1.6068689363025972e-05, + "loss": 1.0039, + "step": 25468 + }, + { + "epoch": 1.88, + "learning_rate": 1.6068372334268502e-05, + "loss": 0.9784, + "step": 25469 + }, + { + "epoch": 1.88, + "learning_rate": 1.6068055295856336e-05, + "loss": 1.0546, + "step": 25470 + }, + { + "epoch": 1.88, + "learning_rate": 1.6067738247789984e-05, + "loss": 0.9737, + "step": 25471 + }, + { + "epoch": 1.88, + "learning_rate": 1.6067421190069947e-05, + "loss": 0.9913, + "step": 25472 + }, + { + "epoch": 1.88, + "learning_rate": 1.6067104122696732e-05, + "loss": 1.1268, + "step": 25473 + }, + { + "epoch": 1.88, + "learning_rate": 1.6066787045670835e-05, + "loss": 1.0869, + "step": 25474 + }, + { + "epoch": 1.88, + "learning_rate": 1.606646995899277e-05, + "loss": 0.986, + "step": 25475 + }, + { + "epoch": 1.88, + "learning_rate": 1.6066152862663037e-05, + "loss": 0.9054, + "step": 25476 + }, + { + "epoch": 1.88, + "learning_rate": 1.6065835756682145e-05, + "loss": 0.9807, + "step": 25477 + }, + { + "epoch": 1.88, + "learning_rate": 1.6065518641050594e-05, + "loss": 0.9866, + "step": 25478 + }, + { + "epoch": 1.88, + "learning_rate": 1.6065201515768888e-05, + "loss": 0.9893, + "step": 25479 + }, + { + "epoch": 1.88, + "learning_rate": 1.606488438083753e-05, + "loss": 0.9984, + "step": 25480 + }, + { + "epoch": 1.88, + "learning_rate": 1.6064567236257033e-05, + "loss": 1.0216, + "step": 25481 + }, + { + "epoch": 1.88, + "learning_rate": 1.6064250082027895e-05, + "loss": 0.979, + "step": 25482 + }, + { + "epoch": 1.88, + "learning_rate": 1.606393291815062e-05, + "loss": 1.0776, + "step": 25483 + }, + { + "epoch": 1.88, + "learning_rate": 1.606361574462572e-05, + "loss": 1.0515, + "step": 25484 + }, + { + "epoch": 1.88, + "learning_rate": 1.6063298561453692e-05, + "loss": 0.9464, + "step": 25485 + }, + { + "epoch": 1.88, + "learning_rate": 1.6062981368635037e-05, + "loss": 0.9999, + "step": 25486 + }, + { + "epoch": 1.88, + "learning_rate": 1.606266416617027e-05, + "loss": 0.9765, + "step": 25487 + }, + { + "epoch": 1.88, + "learning_rate": 1.606234695405989e-05, + "loss": 1.1051, + "step": 25488 + }, + { + "epoch": 1.88, + "learning_rate": 1.6062029732304404e-05, + "loss": 0.9199, + "step": 25489 + }, + { + "epoch": 1.88, + "learning_rate": 1.6061712500904313e-05, + "loss": 0.9684, + "step": 25490 + }, + { + "epoch": 1.88, + "learning_rate": 1.6061395259860128e-05, + "loss": 1.0068, + "step": 25491 + }, + { + "epoch": 1.88, + "learning_rate": 1.6061078009172345e-05, + "loss": 0.894, + "step": 25492 + }, + { + "epoch": 1.88, + "learning_rate": 1.6060760748841477e-05, + "loss": 0.9996, + "step": 25493 + }, + { + "epoch": 1.88, + "learning_rate": 1.6060443478868024e-05, + "loss": 1.0199, + "step": 25494 + }, + { + "epoch": 1.88, + "learning_rate": 1.6060126199252493e-05, + "loss": 1.0492, + "step": 25495 + }, + { + "epoch": 1.88, + "learning_rate": 1.605980890999539e-05, + "loss": 1.0627, + "step": 25496 + }, + { + "epoch": 1.88, + "learning_rate": 1.6059491611097212e-05, + "loss": 0.9921, + "step": 25497 + }, + { + "epoch": 1.88, + "learning_rate": 1.6059174302558474e-05, + "loss": 1.1021, + "step": 25498 + }, + { + "epoch": 1.88, + "learning_rate": 1.6058856984379674e-05, + "loss": 1.0953, + "step": 25499 + }, + { + "epoch": 1.88, + "learning_rate": 1.605853965656132e-05, + "loss": 0.9476, + "step": 25500 + }, + { + "epoch": 1.88, + "learning_rate": 1.605822231910392e-05, + "loss": 1.1383, + "step": 25501 + }, + { + "epoch": 1.88, + "learning_rate": 1.6057904972007968e-05, + "loss": 1.0654, + "step": 25502 + }, + { + "epoch": 1.88, + "learning_rate": 1.605758761527398e-05, + "loss": 1.003, + "step": 25503 + }, + { + "epoch": 1.88, + "learning_rate": 1.605727024890246e-05, + "loss": 1.0468, + "step": 25504 + }, + { + "epoch": 1.88, + "learning_rate": 1.6056952872893904e-05, + "loss": 0.9546, + "step": 25505 + }, + { + "epoch": 1.88, + "learning_rate": 1.6056635487248825e-05, + "loss": 1.0175, + "step": 25506 + }, + { + "epoch": 1.88, + "learning_rate": 1.6056318091967724e-05, + "loss": 0.9603, + "step": 25507 + }, + { + "epoch": 1.88, + "learning_rate": 1.605600068705111e-05, + "loss": 1.1331, + "step": 25508 + }, + { + "epoch": 1.88, + "learning_rate": 1.6055683272499484e-05, + "loss": 0.9177, + "step": 25509 + }, + { + "epoch": 1.88, + "learning_rate": 1.6055365848313352e-05, + "loss": 1.0447, + "step": 25510 + }, + { + "epoch": 1.88, + "learning_rate": 1.605504841449322e-05, + "loss": 1.047, + "step": 25511 + }, + { + "epoch": 1.89, + "learning_rate": 1.6054730971039593e-05, + "loss": 1.0127, + "step": 25512 + }, + { + "epoch": 1.89, + "learning_rate": 1.6054413517952975e-05, + "loss": 0.9969, + "step": 25513 + }, + { + "epoch": 1.89, + "learning_rate": 1.6054096055233875e-05, + "loss": 1.0117, + "step": 25514 + }, + { + "epoch": 1.89, + "learning_rate": 1.605377858288279e-05, + "loss": 0.9665, + "step": 25515 + }, + { + "epoch": 1.89, + "learning_rate": 1.6053461100900233e-05, + "loss": 1.1236, + "step": 25516 + }, + { + "epoch": 1.89, + "learning_rate": 1.6053143609286704e-05, + "loss": 1.0713, + "step": 25517 + }, + { + "epoch": 1.89, + "learning_rate": 1.6052826108042713e-05, + "loss": 0.9898, + "step": 25518 + }, + { + "epoch": 1.89, + "learning_rate": 1.605250859716876e-05, + "loss": 1.057, + "step": 25519 + }, + { + "epoch": 1.89, + "learning_rate": 1.6052191076665352e-05, + "loss": 0.9827, + "step": 25520 + }, + { + "epoch": 1.89, + "learning_rate": 1.6051873546532994e-05, + "loss": 1.0869, + "step": 25521 + }, + { + "epoch": 1.89, + "learning_rate": 1.6051556006772195e-05, + "loss": 1.0351, + "step": 25522 + }, + { + "epoch": 1.89, + "learning_rate": 1.6051238457383453e-05, + "loss": 0.9846, + "step": 25523 + }, + { + "epoch": 1.89, + "learning_rate": 1.6050920898367278e-05, + "loss": 1.0272, + "step": 25524 + }, + { + "epoch": 1.89, + "learning_rate": 1.6050603329724177e-05, + "loss": 1.009, + "step": 25525 + }, + { + "epoch": 1.89, + "learning_rate": 1.605028575145465e-05, + "loss": 1.069, + "step": 25526 + }, + { + "epoch": 1.89, + "learning_rate": 1.6049968163559206e-05, + "loss": 0.995, + "step": 25527 + }, + { + "epoch": 1.89, + "learning_rate": 1.6049650566038352e-05, + "loss": 1.0858, + "step": 25528 + }, + { + "epoch": 1.89, + "learning_rate": 1.604933295889259e-05, + "loss": 1.038, + "step": 25529 + }, + { + "epoch": 1.89, + "learning_rate": 1.6049015342122424e-05, + "loss": 1.0173, + "step": 25530 + }, + { + "epoch": 1.89, + "learning_rate": 1.604869771572836e-05, + "loss": 0.9603, + "step": 25531 + }, + { + "epoch": 1.89, + "learning_rate": 1.6048380079710905e-05, + "loss": 1.0186, + "step": 25532 + }, + { + "epoch": 1.89, + "learning_rate": 1.6048062434070564e-05, + "loss": 1.0566, + "step": 25533 + }, + { + "epoch": 1.89, + "learning_rate": 1.6047744778807844e-05, + "loss": 0.9855, + "step": 25534 + }, + { + "epoch": 1.89, + "learning_rate": 1.6047427113923247e-05, + "loss": 1.0312, + "step": 25535 + }, + { + "epoch": 1.89, + "learning_rate": 1.604710943941728e-05, + "loss": 1.0408, + "step": 25536 + }, + { + "epoch": 1.89, + "learning_rate": 1.604679175529045e-05, + "loss": 1.0045, + "step": 25537 + }, + { + "epoch": 1.89, + "learning_rate": 1.6046474061543254e-05, + "loss": 1.0891, + "step": 25538 + }, + { + "epoch": 1.89, + "learning_rate": 1.6046156358176212e-05, + "loss": 0.9901, + "step": 25539 + }, + { + "epoch": 1.89, + "learning_rate": 1.604583864518982e-05, + "loss": 0.9572, + "step": 25540 + }, + { + "epoch": 1.89, + "learning_rate": 1.6045520922584582e-05, + "loss": 1.041, + "step": 25541 + }, + { + "epoch": 1.89, + "learning_rate": 1.604520319036101e-05, + "loss": 1.0784, + "step": 25542 + }, + { + "epoch": 1.89, + "learning_rate": 1.6044885448519604e-05, + "loss": 0.9991, + "step": 25543 + }, + { + "epoch": 1.89, + "learning_rate": 1.6044567697060877e-05, + "loss": 1.0011, + "step": 25544 + }, + { + "epoch": 1.89, + "learning_rate": 1.6044249935985324e-05, + "loss": 1.0031, + "step": 25545 + }, + { + "epoch": 1.89, + "learning_rate": 1.6043932165293457e-05, + "loss": 1.0082, + "step": 25546 + }, + { + "epoch": 1.89, + "learning_rate": 1.604361438498578e-05, + "loss": 0.8597, + "step": 25547 + }, + { + "epoch": 1.89, + "learning_rate": 1.6043296595062802e-05, + "loss": 1.0415, + "step": 25548 + }, + { + "epoch": 1.89, + "learning_rate": 1.6042978795525025e-05, + "loss": 0.9545, + "step": 25549 + }, + { + "epoch": 1.89, + "learning_rate": 1.6042660986372952e-05, + "loss": 1.0938, + "step": 25550 + }, + { + "epoch": 1.89, + "learning_rate": 1.6042343167607093e-05, + "loss": 0.9593, + "step": 25551 + }, + { + "epoch": 1.89, + "learning_rate": 1.6042025339227952e-05, + "loss": 0.9699, + "step": 25552 + }, + { + "epoch": 1.89, + "learning_rate": 1.6041707501236038e-05, + "loss": 0.9864, + "step": 25553 + }, + { + "epoch": 1.89, + "learning_rate": 1.6041389653631848e-05, + "loss": 1.1086, + "step": 25554 + }, + { + "epoch": 1.89, + "learning_rate": 1.6041071796415897e-05, + "loss": 1.0834, + "step": 25555 + }, + { + "epoch": 1.89, + "learning_rate": 1.6040753929588687e-05, + "loss": 1.0454, + "step": 25556 + }, + { + "epoch": 1.89, + "learning_rate": 1.6040436053150726e-05, + "loss": 0.9482, + "step": 25557 + }, + { + "epoch": 1.89, + "learning_rate": 1.604011816710252e-05, + "loss": 1.0131, + "step": 25558 + }, + { + "epoch": 1.89, + "learning_rate": 1.6039800271444564e-05, + "loss": 1.0056, + "step": 25559 + }, + { + "epoch": 1.89, + "learning_rate": 1.603948236617738e-05, + "loss": 0.9338, + "step": 25560 + }, + { + "epoch": 1.89, + "learning_rate": 1.603916445130146e-05, + "loss": 1.1199, + "step": 25561 + }, + { + "epoch": 1.89, + "learning_rate": 1.6038846526817317e-05, + "loss": 1.0185, + "step": 25562 + }, + { + "epoch": 1.89, + "learning_rate": 1.603852859272546e-05, + "loss": 1.0911, + "step": 25563 + }, + { + "epoch": 1.89, + "learning_rate": 1.6038210649026384e-05, + "loss": 0.966, + "step": 25564 + }, + { + "epoch": 1.89, + "learning_rate": 1.6037892695720607e-05, + "loss": 1.0148, + "step": 25565 + }, + { + "epoch": 1.89, + "learning_rate": 1.6037574732808626e-05, + "loss": 1.0262, + "step": 25566 + }, + { + "epoch": 1.89, + "learning_rate": 1.6037256760290952e-05, + "loss": 1.0283, + "step": 25567 + }, + { + "epoch": 1.89, + "learning_rate": 1.6036938778168084e-05, + "loss": 0.9646, + "step": 25568 + }, + { + "epoch": 1.89, + "learning_rate": 1.603662078644054e-05, + "loss": 1.0805, + "step": 25569 + }, + { + "epoch": 1.89, + "learning_rate": 1.6036302785108814e-05, + "loss": 1.0271, + "step": 25570 + }, + { + "epoch": 1.89, + "learning_rate": 1.6035984774173416e-05, + "loss": 0.9756, + "step": 25571 + }, + { + "epoch": 1.89, + "learning_rate": 1.6035666753634856e-05, + "loss": 0.9969, + "step": 25572 + }, + { + "epoch": 1.89, + "learning_rate": 1.6035348723493632e-05, + "loss": 0.9731, + "step": 25573 + }, + { + "epoch": 1.89, + "learning_rate": 1.603503068375026e-05, + "loss": 0.988, + "step": 25574 + }, + { + "epoch": 1.89, + "learning_rate": 1.6034712634405235e-05, + "loss": 1.1356, + "step": 25575 + }, + { + "epoch": 1.89, + "learning_rate": 1.6034394575459074e-05, + "loss": 0.9662, + "step": 25576 + }, + { + "epoch": 1.89, + "learning_rate": 1.6034076506912275e-05, + "loss": 0.9696, + "step": 25577 + }, + { + "epoch": 1.89, + "learning_rate": 1.6033758428765348e-05, + "loss": 1.0325, + "step": 25578 + }, + { + "epoch": 1.89, + "learning_rate": 1.6033440341018796e-05, + "loss": 0.9316, + "step": 25579 + }, + { + "epoch": 1.89, + "learning_rate": 1.6033122243673128e-05, + "loss": 1.0413, + "step": 25580 + }, + { + "epoch": 1.89, + "learning_rate": 1.6032804136728845e-05, + "loss": 1.0153, + "step": 25581 + }, + { + "epoch": 1.89, + "learning_rate": 1.603248602018646e-05, + "loss": 0.9998, + "step": 25582 + }, + { + "epoch": 1.89, + "learning_rate": 1.6032167894046476e-05, + "loss": 0.8985, + "step": 25583 + }, + { + "epoch": 1.89, + "learning_rate": 1.60318497583094e-05, + "loss": 0.986, + "step": 25584 + }, + { + "epoch": 1.89, + "learning_rate": 1.6031531612975733e-05, + "loss": 1.0291, + "step": 25585 + }, + { + "epoch": 1.89, + "learning_rate": 1.603121345804599e-05, + "loss": 1.0094, + "step": 25586 + }, + { + "epoch": 1.89, + "learning_rate": 1.6030895293520673e-05, + "loss": 0.9481, + "step": 25587 + }, + { + "epoch": 1.89, + "learning_rate": 1.6030577119400284e-05, + "loss": 0.8777, + "step": 25588 + }, + { + "epoch": 1.89, + "learning_rate": 1.6030258935685336e-05, + "loss": 0.9278, + "step": 25589 + }, + { + "epoch": 1.89, + "learning_rate": 1.6029940742376334e-05, + "loss": 1.071, + "step": 25590 + }, + { + "epoch": 1.89, + "learning_rate": 1.602962253947378e-05, + "loss": 1.0877, + "step": 25591 + }, + { + "epoch": 1.89, + "learning_rate": 1.6029304326978184e-05, + "loss": 0.9513, + "step": 25592 + }, + { + "epoch": 1.89, + "learning_rate": 1.602898610489005e-05, + "loss": 0.9554, + "step": 25593 + }, + { + "epoch": 1.89, + "learning_rate": 1.6028667873209885e-05, + "loss": 1.0509, + "step": 25594 + }, + { + "epoch": 1.89, + "learning_rate": 1.6028349631938195e-05, + "loss": 1.1038, + "step": 25595 + }, + { + "epoch": 1.89, + "learning_rate": 1.602803138107549e-05, + "loss": 1.0172, + "step": 25596 + }, + { + "epoch": 1.89, + "learning_rate": 1.6027713120622274e-05, + "loss": 1.0957, + "step": 25597 + }, + { + "epoch": 1.89, + "learning_rate": 1.602739485057905e-05, + "loss": 1.0927, + "step": 25598 + }, + { + "epoch": 1.89, + "learning_rate": 1.6027076570946324e-05, + "loss": 1.0314, + "step": 25599 + }, + { + "epoch": 1.89, + "learning_rate": 1.602675828172461e-05, + "loss": 0.9387, + "step": 25600 + }, + { + "epoch": 1.89, + "learning_rate": 1.602643998291441e-05, + "loss": 1.0431, + "step": 25601 + }, + { + "epoch": 1.89, + "learning_rate": 1.6026121674516225e-05, + "loss": 0.9708, + "step": 25602 + }, + { + "epoch": 1.89, + "learning_rate": 1.602580335653057e-05, + "loss": 1.0776, + "step": 25603 + }, + { + "epoch": 1.89, + "learning_rate": 1.6025485028957952e-05, + "loss": 1.119, + "step": 25604 + }, + { + "epoch": 1.89, + "learning_rate": 1.6025166691798866e-05, + "loss": 0.9776, + "step": 25605 + }, + { + "epoch": 1.89, + "learning_rate": 1.602484834505383e-05, + "loss": 1.0726, + "step": 25606 + }, + { + "epoch": 1.89, + "learning_rate": 1.602452998872335e-05, + "loss": 0.9867, + "step": 25607 + }, + { + "epoch": 1.89, + "learning_rate": 1.6024211622807925e-05, + "loss": 1.0962, + "step": 25608 + }, + { + "epoch": 1.89, + "learning_rate": 1.6023893247308066e-05, + "loss": 1.0068, + "step": 25609 + }, + { + "epoch": 1.89, + "learning_rate": 1.602357486222428e-05, + "loss": 1.0614, + "step": 25610 + }, + { + "epoch": 1.89, + "learning_rate": 1.602325646755707e-05, + "loss": 0.9639, + "step": 25611 + }, + { + "epoch": 1.89, + "learning_rate": 1.6022938063306947e-05, + "loss": 0.9319, + "step": 25612 + }, + { + "epoch": 1.89, + "learning_rate": 1.6022619649474415e-05, + "loss": 0.9794, + "step": 25613 + }, + { + "epoch": 1.89, + "learning_rate": 1.602230122605998e-05, + "loss": 1.0621, + "step": 25614 + }, + { + "epoch": 1.89, + "learning_rate": 1.6021982793064156e-05, + "loss": 0.9854, + "step": 25615 + }, + { + "epoch": 1.89, + "learning_rate": 1.602166435048744e-05, + "loss": 0.9714, + "step": 25616 + }, + { + "epoch": 1.89, + "learning_rate": 1.6021345898330338e-05, + "loss": 1.0341, + "step": 25617 + }, + { + "epoch": 1.89, + "learning_rate": 1.6021027436593363e-05, + "loss": 0.9984, + "step": 25618 + }, + { + "epoch": 1.89, + "learning_rate": 1.6020708965277025e-05, + "loss": 1.0262, + "step": 25619 + }, + { + "epoch": 1.89, + "learning_rate": 1.6020390484381814e-05, + "loss": 1.124, + "step": 25620 + }, + { + "epoch": 1.89, + "learning_rate": 1.6020071993908257e-05, + "loss": 0.9589, + "step": 25621 + }, + { + "epoch": 1.89, + "learning_rate": 1.6019753493856848e-05, + "loss": 0.9704, + "step": 25622 + }, + { + "epoch": 1.89, + "learning_rate": 1.60194349842281e-05, + "loss": 1.0459, + "step": 25623 + }, + { + "epoch": 1.89, + "learning_rate": 1.601911646502252e-05, + "loss": 1.0345, + "step": 25624 + }, + { + "epoch": 1.89, + "learning_rate": 1.6018797936240606e-05, + "loss": 1.0038, + "step": 25625 + }, + { + "epoch": 1.89, + "learning_rate": 1.601847939788287e-05, + "loss": 1.0515, + "step": 25626 + }, + { + "epoch": 1.89, + "learning_rate": 1.6018160849949823e-05, + "loss": 0.989, + "step": 25627 + }, + { + "epoch": 1.89, + "learning_rate": 1.6017842292441966e-05, + "loss": 0.8969, + "step": 25628 + }, + { + "epoch": 1.89, + "learning_rate": 1.601752372535981e-05, + "loss": 1.0073, + "step": 25629 + }, + { + "epoch": 1.89, + "learning_rate": 1.6017205148703857e-05, + "loss": 0.9604, + "step": 25630 + }, + { + "epoch": 1.89, + "learning_rate": 1.6016886562474615e-05, + "loss": 0.9476, + "step": 25631 + }, + { + "epoch": 1.89, + "learning_rate": 1.6016567966672602e-05, + "loss": 0.965, + "step": 25632 + }, + { + "epoch": 1.89, + "learning_rate": 1.6016249361298307e-05, + "loss": 1.0849, + "step": 25633 + }, + { + "epoch": 1.89, + "learning_rate": 1.6015930746352245e-05, + "loss": 0.9272, + "step": 25634 + }, + { + "epoch": 1.89, + "learning_rate": 1.6015612121834928e-05, + "loss": 1.0053, + "step": 25635 + }, + { + "epoch": 1.89, + "learning_rate": 1.6015293487746853e-05, + "loss": 0.9622, + "step": 25636 + }, + { + "epoch": 1.89, + "learning_rate": 1.601497484408854e-05, + "loss": 1.0236, + "step": 25637 + }, + { + "epoch": 1.89, + "learning_rate": 1.601465619086048e-05, + "loss": 1.0289, + "step": 25638 + }, + { + "epoch": 1.89, + "learning_rate": 1.6014337528063194e-05, + "loss": 1.0317, + "step": 25639 + }, + { + "epoch": 1.89, + "learning_rate": 1.6014018855697178e-05, + "loss": 0.986, + "step": 25640 + }, + { + "epoch": 1.89, + "learning_rate": 1.6013700173762946e-05, + "loss": 0.9957, + "step": 25641 + }, + { + "epoch": 1.89, + "learning_rate": 1.6013381482261007e-05, + "loss": 1.0774, + "step": 25642 + }, + { + "epoch": 1.89, + "learning_rate": 1.6013062781191857e-05, + "loss": 0.9448, + "step": 25643 + }, + { + "epoch": 1.89, + "learning_rate": 1.6012744070556015e-05, + "loss": 1.0268, + "step": 25644 + }, + { + "epoch": 1.89, + "learning_rate": 1.601242535035398e-05, + "loss": 1.0066, + "step": 25645 + }, + { + "epoch": 1.89, + "learning_rate": 1.6012106620586267e-05, + "loss": 0.9213, + "step": 25646 + }, + { + "epoch": 1.9, + "learning_rate": 1.6011787881253376e-05, + "loss": 1.0601, + "step": 25647 + }, + { + "epoch": 1.9, + "learning_rate": 1.6011469132355816e-05, + "loss": 0.9944, + "step": 25648 + }, + { + "epoch": 1.9, + "learning_rate": 1.6011150373894092e-05, + "loss": 1.1394, + "step": 25649 + }, + { + "epoch": 1.9, + "learning_rate": 1.6010831605868716e-05, + "loss": 1.0519, + "step": 25650 + }, + { + "epoch": 1.9, + "learning_rate": 1.6010512828280195e-05, + "loss": 0.9911, + "step": 25651 + }, + { + "epoch": 1.9, + "learning_rate": 1.601019404112903e-05, + "loss": 0.9206, + "step": 25652 + }, + { + "epoch": 1.9, + "learning_rate": 1.6009875244415737e-05, + "loss": 1.0722, + "step": 25653 + }, + { + "epoch": 1.9, + "learning_rate": 1.6009556438140816e-05, + "loss": 0.9885, + "step": 25654 + }, + { + "epoch": 1.9, + "learning_rate": 1.600923762230478e-05, + "loss": 1.0416, + "step": 25655 + }, + { + "epoch": 1.9, + "learning_rate": 1.6008918796908127e-05, + "loss": 1.0042, + "step": 25656 + }, + { + "epoch": 1.9, + "learning_rate": 1.6008599961951373e-05, + "loss": 1.0023, + "step": 25657 + }, + { + "epoch": 1.9, + "learning_rate": 1.600828111743502e-05, + "loss": 1.0758, + "step": 25658 + }, + { + "epoch": 1.9, + "learning_rate": 1.600796226335958e-05, + "loss": 1.2027, + "step": 25659 + }, + { + "epoch": 1.9, + "learning_rate": 1.600764339972556e-05, + "loss": 1.1004, + "step": 25660 + }, + { + "epoch": 1.9, + "learning_rate": 1.6007324526533465e-05, + "loss": 0.9188, + "step": 25661 + }, + { + "epoch": 1.9, + "learning_rate": 1.6007005643783798e-05, + "loss": 1.0722, + "step": 25662 + }, + { + "epoch": 1.9, + "learning_rate": 1.6006686751477076e-05, + "loss": 1.0313, + "step": 25663 + }, + { + "epoch": 1.9, + "learning_rate": 1.60063678496138e-05, + "loss": 0.9906, + "step": 25664 + }, + { + "epoch": 1.9, + "learning_rate": 1.6006048938194475e-05, + "loss": 0.9818, + "step": 25665 + }, + { + "epoch": 1.9, + "learning_rate": 1.6005730017219618e-05, + "loss": 0.9484, + "step": 25666 + }, + { + "epoch": 1.9, + "learning_rate": 1.6005411086689727e-05, + "loss": 0.9594, + "step": 25667 + }, + { + "epoch": 1.9, + "learning_rate": 1.6005092146605315e-05, + "loss": 0.9052, + "step": 25668 + }, + { + "epoch": 1.9, + "learning_rate": 1.6004773196966886e-05, + "loss": 0.9453, + "step": 25669 + }, + { + "epoch": 1.9, + "learning_rate": 1.6004454237774945e-05, + "loss": 0.968, + "step": 25670 + }, + { + "epoch": 1.9, + "learning_rate": 1.600413526903001e-05, + "loss": 1.1254, + "step": 25671 + }, + { + "epoch": 1.9, + "learning_rate": 1.6003816290732578e-05, + "loss": 0.995, + "step": 25672 + }, + { + "epoch": 1.9, + "learning_rate": 1.600349730288316e-05, + "loss": 0.9699, + "step": 25673 + }, + { + "epoch": 1.9, + "learning_rate": 1.6003178305482262e-05, + "loss": 0.9689, + "step": 25674 + }, + { + "epoch": 1.9, + "learning_rate": 1.6002859298530396e-05, + "loss": 1.0269, + "step": 25675 + }, + { + "epoch": 1.9, + "learning_rate": 1.6002540282028067e-05, + "loss": 1.0441, + "step": 25676 + }, + { + "epoch": 1.9, + "learning_rate": 1.6002221255975784e-05, + "loss": 1.0576, + "step": 25677 + }, + { + "epoch": 1.9, + "learning_rate": 1.6001902220374048e-05, + "loss": 0.906, + "step": 25678 + }, + { + "epoch": 1.9, + "learning_rate": 1.6001583175223374e-05, + "loss": 1.0394, + "step": 25679 + }, + { + "epoch": 1.9, + "learning_rate": 1.6001264120524268e-05, + "loss": 0.991, + "step": 25680 + }, + { + "epoch": 1.9, + "learning_rate": 1.6000945056277236e-05, + "loss": 1.058, + "step": 25681 + }, + { + "epoch": 1.9, + "learning_rate": 1.6000625982482784e-05, + "loss": 0.9368, + "step": 25682 + }, + { + "epoch": 1.9, + "learning_rate": 1.6000306899141425e-05, + "loss": 0.854, + "step": 25683 + }, + { + "epoch": 1.9, + "learning_rate": 1.599998780625366e-05, + "loss": 1.0273, + "step": 25684 + }, + { + "epoch": 1.9, + "learning_rate": 1.5999668703820002e-05, + "loss": 0.9145, + "step": 25685 + }, + { + "epoch": 1.9, + "learning_rate": 1.599934959184096e-05, + "loss": 1.0723, + "step": 25686 + }, + { + "epoch": 1.9, + "learning_rate": 1.5999030470317035e-05, + "loss": 1.0152, + "step": 25687 + }, + { + "epoch": 1.9, + "learning_rate": 1.5998711339248743e-05, + "loss": 1.0682, + "step": 25688 + }, + { + "epoch": 1.9, + "learning_rate": 1.5998392198636582e-05, + "loss": 1.0369, + "step": 25689 + }, + { + "epoch": 1.9, + "learning_rate": 1.5998073048481067e-05, + "loss": 0.9955, + "step": 25690 + }, + { + "epoch": 1.9, + "learning_rate": 1.5997753888782702e-05, + "loss": 1.1077, + "step": 25691 + }, + { + "epoch": 1.9, + "learning_rate": 1.5997434719541997e-05, + "loss": 1.048, + "step": 25692 + }, + { + "epoch": 1.9, + "learning_rate": 1.5997115540759458e-05, + "loss": 0.9213, + "step": 25693 + }, + { + "epoch": 1.9, + "learning_rate": 1.5996796352435598e-05, + "loss": 1.0936, + "step": 25694 + }, + { + "epoch": 1.9, + "learning_rate": 1.5996477154570918e-05, + "loss": 0.9612, + "step": 25695 + }, + { + "epoch": 1.9, + "learning_rate": 1.5996157947165926e-05, + "loss": 1.0396, + "step": 25696 + }, + { + "epoch": 1.9, + "learning_rate": 1.5995838730221138e-05, + "loss": 0.8837, + "step": 25697 + }, + { + "epoch": 1.9, + "learning_rate": 1.5995519503737055e-05, + "loss": 0.9686, + "step": 25698 + }, + { + "epoch": 1.9, + "learning_rate": 1.5995200267714188e-05, + "loss": 0.8522, + "step": 25699 + }, + { + "epoch": 1.9, + "learning_rate": 1.599488102215304e-05, + "loss": 1.0004, + "step": 25700 + }, + { + "epoch": 1.9, + "learning_rate": 1.599456176705412e-05, + "loss": 1.0686, + "step": 25701 + }, + { + "epoch": 1.9, + "learning_rate": 1.599424250241794e-05, + "loss": 1.0457, + "step": 25702 + }, + { + "epoch": 1.9, + "learning_rate": 1.5993923228245004e-05, + "loss": 0.8777, + "step": 25703 + }, + { + "epoch": 1.9, + "learning_rate": 1.5993603944535825e-05, + "loss": 0.9818, + "step": 25704 + }, + { + "epoch": 1.9, + "learning_rate": 1.599328465129091e-05, + "loss": 1.0424, + "step": 25705 + }, + { + "epoch": 1.9, + "learning_rate": 1.599296534851076e-05, + "loss": 1.076, + "step": 25706 + }, + { + "epoch": 1.9, + "learning_rate": 1.599264603619589e-05, + "loss": 1.0849, + "step": 25707 + }, + { + "epoch": 1.9, + "learning_rate": 1.5992326714346805e-05, + "loss": 1.0371, + "step": 25708 + }, + { + "epoch": 1.9, + "learning_rate": 1.5992007382964014e-05, + "loss": 1.0035, + "step": 25709 + }, + { + "epoch": 1.9, + "learning_rate": 1.5991688042048025e-05, + "loss": 1.0531, + "step": 25710 + }, + { + "epoch": 1.9, + "learning_rate": 1.599136869159935e-05, + "loss": 1.0915, + "step": 25711 + }, + { + "epoch": 1.9, + "learning_rate": 1.5991049331618485e-05, + "loss": 0.9721, + "step": 25712 + }, + { + "epoch": 1.9, + "learning_rate": 1.5990729962105953e-05, + "loss": 1.0015, + "step": 25713 + }, + { + "epoch": 1.9, + "learning_rate": 1.5990410583062254e-05, + "loss": 0.9987, + "step": 25714 + }, + { + "epoch": 1.9, + "learning_rate": 1.5990091194487896e-05, + "loss": 1.0066, + "step": 25715 + }, + { + "epoch": 1.9, + "learning_rate": 1.598977179638339e-05, + "loss": 1.0394, + "step": 25716 + }, + { + "epoch": 1.9, + "learning_rate": 1.598945238874924e-05, + "loss": 1.0113, + "step": 25717 + }, + { + "epoch": 1.9, + "learning_rate": 1.598913297158596e-05, + "loss": 1.0549, + "step": 25718 + }, + { + "epoch": 1.9, + "learning_rate": 1.5988813544894056e-05, + "loss": 1.036, + "step": 25719 + }, + { + "epoch": 1.9, + "learning_rate": 1.5988494108674034e-05, + "loss": 1.111, + "step": 25720 + }, + { + "epoch": 1.9, + "learning_rate": 1.5988174662926402e-05, + "loss": 1.0715, + "step": 25721 + }, + { + "epoch": 1.9, + "learning_rate": 1.598785520765167e-05, + "loss": 1.0536, + "step": 25722 + }, + { + "epoch": 1.9, + "learning_rate": 1.5987535742850346e-05, + "loss": 0.9477, + "step": 25723 + }, + { + "epoch": 1.9, + "learning_rate": 1.598721626852294e-05, + "loss": 0.9509, + "step": 25724 + }, + { + "epoch": 1.9, + "learning_rate": 1.5986896784669957e-05, + "loss": 1.008, + "step": 25725 + }, + { + "epoch": 1.9, + "learning_rate": 1.598657729129191e-05, + "loss": 1.0035, + "step": 25726 + }, + { + "epoch": 1.9, + "learning_rate": 1.59862577883893e-05, + "loss": 0.9915, + "step": 25727 + }, + { + "epoch": 1.9, + "learning_rate": 1.5985938275962643e-05, + "loss": 0.9569, + "step": 25728 + }, + { + "epoch": 1.9, + "learning_rate": 1.598561875401244e-05, + "loss": 0.9595, + "step": 25729 + }, + { + "epoch": 1.9, + "learning_rate": 1.598529922253921e-05, + "loss": 1.0651, + "step": 25730 + }, + { + "epoch": 1.9, + "learning_rate": 1.5984979681543448e-05, + "loss": 1.0738, + "step": 25731 + }, + { + "epoch": 1.9, + "learning_rate": 1.598466013102567e-05, + "loss": 0.9711, + "step": 25732 + }, + { + "epoch": 1.9, + "learning_rate": 1.5984340570986385e-05, + "loss": 0.9643, + "step": 25733 + }, + { + "epoch": 1.9, + "learning_rate": 1.59840210014261e-05, + "loss": 0.9426, + "step": 25734 + }, + { + "epoch": 1.9, + "learning_rate": 1.5983701422345322e-05, + "loss": 1.0791, + "step": 25735 + }, + { + "epoch": 1.9, + "learning_rate": 1.5983381833744564e-05, + "loss": 0.8964, + "step": 25736 + }, + { + "epoch": 1.9, + "learning_rate": 1.5983062235624327e-05, + "loss": 0.9557, + "step": 25737 + }, + { + "epoch": 1.9, + "learning_rate": 1.5982742627985125e-05, + "loss": 0.9434, + "step": 25738 + }, + { + "epoch": 1.9, + "learning_rate": 1.5982423010827464e-05, + "loss": 1.0144, + "step": 25739 + }, + { + "epoch": 1.9, + "learning_rate": 1.5982103384151853e-05, + "loss": 0.9321, + "step": 25740 + }, + { + "epoch": 1.9, + "learning_rate": 1.59817837479588e-05, + "loss": 0.9393, + "step": 25741 + }, + { + "epoch": 1.9, + "learning_rate": 1.598146410224882e-05, + "loss": 1.0043, + "step": 25742 + }, + { + "epoch": 1.9, + "learning_rate": 1.5981144447022413e-05, + "loss": 0.9761, + "step": 25743 + }, + { + "epoch": 1.9, + "learning_rate": 1.598082478228009e-05, + "loss": 0.9663, + "step": 25744 + }, + { + "epoch": 1.9, + "learning_rate": 1.5980505108022363e-05, + "loss": 1.2242, + "step": 25745 + }, + { + "epoch": 1.9, + "learning_rate": 1.5980185424249735e-05, + "loss": 0.8861, + "step": 25746 + }, + { + "epoch": 1.9, + "learning_rate": 1.597986573096272e-05, + "loss": 0.9884, + "step": 25747 + }, + { + "epoch": 1.9, + "learning_rate": 1.5979546028161824e-05, + "loss": 1.0049, + "step": 25748 + }, + { + "epoch": 1.9, + "learning_rate": 1.5979226315847553e-05, + "loss": 0.99, + "step": 25749 + }, + { + "epoch": 1.9, + "learning_rate": 1.597890659402042e-05, + "loss": 1.0518, + "step": 25750 + }, + { + "epoch": 1.9, + "learning_rate": 1.597858686268093e-05, + "loss": 0.9569, + "step": 25751 + }, + { + "epoch": 1.9, + "learning_rate": 1.5978267121829598e-05, + "loss": 1.0254, + "step": 25752 + }, + { + "epoch": 1.9, + "learning_rate": 1.5977947371466926e-05, + "loss": 0.9501, + "step": 25753 + }, + { + "epoch": 1.9, + "learning_rate": 1.5977627611593433e-05, + "loss": 0.9717, + "step": 25754 + }, + { + "epoch": 1.9, + "learning_rate": 1.597730784220961e-05, + "loss": 1.0933, + "step": 25755 + }, + { + "epoch": 1.9, + "learning_rate": 1.597698806331598e-05, + "loss": 0.994, + "step": 25756 + }, + { + "epoch": 1.9, + "learning_rate": 1.5976668274913045e-05, + "loss": 1.0377, + "step": 25757 + }, + { + "epoch": 1.9, + "learning_rate": 1.5976348477001318e-05, + "loss": 1.1753, + "step": 25758 + }, + { + "epoch": 1.9, + "learning_rate": 1.5976028669581304e-05, + "loss": 0.9681, + "step": 25759 + }, + { + "epoch": 1.9, + "learning_rate": 1.5975708852653514e-05, + "loss": 0.9588, + "step": 25760 + }, + { + "epoch": 1.9, + "learning_rate": 1.597538902621846e-05, + "loss": 0.9961, + "step": 25761 + }, + { + "epoch": 1.9, + "learning_rate": 1.5975069190276645e-05, + "loss": 1.0265, + "step": 25762 + }, + { + "epoch": 1.9, + "learning_rate": 1.5974749344828585e-05, + "loss": 1.0246, + "step": 25763 + }, + { + "epoch": 1.9, + "learning_rate": 1.597442948987478e-05, + "loss": 1.0585, + "step": 25764 + }, + { + "epoch": 1.9, + "learning_rate": 1.597410962541574e-05, + "loss": 0.9453, + "step": 25765 + }, + { + "epoch": 1.9, + "learning_rate": 1.5973789751451984e-05, + "loss": 1.0516, + "step": 25766 + }, + { + "epoch": 1.9, + "learning_rate": 1.597346986798401e-05, + "loss": 1.0916, + "step": 25767 + }, + { + "epoch": 1.9, + "learning_rate": 1.5973149975012333e-05, + "loss": 1.1491, + "step": 25768 + }, + { + "epoch": 1.9, + "learning_rate": 1.5972830072537458e-05, + "loss": 0.9531, + "step": 25769 + }, + { + "epoch": 1.9, + "learning_rate": 1.5972510160559897e-05, + "loss": 1.109, + "step": 25770 + }, + { + "epoch": 1.9, + "learning_rate": 1.5972190239080157e-05, + "loss": 1.0215, + "step": 25771 + }, + { + "epoch": 1.9, + "learning_rate": 1.5971870308098748e-05, + "loss": 1.0687, + "step": 25772 + }, + { + "epoch": 1.9, + "learning_rate": 1.5971550367616176e-05, + "loss": 1.0819, + "step": 25773 + }, + { + "epoch": 1.9, + "learning_rate": 1.5971230417632958e-05, + "loss": 1.0953, + "step": 25774 + }, + { + "epoch": 1.9, + "learning_rate": 1.5970910458149595e-05, + "loss": 1.024, + "step": 25775 + }, + { + "epoch": 1.9, + "learning_rate": 1.59705904891666e-05, + "loss": 0.9235, + "step": 25776 + }, + { + "epoch": 1.9, + "learning_rate": 1.597027051068448e-05, + "loss": 1.0555, + "step": 25777 + }, + { + "epoch": 1.9, + "learning_rate": 1.5969950522703745e-05, + "loss": 1.0646, + "step": 25778 + }, + { + "epoch": 1.9, + "learning_rate": 1.5969630525224903e-05, + "loss": 1.0186, + "step": 25779 + }, + { + "epoch": 1.9, + "learning_rate": 1.5969310518248466e-05, + "loss": 1.1239, + "step": 25780 + }, + { + "epoch": 1.9, + "learning_rate": 1.596899050177494e-05, + "loss": 0.9304, + "step": 25781 + }, + { + "epoch": 1.91, + "learning_rate": 1.596867047580484e-05, + "loss": 1.0791, + "step": 25782 + }, + { + "epoch": 1.91, + "learning_rate": 1.5968350440338668e-05, + "loss": 1.0005, + "step": 25783 + }, + { + "epoch": 1.91, + "learning_rate": 1.5968030395376934e-05, + "loss": 0.9955, + "step": 25784 + }, + { + "epoch": 1.91, + "learning_rate": 1.596771034092015e-05, + "loss": 0.98, + "step": 25785 + }, + { + "epoch": 1.91, + "learning_rate": 1.5967390276968825e-05, + "loss": 0.9944, + "step": 25786 + }, + { + "epoch": 1.91, + "learning_rate": 1.5967070203523463e-05, + "loss": 0.9817, + "step": 25787 + }, + { + "epoch": 1.91, + "learning_rate": 1.5966750120584584e-05, + "loss": 0.8698, + "step": 25788 + }, + { + "epoch": 1.91, + "learning_rate": 1.5966430028152687e-05, + "loss": 1.0414, + "step": 25789 + }, + { + "epoch": 1.91, + "learning_rate": 1.5966109926228286e-05, + "loss": 1.014, + "step": 25790 + }, + { + "epoch": 1.91, + "learning_rate": 1.5965789814811894e-05, + "loss": 1.0435, + "step": 25791 + }, + { + "epoch": 1.91, + "learning_rate": 1.596546969390401e-05, + "loss": 1.1, + "step": 25792 + }, + { + "epoch": 1.91, + "learning_rate": 1.5965149563505153e-05, + "loss": 0.9318, + "step": 25793 + }, + { + "epoch": 1.91, + "learning_rate": 1.5964829423615823e-05, + "loss": 0.9841, + "step": 25794 + }, + { + "epoch": 1.91, + "learning_rate": 1.596450927423654e-05, + "loss": 0.9682, + "step": 25795 + }, + { + "epoch": 1.91, + "learning_rate": 1.5964189115367804e-05, + "loss": 1.1927, + "step": 25796 + }, + { + "epoch": 1.91, + "learning_rate": 1.596386894701013e-05, + "loss": 1.0814, + "step": 25797 + }, + { + "epoch": 1.91, + "learning_rate": 1.5963548769164024e-05, + "loss": 1.0056, + "step": 25798 + }, + { + "epoch": 1.91, + "learning_rate": 1.596322858183e-05, + "loss": 1.2012, + "step": 25799 + }, + { + "epoch": 1.91, + "learning_rate": 1.5962908385008568e-05, + "loss": 1.0893, + "step": 25800 + }, + { + "epoch": 1.91, + "learning_rate": 1.596258817870023e-05, + "loss": 0.9288, + "step": 25801 + }, + { + "epoch": 1.91, + "learning_rate": 1.59622679629055e-05, + "loss": 1.0358, + "step": 25802 + }, + { + "epoch": 1.91, + "learning_rate": 1.5961947737624885e-05, + "loss": 0.9236, + "step": 25803 + }, + { + "epoch": 1.91, + "learning_rate": 1.59616275028589e-05, + "loss": 0.952, + "step": 25804 + }, + { + "epoch": 1.91, + "learning_rate": 1.596130725860805e-05, + "loss": 0.9449, + "step": 25805 + }, + { + "epoch": 1.91, + "learning_rate": 1.5960987004872845e-05, + "loss": 1.0293, + "step": 25806 + }, + { + "epoch": 1.91, + "learning_rate": 1.5960666741653793e-05, + "loss": 0.9184, + "step": 25807 + }, + { + "epoch": 1.91, + "learning_rate": 1.5960346468951405e-05, + "loss": 0.9812, + "step": 25808 + }, + { + "epoch": 1.91, + "learning_rate": 1.5960026186766195e-05, + "loss": 0.9659, + "step": 25809 + }, + { + "epoch": 1.91, + "learning_rate": 1.5959705895098668e-05, + "loss": 1.0904, + "step": 25810 + }, + { + "epoch": 1.91, + "learning_rate": 1.5959385593949332e-05, + "loss": 0.9514, + "step": 25811 + }, + { + "epoch": 1.91, + "learning_rate": 1.5959065283318703e-05, + "loss": 0.9328, + "step": 25812 + }, + { + "epoch": 1.91, + "learning_rate": 1.595874496320728e-05, + "loss": 1.0409, + "step": 25813 + }, + { + "epoch": 1.91, + "learning_rate": 1.5958424633615585e-05, + "loss": 1.0291, + "step": 25814 + }, + { + "epoch": 1.91, + "learning_rate": 1.5958104294544116e-05, + "loss": 1.0012, + "step": 25815 + }, + { + "epoch": 1.91, + "learning_rate": 1.5957783945993394e-05, + "loss": 1.1027, + "step": 25816 + }, + { + "epoch": 1.91, + "learning_rate": 1.595746358796392e-05, + "loss": 1.1377, + "step": 25817 + }, + { + "epoch": 1.91, + "learning_rate": 1.5957143220456202e-05, + "loss": 1.0838, + "step": 25818 + }, + { + "epoch": 1.91, + "learning_rate": 1.5956822843470763e-05, + "loss": 0.8899, + "step": 25819 + }, + { + "epoch": 1.91, + "learning_rate": 1.59565024570081e-05, + "loss": 1.0968, + "step": 25820 + }, + { + "epoch": 1.91, + "learning_rate": 1.5956182061068726e-05, + "loss": 0.937, + "step": 25821 + }, + { + "epoch": 1.91, + "learning_rate": 1.595586165565315e-05, + "loss": 0.9569, + "step": 25822 + }, + { + "epoch": 1.91, + "learning_rate": 1.595554124076189e-05, + "loss": 0.9776, + "step": 25823 + }, + { + "epoch": 1.91, + "learning_rate": 1.595522081639544e-05, + "loss": 0.9909, + "step": 25824 + }, + { + "epoch": 1.91, + "learning_rate": 1.5954900382554325e-05, + "loss": 0.9926, + "step": 25825 + }, + { + "epoch": 1.91, + "learning_rate": 1.5954579939239047e-05, + "loss": 0.9304, + "step": 25826 + }, + { + "epoch": 1.91, + "learning_rate": 1.5954259486450122e-05, + "loss": 1.0394, + "step": 25827 + }, + { + "epoch": 1.91, + "learning_rate": 1.595393902418805e-05, + "loss": 0.9761, + "step": 25828 + }, + { + "epoch": 1.91, + "learning_rate": 1.595361855245335e-05, + "loss": 0.9786, + "step": 25829 + }, + { + "epoch": 1.91, + "learning_rate": 1.5953298071246523e-05, + "loss": 1.0153, + "step": 25830 + }, + { + "epoch": 1.91, + "learning_rate": 1.5952977580568086e-05, + "loss": 1.1171, + "step": 25831 + }, + { + "epoch": 1.91, + "learning_rate": 1.595265708041855e-05, + "loss": 1.0975, + "step": 25832 + }, + { + "epoch": 1.91, + "learning_rate": 1.595233657079842e-05, + "loss": 1.048, + "step": 25833 + }, + { + "epoch": 1.91, + "learning_rate": 1.5952016051708207e-05, + "loss": 1.0733, + "step": 25834 + }, + { + "epoch": 1.91, + "learning_rate": 1.5951695523148417e-05, + "loss": 0.9574, + "step": 25835 + }, + { + "epoch": 1.91, + "learning_rate": 1.595137498511957e-05, + "loss": 0.9899, + "step": 25836 + }, + { + "epoch": 1.91, + "learning_rate": 1.5951054437622172e-05, + "loss": 1.0849, + "step": 25837 + }, + { + "epoch": 1.91, + "learning_rate": 1.5950733880656726e-05, + "loss": 0.9479, + "step": 25838 + }, + { + "epoch": 1.91, + "learning_rate": 1.5950413314223753e-05, + "loss": 1.028, + "step": 25839 + }, + { + "epoch": 1.91, + "learning_rate": 1.5950092738323756e-05, + "loss": 1.0386, + "step": 25840 + }, + { + "epoch": 1.91, + "learning_rate": 1.5949772152957245e-05, + "loss": 1.0999, + "step": 25841 + }, + { + "epoch": 1.91, + "learning_rate": 1.5949451558124733e-05, + "loss": 0.9903, + "step": 25842 + }, + { + "epoch": 1.91, + "learning_rate": 1.5949130953826727e-05, + "loss": 1.0373, + "step": 25843 + }, + { + "epoch": 1.91, + "learning_rate": 1.594881034006374e-05, + "loss": 1.0736, + "step": 25844 + }, + { + "epoch": 1.91, + "learning_rate": 1.594848971683628e-05, + "loss": 0.97, + "step": 25845 + }, + { + "epoch": 1.91, + "learning_rate": 1.594816908414486e-05, + "loss": 1.1084, + "step": 25846 + }, + { + "epoch": 1.91, + "learning_rate": 1.594784844198999e-05, + "loss": 1.0597, + "step": 25847 + }, + { + "epoch": 1.91, + "learning_rate": 1.5947527790372172e-05, + "loss": 0.9541, + "step": 25848 + }, + { + "epoch": 1.91, + "learning_rate": 1.5947207129291927e-05, + "loss": 0.8454, + "step": 25849 + }, + { + "epoch": 1.91, + "learning_rate": 1.594688645874976e-05, + "loss": 0.9587, + "step": 25850 + }, + { + "epoch": 1.91, + "learning_rate": 1.5946565778746183e-05, + "loss": 0.9471, + "step": 25851 + }, + { + "epoch": 1.91, + "learning_rate": 1.59462450892817e-05, + "loss": 1.0119, + "step": 25852 + }, + { + "epoch": 1.91, + "learning_rate": 1.594592439035683e-05, + "loss": 1.089, + "step": 25853 + }, + { + "epoch": 1.91, + "learning_rate": 1.5945603681972074e-05, + "loss": 1.0073, + "step": 25854 + }, + { + "epoch": 1.91, + "learning_rate": 1.5945282964127953e-05, + "loss": 1.088, + "step": 25855 + }, + { + "epoch": 1.91, + "learning_rate": 1.594496223682497e-05, + "loss": 0.9406, + "step": 25856 + }, + { + "epoch": 1.91, + "learning_rate": 1.5944641500063643e-05, + "loss": 1.033, + "step": 25857 + }, + { + "epoch": 1.91, + "learning_rate": 1.594432075384447e-05, + "loss": 0.9737, + "step": 25858 + }, + { + "epoch": 1.91, + "learning_rate": 1.5943999998167968e-05, + "loss": 0.9648, + "step": 25859 + }, + { + "epoch": 1.91, + "learning_rate": 1.5943679233034648e-05, + "loss": 1.0671, + "step": 25860 + }, + { + "epoch": 1.91, + "learning_rate": 1.5943358458445018e-05, + "loss": 0.877, + "step": 25861 + }, + { + "epoch": 1.91, + "learning_rate": 1.5943037674399593e-05, + "loss": 1.0705, + "step": 25862 + }, + { + "epoch": 1.91, + "learning_rate": 1.594271688089888e-05, + "loss": 0.9935, + "step": 25863 + }, + { + "epoch": 1.91, + "learning_rate": 1.5942396077943386e-05, + "loss": 0.9841, + "step": 25864 + }, + { + "epoch": 1.91, + "learning_rate": 1.5942075265533624e-05, + "loss": 0.9308, + "step": 25865 + }, + { + "epoch": 1.91, + "learning_rate": 1.594175444367011e-05, + "loss": 1.126, + "step": 25866 + }, + { + "epoch": 1.91, + "learning_rate": 1.5941433612353348e-05, + "loss": 0.9667, + "step": 25867 + }, + { + "epoch": 1.91, + "learning_rate": 1.594111277158385e-05, + "loss": 1.0833, + "step": 25868 + }, + { + "epoch": 1.91, + "learning_rate": 1.5940791921362125e-05, + "loss": 1.086, + "step": 25869 + }, + { + "epoch": 1.91, + "learning_rate": 1.5940471061688686e-05, + "loss": 0.9899, + "step": 25870 + }, + { + "epoch": 1.91, + "learning_rate": 1.5940150192564044e-05, + "loss": 1.009, + "step": 25871 + }, + { + "epoch": 1.91, + "learning_rate": 1.5939829313988705e-05, + "loss": 0.9363, + "step": 25872 + }, + { + "epoch": 1.91, + "learning_rate": 1.5939508425963187e-05, + "loss": 1.0131, + "step": 25873 + }, + { + "epoch": 1.91, + "learning_rate": 1.5939187528487992e-05, + "loss": 0.8894, + "step": 25874 + }, + { + "epoch": 1.91, + "learning_rate": 1.5938866621563637e-05, + "loss": 0.9122, + "step": 25875 + }, + { + "epoch": 1.91, + "learning_rate": 1.593854570519063e-05, + "loss": 1.002, + "step": 25876 + }, + { + "epoch": 1.91, + "learning_rate": 1.593822477936948e-05, + "loss": 1.0758, + "step": 25877 + }, + { + "epoch": 1.91, + "learning_rate": 1.59379038441007e-05, + "loss": 0.9436, + "step": 25878 + }, + { + "epoch": 1.91, + "learning_rate": 1.59375828993848e-05, + "loss": 1.0986, + "step": 25879 + }, + { + "epoch": 1.91, + "learning_rate": 1.593726194522229e-05, + "loss": 1.0436, + "step": 25880 + }, + { + "epoch": 1.91, + "learning_rate": 1.593694098161368e-05, + "loss": 0.9946, + "step": 25881 + }, + { + "epoch": 1.91, + "learning_rate": 1.5936620008559485e-05, + "loss": 0.9723, + "step": 25882 + }, + { + "epoch": 1.91, + "learning_rate": 1.593629902606021e-05, + "loss": 0.9206, + "step": 25883 + }, + { + "epoch": 1.91, + "learning_rate": 1.5935978034116374e-05, + "loss": 1.0301, + "step": 25884 + }, + { + "epoch": 1.91, + "learning_rate": 1.593565703272848e-05, + "loss": 1.1035, + "step": 25885 + }, + { + "epoch": 1.91, + "learning_rate": 1.5935336021897033e-05, + "loss": 0.9931, + "step": 25886 + }, + { + "epoch": 1.91, + "learning_rate": 1.593501500162256e-05, + "loss": 0.9897, + "step": 25887 + }, + { + "epoch": 1.91, + "learning_rate": 1.593469397190556e-05, + "loss": 1.0538, + "step": 25888 + }, + { + "epoch": 1.91, + "learning_rate": 1.5934372932746543e-05, + "loss": 0.9824, + "step": 25889 + }, + { + "epoch": 1.91, + "learning_rate": 1.5934051884146028e-05, + "loss": 0.9974, + "step": 25890 + }, + { + "epoch": 1.91, + "learning_rate": 1.593373082610452e-05, + "loss": 1.0772, + "step": 25891 + }, + { + "epoch": 1.91, + "learning_rate": 1.593340975862253e-05, + "loss": 0.8943, + "step": 25892 + }, + { + "epoch": 1.91, + "learning_rate": 1.593308868170057e-05, + "loss": 0.9726, + "step": 25893 + }, + { + "epoch": 1.91, + "learning_rate": 1.5932767595339158e-05, + "loss": 1.017, + "step": 25894 + }, + { + "epoch": 1.91, + "learning_rate": 1.5932446499538788e-05, + "loss": 1.0195, + "step": 25895 + }, + { + "epoch": 1.91, + "learning_rate": 1.5932125394299984e-05, + "loss": 1.0214, + "step": 25896 + }, + { + "epoch": 1.91, + "learning_rate": 1.5931804279623255e-05, + "loss": 0.9755, + "step": 25897 + }, + { + "epoch": 1.91, + "learning_rate": 1.5931483155509108e-05, + "loss": 1.0535, + "step": 25898 + }, + { + "epoch": 1.91, + "learning_rate": 1.593116202195806e-05, + "loss": 0.9215, + "step": 25899 + }, + { + "epoch": 1.91, + "learning_rate": 1.5930840878970612e-05, + "loss": 1.0545, + "step": 25900 + }, + { + "epoch": 1.91, + "learning_rate": 1.5930519726547286e-05, + "loss": 1.0811, + "step": 25901 + }, + { + "epoch": 1.91, + "learning_rate": 1.5930198564688585e-05, + "loss": 1.146, + "step": 25902 + }, + { + "epoch": 1.91, + "learning_rate": 1.5929877393395025e-05, + "loss": 1.0775, + "step": 25903 + }, + { + "epoch": 1.91, + "learning_rate": 1.5929556212667114e-05, + "loss": 1.0892, + "step": 25904 + }, + { + "epoch": 1.91, + "learning_rate": 1.5929235022505363e-05, + "loss": 0.9771, + "step": 25905 + }, + { + "epoch": 1.91, + "learning_rate": 1.5928913822910287e-05, + "loss": 0.9674, + "step": 25906 + }, + { + "epoch": 1.91, + "learning_rate": 1.592859261388239e-05, + "loss": 1.1286, + "step": 25907 + }, + { + "epoch": 1.91, + "learning_rate": 1.592827139542219e-05, + "loss": 1.0817, + "step": 25908 + }, + { + "epoch": 1.91, + "learning_rate": 1.5927950167530192e-05, + "loss": 0.9498, + "step": 25909 + }, + { + "epoch": 1.91, + "learning_rate": 1.592762893020691e-05, + "loss": 0.9768, + "step": 25910 + }, + { + "epoch": 1.91, + "learning_rate": 1.592730768345286e-05, + "loss": 0.9852, + "step": 25911 + }, + { + "epoch": 1.91, + "learning_rate": 1.5926986427268542e-05, + "loss": 1.0243, + "step": 25912 + }, + { + "epoch": 1.91, + "learning_rate": 1.5926665161654478e-05, + "loss": 0.9716, + "step": 25913 + }, + { + "epoch": 1.91, + "learning_rate": 1.5926343886611173e-05, + "loss": 1.0133, + "step": 25914 + }, + { + "epoch": 1.91, + "learning_rate": 1.592602260213914e-05, + "loss": 1.088, + "step": 25915 + }, + { + "epoch": 1.91, + "learning_rate": 1.592570130823889e-05, + "loss": 1.0003, + "step": 25916 + }, + { + "epoch": 1.91, + "learning_rate": 1.5925380004910935e-05, + "loss": 1.0898, + "step": 25917 + }, + { + "epoch": 1.92, + "learning_rate": 1.5925058692155784e-05, + "loss": 1.0548, + "step": 25918 + }, + { + "epoch": 1.92, + "learning_rate": 1.5924737369973947e-05, + "loss": 1.0061, + "step": 25919 + }, + { + "epoch": 1.92, + "learning_rate": 1.592441603836594e-05, + "loss": 0.9655, + "step": 25920 + }, + { + "epoch": 1.92, + "learning_rate": 1.592409469733227e-05, + "loss": 1.0521, + "step": 25921 + }, + { + "epoch": 1.92, + "learning_rate": 1.5923773346873454e-05, + "loss": 1.0972, + "step": 25922 + }, + { + "epoch": 1.92, + "learning_rate": 1.5923451986989997e-05, + "loss": 1.1099, + "step": 25923 + }, + { + "epoch": 1.92, + "learning_rate": 1.592313061768241e-05, + "loss": 0.9849, + "step": 25924 + }, + { + "epoch": 1.92, + "learning_rate": 1.592280923895121e-05, + "loss": 1.0574, + "step": 25925 + }, + { + "epoch": 1.92, + "learning_rate": 1.5922487850796907e-05, + "loss": 1.0505, + "step": 25926 + }, + { + "epoch": 1.92, + "learning_rate": 1.592216645322001e-05, + "loss": 0.947, + "step": 25927 + }, + { + "epoch": 1.92, + "learning_rate": 1.5921845046221026e-05, + "loss": 0.9856, + "step": 25928 + }, + { + "epoch": 1.92, + "learning_rate": 1.5921523629800475e-05, + "loss": 0.9516, + "step": 25929 + }, + { + "epoch": 1.92, + "learning_rate": 1.5921202203958864e-05, + "loss": 1.0209, + "step": 25930 + }, + { + "epoch": 1.92, + "learning_rate": 1.592088076869671e-05, + "loss": 0.9832, + "step": 25931 + }, + { + "epoch": 1.92, + "learning_rate": 1.5920559324014513e-05, + "loss": 1.0388, + "step": 25932 + }, + { + "epoch": 1.92, + "learning_rate": 1.5920237869912793e-05, + "loss": 0.9904, + "step": 25933 + }, + { + "epoch": 1.92, + "learning_rate": 1.591991640639206e-05, + "loss": 0.9685, + "step": 25934 + }, + { + "epoch": 1.92, + "learning_rate": 1.591959493345282e-05, + "loss": 1.0227, + "step": 25935 + }, + { + "epoch": 1.92, + "learning_rate": 1.5919273451095593e-05, + "loss": 0.9652, + "step": 25936 + }, + { + "epoch": 1.92, + "learning_rate": 1.5918951959320887e-05, + "loss": 0.9327, + "step": 25937 + }, + { + "epoch": 1.92, + "learning_rate": 1.591863045812921e-05, + "loss": 1.0154, + "step": 25938 + }, + { + "epoch": 1.92, + "learning_rate": 1.5918308947521082e-05, + "loss": 1.1001, + "step": 25939 + }, + { + "epoch": 1.92, + "learning_rate": 1.5917987427497007e-05, + "loss": 0.9945, + "step": 25940 + }, + { + "epoch": 1.92, + "learning_rate": 1.59176658980575e-05, + "loss": 1.0171, + "step": 25941 + }, + { + "epoch": 1.92, + "learning_rate": 1.591734435920307e-05, + "loss": 0.9579, + "step": 25942 + }, + { + "epoch": 1.92, + "learning_rate": 1.5917022810934226e-05, + "loss": 0.9583, + "step": 25943 + }, + { + "epoch": 1.92, + "learning_rate": 1.5916701253251486e-05, + "loss": 0.9654, + "step": 25944 + }, + { + "epoch": 1.92, + "learning_rate": 1.5916379686155358e-05, + "loss": 0.9806, + "step": 25945 + }, + { + "epoch": 1.92, + "learning_rate": 1.5916058109646355e-05, + "loss": 1.0724, + "step": 25946 + }, + { + "epoch": 1.92, + "learning_rate": 1.591573652372499e-05, + "loss": 1.018, + "step": 25947 + }, + { + "epoch": 1.92, + "learning_rate": 1.591541492839177e-05, + "loss": 1.024, + "step": 25948 + }, + { + "epoch": 1.92, + "learning_rate": 1.5915093323647214e-05, + "loss": 1.022, + "step": 25949 + }, + { + "epoch": 1.92, + "learning_rate": 1.5914771709491828e-05, + "loss": 1.0497, + "step": 25950 + }, + { + "epoch": 1.92, + "learning_rate": 1.5914450085926124e-05, + "loss": 1.0175, + "step": 25951 + }, + { + "epoch": 1.92, + "learning_rate": 1.591412845295061e-05, + "loss": 1.029, + "step": 25952 + }, + { + "epoch": 1.92, + "learning_rate": 1.5913806810565807e-05, + "loss": 1.0343, + "step": 25953 + }, + { + "epoch": 1.92, + "learning_rate": 1.591348515877222e-05, + "loss": 0.9755, + "step": 25954 + }, + { + "epoch": 1.92, + "learning_rate": 1.5913163497570366e-05, + "loss": 1.0227, + "step": 25955 + }, + { + "epoch": 1.92, + "learning_rate": 1.5912841826960753e-05, + "loss": 1.0485, + "step": 25956 + }, + { + "epoch": 1.92, + "learning_rate": 1.5912520146943887e-05, + "loss": 0.9707, + "step": 25957 + }, + { + "epoch": 1.92, + "learning_rate": 1.591219845752029e-05, + "loss": 1.0492, + "step": 25958 + }, + { + "epoch": 1.92, + "learning_rate": 1.5911876758690474e-05, + "loss": 1.0471, + "step": 25959 + }, + { + "epoch": 1.92, + "learning_rate": 1.5911555050454943e-05, + "loss": 1.0235, + "step": 25960 + }, + { + "epoch": 1.92, + "learning_rate": 1.5911233332814212e-05, + "loss": 1.0496, + "step": 25961 + }, + { + "epoch": 1.92, + "learning_rate": 1.5910911605768796e-05, + "loss": 1.0459, + "step": 25962 + }, + { + "epoch": 1.92, + "learning_rate": 1.59105898693192e-05, + "loss": 0.9478, + "step": 25963 + }, + { + "epoch": 1.92, + "learning_rate": 1.5910268123465943e-05, + "loss": 0.948, + "step": 25964 + }, + { + "epoch": 1.92, + "learning_rate": 1.5909946368209533e-05, + "loss": 1.0633, + "step": 25965 + }, + { + "epoch": 1.92, + "learning_rate": 1.590962460355048e-05, + "loss": 0.9847, + "step": 25966 + }, + { + "epoch": 1.92, + "learning_rate": 1.5909302829489305e-05, + "loss": 1.0819, + "step": 25967 + }, + { + "epoch": 1.92, + "learning_rate": 1.5908981046026508e-05, + "loss": 1.0557, + "step": 25968 + }, + { + "epoch": 1.92, + "learning_rate": 1.590865925316261e-05, + "loss": 1.0759, + "step": 25969 + }, + { + "epoch": 1.92, + "learning_rate": 1.590833745089812e-05, + "loss": 1.0228, + "step": 25970 + }, + { + "epoch": 1.92, + "learning_rate": 1.5908015639233548e-05, + "loss": 1.0435, + "step": 25971 + }, + { + "epoch": 1.92, + "learning_rate": 1.5907693818169406e-05, + "loss": 0.9654, + "step": 25972 + }, + { + "epoch": 1.92, + "learning_rate": 1.5907371987706214e-05, + "loss": 0.9342, + "step": 25973 + }, + { + "epoch": 1.92, + "learning_rate": 1.590705014784447e-05, + "loss": 1.0317, + "step": 25974 + }, + { + "epoch": 1.92, + "learning_rate": 1.59067282985847e-05, + "loss": 0.999, + "step": 25975 + }, + { + "epoch": 1.92, + "learning_rate": 1.5906406439927405e-05, + "loss": 0.9547, + "step": 25976 + }, + { + "epoch": 1.92, + "learning_rate": 1.5906084571873105e-05, + "loss": 0.9246, + "step": 25977 + }, + { + "epoch": 1.92, + "learning_rate": 1.590576269442231e-05, + "loss": 1.0617, + "step": 25978 + }, + { + "epoch": 1.92, + "learning_rate": 1.5905440807575528e-05, + "loss": 1.0042, + "step": 25979 + }, + { + "epoch": 1.92, + "learning_rate": 1.5905118911333275e-05, + "loss": 0.988, + "step": 25980 + }, + { + "epoch": 1.92, + "learning_rate": 1.5904797005696062e-05, + "loss": 1.1683, + "step": 25981 + }, + { + "epoch": 1.92, + "learning_rate": 1.5904475090664402e-05, + "loss": 0.9575, + "step": 25982 + }, + { + "epoch": 1.92, + "learning_rate": 1.5904153166238806e-05, + "loss": 1.073, + "step": 25983 + }, + { + "epoch": 1.92, + "learning_rate": 1.5903831232419786e-05, + "loss": 1.0353, + "step": 25984 + }, + { + "epoch": 1.92, + "learning_rate": 1.5903509289207856e-05, + "loss": 0.9691, + "step": 25985 + }, + { + "epoch": 1.92, + "learning_rate": 1.590318733660353e-05, + "loss": 1.1189, + "step": 25986 + }, + { + "epoch": 1.92, + "learning_rate": 1.5902865374607315e-05, + "loss": 0.9141, + "step": 25987 + }, + { + "epoch": 1.92, + "learning_rate": 1.5902543403219724e-05, + "loss": 0.9461, + "step": 25988 + }, + { + "epoch": 1.92, + "learning_rate": 1.5902221422441276e-05, + "loss": 1.0307, + "step": 25989 + }, + { + "epoch": 1.92, + "learning_rate": 1.590189943227247e-05, + "loss": 0.9542, + "step": 25990 + }, + { + "epoch": 1.92, + "learning_rate": 1.5901577432713833e-05, + "loss": 1.0436, + "step": 25991 + }, + { + "epoch": 1.92, + "learning_rate": 1.590125542376587e-05, + "loss": 1.0588, + "step": 25992 + }, + { + "epoch": 1.92, + "learning_rate": 1.590093340542909e-05, + "loss": 1.0032, + "step": 25993 + }, + { + "epoch": 1.92, + "learning_rate": 1.5900611377704014e-05, + "loss": 1.0254, + "step": 25994 + }, + { + "epoch": 1.92, + "learning_rate": 1.590028934059115e-05, + "loss": 0.9816, + "step": 25995 + }, + { + "epoch": 1.92, + "learning_rate": 1.5899967294091005e-05, + "loss": 0.8989, + "step": 25996 + }, + { + "epoch": 1.92, + "learning_rate": 1.58996452382041e-05, + "loss": 1.0625, + "step": 25997 + }, + { + "epoch": 1.92, + "learning_rate": 1.5899323172930946e-05, + "loss": 0.9589, + "step": 25998 + }, + { + "epoch": 1.92, + "learning_rate": 1.589900109827205e-05, + "loss": 0.9692, + "step": 25999 + }, + { + "epoch": 1.92, + "learning_rate": 1.5898679014227927e-05, + "loss": 1.0263, + "step": 26000 + }, + { + "epoch": 1.92, + "learning_rate": 1.5898356920799088e-05, + "loss": 1.0184, + "step": 26001 + }, + { + "epoch": 1.92, + "learning_rate": 1.5898034817986055e-05, + "loss": 1.0361, + "step": 26002 + }, + { + "epoch": 1.92, + "learning_rate": 1.5897712705789327e-05, + "loss": 1.0143, + "step": 26003 + }, + { + "epoch": 1.92, + "learning_rate": 1.5897390584209426e-05, + "loss": 0.996, + "step": 26004 + }, + { + "epoch": 1.92, + "learning_rate": 1.589706845324686e-05, + "loss": 0.9783, + "step": 26005 + }, + { + "epoch": 1.92, + "learning_rate": 1.5896746312902142e-05, + "loss": 1.0086, + "step": 26006 + }, + { + "epoch": 1.92, + "learning_rate": 1.5896424163175783e-05, + "loss": 1.0105, + "step": 26007 + }, + { + "epoch": 1.92, + "learning_rate": 1.58961020040683e-05, + "loss": 1.008, + "step": 26008 + }, + { + "epoch": 1.92, + "learning_rate": 1.58957798355802e-05, + "loss": 0.9749, + "step": 26009 + }, + { + "epoch": 1.92, + "learning_rate": 1.5895457657712003e-05, + "loss": 1.058, + "step": 26010 + }, + { + "epoch": 1.92, + "learning_rate": 1.5895135470464212e-05, + "loss": 1.0352, + "step": 26011 + }, + { + "epoch": 1.92, + "learning_rate": 1.589481327383735e-05, + "loss": 1.0924, + "step": 26012 + }, + { + "epoch": 1.92, + "learning_rate": 1.5894491067831923e-05, + "loss": 1.1477, + "step": 26013 + }, + { + "epoch": 1.92, + "learning_rate": 1.5894168852448442e-05, + "loss": 0.9238, + "step": 26014 + }, + { + "epoch": 1.92, + "learning_rate": 1.5893846627687426e-05, + "loss": 1.0275, + "step": 26015 + }, + { + "epoch": 1.92, + "learning_rate": 1.5893524393549385e-05, + "loss": 0.9502, + "step": 26016 + }, + { + "epoch": 1.92, + "learning_rate": 1.589320215003483e-05, + "loss": 0.9818, + "step": 26017 + }, + { + "epoch": 1.92, + "learning_rate": 1.589287989714427e-05, + "loss": 1.0229, + "step": 26018 + }, + { + "epoch": 1.92, + "learning_rate": 1.589255763487823e-05, + "loss": 1.0006, + "step": 26019 + }, + { + "epoch": 1.92, + "learning_rate": 1.589223536323721e-05, + "loss": 1.0371, + "step": 26020 + }, + { + "epoch": 1.92, + "learning_rate": 1.589191308222173e-05, + "loss": 1.0293, + "step": 26021 + }, + { + "epoch": 1.92, + "learning_rate": 1.58915907918323e-05, + "loss": 1.0193, + "step": 26022 + }, + { + "epoch": 1.92, + "learning_rate": 1.5891268492069432e-05, + "loss": 1.0093, + "step": 26023 + }, + { + "epoch": 1.92, + "learning_rate": 1.5890946182933644e-05, + "loss": 1.0317, + "step": 26024 + }, + { + "epoch": 1.92, + "learning_rate": 1.589062386442544e-05, + "loss": 0.85, + "step": 26025 + }, + { + "epoch": 1.92, + "learning_rate": 1.589030153654534e-05, + "loss": 0.9664, + "step": 26026 + }, + { + "epoch": 1.92, + "learning_rate": 1.5889979199293857e-05, + "loss": 0.954, + "step": 26027 + }, + { + "epoch": 1.92, + "learning_rate": 1.5889656852671498e-05, + "loss": 1.0445, + "step": 26028 + }, + { + "epoch": 1.92, + "learning_rate": 1.5889334496678782e-05, + "loss": 1.0139, + "step": 26029 + }, + { + "epoch": 1.92, + "learning_rate": 1.588901213131622e-05, + "loss": 1.0659, + "step": 26030 + }, + { + "epoch": 1.92, + "learning_rate": 1.5888689756584315e-05, + "loss": 1.0309, + "step": 26031 + }, + { + "epoch": 1.92, + "learning_rate": 1.5888367372483597e-05, + "loss": 1.0776, + "step": 26032 + }, + { + "epoch": 1.92, + "learning_rate": 1.5888044979014566e-05, + "loss": 0.9736, + "step": 26033 + }, + { + "epoch": 1.92, + "learning_rate": 1.5887722576177747e-05, + "loss": 1.0452, + "step": 26034 + }, + { + "epoch": 1.92, + "learning_rate": 1.5887400163973642e-05, + "loss": 0.9928, + "step": 26035 + }, + { + "epoch": 1.92, + "learning_rate": 1.5887077742402765e-05, + "loss": 1.072, + "step": 26036 + }, + { + "epoch": 1.92, + "learning_rate": 1.5886755311465635e-05, + "loss": 1.034, + "step": 26037 + }, + { + "epoch": 1.92, + "learning_rate": 1.588643287116276e-05, + "loss": 0.8904, + "step": 26038 + }, + { + "epoch": 1.92, + "learning_rate": 1.588611042149465e-05, + "loss": 1.0397, + "step": 26039 + }, + { + "epoch": 1.92, + "learning_rate": 1.588578796246183e-05, + "loss": 0.9939, + "step": 26040 + }, + { + "epoch": 1.92, + "learning_rate": 1.58854654940648e-05, + "loss": 1.1304, + "step": 26041 + }, + { + "epoch": 1.92, + "learning_rate": 1.5885143016304082e-05, + "loss": 0.9989, + "step": 26042 + }, + { + "epoch": 1.92, + "learning_rate": 1.5884820529180184e-05, + "loss": 1.0411, + "step": 26043 + }, + { + "epoch": 1.92, + "learning_rate": 1.5884498032693627e-05, + "loss": 0.9385, + "step": 26044 + }, + { + "epoch": 1.92, + "learning_rate": 1.588417552684491e-05, + "loss": 1.0521, + "step": 26045 + }, + { + "epoch": 1.92, + "learning_rate": 1.5883853011634557e-05, + "loss": 0.9872, + "step": 26046 + }, + { + "epoch": 1.92, + "learning_rate": 1.5883530487063077e-05, + "loss": 0.9411, + "step": 26047 + }, + { + "epoch": 1.92, + "learning_rate": 1.5883207953130983e-05, + "loss": 1.0146, + "step": 26048 + }, + { + "epoch": 1.92, + "learning_rate": 1.588288540983879e-05, + "loss": 1.0199, + "step": 26049 + }, + { + "epoch": 1.92, + "learning_rate": 1.588256285718701e-05, + "loss": 0.9582, + "step": 26050 + }, + { + "epoch": 1.92, + "learning_rate": 1.588224029517616e-05, + "loss": 1.004, + "step": 26051 + }, + { + "epoch": 1.92, + "learning_rate": 1.5881917723806748e-05, + "loss": 1.0252, + "step": 26052 + }, + { + "epoch": 1.93, + "learning_rate": 1.5881595143079288e-05, + "loss": 1.0385, + "step": 26053 + }, + { + "epoch": 1.93, + "learning_rate": 1.58812725529943e-05, + "loss": 1.033, + "step": 26054 + }, + { + "epoch": 1.93, + "learning_rate": 1.5880949953552283e-05, + "loss": 0.9462, + "step": 26055 + }, + { + "epoch": 1.93, + "learning_rate": 1.5880627344753762e-05, + "loss": 1.0471, + "step": 26056 + }, + { + "epoch": 1.93, + "learning_rate": 1.5880304726599247e-05, + "loss": 0.9678, + "step": 26057 + }, + { + "epoch": 1.93, + "learning_rate": 1.5879982099089253e-05, + "loss": 0.9023, + "step": 26058 + }, + { + "epoch": 1.93, + "learning_rate": 1.587965946222429e-05, + "loss": 1.0913, + "step": 26059 + }, + { + "epoch": 1.93, + "learning_rate": 1.5879336816004874e-05, + "loss": 0.9637, + "step": 26060 + }, + { + "epoch": 1.93, + "learning_rate": 1.5879014160431513e-05, + "loss": 1.0138, + "step": 26061 + }, + { + "epoch": 1.93, + "learning_rate": 1.587869149550473e-05, + "loss": 1.014, + "step": 26062 + }, + { + "epoch": 1.93, + "learning_rate": 1.587836882122503e-05, + "loss": 1.0881, + "step": 26063 + }, + { + "epoch": 1.93, + "learning_rate": 1.5878046137592932e-05, + "loss": 1.0173, + "step": 26064 + }, + { + "epoch": 1.93, + "learning_rate": 1.5877723444608944e-05, + "loss": 0.9999, + "step": 26065 + }, + { + "epoch": 1.93, + "learning_rate": 1.5877400742273584e-05, + "loss": 0.9194, + "step": 26066 + }, + { + "epoch": 1.93, + "learning_rate": 1.5877078030587363e-05, + "loss": 0.8665, + "step": 26067 + }, + { + "epoch": 1.93, + "learning_rate": 1.5876755309550796e-05, + "loss": 0.9115, + "step": 26068 + }, + { + "epoch": 1.93, + "learning_rate": 1.5876432579164392e-05, + "loss": 1.0454, + "step": 26069 + }, + { + "epoch": 1.93, + "learning_rate": 1.587610983942867e-05, + "loss": 1.0597, + "step": 26070 + }, + { + "epoch": 1.93, + "learning_rate": 1.587578709034414e-05, + "loss": 0.9784, + "step": 26071 + }, + { + "epoch": 1.93, + "learning_rate": 1.587546433191132e-05, + "loss": 1.0182, + "step": 26072 + }, + { + "epoch": 1.93, + "learning_rate": 1.5875141564130714e-05, + "loss": 1.0356, + "step": 26073 + }, + { + "epoch": 1.93, + "learning_rate": 1.587481878700285e-05, + "loss": 1.135, + "step": 26074 + }, + { + "epoch": 1.93, + "learning_rate": 1.5874496000528228e-05, + "loss": 0.9573, + "step": 26075 + }, + { + "epoch": 1.93, + "learning_rate": 1.587417320470737e-05, + "loss": 0.9748, + "step": 26076 + }, + { + "epoch": 1.93, + "learning_rate": 1.5873850399540783e-05, + "loss": 1.0298, + "step": 26077 + }, + { + "epoch": 1.93, + "learning_rate": 1.5873527585028984e-05, + "loss": 0.9465, + "step": 26078 + }, + { + "epoch": 1.93, + "learning_rate": 1.587320476117249e-05, + "loss": 0.9208, + "step": 26079 + }, + { + "epoch": 1.93, + "learning_rate": 1.587288192797181e-05, + "loss": 1.1745, + "step": 26080 + }, + { + "epoch": 1.93, + "learning_rate": 1.587255908542746e-05, + "loss": 1.0224, + "step": 26081 + }, + { + "epoch": 1.93, + "learning_rate": 1.5872236233539952e-05, + "loss": 1.0003, + "step": 26082 + }, + { + "epoch": 1.93, + "learning_rate": 1.58719133723098e-05, + "loss": 1.0065, + "step": 26083 + }, + { + "epoch": 1.93, + "learning_rate": 1.587159050173752e-05, + "loss": 1.0713, + "step": 26084 + }, + { + "epoch": 1.93, + "learning_rate": 1.5871267621823615e-05, + "loss": 0.945, + "step": 26085 + }, + { + "epoch": 1.93, + "learning_rate": 1.5870944732568612e-05, + "loss": 1.0357, + "step": 26086 + }, + { + "epoch": 1.93, + "learning_rate": 1.5870621833973025e-05, + "loss": 0.9766, + "step": 26087 + }, + { + "epoch": 1.93, + "learning_rate": 1.5870298926037358e-05, + "loss": 1.1042, + "step": 26088 + }, + { + "epoch": 1.93, + "learning_rate": 1.5869976008762125e-05, + "loss": 1.0283, + "step": 26089 + }, + { + "epoch": 1.93, + "learning_rate": 1.5869653082147854e-05, + "loss": 0.9802, + "step": 26090 + }, + { + "epoch": 1.93, + "learning_rate": 1.5869330146195043e-05, + "loss": 0.9057, + "step": 26091 + }, + { + "epoch": 1.93, + "learning_rate": 1.586900720090421e-05, + "loss": 1.0242, + "step": 26092 + }, + { + "epoch": 1.93, + "learning_rate": 1.5868684246275875e-05, + "loss": 0.9887, + "step": 26093 + }, + { + "epoch": 1.93, + "learning_rate": 1.5868361282310547e-05, + "loss": 1.0478, + "step": 26094 + }, + { + "epoch": 1.93, + "learning_rate": 1.5868038309008737e-05, + "loss": 1.0515, + "step": 26095 + }, + { + "epoch": 1.93, + "learning_rate": 1.5867715326370964e-05, + "loss": 0.9593, + "step": 26096 + }, + { + "epoch": 1.93, + "learning_rate": 1.586739233439774e-05, + "loss": 0.9833, + "step": 26097 + }, + { + "epoch": 1.93, + "learning_rate": 1.586706933308958e-05, + "loss": 1.1057, + "step": 26098 + }, + { + "epoch": 1.93, + "learning_rate": 1.5866746322446992e-05, + "loss": 1.0229, + "step": 26099 + }, + { + "epoch": 1.93, + "learning_rate": 1.5866423302470503e-05, + "loss": 1.0222, + "step": 26100 + }, + { + "epoch": 1.93, + "learning_rate": 1.586610027316061e-05, + "loss": 1.0715, + "step": 26101 + }, + { + "epoch": 1.93, + "learning_rate": 1.586577723451784e-05, + "loss": 1.0305, + "step": 26102 + }, + { + "epoch": 1.93, + "learning_rate": 1.58654541865427e-05, + "loss": 1.0435, + "step": 26103 + }, + { + "epoch": 1.93, + "learning_rate": 1.5865131129235706e-05, + "loss": 0.9868, + "step": 26104 + }, + { + "epoch": 1.93, + "learning_rate": 1.586480806259737e-05, + "loss": 0.9782, + "step": 26105 + }, + { + "epoch": 1.93, + "learning_rate": 1.5864484986628215e-05, + "loss": 0.9413, + "step": 26106 + }, + { + "epoch": 1.93, + "learning_rate": 1.586416190132874e-05, + "loss": 0.9695, + "step": 26107 + }, + { + "epoch": 1.93, + "learning_rate": 1.5863838806699477e-05, + "loss": 0.945, + "step": 26108 + }, + { + "epoch": 1.93, + "learning_rate": 1.5863515702740925e-05, + "loss": 0.9056, + "step": 26109 + }, + { + "epoch": 1.93, + "learning_rate": 1.5863192589453605e-05, + "loss": 1.0181, + "step": 26110 + }, + { + "epoch": 1.93, + "learning_rate": 1.586286946683803e-05, + "loss": 1.0687, + "step": 26111 + }, + { + "epoch": 1.93, + "learning_rate": 1.586254633489471e-05, + "loss": 0.9863, + "step": 26112 + }, + { + "epoch": 1.93, + "learning_rate": 1.5862223193624164e-05, + "loss": 1.0287, + "step": 26113 + }, + { + "epoch": 1.93, + "learning_rate": 1.5861900043026907e-05, + "loss": 0.9652, + "step": 26114 + }, + { + "epoch": 1.93, + "learning_rate": 1.5861576883103447e-05, + "loss": 0.9385, + "step": 26115 + }, + { + "epoch": 1.93, + "learning_rate": 1.5861253713854304e-05, + "loss": 1.043, + "step": 26116 + }, + { + "epoch": 1.93, + "learning_rate": 1.586093053527999e-05, + "loss": 0.8922, + "step": 26117 + }, + { + "epoch": 1.93, + "learning_rate": 1.5860607347381023e-05, + "loss": 0.944, + "step": 26118 + }, + { + "epoch": 1.93, + "learning_rate": 1.586028415015791e-05, + "loss": 0.9791, + "step": 26119 + }, + { + "epoch": 1.93, + "learning_rate": 1.585996094361117e-05, + "loss": 1.0655, + "step": 26120 + }, + { + "epoch": 1.93, + "learning_rate": 1.5859637727741314e-05, + "loss": 0.9732, + "step": 26121 + }, + { + "epoch": 1.93, + "learning_rate": 1.585931450254886e-05, + "loss": 0.9575, + "step": 26122 + }, + { + "epoch": 1.93, + "learning_rate": 1.5858991268034317e-05, + "loss": 1.0185, + "step": 26123 + }, + { + "epoch": 1.93, + "learning_rate": 1.5858668024198207e-05, + "loss": 1.0716, + "step": 26124 + }, + { + "epoch": 1.93, + "learning_rate": 1.585834477104104e-05, + "loss": 0.9975, + "step": 26125 + }, + { + "epoch": 1.93, + "learning_rate": 1.5858021508563332e-05, + "loss": 0.9484, + "step": 26126 + }, + { + "epoch": 1.93, + "learning_rate": 1.585769823676559e-05, + "loss": 0.9904, + "step": 26127 + }, + { + "epoch": 1.93, + "learning_rate": 1.585737495564834e-05, + "loss": 1.0303, + "step": 26128 + }, + { + "epoch": 1.93, + "learning_rate": 1.5857051665212084e-05, + "loss": 1.0551, + "step": 26129 + }, + { + "epoch": 1.93, + "learning_rate": 1.5856728365457342e-05, + "loss": 0.9817, + "step": 26130 + }, + { + "epoch": 1.93, + "learning_rate": 1.5856405056384635e-05, + "loss": 0.9646, + "step": 26131 + }, + { + "epoch": 1.93, + "learning_rate": 1.5856081737994467e-05, + "loss": 0.9978, + "step": 26132 + }, + { + "epoch": 1.93, + "learning_rate": 1.585575841028736e-05, + "loss": 0.9407, + "step": 26133 + }, + { + "epoch": 1.93, + "learning_rate": 1.5855435073263823e-05, + "loss": 1.0126, + "step": 26134 + }, + { + "epoch": 1.93, + "learning_rate": 1.5855111726924373e-05, + "loss": 1.0901, + "step": 26135 + }, + { + "epoch": 1.93, + "learning_rate": 1.5854788371269526e-05, + "loss": 1.0394, + "step": 26136 + }, + { + "epoch": 1.93, + "learning_rate": 1.585446500629979e-05, + "loss": 0.9737, + "step": 26137 + }, + { + "epoch": 1.93, + "learning_rate": 1.585414163201569e-05, + "loss": 1.0437, + "step": 26138 + }, + { + "epoch": 1.93, + "learning_rate": 1.585381824841773e-05, + "loss": 1.031, + "step": 26139 + }, + { + "epoch": 1.93, + "learning_rate": 1.5853494855506428e-05, + "loss": 0.9137, + "step": 26140 + }, + { + "epoch": 1.93, + "learning_rate": 1.5853171453282302e-05, + "loss": 1.1009, + "step": 26141 + }, + { + "epoch": 1.93, + "learning_rate": 1.5852848041745863e-05, + "loss": 1.0769, + "step": 26142 + }, + { + "epoch": 1.93, + "learning_rate": 1.5852524620897627e-05, + "loss": 1.0971, + "step": 26143 + }, + { + "epoch": 1.93, + "learning_rate": 1.5852201190738104e-05, + "loss": 0.9731, + "step": 26144 + }, + { + "epoch": 1.93, + "learning_rate": 1.585187775126782e-05, + "loss": 1.0804, + "step": 26145 + }, + { + "epoch": 1.93, + "learning_rate": 1.585155430248728e-05, + "loss": 1.0524, + "step": 26146 + }, + { + "epoch": 1.93, + "learning_rate": 1.5851230844397e-05, + "loss": 1.0396, + "step": 26147 + }, + { + "epoch": 1.93, + "learning_rate": 1.5850907376997495e-05, + "loss": 1.0681, + "step": 26148 + }, + { + "epoch": 1.93, + "learning_rate": 1.585058390028928e-05, + "loss": 1.1086, + "step": 26149 + }, + { + "epoch": 1.93, + "learning_rate": 1.5850260414272868e-05, + "loss": 1.0703, + "step": 26150 + }, + { + "epoch": 1.93, + "learning_rate": 1.584993691894878e-05, + "loss": 1.0809, + "step": 26151 + }, + { + "epoch": 1.93, + "learning_rate": 1.5849613414317523e-05, + "loss": 1.0584, + "step": 26152 + }, + { + "epoch": 1.93, + "learning_rate": 1.584928990037961e-05, + "loss": 1.0695, + "step": 26153 + }, + { + "epoch": 1.93, + "learning_rate": 1.584896637713557e-05, + "loss": 1.0478, + "step": 26154 + }, + { + "epoch": 1.93, + "learning_rate": 1.58486428445859e-05, + "loss": 0.9935, + "step": 26155 + }, + { + "epoch": 1.93, + "learning_rate": 1.584831930273113e-05, + "loss": 0.9865, + "step": 26156 + }, + { + "epoch": 1.93, + "learning_rate": 1.5847995751571763e-05, + "loss": 0.9923, + "step": 26157 + }, + { + "epoch": 1.93, + "learning_rate": 1.584767219110832e-05, + "loss": 1.0225, + "step": 26158 + }, + { + "epoch": 1.93, + "learning_rate": 1.5847348621341314e-05, + "loss": 1.057, + "step": 26159 + }, + { + "epoch": 1.93, + "learning_rate": 1.584702504227126e-05, + "loss": 0.9918, + "step": 26160 + }, + { + "epoch": 1.93, + "learning_rate": 1.5846701453898673e-05, + "loss": 1.0815, + "step": 26161 + }, + { + "epoch": 1.93, + "learning_rate": 1.5846377856224066e-05, + "loss": 0.9997, + "step": 26162 + }, + { + "epoch": 1.93, + "learning_rate": 1.5846054249247957e-05, + "loss": 0.9572, + "step": 26163 + }, + { + "epoch": 1.93, + "learning_rate": 1.5845730632970857e-05, + "loss": 1.0272, + "step": 26164 + }, + { + "epoch": 1.93, + "learning_rate": 1.5845407007393288e-05, + "loss": 0.9166, + "step": 26165 + }, + { + "epoch": 1.93, + "learning_rate": 1.5845083372515758e-05, + "loss": 1.0227, + "step": 26166 + }, + { + "epoch": 1.93, + "learning_rate": 1.5844759728338782e-05, + "loss": 0.9545, + "step": 26167 + }, + { + "epoch": 1.93, + "learning_rate": 1.5844436074862878e-05, + "loss": 1.0278, + "step": 26168 + }, + { + "epoch": 1.93, + "learning_rate": 1.584411241208856e-05, + "loss": 0.9944, + "step": 26169 + }, + { + "epoch": 1.93, + "learning_rate": 1.5843788740016344e-05, + "loss": 1.0712, + "step": 26170 + }, + { + "epoch": 1.93, + "learning_rate": 1.5843465058646744e-05, + "loss": 1.0823, + "step": 26171 + }, + { + "epoch": 1.93, + "learning_rate": 1.5843141367980274e-05, + "loss": 0.9458, + "step": 26172 + }, + { + "epoch": 1.93, + "learning_rate": 1.5842817668017447e-05, + "loss": 1.0163, + "step": 26173 + }, + { + "epoch": 1.93, + "learning_rate": 1.5842493958758782e-05, + "loss": 1.072, + "step": 26174 + }, + { + "epoch": 1.93, + "learning_rate": 1.584217024020479e-05, + "loss": 1.0265, + "step": 26175 + }, + { + "epoch": 1.93, + "learning_rate": 1.5841846512355993e-05, + "loss": 1.0461, + "step": 26176 + }, + { + "epoch": 1.93, + "learning_rate": 1.5841522775212903e-05, + "loss": 0.9032, + "step": 26177 + }, + { + "epoch": 1.93, + "learning_rate": 1.5841199028776034e-05, + "loss": 1.0406, + "step": 26178 + }, + { + "epoch": 1.93, + "learning_rate": 1.5840875273045896e-05, + "loss": 1.0394, + "step": 26179 + }, + { + "epoch": 1.93, + "learning_rate": 1.584055150802301e-05, + "loss": 0.9934, + "step": 26180 + }, + { + "epoch": 1.93, + "learning_rate": 1.5840227733707893e-05, + "loss": 1.0422, + "step": 26181 + }, + { + "epoch": 1.93, + "learning_rate": 1.5839903950101057e-05, + "loss": 0.9912, + "step": 26182 + }, + { + "epoch": 1.93, + "learning_rate": 1.5839580157203015e-05, + "loss": 0.9993, + "step": 26183 + }, + { + "epoch": 1.93, + "learning_rate": 1.583925635501429e-05, + "loss": 1.15, + "step": 26184 + }, + { + "epoch": 1.93, + "learning_rate": 1.5838932543535387e-05, + "loss": 1.0827, + "step": 26185 + }, + { + "epoch": 1.93, + "learning_rate": 1.5838608722766827e-05, + "loss": 1.1101, + "step": 26186 + }, + { + "epoch": 1.93, + "learning_rate": 1.583828489270912e-05, + "loss": 1.0197, + "step": 26187 + }, + { + "epoch": 1.94, + "learning_rate": 1.583796105336279e-05, + "loss": 1.0642, + "step": 26188 + }, + { + "epoch": 1.94, + "learning_rate": 1.5837637204728348e-05, + "loss": 1.0092, + "step": 26189 + }, + { + "epoch": 1.94, + "learning_rate": 1.5837313346806304e-05, + "loss": 1.058, + "step": 26190 + }, + { + "epoch": 1.94, + "learning_rate": 1.583698947959718e-05, + "loss": 0.9835, + "step": 26191 + }, + { + "epoch": 1.94, + "learning_rate": 1.583666560310149e-05, + "loss": 0.9756, + "step": 26192 + }, + { + "epoch": 1.94, + "learning_rate": 1.5836341717319753e-05, + "loss": 1.0614, + "step": 26193 + }, + { + "epoch": 1.94, + "learning_rate": 1.5836017822252475e-05, + "loss": 0.9022, + "step": 26194 + }, + { + "epoch": 1.94, + "learning_rate": 1.5835693917900178e-05, + "loss": 1.0321, + "step": 26195 + }, + { + "epoch": 1.94, + "learning_rate": 1.583537000426337e-05, + "loss": 1.0307, + "step": 26196 + }, + { + "epoch": 1.94, + "learning_rate": 1.583504608134258e-05, + "loss": 1.0038, + "step": 26197 + }, + { + "epoch": 1.94, + "learning_rate": 1.5834722149138307e-05, + "loss": 0.9142, + "step": 26198 + }, + { + "epoch": 1.94, + "learning_rate": 1.583439820765108e-05, + "loss": 1.0323, + "step": 26199 + }, + { + "epoch": 1.94, + "learning_rate": 1.5834074256881407e-05, + "loss": 1.0033, + "step": 26200 + }, + { + "epoch": 1.94, + "learning_rate": 1.5833750296829805e-05, + "loss": 1.009, + "step": 26201 + }, + { + "epoch": 1.94, + "learning_rate": 1.5833426327496793e-05, + "loss": 1.0439, + "step": 26202 + }, + { + "epoch": 1.94, + "learning_rate": 1.583310234888288e-05, + "loss": 1.0608, + "step": 26203 + }, + { + "epoch": 1.94, + "learning_rate": 1.5832778360988588e-05, + "loss": 0.876, + "step": 26204 + }, + { + "epoch": 1.94, + "learning_rate": 1.5832454363814426e-05, + "loss": 1.0225, + "step": 26205 + }, + { + "epoch": 1.94, + "learning_rate": 1.5832130357360913e-05, + "loss": 0.9488, + "step": 26206 + }, + { + "epoch": 1.94, + "learning_rate": 1.5831806341628563e-05, + "loss": 0.9244, + "step": 26207 + }, + { + "epoch": 1.94, + "learning_rate": 1.5831482316617893e-05, + "loss": 0.9449, + "step": 26208 + }, + { + "epoch": 1.94, + "learning_rate": 1.583115828232942e-05, + "loss": 1.0174, + "step": 26209 + }, + { + "epoch": 1.94, + "learning_rate": 1.5830834238763654e-05, + "loss": 0.9953, + "step": 26210 + }, + { + "epoch": 1.94, + "learning_rate": 1.5830510185921117e-05, + "loss": 0.9636, + "step": 26211 + }, + { + "epoch": 1.94, + "learning_rate": 1.5830186123802323e-05, + "loss": 0.9709, + "step": 26212 + }, + { + "epoch": 1.94, + "learning_rate": 1.5829862052407784e-05, + "loss": 0.8337, + "step": 26213 + }, + { + "epoch": 1.94, + "learning_rate": 1.582953797173802e-05, + "loss": 1.1058, + "step": 26214 + }, + { + "epoch": 1.94, + "learning_rate": 1.5829213881793543e-05, + "loss": 1.0795, + "step": 26215 + }, + { + "epoch": 1.94, + "learning_rate": 1.582888978257487e-05, + "loss": 1.093, + "step": 26216 + }, + { + "epoch": 1.94, + "learning_rate": 1.5828565674082518e-05, + "loss": 1.0372, + "step": 26217 + }, + { + "epoch": 1.94, + "learning_rate": 1.5828241556317e-05, + "loss": 1.0357, + "step": 26218 + }, + { + "epoch": 1.94, + "learning_rate": 1.5827917429278833e-05, + "loss": 1.0051, + "step": 26219 + }, + { + "epoch": 1.94, + "learning_rate": 1.5827593292968532e-05, + "loss": 1.0532, + "step": 26220 + }, + { + "epoch": 1.94, + "learning_rate": 1.5827269147386615e-05, + "loss": 0.9463, + "step": 26221 + }, + { + "epoch": 1.94, + "learning_rate": 1.5826944992533597e-05, + "loss": 1.018, + "step": 26222 + }, + { + "epoch": 1.94, + "learning_rate": 1.582662082840999e-05, + "loss": 0.8979, + "step": 26223 + }, + { + "epoch": 1.94, + "learning_rate": 1.5826296655016315e-05, + "loss": 0.9989, + "step": 26224 + }, + { + "epoch": 1.94, + "learning_rate": 1.5825972472353084e-05, + "loss": 1.0161, + "step": 26225 + }, + { + "epoch": 1.94, + "learning_rate": 1.5825648280420812e-05, + "loss": 0.9385, + "step": 26226 + }, + { + "epoch": 1.94, + "learning_rate": 1.582532407922002e-05, + "loss": 1.0549, + "step": 26227 + }, + { + "epoch": 1.94, + "learning_rate": 1.582499986875122e-05, + "loss": 0.9942, + "step": 26228 + }, + { + "epoch": 1.94, + "learning_rate": 1.5824675649014933e-05, + "loss": 0.9755, + "step": 26229 + }, + { + "epoch": 1.94, + "learning_rate": 1.5824351420011665e-05, + "loss": 1.0389, + "step": 26230 + }, + { + "epoch": 1.94, + "learning_rate": 1.582402718174194e-05, + "loss": 0.9833, + "step": 26231 + }, + { + "epoch": 1.94, + "learning_rate": 1.582370293420627e-05, + "loss": 1.0132, + "step": 26232 + }, + { + "epoch": 1.94, + "learning_rate": 1.582337867740517e-05, + "loss": 0.9809, + "step": 26233 + }, + { + "epoch": 1.94, + "learning_rate": 1.582305441133916e-05, + "loss": 1.0525, + "step": 26234 + }, + { + "epoch": 1.94, + "learning_rate": 1.5822730136008753e-05, + "loss": 0.9652, + "step": 26235 + }, + { + "epoch": 1.94, + "learning_rate": 1.5822405851414466e-05, + "loss": 0.9641, + "step": 26236 + }, + { + "epoch": 1.94, + "learning_rate": 1.5822081557556815e-05, + "loss": 0.9521, + "step": 26237 + }, + { + "epoch": 1.94, + "learning_rate": 1.5821757254436314e-05, + "loss": 1.0696, + "step": 26238 + }, + { + "epoch": 1.94, + "learning_rate": 1.582143294205348e-05, + "loss": 1.0337, + "step": 26239 + }, + { + "epoch": 1.94, + "learning_rate": 1.5821108620408834e-05, + "loss": 0.9222, + "step": 26240 + }, + { + "epoch": 1.94, + "learning_rate": 1.5820784289502884e-05, + "loss": 1.0035, + "step": 26241 + }, + { + "epoch": 1.94, + "learning_rate": 1.582045994933615e-05, + "loss": 1.0005, + "step": 26242 + }, + { + "epoch": 1.94, + "learning_rate": 1.5820135599909147e-05, + "loss": 0.9668, + "step": 26243 + }, + { + "epoch": 1.94, + "learning_rate": 1.581981124122239e-05, + "loss": 1.0235, + "step": 26244 + }, + { + "epoch": 1.94, + "learning_rate": 1.5819486873276398e-05, + "loss": 1.0474, + "step": 26245 + }, + { + "epoch": 1.94, + "learning_rate": 1.5819162496071688e-05, + "loss": 0.9633, + "step": 26246 + }, + { + "epoch": 1.94, + "learning_rate": 1.5818838109608773e-05, + "loss": 0.9303, + "step": 26247 + }, + { + "epoch": 1.94, + "learning_rate": 1.5818513713888167e-05, + "loss": 0.9537, + "step": 26248 + }, + { + "epoch": 1.94, + "learning_rate": 1.5818189308910392e-05, + "loss": 0.9623, + "step": 26249 + }, + { + "epoch": 1.94, + "learning_rate": 1.581786489467596e-05, + "loss": 1.0645, + "step": 26250 + }, + { + "epoch": 1.94, + "learning_rate": 1.5817540471185386e-05, + "loss": 1.0045, + "step": 26251 + }, + { + "epoch": 1.94, + "learning_rate": 1.5817216038439194e-05, + "loss": 1.0615, + "step": 26252 + }, + { + "epoch": 1.94, + "learning_rate": 1.581689159643789e-05, + "loss": 1.0485, + "step": 26253 + }, + { + "epoch": 1.94, + "learning_rate": 1.5816567145181993e-05, + "loss": 0.9889, + "step": 26254 + }, + { + "epoch": 1.94, + "learning_rate": 1.5816242684672026e-05, + "loss": 0.9669, + "step": 26255 + }, + { + "epoch": 1.94, + "learning_rate": 1.5815918214908496e-05, + "loss": 0.9714, + "step": 26256 + }, + { + "epoch": 1.94, + "learning_rate": 1.5815593735891927e-05, + "loss": 0.9632, + "step": 26257 + }, + { + "epoch": 1.94, + "learning_rate": 1.5815269247622828e-05, + "loss": 0.9889, + "step": 26258 + }, + { + "epoch": 1.94, + "learning_rate": 1.581494475010172e-05, + "loss": 0.9519, + "step": 26259 + }, + { + "epoch": 1.94, + "learning_rate": 1.581462024332912e-05, + "loss": 1.0701, + "step": 26260 + }, + { + "epoch": 1.94, + "learning_rate": 1.5814295727305536e-05, + "loss": 1.0019, + "step": 26261 + }, + { + "epoch": 1.94, + "learning_rate": 1.58139712020315e-05, + "loss": 1.0483, + "step": 26262 + }, + { + "epoch": 1.94, + "learning_rate": 1.5813646667507516e-05, + "loss": 0.9429, + "step": 26263 + }, + { + "epoch": 1.94, + "learning_rate": 1.58133221237341e-05, + "loss": 1.0787, + "step": 26264 + }, + { + "epoch": 1.94, + "learning_rate": 1.581299757071177e-05, + "loss": 0.9116, + "step": 26265 + }, + { + "epoch": 1.94, + "learning_rate": 1.5812673008441047e-05, + "loss": 1.0831, + "step": 26266 + }, + { + "epoch": 1.94, + "learning_rate": 1.5812348436922446e-05, + "loss": 1.0026, + "step": 26267 + }, + { + "epoch": 1.94, + "learning_rate": 1.581202385615648e-05, + "loss": 0.9677, + "step": 26268 + }, + { + "epoch": 1.94, + "learning_rate": 1.5811699266143668e-05, + "loss": 1.0135, + "step": 26269 + }, + { + "epoch": 1.94, + "learning_rate": 1.5811374666884528e-05, + "loss": 0.9945, + "step": 26270 + }, + { + "epoch": 1.94, + "learning_rate": 1.581105005837957e-05, + "loss": 0.9908, + "step": 26271 + }, + { + "epoch": 1.94, + "learning_rate": 1.5810725440629317e-05, + "loss": 0.946, + "step": 26272 + }, + { + "epoch": 1.94, + "learning_rate": 1.581040081363428e-05, + "loss": 1.0325, + "step": 26273 + }, + { + "epoch": 1.94, + "learning_rate": 1.5810076177394978e-05, + "loss": 0.9373, + "step": 26274 + }, + { + "epoch": 1.94, + "learning_rate": 1.5809751531911932e-05, + "loss": 0.967, + "step": 26275 + }, + { + "epoch": 1.94, + "learning_rate": 1.5809426877185652e-05, + "loss": 0.962, + "step": 26276 + }, + { + "epoch": 1.94, + "learning_rate": 1.580910221321666e-05, + "loss": 0.9883, + "step": 26277 + }, + { + "epoch": 1.94, + "learning_rate": 1.5808777540005466e-05, + "loss": 0.8993, + "step": 26278 + }, + { + "epoch": 1.94, + "learning_rate": 1.5808452857552593e-05, + "loss": 0.9846, + "step": 26279 + }, + { + "epoch": 1.94, + "learning_rate": 1.5808128165858553e-05, + "loss": 0.9796, + "step": 26280 + }, + { + "epoch": 1.94, + "learning_rate": 1.5807803464923862e-05, + "loss": 0.9476, + "step": 26281 + }, + { + "epoch": 1.94, + "learning_rate": 1.5807478754749044e-05, + "loss": 1.0133, + "step": 26282 + }, + { + "epoch": 1.94, + "learning_rate": 1.5807154035334607e-05, + "loss": 0.8923, + "step": 26283 + }, + { + "epoch": 1.94, + "learning_rate": 1.580682930668107e-05, + "loss": 0.9709, + "step": 26284 + }, + { + "epoch": 1.94, + "learning_rate": 1.5806504568788953e-05, + "loss": 1.0875, + "step": 26285 + }, + { + "epoch": 1.94, + "learning_rate": 1.5806179821658768e-05, + "loss": 1.0128, + "step": 26286 + }, + { + "epoch": 1.94, + "learning_rate": 1.5805855065291035e-05, + "loss": 0.96, + "step": 26287 + }, + { + "epoch": 1.94, + "learning_rate": 1.580553029968627e-05, + "loss": 0.9696, + "step": 26288 + }, + { + "epoch": 1.94, + "learning_rate": 1.580520552484499e-05, + "loss": 1.0333, + "step": 26289 + }, + { + "epoch": 1.94, + "learning_rate": 1.580488074076771e-05, + "loss": 1.095, + "step": 26290 + }, + { + "epoch": 1.94, + "learning_rate": 1.580455594745495e-05, + "loss": 0.9545, + "step": 26291 + }, + { + "epoch": 1.94, + "learning_rate": 1.580423114490722e-05, + "loss": 1.0315, + "step": 26292 + }, + { + "epoch": 1.94, + "learning_rate": 1.5803906333125044e-05, + "loss": 0.9095, + "step": 26293 + }, + { + "epoch": 1.94, + "learning_rate": 1.5803581512108937e-05, + "loss": 0.983, + "step": 26294 + }, + { + "epoch": 1.94, + "learning_rate": 1.580325668185941e-05, + "loss": 1.045, + "step": 26295 + }, + { + "epoch": 1.94, + "learning_rate": 1.5802931842376992e-05, + "loss": 0.9034, + "step": 26296 + }, + { + "epoch": 1.94, + "learning_rate": 1.5802606993662187e-05, + "loss": 1.0582, + "step": 26297 + }, + { + "epoch": 1.94, + "learning_rate": 1.5802282135715522e-05, + "loss": 0.9708, + "step": 26298 + }, + { + "epoch": 1.94, + "learning_rate": 1.5801957268537506e-05, + "loss": 1.0272, + "step": 26299 + }, + { + "epoch": 1.94, + "learning_rate": 1.5801632392128657e-05, + "loss": 0.8952, + "step": 26300 + }, + { + "epoch": 1.94, + "learning_rate": 1.5801307506489497e-05, + "loss": 1.0241, + "step": 26301 + }, + { + "epoch": 1.94, + "learning_rate": 1.580098261162054e-05, + "loss": 0.9153, + "step": 26302 + }, + { + "epoch": 1.94, + "learning_rate": 1.58006577075223e-05, + "loss": 1.0687, + "step": 26303 + }, + { + "epoch": 1.94, + "learning_rate": 1.5800332794195302e-05, + "loss": 0.9732, + "step": 26304 + }, + { + "epoch": 1.94, + "learning_rate": 1.5800007871640054e-05, + "loss": 1.0227, + "step": 26305 + }, + { + "epoch": 1.94, + "learning_rate": 1.5799682939857073e-05, + "loss": 1.0566, + "step": 26306 + }, + { + "epoch": 1.94, + "learning_rate": 1.5799357998846885e-05, + "loss": 0.9166, + "step": 26307 + }, + { + "epoch": 1.94, + "learning_rate": 1.579903304861e-05, + "loss": 1.047, + "step": 26308 + }, + { + "epoch": 1.94, + "learning_rate": 1.579870808914693e-05, + "loss": 0.9407, + "step": 26309 + }, + { + "epoch": 1.94, + "learning_rate": 1.5798383120458207e-05, + "loss": 0.8982, + "step": 26310 + }, + { + "epoch": 1.94, + "learning_rate": 1.5798058142544338e-05, + "loss": 1.1142, + "step": 26311 + }, + { + "epoch": 1.94, + "learning_rate": 1.579773315540584e-05, + "loss": 0.9734, + "step": 26312 + }, + { + "epoch": 1.94, + "learning_rate": 1.579740815904323e-05, + "loss": 0.984, + "step": 26313 + }, + { + "epoch": 1.94, + "learning_rate": 1.5797083153457025e-05, + "loss": 0.9898, + "step": 26314 + }, + { + "epoch": 1.94, + "learning_rate": 1.5796758138647746e-05, + "loss": 1.0367, + "step": 26315 + }, + { + "epoch": 1.94, + "learning_rate": 1.579643311461591e-05, + "loss": 1.0083, + "step": 26316 + }, + { + "epoch": 1.94, + "learning_rate": 1.5796108081362032e-05, + "loss": 0.883, + "step": 26317 + }, + { + "epoch": 1.94, + "learning_rate": 1.5795783038886624e-05, + "loss": 1.0517, + "step": 26318 + }, + { + "epoch": 1.94, + "learning_rate": 1.5795457987190212e-05, + "loss": 0.9872, + "step": 26319 + }, + { + "epoch": 1.94, + "learning_rate": 1.5795132926273308e-05, + "loss": 1.0252, + "step": 26320 + }, + { + "epoch": 1.94, + "learning_rate": 1.5794807856136433e-05, + "loss": 1.173, + "step": 26321 + }, + { + "epoch": 1.94, + "learning_rate": 1.57944827767801e-05, + "loss": 0.9892, + "step": 26322 + }, + { + "epoch": 1.94, + "learning_rate": 1.5794157688204825e-05, + "loss": 0.9045, + "step": 26323 + }, + { + "epoch": 1.95, + "learning_rate": 1.579383259041113e-05, + "loss": 1.0394, + "step": 26324 + }, + { + "epoch": 1.95, + "learning_rate": 1.579350748339953e-05, + "loss": 0.9607, + "step": 26325 + }, + { + "epoch": 1.95, + "learning_rate": 1.579318236717054e-05, + "loss": 0.9931, + "step": 26326 + }, + { + "epoch": 1.95, + "learning_rate": 1.5792857241724685e-05, + "loss": 1.0568, + "step": 26327 + }, + { + "epoch": 1.95, + "learning_rate": 1.5792532107062477e-05, + "loss": 0.9617, + "step": 26328 + }, + { + "epoch": 1.95, + "learning_rate": 1.5792206963184428e-05, + "loss": 1.1749, + "step": 26329 + }, + { + "epoch": 1.95, + "learning_rate": 1.5791881810091064e-05, + "loss": 0.9687, + "step": 26330 + }, + { + "epoch": 1.95, + "learning_rate": 1.57915566477829e-05, + "loss": 0.9168, + "step": 26331 + }, + { + "epoch": 1.95, + "learning_rate": 1.579123147626045e-05, + "loss": 0.9356, + "step": 26332 + }, + { + "epoch": 1.95, + "learning_rate": 1.5790906295524237e-05, + "loss": 1.04, + "step": 26333 + }, + { + "epoch": 1.95, + "learning_rate": 1.579058110557477e-05, + "loss": 1.0063, + "step": 26334 + }, + { + "epoch": 1.95, + "learning_rate": 1.5790255906412576e-05, + "loss": 0.904, + "step": 26335 + }, + { + "epoch": 1.95, + "learning_rate": 1.5789930698038163e-05, + "loss": 0.9626, + "step": 26336 + }, + { + "epoch": 1.95, + "learning_rate": 1.5789605480452056e-05, + "loss": 1.1014, + "step": 26337 + }, + { + "epoch": 1.95, + "learning_rate": 1.578928025365477e-05, + "loss": 1.0978, + "step": 26338 + }, + { + "epoch": 1.95, + "learning_rate": 1.578895501764682e-05, + "loss": 0.9567, + "step": 26339 + }, + { + "epoch": 1.95, + "learning_rate": 1.5788629772428724e-05, + "loss": 0.993, + "step": 26340 + }, + { + "epoch": 1.95, + "learning_rate": 1.5788304518001007e-05, + "loss": 1.0478, + "step": 26341 + }, + { + "epoch": 1.95, + "learning_rate": 1.578797925436417e-05, + "loss": 1.0558, + "step": 26342 + }, + { + "epoch": 1.95, + "learning_rate": 1.5787653981518753e-05, + "loss": 0.9318, + "step": 26343 + }, + { + "epoch": 1.95, + "learning_rate": 1.5787328699465256e-05, + "loss": 1.1095, + "step": 26344 + }, + { + "epoch": 1.95, + "learning_rate": 1.57870034082042e-05, + "loss": 0.9848, + "step": 26345 + }, + { + "epoch": 1.95, + "learning_rate": 1.5786678107736105e-05, + "loss": 0.9479, + "step": 26346 + }, + { + "epoch": 1.95, + "learning_rate": 1.5786352798061487e-05, + "loss": 0.9898, + "step": 26347 + }, + { + "epoch": 1.95, + "learning_rate": 1.5786027479180868e-05, + "loss": 1.0904, + "step": 26348 + }, + { + "epoch": 1.95, + "learning_rate": 1.5785702151094763e-05, + "loss": 1.0089, + "step": 26349 + }, + { + "epoch": 1.95, + "learning_rate": 1.5785376813803683e-05, + "loss": 0.9461, + "step": 26350 + }, + { + "epoch": 1.95, + "learning_rate": 1.5785051467308153e-05, + "loss": 1.1691, + "step": 26351 + }, + { + "epoch": 1.95, + "learning_rate": 1.5784726111608692e-05, + "loss": 0.9848, + "step": 26352 + }, + { + "epoch": 1.95, + "learning_rate": 1.5784400746705812e-05, + "loss": 0.8885, + "step": 26353 + }, + { + "epoch": 1.95, + "learning_rate": 1.5784075372600034e-05, + "loss": 1.0793, + "step": 26354 + }, + { + "epoch": 1.95, + "learning_rate": 1.5783749989291873e-05, + "loss": 0.8879, + "step": 26355 + }, + { + "epoch": 1.95, + "learning_rate": 1.578342459678185e-05, + "loss": 0.9943, + "step": 26356 + }, + { + "epoch": 1.95, + "learning_rate": 1.5783099195070484e-05, + "loss": 1.051, + "step": 26357 + }, + { + "epoch": 1.95, + "learning_rate": 1.5782773784158285e-05, + "loss": 0.9861, + "step": 26358 + }, + { + "epoch": 1.95, + "learning_rate": 1.578244836404578e-05, + "loss": 0.9508, + "step": 26359 + }, + { + "epoch": 1.95, + "learning_rate": 1.5782122934733478e-05, + "loss": 0.8779, + "step": 26360 + }, + { + "epoch": 1.95, + "learning_rate": 1.5781797496221906e-05, + "loss": 0.9976, + "step": 26361 + }, + { + "epoch": 1.95, + "learning_rate": 1.5781472048511574e-05, + "loss": 0.9358, + "step": 26362 + }, + { + "epoch": 1.95, + "learning_rate": 1.5781146591603002e-05, + "loss": 0.9365, + "step": 26363 + }, + { + "epoch": 1.95, + "learning_rate": 1.578082112549671e-05, + "loss": 0.9894, + "step": 26364 + }, + { + "epoch": 1.95, + "learning_rate": 1.5780495650193212e-05, + "loss": 1.0472, + "step": 26365 + }, + { + "epoch": 1.95, + "learning_rate": 1.578017016569303e-05, + "loss": 0.9055, + "step": 26366 + }, + { + "epoch": 1.95, + "learning_rate": 1.5779844671996683e-05, + "loss": 1.0119, + "step": 26367 + }, + { + "epoch": 1.95, + "learning_rate": 1.5779519169104682e-05, + "loss": 0.9606, + "step": 26368 + }, + { + "epoch": 1.95, + "learning_rate": 1.577919365701755e-05, + "loss": 0.9562, + "step": 26369 + }, + { + "epoch": 1.95, + "learning_rate": 1.57788681357358e-05, + "loss": 1.025, + "step": 26370 + }, + { + "epoch": 1.95, + "learning_rate": 1.577854260525996e-05, + "loss": 1.051, + "step": 26371 + }, + { + "epoch": 1.95, + "learning_rate": 1.5778217065590536e-05, + "loss": 1.0518, + "step": 26372 + }, + { + "epoch": 1.95, + "learning_rate": 1.577789151672805e-05, + "loss": 0.921, + "step": 26373 + }, + { + "epoch": 1.95, + "learning_rate": 1.5777565958673025e-05, + "loss": 0.9385, + "step": 26374 + }, + { + "epoch": 1.95, + "learning_rate": 1.5777240391425978e-05, + "loss": 0.9748, + "step": 26375 + }, + { + "epoch": 1.95, + "learning_rate": 1.577691481498742e-05, + "loss": 0.996, + "step": 26376 + }, + { + "epoch": 1.95, + "learning_rate": 1.577658922935787e-05, + "loss": 1.0081, + "step": 26377 + }, + { + "epoch": 1.95, + "learning_rate": 1.5776263634537855e-05, + "loss": 1.1773, + "step": 26378 + }, + { + "epoch": 1.95, + "learning_rate": 1.5775938030527887e-05, + "loss": 1.0902, + "step": 26379 + }, + { + "epoch": 1.95, + "learning_rate": 1.5775612417328483e-05, + "loss": 0.9893, + "step": 26380 + }, + { + "epoch": 1.95, + "learning_rate": 1.5775286794940163e-05, + "loss": 1.1346, + "step": 26381 + }, + { + "epoch": 1.95, + "learning_rate": 1.5774961163363444e-05, + "loss": 0.9191, + "step": 26382 + }, + { + "epoch": 1.95, + "learning_rate": 1.5774635522598843e-05, + "loss": 0.9703, + "step": 26383 + }, + { + "epoch": 1.95, + "learning_rate": 1.5774309872646882e-05, + "loss": 0.9301, + "step": 26384 + }, + { + "epoch": 1.95, + "learning_rate": 1.5773984213508073e-05, + "loss": 1.01, + "step": 26385 + }, + { + "epoch": 1.95, + "learning_rate": 1.577365854518294e-05, + "loss": 0.9996, + "step": 26386 + }, + { + "epoch": 1.95, + "learning_rate": 1.5773332867671997e-05, + "loss": 1.0275, + "step": 26387 + }, + { + "epoch": 1.95, + "learning_rate": 1.5773007180975764e-05, + "loss": 1.1375, + "step": 26388 + }, + { + "epoch": 1.95, + "learning_rate": 1.5772681485094763e-05, + "loss": 1.0174, + "step": 26389 + }, + { + "epoch": 1.95, + "learning_rate": 1.5772355780029506e-05, + "loss": 1.0393, + "step": 26390 + }, + { + "epoch": 1.95, + "learning_rate": 1.5772030065780514e-05, + "loss": 0.8936, + "step": 26391 + }, + { + "epoch": 1.95, + "learning_rate": 1.5771704342348302e-05, + "loss": 1.0667, + "step": 26392 + }, + { + "epoch": 1.95, + "learning_rate": 1.5771378609733393e-05, + "loss": 1.0301, + "step": 26393 + }, + { + "epoch": 1.95, + "learning_rate": 1.57710528679363e-05, + "loss": 1.0774, + "step": 26394 + }, + { + "epoch": 1.95, + "learning_rate": 1.5770727116957552e-05, + "loss": 0.9554, + "step": 26395 + }, + { + "epoch": 1.95, + "learning_rate": 1.5770401356797657e-05, + "loss": 1.0349, + "step": 26396 + }, + { + "epoch": 1.95, + "learning_rate": 1.577007558745713e-05, + "loss": 0.9893, + "step": 26397 + }, + { + "epoch": 1.95, + "learning_rate": 1.5769749808936502e-05, + "loss": 0.9789, + "step": 26398 + }, + { + "epoch": 1.95, + "learning_rate": 1.5769424021236286e-05, + "loss": 0.9494, + "step": 26399 + }, + { + "epoch": 1.95, + "learning_rate": 1.5769098224356994e-05, + "loss": 1.0335, + "step": 26400 + }, + { + "epoch": 1.95, + "learning_rate": 1.576877241829915e-05, + "loss": 1.0935, + "step": 26401 + }, + { + "epoch": 1.95, + "learning_rate": 1.5768446603063273e-05, + "loss": 0.9399, + "step": 26402 + }, + { + "epoch": 1.95, + "learning_rate": 1.5768120778649877e-05, + "loss": 0.9632, + "step": 26403 + }, + { + "epoch": 1.95, + "learning_rate": 1.5767794945059486e-05, + "loss": 1.1308, + "step": 26404 + }, + { + "epoch": 1.95, + "learning_rate": 1.5767469102292618e-05, + "loss": 0.9535, + "step": 26405 + }, + { + "epoch": 1.95, + "learning_rate": 1.5767143250349784e-05, + "loss": 1.0369, + "step": 26406 + }, + { + "epoch": 1.95, + "learning_rate": 1.576681738923151e-05, + "loss": 0.9869, + "step": 26407 + }, + { + "epoch": 1.95, + "learning_rate": 1.5766491518938317e-05, + "loss": 1.0196, + "step": 26408 + }, + { + "epoch": 1.95, + "learning_rate": 1.576616563947071e-05, + "loss": 0.9316, + "step": 26409 + }, + { + "epoch": 1.95, + "learning_rate": 1.576583975082922e-05, + "loss": 0.9722, + "step": 26410 + }, + { + "epoch": 1.95, + "learning_rate": 1.576551385301436e-05, + "loss": 1.2065, + "step": 26411 + }, + { + "epoch": 1.95, + "learning_rate": 1.576518794602665e-05, + "loss": 1.0399, + "step": 26412 + }, + { + "epoch": 1.95, + "learning_rate": 1.5764862029866614e-05, + "loss": 0.9169, + "step": 26413 + }, + { + "epoch": 1.95, + "learning_rate": 1.5764536104534758e-05, + "loss": 1.0405, + "step": 26414 + }, + { + "epoch": 1.95, + "learning_rate": 1.576421017003161e-05, + "loss": 1.1076, + "step": 26415 + }, + { + "epoch": 1.95, + "learning_rate": 1.5763884226357686e-05, + "loss": 0.9966, + "step": 26416 + }, + { + "epoch": 1.95, + "learning_rate": 1.5763558273513506e-05, + "loss": 1.0569, + "step": 26417 + }, + { + "epoch": 1.95, + "learning_rate": 1.5763232311499584e-05, + "loss": 0.9711, + "step": 26418 + }, + { + "epoch": 1.95, + "learning_rate": 1.5762906340316446e-05, + "loss": 0.9969, + "step": 26419 + }, + { + "epoch": 1.95, + "learning_rate": 1.5762580359964604e-05, + "loss": 0.9583, + "step": 26420 + }, + { + "epoch": 1.95, + "learning_rate": 1.576225437044458e-05, + "loss": 1.0473, + "step": 26421 + }, + { + "epoch": 1.95, + "learning_rate": 1.576192837175689e-05, + "loss": 1.016, + "step": 26422 + }, + { + "epoch": 1.95, + "learning_rate": 1.5761602363902055e-05, + "loss": 1.1774, + "step": 26423 + }, + { + "epoch": 1.95, + "learning_rate": 1.5761276346880597e-05, + "loss": 1.0706, + "step": 26424 + }, + { + "epoch": 1.95, + "learning_rate": 1.5760950320693023e-05, + "loss": 0.9628, + "step": 26425 + }, + { + "epoch": 1.95, + "learning_rate": 1.5760624285339865e-05, + "loss": 0.9694, + "step": 26426 + }, + { + "epoch": 1.95, + "learning_rate": 1.5760298240821638e-05, + "loss": 0.9923, + "step": 26427 + }, + { + "epoch": 1.95, + "learning_rate": 1.5759972187138853e-05, + "loss": 0.9597, + "step": 26428 + }, + { + "epoch": 1.95, + "learning_rate": 1.575964612429204e-05, + "loss": 0.9748, + "step": 26429 + }, + { + "epoch": 1.95, + "learning_rate": 1.575932005228171e-05, + "loss": 0.8977, + "step": 26430 + }, + { + "epoch": 1.95, + "learning_rate": 1.5758993971108383e-05, + "loss": 1.015, + "step": 26431 + }, + { + "epoch": 1.95, + "learning_rate": 1.5758667880772582e-05, + "loss": 1.0292, + "step": 26432 + }, + { + "epoch": 1.95, + "learning_rate": 1.575834178127482e-05, + "loss": 1.0265, + "step": 26433 + }, + { + "epoch": 1.95, + "learning_rate": 1.575801567261562e-05, + "loss": 0.974, + "step": 26434 + }, + { + "epoch": 1.95, + "learning_rate": 1.5757689554795498e-05, + "loss": 1.0928, + "step": 26435 + }, + { + "epoch": 1.95, + "learning_rate": 1.5757363427814978e-05, + "loss": 0.9549, + "step": 26436 + }, + { + "epoch": 1.95, + "learning_rate": 1.5757037291674572e-05, + "loss": 1.1052, + "step": 26437 + }, + { + "epoch": 1.95, + "learning_rate": 1.5756711146374803e-05, + "loss": 1.0773, + "step": 26438 + }, + { + "epoch": 1.95, + "learning_rate": 1.5756384991916188e-05, + "loss": 1.0247, + "step": 26439 + }, + { + "epoch": 1.95, + "learning_rate": 1.5756058828299247e-05, + "loss": 1.13, + "step": 26440 + }, + { + "epoch": 1.95, + "learning_rate": 1.57557326555245e-05, + "loss": 1.0289, + "step": 26441 + }, + { + "epoch": 1.95, + "learning_rate": 1.5755406473592467e-05, + "loss": 1.0643, + "step": 26442 + }, + { + "epoch": 1.95, + "learning_rate": 1.5755080282503658e-05, + "loss": 0.9346, + "step": 26443 + }, + { + "epoch": 1.95, + "learning_rate": 1.5754754082258603e-05, + "loss": 1.0637, + "step": 26444 + }, + { + "epoch": 1.95, + "learning_rate": 1.5754427872857813e-05, + "loss": 0.9218, + "step": 26445 + }, + { + "epoch": 1.95, + "learning_rate": 1.5754101654301816e-05, + "loss": 1.0518, + "step": 26446 + }, + { + "epoch": 1.95, + "learning_rate": 1.5753775426591125e-05, + "loss": 1.1321, + "step": 26447 + }, + { + "epoch": 1.95, + "learning_rate": 1.5753449189726254e-05, + "loss": 0.995, + "step": 26448 + }, + { + "epoch": 1.95, + "learning_rate": 1.575312294370773e-05, + "loss": 1.0237, + "step": 26449 + }, + { + "epoch": 1.95, + "learning_rate": 1.5752796688536073e-05, + "loss": 1.1045, + "step": 26450 + }, + { + "epoch": 1.95, + "learning_rate": 1.5752470424211796e-05, + "loss": 1.0489, + "step": 26451 + }, + { + "epoch": 1.95, + "learning_rate": 1.5752144150735423e-05, + "loss": 0.917, + "step": 26452 + }, + { + "epoch": 1.95, + "learning_rate": 1.575181786810747e-05, + "loss": 1.0866, + "step": 26453 + }, + { + "epoch": 1.95, + "learning_rate": 1.5751491576328456e-05, + "loss": 1.0281, + "step": 26454 + }, + { + "epoch": 1.95, + "learning_rate": 1.5751165275398902e-05, + "loss": 0.97, + "step": 26455 + }, + { + "epoch": 1.95, + "learning_rate": 1.5750838965319325e-05, + "loss": 0.9989, + "step": 26456 + }, + { + "epoch": 1.95, + "learning_rate": 1.5750512646090248e-05, + "loss": 0.965, + "step": 26457 + }, + { + "epoch": 1.95, + "learning_rate": 1.5750186317712188e-05, + "loss": 1.0191, + "step": 26458 + }, + { + "epoch": 1.96, + "learning_rate": 1.574985998018566e-05, + "loss": 1.0302, + "step": 26459 + }, + { + "epoch": 1.96, + "learning_rate": 1.574953363351119e-05, + "loss": 0.9833, + "step": 26460 + }, + { + "epoch": 1.96, + "learning_rate": 1.5749207277689293e-05, + "loss": 1.0003, + "step": 26461 + }, + { + "epoch": 1.96, + "learning_rate": 1.5748880912720487e-05, + "loss": 0.9573, + "step": 26462 + }, + { + "epoch": 1.96, + "learning_rate": 1.5748554538605298e-05, + "loss": 1.0122, + "step": 26463 + }, + { + "epoch": 1.96, + "learning_rate": 1.574822815534424e-05, + "loss": 0.932, + "step": 26464 + }, + { + "epoch": 1.96, + "learning_rate": 1.5747901762937834e-05, + "loss": 1.0444, + "step": 26465 + }, + { + "epoch": 1.96, + "learning_rate": 1.5747575361386597e-05, + "loss": 1.0386, + "step": 26466 + }, + { + "epoch": 1.96, + "learning_rate": 1.574724895069105e-05, + "loss": 0.9942, + "step": 26467 + }, + { + "epoch": 1.96, + "learning_rate": 1.5746922530851713e-05, + "loss": 1.0176, + "step": 26468 + }, + { + "epoch": 1.96, + "learning_rate": 1.5746596101869104e-05, + "loss": 0.9363, + "step": 26469 + }, + { + "epoch": 1.96, + "learning_rate": 1.5746269663743742e-05, + "loss": 0.9538, + "step": 26470 + }, + { + "epoch": 1.96, + "learning_rate": 1.5745943216476148e-05, + "loss": 1.0237, + "step": 26471 + }, + { + "epoch": 1.96, + "learning_rate": 1.574561676006684e-05, + "loss": 0.9246, + "step": 26472 + }, + { + "epoch": 1.96, + "learning_rate": 1.5745290294516336e-05, + "loss": 1.0155, + "step": 26473 + }, + { + "epoch": 1.96, + "learning_rate": 1.5744963819825163e-05, + "loss": 1.0879, + "step": 26474 + }, + { + "epoch": 1.96, + "learning_rate": 1.574463733599383e-05, + "loss": 0.9955, + "step": 26475 + }, + { + "epoch": 1.96, + "learning_rate": 1.5744310843022865e-05, + "loss": 1.0827, + "step": 26476 + }, + { + "epoch": 1.96, + "learning_rate": 1.574398434091278e-05, + "loss": 1.0034, + "step": 26477 + }, + { + "epoch": 1.96, + "learning_rate": 1.57436578296641e-05, + "loss": 1.0445, + "step": 26478 + }, + { + "epoch": 1.96, + "learning_rate": 1.574333130927734e-05, + "loss": 0.8948, + "step": 26479 + }, + { + "epoch": 1.96, + "learning_rate": 1.5743004779753025e-05, + "loss": 1.0241, + "step": 26480 + }, + { + "epoch": 1.96, + "learning_rate": 1.574267824109167e-05, + "loss": 0.9562, + "step": 26481 + }, + { + "epoch": 1.96, + "learning_rate": 1.5742351693293797e-05, + "loss": 1.0639, + "step": 26482 + }, + { + "epoch": 1.96, + "learning_rate": 1.5742025136359925e-05, + "loss": 1.0664, + "step": 26483 + }, + { + "epoch": 1.96, + "learning_rate": 1.574169857029057e-05, + "loss": 0.9485, + "step": 26484 + }, + { + "epoch": 1.96, + "learning_rate": 1.574137199508626e-05, + "loss": 0.8953, + "step": 26485 + }, + { + "epoch": 1.96, + "learning_rate": 1.5741045410747506e-05, + "loss": 1.0599, + "step": 26486 + }, + { + "epoch": 1.96, + "learning_rate": 1.5740718817274832e-05, + "loss": 0.9294, + "step": 26487 + }, + { + "epoch": 1.96, + "learning_rate": 1.5740392214668755e-05, + "loss": 1.0132, + "step": 26488 + }, + { + "epoch": 1.96, + "learning_rate": 1.5740065602929795e-05, + "loss": 1.0792, + "step": 26489 + }, + { + "epoch": 1.96, + "learning_rate": 1.5739738982058474e-05, + "loss": 1.0401, + "step": 26490 + }, + { + "epoch": 1.96, + "learning_rate": 1.5739412352055312e-05, + "loss": 1.0568, + "step": 26491 + }, + { + "epoch": 1.96, + "learning_rate": 1.5739085712920827e-05, + "loss": 0.9902, + "step": 26492 + }, + { + "epoch": 1.96, + "learning_rate": 1.573875906465554e-05, + "loss": 0.9898, + "step": 26493 + }, + { + "epoch": 1.96, + "learning_rate": 1.5738432407259966e-05, + "loss": 1.0444, + "step": 26494 + }, + { + "epoch": 1.96, + "learning_rate": 1.5738105740734627e-05, + "loss": 0.8962, + "step": 26495 + }, + { + "epoch": 1.96, + "learning_rate": 1.5737779065080048e-05, + "loss": 1.0771, + "step": 26496 + }, + { + "epoch": 1.96, + "learning_rate": 1.573745238029674e-05, + "loss": 0.9877, + "step": 26497 + }, + { + "epoch": 1.96, + "learning_rate": 1.573712568638523e-05, + "loss": 0.9961, + "step": 26498 + }, + { + "epoch": 1.96, + "learning_rate": 1.5736798983346037e-05, + "loss": 1.0725, + "step": 26499 + }, + { + "epoch": 1.96, + "learning_rate": 1.573647227117968e-05, + "loss": 1.0154, + "step": 26500 + }, + { + "epoch": 1.96, + "learning_rate": 1.573614554988667e-05, + "loss": 0.9978, + "step": 26501 + }, + { + "epoch": 1.96, + "learning_rate": 1.5735818819467543e-05, + "loss": 1.0642, + "step": 26502 + }, + { + "epoch": 1.96, + "learning_rate": 1.5735492079922807e-05, + "loss": 1.1046, + "step": 26503 + }, + { + "epoch": 1.96, + "learning_rate": 1.5735165331252985e-05, + "loss": 1.1026, + "step": 26504 + }, + { + "epoch": 1.96, + "learning_rate": 1.5734838573458596e-05, + "loss": 0.8815, + "step": 26505 + }, + { + "epoch": 1.96, + "learning_rate": 1.573451180654016e-05, + "loss": 0.9749, + "step": 26506 + }, + { + "epoch": 1.96, + "learning_rate": 1.5734185030498202e-05, + "loss": 1.0479, + "step": 26507 + }, + { + "epoch": 1.96, + "learning_rate": 1.5733858245333234e-05, + "loss": 0.97, + "step": 26508 + }, + { + "epoch": 1.96, + "learning_rate": 1.5733531451045782e-05, + "loss": 1.1179, + "step": 26509 + }, + { + "epoch": 1.96, + "learning_rate": 1.5733204647636365e-05, + "loss": 1.0239, + "step": 26510 + }, + { + "epoch": 1.96, + "learning_rate": 1.5732877835105497e-05, + "loss": 0.9389, + "step": 26511 + }, + { + "epoch": 1.96, + "learning_rate": 1.5732551013453704e-05, + "loss": 1.1396, + "step": 26512 + }, + { + "epoch": 1.96, + "learning_rate": 1.5732224182681507e-05, + "loss": 1.0861, + "step": 26513 + }, + { + "epoch": 1.96, + "learning_rate": 1.573189734278942e-05, + "loss": 1.022, + "step": 26514 + }, + { + "epoch": 1.96, + "learning_rate": 1.5731570493777967e-05, + "loss": 1.0498, + "step": 26515 + }, + { + "epoch": 1.96, + "learning_rate": 1.573124363564767e-05, + "loss": 0.9797, + "step": 26516 + }, + { + "epoch": 1.96, + "learning_rate": 1.573091676839904e-05, + "loss": 1.1095, + "step": 26517 + }, + { + "epoch": 1.96, + "learning_rate": 1.5730589892032607e-05, + "loss": 1.0776, + "step": 26518 + }, + { + "epoch": 1.96, + "learning_rate": 1.5730263006548886e-05, + "loss": 1.0291, + "step": 26519 + }, + { + "epoch": 1.96, + "learning_rate": 1.5729936111948406e-05, + "loss": 0.9072, + "step": 26520 + }, + { + "epoch": 1.96, + "learning_rate": 1.5729609208231673e-05, + "loss": 1.0521, + "step": 26521 + }, + { + "epoch": 1.96, + "learning_rate": 1.5729282295399213e-05, + "loss": 1.092, + "step": 26522 + }, + { + "epoch": 1.96, + "learning_rate": 1.572895537345155e-05, + "loss": 1.045, + "step": 26523 + }, + { + "epoch": 1.96, + "learning_rate": 1.5728628442389198e-05, + "loss": 1.1685, + "step": 26524 + }, + { + "epoch": 1.96, + "learning_rate": 1.572830150221268e-05, + "loss": 0.9224, + "step": 26525 + }, + { + "epoch": 1.96, + "learning_rate": 1.572797455292252e-05, + "loss": 0.99, + "step": 26526 + }, + { + "epoch": 1.96, + "learning_rate": 1.5727647594519233e-05, + "loss": 0.9409, + "step": 26527 + }, + { + "epoch": 1.96, + "learning_rate": 1.5727320627003338e-05, + "loss": 0.9539, + "step": 26528 + }, + { + "epoch": 1.96, + "learning_rate": 1.572699365037536e-05, + "loss": 1.0558, + "step": 26529 + }, + { + "epoch": 1.96, + "learning_rate": 1.5726666664635817e-05, + "loss": 1.041, + "step": 26530 + }, + { + "epoch": 1.96, + "learning_rate": 1.572633966978523e-05, + "loss": 1.0112, + "step": 26531 + }, + { + "epoch": 1.96, + "learning_rate": 1.5726012665824117e-05, + "loss": 0.8766, + "step": 26532 + }, + { + "epoch": 1.96, + "learning_rate": 1.5725685652753e-05, + "loss": 1.0528, + "step": 26533 + }, + { + "epoch": 1.96, + "learning_rate": 1.57253586305724e-05, + "loss": 0.9925, + "step": 26534 + }, + { + "epoch": 1.96, + "learning_rate": 1.5725031599282835e-05, + "loss": 1.0308, + "step": 26535 + }, + { + "epoch": 1.96, + "learning_rate": 1.5724704558884827e-05, + "loss": 0.902, + "step": 26536 + }, + { + "epoch": 1.96, + "learning_rate": 1.5724377509378898e-05, + "loss": 0.9943, + "step": 26537 + }, + { + "epoch": 1.96, + "learning_rate": 1.5724050450765565e-05, + "loss": 1.1135, + "step": 26538 + }, + { + "epoch": 1.96, + "learning_rate": 1.5723723383045347e-05, + "loss": 0.9166, + "step": 26539 + }, + { + "epoch": 1.96, + "learning_rate": 1.5723396306218773e-05, + "loss": 0.9295, + "step": 26540 + }, + { + "epoch": 1.96, + "learning_rate": 1.572306922028635e-05, + "loss": 0.9991, + "step": 26541 + }, + { + "epoch": 1.96, + "learning_rate": 1.572274212524861e-05, + "loss": 0.9874, + "step": 26542 + }, + { + "epoch": 1.96, + "learning_rate": 1.572241502110607e-05, + "loss": 1.0137, + "step": 26543 + }, + { + "epoch": 1.96, + "learning_rate": 1.5722087907859248e-05, + "loss": 0.9525, + "step": 26544 + }, + { + "epoch": 1.96, + "learning_rate": 1.5721760785508665e-05, + "loss": 1.0283, + "step": 26545 + }, + { + "epoch": 1.96, + "learning_rate": 1.5721433654054846e-05, + "loss": 0.8964, + "step": 26546 + }, + { + "epoch": 1.96, + "learning_rate": 1.5721106513498306e-05, + "loss": 0.9775, + "step": 26547 + }, + { + "epoch": 1.96, + "learning_rate": 1.5720779363839566e-05, + "loss": 0.966, + "step": 26548 + }, + { + "epoch": 1.96, + "learning_rate": 1.572045220507915e-05, + "loss": 0.9832, + "step": 26549 + }, + { + "epoch": 1.96, + "learning_rate": 1.5720125037217575e-05, + "loss": 1.0593, + "step": 26550 + }, + { + "epoch": 1.96, + "learning_rate": 1.5719797860255362e-05, + "loss": 1.1111, + "step": 26551 + }, + { + "epoch": 1.96, + "learning_rate": 1.5719470674193034e-05, + "loss": 0.9725, + "step": 26552 + }, + { + "epoch": 1.96, + "learning_rate": 1.571914347903111e-05, + "loss": 0.9818, + "step": 26553 + }, + { + "epoch": 1.96, + "learning_rate": 1.571881627477011e-05, + "loss": 0.9887, + "step": 26554 + }, + { + "epoch": 1.96, + "learning_rate": 1.5718489061410555e-05, + "loss": 1.0076, + "step": 26555 + }, + { + "epoch": 1.96, + "learning_rate": 1.5718161838952966e-05, + "loss": 1.0199, + "step": 26556 + }, + { + "epoch": 1.96, + "learning_rate": 1.571783460739786e-05, + "loss": 1.1002, + "step": 26557 + }, + { + "epoch": 1.96, + "learning_rate": 1.5717507366745766e-05, + "loss": 1.0009, + "step": 26558 + }, + { + "epoch": 1.96, + "learning_rate": 1.57171801169972e-05, + "loss": 0.936, + "step": 26559 + }, + { + "epoch": 1.96, + "learning_rate": 1.571685285815268e-05, + "loss": 0.9302, + "step": 26560 + }, + { + "epoch": 1.96, + "learning_rate": 1.5716525590212726e-05, + "loss": 1.0638, + "step": 26561 + }, + { + "epoch": 1.96, + "learning_rate": 1.5716198313177867e-05, + "loss": 1.0559, + "step": 26562 + }, + { + "epoch": 1.96, + "learning_rate": 1.5715871027048615e-05, + "loss": 0.9826, + "step": 26563 + }, + { + "epoch": 1.96, + "learning_rate": 1.5715543731825492e-05, + "loss": 1.0279, + "step": 26564 + }, + { + "epoch": 1.96, + "learning_rate": 1.5715216427509026e-05, + "loss": 1.0602, + "step": 26565 + }, + { + "epoch": 1.96, + "learning_rate": 1.571488911409973e-05, + "loss": 1.0768, + "step": 26566 + }, + { + "epoch": 1.96, + "learning_rate": 1.5714561791598128e-05, + "loss": 1.0773, + "step": 26567 + }, + { + "epoch": 1.96, + "learning_rate": 1.5714234460004738e-05, + "loss": 1.0628, + "step": 26568 + }, + { + "epoch": 1.96, + "learning_rate": 1.5713907119320086e-05, + "loss": 1.0069, + "step": 26569 + }, + { + "epoch": 1.96, + "learning_rate": 1.5713579769544683e-05, + "loss": 1.0247, + "step": 26570 + }, + { + "epoch": 1.96, + "learning_rate": 1.5713252410679062e-05, + "loss": 1.0636, + "step": 26571 + }, + { + "epoch": 1.96, + "learning_rate": 1.5712925042723737e-05, + "loss": 1.0152, + "step": 26572 + }, + { + "epoch": 1.96, + "learning_rate": 1.571259766567923e-05, + "loss": 0.9734, + "step": 26573 + }, + { + "epoch": 1.96, + "learning_rate": 1.5712270279546063e-05, + "loss": 1.0972, + "step": 26574 + }, + { + "epoch": 1.96, + "learning_rate": 1.5711942884324755e-05, + "loss": 1.0785, + "step": 26575 + }, + { + "epoch": 1.96, + "learning_rate": 1.5711615480015826e-05, + "loss": 1.0265, + "step": 26576 + }, + { + "epoch": 1.96, + "learning_rate": 1.57112880666198e-05, + "loss": 0.9918, + "step": 26577 + }, + { + "epoch": 1.96, + "learning_rate": 1.5710960644137196e-05, + "loss": 1.1102, + "step": 26578 + }, + { + "epoch": 1.96, + "learning_rate": 1.5710633212568534e-05, + "loss": 1.0393, + "step": 26579 + }, + { + "epoch": 1.96, + "learning_rate": 1.5710305771914337e-05, + "loss": 1.0382, + "step": 26580 + }, + { + "epoch": 1.96, + "learning_rate": 1.5709978322175124e-05, + "loss": 0.9928, + "step": 26581 + }, + { + "epoch": 1.96, + "learning_rate": 1.570965086335142e-05, + "loss": 0.9077, + "step": 26582 + }, + { + "epoch": 1.96, + "learning_rate": 1.570932339544374e-05, + "loss": 0.8935, + "step": 26583 + }, + { + "epoch": 1.96, + "learning_rate": 1.5708995918452612e-05, + "loss": 0.929, + "step": 26584 + }, + { + "epoch": 1.96, + "learning_rate": 1.5708668432378552e-05, + "loss": 1.0387, + "step": 26585 + }, + { + "epoch": 1.96, + "learning_rate": 1.5708340937222084e-05, + "loss": 0.9725, + "step": 26586 + }, + { + "epoch": 1.96, + "learning_rate": 1.5708013432983725e-05, + "loss": 1.0488, + "step": 26587 + }, + { + "epoch": 1.96, + "learning_rate": 1.5707685919663997e-05, + "loss": 0.9808, + "step": 26588 + }, + { + "epoch": 1.96, + "learning_rate": 1.5707358397263423e-05, + "loss": 1.0059, + "step": 26589 + }, + { + "epoch": 1.96, + "learning_rate": 1.5707030865782522e-05, + "loss": 1.041, + "step": 26590 + }, + { + "epoch": 1.96, + "learning_rate": 1.570670332522182e-05, + "loss": 1.1005, + "step": 26591 + }, + { + "epoch": 1.96, + "learning_rate": 1.5706375775581833e-05, + "loss": 1.0768, + "step": 26592 + }, + { + "epoch": 1.96, + "learning_rate": 1.5706048216863084e-05, + "loss": 0.9118, + "step": 26593 + }, + { + "epoch": 1.97, + "learning_rate": 1.5705720649066093e-05, + "loss": 1.0323, + "step": 26594 + }, + { + "epoch": 1.97, + "learning_rate": 1.5705393072191383e-05, + "loss": 1.102, + "step": 26595 + }, + { + "epoch": 1.97, + "learning_rate": 1.5705065486239474e-05, + "loss": 1.1553, + "step": 26596 + }, + { + "epoch": 1.97, + "learning_rate": 1.570473789121089e-05, + "loss": 1.0242, + "step": 26597 + }, + { + "epoch": 1.97, + "learning_rate": 1.5704410287106146e-05, + "loss": 1.0349, + "step": 26598 + }, + { + "epoch": 1.97, + "learning_rate": 1.5704082673925767e-05, + "loss": 0.9551, + "step": 26599 + }, + { + "epoch": 1.97, + "learning_rate": 1.5703755051670274e-05, + "loss": 1.0735, + "step": 26600 + }, + { + "epoch": 1.97, + "learning_rate": 1.5703427420340188e-05, + "loss": 0.9931, + "step": 26601 + }, + { + "epoch": 1.97, + "learning_rate": 1.5703099779936035e-05, + "loss": 1.1307, + "step": 26602 + }, + { + "epoch": 1.97, + "learning_rate": 1.570277213045833e-05, + "loss": 1.1031, + "step": 26603 + }, + { + "epoch": 1.97, + "learning_rate": 1.5702444471907596e-05, + "loss": 0.9567, + "step": 26604 + }, + { + "epoch": 1.97, + "learning_rate": 1.570211680428435e-05, + "loss": 1.0231, + "step": 26605 + }, + { + "epoch": 1.97, + "learning_rate": 1.5701789127589124e-05, + "loss": 0.9186, + "step": 26606 + }, + { + "epoch": 1.97, + "learning_rate": 1.570146144182243e-05, + "loss": 0.9607, + "step": 26607 + }, + { + "epoch": 1.97, + "learning_rate": 1.570113374698479e-05, + "loss": 0.9669, + "step": 26608 + }, + { + "epoch": 1.97, + "learning_rate": 1.570080604307673e-05, + "loss": 1.0052, + "step": 26609 + }, + { + "epoch": 1.97, + "learning_rate": 1.570047833009877e-05, + "loss": 1.0513, + "step": 26610 + }, + { + "epoch": 1.97, + "learning_rate": 1.5700150608051433e-05, + "loss": 1.0704, + "step": 26611 + }, + { + "epoch": 1.97, + "learning_rate": 1.5699822876935235e-05, + "loss": 1.0163, + "step": 26612 + }, + { + "epoch": 1.97, + "learning_rate": 1.56994951367507e-05, + "loss": 0.9404, + "step": 26613 + }, + { + "epoch": 1.97, + "learning_rate": 1.569916738749835e-05, + "loss": 0.9735, + "step": 26614 + }, + { + "epoch": 1.97, + "learning_rate": 1.5698839629178705e-05, + "loss": 1.0046, + "step": 26615 + }, + { + "epoch": 1.97, + "learning_rate": 1.569851186179229e-05, + "loss": 0.97, + "step": 26616 + }, + { + "epoch": 1.97, + "learning_rate": 1.5698184085339624e-05, + "loss": 1.0327, + "step": 26617 + }, + { + "epoch": 1.97, + "learning_rate": 1.5697856299821227e-05, + "loss": 1.0177, + "step": 26618 + }, + { + "epoch": 1.97, + "learning_rate": 1.5697528505237626e-05, + "loss": 1.135, + "step": 26619 + }, + { + "epoch": 1.97, + "learning_rate": 1.5697200701589333e-05, + "loss": 1.0789, + "step": 26620 + }, + { + "epoch": 1.97, + "learning_rate": 1.569687288887688e-05, + "loss": 1.0647, + "step": 26621 + }, + { + "epoch": 1.97, + "learning_rate": 1.569654506710078e-05, + "loss": 1.0877, + "step": 26622 + }, + { + "epoch": 1.97, + "learning_rate": 1.5696217236261562e-05, + "loss": 0.9773, + "step": 26623 + }, + { + "epoch": 1.97, + "learning_rate": 1.569588939635974e-05, + "loss": 0.9965, + "step": 26624 + }, + { + "epoch": 1.97, + "learning_rate": 1.569556154739584e-05, + "loss": 1.0307, + "step": 26625 + }, + { + "epoch": 1.97, + "learning_rate": 1.5695233689370387e-05, + "loss": 1.0858, + "step": 26626 + }, + { + "epoch": 1.97, + "learning_rate": 1.5694905822283896e-05, + "loss": 1.0336, + "step": 26627 + }, + { + "epoch": 1.97, + "learning_rate": 1.569457794613689e-05, + "loss": 1.0248, + "step": 26628 + }, + { + "epoch": 1.97, + "learning_rate": 1.5694250060929892e-05, + "loss": 1.0493, + "step": 26629 + }, + { + "epoch": 1.97, + "learning_rate": 1.569392216666343e-05, + "loss": 1.112, + "step": 26630 + }, + { + "epoch": 1.97, + "learning_rate": 1.569359426333801e-05, + "loss": 0.9003, + "step": 26631 + }, + { + "epoch": 1.97, + "learning_rate": 1.569326635095417e-05, + "loss": 1.0122, + "step": 26632 + }, + { + "epoch": 1.97, + "learning_rate": 1.569293842951242e-05, + "loss": 1.0402, + "step": 26633 + }, + { + "epoch": 1.97, + "learning_rate": 1.569261049901329e-05, + "loss": 0.9081, + "step": 26634 + }, + { + "epoch": 1.97, + "learning_rate": 1.5692282559457296e-05, + "loss": 1.0155, + "step": 26635 + }, + { + "epoch": 1.97, + "learning_rate": 1.569195461084496e-05, + "loss": 1.0328, + "step": 26636 + }, + { + "epoch": 1.97, + "learning_rate": 1.569162665317681e-05, + "loss": 1.0267, + "step": 26637 + }, + { + "epoch": 1.97, + "learning_rate": 1.569129868645336e-05, + "loss": 0.9717, + "step": 26638 + }, + { + "epoch": 1.97, + "learning_rate": 1.5690970710675138e-05, + "loss": 1.0684, + "step": 26639 + }, + { + "epoch": 1.97, + "learning_rate": 1.569064272584266e-05, + "loss": 1.0419, + "step": 26640 + }, + { + "epoch": 1.97, + "learning_rate": 1.5690314731956453e-05, + "loss": 1.0105, + "step": 26641 + }, + { + "epoch": 1.97, + "learning_rate": 1.5689986729017038e-05, + "loss": 0.9094, + "step": 26642 + }, + { + "epoch": 1.97, + "learning_rate": 1.5689658717024935e-05, + "loss": 1.0303, + "step": 26643 + }, + { + "epoch": 1.97, + "learning_rate": 1.5689330695980665e-05, + "loss": 0.9672, + "step": 26644 + }, + { + "epoch": 1.97, + "learning_rate": 1.568900266588475e-05, + "loss": 1.029, + "step": 26645 + }, + { + "epoch": 1.97, + "learning_rate": 1.5688674626737716e-05, + "loss": 1.066, + "step": 26646 + }, + { + "epoch": 1.97, + "learning_rate": 1.568834657854008e-05, + "loss": 1.0193, + "step": 26647 + }, + { + "epoch": 1.97, + "learning_rate": 1.568801852129237e-05, + "loss": 1.0028, + "step": 26648 + }, + { + "epoch": 1.97, + "learning_rate": 1.5687690454995096e-05, + "loss": 0.9513, + "step": 26649 + }, + { + "epoch": 1.97, + "learning_rate": 1.5687362379648794e-05, + "loss": 1.0026, + "step": 26650 + }, + { + "epoch": 1.97, + "learning_rate": 1.5687034295253978e-05, + "loss": 0.866, + "step": 26651 + }, + { + "epoch": 1.97, + "learning_rate": 1.5686706201811173e-05, + "loss": 0.9891, + "step": 26652 + }, + { + "epoch": 1.97, + "learning_rate": 1.5686378099320897e-05, + "loss": 0.9996, + "step": 26653 + }, + { + "epoch": 1.97, + "learning_rate": 1.568604998778368e-05, + "loss": 0.9703, + "step": 26654 + }, + { + "epoch": 1.97, + "learning_rate": 1.568572186720003e-05, + "loss": 1.0073, + "step": 26655 + }, + { + "epoch": 1.97, + "learning_rate": 1.5685393737570488e-05, + "loss": 0.9614, + "step": 26656 + }, + { + "epoch": 1.97, + "learning_rate": 1.5685065598895558e-05, + "loss": 0.9385, + "step": 26657 + }, + { + "epoch": 1.97, + "learning_rate": 1.5684737451175776e-05, + "loss": 1.1454, + "step": 26658 + }, + { + "epoch": 1.97, + "learning_rate": 1.5684409294411654e-05, + "loss": 0.8849, + "step": 26659 + }, + { + "epoch": 1.97, + "learning_rate": 1.5684081128603717e-05, + "loss": 1.1048, + "step": 26660 + }, + { + "epoch": 1.97, + "learning_rate": 1.5683752953752494e-05, + "loss": 0.9271, + "step": 26661 + }, + { + "epoch": 1.97, + "learning_rate": 1.56834247698585e-05, + "loss": 1.1306, + "step": 26662 + }, + { + "epoch": 1.97, + "learning_rate": 1.5683096576922253e-05, + "loss": 1.1245, + "step": 26663 + }, + { + "epoch": 1.97, + "learning_rate": 1.5682768374944285e-05, + "loss": 0.9727, + "step": 26664 + }, + { + "epoch": 1.97, + "learning_rate": 1.5682440163925114e-05, + "loss": 1.0658, + "step": 26665 + }, + { + "epoch": 1.97, + "learning_rate": 1.568211194386526e-05, + "loss": 0.8522, + "step": 26666 + }, + { + "epoch": 1.97, + "learning_rate": 1.568178371476525e-05, + "loss": 0.9589, + "step": 26667 + }, + { + "epoch": 1.97, + "learning_rate": 1.56814554766256e-05, + "loss": 0.9148, + "step": 26668 + }, + { + "epoch": 1.97, + "learning_rate": 1.5681127229446838e-05, + "loss": 1.071, + "step": 26669 + }, + { + "epoch": 1.97, + "learning_rate": 1.5680798973229484e-05, + "loss": 0.9658, + "step": 26670 + }, + { + "epoch": 1.97, + "learning_rate": 1.568047070797406e-05, + "loss": 0.9917, + "step": 26671 + }, + { + "epoch": 1.97, + "learning_rate": 1.5680142433681087e-05, + "loss": 1.0601, + "step": 26672 + }, + { + "epoch": 1.97, + "learning_rate": 1.5679814150351086e-05, + "loss": 0.9895, + "step": 26673 + }, + { + "epoch": 1.97, + "learning_rate": 1.5679485857984585e-05, + "loss": 0.9453, + "step": 26674 + }, + { + "epoch": 1.97, + "learning_rate": 1.5679157556582107e-05, + "loss": 1.0941, + "step": 26675 + }, + { + "epoch": 1.97, + "learning_rate": 1.5678829246144165e-05, + "loss": 0.988, + "step": 26676 + }, + { + "epoch": 1.97, + "learning_rate": 1.567850092667129e-05, + "loss": 0.9651, + "step": 26677 + }, + { + "epoch": 1.97, + "learning_rate": 1.5678172598164e-05, + "loss": 1.004, + "step": 26678 + }, + { + "epoch": 1.97, + "learning_rate": 1.567784426062282e-05, + "loss": 1.1161, + "step": 26679 + }, + { + "epoch": 1.97, + "learning_rate": 1.5677515914048267e-05, + "loss": 0.9723, + "step": 26680 + }, + { + "epoch": 1.97, + "learning_rate": 1.567718755844087e-05, + "loss": 1.0162, + "step": 26681 + }, + { + "epoch": 1.97, + "learning_rate": 1.5676859193801152e-05, + "loss": 1.0188, + "step": 26682 + }, + { + "epoch": 1.97, + "learning_rate": 1.5676530820129626e-05, + "loss": 1.0327, + "step": 26683 + }, + { + "epoch": 1.97, + "learning_rate": 1.5676202437426828e-05, + "loss": 1.0078, + "step": 26684 + }, + { + "epoch": 1.97, + "learning_rate": 1.5675874045693266e-05, + "loss": 1.1546, + "step": 26685 + }, + { + "epoch": 1.97, + "learning_rate": 1.5675545644929476e-05, + "loss": 0.9929, + "step": 26686 + }, + { + "epoch": 1.97, + "learning_rate": 1.567521723513597e-05, + "loss": 1.0749, + "step": 26687 + }, + { + "epoch": 1.97, + "learning_rate": 1.5674888816313273e-05, + "loss": 1.0119, + "step": 26688 + }, + { + "epoch": 1.97, + "learning_rate": 1.5674560388461915e-05, + "loss": 0.8981, + "step": 26689 + }, + { + "epoch": 1.97, + "learning_rate": 1.5674231951582408e-05, + "loss": 1.0827, + "step": 26690 + }, + { + "epoch": 1.97, + "learning_rate": 1.5673903505675282e-05, + "loss": 0.979, + "step": 26691 + }, + { + "epoch": 1.97, + "learning_rate": 1.567357505074105e-05, + "loss": 1.0771, + "step": 26692 + }, + { + "epoch": 1.97, + "learning_rate": 1.5673246586780248e-05, + "loss": 0.904, + "step": 26693 + }, + { + "epoch": 1.97, + "learning_rate": 1.567291811379339e-05, + "loss": 1.0695, + "step": 26694 + }, + { + "epoch": 1.97, + "learning_rate": 1.5672589631781004e-05, + "loss": 0.9797, + "step": 26695 + }, + { + "epoch": 1.97, + "learning_rate": 1.5672261140743603e-05, + "loss": 1.0639, + "step": 26696 + }, + { + "epoch": 1.97, + "learning_rate": 1.567193264068172e-05, + "loss": 0.9764, + "step": 26697 + }, + { + "epoch": 1.97, + "learning_rate": 1.567160413159587e-05, + "loss": 1.0246, + "step": 26698 + }, + { + "epoch": 1.97, + "learning_rate": 1.5671275613486582e-05, + "loss": 1.0682, + "step": 26699 + }, + { + "epoch": 1.97, + "learning_rate": 1.5670947086354377e-05, + "loss": 0.9138, + "step": 26700 + }, + { + "epoch": 1.97, + "learning_rate": 1.567061855019977e-05, + "loss": 1.0552, + "step": 26701 + }, + { + "epoch": 1.97, + "learning_rate": 1.5670290005023296e-05, + "loss": 0.9054, + "step": 26702 + }, + { + "epoch": 1.97, + "learning_rate": 1.5669961450825467e-05, + "loss": 0.9333, + "step": 26703 + }, + { + "epoch": 1.97, + "learning_rate": 1.5669632887606812e-05, + "loss": 1.0596, + "step": 26704 + }, + { + "epoch": 1.97, + "learning_rate": 1.5669304315367853e-05, + "loss": 1.0148, + "step": 26705 + }, + { + "epoch": 1.97, + "learning_rate": 1.5668975734109114e-05, + "loss": 1.0185, + "step": 26706 + }, + { + "epoch": 1.97, + "learning_rate": 1.5668647143831114e-05, + "loss": 1.0452, + "step": 26707 + }, + { + "epoch": 1.97, + "learning_rate": 1.5668318544534374e-05, + "loss": 1.0123, + "step": 26708 + }, + { + "epoch": 1.97, + "learning_rate": 1.5667989936219424e-05, + "loss": 1.0579, + "step": 26709 + }, + { + "epoch": 1.97, + "learning_rate": 1.5667661318886785e-05, + "loss": 0.9582, + "step": 26710 + }, + { + "epoch": 1.97, + "learning_rate": 1.5667332692536973e-05, + "loss": 0.9566, + "step": 26711 + }, + { + "epoch": 1.97, + "learning_rate": 1.566700405717052e-05, + "loss": 0.9591, + "step": 26712 + }, + { + "epoch": 1.97, + "learning_rate": 1.5666675412787943e-05, + "loss": 1.0693, + "step": 26713 + }, + { + "epoch": 1.97, + "learning_rate": 1.5666346759389766e-05, + "loss": 0.9514, + "step": 26714 + }, + { + "epoch": 1.97, + "learning_rate": 1.5666018096976512e-05, + "loss": 1.0077, + "step": 26715 + }, + { + "epoch": 1.97, + "learning_rate": 1.5665689425548704e-05, + "loss": 1.0127, + "step": 26716 + }, + { + "epoch": 1.97, + "learning_rate": 1.566536074510687e-05, + "loss": 1.0401, + "step": 26717 + }, + { + "epoch": 1.97, + "learning_rate": 1.5665032055651522e-05, + "loss": 0.9424, + "step": 26718 + }, + { + "epoch": 1.97, + "learning_rate": 1.5664703357183193e-05, + "loss": 1.0502, + "step": 26719 + }, + { + "epoch": 1.97, + "learning_rate": 1.56643746497024e-05, + "loss": 0.9328, + "step": 26720 + }, + { + "epoch": 1.97, + "learning_rate": 1.5664045933209667e-05, + "loss": 0.9387, + "step": 26721 + }, + { + "epoch": 1.97, + "learning_rate": 1.566371720770552e-05, + "loss": 1.0345, + "step": 26722 + }, + { + "epoch": 1.97, + "learning_rate": 1.566338847319048e-05, + "loss": 1.1526, + "step": 26723 + }, + { + "epoch": 1.97, + "learning_rate": 1.5663059729665067e-05, + "loss": 0.9069, + "step": 26724 + }, + { + "epoch": 1.97, + "learning_rate": 1.5662730977129812e-05, + "loss": 1.0623, + "step": 26725 + }, + { + "epoch": 1.97, + "learning_rate": 1.566240221558523e-05, + "loss": 0.9524, + "step": 26726 + }, + { + "epoch": 1.97, + "learning_rate": 1.5662073445031846e-05, + "loss": 0.9418, + "step": 26727 + }, + { + "epoch": 1.97, + "learning_rate": 1.5661744665470188e-05, + "loss": 0.9736, + "step": 26728 + }, + { + "epoch": 1.97, + "learning_rate": 1.5661415876900772e-05, + "loss": 1.0279, + "step": 26729 + }, + { + "epoch": 1.98, + "learning_rate": 1.5661087079324123e-05, + "loss": 0.9822, + "step": 26730 + }, + { + "epoch": 1.98, + "learning_rate": 1.566075827274077e-05, + "loss": 1.0083, + "step": 26731 + }, + { + "epoch": 1.98, + "learning_rate": 1.5660429457151226e-05, + "loss": 1.0342, + "step": 26732 + }, + { + "epoch": 1.98, + "learning_rate": 1.5660100632556025e-05, + "loss": 1.0176, + "step": 26733 + }, + { + "epoch": 1.98, + "learning_rate": 1.5659771798955682e-05, + "loss": 1.0168, + "step": 26734 + }, + { + "epoch": 1.98, + "learning_rate": 1.5659442956350724e-05, + "loss": 0.9893, + "step": 26735 + }, + { + "epoch": 1.98, + "learning_rate": 1.5659114104741674e-05, + "loss": 1.0018, + "step": 26736 + }, + { + "epoch": 1.98, + "learning_rate": 1.5658785244129053e-05, + "loss": 1.0108, + "step": 26737 + }, + { + "epoch": 1.98, + "learning_rate": 1.5658456374513387e-05, + "loss": 1.0915, + "step": 26738 + }, + { + "epoch": 1.98, + "learning_rate": 1.5658127495895197e-05, + "loss": 0.8885, + "step": 26739 + }, + { + "epoch": 1.98, + "learning_rate": 1.5657798608275006e-05, + "loss": 0.8687, + "step": 26740 + }, + { + "epoch": 1.98, + "learning_rate": 1.565746971165334e-05, + "loss": 1.0278, + "step": 26741 + }, + { + "epoch": 1.98, + "learning_rate": 1.565714080603072e-05, + "loss": 0.9304, + "step": 26742 + }, + { + "epoch": 1.98, + "learning_rate": 1.5656811891407667e-05, + "loss": 0.9357, + "step": 26743 + }, + { + "epoch": 1.98, + "learning_rate": 1.5656482967784713e-05, + "loss": 1.1244, + "step": 26744 + }, + { + "epoch": 1.98, + "learning_rate": 1.5656154035162372e-05, + "loss": 0.9354, + "step": 26745 + }, + { + "epoch": 1.98, + "learning_rate": 1.5655825093541174e-05, + "loss": 0.9897, + "step": 26746 + }, + { + "epoch": 1.98, + "learning_rate": 1.5655496142921636e-05, + "loss": 1.1162, + "step": 26747 + }, + { + "epoch": 1.98, + "learning_rate": 1.5655167183304284e-05, + "loss": 0.9269, + "step": 26748 + }, + { + "epoch": 1.98, + "learning_rate": 1.5654838214689644e-05, + "loss": 1.0848, + "step": 26749 + }, + { + "epoch": 1.98, + "learning_rate": 1.5654509237078236e-05, + "loss": 1.1235, + "step": 26750 + }, + { + "epoch": 1.98, + "learning_rate": 1.5654180250470585e-05, + "loss": 1.0144, + "step": 26751 + }, + { + "epoch": 1.98, + "learning_rate": 1.5653851254867215e-05, + "loss": 1.0698, + "step": 26752 + }, + { + "epoch": 1.98, + "learning_rate": 1.5653522250268646e-05, + "loss": 1.0912, + "step": 26753 + }, + { + "epoch": 1.98, + "learning_rate": 1.565319323667541e-05, + "loss": 1.0141, + "step": 26754 + }, + { + "epoch": 1.98, + "learning_rate": 1.565286421408802e-05, + "loss": 1.0624, + "step": 26755 + }, + { + "epoch": 1.98, + "learning_rate": 1.5652535182507006e-05, + "loss": 1.0931, + "step": 26756 + }, + { + "epoch": 1.98, + "learning_rate": 1.5652206141932885e-05, + "loss": 1.0137, + "step": 26757 + }, + { + "epoch": 1.98, + "learning_rate": 1.5651877092366187e-05, + "loss": 0.9149, + "step": 26758 + }, + { + "epoch": 1.98, + "learning_rate": 1.5651548033807435e-05, + "loss": 1.054, + "step": 26759 + }, + { + "epoch": 1.98, + "learning_rate": 1.565121896625715e-05, + "loss": 0.9184, + "step": 26760 + }, + { + "epoch": 1.98, + "learning_rate": 1.5650889889715855e-05, + "loss": 1.0376, + "step": 26761 + }, + { + "epoch": 1.98, + "learning_rate": 1.5650560804184077e-05, + "loss": 1.0068, + "step": 26762 + }, + { + "epoch": 1.98, + "learning_rate": 1.565023170966234e-05, + "loss": 0.9294, + "step": 26763 + }, + { + "epoch": 1.98, + "learning_rate": 1.564990260615116e-05, + "loss": 0.9651, + "step": 26764 + }, + { + "epoch": 1.98, + "learning_rate": 1.5649573493651065e-05, + "loss": 0.93, + "step": 26765 + }, + { + "epoch": 1.98, + "learning_rate": 1.5649244372162583e-05, + "loss": 1.0107, + "step": 26766 + }, + { + "epoch": 1.98, + "learning_rate": 1.5648915241686235e-05, + "loss": 0.9411, + "step": 26767 + }, + { + "epoch": 1.98, + "learning_rate": 1.564858610222254e-05, + "loss": 0.895, + "step": 26768 + }, + { + "epoch": 1.98, + "learning_rate": 1.564825695377203e-05, + "loss": 1.0338, + "step": 26769 + }, + { + "epoch": 1.98, + "learning_rate": 1.564792779633522e-05, + "loss": 0.9982, + "step": 26770 + }, + { + "epoch": 1.98, + "learning_rate": 1.564759862991264e-05, + "loss": 1.0049, + "step": 26771 + }, + { + "epoch": 1.98, + "learning_rate": 1.564726945450481e-05, + "loss": 1.0189, + "step": 26772 + }, + { + "epoch": 1.98, + "learning_rate": 1.5646940270112258e-05, + "loss": 1.0396, + "step": 26773 + }, + { + "epoch": 1.98, + "learning_rate": 1.5646611076735503e-05, + "loss": 0.9313, + "step": 26774 + }, + { + "epoch": 1.98, + "learning_rate": 1.564628187437507e-05, + "loss": 1.003, + "step": 26775 + }, + { + "epoch": 1.98, + "learning_rate": 1.5645952663031483e-05, + "loss": 1.0916, + "step": 26776 + }, + { + "epoch": 1.98, + "learning_rate": 1.5645623442705264e-05, + "loss": 0.9032, + "step": 26777 + }, + { + "epoch": 1.98, + "learning_rate": 1.564529421339694e-05, + "loss": 0.9736, + "step": 26778 + }, + { + "epoch": 1.98, + "learning_rate": 1.5644964975107034e-05, + "loss": 1.0025, + "step": 26779 + }, + { + "epoch": 1.98, + "learning_rate": 1.5644635727836074e-05, + "loss": 1.0419, + "step": 26780 + }, + { + "epoch": 1.98, + "learning_rate": 1.5644306471584575e-05, + "loss": 1.0153, + "step": 26781 + }, + { + "epoch": 1.98, + "learning_rate": 1.5643977206353067e-05, + "loss": 1.0219, + "step": 26782 + }, + { + "epoch": 1.98, + "learning_rate": 1.564364793214207e-05, + "loss": 1.0116, + "step": 26783 + }, + { + "epoch": 1.98, + "learning_rate": 1.5643318648952113e-05, + "loss": 1.0232, + "step": 26784 + }, + { + "epoch": 1.98, + "learning_rate": 1.5642989356783714e-05, + "loss": 0.9442, + "step": 26785 + }, + { + "epoch": 1.98, + "learning_rate": 1.56426600556374e-05, + "loss": 0.9434, + "step": 26786 + }, + { + "epoch": 1.98, + "learning_rate": 1.5642330745513693e-05, + "loss": 0.9655, + "step": 26787 + }, + { + "epoch": 1.98, + "learning_rate": 1.5642001426413123e-05, + "loss": 0.9233, + "step": 26788 + }, + { + "epoch": 1.98, + "learning_rate": 1.564167209833621e-05, + "loss": 1.0188, + "step": 26789 + }, + { + "epoch": 1.98, + "learning_rate": 1.5641342761283472e-05, + "loss": 0.9497, + "step": 26790 + }, + { + "epoch": 1.98, + "learning_rate": 1.564101341525544e-05, + "loss": 1.0115, + "step": 26791 + }, + { + "epoch": 1.98, + "learning_rate": 1.5640684060252635e-05, + "loss": 1.0105, + "step": 26792 + }, + { + "epoch": 1.98, + "learning_rate": 1.5640354696275585e-05, + "loss": 1.0711, + "step": 26793 + }, + { + "epoch": 1.98, + "learning_rate": 1.5640025323324813e-05, + "loss": 0.8763, + "step": 26794 + }, + { + "epoch": 1.98, + "learning_rate": 1.5639695941400835e-05, + "loss": 1.0095, + "step": 26795 + }, + { + "epoch": 1.98, + "learning_rate": 1.563936655050419e-05, + "loss": 0.9168, + "step": 26796 + }, + { + "epoch": 1.98, + "learning_rate": 1.5639037150635386e-05, + "loss": 1.0892, + "step": 26797 + }, + { + "epoch": 1.98, + "learning_rate": 1.5638707741794957e-05, + "loss": 1.0771, + "step": 26798 + }, + { + "epoch": 1.98, + "learning_rate": 1.5638378323983427e-05, + "loss": 0.8682, + "step": 26799 + }, + { + "epoch": 1.98, + "learning_rate": 1.5638048897201316e-05, + "loss": 1.0097, + "step": 26800 + }, + { + "epoch": 1.98, + "learning_rate": 1.563771946144915e-05, + "loss": 0.9765, + "step": 26801 + }, + { + "epoch": 1.98, + "learning_rate": 1.5637390016727454e-05, + "loss": 0.9585, + "step": 26802 + }, + { + "epoch": 1.98, + "learning_rate": 1.563706056303675e-05, + "loss": 0.968, + "step": 26803 + }, + { + "epoch": 1.98, + "learning_rate": 1.5636731100377565e-05, + "loss": 0.9613, + "step": 26804 + }, + { + "epoch": 1.98, + "learning_rate": 1.563640162875042e-05, + "loss": 0.92, + "step": 26805 + }, + { + "epoch": 1.98, + "learning_rate": 1.5636072148155838e-05, + "loss": 0.9849, + "step": 26806 + }, + { + "epoch": 1.98, + "learning_rate": 1.563574265859435e-05, + "loss": 1.0269, + "step": 26807 + }, + { + "epoch": 1.98, + "learning_rate": 1.5635413160066474e-05, + "loss": 1.0296, + "step": 26808 + }, + { + "epoch": 1.98, + "learning_rate": 1.5635083652572735e-05, + "loss": 1.0055, + "step": 26809 + }, + { + "epoch": 1.98, + "learning_rate": 1.563475413611366e-05, + "loss": 1.0947, + "step": 26810 + }, + { + "epoch": 1.98, + "learning_rate": 1.5634424610689772e-05, + "loss": 1.101, + "step": 26811 + }, + { + "epoch": 1.98, + "learning_rate": 1.5634095076301598e-05, + "loss": 0.9351, + "step": 26812 + }, + { + "epoch": 1.98, + "learning_rate": 1.5633765532949658e-05, + "loss": 1.0437, + "step": 26813 + }, + { + "epoch": 1.98, + "learning_rate": 1.5633435980634476e-05, + "loss": 0.9585, + "step": 26814 + }, + { + "epoch": 1.98, + "learning_rate": 1.5633106419356577e-05, + "loss": 1.034, + "step": 26815 + }, + { + "epoch": 1.98, + "learning_rate": 1.5632776849116487e-05, + "loss": 1.0809, + "step": 26816 + }, + { + "epoch": 1.98, + "learning_rate": 1.563244726991473e-05, + "loss": 1.0341, + "step": 26817 + }, + { + "epoch": 1.98, + "learning_rate": 1.563211768175183e-05, + "loss": 0.9901, + "step": 26818 + }, + { + "epoch": 1.98, + "learning_rate": 1.563178808462831e-05, + "loss": 0.9686, + "step": 26819 + }, + { + "epoch": 1.98, + "learning_rate": 1.56314584785447e-05, + "loss": 0.9938, + "step": 26820 + }, + { + "epoch": 1.98, + "learning_rate": 1.5631128863501514e-05, + "loss": 0.9734, + "step": 26821 + }, + { + "epoch": 1.98, + "learning_rate": 1.5630799239499288e-05, + "loss": 0.8379, + "step": 26822 + }, + { + "epoch": 1.98, + "learning_rate": 1.563046960653854e-05, + "loss": 0.936, + "step": 26823 + }, + { + "epoch": 1.98, + "learning_rate": 1.563013996461979e-05, + "loss": 0.9699, + "step": 26824 + }, + { + "epoch": 1.98, + "learning_rate": 1.5629810313743573e-05, + "loss": 0.9933, + "step": 26825 + }, + { + "epoch": 1.98, + "learning_rate": 1.5629480653910406e-05, + "loss": 0.8257, + "step": 26826 + }, + { + "epoch": 1.98, + "learning_rate": 1.5629150985120818e-05, + "loss": 0.9566, + "step": 26827 + }, + { + "epoch": 1.98, + "learning_rate": 1.562882130737533e-05, + "loss": 0.9871, + "step": 26828 + }, + { + "epoch": 1.98, + "learning_rate": 1.562849162067447e-05, + "loss": 0.9784, + "step": 26829 + }, + { + "epoch": 1.98, + "learning_rate": 1.562816192501876e-05, + "loss": 1.0765, + "step": 26830 + }, + { + "epoch": 1.98, + "learning_rate": 1.5627832220408724e-05, + "loss": 1.0663, + "step": 26831 + }, + { + "epoch": 1.98, + "learning_rate": 1.5627502506844886e-05, + "loss": 0.9854, + "step": 26832 + }, + { + "epoch": 1.98, + "learning_rate": 1.562717278432777e-05, + "loss": 1.0474, + "step": 26833 + }, + { + "epoch": 1.98, + "learning_rate": 1.5626843052857907e-05, + "loss": 1.0227, + "step": 26834 + }, + { + "epoch": 1.98, + "learning_rate": 1.5626513312435816e-05, + "loss": 1.0551, + "step": 26835 + }, + { + "epoch": 1.98, + "learning_rate": 1.5626183563062024e-05, + "loss": 0.9282, + "step": 26836 + }, + { + "epoch": 1.98, + "learning_rate": 1.5625853804737052e-05, + "loss": 1.0309, + "step": 26837 + }, + { + "epoch": 1.98, + "learning_rate": 1.5625524037461432e-05, + "loss": 1.0338, + "step": 26838 + }, + { + "epoch": 1.98, + "learning_rate": 1.5625194261235675e-05, + "loss": 1.105, + "step": 26839 + }, + { + "epoch": 1.98, + "learning_rate": 1.562486447606032e-05, + "loss": 0.9162, + "step": 26840 + }, + { + "epoch": 1.98, + "learning_rate": 1.5624534681935888e-05, + "loss": 0.9894, + "step": 26841 + }, + { + "epoch": 1.98, + "learning_rate": 1.56242048788629e-05, + "loss": 1.0026, + "step": 26842 + }, + { + "epoch": 1.98, + "learning_rate": 1.562387506684188e-05, + "loss": 1.0313, + "step": 26843 + }, + { + "epoch": 1.98, + "learning_rate": 1.5623545245873356e-05, + "loss": 1.0289, + "step": 26844 + }, + { + "epoch": 1.98, + "learning_rate": 1.5623215415957854e-05, + "loss": 1.0672, + "step": 26845 + }, + { + "epoch": 1.98, + "learning_rate": 1.5622885577095895e-05, + "loss": 1.0134, + "step": 26846 + }, + { + "epoch": 1.98, + "learning_rate": 1.562255572928801e-05, + "loss": 1.0736, + "step": 26847 + }, + { + "epoch": 1.98, + "learning_rate": 1.5622225872534717e-05, + "loss": 0.9457, + "step": 26848 + }, + { + "epoch": 1.98, + "learning_rate": 1.5621896006836538e-05, + "loss": 0.981, + "step": 26849 + }, + { + "epoch": 1.98, + "learning_rate": 1.5621566132194006e-05, + "loss": 0.9982, + "step": 26850 + }, + { + "epoch": 1.98, + "learning_rate": 1.5621236248607643e-05, + "loss": 1.0036, + "step": 26851 + }, + { + "epoch": 1.98, + "learning_rate": 1.5620906356077975e-05, + "loss": 0.9924, + "step": 26852 + }, + { + "epoch": 1.98, + "learning_rate": 1.5620576454605526e-05, + "loss": 0.999, + "step": 26853 + }, + { + "epoch": 1.98, + "learning_rate": 1.5620246544190817e-05, + "loss": 0.9848, + "step": 26854 + }, + { + "epoch": 1.98, + "learning_rate": 1.5619916624834375e-05, + "loss": 1.0437, + "step": 26855 + }, + { + "epoch": 1.98, + "learning_rate": 1.5619586696536728e-05, + "loss": 0.9994, + "step": 26856 + }, + { + "epoch": 1.98, + "learning_rate": 1.5619256759298404e-05, + "loss": 1.0488, + "step": 26857 + }, + { + "epoch": 1.98, + "learning_rate": 1.5618926813119918e-05, + "loss": 1.0548, + "step": 26858 + }, + { + "epoch": 1.98, + "learning_rate": 1.56185968580018e-05, + "loss": 1.0298, + "step": 26859 + }, + { + "epoch": 1.98, + "learning_rate": 1.5618266893944576e-05, + "loss": 1.0511, + "step": 26860 + }, + { + "epoch": 1.98, + "learning_rate": 1.561793692094877e-05, + "loss": 1.0931, + "step": 26861 + }, + { + "epoch": 1.98, + "learning_rate": 1.5617606939014902e-05, + "loss": 1.0894, + "step": 26862 + }, + { + "epoch": 1.98, + "learning_rate": 1.5617276948143505e-05, + "loss": 0.9556, + "step": 26863 + }, + { + "epoch": 1.98, + "learning_rate": 1.56169469483351e-05, + "loss": 1.1064, + "step": 26864 + }, + { + "epoch": 1.99, + "learning_rate": 1.5616616939590215e-05, + "loss": 0.9458, + "step": 26865 + }, + { + "epoch": 1.99, + "learning_rate": 1.5616286921909372e-05, + "loss": 1.0436, + "step": 26866 + }, + { + "epoch": 1.99, + "learning_rate": 1.5615956895293097e-05, + "loss": 1.0516, + "step": 26867 + }, + { + "epoch": 1.99, + "learning_rate": 1.5615626859741915e-05, + "loss": 0.9726, + "step": 26868 + }, + { + "epoch": 1.99, + "learning_rate": 1.561529681525635e-05, + "loss": 0.9071, + "step": 26869 + }, + { + "epoch": 1.99, + "learning_rate": 1.561496676183693e-05, + "loss": 0.9266, + "step": 26870 + }, + { + "epoch": 1.99, + "learning_rate": 1.5614636699484175e-05, + "loss": 1.0371, + "step": 26871 + }, + { + "epoch": 1.99, + "learning_rate": 1.5614306628198614e-05, + "loss": 0.9225, + "step": 26872 + }, + { + "epoch": 1.99, + "learning_rate": 1.5613976547980777e-05, + "loss": 0.896, + "step": 26873 + }, + { + "epoch": 1.99, + "learning_rate": 1.5613646458831176e-05, + "loss": 0.943, + "step": 26874 + }, + { + "epoch": 1.99, + "learning_rate": 1.561331636075035e-05, + "loss": 0.935, + "step": 26875 + }, + { + "epoch": 1.99, + "learning_rate": 1.5612986253738815e-05, + "loss": 1.004, + "step": 26876 + }, + { + "epoch": 1.99, + "learning_rate": 1.56126561377971e-05, + "loss": 0.9382, + "step": 26877 + }, + { + "epoch": 1.99, + "learning_rate": 1.5612326012925727e-05, + "loss": 1.0167, + "step": 26878 + }, + { + "epoch": 1.99, + "learning_rate": 1.5611995879125225e-05, + "loss": 1.0887, + "step": 26879 + }, + { + "epoch": 1.99, + "learning_rate": 1.561166573639612e-05, + "loss": 0.9475, + "step": 26880 + }, + { + "epoch": 1.99, + "learning_rate": 1.561133558473893e-05, + "loss": 0.8497, + "step": 26881 + }, + { + "epoch": 1.99, + "learning_rate": 1.561100542415419e-05, + "loss": 0.9665, + "step": 26882 + }, + { + "epoch": 1.99, + "learning_rate": 1.5610675254642418e-05, + "loss": 0.9075, + "step": 26883 + }, + { + "epoch": 1.99, + "learning_rate": 1.5610345076204144e-05, + "loss": 1.0319, + "step": 26884 + }, + { + "epoch": 1.99, + "learning_rate": 1.561001488883989e-05, + "loss": 0.9848, + "step": 26885 + }, + { + "epoch": 1.99, + "learning_rate": 1.5609684692550184e-05, + "loss": 0.9493, + "step": 26886 + }, + { + "epoch": 1.99, + "learning_rate": 1.5609354487335552e-05, + "loss": 0.966, + "step": 26887 + }, + { + "epoch": 1.99, + "learning_rate": 1.5609024273196513e-05, + "loss": 0.9758, + "step": 26888 + }, + { + "epoch": 1.99, + "learning_rate": 1.56086940501336e-05, + "loss": 1.0329, + "step": 26889 + }, + { + "epoch": 1.99, + "learning_rate": 1.5608363818147328e-05, + "loss": 1.0183, + "step": 26890 + }, + { + "epoch": 1.99, + "learning_rate": 1.5608033577238236e-05, + "loss": 1.0027, + "step": 26891 + }, + { + "epoch": 1.99, + "learning_rate": 1.560770332740684e-05, + "loss": 0.9986, + "step": 26892 + }, + { + "epoch": 1.99, + "learning_rate": 1.560737306865367e-05, + "loss": 1.0371, + "step": 26893 + }, + { + "epoch": 1.99, + "learning_rate": 1.5607042800979247e-05, + "loss": 1.1131, + "step": 26894 + }, + { + "epoch": 1.99, + "learning_rate": 1.5606712524384105e-05, + "loss": 1.0822, + "step": 26895 + }, + { + "epoch": 1.99, + "learning_rate": 1.5606382238868757e-05, + "loss": 0.9962, + "step": 26896 + }, + { + "epoch": 1.99, + "learning_rate": 1.560605194443374e-05, + "loss": 1.0896, + "step": 26897 + }, + { + "epoch": 1.99, + "learning_rate": 1.5605721641079572e-05, + "loss": 0.9752, + "step": 26898 + }, + { + "epoch": 1.99, + "learning_rate": 1.560539132880678e-05, + "loss": 0.9781, + "step": 26899 + }, + { + "epoch": 1.99, + "learning_rate": 1.560506100761589e-05, + "loss": 0.989, + "step": 26900 + }, + { + "epoch": 1.99, + "learning_rate": 1.5604730677507433e-05, + "loss": 1.0221, + "step": 26901 + }, + { + "epoch": 1.99, + "learning_rate": 1.5604400338481927e-05, + "loss": 1.0077, + "step": 26902 + }, + { + "epoch": 1.99, + "learning_rate": 1.56040699905399e-05, + "loss": 1.0278, + "step": 26903 + }, + { + "epoch": 1.99, + "learning_rate": 1.560373963368188e-05, + "loss": 0.9972, + "step": 26904 + }, + { + "epoch": 1.99, + "learning_rate": 1.5603409267908385e-05, + "loss": 1.0222, + "step": 26905 + }, + { + "epoch": 1.99, + "learning_rate": 1.5603078893219953e-05, + "loss": 0.9218, + "step": 26906 + }, + { + "epoch": 1.99, + "learning_rate": 1.5602748509617095e-05, + "loss": 0.888, + "step": 26907 + }, + { + "epoch": 1.99, + "learning_rate": 1.5602418117100347e-05, + "loss": 1.0263, + "step": 26908 + }, + { + "epoch": 1.99, + "learning_rate": 1.5602087715670235e-05, + "loss": 0.971, + "step": 26909 + }, + { + "epoch": 1.99, + "learning_rate": 1.560175730532728e-05, + "loss": 1.1152, + "step": 26910 + }, + { + "epoch": 1.99, + "learning_rate": 1.560142688607201e-05, + "loss": 0.9507, + "step": 26911 + }, + { + "epoch": 1.99, + "learning_rate": 1.560109645790495e-05, + "loss": 1.0575, + "step": 26912 + }, + { + "epoch": 1.99, + "learning_rate": 1.5600766020826624e-05, + "loss": 1.0138, + "step": 26913 + }, + { + "epoch": 1.99, + "learning_rate": 1.560043557483756e-05, + "loss": 0.89, + "step": 26914 + }, + { + "epoch": 1.99, + "learning_rate": 1.5600105119938283e-05, + "loss": 1.0116, + "step": 26915 + }, + { + "epoch": 1.99, + "learning_rate": 1.5599774656129317e-05, + "loss": 1.0214, + "step": 26916 + }, + { + "epoch": 1.99, + "learning_rate": 1.5599444183411194e-05, + "loss": 1.026, + "step": 26917 + }, + { + "epoch": 1.99, + "learning_rate": 1.5599113701784432e-05, + "loss": 1.01, + "step": 26918 + }, + { + "epoch": 1.99, + "learning_rate": 1.559878321124956e-05, + "loss": 0.9494, + "step": 26919 + }, + { + "epoch": 1.99, + "learning_rate": 1.5598452711807107e-05, + "loss": 0.938, + "step": 26920 + }, + { + "epoch": 1.99, + "learning_rate": 1.5598122203457594e-05, + "loss": 1.0913, + "step": 26921 + }, + { + "epoch": 1.99, + "learning_rate": 1.559779168620155e-05, + "loss": 1.0758, + "step": 26922 + }, + { + "epoch": 1.99, + "learning_rate": 1.5597461160039497e-05, + "loss": 0.971, + "step": 26923 + }, + { + "epoch": 1.99, + "learning_rate": 1.5597130624971966e-05, + "loss": 1.0162, + "step": 26924 + }, + { + "epoch": 1.99, + "learning_rate": 1.559680008099948e-05, + "loss": 1.085, + "step": 26925 + }, + { + "epoch": 1.99, + "learning_rate": 1.5596469528122563e-05, + "loss": 1.0372, + "step": 26926 + }, + { + "epoch": 1.99, + "learning_rate": 1.5596138966341745e-05, + "loss": 1.0057, + "step": 26927 + }, + { + "epoch": 1.99, + "learning_rate": 1.559580839565755e-05, + "loss": 1.132, + "step": 26928 + }, + { + "epoch": 1.99, + "learning_rate": 1.5595477816070502e-05, + "loss": 1.0347, + "step": 26929 + }, + { + "epoch": 1.99, + "learning_rate": 1.5595147227581132e-05, + "loss": 0.9959, + "step": 26930 + }, + { + "epoch": 1.99, + "learning_rate": 1.5594816630189957e-05, + "loss": 1.0579, + "step": 26931 + }, + { + "epoch": 1.99, + "learning_rate": 1.5594486023897516e-05, + "loss": 1.0092, + "step": 26932 + }, + { + "epoch": 1.99, + "learning_rate": 1.5594155408704325e-05, + "loss": 1.02, + "step": 26933 + }, + { + "epoch": 1.99, + "learning_rate": 1.559382478461091e-05, + "loss": 1.0023, + "step": 26934 + }, + { + "epoch": 1.99, + "learning_rate": 1.5593494151617802e-05, + "loss": 1.0292, + "step": 26935 + }, + { + "epoch": 1.99, + "learning_rate": 1.5593163509725522e-05, + "loss": 0.9852, + "step": 26936 + }, + { + "epoch": 1.99, + "learning_rate": 1.5592832858934602e-05, + "loss": 1.0174, + "step": 26937 + }, + { + "epoch": 1.99, + "learning_rate": 1.5592502199245567e-05, + "loss": 1.0391, + "step": 26938 + }, + { + "epoch": 1.99, + "learning_rate": 1.559217153065894e-05, + "loss": 1.0647, + "step": 26939 + }, + { + "epoch": 1.99, + "learning_rate": 1.5591840853175245e-05, + "loss": 0.9742, + "step": 26940 + }, + { + "epoch": 1.99, + "learning_rate": 1.5591510166795013e-05, + "loss": 1.0431, + "step": 26941 + }, + { + "epoch": 1.99, + "learning_rate": 1.5591179471518773e-05, + "loss": 0.9852, + "step": 26942 + }, + { + "epoch": 1.99, + "learning_rate": 1.559084876734704e-05, + "loss": 0.9446, + "step": 26943 + }, + { + "epoch": 1.99, + "learning_rate": 1.5590518054280346e-05, + "loss": 1.0092, + "step": 26944 + }, + { + "epoch": 1.99, + "learning_rate": 1.559018733231922e-05, + "loss": 1.0185, + "step": 26945 + }, + { + "epoch": 1.99, + "learning_rate": 1.558985660146419e-05, + "loss": 1.0279, + "step": 26946 + }, + { + "epoch": 1.99, + "learning_rate": 1.558952586171577e-05, + "loss": 1.0151, + "step": 26947 + }, + { + "epoch": 1.99, + "learning_rate": 1.5589195113074497e-05, + "loss": 0.9723, + "step": 26948 + }, + { + "epoch": 1.99, + "learning_rate": 1.5588864355540898e-05, + "loss": 1.0704, + "step": 26949 + }, + { + "epoch": 1.99, + "learning_rate": 1.5588533589115496e-05, + "loss": 0.969, + "step": 26950 + }, + { + "epoch": 1.99, + "learning_rate": 1.5588202813798813e-05, + "loss": 1.0396, + "step": 26951 + }, + { + "epoch": 1.99, + "learning_rate": 1.558787202959138e-05, + "loss": 0.9738, + "step": 26952 + }, + { + "epoch": 1.99, + "learning_rate": 1.5587541236493725e-05, + "loss": 0.9332, + "step": 26953 + }, + { + "epoch": 1.99, + "learning_rate": 1.5587210434506373e-05, + "loss": 0.9385, + "step": 26954 + }, + { + "epoch": 1.99, + "learning_rate": 1.5586879623629844e-05, + "loss": 1.0025, + "step": 26955 + }, + { + "epoch": 1.99, + "learning_rate": 1.558654880386467e-05, + "loss": 1.0627, + "step": 26956 + }, + { + "epoch": 1.99, + "learning_rate": 1.5586217975211383e-05, + "loss": 0.948, + "step": 26957 + }, + { + "epoch": 1.99, + "learning_rate": 1.5585887137670498e-05, + "loss": 1.0053, + "step": 26958 + }, + { + "epoch": 1.99, + "learning_rate": 1.5585556291242546e-05, + "loss": 0.9868, + "step": 26959 + }, + { + "epoch": 1.99, + "learning_rate": 1.5585225435928054e-05, + "loss": 0.9124, + "step": 26960 + }, + { + "epoch": 1.99, + "learning_rate": 1.5584894571727555e-05, + "loss": 1.0442, + "step": 26961 + }, + { + "epoch": 1.99, + "learning_rate": 1.5584563698641562e-05, + "loss": 1.1521, + "step": 26962 + }, + { + "epoch": 1.99, + "learning_rate": 1.5584232816670607e-05, + "loss": 0.8897, + "step": 26963 + }, + { + "epoch": 1.99, + "learning_rate": 1.5583901925815222e-05, + "loss": 0.9631, + "step": 26964 + }, + { + "epoch": 1.99, + "learning_rate": 1.5583571026075925e-05, + "loss": 1.0962, + "step": 26965 + }, + { + "epoch": 1.99, + "learning_rate": 1.5583240117453247e-05, + "loss": 0.9485, + "step": 26966 + }, + { + "epoch": 1.99, + "learning_rate": 1.5582909199947715e-05, + "loss": 1.01, + "step": 26967 + }, + { + "epoch": 1.99, + "learning_rate": 1.5582578273559853e-05, + "loss": 1.099, + "step": 26968 + }, + { + "epoch": 1.99, + "learning_rate": 1.5582247338290194e-05, + "loss": 1.0355, + "step": 26969 + }, + { + "epoch": 1.99, + "learning_rate": 1.5581916394139255e-05, + "loss": 0.956, + "step": 26970 + }, + { + "epoch": 1.99, + "learning_rate": 1.5581585441107568e-05, + "loss": 1.0625, + "step": 26971 + }, + { + "epoch": 1.99, + "learning_rate": 1.5581254479195656e-05, + "loss": 1.0537, + "step": 26972 + }, + { + "epoch": 1.99, + "learning_rate": 1.558092350840405e-05, + "loss": 0.98, + "step": 26973 + }, + { + "epoch": 1.99, + "learning_rate": 1.5580592528733273e-05, + "loss": 1.0758, + "step": 26974 + }, + { + "epoch": 1.99, + "learning_rate": 1.5580261540183855e-05, + "loss": 1.0459, + "step": 26975 + }, + { + "epoch": 1.99, + "learning_rate": 1.5579930542756322e-05, + "loss": 0.9996, + "step": 26976 + }, + { + "epoch": 1.99, + "learning_rate": 1.5579599536451197e-05, + "loss": 0.9602, + "step": 26977 + }, + { + "epoch": 1.99, + "learning_rate": 1.557926852126901e-05, + "loss": 0.9684, + "step": 26978 + }, + { + "epoch": 1.99, + "learning_rate": 1.5578937497210284e-05, + "loss": 1.001, + "step": 26979 + }, + { + "epoch": 1.99, + "learning_rate": 1.5578606464275553e-05, + "loss": 1.0174, + "step": 26980 + }, + { + "epoch": 1.99, + "learning_rate": 1.5578275422465337e-05, + "loss": 1.0751, + "step": 26981 + }, + { + "epoch": 1.99, + "learning_rate": 1.557794437178016e-05, + "loss": 1.0968, + "step": 26982 + }, + { + "epoch": 1.99, + "learning_rate": 1.557761331222056e-05, + "loss": 0.9887, + "step": 26983 + }, + { + "epoch": 1.99, + "learning_rate": 1.5577282243787055e-05, + "loss": 0.9888, + "step": 26984 + }, + { + "epoch": 1.99, + "learning_rate": 1.557695116648017e-05, + "loss": 1.0886, + "step": 26985 + }, + { + "epoch": 1.99, + "learning_rate": 1.557662008030044e-05, + "loss": 0.9547, + "step": 26986 + }, + { + "epoch": 1.99, + "learning_rate": 1.5576288985248384e-05, + "loss": 0.9956, + "step": 26987 + }, + { + "epoch": 1.99, + "learning_rate": 1.5575957881324538e-05, + "loss": 1.0166, + "step": 26988 + }, + { + "epoch": 1.99, + "learning_rate": 1.5575626768529417e-05, + "loss": 1.0206, + "step": 26989 + }, + { + "epoch": 1.99, + "learning_rate": 1.5575295646863554e-05, + "loss": 0.9041, + "step": 26990 + }, + { + "epoch": 1.99, + "learning_rate": 1.5574964516327477e-05, + "loss": 0.9991, + "step": 26991 + }, + { + "epoch": 1.99, + "learning_rate": 1.557463337692171e-05, + "loss": 0.9968, + "step": 26992 + }, + { + "epoch": 1.99, + "learning_rate": 1.5574302228646783e-05, + "loss": 1.0271, + "step": 26993 + }, + { + "epoch": 1.99, + "learning_rate": 1.557397107150322e-05, + "loss": 0.9972, + "step": 26994 + }, + { + "epoch": 1.99, + "learning_rate": 1.5573639905491546e-05, + "loss": 1.0123, + "step": 26995 + }, + { + "epoch": 1.99, + "learning_rate": 1.5573308730612294e-05, + "loss": 0.995, + "step": 26996 + }, + { + "epoch": 1.99, + "learning_rate": 1.5572977546865987e-05, + "loss": 0.9692, + "step": 26997 + }, + { + "epoch": 1.99, + "learning_rate": 1.5572646354253153e-05, + "loss": 0.9737, + "step": 26998 + }, + { + "epoch": 1.99, + "learning_rate": 1.557231515277432e-05, + "loss": 1.0621, + "step": 26999 + }, + { + "epoch": 2.0, + "learning_rate": 1.5571983942430005e-05, + "loss": 0.9535, + "step": 27000 + }, + { + "epoch": 2.0, + "learning_rate": 1.5571652723220753e-05, + "loss": 0.9986, + "step": 27001 + }, + { + "epoch": 2.0, + "learning_rate": 1.5571321495147077e-05, + "loss": 0.9144, + "step": 27002 + }, + { + "epoch": 2.0, + "learning_rate": 1.557099025820951e-05, + "loss": 1.06, + "step": 27003 + }, + { + "epoch": 2.0, + "learning_rate": 1.5570659012408573e-05, + "loss": 0.9891, + "step": 27004 + }, + { + "epoch": 2.0, + "learning_rate": 1.5570327757744798e-05, + "loss": 1.0113, + "step": 27005 + }, + { + "epoch": 2.0, + "learning_rate": 1.5569996494218718e-05, + "loss": 1.0044, + "step": 27006 + }, + { + "epoch": 2.0, + "learning_rate": 1.5569665221830846e-05, + "loss": 1.0115, + "step": 27007 + }, + { + "epoch": 2.0, + "learning_rate": 1.556933394058172e-05, + "loss": 1.001, + "step": 27008 + }, + { + "epoch": 2.0, + "learning_rate": 1.5569002650471862e-05, + "loss": 1.0363, + "step": 27009 + }, + { + "epoch": 2.0, + "learning_rate": 1.55686713515018e-05, + "loss": 1.113, + "step": 27010 + }, + { + "epoch": 2.0, + "learning_rate": 1.556834004367206e-05, + "loss": 1.0355, + "step": 27011 + }, + { + "epoch": 2.0, + "learning_rate": 1.5568008726983177e-05, + "loss": 1.0248, + "step": 27012 + }, + { + "epoch": 2.0, + "learning_rate": 1.5567677401435667e-05, + "loss": 0.9844, + "step": 27013 + }, + { + "epoch": 2.0, + "learning_rate": 1.5567346067030062e-05, + "loss": 0.9642, + "step": 27014 + }, + { + "epoch": 2.0, + "learning_rate": 1.556701472376689e-05, + "loss": 1.0382, + "step": 27015 + }, + { + "epoch": 2.0, + "learning_rate": 1.5566683371646674e-05, + "loss": 0.8885, + "step": 27016 + }, + { + "epoch": 2.0, + "learning_rate": 1.5566352010669948e-05, + "loss": 1.0528, + "step": 27017 + }, + { + "epoch": 2.0, + "learning_rate": 1.5566020640837235e-05, + "loss": 1.0039, + "step": 27018 + }, + { + "epoch": 2.0, + "learning_rate": 1.5565689262149063e-05, + "loss": 0.9833, + "step": 27019 + }, + { + "epoch": 2.0, + "learning_rate": 1.5565357874605958e-05, + "loss": 0.9518, + "step": 27020 + }, + { + "epoch": 2.0, + "learning_rate": 1.5565026478208448e-05, + "loss": 0.9942, + "step": 27021 + }, + { + "epoch": 2.0, + "learning_rate": 1.5564695072957062e-05, + "loss": 1.0139, + "step": 27022 + }, + { + "epoch": 2.0, + "learning_rate": 1.5564363658852325e-05, + "loss": 0.9917, + "step": 27023 + }, + { + "epoch": 2.0, + "learning_rate": 1.556403223589477e-05, + "loss": 0.9704, + "step": 27024 + }, + { + "epoch": 2.0, + "learning_rate": 1.5563700804084912e-05, + "loss": 0.9517, + "step": 27025 + }, + { + "epoch": 2.0, + "learning_rate": 1.556336936342329e-05, + "loss": 1.0764, + "step": 27026 + }, + { + "epoch": 2.0, + "learning_rate": 1.5563037913910425e-05, + "loss": 1.0333, + "step": 27027 + }, + { + "epoch": 2.0, + "learning_rate": 1.5562706455546844e-05, + "loss": 1.0339, + "step": 27028 + }, + { + "epoch": 2.0, + "learning_rate": 1.556237498833308e-05, + "loss": 1.0121, + "step": 27029 + }, + { + "epoch": 2.0, + "learning_rate": 1.5562043512269655e-05, + "loss": 1.1635, + "step": 27030 + }, + { + "epoch": 2.0, + "learning_rate": 1.5561712027357097e-05, + "loss": 0.9138, + "step": 27031 + }, + { + "epoch": 2.0, + "learning_rate": 1.556138053359594e-05, + "loss": 0.8833, + "step": 27032 + }, + { + "epoch": 2.0, + "learning_rate": 1.55610490309867e-05, + "loss": 0.969, + "step": 27033 + }, + { + "epoch": 2.0, + "learning_rate": 1.5560717519529916e-05, + "loss": 1.0632, + "step": 27034 + }, + { + "epoch": 2.0, + "learning_rate": 1.556038599922611e-05, + "loss": 1.0502, + "step": 27035 + }, + { + "epoch": 2.0, + "learning_rate": 1.5560054470075806e-05, + "loss": 1.0085, + "step": 27036 + }, + { + "epoch": 2.0, + "learning_rate": 1.5559722932079537e-05, + "loss": 0.9567, + "step": 27037 + }, + { + "epoch": 2.0, + "learning_rate": 1.555939138523783e-05, + "loss": 1.1171, + "step": 27038 + }, + { + "epoch": 2.0, + "learning_rate": 1.5559059829551203e-05, + "loss": 0.9394, + "step": 27039 + }, + { + "epoch": 2.0, + "learning_rate": 1.5558728265020198e-05, + "loss": 1.023, + "step": 27040 + }, + { + "epoch": 2.0, + "learning_rate": 1.5558396691645335e-05, + "loss": 0.9868, + "step": 27041 + }, + { + "epoch": 2.0, + "learning_rate": 1.555806510942714e-05, + "loss": 1.1336, + "step": 27042 + }, + { + "epoch": 2.0, + "learning_rate": 1.5557733518366144e-05, + "loss": 1.0172, + "step": 27043 + }, + { + "epoch": 2.0, + "learning_rate": 1.5557401918462877e-05, + "loss": 1.0553, + "step": 27044 + }, + { + "epoch": 2.0, + "learning_rate": 1.555707030971786e-05, + "loss": 1.0027, + "step": 27045 + }, + { + "epoch": 2.0, + "learning_rate": 1.5556738692131624e-05, + "loss": 0.8624, + "step": 27046 + }, + { + "epoch": 2.0, + "learning_rate": 1.5556407065704696e-05, + "loss": 0.9529, + "step": 27047 + }, + { + "epoch": 2.0, + "learning_rate": 1.5556075430437604e-05, + "loss": 1.0623, + "step": 27048 + }, + { + "epoch": 2.0, + "learning_rate": 1.555574378633088e-05, + "loss": 0.9202, + "step": 27049 + }, + { + "epoch": 2.0, + "learning_rate": 1.5555412133385036e-05, + "loss": 1.026, + "step": 27050 + }, + { + "epoch": 2.0, + "learning_rate": 1.5555080471600618e-05, + "loss": 1.0164, + "step": 27051 + }, + { + "epoch": 2.0, + "learning_rate": 1.5554748800978146e-05, + "loss": 1.0548, + "step": 27052 + }, + { + "epoch": 2.0, + "learning_rate": 1.555441712151815e-05, + "loss": 1.0009, + "step": 27053 + }, + { + "epoch": 2.0, + "learning_rate": 1.5554085433221153e-05, + "loss": 0.9989, + "step": 27054 + }, + { + "epoch": 2.0, + "learning_rate": 1.5553753736087686e-05, + "loss": 0.9421, + "step": 27055 + }, + { + "epoch": 2.0, + "learning_rate": 1.555342203011828e-05, + "loss": 0.991, + "step": 27056 + }, + { + "epoch": 2.0, + "learning_rate": 1.5553090315313454e-05, + "loss": 1.0166, + "step": 27057 + }, + { + "epoch": 2.0, + "learning_rate": 1.5552758591673743e-05, + "loss": 1.0112, + "step": 27058 + }, + { + "epoch": 2.0, + "learning_rate": 1.555242685919967e-05, + "loss": 0.8991, + "step": 27059 + }, + { + "epoch": 2.0, + "learning_rate": 1.555209511789177e-05, + "loss": 0.9368, + "step": 27060 + }, + { + "epoch": 2.0, + "learning_rate": 1.5551763367750563e-05, + "loss": 1.015, + "step": 27061 + }, + { + "epoch": 2.0, + "learning_rate": 1.555143160877658e-05, + "loss": 0.9951, + "step": 27062 + }, + { + "epoch": 2.0, + "learning_rate": 1.555109984097035e-05, + "loss": 1.0454, + "step": 27063 + }, + { + "epoch": 2.0, + "learning_rate": 1.5550768064332396e-05, + "loss": 0.9116, + "step": 27064 + }, + { + "epoch": 2.0, + "learning_rate": 1.555043627886325e-05, + "loss": 1.0048, + "step": 27065 + }, + { + "epoch": 2.0, + "learning_rate": 1.5550104484563446e-05, + "loss": 1.0825, + "step": 27066 + }, + { + "epoch": 2.0, + "learning_rate": 1.5549772681433498e-05, + "loss": 0.9646, + "step": 27067 + }, + { + "epoch": 2.0, + "learning_rate": 1.5549440869473947e-05, + "loss": 1.0928, + "step": 27068 + }, + { + "epoch": 2.0, + "learning_rate": 1.554910904868531e-05, + "loss": 1.1226, + "step": 27069 + }, + { + "epoch": 2.0, + "learning_rate": 1.554877721906812e-05, + "loss": 1.0746, + "step": 27070 + }, + { + "epoch": 2.0, + "learning_rate": 1.5548445380622906e-05, + "loss": 0.9853, + "step": 27071 + }, + { + "epoch": 2.0, + "learning_rate": 1.5548113533350196e-05, + "loss": 1.0096, + "step": 27072 + }, + { + "epoch": 2.0, + "learning_rate": 1.5547781677250514e-05, + "loss": 0.9879, + "step": 27073 + }, + { + "epoch": 2.0, + "learning_rate": 1.5547449812324394e-05, + "loss": 0.9769, + "step": 27074 + }, + { + "epoch": 2.0, + "learning_rate": 1.554711793857236e-05, + "loss": 0.8638, + "step": 27075 + }, + { + "epoch": 2.0, + "learning_rate": 1.554678605599494e-05, + "loss": 1.0427, + "step": 27076 + }, + { + "epoch": 2.0, + "learning_rate": 1.5546454164592662e-05, + "loss": 1.0292, + "step": 27077 + }, + { + "epoch": 2.0, + "learning_rate": 1.5546122264366058e-05, + "loss": 0.9833, + "step": 27078 + }, + { + "epoch": 2.0, + "learning_rate": 1.554579035531565e-05, + "loss": 1.0319, + "step": 27079 + }, + { + "epoch": 2.0, + "learning_rate": 1.5545458437441967e-05, + "loss": 0.9759, + "step": 27080 + }, + { + "epoch": 2.0, + "learning_rate": 1.5545126510745542e-05, + "loss": 1.0245, + "step": 27081 + }, + { + "epoch": 2.0, + "learning_rate": 1.55447945752269e-05, + "loss": 0.9817, + "step": 27082 + }, + { + "epoch": 2.0, + "learning_rate": 1.5544462630886568e-05, + "loss": 0.9903, + "step": 27083 + }, + { + "epoch": 2.0, + "learning_rate": 1.5544130677725076e-05, + "loss": 0.9717, + "step": 27084 + }, + { + "epoch": 2.0, + "learning_rate": 1.5543798715742952e-05, + "loss": 1.2064, + "step": 27085 + }, + { + "epoch": 2.0, + "learning_rate": 1.554346674494072e-05, + "loss": 1.0416, + "step": 27086 + }, + { + "epoch": 2.0, + "learning_rate": 1.5543134765318913e-05, + "loss": 1.0341, + "step": 27087 + }, + { + "epoch": 2.0, + "learning_rate": 1.5542802776878062e-05, + "loss": 1.0697, + "step": 27088 + }, + { + "epoch": 2.0, + "learning_rate": 1.554247077961869e-05, + "loss": 1.0082, + "step": 27089 + }, + { + "epoch": 2.0, + "learning_rate": 1.554213877354132e-05, + "loss": 0.9768, + "step": 27090 + }, + { + "epoch": 2.0, + "learning_rate": 1.5541806758646493e-05, + "loss": 1.0444, + "step": 27091 + }, + { + "epoch": 2.0, + "learning_rate": 1.5541474734934732e-05, + "loss": 1.0184, + "step": 27092 + }, + { + "epoch": 2.0, + "learning_rate": 1.5541142702406558e-05, + "loss": 1.0656, + "step": 27093 + }, + { + "epoch": 2.0, + "learning_rate": 1.554081066106251e-05, + "loss": 0.9643, + "step": 27094 + }, + { + "epoch": 2.0, + "learning_rate": 1.554047861090311e-05, + "loss": 1.0571, + "step": 27095 + }, + { + "epoch": 2.0, + "learning_rate": 1.5540146551928884e-05, + "loss": 0.986, + "step": 27096 + }, + { + "epoch": 2.0, + "learning_rate": 1.5539814484140368e-05, + "loss": 1.0954, + "step": 27097 + }, + { + "epoch": 2.0, + "learning_rate": 1.5539482407538088e-05, + "loss": 1.0359, + "step": 27098 + }, + { + "epoch": 2.0, + "learning_rate": 1.553915032212257e-05, + "loss": 1.0873, + "step": 27099 + }, + { + "epoch": 2.0, + "learning_rate": 1.553881822789434e-05, + "loss": 1.0684, + "step": 27100 + }, + { + "epoch": 2.0, + "learning_rate": 1.5538486124853932e-05, + "loss": 0.9138, + "step": 27101 + }, + { + "epoch": 2.0, + "learning_rate": 1.553815401300187e-05, + "loss": 1.0768, + "step": 27102 + }, + { + "epoch": 2.0, + "learning_rate": 1.5537821892338686e-05, + "loss": 0.9811, + "step": 27103 + }, + { + "epoch": 2.0, + "learning_rate": 1.5537489762864904e-05, + "loss": 0.9573, + "step": 27104 + }, + { + "epoch": 2.0, + "learning_rate": 1.5537157624581062e-05, + "loss": 0.8854, + "step": 27105 + }, + { + "epoch": 2.0, + "learning_rate": 1.5536825477487673e-05, + "loss": 1.113, + "step": 27106 + }, + { + "epoch": 2.0, + "learning_rate": 1.553649332158528e-05, + "loss": 0.9633, + "step": 27107 + }, + { + "epoch": 2.0, + "learning_rate": 1.55361611568744e-05, + "loss": 1.0339, + "step": 27108 + }, + { + "epoch": 2.0, + "learning_rate": 1.5535828983355573e-05, + "loss": 0.9768, + "step": 27109 + }, + { + "epoch": 2.0, + "learning_rate": 1.5535496801029314e-05, + "loss": 0.948, + "step": 27110 + }, + { + "epoch": 2.0, + "learning_rate": 1.5535164609896165e-05, + "loss": 1.0929, + "step": 27111 + }, + { + "epoch": 2.0, + "learning_rate": 1.5534832409956645e-05, + "loss": 1.0495, + "step": 27112 + }, + { + "epoch": 2.0, + "learning_rate": 1.5534500201211286e-05, + "loss": 0.9015, + "step": 27113 + }, + { + "epoch": 2.0, + "learning_rate": 1.5534167983660616e-05, + "loss": 0.9889, + "step": 27114 + }, + { + "epoch": 2.0, + "learning_rate": 1.5533835757305168e-05, + "loss": 0.9493, + "step": 27115 + }, + { + "epoch": 2.0, + "learning_rate": 1.5533503522145462e-05, + "loss": 1.0664, + "step": 27116 + }, + { + "epoch": 2.0, + "learning_rate": 1.553317127818203e-05, + "loss": 0.9772, + "step": 27117 + }, + { + "epoch": 2.0, + "learning_rate": 1.5532839025415408e-05, + "loss": 0.967, + "step": 27118 + }, + { + "epoch": 2.0, + "learning_rate": 1.5532506763846113e-05, + "loss": 1.0643, + "step": 27119 + }, + { + "epoch": 2.0, + "learning_rate": 1.553217449347468e-05, + "loss": 1.1379, + "step": 27120 + }, + { + "epoch": 2.0, + "learning_rate": 1.5531842214301637e-05, + "loss": 0.9891, + "step": 27121 + }, + { + "epoch": 2.0, + "learning_rate": 1.5531509926327513e-05, + "loss": 1.0467, + "step": 27122 + }, + { + "epoch": 2.0, + "learning_rate": 1.5531177629552835e-05, + "loss": 1.1481, + "step": 27123 + }, + { + "epoch": 2.0, + "learning_rate": 1.5530845323978132e-05, + "loss": 0.8663, + "step": 27124 + }, + { + "epoch": 2.0, + "learning_rate": 1.5530513009603933e-05, + "loss": 1.0312, + "step": 27125 + }, + { + "epoch": 2.0, + "learning_rate": 1.553018068643077e-05, + "loss": 0.9408, + "step": 27126 + }, + { + "epoch": 2.0, + "learning_rate": 1.5529848354459165e-05, + "loss": 1.0249, + "step": 27127 + }, + { + "epoch": 2.0, + "learning_rate": 1.5529516013689653e-05, + "loss": 1.0368, + "step": 27128 + }, + { + "epoch": 2.0, + "learning_rate": 1.552918366412276e-05, + "loss": 0.9256, + "step": 27129 + }, + { + "epoch": 2.0, + "learning_rate": 1.5528851305759008e-05, + "loss": 1.0391, + "step": 27130 + }, + { + "epoch": 2.0, + "learning_rate": 1.5528518938598938e-05, + "loss": 1.0588, + "step": 27131 + }, + { + "epoch": 2.0, + "learning_rate": 1.5528186562643075e-05, + "loss": 1.0929, + "step": 27132 + }, + { + "epoch": 2.0, + "learning_rate": 1.5527854177891947e-05, + "loss": 1.0298, + "step": 27133 + }, + { + "epoch": 2.0, + "learning_rate": 1.552752178434608e-05, + "loss": 1.0974, + "step": 27134 + }, + { + "epoch": 2.0, + "learning_rate": 1.5527189382006004e-05, + "loss": 1.0556, + "step": 27135 + }, + { + "epoch": 2.01, + "learning_rate": 1.552685697087225e-05, + "loss": 1.0299, + "step": 27136 + }, + { + "epoch": 2.01, + "learning_rate": 1.5526524550945347e-05, + "loss": 1.0566, + "step": 27137 + }, + { + "epoch": 2.01, + "learning_rate": 1.552619212222582e-05, + "loss": 0.9144, + "step": 27138 + }, + { + "epoch": 2.01, + "learning_rate": 1.5525859684714202e-05, + "loss": 1.0154, + "step": 27139 + }, + { + "epoch": 2.01, + "learning_rate": 1.552552723841102e-05, + "loss": 1.0267, + "step": 27140 + }, + { + "epoch": 2.01, + "learning_rate": 1.55251947833168e-05, + "loss": 0.9293, + "step": 27141 + }, + { + "epoch": 2.01, + "learning_rate": 1.552486231943208e-05, + "loss": 1.1594, + "step": 27142 + }, + { + "epoch": 2.01, + "learning_rate": 1.5524529846757376e-05, + "loss": 1.0623, + "step": 27143 + }, + { + "epoch": 2.01, + "learning_rate": 1.552419736529323e-05, + "loss": 0.9972, + "step": 27144 + }, + { + "epoch": 2.01, + "learning_rate": 1.5523864875040163e-05, + "loss": 0.9851, + "step": 27145 + }, + { + "epoch": 2.01, + "learning_rate": 1.5523532375998707e-05, + "loss": 0.9537, + "step": 27146 + }, + { + "epoch": 2.01, + "learning_rate": 1.5523199868169392e-05, + "loss": 0.9474, + "step": 27147 + }, + { + "epoch": 2.01, + "learning_rate": 1.552286735155274e-05, + "loss": 1.0683, + "step": 27148 + }, + { + "epoch": 2.01, + "learning_rate": 1.5522534826149288e-05, + "loss": 1.0712, + "step": 27149 + }, + { + "epoch": 2.01, + "learning_rate": 1.552220229195956e-05, + "loss": 0.9696, + "step": 27150 + }, + { + "epoch": 2.01, + "learning_rate": 1.552186974898409e-05, + "loss": 1.0192, + "step": 27151 + }, + { + "epoch": 2.01, + "learning_rate": 1.55215371972234e-05, + "loss": 1.0118, + "step": 27152 + }, + { + "epoch": 2.01, + "learning_rate": 1.5521204636678028e-05, + "loss": 1.0278, + "step": 27153 + }, + { + "epoch": 2.01, + "learning_rate": 1.5520872067348497e-05, + "loss": 0.9911, + "step": 27154 + }, + { + "epoch": 2.01, + "learning_rate": 1.5520539489235336e-05, + "loss": 0.9241, + "step": 27155 + }, + { + "epoch": 2.01, + "learning_rate": 1.552020690233908e-05, + "loss": 0.9408, + "step": 27156 + }, + { + "epoch": 2.01, + "learning_rate": 1.551987430666025e-05, + "loss": 1.0372, + "step": 27157 + }, + { + "epoch": 2.01, + "learning_rate": 1.5519541702199378e-05, + "loss": 0.9171, + "step": 27158 + }, + { + "epoch": 2.01, + "learning_rate": 1.5519209088956996e-05, + "loss": 0.9489, + "step": 27159 + }, + { + "epoch": 2.01, + "learning_rate": 1.5518876466933633e-05, + "loss": 1.0058, + "step": 27160 + }, + { + "epoch": 2.01, + "learning_rate": 1.5518543836129812e-05, + "loss": 0.9052, + "step": 27161 + }, + { + "epoch": 2.01, + "learning_rate": 1.551821119654607e-05, + "loss": 1.1175, + "step": 27162 + }, + { + "epoch": 2.01, + "learning_rate": 1.551787854818293e-05, + "loss": 1.0094, + "step": 27163 + }, + { + "epoch": 2.01, + "learning_rate": 1.551754589104093e-05, + "loss": 0.9417, + "step": 27164 + }, + { + "epoch": 2.01, + "learning_rate": 1.5517213225120588e-05, + "loss": 1.0298, + "step": 27165 + }, + { + "epoch": 2.01, + "learning_rate": 1.5516880550422442e-05, + "loss": 1.0261, + "step": 27166 + }, + { + "epoch": 2.01, + "learning_rate": 1.5516547866947015e-05, + "loss": 0.9858, + "step": 27167 + }, + { + "epoch": 2.01, + "learning_rate": 1.5516215174694843e-05, + "loss": 1.0017, + "step": 27168 + }, + { + "epoch": 2.01, + "learning_rate": 1.5515882473666447e-05, + "loss": 1.0713, + "step": 27169 + }, + { + "epoch": 2.01, + "learning_rate": 1.5515549763862364e-05, + "loss": 1.0121, + "step": 27170 + }, + { + "epoch": 2.01, + "learning_rate": 1.551521704528312e-05, + "loss": 1.0005, + "step": 27171 + }, + { + "epoch": 2.01, + "learning_rate": 1.5514884317929242e-05, + "loss": 1.116, + "step": 27172 + }, + { + "epoch": 2.01, + "learning_rate": 1.5514551581801264e-05, + "loss": 1.0022, + "step": 27173 + }, + { + "epoch": 2.01, + "learning_rate": 1.5514218836899716e-05, + "loss": 1.139, + "step": 27174 + }, + { + "epoch": 2.01, + "learning_rate": 1.551388608322512e-05, + "loss": 1.0125, + "step": 27175 + }, + { + "epoch": 2.01, + "learning_rate": 1.5513553320778013e-05, + "loss": 1.0924, + "step": 27176 + }, + { + "epoch": 2.01, + "learning_rate": 1.5513220549558917e-05, + "loss": 0.9927, + "step": 27177 + }, + { + "epoch": 2.01, + "learning_rate": 1.5512887769568372e-05, + "loss": 0.9549, + "step": 27178 + }, + { + "epoch": 2.01, + "learning_rate": 1.55125549808069e-05, + "loss": 1.0061, + "step": 27179 + }, + { + "epoch": 2.01, + "learning_rate": 1.551222218327503e-05, + "loss": 0.9706, + "step": 27180 + }, + { + "epoch": 2.01, + "learning_rate": 1.5511889376973293e-05, + "loss": 1.0518, + "step": 27181 + }, + { + "epoch": 2.01, + "learning_rate": 1.551155656190222e-05, + "loss": 1.007, + "step": 27182 + }, + { + "epoch": 2.01, + "learning_rate": 1.551122373806234e-05, + "loss": 1.1041, + "step": 27183 + }, + { + "epoch": 2.01, + "learning_rate": 1.5510890905454185e-05, + "loss": 0.9697, + "step": 27184 + }, + { + "epoch": 2.01, + "learning_rate": 1.5510558064078275e-05, + "loss": 1.0379, + "step": 27185 + }, + { + "epoch": 2.01, + "learning_rate": 1.551022521393515e-05, + "loss": 1.0484, + "step": 27186 + }, + { + "epoch": 2.01, + "learning_rate": 1.5509892355025335e-05, + "loss": 1.1335, + "step": 27187 + }, + { + "epoch": 2.01, + "learning_rate": 1.550955948734936e-05, + "loss": 1.0247, + "step": 27188 + }, + { + "epoch": 2.01, + "learning_rate": 1.5509226610907756e-05, + "loss": 1.0571, + "step": 27189 + }, + { + "epoch": 2.01, + "learning_rate": 1.5508893725701045e-05, + "loss": 1.0601, + "step": 27190 + }, + { + "epoch": 2.01, + "learning_rate": 1.550856083172977e-05, + "loss": 1.0373, + "step": 27191 + }, + { + "epoch": 2.01, + "learning_rate": 1.550822792899445e-05, + "loss": 1.052, + "step": 27192 + }, + { + "epoch": 2.01, + "learning_rate": 1.550789501749562e-05, + "loss": 1.0448, + "step": 27193 + }, + { + "epoch": 2.01, + "learning_rate": 1.550756209723381e-05, + "loss": 1.0065, + "step": 27194 + }, + { + "epoch": 2.01, + "learning_rate": 1.5507229168209544e-05, + "loss": 0.8223, + "step": 27195 + }, + { + "epoch": 2.01, + "learning_rate": 1.5506896230423354e-05, + "loss": 0.9968, + "step": 27196 + }, + { + "epoch": 2.01, + "learning_rate": 1.5506563283875774e-05, + "loss": 0.9536, + "step": 27197 + }, + { + "epoch": 2.01, + "learning_rate": 1.5506230328567325e-05, + "loss": 0.9931, + "step": 27198 + }, + { + "epoch": 2.01, + "learning_rate": 1.5505897364498547e-05, + "loss": 1.0032, + "step": 27199 + }, + { + "epoch": 2.01, + "learning_rate": 1.5505564391669964e-05, + "loss": 0.9477, + "step": 27200 + }, + { + "epoch": 2.01, + "learning_rate": 1.5505231410082108e-05, + "loss": 0.9819, + "step": 27201 + }, + { + "epoch": 2.01, + "learning_rate": 1.5504898419735507e-05, + "loss": 0.9694, + "step": 27202 + }, + { + "epoch": 2.01, + "learning_rate": 1.550456542063069e-05, + "loss": 1.0808, + "step": 27203 + }, + { + "epoch": 2.01, + "learning_rate": 1.5504232412768192e-05, + "loss": 1.0679, + "step": 27204 + }, + { + "epoch": 2.01, + "learning_rate": 1.5503899396148537e-05, + "loss": 0.9787, + "step": 27205 + }, + { + "epoch": 2.01, + "learning_rate": 1.5503566370772256e-05, + "loss": 0.929, + "step": 27206 + }, + { + "epoch": 2.01, + "learning_rate": 1.5503233336639877e-05, + "loss": 0.9231, + "step": 27207 + }, + { + "epoch": 2.01, + "learning_rate": 1.5502900293751938e-05, + "loss": 0.9922, + "step": 27208 + }, + { + "epoch": 2.01, + "learning_rate": 1.5502567242108958e-05, + "loss": 0.9798, + "step": 27209 + }, + { + "epoch": 2.01, + "learning_rate": 1.5502234181711476e-05, + "loss": 1.0199, + "step": 27210 + }, + { + "epoch": 2.01, + "learning_rate": 1.5501901112560017e-05, + "loss": 1.0288, + "step": 27211 + }, + { + "epoch": 2.01, + "learning_rate": 1.5501568034655113e-05, + "loss": 1.1368, + "step": 27212 + }, + { + "epoch": 2.01, + "learning_rate": 1.5501234947997295e-05, + "loss": 1.0898, + "step": 27213 + }, + { + "epoch": 2.01, + "learning_rate": 1.5500901852587083e-05, + "loss": 0.9319, + "step": 27214 + }, + { + "epoch": 2.01, + "learning_rate": 1.550056874842502e-05, + "loss": 1.0421, + "step": 27215 + }, + { + "epoch": 2.01, + "learning_rate": 1.5500235635511636e-05, + "loss": 1.1525, + "step": 27216 + }, + { + "epoch": 2.01, + "learning_rate": 1.5499902513847448e-05, + "loss": 1.058, + "step": 27217 + }, + { + "epoch": 2.01, + "learning_rate": 1.5499569383432996e-05, + "loss": 1.0446, + "step": 27218 + }, + { + "epoch": 2.01, + "learning_rate": 1.549923624426881e-05, + "loss": 1.0436, + "step": 27219 + }, + { + "epoch": 2.01, + "learning_rate": 1.5498903096355415e-05, + "loss": 0.9661, + "step": 27220 + }, + { + "epoch": 2.01, + "learning_rate": 1.5498569939693347e-05, + "loss": 1.1413, + "step": 27221 + }, + { + "epoch": 2.01, + "learning_rate": 1.549823677428313e-05, + "loss": 0.9502, + "step": 27222 + }, + { + "epoch": 2.01, + "learning_rate": 1.5497903600125297e-05, + "loss": 0.9229, + "step": 27223 + }, + { + "epoch": 2.01, + "learning_rate": 1.549757041722038e-05, + "loss": 0.9226, + "step": 27224 + }, + { + "epoch": 2.01, + "learning_rate": 1.5497237225568905e-05, + "loss": 1.1637, + "step": 27225 + }, + { + "epoch": 2.01, + "learning_rate": 1.5496904025171405e-05, + "loss": 1.0624, + "step": 27226 + }, + { + "epoch": 2.01, + "learning_rate": 1.549657081602841e-05, + "loss": 1.0534, + "step": 27227 + }, + { + "epoch": 2.01, + "learning_rate": 1.5496237598140448e-05, + "loss": 0.8315, + "step": 27228 + }, + { + "epoch": 2.01, + "learning_rate": 1.549590437150805e-05, + "loss": 0.9404, + "step": 27229 + }, + { + "epoch": 2.01, + "learning_rate": 1.549557113613175e-05, + "loss": 0.9971, + "step": 27230 + }, + { + "epoch": 2.01, + "learning_rate": 1.5495237892012075e-05, + "loss": 1.0292, + "step": 27231 + }, + { + "epoch": 2.01, + "learning_rate": 1.549490463914955e-05, + "loss": 1.0244, + "step": 27232 + }, + { + "epoch": 2.01, + "learning_rate": 1.5494571377544712e-05, + "loss": 1.0738, + "step": 27233 + }, + { + "epoch": 2.01, + "learning_rate": 1.5494238107198093e-05, + "loss": 0.9195, + "step": 27234 + }, + { + "epoch": 2.01, + "learning_rate": 1.549390482811022e-05, + "loss": 0.9953, + "step": 27235 + }, + { + "epoch": 2.01, + "learning_rate": 1.549357154028162e-05, + "loss": 1.0322, + "step": 27236 + }, + { + "epoch": 2.01, + "learning_rate": 1.5493238243712826e-05, + "loss": 0.9594, + "step": 27237 + }, + { + "epoch": 2.01, + "learning_rate": 1.5492904938404373e-05, + "loss": 0.9951, + "step": 27238 + }, + { + "epoch": 2.01, + "learning_rate": 1.5492571624356783e-05, + "loss": 0.91, + "step": 27239 + }, + { + "epoch": 2.01, + "learning_rate": 1.549223830157059e-05, + "loss": 1.035, + "step": 27240 + }, + { + "epoch": 2.01, + "learning_rate": 1.5491904970046327e-05, + "loss": 1.0388, + "step": 27241 + }, + { + "epoch": 2.01, + "learning_rate": 1.5491571629784517e-05, + "loss": 0.9397, + "step": 27242 + }, + { + "epoch": 2.01, + "learning_rate": 1.5491238280785702e-05, + "loss": 1.0239, + "step": 27243 + }, + { + "epoch": 2.01, + "learning_rate": 1.54909049230504e-05, + "loss": 1.0171, + "step": 27244 + }, + { + "epoch": 2.01, + "learning_rate": 1.5490571556579147e-05, + "loss": 0.9488, + "step": 27245 + }, + { + "epoch": 2.01, + "learning_rate": 1.5490238181372475e-05, + "loss": 0.9798, + "step": 27246 + }, + { + "epoch": 2.01, + "learning_rate": 1.548990479743091e-05, + "loss": 1.0257, + "step": 27247 + }, + { + "epoch": 2.01, + "learning_rate": 1.548957140475499e-05, + "loss": 0.928, + "step": 27248 + }, + { + "epoch": 2.01, + "learning_rate": 1.548923800334524e-05, + "loss": 0.9792, + "step": 27249 + }, + { + "epoch": 2.01, + "learning_rate": 1.548890459320219e-05, + "loss": 1.0653, + "step": 27250 + }, + { + "epoch": 2.01, + "learning_rate": 1.5488571174326367e-05, + "loss": 1.0443, + "step": 27251 + }, + { + "epoch": 2.01, + "learning_rate": 1.548823774671831e-05, + "loss": 0.9929, + "step": 27252 + }, + { + "epoch": 2.01, + "learning_rate": 1.5487904310378543e-05, + "loss": 0.9578, + "step": 27253 + }, + { + "epoch": 2.01, + "learning_rate": 1.54875708653076e-05, + "loss": 1.0349, + "step": 27254 + }, + { + "epoch": 2.01, + "learning_rate": 1.5487237411506008e-05, + "loss": 0.977, + "step": 27255 + }, + { + "epoch": 2.01, + "learning_rate": 1.5486903948974305e-05, + "loss": 0.9641, + "step": 27256 + }, + { + "epoch": 2.01, + "learning_rate": 1.5486570477713012e-05, + "loss": 1.0402, + "step": 27257 + }, + { + "epoch": 2.01, + "learning_rate": 1.5486236997722666e-05, + "loss": 0.9535, + "step": 27258 + }, + { + "epoch": 2.01, + "learning_rate": 1.5485903509003794e-05, + "loss": 1.0054, + "step": 27259 + }, + { + "epoch": 2.01, + "learning_rate": 1.548557001155693e-05, + "loss": 0.9975, + "step": 27260 + }, + { + "epoch": 2.01, + "learning_rate": 1.5485236505382598e-05, + "loss": 1.0054, + "step": 27261 + }, + { + "epoch": 2.01, + "learning_rate": 1.5484902990481335e-05, + "loss": 0.9928, + "step": 27262 + }, + { + "epoch": 2.01, + "learning_rate": 1.5484569466853674e-05, + "loss": 1.0509, + "step": 27263 + }, + { + "epoch": 2.01, + "learning_rate": 1.5484235934500137e-05, + "loss": 0.8481, + "step": 27264 + }, + { + "epoch": 2.01, + "learning_rate": 1.5483902393421262e-05, + "loss": 0.9897, + "step": 27265 + }, + { + "epoch": 2.01, + "learning_rate": 1.5483568843617573e-05, + "loss": 1.0016, + "step": 27266 + }, + { + "epoch": 2.01, + "learning_rate": 1.548323528508961e-05, + "loss": 1.0248, + "step": 27267 + }, + { + "epoch": 2.01, + "learning_rate": 1.548290171783789e-05, + "loss": 0.9345, + "step": 27268 + }, + { + "epoch": 2.01, + "learning_rate": 1.548256814186296e-05, + "loss": 1.1094, + "step": 27269 + }, + { + "epoch": 2.01, + "learning_rate": 1.5482234557165342e-05, + "loss": 0.9566, + "step": 27270 + }, + { + "epoch": 2.02, + "learning_rate": 1.548190096374556e-05, + "loss": 0.9979, + "step": 27271 + }, + { + "epoch": 2.02, + "learning_rate": 1.548156736160416e-05, + "loss": 1.0299, + "step": 27272 + }, + { + "epoch": 2.02, + "learning_rate": 1.548123375074166e-05, + "loss": 0.9761, + "step": 27273 + }, + { + "epoch": 2.02, + "learning_rate": 1.5480900131158596e-05, + "loss": 0.909, + "step": 27274 + }, + { + "epoch": 2.02, + "learning_rate": 1.5480566502855498e-05, + "loss": 1.0013, + "step": 27275 + }, + { + "epoch": 2.02, + "learning_rate": 1.5480232865832897e-05, + "loss": 0.9842, + "step": 27276 + }, + { + "epoch": 2.02, + "learning_rate": 1.547989922009133e-05, + "loss": 1.0013, + "step": 27277 + }, + { + "epoch": 2.02, + "learning_rate": 1.5479565565631313e-05, + "loss": 0.988, + "step": 27278 + }, + { + "epoch": 2.02, + "learning_rate": 1.5479231902453392e-05, + "loss": 1.0187, + "step": 27279 + }, + { + "epoch": 2.02, + "learning_rate": 1.547889823055809e-05, + "loss": 1.0799, + "step": 27280 + }, + { + "epoch": 2.02, + "learning_rate": 1.5478564549945935e-05, + "loss": 0.9966, + "step": 27281 + }, + { + "epoch": 2.02, + "learning_rate": 1.5478230860617468e-05, + "loss": 0.9168, + "step": 27282 + }, + { + "epoch": 2.02, + "learning_rate": 1.547789716257321e-05, + "loss": 1.0247, + "step": 27283 + }, + { + "epoch": 2.02, + "learning_rate": 1.54775634558137e-05, + "loss": 1.0201, + "step": 27284 + }, + { + "epoch": 2.02, + "learning_rate": 1.547722974033946e-05, + "loss": 1.0898, + "step": 27285 + }, + { + "epoch": 2.02, + "learning_rate": 1.5476896016151028e-05, + "loss": 1.0484, + "step": 27286 + }, + { + "epoch": 2.02, + "learning_rate": 1.5476562283248935e-05, + "loss": 1.0964, + "step": 27287 + }, + { + "epoch": 2.02, + "learning_rate": 1.5476228541633706e-05, + "loss": 0.9497, + "step": 27288 + }, + { + "epoch": 2.02, + "learning_rate": 1.547589479130588e-05, + "loss": 1.0299, + "step": 27289 + }, + { + "epoch": 2.02, + "learning_rate": 1.547556103226598e-05, + "loss": 1.0338, + "step": 27290 + }, + { + "epoch": 2.02, + "learning_rate": 1.547522726451454e-05, + "loss": 1.0181, + "step": 27291 + }, + { + "epoch": 2.02, + "learning_rate": 1.5474893488052094e-05, + "loss": 0.97, + "step": 27292 + }, + { + "epoch": 2.02, + "learning_rate": 1.547455970287917e-05, + "loss": 0.9701, + "step": 27293 + }, + { + "epoch": 2.02, + "learning_rate": 1.5474225908996304e-05, + "loss": 0.899, + "step": 27294 + }, + { + "epoch": 2.02, + "learning_rate": 1.5473892106404017e-05, + "loss": 1.0046, + "step": 27295 + }, + { + "epoch": 2.02, + "learning_rate": 1.5473558295102852e-05, + "loss": 0.9678, + "step": 27296 + }, + { + "epoch": 2.02, + "learning_rate": 1.5473224475093328e-05, + "loss": 1.0216, + "step": 27297 + }, + { + "epoch": 2.02, + "learning_rate": 1.5472890646375983e-05, + "loss": 1.0393, + "step": 27298 + }, + { + "epoch": 2.02, + "learning_rate": 1.547255680895135e-05, + "loss": 1.0614, + "step": 27299 + }, + { + "epoch": 2.02, + "learning_rate": 1.5472222962819957e-05, + "loss": 1.0326, + "step": 27300 + }, + { + "epoch": 2.02, + "learning_rate": 1.547188910798233e-05, + "loss": 0.9699, + "step": 27301 + }, + { + "epoch": 2.02, + "learning_rate": 1.5471555244439014e-05, + "loss": 0.9317, + "step": 27302 + }, + { + "epoch": 2.02, + "learning_rate": 1.547122137219053e-05, + "loss": 0.931, + "step": 27303 + }, + { + "epoch": 2.02, + "learning_rate": 1.5470887491237407e-05, + "loss": 1.0395, + "step": 27304 + }, + { + "epoch": 2.02, + "learning_rate": 1.5470553601580182e-05, + "loss": 0.9554, + "step": 27305 + }, + { + "epoch": 2.02, + "learning_rate": 1.547021970321939e-05, + "loss": 1.0135, + "step": 27306 + }, + { + "epoch": 2.02, + "learning_rate": 1.546988579615555e-05, + "loss": 0.8403, + "step": 27307 + }, + { + "epoch": 2.02, + "learning_rate": 1.54695518803892e-05, + "loss": 1.0666, + "step": 27308 + }, + { + "epoch": 2.02, + "learning_rate": 1.5469217955920873e-05, + "loss": 0.9999, + "step": 27309 + }, + { + "epoch": 2.02, + "learning_rate": 1.5468884022751098e-05, + "loss": 0.8872, + "step": 27310 + }, + { + "epoch": 2.02, + "learning_rate": 1.5468550080880408e-05, + "loss": 1.0285, + "step": 27311 + }, + { + "epoch": 2.02, + "learning_rate": 1.5468216130309333e-05, + "loss": 0.9305, + "step": 27312 + }, + { + "epoch": 2.02, + "learning_rate": 1.5467882171038404e-05, + "loss": 1.0338, + "step": 27313 + }, + { + "epoch": 2.02, + "learning_rate": 1.546754820306815e-05, + "loss": 0.9882, + "step": 27314 + }, + { + "epoch": 2.02, + "learning_rate": 1.546721422639911e-05, + "loss": 0.938, + "step": 27315 + }, + { + "epoch": 2.02, + "learning_rate": 1.546688024103181e-05, + "loss": 1.0186, + "step": 27316 + }, + { + "epoch": 2.02, + "learning_rate": 1.546654624696678e-05, + "loss": 1.003, + "step": 27317 + }, + { + "epoch": 2.02, + "learning_rate": 1.546621224420455e-05, + "loss": 0.9311, + "step": 27318 + }, + { + "epoch": 2.02, + "learning_rate": 1.5465878232745657e-05, + "loss": 1.0913, + "step": 27319 + }, + { + "epoch": 2.02, + "learning_rate": 1.546554421259063e-05, + "loss": 0.9315, + "step": 27320 + }, + { + "epoch": 2.02, + "learning_rate": 1.5465210183740003e-05, + "loss": 1.0309, + "step": 27321 + }, + { + "epoch": 2.02, + "learning_rate": 1.54648761461943e-05, + "loss": 1.0248, + "step": 27322 + }, + { + "epoch": 2.02, + "learning_rate": 1.546454209995406e-05, + "loss": 0.9442, + "step": 27323 + }, + { + "epoch": 2.02, + "learning_rate": 1.5464208045019813e-05, + "loss": 1.0039, + "step": 27324 + }, + { + "epoch": 2.02, + "learning_rate": 1.5463873981392087e-05, + "loss": 1.0893, + "step": 27325 + }, + { + "epoch": 2.02, + "learning_rate": 1.5463539909071418e-05, + "loss": 0.9942, + "step": 27326 + }, + { + "epoch": 2.02, + "learning_rate": 1.5463205828058333e-05, + "loss": 1.0345, + "step": 27327 + }, + { + "epoch": 2.02, + "learning_rate": 1.5462871738353363e-05, + "loss": 1.1008, + "step": 27328 + }, + { + "epoch": 2.02, + "learning_rate": 1.546253763995705e-05, + "loss": 0.9847, + "step": 27329 + }, + { + "epoch": 2.02, + "learning_rate": 1.5462203532869912e-05, + "loss": 0.9351, + "step": 27330 + }, + { + "epoch": 2.02, + "learning_rate": 1.5461869417092485e-05, + "loss": 1.0309, + "step": 27331 + }, + { + "epoch": 2.02, + "learning_rate": 1.5461535292625306e-05, + "loss": 1.0438, + "step": 27332 + }, + { + "epoch": 2.02, + "learning_rate": 1.54612011594689e-05, + "loss": 1.038, + "step": 27333 + }, + { + "epoch": 2.02, + "learning_rate": 1.54608670176238e-05, + "loss": 1.0584, + "step": 27334 + }, + { + "epoch": 2.02, + "learning_rate": 1.546053286709054e-05, + "loss": 1.0213, + "step": 27335 + }, + { + "epoch": 2.02, + "learning_rate": 1.5460198707869655e-05, + "loss": 1.0014, + "step": 27336 + }, + { + "epoch": 2.02, + "learning_rate": 1.5459864539961666e-05, + "loss": 1.0197, + "step": 27337 + }, + { + "epoch": 2.02, + "learning_rate": 1.5459530363367114e-05, + "loss": 1.0888, + "step": 27338 + }, + { + "epoch": 2.02, + "learning_rate": 1.5459196178086524e-05, + "loss": 0.9646, + "step": 27339 + }, + { + "epoch": 2.02, + "learning_rate": 1.545886198412043e-05, + "loss": 1.0686, + "step": 27340 + }, + { + "epoch": 2.02, + "learning_rate": 1.5458527781469373e-05, + "loss": 1.1266, + "step": 27341 + }, + { + "epoch": 2.02, + "learning_rate": 1.5458193570133867e-05, + "loss": 1.0822, + "step": 27342 + }, + { + "epoch": 2.02, + "learning_rate": 1.5457859350114463e-05, + "loss": 1.0821, + "step": 27343 + }, + { + "epoch": 2.02, + "learning_rate": 1.5457525121411674e-05, + "loss": 0.9508, + "step": 27344 + }, + { + "epoch": 2.02, + "learning_rate": 1.5457190884026045e-05, + "loss": 1.0128, + "step": 27345 + }, + { + "epoch": 2.02, + "learning_rate": 1.54568566379581e-05, + "loss": 1.0448, + "step": 27346 + }, + { + "epoch": 2.02, + "learning_rate": 1.5456522383208378e-05, + "loss": 1.0621, + "step": 27347 + }, + { + "epoch": 2.02, + "learning_rate": 1.5456188119777404e-05, + "loss": 0.9136, + "step": 27348 + }, + { + "epoch": 2.02, + "learning_rate": 1.545585384766571e-05, + "loss": 1.0301, + "step": 27349 + }, + { + "epoch": 2.02, + "learning_rate": 1.545551956687384e-05, + "loss": 1.1516, + "step": 27350 + }, + { + "epoch": 2.02, + "learning_rate": 1.5455185277402307e-05, + "loss": 0.9912, + "step": 27351 + }, + { + "epoch": 2.02, + "learning_rate": 1.545485097925166e-05, + "loss": 1.0393, + "step": 27352 + }, + { + "epoch": 2.02, + "learning_rate": 1.545451667242242e-05, + "loss": 1.0043, + "step": 27353 + }, + { + "epoch": 2.02, + "learning_rate": 1.545418235691512e-05, + "loss": 1.0274, + "step": 27354 + }, + { + "epoch": 2.02, + "learning_rate": 1.5453848032730294e-05, + "loss": 1.0118, + "step": 27355 + }, + { + "epoch": 2.02, + "learning_rate": 1.5453513699868477e-05, + "loss": 1.0418, + "step": 27356 + }, + { + "epoch": 2.02, + "learning_rate": 1.5453179358330196e-05, + "loss": 0.9848, + "step": 27357 + }, + { + "epoch": 2.02, + "learning_rate": 1.545284500811598e-05, + "loss": 0.9939, + "step": 27358 + }, + { + "epoch": 2.02, + "learning_rate": 1.5452510649226372e-05, + "loss": 1.1044, + "step": 27359 + }, + { + "epoch": 2.02, + "learning_rate": 1.5452176281661896e-05, + "loss": 1.0743, + "step": 27360 + }, + { + "epoch": 2.02, + "learning_rate": 1.5451841905423084e-05, + "loss": 1.0537, + "step": 27361 + }, + { + "epoch": 2.02, + "learning_rate": 1.5451507520510472e-05, + "loss": 0.9914, + "step": 27362 + }, + { + "epoch": 2.02, + "learning_rate": 1.5451173126924585e-05, + "loss": 0.9911, + "step": 27363 + }, + { + "epoch": 2.02, + "learning_rate": 1.5450838724665965e-05, + "loss": 1.0619, + "step": 27364 + }, + { + "epoch": 2.02, + "learning_rate": 1.5450504313735137e-05, + "loss": 1.0849, + "step": 27365 + }, + { + "epoch": 2.02, + "learning_rate": 1.5450169894132633e-05, + "loss": 1.0562, + "step": 27366 + }, + { + "epoch": 2.02, + "learning_rate": 1.5449835465858988e-05, + "loss": 0.9446, + "step": 27367 + }, + { + "epoch": 2.02, + "learning_rate": 1.5449501028914735e-05, + "loss": 1.0594, + "step": 27368 + }, + { + "epoch": 2.02, + "learning_rate": 1.5449166583300398e-05, + "loss": 1.1153, + "step": 27369 + }, + { + "epoch": 2.02, + "learning_rate": 1.544883212901652e-05, + "loss": 0.9465, + "step": 27370 + }, + { + "epoch": 2.02, + "learning_rate": 1.5448497666063628e-05, + "loss": 1.0062, + "step": 27371 + }, + { + "epoch": 2.02, + "learning_rate": 1.544816319444225e-05, + "loss": 0.9826, + "step": 27372 + }, + { + "epoch": 2.02, + "learning_rate": 1.5447828714152924e-05, + "loss": 1.0506, + "step": 27373 + }, + { + "epoch": 2.02, + "learning_rate": 1.5447494225196182e-05, + "loss": 1.0346, + "step": 27374 + }, + { + "epoch": 2.02, + "learning_rate": 1.5447159727572555e-05, + "loss": 0.9981, + "step": 27375 + }, + { + "epoch": 2.02, + "learning_rate": 1.544682522128257e-05, + "loss": 1.0591, + "step": 27376 + }, + { + "epoch": 2.02, + "learning_rate": 1.544649070632677e-05, + "loss": 0.924, + "step": 27377 + }, + { + "epoch": 2.02, + "learning_rate": 1.5446156182705675e-05, + "loss": 0.9261, + "step": 27378 + }, + { + "epoch": 2.02, + "learning_rate": 1.544582165041983e-05, + "loss": 1.0124, + "step": 27379 + }, + { + "epoch": 2.02, + "learning_rate": 1.544548710946976e-05, + "loss": 1.1471, + "step": 27380 + }, + { + "epoch": 2.02, + "learning_rate": 1.5445152559855992e-05, + "loss": 1.0152, + "step": 27381 + }, + { + "epoch": 2.02, + "learning_rate": 1.544481800157907e-05, + "loss": 1.0754, + "step": 27382 + }, + { + "epoch": 2.02, + "learning_rate": 1.5444483434639518e-05, + "loss": 1.0401, + "step": 27383 + }, + { + "epoch": 2.02, + "learning_rate": 1.5444148859037872e-05, + "loss": 1.1155, + "step": 27384 + }, + { + "epoch": 2.02, + "learning_rate": 1.544381427477466e-05, + "loss": 0.9658, + "step": 27385 + }, + { + "epoch": 2.02, + "learning_rate": 1.544347968185042e-05, + "loss": 0.9784, + "step": 27386 + }, + { + "epoch": 2.02, + "learning_rate": 1.544314508026568e-05, + "loss": 1.0044, + "step": 27387 + }, + { + "epoch": 2.02, + "learning_rate": 1.5442810470020976e-05, + "loss": 1.0665, + "step": 27388 + }, + { + "epoch": 2.02, + "learning_rate": 1.5442475851116837e-05, + "loss": 1.0424, + "step": 27389 + }, + { + "epoch": 2.02, + "learning_rate": 1.54421412235538e-05, + "loss": 1.1354, + "step": 27390 + }, + { + "epoch": 2.02, + "learning_rate": 1.544180658733239e-05, + "loss": 1.0066, + "step": 27391 + }, + { + "epoch": 2.02, + "learning_rate": 1.544147194245315e-05, + "loss": 1.0327, + "step": 27392 + }, + { + "epoch": 2.02, + "learning_rate": 1.5441137288916597e-05, + "loss": 1.066, + "step": 27393 + }, + { + "epoch": 2.02, + "learning_rate": 1.5440802626723278e-05, + "loss": 1.0247, + "step": 27394 + }, + { + "epoch": 2.02, + "learning_rate": 1.544046795587372e-05, + "loss": 1.0668, + "step": 27395 + }, + { + "epoch": 2.02, + "learning_rate": 1.544013327636845e-05, + "loss": 1.0768, + "step": 27396 + }, + { + "epoch": 2.02, + "learning_rate": 1.543979858820801e-05, + "loss": 1.0615, + "step": 27397 + }, + { + "epoch": 2.02, + "learning_rate": 1.543946389139293e-05, + "loss": 1.1073, + "step": 27398 + }, + { + "epoch": 2.02, + "learning_rate": 1.543912918592374e-05, + "loss": 1.0907, + "step": 27399 + }, + { + "epoch": 2.02, + "learning_rate": 1.5438794471800972e-05, + "loss": 1.0373, + "step": 27400 + }, + { + "epoch": 2.02, + "learning_rate": 1.543845974902516e-05, + "loss": 1.0624, + "step": 27401 + }, + { + "epoch": 2.02, + "learning_rate": 1.5438125017596837e-05, + "loss": 0.9941, + "step": 27402 + }, + { + "epoch": 2.02, + "learning_rate": 1.5437790277516534e-05, + "loss": 1.0531, + "step": 27403 + }, + { + "epoch": 2.02, + "learning_rate": 1.5437455528784782e-05, + "loss": 0.9843, + "step": 27404 + }, + { + "epoch": 2.02, + "learning_rate": 1.5437120771402117e-05, + "loss": 1.0032, + "step": 27405 + }, + { + "epoch": 2.03, + "learning_rate": 1.5436786005369077e-05, + "loss": 1.042, + "step": 27406 + }, + { + "epoch": 2.03, + "learning_rate": 1.5436451230686184e-05, + "loss": 0.9256, + "step": 27407 + }, + { + "epoch": 2.03, + "learning_rate": 1.5436116447353973e-05, + "loss": 1.0676, + "step": 27408 + }, + { + "epoch": 2.03, + "learning_rate": 1.543578165537298e-05, + "loss": 1.081, + "step": 27409 + }, + { + "epoch": 2.03, + "learning_rate": 1.5435446854743736e-05, + "loss": 1.0752, + "step": 27410 + }, + { + "epoch": 2.03, + "learning_rate": 1.5435112045466774e-05, + "loss": 0.9511, + "step": 27411 + }, + { + "epoch": 2.03, + "learning_rate": 1.5434777227542625e-05, + "loss": 0.9942, + "step": 27412 + }, + { + "epoch": 2.03, + "learning_rate": 1.5434442400971825e-05, + "loss": 1.0714, + "step": 27413 + }, + { + "epoch": 2.03, + "learning_rate": 1.5434107565754906e-05, + "loss": 0.8869, + "step": 27414 + }, + { + "epoch": 2.03, + "learning_rate": 1.5433772721892396e-05, + "loss": 1.0946, + "step": 27415 + }, + { + "epoch": 2.03, + "learning_rate": 1.5433437869384834e-05, + "loss": 0.9891, + "step": 27416 + }, + { + "epoch": 2.03, + "learning_rate": 1.5433103008232748e-05, + "loss": 1.057, + "step": 27417 + }, + { + "epoch": 2.03, + "learning_rate": 1.5432768138436678e-05, + "loss": 1.1043, + "step": 27418 + }, + { + "epoch": 2.03, + "learning_rate": 1.5432433259997148e-05, + "loss": 0.9612, + "step": 27419 + }, + { + "epoch": 2.03, + "learning_rate": 1.5432098372914694e-05, + "loss": 1.0747, + "step": 27420 + }, + { + "epoch": 2.03, + "learning_rate": 1.5431763477189847e-05, + "loss": 0.9954, + "step": 27421 + }, + { + "epoch": 2.03, + "learning_rate": 1.5431428572823147e-05, + "loss": 1.1115, + "step": 27422 + }, + { + "epoch": 2.03, + "learning_rate": 1.5431093659815118e-05, + "loss": 1.0836, + "step": 27423 + }, + { + "epoch": 2.03, + "learning_rate": 1.54307587381663e-05, + "loss": 0.9818, + "step": 27424 + }, + { + "epoch": 2.03, + "learning_rate": 1.5430423807877216e-05, + "loss": 0.9995, + "step": 27425 + }, + { + "epoch": 2.03, + "learning_rate": 1.5430088868948414e-05, + "loss": 0.981, + "step": 27426 + }, + { + "epoch": 2.03, + "learning_rate": 1.5429753921380414e-05, + "loss": 1.0297, + "step": 27427 + }, + { + "epoch": 2.03, + "learning_rate": 1.5429418965173756e-05, + "loss": 0.9688, + "step": 27428 + }, + { + "epoch": 2.03, + "learning_rate": 1.5429084000328968e-05, + "loss": 0.959, + "step": 27429 + }, + { + "epoch": 2.03, + "learning_rate": 1.5428749026846585e-05, + "loss": 1.1079, + "step": 27430 + }, + { + "epoch": 2.03, + "learning_rate": 1.542841404472714e-05, + "loss": 1.0657, + "step": 27431 + }, + { + "epoch": 2.03, + "learning_rate": 1.5428079053971162e-05, + "loss": 1.0357, + "step": 27432 + }, + { + "epoch": 2.03, + "learning_rate": 1.5427744054579193e-05, + "loss": 1.0275, + "step": 27433 + }, + { + "epoch": 2.03, + "learning_rate": 1.542740904655176e-05, + "loss": 1.0048, + "step": 27434 + }, + { + "epoch": 2.03, + "learning_rate": 1.5427074029889396e-05, + "loss": 1.0188, + "step": 27435 + }, + { + "epoch": 2.03, + "learning_rate": 1.5426739004592636e-05, + "loss": 0.9849, + "step": 27436 + }, + { + "epoch": 2.03, + "learning_rate": 1.542640397066201e-05, + "loss": 0.9925, + "step": 27437 + }, + { + "epoch": 2.03, + "learning_rate": 1.5426068928098056e-05, + "loss": 1.0207, + "step": 27438 + }, + { + "epoch": 2.03, + "learning_rate": 1.54257338769013e-05, + "loss": 0.9854, + "step": 27439 + }, + { + "epoch": 2.03, + "learning_rate": 1.542539881707228e-05, + "loss": 0.9433, + "step": 27440 + }, + { + "epoch": 2.03, + "learning_rate": 1.542506374861153e-05, + "loss": 1.0645, + "step": 27441 + }, + { + "epoch": 2.03, + "learning_rate": 1.542472867151958e-05, + "loss": 1.0711, + "step": 27442 + }, + { + "epoch": 2.03, + "learning_rate": 1.5424393585796967e-05, + "loss": 1.0839, + "step": 27443 + }, + { + "epoch": 2.03, + "learning_rate": 1.5424058491444215e-05, + "loss": 1.0658, + "step": 27444 + }, + { + "epoch": 2.03, + "learning_rate": 1.542372338846187e-05, + "loss": 1.0064, + "step": 27445 + }, + { + "epoch": 2.03, + "learning_rate": 1.5423388276850458e-05, + "loss": 1.0565, + "step": 27446 + }, + { + "epoch": 2.03, + "learning_rate": 1.542305315661051e-05, + "loss": 1.0125, + "step": 27447 + }, + { + "epoch": 2.03, + "learning_rate": 1.542271802774256e-05, + "loss": 0.9907, + "step": 27448 + }, + { + "epoch": 2.03, + "learning_rate": 1.542238289024715e-05, + "loss": 0.9716, + "step": 27449 + }, + { + "epoch": 2.03, + "learning_rate": 1.54220477441248e-05, + "loss": 0.9991, + "step": 27450 + }, + { + "epoch": 2.03, + "learning_rate": 1.5421712589376053e-05, + "loss": 0.9565, + "step": 27451 + }, + { + "epoch": 2.03, + "learning_rate": 1.542137742600144e-05, + "loss": 1.0842, + "step": 27452 + }, + { + "epoch": 2.03, + "learning_rate": 1.542104225400149e-05, + "loss": 1.043, + "step": 27453 + }, + { + "epoch": 2.03, + "learning_rate": 1.5420707073376742e-05, + "loss": 0.9899, + "step": 27454 + }, + { + "epoch": 2.03, + "learning_rate": 1.5420371884127725e-05, + "loss": 1.0061, + "step": 27455 + }, + { + "epoch": 2.03, + "learning_rate": 1.5420036686254974e-05, + "loss": 1.1019, + "step": 27456 + }, + { + "epoch": 2.03, + "learning_rate": 1.5419701479759024e-05, + "loss": 1.0603, + "step": 27457 + }, + { + "epoch": 2.03, + "learning_rate": 1.54193662646404e-05, + "loss": 0.9915, + "step": 27458 + }, + { + "epoch": 2.03, + "learning_rate": 1.5419031040899647e-05, + "loss": 0.9985, + "step": 27459 + }, + { + "epoch": 2.03, + "learning_rate": 1.5418695808537294e-05, + "loss": 0.969, + "step": 27460 + }, + { + "epoch": 2.03, + "learning_rate": 1.5418360567553873e-05, + "loss": 1.0499, + "step": 27461 + }, + { + "epoch": 2.03, + "learning_rate": 1.541802531794992e-05, + "loss": 1.134, + "step": 27462 + }, + { + "epoch": 2.03, + "learning_rate": 1.541769005972596e-05, + "loss": 1.0418, + "step": 27463 + }, + { + "epoch": 2.03, + "learning_rate": 1.5417354792882537e-05, + "loss": 0.9667, + "step": 27464 + }, + { + "epoch": 2.03, + "learning_rate": 1.541701951742018e-05, + "loss": 0.9869, + "step": 27465 + }, + { + "epoch": 2.03, + "learning_rate": 1.5416684233339417e-05, + "loss": 0.9409, + "step": 27466 + }, + { + "epoch": 2.03, + "learning_rate": 1.5416348940640796e-05, + "loss": 0.9422, + "step": 27467 + }, + { + "epoch": 2.03, + "learning_rate": 1.5416013639324834e-05, + "loss": 1.0069, + "step": 27468 + }, + { + "epoch": 2.03, + "learning_rate": 1.5415678329392077e-05, + "loss": 1.0412, + "step": 27469 + }, + { + "epoch": 2.03, + "learning_rate": 1.541534301084305e-05, + "loss": 0.9893, + "step": 27470 + }, + { + "epoch": 2.03, + "learning_rate": 1.541500768367829e-05, + "loss": 1.042, + "step": 27471 + }, + { + "epoch": 2.03, + "learning_rate": 1.541467234789833e-05, + "loss": 1.0062, + "step": 27472 + }, + { + "epoch": 2.03, + "learning_rate": 1.5414337003503705e-05, + "loss": 0.9018, + "step": 27473 + }, + { + "epoch": 2.03, + "learning_rate": 1.541400165049495e-05, + "loss": 1.0159, + "step": 27474 + }, + { + "epoch": 2.03, + "learning_rate": 1.541366628887259e-05, + "loss": 0.9722, + "step": 27475 + }, + { + "epoch": 2.03, + "learning_rate": 1.5413330918637164e-05, + "loss": 0.9866, + "step": 27476 + }, + { + "epoch": 2.03, + "learning_rate": 1.541299553978921e-05, + "loss": 1.0938, + "step": 27477 + }, + { + "epoch": 2.03, + "learning_rate": 1.5412660152329256e-05, + "loss": 1.0393, + "step": 27478 + }, + { + "epoch": 2.03, + "learning_rate": 1.541232475625784e-05, + "loss": 0.9485, + "step": 27479 + }, + { + "epoch": 2.03, + "learning_rate": 1.541198935157549e-05, + "loss": 0.9164, + "step": 27480 + }, + { + "epoch": 2.03, + "learning_rate": 1.541165393828274e-05, + "loss": 1.0507, + "step": 27481 + }, + { + "epoch": 2.03, + "learning_rate": 1.5411318516380133e-05, + "loss": 0.9912, + "step": 27482 + }, + { + "epoch": 2.03, + "learning_rate": 1.541098308586819e-05, + "loss": 1.0404, + "step": 27483 + }, + { + "epoch": 2.03, + "learning_rate": 1.5410647646747453e-05, + "loss": 0.9482, + "step": 27484 + }, + { + "epoch": 2.03, + "learning_rate": 1.541031219901845e-05, + "loss": 1.0013, + "step": 27485 + }, + { + "epoch": 2.03, + "learning_rate": 1.540997674268172e-05, + "loss": 1.0529, + "step": 27486 + }, + { + "epoch": 2.03, + "learning_rate": 1.5409641277737795e-05, + "loss": 0.9529, + "step": 27487 + }, + { + "epoch": 2.03, + "learning_rate": 1.5409305804187207e-05, + "loss": 0.8696, + "step": 27488 + }, + { + "epoch": 2.03, + "learning_rate": 1.540897032203049e-05, + "loss": 1.0341, + "step": 27489 + }, + { + "epoch": 2.03, + "learning_rate": 1.540863483126818e-05, + "loss": 1.1093, + "step": 27490 + }, + { + "epoch": 2.03, + "learning_rate": 1.5408299331900807e-05, + "loss": 0.9226, + "step": 27491 + }, + { + "epoch": 2.03, + "learning_rate": 1.540796382392891e-05, + "loss": 0.9705, + "step": 27492 + }, + { + "epoch": 2.03, + "learning_rate": 1.540762830735302e-05, + "loss": 1.0807, + "step": 27493 + }, + { + "epoch": 2.03, + "learning_rate": 1.5407292782173668e-05, + "loss": 1.072, + "step": 27494 + }, + { + "epoch": 2.03, + "learning_rate": 1.5406957248391394e-05, + "loss": 1.0051, + "step": 27495 + }, + { + "epoch": 2.03, + "learning_rate": 1.5406621706006725e-05, + "loss": 0.9112, + "step": 27496 + }, + { + "epoch": 2.03, + "learning_rate": 1.54062861550202e-05, + "loss": 0.9027, + "step": 27497 + }, + { + "epoch": 2.03, + "learning_rate": 1.5405950595432352e-05, + "loss": 0.9819, + "step": 27498 + }, + { + "epoch": 2.03, + "learning_rate": 1.540561502724371e-05, + "loss": 0.9958, + "step": 27499 + }, + { + "epoch": 2.03, + "learning_rate": 1.5405279450454814e-05, + "loss": 0.9716, + "step": 27500 + }, + { + "epoch": 2.03, + "learning_rate": 1.54049438650662e-05, + "loss": 0.9543, + "step": 27501 + }, + { + "epoch": 2.03, + "learning_rate": 1.5404608271078394e-05, + "loss": 1.0196, + "step": 27502 + }, + { + "epoch": 2.03, + "learning_rate": 1.5404272668491932e-05, + "loss": 1.0809, + "step": 27503 + }, + { + "epoch": 2.03, + "learning_rate": 1.5403937057307354e-05, + "loss": 0.9676, + "step": 27504 + }, + { + "epoch": 2.03, + "learning_rate": 1.5403601437525188e-05, + "loss": 0.9862, + "step": 27505 + }, + { + "epoch": 2.03, + "learning_rate": 1.5403265809145963e-05, + "loss": 1.1335, + "step": 27506 + }, + { + "epoch": 2.03, + "learning_rate": 1.5402930172170224e-05, + "loss": 1.009, + "step": 27507 + }, + { + "epoch": 2.03, + "learning_rate": 1.5402594526598504e-05, + "loss": 0.9537, + "step": 27508 + }, + { + "epoch": 2.03, + "learning_rate": 1.5402258872431327e-05, + "loss": 1.0489, + "step": 27509 + }, + { + "epoch": 2.03, + "learning_rate": 1.540192320966924e-05, + "loss": 0.9566, + "step": 27510 + }, + { + "epoch": 2.03, + "learning_rate": 1.5401587538312765e-05, + "loss": 1.0524, + "step": 27511 + }, + { + "epoch": 2.03, + "learning_rate": 1.5401251858362444e-05, + "loss": 1.0133, + "step": 27512 + }, + { + "epoch": 2.03, + "learning_rate": 1.5400916169818807e-05, + "loss": 0.9203, + "step": 27513 + }, + { + "epoch": 2.03, + "learning_rate": 1.5400580472682394e-05, + "loss": 0.9858, + "step": 27514 + }, + { + "epoch": 2.03, + "learning_rate": 1.540024476695373e-05, + "loss": 0.9136, + "step": 27515 + }, + { + "epoch": 2.03, + "learning_rate": 1.5399909052633356e-05, + "loss": 1.0161, + "step": 27516 + }, + { + "epoch": 2.03, + "learning_rate": 1.5399573329721802e-05, + "loss": 1.0121, + "step": 27517 + }, + { + "epoch": 2.03, + "learning_rate": 1.5399237598219605e-05, + "loss": 1.0529, + "step": 27518 + }, + { + "epoch": 2.03, + "learning_rate": 1.53989018581273e-05, + "loss": 1.0459, + "step": 27519 + }, + { + "epoch": 2.03, + "learning_rate": 1.5398566109445417e-05, + "loss": 1.13, + "step": 27520 + }, + { + "epoch": 2.03, + "learning_rate": 1.5398230352174493e-05, + "loss": 1.0427, + "step": 27521 + }, + { + "epoch": 2.03, + "learning_rate": 1.5397894586315065e-05, + "loss": 1.0241, + "step": 27522 + }, + { + "epoch": 2.03, + "learning_rate": 1.5397558811867658e-05, + "loss": 0.9362, + "step": 27523 + }, + { + "epoch": 2.03, + "learning_rate": 1.539722302883282e-05, + "loss": 1.0433, + "step": 27524 + }, + { + "epoch": 2.03, + "learning_rate": 1.539688723721107e-05, + "loss": 1.093, + "step": 27525 + }, + { + "epoch": 2.03, + "learning_rate": 1.5396551437002953e-05, + "loss": 0.961, + "step": 27526 + }, + { + "epoch": 2.03, + "learning_rate": 1.5396215628209e-05, + "loss": 0.9797, + "step": 27527 + }, + { + "epoch": 2.03, + "learning_rate": 1.5395879810829745e-05, + "loss": 1.0398, + "step": 27528 + }, + { + "epoch": 2.03, + "learning_rate": 1.539554398486572e-05, + "loss": 1.01, + "step": 27529 + }, + { + "epoch": 2.03, + "learning_rate": 1.5395208150317465e-05, + "loss": 1.0304, + "step": 27530 + }, + { + "epoch": 2.03, + "learning_rate": 1.5394872307185508e-05, + "loss": 1.1129, + "step": 27531 + }, + { + "epoch": 2.03, + "learning_rate": 1.539453645547039e-05, + "loss": 1.065, + "step": 27532 + }, + { + "epoch": 2.03, + "learning_rate": 1.539420059517264e-05, + "loss": 1.0233, + "step": 27533 + }, + { + "epoch": 2.03, + "learning_rate": 1.539386472629279e-05, + "loss": 1.0452, + "step": 27534 + }, + { + "epoch": 2.03, + "learning_rate": 1.5393528848831385e-05, + "loss": 1.0189, + "step": 27535 + }, + { + "epoch": 2.03, + "learning_rate": 1.5393192962788953e-05, + "loss": 1.0215, + "step": 27536 + }, + { + "epoch": 2.03, + "learning_rate": 1.5392857068166022e-05, + "loss": 1.0069, + "step": 27537 + }, + { + "epoch": 2.03, + "learning_rate": 1.5392521164963135e-05, + "loss": 0.8613, + "step": 27538 + }, + { + "epoch": 2.03, + "learning_rate": 1.5392185253180825e-05, + "loss": 0.9156, + "step": 27539 + }, + { + "epoch": 2.03, + "learning_rate": 1.5391849332819626e-05, + "loss": 1.0708, + "step": 27540 + }, + { + "epoch": 2.03, + "learning_rate": 1.539151340388007e-05, + "loss": 1.1109, + "step": 27541 + }, + { + "epoch": 2.04, + "learning_rate": 1.5391177466362693e-05, + "loss": 1.055, + "step": 27542 + }, + { + "epoch": 2.04, + "learning_rate": 1.5390841520268033e-05, + "loss": 1.0322, + "step": 27543 + }, + { + "epoch": 2.04, + "learning_rate": 1.5390505565596616e-05, + "loss": 0.8776, + "step": 27544 + }, + { + "epoch": 2.04, + "learning_rate": 1.5390169602348987e-05, + "loss": 1.0241, + "step": 27545 + }, + { + "epoch": 2.04, + "learning_rate": 1.5389833630525674e-05, + "loss": 1.1105, + "step": 27546 + }, + { + "epoch": 2.04, + "learning_rate": 1.538949765012721e-05, + "loss": 1.1313, + "step": 27547 + }, + { + "epoch": 2.04, + "learning_rate": 1.5389161661154137e-05, + "loss": 1.0334, + "step": 27548 + }, + { + "epoch": 2.04, + "learning_rate": 1.5388825663606983e-05, + "loss": 1.0264, + "step": 27549 + }, + { + "epoch": 2.04, + "learning_rate": 1.5388489657486283e-05, + "loss": 0.9569, + "step": 27550 + }, + { + "epoch": 2.04, + "learning_rate": 1.5388153642792576e-05, + "loss": 0.9844, + "step": 27551 + }, + { + "epoch": 2.04, + "learning_rate": 1.5387817619526388e-05, + "loss": 0.9676, + "step": 27552 + }, + { + "epoch": 2.04, + "learning_rate": 1.5387481587688263e-05, + "loss": 0.9216, + "step": 27553 + }, + { + "epoch": 2.04, + "learning_rate": 1.538714554727873e-05, + "loss": 1.0373, + "step": 27554 + }, + { + "epoch": 2.04, + "learning_rate": 1.538680949829833e-05, + "loss": 0.9953, + "step": 27555 + }, + { + "epoch": 2.04, + "learning_rate": 1.538647344074759e-05, + "loss": 0.9945, + "step": 27556 + }, + { + "epoch": 2.04, + "learning_rate": 1.538613737462705e-05, + "loss": 0.9891, + "step": 27557 + }, + { + "epoch": 2.04, + "learning_rate": 1.538580129993724e-05, + "loss": 1.0094, + "step": 27558 + }, + { + "epoch": 2.04, + "learning_rate": 1.5385465216678697e-05, + "loss": 1.1244, + "step": 27559 + }, + { + "epoch": 2.04, + "learning_rate": 1.5385129124851957e-05, + "loss": 0.9906, + "step": 27560 + }, + { + "epoch": 2.04, + "learning_rate": 1.5384793024457554e-05, + "loss": 1.0308, + "step": 27561 + }, + { + "epoch": 2.04, + "learning_rate": 1.538445691549602e-05, + "loss": 1.0373, + "step": 27562 + }, + { + "epoch": 2.04, + "learning_rate": 1.5384120797967895e-05, + "loss": 0.9534, + "step": 27563 + }, + { + "epoch": 2.04, + "learning_rate": 1.538378467187371e-05, + "loss": 1.0428, + "step": 27564 + }, + { + "epoch": 2.04, + "learning_rate": 1.5383448537213998e-05, + "loss": 1.054, + "step": 27565 + }, + { + "epoch": 2.04, + "learning_rate": 1.53831123939893e-05, + "loss": 1.0318, + "step": 27566 + }, + { + "epoch": 2.04, + "learning_rate": 1.538277624220015e-05, + "loss": 0.9697, + "step": 27567 + }, + { + "epoch": 2.04, + "learning_rate": 1.5382440081847072e-05, + "loss": 1.062, + "step": 27568 + }, + { + "epoch": 2.04, + "learning_rate": 1.5382103912930614e-05, + "loss": 1.1463, + "step": 27569 + }, + { + "epoch": 2.04, + "learning_rate": 1.5381767735451304e-05, + "loss": 1.0722, + "step": 27570 + }, + { + "epoch": 2.04, + "learning_rate": 1.538143154940968e-05, + "loss": 1.1051, + "step": 27571 + }, + { + "epoch": 2.04, + "learning_rate": 1.538109535480627e-05, + "loss": 1.1008, + "step": 27572 + }, + { + "epoch": 2.04, + "learning_rate": 1.538075915164162e-05, + "loss": 0.9429, + "step": 27573 + }, + { + "epoch": 2.04, + "learning_rate": 1.538042293991626e-05, + "loss": 1.0124, + "step": 27574 + }, + { + "epoch": 2.04, + "learning_rate": 1.538008671963072e-05, + "loss": 0.9363, + "step": 27575 + }, + { + "epoch": 2.04, + "learning_rate": 1.537975049078554e-05, + "loss": 0.9549, + "step": 27576 + }, + { + "epoch": 2.04, + "learning_rate": 1.537941425338126e-05, + "loss": 0.9705, + "step": 27577 + }, + { + "epoch": 2.04, + "learning_rate": 1.53790780074184e-05, + "loss": 0.9802, + "step": 27578 + }, + { + "epoch": 2.04, + "learning_rate": 1.5378741752897508e-05, + "loss": 1.0822, + "step": 27579 + }, + { + "epoch": 2.04, + "learning_rate": 1.537840548981911e-05, + "loss": 1.0481, + "step": 27580 + }, + { + "epoch": 2.04, + "learning_rate": 1.5378069218183754e-05, + "loss": 1.0498, + "step": 27581 + }, + { + "epoch": 2.04, + "learning_rate": 1.5377732937991964e-05, + "loss": 1.0489, + "step": 27582 + }, + { + "epoch": 2.04, + "learning_rate": 1.5377396649244275e-05, + "loss": 1.1189, + "step": 27583 + }, + { + "epoch": 2.04, + "learning_rate": 1.5377060351941227e-05, + "loss": 1.0356, + "step": 27584 + }, + { + "epoch": 2.04, + "learning_rate": 1.5376724046083353e-05, + "loss": 0.8921, + "step": 27585 + }, + { + "epoch": 2.04, + "learning_rate": 1.537638773167119e-05, + "loss": 1.0648, + "step": 27586 + }, + { + "epoch": 2.04, + "learning_rate": 1.5376051408705265e-05, + "loss": 1.055, + "step": 27587 + }, + { + "epoch": 2.04, + "learning_rate": 1.5375715077186125e-05, + "loss": 1.0239, + "step": 27588 + }, + { + "epoch": 2.04, + "learning_rate": 1.5375378737114297e-05, + "loss": 0.9981, + "step": 27589 + }, + { + "epoch": 2.04, + "learning_rate": 1.537504238849032e-05, + "loss": 1.1127, + "step": 27590 + }, + { + "epoch": 2.04, + "learning_rate": 1.5374706031314723e-05, + "loss": 1.0, + "step": 27591 + }, + { + "epoch": 2.04, + "learning_rate": 1.5374369665588046e-05, + "loss": 0.9384, + "step": 27592 + }, + { + "epoch": 2.04, + "learning_rate": 1.537403329131083e-05, + "loss": 1.0603, + "step": 27593 + }, + { + "epoch": 2.04, + "learning_rate": 1.53736969084836e-05, + "loss": 0.9988, + "step": 27594 + }, + { + "epoch": 2.04, + "learning_rate": 1.5373360517106894e-05, + "loss": 1.0131, + "step": 27595 + }, + { + "epoch": 2.04, + "learning_rate": 1.537302411718125e-05, + "loss": 1.0798, + "step": 27596 + }, + { + "epoch": 2.04, + "learning_rate": 1.53726877087072e-05, + "loss": 1.2183, + "step": 27597 + }, + { + "epoch": 2.04, + "learning_rate": 1.5372351291685284e-05, + "loss": 1.1048, + "step": 27598 + }, + { + "epoch": 2.04, + "learning_rate": 1.537201486611603e-05, + "loss": 0.9897, + "step": 27599 + }, + { + "epoch": 2.04, + "learning_rate": 1.537167843199998e-05, + "loss": 1.0696, + "step": 27600 + }, + { + "epoch": 2.04, + "learning_rate": 1.5371341989337666e-05, + "loss": 1.0862, + "step": 27601 + }, + { + "epoch": 2.04, + "learning_rate": 1.537100553812962e-05, + "loss": 1.0416, + "step": 27602 + }, + { + "epoch": 2.04, + "learning_rate": 1.5370669078376382e-05, + "loss": 1.0211, + "step": 27603 + }, + { + "epoch": 2.04, + "learning_rate": 1.537033261007849e-05, + "loss": 0.9751, + "step": 27604 + }, + { + "epoch": 2.04, + "learning_rate": 1.536999613323648e-05, + "loss": 1.0191, + "step": 27605 + }, + { + "epoch": 2.04, + "learning_rate": 1.5369659647850874e-05, + "loss": 0.9806, + "step": 27606 + }, + { + "epoch": 2.04, + "learning_rate": 1.536932315392222e-05, + "loss": 1.0107, + "step": 27607 + }, + { + "epoch": 2.04, + "learning_rate": 1.5368986651451047e-05, + "loss": 1.0071, + "step": 27608 + }, + { + "epoch": 2.04, + "learning_rate": 1.5368650140437894e-05, + "loss": 1.1095, + "step": 27609 + }, + { + "epoch": 2.04, + "learning_rate": 1.53683136208833e-05, + "loss": 0.9699, + "step": 27610 + }, + { + "epoch": 2.04, + "learning_rate": 1.536797709278779e-05, + "loss": 0.9704, + "step": 27611 + }, + { + "epoch": 2.04, + "learning_rate": 1.536764055615191e-05, + "loss": 1.052, + "step": 27612 + }, + { + "epoch": 2.04, + "learning_rate": 1.5367304010976187e-05, + "loss": 0.9928, + "step": 27613 + }, + { + "epoch": 2.04, + "learning_rate": 1.5366967457261162e-05, + "loss": 0.9657, + "step": 27614 + }, + { + "epoch": 2.04, + "learning_rate": 1.5366630895007367e-05, + "loss": 0.988, + "step": 27615 + }, + { + "epoch": 2.04, + "learning_rate": 1.5366294324215344e-05, + "loss": 1.0199, + "step": 27616 + }, + { + "epoch": 2.04, + "learning_rate": 1.5365957744885616e-05, + "loss": 0.9171, + "step": 27617 + }, + { + "epoch": 2.04, + "learning_rate": 1.536562115701873e-05, + "loss": 0.9576, + "step": 27618 + }, + { + "epoch": 2.04, + "learning_rate": 1.5365284560615217e-05, + "loss": 1.0503, + "step": 27619 + }, + { + "epoch": 2.04, + "learning_rate": 1.5364947955675617e-05, + "loss": 0.9595, + "step": 27620 + }, + { + "epoch": 2.04, + "learning_rate": 1.5364611342200452e-05, + "loss": 1.11, + "step": 27621 + }, + { + "epoch": 2.04, + "learning_rate": 1.5364274720190275e-05, + "loss": 1.0944, + "step": 27622 + }, + { + "epoch": 2.04, + "learning_rate": 1.536393808964561e-05, + "loss": 1.0889, + "step": 27623 + }, + { + "epoch": 2.04, + "learning_rate": 1.5363601450567002e-05, + "loss": 0.9972, + "step": 27624 + }, + { + "epoch": 2.04, + "learning_rate": 1.5363264802954973e-05, + "loss": 0.9029, + "step": 27625 + }, + { + "epoch": 2.04, + "learning_rate": 1.5362928146810073e-05, + "loss": 0.9422, + "step": 27626 + }, + { + "epoch": 2.04, + "learning_rate": 1.5362591482132824e-05, + "loss": 1.1103, + "step": 27627 + }, + { + "epoch": 2.04, + "learning_rate": 1.5362254808923774e-05, + "loss": 0.9051, + "step": 27628 + }, + { + "epoch": 2.04, + "learning_rate": 1.5361918127183453e-05, + "loss": 1.1043, + "step": 27629 + }, + { + "epoch": 2.04, + "learning_rate": 1.5361581436912395e-05, + "loss": 1.0174, + "step": 27630 + }, + { + "epoch": 2.04, + "learning_rate": 1.536124473811114e-05, + "loss": 0.9314, + "step": 27631 + }, + { + "epoch": 2.04, + "learning_rate": 1.536090803078022e-05, + "loss": 0.9447, + "step": 27632 + }, + { + "epoch": 2.04, + "learning_rate": 1.5360571314920172e-05, + "loss": 0.9818, + "step": 27633 + }, + { + "epoch": 2.04, + "learning_rate": 1.5360234590531532e-05, + "loss": 0.9433, + "step": 27634 + }, + { + "epoch": 2.04, + "learning_rate": 1.5359897857614835e-05, + "loss": 1.0607, + "step": 27635 + }, + { + "epoch": 2.04, + "learning_rate": 1.5359561116170617e-05, + "loss": 0.9973, + "step": 27636 + }, + { + "epoch": 2.04, + "learning_rate": 1.5359224366199414e-05, + "loss": 0.9536, + "step": 27637 + }, + { + "epoch": 2.04, + "learning_rate": 1.5358887607701763e-05, + "loss": 1.0765, + "step": 27638 + }, + { + "epoch": 2.04, + "learning_rate": 1.5358550840678197e-05, + "loss": 1.0552, + "step": 27639 + }, + { + "epoch": 2.04, + "learning_rate": 1.5358214065129254e-05, + "loss": 1.0116, + "step": 27640 + }, + { + "epoch": 2.04, + "learning_rate": 1.5357877281055467e-05, + "loss": 1.0562, + "step": 27641 + }, + { + "epoch": 2.04, + "learning_rate": 1.5357540488457378e-05, + "loss": 1.0085, + "step": 27642 + }, + { + "epoch": 2.04, + "learning_rate": 1.5357203687335518e-05, + "loss": 1.0472, + "step": 27643 + }, + { + "epoch": 2.04, + "learning_rate": 1.535686687769042e-05, + "loss": 1.0628, + "step": 27644 + }, + { + "epoch": 2.04, + "learning_rate": 1.5356530059522628e-05, + "loss": 1.0205, + "step": 27645 + }, + { + "epoch": 2.04, + "learning_rate": 1.535619323283267e-05, + "loss": 0.9986, + "step": 27646 + }, + { + "epoch": 2.04, + "learning_rate": 1.5355856397621083e-05, + "loss": 1.0925, + "step": 27647 + }, + { + "epoch": 2.04, + "learning_rate": 1.5355519553888408e-05, + "loss": 1.0856, + "step": 27648 + }, + { + "epoch": 2.04, + "learning_rate": 1.535518270163518e-05, + "loss": 1.074, + "step": 27649 + }, + { + "epoch": 2.04, + "learning_rate": 1.5354845840861933e-05, + "loss": 0.9614, + "step": 27650 + }, + { + "epoch": 2.04, + "learning_rate": 1.5354508971569198e-05, + "loss": 1.0828, + "step": 27651 + }, + { + "epoch": 2.04, + "learning_rate": 1.535417209375752e-05, + "loss": 0.998, + "step": 27652 + }, + { + "epoch": 2.04, + "learning_rate": 1.5353835207427433e-05, + "loss": 0.9475, + "step": 27653 + }, + { + "epoch": 2.04, + "learning_rate": 1.5353498312579463e-05, + "loss": 1.0283, + "step": 27654 + }, + { + "epoch": 2.04, + "learning_rate": 1.535316140921416e-05, + "loss": 0.9275, + "step": 27655 + }, + { + "epoch": 2.04, + "learning_rate": 1.535282449733205e-05, + "loss": 1.1122, + "step": 27656 + }, + { + "epoch": 2.04, + "learning_rate": 1.5352487576933677e-05, + "loss": 0.9908, + "step": 27657 + }, + { + "epoch": 2.04, + "learning_rate": 1.535215064801957e-05, + "loss": 1.1397, + "step": 27658 + }, + { + "epoch": 2.04, + "learning_rate": 1.5351813710590268e-05, + "loss": 1.0157, + "step": 27659 + }, + { + "epoch": 2.04, + "learning_rate": 1.5351476764646306e-05, + "loss": 1.0541, + "step": 27660 + }, + { + "epoch": 2.04, + "learning_rate": 1.5351139810188226e-05, + "loss": 1.0246, + "step": 27661 + }, + { + "epoch": 2.04, + "learning_rate": 1.5350802847216554e-05, + "loss": 0.9288, + "step": 27662 + }, + { + "epoch": 2.04, + "learning_rate": 1.5350465875731835e-05, + "loss": 0.9458, + "step": 27663 + }, + { + "epoch": 2.04, + "learning_rate": 1.5350128895734598e-05, + "loss": 1.1066, + "step": 27664 + }, + { + "epoch": 2.04, + "learning_rate": 1.5349791907225384e-05, + "loss": 1.007, + "step": 27665 + }, + { + "epoch": 2.04, + "learning_rate": 1.5349454910204726e-05, + "loss": 0.966, + "step": 27666 + }, + { + "epoch": 2.04, + "learning_rate": 1.534911790467316e-05, + "loss": 1.0009, + "step": 27667 + }, + { + "epoch": 2.04, + "learning_rate": 1.534878089063123e-05, + "loss": 1.0046, + "step": 27668 + }, + { + "epoch": 2.04, + "learning_rate": 1.5348443868079464e-05, + "loss": 0.9519, + "step": 27669 + }, + { + "epoch": 2.04, + "learning_rate": 1.5348106837018398e-05, + "loss": 1.0061, + "step": 27670 + }, + { + "epoch": 2.04, + "learning_rate": 1.5347769797448575e-05, + "loss": 1.0564, + "step": 27671 + }, + { + "epoch": 2.04, + "learning_rate": 1.5347432749370522e-05, + "loss": 1.0095, + "step": 27672 + }, + { + "epoch": 2.04, + "learning_rate": 1.5347095692784782e-05, + "loss": 0.923, + "step": 27673 + }, + { + "epoch": 2.04, + "learning_rate": 1.534675862769189e-05, + "loss": 1.1399, + "step": 27674 + }, + { + "epoch": 2.04, + "learning_rate": 1.5346421554092383e-05, + "loss": 0.9609, + "step": 27675 + }, + { + "epoch": 2.04, + "learning_rate": 1.534608447198679e-05, + "loss": 1.079, + "step": 27676 + }, + { + "epoch": 2.05, + "learning_rate": 1.5345747381375656e-05, + "loss": 1.0634, + "step": 27677 + }, + { + "epoch": 2.05, + "learning_rate": 1.5345410282259514e-05, + "loss": 0.9913, + "step": 27678 + }, + { + "epoch": 2.05, + "learning_rate": 1.5345073174638904e-05, + "loss": 1.091, + "step": 27679 + }, + { + "epoch": 2.05, + "learning_rate": 1.5344736058514358e-05, + "loss": 0.9816, + "step": 27680 + }, + { + "epoch": 2.05, + "learning_rate": 1.5344398933886414e-05, + "loss": 1.1049, + "step": 27681 + }, + { + "epoch": 2.05, + "learning_rate": 1.53440618007556e-05, + "loss": 1.0415, + "step": 27682 + }, + { + "epoch": 2.05, + "learning_rate": 1.5343724659122468e-05, + "loss": 1.0854, + "step": 27683 + }, + { + "epoch": 2.05, + "learning_rate": 1.5343387508987543e-05, + "loss": 0.8917, + "step": 27684 + }, + { + "epoch": 2.05, + "learning_rate": 1.534305035035137e-05, + "loss": 1.0908, + "step": 27685 + }, + { + "epoch": 2.05, + "learning_rate": 1.5342713183214474e-05, + "loss": 1.0576, + "step": 27686 + }, + { + "epoch": 2.05, + "learning_rate": 1.5342376007577403e-05, + "loss": 1.0768, + "step": 27687 + }, + { + "epoch": 2.05, + "learning_rate": 1.5342038823440685e-05, + "loss": 0.9707, + "step": 27688 + }, + { + "epoch": 2.05, + "learning_rate": 1.5341701630804862e-05, + "loss": 0.9954, + "step": 27689 + }, + { + "epoch": 2.05, + "learning_rate": 1.5341364429670463e-05, + "loss": 1.0171, + "step": 27690 + }, + { + "epoch": 2.05, + "learning_rate": 1.5341027220038033e-05, + "loss": 0.9973, + "step": 27691 + }, + { + "epoch": 2.05, + "learning_rate": 1.5340690001908107e-05, + "loss": 0.9979, + "step": 27692 + }, + { + "epoch": 2.05, + "learning_rate": 1.534035277528122e-05, + "loss": 1.0703, + "step": 27693 + }, + { + "epoch": 2.05, + "learning_rate": 1.5340015540157904e-05, + "loss": 1.0342, + "step": 27694 + }, + { + "epoch": 2.05, + "learning_rate": 1.53396782965387e-05, + "loss": 0.961, + "step": 27695 + }, + { + "epoch": 2.05, + "learning_rate": 1.5339341044424147e-05, + "loss": 1.0345, + "step": 27696 + }, + { + "epoch": 2.05, + "learning_rate": 1.533900378381478e-05, + "loss": 0.9726, + "step": 27697 + }, + { + "epoch": 2.05, + "learning_rate": 1.5338666514711134e-05, + "loss": 1.1191, + "step": 27698 + }, + { + "epoch": 2.05, + "learning_rate": 1.533832923711374e-05, + "loss": 1.0327, + "step": 27699 + }, + { + "epoch": 2.05, + "learning_rate": 1.5337991951023145e-05, + "loss": 0.8853, + "step": 27700 + }, + { + "epoch": 2.05, + "learning_rate": 1.5337654656439883e-05, + "loss": 1.052, + "step": 27701 + }, + { + "epoch": 2.05, + "learning_rate": 1.5337317353364486e-05, + "loss": 0.9773, + "step": 27702 + }, + { + "epoch": 2.05, + "learning_rate": 1.5336980041797492e-05, + "loss": 0.9902, + "step": 27703 + }, + { + "epoch": 2.05, + "learning_rate": 1.5336642721739444e-05, + "loss": 1.0323, + "step": 27704 + }, + { + "epoch": 2.05, + "learning_rate": 1.533630539319087e-05, + "loss": 1.0223, + "step": 27705 + }, + { + "epoch": 2.05, + "learning_rate": 1.533596805615231e-05, + "loss": 1.0781, + "step": 27706 + }, + { + "epoch": 2.05, + "learning_rate": 1.5335630710624303e-05, + "loss": 0.9538, + "step": 27707 + }, + { + "epoch": 2.05, + "learning_rate": 1.5335293356607386e-05, + "loss": 1.0504, + "step": 27708 + }, + { + "epoch": 2.05, + "learning_rate": 1.533495599410209e-05, + "loss": 0.9969, + "step": 27709 + }, + { + "epoch": 2.05, + "learning_rate": 1.5334618623108956e-05, + "loss": 1.0638, + "step": 27710 + }, + { + "epoch": 2.05, + "learning_rate": 1.533428124362852e-05, + "loss": 1.011, + "step": 27711 + }, + { + "epoch": 2.05, + "learning_rate": 1.533394385566132e-05, + "loss": 1.0508, + "step": 27712 + }, + { + "epoch": 2.05, + "learning_rate": 1.533360645920789e-05, + "loss": 0.9611, + "step": 27713 + }, + { + "epoch": 2.05, + "learning_rate": 1.533326905426877e-05, + "loss": 1.091, + "step": 27714 + }, + { + "epoch": 2.05, + "learning_rate": 1.5332931640844492e-05, + "loss": 0.9195, + "step": 27715 + }, + { + "epoch": 2.05, + "learning_rate": 1.53325942189356e-05, + "loss": 0.9675, + "step": 27716 + }, + { + "epoch": 2.05, + "learning_rate": 1.533225678854263e-05, + "loss": 1.0217, + "step": 27717 + }, + { + "epoch": 2.05, + "learning_rate": 1.533191934966611e-05, + "loss": 0.9719, + "step": 27718 + }, + { + "epoch": 2.05, + "learning_rate": 1.5331581902306582e-05, + "loss": 1.0855, + "step": 27719 + }, + { + "epoch": 2.05, + "learning_rate": 1.533124444646459e-05, + "loss": 1.1109, + "step": 27720 + }, + { + "epoch": 2.05, + "learning_rate": 1.5330906982140658e-05, + "loss": 0.987, + "step": 27721 + }, + { + "epoch": 2.05, + "learning_rate": 1.533056950933533e-05, + "loss": 1.0086, + "step": 27722 + }, + { + "epoch": 2.05, + "learning_rate": 1.5330232028049142e-05, + "loss": 1.0125, + "step": 27723 + }, + { + "epoch": 2.05, + "learning_rate": 1.5329894538282632e-05, + "loss": 1.0627, + "step": 27724 + }, + { + "epoch": 2.05, + "learning_rate": 1.5329557040036336e-05, + "loss": 1.0233, + "step": 27725 + }, + { + "epoch": 2.05, + "learning_rate": 1.5329219533310796e-05, + "loss": 0.9749, + "step": 27726 + }, + { + "epoch": 2.05, + "learning_rate": 1.5328882018106538e-05, + "loss": 0.9309, + "step": 27727 + }, + { + "epoch": 2.05, + "learning_rate": 1.5328544494424107e-05, + "loss": 1.0624, + "step": 27728 + }, + { + "epoch": 2.05, + "learning_rate": 1.5328206962264038e-05, + "loss": 1.0773, + "step": 27729 + }, + { + "epoch": 2.05, + "learning_rate": 1.5327869421626866e-05, + "loss": 1.1409, + "step": 27730 + }, + { + "epoch": 2.05, + "learning_rate": 1.532753187251313e-05, + "loss": 0.9405, + "step": 27731 + }, + { + "epoch": 2.05, + "learning_rate": 1.5327194314923368e-05, + "loss": 0.9959, + "step": 27732 + }, + { + "epoch": 2.05, + "learning_rate": 1.5326856748858115e-05, + "loss": 0.8288, + "step": 27733 + }, + { + "epoch": 2.05, + "learning_rate": 1.5326519174317912e-05, + "loss": 0.9795, + "step": 27734 + }, + { + "epoch": 2.05, + "learning_rate": 1.5326181591303293e-05, + "loss": 0.9124, + "step": 27735 + }, + { + "epoch": 2.05, + "learning_rate": 1.5325843999814793e-05, + "loss": 0.9627, + "step": 27736 + }, + { + "epoch": 2.05, + "learning_rate": 1.5325506399852952e-05, + "loss": 1.105, + "step": 27737 + }, + { + "epoch": 2.05, + "learning_rate": 1.5325168791418308e-05, + "loss": 1.0117, + "step": 27738 + }, + { + "epoch": 2.05, + "learning_rate": 1.5324831174511397e-05, + "loss": 0.9644, + "step": 27739 + }, + { + "epoch": 2.05, + "learning_rate": 1.5324493549132754e-05, + "loss": 0.8752, + "step": 27740 + }, + { + "epoch": 2.05, + "learning_rate": 1.5324155915282917e-05, + "loss": 0.9711, + "step": 27741 + }, + { + "epoch": 2.05, + "learning_rate": 1.532381827296243e-05, + "loss": 0.8924, + "step": 27742 + }, + { + "epoch": 2.05, + "learning_rate": 1.5323480622171817e-05, + "loss": 1.053, + "step": 27743 + }, + { + "epoch": 2.05, + "learning_rate": 1.5323142962911622e-05, + "loss": 0.9352, + "step": 27744 + }, + { + "epoch": 2.05, + "learning_rate": 1.532280529518239e-05, + "loss": 1.0586, + "step": 27745 + }, + { + "epoch": 2.05, + "learning_rate": 1.5322467618984645e-05, + "loss": 1.0024, + "step": 27746 + }, + { + "epoch": 2.05, + "learning_rate": 1.5322129934318933e-05, + "loss": 1.0447, + "step": 27747 + }, + { + "epoch": 2.05, + "learning_rate": 1.5321792241185785e-05, + "loss": 0.8282, + "step": 27748 + }, + { + "epoch": 2.05, + "learning_rate": 1.5321454539585744e-05, + "loss": 1.0825, + "step": 27749 + }, + { + "epoch": 2.05, + "learning_rate": 1.5321116829519344e-05, + "loss": 0.9602, + "step": 27750 + }, + { + "epoch": 2.05, + "learning_rate": 1.5320779110987127e-05, + "loss": 0.9695, + "step": 27751 + }, + { + "epoch": 2.05, + "learning_rate": 1.5320441383989622e-05, + "loss": 1.0412, + "step": 27752 + }, + { + "epoch": 2.05, + "learning_rate": 1.5320103648527375e-05, + "loss": 0.9918, + "step": 27753 + }, + { + "epoch": 2.05, + "learning_rate": 1.5319765904600918e-05, + "loss": 0.9999, + "step": 27754 + }, + { + "epoch": 2.05, + "learning_rate": 1.5319428152210788e-05, + "loss": 0.9542, + "step": 27755 + }, + { + "epoch": 2.05, + "learning_rate": 1.5319090391357522e-05, + "loss": 0.9945, + "step": 27756 + }, + { + "epoch": 2.05, + "learning_rate": 1.5318752622041662e-05, + "loss": 0.9049, + "step": 27757 + }, + { + "epoch": 2.05, + "learning_rate": 1.5318414844263743e-05, + "loss": 1.0428, + "step": 27758 + }, + { + "epoch": 2.05, + "learning_rate": 1.53180770580243e-05, + "loss": 0.8369, + "step": 27759 + }, + { + "epoch": 2.05, + "learning_rate": 1.5317739263323875e-05, + "loss": 0.9803, + "step": 27760 + }, + { + "epoch": 2.05, + "learning_rate": 1.5317401460163e-05, + "loss": 1.0634, + "step": 27761 + }, + { + "epoch": 2.05, + "learning_rate": 1.531706364854222e-05, + "loss": 1.0743, + "step": 27762 + }, + { + "epoch": 2.05, + "learning_rate": 1.5316725828462065e-05, + "loss": 0.9915, + "step": 27763 + }, + { + "epoch": 2.05, + "learning_rate": 1.5316387999923078e-05, + "loss": 1.0728, + "step": 27764 + }, + { + "epoch": 2.05, + "learning_rate": 1.531605016292579e-05, + "loss": 0.8782, + "step": 27765 + }, + { + "epoch": 2.05, + "learning_rate": 1.531571231747074e-05, + "loss": 1.0453, + "step": 27766 + }, + { + "epoch": 2.05, + "learning_rate": 1.5315374463558474e-05, + "loss": 1.0105, + "step": 27767 + }, + { + "epoch": 2.05, + "learning_rate": 1.531503660118952e-05, + "loss": 0.8758, + "step": 27768 + }, + { + "epoch": 2.05, + "learning_rate": 1.531469873036442e-05, + "loss": 1.0564, + "step": 27769 + }, + { + "epoch": 2.05, + "learning_rate": 1.5314360851083707e-05, + "loss": 0.9617, + "step": 27770 + }, + { + "epoch": 2.05, + "learning_rate": 1.5314022963347925e-05, + "loss": 0.8803, + "step": 27771 + }, + { + "epoch": 2.05, + "learning_rate": 1.5313685067157608e-05, + "loss": 1.0196, + "step": 27772 + }, + { + "epoch": 2.05, + "learning_rate": 1.5313347162513296e-05, + "loss": 1.0508, + "step": 27773 + }, + { + "epoch": 2.05, + "learning_rate": 1.531300924941552e-05, + "loss": 0.8962, + "step": 27774 + }, + { + "epoch": 2.05, + "learning_rate": 1.5312671327864828e-05, + "loss": 0.9961, + "step": 27775 + }, + { + "epoch": 2.05, + "learning_rate": 1.5312333397861746e-05, + "loss": 1.0355, + "step": 27776 + }, + { + "epoch": 2.05, + "learning_rate": 1.5311995459406823e-05, + "loss": 1.0799, + "step": 27777 + }, + { + "epoch": 2.05, + "learning_rate": 1.531165751250059e-05, + "loss": 0.9408, + "step": 27778 + }, + { + "epoch": 2.05, + "learning_rate": 1.5311319557143584e-05, + "loss": 1.0208, + "step": 27779 + }, + { + "epoch": 2.05, + "learning_rate": 1.531098159333634e-05, + "loss": 1.1176, + "step": 27780 + }, + { + "epoch": 2.05, + "learning_rate": 1.5310643621079407e-05, + "loss": 0.9426, + "step": 27781 + }, + { + "epoch": 2.05, + "learning_rate": 1.531030564037332e-05, + "loss": 1.0026, + "step": 27782 + }, + { + "epoch": 2.05, + "learning_rate": 1.5309967651218605e-05, + "loss": 0.9901, + "step": 27783 + }, + { + "epoch": 2.05, + "learning_rate": 1.530962965361581e-05, + "loss": 1.1079, + "step": 27784 + }, + { + "epoch": 2.05, + "learning_rate": 1.5309291647565468e-05, + "loss": 1.0955, + "step": 27785 + }, + { + "epoch": 2.05, + "learning_rate": 1.530895363306812e-05, + "loss": 1.0339, + "step": 27786 + }, + { + "epoch": 2.05, + "learning_rate": 1.5308615610124305e-05, + "loss": 1.0551, + "step": 27787 + }, + { + "epoch": 2.05, + "learning_rate": 1.5308277578734557e-05, + "loss": 1.0629, + "step": 27788 + }, + { + "epoch": 2.05, + "learning_rate": 1.5307939538899413e-05, + "loss": 0.9913, + "step": 27789 + }, + { + "epoch": 2.05, + "learning_rate": 1.5307601490619418e-05, + "loss": 1.0008, + "step": 27790 + }, + { + "epoch": 2.05, + "learning_rate": 1.5307263433895104e-05, + "loss": 1.0519, + "step": 27791 + }, + { + "epoch": 2.05, + "learning_rate": 1.5306925368727007e-05, + "loss": 1.1351, + "step": 27792 + }, + { + "epoch": 2.05, + "learning_rate": 1.5306587295115665e-05, + "loss": 0.951, + "step": 27793 + }, + { + "epoch": 2.05, + "learning_rate": 1.5306249213061627e-05, + "loss": 1.0517, + "step": 27794 + }, + { + "epoch": 2.05, + "learning_rate": 1.530591112256542e-05, + "loss": 1.0291, + "step": 27795 + }, + { + "epoch": 2.05, + "learning_rate": 1.530557302362758e-05, + "loss": 1.0788, + "step": 27796 + }, + { + "epoch": 2.05, + "learning_rate": 1.530523491624865e-05, + "loss": 0.8919, + "step": 27797 + }, + { + "epoch": 2.05, + "learning_rate": 1.530489680042917e-05, + "loss": 0.987, + "step": 27798 + }, + { + "epoch": 2.05, + "learning_rate": 1.5304558676169676e-05, + "loss": 0.9694, + "step": 27799 + }, + { + "epoch": 2.05, + "learning_rate": 1.5304220543470698e-05, + "loss": 0.9207, + "step": 27800 + }, + { + "epoch": 2.05, + "learning_rate": 1.5303882402332788e-05, + "loss": 0.9989, + "step": 27801 + }, + { + "epoch": 2.05, + "learning_rate": 1.5303544252756475e-05, + "loss": 1.1503, + "step": 27802 + }, + { + "epoch": 2.05, + "learning_rate": 1.5303206094742298e-05, + "loss": 1.019, + "step": 27803 + }, + { + "epoch": 2.05, + "learning_rate": 1.5302867928290795e-05, + "loss": 0.8967, + "step": 27804 + }, + { + "epoch": 2.05, + "learning_rate": 1.530252975340251e-05, + "loss": 0.9826, + "step": 27805 + }, + { + "epoch": 2.05, + "learning_rate": 1.530219157007797e-05, + "loss": 0.9748, + "step": 27806 + }, + { + "epoch": 2.05, + "learning_rate": 1.5301853378317723e-05, + "loss": 1.0922, + "step": 27807 + }, + { + "epoch": 2.05, + "learning_rate": 1.53015151781223e-05, + "loss": 1.0381, + "step": 27808 + }, + { + "epoch": 2.05, + "learning_rate": 1.5301176969492244e-05, + "loss": 0.9732, + "step": 27809 + }, + { + "epoch": 2.05, + "learning_rate": 1.5300838752428093e-05, + "loss": 0.9078, + "step": 27810 + }, + { + "epoch": 2.05, + "learning_rate": 1.530050052693038e-05, + "loss": 1.1152, + "step": 27811 + }, + { + "epoch": 2.06, + "learning_rate": 1.5300162292999653e-05, + "loss": 1.1156, + "step": 27812 + }, + { + "epoch": 2.06, + "learning_rate": 1.5299824050636437e-05, + "loss": 0.9254, + "step": 27813 + }, + { + "epoch": 2.06, + "learning_rate": 1.529948579984128e-05, + "loss": 0.9336, + "step": 27814 + }, + { + "epoch": 2.06, + "learning_rate": 1.5299147540614715e-05, + "loss": 1.0993, + "step": 27815 + }, + { + "epoch": 2.06, + "learning_rate": 1.5298809272957283e-05, + "loss": 0.8496, + "step": 27816 + }, + { + "epoch": 2.06, + "learning_rate": 1.529847099686952e-05, + "loss": 0.8678, + "step": 27817 + }, + { + "epoch": 2.06, + "learning_rate": 1.5298132712351964e-05, + "loss": 0.9499, + "step": 27818 + }, + { + "epoch": 2.06, + "learning_rate": 1.529779441940516e-05, + "loss": 0.9979, + "step": 27819 + }, + { + "epoch": 2.06, + "learning_rate": 1.529745611802964e-05, + "loss": 0.9293, + "step": 27820 + }, + { + "epoch": 2.06, + "learning_rate": 1.529711780822594e-05, + "loss": 1.0322, + "step": 27821 + }, + { + "epoch": 2.06, + "learning_rate": 1.5296779489994602e-05, + "loss": 0.9962, + "step": 27822 + }, + { + "epoch": 2.06, + "learning_rate": 1.5296441163336162e-05, + "loss": 1.0737, + "step": 27823 + }, + { + "epoch": 2.06, + "learning_rate": 1.5296102828251164e-05, + "loss": 0.9775, + "step": 27824 + }, + { + "epoch": 2.06, + "learning_rate": 1.5295764484740135e-05, + "loss": 0.9334, + "step": 27825 + }, + { + "epoch": 2.06, + "learning_rate": 1.5295426132803626e-05, + "loss": 1.0129, + "step": 27826 + }, + { + "epoch": 2.06, + "learning_rate": 1.5295087772442172e-05, + "loss": 1.047, + "step": 27827 + }, + { + "epoch": 2.06, + "learning_rate": 1.5294749403656307e-05, + "loss": 1.0489, + "step": 27828 + }, + { + "epoch": 2.06, + "learning_rate": 1.529441102644657e-05, + "loss": 1.0331, + "step": 27829 + }, + { + "epoch": 2.06, + "learning_rate": 1.5294072640813503e-05, + "loss": 1.0283, + "step": 27830 + }, + { + "epoch": 2.06, + "learning_rate": 1.529373424675764e-05, + "loss": 0.9643, + "step": 27831 + }, + { + "epoch": 2.06, + "learning_rate": 1.5293395844279522e-05, + "loss": 0.9371, + "step": 27832 + }, + { + "epoch": 2.06, + "learning_rate": 1.529305743337969e-05, + "loss": 1.1034, + "step": 27833 + }, + { + "epoch": 2.06, + "learning_rate": 1.5292719014058676e-05, + "loss": 0.9758, + "step": 27834 + }, + { + "epoch": 2.06, + "learning_rate": 1.529238058631702e-05, + "loss": 0.9614, + "step": 27835 + }, + { + "epoch": 2.06, + "learning_rate": 1.5292042150155265e-05, + "loss": 0.9996, + "step": 27836 + }, + { + "epoch": 2.06, + "learning_rate": 1.529170370557395e-05, + "loss": 0.9315, + "step": 27837 + }, + { + "epoch": 2.06, + "learning_rate": 1.5291365252573605e-05, + "loss": 1.0344, + "step": 27838 + }, + { + "epoch": 2.06, + "learning_rate": 1.5291026791154774e-05, + "loss": 1.031, + "step": 27839 + }, + { + "epoch": 2.06, + "learning_rate": 1.5290688321317998e-05, + "loss": 0.9948, + "step": 27840 + }, + { + "epoch": 2.06, + "learning_rate": 1.529034984306381e-05, + "loss": 0.9293, + "step": 27841 + }, + { + "epoch": 2.06, + "learning_rate": 1.5290011356392752e-05, + "loss": 0.9202, + "step": 27842 + }, + { + "epoch": 2.06, + "learning_rate": 1.528967286130536e-05, + "loss": 1.0605, + "step": 27843 + }, + { + "epoch": 2.06, + "learning_rate": 1.5289334357802176e-05, + "loss": 1.017, + "step": 27844 + }, + { + "epoch": 2.06, + "learning_rate": 1.5288995845883735e-05, + "loss": 1.041, + "step": 27845 + }, + { + "epoch": 2.06, + "learning_rate": 1.528865732555058e-05, + "loss": 1.0411, + "step": 27846 + }, + { + "epoch": 2.06, + "learning_rate": 1.5288318796803246e-05, + "loss": 0.9433, + "step": 27847 + }, + { + "epoch": 2.06, + "learning_rate": 1.528798025964227e-05, + "loss": 1.0119, + "step": 27848 + }, + { + "epoch": 2.06, + "learning_rate": 1.5287641714068197e-05, + "loss": 0.9974, + "step": 27849 + }, + { + "epoch": 2.06, + "learning_rate": 1.528730316008156e-05, + "loss": 0.9405, + "step": 27850 + }, + { + "epoch": 2.06, + "learning_rate": 1.52869645976829e-05, + "loss": 0.9363, + "step": 27851 + }, + { + "epoch": 2.06, + "learning_rate": 1.5286626026872752e-05, + "loss": 0.9782, + "step": 27852 + }, + { + "epoch": 2.06, + "learning_rate": 1.528628744765166e-05, + "loss": 0.9525, + "step": 27853 + }, + { + "epoch": 2.06, + "learning_rate": 1.5285948860020158e-05, + "loss": 0.9374, + "step": 27854 + }, + { + "epoch": 2.06, + "learning_rate": 1.528561026397879e-05, + "loss": 1.0528, + "step": 27855 + }, + { + "epoch": 2.06, + "learning_rate": 1.5285271659528088e-05, + "loss": 1.0729, + "step": 27856 + }, + { + "epoch": 2.06, + "learning_rate": 1.5284933046668602e-05, + "loss": 0.9154, + "step": 27857 + }, + { + "epoch": 2.06, + "learning_rate": 1.5284594425400857e-05, + "loss": 0.9756, + "step": 27858 + }, + { + "epoch": 2.06, + "learning_rate": 1.52842557957254e-05, + "loss": 1.0857, + "step": 27859 + }, + { + "epoch": 2.06, + "learning_rate": 1.5283917157642763e-05, + "loss": 1.0632, + "step": 27860 + }, + { + "epoch": 2.06, + "learning_rate": 1.5283578511153498e-05, + "loss": 0.9492, + "step": 27861 + }, + { + "epoch": 2.06, + "learning_rate": 1.528323985625813e-05, + "loss": 0.9436, + "step": 27862 + }, + { + "epoch": 2.06, + "learning_rate": 1.5282901192957204e-05, + "loss": 0.9786, + "step": 27863 + }, + { + "epoch": 2.06, + "learning_rate": 1.5282562521251253e-05, + "loss": 0.9346, + "step": 27864 + }, + { + "epoch": 2.06, + "learning_rate": 1.5282223841140824e-05, + "loss": 1.0174, + "step": 27865 + }, + { + "epoch": 2.06, + "learning_rate": 1.5281885152626456e-05, + "loss": 1.0776, + "step": 27866 + }, + { + "epoch": 2.06, + "learning_rate": 1.528154645570868e-05, + "loss": 0.9973, + "step": 27867 + }, + { + "epoch": 2.06, + "learning_rate": 1.528120775038804e-05, + "loss": 0.9395, + "step": 27868 + }, + { + "epoch": 2.06, + "learning_rate": 1.5280869036665075e-05, + "loss": 1.0062, + "step": 27869 + }, + { + "epoch": 2.06, + "learning_rate": 1.5280530314540324e-05, + "loss": 0.9558, + "step": 27870 + }, + { + "epoch": 2.06, + "learning_rate": 1.5280191584014325e-05, + "loss": 0.9022, + "step": 27871 + }, + { + "epoch": 2.06, + "learning_rate": 1.5279852845087612e-05, + "loss": 1.0583, + "step": 27872 + }, + { + "epoch": 2.06, + "learning_rate": 1.527951409776073e-05, + "loss": 1.0237, + "step": 27873 + }, + { + "epoch": 2.06, + "learning_rate": 1.527917534203422e-05, + "loss": 1.0188, + "step": 27874 + }, + { + "epoch": 2.06, + "learning_rate": 1.5278836577908617e-05, + "loss": 1.0092, + "step": 27875 + }, + { + "epoch": 2.06, + "learning_rate": 1.5278497805384457e-05, + "loss": 0.9823, + "step": 27876 + }, + { + "epoch": 2.06, + "learning_rate": 1.5278159024462288e-05, + "loss": 0.9233, + "step": 27877 + }, + { + "epoch": 2.06, + "learning_rate": 1.527782023514264e-05, + "loss": 0.9926, + "step": 27878 + }, + { + "epoch": 2.06, + "learning_rate": 1.5277481437426058e-05, + "loss": 1.0256, + "step": 27879 + }, + { + "epoch": 2.06, + "learning_rate": 1.5277142631313074e-05, + "loss": 1.1036, + "step": 27880 + }, + { + "epoch": 2.06, + "learning_rate": 1.5276803816804236e-05, + "loss": 1.0257, + "step": 27881 + }, + { + "epoch": 2.06, + "learning_rate": 1.5276464993900075e-05, + "loss": 0.8689, + "step": 27882 + }, + { + "epoch": 2.06, + "learning_rate": 1.5276126162601136e-05, + "loss": 1.0131, + "step": 27883 + }, + { + "epoch": 2.06, + "learning_rate": 1.5275787322907955e-05, + "loss": 1.0376, + "step": 27884 + }, + { + "epoch": 2.06, + "learning_rate": 1.527544847482107e-05, + "loss": 0.9236, + "step": 27885 + }, + { + "epoch": 2.06, + "learning_rate": 1.5275109618341028e-05, + "loss": 0.9388, + "step": 27886 + }, + { + "epoch": 2.06, + "learning_rate": 1.5274770753468352e-05, + "loss": 1.107, + "step": 27887 + }, + { + "epoch": 2.06, + "learning_rate": 1.5274431880203603e-05, + "loss": 1.0036, + "step": 27888 + }, + { + "epoch": 2.06, + "learning_rate": 1.5274092998547302e-05, + "loss": 1.0515, + "step": 27889 + }, + { + "epoch": 2.06, + "learning_rate": 1.527375410849999e-05, + "loss": 1.0564, + "step": 27890 + }, + { + "epoch": 2.06, + "learning_rate": 1.527341521006222e-05, + "loss": 1.0324, + "step": 27891 + }, + { + "epoch": 2.06, + "learning_rate": 1.5273076303234518e-05, + "loss": 0.9215, + "step": 27892 + }, + { + "epoch": 2.06, + "learning_rate": 1.5272737388017426e-05, + "loss": 0.9965, + "step": 27893 + }, + { + "epoch": 2.06, + "learning_rate": 1.5272398464411486e-05, + "loss": 1.0272, + "step": 27894 + }, + { + "epoch": 2.06, + "learning_rate": 1.5272059532417233e-05, + "loss": 1.032, + "step": 27895 + }, + { + "epoch": 2.06, + "learning_rate": 1.527172059203521e-05, + "loss": 1.1004, + "step": 27896 + }, + { + "epoch": 2.06, + "learning_rate": 1.5271381643265954e-05, + "loss": 1.0067, + "step": 27897 + }, + { + "epoch": 2.06, + "learning_rate": 1.5271042686110008e-05, + "loss": 0.9732, + "step": 27898 + }, + { + "epoch": 2.06, + "learning_rate": 1.5270703720567907e-05, + "loss": 1.0024, + "step": 27899 + }, + { + "epoch": 2.06, + "learning_rate": 1.527036474664019e-05, + "loss": 0.8568, + "step": 27900 + }, + { + "epoch": 2.06, + "learning_rate": 1.52700257643274e-05, + "loss": 1.0999, + "step": 27901 + }, + { + "epoch": 2.06, + "learning_rate": 1.5269686773630074e-05, + "loss": 0.9944, + "step": 27902 + }, + { + "epoch": 2.06, + "learning_rate": 1.526934777454875e-05, + "loss": 0.9726, + "step": 27903 + }, + { + "epoch": 2.06, + "learning_rate": 1.5269008767083975e-05, + "loss": 1.0325, + "step": 27904 + }, + { + "epoch": 2.06, + "learning_rate": 1.5268669751236276e-05, + "loss": 1.0311, + "step": 27905 + }, + { + "epoch": 2.06, + "learning_rate": 1.5268330727006203e-05, + "loss": 0.9992, + "step": 27906 + }, + { + "epoch": 2.06, + "learning_rate": 1.526799169439429e-05, + "loss": 1.0417, + "step": 27907 + }, + { + "epoch": 2.06, + "learning_rate": 1.5267652653401075e-05, + "loss": 1.0034, + "step": 27908 + }, + { + "epoch": 2.06, + "learning_rate": 1.52673136040271e-05, + "loss": 1.1062, + "step": 27909 + }, + { + "epoch": 2.06, + "learning_rate": 1.5266974546272907e-05, + "loss": 0.9977, + "step": 27910 + }, + { + "epoch": 2.06, + "learning_rate": 1.526663548013903e-05, + "loss": 1.0648, + "step": 27911 + }, + { + "epoch": 2.06, + "learning_rate": 1.5266296405626015e-05, + "loss": 0.9728, + "step": 27912 + }, + { + "epoch": 2.06, + "learning_rate": 1.5265957322734398e-05, + "loss": 0.991, + "step": 27913 + }, + { + "epoch": 2.06, + "learning_rate": 1.5265618231464717e-05, + "loss": 0.9844, + "step": 27914 + }, + { + "epoch": 2.06, + "learning_rate": 1.526527913181751e-05, + "loss": 0.9221, + "step": 27915 + }, + { + "epoch": 2.06, + "learning_rate": 1.5264940023793324e-05, + "loss": 1.064, + "step": 27916 + }, + { + "epoch": 2.06, + "learning_rate": 1.526460090739269e-05, + "loss": 1.0907, + "step": 27917 + }, + { + "epoch": 2.06, + "learning_rate": 1.526426178261615e-05, + "loss": 1.0273, + "step": 27918 + }, + { + "epoch": 2.06, + "learning_rate": 1.526392264946425e-05, + "loss": 1.0149, + "step": 27919 + }, + { + "epoch": 2.06, + "learning_rate": 1.526358350793752e-05, + "loss": 0.9434, + "step": 27920 + }, + { + "epoch": 2.06, + "learning_rate": 1.5263244358036507e-05, + "loss": 0.8722, + "step": 27921 + }, + { + "epoch": 2.06, + "learning_rate": 1.526290519976175e-05, + "loss": 0.9915, + "step": 27922 + }, + { + "epoch": 2.06, + "learning_rate": 1.526256603311378e-05, + "loss": 1.0714, + "step": 27923 + }, + { + "epoch": 2.06, + "learning_rate": 1.5262226858093146e-05, + "loss": 1.0477, + "step": 27924 + }, + { + "epoch": 2.06, + "learning_rate": 1.5261887674700382e-05, + "loss": 0.9576, + "step": 27925 + }, + { + "epoch": 2.06, + "learning_rate": 1.5261548482936032e-05, + "loss": 1.025, + "step": 27926 + }, + { + "epoch": 2.06, + "learning_rate": 1.5261209282800632e-05, + "loss": 1.0833, + "step": 27927 + }, + { + "epoch": 2.06, + "learning_rate": 1.5260870074294726e-05, + "loss": 1.0657, + "step": 27928 + }, + { + "epoch": 2.06, + "learning_rate": 1.5260530857418848e-05, + "loss": 0.9637, + "step": 27929 + }, + { + "epoch": 2.06, + "learning_rate": 1.5260191632173542e-05, + "loss": 1.0959, + "step": 27930 + }, + { + "epoch": 2.06, + "learning_rate": 1.5259852398559344e-05, + "loss": 1.1054, + "step": 27931 + }, + { + "epoch": 2.06, + "learning_rate": 1.5259513156576802e-05, + "loss": 0.9798, + "step": 27932 + }, + { + "epoch": 2.06, + "learning_rate": 1.5259173906226446e-05, + "loss": 1.1359, + "step": 27933 + }, + { + "epoch": 2.06, + "learning_rate": 1.525883464750882e-05, + "loss": 0.9887, + "step": 27934 + }, + { + "epoch": 2.06, + "learning_rate": 1.5258495380424463e-05, + "loss": 1.0081, + "step": 27935 + }, + { + "epoch": 2.06, + "learning_rate": 1.5258156104973915e-05, + "loss": 0.9668, + "step": 27936 + }, + { + "epoch": 2.06, + "learning_rate": 1.5257816821157714e-05, + "loss": 1.1132, + "step": 27937 + }, + { + "epoch": 2.06, + "learning_rate": 1.5257477528976405e-05, + "loss": 0.9702, + "step": 27938 + }, + { + "epoch": 2.06, + "learning_rate": 1.5257138228430524e-05, + "loss": 0.9702, + "step": 27939 + }, + { + "epoch": 2.06, + "learning_rate": 1.525679891952061e-05, + "loss": 0.911, + "step": 27940 + }, + { + "epoch": 2.06, + "learning_rate": 1.5256459602247207e-05, + "loss": 0.9217, + "step": 27941 + }, + { + "epoch": 2.06, + "learning_rate": 1.5256120276610848e-05, + "loss": 0.9707, + "step": 27942 + }, + { + "epoch": 2.06, + "learning_rate": 1.5255780942612078e-05, + "loss": 0.9813, + "step": 27943 + }, + { + "epoch": 2.06, + "learning_rate": 1.5255441600251435e-05, + "loss": 1.0603, + "step": 27944 + }, + { + "epoch": 2.06, + "learning_rate": 1.5255102249529463e-05, + "loss": 1.0556, + "step": 27945 + }, + { + "epoch": 2.06, + "learning_rate": 1.5254762890446695e-05, + "loss": 0.9758, + "step": 27946 + }, + { + "epoch": 2.06, + "learning_rate": 1.5254423523003677e-05, + "loss": 1.0093, + "step": 27947 + }, + { + "epoch": 2.07, + "learning_rate": 1.5254084147200945e-05, + "loss": 1.0541, + "step": 27948 + }, + { + "epoch": 2.07, + "learning_rate": 1.5253744763039037e-05, + "loss": 1.0104, + "step": 27949 + }, + { + "epoch": 2.07, + "learning_rate": 1.5253405370518502e-05, + "loss": 0.9543, + "step": 27950 + }, + { + "epoch": 2.07, + "learning_rate": 1.5253065969639873e-05, + "loss": 1.0235, + "step": 27951 + }, + { + "epoch": 2.07, + "learning_rate": 1.525272656040369e-05, + "loss": 1.009, + "step": 27952 + }, + { + "epoch": 2.07, + "learning_rate": 1.5252387142810495e-05, + "loss": 0.8599, + "step": 27953 + }, + { + "epoch": 2.07, + "learning_rate": 1.5252047716860825e-05, + "loss": 1.0158, + "step": 27954 + }, + { + "epoch": 2.07, + "learning_rate": 1.5251708282555224e-05, + "loss": 0.9105, + "step": 27955 + }, + { + "epoch": 2.07, + "learning_rate": 1.525136883989423e-05, + "loss": 1.0039, + "step": 27956 + }, + { + "epoch": 2.07, + "learning_rate": 1.5251029388878382e-05, + "loss": 1.1428, + "step": 27957 + }, + { + "epoch": 2.07, + "learning_rate": 1.5250689929508223e-05, + "loss": 1.1165, + "step": 27958 + }, + { + "epoch": 2.07, + "learning_rate": 1.5250350461784291e-05, + "loss": 1.033, + "step": 27959 + }, + { + "epoch": 2.07, + "learning_rate": 1.5250010985707129e-05, + "loss": 0.9112, + "step": 27960 + }, + { + "epoch": 2.07, + "learning_rate": 1.5249671501277273e-05, + "loss": 0.9412, + "step": 27961 + }, + { + "epoch": 2.07, + "learning_rate": 1.5249332008495263e-05, + "loss": 0.9876, + "step": 27962 + }, + { + "epoch": 2.07, + "learning_rate": 1.5248992507361642e-05, + "loss": 0.9329, + "step": 27963 + }, + { + "epoch": 2.07, + "learning_rate": 1.524865299787695e-05, + "loss": 0.9882, + "step": 27964 + }, + { + "epoch": 2.07, + "learning_rate": 1.5248313480041725e-05, + "loss": 1.0594, + "step": 27965 + }, + { + "epoch": 2.07, + "learning_rate": 1.5247973953856506e-05, + "loss": 1.005, + "step": 27966 + }, + { + "epoch": 2.07, + "learning_rate": 1.524763441932184e-05, + "loss": 1.0007, + "step": 27967 + }, + { + "epoch": 2.07, + "learning_rate": 1.524729487643826e-05, + "loss": 0.9405, + "step": 27968 + }, + { + "epoch": 2.07, + "learning_rate": 1.5246955325206311e-05, + "loss": 0.9536, + "step": 27969 + }, + { + "epoch": 2.07, + "learning_rate": 1.5246615765626528e-05, + "loss": 1.0311, + "step": 27970 + }, + { + "epoch": 2.07, + "learning_rate": 1.524627619769946e-05, + "loss": 0.9724, + "step": 27971 + }, + { + "epoch": 2.07, + "learning_rate": 1.5245936621425635e-05, + "loss": 1.0142, + "step": 27972 + }, + { + "epoch": 2.07, + "learning_rate": 1.5245597036805601e-05, + "loss": 0.925, + "step": 27973 + }, + { + "epoch": 2.07, + "learning_rate": 1.5245257443839898e-05, + "loss": 0.9125, + "step": 27974 + }, + { + "epoch": 2.07, + "learning_rate": 1.524491784252907e-05, + "loss": 1.0407, + "step": 27975 + }, + { + "epoch": 2.07, + "learning_rate": 1.5244578232873646e-05, + "loss": 1.0204, + "step": 27976 + }, + { + "epoch": 2.07, + "learning_rate": 1.5244238614874175e-05, + "loss": 1.0417, + "step": 27977 + }, + { + "epoch": 2.07, + "learning_rate": 1.5243898988531198e-05, + "loss": 0.9349, + "step": 27978 + }, + { + "epoch": 2.07, + "learning_rate": 1.5243559353845254e-05, + "loss": 1.0633, + "step": 27979 + }, + { + "epoch": 2.07, + "learning_rate": 1.5243219710816877e-05, + "loss": 1.003, + "step": 27980 + }, + { + "epoch": 2.07, + "learning_rate": 1.5242880059446617e-05, + "loss": 1.0426, + "step": 27981 + }, + { + "epoch": 2.07, + "learning_rate": 1.5242540399735007e-05, + "loss": 1.0413, + "step": 27982 + }, + { + "epoch": 2.07, + "learning_rate": 1.524220073168259e-05, + "loss": 1.0358, + "step": 27983 + }, + { + "epoch": 2.07, + "learning_rate": 1.5241861055289908e-05, + "loss": 1.0517, + "step": 27984 + }, + { + "epoch": 2.07, + "learning_rate": 1.52415213705575e-05, + "loss": 1.0349, + "step": 27985 + }, + { + "epoch": 2.07, + "learning_rate": 1.5241181677485903e-05, + "loss": 1.0811, + "step": 27986 + }, + { + "epoch": 2.07, + "learning_rate": 1.5240841976075666e-05, + "loss": 1.0072, + "step": 27987 + }, + { + "epoch": 2.07, + "learning_rate": 1.5240502266327324e-05, + "loss": 0.945, + "step": 27988 + }, + { + "epoch": 2.07, + "learning_rate": 1.5240162548241416e-05, + "loss": 0.9922, + "step": 27989 + }, + { + "epoch": 2.07, + "learning_rate": 1.5239822821818484e-05, + "loss": 1.1341, + "step": 27990 + }, + { + "epoch": 2.07, + "learning_rate": 1.5239483087059068e-05, + "loss": 1.0399, + "step": 27991 + }, + { + "epoch": 2.07, + "learning_rate": 1.5239143343963712e-05, + "loss": 0.9733, + "step": 27992 + }, + { + "epoch": 2.07, + "learning_rate": 1.5238803592532954e-05, + "loss": 1.0688, + "step": 27993 + }, + { + "epoch": 2.07, + "learning_rate": 1.5238463832767332e-05, + "loss": 1.1001, + "step": 27994 + }, + { + "epoch": 2.07, + "learning_rate": 1.523812406466739e-05, + "loss": 1.1116, + "step": 27995 + }, + { + "epoch": 2.07, + "learning_rate": 1.5237784288233668e-05, + "loss": 0.9709, + "step": 27996 + }, + { + "epoch": 2.07, + "learning_rate": 1.5237444503466705e-05, + "loss": 1.0314, + "step": 27997 + }, + { + "epoch": 2.07, + "learning_rate": 1.5237104710367045e-05, + "loss": 0.9221, + "step": 27998 + }, + { + "epoch": 2.07, + "learning_rate": 1.5236764908935224e-05, + "loss": 1.04, + "step": 27999 + }, + { + "epoch": 2.07, + "learning_rate": 1.5236425099171788e-05, + "loss": 1.0977, + "step": 28000 + }, + { + "epoch": 2.07, + "learning_rate": 1.523608528107727e-05, + "loss": 1.0708, + "step": 28001 + }, + { + "epoch": 2.07, + "learning_rate": 1.523574545465222e-05, + "loss": 1.1207, + "step": 28002 + }, + { + "epoch": 2.07, + "learning_rate": 1.523540561989717e-05, + "loss": 1.0268, + "step": 28003 + }, + { + "epoch": 2.07, + "learning_rate": 1.5235065776812668e-05, + "loss": 1.0773, + "step": 28004 + }, + { + "epoch": 2.07, + "learning_rate": 1.5234725925399248e-05, + "loss": 1.0068, + "step": 28005 + }, + { + "epoch": 2.07, + "learning_rate": 1.5234386065657453e-05, + "loss": 1.0204, + "step": 28006 + }, + { + "epoch": 2.07, + "learning_rate": 1.523404619758783e-05, + "loss": 0.9521, + "step": 28007 + }, + { + "epoch": 2.07, + "learning_rate": 1.5233706321190913e-05, + "loss": 0.9749, + "step": 28008 + }, + { + "epoch": 2.07, + "learning_rate": 1.523336643646724e-05, + "loss": 0.9892, + "step": 28009 + }, + { + "epoch": 2.07, + "learning_rate": 1.523302654341736e-05, + "loss": 0.9794, + "step": 28010 + }, + { + "epoch": 2.07, + "learning_rate": 1.5232686642041808e-05, + "loss": 0.9763, + "step": 28011 + }, + { + "epoch": 2.07, + "learning_rate": 1.5232346732341126e-05, + "loss": 1.0141, + "step": 28012 + }, + { + "epoch": 2.07, + "learning_rate": 1.5232006814315856e-05, + "loss": 0.9566, + "step": 28013 + }, + { + "epoch": 2.07, + "learning_rate": 1.5231666887966538e-05, + "loss": 1.0658, + "step": 28014 + }, + { + "epoch": 2.07, + "learning_rate": 1.523132695329371e-05, + "loss": 0.9415, + "step": 28015 + }, + { + "epoch": 2.07, + "learning_rate": 1.5230987010297921e-05, + "loss": 0.9976, + "step": 28016 + }, + { + "epoch": 2.07, + "learning_rate": 1.5230647058979703e-05, + "loss": 1.0526, + "step": 28017 + }, + { + "epoch": 2.07, + "learning_rate": 1.5230307099339604e-05, + "loss": 1.011, + "step": 28018 + }, + { + "epoch": 2.07, + "learning_rate": 1.5229967131378155e-05, + "loss": 0.8928, + "step": 28019 + }, + { + "epoch": 2.07, + "learning_rate": 1.522962715509591e-05, + "loss": 1.0586, + "step": 28020 + }, + { + "epoch": 2.07, + "learning_rate": 1.5229287170493397e-05, + "loss": 0.9329, + "step": 28021 + }, + { + "epoch": 2.07, + "learning_rate": 1.5228947177571165e-05, + "loss": 1.1124, + "step": 28022 + }, + { + "epoch": 2.07, + "learning_rate": 1.5228607176329753e-05, + "loss": 1.0231, + "step": 28023 + }, + { + "epoch": 2.07, + "learning_rate": 1.52282671667697e-05, + "loss": 1.0404, + "step": 28024 + }, + { + "epoch": 2.07, + "learning_rate": 1.5227927148891554e-05, + "loss": 1.056, + "step": 28025 + }, + { + "epoch": 2.07, + "learning_rate": 1.5227587122695849e-05, + "loss": 0.9419, + "step": 28026 + }, + { + "epoch": 2.07, + "learning_rate": 1.5227247088183124e-05, + "loss": 1.0306, + "step": 28027 + }, + { + "epoch": 2.07, + "learning_rate": 1.5226907045353928e-05, + "loss": 0.9399, + "step": 28028 + }, + { + "epoch": 2.07, + "learning_rate": 1.5226566994208796e-05, + "loss": 0.9257, + "step": 28029 + }, + { + "epoch": 2.07, + "learning_rate": 1.5226226934748269e-05, + "loss": 0.9773, + "step": 28030 + }, + { + "epoch": 2.07, + "learning_rate": 1.522588686697289e-05, + "loss": 0.9694, + "step": 28031 + }, + { + "epoch": 2.07, + "learning_rate": 1.52255467908832e-05, + "loss": 0.9669, + "step": 28032 + }, + { + "epoch": 2.07, + "learning_rate": 1.522520670647974e-05, + "loss": 0.9502, + "step": 28033 + }, + { + "epoch": 2.07, + "learning_rate": 1.5224866613763053e-05, + "loss": 1.0251, + "step": 28034 + }, + { + "epoch": 2.07, + "learning_rate": 1.5224526512733677e-05, + "loss": 1.0117, + "step": 28035 + }, + { + "epoch": 2.07, + "learning_rate": 1.5224186403392152e-05, + "loss": 0.9177, + "step": 28036 + }, + { + "epoch": 2.07, + "learning_rate": 1.5223846285739025e-05, + "loss": 0.9376, + "step": 28037 + }, + { + "epoch": 2.07, + "learning_rate": 1.5223506159774833e-05, + "loss": 1.0339, + "step": 28038 + }, + { + "epoch": 2.07, + "learning_rate": 1.5223166025500114e-05, + "loss": 0.9545, + "step": 28039 + }, + { + "epoch": 2.07, + "learning_rate": 1.5222825882915414e-05, + "loss": 1.0134, + "step": 28040 + }, + { + "epoch": 2.07, + "learning_rate": 1.522248573202127e-05, + "loss": 0.997, + "step": 28041 + }, + { + "epoch": 2.07, + "learning_rate": 1.522214557281823e-05, + "loss": 1.0353, + "step": 28042 + }, + { + "epoch": 2.07, + "learning_rate": 1.5221805405306828e-05, + "loss": 0.9743, + "step": 28043 + }, + { + "epoch": 2.07, + "learning_rate": 1.5221465229487611e-05, + "loss": 1.0366, + "step": 28044 + }, + { + "epoch": 2.07, + "learning_rate": 1.5221125045361119e-05, + "loss": 1.0066, + "step": 28045 + }, + { + "epoch": 2.07, + "learning_rate": 1.5220784852927888e-05, + "loss": 0.9736, + "step": 28046 + }, + { + "epoch": 2.07, + "learning_rate": 1.5220444652188464e-05, + "loss": 1.078, + "step": 28047 + }, + { + "epoch": 2.07, + "learning_rate": 1.5220104443143387e-05, + "loss": 0.9382, + "step": 28048 + }, + { + "epoch": 2.07, + "learning_rate": 1.5219764225793201e-05, + "loss": 1.0295, + "step": 28049 + }, + { + "epoch": 2.07, + "learning_rate": 1.521942400013844e-05, + "loss": 0.9896, + "step": 28050 + }, + { + "epoch": 2.07, + "learning_rate": 1.5219083766179654e-05, + "loss": 1.0498, + "step": 28051 + }, + { + "epoch": 2.07, + "learning_rate": 1.5218743523917379e-05, + "loss": 1.0579, + "step": 28052 + }, + { + "epoch": 2.07, + "learning_rate": 1.521840327335216e-05, + "loss": 1.0052, + "step": 28053 + }, + { + "epoch": 2.07, + "learning_rate": 1.5218063014484534e-05, + "loss": 1.0118, + "step": 28054 + }, + { + "epoch": 2.07, + "learning_rate": 1.5217722747315047e-05, + "loss": 0.9896, + "step": 28055 + }, + { + "epoch": 2.07, + "learning_rate": 1.5217382471844233e-05, + "loss": 1.0165, + "step": 28056 + }, + { + "epoch": 2.07, + "learning_rate": 1.5217042188072641e-05, + "loss": 0.9439, + "step": 28057 + }, + { + "epoch": 2.07, + "learning_rate": 1.5216701896000807e-05, + "loss": 1.0307, + "step": 28058 + }, + { + "epoch": 2.07, + "learning_rate": 1.5216361595629278e-05, + "loss": 1.0381, + "step": 28059 + }, + { + "epoch": 2.07, + "learning_rate": 1.5216021286958591e-05, + "loss": 1.05, + "step": 28060 + }, + { + "epoch": 2.07, + "learning_rate": 1.5215680969989288e-05, + "loss": 0.9302, + "step": 28061 + }, + { + "epoch": 2.07, + "learning_rate": 1.5215340644721912e-05, + "loss": 0.9875, + "step": 28062 + }, + { + "epoch": 2.07, + "learning_rate": 1.5215000311157007e-05, + "loss": 1.0113, + "step": 28063 + }, + { + "epoch": 2.07, + "learning_rate": 1.5214659969295107e-05, + "loss": 1.0729, + "step": 28064 + }, + { + "epoch": 2.07, + "learning_rate": 1.5214319619136759e-05, + "loss": 0.9627, + "step": 28065 + }, + { + "epoch": 2.07, + "learning_rate": 1.52139792606825e-05, + "loss": 1.0622, + "step": 28066 + }, + { + "epoch": 2.07, + "learning_rate": 1.521363889393288e-05, + "loss": 0.9827, + "step": 28067 + }, + { + "epoch": 2.07, + "learning_rate": 1.5213298518888433e-05, + "loss": 1.0569, + "step": 28068 + }, + { + "epoch": 2.07, + "learning_rate": 1.52129581355497e-05, + "loss": 0.9208, + "step": 28069 + }, + { + "epoch": 2.07, + "learning_rate": 1.5212617743917229e-05, + "loss": 0.937, + "step": 28070 + }, + { + "epoch": 2.07, + "learning_rate": 1.5212277343991554e-05, + "loss": 0.9801, + "step": 28071 + }, + { + "epoch": 2.07, + "learning_rate": 1.5211936935773224e-05, + "loss": 1.0125, + "step": 28072 + }, + { + "epoch": 2.07, + "learning_rate": 1.5211596519262777e-05, + "loss": 1.071, + "step": 28073 + }, + { + "epoch": 2.07, + "learning_rate": 1.5211256094460751e-05, + "loss": 0.8516, + "step": 28074 + }, + { + "epoch": 2.07, + "learning_rate": 1.5210915661367694e-05, + "loss": 1.0712, + "step": 28075 + }, + { + "epoch": 2.07, + "learning_rate": 1.5210575219984143e-05, + "loss": 1.0162, + "step": 28076 + }, + { + "epoch": 2.07, + "learning_rate": 1.5210234770310641e-05, + "loss": 0.9478, + "step": 28077 + }, + { + "epoch": 2.07, + "learning_rate": 1.5209894312347731e-05, + "loss": 1.0569, + "step": 28078 + }, + { + "epoch": 2.07, + "learning_rate": 1.5209553846095955e-05, + "loss": 0.9839, + "step": 28079 + }, + { + "epoch": 2.07, + "learning_rate": 1.5209213371555848e-05, + "loss": 0.8902, + "step": 28080 + }, + { + "epoch": 2.07, + "learning_rate": 1.5208872888727966e-05, + "loss": 1.0346, + "step": 28081 + }, + { + "epoch": 2.07, + "learning_rate": 1.5208532397612837e-05, + "loss": 0.9888, + "step": 28082 + }, + { + "epoch": 2.08, + "learning_rate": 1.5208191898211006e-05, + "loss": 1.0412, + "step": 28083 + }, + { + "epoch": 2.08, + "learning_rate": 1.5207851390523019e-05, + "loss": 0.962, + "step": 28084 + }, + { + "epoch": 2.08, + "learning_rate": 1.5207510874549414e-05, + "loss": 0.9618, + "step": 28085 + }, + { + "epoch": 2.08, + "learning_rate": 1.520717035029073e-05, + "loss": 0.9286, + "step": 28086 + }, + { + "epoch": 2.08, + "learning_rate": 1.5206829817747517e-05, + "loss": 1.0629, + "step": 28087 + }, + { + "epoch": 2.08, + "learning_rate": 1.520648927692031e-05, + "loss": 1.066, + "step": 28088 + }, + { + "epoch": 2.08, + "learning_rate": 1.5206148727809656e-05, + "loss": 1.0159, + "step": 28089 + }, + { + "epoch": 2.08, + "learning_rate": 1.520580817041609e-05, + "loss": 1.0511, + "step": 28090 + }, + { + "epoch": 2.08, + "learning_rate": 1.5205467604740159e-05, + "loss": 0.9967, + "step": 28091 + }, + { + "epoch": 2.08, + "learning_rate": 1.5205127030782407e-05, + "loss": 0.9822, + "step": 28092 + }, + { + "epoch": 2.08, + "learning_rate": 1.5204786448543367e-05, + "loss": 1.0504, + "step": 28093 + }, + { + "epoch": 2.08, + "learning_rate": 1.520444585802359e-05, + "loss": 1.041, + "step": 28094 + }, + { + "epoch": 2.08, + "learning_rate": 1.5204105259223611e-05, + "loss": 1.0605, + "step": 28095 + }, + { + "epoch": 2.08, + "learning_rate": 1.5203764652143978e-05, + "loss": 0.9595, + "step": 28096 + }, + { + "epoch": 2.08, + "learning_rate": 1.5203424036785226e-05, + "loss": 0.9821, + "step": 28097 + }, + { + "epoch": 2.08, + "learning_rate": 1.5203083413147905e-05, + "loss": 0.9454, + "step": 28098 + }, + { + "epoch": 2.08, + "learning_rate": 1.5202742781232554e-05, + "loss": 1.0408, + "step": 28099 + }, + { + "epoch": 2.08, + "learning_rate": 1.520240214103971e-05, + "loss": 0.9655, + "step": 28100 + }, + { + "epoch": 2.08, + "learning_rate": 1.5202061492569918e-05, + "loss": 1.0519, + "step": 28101 + }, + { + "epoch": 2.08, + "learning_rate": 1.5201720835823723e-05, + "loss": 0.9559, + "step": 28102 + }, + { + "epoch": 2.08, + "learning_rate": 1.5201380170801663e-05, + "loss": 0.9616, + "step": 28103 + }, + { + "epoch": 2.08, + "learning_rate": 1.5201039497504284e-05, + "loss": 0.9256, + "step": 28104 + }, + { + "epoch": 2.08, + "learning_rate": 1.5200698815932123e-05, + "loss": 0.9136, + "step": 28105 + }, + { + "epoch": 2.08, + "learning_rate": 1.5200358126085726e-05, + "loss": 1.0138, + "step": 28106 + }, + { + "epoch": 2.08, + "learning_rate": 1.5200017427965633e-05, + "loss": 0.9697, + "step": 28107 + }, + { + "epoch": 2.08, + "learning_rate": 1.5199676721572384e-05, + "loss": 0.994, + "step": 28108 + }, + { + "epoch": 2.08, + "learning_rate": 1.519933600690653e-05, + "loss": 1.0452, + "step": 28109 + }, + { + "epoch": 2.08, + "learning_rate": 1.5198995283968603e-05, + "loss": 0.9291, + "step": 28110 + }, + { + "epoch": 2.08, + "learning_rate": 1.5198654552759148e-05, + "loss": 0.9887, + "step": 28111 + }, + { + "epoch": 2.08, + "learning_rate": 1.5198313813278708e-05, + "loss": 1.0043, + "step": 28112 + }, + { + "epoch": 2.08, + "learning_rate": 1.5197973065527824e-05, + "loss": 1.079, + "step": 28113 + }, + { + "epoch": 2.08, + "learning_rate": 1.5197632309507044e-05, + "loss": 1.0609, + "step": 28114 + }, + { + "epoch": 2.08, + "learning_rate": 1.5197291545216902e-05, + "loss": 1.0213, + "step": 28115 + }, + { + "epoch": 2.08, + "learning_rate": 1.5196950772657946e-05, + "loss": 1.0644, + "step": 28116 + }, + { + "epoch": 2.08, + "learning_rate": 1.5196609991830711e-05, + "loss": 1.0742, + "step": 28117 + }, + { + "epoch": 2.08, + "learning_rate": 1.519626920273575e-05, + "loss": 0.9343, + "step": 28118 + }, + { + "epoch": 2.08, + "learning_rate": 1.5195928405373597e-05, + "loss": 1.0625, + "step": 28119 + }, + { + "epoch": 2.08, + "learning_rate": 1.5195587599744794e-05, + "loss": 1.0007, + "step": 28120 + }, + { + "epoch": 2.08, + "learning_rate": 1.5195246785849884e-05, + "loss": 1.132, + "step": 28121 + }, + { + "epoch": 2.08, + "learning_rate": 1.5194905963689418e-05, + "loss": 1.0461, + "step": 28122 + }, + { + "epoch": 2.08, + "learning_rate": 1.5194565133263923e-05, + "loss": 0.9159, + "step": 28123 + }, + { + "epoch": 2.08, + "learning_rate": 1.519422429457395e-05, + "loss": 1.0014, + "step": 28124 + }, + { + "epoch": 2.08, + "learning_rate": 1.5193883447620045e-05, + "loss": 1.0134, + "step": 28125 + }, + { + "epoch": 2.08, + "learning_rate": 1.5193542592402745e-05, + "loss": 0.9529, + "step": 28126 + }, + { + "epoch": 2.08, + "learning_rate": 1.5193201728922588e-05, + "loss": 1.0245, + "step": 28127 + }, + { + "epoch": 2.08, + "learning_rate": 1.519286085718013e-05, + "loss": 0.9951, + "step": 28128 + }, + { + "epoch": 2.08, + "learning_rate": 1.5192519977175896e-05, + "loss": 0.9632, + "step": 28129 + }, + { + "epoch": 2.08, + "learning_rate": 1.5192179088910444e-05, + "loss": 1.0194, + "step": 28130 + }, + { + "epoch": 2.08, + "learning_rate": 1.5191838192384304e-05, + "loss": 0.9101, + "step": 28131 + }, + { + "epoch": 2.08, + "learning_rate": 1.5191497287598027e-05, + "loss": 1.0121, + "step": 28132 + }, + { + "epoch": 2.08, + "learning_rate": 1.519115637455215e-05, + "loss": 0.9749, + "step": 28133 + }, + { + "epoch": 2.08, + "learning_rate": 1.5190815453247218e-05, + "loss": 1.0757, + "step": 28134 + }, + { + "epoch": 2.08, + "learning_rate": 1.5190474523683773e-05, + "loss": 0.8869, + "step": 28135 + }, + { + "epoch": 2.08, + "learning_rate": 1.5190133585862358e-05, + "loss": 0.9724, + "step": 28136 + }, + { + "epoch": 2.08, + "learning_rate": 1.5189792639783516e-05, + "loss": 1.0733, + "step": 28137 + }, + { + "epoch": 2.08, + "learning_rate": 1.5189451685447784e-05, + "loss": 1.053, + "step": 28138 + }, + { + "epoch": 2.08, + "learning_rate": 1.518911072285571e-05, + "loss": 1.0592, + "step": 28139 + }, + { + "epoch": 2.08, + "learning_rate": 1.5188769752007839e-05, + "loss": 1.0705, + "step": 28140 + }, + { + "epoch": 2.08, + "learning_rate": 1.5188428772904706e-05, + "loss": 1.0658, + "step": 28141 + }, + { + "epoch": 2.08, + "learning_rate": 1.5188087785546857e-05, + "loss": 1.0202, + "step": 28142 + }, + { + "epoch": 2.08, + "learning_rate": 1.5187746789934837e-05, + "loss": 0.9175, + "step": 28143 + }, + { + "epoch": 2.08, + "learning_rate": 1.5187405786069182e-05, + "loss": 1.0437, + "step": 28144 + }, + { + "epoch": 2.08, + "learning_rate": 1.5187064773950442e-05, + "loss": 0.9049, + "step": 28145 + }, + { + "epoch": 2.08, + "learning_rate": 1.5186723753579156e-05, + "loss": 0.9904, + "step": 28146 + }, + { + "epoch": 2.08, + "learning_rate": 1.5186382724955865e-05, + "loss": 1.0639, + "step": 28147 + }, + { + "epoch": 2.08, + "learning_rate": 1.5186041688081115e-05, + "loss": 1.1278, + "step": 28148 + }, + { + "epoch": 2.08, + "learning_rate": 1.5185700642955445e-05, + "loss": 0.9536, + "step": 28149 + }, + { + "epoch": 2.08, + "learning_rate": 1.51853595895794e-05, + "loss": 1.0733, + "step": 28150 + }, + { + "epoch": 2.08, + "learning_rate": 1.5185018527953524e-05, + "loss": 0.9524, + "step": 28151 + }, + { + "epoch": 2.08, + "learning_rate": 1.5184677458078355e-05, + "loss": 1.0531, + "step": 28152 + }, + { + "epoch": 2.08, + "learning_rate": 1.5184336379954441e-05, + "loss": 0.925, + "step": 28153 + }, + { + "epoch": 2.08, + "learning_rate": 1.518399529358232e-05, + "loss": 0.9695, + "step": 28154 + }, + { + "epoch": 2.08, + "learning_rate": 1.5183654198962536e-05, + "loss": 0.9847, + "step": 28155 + }, + { + "epoch": 2.08, + "learning_rate": 1.5183313096095635e-05, + "loss": 1.0038, + "step": 28156 + }, + { + "epoch": 2.08, + "learning_rate": 1.5182971984982159e-05, + "loss": 0.9313, + "step": 28157 + }, + { + "epoch": 2.08, + "learning_rate": 1.5182630865622644e-05, + "loss": 0.9556, + "step": 28158 + }, + { + "epoch": 2.08, + "learning_rate": 1.5182289738017639e-05, + "loss": 0.9333, + "step": 28159 + }, + { + "epoch": 2.08, + "learning_rate": 1.5181948602167685e-05, + "loss": 1.0993, + "step": 28160 + }, + { + "epoch": 2.08, + "learning_rate": 1.5181607458073324e-05, + "loss": 1.0535, + "step": 28161 + }, + { + "epoch": 2.08, + "learning_rate": 1.51812663057351e-05, + "loss": 0.8593, + "step": 28162 + }, + { + "epoch": 2.08, + "learning_rate": 1.5180925145153555e-05, + "loss": 1.022, + "step": 28163 + }, + { + "epoch": 2.08, + "learning_rate": 1.5180583976329234e-05, + "loss": 0.981, + "step": 28164 + }, + { + "epoch": 2.08, + "learning_rate": 1.5180242799262678e-05, + "loss": 1.1345, + "step": 28165 + }, + { + "epoch": 2.08, + "learning_rate": 1.5179901613954428e-05, + "loss": 1.0569, + "step": 28166 + }, + { + "epoch": 2.08, + "learning_rate": 1.517956042040503e-05, + "loss": 0.9901, + "step": 28167 + }, + { + "epoch": 2.08, + "learning_rate": 1.5179219218615027e-05, + "loss": 1.1266, + "step": 28168 + }, + { + "epoch": 2.08, + "learning_rate": 1.5178878008584958e-05, + "loss": 0.872, + "step": 28169 + }, + { + "epoch": 2.08, + "learning_rate": 1.5178536790315368e-05, + "loss": 1.053, + "step": 28170 + }, + { + "epoch": 2.08, + "learning_rate": 1.5178195563806801e-05, + "loss": 1.0493, + "step": 28171 + }, + { + "epoch": 2.08, + "learning_rate": 1.5177854329059798e-05, + "loss": 0.9982, + "step": 28172 + }, + { + "epoch": 2.08, + "learning_rate": 1.5177513086074904e-05, + "loss": 1.0953, + "step": 28173 + }, + { + "epoch": 2.08, + "learning_rate": 1.517717183485266e-05, + "loss": 1.0064, + "step": 28174 + }, + { + "epoch": 2.08, + "learning_rate": 1.5176830575393613e-05, + "loss": 0.9205, + "step": 28175 + }, + { + "epoch": 2.08, + "learning_rate": 1.5176489307698297e-05, + "loss": 1.1185, + "step": 28176 + }, + { + "epoch": 2.08, + "learning_rate": 1.5176148031767266e-05, + "loss": 0.8797, + "step": 28177 + }, + { + "epoch": 2.08, + "learning_rate": 1.5175806747601052e-05, + "loss": 0.9505, + "step": 28178 + }, + { + "epoch": 2.08, + "learning_rate": 1.5175465455200207e-05, + "loss": 1.0064, + "step": 28179 + }, + { + "epoch": 2.08, + "learning_rate": 1.517512415456527e-05, + "loss": 1.092, + "step": 28180 + }, + { + "epoch": 2.08, + "learning_rate": 1.5174782845696783e-05, + "loss": 1.0919, + "step": 28181 + }, + { + "epoch": 2.08, + "learning_rate": 1.5174441528595292e-05, + "loss": 0.9891, + "step": 28182 + }, + { + "epoch": 2.08, + "learning_rate": 1.5174100203261339e-05, + "loss": 1.1535, + "step": 28183 + }, + { + "epoch": 2.08, + "learning_rate": 1.5173758869695467e-05, + "loss": 0.9823, + "step": 28184 + }, + { + "epoch": 2.08, + "learning_rate": 1.5173417527898218e-05, + "loss": 0.9679, + "step": 28185 + }, + { + "epoch": 2.08, + "learning_rate": 1.5173076177870134e-05, + "loss": 1.0021, + "step": 28186 + }, + { + "epoch": 2.08, + "learning_rate": 1.5172734819611763e-05, + "loss": 0.9279, + "step": 28187 + }, + { + "epoch": 2.08, + "learning_rate": 1.5172393453123641e-05, + "loss": 0.9122, + "step": 28188 + }, + { + "epoch": 2.08, + "learning_rate": 1.5172052078406318e-05, + "loss": 1.0308, + "step": 28189 + }, + { + "epoch": 2.08, + "learning_rate": 1.5171710695460333e-05, + "loss": 1.125, + "step": 28190 + }, + { + "epoch": 2.08, + "learning_rate": 1.5171369304286229e-05, + "loss": 1.0408, + "step": 28191 + }, + { + "epoch": 2.08, + "learning_rate": 1.5171027904884552e-05, + "loss": 0.8573, + "step": 28192 + }, + { + "epoch": 2.08, + "learning_rate": 1.5170686497255847e-05, + "loss": 0.9768, + "step": 28193 + }, + { + "epoch": 2.08, + "learning_rate": 1.517034508140065e-05, + "loss": 0.9581, + "step": 28194 + }, + { + "epoch": 2.08, + "learning_rate": 1.5170003657319506e-05, + "loss": 1.0131, + "step": 28195 + }, + { + "epoch": 2.08, + "learning_rate": 1.5169662225012962e-05, + "loss": 1.052, + "step": 28196 + }, + { + "epoch": 2.08, + "learning_rate": 1.516932078448156e-05, + "loss": 0.8824, + "step": 28197 + }, + { + "epoch": 2.08, + "learning_rate": 1.5168979335725845e-05, + "loss": 1.0343, + "step": 28198 + }, + { + "epoch": 2.08, + "learning_rate": 1.5168637878746352e-05, + "loss": 0.8992, + "step": 28199 + }, + { + "epoch": 2.08, + "learning_rate": 1.5168296413543636e-05, + "loss": 1.0548, + "step": 28200 + }, + { + "epoch": 2.08, + "learning_rate": 1.5167954940118228e-05, + "loss": 1.0342, + "step": 28201 + }, + { + "epoch": 2.08, + "learning_rate": 1.5167613458470683e-05, + "loss": 0.9152, + "step": 28202 + }, + { + "epoch": 2.08, + "learning_rate": 1.516727196860154e-05, + "loss": 1.0149, + "step": 28203 + }, + { + "epoch": 2.08, + "learning_rate": 1.5166930470511338e-05, + "loss": 1.0464, + "step": 28204 + }, + { + "epoch": 2.08, + "learning_rate": 1.5166588964200623e-05, + "loss": 1.1126, + "step": 28205 + }, + { + "epoch": 2.08, + "learning_rate": 1.5166247449669941e-05, + "loss": 1.0562, + "step": 28206 + }, + { + "epoch": 2.08, + "learning_rate": 1.516590592691983e-05, + "loss": 1.0737, + "step": 28207 + }, + { + "epoch": 2.08, + "learning_rate": 1.5165564395950841e-05, + "loss": 1.012, + "step": 28208 + }, + { + "epoch": 2.08, + "learning_rate": 1.5165222856763508e-05, + "loss": 1.0416, + "step": 28209 + }, + { + "epoch": 2.08, + "learning_rate": 1.5164881309358383e-05, + "loss": 1.0367, + "step": 28210 + }, + { + "epoch": 2.08, + "learning_rate": 1.5164539753736005e-05, + "loss": 0.966, + "step": 28211 + }, + { + "epoch": 2.08, + "learning_rate": 1.5164198189896919e-05, + "loss": 1.0417, + "step": 28212 + }, + { + "epoch": 2.08, + "learning_rate": 1.5163856617841666e-05, + "loss": 1.0978, + "step": 28213 + }, + { + "epoch": 2.08, + "learning_rate": 1.5163515037570791e-05, + "loss": 1.1654, + "step": 28214 + }, + { + "epoch": 2.08, + "learning_rate": 1.5163173449084838e-05, + "loss": 0.9798, + "step": 28215 + }, + { + "epoch": 2.08, + "learning_rate": 1.516283185238435e-05, + "loss": 0.8285, + "step": 28216 + }, + { + "epoch": 2.08, + "learning_rate": 1.516249024746987e-05, + "loss": 0.9787, + "step": 28217 + }, + { + "epoch": 2.09, + "learning_rate": 1.5162148634341942e-05, + "loss": 1.0632, + "step": 28218 + }, + { + "epoch": 2.09, + "learning_rate": 1.5161807013001109e-05, + "loss": 0.9475, + "step": 28219 + }, + { + "epoch": 2.09, + "learning_rate": 1.5161465383447915e-05, + "loss": 0.9239, + "step": 28220 + }, + { + "epoch": 2.09, + "learning_rate": 1.5161123745682905e-05, + "loss": 0.9863, + "step": 28221 + }, + { + "epoch": 2.09, + "learning_rate": 1.5160782099706621e-05, + "loss": 1.0086, + "step": 28222 + }, + { + "epoch": 2.09, + "learning_rate": 1.5160440445519604e-05, + "loss": 1.1022, + "step": 28223 + }, + { + "epoch": 2.09, + "learning_rate": 1.5160098783122404e-05, + "loss": 0.9994, + "step": 28224 + }, + { + "epoch": 2.09, + "learning_rate": 1.5159757112515556e-05, + "loss": 1.0717, + "step": 28225 + }, + { + "epoch": 2.09, + "learning_rate": 1.5159415433699609e-05, + "loss": 0.9747, + "step": 28226 + }, + { + "epoch": 2.09, + "learning_rate": 1.5159073746675108e-05, + "loss": 1.0263, + "step": 28227 + }, + { + "epoch": 2.09, + "learning_rate": 1.5158732051442593e-05, + "loss": 1.0951, + "step": 28228 + }, + { + "epoch": 2.09, + "learning_rate": 1.5158390348002607e-05, + "loss": 1.0576, + "step": 28229 + }, + { + "epoch": 2.09, + "learning_rate": 1.51580486363557e-05, + "loss": 0.9418, + "step": 28230 + }, + { + "epoch": 2.09, + "learning_rate": 1.515770691650241e-05, + "loss": 1.0051, + "step": 28231 + }, + { + "epoch": 2.09, + "learning_rate": 1.5157365188443285e-05, + "loss": 0.9874, + "step": 28232 + }, + { + "epoch": 2.09, + "learning_rate": 1.5157023452178861e-05, + "loss": 1.0076, + "step": 28233 + }, + { + "epoch": 2.09, + "learning_rate": 1.5156681707709688e-05, + "loss": 0.9981, + "step": 28234 + }, + { + "epoch": 2.09, + "learning_rate": 1.515633995503631e-05, + "loss": 0.9828, + "step": 28235 + }, + { + "epoch": 2.09, + "learning_rate": 1.5155998194159267e-05, + "loss": 1.0265, + "step": 28236 + }, + { + "epoch": 2.09, + "learning_rate": 1.5155656425079105e-05, + "loss": 0.9339, + "step": 28237 + }, + { + "epoch": 2.09, + "learning_rate": 1.5155314647796367e-05, + "loss": 0.9952, + "step": 28238 + }, + { + "epoch": 2.09, + "learning_rate": 1.5154972862311597e-05, + "loss": 1.0228, + "step": 28239 + }, + { + "epoch": 2.09, + "learning_rate": 1.5154631068625341e-05, + "loss": 1.0608, + "step": 28240 + }, + { + "epoch": 2.09, + "learning_rate": 1.5154289266738137e-05, + "loss": 1.0357, + "step": 28241 + }, + { + "epoch": 2.09, + "learning_rate": 1.5153947456650535e-05, + "loss": 0.9847, + "step": 28242 + }, + { + "epoch": 2.09, + "learning_rate": 1.5153605638363078e-05, + "loss": 0.9852, + "step": 28243 + }, + { + "epoch": 2.09, + "learning_rate": 1.5153263811876304e-05, + "loss": 0.9879, + "step": 28244 + }, + { + "epoch": 2.09, + "learning_rate": 1.5152921977190763e-05, + "loss": 0.9669, + "step": 28245 + }, + { + "epoch": 2.09, + "learning_rate": 1.5152580134306996e-05, + "loss": 1.0454, + "step": 28246 + }, + { + "epoch": 2.09, + "learning_rate": 1.5152238283225548e-05, + "loss": 1.0437, + "step": 28247 + }, + { + "epoch": 2.09, + "learning_rate": 1.5151896423946965e-05, + "loss": 0.9657, + "step": 28248 + }, + { + "epoch": 2.09, + "learning_rate": 1.5151554556471788e-05, + "loss": 0.9106, + "step": 28249 + }, + { + "epoch": 2.09, + "learning_rate": 1.5151212680800562e-05, + "loss": 1.0276, + "step": 28250 + }, + { + "epoch": 2.09, + "learning_rate": 1.5150870796933827e-05, + "loss": 1.0294, + "step": 28251 + }, + { + "epoch": 2.09, + "learning_rate": 1.5150528904872133e-05, + "loss": 1.0495, + "step": 28252 + }, + { + "epoch": 2.09, + "learning_rate": 1.515018700461602e-05, + "loss": 1.0062, + "step": 28253 + }, + { + "epoch": 2.09, + "learning_rate": 1.5149845096166032e-05, + "loss": 0.9539, + "step": 28254 + }, + { + "epoch": 2.09, + "learning_rate": 1.5149503179522715e-05, + "loss": 0.9927, + "step": 28255 + }, + { + "epoch": 2.09, + "learning_rate": 1.5149161254686613e-05, + "loss": 1.12, + "step": 28256 + }, + { + "epoch": 2.09, + "learning_rate": 1.5148819321658268e-05, + "loss": 0.9718, + "step": 28257 + }, + { + "epoch": 2.09, + "learning_rate": 1.5148477380438227e-05, + "loss": 0.9828, + "step": 28258 + }, + { + "epoch": 2.09, + "learning_rate": 1.514813543102703e-05, + "loss": 0.969, + "step": 28259 + }, + { + "epoch": 2.09, + "learning_rate": 1.5147793473425225e-05, + "loss": 1.0771, + "step": 28260 + }, + { + "epoch": 2.09, + "learning_rate": 1.5147451507633353e-05, + "loss": 1.0014, + "step": 28261 + }, + { + "epoch": 2.09, + "learning_rate": 1.5147109533651959e-05, + "loss": 0.9738, + "step": 28262 + }, + { + "epoch": 2.09, + "learning_rate": 1.514676755148159e-05, + "loss": 0.9875, + "step": 28263 + }, + { + "epoch": 2.09, + "learning_rate": 1.5146425561122784e-05, + "loss": 0.9664, + "step": 28264 + }, + { + "epoch": 2.09, + "learning_rate": 1.514608356257609e-05, + "loss": 0.9975, + "step": 28265 + }, + { + "epoch": 2.09, + "learning_rate": 1.5145741555842051e-05, + "loss": 1.0295, + "step": 28266 + }, + { + "epoch": 2.09, + "learning_rate": 1.5145399540921211e-05, + "loss": 1.0435, + "step": 28267 + }, + { + "epoch": 2.09, + "learning_rate": 1.5145057517814113e-05, + "loss": 0.9598, + "step": 28268 + }, + { + "epoch": 2.09, + "learning_rate": 1.5144715486521304e-05, + "loss": 0.9346, + "step": 28269 + }, + { + "epoch": 2.09, + "learning_rate": 1.5144373447043323e-05, + "loss": 1.104, + "step": 28270 + }, + { + "epoch": 2.09, + "learning_rate": 1.5144031399380721e-05, + "loss": 0.8946, + "step": 28271 + }, + { + "epoch": 2.09, + "learning_rate": 1.5143689343534036e-05, + "loss": 1.0869, + "step": 28272 + }, + { + "epoch": 2.09, + "learning_rate": 1.5143347279503816e-05, + "loss": 1.0162, + "step": 28273 + }, + { + "epoch": 2.09, + "learning_rate": 1.5143005207290605e-05, + "loss": 1.0078, + "step": 28274 + }, + { + "epoch": 2.09, + "learning_rate": 1.5142663126894945e-05, + "loss": 0.9605, + "step": 28275 + }, + { + "epoch": 2.09, + "learning_rate": 1.514232103831738e-05, + "loss": 0.9692, + "step": 28276 + }, + { + "epoch": 2.09, + "learning_rate": 1.5141978941558458e-05, + "loss": 1.0462, + "step": 28277 + }, + { + "epoch": 2.09, + "learning_rate": 1.514163683661872e-05, + "loss": 1.0118, + "step": 28278 + }, + { + "epoch": 2.09, + "learning_rate": 1.5141294723498713e-05, + "loss": 1.0245, + "step": 28279 + }, + { + "epoch": 2.09, + "learning_rate": 1.514095260219898e-05, + "loss": 0.9222, + "step": 28280 + }, + { + "epoch": 2.09, + "learning_rate": 1.5140610472720063e-05, + "loss": 1.0506, + "step": 28281 + }, + { + "epoch": 2.09, + "learning_rate": 1.5140268335062507e-05, + "loss": 0.9317, + "step": 28282 + }, + { + "epoch": 2.09, + "learning_rate": 1.513992618922686e-05, + "loss": 0.8892, + "step": 28283 + }, + { + "epoch": 2.09, + "learning_rate": 1.5139584035213663e-05, + "loss": 1.0394, + "step": 28284 + }, + { + "epoch": 2.09, + "learning_rate": 1.5139241873023463e-05, + "loss": 1.0818, + "step": 28285 + }, + { + "epoch": 2.09, + "learning_rate": 1.51388997026568e-05, + "loss": 0.994, + "step": 28286 + }, + { + "epoch": 2.09, + "learning_rate": 1.5138557524114223e-05, + "loss": 1.0211, + "step": 28287 + }, + { + "epoch": 2.09, + "learning_rate": 1.5138215337396274e-05, + "loss": 1.1109, + "step": 28288 + }, + { + "epoch": 2.09, + "learning_rate": 1.5137873142503499e-05, + "loss": 0.9856, + "step": 28289 + }, + { + "epoch": 2.09, + "learning_rate": 1.513753093943644e-05, + "loss": 0.9366, + "step": 28290 + }, + { + "epoch": 2.09, + "learning_rate": 1.5137188728195641e-05, + "loss": 1.0358, + "step": 28291 + }, + { + "epoch": 2.09, + "learning_rate": 1.5136846508781652e-05, + "loss": 0.9476, + "step": 28292 + }, + { + "epoch": 2.09, + "learning_rate": 1.5136504281195015e-05, + "loss": 1.1678, + "step": 28293 + }, + { + "epoch": 2.09, + "learning_rate": 1.5136162045436269e-05, + "loss": 1.0201, + "step": 28294 + }, + { + "epoch": 2.09, + "learning_rate": 1.5135819801505962e-05, + "loss": 1.0657, + "step": 28295 + }, + { + "epoch": 2.09, + "learning_rate": 1.5135477549404645e-05, + "loss": 0.9322, + "step": 28296 + }, + { + "epoch": 2.09, + "learning_rate": 1.513513528913285e-05, + "loss": 0.9989, + "step": 28297 + }, + { + "epoch": 2.09, + "learning_rate": 1.5134793020691131e-05, + "loss": 0.9466, + "step": 28298 + }, + { + "epoch": 2.09, + "learning_rate": 1.513445074408003e-05, + "loss": 0.9818, + "step": 28299 + }, + { + "epoch": 2.09, + "learning_rate": 1.5134108459300092e-05, + "loss": 1.0534, + "step": 28300 + }, + { + "epoch": 2.09, + "learning_rate": 1.5133766166351858e-05, + "loss": 1.0336, + "step": 28301 + }, + { + "epoch": 2.09, + "learning_rate": 1.5133423865235876e-05, + "loss": 1.033, + "step": 28302 + }, + { + "epoch": 2.09, + "learning_rate": 1.5133081555952691e-05, + "loss": 0.9573, + "step": 28303 + }, + { + "epoch": 2.09, + "learning_rate": 1.513273923850285e-05, + "loss": 1.0367, + "step": 28304 + }, + { + "epoch": 2.09, + "learning_rate": 1.5132396912886891e-05, + "loss": 1.1158, + "step": 28305 + }, + { + "epoch": 2.09, + "learning_rate": 1.5132054579105363e-05, + "loss": 1.0425, + "step": 28306 + }, + { + "epoch": 2.09, + "learning_rate": 1.5131712237158809e-05, + "loss": 0.9572, + "step": 28307 + }, + { + "epoch": 2.09, + "learning_rate": 1.5131369887047775e-05, + "loss": 0.9881, + "step": 28308 + }, + { + "epoch": 2.09, + "learning_rate": 1.5131027528772804e-05, + "loss": 1.1024, + "step": 28309 + }, + { + "epoch": 2.09, + "learning_rate": 1.5130685162334442e-05, + "loss": 1.0927, + "step": 28310 + }, + { + "epoch": 2.09, + "learning_rate": 1.5130342787733232e-05, + "loss": 0.9758, + "step": 28311 + }, + { + "epoch": 2.09, + "learning_rate": 1.5130000404969722e-05, + "loss": 1.0492, + "step": 28312 + }, + { + "epoch": 2.09, + "learning_rate": 1.5129658014044455e-05, + "loss": 1.1057, + "step": 28313 + }, + { + "epoch": 2.09, + "learning_rate": 1.5129315614957973e-05, + "loss": 1.0301, + "step": 28314 + }, + { + "epoch": 2.09, + "learning_rate": 1.5128973207710827e-05, + "loss": 1.0257, + "step": 28315 + }, + { + "epoch": 2.09, + "learning_rate": 1.5128630792303556e-05, + "loss": 1.056, + "step": 28316 + }, + { + "epoch": 2.09, + "learning_rate": 1.5128288368736706e-05, + "loss": 0.8796, + "step": 28317 + }, + { + "epoch": 2.09, + "learning_rate": 1.5127945937010822e-05, + "loss": 1.0208, + "step": 28318 + }, + { + "epoch": 2.09, + "learning_rate": 1.5127603497126451e-05, + "loss": 0.9767, + "step": 28319 + }, + { + "epoch": 2.09, + "learning_rate": 1.5127261049084135e-05, + "loss": 0.9527, + "step": 28320 + }, + { + "epoch": 2.09, + "learning_rate": 1.5126918592884421e-05, + "loss": 1.0418, + "step": 28321 + }, + { + "epoch": 2.09, + "learning_rate": 1.5126576128527852e-05, + "loss": 0.8947, + "step": 28322 + }, + { + "epoch": 2.09, + "learning_rate": 1.5126233656014972e-05, + "loss": 0.9854, + "step": 28323 + }, + { + "epoch": 2.09, + "learning_rate": 1.5125891175346334e-05, + "loss": 1.0853, + "step": 28324 + }, + { + "epoch": 2.09, + "learning_rate": 1.5125548686522473e-05, + "loss": 0.9365, + "step": 28325 + }, + { + "epoch": 2.09, + "learning_rate": 1.5125206189543936e-05, + "loss": 0.9459, + "step": 28326 + }, + { + "epoch": 2.09, + "learning_rate": 1.512486368441127e-05, + "loss": 1.0878, + "step": 28327 + }, + { + "epoch": 2.09, + "learning_rate": 1.5124521171125019e-05, + "loss": 1.0216, + "step": 28328 + }, + { + "epoch": 2.09, + "learning_rate": 1.5124178649685727e-05, + "loss": 1.1153, + "step": 28329 + }, + { + "epoch": 2.09, + "learning_rate": 1.5123836120093942e-05, + "loss": 1.1026, + "step": 28330 + }, + { + "epoch": 2.09, + "learning_rate": 1.5123493582350206e-05, + "loss": 1.051, + "step": 28331 + }, + { + "epoch": 2.09, + "learning_rate": 1.5123151036455065e-05, + "loss": 1.0931, + "step": 28332 + }, + { + "epoch": 2.09, + "learning_rate": 1.5122808482409066e-05, + "loss": 1.1187, + "step": 28333 + }, + { + "epoch": 2.09, + "learning_rate": 1.512246592021275e-05, + "loss": 1.0814, + "step": 28334 + }, + { + "epoch": 2.09, + "learning_rate": 1.5122123349866664e-05, + "loss": 1.052, + "step": 28335 + }, + { + "epoch": 2.09, + "learning_rate": 1.5121780771371355e-05, + "loss": 1.1231, + "step": 28336 + }, + { + "epoch": 2.09, + "learning_rate": 1.5121438184727363e-05, + "loss": 1.016, + "step": 28337 + }, + { + "epoch": 2.09, + "learning_rate": 1.512109558993524e-05, + "loss": 1.001, + "step": 28338 + }, + { + "epoch": 2.09, + "learning_rate": 1.5120752986995524e-05, + "loss": 0.9453, + "step": 28339 + }, + { + "epoch": 2.09, + "learning_rate": 1.5120410375908766e-05, + "loss": 0.9401, + "step": 28340 + }, + { + "epoch": 2.09, + "learning_rate": 1.5120067756675505e-05, + "loss": 1.078, + "step": 28341 + }, + { + "epoch": 2.09, + "learning_rate": 1.5119725129296292e-05, + "loss": 0.9594, + "step": 28342 + }, + { + "epoch": 2.09, + "learning_rate": 1.5119382493771669e-05, + "loss": 0.9689, + "step": 28343 + }, + { + "epoch": 2.09, + "learning_rate": 1.511903985010218e-05, + "loss": 1.0458, + "step": 28344 + }, + { + "epoch": 2.09, + "learning_rate": 1.5118697198288375e-05, + "loss": 1.0254, + "step": 28345 + }, + { + "epoch": 2.09, + "learning_rate": 1.5118354538330793e-05, + "loss": 1.0277, + "step": 28346 + }, + { + "epoch": 2.09, + "learning_rate": 1.5118011870229983e-05, + "loss": 0.9293, + "step": 28347 + }, + { + "epoch": 2.09, + "learning_rate": 1.5117669193986489e-05, + "loss": 0.9422, + "step": 28348 + }, + { + "epoch": 2.09, + "learning_rate": 1.5117326509600858e-05, + "loss": 0.9472, + "step": 28349 + }, + { + "epoch": 2.09, + "learning_rate": 1.5116983817073632e-05, + "loss": 0.8605, + "step": 28350 + }, + { + "epoch": 2.09, + "learning_rate": 1.5116641116405357e-05, + "loss": 0.9472, + "step": 28351 + }, + { + "epoch": 2.09, + "learning_rate": 1.5116298407596583e-05, + "loss": 0.9624, + "step": 28352 + }, + { + "epoch": 2.09, + "learning_rate": 1.511595569064785e-05, + "loss": 1.0067, + "step": 28353 + }, + { + "epoch": 2.1, + "learning_rate": 1.5115612965559705e-05, + "loss": 0.926, + "step": 28354 + }, + { + "epoch": 2.1, + "learning_rate": 1.5115270232332693e-05, + "loss": 1.0458, + "step": 28355 + }, + { + "epoch": 2.1, + "learning_rate": 1.5114927490967354e-05, + "loss": 1.0635, + "step": 28356 + }, + { + "epoch": 2.1, + "learning_rate": 1.5114584741464247e-05, + "loss": 0.9377, + "step": 28357 + }, + { + "epoch": 2.1, + "learning_rate": 1.5114241983823903e-05, + "loss": 1.0297, + "step": 28358 + }, + { + "epoch": 2.1, + "learning_rate": 1.5113899218046877e-05, + "loss": 1.0102, + "step": 28359 + }, + { + "epoch": 2.1, + "learning_rate": 1.5113556444133706e-05, + "loss": 0.976, + "step": 28360 + }, + { + "epoch": 2.1, + "learning_rate": 1.5113213662084945e-05, + "loss": 1.0469, + "step": 28361 + }, + { + "epoch": 2.1, + "learning_rate": 1.511287087190113e-05, + "loss": 1.0221, + "step": 28362 + }, + { + "epoch": 2.1, + "learning_rate": 1.5112528073582812e-05, + "loss": 1.0255, + "step": 28363 + }, + { + "epoch": 2.1, + "learning_rate": 1.5112185267130537e-05, + "loss": 0.9953, + "step": 28364 + }, + { + "epoch": 2.1, + "learning_rate": 1.5111842452544845e-05, + "loss": 0.9692, + "step": 28365 + }, + { + "epoch": 2.1, + "learning_rate": 1.5111499629826284e-05, + "loss": 1.0186, + "step": 28366 + }, + { + "epoch": 2.1, + "learning_rate": 1.5111156798975403e-05, + "loss": 1.0073, + "step": 28367 + }, + { + "epoch": 2.1, + "learning_rate": 1.5110813959992744e-05, + "loss": 0.8042, + "step": 28368 + }, + { + "epoch": 2.1, + "learning_rate": 1.5110471112878856e-05, + "loss": 0.9972, + "step": 28369 + }, + { + "epoch": 2.1, + "learning_rate": 1.5110128257634278e-05, + "loss": 0.9357, + "step": 28370 + }, + { + "epoch": 2.1, + "learning_rate": 1.5109785394259562e-05, + "loss": 0.9714, + "step": 28371 + }, + { + "epoch": 2.1, + "learning_rate": 1.5109442522755248e-05, + "loss": 1.0819, + "step": 28372 + }, + { + "epoch": 2.1, + "learning_rate": 1.5109099643121884e-05, + "loss": 0.9507, + "step": 28373 + }, + { + "epoch": 2.1, + "learning_rate": 1.5108756755360014e-05, + "loss": 1.0897, + "step": 28374 + }, + { + "epoch": 2.1, + "learning_rate": 1.510841385947019e-05, + "loss": 0.9137, + "step": 28375 + }, + { + "epoch": 2.1, + "learning_rate": 1.5108070955452948e-05, + "loss": 1.0679, + "step": 28376 + }, + { + "epoch": 2.1, + "learning_rate": 1.510772804330884e-05, + "loss": 0.9651, + "step": 28377 + }, + { + "epoch": 2.1, + "learning_rate": 1.5107385123038411e-05, + "loss": 0.951, + "step": 28378 + }, + { + "epoch": 2.1, + "learning_rate": 1.5107042194642202e-05, + "loss": 0.9026, + "step": 28379 + }, + { + "epoch": 2.1, + "learning_rate": 1.5106699258120768e-05, + "loss": 0.8333, + "step": 28380 + }, + { + "epoch": 2.1, + "learning_rate": 1.5106356313474644e-05, + "loss": 1.0877, + "step": 28381 + }, + { + "epoch": 2.1, + "learning_rate": 1.510601336070438e-05, + "loss": 1.1402, + "step": 28382 + }, + { + "epoch": 2.1, + "learning_rate": 1.5105670399810523e-05, + "loss": 1.1119, + "step": 28383 + }, + { + "epoch": 2.1, + "learning_rate": 1.5105327430793616e-05, + "loss": 0.9772, + "step": 28384 + }, + { + "epoch": 2.1, + "learning_rate": 1.5104984453654207e-05, + "loss": 1.003, + "step": 28385 + }, + { + "epoch": 2.1, + "learning_rate": 1.5104641468392841e-05, + "loss": 0.9857, + "step": 28386 + }, + { + "epoch": 2.1, + "learning_rate": 1.5104298475010064e-05, + "loss": 1.0625, + "step": 28387 + }, + { + "epoch": 2.1, + "learning_rate": 1.510395547350642e-05, + "loss": 1.1231, + "step": 28388 + }, + { + "epoch": 2.1, + "learning_rate": 1.5103612463882456e-05, + "loss": 1.0341, + "step": 28389 + }, + { + "epoch": 2.1, + "learning_rate": 1.510326944613872e-05, + "loss": 1.149, + "step": 28390 + }, + { + "epoch": 2.1, + "learning_rate": 1.5102926420275751e-05, + "loss": 1.0349, + "step": 28391 + }, + { + "epoch": 2.1, + "learning_rate": 1.5102583386294102e-05, + "loss": 0.9366, + "step": 28392 + }, + { + "epoch": 2.1, + "learning_rate": 1.5102240344194315e-05, + "loss": 1.0151, + "step": 28393 + }, + { + "epoch": 2.1, + "learning_rate": 1.5101897293976939e-05, + "loss": 1.0419, + "step": 28394 + }, + { + "epoch": 2.1, + "learning_rate": 1.5101554235642513e-05, + "loss": 0.9566, + "step": 28395 + }, + { + "epoch": 2.1, + "learning_rate": 1.5101211169191593e-05, + "loss": 1.0419, + "step": 28396 + }, + { + "epoch": 2.1, + "learning_rate": 1.5100868094624712e-05, + "loss": 1.0154, + "step": 28397 + }, + { + "epoch": 2.1, + "learning_rate": 1.5100525011942427e-05, + "loss": 0.9703, + "step": 28398 + }, + { + "epoch": 2.1, + "learning_rate": 1.510018192114528e-05, + "loss": 1.0319, + "step": 28399 + }, + { + "epoch": 2.1, + "learning_rate": 1.5099838822233816e-05, + "loss": 0.9853, + "step": 28400 + }, + { + "epoch": 2.1, + "learning_rate": 1.509949571520858e-05, + "loss": 1.0531, + "step": 28401 + }, + { + "epoch": 2.1, + "learning_rate": 1.509915260007012e-05, + "loss": 1.0283, + "step": 28402 + }, + { + "epoch": 2.1, + "learning_rate": 1.5098809476818981e-05, + "loss": 1.0168, + "step": 28403 + }, + { + "epoch": 2.1, + "learning_rate": 1.509846634545571e-05, + "loss": 1.0704, + "step": 28404 + }, + { + "epoch": 2.1, + "learning_rate": 1.509812320598085e-05, + "loss": 1.0072, + "step": 28405 + }, + { + "epoch": 2.1, + "learning_rate": 1.5097780058394951e-05, + "loss": 0.9506, + "step": 28406 + }, + { + "epoch": 2.1, + "learning_rate": 1.5097436902698554e-05, + "loss": 0.9795, + "step": 28407 + }, + { + "epoch": 2.1, + "learning_rate": 1.5097093738892214e-05, + "loss": 0.9567, + "step": 28408 + }, + { + "epoch": 2.1, + "learning_rate": 1.5096750566976468e-05, + "loss": 1.0236, + "step": 28409 + }, + { + "epoch": 2.1, + "learning_rate": 1.5096407386951863e-05, + "loss": 0.9869, + "step": 28410 + }, + { + "epoch": 2.1, + "learning_rate": 1.5096064198818945e-05, + "loss": 0.9614, + "step": 28411 + }, + { + "epoch": 2.1, + "learning_rate": 1.5095721002578266e-05, + "loss": 0.8931, + "step": 28412 + }, + { + "epoch": 2.1, + "learning_rate": 1.5095377798230365e-05, + "loss": 0.8949, + "step": 28413 + }, + { + "epoch": 2.1, + "learning_rate": 1.5095034585775792e-05, + "loss": 1.077, + "step": 28414 + }, + { + "epoch": 2.1, + "learning_rate": 1.5094691365215091e-05, + "loss": 1.099, + "step": 28415 + }, + { + "epoch": 2.1, + "learning_rate": 1.5094348136548811e-05, + "loss": 0.8888, + "step": 28416 + }, + { + "epoch": 2.1, + "learning_rate": 1.5094004899777495e-05, + "loss": 0.9251, + "step": 28417 + }, + { + "epoch": 2.1, + "learning_rate": 1.509366165490169e-05, + "loss": 1.1251, + "step": 28418 + }, + { + "epoch": 2.1, + "learning_rate": 1.5093318401921942e-05, + "loss": 1.0524, + "step": 28419 + }, + { + "epoch": 2.1, + "learning_rate": 1.5092975140838797e-05, + "loss": 1.0638, + "step": 28420 + }, + { + "epoch": 2.1, + "learning_rate": 1.50926318716528e-05, + "loss": 0.9633, + "step": 28421 + }, + { + "epoch": 2.1, + "learning_rate": 1.5092288594364501e-05, + "loss": 0.9876, + "step": 28422 + }, + { + "epoch": 2.1, + "learning_rate": 1.5091945308974443e-05, + "loss": 1.012, + "step": 28423 + }, + { + "epoch": 2.1, + "learning_rate": 1.5091602015483173e-05, + "loss": 0.9513, + "step": 28424 + }, + { + "epoch": 2.1, + "learning_rate": 1.5091258713891234e-05, + "loss": 1.0111, + "step": 28425 + }, + { + "epoch": 2.1, + "learning_rate": 1.509091540419918e-05, + "loss": 0.978, + "step": 28426 + }, + { + "epoch": 2.1, + "learning_rate": 1.5090572086407551e-05, + "loss": 0.9539, + "step": 28427 + }, + { + "epoch": 2.1, + "learning_rate": 1.5090228760516893e-05, + "loss": 1.0511, + "step": 28428 + }, + { + "epoch": 2.1, + "learning_rate": 1.5089885426527757e-05, + "loss": 1.1345, + "step": 28429 + }, + { + "epoch": 2.1, + "learning_rate": 1.5089542084440684e-05, + "loss": 1.0583, + "step": 28430 + }, + { + "epoch": 2.1, + "learning_rate": 1.508919873425622e-05, + "loss": 0.9417, + "step": 28431 + }, + { + "epoch": 2.1, + "learning_rate": 1.5088855375974917e-05, + "loss": 0.9657, + "step": 28432 + }, + { + "epoch": 2.1, + "learning_rate": 1.5088512009597316e-05, + "loss": 1.0025, + "step": 28433 + }, + { + "epoch": 2.1, + "learning_rate": 1.5088168635123967e-05, + "loss": 1.0789, + "step": 28434 + }, + { + "epoch": 2.1, + "learning_rate": 1.5087825252555413e-05, + "loss": 0.9126, + "step": 28435 + }, + { + "epoch": 2.1, + "learning_rate": 1.5087481861892204e-05, + "loss": 0.9276, + "step": 28436 + }, + { + "epoch": 2.1, + "learning_rate": 1.5087138463134883e-05, + "loss": 1.0609, + "step": 28437 + }, + { + "epoch": 2.1, + "learning_rate": 1.5086795056283997e-05, + "loss": 1.0235, + "step": 28438 + }, + { + "epoch": 2.1, + "learning_rate": 1.5086451641340091e-05, + "loss": 1.0744, + "step": 28439 + }, + { + "epoch": 2.1, + "learning_rate": 1.5086108218303714e-05, + "loss": 1.043, + "step": 28440 + }, + { + "epoch": 2.1, + "learning_rate": 1.5085764787175414e-05, + "loss": 1.0787, + "step": 28441 + }, + { + "epoch": 2.1, + "learning_rate": 1.5085421347955736e-05, + "loss": 0.9235, + "step": 28442 + }, + { + "epoch": 2.1, + "learning_rate": 1.5085077900645223e-05, + "loss": 0.8584, + "step": 28443 + }, + { + "epoch": 2.1, + "learning_rate": 1.5084734445244422e-05, + "loss": 0.9764, + "step": 28444 + }, + { + "epoch": 2.1, + "learning_rate": 1.5084390981753885e-05, + "loss": 0.923, + "step": 28445 + }, + { + "epoch": 2.1, + "learning_rate": 1.5084047510174151e-05, + "loss": 1.0065, + "step": 28446 + }, + { + "epoch": 2.1, + "learning_rate": 1.5083704030505773e-05, + "loss": 1.0463, + "step": 28447 + }, + { + "epoch": 2.1, + "learning_rate": 1.5083360542749294e-05, + "loss": 0.9482, + "step": 28448 + }, + { + "epoch": 2.1, + "learning_rate": 1.508301704690526e-05, + "loss": 1.0015, + "step": 28449 + }, + { + "epoch": 2.1, + "learning_rate": 1.5082673542974218e-05, + "loss": 1.0001, + "step": 28450 + }, + { + "epoch": 2.1, + "learning_rate": 1.508233003095672e-05, + "loss": 1.0446, + "step": 28451 + }, + { + "epoch": 2.1, + "learning_rate": 1.5081986510853301e-05, + "loss": 0.9789, + "step": 28452 + }, + { + "epoch": 2.1, + "learning_rate": 1.508164298266452e-05, + "loss": 1.0562, + "step": 28453 + }, + { + "epoch": 2.1, + "learning_rate": 1.5081299446390913e-05, + "loss": 0.9405, + "step": 28454 + }, + { + "epoch": 2.1, + "learning_rate": 1.5080955902033036e-05, + "loss": 0.9541, + "step": 28455 + }, + { + "epoch": 2.1, + "learning_rate": 1.5080612349591426e-05, + "loss": 1.0674, + "step": 28456 + }, + { + "epoch": 2.1, + "learning_rate": 1.5080268789066637e-05, + "loss": 1.0996, + "step": 28457 + }, + { + "epoch": 2.1, + "learning_rate": 1.5079925220459213e-05, + "loss": 0.9713, + "step": 28458 + }, + { + "epoch": 2.1, + "learning_rate": 1.50795816437697e-05, + "loss": 0.9089, + "step": 28459 + }, + { + "epoch": 2.1, + "learning_rate": 1.5079238058998647e-05, + "loss": 0.9789, + "step": 28460 + }, + { + "epoch": 2.1, + "learning_rate": 1.5078894466146596e-05, + "loss": 1.0126, + "step": 28461 + }, + { + "epoch": 2.1, + "learning_rate": 1.5078550865214096e-05, + "loss": 1.1139, + "step": 28462 + }, + { + "epoch": 2.1, + "learning_rate": 1.5078207256201698e-05, + "loss": 0.9948, + "step": 28463 + }, + { + "epoch": 2.1, + "learning_rate": 1.5077863639109942e-05, + "loss": 1.0083, + "step": 28464 + }, + { + "epoch": 2.1, + "learning_rate": 1.5077520013939381e-05, + "loss": 1.0208, + "step": 28465 + }, + { + "epoch": 2.1, + "learning_rate": 1.5077176380690555e-05, + "loss": 1.015, + "step": 28466 + }, + { + "epoch": 2.1, + "learning_rate": 1.5076832739364018e-05, + "loss": 0.9572, + "step": 28467 + }, + { + "epoch": 2.1, + "learning_rate": 1.5076489089960307e-05, + "loss": 1.0082, + "step": 28468 + }, + { + "epoch": 2.1, + "learning_rate": 1.5076145432479979e-05, + "loss": 1.0461, + "step": 28469 + }, + { + "epoch": 2.1, + "learning_rate": 1.5075801766923573e-05, + "loss": 1.0093, + "step": 28470 + }, + { + "epoch": 2.1, + "learning_rate": 1.5075458093291641e-05, + "loss": 1.0449, + "step": 28471 + }, + { + "epoch": 2.1, + "learning_rate": 1.5075114411584726e-05, + "loss": 0.9399, + "step": 28472 + }, + { + "epoch": 2.1, + "learning_rate": 1.507477072180338e-05, + "loss": 1.1391, + "step": 28473 + }, + { + "epoch": 2.1, + "learning_rate": 1.5074427023948145e-05, + "loss": 1.0508, + "step": 28474 + }, + { + "epoch": 2.1, + "learning_rate": 1.507408331801957e-05, + "loss": 0.9629, + "step": 28475 + }, + { + "epoch": 2.1, + "learning_rate": 1.5073739604018198e-05, + "loss": 1.0895, + "step": 28476 + }, + { + "epoch": 2.1, + "learning_rate": 1.507339588194458e-05, + "loss": 1.0604, + "step": 28477 + }, + { + "epoch": 2.1, + "learning_rate": 1.5073052151799261e-05, + "loss": 1.0212, + "step": 28478 + }, + { + "epoch": 2.1, + "learning_rate": 1.5072708413582792e-05, + "loss": 0.9854, + "step": 28479 + }, + { + "epoch": 2.1, + "learning_rate": 1.5072364667295713e-05, + "loss": 1.0601, + "step": 28480 + }, + { + "epoch": 2.1, + "learning_rate": 1.5072020912938577e-05, + "loss": 1.0692, + "step": 28481 + }, + { + "epoch": 2.1, + "learning_rate": 1.5071677150511926e-05, + "loss": 0.9817, + "step": 28482 + }, + { + "epoch": 2.1, + "learning_rate": 1.507133338001631e-05, + "loss": 0.9313, + "step": 28483 + }, + { + "epoch": 2.1, + "learning_rate": 1.5070989601452277e-05, + "loss": 1.0245, + "step": 28484 + }, + { + "epoch": 2.1, + "learning_rate": 1.507064581482037e-05, + "loss": 1.0084, + "step": 28485 + }, + { + "epoch": 2.1, + "learning_rate": 1.507030202012114e-05, + "loss": 1.03, + "step": 28486 + }, + { + "epoch": 2.1, + "learning_rate": 1.5069958217355129e-05, + "loss": 1.0396, + "step": 28487 + }, + { + "epoch": 2.1, + "learning_rate": 1.506961440652289e-05, + "loss": 0.9437, + "step": 28488 + }, + { + "epoch": 2.11, + "learning_rate": 1.5069270587624965e-05, + "loss": 0.9924, + "step": 28489 + }, + { + "epoch": 2.11, + "learning_rate": 1.5068926760661904e-05, + "loss": 0.9418, + "step": 28490 + }, + { + "epoch": 2.11, + "learning_rate": 1.5068582925634251e-05, + "loss": 1.0296, + "step": 28491 + }, + { + "epoch": 2.11, + "learning_rate": 1.5068239082542557e-05, + "loss": 0.9169, + "step": 28492 + }, + { + "epoch": 2.11, + "learning_rate": 1.5067895231387368e-05, + "loss": 1.1187, + "step": 28493 + }, + { + "epoch": 2.11, + "learning_rate": 1.5067551372169228e-05, + "loss": 0.9699, + "step": 28494 + }, + { + "epoch": 2.11, + "learning_rate": 1.5067207504888687e-05, + "loss": 0.9147, + "step": 28495 + }, + { + "epoch": 2.11, + "learning_rate": 1.5066863629546293e-05, + "loss": 1.0393, + "step": 28496 + }, + { + "epoch": 2.11, + "learning_rate": 1.506651974614259e-05, + "loss": 1.0305, + "step": 28497 + }, + { + "epoch": 2.11, + "learning_rate": 1.5066175854678128e-05, + "loss": 1.0492, + "step": 28498 + }, + { + "epoch": 2.11, + "learning_rate": 1.5065831955153449e-05, + "loss": 1.0543, + "step": 28499 + }, + { + "epoch": 2.11, + "learning_rate": 1.5065488047569108e-05, + "loss": 1.0691, + "step": 28500 + }, + { + "epoch": 2.11, + "learning_rate": 1.5065144131925644e-05, + "loss": 0.9548, + "step": 28501 + }, + { + "epoch": 2.11, + "learning_rate": 1.5064800208223616e-05, + "loss": 0.9918, + "step": 28502 + }, + { + "epoch": 2.11, + "learning_rate": 1.5064456276463555e-05, + "loss": 1.0807, + "step": 28503 + }, + { + "epoch": 2.11, + "learning_rate": 1.506411233664602e-05, + "loss": 1.1011, + "step": 28504 + }, + { + "epoch": 2.11, + "learning_rate": 1.5063768388771554e-05, + "loss": 0.9563, + "step": 28505 + }, + { + "epoch": 2.11, + "learning_rate": 1.5063424432840705e-05, + "loss": 1.0186, + "step": 28506 + }, + { + "epoch": 2.11, + "learning_rate": 1.5063080468854019e-05, + "loss": 0.9913, + "step": 28507 + }, + { + "epoch": 2.11, + "learning_rate": 1.5062736496812046e-05, + "loss": 1.0321, + "step": 28508 + }, + { + "epoch": 2.11, + "learning_rate": 1.5062392516715329e-05, + "loss": 1.0536, + "step": 28509 + }, + { + "epoch": 2.11, + "learning_rate": 1.5062048528564423e-05, + "loss": 0.9933, + "step": 28510 + }, + { + "epoch": 2.11, + "learning_rate": 1.5061704532359866e-05, + "loss": 0.9128, + "step": 28511 + }, + { + "epoch": 2.11, + "learning_rate": 1.5061360528102213e-05, + "loss": 0.929, + "step": 28512 + }, + { + "epoch": 2.11, + "learning_rate": 1.5061016515792006e-05, + "loss": 1.0885, + "step": 28513 + }, + { + "epoch": 2.11, + "learning_rate": 1.5060672495429794e-05, + "loss": 1.0504, + "step": 28514 + }, + { + "epoch": 2.11, + "learning_rate": 1.5060328467016124e-05, + "loss": 1.0297, + "step": 28515 + }, + { + "epoch": 2.11, + "learning_rate": 1.5059984430551544e-05, + "loss": 1.051, + "step": 28516 + }, + { + "epoch": 2.11, + "learning_rate": 1.50596403860366e-05, + "loss": 0.9649, + "step": 28517 + }, + { + "epoch": 2.11, + "learning_rate": 1.5059296333471842e-05, + "loss": 0.9271, + "step": 28518 + }, + { + "epoch": 2.11, + "learning_rate": 1.5058952272857816e-05, + "loss": 1.0258, + "step": 28519 + }, + { + "epoch": 2.11, + "learning_rate": 1.5058608204195068e-05, + "loss": 0.9581, + "step": 28520 + }, + { + "epoch": 2.11, + "learning_rate": 1.505826412748415e-05, + "loss": 0.96, + "step": 28521 + }, + { + "epoch": 2.11, + "learning_rate": 1.5057920042725604e-05, + "loss": 1.0035, + "step": 28522 + }, + { + "epoch": 2.11, + "learning_rate": 1.505757594991998e-05, + "loss": 0.9007, + "step": 28523 + }, + { + "epoch": 2.11, + "learning_rate": 1.5057231849067826e-05, + "loss": 1.0608, + "step": 28524 + }, + { + "epoch": 2.11, + "learning_rate": 1.5056887740169686e-05, + "loss": 0.879, + "step": 28525 + }, + { + "epoch": 2.11, + "learning_rate": 1.505654362322611e-05, + "loss": 1.1071, + "step": 28526 + }, + { + "epoch": 2.11, + "learning_rate": 1.5056199498237645e-05, + "loss": 0.8458, + "step": 28527 + }, + { + "epoch": 2.11, + "learning_rate": 1.5055855365204841e-05, + "loss": 1.0943, + "step": 28528 + }, + { + "epoch": 2.11, + "learning_rate": 1.5055511224128243e-05, + "loss": 1.0251, + "step": 28529 + }, + { + "epoch": 2.11, + "learning_rate": 1.50551670750084e-05, + "loss": 1.0586, + "step": 28530 + }, + { + "epoch": 2.11, + "learning_rate": 1.5054822917845858e-05, + "loss": 0.993, + "step": 28531 + }, + { + "epoch": 2.11, + "learning_rate": 1.5054478752641164e-05, + "loss": 1.0379, + "step": 28532 + }, + { + "epoch": 2.11, + "learning_rate": 1.5054134579394868e-05, + "loss": 0.9825, + "step": 28533 + }, + { + "epoch": 2.11, + "learning_rate": 1.5053790398107512e-05, + "loss": 0.9409, + "step": 28534 + }, + { + "epoch": 2.11, + "learning_rate": 1.5053446208779652e-05, + "loss": 1.0235, + "step": 28535 + }, + { + "epoch": 2.11, + "learning_rate": 1.5053102011411831e-05, + "loss": 0.9795, + "step": 28536 + }, + { + "epoch": 2.11, + "learning_rate": 1.5052757806004596e-05, + "loss": 0.9699, + "step": 28537 + }, + { + "epoch": 2.11, + "learning_rate": 1.5052413592558497e-05, + "loss": 0.8766, + "step": 28538 + }, + { + "epoch": 2.11, + "learning_rate": 1.505206937107408e-05, + "loss": 0.8618, + "step": 28539 + }, + { + "epoch": 2.11, + "learning_rate": 1.5051725141551891e-05, + "loss": 1.003, + "step": 28540 + }, + { + "epoch": 2.11, + "learning_rate": 1.505138090399248e-05, + "loss": 1.0384, + "step": 28541 + }, + { + "epoch": 2.11, + "learning_rate": 1.5051036658396396e-05, + "loss": 1.0504, + "step": 28542 + }, + { + "epoch": 2.11, + "learning_rate": 1.5050692404764185e-05, + "loss": 1.0135, + "step": 28543 + }, + { + "epoch": 2.11, + "learning_rate": 1.5050348143096391e-05, + "loss": 1.0385, + "step": 28544 + }, + { + "epoch": 2.11, + "learning_rate": 1.5050003873393571e-05, + "loss": 0.9654, + "step": 28545 + }, + { + "epoch": 2.11, + "learning_rate": 1.5049659595656261e-05, + "loss": 1.0066, + "step": 28546 + }, + { + "epoch": 2.11, + "learning_rate": 1.5049315309885018e-05, + "loss": 0.9864, + "step": 28547 + }, + { + "epoch": 2.11, + "learning_rate": 1.5048971016080387e-05, + "loss": 0.9733, + "step": 28548 + }, + { + "epoch": 2.11, + "learning_rate": 1.5048626714242916e-05, + "loss": 1.0101, + "step": 28549 + }, + { + "epoch": 2.11, + "learning_rate": 1.5048282404373148e-05, + "loss": 0.987, + "step": 28550 + }, + { + "epoch": 2.11, + "learning_rate": 1.5047938086471643e-05, + "loss": 1.0393, + "step": 28551 + }, + { + "epoch": 2.11, + "learning_rate": 1.5047593760538934e-05, + "loss": 1.091, + "step": 28552 + }, + { + "epoch": 2.11, + "learning_rate": 1.5047249426575577e-05, + "loss": 1.0318, + "step": 28553 + }, + { + "epoch": 2.11, + "learning_rate": 1.5046905084582117e-05, + "loss": 1.056, + "step": 28554 + }, + { + "epoch": 2.11, + "learning_rate": 1.5046560734559106e-05, + "loss": 1.0118, + "step": 28555 + }, + { + "epoch": 2.11, + "learning_rate": 1.5046216376507086e-05, + "loss": 0.9957, + "step": 28556 + }, + { + "epoch": 2.11, + "learning_rate": 1.5045872010426613e-05, + "loss": 1.0435, + "step": 28557 + }, + { + "epoch": 2.11, + "learning_rate": 1.5045527636318227e-05, + "loss": 0.8563, + "step": 28558 + }, + { + "epoch": 2.11, + "learning_rate": 1.5045183254182476e-05, + "loss": 0.9374, + "step": 28559 + }, + { + "epoch": 2.11, + "learning_rate": 1.5044838864019914e-05, + "loss": 0.9771, + "step": 28560 + }, + { + "epoch": 2.11, + "learning_rate": 1.5044494465831085e-05, + "loss": 0.9453, + "step": 28561 + }, + { + "epoch": 2.11, + "learning_rate": 1.5044150059616536e-05, + "loss": 1.0404, + "step": 28562 + }, + { + "epoch": 2.11, + "learning_rate": 1.5043805645376819e-05, + "loss": 1.0557, + "step": 28563 + }, + { + "epoch": 2.11, + "learning_rate": 1.5043461223112474e-05, + "loss": 1.0165, + "step": 28564 + }, + { + "epoch": 2.11, + "learning_rate": 1.5043116792824059e-05, + "loss": 0.9947, + "step": 28565 + }, + { + "epoch": 2.11, + "learning_rate": 1.5042772354512116e-05, + "loss": 1.0116, + "step": 28566 + }, + { + "epoch": 2.11, + "learning_rate": 1.50424279081772e-05, + "loss": 1.0217, + "step": 28567 + }, + { + "epoch": 2.11, + "learning_rate": 1.5042083453819846e-05, + "loss": 0.9183, + "step": 28568 + }, + { + "epoch": 2.11, + "learning_rate": 1.5041738991440614e-05, + "loss": 0.9159, + "step": 28569 + }, + { + "epoch": 2.11, + "learning_rate": 1.504139452104004e-05, + "loss": 0.8857, + "step": 28570 + }, + { + "epoch": 2.11, + "learning_rate": 1.5041050042618689e-05, + "loss": 1.078, + "step": 28571 + }, + { + "epoch": 2.11, + "learning_rate": 1.504070555617709e-05, + "loss": 0.9519, + "step": 28572 + }, + { + "epoch": 2.11, + "learning_rate": 1.5040361061715807e-05, + "loss": 1.0145, + "step": 28573 + }, + { + "epoch": 2.11, + "learning_rate": 1.5040016559235379e-05, + "loss": 1.0401, + "step": 28574 + }, + { + "epoch": 2.11, + "learning_rate": 1.5039672048736358e-05, + "loss": 1.0188, + "step": 28575 + }, + { + "epoch": 2.11, + "learning_rate": 1.5039327530219292e-05, + "loss": 1.0042, + "step": 28576 + }, + { + "epoch": 2.11, + "learning_rate": 1.5038983003684729e-05, + "loss": 0.9884, + "step": 28577 + }, + { + "epoch": 2.11, + "learning_rate": 1.5038638469133214e-05, + "loss": 1.068, + "step": 28578 + }, + { + "epoch": 2.11, + "learning_rate": 1.5038293926565298e-05, + "loss": 1.0936, + "step": 28579 + }, + { + "epoch": 2.11, + "learning_rate": 1.5037949375981526e-05, + "loss": 0.9768, + "step": 28580 + }, + { + "epoch": 2.11, + "learning_rate": 1.503760481738245e-05, + "loss": 0.9243, + "step": 28581 + }, + { + "epoch": 2.11, + "learning_rate": 1.5037260250768615e-05, + "loss": 1.0535, + "step": 28582 + }, + { + "epoch": 2.11, + "learning_rate": 1.5036915676140575e-05, + "loss": 1.0752, + "step": 28583 + }, + { + "epoch": 2.11, + "learning_rate": 1.5036571093498871e-05, + "loss": 1.0405, + "step": 28584 + }, + { + "epoch": 2.11, + "learning_rate": 1.5036226502844056e-05, + "loss": 1.013, + "step": 28585 + }, + { + "epoch": 2.11, + "learning_rate": 1.5035881904176677e-05, + "loss": 0.9002, + "step": 28586 + }, + { + "epoch": 2.11, + "learning_rate": 1.503553729749728e-05, + "loss": 0.898, + "step": 28587 + }, + { + "epoch": 2.11, + "learning_rate": 1.503519268280642e-05, + "loss": 0.9237, + "step": 28588 + }, + { + "epoch": 2.11, + "learning_rate": 1.5034848060104635e-05, + "loss": 0.8805, + "step": 28589 + }, + { + "epoch": 2.11, + "learning_rate": 1.5034503429392481e-05, + "loss": 1.0681, + "step": 28590 + }, + { + "epoch": 2.11, + "learning_rate": 1.5034158790670503e-05, + "loss": 1.0378, + "step": 28591 + }, + { + "epoch": 2.11, + "learning_rate": 1.5033814143939251e-05, + "loss": 0.9867, + "step": 28592 + }, + { + "epoch": 2.11, + "learning_rate": 1.5033469489199275e-05, + "loss": 1.0128, + "step": 28593 + }, + { + "epoch": 2.11, + "learning_rate": 1.5033124826451118e-05, + "loss": 0.9771, + "step": 28594 + }, + { + "epoch": 2.11, + "learning_rate": 1.5032780155695333e-05, + "loss": 1.0032, + "step": 28595 + }, + { + "epoch": 2.11, + "learning_rate": 1.5032435476932468e-05, + "loss": 1.0218, + "step": 28596 + }, + { + "epoch": 2.11, + "learning_rate": 1.5032090790163066e-05, + "loss": 0.994, + "step": 28597 + }, + { + "epoch": 2.11, + "learning_rate": 1.5031746095387685e-05, + "loss": 0.9183, + "step": 28598 + }, + { + "epoch": 2.11, + "learning_rate": 1.5031401392606863e-05, + "loss": 0.9146, + "step": 28599 + }, + { + "epoch": 2.11, + "learning_rate": 1.5031056681821157e-05, + "loss": 0.9583, + "step": 28600 + }, + { + "epoch": 2.11, + "learning_rate": 1.5030711963031106e-05, + "loss": 1.0134, + "step": 28601 + }, + { + "epoch": 2.11, + "learning_rate": 1.5030367236237271e-05, + "loss": 0.9554, + "step": 28602 + }, + { + "epoch": 2.11, + "learning_rate": 1.503002250144019e-05, + "loss": 1.0502, + "step": 28603 + }, + { + "epoch": 2.11, + "learning_rate": 1.5029677758640416e-05, + "loss": 1.0298, + "step": 28604 + }, + { + "epoch": 2.11, + "learning_rate": 1.5029333007838498e-05, + "loss": 0.9708, + "step": 28605 + }, + { + "epoch": 2.11, + "learning_rate": 1.5028988249034983e-05, + "loss": 1.0699, + "step": 28606 + }, + { + "epoch": 2.11, + "learning_rate": 1.5028643482230419e-05, + "loss": 1.0244, + "step": 28607 + }, + { + "epoch": 2.11, + "learning_rate": 1.5028298707425356e-05, + "loss": 1.0466, + "step": 28608 + }, + { + "epoch": 2.11, + "learning_rate": 1.5027953924620337e-05, + "loss": 0.9233, + "step": 28609 + }, + { + "epoch": 2.11, + "learning_rate": 1.5027609133815921e-05, + "loss": 0.9515, + "step": 28610 + }, + { + "epoch": 2.11, + "learning_rate": 1.5027264335012648e-05, + "loss": 0.9897, + "step": 28611 + }, + { + "epoch": 2.11, + "learning_rate": 1.5026919528211071e-05, + "loss": 0.9149, + "step": 28612 + }, + { + "epoch": 2.11, + "learning_rate": 1.5026574713411735e-05, + "loss": 1.0153, + "step": 28613 + }, + { + "epoch": 2.11, + "learning_rate": 1.502622989061519e-05, + "loss": 1.0394, + "step": 28614 + }, + { + "epoch": 2.11, + "learning_rate": 1.5025885059821987e-05, + "loss": 1.0215, + "step": 28615 + }, + { + "epoch": 2.11, + "learning_rate": 1.5025540221032675e-05, + "loss": 0.942, + "step": 28616 + }, + { + "epoch": 2.11, + "learning_rate": 1.5025195374247797e-05, + "loss": 1.0669, + "step": 28617 + }, + { + "epoch": 2.11, + "learning_rate": 1.5024850519467906e-05, + "loss": 0.9986, + "step": 28618 + }, + { + "epoch": 2.11, + "learning_rate": 1.5024505656693548e-05, + "loss": 0.9776, + "step": 28619 + }, + { + "epoch": 2.11, + "learning_rate": 1.5024160785925275e-05, + "loss": 1.0437, + "step": 28620 + }, + { + "epoch": 2.11, + "learning_rate": 1.5023815907163634e-05, + "loss": 1.0257, + "step": 28621 + }, + { + "epoch": 2.11, + "learning_rate": 1.5023471020409174e-05, + "loss": 0.9755, + "step": 28622 + }, + { + "epoch": 2.11, + "learning_rate": 1.5023126125662443e-05, + "loss": 0.99, + "step": 28623 + }, + { + "epoch": 2.12, + "learning_rate": 1.5022781222923994e-05, + "loss": 1.0418, + "step": 28624 + }, + { + "epoch": 2.12, + "learning_rate": 1.5022436312194366e-05, + "loss": 0.8762, + "step": 28625 + }, + { + "epoch": 2.12, + "learning_rate": 1.5022091393474116e-05, + "loss": 0.9976, + "step": 28626 + }, + { + "epoch": 2.12, + "learning_rate": 1.5021746466763792e-05, + "loss": 1.0215, + "step": 28627 + }, + { + "epoch": 2.12, + "learning_rate": 1.502140153206394e-05, + "loss": 1.0136, + "step": 28628 + }, + { + "epoch": 2.12, + "learning_rate": 1.5021056589375107e-05, + "loss": 1.0825, + "step": 28629 + }, + { + "epoch": 2.12, + "learning_rate": 1.5020711638697849e-05, + "loss": 0.9688, + "step": 28630 + }, + { + "epoch": 2.12, + "learning_rate": 1.502036668003271e-05, + "loss": 0.9462, + "step": 28631 + }, + { + "epoch": 2.12, + "learning_rate": 1.502002171338024e-05, + "loss": 1.0263, + "step": 28632 + }, + { + "epoch": 2.12, + "learning_rate": 1.5019676738740985e-05, + "loss": 1.0404, + "step": 28633 + }, + { + "epoch": 2.12, + "learning_rate": 1.5019331756115495e-05, + "loss": 1.0331, + "step": 28634 + }, + { + "epoch": 2.12, + "learning_rate": 1.5018986765504325e-05, + "loss": 0.9587, + "step": 28635 + }, + { + "epoch": 2.12, + "learning_rate": 1.5018641766908016e-05, + "loss": 1.0847, + "step": 28636 + }, + { + "epoch": 2.12, + "learning_rate": 1.5018296760327118e-05, + "loss": 1.0748, + "step": 28637 + }, + { + "epoch": 2.12, + "learning_rate": 1.5017951745762185e-05, + "loss": 1.0555, + "step": 28638 + }, + { + "epoch": 2.12, + "learning_rate": 1.501760672321376e-05, + "loss": 0.9031, + "step": 28639 + }, + { + "epoch": 2.12, + "learning_rate": 1.5017261692682397e-05, + "loss": 1.0538, + "step": 28640 + }, + { + "epoch": 2.12, + "learning_rate": 1.5016916654168638e-05, + "loss": 0.9306, + "step": 28641 + }, + { + "epoch": 2.12, + "learning_rate": 1.5016571607673041e-05, + "loss": 1.0442, + "step": 28642 + }, + { + "epoch": 2.12, + "learning_rate": 1.5016226553196149e-05, + "loss": 0.9893, + "step": 28643 + }, + { + "epoch": 2.12, + "learning_rate": 1.501588149073851e-05, + "loss": 0.8948, + "step": 28644 + }, + { + "epoch": 2.12, + "learning_rate": 1.5015536420300678e-05, + "loss": 0.9634, + "step": 28645 + }, + { + "epoch": 2.12, + "learning_rate": 1.5015191341883198e-05, + "loss": 0.991, + "step": 28646 + }, + { + "epoch": 2.12, + "learning_rate": 1.501484625548662e-05, + "loss": 0.9806, + "step": 28647 + }, + { + "epoch": 2.12, + "learning_rate": 1.5014501161111494e-05, + "loss": 0.9818, + "step": 28648 + }, + { + "epoch": 2.12, + "learning_rate": 1.501415605875837e-05, + "loss": 1.0148, + "step": 28649 + }, + { + "epoch": 2.12, + "learning_rate": 1.5013810948427794e-05, + "loss": 0.9645, + "step": 28650 + }, + { + "epoch": 2.12, + "learning_rate": 1.5013465830120315e-05, + "loss": 0.8694, + "step": 28651 + }, + { + "epoch": 2.12, + "learning_rate": 1.5013120703836488e-05, + "loss": 1.1008, + "step": 28652 + }, + { + "epoch": 2.12, + "learning_rate": 1.5012775569576854e-05, + "loss": 0.9473, + "step": 28653 + }, + { + "epoch": 2.12, + "learning_rate": 1.5012430427341967e-05, + "loss": 0.9591, + "step": 28654 + }, + { + "epoch": 2.12, + "learning_rate": 1.5012085277132376e-05, + "loss": 0.9767, + "step": 28655 + }, + { + "epoch": 2.12, + "learning_rate": 1.5011740118948625e-05, + "loss": 1.0175, + "step": 28656 + }, + { + "epoch": 2.12, + "learning_rate": 1.5011394952791269e-05, + "loss": 0.9002, + "step": 28657 + }, + { + "epoch": 2.12, + "learning_rate": 1.5011049778660858e-05, + "loss": 1.0514, + "step": 28658 + }, + { + "epoch": 2.12, + "learning_rate": 1.5010704596557937e-05, + "loss": 1.0691, + "step": 28659 + }, + { + "epoch": 2.12, + "learning_rate": 1.5010359406483055e-05, + "loss": 1.0209, + "step": 28660 + }, + { + "epoch": 2.12, + "learning_rate": 1.5010014208436765e-05, + "loss": 1.1035, + "step": 28661 + }, + { + "epoch": 2.12, + "learning_rate": 1.5009669002419614e-05, + "loss": 0.9949, + "step": 28662 + }, + { + "epoch": 2.12, + "learning_rate": 1.500932378843215e-05, + "loss": 1.0066, + "step": 28663 + }, + { + "epoch": 2.12, + "learning_rate": 1.5008978566474923e-05, + "loss": 0.9647, + "step": 28664 + }, + { + "epoch": 2.12, + "learning_rate": 1.5008633336548486e-05, + "loss": 1.0016, + "step": 28665 + }, + { + "epoch": 2.12, + "learning_rate": 1.500828809865338e-05, + "loss": 0.96, + "step": 28666 + }, + { + "epoch": 2.12, + "learning_rate": 1.5007942852790165e-05, + "loss": 1.0414, + "step": 28667 + }, + { + "epoch": 2.12, + "learning_rate": 1.5007597598959381e-05, + "loss": 0.8963, + "step": 28668 + }, + { + "epoch": 2.12, + "learning_rate": 1.5007252337161581e-05, + "loss": 1.1054, + "step": 28669 + }, + { + "epoch": 2.12, + "learning_rate": 1.5006907067397317e-05, + "loss": 1.0485, + "step": 28670 + }, + { + "epoch": 2.12, + "learning_rate": 1.5006561789667134e-05, + "loss": 1.0963, + "step": 28671 + }, + { + "epoch": 2.12, + "learning_rate": 1.500621650397158e-05, + "loss": 1.0134, + "step": 28672 + }, + { + "epoch": 2.12, + "learning_rate": 1.5005871210311212e-05, + "loss": 1.059, + "step": 28673 + }, + { + "epoch": 2.12, + "learning_rate": 1.5005525908686567e-05, + "loss": 1.0351, + "step": 28674 + }, + { + "epoch": 2.12, + "learning_rate": 1.5005180599098208e-05, + "loss": 1.0217, + "step": 28675 + }, + { + "epoch": 2.12, + "learning_rate": 1.5004835281546677e-05, + "loss": 1.1013, + "step": 28676 + }, + { + "epoch": 2.12, + "learning_rate": 1.5004489956032525e-05, + "loss": 0.9964, + "step": 28677 + }, + { + "epoch": 2.12, + "learning_rate": 1.50041446225563e-05, + "loss": 1.1258, + "step": 28678 + }, + { + "epoch": 2.12, + "learning_rate": 1.5003799281118554e-05, + "loss": 1.0184, + "step": 28679 + }, + { + "epoch": 2.12, + "learning_rate": 1.5003453931719834e-05, + "loss": 0.9434, + "step": 28680 + }, + { + "epoch": 2.12, + "learning_rate": 1.5003108574360692e-05, + "loss": 0.9253, + "step": 28681 + }, + { + "epoch": 2.12, + "learning_rate": 1.5002763209041674e-05, + "loss": 1.0419, + "step": 28682 + }, + { + "epoch": 2.12, + "learning_rate": 1.5002417835763332e-05, + "loss": 1.1186, + "step": 28683 + }, + { + "epoch": 2.12, + "learning_rate": 1.5002072454526213e-05, + "loss": 1.0698, + "step": 28684 + }, + { + "epoch": 2.12, + "learning_rate": 1.5001727065330872e-05, + "loss": 0.9912, + "step": 28685 + }, + { + "epoch": 2.12, + "learning_rate": 1.500138166817785e-05, + "loss": 1.0163, + "step": 28686 + }, + { + "epoch": 2.12, + "learning_rate": 1.5001036263067702e-05, + "loss": 0.985, + "step": 28687 + }, + { + "epoch": 2.12, + "learning_rate": 1.500069085000098e-05, + "loss": 1.141, + "step": 28688 + }, + { + "epoch": 2.12, + "learning_rate": 1.5000345428978231e-05, + "loss": 1.0372, + "step": 28689 + }, + { + "epoch": 2.12, + "learning_rate": 1.5000000000000002e-05, + "loss": 0.928, + "step": 28690 + }, + { + "epoch": 2.12, + "learning_rate": 1.4999654563066843e-05, + "loss": 0.9946, + "step": 28691 + }, + { + "epoch": 2.12, + "learning_rate": 1.4999309118179308e-05, + "loss": 0.9605, + "step": 28692 + }, + { + "epoch": 2.12, + "learning_rate": 1.4998963665337943e-05, + "loss": 1.1292, + "step": 28693 + }, + { + "epoch": 2.12, + "learning_rate": 1.4998618204543296e-05, + "loss": 0.9785, + "step": 28694 + }, + { + "epoch": 2.12, + "learning_rate": 1.4998272735795922e-05, + "loss": 1.048, + "step": 28695 + }, + { + "epoch": 2.12, + "learning_rate": 1.4997927259096365e-05, + "loss": 1.0206, + "step": 28696 + }, + { + "epoch": 2.12, + "learning_rate": 1.499758177444518e-05, + "loss": 1.0378, + "step": 28697 + }, + { + "epoch": 2.12, + "learning_rate": 1.4997236281842913e-05, + "loss": 0.9663, + "step": 28698 + }, + { + "epoch": 2.12, + "learning_rate": 1.4996890781290114e-05, + "loss": 0.9388, + "step": 28699 + }, + { + "epoch": 2.12, + "learning_rate": 1.4996545272787333e-05, + "loss": 0.9829, + "step": 28700 + }, + { + "epoch": 2.12, + "learning_rate": 1.4996199756335118e-05, + "loss": 1.0518, + "step": 28701 + }, + { + "epoch": 2.12, + "learning_rate": 1.4995854231934025e-05, + "loss": 1.0532, + "step": 28702 + }, + { + "epoch": 2.12, + "learning_rate": 1.4995508699584599e-05, + "loss": 1.0323, + "step": 28703 + }, + { + "epoch": 2.12, + "learning_rate": 1.4995163159287387e-05, + "loss": 1.152, + "step": 28704 + }, + { + "epoch": 2.12, + "learning_rate": 1.4994817611042942e-05, + "loss": 0.9313, + "step": 28705 + }, + { + "epoch": 2.12, + "learning_rate": 1.4994472054851816e-05, + "loss": 0.9902, + "step": 28706 + }, + { + "epoch": 2.12, + "learning_rate": 1.4994126490714556e-05, + "loss": 1.0149, + "step": 28707 + }, + { + "epoch": 2.12, + "learning_rate": 1.4993780918631713e-05, + "loss": 0.8756, + "step": 28708 + }, + { + "epoch": 2.12, + "learning_rate": 1.4993435338603835e-05, + "loss": 1.0792, + "step": 28709 + }, + { + "epoch": 2.12, + "learning_rate": 1.4993089750631474e-05, + "loss": 1.0328, + "step": 28710 + }, + { + "epoch": 2.12, + "learning_rate": 1.4992744154715176e-05, + "loss": 0.9686, + "step": 28711 + }, + { + "epoch": 2.12, + "learning_rate": 1.4992398550855494e-05, + "loss": 1.0154, + "step": 28712 + }, + { + "epoch": 2.12, + "learning_rate": 1.4992052939052977e-05, + "loss": 1.0276, + "step": 28713 + }, + { + "epoch": 2.12, + "learning_rate": 1.4991707319308177e-05, + "loss": 1.0755, + "step": 28714 + }, + { + "epoch": 2.12, + "learning_rate": 1.499136169162164e-05, + "loss": 0.9994, + "step": 28715 + }, + { + "epoch": 2.12, + "learning_rate": 1.4991016055993919e-05, + "loss": 1.0734, + "step": 28716 + }, + { + "epoch": 2.12, + "learning_rate": 1.4990670412425565e-05, + "loss": 0.981, + "step": 28717 + }, + { + "epoch": 2.12, + "learning_rate": 1.4990324760917125e-05, + "loss": 1.1413, + "step": 28718 + }, + { + "epoch": 2.12, + "learning_rate": 1.4989979101469146e-05, + "loss": 1.0034, + "step": 28719 + }, + { + "epoch": 2.12, + "learning_rate": 1.4989633434082186e-05, + "loss": 1.0874, + "step": 28720 + }, + { + "epoch": 2.12, + "learning_rate": 1.4989287758756788e-05, + "loss": 0.9776, + "step": 28721 + }, + { + "epoch": 2.12, + "learning_rate": 1.4988942075493507e-05, + "loss": 0.9387, + "step": 28722 + }, + { + "epoch": 2.12, + "learning_rate": 1.4988596384292887e-05, + "loss": 0.9403, + "step": 28723 + }, + { + "epoch": 2.12, + "learning_rate": 1.4988250685155483e-05, + "loss": 1.0203, + "step": 28724 + }, + { + "epoch": 2.12, + "learning_rate": 1.4987904978081841e-05, + "loss": 1.08, + "step": 28725 + }, + { + "epoch": 2.12, + "learning_rate": 1.4987559263072523e-05, + "loss": 0.9863, + "step": 28726 + }, + { + "epoch": 2.12, + "learning_rate": 1.498721354012806e-05, + "loss": 1.0282, + "step": 28727 + }, + { + "epoch": 2.12, + "learning_rate": 1.4986867809249015e-05, + "loss": 0.9661, + "step": 28728 + }, + { + "epoch": 2.12, + "learning_rate": 1.4986522070435934e-05, + "loss": 1.0223, + "step": 28729 + }, + { + "epoch": 2.12, + "learning_rate": 1.4986176323689368e-05, + "loss": 1.0177, + "step": 28730 + }, + { + "epoch": 2.12, + "learning_rate": 1.4985830569009868e-05, + "loss": 1.0658, + "step": 28731 + }, + { + "epoch": 2.12, + "learning_rate": 1.498548480639798e-05, + "loss": 1.041, + "step": 28732 + }, + { + "epoch": 2.12, + "learning_rate": 1.4985139035854259e-05, + "loss": 1.1146, + "step": 28733 + }, + { + "epoch": 2.12, + "learning_rate": 1.4984793257379254e-05, + "loss": 1.1125, + "step": 28734 + }, + { + "epoch": 2.12, + "learning_rate": 1.498444747097351e-05, + "loss": 1.084, + "step": 28735 + }, + { + "epoch": 2.12, + "learning_rate": 1.4984101676637588e-05, + "loss": 1.0366, + "step": 28736 + }, + { + "epoch": 2.12, + "learning_rate": 1.4983755874372027e-05, + "loss": 0.931, + "step": 28737 + }, + { + "epoch": 2.12, + "learning_rate": 1.4983410064177384e-05, + "loss": 1.041, + "step": 28738 + }, + { + "epoch": 2.12, + "learning_rate": 1.4983064246054207e-05, + "loss": 1.0432, + "step": 28739 + }, + { + "epoch": 2.12, + "learning_rate": 1.4982718420003044e-05, + "loss": 1.0773, + "step": 28740 + }, + { + "epoch": 2.12, + "learning_rate": 1.4982372586024448e-05, + "loss": 0.9778, + "step": 28741 + }, + { + "epoch": 2.12, + "learning_rate": 1.4982026744118967e-05, + "loss": 0.9596, + "step": 28742 + }, + { + "epoch": 2.12, + "learning_rate": 1.4981680894287153e-05, + "loss": 0.9586, + "step": 28743 + }, + { + "epoch": 2.12, + "learning_rate": 1.4981335036529559e-05, + "loss": 1.0386, + "step": 28744 + }, + { + "epoch": 2.12, + "learning_rate": 1.4980989170846731e-05, + "loss": 1.0054, + "step": 28745 + }, + { + "epoch": 2.12, + "learning_rate": 1.4980643297239219e-05, + "loss": 1.0032, + "step": 28746 + }, + { + "epoch": 2.12, + "learning_rate": 1.4980297415707577e-05, + "loss": 1.0275, + "step": 28747 + }, + { + "epoch": 2.12, + "learning_rate": 1.497995152625235e-05, + "loss": 1.0084, + "step": 28748 + }, + { + "epoch": 2.12, + "learning_rate": 1.4979605628874094e-05, + "loss": 1.038, + "step": 28749 + }, + { + "epoch": 2.12, + "learning_rate": 1.4979259723573355e-05, + "loss": 1.0583, + "step": 28750 + }, + { + "epoch": 2.12, + "learning_rate": 1.4978913810350687e-05, + "loss": 0.9238, + "step": 28751 + }, + { + "epoch": 2.12, + "learning_rate": 1.4978567889206637e-05, + "loss": 1.0821, + "step": 28752 + }, + { + "epoch": 2.12, + "learning_rate": 1.4978221960141757e-05, + "loss": 1.0644, + "step": 28753 + }, + { + "epoch": 2.12, + "learning_rate": 1.4977876023156599e-05, + "loss": 1.1081, + "step": 28754 + }, + { + "epoch": 2.12, + "learning_rate": 1.497753007825171e-05, + "loss": 1.1674, + "step": 28755 + }, + { + "epoch": 2.12, + "learning_rate": 1.497718412542764e-05, + "loss": 1.0835, + "step": 28756 + }, + { + "epoch": 2.12, + "learning_rate": 1.4976838164684944e-05, + "loss": 0.9898, + "step": 28757 + }, + { + "epoch": 2.12, + "learning_rate": 1.4976492196024166e-05, + "loss": 1.025, + "step": 28758 + }, + { + "epoch": 2.12, + "learning_rate": 1.4976146219445863e-05, + "loss": 1.0579, + "step": 28759 + }, + { + "epoch": 2.13, + "learning_rate": 1.4975800234950582e-05, + "loss": 1.0804, + "step": 28760 + }, + { + "epoch": 2.13, + "learning_rate": 1.4975454242538874e-05, + "loss": 1.0492, + "step": 28761 + }, + { + "epoch": 2.13, + "learning_rate": 1.4975108242211289e-05, + "loss": 1.0147, + "step": 28762 + }, + { + "epoch": 2.13, + "learning_rate": 1.4974762233968378e-05, + "loss": 0.975, + "step": 28763 + }, + { + "epoch": 2.13, + "learning_rate": 1.4974416217810694e-05, + "loss": 0.9642, + "step": 28764 + }, + { + "epoch": 2.13, + "learning_rate": 1.4974070193738784e-05, + "loss": 0.8797, + "step": 28765 + }, + { + "epoch": 2.13, + "learning_rate": 1.4973724161753195e-05, + "loss": 0.9684, + "step": 28766 + }, + { + "epoch": 2.13, + "learning_rate": 1.4973378121854488e-05, + "loss": 1.0482, + "step": 28767 + }, + { + "epoch": 2.13, + "learning_rate": 1.4973032074043203e-05, + "loss": 0.9892, + "step": 28768 + }, + { + "epoch": 2.13, + "learning_rate": 1.49726860183199e-05, + "loss": 1.0104, + "step": 28769 + }, + { + "epoch": 2.13, + "learning_rate": 1.497233995468512e-05, + "loss": 1.0868, + "step": 28770 + }, + { + "epoch": 2.13, + "learning_rate": 1.4971993883139422e-05, + "loss": 0.9711, + "step": 28771 + }, + { + "epoch": 2.13, + "learning_rate": 1.4971647803683349e-05, + "loss": 1.0673, + "step": 28772 + }, + { + "epoch": 2.13, + "learning_rate": 1.4971301716317459e-05, + "loss": 1.085, + "step": 28773 + }, + { + "epoch": 2.13, + "learning_rate": 1.4970955621042298e-05, + "loss": 0.942, + "step": 28774 + }, + { + "epoch": 2.13, + "learning_rate": 1.4970609517858417e-05, + "loss": 0.9807, + "step": 28775 + }, + { + "epoch": 2.13, + "learning_rate": 1.4970263406766368e-05, + "loss": 0.9901, + "step": 28776 + }, + { + "epoch": 2.13, + "learning_rate": 1.49699172877667e-05, + "loss": 0.9936, + "step": 28777 + }, + { + "epoch": 2.13, + "learning_rate": 1.4969571160859966e-05, + "loss": 1.0295, + "step": 28778 + }, + { + "epoch": 2.13, + "learning_rate": 1.4969225026046714e-05, + "loss": 0.9687, + "step": 28779 + }, + { + "epoch": 2.13, + "learning_rate": 1.4968878883327496e-05, + "loss": 0.9707, + "step": 28780 + }, + { + "epoch": 2.13, + "learning_rate": 1.4968532732702866e-05, + "loss": 1.0739, + "step": 28781 + }, + { + "epoch": 2.13, + "learning_rate": 1.4968186574173369e-05, + "loss": 1.1339, + "step": 28782 + }, + { + "epoch": 2.13, + "learning_rate": 1.4967840407739559e-05, + "loss": 0.9817, + "step": 28783 + }, + { + "epoch": 2.13, + "learning_rate": 1.4967494233401984e-05, + "loss": 0.9983, + "step": 28784 + }, + { + "epoch": 2.13, + "learning_rate": 1.49671480511612e-05, + "loss": 1.0283, + "step": 28785 + }, + { + "epoch": 2.13, + "learning_rate": 1.496680186101775e-05, + "loss": 0.9122, + "step": 28786 + }, + { + "epoch": 2.13, + "learning_rate": 1.4966455662972193e-05, + "loss": 1.0344, + "step": 28787 + }, + { + "epoch": 2.13, + "learning_rate": 1.4966109457025075e-05, + "loss": 0.9222, + "step": 28788 + }, + { + "epoch": 2.13, + "learning_rate": 1.4965763243176947e-05, + "loss": 1.0036, + "step": 28789 + }, + { + "epoch": 2.13, + "learning_rate": 1.496541702142836e-05, + "loss": 1.0679, + "step": 28790 + }, + { + "epoch": 2.13, + "learning_rate": 1.496507079177987e-05, + "loss": 0.995, + "step": 28791 + }, + { + "epoch": 2.13, + "learning_rate": 1.4964724554232023e-05, + "loss": 0.9961, + "step": 28792 + }, + { + "epoch": 2.13, + "learning_rate": 1.4964378308785366e-05, + "loss": 0.9783, + "step": 28793 + }, + { + "epoch": 2.13, + "learning_rate": 1.4964032055440456e-05, + "loss": 1.0083, + "step": 28794 + }, + { + "epoch": 2.13, + "learning_rate": 1.4963685794197842e-05, + "loss": 1.0488, + "step": 28795 + }, + { + "epoch": 2.13, + "learning_rate": 1.4963339525058073e-05, + "loss": 0.9884, + "step": 28796 + }, + { + "epoch": 2.13, + "learning_rate": 1.4962993248021704e-05, + "loss": 1.0201, + "step": 28797 + }, + { + "epoch": 2.13, + "learning_rate": 1.4962646963089284e-05, + "loss": 1.006, + "step": 28798 + }, + { + "epoch": 2.13, + "learning_rate": 1.4962300670261363e-05, + "loss": 1.0248, + "step": 28799 + }, + { + "epoch": 2.13, + "learning_rate": 1.4961954369538494e-05, + "loss": 1.0595, + "step": 28800 + }, + { + "epoch": 2.13, + "learning_rate": 1.4961608060921226e-05, + "loss": 0.8414, + "step": 28801 + }, + { + "epoch": 2.13, + "learning_rate": 1.496126174441011e-05, + "loss": 0.9899, + "step": 28802 + }, + { + "epoch": 2.13, + "learning_rate": 1.4960915420005697e-05, + "loss": 0.9172, + "step": 28803 + }, + { + "epoch": 2.13, + "learning_rate": 1.4960569087708541e-05, + "loss": 1.0678, + "step": 28804 + }, + { + "epoch": 2.13, + "learning_rate": 1.4960222747519186e-05, + "loss": 1.1136, + "step": 28805 + }, + { + "epoch": 2.13, + "learning_rate": 1.4959876399438191e-05, + "loss": 0.9644, + "step": 28806 + }, + { + "epoch": 2.13, + "learning_rate": 1.4959530043466104e-05, + "loss": 0.9761, + "step": 28807 + }, + { + "epoch": 2.13, + "learning_rate": 1.4959183679603472e-05, + "loss": 1.0501, + "step": 28808 + }, + { + "epoch": 2.13, + "learning_rate": 1.4958837307850853e-05, + "loss": 0.9957, + "step": 28809 + }, + { + "epoch": 2.13, + "learning_rate": 1.4958490928208797e-05, + "loss": 1.0147, + "step": 28810 + }, + { + "epoch": 2.13, + "learning_rate": 1.4958144540677851e-05, + "loss": 1.0901, + "step": 28811 + }, + { + "epoch": 2.13, + "learning_rate": 1.4957798145258567e-05, + "loss": 1.0609, + "step": 28812 + }, + { + "epoch": 2.13, + "learning_rate": 1.4957451741951495e-05, + "loss": 1.0548, + "step": 28813 + }, + { + "epoch": 2.13, + "learning_rate": 1.495710533075719e-05, + "loss": 1.0252, + "step": 28814 + }, + { + "epoch": 2.13, + "learning_rate": 1.4956758911676202e-05, + "loss": 1.0714, + "step": 28815 + }, + { + "epoch": 2.13, + "learning_rate": 1.4956412484709079e-05, + "loss": 0.9899, + "step": 28816 + }, + { + "epoch": 2.13, + "learning_rate": 1.4956066049856377e-05, + "loss": 1.0251, + "step": 28817 + }, + { + "epoch": 2.13, + "learning_rate": 1.4955719607118646e-05, + "loss": 0.9271, + "step": 28818 + }, + { + "epoch": 2.13, + "learning_rate": 1.4955373156496434e-05, + "loss": 0.9354, + "step": 28819 + }, + { + "epoch": 2.13, + "learning_rate": 1.4955026697990298e-05, + "loss": 1.147, + "step": 28820 + }, + { + "epoch": 2.13, + "learning_rate": 1.495468023160078e-05, + "loss": 0.9769, + "step": 28821 + }, + { + "epoch": 2.13, + "learning_rate": 1.4954333757328438e-05, + "loss": 1.0076, + "step": 28822 + }, + { + "epoch": 2.13, + "learning_rate": 1.4953987275173822e-05, + "loss": 1.0957, + "step": 28823 + }, + { + "epoch": 2.13, + "learning_rate": 1.4953640785137486e-05, + "loss": 0.8897, + "step": 28824 + }, + { + "epoch": 2.13, + "learning_rate": 1.4953294287219975e-05, + "loss": 1.1, + "step": 28825 + }, + { + "epoch": 2.13, + "learning_rate": 1.4952947781421845e-05, + "loss": 1.0272, + "step": 28826 + }, + { + "epoch": 2.13, + "learning_rate": 1.4952601267743644e-05, + "loss": 0.9472, + "step": 28827 + }, + { + "epoch": 2.13, + "learning_rate": 1.495225474618593e-05, + "loss": 0.9401, + "step": 28828 + }, + { + "epoch": 2.13, + "learning_rate": 1.4951908216749246e-05, + "loss": 1.0083, + "step": 28829 + }, + { + "epoch": 2.13, + "learning_rate": 1.495156167943415e-05, + "loss": 1.1074, + "step": 28830 + }, + { + "epoch": 2.13, + "learning_rate": 1.4951215134241186e-05, + "loss": 0.8488, + "step": 28831 + }, + { + "epoch": 2.13, + "learning_rate": 1.4950868581170911e-05, + "loss": 1.0167, + "step": 28832 + }, + { + "epoch": 2.13, + "learning_rate": 1.4950522020223874e-05, + "loss": 0.8975, + "step": 28833 + }, + { + "epoch": 2.13, + "learning_rate": 1.495017545140063e-05, + "loss": 1.0327, + "step": 28834 + }, + { + "epoch": 2.13, + "learning_rate": 1.4949828874701724e-05, + "loss": 0.9918, + "step": 28835 + }, + { + "epoch": 2.13, + "learning_rate": 1.4949482290127714e-05, + "loss": 0.9409, + "step": 28836 + }, + { + "epoch": 2.13, + "learning_rate": 1.4949135697679148e-05, + "loss": 1.0214, + "step": 28837 + }, + { + "epoch": 2.13, + "learning_rate": 1.4948789097356578e-05, + "loss": 0.9103, + "step": 28838 + }, + { + "epoch": 2.13, + "learning_rate": 1.4948442489160555e-05, + "loss": 0.9985, + "step": 28839 + }, + { + "epoch": 2.13, + "learning_rate": 1.4948095873091631e-05, + "loss": 1.0288, + "step": 28840 + }, + { + "epoch": 2.13, + "learning_rate": 1.4947749249150359e-05, + "loss": 1.0232, + "step": 28841 + }, + { + "epoch": 2.13, + "learning_rate": 1.4947402617337282e-05, + "loss": 0.9935, + "step": 28842 + }, + { + "epoch": 2.13, + "learning_rate": 1.4947055977652966e-05, + "loss": 0.9197, + "step": 28843 + }, + { + "epoch": 2.13, + "learning_rate": 1.494670933009795e-05, + "loss": 1.0907, + "step": 28844 + }, + { + "epoch": 2.13, + "learning_rate": 1.4946362674672794e-05, + "loss": 1.0153, + "step": 28845 + }, + { + "epoch": 2.13, + "learning_rate": 1.4946016011378044e-05, + "loss": 1.0256, + "step": 28846 + }, + { + "epoch": 2.13, + "learning_rate": 1.4945669340214251e-05, + "loss": 1.0238, + "step": 28847 + }, + { + "epoch": 2.13, + "learning_rate": 1.4945322661181973e-05, + "loss": 0.9188, + "step": 28848 + }, + { + "epoch": 2.13, + "learning_rate": 1.4944975974281755e-05, + "loss": 1.027, + "step": 28849 + }, + { + "epoch": 2.13, + "learning_rate": 1.494462927951415e-05, + "loss": 1.1884, + "step": 28850 + }, + { + "epoch": 2.13, + "learning_rate": 1.4944282576879715e-05, + "loss": 0.9654, + "step": 28851 + }, + { + "epoch": 2.13, + "learning_rate": 1.4943935866378992e-05, + "loss": 1.1021, + "step": 28852 + }, + { + "epoch": 2.13, + "learning_rate": 1.494358914801254e-05, + "loss": 1.0416, + "step": 28853 + }, + { + "epoch": 2.13, + "learning_rate": 1.4943242421780906e-05, + "loss": 1.0157, + "step": 28854 + }, + { + "epoch": 2.13, + "learning_rate": 1.4942895687684647e-05, + "loss": 1.1164, + "step": 28855 + }, + { + "epoch": 2.13, + "learning_rate": 1.4942548945724311e-05, + "loss": 1.0603, + "step": 28856 + }, + { + "epoch": 2.13, + "learning_rate": 1.4942202195900453e-05, + "loss": 0.9257, + "step": 28857 + }, + { + "epoch": 2.13, + "learning_rate": 1.4941855438213617e-05, + "loss": 1.1363, + "step": 28858 + }, + { + "epoch": 2.13, + "learning_rate": 1.4941508672664364e-05, + "loss": 1.0354, + "step": 28859 + }, + { + "epoch": 2.13, + "learning_rate": 1.4941161899253235e-05, + "loss": 1.0449, + "step": 28860 + }, + { + "epoch": 2.13, + "learning_rate": 1.4940815117980795e-05, + "loss": 1.1278, + "step": 28861 + }, + { + "epoch": 2.13, + "learning_rate": 1.4940468328847587e-05, + "loss": 1.0206, + "step": 28862 + }, + { + "epoch": 2.13, + "learning_rate": 1.4940121531854163e-05, + "loss": 1.0171, + "step": 28863 + }, + { + "epoch": 2.13, + "learning_rate": 1.4939774727001077e-05, + "loss": 1.0189, + "step": 28864 + }, + { + "epoch": 2.13, + "learning_rate": 1.493942791428888e-05, + "loss": 1.1081, + "step": 28865 + }, + { + "epoch": 2.13, + "learning_rate": 1.4939081093718124e-05, + "loss": 0.9011, + "step": 28866 + }, + { + "epoch": 2.13, + "learning_rate": 1.4938734265289363e-05, + "loss": 1.0078, + "step": 28867 + }, + { + "epoch": 2.13, + "learning_rate": 1.4938387429003143e-05, + "loss": 1.0188, + "step": 28868 + }, + { + "epoch": 2.13, + "learning_rate": 1.493804058486002e-05, + "loss": 1.0648, + "step": 28869 + }, + { + "epoch": 2.13, + "learning_rate": 1.4937693732860545e-05, + "loss": 1.0649, + "step": 28870 + }, + { + "epoch": 2.13, + "learning_rate": 1.4937346873005272e-05, + "loss": 0.8983, + "step": 28871 + }, + { + "epoch": 2.13, + "learning_rate": 1.4937000005294748e-05, + "loss": 1.0941, + "step": 28872 + }, + { + "epoch": 2.13, + "learning_rate": 1.493665312972953e-05, + "loss": 0.9746, + "step": 28873 + }, + { + "epoch": 2.13, + "learning_rate": 1.4936306246310164e-05, + "loss": 0.9982, + "step": 28874 + }, + { + "epoch": 2.13, + "learning_rate": 1.493595935503721e-05, + "loss": 0.9901, + "step": 28875 + }, + { + "epoch": 2.13, + "learning_rate": 1.4935612455911212e-05, + "loss": 0.9373, + "step": 28876 + }, + { + "epoch": 2.13, + "learning_rate": 1.4935265548932729e-05, + "loss": 1.0535, + "step": 28877 + }, + { + "epoch": 2.13, + "learning_rate": 1.4934918634102304e-05, + "loss": 1.0247, + "step": 28878 + }, + { + "epoch": 2.13, + "learning_rate": 1.4934571711420497e-05, + "loss": 1.0404, + "step": 28879 + }, + { + "epoch": 2.13, + "learning_rate": 1.4934224780887855e-05, + "loss": 0.9813, + "step": 28880 + }, + { + "epoch": 2.13, + "learning_rate": 1.4933877842504936e-05, + "loss": 1.0303, + "step": 28881 + }, + { + "epoch": 2.13, + "learning_rate": 1.4933530896272284e-05, + "loss": 1.0395, + "step": 28882 + }, + { + "epoch": 2.13, + "learning_rate": 1.4933183942190456e-05, + "loss": 0.974, + "step": 28883 + }, + { + "epoch": 2.13, + "learning_rate": 1.4932836980260002e-05, + "loss": 1.0512, + "step": 28884 + }, + { + "epoch": 2.13, + "learning_rate": 1.4932490010481478e-05, + "loss": 1.082, + "step": 28885 + }, + { + "epoch": 2.13, + "learning_rate": 1.493214303285543e-05, + "loss": 1.0369, + "step": 28886 + }, + { + "epoch": 2.13, + "learning_rate": 1.4931796047382414e-05, + "loss": 0.9582, + "step": 28887 + }, + { + "epoch": 2.13, + "learning_rate": 1.4931449054062984e-05, + "loss": 1.0046, + "step": 28888 + }, + { + "epoch": 2.13, + "learning_rate": 1.4931102052897684e-05, + "loss": 1.0224, + "step": 28889 + }, + { + "epoch": 2.13, + "learning_rate": 1.4930755043887076e-05, + "loss": 1.1019, + "step": 28890 + }, + { + "epoch": 2.13, + "learning_rate": 1.4930408027031703e-05, + "loss": 1.024, + "step": 28891 + }, + { + "epoch": 2.13, + "learning_rate": 1.4930061002332124e-05, + "loss": 1.0121, + "step": 28892 + }, + { + "epoch": 2.13, + "learning_rate": 1.4929713969788888e-05, + "loss": 0.9443, + "step": 28893 + }, + { + "epoch": 2.13, + "learning_rate": 1.4929366929402544e-05, + "loss": 1.0225, + "step": 28894 + }, + { + "epoch": 2.14, + "learning_rate": 1.4929019881173651e-05, + "loss": 0.9542, + "step": 28895 + }, + { + "epoch": 2.14, + "learning_rate": 1.4928672825102761e-05, + "loss": 1.035, + "step": 28896 + }, + { + "epoch": 2.14, + "learning_rate": 1.492832576119042e-05, + "loss": 0.9377, + "step": 28897 + }, + { + "epoch": 2.14, + "learning_rate": 1.4927978689437182e-05, + "loss": 0.9689, + "step": 28898 + }, + { + "epoch": 2.14, + "learning_rate": 1.49276316098436e-05, + "loss": 0.9592, + "step": 28899 + }, + { + "epoch": 2.14, + "learning_rate": 1.492728452241023e-05, + "loss": 0.991, + "step": 28900 + }, + { + "epoch": 2.14, + "learning_rate": 1.4926937427137618e-05, + "loss": 0.9963, + "step": 28901 + }, + { + "epoch": 2.14, + "learning_rate": 1.492659032402632e-05, + "loss": 1.0467, + "step": 28902 + }, + { + "epoch": 2.14, + "learning_rate": 1.4926243213076888e-05, + "loss": 1.0267, + "step": 28903 + }, + { + "epoch": 2.14, + "learning_rate": 1.4925896094289873e-05, + "loss": 0.9113, + "step": 28904 + }, + { + "epoch": 2.14, + "learning_rate": 1.4925548967665825e-05, + "loss": 1.0946, + "step": 28905 + }, + { + "epoch": 2.14, + "learning_rate": 1.4925201833205306e-05, + "loss": 1.1605, + "step": 28906 + }, + { + "epoch": 2.14, + "learning_rate": 1.4924854690908855e-05, + "loss": 0.9181, + "step": 28907 + }, + { + "epoch": 2.14, + "learning_rate": 1.4924507540777033e-05, + "loss": 1.0723, + "step": 28908 + }, + { + "epoch": 2.14, + "learning_rate": 1.4924160382810388e-05, + "loss": 1.0542, + "step": 28909 + }, + { + "epoch": 2.14, + "learning_rate": 1.4923813217009477e-05, + "loss": 0.9543, + "step": 28910 + }, + { + "epoch": 2.14, + "learning_rate": 1.4923466043374849e-05, + "loss": 0.9828, + "step": 28911 + }, + { + "epoch": 2.14, + "learning_rate": 1.4923118861907058e-05, + "loss": 0.9228, + "step": 28912 + }, + { + "epoch": 2.14, + "learning_rate": 1.4922771672606654e-05, + "loss": 0.9384, + "step": 28913 + }, + { + "epoch": 2.14, + "learning_rate": 1.4922424475474191e-05, + "loss": 0.9086, + "step": 28914 + }, + { + "epoch": 2.14, + "learning_rate": 1.4922077270510219e-05, + "loss": 0.9099, + "step": 28915 + }, + { + "epoch": 2.14, + "learning_rate": 1.4921730057715294e-05, + "loss": 0.9775, + "step": 28916 + }, + { + "epoch": 2.14, + "learning_rate": 1.4921382837089966e-05, + "loss": 0.9634, + "step": 28917 + }, + { + "epoch": 2.14, + "learning_rate": 1.4921035608634791e-05, + "loss": 0.9469, + "step": 28918 + }, + { + "epoch": 2.14, + "learning_rate": 1.4920688372350317e-05, + "loss": 1.0373, + "step": 28919 + }, + { + "epoch": 2.14, + "learning_rate": 1.4920341128237099e-05, + "loss": 1.0516, + "step": 28920 + }, + { + "epoch": 2.14, + "learning_rate": 1.4919993876295686e-05, + "loss": 1.0144, + "step": 28921 + }, + { + "epoch": 2.14, + "learning_rate": 1.4919646616526635e-05, + "loss": 0.9793, + "step": 28922 + }, + { + "epoch": 2.14, + "learning_rate": 1.4919299348930498e-05, + "loss": 0.9153, + "step": 28923 + }, + { + "epoch": 2.14, + "learning_rate": 1.4918952073507826e-05, + "loss": 0.9581, + "step": 28924 + }, + { + "epoch": 2.14, + "learning_rate": 1.4918604790259169e-05, + "loss": 0.8841, + "step": 28925 + }, + { + "epoch": 2.14, + "learning_rate": 1.4918257499185084e-05, + "loss": 1.0682, + "step": 28926 + }, + { + "epoch": 2.14, + "learning_rate": 1.4917910200286121e-05, + "loss": 0.9809, + "step": 28927 + }, + { + "epoch": 2.14, + "learning_rate": 1.4917562893562834e-05, + "loss": 1.0318, + "step": 28928 + }, + { + "epoch": 2.14, + "learning_rate": 1.4917215579015774e-05, + "loss": 1.0203, + "step": 28929 + }, + { + "epoch": 2.14, + "learning_rate": 1.4916868256645494e-05, + "loss": 1.0886, + "step": 28930 + }, + { + "epoch": 2.14, + "learning_rate": 1.4916520926452547e-05, + "loss": 0.9505, + "step": 28931 + }, + { + "epoch": 2.14, + "learning_rate": 1.4916173588437487e-05, + "loss": 1.008, + "step": 28932 + }, + { + "epoch": 2.14, + "learning_rate": 1.4915826242600866e-05, + "loss": 0.9583, + "step": 28933 + }, + { + "epoch": 2.14, + "learning_rate": 1.4915478888943234e-05, + "loss": 0.9808, + "step": 28934 + }, + { + "epoch": 2.14, + "learning_rate": 1.4915131527465145e-05, + "loss": 1.0073, + "step": 28935 + }, + { + "epoch": 2.14, + "learning_rate": 1.4914784158167153e-05, + "loss": 1.0616, + "step": 28936 + }, + { + "epoch": 2.14, + "learning_rate": 1.4914436781049808e-05, + "loss": 0.8427, + "step": 28937 + }, + { + "epoch": 2.14, + "learning_rate": 1.4914089396113667e-05, + "loss": 0.9795, + "step": 28938 + }, + { + "epoch": 2.14, + "learning_rate": 1.4913742003359278e-05, + "loss": 1.1065, + "step": 28939 + }, + { + "epoch": 2.14, + "learning_rate": 1.4913394602787196e-05, + "loss": 1.0979, + "step": 28940 + }, + { + "epoch": 2.14, + "learning_rate": 1.4913047194397976e-05, + "loss": 1.1266, + "step": 28941 + }, + { + "epoch": 2.14, + "learning_rate": 1.4912699778192165e-05, + "loss": 1.0575, + "step": 28942 + }, + { + "epoch": 2.14, + "learning_rate": 1.491235235417032e-05, + "loss": 1.0645, + "step": 28943 + }, + { + "epoch": 2.14, + "learning_rate": 1.4912004922332992e-05, + "loss": 1.0074, + "step": 28944 + }, + { + "epoch": 2.14, + "learning_rate": 1.4911657482680736e-05, + "loss": 0.9937, + "step": 28945 + }, + { + "epoch": 2.14, + "learning_rate": 1.4911310035214102e-05, + "loss": 0.9848, + "step": 28946 + }, + { + "epoch": 2.14, + "learning_rate": 1.4910962579933646e-05, + "loss": 0.9318, + "step": 28947 + }, + { + "epoch": 2.14, + "learning_rate": 1.4910615116839916e-05, + "loss": 1.0365, + "step": 28948 + }, + { + "epoch": 2.14, + "learning_rate": 1.4910267645933471e-05, + "loss": 1.0279, + "step": 28949 + }, + { + "epoch": 2.14, + "learning_rate": 1.4909920167214857e-05, + "loss": 1.0723, + "step": 28950 + }, + { + "epoch": 2.14, + "learning_rate": 1.4909572680684633e-05, + "loss": 1.0885, + "step": 28951 + }, + { + "epoch": 2.14, + "learning_rate": 1.4909225186343348e-05, + "loss": 1.0019, + "step": 28952 + }, + { + "epoch": 2.14, + "learning_rate": 1.4908877684191557e-05, + "loss": 0.9793, + "step": 28953 + }, + { + "epoch": 2.14, + "learning_rate": 1.490853017422981e-05, + "loss": 1.0499, + "step": 28954 + }, + { + "epoch": 2.14, + "learning_rate": 1.4908182656458664e-05, + "loss": 1.1246, + "step": 28955 + }, + { + "epoch": 2.14, + "learning_rate": 1.4907835130878665e-05, + "loss": 1.0002, + "step": 28956 + }, + { + "epoch": 2.14, + "learning_rate": 1.4907487597490375e-05, + "loss": 0.9359, + "step": 28957 + }, + { + "epoch": 2.14, + "learning_rate": 1.4907140056294341e-05, + "loss": 1.0335, + "step": 28958 + }, + { + "epoch": 2.14, + "learning_rate": 1.4906792507291116e-05, + "loss": 1.0407, + "step": 28959 + }, + { + "epoch": 2.14, + "learning_rate": 1.4906444950481256e-05, + "loss": 1.0283, + "step": 28960 + }, + { + "epoch": 2.14, + "learning_rate": 1.4906097385865313e-05, + "loss": 1.1072, + "step": 28961 + }, + { + "epoch": 2.14, + "learning_rate": 1.4905749813443836e-05, + "loss": 0.9892, + "step": 28962 + }, + { + "epoch": 2.14, + "learning_rate": 1.4905402233217385e-05, + "loss": 1.0685, + "step": 28963 + }, + { + "epoch": 2.14, + "learning_rate": 1.4905054645186507e-05, + "loss": 0.9791, + "step": 28964 + }, + { + "epoch": 2.14, + "learning_rate": 1.4904707049351759e-05, + "loss": 1.035, + "step": 28965 + }, + { + "epoch": 2.14, + "learning_rate": 1.4904359445713689e-05, + "loss": 1.0011, + "step": 28966 + }, + { + "epoch": 2.14, + "learning_rate": 1.4904011834272858e-05, + "loss": 1.0372, + "step": 28967 + }, + { + "epoch": 2.14, + "learning_rate": 1.490366421502981e-05, + "loss": 0.9925, + "step": 28968 + }, + { + "epoch": 2.14, + "learning_rate": 1.4903316587985105e-05, + "loss": 0.9904, + "step": 28969 + }, + { + "epoch": 2.14, + "learning_rate": 1.4902968953139291e-05, + "loss": 0.9129, + "step": 28970 + }, + { + "epoch": 2.14, + "learning_rate": 1.4902621310492928e-05, + "loss": 0.9459, + "step": 28971 + }, + { + "epoch": 2.14, + "learning_rate": 1.490227366004656e-05, + "loss": 1.0682, + "step": 28972 + }, + { + "epoch": 2.14, + "learning_rate": 1.4901926001800748e-05, + "loss": 1.0744, + "step": 28973 + }, + { + "epoch": 2.14, + "learning_rate": 1.490157833575604e-05, + "loss": 1.0155, + "step": 28974 + }, + { + "epoch": 2.14, + "learning_rate": 1.4901230661912991e-05, + "loss": 0.9978, + "step": 28975 + }, + { + "epoch": 2.14, + "learning_rate": 1.4900882980272154e-05, + "loss": 0.9786, + "step": 28976 + }, + { + "epoch": 2.14, + "learning_rate": 1.4900535290834083e-05, + "loss": 0.9967, + "step": 28977 + }, + { + "epoch": 2.14, + "learning_rate": 1.4900187593599329e-05, + "loss": 0.9033, + "step": 28978 + }, + { + "epoch": 2.14, + "learning_rate": 1.489983988856845e-05, + "loss": 0.9998, + "step": 28979 + }, + { + "epoch": 2.14, + "learning_rate": 1.4899492175741994e-05, + "loss": 1.0195, + "step": 28980 + }, + { + "epoch": 2.14, + "learning_rate": 1.4899144455120518e-05, + "loss": 1.0285, + "step": 28981 + }, + { + "epoch": 2.14, + "learning_rate": 1.489879672670457e-05, + "loss": 0.9992, + "step": 28982 + }, + { + "epoch": 2.14, + "learning_rate": 1.4898448990494707e-05, + "loss": 0.9651, + "step": 28983 + }, + { + "epoch": 2.14, + "learning_rate": 1.4898101246491482e-05, + "loss": 1.0327, + "step": 28984 + }, + { + "epoch": 2.14, + "learning_rate": 1.4897753494695449e-05, + "loss": 0.9387, + "step": 28985 + }, + { + "epoch": 2.14, + "learning_rate": 1.4897405735107158e-05, + "loss": 0.9489, + "step": 28986 + }, + { + "epoch": 2.14, + "learning_rate": 1.4897057967727166e-05, + "loss": 0.8941, + "step": 28987 + }, + { + "epoch": 2.14, + "learning_rate": 1.4896710192556027e-05, + "loss": 0.9169, + "step": 28988 + }, + { + "epoch": 2.14, + "learning_rate": 1.4896362409594289e-05, + "loss": 1.1294, + "step": 28989 + }, + { + "epoch": 2.14, + "learning_rate": 1.4896014618842511e-05, + "loss": 1.0244, + "step": 28990 + }, + { + "epoch": 2.14, + "learning_rate": 1.489566682030124e-05, + "loss": 1.0179, + "step": 28991 + }, + { + "epoch": 2.14, + "learning_rate": 1.4895319013971037e-05, + "loss": 1.0151, + "step": 28992 + }, + { + "epoch": 2.14, + "learning_rate": 1.489497119985245e-05, + "loss": 1.0799, + "step": 28993 + }, + { + "epoch": 2.14, + "learning_rate": 1.4894623377946034e-05, + "loss": 1.0227, + "step": 28994 + }, + { + "epoch": 2.14, + "learning_rate": 1.4894275548252343e-05, + "loss": 1.0401, + "step": 28995 + }, + { + "epoch": 2.14, + "learning_rate": 1.4893927710771926e-05, + "loss": 1.005, + "step": 28996 + }, + { + "epoch": 2.14, + "learning_rate": 1.4893579865505345e-05, + "loss": 0.9781, + "step": 28997 + }, + { + "epoch": 2.14, + "learning_rate": 1.4893232012453147e-05, + "loss": 1.108, + "step": 28998 + }, + { + "epoch": 2.14, + "learning_rate": 1.4892884151615884e-05, + "loss": 0.9902, + "step": 28999 + }, + { + "epoch": 2.14, + "learning_rate": 1.4892536282994116e-05, + "loss": 0.9578, + "step": 29000 + }, + { + "epoch": 2.14, + "learning_rate": 1.489218840658839e-05, + "loss": 0.9824, + "step": 29001 + }, + { + "epoch": 2.14, + "learning_rate": 1.4891840522399266e-05, + "loss": 0.9878, + "step": 29002 + }, + { + "epoch": 2.14, + "learning_rate": 1.489149263042729e-05, + "loss": 0.917, + "step": 29003 + }, + { + "epoch": 2.14, + "learning_rate": 1.489114473067302e-05, + "loss": 0.9241, + "step": 29004 + }, + { + "epoch": 2.14, + "learning_rate": 1.4890796823137008e-05, + "loss": 0.9147, + "step": 29005 + }, + { + "epoch": 2.14, + "learning_rate": 1.489044890781981e-05, + "loss": 0.9878, + "step": 29006 + }, + { + "epoch": 2.14, + "learning_rate": 1.4890100984721978e-05, + "loss": 0.791, + "step": 29007 + }, + { + "epoch": 2.14, + "learning_rate": 1.4889753053844064e-05, + "loss": 1.0676, + "step": 29008 + }, + { + "epoch": 2.14, + "learning_rate": 1.4889405115186624e-05, + "loss": 1.0823, + "step": 29009 + }, + { + "epoch": 2.14, + "learning_rate": 1.488905716875021e-05, + "loss": 1.0099, + "step": 29010 + }, + { + "epoch": 2.14, + "learning_rate": 1.4888709214535375e-05, + "loss": 1.0693, + "step": 29011 + }, + { + "epoch": 2.14, + "learning_rate": 1.4888361252542676e-05, + "loss": 0.9691, + "step": 29012 + }, + { + "epoch": 2.14, + "learning_rate": 1.488801328277266e-05, + "loss": 1.0153, + "step": 29013 + }, + { + "epoch": 2.14, + "learning_rate": 1.4887665305225887e-05, + "loss": 0.9009, + "step": 29014 + }, + { + "epoch": 2.14, + "learning_rate": 1.4887317319902908e-05, + "loss": 0.91, + "step": 29015 + }, + { + "epoch": 2.14, + "learning_rate": 1.4886969326804281e-05, + "loss": 0.9372, + "step": 29016 + }, + { + "epoch": 2.14, + "learning_rate": 1.488662132593055e-05, + "loss": 1.0456, + "step": 29017 + }, + { + "epoch": 2.14, + "learning_rate": 1.4886273317282279e-05, + "loss": 1.076, + "step": 29018 + }, + { + "epoch": 2.14, + "learning_rate": 1.4885925300860014e-05, + "loss": 1.1009, + "step": 29019 + }, + { + "epoch": 2.14, + "learning_rate": 1.4885577276664314e-05, + "loss": 1.0264, + "step": 29020 + }, + { + "epoch": 2.14, + "learning_rate": 1.4885229244695727e-05, + "loss": 1.1947, + "step": 29021 + }, + { + "epoch": 2.14, + "learning_rate": 1.4884881204954812e-05, + "loss": 1.0154, + "step": 29022 + }, + { + "epoch": 2.14, + "learning_rate": 1.488453315744212e-05, + "loss": 1.0543, + "step": 29023 + }, + { + "epoch": 2.14, + "learning_rate": 1.4884185102158209e-05, + "loss": 0.9846, + "step": 29024 + }, + { + "epoch": 2.14, + "learning_rate": 1.4883837039103628e-05, + "loss": 1.0329, + "step": 29025 + }, + { + "epoch": 2.14, + "learning_rate": 1.4883488968278932e-05, + "loss": 0.9078, + "step": 29026 + }, + { + "epoch": 2.14, + "learning_rate": 1.4883140889684672e-05, + "loss": 0.9599, + "step": 29027 + }, + { + "epoch": 2.14, + "learning_rate": 1.4882792803321408e-05, + "loss": 0.9936, + "step": 29028 + }, + { + "epoch": 2.14, + "learning_rate": 1.4882444709189689e-05, + "loss": 1.1699, + "step": 29029 + }, + { + "epoch": 2.15, + "learning_rate": 1.488209660729007e-05, + "loss": 1.1097, + "step": 29030 + }, + { + "epoch": 2.15, + "learning_rate": 1.4881748497623108e-05, + "loss": 1.0219, + "step": 29031 + }, + { + "epoch": 2.15, + "learning_rate": 1.4881400380189352e-05, + "loss": 0.9487, + "step": 29032 + }, + { + "epoch": 2.15, + "learning_rate": 1.4881052254989356e-05, + "loss": 0.9419, + "step": 29033 + }, + { + "epoch": 2.15, + "learning_rate": 1.4880704122023679e-05, + "loss": 1.0585, + "step": 29034 + }, + { + "epoch": 2.15, + "learning_rate": 1.4880355981292871e-05, + "loss": 1.0104, + "step": 29035 + }, + { + "epoch": 2.15, + "learning_rate": 1.4880007832797484e-05, + "loss": 0.9841, + "step": 29036 + }, + { + "epoch": 2.15, + "learning_rate": 1.4879659676538077e-05, + "loss": 0.9847, + "step": 29037 + }, + { + "epoch": 2.15, + "learning_rate": 1.48793115125152e-05, + "loss": 1.0232, + "step": 29038 + }, + { + "epoch": 2.15, + "learning_rate": 1.4878963340729406e-05, + "loss": 1.0429, + "step": 29039 + }, + { + "epoch": 2.15, + "learning_rate": 1.4878615161181255e-05, + "loss": 1.0537, + "step": 29040 + }, + { + "epoch": 2.15, + "learning_rate": 1.4878266973871296e-05, + "loss": 0.9549, + "step": 29041 + }, + { + "epoch": 2.15, + "learning_rate": 1.4877918778800084e-05, + "loss": 1.0186, + "step": 29042 + }, + { + "epoch": 2.15, + "learning_rate": 1.487757057596817e-05, + "loss": 0.9383, + "step": 29043 + }, + { + "epoch": 2.15, + "learning_rate": 1.4877222365376116e-05, + "loss": 1.015, + "step": 29044 + }, + { + "epoch": 2.15, + "learning_rate": 1.487687414702447e-05, + "loss": 1.0556, + "step": 29045 + }, + { + "epoch": 2.15, + "learning_rate": 1.4876525920913784e-05, + "loss": 0.9693, + "step": 29046 + }, + { + "epoch": 2.15, + "learning_rate": 1.4876177687044617e-05, + "loss": 1.065, + "step": 29047 + }, + { + "epoch": 2.15, + "learning_rate": 1.487582944541752e-05, + "loss": 0.9691, + "step": 29048 + }, + { + "epoch": 2.15, + "learning_rate": 1.487548119603305e-05, + "loss": 0.9628, + "step": 29049 + }, + { + "epoch": 2.15, + "learning_rate": 1.4875132938891756e-05, + "loss": 0.9428, + "step": 29050 + }, + { + "epoch": 2.15, + "learning_rate": 1.4874784673994197e-05, + "loss": 1.0608, + "step": 29051 + }, + { + "epoch": 2.15, + "learning_rate": 1.4874436401340925e-05, + "loss": 1.0395, + "step": 29052 + }, + { + "epoch": 2.15, + "learning_rate": 1.4874088120932499e-05, + "loss": 1.0891, + "step": 29053 + }, + { + "epoch": 2.15, + "learning_rate": 1.4873739832769463e-05, + "loss": 1.0159, + "step": 29054 + }, + { + "epoch": 2.15, + "learning_rate": 1.487339153685238e-05, + "loss": 0.9533, + "step": 29055 + }, + { + "epoch": 2.15, + "learning_rate": 1.4873043233181796e-05, + "loss": 1.034, + "step": 29056 + }, + { + "epoch": 2.15, + "learning_rate": 1.4872694921758275e-05, + "loss": 1.0701, + "step": 29057 + }, + { + "epoch": 2.15, + "learning_rate": 1.4872346602582362e-05, + "loss": 0.9838, + "step": 29058 + }, + { + "epoch": 2.15, + "learning_rate": 1.4871998275654618e-05, + "loss": 0.9404, + "step": 29059 + }, + { + "epoch": 2.15, + "learning_rate": 1.4871649940975595e-05, + "loss": 0.9844, + "step": 29060 + }, + { + "epoch": 2.15, + "learning_rate": 1.4871301598545844e-05, + "loss": 0.9798, + "step": 29061 + }, + { + "epoch": 2.15, + "learning_rate": 1.4870953248365922e-05, + "loss": 1.0121, + "step": 29062 + }, + { + "epoch": 2.15, + "learning_rate": 1.4870604890436386e-05, + "loss": 1.0983, + "step": 29063 + }, + { + "epoch": 2.15, + "learning_rate": 1.4870256524757785e-05, + "loss": 1.0829, + "step": 29064 + }, + { + "epoch": 2.15, + "learning_rate": 1.486990815133068e-05, + "loss": 1.0014, + "step": 29065 + }, + { + "epoch": 2.15, + "learning_rate": 1.4869559770155615e-05, + "loss": 0.902, + "step": 29066 + }, + { + "epoch": 2.15, + "learning_rate": 1.4869211381233153e-05, + "loss": 1.0573, + "step": 29067 + }, + { + "epoch": 2.15, + "learning_rate": 1.4868862984563844e-05, + "loss": 0.9639, + "step": 29068 + }, + { + "epoch": 2.15, + "learning_rate": 1.4868514580148243e-05, + "loss": 1.0766, + "step": 29069 + }, + { + "epoch": 2.15, + "learning_rate": 1.4868166167986906e-05, + "loss": 0.941, + "step": 29070 + }, + { + "epoch": 2.15, + "learning_rate": 1.4867817748080387e-05, + "loss": 1.076, + "step": 29071 + }, + { + "epoch": 2.15, + "learning_rate": 1.486746932042924e-05, + "loss": 0.9239, + "step": 29072 + }, + { + "epoch": 2.15, + "learning_rate": 1.4867120885034017e-05, + "loss": 0.9816, + "step": 29073 + }, + { + "epoch": 2.15, + "learning_rate": 1.4866772441895274e-05, + "loss": 1.077, + "step": 29074 + }, + { + "epoch": 2.15, + "learning_rate": 1.4866423991013568e-05, + "loss": 0.9666, + "step": 29075 + }, + { + "epoch": 2.15, + "learning_rate": 1.4866075532389448e-05, + "loss": 1.1072, + "step": 29076 + }, + { + "epoch": 2.15, + "learning_rate": 1.4865727066023475e-05, + "loss": 0.884, + "step": 29077 + }, + { + "epoch": 2.15, + "learning_rate": 1.4865378591916198e-05, + "loss": 0.9701, + "step": 29078 + }, + { + "epoch": 2.15, + "learning_rate": 1.4865030110068172e-05, + "loss": 1.0461, + "step": 29079 + }, + { + "epoch": 2.15, + "learning_rate": 1.4864681620479955e-05, + "loss": 1.0266, + "step": 29080 + }, + { + "epoch": 2.15, + "learning_rate": 1.48643331231521e-05, + "loss": 0.9478, + "step": 29081 + }, + { + "epoch": 2.15, + "learning_rate": 1.486398461808516e-05, + "loss": 1.038, + "step": 29082 + }, + { + "epoch": 2.15, + "learning_rate": 1.4863636105279688e-05, + "loss": 1.0741, + "step": 29083 + }, + { + "epoch": 2.15, + "learning_rate": 1.486328758473624e-05, + "loss": 1.0698, + "step": 29084 + }, + { + "epoch": 2.15, + "learning_rate": 1.4862939056455374e-05, + "loss": 1.0306, + "step": 29085 + }, + { + "epoch": 2.15, + "learning_rate": 1.486259052043764e-05, + "loss": 1.0012, + "step": 29086 + }, + { + "epoch": 2.15, + "learning_rate": 1.4862241976683593e-05, + "loss": 1.0293, + "step": 29087 + }, + { + "epoch": 2.15, + "learning_rate": 1.486189342519379e-05, + "loss": 1.0424, + "step": 29088 + }, + { + "epoch": 2.15, + "learning_rate": 1.4861544865968784e-05, + "loss": 1.0698, + "step": 29089 + }, + { + "epoch": 2.15, + "learning_rate": 1.4861196299009129e-05, + "loss": 1.005, + "step": 29090 + }, + { + "epoch": 2.15, + "learning_rate": 1.4860847724315382e-05, + "loss": 0.9455, + "step": 29091 + }, + { + "epoch": 2.15, + "learning_rate": 1.4860499141888093e-05, + "loss": 0.9955, + "step": 29092 + }, + { + "epoch": 2.15, + "learning_rate": 1.4860150551727821e-05, + "loss": 1.0753, + "step": 29093 + }, + { + "epoch": 2.15, + "learning_rate": 1.485980195383512e-05, + "loss": 1.0163, + "step": 29094 + }, + { + "epoch": 2.15, + "learning_rate": 1.4859453348210542e-05, + "loss": 0.976, + "step": 29095 + }, + { + "epoch": 2.15, + "learning_rate": 1.4859104734854644e-05, + "loss": 0.9978, + "step": 29096 + }, + { + "epoch": 2.15, + "learning_rate": 1.4858756113767979e-05, + "loss": 0.9512, + "step": 29097 + }, + { + "epoch": 2.15, + "learning_rate": 1.4858407484951105e-05, + "loss": 1.0216, + "step": 29098 + }, + { + "epoch": 2.15, + "learning_rate": 1.485805884840457e-05, + "loss": 1.0119, + "step": 29099 + }, + { + "epoch": 2.15, + "learning_rate": 1.4857710204128939e-05, + "loss": 0.9162, + "step": 29100 + }, + { + "epoch": 2.15, + "learning_rate": 1.4857361552124757e-05, + "loss": 0.8901, + "step": 29101 + }, + { + "epoch": 2.15, + "learning_rate": 1.4857012892392581e-05, + "loss": 1.0524, + "step": 29102 + }, + { + "epoch": 2.15, + "learning_rate": 1.4856664224932968e-05, + "loss": 0.9089, + "step": 29103 + }, + { + "epoch": 2.15, + "learning_rate": 1.4856315549746473e-05, + "loss": 1.001, + "step": 29104 + }, + { + "epoch": 2.15, + "learning_rate": 1.4855966866833647e-05, + "loss": 1.0365, + "step": 29105 + }, + { + "epoch": 2.15, + "learning_rate": 1.485561817619505e-05, + "loss": 1.0733, + "step": 29106 + }, + { + "epoch": 2.15, + "learning_rate": 1.4855269477831232e-05, + "loss": 1.0744, + "step": 29107 + }, + { + "epoch": 2.15, + "learning_rate": 1.4854920771742749e-05, + "loss": 1.0051, + "step": 29108 + }, + { + "epoch": 2.15, + "learning_rate": 1.4854572057930159e-05, + "loss": 1.0011, + "step": 29109 + }, + { + "epoch": 2.15, + "learning_rate": 1.4854223336394014e-05, + "loss": 0.8849, + "step": 29110 + }, + { + "epoch": 2.15, + "learning_rate": 1.4853874607134869e-05, + "loss": 0.9532, + "step": 29111 + }, + { + "epoch": 2.15, + "learning_rate": 1.4853525870153278e-05, + "loss": 0.837, + "step": 29112 + }, + { + "epoch": 2.15, + "learning_rate": 1.4853177125449795e-05, + "loss": 0.9014, + "step": 29113 + }, + { + "epoch": 2.15, + "learning_rate": 1.485282837302498e-05, + "loss": 0.9114, + "step": 29114 + }, + { + "epoch": 2.15, + "learning_rate": 1.4852479612879381e-05, + "loss": 0.9614, + "step": 29115 + }, + { + "epoch": 2.15, + "learning_rate": 1.485213084501356e-05, + "loss": 0.9419, + "step": 29116 + }, + { + "epoch": 2.15, + "learning_rate": 1.4851782069428065e-05, + "loss": 1.0847, + "step": 29117 + }, + { + "epoch": 2.15, + "learning_rate": 1.4851433286123457e-05, + "loss": 1.0247, + "step": 29118 + }, + { + "epoch": 2.15, + "learning_rate": 1.4851084495100287e-05, + "loss": 0.9797, + "step": 29119 + }, + { + "epoch": 2.15, + "learning_rate": 1.4850735696359114e-05, + "loss": 1.064, + "step": 29120 + }, + { + "epoch": 2.15, + "learning_rate": 1.4850386889900483e-05, + "loss": 1.0711, + "step": 29121 + }, + { + "epoch": 2.15, + "learning_rate": 1.4850038075724961e-05, + "loss": 1.0717, + "step": 29122 + }, + { + "epoch": 2.15, + "learning_rate": 1.4849689253833095e-05, + "loss": 1.0348, + "step": 29123 + }, + { + "epoch": 2.15, + "learning_rate": 1.4849340424225447e-05, + "loss": 0.9985, + "step": 29124 + }, + { + "epoch": 2.15, + "learning_rate": 1.4848991586902561e-05, + "loss": 0.9487, + "step": 29125 + }, + { + "epoch": 2.15, + "learning_rate": 1.4848642741865005e-05, + "loss": 1.0427, + "step": 29126 + }, + { + "epoch": 2.15, + "learning_rate": 1.4848293889113326e-05, + "loss": 1.0341, + "step": 29127 + }, + { + "epoch": 2.15, + "learning_rate": 1.4847945028648081e-05, + "loss": 1.0396, + "step": 29128 + }, + { + "epoch": 2.15, + "learning_rate": 1.4847596160469823e-05, + "loss": 0.993, + "step": 29129 + }, + { + "epoch": 2.15, + "learning_rate": 1.4847247284579112e-05, + "loss": 0.9676, + "step": 29130 + }, + { + "epoch": 2.15, + "learning_rate": 1.4846898400976498e-05, + "loss": 1.042, + "step": 29131 + }, + { + "epoch": 2.15, + "learning_rate": 1.4846549509662538e-05, + "loss": 0.9025, + "step": 29132 + }, + { + "epoch": 2.15, + "learning_rate": 1.4846200610637785e-05, + "loss": 0.994, + "step": 29133 + }, + { + "epoch": 2.15, + "learning_rate": 1.4845851703902798e-05, + "loss": 0.9952, + "step": 29134 + }, + { + "epoch": 2.15, + "learning_rate": 1.484550278945813e-05, + "loss": 1.0255, + "step": 29135 + }, + { + "epoch": 2.15, + "learning_rate": 1.4845153867304337e-05, + "loss": 1.0485, + "step": 29136 + }, + { + "epoch": 2.15, + "learning_rate": 1.4844804937441973e-05, + "loss": 1.0832, + "step": 29137 + }, + { + "epoch": 2.15, + "learning_rate": 1.4844455999871594e-05, + "loss": 0.923, + "step": 29138 + }, + { + "epoch": 2.15, + "learning_rate": 1.4844107054593755e-05, + "loss": 0.8888, + "step": 29139 + }, + { + "epoch": 2.15, + "learning_rate": 1.4843758101609012e-05, + "loss": 1.0825, + "step": 29140 + }, + { + "epoch": 2.15, + "learning_rate": 1.4843409140917917e-05, + "loss": 0.9242, + "step": 29141 + }, + { + "epoch": 2.15, + "learning_rate": 1.4843060172521027e-05, + "loss": 1.0869, + "step": 29142 + }, + { + "epoch": 2.15, + "learning_rate": 1.48427111964189e-05, + "loss": 1.0018, + "step": 29143 + }, + { + "epoch": 2.15, + "learning_rate": 1.484236221261209e-05, + "loss": 1.0069, + "step": 29144 + }, + { + "epoch": 2.15, + "learning_rate": 1.4842013221101146e-05, + "loss": 1.0333, + "step": 29145 + }, + { + "epoch": 2.15, + "learning_rate": 1.4841664221886632e-05, + "loss": 1.0859, + "step": 29146 + }, + { + "epoch": 2.15, + "learning_rate": 1.4841315214969099e-05, + "loss": 1.0204, + "step": 29147 + }, + { + "epoch": 2.15, + "learning_rate": 1.48409662003491e-05, + "loss": 0.9813, + "step": 29148 + }, + { + "epoch": 2.15, + "learning_rate": 1.4840617178027197e-05, + "loss": 1.0467, + "step": 29149 + }, + { + "epoch": 2.15, + "learning_rate": 1.4840268148003937e-05, + "loss": 0.9976, + "step": 29150 + }, + { + "epoch": 2.15, + "learning_rate": 1.4839919110279883e-05, + "loss": 0.9045, + "step": 29151 + }, + { + "epoch": 2.15, + "learning_rate": 1.4839570064855585e-05, + "loss": 0.795, + "step": 29152 + }, + { + "epoch": 2.15, + "learning_rate": 1.4839221011731604e-05, + "loss": 0.9997, + "step": 29153 + }, + { + "epoch": 2.15, + "learning_rate": 1.4838871950908487e-05, + "loss": 0.989, + "step": 29154 + }, + { + "epoch": 2.15, + "learning_rate": 1.4838522882386797e-05, + "loss": 1.0319, + "step": 29155 + }, + { + "epoch": 2.15, + "learning_rate": 1.4838173806167086e-05, + "loss": 1.1482, + "step": 29156 + }, + { + "epoch": 2.15, + "learning_rate": 1.483782472224991e-05, + "loss": 0.9685, + "step": 29157 + }, + { + "epoch": 2.15, + "learning_rate": 1.4837475630635822e-05, + "loss": 0.9921, + "step": 29158 + }, + { + "epoch": 2.15, + "learning_rate": 1.483712653132538e-05, + "loss": 1.0972, + "step": 29159 + }, + { + "epoch": 2.15, + "learning_rate": 1.483677742431914e-05, + "loss": 1.0244, + "step": 29160 + }, + { + "epoch": 2.15, + "learning_rate": 1.4836428309617656e-05, + "loss": 0.8836, + "step": 29161 + }, + { + "epoch": 2.15, + "learning_rate": 1.4836079187221484e-05, + "loss": 0.995, + "step": 29162 + }, + { + "epoch": 2.15, + "learning_rate": 1.483573005713118e-05, + "loss": 0.9486, + "step": 29163 + }, + { + "epoch": 2.15, + "learning_rate": 1.4835380919347295e-05, + "loss": 0.9747, + "step": 29164 + }, + { + "epoch": 2.15, + "learning_rate": 1.483503177387039e-05, + "loss": 0.9675, + "step": 29165 + }, + { + "epoch": 2.16, + "learning_rate": 1.483468262070102e-05, + "loss": 1.0106, + "step": 29166 + }, + { + "epoch": 2.16, + "learning_rate": 1.4834333459839738e-05, + "loss": 1.0224, + "step": 29167 + }, + { + "epoch": 2.16, + "learning_rate": 1.4833984291287102e-05, + "loss": 1.0545, + "step": 29168 + }, + { + "epoch": 2.16, + "learning_rate": 1.4833635115043665e-05, + "loss": 1.0737, + "step": 29169 + }, + { + "epoch": 2.16, + "learning_rate": 1.4833285931109981e-05, + "loss": 0.9648, + "step": 29170 + }, + { + "epoch": 2.16, + "learning_rate": 1.4832936739486614e-05, + "loss": 1.0531, + "step": 29171 + }, + { + "epoch": 2.16, + "learning_rate": 1.483258754017411e-05, + "loss": 1.0602, + "step": 29172 + }, + { + "epoch": 2.16, + "learning_rate": 1.483223833317303e-05, + "loss": 0.9042, + "step": 29173 + }, + { + "epoch": 2.16, + "learning_rate": 1.4831889118483923e-05, + "loss": 0.9536, + "step": 29174 + }, + { + "epoch": 2.16, + "learning_rate": 1.4831539896107358e-05, + "loss": 1.0015, + "step": 29175 + }, + { + "epoch": 2.16, + "learning_rate": 1.4831190666043877e-05, + "loss": 1.0047, + "step": 29176 + }, + { + "epoch": 2.16, + "learning_rate": 1.4830841428294044e-05, + "loss": 1.033, + "step": 29177 + }, + { + "epoch": 2.16, + "learning_rate": 1.4830492182858406e-05, + "loss": 0.9223, + "step": 29178 + }, + { + "epoch": 2.16, + "learning_rate": 1.4830142929737528e-05, + "loss": 1.0407, + "step": 29179 + }, + { + "epoch": 2.16, + "learning_rate": 1.482979366893196e-05, + "loss": 1.0444, + "step": 29180 + }, + { + "epoch": 2.16, + "learning_rate": 1.482944440044226e-05, + "loss": 1.0415, + "step": 29181 + }, + { + "epoch": 2.16, + "learning_rate": 1.4829095124268981e-05, + "loss": 0.999, + "step": 29182 + }, + { + "epoch": 2.16, + "learning_rate": 1.4828745840412686e-05, + "loss": 1.006, + "step": 29183 + }, + { + "epoch": 2.16, + "learning_rate": 1.4828396548873922e-05, + "loss": 1.1049, + "step": 29184 + }, + { + "epoch": 2.16, + "learning_rate": 1.4828047249653249e-05, + "loss": 0.9792, + "step": 29185 + }, + { + "epoch": 2.16, + "learning_rate": 1.482769794275122e-05, + "loss": 0.9683, + "step": 29186 + }, + { + "epoch": 2.16, + "learning_rate": 1.4827348628168394e-05, + "loss": 0.9352, + "step": 29187 + }, + { + "epoch": 2.16, + "learning_rate": 1.4826999305905325e-05, + "loss": 1.022, + "step": 29188 + }, + { + "epoch": 2.16, + "learning_rate": 1.482664997596257e-05, + "loss": 0.9118, + "step": 29189 + }, + { + "epoch": 2.16, + "learning_rate": 1.4826300638340682e-05, + "loss": 1.0451, + "step": 29190 + }, + { + "epoch": 2.16, + "learning_rate": 1.4825951293040224e-05, + "loss": 1.0171, + "step": 29191 + }, + { + "epoch": 2.16, + "learning_rate": 1.4825601940061741e-05, + "loss": 0.9967, + "step": 29192 + }, + { + "epoch": 2.16, + "learning_rate": 1.4825252579405795e-05, + "loss": 0.9153, + "step": 29193 + }, + { + "epoch": 2.16, + "learning_rate": 1.4824903211072946e-05, + "loss": 1.0457, + "step": 29194 + }, + { + "epoch": 2.16, + "learning_rate": 1.482455383506374e-05, + "loss": 0.974, + "step": 29195 + }, + { + "epoch": 2.16, + "learning_rate": 1.482420445137874e-05, + "loss": 1.0412, + "step": 29196 + }, + { + "epoch": 2.16, + "learning_rate": 1.4823855060018497e-05, + "loss": 1.029, + "step": 29197 + }, + { + "epoch": 2.16, + "learning_rate": 1.4823505660983573e-05, + "loss": 1.0334, + "step": 29198 + }, + { + "epoch": 2.16, + "learning_rate": 1.4823156254274517e-05, + "loss": 1.0892, + "step": 29199 + }, + { + "epoch": 2.16, + "learning_rate": 1.4822806839891892e-05, + "loss": 0.9163, + "step": 29200 + }, + { + "epoch": 2.16, + "learning_rate": 1.4822457417836247e-05, + "loss": 1.0862, + "step": 29201 + }, + { + "epoch": 2.16, + "learning_rate": 1.4822107988108145e-05, + "loss": 1.0595, + "step": 29202 + }, + { + "epoch": 2.16, + "learning_rate": 1.4821758550708136e-05, + "loss": 1.0166, + "step": 29203 + }, + { + "epoch": 2.16, + "learning_rate": 1.4821409105636782e-05, + "loss": 0.9846, + "step": 29204 + }, + { + "epoch": 2.16, + "learning_rate": 1.482105965289463e-05, + "loss": 1.0872, + "step": 29205 + }, + { + "epoch": 2.16, + "learning_rate": 1.4820710192482241e-05, + "loss": 0.9505, + "step": 29206 + }, + { + "epoch": 2.16, + "learning_rate": 1.4820360724400173e-05, + "loss": 1.0533, + "step": 29207 + }, + { + "epoch": 2.16, + "learning_rate": 1.482001124864898e-05, + "loss": 0.8963, + "step": 29208 + }, + { + "epoch": 2.16, + "learning_rate": 1.4819661765229215e-05, + "loss": 1.0302, + "step": 29209 + }, + { + "epoch": 2.16, + "learning_rate": 1.4819312274141444e-05, + "loss": 1.0744, + "step": 29210 + }, + { + "epoch": 2.16, + "learning_rate": 1.4818962775386209e-05, + "loss": 1.0751, + "step": 29211 + }, + { + "epoch": 2.16, + "learning_rate": 1.4818613268964079e-05, + "loss": 0.9602, + "step": 29212 + }, + { + "epoch": 2.16, + "learning_rate": 1.4818263754875602e-05, + "loss": 1.0138, + "step": 29213 + }, + { + "epoch": 2.16, + "learning_rate": 1.4817914233121335e-05, + "loss": 0.9798, + "step": 29214 + }, + { + "epoch": 2.16, + "learning_rate": 1.4817564703701837e-05, + "loss": 0.9756, + "step": 29215 + }, + { + "epoch": 2.16, + "learning_rate": 1.4817215166617663e-05, + "loss": 1.1273, + "step": 29216 + }, + { + "epoch": 2.16, + "learning_rate": 1.4816865621869365e-05, + "loss": 1.1326, + "step": 29217 + }, + { + "epoch": 2.16, + "learning_rate": 1.4816516069457505e-05, + "loss": 1.0311, + "step": 29218 + }, + { + "epoch": 2.16, + "learning_rate": 1.4816166509382635e-05, + "loss": 0.9946, + "step": 29219 + }, + { + "epoch": 2.16, + "learning_rate": 1.4815816941645316e-05, + "loss": 0.9528, + "step": 29220 + }, + { + "epoch": 2.16, + "learning_rate": 1.48154673662461e-05, + "loss": 0.8738, + "step": 29221 + }, + { + "epoch": 2.16, + "learning_rate": 1.4815117783185546e-05, + "loss": 0.9493, + "step": 29222 + }, + { + "epoch": 2.16, + "learning_rate": 1.4814768192464204e-05, + "loss": 0.9845, + "step": 29223 + }, + { + "epoch": 2.16, + "learning_rate": 1.481441859408264e-05, + "loss": 1.0286, + "step": 29224 + }, + { + "epoch": 2.16, + "learning_rate": 1.4814068988041402e-05, + "loss": 1.0368, + "step": 29225 + }, + { + "epoch": 2.16, + "learning_rate": 1.481371937434105e-05, + "loss": 1.0783, + "step": 29226 + }, + { + "epoch": 2.16, + "learning_rate": 1.4813369752982137e-05, + "loss": 1.0338, + "step": 29227 + }, + { + "epoch": 2.16, + "learning_rate": 1.4813020123965225e-05, + "loss": 0.9446, + "step": 29228 + }, + { + "epoch": 2.16, + "learning_rate": 1.4812670487290862e-05, + "loss": 1.1174, + "step": 29229 + }, + { + "epoch": 2.16, + "learning_rate": 1.4812320842959613e-05, + "loss": 0.9873, + "step": 29230 + }, + { + "epoch": 2.16, + "learning_rate": 1.4811971190972032e-05, + "loss": 1.0283, + "step": 29231 + }, + { + "epoch": 2.16, + "learning_rate": 1.481162153132867e-05, + "loss": 1.0709, + "step": 29232 + }, + { + "epoch": 2.16, + "learning_rate": 1.481127186403009e-05, + "loss": 0.9532, + "step": 29233 + }, + { + "epoch": 2.16, + "learning_rate": 1.4810922189076844e-05, + "loss": 1.0535, + "step": 29234 + }, + { + "epoch": 2.16, + "learning_rate": 1.4810572506469487e-05, + "loss": 1.0224, + "step": 29235 + }, + { + "epoch": 2.16, + "learning_rate": 1.4810222816208581e-05, + "loss": 0.8581, + "step": 29236 + }, + { + "epoch": 2.16, + "learning_rate": 1.4809873118294678e-05, + "loss": 1.0474, + "step": 29237 + }, + { + "epoch": 2.16, + "learning_rate": 1.4809523412728337e-05, + "loss": 0.999, + "step": 29238 + }, + { + "epoch": 2.16, + "learning_rate": 1.4809173699510111e-05, + "loss": 0.9073, + "step": 29239 + }, + { + "epoch": 2.16, + "learning_rate": 1.480882397864056e-05, + "loss": 1.02, + "step": 29240 + }, + { + "epoch": 2.16, + "learning_rate": 1.4808474250120242e-05, + "loss": 1.0484, + "step": 29241 + }, + { + "epoch": 2.16, + "learning_rate": 1.4808124513949704e-05, + "loss": 0.9284, + "step": 29242 + }, + { + "epoch": 2.16, + "learning_rate": 1.4807774770129513e-05, + "loss": 1.1287, + "step": 29243 + }, + { + "epoch": 2.16, + "learning_rate": 1.4807425018660217e-05, + "loss": 0.9465, + "step": 29244 + }, + { + "epoch": 2.16, + "learning_rate": 1.4807075259542383e-05, + "loss": 1.0048, + "step": 29245 + }, + { + "epoch": 2.16, + "learning_rate": 1.4806725492776555e-05, + "loss": 0.9573, + "step": 29246 + }, + { + "epoch": 2.16, + "learning_rate": 1.4806375718363297e-05, + "loss": 1.0758, + "step": 29247 + }, + { + "epoch": 2.16, + "learning_rate": 1.4806025936303166e-05, + "loss": 0.9316, + "step": 29248 + }, + { + "epoch": 2.16, + "learning_rate": 1.4805676146596714e-05, + "loss": 1.013, + "step": 29249 + }, + { + "epoch": 2.16, + "learning_rate": 1.4805326349244505e-05, + "loss": 1.0399, + "step": 29250 + }, + { + "epoch": 2.16, + "learning_rate": 1.4804976544247086e-05, + "loss": 0.9625, + "step": 29251 + }, + { + "epoch": 2.16, + "learning_rate": 1.4804626731605018e-05, + "loss": 0.9882, + "step": 29252 + }, + { + "epoch": 2.16, + "learning_rate": 1.480427691131886e-05, + "loss": 0.9989, + "step": 29253 + }, + { + "epoch": 2.16, + "learning_rate": 1.4803927083389162e-05, + "loss": 0.926, + "step": 29254 + }, + { + "epoch": 2.16, + "learning_rate": 1.4803577247816489e-05, + "loss": 0.9085, + "step": 29255 + }, + { + "epoch": 2.16, + "learning_rate": 1.4803227404601392e-05, + "loss": 1.0798, + "step": 29256 + }, + { + "epoch": 2.16, + "learning_rate": 1.4802877553744429e-05, + "loss": 1.0171, + "step": 29257 + }, + { + "epoch": 2.16, + "learning_rate": 1.4802527695246154e-05, + "loss": 0.9491, + "step": 29258 + }, + { + "epoch": 2.16, + "learning_rate": 1.4802177829107132e-05, + "loss": 1.0224, + "step": 29259 + }, + { + "epoch": 2.16, + "learning_rate": 1.480182795532791e-05, + "loss": 0.94, + "step": 29260 + }, + { + "epoch": 2.16, + "learning_rate": 1.4801478073909048e-05, + "loss": 1.073, + "step": 29261 + }, + { + "epoch": 2.16, + "learning_rate": 1.4801128184851103e-05, + "loss": 0.9446, + "step": 29262 + }, + { + "epoch": 2.16, + "learning_rate": 1.4800778288154634e-05, + "loss": 0.9859, + "step": 29263 + }, + { + "epoch": 2.16, + "learning_rate": 1.4800428383820191e-05, + "loss": 0.988, + "step": 29264 + }, + { + "epoch": 2.16, + "learning_rate": 1.480007847184834e-05, + "loss": 1.0802, + "step": 29265 + }, + { + "epoch": 2.16, + "learning_rate": 1.479972855223963e-05, + "loss": 1.0384, + "step": 29266 + }, + { + "epoch": 2.16, + "learning_rate": 1.4799378624994623e-05, + "loss": 0.9228, + "step": 29267 + }, + { + "epoch": 2.16, + "learning_rate": 1.4799028690113873e-05, + "loss": 0.9778, + "step": 29268 + }, + { + "epoch": 2.16, + "learning_rate": 1.4798678747597936e-05, + "loss": 0.9828, + "step": 29269 + }, + { + "epoch": 2.16, + "learning_rate": 1.4798328797447368e-05, + "loss": 1.0316, + "step": 29270 + }, + { + "epoch": 2.16, + "learning_rate": 1.479797883966273e-05, + "loss": 1.0013, + "step": 29271 + }, + { + "epoch": 2.16, + "learning_rate": 1.4797628874244574e-05, + "loss": 1.0985, + "step": 29272 + }, + { + "epoch": 2.16, + "learning_rate": 1.4797278901193462e-05, + "loss": 1.0348, + "step": 29273 + }, + { + "epoch": 2.16, + "learning_rate": 1.4796928920509946e-05, + "loss": 1.0688, + "step": 29274 + }, + { + "epoch": 2.16, + "learning_rate": 1.4796578932194585e-05, + "loss": 0.9235, + "step": 29275 + }, + { + "epoch": 2.16, + "learning_rate": 1.4796228936247933e-05, + "loss": 1.0123, + "step": 29276 + }, + { + "epoch": 2.16, + "learning_rate": 1.4795878932670553e-05, + "loss": 0.9958, + "step": 29277 + }, + { + "epoch": 2.16, + "learning_rate": 1.4795528921462999e-05, + "loss": 0.9692, + "step": 29278 + }, + { + "epoch": 2.16, + "learning_rate": 1.4795178902625824e-05, + "loss": 1.1268, + "step": 29279 + }, + { + "epoch": 2.16, + "learning_rate": 1.479482887615959e-05, + "loss": 1.0011, + "step": 29280 + }, + { + "epoch": 2.16, + "learning_rate": 1.4794478842064852e-05, + "loss": 1.0125, + "step": 29281 + }, + { + "epoch": 2.16, + "learning_rate": 1.4794128800342165e-05, + "loss": 1.0587, + "step": 29282 + }, + { + "epoch": 2.16, + "learning_rate": 1.479377875099209e-05, + "loss": 1.0395, + "step": 29283 + }, + { + "epoch": 2.16, + "learning_rate": 1.4793428694015178e-05, + "loss": 0.9962, + "step": 29284 + }, + { + "epoch": 2.16, + "learning_rate": 1.4793078629411993e-05, + "loss": 0.9371, + "step": 29285 + }, + { + "epoch": 2.16, + "learning_rate": 1.4792728557183085e-05, + "loss": 0.9942, + "step": 29286 + }, + { + "epoch": 2.16, + "learning_rate": 1.4792378477329016e-05, + "loss": 0.9471, + "step": 29287 + }, + { + "epoch": 2.16, + "learning_rate": 1.4792028389850344e-05, + "loss": 1.0539, + "step": 29288 + }, + { + "epoch": 2.16, + "learning_rate": 1.4791678294747622e-05, + "loss": 1.01, + "step": 29289 + }, + { + "epoch": 2.16, + "learning_rate": 1.4791328192021409e-05, + "loss": 0.988, + "step": 29290 + }, + { + "epoch": 2.16, + "learning_rate": 1.4790978081672258e-05, + "loss": 1.0319, + "step": 29291 + }, + { + "epoch": 2.16, + "learning_rate": 1.4790627963700732e-05, + "loss": 1.0373, + "step": 29292 + }, + { + "epoch": 2.16, + "learning_rate": 1.4790277838107385e-05, + "loss": 0.936, + "step": 29293 + }, + { + "epoch": 2.16, + "learning_rate": 1.4789927704892775e-05, + "loss": 0.9976, + "step": 29294 + }, + { + "epoch": 2.16, + "learning_rate": 1.4789577564057458e-05, + "loss": 1.0148, + "step": 29295 + }, + { + "epoch": 2.16, + "learning_rate": 1.478922741560199e-05, + "loss": 1.0841, + "step": 29296 + }, + { + "epoch": 2.16, + "learning_rate": 1.4788877259526935e-05, + "loss": 1.0098, + "step": 29297 + }, + { + "epoch": 2.16, + "learning_rate": 1.4788527095832841e-05, + "loss": 0.9741, + "step": 29298 + }, + { + "epoch": 2.16, + "learning_rate": 1.4788176924520268e-05, + "loss": 0.95, + "step": 29299 + }, + { + "epoch": 2.16, + "learning_rate": 1.4787826745589775e-05, + "loss": 1.0246, + "step": 29300 + }, + { + "epoch": 2.17, + "learning_rate": 1.4787476559041918e-05, + "loss": 1.0744, + "step": 29301 + }, + { + "epoch": 2.17, + "learning_rate": 1.4787126364877256e-05, + "loss": 1.0261, + "step": 29302 + }, + { + "epoch": 2.17, + "learning_rate": 1.4786776163096341e-05, + "loss": 1.0518, + "step": 29303 + }, + { + "epoch": 2.17, + "learning_rate": 1.4786425953699736e-05, + "loss": 1.0449, + "step": 29304 + }, + { + "epoch": 2.17, + "learning_rate": 1.4786075736687996e-05, + "loss": 1.0637, + "step": 29305 + }, + { + "epoch": 2.17, + "learning_rate": 1.478572551206168e-05, + "loss": 0.9311, + "step": 29306 + }, + { + "epoch": 2.17, + "learning_rate": 1.478537527982134e-05, + "loss": 1.099, + "step": 29307 + }, + { + "epoch": 2.17, + "learning_rate": 1.4785025039967536e-05, + "loss": 1.0464, + "step": 29308 + }, + { + "epoch": 2.17, + "learning_rate": 1.4784674792500825e-05, + "loss": 1.0058, + "step": 29309 + }, + { + "epoch": 2.17, + "learning_rate": 1.4784324537421769e-05, + "loss": 0.9079, + "step": 29310 + }, + { + "epoch": 2.17, + "learning_rate": 1.4783974274730916e-05, + "loss": 1.0272, + "step": 29311 + }, + { + "epoch": 2.17, + "learning_rate": 1.4783624004428831e-05, + "loss": 0.9877, + "step": 29312 + }, + { + "epoch": 2.17, + "learning_rate": 1.478327372651607e-05, + "loss": 1.0458, + "step": 29313 + }, + { + "epoch": 2.17, + "learning_rate": 1.4782923440993186e-05, + "loss": 1.0413, + "step": 29314 + }, + { + "epoch": 2.17, + "learning_rate": 1.4782573147860744e-05, + "loss": 1.0092, + "step": 29315 + }, + { + "epoch": 2.17, + "learning_rate": 1.4782222847119295e-05, + "loss": 0.968, + "step": 29316 + }, + { + "epoch": 2.17, + "learning_rate": 1.4781872538769392e-05, + "loss": 1.033, + "step": 29317 + }, + { + "epoch": 2.17, + "learning_rate": 1.4781522222811604e-05, + "loss": 1.1222, + "step": 29318 + }, + { + "epoch": 2.17, + "learning_rate": 1.478117189924648e-05, + "loss": 0.9325, + "step": 29319 + }, + { + "epoch": 2.17, + "learning_rate": 1.4780821568074584e-05, + "loss": 1.0459, + "step": 29320 + }, + { + "epoch": 2.17, + "learning_rate": 1.4780471229296465e-05, + "loss": 1.097, + "step": 29321 + }, + { + "epoch": 2.17, + "learning_rate": 1.4780120882912686e-05, + "loss": 0.9536, + "step": 29322 + }, + { + "epoch": 2.17, + "learning_rate": 1.47797705289238e-05, + "loss": 1.0076, + "step": 29323 + }, + { + "epoch": 2.17, + "learning_rate": 1.4779420167330374e-05, + "loss": 0.9217, + "step": 29324 + }, + { + "epoch": 2.17, + "learning_rate": 1.4779069798132955e-05, + "loss": 1.0282, + "step": 29325 + }, + { + "epoch": 2.17, + "learning_rate": 1.4778719421332106e-05, + "loss": 0.9722, + "step": 29326 + }, + { + "epoch": 2.17, + "learning_rate": 1.477836903692838e-05, + "loss": 0.925, + "step": 29327 + }, + { + "epoch": 2.17, + "learning_rate": 1.4778018644922338e-05, + "loss": 0.9038, + "step": 29328 + }, + { + "epoch": 2.17, + "learning_rate": 1.4777668245314538e-05, + "loss": 0.9509, + "step": 29329 + }, + { + "epoch": 2.17, + "learning_rate": 1.4777317838105537e-05, + "loss": 1.0701, + "step": 29330 + }, + { + "epoch": 2.17, + "learning_rate": 1.477696742329589e-05, + "loss": 1.0283, + "step": 29331 + }, + { + "epoch": 2.17, + "learning_rate": 1.4776617000886156e-05, + "loss": 1.0454, + "step": 29332 + }, + { + "epoch": 2.17, + "learning_rate": 1.477626657087689e-05, + "loss": 1.0054, + "step": 29333 + }, + { + "epoch": 2.17, + "learning_rate": 1.4775916133268658e-05, + "loss": 1.0756, + "step": 29334 + }, + { + "epoch": 2.17, + "learning_rate": 1.4775565688062008e-05, + "loss": 0.9364, + "step": 29335 + }, + { + "epoch": 2.17, + "learning_rate": 1.4775215235257502e-05, + "loss": 1.1279, + "step": 29336 + }, + { + "epoch": 2.17, + "learning_rate": 1.4774864774855697e-05, + "loss": 1.0119, + "step": 29337 + }, + { + "epoch": 2.17, + "learning_rate": 1.4774514306857152e-05, + "loss": 1.0126, + "step": 29338 + }, + { + "epoch": 2.17, + "learning_rate": 1.4774163831262422e-05, + "loss": 0.9194, + "step": 29339 + }, + { + "epoch": 2.17, + "learning_rate": 1.4773813348072065e-05, + "loss": 1.1387, + "step": 29340 + }, + { + "epoch": 2.17, + "learning_rate": 1.477346285728664e-05, + "loss": 1.0796, + "step": 29341 + }, + { + "epoch": 2.17, + "learning_rate": 1.4773112358906705e-05, + "loss": 1.0279, + "step": 29342 + }, + { + "epoch": 2.17, + "learning_rate": 1.4772761852932815e-05, + "loss": 1.0981, + "step": 29343 + }, + { + "epoch": 2.17, + "learning_rate": 1.4772411339365529e-05, + "loss": 1.0688, + "step": 29344 + }, + { + "epoch": 2.17, + "learning_rate": 1.4772060818205405e-05, + "loss": 1.0333, + "step": 29345 + }, + { + "epoch": 2.17, + "learning_rate": 1.4771710289452998e-05, + "loss": 0.9626, + "step": 29346 + }, + { + "epoch": 2.17, + "learning_rate": 1.4771359753108874e-05, + "loss": 1.1067, + "step": 29347 + }, + { + "epoch": 2.17, + "learning_rate": 1.4771009209173581e-05, + "loss": 0.9699, + "step": 29348 + }, + { + "epoch": 2.17, + "learning_rate": 1.477065865764768e-05, + "loss": 1.0819, + "step": 29349 + }, + { + "epoch": 2.17, + "learning_rate": 1.4770308098531731e-05, + "loss": 0.9701, + "step": 29350 + }, + { + "epoch": 2.17, + "learning_rate": 1.4769957531826292e-05, + "loss": 1.01, + "step": 29351 + }, + { + "epoch": 2.17, + "learning_rate": 1.4769606957531915e-05, + "loss": 1.0627, + "step": 29352 + }, + { + "epoch": 2.17, + "learning_rate": 1.4769256375649165e-05, + "loss": 0.9542, + "step": 29353 + }, + { + "epoch": 2.17, + "learning_rate": 1.4768905786178594e-05, + "loss": 0.9633, + "step": 29354 + }, + { + "epoch": 2.17, + "learning_rate": 1.4768555189120763e-05, + "loss": 1.0296, + "step": 29355 + }, + { + "epoch": 2.17, + "learning_rate": 1.4768204584476227e-05, + "loss": 0.9767, + "step": 29356 + }, + { + "epoch": 2.17, + "learning_rate": 1.4767853972245546e-05, + "loss": 0.9489, + "step": 29357 + }, + { + "epoch": 2.17, + "learning_rate": 1.476750335242928e-05, + "loss": 0.9324, + "step": 29358 + }, + { + "epoch": 2.17, + "learning_rate": 1.4767152725027985e-05, + "loss": 1.0046, + "step": 29359 + }, + { + "epoch": 2.17, + "learning_rate": 1.4766802090042214e-05, + "loss": 1.0508, + "step": 29360 + }, + { + "epoch": 2.17, + "learning_rate": 1.4766451447472531e-05, + "loss": 0.9548, + "step": 29361 + }, + { + "epoch": 2.17, + "learning_rate": 1.4766100797319495e-05, + "loss": 0.9631, + "step": 29362 + }, + { + "epoch": 2.17, + "learning_rate": 1.4765750139583659e-05, + "loss": 1.026, + "step": 29363 + }, + { + "epoch": 2.17, + "learning_rate": 1.476539947426558e-05, + "loss": 1.1101, + "step": 29364 + }, + { + "epoch": 2.17, + "learning_rate": 1.4765048801365823e-05, + "loss": 1.0262, + "step": 29365 + }, + { + "epoch": 2.17, + "learning_rate": 1.4764698120884938e-05, + "loss": 0.9537, + "step": 29366 + }, + { + "epoch": 2.17, + "learning_rate": 1.4764347432823487e-05, + "loss": 0.9859, + "step": 29367 + }, + { + "epoch": 2.17, + "learning_rate": 1.4763996737182027e-05, + "loss": 1.0488, + "step": 29368 + }, + { + "epoch": 2.17, + "learning_rate": 1.4763646033961118e-05, + "loss": 1.0145, + "step": 29369 + }, + { + "epoch": 2.17, + "learning_rate": 1.4763295323161316e-05, + "loss": 0.9863, + "step": 29370 + }, + { + "epoch": 2.17, + "learning_rate": 1.476294460478318e-05, + "loss": 1.007, + "step": 29371 + }, + { + "epoch": 2.17, + "learning_rate": 1.4762593878827267e-05, + "loss": 1.0078, + "step": 29372 + }, + { + "epoch": 2.17, + "learning_rate": 1.4762243145294135e-05, + "loss": 0.9742, + "step": 29373 + }, + { + "epoch": 2.17, + "learning_rate": 1.4761892404184339e-05, + "loss": 1.0424, + "step": 29374 + }, + { + "epoch": 2.17, + "learning_rate": 1.4761541655498445e-05, + "loss": 1.0709, + "step": 29375 + }, + { + "epoch": 2.17, + "learning_rate": 1.4761190899237004e-05, + "loss": 1.0362, + "step": 29376 + }, + { + "epoch": 2.17, + "learning_rate": 1.4760840135400579e-05, + "loss": 0.9852, + "step": 29377 + }, + { + "epoch": 2.17, + "learning_rate": 1.476048936398972e-05, + "loss": 1.0814, + "step": 29378 + }, + { + "epoch": 2.17, + "learning_rate": 1.4760138585004994e-05, + "loss": 0.9494, + "step": 29379 + }, + { + "epoch": 2.17, + "learning_rate": 1.4759787798446956e-05, + "loss": 1.0079, + "step": 29380 + }, + { + "epoch": 2.17, + "learning_rate": 1.4759437004316164e-05, + "loss": 1.009, + "step": 29381 + }, + { + "epoch": 2.17, + "learning_rate": 1.4759086202613175e-05, + "loss": 0.9836, + "step": 29382 + }, + { + "epoch": 2.17, + "learning_rate": 1.475873539333855e-05, + "loss": 1.0625, + "step": 29383 + }, + { + "epoch": 2.17, + "learning_rate": 1.4758384576492842e-05, + "loss": 1.081, + "step": 29384 + }, + { + "epoch": 2.17, + "learning_rate": 1.4758033752076612e-05, + "loss": 1.049, + "step": 29385 + }, + { + "epoch": 2.17, + "learning_rate": 1.475768292009042e-05, + "loss": 1.0579, + "step": 29386 + }, + { + "epoch": 2.17, + "learning_rate": 1.4757332080534823e-05, + "loss": 1.055, + "step": 29387 + }, + { + "epoch": 2.17, + "learning_rate": 1.4756981233410377e-05, + "loss": 0.961, + "step": 29388 + }, + { + "epoch": 2.17, + "learning_rate": 1.4756630378717642e-05, + "loss": 1.0442, + "step": 29389 + }, + { + "epoch": 2.17, + "learning_rate": 1.475627951645718e-05, + "loss": 1.019, + "step": 29390 + }, + { + "epoch": 2.17, + "learning_rate": 1.475592864662954e-05, + "loss": 1.0578, + "step": 29391 + }, + { + "epoch": 2.17, + "learning_rate": 1.475557776923529e-05, + "loss": 1.0073, + "step": 29392 + }, + { + "epoch": 2.17, + "learning_rate": 1.475522688427498e-05, + "loss": 0.9818, + "step": 29393 + }, + { + "epoch": 2.17, + "learning_rate": 1.4754875991749173e-05, + "loss": 1.0006, + "step": 29394 + }, + { + "epoch": 2.17, + "learning_rate": 1.4754525091658427e-05, + "loss": 1.0329, + "step": 29395 + }, + { + "epoch": 2.17, + "learning_rate": 1.47541741840033e-05, + "loss": 1.0528, + "step": 29396 + }, + { + "epoch": 2.17, + "learning_rate": 1.475382326878435e-05, + "loss": 1.0704, + "step": 29397 + }, + { + "epoch": 2.17, + "learning_rate": 1.4753472346002134e-05, + "loss": 1.1037, + "step": 29398 + }, + { + "epoch": 2.17, + "learning_rate": 1.4753121415657213e-05, + "loss": 1.0062, + "step": 29399 + }, + { + "epoch": 2.17, + "learning_rate": 1.4752770477750145e-05, + "loss": 0.9942, + "step": 29400 + }, + { + "epoch": 2.17, + "learning_rate": 1.4752419532281484e-05, + "loss": 0.9834, + "step": 29401 + }, + { + "epoch": 2.17, + "learning_rate": 1.4752068579251792e-05, + "loss": 0.9795, + "step": 29402 + }, + { + "epoch": 2.17, + "learning_rate": 1.4751717618661627e-05, + "loss": 0.9628, + "step": 29403 + }, + { + "epoch": 2.17, + "learning_rate": 1.475136665051155e-05, + "loss": 1.0928, + "step": 29404 + }, + { + "epoch": 2.17, + "learning_rate": 1.4751015674802114e-05, + "loss": 1.0194, + "step": 29405 + }, + { + "epoch": 2.17, + "learning_rate": 1.475066469153388e-05, + "loss": 0.9468, + "step": 29406 + }, + { + "epoch": 2.17, + "learning_rate": 1.4750313700707407e-05, + "loss": 1.0718, + "step": 29407 + }, + { + "epoch": 2.17, + "learning_rate": 1.4749962702323252e-05, + "loss": 0.9538, + "step": 29408 + }, + { + "epoch": 2.17, + "learning_rate": 1.4749611696381978e-05, + "loss": 1.0694, + "step": 29409 + }, + { + "epoch": 2.17, + "learning_rate": 1.4749260682884137e-05, + "loss": 1.0424, + "step": 29410 + }, + { + "epoch": 2.17, + "learning_rate": 1.4748909661830288e-05, + "loss": 0.9632, + "step": 29411 + }, + { + "epoch": 2.17, + "learning_rate": 1.4748558633220996e-05, + "loss": 0.9439, + "step": 29412 + }, + { + "epoch": 2.17, + "learning_rate": 1.4748207597056812e-05, + "loss": 1.052, + "step": 29413 + }, + { + "epoch": 2.17, + "learning_rate": 1.4747856553338297e-05, + "loss": 0.9807, + "step": 29414 + }, + { + "epoch": 2.17, + "learning_rate": 1.4747505502066012e-05, + "loss": 0.9259, + "step": 29415 + }, + { + "epoch": 2.17, + "learning_rate": 1.4747154443240514e-05, + "loss": 1.1366, + "step": 29416 + }, + { + "epoch": 2.17, + "learning_rate": 1.474680337686236e-05, + "loss": 0.9024, + "step": 29417 + }, + { + "epoch": 2.17, + "learning_rate": 1.4746452302932115e-05, + "loss": 0.9949, + "step": 29418 + }, + { + "epoch": 2.17, + "learning_rate": 1.4746101221450327e-05, + "loss": 0.9776, + "step": 29419 + }, + { + "epoch": 2.17, + "learning_rate": 1.474575013241756e-05, + "loss": 1.021, + "step": 29420 + }, + { + "epoch": 2.17, + "learning_rate": 1.4745399035834371e-05, + "loss": 0.9744, + "step": 29421 + }, + { + "epoch": 2.17, + "learning_rate": 1.4745047931701324e-05, + "loss": 1.0135, + "step": 29422 + }, + { + "epoch": 2.17, + "learning_rate": 1.474469682001897e-05, + "loss": 1.0336, + "step": 29423 + }, + { + "epoch": 2.17, + "learning_rate": 1.4744345700787874e-05, + "loss": 0.9937, + "step": 29424 + }, + { + "epoch": 2.17, + "learning_rate": 1.4743994574008589e-05, + "loss": 1.0052, + "step": 29425 + }, + { + "epoch": 2.17, + "learning_rate": 1.4743643439681679e-05, + "loss": 0.9056, + "step": 29426 + }, + { + "epoch": 2.17, + "learning_rate": 1.47432922978077e-05, + "loss": 1.0026, + "step": 29427 + }, + { + "epoch": 2.17, + "learning_rate": 1.4742941148387212e-05, + "loss": 1.0433, + "step": 29428 + }, + { + "epoch": 2.17, + "learning_rate": 1.474258999142077e-05, + "loss": 1.0934, + "step": 29429 + }, + { + "epoch": 2.17, + "learning_rate": 1.4742238826908937e-05, + "loss": 0.9296, + "step": 29430 + }, + { + "epoch": 2.17, + "learning_rate": 1.4741887654852267e-05, + "loss": 0.9707, + "step": 29431 + }, + { + "epoch": 2.17, + "learning_rate": 1.4741536475251326e-05, + "loss": 1.0343, + "step": 29432 + }, + { + "epoch": 2.17, + "learning_rate": 1.4741185288106664e-05, + "loss": 0.9991, + "step": 29433 + }, + { + "epoch": 2.17, + "learning_rate": 1.4740834093418846e-05, + "loss": 1.0764, + "step": 29434 + }, + { + "epoch": 2.17, + "learning_rate": 1.4740482891188425e-05, + "loss": 0.974, + "step": 29435 + }, + { + "epoch": 2.18, + "learning_rate": 1.4740131681415971e-05, + "loss": 1.1459, + "step": 29436 + }, + { + "epoch": 2.18, + "learning_rate": 1.4739780464102031e-05, + "loss": 1.0306, + "step": 29437 + }, + { + "epoch": 2.18, + "learning_rate": 1.4739429239247167e-05, + "loss": 1.0577, + "step": 29438 + }, + { + "epoch": 2.18, + "learning_rate": 1.4739078006851942e-05, + "loss": 0.9715, + "step": 29439 + }, + { + "epoch": 2.18, + "learning_rate": 1.4738726766916908e-05, + "loss": 1.0471, + "step": 29440 + }, + { + "epoch": 2.18, + "learning_rate": 1.473837551944263e-05, + "loss": 1.0582, + "step": 29441 + }, + { + "epoch": 2.18, + "learning_rate": 1.4738024264429665e-05, + "loss": 0.9558, + "step": 29442 + }, + { + "epoch": 2.18, + "learning_rate": 1.4737673001878569e-05, + "loss": 0.9767, + "step": 29443 + }, + { + "epoch": 2.18, + "learning_rate": 1.4737321731789906e-05, + "loss": 0.9443, + "step": 29444 + }, + { + "epoch": 2.18, + "learning_rate": 1.4736970454164227e-05, + "loss": 1.0017, + "step": 29445 + }, + { + "epoch": 2.18, + "learning_rate": 1.4736619169002102e-05, + "loss": 1.0015, + "step": 29446 + }, + { + "epoch": 2.18, + "learning_rate": 1.473626787630408e-05, + "loss": 0.9258, + "step": 29447 + }, + { + "epoch": 2.18, + "learning_rate": 1.4735916576070724e-05, + "loss": 1.0874, + "step": 29448 + }, + { + "epoch": 2.18, + "learning_rate": 1.4735565268302595e-05, + "loss": 0.9668, + "step": 29449 + }, + { + "epoch": 2.18, + "learning_rate": 1.4735213953000245e-05, + "loss": 0.8945, + "step": 29450 + }, + { + "epoch": 2.18, + "learning_rate": 1.473486263016424e-05, + "loss": 1.066, + "step": 29451 + }, + { + "epoch": 2.18, + "learning_rate": 1.4734511299795134e-05, + "loss": 1.0747, + "step": 29452 + }, + { + "epoch": 2.18, + "learning_rate": 1.4734159961893491e-05, + "loss": 0.9928, + "step": 29453 + }, + { + "epoch": 2.18, + "learning_rate": 1.4733808616459864e-05, + "loss": 1.0236, + "step": 29454 + }, + { + "epoch": 2.18, + "learning_rate": 1.473345726349482e-05, + "loss": 0.9159, + "step": 29455 + }, + { + "epoch": 2.18, + "learning_rate": 1.473310590299891e-05, + "loss": 1.1031, + "step": 29456 + }, + { + "epoch": 2.18, + "learning_rate": 1.4732754534972697e-05, + "loss": 1.0717, + "step": 29457 + }, + { + "epoch": 2.18, + "learning_rate": 1.473240315941674e-05, + "loss": 0.9845, + "step": 29458 + }, + { + "epoch": 2.18, + "learning_rate": 1.4732051776331595e-05, + "loss": 1.0599, + "step": 29459 + }, + { + "epoch": 2.18, + "learning_rate": 1.4731700385717825e-05, + "loss": 0.8675, + "step": 29460 + }, + { + "epoch": 2.18, + "learning_rate": 1.4731348987575988e-05, + "loss": 1.0142, + "step": 29461 + }, + { + "epoch": 2.18, + "learning_rate": 1.4730997581906641e-05, + "loss": 1.0002, + "step": 29462 + }, + { + "epoch": 2.18, + "learning_rate": 1.4730646168710345e-05, + "loss": 1.0066, + "step": 29463 + }, + { + "epoch": 2.18, + "learning_rate": 1.4730294747987657e-05, + "loss": 0.9622, + "step": 29464 + }, + { + "epoch": 2.18, + "learning_rate": 1.4729943319739142e-05, + "loss": 1.0652, + "step": 29465 + }, + { + "epoch": 2.18, + "learning_rate": 1.4729591883965353e-05, + "loss": 1.1017, + "step": 29466 + }, + { + "epoch": 2.18, + "learning_rate": 1.472924044066685e-05, + "loss": 1.0686, + "step": 29467 + }, + { + "epoch": 2.18, + "learning_rate": 1.4728888989844192e-05, + "loss": 1.0642, + "step": 29468 + }, + { + "epoch": 2.18, + "learning_rate": 1.4728537531497942e-05, + "loss": 1.0782, + "step": 29469 + }, + { + "epoch": 2.18, + "learning_rate": 1.4728186065628654e-05, + "loss": 0.9398, + "step": 29470 + }, + { + "epoch": 2.18, + "learning_rate": 1.4727834592236891e-05, + "loss": 0.9594, + "step": 29471 + }, + { + "epoch": 2.18, + "learning_rate": 1.472748311132321e-05, + "loss": 1.1136, + "step": 29472 + }, + { + "epoch": 2.18, + "learning_rate": 1.4727131622888173e-05, + "loss": 0.9887, + "step": 29473 + }, + { + "epoch": 2.18, + "learning_rate": 1.4726780126932337e-05, + "loss": 0.8953, + "step": 29474 + }, + { + "epoch": 2.18, + "learning_rate": 1.472642862345626e-05, + "loss": 1.036, + "step": 29475 + }, + { + "epoch": 2.18, + "learning_rate": 1.4726077112460502e-05, + "loss": 0.9203, + "step": 29476 + }, + { + "epoch": 2.18, + "learning_rate": 1.4725725593945624e-05, + "loss": 1.0224, + "step": 29477 + }, + { + "epoch": 2.18, + "learning_rate": 1.4725374067912183e-05, + "loss": 0.8985, + "step": 29478 + }, + { + "epoch": 2.18, + "learning_rate": 1.4725022534360742e-05, + "loss": 1.0112, + "step": 29479 + }, + { + "epoch": 2.18, + "learning_rate": 1.4724670993291854e-05, + "loss": 0.9697, + "step": 29480 + }, + { + "epoch": 2.18, + "learning_rate": 1.4724319444706086e-05, + "loss": 1.0808, + "step": 29481 + }, + { + "epoch": 2.18, + "learning_rate": 1.4723967888603992e-05, + "loss": 0.9252, + "step": 29482 + }, + { + "epoch": 2.18, + "learning_rate": 1.4723616324986131e-05, + "loss": 1.0282, + "step": 29483 + }, + { + "epoch": 2.18, + "learning_rate": 1.4723264753853065e-05, + "loss": 0.8902, + "step": 29484 + }, + { + "epoch": 2.18, + "learning_rate": 1.4722913175205353e-05, + "loss": 1.0647, + "step": 29485 + }, + { + "epoch": 2.18, + "learning_rate": 1.4722561589043554e-05, + "loss": 0.9105, + "step": 29486 + }, + { + "epoch": 2.18, + "learning_rate": 1.4722209995368229e-05, + "loss": 1.068, + "step": 29487 + }, + { + "epoch": 2.18, + "learning_rate": 1.472185839417993e-05, + "loss": 1.077, + "step": 29488 + }, + { + "epoch": 2.18, + "learning_rate": 1.4721506785479227e-05, + "loss": 1.1319, + "step": 29489 + }, + { + "epoch": 2.18, + "learning_rate": 1.472115516926667e-05, + "loss": 0.9763, + "step": 29490 + }, + { + "epoch": 2.18, + "learning_rate": 1.4720803545542826e-05, + "loss": 1.0335, + "step": 29491 + }, + { + "epoch": 2.18, + "learning_rate": 1.472045191430825e-05, + "loss": 0.9941, + "step": 29492 + }, + { + "epoch": 2.18, + "learning_rate": 1.4720100275563502e-05, + "loss": 0.9414, + "step": 29493 + }, + { + "epoch": 2.18, + "learning_rate": 1.4719748629309142e-05, + "loss": 0.9725, + "step": 29494 + }, + { + "epoch": 2.18, + "learning_rate": 1.471939697554573e-05, + "loss": 0.9878, + "step": 29495 + }, + { + "epoch": 2.18, + "learning_rate": 1.4719045314273825e-05, + "loss": 1.0126, + "step": 29496 + }, + { + "epoch": 2.18, + "learning_rate": 1.4718693645493986e-05, + "loss": 1.0931, + "step": 29497 + }, + { + "epoch": 2.18, + "learning_rate": 1.4718341969206774e-05, + "loss": 0.9778, + "step": 29498 + }, + { + "epoch": 2.18, + "learning_rate": 1.4717990285412744e-05, + "loss": 0.8998, + "step": 29499 + }, + { + "epoch": 2.18, + "learning_rate": 1.4717638594112462e-05, + "loss": 0.8866, + "step": 29500 + }, + { + "epoch": 2.18, + "learning_rate": 1.4717286895306484e-05, + "loss": 0.9839, + "step": 29501 + }, + { + "epoch": 2.18, + "learning_rate": 1.4716935188995374e-05, + "loss": 0.9366, + "step": 29502 + }, + { + "epoch": 2.18, + "learning_rate": 1.471658347517968e-05, + "loss": 1.0236, + "step": 29503 + }, + { + "epoch": 2.18, + "learning_rate": 1.4716231753859975e-05, + "loss": 1.0326, + "step": 29504 + }, + { + "epoch": 2.18, + "learning_rate": 1.4715880025036811e-05, + "loss": 1.0225, + "step": 29505 + }, + { + "epoch": 2.18, + "learning_rate": 1.4715528288710748e-05, + "loss": 0.9354, + "step": 29506 + }, + { + "epoch": 2.18, + "learning_rate": 1.4715176544882348e-05, + "loss": 0.909, + "step": 29507 + }, + { + "epoch": 2.18, + "learning_rate": 1.4714824793552169e-05, + "loss": 1.0752, + "step": 29508 + }, + { + "epoch": 2.18, + "learning_rate": 1.471447303472077e-05, + "loss": 0.9604, + "step": 29509 + }, + { + "epoch": 2.18, + "learning_rate": 1.4714121268388714e-05, + "loss": 1.0132, + "step": 29510 + }, + { + "epoch": 2.18, + "learning_rate": 1.4713769494556558e-05, + "loss": 0.819, + "step": 29511 + }, + { + "epoch": 2.18, + "learning_rate": 1.4713417713224864e-05, + "loss": 0.9354, + "step": 29512 + }, + { + "epoch": 2.18, + "learning_rate": 1.4713065924394188e-05, + "loss": 1.0367, + "step": 29513 + }, + { + "epoch": 2.18, + "learning_rate": 1.4712714128065094e-05, + "loss": 1.0378, + "step": 29514 + }, + { + "epoch": 2.18, + "learning_rate": 1.4712362324238135e-05, + "loss": 0.9011, + "step": 29515 + }, + { + "epoch": 2.18, + "learning_rate": 1.4712010512913877e-05, + "loss": 0.9285, + "step": 29516 + }, + { + "epoch": 2.18, + "learning_rate": 1.4711658694092877e-05, + "loss": 0.9126, + "step": 29517 + }, + { + "epoch": 2.18, + "learning_rate": 1.4711306867775697e-05, + "loss": 1.0431, + "step": 29518 + }, + { + "epoch": 2.18, + "learning_rate": 1.4710955033962893e-05, + "loss": 0.9852, + "step": 29519 + }, + { + "epoch": 2.18, + "learning_rate": 1.471060319265503e-05, + "loss": 0.9844, + "step": 29520 + }, + { + "epoch": 2.18, + "learning_rate": 1.4710251343852662e-05, + "loss": 1.0225, + "step": 29521 + }, + { + "epoch": 2.18, + "learning_rate": 1.4709899487556354e-05, + "loss": 1.0632, + "step": 29522 + }, + { + "epoch": 2.18, + "learning_rate": 1.470954762376666e-05, + "loss": 1.0188, + "step": 29523 + }, + { + "epoch": 2.18, + "learning_rate": 1.4709195752484144e-05, + "loss": 1.078, + "step": 29524 + }, + { + "epoch": 2.18, + "learning_rate": 1.4708843873709367e-05, + "loss": 0.799, + "step": 29525 + }, + { + "epoch": 2.18, + "learning_rate": 1.4708491987442885e-05, + "loss": 0.9008, + "step": 29526 + }, + { + "epoch": 2.18, + "learning_rate": 1.470814009368526e-05, + "loss": 0.9828, + "step": 29527 + }, + { + "epoch": 2.18, + "learning_rate": 1.4707788192437052e-05, + "loss": 1.1393, + "step": 29528 + }, + { + "epoch": 2.18, + "learning_rate": 1.4707436283698818e-05, + "loss": 0.9861, + "step": 29529 + }, + { + "epoch": 2.18, + "learning_rate": 1.4707084367471122e-05, + "loss": 1.0065, + "step": 29530 + }, + { + "epoch": 2.18, + "learning_rate": 1.4706732443754524e-05, + "loss": 0.9443, + "step": 29531 + }, + { + "epoch": 2.18, + "learning_rate": 1.4706380512549581e-05, + "loss": 1.0681, + "step": 29532 + }, + { + "epoch": 2.18, + "learning_rate": 1.4706028573856852e-05, + "loss": 0.9764, + "step": 29533 + }, + { + "epoch": 2.18, + "learning_rate": 1.4705676627676901e-05, + "loss": 1.1243, + "step": 29534 + }, + { + "epoch": 2.18, + "learning_rate": 1.4705324674010285e-05, + "loss": 0.9905, + "step": 29535 + }, + { + "epoch": 2.18, + "learning_rate": 1.4704972712857564e-05, + "loss": 0.959, + "step": 29536 + }, + { + "epoch": 2.18, + "learning_rate": 1.4704620744219298e-05, + "loss": 1.018, + "step": 29537 + }, + { + "epoch": 2.18, + "learning_rate": 1.4704268768096051e-05, + "loss": 1.091, + "step": 29538 + }, + { + "epoch": 2.18, + "learning_rate": 1.4703916784488375e-05, + "loss": 0.9507, + "step": 29539 + }, + { + "epoch": 2.18, + "learning_rate": 1.4703564793396838e-05, + "loss": 1.1385, + "step": 29540 + }, + { + "epoch": 2.18, + "learning_rate": 1.4703212794821999e-05, + "loss": 1.0024, + "step": 29541 + }, + { + "epoch": 2.18, + "learning_rate": 1.470286078876441e-05, + "loss": 0.9476, + "step": 29542 + }, + { + "epoch": 2.18, + "learning_rate": 1.4702508775224642e-05, + "loss": 1.0647, + "step": 29543 + }, + { + "epoch": 2.18, + "learning_rate": 1.4702156754203245e-05, + "loss": 1.0048, + "step": 29544 + }, + { + "epoch": 2.18, + "learning_rate": 1.4701804725700788e-05, + "loss": 1.0353, + "step": 29545 + }, + { + "epoch": 2.18, + "learning_rate": 1.4701452689717823e-05, + "loss": 0.9725, + "step": 29546 + }, + { + "epoch": 2.18, + "learning_rate": 1.4701100646254918e-05, + "loss": 0.9125, + "step": 29547 + }, + { + "epoch": 2.18, + "learning_rate": 1.4700748595312629e-05, + "loss": 0.9511, + "step": 29548 + }, + { + "epoch": 2.18, + "learning_rate": 1.4700396536891515e-05, + "loss": 0.965, + "step": 29549 + }, + { + "epoch": 2.18, + "learning_rate": 1.4700044470992137e-05, + "loss": 0.9877, + "step": 29550 + }, + { + "epoch": 2.18, + "learning_rate": 1.4699692397615056e-05, + "loss": 0.9949, + "step": 29551 + }, + { + "epoch": 2.18, + "learning_rate": 1.4699340316760831e-05, + "loss": 0.9241, + "step": 29552 + }, + { + "epoch": 2.18, + "learning_rate": 1.4698988228430022e-05, + "loss": 0.9861, + "step": 29553 + }, + { + "epoch": 2.18, + "learning_rate": 1.469863613262319e-05, + "loss": 1.1127, + "step": 29554 + }, + { + "epoch": 2.18, + "learning_rate": 1.4698284029340897e-05, + "loss": 0.8584, + "step": 29555 + }, + { + "epoch": 2.18, + "learning_rate": 1.46979319185837e-05, + "loss": 0.9703, + "step": 29556 + }, + { + "epoch": 2.18, + "learning_rate": 1.4697579800352163e-05, + "loss": 1.0925, + "step": 29557 + }, + { + "epoch": 2.18, + "learning_rate": 1.469722767464684e-05, + "loss": 1.0546, + "step": 29558 + }, + { + "epoch": 2.18, + "learning_rate": 1.4696875541468298e-05, + "loss": 0.9915, + "step": 29559 + }, + { + "epoch": 2.18, + "learning_rate": 1.4696523400817094e-05, + "loss": 0.9716, + "step": 29560 + }, + { + "epoch": 2.18, + "learning_rate": 1.4696171252693786e-05, + "loss": 1.0348, + "step": 29561 + }, + { + "epoch": 2.18, + "learning_rate": 1.4695819097098939e-05, + "loss": 1.072, + "step": 29562 + }, + { + "epoch": 2.18, + "learning_rate": 1.469546693403311e-05, + "loss": 1.0143, + "step": 29563 + }, + { + "epoch": 2.18, + "learning_rate": 1.4695114763496858e-05, + "loss": 1.0258, + "step": 29564 + }, + { + "epoch": 2.18, + "learning_rate": 1.469476258549075e-05, + "loss": 0.945, + "step": 29565 + }, + { + "epoch": 2.18, + "learning_rate": 1.469441040001534e-05, + "loss": 0.9268, + "step": 29566 + }, + { + "epoch": 2.18, + "learning_rate": 1.4694058207071189e-05, + "loss": 0.9434, + "step": 29567 + }, + { + "epoch": 2.18, + "learning_rate": 1.469370600665886e-05, + "loss": 1.0339, + "step": 29568 + }, + { + "epoch": 2.18, + "learning_rate": 1.4693353798778914e-05, + "loss": 0.9593, + "step": 29569 + }, + { + "epoch": 2.18, + "learning_rate": 1.4693001583431906e-05, + "loss": 0.9322, + "step": 29570 + }, + { + "epoch": 2.18, + "learning_rate": 1.46926493606184e-05, + "loss": 1.0304, + "step": 29571 + }, + { + "epoch": 2.19, + "learning_rate": 1.4692297130338955e-05, + "loss": 1.0876, + "step": 29572 + }, + { + "epoch": 2.19, + "learning_rate": 1.4691944892594134e-05, + "loss": 0.8953, + "step": 29573 + }, + { + "epoch": 2.19, + "learning_rate": 1.4691592647384496e-05, + "loss": 0.959, + "step": 29574 + }, + { + "epoch": 2.19, + "learning_rate": 1.4691240394710601e-05, + "loss": 1.0467, + "step": 29575 + }, + { + "epoch": 2.19, + "learning_rate": 1.4690888134573009e-05, + "loss": 0.9621, + "step": 29576 + }, + { + "epoch": 2.19, + "learning_rate": 1.4690535866972284e-05, + "loss": 1.022, + "step": 29577 + }, + { + "epoch": 2.19, + "learning_rate": 1.4690183591908982e-05, + "loss": 0.979, + "step": 29578 + }, + { + "epoch": 2.19, + "learning_rate": 1.4689831309383664e-05, + "loss": 1.065, + "step": 29579 + }, + { + "epoch": 2.19, + "learning_rate": 1.468947901939689e-05, + "loss": 1.0154, + "step": 29580 + }, + { + "epoch": 2.19, + "learning_rate": 1.4689126721949224e-05, + "loss": 0.9651, + "step": 29581 + }, + { + "epoch": 2.19, + "learning_rate": 1.4688774417041223e-05, + "loss": 0.9303, + "step": 29582 + }, + { + "epoch": 2.19, + "learning_rate": 1.4688422104673452e-05, + "loss": 0.8048, + "step": 29583 + }, + { + "epoch": 2.19, + "learning_rate": 1.4688069784846466e-05, + "loss": 1.0197, + "step": 29584 + }, + { + "epoch": 2.19, + "learning_rate": 1.4687717457560828e-05, + "loss": 0.9858, + "step": 29585 + }, + { + "epoch": 2.19, + "learning_rate": 1.4687365122817101e-05, + "loss": 1.102, + "step": 29586 + }, + { + "epoch": 2.19, + "learning_rate": 1.468701278061584e-05, + "loss": 1.0634, + "step": 29587 + }, + { + "epoch": 2.19, + "learning_rate": 1.468666043095761e-05, + "loss": 0.9984, + "step": 29588 + }, + { + "epoch": 2.19, + "learning_rate": 1.468630807384297e-05, + "loss": 0.9137, + "step": 29589 + }, + { + "epoch": 2.19, + "learning_rate": 1.4685955709272484e-05, + "loss": 1.028, + "step": 29590 + }, + { + "epoch": 2.19, + "learning_rate": 1.4685603337246705e-05, + "loss": 0.9598, + "step": 29591 + }, + { + "epoch": 2.19, + "learning_rate": 1.4685250957766201e-05, + "loss": 0.9476, + "step": 29592 + }, + { + "epoch": 2.19, + "learning_rate": 1.468489857083153e-05, + "loss": 1.0119, + "step": 29593 + }, + { + "epoch": 2.19, + "learning_rate": 1.468454617644325e-05, + "loss": 1.0455, + "step": 29594 + }, + { + "epoch": 2.19, + "learning_rate": 1.4684193774601925e-05, + "loss": 0.9361, + "step": 29595 + }, + { + "epoch": 2.19, + "learning_rate": 1.4683841365308116e-05, + "loss": 1.0738, + "step": 29596 + }, + { + "epoch": 2.19, + "learning_rate": 1.4683488948562382e-05, + "loss": 1.0053, + "step": 29597 + }, + { + "epoch": 2.19, + "learning_rate": 1.4683136524365282e-05, + "loss": 1.095, + "step": 29598 + }, + { + "epoch": 2.19, + "learning_rate": 1.4682784092717379e-05, + "loss": 0.9208, + "step": 29599 + }, + { + "epoch": 2.19, + "learning_rate": 1.4682431653619235e-05, + "loss": 1.0816, + "step": 29600 + }, + { + "epoch": 2.19, + "learning_rate": 1.468207920707141e-05, + "loss": 0.9617, + "step": 29601 + }, + { + "epoch": 2.19, + "learning_rate": 1.4681726753074464e-05, + "loss": 1.0664, + "step": 29602 + }, + { + "epoch": 2.19, + "learning_rate": 1.4681374291628956e-05, + "loss": 0.9552, + "step": 29603 + }, + { + "epoch": 2.19, + "learning_rate": 1.4681021822735448e-05, + "loss": 1.0275, + "step": 29604 + }, + { + "epoch": 2.19, + "learning_rate": 1.4680669346394503e-05, + "loss": 1.0935, + "step": 29605 + }, + { + "epoch": 2.19, + "learning_rate": 1.468031686260668e-05, + "loss": 1.1453, + "step": 29606 + }, + { + "epoch": 2.19, + "learning_rate": 1.467996437137254e-05, + "loss": 0.8929, + "step": 29607 + }, + { + "epoch": 2.19, + "learning_rate": 1.4679611872692642e-05, + "loss": 1.0492, + "step": 29608 + }, + { + "epoch": 2.19, + "learning_rate": 1.4679259366567551e-05, + "loss": 1.0614, + "step": 29609 + }, + { + "epoch": 2.19, + "learning_rate": 1.4678906852997823e-05, + "loss": 0.9483, + "step": 29610 + }, + { + "epoch": 2.19, + "learning_rate": 1.4678554331984021e-05, + "loss": 1.0431, + "step": 29611 + }, + { + "epoch": 2.19, + "learning_rate": 1.4678201803526708e-05, + "loss": 0.9144, + "step": 29612 + }, + { + "epoch": 2.19, + "learning_rate": 1.467784926762644e-05, + "loss": 1.1241, + "step": 29613 + }, + { + "epoch": 2.19, + "learning_rate": 1.4677496724283784e-05, + "loss": 0.9359, + "step": 29614 + }, + { + "epoch": 2.19, + "learning_rate": 1.4677144173499296e-05, + "loss": 0.9579, + "step": 29615 + }, + { + "epoch": 2.19, + "learning_rate": 1.467679161527354e-05, + "loss": 0.9977, + "step": 29616 + }, + { + "epoch": 2.19, + "learning_rate": 1.4676439049607071e-05, + "loss": 1.1671, + "step": 29617 + }, + { + "epoch": 2.19, + "learning_rate": 1.467608647650046e-05, + "loss": 0.9706, + "step": 29618 + }, + { + "epoch": 2.19, + "learning_rate": 1.4675733895954258e-05, + "loss": 1.0568, + "step": 29619 + }, + { + "epoch": 2.19, + "learning_rate": 1.4675381307969035e-05, + "loss": 0.9264, + "step": 29620 + }, + { + "epoch": 2.19, + "learning_rate": 1.4675028712545341e-05, + "loss": 0.9449, + "step": 29621 + }, + { + "epoch": 2.19, + "learning_rate": 1.4674676109683748e-05, + "loss": 0.9963, + "step": 29622 + }, + { + "epoch": 2.19, + "learning_rate": 1.467432349938481e-05, + "loss": 0.9969, + "step": 29623 + }, + { + "epoch": 2.19, + "learning_rate": 1.4673970881649093e-05, + "loss": 0.9804, + "step": 29624 + }, + { + "epoch": 2.19, + "learning_rate": 1.4673618256477152e-05, + "loss": 0.8641, + "step": 29625 + }, + { + "epoch": 2.19, + "learning_rate": 1.4673265623869552e-05, + "loss": 0.9878, + "step": 29626 + }, + { + "epoch": 2.19, + "learning_rate": 1.4672912983826855e-05, + "loss": 1.0206, + "step": 29627 + }, + { + "epoch": 2.19, + "learning_rate": 1.4672560336349619e-05, + "loss": 0.9879, + "step": 29628 + }, + { + "epoch": 2.19, + "learning_rate": 1.4672207681438403e-05, + "loss": 1.0051, + "step": 29629 + }, + { + "epoch": 2.19, + "learning_rate": 1.4671855019093777e-05, + "loss": 1.0147, + "step": 29630 + }, + { + "epoch": 2.19, + "learning_rate": 1.4671502349316293e-05, + "loss": 0.9935, + "step": 29631 + }, + { + "epoch": 2.19, + "learning_rate": 1.4671149672106517e-05, + "loss": 1.0321, + "step": 29632 + }, + { + "epoch": 2.19, + "learning_rate": 1.467079698746501e-05, + "loss": 0.9075, + "step": 29633 + }, + { + "epoch": 2.19, + "learning_rate": 1.467044429539233e-05, + "loss": 0.9488, + "step": 29634 + }, + { + "epoch": 2.19, + "learning_rate": 1.467009159588904e-05, + "loss": 0.9439, + "step": 29635 + }, + { + "epoch": 2.19, + "learning_rate": 1.46697388889557e-05, + "loss": 1.1412, + "step": 29636 + }, + { + "epoch": 2.19, + "learning_rate": 1.4669386174592874e-05, + "loss": 1.072, + "step": 29637 + }, + { + "epoch": 2.19, + "learning_rate": 1.4669033452801123e-05, + "loss": 0.9623, + "step": 29638 + }, + { + "epoch": 2.19, + "learning_rate": 1.4668680723581003e-05, + "loss": 0.9871, + "step": 29639 + }, + { + "epoch": 2.19, + "learning_rate": 1.4668327986933082e-05, + "loss": 0.9576, + "step": 29640 + }, + { + "epoch": 2.19, + "learning_rate": 1.4667975242857915e-05, + "loss": 1.0091, + "step": 29641 + }, + { + "epoch": 2.19, + "learning_rate": 1.4667622491356068e-05, + "loss": 1.0642, + "step": 29642 + }, + { + "epoch": 2.19, + "learning_rate": 1.46672697324281e-05, + "loss": 0.8903, + "step": 29643 + }, + { + "epoch": 2.19, + "learning_rate": 1.4666916966074572e-05, + "loss": 1.0613, + "step": 29644 + }, + { + "epoch": 2.19, + "learning_rate": 1.4666564192296047e-05, + "loss": 1.0508, + "step": 29645 + }, + { + "epoch": 2.19, + "learning_rate": 1.4666211411093082e-05, + "loss": 1.0049, + "step": 29646 + }, + { + "epoch": 2.19, + "learning_rate": 1.4665858622466247e-05, + "loss": 1.0001, + "step": 29647 + }, + { + "epoch": 2.19, + "learning_rate": 1.4665505826416093e-05, + "loss": 0.9316, + "step": 29648 + }, + { + "epoch": 2.19, + "learning_rate": 1.4665153022943188e-05, + "loss": 1.0257, + "step": 29649 + }, + { + "epoch": 2.19, + "learning_rate": 1.4664800212048089e-05, + "loss": 1.0316, + "step": 29650 + }, + { + "epoch": 2.19, + "learning_rate": 1.4664447393731363e-05, + "loss": 1.0622, + "step": 29651 + }, + { + "epoch": 2.19, + "learning_rate": 1.4664094567993568e-05, + "loss": 1.0778, + "step": 29652 + }, + { + "epoch": 2.19, + "learning_rate": 1.4663741734835263e-05, + "loss": 0.9429, + "step": 29653 + }, + { + "epoch": 2.19, + "learning_rate": 1.466338889425701e-05, + "loss": 1.0203, + "step": 29654 + }, + { + "epoch": 2.19, + "learning_rate": 1.4663036046259378e-05, + "loss": 1.0455, + "step": 29655 + }, + { + "epoch": 2.19, + "learning_rate": 1.4662683190842916e-05, + "loss": 1.0263, + "step": 29656 + }, + { + "epoch": 2.19, + "learning_rate": 1.4662330328008197e-05, + "loss": 0.9373, + "step": 29657 + }, + { + "epoch": 2.19, + "learning_rate": 1.4661977457755771e-05, + "loss": 1.0544, + "step": 29658 + }, + { + "epoch": 2.19, + "learning_rate": 1.466162458008621e-05, + "loss": 1.0022, + "step": 29659 + }, + { + "epoch": 2.19, + "learning_rate": 1.466127169500007e-05, + "loss": 0.9745, + "step": 29660 + }, + { + "epoch": 2.19, + "learning_rate": 1.4660918802497916e-05, + "loss": 1.0508, + "step": 29661 + }, + { + "epoch": 2.19, + "learning_rate": 1.4660565902580303e-05, + "loss": 0.9335, + "step": 29662 + }, + { + "epoch": 2.19, + "learning_rate": 1.46602129952478e-05, + "loss": 1.1147, + "step": 29663 + }, + { + "epoch": 2.19, + "learning_rate": 1.465986008050096e-05, + "loss": 0.9218, + "step": 29664 + }, + { + "epoch": 2.19, + "learning_rate": 1.4659507158340353e-05, + "loss": 1.0649, + "step": 29665 + }, + { + "epoch": 2.19, + "learning_rate": 1.4659154228766534e-05, + "loss": 1.0953, + "step": 29666 + }, + { + "epoch": 2.19, + "learning_rate": 1.4658801291780072e-05, + "loss": 0.937, + "step": 29667 + }, + { + "epoch": 2.19, + "learning_rate": 1.4658448347381518e-05, + "loss": 0.9749, + "step": 29668 + }, + { + "epoch": 2.19, + "learning_rate": 1.4658095395571443e-05, + "loss": 1.022, + "step": 29669 + }, + { + "epoch": 2.19, + "learning_rate": 1.4657742436350404e-05, + "loss": 0.9714, + "step": 29670 + }, + { + "epoch": 2.19, + "learning_rate": 1.4657389469718966e-05, + "loss": 1.0339, + "step": 29671 + }, + { + "epoch": 2.19, + "learning_rate": 1.4657036495677683e-05, + "loss": 0.975, + "step": 29672 + }, + { + "epoch": 2.19, + "learning_rate": 1.4656683514227126e-05, + "loss": 1.0767, + "step": 29673 + }, + { + "epoch": 2.19, + "learning_rate": 1.4656330525367848e-05, + "loss": 0.8745, + "step": 29674 + }, + { + "epoch": 2.19, + "learning_rate": 1.4655977529100418e-05, + "loss": 1.0351, + "step": 29675 + }, + { + "epoch": 2.19, + "learning_rate": 1.465562452542539e-05, + "loss": 0.9616, + "step": 29676 + }, + { + "epoch": 2.19, + "learning_rate": 1.4655271514343335e-05, + "loss": 1.0097, + "step": 29677 + }, + { + "epoch": 2.19, + "learning_rate": 1.4654918495854809e-05, + "loss": 0.8971, + "step": 29678 + }, + { + "epoch": 2.19, + "learning_rate": 1.465456546996037e-05, + "loss": 1.0125, + "step": 29679 + }, + { + "epoch": 2.19, + "learning_rate": 1.4654212436660591e-05, + "loss": 0.9044, + "step": 29680 + }, + { + "epoch": 2.19, + "learning_rate": 1.4653859395956022e-05, + "loss": 1.101, + "step": 29681 + }, + { + "epoch": 2.19, + "learning_rate": 1.465350634784723e-05, + "loss": 0.9838, + "step": 29682 + }, + { + "epoch": 2.19, + "learning_rate": 1.4653153292334776e-05, + "loss": 1.0263, + "step": 29683 + }, + { + "epoch": 2.19, + "learning_rate": 1.4652800229419222e-05, + "loss": 0.9872, + "step": 29684 + }, + { + "epoch": 2.19, + "learning_rate": 1.4652447159101128e-05, + "loss": 1.0521, + "step": 29685 + }, + { + "epoch": 2.19, + "learning_rate": 1.4652094081381059e-05, + "loss": 1.0598, + "step": 29686 + }, + { + "epoch": 2.19, + "learning_rate": 1.4651740996259573e-05, + "loss": 0.9887, + "step": 29687 + }, + { + "epoch": 2.19, + "learning_rate": 1.4651387903737234e-05, + "loss": 1.0526, + "step": 29688 + }, + { + "epoch": 2.19, + "learning_rate": 1.4651034803814606e-05, + "loss": 1.044, + "step": 29689 + }, + { + "epoch": 2.19, + "learning_rate": 1.4650681696492246e-05, + "loss": 1.0337, + "step": 29690 + }, + { + "epoch": 2.19, + "learning_rate": 1.4650328581770717e-05, + "loss": 1.0541, + "step": 29691 + }, + { + "epoch": 2.19, + "learning_rate": 1.4649975459650583e-05, + "loss": 0.9321, + "step": 29692 + }, + { + "epoch": 2.19, + "learning_rate": 1.4649622330132402e-05, + "loss": 1.079, + "step": 29693 + }, + { + "epoch": 2.19, + "learning_rate": 1.4649269193216742e-05, + "loss": 1.0733, + "step": 29694 + }, + { + "epoch": 2.19, + "learning_rate": 1.464891604890416e-05, + "loss": 1.0636, + "step": 29695 + }, + { + "epoch": 2.19, + "learning_rate": 1.464856289719522e-05, + "loss": 0.9853, + "step": 29696 + }, + { + "epoch": 2.19, + "learning_rate": 1.4648209738090483e-05, + "loss": 0.9451, + "step": 29697 + }, + { + "epoch": 2.19, + "learning_rate": 1.464785657159051e-05, + "loss": 1.0156, + "step": 29698 + }, + { + "epoch": 2.19, + "learning_rate": 1.4647503397695863e-05, + "loss": 0.8938, + "step": 29699 + }, + { + "epoch": 2.19, + "learning_rate": 1.4647150216407107e-05, + "loss": 0.8899, + "step": 29700 + }, + { + "epoch": 2.19, + "learning_rate": 1.4646797027724798e-05, + "loss": 0.979, + "step": 29701 + }, + { + "epoch": 2.19, + "learning_rate": 1.4646443831649504e-05, + "loss": 0.944, + "step": 29702 + }, + { + "epoch": 2.19, + "learning_rate": 1.4646090628181781e-05, + "loss": 1.0034, + "step": 29703 + }, + { + "epoch": 2.19, + "learning_rate": 1.4645737417322199e-05, + "loss": 1.0692, + "step": 29704 + }, + { + "epoch": 2.19, + "learning_rate": 1.4645384199071311e-05, + "loss": 0.9844, + "step": 29705 + }, + { + "epoch": 2.19, + "learning_rate": 1.4645030973429685e-05, + "loss": 1.0696, + "step": 29706 + }, + { + "epoch": 2.2, + "learning_rate": 1.4644677740397882e-05, + "loss": 1.114, + "step": 29707 + }, + { + "epoch": 2.2, + "learning_rate": 1.4644324499976466e-05, + "loss": 1.0646, + "step": 29708 + }, + { + "epoch": 2.2, + "learning_rate": 1.4643971252165992e-05, + "loss": 1.0717, + "step": 29709 + }, + { + "epoch": 2.2, + "learning_rate": 1.4643617996967025e-05, + "loss": 1.0227, + "step": 29710 + }, + { + "epoch": 2.2, + "learning_rate": 1.464326473438013e-05, + "loss": 0.8066, + "step": 29711 + }, + { + "epoch": 2.2, + "learning_rate": 1.464291146440587e-05, + "loss": 0.9988, + "step": 29712 + }, + { + "epoch": 2.2, + "learning_rate": 1.46425581870448e-05, + "loss": 1.0478, + "step": 29713 + }, + { + "epoch": 2.2, + "learning_rate": 1.4642204902297488e-05, + "loss": 0.9902, + "step": 29714 + }, + { + "epoch": 2.2, + "learning_rate": 1.4641851610164492e-05, + "loss": 1.0933, + "step": 29715 + }, + { + "epoch": 2.2, + "learning_rate": 1.464149831064638e-05, + "loss": 0.8972, + "step": 29716 + }, + { + "epoch": 2.2, + "learning_rate": 1.464114500374371e-05, + "loss": 0.8831, + "step": 29717 + }, + { + "epoch": 2.2, + "learning_rate": 1.4640791689457046e-05, + "loss": 1.0011, + "step": 29718 + }, + { + "epoch": 2.2, + "learning_rate": 1.4640438367786947e-05, + "loss": 1.0093, + "step": 29719 + }, + { + "epoch": 2.2, + "learning_rate": 1.4640085038733978e-05, + "loss": 1.0581, + "step": 29720 + }, + { + "epoch": 2.2, + "learning_rate": 1.4639731702298696e-05, + "loss": 0.9714, + "step": 29721 + }, + { + "epoch": 2.2, + "learning_rate": 1.4639378358481673e-05, + "loss": 1.1122, + "step": 29722 + }, + { + "epoch": 2.2, + "learning_rate": 1.463902500728346e-05, + "loss": 1.0739, + "step": 29723 + }, + { + "epoch": 2.2, + "learning_rate": 1.463867164870463e-05, + "loss": 0.9413, + "step": 29724 + }, + { + "epoch": 2.2, + "learning_rate": 1.4638318282745736e-05, + "loss": 1.0572, + "step": 29725 + }, + { + "epoch": 2.2, + "learning_rate": 1.4637964909407345e-05, + "loss": 1.0376, + "step": 29726 + }, + { + "epoch": 2.2, + "learning_rate": 1.4637611528690019e-05, + "loss": 0.9603, + "step": 29727 + }, + { + "epoch": 2.2, + "learning_rate": 1.463725814059432e-05, + "loss": 1.0042, + "step": 29728 + }, + { + "epoch": 2.2, + "learning_rate": 1.4636904745120806e-05, + "loss": 0.9794, + "step": 29729 + }, + { + "epoch": 2.2, + "learning_rate": 1.4636551342270046e-05, + "loss": 1.131, + "step": 29730 + }, + { + "epoch": 2.2, + "learning_rate": 1.4636197932042598e-05, + "loss": 0.9307, + "step": 29731 + }, + { + "epoch": 2.2, + "learning_rate": 1.4635844514439027e-05, + "loss": 0.9626, + "step": 29732 + }, + { + "epoch": 2.2, + "learning_rate": 1.4635491089459892e-05, + "loss": 1.0066, + "step": 29733 + }, + { + "epoch": 2.2, + "learning_rate": 1.4635137657105758e-05, + "loss": 1.1215, + "step": 29734 + }, + { + "epoch": 2.2, + "learning_rate": 1.4634784217377184e-05, + "loss": 1.0534, + "step": 29735 + }, + { + "epoch": 2.2, + "learning_rate": 1.4634430770274737e-05, + "loss": 1.1481, + "step": 29736 + }, + { + "epoch": 2.2, + "learning_rate": 1.4634077315798978e-05, + "loss": 1.0355, + "step": 29737 + }, + { + "epoch": 2.2, + "learning_rate": 1.4633723853950465e-05, + "loss": 1.0672, + "step": 29738 + }, + { + "epoch": 2.2, + "learning_rate": 1.4633370384729765e-05, + "loss": 0.9707, + "step": 29739 + }, + { + "epoch": 2.2, + "learning_rate": 1.463301690813744e-05, + "loss": 0.9214, + "step": 29740 + }, + { + "epoch": 2.2, + "learning_rate": 1.4632663424174053e-05, + "loss": 1.0642, + "step": 29741 + }, + { + "epoch": 2.2, + "learning_rate": 1.4632309932840161e-05, + "loss": 1.0491, + "step": 29742 + }, + { + "epoch": 2.2, + "learning_rate": 1.463195643413633e-05, + "loss": 0.8491, + "step": 29743 + }, + { + "epoch": 2.2, + "learning_rate": 1.4631602928063124e-05, + "loss": 1.0013, + "step": 29744 + }, + { + "epoch": 2.2, + "learning_rate": 1.4631249414621107e-05, + "loss": 0.973, + "step": 29745 + }, + { + "epoch": 2.2, + "learning_rate": 1.4630895893810835e-05, + "loss": 0.9828, + "step": 29746 + }, + { + "epoch": 2.2, + "learning_rate": 1.4630542365632874e-05, + "loss": 0.9897, + "step": 29747 + }, + { + "epoch": 2.2, + "learning_rate": 1.4630188830087785e-05, + "loss": 1.0086, + "step": 29748 + }, + { + "epoch": 2.2, + "learning_rate": 1.4629835287176137e-05, + "loss": 1.1252, + "step": 29749 + }, + { + "epoch": 2.2, + "learning_rate": 1.4629481736898484e-05, + "loss": 1.0719, + "step": 29750 + }, + { + "epoch": 2.2, + "learning_rate": 1.4629128179255392e-05, + "loss": 1.0012, + "step": 29751 + }, + { + "epoch": 2.2, + "learning_rate": 1.4628774614247421e-05, + "loss": 1.0117, + "step": 29752 + }, + { + "epoch": 2.2, + "learning_rate": 1.4628421041875138e-05, + "loss": 1.045, + "step": 29753 + }, + { + "epoch": 2.2, + "learning_rate": 1.4628067462139104e-05, + "loss": 1.0372, + "step": 29754 + }, + { + "epoch": 2.2, + "learning_rate": 1.4627713875039881e-05, + "loss": 0.9617, + "step": 29755 + }, + { + "epoch": 2.2, + "learning_rate": 1.462736028057803e-05, + "loss": 1.1369, + "step": 29756 + }, + { + "epoch": 2.2, + "learning_rate": 1.4627006678754117e-05, + "loss": 1.041, + "step": 29757 + }, + { + "epoch": 2.2, + "learning_rate": 1.46266530695687e-05, + "loss": 1.0574, + "step": 29758 + }, + { + "epoch": 2.2, + "learning_rate": 1.4626299453022346e-05, + "loss": 1.0109, + "step": 29759 + }, + { + "epoch": 2.2, + "learning_rate": 1.4625945829115615e-05, + "loss": 1.0009, + "step": 29760 + }, + { + "epoch": 2.2, + "learning_rate": 1.462559219784907e-05, + "loss": 0.9653, + "step": 29761 + }, + { + "epoch": 2.2, + "learning_rate": 1.4625238559223274e-05, + "loss": 0.9659, + "step": 29762 + }, + { + "epoch": 2.2, + "learning_rate": 1.462488491323879e-05, + "loss": 0.9658, + "step": 29763 + }, + { + "epoch": 2.2, + "learning_rate": 1.4624531259896182e-05, + "loss": 0.9701, + "step": 29764 + }, + { + "epoch": 2.2, + "learning_rate": 1.462417759919601e-05, + "loss": 0.9476, + "step": 29765 + }, + { + "epoch": 2.2, + "learning_rate": 1.4623823931138836e-05, + "loss": 0.936, + "step": 29766 + }, + { + "epoch": 2.2, + "learning_rate": 1.4623470255725225e-05, + "loss": 1.0069, + "step": 29767 + }, + { + "epoch": 2.2, + "learning_rate": 1.4623116572955738e-05, + "loss": 1.1417, + "step": 29768 + }, + { + "epoch": 2.2, + "learning_rate": 1.462276288283094e-05, + "loss": 0.9297, + "step": 29769 + }, + { + "epoch": 2.2, + "learning_rate": 1.4622409185351392e-05, + "loss": 1.0769, + "step": 29770 + }, + { + "epoch": 2.2, + "learning_rate": 1.4622055480517658e-05, + "loss": 1.0585, + "step": 29771 + }, + { + "epoch": 2.2, + "learning_rate": 1.4621701768330299e-05, + "loss": 0.9348, + "step": 29772 + }, + { + "epoch": 2.2, + "learning_rate": 1.4621348048789884e-05, + "loss": 0.9796, + "step": 29773 + }, + { + "epoch": 2.2, + "learning_rate": 1.4620994321896962e-05, + "loss": 0.8984, + "step": 29774 + }, + { + "epoch": 2.2, + "learning_rate": 1.4620640587652109e-05, + "loss": 0.9675, + "step": 29775 + }, + { + "epoch": 2.2, + "learning_rate": 1.462028684605588e-05, + "loss": 0.9646, + "step": 29776 + }, + { + "epoch": 2.2, + "learning_rate": 1.4619933097108843e-05, + "loss": 0.9703, + "step": 29777 + }, + { + "epoch": 2.2, + "learning_rate": 1.4619579340811556e-05, + "loss": 1.0492, + "step": 29778 + }, + { + "epoch": 2.2, + "learning_rate": 1.4619225577164587e-05, + "loss": 1.019, + "step": 29779 + }, + { + "epoch": 2.2, + "learning_rate": 1.4618871806168494e-05, + "loss": 1.0705, + "step": 29780 + }, + { + "epoch": 2.2, + "learning_rate": 1.4618518027823843e-05, + "loss": 0.9802, + "step": 29781 + }, + { + "epoch": 2.2, + "learning_rate": 1.4618164242131197e-05, + "loss": 0.9965, + "step": 29782 + }, + { + "epoch": 2.2, + "learning_rate": 1.4617810449091117e-05, + "loss": 1.0477, + "step": 29783 + }, + { + "epoch": 2.2, + "learning_rate": 1.4617456648704167e-05, + "loss": 0.9632, + "step": 29784 + }, + { + "epoch": 2.2, + "learning_rate": 1.4617102840970907e-05, + "loss": 0.9123, + "step": 29785 + }, + { + "epoch": 2.2, + "learning_rate": 1.4616749025891906e-05, + "loss": 1.0149, + "step": 29786 + }, + { + "epoch": 2.2, + "learning_rate": 1.4616395203467721e-05, + "loss": 1.0412, + "step": 29787 + }, + { + "epoch": 2.2, + "learning_rate": 1.461604137369892e-05, + "loss": 0.9464, + "step": 29788 + }, + { + "epoch": 2.2, + "learning_rate": 1.461568753658606e-05, + "loss": 1.169, + "step": 29789 + }, + { + "epoch": 2.2, + "learning_rate": 1.4615333692129707e-05, + "loss": 0.9502, + "step": 29790 + }, + { + "epoch": 2.2, + "learning_rate": 1.4614979840330427e-05, + "loss": 1.0581, + "step": 29791 + }, + { + "epoch": 2.2, + "learning_rate": 1.4614625981188779e-05, + "loss": 0.9788, + "step": 29792 + }, + { + "epoch": 2.2, + "learning_rate": 1.4614272114705325e-05, + "loss": 0.9511, + "step": 29793 + }, + { + "epoch": 2.2, + "learning_rate": 1.4613918240880634e-05, + "loss": 1.0378, + "step": 29794 + }, + { + "epoch": 2.2, + "learning_rate": 1.4613564359715262e-05, + "loss": 1.0326, + "step": 29795 + }, + { + "epoch": 2.2, + "learning_rate": 1.4613210471209778e-05, + "loss": 0.9874, + "step": 29796 + }, + { + "epoch": 2.2, + "learning_rate": 1.4612856575364737e-05, + "loss": 1.0449, + "step": 29797 + }, + { + "epoch": 2.2, + "learning_rate": 1.4612502672180713e-05, + "loss": 1.0379, + "step": 29798 + }, + { + "epoch": 2.2, + "learning_rate": 1.4612148761658259e-05, + "loss": 0.9664, + "step": 29799 + }, + { + "epoch": 2.2, + "learning_rate": 1.4611794843797944e-05, + "loss": 1.0241, + "step": 29800 + }, + { + "epoch": 2.2, + "learning_rate": 1.461144091860033e-05, + "loss": 1.0857, + "step": 29801 + }, + { + "epoch": 2.2, + "learning_rate": 1.461108698606598e-05, + "loss": 1.1107, + "step": 29802 + }, + { + "epoch": 2.2, + "learning_rate": 1.4610733046195455e-05, + "loss": 1.0238, + "step": 29803 + }, + { + "epoch": 2.2, + "learning_rate": 1.461037909898932e-05, + "loss": 1.0522, + "step": 29804 + }, + { + "epoch": 2.2, + "learning_rate": 1.4610025144448136e-05, + "loss": 0.9674, + "step": 29805 + }, + { + "epoch": 2.2, + "learning_rate": 1.4609671182572474e-05, + "loss": 1.0072, + "step": 29806 + }, + { + "epoch": 2.2, + "learning_rate": 1.4609317213362884e-05, + "loss": 1.0087, + "step": 29807 + }, + { + "epoch": 2.2, + "learning_rate": 1.460896323681994e-05, + "loss": 0.9582, + "step": 29808 + }, + { + "epoch": 2.2, + "learning_rate": 1.46086092529442e-05, + "loss": 0.9198, + "step": 29809 + }, + { + "epoch": 2.2, + "learning_rate": 1.460825526173623e-05, + "loss": 1.0554, + "step": 29810 + }, + { + "epoch": 2.2, + "learning_rate": 1.4607901263196592e-05, + "loss": 0.9024, + "step": 29811 + }, + { + "epoch": 2.2, + "learning_rate": 1.4607547257325849e-05, + "loss": 1.0132, + "step": 29812 + }, + { + "epoch": 2.2, + "learning_rate": 1.4607193244124562e-05, + "loss": 0.957, + "step": 29813 + }, + { + "epoch": 2.2, + "learning_rate": 1.4606839223593298e-05, + "loss": 0.9687, + "step": 29814 + }, + { + "epoch": 2.2, + "learning_rate": 1.4606485195732618e-05, + "loss": 0.9856, + "step": 29815 + }, + { + "epoch": 2.2, + "learning_rate": 1.4606131160543088e-05, + "loss": 0.9819, + "step": 29816 + }, + { + "epoch": 2.2, + "learning_rate": 1.4605777118025266e-05, + "loss": 0.9642, + "step": 29817 + }, + { + "epoch": 2.2, + "learning_rate": 1.4605423068179722e-05, + "loss": 1.154, + "step": 29818 + }, + { + "epoch": 2.2, + "learning_rate": 1.4605069011007014e-05, + "loss": 0.9118, + "step": 29819 + }, + { + "epoch": 2.2, + "learning_rate": 1.4604714946507708e-05, + "loss": 1.0466, + "step": 29820 + }, + { + "epoch": 2.2, + "learning_rate": 1.4604360874682363e-05, + "loss": 1.0649, + "step": 29821 + }, + { + "epoch": 2.2, + "learning_rate": 1.4604006795531551e-05, + "loss": 1.0325, + "step": 29822 + }, + { + "epoch": 2.2, + "learning_rate": 1.4603652709055826e-05, + "loss": 1.0558, + "step": 29823 + }, + { + "epoch": 2.2, + "learning_rate": 1.4603298615255757e-05, + "loss": 0.9916, + "step": 29824 + }, + { + "epoch": 2.2, + "learning_rate": 1.4602944514131905e-05, + "loss": 1.0847, + "step": 29825 + }, + { + "epoch": 2.2, + "learning_rate": 1.4602590405684835e-05, + "loss": 1.0306, + "step": 29826 + }, + { + "epoch": 2.2, + "learning_rate": 1.4602236289915108e-05, + "loss": 1.093, + "step": 29827 + }, + { + "epoch": 2.2, + "learning_rate": 1.4601882166823292e-05, + "loss": 0.9934, + "step": 29828 + }, + { + "epoch": 2.2, + "learning_rate": 1.4601528036409947e-05, + "loss": 1.0287, + "step": 29829 + }, + { + "epoch": 2.2, + "learning_rate": 1.460117389867563e-05, + "loss": 1.1125, + "step": 29830 + }, + { + "epoch": 2.2, + "learning_rate": 1.4600819753620918e-05, + "loss": 0.9806, + "step": 29831 + }, + { + "epoch": 2.2, + "learning_rate": 1.4600465601246367e-05, + "loss": 0.9944, + "step": 29832 + }, + { + "epoch": 2.2, + "learning_rate": 1.460011144155254e-05, + "loss": 1.1062, + "step": 29833 + }, + { + "epoch": 2.2, + "learning_rate": 1.4599757274540003e-05, + "loss": 1.0017, + "step": 29834 + }, + { + "epoch": 2.2, + "learning_rate": 1.4599403100209315e-05, + "loss": 1.0439, + "step": 29835 + }, + { + "epoch": 2.2, + "learning_rate": 1.4599048918561046e-05, + "loss": 1.0383, + "step": 29836 + }, + { + "epoch": 2.2, + "learning_rate": 1.459869472959575e-05, + "loss": 0.9724, + "step": 29837 + }, + { + "epoch": 2.2, + "learning_rate": 1.4598340533314002e-05, + "loss": 1.0818, + "step": 29838 + }, + { + "epoch": 2.2, + "learning_rate": 1.4597986329716358e-05, + "loss": 1.1289, + "step": 29839 + }, + { + "epoch": 2.2, + "learning_rate": 1.4597632118803384e-05, + "loss": 0.8626, + "step": 29840 + }, + { + "epoch": 2.2, + "learning_rate": 1.4597277900575643e-05, + "loss": 0.9603, + "step": 29841 + }, + { + "epoch": 2.21, + "learning_rate": 1.4596923675033699e-05, + "loss": 1.0323, + "step": 29842 + }, + { + "epoch": 2.21, + "learning_rate": 1.4596569442178114e-05, + "loss": 0.9979, + "step": 29843 + }, + { + "epoch": 2.21, + "learning_rate": 1.4596215202009454e-05, + "loss": 0.9617, + "step": 29844 + }, + { + "epoch": 2.21, + "learning_rate": 1.4595860954528283e-05, + "loss": 1.0098, + "step": 29845 + }, + { + "epoch": 2.21, + "learning_rate": 1.4595506699735159e-05, + "loss": 0.9741, + "step": 29846 + }, + { + "epoch": 2.21, + "learning_rate": 1.459515243763065e-05, + "loss": 1.0519, + "step": 29847 + }, + { + "epoch": 2.21, + "learning_rate": 1.4594798168215323e-05, + "loss": 1.0406, + "step": 29848 + }, + { + "epoch": 2.21, + "learning_rate": 1.4594443891489735e-05, + "loss": 1.0014, + "step": 29849 + }, + { + "epoch": 2.21, + "learning_rate": 1.4594089607454453e-05, + "loss": 1.0539, + "step": 29850 + }, + { + "epoch": 2.21, + "learning_rate": 1.4593735316110041e-05, + "loss": 1.0444, + "step": 29851 + }, + { + "epoch": 2.21, + "learning_rate": 1.459338101745706e-05, + "loss": 0.9379, + "step": 29852 + }, + { + "epoch": 2.21, + "learning_rate": 1.4593026711496077e-05, + "loss": 0.9503, + "step": 29853 + }, + { + "epoch": 2.21, + "learning_rate": 1.4592672398227655e-05, + "loss": 1.1278, + "step": 29854 + }, + { + "epoch": 2.21, + "learning_rate": 1.4592318077652357e-05, + "loss": 0.9686, + "step": 29855 + }, + { + "epoch": 2.21, + "learning_rate": 1.4591963749770743e-05, + "loss": 1.0746, + "step": 29856 + }, + { + "epoch": 2.21, + "learning_rate": 1.4591609414583385e-05, + "loss": 1.0176, + "step": 29857 + }, + { + "epoch": 2.21, + "learning_rate": 1.459125507209084e-05, + "loss": 0.94, + "step": 29858 + }, + { + "epoch": 2.21, + "learning_rate": 1.4590900722293672e-05, + "loss": 1.0484, + "step": 29859 + }, + { + "epoch": 2.21, + "learning_rate": 1.4590546365192449e-05, + "loss": 0.9787, + "step": 29860 + }, + { + "epoch": 2.21, + "learning_rate": 1.4590192000787733e-05, + "loss": 0.8995, + "step": 29861 + }, + { + "epoch": 2.21, + "learning_rate": 1.4589837629080084e-05, + "loss": 1.0336, + "step": 29862 + }, + { + "epoch": 2.21, + "learning_rate": 1.458948325007007e-05, + "loss": 1.0092, + "step": 29863 + }, + { + "epoch": 2.21, + "learning_rate": 1.4589128863758255e-05, + "loss": 0.9398, + "step": 29864 + }, + { + "epoch": 2.21, + "learning_rate": 1.4588774470145204e-05, + "loss": 1.1086, + "step": 29865 + }, + { + "epoch": 2.21, + "learning_rate": 1.4588420069231474e-05, + "loss": 1.0022, + "step": 29866 + }, + { + "epoch": 2.21, + "learning_rate": 1.4588065661017639e-05, + "loss": 0.9438, + "step": 29867 + }, + { + "epoch": 2.21, + "learning_rate": 1.4587711245504252e-05, + "loss": 0.9833, + "step": 29868 + }, + { + "epoch": 2.21, + "learning_rate": 1.4587356822691883e-05, + "loss": 1.1523, + "step": 29869 + }, + { + "epoch": 2.21, + "learning_rate": 1.4587002392581096e-05, + "loss": 0.9978, + "step": 29870 + }, + { + "epoch": 2.21, + "learning_rate": 1.4586647955172453e-05, + "loss": 1.0085, + "step": 29871 + }, + { + "epoch": 2.21, + "learning_rate": 1.4586293510466519e-05, + "loss": 0.9453, + "step": 29872 + }, + { + "epoch": 2.21, + "learning_rate": 1.4585939058463858e-05, + "loss": 1.0219, + "step": 29873 + }, + { + "epoch": 2.21, + "learning_rate": 1.4585584599165033e-05, + "loss": 0.9326, + "step": 29874 + }, + { + "epoch": 2.21, + "learning_rate": 1.458523013257061e-05, + "loss": 1.0707, + "step": 29875 + }, + { + "epoch": 2.21, + "learning_rate": 1.458487565868115e-05, + "loss": 0.9014, + "step": 29876 + }, + { + "epoch": 2.21, + "learning_rate": 1.458452117749722e-05, + "loss": 0.9339, + "step": 29877 + }, + { + "epoch": 2.21, + "learning_rate": 1.4584166689019378e-05, + "loss": 1.0644, + "step": 29878 + }, + { + "epoch": 2.21, + "learning_rate": 1.4583812193248197e-05, + "loss": 1.0141, + "step": 29879 + }, + { + "epoch": 2.21, + "learning_rate": 1.4583457690184232e-05, + "loss": 1.0504, + "step": 29880 + }, + { + "epoch": 2.21, + "learning_rate": 1.4583103179828055e-05, + "loss": 0.9724, + "step": 29881 + }, + { + "epoch": 2.21, + "learning_rate": 1.4582748662180226e-05, + "loss": 0.9903, + "step": 29882 + }, + { + "epoch": 2.21, + "learning_rate": 1.458239413724131e-05, + "loss": 0.9693, + "step": 29883 + }, + { + "epoch": 2.21, + "learning_rate": 1.4582039605011869e-05, + "loss": 0.9474, + "step": 29884 + }, + { + "epoch": 2.21, + "learning_rate": 1.4581685065492468e-05, + "loss": 1.0471, + "step": 29885 + }, + { + "epoch": 2.21, + "learning_rate": 1.4581330518683674e-05, + "loss": 0.9417, + "step": 29886 + }, + { + "epoch": 2.21, + "learning_rate": 1.4580975964586046e-05, + "loss": 0.9159, + "step": 29887 + }, + { + "epoch": 2.21, + "learning_rate": 1.4580621403200153e-05, + "loss": 1.0698, + "step": 29888 + }, + { + "epoch": 2.21, + "learning_rate": 1.4580266834526555e-05, + "loss": 1.004, + "step": 29889 + }, + { + "epoch": 2.21, + "learning_rate": 1.457991225856582e-05, + "loss": 0.9985, + "step": 29890 + }, + { + "epoch": 2.21, + "learning_rate": 1.4579557675318509e-05, + "loss": 1.114, + "step": 29891 + }, + { + "epoch": 2.21, + "learning_rate": 1.4579203084785187e-05, + "loss": 0.9854, + "step": 29892 + }, + { + "epoch": 2.21, + "learning_rate": 1.4578848486966419e-05, + "loss": 1.0168, + "step": 29893 + }, + { + "epoch": 2.21, + "learning_rate": 1.457849388186277e-05, + "loss": 0.9361, + "step": 29894 + }, + { + "epoch": 2.21, + "learning_rate": 1.45781392694748e-05, + "loss": 1.0003, + "step": 29895 + }, + { + "epoch": 2.21, + "learning_rate": 1.457778464980308e-05, + "loss": 0.9252, + "step": 29896 + }, + { + "epoch": 2.21, + "learning_rate": 1.4577430022848164e-05, + "loss": 0.9173, + "step": 29897 + }, + { + "epoch": 2.21, + "learning_rate": 1.4577075388610627e-05, + "loss": 0.9915, + "step": 29898 + }, + { + "epoch": 2.21, + "learning_rate": 1.4576720747091027e-05, + "loss": 1.042, + "step": 29899 + }, + { + "epoch": 2.21, + "learning_rate": 1.4576366098289929e-05, + "loss": 0.9989, + "step": 29900 + }, + { + "epoch": 2.21, + "learning_rate": 1.4576011442207898e-05, + "loss": 1.0324, + "step": 29901 + }, + { + "epoch": 2.21, + "learning_rate": 1.4575656778845499e-05, + "loss": 1.0785, + "step": 29902 + }, + { + "epoch": 2.21, + "learning_rate": 1.4575302108203295e-05, + "loss": 0.893, + "step": 29903 + }, + { + "epoch": 2.21, + "learning_rate": 1.4574947430281853e-05, + "loss": 0.9597, + "step": 29904 + }, + { + "epoch": 2.21, + "learning_rate": 1.4574592745081735e-05, + "loss": 0.9388, + "step": 29905 + }, + { + "epoch": 2.21, + "learning_rate": 1.4574238052603505e-05, + "loss": 1.0391, + "step": 29906 + }, + { + "epoch": 2.21, + "learning_rate": 1.4573883352847723e-05, + "loss": 1.0156, + "step": 29907 + }, + { + "epoch": 2.21, + "learning_rate": 1.4573528645814965e-05, + "loss": 0.9, + "step": 29908 + }, + { + "epoch": 2.21, + "learning_rate": 1.4573173931505782e-05, + "loss": 1.0716, + "step": 29909 + }, + { + "epoch": 2.21, + "learning_rate": 1.457281920992075e-05, + "loss": 0.9442, + "step": 29910 + }, + { + "epoch": 2.21, + "learning_rate": 1.4572464481060424e-05, + "loss": 0.8653, + "step": 29911 + }, + { + "epoch": 2.21, + "learning_rate": 1.4572109744925375e-05, + "loss": 0.9134, + "step": 29912 + }, + { + "epoch": 2.21, + "learning_rate": 1.4571755001516165e-05, + "loss": 1.09, + "step": 29913 + }, + { + "epoch": 2.21, + "learning_rate": 1.4571400250833358e-05, + "loss": 1.0723, + "step": 29914 + }, + { + "epoch": 2.21, + "learning_rate": 1.4571045492877516e-05, + "loss": 0.9022, + "step": 29915 + }, + { + "epoch": 2.21, + "learning_rate": 1.457069072764921e-05, + "loss": 0.9383, + "step": 29916 + }, + { + "epoch": 2.21, + "learning_rate": 1.4570335955148997e-05, + "loss": 0.9679, + "step": 29917 + }, + { + "epoch": 2.21, + "learning_rate": 1.4569981175377446e-05, + "loss": 0.9669, + "step": 29918 + }, + { + "epoch": 2.21, + "learning_rate": 1.456962638833512e-05, + "loss": 0.9602, + "step": 29919 + }, + { + "epoch": 2.21, + "learning_rate": 1.4569271594022585e-05, + "loss": 1.0239, + "step": 29920 + }, + { + "epoch": 2.21, + "learning_rate": 1.4568916792440404e-05, + "loss": 1.05, + "step": 29921 + }, + { + "epoch": 2.21, + "learning_rate": 1.456856198358914e-05, + "loss": 0.9773, + "step": 29922 + }, + { + "epoch": 2.21, + "learning_rate": 1.4568207167469362e-05, + "loss": 1.0383, + "step": 29923 + }, + { + "epoch": 2.21, + "learning_rate": 1.4567852344081633e-05, + "loss": 1.0708, + "step": 29924 + }, + { + "epoch": 2.21, + "learning_rate": 1.4567497513426511e-05, + "loss": 1.0884, + "step": 29925 + }, + { + "epoch": 2.21, + "learning_rate": 1.4567142675504569e-05, + "loss": 1.0514, + "step": 29926 + }, + { + "epoch": 2.21, + "learning_rate": 1.4566787830316366e-05, + "loss": 0.9386, + "step": 29927 + }, + { + "epoch": 2.21, + "learning_rate": 1.4566432977862472e-05, + "loss": 1.0126, + "step": 29928 + }, + { + "epoch": 2.21, + "learning_rate": 1.4566078118143448e-05, + "loss": 1.0109, + "step": 29929 + }, + { + "epoch": 2.21, + "learning_rate": 1.4565723251159858e-05, + "loss": 1.0165, + "step": 29930 + }, + { + "epoch": 2.21, + "learning_rate": 1.4565368376912265e-05, + "loss": 0.9902, + "step": 29931 + }, + { + "epoch": 2.21, + "learning_rate": 1.4565013495401242e-05, + "loss": 1.0235, + "step": 29932 + }, + { + "epoch": 2.21, + "learning_rate": 1.4564658606627344e-05, + "loss": 0.9546, + "step": 29933 + }, + { + "epoch": 2.21, + "learning_rate": 1.4564303710591139e-05, + "loss": 1.0458, + "step": 29934 + }, + { + "epoch": 2.21, + "learning_rate": 1.4563948807293196e-05, + "loss": 1.0544, + "step": 29935 + }, + { + "epoch": 2.21, + "learning_rate": 1.4563593896734074e-05, + "loss": 1.1064, + "step": 29936 + }, + { + "epoch": 2.21, + "learning_rate": 1.4563238978914336e-05, + "loss": 0.9379, + "step": 29937 + }, + { + "epoch": 2.21, + "learning_rate": 1.4562884053834553e-05, + "loss": 1.0302, + "step": 29938 + }, + { + "epoch": 2.21, + "learning_rate": 1.4562529121495286e-05, + "loss": 1.0314, + "step": 29939 + }, + { + "epoch": 2.21, + "learning_rate": 1.4562174181897101e-05, + "loss": 1.0993, + "step": 29940 + }, + { + "epoch": 2.21, + "learning_rate": 1.4561819235040564e-05, + "loss": 1.0582, + "step": 29941 + }, + { + "epoch": 2.21, + "learning_rate": 1.4561464280926237e-05, + "loss": 1.0511, + "step": 29942 + }, + { + "epoch": 2.21, + "learning_rate": 1.4561109319554682e-05, + "loss": 1.0361, + "step": 29943 + }, + { + "epoch": 2.21, + "learning_rate": 1.4560754350926472e-05, + "loss": 1.0871, + "step": 29944 + }, + { + "epoch": 2.21, + "learning_rate": 1.4560399375042165e-05, + "loss": 0.9341, + "step": 29945 + }, + { + "epoch": 2.21, + "learning_rate": 1.4560044391902326e-05, + "loss": 1.0425, + "step": 29946 + }, + { + "epoch": 2.21, + "learning_rate": 1.4559689401507524e-05, + "loss": 1.0213, + "step": 29947 + }, + { + "epoch": 2.21, + "learning_rate": 1.455933440385832e-05, + "loss": 1.174, + "step": 29948 + }, + { + "epoch": 2.21, + "learning_rate": 1.4558979398955281e-05, + "loss": 1.116, + "step": 29949 + }, + { + "epoch": 2.21, + "learning_rate": 1.4558624386798971e-05, + "loss": 0.9721, + "step": 29950 + }, + { + "epoch": 2.21, + "learning_rate": 1.4558269367389957e-05, + "loss": 0.9924, + "step": 29951 + }, + { + "epoch": 2.21, + "learning_rate": 1.4557914340728798e-05, + "loss": 1.0379, + "step": 29952 + }, + { + "epoch": 2.21, + "learning_rate": 1.4557559306816064e-05, + "loss": 1.0217, + "step": 29953 + }, + { + "epoch": 2.21, + "learning_rate": 1.4557204265652319e-05, + "loss": 0.9554, + "step": 29954 + }, + { + "epoch": 2.21, + "learning_rate": 1.4556849217238127e-05, + "loss": 0.9437, + "step": 29955 + }, + { + "epoch": 2.21, + "learning_rate": 1.4556494161574051e-05, + "loss": 1.0343, + "step": 29956 + }, + { + "epoch": 2.21, + "learning_rate": 1.4556139098660662e-05, + "loss": 0.9638, + "step": 29957 + }, + { + "epoch": 2.21, + "learning_rate": 1.4555784028498517e-05, + "loss": 1.0615, + "step": 29958 + }, + { + "epoch": 2.21, + "learning_rate": 1.4555428951088186e-05, + "loss": 0.9957, + "step": 29959 + }, + { + "epoch": 2.21, + "learning_rate": 1.4555073866430234e-05, + "loss": 1.0342, + "step": 29960 + }, + { + "epoch": 2.21, + "learning_rate": 1.4554718774525225e-05, + "loss": 1.0714, + "step": 29961 + }, + { + "epoch": 2.21, + "learning_rate": 1.4554363675373722e-05, + "loss": 0.9783, + "step": 29962 + }, + { + "epoch": 2.21, + "learning_rate": 1.4554008568976292e-05, + "loss": 1.0097, + "step": 29963 + }, + { + "epoch": 2.21, + "learning_rate": 1.4553653455333497e-05, + "loss": 1.1201, + "step": 29964 + }, + { + "epoch": 2.21, + "learning_rate": 1.4553298334445909e-05, + "loss": 1.0746, + "step": 29965 + }, + { + "epoch": 2.21, + "learning_rate": 1.4552943206314085e-05, + "loss": 0.9866, + "step": 29966 + }, + { + "epoch": 2.21, + "learning_rate": 1.4552588070938595e-05, + "loss": 1.123, + "step": 29967 + }, + { + "epoch": 2.21, + "learning_rate": 1.4552232928320002e-05, + "loss": 1.0474, + "step": 29968 + }, + { + "epoch": 2.21, + "learning_rate": 1.4551877778458872e-05, + "loss": 1.0487, + "step": 29969 + }, + { + "epoch": 2.21, + "learning_rate": 1.4551522621355768e-05, + "loss": 1.0144, + "step": 29970 + }, + { + "epoch": 2.21, + "learning_rate": 1.4551167457011262e-05, + "loss": 1.0456, + "step": 29971 + }, + { + "epoch": 2.21, + "learning_rate": 1.4550812285425908e-05, + "loss": 1.0739, + "step": 29972 + }, + { + "epoch": 2.21, + "learning_rate": 1.455045710660028e-05, + "loss": 1.0191, + "step": 29973 + }, + { + "epoch": 2.21, + "learning_rate": 1.4550101920534939e-05, + "loss": 1.0819, + "step": 29974 + }, + { + "epoch": 2.21, + "learning_rate": 1.454974672723045e-05, + "loss": 1.0155, + "step": 29975 + }, + { + "epoch": 2.21, + "learning_rate": 1.4549391526687377e-05, + "loss": 1.0537, + "step": 29976 + }, + { + "epoch": 2.21, + "learning_rate": 1.4549036318906291e-05, + "loss": 1.031, + "step": 29977 + }, + { + "epoch": 2.22, + "learning_rate": 1.4548681103887749e-05, + "loss": 1.0125, + "step": 29978 + }, + { + "epoch": 2.22, + "learning_rate": 1.4548325881632326e-05, + "loss": 1.0895, + "step": 29979 + }, + { + "epoch": 2.22, + "learning_rate": 1.454797065214058e-05, + "loss": 1.094, + "step": 29980 + }, + { + "epoch": 2.22, + "learning_rate": 1.4547615415413076e-05, + "loss": 1.046, + "step": 29981 + }, + { + "epoch": 2.22, + "learning_rate": 1.4547260171450381e-05, + "loss": 1.0225, + "step": 29982 + }, + { + "epoch": 2.22, + "learning_rate": 1.4546904920253063e-05, + "loss": 1.0747, + "step": 29983 + }, + { + "epoch": 2.22, + "learning_rate": 1.4546549661821681e-05, + "loss": 0.9774, + "step": 29984 + }, + { + "epoch": 2.22, + "learning_rate": 1.4546194396156805e-05, + "loss": 0.917, + "step": 29985 + }, + { + "epoch": 2.22, + "learning_rate": 1.4545839123258998e-05, + "loss": 0.913, + "step": 29986 + }, + { + "epoch": 2.22, + "learning_rate": 1.4545483843128825e-05, + "loss": 1.0265, + "step": 29987 + }, + { + "epoch": 2.22, + "learning_rate": 1.4545128555766854e-05, + "loss": 0.9214, + "step": 29988 + }, + { + "epoch": 2.22, + "learning_rate": 1.4544773261173649e-05, + "loss": 1.0129, + "step": 29989 + }, + { + "epoch": 2.22, + "learning_rate": 1.4544417959349774e-05, + "loss": 1.0483, + "step": 29990 + }, + { + "epoch": 2.22, + "learning_rate": 1.4544062650295794e-05, + "loss": 1.0387, + "step": 29991 + }, + { + "epoch": 2.22, + "learning_rate": 1.4543707334012278e-05, + "loss": 1.0542, + "step": 29992 + }, + { + "epoch": 2.22, + "learning_rate": 1.4543352010499786e-05, + "loss": 1.0684, + "step": 29993 + }, + { + "epoch": 2.22, + "learning_rate": 1.4542996679758887e-05, + "loss": 0.9651, + "step": 29994 + }, + { + "epoch": 2.22, + "learning_rate": 1.4542641341790145e-05, + "loss": 0.9844, + "step": 29995 + }, + { + "epoch": 2.22, + "learning_rate": 1.4542285996594126e-05, + "loss": 1.0227, + "step": 29996 + }, + { + "epoch": 2.22, + "learning_rate": 1.4541930644171393e-05, + "loss": 1.092, + "step": 29997 + }, + { + "epoch": 2.22, + "learning_rate": 1.4541575284522517e-05, + "loss": 1.0503, + "step": 29998 + }, + { + "epoch": 2.22, + "learning_rate": 1.4541219917648057e-05, + "loss": 1.0096, + "step": 29999 + }, + { + "epoch": 2.22, + "learning_rate": 1.4540864543548582e-05, + "loss": 0.9499, + "step": 30000 + }, + { + "epoch": 2.22, + "learning_rate": 1.4540509162224657e-05, + "loss": 0.9613, + "step": 30001 + }, + { + "epoch": 2.22, + "learning_rate": 1.4540153773676846e-05, + "loss": 1.0205, + "step": 30002 + }, + { + "epoch": 2.22, + "learning_rate": 1.4539798377905713e-05, + "loss": 1.0097, + "step": 30003 + }, + { + "epoch": 2.22, + "learning_rate": 1.453944297491183e-05, + "loss": 1.1057, + "step": 30004 + }, + { + "epoch": 2.22, + "learning_rate": 1.4539087564695755e-05, + "loss": 0.9333, + "step": 30005 + }, + { + "epoch": 2.22, + "learning_rate": 1.453873214725806e-05, + "loss": 1.0247, + "step": 30006 + }, + { + "epoch": 2.22, + "learning_rate": 1.4538376722599305e-05, + "loss": 1.0432, + "step": 30007 + }, + { + "epoch": 2.22, + "learning_rate": 1.4538021290720059e-05, + "loss": 0.946, + "step": 30008 + }, + { + "epoch": 2.22, + "learning_rate": 1.4537665851620883e-05, + "loss": 0.9638, + "step": 30009 + }, + { + "epoch": 2.22, + "learning_rate": 1.453731040530235e-05, + "loss": 1.0071, + "step": 30010 + }, + { + "epoch": 2.22, + "learning_rate": 1.4536954951765016e-05, + "loss": 1.0728, + "step": 30011 + }, + { + "epoch": 2.22, + "learning_rate": 1.4536599491009456e-05, + "loss": 0.8822, + "step": 30012 + }, + { + "epoch": 2.22, + "learning_rate": 1.4536244023036228e-05, + "loss": 0.9453, + "step": 30013 + }, + { + "epoch": 2.22, + "learning_rate": 1.4535888547845902e-05, + "loss": 0.97, + "step": 30014 + }, + { + "epoch": 2.22, + "learning_rate": 1.453553306543904e-05, + "loss": 0.9925, + "step": 30015 + }, + { + "epoch": 2.22, + "learning_rate": 1.4535177575816214e-05, + "loss": 1.0772, + "step": 30016 + }, + { + "epoch": 2.22, + "learning_rate": 1.4534822078977984e-05, + "loss": 1.062, + "step": 30017 + }, + { + "epoch": 2.22, + "learning_rate": 1.4534466574924917e-05, + "loss": 1.0797, + "step": 30018 + }, + { + "epoch": 2.22, + "learning_rate": 1.4534111063657577e-05, + "loss": 1.0283, + "step": 30019 + }, + { + "epoch": 2.22, + "learning_rate": 1.453375554517653e-05, + "loss": 1.0133, + "step": 30020 + }, + { + "epoch": 2.22, + "learning_rate": 1.4533400019482344e-05, + "loss": 1.0344, + "step": 30021 + }, + { + "epoch": 2.22, + "learning_rate": 1.4533044486575585e-05, + "loss": 1.036, + "step": 30022 + }, + { + "epoch": 2.22, + "learning_rate": 1.4532688946456817e-05, + "loss": 1.0119, + "step": 30023 + }, + { + "epoch": 2.22, + "learning_rate": 1.4532333399126604e-05, + "loss": 1.0406, + "step": 30024 + }, + { + "epoch": 2.22, + "learning_rate": 1.4531977844585515e-05, + "loss": 1.0581, + "step": 30025 + }, + { + "epoch": 2.22, + "learning_rate": 1.4531622282834115e-05, + "loss": 0.9854, + "step": 30026 + }, + { + "epoch": 2.22, + "learning_rate": 1.4531266713872967e-05, + "loss": 0.9487, + "step": 30027 + }, + { + "epoch": 2.22, + "learning_rate": 1.4530911137702638e-05, + "loss": 1.0088, + "step": 30028 + }, + { + "epoch": 2.22, + "learning_rate": 1.4530555554323693e-05, + "loss": 1.0819, + "step": 30029 + }, + { + "epoch": 2.22, + "learning_rate": 1.4530199963736702e-05, + "loss": 0.9858, + "step": 30030 + }, + { + "epoch": 2.22, + "learning_rate": 1.4529844365942224e-05, + "loss": 1.1318, + "step": 30031 + }, + { + "epoch": 2.22, + "learning_rate": 1.4529488760940833e-05, + "loss": 1.0143, + "step": 30032 + }, + { + "epoch": 2.22, + "learning_rate": 1.4529133148733088e-05, + "loss": 1.1099, + "step": 30033 + }, + { + "epoch": 2.22, + "learning_rate": 1.4528777529319558e-05, + "loss": 1.0497, + "step": 30034 + }, + { + "epoch": 2.22, + "learning_rate": 1.4528421902700807e-05, + "loss": 1.0463, + "step": 30035 + }, + { + "epoch": 2.22, + "learning_rate": 1.4528066268877398e-05, + "loss": 0.984, + "step": 30036 + }, + { + "epoch": 2.22, + "learning_rate": 1.4527710627849906e-05, + "loss": 0.8999, + "step": 30037 + }, + { + "epoch": 2.22, + "learning_rate": 1.452735497961889e-05, + "loss": 1.0389, + "step": 30038 + }, + { + "epoch": 2.22, + "learning_rate": 1.4526999324184914e-05, + "loss": 1.0755, + "step": 30039 + }, + { + "epoch": 2.22, + "learning_rate": 1.452664366154855e-05, + "loss": 0.9793, + "step": 30040 + }, + { + "epoch": 2.22, + "learning_rate": 1.4526287991710359e-05, + "loss": 1.0635, + "step": 30041 + }, + { + "epoch": 2.22, + "learning_rate": 1.452593231467091e-05, + "loss": 1.0636, + "step": 30042 + }, + { + "epoch": 2.22, + "learning_rate": 1.4525576630430764e-05, + "loss": 1.1416, + "step": 30043 + }, + { + "epoch": 2.22, + "learning_rate": 1.4525220938990494e-05, + "loss": 0.8731, + "step": 30044 + }, + { + "epoch": 2.22, + "learning_rate": 1.4524865240350664e-05, + "loss": 1.0391, + "step": 30045 + }, + { + "epoch": 2.22, + "learning_rate": 1.4524509534511833e-05, + "loss": 1.0404, + "step": 30046 + }, + { + "epoch": 2.22, + "learning_rate": 1.4524153821474574e-05, + "loss": 0.9943, + "step": 30047 + }, + { + "epoch": 2.22, + "learning_rate": 1.452379810123945e-05, + "loss": 0.9028, + "step": 30048 + }, + { + "epoch": 2.22, + "learning_rate": 1.4523442373807031e-05, + "loss": 0.9871, + "step": 30049 + }, + { + "epoch": 2.22, + "learning_rate": 1.4523086639177875e-05, + "loss": 0.8885, + "step": 30050 + }, + { + "epoch": 2.22, + "learning_rate": 1.4522730897352557e-05, + "loss": 0.9656, + "step": 30051 + }, + { + "epoch": 2.22, + "learning_rate": 1.4522375148331637e-05, + "loss": 0.9243, + "step": 30052 + }, + { + "epoch": 2.22, + "learning_rate": 1.4522019392115684e-05, + "loss": 1.1774, + "step": 30053 + }, + { + "epoch": 2.22, + "learning_rate": 1.4521663628705264e-05, + "loss": 0.9484, + "step": 30054 + }, + { + "epoch": 2.22, + "learning_rate": 1.452130785810094e-05, + "loss": 1.1277, + "step": 30055 + }, + { + "epoch": 2.22, + "learning_rate": 1.4520952080303281e-05, + "loss": 0.9662, + "step": 30056 + }, + { + "epoch": 2.22, + "learning_rate": 1.4520596295312852e-05, + "loss": 1.0431, + "step": 30057 + }, + { + "epoch": 2.22, + "learning_rate": 1.4520240503130216e-05, + "loss": 1.0151, + "step": 30058 + }, + { + "epoch": 2.22, + "learning_rate": 1.4519884703755944e-05, + "loss": 1.164, + "step": 30059 + }, + { + "epoch": 2.22, + "learning_rate": 1.45195288971906e-05, + "loss": 0.9332, + "step": 30060 + }, + { + "epoch": 2.22, + "learning_rate": 1.4519173083434752e-05, + "loss": 0.9679, + "step": 30061 + }, + { + "epoch": 2.22, + "learning_rate": 1.451881726248896e-05, + "loss": 1.0743, + "step": 30062 + }, + { + "epoch": 2.22, + "learning_rate": 1.4518461434353801e-05, + "loss": 1.1102, + "step": 30063 + }, + { + "epoch": 2.22, + "learning_rate": 1.4518105599029828e-05, + "loss": 0.9982, + "step": 30064 + }, + { + "epoch": 2.22, + "learning_rate": 1.4517749756517618e-05, + "loss": 0.9205, + "step": 30065 + }, + { + "epoch": 2.22, + "learning_rate": 1.4517393906817729e-05, + "loss": 0.9463, + "step": 30066 + }, + { + "epoch": 2.22, + "learning_rate": 1.4517038049930735e-05, + "loss": 1.0093, + "step": 30067 + }, + { + "epoch": 2.22, + "learning_rate": 1.4516682185857196e-05, + "loss": 1.0499, + "step": 30068 + }, + { + "epoch": 2.22, + "learning_rate": 1.4516326314597679e-05, + "loss": 0.9359, + "step": 30069 + }, + { + "epoch": 2.22, + "learning_rate": 1.4515970436152753e-05, + "loss": 0.9705, + "step": 30070 + }, + { + "epoch": 2.22, + "learning_rate": 1.4515614550522983e-05, + "loss": 1.1537, + "step": 30071 + }, + { + "epoch": 2.22, + "learning_rate": 1.4515258657708935e-05, + "loss": 1.0354, + "step": 30072 + }, + { + "epoch": 2.22, + "learning_rate": 1.4514902757711176e-05, + "loss": 0.9634, + "step": 30073 + }, + { + "epoch": 2.22, + "learning_rate": 1.4514546850530267e-05, + "loss": 0.9946, + "step": 30074 + }, + { + "epoch": 2.22, + "learning_rate": 1.4514190936166783e-05, + "loss": 1.1129, + "step": 30075 + }, + { + "epoch": 2.22, + "learning_rate": 1.4513835014621282e-05, + "loss": 0.9387, + "step": 30076 + }, + { + "epoch": 2.22, + "learning_rate": 1.4513479085894338e-05, + "loss": 1.0786, + "step": 30077 + }, + { + "epoch": 2.22, + "learning_rate": 1.451312314998651e-05, + "loss": 0.9861, + "step": 30078 + }, + { + "epoch": 2.22, + "learning_rate": 1.4512767206898369e-05, + "loss": 0.9715, + "step": 30079 + }, + { + "epoch": 2.22, + "learning_rate": 1.4512411256630478e-05, + "loss": 1.015, + "step": 30080 + }, + { + "epoch": 2.22, + "learning_rate": 1.4512055299183409e-05, + "loss": 1.0477, + "step": 30081 + }, + { + "epoch": 2.22, + "learning_rate": 1.4511699334557725e-05, + "loss": 0.953, + "step": 30082 + }, + { + "epoch": 2.22, + "learning_rate": 1.4511343362753986e-05, + "loss": 1.0152, + "step": 30083 + }, + { + "epoch": 2.22, + "learning_rate": 1.451098738377277e-05, + "loss": 1.053, + "step": 30084 + }, + { + "epoch": 2.22, + "learning_rate": 1.4510631397614636e-05, + "loss": 0.9823, + "step": 30085 + }, + { + "epoch": 2.22, + "learning_rate": 1.4510275404280153e-05, + "loss": 0.8971, + "step": 30086 + }, + { + "epoch": 2.22, + "learning_rate": 1.4509919403769886e-05, + "loss": 1.0087, + "step": 30087 + }, + { + "epoch": 2.22, + "learning_rate": 1.4509563396084398e-05, + "loss": 0.9994, + "step": 30088 + }, + { + "epoch": 2.22, + "learning_rate": 1.4509207381224262e-05, + "loss": 1.1112, + "step": 30089 + }, + { + "epoch": 2.22, + "learning_rate": 1.450885135919004e-05, + "loss": 0.9884, + "step": 30090 + }, + { + "epoch": 2.22, + "learning_rate": 1.4508495329982305e-05, + "loss": 1.0075, + "step": 30091 + }, + { + "epoch": 2.22, + "learning_rate": 1.4508139293601614e-05, + "loss": 0.9607, + "step": 30092 + }, + { + "epoch": 2.22, + "learning_rate": 1.4507783250048538e-05, + "loss": 1.0057, + "step": 30093 + }, + { + "epoch": 2.22, + "learning_rate": 1.4507427199323646e-05, + "loss": 1.0011, + "step": 30094 + }, + { + "epoch": 2.22, + "learning_rate": 1.4507071141427495e-05, + "loss": 0.9925, + "step": 30095 + }, + { + "epoch": 2.22, + "learning_rate": 1.4506715076360665e-05, + "loss": 0.9646, + "step": 30096 + }, + { + "epoch": 2.22, + "learning_rate": 1.4506359004123714e-05, + "loss": 0.8527, + "step": 30097 + }, + { + "epoch": 2.22, + "learning_rate": 1.450600292471721e-05, + "loss": 0.9641, + "step": 30098 + }, + { + "epoch": 2.22, + "learning_rate": 1.450564683814172e-05, + "loss": 1.0, + "step": 30099 + }, + { + "epoch": 2.22, + "learning_rate": 1.4505290744397813e-05, + "loss": 1.0123, + "step": 30100 + }, + { + "epoch": 2.22, + "learning_rate": 1.4504934643486048e-05, + "loss": 0.9325, + "step": 30101 + }, + { + "epoch": 2.22, + "learning_rate": 1.4504578535407002e-05, + "loss": 1.004, + "step": 30102 + }, + { + "epoch": 2.22, + "learning_rate": 1.4504222420161231e-05, + "loss": 1.0888, + "step": 30103 + }, + { + "epoch": 2.22, + "learning_rate": 1.450386629774931e-05, + "loss": 1.0651, + "step": 30104 + }, + { + "epoch": 2.22, + "learning_rate": 1.4503510168171797e-05, + "loss": 1.1307, + "step": 30105 + }, + { + "epoch": 2.22, + "learning_rate": 1.4503154031429269e-05, + "loss": 1.0495, + "step": 30106 + }, + { + "epoch": 2.22, + "learning_rate": 1.4502797887522286e-05, + "loss": 0.9836, + "step": 30107 + }, + { + "epoch": 2.22, + "learning_rate": 1.4502441736451416e-05, + "loss": 0.9983, + "step": 30108 + }, + { + "epoch": 2.22, + "learning_rate": 1.4502085578217225e-05, + "loss": 0.9272, + "step": 30109 + }, + { + "epoch": 2.22, + "learning_rate": 1.4501729412820282e-05, + "loss": 1.0731, + "step": 30110 + }, + { + "epoch": 2.22, + "learning_rate": 1.4501373240261149e-05, + "loss": 1.0338, + "step": 30111 + }, + { + "epoch": 2.22, + "learning_rate": 1.45010170605404e-05, + "loss": 1.065, + "step": 30112 + }, + { + "epoch": 2.23, + "learning_rate": 1.4500660873658592e-05, + "loss": 0.9661, + "step": 30113 + }, + { + "epoch": 2.23, + "learning_rate": 1.4500304679616299e-05, + "loss": 0.9757, + "step": 30114 + }, + { + "epoch": 2.23, + "learning_rate": 1.4499948478414086e-05, + "loss": 0.9855, + "step": 30115 + }, + { + "epoch": 2.23, + "learning_rate": 1.4499592270052519e-05, + "loss": 1.0723, + "step": 30116 + }, + { + "epoch": 2.23, + "learning_rate": 1.4499236054532164e-05, + "loss": 1.0521, + "step": 30117 + }, + { + "epoch": 2.23, + "learning_rate": 1.4498879831853592e-05, + "loss": 0.9419, + "step": 30118 + }, + { + "epoch": 2.23, + "learning_rate": 1.4498523602017364e-05, + "loss": 0.9824, + "step": 30119 + }, + { + "epoch": 2.23, + "learning_rate": 1.4498167365024052e-05, + "loss": 1.1224, + "step": 30120 + }, + { + "epoch": 2.23, + "learning_rate": 1.4497811120874216e-05, + "loss": 1.0245, + "step": 30121 + }, + { + "epoch": 2.23, + "learning_rate": 1.449745486956843e-05, + "loss": 0.9619, + "step": 30122 + }, + { + "epoch": 2.23, + "learning_rate": 1.4497098611107255e-05, + "loss": 1.072, + "step": 30123 + }, + { + "epoch": 2.23, + "learning_rate": 1.4496742345491264e-05, + "loss": 1.0874, + "step": 30124 + }, + { + "epoch": 2.23, + "learning_rate": 1.4496386072721017e-05, + "loss": 0.9947, + "step": 30125 + }, + { + "epoch": 2.23, + "learning_rate": 1.4496029792797086e-05, + "loss": 1.0247, + "step": 30126 + }, + { + "epoch": 2.23, + "learning_rate": 1.4495673505720034e-05, + "loss": 1.0601, + "step": 30127 + }, + { + "epoch": 2.23, + "learning_rate": 1.4495317211490432e-05, + "loss": 0.9945, + "step": 30128 + }, + { + "epoch": 2.23, + "learning_rate": 1.4494960910108845e-05, + "loss": 1.0591, + "step": 30129 + }, + { + "epoch": 2.23, + "learning_rate": 1.449460460157584e-05, + "loss": 0.9834, + "step": 30130 + }, + { + "epoch": 2.23, + "learning_rate": 1.4494248285891979e-05, + "loss": 0.9247, + "step": 30131 + }, + { + "epoch": 2.23, + "learning_rate": 1.4493891963057835e-05, + "loss": 1.0801, + "step": 30132 + }, + { + "epoch": 2.23, + "learning_rate": 1.4493535633073973e-05, + "loss": 1.1414, + "step": 30133 + }, + { + "epoch": 2.23, + "learning_rate": 1.4493179295940962e-05, + "loss": 0.951, + "step": 30134 + }, + { + "epoch": 2.23, + "learning_rate": 1.4492822951659366e-05, + "loss": 1.0542, + "step": 30135 + }, + { + "epoch": 2.23, + "learning_rate": 1.4492466600229755e-05, + "loss": 0.9461, + "step": 30136 + }, + { + "epoch": 2.23, + "learning_rate": 1.449211024165269e-05, + "loss": 1.0379, + "step": 30137 + }, + { + "epoch": 2.23, + "learning_rate": 1.4491753875928745e-05, + "loss": 0.9614, + "step": 30138 + }, + { + "epoch": 2.23, + "learning_rate": 1.4491397503058485e-05, + "loss": 1.031, + "step": 30139 + }, + { + "epoch": 2.23, + "learning_rate": 1.449104112304247e-05, + "loss": 1.0064, + "step": 30140 + }, + { + "epoch": 2.23, + "learning_rate": 1.449068473588128e-05, + "loss": 1.088, + "step": 30141 + }, + { + "epoch": 2.23, + "learning_rate": 1.4490328341575468e-05, + "loss": 0.9632, + "step": 30142 + }, + { + "epoch": 2.23, + "learning_rate": 1.4489971940125611e-05, + "loss": 0.9598, + "step": 30143 + }, + { + "epoch": 2.23, + "learning_rate": 1.4489615531532274e-05, + "loss": 1.0152, + "step": 30144 + }, + { + "epoch": 2.23, + "learning_rate": 1.4489259115796022e-05, + "loss": 0.9753, + "step": 30145 + }, + { + "epoch": 2.23, + "learning_rate": 1.4488902692917423e-05, + "loss": 1.0574, + "step": 30146 + }, + { + "epoch": 2.23, + "learning_rate": 1.4488546262897046e-05, + "loss": 1.0461, + "step": 30147 + }, + { + "epoch": 2.23, + "learning_rate": 1.4488189825735452e-05, + "loss": 1.0205, + "step": 30148 + }, + { + "epoch": 2.23, + "learning_rate": 1.4487833381433216e-05, + "loss": 1.0419, + "step": 30149 + }, + { + "epoch": 2.23, + "learning_rate": 1.4487476929990898e-05, + "loss": 1.029, + "step": 30150 + }, + { + "epoch": 2.23, + "learning_rate": 1.448712047140907e-05, + "loss": 1.1707, + "step": 30151 + }, + { + "epoch": 2.23, + "learning_rate": 1.4486764005688297e-05, + "loss": 1.0147, + "step": 30152 + }, + { + "epoch": 2.23, + "learning_rate": 1.4486407532829148e-05, + "loss": 0.9261, + "step": 30153 + }, + { + "epoch": 2.23, + "learning_rate": 1.4486051052832189e-05, + "loss": 1.0761, + "step": 30154 + }, + { + "epoch": 2.23, + "learning_rate": 1.4485694565697985e-05, + "loss": 1.0961, + "step": 30155 + }, + { + "epoch": 2.23, + "learning_rate": 1.4485338071427106e-05, + "loss": 1.0023, + "step": 30156 + }, + { + "epoch": 2.23, + "learning_rate": 1.448498157002012e-05, + "loss": 1.0515, + "step": 30157 + }, + { + "epoch": 2.23, + "learning_rate": 1.448462506147759e-05, + "loss": 1.0472, + "step": 30158 + }, + { + "epoch": 2.23, + "learning_rate": 1.4484268545800086e-05, + "loss": 0.965, + "step": 30159 + }, + { + "epoch": 2.23, + "learning_rate": 1.4483912022988176e-05, + "loss": 0.9068, + "step": 30160 + }, + { + "epoch": 2.23, + "learning_rate": 1.4483555493042425e-05, + "loss": 1.0429, + "step": 30161 + }, + { + "epoch": 2.23, + "learning_rate": 1.4483198955963402e-05, + "loss": 1.0111, + "step": 30162 + }, + { + "epoch": 2.23, + "learning_rate": 1.4482842411751673e-05, + "loss": 0.9283, + "step": 30163 + }, + { + "epoch": 2.23, + "learning_rate": 1.4482485860407807e-05, + "loss": 1.0115, + "step": 30164 + }, + { + "epoch": 2.23, + "learning_rate": 1.448212930193237e-05, + "loss": 0.9257, + "step": 30165 + }, + { + "epoch": 2.23, + "learning_rate": 1.448177273632593e-05, + "loss": 1.0298, + "step": 30166 + }, + { + "epoch": 2.23, + "learning_rate": 1.4481416163589052e-05, + "loss": 1.028, + "step": 30167 + }, + { + "epoch": 2.23, + "learning_rate": 1.4481059583722305e-05, + "loss": 1.0205, + "step": 30168 + }, + { + "epoch": 2.23, + "learning_rate": 1.4480702996726258e-05, + "loss": 0.9308, + "step": 30169 + }, + { + "epoch": 2.23, + "learning_rate": 1.4480346402601475e-05, + "loss": 0.981, + "step": 30170 + }, + { + "epoch": 2.23, + "learning_rate": 1.4479989801348527e-05, + "loss": 0.9591, + "step": 30171 + }, + { + "epoch": 2.23, + "learning_rate": 1.4479633192967976e-05, + "loss": 1.0337, + "step": 30172 + }, + { + "epoch": 2.23, + "learning_rate": 1.4479276577460395e-05, + "loss": 0.9676, + "step": 30173 + }, + { + "epoch": 2.23, + "learning_rate": 1.447891995482635e-05, + "loss": 1.1283, + "step": 30174 + }, + { + "epoch": 2.23, + "learning_rate": 1.4478563325066408e-05, + "loss": 0.8834, + "step": 30175 + }, + { + "epoch": 2.23, + "learning_rate": 1.4478206688181131e-05, + "loss": 1.0423, + "step": 30176 + }, + { + "epoch": 2.23, + "learning_rate": 1.4477850044171095e-05, + "loss": 0.919, + "step": 30177 + }, + { + "epoch": 2.23, + "learning_rate": 1.4477493393036863e-05, + "loss": 0.974, + "step": 30178 + }, + { + "epoch": 2.23, + "learning_rate": 1.4477136734779002e-05, + "loss": 1.0186, + "step": 30179 + }, + { + "epoch": 2.23, + "learning_rate": 1.4476780069398082e-05, + "loss": 0.9883, + "step": 30180 + }, + { + "epoch": 2.23, + "learning_rate": 1.447642339689467e-05, + "loss": 0.9184, + "step": 30181 + }, + { + "epoch": 2.23, + "learning_rate": 1.4476066717269328e-05, + "loss": 0.847, + "step": 30182 + }, + { + "epoch": 2.23, + "learning_rate": 1.4475710030522632e-05, + "loss": 0.9522, + "step": 30183 + }, + { + "epoch": 2.23, + "learning_rate": 1.4475353336655142e-05, + "loss": 1.0481, + "step": 30184 + }, + { + "epoch": 2.23, + "learning_rate": 1.4474996635667432e-05, + "loss": 1.0467, + "step": 30185 + }, + { + "epoch": 2.23, + "learning_rate": 1.4474639927560068e-05, + "loss": 1.0777, + "step": 30186 + }, + { + "epoch": 2.23, + "learning_rate": 1.4474283212333611e-05, + "loss": 0.9823, + "step": 30187 + }, + { + "epoch": 2.23, + "learning_rate": 1.4473926489988636e-05, + "loss": 0.9173, + "step": 30188 + }, + { + "epoch": 2.23, + "learning_rate": 1.4473569760525704e-05, + "loss": 1.0112, + "step": 30189 + }, + { + "epoch": 2.23, + "learning_rate": 1.4473213023945394e-05, + "loss": 0.9188, + "step": 30190 + }, + { + "epoch": 2.23, + "learning_rate": 1.4472856280248262e-05, + "loss": 1.0035, + "step": 30191 + }, + { + "epoch": 2.23, + "learning_rate": 1.447249952943488e-05, + "loss": 1.0611, + "step": 30192 + }, + { + "epoch": 2.23, + "learning_rate": 1.4472142771505814e-05, + "loss": 1.0439, + "step": 30193 + }, + { + "epoch": 2.23, + "learning_rate": 1.4471786006461636e-05, + "loss": 1.0762, + "step": 30194 + }, + { + "epoch": 2.23, + "learning_rate": 1.4471429234302907e-05, + "loss": 1.0722, + "step": 30195 + }, + { + "epoch": 2.23, + "learning_rate": 1.44710724550302e-05, + "loss": 0.9483, + "step": 30196 + }, + { + "epoch": 2.23, + "learning_rate": 1.447071566864408e-05, + "loss": 0.908, + "step": 30197 + }, + { + "epoch": 2.23, + "learning_rate": 1.4470358875145117e-05, + "loss": 0.9312, + "step": 30198 + }, + { + "epoch": 2.23, + "learning_rate": 1.4470002074533874e-05, + "loss": 1.001, + "step": 30199 + }, + { + "epoch": 2.23, + "learning_rate": 1.4469645266810924e-05, + "loss": 0.9654, + "step": 30200 + }, + { + "epoch": 2.23, + "learning_rate": 1.4469288451976829e-05, + "loss": 0.9808, + "step": 30201 + }, + { + "epoch": 2.23, + "learning_rate": 1.4468931630032164e-05, + "loss": 0.9174, + "step": 30202 + }, + { + "epoch": 2.23, + "learning_rate": 1.446857480097749e-05, + "loss": 0.9635, + "step": 30203 + }, + { + "epoch": 2.23, + "learning_rate": 1.446821796481338e-05, + "loss": 1.0422, + "step": 30204 + }, + { + "epoch": 2.23, + "learning_rate": 1.44678611215404e-05, + "loss": 1.0065, + "step": 30205 + }, + { + "epoch": 2.23, + "learning_rate": 1.4467504271159113e-05, + "loss": 0.9874, + "step": 30206 + }, + { + "epoch": 2.23, + "learning_rate": 1.4467147413670091e-05, + "loss": 1.0239, + "step": 30207 + }, + { + "epoch": 2.23, + "learning_rate": 1.4466790549073903e-05, + "loss": 0.9034, + "step": 30208 + }, + { + "epoch": 2.23, + "learning_rate": 1.4466433677371114e-05, + "loss": 1.0086, + "step": 30209 + }, + { + "epoch": 2.23, + "learning_rate": 1.4466076798562295e-05, + "loss": 1.0458, + "step": 30210 + }, + { + "epoch": 2.23, + "learning_rate": 1.4465719912648009e-05, + "loss": 1.0736, + "step": 30211 + }, + { + "epoch": 2.23, + "learning_rate": 1.4465363019628829e-05, + "loss": 0.9849, + "step": 30212 + }, + { + "epoch": 2.23, + "learning_rate": 1.446500611950532e-05, + "loss": 1.0875, + "step": 30213 + }, + { + "epoch": 2.23, + "learning_rate": 1.446464921227805e-05, + "loss": 0.9274, + "step": 30214 + }, + { + "epoch": 2.23, + "learning_rate": 1.4464292297947586e-05, + "loss": 0.9125, + "step": 30215 + }, + { + "epoch": 2.23, + "learning_rate": 1.4463935376514498e-05, + "loss": 1.0437, + "step": 30216 + }, + { + "epoch": 2.23, + "learning_rate": 1.4463578447979349e-05, + "loss": 1.0729, + "step": 30217 + }, + { + "epoch": 2.23, + "learning_rate": 1.4463221512342716e-05, + "loss": 0.9984, + "step": 30218 + }, + { + "epoch": 2.23, + "learning_rate": 1.4462864569605157e-05, + "loss": 1.0552, + "step": 30219 + }, + { + "epoch": 2.23, + "learning_rate": 1.4462507619767248e-05, + "loss": 0.9392, + "step": 30220 + }, + { + "epoch": 2.23, + "learning_rate": 1.4462150662829548e-05, + "loss": 0.8384, + "step": 30221 + }, + { + "epoch": 2.23, + "learning_rate": 1.4461793698792637e-05, + "loss": 1.0494, + "step": 30222 + }, + { + "epoch": 2.23, + "learning_rate": 1.4461436727657072e-05, + "loss": 1.0654, + "step": 30223 + }, + { + "epoch": 2.23, + "learning_rate": 1.4461079749423425e-05, + "loss": 0.9612, + "step": 30224 + }, + { + "epoch": 2.23, + "learning_rate": 1.4460722764092266e-05, + "loss": 1.0542, + "step": 30225 + }, + { + "epoch": 2.23, + "learning_rate": 1.446036577166416e-05, + "loss": 0.9334, + "step": 30226 + }, + { + "epoch": 2.23, + "learning_rate": 1.4460008772139674e-05, + "loss": 1.0007, + "step": 30227 + }, + { + "epoch": 2.23, + "learning_rate": 1.4459651765519378e-05, + "loss": 1.0256, + "step": 30228 + }, + { + "epoch": 2.23, + "learning_rate": 1.4459294751803841e-05, + "loss": 1.0201, + "step": 30229 + }, + { + "epoch": 2.23, + "learning_rate": 1.4458937730993632e-05, + "loss": 0.9592, + "step": 30230 + }, + { + "epoch": 2.23, + "learning_rate": 1.4458580703089316e-05, + "loss": 0.9741, + "step": 30231 + }, + { + "epoch": 2.23, + "learning_rate": 1.4458223668091458e-05, + "loss": 1.1059, + "step": 30232 + }, + { + "epoch": 2.23, + "learning_rate": 1.4457866626000634e-05, + "loss": 0.9434, + "step": 30233 + }, + { + "epoch": 2.23, + "learning_rate": 1.4457509576817406e-05, + "loss": 0.98, + "step": 30234 + }, + { + "epoch": 2.23, + "learning_rate": 1.4457152520542344e-05, + "loss": 0.9007, + "step": 30235 + }, + { + "epoch": 2.23, + "learning_rate": 1.4456795457176018e-05, + "loss": 1.0317, + "step": 30236 + }, + { + "epoch": 2.23, + "learning_rate": 1.4456438386718992e-05, + "loss": 0.9796, + "step": 30237 + }, + { + "epoch": 2.23, + "learning_rate": 1.4456081309171839e-05, + "loss": 0.9574, + "step": 30238 + }, + { + "epoch": 2.23, + "learning_rate": 1.445572422453512e-05, + "loss": 1.013, + "step": 30239 + }, + { + "epoch": 2.23, + "learning_rate": 1.445536713280941e-05, + "loss": 1.0048, + "step": 30240 + }, + { + "epoch": 2.23, + "learning_rate": 1.4455010033995276e-05, + "loss": 1.0054, + "step": 30241 + }, + { + "epoch": 2.23, + "learning_rate": 1.4454652928093283e-05, + "loss": 1.1523, + "step": 30242 + }, + { + "epoch": 2.23, + "learning_rate": 1.4454295815104003e-05, + "loss": 0.9731, + "step": 30243 + }, + { + "epoch": 2.23, + "learning_rate": 1.4453938695027999e-05, + "loss": 1.1032, + "step": 30244 + }, + { + "epoch": 2.23, + "learning_rate": 1.4453581567865844e-05, + "loss": 1.0444, + "step": 30245 + }, + { + "epoch": 2.23, + "learning_rate": 1.4453224433618102e-05, + "loss": 0.9294, + "step": 30246 + }, + { + "epoch": 2.23, + "learning_rate": 1.4452867292285349e-05, + "loss": 1.1164, + "step": 30247 + }, + { + "epoch": 2.24, + "learning_rate": 1.4452510143868143e-05, + "loss": 0.974, + "step": 30248 + }, + { + "epoch": 2.24, + "learning_rate": 1.4452152988367058e-05, + "loss": 0.917, + "step": 30249 + }, + { + "epoch": 2.24, + "learning_rate": 1.4451795825782663e-05, + "loss": 1.0426, + "step": 30250 + }, + { + "epoch": 2.24, + "learning_rate": 1.4451438656115527e-05, + "loss": 0.9248, + "step": 30251 + }, + { + "epoch": 2.24, + "learning_rate": 1.4451081479366212e-05, + "loss": 1.0721, + "step": 30252 + }, + { + "epoch": 2.24, + "learning_rate": 1.445072429553529e-05, + "loss": 0.9438, + "step": 30253 + }, + { + "epoch": 2.24, + "learning_rate": 1.4450367104623329e-05, + "loss": 1.0813, + "step": 30254 + }, + { + "epoch": 2.24, + "learning_rate": 1.4450009906630897e-05, + "loss": 0.9567, + "step": 30255 + }, + { + "epoch": 2.24, + "learning_rate": 1.4449652701558562e-05, + "loss": 1.0414, + "step": 30256 + }, + { + "epoch": 2.24, + "learning_rate": 1.4449295489406898e-05, + "loss": 1.0273, + "step": 30257 + }, + { + "epoch": 2.24, + "learning_rate": 1.4448938270176465e-05, + "loss": 1.1264, + "step": 30258 + }, + { + "epoch": 2.24, + "learning_rate": 1.444858104386784e-05, + "loss": 1.0194, + "step": 30259 + }, + { + "epoch": 2.24, + "learning_rate": 1.4448223810481581e-05, + "loss": 1.0419, + "step": 30260 + }, + { + "epoch": 2.24, + "learning_rate": 1.4447866570018264e-05, + "loss": 0.9551, + "step": 30261 + }, + { + "epoch": 2.24, + "learning_rate": 1.4447509322478453e-05, + "loss": 0.9054, + "step": 30262 + }, + { + "epoch": 2.24, + "learning_rate": 1.4447152067862718e-05, + "loss": 0.9069, + "step": 30263 + }, + { + "epoch": 2.24, + "learning_rate": 1.4446794806171626e-05, + "loss": 1.0526, + "step": 30264 + }, + { + "epoch": 2.24, + "learning_rate": 1.4446437537405751e-05, + "loss": 1.0403, + "step": 30265 + }, + { + "epoch": 2.24, + "learning_rate": 1.4446080261565655e-05, + "loss": 0.8839, + "step": 30266 + }, + { + "epoch": 2.24, + "learning_rate": 1.4445722978651913e-05, + "loss": 0.9054, + "step": 30267 + }, + { + "epoch": 2.24, + "learning_rate": 1.4445365688665086e-05, + "loss": 1.0179, + "step": 30268 + }, + { + "epoch": 2.24, + "learning_rate": 1.4445008391605747e-05, + "loss": 1.0776, + "step": 30269 + }, + { + "epoch": 2.24, + "learning_rate": 1.4444651087474461e-05, + "loss": 0.9891, + "step": 30270 + }, + { + "epoch": 2.24, + "learning_rate": 1.4444293776271802e-05, + "loss": 1.0304, + "step": 30271 + }, + { + "epoch": 2.24, + "learning_rate": 1.444393645799833e-05, + "loss": 1.0149, + "step": 30272 + }, + { + "epoch": 2.24, + "learning_rate": 1.4443579132654625e-05, + "loss": 0.9542, + "step": 30273 + }, + { + "epoch": 2.24, + "learning_rate": 1.4443221800241245e-05, + "loss": 0.8948, + "step": 30274 + }, + { + "epoch": 2.24, + "learning_rate": 1.4442864460758766e-05, + "loss": 1.0766, + "step": 30275 + }, + { + "epoch": 2.24, + "learning_rate": 1.444250711420775e-05, + "loss": 0.9685, + "step": 30276 + }, + { + "epoch": 2.24, + "learning_rate": 1.444214976058877e-05, + "loss": 1.0217, + "step": 30277 + }, + { + "epoch": 2.24, + "learning_rate": 1.4441792399902394e-05, + "loss": 0.8711, + "step": 30278 + }, + { + "epoch": 2.24, + "learning_rate": 1.4441435032149189e-05, + "loss": 1.0365, + "step": 30279 + }, + { + "epoch": 2.24, + "learning_rate": 1.4441077657329726e-05, + "loss": 0.9882, + "step": 30280 + }, + { + "epoch": 2.24, + "learning_rate": 1.4440720275444569e-05, + "loss": 0.978, + "step": 30281 + }, + { + "epoch": 2.24, + "learning_rate": 1.444036288649429e-05, + "loss": 0.9886, + "step": 30282 + }, + { + "epoch": 2.24, + "learning_rate": 1.4440005490479459e-05, + "loss": 0.9344, + "step": 30283 + }, + { + "epoch": 2.24, + "learning_rate": 1.4439648087400642e-05, + "loss": 0.9891, + "step": 30284 + }, + { + "epoch": 2.24, + "learning_rate": 1.4439290677258409e-05, + "loss": 0.9446, + "step": 30285 + }, + { + "epoch": 2.24, + "learning_rate": 1.4438933260053325e-05, + "loss": 0.8994, + "step": 30286 + }, + { + "epoch": 2.24, + "learning_rate": 1.4438575835785967e-05, + "loss": 1.0559, + "step": 30287 + }, + { + "epoch": 2.24, + "learning_rate": 1.4438218404456893e-05, + "loss": 0.9489, + "step": 30288 + }, + { + "epoch": 2.24, + "learning_rate": 1.4437860966066679e-05, + "loss": 0.9587, + "step": 30289 + }, + { + "epoch": 2.24, + "learning_rate": 1.4437503520615894e-05, + "loss": 1.0579, + "step": 30290 + }, + { + "epoch": 2.24, + "learning_rate": 1.4437146068105102e-05, + "loss": 0.9674, + "step": 30291 + }, + { + "epoch": 2.24, + "learning_rate": 1.4436788608534876e-05, + "loss": 0.97, + "step": 30292 + }, + { + "epoch": 2.24, + "learning_rate": 1.4436431141905779e-05, + "loss": 0.9917, + "step": 30293 + }, + { + "epoch": 2.24, + "learning_rate": 1.4436073668218387e-05, + "loss": 1.0486, + "step": 30294 + }, + { + "epoch": 2.24, + "learning_rate": 1.4435716187473265e-05, + "loss": 0.9686, + "step": 30295 + }, + { + "epoch": 2.24, + "learning_rate": 1.443535869967098e-05, + "loss": 1.0571, + "step": 30296 + }, + { + "epoch": 2.24, + "learning_rate": 1.4435001204812105e-05, + "loss": 1.0269, + "step": 30297 + }, + { + "epoch": 2.24, + "learning_rate": 1.4434643702897206e-05, + "loss": 1.0115, + "step": 30298 + }, + { + "epoch": 2.24, + "learning_rate": 1.4434286193926852e-05, + "loss": 1.0417, + "step": 30299 + }, + { + "epoch": 2.24, + "learning_rate": 1.4433928677901612e-05, + "loss": 1.1271, + "step": 30300 + }, + { + "epoch": 2.24, + "learning_rate": 1.4433571154822055e-05, + "loss": 1.1414, + "step": 30301 + }, + { + "epoch": 2.24, + "learning_rate": 1.4433213624688748e-05, + "loss": 0.9087, + "step": 30302 + }, + { + "epoch": 2.24, + "learning_rate": 1.4432856087502263e-05, + "loss": 0.9581, + "step": 30303 + }, + { + "epoch": 2.24, + "learning_rate": 1.4432498543263169e-05, + "loss": 1.0272, + "step": 30304 + }, + { + "epoch": 2.24, + "learning_rate": 1.4432140991972031e-05, + "loss": 0.8765, + "step": 30305 + }, + { + "epoch": 2.24, + "learning_rate": 1.4431783433629423e-05, + "loss": 1.0355, + "step": 30306 + }, + { + "epoch": 2.24, + "learning_rate": 1.4431425868235909e-05, + "loss": 0.9624, + "step": 30307 + }, + { + "epoch": 2.24, + "learning_rate": 1.443106829579206e-05, + "loss": 0.9003, + "step": 30308 + }, + { + "epoch": 2.24, + "learning_rate": 1.4430710716298441e-05, + "loss": 1.0, + "step": 30309 + }, + { + "epoch": 2.24, + "learning_rate": 1.443035312975563e-05, + "loss": 0.9792, + "step": 30310 + }, + { + "epoch": 2.24, + "learning_rate": 1.442999553616419e-05, + "loss": 0.9851, + "step": 30311 + }, + { + "epoch": 2.24, + "learning_rate": 1.442963793552469e-05, + "loss": 1.0157, + "step": 30312 + }, + { + "epoch": 2.24, + "learning_rate": 1.4429280327837696e-05, + "loss": 1.0446, + "step": 30313 + }, + { + "epoch": 2.24, + "learning_rate": 1.4428922713103784e-05, + "loss": 1.0598, + "step": 30314 + }, + { + "epoch": 2.24, + "learning_rate": 1.4428565091323518e-05, + "loss": 1.1065, + "step": 30315 + }, + { + "epoch": 2.24, + "learning_rate": 1.442820746249747e-05, + "loss": 1.091, + "step": 30316 + }, + { + "epoch": 2.24, + "learning_rate": 1.4427849826626206e-05, + "loss": 0.9748, + "step": 30317 + }, + { + "epoch": 2.24, + "learning_rate": 1.4427492183710296e-05, + "loss": 0.9559, + "step": 30318 + }, + { + "epoch": 2.24, + "learning_rate": 1.442713453375031e-05, + "loss": 1.108, + "step": 30319 + }, + { + "epoch": 2.24, + "learning_rate": 1.4426776876746815e-05, + "loss": 0.9909, + "step": 30320 + }, + { + "epoch": 2.24, + "learning_rate": 1.4426419212700381e-05, + "loss": 0.9752, + "step": 30321 + }, + { + "epoch": 2.24, + "learning_rate": 1.4426061541611578e-05, + "loss": 1.107, + "step": 30322 + }, + { + "epoch": 2.24, + "learning_rate": 1.4425703863480975e-05, + "loss": 0.9744, + "step": 30323 + }, + { + "epoch": 2.24, + "learning_rate": 1.4425346178309141e-05, + "loss": 0.9151, + "step": 30324 + }, + { + "epoch": 2.24, + "learning_rate": 1.4424988486096646e-05, + "loss": 1.059, + "step": 30325 + }, + { + "epoch": 2.24, + "learning_rate": 1.4424630786844054e-05, + "loss": 1.0001, + "step": 30326 + }, + { + "epoch": 2.24, + "learning_rate": 1.4424273080551938e-05, + "loss": 1.0163, + "step": 30327 + }, + { + "epoch": 2.24, + "learning_rate": 1.442391536722087e-05, + "loss": 1.0624, + "step": 30328 + }, + { + "epoch": 2.24, + "learning_rate": 1.4423557646851412e-05, + "loss": 0.9698, + "step": 30329 + }, + { + "epoch": 2.24, + "learning_rate": 1.4423199919444142e-05, + "loss": 0.9951, + "step": 30330 + }, + { + "epoch": 2.24, + "learning_rate": 1.4422842184999619e-05, + "loss": 0.9904, + "step": 30331 + }, + { + "epoch": 2.24, + "learning_rate": 1.442248444351842e-05, + "loss": 0.9838, + "step": 30332 + }, + { + "epoch": 2.24, + "learning_rate": 1.442212669500111e-05, + "loss": 1.0263, + "step": 30333 + }, + { + "epoch": 2.24, + "learning_rate": 1.4421768939448262e-05, + "loss": 0.9862, + "step": 30334 + }, + { + "epoch": 2.24, + "learning_rate": 1.4421411176860443e-05, + "loss": 1.0765, + "step": 30335 + }, + { + "epoch": 2.24, + "learning_rate": 1.442105340723822e-05, + "loss": 0.9934, + "step": 30336 + }, + { + "epoch": 2.24, + "learning_rate": 1.4420695630582165e-05, + "loss": 0.9815, + "step": 30337 + }, + { + "epoch": 2.24, + "learning_rate": 1.4420337846892847e-05, + "loss": 1.0011, + "step": 30338 + }, + { + "epoch": 2.24, + "learning_rate": 1.4419980056170832e-05, + "loss": 1.0579, + "step": 30339 + }, + { + "epoch": 2.24, + "learning_rate": 1.4419622258416696e-05, + "loss": 1.1045, + "step": 30340 + }, + { + "epoch": 2.24, + "learning_rate": 1.4419264453631003e-05, + "loss": 1.0289, + "step": 30341 + }, + { + "epoch": 2.24, + "learning_rate": 1.4418906641814323e-05, + "loss": 1.0708, + "step": 30342 + }, + { + "epoch": 2.24, + "learning_rate": 1.4418548822967227e-05, + "loss": 1.0851, + "step": 30343 + }, + { + "epoch": 2.24, + "learning_rate": 1.441819099709028e-05, + "loss": 1.0503, + "step": 30344 + }, + { + "epoch": 2.24, + "learning_rate": 1.4417833164184058e-05, + "loss": 1.0784, + "step": 30345 + }, + { + "epoch": 2.24, + "learning_rate": 1.4417475324249124e-05, + "loss": 1.0833, + "step": 30346 + }, + { + "epoch": 2.24, + "learning_rate": 1.4417117477286052e-05, + "loss": 0.8857, + "step": 30347 + }, + { + "epoch": 2.24, + "learning_rate": 1.4416759623295406e-05, + "loss": 0.9399, + "step": 30348 + }, + { + "epoch": 2.24, + "learning_rate": 1.441640176227776e-05, + "loss": 0.9803, + "step": 30349 + }, + { + "epoch": 2.24, + "learning_rate": 1.4416043894233682e-05, + "loss": 1.0145, + "step": 30350 + }, + { + "epoch": 2.24, + "learning_rate": 1.4415686019163744e-05, + "loss": 0.9579, + "step": 30351 + }, + { + "epoch": 2.24, + "learning_rate": 1.4415328137068509e-05, + "loss": 1.0083, + "step": 30352 + }, + { + "epoch": 2.24, + "learning_rate": 1.4414970247948554e-05, + "loss": 1.0161, + "step": 30353 + }, + { + "epoch": 2.24, + "learning_rate": 1.4414612351804442e-05, + "loss": 1.011, + "step": 30354 + }, + { + "epoch": 2.24, + "learning_rate": 1.4414254448636746e-05, + "loss": 1.0543, + "step": 30355 + }, + { + "epoch": 2.24, + "learning_rate": 1.4413896538446032e-05, + "loss": 0.9258, + "step": 30356 + }, + { + "epoch": 2.24, + "learning_rate": 1.4413538621232873e-05, + "loss": 0.9713, + "step": 30357 + }, + { + "epoch": 2.24, + "learning_rate": 1.4413180696997837e-05, + "loss": 0.9934, + "step": 30358 + }, + { + "epoch": 2.24, + "learning_rate": 1.4412822765741497e-05, + "loss": 0.9983, + "step": 30359 + }, + { + "epoch": 2.24, + "learning_rate": 1.4412464827464414e-05, + "loss": 0.9328, + "step": 30360 + }, + { + "epoch": 2.24, + "learning_rate": 1.4412106882167164e-05, + "loss": 1.0164, + "step": 30361 + }, + { + "epoch": 2.24, + "learning_rate": 1.441174892985032e-05, + "loss": 0.8692, + "step": 30362 + }, + { + "epoch": 2.24, + "learning_rate": 1.4411390970514444e-05, + "loss": 0.9711, + "step": 30363 + }, + { + "epoch": 2.24, + "learning_rate": 1.4411033004160104e-05, + "loss": 0.9645, + "step": 30364 + }, + { + "epoch": 2.24, + "learning_rate": 1.4410675030787878e-05, + "loss": 1.0253, + "step": 30365 + }, + { + "epoch": 2.24, + "learning_rate": 1.4410317050398326e-05, + "loss": 1.0656, + "step": 30366 + }, + { + "epoch": 2.24, + "learning_rate": 1.4409959062992026e-05, + "loss": 0.9846, + "step": 30367 + }, + { + "epoch": 2.24, + "learning_rate": 1.4409601068569546e-05, + "loss": 0.9153, + "step": 30368 + }, + { + "epoch": 2.24, + "learning_rate": 1.440924306713145e-05, + "loss": 1.0577, + "step": 30369 + }, + { + "epoch": 2.24, + "learning_rate": 1.4408885058678315e-05, + "loss": 1.0609, + "step": 30370 + }, + { + "epoch": 2.24, + "learning_rate": 1.4408527043210706e-05, + "loss": 1.0887, + "step": 30371 + }, + { + "epoch": 2.24, + "learning_rate": 1.4408169020729194e-05, + "loss": 1.062, + "step": 30372 + }, + { + "epoch": 2.24, + "learning_rate": 1.4407810991234347e-05, + "loss": 0.9452, + "step": 30373 + }, + { + "epoch": 2.24, + "learning_rate": 1.4407452954726735e-05, + "loss": 0.9919, + "step": 30374 + }, + { + "epoch": 2.24, + "learning_rate": 1.4407094911206931e-05, + "loss": 1.0594, + "step": 30375 + }, + { + "epoch": 2.24, + "learning_rate": 1.44067368606755e-05, + "loss": 1.0197, + "step": 30376 + }, + { + "epoch": 2.24, + "learning_rate": 1.4406378803133011e-05, + "loss": 0.9744, + "step": 30377 + }, + { + "epoch": 2.24, + "learning_rate": 1.440602073858004e-05, + "loss": 1.0575, + "step": 30378 + }, + { + "epoch": 2.24, + "learning_rate": 1.4405662667017154e-05, + "loss": 0.9541, + "step": 30379 + }, + { + "epoch": 2.24, + "learning_rate": 1.440530458844492e-05, + "loss": 1.0367, + "step": 30380 + }, + { + "epoch": 2.24, + "learning_rate": 1.4404946502863909e-05, + "loss": 0.8388, + "step": 30381 + }, + { + "epoch": 2.24, + "learning_rate": 1.4404588410274694e-05, + "loss": 1.0423, + "step": 30382 + }, + { + "epoch": 2.24, + "learning_rate": 1.4404230310677838e-05, + "loss": 0.9898, + "step": 30383 + }, + { + "epoch": 2.25, + "learning_rate": 1.4403872204073916e-05, + "loss": 0.9952, + "step": 30384 + }, + { + "epoch": 2.25, + "learning_rate": 1.4403514090463497e-05, + "loss": 1.0473, + "step": 30385 + }, + { + "epoch": 2.25, + "learning_rate": 1.440315596984715e-05, + "loss": 1.0153, + "step": 30386 + }, + { + "epoch": 2.25, + "learning_rate": 1.4402797842225445e-05, + "loss": 0.8797, + "step": 30387 + }, + { + "epoch": 2.25, + "learning_rate": 1.440243970759895e-05, + "loss": 1.0076, + "step": 30388 + }, + { + "epoch": 2.25, + "learning_rate": 1.4402081565968237e-05, + "loss": 0.9638, + "step": 30389 + }, + { + "epoch": 2.25, + "learning_rate": 1.4401723417333878e-05, + "loss": 1.0919, + "step": 30390 + }, + { + "epoch": 2.25, + "learning_rate": 1.4401365261696436e-05, + "loss": 1.0702, + "step": 30391 + }, + { + "epoch": 2.25, + "learning_rate": 1.4401007099056487e-05, + "loss": 1.0906, + "step": 30392 + }, + { + "epoch": 2.25, + "learning_rate": 1.4400648929414598e-05, + "loss": 0.8578, + "step": 30393 + }, + { + "epoch": 2.25, + "learning_rate": 1.4400290752771339e-05, + "loss": 0.9808, + "step": 30394 + }, + { + "epoch": 2.25, + "learning_rate": 1.4399932569127282e-05, + "loss": 1.0419, + "step": 30395 + }, + { + "epoch": 2.25, + "learning_rate": 1.4399574378482994e-05, + "loss": 0.9646, + "step": 30396 + }, + { + "epoch": 2.25, + "learning_rate": 1.4399216180839047e-05, + "loss": 0.934, + "step": 30397 + }, + { + "epoch": 2.25, + "learning_rate": 1.439885797619601e-05, + "loss": 0.9448, + "step": 30398 + }, + { + "epoch": 2.25, + "learning_rate": 1.439849976455445e-05, + "loss": 1.126, + "step": 30399 + }, + { + "epoch": 2.25, + "learning_rate": 1.4398141545914946e-05, + "loss": 0.9624, + "step": 30400 + }, + { + "epoch": 2.25, + "learning_rate": 1.4397783320278056e-05, + "loss": 1.0619, + "step": 30401 + }, + { + "epoch": 2.25, + "learning_rate": 1.439742508764436e-05, + "loss": 0.9528, + "step": 30402 + }, + { + "epoch": 2.25, + "learning_rate": 1.4397066848014418e-05, + "loss": 0.9171, + "step": 30403 + }, + { + "epoch": 2.25, + "learning_rate": 1.4396708601388811e-05, + "loss": 1.1407, + "step": 30404 + }, + { + "epoch": 2.25, + "learning_rate": 1.4396350347768099e-05, + "loss": 1.0061, + "step": 30405 + }, + { + "epoch": 2.25, + "learning_rate": 1.4395992087152859e-05, + "loss": 1.0438, + "step": 30406 + }, + { + "epoch": 2.25, + "learning_rate": 1.4395633819543657e-05, + "loss": 0.9398, + "step": 30407 + }, + { + "epoch": 2.25, + "learning_rate": 1.4395275544941068e-05, + "loss": 1.04, + "step": 30408 + }, + { + "epoch": 2.25, + "learning_rate": 1.4394917263345655e-05, + "loss": 1.0777, + "step": 30409 + }, + { + "epoch": 2.25, + "learning_rate": 1.4394558974757995e-05, + "loss": 1.0326, + "step": 30410 + }, + { + "epoch": 2.25, + "learning_rate": 1.4394200679178651e-05, + "loss": 0.9586, + "step": 30411 + }, + { + "epoch": 2.25, + "learning_rate": 1.43938423766082e-05, + "loss": 0.8857, + "step": 30412 + }, + { + "epoch": 2.25, + "learning_rate": 1.4393484067047205e-05, + "loss": 1.0105, + "step": 30413 + }, + { + "epoch": 2.25, + "learning_rate": 1.4393125750496243e-05, + "loss": 1.023, + "step": 30414 + }, + { + "epoch": 2.25, + "learning_rate": 1.4392767426955877e-05, + "loss": 1.0343, + "step": 30415 + }, + { + "epoch": 2.25, + "learning_rate": 1.4392409096426682e-05, + "loss": 1.0352, + "step": 30416 + }, + { + "epoch": 2.25, + "learning_rate": 1.4392050758909229e-05, + "loss": 1.1177, + "step": 30417 + }, + { + "epoch": 2.25, + "learning_rate": 1.4391692414404086e-05, + "loss": 1.0539, + "step": 30418 + }, + { + "epoch": 2.25, + "learning_rate": 1.439133406291182e-05, + "loss": 0.9683, + "step": 30419 + }, + { + "epoch": 2.25, + "learning_rate": 1.439097570443301e-05, + "loss": 1.0975, + "step": 30420 + }, + { + "epoch": 2.25, + "learning_rate": 1.4390617338968214e-05, + "loss": 0.8508, + "step": 30421 + }, + { + "epoch": 2.25, + "learning_rate": 1.4390258966518014e-05, + "loss": 0.8924, + "step": 30422 + }, + { + "epoch": 2.25, + "learning_rate": 1.438990058708297e-05, + "loss": 1.0383, + "step": 30423 + }, + { + "epoch": 2.25, + "learning_rate": 1.438954220066366e-05, + "loss": 1.0146, + "step": 30424 + }, + { + "epoch": 2.25, + "learning_rate": 1.4389183807260649e-05, + "loss": 1.0742, + "step": 30425 + }, + { + "epoch": 2.25, + "learning_rate": 1.4388825406874514e-05, + "loss": 1.0287, + "step": 30426 + }, + { + "epoch": 2.25, + "learning_rate": 1.4388466999505815e-05, + "loss": 1.009, + "step": 30427 + }, + { + "epoch": 2.25, + "learning_rate": 1.4388108585155135e-05, + "loss": 1.0396, + "step": 30428 + }, + { + "epoch": 2.25, + "learning_rate": 1.438775016382303e-05, + "loss": 0.8883, + "step": 30429 + }, + { + "epoch": 2.25, + "learning_rate": 1.438739173551008e-05, + "loss": 1.0001, + "step": 30430 + }, + { + "epoch": 2.25, + "learning_rate": 1.438703330021685e-05, + "loss": 0.9172, + "step": 30431 + }, + { + "epoch": 2.25, + "learning_rate": 1.4386674857943917e-05, + "loss": 0.9523, + "step": 30432 + }, + { + "epoch": 2.25, + "learning_rate": 1.4386316408691842e-05, + "loss": 1.0047, + "step": 30433 + }, + { + "epoch": 2.25, + "learning_rate": 1.4385957952461205e-05, + "loss": 0.9733, + "step": 30434 + }, + { + "epoch": 2.25, + "learning_rate": 1.4385599489252568e-05, + "loss": 1.0362, + "step": 30435 + }, + { + "epoch": 2.25, + "learning_rate": 1.438524101906651e-05, + "loss": 0.988, + "step": 30436 + }, + { + "epoch": 2.25, + "learning_rate": 1.4384882541903593e-05, + "loss": 0.991, + "step": 30437 + }, + { + "epoch": 2.25, + "learning_rate": 1.4384524057764391e-05, + "loss": 1.1389, + "step": 30438 + }, + { + "epoch": 2.25, + "learning_rate": 1.4384165566649473e-05, + "loss": 0.8424, + "step": 30439 + }, + { + "epoch": 2.25, + "learning_rate": 1.438380706855941e-05, + "loss": 1.0253, + "step": 30440 + }, + { + "epoch": 2.25, + "learning_rate": 1.4383448563494775e-05, + "loss": 0.9565, + "step": 30441 + }, + { + "epoch": 2.25, + "learning_rate": 1.4383090051456132e-05, + "loss": 0.9971, + "step": 30442 + }, + { + "epoch": 2.25, + "learning_rate": 1.438273153244406e-05, + "loss": 1.0093, + "step": 30443 + }, + { + "epoch": 2.25, + "learning_rate": 1.4382373006459123e-05, + "loss": 1.0965, + "step": 30444 + }, + { + "epoch": 2.25, + "learning_rate": 1.4382014473501892e-05, + "loss": 0.9435, + "step": 30445 + }, + { + "epoch": 2.25, + "learning_rate": 1.4381655933572942e-05, + "loss": 0.9431, + "step": 30446 + }, + { + "epoch": 2.25, + "learning_rate": 1.4381297386672838e-05, + "loss": 1.0097, + "step": 30447 + }, + { + "epoch": 2.25, + "learning_rate": 1.4380938832802152e-05, + "loss": 1.0285, + "step": 30448 + }, + { + "epoch": 2.25, + "learning_rate": 1.4380580271961457e-05, + "loss": 1.0703, + "step": 30449 + }, + { + "epoch": 2.25, + "learning_rate": 1.4380221704151317e-05, + "loss": 1.0448, + "step": 30450 + }, + { + "epoch": 2.25, + "learning_rate": 1.4379863129372314e-05, + "loss": 0.9739, + "step": 30451 + }, + { + "epoch": 2.25, + "learning_rate": 1.4379504547625004e-05, + "loss": 1.0069, + "step": 30452 + }, + { + "epoch": 2.25, + "learning_rate": 1.4379145958909971e-05, + "loss": 0.9616, + "step": 30453 + }, + { + "epoch": 2.25, + "learning_rate": 1.4378787363227775e-05, + "loss": 0.9942, + "step": 30454 + }, + { + "epoch": 2.25, + "learning_rate": 1.4378428760578992e-05, + "loss": 0.9858, + "step": 30455 + }, + { + "epoch": 2.25, + "learning_rate": 1.4378070150964194e-05, + "loss": 0.9962, + "step": 30456 + }, + { + "epoch": 2.25, + "learning_rate": 1.4377711534383948e-05, + "loss": 1.0054, + "step": 30457 + }, + { + "epoch": 2.25, + "learning_rate": 1.4377352910838824e-05, + "loss": 1.0552, + "step": 30458 + }, + { + "epoch": 2.25, + "learning_rate": 1.4376994280329398e-05, + "loss": 1.0051, + "step": 30459 + }, + { + "epoch": 2.25, + "learning_rate": 1.4376635642856231e-05, + "loss": 0.9213, + "step": 30460 + }, + { + "epoch": 2.25, + "learning_rate": 1.4376276998419903e-05, + "loss": 1.0638, + "step": 30461 + }, + { + "epoch": 2.25, + "learning_rate": 1.4375918347020979e-05, + "loss": 0.981, + "step": 30462 + }, + { + "epoch": 2.25, + "learning_rate": 1.4375559688660032e-05, + "loss": 0.9972, + "step": 30463 + }, + { + "epoch": 2.25, + "learning_rate": 1.4375201023337633e-05, + "loss": 1.0282, + "step": 30464 + }, + { + "epoch": 2.25, + "learning_rate": 1.4374842351054352e-05, + "loss": 1.0207, + "step": 30465 + }, + { + "epoch": 2.25, + "learning_rate": 1.437448367181076e-05, + "loss": 0.9615, + "step": 30466 + }, + { + "epoch": 2.25, + "learning_rate": 1.4374124985607426e-05, + "loss": 1.0451, + "step": 30467 + }, + { + "epoch": 2.25, + "learning_rate": 1.437376629244492e-05, + "loss": 0.9384, + "step": 30468 + }, + { + "epoch": 2.25, + "learning_rate": 1.4373407592323819e-05, + "loss": 0.9227, + "step": 30469 + }, + { + "epoch": 2.25, + "learning_rate": 1.4373048885244683e-05, + "loss": 0.9839, + "step": 30470 + }, + { + "epoch": 2.25, + "learning_rate": 1.4372690171208093e-05, + "loss": 0.9416, + "step": 30471 + }, + { + "epoch": 2.25, + "learning_rate": 1.4372331450214614e-05, + "loss": 0.9648, + "step": 30472 + }, + { + "epoch": 2.25, + "learning_rate": 1.4371972722264817e-05, + "loss": 1.1412, + "step": 30473 + }, + { + "epoch": 2.25, + "learning_rate": 1.4371613987359278e-05, + "loss": 1.0208, + "step": 30474 + }, + { + "epoch": 2.25, + "learning_rate": 1.437125524549856e-05, + "loss": 0.8655, + "step": 30475 + }, + { + "epoch": 2.25, + "learning_rate": 1.437089649668324e-05, + "loss": 0.954, + "step": 30476 + }, + { + "epoch": 2.25, + "learning_rate": 1.4370537740913885e-05, + "loss": 0.9999, + "step": 30477 + }, + { + "epoch": 2.25, + "learning_rate": 1.4370178978191068e-05, + "loss": 1.1077, + "step": 30478 + }, + { + "epoch": 2.25, + "learning_rate": 1.4369820208515357e-05, + "loss": 0.9448, + "step": 30479 + }, + { + "epoch": 2.25, + "learning_rate": 1.4369461431887324e-05, + "loss": 1.0093, + "step": 30480 + }, + { + "epoch": 2.25, + "learning_rate": 1.4369102648307542e-05, + "loss": 1.1609, + "step": 30481 + }, + { + "epoch": 2.25, + "learning_rate": 1.4368743857776579e-05, + "loss": 1.0123, + "step": 30482 + }, + { + "epoch": 2.25, + "learning_rate": 1.4368385060295009e-05, + "loss": 0.925, + "step": 30483 + }, + { + "epoch": 2.25, + "learning_rate": 1.4368026255863398e-05, + "loss": 1.0141, + "step": 30484 + }, + { + "epoch": 2.25, + "learning_rate": 1.4367667444482322e-05, + "loss": 0.9649, + "step": 30485 + }, + { + "epoch": 2.25, + "learning_rate": 1.436730862615235e-05, + "loss": 1.0218, + "step": 30486 + }, + { + "epoch": 2.25, + "learning_rate": 1.436694980087405e-05, + "loss": 1.0264, + "step": 30487 + }, + { + "epoch": 2.25, + "learning_rate": 1.4366590968647998e-05, + "loss": 1.0667, + "step": 30488 + }, + { + "epoch": 2.25, + "learning_rate": 1.4366232129474762e-05, + "loss": 1.0665, + "step": 30489 + }, + { + "epoch": 2.25, + "learning_rate": 1.4365873283354912e-05, + "loss": 0.971, + "step": 30490 + }, + { + "epoch": 2.25, + "learning_rate": 1.436551443028902e-05, + "loss": 1.1797, + "step": 30491 + }, + { + "epoch": 2.25, + "learning_rate": 1.4365155570277655e-05, + "loss": 0.8663, + "step": 30492 + }, + { + "epoch": 2.25, + "learning_rate": 1.4364796703321394e-05, + "loss": 1.1667, + "step": 30493 + }, + { + "epoch": 2.25, + "learning_rate": 1.4364437829420803e-05, + "loss": 1.0438, + "step": 30494 + }, + { + "epoch": 2.25, + "learning_rate": 1.4364078948576451e-05, + "loss": 1.0763, + "step": 30495 + }, + { + "epoch": 2.25, + "learning_rate": 1.4363720060788918e-05, + "loss": 1.0152, + "step": 30496 + }, + { + "epoch": 2.25, + "learning_rate": 1.4363361166058762e-05, + "loss": 1.0487, + "step": 30497 + }, + { + "epoch": 2.25, + "learning_rate": 1.4363002264386566e-05, + "loss": 1.0024, + "step": 30498 + }, + { + "epoch": 2.25, + "learning_rate": 1.4362643355772889e-05, + "loss": 1.0063, + "step": 30499 + }, + { + "epoch": 2.25, + "learning_rate": 1.4362284440218314e-05, + "loss": 0.9819, + "step": 30500 + }, + { + "epoch": 2.25, + "learning_rate": 1.4361925517723406e-05, + "loss": 1.0064, + "step": 30501 + }, + { + "epoch": 2.25, + "learning_rate": 1.436156658828874e-05, + "loss": 0.9542, + "step": 30502 + }, + { + "epoch": 2.25, + "learning_rate": 1.436120765191488e-05, + "loss": 1.038, + "step": 30503 + }, + { + "epoch": 2.25, + "learning_rate": 1.4360848708602401e-05, + "loss": 0.9262, + "step": 30504 + }, + { + "epoch": 2.25, + "learning_rate": 1.4360489758351875e-05, + "loss": 0.9939, + "step": 30505 + }, + { + "epoch": 2.25, + "learning_rate": 1.4360130801163874e-05, + "loss": 1.0431, + "step": 30506 + }, + { + "epoch": 2.25, + "learning_rate": 1.4359771837038963e-05, + "loss": 1.0133, + "step": 30507 + }, + { + "epoch": 2.25, + "learning_rate": 1.435941286597772e-05, + "loss": 0.9148, + "step": 30508 + }, + { + "epoch": 2.25, + "learning_rate": 1.4359053887980711e-05, + "loss": 1.0054, + "step": 30509 + }, + { + "epoch": 2.25, + "learning_rate": 1.4358694903048516e-05, + "loss": 1.0573, + "step": 30510 + }, + { + "epoch": 2.25, + "learning_rate": 1.4358335911181693e-05, + "loss": 1.0175, + "step": 30511 + }, + { + "epoch": 2.25, + "learning_rate": 1.4357976912380827e-05, + "loss": 1.0243, + "step": 30512 + }, + { + "epoch": 2.25, + "learning_rate": 1.4357617906646475e-05, + "loss": 1.0026, + "step": 30513 + }, + { + "epoch": 2.25, + "learning_rate": 1.435725889397922e-05, + "loss": 0.9343, + "step": 30514 + }, + { + "epoch": 2.25, + "learning_rate": 1.4356899874379627e-05, + "loss": 0.97, + "step": 30515 + }, + { + "epoch": 2.25, + "learning_rate": 1.4356540847848266e-05, + "loss": 0.9178, + "step": 30516 + }, + { + "epoch": 2.25, + "learning_rate": 1.4356181814385712e-05, + "loss": 1.0113, + "step": 30517 + }, + { + "epoch": 2.25, + "learning_rate": 1.4355822773992536e-05, + "loss": 1.0374, + "step": 30518 + }, + { + "epoch": 2.26, + "learning_rate": 1.4355463726669307e-05, + "loss": 0.9433, + "step": 30519 + }, + { + "epoch": 2.26, + "learning_rate": 1.43551046724166e-05, + "loss": 1.0593, + "step": 30520 + }, + { + "epoch": 2.26, + "learning_rate": 1.4354745611234985e-05, + "loss": 1.0672, + "step": 30521 + }, + { + "epoch": 2.26, + "learning_rate": 1.435438654312503e-05, + "loss": 1.1577, + "step": 30522 + }, + { + "epoch": 2.26, + "learning_rate": 1.4354027468087309e-05, + "loss": 0.95, + "step": 30523 + }, + { + "epoch": 2.26, + "learning_rate": 1.435366838612239e-05, + "loss": 1.062, + "step": 30524 + }, + { + "epoch": 2.26, + "learning_rate": 1.4353309297230849e-05, + "loss": 0.9252, + "step": 30525 + }, + { + "epoch": 2.26, + "learning_rate": 1.4352950201413254e-05, + "loss": 1.0521, + "step": 30526 + }, + { + "epoch": 2.26, + "learning_rate": 1.4352591098670176e-05, + "loss": 1.0846, + "step": 30527 + }, + { + "epoch": 2.26, + "learning_rate": 1.4352231989002192e-05, + "loss": 1.0021, + "step": 30528 + }, + { + "epoch": 2.26, + "learning_rate": 1.4351872872409867e-05, + "loss": 0.9814, + "step": 30529 + }, + { + "epoch": 2.26, + "learning_rate": 1.4351513748893776e-05, + "loss": 1.0002, + "step": 30530 + }, + { + "epoch": 2.26, + "learning_rate": 1.4351154618454486e-05, + "loss": 1.0249, + "step": 30531 + }, + { + "epoch": 2.26, + "learning_rate": 1.4350795481092572e-05, + "loss": 0.9844, + "step": 30532 + }, + { + "epoch": 2.26, + "learning_rate": 1.4350436336808607e-05, + "loss": 1.0464, + "step": 30533 + }, + { + "epoch": 2.26, + "learning_rate": 1.435007718560316e-05, + "loss": 1.0298, + "step": 30534 + }, + { + "epoch": 2.26, + "learning_rate": 1.43497180274768e-05, + "loss": 0.9679, + "step": 30535 + }, + { + "epoch": 2.26, + "learning_rate": 1.43493588624301e-05, + "loss": 1.0285, + "step": 30536 + }, + { + "epoch": 2.26, + "learning_rate": 1.4348999690463633e-05, + "loss": 0.988, + "step": 30537 + }, + { + "epoch": 2.26, + "learning_rate": 1.4348640511577974e-05, + "loss": 1.0512, + "step": 30538 + }, + { + "epoch": 2.26, + "learning_rate": 1.4348281325773685e-05, + "loss": 0.9742, + "step": 30539 + }, + { + "epoch": 2.26, + "learning_rate": 1.4347922133051345e-05, + "loss": 1.0309, + "step": 30540 + }, + { + "epoch": 2.26, + "learning_rate": 1.4347562933411523e-05, + "loss": 0.9739, + "step": 30541 + }, + { + "epoch": 2.26, + "learning_rate": 1.434720372685479e-05, + "loss": 0.9853, + "step": 30542 + }, + { + "epoch": 2.26, + "learning_rate": 1.434684451338172e-05, + "loss": 1.0275, + "step": 30543 + }, + { + "epoch": 2.26, + "learning_rate": 1.4346485292992877e-05, + "loss": 1.024, + "step": 30544 + }, + { + "epoch": 2.26, + "learning_rate": 1.4346126065688844e-05, + "loss": 0.9404, + "step": 30545 + }, + { + "epoch": 2.26, + "learning_rate": 1.4345766831470185e-05, + "loss": 0.9012, + "step": 30546 + }, + { + "epoch": 2.26, + "learning_rate": 1.4345407590337472e-05, + "loss": 1.0418, + "step": 30547 + }, + { + "epoch": 2.26, + "learning_rate": 1.4345048342291276e-05, + "loss": 0.9726, + "step": 30548 + }, + { + "epoch": 2.26, + "learning_rate": 1.4344689087332178e-05, + "loss": 1.0858, + "step": 30549 + }, + { + "epoch": 2.26, + "learning_rate": 1.4344329825460734e-05, + "loss": 0.8982, + "step": 30550 + }, + { + "epoch": 2.26, + "learning_rate": 1.4343970556677527e-05, + "loss": 1.1438, + "step": 30551 + }, + { + "epoch": 2.26, + "learning_rate": 1.4343611280983124e-05, + "loss": 1.0234, + "step": 30552 + }, + { + "epoch": 2.26, + "learning_rate": 1.4343251998378099e-05, + "loss": 0.9487, + "step": 30553 + }, + { + "epoch": 2.26, + "learning_rate": 1.4342892708863016e-05, + "loss": 0.9619, + "step": 30554 + }, + { + "epoch": 2.26, + "learning_rate": 1.4342533412438459e-05, + "loss": 0.8963, + "step": 30555 + }, + { + "epoch": 2.26, + "learning_rate": 1.4342174109104991e-05, + "loss": 1.1564, + "step": 30556 + }, + { + "epoch": 2.26, + "learning_rate": 1.434181479886319e-05, + "loss": 1.0359, + "step": 30557 + }, + { + "epoch": 2.26, + "learning_rate": 1.434145548171362e-05, + "loss": 1.0107, + "step": 30558 + }, + { + "epoch": 2.26, + "learning_rate": 1.434109615765686e-05, + "loss": 1.0449, + "step": 30559 + }, + { + "epoch": 2.26, + "learning_rate": 1.4340736826693474e-05, + "loss": 0.9441, + "step": 30560 + }, + { + "epoch": 2.26, + "learning_rate": 1.4340377488824038e-05, + "loss": 0.9791, + "step": 30561 + }, + { + "epoch": 2.26, + "learning_rate": 1.4340018144049124e-05, + "loss": 1.0303, + "step": 30562 + }, + { + "epoch": 2.26, + "learning_rate": 1.4339658792369306e-05, + "loss": 1.0025, + "step": 30563 + }, + { + "epoch": 2.26, + "learning_rate": 1.433929943378515e-05, + "loss": 1.0522, + "step": 30564 + }, + { + "epoch": 2.26, + "learning_rate": 1.433894006829723e-05, + "loss": 1.1242, + "step": 30565 + }, + { + "epoch": 2.26, + "learning_rate": 1.4338580695906122e-05, + "loss": 1.0241, + "step": 30566 + }, + { + "epoch": 2.26, + "learning_rate": 1.4338221316612391e-05, + "loss": 1.0227, + "step": 30567 + }, + { + "epoch": 2.26, + "learning_rate": 1.4337861930416614e-05, + "loss": 1.0852, + "step": 30568 + }, + { + "epoch": 2.26, + "learning_rate": 1.4337502537319363e-05, + "loss": 0.9839, + "step": 30569 + }, + { + "epoch": 2.26, + "learning_rate": 1.4337143137321201e-05, + "loss": 0.8874, + "step": 30570 + }, + { + "epoch": 2.26, + "learning_rate": 1.4336783730422712e-05, + "loss": 0.9553, + "step": 30571 + }, + { + "epoch": 2.26, + "learning_rate": 1.4336424316624457e-05, + "loss": 1.0402, + "step": 30572 + }, + { + "epoch": 2.26, + "learning_rate": 1.4336064895927018e-05, + "loss": 1.0911, + "step": 30573 + }, + { + "epoch": 2.26, + "learning_rate": 1.4335705468330958e-05, + "loss": 1.1098, + "step": 30574 + }, + { + "epoch": 2.26, + "learning_rate": 1.4335346033836856e-05, + "loss": 0.9922, + "step": 30575 + }, + { + "epoch": 2.26, + "learning_rate": 1.4334986592445276e-05, + "loss": 1.0983, + "step": 30576 + }, + { + "epoch": 2.26, + "learning_rate": 1.43346271441568e-05, + "loss": 0.9748, + "step": 30577 + }, + { + "epoch": 2.26, + "learning_rate": 1.4334267688971992e-05, + "loss": 0.9973, + "step": 30578 + }, + { + "epoch": 2.26, + "learning_rate": 1.4333908226891426e-05, + "loss": 1.0865, + "step": 30579 + }, + { + "epoch": 2.26, + "learning_rate": 1.4333548757915675e-05, + "loss": 0.9817, + "step": 30580 + }, + { + "epoch": 2.26, + "learning_rate": 1.4333189282045307e-05, + "loss": 0.9219, + "step": 30581 + }, + { + "epoch": 2.26, + "learning_rate": 1.4332829799280896e-05, + "loss": 0.9632, + "step": 30582 + }, + { + "epoch": 2.26, + "learning_rate": 1.433247030962302e-05, + "loss": 1.0739, + "step": 30583 + }, + { + "epoch": 2.26, + "learning_rate": 1.4332110813072242e-05, + "loss": 0.9933, + "step": 30584 + }, + { + "epoch": 2.26, + "learning_rate": 1.4331751309629138e-05, + "loss": 1.1533, + "step": 30585 + }, + { + "epoch": 2.26, + "learning_rate": 1.433139179929428e-05, + "loss": 1.1413, + "step": 30586 + }, + { + "epoch": 2.26, + "learning_rate": 1.4331032282068241e-05, + "loss": 1.0007, + "step": 30587 + }, + { + "epoch": 2.26, + "learning_rate": 1.4330672757951592e-05, + "loss": 1.0785, + "step": 30588 + }, + { + "epoch": 2.26, + "learning_rate": 1.43303132269449e-05, + "loss": 1.1295, + "step": 30589 + }, + { + "epoch": 2.26, + "learning_rate": 1.4329953689048746e-05, + "loss": 1.0072, + "step": 30590 + }, + { + "epoch": 2.26, + "learning_rate": 1.4329594144263695e-05, + "loss": 0.965, + "step": 30591 + }, + { + "epoch": 2.26, + "learning_rate": 1.4329234592590324e-05, + "loss": 1.0163, + "step": 30592 + }, + { + "epoch": 2.26, + "learning_rate": 1.4328875034029199e-05, + "loss": 1.0806, + "step": 30593 + }, + { + "epoch": 2.26, + "learning_rate": 1.43285154685809e-05, + "loss": 0.8955, + "step": 30594 + }, + { + "epoch": 2.26, + "learning_rate": 1.4328155896245995e-05, + "loss": 1.0042, + "step": 30595 + }, + { + "epoch": 2.26, + "learning_rate": 1.4327796317025053e-05, + "loss": 0.9819, + "step": 30596 + }, + { + "epoch": 2.26, + "learning_rate": 1.432743673091865e-05, + "loss": 1.0048, + "step": 30597 + }, + { + "epoch": 2.26, + "learning_rate": 1.432707713792736e-05, + "loss": 0.9872, + "step": 30598 + }, + { + "epoch": 2.26, + "learning_rate": 1.4326717538051746e-05, + "loss": 1.0109, + "step": 30599 + }, + { + "epoch": 2.26, + "learning_rate": 1.4326357931292391e-05, + "loss": 1.1105, + "step": 30600 + }, + { + "epoch": 2.26, + "learning_rate": 1.4325998317649858e-05, + "loss": 1.0706, + "step": 30601 + }, + { + "epoch": 2.26, + "learning_rate": 1.4325638697124729e-05, + "loss": 0.9083, + "step": 30602 + }, + { + "epoch": 2.26, + "learning_rate": 1.4325279069717568e-05, + "loss": 1.0642, + "step": 30603 + }, + { + "epoch": 2.26, + "learning_rate": 1.432491943542895e-05, + "loss": 0.9985, + "step": 30604 + }, + { + "epoch": 2.26, + "learning_rate": 1.432455979425945e-05, + "loss": 0.8921, + "step": 30605 + }, + { + "epoch": 2.26, + "learning_rate": 1.4324200146209633e-05, + "loss": 0.9236, + "step": 30606 + }, + { + "epoch": 2.26, + "learning_rate": 1.4323840491280076e-05, + "loss": 1.0873, + "step": 30607 + }, + { + "epoch": 2.26, + "learning_rate": 1.4323480829471352e-05, + "loss": 0.9177, + "step": 30608 + }, + { + "epoch": 2.26, + "learning_rate": 1.432312116078403e-05, + "loss": 0.9997, + "step": 30609 + }, + { + "epoch": 2.26, + "learning_rate": 1.4322761485218688e-05, + "loss": 0.881, + "step": 30610 + }, + { + "epoch": 2.26, + "learning_rate": 1.4322401802775889e-05, + "loss": 0.9767, + "step": 30611 + }, + { + "epoch": 2.26, + "learning_rate": 1.4322042113456215e-05, + "loss": 1.0377, + "step": 30612 + }, + { + "epoch": 2.26, + "learning_rate": 1.4321682417260232e-05, + "loss": 0.9478, + "step": 30613 + }, + { + "epoch": 2.26, + "learning_rate": 1.4321322714188514e-05, + "loss": 0.9445, + "step": 30614 + }, + { + "epoch": 2.26, + "learning_rate": 1.4320963004241635e-05, + "loss": 0.9578, + "step": 30615 + }, + { + "epoch": 2.26, + "learning_rate": 1.4320603287420164e-05, + "loss": 0.987, + "step": 30616 + }, + { + "epoch": 2.26, + "learning_rate": 1.4320243563724677e-05, + "loss": 0.9164, + "step": 30617 + }, + { + "epoch": 2.26, + "learning_rate": 1.4319883833155744e-05, + "loss": 0.9701, + "step": 30618 + }, + { + "epoch": 2.26, + "learning_rate": 1.4319524095713933e-05, + "loss": 0.9083, + "step": 30619 + }, + { + "epoch": 2.26, + "learning_rate": 1.4319164351399826e-05, + "loss": 1.1247, + "step": 30620 + }, + { + "epoch": 2.26, + "learning_rate": 1.4318804600213987e-05, + "loss": 0.9262, + "step": 30621 + }, + { + "epoch": 2.26, + "learning_rate": 1.4318444842156996e-05, + "loss": 1.0592, + "step": 30622 + }, + { + "epoch": 2.26, + "learning_rate": 1.4318085077229415e-05, + "loss": 1.0506, + "step": 30623 + }, + { + "epoch": 2.26, + "learning_rate": 1.431772530543183e-05, + "loss": 0.9867, + "step": 30624 + }, + { + "epoch": 2.26, + "learning_rate": 1.43173655267648e-05, + "loss": 1.1496, + "step": 30625 + }, + { + "epoch": 2.26, + "learning_rate": 1.4317005741228907e-05, + "loss": 0.9449, + "step": 30626 + }, + { + "epoch": 2.26, + "learning_rate": 1.4316645948824714e-05, + "loss": 0.9639, + "step": 30627 + }, + { + "epoch": 2.26, + "learning_rate": 1.4316286149552805e-05, + "loss": 1.0457, + "step": 30628 + }, + { + "epoch": 2.26, + "learning_rate": 1.4315926343413743e-05, + "loss": 0.9487, + "step": 30629 + }, + { + "epoch": 2.26, + "learning_rate": 1.4315566530408106e-05, + "loss": 0.896, + "step": 30630 + }, + { + "epoch": 2.26, + "learning_rate": 1.431520671053646e-05, + "loss": 0.9766, + "step": 30631 + }, + { + "epoch": 2.26, + "learning_rate": 1.4314846883799388e-05, + "loss": 1.0129, + "step": 30632 + }, + { + "epoch": 2.26, + "learning_rate": 1.4314487050197454e-05, + "loss": 0.9811, + "step": 30633 + }, + { + "epoch": 2.26, + "learning_rate": 1.431412720973123e-05, + "loss": 1.0374, + "step": 30634 + }, + { + "epoch": 2.26, + "learning_rate": 1.4313767362401296e-05, + "loss": 1.159, + "step": 30635 + }, + { + "epoch": 2.26, + "learning_rate": 1.4313407508208217e-05, + "loss": 0.9281, + "step": 30636 + }, + { + "epoch": 2.26, + "learning_rate": 1.4313047647152569e-05, + "loss": 1.0067, + "step": 30637 + }, + { + "epoch": 2.26, + "learning_rate": 1.4312687779234921e-05, + "loss": 1.0289, + "step": 30638 + }, + { + "epoch": 2.26, + "learning_rate": 1.4312327904455854e-05, + "loss": 0.8515, + "step": 30639 + }, + { + "epoch": 2.26, + "learning_rate": 1.4311968022815932e-05, + "loss": 0.8827, + "step": 30640 + }, + { + "epoch": 2.26, + "learning_rate": 1.4311608134315731e-05, + "loss": 0.9464, + "step": 30641 + }, + { + "epoch": 2.26, + "learning_rate": 1.4311248238955823e-05, + "loss": 1.0116, + "step": 30642 + }, + { + "epoch": 2.26, + "learning_rate": 1.4310888336736782e-05, + "loss": 0.9802, + "step": 30643 + }, + { + "epoch": 2.26, + "learning_rate": 1.4310528427659176e-05, + "loss": 1.0066, + "step": 30644 + }, + { + "epoch": 2.26, + "learning_rate": 1.4310168511723586e-05, + "loss": 1.0204, + "step": 30645 + }, + { + "epoch": 2.26, + "learning_rate": 1.4309808588930573e-05, + "loss": 1.0564, + "step": 30646 + }, + { + "epoch": 2.26, + "learning_rate": 1.4309448659280722e-05, + "loss": 1.152, + "step": 30647 + }, + { + "epoch": 2.26, + "learning_rate": 1.4309088722774595e-05, + "loss": 0.9669, + "step": 30648 + }, + { + "epoch": 2.26, + "learning_rate": 1.4308728779412775e-05, + "loss": 1.0004, + "step": 30649 + }, + { + "epoch": 2.26, + "learning_rate": 1.4308368829195825e-05, + "loss": 1.0131, + "step": 30650 + }, + { + "epoch": 2.26, + "learning_rate": 1.4308008872124324e-05, + "loss": 1.0895, + "step": 30651 + }, + { + "epoch": 2.26, + "learning_rate": 1.4307648908198841e-05, + "loss": 1.1362, + "step": 30652 + }, + { + "epoch": 2.26, + "learning_rate": 1.4307288937419952e-05, + "loss": 1.0807, + "step": 30653 + }, + { + "epoch": 2.27, + "learning_rate": 1.4306928959788227e-05, + "loss": 0.9667, + "step": 30654 + }, + { + "epoch": 2.27, + "learning_rate": 1.4306568975304237e-05, + "loss": 1.1659, + "step": 30655 + }, + { + "epoch": 2.27, + "learning_rate": 1.4306208983968561e-05, + "loss": 0.9458, + "step": 30656 + }, + { + "epoch": 2.27, + "learning_rate": 1.4305848985781768e-05, + "loss": 1.0079, + "step": 30657 + }, + { + "epoch": 2.27, + "learning_rate": 1.430548898074443e-05, + "loss": 0.9997, + "step": 30658 + }, + { + "epoch": 2.27, + "learning_rate": 1.430512896885712e-05, + "loss": 1.0163, + "step": 30659 + }, + { + "epoch": 2.27, + "learning_rate": 1.4304768950120411e-05, + "loss": 0.9386, + "step": 30660 + }, + { + "epoch": 2.27, + "learning_rate": 1.430440892453488e-05, + "loss": 0.9278, + "step": 30661 + }, + { + "epoch": 2.27, + "learning_rate": 1.430404889210109e-05, + "loss": 1.0737, + "step": 30662 + }, + { + "epoch": 2.27, + "learning_rate": 1.4303688852819626e-05, + "loss": 1.0096, + "step": 30663 + }, + { + "epoch": 2.27, + "learning_rate": 1.4303328806691049e-05, + "loss": 0.9912, + "step": 30664 + }, + { + "epoch": 2.27, + "learning_rate": 1.430296875371594e-05, + "loss": 1.0524, + "step": 30665 + }, + { + "epoch": 2.27, + "learning_rate": 1.4302608693894868e-05, + "loss": 0.9724, + "step": 30666 + }, + { + "epoch": 2.27, + "learning_rate": 1.430224862722841e-05, + "loss": 1.0167, + "step": 30667 + }, + { + "epoch": 2.27, + "learning_rate": 1.4301888553717133e-05, + "loss": 1.0783, + "step": 30668 + }, + { + "epoch": 2.27, + "learning_rate": 1.4301528473361615e-05, + "loss": 0.9398, + "step": 30669 + }, + { + "epoch": 2.27, + "learning_rate": 1.4301168386162424e-05, + "loss": 1.0953, + "step": 30670 + }, + { + "epoch": 2.27, + "learning_rate": 1.4300808292120141e-05, + "loss": 0.9455, + "step": 30671 + }, + { + "epoch": 2.27, + "learning_rate": 1.4300448191235327e-05, + "loss": 0.9858, + "step": 30672 + }, + { + "epoch": 2.27, + "learning_rate": 1.4300088083508568e-05, + "loss": 0.902, + "step": 30673 + }, + { + "epoch": 2.27, + "learning_rate": 1.4299727968940425e-05, + "loss": 1.0626, + "step": 30674 + }, + { + "epoch": 2.27, + "learning_rate": 1.429936784753148e-05, + "loss": 0.9057, + "step": 30675 + }, + { + "epoch": 2.27, + "learning_rate": 1.4299007719282299e-05, + "loss": 0.9749, + "step": 30676 + }, + { + "epoch": 2.27, + "learning_rate": 1.429864758419346e-05, + "loss": 1.0118, + "step": 30677 + }, + { + "epoch": 2.27, + "learning_rate": 1.4298287442265532e-05, + "loss": 1.0776, + "step": 30678 + }, + { + "epoch": 2.27, + "learning_rate": 1.4297927293499094e-05, + "loss": 0.9203, + "step": 30679 + }, + { + "epoch": 2.27, + "learning_rate": 1.4297567137894714e-05, + "loss": 0.9943, + "step": 30680 + }, + { + "epoch": 2.27, + "learning_rate": 1.4297206975452965e-05, + "loss": 0.9559, + "step": 30681 + }, + { + "epoch": 2.27, + "learning_rate": 1.429684680617442e-05, + "loss": 0.9634, + "step": 30682 + }, + { + "epoch": 2.27, + "learning_rate": 1.4296486630059657e-05, + "loss": 1.0195, + "step": 30683 + }, + { + "epoch": 2.27, + "learning_rate": 1.4296126447109243e-05, + "loss": 1.0402, + "step": 30684 + }, + { + "epoch": 2.27, + "learning_rate": 1.4295766257323752e-05, + "loss": 0.9975, + "step": 30685 + }, + { + "epoch": 2.27, + "learning_rate": 1.429540606070376e-05, + "loss": 1.0226, + "step": 30686 + }, + { + "epoch": 2.27, + "learning_rate": 1.429504585724984e-05, + "loss": 1.0005, + "step": 30687 + }, + { + "epoch": 2.27, + "learning_rate": 1.4294685646962558e-05, + "loss": 0.9905, + "step": 30688 + }, + { + "epoch": 2.27, + "learning_rate": 1.42943254298425e-05, + "loss": 1.035, + "step": 30689 + }, + { + "epoch": 2.27, + "learning_rate": 1.4293965205890228e-05, + "loss": 1.0083, + "step": 30690 + }, + { + "epoch": 2.27, + "learning_rate": 1.4293604975106317e-05, + "loss": 0.9383, + "step": 30691 + }, + { + "epoch": 2.27, + "learning_rate": 1.4293244737491342e-05, + "loss": 0.981, + "step": 30692 + }, + { + "epoch": 2.27, + "learning_rate": 1.4292884493045877e-05, + "loss": 1.1006, + "step": 30693 + }, + { + "epoch": 2.27, + "learning_rate": 1.4292524241770496e-05, + "loss": 1.0528, + "step": 30694 + }, + { + "epoch": 2.27, + "learning_rate": 1.429216398366577e-05, + "loss": 1.0306, + "step": 30695 + }, + { + "epoch": 2.27, + "learning_rate": 1.4291803718732271e-05, + "loss": 1.0502, + "step": 30696 + }, + { + "epoch": 2.27, + "learning_rate": 1.4291443446970575e-05, + "loss": 0.9201, + "step": 30697 + }, + { + "epoch": 2.27, + "learning_rate": 1.4291083168381252e-05, + "loss": 0.8998, + "step": 30698 + }, + { + "epoch": 2.27, + "learning_rate": 1.429072288296488e-05, + "loss": 1.066, + "step": 30699 + }, + { + "epoch": 2.27, + "learning_rate": 1.4290362590722028e-05, + "loss": 1.1243, + "step": 30700 + }, + { + "epoch": 2.27, + "learning_rate": 1.4290002291653269e-05, + "loss": 0.9834, + "step": 30701 + }, + { + "epoch": 2.27, + "learning_rate": 1.428964198575918e-05, + "loss": 1.0379, + "step": 30702 + }, + { + "epoch": 2.27, + "learning_rate": 1.428928167304033e-05, + "loss": 0.9444, + "step": 30703 + }, + { + "epoch": 2.27, + "learning_rate": 1.4288921353497295e-05, + "loss": 0.9445, + "step": 30704 + }, + { + "epoch": 2.27, + "learning_rate": 1.4288561027130647e-05, + "loss": 1.0456, + "step": 30705 + }, + { + "epoch": 2.27, + "learning_rate": 1.4288200693940961e-05, + "loss": 0.8936, + "step": 30706 + }, + { + "epoch": 2.27, + "learning_rate": 1.4287840353928807e-05, + "loss": 0.9308, + "step": 30707 + }, + { + "epoch": 2.27, + "learning_rate": 1.4287480007094765e-05, + "loss": 1.0342, + "step": 30708 + }, + { + "epoch": 2.27, + "learning_rate": 1.42871196534394e-05, + "loss": 1.0266, + "step": 30709 + }, + { + "epoch": 2.27, + "learning_rate": 1.428675929296329e-05, + "loss": 1.14, + "step": 30710 + }, + { + "epoch": 2.27, + "learning_rate": 1.4286398925667004e-05, + "loss": 0.9948, + "step": 30711 + }, + { + "epoch": 2.27, + "learning_rate": 1.4286038551551122e-05, + "loss": 1.1337, + "step": 30712 + }, + { + "epoch": 2.27, + "learning_rate": 1.4285678170616213e-05, + "loss": 0.9936, + "step": 30713 + }, + { + "epoch": 2.27, + "learning_rate": 1.4285317782862853e-05, + "loss": 0.9464, + "step": 30714 + }, + { + "epoch": 2.27, + "learning_rate": 1.428495738829161e-05, + "loss": 1.0705, + "step": 30715 + }, + { + "epoch": 2.27, + "learning_rate": 1.4284596986903065e-05, + "loss": 0.8772, + "step": 30716 + }, + { + "epoch": 2.27, + "learning_rate": 1.4284236578697788e-05, + "loss": 0.9585, + "step": 30717 + }, + { + "epoch": 2.27, + "learning_rate": 1.4283876163676349e-05, + "loss": 0.974, + "step": 30718 + }, + { + "epoch": 2.27, + "learning_rate": 1.4283515741839325e-05, + "loss": 1.0044, + "step": 30719 + }, + { + "epoch": 2.27, + "learning_rate": 1.4283155313187289e-05, + "loss": 1.1209, + "step": 30720 + }, + { + "epoch": 2.27, + "learning_rate": 1.4282794877720814e-05, + "loss": 1.0168, + "step": 30721 + }, + { + "epoch": 2.27, + "learning_rate": 1.4282434435440474e-05, + "loss": 1.0581, + "step": 30722 + }, + { + "epoch": 2.27, + "learning_rate": 1.428207398634684e-05, + "loss": 1.0516, + "step": 30723 + }, + { + "epoch": 2.27, + "learning_rate": 1.4281713530440492e-05, + "loss": 1.0071, + "step": 30724 + }, + { + "epoch": 2.27, + "learning_rate": 1.4281353067721992e-05, + "loss": 0.96, + "step": 30725 + }, + { + "epoch": 2.27, + "learning_rate": 1.4280992598191927e-05, + "loss": 0.9752, + "step": 30726 + }, + { + "epoch": 2.27, + "learning_rate": 1.4280632121850864e-05, + "loss": 0.9268, + "step": 30727 + }, + { + "epoch": 2.27, + "learning_rate": 1.4280271638699375e-05, + "loss": 1.1013, + "step": 30728 + }, + { + "epoch": 2.27, + "learning_rate": 1.4279911148738032e-05, + "loss": 0.9536, + "step": 30729 + }, + { + "epoch": 2.27, + "learning_rate": 1.4279550651967415e-05, + "loss": 1.0405, + "step": 30730 + }, + { + "epoch": 2.27, + "learning_rate": 1.4279190148388092e-05, + "loss": 0.9962, + "step": 30731 + }, + { + "epoch": 2.27, + "learning_rate": 1.4278829638000643e-05, + "loss": 1.0771, + "step": 30732 + }, + { + "epoch": 2.27, + "learning_rate": 1.4278469120805632e-05, + "loss": 1.0644, + "step": 30733 + }, + { + "epoch": 2.27, + "learning_rate": 1.427810859680364e-05, + "loss": 1.0696, + "step": 30734 + }, + { + "epoch": 2.27, + "learning_rate": 1.4277748065995236e-05, + "loss": 1.0518, + "step": 30735 + }, + { + "epoch": 2.27, + "learning_rate": 1.4277387528381002e-05, + "loss": 0.9214, + "step": 30736 + }, + { + "epoch": 2.27, + "learning_rate": 1.4277026983961502e-05, + "loss": 1.0132, + "step": 30737 + }, + { + "epoch": 2.27, + "learning_rate": 1.4276666432737312e-05, + "loss": 1.0632, + "step": 30738 + }, + { + "epoch": 2.27, + "learning_rate": 1.427630587470901e-05, + "loss": 0.9759, + "step": 30739 + }, + { + "epoch": 2.27, + "learning_rate": 1.4275945309877161e-05, + "loss": 0.9858, + "step": 30740 + }, + { + "epoch": 2.27, + "learning_rate": 1.427558473824235e-05, + "loss": 1.0287, + "step": 30741 + }, + { + "epoch": 2.27, + "learning_rate": 1.4275224159805143e-05, + "loss": 1.028, + "step": 30742 + }, + { + "epoch": 2.27, + "learning_rate": 1.4274863574566116e-05, + "loss": 1.0054, + "step": 30743 + }, + { + "epoch": 2.27, + "learning_rate": 1.4274502982525843e-05, + "loss": 0.9376, + "step": 30744 + }, + { + "epoch": 2.27, + "learning_rate": 1.4274142383684894e-05, + "loss": 0.9942, + "step": 30745 + }, + { + "epoch": 2.27, + "learning_rate": 1.4273781778043847e-05, + "loss": 0.964, + "step": 30746 + }, + { + "epoch": 2.27, + "learning_rate": 1.4273421165603276e-05, + "loss": 1.0267, + "step": 30747 + }, + { + "epoch": 2.27, + "learning_rate": 1.427306054636375e-05, + "loss": 0.9659, + "step": 30748 + }, + { + "epoch": 2.27, + "learning_rate": 1.427269992032585e-05, + "loss": 0.952, + "step": 30749 + }, + { + "epoch": 2.27, + "learning_rate": 1.427233928749014e-05, + "loss": 0.8814, + "step": 30750 + }, + { + "epoch": 2.27, + "learning_rate": 1.4271978647857203e-05, + "loss": 1.0039, + "step": 30751 + }, + { + "epoch": 2.27, + "learning_rate": 1.4271618001427608e-05, + "loss": 1.0223, + "step": 30752 + }, + { + "epoch": 2.27, + "learning_rate": 1.427125734820193e-05, + "loss": 1.018, + "step": 30753 + }, + { + "epoch": 2.27, + "learning_rate": 1.4270896688180742e-05, + "loss": 0.978, + "step": 30754 + }, + { + "epoch": 2.27, + "learning_rate": 1.4270536021364622e-05, + "loss": 0.9293, + "step": 30755 + }, + { + "epoch": 2.27, + "learning_rate": 1.4270175347754136e-05, + "loss": 1.0361, + "step": 30756 + }, + { + "epoch": 2.27, + "learning_rate": 1.4269814667349866e-05, + "loss": 1.1204, + "step": 30757 + }, + { + "epoch": 2.27, + "learning_rate": 1.4269453980152379e-05, + "loss": 0.9675, + "step": 30758 + }, + { + "epoch": 2.27, + "learning_rate": 1.4269093286162252e-05, + "loss": 0.996, + "step": 30759 + }, + { + "epoch": 2.27, + "learning_rate": 1.4268732585380059e-05, + "loss": 1.0392, + "step": 30760 + }, + { + "epoch": 2.27, + "learning_rate": 1.4268371877806375e-05, + "loss": 1.0456, + "step": 30761 + }, + { + "epoch": 2.27, + "learning_rate": 1.4268011163441768e-05, + "loss": 0.9359, + "step": 30762 + }, + { + "epoch": 2.27, + "learning_rate": 1.4267650442286823e-05, + "loss": 0.9935, + "step": 30763 + }, + { + "epoch": 2.27, + "learning_rate": 1.4267289714342103e-05, + "loss": 1.0853, + "step": 30764 + }, + { + "epoch": 2.27, + "learning_rate": 1.4266928979608188e-05, + "loss": 1.064, + "step": 30765 + }, + { + "epoch": 2.27, + "learning_rate": 1.4266568238085648e-05, + "loss": 0.9308, + "step": 30766 + }, + { + "epoch": 2.27, + "learning_rate": 1.4266207489775062e-05, + "loss": 1.1068, + "step": 30767 + }, + { + "epoch": 2.27, + "learning_rate": 1.4265846734676996e-05, + "loss": 0.9781, + "step": 30768 + }, + { + "epoch": 2.27, + "learning_rate": 1.4265485972792032e-05, + "loss": 1.1285, + "step": 30769 + }, + { + "epoch": 2.27, + "learning_rate": 1.426512520412074e-05, + "loss": 1.113, + "step": 30770 + }, + { + "epoch": 2.27, + "learning_rate": 1.4264764428663696e-05, + "loss": 0.9777, + "step": 30771 + }, + { + "epoch": 2.27, + "learning_rate": 1.4264403646421473e-05, + "loss": 1.0719, + "step": 30772 + }, + { + "epoch": 2.27, + "learning_rate": 1.4264042857394645e-05, + "loss": 0.9642, + "step": 30773 + }, + { + "epoch": 2.27, + "learning_rate": 1.4263682061583784e-05, + "loss": 1.0263, + "step": 30774 + }, + { + "epoch": 2.27, + "learning_rate": 1.4263321258989467e-05, + "loss": 0.9378, + "step": 30775 + }, + { + "epoch": 2.27, + "learning_rate": 1.4262960449612264e-05, + "loss": 0.9732, + "step": 30776 + }, + { + "epoch": 2.27, + "learning_rate": 1.4262599633452756e-05, + "loss": 1.0005, + "step": 30777 + }, + { + "epoch": 2.27, + "learning_rate": 1.426223881051151e-05, + "loss": 0.9522, + "step": 30778 + }, + { + "epoch": 2.27, + "learning_rate": 1.4261877980789104e-05, + "loss": 1.0874, + "step": 30779 + }, + { + "epoch": 2.27, + "learning_rate": 1.4261517144286112e-05, + "loss": 0.9712, + "step": 30780 + }, + { + "epoch": 2.27, + "learning_rate": 1.4261156301003107e-05, + "loss": 0.8814, + "step": 30781 + }, + { + "epoch": 2.27, + "learning_rate": 1.4260795450940662e-05, + "loss": 0.9511, + "step": 30782 + }, + { + "epoch": 2.27, + "learning_rate": 1.4260434594099354e-05, + "loss": 0.9609, + "step": 30783 + }, + { + "epoch": 2.27, + "learning_rate": 1.4260073730479755e-05, + "loss": 0.9407, + "step": 30784 + }, + { + "epoch": 2.27, + "learning_rate": 1.425971286008244e-05, + "loss": 1.0254, + "step": 30785 + }, + { + "epoch": 2.27, + "learning_rate": 1.4259351982907983e-05, + "loss": 1.047, + "step": 30786 + }, + { + "epoch": 2.27, + "learning_rate": 1.4258991098956956e-05, + "loss": 0.9483, + "step": 30787 + }, + { + "epoch": 2.27, + "learning_rate": 1.4258630208229938e-05, + "loss": 0.9845, + "step": 30788 + }, + { + "epoch": 2.27, + "learning_rate": 1.42582693107275e-05, + "loss": 1.1284, + "step": 30789 + }, + { + "epoch": 2.28, + "learning_rate": 1.4257908406450211e-05, + "loss": 1.0066, + "step": 30790 + }, + { + "epoch": 2.28, + "learning_rate": 1.4257547495398658e-05, + "loss": 0.9538, + "step": 30791 + }, + { + "epoch": 2.28, + "learning_rate": 1.4257186577573406e-05, + "loss": 1.0023, + "step": 30792 + }, + { + "epoch": 2.28, + "learning_rate": 1.4256825652975026e-05, + "loss": 0.9274, + "step": 30793 + }, + { + "epoch": 2.28, + "learning_rate": 1.4256464721604101e-05, + "loss": 0.9847, + "step": 30794 + }, + { + "epoch": 2.28, + "learning_rate": 1.4256103783461202e-05, + "loss": 0.9496, + "step": 30795 + }, + { + "epoch": 2.28, + "learning_rate": 1.4255742838546902e-05, + "loss": 0.9834, + "step": 30796 + }, + { + "epoch": 2.28, + "learning_rate": 1.4255381886861777e-05, + "loss": 1.0476, + "step": 30797 + }, + { + "epoch": 2.28, + "learning_rate": 1.4255020928406401e-05, + "loss": 0.9965, + "step": 30798 + }, + { + "epoch": 2.28, + "learning_rate": 1.4254659963181343e-05, + "loss": 1.0418, + "step": 30799 + }, + { + "epoch": 2.28, + "learning_rate": 1.4254298991187185e-05, + "loss": 0.8534, + "step": 30800 + }, + { + "epoch": 2.28, + "learning_rate": 1.42539380124245e-05, + "loss": 1.0156, + "step": 30801 + }, + { + "epoch": 2.28, + "learning_rate": 1.4253577026893859e-05, + "loss": 1.015, + "step": 30802 + }, + { + "epoch": 2.28, + "learning_rate": 1.4253216034595836e-05, + "loss": 1.106, + "step": 30803 + }, + { + "epoch": 2.28, + "learning_rate": 1.425285503553101e-05, + "loss": 0.9514, + "step": 30804 + }, + { + "epoch": 2.28, + "learning_rate": 1.425249402969995e-05, + "loss": 0.9657, + "step": 30805 + }, + { + "epoch": 2.28, + "learning_rate": 1.4252133017103236e-05, + "loss": 0.992, + "step": 30806 + }, + { + "epoch": 2.28, + "learning_rate": 1.4251771997741435e-05, + "loss": 0.8517, + "step": 30807 + }, + { + "epoch": 2.28, + "learning_rate": 1.4251410971615129e-05, + "loss": 1.009, + "step": 30808 + }, + { + "epoch": 2.28, + "learning_rate": 1.4251049938724886e-05, + "loss": 0.9734, + "step": 30809 + }, + { + "epoch": 2.28, + "learning_rate": 1.4250688899071285e-05, + "loss": 0.9787, + "step": 30810 + }, + { + "epoch": 2.28, + "learning_rate": 1.4250327852654902e-05, + "loss": 1.1518, + "step": 30811 + }, + { + "epoch": 2.28, + "learning_rate": 1.4249966799476304e-05, + "loss": 0.9914, + "step": 30812 + }, + { + "epoch": 2.28, + "learning_rate": 1.4249605739536069e-05, + "loss": 0.9725, + "step": 30813 + }, + { + "epoch": 2.28, + "learning_rate": 1.4249244672834777e-05, + "loss": 1.0539, + "step": 30814 + }, + { + "epoch": 2.28, + "learning_rate": 1.4248883599372994e-05, + "loss": 1.0147, + "step": 30815 + }, + { + "epoch": 2.28, + "learning_rate": 1.4248522519151297e-05, + "loss": 1.0715, + "step": 30816 + }, + { + "epoch": 2.28, + "learning_rate": 1.4248161432170262e-05, + "loss": 0.9111, + "step": 30817 + }, + { + "epoch": 2.28, + "learning_rate": 1.4247800338430463e-05, + "loss": 1.0178, + "step": 30818 + }, + { + "epoch": 2.28, + "learning_rate": 1.4247439237932477e-05, + "loss": 1.0628, + "step": 30819 + }, + { + "epoch": 2.28, + "learning_rate": 1.4247078130676876e-05, + "loss": 1.1145, + "step": 30820 + }, + { + "epoch": 2.28, + "learning_rate": 1.4246717016664232e-05, + "loss": 1.0333, + "step": 30821 + }, + { + "epoch": 2.28, + "learning_rate": 1.4246355895895124e-05, + "loss": 0.9735, + "step": 30822 + }, + { + "epoch": 2.28, + "learning_rate": 1.424599476837012e-05, + "loss": 0.928, + "step": 30823 + }, + { + "epoch": 2.28, + "learning_rate": 1.4245633634089804e-05, + "loss": 0.9056, + "step": 30824 + }, + { + "epoch": 2.28, + "learning_rate": 1.4245272493054742e-05, + "loss": 1.0486, + "step": 30825 + }, + { + "epoch": 2.28, + "learning_rate": 1.4244911345265515e-05, + "loss": 0.9064, + "step": 30826 + }, + { + "epoch": 2.28, + "learning_rate": 1.4244550190722693e-05, + "loss": 1.0162, + "step": 30827 + }, + { + "epoch": 2.28, + "learning_rate": 1.4244189029426854e-05, + "loss": 0.9617, + "step": 30828 + }, + { + "epoch": 2.28, + "learning_rate": 1.4243827861378568e-05, + "loss": 1.1086, + "step": 30829 + }, + { + "epoch": 2.28, + "learning_rate": 1.4243466686578416e-05, + "loss": 1.0314, + "step": 30830 + }, + { + "epoch": 2.28, + "learning_rate": 1.4243105505026969e-05, + "loss": 0.97, + "step": 30831 + }, + { + "epoch": 2.28, + "learning_rate": 1.4242744316724801e-05, + "loss": 0.9627, + "step": 30832 + }, + { + "epoch": 2.28, + "learning_rate": 1.4242383121672484e-05, + "loss": 0.9679, + "step": 30833 + }, + { + "epoch": 2.28, + "learning_rate": 1.4242021919870604e-05, + "loss": 0.9869, + "step": 30834 + }, + { + "epoch": 2.28, + "learning_rate": 1.4241660711319721e-05, + "loss": 1.0415, + "step": 30835 + }, + { + "epoch": 2.28, + "learning_rate": 1.4241299496020418e-05, + "loss": 1.0933, + "step": 30836 + }, + { + "epoch": 2.28, + "learning_rate": 1.4240938273973268e-05, + "loss": 0.9532, + "step": 30837 + }, + { + "epoch": 2.28, + "learning_rate": 1.424057704517885e-05, + "loss": 1.1551, + "step": 30838 + }, + { + "epoch": 2.28, + "learning_rate": 1.4240215809637729e-05, + "loss": 0.9669, + "step": 30839 + }, + { + "epoch": 2.28, + "learning_rate": 1.4239854567350488e-05, + "loss": 0.9637, + "step": 30840 + }, + { + "epoch": 2.28, + "learning_rate": 1.4239493318317701e-05, + "loss": 1.1651, + "step": 30841 + }, + { + "epoch": 2.28, + "learning_rate": 1.4239132062539935e-05, + "loss": 0.9431, + "step": 30842 + }, + { + "epoch": 2.28, + "learning_rate": 1.4238770800017777e-05, + "loss": 0.9267, + "step": 30843 + }, + { + "epoch": 2.28, + "learning_rate": 1.423840953075179e-05, + "loss": 0.9981, + "step": 30844 + }, + { + "epoch": 2.28, + "learning_rate": 1.4238048254742558e-05, + "loss": 0.9926, + "step": 30845 + }, + { + "epoch": 2.28, + "learning_rate": 1.4237686971990647e-05, + "loss": 0.877, + "step": 30846 + }, + { + "epoch": 2.28, + "learning_rate": 1.4237325682496641e-05, + "loss": 1.1308, + "step": 30847 + }, + { + "epoch": 2.28, + "learning_rate": 1.4236964386261111e-05, + "loss": 0.956, + "step": 30848 + }, + { + "epoch": 2.28, + "learning_rate": 1.423660308328463e-05, + "loss": 1.0936, + "step": 30849 + }, + { + "epoch": 2.28, + "learning_rate": 1.4236241773567774e-05, + "loss": 0.909, + "step": 30850 + }, + { + "epoch": 2.28, + "learning_rate": 1.4235880457111117e-05, + "loss": 1.0035, + "step": 30851 + }, + { + "epoch": 2.28, + "learning_rate": 1.4235519133915236e-05, + "loss": 0.9842, + "step": 30852 + }, + { + "epoch": 2.28, + "learning_rate": 1.4235157803980703e-05, + "loss": 1.0896, + "step": 30853 + }, + { + "epoch": 2.28, + "learning_rate": 1.4234796467308096e-05, + "loss": 1.0565, + "step": 30854 + }, + { + "epoch": 2.28, + "learning_rate": 1.4234435123897989e-05, + "loss": 1.0244, + "step": 30855 + }, + { + "epoch": 2.28, + "learning_rate": 1.4234073773750954e-05, + "loss": 1.0623, + "step": 30856 + }, + { + "epoch": 2.28, + "learning_rate": 1.4233712416867572e-05, + "loss": 1.0273, + "step": 30857 + }, + { + "epoch": 2.28, + "learning_rate": 1.4233351053248412e-05, + "loss": 1.1314, + "step": 30858 + }, + { + "epoch": 2.28, + "learning_rate": 1.4232989682894051e-05, + "loss": 0.9721, + "step": 30859 + }, + { + "epoch": 2.28, + "learning_rate": 1.4232628305805064e-05, + "loss": 1.0352, + "step": 30860 + }, + { + "epoch": 2.28, + "learning_rate": 1.4232266921982025e-05, + "loss": 1.0519, + "step": 30861 + }, + { + "epoch": 2.28, + "learning_rate": 1.423190553142551e-05, + "loss": 1.065, + "step": 30862 + }, + { + "epoch": 2.28, + "learning_rate": 1.4231544134136095e-05, + "loss": 1.0179, + "step": 30863 + }, + { + "epoch": 2.28, + "learning_rate": 1.4231182730114353e-05, + "loss": 1.0146, + "step": 30864 + }, + { + "epoch": 2.28, + "learning_rate": 1.423082131936086e-05, + "loss": 0.9259, + "step": 30865 + }, + { + "epoch": 2.28, + "learning_rate": 1.4230459901876191e-05, + "loss": 1.0612, + "step": 30866 + }, + { + "epoch": 2.28, + "learning_rate": 1.4230098477660923e-05, + "loss": 1.0158, + "step": 30867 + }, + { + "epoch": 2.28, + "learning_rate": 1.4229737046715625e-05, + "loss": 1.0208, + "step": 30868 + }, + { + "epoch": 2.28, + "learning_rate": 1.4229375609040878e-05, + "loss": 0.995, + "step": 30869 + }, + { + "epoch": 2.28, + "learning_rate": 1.4229014164637254e-05, + "loss": 0.9922, + "step": 30870 + }, + { + "epoch": 2.28, + "learning_rate": 1.422865271350533e-05, + "loss": 1.0602, + "step": 30871 + }, + { + "epoch": 2.28, + "learning_rate": 1.4228291255645678e-05, + "loss": 0.9481, + "step": 30872 + }, + { + "epoch": 2.28, + "learning_rate": 1.4227929791058878e-05, + "loss": 0.9833, + "step": 30873 + }, + { + "epoch": 2.28, + "learning_rate": 1.42275683197455e-05, + "loss": 0.8569, + "step": 30874 + }, + { + "epoch": 2.28, + "learning_rate": 1.4227206841706123e-05, + "loss": 0.9754, + "step": 30875 + }, + { + "epoch": 2.28, + "learning_rate": 1.422684535694132e-05, + "loss": 1.1006, + "step": 30876 + }, + { + "epoch": 2.28, + "learning_rate": 1.4226483865451665e-05, + "loss": 0.9576, + "step": 30877 + }, + { + "epoch": 2.28, + "learning_rate": 1.4226122367237735e-05, + "loss": 1.146, + "step": 30878 + }, + { + "epoch": 2.28, + "learning_rate": 1.4225760862300107e-05, + "loss": 1.0092, + "step": 30879 + }, + { + "epoch": 2.28, + "learning_rate": 1.4225399350639351e-05, + "loss": 1.0131, + "step": 30880 + }, + { + "epoch": 2.28, + "learning_rate": 1.4225037832256047e-05, + "loss": 1.0204, + "step": 30881 + }, + { + "epoch": 2.28, + "learning_rate": 1.4224676307150766e-05, + "loss": 1.0826, + "step": 30882 + }, + { + "epoch": 2.28, + "learning_rate": 1.4224314775324089e-05, + "loss": 0.9534, + "step": 30883 + }, + { + "epoch": 2.28, + "learning_rate": 1.4223953236776584e-05, + "loss": 1.0384, + "step": 30884 + }, + { + "epoch": 2.28, + "learning_rate": 1.4223591691508834e-05, + "loss": 0.9582, + "step": 30885 + }, + { + "epoch": 2.28, + "learning_rate": 1.4223230139521407e-05, + "loss": 1.0156, + "step": 30886 + }, + { + "epoch": 2.28, + "learning_rate": 1.422286858081488e-05, + "loss": 0.9751, + "step": 30887 + }, + { + "epoch": 2.28, + "learning_rate": 1.4222507015389832e-05, + "loss": 0.924, + "step": 30888 + }, + { + "epoch": 2.28, + "learning_rate": 1.4222145443246836e-05, + "loss": 1.0157, + "step": 30889 + }, + { + "epoch": 2.28, + "learning_rate": 1.4221783864386465e-05, + "loss": 0.9869, + "step": 30890 + }, + { + "epoch": 2.28, + "learning_rate": 1.4221422278809296e-05, + "loss": 1.0246, + "step": 30891 + }, + { + "epoch": 2.28, + "learning_rate": 1.4221060686515907e-05, + "loss": 0.9754, + "step": 30892 + }, + { + "epoch": 2.28, + "learning_rate": 1.422069908750687e-05, + "loss": 0.9519, + "step": 30893 + }, + { + "epoch": 2.28, + "learning_rate": 1.422033748178276e-05, + "loss": 0.93, + "step": 30894 + }, + { + "epoch": 2.28, + "learning_rate": 1.4219975869344155e-05, + "loss": 1.0243, + "step": 30895 + }, + { + "epoch": 2.28, + "learning_rate": 1.4219614250191629e-05, + "loss": 1.0482, + "step": 30896 + }, + { + "epoch": 2.28, + "learning_rate": 1.4219252624325757e-05, + "loss": 0.9931, + "step": 30897 + }, + { + "epoch": 2.28, + "learning_rate": 1.4218890991747115e-05, + "loss": 0.9871, + "step": 30898 + }, + { + "epoch": 2.28, + "learning_rate": 1.4218529352456275e-05, + "loss": 1.0601, + "step": 30899 + }, + { + "epoch": 2.28, + "learning_rate": 1.4218167706453817e-05, + "loss": 1.0127, + "step": 30900 + }, + { + "epoch": 2.28, + "learning_rate": 1.4217806053740313e-05, + "loss": 0.8877, + "step": 30901 + }, + { + "epoch": 2.28, + "learning_rate": 1.4217444394316342e-05, + "loss": 0.9325, + "step": 30902 + }, + { + "epoch": 2.28, + "learning_rate": 1.4217082728182474e-05, + "loss": 1.033, + "step": 30903 + }, + { + "epoch": 2.28, + "learning_rate": 1.4216721055339293e-05, + "loss": 1.032, + "step": 30904 + }, + { + "epoch": 2.28, + "learning_rate": 1.4216359375787366e-05, + "loss": 1.0342, + "step": 30905 + }, + { + "epoch": 2.28, + "learning_rate": 1.4215997689527272e-05, + "loss": 1.0054, + "step": 30906 + }, + { + "epoch": 2.28, + "learning_rate": 1.4215635996559584e-05, + "loss": 1.043, + "step": 30907 + }, + { + "epoch": 2.28, + "learning_rate": 1.4215274296884882e-05, + "loss": 1.0552, + "step": 30908 + }, + { + "epoch": 2.28, + "learning_rate": 1.4214912590503738e-05, + "loss": 0.9488, + "step": 30909 + }, + { + "epoch": 2.28, + "learning_rate": 1.4214550877416728e-05, + "loss": 0.905, + "step": 30910 + }, + { + "epoch": 2.28, + "learning_rate": 1.4214189157624427e-05, + "loss": 1.0347, + "step": 30911 + }, + { + "epoch": 2.28, + "learning_rate": 1.4213827431127414e-05, + "loss": 0.9542, + "step": 30912 + }, + { + "epoch": 2.28, + "learning_rate": 1.421346569792626e-05, + "loss": 0.9685, + "step": 30913 + }, + { + "epoch": 2.28, + "learning_rate": 1.4213103958021546e-05, + "loss": 1.037, + "step": 30914 + }, + { + "epoch": 2.28, + "learning_rate": 1.4212742211413838e-05, + "loss": 0.9751, + "step": 30915 + }, + { + "epoch": 2.28, + "learning_rate": 1.4212380458103722e-05, + "loss": 1.0997, + "step": 30916 + }, + { + "epoch": 2.28, + "learning_rate": 1.4212018698091768e-05, + "loss": 1.1289, + "step": 30917 + }, + { + "epoch": 2.28, + "learning_rate": 1.421165693137855e-05, + "loss": 1.0242, + "step": 30918 + }, + { + "epoch": 2.28, + "learning_rate": 1.4211295157964648e-05, + "loss": 1.0235, + "step": 30919 + }, + { + "epoch": 2.28, + "learning_rate": 1.4210933377850635e-05, + "loss": 0.9429, + "step": 30920 + }, + { + "epoch": 2.28, + "learning_rate": 1.4210571591037088e-05, + "loss": 1.0364, + "step": 30921 + }, + { + "epoch": 2.28, + "learning_rate": 1.421020979752458e-05, + "loss": 0.9099, + "step": 30922 + }, + { + "epoch": 2.28, + "learning_rate": 1.4209847997313693e-05, + "loss": 1.0587, + "step": 30923 + }, + { + "epoch": 2.28, + "learning_rate": 1.4209486190404995e-05, + "loss": 0.9987, + "step": 30924 + }, + { + "epoch": 2.29, + "learning_rate": 1.4209124376799064e-05, + "loss": 1.0726, + "step": 30925 + }, + { + "epoch": 2.29, + "learning_rate": 1.4208762556496479e-05, + "loss": 1.006, + "step": 30926 + }, + { + "epoch": 2.29, + "learning_rate": 1.420840072949781e-05, + "loss": 0.9245, + "step": 30927 + }, + { + "epoch": 2.29, + "learning_rate": 1.4208038895803635e-05, + "loss": 1.0183, + "step": 30928 + }, + { + "epoch": 2.29, + "learning_rate": 1.4207677055414534e-05, + "loss": 1.0409, + "step": 30929 + }, + { + "epoch": 2.29, + "learning_rate": 1.4207315208331077e-05, + "loss": 1.0227, + "step": 30930 + }, + { + "epoch": 2.29, + "learning_rate": 1.4206953354553839e-05, + "loss": 1.0204, + "step": 30931 + }, + { + "epoch": 2.29, + "learning_rate": 1.4206591494083401e-05, + "loss": 0.9773, + "step": 30932 + }, + { + "epoch": 2.29, + "learning_rate": 1.4206229626920339e-05, + "loss": 0.9068, + "step": 30933 + }, + { + "epoch": 2.29, + "learning_rate": 1.4205867753065222e-05, + "loss": 1.0649, + "step": 30934 + }, + { + "epoch": 2.29, + "learning_rate": 1.4205505872518632e-05, + "loss": 1.1099, + "step": 30935 + }, + { + "epoch": 2.29, + "learning_rate": 1.420514398528114e-05, + "loss": 1.1251, + "step": 30936 + }, + { + "epoch": 2.29, + "learning_rate": 1.4204782091353324e-05, + "loss": 0.9928, + "step": 30937 + }, + { + "epoch": 2.29, + "learning_rate": 1.4204420190735762e-05, + "loss": 0.9471, + "step": 30938 + }, + { + "epoch": 2.29, + "learning_rate": 1.4204058283429025e-05, + "loss": 1.0122, + "step": 30939 + }, + { + "epoch": 2.29, + "learning_rate": 1.4203696369433694e-05, + "loss": 1.0183, + "step": 30940 + }, + { + "epoch": 2.29, + "learning_rate": 1.420333444875034e-05, + "loss": 1.0302, + "step": 30941 + }, + { + "epoch": 2.29, + "learning_rate": 1.4202972521379543e-05, + "loss": 1.0245, + "step": 30942 + }, + { + "epoch": 2.29, + "learning_rate": 1.4202610587321879e-05, + "loss": 0.963, + "step": 30943 + }, + { + "epoch": 2.29, + "learning_rate": 1.4202248646577917e-05, + "loss": 0.9644, + "step": 30944 + }, + { + "epoch": 2.29, + "learning_rate": 1.4201886699148239e-05, + "loss": 0.9434, + "step": 30945 + }, + { + "epoch": 2.29, + "learning_rate": 1.420152474503342e-05, + "loss": 1.0545, + "step": 30946 + }, + { + "epoch": 2.29, + "learning_rate": 1.4201162784234035e-05, + "loss": 0.9731, + "step": 30947 + }, + { + "epoch": 2.29, + "learning_rate": 1.420080081675066e-05, + "loss": 1.1437, + "step": 30948 + }, + { + "epoch": 2.29, + "learning_rate": 1.4200438842583872e-05, + "loss": 1.014, + "step": 30949 + }, + { + "epoch": 2.29, + "learning_rate": 1.4200076861734242e-05, + "loss": 0.9878, + "step": 30950 + }, + { + "epoch": 2.29, + "learning_rate": 1.4199714874202358e-05, + "loss": 0.9864, + "step": 30951 + }, + { + "epoch": 2.29, + "learning_rate": 1.4199352879988779e-05, + "loss": 1.073, + "step": 30952 + }, + { + "epoch": 2.29, + "learning_rate": 1.4198990879094095e-05, + "loss": 1.0296, + "step": 30953 + }, + { + "epoch": 2.29, + "learning_rate": 1.4198628871518874e-05, + "loss": 1.0847, + "step": 30954 + }, + { + "epoch": 2.29, + "learning_rate": 1.4198266857263695e-05, + "loss": 1.0602, + "step": 30955 + }, + { + "epoch": 2.29, + "learning_rate": 1.4197904836329134e-05, + "loss": 0.9971, + "step": 30956 + }, + { + "epoch": 2.29, + "learning_rate": 1.4197542808715766e-05, + "loss": 1.1003, + "step": 30957 + }, + { + "epoch": 2.29, + "learning_rate": 1.4197180774424167e-05, + "loss": 0.9019, + "step": 30958 + }, + { + "epoch": 2.29, + "learning_rate": 1.4196818733454916e-05, + "loss": 0.9592, + "step": 30959 + }, + { + "epoch": 2.29, + "learning_rate": 1.4196456685808584e-05, + "loss": 1.0189, + "step": 30960 + }, + { + "epoch": 2.29, + "learning_rate": 1.419609463148575e-05, + "loss": 1.026, + "step": 30961 + }, + { + "epoch": 2.29, + "learning_rate": 1.4195732570486991e-05, + "loss": 1.0653, + "step": 30962 + }, + { + "epoch": 2.29, + "learning_rate": 1.419537050281288e-05, + "loss": 0.9742, + "step": 30963 + }, + { + "epoch": 2.29, + "learning_rate": 1.4195008428463992e-05, + "loss": 0.9814, + "step": 30964 + }, + { + "epoch": 2.29, + "learning_rate": 1.419464634744091e-05, + "loss": 1.0469, + "step": 30965 + }, + { + "epoch": 2.29, + "learning_rate": 1.4194284259744202e-05, + "loss": 0.988, + "step": 30966 + }, + { + "epoch": 2.29, + "learning_rate": 1.4193922165374452e-05, + "loss": 0.9418, + "step": 30967 + }, + { + "epoch": 2.29, + "learning_rate": 1.4193560064332228e-05, + "loss": 0.972, + "step": 30968 + }, + { + "epoch": 2.29, + "learning_rate": 1.4193197956618113e-05, + "loss": 1.0534, + "step": 30969 + }, + { + "epoch": 2.29, + "learning_rate": 1.4192835842232677e-05, + "loss": 0.9944, + "step": 30970 + }, + { + "epoch": 2.29, + "learning_rate": 1.4192473721176502e-05, + "loss": 1.1168, + "step": 30971 + }, + { + "epoch": 2.29, + "learning_rate": 1.4192111593450159e-05, + "loss": 0.9943, + "step": 30972 + }, + { + "epoch": 2.29, + "learning_rate": 1.419174945905423e-05, + "loss": 1.0859, + "step": 30973 + }, + { + "epoch": 2.29, + "learning_rate": 1.4191387317989283e-05, + "loss": 1.0096, + "step": 30974 + }, + { + "epoch": 2.29, + "learning_rate": 1.4191025170255901e-05, + "loss": 1.1211, + "step": 30975 + }, + { + "epoch": 2.29, + "learning_rate": 1.4190663015854656e-05, + "loss": 0.9384, + "step": 30976 + }, + { + "epoch": 2.29, + "learning_rate": 1.4190300854786128e-05, + "loss": 0.9629, + "step": 30977 + }, + { + "epoch": 2.29, + "learning_rate": 1.418993868705089e-05, + "loss": 1.0621, + "step": 30978 + }, + { + "epoch": 2.29, + "learning_rate": 1.4189576512649522e-05, + "loss": 0.8486, + "step": 30979 + }, + { + "epoch": 2.29, + "learning_rate": 1.4189214331582597e-05, + "loss": 1.0186, + "step": 30980 + }, + { + "epoch": 2.29, + "learning_rate": 1.4188852143850688e-05, + "loss": 0.9157, + "step": 30981 + }, + { + "epoch": 2.29, + "learning_rate": 1.418848994945438e-05, + "loss": 1.1002, + "step": 30982 + }, + { + "epoch": 2.29, + "learning_rate": 1.4188127748394246e-05, + "loss": 1.0291, + "step": 30983 + }, + { + "epoch": 2.29, + "learning_rate": 1.4187765540670855e-05, + "loss": 1.0296, + "step": 30984 + }, + { + "epoch": 2.29, + "learning_rate": 1.4187403326284791e-05, + "loss": 0.9795, + "step": 30985 + }, + { + "epoch": 2.29, + "learning_rate": 1.4187041105236627e-05, + "loss": 1.0136, + "step": 30986 + }, + { + "epoch": 2.29, + "learning_rate": 1.4186678877526942e-05, + "loss": 1.1006, + "step": 30987 + }, + { + "epoch": 2.29, + "learning_rate": 1.4186316643156315e-05, + "loss": 0.9806, + "step": 30988 + }, + { + "epoch": 2.29, + "learning_rate": 1.4185954402125312e-05, + "loss": 0.9996, + "step": 30989 + }, + { + "epoch": 2.29, + "learning_rate": 1.4185592154434518e-05, + "loss": 1.0984, + "step": 30990 + }, + { + "epoch": 2.29, + "learning_rate": 1.4185229900084505e-05, + "loss": 0.9751, + "step": 30991 + }, + { + "epoch": 2.29, + "learning_rate": 1.4184867639075852e-05, + "loss": 0.991, + "step": 30992 + }, + { + "epoch": 2.29, + "learning_rate": 1.4184505371409134e-05, + "loss": 1.0597, + "step": 30993 + }, + { + "epoch": 2.29, + "learning_rate": 1.418414309708493e-05, + "loss": 1.0597, + "step": 30994 + }, + { + "epoch": 2.29, + "learning_rate": 1.4183780816103812e-05, + "loss": 1.0601, + "step": 30995 + }, + { + "epoch": 2.29, + "learning_rate": 1.418341852846636e-05, + "loss": 0.9404, + "step": 30996 + }, + { + "epoch": 2.29, + "learning_rate": 1.4183056234173147e-05, + "loss": 1.0123, + "step": 30997 + }, + { + "epoch": 2.29, + "learning_rate": 1.4182693933224756e-05, + "loss": 0.9649, + "step": 30998 + }, + { + "epoch": 2.29, + "learning_rate": 1.4182331625621753e-05, + "loss": 1.1099, + "step": 30999 + }, + { + "epoch": 2.29, + "learning_rate": 1.4181969311364725e-05, + "loss": 0.9754, + "step": 31000 + }, + { + "epoch": 2.29, + "learning_rate": 1.418160699045424e-05, + "loss": 1.0841, + "step": 31001 + }, + { + "epoch": 2.29, + "learning_rate": 1.418124466289088e-05, + "loss": 0.9612, + "step": 31002 + }, + { + "epoch": 2.29, + "learning_rate": 1.4180882328675219e-05, + "loss": 0.9344, + "step": 31003 + }, + { + "epoch": 2.29, + "learning_rate": 1.4180519987807832e-05, + "loss": 0.9738, + "step": 31004 + }, + { + "epoch": 2.29, + "learning_rate": 1.41801576402893e-05, + "loss": 1.0138, + "step": 31005 + }, + { + "epoch": 2.29, + "learning_rate": 1.4179795286120197e-05, + "loss": 1.031, + "step": 31006 + }, + { + "epoch": 2.29, + "learning_rate": 1.4179432925301102e-05, + "loss": 0.9491, + "step": 31007 + }, + { + "epoch": 2.29, + "learning_rate": 1.4179070557832586e-05, + "loss": 1.0169, + "step": 31008 + }, + { + "epoch": 2.29, + "learning_rate": 1.4178708183715228e-05, + "loss": 0.9658, + "step": 31009 + }, + { + "epoch": 2.29, + "learning_rate": 1.4178345802949606e-05, + "loss": 0.9656, + "step": 31010 + }, + { + "epoch": 2.29, + "learning_rate": 1.4177983415536294e-05, + "loss": 0.9478, + "step": 31011 + }, + { + "epoch": 2.29, + "learning_rate": 1.4177621021475872e-05, + "loss": 1.0354, + "step": 31012 + }, + { + "epoch": 2.29, + "learning_rate": 1.4177258620768913e-05, + "loss": 0.9957, + "step": 31013 + }, + { + "epoch": 2.29, + "learning_rate": 1.4176896213415998e-05, + "loss": 0.9102, + "step": 31014 + }, + { + "epoch": 2.29, + "learning_rate": 1.4176533799417697e-05, + "loss": 1.0684, + "step": 31015 + }, + { + "epoch": 2.29, + "learning_rate": 1.4176171378774596e-05, + "loss": 1.0259, + "step": 31016 + }, + { + "epoch": 2.29, + "learning_rate": 1.4175808951487262e-05, + "loss": 0.9585, + "step": 31017 + }, + { + "epoch": 2.29, + "learning_rate": 1.4175446517556278e-05, + "loss": 1.0326, + "step": 31018 + }, + { + "epoch": 2.29, + "learning_rate": 1.4175084076982217e-05, + "loss": 1.0026, + "step": 31019 + }, + { + "epoch": 2.29, + "learning_rate": 1.4174721629765658e-05, + "loss": 0.9452, + "step": 31020 + }, + { + "epoch": 2.29, + "learning_rate": 1.4174359175907173e-05, + "loss": 0.983, + "step": 31021 + }, + { + "epoch": 2.29, + "learning_rate": 1.4173996715407345e-05, + "loss": 0.9539, + "step": 31022 + }, + { + "epoch": 2.29, + "learning_rate": 1.4173634248266748e-05, + "loss": 1.0058, + "step": 31023 + }, + { + "epoch": 2.29, + "learning_rate": 1.4173271774485959e-05, + "loss": 1.0982, + "step": 31024 + }, + { + "epoch": 2.29, + "learning_rate": 1.4172909294065554e-05, + "loss": 0.9748, + "step": 31025 + }, + { + "epoch": 2.29, + "learning_rate": 1.4172546807006112e-05, + "loss": 1.0365, + "step": 31026 + }, + { + "epoch": 2.29, + "learning_rate": 1.4172184313308205e-05, + "loss": 1.0842, + "step": 31027 + }, + { + "epoch": 2.29, + "learning_rate": 1.4171821812972412e-05, + "loss": 0.9904, + "step": 31028 + }, + { + "epoch": 2.29, + "learning_rate": 1.4171459305999312e-05, + "loss": 1.0237, + "step": 31029 + }, + { + "epoch": 2.29, + "learning_rate": 1.4171096792389478e-05, + "loss": 1.0086, + "step": 31030 + }, + { + "epoch": 2.29, + "learning_rate": 1.417073427214349e-05, + "loss": 1.0471, + "step": 31031 + }, + { + "epoch": 2.29, + "learning_rate": 1.4170371745261923e-05, + "loss": 1.0596, + "step": 31032 + }, + { + "epoch": 2.29, + "learning_rate": 1.4170009211745353e-05, + "loss": 0.9799, + "step": 31033 + }, + { + "epoch": 2.29, + "learning_rate": 1.4169646671594362e-05, + "loss": 1.0251, + "step": 31034 + }, + { + "epoch": 2.29, + "learning_rate": 1.416928412480952e-05, + "loss": 1.023, + "step": 31035 + }, + { + "epoch": 2.29, + "learning_rate": 1.4168921571391406e-05, + "loss": 1.0174, + "step": 31036 + }, + { + "epoch": 2.29, + "learning_rate": 1.4168559011340599e-05, + "loss": 1.0517, + "step": 31037 + }, + { + "epoch": 2.29, + "learning_rate": 1.416819644465767e-05, + "loss": 0.9548, + "step": 31038 + }, + { + "epoch": 2.29, + "learning_rate": 1.4167833871343207e-05, + "loss": 1.0464, + "step": 31039 + }, + { + "epoch": 2.29, + "learning_rate": 1.4167471291397774e-05, + "loss": 1.1213, + "step": 31040 + }, + { + "epoch": 2.29, + "learning_rate": 1.4167108704821959e-05, + "loss": 0.9733, + "step": 31041 + }, + { + "epoch": 2.29, + "learning_rate": 1.416674611161633e-05, + "loss": 0.8814, + "step": 31042 + }, + { + "epoch": 2.29, + "learning_rate": 1.416638351178147e-05, + "loss": 0.9496, + "step": 31043 + }, + { + "epoch": 2.29, + "learning_rate": 1.4166020905317953e-05, + "loss": 1.0069, + "step": 31044 + }, + { + "epoch": 2.29, + "learning_rate": 1.4165658292226354e-05, + "loss": 1.0463, + "step": 31045 + }, + { + "epoch": 2.29, + "learning_rate": 1.4165295672507253e-05, + "loss": 1.0653, + "step": 31046 + }, + { + "epoch": 2.29, + "learning_rate": 1.4164933046161226e-05, + "loss": 1.056, + "step": 31047 + }, + { + "epoch": 2.29, + "learning_rate": 1.4164570413188851e-05, + "loss": 1.0435, + "step": 31048 + }, + { + "epoch": 2.29, + "learning_rate": 1.4164207773590706e-05, + "loss": 0.8519, + "step": 31049 + }, + { + "epoch": 2.29, + "learning_rate": 1.4163845127367361e-05, + "loss": 0.9205, + "step": 31050 + }, + { + "epoch": 2.29, + "learning_rate": 1.4163482474519404e-05, + "loss": 0.9683, + "step": 31051 + }, + { + "epoch": 2.29, + "learning_rate": 1.4163119815047401e-05, + "loss": 1.0489, + "step": 31052 + }, + { + "epoch": 2.29, + "learning_rate": 1.4162757148951935e-05, + "loss": 0.9251, + "step": 31053 + }, + { + "epoch": 2.29, + "learning_rate": 1.4162394476233585e-05, + "loss": 1.0842, + "step": 31054 + }, + { + "epoch": 2.29, + "learning_rate": 1.4162031796892922e-05, + "loss": 1.0426, + "step": 31055 + }, + { + "epoch": 2.29, + "learning_rate": 1.4161669110930527e-05, + "loss": 0.9528, + "step": 31056 + }, + { + "epoch": 2.29, + "learning_rate": 1.4161306418346978e-05, + "loss": 0.9416, + "step": 31057 + }, + { + "epoch": 2.29, + "learning_rate": 1.4160943719142845e-05, + "loss": 1.0743, + "step": 31058 + }, + { + "epoch": 2.29, + "learning_rate": 1.4160581013318715e-05, + "loss": 1.0268, + "step": 31059 + }, + { + "epoch": 2.3, + "learning_rate": 1.4160218300875156e-05, + "loss": 1.0357, + "step": 31060 + }, + { + "epoch": 2.3, + "learning_rate": 1.4159855581812752e-05, + "loss": 1.0762, + "step": 31061 + }, + { + "epoch": 2.3, + "learning_rate": 1.4159492856132075e-05, + "loss": 1.0757, + "step": 31062 + }, + { + "epoch": 2.3, + "learning_rate": 1.4159130123833708e-05, + "loss": 1.0131, + "step": 31063 + }, + { + "epoch": 2.3, + "learning_rate": 1.4158767384918221e-05, + "loss": 1.0548, + "step": 31064 + }, + { + "epoch": 2.3, + "learning_rate": 1.4158404639386199e-05, + "loss": 1.0849, + "step": 31065 + }, + { + "epoch": 2.3, + "learning_rate": 1.415804188723821e-05, + "loss": 1.1121, + "step": 31066 + }, + { + "epoch": 2.3, + "learning_rate": 1.4157679128474838e-05, + "loss": 1.0204, + "step": 31067 + }, + { + "epoch": 2.3, + "learning_rate": 1.4157316363096656e-05, + "loss": 0.8503, + "step": 31068 + }, + { + "epoch": 2.3, + "learning_rate": 1.4156953591104246e-05, + "loss": 1.0217, + "step": 31069 + }, + { + "epoch": 2.3, + "learning_rate": 1.415659081249818e-05, + "loss": 0.9434, + "step": 31070 + }, + { + "epoch": 2.3, + "learning_rate": 1.4156228027279038e-05, + "loss": 1.0013, + "step": 31071 + }, + { + "epoch": 2.3, + "learning_rate": 1.4155865235447397e-05, + "loss": 1.0157, + "step": 31072 + }, + { + "epoch": 2.3, + "learning_rate": 1.4155502437003836e-05, + "loss": 0.9713, + "step": 31073 + }, + { + "epoch": 2.3, + "learning_rate": 1.4155139631948926e-05, + "loss": 0.999, + "step": 31074 + }, + { + "epoch": 2.3, + "learning_rate": 1.4154776820283252e-05, + "loss": 0.9888, + "step": 31075 + }, + { + "epoch": 2.3, + "learning_rate": 1.4154414002007385e-05, + "loss": 0.998, + "step": 31076 + }, + { + "epoch": 2.3, + "learning_rate": 1.4154051177121906e-05, + "loss": 0.9708, + "step": 31077 + }, + { + "epoch": 2.3, + "learning_rate": 1.415368834562739e-05, + "loss": 1.0569, + "step": 31078 + }, + { + "epoch": 2.3, + "learning_rate": 1.4153325507524416e-05, + "loss": 1.0183, + "step": 31079 + }, + { + "epoch": 2.3, + "learning_rate": 1.4152962662813558e-05, + "loss": 1.0809, + "step": 31080 + }, + { + "epoch": 2.3, + "learning_rate": 1.4152599811495402e-05, + "loss": 1.0776, + "step": 31081 + }, + { + "epoch": 2.3, + "learning_rate": 1.4152236953570515e-05, + "loss": 1.0935, + "step": 31082 + }, + { + "epoch": 2.3, + "learning_rate": 1.4151874089039476e-05, + "loss": 1.0772, + "step": 31083 + }, + { + "epoch": 2.3, + "learning_rate": 1.4151511217902871e-05, + "loss": 1.0384, + "step": 31084 + }, + { + "epoch": 2.3, + "learning_rate": 1.4151148340161264e-05, + "loss": 1.0249, + "step": 31085 + }, + { + "epoch": 2.3, + "learning_rate": 1.4150785455815245e-05, + "loss": 1.0299, + "step": 31086 + }, + { + "epoch": 2.3, + "learning_rate": 1.4150422564865384e-05, + "loss": 1.0631, + "step": 31087 + }, + { + "epoch": 2.3, + "learning_rate": 1.415005966731226e-05, + "loss": 0.9507, + "step": 31088 + }, + { + "epoch": 2.3, + "learning_rate": 1.414969676315645e-05, + "loss": 0.899, + "step": 31089 + }, + { + "epoch": 2.3, + "learning_rate": 1.414933385239853e-05, + "loss": 0.8692, + "step": 31090 + }, + { + "epoch": 2.3, + "learning_rate": 1.4148970935039083e-05, + "loss": 0.979, + "step": 31091 + }, + { + "epoch": 2.3, + "learning_rate": 1.4148608011078682e-05, + "loss": 0.9594, + "step": 31092 + }, + { + "epoch": 2.3, + "learning_rate": 1.4148245080517904e-05, + "loss": 0.9281, + "step": 31093 + }, + { + "epoch": 2.3, + "learning_rate": 1.4147882143357327e-05, + "loss": 1.1267, + "step": 31094 + }, + { + "epoch": 2.3, + "learning_rate": 1.4147519199597529e-05, + "loss": 1.1085, + "step": 31095 + }, + { + "epoch": 2.3, + "learning_rate": 1.4147156249239088e-05, + "loss": 0.99, + "step": 31096 + }, + { + "epoch": 2.3, + "learning_rate": 1.4146793292282578e-05, + "loss": 0.9422, + "step": 31097 + }, + { + "epoch": 2.3, + "learning_rate": 1.4146430328728582e-05, + "loss": 0.9505, + "step": 31098 + }, + { + "epoch": 2.3, + "learning_rate": 1.4146067358577673e-05, + "loss": 1.0545, + "step": 31099 + }, + { + "epoch": 2.3, + "learning_rate": 1.4145704381830432e-05, + "loss": 1.112, + "step": 31100 + }, + { + "epoch": 2.3, + "learning_rate": 1.4145341398487435e-05, + "loss": 0.9702, + "step": 31101 + }, + { + "epoch": 2.3, + "learning_rate": 1.414497840854926e-05, + "loss": 0.9148, + "step": 31102 + }, + { + "epoch": 2.3, + "learning_rate": 1.414461541201648e-05, + "loss": 0.9719, + "step": 31103 + }, + { + "epoch": 2.3, + "learning_rate": 1.4144252408889678e-05, + "loss": 0.9036, + "step": 31104 + }, + { + "epoch": 2.3, + "learning_rate": 1.4143889399169428e-05, + "loss": 0.9273, + "step": 31105 + }, + { + "epoch": 2.3, + "learning_rate": 1.4143526382856313e-05, + "loss": 1.057, + "step": 31106 + }, + { + "epoch": 2.3, + "learning_rate": 1.4143163359950902e-05, + "loss": 1.0073, + "step": 31107 + }, + { + "epoch": 2.3, + "learning_rate": 1.4142800330453781e-05, + "loss": 0.9748, + "step": 31108 + }, + { + "epoch": 2.3, + "learning_rate": 1.4142437294365521e-05, + "loss": 0.9872, + "step": 31109 + }, + { + "epoch": 2.3, + "learning_rate": 1.4142074251686708e-05, + "loss": 0.948, + "step": 31110 + }, + { + "epoch": 2.3, + "learning_rate": 1.4141711202417909e-05, + "loss": 1.055, + "step": 31111 + }, + { + "epoch": 2.3, + "learning_rate": 1.414134814655971e-05, + "loss": 1.0053, + "step": 31112 + }, + { + "epoch": 2.3, + "learning_rate": 1.4140985084112684e-05, + "loss": 1.1216, + "step": 31113 + }, + { + "epoch": 2.3, + "learning_rate": 1.414062201507741e-05, + "loss": 0.989, + "step": 31114 + }, + { + "epoch": 2.3, + "learning_rate": 1.4140258939454463e-05, + "loss": 1.0231, + "step": 31115 + }, + { + "epoch": 2.3, + "learning_rate": 1.4139895857244427e-05, + "loss": 1.0261, + "step": 31116 + }, + { + "epoch": 2.3, + "learning_rate": 1.4139532768447873e-05, + "loss": 0.9297, + "step": 31117 + }, + { + "epoch": 2.3, + "learning_rate": 1.4139169673065386e-05, + "loss": 0.9614, + "step": 31118 + }, + { + "epoch": 2.3, + "learning_rate": 1.4138806571097534e-05, + "loss": 1.0494, + "step": 31119 + }, + { + "epoch": 2.3, + "learning_rate": 1.4138443462544903e-05, + "loss": 1.0385, + "step": 31120 + }, + { + "epoch": 2.3, + "learning_rate": 1.4138080347408068e-05, + "loss": 1.0258, + "step": 31121 + }, + { + "epoch": 2.3, + "learning_rate": 1.4137717225687606e-05, + "loss": 0.9824, + "step": 31122 + }, + { + "epoch": 2.3, + "learning_rate": 1.4137354097384093e-05, + "loss": 0.9593, + "step": 31123 + }, + { + "epoch": 2.3, + "learning_rate": 1.4136990962498112e-05, + "loss": 1.0414, + "step": 31124 + }, + { + "epoch": 2.3, + "learning_rate": 1.4136627821030233e-05, + "loss": 1.0427, + "step": 31125 + }, + { + "epoch": 2.3, + "learning_rate": 1.4136264672981041e-05, + "loss": 1.0124, + "step": 31126 + }, + { + "epoch": 2.3, + "learning_rate": 1.4135901518351112e-05, + "loss": 1.0788, + "step": 31127 + }, + { + "epoch": 2.3, + "learning_rate": 1.4135538357141023e-05, + "loss": 0.8976, + "step": 31128 + }, + { + "epoch": 2.3, + "learning_rate": 1.413517518935135e-05, + "loss": 0.9174, + "step": 31129 + }, + { + "epoch": 2.3, + "learning_rate": 1.4134812014982673e-05, + "loss": 0.96, + "step": 31130 + }, + { + "epoch": 2.3, + "learning_rate": 1.413444883403557e-05, + "loss": 1.1586, + "step": 31131 + }, + { + "epoch": 2.3, + "learning_rate": 1.4134085646510617e-05, + "loss": 0.9474, + "step": 31132 + }, + { + "epoch": 2.3, + "learning_rate": 1.4133722452408393e-05, + "loss": 0.955, + "step": 31133 + }, + { + "epoch": 2.3, + "learning_rate": 1.4133359251729476e-05, + "loss": 1.0999, + "step": 31134 + }, + { + "epoch": 2.3, + "learning_rate": 1.4132996044474443e-05, + "loss": 0.886, + "step": 31135 + }, + { + "epoch": 2.3, + "learning_rate": 1.4132632830643876e-05, + "loss": 1.0389, + "step": 31136 + }, + { + "epoch": 2.3, + "learning_rate": 1.4132269610238343e-05, + "loss": 0.9356, + "step": 31137 + }, + { + "epoch": 2.3, + "learning_rate": 1.4131906383258433e-05, + "loss": 0.9888, + "step": 31138 + }, + { + "epoch": 2.3, + "learning_rate": 1.4131543149704718e-05, + "loss": 1.027, + "step": 31139 + }, + { + "epoch": 2.3, + "learning_rate": 1.4131179909577777e-05, + "loss": 1.0399, + "step": 31140 + }, + { + "epoch": 2.3, + "learning_rate": 1.4130816662878188e-05, + "loss": 0.9682, + "step": 31141 + }, + { + "epoch": 2.3, + "learning_rate": 1.4130453409606527e-05, + "loss": 0.8705, + "step": 31142 + }, + { + "epoch": 2.3, + "learning_rate": 1.4130090149763375e-05, + "loss": 1.0372, + "step": 31143 + }, + { + "epoch": 2.3, + "learning_rate": 1.4129726883349308e-05, + "loss": 0.9437, + "step": 31144 + }, + { + "epoch": 2.3, + "learning_rate": 1.4129363610364906e-05, + "loss": 1.0323, + "step": 31145 + }, + { + "epoch": 2.3, + "learning_rate": 1.4129000330810743e-05, + "loss": 1.1044, + "step": 31146 + }, + { + "epoch": 2.3, + "learning_rate": 1.4128637044687405e-05, + "loss": 1.0266, + "step": 31147 + }, + { + "epoch": 2.3, + "learning_rate": 1.4128273751995459e-05, + "loss": 1.0429, + "step": 31148 + }, + { + "epoch": 2.3, + "learning_rate": 1.4127910452735492e-05, + "loss": 0.9395, + "step": 31149 + }, + { + "epoch": 2.3, + "learning_rate": 1.4127547146908076e-05, + "loss": 0.98, + "step": 31150 + }, + { + "epoch": 2.3, + "learning_rate": 1.4127183834513791e-05, + "loss": 1.009, + "step": 31151 + }, + { + "epoch": 2.3, + "learning_rate": 1.4126820515553217e-05, + "loss": 1.0252, + "step": 31152 + }, + { + "epoch": 2.3, + "learning_rate": 1.4126457190026932e-05, + "loss": 1.0051, + "step": 31153 + }, + { + "epoch": 2.3, + "learning_rate": 1.412609385793551e-05, + "loss": 0.8877, + "step": 31154 + }, + { + "epoch": 2.3, + "learning_rate": 1.4125730519279535e-05, + "loss": 1.0767, + "step": 31155 + }, + { + "epoch": 2.3, + "learning_rate": 1.4125367174059577e-05, + "loss": 0.9679, + "step": 31156 + }, + { + "epoch": 2.3, + "learning_rate": 1.4125003822276227e-05, + "loss": 1.0423, + "step": 31157 + }, + { + "epoch": 2.3, + "learning_rate": 1.4124640463930047e-05, + "loss": 1.0963, + "step": 31158 + }, + { + "epoch": 2.3, + "learning_rate": 1.4124277099021626e-05, + "loss": 0.9252, + "step": 31159 + }, + { + "epoch": 2.3, + "learning_rate": 1.4123913727551538e-05, + "loss": 1.0811, + "step": 31160 + }, + { + "epoch": 2.3, + "learning_rate": 1.4123550349520363e-05, + "loss": 1.0024, + "step": 31161 + }, + { + "epoch": 2.3, + "learning_rate": 1.4123186964928679e-05, + "loss": 1.0052, + "step": 31162 + }, + { + "epoch": 2.3, + "learning_rate": 1.4122823573777062e-05, + "loss": 1.0136, + "step": 31163 + }, + { + "epoch": 2.3, + "learning_rate": 1.4122460176066093e-05, + "loss": 1.0367, + "step": 31164 + }, + { + "epoch": 2.3, + "learning_rate": 1.4122096771796347e-05, + "loss": 0.9516, + "step": 31165 + }, + { + "epoch": 2.3, + "learning_rate": 1.4121733360968407e-05, + "loss": 0.9562, + "step": 31166 + }, + { + "epoch": 2.3, + "learning_rate": 1.4121369943582848e-05, + "loss": 1.0221, + "step": 31167 + }, + { + "epoch": 2.3, + "learning_rate": 1.4121006519640246e-05, + "loss": 0.955, + "step": 31168 + }, + { + "epoch": 2.3, + "learning_rate": 1.4120643089141184e-05, + "loss": 1.0387, + "step": 31169 + }, + { + "epoch": 2.3, + "learning_rate": 1.4120279652086234e-05, + "loss": 0.9607, + "step": 31170 + }, + { + "epoch": 2.3, + "learning_rate": 1.4119916208475983e-05, + "loss": 1.0424, + "step": 31171 + }, + { + "epoch": 2.3, + "learning_rate": 1.4119552758311e-05, + "loss": 1.1255, + "step": 31172 + }, + { + "epoch": 2.3, + "learning_rate": 1.411918930159187e-05, + "loss": 1.0483, + "step": 31173 + }, + { + "epoch": 2.3, + "learning_rate": 1.4118825838319166e-05, + "loss": 1.0, + "step": 31174 + }, + { + "epoch": 2.3, + "learning_rate": 1.4118462368493476e-05, + "loss": 1.1291, + "step": 31175 + }, + { + "epoch": 2.3, + "learning_rate": 1.4118098892115364e-05, + "loss": 1.0733, + "step": 31176 + }, + { + "epoch": 2.3, + "learning_rate": 1.4117735409185418e-05, + "loss": 0.963, + "step": 31177 + }, + { + "epoch": 2.3, + "learning_rate": 1.4117371919704212e-05, + "loss": 0.995, + "step": 31178 + }, + { + "epoch": 2.3, + "learning_rate": 1.4117008423672328e-05, + "loss": 1.0498, + "step": 31179 + }, + { + "epoch": 2.3, + "learning_rate": 1.4116644921090344e-05, + "loss": 1.0589, + "step": 31180 + }, + { + "epoch": 2.3, + "learning_rate": 1.4116281411958835e-05, + "loss": 1.0279, + "step": 31181 + }, + { + "epoch": 2.3, + "learning_rate": 1.4115917896278379e-05, + "loss": 0.9985, + "step": 31182 + }, + { + "epoch": 2.3, + "learning_rate": 1.4115554374049563e-05, + "loss": 1.0287, + "step": 31183 + }, + { + "epoch": 2.3, + "learning_rate": 1.4115190845272953e-05, + "loss": 1.0337, + "step": 31184 + }, + { + "epoch": 2.3, + "learning_rate": 1.4114827309949135e-05, + "loss": 0.9612, + "step": 31185 + }, + { + "epoch": 2.3, + "learning_rate": 1.4114463768078688e-05, + "loss": 1.0137, + "step": 31186 + }, + { + "epoch": 2.3, + "learning_rate": 1.4114100219662183e-05, + "loss": 0.9035, + "step": 31187 + }, + { + "epoch": 2.3, + "learning_rate": 1.4113736664700208e-05, + "loss": 0.9233, + "step": 31188 + }, + { + "epoch": 2.3, + "learning_rate": 1.4113373103193333e-05, + "loss": 1.0828, + "step": 31189 + }, + { + "epoch": 2.3, + "learning_rate": 1.4113009535142143e-05, + "loss": 0.9866, + "step": 31190 + }, + { + "epoch": 2.3, + "learning_rate": 1.4112645960547213e-05, + "loss": 1.0569, + "step": 31191 + }, + { + "epoch": 2.3, + "learning_rate": 1.4112282379409121e-05, + "loss": 0.9426, + "step": 31192 + }, + { + "epoch": 2.3, + "learning_rate": 1.4111918791728449e-05, + "loss": 0.9634, + "step": 31193 + }, + { + "epoch": 2.3, + "learning_rate": 1.4111555197505773e-05, + "loss": 0.9934, + "step": 31194 + }, + { + "epoch": 2.3, + "learning_rate": 1.4111191596741669e-05, + "loss": 0.9822, + "step": 31195 + }, + { + "epoch": 2.31, + "learning_rate": 1.411082798943672e-05, + "loss": 0.8654, + "step": 31196 + }, + { + "epoch": 2.31, + "learning_rate": 1.41104643755915e-05, + "loss": 1.0219, + "step": 31197 + }, + { + "epoch": 2.31, + "learning_rate": 1.4110100755206592e-05, + "loss": 0.9578, + "step": 31198 + }, + { + "epoch": 2.31, + "learning_rate": 1.4109737128282572e-05, + "loss": 0.9439, + "step": 31199 + }, + { + "epoch": 2.31, + "learning_rate": 1.410937349482002e-05, + "loss": 0.9535, + "step": 31200 + }, + { + "epoch": 2.31, + "learning_rate": 1.410900985481951e-05, + "loss": 0.997, + "step": 31201 + }, + { + "epoch": 2.31, + "learning_rate": 1.4108646208281629e-05, + "loss": 0.9935, + "step": 31202 + }, + { + "epoch": 2.31, + "learning_rate": 1.4108282555206948e-05, + "loss": 0.9795, + "step": 31203 + }, + { + "epoch": 2.31, + "learning_rate": 1.4107918895596052e-05, + "loss": 1.0456, + "step": 31204 + }, + { + "epoch": 2.31, + "learning_rate": 1.410755522944951e-05, + "loss": 1.0274, + "step": 31205 + }, + { + "epoch": 2.31, + "learning_rate": 1.410719155676791e-05, + "loss": 1.0362, + "step": 31206 + }, + { + "epoch": 2.31, + "learning_rate": 1.4106827877551825e-05, + "loss": 0.8939, + "step": 31207 + }, + { + "epoch": 2.31, + "learning_rate": 1.4106464191801837e-05, + "loss": 1.0142, + "step": 31208 + }, + { + "epoch": 2.31, + "learning_rate": 1.4106100499518523e-05, + "loss": 0.9603, + "step": 31209 + }, + { + "epoch": 2.31, + "learning_rate": 1.410573680070246e-05, + "loss": 1.0898, + "step": 31210 + }, + { + "epoch": 2.31, + "learning_rate": 1.410537309535423e-05, + "loss": 1.1097, + "step": 31211 + }, + { + "epoch": 2.31, + "learning_rate": 1.410500938347441e-05, + "loss": 0.9948, + "step": 31212 + }, + { + "epoch": 2.31, + "learning_rate": 1.410464566506358e-05, + "loss": 1.0276, + "step": 31213 + }, + { + "epoch": 2.31, + "learning_rate": 1.4104281940122318e-05, + "loss": 0.9826, + "step": 31214 + }, + { + "epoch": 2.31, + "learning_rate": 1.41039182086512e-05, + "loss": 0.9942, + "step": 31215 + }, + { + "epoch": 2.31, + "learning_rate": 1.4103554470650809e-05, + "loss": 0.9878, + "step": 31216 + }, + { + "epoch": 2.31, + "learning_rate": 1.4103190726121718e-05, + "loss": 0.9629, + "step": 31217 + }, + { + "epoch": 2.31, + "learning_rate": 1.4102826975064513e-05, + "loss": 1.0381, + "step": 31218 + }, + { + "epoch": 2.31, + "learning_rate": 1.4102463217479764e-05, + "loss": 1.0513, + "step": 31219 + }, + { + "epoch": 2.31, + "learning_rate": 1.4102099453368058e-05, + "loss": 1.0351, + "step": 31220 + }, + { + "epoch": 2.31, + "learning_rate": 1.4101735682729972e-05, + "loss": 0.992, + "step": 31221 + }, + { + "epoch": 2.31, + "learning_rate": 1.4101371905566084e-05, + "loss": 0.9596, + "step": 31222 + }, + { + "epoch": 2.31, + "learning_rate": 1.4101008121876968e-05, + "loss": 0.9364, + "step": 31223 + }, + { + "epoch": 2.31, + "learning_rate": 1.4100644331663208e-05, + "loss": 1.0255, + "step": 31224 + }, + { + "epoch": 2.31, + "learning_rate": 1.4100280534925384e-05, + "loss": 1.0808, + "step": 31225 + }, + { + "epoch": 2.31, + "learning_rate": 1.4099916731664069e-05, + "loss": 0.9187, + "step": 31226 + }, + { + "epoch": 2.31, + "learning_rate": 1.4099552921879846e-05, + "loss": 0.9573, + "step": 31227 + }, + { + "epoch": 2.31, + "learning_rate": 1.4099189105573294e-05, + "loss": 0.9229, + "step": 31228 + }, + { + "epoch": 2.31, + "learning_rate": 1.409882528274499e-05, + "loss": 0.8889, + "step": 31229 + }, + { + "epoch": 2.31, + "learning_rate": 1.4098461453395514e-05, + "loss": 0.9339, + "step": 31230 + }, + { + "epoch": 2.31, + "learning_rate": 1.4098097617525447e-05, + "loss": 1.0254, + "step": 31231 + }, + { + "epoch": 2.31, + "learning_rate": 1.409773377513536e-05, + "loss": 0.9329, + "step": 31232 + }, + { + "epoch": 2.31, + "learning_rate": 1.4097369926225839e-05, + "loss": 1.0617, + "step": 31233 + }, + { + "epoch": 2.31, + "learning_rate": 1.409700607079746e-05, + "loss": 0.9303, + "step": 31234 + }, + { + "epoch": 2.31, + "learning_rate": 1.4096642208850808e-05, + "loss": 0.9009, + "step": 31235 + }, + { + "epoch": 2.31, + "learning_rate": 1.4096278340386454e-05, + "loss": 0.9962, + "step": 31236 + }, + { + "epoch": 2.31, + "learning_rate": 1.4095914465404979e-05, + "loss": 0.9184, + "step": 31237 + }, + { + "epoch": 2.31, + "learning_rate": 1.4095550583906962e-05, + "loss": 1.1564, + "step": 31238 + }, + { + "epoch": 2.31, + "learning_rate": 1.4095186695892984e-05, + "loss": 1.0449, + "step": 31239 + }, + { + "epoch": 2.31, + "learning_rate": 1.4094822801363624e-05, + "loss": 1.0049, + "step": 31240 + }, + { + "epoch": 2.31, + "learning_rate": 1.4094458900319458e-05, + "loss": 1.0548, + "step": 31241 + }, + { + "epoch": 2.31, + "learning_rate": 1.4094094992761064e-05, + "loss": 0.9051, + "step": 31242 + }, + { + "epoch": 2.31, + "learning_rate": 1.4093731078689029e-05, + "loss": 0.9693, + "step": 31243 + }, + { + "epoch": 2.31, + "learning_rate": 1.4093367158103922e-05, + "loss": 0.8629, + "step": 31244 + }, + { + "epoch": 2.31, + "learning_rate": 1.4093003231006327e-05, + "loss": 0.9897, + "step": 31245 + }, + { + "epoch": 2.31, + "learning_rate": 1.4092639297396821e-05, + "loss": 1.0067, + "step": 31246 + }, + { + "epoch": 2.31, + "learning_rate": 1.4092275357275989e-05, + "loss": 0.9887, + "step": 31247 + }, + { + "epoch": 2.31, + "learning_rate": 1.4091911410644402e-05, + "loss": 1.0478, + "step": 31248 + }, + { + "epoch": 2.31, + "learning_rate": 1.4091547457502643e-05, + "loss": 0.9644, + "step": 31249 + }, + { + "epoch": 2.31, + "learning_rate": 1.4091183497851292e-05, + "loss": 0.9566, + "step": 31250 + }, + { + "epoch": 2.31, + "learning_rate": 1.4090819531690927e-05, + "loss": 1.0415, + "step": 31251 + }, + { + "epoch": 2.31, + "learning_rate": 1.4090455559022123e-05, + "loss": 0.9658, + "step": 31252 + }, + { + "epoch": 2.31, + "learning_rate": 1.4090091579845467e-05, + "loss": 1.0174, + "step": 31253 + }, + { + "epoch": 2.31, + "learning_rate": 1.408972759416153e-05, + "loss": 0.9712, + "step": 31254 + }, + { + "epoch": 2.31, + "learning_rate": 1.4089363601970898e-05, + "loss": 0.8953, + "step": 31255 + }, + { + "epoch": 2.31, + "learning_rate": 1.4088999603274145e-05, + "loss": 1.0058, + "step": 31256 + }, + { + "epoch": 2.31, + "learning_rate": 1.4088635598071854e-05, + "loss": 1.0287, + "step": 31257 + }, + { + "epoch": 2.31, + "learning_rate": 1.40882715863646e-05, + "loss": 0.997, + "step": 31258 + }, + { + "epoch": 2.31, + "learning_rate": 1.4087907568152965e-05, + "loss": 1.0184, + "step": 31259 + }, + { + "epoch": 2.31, + "learning_rate": 1.4087543543437528e-05, + "loss": 1.0052, + "step": 31260 + }, + { + "epoch": 2.31, + "learning_rate": 1.4087179512218869e-05, + "loss": 1.0209, + "step": 31261 + }, + { + "epoch": 2.31, + "learning_rate": 1.4086815474497564e-05, + "loss": 1.014, + "step": 31262 + }, + { + "epoch": 2.31, + "learning_rate": 1.4086451430274195e-05, + "loss": 1.0251, + "step": 31263 + }, + { + "epoch": 2.31, + "learning_rate": 1.4086087379549338e-05, + "loss": 1.0458, + "step": 31264 + }, + { + "epoch": 2.31, + "learning_rate": 1.4085723322323576e-05, + "loss": 1.0171, + "step": 31265 + }, + { + "epoch": 2.31, + "learning_rate": 1.4085359258597487e-05, + "loss": 1.0158, + "step": 31266 + }, + { + "epoch": 2.31, + "learning_rate": 1.4084995188371652e-05, + "loss": 0.9541, + "step": 31267 + }, + { + "epoch": 2.31, + "learning_rate": 1.4084631111646646e-05, + "loss": 1.0898, + "step": 31268 + }, + { + "epoch": 2.31, + "learning_rate": 1.408426702842305e-05, + "loss": 1.0411, + "step": 31269 + }, + { + "epoch": 2.31, + "learning_rate": 1.4083902938701444e-05, + "loss": 1.1159, + "step": 31270 + }, + { + "epoch": 2.31, + "learning_rate": 1.4083538842482407e-05, + "loss": 1.0045, + "step": 31271 + }, + { + "epoch": 2.31, + "learning_rate": 1.4083174739766516e-05, + "loss": 1.1038, + "step": 31272 + }, + { + "epoch": 2.31, + "learning_rate": 1.4082810630554358e-05, + "loss": 1.1371, + "step": 31273 + }, + { + "epoch": 2.31, + "learning_rate": 1.40824465148465e-05, + "loss": 1.0955, + "step": 31274 + }, + { + "epoch": 2.31, + "learning_rate": 1.4082082392643533e-05, + "loss": 0.8896, + "step": 31275 + }, + { + "epoch": 2.31, + "learning_rate": 1.4081718263946027e-05, + "loss": 1.0148, + "step": 31276 + }, + { + "epoch": 2.31, + "learning_rate": 1.408135412875457e-05, + "loss": 0.956, + "step": 31277 + }, + { + "epoch": 2.31, + "learning_rate": 1.4080989987069736e-05, + "loss": 1.0192, + "step": 31278 + }, + { + "epoch": 2.31, + "learning_rate": 1.4080625838892102e-05, + "loss": 1.0442, + "step": 31279 + }, + { + "epoch": 2.31, + "learning_rate": 1.4080261684222254e-05, + "loss": 1.0495, + "step": 31280 + }, + { + "epoch": 2.31, + "learning_rate": 1.4079897523060769e-05, + "loss": 1.0603, + "step": 31281 + }, + { + "epoch": 2.31, + "learning_rate": 1.407953335540822e-05, + "loss": 0.9335, + "step": 31282 + }, + { + "epoch": 2.31, + "learning_rate": 1.4079169181265196e-05, + "loss": 0.9278, + "step": 31283 + }, + { + "epoch": 2.31, + "learning_rate": 1.407880500063227e-05, + "loss": 1.0595, + "step": 31284 + }, + { + "epoch": 2.31, + "learning_rate": 1.4078440813510026e-05, + "loss": 0.9895, + "step": 31285 + }, + { + "epoch": 2.31, + "learning_rate": 1.4078076619899039e-05, + "loss": 0.8432, + "step": 31286 + }, + { + "epoch": 2.31, + "learning_rate": 1.4077712419799893e-05, + "loss": 0.8965, + "step": 31287 + }, + { + "epoch": 2.31, + "learning_rate": 1.4077348213213164e-05, + "loss": 1.0564, + "step": 31288 + }, + { + "epoch": 2.31, + "learning_rate": 1.407698400013943e-05, + "loss": 1.1091, + "step": 31289 + }, + { + "epoch": 2.31, + "learning_rate": 1.4076619780579274e-05, + "loss": 0.9782, + "step": 31290 + }, + { + "epoch": 2.31, + "learning_rate": 1.4076255554533276e-05, + "loss": 1.0586, + "step": 31291 + }, + { + "epoch": 2.31, + "learning_rate": 1.4075891322002013e-05, + "loss": 1.0172, + "step": 31292 + }, + { + "epoch": 2.31, + "learning_rate": 1.4075527082986062e-05, + "loss": 0.9838, + "step": 31293 + }, + { + "epoch": 2.31, + "learning_rate": 1.407516283748601e-05, + "loss": 0.9952, + "step": 31294 + }, + { + "epoch": 2.31, + "learning_rate": 1.407479858550243e-05, + "loss": 0.9325, + "step": 31295 + }, + { + "epoch": 2.31, + "learning_rate": 1.4074434327035904e-05, + "loss": 0.8885, + "step": 31296 + }, + { + "epoch": 2.31, + "learning_rate": 1.4074070062087013e-05, + "loss": 0.9704, + "step": 31297 + }, + { + "epoch": 2.31, + "learning_rate": 1.4073705790656336e-05, + "loss": 1.0294, + "step": 31298 + }, + { + "epoch": 2.31, + "learning_rate": 1.4073341512744447e-05, + "loss": 0.9843, + "step": 31299 + }, + { + "epoch": 2.31, + "learning_rate": 1.4072977228351932e-05, + "loss": 0.985, + "step": 31300 + }, + { + "epoch": 2.31, + "learning_rate": 1.4072612937479367e-05, + "loss": 0.9468, + "step": 31301 + }, + { + "epoch": 2.31, + "learning_rate": 1.4072248640127335e-05, + "loss": 0.9715, + "step": 31302 + }, + { + "epoch": 2.31, + "learning_rate": 1.4071884336296413e-05, + "loss": 1.0323, + "step": 31303 + }, + { + "epoch": 2.31, + "learning_rate": 1.4071520025987181e-05, + "loss": 1.0662, + "step": 31304 + }, + { + "epoch": 2.31, + "learning_rate": 1.4071155709200218e-05, + "loss": 0.9458, + "step": 31305 + }, + { + "epoch": 2.31, + "learning_rate": 1.4070791385936108e-05, + "loss": 1.0172, + "step": 31306 + }, + { + "epoch": 2.31, + "learning_rate": 1.4070427056195424e-05, + "loss": 0.9169, + "step": 31307 + }, + { + "epoch": 2.31, + "learning_rate": 1.407006271997875e-05, + "loss": 0.9739, + "step": 31308 + }, + { + "epoch": 2.31, + "learning_rate": 1.4069698377286662e-05, + "loss": 0.9137, + "step": 31309 + }, + { + "epoch": 2.31, + "learning_rate": 1.4069334028119746e-05, + "loss": 1.0237, + "step": 31310 + }, + { + "epoch": 2.31, + "learning_rate": 1.4068969672478577e-05, + "loss": 1.0229, + "step": 31311 + }, + { + "epoch": 2.31, + "learning_rate": 1.4068605310363734e-05, + "loss": 1.014, + "step": 31312 + }, + { + "epoch": 2.31, + "learning_rate": 1.4068240941775796e-05, + "loss": 1.0326, + "step": 31313 + }, + { + "epoch": 2.31, + "learning_rate": 1.4067876566715351e-05, + "loss": 0.994, + "step": 31314 + }, + { + "epoch": 2.31, + "learning_rate": 1.4067512185182968e-05, + "loss": 0.9433, + "step": 31315 + }, + { + "epoch": 2.31, + "learning_rate": 1.4067147797179232e-05, + "loss": 1.0515, + "step": 31316 + }, + { + "epoch": 2.31, + "learning_rate": 1.4066783402704722e-05, + "loss": 0.9411, + "step": 31317 + }, + { + "epoch": 2.31, + "learning_rate": 1.406641900176002e-05, + "loss": 0.9793, + "step": 31318 + }, + { + "epoch": 2.31, + "learning_rate": 1.4066054594345699e-05, + "loss": 1.0544, + "step": 31319 + }, + { + "epoch": 2.31, + "learning_rate": 1.4065690180462347e-05, + "loss": 1.1047, + "step": 31320 + }, + { + "epoch": 2.31, + "learning_rate": 1.406532576011054e-05, + "loss": 0.932, + "step": 31321 + }, + { + "epoch": 2.31, + "learning_rate": 1.4064961333290858e-05, + "loss": 1.1208, + "step": 31322 + }, + { + "epoch": 2.31, + "learning_rate": 1.4064596900003877e-05, + "loss": 1.0004, + "step": 31323 + }, + { + "epoch": 2.31, + "learning_rate": 1.4064232460250184e-05, + "loss": 1.1, + "step": 31324 + }, + { + "epoch": 2.31, + "learning_rate": 1.4063868014030355e-05, + "loss": 0.8948, + "step": 31325 + }, + { + "epoch": 2.31, + "learning_rate": 1.4063503561344973e-05, + "loss": 1.0298, + "step": 31326 + }, + { + "epoch": 2.31, + "learning_rate": 1.4063139102194609e-05, + "loss": 1.0155, + "step": 31327 + }, + { + "epoch": 2.31, + "learning_rate": 1.4062774636579852e-05, + "loss": 0.8791, + "step": 31328 + }, + { + "epoch": 2.31, + "learning_rate": 1.4062410164501276e-05, + "loss": 1.0588, + "step": 31329 + }, + { + "epoch": 2.31, + "learning_rate": 1.406204568595947e-05, + "loss": 0.9148, + "step": 31330 + }, + { + "epoch": 2.32, + "learning_rate": 1.4061681200955e-05, + "loss": 0.9912, + "step": 31331 + }, + { + "epoch": 2.32, + "learning_rate": 1.4061316709488457e-05, + "loss": 1.0406, + "step": 31332 + }, + { + "epoch": 2.32, + "learning_rate": 1.4060952211560417e-05, + "loss": 1.0483, + "step": 31333 + }, + { + "epoch": 2.32, + "learning_rate": 1.406058770717146e-05, + "loss": 1.0424, + "step": 31334 + }, + { + "epoch": 2.32, + "learning_rate": 1.4060223196322167e-05, + "loss": 0.8763, + "step": 31335 + }, + { + "epoch": 2.32, + "learning_rate": 1.4059858679013114e-05, + "loss": 1.0204, + "step": 31336 + }, + { + "epoch": 2.32, + "learning_rate": 1.4059494155244886e-05, + "loss": 0.9586, + "step": 31337 + }, + { + "epoch": 2.32, + "learning_rate": 1.405912962501806e-05, + "loss": 1.0424, + "step": 31338 + }, + { + "epoch": 2.32, + "learning_rate": 1.405876508833322e-05, + "loss": 1.0599, + "step": 31339 + }, + { + "epoch": 2.32, + "learning_rate": 1.4058400545190938e-05, + "loss": 0.95, + "step": 31340 + }, + { + "epoch": 2.32, + "learning_rate": 1.4058035995591802e-05, + "loss": 0.9148, + "step": 31341 + }, + { + "epoch": 2.32, + "learning_rate": 1.4057671439536389e-05, + "loss": 1.0386, + "step": 31342 + }, + { + "epoch": 2.32, + "learning_rate": 1.4057306877025276e-05, + "loss": 0.9563, + "step": 31343 + }, + { + "epoch": 2.32, + "learning_rate": 1.4056942308059046e-05, + "loss": 1.0935, + "step": 31344 + }, + { + "epoch": 2.32, + "learning_rate": 1.4056577732638284e-05, + "loss": 0.9225, + "step": 31345 + }, + { + "epoch": 2.32, + "learning_rate": 1.405621315076356e-05, + "loss": 1.0625, + "step": 31346 + }, + { + "epoch": 2.32, + "learning_rate": 1.405584856243546e-05, + "loss": 1.0172, + "step": 31347 + }, + { + "epoch": 2.32, + "learning_rate": 1.4055483967654562e-05, + "loss": 1.0156, + "step": 31348 + }, + { + "epoch": 2.32, + "learning_rate": 1.4055119366421451e-05, + "loss": 0.9771, + "step": 31349 + }, + { + "epoch": 2.32, + "learning_rate": 1.4054754758736697e-05, + "loss": 1.1069, + "step": 31350 + }, + { + "epoch": 2.32, + "learning_rate": 1.405439014460089e-05, + "loss": 1.0652, + "step": 31351 + }, + { + "epoch": 2.32, + "learning_rate": 1.4054025524014605e-05, + "loss": 1.015, + "step": 31352 + }, + { + "epoch": 2.32, + "learning_rate": 1.4053660896978427e-05, + "loss": 1.0004, + "step": 31353 + }, + { + "epoch": 2.32, + "learning_rate": 1.4053296263492928e-05, + "loss": 1.0447, + "step": 31354 + }, + { + "epoch": 2.32, + "learning_rate": 1.4052931623558697e-05, + "loss": 1.0988, + "step": 31355 + }, + { + "epoch": 2.32, + "learning_rate": 1.4052566977176306e-05, + "loss": 1.056, + "step": 31356 + }, + { + "epoch": 2.32, + "learning_rate": 1.405220232434634e-05, + "loss": 1.0206, + "step": 31357 + }, + { + "epoch": 2.32, + "learning_rate": 1.4051837665069377e-05, + "loss": 1.008, + "step": 31358 + }, + { + "epoch": 2.32, + "learning_rate": 1.4051472999346e-05, + "loss": 1.0458, + "step": 31359 + }, + { + "epoch": 2.32, + "learning_rate": 1.4051108327176786e-05, + "loss": 0.9364, + "step": 31360 + }, + { + "epoch": 2.32, + "learning_rate": 1.4050743648562318e-05, + "loss": 1.0111, + "step": 31361 + }, + { + "epoch": 2.32, + "learning_rate": 1.4050378963503176e-05, + "loss": 0.9514, + "step": 31362 + }, + { + "epoch": 2.32, + "learning_rate": 1.4050014271999936e-05, + "loss": 1.0283, + "step": 31363 + }, + { + "epoch": 2.32, + "learning_rate": 1.404964957405318e-05, + "loss": 0.9413, + "step": 31364 + }, + { + "epoch": 2.32, + "learning_rate": 1.4049284869663495e-05, + "loss": 1.0342, + "step": 31365 + }, + { + "epoch": 2.32, + "learning_rate": 1.404892015883145e-05, + "loss": 0.9943, + "step": 31366 + }, + { + "epoch": 2.32, + "learning_rate": 1.4048555441557635e-05, + "loss": 1.0224, + "step": 31367 + }, + { + "epoch": 2.32, + "learning_rate": 1.4048190717842625e-05, + "loss": 0.9731, + "step": 31368 + }, + { + "epoch": 2.32, + "learning_rate": 1.4047825987687001e-05, + "loss": 1.0696, + "step": 31369 + }, + { + "epoch": 2.32, + "learning_rate": 1.4047461251091344e-05, + "loss": 1.0364, + "step": 31370 + }, + { + "epoch": 2.32, + "learning_rate": 1.4047096508056232e-05, + "loss": 1.0577, + "step": 31371 + }, + { + "epoch": 2.32, + "learning_rate": 1.4046731758582252e-05, + "loss": 0.9185, + "step": 31372 + }, + { + "epoch": 2.32, + "learning_rate": 1.4046367002669978e-05, + "loss": 0.9939, + "step": 31373 + }, + { + "epoch": 2.32, + "learning_rate": 1.4046002240319989e-05, + "loss": 1.0115, + "step": 31374 + }, + { + "epoch": 2.32, + "learning_rate": 1.4045637471532872e-05, + "loss": 0.9816, + "step": 31375 + }, + { + "epoch": 2.32, + "learning_rate": 1.40452726963092e-05, + "loss": 1.0667, + "step": 31376 + }, + { + "epoch": 2.32, + "learning_rate": 1.4044907914649559e-05, + "loss": 0.9635, + "step": 31377 + }, + { + "epoch": 2.32, + "learning_rate": 1.4044543126554529e-05, + "loss": 0.9739, + "step": 31378 + }, + { + "epoch": 2.32, + "learning_rate": 1.4044178332024688e-05, + "loss": 0.9099, + "step": 31379 + }, + { + "epoch": 2.32, + "learning_rate": 1.4043813531060615e-05, + "loss": 0.9687, + "step": 31380 + }, + { + "epoch": 2.32, + "learning_rate": 1.4043448723662893e-05, + "loss": 0.9601, + "step": 31381 + }, + { + "epoch": 2.32, + "learning_rate": 1.4043083909832105e-05, + "loss": 0.8968, + "step": 31382 + }, + { + "epoch": 2.32, + "learning_rate": 1.4042719089568824e-05, + "loss": 1.0197, + "step": 31383 + }, + { + "epoch": 2.32, + "learning_rate": 1.4042354262873638e-05, + "loss": 1.0361, + "step": 31384 + }, + { + "epoch": 2.32, + "learning_rate": 1.4041989429747123e-05, + "loss": 0.9951, + "step": 31385 + }, + { + "epoch": 2.32, + "learning_rate": 1.4041624590189862e-05, + "loss": 1.0089, + "step": 31386 + }, + { + "epoch": 2.32, + "learning_rate": 1.4041259744202433e-05, + "loss": 1.0579, + "step": 31387 + }, + { + "epoch": 2.32, + "learning_rate": 1.4040894891785416e-05, + "loss": 0.9701, + "step": 31388 + }, + { + "epoch": 2.32, + "learning_rate": 1.4040530032939396e-05, + "loss": 1.0079, + "step": 31389 + }, + { + "epoch": 2.32, + "learning_rate": 1.404016516766495e-05, + "loss": 1.1428, + "step": 31390 + }, + { + "epoch": 2.32, + "learning_rate": 1.4039800295962655e-05, + "loss": 1.0157, + "step": 31391 + }, + { + "epoch": 2.32, + "learning_rate": 1.40394354178331e-05, + "loss": 0.9133, + "step": 31392 + }, + { + "epoch": 2.32, + "learning_rate": 1.403907053327686e-05, + "loss": 0.9697, + "step": 31393 + }, + { + "epoch": 2.32, + "learning_rate": 1.4038705642294517e-05, + "loss": 1.0487, + "step": 31394 + }, + { + "epoch": 2.32, + "learning_rate": 1.4038340744886649e-05, + "loss": 1.0305, + "step": 31395 + }, + { + "epoch": 2.32, + "learning_rate": 1.4037975841053842e-05, + "loss": 1.0321, + "step": 31396 + }, + { + "epoch": 2.32, + "learning_rate": 1.4037610930796667e-05, + "loss": 1.0932, + "step": 31397 + }, + { + "epoch": 2.32, + "learning_rate": 1.4037246014115717e-05, + "loss": 0.9688, + "step": 31398 + }, + { + "epoch": 2.32, + "learning_rate": 1.4036881091011563e-05, + "loss": 0.9208, + "step": 31399 + }, + { + "epoch": 2.32, + "learning_rate": 1.4036516161484794e-05, + "loss": 1.0598, + "step": 31400 + }, + { + "epoch": 2.32, + "learning_rate": 1.4036151225535979e-05, + "loss": 0.9362, + "step": 31401 + }, + { + "epoch": 2.32, + "learning_rate": 1.4035786283165708e-05, + "loss": 0.9336, + "step": 31402 + }, + { + "epoch": 2.32, + "learning_rate": 1.4035421334374559e-05, + "loss": 1.1027, + "step": 31403 + }, + { + "epoch": 2.32, + "learning_rate": 1.4035056379163115e-05, + "loss": 0.9965, + "step": 31404 + }, + { + "epoch": 2.32, + "learning_rate": 1.4034691417531949e-05, + "loss": 1.078, + "step": 31405 + }, + { + "epoch": 2.32, + "learning_rate": 1.4034326449481649e-05, + "loss": 1.0233, + "step": 31406 + }, + { + "epoch": 2.32, + "learning_rate": 1.4033961475012792e-05, + "loss": 0.8283, + "step": 31407 + }, + { + "epoch": 2.32, + "learning_rate": 1.4033596494125963e-05, + "loss": 1.0716, + "step": 31408 + }, + { + "epoch": 2.32, + "learning_rate": 1.4033231506821738e-05, + "loss": 0.98, + "step": 31409 + }, + { + "epoch": 2.32, + "learning_rate": 1.40328665131007e-05, + "loss": 1.0581, + "step": 31410 + }, + { + "epoch": 2.32, + "learning_rate": 1.4032501512963425e-05, + "loss": 0.9714, + "step": 31411 + }, + { + "epoch": 2.32, + "learning_rate": 1.4032136506410503e-05, + "loss": 0.9959, + "step": 31412 + }, + { + "epoch": 2.32, + "learning_rate": 1.4031771493442507e-05, + "loss": 1.02, + "step": 31413 + }, + { + "epoch": 2.32, + "learning_rate": 1.4031406474060023e-05, + "loss": 1.0146, + "step": 31414 + }, + { + "epoch": 2.32, + "learning_rate": 1.4031041448263624e-05, + "loss": 0.8789, + "step": 31415 + }, + { + "epoch": 2.32, + "learning_rate": 1.4030676416053899e-05, + "loss": 0.9753, + "step": 31416 + }, + { + "epoch": 2.32, + "learning_rate": 1.4030311377431425e-05, + "loss": 1.0343, + "step": 31417 + }, + { + "epoch": 2.32, + "learning_rate": 1.4029946332396786e-05, + "loss": 1.1098, + "step": 31418 + }, + { + "epoch": 2.32, + "learning_rate": 1.4029581280950554e-05, + "loss": 1.0939, + "step": 31419 + }, + { + "epoch": 2.32, + "learning_rate": 1.4029216223093323e-05, + "loss": 1.0279, + "step": 31420 + }, + { + "epoch": 2.32, + "learning_rate": 1.402885115882566e-05, + "loss": 1.066, + "step": 31421 + }, + { + "epoch": 2.32, + "learning_rate": 1.4028486088148157e-05, + "loss": 0.9542, + "step": 31422 + }, + { + "epoch": 2.32, + "learning_rate": 1.4028121011061386e-05, + "loss": 0.9737, + "step": 31423 + }, + { + "epoch": 2.32, + "learning_rate": 1.4027755927565938e-05, + "loss": 0.978, + "step": 31424 + }, + { + "epoch": 2.32, + "learning_rate": 1.4027390837662383e-05, + "loss": 0.9661, + "step": 31425 + }, + { + "epoch": 2.32, + "learning_rate": 1.4027025741351308e-05, + "loss": 0.9671, + "step": 31426 + }, + { + "epoch": 2.32, + "learning_rate": 1.4026660638633294e-05, + "loss": 0.8825, + "step": 31427 + }, + { + "epoch": 2.32, + "learning_rate": 1.402629552950892e-05, + "loss": 0.9138, + "step": 31428 + }, + { + "epoch": 2.32, + "learning_rate": 1.4025930413978769e-05, + "loss": 1.0829, + "step": 31429 + }, + { + "epoch": 2.32, + "learning_rate": 1.4025565292043417e-05, + "loss": 0.853, + "step": 31430 + }, + { + "epoch": 2.32, + "learning_rate": 1.4025200163703448e-05, + "loss": 0.92, + "step": 31431 + }, + { + "epoch": 2.32, + "learning_rate": 1.4024835028959447e-05, + "loss": 0.979, + "step": 31432 + }, + { + "epoch": 2.32, + "learning_rate": 1.4024469887811986e-05, + "loss": 0.9908, + "step": 31433 + }, + { + "epoch": 2.32, + "learning_rate": 1.4024104740261655e-05, + "loss": 1.029, + "step": 31434 + }, + { + "epoch": 2.32, + "learning_rate": 1.402373958630903e-05, + "loss": 1.0159, + "step": 31435 + }, + { + "epoch": 2.32, + "learning_rate": 1.4023374425954693e-05, + "loss": 0.9121, + "step": 31436 + }, + { + "epoch": 2.32, + "learning_rate": 1.4023009259199225e-05, + "loss": 0.9557, + "step": 31437 + }, + { + "epoch": 2.32, + "learning_rate": 1.4022644086043205e-05, + "loss": 1.0215, + "step": 31438 + }, + { + "epoch": 2.32, + "learning_rate": 1.4022278906487217e-05, + "loss": 0.8484, + "step": 31439 + }, + { + "epoch": 2.32, + "learning_rate": 1.4021913720531839e-05, + "loss": 1.0627, + "step": 31440 + }, + { + "epoch": 2.32, + "learning_rate": 1.4021548528177658e-05, + "loss": 0.986, + "step": 31441 + }, + { + "epoch": 2.32, + "learning_rate": 1.4021183329425246e-05, + "loss": 0.9958, + "step": 31442 + }, + { + "epoch": 2.32, + "learning_rate": 1.4020818124275192e-05, + "loss": 1.048, + "step": 31443 + }, + { + "epoch": 2.32, + "learning_rate": 1.4020452912728073e-05, + "loss": 1.0471, + "step": 31444 + }, + { + "epoch": 2.32, + "learning_rate": 1.4020087694784471e-05, + "loss": 0.9577, + "step": 31445 + }, + { + "epoch": 2.32, + "learning_rate": 1.401972247044497e-05, + "loss": 0.9696, + "step": 31446 + }, + { + "epoch": 2.32, + "learning_rate": 1.4019357239710145e-05, + "loss": 0.9439, + "step": 31447 + }, + { + "epoch": 2.32, + "learning_rate": 1.401899200258058e-05, + "loss": 1.1282, + "step": 31448 + }, + { + "epoch": 2.32, + "learning_rate": 1.4018626759056857e-05, + "loss": 0.9785, + "step": 31449 + }, + { + "epoch": 2.32, + "learning_rate": 1.4018261509139554e-05, + "loss": 1.1276, + "step": 31450 + }, + { + "epoch": 2.32, + "learning_rate": 1.4017896252829258e-05, + "loss": 1.0135, + "step": 31451 + }, + { + "epoch": 2.32, + "learning_rate": 1.4017530990126544e-05, + "loss": 1.0798, + "step": 31452 + }, + { + "epoch": 2.32, + "learning_rate": 1.4017165721031995e-05, + "loss": 1.0011, + "step": 31453 + }, + { + "epoch": 2.32, + "learning_rate": 1.4016800445546194e-05, + "loss": 0.9996, + "step": 31454 + }, + { + "epoch": 2.32, + "learning_rate": 1.4016435163669724e-05, + "loss": 0.9166, + "step": 31455 + }, + { + "epoch": 2.32, + "learning_rate": 1.401606987540316e-05, + "loss": 0.995, + "step": 31456 + }, + { + "epoch": 2.32, + "learning_rate": 1.401570458074709e-05, + "loss": 1.0012, + "step": 31457 + }, + { + "epoch": 2.32, + "learning_rate": 1.4015339279702085e-05, + "loss": 1.0573, + "step": 31458 + }, + { + "epoch": 2.32, + "learning_rate": 1.4014973972268737e-05, + "loss": 1.0496, + "step": 31459 + }, + { + "epoch": 2.32, + "learning_rate": 1.401460865844762e-05, + "loss": 0.87, + "step": 31460 + }, + { + "epoch": 2.32, + "learning_rate": 1.4014243338239322e-05, + "loss": 0.9868, + "step": 31461 + }, + { + "epoch": 2.32, + "learning_rate": 1.4013878011644417e-05, + "loss": 1.0909, + "step": 31462 + }, + { + "epoch": 2.32, + "learning_rate": 1.401351267866349e-05, + "loss": 0.9911, + "step": 31463 + }, + { + "epoch": 2.32, + "learning_rate": 1.4013147339297121e-05, + "loss": 1.0706, + "step": 31464 + }, + { + "epoch": 2.32, + "learning_rate": 1.4012781993545897e-05, + "loss": 0.8845, + "step": 31465 + }, + { + "epoch": 2.33, + "learning_rate": 1.4012416641410389e-05, + "loss": 0.9769, + "step": 31466 + }, + { + "epoch": 2.33, + "learning_rate": 1.4012051282891186e-05, + "loss": 0.9684, + "step": 31467 + }, + { + "epoch": 2.33, + "learning_rate": 1.4011685917988866e-05, + "loss": 1.0296, + "step": 31468 + }, + { + "epoch": 2.33, + "learning_rate": 1.4011320546704013e-05, + "loss": 1.0343, + "step": 31469 + }, + { + "epoch": 2.33, + "learning_rate": 1.4010955169037202e-05, + "loss": 0.9934, + "step": 31470 + }, + { + "epoch": 2.33, + "learning_rate": 1.4010589784989021e-05, + "loss": 1.0006, + "step": 31471 + }, + { + "epoch": 2.33, + "learning_rate": 1.4010224394560048e-05, + "loss": 0.9669, + "step": 31472 + }, + { + "epoch": 2.33, + "learning_rate": 1.4009858997750869e-05, + "loss": 1.1148, + "step": 31473 + }, + { + "epoch": 2.33, + "learning_rate": 1.400949359456206e-05, + "loss": 1.0569, + "step": 31474 + }, + { + "epoch": 2.33, + "learning_rate": 1.4009128184994202e-05, + "loss": 1.1047, + "step": 31475 + }, + { + "epoch": 2.33, + "learning_rate": 1.4008762769047879e-05, + "loss": 0.9452, + "step": 31476 + }, + { + "epoch": 2.33, + "learning_rate": 1.4008397346723675e-05, + "loss": 0.9005, + "step": 31477 + }, + { + "epoch": 2.33, + "learning_rate": 1.4008031918022165e-05, + "loss": 0.9791, + "step": 31478 + }, + { + "epoch": 2.33, + "learning_rate": 1.4007666482943934e-05, + "loss": 0.907, + "step": 31479 + }, + { + "epoch": 2.33, + "learning_rate": 1.4007301041489563e-05, + "loss": 1.0162, + "step": 31480 + }, + { + "epoch": 2.33, + "learning_rate": 1.4006935593659635e-05, + "loss": 0.9367, + "step": 31481 + }, + { + "epoch": 2.33, + "learning_rate": 1.4006570139454727e-05, + "loss": 1.0652, + "step": 31482 + }, + { + "epoch": 2.33, + "learning_rate": 1.4006204678875426e-05, + "loss": 1.1311, + "step": 31483 + }, + { + "epoch": 2.33, + "learning_rate": 1.400583921192231e-05, + "loss": 1.0027, + "step": 31484 + }, + { + "epoch": 2.33, + "learning_rate": 1.4005473738595958e-05, + "loss": 1.0938, + "step": 31485 + }, + { + "epoch": 2.33, + "learning_rate": 1.4005108258896958e-05, + "loss": 0.948, + "step": 31486 + }, + { + "epoch": 2.33, + "learning_rate": 1.4004742772825885e-05, + "loss": 1.0446, + "step": 31487 + }, + { + "epoch": 2.33, + "learning_rate": 1.4004377280383328e-05, + "loss": 1.0666, + "step": 31488 + }, + { + "epoch": 2.33, + "learning_rate": 1.4004011781569859e-05, + "loss": 1.0632, + "step": 31489 + }, + { + "epoch": 2.33, + "learning_rate": 1.400364627638607e-05, + "loss": 1.0395, + "step": 31490 + }, + { + "epoch": 2.33, + "learning_rate": 1.4003280764832534e-05, + "loss": 0.9396, + "step": 31491 + }, + { + "epoch": 2.33, + "learning_rate": 1.4002915246909833e-05, + "loss": 0.9922, + "step": 31492 + }, + { + "epoch": 2.33, + "learning_rate": 1.4002549722618557e-05, + "loss": 0.9558, + "step": 31493 + }, + { + "epoch": 2.33, + "learning_rate": 1.4002184191959281e-05, + "loss": 0.8321, + "step": 31494 + }, + { + "epoch": 2.33, + "learning_rate": 1.4001818654932581e-05, + "loss": 0.9669, + "step": 31495 + }, + { + "epoch": 2.33, + "learning_rate": 1.4001453111539051e-05, + "loss": 1.0314, + "step": 31496 + }, + { + "epoch": 2.33, + "learning_rate": 1.4001087561779262e-05, + "loss": 0.9838, + "step": 31497 + }, + { + "epoch": 2.33, + "learning_rate": 1.4000722005653804e-05, + "loss": 0.9349, + "step": 31498 + }, + { + "epoch": 2.33, + "learning_rate": 1.4000356443163252e-05, + "loss": 1.0615, + "step": 31499 + }, + { + "epoch": 2.33, + "learning_rate": 1.399999087430819e-05, + "loss": 0.9568, + "step": 31500 + }, + { + "epoch": 2.33, + "learning_rate": 1.39996252990892e-05, + "loss": 1.0273, + "step": 31501 + }, + { + "epoch": 2.33, + "learning_rate": 1.3999259717506864e-05, + "loss": 0.941, + "step": 31502 + }, + { + "epoch": 2.33, + "learning_rate": 1.3998894129561764e-05, + "loss": 1.0707, + "step": 31503 + }, + { + "epoch": 2.33, + "learning_rate": 1.3998528535254478e-05, + "loss": 0.9774, + "step": 31504 + }, + { + "epoch": 2.33, + "learning_rate": 1.3998162934585593e-05, + "loss": 0.9853, + "step": 31505 + }, + { + "epoch": 2.33, + "learning_rate": 1.3997797327555686e-05, + "loss": 0.9282, + "step": 31506 + }, + { + "epoch": 2.33, + "learning_rate": 1.3997431714165342e-05, + "loss": 1.0897, + "step": 31507 + }, + { + "epoch": 2.33, + "learning_rate": 1.3997066094415141e-05, + "loss": 1.0341, + "step": 31508 + }, + { + "epoch": 2.33, + "learning_rate": 1.3996700468305664e-05, + "loss": 0.8929, + "step": 31509 + }, + { + "epoch": 2.33, + "learning_rate": 1.3996334835837496e-05, + "loss": 0.9169, + "step": 31510 + }, + { + "epoch": 2.33, + "learning_rate": 1.3995969197011212e-05, + "loss": 1.0066, + "step": 31511 + }, + { + "epoch": 2.33, + "learning_rate": 1.3995603551827403e-05, + "loss": 1.1165, + "step": 31512 + }, + { + "epoch": 2.33, + "learning_rate": 1.3995237900286645e-05, + "loss": 1.0332, + "step": 31513 + }, + { + "epoch": 2.33, + "learning_rate": 1.3994872242389519e-05, + "loss": 0.9944, + "step": 31514 + }, + { + "epoch": 2.33, + "learning_rate": 1.3994506578136608e-05, + "loss": 1.1214, + "step": 31515 + }, + { + "epoch": 2.33, + "learning_rate": 1.3994140907528498e-05, + "loss": 0.9417, + "step": 31516 + }, + { + "epoch": 2.33, + "learning_rate": 1.3993775230565763e-05, + "loss": 0.9047, + "step": 31517 + }, + { + "epoch": 2.33, + "learning_rate": 1.3993409547248988e-05, + "loss": 0.9581, + "step": 31518 + }, + { + "epoch": 2.33, + "learning_rate": 1.3993043857578759e-05, + "loss": 1.0802, + "step": 31519 + }, + { + "epoch": 2.33, + "learning_rate": 1.3992678161555653e-05, + "loss": 0.971, + "step": 31520 + }, + { + "epoch": 2.33, + "learning_rate": 1.3992312459180253e-05, + "loss": 1.0846, + "step": 31521 + }, + { + "epoch": 2.33, + "learning_rate": 1.3991946750453143e-05, + "loss": 1.1274, + "step": 31522 + }, + { + "epoch": 2.33, + "learning_rate": 1.39915810353749e-05, + "loss": 0.9608, + "step": 31523 + }, + { + "epoch": 2.33, + "learning_rate": 1.399121531394611e-05, + "loss": 1.0599, + "step": 31524 + }, + { + "epoch": 2.33, + "learning_rate": 1.3990849586167353e-05, + "loss": 0.9436, + "step": 31525 + }, + { + "epoch": 2.33, + "learning_rate": 1.399048385203921e-05, + "loss": 1.0445, + "step": 31526 + }, + { + "epoch": 2.33, + "learning_rate": 1.3990118111562265e-05, + "loss": 0.97, + "step": 31527 + }, + { + "epoch": 2.33, + "learning_rate": 1.3989752364737101e-05, + "loss": 1.0684, + "step": 31528 + }, + { + "epoch": 2.33, + "learning_rate": 1.3989386611564295e-05, + "loss": 1.0552, + "step": 31529 + }, + { + "epoch": 2.33, + "learning_rate": 1.3989020852044437e-05, + "loss": 1.0055, + "step": 31530 + }, + { + "epoch": 2.33, + "learning_rate": 1.39886550861781e-05, + "loss": 0.9548, + "step": 31531 + }, + { + "epoch": 2.33, + "learning_rate": 1.398828931396587e-05, + "loss": 1.1049, + "step": 31532 + }, + { + "epoch": 2.33, + "learning_rate": 1.398792353540833e-05, + "loss": 1.0492, + "step": 31533 + }, + { + "epoch": 2.33, + "learning_rate": 1.398755775050606e-05, + "loss": 0.8875, + "step": 31534 + }, + { + "epoch": 2.33, + "learning_rate": 1.3987191959259643e-05, + "loss": 1.0162, + "step": 31535 + }, + { + "epoch": 2.33, + "learning_rate": 1.3986826161669663e-05, + "loss": 0.9983, + "step": 31536 + }, + { + "epoch": 2.33, + "learning_rate": 1.3986460357736696e-05, + "loss": 0.9492, + "step": 31537 + }, + { + "epoch": 2.33, + "learning_rate": 1.398609454746133e-05, + "loss": 1.0385, + "step": 31538 + }, + { + "epoch": 2.33, + "learning_rate": 1.398572873084414e-05, + "loss": 0.9667, + "step": 31539 + }, + { + "epoch": 2.33, + "learning_rate": 1.3985362907885718e-05, + "loss": 0.9971, + "step": 31540 + }, + { + "epoch": 2.33, + "learning_rate": 1.398499707858664e-05, + "loss": 1.0831, + "step": 31541 + }, + { + "epoch": 2.33, + "learning_rate": 1.3984631242947484e-05, + "loss": 1.0591, + "step": 31542 + }, + { + "epoch": 2.33, + "learning_rate": 1.3984265400968841e-05, + "loss": 1.0523, + "step": 31543 + }, + { + "epoch": 2.33, + "learning_rate": 1.3983899552651287e-05, + "loss": 0.9599, + "step": 31544 + }, + { + "epoch": 2.33, + "learning_rate": 1.398353369799541e-05, + "loss": 1.1051, + "step": 31545 + }, + { + "epoch": 2.33, + "learning_rate": 1.3983167837001782e-05, + "loss": 1.0837, + "step": 31546 + }, + { + "epoch": 2.33, + "learning_rate": 1.3982801969670994e-05, + "loss": 0.9564, + "step": 31547 + }, + { + "epoch": 2.33, + "learning_rate": 1.3982436096003623e-05, + "loss": 0.9325, + "step": 31548 + }, + { + "epoch": 2.33, + "learning_rate": 1.3982070216000257e-05, + "loss": 0.9188, + "step": 31549 + }, + { + "epoch": 2.33, + "learning_rate": 1.398170432966147e-05, + "loss": 1.0226, + "step": 31550 + }, + { + "epoch": 2.33, + "learning_rate": 1.3981338436987852e-05, + "loss": 0.9944, + "step": 31551 + }, + { + "epoch": 2.33, + "learning_rate": 1.3980972537979978e-05, + "loss": 0.9233, + "step": 31552 + }, + { + "epoch": 2.33, + "learning_rate": 1.3980606632638437e-05, + "loss": 1.0681, + "step": 31553 + }, + { + "epoch": 2.33, + "learning_rate": 1.3980240720963804e-05, + "loss": 0.954, + "step": 31554 + }, + { + "epoch": 2.33, + "learning_rate": 1.397987480295667e-05, + "loss": 1.1113, + "step": 31555 + }, + { + "epoch": 2.33, + "learning_rate": 1.3979508878617607e-05, + "loss": 0.9993, + "step": 31556 + }, + { + "epoch": 2.33, + "learning_rate": 1.3979142947947207e-05, + "loss": 1.031, + "step": 31557 + }, + { + "epoch": 2.33, + "learning_rate": 1.3978777010946044e-05, + "loss": 0.8914, + "step": 31558 + }, + { + "epoch": 2.33, + "learning_rate": 1.3978411067614705e-05, + "loss": 1.1048, + "step": 31559 + }, + { + "epoch": 2.33, + "learning_rate": 1.3978045117953772e-05, + "loss": 1.0322, + "step": 31560 + }, + { + "epoch": 2.33, + "learning_rate": 1.3977679161963824e-05, + "loss": 0.9917, + "step": 31561 + }, + { + "epoch": 2.33, + "learning_rate": 1.3977313199645445e-05, + "loss": 0.9589, + "step": 31562 + }, + { + "epoch": 2.33, + "learning_rate": 1.397694723099922e-05, + "loss": 0.9961, + "step": 31563 + }, + { + "epoch": 2.33, + "learning_rate": 1.3976581256025728e-05, + "loss": 0.9864, + "step": 31564 + }, + { + "epoch": 2.33, + "learning_rate": 1.3976215274725551e-05, + "loss": 0.9447, + "step": 31565 + }, + { + "epoch": 2.33, + "learning_rate": 1.3975849287099273e-05, + "loss": 1.0248, + "step": 31566 + }, + { + "epoch": 2.33, + "learning_rate": 1.3975483293147477e-05, + "loss": 1.1192, + "step": 31567 + }, + { + "epoch": 2.33, + "learning_rate": 1.3975117292870743e-05, + "loss": 0.9077, + "step": 31568 + }, + { + "epoch": 2.33, + "learning_rate": 1.3974751286269655e-05, + "loss": 0.9985, + "step": 31569 + }, + { + "epoch": 2.33, + "learning_rate": 1.3974385273344793e-05, + "loss": 1.0704, + "step": 31570 + }, + { + "epoch": 2.33, + "learning_rate": 1.3974019254096742e-05, + "loss": 1.1381, + "step": 31571 + }, + { + "epoch": 2.33, + "learning_rate": 1.397365322852608e-05, + "loss": 0.979, + "step": 31572 + }, + { + "epoch": 2.33, + "learning_rate": 1.3973287196633396e-05, + "loss": 0.9827, + "step": 31573 + }, + { + "epoch": 2.33, + "learning_rate": 1.3972921158419268e-05, + "loss": 1.0913, + "step": 31574 + }, + { + "epoch": 2.33, + "learning_rate": 1.397255511388428e-05, + "loss": 0.9935, + "step": 31575 + }, + { + "epoch": 2.33, + "learning_rate": 1.3972189063029014e-05, + "loss": 0.9771, + "step": 31576 + }, + { + "epoch": 2.33, + "learning_rate": 1.397182300585405e-05, + "loss": 0.999, + "step": 31577 + }, + { + "epoch": 2.33, + "learning_rate": 1.3971456942359976e-05, + "loss": 0.8762, + "step": 31578 + }, + { + "epoch": 2.33, + "learning_rate": 1.3971090872547369e-05, + "loss": 1.0297, + "step": 31579 + }, + { + "epoch": 2.33, + "learning_rate": 1.3970724796416812e-05, + "loss": 1.1157, + "step": 31580 + }, + { + "epoch": 2.33, + "learning_rate": 1.397035871396889e-05, + "loss": 0.949, + "step": 31581 + }, + { + "epoch": 2.33, + "learning_rate": 1.3969992625204181e-05, + "loss": 0.9546, + "step": 31582 + }, + { + "epoch": 2.33, + "learning_rate": 1.3969626530123275e-05, + "loss": 0.9257, + "step": 31583 + }, + { + "epoch": 2.33, + "learning_rate": 1.3969260428726745e-05, + "loss": 1.0326, + "step": 31584 + }, + { + "epoch": 2.33, + "learning_rate": 1.3968894321015185e-05, + "loss": 0.8928, + "step": 31585 + }, + { + "epoch": 2.33, + "learning_rate": 1.3968528206989165e-05, + "loss": 1.1334, + "step": 31586 + }, + { + "epoch": 2.33, + "learning_rate": 1.3968162086649277e-05, + "loss": 0.9388, + "step": 31587 + }, + { + "epoch": 2.33, + "learning_rate": 1.3967795959996097e-05, + "loss": 1.1192, + "step": 31588 + }, + { + "epoch": 2.33, + "learning_rate": 1.3967429827030212e-05, + "loss": 1.0914, + "step": 31589 + }, + { + "epoch": 2.33, + "learning_rate": 1.3967063687752204e-05, + "loss": 0.979, + "step": 31590 + }, + { + "epoch": 2.33, + "learning_rate": 1.3966697542162653e-05, + "loss": 0.9939, + "step": 31591 + }, + { + "epoch": 2.33, + "learning_rate": 1.3966331390262145e-05, + "loss": 1.0904, + "step": 31592 + }, + { + "epoch": 2.33, + "learning_rate": 1.3965965232051258e-05, + "loss": 1.0255, + "step": 31593 + }, + { + "epoch": 2.33, + "learning_rate": 1.3965599067530577e-05, + "loss": 0.889, + "step": 31594 + }, + { + "epoch": 2.33, + "learning_rate": 1.3965232896700685e-05, + "loss": 0.9525, + "step": 31595 + }, + { + "epoch": 2.33, + "learning_rate": 1.3964866719562169e-05, + "loss": 1.0545, + "step": 31596 + }, + { + "epoch": 2.33, + "learning_rate": 1.3964500536115599e-05, + "loss": 0.9038, + "step": 31597 + }, + { + "epoch": 2.33, + "learning_rate": 1.3964134346361571e-05, + "loss": 0.9697, + "step": 31598 + }, + { + "epoch": 2.33, + "learning_rate": 1.3963768150300658e-05, + "loss": 0.938, + "step": 31599 + }, + { + "epoch": 2.33, + "learning_rate": 1.396340194793345e-05, + "loss": 1.1613, + "step": 31600 + }, + { + "epoch": 2.33, + "learning_rate": 1.3963035739260521e-05, + "loss": 1.0587, + "step": 31601 + }, + { + "epoch": 2.34, + "learning_rate": 1.3962669524282465e-05, + "loss": 0.9911, + "step": 31602 + }, + { + "epoch": 2.34, + "learning_rate": 1.3962303302999855e-05, + "loss": 0.9811, + "step": 31603 + }, + { + "epoch": 2.34, + "learning_rate": 1.3961937075413278e-05, + "loss": 1.0853, + "step": 31604 + }, + { + "epoch": 2.34, + "learning_rate": 1.3961570841523318e-05, + "loss": 1.0301, + "step": 31605 + }, + { + "epoch": 2.34, + "learning_rate": 1.3961204601330553e-05, + "loss": 0.933, + "step": 31606 + }, + { + "epoch": 2.34, + "learning_rate": 1.3960838354835569e-05, + "loss": 1.0367, + "step": 31607 + }, + { + "epoch": 2.34, + "learning_rate": 1.3960472102038947e-05, + "loss": 1.0312, + "step": 31608 + }, + { + "epoch": 2.34, + "learning_rate": 1.3960105842941269e-05, + "loss": 1.0542, + "step": 31609 + }, + { + "epoch": 2.34, + "learning_rate": 1.3959739577543123e-05, + "loss": 1.0374, + "step": 31610 + }, + { + "epoch": 2.34, + "learning_rate": 1.3959373305845085e-05, + "loss": 1.0342, + "step": 31611 + }, + { + "epoch": 2.34, + "learning_rate": 1.3959007027847743e-05, + "loss": 0.9972, + "step": 31612 + }, + { + "epoch": 2.34, + "learning_rate": 1.3958640743551674e-05, + "loss": 0.9313, + "step": 31613 + }, + { + "epoch": 2.34, + "learning_rate": 1.3958274452957469e-05, + "loss": 1.0398, + "step": 31614 + }, + { + "epoch": 2.34, + "learning_rate": 1.3957908156065705e-05, + "loss": 0.9965, + "step": 31615 + }, + { + "epoch": 2.34, + "learning_rate": 1.3957541852876965e-05, + "loss": 0.9885, + "step": 31616 + }, + { + "epoch": 2.34, + "learning_rate": 1.395717554339183e-05, + "loss": 1.02, + "step": 31617 + }, + { + "epoch": 2.34, + "learning_rate": 1.3956809227610888e-05, + "loss": 1.0966, + "step": 31618 + }, + { + "epoch": 2.34, + "learning_rate": 1.3956442905534717e-05, + "loss": 0.9797, + "step": 31619 + }, + { + "epoch": 2.34, + "learning_rate": 1.3956076577163904e-05, + "loss": 1.0447, + "step": 31620 + }, + { + "epoch": 2.34, + "learning_rate": 1.395571024249903e-05, + "loss": 1.1162, + "step": 31621 + }, + { + "epoch": 2.34, + "learning_rate": 1.3955343901540677e-05, + "loss": 1.0336, + "step": 31622 + }, + { + "epoch": 2.34, + "learning_rate": 1.3954977554289426e-05, + "loss": 1.0643, + "step": 31623 + }, + { + "epoch": 2.34, + "learning_rate": 1.3954611200745867e-05, + "loss": 1.0092, + "step": 31624 + }, + { + "epoch": 2.34, + "learning_rate": 1.3954244840910574e-05, + "loss": 1.002, + "step": 31625 + }, + { + "epoch": 2.34, + "learning_rate": 1.3953878474784137e-05, + "loss": 1.0336, + "step": 31626 + }, + { + "epoch": 2.34, + "learning_rate": 1.3953512102367132e-05, + "loss": 0.9369, + "step": 31627 + }, + { + "epoch": 2.34, + "learning_rate": 1.395314572366015e-05, + "loss": 0.9793, + "step": 31628 + }, + { + "epoch": 2.34, + "learning_rate": 1.3952779338663766e-05, + "loss": 0.9719, + "step": 31629 + }, + { + "epoch": 2.34, + "learning_rate": 1.395241294737857e-05, + "loss": 1.0511, + "step": 31630 + }, + { + "epoch": 2.34, + "learning_rate": 1.3952046549805141e-05, + "loss": 1.0811, + "step": 31631 + }, + { + "epoch": 2.34, + "learning_rate": 1.3951680145944062e-05, + "loss": 1.0536, + "step": 31632 + }, + { + "epoch": 2.34, + "learning_rate": 1.3951313735795915e-05, + "loss": 1.0654, + "step": 31633 + }, + { + "epoch": 2.34, + "learning_rate": 1.3950947319361285e-05, + "loss": 1.0553, + "step": 31634 + }, + { + "epoch": 2.34, + "learning_rate": 1.3950580896640755e-05, + "loss": 1.0737, + "step": 31635 + }, + { + "epoch": 2.34, + "learning_rate": 1.3950214467634905e-05, + "loss": 0.9879, + "step": 31636 + }, + { + "epoch": 2.34, + "learning_rate": 1.3949848032344324e-05, + "loss": 0.9646, + "step": 31637 + }, + { + "epoch": 2.34, + "learning_rate": 1.394948159076959e-05, + "loss": 0.9154, + "step": 31638 + }, + { + "epoch": 2.34, + "learning_rate": 1.394911514291129e-05, + "loss": 0.9726, + "step": 31639 + }, + { + "epoch": 2.34, + "learning_rate": 1.394874868877e-05, + "loss": 1.084, + "step": 31640 + }, + { + "epoch": 2.34, + "learning_rate": 1.3948382228346309e-05, + "loss": 0.9983, + "step": 31641 + }, + { + "epoch": 2.34, + "learning_rate": 1.3948015761640798e-05, + "loss": 1.0889, + "step": 31642 + }, + { + "epoch": 2.34, + "learning_rate": 1.394764928865405e-05, + "loss": 0.9474, + "step": 31643 + }, + { + "epoch": 2.34, + "learning_rate": 1.3947282809386648e-05, + "loss": 0.9678, + "step": 31644 + }, + { + "epoch": 2.34, + "learning_rate": 1.3946916323839177e-05, + "loss": 1.0805, + "step": 31645 + }, + { + "epoch": 2.34, + "learning_rate": 1.3946549832012217e-05, + "loss": 0.8994, + "step": 31646 + }, + { + "epoch": 2.34, + "learning_rate": 1.3946183333906355e-05, + "loss": 0.9504, + "step": 31647 + }, + { + "epoch": 2.34, + "learning_rate": 1.3945816829522169e-05, + "loss": 1.0356, + "step": 31648 + }, + { + "epoch": 2.34, + "learning_rate": 1.3945450318860248e-05, + "loss": 1.0819, + "step": 31649 + }, + { + "epoch": 2.34, + "learning_rate": 1.3945083801921168e-05, + "loss": 1.0991, + "step": 31650 + }, + { + "epoch": 2.34, + "learning_rate": 1.3944717278705519e-05, + "loss": 1.0626, + "step": 31651 + }, + { + "epoch": 2.34, + "learning_rate": 1.3944350749213884e-05, + "loss": 0.9158, + "step": 31652 + }, + { + "epoch": 2.34, + "learning_rate": 1.394398421344684e-05, + "loss": 1.0645, + "step": 31653 + }, + { + "epoch": 2.34, + "learning_rate": 1.3943617671404972e-05, + "loss": 0.9283, + "step": 31654 + }, + { + "epoch": 2.34, + "learning_rate": 1.3943251123088868e-05, + "loss": 1.0038, + "step": 31655 + }, + { + "epoch": 2.34, + "learning_rate": 1.3942884568499105e-05, + "loss": 1.0012, + "step": 31656 + }, + { + "epoch": 2.34, + "learning_rate": 1.3942518007636271e-05, + "loss": 0.9686, + "step": 31657 + }, + { + "epoch": 2.34, + "learning_rate": 1.3942151440500946e-05, + "loss": 1.119, + "step": 31658 + }, + { + "epoch": 2.34, + "learning_rate": 1.3941784867093715e-05, + "loss": 1.0515, + "step": 31659 + }, + { + "epoch": 2.34, + "learning_rate": 1.394141828741516e-05, + "loss": 0.9301, + "step": 31660 + }, + { + "epoch": 2.34, + "learning_rate": 1.3941051701465868e-05, + "loss": 1.0919, + "step": 31661 + }, + { + "epoch": 2.34, + "learning_rate": 1.3940685109246417e-05, + "loss": 0.9595, + "step": 31662 + }, + { + "epoch": 2.34, + "learning_rate": 1.3940318510757391e-05, + "loss": 1.0079, + "step": 31663 + }, + { + "epoch": 2.34, + "learning_rate": 1.3939951905999376e-05, + "loss": 1.0133, + "step": 31664 + }, + { + "epoch": 2.34, + "learning_rate": 1.3939585294972954e-05, + "loss": 0.9921, + "step": 31665 + }, + { + "epoch": 2.34, + "learning_rate": 1.3939218677678707e-05, + "loss": 1.0402, + "step": 31666 + }, + { + "epoch": 2.34, + "learning_rate": 1.393885205411722e-05, + "loss": 1.0748, + "step": 31667 + }, + { + "epoch": 2.34, + "learning_rate": 1.3938485424289074e-05, + "loss": 0.9368, + "step": 31668 + }, + { + "epoch": 2.34, + "learning_rate": 1.393811878819486e-05, + "loss": 1.0458, + "step": 31669 + }, + { + "epoch": 2.34, + "learning_rate": 1.3937752145835148e-05, + "loss": 1.0189, + "step": 31670 + }, + { + "epoch": 2.34, + "learning_rate": 1.3937385497210536e-05, + "loss": 1.0344, + "step": 31671 + }, + { + "epoch": 2.34, + "learning_rate": 1.3937018842321594e-05, + "loss": 0.9812, + "step": 31672 + }, + { + "epoch": 2.34, + "learning_rate": 1.3936652181168914e-05, + "loss": 1.0564, + "step": 31673 + }, + { + "epoch": 2.34, + "learning_rate": 1.3936285513753077e-05, + "loss": 1.0622, + "step": 31674 + }, + { + "epoch": 2.34, + "learning_rate": 1.3935918840074666e-05, + "loss": 1.0445, + "step": 31675 + }, + { + "epoch": 2.34, + "learning_rate": 1.393555216013426e-05, + "loss": 1.0539, + "step": 31676 + }, + { + "epoch": 2.34, + "learning_rate": 1.3935185473932453e-05, + "loss": 0.9474, + "step": 31677 + }, + { + "epoch": 2.34, + "learning_rate": 1.3934818781469819e-05, + "loss": 0.9917, + "step": 31678 + }, + { + "epoch": 2.34, + "learning_rate": 1.3934452082746946e-05, + "loss": 1.0103, + "step": 31679 + }, + { + "epoch": 2.34, + "learning_rate": 1.3934085377764414e-05, + "loss": 1.1506, + "step": 31680 + }, + { + "epoch": 2.34, + "learning_rate": 1.3933718666522809e-05, + "loss": 1.0459, + "step": 31681 + }, + { + "epoch": 2.34, + "learning_rate": 1.3933351949022716e-05, + "loss": 0.9759, + "step": 31682 + }, + { + "epoch": 2.34, + "learning_rate": 1.3932985225264712e-05, + "loss": 1.1043, + "step": 31683 + }, + { + "epoch": 2.34, + "learning_rate": 1.393261849524939e-05, + "loss": 1.0678, + "step": 31684 + }, + { + "epoch": 2.34, + "learning_rate": 1.3932251758977326e-05, + "loss": 1.0182, + "step": 31685 + }, + { + "epoch": 2.34, + "learning_rate": 1.3931885016449106e-05, + "loss": 0.9013, + "step": 31686 + }, + { + "epoch": 2.34, + "learning_rate": 1.3931518267665315e-05, + "loss": 1.047, + "step": 31687 + }, + { + "epoch": 2.34, + "learning_rate": 1.393115151262653e-05, + "loss": 1.0685, + "step": 31688 + }, + { + "epoch": 2.34, + "learning_rate": 1.3930784751333342e-05, + "loss": 1.0923, + "step": 31689 + }, + { + "epoch": 2.34, + "learning_rate": 1.3930417983786333e-05, + "loss": 0.9593, + "step": 31690 + }, + { + "epoch": 2.34, + "learning_rate": 1.3930051209986081e-05, + "loss": 0.9569, + "step": 31691 + }, + { + "epoch": 2.34, + "learning_rate": 1.392968442993318e-05, + "loss": 0.9934, + "step": 31692 + }, + { + "epoch": 2.34, + "learning_rate": 1.3929317643628202e-05, + "loss": 1.0362, + "step": 31693 + }, + { + "epoch": 2.34, + "learning_rate": 1.3928950851071736e-05, + "loss": 0.9954, + "step": 31694 + }, + { + "epoch": 2.34, + "learning_rate": 1.3928584052264368e-05, + "loss": 1.0115, + "step": 31695 + }, + { + "epoch": 2.34, + "learning_rate": 1.3928217247206679e-05, + "loss": 0.9436, + "step": 31696 + }, + { + "epoch": 2.34, + "learning_rate": 1.392785043589925e-05, + "loss": 0.9325, + "step": 31697 + }, + { + "epoch": 2.34, + "learning_rate": 1.392748361834267e-05, + "loss": 0.9876, + "step": 31698 + }, + { + "epoch": 2.34, + "learning_rate": 1.392711679453752e-05, + "loss": 1.1043, + "step": 31699 + }, + { + "epoch": 2.34, + "learning_rate": 1.3926749964484381e-05, + "loss": 1.0334, + "step": 31700 + }, + { + "epoch": 2.34, + "learning_rate": 1.392638312818384e-05, + "loss": 1.0678, + "step": 31701 + }, + { + "epoch": 2.34, + "learning_rate": 1.3926016285636478e-05, + "loss": 1.0325, + "step": 31702 + }, + { + "epoch": 2.34, + "learning_rate": 1.3925649436842884e-05, + "loss": 1.0779, + "step": 31703 + }, + { + "epoch": 2.34, + "learning_rate": 1.3925282581803634e-05, + "loss": 0.976, + "step": 31704 + }, + { + "epoch": 2.34, + "learning_rate": 1.3924915720519316e-05, + "loss": 1.0312, + "step": 31705 + }, + { + "epoch": 2.34, + "learning_rate": 1.3924548852990516e-05, + "loss": 1.0189, + "step": 31706 + }, + { + "epoch": 2.34, + "learning_rate": 1.3924181979217812e-05, + "loss": 1.0338, + "step": 31707 + }, + { + "epoch": 2.34, + "learning_rate": 1.3923815099201793e-05, + "loss": 1.0303, + "step": 31708 + }, + { + "epoch": 2.34, + "learning_rate": 1.3923448212943039e-05, + "loss": 1.0095, + "step": 31709 + }, + { + "epoch": 2.34, + "learning_rate": 1.3923081320442137e-05, + "loss": 0.9507, + "step": 31710 + }, + { + "epoch": 2.34, + "learning_rate": 1.3922714421699666e-05, + "loss": 1.0828, + "step": 31711 + }, + { + "epoch": 2.34, + "learning_rate": 1.3922347516716214e-05, + "loss": 0.9886, + "step": 31712 + }, + { + "epoch": 2.34, + "learning_rate": 1.3921980605492363e-05, + "loss": 0.9901, + "step": 31713 + }, + { + "epoch": 2.34, + "learning_rate": 1.3921613688028696e-05, + "loss": 0.9828, + "step": 31714 + }, + { + "epoch": 2.34, + "learning_rate": 1.3921246764325798e-05, + "loss": 0.9757, + "step": 31715 + }, + { + "epoch": 2.34, + "learning_rate": 1.3920879834384253e-05, + "loss": 1.0148, + "step": 31716 + }, + { + "epoch": 2.34, + "learning_rate": 1.3920512898204643e-05, + "loss": 1.0423, + "step": 31717 + }, + { + "epoch": 2.34, + "learning_rate": 1.3920145955787558e-05, + "loss": 0.9849, + "step": 31718 + }, + { + "epoch": 2.34, + "learning_rate": 1.3919779007133572e-05, + "loss": 1.0149, + "step": 31719 + }, + { + "epoch": 2.34, + "learning_rate": 1.3919412052243275e-05, + "loss": 0.9536, + "step": 31720 + }, + { + "epoch": 2.34, + "learning_rate": 1.391904509111725e-05, + "loss": 1.0528, + "step": 31721 + }, + { + "epoch": 2.34, + "learning_rate": 1.391867812375608e-05, + "loss": 1.1292, + "step": 31722 + }, + { + "epoch": 2.34, + "learning_rate": 1.3918311150160346e-05, + "loss": 0.9469, + "step": 31723 + }, + { + "epoch": 2.34, + "learning_rate": 1.391794417033064e-05, + "loss": 0.973, + "step": 31724 + }, + { + "epoch": 2.34, + "learning_rate": 1.3917577184267536e-05, + "loss": 0.954, + "step": 31725 + }, + { + "epoch": 2.34, + "learning_rate": 1.3917210191971627e-05, + "loss": 0.9609, + "step": 31726 + }, + { + "epoch": 2.34, + "learning_rate": 1.3916843193443492e-05, + "loss": 0.9919, + "step": 31727 + }, + { + "epoch": 2.34, + "learning_rate": 1.3916476188683713e-05, + "loss": 0.9989, + "step": 31728 + }, + { + "epoch": 2.34, + "learning_rate": 1.391610917769288e-05, + "loss": 1.0046, + "step": 31729 + }, + { + "epoch": 2.34, + "learning_rate": 1.3915742160471572e-05, + "loss": 1.1163, + "step": 31730 + }, + { + "epoch": 2.34, + "learning_rate": 1.3915375137020373e-05, + "loss": 0.9901, + "step": 31731 + }, + { + "epoch": 2.34, + "learning_rate": 1.3915008107339869e-05, + "loss": 0.9679, + "step": 31732 + }, + { + "epoch": 2.34, + "learning_rate": 1.3914641071430644e-05, + "loss": 0.9049, + "step": 31733 + }, + { + "epoch": 2.34, + "learning_rate": 1.391427402929328e-05, + "loss": 1.159, + "step": 31734 + }, + { + "epoch": 2.34, + "learning_rate": 1.391390698092836e-05, + "loss": 0.9885, + "step": 31735 + }, + { + "epoch": 2.34, + "learning_rate": 1.3913539926336473e-05, + "loss": 0.9779, + "step": 31736 + }, + { + "epoch": 2.35, + "learning_rate": 1.39131728655182e-05, + "loss": 1.0792, + "step": 31737 + }, + { + "epoch": 2.35, + "learning_rate": 1.3912805798474123e-05, + "loss": 0.9252, + "step": 31738 + }, + { + "epoch": 2.35, + "learning_rate": 1.3912438725204828e-05, + "loss": 0.9585, + "step": 31739 + }, + { + "epoch": 2.35, + "learning_rate": 1.3912071645710901e-05, + "loss": 1.0201, + "step": 31740 + }, + { + "epoch": 2.35, + "learning_rate": 1.3911704559992923e-05, + "loss": 0.9466, + "step": 31741 + }, + { + "epoch": 2.35, + "learning_rate": 1.3911337468051479e-05, + "loss": 1.0293, + "step": 31742 + }, + { + "epoch": 2.35, + "learning_rate": 1.391097036988715e-05, + "loss": 0.9656, + "step": 31743 + }, + { + "epoch": 2.35, + "learning_rate": 1.391060326550053e-05, + "loss": 0.9258, + "step": 31744 + }, + { + "epoch": 2.35, + "learning_rate": 1.391023615489219e-05, + "loss": 1.0154, + "step": 31745 + }, + { + "epoch": 2.35, + "learning_rate": 1.3909869038062724e-05, + "loss": 0.9632, + "step": 31746 + }, + { + "epoch": 2.35, + "learning_rate": 1.390950191501271e-05, + "loss": 0.9132, + "step": 31747 + }, + { + "epoch": 2.35, + "learning_rate": 1.3909134785742734e-05, + "loss": 0.9526, + "step": 31748 + }, + { + "epoch": 2.35, + "learning_rate": 1.3908767650253383e-05, + "loss": 0.8765, + "step": 31749 + }, + { + "epoch": 2.35, + "learning_rate": 1.3908400508545236e-05, + "loss": 1.0574, + "step": 31750 + }, + { + "epoch": 2.35, + "learning_rate": 1.3908033360618882e-05, + "loss": 1.0155, + "step": 31751 + }, + { + "epoch": 2.35, + "learning_rate": 1.3907666206474902e-05, + "loss": 1.0664, + "step": 31752 + }, + { + "epoch": 2.35, + "learning_rate": 1.390729904611388e-05, + "loss": 0.8638, + "step": 31753 + }, + { + "epoch": 2.35, + "learning_rate": 1.39069318795364e-05, + "loss": 0.9078, + "step": 31754 + }, + { + "epoch": 2.35, + "learning_rate": 1.3906564706743053e-05, + "loss": 1.0182, + "step": 31755 + }, + { + "epoch": 2.35, + "learning_rate": 1.3906197527734412e-05, + "loss": 1.0288, + "step": 31756 + }, + { + "epoch": 2.35, + "learning_rate": 1.3905830342511071e-05, + "loss": 0.9832, + "step": 31757 + }, + { + "epoch": 2.35, + "learning_rate": 1.3905463151073605e-05, + "loss": 1.0813, + "step": 31758 + }, + { + "epoch": 2.35, + "learning_rate": 1.3905095953422604e-05, + "loss": 1.0605, + "step": 31759 + }, + { + "epoch": 2.35, + "learning_rate": 1.3904728749558651e-05, + "loss": 1.1238, + "step": 31760 + }, + { + "epoch": 2.35, + "learning_rate": 1.3904361539482333e-05, + "loss": 1.0095, + "step": 31761 + }, + { + "epoch": 2.35, + "learning_rate": 1.390399432319423e-05, + "loss": 0.9622, + "step": 31762 + }, + { + "epoch": 2.35, + "learning_rate": 1.390362710069493e-05, + "loss": 0.9594, + "step": 31763 + }, + { + "epoch": 2.35, + "learning_rate": 1.3903259871985014e-05, + "loss": 1.1422, + "step": 31764 + }, + { + "epoch": 2.35, + "learning_rate": 1.3902892637065067e-05, + "loss": 1.1032, + "step": 31765 + }, + { + "epoch": 2.35, + "learning_rate": 1.3902525395935671e-05, + "loss": 1.0221, + "step": 31766 + }, + { + "epoch": 2.35, + "learning_rate": 1.3902158148597418e-05, + "loss": 1.1393, + "step": 31767 + }, + { + "epoch": 2.35, + "learning_rate": 1.3901790895050881e-05, + "loss": 1.0865, + "step": 31768 + }, + { + "epoch": 2.35, + "learning_rate": 1.3901423635296655e-05, + "loss": 0.9128, + "step": 31769 + }, + { + "epoch": 2.35, + "learning_rate": 1.390105636933532e-05, + "loss": 0.9193, + "step": 31770 + }, + { + "epoch": 2.35, + "learning_rate": 1.390068909716746e-05, + "loss": 1.063, + "step": 31771 + }, + { + "epoch": 2.35, + "learning_rate": 1.3900321818793655e-05, + "loss": 0.9644, + "step": 31772 + }, + { + "epoch": 2.35, + "learning_rate": 1.38999545342145e-05, + "loss": 1.0866, + "step": 31773 + }, + { + "epoch": 2.35, + "learning_rate": 1.389958724343057e-05, + "loss": 0.9878, + "step": 31774 + }, + { + "epoch": 2.35, + "learning_rate": 1.3899219946442454e-05, + "loss": 0.8643, + "step": 31775 + }, + { + "epoch": 2.35, + "learning_rate": 1.3898852643250734e-05, + "loss": 1.0504, + "step": 31776 + }, + { + "epoch": 2.35, + "learning_rate": 1.3898485333855996e-05, + "loss": 1.0207, + "step": 31777 + }, + { + "epoch": 2.35, + "learning_rate": 1.389811801825882e-05, + "loss": 0.9999, + "step": 31778 + }, + { + "epoch": 2.35, + "learning_rate": 1.3897750696459797e-05, + "loss": 1.0125, + "step": 31779 + }, + { + "epoch": 2.35, + "learning_rate": 1.3897383368459507e-05, + "loss": 0.9515, + "step": 31780 + }, + { + "epoch": 2.35, + "learning_rate": 1.3897016034258539e-05, + "loss": 0.9105, + "step": 31781 + }, + { + "epoch": 2.35, + "learning_rate": 1.3896648693857469e-05, + "loss": 0.9914, + "step": 31782 + }, + { + "epoch": 2.35, + "learning_rate": 1.389628134725689e-05, + "loss": 1.0736, + "step": 31783 + }, + { + "epoch": 2.35, + "learning_rate": 1.3895913994457386e-05, + "loss": 0.993, + "step": 31784 + }, + { + "epoch": 2.35, + "learning_rate": 1.3895546635459533e-05, + "loss": 1.1079, + "step": 31785 + }, + { + "epoch": 2.35, + "learning_rate": 1.3895179270263925e-05, + "loss": 1.0443, + "step": 31786 + }, + { + "epoch": 2.35, + "learning_rate": 1.389481189887114e-05, + "loss": 0.9436, + "step": 31787 + }, + { + "epoch": 2.35, + "learning_rate": 1.3894444521281768e-05, + "loss": 0.9243, + "step": 31788 + }, + { + "epoch": 2.35, + "learning_rate": 1.3894077137496385e-05, + "loss": 0.9881, + "step": 31789 + }, + { + "epoch": 2.35, + "learning_rate": 1.3893709747515587e-05, + "loss": 1.0227, + "step": 31790 + }, + { + "epoch": 2.35, + "learning_rate": 1.3893342351339949e-05, + "loss": 1.0789, + "step": 31791 + }, + { + "epoch": 2.35, + "learning_rate": 1.3892974948970062e-05, + "loss": 0.9761, + "step": 31792 + }, + { + "epoch": 2.35, + "learning_rate": 1.3892607540406504e-05, + "loss": 1.0221, + "step": 31793 + }, + { + "epoch": 2.35, + "learning_rate": 1.3892240125649865e-05, + "loss": 0.9668, + "step": 31794 + }, + { + "epoch": 2.35, + "learning_rate": 1.3891872704700726e-05, + "loss": 1.0788, + "step": 31795 + }, + { + "epoch": 2.35, + "learning_rate": 1.3891505277559676e-05, + "loss": 1.0267, + "step": 31796 + }, + { + "epoch": 2.35, + "learning_rate": 1.3891137844227294e-05, + "loss": 1.0422, + "step": 31797 + }, + { + "epoch": 2.35, + "learning_rate": 1.3890770404704169e-05, + "loss": 0.9474, + "step": 31798 + }, + { + "epoch": 2.35, + "learning_rate": 1.3890402958990883e-05, + "loss": 1.0687, + "step": 31799 + }, + { + "epoch": 2.35, + "learning_rate": 1.3890035507088022e-05, + "loss": 0.8821, + "step": 31800 + }, + { + "epoch": 2.35, + "learning_rate": 1.388966804899617e-05, + "loss": 0.8279, + "step": 31801 + }, + { + "epoch": 2.35, + "learning_rate": 1.3889300584715915e-05, + "loss": 1.0254, + "step": 31802 + }, + { + "epoch": 2.35, + "learning_rate": 1.3888933114247833e-05, + "loss": 0.9922, + "step": 31803 + }, + { + "epoch": 2.35, + "learning_rate": 1.388856563759252e-05, + "loss": 0.8915, + "step": 31804 + }, + { + "epoch": 2.35, + "learning_rate": 1.388819815475055e-05, + "loss": 0.9163, + "step": 31805 + }, + { + "epoch": 2.35, + "learning_rate": 1.3887830665722512e-05, + "loss": 1.0835, + "step": 31806 + }, + { + "epoch": 2.35, + "learning_rate": 1.3887463170508992e-05, + "loss": 1.0725, + "step": 31807 + }, + { + "epoch": 2.35, + "learning_rate": 1.3887095669110576e-05, + "loss": 0.9966, + "step": 31808 + }, + { + "epoch": 2.35, + "learning_rate": 1.3886728161527843e-05, + "loss": 1.0479, + "step": 31809 + }, + { + "epoch": 2.35, + "learning_rate": 1.3886360647761382e-05, + "loss": 0.9502, + "step": 31810 + }, + { + "epoch": 2.35, + "learning_rate": 1.3885993127811781e-05, + "loss": 1.0537, + "step": 31811 + }, + { + "epoch": 2.35, + "learning_rate": 1.3885625601679618e-05, + "loss": 1.0398, + "step": 31812 + }, + { + "epoch": 2.35, + "learning_rate": 1.3885258069365477e-05, + "loss": 1.071, + "step": 31813 + }, + { + "epoch": 2.35, + "learning_rate": 1.388489053086995e-05, + "loss": 0.9718, + "step": 31814 + }, + { + "epoch": 2.35, + "learning_rate": 1.3884522986193616e-05, + "loss": 1.0555, + "step": 31815 + }, + { + "epoch": 2.35, + "learning_rate": 1.3884155435337061e-05, + "loss": 1.0646, + "step": 31816 + }, + { + "epoch": 2.35, + "learning_rate": 1.3883787878300869e-05, + "loss": 0.9854, + "step": 31817 + }, + { + "epoch": 2.35, + "learning_rate": 1.3883420315085633e-05, + "loss": 1.048, + "step": 31818 + }, + { + "epoch": 2.35, + "learning_rate": 1.3883052745691923e-05, + "loss": 1.0629, + "step": 31819 + }, + { + "epoch": 2.35, + "learning_rate": 1.3882685170120337e-05, + "loss": 1.1007, + "step": 31820 + }, + { + "epoch": 2.35, + "learning_rate": 1.388231758837145e-05, + "loss": 0.9642, + "step": 31821 + }, + { + "epoch": 2.35, + "learning_rate": 1.3881950000445855e-05, + "loss": 0.9583, + "step": 31822 + }, + { + "epoch": 2.35, + "learning_rate": 1.388158240634413e-05, + "loss": 0.9302, + "step": 31823 + }, + { + "epoch": 2.35, + "learning_rate": 1.3881214806066865e-05, + "loss": 0.9925, + "step": 31824 + }, + { + "epoch": 2.35, + "learning_rate": 1.388084719961464e-05, + "loss": 1.0644, + "step": 31825 + }, + { + "epoch": 2.35, + "learning_rate": 1.3880479586988047e-05, + "loss": 1.0174, + "step": 31826 + }, + { + "epoch": 2.35, + "learning_rate": 1.3880111968187662e-05, + "loss": 1.0796, + "step": 31827 + }, + { + "epoch": 2.35, + "learning_rate": 1.3879744343214076e-05, + "loss": 1.0367, + "step": 31828 + }, + { + "epoch": 2.35, + "learning_rate": 1.3879376712067873e-05, + "loss": 0.9893, + "step": 31829 + }, + { + "epoch": 2.35, + "learning_rate": 1.3879009074749638e-05, + "loss": 1.0729, + "step": 31830 + }, + { + "epoch": 2.35, + "learning_rate": 1.3878641431259956e-05, + "loss": 1.0177, + "step": 31831 + }, + { + "epoch": 2.35, + "learning_rate": 1.3878273781599407e-05, + "loss": 1.0539, + "step": 31832 + }, + { + "epoch": 2.35, + "learning_rate": 1.3877906125768583e-05, + "loss": 1.053, + "step": 31833 + }, + { + "epoch": 2.35, + "learning_rate": 1.3877538463768065e-05, + "loss": 0.9733, + "step": 31834 + }, + { + "epoch": 2.35, + "learning_rate": 1.3877170795598438e-05, + "loss": 0.9687, + "step": 31835 + }, + { + "epoch": 2.35, + "learning_rate": 1.3876803121260289e-05, + "loss": 0.904, + "step": 31836 + }, + { + "epoch": 2.35, + "learning_rate": 1.38764354407542e-05, + "loss": 1.0086, + "step": 31837 + }, + { + "epoch": 2.35, + "learning_rate": 1.387606775408076e-05, + "loss": 1.0593, + "step": 31838 + }, + { + "epoch": 2.35, + "learning_rate": 1.3875700061240552e-05, + "loss": 1.0081, + "step": 31839 + }, + { + "epoch": 2.35, + "learning_rate": 1.3875332362234155e-05, + "loss": 0.9749, + "step": 31840 + }, + { + "epoch": 2.35, + "learning_rate": 1.3874964657062166e-05, + "loss": 1.0096, + "step": 31841 + }, + { + "epoch": 2.35, + "learning_rate": 1.387459694572516e-05, + "loss": 1.008, + "step": 31842 + }, + { + "epoch": 2.35, + "learning_rate": 1.3874229228223728e-05, + "loss": 1.0128, + "step": 31843 + }, + { + "epoch": 2.35, + "learning_rate": 1.3873861504558451e-05, + "loss": 1.028, + "step": 31844 + }, + { + "epoch": 2.35, + "learning_rate": 1.3873493774729916e-05, + "loss": 1.0766, + "step": 31845 + }, + { + "epoch": 2.35, + "learning_rate": 1.3873126038738707e-05, + "loss": 0.9504, + "step": 31846 + }, + { + "epoch": 2.35, + "learning_rate": 1.3872758296585414e-05, + "loss": 0.9637, + "step": 31847 + }, + { + "epoch": 2.35, + "learning_rate": 1.3872390548270616e-05, + "loss": 0.9877, + "step": 31848 + }, + { + "epoch": 2.35, + "learning_rate": 1.38720227937949e-05, + "loss": 1.0133, + "step": 31849 + }, + { + "epoch": 2.35, + "learning_rate": 1.3871655033158848e-05, + "loss": 0.9258, + "step": 31850 + }, + { + "epoch": 2.35, + "learning_rate": 1.3871287266363051e-05, + "loss": 0.9384, + "step": 31851 + }, + { + "epoch": 2.35, + "learning_rate": 1.3870919493408091e-05, + "loss": 1.0467, + "step": 31852 + }, + { + "epoch": 2.35, + "learning_rate": 1.3870551714294553e-05, + "loss": 0.9011, + "step": 31853 + }, + { + "epoch": 2.35, + "learning_rate": 1.3870183929023023e-05, + "loss": 0.9833, + "step": 31854 + }, + { + "epoch": 2.35, + "learning_rate": 1.3869816137594086e-05, + "loss": 1.0336, + "step": 31855 + }, + { + "epoch": 2.35, + "learning_rate": 1.3869448340008326e-05, + "loss": 1.0402, + "step": 31856 + }, + { + "epoch": 2.35, + "learning_rate": 1.3869080536266331e-05, + "loss": 1.001, + "step": 31857 + }, + { + "epoch": 2.35, + "learning_rate": 1.3868712726368683e-05, + "loss": 0.9932, + "step": 31858 + }, + { + "epoch": 2.35, + "learning_rate": 1.3868344910315969e-05, + "loss": 0.8619, + "step": 31859 + }, + { + "epoch": 2.35, + "learning_rate": 1.3867977088108773e-05, + "loss": 0.9351, + "step": 31860 + }, + { + "epoch": 2.35, + "learning_rate": 1.386760925974768e-05, + "loss": 0.9265, + "step": 31861 + }, + { + "epoch": 2.35, + "learning_rate": 1.3867241425233276e-05, + "loss": 0.9022, + "step": 31862 + }, + { + "epoch": 2.35, + "learning_rate": 1.3866873584566147e-05, + "loss": 1.0551, + "step": 31863 + }, + { + "epoch": 2.35, + "learning_rate": 1.3866505737746878e-05, + "loss": 1.029, + "step": 31864 + }, + { + "epoch": 2.35, + "learning_rate": 1.3866137884776051e-05, + "loss": 0.9248, + "step": 31865 + }, + { + "epoch": 2.35, + "learning_rate": 1.3865770025654258e-05, + "loss": 1.0193, + "step": 31866 + }, + { + "epoch": 2.35, + "learning_rate": 1.386540216038208e-05, + "loss": 0.9881, + "step": 31867 + }, + { + "epoch": 2.35, + "learning_rate": 1.38650342889601e-05, + "loss": 1.0676, + "step": 31868 + }, + { + "epoch": 2.35, + "learning_rate": 1.3864666411388907e-05, + "loss": 1.0066, + "step": 31869 + }, + { + "epoch": 2.35, + "learning_rate": 1.3864298527669083e-05, + "loss": 1.0674, + "step": 31870 + }, + { + "epoch": 2.35, + "learning_rate": 1.3863930637801217e-05, + "loss": 1.1185, + "step": 31871 + }, + { + "epoch": 2.36, + "learning_rate": 1.3863562741785892e-05, + "loss": 1.0198, + "step": 31872 + }, + { + "epoch": 2.36, + "learning_rate": 1.3863194839623694e-05, + "loss": 1.0156, + "step": 31873 + }, + { + "epoch": 2.36, + "learning_rate": 1.3862826931315207e-05, + "loss": 0.9915, + "step": 31874 + }, + { + "epoch": 2.36, + "learning_rate": 1.386245901686102e-05, + "loss": 1.0422, + "step": 31875 + }, + { + "epoch": 2.36, + "learning_rate": 1.3862091096261718e-05, + "loss": 1.0448, + "step": 31876 + }, + { + "epoch": 2.36, + "learning_rate": 1.386172316951788e-05, + "loss": 1.0244, + "step": 31877 + }, + { + "epoch": 2.36, + "learning_rate": 1.3861355236630098e-05, + "loss": 1.049, + "step": 31878 + }, + { + "epoch": 2.36, + "learning_rate": 1.3860987297598956e-05, + "loss": 0.9975, + "step": 31879 + }, + { + "epoch": 2.36, + "learning_rate": 1.3860619352425035e-05, + "loss": 1.0135, + "step": 31880 + }, + { + "epoch": 2.36, + "learning_rate": 1.3860251401108926e-05, + "loss": 1.0556, + "step": 31881 + }, + { + "epoch": 2.36, + "learning_rate": 1.3859883443651212e-05, + "loss": 1.021, + "step": 31882 + }, + { + "epoch": 2.36, + "learning_rate": 1.3859515480052478e-05, + "loss": 1.0215, + "step": 31883 + }, + { + "epoch": 2.36, + "learning_rate": 1.3859147510313311e-05, + "loss": 1.0194, + "step": 31884 + }, + { + "epoch": 2.36, + "learning_rate": 1.3858779534434294e-05, + "loss": 1.0736, + "step": 31885 + }, + { + "epoch": 2.36, + "learning_rate": 1.3858411552416017e-05, + "loss": 1.0466, + "step": 31886 + }, + { + "epoch": 2.36, + "learning_rate": 1.3858043564259061e-05, + "loss": 0.9605, + "step": 31887 + }, + { + "epoch": 2.36, + "learning_rate": 1.385767556996401e-05, + "loss": 0.9935, + "step": 31888 + }, + { + "epoch": 2.36, + "learning_rate": 1.3857307569531455e-05, + "loss": 1.023, + "step": 31889 + }, + { + "epoch": 2.36, + "learning_rate": 1.3856939562961979e-05, + "loss": 1.0792, + "step": 31890 + }, + { + "epoch": 2.36, + "learning_rate": 1.3856571550256166e-05, + "loss": 1.0253, + "step": 31891 + }, + { + "epoch": 2.36, + "learning_rate": 1.3856203531414605e-05, + "loss": 0.9306, + "step": 31892 + }, + { + "epoch": 2.36, + "learning_rate": 1.3855835506437878e-05, + "loss": 1.0061, + "step": 31893 + }, + { + "epoch": 2.36, + "learning_rate": 1.3855467475326573e-05, + "loss": 1.0529, + "step": 31894 + }, + { + "epoch": 2.36, + "learning_rate": 1.3855099438081274e-05, + "loss": 1.0297, + "step": 31895 + }, + { + "epoch": 2.36, + "learning_rate": 1.3854731394702567e-05, + "loss": 0.9012, + "step": 31896 + }, + { + "epoch": 2.36, + "learning_rate": 1.3854363345191035e-05, + "loss": 1.0589, + "step": 31897 + }, + { + "epoch": 2.36, + "learning_rate": 1.385399528954727e-05, + "loss": 0.9228, + "step": 31898 + }, + { + "epoch": 2.36, + "learning_rate": 1.3853627227771852e-05, + "loss": 0.9761, + "step": 31899 + }, + { + "epoch": 2.36, + "learning_rate": 1.3853259159865366e-05, + "loss": 0.9629, + "step": 31900 + }, + { + "epoch": 2.36, + "learning_rate": 1.3852891085828401e-05, + "loss": 1.0681, + "step": 31901 + }, + { + "epoch": 2.36, + "learning_rate": 1.3852523005661544e-05, + "loss": 0.8935, + "step": 31902 + }, + { + "epoch": 2.36, + "learning_rate": 1.3852154919365375e-05, + "loss": 1.0937, + "step": 31903 + }, + { + "epoch": 2.36, + "learning_rate": 1.3851786826940487e-05, + "loss": 0.9985, + "step": 31904 + }, + { + "epoch": 2.36, + "learning_rate": 1.3851418728387455e-05, + "loss": 0.9237, + "step": 31905 + }, + { + "epoch": 2.36, + "learning_rate": 1.3851050623706876e-05, + "loss": 1.1212, + "step": 31906 + }, + { + "epoch": 2.36, + "learning_rate": 1.3850682512899327e-05, + "loss": 0.9267, + "step": 31907 + }, + { + "epoch": 2.36, + "learning_rate": 1.3850314395965399e-05, + "loss": 1.0591, + "step": 31908 + }, + { + "epoch": 2.36, + "learning_rate": 1.3849946272905674e-05, + "loss": 1.0176, + "step": 31909 + }, + { + "epoch": 2.36, + "learning_rate": 1.3849578143720744e-05, + "loss": 0.9972, + "step": 31910 + }, + { + "epoch": 2.36, + "learning_rate": 1.3849210008411185e-05, + "loss": 1.0115, + "step": 31911 + }, + { + "epoch": 2.36, + "learning_rate": 1.384884186697759e-05, + "loss": 0.9283, + "step": 31912 + }, + { + "epoch": 2.36, + "learning_rate": 1.3848473719420543e-05, + "loss": 0.9321, + "step": 31913 + }, + { + "epoch": 2.36, + "learning_rate": 1.3848105565740633e-05, + "loss": 1.0983, + "step": 31914 + }, + { + "epoch": 2.36, + "learning_rate": 1.3847737405938437e-05, + "loss": 0.9848, + "step": 31915 + }, + { + "epoch": 2.36, + "learning_rate": 1.3847369240014547e-05, + "loss": 1.0848, + "step": 31916 + }, + { + "epoch": 2.36, + "learning_rate": 1.3847001067969548e-05, + "loss": 0.9241, + "step": 31917 + }, + { + "epoch": 2.36, + "learning_rate": 1.3846632889804026e-05, + "loss": 0.9569, + "step": 31918 + }, + { + "epoch": 2.36, + "learning_rate": 1.3846264705518565e-05, + "loss": 1.0188, + "step": 31919 + }, + { + "epoch": 2.36, + "learning_rate": 1.3845896515113752e-05, + "loss": 0.9709, + "step": 31920 + }, + { + "epoch": 2.36, + "learning_rate": 1.384552831859017e-05, + "loss": 0.9316, + "step": 31921 + }, + { + "epoch": 2.36, + "learning_rate": 1.3845160115948412e-05, + "loss": 0.982, + "step": 31922 + }, + { + "epoch": 2.36, + "learning_rate": 1.384479190718906e-05, + "loss": 0.8777, + "step": 31923 + }, + { + "epoch": 2.36, + "learning_rate": 1.3844423692312697e-05, + "loss": 1.0215, + "step": 31924 + }, + { + "epoch": 2.36, + "learning_rate": 1.384405547131991e-05, + "loss": 1.0051, + "step": 31925 + }, + { + "epoch": 2.36, + "learning_rate": 1.3843687244211288e-05, + "loss": 0.9357, + "step": 31926 + }, + { + "epoch": 2.36, + "learning_rate": 1.384331901098741e-05, + "loss": 0.9274, + "step": 31927 + }, + { + "epoch": 2.36, + "learning_rate": 1.3842950771648871e-05, + "loss": 0.949, + "step": 31928 + }, + { + "epoch": 2.36, + "learning_rate": 1.3842582526196248e-05, + "loss": 1.0031, + "step": 31929 + }, + { + "epoch": 2.36, + "learning_rate": 1.3842214274630138e-05, + "loss": 0.9338, + "step": 31930 + }, + { + "epoch": 2.36, + "learning_rate": 1.3841846016951112e-05, + "loss": 0.9331, + "step": 31931 + }, + { + "epoch": 2.36, + "learning_rate": 1.3841477753159769e-05, + "loss": 1.0665, + "step": 31932 + }, + { + "epoch": 2.36, + "learning_rate": 1.384110948325669e-05, + "loss": 1.0542, + "step": 31933 + }, + { + "epoch": 2.36, + "learning_rate": 1.3840741207242458e-05, + "loss": 0.9732, + "step": 31934 + }, + { + "epoch": 2.36, + "learning_rate": 1.3840372925117664e-05, + "loss": 0.9795, + "step": 31935 + }, + { + "epoch": 2.36, + "learning_rate": 1.3840004636882891e-05, + "loss": 1.0464, + "step": 31936 + }, + { + "epoch": 2.36, + "learning_rate": 1.3839636342538727e-05, + "loss": 1.0583, + "step": 31937 + }, + { + "epoch": 2.36, + "learning_rate": 1.3839268042085752e-05, + "loss": 0.9855, + "step": 31938 + }, + { + "epoch": 2.36, + "learning_rate": 1.3838899735524561e-05, + "loss": 0.9296, + "step": 31939 + }, + { + "epoch": 2.36, + "learning_rate": 1.3838531422855735e-05, + "loss": 0.9992, + "step": 31940 + }, + { + "epoch": 2.36, + "learning_rate": 1.3838163104079857e-05, + "loss": 1.0658, + "step": 31941 + }, + { + "epoch": 2.36, + "learning_rate": 1.383779477919752e-05, + "loss": 0.9324, + "step": 31942 + }, + { + "epoch": 2.36, + "learning_rate": 1.3837426448209305e-05, + "loss": 1.038, + "step": 31943 + }, + { + "epoch": 2.36, + "learning_rate": 1.38370581111158e-05, + "loss": 0.9679, + "step": 31944 + }, + { + "epoch": 2.36, + "learning_rate": 1.383668976791759e-05, + "loss": 1.0852, + "step": 31945 + }, + { + "epoch": 2.36, + "learning_rate": 1.3836321418615262e-05, + "loss": 0.9274, + "step": 31946 + }, + { + "epoch": 2.36, + "learning_rate": 1.3835953063209399e-05, + "loss": 1.0698, + "step": 31947 + }, + { + "epoch": 2.36, + "learning_rate": 1.3835584701700592e-05, + "loss": 0.9832, + "step": 31948 + }, + { + "epoch": 2.36, + "learning_rate": 1.3835216334089425e-05, + "loss": 1.0624, + "step": 31949 + }, + { + "epoch": 2.36, + "learning_rate": 1.3834847960376481e-05, + "loss": 0.9901, + "step": 31950 + }, + { + "epoch": 2.36, + "learning_rate": 1.3834479580562354e-05, + "loss": 1.0795, + "step": 31951 + }, + { + "epoch": 2.36, + "learning_rate": 1.383411119464762e-05, + "loss": 1.0208, + "step": 31952 + }, + { + "epoch": 2.36, + "learning_rate": 1.3833742802632872e-05, + "loss": 0.9078, + "step": 31953 + }, + { + "epoch": 2.36, + "learning_rate": 1.3833374404518693e-05, + "loss": 1.0256, + "step": 31954 + }, + { + "epoch": 2.36, + "learning_rate": 1.3833006000305671e-05, + "loss": 0.8685, + "step": 31955 + }, + { + "epoch": 2.36, + "learning_rate": 1.3832637589994389e-05, + "loss": 1.1088, + "step": 31956 + }, + { + "epoch": 2.36, + "learning_rate": 1.383226917358544e-05, + "loss": 1.0506, + "step": 31957 + }, + { + "epoch": 2.36, + "learning_rate": 1.38319007510794e-05, + "loss": 0.8654, + "step": 31958 + }, + { + "epoch": 2.36, + "learning_rate": 1.3831532322476867e-05, + "loss": 0.9847, + "step": 31959 + }, + { + "epoch": 2.36, + "learning_rate": 1.3831163887778415e-05, + "loss": 1.0468, + "step": 31960 + }, + { + "epoch": 2.36, + "learning_rate": 1.3830795446984642e-05, + "loss": 1.001, + "step": 31961 + }, + { + "epoch": 2.36, + "learning_rate": 1.3830427000096122e-05, + "loss": 1.0478, + "step": 31962 + }, + { + "epoch": 2.36, + "learning_rate": 1.3830058547113451e-05, + "loss": 1.0027, + "step": 31963 + }, + { + "epoch": 2.36, + "learning_rate": 1.3829690088037211e-05, + "loss": 0.9656, + "step": 31964 + }, + { + "epoch": 2.36, + "learning_rate": 1.382932162286799e-05, + "loss": 0.9033, + "step": 31965 + }, + { + "epoch": 2.36, + "learning_rate": 1.382895315160637e-05, + "loss": 1.0312, + "step": 31966 + }, + { + "epoch": 2.36, + "learning_rate": 1.3828584674252944e-05, + "loss": 1.0217, + "step": 31967 + }, + { + "epoch": 2.36, + "learning_rate": 1.3828216190808293e-05, + "loss": 0.906, + "step": 31968 + }, + { + "epoch": 2.36, + "learning_rate": 1.3827847701273002e-05, + "loss": 1.0249, + "step": 31969 + }, + { + "epoch": 2.36, + "learning_rate": 1.3827479205647665e-05, + "loss": 0.9657, + "step": 31970 + }, + { + "epoch": 2.36, + "learning_rate": 1.3827110703932862e-05, + "loss": 1.0165, + "step": 31971 + }, + { + "epoch": 2.36, + "learning_rate": 1.3826742196129179e-05, + "loss": 0.9706, + "step": 31972 + }, + { + "epoch": 2.36, + "learning_rate": 1.3826373682237203e-05, + "loss": 1.0199, + "step": 31973 + }, + { + "epoch": 2.36, + "learning_rate": 1.3826005162257523e-05, + "loss": 1.0858, + "step": 31974 + }, + { + "epoch": 2.36, + "learning_rate": 1.3825636636190724e-05, + "loss": 0.9239, + "step": 31975 + }, + { + "epoch": 2.36, + "learning_rate": 1.382526810403739e-05, + "loss": 0.9889, + "step": 31976 + }, + { + "epoch": 2.36, + "learning_rate": 1.3824899565798111e-05, + "loss": 0.93, + "step": 31977 + }, + { + "epoch": 2.36, + "learning_rate": 1.382453102147347e-05, + "loss": 1.0887, + "step": 31978 + }, + { + "epoch": 2.36, + "learning_rate": 1.3824162471064058e-05, + "loss": 1.0112, + "step": 31979 + }, + { + "epoch": 2.36, + "learning_rate": 1.3823793914570456e-05, + "loss": 1.0365, + "step": 31980 + }, + { + "epoch": 2.36, + "learning_rate": 1.3823425351993251e-05, + "loss": 1.0439, + "step": 31981 + }, + { + "epoch": 2.36, + "learning_rate": 1.3823056783333033e-05, + "loss": 1.0069, + "step": 31982 + }, + { + "epoch": 2.36, + "learning_rate": 1.3822688208590387e-05, + "loss": 1.1171, + "step": 31983 + }, + { + "epoch": 2.36, + "learning_rate": 1.3822319627765897e-05, + "loss": 1.0048, + "step": 31984 + }, + { + "epoch": 2.36, + "learning_rate": 1.3821951040860153e-05, + "loss": 0.9389, + "step": 31985 + }, + { + "epoch": 2.36, + "learning_rate": 1.3821582447873735e-05, + "loss": 1.0549, + "step": 31986 + }, + { + "epoch": 2.36, + "learning_rate": 1.382121384880724e-05, + "loss": 0.9516, + "step": 31987 + }, + { + "epoch": 2.36, + "learning_rate": 1.3820845243661245e-05, + "loss": 1.0029, + "step": 31988 + }, + { + "epoch": 2.36, + "learning_rate": 1.3820476632436342e-05, + "loss": 0.9057, + "step": 31989 + }, + { + "epoch": 2.36, + "learning_rate": 1.3820108015133113e-05, + "loss": 1.0947, + "step": 31990 + }, + { + "epoch": 2.36, + "learning_rate": 1.3819739391752148e-05, + "loss": 1.0127, + "step": 31991 + }, + { + "epoch": 2.36, + "learning_rate": 1.3819370762294033e-05, + "loss": 0.9858, + "step": 31992 + }, + { + "epoch": 2.36, + "learning_rate": 1.3819002126759351e-05, + "loss": 0.9926, + "step": 31993 + }, + { + "epoch": 2.36, + "learning_rate": 1.3818633485148694e-05, + "loss": 1.0121, + "step": 31994 + }, + { + "epoch": 2.36, + "learning_rate": 1.3818264837462645e-05, + "loss": 1.0389, + "step": 31995 + }, + { + "epoch": 2.36, + "learning_rate": 1.3817896183701791e-05, + "loss": 0.9715, + "step": 31996 + }, + { + "epoch": 2.36, + "learning_rate": 1.3817527523866718e-05, + "loss": 0.952, + "step": 31997 + }, + { + "epoch": 2.36, + "learning_rate": 1.3817158857958018e-05, + "loss": 1.057, + "step": 31998 + }, + { + "epoch": 2.36, + "learning_rate": 1.3816790185976267e-05, + "loss": 0.9611, + "step": 31999 + }, + { + "epoch": 2.36, + "learning_rate": 1.381642150792206e-05, + "loss": 1.0037, + "step": 32000 + }, + { + "epoch": 2.36, + "learning_rate": 1.381605282379598e-05, + "loss": 1.0261, + "step": 32001 + }, + { + "epoch": 2.36, + "learning_rate": 1.3815684133598617e-05, + "loss": 0.9733, + "step": 32002 + }, + { + "epoch": 2.36, + "learning_rate": 1.381531543733055e-05, + "loss": 0.9856, + "step": 32003 + }, + { + "epoch": 2.36, + "learning_rate": 1.3814946734992375e-05, + "loss": 0.9362, + "step": 32004 + }, + { + "epoch": 2.36, + "learning_rate": 1.3814578026584672e-05, + "loss": 1.0608, + "step": 32005 + }, + { + "epoch": 2.36, + "learning_rate": 1.3814209312108033e-05, + "loss": 0.8748, + "step": 32006 + }, + { + "epoch": 2.36, + "learning_rate": 1.381384059156304e-05, + "loss": 0.9821, + "step": 32007 + }, + { + "epoch": 2.37, + "learning_rate": 1.3813471864950282e-05, + "loss": 1.0279, + "step": 32008 + }, + { + "epoch": 2.37, + "learning_rate": 1.3813103132270342e-05, + "loss": 0.9608, + "step": 32009 + }, + { + "epoch": 2.37, + "learning_rate": 1.3812734393523815e-05, + "loss": 1.0425, + "step": 32010 + }, + { + "epoch": 2.37, + "learning_rate": 1.3812365648711276e-05, + "loss": 0.9993, + "step": 32011 + }, + { + "epoch": 2.37, + "learning_rate": 1.381199689783332e-05, + "loss": 0.9835, + "step": 32012 + }, + { + "epoch": 2.37, + "learning_rate": 1.3811628140890532e-05, + "loss": 1.0558, + "step": 32013 + }, + { + "epoch": 2.37, + "learning_rate": 1.3811259377883497e-05, + "loss": 1.1176, + "step": 32014 + }, + { + "epoch": 2.37, + "learning_rate": 1.3810890608812804e-05, + "loss": 0.9981, + "step": 32015 + }, + { + "epoch": 2.37, + "learning_rate": 1.3810521833679038e-05, + "loss": 1.0127, + "step": 32016 + }, + { + "epoch": 2.37, + "learning_rate": 1.3810153052482788e-05, + "loss": 1.0082, + "step": 32017 + }, + { + "epoch": 2.37, + "learning_rate": 1.380978426522464e-05, + "loss": 1.0448, + "step": 32018 + }, + { + "epoch": 2.37, + "learning_rate": 1.3809415471905173e-05, + "loss": 1.1561, + "step": 32019 + }, + { + "epoch": 2.37, + "learning_rate": 1.3809046672524989e-05, + "loss": 0.9572, + "step": 32020 + }, + { + "epoch": 2.37, + "learning_rate": 1.3808677867084659e-05, + "loss": 1.0142, + "step": 32021 + }, + { + "epoch": 2.37, + "learning_rate": 1.3808309055584781e-05, + "loss": 1.0859, + "step": 32022 + }, + { + "epoch": 2.37, + "learning_rate": 1.3807940238025936e-05, + "loss": 1.0394, + "step": 32023 + }, + { + "epoch": 2.37, + "learning_rate": 1.3807571414408713e-05, + "loss": 0.9832, + "step": 32024 + }, + { + "epoch": 2.37, + "learning_rate": 1.3807202584733699e-05, + "loss": 1.0754, + "step": 32025 + }, + { + "epoch": 2.37, + "learning_rate": 1.380683374900148e-05, + "loss": 0.9678, + "step": 32026 + }, + { + "epoch": 2.37, + "learning_rate": 1.3806464907212645e-05, + "loss": 1.0521, + "step": 32027 + }, + { + "epoch": 2.37, + "learning_rate": 1.3806096059367778e-05, + "loss": 1.076, + "step": 32028 + }, + { + "epoch": 2.37, + "learning_rate": 1.3805727205467463e-05, + "loss": 0.9834, + "step": 32029 + }, + { + "epoch": 2.37, + "learning_rate": 1.3805358345512295e-05, + "loss": 1.032, + "step": 32030 + }, + { + "epoch": 2.37, + "learning_rate": 1.3804989479502853e-05, + "loss": 1.0533, + "step": 32031 + }, + { + "epoch": 2.37, + "learning_rate": 1.3804620607439727e-05, + "loss": 1.095, + "step": 32032 + }, + { + "epoch": 2.37, + "learning_rate": 1.3804251729323506e-05, + "loss": 1.0541, + "step": 32033 + }, + { + "epoch": 2.37, + "learning_rate": 1.3803882845154775e-05, + "loss": 0.9681, + "step": 32034 + }, + { + "epoch": 2.37, + "learning_rate": 1.380351395493412e-05, + "loss": 0.9724, + "step": 32035 + }, + { + "epoch": 2.37, + "learning_rate": 1.3803145058662129e-05, + "loss": 0.9659, + "step": 32036 + }, + { + "epoch": 2.37, + "learning_rate": 1.3802776156339389e-05, + "loss": 0.9263, + "step": 32037 + }, + { + "epoch": 2.37, + "learning_rate": 1.3802407247966482e-05, + "loss": 0.9892, + "step": 32038 + }, + { + "epoch": 2.37, + "learning_rate": 1.3802038333544005e-05, + "loss": 0.9626, + "step": 32039 + }, + { + "epoch": 2.37, + "learning_rate": 1.3801669413072538e-05, + "loss": 1.0099, + "step": 32040 + }, + { + "epoch": 2.37, + "learning_rate": 1.3801300486552668e-05, + "loss": 1.058, + "step": 32041 + }, + { + "epoch": 2.37, + "learning_rate": 1.3800931553984984e-05, + "loss": 1.002, + "step": 32042 + }, + { + "epoch": 2.37, + "learning_rate": 1.3800562615370074e-05, + "loss": 1.0168, + "step": 32043 + }, + { + "epoch": 2.37, + "learning_rate": 1.3800193670708524e-05, + "loss": 0.9778, + "step": 32044 + }, + { + "epoch": 2.37, + "learning_rate": 1.3799824720000918e-05, + "loss": 0.9701, + "step": 32045 + }, + { + "epoch": 2.37, + "learning_rate": 1.3799455763247844e-05, + "loss": 1.0571, + "step": 32046 + }, + { + "epoch": 2.37, + "learning_rate": 1.3799086800449892e-05, + "loss": 1.032, + "step": 32047 + }, + { + "epoch": 2.37, + "learning_rate": 1.379871783160765e-05, + "loss": 0.9081, + "step": 32048 + }, + { + "epoch": 2.37, + "learning_rate": 1.3798348856721697e-05, + "loss": 1.0158, + "step": 32049 + }, + { + "epoch": 2.37, + "learning_rate": 1.3797979875792626e-05, + "loss": 1.1202, + "step": 32050 + }, + { + "epoch": 2.37, + "learning_rate": 1.3797610888821027e-05, + "loss": 0.9579, + "step": 32051 + }, + { + "epoch": 2.37, + "learning_rate": 1.379724189580748e-05, + "loss": 1.0522, + "step": 32052 + }, + { + "epoch": 2.37, + "learning_rate": 1.3796872896752579e-05, + "loss": 1.1079, + "step": 32053 + }, + { + "epoch": 2.37, + "learning_rate": 1.3796503891656905e-05, + "loss": 1.0434, + "step": 32054 + }, + { + "epoch": 2.37, + "learning_rate": 1.3796134880521048e-05, + "loss": 1.0362, + "step": 32055 + }, + { + "epoch": 2.37, + "learning_rate": 1.3795765863345594e-05, + "loss": 0.9573, + "step": 32056 + }, + { + "epoch": 2.37, + "learning_rate": 1.3795396840131132e-05, + "loss": 1.0831, + "step": 32057 + }, + { + "epoch": 2.37, + "learning_rate": 1.3795027810878247e-05, + "loss": 0.9546, + "step": 32058 + }, + { + "epoch": 2.37, + "learning_rate": 1.3794658775587528e-05, + "loss": 1.0376, + "step": 32059 + }, + { + "epoch": 2.37, + "learning_rate": 1.3794289734259561e-05, + "loss": 1.0067, + "step": 32060 + }, + { + "epoch": 2.37, + "learning_rate": 1.3793920686894933e-05, + "loss": 0.9816, + "step": 32061 + }, + { + "epoch": 2.37, + "learning_rate": 1.379355163349423e-05, + "loss": 1.0349, + "step": 32062 + }, + { + "epoch": 2.37, + "learning_rate": 1.3793182574058046e-05, + "loss": 0.8746, + "step": 32063 + }, + { + "epoch": 2.37, + "learning_rate": 1.3792813508586957e-05, + "loss": 1.0275, + "step": 32064 + }, + { + "epoch": 2.37, + "learning_rate": 1.379244443708156e-05, + "loss": 1.0319, + "step": 32065 + }, + { + "epoch": 2.37, + "learning_rate": 1.3792075359542435e-05, + "loss": 1.0639, + "step": 32066 + }, + { + "epoch": 2.37, + "learning_rate": 1.3791706275970173e-05, + "loss": 0.9618, + "step": 32067 + }, + { + "epoch": 2.37, + "learning_rate": 1.379133718636536e-05, + "loss": 0.9049, + "step": 32068 + }, + { + "epoch": 2.37, + "learning_rate": 1.3790968090728587e-05, + "loss": 1.0038, + "step": 32069 + }, + { + "epoch": 2.37, + "learning_rate": 1.3790598989060433e-05, + "loss": 1.0364, + "step": 32070 + }, + { + "epoch": 2.37, + "learning_rate": 1.3790229881361495e-05, + "loss": 0.9227, + "step": 32071 + }, + { + "epoch": 2.37, + "learning_rate": 1.3789860767632352e-05, + "loss": 0.9869, + "step": 32072 + }, + { + "epoch": 2.37, + "learning_rate": 1.3789491647873598e-05, + "loss": 0.9707, + "step": 32073 + }, + { + "epoch": 2.37, + "learning_rate": 1.3789122522085815e-05, + "loss": 1.0295, + "step": 32074 + }, + { + "epoch": 2.37, + "learning_rate": 1.3788753390269594e-05, + "loss": 0.947, + "step": 32075 + }, + { + "epoch": 2.37, + "learning_rate": 1.3788384252425518e-05, + "loss": 1.0945, + "step": 32076 + }, + { + "epoch": 2.37, + "learning_rate": 1.3788015108554177e-05, + "loss": 1.0298, + "step": 32077 + }, + { + "epoch": 2.37, + "learning_rate": 1.378764595865616e-05, + "loss": 0.9304, + "step": 32078 + }, + { + "epoch": 2.37, + "learning_rate": 1.378727680273205e-05, + "loss": 0.953, + "step": 32079 + }, + { + "epoch": 2.37, + "learning_rate": 1.3786907640782438e-05, + "loss": 1.0142, + "step": 32080 + }, + { + "epoch": 2.37, + "learning_rate": 1.3786538472807913e-05, + "loss": 0.982, + "step": 32081 + }, + { + "epoch": 2.37, + "learning_rate": 1.3786169298809058e-05, + "loss": 1.0655, + "step": 32082 + }, + { + "epoch": 2.37, + "learning_rate": 1.3785800118786458e-05, + "loss": 0.9555, + "step": 32083 + }, + { + "epoch": 2.37, + "learning_rate": 1.3785430932740708e-05, + "loss": 0.8943, + "step": 32084 + }, + { + "epoch": 2.37, + "learning_rate": 1.378506174067239e-05, + "loss": 1.03, + "step": 32085 + }, + { + "epoch": 2.37, + "learning_rate": 1.3784692542582094e-05, + "loss": 0.8932, + "step": 32086 + }, + { + "epoch": 2.37, + "learning_rate": 1.3784323338470404e-05, + "loss": 1.0252, + "step": 32087 + }, + { + "epoch": 2.37, + "learning_rate": 1.3783954128337914e-05, + "loss": 0.968, + "step": 32088 + }, + { + "epoch": 2.37, + "learning_rate": 1.3783584912185205e-05, + "loss": 1.0282, + "step": 32089 + }, + { + "epoch": 2.37, + "learning_rate": 1.3783215690012863e-05, + "loss": 1.0294, + "step": 32090 + }, + { + "epoch": 2.37, + "learning_rate": 1.3782846461821484e-05, + "loss": 1.1198, + "step": 32091 + }, + { + "epoch": 2.37, + "learning_rate": 1.378247722761165e-05, + "loss": 0.9706, + "step": 32092 + }, + { + "epoch": 2.37, + "learning_rate": 1.3782107987383946e-05, + "loss": 0.9783, + "step": 32093 + }, + { + "epoch": 2.37, + "learning_rate": 1.3781738741138963e-05, + "loss": 0.9295, + "step": 32094 + }, + { + "epoch": 2.37, + "learning_rate": 1.378136948887729e-05, + "loss": 0.9808, + "step": 32095 + }, + { + "epoch": 2.37, + "learning_rate": 1.3781000230599511e-05, + "loss": 1.0581, + "step": 32096 + }, + { + "epoch": 2.37, + "learning_rate": 1.3780630966306211e-05, + "loss": 0.9121, + "step": 32097 + }, + { + "epoch": 2.37, + "learning_rate": 1.3780261695997987e-05, + "loss": 0.9766, + "step": 32098 + }, + { + "epoch": 2.37, + "learning_rate": 1.3779892419675418e-05, + "loss": 1.0697, + "step": 32099 + }, + { + "epoch": 2.37, + "learning_rate": 1.3779523137339095e-05, + "loss": 1.0737, + "step": 32100 + }, + { + "epoch": 2.37, + "learning_rate": 1.3779153848989605e-05, + "loss": 1.0431, + "step": 32101 + }, + { + "epoch": 2.37, + "learning_rate": 1.3778784554627535e-05, + "loss": 1.0414, + "step": 32102 + }, + { + "epoch": 2.37, + "learning_rate": 1.3778415254253473e-05, + "loss": 0.9788, + "step": 32103 + }, + { + "epoch": 2.37, + "learning_rate": 1.3778045947868005e-05, + "loss": 0.9077, + "step": 32104 + }, + { + "epoch": 2.37, + "learning_rate": 1.377767663547172e-05, + "loss": 1.0144, + "step": 32105 + }, + { + "epoch": 2.37, + "learning_rate": 1.3777307317065207e-05, + "loss": 1.0811, + "step": 32106 + }, + { + "epoch": 2.37, + "learning_rate": 1.377693799264905e-05, + "loss": 1.1296, + "step": 32107 + }, + { + "epoch": 2.37, + "learning_rate": 1.3776568662223842e-05, + "loss": 0.9879, + "step": 32108 + }, + { + "epoch": 2.37, + "learning_rate": 1.3776199325790165e-05, + "loss": 0.9945, + "step": 32109 + }, + { + "epoch": 2.37, + "learning_rate": 1.3775829983348611e-05, + "loss": 0.9603, + "step": 32110 + }, + { + "epoch": 2.37, + "learning_rate": 1.3775460634899762e-05, + "loss": 1.0739, + "step": 32111 + }, + { + "epoch": 2.37, + "learning_rate": 1.3775091280444213e-05, + "loss": 0.973, + "step": 32112 + }, + { + "epoch": 2.37, + "learning_rate": 1.3774721919982545e-05, + "loss": 1.0091, + "step": 32113 + }, + { + "epoch": 2.37, + "learning_rate": 1.3774352553515348e-05, + "loss": 0.9742, + "step": 32114 + }, + { + "epoch": 2.37, + "learning_rate": 1.3773983181043212e-05, + "loss": 1.013, + "step": 32115 + }, + { + "epoch": 2.37, + "learning_rate": 1.3773613802566723e-05, + "loss": 1.0724, + "step": 32116 + }, + { + "epoch": 2.37, + "learning_rate": 1.3773244418086466e-05, + "loss": 1.0671, + "step": 32117 + }, + { + "epoch": 2.37, + "learning_rate": 1.3772875027603035e-05, + "loss": 1.0674, + "step": 32118 + }, + { + "epoch": 2.37, + "learning_rate": 1.377250563111701e-05, + "loss": 1.0063, + "step": 32119 + }, + { + "epoch": 2.37, + "learning_rate": 1.3772136228628985e-05, + "loss": 0.991, + "step": 32120 + }, + { + "epoch": 2.37, + "learning_rate": 1.3771766820139546e-05, + "loss": 0.9772, + "step": 32121 + }, + { + "epoch": 2.37, + "learning_rate": 1.3771397405649278e-05, + "loss": 1.0541, + "step": 32122 + }, + { + "epoch": 2.37, + "learning_rate": 1.3771027985158769e-05, + "loss": 1.0434, + "step": 32123 + }, + { + "epoch": 2.37, + "learning_rate": 1.377065855866861e-05, + "loss": 1.0225, + "step": 32124 + }, + { + "epoch": 2.37, + "learning_rate": 1.3770289126179389e-05, + "loss": 1.021, + "step": 32125 + }, + { + "epoch": 2.37, + "learning_rate": 1.3769919687691688e-05, + "loss": 1.0108, + "step": 32126 + }, + { + "epoch": 2.37, + "learning_rate": 1.3769550243206103e-05, + "loss": 1.0355, + "step": 32127 + }, + { + "epoch": 2.37, + "learning_rate": 1.3769180792723216e-05, + "loss": 0.8697, + "step": 32128 + }, + { + "epoch": 2.37, + "learning_rate": 1.3768811336243614e-05, + "loss": 1.0189, + "step": 32129 + }, + { + "epoch": 2.37, + "learning_rate": 1.3768441873767891e-05, + "loss": 1.0244, + "step": 32130 + }, + { + "epoch": 2.37, + "learning_rate": 1.3768072405296629e-05, + "loss": 1.1125, + "step": 32131 + }, + { + "epoch": 2.37, + "learning_rate": 1.3767702930830416e-05, + "loss": 0.9388, + "step": 32132 + }, + { + "epoch": 2.37, + "learning_rate": 1.3767333450369844e-05, + "loss": 1.0512, + "step": 32133 + }, + { + "epoch": 2.37, + "learning_rate": 1.3766963963915499e-05, + "loss": 0.9461, + "step": 32134 + }, + { + "epoch": 2.37, + "learning_rate": 1.3766594471467964e-05, + "loss": 1.0947, + "step": 32135 + }, + { + "epoch": 2.37, + "learning_rate": 1.3766224973027835e-05, + "loss": 1.1087, + "step": 32136 + }, + { + "epoch": 2.37, + "learning_rate": 1.3765855468595694e-05, + "loss": 1.0261, + "step": 32137 + }, + { + "epoch": 2.37, + "learning_rate": 1.3765485958172133e-05, + "loss": 1.0468, + "step": 32138 + }, + { + "epoch": 2.37, + "learning_rate": 1.3765116441757738e-05, + "loss": 0.9235, + "step": 32139 + }, + { + "epoch": 2.37, + "learning_rate": 1.3764746919353091e-05, + "loss": 1.0416, + "step": 32140 + }, + { + "epoch": 2.37, + "learning_rate": 1.3764377390958792e-05, + "loss": 0.908, + "step": 32141 + }, + { + "epoch": 2.37, + "learning_rate": 1.3764007856575419e-05, + "loss": 1.0275, + "step": 32142 + }, + { + "epoch": 2.38, + "learning_rate": 1.3763638316203564e-05, + "loss": 1.0218, + "step": 32143 + }, + { + "epoch": 2.38, + "learning_rate": 1.3763268769843814e-05, + "loss": 0.9795, + "step": 32144 + }, + { + "epoch": 2.38, + "learning_rate": 1.3762899217496759e-05, + "loss": 1.0319, + "step": 32145 + }, + { + "epoch": 2.38, + "learning_rate": 1.3762529659162983e-05, + "loss": 0.8972, + "step": 32146 + }, + { + "epoch": 2.38, + "learning_rate": 1.3762160094843078e-05, + "loss": 1.0572, + "step": 32147 + }, + { + "epoch": 2.38, + "learning_rate": 1.376179052453763e-05, + "loss": 1.0421, + "step": 32148 + }, + { + "epoch": 2.38, + "learning_rate": 1.3761420948247226e-05, + "loss": 1.073, + "step": 32149 + }, + { + "epoch": 2.38, + "learning_rate": 1.3761051365972456e-05, + "loss": 1.0136, + "step": 32150 + }, + { + "epoch": 2.38, + "learning_rate": 1.3760681777713907e-05, + "loss": 0.9978, + "step": 32151 + }, + { + "epoch": 2.38, + "learning_rate": 1.3760312183472165e-05, + "loss": 1.0145, + "step": 32152 + }, + { + "epoch": 2.38, + "learning_rate": 1.3759942583247823e-05, + "loss": 0.9308, + "step": 32153 + }, + { + "epoch": 2.38, + "learning_rate": 1.3759572977041464e-05, + "loss": 0.9979, + "step": 32154 + }, + { + "epoch": 2.38, + "learning_rate": 1.3759203364853679e-05, + "loss": 1.1138, + "step": 32155 + }, + { + "epoch": 2.38, + "learning_rate": 1.3758833746685054e-05, + "loss": 0.9414, + "step": 32156 + }, + { + "epoch": 2.38, + "learning_rate": 1.3758464122536181e-05, + "loss": 1.0406, + "step": 32157 + }, + { + "epoch": 2.38, + "learning_rate": 1.3758094492407645e-05, + "loss": 1.0288, + "step": 32158 + }, + { + "epoch": 2.38, + "learning_rate": 1.3757724856300034e-05, + "loss": 1.1677, + "step": 32159 + }, + { + "epoch": 2.38, + "learning_rate": 1.3757355214213934e-05, + "loss": 1.101, + "step": 32160 + }, + { + "epoch": 2.38, + "learning_rate": 1.3756985566149938e-05, + "loss": 1.0051, + "step": 32161 + }, + { + "epoch": 2.38, + "learning_rate": 1.3756615912108631e-05, + "loss": 0.9761, + "step": 32162 + }, + { + "epoch": 2.38, + "learning_rate": 1.3756246252090603e-05, + "loss": 0.859, + "step": 32163 + }, + { + "epoch": 2.38, + "learning_rate": 1.3755876586096438e-05, + "loss": 0.9898, + "step": 32164 + }, + { + "epoch": 2.38, + "learning_rate": 1.375550691412673e-05, + "loss": 0.9998, + "step": 32165 + }, + { + "epoch": 2.38, + "learning_rate": 1.3755137236182065e-05, + "loss": 1.1313, + "step": 32166 + }, + { + "epoch": 2.38, + "learning_rate": 1.375476755226303e-05, + "loss": 0.9767, + "step": 32167 + }, + { + "epoch": 2.38, + "learning_rate": 1.375439786237021e-05, + "loss": 1.0227, + "step": 32168 + }, + { + "epoch": 2.38, + "learning_rate": 1.3754028166504201e-05, + "loss": 0.8984, + "step": 32169 + }, + { + "epoch": 2.38, + "learning_rate": 1.3753658464665582e-05, + "loss": 0.9644, + "step": 32170 + }, + { + "epoch": 2.38, + "learning_rate": 1.375328875685495e-05, + "loss": 1.0107, + "step": 32171 + }, + { + "epoch": 2.38, + "learning_rate": 1.3752919043072888e-05, + "loss": 1.0653, + "step": 32172 + }, + { + "epoch": 2.38, + "learning_rate": 1.3752549323319986e-05, + "loss": 0.96, + "step": 32173 + }, + { + "epoch": 2.38, + "learning_rate": 1.3752179597596831e-05, + "loss": 0.9636, + "step": 32174 + }, + { + "epoch": 2.38, + "learning_rate": 1.3751809865904014e-05, + "loss": 0.9469, + "step": 32175 + }, + { + "epoch": 2.38, + "learning_rate": 1.375144012824212e-05, + "loss": 0.9331, + "step": 32176 + }, + { + "epoch": 2.38, + "learning_rate": 1.3751070384611738e-05, + "loss": 1.0191, + "step": 32177 + }, + { + "epoch": 2.38, + "learning_rate": 1.3750700635013457e-05, + "loss": 0.9902, + "step": 32178 + }, + { + "epoch": 2.38, + "learning_rate": 1.3750330879447865e-05, + "loss": 1.0878, + "step": 32179 + }, + { + "epoch": 2.38, + "learning_rate": 1.3749961117915547e-05, + "loss": 1.0929, + "step": 32180 + }, + { + "epoch": 2.38, + "learning_rate": 1.37495913504171e-05, + "loss": 1.114, + "step": 32181 + }, + { + "epoch": 2.38, + "learning_rate": 1.3749221576953102e-05, + "loss": 0.9875, + "step": 32182 + }, + { + "epoch": 2.38, + "learning_rate": 1.374885179752415e-05, + "loss": 1.0487, + "step": 32183 + }, + { + "epoch": 2.38, + "learning_rate": 1.3748482012130824e-05, + "loss": 0.911, + "step": 32184 + }, + { + "epoch": 2.38, + "learning_rate": 1.3748112220773721e-05, + "loss": 1.0268, + "step": 32185 + }, + { + "epoch": 2.38, + "learning_rate": 1.3747742423453424e-05, + "loss": 0.9165, + "step": 32186 + }, + { + "epoch": 2.38, + "learning_rate": 1.3747372620170522e-05, + "loss": 0.9762, + "step": 32187 + }, + { + "epoch": 2.38, + "learning_rate": 1.3747002810925603e-05, + "loss": 1.0142, + "step": 32188 + }, + { + "epoch": 2.38, + "learning_rate": 1.3746632995719257e-05, + "loss": 0.9584, + "step": 32189 + }, + { + "epoch": 2.38, + "learning_rate": 1.3746263174552072e-05, + "loss": 0.8967, + "step": 32190 + }, + { + "epoch": 2.38, + "learning_rate": 1.3745893347424635e-05, + "loss": 0.9652, + "step": 32191 + }, + { + "epoch": 2.38, + "learning_rate": 1.3745523514337535e-05, + "loss": 1.0696, + "step": 32192 + }, + { + "epoch": 2.38, + "learning_rate": 1.374515367529136e-05, + "loss": 0.9514, + "step": 32193 + }, + { + "epoch": 2.38, + "learning_rate": 1.3744783830286704e-05, + "loss": 1.0228, + "step": 32194 + }, + { + "epoch": 2.38, + "learning_rate": 1.3744413979324144e-05, + "loss": 1.0455, + "step": 32195 + }, + { + "epoch": 2.38, + "learning_rate": 1.374404412240428e-05, + "loss": 0.9936, + "step": 32196 + }, + { + "epoch": 2.38, + "learning_rate": 1.3743674259527692e-05, + "loss": 1.0345, + "step": 32197 + }, + { + "epoch": 2.38, + "learning_rate": 1.3743304390694974e-05, + "loss": 1.0739, + "step": 32198 + }, + { + "epoch": 2.38, + "learning_rate": 1.3742934515906709e-05, + "loss": 0.943, + "step": 32199 + }, + { + "epoch": 2.38, + "learning_rate": 1.3742564635163493e-05, + "loss": 1.1216, + "step": 32200 + }, + { + "epoch": 2.38, + "learning_rate": 1.3742194748465907e-05, + "loss": 0.9346, + "step": 32201 + }, + { + "epoch": 2.38, + "learning_rate": 1.3741824855814545e-05, + "loss": 0.9952, + "step": 32202 + }, + { + "epoch": 2.38, + "learning_rate": 1.374145495720999e-05, + "loss": 1.0648, + "step": 32203 + }, + { + "epoch": 2.38, + "learning_rate": 1.374108505265284e-05, + "loss": 0.8924, + "step": 32204 + }, + { + "epoch": 2.38, + "learning_rate": 1.374071514214367e-05, + "loss": 0.9893, + "step": 32205 + }, + { + "epoch": 2.38, + "learning_rate": 1.374034522568308e-05, + "loss": 1.203, + "step": 32206 + }, + { + "epoch": 2.38, + "learning_rate": 1.3739975303271653e-05, + "loss": 0.9857, + "step": 32207 + }, + { + "epoch": 2.38, + "learning_rate": 1.373960537490998e-05, + "loss": 1.0628, + "step": 32208 + }, + { + "epoch": 2.38, + "learning_rate": 1.3739235440598646e-05, + "loss": 1.0166, + "step": 32209 + }, + { + "epoch": 2.38, + "learning_rate": 1.3738865500338243e-05, + "loss": 1.0679, + "step": 32210 + }, + { + "epoch": 2.38, + "learning_rate": 1.3738495554129356e-05, + "loss": 1.0606, + "step": 32211 + }, + { + "epoch": 2.38, + "learning_rate": 1.373812560197258e-05, + "loss": 0.9307, + "step": 32212 + }, + { + "epoch": 2.38, + "learning_rate": 1.37377556438685e-05, + "loss": 1.0576, + "step": 32213 + }, + { + "epoch": 2.38, + "learning_rate": 1.3737385679817704e-05, + "loss": 0.902, + "step": 32214 + }, + { + "epoch": 2.38, + "learning_rate": 1.3737015709820778e-05, + "loss": 1.0629, + "step": 32215 + }, + { + "epoch": 2.38, + "learning_rate": 1.3736645733878315e-05, + "loss": 1.028, + "step": 32216 + }, + { + "epoch": 2.38, + "learning_rate": 1.3736275751990901e-05, + "loss": 1.0631, + "step": 32217 + }, + { + "epoch": 2.38, + "learning_rate": 1.3735905764159127e-05, + "loss": 1.0787, + "step": 32218 + }, + { + "epoch": 2.38, + "learning_rate": 1.3735535770383579e-05, + "loss": 1.0127, + "step": 32219 + }, + { + "epoch": 2.38, + "learning_rate": 1.373516577066485e-05, + "loss": 0.9182, + "step": 32220 + }, + { + "epoch": 2.38, + "learning_rate": 1.3734795765003521e-05, + "loss": 1.0302, + "step": 32221 + }, + { + "epoch": 2.38, + "learning_rate": 1.373442575340019e-05, + "loss": 1.0121, + "step": 32222 + }, + { + "epoch": 2.38, + "learning_rate": 1.3734055735855441e-05, + "loss": 0.9384, + "step": 32223 + }, + { + "epoch": 2.38, + "learning_rate": 1.3733685712369861e-05, + "loss": 1.0457, + "step": 32224 + }, + { + "epoch": 2.38, + "learning_rate": 1.373331568294404e-05, + "loss": 1.0252, + "step": 32225 + }, + { + "epoch": 2.38, + "learning_rate": 1.3732945647578569e-05, + "loss": 0.8555, + "step": 32226 + }, + { + "epoch": 2.38, + "learning_rate": 1.373257560627403e-05, + "loss": 1.0534, + "step": 32227 + }, + { + "epoch": 2.38, + "learning_rate": 1.3732205559031023e-05, + "loss": 1.0483, + "step": 32228 + }, + { + "epoch": 2.38, + "learning_rate": 1.3731835505850125e-05, + "loss": 1.0895, + "step": 32229 + }, + { + "epoch": 2.38, + "learning_rate": 1.3731465446731934e-05, + "loss": 0.9563, + "step": 32230 + }, + { + "epoch": 2.38, + "learning_rate": 1.3731095381677033e-05, + "loss": 0.9245, + "step": 32231 + }, + { + "epoch": 2.38, + "learning_rate": 1.3730725310686012e-05, + "loss": 1.0063, + "step": 32232 + }, + { + "epoch": 2.38, + "learning_rate": 1.3730355233759462e-05, + "loss": 1.0704, + "step": 32233 + }, + { + "epoch": 2.38, + "learning_rate": 1.372998515089797e-05, + "loss": 1.0107, + "step": 32234 + }, + { + "epoch": 2.38, + "learning_rate": 1.3729615062102126e-05, + "loss": 0.9721, + "step": 32235 + }, + { + "epoch": 2.38, + "learning_rate": 1.3729244967372515e-05, + "loss": 0.9181, + "step": 32236 + }, + { + "epoch": 2.38, + "learning_rate": 1.372887486670973e-05, + "loss": 1.0974, + "step": 32237 + }, + { + "epoch": 2.38, + "learning_rate": 1.3728504760114358e-05, + "loss": 1.0625, + "step": 32238 + }, + { + "epoch": 2.38, + "learning_rate": 1.372813464758699e-05, + "loss": 1.032, + "step": 32239 + }, + { + "epoch": 2.38, + "learning_rate": 1.3727764529128212e-05, + "loss": 1.068, + "step": 32240 + }, + { + "epoch": 2.38, + "learning_rate": 1.3727394404738614e-05, + "loss": 1.0691, + "step": 32241 + }, + { + "epoch": 2.38, + "learning_rate": 1.3727024274418783e-05, + "loss": 0.9535, + "step": 32242 + }, + { + "epoch": 2.38, + "learning_rate": 1.3726654138169313e-05, + "loss": 1.0677, + "step": 32243 + }, + { + "epoch": 2.38, + "learning_rate": 1.3726283995990789e-05, + "loss": 0.9975, + "step": 32244 + }, + { + "epoch": 2.38, + "learning_rate": 1.3725913847883798e-05, + "loss": 0.9513, + "step": 32245 + }, + { + "epoch": 2.38, + "learning_rate": 1.3725543693848933e-05, + "loss": 1.123, + "step": 32246 + }, + { + "epoch": 2.38, + "learning_rate": 1.3725173533886782e-05, + "loss": 0.9687, + "step": 32247 + }, + { + "epoch": 2.38, + "learning_rate": 1.372480336799793e-05, + "loss": 1.0236, + "step": 32248 + }, + { + "epoch": 2.38, + "learning_rate": 1.3724433196182973e-05, + "loss": 0.962, + "step": 32249 + }, + { + "epoch": 2.38, + "learning_rate": 1.3724063018442494e-05, + "loss": 1.012, + "step": 32250 + }, + { + "epoch": 2.38, + "learning_rate": 1.3723692834777087e-05, + "loss": 0.9567, + "step": 32251 + }, + { + "epoch": 2.38, + "learning_rate": 1.3723322645187333e-05, + "loss": 0.9888, + "step": 32252 + }, + { + "epoch": 2.38, + "learning_rate": 1.372295244967383e-05, + "loss": 1.0511, + "step": 32253 + }, + { + "epoch": 2.38, + "learning_rate": 1.372258224823716e-05, + "loss": 0.9367, + "step": 32254 + }, + { + "epoch": 2.38, + "learning_rate": 1.3722212040877918e-05, + "loss": 0.9547, + "step": 32255 + }, + { + "epoch": 2.38, + "learning_rate": 1.3721841827596687e-05, + "loss": 1.018, + "step": 32256 + }, + { + "epoch": 2.38, + "learning_rate": 1.3721471608394059e-05, + "loss": 0.9178, + "step": 32257 + }, + { + "epoch": 2.38, + "learning_rate": 1.3721101383270626e-05, + "loss": 1.0327, + "step": 32258 + }, + { + "epoch": 2.38, + "learning_rate": 1.3720731152226971e-05, + "loss": 0.9805, + "step": 32259 + }, + { + "epoch": 2.38, + "learning_rate": 1.3720360915263688e-05, + "loss": 0.9743, + "step": 32260 + }, + { + "epoch": 2.38, + "learning_rate": 1.3719990672381362e-05, + "loss": 1.0313, + "step": 32261 + }, + { + "epoch": 2.38, + "learning_rate": 1.3719620423580583e-05, + "loss": 1.0296, + "step": 32262 + }, + { + "epoch": 2.38, + "learning_rate": 1.3719250168861946e-05, + "loss": 0.9638, + "step": 32263 + }, + { + "epoch": 2.38, + "learning_rate": 1.3718879908226032e-05, + "loss": 1.0134, + "step": 32264 + }, + { + "epoch": 2.38, + "learning_rate": 1.3718509641673435e-05, + "loss": 0.9113, + "step": 32265 + }, + { + "epoch": 2.38, + "learning_rate": 1.371813936920474e-05, + "loss": 0.8804, + "step": 32266 + }, + { + "epoch": 2.38, + "learning_rate": 1.3717769090820539e-05, + "loss": 0.9868, + "step": 32267 + }, + { + "epoch": 2.38, + "learning_rate": 1.3717398806521421e-05, + "loss": 0.9999, + "step": 32268 + }, + { + "epoch": 2.38, + "learning_rate": 1.3717028516307975e-05, + "loss": 1.041, + "step": 32269 + }, + { + "epoch": 2.38, + "learning_rate": 1.3716658220180789e-05, + "loss": 0.9871, + "step": 32270 + }, + { + "epoch": 2.38, + "learning_rate": 1.3716287918140454e-05, + "loss": 0.923, + "step": 32271 + }, + { + "epoch": 2.38, + "learning_rate": 1.3715917610187556e-05, + "loss": 0.9764, + "step": 32272 + }, + { + "epoch": 2.38, + "learning_rate": 1.371554729632269e-05, + "loss": 1.002, + "step": 32273 + }, + { + "epoch": 2.38, + "learning_rate": 1.3715176976546438e-05, + "loss": 1.0942, + "step": 32274 + }, + { + "epoch": 2.38, + "learning_rate": 1.3714806650859394e-05, + "loss": 0.9314, + "step": 32275 + }, + { + "epoch": 2.38, + "learning_rate": 1.3714436319262146e-05, + "loss": 1.0532, + "step": 32276 + }, + { + "epoch": 2.38, + "learning_rate": 1.3714065981755283e-05, + "loss": 1.0065, + "step": 32277 + }, + { + "epoch": 2.39, + "learning_rate": 1.3713695638339393e-05, + "loss": 0.9778, + "step": 32278 + }, + { + "epoch": 2.39, + "learning_rate": 1.3713325289015066e-05, + "loss": 0.9768, + "step": 32279 + }, + { + "epoch": 2.39, + "learning_rate": 1.3712954933782894e-05, + "loss": 1.0826, + "step": 32280 + }, + { + "epoch": 2.39, + "learning_rate": 1.371258457264346e-05, + "loss": 1.0208, + "step": 32281 + }, + { + "epoch": 2.39, + "learning_rate": 1.3712214205597364e-05, + "loss": 1.0345, + "step": 32282 + }, + { + "epoch": 2.39, + "learning_rate": 1.3711843832645182e-05, + "loss": 0.9413, + "step": 32283 + }, + { + "epoch": 2.39, + "learning_rate": 1.3711473453787512e-05, + "loss": 1.1465, + "step": 32284 + }, + { + "epoch": 2.39, + "learning_rate": 1.371110306902494e-05, + "loss": 1.0389, + "step": 32285 + }, + { + "epoch": 2.39, + "learning_rate": 1.3710732678358057e-05, + "loss": 1.035, + "step": 32286 + }, + { + "epoch": 2.39, + "learning_rate": 1.3710362281787451e-05, + "loss": 0.9355, + "step": 32287 + }, + { + "epoch": 2.39, + "learning_rate": 1.3709991879313711e-05, + "loss": 0.9855, + "step": 32288 + }, + { + "epoch": 2.39, + "learning_rate": 1.3709621470937426e-05, + "loss": 0.9995, + "step": 32289 + }, + { + "epoch": 2.39, + "learning_rate": 1.370925105665919e-05, + "loss": 0.859, + "step": 32290 + }, + { + "epoch": 2.39, + "learning_rate": 1.3708880636479585e-05, + "loss": 0.9876, + "step": 32291 + }, + { + "epoch": 2.39, + "learning_rate": 1.3708510210399206e-05, + "loss": 1.0053, + "step": 32292 + }, + { + "epoch": 2.39, + "learning_rate": 1.3708139778418637e-05, + "loss": 0.8946, + "step": 32293 + }, + { + "epoch": 2.39, + "learning_rate": 1.3707769340538475e-05, + "loss": 0.9881, + "step": 32294 + }, + { + "epoch": 2.39, + "learning_rate": 1.3707398896759304e-05, + "loss": 1.0812, + "step": 32295 + }, + { + "epoch": 2.39, + "learning_rate": 1.3707028447081712e-05, + "loss": 1.1171, + "step": 32296 + }, + { + "epoch": 2.39, + "learning_rate": 1.3706657991506293e-05, + "loss": 1.0331, + "step": 32297 + }, + { + "epoch": 2.39, + "learning_rate": 1.3706287530033636e-05, + "loss": 1.0281, + "step": 32298 + }, + { + "epoch": 2.39, + "learning_rate": 1.3705917062664324e-05, + "loss": 0.9528, + "step": 32299 + }, + { + "epoch": 2.39, + "learning_rate": 1.3705546589398954e-05, + "loss": 0.9543, + "step": 32300 + }, + { + "epoch": 2.39, + "learning_rate": 1.3705176110238112e-05, + "loss": 1.0298, + "step": 32301 + }, + { + "epoch": 2.39, + "learning_rate": 1.3704805625182389e-05, + "loss": 1.0003, + "step": 32302 + }, + { + "epoch": 2.39, + "learning_rate": 1.370443513423237e-05, + "loss": 0.934, + "step": 32303 + }, + { + "epoch": 2.39, + "learning_rate": 1.370406463738865e-05, + "loss": 0.9587, + "step": 32304 + }, + { + "epoch": 2.39, + "learning_rate": 1.3703694134651815e-05, + "loss": 1.0014, + "step": 32305 + }, + { + "epoch": 2.39, + "learning_rate": 1.3703323626022457e-05, + "loss": 1.0269, + "step": 32306 + }, + { + "epoch": 2.39, + "learning_rate": 1.3702953111501165e-05, + "loss": 0.9619, + "step": 32307 + }, + { + "epoch": 2.39, + "learning_rate": 1.3702582591088524e-05, + "loss": 0.9196, + "step": 32308 + }, + { + "epoch": 2.39, + "learning_rate": 1.370221206478513e-05, + "loss": 0.9048, + "step": 32309 + }, + { + "epoch": 2.39, + "learning_rate": 1.3701841532591568e-05, + "loss": 1.0482, + "step": 32310 + }, + { + "epoch": 2.39, + "learning_rate": 1.370147099450843e-05, + "loss": 0.92, + "step": 32311 + }, + { + "epoch": 2.39, + "learning_rate": 1.3701100450536305e-05, + "loss": 1.0442, + "step": 32312 + }, + { + "epoch": 2.39, + "learning_rate": 1.370072990067578e-05, + "loss": 0.8999, + "step": 32313 + }, + { + "epoch": 2.39, + "learning_rate": 1.3700359344927447e-05, + "loss": 0.9766, + "step": 32314 + }, + { + "epoch": 2.39, + "learning_rate": 1.3699988783291898e-05, + "loss": 1.0066, + "step": 32315 + }, + { + "epoch": 2.39, + "learning_rate": 1.3699618215769718e-05, + "loss": 0.8601, + "step": 32316 + }, + { + "epoch": 2.39, + "learning_rate": 1.3699247642361499e-05, + "loss": 1.0266, + "step": 32317 + }, + { + "epoch": 2.39, + "learning_rate": 1.3698877063067831e-05, + "loss": 0.9031, + "step": 32318 + }, + { + "epoch": 2.39, + "learning_rate": 1.3698506477889298e-05, + "loss": 0.9898, + "step": 32319 + }, + { + "epoch": 2.39, + "learning_rate": 1.3698135886826498e-05, + "loss": 0.9693, + "step": 32320 + }, + { + "epoch": 2.39, + "learning_rate": 1.3697765289880013e-05, + "loss": 1.1282, + "step": 32321 + }, + { + "epoch": 2.39, + "learning_rate": 1.3697394687050442e-05, + "loss": 1.08, + "step": 32322 + }, + { + "epoch": 2.39, + "learning_rate": 1.3697024078338364e-05, + "loss": 0.9388, + "step": 32323 + }, + { + "epoch": 2.39, + "learning_rate": 1.3696653463744376e-05, + "loss": 0.9858, + "step": 32324 + }, + { + "epoch": 2.39, + "learning_rate": 1.3696282843269067e-05, + "loss": 1.0176, + "step": 32325 + }, + { + "epoch": 2.39, + "learning_rate": 1.3695912216913023e-05, + "loss": 0.9675, + "step": 32326 + }, + { + "epoch": 2.39, + "learning_rate": 1.3695541584676835e-05, + "loss": 1.0717, + "step": 32327 + }, + { + "epoch": 2.39, + "learning_rate": 1.3695170946561094e-05, + "loss": 1.0996, + "step": 32328 + }, + { + "epoch": 2.39, + "learning_rate": 1.3694800302566387e-05, + "loss": 1.0482, + "step": 32329 + }, + { + "epoch": 2.39, + "learning_rate": 1.3694429652693306e-05, + "loss": 1.091, + "step": 32330 + }, + { + "epoch": 2.39, + "learning_rate": 1.369405899694244e-05, + "loss": 1.0307, + "step": 32331 + }, + { + "epoch": 2.39, + "learning_rate": 1.3693688335314383e-05, + "loss": 0.9124, + "step": 32332 + }, + { + "epoch": 2.39, + "learning_rate": 1.3693317667809717e-05, + "loss": 1.0978, + "step": 32333 + }, + { + "epoch": 2.39, + "learning_rate": 1.3692946994429036e-05, + "loss": 1.0372, + "step": 32334 + }, + { + "epoch": 2.39, + "learning_rate": 1.3692576315172929e-05, + "loss": 0.9862, + "step": 32335 + }, + { + "epoch": 2.39, + "learning_rate": 1.3692205630041988e-05, + "loss": 0.9125, + "step": 32336 + }, + { + "epoch": 2.39, + "learning_rate": 1.3691834939036798e-05, + "loss": 0.9987, + "step": 32337 + }, + { + "epoch": 2.39, + "learning_rate": 1.3691464242157952e-05, + "loss": 0.9718, + "step": 32338 + }, + { + "epoch": 2.39, + "learning_rate": 1.369109353940604e-05, + "loss": 1.0758, + "step": 32339 + }, + { + "epoch": 2.39, + "learning_rate": 1.369072283078165e-05, + "loss": 1.0905, + "step": 32340 + }, + { + "epoch": 2.39, + "learning_rate": 1.3690352116285374e-05, + "loss": 0.9725, + "step": 32341 + }, + { + "epoch": 2.39, + "learning_rate": 1.36899813959178e-05, + "loss": 0.8898, + "step": 32342 + }, + { + "epoch": 2.39, + "learning_rate": 1.3689610669679518e-05, + "loss": 1.0274, + "step": 32343 + }, + { + "epoch": 2.39, + "learning_rate": 1.368923993757112e-05, + "loss": 0.9532, + "step": 32344 + }, + { + "epoch": 2.39, + "learning_rate": 1.3688869199593193e-05, + "loss": 1.0125, + "step": 32345 + }, + { + "epoch": 2.39, + "learning_rate": 1.3688498455746327e-05, + "loss": 1.0558, + "step": 32346 + }, + { + "epoch": 2.39, + "learning_rate": 1.3688127706031114e-05, + "loss": 1.0381, + "step": 32347 + }, + { + "epoch": 2.39, + "learning_rate": 1.3687756950448141e-05, + "loss": 1.0248, + "step": 32348 + }, + { + "epoch": 2.39, + "learning_rate": 1.3687386188998001e-05, + "loss": 1.0262, + "step": 32349 + }, + { + "epoch": 2.39, + "learning_rate": 1.368701542168128e-05, + "loss": 0.939, + "step": 32350 + }, + { + "epoch": 2.39, + "learning_rate": 1.3686644648498574e-05, + "loss": 0.9628, + "step": 32351 + }, + { + "epoch": 2.39, + "learning_rate": 1.3686273869450466e-05, + "loss": 1.0079, + "step": 32352 + }, + { + "epoch": 2.39, + "learning_rate": 1.3685903084537553e-05, + "loss": 1.025, + "step": 32353 + }, + { + "epoch": 2.39, + "learning_rate": 1.3685532293760417e-05, + "loss": 0.9142, + "step": 32354 + }, + { + "epoch": 2.39, + "learning_rate": 1.3685161497119653e-05, + "loss": 1.0589, + "step": 32355 + }, + { + "epoch": 2.39, + "learning_rate": 1.3684790694615849e-05, + "loss": 0.9717, + "step": 32356 + }, + { + "epoch": 2.39, + "learning_rate": 1.3684419886249596e-05, + "loss": 0.9075, + "step": 32357 + }, + { + "epoch": 2.39, + "learning_rate": 1.3684049072021484e-05, + "loss": 1.0698, + "step": 32358 + }, + { + "epoch": 2.39, + "learning_rate": 1.3683678251932104e-05, + "loss": 1.0398, + "step": 32359 + }, + { + "epoch": 2.39, + "learning_rate": 1.3683307425982042e-05, + "loss": 1.0396, + "step": 32360 + }, + { + "epoch": 2.39, + "learning_rate": 1.3682936594171892e-05, + "loss": 1.054, + "step": 32361 + }, + { + "epoch": 2.39, + "learning_rate": 1.3682565756502241e-05, + "loss": 1.0331, + "step": 32362 + }, + { + "epoch": 2.39, + "learning_rate": 1.3682194912973686e-05, + "loss": 1.0032, + "step": 32363 + }, + { + "epoch": 2.39, + "learning_rate": 1.3681824063586806e-05, + "loss": 1.0339, + "step": 32364 + }, + { + "epoch": 2.39, + "learning_rate": 1.36814532083422e-05, + "loss": 0.9973, + "step": 32365 + }, + { + "epoch": 2.39, + "learning_rate": 1.3681082347240451e-05, + "loss": 0.9489, + "step": 32366 + }, + { + "epoch": 2.39, + "learning_rate": 1.3680711480282155e-05, + "loss": 1.0214, + "step": 32367 + }, + { + "epoch": 2.39, + "learning_rate": 1.3680340607467897e-05, + "loss": 1.0781, + "step": 32368 + }, + { + "epoch": 2.39, + "learning_rate": 1.3679969728798272e-05, + "loss": 0.9217, + "step": 32369 + }, + { + "epoch": 2.39, + "learning_rate": 1.367959884427387e-05, + "loss": 0.9644, + "step": 32370 + }, + { + "epoch": 2.39, + "learning_rate": 1.3679227953895276e-05, + "loss": 0.9007, + "step": 32371 + }, + { + "epoch": 2.39, + "learning_rate": 1.3678857057663084e-05, + "loss": 1.0685, + "step": 32372 + }, + { + "epoch": 2.39, + "learning_rate": 1.3678486155577883e-05, + "loss": 1.0806, + "step": 32373 + }, + { + "epoch": 2.39, + "learning_rate": 1.3678115247640262e-05, + "loss": 1.0467, + "step": 32374 + }, + { + "epoch": 2.39, + "learning_rate": 1.3677744333850813e-05, + "loss": 1.1173, + "step": 32375 + }, + { + "epoch": 2.39, + "learning_rate": 1.3677373414210124e-05, + "loss": 0.9925, + "step": 32376 + }, + { + "epoch": 2.39, + "learning_rate": 1.367700248871879e-05, + "loss": 0.9945, + "step": 32377 + }, + { + "epoch": 2.39, + "learning_rate": 1.3676631557377393e-05, + "loss": 1.1092, + "step": 32378 + }, + { + "epoch": 2.39, + "learning_rate": 1.3676260620186532e-05, + "loss": 1.0577, + "step": 32379 + }, + { + "epoch": 2.39, + "learning_rate": 1.367588967714679e-05, + "loss": 1.089, + "step": 32380 + }, + { + "epoch": 2.39, + "learning_rate": 1.3675518728258763e-05, + "loss": 1.0513, + "step": 32381 + }, + { + "epoch": 2.39, + "learning_rate": 1.3675147773523038e-05, + "loss": 0.9861, + "step": 32382 + }, + { + "epoch": 2.39, + "learning_rate": 1.3674776812940202e-05, + "loss": 0.9504, + "step": 32383 + }, + { + "epoch": 2.39, + "learning_rate": 1.3674405846510852e-05, + "loss": 0.987, + "step": 32384 + }, + { + "epoch": 2.39, + "learning_rate": 1.3674034874235573e-05, + "loss": 0.96, + "step": 32385 + }, + { + "epoch": 2.39, + "learning_rate": 1.3673663896114957e-05, + "loss": 1.0946, + "step": 32386 + }, + { + "epoch": 2.39, + "learning_rate": 1.3673292912149596e-05, + "loss": 0.9429, + "step": 32387 + }, + { + "epoch": 2.39, + "learning_rate": 1.3672921922340078e-05, + "loss": 1.0572, + "step": 32388 + }, + { + "epoch": 2.39, + "learning_rate": 1.3672550926686994e-05, + "loss": 1.0217, + "step": 32389 + }, + { + "epoch": 2.39, + "learning_rate": 1.3672179925190934e-05, + "loss": 0.8634, + "step": 32390 + }, + { + "epoch": 2.39, + "learning_rate": 1.3671808917852486e-05, + "loss": 0.9843, + "step": 32391 + }, + { + "epoch": 2.39, + "learning_rate": 1.3671437904672247e-05, + "loss": 1.0421, + "step": 32392 + }, + { + "epoch": 2.39, + "learning_rate": 1.36710668856508e-05, + "loss": 1.0883, + "step": 32393 + }, + { + "epoch": 2.39, + "learning_rate": 1.367069586078874e-05, + "loss": 1.0585, + "step": 32394 + }, + { + "epoch": 2.39, + "learning_rate": 1.367032483008665e-05, + "loss": 1.0133, + "step": 32395 + }, + { + "epoch": 2.39, + "learning_rate": 1.366995379354513e-05, + "loss": 1.0397, + "step": 32396 + }, + { + "epoch": 2.39, + "learning_rate": 1.3669582751164763e-05, + "loss": 1.0177, + "step": 32397 + }, + { + "epoch": 2.39, + "learning_rate": 1.3669211702946146e-05, + "loss": 1.1249, + "step": 32398 + }, + { + "epoch": 2.39, + "learning_rate": 1.3668840648889865e-05, + "loss": 0.9527, + "step": 32399 + }, + { + "epoch": 2.39, + "learning_rate": 1.3668469588996511e-05, + "loss": 0.9831, + "step": 32400 + }, + { + "epoch": 2.39, + "learning_rate": 1.3668098523266673e-05, + "loss": 1.1265, + "step": 32401 + }, + { + "epoch": 2.39, + "learning_rate": 1.3667727451700945e-05, + "loss": 1.0237, + "step": 32402 + }, + { + "epoch": 2.39, + "learning_rate": 1.366735637429991e-05, + "loss": 1.045, + "step": 32403 + }, + { + "epoch": 2.39, + "learning_rate": 1.366698529106417e-05, + "loss": 0.9736, + "step": 32404 + }, + { + "epoch": 2.39, + "learning_rate": 1.3666614201994306e-05, + "loss": 1.0766, + "step": 32405 + }, + { + "epoch": 2.39, + "learning_rate": 1.3666243107090912e-05, + "loss": 0.9421, + "step": 32406 + }, + { + "epoch": 2.39, + "learning_rate": 1.3665872006354574e-05, + "loss": 0.8946, + "step": 32407 + }, + { + "epoch": 2.39, + "learning_rate": 1.3665500899785891e-05, + "loss": 1.1052, + "step": 32408 + }, + { + "epoch": 2.39, + "learning_rate": 1.3665129787385448e-05, + "loss": 0.8307, + "step": 32409 + }, + { + "epoch": 2.39, + "learning_rate": 1.3664758669153835e-05, + "loss": 0.989, + "step": 32410 + }, + { + "epoch": 2.39, + "learning_rate": 1.3664387545091641e-05, + "loss": 0.9523, + "step": 32411 + }, + { + "epoch": 2.39, + "learning_rate": 1.3664016415199461e-05, + "loss": 1.124, + "step": 32412 + }, + { + "epoch": 2.39, + "learning_rate": 1.3663645279477881e-05, + "loss": 0.9427, + "step": 32413 + }, + { + "epoch": 2.4, + "learning_rate": 1.3663274137927498e-05, + "loss": 0.9539, + "step": 32414 + }, + { + "epoch": 2.4, + "learning_rate": 1.3662902990548894e-05, + "loss": 1.0097, + "step": 32415 + }, + { + "epoch": 2.4, + "learning_rate": 1.3662531837342667e-05, + "loss": 1.0577, + "step": 32416 + }, + { + "epoch": 2.4, + "learning_rate": 1.3662160678309402e-05, + "loss": 0.9199, + "step": 32417 + }, + { + "epoch": 2.4, + "learning_rate": 1.3661789513449692e-05, + "loss": 0.9692, + "step": 32418 + }, + { + "epoch": 2.4, + "learning_rate": 1.366141834276413e-05, + "loss": 1.1226, + "step": 32419 + }, + { + "epoch": 2.4, + "learning_rate": 1.3661047166253303e-05, + "loss": 1.127, + "step": 32420 + }, + { + "epoch": 2.4, + "learning_rate": 1.36606759839178e-05, + "loss": 0.9341, + "step": 32421 + }, + { + "epoch": 2.4, + "learning_rate": 1.3660304795758217e-05, + "loss": 1.0527, + "step": 32422 + }, + { + "epoch": 2.4, + "learning_rate": 1.3659933601775136e-05, + "loss": 1.0723, + "step": 32423 + }, + { + "epoch": 2.4, + "learning_rate": 1.365956240196916e-05, + "loss": 1.0402, + "step": 32424 + }, + { + "epoch": 2.4, + "learning_rate": 1.3659191196340867e-05, + "loss": 1.0946, + "step": 32425 + }, + { + "epoch": 2.4, + "learning_rate": 1.3658819984890857e-05, + "loss": 1.071, + "step": 32426 + }, + { + "epoch": 2.4, + "learning_rate": 1.3658448767619711e-05, + "loss": 0.9767, + "step": 32427 + }, + { + "epoch": 2.4, + "learning_rate": 1.3658077544528031e-05, + "loss": 1.0336, + "step": 32428 + }, + { + "epoch": 2.4, + "learning_rate": 1.3657706315616404e-05, + "loss": 1.0792, + "step": 32429 + }, + { + "epoch": 2.4, + "learning_rate": 1.3657335080885412e-05, + "loss": 0.9334, + "step": 32430 + }, + { + "epoch": 2.4, + "learning_rate": 1.3656963840335658e-05, + "loss": 1.026, + "step": 32431 + }, + { + "epoch": 2.4, + "learning_rate": 1.3656592593967723e-05, + "loss": 1.1292, + "step": 32432 + }, + { + "epoch": 2.4, + "learning_rate": 1.3656221341782201e-05, + "loss": 1.018, + "step": 32433 + }, + { + "epoch": 2.4, + "learning_rate": 1.3655850083779686e-05, + "loss": 0.9979, + "step": 32434 + }, + { + "epoch": 2.4, + "learning_rate": 1.3655478819960764e-05, + "loss": 1.0013, + "step": 32435 + }, + { + "epoch": 2.4, + "learning_rate": 1.3655107550326028e-05, + "loss": 1.1235, + "step": 32436 + }, + { + "epoch": 2.4, + "learning_rate": 1.3654736274876068e-05, + "loss": 1.0281, + "step": 32437 + }, + { + "epoch": 2.4, + "learning_rate": 1.3654364993611476e-05, + "loss": 0.9646, + "step": 32438 + }, + { + "epoch": 2.4, + "learning_rate": 1.365399370653284e-05, + "loss": 0.8914, + "step": 32439 + }, + { + "epoch": 2.4, + "learning_rate": 1.3653622413640753e-05, + "loss": 0.9762, + "step": 32440 + }, + { + "epoch": 2.4, + "learning_rate": 1.3653251114935806e-05, + "loss": 1.1433, + "step": 32441 + }, + { + "epoch": 2.4, + "learning_rate": 1.3652879810418585e-05, + "loss": 0.938, + "step": 32442 + }, + { + "epoch": 2.4, + "learning_rate": 1.3652508500089687e-05, + "loss": 0.9254, + "step": 32443 + }, + { + "epoch": 2.4, + "learning_rate": 1.36521371839497e-05, + "loss": 1.0472, + "step": 32444 + }, + { + "epoch": 2.4, + "learning_rate": 1.3651765861999216e-05, + "loss": 1.021, + "step": 32445 + }, + { + "epoch": 2.4, + "learning_rate": 1.3651394534238823e-05, + "loss": 0.9696, + "step": 32446 + }, + { + "epoch": 2.4, + "learning_rate": 1.3651023200669118e-05, + "loss": 1.0668, + "step": 32447 + }, + { + "epoch": 2.4, + "learning_rate": 1.3650651861290682e-05, + "loss": 1.016, + "step": 32448 + }, + { + "epoch": 2.4, + "learning_rate": 1.3650280516104114e-05, + "loss": 1.0875, + "step": 32449 + }, + { + "epoch": 2.4, + "learning_rate": 1.364990916511e-05, + "loss": 0.946, + "step": 32450 + }, + { + "epoch": 2.4, + "learning_rate": 1.3649537808308935e-05, + "loss": 0.8994, + "step": 32451 + }, + { + "epoch": 2.4, + "learning_rate": 1.3649166445701505e-05, + "loss": 0.9694, + "step": 32452 + }, + { + "epoch": 2.4, + "learning_rate": 1.3648795077288306e-05, + "loss": 0.9685, + "step": 32453 + }, + { + "epoch": 2.4, + "learning_rate": 1.3648423703069924e-05, + "loss": 0.9465, + "step": 32454 + }, + { + "epoch": 2.4, + "learning_rate": 1.3648052323046954e-05, + "loss": 1.1199, + "step": 32455 + }, + { + "epoch": 2.4, + "learning_rate": 1.3647680937219984e-05, + "loss": 1.0443, + "step": 32456 + }, + { + "epoch": 2.4, + "learning_rate": 1.3647309545589607e-05, + "loss": 1.0071, + "step": 32457 + }, + { + "epoch": 2.4, + "learning_rate": 1.364693814815641e-05, + "loss": 0.9966, + "step": 32458 + }, + { + "epoch": 2.4, + "learning_rate": 1.364656674492099e-05, + "loss": 1.1879, + "step": 32459 + }, + { + "epoch": 2.4, + "learning_rate": 1.3646195335883932e-05, + "loss": 1.0235, + "step": 32460 + }, + { + "epoch": 2.4, + "learning_rate": 1.3645823921045831e-05, + "loss": 0.906, + "step": 32461 + }, + { + "epoch": 2.4, + "learning_rate": 1.3645452500407272e-05, + "loss": 1.0281, + "step": 32462 + }, + { + "epoch": 2.4, + "learning_rate": 1.3645081073968855e-05, + "loss": 1.043, + "step": 32463 + }, + { + "epoch": 2.4, + "learning_rate": 1.3644709641731164e-05, + "loss": 0.97, + "step": 32464 + }, + { + "epoch": 2.4, + "learning_rate": 1.3644338203694796e-05, + "loss": 0.9993, + "step": 32465 + }, + { + "epoch": 2.4, + "learning_rate": 1.3643966759860333e-05, + "loss": 0.9704, + "step": 32466 + }, + { + "epoch": 2.4, + "learning_rate": 1.3643595310228374e-05, + "loss": 1.0459, + "step": 32467 + }, + { + "epoch": 2.4, + "learning_rate": 1.3643223854799506e-05, + "loss": 1.022, + "step": 32468 + }, + { + "epoch": 2.4, + "learning_rate": 1.364285239357432e-05, + "loss": 1.0015, + "step": 32469 + }, + { + "epoch": 2.4, + "learning_rate": 1.3642480926553407e-05, + "loss": 0.9923, + "step": 32470 + }, + { + "epoch": 2.4, + "learning_rate": 1.3642109453737361e-05, + "loss": 1.0035, + "step": 32471 + }, + { + "epoch": 2.4, + "learning_rate": 1.3641737975126772e-05, + "loss": 0.9991, + "step": 32472 + }, + { + "epoch": 2.4, + "learning_rate": 1.3641366490722229e-05, + "loss": 0.9138, + "step": 32473 + }, + { + "epoch": 2.4, + "learning_rate": 1.3640995000524322e-05, + "loss": 0.8596, + "step": 32474 + }, + { + "epoch": 2.4, + "learning_rate": 1.3640623504533648e-05, + "loss": 1.0225, + "step": 32475 + }, + { + "epoch": 2.4, + "learning_rate": 1.3640252002750791e-05, + "loss": 1.1129, + "step": 32476 + }, + { + "epoch": 2.4, + "learning_rate": 1.3639880495176347e-05, + "loss": 1.0343, + "step": 32477 + }, + { + "epoch": 2.4, + "learning_rate": 1.3639508981810904e-05, + "loss": 1.073, + "step": 32478 + }, + { + "epoch": 2.4, + "learning_rate": 1.3639137462655056e-05, + "loss": 0.973, + "step": 32479 + }, + { + "epoch": 2.4, + "learning_rate": 1.3638765937709388e-05, + "loss": 1.1001, + "step": 32480 + }, + { + "epoch": 2.4, + "learning_rate": 1.36383944069745e-05, + "loss": 0.9831, + "step": 32481 + }, + { + "epoch": 2.4, + "learning_rate": 1.3638022870450975e-05, + "loss": 0.9862, + "step": 32482 + }, + { + "epoch": 2.4, + "learning_rate": 1.3637651328139413e-05, + "loss": 0.9893, + "step": 32483 + }, + { + "epoch": 2.4, + "learning_rate": 1.3637279780040395e-05, + "loss": 1.036, + "step": 32484 + }, + { + "epoch": 2.4, + "learning_rate": 1.3636908226154517e-05, + "loss": 1.0381, + "step": 32485 + }, + { + "epoch": 2.4, + "learning_rate": 1.3636536666482375e-05, + "loss": 0.9005, + "step": 32486 + }, + { + "epoch": 2.4, + "learning_rate": 1.3636165101024548e-05, + "loss": 1.0434, + "step": 32487 + }, + { + "epoch": 2.4, + "learning_rate": 1.3635793529781637e-05, + "loss": 1.0831, + "step": 32488 + }, + { + "epoch": 2.4, + "learning_rate": 1.3635421952754234e-05, + "loss": 1.0032, + "step": 32489 + }, + { + "epoch": 2.4, + "learning_rate": 1.3635050369942921e-05, + "loss": 1.0405, + "step": 32490 + }, + { + "epoch": 2.4, + "learning_rate": 1.3634678781348298e-05, + "loss": 0.9212, + "step": 32491 + }, + { + "epoch": 2.4, + "learning_rate": 1.3634307186970954e-05, + "loss": 1.0207, + "step": 32492 + }, + { + "epoch": 2.4, + "learning_rate": 1.363393558681148e-05, + "loss": 1.0533, + "step": 32493 + }, + { + "epoch": 2.4, + "learning_rate": 1.3633563980870465e-05, + "loss": 0.9404, + "step": 32494 + }, + { + "epoch": 2.4, + "learning_rate": 1.36331923691485e-05, + "loss": 1.0529, + "step": 32495 + }, + { + "epoch": 2.4, + "learning_rate": 1.363282075164618e-05, + "loss": 1.0089, + "step": 32496 + }, + { + "epoch": 2.4, + "learning_rate": 1.3632449128364092e-05, + "loss": 1.0448, + "step": 32497 + }, + { + "epoch": 2.4, + "learning_rate": 1.3632077499302834e-05, + "loss": 1.0291, + "step": 32498 + }, + { + "epoch": 2.4, + "learning_rate": 1.3631705864462987e-05, + "loss": 0.9191, + "step": 32499 + }, + { + "epoch": 2.4, + "learning_rate": 1.3631334223845152e-05, + "loss": 1.0064, + "step": 32500 + }, + { + "epoch": 2.4, + "learning_rate": 1.3630962577449914e-05, + "loss": 1.0923, + "step": 32501 + }, + { + "epoch": 2.4, + "learning_rate": 1.3630590925277867e-05, + "loss": 0.8725, + "step": 32502 + }, + { + "epoch": 2.4, + "learning_rate": 1.3630219267329603e-05, + "loss": 0.9989, + "step": 32503 + }, + { + "epoch": 2.4, + "learning_rate": 1.362984760360571e-05, + "loss": 1.0039, + "step": 32504 + }, + { + "epoch": 2.4, + "learning_rate": 1.3629475934106781e-05, + "loss": 0.9061, + "step": 32505 + }, + { + "epoch": 2.4, + "learning_rate": 1.3629104258833411e-05, + "loss": 1.0352, + "step": 32506 + }, + { + "epoch": 2.4, + "learning_rate": 1.3628732577786186e-05, + "loss": 0.9968, + "step": 32507 + }, + { + "epoch": 2.4, + "learning_rate": 1.3628360890965699e-05, + "loss": 0.8756, + "step": 32508 + }, + { + "epoch": 2.4, + "learning_rate": 1.3627989198372542e-05, + "loss": 1.0266, + "step": 32509 + }, + { + "epoch": 2.4, + "learning_rate": 1.362761750000731e-05, + "loss": 0.9776, + "step": 32510 + }, + { + "epoch": 2.4, + "learning_rate": 1.3627245795870582e-05, + "loss": 1.0673, + "step": 32511 + }, + { + "epoch": 2.4, + "learning_rate": 1.3626874085962967e-05, + "loss": 1.1237, + "step": 32512 + }, + { + "epoch": 2.4, + "learning_rate": 1.3626502370285042e-05, + "loss": 1.1171, + "step": 32513 + }, + { + "epoch": 2.4, + "learning_rate": 1.3626130648837406e-05, + "loss": 0.911, + "step": 32514 + }, + { + "epoch": 2.4, + "learning_rate": 1.3625758921620643e-05, + "loss": 0.9786, + "step": 32515 + }, + { + "epoch": 2.4, + "learning_rate": 1.3625387188635355e-05, + "loss": 0.9867, + "step": 32516 + }, + { + "epoch": 2.4, + "learning_rate": 1.3625015449882124e-05, + "loss": 1.0891, + "step": 32517 + }, + { + "epoch": 2.4, + "learning_rate": 1.3624643705361548e-05, + "loss": 0.9484, + "step": 32518 + }, + { + "epoch": 2.4, + "learning_rate": 1.3624271955074216e-05, + "loss": 1.0413, + "step": 32519 + }, + { + "epoch": 2.4, + "learning_rate": 1.3623900199020718e-05, + "loss": 1.052, + "step": 32520 + }, + { + "epoch": 2.4, + "learning_rate": 1.3623528437201647e-05, + "loss": 0.9417, + "step": 32521 + }, + { + "epoch": 2.4, + "learning_rate": 1.3623156669617595e-05, + "loss": 0.9722, + "step": 32522 + }, + { + "epoch": 2.4, + "learning_rate": 1.362278489626915e-05, + "loss": 1.0827, + "step": 32523 + }, + { + "epoch": 2.4, + "learning_rate": 1.3622413117156909e-05, + "loss": 1.028, + "step": 32524 + }, + { + "epoch": 2.4, + "learning_rate": 1.3622041332281457e-05, + "loss": 0.9405, + "step": 32525 + }, + { + "epoch": 2.4, + "learning_rate": 1.3621669541643393e-05, + "loss": 1.0624, + "step": 32526 + }, + { + "epoch": 2.4, + "learning_rate": 1.36212977452433e-05, + "loss": 0.9142, + "step": 32527 + }, + { + "epoch": 2.4, + "learning_rate": 1.362092594308178e-05, + "loss": 0.9561, + "step": 32528 + }, + { + "epoch": 2.4, + "learning_rate": 1.3620554135159413e-05, + "loss": 1.0982, + "step": 32529 + }, + { + "epoch": 2.4, + "learning_rate": 1.3620182321476798e-05, + "loss": 1.0203, + "step": 32530 + }, + { + "epoch": 2.4, + "learning_rate": 1.3619810502034527e-05, + "loss": 1.0158, + "step": 32531 + }, + { + "epoch": 2.4, + "learning_rate": 1.3619438676833187e-05, + "loss": 1.0264, + "step": 32532 + }, + { + "epoch": 2.4, + "learning_rate": 1.3619066845873373e-05, + "loss": 1.0329, + "step": 32533 + }, + { + "epoch": 2.4, + "learning_rate": 1.3618695009155674e-05, + "loss": 0.9792, + "step": 32534 + }, + { + "epoch": 2.4, + "learning_rate": 1.3618323166680684e-05, + "loss": 1.0691, + "step": 32535 + }, + { + "epoch": 2.4, + "learning_rate": 1.3617951318448992e-05, + "loss": 1.0775, + "step": 32536 + }, + { + "epoch": 2.4, + "learning_rate": 1.3617579464461195e-05, + "loss": 1.0936, + "step": 32537 + }, + { + "epoch": 2.4, + "learning_rate": 1.361720760471788e-05, + "loss": 0.9958, + "step": 32538 + }, + { + "epoch": 2.4, + "learning_rate": 1.3616835739219636e-05, + "loss": 1.1442, + "step": 32539 + }, + { + "epoch": 2.4, + "learning_rate": 1.361646386796706e-05, + "loss": 1.1158, + "step": 32540 + }, + { + "epoch": 2.4, + "learning_rate": 1.3616091990960742e-05, + "loss": 0.9407, + "step": 32541 + }, + { + "epoch": 2.4, + "learning_rate": 1.3615720108201273e-05, + "loss": 0.9889, + "step": 32542 + }, + { + "epoch": 2.4, + "learning_rate": 1.3615348219689246e-05, + "loss": 1.0218, + "step": 32543 + }, + { + "epoch": 2.4, + "learning_rate": 1.361497632542525e-05, + "loss": 1.1272, + "step": 32544 + }, + { + "epoch": 2.4, + "learning_rate": 1.3614604425409881e-05, + "loss": 1.0833, + "step": 32545 + }, + { + "epoch": 2.4, + "learning_rate": 1.3614232519643724e-05, + "loss": 1.0867, + "step": 32546 + }, + { + "epoch": 2.4, + "learning_rate": 1.3613860608127378e-05, + "loss": 0.9866, + "step": 32547 + }, + { + "epoch": 2.4, + "learning_rate": 1.3613488690861429e-05, + "loss": 1.0214, + "step": 32548 + }, + { + "epoch": 2.41, + "learning_rate": 1.3613116767846477e-05, + "loss": 0.9098, + "step": 32549 + }, + { + "epoch": 2.41, + "learning_rate": 1.36127448390831e-05, + "loss": 0.9357, + "step": 32550 + }, + { + "epoch": 2.41, + "learning_rate": 1.3612372904571904e-05, + "loss": 1.0582, + "step": 32551 + }, + { + "epoch": 2.41, + "learning_rate": 1.3612000964313469e-05, + "loss": 1.11, + "step": 32552 + }, + { + "epoch": 2.41, + "learning_rate": 1.3611629018308397e-05, + "loss": 0.8947, + "step": 32553 + }, + { + "epoch": 2.41, + "learning_rate": 1.3611257066557273e-05, + "loss": 1.0284, + "step": 32554 + }, + { + "epoch": 2.41, + "learning_rate": 1.361088510906069e-05, + "loss": 1.0035, + "step": 32555 + }, + { + "epoch": 2.41, + "learning_rate": 1.361051314581924e-05, + "loss": 0.9062, + "step": 32556 + }, + { + "epoch": 2.41, + "learning_rate": 1.3610141176833516e-05, + "loss": 1.0384, + "step": 32557 + }, + { + "epoch": 2.41, + "learning_rate": 1.3609769202104108e-05, + "loss": 1.0111, + "step": 32558 + }, + { + "epoch": 2.41, + "learning_rate": 1.3609397221631614e-05, + "loss": 1.0346, + "step": 32559 + }, + { + "epoch": 2.41, + "learning_rate": 1.3609025235416615e-05, + "loss": 0.8771, + "step": 32560 + }, + { + "epoch": 2.41, + "learning_rate": 1.360865324345971e-05, + "loss": 0.9965, + "step": 32561 + }, + { + "epoch": 2.41, + "learning_rate": 1.360828124576149e-05, + "loss": 1.0681, + "step": 32562 + }, + { + "epoch": 2.41, + "learning_rate": 1.3607909242322545e-05, + "loss": 1.0424, + "step": 32563 + }, + { + "epoch": 2.41, + "learning_rate": 1.3607537233143468e-05, + "loss": 1.0712, + "step": 32564 + }, + { + "epoch": 2.41, + "learning_rate": 1.3607165218224854e-05, + "loss": 1.081, + "step": 32565 + }, + { + "epoch": 2.41, + "learning_rate": 1.3606793197567288e-05, + "loss": 1.0618, + "step": 32566 + }, + { + "epoch": 2.41, + "learning_rate": 1.3606421171171369e-05, + "loss": 1.0292, + "step": 32567 + }, + { + "epoch": 2.41, + "learning_rate": 1.3606049139037682e-05, + "loss": 1.0201, + "step": 32568 + }, + { + "epoch": 2.41, + "learning_rate": 1.3605677101166825e-05, + "loss": 0.9322, + "step": 32569 + }, + { + "epoch": 2.41, + "learning_rate": 1.3605305057559386e-05, + "loss": 1.0072, + "step": 32570 + }, + { + "epoch": 2.41, + "learning_rate": 1.360493300821596e-05, + "loss": 1.0367, + "step": 32571 + }, + { + "epoch": 2.41, + "learning_rate": 1.3604560953137133e-05, + "loss": 1.0473, + "step": 32572 + }, + { + "epoch": 2.41, + "learning_rate": 1.3604188892323504e-05, + "loss": 1.0179, + "step": 32573 + }, + { + "epoch": 2.41, + "learning_rate": 1.3603816825775662e-05, + "loss": 1.0083, + "step": 32574 + }, + { + "epoch": 2.41, + "learning_rate": 1.3603444753494199e-05, + "loss": 1.0618, + "step": 32575 + }, + { + "epoch": 2.41, + "learning_rate": 1.3603072675479707e-05, + "loss": 1.0048, + "step": 32576 + }, + { + "epoch": 2.41, + "learning_rate": 1.3602700591732778e-05, + "loss": 0.9938, + "step": 32577 + }, + { + "epoch": 2.41, + "learning_rate": 1.3602328502254004e-05, + "loss": 1.0654, + "step": 32578 + }, + { + "epoch": 2.41, + "learning_rate": 1.3601956407043974e-05, + "loss": 1.1007, + "step": 32579 + }, + { + "epoch": 2.41, + "learning_rate": 1.3601584306103285e-05, + "loss": 1.1275, + "step": 32580 + }, + { + "epoch": 2.41, + "learning_rate": 1.3601212199432529e-05, + "loss": 0.9644, + "step": 32581 + }, + { + "epoch": 2.41, + "learning_rate": 1.3600840087032295e-05, + "loss": 1.0081, + "step": 32582 + }, + { + "epoch": 2.41, + "learning_rate": 1.3600467968903175e-05, + "loss": 1.0445, + "step": 32583 + }, + { + "epoch": 2.41, + "learning_rate": 1.3600095845045762e-05, + "loss": 1.0352, + "step": 32584 + }, + { + "epoch": 2.41, + "learning_rate": 1.3599723715460649e-05, + "loss": 0.995, + "step": 32585 + }, + { + "epoch": 2.41, + "learning_rate": 1.3599351580148424e-05, + "loss": 0.9745, + "step": 32586 + }, + { + "epoch": 2.41, + "learning_rate": 1.3598979439109685e-05, + "loss": 0.9734, + "step": 32587 + }, + { + "epoch": 2.41, + "learning_rate": 1.359860729234502e-05, + "loss": 0.9772, + "step": 32588 + }, + { + "epoch": 2.41, + "learning_rate": 1.3598235139855023e-05, + "loss": 0.9449, + "step": 32589 + }, + { + "epoch": 2.41, + "learning_rate": 1.3597862981640286e-05, + "loss": 1.0228, + "step": 32590 + }, + { + "epoch": 2.41, + "learning_rate": 1.35974908177014e-05, + "loss": 1.011, + "step": 32591 + }, + { + "epoch": 2.41, + "learning_rate": 1.3597118648038956e-05, + "loss": 0.9586, + "step": 32592 + }, + { + "epoch": 2.41, + "learning_rate": 1.3596746472653548e-05, + "loss": 1.0112, + "step": 32593 + }, + { + "epoch": 2.41, + "learning_rate": 1.359637429154577e-05, + "loss": 1.0975, + "step": 32594 + }, + { + "epoch": 2.41, + "learning_rate": 1.359600210471621e-05, + "loss": 0.996, + "step": 32595 + }, + { + "epoch": 2.41, + "learning_rate": 1.3595629912165464e-05, + "loss": 1.0326, + "step": 32596 + }, + { + "epoch": 2.41, + "learning_rate": 1.359525771389412e-05, + "loss": 1.0205, + "step": 32597 + }, + { + "epoch": 2.41, + "learning_rate": 1.3594885509902775e-05, + "loss": 1.0084, + "step": 32598 + }, + { + "epoch": 2.41, + "learning_rate": 1.3594513300192016e-05, + "loss": 1.161, + "step": 32599 + }, + { + "epoch": 2.41, + "learning_rate": 1.359414108476244e-05, + "loss": 0.9165, + "step": 32600 + }, + { + "epoch": 2.41, + "learning_rate": 1.3593768863614636e-05, + "loss": 0.9656, + "step": 32601 + }, + { + "epoch": 2.41, + "learning_rate": 1.3593396636749197e-05, + "loss": 0.9816, + "step": 32602 + }, + { + "epoch": 2.41, + "learning_rate": 1.3593024404166713e-05, + "loss": 0.9713, + "step": 32603 + }, + { + "epoch": 2.41, + "learning_rate": 1.359265216586778e-05, + "loss": 0.9673, + "step": 32604 + }, + { + "epoch": 2.41, + "learning_rate": 1.3592279921852992e-05, + "loss": 0.8932, + "step": 32605 + }, + { + "epoch": 2.41, + "learning_rate": 1.3591907672122938e-05, + "loss": 0.9973, + "step": 32606 + }, + { + "epoch": 2.41, + "learning_rate": 1.3591535416678205e-05, + "loss": 1.0331, + "step": 32607 + }, + { + "epoch": 2.41, + "learning_rate": 1.3591163155519396e-05, + "loss": 0.9645, + "step": 32608 + }, + { + "epoch": 2.41, + "learning_rate": 1.3590790888647095e-05, + "loss": 0.9655, + "step": 32609 + }, + { + "epoch": 2.41, + "learning_rate": 1.3590418616061898e-05, + "loss": 1.0944, + "step": 32610 + }, + { + "epoch": 2.41, + "learning_rate": 1.3590046337764392e-05, + "loss": 1.0844, + "step": 32611 + }, + { + "epoch": 2.41, + "learning_rate": 1.358967405375518e-05, + "loss": 0.9526, + "step": 32612 + }, + { + "epoch": 2.41, + "learning_rate": 1.3589301764034843e-05, + "loss": 0.9711, + "step": 32613 + }, + { + "epoch": 2.41, + "learning_rate": 1.3588929468603982e-05, + "loss": 1.0378, + "step": 32614 + }, + { + "epoch": 2.41, + "learning_rate": 1.3588557167463186e-05, + "loss": 1.1094, + "step": 32615 + }, + { + "epoch": 2.41, + "learning_rate": 1.3588184860613045e-05, + "loss": 0.9676, + "step": 32616 + }, + { + "epoch": 2.41, + "learning_rate": 1.3587812548054154e-05, + "loss": 0.994, + "step": 32617 + }, + { + "epoch": 2.41, + "learning_rate": 1.3587440229787104e-05, + "loss": 0.991, + "step": 32618 + }, + { + "epoch": 2.41, + "learning_rate": 1.3587067905812486e-05, + "loss": 1.0131, + "step": 32619 + }, + { + "epoch": 2.41, + "learning_rate": 1.3586695576130897e-05, + "loss": 1.0408, + "step": 32620 + }, + { + "epoch": 2.41, + "learning_rate": 1.3586323240742925e-05, + "loss": 0.9912, + "step": 32621 + }, + { + "epoch": 2.41, + "learning_rate": 1.3585950899649165e-05, + "loss": 1.0328, + "step": 32622 + }, + { + "epoch": 2.41, + "learning_rate": 1.3585578552850208e-05, + "loss": 0.9905, + "step": 32623 + }, + { + "epoch": 2.41, + "learning_rate": 1.3585206200346646e-05, + "loss": 0.8036, + "step": 32624 + }, + { + "epoch": 2.41, + "learning_rate": 1.3584833842139077e-05, + "loss": 1.0903, + "step": 32625 + }, + { + "epoch": 2.41, + "learning_rate": 1.3584461478228083e-05, + "loss": 1.0312, + "step": 32626 + }, + { + "epoch": 2.41, + "learning_rate": 1.3584089108614261e-05, + "loss": 1.0029, + "step": 32627 + }, + { + "epoch": 2.41, + "learning_rate": 1.358371673329821e-05, + "loss": 1.0549, + "step": 32628 + }, + { + "epoch": 2.41, + "learning_rate": 1.3583344352280514e-05, + "loss": 1.0344, + "step": 32629 + }, + { + "epoch": 2.41, + "learning_rate": 1.358297196556177e-05, + "loss": 1.1311, + "step": 32630 + }, + { + "epoch": 2.41, + "learning_rate": 1.3582599573142567e-05, + "loss": 1.0171, + "step": 32631 + }, + { + "epoch": 2.41, + "learning_rate": 1.3582227175023501e-05, + "loss": 1.0571, + "step": 32632 + }, + { + "epoch": 2.41, + "learning_rate": 1.3581854771205159e-05, + "loss": 1.0678, + "step": 32633 + }, + { + "epoch": 2.41, + "learning_rate": 1.3581482361688143e-05, + "loss": 0.9846, + "step": 32634 + }, + { + "epoch": 2.41, + "learning_rate": 1.3581109946473037e-05, + "loss": 1.0686, + "step": 32635 + }, + { + "epoch": 2.41, + "learning_rate": 1.3580737525560435e-05, + "loss": 1.0616, + "step": 32636 + }, + { + "epoch": 2.41, + "learning_rate": 1.3580365098950932e-05, + "loss": 0.96, + "step": 32637 + }, + { + "epoch": 2.41, + "learning_rate": 1.3579992666645118e-05, + "loss": 1.0704, + "step": 32638 + }, + { + "epoch": 2.41, + "learning_rate": 1.357962022864359e-05, + "loss": 1.0274, + "step": 32639 + }, + { + "epoch": 2.41, + "learning_rate": 1.3579247784946934e-05, + "loss": 1.056, + "step": 32640 + }, + { + "epoch": 2.41, + "learning_rate": 1.3578875335555747e-05, + "loss": 0.9993, + "step": 32641 + }, + { + "epoch": 2.41, + "learning_rate": 1.3578502880470622e-05, + "loss": 0.9912, + "step": 32642 + }, + { + "epoch": 2.41, + "learning_rate": 1.3578130419692152e-05, + "loss": 0.9709, + "step": 32643 + }, + { + "epoch": 2.41, + "learning_rate": 1.3577757953220924e-05, + "loss": 1.0126, + "step": 32644 + }, + { + "epoch": 2.41, + "learning_rate": 1.3577385481057536e-05, + "loss": 0.9727, + "step": 32645 + }, + { + "epoch": 2.41, + "learning_rate": 1.3577013003202577e-05, + "loss": 0.9951, + "step": 32646 + }, + { + "epoch": 2.41, + "learning_rate": 1.3576640519656645e-05, + "loss": 0.9602, + "step": 32647 + }, + { + "epoch": 2.41, + "learning_rate": 1.3576268030420326e-05, + "loss": 1.0875, + "step": 32648 + }, + { + "epoch": 2.41, + "learning_rate": 1.3575895535494217e-05, + "loss": 0.9266, + "step": 32649 + }, + { + "epoch": 2.41, + "learning_rate": 1.3575523034878908e-05, + "loss": 1.0089, + "step": 32650 + }, + { + "epoch": 2.41, + "learning_rate": 1.3575150528574993e-05, + "loss": 0.9951, + "step": 32651 + }, + { + "epoch": 2.41, + "learning_rate": 1.3574778016583068e-05, + "loss": 1.0182, + "step": 32652 + }, + { + "epoch": 2.41, + "learning_rate": 1.357440549890372e-05, + "loss": 0.899, + "step": 32653 + }, + { + "epoch": 2.41, + "learning_rate": 1.3574032975537544e-05, + "loss": 1.0226, + "step": 32654 + }, + { + "epoch": 2.41, + "learning_rate": 1.3573660446485133e-05, + "loss": 0.9657, + "step": 32655 + }, + { + "epoch": 2.41, + "learning_rate": 1.3573287911747078e-05, + "loss": 1.1299, + "step": 32656 + }, + { + "epoch": 2.41, + "learning_rate": 1.3572915371323973e-05, + "loss": 0.9181, + "step": 32657 + }, + { + "epoch": 2.41, + "learning_rate": 1.3572542825216413e-05, + "loss": 1.0908, + "step": 32658 + }, + { + "epoch": 2.41, + "learning_rate": 1.3572170273424986e-05, + "loss": 0.9463, + "step": 32659 + }, + { + "epoch": 2.41, + "learning_rate": 1.3571797715950288e-05, + "loss": 0.9839, + "step": 32660 + }, + { + "epoch": 2.41, + "learning_rate": 1.3571425152792914e-05, + "loss": 0.9536, + "step": 32661 + }, + { + "epoch": 2.41, + "learning_rate": 1.357105258395345e-05, + "loss": 1.0026, + "step": 32662 + }, + { + "epoch": 2.41, + "learning_rate": 1.3570680009432495e-05, + "loss": 0.9975, + "step": 32663 + }, + { + "epoch": 2.41, + "learning_rate": 1.3570307429230636e-05, + "loss": 0.9247, + "step": 32664 + }, + { + "epoch": 2.41, + "learning_rate": 1.356993484334847e-05, + "loss": 1.0805, + "step": 32665 + }, + { + "epoch": 2.41, + "learning_rate": 1.3569562251786592e-05, + "loss": 1.0719, + "step": 32666 + }, + { + "epoch": 2.41, + "learning_rate": 1.356918965454559e-05, + "loss": 1.071, + "step": 32667 + }, + { + "epoch": 2.41, + "learning_rate": 1.3568817051626056e-05, + "loss": 1.1124, + "step": 32668 + }, + { + "epoch": 2.41, + "learning_rate": 1.3568444443028585e-05, + "loss": 0.9633, + "step": 32669 + }, + { + "epoch": 2.41, + "learning_rate": 1.356807182875377e-05, + "loss": 1.0054, + "step": 32670 + }, + { + "epoch": 2.41, + "learning_rate": 1.356769920880221e-05, + "loss": 0.9701, + "step": 32671 + }, + { + "epoch": 2.41, + "learning_rate": 1.3567326583174485e-05, + "loss": 0.999, + "step": 32672 + }, + { + "epoch": 2.41, + "learning_rate": 1.3566953951871198e-05, + "loss": 0.9855, + "step": 32673 + }, + { + "epoch": 2.41, + "learning_rate": 1.3566581314892934e-05, + "loss": 1.0208, + "step": 32674 + }, + { + "epoch": 2.41, + "learning_rate": 1.3566208672240295e-05, + "loss": 0.9375, + "step": 32675 + }, + { + "epoch": 2.41, + "learning_rate": 1.3565836023913863e-05, + "loss": 0.9999, + "step": 32676 + }, + { + "epoch": 2.41, + "learning_rate": 1.3565463369914243e-05, + "loss": 0.961, + "step": 32677 + }, + { + "epoch": 2.41, + "learning_rate": 1.3565090710242016e-05, + "loss": 1.0561, + "step": 32678 + }, + { + "epoch": 2.41, + "learning_rate": 1.3564718044897786e-05, + "loss": 1.0065, + "step": 32679 + }, + { + "epoch": 2.41, + "learning_rate": 1.3564345373882142e-05, + "loss": 1.011, + "step": 32680 + }, + { + "epoch": 2.41, + "learning_rate": 1.356397269719567e-05, + "loss": 0.9899, + "step": 32681 + }, + { + "epoch": 2.41, + "learning_rate": 1.3563600014838972e-05, + "loss": 0.9816, + "step": 32682 + }, + { + "epoch": 2.41, + "learning_rate": 1.3563227326812634e-05, + "loss": 0.9907, + "step": 32683 + }, + { + "epoch": 2.41, + "learning_rate": 1.3562854633117254e-05, + "loss": 1.0081, + "step": 32684 + }, + { + "epoch": 2.42, + "learning_rate": 1.3562481933753425e-05, + "loss": 1.0942, + "step": 32685 + }, + { + "epoch": 2.42, + "learning_rate": 1.3562109228721737e-05, + "loss": 0.9953, + "step": 32686 + }, + { + "epoch": 2.42, + "learning_rate": 1.3561736518022783e-05, + "loss": 1.1188, + "step": 32687 + }, + { + "epoch": 2.42, + "learning_rate": 1.3561363801657158e-05, + "loss": 1.0436, + "step": 32688 + }, + { + "epoch": 2.42, + "learning_rate": 1.3560991079625454e-05, + "loss": 1.0009, + "step": 32689 + }, + { + "epoch": 2.42, + "learning_rate": 1.3560618351928266e-05, + "loss": 1.0851, + "step": 32690 + }, + { + "epoch": 2.42, + "learning_rate": 1.3560245618566182e-05, + "loss": 1.0021, + "step": 32691 + }, + { + "epoch": 2.42, + "learning_rate": 1.3559872879539802e-05, + "loss": 1.055, + "step": 32692 + }, + { + "epoch": 2.42, + "learning_rate": 1.355950013484971e-05, + "loss": 1.0986, + "step": 32693 + }, + { + "epoch": 2.42, + "learning_rate": 1.355912738449651e-05, + "loss": 0.9703, + "step": 32694 + }, + { + "epoch": 2.42, + "learning_rate": 1.3558754628480783e-05, + "loss": 1.0107, + "step": 32695 + }, + { + "epoch": 2.42, + "learning_rate": 1.3558381866803133e-05, + "loss": 1.0946, + "step": 32696 + }, + { + "epoch": 2.42, + "learning_rate": 1.3558009099464147e-05, + "loss": 0.9914, + "step": 32697 + }, + { + "epoch": 2.42, + "learning_rate": 1.3557636326464418e-05, + "loss": 1.0269, + "step": 32698 + }, + { + "epoch": 2.42, + "learning_rate": 1.3557263547804544e-05, + "loss": 1.0748, + "step": 32699 + }, + { + "epoch": 2.42, + "learning_rate": 1.3556890763485114e-05, + "loss": 1.0017, + "step": 32700 + }, + { + "epoch": 2.42, + "learning_rate": 1.3556517973506718e-05, + "loss": 1.018, + "step": 32701 + }, + { + "epoch": 2.42, + "learning_rate": 1.3556145177869957e-05, + "loss": 1.048, + "step": 32702 + }, + { + "epoch": 2.42, + "learning_rate": 1.3555772376575415e-05, + "loss": 0.9696, + "step": 32703 + }, + { + "epoch": 2.42, + "learning_rate": 1.3555399569623695e-05, + "loss": 1.0295, + "step": 32704 + }, + { + "epoch": 2.42, + "learning_rate": 1.3555026757015382e-05, + "loss": 1.1441, + "step": 32705 + }, + { + "epoch": 2.42, + "learning_rate": 1.3554653938751075e-05, + "loss": 0.9578, + "step": 32706 + }, + { + "epoch": 2.42, + "learning_rate": 1.355428111483136e-05, + "loss": 1.026, + "step": 32707 + }, + { + "epoch": 2.42, + "learning_rate": 1.355390828525684e-05, + "loss": 1.085, + "step": 32708 + }, + { + "epoch": 2.42, + "learning_rate": 1.3553535450028099e-05, + "loss": 1.0789, + "step": 32709 + }, + { + "epoch": 2.42, + "learning_rate": 1.3553162609145737e-05, + "loss": 0.9619, + "step": 32710 + }, + { + "epoch": 2.42, + "learning_rate": 1.355278976261034e-05, + "loss": 0.9978, + "step": 32711 + }, + { + "epoch": 2.42, + "learning_rate": 1.355241691042251e-05, + "loss": 0.9668, + "step": 32712 + }, + { + "epoch": 2.42, + "learning_rate": 1.3552044052582832e-05, + "loss": 0.9431, + "step": 32713 + }, + { + "epoch": 2.42, + "learning_rate": 1.3551671189091905e-05, + "loss": 1.0401, + "step": 32714 + }, + { + "epoch": 2.42, + "learning_rate": 1.3551298319950317e-05, + "loss": 0.9587, + "step": 32715 + }, + { + "epoch": 2.42, + "learning_rate": 1.3550925445158666e-05, + "loss": 0.9824, + "step": 32716 + }, + { + "epoch": 2.42, + "learning_rate": 1.3550552564717543e-05, + "loss": 1.0513, + "step": 32717 + }, + { + "epoch": 2.42, + "learning_rate": 1.3550179678627544e-05, + "loss": 0.9186, + "step": 32718 + }, + { + "epoch": 2.42, + "learning_rate": 1.3549806786889258e-05, + "loss": 0.9975, + "step": 32719 + }, + { + "epoch": 2.42, + "learning_rate": 1.354943388950328e-05, + "loss": 0.9279, + "step": 32720 + }, + { + "epoch": 2.42, + "learning_rate": 1.35490609864702e-05, + "loss": 1.156, + "step": 32721 + }, + { + "epoch": 2.42, + "learning_rate": 1.3548688077790622e-05, + "loss": 0.9501, + "step": 32722 + }, + { + "epoch": 2.42, + "learning_rate": 1.3548315163465127e-05, + "loss": 1.0379, + "step": 32723 + }, + { + "epoch": 2.42, + "learning_rate": 1.3547942243494316e-05, + "loss": 1.0433, + "step": 32724 + }, + { + "epoch": 2.42, + "learning_rate": 1.3547569317878777e-05, + "loss": 1.012, + "step": 32725 + }, + { + "epoch": 2.42, + "learning_rate": 1.3547196386619107e-05, + "loss": 0.9531, + "step": 32726 + }, + { + "epoch": 2.42, + "learning_rate": 1.35468234497159e-05, + "loss": 0.9653, + "step": 32727 + }, + { + "epoch": 2.42, + "learning_rate": 1.3546450507169745e-05, + "loss": 1.014, + "step": 32728 + }, + { + "epoch": 2.42, + "learning_rate": 1.3546077558981241e-05, + "loss": 1.1186, + "step": 32729 + }, + { + "epoch": 2.42, + "learning_rate": 1.3545704605150977e-05, + "loss": 0.9731, + "step": 32730 + }, + { + "epoch": 2.42, + "learning_rate": 1.3545331645679547e-05, + "loss": 1.0629, + "step": 32731 + }, + { + "epoch": 2.42, + "learning_rate": 1.3544958680567546e-05, + "loss": 1.0001, + "step": 32732 + }, + { + "epoch": 2.42, + "learning_rate": 1.3544585709815566e-05, + "loss": 0.9487, + "step": 32733 + }, + { + "epoch": 2.42, + "learning_rate": 1.35442127334242e-05, + "loss": 1.0546, + "step": 32734 + }, + { + "epoch": 2.42, + "learning_rate": 1.3543839751394041e-05, + "loss": 1.0241, + "step": 32735 + }, + { + "epoch": 2.42, + "learning_rate": 1.3543466763725689e-05, + "loss": 1.022, + "step": 32736 + }, + { + "epoch": 2.42, + "learning_rate": 1.3543093770419729e-05, + "loss": 1.0506, + "step": 32737 + }, + { + "epoch": 2.42, + "learning_rate": 1.3542720771476754e-05, + "loss": 0.9475, + "step": 32738 + }, + { + "epoch": 2.42, + "learning_rate": 1.3542347766897365e-05, + "loss": 0.9776, + "step": 32739 + }, + { + "epoch": 2.42, + "learning_rate": 1.354197475668215e-05, + "loss": 0.9829, + "step": 32740 + }, + { + "epoch": 2.42, + "learning_rate": 1.3541601740831704e-05, + "loss": 0.9483, + "step": 32741 + }, + { + "epoch": 2.42, + "learning_rate": 1.3541228719346622e-05, + "loss": 0.9552, + "step": 32742 + }, + { + "epoch": 2.42, + "learning_rate": 1.3540855692227494e-05, + "loss": 1.0233, + "step": 32743 + }, + { + "epoch": 2.42, + "learning_rate": 1.3540482659474914e-05, + "loss": 0.9598, + "step": 32744 + }, + { + "epoch": 2.42, + "learning_rate": 1.354010962108948e-05, + "loss": 1.0685, + "step": 32745 + }, + { + "epoch": 2.42, + "learning_rate": 1.3539736577071782e-05, + "loss": 1.0283, + "step": 32746 + }, + { + "epoch": 2.42, + "learning_rate": 1.3539363527422413e-05, + "loss": 1.0394, + "step": 32747 + }, + { + "epoch": 2.42, + "learning_rate": 1.3538990472141967e-05, + "loss": 1.0909, + "step": 32748 + }, + { + "epoch": 2.42, + "learning_rate": 1.3538617411231036e-05, + "loss": 1.0635, + "step": 32749 + }, + { + "epoch": 2.42, + "learning_rate": 1.3538244344690216e-05, + "loss": 0.9576, + "step": 32750 + }, + { + "epoch": 2.42, + "learning_rate": 1.3537871272520102e-05, + "loss": 1.0497, + "step": 32751 + }, + { + "epoch": 2.42, + "learning_rate": 1.3537498194721284e-05, + "loss": 1.0258, + "step": 32752 + }, + { + "epoch": 2.42, + "learning_rate": 1.3537125111294359e-05, + "loss": 1.0189, + "step": 32753 + }, + { + "epoch": 2.42, + "learning_rate": 1.3536752022239915e-05, + "loss": 0.9028, + "step": 32754 + }, + { + "epoch": 2.42, + "learning_rate": 1.3536378927558553e-05, + "loss": 0.96, + "step": 32755 + }, + { + "epoch": 2.42, + "learning_rate": 1.353600582725086e-05, + "loss": 0.9151, + "step": 32756 + }, + { + "epoch": 2.42, + "learning_rate": 1.3535632721317432e-05, + "loss": 1.0133, + "step": 32757 + }, + { + "epoch": 2.42, + "learning_rate": 1.3535259609758865e-05, + "loss": 0.9761, + "step": 32758 + }, + { + "epoch": 2.42, + "learning_rate": 1.3534886492575749e-05, + "loss": 0.9718, + "step": 32759 + }, + { + "epoch": 2.42, + "learning_rate": 1.3534513369768678e-05, + "loss": 1.0455, + "step": 32760 + }, + { + "epoch": 2.42, + "learning_rate": 1.353414024133825e-05, + "loss": 0.9253, + "step": 32761 + }, + { + "epoch": 2.42, + "learning_rate": 1.3533767107285051e-05, + "loss": 0.9918, + "step": 32762 + }, + { + "epoch": 2.42, + "learning_rate": 1.3533393967609682e-05, + "loss": 1.0184, + "step": 32763 + }, + { + "epoch": 2.42, + "learning_rate": 1.3533020822312737e-05, + "loss": 1.0777, + "step": 32764 + }, + { + "epoch": 2.42, + "learning_rate": 1.3532647671394803e-05, + "loss": 1.0154, + "step": 32765 + }, + { + "epoch": 2.42, + "learning_rate": 1.3532274514856476e-05, + "loss": 0.9803, + "step": 32766 + }, + { + "epoch": 2.42, + "learning_rate": 1.3531901352698354e-05, + "loss": 1.061, + "step": 32767 + }, + { + "epoch": 2.42, + "learning_rate": 1.3531528184921024e-05, + "loss": 1.0745, + "step": 32768 + }, + { + "epoch": 2.42, + "learning_rate": 1.3531155011525083e-05, + "loss": 0.8545, + "step": 32769 + }, + { + "epoch": 2.42, + "learning_rate": 1.3530781832511125e-05, + "loss": 1.0013, + "step": 32770 + }, + { + "epoch": 2.42, + "learning_rate": 1.3530408647879747e-05, + "loss": 1.0504, + "step": 32771 + }, + { + "epoch": 2.42, + "learning_rate": 1.3530035457631537e-05, + "loss": 1.0508, + "step": 32772 + }, + { + "epoch": 2.42, + "learning_rate": 1.3529662261767092e-05, + "loss": 1.0637, + "step": 32773 + }, + { + "epoch": 2.42, + "learning_rate": 1.3529289060287005e-05, + "loss": 1.0335, + "step": 32774 + }, + { + "epoch": 2.42, + "learning_rate": 1.352891585319187e-05, + "loss": 1.0075, + "step": 32775 + }, + { + "epoch": 2.42, + "learning_rate": 1.3528542640482278e-05, + "loss": 0.9895, + "step": 32776 + }, + { + "epoch": 2.42, + "learning_rate": 1.3528169422158828e-05, + "loss": 0.9774, + "step": 32777 + }, + { + "epoch": 2.42, + "learning_rate": 1.3527796198222107e-05, + "loss": 1.0112, + "step": 32778 + }, + { + "epoch": 2.42, + "learning_rate": 1.3527422968672717e-05, + "loss": 1.0974, + "step": 32779 + }, + { + "epoch": 2.42, + "learning_rate": 1.3527049733511244e-05, + "loss": 0.9555, + "step": 32780 + }, + { + "epoch": 2.42, + "learning_rate": 1.352667649273829e-05, + "loss": 1.0251, + "step": 32781 + }, + { + "epoch": 2.42, + "learning_rate": 1.352630324635444e-05, + "loss": 0.9558, + "step": 32782 + }, + { + "epoch": 2.42, + "learning_rate": 1.3525929994360291e-05, + "loss": 1.0186, + "step": 32783 + }, + { + "epoch": 2.42, + "learning_rate": 1.3525556736756441e-05, + "loss": 1.0337, + "step": 32784 + }, + { + "epoch": 2.42, + "learning_rate": 1.352518347354348e-05, + "loss": 1.088, + "step": 32785 + }, + { + "epoch": 2.42, + "learning_rate": 1.3524810204722004e-05, + "loss": 1.0723, + "step": 32786 + }, + { + "epoch": 2.42, + "learning_rate": 1.3524436930292602e-05, + "loss": 0.9782, + "step": 32787 + }, + { + "epoch": 2.42, + "learning_rate": 1.3524063650255874e-05, + "loss": 1.0106, + "step": 32788 + }, + { + "epoch": 2.42, + "learning_rate": 1.3523690364612408e-05, + "loss": 1.077, + "step": 32789 + }, + { + "epoch": 2.42, + "learning_rate": 1.3523317073362802e-05, + "loss": 1.0028, + "step": 32790 + }, + { + "epoch": 2.42, + "learning_rate": 1.3522943776507652e-05, + "loss": 1.042, + "step": 32791 + }, + { + "epoch": 2.42, + "learning_rate": 1.3522570474047547e-05, + "loss": 1.037, + "step": 32792 + }, + { + "epoch": 2.42, + "learning_rate": 1.3522197165983081e-05, + "loss": 0.9787, + "step": 32793 + }, + { + "epoch": 2.42, + "learning_rate": 1.3521823852314854e-05, + "loss": 1.0139, + "step": 32794 + }, + { + "epoch": 2.42, + "learning_rate": 1.3521450533043452e-05, + "loss": 0.9594, + "step": 32795 + }, + { + "epoch": 2.42, + "learning_rate": 1.3521077208169472e-05, + "loss": 0.9992, + "step": 32796 + }, + { + "epoch": 2.42, + "learning_rate": 1.3520703877693511e-05, + "loss": 0.9111, + "step": 32797 + }, + { + "epoch": 2.42, + "learning_rate": 1.352033054161616e-05, + "loss": 0.9868, + "step": 32798 + }, + { + "epoch": 2.42, + "learning_rate": 1.3519957199938012e-05, + "loss": 1.0, + "step": 32799 + }, + { + "epoch": 2.42, + "learning_rate": 1.3519583852659665e-05, + "loss": 0.9935, + "step": 32800 + }, + { + "epoch": 2.42, + "learning_rate": 1.3519210499781707e-05, + "loss": 0.9669, + "step": 32801 + }, + { + "epoch": 2.42, + "learning_rate": 1.3518837141304741e-05, + "loss": 0.9373, + "step": 32802 + }, + { + "epoch": 2.42, + "learning_rate": 1.3518463777229352e-05, + "loss": 0.9806, + "step": 32803 + }, + { + "epoch": 2.42, + "learning_rate": 1.3518090407556136e-05, + "loss": 0.9748, + "step": 32804 + }, + { + "epoch": 2.42, + "learning_rate": 1.3517717032285688e-05, + "loss": 0.9806, + "step": 32805 + }, + { + "epoch": 2.42, + "learning_rate": 1.3517343651418606e-05, + "loss": 1.0643, + "step": 32806 + }, + { + "epoch": 2.42, + "learning_rate": 1.3516970264955478e-05, + "loss": 1.0318, + "step": 32807 + }, + { + "epoch": 2.42, + "learning_rate": 1.3516596872896903e-05, + "loss": 0.9836, + "step": 32808 + }, + { + "epoch": 2.42, + "learning_rate": 1.351622347524347e-05, + "loss": 1.0055, + "step": 32809 + }, + { + "epoch": 2.42, + "learning_rate": 1.3515850071995778e-05, + "loss": 1.0153, + "step": 32810 + }, + { + "epoch": 2.42, + "learning_rate": 1.351547666315442e-05, + "loss": 0.9701, + "step": 32811 + }, + { + "epoch": 2.42, + "learning_rate": 1.3515103248719985e-05, + "loss": 1.0287, + "step": 32812 + }, + { + "epoch": 2.42, + "learning_rate": 1.3514729828693072e-05, + "loss": 1.1254, + "step": 32813 + }, + { + "epoch": 2.42, + "learning_rate": 1.3514356403074276e-05, + "loss": 1.1653, + "step": 32814 + }, + { + "epoch": 2.42, + "learning_rate": 1.3513982971864188e-05, + "loss": 1.0448, + "step": 32815 + }, + { + "epoch": 2.42, + "learning_rate": 1.3513609535063403e-05, + "loss": 0.9401, + "step": 32816 + }, + { + "epoch": 2.42, + "learning_rate": 1.3513236092672518e-05, + "loss": 1.0055, + "step": 32817 + }, + { + "epoch": 2.42, + "learning_rate": 1.3512862644692121e-05, + "loss": 1.0353, + "step": 32818 + }, + { + "epoch": 2.42, + "learning_rate": 1.3512489191122811e-05, + "loss": 1.1076, + "step": 32819 + }, + { + "epoch": 2.43, + "learning_rate": 1.3512115731965183e-05, + "loss": 1.0996, + "step": 32820 + }, + { + "epoch": 2.43, + "learning_rate": 1.3511742267219828e-05, + "loss": 0.9871, + "step": 32821 + }, + { + "epoch": 2.43, + "learning_rate": 1.3511368796887343e-05, + "loss": 1.0059, + "step": 32822 + }, + { + "epoch": 2.43, + "learning_rate": 1.3510995320968317e-05, + "loss": 0.9951, + "step": 32823 + }, + { + "epoch": 2.43, + "learning_rate": 1.3510621839463349e-05, + "loss": 1.1176, + "step": 32824 + }, + { + "epoch": 2.43, + "learning_rate": 1.351024835237303e-05, + "loss": 0.9617, + "step": 32825 + }, + { + "epoch": 2.43, + "learning_rate": 1.350987485969796e-05, + "loss": 0.9572, + "step": 32826 + }, + { + "epoch": 2.43, + "learning_rate": 1.3509501361438726e-05, + "loss": 0.9298, + "step": 32827 + }, + { + "epoch": 2.43, + "learning_rate": 1.3509127857595927e-05, + "loss": 1.0477, + "step": 32828 + }, + { + "epoch": 2.43, + "learning_rate": 1.3508754348170155e-05, + "loss": 1.0617, + "step": 32829 + }, + { + "epoch": 2.43, + "learning_rate": 1.3508380833162006e-05, + "loss": 1.1292, + "step": 32830 + }, + { + "epoch": 2.43, + "learning_rate": 1.3508007312572074e-05, + "loss": 1.1575, + "step": 32831 + }, + { + "epoch": 2.43, + "learning_rate": 1.3507633786400951e-05, + "loss": 1.0013, + "step": 32832 + }, + { + "epoch": 2.43, + "learning_rate": 1.3507260254649235e-05, + "loss": 0.9567, + "step": 32833 + }, + { + "epoch": 2.43, + "learning_rate": 1.3506886717317513e-05, + "loss": 1.0781, + "step": 32834 + }, + { + "epoch": 2.43, + "learning_rate": 1.350651317440639e-05, + "loss": 1.0805, + "step": 32835 + }, + { + "epoch": 2.43, + "learning_rate": 1.3506139625916453e-05, + "loss": 0.9129, + "step": 32836 + }, + { + "epoch": 2.43, + "learning_rate": 1.3505766071848296e-05, + "loss": 0.9423, + "step": 32837 + }, + { + "epoch": 2.43, + "learning_rate": 1.3505392512202518e-05, + "loss": 0.999, + "step": 32838 + }, + { + "epoch": 2.43, + "learning_rate": 1.350501894697971e-05, + "loss": 1.0079, + "step": 32839 + }, + { + "epoch": 2.43, + "learning_rate": 1.3504645376180467e-05, + "loss": 1.0127, + "step": 32840 + }, + { + "epoch": 2.43, + "learning_rate": 1.3504271799805385e-05, + "loss": 1.012, + "step": 32841 + }, + { + "epoch": 2.43, + "learning_rate": 1.3503898217855052e-05, + "loss": 1.1292, + "step": 32842 + }, + { + "epoch": 2.43, + "learning_rate": 1.3503524630330072e-05, + "loss": 0.9811, + "step": 32843 + }, + { + "epoch": 2.43, + "learning_rate": 1.350315103723103e-05, + "loss": 0.8923, + "step": 32844 + }, + { + "epoch": 2.43, + "learning_rate": 1.3502777438558528e-05, + "loss": 0.9137, + "step": 32845 + }, + { + "epoch": 2.43, + "learning_rate": 1.3502403834313156e-05, + "loss": 1.0753, + "step": 32846 + }, + { + "epoch": 2.43, + "learning_rate": 1.3502030224495508e-05, + "loss": 1.0492, + "step": 32847 + }, + { + "epoch": 2.43, + "learning_rate": 1.3501656609106183e-05, + "loss": 0.8865, + "step": 32848 + }, + { + "epoch": 2.43, + "learning_rate": 1.3501282988145775e-05, + "loss": 1.0271, + "step": 32849 + }, + { + "epoch": 2.43, + "learning_rate": 1.3500909361614868e-05, + "loss": 1.0444, + "step": 32850 + }, + { + "epoch": 2.43, + "learning_rate": 1.3500535729514071e-05, + "loss": 0.9962, + "step": 32851 + }, + { + "epoch": 2.43, + "learning_rate": 1.3500162091843966e-05, + "loss": 0.9887, + "step": 32852 + }, + { + "epoch": 2.43, + "learning_rate": 1.349978844860516e-05, + "loss": 0.9565, + "step": 32853 + }, + { + "epoch": 2.43, + "learning_rate": 1.3499414799798234e-05, + "loss": 1.0015, + "step": 32854 + }, + { + "epoch": 2.43, + "learning_rate": 1.3499041145423793e-05, + "loss": 0.9856, + "step": 32855 + }, + { + "epoch": 2.43, + "learning_rate": 1.3498667485482426e-05, + "loss": 0.9936, + "step": 32856 + }, + { + "epoch": 2.43, + "learning_rate": 1.349829381997473e-05, + "loss": 1.0828, + "step": 32857 + }, + { + "epoch": 2.43, + "learning_rate": 1.34979201489013e-05, + "loss": 0.8979, + "step": 32858 + }, + { + "epoch": 2.43, + "learning_rate": 1.3497546472262728e-05, + "loss": 0.933, + "step": 32859 + }, + { + "epoch": 2.43, + "learning_rate": 1.3497172790059607e-05, + "loss": 1.0799, + "step": 32860 + }, + { + "epoch": 2.43, + "learning_rate": 1.3496799102292539e-05, + "loss": 0.9841, + "step": 32861 + }, + { + "epoch": 2.43, + "learning_rate": 1.3496425408962108e-05, + "loss": 0.9995, + "step": 32862 + }, + { + "epoch": 2.43, + "learning_rate": 1.3496051710068918e-05, + "loss": 1.0443, + "step": 32863 + }, + { + "epoch": 2.43, + "learning_rate": 1.349567800561356e-05, + "loss": 1.1411, + "step": 32864 + }, + { + "epoch": 2.43, + "learning_rate": 1.3495304295596628e-05, + "loss": 0.9285, + "step": 32865 + }, + { + "epoch": 2.43, + "learning_rate": 1.3494930580018715e-05, + "loss": 1.0478, + "step": 32866 + }, + { + "epoch": 2.43, + "learning_rate": 1.3494556858880422e-05, + "loss": 1.0008, + "step": 32867 + }, + { + "epoch": 2.43, + "learning_rate": 1.3494183132182333e-05, + "loss": 1.0449, + "step": 32868 + }, + { + "epoch": 2.43, + "learning_rate": 1.3493809399925053e-05, + "loss": 0.9412, + "step": 32869 + }, + { + "epoch": 2.43, + "learning_rate": 1.3493435662109166e-05, + "loss": 1.0733, + "step": 32870 + }, + { + "epoch": 2.43, + "learning_rate": 1.349306191873528e-05, + "loss": 0.9187, + "step": 32871 + }, + { + "epoch": 2.43, + "learning_rate": 1.3492688169803978e-05, + "loss": 0.9314, + "step": 32872 + }, + { + "epoch": 2.43, + "learning_rate": 1.3492314415315861e-05, + "loss": 1.0041, + "step": 32873 + }, + { + "epoch": 2.43, + "learning_rate": 1.349194065527152e-05, + "loss": 0.9631, + "step": 32874 + }, + { + "epoch": 2.43, + "learning_rate": 1.3491566889671558e-05, + "loss": 0.9829, + "step": 32875 + }, + { + "epoch": 2.43, + "learning_rate": 1.3491193118516555e-05, + "loss": 0.9962, + "step": 32876 + }, + { + "epoch": 2.43, + "learning_rate": 1.3490819341807115e-05, + "loss": 0.8946, + "step": 32877 + }, + { + "epoch": 2.43, + "learning_rate": 1.3490445559543836e-05, + "loss": 0.9754, + "step": 32878 + }, + { + "epoch": 2.43, + "learning_rate": 1.3490071771727302e-05, + "loss": 1.0167, + "step": 32879 + }, + { + "epoch": 2.43, + "learning_rate": 1.348969797835812e-05, + "loss": 0.9585, + "step": 32880 + }, + { + "epoch": 2.43, + "learning_rate": 1.3489324179436876e-05, + "loss": 1.0939, + "step": 32881 + }, + { + "epoch": 2.43, + "learning_rate": 1.3488950374964165e-05, + "loss": 0.9351, + "step": 32882 + }, + { + "epoch": 2.43, + "learning_rate": 1.3488576564940585e-05, + "loss": 1.1186, + "step": 32883 + }, + { + "epoch": 2.43, + "learning_rate": 1.348820274936673e-05, + "loss": 0.9527, + "step": 32884 + }, + { + "epoch": 2.43, + "learning_rate": 1.3487828928243195e-05, + "loss": 0.989, + "step": 32885 + }, + { + "epoch": 2.43, + "learning_rate": 1.3487455101570575e-05, + "loss": 0.965, + "step": 32886 + }, + { + "epoch": 2.43, + "learning_rate": 1.348708126934946e-05, + "loss": 1.1067, + "step": 32887 + }, + { + "epoch": 2.43, + "learning_rate": 1.3486707431580453e-05, + "loss": 0.9948, + "step": 32888 + }, + { + "epoch": 2.43, + "learning_rate": 1.348633358826414e-05, + "loss": 0.9778, + "step": 32889 + }, + { + "epoch": 2.43, + "learning_rate": 1.3485959739401122e-05, + "loss": 1.0607, + "step": 32890 + }, + { + "epoch": 2.43, + "learning_rate": 1.3485585884991993e-05, + "loss": 0.9605, + "step": 32891 + }, + { + "epoch": 2.43, + "learning_rate": 1.3485212025037346e-05, + "loss": 1.0379, + "step": 32892 + }, + { + "epoch": 2.43, + "learning_rate": 1.3484838159537774e-05, + "loss": 1.0431, + "step": 32893 + }, + { + "epoch": 2.43, + "learning_rate": 1.3484464288493876e-05, + "loss": 1.046, + "step": 32894 + }, + { + "epoch": 2.43, + "learning_rate": 1.348409041190625e-05, + "loss": 1.0434, + "step": 32895 + }, + { + "epoch": 2.43, + "learning_rate": 1.3483716529775481e-05, + "loss": 1.1152, + "step": 32896 + }, + { + "epoch": 2.43, + "learning_rate": 1.3483342642102169e-05, + "loss": 1.0469, + "step": 32897 + }, + { + "epoch": 2.43, + "learning_rate": 1.3482968748886911e-05, + "loss": 0.9906, + "step": 32898 + }, + { + "epoch": 2.43, + "learning_rate": 1.3482594850130295e-05, + "loss": 0.9696, + "step": 32899 + }, + { + "epoch": 2.43, + "learning_rate": 1.3482220945832926e-05, + "loss": 1.0389, + "step": 32900 + }, + { + "epoch": 2.43, + "learning_rate": 1.348184703599539e-05, + "loss": 0.8996, + "step": 32901 + }, + { + "epoch": 2.43, + "learning_rate": 1.3481473120618285e-05, + "loss": 0.9929, + "step": 32902 + }, + { + "epoch": 2.43, + "learning_rate": 1.3481099199702206e-05, + "loss": 1.0616, + "step": 32903 + }, + { + "epoch": 2.43, + "learning_rate": 1.348072527324775e-05, + "loss": 0.9507, + "step": 32904 + }, + { + "epoch": 2.43, + "learning_rate": 1.348035134125551e-05, + "loss": 1.1111, + "step": 32905 + }, + { + "epoch": 2.43, + "learning_rate": 1.3479977403726081e-05, + "loss": 1.0291, + "step": 32906 + }, + { + "epoch": 2.43, + "learning_rate": 1.3479603460660056e-05, + "loss": 0.9929, + "step": 32907 + }, + { + "epoch": 2.43, + "learning_rate": 1.3479229512058031e-05, + "loss": 0.9929, + "step": 32908 + }, + { + "epoch": 2.43, + "learning_rate": 1.3478855557920603e-05, + "loss": 0.8892, + "step": 32909 + }, + { + "epoch": 2.43, + "learning_rate": 1.3478481598248365e-05, + "loss": 1.041, + "step": 32910 + }, + { + "epoch": 2.43, + "learning_rate": 1.3478107633041913e-05, + "loss": 1.0545, + "step": 32911 + }, + { + "epoch": 2.43, + "learning_rate": 1.3477733662301841e-05, + "loss": 0.9889, + "step": 32912 + }, + { + "epoch": 2.43, + "learning_rate": 1.3477359686028745e-05, + "loss": 1.0663, + "step": 32913 + }, + { + "epoch": 2.43, + "learning_rate": 1.3476985704223222e-05, + "loss": 0.9149, + "step": 32914 + }, + { + "epoch": 2.43, + "learning_rate": 1.3476611716885861e-05, + "loss": 0.9113, + "step": 32915 + }, + { + "epoch": 2.43, + "learning_rate": 1.3476237724017262e-05, + "loss": 0.93, + "step": 32916 + }, + { + "epoch": 2.43, + "learning_rate": 1.3475863725618018e-05, + "loss": 1.0781, + "step": 32917 + }, + { + "epoch": 2.43, + "learning_rate": 1.3475489721688724e-05, + "loss": 0.9696, + "step": 32918 + }, + { + "epoch": 2.43, + "learning_rate": 1.3475115712229973e-05, + "loss": 1.0833, + "step": 32919 + }, + { + "epoch": 2.43, + "learning_rate": 1.3474741697242369e-05, + "loss": 1.0376, + "step": 32920 + }, + { + "epoch": 2.43, + "learning_rate": 1.3474367676726496e-05, + "loss": 1.0065, + "step": 32921 + }, + { + "epoch": 2.43, + "learning_rate": 1.3473993650682956e-05, + "loss": 0.9726, + "step": 32922 + }, + { + "epoch": 2.43, + "learning_rate": 1.3473619619112342e-05, + "loss": 0.8998, + "step": 32923 + }, + { + "epoch": 2.43, + "learning_rate": 1.3473245582015249e-05, + "loss": 1.0068, + "step": 32924 + }, + { + "epoch": 2.43, + "learning_rate": 1.347287153939227e-05, + "loss": 0.9508, + "step": 32925 + }, + { + "epoch": 2.43, + "learning_rate": 1.3472497491244e-05, + "loss": 0.9972, + "step": 32926 + }, + { + "epoch": 2.43, + "learning_rate": 1.3472123437571041e-05, + "loss": 1.0431, + "step": 32927 + }, + { + "epoch": 2.43, + "learning_rate": 1.3471749378373982e-05, + "loss": 0.9013, + "step": 32928 + }, + { + "epoch": 2.43, + "learning_rate": 1.3471375313653416e-05, + "loss": 0.9091, + "step": 32929 + }, + { + "epoch": 2.43, + "learning_rate": 1.3471001243409947e-05, + "loss": 0.9382, + "step": 32930 + }, + { + "epoch": 2.43, + "learning_rate": 1.347062716764416e-05, + "loss": 1.0113, + "step": 32931 + }, + { + "epoch": 2.43, + "learning_rate": 1.3470253086356658e-05, + "loss": 1.06, + "step": 32932 + }, + { + "epoch": 2.43, + "learning_rate": 1.3469878999548032e-05, + "loss": 1.0678, + "step": 32933 + }, + { + "epoch": 2.43, + "learning_rate": 1.3469504907218875e-05, + "loss": 0.9298, + "step": 32934 + }, + { + "epoch": 2.43, + "learning_rate": 1.346913080936979e-05, + "loss": 0.9868, + "step": 32935 + }, + { + "epoch": 2.43, + "learning_rate": 1.3468756706001365e-05, + "loss": 1.0085, + "step": 32936 + }, + { + "epoch": 2.43, + "learning_rate": 1.34683825971142e-05, + "loss": 1.0036, + "step": 32937 + }, + { + "epoch": 2.43, + "learning_rate": 1.3468008482708883e-05, + "loss": 1.0502, + "step": 32938 + }, + { + "epoch": 2.43, + "learning_rate": 1.346763436278602e-05, + "loss": 1.0404, + "step": 32939 + }, + { + "epoch": 2.43, + "learning_rate": 1.3467260237346195e-05, + "loss": 1.0734, + "step": 32940 + }, + { + "epoch": 2.43, + "learning_rate": 1.3466886106390013e-05, + "loss": 0.9627, + "step": 32941 + }, + { + "epoch": 2.43, + "learning_rate": 1.3466511969918065e-05, + "loss": 1.0652, + "step": 32942 + }, + { + "epoch": 2.43, + "learning_rate": 1.3466137827930946e-05, + "loss": 1.045, + "step": 32943 + }, + { + "epoch": 2.43, + "learning_rate": 1.3465763680429247e-05, + "loss": 0.9784, + "step": 32944 + }, + { + "epoch": 2.43, + "learning_rate": 1.346538952741357e-05, + "loss": 0.994, + "step": 32945 + }, + { + "epoch": 2.43, + "learning_rate": 1.3465015368884507e-05, + "loss": 1.0673, + "step": 32946 + }, + { + "epoch": 2.43, + "learning_rate": 1.3464641204842657e-05, + "loss": 1.1449, + "step": 32947 + }, + { + "epoch": 2.43, + "learning_rate": 1.346426703528861e-05, + "loss": 0.9231, + "step": 32948 + }, + { + "epoch": 2.43, + "learning_rate": 1.3463892860222964e-05, + "loss": 0.9141, + "step": 32949 + }, + { + "epoch": 2.43, + "learning_rate": 1.3463518679646313e-05, + "loss": 0.8839, + "step": 32950 + }, + { + "epoch": 2.43, + "learning_rate": 1.3463144493559256e-05, + "loss": 0.9543, + "step": 32951 + }, + { + "epoch": 2.43, + "learning_rate": 1.3462770301962384e-05, + "loss": 1.0527, + "step": 32952 + }, + { + "epoch": 2.43, + "learning_rate": 1.3462396104856295e-05, + "loss": 0.9393, + "step": 32953 + }, + { + "epoch": 2.43, + "learning_rate": 1.3462021902241582e-05, + "loss": 1.0767, + "step": 32954 + }, + { + "epoch": 2.44, + "learning_rate": 1.3461647694118845e-05, + "loss": 1.0431, + "step": 32955 + }, + { + "epoch": 2.44, + "learning_rate": 1.346127348048867e-05, + "loss": 1.0189, + "step": 32956 + }, + { + "epoch": 2.44, + "learning_rate": 1.3460899261351663e-05, + "loss": 0.9935, + "step": 32957 + }, + { + "epoch": 2.44, + "learning_rate": 1.3460525036708412e-05, + "loss": 0.8094, + "step": 32958 + }, + { + "epoch": 2.44, + "learning_rate": 1.3460150806559518e-05, + "loss": 0.9878, + "step": 32959 + }, + { + "epoch": 2.44, + "learning_rate": 1.3459776570905572e-05, + "loss": 0.9058, + "step": 32960 + }, + { + "epoch": 2.44, + "learning_rate": 1.3459402329747173e-05, + "loss": 0.9725, + "step": 32961 + }, + { + "epoch": 2.44, + "learning_rate": 1.345902808308491e-05, + "loss": 1.0475, + "step": 32962 + }, + { + "epoch": 2.44, + "learning_rate": 1.3458653830919387e-05, + "loss": 1.0412, + "step": 32963 + }, + { + "epoch": 2.44, + "learning_rate": 1.3458279573251194e-05, + "loss": 1.1567, + "step": 32964 + }, + { + "epoch": 2.44, + "learning_rate": 1.3457905310080928e-05, + "loss": 0.9623, + "step": 32965 + }, + { + "epoch": 2.44, + "learning_rate": 1.3457531041409182e-05, + "loss": 0.9872, + "step": 32966 + }, + { + "epoch": 2.44, + "learning_rate": 1.3457156767236557e-05, + "loss": 0.9087, + "step": 32967 + }, + { + "epoch": 2.44, + "learning_rate": 1.345678248756364e-05, + "loss": 0.9939, + "step": 32968 + }, + { + "epoch": 2.44, + "learning_rate": 1.3456408202391036e-05, + "loss": 0.9067, + "step": 32969 + }, + { + "epoch": 2.44, + "learning_rate": 1.3456033911719335e-05, + "loss": 1.0098, + "step": 32970 + }, + { + "epoch": 2.44, + "learning_rate": 1.3455659615549134e-05, + "loss": 0.9657, + "step": 32971 + }, + { + "epoch": 2.44, + "learning_rate": 1.3455285313881025e-05, + "loss": 0.8034, + "step": 32972 + }, + { + "epoch": 2.44, + "learning_rate": 1.3454911006715612e-05, + "loss": 0.9788, + "step": 32973 + }, + { + "epoch": 2.44, + "learning_rate": 1.3454536694053479e-05, + "loss": 1.063, + "step": 32974 + }, + { + "epoch": 2.44, + "learning_rate": 1.3454162375895233e-05, + "loss": 0.9691, + "step": 32975 + }, + { + "epoch": 2.44, + "learning_rate": 1.3453788052241459e-05, + "loss": 0.9653, + "step": 32976 + }, + { + "epoch": 2.44, + "learning_rate": 1.345341372309276e-05, + "loss": 0.9201, + "step": 32977 + }, + { + "epoch": 2.44, + "learning_rate": 1.345303938844973e-05, + "loss": 0.9867, + "step": 32978 + }, + { + "epoch": 2.44, + "learning_rate": 1.3452665048312964e-05, + "loss": 1.0879, + "step": 32979 + }, + { + "epoch": 2.44, + "learning_rate": 1.3452290702683055e-05, + "loss": 1.0294, + "step": 32980 + }, + { + "epoch": 2.44, + "learning_rate": 1.34519163515606e-05, + "loss": 1.0449, + "step": 32981 + }, + { + "epoch": 2.44, + "learning_rate": 1.34515419949462e-05, + "loss": 1.0133, + "step": 32982 + }, + { + "epoch": 2.44, + "learning_rate": 1.345116763284044e-05, + "loss": 0.9943, + "step": 32983 + }, + { + "epoch": 2.44, + "learning_rate": 1.3450793265243926e-05, + "loss": 1.0536, + "step": 32984 + }, + { + "epoch": 2.44, + "learning_rate": 1.345041889215725e-05, + "loss": 1.0875, + "step": 32985 + }, + { + "epoch": 2.44, + "learning_rate": 1.3450044513581006e-05, + "loss": 0.9857, + "step": 32986 + }, + { + "epoch": 2.44, + "learning_rate": 1.344967012951579e-05, + "loss": 0.9499, + "step": 32987 + }, + { + "epoch": 2.44, + "learning_rate": 1.3449295739962196e-05, + "loss": 1.0385, + "step": 32988 + }, + { + "epoch": 2.44, + "learning_rate": 1.3448921344920827e-05, + "loss": 1.022, + "step": 32989 + }, + { + "epoch": 2.44, + "learning_rate": 1.3448546944392269e-05, + "loss": 1.0195, + "step": 32990 + }, + { + "epoch": 2.44, + "learning_rate": 1.3448172538377122e-05, + "loss": 0.9894, + "step": 32991 + }, + { + "epoch": 2.44, + "learning_rate": 1.3447798126875986e-05, + "loss": 1.1371, + "step": 32992 + }, + { + "epoch": 2.44, + "learning_rate": 1.3447423709889449e-05, + "loss": 0.9679, + "step": 32993 + }, + { + "epoch": 2.44, + "learning_rate": 1.3447049287418112e-05, + "loss": 1.0162, + "step": 32994 + }, + { + "epoch": 2.44, + "learning_rate": 1.3446674859462565e-05, + "loss": 1.0211, + "step": 32995 + }, + { + "epoch": 2.44, + "learning_rate": 1.3446300426023414e-05, + "loss": 0.8985, + "step": 32996 + }, + { + "epoch": 2.44, + "learning_rate": 1.3445925987101242e-05, + "loss": 0.9564, + "step": 32997 + }, + { + "epoch": 2.44, + "learning_rate": 1.3445551542696657e-05, + "loss": 1.0086, + "step": 32998 + }, + { + "epoch": 2.44, + "learning_rate": 1.3445177092810244e-05, + "loss": 1.0157, + "step": 32999 + }, + { + "epoch": 2.44, + "learning_rate": 1.3444802637442606e-05, + "loss": 1.0238, + "step": 33000 + }, + { + "epoch": 2.44, + "learning_rate": 1.3444428176594336e-05, + "loss": 0.9074, + "step": 33001 + }, + { + "epoch": 2.44, + "learning_rate": 1.3444053710266029e-05, + "loss": 1.0544, + "step": 33002 + }, + { + "epoch": 2.44, + "learning_rate": 1.3443679238458283e-05, + "loss": 0.9818, + "step": 33003 + }, + { + "epoch": 2.44, + "learning_rate": 1.3443304761171694e-05, + "loss": 1.0713, + "step": 33004 + }, + { + "epoch": 2.44, + "learning_rate": 1.3442930278406853e-05, + "loss": 0.9958, + "step": 33005 + }, + { + "epoch": 2.44, + "learning_rate": 1.3442555790164362e-05, + "loss": 0.9757, + "step": 33006 + }, + { + "epoch": 2.44, + "learning_rate": 1.3442181296444812e-05, + "loss": 0.9941, + "step": 33007 + }, + { + "epoch": 2.44, + "learning_rate": 1.3441806797248801e-05, + "loss": 1.1511, + "step": 33008 + }, + { + "epoch": 2.44, + "learning_rate": 1.3441432292576926e-05, + "loss": 1.0129, + "step": 33009 + }, + { + "epoch": 2.44, + "learning_rate": 1.344105778242978e-05, + "loss": 1.0064, + "step": 33010 + }, + { + "epoch": 2.44, + "learning_rate": 1.3440683266807962e-05, + "loss": 0.9772, + "step": 33011 + }, + { + "epoch": 2.44, + "learning_rate": 1.3440308745712065e-05, + "loss": 1.0531, + "step": 33012 + }, + { + "epoch": 2.44, + "learning_rate": 1.3439934219142685e-05, + "loss": 0.9794, + "step": 33013 + }, + { + "epoch": 2.44, + "learning_rate": 1.343955968710042e-05, + "loss": 0.9935, + "step": 33014 + }, + { + "epoch": 2.44, + "learning_rate": 1.3439185149585865e-05, + "loss": 1.0617, + "step": 33015 + }, + { + "epoch": 2.44, + "learning_rate": 1.3438810606599614e-05, + "loss": 0.9625, + "step": 33016 + }, + { + "epoch": 2.44, + "learning_rate": 1.3438436058142268e-05, + "loss": 0.9964, + "step": 33017 + }, + { + "epoch": 2.44, + "learning_rate": 1.343806150421442e-05, + "loss": 1.056, + "step": 33018 + }, + { + "epoch": 2.44, + "learning_rate": 1.3437686944816659e-05, + "loss": 0.9715, + "step": 33019 + }, + { + "epoch": 2.44, + "learning_rate": 1.3437312379949594e-05, + "loss": 1.0488, + "step": 33020 + }, + { + "epoch": 2.44, + "learning_rate": 1.3436937809613808e-05, + "loss": 1.0205, + "step": 33021 + }, + { + "epoch": 2.44, + "learning_rate": 1.3436563233809908e-05, + "loss": 1.0698, + "step": 33022 + }, + { + "epoch": 2.44, + "learning_rate": 1.3436188652538483e-05, + "loss": 1.051, + "step": 33023 + }, + { + "epoch": 2.44, + "learning_rate": 1.3435814065800131e-05, + "loss": 0.9698, + "step": 33024 + }, + { + "epoch": 2.44, + "learning_rate": 1.3435439473595447e-05, + "loss": 0.9547, + "step": 33025 + }, + { + "epoch": 2.44, + "learning_rate": 1.343506487592503e-05, + "loss": 1.0217, + "step": 33026 + }, + { + "epoch": 2.44, + "learning_rate": 1.3434690272789474e-05, + "loss": 1.1075, + "step": 33027 + }, + { + "epoch": 2.44, + "learning_rate": 1.3434315664189373e-05, + "loss": 0.9857, + "step": 33028 + }, + { + "epoch": 2.44, + "learning_rate": 1.3433941050125325e-05, + "loss": 1.1142, + "step": 33029 + }, + { + "epoch": 2.44, + "learning_rate": 1.3433566430597932e-05, + "loss": 1.0846, + "step": 33030 + }, + { + "epoch": 2.44, + "learning_rate": 1.3433191805607775e-05, + "loss": 1.0442, + "step": 33031 + }, + { + "epoch": 2.44, + "learning_rate": 1.3432817175155464e-05, + "loss": 0.9499, + "step": 33032 + }, + { + "epoch": 2.44, + "learning_rate": 1.3432442539241588e-05, + "loss": 1.056, + "step": 33033 + }, + { + "epoch": 2.44, + "learning_rate": 1.3432067897866748e-05, + "loss": 0.8481, + "step": 33034 + }, + { + "epoch": 2.44, + "learning_rate": 1.3431693251031532e-05, + "loss": 1.1261, + "step": 33035 + }, + { + "epoch": 2.44, + "learning_rate": 1.3431318598736544e-05, + "loss": 0.9018, + "step": 33036 + }, + { + "epoch": 2.44, + "learning_rate": 1.3430943940982379e-05, + "loss": 1.0986, + "step": 33037 + }, + { + "epoch": 2.44, + "learning_rate": 1.3430569277769627e-05, + "loss": 1.034, + "step": 33038 + }, + { + "epoch": 2.44, + "learning_rate": 1.3430194609098893e-05, + "loss": 0.9493, + "step": 33039 + }, + { + "epoch": 2.44, + "learning_rate": 1.3429819934970764e-05, + "loss": 0.999, + "step": 33040 + }, + { + "epoch": 2.44, + "learning_rate": 1.3429445255385845e-05, + "loss": 0.9777, + "step": 33041 + }, + { + "epoch": 2.44, + "learning_rate": 1.3429070570344725e-05, + "loss": 0.9911, + "step": 33042 + }, + { + "epoch": 2.44, + "learning_rate": 1.3428695879848003e-05, + "loss": 1.0044, + "step": 33043 + }, + { + "epoch": 2.44, + "learning_rate": 1.3428321183896273e-05, + "loss": 1.055, + "step": 33044 + }, + { + "epoch": 2.44, + "learning_rate": 1.342794648249014e-05, + "loss": 1.0235, + "step": 33045 + }, + { + "epoch": 2.44, + "learning_rate": 1.3427571775630186e-05, + "loss": 0.9638, + "step": 33046 + }, + { + "epoch": 2.44, + "learning_rate": 1.3427197063317018e-05, + "loss": 0.9069, + "step": 33047 + }, + { + "epoch": 2.44, + "learning_rate": 1.3426822345551226e-05, + "loss": 0.9334, + "step": 33048 + }, + { + "epoch": 2.44, + "learning_rate": 1.3426447622333412e-05, + "loss": 1.0136, + "step": 33049 + }, + { + "epoch": 2.44, + "learning_rate": 1.3426072893664164e-05, + "loss": 0.9634, + "step": 33050 + }, + { + "epoch": 2.44, + "learning_rate": 1.3425698159544087e-05, + "loss": 0.9488, + "step": 33051 + }, + { + "epoch": 2.44, + "learning_rate": 1.342532341997377e-05, + "loss": 0.8363, + "step": 33052 + }, + { + "epoch": 2.44, + "learning_rate": 1.3424948674953817e-05, + "loss": 1.0491, + "step": 33053 + }, + { + "epoch": 2.44, + "learning_rate": 1.3424573924484818e-05, + "loss": 1.0124, + "step": 33054 + }, + { + "epoch": 2.44, + "learning_rate": 1.3424199168567371e-05, + "loss": 1.0024, + "step": 33055 + }, + { + "epoch": 2.44, + "learning_rate": 1.3423824407202071e-05, + "loss": 1.0536, + "step": 33056 + }, + { + "epoch": 2.44, + "learning_rate": 1.3423449640389517e-05, + "loss": 1.0647, + "step": 33057 + }, + { + "epoch": 2.44, + "learning_rate": 1.34230748681303e-05, + "loss": 0.9273, + "step": 33058 + }, + { + "epoch": 2.44, + "learning_rate": 1.3422700090425022e-05, + "loss": 1.0245, + "step": 33059 + }, + { + "epoch": 2.44, + "learning_rate": 1.3422325307274275e-05, + "loss": 0.9929, + "step": 33060 + }, + { + "epoch": 2.44, + "learning_rate": 1.3421950518678663e-05, + "loss": 1.1008, + "step": 33061 + }, + { + "epoch": 2.44, + "learning_rate": 1.342157572463877e-05, + "loss": 0.9651, + "step": 33062 + }, + { + "epoch": 2.44, + "learning_rate": 1.3421200925155204e-05, + "loss": 1.0045, + "step": 33063 + }, + { + "epoch": 2.44, + "learning_rate": 1.3420826120228556e-05, + "loss": 1.0751, + "step": 33064 + }, + { + "epoch": 2.44, + "learning_rate": 1.3420451309859422e-05, + "loss": 1.0029, + "step": 33065 + }, + { + "epoch": 2.44, + "learning_rate": 1.3420076494048397e-05, + "loss": 1.1135, + "step": 33066 + }, + { + "epoch": 2.44, + "learning_rate": 1.3419701672796082e-05, + "loss": 0.9811, + "step": 33067 + }, + { + "epoch": 2.44, + "learning_rate": 1.3419326846103068e-05, + "loss": 1.027, + "step": 33068 + }, + { + "epoch": 2.44, + "learning_rate": 1.3418952013969955e-05, + "loss": 0.9005, + "step": 33069 + }, + { + "epoch": 2.44, + "learning_rate": 1.3418577176397336e-05, + "loss": 0.9436, + "step": 33070 + }, + { + "epoch": 2.44, + "learning_rate": 1.3418202333385812e-05, + "loss": 1.0377, + "step": 33071 + }, + { + "epoch": 2.44, + "learning_rate": 1.3417827484935977e-05, + "loss": 0.9341, + "step": 33072 + }, + { + "epoch": 2.44, + "learning_rate": 1.3417452631048426e-05, + "loss": 0.9631, + "step": 33073 + }, + { + "epoch": 2.44, + "learning_rate": 1.3417077771723759e-05, + "loss": 0.9815, + "step": 33074 + }, + { + "epoch": 2.44, + "learning_rate": 1.3416702906962572e-05, + "loss": 1.0749, + "step": 33075 + }, + { + "epoch": 2.44, + "learning_rate": 1.3416328036765454e-05, + "loss": 1.0663, + "step": 33076 + }, + { + "epoch": 2.44, + "learning_rate": 1.3415953161133009e-05, + "loss": 1.1501, + "step": 33077 + }, + { + "epoch": 2.44, + "learning_rate": 1.341557828006583e-05, + "loss": 1.0178, + "step": 33078 + }, + { + "epoch": 2.44, + "learning_rate": 1.341520339356452e-05, + "loss": 0.9679, + "step": 33079 + }, + { + "epoch": 2.44, + "learning_rate": 1.3414828501629665e-05, + "loss": 1.0378, + "step": 33080 + }, + { + "epoch": 2.44, + "learning_rate": 1.341445360426187e-05, + "loss": 1.0369, + "step": 33081 + }, + { + "epoch": 2.44, + "learning_rate": 1.3414078701461728e-05, + "loss": 0.9403, + "step": 33082 + }, + { + "epoch": 2.44, + "learning_rate": 1.3413703793229833e-05, + "loss": 1.0875, + "step": 33083 + }, + { + "epoch": 2.44, + "learning_rate": 1.3413328879566786e-05, + "loss": 0.9919, + "step": 33084 + }, + { + "epoch": 2.44, + "learning_rate": 1.3412953960473178e-05, + "loss": 0.9775, + "step": 33085 + }, + { + "epoch": 2.44, + "learning_rate": 1.3412579035949613e-05, + "loss": 0.9511, + "step": 33086 + }, + { + "epoch": 2.44, + "learning_rate": 1.3412204105996683e-05, + "loss": 0.9612, + "step": 33087 + }, + { + "epoch": 2.44, + "learning_rate": 1.3411829170614987e-05, + "loss": 0.9183, + "step": 33088 + }, + { + "epoch": 2.44, + "learning_rate": 1.3411454229805117e-05, + "loss": 1.0567, + "step": 33089 + }, + { + "epoch": 2.44, + "learning_rate": 1.3411079283567671e-05, + "loss": 1.0178, + "step": 33090 + }, + { + "epoch": 2.45, + "learning_rate": 1.341070433190325e-05, + "loss": 1.0448, + "step": 33091 + }, + { + "epoch": 2.45, + "learning_rate": 1.3410329374812445e-05, + "loss": 1.0739, + "step": 33092 + }, + { + "epoch": 2.45, + "learning_rate": 1.3409954412295856e-05, + "loss": 0.885, + "step": 33093 + }, + { + "epoch": 2.45, + "learning_rate": 1.3409579444354077e-05, + "loss": 1.0403, + "step": 33094 + }, + { + "epoch": 2.45, + "learning_rate": 1.3409204470987706e-05, + "loss": 0.9895, + "step": 33095 + }, + { + "epoch": 2.45, + "learning_rate": 1.3408829492197341e-05, + "loss": 0.9548, + "step": 33096 + }, + { + "epoch": 2.45, + "learning_rate": 1.3408454507983575e-05, + "loss": 0.9633, + "step": 33097 + }, + { + "epoch": 2.45, + "learning_rate": 1.3408079518347008e-05, + "loss": 0.9519, + "step": 33098 + }, + { + "epoch": 2.45, + "learning_rate": 1.3407704523288233e-05, + "loss": 0.9741, + "step": 33099 + }, + { + "epoch": 2.45, + "learning_rate": 1.3407329522807853e-05, + "loss": 0.9407, + "step": 33100 + }, + { + "epoch": 2.45, + "learning_rate": 1.3406954516906459e-05, + "loss": 1.0919, + "step": 33101 + }, + { + "epoch": 2.45, + "learning_rate": 1.340657950558465e-05, + "loss": 0.9387, + "step": 33102 + }, + { + "epoch": 2.45, + "learning_rate": 1.3406204488843017e-05, + "loss": 0.9506, + "step": 33103 + }, + { + "epoch": 2.45, + "learning_rate": 1.3405829466682166e-05, + "loss": 1.0039, + "step": 33104 + }, + { + "epoch": 2.45, + "learning_rate": 1.3405454439102686e-05, + "loss": 0.9676, + "step": 33105 + }, + { + "epoch": 2.45, + "learning_rate": 1.3405079406105178e-05, + "loss": 1.1407, + "step": 33106 + }, + { + "epoch": 2.45, + "learning_rate": 1.3404704367690238e-05, + "loss": 0.9208, + "step": 33107 + }, + { + "epoch": 2.45, + "learning_rate": 1.3404329323858462e-05, + "loss": 1.0734, + "step": 33108 + }, + { + "epoch": 2.45, + "learning_rate": 1.3403954274610445e-05, + "loss": 0.9822, + "step": 33109 + }, + { + "epoch": 2.45, + "learning_rate": 1.340357921994679e-05, + "loss": 1.0125, + "step": 33110 + }, + { + "epoch": 2.45, + "learning_rate": 1.3403204159868084e-05, + "loss": 0.9034, + "step": 33111 + }, + { + "epoch": 2.45, + "learning_rate": 1.3402829094374931e-05, + "loss": 1.0109, + "step": 33112 + }, + { + "epoch": 2.45, + "learning_rate": 1.3402454023467925e-05, + "loss": 0.9815, + "step": 33113 + }, + { + "epoch": 2.45, + "learning_rate": 1.3402078947147666e-05, + "loss": 0.9634, + "step": 33114 + }, + { + "epoch": 2.45, + "learning_rate": 1.3401703865414744e-05, + "loss": 0.9995, + "step": 33115 + }, + { + "epoch": 2.45, + "learning_rate": 1.3401328778269762e-05, + "loss": 1.0424, + "step": 33116 + }, + { + "epoch": 2.45, + "learning_rate": 1.3400953685713314e-05, + "loss": 0.9891, + "step": 33117 + }, + { + "epoch": 2.45, + "learning_rate": 1.3400578587745998e-05, + "loss": 1.0392, + "step": 33118 + }, + { + "epoch": 2.45, + "learning_rate": 1.3400203484368409e-05, + "loss": 0.9564, + "step": 33119 + }, + { + "epoch": 2.45, + "learning_rate": 1.339982837558115e-05, + "loss": 0.9772, + "step": 33120 + }, + { + "epoch": 2.45, + "learning_rate": 1.3399453261384805e-05, + "loss": 0.9144, + "step": 33121 + }, + { + "epoch": 2.45, + "learning_rate": 1.3399078141779985e-05, + "loss": 0.9403, + "step": 33122 + }, + { + "epoch": 2.45, + "learning_rate": 1.3398703016767274e-05, + "loss": 1.0719, + "step": 33123 + }, + { + "epoch": 2.45, + "learning_rate": 1.3398327886347279e-05, + "loss": 1.0368, + "step": 33124 + }, + { + "epoch": 2.45, + "learning_rate": 1.339795275052059e-05, + "loss": 1.0558, + "step": 33125 + }, + { + "epoch": 2.45, + "learning_rate": 1.3397577609287811e-05, + "loss": 0.9009, + "step": 33126 + }, + { + "epoch": 2.45, + "learning_rate": 1.3397202462649533e-05, + "loss": 1.0617, + "step": 33127 + }, + { + "epoch": 2.45, + "learning_rate": 1.3396827310606356e-05, + "loss": 1.0157, + "step": 33128 + }, + { + "epoch": 2.45, + "learning_rate": 1.3396452153158873e-05, + "loss": 0.9856, + "step": 33129 + }, + { + "epoch": 2.45, + "learning_rate": 1.3396076990307682e-05, + "loss": 1.0318, + "step": 33130 + }, + { + "epoch": 2.45, + "learning_rate": 1.3395701822053384e-05, + "loss": 1.0435, + "step": 33131 + }, + { + "epoch": 2.45, + "learning_rate": 1.3395326648396569e-05, + "loss": 0.925, + "step": 33132 + }, + { + "epoch": 2.45, + "learning_rate": 1.3394951469337843e-05, + "loss": 0.9929, + "step": 33133 + }, + { + "epoch": 2.45, + "learning_rate": 1.3394576284877797e-05, + "loss": 0.9257, + "step": 33134 + }, + { + "epoch": 2.45, + "learning_rate": 1.3394201095017026e-05, + "loss": 0.9915, + "step": 33135 + }, + { + "epoch": 2.45, + "learning_rate": 1.339382589975613e-05, + "loss": 1.0526, + "step": 33136 + }, + { + "epoch": 2.45, + "learning_rate": 1.3393450699095707e-05, + "loss": 0.9989, + "step": 33137 + }, + { + "epoch": 2.45, + "learning_rate": 1.3393075493036351e-05, + "loss": 1.0039, + "step": 33138 + }, + { + "epoch": 2.45, + "learning_rate": 1.3392700281578663e-05, + "loss": 1.0324, + "step": 33139 + }, + { + "epoch": 2.45, + "learning_rate": 1.3392325064723235e-05, + "loss": 1.0135, + "step": 33140 + }, + { + "epoch": 2.45, + "learning_rate": 1.3391949842470669e-05, + "loss": 0.9362, + "step": 33141 + }, + { + "epoch": 2.45, + "learning_rate": 1.3391574614821554e-05, + "loss": 1.0216, + "step": 33142 + }, + { + "epoch": 2.45, + "learning_rate": 1.3391199381776496e-05, + "loss": 0.9547, + "step": 33143 + }, + { + "epoch": 2.45, + "learning_rate": 1.3390824143336087e-05, + "loss": 1.0176, + "step": 33144 + }, + { + "epoch": 2.45, + "learning_rate": 1.3390448899500927e-05, + "loss": 1.017, + "step": 33145 + }, + { + "epoch": 2.45, + "learning_rate": 1.339007365027161e-05, + "loss": 0.9712, + "step": 33146 + }, + { + "epoch": 2.45, + "learning_rate": 1.3389698395648735e-05, + "loss": 1.0925, + "step": 33147 + }, + { + "epoch": 2.45, + "learning_rate": 1.3389323135632901e-05, + "loss": 0.8771, + "step": 33148 + }, + { + "epoch": 2.45, + "learning_rate": 1.33889478702247e-05, + "loss": 0.9506, + "step": 33149 + }, + { + "epoch": 2.45, + "learning_rate": 1.338857259942473e-05, + "loss": 0.9496, + "step": 33150 + }, + { + "epoch": 2.45, + "learning_rate": 1.3388197323233592e-05, + "loss": 0.9573, + "step": 33151 + }, + { + "epoch": 2.45, + "learning_rate": 1.3387822041651877e-05, + "loss": 1.111, + "step": 33152 + }, + { + "epoch": 2.45, + "learning_rate": 1.3387446754680189e-05, + "loss": 0.976, + "step": 33153 + }, + { + "epoch": 2.45, + "learning_rate": 1.3387071462319119e-05, + "loss": 1.1035, + "step": 33154 + }, + { + "epoch": 2.45, + "learning_rate": 1.3386696164569268e-05, + "loss": 0.9734, + "step": 33155 + }, + { + "epoch": 2.45, + "learning_rate": 1.3386320861431232e-05, + "loss": 1.084, + "step": 33156 + }, + { + "epoch": 2.45, + "learning_rate": 1.338594555290561e-05, + "loss": 1.0769, + "step": 33157 + }, + { + "epoch": 2.45, + "learning_rate": 1.3385570238992995e-05, + "loss": 0.9841, + "step": 33158 + }, + { + "epoch": 2.45, + "learning_rate": 1.3385194919693987e-05, + "loss": 1.002, + "step": 33159 + }, + { + "epoch": 2.45, + "learning_rate": 1.338481959500918e-05, + "loss": 1.0205, + "step": 33160 + }, + { + "epoch": 2.45, + "learning_rate": 1.3384444264939175e-05, + "loss": 0.9354, + "step": 33161 + }, + { + "epoch": 2.45, + "learning_rate": 1.3384068929484568e-05, + "loss": 0.9865, + "step": 33162 + }, + { + "epoch": 2.45, + "learning_rate": 1.3383693588645956e-05, + "loss": 1.1491, + "step": 33163 + }, + { + "epoch": 2.45, + "learning_rate": 1.3383318242423933e-05, + "loss": 1.1196, + "step": 33164 + }, + { + "epoch": 2.45, + "learning_rate": 1.3382942890819104e-05, + "loss": 0.9915, + "step": 33165 + }, + { + "epoch": 2.45, + "learning_rate": 1.3382567533832059e-05, + "loss": 1.0445, + "step": 33166 + }, + { + "epoch": 2.45, + "learning_rate": 1.3382192171463397e-05, + "loss": 1.1246, + "step": 33167 + }, + { + "epoch": 2.45, + "learning_rate": 1.3381816803713714e-05, + "loss": 0.9309, + "step": 33168 + }, + { + "epoch": 2.45, + "learning_rate": 1.3381441430583613e-05, + "loss": 1.0009, + "step": 33169 + }, + { + "epoch": 2.45, + "learning_rate": 1.3381066052073683e-05, + "loss": 0.9818, + "step": 33170 + }, + { + "epoch": 2.45, + "learning_rate": 1.3380690668184529e-05, + "loss": 0.9518, + "step": 33171 + }, + { + "epoch": 2.45, + "learning_rate": 1.338031527891674e-05, + "loss": 0.9239, + "step": 33172 + }, + { + "epoch": 2.45, + "learning_rate": 1.337993988427092e-05, + "loss": 0.9741, + "step": 33173 + }, + { + "epoch": 2.45, + "learning_rate": 1.3379564484247665e-05, + "loss": 1.0161, + "step": 33174 + }, + { + "epoch": 2.45, + "learning_rate": 1.337918907884757e-05, + "loss": 0.9886, + "step": 33175 + }, + { + "epoch": 2.45, + "learning_rate": 1.3378813668071235e-05, + "loss": 1.0867, + "step": 33176 + }, + { + "epoch": 2.45, + "learning_rate": 1.3378438251919255e-05, + "loss": 0.9815, + "step": 33177 + }, + { + "epoch": 2.45, + "learning_rate": 1.3378062830392227e-05, + "loss": 0.9324, + "step": 33178 + }, + { + "epoch": 2.45, + "learning_rate": 1.3377687403490751e-05, + "loss": 1.0876, + "step": 33179 + }, + { + "epoch": 2.45, + "learning_rate": 1.3377311971215421e-05, + "loss": 0.8439, + "step": 33180 + }, + { + "epoch": 2.45, + "learning_rate": 1.3376936533566837e-05, + "loss": 1.0685, + "step": 33181 + }, + { + "epoch": 2.45, + "learning_rate": 1.3376561090545594e-05, + "loss": 0.9306, + "step": 33182 + }, + { + "epoch": 2.45, + "learning_rate": 1.3376185642152292e-05, + "loss": 0.9985, + "step": 33183 + }, + { + "epoch": 2.45, + "learning_rate": 1.3375810188387526e-05, + "loss": 0.9536, + "step": 33184 + }, + { + "epoch": 2.45, + "learning_rate": 1.3375434729251897e-05, + "loss": 1.0174, + "step": 33185 + }, + { + "epoch": 2.45, + "learning_rate": 1.3375059264745997e-05, + "loss": 0.9408, + "step": 33186 + }, + { + "epoch": 2.45, + "learning_rate": 1.3374683794870427e-05, + "loss": 1.0093, + "step": 33187 + }, + { + "epoch": 2.45, + "learning_rate": 1.3374308319625784e-05, + "loss": 1.0125, + "step": 33188 + }, + { + "epoch": 2.45, + "learning_rate": 1.3373932839012662e-05, + "loss": 1.096, + "step": 33189 + }, + { + "epoch": 2.45, + "learning_rate": 1.3373557353031663e-05, + "loss": 0.9888, + "step": 33190 + }, + { + "epoch": 2.45, + "learning_rate": 1.3373181861683382e-05, + "loss": 1.0293, + "step": 33191 + }, + { + "epoch": 2.45, + "learning_rate": 1.3372806364968418e-05, + "loss": 1.0875, + "step": 33192 + }, + { + "epoch": 2.45, + "learning_rate": 1.3372430862887367e-05, + "loss": 0.9901, + "step": 33193 + }, + { + "epoch": 2.45, + "learning_rate": 1.3372055355440827e-05, + "loss": 1.0434, + "step": 33194 + }, + { + "epoch": 2.45, + "learning_rate": 1.3371679842629394e-05, + "loss": 0.9943, + "step": 33195 + }, + { + "epoch": 2.45, + "learning_rate": 1.337130432445367e-05, + "loss": 0.9198, + "step": 33196 + }, + { + "epoch": 2.45, + "learning_rate": 1.3370928800914244e-05, + "loss": 1.0373, + "step": 33197 + }, + { + "epoch": 2.45, + "learning_rate": 1.337055327201172e-05, + "loss": 1.058, + "step": 33198 + }, + { + "epoch": 2.45, + "learning_rate": 1.3370177737746695e-05, + "loss": 1.0647, + "step": 33199 + }, + { + "epoch": 2.45, + "learning_rate": 1.3369802198119765e-05, + "loss": 1.0521, + "step": 33200 + }, + { + "epoch": 2.45, + "learning_rate": 1.3369426653131528e-05, + "loss": 1.1411, + "step": 33201 + }, + { + "epoch": 2.45, + "learning_rate": 1.3369051102782583e-05, + "loss": 0.9289, + "step": 33202 + }, + { + "epoch": 2.45, + "learning_rate": 1.3368675547073522e-05, + "loss": 1.0035, + "step": 33203 + }, + { + "epoch": 2.45, + "learning_rate": 1.3368299986004951e-05, + "loss": 0.9369, + "step": 33204 + }, + { + "epoch": 2.45, + "learning_rate": 1.336792441957746e-05, + "loss": 1.0589, + "step": 33205 + }, + { + "epoch": 2.45, + "learning_rate": 1.3367548847791652e-05, + "loss": 0.9467, + "step": 33206 + }, + { + "epoch": 2.45, + "learning_rate": 1.3367173270648117e-05, + "loss": 0.8362, + "step": 33207 + }, + { + "epoch": 2.45, + "learning_rate": 1.3366797688147463e-05, + "loss": 0.8956, + "step": 33208 + }, + { + "epoch": 2.45, + "learning_rate": 1.3366422100290278e-05, + "loss": 1.0361, + "step": 33209 + }, + { + "epoch": 2.45, + "learning_rate": 1.3366046507077166e-05, + "loss": 0.8635, + "step": 33210 + }, + { + "epoch": 2.45, + "learning_rate": 1.336567090850872e-05, + "loss": 0.9677, + "step": 33211 + }, + { + "epoch": 2.45, + "learning_rate": 1.3365295304585542e-05, + "loss": 0.9869, + "step": 33212 + }, + { + "epoch": 2.45, + "learning_rate": 1.3364919695308227e-05, + "loss": 1.0141, + "step": 33213 + }, + { + "epoch": 2.45, + "learning_rate": 1.3364544080677373e-05, + "loss": 0.9731, + "step": 33214 + }, + { + "epoch": 2.45, + "learning_rate": 1.3364168460693575e-05, + "loss": 0.9619, + "step": 33215 + }, + { + "epoch": 2.45, + "learning_rate": 1.3363792835357439e-05, + "loss": 1.0542, + "step": 33216 + }, + { + "epoch": 2.45, + "learning_rate": 1.3363417204669552e-05, + "loss": 0.9601, + "step": 33217 + }, + { + "epoch": 2.45, + "learning_rate": 1.3363041568630517e-05, + "loss": 1.0099, + "step": 33218 + }, + { + "epoch": 2.45, + "learning_rate": 1.3362665927240927e-05, + "loss": 0.964, + "step": 33219 + }, + { + "epoch": 2.45, + "learning_rate": 1.3362290280501391e-05, + "loss": 0.9178, + "step": 33220 + }, + { + "epoch": 2.45, + "learning_rate": 1.3361914628412493e-05, + "loss": 0.9362, + "step": 33221 + }, + { + "epoch": 2.45, + "learning_rate": 1.3361538970974841e-05, + "loss": 1.056, + "step": 33222 + }, + { + "epoch": 2.45, + "learning_rate": 1.3361163308189028e-05, + "loss": 1.0018, + "step": 33223 + }, + { + "epoch": 2.45, + "learning_rate": 1.3360787640055653e-05, + "loss": 1.0036, + "step": 33224 + }, + { + "epoch": 2.45, + "learning_rate": 1.336041196657531e-05, + "loss": 1.034, + "step": 33225 + }, + { + "epoch": 2.46, + "learning_rate": 1.3360036287748603e-05, + "loss": 0.9294, + "step": 33226 + }, + { + "epoch": 2.46, + "learning_rate": 1.3359660603576123e-05, + "loss": 1.0352, + "step": 33227 + }, + { + "epoch": 2.46, + "learning_rate": 1.3359284914058475e-05, + "loss": 1.0022, + "step": 33228 + }, + { + "epoch": 2.46, + "learning_rate": 1.3358909219196248e-05, + "loss": 0.9052, + "step": 33229 + }, + { + "epoch": 2.46, + "learning_rate": 1.335853351899005e-05, + "loss": 0.9062, + "step": 33230 + }, + { + "epoch": 2.46, + "learning_rate": 1.3358157813440467e-05, + "loss": 0.912, + "step": 33231 + }, + { + "epoch": 2.46, + "learning_rate": 1.3357782102548107e-05, + "loss": 0.9997, + "step": 33232 + }, + { + "epoch": 2.46, + "learning_rate": 1.3357406386313565e-05, + "loss": 0.9963, + "step": 33233 + }, + { + "epoch": 2.46, + "learning_rate": 1.3357030664737433e-05, + "loss": 0.9776, + "step": 33234 + }, + { + "epoch": 2.46, + "learning_rate": 1.3356654937820317e-05, + "loss": 1.0197, + "step": 33235 + }, + { + "epoch": 2.46, + "learning_rate": 1.3356279205562808e-05, + "loss": 1.0084, + "step": 33236 + }, + { + "epoch": 2.46, + "learning_rate": 1.335590346796551e-05, + "loss": 1.0768, + "step": 33237 + }, + { + "epoch": 2.46, + "learning_rate": 1.3355527725029013e-05, + "loss": 0.9565, + "step": 33238 + }, + { + "epoch": 2.46, + "learning_rate": 1.3355151976753924e-05, + "loss": 0.9516, + "step": 33239 + }, + { + "epoch": 2.46, + "learning_rate": 1.3354776223140836e-05, + "loss": 1.0026, + "step": 33240 + }, + { + "epoch": 2.46, + "learning_rate": 1.3354400464190347e-05, + "loss": 1.0359, + "step": 33241 + }, + { + "epoch": 2.46, + "learning_rate": 1.3354024699903052e-05, + "loss": 0.9315, + "step": 33242 + }, + { + "epoch": 2.46, + "learning_rate": 1.3353648930279555e-05, + "loss": 0.9632, + "step": 33243 + }, + { + "epoch": 2.46, + "learning_rate": 1.3353273155320446e-05, + "loss": 1.0916, + "step": 33244 + }, + { + "epoch": 2.46, + "learning_rate": 1.3352897375026332e-05, + "loss": 0.9786, + "step": 33245 + }, + { + "epoch": 2.46, + "learning_rate": 1.3352521589397802e-05, + "loss": 1.0153, + "step": 33246 + }, + { + "epoch": 2.46, + "learning_rate": 1.3352145798435464e-05, + "loss": 1.0316, + "step": 33247 + }, + { + "epoch": 2.46, + "learning_rate": 1.3351770002139903e-05, + "loss": 0.9962, + "step": 33248 + }, + { + "epoch": 2.46, + "learning_rate": 1.3351394200511729e-05, + "loss": 1.0686, + "step": 33249 + }, + { + "epoch": 2.46, + "learning_rate": 1.3351018393551532e-05, + "loss": 0.9469, + "step": 33250 + }, + { + "epoch": 2.46, + "learning_rate": 1.3350642581259916e-05, + "loss": 1.037, + "step": 33251 + }, + { + "epoch": 2.46, + "learning_rate": 1.3350266763637473e-05, + "loss": 1.0398, + "step": 33252 + }, + { + "epoch": 2.46, + "learning_rate": 1.3349890940684805e-05, + "loss": 0.8909, + "step": 33253 + }, + { + "epoch": 2.46, + "learning_rate": 1.3349515112402505e-05, + "loss": 0.9921, + "step": 33254 + }, + { + "epoch": 2.46, + "learning_rate": 1.3349139278791178e-05, + "loss": 1.0243, + "step": 33255 + }, + { + "epoch": 2.46, + "learning_rate": 1.3348763439851415e-05, + "loss": 0.9734, + "step": 33256 + }, + { + "epoch": 2.46, + "learning_rate": 1.334838759558382e-05, + "loss": 1.0147, + "step": 33257 + }, + { + "epoch": 2.46, + "learning_rate": 1.3348011745988988e-05, + "loss": 1.0366, + "step": 33258 + }, + { + "epoch": 2.46, + "learning_rate": 1.3347635891067517e-05, + "loss": 0.9803, + "step": 33259 + }, + { + "epoch": 2.46, + "learning_rate": 1.3347260030820006e-05, + "loss": 0.9066, + "step": 33260 + }, + { + "epoch": 2.46, + "learning_rate": 1.3346884165247052e-05, + "loss": 1.0575, + "step": 33261 + }, + { + "epoch": 2.46, + "learning_rate": 1.3346508294349252e-05, + "loss": 1.0328, + "step": 33262 + }, + { + "epoch": 2.46, + "learning_rate": 1.3346132418127206e-05, + "loss": 1.0542, + "step": 33263 + }, + { + "epoch": 2.46, + "learning_rate": 1.334575653658151e-05, + "loss": 1.0193, + "step": 33264 + }, + { + "epoch": 2.46, + "learning_rate": 1.3345380649712764e-05, + "loss": 1.0831, + "step": 33265 + }, + { + "epoch": 2.46, + "learning_rate": 1.3345004757521564e-05, + "loss": 1.0007, + "step": 33266 + }, + { + "epoch": 2.46, + "learning_rate": 1.3344628860008512e-05, + "loss": 0.9297, + "step": 33267 + }, + { + "epoch": 2.46, + "learning_rate": 1.3344252957174199e-05, + "loss": 1.0527, + "step": 33268 + }, + { + "epoch": 2.46, + "learning_rate": 1.3343877049019233e-05, + "loss": 0.9752, + "step": 33269 + }, + { + "epoch": 2.46, + "learning_rate": 1.3343501135544204e-05, + "loss": 0.9639, + "step": 33270 + }, + { + "epoch": 2.46, + "learning_rate": 1.3343125216749713e-05, + "loss": 0.9859, + "step": 33271 + }, + { + "epoch": 2.46, + "learning_rate": 1.3342749292636357e-05, + "loss": 0.9391, + "step": 33272 + }, + { + "epoch": 2.46, + "learning_rate": 1.3342373363204735e-05, + "loss": 0.9626, + "step": 33273 + }, + { + "epoch": 2.46, + "learning_rate": 1.3341997428455442e-05, + "loss": 1.0164, + "step": 33274 + }, + { + "epoch": 2.46, + "learning_rate": 1.3341621488389082e-05, + "loss": 1.1481, + "step": 33275 + }, + { + "epoch": 2.46, + "learning_rate": 1.334124554300625e-05, + "loss": 1.0222, + "step": 33276 + }, + { + "epoch": 2.46, + "learning_rate": 1.3340869592307541e-05, + "loss": 0.9777, + "step": 33277 + }, + { + "epoch": 2.46, + "learning_rate": 1.3340493636293559e-05, + "loss": 1.1121, + "step": 33278 + }, + { + "epoch": 2.46, + "learning_rate": 1.33401176749649e-05, + "loss": 0.9889, + "step": 33279 + }, + { + "epoch": 2.46, + "learning_rate": 1.3339741708322162e-05, + "loss": 1.0487, + "step": 33280 + }, + { + "epoch": 2.46, + "learning_rate": 1.3339365736365938e-05, + "loss": 1.0211, + "step": 33281 + }, + { + "epoch": 2.46, + "learning_rate": 1.3338989759096837e-05, + "loss": 0.9601, + "step": 33282 + }, + { + "epoch": 2.46, + "learning_rate": 1.3338613776515447e-05, + "loss": 1.0441, + "step": 33283 + }, + { + "epoch": 2.46, + "learning_rate": 1.3338237788622372e-05, + "loss": 1.083, + "step": 33284 + }, + { + "epoch": 2.46, + "learning_rate": 1.333786179541821e-05, + "loss": 0.9784, + "step": 33285 + }, + { + "epoch": 2.46, + "learning_rate": 1.3337485796903553e-05, + "loss": 1.0667, + "step": 33286 + }, + { + "epoch": 2.46, + "learning_rate": 1.3337109793079008e-05, + "loss": 0.9148, + "step": 33287 + }, + { + "epoch": 2.46, + "learning_rate": 1.3336733783945168e-05, + "loss": 0.9953, + "step": 33288 + }, + { + "epoch": 2.46, + "learning_rate": 1.3336357769502633e-05, + "loss": 0.9067, + "step": 33289 + }, + { + "epoch": 2.46, + "learning_rate": 1.3335981749752e-05, + "loss": 1.02, + "step": 33290 + }, + { + "epoch": 2.46, + "learning_rate": 1.3335605724693865e-05, + "loss": 1.0672, + "step": 33291 + }, + { + "epoch": 2.46, + "learning_rate": 1.3335229694328833e-05, + "loss": 0.993, + "step": 33292 + }, + { + "epoch": 2.46, + "learning_rate": 1.3334853658657493e-05, + "loss": 1.0134, + "step": 33293 + }, + { + "epoch": 2.46, + "learning_rate": 1.3334477617680455e-05, + "loss": 0.974, + "step": 33294 + }, + { + "epoch": 2.46, + "learning_rate": 1.3334101571398305e-05, + "loss": 0.9739, + "step": 33295 + }, + { + "epoch": 2.46, + "learning_rate": 1.333372551981165e-05, + "loss": 1.0303, + "step": 33296 + }, + { + "epoch": 2.46, + "learning_rate": 1.3333349462921087e-05, + "loss": 1.0379, + "step": 33297 + }, + { + "epoch": 2.46, + "learning_rate": 1.3332973400727214e-05, + "loss": 0.9759, + "step": 33298 + }, + { + "epoch": 2.46, + "learning_rate": 1.3332597333230624e-05, + "loss": 0.9244, + "step": 33299 + }, + { + "epoch": 2.46, + "learning_rate": 1.3332221260431921e-05, + "loss": 1.0527, + "step": 33300 + }, + { + "epoch": 2.46, + "learning_rate": 1.33318451823317e-05, + "loss": 0.9181, + "step": 33301 + }, + { + "epoch": 2.46, + "learning_rate": 1.3331469098930563e-05, + "loss": 1.0611, + "step": 33302 + }, + { + "epoch": 2.46, + "learning_rate": 1.3331093010229103e-05, + "loss": 0.9891, + "step": 33303 + }, + { + "epoch": 2.46, + "learning_rate": 1.3330716916227924e-05, + "loss": 1.0182, + "step": 33304 + }, + { + "epoch": 2.46, + "learning_rate": 1.3330340816927622e-05, + "loss": 0.9796, + "step": 33305 + }, + { + "epoch": 2.46, + "learning_rate": 1.3329964712328795e-05, + "loss": 0.9193, + "step": 33306 + }, + { + "epoch": 2.46, + "learning_rate": 1.3329588602432044e-05, + "loss": 0.9184, + "step": 33307 + }, + { + "epoch": 2.46, + "learning_rate": 1.3329212487237967e-05, + "loss": 0.8999, + "step": 33308 + }, + { + "epoch": 2.46, + "learning_rate": 1.3328836366747154e-05, + "loss": 0.9795, + "step": 33309 + }, + { + "epoch": 2.46, + "learning_rate": 1.3328460240960213e-05, + "loss": 0.9727, + "step": 33310 + }, + { + "epoch": 2.46, + "learning_rate": 1.3328084109877739e-05, + "loss": 1.0791, + "step": 33311 + }, + { + "epoch": 2.46, + "learning_rate": 1.3327707973500332e-05, + "loss": 1.0787, + "step": 33312 + }, + { + "epoch": 2.46, + "learning_rate": 1.3327331831828587e-05, + "loss": 1.0229, + "step": 33313 + }, + { + "epoch": 2.46, + "learning_rate": 1.3326955684863107e-05, + "loss": 1.0126, + "step": 33314 + }, + { + "epoch": 2.46, + "learning_rate": 1.3326579532604486e-05, + "loss": 0.9477, + "step": 33315 + }, + { + "epoch": 2.46, + "learning_rate": 1.3326203375053329e-05, + "loss": 1.0446, + "step": 33316 + }, + { + "epoch": 2.46, + "learning_rate": 1.3325827212210225e-05, + "loss": 1.008, + "step": 33317 + }, + { + "epoch": 2.46, + "learning_rate": 1.3325451044075779e-05, + "loss": 1.0671, + "step": 33318 + }, + { + "epoch": 2.46, + "learning_rate": 1.3325074870650587e-05, + "loss": 0.9385, + "step": 33319 + }, + { + "epoch": 2.46, + "learning_rate": 1.3324698691935252e-05, + "loss": 1.049, + "step": 33320 + }, + { + "epoch": 2.46, + "learning_rate": 1.3324322507930365e-05, + "loss": 0.975, + "step": 33321 + }, + { + "epoch": 2.46, + "learning_rate": 1.3323946318636531e-05, + "loss": 0.9974, + "step": 33322 + }, + { + "epoch": 2.46, + "learning_rate": 1.3323570124054343e-05, + "loss": 1.0013, + "step": 33323 + }, + { + "epoch": 2.46, + "learning_rate": 1.3323193924184405e-05, + "loss": 1.0359, + "step": 33324 + }, + { + "epoch": 2.46, + "learning_rate": 1.3322817719027314e-05, + "loss": 0.996, + "step": 33325 + }, + { + "epoch": 2.46, + "learning_rate": 1.3322441508583666e-05, + "loss": 0.9331, + "step": 33326 + }, + { + "epoch": 2.46, + "learning_rate": 1.332206529285406e-05, + "loss": 1.0479, + "step": 33327 + }, + { + "epoch": 2.46, + "learning_rate": 1.3321689071839099e-05, + "loss": 1.0701, + "step": 33328 + }, + { + "epoch": 2.46, + "learning_rate": 1.3321312845539374e-05, + "loss": 1.0636, + "step": 33329 + }, + { + "epoch": 2.46, + "learning_rate": 1.332093661395549e-05, + "loss": 0.9305, + "step": 33330 + }, + { + "epoch": 2.46, + "learning_rate": 1.3320560377088043e-05, + "loss": 1.0181, + "step": 33331 + }, + { + "epoch": 2.46, + "learning_rate": 1.3320184134937632e-05, + "loss": 0.945, + "step": 33332 + }, + { + "epoch": 2.46, + "learning_rate": 1.3319807887504856e-05, + "loss": 1.035, + "step": 33333 + }, + { + "epoch": 2.46, + "learning_rate": 1.3319431634790314e-05, + "loss": 0.9733, + "step": 33334 + }, + { + "epoch": 2.46, + "learning_rate": 1.3319055376794603e-05, + "loss": 1.0374, + "step": 33335 + }, + { + "epoch": 2.46, + "learning_rate": 1.331867911351832e-05, + "loss": 1.0123, + "step": 33336 + }, + { + "epoch": 2.46, + "learning_rate": 1.331830284496207e-05, + "loss": 1.032, + "step": 33337 + }, + { + "epoch": 2.46, + "learning_rate": 1.3317926571126445e-05, + "loss": 0.9951, + "step": 33338 + }, + { + "epoch": 2.46, + "learning_rate": 1.3317550292012046e-05, + "loss": 1.0153, + "step": 33339 + }, + { + "epoch": 2.46, + "learning_rate": 1.331717400761947e-05, + "loss": 1.0795, + "step": 33340 + }, + { + "epoch": 2.46, + "learning_rate": 1.3316797717949323e-05, + "loss": 1.0089, + "step": 33341 + }, + { + "epoch": 2.46, + "learning_rate": 1.3316421423002195e-05, + "loss": 1.0635, + "step": 33342 + }, + { + "epoch": 2.46, + "learning_rate": 1.3316045122778688e-05, + "loss": 1.1972, + "step": 33343 + }, + { + "epoch": 2.46, + "learning_rate": 1.3315668817279403e-05, + "loss": 1.0179, + "step": 33344 + }, + { + "epoch": 2.46, + "learning_rate": 1.3315292506504936e-05, + "loss": 1.005, + "step": 33345 + }, + { + "epoch": 2.46, + "learning_rate": 1.3314916190455884e-05, + "loss": 0.899, + "step": 33346 + }, + { + "epoch": 2.46, + "learning_rate": 1.331453986913285e-05, + "loss": 0.9953, + "step": 33347 + }, + { + "epoch": 2.46, + "learning_rate": 1.3314163542536426e-05, + "loss": 1.0248, + "step": 33348 + }, + { + "epoch": 2.46, + "learning_rate": 1.331378721066722e-05, + "loss": 1.0671, + "step": 33349 + }, + { + "epoch": 2.46, + "learning_rate": 1.3313410873525823e-05, + "loss": 0.9976, + "step": 33350 + }, + { + "epoch": 2.46, + "learning_rate": 1.3313034531112839e-05, + "loss": 0.9419, + "step": 33351 + }, + { + "epoch": 2.46, + "learning_rate": 1.3312658183428861e-05, + "loss": 1.066, + "step": 33352 + }, + { + "epoch": 2.46, + "learning_rate": 1.3312281830474498e-05, + "loss": 0.9148, + "step": 33353 + }, + { + "epoch": 2.46, + "learning_rate": 1.3311905472250337e-05, + "loss": 1.0886, + "step": 33354 + }, + { + "epoch": 2.46, + "learning_rate": 1.3311529108756985e-05, + "loss": 1.0211, + "step": 33355 + }, + { + "epoch": 2.46, + "learning_rate": 1.3311152739995033e-05, + "loss": 1.0006, + "step": 33356 + }, + { + "epoch": 2.46, + "learning_rate": 1.3310776365965085e-05, + "loss": 0.9257, + "step": 33357 + }, + { + "epoch": 2.46, + "learning_rate": 1.331039998666774e-05, + "loss": 0.9935, + "step": 33358 + }, + { + "epoch": 2.46, + "learning_rate": 1.3310023602103597e-05, + "loss": 0.9954, + "step": 33359 + }, + { + "epoch": 2.46, + "learning_rate": 1.3309647212273254e-05, + "loss": 0.9869, + "step": 33360 + }, + { + "epoch": 2.47, + "learning_rate": 1.3309270817177309e-05, + "loss": 0.9523, + "step": 33361 + }, + { + "epoch": 2.47, + "learning_rate": 1.3308894416816361e-05, + "loss": 0.9566, + "step": 33362 + }, + { + "epoch": 2.47, + "learning_rate": 1.3308518011191012e-05, + "loss": 1.1251, + "step": 33363 + }, + { + "epoch": 2.47, + "learning_rate": 1.3308141600301854e-05, + "loss": 0.9923, + "step": 33364 + }, + { + "epoch": 2.47, + "learning_rate": 1.3307765184149492e-05, + "loss": 1.0397, + "step": 33365 + }, + { + "epoch": 2.47, + "learning_rate": 1.3307388762734524e-05, + "loss": 0.9887, + "step": 33366 + }, + { + "epoch": 2.47, + "learning_rate": 1.3307012336057545e-05, + "loss": 1.0581, + "step": 33367 + }, + { + "epoch": 2.47, + "learning_rate": 1.3306635904119158e-05, + "loss": 0.8681, + "step": 33368 + }, + { + "epoch": 2.47, + "learning_rate": 1.3306259466919963e-05, + "loss": 0.9702, + "step": 33369 + }, + { + "epoch": 2.47, + "learning_rate": 1.3305883024460552e-05, + "loss": 1.0708, + "step": 33370 + }, + { + "epoch": 2.47, + "learning_rate": 1.3305506576741532e-05, + "loss": 0.9592, + "step": 33371 + }, + { + "epoch": 2.47, + "learning_rate": 1.3305130123763499e-05, + "loss": 1.0604, + "step": 33372 + }, + { + "epoch": 2.47, + "learning_rate": 1.3304753665527051e-05, + "loss": 0.9486, + "step": 33373 + }, + { + "epoch": 2.47, + "learning_rate": 1.3304377202032784e-05, + "loss": 0.997, + "step": 33374 + }, + { + "epoch": 2.47, + "learning_rate": 1.3304000733281301e-05, + "loss": 0.925, + "step": 33375 + }, + { + "epoch": 2.47, + "learning_rate": 1.33036242592732e-05, + "loss": 1.0684, + "step": 33376 + }, + { + "epoch": 2.47, + "learning_rate": 1.3303247780009084e-05, + "loss": 0.9718, + "step": 33377 + }, + { + "epoch": 2.47, + "learning_rate": 1.3302871295489541e-05, + "loss": 1.0185, + "step": 33378 + }, + { + "epoch": 2.47, + "learning_rate": 1.3302494805715182e-05, + "loss": 0.9233, + "step": 33379 + }, + { + "epoch": 2.47, + "learning_rate": 1.33021183106866e-05, + "loss": 1.1235, + "step": 33380 + }, + { + "epoch": 2.47, + "learning_rate": 1.3301741810404395e-05, + "loss": 0.9785, + "step": 33381 + }, + { + "epoch": 2.47, + "learning_rate": 1.3301365304869167e-05, + "loss": 0.9519, + "step": 33382 + }, + { + "epoch": 2.47, + "learning_rate": 1.3300988794081513e-05, + "loss": 0.9913, + "step": 33383 + }, + { + "epoch": 2.47, + "learning_rate": 1.3300612278042033e-05, + "loss": 1.084, + "step": 33384 + }, + { + "epoch": 2.47, + "learning_rate": 1.3300235756751325e-05, + "loss": 0.9887, + "step": 33385 + }, + { + "epoch": 2.47, + "learning_rate": 1.3299859230209993e-05, + "loss": 0.88, + "step": 33386 + }, + { + "epoch": 2.47, + "learning_rate": 1.3299482698418626e-05, + "loss": 1.0034, + "step": 33387 + }, + { + "epoch": 2.47, + "learning_rate": 1.3299106161377834e-05, + "loss": 0.9634, + "step": 33388 + }, + { + "epoch": 2.47, + "learning_rate": 1.3298729619088214e-05, + "loss": 0.9925, + "step": 33389 + }, + { + "epoch": 2.47, + "learning_rate": 1.3298353071550356e-05, + "loss": 0.9703, + "step": 33390 + }, + { + "epoch": 2.47, + "learning_rate": 1.3297976518764868e-05, + "loss": 1.0774, + "step": 33391 + }, + { + "epoch": 2.47, + "learning_rate": 1.3297599960732348e-05, + "loss": 1.0545, + "step": 33392 + }, + { + "epoch": 2.47, + "learning_rate": 1.3297223397453392e-05, + "loss": 0.9543, + "step": 33393 + }, + { + "epoch": 2.47, + "learning_rate": 1.3296846828928601e-05, + "loss": 0.9934, + "step": 33394 + }, + { + "epoch": 2.47, + "learning_rate": 1.329647025515857e-05, + "loss": 0.9182, + "step": 33395 + }, + { + "epoch": 2.47, + "learning_rate": 1.3296093676143908e-05, + "loss": 0.9346, + "step": 33396 + }, + { + "epoch": 2.47, + "learning_rate": 1.3295717091885204e-05, + "loss": 1.0435, + "step": 33397 + }, + { + "epoch": 2.47, + "learning_rate": 1.3295340502383065e-05, + "loss": 1.0124, + "step": 33398 + }, + { + "epoch": 2.47, + "learning_rate": 1.3294963907638083e-05, + "loss": 1.0347, + "step": 33399 + }, + { + "epoch": 2.47, + "learning_rate": 1.3294587307650866e-05, + "loss": 0.9838, + "step": 33400 + }, + { + "epoch": 2.47, + "learning_rate": 1.3294210702422002e-05, + "loss": 0.9685, + "step": 33401 + }, + { + "epoch": 2.47, + "learning_rate": 1.3293834091952099e-05, + "loss": 1.0401, + "step": 33402 + }, + { + "epoch": 2.47, + "learning_rate": 1.3293457476241752e-05, + "loss": 0.8935, + "step": 33403 + }, + { + "epoch": 2.47, + "learning_rate": 1.3293080855291561e-05, + "loss": 1.0965, + "step": 33404 + }, + { + "epoch": 2.47, + "learning_rate": 1.3292704229102124e-05, + "loss": 0.9746, + "step": 33405 + }, + { + "epoch": 2.47, + "learning_rate": 1.3292327597674044e-05, + "loss": 1.0485, + "step": 33406 + }, + { + "epoch": 2.47, + "learning_rate": 1.3291950961007916e-05, + "loss": 1.0115, + "step": 33407 + }, + { + "epoch": 2.47, + "learning_rate": 1.3291574319104344e-05, + "loss": 1.017, + "step": 33408 + }, + { + "epoch": 2.47, + "learning_rate": 1.3291197671963923e-05, + "loss": 0.9335, + "step": 33409 + }, + { + "epoch": 2.47, + "learning_rate": 1.3290821019587255e-05, + "loss": 0.9572, + "step": 33410 + }, + { + "epoch": 2.47, + "learning_rate": 1.3290444361974935e-05, + "loss": 1.0016, + "step": 33411 + }, + { + "epoch": 2.47, + "learning_rate": 1.3290067699127564e-05, + "loss": 1.0667, + "step": 33412 + }, + { + "epoch": 2.47, + "learning_rate": 1.3289691031045746e-05, + "loss": 1.0815, + "step": 33413 + }, + { + "epoch": 2.47, + "learning_rate": 1.3289314357730074e-05, + "loss": 1.0086, + "step": 33414 + }, + { + "epoch": 2.47, + "learning_rate": 1.3288937679181151e-05, + "loss": 0.9836, + "step": 33415 + }, + { + "epoch": 2.47, + "learning_rate": 1.3288560995399578e-05, + "loss": 1.0536, + "step": 33416 + }, + { + "epoch": 2.47, + "learning_rate": 1.3288184306385944e-05, + "loss": 0.9768, + "step": 33417 + }, + { + "epoch": 2.47, + "learning_rate": 1.3287807612140864e-05, + "loss": 1.0218, + "step": 33418 + }, + { + "epoch": 2.47, + "learning_rate": 1.3287430912664925e-05, + "loss": 0.9992, + "step": 33419 + }, + { + "epoch": 2.47, + "learning_rate": 1.3287054207958731e-05, + "loss": 0.9871, + "step": 33420 + }, + { + "epoch": 2.47, + "learning_rate": 1.3286677498022882e-05, + "loss": 0.9084, + "step": 33421 + }, + { + "epoch": 2.47, + "learning_rate": 1.3286300782857976e-05, + "loss": 0.9685, + "step": 33422 + }, + { + "epoch": 2.47, + "learning_rate": 1.328592406246461e-05, + "loss": 1.0165, + "step": 33423 + }, + { + "epoch": 2.47, + "learning_rate": 1.3285547336843387e-05, + "loss": 1.0063, + "step": 33424 + }, + { + "epoch": 2.47, + "learning_rate": 1.3285170605994907e-05, + "loss": 1.037, + "step": 33425 + }, + { + "epoch": 2.47, + "learning_rate": 1.3284793869919765e-05, + "loss": 1.0419, + "step": 33426 + }, + { + "epoch": 2.47, + "learning_rate": 1.3284417128618562e-05, + "loss": 1.0525, + "step": 33427 + }, + { + "epoch": 2.47, + "learning_rate": 1.3284040382091903e-05, + "loss": 0.947, + "step": 33428 + }, + { + "epoch": 2.47, + "learning_rate": 1.3283663630340383e-05, + "loss": 0.9278, + "step": 33429 + }, + { + "epoch": 2.47, + "learning_rate": 1.3283286873364597e-05, + "loss": 1.0869, + "step": 33430 + }, + { + "epoch": 2.47, + "learning_rate": 1.328291011116515e-05, + "loss": 1.0571, + "step": 33431 + }, + { + "epoch": 2.47, + "learning_rate": 1.3282533343742638e-05, + "loss": 0.9794, + "step": 33432 + }, + { + "epoch": 2.47, + "learning_rate": 1.3282156571097664e-05, + "loss": 1.0631, + "step": 33433 + }, + { + "epoch": 2.47, + "learning_rate": 1.3281779793230829e-05, + "loss": 1.0168, + "step": 33434 + }, + { + "epoch": 2.47, + "learning_rate": 1.3281403010142724e-05, + "loss": 0.9861, + "step": 33435 + }, + { + "epoch": 2.47, + "learning_rate": 1.3281026221833956e-05, + "loss": 0.9407, + "step": 33436 + }, + { + "epoch": 2.47, + "learning_rate": 1.3280649428305126e-05, + "loss": 1.0231, + "step": 33437 + }, + { + "epoch": 2.47, + "learning_rate": 1.3280272629556824e-05, + "loss": 1.0287, + "step": 33438 + }, + { + "epoch": 2.47, + "learning_rate": 1.3279895825589657e-05, + "loss": 0.9365, + "step": 33439 + }, + { + "epoch": 2.47, + "learning_rate": 1.3279519016404224e-05, + "loss": 1.066, + "step": 33440 + }, + { + "epoch": 2.47, + "learning_rate": 1.3279142202001123e-05, + "loss": 1.004, + "step": 33441 + }, + { + "epoch": 2.47, + "learning_rate": 1.3278765382380949e-05, + "loss": 1.0068, + "step": 33442 + }, + { + "epoch": 2.47, + "learning_rate": 1.3278388557544313e-05, + "loss": 1.1243, + "step": 33443 + }, + { + "epoch": 2.47, + "learning_rate": 1.3278011727491805e-05, + "loss": 1.0496, + "step": 33444 + }, + { + "epoch": 2.47, + "learning_rate": 1.3277634892224028e-05, + "loss": 0.9097, + "step": 33445 + }, + { + "epoch": 2.47, + "learning_rate": 1.3277258051741578e-05, + "loss": 1.0227, + "step": 33446 + }, + { + "epoch": 2.47, + "learning_rate": 1.327688120604506e-05, + "loss": 1.0259, + "step": 33447 + }, + { + "epoch": 2.47, + "learning_rate": 1.3276504355135068e-05, + "loss": 1.0141, + "step": 33448 + }, + { + "epoch": 2.47, + "learning_rate": 1.3276127499012207e-05, + "loss": 0.9909, + "step": 33449 + }, + { + "epoch": 2.47, + "learning_rate": 1.3275750637677075e-05, + "loss": 1.0026, + "step": 33450 + }, + { + "epoch": 2.47, + "learning_rate": 1.3275373771130268e-05, + "loss": 1.0503, + "step": 33451 + }, + { + "epoch": 2.47, + "learning_rate": 1.327499689937239e-05, + "loss": 1.0047, + "step": 33452 + }, + { + "epoch": 2.47, + "learning_rate": 1.3274620022404036e-05, + "loss": 1.0209, + "step": 33453 + }, + { + "epoch": 2.47, + "learning_rate": 1.3274243140225811e-05, + "loss": 0.9685, + "step": 33454 + }, + { + "epoch": 2.47, + "learning_rate": 1.3273866252838312e-05, + "loss": 0.9252, + "step": 33455 + }, + { + "epoch": 2.47, + "learning_rate": 1.327348936024214e-05, + "loss": 1.0136, + "step": 33456 + }, + { + "epoch": 2.47, + "learning_rate": 1.3273112462437891e-05, + "loss": 0.8984, + "step": 33457 + }, + { + "epoch": 2.47, + "learning_rate": 1.3272735559426166e-05, + "loss": 1.0789, + "step": 33458 + }, + { + "epoch": 2.47, + "learning_rate": 1.3272358651207568e-05, + "loss": 1.0361, + "step": 33459 + }, + { + "epoch": 2.47, + "learning_rate": 1.3271981737782693e-05, + "loss": 1.1452, + "step": 33460 + }, + { + "epoch": 2.47, + "learning_rate": 1.3271604819152142e-05, + "loss": 0.9408, + "step": 33461 + }, + { + "epoch": 2.47, + "learning_rate": 1.3271227895316514e-05, + "loss": 1.0232, + "step": 33462 + }, + { + "epoch": 2.47, + "learning_rate": 1.327085096627641e-05, + "loss": 1.0292, + "step": 33463 + }, + { + "epoch": 2.47, + "learning_rate": 1.3270474032032428e-05, + "loss": 0.98, + "step": 33464 + }, + { + "epoch": 2.47, + "learning_rate": 1.3270097092585169e-05, + "loss": 1.0442, + "step": 33465 + }, + { + "epoch": 2.47, + "learning_rate": 1.3269720147935235e-05, + "loss": 1.0536, + "step": 33466 + }, + { + "epoch": 2.47, + "learning_rate": 1.326934319808322e-05, + "loss": 1.1378, + "step": 33467 + }, + { + "epoch": 2.47, + "learning_rate": 1.3268966243029726e-05, + "loss": 1.0895, + "step": 33468 + }, + { + "epoch": 2.47, + "learning_rate": 1.3268589282775355e-05, + "loss": 1.0847, + "step": 33469 + }, + { + "epoch": 2.47, + "learning_rate": 1.3268212317320703e-05, + "loss": 0.9702, + "step": 33470 + }, + { + "epoch": 2.47, + "learning_rate": 1.3267835346666376e-05, + "loss": 0.9884, + "step": 33471 + }, + { + "epoch": 2.47, + "learning_rate": 1.3267458370812967e-05, + "loss": 1.0068, + "step": 33472 + }, + { + "epoch": 2.47, + "learning_rate": 1.326708138976108e-05, + "loss": 0.9909, + "step": 33473 + }, + { + "epoch": 2.47, + "learning_rate": 1.3266704403511311e-05, + "loss": 1.0789, + "step": 33474 + }, + { + "epoch": 2.47, + "learning_rate": 1.3266327412064264e-05, + "loss": 1.0367, + "step": 33475 + }, + { + "epoch": 2.47, + "learning_rate": 1.3265950415420536e-05, + "loss": 1.0821, + "step": 33476 + }, + { + "epoch": 2.47, + "learning_rate": 1.3265573413580728e-05, + "loss": 1.0455, + "step": 33477 + }, + { + "epoch": 2.47, + "learning_rate": 1.326519640654544e-05, + "loss": 1.0572, + "step": 33478 + }, + { + "epoch": 2.47, + "learning_rate": 1.3264819394315271e-05, + "loss": 1.0991, + "step": 33479 + }, + { + "epoch": 2.47, + "learning_rate": 1.3264442376890818e-05, + "loss": 1.0306, + "step": 33480 + }, + { + "epoch": 2.47, + "learning_rate": 1.3264065354272689e-05, + "loss": 0.9793, + "step": 33481 + }, + { + "epoch": 2.47, + "learning_rate": 1.3263688326461475e-05, + "loss": 1.0001, + "step": 33482 + }, + { + "epoch": 2.47, + "learning_rate": 1.326331129345778e-05, + "loss": 0.9988, + "step": 33483 + }, + { + "epoch": 2.47, + "learning_rate": 1.3262934255262206e-05, + "loss": 0.9913, + "step": 33484 + }, + { + "epoch": 2.47, + "learning_rate": 1.3262557211875347e-05, + "loss": 1.0953, + "step": 33485 + }, + { + "epoch": 2.47, + "learning_rate": 1.3262180163297808e-05, + "loss": 0.9218, + "step": 33486 + }, + { + "epoch": 2.47, + "learning_rate": 1.3261803109530187e-05, + "loss": 1.0803, + "step": 33487 + }, + { + "epoch": 2.47, + "learning_rate": 1.3261426050573082e-05, + "loss": 0.9492, + "step": 33488 + }, + { + "epoch": 2.47, + "learning_rate": 1.3261048986427096e-05, + "loss": 0.9897, + "step": 33489 + }, + { + "epoch": 2.47, + "learning_rate": 1.3260671917092828e-05, + "loss": 0.9896, + "step": 33490 + }, + { + "epoch": 2.47, + "learning_rate": 1.3260294842570877e-05, + "loss": 0.8692, + "step": 33491 + }, + { + "epoch": 2.47, + "learning_rate": 1.3259917762861846e-05, + "loss": 0.9974, + "step": 33492 + }, + { + "epoch": 2.47, + "learning_rate": 1.325954067796633e-05, + "loss": 0.9781, + "step": 33493 + }, + { + "epoch": 2.47, + "learning_rate": 1.3259163587884935e-05, + "loss": 1.1433, + "step": 33494 + }, + { + "epoch": 2.47, + "learning_rate": 1.3258786492618254e-05, + "loss": 0.9913, + "step": 33495 + }, + { + "epoch": 2.47, + "learning_rate": 1.3258409392166891e-05, + "loss": 1.0545, + "step": 33496 + }, + { + "epoch": 2.48, + "learning_rate": 1.3258032286531444e-05, + "loss": 0.8955, + "step": 33497 + }, + { + "epoch": 2.48, + "learning_rate": 1.3257655175712516e-05, + "loss": 1.0496, + "step": 33498 + }, + { + "epoch": 2.48, + "learning_rate": 1.3257278059710702e-05, + "loss": 1.0994, + "step": 33499 + }, + { + "epoch": 2.48, + "learning_rate": 1.325690093852661e-05, + "loss": 0.9301, + "step": 33500 + }, + { + "epoch": 2.48, + "learning_rate": 1.3256523812160832e-05, + "loss": 0.9936, + "step": 33501 + }, + { + "epoch": 2.48, + "learning_rate": 1.3256146680613973e-05, + "loss": 0.9567, + "step": 33502 + }, + { + "epoch": 2.48, + "learning_rate": 1.3255769543886631e-05, + "loss": 0.9594, + "step": 33503 + }, + { + "epoch": 2.48, + "learning_rate": 1.3255392401979407e-05, + "loss": 0.9521, + "step": 33504 + }, + { + "epoch": 2.48, + "learning_rate": 1.3255015254892897e-05, + "loss": 0.9804, + "step": 33505 + }, + { + "epoch": 2.48, + "learning_rate": 1.3254638102627709e-05, + "loss": 1.0223, + "step": 33506 + }, + { + "epoch": 2.48, + "learning_rate": 1.3254260945184433e-05, + "loss": 0.9462, + "step": 33507 + }, + { + "epoch": 2.48, + "learning_rate": 1.325388378256368e-05, + "loss": 0.9406, + "step": 33508 + }, + { + "epoch": 2.48, + "learning_rate": 1.3253506614766038e-05, + "loss": 0.9697, + "step": 33509 + }, + { + "epoch": 2.48, + "learning_rate": 1.3253129441792119e-05, + "loss": 1.0034, + "step": 33510 + }, + { + "epoch": 2.48, + "learning_rate": 1.3252752263642515e-05, + "loss": 0.9785, + "step": 33511 + }, + { + "epoch": 2.48, + "learning_rate": 1.325237508031783e-05, + "loss": 1.1174, + "step": 33512 + }, + { + "epoch": 2.48, + "learning_rate": 1.3251997891818662e-05, + "loss": 0.9663, + "step": 33513 + }, + { + "epoch": 2.48, + "learning_rate": 1.3251620698145612e-05, + "loss": 1.036, + "step": 33514 + }, + { + "epoch": 2.48, + "learning_rate": 1.325124349929928e-05, + "loss": 0.93, + "step": 33515 + }, + { + "epoch": 2.48, + "learning_rate": 1.3250866295280265e-05, + "loss": 1.0124, + "step": 33516 + }, + { + "epoch": 2.48, + "learning_rate": 1.3250489086089169e-05, + "loss": 1.0438, + "step": 33517 + }, + { + "epoch": 2.48, + "learning_rate": 1.3250111871726592e-05, + "loss": 1.0582, + "step": 33518 + }, + { + "epoch": 2.48, + "learning_rate": 1.3249734652193134e-05, + "loss": 0.9968, + "step": 33519 + }, + { + "epoch": 2.48, + "learning_rate": 1.3249357427489393e-05, + "loss": 1.0958, + "step": 33520 + }, + { + "epoch": 2.48, + "learning_rate": 1.324898019761597e-05, + "loss": 1.0679, + "step": 33521 + }, + { + "epoch": 2.48, + "learning_rate": 1.3248602962573472e-05, + "loss": 1.0617, + "step": 33522 + }, + { + "epoch": 2.48, + "learning_rate": 1.3248225722362487e-05, + "loss": 1.09, + "step": 33523 + }, + { + "epoch": 2.48, + "learning_rate": 1.3247848476983624e-05, + "loss": 1.1035, + "step": 33524 + }, + { + "epoch": 2.48, + "learning_rate": 1.3247471226437478e-05, + "loss": 1.0023, + "step": 33525 + }, + { + "epoch": 2.48, + "learning_rate": 1.3247093970724653e-05, + "loss": 1.0737, + "step": 33526 + }, + { + "epoch": 2.48, + "learning_rate": 1.3246716709845748e-05, + "loss": 1.1133, + "step": 33527 + }, + { + "epoch": 2.48, + "learning_rate": 1.3246339443801363e-05, + "loss": 0.9937, + "step": 33528 + }, + { + "epoch": 2.48, + "learning_rate": 1.32459621725921e-05, + "loss": 1.0189, + "step": 33529 + }, + { + "epoch": 2.48, + "learning_rate": 1.3245584896218557e-05, + "loss": 1.0006, + "step": 33530 + }, + { + "epoch": 2.48, + "learning_rate": 1.3245207614681334e-05, + "loss": 0.959, + "step": 33531 + }, + { + "epoch": 2.48, + "learning_rate": 1.3244830327981033e-05, + "loss": 0.9721, + "step": 33532 + }, + { + "epoch": 2.48, + "learning_rate": 1.3244453036118254e-05, + "loss": 0.9375, + "step": 33533 + }, + { + "epoch": 2.48, + "learning_rate": 1.3244075739093593e-05, + "loss": 0.9637, + "step": 33534 + }, + { + "epoch": 2.48, + "learning_rate": 1.3243698436907657e-05, + "loss": 1.0624, + "step": 33535 + }, + { + "epoch": 2.48, + "learning_rate": 1.3243321129561041e-05, + "loss": 1.0021, + "step": 33536 + }, + { + "epoch": 2.48, + "learning_rate": 1.3242943817054351e-05, + "loss": 1.0247, + "step": 33537 + }, + { + "epoch": 2.48, + "learning_rate": 1.3242566499388182e-05, + "loss": 0.9841, + "step": 33538 + }, + { + "epoch": 2.48, + "learning_rate": 1.3242189176563137e-05, + "loss": 1.0035, + "step": 33539 + }, + { + "epoch": 2.48, + "learning_rate": 1.3241811848579815e-05, + "loss": 1.0536, + "step": 33540 + }, + { + "epoch": 2.48, + "learning_rate": 1.3241434515438818e-05, + "loss": 1.0194, + "step": 33541 + }, + { + "epoch": 2.48, + "learning_rate": 1.3241057177140742e-05, + "loss": 1.0623, + "step": 33542 + }, + { + "epoch": 2.48, + "learning_rate": 1.3240679833686193e-05, + "loss": 0.9077, + "step": 33543 + }, + { + "epoch": 2.48, + "learning_rate": 1.3240302485075765e-05, + "loss": 0.9722, + "step": 33544 + }, + { + "epoch": 2.48, + "learning_rate": 1.3239925131310065e-05, + "loss": 0.9296, + "step": 33545 + }, + { + "epoch": 2.48, + "learning_rate": 1.323954777238969e-05, + "loss": 1.1597, + "step": 33546 + }, + { + "epoch": 2.48, + "learning_rate": 1.323917040831524e-05, + "loss": 0.9967, + "step": 33547 + }, + { + "epoch": 2.48, + "learning_rate": 1.3238793039087316e-05, + "loss": 1.0227, + "step": 33548 + }, + { + "epoch": 2.48, + "learning_rate": 1.3238415664706521e-05, + "loss": 1.0166, + "step": 33549 + }, + { + "epoch": 2.48, + "learning_rate": 1.3238038285173454e-05, + "loss": 1.0216, + "step": 33550 + }, + { + "epoch": 2.48, + "learning_rate": 1.3237660900488711e-05, + "loss": 0.9687, + "step": 33551 + }, + { + "epoch": 2.48, + "learning_rate": 1.3237283510652896e-05, + "loss": 0.9034, + "step": 33552 + }, + { + "epoch": 2.48, + "learning_rate": 1.323690611566661e-05, + "loss": 1.029, + "step": 33553 + }, + { + "epoch": 2.48, + "learning_rate": 1.3236528715530453e-05, + "loss": 0.9512, + "step": 33554 + }, + { + "epoch": 2.48, + "learning_rate": 1.3236151310245024e-05, + "loss": 1.0396, + "step": 33555 + }, + { + "epoch": 2.48, + "learning_rate": 1.3235773899810927e-05, + "loss": 1.1074, + "step": 33556 + }, + { + "epoch": 2.48, + "learning_rate": 1.3235396484228757e-05, + "loss": 1.0304, + "step": 33557 + }, + { + "epoch": 2.48, + "learning_rate": 1.3235019063499119e-05, + "loss": 0.9443, + "step": 33558 + }, + { + "epoch": 2.48, + "learning_rate": 1.3234641637622615e-05, + "loss": 1.007, + "step": 33559 + }, + { + "epoch": 2.48, + "learning_rate": 1.3234264206599838e-05, + "loss": 1.0668, + "step": 33560 + }, + { + "epoch": 2.48, + "learning_rate": 1.3233886770431394e-05, + "loss": 1.0377, + "step": 33561 + }, + { + "epoch": 2.48, + "learning_rate": 1.3233509329117885e-05, + "loss": 0.9577, + "step": 33562 + }, + { + "epoch": 2.48, + "learning_rate": 1.3233131882659907e-05, + "loss": 1.0182, + "step": 33563 + }, + { + "epoch": 2.48, + "learning_rate": 1.323275443105806e-05, + "loss": 1.1422, + "step": 33564 + }, + { + "epoch": 2.48, + "learning_rate": 1.3232376974312948e-05, + "loss": 0.9679, + "step": 33565 + }, + { + "epoch": 2.48, + "learning_rate": 1.323199951242517e-05, + "loss": 1.0402, + "step": 33566 + }, + { + "epoch": 2.48, + "learning_rate": 1.3231622045395332e-05, + "loss": 0.9081, + "step": 33567 + }, + { + "epoch": 2.48, + "learning_rate": 1.3231244573224026e-05, + "loss": 0.9938, + "step": 33568 + }, + { + "epoch": 2.48, + "learning_rate": 1.3230867095911857e-05, + "loss": 1.0549, + "step": 33569 + }, + { + "epoch": 2.48, + "learning_rate": 1.3230489613459422e-05, + "loss": 0.9311, + "step": 33570 + }, + { + "epoch": 2.48, + "learning_rate": 1.3230112125867329e-05, + "loss": 1.0338, + "step": 33571 + }, + { + "epoch": 2.48, + "learning_rate": 1.3229734633136168e-05, + "loss": 0.9876, + "step": 33572 + }, + { + "epoch": 2.48, + "learning_rate": 1.322935713526655e-05, + "loss": 0.9782, + "step": 33573 + }, + { + "epoch": 2.48, + "learning_rate": 1.3228979632259069e-05, + "loss": 1.1199, + "step": 33574 + }, + { + "epoch": 2.48, + "learning_rate": 1.322860212411433e-05, + "loss": 1.016, + "step": 33575 + }, + { + "epoch": 2.48, + "learning_rate": 1.3228224610832925e-05, + "loss": 0.9293, + "step": 33576 + }, + { + "epoch": 2.48, + "learning_rate": 1.3227847092415466e-05, + "loss": 1.1754, + "step": 33577 + }, + { + "epoch": 2.48, + "learning_rate": 1.322746956886255e-05, + "loss": 1.0688, + "step": 33578 + }, + { + "epoch": 2.48, + "learning_rate": 1.3227092040174772e-05, + "loss": 0.9224, + "step": 33579 + }, + { + "epoch": 2.48, + "learning_rate": 1.3226714506352738e-05, + "loss": 1.1123, + "step": 33580 + }, + { + "epoch": 2.48, + "learning_rate": 1.3226336967397048e-05, + "loss": 1.0107, + "step": 33581 + }, + { + "epoch": 2.48, + "learning_rate": 1.3225959423308301e-05, + "loss": 1.0492, + "step": 33582 + }, + { + "epoch": 2.48, + "learning_rate": 1.32255818740871e-05, + "loss": 0.9656, + "step": 33583 + }, + { + "epoch": 2.48, + "learning_rate": 1.3225204319734044e-05, + "loss": 1.0378, + "step": 33584 + }, + { + "epoch": 2.48, + "learning_rate": 1.3224826760249736e-05, + "loss": 0.9505, + "step": 33585 + }, + { + "epoch": 2.48, + "learning_rate": 1.322444919563477e-05, + "loss": 0.9982, + "step": 33586 + }, + { + "epoch": 2.48, + "learning_rate": 1.3224071625889756e-05, + "loss": 0.9289, + "step": 33587 + }, + { + "epoch": 2.48, + "learning_rate": 1.322369405101529e-05, + "loss": 1.0045, + "step": 33588 + }, + { + "epoch": 2.48, + "learning_rate": 1.3223316471011969e-05, + "loss": 1.0402, + "step": 33589 + }, + { + "epoch": 2.48, + "learning_rate": 1.3222938885880401e-05, + "loss": 0.9969, + "step": 33590 + }, + { + "epoch": 2.48, + "learning_rate": 1.3222561295621181e-05, + "loss": 1.0714, + "step": 33591 + }, + { + "epoch": 2.48, + "learning_rate": 1.3222183700234914e-05, + "loss": 0.8344, + "step": 33592 + }, + { + "epoch": 2.48, + "learning_rate": 1.3221806099722196e-05, + "loss": 0.999, + "step": 33593 + }, + { + "epoch": 2.48, + "learning_rate": 1.3221428494083634e-05, + "loss": 1.1031, + "step": 33594 + }, + { + "epoch": 2.48, + "learning_rate": 1.3221050883319823e-05, + "loss": 1.0813, + "step": 33595 + }, + { + "epoch": 2.48, + "learning_rate": 1.322067326743137e-05, + "loss": 1.0626, + "step": 33596 + }, + { + "epoch": 2.48, + "learning_rate": 1.3220295646418868e-05, + "loss": 1.1162, + "step": 33597 + }, + { + "epoch": 2.48, + "learning_rate": 1.3219918020282925e-05, + "loss": 0.9702, + "step": 33598 + }, + { + "epoch": 2.48, + "learning_rate": 1.3219540389024135e-05, + "loss": 1.0576, + "step": 33599 + }, + { + "epoch": 2.48, + "learning_rate": 1.3219162752643104e-05, + "loss": 1.0384, + "step": 33600 + }, + { + "epoch": 2.48, + "learning_rate": 1.3218785111140428e-05, + "loss": 1.0981, + "step": 33601 + }, + { + "epoch": 2.48, + "learning_rate": 1.3218407464516717e-05, + "loss": 1.0746, + "step": 33602 + }, + { + "epoch": 2.48, + "learning_rate": 1.3218029812772559e-05, + "loss": 0.9852, + "step": 33603 + }, + { + "epoch": 2.48, + "learning_rate": 1.3217652155908566e-05, + "loss": 0.9984, + "step": 33604 + }, + { + "epoch": 2.48, + "learning_rate": 1.3217274493925334e-05, + "loss": 1.0059, + "step": 33605 + }, + { + "epoch": 2.48, + "learning_rate": 1.3216896826823466e-05, + "loss": 1.1218, + "step": 33606 + }, + { + "epoch": 2.48, + "learning_rate": 1.3216519154603557e-05, + "loss": 0.9907, + "step": 33607 + }, + { + "epoch": 2.48, + "learning_rate": 1.3216141477266216e-05, + "loss": 1.0565, + "step": 33608 + }, + { + "epoch": 2.48, + "learning_rate": 1.3215763794812038e-05, + "loss": 0.9738, + "step": 33609 + }, + { + "epoch": 2.48, + "learning_rate": 1.3215386107241626e-05, + "loss": 1.095, + "step": 33610 + }, + { + "epoch": 2.48, + "learning_rate": 1.321500841455558e-05, + "loss": 0.9081, + "step": 33611 + }, + { + "epoch": 2.48, + "learning_rate": 1.3214630716754502e-05, + "loss": 1.0739, + "step": 33612 + }, + { + "epoch": 2.48, + "learning_rate": 1.3214253013838992e-05, + "loss": 0.9482, + "step": 33613 + }, + { + "epoch": 2.48, + "learning_rate": 1.3213875305809652e-05, + "loss": 1.0784, + "step": 33614 + }, + { + "epoch": 2.48, + "learning_rate": 1.3213497592667085e-05, + "loss": 0.9951, + "step": 33615 + }, + { + "epoch": 2.48, + "learning_rate": 1.3213119874411885e-05, + "loss": 1.0218, + "step": 33616 + }, + { + "epoch": 2.48, + "learning_rate": 1.3212742151044658e-05, + "loss": 1.073, + "step": 33617 + }, + { + "epoch": 2.48, + "learning_rate": 1.3212364422566007e-05, + "loss": 1.0674, + "step": 33618 + }, + { + "epoch": 2.48, + "learning_rate": 1.3211986688976528e-05, + "loss": 1.0685, + "step": 33619 + }, + { + "epoch": 2.48, + "learning_rate": 1.3211608950276823e-05, + "loss": 1.0276, + "step": 33620 + }, + { + "epoch": 2.48, + "learning_rate": 1.3211231206467496e-05, + "loss": 1.0035, + "step": 33621 + }, + { + "epoch": 2.48, + "learning_rate": 1.3210853457549147e-05, + "loss": 0.958, + "step": 33622 + }, + { + "epoch": 2.48, + "learning_rate": 1.3210475703522372e-05, + "loss": 1.054, + "step": 33623 + }, + { + "epoch": 2.48, + "learning_rate": 1.3210097944387781e-05, + "loss": 1.0019, + "step": 33624 + }, + { + "epoch": 2.48, + "learning_rate": 1.3209720180145969e-05, + "loss": 0.9557, + "step": 33625 + }, + { + "epoch": 2.48, + "learning_rate": 1.3209342410797534e-05, + "loss": 0.9252, + "step": 33626 + }, + { + "epoch": 2.48, + "learning_rate": 1.3208964636343087e-05, + "loss": 1.1057, + "step": 33627 + }, + { + "epoch": 2.48, + "learning_rate": 1.320858685678322e-05, + "loss": 0.939, + "step": 33628 + }, + { + "epoch": 2.48, + "learning_rate": 1.3208209072118536e-05, + "loss": 1.0439, + "step": 33629 + }, + { + "epoch": 2.48, + "learning_rate": 1.3207831282349641e-05, + "loss": 0.9589, + "step": 33630 + }, + { + "epoch": 2.48, + "learning_rate": 1.3207453487477129e-05, + "loss": 0.9968, + "step": 33631 + }, + { + "epoch": 2.49, + "learning_rate": 1.3207075687501608e-05, + "loss": 0.9918, + "step": 33632 + }, + { + "epoch": 2.49, + "learning_rate": 1.3206697882423671e-05, + "loss": 1.0512, + "step": 33633 + }, + { + "epoch": 2.49, + "learning_rate": 1.3206320072243925e-05, + "loss": 1.0292, + "step": 33634 + }, + { + "epoch": 2.49, + "learning_rate": 1.320594225696297e-05, + "loss": 0.8177, + "step": 33635 + }, + { + "epoch": 2.49, + "learning_rate": 1.3205564436581409e-05, + "loss": 1.0544, + "step": 33636 + }, + { + "epoch": 2.49, + "learning_rate": 1.3205186611099838e-05, + "loss": 1.0846, + "step": 33637 + }, + { + "epoch": 2.49, + "learning_rate": 1.3204808780518861e-05, + "loss": 0.9675, + "step": 33638 + }, + { + "epoch": 2.49, + "learning_rate": 1.3204430944839082e-05, + "loss": 1.0666, + "step": 33639 + }, + { + "epoch": 2.49, + "learning_rate": 1.3204053104061094e-05, + "loss": 0.9897, + "step": 33640 + }, + { + "epoch": 2.49, + "learning_rate": 1.3203675258185509e-05, + "loss": 0.9795, + "step": 33641 + }, + { + "epoch": 2.49, + "learning_rate": 1.3203297407212919e-05, + "loss": 0.9977, + "step": 33642 + }, + { + "epoch": 2.49, + "learning_rate": 1.3202919551143931e-05, + "loss": 1.1543, + "step": 33643 + }, + { + "epoch": 2.49, + "learning_rate": 1.3202541689979142e-05, + "loss": 0.8913, + "step": 33644 + }, + { + "epoch": 2.49, + "learning_rate": 1.3202163823719156e-05, + "loss": 1.0208, + "step": 33645 + }, + { + "epoch": 2.49, + "learning_rate": 1.3201785952364573e-05, + "loss": 0.9992, + "step": 33646 + }, + { + "epoch": 2.49, + "learning_rate": 1.3201408075915994e-05, + "loss": 0.9231, + "step": 33647 + }, + { + "epoch": 2.49, + "learning_rate": 1.320103019437402e-05, + "loss": 1.0466, + "step": 33648 + }, + { + "epoch": 2.49, + "learning_rate": 1.3200652307739254e-05, + "loss": 0.9976, + "step": 33649 + }, + { + "epoch": 2.49, + "learning_rate": 1.3200274416012294e-05, + "loss": 0.9296, + "step": 33650 + }, + { + "epoch": 2.49, + "learning_rate": 1.3199896519193744e-05, + "loss": 1.1133, + "step": 33651 + }, + { + "epoch": 2.49, + "learning_rate": 1.3199518617284205e-05, + "loss": 0.9649, + "step": 33652 + }, + { + "epoch": 2.49, + "learning_rate": 1.3199140710284279e-05, + "loss": 1.075, + "step": 33653 + }, + { + "epoch": 2.49, + "learning_rate": 1.3198762798194565e-05, + "loss": 0.9913, + "step": 33654 + }, + { + "epoch": 2.49, + "learning_rate": 1.3198384881015665e-05, + "loss": 1.0462, + "step": 33655 + }, + { + "epoch": 2.49, + "learning_rate": 1.3198006958748178e-05, + "loss": 1.0946, + "step": 33656 + }, + { + "epoch": 2.49, + "learning_rate": 1.3197629031392711e-05, + "loss": 1.0, + "step": 33657 + }, + { + "epoch": 2.49, + "learning_rate": 1.319725109894986e-05, + "loss": 1.0148, + "step": 33658 + }, + { + "epoch": 2.49, + "learning_rate": 1.319687316142023e-05, + "loss": 0.9959, + "step": 33659 + }, + { + "epoch": 2.49, + "learning_rate": 1.3196495218804417e-05, + "loss": 0.9939, + "step": 33660 + }, + { + "epoch": 2.49, + "learning_rate": 1.319611727110303e-05, + "loss": 1.0055, + "step": 33661 + }, + { + "epoch": 2.49, + "learning_rate": 1.3195739318316665e-05, + "loss": 0.8958, + "step": 33662 + }, + { + "epoch": 2.49, + "learning_rate": 1.3195361360445923e-05, + "loss": 0.9437, + "step": 33663 + }, + { + "epoch": 2.49, + "learning_rate": 1.3194983397491408e-05, + "loss": 0.9995, + "step": 33664 + }, + { + "epoch": 2.49, + "learning_rate": 1.3194605429453722e-05, + "loss": 0.9426, + "step": 33665 + }, + { + "epoch": 2.49, + "learning_rate": 1.3194227456333461e-05, + "loss": 1.006, + "step": 33666 + }, + { + "epoch": 2.49, + "learning_rate": 1.3193849478131232e-05, + "loss": 0.9531, + "step": 33667 + }, + { + "epoch": 2.49, + "learning_rate": 1.3193471494847631e-05, + "loss": 1.0127, + "step": 33668 + }, + { + "epoch": 2.49, + "learning_rate": 1.3193093506483266e-05, + "loss": 0.9649, + "step": 33669 + }, + { + "epoch": 2.49, + "learning_rate": 1.3192715513038731e-05, + "loss": 1.0579, + "step": 33670 + }, + { + "epoch": 2.49, + "learning_rate": 1.319233751451464e-05, + "loss": 0.9191, + "step": 33671 + }, + { + "epoch": 2.49, + "learning_rate": 1.3191959510911576e-05, + "loss": 1.0048, + "step": 33672 + }, + { + "epoch": 2.49, + "learning_rate": 1.3191581502230155e-05, + "loss": 1.0432, + "step": 33673 + }, + { + "epoch": 2.49, + "learning_rate": 1.3191203488470971e-05, + "loss": 0.9322, + "step": 33674 + }, + { + "epoch": 2.49, + "learning_rate": 1.3190825469634632e-05, + "loss": 0.9814, + "step": 33675 + }, + { + "epoch": 2.49, + "learning_rate": 1.319044744572173e-05, + "loss": 0.9888, + "step": 33676 + }, + { + "epoch": 2.49, + "learning_rate": 1.3190069416732876e-05, + "loss": 1.083, + "step": 33677 + }, + { + "epoch": 2.49, + "learning_rate": 1.3189691382668663e-05, + "loss": 1.0014, + "step": 33678 + }, + { + "epoch": 2.49, + "learning_rate": 1.3189313343529702e-05, + "loss": 0.9925, + "step": 33679 + }, + { + "epoch": 2.49, + "learning_rate": 1.3188935299316584e-05, + "loss": 1.0119, + "step": 33680 + }, + { + "epoch": 2.49, + "learning_rate": 1.3188557250029918e-05, + "loss": 1.0682, + "step": 33681 + }, + { + "epoch": 2.49, + "learning_rate": 1.3188179195670306e-05, + "loss": 1.1173, + "step": 33682 + }, + { + "epoch": 2.49, + "learning_rate": 1.318780113623834e-05, + "loss": 0.9612, + "step": 33683 + }, + { + "epoch": 2.49, + "learning_rate": 1.3187423071734632e-05, + "loss": 1.0343, + "step": 33684 + }, + { + "epoch": 2.49, + "learning_rate": 1.3187045002159779e-05, + "loss": 1.0484, + "step": 33685 + }, + { + "epoch": 2.49, + "learning_rate": 1.3186666927514385e-05, + "loss": 1.1235, + "step": 33686 + }, + { + "epoch": 2.49, + "learning_rate": 1.3186288847799044e-05, + "loss": 0.9166, + "step": 33687 + }, + { + "epoch": 2.49, + "learning_rate": 1.318591076301437e-05, + "loss": 0.9523, + "step": 33688 + }, + { + "epoch": 2.49, + "learning_rate": 1.3185532673160953e-05, + "loss": 0.9978, + "step": 33689 + }, + { + "epoch": 2.49, + "learning_rate": 1.3185154578239401e-05, + "loss": 1.0364, + "step": 33690 + }, + { + "epoch": 2.49, + "learning_rate": 1.3184776478250312e-05, + "loss": 0.9418, + "step": 33691 + }, + { + "epoch": 2.49, + "learning_rate": 1.3184398373194292e-05, + "loss": 0.9312, + "step": 33692 + }, + { + "epoch": 2.49, + "learning_rate": 1.3184020263071936e-05, + "loss": 0.9848, + "step": 33693 + }, + { + "epoch": 2.49, + "learning_rate": 1.318364214788385e-05, + "loss": 0.9012, + "step": 33694 + }, + { + "epoch": 2.49, + "learning_rate": 1.3183264027630636e-05, + "loss": 0.9269, + "step": 33695 + }, + { + "epoch": 2.49, + "learning_rate": 1.3182885902312893e-05, + "loss": 0.9823, + "step": 33696 + }, + { + "epoch": 2.49, + "learning_rate": 1.3182507771931225e-05, + "loss": 1.0153, + "step": 33697 + }, + { + "epoch": 2.49, + "learning_rate": 1.3182129636486237e-05, + "loss": 1.0843, + "step": 33698 + }, + { + "epoch": 2.49, + "learning_rate": 1.318175149597852e-05, + "loss": 0.942, + "step": 33699 + }, + { + "epoch": 2.49, + "learning_rate": 1.3181373350408686e-05, + "loss": 0.9145, + "step": 33700 + }, + { + "epoch": 2.49, + "learning_rate": 1.3180995199777331e-05, + "loss": 1.0577, + "step": 33701 + }, + { + "epoch": 2.49, + "learning_rate": 1.3180617044085058e-05, + "loss": 0.9136, + "step": 33702 + }, + { + "epoch": 2.49, + "learning_rate": 1.3180238883332468e-05, + "loss": 0.9296, + "step": 33703 + }, + { + "epoch": 2.49, + "learning_rate": 1.3179860717520162e-05, + "loss": 0.8724, + "step": 33704 + }, + { + "epoch": 2.49, + "learning_rate": 1.3179482546648746e-05, + "loss": 0.9581, + "step": 33705 + }, + { + "epoch": 2.49, + "learning_rate": 1.3179104370718818e-05, + "loss": 0.8946, + "step": 33706 + }, + { + "epoch": 2.49, + "learning_rate": 1.3178726189730982e-05, + "loss": 1.0287, + "step": 33707 + }, + { + "epoch": 2.49, + "learning_rate": 1.3178348003685835e-05, + "loss": 1.0363, + "step": 33708 + }, + { + "epoch": 2.49, + "learning_rate": 1.3177969812583987e-05, + "loss": 1.1078, + "step": 33709 + }, + { + "epoch": 2.49, + "learning_rate": 1.3177591616426031e-05, + "loss": 1.0172, + "step": 33710 + }, + { + "epoch": 2.49, + "learning_rate": 1.3177213415212568e-05, + "loss": 0.9589, + "step": 33711 + }, + { + "epoch": 2.49, + "learning_rate": 1.317683520894421e-05, + "loss": 0.9796, + "step": 33712 + }, + { + "epoch": 2.49, + "learning_rate": 1.3176456997621549e-05, + "loss": 1.0661, + "step": 33713 + }, + { + "epoch": 2.49, + "learning_rate": 1.3176078781245195e-05, + "loss": 1.0238, + "step": 33714 + }, + { + "epoch": 2.49, + "learning_rate": 1.317570055981574e-05, + "loss": 1.0924, + "step": 33715 + }, + { + "epoch": 2.49, + "learning_rate": 1.3175322333333793e-05, + "loss": 1.1064, + "step": 33716 + }, + { + "epoch": 2.49, + "learning_rate": 1.3174944101799954e-05, + "loss": 1.0197, + "step": 33717 + }, + { + "epoch": 2.49, + "learning_rate": 1.3174565865214826e-05, + "loss": 1.0254, + "step": 33718 + }, + { + "epoch": 2.49, + "learning_rate": 1.3174187623579004e-05, + "loss": 1.0182, + "step": 33719 + }, + { + "epoch": 2.49, + "learning_rate": 1.31738093768931e-05, + "loss": 0.9704, + "step": 33720 + }, + { + "epoch": 2.49, + "learning_rate": 1.3173431125157707e-05, + "loss": 1.0618, + "step": 33721 + }, + { + "epoch": 2.49, + "learning_rate": 1.3173052868373435e-05, + "loss": 0.9138, + "step": 33722 + }, + { + "epoch": 2.49, + "learning_rate": 1.3172674606540877e-05, + "loss": 1.0427, + "step": 33723 + }, + { + "epoch": 2.49, + "learning_rate": 1.317229633966064e-05, + "loss": 0.991, + "step": 33724 + }, + { + "epoch": 2.49, + "learning_rate": 1.3171918067733324e-05, + "loss": 0.9631, + "step": 33725 + }, + { + "epoch": 2.49, + "learning_rate": 1.3171539790759535e-05, + "loss": 0.9841, + "step": 33726 + }, + { + "epoch": 2.49, + "learning_rate": 1.3171161508739869e-05, + "loss": 1.0573, + "step": 33727 + }, + { + "epoch": 2.49, + "learning_rate": 1.317078322167493e-05, + "loss": 1.0989, + "step": 33728 + }, + { + "epoch": 2.49, + "learning_rate": 1.3170404929565323e-05, + "loss": 0.9592, + "step": 33729 + }, + { + "epoch": 2.49, + "learning_rate": 1.3170026632411643e-05, + "loss": 0.9171, + "step": 33730 + }, + { + "epoch": 2.49, + "learning_rate": 1.3169648330214502e-05, + "loss": 0.9285, + "step": 33731 + }, + { + "epoch": 2.49, + "learning_rate": 1.3169270022974492e-05, + "loss": 1.0623, + "step": 33732 + }, + { + "epoch": 2.49, + "learning_rate": 1.3168891710692217e-05, + "loss": 1.0633, + "step": 33733 + }, + { + "epoch": 2.49, + "learning_rate": 1.3168513393368282e-05, + "loss": 0.9841, + "step": 33734 + }, + { + "epoch": 2.49, + "learning_rate": 1.316813507100329e-05, + "loss": 1.022, + "step": 33735 + }, + { + "epoch": 2.49, + "learning_rate": 1.316775674359784e-05, + "loss": 0.8976, + "step": 33736 + }, + { + "epoch": 2.49, + "learning_rate": 1.3167378411152534e-05, + "loss": 1.0332, + "step": 33737 + }, + { + "epoch": 2.49, + "learning_rate": 1.3167000073667971e-05, + "loss": 0.9909, + "step": 33738 + }, + { + "epoch": 2.49, + "learning_rate": 1.316662173114476e-05, + "loss": 0.9918, + "step": 33739 + }, + { + "epoch": 2.49, + "learning_rate": 1.3166243383583495e-05, + "loss": 0.9936, + "step": 33740 + }, + { + "epoch": 2.49, + "learning_rate": 1.3165865030984785e-05, + "loss": 1.0507, + "step": 33741 + }, + { + "epoch": 2.49, + "learning_rate": 1.3165486673349229e-05, + "loss": 0.9758, + "step": 33742 + }, + { + "epoch": 2.49, + "learning_rate": 1.3165108310677428e-05, + "loss": 1.0652, + "step": 33743 + }, + { + "epoch": 2.49, + "learning_rate": 1.3164729942969985e-05, + "loss": 1.0429, + "step": 33744 + }, + { + "epoch": 2.49, + "learning_rate": 1.3164351570227503e-05, + "loss": 0.9721, + "step": 33745 + }, + { + "epoch": 2.49, + "learning_rate": 1.3163973192450582e-05, + "loss": 1.0877, + "step": 33746 + }, + { + "epoch": 2.49, + "learning_rate": 1.3163594809639828e-05, + "loss": 1.0204, + "step": 33747 + }, + { + "epoch": 2.49, + "learning_rate": 1.3163216421795836e-05, + "loss": 1.0069, + "step": 33748 + }, + { + "epoch": 2.49, + "learning_rate": 1.3162838028919214e-05, + "loss": 0.9878, + "step": 33749 + }, + { + "epoch": 2.49, + "learning_rate": 1.316245963101056e-05, + "loss": 1.0757, + "step": 33750 + }, + { + "epoch": 2.49, + "learning_rate": 1.316208122807048e-05, + "loss": 1.0027, + "step": 33751 + }, + { + "epoch": 2.49, + "learning_rate": 1.3161702820099573e-05, + "loss": 1.0904, + "step": 33752 + }, + { + "epoch": 2.49, + "learning_rate": 1.3161324407098444e-05, + "loss": 0.9141, + "step": 33753 + }, + { + "epoch": 2.49, + "learning_rate": 1.3160945989067688e-05, + "loss": 1.0167, + "step": 33754 + }, + { + "epoch": 2.49, + "learning_rate": 1.316056756600792e-05, + "loss": 1.0926, + "step": 33755 + }, + { + "epoch": 2.49, + "learning_rate": 1.3160189137919728e-05, + "loss": 1.0169, + "step": 33756 + }, + { + "epoch": 2.49, + "learning_rate": 1.3159810704803725e-05, + "loss": 0.8905, + "step": 33757 + }, + { + "epoch": 2.49, + "learning_rate": 1.3159432266660504e-05, + "loss": 1.0033, + "step": 33758 + }, + { + "epoch": 2.49, + "learning_rate": 1.3159053823490673e-05, + "loss": 1.0232, + "step": 33759 + }, + { + "epoch": 2.49, + "learning_rate": 1.3158675375294832e-05, + "loss": 1.1039, + "step": 33760 + }, + { + "epoch": 2.49, + "learning_rate": 1.3158296922073583e-05, + "loss": 0.9892, + "step": 33761 + }, + { + "epoch": 2.49, + "learning_rate": 1.3157918463827528e-05, + "loss": 1.0116, + "step": 33762 + }, + { + "epoch": 2.49, + "learning_rate": 1.3157540000557273e-05, + "loss": 0.8787, + "step": 33763 + }, + { + "epoch": 2.49, + "learning_rate": 1.3157161532263417e-05, + "loss": 0.9527, + "step": 33764 + }, + { + "epoch": 2.49, + "learning_rate": 1.3156783058946562e-05, + "loss": 0.9409, + "step": 33765 + }, + { + "epoch": 2.49, + "learning_rate": 1.315640458060731e-05, + "loss": 0.9712, + "step": 33766 + }, + { + "epoch": 2.5, + "learning_rate": 1.3156026097246262e-05, + "loss": 1.0752, + "step": 33767 + }, + { + "epoch": 2.5, + "learning_rate": 1.3155647608864022e-05, + "loss": 0.9432, + "step": 33768 + }, + { + "epoch": 2.5, + "learning_rate": 1.3155269115461193e-05, + "loss": 0.9822, + "step": 33769 + }, + { + "epoch": 2.5, + "learning_rate": 1.3154890617038373e-05, + "loss": 1.0106, + "step": 33770 + }, + { + "epoch": 2.5, + "learning_rate": 1.3154512113596172e-05, + "loss": 1.017, + "step": 33771 + }, + { + "epoch": 2.5, + "learning_rate": 1.3154133605135181e-05, + "loss": 0.9853, + "step": 33772 + }, + { + "epoch": 2.5, + "learning_rate": 1.3153755091656015e-05, + "loss": 0.9689, + "step": 33773 + }, + { + "epoch": 2.5, + "learning_rate": 1.3153376573159267e-05, + "loss": 1.006, + "step": 33774 + }, + { + "epoch": 2.5, + "learning_rate": 1.3152998049645542e-05, + "loss": 1.0408, + "step": 33775 + }, + { + "epoch": 2.5, + "learning_rate": 1.3152619521115444e-05, + "loss": 1.0842, + "step": 33776 + }, + { + "epoch": 2.5, + "learning_rate": 1.3152240987569571e-05, + "loss": 1.0178, + "step": 33777 + }, + { + "epoch": 2.5, + "learning_rate": 1.3151862449008528e-05, + "loss": 1.0416, + "step": 33778 + }, + { + "epoch": 2.5, + "learning_rate": 1.3151483905432917e-05, + "loss": 0.9884, + "step": 33779 + }, + { + "epoch": 2.5, + "learning_rate": 1.3151105356843341e-05, + "loss": 0.9219, + "step": 33780 + }, + { + "epoch": 2.5, + "learning_rate": 1.31507268032404e-05, + "loss": 0.9908, + "step": 33781 + }, + { + "epoch": 2.5, + "learning_rate": 1.3150348244624697e-05, + "loss": 1.1242, + "step": 33782 + }, + { + "epoch": 2.5, + "learning_rate": 1.3149969680996839e-05, + "loss": 1.108, + "step": 33783 + }, + { + "epoch": 2.5, + "learning_rate": 1.3149591112357422e-05, + "loss": 0.9189, + "step": 33784 + }, + { + "epoch": 2.5, + "learning_rate": 1.314921253870705e-05, + "loss": 0.9843, + "step": 33785 + }, + { + "epoch": 2.5, + "learning_rate": 1.3148833960046328e-05, + "loss": 0.8878, + "step": 33786 + }, + { + "epoch": 2.5, + "learning_rate": 1.3148455376375854e-05, + "loss": 0.9839, + "step": 33787 + }, + { + "epoch": 2.5, + "learning_rate": 1.3148076787696234e-05, + "loss": 0.9519, + "step": 33788 + }, + { + "epoch": 2.5, + "learning_rate": 1.3147698194008067e-05, + "loss": 1.0408, + "step": 33789 + }, + { + "epoch": 2.5, + "learning_rate": 1.3147319595311961e-05, + "loss": 1.0238, + "step": 33790 + }, + { + "epoch": 2.5, + "learning_rate": 1.314694099160851e-05, + "loss": 1.0093, + "step": 33791 + }, + { + "epoch": 2.5, + "learning_rate": 1.3146562382898323e-05, + "loss": 0.9596, + "step": 33792 + }, + { + "epoch": 2.5, + "learning_rate": 1.3146183769182001e-05, + "loss": 1.0055, + "step": 33793 + }, + { + "epoch": 2.5, + "learning_rate": 1.3145805150460147e-05, + "loss": 0.9954, + "step": 33794 + }, + { + "epoch": 2.5, + "learning_rate": 1.3145426526733358e-05, + "loss": 1.0688, + "step": 33795 + }, + { + "epoch": 2.5, + "learning_rate": 1.3145047898002244e-05, + "loss": 0.9108, + "step": 33796 + }, + { + "epoch": 2.5, + "learning_rate": 1.31446692642674e-05, + "loss": 0.951, + "step": 33797 + }, + { + "epoch": 2.5, + "learning_rate": 1.3144290625529435e-05, + "loss": 0.8888, + "step": 33798 + }, + { + "epoch": 2.5, + "learning_rate": 1.3143911981788947e-05, + "loss": 1.0422, + "step": 33799 + }, + { + "epoch": 2.5, + "learning_rate": 1.314353333304654e-05, + "loss": 0.9282, + "step": 33800 + }, + { + "epoch": 2.5, + "learning_rate": 1.3143154679302818e-05, + "loss": 0.9164, + "step": 33801 + }, + { + "epoch": 2.5, + "learning_rate": 1.3142776020558382e-05, + "loss": 0.9112, + "step": 33802 + }, + { + "epoch": 2.5, + "learning_rate": 1.3142397356813833e-05, + "loss": 0.9279, + "step": 33803 + }, + { + "epoch": 2.5, + "learning_rate": 1.3142018688069777e-05, + "loss": 0.9983, + "step": 33804 + }, + { + "epoch": 2.5, + "learning_rate": 1.3141640014326811e-05, + "loss": 0.9092, + "step": 33805 + }, + { + "epoch": 2.5, + "learning_rate": 1.3141261335585545e-05, + "loss": 1.049, + "step": 33806 + }, + { + "epoch": 2.5, + "learning_rate": 1.3140882651846572e-05, + "loss": 0.9796, + "step": 33807 + }, + { + "epoch": 2.5, + "learning_rate": 1.3140503963110502e-05, + "loss": 0.9371, + "step": 33808 + }, + { + "epoch": 2.5, + "learning_rate": 1.3140125269377934e-05, + "loss": 0.9664, + "step": 33809 + }, + { + "epoch": 2.5, + "learning_rate": 1.3139746570649474e-05, + "loss": 0.9876, + "step": 33810 + }, + { + "epoch": 2.5, + "learning_rate": 1.313936786692572e-05, + "loss": 0.9549, + "step": 33811 + }, + { + "epoch": 2.5, + "learning_rate": 1.313898915820728e-05, + "loss": 1.066, + "step": 33812 + }, + { + "epoch": 2.5, + "learning_rate": 1.3138610444494751e-05, + "loss": 1.0695, + "step": 33813 + }, + { + "epoch": 2.5, + "learning_rate": 1.3138231725788736e-05, + "loss": 0.9889, + "step": 33814 + }, + { + "epoch": 2.5, + "learning_rate": 1.313785300208984e-05, + "loss": 1.0023, + "step": 33815 + }, + { + "epoch": 2.5, + "learning_rate": 1.3137474273398666e-05, + "loss": 1.0335, + "step": 33816 + }, + { + "epoch": 2.5, + "learning_rate": 1.3137095539715814e-05, + "loss": 0.9675, + "step": 33817 + }, + { + "epoch": 2.5, + "learning_rate": 1.3136716801041889e-05, + "loss": 1.0406, + "step": 33818 + }, + { + "epoch": 2.5, + "learning_rate": 1.313633805737749e-05, + "loss": 0.9755, + "step": 33819 + }, + { + "epoch": 2.5, + "learning_rate": 1.3135959308723225e-05, + "loss": 0.9079, + "step": 33820 + }, + { + "epoch": 2.5, + "learning_rate": 1.3135580555079695e-05, + "loss": 1.0017, + "step": 33821 + }, + { + "epoch": 2.5, + "learning_rate": 1.3135201796447498e-05, + "loss": 1.026, + "step": 33822 + }, + { + "epoch": 2.5, + "learning_rate": 1.313482303282724e-05, + "loss": 1.0857, + "step": 33823 + }, + { + "epoch": 2.5, + "learning_rate": 1.3134444264219525e-05, + "loss": 0.8502, + "step": 33824 + }, + { + "epoch": 2.5, + "learning_rate": 1.3134065490624952e-05, + "loss": 0.9867, + "step": 33825 + }, + { + "epoch": 2.5, + "learning_rate": 1.3133686712044126e-05, + "loss": 0.9694, + "step": 33826 + }, + { + "epoch": 2.5, + "learning_rate": 1.313330792847765e-05, + "loss": 1.018, + "step": 33827 + }, + { + "epoch": 2.5, + "learning_rate": 1.3132929139926126e-05, + "loss": 1.0042, + "step": 33828 + }, + { + "epoch": 2.5, + "learning_rate": 1.3132550346390156e-05, + "loss": 1.0053, + "step": 33829 + }, + { + "epoch": 2.5, + "learning_rate": 1.3132171547870345e-05, + "loss": 0.9573, + "step": 33830 + }, + { + "epoch": 2.5, + "learning_rate": 1.3131792744367294e-05, + "loss": 0.917, + "step": 33831 + }, + { + "epoch": 2.5, + "learning_rate": 1.3131413935881602e-05, + "loss": 0.8875, + "step": 33832 + }, + { + "epoch": 2.5, + "learning_rate": 1.3131035122413878e-05, + "loss": 0.9365, + "step": 33833 + }, + { + "epoch": 2.5, + "learning_rate": 1.313065630396472e-05, + "loss": 0.9303, + "step": 33834 + }, + { + "epoch": 2.5, + "learning_rate": 1.3130277480534734e-05, + "loss": 0.9836, + "step": 33835 + }, + { + "epoch": 2.5, + "learning_rate": 1.3129898652124519e-05, + "loss": 0.9561, + "step": 33836 + }, + { + "epoch": 2.5, + "learning_rate": 1.3129519818734683e-05, + "loss": 0.9707, + "step": 33837 + }, + { + "epoch": 2.5, + "learning_rate": 1.3129140980365827e-05, + "loss": 1.0247, + "step": 33838 + }, + { + "epoch": 2.5, + "learning_rate": 1.3128762137018551e-05, + "loss": 0.915, + "step": 33839 + }, + { + "epoch": 2.5, + "learning_rate": 1.3128383288693458e-05, + "loss": 1.026, + "step": 33840 + }, + { + "epoch": 2.5, + "learning_rate": 1.3128004435391155e-05, + "loss": 1.022, + "step": 33841 + }, + { + "epoch": 2.5, + "learning_rate": 1.3127625577112237e-05, + "loss": 1.0794, + "step": 33842 + }, + { + "epoch": 2.5, + "learning_rate": 1.3127246713857314e-05, + "loss": 1.0302, + "step": 33843 + }, + { + "epoch": 2.5, + "learning_rate": 1.3126867845626985e-05, + "loss": 0.9265, + "step": 33844 + }, + { + "epoch": 2.5, + "learning_rate": 1.3126488972421856e-05, + "loss": 1.0102, + "step": 33845 + }, + { + "epoch": 2.5, + "learning_rate": 1.3126110094242526e-05, + "loss": 1.0375, + "step": 33846 + }, + { + "epoch": 2.5, + "learning_rate": 1.3125731211089601e-05, + "loss": 1.1279, + "step": 33847 + }, + { + "epoch": 2.5, + "learning_rate": 1.312535232296368e-05, + "loss": 0.9317, + "step": 33848 + }, + { + "epoch": 2.5, + "learning_rate": 1.3124973429865374e-05, + "loss": 1.0503, + "step": 33849 + }, + { + "epoch": 2.5, + "learning_rate": 1.3124594531795274e-05, + "loss": 0.9787, + "step": 33850 + }, + { + "epoch": 2.5, + "learning_rate": 1.3124215628753991e-05, + "loss": 0.9674, + "step": 33851 + }, + { + "epoch": 2.5, + "learning_rate": 1.3123836720742125e-05, + "loss": 1.0492, + "step": 33852 + }, + { + "epoch": 2.5, + "learning_rate": 1.312345780776028e-05, + "loss": 0.9059, + "step": 33853 + }, + { + "epoch": 2.5, + "learning_rate": 1.3123078889809056e-05, + "loss": 1.1121, + "step": 33854 + }, + { + "epoch": 2.5, + "learning_rate": 1.3122699966889061e-05, + "loss": 1.0781, + "step": 33855 + }, + { + "epoch": 2.5, + "learning_rate": 1.3122321039000894e-05, + "loss": 1.0509, + "step": 33856 + }, + { + "epoch": 2.5, + "learning_rate": 1.312194210614516e-05, + "loss": 0.9136, + "step": 33857 + }, + { + "epoch": 2.5, + "learning_rate": 1.312156316832246e-05, + "loss": 0.9828, + "step": 33858 + }, + { + "epoch": 2.5, + "learning_rate": 1.3121184225533398e-05, + "loss": 1.0243, + "step": 33859 + }, + { + "epoch": 2.5, + "learning_rate": 1.3120805277778575e-05, + "loss": 0.9242, + "step": 33860 + }, + { + "epoch": 2.5, + "learning_rate": 1.3120426325058597e-05, + "loss": 1.0855, + "step": 33861 + }, + { + "epoch": 2.5, + "learning_rate": 1.3120047367374064e-05, + "loss": 1.0272, + "step": 33862 + }, + { + "epoch": 2.5, + "learning_rate": 1.3119668404725583e-05, + "loss": 1.0521, + "step": 33863 + }, + { + "epoch": 2.5, + "learning_rate": 1.3119289437113751e-05, + "loss": 0.9808, + "step": 33864 + }, + { + "epoch": 2.5, + "learning_rate": 1.3118910464539177e-05, + "loss": 0.9491, + "step": 33865 + }, + { + "epoch": 2.5, + "learning_rate": 1.3118531487002457e-05, + "loss": 0.9407, + "step": 33866 + }, + { + "epoch": 2.5, + "learning_rate": 1.3118152504504205e-05, + "loss": 1.1104, + "step": 33867 + }, + { + "epoch": 2.5, + "learning_rate": 1.3117773517045011e-05, + "loss": 0.978, + "step": 33868 + }, + { + "epoch": 2.5, + "learning_rate": 1.3117394524625486e-05, + "loss": 1.058, + "step": 33869 + }, + { + "epoch": 2.5, + "learning_rate": 1.3117015527246228e-05, + "loss": 1.0052, + "step": 33870 + }, + { + "epoch": 2.5, + "learning_rate": 1.3116636524907848e-05, + "loss": 0.9689, + "step": 33871 + }, + { + "epoch": 2.5, + "learning_rate": 1.311625751761094e-05, + "loss": 0.9593, + "step": 33872 + }, + { + "epoch": 2.5, + "learning_rate": 1.3115878505356112e-05, + "loss": 1.0599, + "step": 33873 + }, + { + "epoch": 2.5, + "learning_rate": 1.3115499488143965e-05, + "loss": 1.0526, + "step": 33874 + }, + { + "epoch": 2.5, + "learning_rate": 1.3115120465975107e-05, + "loss": 0.8724, + "step": 33875 + }, + { + "epoch": 2.5, + "learning_rate": 1.3114741438850132e-05, + "loss": 0.9302, + "step": 33876 + }, + { + "epoch": 2.5, + "learning_rate": 1.311436240676965e-05, + "loss": 1.0634, + "step": 33877 + }, + { + "epoch": 2.5, + "learning_rate": 1.3113983369734263e-05, + "loss": 1.1238, + "step": 33878 + }, + { + "epoch": 2.5, + "learning_rate": 1.311360432774457e-05, + "loss": 1.0106, + "step": 33879 + }, + { + "epoch": 2.5, + "learning_rate": 1.311322528080118e-05, + "loss": 1.0588, + "step": 33880 + }, + { + "epoch": 2.5, + "learning_rate": 1.311284622890469e-05, + "loss": 1.012, + "step": 33881 + }, + { + "epoch": 2.5, + "learning_rate": 1.3112467172055709e-05, + "loss": 1.1527, + "step": 33882 + }, + { + "epoch": 2.5, + "learning_rate": 1.3112088110254836e-05, + "loss": 0.946, + "step": 33883 + }, + { + "epoch": 2.5, + "learning_rate": 1.3111709043502677e-05, + "loss": 1.0199, + "step": 33884 + }, + { + "epoch": 2.5, + "learning_rate": 1.3111329971799833e-05, + "loss": 0.9864, + "step": 33885 + }, + { + "epoch": 2.5, + "learning_rate": 1.3110950895146907e-05, + "loss": 0.9494, + "step": 33886 + }, + { + "epoch": 2.5, + "learning_rate": 1.3110571813544502e-05, + "loss": 1.0099, + "step": 33887 + }, + { + "epoch": 2.5, + "learning_rate": 1.3110192726993224e-05, + "loss": 1.0503, + "step": 33888 + }, + { + "epoch": 2.5, + "learning_rate": 1.3109813635493668e-05, + "loss": 0.9652, + "step": 33889 + }, + { + "epoch": 2.5, + "learning_rate": 1.3109434539046448e-05, + "loss": 1.0207, + "step": 33890 + }, + { + "epoch": 2.5, + "learning_rate": 1.310905543765216e-05, + "loss": 1.0477, + "step": 33891 + }, + { + "epoch": 2.5, + "learning_rate": 1.3108676331311411e-05, + "loss": 1.0488, + "step": 33892 + }, + { + "epoch": 2.5, + "learning_rate": 1.3108297220024802e-05, + "loss": 1.1276, + "step": 33893 + }, + { + "epoch": 2.5, + "learning_rate": 1.3107918103792937e-05, + "loss": 0.9362, + "step": 33894 + }, + { + "epoch": 2.5, + "learning_rate": 1.3107538982616419e-05, + "loss": 1.0915, + "step": 33895 + }, + { + "epoch": 2.5, + "learning_rate": 1.310715985649585e-05, + "loss": 1.0045, + "step": 33896 + }, + { + "epoch": 2.5, + "learning_rate": 1.3106780725431833e-05, + "loss": 0.8578, + "step": 33897 + }, + { + "epoch": 2.5, + "learning_rate": 1.3106401589424974e-05, + "loss": 0.998, + "step": 33898 + }, + { + "epoch": 2.5, + "learning_rate": 1.3106022448475875e-05, + "loss": 1.0469, + "step": 33899 + }, + { + "epoch": 2.5, + "learning_rate": 1.3105643302585139e-05, + "loss": 1.1113, + "step": 33900 + }, + { + "epoch": 2.5, + "learning_rate": 1.3105264151753365e-05, + "loss": 1.1048, + "step": 33901 + }, + { + "epoch": 2.5, + "learning_rate": 1.3104884995981164e-05, + "loss": 0.9717, + "step": 33902 + }, + { + "epoch": 2.51, + "learning_rate": 1.3104505835269132e-05, + "loss": 0.9656, + "step": 33903 + }, + { + "epoch": 2.51, + "learning_rate": 1.310412666961788e-05, + "loss": 1.0775, + "step": 33904 + }, + { + "epoch": 2.51, + "learning_rate": 1.3103747499028005e-05, + "loss": 1.0557, + "step": 33905 + }, + { + "epoch": 2.51, + "learning_rate": 1.3103368323500112e-05, + "loss": 1.0146, + "step": 33906 + }, + { + "epoch": 2.51, + "learning_rate": 1.3102989143034803e-05, + "loss": 0.9652, + "step": 33907 + }, + { + "epoch": 2.51, + "learning_rate": 1.3102609957632685e-05, + "loss": 0.9653, + "step": 33908 + }, + { + "epoch": 2.51, + "learning_rate": 1.3102230767294357e-05, + "loss": 0.9618, + "step": 33909 + }, + { + "epoch": 2.51, + "learning_rate": 1.3101851572020428e-05, + "loss": 1.0923, + "step": 33910 + }, + { + "epoch": 2.51, + "learning_rate": 1.3101472371811492e-05, + "loss": 1.0967, + "step": 33911 + }, + { + "epoch": 2.51, + "learning_rate": 1.3101093166668162e-05, + "loss": 0.9139, + "step": 33912 + }, + { + "epoch": 2.51, + "learning_rate": 1.3100713956591033e-05, + "loss": 1.0833, + "step": 33913 + }, + { + "epoch": 2.51, + "learning_rate": 1.3100334741580718e-05, + "loss": 0.9651, + "step": 33914 + }, + { + "epoch": 2.51, + "learning_rate": 1.3099955521637812e-05, + "loss": 0.9375, + "step": 33915 + }, + { + "epoch": 2.51, + "learning_rate": 1.3099576296762921e-05, + "loss": 0.9679, + "step": 33916 + }, + { + "epoch": 2.51, + "learning_rate": 1.3099197066956648e-05, + "loss": 1.081, + "step": 33917 + }, + { + "epoch": 2.51, + "learning_rate": 1.3098817832219597e-05, + "loss": 0.9978, + "step": 33918 + }, + { + "epoch": 2.51, + "learning_rate": 1.3098438592552371e-05, + "loss": 0.9977, + "step": 33919 + }, + { + "epoch": 2.51, + "learning_rate": 1.3098059347955573e-05, + "loss": 1.0772, + "step": 33920 + }, + { + "epoch": 2.51, + "learning_rate": 1.3097680098429808e-05, + "loss": 0.9926, + "step": 33921 + }, + { + "epoch": 2.51, + "learning_rate": 1.309730084397568e-05, + "loss": 1.0171, + "step": 33922 + }, + { + "epoch": 2.51, + "learning_rate": 1.3096921584593787e-05, + "loss": 0.9127, + "step": 33923 + }, + { + "epoch": 2.51, + "learning_rate": 1.309654232028474e-05, + "loss": 0.9363, + "step": 33924 + }, + { + "epoch": 2.51, + "learning_rate": 1.3096163051049136e-05, + "loss": 1.0121, + "step": 33925 + }, + { + "epoch": 2.51, + "learning_rate": 1.3095783776887583e-05, + "loss": 0.9663, + "step": 33926 + }, + { + "epoch": 2.51, + "learning_rate": 1.309540449780068e-05, + "loss": 1.0418, + "step": 33927 + }, + { + "epoch": 2.51, + "learning_rate": 1.3095025213789035e-05, + "loss": 1.0425, + "step": 33928 + }, + { + "epoch": 2.51, + "learning_rate": 1.3094645924853243e-05, + "loss": 0.9671, + "step": 33929 + }, + { + "epoch": 2.51, + "learning_rate": 1.309426663099392e-05, + "loss": 0.8069, + "step": 33930 + }, + { + "epoch": 2.51, + "learning_rate": 1.3093887332211661e-05, + "loss": 1.0818, + "step": 33931 + }, + { + "epoch": 2.51, + "learning_rate": 1.3093508028507071e-05, + "loss": 1.052, + "step": 33932 + }, + { + "epoch": 2.51, + "learning_rate": 1.3093128719880757e-05, + "loss": 0.9853, + "step": 33933 + }, + { + "epoch": 2.51, + "learning_rate": 1.3092749406333316e-05, + "loss": 0.9843, + "step": 33934 + }, + { + "epoch": 2.51, + "learning_rate": 1.3092370087865357e-05, + "loss": 0.9864, + "step": 33935 + }, + { + "epoch": 2.51, + "learning_rate": 1.3091990764477478e-05, + "loss": 1.0145, + "step": 33936 + }, + { + "epoch": 2.51, + "learning_rate": 1.309161143617029e-05, + "loss": 1.098, + "step": 33937 + }, + { + "epoch": 2.51, + "learning_rate": 1.3091232102944388e-05, + "loss": 0.9388, + "step": 33938 + }, + { + "epoch": 2.51, + "learning_rate": 1.3090852764800386e-05, + "loss": 1.0416, + "step": 33939 + }, + { + "epoch": 2.51, + "learning_rate": 1.3090473421738876e-05, + "loss": 1.0373, + "step": 33940 + }, + { + "epoch": 2.51, + "learning_rate": 1.3090094073760472e-05, + "loss": 1.0531, + "step": 33941 + }, + { + "epoch": 2.51, + "learning_rate": 1.308971472086577e-05, + "loss": 1.0524, + "step": 33942 + }, + { + "epoch": 2.51, + "learning_rate": 1.3089335363055375e-05, + "loss": 0.9852, + "step": 33943 + }, + { + "epoch": 2.51, + "learning_rate": 1.3088956000329893e-05, + "loss": 0.9756, + "step": 33944 + }, + { + "epoch": 2.51, + "learning_rate": 1.3088576632689926e-05, + "loss": 1.0118, + "step": 33945 + }, + { + "epoch": 2.51, + "learning_rate": 1.3088197260136078e-05, + "loss": 1.1004, + "step": 33946 + }, + { + "epoch": 2.51, + "learning_rate": 1.3087817882668952e-05, + "loss": 1.0914, + "step": 33947 + }, + { + "epoch": 2.51, + "learning_rate": 1.3087438500289151e-05, + "loss": 1.0501, + "step": 33948 + }, + { + "epoch": 2.51, + "learning_rate": 1.308705911299728e-05, + "loss": 1.045, + "step": 33949 + }, + { + "epoch": 2.51, + "learning_rate": 1.3086679720793942e-05, + "loss": 1.0278, + "step": 33950 + }, + { + "epoch": 2.51, + "learning_rate": 1.3086300323679741e-05, + "loss": 0.9999, + "step": 33951 + }, + { + "epoch": 2.51, + "learning_rate": 1.3085920921655282e-05, + "loss": 1.1188, + "step": 33952 + }, + { + "epoch": 2.51, + "learning_rate": 1.3085541514721167e-05, + "loss": 1.0665, + "step": 33953 + }, + { + "epoch": 2.51, + "learning_rate": 1.3085162102877997e-05, + "loss": 1.0658, + "step": 33954 + }, + { + "epoch": 2.51, + "learning_rate": 1.308478268612638e-05, + "loss": 1.07, + "step": 33955 + }, + { + "epoch": 2.51, + "learning_rate": 1.3084403264466916e-05, + "loss": 0.9667, + "step": 33956 + }, + { + "epoch": 2.51, + "learning_rate": 1.3084023837900213e-05, + "loss": 0.8979, + "step": 33957 + }, + { + "epoch": 2.51, + "learning_rate": 1.3083644406426869e-05, + "loss": 1.0324, + "step": 33958 + }, + { + "epoch": 2.51, + "learning_rate": 1.3083264970047493e-05, + "loss": 0.9479, + "step": 33959 + }, + { + "epoch": 2.51, + "learning_rate": 1.3082885528762686e-05, + "loss": 0.9694, + "step": 33960 + }, + { + "epoch": 2.51, + "learning_rate": 1.3082506082573056e-05, + "loss": 0.9973, + "step": 33961 + }, + { + "epoch": 2.51, + "learning_rate": 1.3082126631479197e-05, + "loss": 0.9951, + "step": 33962 + }, + { + "epoch": 2.51, + "learning_rate": 1.3081747175481722e-05, + "loss": 0.9758, + "step": 33963 + }, + { + "epoch": 2.51, + "learning_rate": 1.308136771458123e-05, + "loss": 1.0511, + "step": 33964 + }, + { + "epoch": 2.51, + "learning_rate": 1.3080988248778328e-05, + "loss": 1.0037, + "step": 33965 + }, + { + "epoch": 2.51, + "learning_rate": 1.3080608778073613e-05, + "loss": 0.915, + "step": 33966 + }, + { + "epoch": 2.51, + "learning_rate": 1.30802293024677e-05, + "loss": 0.8402, + "step": 33967 + }, + { + "epoch": 2.51, + "learning_rate": 1.307984982196118e-05, + "loss": 0.9452, + "step": 33968 + }, + { + "epoch": 2.51, + "learning_rate": 1.3079470336554668e-05, + "loss": 1.0434, + "step": 33969 + }, + { + "epoch": 2.51, + "learning_rate": 1.3079090846248762e-05, + "loss": 1.0245, + "step": 33970 + }, + { + "epoch": 2.51, + "learning_rate": 1.3078711351044064e-05, + "loss": 0.9812, + "step": 33971 + }, + { + "epoch": 2.51, + "learning_rate": 1.3078331850941182e-05, + "loss": 0.9664, + "step": 33972 + }, + { + "epoch": 2.51, + "learning_rate": 1.3077952345940718e-05, + "loss": 0.9256, + "step": 33973 + }, + { + "epoch": 2.51, + "learning_rate": 1.3077572836043276e-05, + "loss": 0.9346, + "step": 33974 + }, + { + "epoch": 2.51, + "learning_rate": 1.307719332124946e-05, + "loss": 1.1072, + "step": 33975 + }, + { + "epoch": 2.51, + "learning_rate": 1.307681380155987e-05, + "loss": 0.9723, + "step": 33976 + }, + { + "epoch": 2.51, + "learning_rate": 1.3076434276975119e-05, + "loss": 0.9689, + "step": 33977 + }, + { + "epoch": 2.51, + "learning_rate": 1.30760547474958e-05, + "loss": 1.035, + "step": 33978 + }, + { + "epoch": 2.51, + "learning_rate": 1.3075675213122523e-05, + "loss": 1.087, + "step": 33979 + }, + { + "epoch": 2.51, + "learning_rate": 1.3075295673855895e-05, + "loss": 1.0814, + "step": 33980 + }, + { + "epoch": 2.51, + "learning_rate": 1.3074916129696509e-05, + "loss": 0.9474, + "step": 33981 + }, + { + "epoch": 2.51, + "learning_rate": 1.307453658064498e-05, + "loss": 1.0125, + "step": 33982 + }, + { + "epoch": 2.51, + "learning_rate": 1.3074157026701904e-05, + "loss": 0.9791, + "step": 33983 + }, + { + "epoch": 2.51, + "learning_rate": 1.3073777467867892e-05, + "loss": 0.9891, + "step": 33984 + }, + { + "epoch": 2.51, + "learning_rate": 1.3073397904143542e-05, + "loss": 1.006, + "step": 33985 + }, + { + "epoch": 2.51, + "learning_rate": 1.307301833552946e-05, + "loss": 1.1591, + "step": 33986 + }, + { + "epoch": 2.51, + "learning_rate": 1.307263876202625e-05, + "loss": 0.9207, + "step": 33987 + }, + { + "epoch": 2.51, + "learning_rate": 1.3072259183634515e-05, + "loss": 1.0531, + "step": 33988 + }, + { + "epoch": 2.51, + "learning_rate": 1.3071879600354863e-05, + "loss": 0.9819, + "step": 33989 + }, + { + "epoch": 2.51, + "learning_rate": 1.307150001218789e-05, + "loss": 1.047, + "step": 33990 + }, + { + "epoch": 2.51, + "learning_rate": 1.3071120419134205e-05, + "loss": 0.9614, + "step": 33991 + }, + { + "epoch": 2.51, + "learning_rate": 1.3070740821194415e-05, + "loss": 0.9389, + "step": 33992 + }, + { + "epoch": 2.51, + "learning_rate": 1.3070361218369115e-05, + "loss": 1.0177, + "step": 33993 + }, + { + "epoch": 2.51, + "learning_rate": 1.3069981610658916e-05, + "loss": 1.056, + "step": 33994 + }, + { + "epoch": 2.51, + "learning_rate": 1.3069601998064423e-05, + "loss": 1.0824, + "step": 33995 + }, + { + "epoch": 2.51, + "learning_rate": 1.3069222380586235e-05, + "loss": 0.9434, + "step": 33996 + }, + { + "epoch": 2.51, + "learning_rate": 1.3068842758224957e-05, + "loss": 1.103, + "step": 33997 + }, + { + "epoch": 2.51, + "learning_rate": 1.3068463130981196e-05, + "loss": 1.1088, + "step": 33998 + }, + { + "epoch": 2.51, + "learning_rate": 1.3068083498855554e-05, + "loss": 1.0422, + "step": 33999 + }, + { + "epoch": 2.51, + "learning_rate": 1.3067703861848634e-05, + "loss": 1.024, + "step": 34000 + }, + { + "epoch": 2.51, + "learning_rate": 1.3067324219961041e-05, + "loss": 0.9597, + "step": 34001 + }, + { + "epoch": 2.51, + "learning_rate": 1.3066944573193383e-05, + "loss": 1.0172, + "step": 34002 + }, + { + "epoch": 2.51, + "learning_rate": 1.3066564921546254e-05, + "loss": 1.0647, + "step": 34003 + }, + { + "epoch": 2.51, + "learning_rate": 1.3066185265020268e-05, + "loss": 0.976, + "step": 34004 + }, + { + "epoch": 2.51, + "learning_rate": 1.3065805603616024e-05, + "loss": 0.9778, + "step": 34005 + }, + { + "epoch": 2.51, + "learning_rate": 1.3065425937334126e-05, + "loss": 0.9218, + "step": 34006 + }, + { + "epoch": 2.51, + "learning_rate": 1.306504626617518e-05, + "loss": 0.9604, + "step": 34007 + }, + { + "epoch": 2.51, + "learning_rate": 1.3064666590139792e-05, + "loss": 1.0048, + "step": 34008 + }, + { + "epoch": 2.51, + "learning_rate": 1.306428690922856e-05, + "loss": 0.9526, + "step": 34009 + }, + { + "epoch": 2.51, + "learning_rate": 1.3063907223442094e-05, + "loss": 0.9778, + "step": 34010 + }, + { + "epoch": 2.51, + "learning_rate": 1.3063527532780991e-05, + "loss": 1.1069, + "step": 34011 + }, + { + "epoch": 2.51, + "learning_rate": 1.3063147837245865e-05, + "loss": 0.9952, + "step": 34012 + }, + { + "epoch": 2.51, + "learning_rate": 1.3062768136837309e-05, + "loss": 0.9176, + "step": 34013 + }, + { + "epoch": 2.51, + "learning_rate": 1.3062388431555937e-05, + "loss": 0.9767, + "step": 34014 + }, + { + "epoch": 2.51, + "learning_rate": 1.3062008721402347e-05, + "loss": 1.0227, + "step": 34015 + }, + { + "epoch": 2.51, + "learning_rate": 1.3061629006377145e-05, + "loss": 1.0166, + "step": 34016 + }, + { + "epoch": 2.51, + "learning_rate": 1.3061249286480938e-05, + "loss": 1.0257, + "step": 34017 + }, + { + "epoch": 2.51, + "learning_rate": 1.3060869561714325e-05, + "loss": 0.9839, + "step": 34018 + }, + { + "epoch": 2.51, + "learning_rate": 1.306048983207791e-05, + "loss": 0.99, + "step": 34019 + }, + { + "epoch": 2.51, + "learning_rate": 1.30601100975723e-05, + "loss": 1.0277, + "step": 34020 + }, + { + "epoch": 2.51, + "learning_rate": 1.3059730358198099e-05, + "loss": 1.128, + "step": 34021 + }, + { + "epoch": 2.51, + "learning_rate": 1.3059350613955912e-05, + "loss": 1.0311, + "step": 34022 + }, + { + "epoch": 2.51, + "learning_rate": 1.3058970864846341e-05, + "loss": 1.0026, + "step": 34023 + }, + { + "epoch": 2.51, + "learning_rate": 1.3058591110869993e-05, + "loss": 1.1185, + "step": 34024 + }, + { + "epoch": 2.51, + "learning_rate": 1.3058211352027467e-05, + "loss": 1.0017, + "step": 34025 + }, + { + "epoch": 2.51, + "learning_rate": 1.3057831588319373e-05, + "loss": 0.971, + "step": 34026 + }, + { + "epoch": 2.51, + "learning_rate": 1.3057451819746314e-05, + "loss": 1.0892, + "step": 34027 + }, + { + "epoch": 2.51, + "learning_rate": 1.3057072046308891e-05, + "loss": 0.9853, + "step": 34028 + }, + { + "epoch": 2.51, + "learning_rate": 1.305669226800771e-05, + "loss": 0.982, + "step": 34029 + }, + { + "epoch": 2.51, + "learning_rate": 1.3056312484843374e-05, + "loss": 1.0234, + "step": 34030 + }, + { + "epoch": 2.51, + "learning_rate": 1.3055932696816491e-05, + "loss": 1.0224, + "step": 34031 + }, + { + "epoch": 2.51, + "learning_rate": 1.3055552903927664e-05, + "loss": 1.115, + "step": 34032 + }, + { + "epoch": 2.51, + "learning_rate": 1.3055173106177492e-05, + "loss": 1.031, + "step": 34033 + }, + { + "epoch": 2.51, + "learning_rate": 1.3054793303566586e-05, + "loss": 1.1085, + "step": 34034 + }, + { + "epoch": 2.51, + "learning_rate": 1.3054413496095546e-05, + "loss": 1.1189, + "step": 34035 + }, + { + "epoch": 2.51, + "learning_rate": 1.305403368376498e-05, + "loss": 1.0372, + "step": 34036 + }, + { + "epoch": 2.51, + "learning_rate": 1.3053653866575488e-05, + "loss": 0.9818, + "step": 34037 + }, + { + "epoch": 2.52, + "learning_rate": 1.3053274044527678e-05, + "loss": 1.0451, + "step": 34038 + }, + { + "epoch": 2.52, + "learning_rate": 1.3052894217622151e-05, + "loss": 0.9706, + "step": 34039 + }, + { + "epoch": 2.52, + "learning_rate": 1.3052514385859514e-05, + "loss": 0.9862, + "step": 34040 + }, + { + "epoch": 2.52, + "learning_rate": 1.3052134549240373e-05, + "loss": 1.0412, + "step": 34041 + }, + { + "epoch": 2.52, + "learning_rate": 1.3051754707765323e-05, + "loss": 1.0539, + "step": 34042 + }, + { + "epoch": 2.52, + "learning_rate": 1.305137486143498e-05, + "loss": 0.9652, + "step": 34043 + }, + { + "epoch": 2.52, + "learning_rate": 1.3050995010249941e-05, + "loss": 0.91, + "step": 34044 + }, + { + "epoch": 2.52, + "learning_rate": 1.3050615154210818e-05, + "loss": 0.9335, + "step": 34045 + }, + { + "epoch": 2.52, + "learning_rate": 1.3050235293318202e-05, + "loss": 1.0594, + "step": 34046 + }, + { + "epoch": 2.52, + "learning_rate": 1.3049855427572712e-05, + "loss": 0.9612, + "step": 34047 + }, + { + "epoch": 2.52, + "learning_rate": 1.304947555697494e-05, + "loss": 1.0545, + "step": 34048 + }, + { + "epoch": 2.52, + "learning_rate": 1.30490956815255e-05, + "loss": 0.9123, + "step": 34049 + }, + { + "epoch": 2.52, + "learning_rate": 1.3048715801224989e-05, + "loss": 0.9598, + "step": 34050 + }, + { + "epoch": 2.52, + "learning_rate": 1.3048335916074017e-05, + "loss": 1.0291, + "step": 34051 + }, + { + "epoch": 2.52, + "learning_rate": 1.3047956026073183e-05, + "loss": 1.0802, + "step": 34052 + }, + { + "epoch": 2.52, + "learning_rate": 1.30475761312231e-05, + "loss": 0.9901, + "step": 34053 + }, + { + "epoch": 2.52, + "learning_rate": 1.3047196231524364e-05, + "loss": 0.9812, + "step": 34054 + }, + { + "epoch": 2.52, + "learning_rate": 1.3046816326977583e-05, + "loss": 0.9465, + "step": 34055 + }, + { + "epoch": 2.52, + "learning_rate": 1.304643641758336e-05, + "loss": 0.9561, + "step": 34056 + }, + { + "epoch": 2.52, + "learning_rate": 1.3046056503342302e-05, + "loss": 1.0182, + "step": 34057 + }, + { + "epoch": 2.52, + "learning_rate": 1.304567658425501e-05, + "loss": 0.8642, + "step": 34058 + }, + { + "epoch": 2.52, + "learning_rate": 1.304529666032209e-05, + "loss": 0.8934, + "step": 34059 + }, + { + "epoch": 2.52, + "learning_rate": 1.3044916731544147e-05, + "loss": 1.084, + "step": 34060 + }, + { + "epoch": 2.52, + "learning_rate": 1.3044536797921786e-05, + "loss": 1.0906, + "step": 34061 + }, + { + "epoch": 2.52, + "learning_rate": 1.3044156859455608e-05, + "loss": 0.983, + "step": 34062 + }, + { + "epoch": 2.52, + "learning_rate": 1.3043776916146223e-05, + "loss": 0.9028, + "step": 34063 + }, + { + "epoch": 2.52, + "learning_rate": 1.3043396967994231e-05, + "loss": 1.0105, + "step": 34064 + }, + { + "epoch": 2.52, + "learning_rate": 1.3043017015000239e-05, + "loss": 1.0058, + "step": 34065 + }, + { + "epoch": 2.52, + "learning_rate": 1.3042637057164848e-05, + "loss": 1.0329, + "step": 34066 + }, + { + "epoch": 2.52, + "learning_rate": 1.3042257094488668e-05, + "loss": 0.9526, + "step": 34067 + }, + { + "epoch": 2.52, + "learning_rate": 1.3041877126972297e-05, + "loss": 0.9291, + "step": 34068 + }, + { + "epoch": 2.52, + "learning_rate": 1.3041497154616348e-05, + "loss": 0.9652, + "step": 34069 + }, + { + "epoch": 2.52, + "learning_rate": 1.3041117177421416e-05, + "loss": 0.9185, + "step": 34070 + }, + { + "epoch": 2.52, + "learning_rate": 1.3040737195388112e-05, + "loss": 1.0236, + "step": 34071 + }, + { + "epoch": 2.52, + "learning_rate": 1.3040357208517037e-05, + "loss": 1.0757, + "step": 34072 + }, + { + "epoch": 2.52, + "learning_rate": 1.3039977216808799e-05, + "loss": 1.0822, + "step": 34073 + }, + { + "epoch": 2.52, + "learning_rate": 1.3039597220264003e-05, + "loss": 0.9468, + "step": 34074 + }, + { + "epoch": 2.52, + "learning_rate": 1.303921721888325e-05, + "loss": 0.9143, + "step": 34075 + }, + { + "epoch": 2.52, + "learning_rate": 1.3038837212667141e-05, + "loss": 0.9274, + "step": 34076 + }, + { + "epoch": 2.52, + "learning_rate": 1.303845720161629e-05, + "loss": 1.0025, + "step": 34077 + }, + { + "epoch": 2.52, + "learning_rate": 1.3038077185731295e-05, + "loss": 0.912, + "step": 34078 + }, + { + "epoch": 2.52, + "learning_rate": 1.3037697165012766e-05, + "loss": 0.996, + "step": 34079 + }, + { + "epoch": 2.52, + "learning_rate": 1.30373171394613e-05, + "loss": 0.918, + "step": 34080 + }, + { + "epoch": 2.52, + "learning_rate": 1.3036937109077508e-05, + "loss": 0.9009, + "step": 34081 + }, + { + "epoch": 2.52, + "learning_rate": 1.3036557073861994e-05, + "loss": 0.9777, + "step": 34082 + }, + { + "epoch": 2.52, + "learning_rate": 1.3036177033815359e-05, + "loss": 0.9243, + "step": 34083 + }, + { + "epoch": 2.52, + "learning_rate": 1.303579698893821e-05, + "loss": 1.0338, + "step": 34084 + }, + { + "epoch": 2.52, + "learning_rate": 1.3035416939231151e-05, + "loss": 0.9955, + "step": 34085 + }, + { + "epoch": 2.52, + "learning_rate": 1.3035036884694786e-05, + "loss": 0.9825, + "step": 34086 + }, + { + "epoch": 2.52, + "learning_rate": 1.3034656825329724e-05, + "loss": 1.0231, + "step": 34087 + }, + { + "epoch": 2.52, + "learning_rate": 1.3034276761136563e-05, + "loss": 1.0367, + "step": 34088 + }, + { + "epoch": 2.52, + "learning_rate": 1.3033896692115913e-05, + "loss": 1.0865, + "step": 34089 + }, + { + "epoch": 2.52, + "learning_rate": 1.3033516618268378e-05, + "loss": 1.0027, + "step": 34090 + }, + { + "epoch": 2.52, + "learning_rate": 1.3033136539594559e-05, + "loss": 1.0251, + "step": 34091 + }, + { + "epoch": 2.52, + "learning_rate": 1.3032756456095068e-05, + "loss": 0.9805, + "step": 34092 + }, + { + "epoch": 2.52, + "learning_rate": 1.3032376367770501e-05, + "loss": 1.0474, + "step": 34093 + }, + { + "epoch": 2.52, + "learning_rate": 1.3031996274621465e-05, + "loss": 0.9694, + "step": 34094 + }, + { + "epoch": 2.52, + "learning_rate": 1.303161617664857e-05, + "loss": 1.0915, + "step": 34095 + }, + { + "epoch": 2.52, + "learning_rate": 1.3031236073852416e-05, + "loss": 1.0131, + "step": 34096 + }, + { + "epoch": 2.52, + "learning_rate": 1.3030855966233605e-05, + "loss": 1.0806, + "step": 34097 + }, + { + "epoch": 2.52, + "learning_rate": 1.303047585379275e-05, + "loss": 0.9759, + "step": 34098 + }, + { + "epoch": 2.52, + "learning_rate": 1.3030095736530447e-05, + "loss": 0.9449, + "step": 34099 + }, + { + "epoch": 2.52, + "learning_rate": 1.302971561444731e-05, + "loss": 1.028, + "step": 34100 + }, + { + "epoch": 2.52, + "learning_rate": 1.3029335487543938e-05, + "loss": 0.947, + "step": 34101 + }, + { + "epoch": 2.52, + "learning_rate": 1.3028955355820935e-05, + "loss": 0.9525, + "step": 34102 + }, + { + "epoch": 2.52, + "learning_rate": 1.3028575219278908e-05, + "loss": 0.9346, + "step": 34103 + }, + { + "epoch": 2.52, + "learning_rate": 1.3028195077918459e-05, + "loss": 0.9668, + "step": 34104 + }, + { + "epoch": 2.52, + "learning_rate": 1.3027814931740197e-05, + "loss": 0.9519, + "step": 34105 + }, + { + "epoch": 2.52, + "learning_rate": 1.3027434780744726e-05, + "loss": 1.0561, + "step": 34106 + }, + { + "epoch": 2.52, + "learning_rate": 1.3027054624932647e-05, + "loss": 1.0212, + "step": 34107 + }, + { + "epoch": 2.52, + "learning_rate": 1.3026674464304567e-05, + "loss": 1.0225, + "step": 34108 + }, + { + "epoch": 2.52, + "learning_rate": 1.3026294298861092e-05, + "loss": 1.1355, + "step": 34109 + }, + { + "epoch": 2.52, + "learning_rate": 1.302591412860283e-05, + "loss": 1.013, + "step": 34110 + }, + { + "epoch": 2.52, + "learning_rate": 1.3025533953530379e-05, + "loss": 1.0035, + "step": 34111 + }, + { + "epoch": 2.52, + "learning_rate": 1.3025153773644346e-05, + "loss": 0.9542, + "step": 34112 + }, + { + "epoch": 2.52, + "learning_rate": 1.3024773588945336e-05, + "loss": 1.0939, + "step": 34113 + }, + { + "epoch": 2.52, + "learning_rate": 1.3024393399433956e-05, + "loss": 0.9904, + "step": 34114 + }, + { + "epoch": 2.52, + "learning_rate": 1.3024013205110809e-05, + "loss": 1.0, + "step": 34115 + }, + { + "epoch": 2.52, + "learning_rate": 1.3023633005976502e-05, + "loss": 0.9566, + "step": 34116 + }, + { + "epoch": 2.52, + "learning_rate": 1.3023252802031633e-05, + "loss": 0.9967, + "step": 34117 + }, + { + "epoch": 2.52, + "learning_rate": 1.3022872593276817e-05, + "loss": 1.0582, + "step": 34118 + }, + { + "epoch": 2.52, + "learning_rate": 1.302249237971265e-05, + "loss": 1.0436, + "step": 34119 + }, + { + "epoch": 2.52, + "learning_rate": 1.3022112161339746e-05, + "loss": 0.9969, + "step": 34120 + }, + { + "epoch": 2.52, + "learning_rate": 1.30217319381587e-05, + "loss": 0.9679, + "step": 34121 + }, + { + "epoch": 2.52, + "learning_rate": 1.3021351710170127e-05, + "loss": 1.0449, + "step": 34122 + }, + { + "epoch": 2.52, + "learning_rate": 1.3020971477374622e-05, + "loss": 1.0854, + "step": 34123 + }, + { + "epoch": 2.52, + "learning_rate": 1.3020591239772795e-05, + "loss": 0.9827, + "step": 34124 + }, + { + "epoch": 2.52, + "learning_rate": 1.3020210997365253e-05, + "loss": 1.0644, + "step": 34125 + }, + { + "epoch": 2.52, + "learning_rate": 1.3019830750152596e-05, + "loss": 0.9088, + "step": 34126 + }, + { + "epoch": 2.52, + "learning_rate": 1.301945049813543e-05, + "loss": 1.0342, + "step": 34127 + }, + { + "epoch": 2.52, + "learning_rate": 1.3019070241314367e-05, + "loss": 1.059, + "step": 34128 + }, + { + "epoch": 2.52, + "learning_rate": 1.3018689979690004e-05, + "loss": 0.9285, + "step": 34129 + }, + { + "epoch": 2.52, + "learning_rate": 1.3018309713262946e-05, + "loss": 0.9521, + "step": 34130 + }, + { + "epoch": 2.52, + "learning_rate": 1.3017929442033804e-05, + "loss": 0.9202, + "step": 34131 + }, + { + "epoch": 2.52, + "learning_rate": 1.3017549166003177e-05, + "loss": 0.9841, + "step": 34132 + }, + { + "epoch": 2.52, + "learning_rate": 1.3017168885171673e-05, + "loss": 1.0106, + "step": 34133 + }, + { + "epoch": 2.52, + "learning_rate": 1.3016788599539896e-05, + "loss": 0.9332, + "step": 34134 + }, + { + "epoch": 2.52, + "learning_rate": 1.3016408309108453e-05, + "loss": 1.1069, + "step": 34135 + }, + { + "epoch": 2.52, + "learning_rate": 1.3016028013877949e-05, + "loss": 0.9647, + "step": 34136 + }, + { + "epoch": 2.52, + "learning_rate": 1.3015647713848983e-05, + "loss": 1.0105, + "step": 34137 + }, + { + "epoch": 2.52, + "learning_rate": 1.3015267409022168e-05, + "loss": 1.0224, + "step": 34138 + }, + { + "epoch": 2.52, + "learning_rate": 1.3014887099398104e-05, + "loss": 1.0439, + "step": 34139 + }, + { + "epoch": 2.52, + "learning_rate": 1.30145067849774e-05, + "loss": 0.8986, + "step": 34140 + }, + { + "epoch": 2.52, + "learning_rate": 1.3014126465760655e-05, + "loss": 1.002, + "step": 34141 + }, + { + "epoch": 2.52, + "learning_rate": 1.301374614174848e-05, + "loss": 1.005, + "step": 34142 + }, + { + "epoch": 2.52, + "learning_rate": 1.3013365812941478e-05, + "loss": 1.0128, + "step": 34143 + }, + { + "epoch": 2.52, + "learning_rate": 1.3012985479340254e-05, + "loss": 0.913, + "step": 34144 + }, + { + "epoch": 2.52, + "learning_rate": 1.3012605140945413e-05, + "loss": 1.0418, + "step": 34145 + }, + { + "epoch": 2.52, + "learning_rate": 1.301222479775756e-05, + "loss": 0.9844, + "step": 34146 + }, + { + "epoch": 2.52, + "learning_rate": 1.3011844449777302e-05, + "loss": 0.9585, + "step": 34147 + }, + { + "epoch": 2.52, + "learning_rate": 1.3011464097005244e-05, + "loss": 0.9597, + "step": 34148 + }, + { + "epoch": 2.52, + "learning_rate": 1.3011083739441987e-05, + "loss": 0.9664, + "step": 34149 + }, + { + "epoch": 2.52, + "learning_rate": 1.3010703377088139e-05, + "loss": 0.9757, + "step": 34150 + }, + { + "epoch": 2.52, + "learning_rate": 1.3010323009944305e-05, + "loss": 1.0606, + "step": 34151 + }, + { + "epoch": 2.52, + "learning_rate": 1.300994263801109e-05, + "loss": 0.9432, + "step": 34152 + }, + { + "epoch": 2.52, + "learning_rate": 1.3009562261289101e-05, + "loss": 0.9102, + "step": 34153 + }, + { + "epoch": 2.52, + "learning_rate": 1.3009181879778938e-05, + "loss": 0.8846, + "step": 34154 + }, + { + "epoch": 2.52, + "learning_rate": 1.3008801493481213e-05, + "loss": 0.9612, + "step": 34155 + }, + { + "epoch": 2.52, + "learning_rate": 1.3008421102396527e-05, + "loss": 1.1167, + "step": 34156 + }, + { + "epoch": 2.52, + "learning_rate": 1.3008040706525488e-05, + "loss": 0.9972, + "step": 34157 + }, + { + "epoch": 2.52, + "learning_rate": 1.3007660305868696e-05, + "loss": 1.0132, + "step": 34158 + }, + { + "epoch": 2.52, + "learning_rate": 1.3007279900426761e-05, + "loss": 1.0203, + "step": 34159 + }, + { + "epoch": 2.52, + "learning_rate": 1.3006899490200287e-05, + "loss": 0.968, + "step": 34160 + }, + { + "epoch": 2.52, + "learning_rate": 1.3006519075189878e-05, + "loss": 0.9549, + "step": 34161 + }, + { + "epoch": 2.52, + "learning_rate": 1.3006138655396139e-05, + "loss": 1.0475, + "step": 34162 + }, + { + "epoch": 2.52, + "learning_rate": 1.3005758230819678e-05, + "loss": 0.8022, + "step": 34163 + }, + { + "epoch": 2.52, + "learning_rate": 1.3005377801461099e-05, + "loss": 1.0439, + "step": 34164 + }, + { + "epoch": 2.52, + "learning_rate": 1.3004997367321006e-05, + "loss": 0.9018, + "step": 34165 + }, + { + "epoch": 2.52, + "learning_rate": 1.3004616928400004e-05, + "loss": 0.9019, + "step": 34166 + }, + { + "epoch": 2.52, + "learning_rate": 1.3004236484698707e-05, + "loss": 0.9607, + "step": 34167 + }, + { + "epoch": 2.52, + "learning_rate": 1.3003856036217703e-05, + "loss": 1.0783, + "step": 34168 + }, + { + "epoch": 2.52, + "learning_rate": 1.3003475582957612e-05, + "loss": 1.0188, + "step": 34169 + }, + { + "epoch": 2.52, + "learning_rate": 1.3003095124919032e-05, + "loss": 1.0388, + "step": 34170 + }, + { + "epoch": 2.52, + "learning_rate": 1.3002714662102573e-05, + "loss": 0.9989, + "step": 34171 + }, + { + "epoch": 2.52, + "learning_rate": 1.3002334194508837e-05, + "loss": 1.0634, + "step": 34172 + }, + { + "epoch": 2.53, + "learning_rate": 1.3001953722138429e-05, + "loss": 0.9336, + "step": 34173 + }, + { + "epoch": 2.53, + "learning_rate": 1.3001573244991956e-05, + "loss": 0.9019, + "step": 34174 + }, + { + "epoch": 2.53, + "learning_rate": 1.3001192763070025e-05, + "loss": 0.9509, + "step": 34175 + }, + { + "epoch": 2.53, + "learning_rate": 1.3000812276373238e-05, + "loss": 1.0007, + "step": 34176 + }, + { + "epoch": 2.53, + "learning_rate": 1.3000431784902201e-05, + "loss": 1.0499, + "step": 34177 + }, + { + "epoch": 2.53, + "learning_rate": 1.3000051288657521e-05, + "loss": 1.0027, + "step": 34178 + }, + { + "epoch": 2.53, + "learning_rate": 1.29996707876398e-05, + "loss": 1.0109, + "step": 34179 + }, + { + "epoch": 2.53, + "learning_rate": 1.2999290281849648e-05, + "loss": 1.0429, + "step": 34180 + }, + { + "epoch": 2.53, + "learning_rate": 1.299890977128767e-05, + "loss": 1.0026, + "step": 34181 + }, + { + "epoch": 2.53, + "learning_rate": 1.2998529255954465e-05, + "loss": 1.0286, + "step": 34182 + }, + { + "epoch": 2.53, + "learning_rate": 1.2998148735850643e-05, + "loss": 1.0499, + "step": 34183 + }, + { + "epoch": 2.53, + "learning_rate": 1.2997768210976812e-05, + "loss": 0.9773, + "step": 34184 + }, + { + "epoch": 2.53, + "learning_rate": 1.2997387681333575e-05, + "loss": 1.0366, + "step": 34185 + }, + { + "epoch": 2.53, + "learning_rate": 1.2997007146921536e-05, + "loss": 0.9219, + "step": 34186 + }, + { + "epoch": 2.53, + "learning_rate": 1.2996626607741301e-05, + "loss": 1.0272, + "step": 34187 + }, + { + "epoch": 2.53, + "learning_rate": 1.2996246063793477e-05, + "loss": 0.9969, + "step": 34188 + }, + { + "epoch": 2.53, + "learning_rate": 1.2995865515078666e-05, + "loss": 1.1562, + "step": 34189 + }, + { + "epoch": 2.53, + "learning_rate": 1.2995484961597478e-05, + "loss": 0.9885, + "step": 34190 + }, + { + "epoch": 2.53, + "learning_rate": 1.2995104403350514e-05, + "loss": 1.1015, + "step": 34191 + }, + { + "epoch": 2.53, + "learning_rate": 1.2994723840338386e-05, + "loss": 0.8618, + "step": 34192 + }, + { + "epoch": 2.53, + "learning_rate": 1.2994343272561692e-05, + "loss": 0.9671, + "step": 34193 + }, + { + "epoch": 2.53, + "learning_rate": 1.299396270002104e-05, + "loss": 1.0334, + "step": 34194 + }, + { + "epoch": 2.53, + "learning_rate": 1.2993582122717039e-05, + "loss": 1.0996, + "step": 34195 + }, + { + "epoch": 2.53, + "learning_rate": 1.299320154065029e-05, + "loss": 1.0404, + "step": 34196 + }, + { + "epoch": 2.53, + "learning_rate": 1.2992820953821401e-05, + "loss": 1.0966, + "step": 34197 + }, + { + "epoch": 2.53, + "learning_rate": 1.2992440362230976e-05, + "loss": 0.9676, + "step": 34198 + }, + { + "epoch": 2.53, + "learning_rate": 1.2992059765879622e-05, + "loss": 1.0652, + "step": 34199 + }, + { + "epoch": 2.53, + "learning_rate": 1.2991679164767943e-05, + "loss": 0.9648, + "step": 34200 + }, + { + "epoch": 2.53, + "learning_rate": 1.2991298558896543e-05, + "loss": 1.0298, + "step": 34201 + }, + { + "epoch": 2.53, + "learning_rate": 1.2990917948266033e-05, + "loss": 0.9825, + "step": 34202 + }, + { + "epoch": 2.53, + "learning_rate": 1.2990537332877013e-05, + "loss": 1.0283, + "step": 34203 + }, + { + "epoch": 2.53, + "learning_rate": 1.2990156712730096e-05, + "loss": 0.9407, + "step": 34204 + }, + { + "epoch": 2.53, + "learning_rate": 1.2989776087825878e-05, + "loss": 0.9744, + "step": 34205 + }, + { + "epoch": 2.53, + "learning_rate": 1.298939545816497e-05, + "loss": 1.085, + "step": 34206 + }, + { + "epoch": 2.53, + "learning_rate": 1.2989014823747975e-05, + "loss": 1.0262, + "step": 34207 + }, + { + "epoch": 2.53, + "learning_rate": 1.2988634184575502e-05, + "loss": 1.0583, + "step": 34208 + }, + { + "epoch": 2.53, + "learning_rate": 1.2988253540648152e-05, + "loss": 1.0537, + "step": 34209 + }, + { + "epoch": 2.53, + "learning_rate": 1.2987872891966537e-05, + "loss": 1.0573, + "step": 34210 + }, + { + "epoch": 2.53, + "learning_rate": 1.2987492238531258e-05, + "loss": 0.9472, + "step": 34211 + }, + { + "epoch": 2.53, + "learning_rate": 1.2987111580342921e-05, + "loss": 0.9919, + "step": 34212 + }, + { + "epoch": 2.53, + "learning_rate": 1.2986730917402133e-05, + "loss": 1.0132, + "step": 34213 + }, + { + "epoch": 2.53, + "learning_rate": 1.29863502497095e-05, + "loss": 1.0598, + "step": 34214 + }, + { + "epoch": 2.53, + "learning_rate": 1.2985969577265624e-05, + "loss": 0.9314, + "step": 34215 + }, + { + "epoch": 2.53, + "learning_rate": 1.2985588900071115e-05, + "loss": 1.0044, + "step": 34216 + }, + { + "epoch": 2.53, + "learning_rate": 1.2985208218126575e-05, + "loss": 1.0344, + "step": 34217 + }, + { + "epoch": 2.53, + "learning_rate": 1.2984827531432612e-05, + "loss": 0.9743, + "step": 34218 + }, + { + "epoch": 2.53, + "learning_rate": 1.298444683998983e-05, + "loss": 1.057, + "step": 34219 + }, + { + "epoch": 2.53, + "learning_rate": 1.2984066143798837e-05, + "loss": 0.9771, + "step": 34220 + }, + { + "epoch": 2.53, + "learning_rate": 1.2983685442860236e-05, + "loss": 1.0289, + "step": 34221 + }, + { + "epoch": 2.53, + "learning_rate": 1.2983304737174638e-05, + "loss": 0.9522, + "step": 34222 + }, + { + "epoch": 2.53, + "learning_rate": 1.2982924026742642e-05, + "loss": 0.9921, + "step": 34223 + }, + { + "epoch": 2.53, + "learning_rate": 1.2982543311564857e-05, + "loss": 1.0513, + "step": 34224 + }, + { + "epoch": 2.53, + "learning_rate": 1.2982162591641889e-05, + "loss": 0.9453, + "step": 34225 + }, + { + "epoch": 2.53, + "learning_rate": 1.2981781866974344e-05, + "loss": 1.0024, + "step": 34226 + }, + { + "epoch": 2.53, + "learning_rate": 1.2981401137562823e-05, + "loss": 0.9502, + "step": 34227 + }, + { + "epoch": 2.53, + "learning_rate": 1.2981020403407938e-05, + "loss": 0.9812, + "step": 34228 + }, + { + "epoch": 2.53, + "learning_rate": 1.298063966451029e-05, + "loss": 1.1026, + "step": 34229 + }, + { + "epoch": 2.53, + "learning_rate": 1.298025892087049e-05, + "loss": 1.0201, + "step": 34230 + }, + { + "epoch": 2.53, + "learning_rate": 1.2979878172489138e-05, + "loss": 1.0612, + "step": 34231 + }, + { + "epoch": 2.53, + "learning_rate": 1.2979497419366844e-05, + "loss": 1.0234, + "step": 34232 + }, + { + "epoch": 2.53, + "learning_rate": 1.2979116661504212e-05, + "loss": 1.0099, + "step": 34233 + }, + { + "epoch": 2.53, + "learning_rate": 1.2978735898901849e-05, + "loss": 0.9607, + "step": 34234 + }, + { + "epoch": 2.53, + "learning_rate": 1.2978355131560359e-05, + "loss": 1.0158, + "step": 34235 + }, + { + "epoch": 2.53, + "learning_rate": 1.2977974359480348e-05, + "loss": 0.9798, + "step": 34236 + }, + { + "epoch": 2.53, + "learning_rate": 1.2977593582662423e-05, + "loss": 1.0201, + "step": 34237 + }, + { + "epoch": 2.53, + "learning_rate": 1.2977212801107186e-05, + "loss": 1.0197, + "step": 34238 + }, + { + "epoch": 2.53, + "learning_rate": 1.297683201481525e-05, + "loss": 0.9001, + "step": 34239 + }, + { + "epoch": 2.53, + "learning_rate": 1.2976451223787217e-05, + "loss": 0.9542, + "step": 34240 + }, + { + "epoch": 2.53, + "learning_rate": 1.2976070428023692e-05, + "loss": 0.8328, + "step": 34241 + }, + { + "epoch": 2.53, + "learning_rate": 1.297568962752528e-05, + "loss": 0.9412, + "step": 34242 + }, + { + "epoch": 2.53, + "learning_rate": 1.2975308822292592e-05, + "loss": 1.0068, + "step": 34243 + }, + { + "epoch": 2.53, + "learning_rate": 1.2974928012326227e-05, + "loss": 1.0637, + "step": 34244 + }, + { + "epoch": 2.53, + "learning_rate": 1.2974547197626794e-05, + "loss": 1.0616, + "step": 34245 + }, + { + "epoch": 2.53, + "learning_rate": 1.2974166378194899e-05, + "loss": 0.9752, + "step": 34246 + }, + { + "epoch": 2.53, + "learning_rate": 1.2973785554031149e-05, + "loss": 0.9439, + "step": 34247 + }, + { + "epoch": 2.53, + "learning_rate": 1.2973404725136147e-05, + "loss": 0.9451, + "step": 34248 + }, + { + "epoch": 2.53, + "learning_rate": 1.2973023891510504e-05, + "loss": 0.971, + "step": 34249 + }, + { + "epoch": 2.53, + "learning_rate": 1.2972643053154818e-05, + "loss": 1.0245, + "step": 34250 + }, + { + "epoch": 2.53, + "learning_rate": 1.2972262210069706e-05, + "loss": 1.1592, + "step": 34251 + }, + { + "epoch": 2.53, + "learning_rate": 1.2971881362255761e-05, + "loss": 1.0963, + "step": 34252 + }, + { + "epoch": 2.53, + "learning_rate": 1.29715005097136e-05, + "loss": 0.9706, + "step": 34253 + }, + { + "epoch": 2.53, + "learning_rate": 1.297111965244382e-05, + "loss": 1.0714, + "step": 34254 + }, + { + "epoch": 2.53, + "learning_rate": 1.2970738790447033e-05, + "loss": 1.0742, + "step": 34255 + }, + { + "epoch": 2.53, + "learning_rate": 1.2970357923723842e-05, + "loss": 1.0806, + "step": 34256 + }, + { + "epoch": 2.53, + "learning_rate": 1.2969977052274855e-05, + "loss": 1.0878, + "step": 34257 + }, + { + "epoch": 2.53, + "learning_rate": 1.2969596176100677e-05, + "loss": 1.0631, + "step": 34258 + }, + { + "epoch": 2.53, + "learning_rate": 1.2969215295201916e-05, + "loss": 0.9255, + "step": 34259 + }, + { + "epoch": 2.53, + "learning_rate": 1.2968834409579172e-05, + "loss": 1.0044, + "step": 34260 + }, + { + "epoch": 2.53, + "learning_rate": 1.296845351923306e-05, + "loss": 0.9325, + "step": 34261 + }, + { + "epoch": 2.53, + "learning_rate": 1.2968072624164176e-05, + "loss": 0.9788, + "step": 34262 + }, + { + "epoch": 2.53, + "learning_rate": 1.2967691724373132e-05, + "loss": 1.0199, + "step": 34263 + }, + { + "epoch": 2.53, + "learning_rate": 1.2967310819860535e-05, + "loss": 1.04, + "step": 34264 + }, + { + "epoch": 2.53, + "learning_rate": 1.2966929910626987e-05, + "loss": 1.0619, + "step": 34265 + }, + { + "epoch": 2.53, + "learning_rate": 1.2966548996673094e-05, + "loss": 0.8745, + "step": 34266 + }, + { + "epoch": 2.53, + "learning_rate": 1.2966168077999468e-05, + "loss": 1.0049, + "step": 34267 + }, + { + "epoch": 2.53, + "learning_rate": 1.2965787154606708e-05, + "loss": 1.0082, + "step": 34268 + }, + { + "epoch": 2.53, + "learning_rate": 1.2965406226495429e-05, + "loss": 1.0927, + "step": 34269 + }, + { + "epoch": 2.53, + "learning_rate": 1.2965025293666225e-05, + "loss": 1.0508, + "step": 34270 + }, + { + "epoch": 2.53, + "learning_rate": 1.2964644356119712e-05, + "loss": 1.1181, + "step": 34271 + }, + { + "epoch": 2.53, + "learning_rate": 1.2964263413856489e-05, + "loss": 0.9459, + "step": 34272 + }, + { + "epoch": 2.53, + "learning_rate": 1.2963882466877166e-05, + "loss": 1.0473, + "step": 34273 + }, + { + "epoch": 2.53, + "learning_rate": 1.2963501515182349e-05, + "loss": 1.009, + "step": 34274 + }, + { + "epoch": 2.53, + "learning_rate": 1.2963120558772644e-05, + "loss": 1.1029, + "step": 34275 + }, + { + "epoch": 2.53, + "learning_rate": 1.2962739597648655e-05, + "loss": 1.0136, + "step": 34276 + }, + { + "epoch": 2.53, + "learning_rate": 1.296235863181099e-05, + "loss": 1.053, + "step": 34277 + }, + { + "epoch": 2.53, + "learning_rate": 1.2961977661260254e-05, + "loss": 1.0031, + "step": 34278 + }, + { + "epoch": 2.53, + "learning_rate": 1.2961596685997058e-05, + "loss": 1.1634, + "step": 34279 + }, + { + "epoch": 2.53, + "learning_rate": 1.2961215706022e-05, + "loss": 1.0686, + "step": 34280 + }, + { + "epoch": 2.53, + "learning_rate": 1.2960834721335692e-05, + "loss": 1.0127, + "step": 34281 + }, + { + "epoch": 2.53, + "learning_rate": 1.2960453731938738e-05, + "loss": 0.9532, + "step": 34282 + }, + { + "epoch": 2.53, + "learning_rate": 1.2960072737831743e-05, + "loss": 1.0721, + "step": 34283 + }, + { + "epoch": 2.53, + "learning_rate": 1.2959691739015319e-05, + "loss": 0.9807, + "step": 34284 + }, + { + "epoch": 2.53, + "learning_rate": 1.2959310735490062e-05, + "loss": 0.8744, + "step": 34285 + }, + { + "epoch": 2.53, + "learning_rate": 1.2958929727256588e-05, + "loss": 1.0806, + "step": 34286 + }, + { + "epoch": 2.53, + "learning_rate": 1.29585487143155e-05, + "loss": 1.0408, + "step": 34287 + }, + { + "epoch": 2.53, + "learning_rate": 1.2958167696667401e-05, + "loss": 1.0234, + "step": 34288 + }, + { + "epoch": 2.53, + "learning_rate": 1.2957786674312897e-05, + "loss": 1.0575, + "step": 34289 + }, + { + "epoch": 2.53, + "learning_rate": 1.2957405647252601e-05, + "loss": 1.0166, + "step": 34290 + }, + { + "epoch": 2.53, + "learning_rate": 1.2957024615487113e-05, + "loss": 1.0069, + "step": 34291 + }, + { + "epoch": 2.53, + "learning_rate": 1.2956643579017044e-05, + "loss": 1.0631, + "step": 34292 + }, + { + "epoch": 2.53, + "learning_rate": 1.2956262537842992e-05, + "loss": 1.0029, + "step": 34293 + }, + { + "epoch": 2.53, + "learning_rate": 1.2955881491965572e-05, + "loss": 1.0817, + "step": 34294 + }, + { + "epoch": 2.53, + "learning_rate": 1.2955500441385387e-05, + "loss": 1.0808, + "step": 34295 + }, + { + "epoch": 2.53, + "learning_rate": 1.2955119386103041e-05, + "loss": 1.0303, + "step": 34296 + }, + { + "epoch": 2.53, + "learning_rate": 1.2954738326119145e-05, + "loss": 1.0178, + "step": 34297 + }, + { + "epoch": 2.53, + "learning_rate": 1.2954357261434303e-05, + "loss": 0.9817, + "step": 34298 + }, + { + "epoch": 2.53, + "learning_rate": 1.2953976192049118e-05, + "loss": 1.0251, + "step": 34299 + }, + { + "epoch": 2.53, + "learning_rate": 1.2953595117964202e-05, + "loss": 0.9988, + "step": 34300 + }, + { + "epoch": 2.53, + "learning_rate": 1.2953214039180156e-05, + "loss": 1.0206, + "step": 34301 + }, + { + "epoch": 2.53, + "learning_rate": 1.295283295569759e-05, + "loss": 1.0583, + "step": 34302 + }, + { + "epoch": 2.53, + "learning_rate": 1.2952451867517107e-05, + "loss": 0.9244, + "step": 34303 + }, + { + "epoch": 2.53, + "learning_rate": 1.2952070774639319e-05, + "loss": 0.9516, + "step": 34304 + }, + { + "epoch": 2.53, + "learning_rate": 1.2951689677064824e-05, + "loss": 1.0505, + "step": 34305 + }, + { + "epoch": 2.53, + "learning_rate": 1.2951308574794235e-05, + "loss": 1.0646, + "step": 34306 + }, + { + "epoch": 2.53, + "learning_rate": 1.2950927467828157e-05, + "loss": 0.9565, + "step": 34307 + }, + { + "epoch": 2.53, + "learning_rate": 1.2950546356167197e-05, + "loss": 1.1151, + "step": 34308 + }, + { + "epoch": 2.54, + "learning_rate": 1.2950165239811959e-05, + "loss": 0.9997, + "step": 34309 + }, + { + "epoch": 2.54, + "learning_rate": 1.2949784118763049e-05, + "loss": 0.9295, + "step": 34310 + }, + { + "epoch": 2.54, + "learning_rate": 1.2949402993021074e-05, + "loss": 0.9618, + "step": 34311 + }, + { + "epoch": 2.54, + "learning_rate": 1.2949021862586644e-05, + "loss": 0.9849, + "step": 34312 + }, + { + "epoch": 2.54, + "learning_rate": 1.294864072746036e-05, + "loss": 0.9554, + "step": 34313 + }, + { + "epoch": 2.54, + "learning_rate": 1.2948259587642831e-05, + "loss": 0.9106, + "step": 34314 + }, + { + "epoch": 2.54, + "learning_rate": 1.2947878443134663e-05, + "loss": 0.9319, + "step": 34315 + }, + { + "epoch": 2.54, + "learning_rate": 1.2947497293936466e-05, + "loss": 1.0244, + "step": 34316 + }, + { + "epoch": 2.54, + "learning_rate": 1.2947116140048838e-05, + "loss": 0.9631, + "step": 34317 + }, + { + "epoch": 2.54, + "learning_rate": 1.2946734981472396e-05, + "loss": 1.0565, + "step": 34318 + }, + { + "epoch": 2.54, + "learning_rate": 1.2946353818207735e-05, + "loss": 1.0547, + "step": 34319 + }, + { + "epoch": 2.54, + "learning_rate": 1.2945972650255472e-05, + "loss": 1.0221, + "step": 34320 + }, + { + "epoch": 2.54, + "learning_rate": 1.2945591477616204e-05, + "loss": 0.937, + "step": 34321 + }, + { + "epoch": 2.54, + "learning_rate": 1.2945210300290546e-05, + "loss": 1.0354, + "step": 34322 + }, + { + "epoch": 2.54, + "learning_rate": 1.2944829118279096e-05, + "loss": 1.0356, + "step": 34323 + }, + { + "epoch": 2.54, + "learning_rate": 1.2944447931582471e-05, + "loss": 0.9724, + "step": 34324 + }, + { + "epoch": 2.54, + "learning_rate": 1.2944066740201268e-05, + "loss": 1.0619, + "step": 34325 + }, + { + "epoch": 2.54, + "learning_rate": 1.2943685544136098e-05, + "loss": 0.9831, + "step": 34326 + }, + { + "epoch": 2.54, + "learning_rate": 1.2943304343387566e-05, + "loss": 0.9656, + "step": 34327 + }, + { + "epoch": 2.54, + "learning_rate": 1.2942923137956279e-05, + "loss": 0.9361, + "step": 34328 + }, + { + "epoch": 2.54, + "learning_rate": 1.2942541927842842e-05, + "loss": 0.9562, + "step": 34329 + }, + { + "epoch": 2.54, + "learning_rate": 1.2942160713047865e-05, + "loss": 0.9234, + "step": 34330 + }, + { + "epoch": 2.54, + "learning_rate": 1.2941779493571951e-05, + "loss": 1.0459, + "step": 34331 + }, + { + "epoch": 2.54, + "learning_rate": 1.2941398269415708e-05, + "loss": 1.0671, + "step": 34332 + }, + { + "epoch": 2.54, + "learning_rate": 1.2941017040579743e-05, + "loss": 1.0667, + "step": 34333 + }, + { + "epoch": 2.54, + "learning_rate": 1.2940635807064661e-05, + "loss": 1.1115, + "step": 34334 + }, + { + "epoch": 2.54, + "learning_rate": 1.2940254568871071e-05, + "loss": 0.9834, + "step": 34335 + }, + { + "epoch": 2.54, + "learning_rate": 1.2939873325999577e-05, + "loss": 0.9563, + "step": 34336 + }, + { + "epoch": 2.54, + "learning_rate": 1.2939492078450789e-05, + "loss": 0.9134, + "step": 34337 + }, + { + "epoch": 2.54, + "learning_rate": 1.2939110826225307e-05, + "loss": 0.9934, + "step": 34338 + }, + { + "epoch": 2.54, + "learning_rate": 1.2938729569323744e-05, + "loss": 1.0472, + "step": 34339 + }, + { + "epoch": 2.54, + "learning_rate": 1.2938348307746705e-05, + "loss": 0.9935, + "step": 34340 + }, + { + "epoch": 2.54, + "learning_rate": 1.2937967041494794e-05, + "loss": 1.0451, + "step": 34341 + }, + { + "epoch": 2.54, + "learning_rate": 1.2937585770568619e-05, + "loss": 0.9489, + "step": 34342 + }, + { + "epoch": 2.54, + "learning_rate": 1.293720449496879e-05, + "loss": 0.9362, + "step": 34343 + }, + { + "epoch": 2.54, + "learning_rate": 1.293682321469591e-05, + "loss": 1.0364, + "step": 34344 + }, + { + "epoch": 2.54, + "learning_rate": 1.2936441929750588e-05, + "loss": 0.9641, + "step": 34345 + }, + { + "epoch": 2.54, + "learning_rate": 1.2936060640133424e-05, + "loss": 0.9211, + "step": 34346 + }, + { + "epoch": 2.54, + "learning_rate": 1.2935679345845031e-05, + "loss": 0.9729, + "step": 34347 + }, + { + "epoch": 2.54, + "learning_rate": 1.2935298046886015e-05, + "loss": 1.0427, + "step": 34348 + }, + { + "epoch": 2.54, + "learning_rate": 1.2934916743256982e-05, + "loss": 1.1375, + "step": 34349 + }, + { + "epoch": 2.54, + "learning_rate": 1.2934535434958537e-05, + "loss": 0.9968, + "step": 34350 + }, + { + "epoch": 2.54, + "learning_rate": 1.293415412199129e-05, + "loss": 0.9057, + "step": 34351 + }, + { + "epoch": 2.54, + "learning_rate": 1.2933772804355846e-05, + "loss": 1.0961, + "step": 34352 + }, + { + "epoch": 2.54, + "learning_rate": 1.293339148205281e-05, + "loss": 0.8448, + "step": 34353 + }, + { + "epoch": 2.54, + "learning_rate": 1.2933010155082792e-05, + "loss": 0.9378, + "step": 34354 + }, + { + "epoch": 2.54, + "learning_rate": 1.2932628823446395e-05, + "loss": 1.0468, + "step": 34355 + }, + { + "epoch": 2.54, + "learning_rate": 1.293224748714423e-05, + "loss": 1.0856, + "step": 34356 + }, + { + "epoch": 2.54, + "learning_rate": 1.29318661461769e-05, + "loss": 0.9226, + "step": 34357 + }, + { + "epoch": 2.54, + "learning_rate": 1.293148480054501e-05, + "loss": 1.0186, + "step": 34358 + }, + { + "epoch": 2.54, + "learning_rate": 1.2931103450249171e-05, + "loss": 0.9396, + "step": 34359 + }, + { + "epoch": 2.54, + "learning_rate": 1.2930722095289988e-05, + "loss": 0.9074, + "step": 34360 + }, + { + "epoch": 2.54, + "learning_rate": 1.2930340735668072e-05, + "loss": 0.9304, + "step": 34361 + }, + { + "epoch": 2.54, + "learning_rate": 1.2929959371384023e-05, + "loss": 0.8888, + "step": 34362 + }, + { + "epoch": 2.54, + "learning_rate": 1.2929578002438453e-05, + "loss": 0.9958, + "step": 34363 + }, + { + "epoch": 2.54, + "learning_rate": 1.2929196628831962e-05, + "loss": 1.0295, + "step": 34364 + }, + { + "epoch": 2.54, + "learning_rate": 1.2928815250565163e-05, + "loss": 1.1497, + "step": 34365 + }, + { + "epoch": 2.54, + "learning_rate": 1.2928433867638662e-05, + "loss": 1.0798, + "step": 34366 + }, + { + "epoch": 2.54, + "learning_rate": 1.2928052480053061e-05, + "loss": 1.0245, + "step": 34367 + }, + { + "epoch": 2.54, + "learning_rate": 1.2927671087808972e-05, + "loss": 1.0182, + "step": 34368 + }, + { + "epoch": 2.54, + "learning_rate": 1.2927289690907003e-05, + "loss": 0.8671, + "step": 34369 + }, + { + "epoch": 2.54, + "learning_rate": 1.2926908289347754e-05, + "loss": 1.0013, + "step": 34370 + }, + { + "epoch": 2.54, + "learning_rate": 1.292652688313184e-05, + "loss": 0.9581, + "step": 34371 + }, + { + "epoch": 2.54, + "learning_rate": 1.2926145472259863e-05, + "loss": 1.043, + "step": 34372 + }, + { + "epoch": 2.54, + "learning_rate": 1.2925764056732426e-05, + "loss": 1.072, + "step": 34373 + }, + { + "epoch": 2.54, + "learning_rate": 1.2925382636550146e-05, + "loss": 1.0618, + "step": 34374 + }, + { + "epoch": 2.54, + "learning_rate": 1.2925001211713622e-05, + "loss": 1.006, + "step": 34375 + }, + { + "epoch": 2.54, + "learning_rate": 1.292461978222346e-05, + "loss": 1.0351, + "step": 34376 + }, + { + "epoch": 2.54, + "learning_rate": 1.2924238348080271e-05, + "loss": 1.0505, + "step": 34377 + }, + { + "epoch": 2.54, + "learning_rate": 1.2923856909284663e-05, + "loss": 0.9442, + "step": 34378 + }, + { + "epoch": 2.54, + "learning_rate": 1.2923475465837239e-05, + "loss": 0.9635, + "step": 34379 + }, + { + "epoch": 2.54, + "learning_rate": 1.2923094017738607e-05, + "loss": 1.008, + "step": 34380 + }, + { + "epoch": 2.54, + "learning_rate": 1.2922712564989375e-05, + "loss": 0.9336, + "step": 34381 + }, + { + "epoch": 2.54, + "learning_rate": 1.2922331107590148e-05, + "loss": 1.1495, + "step": 34382 + }, + { + "epoch": 2.54, + "learning_rate": 1.2921949645541534e-05, + "loss": 1.0486, + "step": 34383 + }, + { + "epoch": 2.54, + "learning_rate": 1.2921568178844142e-05, + "loss": 1.0182, + "step": 34384 + }, + { + "epoch": 2.54, + "learning_rate": 1.2921186707498575e-05, + "loss": 0.9958, + "step": 34385 + }, + { + "epoch": 2.54, + "learning_rate": 1.2920805231505442e-05, + "loss": 0.9741, + "step": 34386 + }, + { + "epoch": 2.54, + "learning_rate": 1.292042375086535e-05, + "loss": 0.9831, + "step": 34387 + }, + { + "epoch": 2.54, + "learning_rate": 1.2920042265578905e-05, + "loss": 0.8993, + "step": 34388 + }, + { + "epoch": 2.54, + "learning_rate": 1.2919660775646715e-05, + "loss": 1.0756, + "step": 34389 + }, + { + "epoch": 2.54, + "learning_rate": 1.2919279281069387e-05, + "loss": 0.976, + "step": 34390 + }, + { + "epoch": 2.54, + "learning_rate": 1.2918897781847529e-05, + "loss": 1.0355, + "step": 34391 + }, + { + "epoch": 2.54, + "learning_rate": 1.2918516277981743e-05, + "loss": 1.05, + "step": 34392 + }, + { + "epoch": 2.54, + "learning_rate": 1.2918134769472641e-05, + "loss": 1.1308, + "step": 34393 + }, + { + "epoch": 2.54, + "learning_rate": 1.2917753256320828e-05, + "loss": 1.1427, + "step": 34394 + }, + { + "epoch": 2.54, + "learning_rate": 1.2917371738526909e-05, + "loss": 1.0989, + "step": 34395 + }, + { + "epoch": 2.54, + "learning_rate": 1.2916990216091498e-05, + "loss": 1.02, + "step": 34396 + }, + { + "epoch": 2.54, + "learning_rate": 1.2916608689015193e-05, + "loss": 0.93, + "step": 34397 + }, + { + "epoch": 2.54, + "learning_rate": 1.2916227157298608e-05, + "loss": 0.9847, + "step": 34398 + }, + { + "epoch": 2.54, + "learning_rate": 1.2915845620942346e-05, + "loss": 1.1152, + "step": 34399 + }, + { + "epoch": 2.54, + "learning_rate": 1.291546407994702e-05, + "loss": 0.91, + "step": 34400 + }, + { + "epoch": 2.54, + "learning_rate": 1.2915082534313227e-05, + "loss": 0.9205, + "step": 34401 + }, + { + "epoch": 2.54, + "learning_rate": 1.2914700984041582e-05, + "loss": 1.0072, + "step": 34402 + }, + { + "epoch": 2.54, + "learning_rate": 1.2914319429132686e-05, + "loss": 1.099, + "step": 34403 + }, + { + "epoch": 2.54, + "learning_rate": 1.2913937869587152e-05, + "loss": 1.0327, + "step": 34404 + }, + { + "epoch": 2.54, + "learning_rate": 1.2913556305405584e-05, + "loss": 1.1541, + "step": 34405 + }, + { + "epoch": 2.54, + "learning_rate": 1.291317473658859e-05, + "loss": 1.0734, + "step": 34406 + }, + { + "epoch": 2.54, + "learning_rate": 1.2912793163136774e-05, + "loss": 1.0406, + "step": 34407 + }, + { + "epoch": 2.54, + "learning_rate": 1.2912411585050751e-05, + "loss": 0.9836, + "step": 34408 + }, + { + "epoch": 2.54, + "learning_rate": 1.2912030002331117e-05, + "loss": 0.9304, + "step": 34409 + }, + { + "epoch": 2.54, + "learning_rate": 1.291164841497849e-05, + "loss": 1.0297, + "step": 34410 + }, + { + "epoch": 2.54, + "learning_rate": 1.2911266822993471e-05, + "loss": 1.1115, + "step": 34411 + }, + { + "epoch": 2.54, + "learning_rate": 1.2910885226376668e-05, + "loss": 1.09, + "step": 34412 + }, + { + "epoch": 2.54, + "learning_rate": 1.2910503625128685e-05, + "loss": 0.9889, + "step": 34413 + }, + { + "epoch": 2.54, + "learning_rate": 1.2910122019250135e-05, + "loss": 1.0502, + "step": 34414 + }, + { + "epoch": 2.54, + "learning_rate": 1.2909740408741623e-05, + "loss": 1.0236, + "step": 34415 + }, + { + "epoch": 2.54, + "learning_rate": 1.2909358793603754e-05, + "loss": 0.9793, + "step": 34416 + }, + { + "epoch": 2.54, + "learning_rate": 1.2908977173837136e-05, + "loss": 0.8516, + "step": 34417 + }, + { + "epoch": 2.54, + "learning_rate": 1.290859554944238e-05, + "loss": 1.1171, + "step": 34418 + }, + { + "epoch": 2.54, + "learning_rate": 1.290821392042009e-05, + "loss": 0.9946, + "step": 34419 + }, + { + "epoch": 2.54, + "learning_rate": 1.2907832286770874e-05, + "loss": 0.9169, + "step": 34420 + }, + { + "epoch": 2.54, + "learning_rate": 1.2907450648495334e-05, + "loss": 1.0152, + "step": 34421 + }, + { + "epoch": 2.54, + "learning_rate": 1.2907069005594086e-05, + "loss": 1.0177, + "step": 34422 + }, + { + "epoch": 2.54, + "learning_rate": 1.290668735806773e-05, + "loss": 0.9746, + "step": 34423 + }, + { + "epoch": 2.54, + "learning_rate": 1.2906305705916874e-05, + "loss": 0.8342, + "step": 34424 + }, + { + "epoch": 2.54, + "learning_rate": 1.2905924049142131e-05, + "loss": 1.1091, + "step": 34425 + }, + { + "epoch": 2.54, + "learning_rate": 1.2905542387744104e-05, + "loss": 0.9776, + "step": 34426 + }, + { + "epoch": 2.54, + "learning_rate": 1.29051607217234e-05, + "loss": 1.045, + "step": 34427 + }, + { + "epoch": 2.54, + "learning_rate": 1.2904779051080627e-05, + "loss": 0.9846, + "step": 34428 + }, + { + "epoch": 2.54, + "learning_rate": 1.2904397375816392e-05, + "loss": 1.0478, + "step": 34429 + }, + { + "epoch": 2.54, + "learning_rate": 1.2904015695931302e-05, + "loss": 0.9483, + "step": 34430 + }, + { + "epoch": 2.54, + "learning_rate": 1.2903634011425967e-05, + "loss": 0.9693, + "step": 34431 + }, + { + "epoch": 2.54, + "learning_rate": 1.2903252322300987e-05, + "loss": 0.8118, + "step": 34432 + }, + { + "epoch": 2.54, + "learning_rate": 1.2902870628556978e-05, + "loss": 0.9481, + "step": 34433 + }, + { + "epoch": 2.54, + "learning_rate": 1.2902488930194541e-05, + "loss": 0.9176, + "step": 34434 + }, + { + "epoch": 2.54, + "learning_rate": 1.2902107227214286e-05, + "loss": 0.9845, + "step": 34435 + }, + { + "epoch": 2.54, + "learning_rate": 1.2901725519616824e-05, + "loss": 1.0422, + "step": 34436 + }, + { + "epoch": 2.54, + "learning_rate": 1.2901343807402753e-05, + "loss": 0.9778, + "step": 34437 + }, + { + "epoch": 2.54, + "learning_rate": 1.290096209057269e-05, + "loss": 0.996, + "step": 34438 + }, + { + "epoch": 2.54, + "learning_rate": 1.2900580369127236e-05, + "loss": 0.9302, + "step": 34439 + }, + { + "epoch": 2.54, + "learning_rate": 1.2900198643066997e-05, + "loss": 1.0695, + "step": 34440 + }, + { + "epoch": 2.54, + "learning_rate": 1.2899816912392586e-05, + "loss": 0.9963, + "step": 34441 + }, + { + "epoch": 2.54, + "learning_rate": 1.2899435177104607e-05, + "loss": 1.0711, + "step": 34442 + }, + { + "epoch": 2.54, + "learning_rate": 1.289905343720367e-05, + "loss": 0.9987, + "step": 34443 + }, + { + "epoch": 2.55, + "learning_rate": 1.289867169269038e-05, + "loss": 1.0228, + "step": 34444 + }, + { + "epoch": 2.55, + "learning_rate": 1.2898289943565343e-05, + "loss": 0.9978, + "step": 34445 + }, + { + "epoch": 2.55, + "learning_rate": 1.2897908189829166e-05, + "loss": 1.0229, + "step": 34446 + }, + { + "epoch": 2.55, + "learning_rate": 1.2897526431482467e-05, + "loss": 0.8618, + "step": 34447 + }, + { + "epoch": 2.55, + "learning_rate": 1.2897144668525838e-05, + "loss": 1.1216, + "step": 34448 + }, + { + "epoch": 2.55, + "learning_rate": 1.2896762900959895e-05, + "loss": 1.0543, + "step": 34449 + }, + { + "epoch": 2.55, + "learning_rate": 1.2896381128785244e-05, + "loss": 0.9646, + "step": 34450 + }, + { + "epoch": 2.55, + "learning_rate": 1.2895999352002494e-05, + "loss": 1.0002, + "step": 34451 + }, + { + "epoch": 2.55, + "learning_rate": 1.2895617570612245e-05, + "loss": 0.9742, + "step": 34452 + }, + { + "epoch": 2.55, + "learning_rate": 1.2895235784615114e-05, + "loss": 0.9466, + "step": 34453 + }, + { + "epoch": 2.55, + "learning_rate": 1.2894853994011702e-05, + "loss": 1.0599, + "step": 34454 + }, + { + "epoch": 2.55, + "learning_rate": 1.289447219880262e-05, + "loss": 1.0286, + "step": 34455 + }, + { + "epoch": 2.55, + "learning_rate": 1.2894090398988476e-05, + "loss": 1.062, + "step": 34456 + }, + { + "epoch": 2.55, + "learning_rate": 1.2893708594569873e-05, + "loss": 0.9249, + "step": 34457 + }, + { + "epoch": 2.55, + "learning_rate": 1.2893326785547422e-05, + "loss": 0.9735, + "step": 34458 + }, + { + "epoch": 2.55, + "learning_rate": 1.289294497192173e-05, + "loss": 0.9474, + "step": 34459 + }, + { + "epoch": 2.55, + "learning_rate": 1.2892563153693403e-05, + "loss": 1.0406, + "step": 34460 + }, + { + "epoch": 2.55, + "learning_rate": 1.289218133086305e-05, + "loss": 0.9843, + "step": 34461 + }, + { + "epoch": 2.55, + "learning_rate": 1.2891799503431277e-05, + "loss": 1.0889, + "step": 34462 + }, + { + "epoch": 2.55, + "learning_rate": 1.2891417671398694e-05, + "loss": 0.9317, + "step": 34463 + }, + { + "epoch": 2.55, + "learning_rate": 1.2891035834765904e-05, + "loss": 1.0613, + "step": 34464 + }, + { + "epoch": 2.55, + "learning_rate": 1.289065399353352e-05, + "loss": 1.0667, + "step": 34465 + }, + { + "epoch": 2.55, + "learning_rate": 1.2890272147702145e-05, + "loss": 0.9404, + "step": 34466 + }, + { + "epoch": 2.55, + "learning_rate": 1.2889890297272392e-05, + "loss": 0.9543, + "step": 34467 + }, + { + "epoch": 2.55, + "learning_rate": 1.288950844224486e-05, + "loss": 1.0516, + "step": 34468 + }, + { + "epoch": 2.55, + "learning_rate": 1.2889126582620164e-05, + "loss": 0.8843, + "step": 34469 + }, + { + "epoch": 2.55, + "learning_rate": 1.2888744718398908e-05, + "loss": 1.0391, + "step": 34470 + }, + { + "epoch": 2.55, + "learning_rate": 1.2888362849581704e-05, + "loss": 0.9217, + "step": 34471 + }, + { + "epoch": 2.55, + "learning_rate": 1.288798097616915e-05, + "loss": 0.9898, + "step": 34472 + }, + { + "epoch": 2.55, + "learning_rate": 1.2887599098161863e-05, + "loss": 0.9677, + "step": 34473 + }, + { + "epoch": 2.55, + "learning_rate": 1.2887217215560446e-05, + "loss": 1.0025, + "step": 34474 + }, + { + "epoch": 2.55, + "learning_rate": 1.288683532836551e-05, + "loss": 1.0031, + "step": 34475 + }, + { + "epoch": 2.55, + "learning_rate": 1.288645343657766e-05, + "loss": 1.0966, + "step": 34476 + }, + { + "epoch": 2.55, + "learning_rate": 1.2886071540197503e-05, + "loss": 1.0521, + "step": 34477 + }, + { + "epoch": 2.55, + "learning_rate": 1.2885689639225647e-05, + "loss": 0.9923, + "step": 34478 + }, + { + "epoch": 2.55, + "learning_rate": 1.28853077336627e-05, + "loss": 0.9505, + "step": 34479 + }, + { + "epoch": 2.55, + "learning_rate": 1.288492582350927e-05, + "loss": 1.0096, + "step": 34480 + }, + { + "epoch": 2.55, + "learning_rate": 1.2884543908765963e-05, + "loss": 0.9826, + "step": 34481 + }, + { + "epoch": 2.55, + "learning_rate": 1.288416198943339e-05, + "loss": 1.0236, + "step": 34482 + }, + { + "epoch": 2.55, + "learning_rate": 1.2883780065512154e-05, + "loss": 0.9775, + "step": 34483 + }, + { + "epoch": 2.55, + "learning_rate": 1.2883398137002869e-05, + "loss": 0.9791, + "step": 34484 + }, + { + "epoch": 2.55, + "learning_rate": 1.2883016203906135e-05, + "loss": 1.0878, + "step": 34485 + }, + { + "epoch": 2.55, + "learning_rate": 1.2882634266222566e-05, + "loss": 1.0669, + "step": 34486 + }, + { + "epoch": 2.55, + "learning_rate": 1.2882252323952766e-05, + "loss": 0.9417, + "step": 34487 + }, + { + "epoch": 2.55, + "learning_rate": 1.2881870377097346e-05, + "loss": 0.9626, + "step": 34488 + }, + { + "epoch": 2.55, + "learning_rate": 1.2881488425656908e-05, + "loss": 1.0612, + "step": 34489 + }, + { + "epoch": 2.55, + "learning_rate": 1.2881106469632063e-05, + "loss": 0.9874, + "step": 34490 + }, + { + "epoch": 2.55, + "learning_rate": 1.288072450902342e-05, + "loss": 0.9791, + "step": 34491 + }, + { + "epoch": 2.55, + "learning_rate": 1.2880342543831587e-05, + "loss": 1.1383, + "step": 34492 + }, + { + "epoch": 2.55, + "learning_rate": 1.2879960574057169e-05, + "loss": 1.004, + "step": 34493 + }, + { + "epoch": 2.55, + "learning_rate": 1.2879578599700778e-05, + "loss": 1.0409, + "step": 34494 + }, + { + "epoch": 2.55, + "learning_rate": 1.2879196620763014e-05, + "loss": 1.0189, + "step": 34495 + }, + { + "epoch": 2.55, + "learning_rate": 1.2878814637244492e-05, + "loss": 1.0239, + "step": 34496 + }, + { + "epoch": 2.55, + "learning_rate": 1.2878432649145816e-05, + "loss": 1.0182, + "step": 34497 + }, + { + "epoch": 2.55, + "learning_rate": 1.2878050656467596e-05, + "loss": 0.8812, + "step": 34498 + }, + { + "epoch": 2.55, + "learning_rate": 1.2877668659210436e-05, + "loss": 1.0128, + "step": 34499 + }, + { + "epoch": 2.55, + "learning_rate": 1.287728665737495e-05, + "loss": 0.9622, + "step": 34500 + }, + { + "epoch": 2.55, + "learning_rate": 1.2876904650961736e-05, + "loss": 1.0215, + "step": 34501 + }, + { + "epoch": 2.55, + "learning_rate": 1.2876522639971415e-05, + "loss": 0.938, + "step": 34502 + }, + { + "epoch": 2.55, + "learning_rate": 1.2876140624404586e-05, + "loss": 0.8714, + "step": 34503 + }, + { + "epoch": 2.55, + "learning_rate": 1.2875758604261856e-05, + "loss": 1.0479, + "step": 34504 + }, + { + "epoch": 2.55, + "learning_rate": 1.2875376579543837e-05, + "loss": 1.0085, + "step": 34505 + }, + { + "epoch": 2.55, + "learning_rate": 1.2874994550251134e-05, + "loss": 1.0763, + "step": 34506 + }, + { + "epoch": 2.55, + "learning_rate": 1.2874612516384354e-05, + "loss": 1.0414, + "step": 34507 + }, + { + "epoch": 2.55, + "learning_rate": 1.2874230477944112e-05, + "loss": 1.0764, + "step": 34508 + }, + { + "epoch": 2.55, + "learning_rate": 1.2873848434931005e-05, + "loss": 0.8974, + "step": 34509 + }, + { + "epoch": 2.55, + "learning_rate": 1.2873466387345648e-05, + "loss": 1.0762, + "step": 34510 + }, + { + "epoch": 2.55, + "learning_rate": 1.2873084335188649e-05, + "loss": 0.9938, + "step": 34511 + }, + { + "epoch": 2.55, + "learning_rate": 1.2872702278460614e-05, + "loss": 1.0983, + "step": 34512 + }, + { + "epoch": 2.55, + "learning_rate": 1.2872320217162146e-05, + "loss": 0.993, + "step": 34513 + }, + { + "epoch": 2.55, + "learning_rate": 1.2871938151293864e-05, + "loss": 0.9778, + "step": 34514 + }, + { + "epoch": 2.55, + "learning_rate": 1.2871556080856364e-05, + "loss": 1.0594, + "step": 34515 + }, + { + "epoch": 2.55, + "learning_rate": 1.2871174005850262e-05, + "loss": 1.0167, + "step": 34516 + }, + { + "epoch": 2.55, + "learning_rate": 1.2870791926276163e-05, + "loss": 0.9298, + "step": 34517 + }, + { + "epoch": 2.55, + "learning_rate": 1.2870409842134675e-05, + "loss": 1.1051, + "step": 34518 + }, + { + "epoch": 2.55, + "learning_rate": 1.2870027753426404e-05, + "loss": 0.9529, + "step": 34519 + }, + { + "epoch": 2.55, + "learning_rate": 1.2869645660151964e-05, + "loss": 1.0409, + "step": 34520 + }, + { + "epoch": 2.55, + "learning_rate": 1.2869263562311956e-05, + "loss": 1.021, + "step": 34521 + }, + { + "epoch": 2.55, + "learning_rate": 1.286888145990699e-05, + "loss": 0.9941, + "step": 34522 + }, + { + "epoch": 2.55, + "learning_rate": 1.2868499352937678e-05, + "loss": 0.9075, + "step": 34523 + }, + { + "epoch": 2.55, + "learning_rate": 1.2868117241404625e-05, + "loss": 1.0395, + "step": 34524 + }, + { + "epoch": 2.55, + "learning_rate": 1.2867735125308434e-05, + "loss": 1.0935, + "step": 34525 + }, + { + "epoch": 2.55, + "learning_rate": 1.2867353004649721e-05, + "loss": 1.0053, + "step": 34526 + }, + { + "epoch": 2.55, + "learning_rate": 1.2866970879429088e-05, + "loss": 0.9864, + "step": 34527 + }, + { + "epoch": 2.55, + "learning_rate": 1.2866588749647147e-05, + "loss": 0.9162, + "step": 34528 + }, + { + "epoch": 2.55, + "learning_rate": 1.2866206615304501e-05, + "loss": 0.9989, + "step": 34529 + }, + { + "epoch": 2.55, + "learning_rate": 1.2865824476401767e-05, + "loss": 0.9584, + "step": 34530 + }, + { + "epoch": 2.55, + "learning_rate": 1.2865442332939545e-05, + "loss": 1.0456, + "step": 34531 + }, + { + "epoch": 2.55, + "learning_rate": 1.2865060184918443e-05, + "loss": 0.9433, + "step": 34532 + }, + { + "epoch": 2.55, + "learning_rate": 1.2864678032339073e-05, + "loss": 0.9944, + "step": 34533 + }, + { + "epoch": 2.55, + "learning_rate": 1.286429587520204e-05, + "loss": 0.9932, + "step": 34534 + }, + { + "epoch": 2.55, + "learning_rate": 1.2863913713507954e-05, + "loss": 1.108, + "step": 34535 + }, + { + "epoch": 2.55, + "learning_rate": 1.2863531547257423e-05, + "loss": 0.972, + "step": 34536 + }, + { + "epoch": 2.55, + "learning_rate": 1.2863149376451054e-05, + "loss": 0.8973, + "step": 34537 + }, + { + "epoch": 2.55, + "learning_rate": 1.2862767201089455e-05, + "loss": 0.9636, + "step": 34538 + }, + { + "epoch": 2.55, + "learning_rate": 1.2862385021173233e-05, + "loss": 1.0447, + "step": 34539 + }, + { + "epoch": 2.55, + "learning_rate": 1.2862002836703e-05, + "loss": 0.9945, + "step": 34540 + }, + { + "epoch": 2.55, + "learning_rate": 1.2861620647679361e-05, + "loss": 0.9889, + "step": 34541 + }, + { + "epoch": 2.55, + "learning_rate": 1.2861238454102922e-05, + "loss": 0.913, + "step": 34542 + }, + { + "epoch": 2.55, + "learning_rate": 1.2860856255974297e-05, + "loss": 1.0409, + "step": 34543 + }, + { + "epoch": 2.55, + "learning_rate": 1.2860474053294087e-05, + "loss": 1.0298, + "step": 34544 + }, + { + "epoch": 2.55, + "learning_rate": 1.2860091846062908e-05, + "loss": 1.0388, + "step": 34545 + }, + { + "epoch": 2.55, + "learning_rate": 1.2859709634281359e-05, + "loss": 0.9252, + "step": 34546 + }, + { + "epoch": 2.55, + "learning_rate": 1.2859327417950056e-05, + "loss": 1.0682, + "step": 34547 + }, + { + "epoch": 2.55, + "learning_rate": 1.2858945197069601e-05, + "loss": 0.9467, + "step": 34548 + }, + { + "epoch": 2.55, + "learning_rate": 1.285856297164061e-05, + "loss": 1.0377, + "step": 34549 + }, + { + "epoch": 2.55, + "learning_rate": 1.2858180741663684e-05, + "loss": 0.9083, + "step": 34550 + }, + { + "epoch": 2.55, + "learning_rate": 1.2857798507139434e-05, + "loss": 0.9666, + "step": 34551 + }, + { + "epoch": 2.55, + "learning_rate": 1.2857416268068464e-05, + "loss": 0.9321, + "step": 34552 + }, + { + "epoch": 2.55, + "learning_rate": 1.2857034024451389e-05, + "loss": 1.0168, + "step": 34553 + }, + { + "epoch": 2.55, + "learning_rate": 1.2856651776288812e-05, + "loss": 1.0626, + "step": 34554 + }, + { + "epoch": 2.55, + "learning_rate": 1.2856269523581344e-05, + "loss": 0.9229, + "step": 34555 + }, + { + "epoch": 2.55, + "learning_rate": 1.285588726632959e-05, + "loss": 1.0101, + "step": 34556 + }, + { + "epoch": 2.55, + "learning_rate": 1.2855505004534163e-05, + "loss": 1.0334, + "step": 34557 + }, + { + "epoch": 2.55, + "learning_rate": 1.2855122738195666e-05, + "loss": 1.0049, + "step": 34558 + }, + { + "epoch": 2.55, + "learning_rate": 1.2854740467314715e-05, + "loss": 0.9186, + "step": 34559 + }, + { + "epoch": 2.55, + "learning_rate": 1.2854358191891908e-05, + "loss": 0.9076, + "step": 34560 + }, + { + "epoch": 2.55, + "learning_rate": 1.2853975911927859e-05, + "loss": 0.9445, + "step": 34561 + }, + { + "epoch": 2.55, + "learning_rate": 1.2853593627423173e-05, + "loss": 0.9774, + "step": 34562 + }, + { + "epoch": 2.55, + "learning_rate": 1.2853211338378464e-05, + "loss": 0.9597, + "step": 34563 + }, + { + "epoch": 2.55, + "learning_rate": 1.2852829044794332e-05, + "loss": 0.9593, + "step": 34564 + }, + { + "epoch": 2.55, + "learning_rate": 1.2852446746671395e-05, + "loss": 0.9556, + "step": 34565 + }, + { + "epoch": 2.55, + "learning_rate": 1.2852064444010251e-05, + "loss": 0.969, + "step": 34566 + }, + { + "epoch": 2.55, + "learning_rate": 1.2851682136811517e-05, + "loss": 1.049, + "step": 34567 + }, + { + "epoch": 2.55, + "learning_rate": 1.2851299825075794e-05, + "loss": 0.9861, + "step": 34568 + }, + { + "epoch": 2.55, + "learning_rate": 1.28509175088037e-05, + "loss": 0.9516, + "step": 34569 + }, + { + "epoch": 2.55, + "learning_rate": 1.2850535187995831e-05, + "loss": 1.0674, + "step": 34570 + }, + { + "epoch": 2.55, + "learning_rate": 1.2850152862652805e-05, + "loss": 0.9285, + "step": 34571 + }, + { + "epoch": 2.55, + "learning_rate": 1.2849770532775223e-05, + "loss": 0.9349, + "step": 34572 + }, + { + "epoch": 2.55, + "learning_rate": 1.28493881983637e-05, + "loss": 0.8795, + "step": 34573 + }, + { + "epoch": 2.55, + "learning_rate": 1.284900585941884e-05, + "loss": 1.0442, + "step": 34574 + }, + { + "epoch": 2.55, + "learning_rate": 1.2848623515941252e-05, + "loss": 0.9921, + "step": 34575 + }, + { + "epoch": 2.55, + "learning_rate": 1.2848241167931543e-05, + "loss": 1.0923, + "step": 34576 + }, + { + "epoch": 2.55, + "learning_rate": 1.2847858815390329e-05, + "loss": 1.0786, + "step": 34577 + }, + { + "epoch": 2.55, + "learning_rate": 1.284747645831821e-05, + "loss": 1.0385, + "step": 34578 + }, + { + "epoch": 2.56, + "learning_rate": 1.2847094096715794e-05, + "loss": 0.9783, + "step": 34579 + }, + { + "epoch": 2.56, + "learning_rate": 1.2846711730583691e-05, + "loss": 1.0409, + "step": 34580 + }, + { + "epoch": 2.56, + "learning_rate": 1.284632935992251e-05, + "loss": 0.9666, + "step": 34581 + }, + { + "epoch": 2.56, + "learning_rate": 1.2845946984732867e-05, + "loss": 0.9494, + "step": 34582 + }, + { + "epoch": 2.56, + "learning_rate": 1.2845564605015357e-05, + "loss": 1.013, + "step": 34583 + }, + { + "epoch": 2.56, + "learning_rate": 1.2845182220770596e-05, + "loss": 0.9788, + "step": 34584 + }, + { + "epoch": 2.56, + "learning_rate": 1.284479983199919e-05, + "loss": 1.0029, + "step": 34585 + }, + { + "epoch": 2.56, + "learning_rate": 1.284441743870175e-05, + "loss": 1.0764, + "step": 34586 + }, + { + "epoch": 2.56, + "learning_rate": 1.2844035040878882e-05, + "loss": 1.0232, + "step": 34587 + }, + { + "epoch": 2.56, + "learning_rate": 1.2843652638531194e-05, + "loss": 1.0097, + "step": 34588 + }, + { + "epoch": 2.56, + "learning_rate": 1.2843270231659295e-05, + "loss": 0.9444, + "step": 34589 + }, + { + "epoch": 2.56, + "learning_rate": 1.2842887820263797e-05, + "loss": 1.1189, + "step": 34590 + }, + { + "epoch": 2.56, + "learning_rate": 1.28425054043453e-05, + "loss": 0.9928, + "step": 34591 + }, + { + "epoch": 2.56, + "learning_rate": 1.2842122983904422e-05, + "loss": 0.9663, + "step": 34592 + }, + { + "epoch": 2.56, + "learning_rate": 1.2841740558941762e-05, + "loss": 0.9944, + "step": 34593 + }, + { + "epoch": 2.56, + "learning_rate": 1.284135812945794e-05, + "loss": 1.0442, + "step": 34594 + }, + { + "epoch": 2.56, + "learning_rate": 1.284097569545355e-05, + "loss": 0.9957, + "step": 34595 + }, + { + "epoch": 2.56, + "learning_rate": 1.2840593256929214e-05, + "loss": 1.0134, + "step": 34596 + }, + { + "epoch": 2.56, + "learning_rate": 1.2840210813885536e-05, + "loss": 1.1032, + "step": 34597 + }, + { + "epoch": 2.56, + "learning_rate": 1.2839828366323122e-05, + "loss": 1.0632, + "step": 34598 + }, + { + "epoch": 2.56, + "learning_rate": 1.2839445914242578e-05, + "loss": 1.0139, + "step": 34599 + }, + { + "epoch": 2.56, + "learning_rate": 1.2839063457644521e-05, + "loss": 1.0207, + "step": 34600 + }, + { + "epoch": 2.56, + "learning_rate": 1.283868099652955e-05, + "loss": 1.0747, + "step": 34601 + }, + { + "epoch": 2.56, + "learning_rate": 1.2838298530898281e-05, + "loss": 0.9433, + "step": 34602 + }, + { + "epoch": 2.56, + "learning_rate": 1.2837916060751319e-05, + "loss": 0.8233, + "step": 34603 + }, + { + "epoch": 2.56, + "learning_rate": 1.2837533586089275e-05, + "loss": 1.0159, + "step": 34604 + }, + { + "epoch": 2.56, + "learning_rate": 1.283715110691275e-05, + "loss": 0.925, + "step": 34605 + }, + { + "epoch": 2.56, + "learning_rate": 1.2836768623222366e-05, + "loss": 1.0704, + "step": 34606 + }, + { + "epoch": 2.56, + "learning_rate": 1.2836386135018718e-05, + "loss": 1.0792, + "step": 34607 + }, + { + "epoch": 2.56, + "learning_rate": 1.2836003642302425e-05, + "loss": 0.9936, + "step": 34608 + }, + { + "epoch": 2.56, + "learning_rate": 1.2835621145074084e-05, + "loss": 0.9487, + "step": 34609 + }, + { + "epoch": 2.56, + "learning_rate": 1.2835238643334316e-05, + "loss": 1.0862, + "step": 34610 + }, + { + "epoch": 2.56, + "learning_rate": 1.283485613708372e-05, + "loss": 0.9534, + "step": 34611 + }, + { + "epoch": 2.56, + "learning_rate": 1.2834473626322912e-05, + "loss": 0.9834, + "step": 34612 + }, + { + "epoch": 2.56, + "learning_rate": 1.2834091111052496e-05, + "loss": 0.9593, + "step": 34613 + }, + { + "epoch": 2.56, + "learning_rate": 1.2833708591273083e-05, + "loss": 1.0425, + "step": 34614 + }, + { + "epoch": 2.56, + "learning_rate": 1.2833326066985277e-05, + "loss": 1.0131, + "step": 34615 + }, + { + "epoch": 2.56, + "learning_rate": 1.2832943538189693e-05, + "loss": 0.9707, + "step": 34616 + }, + { + "epoch": 2.56, + "learning_rate": 1.2832561004886933e-05, + "loss": 1.0265, + "step": 34617 + }, + { + "epoch": 2.56, + "learning_rate": 1.283217846707761e-05, + "loss": 0.9818, + "step": 34618 + }, + { + "epoch": 2.56, + "learning_rate": 1.2831795924762331e-05, + "loss": 0.9188, + "step": 34619 + }, + { + "epoch": 2.56, + "learning_rate": 1.2831413377941706e-05, + "loss": 1.0194, + "step": 34620 + }, + { + "epoch": 2.56, + "learning_rate": 1.2831030826616341e-05, + "loss": 0.918, + "step": 34621 + }, + { + "epoch": 2.56, + "learning_rate": 1.2830648270786852e-05, + "loss": 0.9825, + "step": 34622 + }, + { + "epoch": 2.56, + "learning_rate": 1.2830265710453835e-05, + "loss": 1.0134, + "step": 34623 + }, + { + "epoch": 2.56, + "learning_rate": 1.2829883145617912e-05, + "loss": 1.0622, + "step": 34624 + }, + { + "epoch": 2.56, + "learning_rate": 1.2829500576279684e-05, + "loss": 1.0225, + "step": 34625 + }, + { + "epoch": 2.56, + "learning_rate": 1.2829118002439758e-05, + "loss": 0.9979, + "step": 34626 + }, + { + "epoch": 2.56, + "learning_rate": 1.2828735424098748e-05, + "loss": 1.1084, + "step": 34627 + }, + { + "epoch": 2.56, + "learning_rate": 1.282835284125726e-05, + "loss": 0.9256, + "step": 34628 + }, + { + "epoch": 2.56, + "learning_rate": 1.2827970253915903e-05, + "loss": 0.9615, + "step": 34629 + }, + { + "epoch": 2.56, + "learning_rate": 1.2827587662075287e-05, + "loss": 0.9863, + "step": 34630 + }, + { + "epoch": 2.56, + "learning_rate": 1.2827205065736016e-05, + "loss": 0.9956, + "step": 34631 + }, + { + "epoch": 2.56, + "learning_rate": 1.2826822464898704e-05, + "loss": 0.8637, + "step": 34632 + }, + { + "epoch": 2.56, + "learning_rate": 1.282643985956396e-05, + "loss": 1.0458, + "step": 34633 + }, + { + "epoch": 2.56, + "learning_rate": 1.2826057249732389e-05, + "loss": 0.9641, + "step": 34634 + }, + { + "epoch": 2.56, + "learning_rate": 1.2825674635404602e-05, + "loss": 1.0045, + "step": 34635 + }, + { + "epoch": 2.56, + "learning_rate": 1.2825292016581206e-05, + "loss": 1.1194, + "step": 34636 + }, + { + "epoch": 2.56, + "learning_rate": 1.2824909393262813e-05, + "loss": 0.9679, + "step": 34637 + }, + { + "epoch": 2.56, + "learning_rate": 1.2824526765450025e-05, + "loss": 1.0584, + "step": 34638 + }, + { + "epoch": 2.56, + "learning_rate": 1.2824144133143459e-05, + "loss": 0.9637, + "step": 34639 + }, + { + "epoch": 2.56, + "learning_rate": 1.2823761496343718e-05, + "loss": 1.0861, + "step": 34640 + }, + { + "epoch": 2.56, + "learning_rate": 1.2823378855051413e-05, + "loss": 1.075, + "step": 34641 + }, + { + "epoch": 2.56, + "learning_rate": 1.2822996209267155e-05, + "loss": 1.0472, + "step": 34642 + }, + { + "epoch": 2.56, + "learning_rate": 1.2822613558991552e-05, + "loss": 0.9608, + "step": 34643 + }, + { + "epoch": 2.56, + "learning_rate": 1.2822230904225208e-05, + "loss": 0.8723, + "step": 34644 + }, + { + "epoch": 2.56, + "learning_rate": 1.2821848244968736e-05, + "loss": 1.0665, + "step": 34645 + }, + { + "epoch": 2.56, + "learning_rate": 1.2821465581222741e-05, + "loss": 1.065, + "step": 34646 + }, + { + "epoch": 2.56, + "learning_rate": 1.282108291298784e-05, + "loss": 0.9225, + "step": 34647 + }, + { + "epoch": 2.56, + "learning_rate": 1.282070024026463e-05, + "loss": 1.0916, + "step": 34648 + }, + { + "epoch": 2.56, + "learning_rate": 1.282031756305373e-05, + "loss": 0.9417, + "step": 34649 + }, + { + "epoch": 2.56, + "learning_rate": 1.2819934881355744e-05, + "loss": 0.988, + "step": 34650 + }, + { + "epoch": 2.56, + "learning_rate": 1.2819552195171282e-05, + "loss": 1.0926, + "step": 34651 + }, + { + "epoch": 2.56, + "learning_rate": 1.2819169504500954e-05, + "loss": 0.9374, + "step": 34652 + }, + { + "epoch": 2.56, + "learning_rate": 1.281878680934537e-05, + "loss": 1.0108, + "step": 34653 + }, + { + "epoch": 2.56, + "learning_rate": 1.281840410970513e-05, + "loss": 0.9797, + "step": 34654 + }, + { + "epoch": 2.56, + "learning_rate": 1.2818021405580857e-05, + "loss": 1.1299, + "step": 34655 + }, + { + "epoch": 2.56, + "learning_rate": 1.2817638696973144e-05, + "loss": 1.0861, + "step": 34656 + }, + { + "epoch": 2.56, + "learning_rate": 1.2817255983882614e-05, + "loss": 1.1153, + "step": 34657 + }, + { + "epoch": 2.56, + "learning_rate": 1.2816873266309868e-05, + "loss": 0.9978, + "step": 34658 + }, + { + "epoch": 2.56, + "learning_rate": 1.2816490544255517e-05, + "loss": 0.9829, + "step": 34659 + }, + { + "epoch": 2.56, + "learning_rate": 1.2816107817720171e-05, + "loss": 0.9452, + "step": 34660 + }, + { + "epoch": 2.56, + "learning_rate": 1.2815725086704439e-05, + "loss": 0.9384, + "step": 34661 + }, + { + "epoch": 2.56, + "learning_rate": 1.2815342351208926e-05, + "loss": 0.961, + "step": 34662 + }, + { + "epoch": 2.56, + "learning_rate": 1.2814959611234246e-05, + "loss": 1.0075, + "step": 34663 + }, + { + "epoch": 2.56, + "learning_rate": 1.2814576866781e-05, + "loss": 1.0419, + "step": 34664 + }, + { + "epoch": 2.56, + "learning_rate": 1.281419411784981e-05, + "loss": 1.0134, + "step": 34665 + }, + { + "epoch": 2.56, + "learning_rate": 1.2813811364441271e-05, + "loss": 1.0761, + "step": 34666 + }, + { + "epoch": 2.56, + "learning_rate": 1.2813428606556002e-05, + "loss": 0.9713, + "step": 34667 + }, + { + "epoch": 2.56, + "learning_rate": 1.2813045844194605e-05, + "loss": 1.0187, + "step": 34668 + }, + { + "epoch": 2.56, + "learning_rate": 1.2812663077357697e-05, + "loss": 0.979, + "step": 34669 + }, + { + "epoch": 2.56, + "learning_rate": 1.2812280306045878e-05, + "loss": 1.0035, + "step": 34670 + }, + { + "epoch": 2.56, + "learning_rate": 1.2811897530259764e-05, + "loss": 0.9569, + "step": 34671 + }, + { + "epoch": 2.56, + "learning_rate": 1.2811514749999962e-05, + "loss": 0.9961, + "step": 34672 + }, + { + "epoch": 2.56, + "learning_rate": 1.2811131965267078e-05, + "loss": 1.0789, + "step": 34673 + }, + { + "epoch": 2.56, + "learning_rate": 1.2810749176061724e-05, + "loss": 0.9498, + "step": 34674 + }, + { + "epoch": 2.56, + "learning_rate": 1.2810366382384509e-05, + "loss": 0.907, + "step": 34675 + }, + { + "epoch": 2.56, + "learning_rate": 1.2809983584236042e-05, + "loss": 1.0216, + "step": 34676 + }, + { + "epoch": 2.56, + "learning_rate": 1.2809600781616928e-05, + "loss": 1.0336, + "step": 34677 + }, + { + "epoch": 2.56, + "learning_rate": 1.2809217974527781e-05, + "loss": 0.9603, + "step": 34678 + }, + { + "epoch": 2.56, + "learning_rate": 1.280883516296921e-05, + "loss": 0.9348, + "step": 34679 + }, + { + "epoch": 2.56, + "learning_rate": 1.2808452346941817e-05, + "loss": 1.0825, + "step": 34680 + }, + { + "epoch": 2.56, + "learning_rate": 1.2808069526446224e-05, + "loss": 1.0688, + "step": 34681 + }, + { + "epoch": 2.56, + "learning_rate": 1.2807686701483028e-05, + "loss": 0.9584, + "step": 34682 + }, + { + "epoch": 2.56, + "learning_rate": 1.2807303872052845e-05, + "loss": 1.0095, + "step": 34683 + }, + { + "epoch": 2.56, + "learning_rate": 1.280692103815628e-05, + "loss": 1.0617, + "step": 34684 + }, + { + "epoch": 2.56, + "learning_rate": 1.2806538199793943e-05, + "loss": 1.0696, + "step": 34685 + }, + { + "epoch": 2.56, + "learning_rate": 1.2806155356966445e-05, + "loss": 1.0128, + "step": 34686 + }, + { + "epoch": 2.56, + "learning_rate": 1.2805772509674395e-05, + "loss": 0.9314, + "step": 34687 + }, + { + "epoch": 2.56, + "learning_rate": 1.2805389657918401e-05, + "loss": 0.9836, + "step": 34688 + }, + { + "epoch": 2.56, + "learning_rate": 1.280500680169907e-05, + "loss": 0.9999, + "step": 34689 + }, + { + "epoch": 2.56, + "learning_rate": 1.2804623941017018e-05, + "loss": 0.9967, + "step": 34690 + }, + { + "epoch": 2.56, + "learning_rate": 1.2804241075872845e-05, + "loss": 0.9928, + "step": 34691 + }, + { + "epoch": 2.56, + "learning_rate": 1.2803858206267166e-05, + "loss": 1.1169, + "step": 34692 + }, + { + "epoch": 2.56, + "learning_rate": 1.2803475332200588e-05, + "loss": 0.9502, + "step": 34693 + }, + { + "epoch": 2.56, + "learning_rate": 1.2803092453673722e-05, + "loss": 0.9962, + "step": 34694 + }, + { + "epoch": 2.56, + "learning_rate": 1.2802709570687172e-05, + "loss": 0.991, + "step": 34695 + }, + { + "epoch": 2.56, + "learning_rate": 1.2802326683241557e-05, + "loss": 1.1059, + "step": 34696 + }, + { + "epoch": 2.56, + "learning_rate": 1.2801943791337476e-05, + "loss": 1.0432, + "step": 34697 + }, + { + "epoch": 2.56, + "learning_rate": 1.2801560894975546e-05, + "loss": 0.938, + "step": 34698 + }, + { + "epoch": 2.56, + "learning_rate": 1.280117799415637e-05, + "loss": 0.9825, + "step": 34699 + }, + { + "epoch": 2.56, + "learning_rate": 1.2800795088880564e-05, + "loss": 1.0404, + "step": 34700 + }, + { + "epoch": 2.56, + "learning_rate": 1.2800412179148729e-05, + "loss": 0.999, + "step": 34701 + }, + { + "epoch": 2.56, + "learning_rate": 1.280002926496148e-05, + "loss": 0.9415, + "step": 34702 + }, + { + "epoch": 2.56, + "learning_rate": 1.2799646346319423e-05, + "loss": 0.9444, + "step": 34703 + }, + { + "epoch": 2.56, + "learning_rate": 1.279926342322317e-05, + "loss": 1.0428, + "step": 34704 + }, + { + "epoch": 2.56, + "learning_rate": 1.2798880495673327e-05, + "loss": 1.1261, + "step": 34705 + }, + { + "epoch": 2.56, + "learning_rate": 1.2798497563670509e-05, + "loss": 0.959, + "step": 34706 + }, + { + "epoch": 2.56, + "learning_rate": 1.2798114627215316e-05, + "loss": 0.9568, + "step": 34707 + }, + { + "epoch": 2.56, + "learning_rate": 1.2797731686308369e-05, + "loss": 1.0965, + "step": 34708 + }, + { + "epoch": 2.56, + "learning_rate": 1.2797348740950266e-05, + "loss": 1.0821, + "step": 34709 + }, + { + "epoch": 2.56, + "learning_rate": 1.2796965791141624e-05, + "loss": 1.1235, + "step": 34710 + }, + { + "epoch": 2.56, + "learning_rate": 1.2796582836883048e-05, + "loss": 1.02, + "step": 34711 + }, + { + "epoch": 2.56, + "learning_rate": 1.2796199878175149e-05, + "loss": 1.0532, + "step": 34712 + }, + { + "epoch": 2.56, + "learning_rate": 1.2795816915018535e-05, + "loss": 1.0431, + "step": 34713 + }, + { + "epoch": 2.56, + "learning_rate": 1.2795433947413817e-05, + "loss": 1.0422, + "step": 34714 + }, + { + "epoch": 2.57, + "learning_rate": 1.2795050975361605e-05, + "loss": 1.0286, + "step": 34715 + }, + { + "epoch": 2.57, + "learning_rate": 1.2794667998862505e-05, + "loss": 0.9926, + "step": 34716 + }, + { + "epoch": 2.57, + "learning_rate": 1.2794285017917128e-05, + "loss": 1.045, + "step": 34717 + }, + { + "epoch": 2.57, + "learning_rate": 1.2793902032526089e-05, + "loss": 1.0608, + "step": 34718 + }, + { + "epoch": 2.57, + "learning_rate": 1.2793519042689985e-05, + "loss": 1.0288, + "step": 34719 + }, + { + "epoch": 2.57, + "learning_rate": 1.2793136048409435e-05, + "loss": 0.9083, + "step": 34720 + }, + { + "epoch": 2.57, + "learning_rate": 1.2792753049685046e-05, + "loss": 1.0936, + "step": 34721 + }, + { + "epoch": 2.57, + "learning_rate": 1.2792370046517424e-05, + "loss": 0.9574, + "step": 34722 + }, + { + "epoch": 2.57, + "learning_rate": 1.2791987038907183e-05, + "loss": 1.0429, + "step": 34723 + }, + { + "epoch": 2.57, + "learning_rate": 1.2791604026854931e-05, + "loss": 1.0435, + "step": 34724 + }, + { + "epoch": 2.57, + "learning_rate": 1.2791221010361276e-05, + "loss": 1.0185, + "step": 34725 + }, + { + "epoch": 2.57, + "learning_rate": 1.279083798942683e-05, + "loss": 1.0126, + "step": 34726 + }, + { + "epoch": 2.57, + "learning_rate": 1.2790454964052199e-05, + "loss": 0.9305, + "step": 34727 + }, + { + "epoch": 2.57, + "learning_rate": 1.2790071934237993e-05, + "loss": 0.9035, + "step": 34728 + }, + { + "epoch": 2.57, + "learning_rate": 1.2789688899984827e-05, + "loss": 0.9979, + "step": 34729 + }, + { + "epoch": 2.57, + "learning_rate": 1.27893058612933e-05, + "loss": 1.0156, + "step": 34730 + }, + { + "epoch": 2.57, + "learning_rate": 1.278892281816403e-05, + "loss": 1.0433, + "step": 34731 + }, + { + "epoch": 2.57, + "learning_rate": 1.2788539770597624e-05, + "loss": 1.0064, + "step": 34732 + }, + { + "epoch": 2.57, + "learning_rate": 1.2788156718594694e-05, + "loss": 0.9558, + "step": 34733 + }, + { + "epoch": 2.57, + "learning_rate": 1.2787773662155842e-05, + "loss": 1.0069, + "step": 34734 + }, + { + "epoch": 2.57, + "learning_rate": 1.2787390601281684e-05, + "loss": 1.0639, + "step": 34735 + }, + { + "epoch": 2.57, + "learning_rate": 1.2787007535972827e-05, + "loss": 1.1063, + "step": 34736 + }, + { + "epoch": 2.57, + "learning_rate": 1.2786624466229881e-05, + "loss": 1.166, + "step": 34737 + }, + { + "epoch": 2.57, + "learning_rate": 1.2786241392053455e-05, + "loss": 0.987, + "step": 34738 + }, + { + "epoch": 2.57, + "learning_rate": 1.2785858313444162e-05, + "loss": 1.0365, + "step": 34739 + }, + { + "epoch": 2.57, + "learning_rate": 1.2785475230402602e-05, + "loss": 0.9809, + "step": 34740 + }, + { + "epoch": 2.57, + "learning_rate": 1.2785092142929395e-05, + "loss": 0.9861, + "step": 34741 + }, + { + "epoch": 2.57, + "learning_rate": 1.2784709051025144e-05, + "loss": 1.0764, + "step": 34742 + }, + { + "epoch": 2.57, + "learning_rate": 1.2784325954690464e-05, + "loss": 1.0092, + "step": 34743 + }, + { + "epoch": 2.57, + "learning_rate": 1.2783942853925959e-05, + "loss": 1.1385, + "step": 34744 + }, + { + "epoch": 2.57, + "learning_rate": 1.2783559748732241e-05, + "loss": 0.9518, + "step": 34745 + }, + { + "epoch": 2.57, + "learning_rate": 1.278317663910992e-05, + "loss": 1.0183, + "step": 34746 + }, + { + "epoch": 2.57, + "learning_rate": 1.2782793525059605e-05, + "loss": 1.0262, + "step": 34747 + }, + { + "epoch": 2.57, + "learning_rate": 1.2782410406581904e-05, + "loss": 0.9811, + "step": 34748 + }, + { + "epoch": 2.57, + "learning_rate": 1.2782027283677429e-05, + "loss": 0.9757, + "step": 34749 + }, + { + "epoch": 2.57, + "learning_rate": 1.2781644156346787e-05, + "loss": 1.0302, + "step": 34750 + }, + { + "epoch": 2.57, + "learning_rate": 1.278126102459059e-05, + "loss": 0.9633, + "step": 34751 + }, + { + "epoch": 2.57, + "learning_rate": 1.2780877888409445e-05, + "loss": 0.9679, + "step": 34752 + }, + { + "epoch": 2.57, + "learning_rate": 1.2780494747803967e-05, + "loss": 1.0081, + "step": 34753 + }, + { + "epoch": 2.57, + "learning_rate": 1.2780111602774756e-05, + "loss": 1.0601, + "step": 34754 + }, + { + "epoch": 2.57, + "learning_rate": 1.2779728453322434e-05, + "loss": 0.9583, + "step": 34755 + }, + { + "epoch": 2.57, + "learning_rate": 1.2779345299447598e-05, + "loss": 0.9772, + "step": 34756 + }, + { + "epoch": 2.57, + "learning_rate": 1.2778962141150866e-05, + "loss": 0.9808, + "step": 34757 + }, + { + "epoch": 2.57, + "learning_rate": 1.2778578978432843e-05, + "loss": 1.0155, + "step": 34758 + }, + { + "epoch": 2.57, + "learning_rate": 1.2778195811294142e-05, + "loss": 1.0599, + "step": 34759 + }, + { + "epoch": 2.57, + "learning_rate": 1.277781263973537e-05, + "loss": 0.9795, + "step": 34760 + }, + { + "epoch": 2.57, + "learning_rate": 1.277742946375714e-05, + "loss": 1.0477, + "step": 34761 + }, + { + "epoch": 2.57, + "learning_rate": 1.2777046283360057e-05, + "loss": 1.0022, + "step": 34762 + }, + { + "epoch": 2.57, + "learning_rate": 1.2776663098544738e-05, + "loss": 0.9409, + "step": 34763 + }, + { + "epoch": 2.57, + "learning_rate": 1.2776279909311782e-05, + "loss": 1.0236, + "step": 34764 + }, + { + "epoch": 2.57, + "learning_rate": 1.2775896715661808e-05, + "loss": 1.0697, + "step": 34765 + }, + { + "epoch": 2.57, + "learning_rate": 1.2775513517595419e-05, + "loss": 0.9138, + "step": 34766 + }, + { + "epoch": 2.57, + "learning_rate": 1.277513031511323e-05, + "loss": 0.9344, + "step": 34767 + }, + { + "epoch": 2.57, + "learning_rate": 1.2774747108215849e-05, + "loss": 1.0173, + "step": 34768 + }, + { + "epoch": 2.57, + "learning_rate": 1.2774363896903884e-05, + "loss": 1.0018, + "step": 34769 + }, + { + "epoch": 2.57, + "learning_rate": 1.2773980681177944e-05, + "loss": 0.9909, + "step": 34770 + }, + { + "epoch": 2.57, + "learning_rate": 1.2773597461038643e-05, + "loss": 1.0035, + "step": 34771 + }, + { + "epoch": 2.57, + "learning_rate": 1.2773214236486586e-05, + "loss": 0.9478, + "step": 34772 + }, + { + "epoch": 2.57, + "learning_rate": 1.2772831007522387e-05, + "loss": 1.0446, + "step": 34773 + }, + { + "epoch": 2.57, + "learning_rate": 1.2772447774146652e-05, + "loss": 1.0232, + "step": 34774 + }, + { + "epoch": 2.57, + "learning_rate": 1.2772064536359991e-05, + "loss": 1.0176, + "step": 34775 + }, + { + "epoch": 2.57, + "learning_rate": 1.277168129416302e-05, + "loss": 1.0709, + "step": 34776 + }, + { + "epoch": 2.57, + "learning_rate": 1.2771298047556338e-05, + "loss": 0.8081, + "step": 34777 + }, + { + "epoch": 2.57, + "learning_rate": 1.2770914796540563e-05, + "loss": 0.9895, + "step": 34778 + }, + { + "epoch": 2.57, + "learning_rate": 1.2770531541116304e-05, + "loss": 0.9784, + "step": 34779 + }, + { + "epoch": 2.57, + "learning_rate": 1.2770148281284164e-05, + "loss": 1.1681, + "step": 34780 + }, + { + "epoch": 2.57, + "learning_rate": 1.2769765017044765e-05, + "loss": 0.8714, + "step": 34781 + }, + { + "epoch": 2.57, + "learning_rate": 1.2769381748398701e-05, + "loss": 1.0286, + "step": 34782 + }, + { + "epoch": 2.57, + "learning_rate": 1.2768998475346596e-05, + "loss": 0.9826, + "step": 34783 + }, + { + "epoch": 2.57, + "learning_rate": 1.2768615197889054e-05, + "loss": 0.9111, + "step": 34784 + }, + { + "epoch": 2.57, + "learning_rate": 1.2768231916026683e-05, + "loss": 1.0064, + "step": 34785 + }, + { + "epoch": 2.57, + "learning_rate": 1.2767848629760095e-05, + "loss": 1.0187, + "step": 34786 + }, + { + "epoch": 2.57, + "learning_rate": 1.27674653390899e-05, + "loss": 0.9781, + "step": 34787 + }, + { + "epoch": 2.57, + "learning_rate": 1.2767082044016705e-05, + "loss": 0.9443, + "step": 34788 + }, + { + "epoch": 2.57, + "learning_rate": 1.2766698744541124e-05, + "loss": 0.9553, + "step": 34789 + }, + { + "epoch": 2.57, + "learning_rate": 1.2766315440663766e-05, + "loss": 0.9302, + "step": 34790 + }, + { + "epoch": 2.57, + "learning_rate": 1.2765932132385237e-05, + "loss": 0.9324, + "step": 34791 + }, + { + "epoch": 2.57, + "learning_rate": 1.2765548819706151e-05, + "loss": 1.0773, + "step": 34792 + }, + { + "epoch": 2.57, + "learning_rate": 1.2765165502627117e-05, + "loss": 0.9644, + "step": 34793 + }, + { + "epoch": 2.57, + "learning_rate": 1.2764782181148746e-05, + "loss": 1.006, + "step": 34794 + }, + { + "epoch": 2.57, + "learning_rate": 1.276439885527164e-05, + "loss": 0.9311, + "step": 34795 + }, + { + "epoch": 2.57, + "learning_rate": 1.2764015524996421e-05, + "loss": 1.0094, + "step": 34796 + }, + { + "epoch": 2.57, + "learning_rate": 1.2763632190323689e-05, + "loss": 0.9958, + "step": 34797 + }, + { + "epoch": 2.57, + "learning_rate": 1.276324885125406e-05, + "loss": 0.9937, + "step": 34798 + }, + { + "epoch": 2.57, + "learning_rate": 1.2762865507788141e-05, + "loss": 0.8794, + "step": 34799 + }, + { + "epoch": 2.57, + "learning_rate": 1.2762482159926542e-05, + "loss": 1.1287, + "step": 34800 + }, + { + "epoch": 2.57, + "learning_rate": 1.2762098807669872e-05, + "loss": 1.1197, + "step": 34801 + }, + { + "epoch": 2.57, + "learning_rate": 1.2761715451018749e-05, + "loss": 0.96, + "step": 34802 + }, + { + "epoch": 2.57, + "learning_rate": 1.2761332089973768e-05, + "loss": 1.0198, + "step": 34803 + }, + { + "epoch": 2.57, + "learning_rate": 1.2760948724535552e-05, + "loss": 0.9516, + "step": 34804 + }, + { + "epoch": 2.57, + "learning_rate": 1.2760565354704703e-05, + "loss": 1.0288, + "step": 34805 + }, + { + "epoch": 2.57, + "learning_rate": 1.2760181980481836e-05, + "loss": 0.9474, + "step": 34806 + }, + { + "epoch": 2.57, + "learning_rate": 1.2759798601867557e-05, + "loss": 0.996, + "step": 34807 + }, + { + "epoch": 2.57, + "learning_rate": 1.2759415218862481e-05, + "loss": 1.0089, + "step": 34808 + }, + { + "epoch": 2.57, + "learning_rate": 1.2759031831467212e-05, + "loss": 1.0017, + "step": 34809 + }, + { + "epoch": 2.57, + "learning_rate": 1.2758648439682367e-05, + "loss": 1.0533, + "step": 34810 + }, + { + "epoch": 2.57, + "learning_rate": 1.2758265043508546e-05, + "loss": 0.9156, + "step": 34811 + }, + { + "epoch": 2.57, + "learning_rate": 1.275788164294637e-05, + "loss": 1.0437, + "step": 34812 + }, + { + "epoch": 2.57, + "learning_rate": 1.275749823799644e-05, + "loss": 1.0012, + "step": 34813 + }, + { + "epoch": 2.57, + "learning_rate": 1.2757114828659373e-05, + "loss": 0.9493, + "step": 34814 + }, + { + "epoch": 2.57, + "learning_rate": 1.2756731414935773e-05, + "loss": 1.0562, + "step": 34815 + }, + { + "epoch": 2.57, + "learning_rate": 1.2756347996826255e-05, + "loss": 1.0678, + "step": 34816 + }, + { + "epoch": 2.57, + "learning_rate": 1.2755964574331423e-05, + "loss": 1.028, + "step": 34817 + }, + { + "epoch": 2.57, + "learning_rate": 1.2755581147451894e-05, + "loss": 1.0132, + "step": 34818 + }, + { + "epoch": 2.57, + "learning_rate": 1.2755197716188273e-05, + "loss": 0.892, + "step": 34819 + }, + { + "epoch": 2.57, + "learning_rate": 1.2754814280541175e-05, + "loss": 0.9619, + "step": 34820 + }, + { + "epoch": 2.57, + "learning_rate": 1.2754430840511205e-05, + "loss": 0.9237, + "step": 34821 + }, + { + "epoch": 2.57, + "learning_rate": 1.2754047396098976e-05, + "loss": 0.9275, + "step": 34822 + }, + { + "epoch": 2.57, + "learning_rate": 1.2753663947305095e-05, + "loss": 1.0957, + "step": 34823 + }, + { + "epoch": 2.57, + "learning_rate": 1.2753280494130176e-05, + "loss": 0.9959, + "step": 34824 + }, + { + "epoch": 2.57, + "learning_rate": 1.2752897036574825e-05, + "loss": 1.0037, + "step": 34825 + }, + { + "epoch": 2.57, + "learning_rate": 1.2752513574639656e-05, + "loss": 1.0925, + "step": 34826 + }, + { + "epoch": 2.57, + "learning_rate": 1.2752130108325273e-05, + "loss": 1.045, + "step": 34827 + }, + { + "epoch": 2.57, + "learning_rate": 1.2751746637632296e-05, + "loss": 0.9906, + "step": 34828 + }, + { + "epoch": 2.57, + "learning_rate": 1.2751363162561325e-05, + "loss": 1.1344, + "step": 34829 + }, + { + "epoch": 2.57, + "learning_rate": 1.2750979683112978e-05, + "loss": 0.9059, + "step": 34830 + }, + { + "epoch": 2.57, + "learning_rate": 1.2750596199287864e-05, + "loss": 1.0724, + "step": 34831 + }, + { + "epoch": 2.57, + "learning_rate": 1.2750212711086585e-05, + "loss": 1.0272, + "step": 34832 + }, + { + "epoch": 2.57, + "learning_rate": 1.2749829218509759e-05, + "loss": 0.9391, + "step": 34833 + }, + { + "epoch": 2.57, + "learning_rate": 1.2749445721557994e-05, + "loss": 0.915, + "step": 34834 + }, + { + "epoch": 2.57, + "learning_rate": 1.2749062220231901e-05, + "loss": 1.0275, + "step": 34835 + }, + { + "epoch": 2.57, + "learning_rate": 1.2748678714532088e-05, + "loss": 0.8858, + "step": 34836 + }, + { + "epoch": 2.57, + "learning_rate": 1.2748295204459167e-05, + "loss": 1.0522, + "step": 34837 + }, + { + "epoch": 2.57, + "learning_rate": 1.2747911690013748e-05, + "loss": 1.0699, + "step": 34838 + }, + { + "epoch": 2.57, + "learning_rate": 1.2747528171196442e-05, + "loss": 0.9507, + "step": 34839 + }, + { + "epoch": 2.57, + "learning_rate": 1.2747144648007856e-05, + "loss": 0.993, + "step": 34840 + }, + { + "epoch": 2.57, + "learning_rate": 1.2746761120448605e-05, + "loss": 0.9641, + "step": 34841 + }, + { + "epoch": 2.57, + "learning_rate": 1.2746377588519292e-05, + "loss": 1.0238, + "step": 34842 + }, + { + "epoch": 2.57, + "learning_rate": 1.2745994052220537e-05, + "loss": 1.0219, + "step": 34843 + }, + { + "epoch": 2.57, + "learning_rate": 1.2745610511552939e-05, + "loss": 1.0102, + "step": 34844 + }, + { + "epoch": 2.57, + "learning_rate": 1.2745226966517116e-05, + "loss": 0.9494, + "step": 34845 + }, + { + "epoch": 2.57, + "learning_rate": 1.2744843417113677e-05, + "loss": 1.1472, + "step": 34846 + }, + { + "epoch": 2.57, + "learning_rate": 1.2744459863343233e-05, + "loss": 1.0426, + "step": 34847 + }, + { + "epoch": 2.57, + "learning_rate": 1.274407630520639e-05, + "loss": 1.0662, + "step": 34848 + }, + { + "epoch": 2.57, + "learning_rate": 1.2743692742703763e-05, + "loss": 1.1205, + "step": 34849 + }, + { + "epoch": 2.58, + "learning_rate": 1.274330917583596e-05, + "loss": 1.038, + "step": 34850 + }, + { + "epoch": 2.58, + "learning_rate": 1.2742925604603591e-05, + "loss": 0.972, + "step": 34851 + }, + { + "epoch": 2.58, + "learning_rate": 1.2742542029007264e-05, + "loss": 0.9559, + "step": 34852 + }, + { + "epoch": 2.58, + "learning_rate": 1.2742158449047595e-05, + "loss": 1.0422, + "step": 34853 + }, + { + "epoch": 2.58, + "learning_rate": 1.2741774864725186e-05, + "loss": 0.9013, + "step": 34854 + }, + { + "epoch": 2.58, + "learning_rate": 1.2741391276040658e-05, + "loss": 0.9963, + "step": 34855 + }, + { + "epoch": 2.58, + "learning_rate": 1.2741007682994613e-05, + "loss": 1.0589, + "step": 34856 + }, + { + "epoch": 2.58, + "learning_rate": 1.2740624085587665e-05, + "loss": 0.9996, + "step": 34857 + }, + { + "epoch": 2.58, + "learning_rate": 1.2740240483820425e-05, + "loss": 0.9141, + "step": 34858 + }, + { + "epoch": 2.58, + "learning_rate": 1.2739856877693501e-05, + "loss": 0.9234, + "step": 34859 + }, + { + "epoch": 2.58, + "learning_rate": 1.27394732672075e-05, + "loss": 0.8646, + "step": 34860 + }, + { + "epoch": 2.58, + "learning_rate": 1.2739089652363039e-05, + "loss": 0.9573, + "step": 34861 + }, + { + "epoch": 2.58, + "learning_rate": 1.2738706033160725e-05, + "loss": 1.0187, + "step": 34862 + }, + { + "epoch": 2.58, + "learning_rate": 1.2738322409601172e-05, + "loss": 0.9557, + "step": 34863 + }, + { + "epoch": 2.58, + "learning_rate": 1.2737938781684982e-05, + "loss": 1.0385, + "step": 34864 + }, + { + "epoch": 2.58, + "learning_rate": 1.2737555149412775e-05, + "loss": 0.9575, + "step": 34865 + }, + { + "epoch": 2.58, + "learning_rate": 1.2737171512785154e-05, + "loss": 0.8649, + "step": 34866 + }, + { + "epoch": 2.58, + "learning_rate": 1.2736787871802734e-05, + "loss": 0.9226, + "step": 34867 + }, + { + "epoch": 2.58, + "learning_rate": 1.2736404226466124e-05, + "loss": 0.9769, + "step": 34868 + }, + { + "epoch": 2.58, + "learning_rate": 1.2736020576775934e-05, + "loss": 1.0025, + "step": 34869 + }, + { + "epoch": 2.58, + "learning_rate": 1.2735636922732772e-05, + "loss": 0.9435, + "step": 34870 + }, + { + "epoch": 2.58, + "learning_rate": 1.2735253264337253e-05, + "loss": 0.9523, + "step": 34871 + }, + { + "epoch": 2.58, + "learning_rate": 1.2734869601589983e-05, + "loss": 1.0051, + "step": 34872 + }, + { + "epoch": 2.58, + "learning_rate": 1.2734485934491576e-05, + "loss": 0.9906, + "step": 34873 + }, + { + "epoch": 2.58, + "learning_rate": 1.273410226304264e-05, + "loss": 1.0996, + "step": 34874 + }, + { + "epoch": 2.58, + "learning_rate": 1.2733718587243789e-05, + "loss": 0.9607, + "step": 34875 + }, + { + "epoch": 2.58, + "learning_rate": 1.2733334907095627e-05, + "loss": 0.9781, + "step": 34876 + }, + { + "epoch": 2.58, + "learning_rate": 1.2732951222598772e-05, + "loss": 1.035, + "step": 34877 + }, + { + "epoch": 2.58, + "learning_rate": 1.273256753375383e-05, + "loss": 1.0873, + "step": 34878 + }, + { + "epoch": 2.58, + "learning_rate": 1.2732183840561412e-05, + "loss": 1.0358, + "step": 34879 + }, + { + "epoch": 2.58, + "learning_rate": 1.2731800143022129e-05, + "loss": 0.9522, + "step": 34880 + }, + { + "epoch": 2.58, + "learning_rate": 1.2731416441136589e-05, + "loss": 0.9288, + "step": 34881 + }, + { + "epoch": 2.58, + "learning_rate": 1.2731032734905404e-05, + "loss": 0.8836, + "step": 34882 + }, + { + "epoch": 2.58, + "learning_rate": 1.2730649024329189e-05, + "loss": 1.0784, + "step": 34883 + }, + { + "epoch": 2.58, + "learning_rate": 1.2730265309408547e-05, + "loss": 1.111, + "step": 34884 + }, + { + "epoch": 2.58, + "learning_rate": 1.2729881590144093e-05, + "loss": 1.1846, + "step": 34885 + }, + { + "epoch": 2.58, + "learning_rate": 1.2729497866536437e-05, + "loss": 1.0182, + "step": 34886 + }, + { + "epoch": 2.58, + "learning_rate": 1.2729114138586188e-05, + "loss": 1.0188, + "step": 34887 + }, + { + "epoch": 2.58, + "learning_rate": 1.2728730406293959e-05, + "loss": 1.0183, + "step": 34888 + }, + { + "epoch": 2.58, + "learning_rate": 1.2728346669660356e-05, + "loss": 0.928, + "step": 34889 + }, + { + "epoch": 2.58, + "learning_rate": 1.2727962928685995e-05, + "loss": 0.964, + "step": 34890 + }, + { + "epoch": 2.58, + "learning_rate": 1.2727579183371482e-05, + "loss": 1.0058, + "step": 34891 + }, + { + "epoch": 2.58, + "learning_rate": 1.2727195433717432e-05, + "loss": 0.9134, + "step": 34892 + }, + { + "epoch": 2.58, + "learning_rate": 1.2726811679724452e-05, + "loss": 1.028, + "step": 34893 + }, + { + "epoch": 2.58, + "learning_rate": 1.2726427921393154e-05, + "loss": 1.0322, + "step": 34894 + }, + { + "epoch": 2.58, + "learning_rate": 1.2726044158724147e-05, + "loss": 1.0538, + "step": 34895 + }, + { + "epoch": 2.58, + "learning_rate": 1.2725660391718045e-05, + "loss": 0.9971, + "step": 34896 + }, + { + "epoch": 2.58, + "learning_rate": 1.2725276620375453e-05, + "loss": 0.9545, + "step": 34897 + }, + { + "epoch": 2.58, + "learning_rate": 1.2724892844696988e-05, + "loss": 0.9648, + "step": 34898 + }, + { + "epoch": 2.58, + "learning_rate": 1.2724509064683255e-05, + "loss": 0.9781, + "step": 34899 + }, + { + "epoch": 2.58, + "learning_rate": 1.2724125280334869e-05, + "loss": 0.938, + "step": 34900 + }, + { + "epoch": 2.58, + "learning_rate": 1.2723741491652438e-05, + "loss": 0.9724, + "step": 34901 + }, + { + "epoch": 2.58, + "learning_rate": 1.2723357698636573e-05, + "loss": 1.0782, + "step": 34902 + }, + { + "epoch": 2.58, + "learning_rate": 1.2722973901287885e-05, + "loss": 1.0774, + "step": 34903 + }, + { + "epoch": 2.58, + "learning_rate": 1.2722590099606984e-05, + "loss": 1.0703, + "step": 34904 + }, + { + "epoch": 2.58, + "learning_rate": 1.272220629359448e-05, + "loss": 0.9205, + "step": 34905 + }, + { + "epoch": 2.58, + "learning_rate": 1.272182248325099e-05, + "loss": 0.9687, + "step": 34906 + }, + { + "epoch": 2.58, + "learning_rate": 1.2721438668577112e-05, + "loss": 0.9109, + "step": 34907 + }, + { + "epoch": 2.58, + "learning_rate": 1.272105484957347e-05, + "loss": 1.011, + "step": 34908 + }, + { + "epoch": 2.58, + "learning_rate": 1.2720671026240665e-05, + "loss": 0.9416, + "step": 34909 + }, + { + "epoch": 2.58, + "learning_rate": 1.2720287198579316e-05, + "loss": 0.9693, + "step": 34910 + }, + { + "epoch": 2.58, + "learning_rate": 1.2719903366590023e-05, + "loss": 0.9921, + "step": 34911 + }, + { + "epoch": 2.58, + "learning_rate": 1.2719519530273407e-05, + "loss": 1.0136, + "step": 34912 + }, + { + "epoch": 2.58, + "learning_rate": 1.271913568963007e-05, + "loss": 1.0016, + "step": 34913 + }, + { + "epoch": 2.58, + "learning_rate": 1.2718751844660634e-05, + "loss": 0.9624, + "step": 34914 + }, + { + "epoch": 2.58, + "learning_rate": 1.2718367995365697e-05, + "loss": 0.8729, + "step": 34915 + }, + { + "epoch": 2.58, + "learning_rate": 1.2717984141745879e-05, + "loss": 0.9203, + "step": 34916 + }, + { + "epoch": 2.58, + "learning_rate": 1.2717600283801784e-05, + "loss": 0.8765, + "step": 34917 + }, + { + "epoch": 2.58, + "learning_rate": 1.2717216421534026e-05, + "loss": 1.0453, + "step": 34918 + }, + { + "epoch": 2.58, + "learning_rate": 1.2716832554943214e-05, + "loss": 0.8788, + "step": 34919 + }, + { + "epoch": 2.58, + "learning_rate": 1.2716448684029965e-05, + "loss": 0.9134, + "step": 34920 + }, + { + "epoch": 2.58, + "learning_rate": 1.2716064808794881e-05, + "loss": 1.0742, + "step": 34921 + }, + { + "epoch": 2.58, + "learning_rate": 1.2715680929238579e-05, + "loss": 0.9711, + "step": 34922 + }, + { + "epoch": 2.58, + "learning_rate": 1.2715297045361666e-05, + "loss": 0.925, + "step": 34923 + }, + { + "epoch": 2.58, + "learning_rate": 1.2714913157164757e-05, + "loss": 1.1004, + "step": 34924 + }, + { + "epoch": 2.58, + "learning_rate": 1.2714529264648459e-05, + "loss": 1.0288, + "step": 34925 + }, + { + "epoch": 2.58, + "learning_rate": 1.2714145367813382e-05, + "loss": 0.9935, + "step": 34926 + }, + { + "epoch": 2.58, + "learning_rate": 1.271376146666014e-05, + "loss": 0.9859, + "step": 34927 + }, + { + "epoch": 2.58, + "learning_rate": 1.2713377561189341e-05, + "loss": 0.9957, + "step": 34928 + }, + { + "epoch": 2.58, + "learning_rate": 1.2712993651401597e-05, + "loss": 1.0029, + "step": 34929 + }, + { + "epoch": 2.58, + "learning_rate": 1.271260973729752e-05, + "loss": 0.951, + "step": 34930 + }, + { + "epoch": 2.58, + "learning_rate": 1.271222581887772e-05, + "loss": 0.9036, + "step": 34931 + }, + { + "epoch": 2.58, + "learning_rate": 1.2711841896142808e-05, + "loss": 1.0647, + "step": 34932 + }, + { + "epoch": 2.58, + "learning_rate": 1.2711457969093393e-05, + "loss": 1.0749, + "step": 34933 + }, + { + "epoch": 2.58, + "learning_rate": 1.2711074037730086e-05, + "loss": 1.0773, + "step": 34934 + }, + { + "epoch": 2.58, + "learning_rate": 1.2710690102053502e-05, + "loss": 0.9864, + "step": 34935 + }, + { + "epoch": 2.58, + "learning_rate": 1.2710306162064248e-05, + "loss": 0.9585, + "step": 34936 + }, + { + "epoch": 2.58, + "learning_rate": 1.2709922217762935e-05, + "loss": 0.959, + "step": 34937 + }, + { + "epoch": 2.58, + "learning_rate": 1.2709538269150174e-05, + "loss": 0.9595, + "step": 34938 + }, + { + "epoch": 2.58, + "learning_rate": 1.2709154316226577e-05, + "loss": 1.0277, + "step": 34939 + }, + { + "epoch": 2.58, + "learning_rate": 1.2708770358992752e-05, + "loss": 1.0047, + "step": 34940 + }, + { + "epoch": 2.58, + "learning_rate": 1.2708386397449318e-05, + "loss": 1.0141, + "step": 34941 + }, + { + "epoch": 2.58, + "learning_rate": 1.2708002431596875e-05, + "loss": 0.9674, + "step": 34942 + }, + { + "epoch": 2.58, + "learning_rate": 1.2707618461436042e-05, + "loss": 1.0414, + "step": 34943 + }, + { + "epoch": 2.58, + "learning_rate": 1.2707234486967424e-05, + "loss": 0.8252, + "step": 34944 + }, + { + "epoch": 2.58, + "learning_rate": 1.2706850508191637e-05, + "loss": 0.993, + "step": 34945 + }, + { + "epoch": 2.58, + "learning_rate": 1.2706466525109287e-05, + "loss": 1.104, + "step": 34946 + }, + { + "epoch": 2.58, + "learning_rate": 1.270608253772099e-05, + "loss": 0.9253, + "step": 34947 + }, + { + "epoch": 2.58, + "learning_rate": 1.270569854602735e-05, + "loss": 1.0092, + "step": 34948 + }, + { + "epoch": 2.58, + "learning_rate": 1.2705314550028986e-05, + "loss": 0.9727, + "step": 34949 + }, + { + "epoch": 2.58, + "learning_rate": 1.2704930549726503e-05, + "loss": 1.0458, + "step": 34950 + }, + { + "epoch": 2.58, + "learning_rate": 1.2704546545120517e-05, + "loss": 0.9914, + "step": 34951 + }, + { + "epoch": 2.58, + "learning_rate": 1.2704162536211638e-05, + "loss": 1.0652, + "step": 34952 + }, + { + "epoch": 2.58, + "learning_rate": 1.2703778523000471e-05, + "loss": 0.9027, + "step": 34953 + }, + { + "epoch": 2.58, + "learning_rate": 1.2703394505487631e-05, + "loss": 1.049, + "step": 34954 + }, + { + "epoch": 2.58, + "learning_rate": 1.2703010483673732e-05, + "loss": 1.0098, + "step": 34955 + }, + { + "epoch": 2.58, + "learning_rate": 1.2702626457559379e-05, + "loss": 0.9892, + "step": 34956 + }, + { + "epoch": 2.58, + "learning_rate": 1.2702242427145188e-05, + "loss": 1.0629, + "step": 34957 + }, + { + "epoch": 2.58, + "learning_rate": 1.2701858392431765e-05, + "loss": 1.0546, + "step": 34958 + }, + { + "epoch": 2.58, + "learning_rate": 1.270147435341973e-05, + "loss": 0.9411, + "step": 34959 + }, + { + "epoch": 2.58, + "learning_rate": 1.270109031010968e-05, + "loss": 1.0369, + "step": 34960 + }, + { + "epoch": 2.58, + "learning_rate": 1.2700706262502243e-05, + "loss": 0.8991, + "step": 34961 + }, + { + "epoch": 2.58, + "learning_rate": 1.2700322210598013e-05, + "loss": 0.8989, + "step": 34962 + }, + { + "epoch": 2.58, + "learning_rate": 1.2699938154397615e-05, + "loss": 0.9968, + "step": 34963 + }, + { + "epoch": 2.58, + "learning_rate": 1.269955409390165e-05, + "loss": 0.9721, + "step": 34964 + }, + { + "epoch": 2.58, + "learning_rate": 1.2699170029110735e-05, + "loss": 0.947, + "step": 34965 + }, + { + "epoch": 2.58, + "learning_rate": 1.2698785960025477e-05, + "loss": 1.0036, + "step": 34966 + }, + { + "epoch": 2.58, + "learning_rate": 1.2698401886646489e-05, + "loss": 1.0389, + "step": 34967 + }, + { + "epoch": 2.58, + "learning_rate": 1.2698017808974384e-05, + "loss": 0.9379, + "step": 34968 + }, + { + "epoch": 2.58, + "learning_rate": 1.269763372700977e-05, + "loss": 1.0569, + "step": 34969 + }, + { + "epoch": 2.58, + "learning_rate": 1.2697249640753262e-05, + "loss": 1.0129, + "step": 34970 + }, + { + "epoch": 2.58, + "learning_rate": 1.2696865550205467e-05, + "loss": 0.9906, + "step": 34971 + }, + { + "epoch": 2.58, + "learning_rate": 1.2696481455366995e-05, + "loss": 0.9924, + "step": 34972 + }, + { + "epoch": 2.58, + "learning_rate": 1.2696097356238463e-05, + "loss": 0.9118, + "step": 34973 + }, + { + "epoch": 2.58, + "learning_rate": 1.2695713252820476e-05, + "loss": 1.0449, + "step": 34974 + }, + { + "epoch": 2.58, + "learning_rate": 1.269532914511365e-05, + "loss": 0.9668, + "step": 34975 + }, + { + "epoch": 2.58, + "learning_rate": 1.2694945033118593e-05, + "loss": 1.0509, + "step": 34976 + }, + { + "epoch": 2.58, + "learning_rate": 1.2694560916835919e-05, + "loss": 1.0618, + "step": 34977 + }, + { + "epoch": 2.58, + "learning_rate": 1.2694176796266233e-05, + "loss": 1.0366, + "step": 34978 + }, + { + "epoch": 2.58, + "learning_rate": 1.2693792671410153e-05, + "loss": 0.9803, + "step": 34979 + }, + { + "epoch": 2.58, + "learning_rate": 1.2693408542268287e-05, + "loss": 1.0083, + "step": 34980 + }, + { + "epoch": 2.58, + "learning_rate": 1.2693024408841246e-05, + "loss": 1.1204, + "step": 34981 + }, + { + "epoch": 2.58, + "learning_rate": 1.2692640271129641e-05, + "loss": 0.9987, + "step": 34982 + }, + { + "epoch": 2.58, + "learning_rate": 1.2692256129134086e-05, + "loss": 0.986, + "step": 34983 + }, + { + "epoch": 2.58, + "learning_rate": 1.2691871982855189e-05, + "loss": 0.9866, + "step": 34984 + }, + { + "epoch": 2.59, + "learning_rate": 1.269148783229356e-05, + "loss": 0.9421, + "step": 34985 + }, + { + "epoch": 2.59, + "learning_rate": 1.2691103677449817e-05, + "loss": 1.0316, + "step": 34986 + }, + { + "epoch": 2.59, + "learning_rate": 1.2690719518324562e-05, + "loss": 1.0133, + "step": 34987 + }, + { + "epoch": 2.59, + "learning_rate": 1.2690335354918415e-05, + "loss": 0.9691, + "step": 34988 + }, + { + "epoch": 2.59, + "learning_rate": 1.2689951187231979e-05, + "loss": 0.9852, + "step": 34989 + }, + { + "epoch": 2.59, + "learning_rate": 1.2689567015265874e-05, + "loss": 0.9917, + "step": 34990 + }, + { + "epoch": 2.59, + "learning_rate": 1.2689182839020701e-05, + "loss": 1.0221, + "step": 34991 + }, + { + "epoch": 2.59, + "learning_rate": 1.2688798658497078e-05, + "loss": 0.9693, + "step": 34992 + }, + { + "epoch": 2.59, + "learning_rate": 1.2688414473695615e-05, + "loss": 0.9626, + "step": 34993 + }, + { + "epoch": 2.59, + "learning_rate": 1.2688030284616926e-05, + "loss": 1.1461, + "step": 34994 + }, + { + "epoch": 2.59, + "learning_rate": 1.2687646091261616e-05, + "loss": 1.0149, + "step": 34995 + }, + { + "epoch": 2.59, + "learning_rate": 1.2687261893630301e-05, + "loss": 0.9285, + "step": 34996 + }, + { + "epoch": 2.59, + "learning_rate": 1.2686877691723589e-05, + "loss": 0.9639, + "step": 34997 + }, + { + "epoch": 2.59, + "learning_rate": 1.2686493485542095e-05, + "loss": 1.0887, + "step": 34998 + }, + { + "epoch": 2.59, + "learning_rate": 1.268610927508643e-05, + "loss": 0.9083, + "step": 34999 + }, + { + "epoch": 2.59, + "learning_rate": 1.2685725060357199e-05, + "loss": 1.0273, + "step": 35000 + }, + { + "epoch": 2.59, + "learning_rate": 1.2685340841355021e-05, + "loss": 0.955, + "step": 35001 + }, + { + "epoch": 2.59, + "learning_rate": 1.2684956618080503e-05, + "loss": 0.9415, + "step": 35002 + }, + { + "epoch": 2.59, + "learning_rate": 1.2684572390534258e-05, + "loss": 1.0587, + "step": 35003 + }, + { + "epoch": 2.59, + "learning_rate": 1.2684188158716895e-05, + "loss": 0.937, + "step": 35004 + }, + { + "epoch": 2.59, + "learning_rate": 1.2683803922629027e-05, + "loss": 1.0784, + "step": 35005 + }, + { + "epoch": 2.59, + "learning_rate": 1.268341968227127e-05, + "loss": 1.049, + "step": 35006 + }, + { + "epoch": 2.59, + "learning_rate": 1.2683035437644224e-05, + "loss": 0.9545, + "step": 35007 + }, + { + "epoch": 2.59, + "learning_rate": 1.2682651188748513e-05, + "loss": 0.9502, + "step": 35008 + }, + { + "epoch": 2.59, + "learning_rate": 1.2682266935584739e-05, + "loss": 1.0553, + "step": 35009 + }, + { + "epoch": 2.59, + "learning_rate": 1.268188267815352e-05, + "loss": 1.0279, + "step": 35010 + }, + { + "epoch": 2.59, + "learning_rate": 1.268149841645546e-05, + "loss": 0.9419, + "step": 35011 + }, + { + "epoch": 2.59, + "learning_rate": 1.2681114150491176e-05, + "loss": 1.055, + "step": 35012 + }, + { + "epoch": 2.59, + "learning_rate": 1.2680729880261275e-05, + "loss": 1.0705, + "step": 35013 + }, + { + "epoch": 2.59, + "learning_rate": 1.2680345605766376e-05, + "loss": 0.9614, + "step": 35014 + }, + { + "epoch": 2.59, + "learning_rate": 1.2679961327007082e-05, + "loss": 1.0151, + "step": 35015 + }, + { + "epoch": 2.59, + "learning_rate": 1.267957704398401e-05, + "loss": 0.9098, + "step": 35016 + }, + { + "epoch": 2.59, + "learning_rate": 1.267919275669777e-05, + "loss": 1.0756, + "step": 35017 + }, + { + "epoch": 2.59, + "learning_rate": 1.2678808465148971e-05, + "loss": 1.0091, + "step": 35018 + }, + { + "epoch": 2.59, + "learning_rate": 1.2678424169338227e-05, + "loss": 1.0035, + "step": 35019 + }, + { + "epoch": 2.59, + "learning_rate": 1.2678039869266145e-05, + "loss": 1.0517, + "step": 35020 + }, + { + "epoch": 2.59, + "learning_rate": 1.2677655564933343e-05, + "loss": 0.9649, + "step": 35021 + }, + { + "epoch": 2.59, + "learning_rate": 1.2677271256340429e-05, + "loss": 1.047, + "step": 35022 + }, + { + "epoch": 2.59, + "learning_rate": 1.2676886943488013e-05, + "loss": 0.9259, + "step": 35023 + }, + { + "epoch": 2.59, + "learning_rate": 1.2676502626376709e-05, + "loss": 1.0272, + "step": 35024 + }, + { + "epoch": 2.59, + "learning_rate": 1.2676118305007128e-05, + "loss": 0.9565, + "step": 35025 + }, + { + "epoch": 2.59, + "learning_rate": 1.2675733979379883e-05, + "loss": 1.0363, + "step": 35026 + }, + { + "epoch": 2.59, + "learning_rate": 1.2675349649495583e-05, + "loss": 1.0007, + "step": 35027 + }, + { + "epoch": 2.59, + "learning_rate": 1.2674965315354838e-05, + "loss": 1.1405, + "step": 35028 + }, + { + "epoch": 2.59, + "learning_rate": 1.2674580976958261e-05, + "loss": 0.9989, + "step": 35029 + }, + { + "epoch": 2.59, + "learning_rate": 1.2674196634306464e-05, + "loss": 0.9692, + "step": 35030 + }, + { + "epoch": 2.59, + "learning_rate": 1.267381228740006e-05, + "loss": 0.9847, + "step": 35031 + }, + { + "epoch": 2.59, + "learning_rate": 1.2673427936239659e-05, + "loss": 1.0422, + "step": 35032 + }, + { + "epoch": 2.59, + "learning_rate": 1.2673043580825872e-05, + "loss": 0.9513, + "step": 35033 + }, + { + "epoch": 2.59, + "learning_rate": 1.2672659221159314e-05, + "loss": 1.1492, + "step": 35034 + }, + { + "epoch": 2.59, + "learning_rate": 1.2672274857240587e-05, + "loss": 1.0709, + "step": 35035 + }, + { + "epoch": 2.59, + "learning_rate": 1.2671890489070312e-05, + "loss": 0.9173, + "step": 35036 + }, + { + "epoch": 2.59, + "learning_rate": 1.2671506116649099e-05, + "loss": 1.1428, + "step": 35037 + }, + { + "epoch": 2.59, + "learning_rate": 1.2671121739977556e-05, + "loss": 0.997, + "step": 35038 + }, + { + "epoch": 2.59, + "learning_rate": 1.2670737359056297e-05, + "loss": 1.0372, + "step": 35039 + }, + { + "epoch": 2.59, + "learning_rate": 1.2670352973885933e-05, + "loss": 0.8612, + "step": 35040 + }, + { + "epoch": 2.59, + "learning_rate": 1.2669968584467079e-05, + "loss": 0.9705, + "step": 35041 + }, + { + "epoch": 2.59, + "learning_rate": 1.2669584190800337e-05, + "loss": 1.0411, + "step": 35042 + }, + { + "epoch": 2.59, + "learning_rate": 1.2669199792886328e-05, + "loss": 0.9168, + "step": 35043 + }, + { + "epoch": 2.59, + "learning_rate": 1.2668815390725659e-05, + "loss": 0.9677, + "step": 35044 + }, + { + "epoch": 2.59, + "learning_rate": 1.2668430984318947e-05, + "loss": 1.1362, + "step": 35045 + }, + { + "epoch": 2.59, + "learning_rate": 1.2668046573666794e-05, + "loss": 1.0494, + "step": 35046 + }, + { + "epoch": 2.59, + "learning_rate": 1.2667662158769823e-05, + "loss": 0.9934, + "step": 35047 + }, + { + "epoch": 2.59, + "learning_rate": 1.2667277739628634e-05, + "loss": 0.9128, + "step": 35048 + }, + { + "epoch": 2.59, + "learning_rate": 1.2666893316243847e-05, + "loss": 0.9804, + "step": 35049 + }, + { + "epoch": 2.59, + "learning_rate": 1.266650888861607e-05, + "loss": 0.9367, + "step": 35050 + }, + { + "epoch": 2.59, + "learning_rate": 1.2666124456745916e-05, + "loss": 0.9854, + "step": 35051 + }, + { + "epoch": 2.59, + "learning_rate": 1.2665740020633996e-05, + "loss": 1.0188, + "step": 35052 + }, + { + "epoch": 2.59, + "learning_rate": 1.2665355580280923e-05, + "loss": 1.0078, + "step": 35053 + }, + { + "epoch": 2.59, + "learning_rate": 1.2664971135687303e-05, + "loss": 1.0243, + "step": 35054 + }, + { + "epoch": 2.59, + "learning_rate": 1.266458668685376e-05, + "loss": 1.0318, + "step": 35055 + }, + { + "epoch": 2.59, + "learning_rate": 1.266420223378089e-05, + "loss": 1.0392, + "step": 35056 + }, + { + "epoch": 2.59, + "learning_rate": 1.2663817776469318e-05, + "loss": 1.0738, + "step": 35057 + }, + { + "epoch": 2.59, + "learning_rate": 1.2663433314919647e-05, + "loss": 0.9939, + "step": 35058 + }, + { + "epoch": 2.59, + "learning_rate": 1.2663048849132491e-05, + "loss": 1.0445, + "step": 35059 + }, + { + "epoch": 2.59, + "learning_rate": 1.2662664379108462e-05, + "loss": 0.8969, + "step": 35060 + }, + { + "epoch": 2.59, + "learning_rate": 1.2662279904848175e-05, + "loss": 0.9915, + "step": 35061 + }, + { + "epoch": 2.59, + "learning_rate": 1.2661895426352238e-05, + "loss": 0.9307, + "step": 35062 + }, + { + "epoch": 2.59, + "learning_rate": 1.2661510943621262e-05, + "loss": 1.0125, + "step": 35063 + }, + { + "epoch": 2.59, + "learning_rate": 1.2661126456655861e-05, + "loss": 0.9565, + "step": 35064 + }, + { + "epoch": 2.59, + "learning_rate": 1.2660741965456647e-05, + "loss": 1.0152, + "step": 35065 + }, + { + "epoch": 2.59, + "learning_rate": 1.266035747002423e-05, + "loss": 1.0475, + "step": 35066 + }, + { + "epoch": 2.59, + "learning_rate": 1.2659972970359222e-05, + "loss": 1.0842, + "step": 35067 + }, + { + "epoch": 2.59, + "learning_rate": 1.2659588466462234e-05, + "loss": 0.9568, + "step": 35068 + }, + { + "epoch": 2.59, + "learning_rate": 1.2659203958333878e-05, + "loss": 1.0361, + "step": 35069 + }, + { + "epoch": 2.59, + "learning_rate": 1.2658819445974768e-05, + "loss": 0.9192, + "step": 35070 + }, + { + "epoch": 2.59, + "learning_rate": 1.2658434929385517e-05, + "loss": 0.9843, + "step": 35071 + }, + { + "epoch": 2.59, + "learning_rate": 1.2658050408566728e-05, + "loss": 1.0357, + "step": 35072 + }, + { + "epoch": 2.59, + "learning_rate": 1.2657665883519024e-05, + "loss": 0.9913, + "step": 35073 + }, + { + "epoch": 2.59, + "learning_rate": 1.265728135424301e-05, + "loss": 1.0113, + "step": 35074 + }, + { + "epoch": 2.59, + "learning_rate": 1.26568968207393e-05, + "loss": 1.0496, + "step": 35075 + }, + { + "epoch": 2.59, + "learning_rate": 1.2656512283008505e-05, + "loss": 1.087, + "step": 35076 + }, + { + "epoch": 2.59, + "learning_rate": 1.2656127741051236e-05, + "loss": 1.0599, + "step": 35077 + }, + { + "epoch": 2.59, + "learning_rate": 1.2655743194868106e-05, + "loss": 0.9865, + "step": 35078 + }, + { + "epoch": 2.59, + "learning_rate": 1.2655358644459728e-05, + "loss": 1.0421, + "step": 35079 + }, + { + "epoch": 2.59, + "learning_rate": 1.2654974089826708e-05, + "loss": 1.0428, + "step": 35080 + }, + { + "epoch": 2.59, + "learning_rate": 1.2654589530969668e-05, + "loss": 0.9424, + "step": 35081 + }, + { + "epoch": 2.59, + "learning_rate": 1.265420496788921e-05, + "loss": 1.0223, + "step": 35082 + }, + { + "epoch": 2.59, + "learning_rate": 1.2653820400585952e-05, + "loss": 1.1292, + "step": 35083 + }, + { + "epoch": 2.59, + "learning_rate": 1.2653435829060503e-05, + "loss": 1.0674, + "step": 35084 + }, + { + "epoch": 2.59, + "learning_rate": 1.2653051253313475e-05, + "loss": 0.9684, + "step": 35085 + }, + { + "epoch": 2.59, + "learning_rate": 1.2652666673345482e-05, + "loss": 0.9562, + "step": 35086 + }, + { + "epoch": 2.59, + "learning_rate": 1.2652282089157133e-05, + "loss": 1.0615, + "step": 35087 + }, + { + "epoch": 2.59, + "learning_rate": 1.2651897500749041e-05, + "loss": 1.0052, + "step": 35088 + }, + { + "epoch": 2.59, + "learning_rate": 1.265151290812182e-05, + "loss": 0.9885, + "step": 35089 + }, + { + "epoch": 2.59, + "learning_rate": 1.2651128311276077e-05, + "loss": 1.0007, + "step": 35090 + }, + { + "epoch": 2.59, + "learning_rate": 1.2650743710212427e-05, + "loss": 1.0145, + "step": 35091 + }, + { + "epoch": 2.59, + "learning_rate": 1.2650359104931486e-05, + "loss": 0.9898, + "step": 35092 + }, + { + "epoch": 2.59, + "learning_rate": 1.2649974495433857e-05, + "loss": 1.0363, + "step": 35093 + }, + { + "epoch": 2.59, + "learning_rate": 1.264958988172016e-05, + "loss": 0.8861, + "step": 35094 + }, + { + "epoch": 2.59, + "learning_rate": 1.2649205263791e-05, + "loss": 1.0187, + "step": 35095 + }, + { + "epoch": 2.59, + "learning_rate": 1.2648820641646993e-05, + "loss": 0.9013, + "step": 35096 + }, + { + "epoch": 2.59, + "learning_rate": 1.2648436015288749e-05, + "loss": 1.0279, + "step": 35097 + }, + { + "epoch": 2.59, + "learning_rate": 1.2648051384716884e-05, + "loss": 0.9266, + "step": 35098 + }, + { + "epoch": 2.59, + "learning_rate": 1.2647666749932007e-05, + "loss": 1.0749, + "step": 35099 + }, + { + "epoch": 2.59, + "learning_rate": 1.2647282110934729e-05, + "loss": 1.056, + "step": 35100 + }, + { + "epoch": 2.59, + "learning_rate": 1.2646897467725665e-05, + "loss": 0.9455, + "step": 35101 + }, + { + "epoch": 2.59, + "learning_rate": 1.2646512820305423e-05, + "loss": 1.11, + "step": 35102 + }, + { + "epoch": 2.59, + "learning_rate": 1.2646128168674616e-05, + "loss": 0.9711, + "step": 35103 + }, + { + "epoch": 2.59, + "learning_rate": 1.2645743512833861e-05, + "loss": 1.1697, + "step": 35104 + }, + { + "epoch": 2.59, + "learning_rate": 1.264535885278376e-05, + "loss": 1.0681, + "step": 35105 + }, + { + "epoch": 2.59, + "learning_rate": 1.2644974188524937e-05, + "loss": 0.9498, + "step": 35106 + }, + { + "epoch": 2.59, + "learning_rate": 1.2644589520057993e-05, + "loss": 0.8633, + "step": 35107 + }, + { + "epoch": 2.59, + "learning_rate": 1.2644204847383547e-05, + "loss": 1.03, + "step": 35108 + }, + { + "epoch": 2.59, + "learning_rate": 1.2643820170502208e-05, + "loss": 1.0436, + "step": 35109 + }, + { + "epoch": 2.59, + "learning_rate": 1.264343548941459e-05, + "loss": 0.9836, + "step": 35110 + }, + { + "epoch": 2.59, + "learning_rate": 1.2643050804121307e-05, + "loss": 1.0082, + "step": 35111 + }, + { + "epoch": 2.59, + "learning_rate": 1.2642666114622964e-05, + "loss": 1.0279, + "step": 35112 + }, + { + "epoch": 2.59, + "learning_rate": 1.2642281420920178e-05, + "loss": 0.9341, + "step": 35113 + }, + { + "epoch": 2.59, + "learning_rate": 1.264189672301356e-05, + "loss": 1.013, + "step": 35114 + }, + { + "epoch": 2.59, + "learning_rate": 1.2641512020903721e-05, + "loss": 1.0006, + "step": 35115 + }, + { + "epoch": 2.59, + "learning_rate": 1.2641127314591274e-05, + "loss": 0.997, + "step": 35116 + }, + { + "epoch": 2.59, + "learning_rate": 1.2640742604076833e-05, + "loss": 1.0228, + "step": 35117 + }, + { + "epoch": 2.59, + "learning_rate": 1.2640357889361007e-05, + "loss": 0.9969, + "step": 35118 + }, + { + "epoch": 2.59, + "learning_rate": 1.2639973170444409e-05, + "loss": 0.9063, + "step": 35119 + }, + { + "epoch": 2.59, + "learning_rate": 1.2639588447327653e-05, + "loss": 1.032, + "step": 35120 + }, + { + "epoch": 2.6, + "learning_rate": 1.263920372001135e-05, + "loss": 1.044, + "step": 35121 + }, + { + "epoch": 2.6, + "learning_rate": 1.2638818988496114e-05, + "loss": 0.9558, + "step": 35122 + }, + { + "epoch": 2.6, + "learning_rate": 1.2638434252782548e-05, + "loss": 0.9914, + "step": 35123 + }, + { + "epoch": 2.6, + "learning_rate": 1.2638049512871276e-05, + "loss": 1.0252, + "step": 35124 + }, + { + "epoch": 2.6, + "learning_rate": 1.26376647687629e-05, + "loss": 1.0053, + "step": 35125 + }, + { + "epoch": 2.6, + "learning_rate": 1.2637280020458043e-05, + "loss": 1.0162, + "step": 35126 + }, + { + "epoch": 2.6, + "learning_rate": 1.2636895267957305e-05, + "loss": 1.0661, + "step": 35127 + }, + { + "epoch": 2.6, + "learning_rate": 1.263651051126131e-05, + "loss": 1.0206, + "step": 35128 + }, + { + "epoch": 2.6, + "learning_rate": 1.2636125750370663e-05, + "loss": 1.0733, + "step": 35129 + }, + { + "epoch": 2.6, + "learning_rate": 1.2635740985285976e-05, + "loss": 0.9882, + "step": 35130 + }, + { + "epoch": 2.6, + "learning_rate": 1.2635356216007866e-05, + "loss": 1.0282, + "step": 35131 + }, + { + "epoch": 2.6, + "learning_rate": 1.2634971442536938e-05, + "loss": 1.0162, + "step": 35132 + }, + { + "epoch": 2.6, + "learning_rate": 1.2634586664873812e-05, + "loss": 0.9477, + "step": 35133 + }, + { + "epoch": 2.6, + "learning_rate": 1.2634201883019092e-05, + "loss": 1.0258, + "step": 35134 + }, + { + "epoch": 2.6, + "learning_rate": 1.2633817096973398e-05, + "loss": 1.0594, + "step": 35135 + }, + { + "epoch": 2.6, + "learning_rate": 1.2633432306737335e-05, + "loss": 0.8768, + "step": 35136 + }, + { + "epoch": 2.6, + "learning_rate": 1.2633047512311523e-05, + "loss": 1.0726, + "step": 35137 + }, + { + "epoch": 2.6, + "learning_rate": 1.263266271369657e-05, + "loss": 0.9933, + "step": 35138 + }, + { + "epoch": 2.6, + "learning_rate": 1.2632277910893087e-05, + "loss": 1.0395, + "step": 35139 + }, + { + "epoch": 2.6, + "learning_rate": 1.2631893103901684e-05, + "loss": 0.9569, + "step": 35140 + }, + { + "epoch": 2.6, + "learning_rate": 1.2631508292722983e-05, + "loss": 0.8959, + "step": 35141 + }, + { + "epoch": 2.6, + "learning_rate": 1.2631123477357586e-05, + "loss": 0.892, + "step": 35142 + }, + { + "epoch": 2.6, + "learning_rate": 1.263073865780611e-05, + "loss": 1.048, + "step": 35143 + }, + { + "epoch": 2.6, + "learning_rate": 1.2630353834069164e-05, + "loss": 1.0395, + "step": 35144 + }, + { + "epoch": 2.6, + "learning_rate": 1.2629969006147367e-05, + "loss": 1.0093, + "step": 35145 + }, + { + "epoch": 2.6, + "learning_rate": 1.2629584174041325e-05, + "loss": 0.9992, + "step": 35146 + }, + { + "epoch": 2.6, + "learning_rate": 1.2629199337751652e-05, + "loss": 1.0138, + "step": 35147 + }, + { + "epoch": 2.6, + "learning_rate": 1.262881449727896e-05, + "loss": 0.9316, + "step": 35148 + }, + { + "epoch": 2.6, + "learning_rate": 1.2628429652623867e-05, + "loss": 1.0111, + "step": 35149 + }, + { + "epoch": 2.6, + "learning_rate": 1.2628044803786972e-05, + "loss": 1.0054, + "step": 35150 + }, + { + "epoch": 2.6, + "learning_rate": 1.26276599507689e-05, + "loss": 0.9485, + "step": 35151 + }, + { + "epoch": 2.6, + "learning_rate": 1.2627275093570258e-05, + "loss": 0.9181, + "step": 35152 + }, + { + "epoch": 2.6, + "learning_rate": 1.2626890232191657e-05, + "loss": 1.0532, + "step": 35153 + }, + { + "epoch": 2.6, + "learning_rate": 1.2626505366633712e-05, + "loss": 0.9791, + "step": 35154 + }, + { + "epoch": 2.6, + "learning_rate": 1.2626120496897036e-05, + "loss": 0.8984, + "step": 35155 + }, + { + "epoch": 2.6, + "learning_rate": 1.2625735622982239e-05, + "loss": 1.1253, + "step": 35156 + }, + { + "epoch": 2.6, + "learning_rate": 1.2625350744889935e-05, + "loss": 0.9512, + "step": 35157 + }, + { + "epoch": 2.6, + "learning_rate": 1.2624965862620735e-05, + "loss": 0.9958, + "step": 35158 + }, + { + "epoch": 2.6, + "learning_rate": 1.2624580976175254e-05, + "loss": 1.0098, + "step": 35159 + }, + { + "epoch": 2.6, + "learning_rate": 1.2624196085554096e-05, + "loss": 1.0362, + "step": 35160 + }, + { + "epoch": 2.6, + "learning_rate": 1.2623811190757884e-05, + "loss": 1.12, + "step": 35161 + }, + { + "epoch": 2.6, + "learning_rate": 1.2623426291787224e-05, + "loss": 1.1026, + "step": 35162 + }, + { + "epoch": 2.6, + "learning_rate": 1.2623041388642734e-05, + "loss": 1.0499, + "step": 35163 + }, + { + "epoch": 2.6, + "learning_rate": 1.262265648132502e-05, + "loss": 0.9739, + "step": 35164 + }, + { + "epoch": 2.6, + "learning_rate": 1.2622271569834697e-05, + "loss": 1.1218, + "step": 35165 + }, + { + "epoch": 2.6, + "learning_rate": 1.2621886654172377e-05, + "loss": 0.9309, + "step": 35166 + }, + { + "epoch": 2.6, + "learning_rate": 1.2621501734338677e-05, + "loss": 1.0498, + "step": 35167 + }, + { + "epoch": 2.6, + "learning_rate": 1.26211168103342e-05, + "loss": 1.0298, + "step": 35168 + }, + { + "epoch": 2.6, + "learning_rate": 1.2620731882159567e-05, + "loss": 1.0036, + "step": 35169 + }, + { + "epoch": 2.6, + "learning_rate": 1.2620346949815385e-05, + "loss": 1.0035, + "step": 35170 + }, + { + "epoch": 2.6, + "learning_rate": 1.261996201330227e-05, + "loss": 0.9737, + "step": 35171 + }, + { + "epoch": 2.6, + "learning_rate": 1.261957707262083e-05, + "loss": 0.9594, + "step": 35172 + }, + { + "epoch": 2.6, + "learning_rate": 1.2619192127771682e-05, + "loss": 0.9717, + "step": 35173 + }, + { + "epoch": 2.6, + "learning_rate": 1.2618807178755437e-05, + "loss": 1.0093, + "step": 35174 + }, + { + "epoch": 2.6, + "learning_rate": 1.2618422225572708e-05, + "loss": 0.9532, + "step": 35175 + }, + { + "epoch": 2.6, + "learning_rate": 1.2618037268224107e-05, + "loss": 0.9291, + "step": 35176 + }, + { + "epoch": 2.6, + "learning_rate": 1.2617652306710244e-05, + "loss": 1.0739, + "step": 35177 + }, + { + "epoch": 2.6, + "learning_rate": 1.2617267341031738e-05, + "loss": 1.058, + "step": 35178 + }, + { + "epoch": 2.6, + "learning_rate": 1.2616882371189192e-05, + "loss": 0.9935, + "step": 35179 + }, + { + "epoch": 2.6, + "learning_rate": 1.2616497397183224e-05, + "loss": 0.9725, + "step": 35180 + }, + { + "epoch": 2.6, + "learning_rate": 1.2616112419014448e-05, + "loss": 1.0997, + "step": 35181 + }, + { + "epoch": 2.6, + "learning_rate": 1.2615727436683476e-05, + "loss": 1.0575, + "step": 35182 + }, + { + "epoch": 2.6, + "learning_rate": 1.261534245019092e-05, + "loss": 1.103, + "step": 35183 + }, + { + "epoch": 2.6, + "learning_rate": 1.2614957459537384e-05, + "loss": 0.9585, + "step": 35184 + }, + { + "epoch": 2.6, + "learning_rate": 1.2614572464723496e-05, + "loss": 1.1088, + "step": 35185 + }, + { + "epoch": 2.6, + "learning_rate": 1.2614187465749858e-05, + "loss": 1.0433, + "step": 35186 + }, + { + "epoch": 2.6, + "learning_rate": 1.2613802462617085e-05, + "loss": 1.0066, + "step": 35187 + }, + { + "epoch": 2.6, + "learning_rate": 1.261341745532579e-05, + "loss": 1.1109, + "step": 35188 + }, + { + "epoch": 2.6, + "learning_rate": 1.2613032443876583e-05, + "loss": 0.8891, + "step": 35189 + }, + { + "epoch": 2.6, + "learning_rate": 1.2612647428270082e-05, + "loss": 0.9989, + "step": 35190 + }, + { + "epoch": 2.6, + "learning_rate": 1.2612262408506893e-05, + "loss": 0.9259, + "step": 35191 + }, + { + "epoch": 2.6, + "learning_rate": 1.2611877384587635e-05, + "loss": 0.9473, + "step": 35192 + }, + { + "epoch": 2.6, + "learning_rate": 1.2611492356512916e-05, + "loss": 1.0435, + "step": 35193 + }, + { + "epoch": 2.6, + "learning_rate": 1.261110732428335e-05, + "loss": 1.1111, + "step": 35194 + }, + { + "epoch": 2.6, + "learning_rate": 1.2610722287899551e-05, + "loss": 1.0399, + "step": 35195 + }, + { + "epoch": 2.6, + "learning_rate": 1.261033724736213e-05, + "loss": 0.9868, + "step": 35196 + }, + { + "epoch": 2.6, + "learning_rate": 1.2609952202671698e-05, + "loss": 1.0329, + "step": 35197 + }, + { + "epoch": 2.6, + "learning_rate": 1.260956715382887e-05, + "loss": 1.0547, + "step": 35198 + }, + { + "epoch": 2.6, + "learning_rate": 1.2609182100834257e-05, + "loss": 0.8319, + "step": 35199 + }, + { + "epoch": 2.6, + "learning_rate": 1.2608797043688474e-05, + "loss": 0.891, + "step": 35200 + }, + { + "epoch": 2.6, + "learning_rate": 1.2608411982392132e-05, + "loss": 1.0083, + "step": 35201 + }, + { + "epoch": 2.6, + "learning_rate": 1.2608026916945844e-05, + "loss": 0.9833, + "step": 35202 + }, + { + "epoch": 2.6, + "learning_rate": 1.260764184735022e-05, + "loss": 1.0991, + "step": 35203 + }, + { + "epoch": 2.6, + "learning_rate": 1.260725677360588e-05, + "loss": 1.0294, + "step": 35204 + }, + { + "epoch": 2.6, + "learning_rate": 1.260687169571343e-05, + "loss": 1.015, + "step": 35205 + }, + { + "epoch": 2.6, + "learning_rate": 1.2606486613673483e-05, + "loss": 1.0754, + "step": 35206 + }, + { + "epoch": 2.6, + "learning_rate": 1.2606101527486653e-05, + "loss": 1.04, + "step": 35207 + }, + { + "epoch": 2.6, + "learning_rate": 1.2605716437153554e-05, + "loss": 1.043, + "step": 35208 + }, + { + "epoch": 2.6, + "learning_rate": 1.2605331342674796e-05, + "loss": 0.9569, + "step": 35209 + }, + { + "epoch": 2.6, + "learning_rate": 1.2604946244050995e-05, + "loss": 1.1234, + "step": 35210 + }, + { + "epoch": 2.6, + "learning_rate": 1.260456114128276e-05, + "loss": 0.9781, + "step": 35211 + }, + { + "epoch": 2.6, + "learning_rate": 1.2604176034370708e-05, + "loss": 1.0039, + "step": 35212 + }, + { + "epoch": 2.6, + "learning_rate": 1.260379092331545e-05, + "loss": 0.9259, + "step": 35213 + }, + { + "epoch": 2.6, + "learning_rate": 1.2603405808117595e-05, + "loss": 0.8357, + "step": 35214 + }, + { + "epoch": 2.6, + "learning_rate": 1.260302068877776e-05, + "loss": 1.007, + "step": 35215 + }, + { + "epoch": 2.6, + "learning_rate": 1.2602635565296554e-05, + "loss": 1.0169, + "step": 35216 + }, + { + "epoch": 2.6, + "learning_rate": 1.2602250437674593e-05, + "loss": 0.9675, + "step": 35217 + }, + { + "epoch": 2.6, + "learning_rate": 1.2601865305912488e-05, + "loss": 0.9933, + "step": 35218 + }, + { + "epoch": 2.6, + "learning_rate": 1.2601480170010854e-05, + "loss": 1.0053, + "step": 35219 + }, + { + "epoch": 2.6, + "learning_rate": 1.2601095029970303e-05, + "loss": 0.9847, + "step": 35220 + }, + { + "epoch": 2.6, + "learning_rate": 1.2600709885791444e-05, + "loss": 0.9288, + "step": 35221 + }, + { + "epoch": 2.6, + "learning_rate": 1.2600324737474896e-05, + "loss": 0.8824, + "step": 35222 + }, + { + "epoch": 2.6, + "learning_rate": 1.2599939585021268e-05, + "loss": 1.0487, + "step": 35223 + }, + { + "epoch": 2.6, + "learning_rate": 1.2599554428431172e-05, + "loss": 0.9504, + "step": 35224 + }, + { + "epoch": 2.6, + "learning_rate": 1.2599169267705222e-05, + "loss": 1.0835, + "step": 35225 + }, + { + "epoch": 2.6, + "learning_rate": 1.2598784102844035e-05, + "loss": 1.0029, + "step": 35226 + }, + { + "epoch": 2.6, + "learning_rate": 1.2598398933848213e-05, + "loss": 1.0316, + "step": 35227 + }, + { + "epoch": 2.6, + "learning_rate": 1.2598013760718382e-05, + "loss": 0.9287, + "step": 35228 + }, + { + "epoch": 2.6, + "learning_rate": 1.2597628583455142e-05, + "loss": 1.0776, + "step": 35229 + }, + { + "epoch": 2.6, + "learning_rate": 1.2597243402059116e-05, + "loss": 1.0231, + "step": 35230 + }, + { + "epoch": 2.6, + "learning_rate": 1.2596858216530911e-05, + "loss": 0.947, + "step": 35231 + }, + { + "epoch": 2.6, + "learning_rate": 1.2596473026871143e-05, + "loss": 1.0188, + "step": 35232 + }, + { + "epoch": 2.6, + "learning_rate": 1.2596087833080426e-05, + "loss": 0.8882, + "step": 35233 + }, + { + "epoch": 2.6, + "learning_rate": 1.2595702635159366e-05, + "loss": 0.9817, + "step": 35234 + }, + { + "epoch": 2.6, + "learning_rate": 1.2595317433108582e-05, + "loss": 0.9434, + "step": 35235 + }, + { + "epoch": 2.6, + "learning_rate": 1.2594932226928684e-05, + "loss": 0.9562, + "step": 35236 + }, + { + "epoch": 2.6, + "learning_rate": 1.2594547016620289e-05, + "loss": 1.1065, + "step": 35237 + }, + { + "epoch": 2.6, + "learning_rate": 1.2594161802184003e-05, + "loss": 0.9712, + "step": 35238 + }, + { + "epoch": 2.6, + "learning_rate": 1.2593776583620445e-05, + "loss": 1.114, + "step": 35239 + }, + { + "epoch": 2.6, + "learning_rate": 1.2593391360930225e-05, + "loss": 1.0477, + "step": 35240 + }, + { + "epoch": 2.6, + "learning_rate": 1.2593006134113956e-05, + "loss": 0.9606, + "step": 35241 + }, + { + "epoch": 2.6, + "learning_rate": 1.2592620903172255e-05, + "loss": 1.0433, + "step": 35242 + }, + { + "epoch": 2.6, + "learning_rate": 1.2592235668105728e-05, + "loss": 1.0745, + "step": 35243 + }, + { + "epoch": 2.6, + "learning_rate": 1.259185042891499e-05, + "loss": 0.9797, + "step": 35244 + }, + { + "epoch": 2.6, + "learning_rate": 1.2591465185600657e-05, + "loss": 1.0487, + "step": 35245 + }, + { + "epoch": 2.6, + "learning_rate": 1.259107993816334e-05, + "loss": 1.0492, + "step": 35246 + }, + { + "epoch": 2.6, + "learning_rate": 1.259069468660365e-05, + "loss": 1.0477, + "step": 35247 + }, + { + "epoch": 2.6, + "learning_rate": 1.2590309430922203e-05, + "loss": 0.9863, + "step": 35248 + }, + { + "epoch": 2.6, + "learning_rate": 1.2589924171119612e-05, + "loss": 1.0255, + "step": 35249 + }, + { + "epoch": 2.6, + "learning_rate": 1.2589538907196486e-05, + "loss": 0.8824, + "step": 35250 + }, + { + "epoch": 2.6, + "learning_rate": 1.2589153639153445e-05, + "loss": 1.0316, + "step": 35251 + }, + { + "epoch": 2.6, + "learning_rate": 1.2588768366991093e-05, + "loss": 0.9062, + "step": 35252 + }, + { + "epoch": 2.6, + "learning_rate": 1.258838309071005e-05, + "loss": 1.0658, + "step": 35253 + }, + { + "epoch": 2.6, + "learning_rate": 1.2587997810310925e-05, + "loss": 1.0224, + "step": 35254 + }, + { + "epoch": 2.6, + "learning_rate": 1.2587612525794336e-05, + "loss": 1.0149, + "step": 35255 + }, + { + "epoch": 2.61, + "learning_rate": 1.2587227237160886e-05, + "loss": 0.945, + "step": 35256 + }, + { + "epoch": 2.61, + "learning_rate": 1.2586841944411199e-05, + "loss": 1.0407, + "step": 35257 + }, + { + "epoch": 2.61, + "learning_rate": 1.2586456647545884e-05, + "loss": 1.0824, + "step": 35258 + }, + { + "epoch": 2.61, + "learning_rate": 1.2586071346565552e-05, + "loss": 1.0486, + "step": 35259 + }, + { + "epoch": 2.61, + "learning_rate": 1.258568604147082e-05, + "loss": 1.0189, + "step": 35260 + }, + { + "epoch": 2.61, + "learning_rate": 1.2585300732262295e-05, + "loss": 1.0222, + "step": 35261 + }, + { + "epoch": 2.61, + "learning_rate": 1.2584915418940595e-05, + "loss": 0.992, + "step": 35262 + }, + { + "epoch": 2.61, + "learning_rate": 1.2584530101506331e-05, + "loss": 1.0122, + "step": 35263 + }, + { + "epoch": 2.61, + "learning_rate": 1.2584144779960117e-05, + "loss": 1.0012, + "step": 35264 + }, + { + "epoch": 2.61, + "learning_rate": 1.2583759454302566e-05, + "loss": 0.9368, + "step": 35265 + }, + { + "epoch": 2.61, + "learning_rate": 1.2583374124534288e-05, + "loss": 0.9837, + "step": 35266 + }, + { + "epoch": 2.61, + "learning_rate": 1.2582988790655903e-05, + "loss": 1.1665, + "step": 35267 + }, + { + "epoch": 2.61, + "learning_rate": 1.2582603452668016e-05, + "loss": 0.9024, + "step": 35268 + }, + { + "epoch": 2.61, + "learning_rate": 1.2582218110571248e-05, + "loss": 0.9592, + "step": 35269 + }, + { + "epoch": 2.61, + "learning_rate": 1.2581832764366206e-05, + "loss": 1.0432, + "step": 35270 + }, + { + "epoch": 2.61, + "learning_rate": 1.2581447414053504e-05, + "loss": 1.0565, + "step": 35271 + }, + { + "epoch": 2.61, + "learning_rate": 1.2581062059633755e-05, + "loss": 0.899, + "step": 35272 + }, + { + "epoch": 2.61, + "learning_rate": 1.2580676701107577e-05, + "loss": 1.0292, + "step": 35273 + }, + { + "epoch": 2.61, + "learning_rate": 1.2580291338475576e-05, + "loss": 1.0199, + "step": 35274 + }, + { + "epoch": 2.61, + "learning_rate": 1.257990597173837e-05, + "loss": 1.0522, + "step": 35275 + }, + { + "epoch": 2.61, + "learning_rate": 1.2579520600896569e-05, + "loss": 1.0639, + "step": 35276 + }, + { + "epoch": 2.61, + "learning_rate": 1.2579135225950789e-05, + "loss": 0.9552, + "step": 35277 + }, + { + "epoch": 2.61, + "learning_rate": 1.2578749846901641e-05, + "loss": 1.0286, + "step": 35278 + }, + { + "epoch": 2.61, + "learning_rate": 1.2578364463749742e-05, + "loss": 1.0279, + "step": 35279 + }, + { + "epoch": 2.61, + "learning_rate": 1.25779790764957e-05, + "loss": 1.0613, + "step": 35280 + }, + { + "epoch": 2.61, + "learning_rate": 1.2577593685140128e-05, + "loss": 1.0664, + "step": 35281 + }, + { + "epoch": 2.61, + "learning_rate": 1.2577208289683642e-05, + "loss": 1.0276, + "step": 35282 + }, + { + "epoch": 2.61, + "learning_rate": 1.2576822890126854e-05, + "loss": 0.9959, + "step": 35283 + }, + { + "epoch": 2.61, + "learning_rate": 1.2576437486470381e-05, + "loss": 0.9599, + "step": 35284 + }, + { + "epoch": 2.61, + "learning_rate": 1.2576052078714831e-05, + "loss": 1.071, + "step": 35285 + }, + { + "epoch": 2.61, + "learning_rate": 1.2575666666860817e-05, + "loss": 0.9518, + "step": 35286 + }, + { + "epoch": 2.61, + "learning_rate": 1.2575281250908958e-05, + "loss": 0.9631, + "step": 35287 + }, + { + "epoch": 2.61, + "learning_rate": 1.2574895830859863e-05, + "loss": 1.0692, + "step": 35288 + }, + { + "epoch": 2.61, + "learning_rate": 1.2574510406714142e-05, + "loss": 0.9121, + "step": 35289 + }, + { + "epoch": 2.61, + "learning_rate": 1.2574124978472416e-05, + "loss": 1.0621, + "step": 35290 + }, + { + "epoch": 2.61, + "learning_rate": 1.257373954613529e-05, + "loss": 1.043, + "step": 35291 + }, + { + "epoch": 2.61, + "learning_rate": 1.2573354109703384e-05, + "loss": 1.1222, + "step": 35292 + }, + { + "epoch": 2.61, + "learning_rate": 1.2572968669177306e-05, + "loss": 1.0223, + "step": 35293 + }, + { + "epoch": 2.61, + "learning_rate": 1.2572583224557673e-05, + "loss": 1.0636, + "step": 35294 + }, + { + "epoch": 2.61, + "learning_rate": 1.2572197775845096e-05, + "loss": 0.938, + "step": 35295 + }, + { + "epoch": 2.61, + "learning_rate": 1.257181232304019e-05, + "loss": 0.952, + "step": 35296 + }, + { + "epoch": 2.61, + "learning_rate": 1.2571426866143568e-05, + "loss": 1.1076, + "step": 35297 + }, + { + "epoch": 2.61, + "learning_rate": 1.2571041405155844e-05, + "loss": 0.8938, + "step": 35298 + }, + { + "epoch": 2.61, + "learning_rate": 1.2570655940077628e-05, + "loss": 1.0159, + "step": 35299 + }, + { + "epoch": 2.61, + "learning_rate": 1.2570270470909536e-05, + "loss": 0.9398, + "step": 35300 + }, + { + "epoch": 2.61, + "learning_rate": 1.2569884997652177e-05, + "loss": 1.0523, + "step": 35301 + }, + { + "epoch": 2.61, + "learning_rate": 1.2569499520306172e-05, + "loss": 0.9712, + "step": 35302 + }, + { + "epoch": 2.61, + "learning_rate": 1.2569114038872127e-05, + "loss": 0.9851, + "step": 35303 + }, + { + "epoch": 2.61, + "learning_rate": 1.256872855335066e-05, + "loss": 1.077, + "step": 35304 + }, + { + "epoch": 2.61, + "learning_rate": 1.256834306374238e-05, + "loss": 1.0312, + "step": 35305 + }, + { + "epoch": 2.61, + "learning_rate": 1.2567957570047908e-05, + "loss": 0.98, + "step": 35306 + }, + { + "epoch": 2.61, + "learning_rate": 1.2567572072267851e-05, + "loss": 0.9709, + "step": 35307 + }, + { + "epoch": 2.61, + "learning_rate": 1.2567186570402823e-05, + "loss": 1.0213, + "step": 35308 + }, + { + "epoch": 2.61, + "learning_rate": 1.2566801064453435e-05, + "loss": 1.0462, + "step": 35309 + }, + { + "epoch": 2.61, + "learning_rate": 1.2566415554420308e-05, + "loss": 1.0588, + "step": 35310 + }, + { + "epoch": 2.61, + "learning_rate": 1.2566030040304045e-05, + "loss": 0.9716, + "step": 35311 + }, + { + "epoch": 2.61, + "learning_rate": 1.2565644522105268e-05, + "loss": 1.1784, + "step": 35312 + }, + { + "epoch": 2.61, + "learning_rate": 1.2565258999824587e-05, + "loss": 0.9704, + "step": 35313 + }, + { + "epoch": 2.61, + "learning_rate": 1.2564873473462615e-05, + "loss": 1.0167, + "step": 35314 + }, + { + "epoch": 2.61, + "learning_rate": 1.2564487943019967e-05, + "loss": 1.0524, + "step": 35315 + }, + { + "epoch": 2.61, + "learning_rate": 1.2564102408497259e-05, + "loss": 0.9298, + "step": 35316 + }, + { + "epoch": 2.61, + "learning_rate": 1.2563716869895095e-05, + "loss": 0.9563, + "step": 35317 + }, + { + "epoch": 2.61, + "learning_rate": 1.2563331327214096e-05, + "loss": 0.9447, + "step": 35318 + }, + { + "epoch": 2.61, + "learning_rate": 1.2562945780454871e-05, + "loss": 0.9632, + "step": 35319 + }, + { + "epoch": 2.61, + "learning_rate": 1.2562560229618041e-05, + "loss": 1.0288, + "step": 35320 + }, + { + "epoch": 2.61, + "learning_rate": 1.2562174674704211e-05, + "loss": 0.9247, + "step": 35321 + }, + { + "epoch": 2.61, + "learning_rate": 1.2561789115713998e-05, + "loss": 0.9794, + "step": 35322 + }, + { + "epoch": 2.61, + "learning_rate": 1.2561403552648015e-05, + "loss": 0.8924, + "step": 35323 + }, + { + "epoch": 2.61, + "learning_rate": 1.2561017985506876e-05, + "loss": 0.9733, + "step": 35324 + }, + { + "epoch": 2.61, + "learning_rate": 1.2560632414291194e-05, + "loss": 0.9833, + "step": 35325 + }, + { + "epoch": 2.61, + "learning_rate": 1.2560246839001584e-05, + "loss": 0.9997, + "step": 35326 + }, + { + "epoch": 2.61, + "learning_rate": 1.2559861259638658e-05, + "loss": 1.0435, + "step": 35327 + }, + { + "epoch": 2.61, + "learning_rate": 1.2559475676203026e-05, + "loss": 1.0027, + "step": 35328 + }, + { + "epoch": 2.61, + "learning_rate": 1.2559090088695308e-05, + "loss": 0.9161, + "step": 35329 + }, + { + "epoch": 2.61, + "learning_rate": 1.2558704497116112e-05, + "loss": 1.0215, + "step": 35330 + }, + { + "epoch": 2.61, + "learning_rate": 1.2558318901466055e-05, + "loss": 0.9473, + "step": 35331 + }, + { + "epoch": 2.61, + "learning_rate": 1.255793330174575e-05, + "loss": 0.9316, + "step": 35332 + }, + { + "epoch": 2.61, + "learning_rate": 1.2557547697955807e-05, + "loss": 1.0501, + "step": 35333 + }, + { + "epoch": 2.61, + "learning_rate": 1.2557162090096847e-05, + "loss": 1.061, + "step": 35334 + }, + { + "epoch": 2.61, + "learning_rate": 1.2556776478169476e-05, + "loss": 0.9501, + "step": 35335 + }, + { + "epoch": 2.61, + "learning_rate": 1.2556390862174308e-05, + "loss": 0.9888, + "step": 35336 + }, + { + "epoch": 2.61, + "learning_rate": 1.2556005242111962e-05, + "loss": 1.0119, + "step": 35337 + }, + { + "epoch": 2.61, + "learning_rate": 1.2555619617983045e-05, + "loss": 0.9685, + "step": 35338 + }, + { + "epoch": 2.61, + "learning_rate": 1.2555233989788179e-05, + "loss": 1.0254, + "step": 35339 + }, + { + "epoch": 2.61, + "learning_rate": 1.2554848357527964e-05, + "loss": 0.9628, + "step": 35340 + }, + { + "epoch": 2.61, + "learning_rate": 1.2554462721203029e-05, + "loss": 0.9969, + "step": 35341 + }, + { + "epoch": 2.61, + "learning_rate": 1.2554077080813978e-05, + "loss": 0.9313, + "step": 35342 + }, + { + "epoch": 2.61, + "learning_rate": 1.2553691436361425e-05, + "loss": 1.0504, + "step": 35343 + }, + { + "epoch": 2.61, + "learning_rate": 1.255330578784599e-05, + "loss": 0.9585, + "step": 35344 + }, + { + "epoch": 2.61, + "learning_rate": 1.2552920135268281e-05, + "loss": 1.0437, + "step": 35345 + }, + { + "epoch": 2.61, + "learning_rate": 1.255253447862891e-05, + "loss": 0.9583, + "step": 35346 + }, + { + "epoch": 2.61, + "learning_rate": 1.2552148817928496e-05, + "loss": 0.8766, + "step": 35347 + }, + { + "epoch": 2.61, + "learning_rate": 1.2551763153167644e-05, + "loss": 0.9385, + "step": 35348 + }, + { + "epoch": 2.61, + "learning_rate": 1.255137748434698e-05, + "loss": 1.002, + "step": 35349 + }, + { + "epoch": 2.61, + "learning_rate": 1.2550991811467105e-05, + "loss": 1.0113, + "step": 35350 + }, + { + "epoch": 2.61, + "learning_rate": 1.2550606134528645e-05, + "loss": 1.0529, + "step": 35351 + }, + { + "epoch": 2.61, + "learning_rate": 1.2550220453532203e-05, + "loss": 1.0606, + "step": 35352 + }, + { + "epoch": 2.61, + "learning_rate": 1.2549834768478398e-05, + "loss": 1.1346, + "step": 35353 + }, + { + "epoch": 2.61, + "learning_rate": 1.2549449079367842e-05, + "loss": 1.038, + "step": 35354 + }, + { + "epoch": 2.61, + "learning_rate": 1.254906338620115e-05, + "loss": 1.1136, + "step": 35355 + }, + { + "epoch": 2.61, + "learning_rate": 1.2548677688978931e-05, + "loss": 1.0512, + "step": 35356 + }, + { + "epoch": 2.61, + "learning_rate": 1.2548291987701808e-05, + "loss": 1.1751, + "step": 35357 + }, + { + "epoch": 2.61, + "learning_rate": 1.2547906282370383e-05, + "loss": 1.0273, + "step": 35358 + }, + { + "epoch": 2.61, + "learning_rate": 1.254752057298528e-05, + "loss": 1.0815, + "step": 35359 + }, + { + "epoch": 2.61, + "learning_rate": 1.2547134859547105e-05, + "loss": 0.9903, + "step": 35360 + }, + { + "epoch": 2.61, + "learning_rate": 1.2546749142056479e-05, + "loss": 1.0548, + "step": 35361 + }, + { + "epoch": 2.61, + "learning_rate": 1.2546363420514007e-05, + "loss": 1.0578, + "step": 35362 + }, + { + "epoch": 2.61, + "learning_rate": 1.2545977694920313e-05, + "loss": 0.9356, + "step": 35363 + }, + { + "epoch": 2.61, + "learning_rate": 1.2545591965275998e-05, + "loss": 1.0515, + "step": 35364 + }, + { + "epoch": 2.61, + "learning_rate": 1.2545206231581689e-05, + "loss": 0.8829, + "step": 35365 + }, + { + "epoch": 2.61, + "learning_rate": 1.2544820493837988e-05, + "loss": 1.0437, + "step": 35366 + }, + { + "epoch": 2.61, + "learning_rate": 1.2544434752045515e-05, + "loss": 0.9814, + "step": 35367 + }, + { + "epoch": 2.61, + "learning_rate": 1.2544049006204883e-05, + "loss": 0.9313, + "step": 35368 + }, + { + "epoch": 2.61, + "learning_rate": 1.2543663256316707e-05, + "loss": 0.984, + "step": 35369 + }, + { + "epoch": 2.61, + "learning_rate": 1.2543277502381598e-05, + "loss": 1.0258, + "step": 35370 + }, + { + "epoch": 2.61, + "learning_rate": 1.2542891744400173e-05, + "loss": 1.0369, + "step": 35371 + }, + { + "epoch": 2.61, + "learning_rate": 1.2542505982373041e-05, + "loss": 0.9547, + "step": 35372 + }, + { + "epoch": 2.61, + "learning_rate": 1.254212021630082e-05, + "loss": 1.0664, + "step": 35373 + }, + { + "epoch": 2.61, + "learning_rate": 1.2541734446184122e-05, + "loss": 1.035, + "step": 35374 + }, + { + "epoch": 2.61, + "learning_rate": 1.254134867202356e-05, + "loss": 1.001, + "step": 35375 + }, + { + "epoch": 2.61, + "learning_rate": 1.2540962893819747e-05, + "loss": 1.0188, + "step": 35376 + }, + { + "epoch": 2.61, + "learning_rate": 1.2540577111573303e-05, + "loss": 1.046, + "step": 35377 + }, + { + "epoch": 2.61, + "learning_rate": 1.2540191325284832e-05, + "loss": 1.0604, + "step": 35378 + }, + { + "epoch": 2.61, + "learning_rate": 1.2539805534954956e-05, + "loss": 1.1036, + "step": 35379 + }, + { + "epoch": 2.61, + "learning_rate": 1.2539419740584285e-05, + "loss": 1.0583, + "step": 35380 + }, + { + "epoch": 2.61, + "learning_rate": 1.2539033942173433e-05, + "loss": 0.9848, + "step": 35381 + }, + { + "epoch": 2.61, + "learning_rate": 1.2538648139723018e-05, + "loss": 1.0499, + "step": 35382 + }, + { + "epoch": 2.61, + "learning_rate": 1.2538262333233643e-05, + "loss": 1.0045, + "step": 35383 + }, + { + "epoch": 2.61, + "learning_rate": 1.2537876522705936e-05, + "loss": 1.1046, + "step": 35384 + }, + { + "epoch": 2.61, + "learning_rate": 1.2537490708140499e-05, + "loss": 0.9447, + "step": 35385 + }, + { + "epoch": 2.61, + "learning_rate": 1.2537104889537952e-05, + "loss": 1.0311, + "step": 35386 + }, + { + "epoch": 2.61, + "learning_rate": 1.2536719066898906e-05, + "loss": 1.0937, + "step": 35387 + }, + { + "epoch": 2.61, + "learning_rate": 1.2536333240223981e-05, + "loss": 0.9307, + "step": 35388 + }, + { + "epoch": 2.61, + "learning_rate": 1.2535947409513782e-05, + "loss": 1.0045, + "step": 35389 + }, + { + "epoch": 2.61, + "learning_rate": 1.2535561574768929e-05, + "loss": 1.0385, + "step": 35390 + }, + { + "epoch": 2.62, + "learning_rate": 1.2535175735990033e-05, + "loss": 1.0065, + "step": 35391 + }, + { + "epoch": 2.62, + "learning_rate": 1.2534789893177709e-05, + "loss": 1.0135, + "step": 35392 + }, + { + "epoch": 2.62, + "learning_rate": 1.2534404046332569e-05, + "loss": 0.986, + "step": 35393 + }, + { + "epoch": 2.62, + "learning_rate": 1.2534018195455231e-05, + "loss": 1.0175, + "step": 35394 + }, + { + "epoch": 2.62, + "learning_rate": 1.2533632340546302e-05, + "loss": 1.0518, + "step": 35395 + }, + { + "epoch": 2.62, + "learning_rate": 1.2533246481606405e-05, + "loss": 0.9889, + "step": 35396 + }, + { + "epoch": 2.62, + "learning_rate": 1.2532860618636145e-05, + "loss": 1.0695, + "step": 35397 + }, + { + "epoch": 2.62, + "learning_rate": 1.2532474751636144e-05, + "loss": 0.8926, + "step": 35398 + }, + { + "epoch": 2.62, + "learning_rate": 1.253208888060701e-05, + "loss": 1.0221, + "step": 35399 + }, + { + "epoch": 2.62, + "learning_rate": 1.253170300554936e-05, + "loss": 0.98, + "step": 35400 + }, + { + "epoch": 2.62, + "learning_rate": 1.2531317126463804e-05, + "loss": 0.9615, + "step": 35401 + }, + { + "epoch": 2.62, + "learning_rate": 1.2530931243350964e-05, + "loss": 1.0003, + "step": 35402 + }, + { + "epoch": 2.62, + "learning_rate": 1.2530545356211442e-05, + "loss": 1.0225, + "step": 35403 + }, + { + "epoch": 2.62, + "learning_rate": 1.253015946504586e-05, + "loss": 0.9649, + "step": 35404 + }, + { + "epoch": 2.62, + "learning_rate": 1.2529773569854833e-05, + "loss": 1.0152, + "step": 35405 + }, + { + "epoch": 2.62, + "learning_rate": 1.2529387670638973e-05, + "loss": 0.9386, + "step": 35406 + }, + { + "epoch": 2.62, + "learning_rate": 1.252900176739889e-05, + "loss": 1.0872, + "step": 35407 + }, + { + "epoch": 2.62, + "learning_rate": 1.2528615860135202e-05, + "loss": 1.0247, + "step": 35408 + }, + { + "epoch": 2.62, + "learning_rate": 1.2528229948848524e-05, + "loss": 1.041, + "step": 35409 + }, + { + "epoch": 2.62, + "learning_rate": 1.252784403353947e-05, + "loss": 1.0198, + "step": 35410 + }, + { + "epoch": 2.62, + "learning_rate": 1.2527458114208649e-05, + "loss": 0.894, + "step": 35411 + }, + { + "epoch": 2.62, + "learning_rate": 1.252707219085668e-05, + "loss": 0.9167, + "step": 35412 + }, + { + "epoch": 2.62, + "learning_rate": 1.2526686263484174e-05, + "loss": 1.0698, + "step": 35413 + }, + { + "epoch": 2.62, + "learning_rate": 1.2526300332091748e-05, + "loss": 0.9754, + "step": 35414 + }, + { + "epoch": 2.62, + "learning_rate": 1.2525914396680012e-05, + "loss": 0.9718, + "step": 35415 + }, + { + "epoch": 2.62, + "learning_rate": 1.2525528457249585e-05, + "loss": 0.9502, + "step": 35416 + }, + { + "epoch": 2.62, + "learning_rate": 1.2525142513801076e-05, + "loss": 1.0846, + "step": 35417 + }, + { + "epoch": 2.62, + "learning_rate": 1.2524756566335104e-05, + "loss": 1.0836, + "step": 35418 + }, + { + "epoch": 2.62, + "learning_rate": 1.2524370614852276e-05, + "loss": 0.9523, + "step": 35419 + }, + { + "epoch": 2.62, + "learning_rate": 1.2523984659353216e-05, + "loss": 0.945, + "step": 35420 + }, + { + "epoch": 2.62, + "learning_rate": 1.2523598699838529e-05, + "loss": 0.988, + "step": 35421 + }, + { + "epoch": 2.62, + "learning_rate": 1.2523212736308834e-05, + "loss": 0.9222, + "step": 35422 + }, + { + "epoch": 2.62, + "learning_rate": 1.2522826768764742e-05, + "loss": 0.969, + "step": 35423 + }, + { + "epoch": 2.62, + "learning_rate": 1.2522440797206871e-05, + "loss": 0.9288, + "step": 35424 + }, + { + "epoch": 2.62, + "learning_rate": 1.2522054821635829e-05, + "loss": 1.0707, + "step": 35425 + }, + { + "epoch": 2.62, + "learning_rate": 1.2521668842052238e-05, + "loss": 1.0717, + "step": 35426 + }, + { + "epoch": 2.62, + "learning_rate": 1.2521282858456705e-05, + "loss": 0.9881, + "step": 35427 + }, + { + "epoch": 2.62, + "learning_rate": 1.252089687084985e-05, + "loss": 1.0416, + "step": 35428 + }, + { + "epoch": 2.62, + "learning_rate": 1.2520510879232284e-05, + "loss": 1.028, + "step": 35429 + }, + { + "epoch": 2.62, + "learning_rate": 1.2520124883604616e-05, + "loss": 0.9025, + "step": 35430 + }, + { + "epoch": 2.62, + "learning_rate": 1.2519738883967472e-05, + "loss": 0.8978, + "step": 35431 + }, + { + "epoch": 2.62, + "learning_rate": 1.2519352880321453e-05, + "loss": 0.9448, + "step": 35432 + }, + { + "epoch": 2.62, + "learning_rate": 1.2518966872667187e-05, + "loss": 1.0619, + "step": 35433 + }, + { + "epoch": 2.62, + "learning_rate": 1.2518580861005277e-05, + "loss": 1.0072, + "step": 35434 + }, + { + "epoch": 2.62, + "learning_rate": 1.251819484533634e-05, + "loss": 0.9251, + "step": 35435 + }, + { + "epoch": 2.62, + "learning_rate": 1.251780882566099e-05, + "loss": 1.0184, + "step": 35436 + }, + { + "epoch": 2.62, + "learning_rate": 1.2517422801979846e-05, + "loss": 1.0972, + "step": 35437 + }, + { + "epoch": 2.62, + "learning_rate": 1.2517036774293516e-05, + "loss": 0.9659, + "step": 35438 + }, + { + "epoch": 2.62, + "learning_rate": 1.251665074260262e-05, + "loss": 0.9764, + "step": 35439 + }, + { + "epoch": 2.62, + "learning_rate": 1.2516264706907762e-05, + "loss": 0.8839, + "step": 35440 + }, + { + "epoch": 2.62, + "learning_rate": 1.2515878667209568e-05, + "loss": 1.0591, + "step": 35441 + }, + { + "epoch": 2.62, + "learning_rate": 1.2515492623508646e-05, + "loss": 1.0633, + "step": 35442 + }, + { + "epoch": 2.62, + "learning_rate": 1.251510657580561e-05, + "loss": 0.9977, + "step": 35443 + }, + { + "epoch": 2.62, + "learning_rate": 1.2514720524101076e-05, + "loss": 0.991, + "step": 35444 + }, + { + "epoch": 2.62, + "learning_rate": 1.2514334468395657e-05, + "loss": 0.9724, + "step": 35445 + }, + { + "epoch": 2.62, + "learning_rate": 1.2513948408689968e-05, + "loss": 1.0269, + "step": 35446 + }, + { + "epoch": 2.62, + "learning_rate": 1.2513562344984626e-05, + "loss": 1.0046, + "step": 35447 + }, + { + "epoch": 2.62, + "learning_rate": 1.2513176277280238e-05, + "loss": 1.0072, + "step": 35448 + }, + { + "epoch": 2.62, + "learning_rate": 1.2512790205577428e-05, + "loss": 1.0413, + "step": 35449 + }, + { + "epoch": 2.62, + "learning_rate": 1.2512404129876798e-05, + "loss": 1.0525, + "step": 35450 + }, + { + "epoch": 2.62, + "learning_rate": 1.2512018050178973e-05, + "loss": 1.0008, + "step": 35451 + }, + { + "epoch": 2.62, + "learning_rate": 1.2511631966484563e-05, + "loss": 0.9127, + "step": 35452 + }, + { + "epoch": 2.62, + "learning_rate": 1.251124587879418e-05, + "loss": 1.1317, + "step": 35453 + }, + { + "epoch": 2.62, + "learning_rate": 1.2510859787108443e-05, + "loss": 1.049, + "step": 35454 + }, + { + "epoch": 2.62, + "learning_rate": 1.2510473691427965e-05, + "loss": 0.9981, + "step": 35455 + }, + { + "epoch": 2.62, + "learning_rate": 1.2510087591753356e-05, + "loss": 0.8891, + "step": 35456 + }, + { + "epoch": 2.62, + "learning_rate": 1.250970148808524e-05, + "loss": 1.0226, + "step": 35457 + }, + { + "epoch": 2.62, + "learning_rate": 1.2509315380424218e-05, + "loss": 1.011, + "step": 35458 + }, + { + "epoch": 2.62, + "learning_rate": 1.2508929268770912e-05, + "loss": 0.9958, + "step": 35459 + }, + { + "epoch": 2.62, + "learning_rate": 1.2508543153125938e-05, + "loss": 0.9884, + "step": 35460 + }, + { + "epoch": 2.62, + "learning_rate": 1.2508157033489908e-05, + "loss": 0.9552, + "step": 35461 + }, + { + "epoch": 2.62, + "learning_rate": 1.2507770909863433e-05, + "loss": 0.9789, + "step": 35462 + }, + { + "epoch": 2.62, + "learning_rate": 1.2507384782247132e-05, + "loss": 0.9816, + "step": 35463 + }, + { + "epoch": 2.62, + "learning_rate": 1.2506998650641618e-05, + "loss": 1.0037, + "step": 35464 + }, + { + "epoch": 2.62, + "learning_rate": 1.2506612515047505e-05, + "loss": 1.0521, + "step": 35465 + }, + { + "epoch": 2.62, + "learning_rate": 1.2506226375465406e-05, + "loss": 0.9035, + "step": 35466 + }, + { + "epoch": 2.62, + "learning_rate": 1.2505840231895938e-05, + "loss": 0.982, + "step": 35467 + }, + { + "epoch": 2.62, + "learning_rate": 1.2505454084339714e-05, + "loss": 1.0453, + "step": 35468 + }, + { + "epoch": 2.62, + "learning_rate": 1.250506793279735e-05, + "loss": 1.0732, + "step": 35469 + }, + { + "epoch": 2.62, + "learning_rate": 1.2504681777269454e-05, + "loss": 0.9267, + "step": 35470 + }, + { + "epoch": 2.62, + "learning_rate": 1.250429561775665e-05, + "loss": 1.0662, + "step": 35471 + }, + { + "epoch": 2.62, + "learning_rate": 1.2503909454259543e-05, + "loss": 1.0166, + "step": 35472 + }, + { + "epoch": 2.62, + "learning_rate": 1.2503523286778755e-05, + "loss": 0.9911, + "step": 35473 + }, + { + "epoch": 2.62, + "learning_rate": 1.2503137115314896e-05, + "loss": 0.9489, + "step": 35474 + }, + { + "epoch": 2.62, + "learning_rate": 1.2502750939868585e-05, + "loss": 1.0721, + "step": 35475 + }, + { + "epoch": 2.62, + "learning_rate": 1.2502364760440431e-05, + "loss": 1.0102, + "step": 35476 + }, + { + "epoch": 2.62, + "learning_rate": 1.250197857703105e-05, + "loss": 1.0501, + "step": 35477 + }, + { + "epoch": 2.62, + "learning_rate": 1.2501592389641055e-05, + "loss": 1.0512, + "step": 35478 + }, + { + "epoch": 2.62, + "learning_rate": 1.2501206198271063e-05, + "loss": 0.9829, + "step": 35479 + }, + { + "epoch": 2.62, + "learning_rate": 1.250082000292169e-05, + "loss": 1.06, + "step": 35480 + }, + { + "epoch": 2.62, + "learning_rate": 1.2500433803593549e-05, + "loss": 1.0855, + "step": 35481 + }, + { + "epoch": 2.62, + "learning_rate": 1.2500047600287251e-05, + "loss": 0.9334, + "step": 35482 + }, + { + "epoch": 2.62, + "learning_rate": 1.2499661393003413e-05, + "loss": 0.9805, + "step": 35483 + }, + { + "epoch": 2.62, + "learning_rate": 1.249927518174265e-05, + "loss": 1.1139, + "step": 35484 + }, + { + "epoch": 2.62, + "learning_rate": 1.2498888966505576e-05, + "loss": 1.0072, + "step": 35485 + }, + { + "epoch": 2.62, + "learning_rate": 1.2498502747292806e-05, + "loss": 0.9971, + "step": 35486 + }, + { + "epoch": 2.62, + "learning_rate": 1.2498116524104953e-05, + "loss": 0.965, + "step": 35487 + }, + { + "epoch": 2.62, + "learning_rate": 1.2497730296942632e-05, + "loss": 1.0198, + "step": 35488 + }, + { + "epoch": 2.62, + "learning_rate": 1.2497344065806457e-05, + "loss": 1.0513, + "step": 35489 + }, + { + "epoch": 2.62, + "learning_rate": 1.2496957830697048e-05, + "loss": 0.9642, + "step": 35490 + }, + { + "epoch": 2.62, + "learning_rate": 1.249657159161501e-05, + "loss": 0.9937, + "step": 35491 + }, + { + "epoch": 2.62, + "learning_rate": 1.2496185348560962e-05, + "loss": 0.9954, + "step": 35492 + }, + { + "epoch": 2.62, + "learning_rate": 1.2495799101535522e-05, + "loss": 0.9823, + "step": 35493 + }, + { + "epoch": 2.62, + "learning_rate": 1.2495412850539301e-05, + "loss": 1.0116, + "step": 35494 + }, + { + "epoch": 2.62, + "learning_rate": 1.2495026595572912e-05, + "loss": 0.9979, + "step": 35495 + }, + { + "epoch": 2.62, + "learning_rate": 1.2494640336636973e-05, + "loss": 1.0787, + "step": 35496 + }, + { + "epoch": 2.62, + "learning_rate": 1.2494254073732098e-05, + "loss": 0.9313, + "step": 35497 + }, + { + "epoch": 2.62, + "learning_rate": 1.2493867806858897e-05, + "loss": 0.979, + "step": 35498 + }, + { + "epoch": 2.62, + "learning_rate": 1.2493481536017988e-05, + "loss": 0.9936, + "step": 35499 + }, + { + "epoch": 2.62, + "learning_rate": 1.249309526120999e-05, + "loss": 0.9756, + "step": 35500 + }, + { + "epoch": 2.62, + "learning_rate": 1.249270898243551e-05, + "loss": 0.9701, + "step": 35501 + }, + { + "epoch": 2.62, + "learning_rate": 1.2492322699695165e-05, + "loss": 1.0455, + "step": 35502 + }, + { + "epoch": 2.62, + "learning_rate": 1.2491936412989573e-05, + "loss": 1.0908, + "step": 35503 + }, + { + "epoch": 2.62, + "learning_rate": 1.2491550122319344e-05, + "loss": 1.0525, + "step": 35504 + }, + { + "epoch": 2.62, + "learning_rate": 1.2491163827685094e-05, + "loss": 0.8352, + "step": 35505 + }, + { + "epoch": 2.62, + "learning_rate": 1.249077752908744e-05, + "loss": 0.9797, + "step": 35506 + }, + { + "epoch": 2.62, + "learning_rate": 1.2490391226526992e-05, + "loss": 1.0303, + "step": 35507 + }, + { + "epoch": 2.62, + "learning_rate": 1.2490004920004368e-05, + "loss": 0.9094, + "step": 35508 + }, + { + "epoch": 2.62, + "learning_rate": 1.2489618609520181e-05, + "loss": 0.9207, + "step": 35509 + }, + { + "epoch": 2.62, + "learning_rate": 1.2489232295075049e-05, + "loss": 0.9785, + "step": 35510 + }, + { + "epoch": 2.62, + "learning_rate": 1.2488845976669581e-05, + "loss": 0.9865, + "step": 35511 + }, + { + "epoch": 2.62, + "learning_rate": 1.2488459654304397e-05, + "loss": 1.0409, + "step": 35512 + }, + { + "epoch": 2.62, + "learning_rate": 1.248807332798011e-05, + "loss": 1.0744, + "step": 35513 + }, + { + "epoch": 2.62, + "learning_rate": 1.2487686997697333e-05, + "loss": 1.0982, + "step": 35514 + }, + { + "epoch": 2.62, + "learning_rate": 1.248730066345668e-05, + "loss": 1.1074, + "step": 35515 + }, + { + "epoch": 2.62, + "learning_rate": 1.248691432525877e-05, + "loss": 0.9205, + "step": 35516 + }, + { + "epoch": 2.62, + "learning_rate": 1.2486527983104212e-05, + "loss": 0.9957, + "step": 35517 + }, + { + "epoch": 2.62, + "learning_rate": 1.2486141636993624e-05, + "loss": 1.0115, + "step": 35518 + }, + { + "epoch": 2.62, + "learning_rate": 1.248575528692762e-05, + "loss": 0.9607, + "step": 35519 + }, + { + "epoch": 2.62, + "learning_rate": 1.2485368932906816e-05, + "loss": 1.0159, + "step": 35520 + }, + { + "epoch": 2.62, + "learning_rate": 1.2484982574931825e-05, + "loss": 1.1008, + "step": 35521 + }, + { + "epoch": 2.62, + "learning_rate": 1.2484596213003263e-05, + "loss": 0.8997, + "step": 35522 + }, + { + "epoch": 2.62, + "learning_rate": 1.2484209847121745e-05, + "loss": 0.9519, + "step": 35523 + }, + { + "epoch": 2.62, + "learning_rate": 1.2483823477287882e-05, + "loss": 1.0195, + "step": 35524 + }, + { + "epoch": 2.62, + "learning_rate": 1.2483437103502294e-05, + "loss": 0.953, + "step": 35525 + }, + { + "epoch": 2.62, + "learning_rate": 1.2483050725765592e-05, + "loss": 0.9886, + "step": 35526 + }, + { + "epoch": 2.63, + "learning_rate": 1.2482664344078393e-05, + "loss": 1.0237, + "step": 35527 + }, + { + "epoch": 2.63, + "learning_rate": 1.2482277958441308e-05, + "loss": 1.0551, + "step": 35528 + }, + { + "epoch": 2.63, + "learning_rate": 1.2481891568854954e-05, + "loss": 1.0858, + "step": 35529 + }, + { + "epoch": 2.63, + "learning_rate": 1.248150517531995e-05, + "loss": 1.0573, + "step": 35530 + }, + { + "epoch": 2.63, + "learning_rate": 1.2481118777836905e-05, + "loss": 1.0339, + "step": 35531 + }, + { + "epoch": 2.63, + "learning_rate": 1.2480732376406432e-05, + "loss": 0.9323, + "step": 35532 + }, + { + "epoch": 2.63, + "learning_rate": 1.2480345971029153e-05, + "loss": 0.9016, + "step": 35533 + }, + { + "epoch": 2.63, + "learning_rate": 1.2479959561705676e-05, + "loss": 1.0092, + "step": 35534 + }, + { + "epoch": 2.63, + "learning_rate": 1.2479573148436623e-05, + "loss": 0.9299, + "step": 35535 + }, + { + "epoch": 2.63, + "learning_rate": 1.24791867312226e-05, + "loss": 1.0257, + "step": 35536 + }, + { + "epoch": 2.63, + "learning_rate": 1.2478800310064232e-05, + "loss": 1.0044, + "step": 35537 + }, + { + "epoch": 2.63, + "learning_rate": 1.2478413884962123e-05, + "loss": 1.0379, + "step": 35538 + }, + { + "epoch": 2.63, + "learning_rate": 1.2478027455916897e-05, + "loss": 0.9924, + "step": 35539 + }, + { + "epoch": 2.63, + "learning_rate": 1.2477641022929161e-05, + "loss": 0.9948, + "step": 35540 + }, + { + "epoch": 2.63, + "learning_rate": 1.2477254585999539e-05, + "loss": 0.9591, + "step": 35541 + }, + { + "epoch": 2.63, + "learning_rate": 1.2476868145128636e-05, + "loss": 1.0324, + "step": 35542 + }, + { + "epoch": 2.63, + "learning_rate": 1.247648170031707e-05, + "loss": 0.9965, + "step": 35543 + }, + { + "epoch": 2.63, + "learning_rate": 1.247609525156546e-05, + "loss": 0.9688, + "step": 35544 + }, + { + "epoch": 2.63, + "learning_rate": 1.2475708798874418e-05, + "loss": 1.0276, + "step": 35545 + }, + { + "epoch": 2.63, + "learning_rate": 1.2475322342244556e-05, + "loss": 0.9008, + "step": 35546 + }, + { + "epoch": 2.63, + "learning_rate": 1.2474935881676494e-05, + "loss": 0.9241, + "step": 35547 + }, + { + "epoch": 2.63, + "learning_rate": 1.2474549417170843e-05, + "loss": 0.9109, + "step": 35548 + }, + { + "epoch": 2.63, + "learning_rate": 1.2474162948728222e-05, + "loss": 1.0067, + "step": 35549 + }, + { + "epoch": 2.63, + "learning_rate": 1.2473776476349239e-05, + "loss": 1.0094, + "step": 35550 + }, + { + "epoch": 2.63, + "learning_rate": 1.2473390000034517e-05, + "loss": 1.0132, + "step": 35551 + }, + { + "epoch": 2.63, + "learning_rate": 1.2473003519784665e-05, + "loss": 0.9519, + "step": 35552 + }, + { + "epoch": 2.63, + "learning_rate": 1.2472617035600301e-05, + "loss": 1.0371, + "step": 35553 + }, + { + "epoch": 2.63, + "learning_rate": 1.2472230547482035e-05, + "loss": 1.0674, + "step": 35554 + }, + { + "epoch": 2.63, + "learning_rate": 1.2471844055430488e-05, + "loss": 0.9866, + "step": 35555 + }, + { + "epoch": 2.63, + "learning_rate": 1.2471457559446271e-05, + "loss": 1.0344, + "step": 35556 + }, + { + "epoch": 2.63, + "learning_rate": 1.2471071059530005e-05, + "loss": 1.017, + "step": 35557 + }, + { + "epoch": 2.63, + "learning_rate": 1.2470684555682296e-05, + "loss": 1.0415, + "step": 35558 + }, + { + "epoch": 2.63, + "learning_rate": 1.2470298047903766e-05, + "loss": 0.9039, + "step": 35559 + }, + { + "epoch": 2.63, + "learning_rate": 1.2469911536195024e-05, + "loss": 1.1728, + "step": 35560 + }, + { + "epoch": 2.63, + "learning_rate": 1.2469525020556692e-05, + "loss": 1.0223, + "step": 35561 + }, + { + "epoch": 2.63, + "learning_rate": 1.2469138500989378e-05, + "loss": 0.917, + "step": 35562 + }, + { + "epoch": 2.63, + "learning_rate": 1.24687519774937e-05, + "loss": 0.9903, + "step": 35563 + }, + { + "epoch": 2.63, + "learning_rate": 1.2468365450070272e-05, + "loss": 0.976, + "step": 35564 + }, + { + "epoch": 2.63, + "learning_rate": 1.2467978918719714e-05, + "loss": 1.0334, + "step": 35565 + }, + { + "epoch": 2.63, + "learning_rate": 1.2467592383442632e-05, + "loss": 0.8905, + "step": 35566 + }, + { + "epoch": 2.63, + "learning_rate": 1.246720584423965e-05, + "loss": 1.0171, + "step": 35567 + }, + { + "epoch": 2.63, + "learning_rate": 1.2466819301111374e-05, + "loss": 1.0364, + "step": 35568 + }, + { + "epoch": 2.63, + "learning_rate": 1.246643275405843e-05, + "loss": 0.9805, + "step": 35569 + }, + { + "epoch": 2.63, + "learning_rate": 1.2466046203081424e-05, + "loss": 0.9603, + "step": 35570 + }, + { + "epoch": 2.63, + "learning_rate": 1.2465659648180975e-05, + "loss": 1.0707, + "step": 35571 + }, + { + "epoch": 2.63, + "learning_rate": 1.2465273089357692e-05, + "loss": 1.1251, + "step": 35572 + }, + { + "epoch": 2.63, + "learning_rate": 1.24648865266122e-05, + "loss": 1.0215, + "step": 35573 + }, + { + "epoch": 2.63, + "learning_rate": 1.2464499959945105e-05, + "loss": 1.0174, + "step": 35574 + }, + { + "epoch": 2.63, + "learning_rate": 1.246411338935703e-05, + "loss": 0.9779, + "step": 35575 + }, + { + "epoch": 2.63, + "learning_rate": 1.246372681484858e-05, + "loss": 0.9257, + "step": 35576 + }, + { + "epoch": 2.63, + "learning_rate": 1.2463340236420381e-05, + "loss": 1.0531, + "step": 35577 + }, + { + "epoch": 2.63, + "learning_rate": 1.2462953654073042e-05, + "loss": 1.026, + "step": 35578 + }, + { + "epoch": 2.63, + "learning_rate": 1.2462567067807176e-05, + "loss": 0.9578, + "step": 35579 + }, + { + "epoch": 2.63, + "learning_rate": 1.2462180477623404e-05, + "loss": 1.0057, + "step": 35580 + }, + { + "epoch": 2.63, + "learning_rate": 1.2461793883522338e-05, + "loss": 1.0102, + "step": 35581 + }, + { + "epoch": 2.63, + "learning_rate": 1.2461407285504593e-05, + "loss": 1.0742, + "step": 35582 + }, + { + "epoch": 2.63, + "learning_rate": 1.2461020683570782e-05, + "loss": 1.0864, + "step": 35583 + }, + { + "epoch": 2.63, + "learning_rate": 1.2460634077721524e-05, + "loss": 1.0426, + "step": 35584 + }, + { + "epoch": 2.63, + "learning_rate": 1.2460247467957431e-05, + "loss": 0.9157, + "step": 35585 + }, + { + "epoch": 2.63, + "learning_rate": 1.2459860854279122e-05, + "loss": 0.9625, + "step": 35586 + }, + { + "epoch": 2.63, + "learning_rate": 1.2459474236687208e-05, + "loss": 1.0273, + "step": 35587 + }, + { + "epoch": 2.63, + "learning_rate": 1.2459087615182308e-05, + "loss": 1.0015, + "step": 35588 + }, + { + "epoch": 2.63, + "learning_rate": 1.245870098976503e-05, + "loss": 1.0068, + "step": 35589 + }, + { + "epoch": 2.63, + "learning_rate": 1.2458314360436e-05, + "loss": 1.0674, + "step": 35590 + }, + { + "epoch": 2.63, + "learning_rate": 1.2457927727195822e-05, + "loss": 0.9209, + "step": 35591 + }, + { + "epoch": 2.63, + "learning_rate": 1.2457541090045118e-05, + "loss": 1.0528, + "step": 35592 + }, + { + "epoch": 2.63, + "learning_rate": 1.24571544489845e-05, + "loss": 1.0832, + "step": 35593 + }, + { + "epoch": 2.63, + "learning_rate": 1.2456767804014587e-05, + "loss": 0.9843, + "step": 35594 + }, + { + "epoch": 2.63, + "learning_rate": 1.2456381155135988e-05, + "loss": 1.0411, + "step": 35595 + }, + { + "epoch": 2.63, + "learning_rate": 1.2455994502349326e-05, + "loss": 0.9999, + "step": 35596 + }, + { + "epoch": 2.63, + "learning_rate": 1.2455607845655212e-05, + "loss": 1.0714, + "step": 35597 + }, + { + "epoch": 2.63, + "learning_rate": 1.2455221185054257e-05, + "loss": 0.9752, + "step": 35598 + }, + { + "epoch": 2.63, + "learning_rate": 1.245483452054708e-05, + "loss": 0.9925, + "step": 35599 + }, + { + "epoch": 2.63, + "learning_rate": 1.2454447852134302e-05, + "loss": 0.9201, + "step": 35600 + }, + { + "epoch": 2.63, + "learning_rate": 1.2454061179816528e-05, + "loss": 1.0069, + "step": 35601 + }, + { + "epoch": 2.63, + "learning_rate": 1.245367450359438e-05, + "loss": 1.0363, + "step": 35602 + }, + { + "epoch": 2.63, + "learning_rate": 1.2453287823468468e-05, + "loss": 1.016, + "step": 35603 + }, + { + "epoch": 2.63, + "learning_rate": 1.2452901139439413e-05, + "loss": 1.03, + "step": 35604 + }, + { + "epoch": 2.63, + "learning_rate": 1.2452514451507824e-05, + "loss": 1.0683, + "step": 35605 + }, + { + "epoch": 2.63, + "learning_rate": 1.2452127759674326e-05, + "loss": 0.918, + "step": 35606 + }, + { + "epoch": 2.63, + "learning_rate": 1.2451741063939522e-05, + "loss": 1.0433, + "step": 35607 + }, + { + "epoch": 2.63, + "learning_rate": 1.2451354364304036e-05, + "loss": 1.0443, + "step": 35608 + }, + { + "epoch": 2.63, + "learning_rate": 1.2450967660768479e-05, + "loss": 0.9987, + "step": 35609 + }, + { + "epoch": 2.63, + "learning_rate": 1.2450580953333469e-05, + "loss": 1.0486, + "step": 35610 + }, + { + "epoch": 2.63, + "learning_rate": 1.2450194241999615e-05, + "loss": 0.9478, + "step": 35611 + }, + { + "epoch": 2.63, + "learning_rate": 1.2449807526767543e-05, + "loss": 0.9983, + "step": 35612 + }, + { + "epoch": 2.63, + "learning_rate": 1.2449420807637858e-05, + "loss": 1.0162, + "step": 35613 + }, + { + "epoch": 2.63, + "learning_rate": 1.2449034084611182e-05, + "loss": 1.0298, + "step": 35614 + }, + { + "epoch": 2.63, + "learning_rate": 1.2448647357688129e-05, + "loss": 1.0787, + "step": 35615 + }, + { + "epoch": 2.63, + "learning_rate": 1.244826062686931e-05, + "loss": 0.9583, + "step": 35616 + }, + { + "epoch": 2.63, + "learning_rate": 1.2447873892155344e-05, + "loss": 1.0078, + "step": 35617 + }, + { + "epoch": 2.63, + "learning_rate": 1.2447487153546846e-05, + "loss": 0.9874, + "step": 35618 + }, + { + "epoch": 2.63, + "learning_rate": 1.244710041104443e-05, + "loss": 0.9748, + "step": 35619 + }, + { + "epoch": 2.63, + "learning_rate": 1.2446713664648715e-05, + "loss": 0.9989, + "step": 35620 + }, + { + "epoch": 2.63, + "learning_rate": 1.2446326914360312e-05, + "loss": 1.0512, + "step": 35621 + }, + { + "epoch": 2.63, + "learning_rate": 1.2445940160179838e-05, + "loss": 1.0501, + "step": 35622 + }, + { + "epoch": 2.63, + "learning_rate": 1.2445553402107905e-05, + "loss": 1.072, + "step": 35623 + }, + { + "epoch": 2.63, + "learning_rate": 1.2445166640145136e-05, + "loss": 0.9837, + "step": 35624 + }, + { + "epoch": 2.63, + "learning_rate": 1.2444779874292144e-05, + "loss": 1.0088, + "step": 35625 + }, + { + "epoch": 2.63, + "learning_rate": 1.2444393104549537e-05, + "loss": 1.0351, + "step": 35626 + }, + { + "epoch": 2.63, + "learning_rate": 1.2444006330917937e-05, + "loss": 1.0293, + "step": 35627 + }, + { + "epoch": 2.63, + "learning_rate": 1.2443619553397957e-05, + "loss": 1.0162, + "step": 35628 + }, + { + "epoch": 2.63, + "learning_rate": 1.2443232771990215e-05, + "loss": 0.9716, + "step": 35629 + }, + { + "epoch": 2.63, + "learning_rate": 1.2442845986695325e-05, + "loss": 1.0473, + "step": 35630 + }, + { + "epoch": 2.63, + "learning_rate": 1.2442459197513898e-05, + "loss": 0.9175, + "step": 35631 + }, + { + "epoch": 2.63, + "learning_rate": 1.244207240444656e-05, + "loss": 1.042, + "step": 35632 + }, + { + "epoch": 2.63, + "learning_rate": 1.2441685607493914e-05, + "loss": 1.1031, + "step": 35633 + }, + { + "epoch": 2.63, + "learning_rate": 1.2441298806656583e-05, + "loss": 1.0114, + "step": 35634 + }, + { + "epoch": 2.63, + "learning_rate": 1.2440912001935182e-05, + "loss": 1.0199, + "step": 35635 + }, + { + "epoch": 2.63, + "learning_rate": 1.2440525193330322e-05, + "loss": 1.0291, + "step": 35636 + }, + { + "epoch": 2.63, + "learning_rate": 1.2440138380842624e-05, + "loss": 1.0327, + "step": 35637 + }, + { + "epoch": 2.63, + "learning_rate": 1.24397515644727e-05, + "loss": 1.067, + "step": 35638 + }, + { + "epoch": 2.63, + "learning_rate": 1.2439364744221165e-05, + "loss": 0.9303, + "step": 35639 + }, + { + "epoch": 2.63, + "learning_rate": 1.2438977920088635e-05, + "loss": 0.9579, + "step": 35640 + }, + { + "epoch": 2.63, + "learning_rate": 1.2438591092075729e-05, + "loss": 1.0403, + "step": 35641 + }, + { + "epoch": 2.63, + "learning_rate": 1.2438204260183058e-05, + "loss": 1.0551, + "step": 35642 + }, + { + "epoch": 2.63, + "learning_rate": 1.243781742441124e-05, + "loss": 0.9427, + "step": 35643 + }, + { + "epoch": 2.63, + "learning_rate": 1.2437430584760889e-05, + "loss": 0.9223, + "step": 35644 + }, + { + "epoch": 2.63, + "learning_rate": 1.2437043741232619e-05, + "loss": 1.0312, + "step": 35645 + }, + { + "epoch": 2.63, + "learning_rate": 1.2436656893827049e-05, + "loss": 1.0008, + "step": 35646 + }, + { + "epoch": 2.63, + "learning_rate": 1.243627004254479e-05, + "loss": 1.0391, + "step": 35647 + }, + { + "epoch": 2.63, + "learning_rate": 1.2435883187386462e-05, + "loss": 1.082, + "step": 35648 + }, + { + "epoch": 2.63, + "learning_rate": 1.2435496328352682e-05, + "loss": 1.0903, + "step": 35649 + }, + { + "epoch": 2.63, + "learning_rate": 1.2435109465444058e-05, + "loss": 1.0661, + "step": 35650 + }, + { + "epoch": 2.63, + "learning_rate": 1.2434722598661211e-05, + "loss": 0.9739, + "step": 35651 + }, + { + "epoch": 2.63, + "learning_rate": 1.2434335728004753e-05, + "loss": 1.044, + "step": 35652 + }, + { + "epoch": 2.63, + "learning_rate": 1.2433948853475306e-05, + "loss": 1.0214, + "step": 35653 + }, + { + "epoch": 2.63, + "learning_rate": 1.2433561975073479e-05, + "loss": 1.0171, + "step": 35654 + }, + { + "epoch": 2.63, + "learning_rate": 1.243317509279989e-05, + "loss": 1.019, + "step": 35655 + }, + { + "epoch": 2.63, + "learning_rate": 1.2432788206655152e-05, + "loss": 1.0285, + "step": 35656 + }, + { + "epoch": 2.63, + "learning_rate": 1.2432401316639887e-05, + "loss": 0.9274, + "step": 35657 + }, + { + "epoch": 2.63, + "learning_rate": 1.2432014422754702e-05, + "loss": 1.0678, + "step": 35658 + }, + { + "epoch": 2.63, + "learning_rate": 1.243162752500022e-05, + "loss": 1.0238, + "step": 35659 + }, + { + "epoch": 2.63, + "learning_rate": 1.243124062337705e-05, + "loss": 1.0518, + "step": 35660 + }, + { + "epoch": 2.63, + "learning_rate": 1.2430853717885816e-05, + "loss": 1.0213, + "step": 35661 + }, + { + "epoch": 2.64, + "learning_rate": 1.2430466808527127e-05, + "loss": 1.0543, + "step": 35662 + }, + { + "epoch": 2.64, + "learning_rate": 1.2430079895301598e-05, + "loss": 1.0008, + "step": 35663 + }, + { + "epoch": 2.64, + "learning_rate": 1.2429692978209847e-05, + "loss": 0.9774, + "step": 35664 + }, + { + "epoch": 2.64, + "learning_rate": 1.2429306057252491e-05, + "loss": 0.9092, + "step": 35665 + }, + { + "epoch": 2.64, + "learning_rate": 1.242891913243014e-05, + "loss": 1.0075, + "step": 35666 + }, + { + "epoch": 2.64, + "learning_rate": 1.2428532203743417e-05, + "loss": 0.9628, + "step": 35667 + }, + { + "epoch": 2.64, + "learning_rate": 1.2428145271192931e-05, + "loss": 0.9224, + "step": 35668 + }, + { + "epoch": 2.64, + "learning_rate": 1.2427758334779303e-05, + "loss": 0.9316, + "step": 35669 + }, + { + "epoch": 2.64, + "learning_rate": 1.2427371394503145e-05, + "loss": 1.0976, + "step": 35670 + }, + { + "epoch": 2.64, + "learning_rate": 1.2426984450365076e-05, + "loss": 0.9887, + "step": 35671 + }, + { + "epoch": 2.64, + "learning_rate": 1.242659750236571e-05, + "loss": 1.117, + "step": 35672 + }, + { + "epoch": 2.64, + "learning_rate": 1.2426210550505658e-05, + "loss": 1.0414, + "step": 35673 + }, + { + "epoch": 2.64, + "learning_rate": 1.2425823594785543e-05, + "loss": 1.0532, + "step": 35674 + }, + { + "epoch": 2.64, + "learning_rate": 1.2425436635205977e-05, + "loss": 1.0718, + "step": 35675 + }, + { + "epoch": 2.64, + "learning_rate": 1.2425049671767573e-05, + "loss": 1.1106, + "step": 35676 + }, + { + "epoch": 2.64, + "learning_rate": 1.2424662704470954e-05, + "loss": 1.0342, + "step": 35677 + }, + { + "epoch": 2.64, + "learning_rate": 1.2424275733316727e-05, + "loss": 1.0065, + "step": 35678 + }, + { + "epoch": 2.64, + "learning_rate": 1.2423888758305516e-05, + "loss": 0.9672, + "step": 35679 + }, + { + "epoch": 2.64, + "learning_rate": 1.2423501779437927e-05, + "loss": 1.0769, + "step": 35680 + }, + { + "epoch": 2.64, + "learning_rate": 1.2423114796714589e-05, + "loss": 0.9641, + "step": 35681 + }, + { + "epoch": 2.64, + "learning_rate": 1.2422727810136107e-05, + "loss": 0.9951, + "step": 35682 + }, + { + "epoch": 2.64, + "learning_rate": 1.2422340819703096e-05, + "loss": 1.0891, + "step": 35683 + }, + { + "epoch": 2.64, + "learning_rate": 1.2421953825416179e-05, + "loss": 1.0117, + "step": 35684 + }, + { + "epoch": 2.64, + "learning_rate": 1.2421566827275968e-05, + "loss": 1.0394, + "step": 35685 + }, + { + "epoch": 2.64, + "learning_rate": 1.2421179825283079e-05, + "loss": 0.9007, + "step": 35686 + }, + { + "epoch": 2.64, + "learning_rate": 1.2420792819438126e-05, + "loss": 0.9927, + "step": 35687 + }, + { + "epoch": 2.64, + "learning_rate": 1.2420405809741726e-05, + "loss": 0.9489, + "step": 35688 + }, + { + "epoch": 2.64, + "learning_rate": 1.2420018796194495e-05, + "loss": 1.0667, + "step": 35689 + }, + { + "epoch": 2.64, + "learning_rate": 1.2419631778797054e-05, + "loss": 0.9112, + "step": 35690 + }, + { + "epoch": 2.64, + "learning_rate": 1.2419244757550008e-05, + "loss": 0.943, + "step": 35691 + }, + { + "epoch": 2.64, + "learning_rate": 1.241885773245398e-05, + "loss": 1.0561, + "step": 35692 + }, + { + "epoch": 2.64, + "learning_rate": 1.2418470703509582e-05, + "loss": 1.008, + "step": 35693 + }, + { + "epoch": 2.64, + "learning_rate": 1.2418083670717436e-05, + "loss": 0.9835, + "step": 35694 + }, + { + "epoch": 2.64, + "learning_rate": 1.2417696634078146e-05, + "loss": 1.0433, + "step": 35695 + }, + { + "epoch": 2.64, + "learning_rate": 1.2417309593592345e-05, + "loss": 0.9481, + "step": 35696 + }, + { + "epoch": 2.64, + "learning_rate": 1.241692254926063e-05, + "loss": 0.9038, + "step": 35697 + }, + { + "epoch": 2.64, + "learning_rate": 1.241653550108363e-05, + "loss": 1.051, + "step": 35698 + }, + { + "epoch": 2.64, + "learning_rate": 1.2416148449061955e-05, + "loss": 0.976, + "step": 35699 + }, + { + "epoch": 2.64, + "learning_rate": 1.2415761393196229e-05, + "loss": 1.0297, + "step": 35700 + }, + { + "epoch": 2.64, + "learning_rate": 1.2415374333487053e-05, + "loss": 0.9593, + "step": 35701 + }, + { + "epoch": 2.64, + "learning_rate": 1.2414987269935055e-05, + "loss": 1.025, + "step": 35702 + }, + { + "epoch": 2.64, + "learning_rate": 1.2414600202540846e-05, + "loss": 1.0283, + "step": 35703 + }, + { + "epoch": 2.64, + "learning_rate": 1.2414213131305041e-05, + "loss": 1.0289, + "step": 35704 + }, + { + "epoch": 2.64, + "learning_rate": 1.2413826056228257e-05, + "loss": 1.0127, + "step": 35705 + }, + { + "epoch": 2.64, + "learning_rate": 1.2413438977311113e-05, + "loss": 0.9203, + "step": 35706 + }, + { + "epoch": 2.64, + "learning_rate": 1.2413051894554222e-05, + "loss": 0.9705, + "step": 35707 + }, + { + "epoch": 2.64, + "learning_rate": 1.24126648079582e-05, + "loss": 1.0642, + "step": 35708 + }, + { + "epoch": 2.64, + "learning_rate": 1.2412277717523661e-05, + "loss": 1.0253, + "step": 35709 + }, + { + "epoch": 2.64, + "learning_rate": 1.2411890623251227e-05, + "loss": 1.1671, + "step": 35710 + }, + { + "epoch": 2.64, + "learning_rate": 1.2411503525141504e-05, + "loss": 0.8727, + "step": 35711 + }, + { + "epoch": 2.64, + "learning_rate": 1.2411116423195115e-05, + "loss": 1.0532, + "step": 35712 + }, + { + "epoch": 2.64, + "learning_rate": 1.2410729317412675e-05, + "loss": 1.0975, + "step": 35713 + }, + { + "epoch": 2.64, + "learning_rate": 1.2410342207794799e-05, + "loss": 1.0397, + "step": 35714 + }, + { + "epoch": 2.64, + "learning_rate": 1.2409955094342103e-05, + "loss": 0.9535, + "step": 35715 + }, + { + "epoch": 2.64, + "learning_rate": 1.2409567977055204e-05, + "loss": 1.0713, + "step": 35716 + }, + { + "epoch": 2.64, + "learning_rate": 1.2409180855934713e-05, + "loss": 0.8987, + "step": 35717 + }, + { + "epoch": 2.64, + "learning_rate": 1.2408793730981254e-05, + "loss": 1.0351, + "step": 35718 + }, + { + "epoch": 2.64, + "learning_rate": 1.2408406602195437e-05, + "loss": 1.0489, + "step": 35719 + }, + { + "epoch": 2.64, + "learning_rate": 1.240801946957788e-05, + "loss": 0.961, + "step": 35720 + }, + { + "epoch": 2.64, + "learning_rate": 1.2407632333129199e-05, + "loss": 0.9863, + "step": 35721 + }, + { + "epoch": 2.64, + "learning_rate": 1.2407245192850009e-05, + "loss": 1.1131, + "step": 35722 + }, + { + "epoch": 2.64, + "learning_rate": 1.2406858048740923e-05, + "loss": 1.0433, + "step": 35723 + }, + { + "epoch": 2.64, + "learning_rate": 1.2406470900802565e-05, + "loss": 1.0403, + "step": 35724 + }, + { + "epoch": 2.64, + "learning_rate": 1.2406083749035543e-05, + "loss": 1.0048, + "step": 35725 + }, + { + "epoch": 2.64, + "learning_rate": 1.240569659344048e-05, + "loss": 0.9475, + "step": 35726 + }, + { + "epoch": 2.64, + "learning_rate": 1.2405309434017982e-05, + "loss": 1.0096, + "step": 35727 + }, + { + "epoch": 2.64, + "learning_rate": 1.2404922270768678e-05, + "loss": 1.0132, + "step": 35728 + }, + { + "epoch": 2.64, + "learning_rate": 1.2404535103693174e-05, + "loss": 0.8704, + "step": 35729 + }, + { + "epoch": 2.64, + "learning_rate": 1.2404147932792088e-05, + "loss": 0.9272, + "step": 35730 + }, + { + "epoch": 2.64, + "learning_rate": 1.240376075806604e-05, + "loss": 0.8865, + "step": 35731 + }, + { + "epoch": 2.64, + "learning_rate": 1.2403373579515638e-05, + "loss": 0.9855, + "step": 35732 + }, + { + "epoch": 2.64, + "learning_rate": 1.2402986397141507e-05, + "loss": 1.0321, + "step": 35733 + }, + { + "epoch": 2.64, + "learning_rate": 1.2402599210944258e-05, + "loss": 0.9788, + "step": 35734 + }, + { + "epoch": 2.64, + "learning_rate": 1.2402212020924506e-05, + "loss": 1.0415, + "step": 35735 + }, + { + "epoch": 2.64, + "learning_rate": 1.2401824827082873e-05, + "loss": 1.0697, + "step": 35736 + }, + { + "epoch": 2.64, + "learning_rate": 1.240143762941997e-05, + "loss": 1.0226, + "step": 35737 + }, + { + "epoch": 2.64, + "learning_rate": 1.2401050427936412e-05, + "loss": 1.0713, + "step": 35738 + }, + { + "epoch": 2.64, + "learning_rate": 1.240066322263282e-05, + "loss": 1.015, + "step": 35739 + }, + { + "epoch": 2.64, + "learning_rate": 1.2400276013509804e-05, + "loss": 1.0056, + "step": 35740 + }, + { + "epoch": 2.64, + "learning_rate": 1.2399888800567984e-05, + "loss": 1.062, + "step": 35741 + }, + { + "epoch": 2.64, + "learning_rate": 1.2399501583807975e-05, + "loss": 1.1347, + "step": 35742 + }, + { + "epoch": 2.64, + "learning_rate": 1.2399114363230397e-05, + "loss": 0.9382, + "step": 35743 + }, + { + "epoch": 2.64, + "learning_rate": 1.2398727138835858e-05, + "loss": 0.9532, + "step": 35744 + }, + { + "epoch": 2.64, + "learning_rate": 1.2398339910624982e-05, + "loss": 0.9759, + "step": 35745 + }, + { + "epoch": 2.64, + "learning_rate": 1.239795267859838e-05, + "loss": 0.9991, + "step": 35746 + }, + { + "epoch": 2.64, + "learning_rate": 1.239756544275667e-05, + "loss": 1.0411, + "step": 35747 + }, + { + "epoch": 2.64, + "learning_rate": 1.2397178203100466e-05, + "loss": 0.9171, + "step": 35748 + }, + { + "epoch": 2.64, + "learning_rate": 1.2396790959630388e-05, + "loss": 1.0908, + "step": 35749 + }, + { + "epoch": 2.64, + "learning_rate": 1.2396403712347047e-05, + "loss": 0.9878, + "step": 35750 + }, + { + "epoch": 2.64, + "learning_rate": 1.2396016461251063e-05, + "loss": 1.0662, + "step": 35751 + }, + { + "epoch": 2.64, + "learning_rate": 1.2395629206343052e-05, + "loss": 0.995, + "step": 35752 + }, + { + "epoch": 2.64, + "learning_rate": 1.2395241947623629e-05, + "loss": 0.924, + "step": 35753 + }, + { + "epoch": 2.64, + "learning_rate": 1.2394854685093409e-05, + "loss": 1.0497, + "step": 35754 + }, + { + "epoch": 2.64, + "learning_rate": 1.239446741875301e-05, + "loss": 1.0049, + "step": 35755 + }, + { + "epoch": 2.64, + "learning_rate": 1.2394080148603051e-05, + "loss": 0.9885, + "step": 35756 + }, + { + "epoch": 2.64, + "learning_rate": 1.2393692874644144e-05, + "loss": 0.9421, + "step": 35757 + }, + { + "epoch": 2.64, + "learning_rate": 1.2393305596876902e-05, + "loss": 0.9732, + "step": 35758 + }, + { + "epoch": 2.64, + "learning_rate": 1.2392918315301947e-05, + "loss": 0.9326, + "step": 35759 + }, + { + "epoch": 2.64, + "learning_rate": 1.2392531029919892e-05, + "loss": 0.9789, + "step": 35760 + }, + { + "epoch": 2.64, + "learning_rate": 1.2392143740731356e-05, + "loss": 1.0748, + "step": 35761 + }, + { + "epoch": 2.64, + "learning_rate": 1.2391756447736954e-05, + "loss": 0.9561, + "step": 35762 + }, + { + "epoch": 2.64, + "learning_rate": 1.23913691509373e-05, + "loss": 0.966, + "step": 35763 + }, + { + "epoch": 2.64, + "learning_rate": 1.239098185033301e-05, + "loss": 0.9855, + "step": 35764 + }, + { + "epoch": 2.64, + "learning_rate": 1.239059454592471e-05, + "loss": 0.9492, + "step": 35765 + }, + { + "epoch": 2.64, + "learning_rate": 1.2390207237713001e-05, + "loss": 0.8689, + "step": 35766 + }, + { + "epoch": 2.64, + "learning_rate": 1.238981992569851e-05, + "loss": 0.9534, + "step": 35767 + }, + { + "epoch": 2.64, + "learning_rate": 1.2389432609881846e-05, + "loss": 0.9756, + "step": 35768 + }, + { + "epoch": 2.64, + "learning_rate": 1.2389045290263633e-05, + "loss": 1.0155, + "step": 35769 + }, + { + "epoch": 2.64, + "learning_rate": 1.238865796684448e-05, + "loss": 1.0434, + "step": 35770 + }, + { + "epoch": 2.64, + "learning_rate": 1.2388270639625008e-05, + "loss": 0.9195, + "step": 35771 + }, + { + "epoch": 2.64, + "learning_rate": 1.238788330860583e-05, + "loss": 1.0892, + "step": 35772 + }, + { + "epoch": 2.64, + "learning_rate": 1.2387495973787565e-05, + "loss": 1.061, + "step": 35773 + }, + { + "epoch": 2.64, + "learning_rate": 1.2387108635170828e-05, + "loss": 0.876, + "step": 35774 + }, + { + "epoch": 2.64, + "learning_rate": 1.2386721292756234e-05, + "loss": 1.0266, + "step": 35775 + }, + { + "epoch": 2.64, + "learning_rate": 1.2386333946544405e-05, + "loss": 0.9581, + "step": 35776 + }, + { + "epoch": 2.64, + "learning_rate": 1.2385946596535947e-05, + "loss": 0.9965, + "step": 35777 + }, + { + "epoch": 2.64, + "learning_rate": 1.2385559242731486e-05, + "loss": 1.1154, + "step": 35778 + }, + { + "epoch": 2.64, + "learning_rate": 1.2385171885131632e-05, + "loss": 0.9612, + "step": 35779 + }, + { + "epoch": 2.64, + "learning_rate": 1.2384784523737005e-05, + "loss": 1.0468, + "step": 35780 + }, + { + "epoch": 2.64, + "learning_rate": 1.2384397158548223e-05, + "loss": 1.0061, + "step": 35781 + }, + { + "epoch": 2.64, + "learning_rate": 1.2384009789565894e-05, + "loss": 1.0407, + "step": 35782 + }, + { + "epoch": 2.64, + "learning_rate": 1.2383622416790642e-05, + "loss": 0.8582, + "step": 35783 + }, + { + "epoch": 2.64, + "learning_rate": 1.238323504022308e-05, + "loss": 1.0731, + "step": 35784 + }, + { + "epoch": 2.64, + "learning_rate": 1.2382847659863826e-05, + "loss": 1.0136, + "step": 35785 + }, + { + "epoch": 2.64, + "learning_rate": 1.2382460275713494e-05, + "loss": 0.9656, + "step": 35786 + }, + { + "epoch": 2.64, + "learning_rate": 1.2382072887772702e-05, + "loss": 1.0218, + "step": 35787 + }, + { + "epoch": 2.64, + "learning_rate": 1.2381685496042068e-05, + "loss": 1.0725, + "step": 35788 + }, + { + "epoch": 2.64, + "learning_rate": 1.2381298100522205e-05, + "loss": 0.975, + "step": 35789 + }, + { + "epoch": 2.64, + "learning_rate": 1.2380910701213731e-05, + "loss": 0.9312, + "step": 35790 + }, + { + "epoch": 2.64, + "learning_rate": 1.2380523298117261e-05, + "loss": 0.9661, + "step": 35791 + }, + { + "epoch": 2.64, + "learning_rate": 1.2380135891233416e-05, + "loss": 1.0493, + "step": 35792 + }, + { + "epoch": 2.64, + "learning_rate": 1.2379748480562808e-05, + "loss": 1.0984, + "step": 35793 + }, + { + "epoch": 2.64, + "learning_rate": 1.2379361066106053e-05, + "loss": 0.9508, + "step": 35794 + }, + { + "epoch": 2.64, + "learning_rate": 1.2378973647863767e-05, + "loss": 0.9677, + "step": 35795 + }, + { + "epoch": 2.64, + "learning_rate": 1.237858622583657e-05, + "loss": 0.9735, + "step": 35796 + }, + { + "epoch": 2.65, + "learning_rate": 1.2378198800025074e-05, + "loss": 0.9759, + "step": 35797 + }, + { + "epoch": 2.65, + "learning_rate": 1.2377811370429902e-05, + "loss": 1.0348, + "step": 35798 + }, + { + "epoch": 2.65, + "learning_rate": 1.2377423937051662e-05, + "loss": 1.131, + "step": 35799 + }, + { + "epoch": 2.65, + "learning_rate": 1.2377036499890977e-05, + "loss": 1.048, + "step": 35800 + }, + { + "epoch": 2.65, + "learning_rate": 1.237664905894846e-05, + "loss": 1.0481, + "step": 35801 + }, + { + "epoch": 2.65, + "learning_rate": 1.2376261614224729e-05, + "loss": 0.8299, + "step": 35802 + }, + { + "epoch": 2.65, + "learning_rate": 1.2375874165720399e-05, + "loss": 1.0909, + "step": 35803 + }, + { + "epoch": 2.65, + "learning_rate": 1.2375486713436089e-05, + "loss": 1.0271, + "step": 35804 + }, + { + "epoch": 2.65, + "learning_rate": 1.237509925737241e-05, + "loss": 1.0527, + "step": 35805 + }, + { + "epoch": 2.65, + "learning_rate": 1.2374711797529986e-05, + "loss": 0.9494, + "step": 35806 + }, + { + "epoch": 2.65, + "learning_rate": 1.2374324333909426e-05, + "loss": 1.0055, + "step": 35807 + }, + { + "epoch": 2.65, + "learning_rate": 1.2373936866511351e-05, + "loss": 0.993, + "step": 35808 + }, + { + "epoch": 2.65, + "learning_rate": 1.2373549395336377e-05, + "loss": 0.9598, + "step": 35809 + }, + { + "epoch": 2.65, + "learning_rate": 1.2373161920385122e-05, + "loss": 1.0686, + "step": 35810 + }, + { + "epoch": 2.65, + "learning_rate": 1.2372774441658195e-05, + "loss": 1.0951, + "step": 35811 + }, + { + "epoch": 2.65, + "learning_rate": 1.2372386959156224e-05, + "loss": 0.9751, + "step": 35812 + }, + { + "epoch": 2.65, + "learning_rate": 1.2371999472879816e-05, + "loss": 1.0648, + "step": 35813 + }, + { + "epoch": 2.65, + "learning_rate": 1.2371611982829592e-05, + "loss": 1.0135, + "step": 35814 + }, + { + "epoch": 2.65, + "learning_rate": 1.2371224489006166e-05, + "loss": 0.9692, + "step": 35815 + }, + { + "epoch": 2.65, + "learning_rate": 1.2370836991410156e-05, + "loss": 1.0037, + "step": 35816 + }, + { + "epoch": 2.65, + "learning_rate": 1.2370449490042177e-05, + "loss": 0.9579, + "step": 35817 + }, + { + "epoch": 2.65, + "learning_rate": 1.237006198490285e-05, + "loss": 1.0999, + "step": 35818 + }, + { + "epoch": 2.65, + "learning_rate": 1.2369674475992785e-05, + "loss": 0.98, + "step": 35819 + }, + { + "epoch": 2.65, + "learning_rate": 1.2369286963312604e-05, + "loss": 1.0168, + "step": 35820 + }, + { + "epoch": 2.65, + "learning_rate": 1.2368899446862921e-05, + "loss": 0.9956, + "step": 35821 + }, + { + "epoch": 2.65, + "learning_rate": 1.2368511926644351e-05, + "loss": 0.9579, + "step": 35822 + }, + { + "epoch": 2.65, + "learning_rate": 1.2368124402657515e-05, + "loss": 0.9867, + "step": 35823 + }, + { + "epoch": 2.65, + "learning_rate": 1.2367736874903027e-05, + "loss": 1.012, + "step": 35824 + }, + { + "epoch": 2.65, + "learning_rate": 1.23673493433815e-05, + "loss": 1.1212, + "step": 35825 + }, + { + "epoch": 2.65, + "learning_rate": 1.2366961808093559e-05, + "loss": 0.944, + "step": 35826 + }, + { + "epoch": 2.65, + "learning_rate": 1.236657426903981e-05, + "loss": 0.9502, + "step": 35827 + }, + { + "epoch": 2.65, + "learning_rate": 1.236618672622088e-05, + "loss": 0.9986, + "step": 35828 + }, + { + "epoch": 2.65, + "learning_rate": 1.2365799179637379e-05, + "loss": 1.0019, + "step": 35829 + }, + { + "epoch": 2.65, + "learning_rate": 1.2365411629289925e-05, + "loss": 0.9671, + "step": 35830 + }, + { + "epoch": 2.65, + "learning_rate": 1.2365024075179138e-05, + "loss": 0.9874, + "step": 35831 + }, + { + "epoch": 2.65, + "learning_rate": 1.2364636517305627e-05, + "loss": 0.9893, + "step": 35832 + }, + { + "epoch": 2.65, + "learning_rate": 1.2364248955670016e-05, + "loss": 1.0349, + "step": 35833 + }, + { + "epoch": 2.65, + "learning_rate": 1.2363861390272916e-05, + "loss": 1.0207, + "step": 35834 + }, + { + "epoch": 2.65, + "learning_rate": 1.236347382111495e-05, + "loss": 1.1033, + "step": 35835 + }, + { + "epoch": 2.65, + "learning_rate": 1.2363086248196727e-05, + "loss": 1.0987, + "step": 35836 + }, + { + "epoch": 2.65, + "learning_rate": 1.236269867151887e-05, + "loss": 1.049, + "step": 35837 + }, + { + "epoch": 2.65, + "learning_rate": 1.2362311091081992e-05, + "loss": 0.9479, + "step": 35838 + }, + { + "epoch": 2.65, + "learning_rate": 1.2361923506886712e-05, + "loss": 1.1169, + "step": 35839 + }, + { + "epoch": 2.65, + "learning_rate": 1.2361535918933647e-05, + "loss": 1.0339, + "step": 35840 + }, + { + "epoch": 2.65, + "learning_rate": 1.236114832722341e-05, + "loss": 0.9798, + "step": 35841 + }, + { + "epoch": 2.65, + "learning_rate": 1.2360760731756621e-05, + "loss": 0.9497, + "step": 35842 + }, + { + "epoch": 2.65, + "learning_rate": 1.2360373132533894e-05, + "loss": 1.0908, + "step": 35843 + }, + { + "epoch": 2.65, + "learning_rate": 1.2359985529555846e-05, + "loss": 1.0482, + "step": 35844 + }, + { + "epoch": 2.65, + "learning_rate": 1.2359597922823098e-05, + "loss": 1.0334, + "step": 35845 + }, + { + "epoch": 2.65, + "learning_rate": 1.2359210312336261e-05, + "loss": 0.9291, + "step": 35846 + }, + { + "epoch": 2.65, + "learning_rate": 1.2358822698095955e-05, + "loss": 1.0256, + "step": 35847 + }, + { + "epoch": 2.65, + "learning_rate": 1.2358435080102795e-05, + "loss": 1.0125, + "step": 35848 + }, + { + "epoch": 2.65, + "learning_rate": 1.2358047458357403e-05, + "loss": 0.9262, + "step": 35849 + }, + { + "epoch": 2.65, + "learning_rate": 1.2357659832860387e-05, + "loss": 0.8733, + "step": 35850 + }, + { + "epoch": 2.65, + "learning_rate": 1.235727220361237e-05, + "loss": 0.9246, + "step": 35851 + }, + { + "epoch": 2.65, + "learning_rate": 1.2356884570613963e-05, + "loss": 0.9512, + "step": 35852 + }, + { + "epoch": 2.65, + "learning_rate": 1.235649693386579e-05, + "loss": 1.0505, + "step": 35853 + }, + { + "epoch": 2.65, + "learning_rate": 1.2356109293368462e-05, + "loss": 1.0132, + "step": 35854 + }, + { + "epoch": 2.65, + "learning_rate": 1.2355721649122598e-05, + "loss": 0.9772, + "step": 35855 + }, + { + "epoch": 2.65, + "learning_rate": 1.2355334001128813e-05, + "loss": 0.9614, + "step": 35856 + }, + { + "epoch": 2.65, + "learning_rate": 1.2354946349387728e-05, + "loss": 1.0579, + "step": 35857 + }, + { + "epoch": 2.65, + "learning_rate": 1.2354558693899958e-05, + "loss": 0.9708, + "step": 35858 + }, + { + "epoch": 2.65, + "learning_rate": 1.2354171034666118e-05, + "loss": 1.0756, + "step": 35859 + }, + { + "epoch": 2.65, + "learning_rate": 1.2353783371686824e-05, + "loss": 1.1469, + "step": 35860 + }, + { + "epoch": 2.65, + "learning_rate": 1.2353395704962697e-05, + "loss": 0.981, + "step": 35861 + }, + { + "epoch": 2.65, + "learning_rate": 1.2353008034494346e-05, + "loss": 1.0175, + "step": 35862 + }, + { + "epoch": 2.65, + "learning_rate": 1.2352620360282399e-05, + "loss": 0.8745, + "step": 35863 + }, + { + "epoch": 2.65, + "learning_rate": 1.2352232682327463e-05, + "loss": 1.1745, + "step": 35864 + }, + { + "epoch": 2.65, + "learning_rate": 1.2351845000630161e-05, + "loss": 1.0008, + "step": 35865 + }, + { + "epoch": 2.65, + "learning_rate": 1.2351457315191101e-05, + "loss": 0.9493, + "step": 35866 + }, + { + "epoch": 2.65, + "learning_rate": 1.2351069626010914e-05, + "loss": 0.9667, + "step": 35867 + }, + { + "epoch": 2.65, + "learning_rate": 1.2350681933090205e-05, + "loss": 0.9001, + "step": 35868 + }, + { + "epoch": 2.65, + "learning_rate": 1.2350294236429596e-05, + "loss": 0.9737, + "step": 35869 + }, + { + "epoch": 2.65, + "learning_rate": 1.2349906536029701e-05, + "loss": 1.0798, + "step": 35870 + }, + { + "epoch": 2.65, + "learning_rate": 1.234951883189114e-05, + "loss": 1.0313, + "step": 35871 + }, + { + "epoch": 2.65, + "learning_rate": 1.2349131124014526e-05, + "loss": 0.9994, + "step": 35872 + }, + { + "epoch": 2.65, + "learning_rate": 1.234874341240048e-05, + "loss": 0.8771, + "step": 35873 + }, + { + "epoch": 2.65, + "learning_rate": 1.2348355697049615e-05, + "loss": 0.9335, + "step": 35874 + }, + { + "epoch": 2.65, + "learning_rate": 1.234796797796255e-05, + "loss": 0.9545, + "step": 35875 + }, + { + "epoch": 2.65, + "learning_rate": 1.2347580255139902e-05, + "loss": 1.0208, + "step": 35876 + }, + { + "epoch": 2.65, + "learning_rate": 1.2347192528582289e-05, + "loss": 0.9985, + "step": 35877 + }, + { + "epoch": 2.65, + "learning_rate": 1.2346804798290326e-05, + "loss": 0.9685, + "step": 35878 + }, + { + "epoch": 2.65, + "learning_rate": 1.2346417064264626e-05, + "loss": 0.937, + "step": 35879 + }, + { + "epoch": 2.65, + "learning_rate": 1.2346029326505815e-05, + "loss": 1.0253, + "step": 35880 + }, + { + "epoch": 2.65, + "learning_rate": 1.2345641585014502e-05, + "loss": 0.9443, + "step": 35881 + }, + { + "epoch": 2.65, + "learning_rate": 1.2345253839791308e-05, + "loss": 0.901, + "step": 35882 + }, + { + "epoch": 2.65, + "learning_rate": 1.234486609083685e-05, + "loss": 0.98, + "step": 35883 + }, + { + "epoch": 2.65, + "learning_rate": 1.2344478338151743e-05, + "loss": 0.9846, + "step": 35884 + }, + { + "epoch": 2.65, + "learning_rate": 1.2344090581736602e-05, + "loss": 1.0344, + "step": 35885 + }, + { + "epoch": 2.65, + "learning_rate": 1.2343702821592047e-05, + "loss": 0.9457, + "step": 35886 + }, + { + "epoch": 2.65, + "learning_rate": 1.2343315057718697e-05, + "loss": 1.0525, + "step": 35887 + }, + { + "epoch": 2.65, + "learning_rate": 1.2342927290117164e-05, + "loss": 0.9392, + "step": 35888 + }, + { + "epoch": 2.65, + "learning_rate": 1.2342539518788069e-05, + "loss": 1.0522, + "step": 35889 + }, + { + "epoch": 2.65, + "learning_rate": 1.2342151743732025e-05, + "loss": 1.0455, + "step": 35890 + }, + { + "epoch": 2.65, + "learning_rate": 1.234176396494965e-05, + "loss": 0.9799, + "step": 35891 + }, + { + "epoch": 2.65, + "learning_rate": 1.2341376182441567e-05, + "loss": 0.988, + "step": 35892 + }, + { + "epoch": 2.65, + "learning_rate": 1.2340988396208382e-05, + "loss": 1.0997, + "step": 35893 + }, + { + "epoch": 2.65, + "learning_rate": 1.2340600606250723e-05, + "loss": 0.9815, + "step": 35894 + }, + { + "epoch": 2.65, + "learning_rate": 1.23402128125692e-05, + "loss": 1.0641, + "step": 35895 + }, + { + "epoch": 2.65, + "learning_rate": 1.2339825015164433e-05, + "loss": 1.0316, + "step": 35896 + }, + { + "epoch": 2.65, + "learning_rate": 1.2339437214037035e-05, + "loss": 0.945, + "step": 35897 + }, + { + "epoch": 2.65, + "learning_rate": 1.233904940918763e-05, + "loss": 0.9897, + "step": 35898 + }, + { + "epoch": 2.65, + "learning_rate": 1.2338661600616825e-05, + "loss": 1.0, + "step": 35899 + }, + { + "epoch": 2.65, + "learning_rate": 1.2338273788325247e-05, + "loss": 1.0715, + "step": 35900 + }, + { + "epoch": 2.65, + "learning_rate": 1.2337885972313508e-05, + "loss": 1.0485, + "step": 35901 + }, + { + "epoch": 2.65, + "learning_rate": 1.2337498152582228e-05, + "loss": 1.0562, + "step": 35902 + }, + { + "epoch": 2.65, + "learning_rate": 1.2337110329132017e-05, + "loss": 0.9387, + "step": 35903 + }, + { + "epoch": 2.65, + "learning_rate": 1.23367225019635e-05, + "loss": 1.0909, + "step": 35904 + }, + { + "epoch": 2.65, + "learning_rate": 1.2336334671077293e-05, + "loss": 1.0775, + "step": 35905 + }, + { + "epoch": 2.65, + "learning_rate": 1.2335946836474008e-05, + "loss": 1.035, + "step": 35906 + }, + { + "epoch": 2.65, + "learning_rate": 1.2335558998154267e-05, + "loss": 1.1016, + "step": 35907 + }, + { + "epoch": 2.65, + "learning_rate": 1.2335171156118683e-05, + "loss": 1.1248, + "step": 35908 + }, + { + "epoch": 2.65, + "learning_rate": 1.2334783310367876e-05, + "loss": 1.0119, + "step": 35909 + }, + { + "epoch": 2.65, + "learning_rate": 1.2334395460902461e-05, + "loss": 1.0357, + "step": 35910 + }, + { + "epoch": 2.65, + "learning_rate": 1.2334007607723058e-05, + "loss": 1.095, + "step": 35911 + }, + { + "epoch": 2.65, + "learning_rate": 1.2333619750830281e-05, + "loss": 0.9114, + "step": 35912 + }, + { + "epoch": 2.65, + "learning_rate": 1.2333231890224749e-05, + "loss": 1.0451, + "step": 35913 + }, + { + "epoch": 2.65, + "learning_rate": 1.2332844025907077e-05, + "loss": 1.0537, + "step": 35914 + }, + { + "epoch": 2.65, + "learning_rate": 1.2332456157877887e-05, + "loss": 1.0196, + "step": 35915 + }, + { + "epoch": 2.65, + "learning_rate": 1.233206828613779e-05, + "loss": 1.0008, + "step": 35916 + }, + { + "epoch": 2.65, + "learning_rate": 1.2331680410687405e-05, + "loss": 0.9814, + "step": 35917 + }, + { + "epoch": 2.65, + "learning_rate": 1.2331292531527352e-05, + "loss": 1.0907, + "step": 35918 + }, + { + "epoch": 2.65, + "learning_rate": 1.2330904648658244e-05, + "loss": 0.924, + "step": 35919 + }, + { + "epoch": 2.65, + "learning_rate": 1.2330516762080701e-05, + "loss": 0.9758, + "step": 35920 + }, + { + "epoch": 2.65, + "learning_rate": 1.233012887179534e-05, + "loss": 0.9941, + "step": 35921 + }, + { + "epoch": 2.65, + "learning_rate": 1.2329740977802777e-05, + "loss": 1.008, + "step": 35922 + }, + { + "epoch": 2.65, + "learning_rate": 1.2329353080103627e-05, + "loss": 1.0242, + "step": 35923 + }, + { + "epoch": 2.65, + "learning_rate": 1.2328965178698513e-05, + "loss": 1.012, + "step": 35924 + }, + { + "epoch": 2.65, + "learning_rate": 1.232857727358805e-05, + "loss": 1.1036, + "step": 35925 + }, + { + "epoch": 2.65, + "learning_rate": 1.232818936477285e-05, + "loss": 1.0818, + "step": 35926 + }, + { + "epoch": 2.65, + "learning_rate": 1.2327801452253534e-05, + "loss": 1.0642, + "step": 35927 + }, + { + "epoch": 2.65, + "learning_rate": 1.2327413536030718e-05, + "loss": 0.9527, + "step": 35928 + }, + { + "epoch": 2.65, + "learning_rate": 1.2327025616105026e-05, + "loss": 1.0382, + "step": 35929 + }, + { + "epoch": 2.65, + "learning_rate": 1.2326637692477066e-05, + "loss": 0.9808, + "step": 35930 + }, + { + "epoch": 2.65, + "learning_rate": 1.2326249765147458e-05, + "loss": 0.9665, + "step": 35931 + }, + { + "epoch": 2.65, + "learning_rate": 1.2325861834116821e-05, + "loss": 0.943, + "step": 35932 + }, + { + "epoch": 2.66, + "learning_rate": 1.2325473899385774e-05, + "loss": 1.0062, + "step": 35933 + }, + { + "epoch": 2.66, + "learning_rate": 1.2325085960954925e-05, + "loss": 1.0436, + "step": 35934 + }, + { + "epoch": 2.66, + "learning_rate": 1.2324698018824903e-05, + "loss": 0.9985, + "step": 35935 + }, + { + "epoch": 2.66, + "learning_rate": 1.2324310072996316e-05, + "loss": 0.9953, + "step": 35936 + }, + { + "epoch": 2.66, + "learning_rate": 1.2323922123469786e-05, + "loss": 0.9724, + "step": 35937 + }, + { + "epoch": 2.66, + "learning_rate": 1.2323534170245929e-05, + "loss": 0.9366, + "step": 35938 + }, + { + "epoch": 2.66, + "learning_rate": 1.2323146213325363e-05, + "loss": 0.9731, + "step": 35939 + }, + { + "epoch": 2.66, + "learning_rate": 1.2322758252708704e-05, + "loss": 1.0389, + "step": 35940 + }, + { + "epoch": 2.66, + "learning_rate": 1.2322370288396572e-05, + "loss": 1.081, + "step": 35941 + }, + { + "epoch": 2.66, + "learning_rate": 1.2321982320389578e-05, + "loss": 1.0468, + "step": 35942 + }, + { + "epoch": 2.66, + "learning_rate": 1.232159434868835e-05, + "loss": 1.0134, + "step": 35943 + }, + { + "epoch": 2.66, + "learning_rate": 1.2321206373293493e-05, + "loss": 0.9708, + "step": 35944 + }, + { + "epoch": 2.66, + "learning_rate": 1.2320818394205632e-05, + "loss": 0.9423, + "step": 35945 + }, + { + "epoch": 2.66, + "learning_rate": 1.2320430411425382e-05, + "loss": 1.0249, + "step": 35946 + }, + { + "epoch": 2.66, + "learning_rate": 1.232004242495336e-05, + "loss": 0.9766, + "step": 35947 + }, + { + "epoch": 2.66, + "learning_rate": 1.2319654434790183e-05, + "loss": 1.1318, + "step": 35948 + }, + { + "epoch": 2.66, + "learning_rate": 1.231926644093647e-05, + "loss": 0.9775, + "step": 35949 + }, + { + "epoch": 2.66, + "learning_rate": 1.2318878443392835e-05, + "loss": 0.9942, + "step": 35950 + }, + { + "epoch": 2.66, + "learning_rate": 1.2318490442159901e-05, + "loss": 1.027, + "step": 35951 + }, + { + "epoch": 2.66, + "learning_rate": 1.2318102437238279e-05, + "loss": 1.0157, + "step": 35952 + }, + { + "epoch": 2.66, + "learning_rate": 1.2317714428628593e-05, + "loss": 0.9803, + "step": 35953 + }, + { + "epoch": 2.66, + "learning_rate": 1.2317326416331452e-05, + "loss": 1.1857, + "step": 35954 + }, + { + "epoch": 2.66, + "learning_rate": 1.231693840034748e-05, + "loss": 0.9496, + "step": 35955 + }, + { + "epoch": 2.66, + "learning_rate": 1.2316550380677293e-05, + "loss": 1.0758, + "step": 35956 + }, + { + "epoch": 2.66, + "learning_rate": 1.2316162357321506e-05, + "loss": 1.0811, + "step": 35957 + }, + { + "epoch": 2.66, + "learning_rate": 1.2315774330280737e-05, + "loss": 1.1023, + "step": 35958 + }, + { + "epoch": 2.66, + "learning_rate": 1.2315386299555608e-05, + "loss": 1.0024, + "step": 35959 + }, + { + "epoch": 2.66, + "learning_rate": 1.2314998265146726e-05, + "loss": 0.9852, + "step": 35960 + }, + { + "epoch": 2.66, + "learning_rate": 1.2314610227054725e-05, + "loss": 1.0446, + "step": 35961 + }, + { + "epoch": 2.66, + "learning_rate": 1.2314222185280202e-05, + "loss": 1.0098, + "step": 35962 + }, + { + "epoch": 2.66, + "learning_rate": 1.2313834139823791e-05, + "loss": 0.9989, + "step": 35963 + }, + { + "epoch": 2.66, + "learning_rate": 1.2313446090686099e-05, + "loss": 1.0394, + "step": 35964 + }, + { + "epoch": 2.66, + "learning_rate": 1.231305803786775e-05, + "loss": 1.0903, + "step": 35965 + }, + { + "epoch": 2.66, + "learning_rate": 1.2312669981369356e-05, + "loss": 1.142, + "step": 35966 + }, + { + "epoch": 2.66, + "learning_rate": 1.231228192119154e-05, + "loss": 1.0808, + "step": 35967 + }, + { + "epoch": 2.66, + "learning_rate": 1.2311893857334915e-05, + "loss": 1.0013, + "step": 35968 + }, + { + "epoch": 2.66, + "learning_rate": 1.2311505789800103e-05, + "loss": 0.9621, + "step": 35969 + }, + { + "epoch": 2.66, + "learning_rate": 1.2311117718587714e-05, + "loss": 1.0573, + "step": 35970 + }, + { + "epoch": 2.66, + "learning_rate": 1.2310729643698372e-05, + "loss": 0.9809, + "step": 35971 + }, + { + "epoch": 2.66, + "learning_rate": 1.2310341565132693e-05, + "loss": 1.0816, + "step": 35972 + }, + { + "epoch": 2.66, + "learning_rate": 1.2309953482891293e-05, + "loss": 0.9434, + "step": 35973 + }, + { + "epoch": 2.66, + "learning_rate": 1.2309565396974787e-05, + "loss": 1.0438, + "step": 35974 + }, + { + "epoch": 2.66, + "learning_rate": 1.2309177307383799e-05, + "loss": 1.0681, + "step": 35975 + }, + { + "epoch": 2.66, + "learning_rate": 1.2308789214118941e-05, + "loss": 0.9793, + "step": 35976 + }, + { + "epoch": 2.66, + "learning_rate": 1.2308401117180835e-05, + "loss": 1.0988, + "step": 35977 + }, + { + "epoch": 2.66, + "learning_rate": 1.2308013016570092e-05, + "loss": 0.9406, + "step": 35978 + }, + { + "epoch": 2.66, + "learning_rate": 1.2307624912287337e-05, + "loss": 0.9788, + "step": 35979 + }, + { + "epoch": 2.66, + "learning_rate": 1.2307236804333183e-05, + "loss": 0.9819, + "step": 35980 + }, + { + "epoch": 2.66, + "learning_rate": 1.2306848692708247e-05, + "loss": 0.8861, + "step": 35981 + }, + { + "epoch": 2.66, + "learning_rate": 1.2306460577413148e-05, + "loss": 1.1309, + "step": 35982 + }, + { + "epoch": 2.66, + "learning_rate": 1.2306072458448504e-05, + "loss": 0.9413, + "step": 35983 + }, + { + "epoch": 2.66, + "learning_rate": 1.230568433581493e-05, + "loss": 0.9774, + "step": 35984 + }, + { + "epoch": 2.66, + "learning_rate": 1.2305296209513047e-05, + "loss": 1.0684, + "step": 35985 + }, + { + "epoch": 2.66, + "learning_rate": 1.2304908079543471e-05, + "loss": 0.9189, + "step": 35986 + }, + { + "epoch": 2.66, + "learning_rate": 1.2304519945906816e-05, + "loss": 0.9928, + "step": 35987 + }, + { + "epoch": 2.66, + "learning_rate": 1.2304131808603707e-05, + "loss": 0.9648, + "step": 35988 + }, + { + "epoch": 2.66, + "learning_rate": 1.2303743667634755e-05, + "loss": 1.0311, + "step": 35989 + }, + { + "epoch": 2.66, + "learning_rate": 1.2303355523000581e-05, + "loss": 1.0423, + "step": 35990 + }, + { + "epoch": 2.66, + "learning_rate": 1.23029673747018e-05, + "loss": 0.9838, + "step": 35991 + }, + { + "epoch": 2.66, + "learning_rate": 1.2302579222739032e-05, + "loss": 1.0181, + "step": 35992 + }, + { + "epoch": 2.66, + "learning_rate": 1.2302191067112891e-05, + "loss": 0.9602, + "step": 35993 + }, + { + "epoch": 2.66, + "learning_rate": 1.2301802907824e-05, + "loss": 0.9888, + "step": 35994 + }, + { + "epoch": 2.66, + "learning_rate": 1.230141474487297e-05, + "loss": 1.0486, + "step": 35995 + }, + { + "epoch": 2.66, + "learning_rate": 1.2301026578260427e-05, + "loss": 0.9198, + "step": 35996 + }, + { + "epoch": 2.66, + "learning_rate": 1.230063840798698e-05, + "loss": 0.9288, + "step": 35997 + }, + { + "epoch": 2.66, + "learning_rate": 1.230025023405325e-05, + "loss": 0.9788, + "step": 35998 + }, + { + "epoch": 2.66, + "learning_rate": 1.2299862056459857e-05, + "loss": 1.0274, + "step": 35999 + }, + { + "epoch": 2.66, + "learning_rate": 1.2299473875207417e-05, + "loss": 1.0034, + "step": 36000 + }, + { + "epoch": 2.66, + "learning_rate": 1.2299085690296544e-05, + "loss": 1.0018, + "step": 36001 + }, + { + "epoch": 2.66, + "learning_rate": 1.229869750172786e-05, + "loss": 0.9597, + "step": 36002 + }, + { + "epoch": 2.66, + "learning_rate": 1.2298309309501982e-05, + "loss": 0.9717, + "step": 36003 + }, + { + "epoch": 2.66, + "learning_rate": 1.2297921113619524e-05, + "loss": 1.0376, + "step": 36004 + }, + { + "epoch": 2.66, + "learning_rate": 1.2297532914081107e-05, + "loss": 1.0975, + "step": 36005 + }, + { + "epoch": 2.66, + "learning_rate": 1.2297144710887353e-05, + "loss": 1.1003, + "step": 36006 + }, + { + "epoch": 2.66, + "learning_rate": 1.2296756504038866e-05, + "loss": 0.9222, + "step": 36007 + }, + { + "epoch": 2.66, + "learning_rate": 1.2296368293536283e-05, + "loss": 1.084, + "step": 36008 + }, + { + "epoch": 2.66, + "learning_rate": 1.2295980079380203e-05, + "loss": 1.0292, + "step": 36009 + }, + { + "epoch": 2.66, + "learning_rate": 1.2295591861571255e-05, + "loss": 1.063, + "step": 36010 + }, + { + "epoch": 2.66, + "learning_rate": 1.229520364011005e-05, + "loss": 0.9446, + "step": 36011 + }, + { + "epoch": 2.66, + "learning_rate": 1.229481541499721e-05, + "loss": 1.0038, + "step": 36012 + }, + { + "epoch": 2.66, + "learning_rate": 1.229442718623335e-05, + "loss": 1.128, + "step": 36013 + }, + { + "epoch": 2.66, + "learning_rate": 1.2294038953819094e-05, + "loss": 1.0123, + "step": 36014 + }, + { + "epoch": 2.66, + "learning_rate": 1.229365071775505e-05, + "loss": 0.8056, + "step": 36015 + }, + { + "epoch": 2.66, + "learning_rate": 1.2293262478041844e-05, + "loss": 1.0937, + "step": 36016 + }, + { + "epoch": 2.66, + "learning_rate": 1.229287423468009e-05, + "loss": 1.0094, + "step": 36017 + }, + { + "epoch": 2.66, + "learning_rate": 1.2292485987670403e-05, + "loss": 0.9826, + "step": 36018 + }, + { + "epoch": 2.66, + "learning_rate": 1.2292097737013407e-05, + "loss": 1.0117, + "step": 36019 + }, + { + "epoch": 2.66, + "learning_rate": 1.2291709482709715e-05, + "loss": 1.0503, + "step": 36020 + }, + { + "epoch": 2.66, + "learning_rate": 1.2291321224759943e-05, + "loss": 1.0293, + "step": 36021 + }, + { + "epoch": 2.66, + "learning_rate": 1.2290932963164714e-05, + "loss": 0.9591, + "step": 36022 + }, + { + "epoch": 2.66, + "learning_rate": 1.2290544697924644e-05, + "loss": 0.9806, + "step": 36023 + }, + { + "epoch": 2.66, + "learning_rate": 1.229015642904035e-05, + "loss": 0.9081, + "step": 36024 + }, + { + "epoch": 2.66, + "learning_rate": 1.2289768156512449e-05, + "loss": 0.9058, + "step": 36025 + }, + { + "epoch": 2.66, + "learning_rate": 1.2289379880341561e-05, + "loss": 0.9623, + "step": 36026 + }, + { + "epoch": 2.66, + "learning_rate": 1.2288991600528303e-05, + "loss": 1.081, + "step": 36027 + }, + { + "epoch": 2.66, + "learning_rate": 1.2288603317073288e-05, + "loss": 0.8669, + "step": 36028 + }, + { + "epoch": 2.66, + "learning_rate": 1.2288215029977145e-05, + "loss": 1.0365, + "step": 36029 + }, + { + "epoch": 2.66, + "learning_rate": 1.2287826739240479e-05, + "loss": 0.9794, + "step": 36030 + }, + { + "epoch": 2.66, + "learning_rate": 1.2287438444863912e-05, + "loss": 1.0477, + "step": 36031 + }, + { + "epoch": 2.66, + "learning_rate": 1.2287050146848067e-05, + "loss": 0.9427, + "step": 36032 + }, + { + "epoch": 2.66, + "learning_rate": 1.2286661845193556e-05, + "loss": 1.0903, + "step": 36033 + }, + { + "epoch": 2.66, + "learning_rate": 1.2286273539901002e-05, + "loss": 0.9739, + "step": 36034 + }, + { + "epoch": 2.66, + "learning_rate": 1.2285885230971015e-05, + "loss": 1.0699, + "step": 36035 + }, + { + "epoch": 2.66, + "learning_rate": 1.228549691840422e-05, + "loss": 1.0737, + "step": 36036 + }, + { + "epoch": 2.66, + "learning_rate": 1.2285108602201232e-05, + "loss": 0.9891, + "step": 36037 + }, + { + "epoch": 2.66, + "learning_rate": 1.2284720282362667e-05, + "loss": 0.8797, + "step": 36038 + }, + { + "epoch": 2.66, + "learning_rate": 1.2284331958889149e-05, + "loss": 1.0071, + "step": 36039 + }, + { + "epoch": 2.66, + "learning_rate": 1.2283943631781288e-05, + "loss": 0.9573, + "step": 36040 + }, + { + "epoch": 2.66, + "learning_rate": 1.2283555301039706e-05, + "loss": 1.03, + "step": 36041 + }, + { + "epoch": 2.66, + "learning_rate": 1.228316696666502e-05, + "loss": 1.0347, + "step": 36042 + }, + { + "epoch": 2.66, + "learning_rate": 1.2282778628657848e-05, + "loss": 0.8833, + "step": 36043 + }, + { + "epoch": 2.66, + "learning_rate": 1.2282390287018805e-05, + "loss": 1.0065, + "step": 36044 + }, + { + "epoch": 2.66, + "learning_rate": 1.228200194174852e-05, + "loss": 0.9417, + "step": 36045 + }, + { + "epoch": 2.66, + "learning_rate": 1.2281613592847594e-05, + "loss": 1.0139, + "step": 36046 + }, + { + "epoch": 2.66, + "learning_rate": 1.228122524031666e-05, + "loss": 0.8961, + "step": 36047 + }, + { + "epoch": 2.66, + "learning_rate": 1.2280836884156324e-05, + "loss": 1.1003, + "step": 36048 + }, + { + "epoch": 2.66, + "learning_rate": 1.2280448524367213e-05, + "loss": 1.0594, + "step": 36049 + }, + { + "epoch": 2.66, + "learning_rate": 1.2280060160949936e-05, + "loss": 0.9742, + "step": 36050 + }, + { + "epoch": 2.66, + "learning_rate": 1.227967179390512e-05, + "loss": 0.9476, + "step": 36051 + }, + { + "epoch": 2.66, + "learning_rate": 1.2279283423233376e-05, + "loss": 1.0399, + "step": 36052 + }, + { + "epoch": 2.66, + "learning_rate": 1.227889504893533e-05, + "loss": 0.9519, + "step": 36053 + }, + { + "epoch": 2.66, + "learning_rate": 1.227850667101159e-05, + "loss": 0.966, + "step": 36054 + }, + { + "epoch": 2.66, + "learning_rate": 1.227811828946278e-05, + "loss": 1.1078, + "step": 36055 + }, + { + "epoch": 2.66, + "learning_rate": 1.2277729904289517e-05, + "loss": 0.9229, + "step": 36056 + }, + { + "epoch": 2.66, + "learning_rate": 1.2277341515492418e-05, + "loss": 1.0493, + "step": 36057 + }, + { + "epoch": 2.66, + "learning_rate": 1.22769531230721e-05, + "loss": 1.0337, + "step": 36058 + }, + { + "epoch": 2.66, + "learning_rate": 1.2276564727029184e-05, + "loss": 0.9763, + "step": 36059 + }, + { + "epoch": 2.66, + "learning_rate": 1.2276176327364282e-05, + "loss": 0.9234, + "step": 36060 + }, + { + "epoch": 2.66, + "learning_rate": 1.227578792407802e-05, + "loss": 1.0638, + "step": 36061 + }, + { + "epoch": 2.66, + "learning_rate": 1.227539951717101e-05, + "loss": 1.0349, + "step": 36062 + }, + { + "epoch": 2.66, + "learning_rate": 1.2275011106643873e-05, + "loss": 0.9979, + "step": 36063 + }, + { + "epoch": 2.66, + "learning_rate": 1.2274622692497227e-05, + "loss": 0.9651, + "step": 36064 + }, + { + "epoch": 2.66, + "learning_rate": 1.2274234274731689e-05, + "loss": 1.0251, + "step": 36065 + }, + { + "epoch": 2.66, + "learning_rate": 1.2273845853347875e-05, + "loss": 0.9291, + "step": 36066 + }, + { + "epoch": 2.66, + "learning_rate": 1.2273457428346404e-05, + "loss": 0.9035, + "step": 36067 + }, + { + "epoch": 2.67, + "learning_rate": 1.2273068999727895e-05, + "loss": 0.9829, + "step": 36068 + }, + { + "epoch": 2.67, + "learning_rate": 1.2272680567492968e-05, + "loss": 0.9777, + "step": 36069 + }, + { + "epoch": 2.67, + "learning_rate": 1.2272292131642236e-05, + "loss": 1.0506, + "step": 36070 + }, + { + "epoch": 2.67, + "learning_rate": 1.227190369217632e-05, + "loss": 1.0535, + "step": 36071 + }, + { + "epoch": 2.67, + "learning_rate": 1.227151524909584e-05, + "loss": 0.9052, + "step": 36072 + }, + { + "epoch": 2.67, + "learning_rate": 1.2271126802401412e-05, + "loss": 0.9477, + "step": 36073 + }, + { + "epoch": 2.67, + "learning_rate": 1.2270738352093653e-05, + "loss": 1.02, + "step": 36074 + }, + { + "epoch": 2.67, + "learning_rate": 1.2270349898173179e-05, + "loss": 1.0531, + "step": 36075 + }, + { + "epoch": 2.67, + "learning_rate": 1.2269961440640615e-05, + "loss": 0.9934, + "step": 36076 + }, + { + "epoch": 2.67, + "learning_rate": 1.226957297949657e-05, + "loss": 0.9928, + "step": 36077 + }, + { + "epoch": 2.67, + "learning_rate": 1.226918451474167e-05, + "loss": 1.0086, + "step": 36078 + }, + { + "epoch": 2.67, + "learning_rate": 1.226879604637653e-05, + "loss": 1.0661, + "step": 36079 + }, + { + "epoch": 2.67, + "learning_rate": 1.2268407574401766e-05, + "loss": 1.047, + "step": 36080 + }, + { + "epoch": 2.67, + "learning_rate": 1.2268019098818e-05, + "loss": 0.9167, + "step": 36081 + }, + { + "epoch": 2.67, + "learning_rate": 1.2267630619625848e-05, + "loss": 0.9779, + "step": 36082 + }, + { + "epoch": 2.67, + "learning_rate": 1.2267242136825928e-05, + "loss": 1.0141, + "step": 36083 + }, + { + "epoch": 2.67, + "learning_rate": 1.2266853650418859e-05, + "loss": 0.9377, + "step": 36084 + }, + { + "epoch": 2.67, + "learning_rate": 1.2266465160405254e-05, + "loss": 0.8921, + "step": 36085 + }, + { + "epoch": 2.67, + "learning_rate": 1.226607666678574e-05, + "loss": 1.06, + "step": 36086 + }, + { + "epoch": 2.67, + "learning_rate": 1.2265688169560928e-05, + "loss": 1.0296, + "step": 36087 + }, + { + "epoch": 2.67, + "learning_rate": 1.2265299668731441e-05, + "loss": 0.945, + "step": 36088 + }, + { + "epoch": 2.67, + "learning_rate": 1.2264911164297892e-05, + "loss": 1.0184, + "step": 36089 + }, + { + "epoch": 2.67, + "learning_rate": 1.2264522656260901e-05, + "loss": 0.8975, + "step": 36090 + }, + { + "epoch": 2.67, + "learning_rate": 1.2264134144621089e-05, + "loss": 1.0763, + "step": 36091 + }, + { + "epoch": 2.67, + "learning_rate": 1.2263745629379075e-05, + "loss": 0.9364, + "step": 36092 + }, + { + "epoch": 2.67, + "learning_rate": 1.226335711053547e-05, + "loss": 1.0352, + "step": 36093 + }, + { + "epoch": 2.67, + "learning_rate": 1.2262968588090898e-05, + "loss": 0.9547, + "step": 36094 + }, + { + "epoch": 2.67, + "learning_rate": 1.2262580062045972e-05, + "loss": 0.9781, + "step": 36095 + }, + { + "epoch": 2.67, + "learning_rate": 1.2262191532401317e-05, + "loss": 1.0383, + "step": 36096 + }, + { + "epoch": 2.67, + "learning_rate": 1.2261802999157544e-05, + "loss": 1.1332, + "step": 36097 + }, + { + "epoch": 2.67, + "learning_rate": 1.226141446231528e-05, + "loss": 0.9435, + "step": 36098 + }, + { + "epoch": 2.67, + "learning_rate": 1.2261025921875134e-05, + "loss": 0.9762, + "step": 36099 + }, + { + "epoch": 2.67, + "learning_rate": 1.226063737783773e-05, + "loss": 0.8859, + "step": 36100 + }, + { + "epoch": 2.67, + "learning_rate": 1.2260248830203683e-05, + "loss": 0.9092, + "step": 36101 + }, + { + "epoch": 2.67, + "learning_rate": 1.2259860278973615e-05, + "loss": 1.0074, + "step": 36102 + }, + { + "epoch": 2.67, + "learning_rate": 1.2259471724148137e-05, + "loss": 1.0321, + "step": 36103 + }, + { + "epoch": 2.67, + "learning_rate": 1.2259083165727876e-05, + "loss": 1.0619, + "step": 36104 + }, + { + "epoch": 2.67, + "learning_rate": 1.2258694603713443e-05, + "loss": 1.0803, + "step": 36105 + }, + { + "epoch": 2.67, + "learning_rate": 1.2258306038105463e-05, + "loss": 1.0361, + "step": 36106 + }, + { + "epoch": 2.67, + "learning_rate": 1.2257917468904546e-05, + "loss": 0.8901, + "step": 36107 + }, + { + "epoch": 2.67, + "learning_rate": 1.2257528896111318e-05, + "loss": 1.0765, + "step": 36108 + }, + { + "epoch": 2.67, + "learning_rate": 1.225714031972639e-05, + "loss": 1.0127, + "step": 36109 + }, + { + "epoch": 2.67, + "learning_rate": 1.2256751739750387e-05, + "loss": 0.918, + "step": 36110 + }, + { + "epoch": 2.67, + "learning_rate": 1.2256363156183926e-05, + "loss": 1.0209, + "step": 36111 + }, + { + "epoch": 2.67, + "learning_rate": 1.2255974569027621e-05, + "loss": 1.0975, + "step": 36112 + }, + { + "epoch": 2.67, + "learning_rate": 1.2255585978282092e-05, + "loss": 1.0299, + "step": 36113 + }, + { + "epoch": 2.67, + "learning_rate": 1.225519738394796e-05, + "loss": 0.9752, + "step": 36114 + }, + { + "epoch": 2.67, + "learning_rate": 1.225480878602584e-05, + "loss": 0.9021, + "step": 36115 + }, + { + "epoch": 2.67, + "learning_rate": 1.2254420184516352e-05, + "loss": 0.9198, + "step": 36116 + }, + { + "epoch": 2.67, + "learning_rate": 1.225403157942011e-05, + "loss": 1.0437, + "step": 36117 + }, + { + "epoch": 2.67, + "learning_rate": 1.2253642970737743e-05, + "loss": 0.9938, + "step": 36118 + }, + { + "epoch": 2.67, + "learning_rate": 1.2253254358469857e-05, + "loss": 1.0226, + "step": 36119 + }, + { + "epoch": 2.67, + "learning_rate": 1.2252865742617077e-05, + "loss": 1.0424, + "step": 36120 + }, + { + "epoch": 2.67, + "learning_rate": 1.2252477123180023e-05, + "loss": 1.065, + "step": 36121 + }, + { + "epoch": 2.67, + "learning_rate": 1.2252088500159304e-05, + "loss": 0.9916, + "step": 36122 + }, + { + "epoch": 2.67, + "learning_rate": 1.225169987355555e-05, + "loss": 0.96, + "step": 36123 + }, + { + "epoch": 2.67, + "learning_rate": 1.2251311243369373e-05, + "loss": 0.962, + "step": 36124 + }, + { + "epoch": 2.67, + "learning_rate": 1.2250922609601388e-05, + "loss": 1.0906, + "step": 36125 + }, + { + "epoch": 2.67, + "learning_rate": 1.225053397225222e-05, + "loss": 0.9229, + "step": 36126 + }, + { + "epoch": 2.67, + "learning_rate": 1.2250145331322485e-05, + "loss": 1.0056, + "step": 36127 + }, + { + "epoch": 2.67, + "learning_rate": 1.22497566868128e-05, + "loss": 0.9683, + "step": 36128 + }, + { + "epoch": 2.67, + "learning_rate": 1.2249368038723784e-05, + "loss": 1.0375, + "step": 36129 + }, + { + "epoch": 2.67, + "learning_rate": 1.2248979387056057e-05, + "loss": 1.0592, + "step": 36130 + }, + { + "epoch": 2.67, + "learning_rate": 1.2248590731810236e-05, + "loss": 0.9114, + "step": 36131 + }, + { + "epoch": 2.67, + "learning_rate": 1.2248202072986936e-05, + "loss": 1.0596, + "step": 36132 + }, + { + "epoch": 2.67, + "learning_rate": 1.2247813410586784e-05, + "loss": 0.9158, + "step": 36133 + }, + { + "epoch": 2.67, + "learning_rate": 1.224742474461039e-05, + "loss": 1.0617, + "step": 36134 + }, + { + "epoch": 2.67, + "learning_rate": 1.2247036075058376e-05, + "loss": 0.9422, + "step": 36135 + }, + { + "epoch": 2.67, + "learning_rate": 1.2246647401931358e-05, + "loss": 0.9513, + "step": 36136 + }, + { + "epoch": 2.67, + "learning_rate": 1.2246258725229958e-05, + "loss": 1.0425, + "step": 36137 + }, + { + "epoch": 2.67, + "learning_rate": 1.224587004495479e-05, + "loss": 0.9471, + "step": 36138 + }, + { + "epoch": 2.67, + "learning_rate": 1.2245481361106481e-05, + "loss": 0.948, + "step": 36139 + }, + { + "epoch": 2.67, + "learning_rate": 1.2245092673685637e-05, + "loss": 0.9825, + "step": 36140 + }, + { + "epoch": 2.67, + "learning_rate": 1.2244703982692887e-05, + "loss": 1.0636, + "step": 36141 + }, + { + "epoch": 2.67, + "learning_rate": 1.2244315288128842e-05, + "loss": 0.9317, + "step": 36142 + }, + { + "epoch": 2.67, + "learning_rate": 1.2243926589994124e-05, + "loss": 1.041, + "step": 36143 + }, + { + "epoch": 2.67, + "learning_rate": 1.2243537888289352e-05, + "loss": 1.0076, + "step": 36144 + }, + { + "epoch": 2.67, + "learning_rate": 1.2243149183015141e-05, + "loss": 1.0274, + "step": 36145 + }, + { + "epoch": 2.67, + "learning_rate": 1.2242760474172113e-05, + "loss": 1.0063, + "step": 36146 + }, + { + "epoch": 2.67, + "learning_rate": 1.2242371761760887e-05, + "loss": 0.9632, + "step": 36147 + }, + { + "epoch": 2.67, + "learning_rate": 1.2241983045782079e-05, + "loss": 1.0084, + "step": 36148 + }, + { + "epoch": 2.67, + "learning_rate": 1.2241594326236308e-05, + "loss": 0.9826, + "step": 36149 + }, + { + "epoch": 2.67, + "learning_rate": 1.2241205603124192e-05, + "loss": 0.9908, + "step": 36150 + }, + { + "epoch": 2.67, + "learning_rate": 1.2240816876446351e-05, + "loss": 1.0811, + "step": 36151 + }, + { + "epoch": 2.67, + "learning_rate": 1.2240428146203398e-05, + "loss": 0.8655, + "step": 36152 + }, + { + "epoch": 2.67, + "learning_rate": 1.224003941239596e-05, + "loss": 0.9273, + "step": 36153 + }, + { + "epoch": 2.67, + "learning_rate": 1.2239650675024651e-05, + "loss": 1.0171, + "step": 36154 + }, + { + "epoch": 2.67, + "learning_rate": 1.223926193409009e-05, + "loss": 0.9813, + "step": 36155 + }, + { + "epoch": 2.67, + "learning_rate": 1.2238873189592892e-05, + "loss": 1.0568, + "step": 36156 + }, + { + "epoch": 2.67, + "learning_rate": 1.2238484441533682e-05, + "loss": 1.0556, + "step": 36157 + }, + { + "epoch": 2.67, + "learning_rate": 1.2238095689913076e-05, + "loss": 0.9993, + "step": 36158 + }, + { + "epoch": 2.67, + "learning_rate": 1.223770693473169e-05, + "loss": 0.9558, + "step": 36159 + }, + { + "epoch": 2.67, + "learning_rate": 1.2237318175990144e-05, + "loss": 0.9901, + "step": 36160 + }, + { + "epoch": 2.67, + "learning_rate": 1.2236929413689057e-05, + "loss": 0.9791, + "step": 36161 + }, + { + "epoch": 2.67, + "learning_rate": 1.2236540647829046e-05, + "loss": 1.0722, + "step": 36162 + }, + { + "epoch": 2.67, + "learning_rate": 1.2236151878410736e-05, + "loss": 0.9838, + "step": 36163 + }, + { + "epoch": 2.67, + "learning_rate": 1.2235763105434736e-05, + "loss": 0.9488, + "step": 36164 + }, + { + "epoch": 2.67, + "learning_rate": 1.223537432890167e-05, + "loss": 1.0848, + "step": 36165 + }, + { + "epoch": 2.67, + "learning_rate": 1.2234985548812155e-05, + "loss": 0.9036, + "step": 36166 + }, + { + "epoch": 2.67, + "learning_rate": 1.2234596765166814e-05, + "loss": 1.0024, + "step": 36167 + }, + { + "epoch": 2.67, + "learning_rate": 1.2234207977966255e-05, + "loss": 0.9283, + "step": 36168 + }, + { + "epoch": 2.67, + "learning_rate": 1.2233819187211107e-05, + "loss": 0.9062, + "step": 36169 + }, + { + "epoch": 2.67, + "learning_rate": 1.2233430392901981e-05, + "loss": 0.9251, + "step": 36170 + }, + { + "epoch": 2.67, + "learning_rate": 1.2233041595039504e-05, + "loss": 0.8888, + "step": 36171 + }, + { + "epoch": 2.67, + "learning_rate": 1.2232652793624285e-05, + "loss": 0.9956, + "step": 36172 + }, + { + "epoch": 2.67, + "learning_rate": 1.2232263988656951e-05, + "loss": 0.9654, + "step": 36173 + }, + { + "epoch": 2.67, + "learning_rate": 1.2231875180138113e-05, + "loss": 1.0369, + "step": 36174 + }, + { + "epoch": 2.67, + "learning_rate": 1.2231486368068399e-05, + "loss": 0.9662, + "step": 36175 + }, + { + "epoch": 2.67, + "learning_rate": 1.223109755244842e-05, + "loss": 1.067, + "step": 36176 + }, + { + "epoch": 2.67, + "learning_rate": 1.2230708733278794e-05, + "loss": 0.9848, + "step": 36177 + }, + { + "epoch": 2.67, + "learning_rate": 1.2230319910560145e-05, + "loss": 1.028, + "step": 36178 + }, + { + "epoch": 2.67, + "learning_rate": 1.2229931084293088e-05, + "loss": 1.1021, + "step": 36179 + }, + { + "epoch": 2.67, + "learning_rate": 1.2229542254478244e-05, + "loss": 0.9969, + "step": 36180 + }, + { + "epoch": 2.67, + "learning_rate": 1.2229153421116227e-05, + "loss": 0.9724, + "step": 36181 + }, + { + "epoch": 2.67, + "learning_rate": 1.2228764584207662e-05, + "loss": 1.0249, + "step": 36182 + }, + { + "epoch": 2.67, + "learning_rate": 1.2228375743753163e-05, + "loss": 0.9935, + "step": 36183 + }, + { + "epoch": 2.67, + "learning_rate": 1.222798689975335e-05, + "loss": 0.9669, + "step": 36184 + }, + { + "epoch": 2.67, + "learning_rate": 1.2227598052208842e-05, + "loss": 0.9925, + "step": 36185 + }, + { + "epoch": 2.67, + "learning_rate": 1.222720920112026e-05, + "loss": 1.0388, + "step": 36186 + }, + { + "epoch": 2.67, + "learning_rate": 1.2226820346488215e-05, + "loss": 0.8853, + "step": 36187 + }, + { + "epoch": 2.67, + "learning_rate": 1.2226431488313336e-05, + "loss": 1.0407, + "step": 36188 + }, + { + "epoch": 2.67, + "learning_rate": 1.222604262659623e-05, + "loss": 0.9517, + "step": 36189 + }, + { + "epoch": 2.67, + "learning_rate": 1.2225653761337527e-05, + "loss": 1.0726, + "step": 36190 + }, + { + "epoch": 2.67, + "learning_rate": 1.2225264892537839e-05, + "loss": 1.0499, + "step": 36191 + }, + { + "epoch": 2.67, + "learning_rate": 1.2224876020197787e-05, + "loss": 0.953, + "step": 36192 + }, + { + "epoch": 2.67, + "learning_rate": 1.2224487144317988e-05, + "loss": 0.9385, + "step": 36193 + }, + { + "epoch": 2.67, + "learning_rate": 1.2224098264899063e-05, + "loss": 1.0158, + "step": 36194 + }, + { + "epoch": 2.67, + "learning_rate": 1.222370938194163e-05, + "loss": 0.9342, + "step": 36195 + }, + { + "epoch": 2.67, + "learning_rate": 1.2223320495446307e-05, + "loss": 0.9755, + "step": 36196 + }, + { + "epoch": 2.67, + "learning_rate": 1.222293160541371e-05, + "loss": 0.8528, + "step": 36197 + }, + { + "epoch": 2.67, + "learning_rate": 1.2222542711844463e-05, + "loss": 0.975, + "step": 36198 + }, + { + "epoch": 2.67, + "learning_rate": 1.2222153814739182e-05, + "loss": 0.8812, + "step": 36199 + }, + { + "epoch": 2.67, + "learning_rate": 1.2221764914098485e-05, + "loss": 0.9742, + "step": 36200 + }, + { + "epoch": 2.67, + "learning_rate": 1.2221376009922993e-05, + "loss": 1.0742, + "step": 36201 + }, + { + "epoch": 2.67, + "learning_rate": 1.2220987102213324e-05, + "loss": 0.9879, + "step": 36202 + }, + { + "epoch": 2.68, + "learning_rate": 1.2220598190970094e-05, + "loss": 1.062, + "step": 36203 + }, + { + "epoch": 2.68, + "learning_rate": 1.222020927619393e-05, + "loss": 0.9822, + "step": 36204 + }, + { + "epoch": 2.68, + "learning_rate": 1.221982035788544e-05, + "loss": 0.9776, + "step": 36205 + }, + { + "epoch": 2.68, + "learning_rate": 1.2219431436045248e-05, + "loss": 1.0519, + "step": 36206 + }, + { + "epoch": 2.68, + "learning_rate": 1.2219042510673971e-05, + "loss": 0.9617, + "step": 36207 + }, + { + "epoch": 2.68, + "learning_rate": 1.2218653581772234e-05, + "loss": 0.9293, + "step": 36208 + }, + { + "epoch": 2.68, + "learning_rate": 1.2218264649340644e-05, + "loss": 1.0249, + "step": 36209 + }, + { + "epoch": 2.68, + "learning_rate": 1.2217875713379832e-05, + "loss": 0.9443, + "step": 36210 + }, + { + "epoch": 2.68, + "learning_rate": 1.2217486773890408e-05, + "loss": 0.9929, + "step": 36211 + }, + { + "epoch": 2.68, + "learning_rate": 1.2217097830872997e-05, + "loss": 0.9948, + "step": 36212 + }, + { + "epoch": 2.68, + "learning_rate": 1.2216708884328212e-05, + "loss": 1.046, + "step": 36213 + }, + { + "epoch": 2.68, + "learning_rate": 1.2216319934256679e-05, + "loss": 1.037, + "step": 36214 + }, + { + "epoch": 2.68, + "learning_rate": 1.221593098065901e-05, + "loss": 0.9978, + "step": 36215 + }, + { + "epoch": 2.68, + "learning_rate": 1.2215542023535829e-05, + "loss": 1.0912, + "step": 36216 + }, + { + "epoch": 2.68, + "learning_rate": 1.2215153062887749e-05, + "loss": 1.0616, + "step": 36217 + }, + { + "epoch": 2.68, + "learning_rate": 1.2214764098715395e-05, + "loss": 1.0316, + "step": 36218 + }, + { + "epoch": 2.68, + "learning_rate": 1.2214375131019381e-05, + "loss": 0.9995, + "step": 36219 + }, + { + "epoch": 2.68, + "learning_rate": 1.2213986159800328e-05, + "loss": 0.9845, + "step": 36220 + }, + { + "epoch": 2.68, + "learning_rate": 1.2213597185058855e-05, + "loss": 0.9865, + "step": 36221 + }, + { + "epoch": 2.68, + "learning_rate": 1.2213208206795581e-05, + "loss": 1.0373, + "step": 36222 + }, + { + "epoch": 2.68, + "learning_rate": 1.2212819225011126e-05, + "loss": 1.0016, + "step": 36223 + }, + { + "epoch": 2.68, + "learning_rate": 1.2212430239706103e-05, + "loss": 1.0887, + "step": 36224 + }, + { + "epoch": 2.68, + "learning_rate": 1.221204125088114e-05, + "loss": 0.9553, + "step": 36225 + }, + { + "epoch": 2.68, + "learning_rate": 1.2211652258536845e-05, + "loss": 1.0707, + "step": 36226 + }, + { + "epoch": 2.68, + "learning_rate": 1.2211263262673846e-05, + "loss": 1.061, + "step": 36227 + }, + { + "epoch": 2.68, + "learning_rate": 1.2210874263292762e-05, + "loss": 1.0179, + "step": 36228 + }, + { + "epoch": 2.68, + "learning_rate": 1.2210485260394205e-05, + "loss": 0.9658, + "step": 36229 + }, + { + "epoch": 2.68, + "learning_rate": 1.2210096253978798e-05, + "loss": 0.9826, + "step": 36230 + }, + { + "epoch": 2.68, + "learning_rate": 1.220970724404716e-05, + "loss": 1.0285, + "step": 36231 + }, + { + "epoch": 2.68, + "learning_rate": 1.2209318230599912e-05, + "loss": 1.0261, + "step": 36232 + }, + { + "epoch": 2.68, + "learning_rate": 1.2208929213637668e-05, + "loss": 1.0085, + "step": 36233 + }, + { + "epoch": 2.68, + "learning_rate": 1.2208540193161047e-05, + "loss": 0.9527, + "step": 36234 + }, + { + "epoch": 2.68, + "learning_rate": 1.2208151169170673e-05, + "loss": 0.9296, + "step": 36235 + }, + { + "epoch": 2.68, + "learning_rate": 1.220776214166716e-05, + "loss": 0.9741, + "step": 36236 + }, + { + "epoch": 2.68, + "learning_rate": 1.2207373110651131e-05, + "loss": 1.0493, + "step": 36237 + }, + { + "epoch": 2.68, + "learning_rate": 1.2206984076123201e-05, + "loss": 1.0045, + "step": 36238 + }, + { + "epoch": 2.68, + "learning_rate": 1.2206595038083993e-05, + "loss": 1.0332, + "step": 36239 + }, + { + "epoch": 2.68, + "learning_rate": 1.2206205996534122e-05, + "loss": 0.9176, + "step": 36240 + }, + { + "epoch": 2.68, + "learning_rate": 1.220581695147421e-05, + "loss": 0.9578, + "step": 36241 + }, + { + "epoch": 2.68, + "learning_rate": 1.2205427902904874e-05, + "loss": 1.004, + "step": 36242 + }, + { + "epoch": 2.68, + "learning_rate": 1.2205038850826736e-05, + "loss": 1.0505, + "step": 36243 + }, + { + "epoch": 2.68, + "learning_rate": 1.220464979524041e-05, + "loss": 1.0101, + "step": 36244 + }, + { + "epoch": 2.68, + "learning_rate": 1.220426073614652e-05, + "loss": 0.9732, + "step": 36245 + }, + { + "epoch": 2.68, + "learning_rate": 1.220387167354568e-05, + "loss": 1.0975, + "step": 36246 + }, + { + "epoch": 2.68, + "learning_rate": 1.2203482607438514e-05, + "loss": 1.0545, + "step": 36247 + }, + { + "epoch": 2.68, + "learning_rate": 1.2203093537825636e-05, + "loss": 1.0879, + "step": 36248 + }, + { + "epoch": 2.68, + "learning_rate": 1.220270446470767e-05, + "loss": 1.0108, + "step": 36249 + }, + { + "epoch": 2.68, + "learning_rate": 1.2202315388085232e-05, + "loss": 1.0487, + "step": 36250 + }, + { + "epoch": 2.68, + "learning_rate": 1.2201926307958945e-05, + "loss": 0.9294, + "step": 36251 + }, + { + "epoch": 2.68, + "learning_rate": 1.220153722432942e-05, + "loss": 0.8559, + "step": 36252 + }, + { + "epoch": 2.68, + "learning_rate": 1.2201148137197283e-05, + "loss": 0.977, + "step": 36253 + }, + { + "epoch": 2.68, + "learning_rate": 1.2200759046563149e-05, + "loss": 1.0017, + "step": 36254 + }, + { + "epoch": 2.68, + "learning_rate": 1.2200369952427642e-05, + "loss": 1.0157, + "step": 36255 + }, + { + "epoch": 2.68, + "learning_rate": 1.2199980854791376e-05, + "loss": 1.0639, + "step": 36256 + }, + { + "epoch": 2.68, + "learning_rate": 1.2199591753654971e-05, + "loss": 0.9631, + "step": 36257 + }, + { + "epoch": 2.68, + "learning_rate": 1.2199202649019047e-05, + "loss": 0.9325, + "step": 36258 + }, + { + "epoch": 2.68, + "learning_rate": 1.2198813540884225e-05, + "loss": 0.9677, + "step": 36259 + }, + { + "epoch": 2.68, + "learning_rate": 1.2198424429251123e-05, + "loss": 0.9624, + "step": 36260 + }, + { + "epoch": 2.68, + "learning_rate": 1.2198035314120357e-05, + "loss": 1.0049, + "step": 36261 + }, + { + "epoch": 2.68, + "learning_rate": 1.2197646195492547e-05, + "loss": 1.0502, + "step": 36262 + }, + { + "epoch": 2.68, + "learning_rate": 1.2197257073368317e-05, + "loss": 1.0289, + "step": 36263 + }, + { + "epoch": 2.68, + "learning_rate": 1.2196867947748278e-05, + "loss": 0.9573, + "step": 36264 + }, + { + "epoch": 2.68, + "learning_rate": 1.2196478818633058e-05, + "loss": 1.008, + "step": 36265 + }, + { + "epoch": 2.68, + "learning_rate": 1.219608968602327e-05, + "loss": 0.9059, + "step": 36266 + }, + { + "epoch": 2.68, + "learning_rate": 1.2195700549919533e-05, + "loss": 0.9532, + "step": 36267 + }, + { + "epoch": 2.68, + "learning_rate": 1.2195311410322468e-05, + "loss": 1.1132, + "step": 36268 + }, + { + "epoch": 2.68, + "learning_rate": 1.2194922267232696e-05, + "loss": 0.8807, + "step": 36269 + }, + { + "epoch": 2.68, + "learning_rate": 1.2194533120650833e-05, + "loss": 1.0792, + "step": 36270 + }, + { + "epoch": 2.68, + "learning_rate": 1.2194143970577498e-05, + "loss": 1.109, + "step": 36271 + }, + { + "epoch": 2.68, + "learning_rate": 1.2193754817013314e-05, + "loss": 0.9031, + "step": 36272 + }, + { + "epoch": 2.68, + "learning_rate": 1.2193365659958897e-05, + "loss": 0.961, + "step": 36273 + }, + { + "epoch": 2.68, + "learning_rate": 1.2192976499414865e-05, + "loss": 1.0242, + "step": 36274 + }, + { + "epoch": 2.68, + "learning_rate": 1.219258733538184e-05, + "loss": 1.0714, + "step": 36275 + }, + { + "epoch": 2.68, + "learning_rate": 1.2192198167860439e-05, + "loss": 1.0665, + "step": 36276 + }, + { + "epoch": 2.68, + "learning_rate": 1.2191808996851284e-05, + "loss": 0.9886, + "step": 36277 + }, + { + "epoch": 2.68, + "learning_rate": 1.2191419822354988e-05, + "loss": 0.9253, + "step": 36278 + }, + { + "epoch": 2.68, + "learning_rate": 1.2191030644372179e-05, + "loss": 0.9174, + "step": 36279 + }, + { + "epoch": 2.68, + "learning_rate": 1.219064146290347e-05, + "loss": 1.1028, + "step": 36280 + }, + { + "epoch": 2.68, + "learning_rate": 1.2190252277949482e-05, + "loss": 0.9621, + "step": 36281 + }, + { + "epoch": 2.68, + "learning_rate": 1.2189863089510834e-05, + "loss": 1.1136, + "step": 36282 + }, + { + "epoch": 2.68, + "learning_rate": 1.2189473897588143e-05, + "loss": 1.0549, + "step": 36283 + }, + { + "epoch": 2.68, + "learning_rate": 1.2189084702182032e-05, + "loss": 1.104, + "step": 36284 + }, + { + "epoch": 2.68, + "learning_rate": 1.2188695503293117e-05, + "loss": 1.0637, + "step": 36285 + }, + { + "epoch": 2.68, + "learning_rate": 1.2188306300922022e-05, + "loss": 1.0952, + "step": 36286 + }, + { + "epoch": 2.68, + "learning_rate": 1.2187917095069359e-05, + "loss": 1.0405, + "step": 36287 + }, + { + "epoch": 2.68, + "learning_rate": 1.2187527885735757e-05, + "loss": 1.0407, + "step": 36288 + }, + { + "epoch": 2.68, + "learning_rate": 1.2187138672921826e-05, + "loss": 1.0075, + "step": 36289 + }, + { + "epoch": 2.68, + "learning_rate": 1.2186749456628188e-05, + "loss": 0.9725, + "step": 36290 + }, + { + "epoch": 2.68, + "learning_rate": 1.2186360236855464e-05, + "loss": 1.0816, + "step": 36291 + }, + { + "epoch": 2.68, + "learning_rate": 1.2185971013604271e-05, + "loss": 0.9604, + "step": 36292 + }, + { + "epoch": 2.68, + "learning_rate": 1.218558178687523e-05, + "loss": 0.9662, + "step": 36293 + }, + { + "epoch": 2.68, + "learning_rate": 1.2185192556668962e-05, + "loss": 0.9917, + "step": 36294 + }, + { + "epoch": 2.68, + "learning_rate": 1.2184803322986079e-05, + "loss": 0.96, + "step": 36295 + }, + { + "epoch": 2.68, + "learning_rate": 1.2184414085827209e-05, + "loss": 0.9803, + "step": 36296 + }, + { + "epoch": 2.68, + "learning_rate": 1.2184024845192966e-05, + "loss": 1.0809, + "step": 36297 + }, + { + "epoch": 2.68, + "learning_rate": 1.2183635601083972e-05, + "loss": 0.9925, + "step": 36298 + }, + { + "epoch": 2.68, + "learning_rate": 1.2183246353500843e-05, + "loss": 1.0545, + "step": 36299 + }, + { + "epoch": 2.68, + "learning_rate": 1.2182857102444204e-05, + "loss": 1.0191, + "step": 36300 + }, + { + "epoch": 2.68, + "learning_rate": 1.2182467847914665e-05, + "loss": 1.1139, + "step": 36301 + }, + { + "epoch": 2.68, + "learning_rate": 1.2182078589912855e-05, + "loss": 0.9086, + "step": 36302 + }, + { + "epoch": 2.68, + "learning_rate": 1.2181689328439386e-05, + "loss": 1.0728, + "step": 36303 + }, + { + "epoch": 2.68, + "learning_rate": 1.2181300063494885e-05, + "loss": 0.9325, + "step": 36304 + }, + { + "epoch": 2.68, + "learning_rate": 1.218091079507996e-05, + "loss": 0.965, + "step": 36305 + }, + { + "epoch": 2.68, + "learning_rate": 1.2180521523195244e-05, + "loss": 0.9716, + "step": 36306 + }, + { + "epoch": 2.68, + "learning_rate": 1.2180132247841347e-05, + "loss": 0.8402, + "step": 36307 + }, + { + "epoch": 2.68, + "learning_rate": 1.2179742969018893e-05, + "loss": 0.9908, + "step": 36308 + }, + { + "epoch": 2.68, + "learning_rate": 1.2179353686728495e-05, + "loss": 0.994, + "step": 36309 + }, + { + "epoch": 2.68, + "learning_rate": 1.2178964400970779e-05, + "loss": 0.9604, + "step": 36310 + }, + { + "epoch": 2.68, + "learning_rate": 1.217857511174636e-05, + "loss": 1.0917, + "step": 36311 + }, + { + "epoch": 2.68, + "learning_rate": 1.2178185819055861e-05, + "loss": 0.9301, + "step": 36312 + }, + { + "epoch": 2.68, + "learning_rate": 1.21777965228999e-05, + "loss": 0.9433, + "step": 36313 + }, + { + "epoch": 2.68, + "learning_rate": 1.2177407223279092e-05, + "loss": 0.946, + "step": 36314 + }, + { + "epoch": 2.68, + "learning_rate": 1.2177017920194063e-05, + "loss": 0.9756, + "step": 36315 + }, + { + "epoch": 2.68, + "learning_rate": 1.2176628613645429e-05, + "loss": 0.9492, + "step": 36316 + }, + { + "epoch": 2.68, + "learning_rate": 1.2176239303633812e-05, + "loss": 0.987, + "step": 36317 + }, + { + "epoch": 2.68, + "learning_rate": 1.217584999015983e-05, + "loss": 0.9347, + "step": 36318 + }, + { + "epoch": 2.68, + "learning_rate": 1.2175460673224098e-05, + "loss": 1.0083, + "step": 36319 + }, + { + "epoch": 2.68, + "learning_rate": 1.2175071352827241e-05, + "loss": 0.8819, + "step": 36320 + }, + { + "epoch": 2.68, + "learning_rate": 1.2174682028969875e-05, + "loss": 0.8962, + "step": 36321 + }, + { + "epoch": 2.68, + "learning_rate": 1.2174292701652623e-05, + "loss": 0.9838, + "step": 36322 + }, + { + "epoch": 2.68, + "learning_rate": 1.2173903370876102e-05, + "loss": 1.037, + "step": 36323 + }, + { + "epoch": 2.68, + "learning_rate": 1.2173514036640932e-05, + "loss": 1.0501, + "step": 36324 + }, + { + "epoch": 2.68, + "learning_rate": 1.2173124698947732e-05, + "loss": 1.0252, + "step": 36325 + }, + { + "epoch": 2.68, + "learning_rate": 1.217273535779712e-05, + "loss": 0.9019, + "step": 36326 + }, + { + "epoch": 2.68, + "learning_rate": 1.2172346013189722e-05, + "loss": 1.0462, + "step": 36327 + }, + { + "epoch": 2.68, + "learning_rate": 1.2171956665126148e-05, + "loss": 0.9175, + "step": 36328 + }, + { + "epoch": 2.68, + "learning_rate": 1.2171567313607025e-05, + "loss": 0.9713, + "step": 36329 + }, + { + "epoch": 2.68, + "learning_rate": 1.2171177958632965e-05, + "loss": 1.05, + "step": 36330 + }, + { + "epoch": 2.68, + "learning_rate": 1.2170788600204598e-05, + "loss": 1.0721, + "step": 36331 + }, + { + "epoch": 2.68, + "learning_rate": 1.2170399238322532e-05, + "loss": 0.9747, + "step": 36332 + }, + { + "epoch": 2.68, + "learning_rate": 1.2170009872987394e-05, + "loss": 0.9983, + "step": 36333 + }, + { + "epoch": 2.68, + "learning_rate": 1.2169620504199802e-05, + "loss": 0.9846, + "step": 36334 + }, + { + "epoch": 2.68, + "learning_rate": 1.2169231131960374e-05, + "loss": 0.9206, + "step": 36335 + }, + { + "epoch": 2.68, + "learning_rate": 1.2168841756269731e-05, + "loss": 0.9501, + "step": 36336 + }, + { + "epoch": 2.68, + "learning_rate": 1.2168452377128493e-05, + "loss": 1.1654, + "step": 36337 + }, + { + "epoch": 2.68, + "learning_rate": 1.2168062994537275e-05, + "loss": 1.0755, + "step": 36338 + }, + { + "epoch": 2.69, + "learning_rate": 1.2167673608496702e-05, + "loss": 1.032, + "step": 36339 + }, + { + "epoch": 2.69, + "learning_rate": 1.216728421900739e-05, + "loss": 1.0727, + "step": 36340 + }, + { + "epoch": 2.69, + "learning_rate": 1.2166894826069962e-05, + "loss": 1.025, + "step": 36341 + }, + { + "epoch": 2.69, + "learning_rate": 1.2166505429685032e-05, + "loss": 1.0605, + "step": 36342 + }, + { + "epoch": 2.69, + "learning_rate": 1.2166116029853225e-05, + "loss": 0.9433, + "step": 36343 + }, + { + "epoch": 2.69, + "learning_rate": 1.216572662657516e-05, + "loss": 1.0082, + "step": 36344 + }, + { + "epoch": 2.69, + "learning_rate": 1.2165337219851453e-05, + "loss": 0.9771, + "step": 36345 + }, + { + "epoch": 2.69, + "learning_rate": 1.2164947809682726e-05, + "loss": 1.0481, + "step": 36346 + }, + { + "epoch": 2.69, + "learning_rate": 1.2164558396069596e-05, + "loss": 0.9616, + "step": 36347 + }, + { + "epoch": 2.69, + "learning_rate": 1.2164168979012687e-05, + "loss": 0.9926, + "step": 36348 + }, + { + "epoch": 2.69, + "learning_rate": 1.2163779558512617e-05, + "loss": 1.1103, + "step": 36349 + }, + { + "epoch": 2.69, + "learning_rate": 1.2163390134570002e-05, + "loss": 1.0132, + "step": 36350 + }, + { + "epoch": 2.69, + "learning_rate": 1.2163000707185463e-05, + "loss": 0.9771, + "step": 36351 + }, + { + "epoch": 2.69, + "learning_rate": 1.2162611276359623e-05, + "loss": 1.0308, + "step": 36352 + }, + { + "epoch": 2.69, + "learning_rate": 1.2162221842093101e-05, + "loss": 1.0416, + "step": 36353 + }, + { + "epoch": 2.69, + "learning_rate": 1.2161832404386513e-05, + "loss": 1.0412, + "step": 36354 + }, + { + "epoch": 2.69, + "learning_rate": 1.2161442963240482e-05, + "loss": 0.9794, + "step": 36355 + }, + { + "epoch": 2.69, + "learning_rate": 1.2161053518655623e-05, + "loss": 1.0687, + "step": 36356 + }, + { + "epoch": 2.69, + "learning_rate": 1.2160664070632561e-05, + "loss": 0.9936, + "step": 36357 + }, + { + "epoch": 2.69, + "learning_rate": 1.2160274619171914e-05, + "loss": 0.9425, + "step": 36358 + }, + { + "epoch": 2.69, + "learning_rate": 1.21598851642743e-05, + "loss": 0.9377, + "step": 36359 + }, + { + "epoch": 2.69, + "learning_rate": 1.215949570594034e-05, + "loss": 0.9945, + "step": 36360 + }, + { + "epoch": 2.69, + "learning_rate": 1.2159106244170654e-05, + "loss": 1.0081, + "step": 36361 + }, + { + "epoch": 2.69, + "learning_rate": 1.2158716778965857e-05, + "loss": 0.9921, + "step": 36362 + }, + { + "epoch": 2.69, + "learning_rate": 1.215832731032658e-05, + "loss": 1.0412, + "step": 36363 + }, + { + "epoch": 2.69, + "learning_rate": 1.215793783825343e-05, + "loss": 0.996, + "step": 36364 + }, + { + "epoch": 2.69, + "learning_rate": 1.2157548362747033e-05, + "loss": 0.9741, + "step": 36365 + }, + { + "epoch": 2.69, + "learning_rate": 1.2157158883808005e-05, + "loss": 0.8763, + "step": 36366 + }, + { + "epoch": 2.69, + "learning_rate": 1.215676940143697e-05, + "loss": 1.0751, + "step": 36367 + }, + { + "epoch": 2.69, + "learning_rate": 1.2156379915634546e-05, + "loss": 1.0233, + "step": 36368 + }, + { + "epoch": 2.69, + "learning_rate": 1.2155990426401352e-05, + "loss": 0.9148, + "step": 36369 + }, + { + "epoch": 2.69, + "learning_rate": 1.2155600933738007e-05, + "loss": 1.0627, + "step": 36370 + }, + { + "epoch": 2.69, + "learning_rate": 1.2155211437645134e-05, + "loss": 1.119, + "step": 36371 + }, + { + "epoch": 2.69, + "learning_rate": 1.2154821938123348e-05, + "loss": 1.0176, + "step": 36372 + }, + { + "epoch": 2.69, + "learning_rate": 1.2154432435173274e-05, + "loss": 1.0608, + "step": 36373 + }, + { + "epoch": 2.69, + "learning_rate": 1.2154042928795526e-05, + "loss": 0.9371, + "step": 36374 + }, + { + "epoch": 2.69, + "learning_rate": 1.2153653418990727e-05, + "loss": 1.0776, + "step": 36375 + }, + { + "epoch": 2.69, + "learning_rate": 1.2153263905759498e-05, + "loss": 1.0504, + "step": 36376 + }, + { + "epoch": 2.69, + "learning_rate": 1.2152874389102457e-05, + "loss": 0.986, + "step": 36377 + }, + { + "epoch": 2.69, + "learning_rate": 1.2152484869020222e-05, + "loss": 1.0291, + "step": 36378 + }, + { + "epoch": 2.69, + "learning_rate": 1.2152095345513414e-05, + "loss": 1.0359, + "step": 36379 + }, + { + "epoch": 2.69, + "learning_rate": 1.2151705818582655e-05, + "loss": 1.0173, + "step": 36380 + }, + { + "epoch": 2.69, + "learning_rate": 1.2151316288228563e-05, + "loss": 1.0683, + "step": 36381 + }, + { + "epoch": 2.69, + "learning_rate": 1.2150926754451757e-05, + "loss": 1.0223, + "step": 36382 + }, + { + "epoch": 2.69, + "learning_rate": 1.2150537217252856e-05, + "loss": 1.0407, + "step": 36383 + }, + { + "epoch": 2.69, + "learning_rate": 1.2150147676632483e-05, + "loss": 1.0215, + "step": 36384 + }, + { + "epoch": 2.69, + "learning_rate": 1.2149758132591251e-05, + "loss": 1.0094, + "step": 36385 + }, + { + "epoch": 2.69, + "learning_rate": 1.214936858512979e-05, + "loss": 0.9015, + "step": 36386 + }, + { + "epoch": 2.69, + "learning_rate": 1.2148979034248711e-05, + "loss": 1.0103, + "step": 36387 + }, + { + "epoch": 2.69, + "learning_rate": 1.214858947994864e-05, + "loss": 1.1723, + "step": 36388 + }, + { + "epoch": 2.69, + "learning_rate": 1.214819992223019e-05, + "loss": 1.0873, + "step": 36389 + }, + { + "epoch": 2.69, + "learning_rate": 1.214781036109399e-05, + "loss": 1.1477, + "step": 36390 + }, + { + "epoch": 2.69, + "learning_rate": 1.2147420796540652e-05, + "loss": 1.0257, + "step": 36391 + }, + { + "epoch": 2.69, + "learning_rate": 1.2147031228570798e-05, + "loss": 1.0177, + "step": 36392 + }, + { + "epoch": 2.69, + "learning_rate": 1.2146641657185048e-05, + "loss": 0.9708, + "step": 36393 + }, + { + "epoch": 2.69, + "learning_rate": 1.214625208238402e-05, + "loss": 1.0094, + "step": 36394 + }, + { + "epoch": 2.69, + "learning_rate": 1.2145862504168338e-05, + "loss": 0.9706, + "step": 36395 + }, + { + "epoch": 2.69, + "learning_rate": 1.2145472922538619e-05, + "loss": 0.9659, + "step": 36396 + }, + { + "epoch": 2.69, + "learning_rate": 1.2145083337495482e-05, + "loss": 1.0831, + "step": 36397 + }, + { + "epoch": 2.69, + "learning_rate": 1.214469374903955e-05, + "loss": 1.0119, + "step": 36398 + }, + { + "epoch": 2.69, + "learning_rate": 1.2144304157171438e-05, + "loss": 1.0195, + "step": 36399 + }, + { + "epoch": 2.69, + "learning_rate": 1.2143914561891771e-05, + "loss": 0.9843, + "step": 36400 + }, + { + "epoch": 2.69, + "learning_rate": 1.214352496320117e-05, + "loss": 1.0701, + "step": 36401 + }, + { + "epoch": 2.69, + "learning_rate": 1.2143135361100247e-05, + "loss": 1.0515, + "step": 36402 + }, + { + "epoch": 2.69, + "learning_rate": 1.2142745755589626e-05, + "loss": 1.0426, + "step": 36403 + }, + { + "epoch": 2.69, + "learning_rate": 1.2142356146669928e-05, + "loss": 1.0278, + "step": 36404 + }, + { + "epoch": 2.69, + "learning_rate": 1.2141966534341773e-05, + "loss": 0.9875, + "step": 36405 + }, + { + "epoch": 2.69, + "learning_rate": 1.214157691860578e-05, + "loss": 1.0942, + "step": 36406 + }, + { + "epoch": 2.69, + "learning_rate": 1.2141187299462567e-05, + "loss": 0.9826, + "step": 36407 + }, + { + "epoch": 2.69, + "learning_rate": 1.2140797676912759e-05, + "loss": 1.1061, + "step": 36408 + }, + { + "epoch": 2.69, + "learning_rate": 1.2140408050956969e-05, + "loss": 0.9318, + "step": 36409 + }, + { + "epoch": 2.69, + "learning_rate": 1.2140018421595823e-05, + "loss": 1.0038, + "step": 36410 + }, + { + "epoch": 2.69, + "learning_rate": 1.2139628788829935e-05, + "loss": 1.0943, + "step": 36411 + }, + { + "epoch": 2.69, + "learning_rate": 1.2139239152659932e-05, + "loss": 1.0415, + "step": 36412 + }, + { + "epoch": 2.69, + "learning_rate": 1.2138849513086429e-05, + "loss": 0.9938, + "step": 36413 + }, + { + "epoch": 2.69, + "learning_rate": 1.2138459870110047e-05, + "loss": 0.9713, + "step": 36414 + }, + { + "epoch": 2.69, + "learning_rate": 1.2138070223731401e-05, + "loss": 1.0238, + "step": 36415 + }, + { + "epoch": 2.69, + "learning_rate": 1.2137680573951124e-05, + "loss": 0.9183, + "step": 36416 + }, + { + "epoch": 2.69, + "learning_rate": 1.2137290920769822e-05, + "loss": 1.0092, + "step": 36417 + }, + { + "epoch": 2.69, + "learning_rate": 1.2136901264188123e-05, + "loss": 1.008, + "step": 36418 + }, + { + "epoch": 2.69, + "learning_rate": 1.2136511604206645e-05, + "loss": 1.0193, + "step": 36419 + }, + { + "epoch": 2.69, + "learning_rate": 1.2136121940826006e-05, + "loss": 0.9843, + "step": 36420 + }, + { + "epoch": 2.69, + "learning_rate": 1.213573227404683e-05, + "loss": 1.0259, + "step": 36421 + }, + { + "epoch": 2.69, + "learning_rate": 1.2135342603869733e-05, + "loss": 0.9628, + "step": 36422 + }, + { + "epoch": 2.69, + "learning_rate": 1.2134952930295336e-05, + "loss": 1.0867, + "step": 36423 + }, + { + "epoch": 2.69, + "learning_rate": 1.213456325332426e-05, + "loss": 0.9757, + "step": 36424 + }, + { + "epoch": 2.69, + "learning_rate": 1.2134173572957124e-05, + "loss": 1.0058, + "step": 36425 + }, + { + "epoch": 2.69, + "learning_rate": 1.2133783889194548e-05, + "loss": 1.03, + "step": 36426 + }, + { + "epoch": 2.69, + "learning_rate": 1.2133394202037151e-05, + "loss": 0.9849, + "step": 36427 + }, + { + "epoch": 2.69, + "learning_rate": 1.2133004511485558e-05, + "loss": 0.958, + "step": 36428 + }, + { + "epoch": 2.69, + "learning_rate": 1.2132614817540385e-05, + "loss": 1.1215, + "step": 36429 + }, + { + "epoch": 2.69, + "learning_rate": 1.213222512020225e-05, + "loss": 0.9937, + "step": 36430 + }, + { + "epoch": 2.69, + "learning_rate": 1.2131835419471778e-05, + "loss": 1.0311, + "step": 36431 + }, + { + "epoch": 2.69, + "learning_rate": 1.2131445715349584e-05, + "loss": 0.9811, + "step": 36432 + }, + { + "epoch": 2.69, + "learning_rate": 1.213105600783629e-05, + "loss": 1.0789, + "step": 36433 + }, + { + "epoch": 2.69, + "learning_rate": 1.2130666296932517e-05, + "loss": 1.0045, + "step": 36434 + }, + { + "epoch": 2.69, + "learning_rate": 1.2130276582638886e-05, + "loss": 0.9738, + "step": 36435 + }, + { + "epoch": 2.69, + "learning_rate": 1.2129886864956013e-05, + "loss": 1.0488, + "step": 36436 + }, + { + "epoch": 2.69, + "learning_rate": 1.2129497143884523e-05, + "loss": 0.9792, + "step": 36437 + }, + { + "epoch": 2.69, + "learning_rate": 1.2129107419425033e-05, + "loss": 1.1043, + "step": 36438 + }, + { + "epoch": 2.69, + "learning_rate": 1.2128717691578164e-05, + "loss": 1.007, + "step": 36439 + }, + { + "epoch": 2.69, + "learning_rate": 1.2128327960344532e-05, + "loss": 1.0662, + "step": 36440 + }, + { + "epoch": 2.69, + "learning_rate": 1.2127938225724765e-05, + "loss": 0.913, + "step": 36441 + }, + { + "epoch": 2.69, + "learning_rate": 1.2127548487719475e-05, + "loss": 1.0053, + "step": 36442 + }, + { + "epoch": 2.69, + "learning_rate": 1.2127158746329288e-05, + "loss": 1.0182, + "step": 36443 + }, + { + "epoch": 2.69, + "learning_rate": 1.2126769001554821e-05, + "loss": 1.0706, + "step": 36444 + }, + { + "epoch": 2.69, + "learning_rate": 1.2126379253396697e-05, + "loss": 0.9727, + "step": 36445 + }, + { + "epoch": 2.69, + "learning_rate": 1.212598950185553e-05, + "loss": 0.9282, + "step": 36446 + }, + { + "epoch": 2.69, + "learning_rate": 1.212559974693195e-05, + "loss": 1.102, + "step": 36447 + }, + { + "epoch": 2.69, + "learning_rate": 1.2125209988626568e-05, + "loss": 1.0743, + "step": 36448 + }, + { + "epoch": 2.69, + "learning_rate": 1.2124820226940006e-05, + "loss": 1.0554, + "step": 36449 + }, + { + "epoch": 2.69, + "learning_rate": 1.2124430461872888e-05, + "loss": 0.9248, + "step": 36450 + }, + { + "epoch": 2.69, + "learning_rate": 1.2124040693425828e-05, + "loss": 1.0155, + "step": 36451 + }, + { + "epoch": 2.69, + "learning_rate": 1.2123650921599452e-05, + "loss": 1.0627, + "step": 36452 + }, + { + "epoch": 2.69, + "learning_rate": 1.2123261146394378e-05, + "loss": 1.0395, + "step": 36453 + }, + { + "epoch": 2.69, + "learning_rate": 1.2122871367811224e-05, + "loss": 0.9996, + "step": 36454 + }, + { + "epoch": 2.69, + "learning_rate": 1.2122481585850613e-05, + "loss": 1.0042, + "step": 36455 + }, + { + "epoch": 2.69, + "learning_rate": 1.2122091800513165e-05, + "loss": 1.0261, + "step": 36456 + }, + { + "epoch": 2.69, + "learning_rate": 1.2121702011799503e-05, + "loss": 1.0209, + "step": 36457 + }, + { + "epoch": 2.69, + "learning_rate": 1.2121312219710238e-05, + "loss": 1.0677, + "step": 36458 + }, + { + "epoch": 2.69, + "learning_rate": 1.2120922424245997e-05, + "loss": 1.1111, + "step": 36459 + }, + { + "epoch": 2.69, + "learning_rate": 1.2120532625407398e-05, + "loss": 1.0045, + "step": 36460 + }, + { + "epoch": 2.69, + "learning_rate": 1.2120142823195064e-05, + "loss": 1.0412, + "step": 36461 + }, + { + "epoch": 2.69, + "learning_rate": 1.211975301760961e-05, + "loss": 1.0158, + "step": 36462 + }, + { + "epoch": 2.69, + "learning_rate": 1.2119363208651662e-05, + "loss": 0.8596, + "step": 36463 + }, + { + "epoch": 2.69, + "learning_rate": 1.2118973396321835e-05, + "loss": 0.9548, + "step": 36464 + }, + { + "epoch": 2.69, + "learning_rate": 1.2118583580620754e-05, + "loss": 1.0317, + "step": 36465 + }, + { + "epoch": 2.69, + "learning_rate": 1.2118193761549038e-05, + "loss": 0.9231, + "step": 36466 + }, + { + "epoch": 2.69, + "learning_rate": 1.2117803939107305e-05, + "loss": 1.0224, + "step": 36467 + }, + { + "epoch": 2.69, + "learning_rate": 1.2117414113296173e-05, + "loss": 1.072, + "step": 36468 + }, + { + "epoch": 2.69, + "learning_rate": 1.211702428411627e-05, + "loss": 1.0796, + "step": 36469 + }, + { + "epoch": 2.69, + "learning_rate": 1.2116634451568207e-05, + "loss": 1.0183, + "step": 36470 + }, + { + "epoch": 2.69, + "learning_rate": 1.2116244615652613e-05, + "loss": 1.0661, + "step": 36471 + }, + { + "epoch": 2.69, + "learning_rate": 1.2115854776370099e-05, + "loss": 1.0451, + "step": 36472 + }, + { + "epoch": 2.69, + "learning_rate": 1.2115464933721297e-05, + "loss": 1.0303, + "step": 36473 + }, + { + "epoch": 2.7, + "learning_rate": 1.2115075087706816e-05, + "loss": 0.9828, + "step": 36474 + }, + { + "epoch": 2.7, + "learning_rate": 1.2114685238327282e-05, + "loss": 1.024, + "step": 36475 + }, + { + "epoch": 2.7, + "learning_rate": 1.2114295385583314e-05, + "loss": 1.0025, + "step": 36476 + }, + { + "epoch": 2.7, + "learning_rate": 1.2113905529475531e-05, + "loss": 0.8817, + "step": 36477 + }, + { + "epoch": 2.7, + "learning_rate": 1.2113515670004557e-05, + "loss": 0.9989, + "step": 36478 + }, + { + "epoch": 2.7, + "learning_rate": 1.2113125807171009e-05, + "loss": 0.9546, + "step": 36479 + }, + { + "epoch": 2.7, + "learning_rate": 1.2112735940975509e-05, + "loss": 1.1122, + "step": 36480 + }, + { + "epoch": 2.7, + "learning_rate": 1.2112346071418672e-05, + "loss": 1.044, + "step": 36481 + }, + { + "epoch": 2.7, + "learning_rate": 1.2111956198501128e-05, + "loss": 0.9035, + "step": 36482 + }, + { + "epoch": 2.7, + "learning_rate": 1.211156632222349e-05, + "loss": 1.0173, + "step": 36483 + }, + { + "epoch": 2.7, + "learning_rate": 1.2111176442586378e-05, + "loss": 1.0084, + "step": 36484 + }, + { + "epoch": 2.7, + "learning_rate": 1.2110786559590417e-05, + "loss": 1.0024, + "step": 36485 + }, + { + "epoch": 2.7, + "learning_rate": 1.2110396673236226e-05, + "loss": 0.9531, + "step": 36486 + }, + { + "epoch": 2.7, + "learning_rate": 1.2110006783524422e-05, + "loss": 1.0725, + "step": 36487 + }, + { + "epoch": 2.7, + "learning_rate": 1.2109616890455626e-05, + "loss": 1.0553, + "step": 36488 + }, + { + "epoch": 2.7, + "learning_rate": 1.210922699403046e-05, + "loss": 0.9781, + "step": 36489 + }, + { + "epoch": 2.7, + "learning_rate": 1.2108837094249549e-05, + "loss": 0.9852, + "step": 36490 + }, + { + "epoch": 2.7, + "learning_rate": 1.2108447191113502e-05, + "loss": 1.1008, + "step": 36491 + }, + { + "epoch": 2.7, + "learning_rate": 1.210805728462295e-05, + "loss": 1.0102, + "step": 36492 + }, + { + "epoch": 2.7, + "learning_rate": 1.2107667374778505e-05, + "loss": 1.1203, + "step": 36493 + }, + { + "epoch": 2.7, + "learning_rate": 1.2107277461580798e-05, + "loss": 0.9557, + "step": 36494 + }, + { + "epoch": 2.7, + "learning_rate": 1.2106887545030437e-05, + "loss": 1.0234, + "step": 36495 + }, + { + "epoch": 2.7, + "learning_rate": 1.210649762512805e-05, + "loss": 1.0529, + "step": 36496 + }, + { + "epoch": 2.7, + "learning_rate": 1.2106107701874255e-05, + "loss": 0.9819, + "step": 36497 + }, + { + "epoch": 2.7, + "learning_rate": 1.2105717775269671e-05, + "loss": 0.9902, + "step": 36498 + }, + { + "epoch": 2.7, + "learning_rate": 1.2105327845314922e-05, + "loss": 0.859, + "step": 36499 + }, + { + "epoch": 2.7, + "learning_rate": 1.2104937912010626e-05, + "loss": 1.0279, + "step": 36500 + }, + { + "epoch": 2.7, + "learning_rate": 1.2104547975357403e-05, + "loss": 1.0184, + "step": 36501 + }, + { + "epoch": 2.7, + "learning_rate": 1.2104158035355877e-05, + "loss": 0.9623, + "step": 36502 + }, + { + "epoch": 2.7, + "learning_rate": 1.2103768092006667e-05, + "loss": 0.8453, + "step": 36503 + }, + { + "epoch": 2.7, + "learning_rate": 1.2103378145310389e-05, + "loss": 0.9949, + "step": 36504 + }, + { + "epoch": 2.7, + "learning_rate": 1.2102988195267667e-05, + "loss": 0.9512, + "step": 36505 + }, + { + "epoch": 2.7, + "learning_rate": 1.210259824187912e-05, + "loss": 0.8875, + "step": 36506 + }, + { + "epoch": 2.7, + "learning_rate": 1.2102208285145369e-05, + "loss": 1.0513, + "step": 36507 + }, + { + "epoch": 2.7, + "learning_rate": 1.2101818325067036e-05, + "loss": 1.0899, + "step": 36508 + }, + { + "epoch": 2.7, + "learning_rate": 1.2101428361644737e-05, + "loss": 1.0921, + "step": 36509 + }, + { + "epoch": 2.7, + "learning_rate": 1.2101038394879102e-05, + "loss": 0.9076, + "step": 36510 + }, + { + "epoch": 2.7, + "learning_rate": 1.210064842477074e-05, + "loss": 1.0072, + "step": 36511 + }, + { + "epoch": 2.7, + "learning_rate": 1.2100258451320278e-05, + "loss": 0.9576, + "step": 36512 + }, + { + "epoch": 2.7, + "learning_rate": 1.2099868474528335e-05, + "loss": 1.0064, + "step": 36513 + }, + { + "epoch": 2.7, + "learning_rate": 1.2099478494395531e-05, + "loss": 0.9503, + "step": 36514 + }, + { + "epoch": 2.7, + "learning_rate": 1.2099088510922486e-05, + "loss": 0.9864, + "step": 36515 + }, + { + "epoch": 2.7, + "learning_rate": 1.2098698524109821e-05, + "loss": 1.1049, + "step": 36516 + }, + { + "epoch": 2.7, + "learning_rate": 1.2098308533958158e-05, + "loss": 1.0061, + "step": 36517 + }, + { + "epoch": 2.7, + "learning_rate": 1.2097918540468116e-05, + "loss": 0.902, + "step": 36518 + }, + { + "epoch": 2.7, + "learning_rate": 1.2097528543640315e-05, + "loss": 0.9349, + "step": 36519 + }, + { + "epoch": 2.7, + "learning_rate": 1.2097138543475377e-05, + "loss": 1.0079, + "step": 36520 + }, + { + "epoch": 2.7, + "learning_rate": 1.2096748539973919e-05, + "loss": 0.9978, + "step": 36521 + }, + { + "epoch": 2.7, + "learning_rate": 1.2096358533136567e-05, + "loss": 0.9191, + "step": 36522 + }, + { + "epoch": 2.7, + "learning_rate": 1.2095968522963939e-05, + "loss": 1.0887, + "step": 36523 + }, + { + "epoch": 2.7, + "learning_rate": 1.2095578509456651e-05, + "loss": 1.0479, + "step": 36524 + }, + { + "epoch": 2.7, + "learning_rate": 1.2095188492615332e-05, + "loss": 0.9219, + "step": 36525 + }, + { + "epoch": 2.7, + "learning_rate": 1.2094798472440595e-05, + "loss": 1.0019, + "step": 36526 + }, + { + "epoch": 2.7, + "learning_rate": 1.2094408448933065e-05, + "loss": 0.9125, + "step": 36527 + }, + { + "epoch": 2.7, + "learning_rate": 1.2094018422093362e-05, + "loss": 1.0625, + "step": 36528 + }, + { + "epoch": 2.7, + "learning_rate": 1.2093628391922105e-05, + "loss": 0.9943, + "step": 36529 + }, + { + "epoch": 2.7, + "learning_rate": 1.2093238358419915e-05, + "loss": 0.9905, + "step": 36530 + }, + { + "epoch": 2.7, + "learning_rate": 1.2092848321587414e-05, + "loss": 1.0056, + "step": 36531 + }, + { + "epoch": 2.7, + "learning_rate": 1.209245828142522e-05, + "loss": 0.8929, + "step": 36532 + }, + { + "epoch": 2.7, + "learning_rate": 1.2092068237933953e-05, + "loss": 0.9912, + "step": 36533 + }, + { + "epoch": 2.7, + "learning_rate": 1.2091678191114236e-05, + "loss": 1.0231, + "step": 36534 + }, + { + "epoch": 2.7, + "learning_rate": 1.209128814096669e-05, + "loss": 1.082, + "step": 36535 + }, + { + "epoch": 2.7, + "learning_rate": 1.2090898087491936e-05, + "loss": 0.9421, + "step": 36536 + }, + { + "epoch": 2.7, + "learning_rate": 1.2090508030690591e-05, + "loss": 0.9165, + "step": 36537 + }, + { + "epoch": 2.7, + "learning_rate": 1.2090117970563278e-05, + "loss": 0.9254, + "step": 36538 + }, + { + "epoch": 2.7, + "learning_rate": 1.2089727907110618e-05, + "loss": 1.0805, + "step": 36539 + }, + { + "epoch": 2.7, + "learning_rate": 1.208933784033323e-05, + "loss": 0.9728, + "step": 36540 + }, + { + "epoch": 2.7, + "learning_rate": 1.2088947770231735e-05, + "loss": 0.9296, + "step": 36541 + }, + { + "epoch": 2.7, + "learning_rate": 1.2088557696806755e-05, + "loss": 0.957, + "step": 36542 + }, + { + "epoch": 2.7, + "learning_rate": 1.208816762005891e-05, + "loss": 1.0162, + "step": 36543 + }, + { + "epoch": 2.7, + "learning_rate": 1.2087777539988818e-05, + "loss": 0.9655, + "step": 36544 + }, + { + "epoch": 2.7, + "learning_rate": 1.2087387456597103e-05, + "loss": 0.9818, + "step": 36545 + }, + { + "epoch": 2.7, + "learning_rate": 1.2086997369884383e-05, + "loss": 1.1235, + "step": 36546 + }, + { + "epoch": 2.7, + "learning_rate": 1.2086607279851283e-05, + "loss": 1.079, + "step": 36547 + }, + { + "epoch": 2.7, + "learning_rate": 1.208621718649842e-05, + "loss": 1.0368, + "step": 36548 + }, + { + "epoch": 2.7, + "learning_rate": 1.2085827089826414e-05, + "loss": 1.015, + "step": 36549 + }, + { + "epoch": 2.7, + "learning_rate": 1.2085436989835888e-05, + "loss": 1.1222, + "step": 36550 + }, + { + "epoch": 2.7, + "learning_rate": 1.2085046886527462e-05, + "loss": 1.0033, + "step": 36551 + }, + { + "epoch": 2.7, + "learning_rate": 1.2084656779901754e-05, + "loss": 0.9827, + "step": 36552 + }, + { + "epoch": 2.7, + "learning_rate": 1.208426666995939e-05, + "loss": 1.1302, + "step": 36553 + }, + { + "epoch": 2.7, + "learning_rate": 1.2083876556700984e-05, + "loss": 0.8984, + "step": 36554 + }, + { + "epoch": 2.7, + "learning_rate": 1.2083486440127163e-05, + "loss": 0.941, + "step": 36555 + }, + { + "epoch": 2.7, + "learning_rate": 1.2083096320238543e-05, + "loss": 1.0402, + "step": 36556 + }, + { + "epoch": 2.7, + "learning_rate": 1.2082706197035747e-05, + "loss": 0.9038, + "step": 36557 + }, + { + "epoch": 2.7, + "learning_rate": 1.2082316070519396e-05, + "loss": 1.1207, + "step": 36558 + }, + { + "epoch": 2.7, + "learning_rate": 1.2081925940690109e-05, + "loss": 1.0328, + "step": 36559 + }, + { + "epoch": 2.7, + "learning_rate": 1.208153580754851e-05, + "loss": 0.9872, + "step": 36560 + }, + { + "epoch": 2.7, + "learning_rate": 1.2081145671095215e-05, + "loss": 0.9429, + "step": 36561 + }, + { + "epoch": 2.7, + "learning_rate": 1.2080755531330847e-05, + "loss": 0.9494, + "step": 36562 + }, + { + "epoch": 2.7, + "learning_rate": 1.2080365388256028e-05, + "loss": 1.0609, + "step": 36563 + }, + { + "epoch": 2.7, + "learning_rate": 1.2079975241871376e-05, + "loss": 1.0407, + "step": 36564 + }, + { + "epoch": 2.7, + "learning_rate": 1.2079585092177514e-05, + "loss": 0.9633, + "step": 36565 + }, + { + "epoch": 2.7, + "learning_rate": 1.2079194939175061e-05, + "loss": 0.9847, + "step": 36566 + }, + { + "epoch": 2.7, + "learning_rate": 1.207880478286464e-05, + "loss": 1.0778, + "step": 36567 + }, + { + "epoch": 2.7, + "learning_rate": 1.2078414623246867e-05, + "loss": 0.9015, + "step": 36568 + }, + { + "epoch": 2.7, + "learning_rate": 1.207802446032237e-05, + "loss": 0.9784, + "step": 36569 + }, + { + "epoch": 2.7, + "learning_rate": 1.2077634294091766e-05, + "loss": 1.0835, + "step": 36570 + }, + { + "epoch": 2.7, + "learning_rate": 1.2077244124555675e-05, + "loss": 0.9689, + "step": 36571 + }, + { + "epoch": 2.7, + "learning_rate": 1.2076853951714716e-05, + "loss": 0.956, + "step": 36572 + }, + { + "epoch": 2.7, + "learning_rate": 1.2076463775569515e-05, + "loss": 1.1031, + "step": 36573 + }, + { + "epoch": 2.7, + "learning_rate": 1.2076073596120688e-05, + "loss": 0.9634, + "step": 36574 + }, + { + "epoch": 2.7, + "learning_rate": 1.2075683413368857e-05, + "loss": 1.0292, + "step": 36575 + }, + { + "epoch": 2.7, + "learning_rate": 1.2075293227314644e-05, + "loss": 0.9492, + "step": 36576 + }, + { + "epoch": 2.7, + "learning_rate": 1.207490303795867e-05, + "loss": 1.0229, + "step": 36577 + }, + { + "epoch": 2.7, + "learning_rate": 1.2074512845301557e-05, + "loss": 1.0766, + "step": 36578 + }, + { + "epoch": 2.7, + "learning_rate": 1.2074122649343918e-05, + "loss": 1.0559, + "step": 36579 + }, + { + "epoch": 2.7, + "learning_rate": 1.2073732450086385e-05, + "loss": 0.9155, + "step": 36580 + }, + { + "epoch": 2.7, + "learning_rate": 1.207334224752957e-05, + "loss": 1.0853, + "step": 36581 + }, + { + "epoch": 2.7, + "learning_rate": 1.2072952041674097e-05, + "loss": 1.1005, + "step": 36582 + }, + { + "epoch": 2.7, + "learning_rate": 1.2072561832520588e-05, + "loss": 1.0221, + "step": 36583 + }, + { + "epoch": 2.7, + "learning_rate": 1.2072171620069664e-05, + "loss": 1.0375, + "step": 36584 + }, + { + "epoch": 2.7, + "learning_rate": 1.2071781404321943e-05, + "loss": 0.991, + "step": 36585 + }, + { + "epoch": 2.7, + "learning_rate": 1.2071391185278047e-05, + "loss": 1.129, + "step": 36586 + }, + { + "epoch": 2.7, + "learning_rate": 1.2071000962938596e-05, + "loss": 1.0475, + "step": 36587 + }, + { + "epoch": 2.7, + "learning_rate": 1.2070610737304218e-05, + "loss": 1.0548, + "step": 36588 + }, + { + "epoch": 2.7, + "learning_rate": 1.2070220508375522e-05, + "loss": 0.9806, + "step": 36589 + }, + { + "epoch": 2.7, + "learning_rate": 1.206983027615314e-05, + "loss": 0.9589, + "step": 36590 + }, + { + "epoch": 2.7, + "learning_rate": 1.2069440040637684e-05, + "loss": 0.9579, + "step": 36591 + }, + { + "epoch": 2.7, + "learning_rate": 1.206904980182978e-05, + "loss": 1.0905, + "step": 36592 + }, + { + "epoch": 2.7, + "learning_rate": 1.2068659559730046e-05, + "loss": 1.0262, + "step": 36593 + }, + { + "epoch": 2.7, + "learning_rate": 1.2068269314339105e-05, + "loss": 1.0404, + "step": 36594 + }, + { + "epoch": 2.7, + "learning_rate": 1.2067879065657575e-05, + "loss": 1.0211, + "step": 36595 + }, + { + "epoch": 2.7, + "learning_rate": 1.2067488813686081e-05, + "loss": 0.9688, + "step": 36596 + }, + { + "epoch": 2.7, + "learning_rate": 1.2067098558425244e-05, + "loss": 0.9539, + "step": 36597 + }, + { + "epoch": 2.7, + "learning_rate": 1.206670829987568e-05, + "loss": 0.9847, + "step": 36598 + }, + { + "epoch": 2.7, + "learning_rate": 1.2066318038038014e-05, + "loss": 1.1291, + "step": 36599 + }, + { + "epoch": 2.7, + "learning_rate": 1.2065927772912865e-05, + "loss": 0.9884, + "step": 36600 + }, + { + "epoch": 2.7, + "learning_rate": 1.2065537504500853e-05, + "loss": 0.9776, + "step": 36601 + }, + { + "epoch": 2.7, + "learning_rate": 1.2065147232802604e-05, + "loss": 1.0536, + "step": 36602 + }, + { + "epoch": 2.7, + "learning_rate": 1.206475695781873e-05, + "loss": 0.9871, + "step": 36603 + }, + { + "epoch": 2.7, + "learning_rate": 1.206436667954986e-05, + "loss": 1.0155, + "step": 36604 + }, + { + "epoch": 2.7, + "learning_rate": 1.2063976397996614e-05, + "loss": 0.8547, + "step": 36605 + }, + { + "epoch": 2.7, + "learning_rate": 1.2063586113159611e-05, + "loss": 0.9436, + "step": 36606 + }, + { + "epoch": 2.7, + "learning_rate": 1.2063195825039469e-05, + "loss": 1.0621, + "step": 36607 + }, + { + "epoch": 2.7, + "learning_rate": 1.2062805533636814e-05, + "loss": 1.0451, + "step": 36608 + }, + { + "epoch": 2.71, + "learning_rate": 1.2062415238952263e-05, + "loss": 1.0474, + "step": 36609 + }, + { + "epoch": 2.71, + "learning_rate": 1.2062024940986441e-05, + "loss": 0.9956, + "step": 36610 + }, + { + "epoch": 2.71, + "learning_rate": 1.2061634639739966e-05, + "loss": 1.0437, + "step": 36611 + }, + { + "epoch": 2.71, + "learning_rate": 1.206124433521346e-05, + "loss": 1.1689, + "step": 36612 + }, + { + "epoch": 2.71, + "learning_rate": 1.2060854027407542e-05, + "loss": 1.0508, + "step": 36613 + }, + { + "epoch": 2.71, + "learning_rate": 1.2060463716322837e-05, + "loss": 1.0138, + "step": 36614 + }, + { + "epoch": 2.71, + "learning_rate": 1.2060073401959962e-05, + "loss": 0.9444, + "step": 36615 + }, + { + "epoch": 2.71, + "learning_rate": 1.2059683084319542e-05, + "loss": 0.9243, + "step": 36616 + }, + { + "epoch": 2.71, + "learning_rate": 1.2059292763402196e-05, + "loss": 0.893, + "step": 36617 + }, + { + "epoch": 2.71, + "learning_rate": 1.2058902439208541e-05, + "loss": 1.0512, + "step": 36618 + }, + { + "epoch": 2.71, + "learning_rate": 1.2058512111739205e-05, + "loss": 0.9774, + "step": 36619 + }, + { + "epoch": 2.71, + "learning_rate": 1.2058121780994804e-05, + "loss": 0.9251, + "step": 36620 + }, + { + "epoch": 2.71, + "learning_rate": 1.205773144697596e-05, + "loss": 0.8126, + "step": 36621 + }, + { + "epoch": 2.71, + "learning_rate": 1.2057341109683295e-05, + "loss": 0.9531, + "step": 36622 + }, + { + "epoch": 2.71, + "learning_rate": 1.205695076911743e-05, + "loss": 0.8896, + "step": 36623 + }, + { + "epoch": 2.71, + "learning_rate": 1.205656042527899e-05, + "loss": 0.9974, + "step": 36624 + }, + { + "epoch": 2.71, + "learning_rate": 1.2056170078168587e-05, + "loss": 1.0862, + "step": 36625 + }, + { + "epoch": 2.71, + "learning_rate": 1.2055779727786848e-05, + "loss": 1.0521, + "step": 36626 + }, + { + "epoch": 2.71, + "learning_rate": 1.2055389374134392e-05, + "loss": 0.943, + "step": 36627 + }, + { + "epoch": 2.71, + "learning_rate": 1.2054999017211841e-05, + "loss": 0.9702, + "step": 36628 + }, + { + "epoch": 2.71, + "learning_rate": 1.2054608657019816e-05, + "loss": 0.9668, + "step": 36629 + }, + { + "epoch": 2.71, + "learning_rate": 1.205421829355894e-05, + "loss": 1.0882, + "step": 36630 + }, + { + "epoch": 2.71, + "learning_rate": 1.2053827926829831e-05, + "loss": 1.0447, + "step": 36631 + }, + { + "epoch": 2.71, + "learning_rate": 1.2053437556833112e-05, + "loss": 1.0151, + "step": 36632 + }, + { + "epoch": 2.71, + "learning_rate": 1.20530471835694e-05, + "loss": 1.0345, + "step": 36633 + }, + { + "epoch": 2.71, + "learning_rate": 1.2052656807039323e-05, + "loss": 0.9945, + "step": 36634 + }, + { + "epoch": 2.71, + "learning_rate": 1.2052266427243498e-05, + "loss": 1.0013, + "step": 36635 + }, + { + "epoch": 2.71, + "learning_rate": 1.2051876044182543e-05, + "loss": 1.0532, + "step": 36636 + }, + { + "epoch": 2.71, + "learning_rate": 1.2051485657857085e-05, + "loss": 1.1046, + "step": 36637 + }, + { + "epoch": 2.71, + "learning_rate": 1.205109526826774e-05, + "loss": 0.9734, + "step": 36638 + }, + { + "epoch": 2.71, + "learning_rate": 1.2050704875415138e-05, + "loss": 0.9755, + "step": 36639 + }, + { + "epoch": 2.71, + "learning_rate": 1.2050314479299887e-05, + "loss": 1.0025, + "step": 36640 + }, + { + "epoch": 2.71, + "learning_rate": 1.2049924079922619e-05, + "loss": 1.0882, + "step": 36641 + }, + { + "epoch": 2.71, + "learning_rate": 1.204953367728395e-05, + "loss": 1.0299, + "step": 36642 + }, + { + "epoch": 2.71, + "learning_rate": 1.2049143271384502e-05, + "loss": 0.9663, + "step": 36643 + }, + { + "epoch": 2.71, + "learning_rate": 1.20487528622249e-05, + "loss": 0.8974, + "step": 36644 + }, + { + "epoch": 2.71, + "learning_rate": 1.204836244980576e-05, + "loss": 0.9914, + "step": 36645 + }, + { + "epoch": 2.71, + "learning_rate": 1.20479720341277e-05, + "loss": 1.0558, + "step": 36646 + }, + { + "epoch": 2.71, + "learning_rate": 1.204758161519135e-05, + "loss": 1.0624, + "step": 36647 + }, + { + "epoch": 2.71, + "learning_rate": 1.2047191192997325e-05, + "loss": 1.1062, + "step": 36648 + }, + { + "epoch": 2.71, + "learning_rate": 1.204680076754625e-05, + "loss": 0.9845, + "step": 36649 + }, + { + "epoch": 2.71, + "learning_rate": 1.2046410338838742e-05, + "loss": 0.9472, + "step": 36650 + }, + { + "epoch": 2.71, + "learning_rate": 1.2046019906875425e-05, + "loss": 0.9413, + "step": 36651 + }, + { + "epoch": 2.71, + "learning_rate": 1.2045629471656918e-05, + "loss": 0.9361, + "step": 36652 + }, + { + "epoch": 2.71, + "learning_rate": 1.204523903318385e-05, + "loss": 0.9815, + "step": 36653 + }, + { + "epoch": 2.71, + "learning_rate": 1.2044848591456832e-05, + "loss": 1.1151, + "step": 36654 + }, + { + "epoch": 2.71, + "learning_rate": 1.2044458146476489e-05, + "loss": 1.0589, + "step": 36655 + }, + { + "epoch": 2.71, + "learning_rate": 1.2044067698243443e-05, + "loss": 0.8985, + "step": 36656 + }, + { + "epoch": 2.71, + "learning_rate": 1.2043677246758314e-05, + "loss": 1.0781, + "step": 36657 + }, + { + "epoch": 2.71, + "learning_rate": 1.2043286792021722e-05, + "loss": 1.0461, + "step": 36658 + }, + { + "epoch": 2.71, + "learning_rate": 1.2042896334034294e-05, + "loss": 1.0011, + "step": 36659 + }, + { + "epoch": 2.71, + "learning_rate": 1.2042505872796641e-05, + "loss": 1.0178, + "step": 36660 + }, + { + "epoch": 2.71, + "learning_rate": 1.2042115408309397e-05, + "loss": 1.0209, + "step": 36661 + }, + { + "epoch": 2.71, + "learning_rate": 1.2041724940573173e-05, + "loss": 1.0594, + "step": 36662 + }, + { + "epoch": 2.71, + "learning_rate": 1.2041334469588597e-05, + "loss": 1.0245, + "step": 36663 + }, + { + "epoch": 2.71, + "learning_rate": 1.2040943995356283e-05, + "loss": 0.903, + "step": 36664 + }, + { + "epoch": 2.71, + "learning_rate": 1.2040553517876858e-05, + "loss": 0.9979, + "step": 36665 + }, + { + "epoch": 2.71, + "learning_rate": 1.2040163037150942e-05, + "loss": 1.0713, + "step": 36666 + }, + { + "epoch": 2.71, + "learning_rate": 1.2039772553179157e-05, + "loss": 0.9969, + "step": 36667 + }, + { + "epoch": 2.71, + "learning_rate": 1.203938206596212e-05, + "loss": 0.8968, + "step": 36668 + }, + { + "epoch": 2.71, + "learning_rate": 1.2038991575500456e-05, + "loss": 0.968, + "step": 36669 + }, + { + "epoch": 2.71, + "learning_rate": 1.2038601081794785e-05, + "loss": 1.0878, + "step": 36670 + }, + { + "epoch": 2.71, + "learning_rate": 1.2038210584845732e-05, + "loss": 1.0799, + "step": 36671 + }, + { + "epoch": 2.71, + "learning_rate": 1.2037820084653913e-05, + "loss": 0.8862, + "step": 36672 + }, + { + "epoch": 2.71, + "learning_rate": 1.203742958121995e-05, + "loss": 1.0725, + "step": 36673 + }, + { + "epoch": 2.71, + "learning_rate": 1.203703907454447e-05, + "loss": 1.0321, + "step": 36674 + }, + { + "epoch": 2.71, + "learning_rate": 1.2036648564628086e-05, + "loss": 0.9169, + "step": 36675 + }, + { + "epoch": 2.71, + "learning_rate": 1.2036258051471426e-05, + "loss": 1.0192, + "step": 36676 + }, + { + "epoch": 2.71, + "learning_rate": 1.2035867535075108e-05, + "loss": 0.9116, + "step": 36677 + }, + { + "epoch": 2.71, + "learning_rate": 1.2035477015439751e-05, + "loss": 1.0132, + "step": 36678 + }, + { + "epoch": 2.71, + "learning_rate": 1.2035086492565982e-05, + "loss": 1.005, + "step": 36679 + }, + { + "epoch": 2.71, + "learning_rate": 1.2034695966454416e-05, + "loss": 0.932, + "step": 36680 + }, + { + "epoch": 2.71, + "learning_rate": 1.2034305437105682e-05, + "loss": 0.97, + "step": 36681 + }, + { + "epoch": 2.71, + "learning_rate": 1.20339149045204e-05, + "loss": 0.9505, + "step": 36682 + }, + { + "epoch": 2.71, + "learning_rate": 1.2033524368699181e-05, + "loss": 0.9042, + "step": 36683 + }, + { + "epoch": 2.71, + "learning_rate": 1.2033133829642658e-05, + "loss": 1.0071, + "step": 36684 + }, + { + "epoch": 2.71, + "learning_rate": 1.2032743287351445e-05, + "loss": 0.925, + "step": 36685 + }, + { + "epoch": 2.71, + "learning_rate": 1.203235274182617e-05, + "loss": 0.9272, + "step": 36686 + }, + { + "epoch": 2.71, + "learning_rate": 1.2031962193067447e-05, + "loss": 1.0124, + "step": 36687 + }, + { + "epoch": 2.71, + "learning_rate": 1.2031571641075906e-05, + "loss": 0.943, + "step": 36688 + }, + { + "epoch": 2.71, + "learning_rate": 1.203118108585216e-05, + "loss": 0.9208, + "step": 36689 + }, + { + "epoch": 2.71, + "learning_rate": 1.2030790527396838e-05, + "loss": 1.0059, + "step": 36690 + }, + { + "epoch": 2.71, + "learning_rate": 1.2030399965710555e-05, + "loss": 1.018, + "step": 36691 + }, + { + "epoch": 2.71, + "learning_rate": 1.2030009400793934e-05, + "loss": 0.9776, + "step": 36692 + }, + { + "epoch": 2.71, + "learning_rate": 1.2029618832647595e-05, + "loss": 0.9, + "step": 36693 + }, + { + "epoch": 2.71, + "learning_rate": 1.2029228261272166e-05, + "loss": 1.0396, + "step": 36694 + }, + { + "epoch": 2.71, + "learning_rate": 1.2028837686668262e-05, + "loss": 1.0796, + "step": 36695 + }, + { + "epoch": 2.71, + "learning_rate": 1.2028447108836505e-05, + "loss": 0.9951, + "step": 36696 + }, + { + "epoch": 2.71, + "learning_rate": 1.2028056527777519e-05, + "loss": 1.0064, + "step": 36697 + }, + { + "epoch": 2.71, + "learning_rate": 1.2027665943491924e-05, + "loss": 0.874, + "step": 36698 + }, + { + "epoch": 2.71, + "learning_rate": 1.2027275355980342e-05, + "loss": 0.9167, + "step": 36699 + }, + { + "epoch": 2.71, + "learning_rate": 1.2026884765243395e-05, + "loss": 1.0584, + "step": 36700 + }, + { + "epoch": 2.71, + "learning_rate": 1.20264941712817e-05, + "loss": 0.8878, + "step": 36701 + }, + { + "epoch": 2.71, + "learning_rate": 1.2026103574095884e-05, + "loss": 0.8912, + "step": 36702 + }, + { + "epoch": 2.71, + "learning_rate": 1.2025712973686565e-05, + "loss": 0.9083, + "step": 36703 + }, + { + "epoch": 2.71, + "learning_rate": 1.2025322370054366e-05, + "loss": 1.0772, + "step": 36704 + }, + { + "epoch": 2.71, + "learning_rate": 1.2024931763199908e-05, + "loss": 1.0891, + "step": 36705 + }, + { + "epoch": 2.71, + "learning_rate": 1.2024541153123815e-05, + "loss": 1.0307, + "step": 36706 + }, + { + "epoch": 2.71, + "learning_rate": 1.2024150539826702e-05, + "loss": 1.0774, + "step": 36707 + }, + { + "epoch": 2.71, + "learning_rate": 1.2023759923309197e-05, + "loss": 0.9245, + "step": 36708 + }, + { + "epoch": 2.71, + "learning_rate": 1.2023369303571918e-05, + "loss": 1.0673, + "step": 36709 + }, + { + "epoch": 2.71, + "learning_rate": 1.202297868061549e-05, + "loss": 0.9499, + "step": 36710 + }, + { + "epoch": 2.71, + "learning_rate": 1.2022588054440527e-05, + "loss": 1.1298, + "step": 36711 + }, + { + "epoch": 2.71, + "learning_rate": 1.2022197425047658e-05, + "loss": 1.0958, + "step": 36712 + }, + { + "epoch": 2.71, + "learning_rate": 1.20218067924375e-05, + "loss": 0.8674, + "step": 36713 + }, + { + "epoch": 2.71, + "learning_rate": 1.2021416156610681e-05, + "loss": 1.0603, + "step": 36714 + }, + { + "epoch": 2.71, + "learning_rate": 1.2021025517567813e-05, + "loss": 1.0857, + "step": 36715 + }, + { + "epoch": 2.71, + "learning_rate": 1.2020634875309525e-05, + "loss": 0.9952, + "step": 36716 + }, + { + "epoch": 2.71, + "learning_rate": 1.2020244229836433e-05, + "loss": 0.9543, + "step": 36717 + }, + { + "epoch": 2.71, + "learning_rate": 1.2019853581149164e-05, + "loss": 1.1241, + "step": 36718 + }, + { + "epoch": 2.71, + "learning_rate": 1.2019462929248339e-05, + "loss": 0.9683, + "step": 36719 + }, + { + "epoch": 2.71, + "learning_rate": 1.2019072274134571e-05, + "loss": 1.022, + "step": 36720 + }, + { + "epoch": 2.71, + "learning_rate": 1.2018681615808491e-05, + "loss": 0.8887, + "step": 36721 + }, + { + "epoch": 2.71, + "learning_rate": 1.201829095427072e-05, + "loss": 1.0038, + "step": 36722 + }, + { + "epoch": 2.71, + "learning_rate": 1.2017900289521873e-05, + "loss": 0.9644, + "step": 36723 + }, + { + "epoch": 2.71, + "learning_rate": 1.201750962156258e-05, + "loss": 0.9034, + "step": 36724 + }, + { + "epoch": 2.71, + "learning_rate": 1.2017118950393454e-05, + "loss": 1.0701, + "step": 36725 + }, + { + "epoch": 2.71, + "learning_rate": 1.2016728276015121e-05, + "loss": 1.0377, + "step": 36726 + }, + { + "epoch": 2.71, + "learning_rate": 1.2016337598428203e-05, + "loss": 0.9248, + "step": 36727 + }, + { + "epoch": 2.71, + "learning_rate": 1.2015946917633322e-05, + "loss": 1.1656, + "step": 36728 + }, + { + "epoch": 2.71, + "learning_rate": 1.2015556233631098e-05, + "loss": 0.9979, + "step": 36729 + }, + { + "epoch": 2.71, + "learning_rate": 1.201516554642215e-05, + "loss": 0.9157, + "step": 36730 + }, + { + "epoch": 2.71, + "learning_rate": 1.2014774856007105e-05, + "loss": 0.9019, + "step": 36731 + }, + { + "epoch": 2.71, + "learning_rate": 1.2014384162386581e-05, + "loss": 0.9804, + "step": 36732 + }, + { + "epoch": 2.71, + "learning_rate": 1.2013993465561202e-05, + "loss": 0.9143, + "step": 36733 + }, + { + "epoch": 2.71, + "learning_rate": 1.2013602765531586e-05, + "loss": 0.9468, + "step": 36734 + }, + { + "epoch": 2.71, + "learning_rate": 1.2013212062298358e-05, + "loss": 1.0272, + "step": 36735 + }, + { + "epoch": 2.71, + "learning_rate": 1.2012821355862138e-05, + "loss": 0.9477, + "step": 36736 + }, + { + "epoch": 2.71, + "learning_rate": 1.2012430646223552e-05, + "loss": 0.9042, + "step": 36737 + }, + { + "epoch": 2.71, + "learning_rate": 1.2012039933383212e-05, + "loss": 1.0611, + "step": 36738 + }, + { + "epoch": 2.71, + "learning_rate": 1.201164921734175e-05, + "loss": 0.9825, + "step": 36739 + }, + { + "epoch": 2.71, + "learning_rate": 1.2011258498099778e-05, + "loss": 0.9635, + "step": 36740 + }, + { + "epoch": 2.71, + "learning_rate": 1.2010867775657926e-05, + "loss": 1.0364, + "step": 36741 + }, + { + "epoch": 2.71, + "learning_rate": 1.201047705001681e-05, + "loss": 0.957, + "step": 36742 + }, + { + "epoch": 2.71, + "learning_rate": 1.2010086321177054e-05, + "loss": 0.8855, + "step": 36743 + }, + { + "epoch": 2.71, + "learning_rate": 1.2009695589139281e-05, + "loss": 1.032, + "step": 36744 + }, + { + "epoch": 2.72, + "learning_rate": 1.2009304853904112e-05, + "loss": 0.9334, + "step": 36745 + }, + { + "epoch": 2.72, + "learning_rate": 1.2008914115472166e-05, + "loss": 0.9582, + "step": 36746 + }, + { + "epoch": 2.72, + "learning_rate": 1.2008523373844066e-05, + "loss": 1.0212, + "step": 36747 + }, + { + "epoch": 2.72, + "learning_rate": 1.2008132629020435e-05, + "loss": 1.0278, + "step": 36748 + }, + { + "epoch": 2.72, + "learning_rate": 1.2007741881001893e-05, + "loss": 1.0625, + "step": 36749 + }, + { + "epoch": 2.72, + "learning_rate": 1.2007351129789062e-05, + "loss": 0.9204, + "step": 36750 + }, + { + "epoch": 2.72, + "learning_rate": 1.2006960375382564e-05, + "loss": 0.94, + "step": 36751 + }, + { + "epoch": 2.72, + "learning_rate": 1.200656961778302e-05, + "loss": 1.0601, + "step": 36752 + }, + { + "epoch": 2.72, + "learning_rate": 1.2006178856991055e-05, + "loss": 0.9123, + "step": 36753 + }, + { + "epoch": 2.72, + "learning_rate": 1.2005788093007287e-05, + "loss": 1.0135, + "step": 36754 + }, + { + "epoch": 2.72, + "learning_rate": 1.2005397325832337e-05, + "loss": 0.9784, + "step": 36755 + }, + { + "epoch": 2.72, + "learning_rate": 1.2005006555466833e-05, + "loss": 0.967, + "step": 36756 + }, + { + "epoch": 2.72, + "learning_rate": 1.200461578191139e-05, + "loss": 1.0317, + "step": 36757 + }, + { + "epoch": 2.72, + "learning_rate": 1.200422500516663e-05, + "loss": 1.0338, + "step": 36758 + }, + { + "epoch": 2.72, + "learning_rate": 1.2003834225233179e-05, + "loss": 1.064, + "step": 36759 + }, + { + "epoch": 2.72, + "learning_rate": 1.2003443442111656e-05, + "loss": 0.9429, + "step": 36760 + }, + { + "epoch": 2.72, + "learning_rate": 1.2003052655802683e-05, + "loss": 0.9543, + "step": 36761 + }, + { + "epoch": 2.72, + "learning_rate": 1.200266186630688e-05, + "loss": 1.0284, + "step": 36762 + }, + { + "epoch": 2.72, + "learning_rate": 1.2002271073624874e-05, + "loss": 0.9729, + "step": 36763 + }, + { + "epoch": 2.72, + "learning_rate": 1.200188027775728e-05, + "loss": 0.9219, + "step": 36764 + }, + { + "epoch": 2.72, + "learning_rate": 1.200148947870473e-05, + "loss": 0.9885, + "step": 36765 + }, + { + "epoch": 2.72, + "learning_rate": 1.200109867646783e-05, + "loss": 0.9549, + "step": 36766 + }, + { + "epoch": 2.72, + "learning_rate": 1.2000707871047215e-05, + "loss": 0.9318, + "step": 36767 + }, + { + "epoch": 2.72, + "learning_rate": 1.20003170624435e-05, + "loss": 1.0713, + "step": 36768 + }, + { + "epoch": 2.72, + "learning_rate": 1.1999926250657313e-05, + "loss": 1.0109, + "step": 36769 + }, + { + "epoch": 2.72, + "learning_rate": 1.1999535435689269e-05, + "loss": 1.0129, + "step": 36770 + }, + { + "epoch": 2.72, + "learning_rate": 1.1999144617539992e-05, + "loss": 1.0569, + "step": 36771 + }, + { + "epoch": 2.72, + "learning_rate": 1.1998753796210107e-05, + "loss": 0.9815, + "step": 36772 + }, + { + "epoch": 2.72, + "learning_rate": 1.1998362971700232e-05, + "loss": 1.0162, + "step": 36773 + }, + { + "epoch": 2.72, + "learning_rate": 1.199797214401099e-05, + "loss": 0.9137, + "step": 36774 + }, + { + "epoch": 2.72, + "learning_rate": 1.1997581313143005e-05, + "loss": 1.0367, + "step": 36775 + }, + { + "epoch": 2.72, + "learning_rate": 1.1997190479096895e-05, + "loss": 0.8929, + "step": 36776 + }, + { + "epoch": 2.72, + "learning_rate": 1.1996799641873282e-05, + "loss": 0.9371, + "step": 36777 + }, + { + "epoch": 2.72, + "learning_rate": 1.1996408801472793e-05, + "loss": 0.8574, + "step": 36778 + }, + { + "epoch": 2.72, + "learning_rate": 1.1996017957896043e-05, + "loss": 0.922, + "step": 36779 + }, + { + "epoch": 2.72, + "learning_rate": 1.199562711114366e-05, + "loss": 0.9343, + "step": 36780 + }, + { + "epoch": 2.72, + "learning_rate": 1.199523626121626e-05, + "loss": 0.957, + "step": 36781 + }, + { + "epoch": 2.72, + "learning_rate": 1.1994845408114467e-05, + "loss": 0.9319, + "step": 36782 + }, + { + "epoch": 2.72, + "learning_rate": 1.1994454551838908e-05, + "loss": 1.0636, + "step": 36783 + }, + { + "epoch": 2.72, + "learning_rate": 1.1994063692390196e-05, + "loss": 1.0957, + "step": 36784 + }, + { + "epoch": 2.72, + "learning_rate": 1.1993672829768959e-05, + "loss": 1.026, + "step": 36785 + }, + { + "epoch": 2.72, + "learning_rate": 1.1993281963975817e-05, + "loss": 1.0541, + "step": 36786 + }, + { + "epoch": 2.72, + "learning_rate": 1.1992891095011391e-05, + "loss": 0.8819, + "step": 36787 + }, + { + "epoch": 2.72, + "learning_rate": 1.1992500222876304e-05, + "loss": 1.0032, + "step": 36788 + }, + { + "epoch": 2.72, + "learning_rate": 1.1992109347571177e-05, + "loss": 1.1039, + "step": 36789 + }, + { + "epoch": 2.72, + "learning_rate": 1.1991718469096636e-05, + "loss": 0.9722, + "step": 36790 + }, + { + "epoch": 2.72, + "learning_rate": 1.1991327587453294e-05, + "loss": 1.0576, + "step": 36791 + }, + { + "epoch": 2.72, + "learning_rate": 1.1990936702641784e-05, + "loss": 0.9958, + "step": 36792 + }, + { + "epoch": 2.72, + "learning_rate": 1.1990545814662719e-05, + "loss": 1.0866, + "step": 36793 + }, + { + "epoch": 2.72, + "learning_rate": 1.1990154923516726e-05, + "loss": 0.9679, + "step": 36794 + }, + { + "epoch": 2.72, + "learning_rate": 1.1989764029204422e-05, + "loss": 0.9644, + "step": 36795 + }, + { + "epoch": 2.72, + "learning_rate": 1.1989373131726434e-05, + "loss": 1.0711, + "step": 36796 + }, + { + "epoch": 2.72, + "learning_rate": 1.1988982231083382e-05, + "loss": 0.9762, + "step": 36797 + }, + { + "epoch": 2.72, + "learning_rate": 1.1988591327275887e-05, + "loss": 0.9648, + "step": 36798 + }, + { + "epoch": 2.72, + "learning_rate": 1.198820042030457e-05, + "loss": 0.8527, + "step": 36799 + }, + { + "epoch": 2.72, + "learning_rate": 1.1987809510170057e-05, + "loss": 1.0044, + "step": 36800 + }, + { + "epoch": 2.72, + "learning_rate": 1.1987418596872966e-05, + "loss": 1.01, + "step": 36801 + }, + { + "epoch": 2.72, + "learning_rate": 1.1987027680413922e-05, + "loss": 0.97, + "step": 36802 + }, + { + "epoch": 2.72, + "learning_rate": 1.1986636760793546e-05, + "loss": 1.1021, + "step": 36803 + }, + { + "epoch": 2.72, + "learning_rate": 1.1986245838012458e-05, + "loss": 1.1275, + "step": 36804 + }, + { + "epoch": 2.72, + "learning_rate": 1.198585491207128e-05, + "loss": 1.0361, + "step": 36805 + }, + { + "epoch": 2.72, + "learning_rate": 1.1985463982970637e-05, + "loss": 0.9589, + "step": 36806 + }, + { + "epoch": 2.72, + "learning_rate": 1.1985073050711147e-05, + "loss": 0.9354, + "step": 36807 + }, + { + "epoch": 2.72, + "learning_rate": 1.1984682115293437e-05, + "loss": 0.9693, + "step": 36808 + }, + { + "epoch": 2.72, + "learning_rate": 1.1984291176718124e-05, + "loss": 1.0669, + "step": 36809 + }, + { + "epoch": 2.72, + "learning_rate": 1.1983900234985834e-05, + "loss": 1.0236, + "step": 36810 + }, + { + "epoch": 2.72, + "learning_rate": 1.1983509290097186e-05, + "loss": 1.0791, + "step": 36811 + }, + { + "epoch": 2.72, + "learning_rate": 1.1983118342052807e-05, + "loss": 1.0275, + "step": 36812 + }, + { + "epoch": 2.72, + "learning_rate": 1.1982727390853308e-05, + "loss": 1.0741, + "step": 36813 + }, + { + "epoch": 2.72, + "learning_rate": 1.1982336436499323e-05, + "loss": 1.0162, + "step": 36814 + }, + { + "epoch": 2.72, + "learning_rate": 1.1981945478991466e-05, + "loss": 0.9667, + "step": 36815 + }, + { + "epoch": 2.72, + "learning_rate": 1.1981554518330365e-05, + "loss": 1.067, + "step": 36816 + }, + { + "epoch": 2.72, + "learning_rate": 1.1981163554516638e-05, + "loss": 0.9509, + "step": 36817 + }, + { + "epoch": 2.72, + "learning_rate": 1.1980772587550908e-05, + "loss": 0.9433, + "step": 36818 + }, + { + "epoch": 2.72, + "learning_rate": 1.1980381617433795e-05, + "loss": 0.9918, + "step": 36819 + }, + { + "epoch": 2.72, + "learning_rate": 1.1979990644165927e-05, + "loss": 0.9385, + "step": 36820 + }, + { + "epoch": 2.72, + "learning_rate": 1.1979599667747921e-05, + "loss": 0.8712, + "step": 36821 + }, + { + "epoch": 2.72, + "learning_rate": 1.1979208688180399e-05, + "loss": 1.0571, + "step": 36822 + }, + { + "epoch": 2.72, + "learning_rate": 1.1978817705463985e-05, + "loss": 0.8553, + "step": 36823 + }, + { + "epoch": 2.72, + "learning_rate": 1.1978426719599298e-05, + "loss": 0.942, + "step": 36824 + }, + { + "epoch": 2.72, + "learning_rate": 1.1978035730586968e-05, + "loss": 1.0035, + "step": 36825 + }, + { + "epoch": 2.72, + "learning_rate": 1.197764473842761e-05, + "loss": 1.0997, + "step": 36826 + }, + { + "epoch": 2.72, + "learning_rate": 1.1977253743121842e-05, + "loss": 0.9928, + "step": 36827 + }, + { + "epoch": 2.72, + "learning_rate": 1.1976862744670296e-05, + "loss": 0.9365, + "step": 36828 + }, + { + "epoch": 2.72, + "learning_rate": 1.1976471743073587e-05, + "loss": 1.1321, + "step": 36829 + }, + { + "epoch": 2.72, + "learning_rate": 1.1976080738332346e-05, + "loss": 1.0453, + "step": 36830 + }, + { + "epoch": 2.72, + "learning_rate": 1.1975689730447184e-05, + "loss": 1.0354, + "step": 36831 + }, + { + "epoch": 2.72, + "learning_rate": 1.1975298719418726e-05, + "loss": 0.9929, + "step": 36832 + }, + { + "epoch": 2.72, + "learning_rate": 1.1974907705247601e-05, + "loss": 1.0422, + "step": 36833 + }, + { + "epoch": 2.72, + "learning_rate": 1.1974516687934421e-05, + "loss": 0.8521, + "step": 36834 + }, + { + "epoch": 2.72, + "learning_rate": 1.1974125667479817e-05, + "loss": 1.0709, + "step": 36835 + }, + { + "epoch": 2.72, + "learning_rate": 1.1973734643884406e-05, + "loss": 1.1311, + "step": 36836 + }, + { + "epoch": 2.72, + "learning_rate": 1.1973343617148812e-05, + "loss": 1.0797, + "step": 36837 + }, + { + "epoch": 2.72, + "learning_rate": 1.1972952587273654e-05, + "loss": 0.8845, + "step": 36838 + }, + { + "epoch": 2.72, + "learning_rate": 1.1972561554259561e-05, + "loss": 1.042, + "step": 36839 + }, + { + "epoch": 2.72, + "learning_rate": 1.1972170518107148e-05, + "loss": 0.964, + "step": 36840 + }, + { + "epoch": 2.72, + "learning_rate": 1.1971779478817043e-05, + "loss": 1.0332, + "step": 36841 + }, + { + "epoch": 2.72, + "learning_rate": 1.1971388436389859e-05, + "loss": 0.9622, + "step": 36842 + }, + { + "epoch": 2.72, + "learning_rate": 1.197099739082623e-05, + "loss": 0.9616, + "step": 36843 + }, + { + "epoch": 2.72, + "learning_rate": 1.1970606342126767e-05, + "loss": 1.04, + "step": 36844 + }, + { + "epoch": 2.72, + "learning_rate": 1.1970215290292101e-05, + "loss": 1.0033, + "step": 36845 + }, + { + "epoch": 2.72, + "learning_rate": 1.1969824235322852e-05, + "loss": 0.9425, + "step": 36846 + }, + { + "epoch": 2.72, + "learning_rate": 1.1969433177219638e-05, + "loss": 1.06, + "step": 36847 + }, + { + "epoch": 2.72, + "learning_rate": 1.1969042115983083e-05, + "loss": 0.9395, + "step": 36848 + }, + { + "epoch": 2.72, + "learning_rate": 1.1968651051613816e-05, + "loss": 0.996, + "step": 36849 + }, + { + "epoch": 2.72, + "learning_rate": 1.196825998411245e-05, + "loss": 0.9788, + "step": 36850 + }, + { + "epoch": 2.72, + "learning_rate": 1.196786891347961e-05, + "loss": 0.9872, + "step": 36851 + }, + { + "epoch": 2.72, + "learning_rate": 1.1967477839715918e-05, + "loss": 0.8959, + "step": 36852 + }, + { + "epoch": 2.72, + "learning_rate": 1.1967086762821998e-05, + "loss": 1.0013, + "step": 36853 + }, + { + "epoch": 2.72, + "learning_rate": 1.1966695682798468e-05, + "loss": 1.0455, + "step": 36854 + }, + { + "epoch": 2.72, + "learning_rate": 1.1966304599645957e-05, + "loss": 0.9564, + "step": 36855 + }, + { + "epoch": 2.72, + "learning_rate": 1.1965913513365082e-05, + "loss": 1.1385, + "step": 36856 + }, + { + "epoch": 2.72, + "learning_rate": 1.1965522423956468e-05, + "loss": 1.0486, + "step": 36857 + }, + { + "epoch": 2.72, + "learning_rate": 1.1965131331420734e-05, + "loss": 1.0918, + "step": 36858 + }, + { + "epoch": 2.72, + "learning_rate": 1.1964740235758509e-05, + "loss": 1.0069, + "step": 36859 + }, + { + "epoch": 2.72, + "learning_rate": 1.1964349136970405e-05, + "loss": 1.0537, + "step": 36860 + }, + { + "epoch": 2.72, + "learning_rate": 1.1963958035057053e-05, + "loss": 0.9967, + "step": 36861 + }, + { + "epoch": 2.72, + "learning_rate": 1.1963566930019068e-05, + "loss": 0.9831, + "step": 36862 + }, + { + "epoch": 2.72, + "learning_rate": 1.196317582185708e-05, + "loss": 1.0819, + "step": 36863 + }, + { + "epoch": 2.72, + "learning_rate": 1.1962784710571704e-05, + "loss": 0.9895, + "step": 36864 + }, + { + "epoch": 2.72, + "learning_rate": 1.1962393596163569e-05, + "loss": 0.9631, + "step": 36865 + }, + { + "epoch": 2.72, + "learning_rate": 1.196200247863329e-05, + "loss": 1.0144, + "step": 36866 + }, + { + "epoch": 2.72, + "learning_rate": 1.1961611357981498e-05, + "loss": 1.0772, + "step": 36867 + }, + { + "epoch": 2.72, + "learning_rate": 1.1961220234208807e-05, + "loss": 1.0512, + "step": 36868 + }, + { + "epoch": 2.72, + "learning_rate": 1.1960829107315844e-05, + "loss": 1.0125, + "step": 36869 + }, + { + "epoch": 2.72, + "learning_rate": 1.196043797730323e-05, + "loss": 1.0607, + "step": 36870 + }, + { + "epoch": 2.72, + "learning_rate": 1.1960046844171587e-05, + "loss": 0.9506, + "step": 36871 + }, + { + "epoch": 2.72, + "learning_rate": 1.1959655707921537e-05, + "loss": 1.0531, + "step": 36872 + }, + { + "epoch": 2.72, + "learning_rate": 1.1959264568553704e-05, + "loss": 0.9938, + "step": 36873 + }, + { + "epoch": 2.72, + "learning_rate": 1.1958873426068705e-05, + "loss": 0.9923, + "step": 36874 + }, + { + "epoch": 2.72, + "learning_rate": 1.195848228046717e-05, + "loss": 0.9306, + "step": 36875 + }, + { + "epoch": 2.72, + "learning_rate": 1.1958091131749716e-05, + "loss": 1.0125, + "step": 36876 + }, + { + "epoch": 2.72, + "learning_rate": 1.195769997991697e-05, + "loss": 1.1278, + "step": 36877 + }, + { + "epoch": 2.72, + "learning_rate": 1.1957308824969548e-05, + "loss": 0.9887, + "step": 36878 + }, + { + "epoch": 2.72, + "learning_rate": 1.195691766690808e-05, + "loss": 1.0299, + "step": 36879 + }, + { + "epoch": 2.73, + "learning_rate": 1.1956526505733179e-05, + "loss": 1.0516, + "step": 36880 + }, + { + "epoch": 2.73, + "learning_rate": 1.1956135341445474e-05, + "loss": 0.9252, + "step": 36881 + }, + { + "epoch": 2.73, + "learning_rate": 1.1955744174045587e-05, + "loss": 1.0483, + "step": 36882 + }, + { + "epoch": 2.73, + "learning_rate": 1.1955353003534139e-05, + "loss": 1.002, + "step": 36883 + }, + { + "epoch": 2.73, + "learning_rate": 1.1954961829911752e-05, + "loss": 1.0635, + "step": 36884 + }, + { + "epoch": 2.73, + "learning_rate": 1.1954570653179046e-05, + "loss": 0.9797, + "step": 36885 + }, + { + "epoch": 2.73, + "learning_rate": 1.1954179473336648e-05, + "loss": 1.0024, + "step": 36886 + }, + { + "epoch": 2.73, + "learning_rate": 1.195378829038518e-05, + "loss": 1.0564, + "step": 36887 + }, + { + "epoch": 2.73, + "learning_rate": 1.1953397104325263e-05, + "loss": 1.0384, + "step": 36888 + }, + { + "epoch": 2.73, + "learning_rate": 1.1953005915157516e-05, + "loss": 1.1164, + "step": 36889 + }, + { + "epoch": 2.73, + "learning_rate": 1.1952614722882567e-05, + "loss": 0.9501, + "step": 36890 + }, + { + "epoch": 2.73, + "learning_rate": 1.1952223527501035e-05, + "loss": 1.0021, + "step": 36891 + }, + { + "epoch": 2.73, + "learning_rate": 1.1951832329013544e-05, + "loss": 1.0085, + "step": 36892 + }, + { + "epoch": 2.73, + "learning_rate": 1.1951441127420714e-05, + "loss": 0.929, + "step": 36893 + }, + { + "epoch": 2.73, + "learning_rate": 1.195104992272317e-05, + "loss": 1.0515, + "step": 36894 + }, + { + "epoch": 2.73, + "learning_rate": 1.1950658714921533e-05, + "loss": 1.0068, + "step": 36895 + }, + { + "epoch": 2.73, + "learning_rate": 1.1950267504016428e-05, + "loss": 1.0063, + "step": 36896 + }, + { + "epoch": 2.73, + "learning_rate": 1.1949876290008474e-05, + "loss": 1.0385, + "step": 36897 + }, + { + "epoch": 2.73, + "learning_rate": 1.1949485072898295e-05, + "loss": 1.0609, + "step": 36898 + }, + { + "epoch": 2.73, + "learning_rate": 1.194909385268651e-05, + "loss": 0.9824, + "step": 36899 + }, + { + "epoch": 2.73, + "learning_rate": 1.1948702629373749e-05, + "loss": 1.062, + "step": 36900 + }, + { + "epoch": 2.73, + "learning_rate": 1.1948311402960628e-05, + "loss": 0.987, + "step": 36901 + }, + { + "epoch": 2.73, + "learning_rate": 1.1947920173447773e-05, + "loss": 0.9203, + "step": 36902 + }, + { + "epoch": 2.73, + "learning_rate": 1.1947528940835802e-05, + "loss": 1.0393, + "step": 36903 + }, + { + "epoch": 2.73, + "learning_rate": 1.1947137705125342e-05, + "loss": 1.0489, + "step": 36904 + }, + { + "epoch": 2.73, + "learning_rate": 1.1946746466317014e-05, + "loss": 1.0632, + "step": 36905 + }, + { + "epoch": 2.73, + "learning_rate": 1.1946355224411442e-05, + "loss": 1.0558, + "step": 36906 + }, + { + "epoch": 2.73, + "learning_rate": 1.1945963979409243e-05, + "loss": 1.0237, + "step": 36907 + }, + { + "epoch": 2.73, + "learning_rate": 1.1945572731311046e-05, + "loss": 1.0058, + "step": 36908 + }, + { + "epoch": 2.73, + "learning_rate": 1.1945181480117467e-05, + "loss": 0.8855, + "step": 36909 + }, + { + "epoch": 2.73, + "learning_rate": 1.1944790225829137e-05, + "loss": 1.1009, + "step": 36910 + }, + { + "epoch": 2.73, + "learning_rate": 1.1944398968446668e-05, + "loss": 0.9886, + "step": 36911 + }, + { + "epoch": 2.73, + "learning_rate": 1.1944007707970693e-05, + "loss": 1.058, + "step": 36912 + }, + { + "epoch": 2.73, + "learning_rate": 1.1943616444401827e-05, + "loss": 1.04, + "step": 36913 + }, + { + "epoch": 2.73, + "learning_rate": 1.1943225177740695e-05, + "loss": 1.0243, + "step": 36914 + }, + { + "epoch": 2.73, + "learning_rate": 1.194283390798792e-05, + "loss": 1.0062, + "step": 36915 + }, + { + "epoch": 2.73, + "learning_rate": 1.1942442635144128e-05, + "loss": 1.0082, + "step": 36916 + }, + { + "epoch": 2.73, + "learning_rate": 1.1942051359209932e-05, + "loss": 1.0961, + "step": 36917 + }, + { + "epoch": 2.73, + "learning_rate": 1.1941660080185961e-05, + "loss": 1.0097, + "step": 36918 + }, + { + "epoch": 2.73, + "learning_rate": 1.1941268798072838e-05, + "loss": 0.8945, + "step": 36919 + }, + { + "epoch": 2.73, + "learning_rate": 1.1940877512871185e-05, + "loss": 1.1079, + "step": 36920 + }, + { + "epoch": 2.73, + "learning_rate": 1.194048622458162e-05, + "loss": 0.9443, + "step": 36921 + }, + { + "epoch": 2.73, + "learning_rate": 1.1940094933204773e-05, + "loss": 0.9944, + "step": 36922 + }, + { + "epoch": 2.73, + "learning_rate": 1.193970363874126e-05, + "loss": 1.0436, + "step": 36923 + }, + { + "epoch": 2.73, + "learning_rate": 1.1939312341191707e-05, + "loss": 1.0037, + "step": 36924 + }, + { + "epoch": 2.73, + "learning_rate": 1.1938921040556736e-05, + "loss": 1.0728, + "step": 36925 + }, + { + "epoch": 2.73, + "learning_rate": 1.193852973683697e-05, + "loss": 1.0115, + "step": 36926 + }, + { + "epoch": 2.73, + "learning_rate": 1.1938138430033031e-05, + "loss": 0.8495, + "step": 36927 + }, + { + "epoch": 2.73, + "learning_rate": 1.193774712014554e-05, + "loss": 1.031, + "step": 36928 + }, + { + "epoch": 2.73, + "learning_rate": 1.1937355807175125e-05, + "loss": 0.9743, + "step": 36929 + }, + { + "epoch": 2.73, + "learning_rate": 1.1936964491122397e-05, + "loss": 1.0955, + "step": 36930 + }, + { + "epoch": 2.73, + "learning_rate": 1.1936573171987993e-05, + "loss": 0.9763, + "step": 36931 + }, + { + "epoch": 2.73, + "learning_rate": 1.1936181849772529e-05, + "loss": 1.0951, + "step": 36932 + }, + { + "epoch": 2.73, + "learning_rate": 1.1935790524476626e-05, + "loss": 1.0233, + "step": 36933 + }, + { + "epoch": 2.73, + "learning_rate": 1.1935399196100905e-05, + "loss": 1.0491, + "step": 36934 + }, + { + "epoch": 2.73, + "learning_rate": 1.1935007864645995e-05, + "loss": 0.9296, + "step": 36935 + }, + { + "epoch": 2.73, + "learning_rate": 1.1934616530112514e-05, + "loss": 0.947, + "step": 36936 + }, + { + "epoch": 2.73, + "learning_rate": 1.1934225192501083e-05, + "loss": 0.9758, + "step": 36937 + }, + { + "epoch": 2.73, + "learning_rate": 1.193383385181233e-05, + "loss": 0.9146, + "step": 36938 + }, + { + "epoch": 2.73, + "learning_rate": 1.1933442508046875e-05, + "loss": 1.0144, + "step": 36939 + }, + { + "epoch": 2.73, + "learning_rate": 1.193305116120534e-05, + "loss": 0.9777, + "step": 36940 + }, + { + "epoch": 2.73, + "learning_rate": 1.193265981128835e-05, + "loss": 0.9666, + "step": 36941 + }, + { + "epoch": 2.73, + "learning_rate": 1.1932268458296522e-05, + "loss": 1.0807, + "step": 36942 + }, + { + "epoch": 2.73, + "learning_rate": 1.1931877102230489e-05, + "loss": 0.9992, + "step": 36943 + }, + { + "epoch": 2.73, + "learning_rate": 1.1931485743090863e-05, + "loss": 0.9434, + "step": 36944 + }, + { + "epoch": 2.73, + "learning_rate": 1.1931094380878272e-05, + "loss": 1.0351, + "step": 36945 + }, + { + "epoch": 2.73, + "learning_rate": 1.1930703015593337e-05, + "loss": 0.9394, + "step": 36946 + }, + { + "epoch": 2.73, + "learning_rate": 1.1930311647236681e-05, + "loss": 0.9762, + "step": 36947 + }, + { + "epoch": 2.73, + "learning_rate": 1.1929920275808924e-05, + "loss": 1.0145, + "step": 36948 + }, + { + "epoch": 2.73, + "learning_rate": 1.1929528901310696e-05, + "loss": 1.0281, + "step": 36949 + }, + { + "epoch": 2.73, + "learning_rate": 1.1929137523742611e-05, + "loss": 1.0496, + "step": 36950 + }, + { + "epoch": 2.73, + "learning_rate": 1.1928746143105299e-05, + "loss": 1.0115, + "step": 36951 + }, + { + "epoch": 2.73, + "learning_rate": 1.1928354759399381e-05, + "loss": 0.9653, + "step": 36952 + }, + { + "epoch": 2.73, + "learning_rate": 1.1927963372625476e-05, + "loss": 0.9132, + "step": 36953 + }, + { + "epoch": 2.73, + "learning_rate": 1.1927571982784207e-05, + "loss": 0.9416, + "step": 36954 + }, + { + "epoch": 2.73, + "learning_rate": 1.1927180589876201e-05, + "loss": 0.964, + "step": 36955 + }, + { + "epoch": 2.73, + "learning_rate": 1.1926789193902075e-05, + "loss": 1.0379, + "step": 36956 + }, + { + "epoch": 2.73, + "learning_rate": 1.1926397794862458e-05, + "loss": 1.0574, + "step": 36957 + }, + { + "epoch": 2.73, + "learning_rate": 1.192600639275797e-05, + "loss": 0.9948, + "step": 36958 + }, + { + "epoch": 2.73, + "learning_rate": 1.1925614987589232e-05, + "loss": 1.0018, + "step": 36959 + }, + { + "epoch": 2.73, + "learning_rate": 1.1925223579356868e-05, + "loss": 0.9581, + "step": 36960 + }, + { + "epoch": 2.73, + "learning_rate": 1.1924832168061503e-05, + "loss": 0.9397, + "step": 36961 + }, + { + "epoch": 2.73, + "learning_rate": 1.1924440753703754e-05, + "loss": 0.9114, + "step": 36962 + }, + { + "epoch": 2.73, + "learning_rate": 1.192404933628425e-05, + "loss": 1.0736, + "step": 36963 + }, + { + "epoch": 2.73, + "learning_rate": 1.192365791580361e-05, + "loss": 0.9765, + "step": 36964 + }, + { + "epoch": 2.73, + "learning_rate": 1.1923266492262459e-05, + "loss": 1.0174, + "step": 36965 + }, + { + "epoch": 2.73, + "learning_rate": 1.1922875065661417e-05, + "loss": 0.9292, + "step": 36966 + }, + { + "epoch": 2.73, + "learning_rate": 1.192248363600111e-05, + "loss": 0.9447, + "step": 36967 + }, + { + "epoch": 2.73, + "learning_rate": 1.1922092203282156e-05, + "loss": 0.9276, + "step": 36968 + }, + { + "epoch": 2.73, + "learning_rate": 1.1921700767505184e-05, + "loss": 0.9168, + "step": 36969 + }, + { + "epoch": 2.73, + "learning_rate": 1.1921309328670812e-05, + "loss": 0.981, + "step": 36970 + }, + { + "epoch": 2.73, + "learning_rate": 1.1920917886779665e-05, + "loss": 0.9852, + "step": 36971 + }, + { + "epoch": 2.73, + "learning_rate": 1.1920526441832365e-05, + "loss": 0.9969, + "step": 36972 + }, + { + "epoch": 2.73, + "learning_rate": 1.1920134993829533e-05, + "loss": 1.1055, + "step": 36973 + }, + { + "epoch": 2.73, + "learning_rate": 1.1919743542771795e-05, + "loss": 1.0123, + "step": 36974 + }, + { + "epoch": 2.73, + "learning_rate": 1.1919352088659774e-05, + "loss": 1.0647, + "step": 36975 + }, + { + "epoch": 2.73, + "learning_rate": 1.191896063149409e-05, + "loss": 1.0921, + "step": 36976 + }, + { + "epoch": 2.73, + "learning_rate": 1.1918569171275368e-05, + "loss": 0.9897, + "step": 36977 + }, + { + "epoch": 2.73, + "learning_rate": 1.1918177708004226e-05, + "loss": 0.9133, + "step": 36978 + }, + { + "epoch": 2.73, + "learning_rate": 1.1917786241681295e-05, + "loss": 0.9464, + "step": 36979 + }, + { + "epoch": 2.73, + "learning_rate": 1.1917394772307193e-05, + "loss": 0.9831, + "step": 36980 + }, + { + "epoch": 2.73, + "learning_rate": 1.1917003299882541e-05, + "loss": 0.8874, + "step": 36981 + }, + { + "epoch": 2.73, + "learning_rate": 1.191661182440797e-05, + "loss": 1.1419, + "step": 36982 + }, + { + "epoch": 2.73, + "learning_rate": 1.1916220345884089e-05, + "loss": 0.9315, + "step": 36983 + }, + { + "epoch": 2.73, + "learning_rate": 1.1915828864311534e-05, + "loss": 0.9234, + "step": 36984 + }, + { + "epoch": 2.73, + "learning_rate": 1.191543737969092e-05, + "loss": 1.0413, + "step": 36985 + }, + { + "epoch": 2.73, + "learning_rate": 1.1915045892022877e-05, + "loss": 1.0615, + "step": 36986 + }, + { + "epoch": 2.73, + "learning_rate": 1.1914654401308017e-05, + "loss": 1.0778, + "step": 36987 + }, + { + "epoch": 2.73, + "learning_rate": 1.1914262907546975e-05, + "loss": 0.9093, + "step": 36988 + }, + { + "epoch": 2.73, + "learning_rate": 1.1913871410740363e-05, + "loss": 0.904, + "step": 36989 + }, + { + "epoch": 2.73, + "learning_rate": 1.1913479910888814e-05, + "loss": 0.9563, + "step": 36990 + }, + { + "epoch": 2.73, + "learning_rate": 1.1913088407992943e-05, + "loss": 1.0075, + "step": 36991 + }, + { + "epoch": 2.73, + "learning_rate": 1.1912696902053378e-05, + "loss": 1.0002, + "step": 36992 + }, + { + "epoch": 2.73, + "learning_rate": 1.1912305393070735e-05, + "loss": 0.9777, + "step": 36993 + }, + { + "epoch": 2.73, + "learning_rate": 1.1911913881045644e-05, + "loss": 1.0507, + "step": 36994 + }, + { + "epoch": 2.73, + "learning_rate": 1.1911522365978728e-05, + "loss": 0.9102, + "step": 36995 + }, + { + "epoch": 2.73, + "learning_rate": 1.1911130847870604e-05, + "loss": 1.0638, + "step": 36996 + }, + { + "epoch": 2.73, + "learning_rate": 1.1910739326721896e-05, + "loss": 1.0536, + "step": 36997 + }, + { + "epoch": 2.73, + "learning_rate": 1.1910347802533235e-05, + "loss": 1.0307, + "step": 36998 + }, + { + "epoch": 2.73, + "learning_rate": 1.1909956275305236e-05, + "loss": 0.9416, + "step": 36999 + }, + { + "epoch": 2.73, + "learning_rate": 1.1909564745038524e-05, + "loss": 1.0333, + "step": 37000 + }, + { + "epoch": 2.73, + "learning_rate": 1.1909173211733717e-05, + "loss": 0.9535, + "step": 37001 + }, + { + "epoch": 2.73, + "learning_rate": 1.190878167539145e-05, + "loss": 0.9431, + "step": 37002 + }, + { + "epoch": 2.73, + "learning_rate": 1.1908390136012333e-05, + "loss": 0.8455, + "step": 37003 + }, + { + "epoch": 2.73, + "learning_rate": 1.1907998593597e-05, + "loss": 1.0254, + "step": 37004 + }, + { + "epoch": 2.73, + "learning_rate": 1.1907607048146061e-05, + "loss": 1.0073, + "step": 37005 + }, + { + "epoch": 2.73, + "learning_rate": 1.1907215499660153e-05, + "loss": 1.0551, + "step": 37006 + }, + { + "epoch": 2.73, + "learning_rate": 1.190682394813989e-05, + "loss": 0.9288, + "step": 37007 + }, + { + "epoch": 2.73, + "learning_rate": 1.1906432393585901e-05, + "loss": 0.927, + "step": 37008 + }, + { + "epoch": 2.73, + "learning_rate": 1.19060408359988e-05, + "loss": 0.9996, + "step": 37009 + }, + { + "epoch": 2.73, + "learning_rate": 1.1905649275379216e-05, + "loss": 1.002, + "step": 37010 + }, + { + "epoch": 2.73, + "learning_rate": 1.1905257711727774e-05, + "loss": 1.0184, + "step": 37011 + }, + { + "epoch": 2.73, + "learning_rate": 1.190486614504509e-05, + "loss": 0.8612, + "step": 37012 + }, + { + "epoch": 2.73, + "learning_rate": 1.1904474575331795e-05, + "loss": 1.0665, + "step": 37013 + }, + { + "epoch": 2.73, + "learning_rate": 1.190408300258851e-05, + "loss": 0.9677, + "step": 37014 + }, + { + "epoch": 2.74, + "learning_rate": 1.190369142681585e-05, + "loss": 1.0844, + "step": 37015 + }, + { + "epoch": 2.74, + "learning_rate": 1.1903299848014449e-05, + "loss": 1.0319, + "step": 37016 + }, + { + "epoch": 2.74, + "learning_rate": 1.1902908266184923e-05, + "loss": 0.9755, + "step": 37017 + }, + { + "epoch": 2.74, + "learning_rate": 1.1902516681327897e-05, + "loss": 1.0328, + "step": 37018 + }, + { + "epoch": 2.74, + "learning_rate": 1.1902125093443996e-05, + "loss": 1.1175, + "step": 37019 + }, + { + "epoch": 2.74, + "learning_rate": 1.1901733502533841e-05, + "loss": 0.9334, + "step": 37020 + }, + { + "epoch": 2.74, + "learning_rate": 1.1901341908598054e-05, + "loss": 1.0921, + "step": 37021 + }, + { + "epoch": 2.74, + "learning_rate": 1.190095031163726e-05, + "loss": 0.9129, + "step": 37022 + }, + { + "epoch": 2.74, + "learning_rate": 1.1900558711652079e-05, + "loss": 1.0078, + "step": 37023 + }, + { + "epoch": 2.74, + "learning_rate": 1.190016710864314e-05, + "loss": 0.9195, + "step": 37024 + }, + { + "epoch": 2.74, + "learning_rate": 1.1899775502611059e-05, + "loss": 0.9773, + "step": 37025 + }, + { + "epoch": 2.74, + "learning_rate": 1.1899383893556466e-05, + "loss": 1.0428, + "step": 37026 + }, + { + "epoch": 2.74, + "learning_rate": 1.1898992281479979e-05, + "loss": 0.9612, + "step": 37027 + }, + { + "epoch": 2.74, + "learning_rate": 1.189860066638222e-05, + "loss": 1.02, + "step": 37028 + }, + { + "epoch": 2.74, + "learning_rate": 1.1898209048263817e-05, + "loss": 1.0378, + "step": 37029 + }, + { + "epoch": 2.74, + "learning_rate": 1.189781742712539e-05, + "loss": 1.0563, + "step": 37030 + }, + { + "epoch": 2.74, + "learning_rate": 1.1897425802967562e-05, + "loss": 0.9867, + "step": 37031 + }, + { + "epoch": 2.74, + "learning_rate": 1.1897034175790956e-05, + "loss": 0.981, + "step": 37032 + }, + { + "epoch": 2.74, + "learning_rate": 1.1896642545596197e-05, + "loss": 0.9475, + "step": 37033 + }, + { + "epoch": 2.74, + "learning_rate": 1.1896250912383907e-05, + "loss": 0.9856, + "step": 37034 + }, + { + "epoch": 2.74, + "learning_rate": 1.1895859276154708e-05, + "loss": 1.0193, + "step": 37035 + }, + { + "epoch": 2.74, + "learning_rate": 1.1895467636909226e-05, + "loss": 0.9748, + "step": 37036 + }, + { + "epoch": 2.74, + "learning_rate": 1.1895075994648082e-05, + "loss": 1.026, + "step": 37037 + }, + { + "epoch": 2.74, + "learning_rate": 1.1894684349371896e-05, + "loss": 1.0959, + "step": 37038 + }, + { + "epoch": 2.74, + "learning_rate": 1.1894292701081299e-05, + "loss": 0.9792, + "step": 37039 + }, + { + "epoch": 2.74, + "learning_rate": 1.1893901049776905e-05, + "loss": 1.0133, + "step": 37040 + }, + { + "epoch": 2.74, + "learning_rate": 1.1893509395459344e-05, + "loss": 0.9669, + "step": 37041 + }, + { + "epoch": 2.74, + "learning_rate": 1.1893117738129235e-05, + "loss": 0.9388, + "step": 37042 + }, + { + "epoch": 2.74, + "learning_rate": 1.1892726077787205e-05, + "loss": 0.9849, + "step": 37043 + }, + { + "epoch": 2.74, + "learning_rate": 1.189233441443387e-05, + "loss": 1.0601, + "step": 37044 + }, + { + "epoch": 2.74, + "learning_rate": 1.1891942748069865e-05, + "loss": 0.9974, + "step": 37045 + }, + { + "epoch": 2.74, + "learning_rate": 1.1891551078695802e-05, + "loss": 0.9848, + "step": 37046 + }, + { + "epoch": 2.74, + "learning_rate": 1.189115940631231e-05, + "loss": 0.9902, + "step": 37047 + }, + { + "epoch": 2.74, + "learning_rate": 1.1890767730920008e-05, + "loss": 0.9653, + "step": 37048 + }, + { + "epoch": 2.74, + "learning_rate": 1.1890376052519526e-05, + "loss": 1.0881, + "step": 37049 + }, + { + "epoch": 2.74, + "learning_rate": 1.1889984371111476e-05, + "loss": 1.0146, + "step": 37050 + }, + { + "epoch": 2.74, + "learning_rate": 1.1889592686696494e-05, + "loss": 1.1121, + "step": 37051 + }, + { + "epoch": 2.74, + "learning_rate": 1.1889200999275193e-05, + "loss": 1.0284, + "step": 37052 + }, + { + "epoch": 2.74, + "learning_rate": 1.1888809308848204e-05, + "loss": 0.8757, + "step": 37053 + }, + { + "epoch": 2.74, + "learning_rate": 1.1888417615416143e-05, + "loss": 0.9557, + "step": 37054 + }, + { + "epoch": 2.74, + "learning_rate": 1.188802591897964e-05, + "loss": 0.997, + "step": 37055 + }, + { + "epoch": 2.74, + "learning_rate": 1.1887634219539313e-05, + "loss": 0.9887, + "step": 37056 + }, + { + "epoch": 2.74, + "learning_rate": 1.1887242517095788e-05, + "loss": 1.0302, + "step": 37057 + }, + { + "epoch": 2.74, + "learning_rate": 1.1886850811649684e-05, + "loss": 1.0382, + "step": 37058 + }, + { + "epoch": 2.74, + "learning_rate": 1.188645910320163e-05, + "loss": 1.0146, + "step": 37059 + }, + { + "epoch": 2.74, + "learning_rate": 1.1886067391752244e-05, + "loss": 1.0423, + "step": 37060 + }, + { + "epoch": 2.74, + "learning_rate": 1.1885675677302155e-05, + "loss": 0.983, + "step": 37061 + }, + { + "epoch": 2.74, + "learning_rate": 1.188528395985198e-05, + "loss": 0.8244, + "step": 37062 + }, + { + "epoch": 2.74, + "learning_rate": 1.1884892239402348e-05, + "loss": 1.0768, + "step": 37063 + }, + { + "epoch": 2.74, + "learning_rate": 1.1884500515953878e-05, + "loss": 0.9528, + "step": 37064 + }, + { + "epoch": 2.74, + "learning_rate": 1.1884108789507196e-05, + "loss": 1.0814, + "step": 37065 + }, + { + "epoch": 2.74, + "learning_rate": 1.1883717060062923e-05, + "loss": 0.9418, + "step": 37066 + }, + { + "epoch": 2.74, + "learning_rate": 1.1883325327621682e-05, + "loss": 0.9933, + "step": 37067 + }, + { + "epoch": 2.74, + "learning_rate": 1.1882933592184098e-05, + "loss": 0.8551, + "step": 37068 + }, + { + "epoch": 2.74, + "learning_rate": 1.1882541853750795e-05, + "loss": 0.9405, + "step": 37069 + }, + { + "epoch": 2.74, + "learning_rate": 1.1882150112322392e-05, + "loss": 0.999, + "step": 37070 + }, + { + "epoch": 2.74, + "learning_rate": 1.1881758367899518e-05, + "loss": 1.0, + "step": 37071 + }, + { + "epoch": 2.74, + "learning_rate": 1.1881366620482791e-05, + "loss": 0.9973, + "step": 37072 + }, + { + "epoch": 2.74, + "learning_rate": 1.1880974870072839e-05, + "loss": 0.9216, + "step": 37073 + }, + { + "epoch": 2.74, + "learning_rate": 1.188058311667028e-05, + "loss": 1.053, + "step": 37074 + }, + { + "epoch": 2.74, + "learning_rate": 1.188019136027574e-05, + "loss": 0.9665, + "step": 37075 + }, + { + "epoch": 2.74, + "learning_rate": 1.1879799600889848e-05, + "loss": 0.9564, + "step": 37076 + }, + { + "epoch": 2.74, + "learning_rate": 1.1879407838513215e-05, + "loss": 0.9977, + "step": 37077 + }, + { + "epoch": 2.74, + "learning_rate": 1.1879016073146477e-05, + "loss": 1.1064, + "step": 37078 + }, + { + "epoch": 2.74, + "learning_rate": 1.1878624304790245e-05, + "loss": 0.966, + "step": 37079 + }, + { + "epoch": 2.74, + "learning_rate": 1.1878232533445153e-05, + "loss": 0.911, + "step": 37080 + }, + { + "epoch": 2.74, + "learning_rate": 1.1877840759111821e-05, + "loss": 1.0153, + "step": 37081 + }, + { + "epoch": 2.74, + "learning_rate": 1.1877448981790867e-05, + "loss": 0.9994, + "step": 37082 + }, + { + "epoch": 2.74, + "learning_rate": 1.1877057201482922e-05, + "loss": 1.0174, + "step": 37083 + }, + { + "epoch": 2.74, + "learning_rate": 1.1876665418188604e-05, + "loss": 0.94, + "step": 37084 + }, + { + "epoch": 2.74, + "learning_rate": 1.1876273631908538e-05, + "loss": 1.0218, + "step": 37085 + }, + { + "epoch": 2.74, + "learning_rate": 1.1875881842643349e-05, + "loss": 0.9065, + "step": 37086 + }, + { + "epoch": 2.74, + "learning_rate": 1.1875490050393657e-05, + "loss": 0.9236, + "step": 37087 + }, + { + "epoch": 2.74, + "learning_rate": 1.1875098255160089e-05, + "loss": 1.1386, + "step": 37088 + }, + { + "epoch": 2.74, + "learning_rate": 1.1874706456943263e-05, + "loss": 1.0786, + "step": 37089 + }, + { + "epoch": 2.74, + "learning_rate": 1.1874314655743811e-05, + "loss": 0.9767, + "step": 37090 + }, + { + "epoch": 2.74, + "learning_rate": 1.1873922851562348e-05, + "loss": 1.0164, + "step": 37091 + }, + { + "epoch": 2.74, + "learning_rate": 1.1873531044399503e-05, + "loss": 1.013, + "step": 37092 + }, + { + "epoch": 2.74, + "learning_rate": 1.1873139234255893e-05, + "loss": 1.0177, + "step": 37093 + }, + { + "epoch": 2.74, + "learning_rate": 1.187274742113215e-05, + "loss": 1.029, + "step": 37094 + }, + { + "epoch": 2.74, + "learning_rate": 1.1872355605028887e-05, + "loss": 0.9197, + "step": 37095 + }, + { + "epoch": 2.74, + "learning_rate": 1.1871963785946738e-05, + "loss": 0.9876, + "step": 37096 + }, + { + "epoch": 2.74, + "learning_rate": 1.1871571963886317e-05, + "loss": 0.9731, + "step": 37097 + }, + { + "epoch": 2.74, + "learning_rate": 1.1871180138848256e-05, + "loss": 0.9977, + "step": 37098 + }, + { + "epoch": 2.74, + "learning_rate": 1.187078831083317e-05, + "loss": 0.944, + "step": 37099 + }, + { + "epoch": 2.74, + "learning_rate": 1.1870396479841691e-05, + "loss": 0.9402, + "step": 37100 + }, + { + "epoch": 2.74, + "learning_rate": 1.1870004645874436e-05, + "loss": 0.9914, + "step": 37101 + }, + { + "epoch": 2.74, + "learning_rate": 1.1869612808932032e-05, + "loss": 1.0662, + "step": 37102 + }, + { + "epoch": 2.74, + "learning_rate": 1.1869220969015098e-05, + "loss": 0.9755, + "step": 37103 + }, + { + "epoch": 2.74, + "learning_rate": 1.1868829126124261e-05, + "loss": 0.8913, + "step": 37104 + }, + { + "epoch": 2.74, + "learning_rate": 1.1868437280260144e-05, + "loss": 0.9911, + "step": 37105 + }, + { + "epoch": 2.74, + "learning_rate": 1.1868045431423369e-05, + "loss": 0.9585, + "step": 37106 + }, + { + "epoch": 2.74, + "learning_rate": 1.1867653579614562e-05, + "loss": 0.9663, + "step": 37107 + }, + { + "epoch": 2.74, + "learning_rate": 1.1867261724834344e-05, + "loss": 0.984, + "step": 37108 + }, + { + "epoch": 2.74, + "learning_rate": 1.1866869867083337e-05, + "loss": 0.9151, + "step": 37109 + }, + { + "epoch": 2.74, + "learning_rate": 1.1866478006362172e-05, + "loss": 0.9934, + "step": 37110 + }, + { + "epoch": 2.74, + "learning_rate": 1.1866086142671465e-05, + "loss": 0.9936, + "step": 37111 + }, + { + "epoch": 2.74, + "learning_rate": 1.1865694276011841e-05, + "loss": 0.9563, + "step": 37112 + }, + { + "epoch": 2.74, + "learning_rate": 1.1865302406383923e-05, + "loss": 1.0569, + "step": 37113 + }, + { + "epoch": 2.74, + "learning_rate": 1.1864910533788338e-05, + "loss": 1.01, + "step": 37114 + }, + { + "epoch": 2.74, + "learning_rate": 1.1864518658225706e-05, + "loss": 0.9437, + "step": 37115 + }, + { + "epoch": 2.74, + "learning_rate": 1.1864126779696652e-05, + "loss": 0.9945, + "step": 37116 + }, + { + "epoch": 2.74, + "learning_rate": 1.1863734898201795e-05, + "loss": 1.0564, + "step": 37117 + }, + { + "epoch": 2.74, + "learning_rate": 1.1863343013741768e-05, + "loss": 1.0951, + "step": 37118 + }, + { + "epoch": 2.74, + "learning_rate": 1.1862951126317186e-05, + "loss": 0.9411, + "step": 37119 + }, + { + "epoch": 2.74, + "learning_rate": 1.1862559235928676e-05, + "loss": 0.8744, + "step": 37120 + }, + { + "epoch": 2.74, + "learning_rate": 1.1862167342576862e-05, + "loss": 0.9671, + "step": 37121 + }, + { + "epoch": 2.74, + "learning_rate": 1.1861775446262365e-05, + "loss": 0.98, + "step": 37122 + }, + { + "epoch": 2.74, + "learning_rate": 1.186138354698581e-05, + "loss": 1.0212, + "step": 37123 + }, + { + "epoch": 2.74, + "learning_rate": 1.186099164474782e-05, + "loss": 1.0269, + "step": 37124 + }, + { + "epoch": 2.74, + "learning_rate": 1.1860599739549018e-05, + "loss": 1.0149, + "step": 37125 + }, + { + "epoch": 2.74, + "learning_rate": 1.1860207831390032e-05, + "loss": 1.0338, + "step": 37126 + }, + { + "epoch": 2.74, + "learning_rate": 1.1859815920271479e-05, + "loss": 0.9473, + "step": 37127 + }, + { + "epoch": 2.74, + "learning_rate": 1.1859424006193987e-05, + "loss": 0.9971, + "step": 37128 + }, + { + "epoch": 2.74, + "learning_rate": 1.1859032089158175e-05, + "loss": 0.9756, + "step": 37129 + }, + { + "epoch": 2.74, + "learning_rate": 1.1858640169164675e-05, + "loss": 1.0404, + "step": 37130 + }, + { + "epoch": 2.74, + "learning_rate": 1.1858248246214101e-05, + "loss": 0.8758, + "step": 37131 + }, + { + "epoch": 2.74, + "learning_rate": 1.1857856320307081e-05, + "loss": 1.0448, + "step": 37132 + }, + { + "epoch": 2.74, + "learning_rate": 1.185746439144424e-05, + "loss": 1.0863, + "step": 37133 + }, + { + "epoch": 2.74, + "learning_rate": 1.1857072459626198e-05, + "loss": 1.0205, + "step": 37134 + }, + { + "epoch": 2.74, + "learning_rate": 1.1856680524853581e-05, + "loss": 1.0271, + "step": 37135 + }, + { + "epoch": 2.74, + "learning_rate": 1.1856288587127012e-05, + "loss": 1.057, + "step": 37136 + }, + { + "epoch": 2.74, + "learning_rate": 1.1855896646447114e-05, + "loss": 1.0521, + "step": 37137 + }, + { + "epoch": 2.74, + "learning_rate": 1.1855504702814508e-05, + "loss": 1.0664, + "step": 37138 + }, + { + "epoch": 2.74, + "learning_rate": 1.1855112756229827e-05, + "loss": 0.9908, + "step": 37139 + }, + { + "epoch": 2.74, + "learning_rate": 1.1854720806693684e-05, + "loss": 1.0892, + "step": 37140 + }, + { + "epoch": 2.74, + "learning_rate": 1.1854328854206709e-05, + "loss": 0.956, + "step": 37141 + }, + { + "epoch": 2.74, + "learning_rate": 1.185393689876952e-05, + "loss": 0.94, + "step": 37142 + }, + { + "epoch": 2.74, + "learning_rate": 1.1853544940382749e-05, + "loss": 1.0416, + "step": 37143 + }, + { + "epoch": 2.74, + "learning_rate": 1.1853152979047009e-05, + "loss": 0.9726, + "step": 37144 + }, + { + "epoch": 2.74, + "learning_rate": 1.1852761014762932e-05, + "loss": 1.0815, + "step": 37145 + }, + { + "epoch": 2.74, + "learning_rate": 1.1852369047531137e-05, + "loss": 1.005, + "step": 37146 + }, + { + "epoch": 2.74, + "learning_rate": 1.1851977077352253e-05, + "loss": 1.0048, + "step": 37147 + }, + { + "epoch": 2.74, + "learning_rate": 1.1851585104226897e-05, + "loss": 1.0118, + "step": 37148 + }, + { + "epoch": 2.74, + "learning_rate": 1.1851193128155699e-05, + "loss": 1.109, + "step": 37149 + }, + { + "epoch": 2.74, + "learning_rate": 1.1850801149139272e-05, + "loss": 0.9202, + "step": 37150 + }, + { + "epoch": 2.75, + "learning_rate": 1.1850409167178255e-05, + "loss": 0.982, + "step": 37151 + }, + { + "epoch": 2.75, + "learning_rate": 1.1850017182273257e-05, + "loss": 0.937, + "step": 37152 + }, + { + "epoch": 2.75, + "learning_rate": 1.1849625194424911e-05, + "loss": 1.0395, + "step": 37153 + }, + { + "epoch": 2.75, + "learning_rate": 1.1849233203633836e-05, + "loss": 1.019, + "step": 37154 + }, + { + "epoch": 2.75, + "learning_rate": 1.1848841209900661e-05, + "loss": 0.9926, + "step": 37155 + }, + { + "epoch": 2.75, + "learning_rate": 1.1848449213226003e-05, + "loss": 1.0647, + "step": 37156 + }, + { + "epoch": 2.75, + "learning_rate": 1.1848057213610491e-05, + "loss": 0.9868, + "step": 37157 + }, + { + "epoch": 2.75, + "learning_rate": 1.1847665211054746e-05, + "loss": 1.0342, + "step": 37158 + }, + { + "epoch": 2.75, + "learning_rate": 1.1847273205559391e-05, + "loss": 0.9219, + "step": 37159 + }, + { + "epoch": 2.75, + "learning_rate": 1.1846881197125051e-05, + "loss": 1.0, + "step": 37160 + }, + { + "epoch": 2.75, + "learning_rate": 1.184648918575235e-05, + "loss": 1.0426, + "step": 37161 + }, + { + "epoch": 2.75, + "learning_rate": 1.1846097171441908e-05, + "loss": 0.9761, + "step": 37162 + }, + { + "epoch": 2.75, + "learning_rate": 1.1845705154194357e-05, + "loss": 0.9534, + "step": 37163 + }, + { + "epoch": 2.75, + "learning_rate": 1.1845313134010313e-05, + "loss": 1.2335, + "step": 37164 + }, + { + "epoch": 2.75, + "learning_rate": 1.1844921110890402e-05, + "loss": 1.0529, + "step": 37165 + }, + { + "epoch": 2.75, + "learning_rate": 1.1844529084835246e-05, + "loss": 0.9538, + "step": 37166 + }, + { + "epoch": 2.75, + "learning_rate": 1.1844137055845475e-05, + "loss": 0.9388, + "step": 37167 + }, + { + "epoch": 2.75, + "learning_rate": 1.1843745023921706e-05, + "loss": 0.9984, + "step": 37168 + }, + { + "epoch": 2.75, + "learning_rate": 1.1843352989064565e-05, + "loss": 0.9423, + "step": 37169 + }, + { + "epoch": 2.75, + "learning_rate": 1.1842960951274675e-05, + "loss": 1.0869, + "step": 37170 + }, + { + "epoch": 2.75, + "learning_rate": 1.1842568910552663e-05, + "loss": 0.9683, + "step": 37171 + }, + { + "epoch": 2.75, + "learning_rate": 1.1842176866899147e-05, + "loss": 1.1179, + "step": 37172 + }, + { + "epoch": 2.75, + "learning_rate": 1.1841784820314757e-05, + "loss": 1.0301, + "step": 37173 + }, + { + "epoch": 2.75, + "learning_rate": 1.184139277080011e-05, + "loss": 1.0224, + "step": 37174 + }, + { + "epoch": 2.75, + "learning_rate": 1.1841000718355838e-05, + "loss": 1.0544, + "step": 37175 + }, + { + "epoch": 2.75, + "learning_rate": 1.1840608662982557e-05, + "loss": 0.8921, + "step": 37176 + }, + { + "epoch": 2.75, + "learning_rate": 1.1840216604680894e-05, + "loss": 0.9954, + "step": 37177 + }, + { + "epoch": 2.75, + "learning_rate": 1.1839824543451474e-05, + "loss": 1.0118, + "step": 37178 + }, + { + "epoch": 2.75, + "learning_rate": 1.183943247929492e-05, + "loss": 1.0572, + "step": 37179 + }, + { + "epoch": 2.75, + "learning_rate": 1.1839040412211852e-05, + "loss": 1.0298, + "step": 37180 + }, + { + "epoch": 2.75, + "learning_rate": 1.1838648342202898e-05, + "loss": 0.9953, + "step": 37181 + }, + { + "epoch": 2.75, + "learning_rate": 1.1838256269268683e-05, + "loss": 0.9434, + "step": 37182 + }, + { + "epoch": 2.75, + "learning_rate": 1.1837864193409827e-05, + "loss": 0.9966, + "step": 37183 + }, + { + "epoch": 2.75, + "learning_rate": 1.1837472114626958e-05, + "loss": 1.1132, + "step": 37184 + }, + { + "epoch": 2.75, + "learning_rate": 1.1837080032920695e-05, + "loss": 0.9848, + "step": 37185 + }, + { + "epoch": 2.75, + "learning_rate": 1.1836687948291666e-05, + "loss": 0.9587, + "step": 37186 + }, + { + "epoch": 2.75, + "learning_rate": 1.183629586074049e-05, + "loss": 1.0464, + "step": 37187 + }, + { + "epoch": 2.75, + "learning_rate": 1.1835903770267794e-05, + "loss": 1.0117, + "step": 37188 + }, + { + "epoch": 2.75, + "learning_rate": 1.18355116768742e-05, + "loss": 1.0831, + "step": 37189 + }, + { + "epoch": 2.75, + "learning_rate": 1.1835119580560338e-05, + "loss": 1.0191, + "step": 37190 + }, + { + "epoch": 2.75, + "learning_rate": 1.1834727481326822e-05, + "loss": 0.9972, + "step": 37191 + }, + { + "epoch": 2.75, + "learning_rate": 1.1834335379174284e-05, + "loss": 1.0532, + "step": 37192 + }, + { + "epoch": 2.75, + "learning_rate": 1.1833943274103341e-05, + "loss": 1.0208, + "step": 37193 + }, + { + "epoch": 2.75, + "learning_rate": 1.1833551166114625e-05, + "loss": 1.0356, + "step": 37194 + }, + { + "epoch": 2.75, + "learning_rate": 1.1833159055208754e-05, + "loss": 1.0957, + "step": 37195 + }, + { + "epoch": 2.75, + "learning_rate": 1.1832766941386353e-05, + "loss": 0.9659, + "step": 37196 + }, + { + "epoch": 2.75, + "learning_rate": 1.1832374824648046e-05, + "loss": 1.1033, + "step": 37197 + }, + { + "epoch": 2.75, + "learning_rate": 1.1831982704994455e-05, + "loss": 0.9766, + "step": 37198 + }, + { + "epoch": 2.75, + "learning_rate": 1.1831590582426207e-05, + "loss": 0.9841, + "step": 37199 + }, + { + "epoch": 2.75, + "learning_rate": 1.1831198456943925e-05, + "loss": 1.0907, + "step": 37200 + }, + { + "epoch": 2.75, + "learning_rate": 1.1830806328548232e-05, + "loss": 0.9728, + "step": 37201 + }, + { + "epoch": 2.75, + "learning_rate": 1.1830414197239753e-05, + "loss": 1.0771, + "step": 37202 + }, + { + "epoch": 2.75, + "learning_rate": 1.183002206301911e-05, + "loss": 0.8911, + "step": 37203 + }, + { + "epoch": 2.75, + "learning_rate": 1.182962992588693e-05, + "loss": 1.1253, + "step": 37204 + }, + { + "epoch": 2.75, + "learning_rate": 1.1829237785843832e-05, + "loss": 0.9351, + "step": 37205 + }, + { + "epoch": 2.75, + "learning_rate": 1.1828845642890447e-05, + "loss": 0.9355, + "step": 37206 + }, + { + "epoch": 2.75, + "learning_rate": 1.182845349702739e-05, + "loss": 0.9127, + "step": 37207 + }, + { + "epoch": 2.75, + "learning_rate": 1.1828061348255293e-05, + "loss": 0.9438, + "step": 37208 + }, + { + "epoch": 2.75, + "learning_rate": 1.1827669196574774e-05, + "loss": 1.01, + "step": 37209 + }, + { + "epoch": 2.75, + "learning_rate": 1.1827277041986462e-05, + "loss": 1.0042, + "step": 37210 + }, + { + "epoch": 2.75, + "learning_rate": 1.1826884884490977e-05, + "loss": 0.9876, + "step": 37211 + }, + { + "epoch": 2.75, + "learning_rate": 1.1826492724088945e-05, + "loss": 1.0019, + "step": 37212 + }, + { + "epoch": 2.75, + "learning_rate": 1.1826100560780989e-05, + "loss": 0.9163, + "step": 37213 + }, + { + "epoch": 2.75, + "learning_rate": 1.1825708394567735e-05, + "loss": 1.0354, + "step": 37214 + }, + { + "epoch": 2.75, + "learning_rate": 1.1825316225449802e-05, + "loss": 0.9644, + "step": 37215 + }, + { + "epoch": 2.75, + "learning_rate": 1.182492405342782e-05, + "loss": 1.0385, + "step": 37216 + }, + { + "epoch": 2.75, + "learning_rate": 1.1824531878502405e-05, + "loss": 0.9229, + "step": 37217 + }, + { + "epoch": 2.75, + "learning_rate": 1.1824139700674191e-05, + "loss": 1.0327, + "step": 37218 + }, + { + "epoch": 2.75, + "learning_rate": 1.182374751994379e-05, + "loss": 1.0567, + "step": 37219 + }, + { + "epoch": 2.75, + "learning_rate": 1.1823355336311841e-05, + "loss": 0.9456, + "step": 37220 + }, + { + "epoch": 2.75, + "learning_rate": 1.1822963149778955e-05, + "loss": 1.0504, + "step": 37221 + }, + { + "epoch": 2.75, + "learning_rate": 1.1822570960345763e-05, + "loss": 0.9913, + "step": 37222 + }, + { + "epoch": 2.75, + "learning_rate": 1.1822178768012886e-05, + "loss": 0.9501, + "step": 37223 + }, + { + "epoch": 2.75, + "learning_rate": 1.1821786572780947e-05, + "loss": 0.9781, + "step": 37224 + }, + { + "epoch": 2.75, + "learning_rate": 1.1821394374650574e-05, + "loss": 0.8956, + "step": 37225 + }, + { + "epoch": 2.75, + "learning_rate": 1.1821002173622387e-05, + "loss": 0.8934, + "step": 37226 + }, + { + "epoch": 2.75, + "learning_rate": 1.1820609969697012e-05, + "loss": 0.9735, + "step": 37227 + }, + { + "epoch": 2.75, + "learning_rate": 1.1820217762875073e-05, + "loss": 1.0577, + "step": 37228 + }, + { + "epoch": 2.75, + "learning_rate": 1.1819825553157194e-05, + "loss": 1.0838, + "step": 37229 + }, + { + "epoch": 2.75, + "learning_rate": 1.1819433340543997e-05, + "loss": 0.9871, + "step": 37230 + }, + { + "epoch": 2.75, + "learning_rate": 1.1819041125036109e-05, + "loss": 0.9894, + "step": 37231 + }, + { + "epoch": 2.75, + "learning_rate": 1.1818648906634152e-05, + "loss": 1.0488, + "step": 37232 + }, + { + "epoch": 2.75, + "learning_rate": 1.1818256685338752e-05, + "loss": 1.0659, + "step": 37233 + }, + { + "epoch": 2.75, + "learning_rate": 1.1817864461150528e-05, + "loss": 1.1125, + "step": 37234 + }, + { + "epoch": 2.75, + "learning_rate": 1.1817472234070111e-05, + "loss": 1.0361, + "step": 37235 + }, + { + "epoch": 2.75, + "learning_rate": 1.1817080004098121e-05, + "loss": 1.1255, + "step": 37236 + }, + { + "epoch": 2.75, + "learning_rate": 1.1816687771235182e-05, + "loss": 1.0111, + "step": 37237 + }, + { + "epoch": 2.75, + "learning_rate": 1.1816295535481919e-05, + "loss": 0.9405, + "step": 37238 + }, + { + "epoch": 2.75, + "learning_rate": 1.1815903296838956e-05, + "loss": 0.9744, + "step": 37239 + }, + { + "epoch": 2.75, + "learning_rate": 1.1815511055306914e-05, + "loss": 1.0009, + "step": 37240 + }, + { + "epoch": 2.75, + "learning_rate": 1.1815118810886423e-05, + "loss": 0.9683, + "step": 37241 + }, + { + "epoch": 2.75, + "learning_rate": 1.1814726563578105e-05, + "loss": 1.0309, + "step": 37242 + }, + { + "epoch": 2.75, + "learning_rate": 1.1814334313382583e-05, + "loss": 1.0013, + "step": 37243 + }, + { + "epoch": 2.75, + "learning_rate": 1.181394206030048e-05, + "loss": 0.9906, + "step": 37244 + }, + { + "epoch": 2.75, + "learning_rate": 1.1813549804332421e-05, + "loss": 1.0499, + "step": 37245 + }, + { + "epoch": 2.75, + "learning_rate": 1.1813157545479028e-05, + "loss": 0.9562, + "step": 37246 + }, + { + "epoch": 2.75, + "learning_rate": 1.181276528374093e-05, + "loss": 0.9799, + "step": 37247 + }, + { + "epoch": 2.75, + "learning_rate": 1.1812373019118745e-05, + "loss": 1.0012, + "step": 37248 + }, + { + "epoch": 2.75, + "learning_rate": 1.1811980751613105e-05, + "loss": 0.922, + "step": 37249 + }, + { + "epoch": 2.75, + "learning_rate": 1.1811588481224628e-05, + "loss": 1.0542, + "step": 37250 + }, + { + "epoch": 2.75, + "learning_rate": 1.1811196207953941e-05, + "loss": 0.9986, + "step": 37251 + }, + { + "epoch": 2.75, + "learning_rate": 1.1810803931801664e-05, + "loss": 0.9228, + "step": 37252 + }, + { + "epoch": 2.75, + "learning_rate": 1.1810411652768426e-05, + "loss": 1.0383, + "step": 37253 + }, + { + "epoch": 2.75, + "learning_rate": 1.1810019370854845e-05, + "loss": 0.9936, + "step": 37254 + }, + { + "epoch": 2.75, + "learning_rate": 1.1809627086061553e-05, + "loss": 1.0057, + "step": 37255 + }, + { + "epoch": 2.75, + "learning_rate": 1.1809234798389168e-05, + "loss": 1.039, + "step": 37256 + }, + { + "epoch": 2.75, + "learning_rate": 1.1808842507838318e-05, + "loss": 1.101, + "step": 37257 + }, + { + "epoch": 2.75, + "learning_rate": 1.1808450214409623e-05, + "loss": 1.0657, + "step": 37258 + }, + { + "epoch": 2.75, + "learning_rate": 1.1808057918103715e-05, + "loss": 1.0117, + "step": 37259 + }, + { + "epoch": 2.75, + "learning_rate": 1.1807665618921208e-05, + "loss": 1.1197, + "step": 37260 + }, + { + "epoch": 2.75, + "learning_rate": 1.1807273316862732e-05, + "loss": 0.9973, + "step": 37261 + }, + { + "epoch": 2.75, + "learning_rate": 1.180688101192891e-05, + "loss": 0.941, + "step": 37262 + }, + { + "epoch": 2.75, + "learning_rate": 1.1806488704120368e-05, + "loss": 0.9386, + "step": 37263 + }, + { + "epoch": 2.75, + "learning_rate": 1.1806096393437725e-05, + "loss": 0.9625, + "step": 37264 + }, + { + "epoch": 2.75, + "learning_rate": 1.1805704079881611e-05, + "loss": 0.9809, + "step": 37265 + }, + { + "epoch": 2.75, + "learning_rate": 1.1805311763452644e-05, + "loss": 0.9089, + "step": 37266 + }, + { + "epoch": 2.75, + "learning_rate": 1.1804919444151456e-05, + "loss": 1.0438, + "step": 37267 + }, + { + "epoch": 2.75, + "learning_rate": 1.1804527121978664e-05, + "loss": 1.0762, + "step": 37268 + }, + { + "epoch": 2.75, + "learning_rate": 1.1804134796934898e-05, + "loss": 0.9337, + "step": 37269 + }, + { + "epoch": 2.75, + "learning_rate": 1.1803742469020778e-05, + "loss": 1.0823, + "step": 37270 + }, + { + "epoch": 2.75, + "learning_rate": 1.1803350138236928e-05, + "loss": 1.0771, + "step": 37271 + }, + { + "epoch": 2.75, + "learning_rate": 1.1802957804583977e-05, + "loss": 0.9881, + "step": 37272 + }, + { + "epoch": 2.75, + "learning_rate": 1.1802565468062543e-05, + "loss": 0.9256, + "step": 37273 + }, + { + "epoch": 2.75, + "learning_rate": 1.1802173128673256e-05, + "loss": 0.9715, + "step": 37274 + }, + { + "epoch": 2.75, + "learning_rate": 1.1801780786416737e-05, + "loss": 0.9279, + "step": 37275 + }, + { + "epoch": 2.75, + "learning_rate": 1.1801388441293607e-05, + "loss": 0.9598, + "step": 37276 + }, + { + "epoch": 2.75, + "learning_rate": 1.1800996093304498e-05, + "loss": 1.0268, + "step": 37277 + }, + { + "epoch": 2.75, + "learning_rate": 1.1800603742450025e-05, + "loss": 1.0088, + "step": 37278 + }, + { + "epoch": 2.75, + "learning_rate": 1.1800211388730822e-05, + "loss": 1.0503, + "step": 37279 + }, + { + "epoch": 2.75, + "learning_rate": 1.1799819032147508e-05, + "loss": 1.0648, + "step": 37280 + }, + { + "epoch": 2.75, + "learning_rate": 1.1799426672700704e-05, + "loss": 0.9382, + "step": 37281 + }, + { + "epoch": 2.75, + "learning_rate": 1.179903431039104e-05, + "loss": 0.894, + "step": 37282 + }, + { + "epoch": 2.75, + "learning_rate": 1.1798641945219138e-05, + "loss": 1.054, + "step": 37283 + }, + { + "epoch": 2.75, + "learning_rate": 1.1798249577185624e-05, + "loss": 0.9664, + "step": 37284 + }, + { + "epoch": 2.75, + "learning_rate": 1.1797857206291117e-05, + "loss": 0.9244, + "step": 37285 + }, + { + "epoch": 2.76, + "learning_rate": 1.179746483253625e-05, + "loss": 1.0635, + "step": 37286 + }, + { + "epoch": 2.76, + "learning_rate": 1.1797072455921636e-05, + "loss": 1.0147, + "step": 37287 + }, + { + "epoch": 2.76, + "learning_rate": 1.179668007644791e-05, + "loss": 0.9398, + "step": 37288 + }, + { + "epoch": 2.76, + "learning_rate": 1.1796287694115694e-05, + "loss": 0.9785, + "step": 37289 + }, + { + "epoch": 2.76, + "learning_rate": 1.1795895308925607e-05, + "loss": 0.939, + "step": 37290 + }, + { + "epoch": 2.76, + "learning_rate": 1.1795502920878272e-05, + "loss": 0.9477, + "step": 37291 + }, + { + "epoch": 2.76, + "learning_rate": 1.1795110529974323e-05, + "loss": 1.1557, + "step": 37292 + }, + { + "epoch": 2.76, + "learning_rate": 1.1794718136214375e-05, + "loss": 1.0102, + "step": 37293 + }, + { + "epoch": 2.76, + "learning_rate": 1.1794325739599063e-05, + "loss": 0.9181, + "step": 37294 + }, + { + "epoch": 2.76, + "learning_rate": 1.1793933340128996e-05, + "loss": 0.9447, + "step": 37295 + }, + { + "epoch": 2.76, + "learning_rate": 1.1793540937804812e-05, + "loss": 1.0932, + "step": 37296 + }, + { + "epoch": 2.76, + "learning_rate": 1.1793148532627128e-05, + "loss": 1.0778, + "step": 37297 + }, + { + "epoch": 2.76, + "learning_rate": 1.1792756124596571e-05, + "loss": 0.9771, + "step": 37298 + }, + { + "epoch": 2.76, + "learning_rate": 1.1792363713713764e-05, + "loss": 0.9814, + "step": 37299 + }, + { + "epoch": 2.76, + "learning_rate": 1.1791971299979335e-05, + "loss": 1.1076, + "step": 37300 + }, + { + "epoch": 2.76, + "learning_rate": 1.1791578883393902e-05, + "loss": 1.1179, + "step": 37301 + }, + { + "epoch": 2.76, + "learning_rate": 1.1791186463958094e-05, + "loss": 0.9377, + "step": 37302 + }, + { + "epoch": 2.76, + "learning_rate": 1.1790794041672531e-05, + "loss": 0.9396, + "step": 37303 + }, + { + "epoch": 2.76, + "learning_rate": 1.1790401616537843e-05, + "loss": 1.0094, + "step": 37304 + }, + { + "epoch": 2.76, + "learning_rate": 1.179000918855465e-05, + "loss": 0.9756, + "step": 37305 + }, + { + "epoch": 2.76, + "learning_rate": 1.178961675772358e-05, + "loss": 1.0114, + "step": 37306 + }, + { + "epoch": 2.76, + "learning_rate": 1.1789224324045257e-05, + "loss": 0.9968, + "step": 37307 + }, + { + "epoch": 2.76, + "learning_rate": 1.1788831887520302e-05, + "loss": 1.1352, + "step": 37308 + }, + { + "epoch": 2.76, + "learning_rate": 1.1788439448149337e-05, + "loss": 1.0049, + "step": 37309 + }, + { + "epoch": 2.76, + "learning_rate": 1.1788047005932996e-05, + "loss": 1.0726, + "step": 37310 + }, + { + "epoch": 2.76, + "learning_rate": 1.1787654560871895e-05, + "loss": 0.9092, + "step": 37311 + }, + { + "epoch": 2.76, + "learning_rate": 1.1787262112966662e-05, + "loss": 0.9706, + "step": 37312 + }, + { + "epoch": 2.76, + "learning_rate": 1.178686966221792e-05, + "loss": 0.9501, + "step": 37313 + }, + { + "epoch": 2.76, + "learning_rate": 1.1786477208626295e-05, + "loss": 1.0569, + "step": 37314 + }, + { + "epoch": 2.76, + "learning_rate": 1.1786084752192408e-05, + "loss": 1.1171, + "step": 37315 + }, + { + "epoch": 2.76, + "learning_rate": 1.1785692292916888e-05, + "loss": 1.0454, + "step": 37316 + }, + { + "epoch": 2.76, + "learning_rate": 1.1785299830800358e-05, + "loss": 0.9793, + "step": 37317 + }, + { + "epoch": 2.76, + "learning_rate": 1.178490736584344e-05, + "loss": 1.09, + "step": 37318 + }, + { + "epoch": 2.76, + "learning_rate": 1.1784514898046759e-05, + "loss": 0.9796, + "step": 37319 + }, + { + "epoch": 2.76, + "learning_rate": 1.1784122427410943e-05, + "loss": 1.0646, + "step": 37320 + }, + { + "epoch": 2.76, + "learning_rate": 1.178372995393661e-05, + "loss": 0.9916, + "step": 37321 + }, + { + "epoch": 2.76, + "learning_rate": 1.178333747762439e-05, + "loss": 0.9959, + "step": 37322 + }, + { + "epoch": 2.76, + "learning_rate": 1.1782944998474905e-05, + "loss": 0.9324, + "step": 37323 + }, + { + "epoch": 2.76, + "learning_rate": 1.1782552516488782e-05, + "loss": 0.9873, + "step": 37324 + }, + { + "epoch": 2.76, + "learning_rate": 1.1782160031666641e-05, + "loss": 0.9858, + "step": 37325 + }, + { + "epoch": 2.76, + "learning_rate": 1.178176754400911e-05, + "loss": 0.9597, + "step": 37326 + }, + { + "epoch": 2.76, + "learning_rate": 1.1781375053516814e-05, + "loss": 1.0308, + "step": 37327 + }, + { + "epoch": 2.76, + "learning_rate": 1.1780982560190373e-05, + "loss": 1.0218, + "step": 37328 + }, + { + "epoch": 2.76, + "learning_rate": 1.1780590064030415e-05, + "loss": 0.9923, + "step": 37329 + }, + { + "epoch": 2.76, + "learning_rate": 1.1780197565037564e-05, + "loss": 0.8984, + "step": 37330 + }, + { + "epoch": 2.76, + "learning_rate": 1.1779805063212443e-05, + "loss": 1.0139, + "step": 37331 + }, + { + "epoch": 2.76, + "learning_rate": 1.1779412558555678e-05, + "loss": 0.9901, + "step": 37332 + }, + { + "epoch": 2.76, + "learning_rate": 1.1779020051067896e-05, + "loss": 0.8532, + "step": 37333 + }, + { + "epoch": 2.76, + "learning_rate": 1.1778627540749715e-05, + "loss": 1.0298, + "step": 37334 + }, + { + "epoch": 2.76, + "learning_rate": 1.1778235027601767e-05, + "loss": 1.0285, + "step": 37335 + }, + { + "epoch": 2.76, + "learning_rate": 1.1777842511624669e-05, + "loss": 1.0903, + "step": 37336 + }, + { + "epoch": 2.76, + "learning_rate": 1.177744999281905e-05, + "loss": 0.9665, + "step": 37337 + }, + { + "epoch": 2.76, + "learning_rate": 1.1777057471185533e-05, + "loss": 1.008, + "step": 37338 + }, + { + "epoch": 2.76, + "learning_rate": 1.1776664946724746e-05, + "loss": 0.9512, + "step": 37339 + }, + { + "epoch": 2.76, + "learning_rate": 1.1776272419437305e-05, + "loss": 1.0111, + "step": 37340 + }, + { + "epoch": 2.76, + "learning_rate": 1.1775879889323845e-05, + "loss": 0.9619, + "step": 37341 + }, + { + "epoch": 2.76, + "learning_rate": 1.177548735638498e-05, + "loss": 1.074, + "step": 37342 + }, + { + "epoch": 2.76, + "learning_rate": 1.1775094820621347e-05, + "loss": 0.9561, + "step": 37343 + }, + { + "epoch": 2.76, + "learning_rate": 1.1774702282033557e-05, + "loss": 1.0477, + "step": 37344 + }, + { + "epoch": 2.76, + "learning_rate": 1.1774309740622248e-05, + "loss": 1.0464, + "step": 37345 + }, + { + "epoch": 2.76, + "learning_rate": 1.1773917196388032e-05, + "loss": 0.9831, + "step": 37346 + }, + { + "epoch": 2.76, + "learning_rate": 1.1773524649331543e-05, + "loss": 1.0709, + "step": 37347 + }, + { + "epoch": 2.76, + "learning_rate": 1.1773132099453398e-05, + "loss": 0.8631, + "step": 37348 + }, + { + "epoch": 2.76, + "learning_rate": 1.1772739546754226e-05, + "loss": 1.11, + "step": 37349 + }, + { + "epoch": 2.76, + "learning_rate": 1.1772346991234652e-05, + "loss": 0.9556, + "step": 37350 + }, + { + "epoch": 2.76, + "learning_rate": 1.1771954432895298e-05, + "loss": 0.9771, + "step": 37351 + }, + { + "epoch": 2.76, + "learning_rate": 1.177156187173679e-05, + "loss": 1.0254, + "step": 37352 + }, + { + "epoch": 2.76, + "learning_rate": 1.1771169307759754e-05, + "loss": 0.9999, + "step": 37353 + }, + { + "epoch": 2.76, + "learning_rate": 1.177077674096481e-05, + "loss": 0.9571, + "step": 37354 + }, + { + "epoch": 2.76, + "learning_rate": 1.1770384171352591e-05, + "loss": 0.952, + "step": 37355 + }, + { + "epoch": 2.76, + "learning_rate": 1.1769991598923712e-05, + "loss": 1.0558, + "step": 37356 + }, + { + "epoch": 2.76, + "learning_rate": 1.1769599023678802e-05, + "loss": 1.0912, + "step": 37357 + }, + { + "epoch": 2.76, + "learning_rate": 1.1769206445618484e-05, + "loss": 1.0406, + "step": 37358 + }, + { + "epoch": 2.76, + "learning_rate": 1.1768813864743387e-05, + "loss": 0.8827, + "step": 37359 + }, + { + "epoch": 2.76, + "learning_rate": 1.1768421281054127e-05, + "loss": 1.0453, + "step": 37360 + }, + { + "epoch": 2.76, + "learning_rate": 1.1768028694551341e-05, + "loss": 1.0285, + "step": 37361 + }, + { + "epoch": 2.76, + "learning_rate": 1.1767636105235642e-05, + "loss": 0.9791, + "step": 37362 + }, + { + "epoch": 2.76, + "learning_rate": 1.1767243513107664e-05, + "loss": 1.0144, + "step": 37363 + }, + { + "epoch": 2.76, + "learning_rate": 1.1766850918168022e-05, + "loss": 0.9988, + "step": 37364 + }, + { + "epoch": 2.76, + "learning_rate": 1.176645832041735e-05, + "loss": 1.0707, + "step": 37365 + }, + { + "epoch": 2.76, + "learning_rate": 1.1766065719856261e-05, + "loss": 1.117, + "step": 37366 + }, + { + "epoch": 2.76, + "learning_rate": 1.1765673116485392e-05, + "loss": 1.0251, + "step": 37367 + }, + { + "epoch": 2.76, + "learning_rate": 1.1765280510305361e-05, + "loss": 1.0953, + "step": 37368 + }, + { + "epoch": 2.76, + "learning_rate": 1.1764887901316794e-05, + "loss": 1.0452, + "step": 37369 + }, + { + "epoch": 2.76, + "learning_rate": 1.1764495289520315e-05, + "loss": 1.055, + "step": 37370 + }, + { + "epoch": 2.76, + "learning_rate": 1.176410267491655e-05, + "loss": 1.0054, + "step": 37371 + }, + { + "epoch": 2.76, + "learning_rate": 1.1763710057506123e-05, + "loss": 1.0045, + "step": 37372 + }, + { + "epoch": 2.76, + "learning_rate": 1.1763317437289659e-05, + "loss": 1.1347, + "step": 37373 + }, + { + "epoch": 2.76, + "learning_rate": 1.1762924814267783e-05, + "loss": 1.021, + "step": 37374 + }, + { + "epoch": 2.76, + "learning_rate": 1.1762532188441116e-05, + "loss": 1.0809, + "step": 37375 + }, + { + "epoch": 2.76, + "learning_rate": 1.1762139559810289e-05, + "loss": 0.9637, + "step": 37376 + }, + { + "epoch": 2.76, + "learning_rate": 1.1761746928375919e-05, + "loss": 0.9809, + "step": 37377 + }, + { + "epoch": 2.76, + "learning_rate": 1.1761354294138638e-05, + "loss": 1.0415, + "step": 37378 + }, + { + "epoch": 2.76, + "learning_rate": 1.1760961657099066e-05, + "loss": 0.9812, + "step": 37379 + }, + { + "epoch": 2.76, + "learning_rate": 1.176056901725783e-05, + "loss": 0.9318, + "step": 37380 + }, + { + "epoch": 2.76, + "learning_rate": 1.1760176374615555e-05, + "loss": 0.9201, + "step": 37381 + }, + { + "epoch": 2.76, + "learning_rate": 1.1759783729172862e-05, + "loss": 1.0776, + "step": 37382 + }, + { + "epoch": 2.76, + "learning_rate": 1.1759391080930378e-05, + "loss": 0.9791, + "step": 37383 + }, + { + "epoch": 2.76, + "learning_rate": 1.1758998429888731e-05, + "loss": 1.0358, + "step": 37384 + }, + { + "epoch": 2.76, + "learning_rate": 1.175860577604854e-05, + "loss": 0.9686, + "step": 37385 + }, + { + "epoch": 2.76, + "learning_rate": 1.1758213119410435e-05, + "loss": 1.0382, + "step": 37386 + }, + { + "epoch": 2.76, + "learning_rate": 1.1757820459975036e-05, + "loss": 1.0043, + "step": 37387 + }, + { + "epoch": 2.76, + "learning_rate": 1.1757427797742971e-05, + "loss": 0.9241, + "step": 37388 + }, + { + "epoch": 2.76, + "learning_rate": 1.175703513271486e-05, + "loss": 1.0285, + "step": 37389 + }, + { + "epoch": 2.76, + "learning_rate": 1.1756642464891335e-05, + "loss": 1.0181, + "step": 37390 + }, + { + "epoch": 2.76, + "learning_rate": 1.1756249794273018e-05, + "loss": 0.9959, + "step": 37391 + }, + { + "epoch": 2.76, + "learning_rate": 1.1755857120860532e-05, + "loss": 1.0691, + "step": 37392 + }, + { + "epoch": 2.76, + "learning_rate": 1.1755464444654499e-05, + "loss": 1.1517, + "step": 37393 + }, + { + "epoch": 2.76, + "learning_rate": 1.175507176565555e-05, + "loss": 0.9624, + "step": 37394 + }, + { + "epoch": 2.76, + "learning_rate": 1.1754679083864304e-05, + "loss": 0.9796, + "step": 37395 + }, + { + "epoch": 2.76, + "learning_rate": 1.1754286399281394e-05, + "loss": 0.9667, + "step": 37396 + }, + { + "epoch": 2.76, + "learning_rate": 1.1753893711907433e-05, + "loss": 1.0173, + "step": 37397 + }, + { + "epoch": 2.76, + "learning_rate": 1.175350102174306e-05, + "loss": 0.9392, + "step": 37398 + }, + { + "epoch": 2.76, + "learning_rate": 1.1753108328788884e-05, + "loss": 0.9265, + "step": 37399 + }, + { + "epoch": 2.76, + "learning_rate": 1.1752715633045541e-05, + "loss": 0.9491, + "step": 37400 + }, + { + "epoch": 2.76, + "learning_rate": 1.1752322934513656e-05, + "loss": 1.0523, + "step": 37401 + }, + { + "epoch": 2.76, + "learning_rate": 1.1751930233193846e-05, + "loss": 0.8618, + "step": 37402 + }, + { + "epoch": 2.76, + "learning_rate": 1.175153752908674e-05, + "loss": 0.9196, + "step": 37403 + }, + { + "epoch": 2.76, + "learning_rate": 1.1751144822192966e-05, + "loss": 0.9446, + "step": 37404 + }, + { + "epoch": 2.76, + "learning_rate": 1.1750752112513141e-05, + "loss": 0.9829, + "step": 37405 + }, + { + "epoch": 2.76, + "learning_rate": 1.17503594000479e-05, + "loss": 1.0321, + "step": 37406 + }, + { + "epoch": 2.76, + "learning_rate": 1.1749966684797858e-05, + "loss": 0.9532, + "step": 37407 + }, + { + "epoch": 2.76, + "learning_rate": 1.1749573966763646e-05, + "loss": 0.9966, + "step": 37408 + }, + { + "epoch": 2.76, + "learning_rate": 1.1749181245945885e-05, + "loss": 0.9557, + "step": 37409 + }, + { + "epoch": 2.76, + "learning_rate": 1.1748788522345205e-05, + "loss": 1.1249, + "step": 37410 + }, + { + "epoch": 2.76, + "learning_rate": 1.1748395795962224e-05, + "loss": 1.015, + "step": 37411 + }, + { + "epoch": 2.76, + "learning_rate": 1.1748003066797574e-05, + "loss": 0.9894, + "step": 37412 + }, + { + "epoch": 2.76, + "learning_rate": 1.1747610334851872e-05, + "loss": 0.9099, + "step": 37413 + }, + { + "epoch": 2.76, + "learning_rate": 1.1747217600125748e-05, + "loss": 1.085, + "step": 37414 + }, + { + "epoch": 2.76, + "learning_rate": 1.1746824862619827e-05, + "loss": 1.013, + "step": 37415 + }, + { + "epoch": 2.76, + "learning_rate": 1.1746432122334731e-05, + "loss": 0.9983, + "step": 37416 + }, + { + "epoch": 2.76, + "learning_rate": 1.1746039379271088e-05, + "loss": 1.0062, + "step": 37417 + }, + { + "epoch": 2.76, + "learning_rate": 1.1745646633429521e-05, + "loss": 1.0532, + "step": 37418 + }, + { + "epoch": 2.76, + "learning_rate": 1.1745253884810655e-05, + "loss": 1.0868, + "step": 37419 + }, + { + "epoch": 2.76, + "learning_rate": 1.1744861133415118e-05, + "loss": 0.9515, + "step": 37420 + }, + { + "epoch": 2.77, + "learning_rate": 1.1744468379243528e-05, + "loss": 0.9824, + "step": 37421 + }, + { + "epoch": 2.77, + "learning_rate": 1.1744075622296514e-05, + "loss": 1.0005, + "step": 37422 + }, + { + "epoch": 2.77, + "learning_rate": 1.1743682862574704e-05, + "loss": 1.0272, + "step": 37423 + }, + { + "epoch": 2.77, + "learning_rate": 1.1743290100078717e-05, + "loss": 1.0548, + "step": 37424 + }, + { + "epoch": 2.77, + "learning_rate": 1.1742897334809179e-05, + "loss": 0.9756, + "step": 37425 + }, + { + "epoch": 2.77, + "learning_rate": 1.1742504566766719e-05, + "loss": 0.9215, + "step": 37426 + }, + { + "epoch": 2.77, + "learning_rate": 1.1742111795951955e-05, + "loss": 0.8674, + "step": 37427 + }, + { + "epoch": 2.77, + "learning_rate": 1.174171902236552e-05, + "loss": 1.0435, + "step": 37428 + }, + { + "epoch": 2.77, + "learning_rate": 1.1741326246008036e-05, + "loss": 0.9035, + "step": 37429 + }, + { + "epoch": 2.77, + "learning_rate": 1.1740933466880122e-05, + "loss": 0.9782, + "step": 37430 + }, + { + "epoch": 2.77, + "learning_rate": 1.1740540684982414e-05, + "loss": 1.0329, + "step": 37431 + }, + { + "epoch": 2.77, + "learning_rate": 1.1740147900315524e-05, + "loss": 0.9897, + "step": 37432 + }, + { + "epoch": 2.77, + "learning_rate": 1.1739755112880089e-05, + "loss": 0.8888, + "step": 37433 + }, + { + "epoch": 2.77, + "learning_rate": 1.1739362322676724e-05, + "loss": 1.0192, + "step": 37434 + }, + { + "epoch": 2.77, + "learning_rate": 1.1738969529706061e-05, + "loss": 1.0806, + "step": 37435 + }, + { + "epoch": 2.77, + "learning_rate": 1.173857673396872e-05, + "loss": 1.0708, + "step": 37436 + }, + { + "epoch": 2.77, + "learning_rate": 1.1738183935465332e-05, + "loss": 1.0366, + "step": 37437 + }, + { + "epoch": 2.77, + "learning_rate": 1.173779113419652e-05, + "loss": 0.9721, + "step": 37438 + }, + { + "epoch": 2.77, + "learning_rate": 1.1737398330162904e-05, + "loss": 1.0032, + "step": 37439 + }, + { + "epoch": 2.77, + "learning_rate": 1.173700552336511e-05, + "loss": 0.9885, + "step": 37440 + }, + { + "epoch": 2.77, + "learning_rate": 1.1736612713803768e-05, + "loss": 0.9451, + "step": 37441 + }, + { + "epoch": 2.77, + "learning_rate": 1.1736219901479498e-05, + "loss": 1.1049, + "step": 37442 + }, + { + "epoch": 2.77, + "learning_rate": 1.1735827086392928e-05, + "loss": 0.9633, + "step": 37443 + }, + { + "epoch": 2.77, + "learning_rate": 1.1735434268544682e-05, + "loss": 1.0674, + "step": 37444 + }, + { + "epoch": 2.77, + "learning_rate": 1.1735041447935384e-05, + "loss": 1.0285, + "step": 37445 + }, + { + "epoch": 2.77, + "learning_rate": 1.1734648624565662e-05, + "loss": 1.0271, + "step": 37446 + }, + { + "epoch": 2.77, + "learning_rate": 1.1734255798436137e-05, + "loss": 1.0482, + "step": 37447 + }, + { + "epoch": 2.77, + "learning_rate": 1.1733862969547436e-05, + "loss": 0.9478, + "step": 37448 + }, + { + "epoch": 2.77, + "learning_rate": 1.1733470137900188e-05, + "loss": 1.0912, + "step": 37449 + }, + { + "epoch": 2.77, + "learning_rate": 1.1733077303495007e-05, + "loss": 1.0917, + "step": 37450 + }, + { + "epoch": 2.77, + "learning_rate": 1.1732684466332529e-05, + "loss": 0.9598, + "step": 37451 + }, + { + "epoch": 2.77, + "learning_rate": 1.1732291626413372e-05, + "loss": 1.013, + "step": 37452 + }, + { + "epoch": 2.77, + "learning_rate": 1.1731898783738168e-05, + "loss": 0.9296, + "step": 37453 + }, + { + "epoch": 2.77, + "learning_rate": 1.1731505938307532e-05, + "loss": 0.9641, + "step": 37454 + }, + { + "epoch": 2.77, + "learning_rate": 1.1731113090122099e-05, + "loss": 1.0994, + "step": 37455 + }, + { + "epoch": 2.77, + "learning_rate": 1.1730720239182488e-05, + "loss": 1.0245, + "step": 37456 + }, + { + "epoch": 2.77, + "learning_rate": 1.1730327385489328e-05, + "loss": 1.0152, + "step": 37457 + }, + { + "epoch": 2.77, + "learning_rate": 1.172993452904324e-05, + "loss": 1.0021, + "step": 37458 + }, + { + "epoch": 2.77, + "learning_rate": 1.1729541669844853e-05, + "loss": 1.065, + "step": 37459 + }, + { + "epoch": 2.77, + "learning_rate": 1.1729148807894788e-05, + "loss": 1.012, + "step": 37460 + }, + { + "epoch": 2.77, + "learning_rate": 1.1728755943193673e-05, + "loss": 0.9402, + "step": 37461 + }, + { + "epoch": 2.77, + "learning_rate": 1.1728363075742132e-05, + "loss": 1.0333, + "step": 37462 + }, + { + "epoch": 2.77, + "learning_rate": 1.1727970205540788e-05, + "loss": 1.0782, + "step": 37463 + }, + { + "epoch": 2.77, + "learning_rate": 1.172757733259027e-05, + "loss": 1.0278, + "step": 37464 + }, + { + "epoch": 2.77, + "learning_rate": 1.1727184456891201e-05, + "loss": 1.0109, + "step": 37465 + }, + { + "epoch": 2.77, + "learning_rate": 1.1726791578444208e-05, + "loss": 1.0494, + "step": 37466 + }, + { + "epoch": 2.77, + "learning_rate": 1.172639869724991e-05, + "loss": 0.9411, + "step": 37467 + }, + { + "epoch": 2.77, + "learning_rate": 1.172600581330894e-05, + "loss": 0.9823, + "step": 37468 + }, + { + "epoch": 2.77, + "learning_rate": 1.172561292662192e-05, + "loss": 0.9962, + "step": 37469 + }, + { + "epoch": 2.77, + "learning_rate": 1.1725220037189472e-05, + "loss": 1.0012, + "step": 37470 + }, + { + "epoch": 2.77, + "learning_rate": 1.1724827145012226e-05, + "loss": 0.9163, + "step": 37471 + }, + { + "epoch": 2.77, + "learning_rate": 1.17244342500908e-05, + "loss": 0.9516, + "step": 37472 + }, + { + "epoch": 2.77, + "learning_rate": 1.1724041352425827e-05, + "loss": 1.0558, + "step": 37473 + }, + { + "epoch": 2.77, + "learning_rate": 1.1723648452017928e-05, + "loss": 0.9864, + "step": 37474 + }, + { + "epoch": 2.77, + "learning_rate": 1.1723255548867731e-05, + "loss": 1.0353, + "step": 37475 + }, + { + "epoch": 2.77, + "learning_rate": 1.1722862642975858e-05, + "loss": 0.9433, + "step": 37476 + }, + { + "epoch": 2.77, + "learning_rate": 1.1722469734342934e-05, + "loss": 0.8861, + "step": 37477 + }, + { + "epoch": 2.77, + "learning_rate": 1.1722076822969588e-05, + "loss": 1.041, + "step": 37478 + }, + { + "epoch": 2.77, + "learning_rate": 1.172168390885644e-05, + "loss": 1.0922, + "step": 37479 + }, + { + "epoch": 2.77, + "learning_rate": 1.172129099200412e-05, + "loss": 1.0317, + "step": 37480 + }, + { + "epoch": 2.77, + "learning_rate": 1.1720898072413248e-05, + "loss": 1.0234, + "step": 37481 + }, + { + "epoch": 2.77, + "learning_rate": 1.1720505150084451e-05, + "loss": 0.9359, + "step": 37482 + }, + { + "epoch": 2.77, + "learning_rate": 1.1720112225018356e-05, + "loss": 0.9479, + "step": 37483 + }, + { + "epoch": 2.77, + "learning_rate": 1.1719719297215588e-05, + "loss": 1.0593, + "step": 37484 + }, + { + "epoch": 2.77, + "learning_rate": 1.1719326366676772e-05, + "loss": 0.9743, + "step": 37485 + }, + { + "epoch": 2.77, + "learning_rate": 1.1718933433402533e-05, + "loss": 0.9689, + "step": 37486 + }, + { + "epoch": 2.77, + "learning_rate": 1.1718540497393492e-05, + "loss": 0.9279, + "step": 37487 + }, + { + "epoch": 2.77, + "learning_rate": 1.1718147558650281e-05, + "loss": 1.1172, + "step": 37488 + }, + { + "epoch": 2.77, + "learning_rate": 1.1717754617173517e-05, + "loss": 0.8696, + "step": 37489 + }, + { + "epoch": 2.77, + "learning_rate": 1.1717361672963837e-05, + "loss": 1.086, + "step": 37490 + }, + { + "epoch": 2.77, + "learning_rate": 1.1716968726021852e-05, + "loss": 0.9336, + "step": 37491 + }, + { + "epoch": 2.77, + "learning_rate": 1.1716575776348197e-05, + "loss": 1.0416, + "step": 37492 + }, + { + "epoch": 2.77, + "learning_rate": 1.1716182823943495e-05, + "loss": 1.0216, + "step": 37493 + }, + { + "epoch": 2.77, + "learning_rate": 1.1715789868808374e-05, + "loss": 0.9668, + "step": 37494 + }, + { + "epoch": 2.77, + "learning_rate": 1.1715396910943452e-05, + "loss": 0.9175, + "step": 37495 + }, + { + "epoch": 2.77, + "learning_rate": 1.171500395034936e-05, + "loss": 1.0668, + "step": 37496 + }, + { + "epoch": 2.77, + "learning_rate": 1.1714610987026718e-05, + "loss": 1.0013, + "step": 37497 + }, + { + "epoch": 2.77, + "learning_rate": 1.1714218020976157e-05, + "loss": 0.9756, + "step": 37498 + }, + { + "epoch": 2.77, + "learning_rate": 1.17138250521983e-05, + "loss": 0.9814, + "step": 37499 + }, + { + "epoch": 2.77, + "learning_rate": 1.1713432080693772e-05, + "loss": 0.9644, + "step": 37500 + }, + { + "epoch": 2.77, + "learning_rate": 1.1713039106463196e-05, + "loss": 1.0201, + "step": 37501 + }, + { + "epoch": 2.77, + "learning_rate": 1.1712646129507198e-05, + "loss": 0.938, + "step": 37502 + }, + { + "epoch": 2.77, + "learning_rate": 1.1712253149826408e-05, + "loss": 1.1156, + "step": 37503 + }, + { + "epoch": 2.77, + "learning_rate": 1.171186016742145e-05, + "loss": 1.0661, + "step": 37504 + }, + { + "epoch": 2.77, + "learning_rate": 1.1711467182292941e-05, + "loss": 1.0187, + "step": 37505 + }, + { + "epoch": 2.77, + "learning_rate": 1.1711074194441517e-05, + "loss": 0.9193, + "step": 37506 + }, + { + "epoch": 2.77, + "learning_rate": 1.1710681203867794e-05, + "loss": 1.0082, + "step": 37507 + }, + { + "epoch": 2.77, + "learning_rate": 1.1710288210572405e-05, + "loss": 1.0191, + "step": 37508 + }, + { + "epoch": 2.77, + "learning_rate": 1.1709895214555969e-05, + "loss": 0.9454, + "step": 37509 + }, + { + "epoch": 2.77, + "learning_rate": 1.1709502215819117e-05, + "loss": 1.0543, + "step": 37510 + }, + { + "epoch": 2.77, + "learning_rate": 1.170910921436247e-05, + "loss": 0.9216, + "step": 37511 + }, + { + "epoch": 2.77, + "learning_rate": 1.1708716210186654e-05, + "loss": 1.0693, + "step": 37512 + }, + { + "epoch": 2.77, + "learning_rate": 1.1708323203292297e-05, + "loss": 0.9775, + "step": 37513 + }, + { + "epoch": 2.77, + "learning_rate": 1.1707930193680023e-05, + "loss": 0.9925, + "step": 37514 + }, + { + "epoch": 2.77, + "learning_rate": 1.1707537181350454e-05, + "loss": 0.9401, + "step": 37515 + }, + { + "epoch": 2.77, + "learning_rate": 1.170714416630422e-05, + "loss": 1.0065, + "step": 37516 + }, + { + "epoch": 2.77, + "learning_rate": 1.1706751148541942e-05, + "loss": 1.0713, + "step": 37517 + }, + { + "epoch": 2.77, + "learning_rate": 1.1706358128064248e-05, + "loss": 1.1138, + "step": 37518 + }, + { + "epoch": 2.77, + "learning_rate": 1.1705965104871761e-05, + "loss": 1.0423, + "step": 37519 + }, + { + "epoch": 2.77, + "learning_rate": 1.1705572078965111e-05, + "loss": 0.937, + "step": 37520 + }, + { + "epoch": 2.77, + "learning_rate": 1.1705179050344916e-05, + "loss": 0.9645, + "step": 37521 + }, + { + "epoch": 2.77, + "learning_rate": 1.170478601901181e-05, + "loss": 0.9802, + "step": 37522 + }, + { + "epoch": 2.77, + "learning_rate": 1.1704392984966412e-05, + "loss": 0.9402, + "step": 37523 + }, + { + "epoch": 2.77, + "learning_rate": 1.170399994820935e-05, + "loss": 0.9969, + "step": 37524 + }, + { + "epoch": 2.77, + "learning_rate": 1.1703606908741248e-05, + "loss": 1.1338, + "step": 37525 + }, + { + "epoch": 2.77, + "learning_rate": 1.170321386656273e-05, + "loss": 0.9738, + "step": 37526 + }, + { + "epoch": 2.77, + "learning_rate": 1.1702820821674426e-05, + "loss": 0.927, + "step": 37527 + }, + { + "epoch": 2.77, + "learning_rate": 1.1702427774076956e-05, + "loss": 0.9681, + "step": 37528 + }, + { + "epoch": 2.77, + "learning_rate": 1.170203472377095e-05, + "loss": 0.9689, + "step": 37529 + }, + { + "epoch": 2.77, + "learning_rate": 1.1701641670757031e-05, + "loss": 0.935, + "step": 37530 + }, + { + "epoch": 2.77, + "learning_rate": 1.170124861503582e-05, + "loss": 1.0016, + "step": 37531 + }, + { + "epoch": 2.77, + "learning_rate": 1.1700855556607952e-05, + "loss": 0.8868, + "step": 37532 + }, + { + "epoch": 2.77, + "learning_rate": 1.1700462495474047e-05, + "loss": 1.0002, + "step": 37533 + }, + { + "epoch": 2.77, + "learning_rate": 1.1700069431634726e-05, + "loss": 0.8525, + "step": 37534 + }, + { + "epoch": 2.77, + "learning_rate": 1.1699676365090625e-05, + "loss": 0.9348, + "step": 37535 + }, + { + "epoch": 2.77, + "learning_rate": 1.1699283295842359e-05, + "loss": 1.031, + "step": 37536 + }, + { + "epoch": 2.77, + "learning_rate": 1.1698890223890557e-05, + "loss": 0.9813, + "step": 37537 + }, + { + "epoch": 2.77, + "learning_rate": 1.1698497149235845e-05, + "loss": 1.1134, + "step": 37538 + }, + { + "epoch": 2.77, + "learning_rate": 1.1698104071878853e-05, + "loss": 1.042, + "step": 37539 + }, + { + "epoch": 2.77, + "learning_rate": 1.1697710991820198e-05, + "loss": 0.9935, + "step": 37540 + }, + { + "epoch": 2.77, + "learning_rate": 1.169731790906051e-05, + "loss": 0.9995, + "step": 37541 + }, + { + "epoch": 2.77, + "learning_rate": 1.1696924823600413e-05, + "loss": 1.1222, + "step": 37542 + }, + { + "epoch": 2.77, + "learning_rate": 1.1696531735440533e-05, + "loss": 1.049, + "step": 37543 + }, + { + "epoch": 2.77, + "learning_rate": 1.1696138644581494e-05, + "loss": 0.9965, + "step": 37544 + }, + { + "epoch": 2.77, + "learning_rate": 1.1695745551023925e-05, + "loss": 1.057, + "step": 37545 + }, + { + "epoch": 2.77, + "learning_rate": 1.1695352454768447e-05, + "loss": 0.9685, + "step": 37546 + }, + { + "epoch": 2.77, + "learning_rate": 1.169495935581569e-05, + "loss": 0.9284, + "step": 37547 + }, + { + "epoch": 2.77, + "learning_rate": 1.1694566254166273e-05, + "loss": 1.1087, + "step": 37548 + }, + { + "epoch": 2.77, + "learning_rate": 1.169417314982083e-05, + "loss": 0.9537, + "step": 37549 + }, + { + "epoch": 2.77, + "learning_rate": 1.169378004277998e-05, + "loss": 1.027, + "step": 37550 + }, + { + "epoch": 2.77, + "learning_rate": 1.1693386933044351e-05, + "loss": 0.9878, + "step": 37551 + }, + { + "epoch": 2.77, + "learning_rate": 1.1692993820614564e-05, + "loss": 0.9769, + "step": 37552 + }, + { + "epoch": 2.77, + "learning_rate": 1.1692600705491252e-05, + "loss": 1.0558, + "step": 37553 + }, + { + "epoch": 2.77, + "learning_rate": 1.1692207587675034e-05, + "loss": 1.0597, + "step": 37554 + }, + { + "epoch": 2.77, + "learning_rate": 1.169181446716654e-05, + "loss": 0.8811, + "step": 37555 + }, + { + "epoch": 2.77, + "learning_rate": 1.1691421343966392e-05, + "loss": 0.964, + "step": 37556 + }, + { + "epoch": 2.78, + "learning_rate": 1.1691028218075217e-05, + "loss": 1.0891, + "step": 37557 + }, + { + "epoch": 2.78, + "learning_rate": 1.1690635089493639e-05, + "loss": 0.9267, + "step": 37558 + }, + { + "epoch": 2.78, + "learning_rate": 1.1690241958222285e-05, + "loss": 1.0003, + "step": 37559 + }, + { + "epoch": 2.78, + "learning_rate": 1.1689848824261785e-05, + "loss": 0.9882, + "step": 37560 + }, + { + "epoch": 2.78, + "learning_rate": 1.1689455687612757e-05, + "loss": 1.0267, + "step": 37561 + }, + { + "epoch": 2.78, + "learning_rate": 1.1689062548275825e-05, + "loss": 1.0623, + "step": 37562 + }, + { + "epoch": 2.78, + "learning_rate": 1.1688669406251622e-05, + "loss": 1.0625, + "step": 37563 + }, + { + "epoch": 2.78, + "learning_rate": 1.168827626154077e-05, + "loss": 1.0756, + "step": 37564 + }, + { + "epoch": 2.78, + "learning_rate": 1.1687883114143897e-05, + "loss": 0.9874, + "step": 37565 + }, + { + "epoch": 2.78, + "learning_rate": 1.1687489964061622e-05, + "loss": 0.9395, + "step": 37566 + }, + { + "epoch": 2.78, + "learning_rate": 1.1687096811294577e-05, + "loss": 1.0282, + "step": 37567 + }, + { + "epoch": 2.78, + "learning_rate": 1.1686703655843384e-05, + "loss": 0.9494, + "step": 37568 + }, + { + "epoch": 2.78, + "learning_rate": 1.168631049770867e-05, + "loss": 0.9902, + "step": 37569 + }, + { + "epoch": 2.78, + "learning_rate": 1.1685917336891062e-05, + "loss": 0.9564, + "step": 37570 + }, + { + "epoch": 2.78, + "learning_rate": 1.168552417339118e-05, + "loss": 1.0554, + "step": 37571 + }, + { + "epoch": 2.78, + "learning_rate": 1.1685131007209656e-05, + "loss": 1.0538, + "step": 37572 + }, + { + "epoch": 2.78, + "learning_rate": 1.168473783834711e-05, + "loss": 1.1092, + "step": 37573 + }, + { + "epoch": 2.78, + "learning_rate": 1.1684344666804173e-05, + "loss": 0.9106, + "step": 37574 + }, + { + "epoch": 2.78, + "learning_rate": 1.1683951492581466e-05, + "loss": 1.0331, + "step": 37575 + }, + { + "epoch": 2.78, + "learning_rate": 1.1683558315679618e-05, + "loss": 0.9746, + "step": 37576 + }, + { + "epoch": 2.78, + "learning_rate": 1.1683165136099253e-05, + "loss": 1.1462, + "step": 37577 + }, + { + "epoch": 2.78, + "learning_rate": 1.1682771953840995e-05, + "loss": 0.9716, + "step": 37578 + }, + { + "epoch": 2.78, + "learning_rate": 1.1682378768905469e-05, + "loss": 1.039, + "step": 37579 + }, + { + "epoch": 2.78, + "learning_rate": 1.1681985581293304e-05, + "loss": 0.9827, + "step": 37580 + }, + { + "epoch": 2.78, + "learning_rate": 1.1681592391005125e-05, + "loss": 0.9678, + "step": 37581 + }, + { + "epoch": 2.78, + "learning_rate": 1.1681199198041558e-05, + "loss": 0.9552, + "step": 37582 + }, + { + "epoch": 2.78, + "learning_rate": 1.1680806002403222e-05, + "loss": 0.9909, + "step": 37583 + }, + { + "epoch": 2.78, + "learning_rate": 1.1680412804090752e-05, + "loss": 0.9796, + "step": 37584 + }, + { + "epoch": 2.78, + "learning_rate": 1.1680019603104767e-05, + "loss": 0.9136, + "step": 37585 + }, + { + "epoch": 2.78, + "learning_rate": 1.1679626399445896e-05, + "loss": 1.0469, + "step": 37586 + }, + { + "epoch": 2.78, + "learning_rate": 1.1679233193114762e-05, + "loss": 0.972, + "step": 37587 + }, + { + "epoch": 2.78, + "learning_rate": 1.1678839984111998e-05, + "loss": 0.955, + "step": 37588 + }, + { + "epoch": 2.78, + "learning_rate": 1.1678446772438217e-05, + "loss": 0.9291, + "step": 37589 + }, + { + "epoch": 2.78, + "learning_rate": 1.1678053558094053e-05, + "loss": 0.9572, + "step": 37590 + }, + { + "epoch": 2.78, + "learning_rate": 1.1677660341080131e-05, + "loss": 1.0644, + "step": 37591 + }, + { + "epoch": 2.78, + "learning_rate": 1.1677267121397073e-05, + "loss": 1.0968, + "step": 37592 + }, + { + "epoch": 2.78, + "learning_rate": 1.167687389904551e-05, + "loss": 0.9592, + "step": 37593 + }, + { + "epoch": 2.78, + "learning_rate": 1.1676480674026062e-05, + "loss": 0.9577, + "step": 37594 + }, + { + "epoch": 2.78, + "learning_rate": 1.1676087446339357e-05, + "loss": 1.1149, + "step": 37595 + }, + { + "epoch": 2.78, + "learning_rate": 1.1675694215986025e-05, + "loss": 1.1274, + "step": 37596 + }, + { + "epoch": 2.78, + "learning_rate": 1.1675300982966683e-05, + "loss": 1.0749, + "step": 37597 + }, + { + "epoch": 2.78, + "learning_rate": 1.1674907747281965e-05, + "loss": 0.9872, + "step": 37598 + }, + { + "epoch": 2.78, + "learning_rate": 1.1674514508932489e-05, + "loss": 1.0252, + "step": 37599 + }, + { + "epoch": 2.78, + "learning_rate": 1.1674121267918886e-05, + "loss": 0.8724, + "step": 37600 + }, + { + "epoch": 2.78, + "learning_rate": 1.167372802424178e-05, + "loss": 1.1011, + "step": 37601 + }, + { + "epoch": 2.78, + "learning_rate": 1.1673334777901796e-05, + "loss": 0.9076, + "step": 37602 + }, + { + "epoch": 2.78, + "learning_rate": 1.167294152889956e-05, + "loss": 0.9289, + "step": 37603 + }, + { + "epoch": 2.78, + "learning_rate": 1.16725482772357e-05, + "loss": 0.9969, + "step": 37604 + }, + { + "epoch": 2.78, + "learning_rate": 1.1672155022910835e-05, + "loss": 1.0193, + "step": 37605 + }, + { + "epoch": 2.78, + "learning_rate": 1.1671761765925603e-05, + "loss": 0.8585, + "step": 37606 + }, + { + "epoch": 2.78, + "learning_rate": 1.1671368506280616e-05, + "loss": 0.8848, + "step": 37607 + }, + { + "epoch": 2.78, + "learning_rate": 1.1670975243976508e-05, + "loss": 1.0691, + "step": 37608 + }, + { + "epoch": 2.78, + "learning_rate": 1.16705819790139e-05, + "loss": 1.0175, + "step": 37609 + }, + { + "epoch": 2.78, + "learning_rate": 1.1670188711393422e-05, + "loss": 0.956, + "step": 37610 + }, + { + "epoch": 2.78, + "learning_rate": 1.1669795441115697e-05, + "loss": 1.0208, + "step": 37611 + }, + { + "epoch": 2.78, + "learning_rate": 1.1669402168181352e-05, + "loss": 1.0836, + "step": 37612 + }, + { + "epoch": 2.78, + "learning_rate": 1.1669008892591012e-05, + "loss": 0.9447, + "step": 37613 + }, + { + "epoch": 2.78, + "learning_rate": 1.1668615614345302e-05, + "loss": 1.0908, + "step": 37614 + }, + { + "epoch": 2.78, + "learning_rate": 1.1668222333444848e-05, + "loss": 1.133, + "step": 37615 + }, + { + "epoch": 2.78, + "learning_rate": 1.1667829049890278e-05, + "loss": 0.9702, + "step": 37616 + }, + { + "epoch": 2.78, + "learning_rate": 1.1667435763682215e-05, + "loss": 0.9785, + "step": 37617 + }, + { + "epoch": 2.78, + "learning_rate": 1.1667042474821287e-05, + "loss": 1.0037, + "step": 37618 + }, + { + "epoch": 2.78, + "learning_rate": 1.1666649183308115e-05, + "loss": 1.0089, + "step": 37619 + }, + { + "epoch": 2.78, + "learning_rate": 1.166625588914333e-05, + "loss": 1.0824, + "step": 37620 + }, + { + "epoch": 2.78, + "learning_rate": 1.1665862592327552e-05, + "loss": 0.8822, + "step": 37621 + }, + { + "epoch": 2.78, + "learning_rate": 1.1665469292861415e-05, + "loss": 1.0667, + "step": 37622 + }, + { + "epoch": 2.78, + "learning_rate": 1.1665075990745539e-05, + "loss": 0.9458, + "step": 37623 + }, + { + "epoch": 2.78, + "learning_rate": 1.166468268598055e-05, + "loss": 0.9285, + "step": 37624 + }, + { + "epoch": 2.78, + "learning_rate": 1.1664289378567077e-05, + "loss": 1.0587, + "step": 37625 + }, + { + "epoch": 2.78, + "learning_rate": 1.166389606850574e-05, + "loss": 0.9349, + "step": 37626 + }, + { + "epoch": 2.78, + "learning_rate": 1.1663502755797172e-05, + "loss": 1.0767, + "step": 37627 + }, + { + "epoch": 2.78, + "learning_rate": 1.166310944044199e-05, + "loss": 1.0336, + "step": 37628 + }, + { + "epoch": 2.78, + "learning_rate": 1.166271612244083e-05, + "loss": 0.9911, + "step": 37629 + }, + { + "epoch": 2.78, + "learning_rate": 1.1662322801794309e-05, + "loss": 0.9839, + "step": 37630 + }, + { + "epoch": 2.78, + "learning_rate": 1.1661929478503058e-05, + "loss": 0.9357, + "step": 37631 + }, + { + "epoch": 2.78, + "learning_rate": 1.1661536152567698e-05, + "loss": 0.963, + "step": 37632 + }, + { + "epoch": 2.78, + "learning_rate": 1.1661142823988863e-05, + "loss": 0.9542, + "step": 37633 + }, + { + "epoch": 2.78, + "learning_rate": 1.1660749492767169e-05, + "loss": 0.9705, + "step": 37634 + }, + { + "epoch": 2.78, + "learning_rate": 1.166035615890325e-05, + "loss": 1.0503, + "step": 37635 + }, + { + "epoch": 2.78, + "learning_rate": 1.1659962822397724e-05, + "loss": 1.0081, + "step": 37636 + }, + { + "epoch": 2.78, + "learning_rate": 1.1659569483251225e-05, + "loss": 0.9461, + "step": 37637 + }, + { + "epoch": 2.78, + "learning_rate": 1.165917614146437e-05, + "loss": 0.9369, + "step": 37638 + }, + { + "epoch": 2.78, + "learning_rate": 1.1658782797037793e-05, + "loss": 0.9817, + "step": 37639 + }, + { + "epoch": 2.78, + "learning_rate": 1.1658389449972115e-05, + "loss": 1.0254, + "step": 37640 + }, + { + "epoch": 2.78, + "learning_rate": 1.1657996100267964e-05, + "loss": 0.9333, + "step": 37641 + }, + { + "epoch": 2.78, + "learning_rate": 1.1657602747925963e-05, + "loss": 1.0204, + "step": 37642 + }, + { + "epoch": 2.78, + "learning_rate": 1.1657209392946742e-05, + "loss": 0.9377, + "step": 37643 + }, + { + "epoch": 2.78, + "learning_rate": 1.1656816035330925e-05, + "loss": 0.9413, + "step": 37644 + }, + { + "epoch": 2.78, + "learning_rate": 1.1656422675079138e-05, + "loss": 1.0364, + "step": 37645 + }, + { + "epoch": 2.78, + "learning_rate": 1.1656029312192002e-05, + "loss": 1.0327, + "step": 37646 + }, + { + "epoch": 2.78, + "learning_rate": 1.165563594667015e-05, + "loss": 0.9108, + "step": 37647 + }, + { + "epoch": 2.78, + "learning_rate": 1.1655242578514204e-05, + "loss": 1.0339, + "step": 37648 + }, + { + "epoch": 2.78, + "learning_rate": 1.165484920772479e-05, + "loss": 0.8971, + "step": 37649 + }, + { + "epoch": 2.78, + "learning_rate": 1.1654455834302535e-05, + "loss": 1.0386, + "step": 37650 + }, + { + "epoch": 2.78, + "learning_rate": 1.1654062458248065e-05, + "loss": 0.9426, + "step": 37651 + }, + { + "epoch": 2.78, + "learning_rate": 1.1653669079562005e-05, + "loss": 1.0985, + "step": 37652 + }, + { + "epoch": 2.78, + "learning_rate": 1.1653275698244984e-05, + "loss": 1.007, + "step": 37653 + }, + { + "epoch": 2.78, + "learning_rate": 1.1652882314297622e-05, + "loss": 1.027, + "step": 37654 + }, + { + "epoch": 2.78, + "learning_rate": 1.1652488927720548e-05, + "loss": 0.9209, + "step": 37655 + }, + { + "epoch": 2.78, + "learning_rate": 1.1652095538514387e-05, + "loss": 1.0532, + "step": 37656 + }, + { + "epoch": 2.78, + "learning_rate": 1.1651702146679768e-05, + "loss": 1.0021, + "step": 37657 + }, + { + "epoch": 2.78, + "learning_rate": 1.1651308752217312e-05, + "loss": 1.0367, + "step": 37658 + }, + { + "epoch": 2.78, + "learning_rate": 1.165091535512765e-05, + "loss": 0.8954, + "step": 37659 + }, + { + "epoch": 2.78, + "learning_rate": 1.1650521955411403e-05, + "loss": 0.9985, + "step": 37660 + }, + { + "epoch": 2.78, + "learning_rate": 1.16501285530692e-05, + "loss": 1.0021, + "step": 37661 + }, + { + "epoch": 2.78, + "learning_rate": 1.1649735148101666e-05, + "loss": 1.0539, + "step": 37662 + }, + { + "epoch": 2.78, + "learning_rate": 1.164934174050943e-05, + "loss": 1.0295, + "step": 37663 + }, + { + "epoch": 2.78, + "learning_rate": 1.164894833029311e-05, + "loss": 1.0875, + "step": 37664 + }, + { + "epoch": 2.78, + "learning_rate": 1.164855491745334e-05, + "loss": 1.0195, + "step": 37665 + }, + { + "epoch": 2.78, + "learning_rate": 1.1648161501990742e-05, + "loss": 1.0619, + "step": 37666 + }, + { + "epoch": 2.78, + "learning_rate": 1.1647768083905943e-05, + "loss": 1.0129, + "step": 37667 + }, + { + "epoch": 2.78, + "learning_rate": 1.1647374663199566e-05, + "loss": 0.974, + "step": 37668 + }, + { + "epoch": 2.78, + "learning_rate": 1.1646981239872241e-05, + "loss": 1.0099, + "step": 37669 + }, + { + "epoch": 2.78, + "learning_rate": 1.1646587813924592e-05, + "loss": 1.0412, + "step": 37670 + }, + { + "epoch": 2.78, + "learning_rate": 1.1646194385357249e-05, + "loss": 1.0139, + "step": 37671 + }, + { + "epoch": 2.78, + "learning_rate": 1.1645800954170832e-05, + "loss": 1.0411, + "step": 37672 + }, + { + "epoch": 2.78, + "learning_rate": 1.1645407520365966e-05, + "loss": 1.049, + "step": 37673 + }, + { + "epoch": 2.78, + "learning_rate": 1.1645014083943285e-05, + "loss": 0.9944, + "step": 37674 + }, + { + "epoch": 2.78, + "learning_rate": 1.1644620644903407e-05, + "loss": 1.0583, + "step": 37675 + }, + { + "epoch": 2.78, + "learning_rate": 1.164422720324696e-05, + "loss": 0.9294, + "step": 37676 + }, + { + "epoch": 2.78, + "learning_rate": 1.1643833758974572e-05, + "loss": 1.0227, + "step": 37677 + }, + { + "epoch": 2.78, + "learning_rate": 1.164344031208687e-05, + "loss": 1.1156, + "step": 37678 + }, + { + "epoch": 2.78, + "learning_rate": 1.164304686258448e-05, + "loss": 0.9139, + "step": 37679 + }, + { + "epoch": 2.78, + "learning_rate": 1.1642653410468022e-05, + "loss": 1.1058, + "step": 37680 + }, + { + "epoch": 2.78, + "learning_rate": 1.1642259955738127e-05, + "loss": 0.9757, + "step": 37681 + }, + { + "epoch": 2.78, + "learning_rate": 1.164186649839542e-05, + "loss": 1.0697, + "step": 37682 + }, + { + "epoch": 2.78, + "learning_rate": 1.1641473038440525e-05, + "loss": 1.0157, + "step": 37683 + }, + { + "epoch": 2.78, + "learning_rate": 1.1641079575874072e-05, + "loss": 1.0813, + "step": 37684 + }, + { + "epoch": 2.78, + "learning_rate": 1.1640686110696685e-05, + "loss": 1.1412, + "step": 37685 + }, + { + "epoch": 2.78, + "learning_rate": 1.1640292642908989e-05, + "loss": 1.0353, + "step": 37686 + }, + { + "epoch": 2.78, + "learning_rate": 1.1639899172511611e-05, + "loss": 1.0356, + "step": 37687 + }, + { + "epoch": 2.78, + "learning_rate": 1.1639505699505178e-05, + "loss": 0.9667, + "step": 37688 + }, + { + "epoch": 2.78, + "learning_rate": 1.1639112223890311e-05, + "loss": 1.0618, + "step": 37689 + }, + { + "epoch": 2.78, + "learning_rate": 1.1638718745667646e-05, + "loss": 0.9843, + "step": 37690 + }, + { + "epoch": 2.78, + "learning_rate": 1.16383252648378e-05, + "loss": 1.0282, + "step": 37691 + }, + { + "epoch": 2.79, + "learning_rate": 1.1637931781401401e-05, + "loss": 0.9108, + "step": 37692 + }, + { + "epoch": 2.79, + "learning_rate": 1.1637538295359076e-05, + "loss": 1.0456, + "step": 37693 + }, + { + "epoch": 2.79, + "learning_rate": 1.1637144806711453e-05, + "loss": 1.1185, + "step": 37694 + }, + { + "epoch": 2.79, + "learning_rate": 1.1636751315459153e-05, + "loss": 1.0276, + "step": 37695 + }, + { + "epoch": 2.79, + "learning_rate": 1.1636357821602809e-05, + "loss": 1.0321, + "step": 37696 + }, + { + "epoch": 2.79, + "learning_rate": 1.163596432514304e-05, + "loss": 0.9686, + "step": 37697 + }, + { + "epoch": 2.79, + "learning_rate": 1.1635570826080475e-05, + "loss": 0.9479, + "step": 37698 + }, + { + "epoch": 2.79, + "learning_rate": 1.163517732441574e-05, + "loss": 1.0558, + "step": 37699 + }, + { + "epoch": 2.79, + "learning_rate": 1.1634783820149466e-05, + "loss": 1.0131, + "step": 37700 + }, + { + "epoch": 2.79, + "learning_rate": 1.1634390313282268e-05, + "loss": 1.0864, + "step": 37701 + }, + { + "epoch": 2.79, + "learning_rate": 1.1633996803814785e-05, + "loss": 1.0524, + "step": 37702 + }, + { + "epoch": 2.79, + "learning_rate": 1.163360329174763e-05, + "loss": 1.0763, + "step": 37703 + }, + { + "epoch": 2.79, + "learning_rate": 1.163320977708144e-05, + "loss": 0.9345, + "step": 37704 + }, + { + "epoch": 2.79, + "learning_rate": 1.1632816259816833e-05, + "loss": 0.997, + "step": 37705 + }, + { + "epoch": 2.79, + "learning_rate": 1.1632422739954441e-05, + "loss": 0.9869, + "step": 37706 + }, + { + "epoch": 2.79, + "learning_rate": 1.1632029217494887e-05, + "loss": 1.0088, + "step": 37707 + }, + { + "epoch": 2.79, + "learning_rate": 1.1631635692438799e-05, + "loss": 0.8821, + "step": 37708 + }, + { + "epoch": 2.79, + "learning_rate": 1.1631242164786802e-05, + "loss": 1.0478, + "step": 37709 + }, + { + "epoch": 2.79, + "learning_rate": 1.1630848634539522e-05, + "loss": 1.0409, + "step": 37710 + }, + { + "epoch": 2.79, + "learning_rate": 1.1630455101697585e-05, + "loss": 0.9887, + "step": 37711 + }, + { + "epoch": 2.79, + "learning_rate": 1.1630061566261618e-05, + "loss": 0.9578, + "step": 37712 + }, + { + "epoch": 2.79, + "learning_rate": 1.1629668028232242e-05, + "loss": 1.0019, + "step": 37713 + }, + { + "epoch": 2.79, + "learning_rate": 1.162927448761009e-05, + "loss": 1.0491, + "step": 37714 + }, + { + "epoch": 2.79, + "learning_rate": 1.1628880944395786e-05, + "loss": 0.9144, + "step": 37715 + }, + { + "epoch": 2.79, + "learning_rate": 1.1628487398589956e-05, + "loss": 0.9871, + "step": 37716 + }, + { + "epoch": 2.79, + "learning_rate": 1.1628093850193224e-05, + "loss": 1.0729, + "step": 37717 + }, + { + "epoch": 2.79, + "learning_rate": 1.1627700299206222e-05, + "loss": 1.0414, + "step": 37718 + }, + { + "epoch": 2.79, + "learning_rate": 1.1627306745629569e-05, + "loss": 0.8764, + "step": 37719 + }, + { + "epoch": 2.79, + "learning_rate": 1.1626913189463895e-05, + "loss": 0.9882, + "step": 37720 + }, + { + "epoch": 2.79, + "learning_rate": 1.1626519630709828e-05, + "loss": 1.0122, + "step": 37721 + }, + { + "epoch": 2.79, + "learning_rate": 1.1626126069367988e-05, + "loss": 1.0384, + "step": 37722 + }, + { + "epoch": 2.79, + "learning_rate": 1.1625732505439005e-05, + "loss": 1.0676, + "step": 37723 + }, + { + "epoch": 2.79, + "learning_rate": 1.1625338938923507e-05, + "loss": 0.9891, + "step": 37724 + }, + { + "epoch": 2.79, + "learning_rate": 1.1624945369822114e-05, + "loss": 1.0659, + "step": 37725 + }, + { + "epoch": 2.79, + "learning_rate": 1.162455179813546e-05, + "loss": 1.0185, + "step": 37726 + }, + { + "epoch": 2.79, + "learning_rate": 1.1624158223864165e-05, + "loss": 1.1566, + "step": 37727 + }, + { + "epoch": 2.79, + "learning_rate": 1.162376464700886e-05, + "loss": 0.9043, + "step": 37728 + }, + { + "epoch": 2.79, + "learning_rate": 1.1623371067570168e-05, + "loss": 0.9138, + "step": 37729 + }, + { + "epoch": 2.79, + "learning_rate": 1.1622977485548712e-05, + "loss": 0.944, + "step": 37730 + }, + { + "epoch": 2.79, + "learning_rate": 1.1622583900945129e-05, + "loss": 1.0944, + "step": 37731 + }, + { + "epoch": 2.79, + "learning_rate": 1.1622190313760031e-05, + "loss": 0.997, + "step": 37732 + }, + { + "epoch": 2.79, + "learning_rate": 1.1621796723994056e-05, + "loss": 1.0061, + "step": 37733 + }, + { + "epoch": 2.79, + "learning_rate": 1.1621403131647824e-05, + "loss": 1.0903, + "step": 37734 + }, + { + "epoch": 2.79, + "learning_rate": 1.1621009536721964e-05, + "loss": 1.052, + "step": 37735 + }, + { + "epoch": 2.79, + "learning_rate": 1.1620615939217098e-05, + "loss": 1.0196, + "step": 37736 + }, + { + "epoch": 2.79, + "learning_rate": 1.162022233913386e-05, + "loss": 1.0981, + "step": 37737 + }, + { + "epoch": 2.79, + "learning_rate": 1.1619828736472867e-05, + "loss": 0.9998, + "step": 37738 + }, + { + "epoch": 2.79, + "learning_rate": 1.1619435131234753e-05, + "loss": 0.982, + "step": 37739 + }, + { + "epoch": 2.79, + "learning_rate": 1.1619041523420138e-05, + "loss": 1.0194, + "step": 37740 + }, + { + "epoch": 2.79, + "learning_rate": 1.1618647913029654e-05, + "loss": 1.0549, + "step": 37741 + }, + { + "epoch": 2.79, + "learning_rate": 1.1618254300063923e-05, + "loss": 1.0211, + "step": 37742 + }, + { + "epoch": 2.79, + "learning_rate": 1.1617860684523573e-05, + "loss": 1.0163, + "step": 37743 + }, + { + "epoch": 2.79, + "learning_rate": 1.1617467066409228e-05, + "loss": 0.9531, + "step": 37744 + }, + { + "epoch": 2.79, + "learning_rate": 1.1617073445721516e-05, + "loss": 1.0848, + "step": 37745 + }, + { + "epoch": 2.79, + "learning_rate": 1.1616679822461063e-05, + "loss": 1.0025, + "step": 37746 + }, + { + "epoch": 2.79, + "learning_rate": 1.1616286196628501e-05, + "loss": 0.9199, + "step": 37747 + }, + { + "epoch": 2.79, + "learning_rate": 1.1615892568224444e-05, + "loss": 0.9289, + "step": 37748 + }, + { + "epoch": 2.79, + "learning_rate": 1.1615498937249529e-05, + "loss": 1.0699, + "step": 37749 + }, + { + "epoch": 2.79, + "learning_rate": 1.1615105303704377e-05, + "loss": 0.9988, + "step": 37750 + }, + { + "epoch": 2.79, + "learning_rate": 1.1614711667589618e-05, + "loss": 0.9926, + "step": 37751 + }, + { + "epoch": 2.79, + "learning_rate": 1.1614318028905871e-05, + "loss": 1.0371, + "step": 37752 + }, + { + "epoch": 2.79, + "learning_rate": 1.161392438765377e-05, + "loss": 1.0672, + "step": 37753 + }, + { + "epoch": 2.79, + "learning_rate": 1.1613530743833938e-05, + "loss": 1.0956, + "step": 37754 + }, + { + "epoch": 2.79, + "learning_rate": 1.1613137097447002e-05, + "loss": 1.0137, + "step": 37755 + }, + { + "epoch": 2.79, + "learning_rate": 1.1612743448493589e-05, + "loss": 0.9746, + "step": 37756 + }, + { + "epoch": 2.79, + "learning_rate": 1.1612349796974325e-05, + "loss": 1.0582, + "step": 37757 + }, + { + "epoch": 2.79, + "learning_rate": 1.1611956142889831e-05, + "loss": 1.1375, + "step": 37758 + }, + { + "epoch": 2.79, + "learning_rate": 1.1611562486240742e-05, + "loss": 1.1509, + "step": 37759 + }, + { + "epoch": 2.79, + "learning_rate": 1.1611168827027676e-05, + "loss": 1.0794, + "step": 37760 + }, + { + "epoch": 2.79, + "learning_rate": 1.1610775165251267e-05, + "loss": 1.0597, + "step": 37761 + }, + { + "epoch": 2.79, + "learning_rate": 1.1610381500912135e-05, + "loss": 1.0529, + "step": 37762 + }, + { + "epoch": 2.79, + "learning_rate": 1.1609987834010912e-05, + "loss": 0.985, + "step": 37763 + }, + { + "epoch": 2.79, + "learning_rate": 1.1609594164548219e-05, + "loss": 0.9076, + "step": 37764 + }, + { + "epoch": 2.79, + "learning_rate": 1.1609200492524689e-05, + "loss": 1.0078, + "step": 37765 + }, + { + "epoch": 2.79, + "learning_rate": 1.1608806817940941e-05, + "loss": 1.0176, + "step": 37766 + }, + { + "epoch": 2.79, + "learning_rate": 1.1608413140797606e-05, + "loss": 0.9711, + "step": 37767 + }, + { + "epoch": 2.79, + "learning_rate": 1.1608019461095304e-05, + "loss": 1.0657, + "step": 37768 + }, + { + "epoch": 2.79, + "learning_rate": 1.160762577883467e-05, + "loss": 1.1074, + "step": 37769 + }, + { + "epoch": 2.79, + "learning_rate": 1.1607232094016326e-05, + "loss": 0.8656, + "step": 37770 + }, + { + "epoch": 2.79, + "learning_rate": 1.1606838406640897e-05, + "loss": 0.984, + "step": 37771 + }, + { + "epoch": 2.79, + "learning_rate": 1.1606444716709012e-05, + "loss": 1.0451, + "step": 37772 + }, + { + "epoch": 2.79, + "learning_rate": 1.1606051024221297e-05, + "loss": 1.0744, + "step": 37773 + }, + { + "epoch": 2.79, + "learning_rate": 1.1605657329178378e-05, + "loss": 0.9399, + "step": 37774 + }, + { + "epoch": 2.79, + "learning_rate": 1.1605263631580879e-05, + "loss": 0.9753, + "step": 37775 + }, + { + "epoch": 2.79, + "learning_rate": 1.1604869931429433e-05, + "loss": 0.9865, + "step": 37776 + }, + { + "epoch": 2.79, + "learning_rate": 1.1604476228724657e-05, + "loss": 1.0943, + "step": 37777 + }, + { + "epoch": 2.79, + "learning_rate": 1.1604082523467183e-05, + "loss": 0.9404, + "step": 37778 + }, + { + "epoch": 2.79, + "learning_rate": 1.1603688815657638e-05, + "loss": 1.0601, + "step": 37779 + }, + { + "epoch": 2.79, + "learning_rate": 1.1603295105296649e-05, + "loss": 0.9858, + "step": 37780 + }, + { + "epoch": 2.79, + "learning_rate": 1.1602901392384836e-05, + "loss": 1.0376, + "step": 37781 + }, + { + "epoch": 2.79, + "learning_rate": 1.1602507676922835e-05, + "loss": 1.0432, + "step": 37782 + }, + { + "epoch": 2.79, + "learning_rate": 1.1602113958911265e-05, + "loss": 1.0599, + "step": 37783 + }, + { + "epoch": 2.79, + "learning_rate": 1.1601720238350754e-05, + "loss": 0.9974, + "step": 37784 + }, + { + "epoch": 2.79, + "learning_rate": 1.1601326515241929e-05, + "loss": 1.0581, + "step": 37785 + }, + { + "epoch": 2.79, + "learning_rate": 1.1600932789585415e-05, + "loss": 0.9544, + "step": 37786 + }, + { + "epoch": 2.79, + "learning_rate": 1.1600539061381842e-05, + "loss": 0.9637, + "step": 37787 + }, + { + "epoch": 2.79, + "learning_rate": 1.1600145330631834e-05, + "loss": 1.0595, + "step": 37788 + }, + { + "epoch": 2.79, + "learning_rate": 1.1599751597336015e-05, + "loss": 0.9881, + "step": 37789 + }, + { + "epoch": 2.79, + "learning_rate": 1.1599357861495017e-05, + "loss": 1.0528, + "step": 37790 + }, + { + "epoch": 2.79, + "learning_rate": 1.1598964123109461e-05, + "loss": 1.0365, + "step": 37791 + }, + { + "epoch": 2.79, + "learning_rate": 1.159857038217998e-05, + "loss": 0.9845, + "step": 37792 + }, + { + "epoch": 2.79, + "learning_rate": 1.1598176638707197e-05, + "loss": 0.9986, + "step": 37793 + }, + { + "epoch": 2.79, + "learning_rate": 1.1597782892691735e-05, + "loss": 1.0474, + "step": 37794 + }, + { + "epoch": 2.79, + "learning_rate": 1.1597389144134223e-05, + "loss": 0.9309, + "step": 37795 + }, + { + "epoch": 2.79, + "learning_rate": 1.1596995393035289e-05, + "loss": 1.029, + "step": 37796 + }, + { + "epoch": 2.79, + "learning_rate": 1.1596601639395556e-05, + "loss": 1.0429, + "step": 37797 + }, + { + "epoch": 2.79, + "learning_rate": 1.1596207883215654e-05, + "loss": 1.0548, + "step": 37798 + }, + { + "epoch": 2.79, + "learning_rate": 1.1595814124496208e-05, + "loss": 1.0292, + "step": 37799 + }, + { + "epoch": 2.79, + "learning_rate": 1.1595420363237844e-05, + "loss": 1.0928, + "step": 37800 + }, + { + "epoch": 2.79, + "learning_rate": 1.1595026599441188e-05, + "loss": 1.0008, + "step": 37801 + }, + { + "epoch": 2.79, + "learning_rate": 1.1594632833106872e-05, + "loss": 1.0642, + "step": 37802 + }, + { + "epoch": 2.79, + "learning_rate": 1.1594239064235516e-05, + "loss": 0.955, + "step": 37803 + }, + { + "epoch": 2.79, + "learning_rate": 1.1593845292827747e-05, + "loss": 1.0353, + "step": 37804 + }, + { + "epoch": 2.79, + "learning_rate": 1.1593451518884194e-05, + "loss": 1.0656, + "step": 37805 + }, + { + "epoch": 2.79, + "learning_rate": 1.1593057742405483e-05, + "loss": 0.966, + "step": 37806 + }, + { + "epoch": 2.79, + "learning_rate": 1.1592663963392237e-05, + "loss": 1.0318, + "step": 37807 + }, + { + "epoch": 2.79, + "learning_rate": 1.1592270181845087e-05, + "loss": 0.922, + "step": 37808 + }, + { + "epoch": 2.79, + "learning_rate": 1.1591876397764657e-05, + "loss": 0.9928, + "step": 37809 + }, + { + "epoch": 2.79, + "learning_rate": 1.1591482611151576e-05, + "loss": 1.0737, + "step": 37810 + }, + { + "epoch": 2.79, + "learning_rate": 1.1591088822006468e-05, + "loss": 0.9869, + "step": 37811 + }, + { + "epoch": 2.79, + "learning_rate": 1.1590695030329963e-05, + "loss": 1.0593, + "step": 37812 + }, + { + "epoch": 2.79, + "learning_rate": 1.1590301236122683e-05, + "loss": 0.8465, + "step": 37813 + }, + { + "epoch": 2.79, + "learning_rate": 1.1589907439385258e-05, + "loss": 0.9798, + "step": 37814 + }, + { + "epoch": 2.79, + "learning_rate": 1.1589513640118311e-05, + "loss": 1.0163, + "step": 37815 + }, + { + "epoch": 2.79, + "learning_rate": 1.1589119838322471e-05, + "loss": 1.0312, + "step": 37816 + }, + { + "epoch": 2.79, + "learning_rate": 1.1588726033998363e-05, + "loss": 0.9182, + "step": 37817 + }, + { + "epoch": 2.79, + "learning_rate": 1.1588332227146614e-05, + "loss": 0.9661, + "step": 37818 + }, + { + "epoch": 2.79, + "learning_rate": 1.1587938417767854e-05, + "loss": 1.0964, + "step": 37819 + }, + { + "epoch": 2.79, + "learning_rate": 1.1587544605862707e-05, + "loss": 0.8962, + "step": 37820 + }, + { + "epoch": 2.79, + "learning_rate": 1.1587150791431798e-05, + "loss": 1.1196, + "step": 37821 + }, + { + "epoch": 2.79, + "learning_rate": 1.1586756974475753e-05, + "loss": 1.0471, + "step": 37822 + }, + { + "epoch": 2.79, + "learning_rate": 1.1586363154995203e-05, + "loss": 1.0917, + "step": 37823 + }, + { + "epoch": 2.79, + "learning_rate": 1.158596933299077e-05, + "loss": 0.93, + "step": 37824 + }, + { + "epoch": 2.79, + "learning_rate": 1.1585575508463083e-05, + "loss": 0.8875, + "step": 37825 + }, + { + "epoch": 2.79, + "learning_rate": 1.1585181681412766e-05, + "loss": 1.0509, + "step": 37826 + }, + { + "epoch": 2.8, + "learning_rate": 1.1584787851840453e-05, + "loss": 1.0254, + "step": 37827 + }, + { + "epoch": 2.8, + "learning_rate": 1.1584394019746763e-05, + "loss": 0.9797, + "step": 37828 + }, + { + "epoch": 2.8, + "learning_rate": 1.1584000185132323e-05, + "loss": 0.9983, + "step": 37829 + }, + { + "epoch": 2.8, + "learning_rate": 1.1583606347997762e-05, + "loss": 0.9169, + "step": 37830 + }, + { + "epoch": 2.8, + "learning_rate": 1.1583212508343709e-05, + "loss": 0.9287, + "step": 37831 + }, + { + "epoch": 2.8, + "learning_rate": 1.1582818666170784e-05, + "loss": 0.9958, + "step": 37832 + }, + { + "epoch": 2.8, + "learning_rate": 1.1582424821479618e-05, + "loss": 0.9668, + "step": 37833 + }, + { + "epoch": 2.8, + "learning_rate": 1.1582030974270836e-05, + "loss": 1.0145, + "step": 37834 + }, + { + "epoch": 2.8, + "learning_rate": 1.1581637124545067e-05, + "loss": 1.0275, + "step": 37835 + }, + { + "epoch": 2.8, + "learning_rate": 1.1581243272302933e-05, + "loss": 0.9361, + "step": 37836 + }, + { + "epoch": 2.8, + "learning_rate": 1.1580849417545068e-05, + "loss": 1.0707, + "step": 37837 + }, + { + "epoch": 2.8, + "learning_rate": 1.158045556027209e-05, + "loss": 0.9997, + "step": 37838 + }, + { + "epoch": 2.8, + "learning_rate": 1.1580061700484632e-05, + "loss": 1.0621, + "step": 37839 + }, + { + "epoch": 2.8, + "learning_rate": 1.157966783818332e-05, + "loss": 1.0429, + "step": 37840 + }, + { + "epoch": 2.8, + "learning_rate": 1.1579273973368776e-05, + "loss": 0.9412, + "step": 37841 + }, + { + "epoch": 2.8, + "learning_rate": 1.1578880106041632e-05, + "loss": 1.0181, + "step": 37842 + }, + { + "epoch": 2.8, + "learning_rate": 1.1578486236202511e-05, + "loss": 1.125, + "step": 37843 + }, + { + "epoch": 2.8, + "learning_rate": 1.157809236385204e-05, + "loss": 1.0308, + "step": 37844 + }, + { + "epoch": 2.8, + "learning_rate": 1.1577698488990849e-05, + "loss": 0.9777, + "step": 37845 + }, + { + "epoch": 2.8, + "learning_rate": 1.157730461161956e-05, + "loss": 1.0417, + "step": 37846 + }, + { + "epoch": 2.8, + "learning_rate": 1.1576910731738803e-05, + "loss": 1.0795, + "step": 37847 + }, + { + "epoch": 2.8, + "learning_rate": 1.1576516849349203e-05, + "loss": 0.8778, + "step": 37848 + }, + { + "epoch": 2.8, + "learning_rate": 1.1576122964451391e-05, + "loss": 0.9246, + "step": 37849 + }, + { + "epoch": 2.8, + "learning_rate": 1.1575729077045987e-05, + "loss": 0.9371, + "step": 37850 + }, + { + "epoch": 2.8, + "learning_rate": 1.157533518713362e-05, + "loss": 1.0075, + "step": 37851 + }, + { + "epoch": 2.8, + "learning_rate": 1.1574941294714916e-05, + "loss": 1.0032, + "step": 37852 + }, + { + "epoch": 2.8, + "learning_rate": 1.1574547399790506e-05, + "loss": 1.0346, + "step": 37853 + }, + { + "epoch": 2.8, + "learning_rate": 1.1574153502361012e-05, + "loss": 0.9595, + "step": 37854 + }, + { + "epoch": 2.8, + "learning_rate": 1.1573759602427065e-05, + "loss": 1.0573, + "step": 37855 + }, + { + "epoch": 2.8, + "learning_rate": 1.1573365699989282e-05, + "loss": 1.0177, + "step": 37856 + }, + { + "epoch": 2.8, + "learning_rate": 1.1572971795048304e-05, + "loss": 1.0143, + "step": 37857 + }, + { + "epoch": 2.8, + "learning_rate": 1.1572577887604745e-05, + "loss": 0.9988, + "step": 37858 + }, + { + "epoch": 2.8, + "learning_rate": 1.1572183977659244e-05, + "loss": 0.936, + "step": 37859 + }, + { + "epoch": 2.8, + "learning_rate": 1.1571790065212414e-05, + "loss": 0.9507, + "step": 37860 + }, + { + "epoch": 2.8, + "learning_rate": 1.1571396150264893e-05, + "loss": 1.0273, + "step": 37861 + }, + { + "epoch": 2.8, + "learning_rate": 1.1571002232817299e-05, + "loss": 1.0274, + "step": 37862 + }, + { + "epoch": 2.8, + "learning_rate": 1.1570608312870267e-05, + "loss": 0.9702, + "step": 37863 + }, + { + "epoch": 2.8, + "learning_rate": 1.1570214390424414e-05, + "loss": 0.8966, + "step": 37864 + }, + { + "epoch": 2.8, + "learning_rate": 1.1569820465480378e-05, + "loss": 0.9942, + "step": 37865 + }, + { + "epoch": 2.8, + "learning_rate": 1.1569426538038776e-05, + "loss": 1.1017, + "step": 37866 + }, + { + "epoch": 2.8, + "learning_rate": 1.1569032608100243e-05, + "loss": 1.0703, + "step": 37867 + }, + { + "epoch": 2.8, + "learning_rate": 1.15686386756654e-05, + "loss": 0.9212, + "step": 37868 + }, + { + "epoch": 2.8, + "learning_rate": 1.1568244740734875e-05, + "loss": 0.9966, + "step": 37869 + }, + { + "epoch": 2.8, + "learning_rate": 1.1567850803309294e-05, + "loss": 0.9672, + "step": 37870 + }, + { + "epoch": 2.8, + "learning_rate": 1.1567456863389286e-05, + "loss": 1.0178, + "step": 37871 + }, + { + "epoch": 2.8, + "learning_rate": 1.1567062920975479e-05, + "loss": 1.1434, + "step": 37872 + }, + { + "epoch": 2.8, + "learning_rate": 1.1566668976068494e-05, + "loss": 1.0734, + "step": 37873 + }, + { + "epoch": 2.8, + "learning_rate": 1.1566275028668961e-05, + "loss": 0.9976, + "step": 37874 + }, + { + "epoch": 2.8, + "learning_rate": 1.1565881078777508e-05, + "loss": 1.1262, + "step": 37875 + }, + { + "epoch": 2.8, + "learning_rate": 1.1565487126394759e-05, + "loss": 0.9793, + "step": 37876 + }, + { + "epoch": 2.8, + "learning_rate": 1.1565093171521344e-05, + "loss": 0.9728, + "step": 37877 + }, + { + "epoch": 2.8, + "learning_rate": 1.1564699214157888e-05, + "loss": 0.9806, + "step": 37878 + }, + { + "epoch": 2.8, + "learning_rate": 1.1564305254305018e-05, + "loss": 1.0795, + "step": 37879 + }, + { + "epoch": 2.8, + "learning_rate": 1.1563911291963361e-05, + "loss": 1.0388, + "step": 37880 + }, + { + "epoch": 2.8, + "learning_rate": 1.156351732713354e-05, + "loss": 0.9813, + "step": 37881 + }, + { + "epoch": 2.8, + "learning_rate": 1.156312335981619e-05, + "loss": 1.0478, + "step": 37882 + }, + { + "epoch": 2.8, + "learning_rate": 1.156272939001193e-05, + "loss": 1.0231, + "step": 37883 + }, + { + "epoch": 2.8, + "learning_rate": 1.1562335417721391e-05, + "loss": 0.9478, + "step": 37884 + }, + { + "epoch": 2.8, + "learning_rate": 1.15619414429452e-05, + "loss": 1.1813, + "step": 37885 + }, + { + "epoch": 2.8, + "learning_rate": 1.1561547465683982e-05, + "loss": 0.9508, + "step": 37886 + }, + { + "epoch": 2.8, + "learning_rate": 1.1561153485938364e-05, + "loss": 0.9867, + "step": 37887 + }, + { + "epoch": 2.8, + "learning_rate": 1.1560759503708973e-05, + "loss": 1.0065, + "step": 37888 + }, + { + "epoch": 2.8, + "learning_rate": 1.1560365518996434e-05, + "loss": 1.0553, + "step": 37889 + }, + { + "epoch": 2.8, + "learning_rate": 1.155997153180138e-05, + "loss": 1.033, + "step": 37890 + }, + { + "epoch": 2.8, + "learning_rate": 1.155957754212443e-05, + "loss": 1.0037, + "step": 37891 + }, + { + "epoch": 2.8, + "learning_rate": 1.1559183549966215e-05, + "loss": 0.9964, + "step": 37892 + }, + { + "epoch": 2.8, + "learning_rate": 1.1558789555327362e-05, + "loss": 0.9388, + "step": 37893 + }, + { + "epoch": 2.8, + "learning_rate": 1.1558395558208498e-05, + "loss": 1.0362, + "step": 37894 + }, + { + "epoch": 2.8, + "learning_rate": 1.1558001558610246e-05, + "loss": 1.0315, + "step": 37895 + }, + { + "epoch": 2.8, + "learning_rate": 1.1557607556533242e-05, + "loss": 0.9674, + "step": 37896 + }, + { + "epoch": 2.8, + "learning_rate": 1.1557213551978101e-05, + "loss": 1.1477, + "step": 37897 + }, + { + "epoch": 2.8, + "learning_rate": 1.1556819544945456e-05, + "loss": 1.0023, + "step": 37898 + }, + { + "epoch": 2.8, + "learning_rate": 1.1556425535435935e-05, + "loss": 1.004, + "step": 37899 + }, + { + "epoch": 2.8, + "learning_rate": 1.1556031523450163e-05, + "loss": 0.9588, + "step": 37900 + }, + { + "epoch": 2.8, + "learning_rate": 1.1555637508988767e-05, + "loss": 1.0354, + "step": 37901 + }, + { + "epoch": 2.8, + "learning_rate": 1.1555243492052373e-05, + "loss": 1.1217, + "step": 37902 + }, + { + "epoch": 2.8, + "learning_rate": 1.1554849472641609e-05, + "loss": 1.0178, + "step": 37903 + }, + { + "epoch": 2.8, + "learning_rate": 1.1554455450757104e-05, + "loss": 0.9619, + "step": 37904 + }, + { + "epoch": 2.8, + "learning_rate": 1.1554061426399479e-05, + "loss": 1.1234, + "step": 37905 + }, + { + "epoch": 2.8, + "learning_rate": 1.1553667399569369e-05, + "loss": 1.0635, + "step": 37906 + }, + { + "epoch": 2.8, + "learning_rate": 1.1553273370267393e-05, + "loss": 1.0072, + "step": 37907 + }, + { + "epoch": 2.8, + "learning_rate": 1.1552879338494183e-05, + "loss": 1.0208, + "step": 37908 + }, + { + "epoch": 2.8, + "learning_rate": 1.1552485304250361e-05, + "loss": 1.0613, + "step": 37909 + }, + { + "epoch": 2.8, + "learning_rate": 1.155209126753656e-05, + "loss": 0.9583, + "step": 37910 + }, + { + "epoch": 2.8, + "learning_rate": 1.1551697228353403e-05, + "loss": 0.9439, + "step": 37911 + }, + { + "epoch": 2.8, + "learning_rate": 1.1551303186701519e-05, + "loss": 0.985, + "step": 37912 + }, + { + "epoch": 2.8, + "learning_rate": 1.1550909142581531e-05, + "loss": 1.0944, + "step": 37913 + }, + { + "epoch": 2.8, + "learning_rate": 1.1550515095994073e-05, + "loss": 0.9525, + "step": 37914 + }, + { + "epoch": 2.8, + "learning_rate": 1.1550121046939766e-05, + "loss": 1.0298, + "step": 37915 + }, + { + "epoch": 2.8, + "learning_rate": 1.1549726995419238e-05, + "loss": 0.9911, + "step": 37916 + }, + { + "epoch": 2.8, + "learning_rate": 1.1549332941433116e-05, + "loss": 0.9692, + "step": 37917 + }, + { + "epoch": 2.8, + "learning_rate": 1.154893888498203e-05, + "loss": 0.9439, + "step": 37918 + }, + { + "epoch": 2.8, + "learning_rate": 1.1548544826066602e-05, + "loss": 0.9245, + "step": 37919 + }, + { + "epoch": 2.8, + "learning_rate": 1.1548150764687462e-05, + "loss": 0.9708, + "step": 37920 + }, + { + "epoch": 2.8, + "learning_rate": 1.1547756700845235e-05, + "loss": 1.0511, + "step": 37921 + }, + { + "epoch": 2.8, + "learning_rate": 1.1547362634540551e-05, + "loss": 0.9829, + "step": 37922 + }, + { + "epoch": 2.8, + "learning_rate": 1.1546968565774033e-05, + "loss": 0.9951, + "step": 37923 + }, + { + "epoch": 2.8, + "learning_rate": 1.1546574494546311e-05, + "loss": 0.9946, + "step": 37924 + }, + { + "epoch": 2.8, + "learning_rate": 1.1546180420858015e-05, + "loss": 0.9489, + "step": 37925 + }, + { + "epoch": 2.8, + "learning_rate": 1.1545786344709764e-05, + "loss": 1.0922, + "step": 37926 + }, + { + "epoch": 2.8, + "learning_rate": 1.154539226610219e-05, + "loss": 0.9758, + "step": 37927 + }, + { + "epoch": 2.8, + "learning_rate": 1.1544998185035918e-05, + "loss": 0.9979, + "step": 37928 + }, + { + "epoch": 2.8, + "learning_rate": 1.1544604101511578e-05, + "loss": 0.9868, + "step": 37929 + }, + { + "epoch": 2.8, + "learning_rate": 1.1544210015529793e-05, + "loss": 1.0746, + "step": 37930 + }, + { + "epoch": 2.8, + "learning_rate": 1.1543815927091193e-05, + "loss": 0.9036, + "step": 37931 + }, + { + "epoch": 2.8, + "learning_rate": 1.1543421836196402e-05, + "loss": 1.1333, + "step": 37932 + }, + { + "epoch": 2.8, + "learning_rate": 1.1543027742846052e-05, + "loss": 0.9834, + "step": 37933 + }, + { + "epoch": 2.8, + "learning_rate": 1.1542633647040766e-05, + "loss": 0.9711, + "step": 37934 + }, + { + "epoch": 2.8, + "learning_rate": 1.1542239548781174e-05, + "loss": 1.1416, + "step": 37935 + }, + { + "epoch": 2.8, + "learning_rate": 1.1541845448067896e-05, + "loss": 1.0295, + "step": 37936 + }, + { + "epoch": 2.8, + "learning_rate": 1.154145134490157e-05, + "loss": 1.1272, + "step": 37937 + }, + { + "epoch": 2.8, + "learning_rate": 1.1541057239282809e-05, + "loss": 1.0376, + "step": 37938 + }, + { + "epoch": 2.8, + "learning_rate": 1.1540663131212256e-05, + "loss": 1.0487, + "step": 37939 + }, + { + "epoch": 2.8, + "learning_rate": 1.1540269020690525e-05, + "loss": 1.0081, + "step": 37940 + }, + { + "epoch": 2.8, + "learning_rate": 1.153987490771825e-05, + "loss": 0.9556, + "step": 37941 + }, + { + "epoch": 2.8, + "learning_rate": 1.1539480792296054e-05, + "loss": 1.0639, + "step": 37942 + }, + { + "epoch": 2.8, + "learning_rate": 1.1539086674424571e-05, + "loss": 1.1334, + "step": 37943 + }, + { + "epoch": 2.8, + "learning_rate": 1.1538692554104419e-05, + "loss": 1.2395, + "step": 37944 + }, + { + "epoch": 2.8, + "learning_rate": 1.153829843133623e-05, + "loss": 1.0372, + "step": 37945 + }, + { + "epoch": 2.8, + "learning_rate": 1.1537904306120629e-05, + "loss": 0.993, + "step": 37946 + }, + { + "epoch": 2.8, + "learning_rate": 1.1537510178458246e-05, + "loss": 0.9884, + "step": 37947 + }, + { + "epoch": 2.8, + "learning_rate": 1.1537116048349703e-05, + "loss": 0.9411, + "step": 37948 + }, + { + "epoch": 2.8, + "learning_rate": 1.1536721915795634e-05, + "loss": 1.1019, + "step": 37949 + }, + { + "epoch": 2.8, + "learning_rate": 1.1536327780796662e-05, + "loss": 0.9638, + "step": 37950 + }, + { + "epoch": 2.8, + "learning_rate": 1.1535933643353414e-05, + "loss": 1.0279, + "step": 37951 + }, + { + "epoch": 2.8, + "learning_rate": 1.1535539503466518e-05, + "loss": 0.9631, + "step": 37952 + }, + { + "epoch": 2.8, + "learning_rate": 1.1535145361136602e-05, + "loss": 0.9594, + "step": 37953 + }, + { + "epoch": 2.8, + "learning_rate": 1.1534751216364288e-05, + "loss": 1.0047, + "step": 37954 + }, + { + "epoch": 2.8, + "learning_rate": 1.1534357069150211e-05, + "loss": 1.0047, + "step": 37955 + }, + { + "epoch": 2.8, + "learning_rate": 1.1533962919494989e-05, + "loss": 1.0375, + "step": 37956 + }, + { + "epoch": 2.8, + "learning_rate": 1.1533568767399256e-05, + "loss": 0.9616, + "step": 37957 + }, + { + "epoch": 2.8, + "learning_rate": 1.1533174612863637e-05, + "loss": 0.9228, + "step": 37958 + }, + { + "epoch": 2.8, + "learning_rate": 1.1532780455888761e-05, + "loss": 0.9388, + "step": 37959 + }, + { + "epoch": 2.8, + "learning_rate": 1.1532386296475252e-05, + "loss": 1.005, + "step": 37960 + }, + { + "epoch": 2.8, + "learning_rate": 1.1531992134623738e-05, + "loss": 1.1012, + "step": 37961 + }, + { + "epoch": 2.8, + "learning_rate": 1.1531597970334847e-05, + "loss": 1.011, + "step": 37962 + }, + { + "epoch": 2.81, + "learning_rate": 1.1531203803609206e-05, + "loss": 0.9103, + "step": 37963 + }, + { + "epoch": 2.81, + "learning_rate": 1.153080963444744e-05, + "loss": 1.0444, + "step": 37964 + }, + { + "epoch": 2.81, + "learning_rate": 1.1530415462850178e-05, + "loss": 0.9011, + "step": 37965 + }, + { + "epoch": 2.81, + "learning_rate": 1.1530021288818047e-05, + "loss": 0.8841, + "step": 37966 + }, + { + "epoch": 2.81, + "learning_rate": 1.1529627112351676e-05, + "loss": 1.0245, + "step": 37967 + }, + { + "epoch": 2.81, + "learning_rate": 1.1529232933451686e-05, + "loss": 0.9693, + "step": 37968 + }, + { + "epoch": 2.81, + "learning_rate": 1.1528838752118712e-05, + "loss": 1.1389, + "step": 37969 + }, + { + "epoch": 2.81, + "learning_rate": 1.1528444568353377e-05, + "loss": 1.025, + "step": 37970 + }, + { + "epoch": 2.81, + "learning_rate": 1.1528050382156306e-05, + "loss": 0.9519, + "step": 37971 + }, + { + "epoch": 2.81, + "learning_rate": 1.1527656193528132e-05, + "loss": 1.0139, + "step": 37972 + }, + { + "epoch": 2.81, + "learning_rate": 1.1527262002469477e-05, + "loss": 1.0944, + "step": 37973 + }, + { + "epoch": 2.81, + "learning_rate": 1.152686780898097e-05, + "loss": 1.0688, + "step": 37974 + }, + { + "epoch": 2.81, + "learning_rate": 1.1526473613063238e-05, + "loss": 0.9523, + "step": 37975 + }, + { + "epoch": 2.81, + "learning_rate": 1.1526079414716911e-05, + "loss": 1.0268, + "step": 37976 + }, + { + "epoch": 2.81, + "learning_rate": 1.1525685213942612e-05, + "loss": 1.0854, + "step": 37977 + }, + { + "epoch": 2.81, + "learning_rate": 1.1525291010740967e-05, + "loss": 1.0468, + "step": 37978 + }, + { + "epoch": 2.81, + "learning_rate": 1.1524896805112609e-05, + "loss": 1.0472, + "step": 37979 + }, + { + "epoch": 2.81, + "learning_rate": 1.1524502597058161e-05, + "loss": 1.0777, + "step": 37980 + }, + { + "epoch": 2.81, + "learning_rate": 1.152410838657825e-05, + "loss": 1.1116, + "step": 37981 + }, + { + "epoch": 2.81, + "learning_rate": 1.1523714173673505e-05, + "loss": 1.0655, + "step": 37982 + }, + { + "epoch": 2.81, + "learning_rate": 1.1523319958344553e-05, + "loss": 1.0711, + "step": 37983 + }, + { + "epoch": 2.81, + "learning_rate": 1.1522925740592022e-05, + "loss": 1.0611, + "step": 37984 + }, + { + "epoch": 2.81, + "learning_rate": 1.1522531520416536e-05, + "loss": 1.0305, + "step": 37985 + }, + { + "epoch": 2.81, + "learning_rate": 1.1522137297818725e-05, + "loss": 0.9765, + "step": 37986 + }, + { + "epoch": 2.81, + "learning_rate": 1.1521743072799213e-05, + "loss": 0.9947, + "step": 37987 + }, + { + "epoch": 2.81, + "learning_rate": 1.1521348845358633e-05, + "loss": 1.0152, + "step": 37988 + }, + { + "epoch": 2.81, + "learning_rate": 1.1520954615497605e-05, + "loss": 0.9765, + "step": 37989 + }, + { + "epoch": 2.81, + "learning_rate": 1.1520560383216765e-05, + "loss": 1.0232, + "step": 37990 + }, + { + "epoch": 2.81, + "learning_rate": 1.1520166148516733e-05, + "loss": 1.0829, + "step": 37991 + }, + { + "epoch": 2.81, + "learning_rate": 1.1519771911398139e-05, + "loss": 1.0297, + "step": 37992 + }, + { + "epoch": 2.81, + "learning_rate": 1.1519377671861606e-05, + "loss": 1.0116, + "step": 37993 + }, + { + "epoch": 2.81, + "learning_rate": 1.1518983429907769e-05, + "loss": 1.0306, + "step": 37994 + }, + { + "epoch": 2.81, + "learning_rate": 1.151858918553725e-05, + "loss": 1.001, + "step": 37995 + }, + { + "epoch": 2.81, + "learning_rate": 1.1518194938750677e-05, + "loss": 0.9893, + "step": 37996 + }, + { + "epoch": 2.81, + "learning_rate": 1.1517800689548675e-05, + "loss": 0.8959, + "step": 37997 + }, + { + "epoch": 2.81, + "learning_rate": 1.1517406437931877e-05, + "loss": 1.0076, + "step": 37998 + }, + { + "epoch": 2.81, + "learning_rate": 1.1517012183900908e-05, + "loss": 1.0596, + "step": 37999 + }, + { + "epoch": 2.81, + "learning_rate": 1.1516617927456394e-05, + "loss": 1.0339, + "step": 38000 + }, + { + "epoch": 2.81, + "learning_rate": 1.151622366859896e-05, + "loss": 1.0324, + "step": 38001 + }, + { + "epoch": 2.81, + "learning_rate": 1.1515829407329237e-05, + "loss": 1.0147, + "step": 38002 + }, + { + "epoch": 2.81, + "learning_rate": 1.151543514364785e-05, + "loss": 1.0393, + "step": 38003 + }, + { + "epoch": 2.81, + "learning_rate": 1.151504087755543e-05, + "loss": 1.0137, + "step": 38004 + }, + { + "epoch": 2.81, + "learning_rate": 1.15146466090526e-05, + "loss": 0.9469, + "step": 38005 + }, + { + "epoch": 2.81, + "learning_rate": 1.1514252338139989e-05, + "loss": 0.9706, + "step": 38006 + }, + { + "epoch": 2.81, + "learning_rate": 1.1513858064818222e-05, + "loss": 0.9646, + "step": 38007 + }, + { + "epoch": 2.81, + "learning_rate": 1.1513463789087935e-05, + "loss": 0.9815, + "step": 38008 + }, + { + "epoch": 2.81, + "learning_rate": 1.1513069510949744e-05, + "loss": 1.0356, + "step": 38009 + }, + { + "epoch": 2.81, + "learning_rate": 1.1512675230404282e-05, + "loss": 1.0342, + "step": 38010 + }, + { + "epoch": 2.81, + "learning_rate": 1.1512280947452173e-05, + "loss": 0.9236, + "step": 38011 + }, + { + "epoch": 2.81, + "learning_rate": 1.151188666209405e-05, + "loss": 1.045, + "step": 38012 + }, + { + "epoch": 2.81, + "learning_rate": 1.1511492374330535e-05, + "loss": 1.11, + "step": 38013 + }, + { + "epoch": 2.81, + "learning_rate": 1.1511098084162258e-05, + "loss": 0.9357, + "step": 38014 + }, + { + "epoch": 2.81, + "learning_rate": 1.1510703791589843e-05, + "loss": 1.0276, + "step": 38015 + }, + { + "epoch": 2.81, + "learning_rate": 1.1510309496613925e-05, + "loss": 1.0749, + "step": 38016 + }, + { + "epoch": 2.81, + "learning_rate": 1.1509915199235121e-05, + "loss": 1.0136, + "step": 38017 + }, + { + "epoch": 2.81, + "learning_rate": 1.1509520899454068e-05, + "loss": 1.0055, + "step": 38018 + }, + { + "epoch": 2.81, + "learning_rate": 1.1509126597271387e-05, + "loss": 1.0082, + "step": 38019 + }, + { + "epoch": 2.81, + "learning_rate": 1.1508732292687705e-05, + "loss": 0.9892, + "step": 38020 + }, + { + "epoch": 2.81, + "learning_rate": 1.1508337985703654e-05, + "loss": 0.9585, + "step": 38021 + }, + { + "epoch": 2.81, + "learning_rate": 1.1507943676319862e-05, + "loss": 1.0471, + "step": 38022 + }, + { + "epoch": 2.81, + "learning_rate": 1.1507549364536948e-05, + "loss": 1.0068, + "step": 38023 + }, + { + "epoch": 2.81, + "learning_rate": 1.1507155050355545e-05, + "loss": 1.0269, + "step": 38024 + }, + { + "epoch": 2.81, + "learning_rate": 1.150676073377628e-05, + "loss": 1.0084, + "step": 38025 + }, + { + "epoch": 2.81, + "learning_rate": 1.150636641479978e-05, + "loss": 0.9642, + "step": 38026 + }, + { + "epoch": 2.81, + "learning_rate": 1.1505972093426677e-05, + "loss": 1.1315, + "step": 38027 + }, + { + "epoch": 2.81, + "learning_rate": 1.150557776965759e-05, + "loss": 1.0622, + "step": 38028 + }, + { + "epoch": 2.81, + "learning_rate": 1.150518344349315e-05, + "loss": 1.0743, + "step": 38029 + }, + { + "epoch": 2.81, + "learning_rate": 1.1504789114933986e-05, + "loss": 0.966, + "step": 38030 + }, + { + "epoch": 2.81, + "learning_rate": 1.1504394783980724e-05, + "loss": 0.9666, + "step": 38031 + }, + { + "epoch": 2.81, + "learning_rate": 1.150400045063399e-05, + "loss": 1.1299, + "step": 38032 + }, + { + "epoch": 2.81, + "learning_rate": 1.1503606114894414e-05, + "loss": 0.9966, + "step": 38033 + }, + { + "epoch": 2.81, + "learning_rate": 1.1503211776762622e-05, + "loss": 0.8713, + "step": 38034 + }, + { + "epoch": 2.81, + "learning_rate": 1.1502817436239241e-05, + "loss": 1.0223, + "step": 38035 + }, + { + "epoch": 2.81, + "learning_rate": 1.15024230933249e-05, + "loss": 1.1061, + "step": 38036 + }, + { + "epoch": 2.81, + "learning_rate": 1.1502028748020228e-05, + "loss": 0.9852, + "step": 38037 + }, + { + "epoch": 2.81, + "learning_rate": 1.1501634400325845e-05, + "loss": 1.106, + "step": 38038 + }, + { + "epoch": 2.81, + "learning_rate": 1.1501240050242386e-05, + "loss": 1.0345, + "step": 38039 + }, + { + "epoch": 2.81, + "learning_rate": 1.1500845697770475e-05, + "loss": 1.0188, + "step": 38040 + }, + { + "epoch": 2.81, + "learning_rate": 1.1500451342910738e-05, + "loss": 1.0632, + "step": 38041 + }, + { + "epoch": 2.81, + "learning_rate": 1.1500056985663806e-05, + "loss": 1.0376, + "step": 38042 + }, + { + "epoch": 2.81, + "learning_rate": 1.1499662626030305e-05, + "loss": 1.0355, + "step": 38043 + }, + { + "epoch": 2.81, + "learning_rate": 1.1499268264010862e-05, + "loss": 1.0273, + "step": 38044 + }, + { + "epoch": 2.81, + "learning_rate": 1.1498873899606104e-05, + "loss": 1.0351, + "step": 38045 + }, + { + "epoch": 2.81, + "learning_rate": 1.149847953281666e-05, + "loss": 0.9865, + "step": 38046 + }, + { + "epoch": 2.81, + "learning_rate": 1.1498085163643159e-05, + "loss": 0.9719, + "step": 38047 + }, + { + "epoch": 2.81, + "learning_rate": 1.1497690792086221e-05, + "loss": 0.9939, + "step": 38048 + }, + { + "epoch": 2.81, + "learning_rate": 1.1497296418146481e-05, + "loss": 1.0241, + "step": 38049 + }, + { + "epoch": 2.81, + "learning_rate": 1.1496902041824563e-05, + "loss": 1.0316, + "step": 38050 + }, + { + "epoch": 2.81, + "learning_rate": 1.1496507663121095e-05, + "loss": 0.9603, + "step": 38051 + }, + { + "epoch": 2.81, + "learning_rate": 1.1496113282036705e-05, + "loss": 0.9911, + "step": 38052 + }, + { + "epoch": 2.81, + "learning_rate": 1.1495718898572022e-05, + "loss": 0.9982, + "step": 38053 + }, + { + "epoch": 2.81, + "learning_rate": 1.1495324512727668e-05, + "loss": 1.0394, + "step": 38054 + }, + { + "epoch": 2.81, + "learning_rate": 1.149493012450428e-05, + "loss": 1.0143, + "step": 38055 + }, + { + "epoch": 2.81, + "learning_rate": 1.1494535733902472e-05, + "loss": 1.0805, + "step": 38056 + }, + { + "epoch": 2.81, + "learning_rate": 1.1494141340922885e-05, + "loss": 1.0057, + "step": 38057 + }, + { + "epoch": 2.81, + "learning_rate": 1.1493746945566137e-05, + "loss": 1.0013, + "step": 38058 + }, + { + "epoch": 2.81, + "learning_rate": 1.1493352547832861e-05, + "loss": 0.9714, + "step": 38059 + }, + { + "epoch": 2.81, + "learning_rate": 1.1492958147723681e-05, + "loss": 0.9958, + "step": 38060 + }, + { + "epoch": 2.81, + "learning_rate": 1.1492563745239227e-05, + "loss": 0.9768, + "step": 38061 + }, + { + "epoch": 2.81, + "learning_rate": 1.1492169340380124e-05, + "loss": 0.9609, + "step": 38062 + }, + { + "epoch": 2.81, + "learning_rate": 1.1491774933147002e-05, + "loss": 1.0306, + "step": 38063 + }, + { + "epoch": 2.81, + "learning_rate": 1.149138052354049e-05, + "loss": 1.0429, + "step": 38064 + }, + { + "epoch": 2.81, + "learning_rate": 1.1490986111561211e-05, + "loss": 1.0151, + "step": 38065 + }, + { + "epoch": 2.81, + "learning_rate": 1.1490591697209792e-05, + "loss": 1.0461, + "step": 38066 + }, + { + "epoch": 2.81, + "learning_rate": 1.1490197280486865e-05, + "loss": 0.9879, + "step": 38067 + }, + { + "epoch": 2.81, + "learning_rate": 1.1489802861393053e-05, + "loss": 1.0037, + "step": 38068 + }, + { + "epoch": 2.81, + "learning_rate": 1.1489408439928991e-05, + "loss": 1.0588, + "step": 38069 + }, + { + "epoch": 2.81, + "learning_rate": 1.1489014016095296e-05, + "loss": 0.8993, + "step": 38070 + }, + { + "epoch": 2.81, + "learning_rate": 1.1488619589892605e-05, + "loss": 1.0066, + "step": 38071 + }, + { + "epoch": 2.81, + "learning_rate": 1.1488225161321538e-05, + "loss": 1.0546, + "step": 38072 + }, + { + "epoch": 2.81, + "learning_rate": 1.148783073038273e-05, + "loss": 0.9777, + "step": 38073 + }, + { + "epoch": 2.81, + "learning_rate": 1.1487436297076804e-05, + "loss": 0.9528, + "step": 38074 + }, + { + "epoch": 2.81, + "learning_rate": 1.1487041861404385e-05, + "loss": 1.0154, + "step": 38075 + }, + { + "epoch": 2.81, + "learning_rate": 1.1486647423366106e-05, + "loss": 0.9692, + "step": 38076 + }, + { + "epoch": 2.81, + "learning_rate": 1.1486252982962591e-05, + "loss": 0.9939, + "step": 38077 + }, + { + "epoch": 2.81, + "learning_rate": 1.148585854019447e-05, + "loss": 0.9645, + "step": 38078 + }, + { + "epoch": 2.81, + "learning_rate": 1.1485464095062369e-05, + "loss": 0.9459, + "step": 38079 + }, + { + "epoch": 2.81, + "learning_rate": 1.1485069647566915e-05, + "loss": 1.0274, + "step": 38080 + }, + { + "epoch": 2.81, + "learning_rate": 1.1484675197708737e-05, + "loss": 0.9768, + "step": 38081 + }, + { + "epoch": 2.81, + "learning_rate": 1.1484280745488464e-05, + "loss": 1.0429, + "step": 38082 + }, + { + "epoch": 2.81, + "learning_rate": 1.1483886290906719e-05, + "loss": 1.0758, + "step": 38083 + }, + { + "epoch": 2.81, + "learning_rate": 1.1483491833964133e-05, + "loss": 1.0236, + "step": 38084 + }, + { + "epoch": 2.81, + "learning_rate": 1.1483097374661333e-05, + "loss": 1.1064, + "step": 38085 + }, + { + "epoch": 2.81, + "learning_rate": 1.1482702912998946e-05, + "loss": 1.0567, + "step": 38086 + }, + { + "epoch": 2.81, + "learning_rate": 1.14823084489776e-05, + "loss": 1.0199, + "step": 38087 + }, + { + "epoch": 2.81, + "learning_rate": 1.1481913982597922e-05, + "loss": 0.8123, + "step": 38088 + }, + { + "epoch": 2.81, + "learning_rate": 1.1481519513860539e-05, + "loss": 1.0548, + "step": 38089 + }, + { + "epoch": 2.81, + "learning_rate": 1.1481125042766081e-05, + "loss": 1.0737, + "step": 38090 + }, + { + "epoch": 2.81, + "learning_rate": 1.1480730569315174e-05, + "loss": 0.9032, + "step": 38091 + }, + { + "epoch": 2.81, + "learning_rate": 1.1480336093508449e-05, + "loss": 0.9964, + "step": 38092 + }, + { + "epoch": 2.81, + "learning_rate": 1.1479941615346526e-05, + "loss": 0.9675, + "step": 38093 + }, + { + "epoch": 2.81, + "learning_rate": 1.147954713483004e-05, + "loss": 1.0235, + "step": 38094 + }, + { + "epoch": 2.81, + "learning_rate": 1.1479152651959613e-05, + "loss": 1.0285, + "step": 38095 + }, + { + "epoch": 2.81, + "learning_rate": 1.1478758166735877e-05, + "loss": 1.0795, + "step": 38096 + }, + { + "epoch": 2.81, + "learning_rate": 1.1478363679159456e-05, + "loss": 0.9883, + "step": 38097 + }, + { + "epoch": 2.82, + "learning_rate": 1.1477969189230982e-05, + "loss": 0.9522, + "step": 38098 + }, + { + "epoch": 2.82, + "learning_rate": 1.147757469695108e-05, + "loss": 0.9592, + "step": 38099 + }, + { + "epoch": 2.82, + "learning_rate": 1.1477180202320377e-05, + "loss": 1.035, + "step": 38100 + }, + { + "epoch": 2.82, + "learning_rate": 1.14767857053395e-05, + "loss": 0.9549, + "step": 38101 + }, + { + "epoch": 2.82, + "learning_rate": 1.1476391206009083e-05, + "loss": 0.9324, + "step": 38102 + }, + { + "epoch": 2.82, + "learning_rate": 1.1475996704329743e-05, + "loss": 0.9462, + "step": 38103 + }, + { + "epoch": 2.82, + "learning_rate": 1.1475602200302116e-05, + "loss": 1.0414, + "step": 38104 + }, + { + "epoch": 2.82, + "learning_rate": 1.1475207693926827e-05, + "loss": 0.9917, + "step": 38105 + }, + { + "epoch": 2.82, + "learning_rate": 1.1474813185204502e-05, + "loss": 0.9377, + "step": 38106 + }, + { + "epoch": 2.82, + "learning_rate": 1.1474418674135773e-05, + "loss": 1.016, + "step": 38107 + }, + { + "epoch": 2.82, + "learning_rate": 1.1474024160721262e-05, + "loss": 1.0608, + "step": 38108 + }, + { + "epoch": 2.82, + "learning_rate": 1.1473629644961602e-05, + "loss": 0.9823, + "step": 38109 + }, + { + "epoch": 2.82, + "learning_rate": 1.1473235126857419e-05, + "loss": 0.9695, + "step": 38110 + }, + { + "epoch": 2.82, + "learning_rate": 1.147284060640934e-05, + "loss": 0.9484, + "step": 38111 + }, + { + "epoch": 2.82, + "learning_rate": 1.147244608361799e-05, + "loss": 0.9231, + "step": 38112 + }, + { + "epoch": 2.82, + "learning_rate": 1.1472051558483999e-05, + "loss": 0.9626, + "step": 38113 + }, + { + "epoch": 2.82, + "learning_rate": 1.1471657031007999e-05, + "loss": 1.0153, + "step": 38114 + }, + { + "epoch": 2.82, + "learning_rate": 1.147126250119061e-05, + "loss": 0.9978, + "step": 38115 + }, + { + "epoch": 2.82, + "learning_rate": 1.1470867969032465e-05, + "loss": 1.0167, + "step": 38116 + }, + { + "epoch": 2.82, + "learning_rate": 1.1470473434534188e-05, + "loss": 1.0959, + "step": 38117 + }, + { + "epoch": 2.82, + "learning_rate": 1.147007889769641e-05, + "loss": 1.0739, + "step": 38118 + }, + { + "epoch": 2.82, + "learning_rate": 1.1469684358519759e-05, + "loss": 1.0269, + "step": 38119 + }, + { + "epoch": 2.82, + "learning_rate": 1.146928981700486e-05, + "loss": 0.8585, + "step": 38120 + }, + { + "epoch": 2.82, + "learning_rate": 1.1468895273152344e-05, + "loss": 0.891, + "step": 38121 + }, + { + "epoch": 2.82, + "learning_rate": 1.1468500726962834e-05, + "loss": 0.935, + "step": 38122 + }, + { + "epoch": 2.82, + "learning_rate": 1.1468106178436961e-05, + "loss": 1.0127, + "step": 38123 + }, + { + "epoch": 2.82, + "learning_rate": 1.146771162757535e-05, + "loss": 1.0854, + "step": 38124 + }, + { + "epoch": 2.82, + "learning_rate": 1.1467317074378637e-05, + "loss": 1.0068, + "step": 38125 + }, + { + "epoch": 2.82, + "learning_rate": 1.1466922518847439e-05, + "loss": 1.0187, + "step": 38126 + }, + { + "epoch": 2.82, + "learning_rate": 1.1466527960982387e-05, + "loss": 1.0197, + "step": 38127 + }, + { + "epoch": 2.82, + "learning_rate": 1.1466133400784114e-05, + "loss": 1.0418, + "step": 38128 + }, + { + "epoch": 2.82, + "learning_rate": 1.1465738838253239e-05, + "loss": 0.8853, + "step": 38129 + }, + { + "epoch": 2.82, + "learning_rate": 1.14653442733904e-05, + "loss": 0.9725, + "step": 38130 + }, + { + "epoch": 2.82, + "learning_rate": 1.1464949706196215e-05, + "loss": 0.9898, + "step": 38131 + }, + { + "epoch": 2.82, + "learning_rate": 1.1464555136671318e-05, + "loss": 1.0148, + "step": 38132 + }, + { + "epoch": 2.82, + "learning_rate": 1.1464160564816336e-05, + "loss": 0.9541, + "step": 38133 + }, + { + "epoch": 2.82, + "learning_rate": 1.146376599063189e-05, + "loss": 1.0316, + "step": 38134 + }, + { + "epoch": 2.82, + "learning_rate": 1.1463371414118618e-05, + "loss": 0.9264, + "step": 38135 + }, + { + "epoch": 2.82, + "learning_rate": 1.1462976835277143e-05, + "loss": 1.0029, + "step": 38136 + }, + { + "epoch": 2.82, + "learning_rate": 1.146258225410809e-05, + "loss": 0.9894, + "step": 38137 + }, + { + "epoch": 2.82, + "learning_rate": 1.1462187670612091e-05, + "loss": 1.0128, + "step": 38138 + }, + { + "epoch": 2.82, + "learning_rate": 1.1461793084789777e-05, + "loss": 0.9099, + "step": 38139 + }, + { + "epoch": 2.82, + "learning_rate": 1.1461398496641763e-05, + "loss": 1.0757, + "step": 38140 + }, + { + "epoch": 2.82, + "learning_rate": 1.146100390616869e-05, + "loss": 1.074, + "step": 38141 + }, + { + "epoch": 2.82, + "learning_rate": 1.1460609313371179e-05, + "loss": 0.99, + "step": 38142 + }, + { + "epoch": 2.82, + "learning_rate": 1.146021471824986e-05, + "loss": 0.8768, + "step": 38143 + }, + { + "epoch": 2.82, + "learning_rate": 1.1459820120805358e-05, + "loss": 1.0188, + "step": 38144 + }, + { + "epoch": 2.82, + "learning_rate": 1.1459425521038305e-05, + "loss": 0.9821, + "step": 38145 + }, + { + "epoch": 2.82, + "learning_rate": 1.1459030918949328e-05, + "loss": 1.0071, + "step": 38146 + }, + { + "epoch": 2.82, + "learning_rate": 1.1458636314539052e-05, + "loss": 1.0199, + "step": 38147 + }, + { + "epoch": 2.82, + "learning_rate": 1.1458241707808107e-05, + "loss": 0.9448, + "step": 38148 + }, + { + "epoch": 2.82, + "learning_rate": 1.1457847098757122e-05, + "loss": 1.0278, + "step": 38149 + }, + { + "epoch": 2.82, + "learning_rate": 1.145745248738672e-05, + "loss": 1.1244, + "step": 38150 + }, + { + "epoch": 2.82, + "learning_rate": 1.1457057873697534e-05, + "loss": 1.0571, + "step": 38151 + }, + { + "epoch": 2.82, + "learning_rate": 1.1456663257690188e-05, + "loss": 1.0323, + "step": 38152 + }, + { + "epoch": 2.82, + "learning_rate": 1.1456268639365315e-05, + "loss": 1.0071, + "step": 38153 + }, + { + "epoch": 2.82, + "learning_rate": 1.1455874018723537e-05, + "loss": 1.0386, + "step": 38154 + }, + { + "epoch": 2.82, + "learning_rate": 1.1455479395765485e-05, + "loss": 1.004, + "step": 38155 + }, + { + "epoch": 2.82, + "learning_rate": 1.1455084770491785e-05, + "loss": 1.0606, + "step": 38156 + }, + { + "epoch": 2.82, + "learning_rate": 1.1454690142903067e-05, + "loss": 1.0786, + "step": 38157 + }, + { + "epoch": 2.82, + "learning_rate": 1.1454295512999958e-05, + "loss": 1.01, + "step": 38158 + }, + { + "epoch": 2.82, + "learning_rate": 1.1453900880783086e-05, + "loss": 1.0258, + "step": 38159 + }, + { + "epoch": 2.82, + "learning_rate": 1.1453506246253077e-05, + "loss": 1.0259, + "step": 38160 + }, + { + "epoch": 2.82, + "learning_rate": 1.1453111609410562e-05, + "loss": 0.9512, + "step": 38161 + }, + { + "epoch": 2.82, + "learning_rate": 1.1452716970256163e-05, + "loss": 0.9333, + "step": 38162 + }, + { + "epoch": 2.82, + "learning_rate": 1.1452322328790516e-05, + "loss": 0.893, + "step": 38163 + }, + { + "epoch": 2.82, + "learning_rate": 1.1451927685014244e-05, + "loss": 0.9612, + "step": 38164 + }, + { + "epoch": 2.82, + "learning_rate": 1.1451533038927975e-05, + "loss": 0.9476, + "step": 38165 + }, + { + "epoch": 2.82, + "learning_rate": 1.145113839053234e-05, + "loss": 1.0748, + "step": 38166 + }, + { + "epoch": 2.82, + "learning_rate": 1.1450743739827963e-05, + "loss": 0.9957, + "step": 38167 + }, + { + "epoch": 2.82, + "learning_rate": 1.1450349086815475e-05, + "loss": 0.9809, + "step": 38168 + }, + { + "epoch": 2.82, + "learning_rate": 1.14499544314955e-05, + "loss": 1.0203, + "step": 38169 + }, + { + "epoch": 2.82, + "learning_rate": 1.144955977386867e-05, + "loss": 1.0183, + "step": 38170 + }, + { + "epoch": 2.82, + "learning_rate": 1.1449165113935613e-05, + "loss": 1.1475, + "step": 38171 + }, + { + "epoch": 2.82, + "learning_rate": 1.1448770451696949e-05, + "loss": 0.9792, + "step": 38172 + }, + { + "epoch": 2.82, + "learning_rate": 1.1448375787153317e-05, + "loss": 1.0725, + "step": 38173 + }, + { + "epoch": 2.82, + "learning_rate": 1.1447981120305336e-05, + "loss": 1.009, + "step": 38174 + }, + { + "epoch": 2.82, + "learning_rate": 1.1447586451153643e-05, + "loss": 1.0037, + "step": 38175 + }, + { + "epoch": 2.82, + "learning_rate": 1.1447191779698854e-05, + "loss": 1.036, + "step": 38176 + }, + { + "epoch": 2.82, + "learning_rate": 1.144679710594161e-05, + "loss": 0.9608, + "step": 38177 + }, + { + "epoch": 2.82, + "learning_rate": 1.1446402429882531e-05, + "loss": 1.0328, + "step": 38178 + }, + { + "epoch": 2.82, + "learning_rate": 1.1446007751522243e-05, + "loss": 1.0485, + "step": 38179 + }, + { + "epoch": 2.82, + "learning_rate": 1.144561307086138e-05, + "loss": 1.0411, + "step": 38180 + }, + { + "epoch": 2.82, + "learning_rate": 1.1445218387900565e-05, + "loss": 0.9732, + "step": 38181 + }, + { + "epoch": 2.82, + "learning_rate": 1.1444823702640432e-05, + "loss": 0.9279, + "step": 38182 + }, + { + "epoch": 2.82, + "learning_rate": 1.1444429015081602e-05, + "loss": 0.9431, + "step": 38183 + }, + { + "epoch": 2.82, + "learning_rate": 1.144403432522471e-05, + "loss": 1.0399, + "step": 38184 + }, + { + "epoch": 2.82, + "learning_rate": 1.1443639633070373e-05, + "loss": 0.9644, + "step": 38185 + }, + { + "epoch": 2.82, + "learning_rate": 1.1443244938619236e-05, + "loss": 1.0009, + "step": 38186 + }, + { + "epoch": 2.82, + "learning_rate": 1.144285024187191e-05, + "loss": 1.0443, + "step": 38187 + }, + { + "epoch": 2.82, + "learning_rate": 1.1442455542829031e-05, + "loss": 0.9888, + "step": 38188 + }, + { + "epoch": 2.82, + "learning_rate": 1.1442060841491225e-05, + "loss": 0.9913, + "step": 38189 + }, + { + "epoch": 2.82, + "learning_rate": 1.1441666137859121e-05, + "loss": 1.1102, + "step": 38190 + }, + { + "epoch": 2.82, + "learning_rate": 1.1441271431933348e-05, + "loss": 0.9519, + "step": 38191 + }, + { + "epoch": 2.82, + "learning_rate": 1.1440876723714534e-05, + "loss": 1.1474, + "step": 38192 + }, + { + "epoch": 2.82, + "learning_rate": 1.1440482013203303e-05, + "loss": 0.9812, + "step": 38193 + }, + { + "epoch": 2.82, + "learning_rate": 1.144008730040029e-05, + "loss": 0.9774, + "step": 38194 + }, + { + "epoch": 2.82, + "learning_rate": 1.1439692585306114e-05, + "loss": 1.0114, + "step": 38195 + }, + { + "epoch": 2.82, + "learning_rate": 1.143929786792141e-05, + "loss": 1.0538, + "step": 38196 + }, + { + "epoch": 2.82, + "learning_rate": 1.1438903148246803e-05, + "loss": 1.1422, + "step": 38197 + }, + { + "epoch": 2.82, + "learning_rate": 1.1438508426282921e-05, + "loss": 1.0573, + "step": 38198 + }, + { + "epoch": 2.82, + "learning_rate": 1.1438113702030394e-05, + "loss": 0.9193, + "step": 38199 + }, + { + "epoch": 2.82, + "learning_rate": 1.1437718975489849e-05, + "loss": 1.0188, + "step": 38200 + }, + { + "epoch": 2.82, + "learning_rate": 1.1437324246661912e-05, + "loss": 1.0759, + "step": 38201 + }, + { + "epoch": 2.82, + "learning_rate": 1.1436929515547215e-05, + "loss": 1.0783, + "step": 38202 + }, + { + "epoch": 2.82, + "learning_rate": 1.143653478214638e-05, + "loss": 1.0015, + "step": 38203 + }, + { + "epoch": 2.82, + "learning_rate": 1.143614004646004e-05, + "loss": 1.031, + "step": 38204 + }, + { + "epoch": 2.82, + "learning_rate": 1.1435745308488826e-05, + "loss": 1.0404, + "step": 38205 + }, + { + "epoch": 2.82, + "learning_rate": 1.143535056823336e-05, + "loss": 1.0351, + "step": 38206 + }, + { + "epoch": 2.82, + "learning_rate": 1.1434955825694268e-05, + "loss": 0.9067, + "step": 38207 + }, + { + "epoch": 2.82, + "learning_rate": 1.1434561080872184e-05, + "loss": 0.9806, + "step": 38208 + }, + { + "epoch": 2.82, + "learning_rate": 1.1434166333767734e-05, + "loss": 0.9344, + "step": 38209 + }, + { + "epoch": 2.82, + "learning_rate": 1.1433771584381545e-05, + "loss": 1.0439, + "step": 38210 + }, + { + "epoch": 2.82, + "learning_rate": 1.1433376832714247e-05, + "loss": 1.0233, + "step": 38211 + }, + { + "epoch": 2.82, + "learning_rate": 1.1432982078766466e-05, + "loss": 0.9511, + "step": 38212 + }, + { + "epoch": 2.82, + "learning_rate": 1.143258732253883e-05, + "loss": 1.0412, + "step": 38213 + }, + { + "epoch": 2.82, + "learning_rate": 1.1432192564031972e-05, + "loss": 1.0659, + "step": 38214 + }, + { + "epoch": 2.82, + "learning_rate": 1.1431797803246514e-05, + "loss": 0.9552, + "step": 38215 + }, + { + "epoch": 2.82, + "learning_rate": 1.1431403040183087e-05, + "loss": 0.9815, + "step": 38216 + }, + { + "epoch": 2.82, + "learning_rate": 1.1431008274842316e-05, + "loss": 0.962, + "step": 38217 + }, + { + "epoch": 2.82, + "learning_rate": 1.1430613507224834e-05, + "loss": 1.0083, + "step": 38218 + }, + { + "epoch": 2.82, + "learning_rate": 1.1430218737331264e-05, + "loss": 1.0373, + "step": 38219 + }, + { + "epoch": 2.82, + "learning_rate": 1.1429823965162237e-05, + "loss": 1.1459, + "step": 38220 + }, + { + "epoch": 2.82, + "learning_rate": 1.142942919071838e-05, + "loss": 0.9836, + "step": 38221 + }, + { + "epoch": 2.82, + "learning_rate": 1.1429034414000324e-05, + "loss": 0.9304, + "step": 38222 + }, + { + "epoch": 2.82, + "learning_rate": 1.1428639635008694e-05, + "loss": 1.0478, + "step": 38223 + }, + { + "epoch": 2.82, + "learning_rate": 1.1428244853744117e-05, + "loss": 1.0834, + "step": 38224 + }, + { + "epoch": 2.82, + "learning_rate": 1.1427850070207225e-05, + "loss": 0.9822, + "step": 38225 + }, + { + "epoch": 2.82, + "learning_rate": 1.1427455284398641e-05, + "loss": 1.0381, + "step": 38226 + }, + { + "epoch": 2.82, + "learning_rate": 1.1427060496318998e-05, + "loss": 0.9386, + "step": 38227 + }, + { + "epoch": 2.82, + "learning_rate": 1.1426665705968922e-05, + "loss": 1.0716, + "step": 38228 + }, + { + "epoch": 2.82, + "learning_rate": 1.1426270913349043e-05, + "loss": 1.1035, + "step": 38229 + }, + { + "epoch": 2.82, + "learning_rate": 1.1425876118459985e-05, + "loss": 1.0812, + "step": 38230 + }, + { + "epoch": 2.82, + "learning_rate": 1.142548132130238e-05, + "loss": 1.0363, + "step": 38231 + }, + { + "epoch": 2.82, + "learning_rate": 1.1425086521876853e-05, + "loss": 0.8863, + "step": 38232 + }, + { + "epoch": 2.83, + "learning_rate": 1.1424691720184035e-05, + "loss": 0.9081, + "step": 38233 + }, + { + "epoch": 2.83, + "learning_rate": 1.1424296916224554e-05, + "loss": 0.8684, + "step": 38234 + }, + { + "epoch": 2.83, + "learning_rate": 1.1423902109999034e-05, + "loss": 1.0441, + "step": 38235 + }, + { + "epoch": 2.83, + "learning_rate": 1.1423507301508106e-05, + "loss": 1.0103, + "step": 38236 + }, + { + "epoch": 2.83, + "learning_rate": 1.1423112490752401e-05, + "loss": 0.981, + "step": 38237 + }, + { + "epoch": 2.83, + "learning_rate": 1.1422717677732542e-05, + "loss": 1.0905, + "step": 38238 + }, + { + "epoch": 2.83, + "learning_rate": 1.1422322862449163e-05, + "loss": 1.1135, + "step": 38239 + }, + { + "epoch": 2.83, + "learning_rate": 1.1421928044902883e-05, + "loss": 1.106, + "step": 38240 + }, + { + "epoch": 2.83, + "learning_rate": 1.1421533225094341e-05, + "loss": 0.9934, + "step": 38241 + }, + { + "epoch": 2.83, + "learning_rate": 1.1421138403024159e-05, + "loss": 0.9392, + "step": 38242 + }, + { + "epoch": 2.83, + "learning_rate": 1.1420743578692965e-05, + "loss": 0.9603, + "step": 38243 + }, + { + "epoch": 2.83, + "learning_rate": 1.1420348752101389e-05, + "loss": 0.9968, + "step": 38244 + }, + { + "epoch": 2.83, + "learning_rate": 1.141995392325006e-05, + "loss": 1.0605, + "step": 38245 + }, + { + "epoch": 2.83, + "learning_rate": 1.14195590921396e-05, + "loss": 0.9079, + "step": 38246 + }, + { + "epoch": 2.83, + "learning_rate": 1.1419164258770645e-05, + "loss": 0.9505, + "step": 38247 + }, + { + "epoch": 2.83, + "learning_rate": 1.1418769423143818e-05, + "loss": 0.9684, + "step": 38248 + }, + { + "epoch": 2.83, + "learning_rate": 1.1418374585259752e-05, + "loss": 0.9516, + "step": 38249 + }, + { + "epoch": 2.83, + "learning_rate": 1.141797974511907e-05, + "loss": 0.9655, + "step": 38250 + }, + { + "epoch": 2.83, + "learning_rate": 1.1417584902722408e-05, + "loss": 1.0742, + "step": 38251 + }, + { + "epoch": 2.83, + "learning_rate": 1.1417190058070382e-05, + "loss": 1.0383, + "step": 38252 + }, + { + "epoch": 2.83, + "learning_rate": 1.141679521116363e-05, + "loss": 0.9825, + "step": 38253 + }, + { + "epoch": 2.83, + "learning_rate": 1.1416400362002776e-05, + "loss": 0.9333, + "step": 38254 + }, + { + "epoch": 2.83, + "learning_rate": 1.1416005510588448e-05, + "loss": 0.9827, + "step": 38255 + }, + { + "epoch": 2.83, + "learning_rate": 1.1415610656921276e-05, + "loss": 0.9678, + "step": 38256 + }, + { + "epoch": 2.83, + "learning_rate": 1.1415215801001888e-05, + "loss": 0.9552, + "step": 38257 + }, + { + "epoch": 2.83, + "learning_rate": 1.1414820942830913e-05, + "loss": 1.0746, + "step": 38258 + }, + { + "epoch": 2.83, + "learning_rate": 1.1414426082408979e-05, + "loss": 0.9649, + "step": 38259 + }, + { + "epoch": 2.83, + "learning_rate": 1.141403121973671e-05, + "loss": 1.1615, + "step": 38260 + }, + { + "epoch": 2.83, + "learning_rate": 1.1413636354814743e-05, + "loss": 0.9408, + "step": 38261 + }, + { + "epoch": 2.83, + "learning_rate": 1.1413241487643696e-05, + "loss": 0.9476, + "step": 38262 + }, + { + "epoch": 2.83, + "learning_rate": 1.1412846618224204e-05, + "loss": 0.8985, + "step": 38263 + }, + { + "epoch": 2.83, + "learning_rate": 1.1412451746556892e-05, + "loss": 1.0054, + "step": 38264 + }, + { + "epoch": 2.83, + "learning_rate": 1.1412056872642392e-05, + "loss": 0.9525, + "step": 38265 + }, + { + "epoch": 2.83, + "learning_rate": 1.1411661996481324e-05, + "loss": 0.9446, + "step": 38266 + }, + { + "epoch": 2.83, + "learning_rate": 1.1411267118074328e-05, + "loss": 0.9611, + "step": 38267 + }, + { + "epoch": 2.83, + "learning_rate": 1.1410872237422023e-05, + "loss": 0.9565, + "step": 38268 + }, + { + "epoch": 2.83, + "learning_rate": 1.1410477354525043e-05, + "loss": 1.0764, + "step": 38269 + }, + { + "epoch": 2.83, + "learning_rate": 1.1410082469384011e-05, + "loss": 0.9724, + "step": 38270 + }, + { + "epoch": 2.83, + "learning_rate": 1.1409687581999558e-05, + "loss": 0.9535, + "step": 38271 + }, + { + "epoch": 2.83, + "learning_rate": 1.1409292692372316e-05, + "loss": 0.875, + "step": 38272 + }, + { + "epoch": 2.83, + "learning_rate": 1.1408897800502905e-05, + "loss": 1.0317, + "step": 38273 + }, + { + "epoch": 2.83, + "learning_rate": 1.140850290639196e-05, + "loss": 1.0226, + "step": 38274 + }, + { + "epoch": 2.83, + "learning_rate": 1.1408108010040104e-05, + "loss": 1.1031, + "step": 38275 + }, + { + "epoch": 2.83, + "learning_rate": 1.1407713111447974e-05, + "loss": 1.0949, + "step": 38276 + }, + { + "epoch": 2.83, + "learning_rate": 1.140731821061619e-05, + "loss": 0.9591, + "step": 38277 + }, + { + "epoch": 2.83, + "learning_rate": 1.1406923307545382e-05, + "loss": 1.0277, + "step": 38278 + }, + { + "epoch": 2.83, + "learning_rate": 1.1406528402236178e-05, + "loss": 1.0323, + "step": 38279 + }, + { + "epoch": 2.83, + "learning_rate": 1.1406133494689211e-05, + "loss": 1.0089, + "step": 38280 + }, + { + "epoch": 2.83, + "learning_rate": 1.1405738584905102e-05, + "loss": 0.9902, + "step": 38281 + }, + { + "epoch": 2.83, + "learning_rate": 1.1405343672884486e-05, + "loss": 1.0598, + "step": 38282 + }, + { + "epoch": 2.83, + "learning_rate": 1.1404948758627984e-05, + "loss": 0.8912, + "step": 38283 + }, + { + "epoch": 2.83, + "learning_rate": 1.1404553842136235e-05, + "loss": 1.1082, + "step": 38284 + }, + { + "epoch": 2.83, + "learning_rate": 1.1404158923409856e-05, + "loss": 0.8736, + "step": 38285 + }, + { + "epoch": 2.83, + "learning_rate": 1.1403764002449483e-05, + "loss": 0.9506, + "step": 38286 + }, + { + "epoch": 2.83, + "learning_rate": 1.140336907925574e-05, + "loss": 0.9989, + "step": 38287 + }, + { + "epoch": 2.83, + "learning_rate": 1.1402974153829257e-05, + "loss": 1.1028, + "step": 38288 + }, + { + "epoch": 2.83, + "learning_rate": 1.1402579226170662e-05, + "loss": 1.0246, + "step": 38289 + }, + { + "epoch": 2.83, + "learning_rate": 1.1402184296280585e-05, + "loss": 1.0005, + "step": 38290 + }, + { + "epoch": 2.83, + "learning_rate": 1.1401789364159652e-05, + "loss": 0.9422, + "step": 38291 + }, + { + "epoch": 2.83, + "learning_rate": 1.1401394429808493e-05, + "loss": 1.0079, + "step": 38292 + }, + { + "epoch": 2.83, + "learning_rate": 1.1400999493227731e-05, + "loss": 1.0215, + "step": 38293 + }, + { + "epoch": 2.83, + "learning_rate": 1.1400604554418004e-05, + "loss": 1.0549, + "step": 38294 + }, + { + "epoch": 2.83, + "learning_rate": 1.1400209613379933e-05, + "loss": 1.007, + "step": 38295 + }, + { + "epoch": 2.83, + "learning_rate": 1.1399814670114152e-05, + "loss": 0.954, + "step": 38296 + }, + { + "epoch": 2.83, + "learning_rate": 1.1399419724621279e-05, + "loss": 1.0271, + "step": 38297 + }, + { + "epoch": 2.83, + "learning_rate": 1.1399024776901958e-05, + "loss": 1.0306, + "step": 38298 + }, + { + "epoch": 2.83, + "learning_rate": 1.1398629826956802e-05, + "loss": 1.0238, + "step": 38299 + }, + { + "epoch": 2.83, + "learning_rate": 1.139823487478645e-05, + "loss": 1.0222, + "step": 38300 + }, + { + "epoch": 2.83, + "learning_rate": 1.1397839920391521e-05, + "loss": 1.0319, + "step": 38301 + }, + { + "epoch": 2.83, + "learning_rate": 1.1397444963772653e-05, + "loss": 1.0062, + "step": 38302 + }, + { + "epoch": 2.83, + "learning_rate": 1.1397050004930468e-05, + "loss": 0.977, + "step": 38303 + }, + { + "epoch": 2.83, + "learning_rate": 1.1396655043865596e-05, + "loss": 1.0223, + "step": 38304 + }, + { + "epoch": 2.83, + "learning_rate": 1.1396260080578668e-05, + "loss": 1.0262, + "step": 38305 + }, + { + "epoch": 2.83, + "learning_rate": 1.139586511507031e-05, + "loss": 1.0819, + "step": 38306 + }, + { + "epoch": 2.83, + "learning_rate": 1.1395470147341148e-05, + "loss": 0.9535, + "step": 38307 + }, + { + "epoch": 2.83, + "learning_rate": 1.1395075177391818e-05, + "loss": 1.0387, + "step": 38308 + }, + { + "epoch": 2.83, + "learning_rate": 1.1394680205222939e-05, + "loss": 0.9983, + "step": 38309 + }, + { + "epoch": 2.83, + "learning_rate": 1.1394285230835144e-05, + "loss": 1.0279, + "step": 38310 + }, + { + "epoch": 2.83, + "learning_rate": 1.1393890254229064e-05, + "loss": 0.991, + "step": 38311 + }, + { + "epoch": 2.83, + "learning_rate": 1.1393495275405322e-05, + "loss": 0.9928, + "step": 38312 + }, + { + "epoch": 2.83, + "learning_rate": 1.1393100294364547e-05, + "loss": 0.9599, + "step": 38313 + }, + { + "epoch": 2.83, + "learning_rate": 1.1392705311107375e-05, + "loss": 0.9789, + "step": 38314 + }, + { + "epoch": 2.83, + "learning_rate": 1.1392310325634424e-05, + "loss": 1.052, + "step": 38315 + }, + { + "epoch": 2.83, + "learning_rate": 1.1391915337946328e-05, + "loss": 1.035, + "step": 38316 + }, + { + "epoch": 2.83, + "learning_rate": 1.1391520348043716e-05, + "loss": 1.1374, + "step": 38317 + }, + { + "epoch": 2.83, + "learning_rate": 1.1391125355927215e-05, + "loss": 1.045, + "step": 38318 + }, + { + "epoch": 2.83, + "learning_rate": 1.1390730361597453e-05, + "loss": 1.0519, + "step": 38319 + }, + { + "epoch": 2.83, + "learning_rate": 1.139033536505506e-05, + "loss": 1.0762, + "step": 38320 + }, + { + "epoch": 2.83, + "learning_rate": 1.138994036630066e-05, + "loss": 0.9971, + "step": 38321 + }, + { + "epoch": 2.83, + "learning_rate": 1.1389545365334888e-05, + "loss": 0.8572, + "step": 38322 + }, + { + "epoch": 2.83, + "learning_rate": 1.1389150362158368e-05, + "loss": 1.0731, + "step": 38323 + }, + { + "epoch": 2.83, + "learning_rate": 1.138875535677173e-05, + "loss": 1.0225, + "step": 38324 + }, + { + "epoch": 2.83, + "learning_rate": 1.13883603491756e-05, + "loss": 1.0368, + "step": 38325 + }, + { + "epoch": 2.83, + "learning_rate": 1.1387965339370612e-05, + "loss": 0.8931, + "step": 38326 + }, + { + "epoch": 2.83, + "learning_rate": 1.138757032735739e-05, + "loss": 1.0112, + "step": 38327 + }, + { + "epoch": 2.83, + "learning_rate": 1.1387175313136563e-05, + "loss": 1.1365, + "step": 38328 + }, + { + "epoch": 2.83, + "learning_rate": 1.1386780296708761e-05, + "loss": 0.9964, + "step": 38329 + }, + { + "epoch": 2.83, + "learning_rate": 1.1386385278074607e-05, + "loss": 0.9479, + "step": 38330 + }, + { + "epoch": 2.83, + "learning_rate": 1.138599025723474e-05, + "loss": 1.0271, + "step": 38331 + }, + { + "epoch": 2.83, + "learning_rate": 1.1385595234189776e-05, + "loss": 1.0502, + "step": 38332 + }, + { + "epoch": 2.83, + "learning_rate": 1.1385200208940357e-05, + "loss": 1.1315, + "step": 38333 + }, + { + "epoch": 2.83, + "learning_rate": 1.1384805181487097e-05, + "loss": 0.9581, + "step": 38334 + }, + { + "epoch": 2.83, + "learning_rate": 1.1384410151830637e-05, + "loss": 0.984, + "step": 38335 + }, + { + "epoch": 2.83, + "learning_rate": 1.13840151199716e-05, + "loss": 0.9526, + "step": 38336 + }, + { + "epoch": 2.83, + "learning_rate": 1.1383620085910613e-05, + "loss": 1.0164, + "step": 38337 + }, + { + "epoch": 2.83, + "learning_rate": 1.1383225049648306e-05, + "loss": 1.1028, + "step": 38338 + }, + { + "epoch": 2.83, + "learning_rate": 1.1382830011185308e-05, + "loss": 0.9679, + "step": 38339 + }, + { + "epoch": 2.83, + "learning_rate": 1.1382434970522248e-05, + "loss": 1.0531, + "step": 38340 + }, + { + "epoch": 2.83, + "learning_rate": 1.1382039927659754e-05, + "loss": 1.038, + "step": 38341 + }, + { + "epoch": 2.83, + "learning_rate": 1.1381644882598453e-05, + "loss": 1.0251, + "step": 38342 + }, + { + "epoch": 2.83, + "learning_rate": 1.1381249835338974e-05, + "loss": 0.9405, + "step": 38343 + }, + { + "epoch": 2.83, + "learning_rate": 1.1380854785881948e-05, + "loss": 0.9578, + "step": 38344 + }, + { + "epoch": 2.83, + "learning_rate": 1.1380459734228005e-05, + "loss": 0.9653, + "step": 38345 + }, + { + "epoch": 2.83, + "learning_rate": 1.1380064680377764e-05, + "loss": 0.951, + "step": 38346 + }, + { + "epoch": 2.83, + "learning_rate": 1.1379669624331864e-05, + "loss": 0.974, + "step": 38347 + }, + { + "epoch": 2.83, + "learning_rate": 1.1379274566090925e-05, + "loss": 1.0702, + "step": 38348 + }, + { + "epoch": 2.83, + "learning_rate": 1.1378879505655584e-05, + "loss": 1.0272, + "step": 38349 + }, + { + "epoch": 2.83, + "learning_rate": 1.1378484443026462e-05, + "loss": 0.9995, + "step": 38350 + }, + { + "epoch": 2.83, + "learning_rate": 1.1378089378204196e-05, + "loss": 0.9099, + "step": 38351 + }, + { + "epoch": 2.83, + "learning_rate": 1.1377694311189405e-05, + "loss": 0.9981, + "step": 38352 + }, + { + "epoch": 2.83, + "learning_rate": 1.1377299241982725e-05, + "loss": 0.9625, + "step": 38353 + }, + { + "epoch": 2.83, + "learning_rate": 1.1376904170584782e-05, + "loss": 0.9808, + "step": 38354 + }, + { + "epoch": 2.83, + "learning_rate": 1.1376509096996202e-05, + "loss": 1.0812, + "step": 38355 + }, + { + "epoch": 2.83, + "learning_rate": 1.1376114021217614e-05, + "loss": 0.8464, + "step": 38356 + }, + { + "epoch": 2.83, + "learning_rate": 1.1375718943249652e-05, + "loss": 1.024, + "step": 38357 + }, + { + "epoch": 2.83, + "learning_rate": 1.137532386309294e-05, + "loss": 1.0523, + "step": 38358 + }, + { + "epoch": 2.83, + "learning_rate": 1.1374928780748106e-05, + "loss": 0.9209, + "step": 38359 + }, + { + "epoch": 2.83, + "learning_rate": 1.137453369621578e-05, + "loss": 0.9525, + "step": 38360 + }, + { + "epoch": 2.83, + "learning_rate": 1.1374138609496592e-05, + "loss": 1.0332, + "step": 38361 + }, + { + "epoch": 2.83, + "learning_rate": 1.1373743520591168e-05, + "loss": 1.0463, + "step": 38362 + }, + { + "epoch": 2.83, + "learning_rate": 1.137334842950014e-05, + "loss": 1.0256, + "step": 38363 + }, + { + "epoch": 2.83, + "learning_rate": 1.1372953336224134e-05, + "loss": 1.0607, + "step": 38364 + }, + { + "epoch": 2.83, + "learning_rate": 1.1372558240763776e-05, + "loss": 1.0959, + "step": 38365 + }, + { + "epoch": 2.83, + "learning_rate": 1.13721631431197e-05, + "loss": 0.9844, + "step": 38366 + }, + { + "epoch": 2.83, + "learning_rate": 1.137176804329253e-05, + "loss": 0.9541, + "step": 38367 + }, + { + "epoch": 2.83, + "learning_rate": 1.1371372941282897e-05, + "loss": 0.9927, + "step": 38368 + }, + { + "epoch": 2.84, + "learning_rate": 1.137097783709143e-05, + "loss": 0.9076, + "step": 38369 + }, + { + "epoch": 2.84, + "learning_rate": 1.1370582730718756e-05, + "loss": 1.0142, + "step": 38370 + }, + { + "epoch": 2.84, + "learning_rate": 1.1370187622165505e-05, + "loss": 1.0312, + "step": 38371 + }, + { + "epoch": 2.84, + "learning_rate": 1.1369792511432306e-05, + "loss": 1.0886, + "step": 38372 + }, + { + "epoch": 2.84, + "learning_rate": 1.1369397398519786e-05, + "loss": 0.95, + "step": 38373 + }, + { + "epoch": 2.84, + "learning_rate": 1.1369002283428574e-05, + "loss": 0.9227, + "step": 38374 + }, + { + "epoch": 2.84, + "learning_rate": 1.1368607166159298e-05, + "loss": 1.0414, + "step": 38375 + }, + { + "epoch": 2.84, + "learning_rate": 1.136821204671259e-05, + "loss": 0.9914, + "step": 38376 + }, + { + "epoch": 2.84, + "learning_rate": 1.1367816925089075e-05, + "loss": 0.9167, + "step": 38377 + }, + { + "epoch": 2.84, + "learning_rate": 1.1367421801289382e-05, + "loss": 0.9858, + "step": 38378 + }, + { + "epoch": 2.84, + "learning_rate": 1.1367026675314142e-05, + "loss": 1.0576, + "step": 38379 + }, + { + "epoch": 2.84, + "learning_rate": 1.136663154716398e-05, + "loss": 1.1247, + "step": 38380 + }, + { + "epoch": 2.84, + "learning_rate": 1.1366236416839531e-05, + "loss": 1.0297, + "step": 38381 + }, + { + "epoch": 2.84, + "learning_rate": 1.1365841284341416e-05, + "loss": 0.9639, + "step": 38382 + }, + { + "epoch": 2.84, + "learning_rate": 1.1365446149670267e-05, + "loss": 0.9577, + "step": 38383 + }, + { + "epoch": 2.84, + "learning_rate": 1.1365051012826715e-05, + "loss": 1.0791, + "step": 38384 + }, + { + "epoch": 2.84, + "learning_rate": 1.1364655873811384e-05, + "loss": 1.023, + "step": 38385 + }, + { + "epoch": 2.84, + "learning_rate": 1.1364260732624906e-05, + "loss": 1.0596, + "step": 38386 + }, + { + "epoch": 2.84, + "learning_rate": 1.1363865589267907e-05, + "loss": 0.9125, + "step": 38387 + }, + { + "epoch": 2.84, + "learning_rate": 1.136347044374102e-05, + "loss": 0.9853, + "step": 38388 + }, + { + "epoch": 2.84, + "learning_rate": 1.136307529604487e-05, + "loss": 1.1108, + "step": 38389 + }, + { + "epoch": 2.84, + "learning_rate": 1.1362680146180088e-05, + "loss": 0.8734, + "step": 38390 + }, + { + "epoch": 2.84, + "learning_rate": 1.1362284994147297e-05, + "loss": 0.8808, + "step": 38391 + }, + { + "epoch": 2.84, + "learning_rate": 1.1361889839947136e-05, + "loss": 0.9687, + "step": 38392 + }, + { + "epoch": 2.84, + "learning_rate": 1.1361494683580223e-05, + "loss": 1.0202, + "step": 38393 + }, + { + "epoch": 2.84, + "learning_rate": 1.1361099525047195e-05, + "loss": 1.0262, + "step": 38394 + }, + { + "epoch": 2.84, + "learning_rate": 1.1360704364348673e-05, + "loss": 1.0473, + "step": 38395 + }, + { + "epoch": 2.84, + "learning_rate": 1.1360309201485294e-05, + "loss": 1.0722, + "step": 38396 + }, + { + "epoch": 2.84, + "learning_rate": 1.135991403645768e-05, + "loss": 0.9863, + "step": 38397 + }, + { + "epoch": 2.84, + "learning_rate": 1.1359518869266464e-05, + "loss": 1.0659, + "step": 38398 + }, + { + "epoch": 2.84, + "learning_rate": 1.1359123699912271e-05, + "loss": 1.1886, + "step": 38399 + }, + { + "epoch": 2.84, + "learning_rate": 1.1358728528395732e-05, + "loss": 0.9659, + "step": 38400 + }, + { + "epoch": 2.84, + "learning_rate": 1.1358333354717477e-05, + "loss": 0.8229, + "step": 38401 + }, + { + "epoch": 2.84, + "learning_rate": 1.1357938178878135e-05, + "loss": 1.1003, + "step": 38402 + }, + { + "epoch": 2.84, + "learning_rate": 1.1357543000878328e-05, + "loss": 1.0114, + "step": 38403 + }, + { + "epoch": 2.84, + "learning_rate": 1.1357147820718692e-05, + "loss": 0.9708, + "step": 38404 + }, + { + "epoch": 2.84, + "learning_rate": 1.135675263839985e-05, + "loss": 0.9913, + "step": 38405 + }, + { + "epoch": 2.84, + "learning_rate": 1.1356357453922438e-05, + "loss": 0.9581, + "step": 38406 + }, + { + "epoch": 2.84, + "learning_rate": 1.1355962267287078e-05, + "loss": 1.0152, + "step": 38407 + }, + { + "epoch": 2.84, + "learning_rate": 1.1355567078494403e-05, + "loss": 1.0513, + "step": 38408 + }, + { + "epoch": 2.84, + "learning_rate": 1.1355171887545037e-05, + "loss": 0.9966, + "step": 38409 + }, + { + "epoch": 2.84, + "learning_rate": 1.1354776694439619e-05, + "loss": 1.0505, + "step": 38410 + }, + { + "epoch": 2.84, + "learning_rate": 1.1354381499178766e-05, + "loss": 0.9779, + "step": 38411 + }, + { + "epoch": 2.84, + "learning_rate": 1.135398630176311e-05, + "loss": 0.9538, + "step": 38412 + }, + { + "epoch": 2.84, + "learning_rate": 1.1353591102193285e-05, + "loss": 0.9264, + "step": 38413 + }, + { + "epoch": 2.84, + "learning_rate": 1.1353195900469913e-05, + "loss": 1.0565, + "step": 38414 + }, + { + "epoch": 2.84, + "learning_rate": 1.1352800696593627e-05, + "loss": 1.0775, + "step": 38415 + }, + { + "epoch": 2.84, + "learning_rate": 1.1352405490565054e-05, + "loss": 1.0163, + "step": 38416 + }, + { + "epoch": 2.84, + "learning_rate": 1.1352010282384823e-05, + "loss": 0.9588, + "step": 38417 + }, + { + "epoch": 2.84, + "learning_rate": 1.1351615072053563e-05, + "loss": 1.0232, + "step": 38418 + }, + { + "epoch": 2.84, + "learning_rate": 1.1351219859571902e-05, + "loss": 0.9224, + "step": 38419 + }, + { + "epoch": 2.84, + "learning_rate": 1.135082464494047e-05, + "loss": 1.1147, + "step": 38420 + }, + { + "epoch": 2.84, + "learning_rate": 1.1350429428159896e-05, + "loss": 0.9297, + "step": 38421 + }, + { + "epoch": 2.84, + "learning_rate": 1.1350034209230809e-05, + "loss": 0.9544, + "step": 38422 + }, + { + "epoch": 2.84, + "learning_rate": 1.1349638988153835e-05, + "loss": 0.9796, + "step": 38423 + }, + { + "epoch": 2.84, + "learning_rate": 1.1349243764929606e-05, + "loss": 0.9665, + "step": 38424 + }, + { + "epoch": 2.84, + "learning_rate": 1.134884853955875e-05, + "loss": 1.0188, + "step": 38425 + }, + { + "epoch": 2.84, + "learning_rate": 1.1348453312041894e-05, + "loss": 0.9225, + "step": 38426 + }, + { + "epoch": 2.84, + "learning_rate": 1.1348058082379667e-05, + "loss": 1.0316, + "step": 38427 + }, + { + "epoch": 2.84, + "learning_rate": 1.13476628505727e-05, + "loss": 0.9526, + "step": 38428 + }, + { + "epoch": 2.84, + "learning_rate": 1.1347267616621622e-05, + "loss": 0.9712, + "step": 38429 + }, + { + "epoch": 2.84, + "learning_rate": 1.134687238052706e-05, + "loss": 1.055, + "step": 38430 + }, + { + "epoch": 2.84, + "learning_rate": 1.1346477142289642e-05, + "loss": 0.9945, + "step": 38431 + }, + { + "epoch": 2.84, + "learning_rate": 1.1346081901909997e-05, + "loss": 1.0272, + "step": 38432 + }, + { + "epoch": 2.84, + "learning_rate": 1.134568665938876e-05, + "loss": 0.9868, + "step": 38433 + }, + { + "epoch": 2.84, + "learning_rate": 1.1345291414726548e-05, + "loss": 1.0244, + "step": 38434 + }, + { + "epoch": 2.84, + "learning_rate": 1.1344896167924002e-05, + "loss": 0.9345, + "step": 38435 + }, + { + "epoch": 2.84, + "learning_rate": 1.1344500918981743e-05, + "loss": 0.8986, + "step": 38436 + }, + { + "epoch": 2.84, + "learning_rate": 1.1344105667900403e-05, + "loss": 0.9646, + "step": 38437 + }, + { + "epoch": 2.84, + "learning_rate": 1.134371041468061e-05, + "loss": 1.0162, + "step": 38438 + }, + { + "epoch": 2.84, + "learning_rate": 1.1343315159322996e-05, + "loss": 1.1221, + "step": 38439 + }, + { + "epoch": 2.84, + "learning_rate": 1.1342919901828183e-05, + "loss": 0.9325, + "step": 38440 + }, + { + "epoch": 2.84, + "learning_rate": 1.1342524642196804e-05, + "loss": 1.0498, + "step": 38441 + }, + { + "epoch": 2.84, + "learning_rate": 1.1342129380429485e-05, + "loss": 1.0452, + "step": 38442 + }, + { + "epoch": 2.84, + "learning_rate": 1.1341734116526863e-05, + "loss": 0.9353, + "step": 38443 + }, + { + "epoch": 2.84, + "learning_rate": 1.1341338850489557e-05, + "loss": 1.0061, + "step": 38444 + }, + { + "epoch": 2.84, + "learning_rate": 1.1340943582318202e-05, + "loss": 0.9827, + "step": 38445 + }, + { + "epoch": 2.84, + "learning_rate": 1.1340548312013425e-05, + "loss": 1.1308, + "step": 38446 + }, + { + "epoch": 2.84, + "learning_rate": 1.1340153039575854e-05, + "loss": 0.9989, + "step": 38447 + }, + { + "epoch": 2.84, + "learning_rate": 1.133975776500612e-05, + "loss": 1.0927, + "step": 38448 + }, + { + "epoch": 2.84, + "learning_rate": 1.1339362488304851e-05, + "loss": 0.9273, + "step": 38449 + }, + { + "epoch": 2.84, + "learning_rate": 1.1338967209472672e-05, + "loss": 0.9665, + "step": 38450 + }, + { + "epoch": 2.84, + "learning_rate": 1.1338571928510218e-05, + "loss": 1.0274, + "step": 38451 + }, + { + "epoch": 2.84, + "learning_rate": 1.1338176645418113e-05, + "loss": 1.0293, + "step": 38452 + }, + { + "epoch": 2.84, + "learning_rate": 1.1337781360196991e-05, + "loss": 0.9507, + "step": 38453 + }, + { + "epoch": 2.84, + "learning_rate": 1.1337386072847475e-05, + "loss": 0.9459, + "step": 38454 + }, + { + "epoch": 2.84, + "learning_rate": 1.13369907833702e-05, + "loss": 1.0748, + "step": 38455 + }, + { + "epoch": 2.84, + "learning_rate": 1.1336595491765788e-05, + "loss": 1.0442, + "step": 38456 + }, + { + "epoch": 2.84, + "learning_rate": 1.1336200198034877e-05, + "loss": 0.9476, + "step": 38457 + }, + { + "epoch": 2.84, + "learning_rate": 1.1335804902178087e-05, + "loss": 1.0493, + "step": 38458 + }, + { + "epoch": 2.84, + "learning_rate": 1.1335409604196053e-05, + "loss": 0.9598, + "step": 38459 + }, + { + "epoch": 2.84, + "learning_rate": 1.1335014304089398e-05, + "loss": 0.9884, + "step": 38460 + }, + { + "epoch": 2.84, + "learning_rate": 1.1334619001858755e-05, + "loss": 1.0532, + "step": 38461 + }, + { + "epoch": 2.84, + "learning_rate": 1.1334223697504754e-05, + "loss": 1.0634, + "step": 38462 + }, + { + "epoch": 2.84, + "learning_rate": 1.1333828391028022e-05, + "loss": 0.8469, + "step": 38463 + }, + { + "epoch": 2.84, + "learning_rate": 1.1333433082429187e-05, + "loss": 1.0462, + "step": 38464 + }, + { + "epoch": 2.84, + "learning_rate": 1.133303777170888e-05, + "loss": 1.0005, + "step": 38465 + }, + { + "epoch": 2.84, + "learning_rate": 1.1332642458867728e-05, + "loss": 1.1939, + "step": 38466 + }, + { + "epoch": 2.84, + "learning_rate": 1.1332247143906361e-05, + "loss": 0.9973, + "step": 38467 + }, + { + "epoch": 2.84, + "learning_rate": 1.133185182682541e-05, + "loss": 1.0508, + "step": 38468 + }, + { + "epoch": 2.84, + "learning_rate": 1.13314565076255e-05, + "loss": 1.0241, + "step": 38469 + }, + { + "epoch": 2.84, + "learning_rate": 1.1331061186307259e-05, + "loss": 1.1307, + "step": 38470 + }, + { + "epoch": 2.84, + "learning_rate": 1.1330665862871324e-05, + "loss": 0.9105, + "step": 38471 + }, + { + "epoch": 2.84, + "learning_rate": 1.1330270537318313e-05, + "loss": 0.8719, + "step": 38472 + }, + { + "epoch": 2.84, + "learning_rate": 1.1329875209648867e-05, + "loss": 0.9567, + "step": 38473 + }, + { + "epoch": 2.84, + "learning_rate": 1.1329479879863604e-05, + "loss": 0.9998, + "step": 38474 + }, + { + "epoch": 2.84, + "learning_rate": 1.1329084547963158e-05, + "loss": 0.9024, + "step": 38475 + }, + { + "epoch": 2.84, + "learning_rate": 1.1328689213948161e-05, + "loss": 0.9492, + "step": 38476 + }, + { + "epoch": 2.84, + "learning_rate": 1.1328293877819234e-05, + "loss": 1.0311, + "step": 38477 + }, + { + "epoch": 2.84, + "learning_rate": 1.1327898539577014e-05, + "loss": 1.0098, + "step": 38478 + }, + { + "epoch": 2.84, + "learning_rate": 1.1327503199222123e-05, + "loss": 1.0247, + "step": 38479 + }, + { + "epoch": 2.84, + "learning_rate": 1.1327107856755194e-05, + "loss": 0.9911, + "step": 38480 + }, + { + "epoch": 2.84, + "learning_rate": 1.1326712512176856e-05, + "loss": 0.8615, + "step": 38481 + }, + { + "epoch": 2.84, + "learning_rate": 1.1326317165487739e-05, + "loss": 1.0732, + "step": 38482 + }, + { + "epoch": 2.84, + "learning_rate": 1.1325921816688468e-05, + "loss": 1.0036, + "step": 38483 + }, + { + "epoch": 2.84, + "learning_rate": 1.1325526465779677e-05, + "loss": 1.0638, + "step": 38484 + }, + { + "epoch": 2.84, + "learning_rate": 1.132513111276199e-05, + "loss": 0.9254, + "step": 38485 + }, + { + "epoch": 2.84, + "learning_rate": 1.132473575763604e-05, + "loss": 1.0243, + "step": 38486 + }, + { + "epoch": 2.84, + "learning_rate": 1.1324340400402454e-05, + "loss": 1.1334, + "step": 38487 + }, + { + "epoch": 2.84, + "learning_rate": 1.1323945041061862e-05, + "loss": 1.0106, + "step": 38488 + }, + { + "epoch": 2.84, + "learning_rate": 1.1323549679614891e-05, + "loss": 1.0993, + "step": 38489 + }, + { + "epoch": 2.84, + "learning_rate": 1.1323154316062173e-05, + "loss": 0.9576, + "step": 38490 + }, + { + "epoch": 2.84, + "learning_rate": 1.1322758950404331e-05, + "loss": 1.0962, + "step": 38491 + }, + { + "epoch": 2.84, + "learning_rate": 1.1322363582642005e-05, + "loss": 0.9613, + "step": 38492 + }, + { + "epoch": 2.84, + "learning_rate": 1.1321968212775812e-05, + "loss": 0.9968, + "step": 38493 + }, + { + "epoch": 2.84, + "learning_rate": 1.1321572840806391e-05, + "loss": 1.0678, + "step": 38494 + }, + { + "epoch": 2.84, + "learning_rate": 1.1321177466734363e-05, + "loss": 1.042, + "step": 38495 + }, + { + "epoch": 2.84, + "learning_rate": 1.1320782090560362e-05, + "loss": 1.0078, + "step": 38496 + }, + { + "epoch": 2.84, + "learning_rate": 1.1320386712285014e-05, + "loss": 1.0186, + "step": 38497 + }, + { + "epoch": 2.84, + "learning_rate": 1.1319991331908953e-05, + "loss": 0.9891, + "step": 38498 + }, + { + "epoch": 2.84, + "learning_rate": 1.13195959494328e-05, + "loss": 1.0355, + "step": 38499 + }, + { + "epoch": 2.84, + "learning_rate": 1.1319200564857194e-05, + "loss": 1.0162, + "step": 38500 + }, + { + "epoch": 2.84, + "learning_rate": 1.1318805178182755e-05, + "loss": 1.0087, + "step": 38501 + }, + { + "epoch": 2.84, + "learning_rate": 1.1318409789410117e-05, + "loss": 1.0257, + "step": 38502 + }, + { + "epoch": 2.84, + "learning_rate": 1.1318014398539905e-05, + "loss": 0.9737, + "step": 38503 + }, + { + "epoch": 2.85, + "learning_rate": 1.1317619005572757e-05, + "loss": 1.0017, + "step": 38504 + }, + { + "epoch": 2.85, + "learning_rate": 1.1317223610509292e-05, + "loss": 1.0103, + "step": 38505 + }, + { + "epoch": 2.85, + "learning_rate": 1.1316828213350143e-05, + "loss": 0.9683, + "step": 38506 + }, + { + "epoch": 2.85, + "learning_rate": 1.131643281409594e-05, + "loss": 1.1019, + "step": 38507 + }, + { + "epoch": 2.85, + "learning_rate": 1.1316037412747315e-05, + "loss": 0.9497, + "step": 38508 + }, + { + "epoch": 2.85, + "learning_rate": 1.1315642009304886e-05, + "loss": 1.0131, + "step": 38509 + }, + { + "epoch": 2.85, + "learning_rate": 1.1315246603769295e-05, + "loss": 1.0215, + "step": 38510 + }, + { + "epoch": 2.85, + "learning_rate": 1.1314851196141162e-05, + "loss": 1.0491, + "step": 38511 + }, + { + "epoch": 2.85, + "learning_rate": 1.1314455786421121e-05, + "loss": 0.9897, + "step": 38512 + }, + { + "epoch": 2.85, + "learning_rate": 1.13140603746098e-05, + "loss": 0.9702, + "step": 38513 + }, + { + "epoch": 2.85, + "learning_rate": 1.1313664960707828e-05, + "loss": 0.9933, + "step": 38514 + }, + { + "epoch": 2.85, + "learning_rate": 1.1313269544715832e-05, + "loss": 1.0685, + "step": 38515 + }, + { + "epoch": 2.85, + "learning_rate": 1.1312874126634445e-05, + "loss": 0.9244, + "step": 38516 + }, + { + "epoch": 2.85, + "learning_rate": 1.1312478706464294e-05, + "loss": 0.9113, + "step": 38517 + }, + { + "epoch": 2.85, + "learning_rate": 1.1312083284206008e-05, + "loss": 0.9974, + "step": 38518 + }, + { + "epoch": 2.85, + "learning_rate": 1.1311687859860214e-05, + "loss": 1.0453, + "step": 38519 + }, + { + "epoch": 2.85, + "learning_rate": 1.1311292433427545e-05, + "loss": 0.9773, + "step": 38520 + }, + { + "epoch": 2.85, + "learning_rate": 1.1310897004908628e-05, + "loss": 1.0139, + "step": 38521 + }, + { + "epoch": 2.85, + "learning_rate": 1.1310501574304094e-05, + "loss": 1.0, + "step": 38522 + }, + { + "epoch": 2.85, + "learning_rate": 1.131010614161457e-05, + "loss": 1.0574, + "step": 38523 + }, + { + "epoch": 2.85, + "learning_rate": 1.1309710706840684e-05, + "loss": 0.9563, + "step": 38524 + }, + { + "epoch": 2.85, + "learning_rate": 1.130931526998307e-05, + "loss": 0.8906, + "step": 38525 + }, + { + "epoch": 2.85, + "learning_rate": 1.130891983104235e-05, + "loss": 0.9915, + "step": 38526 + }, + { + "epoch": 2.85, + "learning_rate": 1.1308524390019162e-05, + "loss": 1.1373, + "step": 38527 + }, + { + "epoch": 2.85, + "learning_rate": 1.1308128946914128e-05, + "loss": 1.1075, + "step": 38528 + }, + { + "epoch": 2.85, + "learning_rate": 1.1307733501727883e-05, + "loss": 1.0278, + "step": 38529 + }, + { + "epoch": 2.85, + "learning_rate": 1.130733805446105e-05, + "loss": 0.9552, + "step": 38530 + }, + { + "epoch": 2.85, + "learning_rate": 1.130694260511426e-05, + "loss": 0.987, + "step": 38531 + }, + { + "epoch": 2.85, + "learning_rate": 1.1306547153688143e-05, + "loss": 1.0107, + "step": 38532 + }, + { + "epoch": 2.85, + "learning_rate": 1.130615170018333e-05, + "loss": 0.9865, + "step": 38533 + }, + { + "epoch": 2.85, + "learning_rate": 1.1305756244600446e-05, + "loss": 1.0258, + "step": 38534 + }, + { + "epoch": 2.85, + "learning_rate": 1.1305360786940124e-05, + "loss": 1.0639, + "step": 38535 + }, + { + "epoch": 2.85, + "learning_rate": 1.130496532720299e-05, + "loss": 0.8347, + "step": 38536 + }, + { + "epoch": 2.85, + "learning_rate": 1.1304569865389678e-05, + "loss": 1.0625, + "step": 38537 + }, + { + "epoch": 2.85, + "learning_rate": 1.1304174401500812e-05, + "loss": 0.9739, + "step": 38538 + }, + { + "epoch": 2.85, + "learning_rate": 1.1303778935537024e-05, + "loss": 1.0304, + "step": 38539 + }, + { + "epoch": 2.85, + "learning_rate": 1.1303383467498938e-05, + "loss": 0.9431, + "step": 38540 + }, + { + "epoch": 2.85, + "learning_rate": 1.1302987997387195e-05, + "loss": 0.9443, + "step": 38541 + }, + { + "epoch": 2.85, + "learning_rate": 1.1302592525202414e-05, + "loss": 1.0268, + "step": 38542 + }, + { + "epoch": 2.85, + "learning_rate": 1.1302197050945225e-05, + "loss": 0.9734, + "step": 38543 + }, + { + "epoch": 2.85, + "learning_rate": 1.130180157461626e-05, + "loss": 0.9869, + "step": 38544 + }, + { + "epoch": 2.85, + "learning_rate": 1.1301406096216148e-05, + "loss": 1.0429, + "step": 38545 + }, + { + "epoch": 2.85, + "learning_rate": 1.1301010615745515e-05, + "loss": 1.0374, + "step": 38546 + }, + { + "epoch": 2.85, + "learning_rate": 1.1300615133204995e-05, + "loss": 0.987, + "step": 38547 + }, + { + "epoch": 2.85, + "learning_rate": 1.1300219648595214e-05, + "loss": 1.0264, + "step": 38548 + }, + { + "epoch": 2.85, + "learning_rate": 1.1299824161916804e-05, + "loss": 1.0767, + "step": 38549 + }, + { + "epoch": 2.85, + "learning_rate": 1.1299428673170389e-05, + "loss": 0.9523, + "step": 38550 + }, + { + "epoch": 2.85, + "learning_rate": 1.1299033182356607e-05, + "loss": 1.0163, + "step": 38551 + }, + { + "epoch": 2.85, + "learning_rate": 1.1298637689476078e-05, + "loss": 1.01, + "step": 38552 + }, + { + "epoch": 2.85, + "learning_rate": 1.1298242194529435e-05, + "loss": 0.9399, + "step": 38553 + }, + { + "epoch": 2.85, + "learning_rate": 1.1297846697517307e-05, + "loss": 0.9945, + "step": 38554 + }, + { + "epoch": 2.85, + "learning_rate": 1.1297451198440325e-05, + "loss": 0.9859, + "step": 38555 + }, + { + "epoch": 2.85, + "learning_rate": 1.1297055697299116e-05, + "loss": 0.9969, + "step": 38556 + }, + { + "epoch": 2.85, + "learning_rate": 1.129666019409431e-05, + "loss": 1.0913, + "step": 38557 + }, + { + "epoch": 2.85, + "learning_rate": 1.1296264688826534e-05, + "loss": 0.9389, + "step": 38558 + }, + { + "epoch": 2.85, + "learning_rate": 1.1295869181496424e-05, + "loss": 0.9947, + "step": 38559 + }, + { + "epoch": 2.85, + "learning_rate": 1.1295473672104603e-05, + "loss": 0.9762, + "step": 38560 + }, + { + "epoch": 2.85, + "learning_rate": 1.1295078160651702e-05, + "loss": 0.955, + "step": 38561 + }, + { + "epoch": 2.85, + "learning_rate": 1.1294682647138348e-05, + "loss": 0.9787, + "step": 38562 + }, + { + "epoch": 2.85, + "learning_rate": 1.1294287131565173e-05, + "loss": 0.9626, + "step": 38563 + }, + { + "epoch": 2.85, + "learning_rate": 1.1293891613932806e-05, + "loss": 1.0466, + "step": 38564 + }, + { + "epoch": 2.85, + "learning_rate": 1.1293496094241879e-05, + "loss": 0.9959, + "step": 38565 + }, + { + "epoch": 2.85, + "learning_rate": 1.1293100572493012e-05, + "loss": 1.1571, + "step": 38566 + }, + { + "epoch": 2.85, + "learning_rate": 1.1292705048686847e-05, + "loss": 1.0839, + "step": 38567 + }, + { + "epoch": 2.85, + "learning_rate": 1.1292309522824e-05, + "loss": 0.9324, + "step": 38568 + }, + { + "epoch": 2.85, + "learning_rate": 1.1291913994905113e-05, + "loss": 1.0133, + "step": 38569 + }, + { + "epoch": 2.85, + "learning_rate": 1.1291518464930809e-05, + "loss": 0.9875, + "step": 38570 + }, + { + "epoch": 2.85, + "learning_rate": 1.1291122932901715e-05, + "loss": 0.9716, + "step": 38571 + }, + { + "epoch": 2.85, + "learning_rate": 1.1290727398818464e-05, + "loss": 0.9337, + "step": 38572 + }, + { + "epoch": 2.85, + "learning_rate": 1.1290331862681684e-05, + "loss": 1.0164, + "step": 38573 + }, + { + "epoch": 2.85, + "learning_rate": 1.1289936324492003e-05, + "loss": 1.1128, + "step": 38574 + }, + { + "epoch": 2.85, + "learning_rate": 1.1289540784250054e-05, + "loss": 0.952, + "step": 38575 + }, + { + "epoch": 2.85, + "learning_rate": 1.1289145241956462e-05, + "loss": 1.0426, + "step": 38576 + }, + { + "epoch": 2.85, + "learning_rate": 1.1288749697611861e-05, + "loss": 1.003, + "step": 38577 + }, + { + "epoch": 2.85, + "learning_rate": 1.1288354151216873e-05, + "loss": 0.9994, + "step": 38578 + }, + { + "epoch": 2.85, + "learning_rate": 1.1287958602772138e-05, + "loss": 1.0393, + "step": 38579 + }, + { + "epoch": 2.85, + "learning_rate": 1.1287563052278276e-05, + "loss": 0.9733, + "step": 38580 + }, + { + "epoch": 2.85, + "learning_rate": 1.1287167499735921e-05, + "loss": 1.0273, + "step": 38581 + }, + { + "epoch": 2.85, + "learning_rate": 1.1286771945145701e-05, + "loss": 1.0784, + "step": 38582 + }, + { + "epoch": 2.85, + "learning_rate": 1.1286376388508243e-05, + "loss": 0.8963, + "step": 38583 + }, + { + "epoch": 2.85, + "learning_rate": 1.1285980829824179e-05, + "loss": 0.9221, + "step": 38584 + }, + { + "epoch": 2.85, + "learning_rate": 1.1285585269094137e-05, + "loss": 1.0524, + "step": 38585 + }, + { + "epoch": 2.85, + "learning_rate": 1.1285189706318752e-05, + "loss": 0.9102, + "step": 38586 + }, + { + "epoch": 2.85, + "learning_rate": 1.1284794141498642e-05, + "loss": 1.0373, + "step": 38587 + }, + { + "epoch": 2.85, + "learning_rate": 1.128439857463445e-05, + "loss": 0.9493, + "step": 38588 + }, + { + "epoch": 2.85, + "learning_rate": 1.1284003005726794e-05, + "loss": 1.0515, + "step": 38589 + }, + { + "epoch": 2.85, + "learning_rate": 1.1283607434776309e-05, + "loss": 1.0739, + "step": 38590 + }, + { + "epoch": 2.85, + "learning_rate": 1.128321186178362e-05, + "loss": 1.0497, + "step": 38591 + }, + { + "epoch": 2.85, + "learning_rate": 1.1282816286749362e-05, + "loss": 0.9899, + "step": 38592 + }, + { + "epoch": 2.85, + "learning_rate": 1.128242070967416e-05, + "loss": 1.0917, + "step": 38593 + }, + { + "epoch": 2.85, + "learning_rate": 1.1282025130558646e-05, + "loss": 0.9421, + "step": 38594 + }, + { + "epoch": 2.85, + "learning_rate": 1.1281629549403449e-05, + "loss": 0.9609, + "step": 38595 + }, + { + "epoch": 2.85, + "learning_rate": 1.1281233966209197e-05, + "loss": 1.0694, + "step": 38596 + }, + { + "epoch": 2.85, + "learning_rate": 1.1280838380976522e-05, + "loss": 0.9774, + "step": 38597 + }, + { + "epoch": 2.85, + "learning_rate": 1.128044279370605e-05, + "loss": 0.983, + "step": 38598 + }, + { + "epoch": 2.85, + "learning_rate": 1.128004720439841e-05, + "loss": 1.0049, + "step": 38599 + }, + { + "epoch": 2.85, + "learning_rate": 1.1279651613054237e-05, + "loss": 0.8682, + "step": 38600 + }, + { + "epoch": 2.85, + "learning_rate": 1.1279256019674151e-05, + "loss": 1.0688, + "step": 38601 + }, + { + "epoch": 2.85, + "learning_rate": 1.1278860424258793e-05, + "loss": 1.0725, + "step": 38602 + }, + { + "epoch": 2.85, + "learning_rate": 1.127846482680878e-05, + "loss": 0.9873, + "step": 38603 + }, + { + "epoch": 2.85, + "learning_rate": 1.1278069227324756e-05, + "loss": 1.0883, + "step": 38604 + }, + { + "epoch": 2.85, + "learning_rate": 1.1277673625807335e-05, + "loss": 1.0057, + "step": 38605 + }, + { + "epoch": 2.85, + "learning_rate": 1.1277278022257157e-05, + "loss": 1.016, + "step": 38606 + }, + { + "epoch": 2.85, + "learning_rate": 1.1276882416674847e-05, + "loss": 1.036, + "step": 38607 + }, + { + "epoch": 2.85, + "learning_rate": 1.1276486809061037e-05, + "loss": 0.9546, + "step": 38608 + }, + { + "epoch": 2.85, + "learning_rate": 1.1276091199416353e-05, + "loss": 1.0046, + "step": 38609 + }, + { + "epoch": 2.85, + "learning_rate": 1.1275695587741426e-05, + "loss": 0.9757, + "step": 38610 + }, + { + "epoch": 2.85, + "learning_rate": 1.1275299974036884e-05, + "loss": 1.009, + "step": 38611 + }, + { + "epoch": 2.85, + "learning_rate": 1.1274904358303362e-05, + "loss": 1.1058, + "step": 38612 + }, + { + "epoch": 2.85, + "learning_rate": 1.127450874054148e-05, + "loss": 1.0832, + "step": 38613 + }, + { + "epoch": 2.85, + "learning_rate": 1.127411312075188e-05, + "loss": 0.9547, + "step": 38614 + }, + { + "epoch": 2.85, + "learning_rate": 1.1273717498935178e-05, + "loss": 1.0267, + "step": 38615 + }, + { + "epoch": 2.85, + "learning_rate": 1.1273321875092013e-05, + "loss": 1.0119, + "step": 38616 + }, + { + "epoch": 2.85, + "learning_rate": 1.127292624922301e-05, + "loss": 0.9503, + "step": 38617 + }, + { + "epoch": 2.85, + "learning_rate": 1.1272530621328799e-05, + "loss": 0.9047, + "step": 38618 + }, + { + "epoch": 2.85, + "learning_rate": 1.1272134991410013e-05, + "loss": 0.9709, + "step": 38619 + }, + { + "epoch": 2.85, + "learning_rate": 1.1271739359467275e-05, + "loss": 0.9918, + "step": 38620 + }, + { + "epoch": 2.85, + "learning_rate": 1.1271343725501218e-05, + "loss": 0.9601, + "step": 38621 + }, + { + "epoch": 2.85, + "learning_rate": 1.1270948089512474e-05, + "loss": 1.0718, + "step": 38622 + }, + { + "epoch": 2.85, + "learning_rate": 1.1270552451501667e-05, + "loss": 0.9714, + "step": 38623 + }, + { + "epoch": 2.85, + "learning_rate": 1.1270156811469431e-05, + "loss": 0.943, + "step": 38624 + }, + { + "epoch": 2.85, + "learning_rate": 1.1269761169416393e-05, + "loss": 1.0681, + "step": 38625 + }, + { + "epoch": 2.85, + "learning_rate": 1.1269365525343183e-05, + "loss": 1.014, + "step": 38626 + }, + { + "epoch": 2.85, + "learning_rate": 1.1268969879250432e-05, + "loss": 1.1045, + "step": 38627 + }, + { + "epoch": 2.85, + "learning_rate": 1.1268574231138765e-05, + "loss": 0.9534, + "step": 38628 + }, + { + "epoch": 2.85, + "learning_rate": 1.1268178581008818e-05, + "loss": 0.9837, + "step": 38629 + }, + { + "epoch": 2.85, + "learning_rate": 1.1267782928861212e-05, + "loss": 1.1093, + "step": 38630 + }, + { + "epoch": 2.85, + "learning_rate": 1.1267387274696589e-05, + "loss": 1.0784, + "step": 38631 + }, + { + "epoch": 2.85, + "learning_rate": 1.1266991618515563e-05, + "loss": 1.0532, + "step": 38632 + }, + { + "epoch": 2.85, + "learning_rate": 1.1266595960318779e-05, + "loss": 1.0511, + "step": 38633 + }, + { + "epoch": 2.85, + "learning_rate": 1.1266200300106854e-05, + "loss": 0.9814, + "step": 38634 + }, + { + "epoch": 2.85, + "learning_rate": 1.1265804637880423e-05, + "loss": 1.0023, + "step": 38635 + }, + { + "epoch": 2.85, + "learning_rate": 1.1265408973640118e-05, + "loss": 1.0833, + "step": 38636 + }, + { + "epoch": 2.85, + "learning_rate": 1.1265013307386563e-05, + "loss": 0.9402, + "step": 38637 + }, + { + "epoch": 2.85, + "learning_rate": 1.1264617639120388e-05, + "loss": 1.16, + "step": 38638 + }, + { + "epoch": 2.86, + "learning_rate": 1.1264221968842228e-05, + "loss": 1.0812, + "step": 38639 + }, + { + "epoch": 2.86, + "learning_rate": 1.1263826296552706e-05, + "loss": 0.9134, + "step": 38640 + }, + { + "epoch": 2.86, + "learning_rate": 1.1263430622252456e-05, + "loss": 1.0286, + "step": 38641 + }, + { + "epoch": 2.86, + "learning_rate": 1.1263034945942106e-05, + "loss": 1.0056, + "step": 38642 + }, + { + "epoch": 2.86, + "learning_rate": 1.1262639267622287e-05, + "loss": 0.8939, + "step": 38643 + }, + { + "epoch": 2.86, + "learning_rate": 1.1262243587293624e-05, + "loss": 0.9765, + "step": 38644 + }, + { + "epoch": 2.86, + "learning_rate": 1.1261847904956753e-05, + "loss": 0.9768, + "step": 38645 + }, + { + "epoch": 2.86, + "learning_rate": 1.1261452220612297e-05, + "loss": 1.0104, + "step": 38646 + }, + { + "epoch": 2.86, + "learning_rate": 1.1261056534260891e-05, + "loss": 1.0093, + "step": 38647 + }, + { + "epoch": 2.86, + "learning_rate": 1.1260660845903157e-05, + "loss": 1.0441, + "step": 38648 + }, + { + "epoch": 2.86, + "learning_rate": 1.1260265155539734e-05, + "loss": 0.9587, + "step": 38649 + }, + { + "epoch": 2.86, + "learning_rate": 1.1259869463171245e-05, + "loss": 0.998, + "step": 38650 + }, + { + "epoch": 2.86, + "learning_rate": 1.1259473768798324e-05, + "loss": 0.9494, + "step": 38651 + }, + { + "epoch": 2.86, + "learning_rate": 1.1259078072421597e-05, + "loss": 1.0241, + "step": 38652 + }, + { + "epoch": 2.86, + "learning_rate": 1.1258682374041698e-05, + "loss": 1.0071, + "step": 38653 + }, + { + "epoch": 2.86, + "learning_rate": 1.1258286673659249e-05, + "loss": 1.0211, + "step": 38654 + }, + { + "epoch": 2.86, + "learning_rate": 1.1257890971274885e-05, + "loss": 0.9713, + "step": 38655 + }, + { + "epoch": 2.86, + "learning_rate": 1.1257495266889236e-05, + "loss": 1.1938, + "step": 38656 + }, + { + "epoch": 2.86, + "learning_rate": 1.125709956050293e-05, + "loss": 1.0818, + "step": 38657 + }, + { + "epoch": 2.86, + "learning_rate": 1.1256703852116593e-05, + "loss": 0.9478, + "step": 38658 + }, + { + "epoch": 2.86, + "learning_rate": 1.1256308141730861e-05, + "loss": 0.9915, + "step": 38659 + }, + { + "epoch": 2.86, + "learning_rate": 1.1255912429346359e-05, + "loss": 1.054, + "step": 38660 + }, + { + "epoch": 2.86, + "learning_rate": 1.125551671496372e-05, + "loss": 0.9603, + "step": 38661 + }, + { + "epoch": 2.86, + "learning_rate": 1.125512099858357e-05, + "loss": 0.9339, + "step": 38662 + }, + { + "epoch": 2.86, + "learning_rate": 1.1254725280206545e-05, + "loss": 0.9293, + "step": 38663 + }, + { + "epoch": 2.86, + "learning_rate": 1.1254329559833266e-05, + "loss": 0.9916, + "step": 38664 + }, + { + "epoch": 2.86, + "learning_rate": 1.1253933837464368e-05, + "loss": 1.0689, + "step": 38665 + }, + { + "epoch": 2.86, + "learning_rate": 1.1253538113100478e-05, + "loss": 0.9227, + "step": 38666 + }, + { + "epoch": 2.86, + "learning_rate": 1.1253142386742229e-05, + "loss": 1.0346, + "step": 38667 + }, + { + "epoch": 2.86, + "learning_rate": 1.1252746658390245e-05, + "loss": 1.0924, + "step": 38668 + }, + { + "epoch": 2.86, + "learning_rate": 1.1252350928045161e-05, + "loss": 1.0756, + "step": 38669 + }, + { + "epoch": 2.86, + "learning_rate": 1.1251955195707605e-05, + "loss": 0.8901, + "step": 38670 + }, + { + "epoch": 2.86, + "learning_rate": 1.1251559461378205e-05, + "loss": 1.0228, + "step": 38671 + }, + { + "epoch": 2.86, + "learning_rate": 1.1251163725057594e-05, + "loss": 1.0465, + "step": 38672 + }, + { + "epoch": 2.86, + "learning_rate": 1.1250767986746396e-05, + "loss": 1.0201, + "step": 38673 + }, + { + "epoch": 2.86, + "learning_rate": 1.1250372246445247e-05, + "loss": 1.0028, + "step": 38674 + }, + { + "epoch": 2.86, + "learning_rate": 1.1249976504154771e-05, + "loss": 0.9221, + "step": 38675 + }, + { + "epoch": 2.86, + "learning_rate": 1.1249580759875602e-05, + "loss": 1.0432, + "step": 38676 + }, + { + "epoch": 2.86, + "learning_rate": 1.1249185013608368e-05, + "loss": 0.9792, + "step": 38677 + }, + { + "epoch": 2.86, + "learning_rate": 1.1248789265353699e-05, + "loss": 1.0153, + "step": 38678 + }, + { + "epoch": 2.86, + "learning_rate": 1.1248393515112221e-05, + "loss": 1.0744, + "step": 38679 + }, + { + "epoch": 2.86, + "learning_rate": 1.124799776288457e-05, + "loss": 0.9739, + "step": 38680 + }, + { + "epoch": 2.86, + "learning_rate": 1.1247602008671374e-05, + "loss": 0.9309, + "step": 38681 + }, + { + "epoch": 2.86, + "learning_rate": 1.124720625247326e-05, + "loss": 1.0132, + "step": 38682 + }, + { + "epoch": 2.86, + "learning_rate": 1.1246810494290856e-05, + "loss": 0.9998, + "step": 38683 + }, + { + "epoch": 2.86, + "learning_rate": 1.1246414734124795e-05, + "loss": 1.1735, + "step": 38684 + }, + { + "epoch": 2.86, + "learning_rate": 1.1246018971975707e-05, + "loss": 0.9791, + "step": 38685 + }, + { + "epoch": 2.86, + "learning_rate": 1.124562320784422e-05, + "loss": 1.0019, + "step": 38686 + }, + { + "epoch": 2.86, + "learning_rate": 1.1245227441730964e-05, + "loss": 0.9644, + "step": 38687 + }, + { + "epoch": 2.86, + "learning_rate": 1.1244831673636571e-05, + "loss": 0.9416, + "step": 38688 + }, + { + "epoch": 2.86, + "learning_rate": 1.1244435903561666e-05, + "loss": 0.9503, + "step": 38689 + }, + { + "epoch": 2.86, + "learning_rate": 1.1244040131506883e-05, + "loss": 1.0184, + "step": 38690 + }, + { + "epoch": 2.86, + "learning_rate": 1.124364435747285e-05, + "loss": 0.9982, + "step": 38691 + }, + { + "epoch": 2.86, + "learning_rate": 1.1243248581460197e-05, + "loss": 1.0089, + "step": 38692 + }, + { + "epoch": 2.86, + "learning_rate": 1.124285280346955e-05, + "loss": 0.986, + "step": 38693 + }, + { + "epoch": 2.86, + "learning_rate": 1.1242457023501548e-05, + "loss": 0.9438, + "step": 38694 + }, + { + "epoch": 2.86, + "learning_rate": 1.1242061241556807e-05, + "loss": 0.9931, + "step": 38695 + }, + { + "epoch": 2.86, + "learning_rate": 1.124166545763597e-05, + "loss": 1.0698, + "step": 38696 + }, + { + "epoch": 2.86, + "learning_rate": 1.1241269671739659e-05, + "loss": 0.9674, + "step": 38697 + }, + { + "epoch": 2.86, + "learning_rate": 1.1240873883868507e-05, + "loss": 0.9285, + "step": 38698 + }, + { + "epoch": 2.86, + "learning_rate": 1.1240478094023139e-05, + "loss": 1.0362, + "step": 38699 + }, + { + "epoch": 2.86, + "learning_rate": 1.1240082302204193e-05, + "loss": 1.0839, + "step": 38700 + }, + { + "epoch": 2.86, + "learning_rate": 1.123968650841229e-05, + "loss": 1.093, + "step": 38701 + }, + { + "epoch": 2.86, + "learning_rate": 1.1239290712648067e-05, + "loss": 0.9504, + "step": 38702 + }, + { + "epoch": 2.86, + "learning_rate": 1.1238894914912147e-05, + "loss": 1.0971, + "step": 38703 + }, + { + "epoch": 2.86, + "learning_rate": 1.1238499115205164e-05, + "loss": 1.0103, + "step": 38704 + }, + { + "epoch": 2.86, + "learning_rate": 1.1238103313527746e-05, + "loss": 0.9876, + "step": 38705 + }, + { + "epoch": 2.86, + "learning_rate": 1.1237707509880524e-05, + "loss": 0.9938, + "step": 38706 + }, + { + "epoch": 2.86, + "learning_rate": 1.1237311704264126e-05, + "loss": 1.0558, + "step": 38707 + }, + { + "epoch": 2.86, + "learning_rate": 1.1236915896679187e-05, + "loss": 1.1782, + "step": 38708 + }, + { + "epoch": 2.86, + "learning_rate": 1.1236520087126329e-05, + "loss": 1.0196, + "step": 38709 + }, + { + "epoch": 2.86, + "learning_rate": 1.1236124275606186e-05, + "loss": 1.0338, + "step": 38710 + }, + { + "epoch": 2.86, + "learning_rate": 1.1235728462119383e-05, + "loss": 1.0762, + "step": 38711 + }, + { + "epoch": 2.86, + "learning_rate": 1.123533264666656e-05, + "loss": 1.0248, + "step": 38712 + }, + { + "epoch": 2.86, + "learning_rate": 1.1234936829248337e-05, + "loss": 0.9099, + "step": 38713 + }, + { + "epoch": 2.86, + "learning_rate": 1.1234541009865348e-05, + "loss": 0.9936, + "step": 38714 + }, + { + "epoch": 2.86, + "learning_rate": 1.1234145188518219e-05, + "loss": 1.0232, + "step": 38715 + }, + { + "epoch": 2.86, + "learning_rate": 1.1233749365207587e-05, + "loss": 0.9606, + "step": 38716 + }, + { + "epoch": 2.86, + "learning_rate": 1.1233353539934076e-05, + "loss": 1.014, + "step": 38717 + }, + { + "epoch": 2.86, + "learning_rate": 1.1232957712698317e-05, + "loss": 1.0393, + "step": 38718 + }, + { + "epoch": 2.86, + "learning_rate": 1.1232561883500941e-05, + "loss": 0.9746, + "step": 38719 + }, + { + "epoch": 2.86, + "learning_rate": 1.1232166052342575e-05, + "loss": 1.0533, + "step": 38720 + }, + { + "epoch": 2.86, + "learning_rate": 1.1231770219223852e-05, + "loss": 1.0114, + "step": 38721 + }, + { + "epoch": 2.86, + "learning_rate": 1.1231374384145397e-05, + "loss": 1.0259, + "step": 38722 + }, + { + "epoch": 2.86, + "learning_rate": 1.1230978547107847e-05, + "loss": 0.8817, + "step": 38723 + }, + { + "epoch": 2.86, + "learning_rate": 1.1230582708111825e-05, + "loss": 1.0181, + "step": 38724 + }, + { + "epoch": 2.86, + "learning_rate": 1.1230186867157964e-05, + "loss": 0.9693, + "step": 38725 + }, + { + "epoch": 2.86, + "learning_rate": 1.1229791024246895e-05, + "loss": 1.1084, + "step": 38726 + }, + { + "epoch": 2.86, + "learning_rate": 1.1229395179379243e-05, + "loss": 0.9741, + "step": 38727 + }, + { + "epoch": 2.86, + "learning_rate": 1.1228999332555643e-05, + "loss": 1.0188, + "step": 38728 + }, + { + "epoch": 2.86, + "learning_rate": 1.1228603483776723e-05, + "loss": 1.106, + "step": 38729 + }, + { + "epoch": 2.86, + "learning_rate": 1.1228207633043112e-05, + "loss": 1.049, + "step": 38730 + }, + { + "epoch": 2.86, + "learning_rate": 1.122781178035544e-05, + "loss": 1.0462, + "step": 38731 + }, + { + "epoch": 2.86, + "learning_rate": 1.1227415925714335e-05, + "loss": 0.9597, + "step": 38732 + }, + { + "epoch": 2.86, + "learning_rate": 1.1227020069120434e-05, + "loss": 0.925, + "step": 38733 + }, + { + "epoch": 2.86, + "learning_rate": 1.1226624210574355e-05, + "loss": 1.0115, + "step": 38734 + }, + { + "epoch": 2.86, + "learning_rate": 1.122622835007674e-05, + "loss": 1.0166, + "step": 38735 + }, + { + "epoch": 2.86, + "learning_rate": 1.122583248762821e-05, + "loss": 0.8783, + "step": 38736 + }, + { + "epoch": 2.86, + "learning_rate": 1.12254366232294e-05, + "loss": 0.9566, + "step": 38737 + }, + { + "epoch": 2.86, + "learning_rate": 1.1225040756880938e-05, + "loss": 1.0016, + "step": 38738 + }, + { + "epoch": 2.86, + "learning_rate": 1.1224644888583454e-05, + "loss": 0.9294, + "step": 38739 + }, + { + "epoch": 2.86, + "learning_rate": 1.1224249018337575e-05, + "loss": 0.9082, + "step": 38740 + }, + { + "epoch": 2.86, + "learning_rate": 1.1223853146143935e-05, + "loss": 1.1329, + "step": 38741 + }, + { + "epoch": 2.86, + "learning_rate": 1.122345727200316e-05, + "loss": 1.1438, + "step": 38742 + }, + { + "epoch": 2.86, + "learning_rate": 1.1223061395915884e-05, + "loss": 1.0002, + "step": 38743 + }, + { + "epoch": 2.86, + "learning_rate": 1.1222665517882733e-05, + "loss": 0.8519, + "step": 38744 + }, + { + "epoch": 2.86, + "learning_rate": 1.1222269637904342e-05, + "loss": 0.9777, + "step": 38745 + }, + { + "epoch": 2.86, + "learning_rate": 1.1221873755981333e-05, + "loss": 1.0344, + "step": 38746 + }, + { + "epoch": 2.86, + "learning_rate": 1.1221477872114348e-05, + "loss": 0.9888, + "step": 38747 + }, + { + "epoch": 2.86, + "learning_rate": 1.1221081986304002e-05, + "loss": 0.9205, + "step": 38748 + }, + { + "epoch": 2.86, + "learning_rate": 1.1220686098550934e-05, + "loss": 1.0477, + "step": 38749 + }, + { + "epoch": 2.86, + "learning_rate": 1.1220290208855773e-05, + "loss": 1.019, + "step": 38750 + }, + { + "epoch": 2.86, + "learning_rate": 1.1219894317219147e-05, + "loss": 1.0128, + "step": 38751 + }, + { + "epoch": 2.86, + "learning_rate": 1.1219498423641687e-05, + "loss": 1.0248, + "step": 38752 + }, + { + "epoch": 2.86, + "learning_rate": 1.1219102528124019e-05, + "loss": 0.976, + "step": 38753 + }, + { + "epoch": 2.86, + "learning_rate": 1.121870663066678e-05, + "loss": 1.1231, + "step": 38754 + }, + { + "epoch": 2.86, + "learning_rate": 1.1218310731270596e-05, + "loss": 1.042, + "step": 38755 + }, + { + "epoch": 2.86, + "learning_rate": 1.1217914829936097e-05, + "loss": 1.1538, + "step": 38756 + }, + { + "epoch": 2.86, + "learning_rate": 1.1217518926663912e-05, + "loss": 1.0304, + "step": 38757 + }, + { + "epoch": 2.86, + "learning_rate": 1.1217123021454673e-05, + "loss": 0.8904, + "step": 38758 + }, + { + "epoch": 2.86, + "learning_rate": 1.1216727114309007e-05, + "loss": 0.9975, + "step": 38759 + }, + { + "epoch": 2.86, + "learning_rate": 1.1216331205227546e-05, + "loss": 1.0256, + "step": 38760 + }, + { + "epoch": 2.86, + "learning_rate": 1.121593529421092e-05, + "loss": 1.0479, + "step": 38761 + }, + { + "epoch": 2.86, + "learning_rate": 1.121553938125976e-05, + "loss": 1.0959, + "step": 38762 + }, + { + "epoch": 2.86, + "learning_rate": 1.1215143466374692e-05, + "loss": 1.002, + "step": 38763 + }, + { + "epoch": 2.86, + "learning_rate": 1.1214747549556346e-05, + "loss": 0.9154, + "step": 38764 + }, + { + "epoch": 2.86, + "learning_rate": 1.1214351630805358e-05, + "loss": 1.062, + "step": 38765 + }, + { + "epoch": 2.86, + "learning_rate": 1.1213955710122353e-05, + "loss": 1.0213, + "step": 38766 + }, + { + "epoch": 2.86, + "learning_rate": 1.1213559787507961e-05, + "loss": 0.9753, + "step": 38767 + }, + { + "epoch": 2.86, + "learning_rate": 1.1213163862962815e-05, + "loss": 0.9673, + "step": 38768 + }, + { + "epoch": 2.86, + "learning_rate": 1.1212767936487541e-05, + "loss": 1.0092, + "step": 38769 + }, + { + "epoch": 2.86, + "learning_rate": 1.1212372008082768e-05, + "loss": 0.9751, + "step": 38770 + }, + { + "epoch": 2.86, + "learning_rate": 1.1211976077749132e-05, + "loss": 1.0272, + "step": 38771 + }, + { + "epoch": 2.86, + "learning_rate": 1.1211580145487258e-05, + "loss": 1.0182, + "step": 38772 + }, + { + "epoch": 2.86, + "learning_rate": 1.121118421129778e-05, + "loss": 0.999, + "step": 38773 + }, + { + "epoch": 2.86, + "learning_rate": 1.121078827518132e-05, + "loss": 0.9951, + "step": 38774 + }, + { + "epoch": 2.87, + "learning_rate": 1.1210392337138515e-05, + "loss": 0.983, + "step": 38775 + }, + { + "epoch": 2.87, + "learning_rate": 1.1209996397169995e-05, + "loss": 0.9496, + "step": 38776 + }, + { + "epoch": 2.87, + "learning_rate": 1.1209600455276388e-05, + "loss": 0.896, + "step": 38777 + }, + { + "epoch": 2.87, + "learning_rate": 1.1209204511458323e-05, + "loss": 1.0573, + "step": 38778 + }, + { + "epoch": 2.87, + "learning_rate": 1.120880856571643e-05, + "loss": 1.1033, + "step": 38779 + }, + { + "epoch": 2.87, + "learning_rate": 1.1208412618051341e-05, + "loss": 0.9784, + "step": 38780 + }, + { + "epoch": 2.87, + "learning_rate": 1.1208016668463684e-05, + "loss": 1.0434, + "step": 38781 + }, + { + "epoch": 2.87, + "learning_rate": 1.1207620716954091e-05, + "loss": 0.9963, + "step": 38782 + }, + { + "epoch": 2.87, + "learning_rate": 1.120722476352319e-05, + "loss": 0.9398, + "step": 38783 + }, + { + "epoch": 2.87, + "learning_rate": 1.1206828808171615e-05, + "loss": 1.029, + "step": 38784 + }, + { + "epoch": 2.87, + "learning_rate": 1.1206432850899987e-05, + "loss": 0.8853, + "step": 38785 + }, + { + "epoch": 2.87, + "learning_rate": 1.1206036891708945e-05, + "loss": 1.0026, + "step": 38786 + }, + { + "epoch": 2.87, + "learning_rate": 1.1205640930599113e-05, + "loss": 1.0665, + "step": 38787 + }, + { + "epoch": 2.87, + "learning_rate": 1.1205244967571127e-05, + "loss": 1.0215, + "step": 38788 + }, + { + "epoch": 2.87, + "learning_rate": 1.1204849002625611e-05, + "loss": 0.9745, + "step": 38789 + }, + { + "epoch": 2.87, + "learning_rate": 1.1204453035763199e-05, + "loss": 0.8927, + "step": 38790 + }, + { + "epoch": 2.87, + "learning_rate": 1.1204057066984516e-05, + "loss": 1.0227, + "step": 38791 + }, + { + "epoch": 2.87, + "learning_rate": 1.12036610962902e-05, + "loss": 0.9847, + "step": 38792 + }, + { + "epoch": 2.87, + "learning_rate": 1.1203265123680874e-05, + "loss": 0.9447, + "step": 38793 + }, + { + "epoch": 2.87, + "learning_rate": 1.1202869149157173e-05, + "loss": 1.1037, + "step": 38794 + }, + { + "epoch": 2.87, + "learning_rate": 1.1202473172719721e-05, + "loss": 1.0483, + "step": 38795 + }, + { + "epoch": 2.87, + "learning_rate": 1.1202077194369155e-05, + "loss": 1.0391, + "step": 38796 + }, + { + "epoch": 2.87, + "learning_rate": 1.1201681214106096e-05, + "loss": 1.04, + "step": 38797 + }, + { + "epoch": 2.87, + "learning_rate": 1.1201285231931185e-05, + "loss": 0.9834, + "step": 38798 + }, + { + "epoch": 2.87, + "learning_rate": 1.1200889247845042e-05, + "loss": 0.9553, + "step": 38799 + }, + { + "epoch": 2.87, + "learning_rate": 1.1200493261848304e-05, + "loss": 1.0014, + "step": 38800 + }, + { + "epoch": 2.87, + "learning_rate": 1.1200097273941595e-05, + "loss": 0.9793, + "step": 38801 + }, + { + "epoch": 2.87, + "learning_rate": 1.1199701284125552e-05, + "loss": 0.9362, + "step": 38802 + }, + { + "epoch": 2.87, + "learning_rate": 1.1199305292400801e-05, + "loss": 0.9624, + "step": 38803 + }, + { + "epoch": 2.87, + "learning_rate": 1.1198909298767972e-05, + "loss": 1.0462, + "step": 38804 + }, + { + "epoch": 2.87, + "learning_rate": 1.1198513303227695e-05, + "loss": 1.01, + "step": 38805 + }, + { + "epoch": 2.87, + "learning_rate": 1.1198117305780602e-05, + "loss": 0.8866, + "step": 38806 + }, + { + "epoch": 2.87, + "learning_rate": 1.119772130642732e-05, + "loss": 0.9873, + "step": 38807 + }, + { + "epoch": 2.87, + "learning_rate": 1.119732530516848e-05, + "loss": 1.0217, + "step": 38808 + }, + { + "epoch": 2.87, + "learning_rate": 1.1196929302004714e-05, + "loss": 0.9843, + "step": 38809 + }, + { + "epoch": 2.87, + "learning_rate": 1.119653329693665e-05, + "loss": 0.9635, + "step": 38810 + }, + { + "epoch": 2.87, + "learning_rate": 1.1196137289964917e-05, + "loss": 1.0313, + "step": 38811 + }, + { + "epoch": 2.87, + "learning_rate": 1.1195741281090151e-05, + "loss": 0.996, + "step": 38812 + }, + { + "epoch": 2.87, + "learning_rate": 1.1195345270312975e-05, + "loss": 1.0299, + "step": 38813 + }, + { + "epoch": 2.87, + "learning_rate": 1.1194949257634023e-05, + "loss": 1.003, + "step": 38814 + }, + { + "epoch": 2.87, + "learning_rate": 1.119455324305392e-05, + "loss": 1.0497, + "step": 38815 + }, + { + "epoch": 2.87, + "learning_rate": 1.1194157226573305e-05, + "loss": 0.9625, + "step": 38816 + }, + { + "epoch": 2.87, + "learning_rate": 1.11937612081928e-05, + "loss": 0.9214, + "step": 38817 + }, + { + "epoch": 2.87, + "learning_rate": 1.1193365187913039e-05, + "loss": 0.9726, + "step": 38818 + }, + { + "epoch": 2.87, + "learning_rate": 1.1192969165734649e-05, + "loss": 1.0549, + "step": 38819 + }, + { + "epoch": 2.87, + "learning_rate": 1.1192573141658266e-05, + "loss": 1.0458, + "step": 38820 + }, + { + "epoch": 2.87, + "learning_rate": 1.119217711568451e-05, + "loss": 0.932, + "step": 38821 + }, + { + "epoch": 2.87, + "learning_rate": 1.1191781087814021e-05, + "loss": 0.9966, + "step": 38822 + }, + { + "epoch": 2.87, + "learning_rate": 1.1191385058047429e-05, + "loss": 1.0534, + "step": 38823 + }, + { + "epoch": 2.87, + "learning_rate": 1.1190989026385354e-05, + "loss": 1.0454, + "step": 38824 + }, + { + "epoch": 2.87, + "learning_rate": 1.1190592992828438e-05, + "loss": 1.0767, + "step": 38825 + }, + { + "epoch": 2.87, + "learning_rate": 1.1190196957377302e-05, + "loss": 0.9565, + "step": 38826 + }, + { + "epoch": 2.87, + "learning_rate": 1.118980092003258e-05, + "loss": 1.0055, + "step": 38827 + }, + { + "epoch": 2.87, + "learning_rate": 1.11894048807949e-05, + "loss": 0.9326, + "step": 38828 + }, + { + "epoch": 2.87, + "learning_rate": 1.1189008839664898e-05, + "loss": 1.0561, + "step": 38829 + }, + { + "epoch": 2.87, + "learning_rate": 1.1188612796643199e-05, + "loss": 1.1286, + "step": 38830 + }, + { + "epoch": 2.87, + "learning_rate": 1.1188216751730434e-05, + "loss": 0.8622, + "step": 38831 + }, + { + "epoch": 2.87, + "learning_rate": 1.1187820704927232e-05, + "loss": 1.0471, + "step": 38832 + }, + { + "epoch": 2.87, + "learning_rate": 1.1187424656234225e-05, + "loss": 1.0999, + "step": 38833 + }, + { + "epoch": 2.87, + "learning_rate": 1.1187028605652042e-05, + "loss": 1.0066, + "step": 38834 + }, + { + "epoch": 2.87, + "learning_rate": 1.1186632553181314e-05, + "loss": 1.0133, + "step": 38835 + }, + { + "epoch": 2.87, + "learning_rate": 1.118623649882267e-05, + "loss": 0.9482, + "step": 38836 + }, + { + "epoch": 2.87, + "learning_rate": 1.1185840442576741e-05, + "loss": 0.9992, + "step": 38837 + }, + { + "epoch": 2.87, + "learning_rate": 1.1185444384444155e-05, + "loss": 0.9225, + "step": 38838 + }, + { + "epoch": 2.87, + "learning_rate": 1.1185048324425547e-05, + "loss": 1.0762, + "step": 38839 + }, + { + "epoch": 2.87, + "learning_rate": 1.1184652262521543e-05, + "loss": 1.0239, + "step": 38840 + }, + { + "epoch": 2.87, + "learning_rate": 1.1184256198732774e-05, + "loss": 1.0607, + "step": 38841 + }, + { + "epoch": 2.87, + "learning_rate": 1.1183860133059869e-05, + "loss": 1.0898, + "step": 38842 + }, + { + "epoch": 2.87, + "learning_rate": 1.1183464065503462e-05, + "loss": 1.0363, + "step": 38843 + }, + { + "epoch": 2.87, + "learning_rate": 1.1183067996064179e-05, + "loss": 1.0124, + "step": 38844 + }, + { + "epoch": 2.87, + "learning_rate": 1.1182671924742651e-05, + "loss": 0.9816, + "step": 38845 + }, + { + "epoch": 2.87, + "learning_rate": 1.118227585153951e-05, + "loss": 1.0187, + "step": 38846 + }, + { + "epoch": 2.87, + "learning_rate": 1.1181879776455385e-05, + "loss": 0.9715, + "step": 38847 + }, + { + "epoch": 2.87, + "learning_rate": 1.1181483699490905e-05, + "loss": 1.0118, + "step": 38848 + }, + { + "epoch": 2.87, + "learning_rate": 1.1181087620646704e-05, + "loss": 0.9946, + "step": 38849 + }, + { + "epoch": 2.87, + "learning_rate": 1.1180691539923407e-05, + "loss": 1.0112, + "step": 38850 + }, + { + "epoch": 2.87, + "learning_rate": 1.118029545732165e-05, + "loss": 1.0255, + "step": 38851 + }, + { + "epoch": 2.87, + "learning_rate": 1.1179899372842055e-05, + "loss": 0.9364, + "step": 38852 + }, + { + "epoch": 2.87, + "learning_rate": 1.117950328648526e-05, + "loss": 0.9487, + "step": 38853 + }, + { + "epoch": 2.87, + "learning_rate": 1.117910719825189e-05, + "loss": 0.9837, + "step": 38854 + }, + { + "epoch": 2.87, + "learning_rate": 1.1178711108142582e-05, + "loss": 1.0142, + "step": 38855 + }, + { + "epoch": 2.87, + "learning_rate": 1.1178315016157957e-05, + "loss": 1.0564, + "step": 38856 + }, + { + "epoch": 2.87, + "learning_rate": 1.1177918922298651e-05, + "loss": 1.146, + "step": 38857 + }, + { + "epoch": 2.87, + "learning_rate": 1.1177522826565292e-05, + "loss": 1.0033, + "step": 38858 + }, + { + "epoch": 2.87, + "learning_rate": 1.1177126728958514e-05, + "loss": 1.0367, + "step": 38859 + }, + { + "epoch": 2.87, + "learning_rate": 1.1176730629478943e-05, + "loss": 0.9693, + "step": 38860 + }, + { + "epoch": 2.87, + "learning_rate": 1.1176334528127209e-05, + "loss": 0.9697, + "step": 38861 + }, + { + "epoch": 2.87, + "learning_rate": 1.1175938424903942e-05, + "loss": 0.9757, + "step": 38862 + }, + { + "epoch": 2.87, + "learning_rate": 1.1175542319809778e-05, + "loss": 0.983, + "step": 38863 + }, + { + "epoch": 2.87, + "learning_rate": 1.1175146212845337e-05, + "loss": 1.0148, + "step": 38864 + }, + { + "epoch": 2.87, + "learning_rate": 1.1174750104011263e-05, + "loss": 0.9703, + "step": 38865 + }, + { + "epoch": 2.87, + "learning_rate": 1.117435399330817e-05, + "loss": 1.0903, + "step": 38866 + }, + { + "epoch": 2.87, + "learning_rate": 1.1173957880736704e-05, + "loss": 1.0116, + "step": 38867 + }, + { + "epoch": 2.87, + "learning_rate": 1.1173561766297484e-05, + "loss": 0.9546, + "step": 38868 + }, + { + "epoch": 2.87, + "learning_rate": 1.1173165649991145e-05, + "loss": 1.0112, + "step": 38869 + }, + { + "epoch": 2.87, + "learning_rate": 1.1172769531818317e-05, + "loss": 1.0, + "step": 38870 + }, + { + "epoch": 2.87, + "learning_rate": 1.1172373411779627e-05, + "loss": 1.0194, + "step": 38871 + }, + { + "epoch": 2.87, + "learning_rate": 1.1171977289875711e-05, + "loss": 1.029, + "step": 38872 + }, + { + "epoch": 2.87, + "learning_rate": 1.1171581166107194e-05, + "loss": 1.0663, + "step": 38873 + }, + { + "epoch": 2.87, + "learning_rate": 1.1171185040474708e-05, + "loss": 1.016, + "step": 38874 + }, + { + "epoch": 2.87, + "learning_rate": 1.1170788912978884e-05, + "loss": 0.9881, + "step": 38875 + }, + { + "epoch": 2.87, + "learning_rate": 1.117039278362035e-05, + "loss": 0.8672, + "step": 38876 + }, + { + "epoch": 2.87, + "learning_rate": 1.1169996652399742e-05, + "loss": 1.0435, + "step": 38877 + }, + { + "epoch": 2.87, + "learning_rate": 1.1169600519317683e-05, + "loss": 1.045, + "step": 38878 + }, + { + "epoch": 2.87, + "learning_rate": 1.1169204384374806e-05, + "loss": 1.0822, + "step": 38879 + }, + { + "epoch": 2.87, + "learning_rate": 1.1168808247571741e-05, + "loss": 1.0304, + "step": 38880 + }, + { + "epoch": 2.87, + "learning_rate": 1.1168412108909119e-05, + "loss": 0.9303, + "step": 38881 + }, + { + "epoch": 2.87, + "learning_rate": 1.1168015968387571e-05, + "loss": 1.0188, + "step": 38882 + }, + { + "epoch": 2.87, + "learning_rate": 1.1167619826007724e-05, + "loss": 0.9535, + "step": 38883 + }, + { + "epoch": 2.87, + "learning_rate": 1.1167223681770216e-05, + "loss": 0.9165, + "step": 38884 + }, + { + "epoch": 2.87, + "learning_rate": 1.1166827535675665e-05, + "loss": 0.9957, + "step": 38885 + }, + { + "epoch": 2.87, + "learning_rate": 1.1166431387724714e-05, + "loss": 0.9709, + "step": 38886 + }, + { + "epoch": 2.87, + "learning_rate": 1.1166035237917986e-05, + "loss": 1.0803, + "step": 38887 + }, + { + "epoch": 2.87, + "learning_rate": 1.1165639086256113e-05, + "loss": 1.0759, + "step": 38888 + }, + { + "epoch": 2.87, + "learning_rate": 1.1165242932739721e-05, + "loss": 1.0231, + "step": 38889 + }, + { + "epoch": 2.87, + "learning_rate": 1.1164846777369449e-05, + "loss": 1.0486, + "step": 38890 + }, + { + "epoch": 2.87, + "learning_rate": 1.1164450620145917e-05, + "loss": 1.0753, + "step": 38891 + }, + { + "epoch": 2.87, + "learning_rate": 1.1164054461069767e-05, + "loss": 1.059, + "step": 38892 + }, + { + "epoch": 2.87, + "learning_rate": 1.1163658300141618e-05, + "loss": 0.9725, + "step": 38893 + }, + { + "epoch": 2.87, + "learning_rate": 1.1163262137362108e-05, + "loss": 1.0097, + "step": 38894 + }, + { + "epoch": 2.87, + "learning_rate": 1.1162865972731863e-05, + "loss": 1.0445, + "step": 38895 + }, + { + "epoch": 2.87, + "learning_rate": 1.1162469806251516e-05, + "loss": 0.9229, + "step": 38896 + }, + { + "epoch": 2.87, + "learning_rate": 1.1162073637921696e-05, + "loss": 1.0136, + "step": 38897 + }, + { + "epoch": 2.87, + "learning_rate": 1.1161677467743032e-05, + "loss": 1.0176, + "step": 38898 + }, + { + "epoch": 2.87, + "learning_rate": 1.1161281295716157e-05, + "loss": 0.9067, + "step": 38899 + }, + { + "epoch": 2.87, + "learning_rate": 1.1160885121841702e-05, + "loss": 1.0829, + "step": 38900 + }, + { + "epoch": 2.87, + "learning_rate": 1.1160488946120291e-05, + "loss": 1.046, + "step": 38901 + }, + { + "epoch": 2.87, + "learning_rate": 1.1160092768552561e-05, + "loss": 0.9389, + "step": 38902 + }, + { + "epoch": 2.87, + "learning_rate": 1.115969658913914e-05, + "loss": 0.9352, + "step": 38903 + }, + { + "epoch": 2.87, + "learning_rate": 1.115930040788066e-05, + "loss": 0.9455, + "step": 38904 + }, + { + "epoch": 2.87, + "learning_rate": 1.1158904224777747e-05, + "loss": 0.9939, + "step": 38905 + }, + { + "epoch": 2.87, + "learning_rate": 1.1158508039831037e-05, + "loss": 1.026, + "step": 38906 + }, + { + "epoch": 2.87, + "learning_rate": 1.1158111853041156e-05, + "loss": 0.9511, + "step": 38907 + }, + { + "epoch": 2.87, + "learning_rate": 1.1157715664408735e-05, + "loss": 0.9633, + "step": 38908 + }, + { + "epoch": 2.87, + "learning_rate": 1.1157319473934403e-05, + "loss": 1.0806, + "step": 38909 + }, + { + "epoch": 2.88, + "learning_rate": 1.1156923281618795e-05, + "loss": 0.9701, + "step": 38910 + }, + { + "epoch": 2.88, + "learning_rate": 1.1156527087462536e-05, + "loss": 0.9976, + "step": 38911 + }, + { + "epoch": 2.88, + "learning_rate": 1.1156130891466261e-05, + "loss": 1.0754, + "step": 38912 + }, + { + "epoch": 2.88, + "learning_rate": 1.1155734693630598e-05, + "loss": 0.9923, + "step": 38913 + }, + { + "epoch": 2.88, + "learning_rate": 1.1155338493956178e-05, + "loss": 0.983, + "step": 38914 + }, + { + "epoch": 2.88, + "learning_rate": 1.1154942292443631e-05, + "loss": 1.1318, + "step": 38915 + }, + { + "epoch": 2.88, + "learning_rate": 1.1154546089093587e-05, + "loss": 1.0343, + "step": 38916 + }, + { + "epoch": 2.88, + "learning_rate": 1.1154149883906678e-05, + "loss": 1.0028, + "step": 38917 + }, + { + "epoch": 2.88, + "learning_rate": 1.1153753676883532e-05, + "loss": 0.9531, + "step": 38918 + }, + { + "epoch": 2.88, + "learning_rate": 1.1153357468024778e-05, + "loss": 0.9929, + "step": 38919 + }, + { + "epoch": 2.88, + "learning_rate": 1.1152961257331052e-05, + "loss": 0.9841, + "step": 38920 + }, + { + "epoch": 2.88, + "learning_rate": 1.115256504480298e-05, + "loss": 1.0646, + "step": 38921 + }, + { + "epoch": 2.88, + "learning_rate": 1.1152168830441195e-05, + "loss": 0.9493, + "step": 38922 + }, + { + "epoch": 2.88, + "learning_rate": 1.1151772614246322e-05, + "loss": 1.0319, + "step": 38923 + }, + { + "epoch": 2.88, + "learning_rate": 1.1151376396218998e-05, + "loss": 0.973, + "step": 38924 + }, + { + "epoch": 2.88, + "learning_rate": 1.1150980176359851e-05, + "loss": 1.0095, + "step": 38925 + }, + { + "epoch": 2.88, + "learning_rate": 1.115058395466951e-05, + "loss": 0.9017, + "step": 38926 + }, + { + "epoch": 2.88, + "learning_rate": 1.1150187731148608e-05, + "loss": 0.9964, + "step": 38927 + }, + { + "epoch": 2.88, + "learning_rate": 1.114979150579777e-05, + "loss": 0.8655, + "step": 38928 + }, + { + "epoch": 2.88, + "learning_rate": 1.1149395278617634e-05, + "loss": 1.0519, + "step": 38929 + }, + { + "epoch": 2.88, + "learning_rate": 1.1148999049608824e-05, + "loss": 1.0434, + "step": 38930 + }, + { + "epoch": 2.88, + "learning_rate": 1.1148602818771975e-05, + "loss": 1.0235, + "step": 38931 + }, + { + "epoch": 2.88, + "learning_rate": 1.1148206586107713e-05, + "loss": 1.0167, + "step": 38932 + }, + { + "epoch": 2.88, + "learning_rate": 1.1147810351616672e-05, + "loss": 1.0195, + "step": 38933 + }, + { + "epoch": 2.88, + "learning_rate": 1.1147414115299484e-05, + "loss": 0.943, + "step": 38934 + }, + { + "epoch": 2.88, + "learning_rate": 1.1147017877156774e-05, + "loss": 1.0579, + "step": 38935 + }, + { + "epoch": 2.88, + "learning_rate": 1.1146621637189174e-05, + "loss": 1.0625, + "step": 38936 + }, + { + "epoch": 2.88, + "learning_rate": 1.1146225395397316e-05, + "loss": 0.9454, + "step": 38937 + }, + { + "epoch": 2.88, + "learning_rate": 1.1145829151781831e-05, + "loss": 0.8369, + "step": 38938 + }, + { + "epoch": 2.88, + "learning_rate": 1.1145432906343347e-05, + "loss": 1.0302, + "step": 38939 + }, + { + "epoch": 2.88, + "learning_rate": 1.1145036659082495e-05, + "loss": 1.0205, + "step": 38940 + }, + { + "epoch": 2.88, + "learning_rate": 1.1144640409999908e-05, + "loss": 1.0067, + "step": 38941 + }, + { + "epoch": 2.88, + "learning_rate": 1.114424415909621e-05, + "loss": 1.0235, + "step": 38942 + }, + { + "epoch": 2.88, + "learning_rate": 1.1143847906372046e-05, + "loss": 1.1175, + "step": 38943 + }, + { + "epoch": 2.88, + "learning_rate": 1.1143451651828028e-05, + "loss": 1.0333, + "step": 38944 + }, + { + "epoch": 2.88, + "learning_rate": 1.1143055395464798e-05, + "loss": 0.9695, + "step": 38945 + }, + { + "epoch": 2.88, + "learning_rate": 1.1142659137282979e-05, + "loss": 1.0097, + "step": 38946 + }, + { + "epoch": 2.88, + "learning_rate": 1.114226287728321e-05, + "loss": 0.9281, + "step": 38947 + }, + { + "epoch": 2.88, + "learning_rate": 1.1141866615466113e-05, + "loss": 0.9929, + "step": 38948 + }, + { + "epoch": 2.88, + "learning_rate": 1.1141470351832326e-05, + "loss": 0.9271, + "step": 38949 + }, + { + "epoch": 2.88, + "learning_rate": 1.1141074086382473e-05, + "loss": 0.9829, + "step": 38950 + }, + { + "epoch": 2.88, + "learning_rate": 1.1140677819117191e-05, + "loss": 1.0, + "step": 38951 + }, + { + "epoch": 2.88, + "learning_rate": 1.1140281550037106e-05, + "loss": 1.0686, + "step": 38952 + }, + { + "epoch": 2.88, + "learning_rate": 1.113988527914285e-05, + "loss": 1.0295, + "step": 38953 + }, + { + "epoch": 2.88, + "learning_rate": 1.1139489006435052e-05, + "loss": 1.0459, + "step": 38954 + }, + { + "epoch": 2.88, + "learning_rate": 1.1139092731914342e-05, + "loss": 1.1095, + "step": 38955 + }, + { + "epoch": 2.88, + "learning_rate": 1.1138696455581353e-05, + "loss": 1.0672, + "step": 38956 + }, + { + "epoch": 2.88, + "learning_rate": 1.1138300177436714e-05, + "loss": 0.968, + "step": 38957 + }, + { + "epoch": 2.88, + "learning_rate": 1.1137903897481054e-05, + "loss": 1.0771, + "step": 38958 + }, + { + "epoch": 2.88, + "learning_rate": 1.113750761571501e-05, + "loss": 1.1697, + "step": 38959 + }, + { + "epoch": 2.88, + "learning_rate": 1.11371113321392e-05, + "loss": 1.016, + "step": 38960 + }, + { + "epoch": 2.88, + "learning_rate": 1.113671504675427e-05, + "loss": 1.0797, + "step": 38961 + }, + { + "epoch": 2.88, + "learning_rate": 1.1136318759560839e-05, + "loss": 1.0972, + "step": 38962 + }, + { + "epoch": 2.88, + "learning_rate": 1.1135922470559541e-05, + "loss": 0.9373, + "step": 38963 + }, + { + "epoch": 2.88, + "learning_rate": 1.1135526179751006e-05, + "loss": 1.0193, + "step": 38964 + }, + { + "epoch": 2.88, + "learning_rate": 1.1135129887135867e-05, + "loss": 0.9979, + "step": 38965 + }, + { + "epoch": 2.88, + "learning_rate": 1.1134733592714751e-05, + "loss": 0.9168, + "step": 38966 + }, + { + "epoch": 2.88, + "learning_rate": 1.1134337296488292e-05, + "loss": 0.9515, + "step": 38967 + }, + { + "epoch": 2.88, + "learning_rate": 1.1133940998457115e-05, + "loss": 1.0969, + "step": 38968 + }, + { + "epoch": 2.88, + "learning_rate": 1.1133544698621858e-05, + "loss": 0.9757, + "step": 38969 + }, + { + "epoch": 2.88, + "learning_rate": 1.1133148396983144e-05, + "loss": 1.0786, + "step": 38970 + }, + { + "epoch": 2.88, + "learning_rate": 1.113275209354161e-05, + "loss": 1.0853, + "step": 38971 + }, + { + "epoch": 2.88, + "learning_rate": 1.1132355788297884e-05, + "loss": 0.938, + "step": 38972 + }, + { + "epoch": 2.88, + "learning_rate": 1.1131959481252595e-05, + "loss": 0.9312, + "step": 38973 + }, + { + "epoch": 2.88, + "learning_rate": 1.1131563172406374e-05, + "loss": 1.0568, + "step": 38974 + }, + { + "epoch": 2.88, + "learning_rate": 1.1131166861759851e-05, + "loss": 0.9619, + "step": 38975 + }, + { + "epoch": 2.88, + "learning_rate": 1.1130770549313661e-05, + "loss": 0.8963, + "step": 38976 + }, + { + "epoch": 2.88, + "learning_rate": 1.113037423506843e-05, + "loss": 0.9474, + "step": 38977 + }, + { + "epoch": 2.88, + "learning_rate": 1.1129977919024789e-05, + "loss": 1.0304, + "step": 38978 + }, + { + "epoch": 2.88, + "learning_rate": 1.112958160118337e-05, + "loss": 0.8359, + "step": 38979 + }, + { + "epoch": 2.88, + "learning_rate": 1.1129185281544801e-05, + "loss": 0.962, + "step": 38980 + }, + { + "epoch": 2.88, + "learning_rate": 1.1128788960109718e-05, + "loss": 0.9415, + "step": 38981 + }, + { + "epoch": 2.88, + "learning_rate": 1.1128392636878747e-05, + "loss": 1.1426, + "step": 38982 + }, + { + "epoch": 2.88, + "learning_rate": 1.1127996311852517e-05, + "loss": 0.9976, + "step": 38983 + }, + { + "epoch": 2.88, + "learning_rate": 1.1127599985031662e-05, + "loss": 1.0523, + "step": 38984 + }, + { + "epoch": 2.88, + "learning_rate": 1.112720365641681e-05, + "loss": 0.9393, + "step": 38985 + }, + { + "epoch": 2.88, + "learning_rate": 1.1126807326008598e-05, + "loss": 1.0467, + "step": 38986 + }, + { + "epoch": 2.88, + "learning_rate": 1.1126410993807646e-05, + "loss": 1.027, + "step": 38987 + }, + { + "epoch": 2.88, + "learning_rate": 1.1126014659814595e-05, + "loss": 1.0478, + "step": 38988 + }, + { + "epoch": 2.88, + "learning_rate": 1.1125618324030065e-05, + "loss": 0.975, + "step": 38989 + }, + { + "epoch": 2.88, + "learning_rate": 1.11252219864547e-05, + "loss": 1.0055, + "step": 38990 + }, + { + "epoch": 2.88, + "learning_rate": 1.1124825647089116e-05, + "loss": 1.0895, + "step": 38991 + }, + { + "epoch": 2.88, + "learning_rate": 1.1124429305933958e-05, + "loss": 1.0427, + "step": 38992 + }, + { + "epoch": 2.88, + "learning_rate": 1.1124032962989841e-05, + "loss": 0.9709, + "step": 38993 + }, + { + "epoch": 2.88, + "learning_rate": 1.1123636618257407e-05, + "loss": 0.9381, + "step": 38994 + }, + { + "epoch": 2.88, + "learning_rate": 1.1123240271737283e-05, + "loss": 0.9609, + "step": 38995 + }, + { + "epoch": 2.88, + "learning_rate": 1.1122843923430102e-05, + "loss": 1.0746, + "step": 38996 + }, + { + "epoch": 2.88, + "learning_rate": 1.1122447573336487e-05, + "loss": 1.0897, + "step": 38997 + }, + { + "epoch": 2.88, + "learning_rate": 1.112205122145708e-05, + "loss": 1.023, + "step": 38998 + }, + { + "epoch": 2.88, + "learning_rate": 1.1121654867792505e-05, + "loss": 1.0144, + "step": 38999 + }, + { + "epoch": 2.88, + "learning_rate": 1.112125851234339e-05, + "loss": 0.9912, + "step": 39000 + }, + { + "epoch": 2.88, + "learning_rate": 1.112086215511037e-05, + "loss": 1.0107, + "step": 39001 + }, + { + "epoch": 2.88, + "learning_rate": 1.1120465796094075e-05, + "loss": 1.0347, + "step": 39002 + }, + { + "epoch": 2.88, + "learning_rate": 1.1120069435295133e-05, + "loss": 1.0241, + "step": 39003 + }, + { + "epoch": 2.88, + "learning_rate": 1.111967307271418e-05, + "loss": 0.9849, + "step": 39004 + }, + { + "epoch": 2.88, + "learning_rate": 1.1119276708351839e-05, + "loss": 0.9397, + "step": 39005 + }, + { + "epoch": 2.88, + "learning_rate": 1.1118880342208746e-05, + "loss": 1.0552, + "step": 39006 + }, + { + "epoch": 2.88, + "learning_rate": 1.111848397428553e-05, + "loss": 0.9543, + "step": 39007 + }, + { + "epoch": 2.88, + "learning_rate": 1.1118087604582826e-05, + "loss": 1.0451, + "step": 39008 + }, + { + "epoch": 2.88, + "learning_rate": 1.1117691233101257e-05, + "loss": 1.0299, + "step": 39009 + }, + { + "epoch": 2.88, + "learning_rate": 1.1117294859841459e-05, + "loss": 0.9833, + "step": 39010 + }, + { + "epoch": 2.88, + "learning_rate": 1.1116898484804059e-05, + "loss": 1.0063, + "step": 39011 + }, + { + "epoch": 2.88, + "learning_rate": 1.111650210798969e-05, + "loss": 0.9084, + "step": 39012 + }, + { + "epoch": 2.88, + "learning_rate": 1.1116105729398982e-05, + "loss": 1.1289, + "step": 39013 + }, + { + "epoch": 2.88, + "learning_rate": 1.1115709349032566e-05, + "loss": 0.9387, + "step": 39014 + }, + { + "epoch": 2.88, + "learning_rate": 1.1115312966891072e-05, + "loss": 1.0606, + "step": 39015 + }, + { + "epoch": 2.88, + "learning_rate": 1.1114916582975133e-05, + "loss": 0.9401, + "step": 39016 + }, + { + "epoch": 2.88, + "learning_rate": 1.1114520197285374e-05, + "loss": 1.0592, + "step": 39017 + }, + { + "epoch": 2.88, + "learning_rate": 1.1114123809822435e-05, + "loss": 0.9991, + "step": 39018 + }, + { + "epoch": 2.88, + "learning_rate": 1.1113727420586938e-05, + "loss": 0.8942, + "step": 39019 + }, + { + "epoch": 2.88, + "learning_rate": 1.1113331029579514e-05, + "loss": 0.9353, + "step": 39020 + }, + { + "epoch": 2.88, + "learning_rate": 1.1112934636800799e-05, + "loss": 1.0001, + "step": 39021 + }, + { + "epoch": 2.88, + "learning_rate": 1.1112538242251421e-05, + "loss": 0.9836, + "step": 39022 + }, + { + "epoch": 2.88, + "learning_rate": 1.111214184593201e-05, + "loss": 1.036, + "step": 39023 + }, + { + "epoch": 2.88, + "learning_rate": 1.1111745447843196e-05, + "loss": 1.0851, + "step": 39024 + }, + { + "epoch": 2.88, + "learning_rate": 1.1111349047985611e-05, + "loss": 0.9575, + "step": 39025 + }, + { + "epoch": 2.88, + "learning_rate": 1.111095264635989e-05, + "loss": 0.9948, + "step": 39026 + }, + { + "epoch": 2.88, + "learning_rate": 1.1110556242966655e-05, + "loss": 0.9802, + "step": 39027 + }, + { + "epoch": 2.88, + "learning_rate": 1.111015983780654e-05, + "loss": 1.0285, + "step": 39028 + }, + { + "epoch": 2.88, + "learning_rate": 1.110976343088018e-05, + "loss": 1.1447, + "step": 39029 + }, + { + "epoch": 2.88, + "learning_rate": 1.11093670221882e-05, + "loss": 0.9937, + "step": 39030 + }, + { + "epoch": 2.88, + "learning_rate": 1.1108970611731234e-05, + "loss": 1.065, + "step": 39031 + }, + { + "epoch": 2.88, + "learning_rate": 1.110857419950991e-05, + "loss": 0.8735, + "step": 39032 + }, + { + "epoch": 2.88, + "learning_rate": 1.1108177785524862e-05, + "loss": 1.0084, + "step": 39033 + }, + { + "epoch": 2.88, + "learning_rate": 1.1107781369776718e-05, + "loss": 1.0519, + "step": 39034 + }, + { + "epoch": 2.88, + "learning_rate": 1.1107384952266111e-05, + "loss": 1.0403, + "step": 39035 + }, + { + "epoch": 2.88, + "learning_rate": 1.1106988532993667e-05, + "loss": 0.9938, + "step": 39036 + }, + { + "epoch": 2.88, + "learning_rate": 1.1106592111960027e-05, + "loss": 1.1017, + "step": 39037 + }, + { + "epoch": 2.88, + "learning_rate": 1.1106195689165808e-05, + "loss": 0.9152, + "step": 39038 + }, + { + "epoch": 2.88, + "learning_rate": 1.1105799264611651e-05, + "loss": 0.9838, + "step": 39039 + }, + { + "epoch": 2.88, + "learning_rate": 1.110540283829818e-05, + "loss": 0.9567, + "step": 39040 + }, + { + "epoch": 2.88, + "learning_rate": 1.1105006410226033e-05, + "loss": 1.0037, + "step": 39041 + }, + { + "epoch": 2.88, + "learning_rate": 1.1104609980395833e-05, + "loss": 0.969, + "step": 39042 + }, + { + "epoch": 2.88, + "learning_rate": 1.1104213548808218e-05, + "loss": 0.9167, + "step": 39043 + }, + { + "epoch": 2.88, + "learning_rate": 1.1103817115463811e-05, + "loss": 0.9457, + "step": 39044 + }, + { + "epoch": 2.89, + "learning_rate": 1.110342068036325e-05, + "loss": 0.9564, + "step": 39045 + }, + { + "epoch": 2.89, + "learning_rate": 1.110302424350716e-05, + "loss": 1.008, + "step": 39046 + }, + { + "epoch": 2.89, + "learning_rate": 1.1102627804896179e-05, + "loss": 1.0311, + "step": 39047 + }, + { + "epoch": 2.89, + "learning_rate": 1.1102231364530927e-05, + "loss": 1.1053, + "step": 39048 + }, + { + "epoch": 2.89, + "learning_rate": 1.1101834922412045e-05, + "loss": 0.9731, + "step": 39049 + }, + { + "epoch": 2.89, + "learning_rate": 1.1101438478540155e-05, + "loss": 0.9955, + "step": 39050 + }, + { + "epoch": 2.89, + "learning_rate": 1.1101042032915896e-05, + "loss": 0.9735, + "step": 39051 + }, + { + "epoch": 2.89, + "learning_rate": 1.1100645585539892e-05, + "loss": 1.0338, + "step": 39052 + }, + { + "epoch": 2.89, + "learning_rate": 1.1100249136412779e-05, + "loss": 0.9941, + "step": 39053 + }, + { + "epoch": 2.89, + "learning_rate": 1.1099852685535181e-05, + "loss": 0.9763, + "step": 39054 + }, + { + "epoch": 2.89, + "learning_rate": 1.1099456232907742e-05, + "loss": 1.0341, + "step": 39055 + }, + { + "epoch": 2.89, + "learning_rate": 1.1099059778531077e-05, + "loss": 1.0035, + "step": 39056 + }, + { + "epoch": 2.89, + "learning_rate": 1.1098663322405825e-05, + "loss": 1.0271, + "step": 39057 + }, + { + "epoch": 2.89, + "learning_rate": 1.1098266864532615e-05, + "loss": 0.8944, + "step": 39058 + }, + { + "epoch": 2.89, + "learning_rate": 1.1097870404912078e-05, + "loss": 1.028, + "step": 39059 + }, + { + "epoch": 2.89, + "learning_rate": 1.1097473943544845e-05, + "loss": 1.0356, + "step": 39060 + }, + { + "epoch": 2.89, + "learning_rate": 1.1097077480431548e-05, + "loss": 1.0391, + "step": 39061 + }, + { + "epoch": 2.89, + "learning_rate": 1.1096681015572814e-05, + "loss": 1.0097, + "step": 39062 + }, + { + "epoch": 2.89, + "learning_rate": 1.1096284548969279e-05, + "loss": 0.9484, + "step": 39063 + }, + { + "epoch": 2.89, + "learning_rate": 1.1095888080621569e-05, + "loss": 0.9952, + "step": 39064 + }, + { + "epoch": 2.89, + "learning_rate": 1.1095491610530318e-05, + "loss": 1.0287, + "step": 39065 + }, + { + "epoch": 2.89, + "learning_rate": 1.1095095138696158e-05, + "loss": 0.8728, + "step": 39066 + }, + { + "epoch": 2.89, + "learning_rate": 1.1094698665119715e-05, + "loss": 1.05, + "step": 39067 + }, + { + "epoch": 2.89, + "learning_rate": 1.1094302189801618e-05, + "loss": 1.0637, + "step": 39068 + }, + { + "epoch": 2.89, + "learning_rate": 1.1093905712742505e-05, + "loss": 0.9936, + "step": 39069 + }, + { + "epoch": 2.89, + "learning_rate": 1.1093509233943003e-05, + "loss": 1.0507, + "step": 39070 + }, + { + "epoch": 2.89, + "learning_rate": 1.1093112753403745e-05, + "loss": 1.0777, + "step": 39071 + }, + { + "epoch": 2.89, + "learning_rate": 1.1092716271125358e-05, + "loss": 0.9775, + "step": 39072 + }, + { + "epoch": 2.89, + "learning_rate": 1.1092319787108479e-05, + "loss": 1.0634, + "step": 39073 + }, + { + "epoch": 2.89, + "learning_rate": 1.1091923301353734e-05, + "loss": 0.9394, + "step": 39074 + }, + { + "epoch": 2.89, + "learning_rate": 1.1091526813861751e-05, + "loss": 1.1004, + "step": 39075 + }, + { + "epoch": 2.89, + "learning_rate": 1.1091130324633168e-05, + "loss": 1.0197, + "step": 39076 + }, + { + "epoch": 2.89, + "learning_rate": 1.109073383366861e-05, + "loss": 1.0259, + "step": 39077 + }, + { + "epoch": 2.89, + "learning_rate": 1.109033734096871e-05, + "loss": 0.976, + "step": 39078 + }, + { + "epoch": 2.89, + "learning_rate": 1.1089940846534099e-05, + "loss": 0.9579, + "step": 39079 + }, + { + "epoch": 2.89, + "learning_rate": 1.1089544350365409e-05, + "loss": 1.0003, + "step": 39080 + }, + { + "epoch": 2.89, + "learning_rate": 1.108914785246327e-05, + "loss": 1.0632, + "step": 39081 + }, + { + "epoch": 2.89, + "learning_rate": 1.108875135282831e-05, + "loss": 0.9956, + "step": 39082 + }, + { + "epoch": 2.89, + "learning_rate": 1.1088354851461166e-05, + "loss": 1.0625, + "step": 39083 + }, + { + "epoch": 2.89, + "learning_rate": 1.1087958348362462e-05, + "loss": 1.0011, + "step": 39084 + }, + { + "epoch": 2.89, + "learning_rate": 1.1087561843532833e-05, + "loss": 0.9923, + "step": 39085 + }, + { + "epoch": 2.89, + "learning_rate": 1.1087165336972907e-05, + "loss": 1.0091, + "step": 39086 + }, + { + "epoch": 2.89, + "learning_rate": 1.1086768828683316e-05, + "loss": 0.9545, + "step": 39087 + }, + { + "epoch": 2.89, + "learning_rate": 1.1086372318664696e-05, + "loss": 0.8703, + "step": 39088 + }, + { + "epoch": 2.89, + "learning_rate": 1.1085975806917669e-05, + "loss": 0.8872, + "step": 39089 + }, + { + "epoch": 2.89, + "learning_rate": 1.1085579293442872e-05, + "loss": 1.0361, + "step": 39090 + }, + { + "epoch": 2.89, + "learning_rate": 1.108518277824093e-05, + "loss": 0.9056, + "step": 39091 + }, + { + "epoch": 2.89, + "learning_rate": 1.108478626131248e-05, + "loss": 1.0018, + "step": 39092 + }, + { + "epoch": 2.89, + "learning_rate": 1.1084389742658154e-05, + "loss": 1.071, + "step": 39093 + }, + { + "epoch": 2.89, + "learning_rate": 1.1083993222278578e-05, + "loss": 1.0654, + "step": 39094 + }, + { + "epoch": 2.89, + "learning_rate": 1.1083596700174382e-05, + "loss": 0.993, + "step": 39095 + }, + { + "epoch": 2.89, + "learning_rate": 1.10832001763462e-05, + "loss": 1.0531, + "step": 39096 + }, + { + "epoch": 2.89, + "learning_rate": 1.1082803650794663e-05, + "loss": 1.0168, + "step": 39097 + }, + { + "epoch": 2.89, + "learning_rate": 1.10824071235204e-05, + "loss": 1.0917, + "step": 39098 + }, + { + "epoch": 2.89, + "learning_rate": 1.1082010594524042e-05, + "loss": 1.0633, + "step": 39099 + }, + { + "epoch": 2.89, + "learning_rate": 1.1081614063806224e-05, + "loss": 1.0814, + "step": 39100 + }, + { + "epoch": 2.89, + "learning_rate": 1.1081217531367567e-05, + "loss": 1.0976, + "step": 39101 + }, + { + "epoch": 2.89, + "learning_rate": 1.1080820997208716e-05, + "loss": 1.0277, + "step": 39102 + }, + { + "epoch": 2.89, + "learning_rate": 1.108042446133029e-05, + "loss": 1.0222, + "step": 39103 + }, + { + "epoch": 2.89, + "learning_rate": 1.1080027923732925e-05, + "loss": 1.0222, + "step": 39104 + }, + { + "epoch": 2.89, + "learning_rate": 1.107963138441725e-05, + "loss": 0.9651, + "step": 39105 + }, + { + "epoch": 2.89, + "learning_rate": 1.1079234843383899e-05, + "loss": 0.9516, + "step": 39106 + }, + { + "epoch": 2.89, + "learning_rate": 1.1078838300633496e-05, + "loss": 1.0016, + "step": 39107 + }, + { + "epoch": 2.89, + "learning_rate": 1.1078441756166682e-05, + "loss": 1.0173, + "step": 39108 + }, + { + "epoch": 2.89, + "learning_rate": 1.107804520998408e-05, + "loss": 0.969, + "step": 39109 + }, + { + "epoch": 2.89, + "learning_rate": 1.1077648662086324e-05, + "loss": 1.0613, + "step": 39110 + }, + { + "epoch": 2.89, + "learning_rate": 1.1077252112474046e-05, + "loss": 0.9835, + "step": 39111 + }, + { + "epoch": 2.89, + "learning_rate": 1.1076855561147877e-05, + "loss": 1.0205, + "step": 39112 + }, + { + "epoch": 2.89, + "learning_rate": 1.107645900810844e-05, + "loss": 0.9301, + "step": 39113 + }, + { + "epoch": 2.89, + "learning_rate": 1.1076062453356377e-05, + "loss": 0.998, + "step": 39114 + }, + { + "epoch": 2.89, + "learning_rate": 1.107566589689231e-05, + "loss": 1.0029, + "step": 39115 + }, + { + "epoch": 2.89, + "learning_rate": 1.1075269338716879e-05, + "loss": 0.9921, + "step": 39116 + }, + { + "epoch": 2.89, + "learning_rate": 1.1074872778830706e-05, + "loss": 1.0208, + "step": 39117 + }, + { + "epoch": 2.89, + "learning_rate": 1.1074476217234427e-05, + "loss": 0.9573, + "step": 39118 + }, + { + "epoch": 2.89, + "learning_rate": 1.107407965392867e-05, + "loss": 1.0644, + "step": 39119 + }, + { + "epoch": 2.89, + "learning_rate": 1.107368308891407e-05, + "loss": 1.0585, + "step": 39120 + }, + { + "epoch": 2.89, + "learning_rate": 1.1073286522191256e-05, + "loss": 1.0434, + "step": 39121 + }, + { + "epoch": 2.89, + "learning_rate": 1.1072889953760857e-05, + "loss": 0.9754, + "step": 39122 + }, + { + "epoch": 2.89, + "learning_rate": 1.1072493383623506e-05, + "loss": 0.9945, + "step": 39123 + }, + { + "epoch": 2.89, + "learning_rate": 1.1072096811779832e-05, + "loss": 1.0205, + "step": 39124 + }, + { + "epoch": 2.89, + "learning_rate": 1.1071700238230469e-05, + "loss": 0.9191, + "step": 39125 + }, + { + "epoch": 2.89, + "learning_rate": 1.1071303662976046e-05, + "loss": 0.9733, + "step": 39126 + }, + { + "epoch": 2.89, + "learning_rate": 1.1070907086017194e-05, + "loss": 1.0449, + "step": 39127 + }, + { + "epoch": 2.89, + "learning_rate": 1.1070510507354544e-05, + "loss": 0.9495, + "step": 39128 + }, + { + "epoch": 2.89, + "learning_rate": 1.1070113926988725e-05, + "loss": 0.9529, + "step": 39129 + }, + { + "epoch": 2.89, + "learning_rate": 1.1069717344920374e-05, + "loss": 1.033, + "step": 39130 + }, + { + "epoch": 2.89, + "learning_rate": 1.1069320761150117e-05, + "loss": 0.9322, + "step": 39131 + }, + { + "epoch": 2.89, + "learning_rate": 1.1068924175678585e-05, + "loss": 1.0171, + "step": 39132 + }, + { + "epoch": 2.89, + "learning_rate": 1.1068527588506411e-05, + "loss": 0.8912, + "step": 39133 + }, + { + "epoch": 2.89, + "learning_rate": 1.1068130999634224e-05, + "loss": 0.8884, + "step": 39134 + }, + { + "epoch": 2.89, + "learning_rate": 1.1067734409062656e-05, + "loss": 1.055, + "step": 39135 + }, + { + "epoch": 2.89, + "learning_rate": 1.1067337816792336e-05, + "loss": 0.9834, + "step": 39136 + }, + { + "epoch": 2.89, + "learning_rate": 1.1066941222823898e-05, + "loss": 1.0612, + "step": 39137 + }, + { + "epoch": 2.89, + "learning_rate": 1.1066544627157972e-05, + "loss": 1.0988, + "step": 39138 + }, + { + "epoch": 2.89, + "learning_rate": 1.1066148029795192e-05, + "loss": 1.041, + "step": 39139 + }, + { + "epoch": 2.89, + "learning_rate": 1.1065751430736182e-05, + "loss": 1.0465, + "step": 39140 + }, + { + "epoch": 2.89, + "learning_rate": 1.1065354829981578e-05, + "loss": 1.0425, + "step": 39141 + }, + { + "epoch": 2.89, + "learning_rate": 1.1064958227532008e-05, + "loss": 1.0868, + "step": 39142 + }, + { + "epoch": 2.89, + "learning_rate": 1.1064561623388108e-05, + "loss": 1.0326, + "step": 39143 + }, + { + "epoch": 2.89, + "learning_rate": 1.1064165017550501e-05, + "loss": 0.9568, + "step": 39144 + }, + { + "epoch": 2.89, + "learning_rate": 1.1063768410019827e-05, + "loss": 0.9192, + "step": 39145 + }, + { + "epoch": 2.89, + "learning_rate": 1.1063371800796709e-05, + "loss": 0.9541, + "step": 39146 + }, + { + "epoch": 2.89, + "learning_rate": 1.1062975189881785e-05, + "loss": 0.9552, + "step": 39147 + }, + { + "epoch": 2.89, + "learning_rate": 1.106257857727568e-05, + "loss": 1.0084, + "step": 39148 + }, + { + "epoch": 2.89, + "learning_rate": 1.1062181962979032e-05, + "loss": 0.9554, + "step": 39149 + }, + { + "epoch": 2.89, + "learning_rate": 1.1061785346992465e-05, + "loss": 0.9079, + "step": 39150 + }, + { + "epoch": 2.89, + "learning_rate": 1.1061388729316614e-05, + "loss": 0.9686, + "step": 39151 + }, + { + "epoch": 2.89, + "learning_rate": 1.1060992109952105e-05, + "loss": 1.0855, + "step": 39152 + }, + { + "epoch": 2.89, + "learning_rate": 1.1060595488899575e-05, + "loss": 1.1004, + "step": 39153 + }, + { + "epoch": 2.89, + "learning_rate": 1.106019886615965e-05, + "loss": 0.8314, + "step": 39154 + }, + { + "epoch": 2.89, + "learning_rate": 1.1059802241732968e-05, + "loss": 0.9435, + "step": 39155 + }, + { + "epoch": 2.89, + "learning_rate": 1.1059405615620153e-05, + "loss": 1.0252, + "step": 39156 + }, + { + "epoch": 2.89, + "learning_rate": 1.1059008987821841e-05, + "loss": 0.9117, + "step": 39157 + }, + { + "epoch": 2.89, + "learning_rate": 1.1058612358338661e-05, + "loss": 1.0408, + "step": 39158 + }, + { + "epoch": 2.89, + "learning_rate": 1.1058215727171244e-05, + "loss": 0.9855, + "step": 39159 + }, + { + "epoch": 2.89, + "learning_rate": 1.1057819094320219e-05, + "loss": 1.0096, + "step": 39160 + }, + { + "epoch": 2.89, + "learning_rate": 1.105742245978622e-05, + "loss": 1.004, + "step": 39161 + }, + { + "epoch": 2.89, + "learning_rate": 1.1057025823569877e-05, + "loss": 0.864, + "step": 39162 + }, + { + "epoch": 2.89, + "learning_rate": 1.1056629185671819e-05, + "loss": 1.0353, + "step": 39163 + }, + { + "epoch": 2.89, + "learning_rate": 1.105623254609268e-05, + "loss": 1.0236, + "step": 39164 + }, + { + "epoch": 2.89, + "learning_rate": 1.1055835904833092e-05, + "loss": 0.9539, + "step": 39165 + }, + { + "epoch": 2.89, + "learning_rate": 1.105543926189368e-05, + "loss": 1.1029, + "step": 39166 + }, + { + "epoch": 2.89, + "learning_rate": 1.1055042617275084e-05, + "loss": 1.0056, + "step": 39167 + }, + { + "epoch": 2.89, + "learning_rate": 1.105464597097793e-05, + "loss": 0.994, + "step": 39168 + }, + { + "epoch": 2.89, + "learning_rate": 1.1054249323002849e-05, + "loss": 1.0211, + "step": 39169 + }, + { + "epoch": 2.89, + "learning_rate": 1.105385267335047e-05, + "loss": 0.9546, + "step": 39170 + }, + { + "epoch": 2.89, + "learning_rate": 1.1053456022021427e-05, + "loss": 0.9287, + "step": 39171 + }, + { + "epoch": 2.89, + "learning_rate": 1.1053059369016352e-05, + "loss": 1.0365, + "step": 39172 + }, + { + "epoch": 2.89, + "learning_rate": 1.1052662714335876e-05, + "loss": 0.8974, + "step": 39173 + }, + { + "epoch": 2.89, + "learning_rate": 1.1052266057980624e-05, + "loss": 1.0679, + "step": 39174 + }, + { + "epoch": 2.89, + "learning_rate": 1.1051869399951236e-05, + "loss": 1.0036, + "step": 39175 + }, + { + "epoch": 2.89, + "learning_rate": 1.1051472740248336e-05, + "loss": 0.9541, + "step": 39176 + }, + { + "epoch": 2.89, + "learning_rate": 1.105107607887256e-05, + "loss": 0.9884, + "step": 39177 + }, + { + "epoch": 2.89, + "learning_rate": 1.1050679415824537e-05, + "loss": 1.0144, + "step": 39178 + }, + { + "epoch": 2.89, + "learning_rate": 1.1050282751104898e-05, + "loss": 0.979, + "step": 39179 + }, + { + "epoch": 2.89, + "learning_rate": 1.1049886084714271e-05, + "loss": 0.905, + "step": 39180 + }, + { + "epoch": 2.9, + "learning_rate": 1.1049489416653294e-05, + "loss": 1.0619, + "step": 39181 + }, + { + "epoch": 2.9, + "learning_rate": 1.1049092746922595e-05, + "loss": 0.9872, + "step": 39182 + }, + { + "epoch": 2.9, + "learning_rate": 1.10486960755228e-05, + "loss": 0.9449, + "step": 39183 + }, + { + "epoch": 2.9, + "learning_rate": 1.1048299402454548e-05, + "loss": 1.028, + "step": 39184 + }, + { + "epoch": 2.9, + "learning_rate": 1.1047902727718462e-05, + "loss": 1.0502, + "step": 39185 + }, + { + "epoch": 2.9, + "learning_rate": 1.1047506051315187e-05, + "loss": 1.0822, + "step": 39186 + }, + { + "epoch": 2.9, + "learning_rate": 1.1047109373245337e-05, + "loss": 0.9781, + "step": 39187 + }, + { + "epoch": 2.9, + "learning_rate": 1.1046712693509554e-05, + "loss": 1.0242, + "step": 39188 + }, + { + "epoch": 2.9, + "learning_rate": 1.1046316012108465e-05, + "loss": 1.0182, + "step": 39189 + }, + { + "epoch": 2.9, + "learning_rate": 1.1045919329042702e-05, + "loss": 0.9032, + "step": 39190 + }, + { + "epoch": 2.9, + "learning_rate": 1.1045522644312896e-05, + "loss": 1.0979, + "step": 39191 + }, + { + "epoch": 2.9, + "learning_rate": 1.104512595791968e-05, + "loss": 1.0151, + "step": 39192 + }, + { + "epoch": 2.9, + "learning_rate": 1.1044729269863681e-05, + "loss": 1.0583, + "step": 39193 + }, + { + "epoch": 2.9, + "learning_rate": 1.1044332580145537e-05, + "loss": 1.0138, + "step": 39194 + }, + { + "epoch": 2.9, + "learning_rate": 1.1043935888765871e-05, + "loss": 0.904, + "step": 39195 + }, + { + "epoch": 2.9, + "learning_rate": 1.1043539195725321e-05, + "loss": 1.0894, + "step": 39196 + }, + { + "epoch": 2.9, + "learning_rate": 1.1043142501024512e-05, + "loss": 1.0116, + "step": 39197 + }, + { + "epoch": 2.9, + "learning_rate": 1.1042745804664081e-05, + "loss": 1.0459, + "step": 39198 + }, + { + "epoch": 2.9, + "learning_rate": 1.1042349106644655e-05, + "loss": 1.0548, + "step": 39199 + }, + { + "epoch": 2.9, + "learning_rate": 1.1041952406966867e-05, + "loss": 1.1295, + "step": 39200 + }, + { + "epoch": 2.9, + "learning_rate": 1.1041555705631345e-05, + "loss": 0.9675, + "step": 39201 + }, + { + "epoch": 2.9, + "learning_rate": 1.1041159002638726e-05, + "loss": 1.0294, + "step": 39202 + }, + { + "epoch": 2.9, + "learning_rate": 1.1040762297989635e-05, + "loss": 1.0492, + "step": 39203 + }, + { + "epoch": 2.9, + "learning_rate": 1.1040365591684709e-05, + "loss": 0.9905, + "step": 39204 + }, + { + "epoch": 2.9, + "learning_rate": 1.1039968883724577e-05, + "loss": 1.0653, + "step": 39205 + }, + { + "epoch": 2.9, + "learning_rate": 1.1039572174109867e-05, + "loss": 0.9548, + "step": 39206 + }, + { + "epoch": 2.9, + "learning_rate": 1.1039175462841212e-05, + "loss": 1.0424, + "step": 39207 + }, + { + "epoch": 2.9, + "learning_rate": 1.1038778749919246e-05, + "loss": 1.0277, + "step": 39208 + }, + { + "epoch": 2.9, + "learning_rate": 1.1038382035344594e-05, + "loss": 1.0297, + "step": 39209 + }, + { + "epoch": 2.9, + "learning_rate": 1.1037985319117895e-05, + "loss": 1.0601, + "step": 39210 + }, + { + "epoch": 2.9, + "learning_rate": 1.1037588601239773e-05, + "loss": 1.0158, + "step": 39211 + }, + { + "epoch": 2.9, + "learning_rate": 1.1037191881710864e-05, + "loss": 0.9522, + "step": 39212 + }, + { + "epoch": 2.9, + "learning_rate": 1.1036795160531799e-05, + "loss": 1.0824, + "step": 39213 + }, + { + "epoch": 2.9, + "learning_rate": 1.1036398437703206e-05, + "loss": 0.8824, + "step": 39214 + }, + { + "epoch": 2.9, + "learning_rate": 1.103600171322572e-05, + "loss": 0.9243, + "step": 39215 + }, + { + "epoch": 2.9, + "learning_rate": 1.1035604987099968e-05, + "loss": 1.066, + "step": 39216 + }, + { + "epoch": 2.9, + "learning_rate": 1.1035208259326585e-05, + "loss": 1.0268, + "step": 39217 + }, + { + "epoch": 2.9, + "learning_rate": 1.1034811529906197e-05, + "loss": 1.0409, + "step": 39218 + }, + { + "epoch": 2.9, + "learning_rate": 1.1034414798839441e-05, + "loss": 1.0524, + "step": 39219 + }, + { + "epoch": 2.9, + "learning_rate": 1.1034018066126946e-05, + "loss": 0.9782, + "step": 39220 + }, + { + "epoch": 2.9, + "learning_rate": 1.1033621331769342e-05, + "loss": 0.9391, + "step": 39221 + }, + { + "epoch": 2.9, + "learning_rate": 1.1033224595767263e-05, + "loss": 0.9624, + "step": 39222 + }, + { + "epoch": 2.9, + "learning_rate": 1.1032827858121336e-05, + "loss": 0.9527, + "step": 39223 + }, + { + "epoch": 2.9, + "learning_rate": 1.1032431118832196e-05, + "loss": 1.0988, + "step": 39224 + }, + { + "epoch": 2.9, + "learning_rate": 1.1032034377900475e-05, + "loss": 1.0644, + "step": 39225 + }, + { + "epoch": 2.9, + "learning_rate": 1.10316376353268e-05, + "loss": 0.9207, + "step": 39226 + }, + { + "epoch": 2.9, + "learning_rate": 1.1031240891111805e-05, + "loss": 0.9441, + "step": 39227 + }, + { + "epoch": 2.9, + "learning_rate": 1.1030844145256116e-05, + "loss": 1.0341, + "step": 39228 + }, + { + "epoch": 2.9, + "learning_rate": 1.1030447397760376e-05, + "loss": 0.8998, + "step": 39229 + }, + { + "epoch": 2.9, + "learning_rate": 1.1030050648625203e-05, + "loss": 1.0237, + "step": 39230 + }, + { + "epoch": 2.9, + "learning_rate": 1.1029653897851238e-05, + "loss": 0.9844, + "step": 39231 + }, + { + "epoch": 2.9, + "learning_rate": 1.1029257145439105e-05, + "loss": 1.0458, + "step": 39232 + }, + { + "epoch": 2.9, + "learning_rate": 1.1028860391389445e-05, + "loss": 1.0762, + "step": 39233 + }, + { + "epoch": 2.9, + "learning_rate": 1.1028463635702877e-05, + "loss": 1.087, + "step": 39234 + }, + { + "epoch": 2.9, + "learning_rate": 1.102806687838004e-05, + "loss": 0.9405, + "step": 39235 + }, + { + "epoch": 2.9, + "learning_rate": 1.1027670119421562e-05, + "loss": 0.9219, + "step": 39236 + }, + { + "epoch": 2.9, + "learning_rate": 1.1027273358828076e-05, + "loss": 1.0297, + "step": 39237 + }, + { + "epoch": 2.9, + "learning_rate": 1.1026876596600213e-05, + "loss": 1.0178, + "step": 39238 + }, + { + "epoch": 2.9, + "learning_rate": 1.1026479832738605e-05, + "loss": 0.9655, + "step": 39239 + }, + { + "epoch": 2.9, + "learning_rate": 1.102608306724388e-05, + "loss": 0.9889, + "step": 39240 + }, + { + "epoch": 2.9, + "learning_rate": 1.1025686300116674e-05, + "loss": 1.0029, + "step": 39241 + }, + { + "epoch": 2.9, + "learning_rate": 1.1025289531357614e-05, + "loss": 0.9939, + "step": 39242 + }, + { + "epoch": 2.9, + "learning_rate": 1.1024892760967335e-05, + "loss": 0.9384, + "step": 39243 + }, + { + "epoch": 2.9, + "learning_rate": 1.1024495988946464e-05, + "loss": 0.9966, + "step": 39244 + }, + { + "epoch": 2.9, + "learning_rate": 1.1024099215295638e-05, + "loss": 0.9596, + "step": 39245 + }, + { + "epoch": 2.9, + "learning_rate": 1.102370244001548e-05, + "loss": 0.9827, + "step": 39246 + }, + { + "epoch": 2.9, + "learning_rate": 1.102330566310663e-05, + "loss": 1.0731, + "step": 39247 + }, + { + "epoch": 2.9, + "learning_rate": 1.1022908884569711e-05, + "loss": 1.0876, + "step": 39248 + }, + { + "epoch": 2.9, + "learning_rate": 1.1022512104405363e-05, + "loss": 0.9391, + "step": 39249 + }, + { + "epoch": 2.9, + "learning_rate": 1.1022115322614211e-05, + "loss": 1.0664, + "step": 39250 + }, + { + "epoch": 2.9, + "learning_rate": 1.1021718539196889e-05, + "loss": 0.9446, + "step": 39251 + }, + { + "epoch": 2.9, + "learning_rate": 1.102132175415403e-05, + "loss": 1.0277, + "step": 39252 + }, + { + "epoch": 2.9, + "learning_rate": 1.1020924967486257e-05, + "loss": 1.101, + "step": 39253 + }, + { + "epoch": 2.9, + "learning_rate": 1.102052817919421e-05, + "loss": 0.985, + "step": 39254 + }, + { + "epoch": 2.9, + "learning_rate": 1.1020131389278517e-05, + "loss": 1.0511, + "step": 39255 + }, + { + "epoch": 2.9, + "learning_rate": 1.1019734597739808e-05, + "loss": 0.8978, + "step": 39256 + }, + { + "epoch": 2.9, + "learning_rate": 1.1019337804578718e-05, + "loss": 0.9757, + "step": 39257 + }, + { + "epoch": 2.9, + "learning_rate": 1.1018941009795874e-05, + "loss": 1.0589, + "step": 39258 + }, + { + "epoch": 2.9, + "learning_rate": 1.1018544213391913e-05, + "loss": 1.0055, + "step": 39259 + }, + { + "epoch": 2.9, + "learning_rate": 1.101814741536746e-05, + "loss": 0.9937, + "step": 39260 + }, + { + "epoch": 2.9, + "learning_rate": 1.101775061572315e-05, + "loss": 1.0029, + "step": 39261 + }, + { + "epoch": 2.9, + "learning_rate": 1.1017353814459613e-05, + "loss": 0.9383, + "step": 39262 + }, + { + "epoch": 2.9, + "learning_rate": 1.101695701157748e-05, + "loss": 0.9998, + "step": 39263 + }, + { + "epoch": 2.9, + "learning_rate": 1.1016560207077383e-05, + "loss": 1.0191, + "step": 39264 + }, + { + "epoch": 2.9, + "learning_rate": 1.1016163400959954e-05, + "loss": 0.9439, + "step": 39265 + }, + { + "epoch": 2.9, + "learning_rate": 1.1015766593225822e-05, + "loss": 1.0242, + "step": 39266 + }, + { + "epoch": 2.9, + "learning_rate": 1.1015369783875623e-05, + "loss": 0.927, + "step": 39267 + }, + { + "epoch": 2.9, + "learning_rate": 1.1014972972909984e-05, + "loss": 1.0267, + "step": 39268 + }, + { + "epoch": 2.9, + "learning_rate": 1.1014576160329536e-05, + "loss": 1.0776, + "step": 39269 + }, + { + "epoch": 2.9, + "learning_rate": 1.1014179346134913e-05, + "loss": 0.997, + "step": 39270 + }, + { + "epoch": 2.9, + "learning_rate": 1.1013782530326744e-05, + "loss": 1.0083, + "step": 39271 + }, + { + "epoch": 2.9, + "learning_rate": 1.1013385712905662e-05, + "loss": 0.9602, + "step": 39272 + }, + { + "epoch": 2.9, + "learning_rate": 1.1012988893872299e-05, + "loss": 1.0494, + "step": 39273 + }, + { + "epoch": 2.9, + "learning_rate": 1.1012592073227284e-05, + "loss": 0.9854, + "step": 39274 + }, + { + "epoch": 2.9, + "learning_rate": 1.101219525097125e-05, + "loss": 0.9594, + "step": 39275 + }, + { + "epoch": 2.9, + "learning_rate": 1.1011798427104827e-05, + "loss": 0.9929, + "step": 39276 + }, + { + "epoch": 2.9, + "learning_rate": 1.1011401601628648e-05, + "loss": 1.0036, + "step": 39277 + }, + { + "epoch": 2.9, + "learning_rate": 1.1011004774543344e-05, + "loss": 0.9534, + "step": 39278 + }, + { + "epoch": 2.9, + "learning_rate": 1.1010607945849546e-05, + "loss": 0.9909, + "step": 39279 + }, + { + "epoch": 2.9, + "learning_rate": 1.1010211115547884e-05, + "loss": 1.0767, + "step": 39280 + }, + { + "epoch": 2.9, + "learning_rate": 1.1009814283638992e-05, + "loss": 0.9869, + "step": 39281 + }, + { + "epoch": 2.9, + "learning_rate": 1.1009417450123498e-05, + "loss": 0.9676, + "step": 39282 + }, + { + "epoch": 2.9, + "learning_rate": 1.1009020615002034e-05, + "loss": 0.9976, + "step": 39283 + }, + { + "epoch": 2.9, + "learning_rate": 1.1008623778275239e-05, + "loss": 1.0334, + "step": 39284 + }, + { + "epoch": 2.9, + "learning_rate": 1.100822693994373e-05, + "loss": 0.9015, + "step": 39285 + }, + { + "epoch": 2.9, + "learning_rate": 1.1007830100008152e-05, + "loss": 1.0806, + "step": 39286 + }, + { + "epoch": 2.9, + "learning_rate": 1.1007433258469128e-05, + "loss": 0.9395, + "step": 39287 + }, + { + "epoch": 2.9, + "learning_rate": 1.1007036415327295e-05, + "loss": 1.0996, + "step": 39288 + }, + { + "epoch": 2.9, + "learning_rate": 1.100663957058328e-05, + "loss": 0.9445, + "step": 39289 + }, + { + "epoch": 2.9, + "learning_rate": 1.1006242724237717e-05, + "loss": 1.0963, + "step": 39290 + }, + { + "epoch": 2.9, + "learning_rate": 1.1005845876291232e-05, + "loss": 1.0073, + "step": 39291 + }, + { + "epoch": 2.9, + "learning_rate": 1.1005449026744464e-05, + "loss": 0.9317, + "step": 39292 + }, + { + "epoch": 2.9, + "learning_rate": 1.1005052175598039e-05, + "loss": 0.9979, + "step": 39293 + }, + { + "epoch": 2.9, + "learning_rate": 1.1004655322852593e-05, + "loss": 1.0145, + "step": 39294 + }, + { + "epoch": 2.9, + "learning_rate": 1.1004258468508752e-05, + "loss": 1.0116, + "step": 39295 + }, + { + "epoch": 2.9, + "learning_rate": 1.1003861612567153e-05, + "loss": 0.9431, + "step": 39296 + }, + { + "epoch": 2.9, + "learning_rate": 1.100346475502842e-05, + "loss": 0.94, + "step": 39297 + }, + { + "epoch": 2.9, + "learning_rate": 1.1003067895893195e-05, + "loss": 0.9998, + "step": 39298 + }, + { + "epoch": 2.9, + "learning_rate": 1.10026710351621e-05, + "loss": 0.9835, + "step": 39299 + }, + { + "epoch": 2.9, + "learning_rate": 1.1002274172835772e-05, + "loss": 1.0462, + "step": 39300 + }, + { + "epoch": 2.9, + "learning_rate": 1.1001877308914836e-05, + "loss": 0.9822, + "step": 39301 + }, + { + "epoch": 2.9, + "learning_rate": 1.100148044339993e-05, + "loss": 1.0304, + "step": 39302 + }, + { + "epoch": 2.9, + "learning_rate": 1.1001083576291682e-05, + "loss": 0.9165, + "step": 39303 + }, + { + "epoch": 2.9, + "learning_rate": 1.1000686707590727e-05, + "loss": 1.0486, + "step": 39304 + }, + { + "epoch": 2.9, + "learning_rate": 1.1000289837297688e-05, + "loss": 1.0311, + "step": 39305 + }, + { + "epoch": 2.9, + "learning_rate": 1.0999892965413207e-05, + "loss": 1.0645, + "step": 39306 + }, + { + "epoch": 2.9, + "learning_rate": 1.099949609193791e-05, + "loss": 0.9842, + "step": 39307 + }, + { + "epoch": 2.9, + "learning_rate": 1.099909921687243e-05, + "loss": 1.0783, + "step": 39308 + }, + { + "epoch": 2.9, + "learning_rate": 1.0998702340217396e-05, + "loss": 0.9547, + "step": 39309 + }, + { + "epoch": 2.9, + "learning_rate": 1.0998305461973441e-05, + "loss": 1.0097, + "step": 39310 + }, + { + "epoch": 2.9, + "learning_rate": 1.0997908582141195e-05, + "loss": 1.0535, + "step": 39311 + }, + { + "epoch": 2.9, + "learning_rate": 1.0997511700721293e-05, + "loss": 0.9305, + "step": 39312 + }, + { + "epoch": 2.9, + "learning_rate": 1.0997114817714362e-05, + "loss": 1.0041, + "step": 39313 + }, + { + "epoch": 2.9, + "learning_rate": 1.0996717933121037e-05, + "loss": 1.0435, + "step": 39314 + }, + { + "epoch": 2.9, + "learning_rate": 1.0996321046941947e-05, + "loss": 1.0761, + "step": 39315 + }, + { + "epoch": 2.91, + "learning_rate": 1.0995924159177726e-05, + "loss": 1.0448, + "step": 39316 + }, + { + "epoch": 2.91, + "learning_rate": 1.0995527269829004e-05, + "loss": 1.1086, + "step": 39317 + }, + { + "epoch": 2.91, + "learning_rate": 1.099513037889641e-05, + "loss": 0.8646, + "step": 39318 + }, + { + "epoch": 2.91, + "learning_rate": 1.099473348638058e-05, + "loss": 0.9505, + "step": 39319 + }, + { + "epoch": 2.91, + "learning_rate": 1.0994336592282143e-05, + "loss": 1.0434, + "step": 39320 + }, + { + "epoch": 2.91, + "learning_rate": 1.099393969660173e-05, + "loss": 1.0134, + "step": 39321 + }, + { + "epoch": 2.91, + "learning_rate": 1.0993542799339973e-05, + "loss": 0.9759, + "step": 39322 + }, + { + "epoch": 2.91, + "learning_rate": 1.0993145900497504e-05, + "loss": 0.9832, + "step": 39323 + }, + { + "epoch": 2.91, + "learning_rate": 1.0992749000074954e-05, + "loss": 0.9442, + "step": 39324 + }, + { + "epoch": 2.91, + "learning_rate": 1.0992352098072953e-05, + "loss": 1.0276, + "step": 39325 + }, + { + "epoch": 2.91, + "learning_rate": 1.0991955194492137e-05, + "loss": 1.0086, + "step": 39326 + }, + { + "epoch": 2.91, + "learning_rate": 1.0991558289333133e-05, + "loss": 1.0465, + "step": 39327 + }, + { + "epoch": 2.91, + "learning_rate": 1.0991161382596574e-05, + "loss": 1.0636, + "step": 39328 + }, + { + "epoch": 2.91, + "learning_rate": 1.0990764474283092e-05, + "loss": 0.9744, + "step": 39329 + }, + { + "epoch": 2.91, + "learning_rate": 1.0990367564393317e-05, + "loss": 1.0287, + "step": 39330 + }, + { + "epoch": 2.91, + "learning_rate": 1.0989970652927881e-05, + "loss": 0.9244, + "step": 39331 + }, + { + "epoch": 2.91, + "learning_rate": 1.0989573739887416e-05, + "loss": 0.9897, + "step": 39332 + }, + { + "epoch": 2.91, + "learning_rate": 1.0989176825272555e-05, + "loss": 1.0091, + "step": 39333 + }, + { + "epoch": 2.91, + "learning_rate": 1.0988779909083924e-05, + "loss": 0.9029, + "step": 39334 + }, + { + "epoch": 2.91, + "learning_rate": 1.0988382991322164e-05, + "loss": 0.9191, + "step": 39335 + }, + { + "epoch": 2.91, + "learning_rate": 1.0987986071987898e-05, + "loss": 0.9539, + "step": 39336 + }, + { + "epoch": 2.91, + "learning_rate": 1.0987589151081761e-05, + "loss": 1.0247, + "step": 39337 + }, + { + "epoch": 2.91, + "learning_rate": 1.0987192228604382e-05, + "loss": 0.9493, + "step": 39338 + }, + { + "epoch": 2.91, + "learning_rate": 1.0986795304556394e-05, + "loss": 1.0219, + "step": 39339 + }, + { + "epoch": 2.91, + "learning_rate": 1.098639837893843e-05, + "loss": 0.9577, + "step": 39340 + }, + { + "epoch": 2.91, + "learning_rate": 1.0986001451751123e-05, + "loss": 0.9976, + "step": 39341 + }, + { + "epoch": 2.91, + "learning_rate": 1.0985604522995097e-05, + "loss": 0.9931, + "step": 39342 + }, + { + "epoch": 2.91, + "learning_rate": 1.098520759267099e-05, + "loss": 1.0181, + "step": 39343 + }, + { + "epoch": 2.91, + "learning_rate": 1.0984810660779432e-05, + "loss": 1.0767, + "step": 39344 + }, + { + "epoch": 2.91, + "learning_rate": 1.0984413727321057e-05, + "loss": 0.9912, + "step": 39345 + }, + { + "epoch": 2.91, + "learning_rate": 1.098401679229649e-05, + "loss": 0.9615, + "step": 39346 + }, + { + "epoch": 2.91, + "learning_rate": 1.098361985570637e-05, + "loss": 1.068, + "step": 39347 + }, + { + "epoch": 2.91, + "learning_rate": 1.0983222917551319e-05, + "loss": 1.0262, + "step": 39348 + }, + { + "epoch": 2.91, + "learning_rate": 1.0982825977831981e-05, + "loss": 1.03, + "step": 39349 + }, + { + "epoch": 2.91, + "learning_rate": 1.0982429036548976e-05, + "loss": 1.0837, + "step": 39350 + }, + { + "epoch": 2.91, + "learning_rate": 1.0982032093702944e-05, + "loss": 1.0425, + "step": 39351 + }, + { + "epoch": 2.91, + "learning_rate": 1.098163514929451e-05, + "loss": 0.9483, + "step": 39352 + }, + { + "epoch": 2.91, + "learning_rate": 1.098123820332431e-05, + "loss": 1.078, + "step": 39353 + }, + { + "epoch": 2.91, + "learning_rate": 1.0980841255792976e-05, + "loss": 1.057, + "step": 39354 + }, + { + "epoch": 2.91, + "learning_rate": 1.0980444306701136e-05, + "loss": 0.9802, + "step": 39355 + }, + { + "epoch": 2.91, + "learning_rate": 1.0980047356049422e-05, + "loss": 0.9628, + "step": 39356 + }, + { + "epoch": 2.91, + "learning_rate": 1.0979650403838466e-05, + "loss": 0.9968, + "step": 39357 + }, + { + "epoch": 2.91, + "learning_rate": 1.0979253450068901e-05, + "loss": 1.0015, + "step": 39358 + }, + { + "epoch": 2.91, + "learning_rate": 1.097885649474136e-05, + "loss": 0.9977, + "step": 39359 + }, + { + "epoch": 2.91, + "learning_rate": 1.097845953785647e-05, + "loss": 1.0028, + "step": 39360 + }, + { + "epoch": 2.91, + "learning_rate": 1.0978062579414865e-05, + "loss": 0.9364, + "step": 39361 + }, + { + "epoch": 2.91, + "learning_rate": 1.0977665619417177e-05, + "loss": 1.0204, + "step": 39362 + }, + { + "epoch": 2.91, + "learning_rate": 1.0977268657864037e-05, + "loss": 0.9956, + "step": 39363 + }, + { + "epoch": 2.91, + "learning_rate": 1.0976871694756077e-05, + "loss": 1.0677, + "step": 39364 + }, + { + "epoch": 2.91, + "learning_rate": 1.0976474730093926e-05, + "loss": 1.0435, + "step": 39365 + }, + { + "epoch": 2.91, + "learning_rate": 1.097607776387822e-05, + "loss": 1.0435, + "step": 39366 + }, + { + "epoch": 2.91, + "learning_rate": 1.0975680796109592e-05, + "loss": 0.9796, + "step": 39367 + }, + { + "epoch": 2.91, + "learning_rate": 1.0975283826788663e-05, + "loss": 0.9516, + "step": 39368 + }, + { + "epoch": 2.91, + "learning_rate": 1.0974886855916074e-05, + "loss": 1.0256, + "step": 39369 + }, + { + "epoch": 2.91, + "learning_rate": 1.0974489883492452e-05, + "loss": 1.0028, + "step": 39370 + }, + { + "epoch": 2.91, + "learning_rate": 1.0974092909518434e-05, + "loss": 0.9541, + "step": 39371 + }, + { + "epoch": 2.91, + "learning_rate": 1.0973695933994646e-05, + "loss": 1.0832, + "step": 39372 + }, + { + "epoch": 2.91, + "learning_rate": 1.0973298956921723e-05, + "loss": 1.0617, + "step": 39373 + }, + { + "epoch": 2.91, + "learning_rate": 1.0972901978300295e-05, + "loss": 0.9799, + "step": 39374 + }, + { + "epoch": 2.91, + "learning_rate": 1.097250499813099e-05, + "loss": 1.0609, + "step": 39375 + }, + { + "epoch": 2.91, + "learning_rate": 1.097210801641445e-05, + "loss": 0.9571, + "step": 39376 + }, + { + "epoch": 2.91, + "learning_rate": 1.0971711033151293e-05, + "loss": 1.0241, + "step": 39377 + }, + { + "epoch": 2.91, + "learning_rate": 1.0971314048342164e-05, + "loss": 0.9696, + "step": 39378 + }, + { + "epoch": 2.91, + "learning_rate": 1.0970917061987686e-05, + "loss": 0.9879, + "step": 39379 + }, + { + "epoch": 2.91, + "learning_rate": 1.097052007408849e-05, + "loss": 1.0865, + "step": 39380 + }, + { + "epoch": 2.91, + "learning_rate": 1.0970123084645212e-05, + "loss": 1.089, + "step": 39381 + }, + { + "epoch": 2.91, + "learning_rate": 1.0969726093658488e-05, + "loss": 0.9217, + "step": 39382 + }, + { + "epoch": 2.91, + "learning_rate": 1.0969329101128937e-05, + "loss": 1.0293, + "step": 39383 + }, + { + "epoch": 2.91, + "learning_rate": 1.0968932107057199e-05, + "loss": 1.1176, + "step": 39384 + }, + { + "epoch": 2.91, + "learning_rate": 1.0968535111443904e-05, + "loss": 1.0714, + "step": 39385 + }, + { + "epoch": 2.91, + "learning_rate": 1.0968138114289684e-05, + "loss": 1.0108, + "step": 39386 + }, + { + "epoch": 2.91, + "learning_rate": 1.0967741115595168e-05, + "loss": 0.977, + "step": 39387 + }, + { + "epoch": 2.91, + "learning_rate": 1.0967344115360991e-05, + "loss": 1.0266, + "step": 39388 + }, + { + "epoch": 2.91, + "learning_rate": 1.0966947113587783e-05, + "loss": 0.9839, + "step": 39389 + }, + { + "epoch": 2.91, + "learning_rate": 1.0966550110276177e-05, + "loss": 0.9126, + "step": 39390 + }, + { + "epoch": 2.91, + "learning_rate": 1.0966153105426803e-05, + "loss": 0.9643, + "step": 39391 + }, + { + "epoch": 2.91, + "learning_rate": 1.0965756099040295e-05, + "loss": 0.9863, + "step": 39392 + }, + { + "epoch": 2.91, + "learning_rate": 1.096535909111728e-05, + "loss": 0.9971, + "step": 39393 + }, + { + "epoch": 2.91, + "learning_rate": 1.0964962081658395e-05, + "loss": 0.9414, + "step": 39394 + }, + { + "epoch": 2.91, + "learning_rate": 1.0964565070664267e-05, + "loss": 0.9587, + "step": 39395 + }, + { + "epoch": 2.91, + "learning_rate": 1.0964168058135532e-05, + "loss": 1.0834, + "step": 39396 + }, + { + "epoch": 2.91, + "learning_rate": 1.0963771044072816e-05, + "loss": 0.9804, + "step": 39397 + }, + { + "epoch": 2.91, + "learning_rate": 1.0963374028476756e-05, + "loss": 1.0336, + "step": 39398 + }, + { + "epoch": 2.91, + "learning_rate": 1.0962977011347982e-05, + "loss": 1.001, + "step": 39399 + }, + { + "epoch": 2.91, + "learning_rate": 1.0962579992687125e-05, + "loss": 0.9882, + "step": 39400 + }, + { + "epoch": 2.91, + "learning_rate": 1.096218297249482e-05, + "loss": 1.0816, + "step": 39401 + }, + { + "epoch": 2.91, + "learning_rate": 1.0961785950771693e-05, + "loss": 0.9669, + "step": 39402 + }, + { + "epoch": 2.91, + "learning_rate": 1.0961388927518376e-05, + "loss": 0.9733, + "step": 39403 + }, + { + "epoch": 2.91, + "learning_rate": 1.0960991902735505e-05, + "loss": 1.0846, + "step": 39404 + }, + { + "epoch": 2.91, + "learning_rate": 1.096059487642371e-05, + "loss": 0.991, + "step": 39405 + }, + { + "epoch": 2.91, + "learning_rate": 1.0960197848583623e-05, + "loss": 1.0089, + "step": 39406 + }, + { + "epoch": 2.91, + "learning_rate": 1.0959800819215872e-05, + "loss": 1.0812, + "step": 39407 + }, + { + "epoch": 2.91, + "learning_rate": 1.0959403788321094e-05, + "loss": 0.9878, + "step": 39408 + }, + { + "epoch": 2.91, + "learning_rate": 1.0959006755899916e-05, + "loss": 0.891, + "step": 39409 + }, + { + "epoch": 2.91, + "learning_rate": 1.0958609721952976e-05, + "loss": 1.1238, + "step": 39410 + }, + { + "epoch": 2.91, + "learning_rate": 1.09582126864809e-05, + "loss": 0.9869, + "step": 39411 + }, + { + "epoch": 2.91, + "learning_rate": 1.0957815649484323e-05, + "loss": 0.9288, + "step": 39412 + }, + { + "epoch": 2.91, + "learning_rate": 1.095741861096387e-05, + "loss": 0.9906, + "step": 39413 + }, + { + "epoch": 2.91, + "learning_rate": 1.095702157092018e-05, + "loss": 1.0074, + "step": 39414 + }, + { + "epoch": 2.91, + "learning_rate": 1.0956624529353883e-05, + "loss": 0.8908, + "step": 39415 + }, + { + "epoch": 2.91, + "learning_rate": 1.0956227486265612e-05, + "loss": 1.042, + "step": 39416 + }, + { + "epoch": 2.91, + "learning_rate": 1.0955830441655992e-05, + "loss": 0.9735, + "step": 39417 + }, + { + "epoch": 2.91, + "learning_rate": 1.0955433395525664e-05, + "loss": 1.0499, + "step": 39418 + }, + { + "epoch": 2.91, + "learning_rate": 1.0955036347875252e-05, + "loss": 1.0627, + "step": 39419 + }, + { + "epoch": 2.91, + "learning_rate": 1.0954639298705395e-05, + "loss": 0.9314, + "step": 39420 + }, + { + "epoch": 2.91, + "learning_rate": 1.0954242248016718e-05, + "loss": 0.9814, + "step": 39421 + }, + { + "epoch": 2.91, + "learning_rate": 1.0953845195809851e-05, + "loss": 1.0708, + "step": 39422 + }, + { + "epoch": 2.91, + "learning_rate": 1.0953448142085437e-05, + "loss": 0.9956, + "step": 39423 + }, + { + "epoch": 2.91, + "learning_rate": 1.0953051086844095e-05, + "loss": 0.9852, + "step": 39424 + }, + { + "epoch": 2.91, + "learning_rate": 1.0952654030086466e-05, + "loss": 1.0114, + "step": 39425 + }, + { + "epoch": 2.91, + "learning_rate": 1.0952256971813175e-05, + "loss": 0.9779, + "step": 39426 + }, + { + "epoch": 2.91, + "learning_rate": 1.095185991202486e-05, + "loss": 1.0506, + "step": 39427 + }, + { + "epoch": 2.91, + "learning_rate": 1.095146285072215e-05, + "loss": 1.0108, + "step": 39428 + }, + { + "epoch": 2.91, + "learning_rate": 1.0951065787905675e-05, + "loss": 1.0959, + "step": 39429 + }, + { + "epoch": 2.91, + "learning_rate": 1.0950668723576065e-05, + "loss": 0.9711, + "step": 39430 + }, + { + "epoch": 2.91, + "learning_rate": 1.095027165773396e-05, + "loss": 1.0887, + "step": 39431 + }, + { + "epoch": 2.91, + "learning_rate": 1.0949874590379981e-05, + "loss": 0.998, + "step": 39432 + }, + { + "epoch": 2.91, + "learning_rate": 1.0949477521514768e-05, + "loss": 1.0746, + "step": 39433 + }, + { + "epoch": 2.91, + "learning_rate": 1.094908045113895e-05, + "loss": 1.0831, + "step": 39434 + }, + { + "epoch": 2.91, + "learning_rate": 1.0948683379253159e-05, + "loss": 1.0151, + "step": 39435 + }, + { + "epoch": 2.91, + "learning_rate": 1.0948286305858022e-05, + "loss": 1.0828, + "step": 39436 + }, + { + "epoch": 2.91, + "learning_rate": 1.094788923095418e-05, + "loss": 0.9988, + "step": 39437 + }, + { + "epoch": 2.91, + "learning_rate": 1.0947492154542257e-05, + "loss": 0.9689, + "step": 39438 + }, + { + "epoch": 2.91, + "learning_rate": 1.094709507662289e-05, + "loss": 0.9066, + "step": 39439 + }, + { + "epoch": 2.91, + "learning_rate": 1.0946697997196706e-05, + "loss": 1.0878, + "step": 39440 + }, + { + "epoch": 2.91, + "learning_rate": 1.094630091626434e-05, + "loss": 1.0488, + "step": 39441 + }, + { + "epoch": 2.91, + "learning_rate": 1.0945903833826424e-05, + "loss": 1.0371, + "step": 39442 + }, + { + "epoch": 2.91, + "learning_rate": 1.0945506749883588e-05, + "loss": 0.9578, + "step": 39443 + }, + { + "epoch": 2.91, + "learning_rate": 1.0945109664436462e-05, + "loss": 0.9816, + "step": 39444 + }, + { + "epoch": 2.91, + "learning_rate": 1.0944712577485683e-05, + "loss": 1.0635, + "step": 39445 + }, + { + "epoch": 2.91, + "learning_rate": 1.0944315489031876e-05, + "loss": 1.1064, + "step": 39446 + }, + { + "epoch": 2.91, + "learning_rate": 1.094391839907568e-05, + "loss": 1.0044, + "step": 39447 + }, + { + "epoch": 2.91, + "learning_rate": 1.0943521307617723e-05, + "loss": 1.0141, + "step": 39448 + }, + { + "epoch": 2.91, + "learning_rate": 1.0943124214658638e-05, + "loss": 0.9301, + "step": 39449 + }, + { + "epoch": 2.91, + "learning_rate": 1.0942727120199052e-05, + "loss": 1.1488, + "step": 39450 + }, + { + "epoch": 2.92, + "learning_rate": 1.0942330024239605e-05, + "loss": 0.9783, + "step": 39451 + }, + { + "epoch": 2.92, + "learning_rate": 1.094193292678092e-05, + "loss": 1.0265, + "step": 39452 + }, + { + "epoch": 2.92, + "learning_rate": 1.0941535827823637e-05, + "loss": 1.0628, + "step": 39453 + }, + { + "epoch": 2.92, + "learning_rate": 1.0941138727368381e-05, + "loss": 1.0728, + "step": 39454 + }, + { + "epoch": 2.92, + "learning_rate": 1.094074162541579e-05, + "loss": 1.0636, + "step": 39455 + }, + { + "epoch": 2.92, + "learning_rate": 1.0940344521966488e-05, + "loss": 1.1022, + "step": 39456 + }, + { + "epoch": 2.92, + "learning_rate": 1.0939947417021117e-05, + "loss": 0.9317, + "step": 39457 + }, + { + "epoch": 2.92, + "learning_rate": 1.09395503105803e-05, + "loss": 0.9881, + "step": 39458 + }, + { + "epoch": 2.92, + "learning_rate": 1.0939153202644674e-05, + "loss": 1.0103, + "step": 39459 + }, + { + "epoch": 2.92, + "learning_rate": 1.0938756093214864e-05, + "loss": 1.0478, + "step": 39460 + }, + { + "epoch": 2.92, + "learning_rate": 1.093835898229151e-05, + "loss": 1.0757, + "step": 39461 + }, + { + "epoch": 2.92, + "learning_rate": 1.093796186987524e-05, + "loss": 1.0584, + "step": 39462 + }, + { + "epoch": 2.92, + "learning_rate": 1.0937564755966685e-05, + "loss": 1.0201, + "step": 39463 + }, + { + "epoch": 2.92, + "learning_rate": 1.0937167640566477e-05, + "loss": 0.9332, + "step": 39464 + }, + { + "epoch": 2.92, + "learning_rate": 1.0936770523675253e-05, + "loss": 0.9458, + "step": 39465 + }, + { + "epoch": 2.92, + "learning_rate": 1.0936373405293636e-05, + "loss": 0.9513, + "step": 39466 + }, + { + "epoch": 2.92, + "learning_rate": 1.0935976285422266e-05, + "loss": 1.008, + "step": 39467 + }, + { + "epoch": 2.92, + "learning_rate": 1.093557916406177e-05, + "loss": 1.0454, + "step": 39468 + }, + { + "epoch": 2.92, + "learning_rate": 1.0935182041212777e-05, + "loss": 1.0906, + "step": 39469 + }, + { + "epoch": 2.92, + "learning_rate": 1.0934784916875927e-05, + "loss": 0.9999, + "step": 39470 + }, + { + "epoch": 2.92, + "learning_rate": 1.0934387791051848e-05, + "loss": 1.0565, + "step": 39471 + }, + { + "epoch": 2.92, + "learning_rate": 1.0933990663741167e-05, + "loss": 1.0129, + "step": 39472 + }, + { + "epoch": 2.92, + "learning_rate": 1.0933593534944526e-05, + "loss": 0.9431, + "step": 39473 + }, + { + "epoch": 2.92, + "learning_rate": 1.0933196404662546e-05, + "loss": 1.1002, + "step": 39474 + }, + { + "epoch": 2.92, + "learning_rate": 1.0932799272895866e-05, + "loss": 1.0507, + "step": 39475 + }, + { + "epoch": 2.92, + "learning_rate": 1.0932402139645117e-05, + "loss": 1.0986, + "step": 39476 + }, + { + "epoch": 2.92, + "learning_rate": 1.0932005004910928e-05, + "loss": 1.1157, + "step": 39477 + }, + { + "epoch": 2.92, + "learning_rate": 1.0931607868693934e-05, + "loss": 1.0634, + "step": 39478 + }, + { + "epoch": 2.92, + "learning_rate": 1.0931210730994762e-05, + "loss": 0.9839, + "step": 39479 + }, + { + "epoch": 2.92, + "learning_rate": 1.0930813591814049e-05, + "loss": 1.0733, + "step": 39480 + }, + { + "epoch": 2.92, + "learning_rate": 1.0930416451152423e-05, + "loss": 0.9284, + "step": 39481 + }, + { + "epoch": 2.92, + "learning_rate": 1.0930019309010522e-05, + "loss": 1.1319, + "step": 39482 + }, + { + "epoch": 2.92, + "learning_rate": 1.092962216538897e-05, + "loss": 1.2127, + "step": 39483 + }, + { + "epoch": 2.92, + "learning_rate": 1.0929225020288403e-05, + "loss": 1.005, + "step": 39484 + }, + { + "epoch": 2.92, + "learning_rate": 1.0928827873709454e-05, + "loss": 0.9945, + "step": 39485 + }, + { + "epoch": 2.92, + "learning_rate": 1.0928430725652752e-05, + "loss": 0.9942, + "step": 39486 + }, + { + "epoch": 2.92, + "learning_rate": 1.092803357611893e-05, + "loss": 0.9532, + "step": 39487 + }, + { + "epoch": 2.92, + "learning_rate": 1.0927636425108621e-05, + "loss": 1.01, + "step": 39488 + }, + { + "epoch": 2.92, + "learning_rate": 1.0927239272622453e-05, + "loss": 0.9394, + "step": 39489 + }, + { + "epoch": 2.92, + "learning_rate": 1.0926842118661062e-05, + "loss": 1.0524, + "step": 39490 + }, + { + "epoch": 2.92, + "learning_rate": 1.0926444963225078e-05, + "loss": 0.957, + "step": 39491 + }, + { + "epoch": 2.92, + "learning_rate": 1.0926047806315134e-05, + "loss": 0.9929, + "step": 39492 + }, + { + "epoch": 2.92, + "learning_rate": 1.092565064793186e-05, + "loss": 1.0417, + "step": 39493 + }, + { + "epoch": 2.92, + "learning_rate": 1.0925253488075892e-05, + "loss": 0.9911, + "step": 39494 + }, + { + "epoch": 2.92, + "learning_rate": 1.0924856326747858e-05, + "loss": 1.0026, + "step": 39495 + }, + { + "epoch": 2.92, + "learning_rate": 1.0924459163948392e-05, + "loss": 0.9351, + "step": 39496 + }, + { + "epoch": 2.92, + "learning_rate": 1.092406199967812e-05, + "loss": 1.137, + "step": 39497 + }, + { + "epoch": 2.92, + "learning_rate": 1.0923664833937682e-05, + "loss": 1.0478, + "step": 39498 + }, + { + "epoch": 2.92, + "learning_rate": 1.0923267666727704e-05, + "loss": 1.0923, + "step": 39499 + }, + { + "epoch": 2.92, + "learning_rate": 1.0922870498048823e-05, + "loss": 1.0342, + "step": 39500 + }, + { + "epoch": 2.92, + "learning_rate": 1.0922473327901665e-05, + "loss": 0.9406, + "step": 39501 + }, + { + "epoch": 2.92, + "learning_rate": 1.092207615628687e-05, + "loss": 0.9304, + "step": 39502 + }, + { + "epoch": 2.92, + "learning_rate": 1.092167898320506e-05, + "loss": 0.9439, + "step": 39503 + }, + { + "epoch": 2.92, + "learning_rate": 1.0921281808656879e-05, + "loss": 0.9504, + "step": 39504 + }, + { + "epoch": 2.92, + "learning_rate": 1.0920884632642946e-05, + "loss": 0.9638, + "step": 39505 + }, + { + "epoch": 2.92, + "learning_rate": 1.0920487455163899e-05, + "loss": 1.0562, + "step": 39506 + }, + { + "epoch": 2.92, + "learning_rate": 1.092009027622037e-05, + "loss": 1.0846, + "step": 39507 + }, + { + "epoch": 2.92, + "learning_rate": 1.0919693095812991e-05, + "loss": 1.0493, + "step": 39508 + }, + { + "epoch": 2.92, + "learning_rate": 1.091929591394239e-05, + "loss": 0.9906, + "step": 39509 + }, + { + "epoch": 2.92, + "learning_rate": 1.0918898730609208e-05, + "loss": 0.9533, + "step": 39510 + }, + { + "epoch": 2.92, + "learning_rate": 1.0918501545814069e-05, + "loss": 1.0723, + "step": 39511 + }, + { + "epoch": 2.92, + "learning_rate": 1.091810435955761e-05, + "loss": 1.0193, + "step": 39512 + }, + { + "epoch": 2.92, + "learning_rate": 1.0917707171840457e-05, + "loss": 1.1597, + "step": 39513 + }, + { + "epoch": 2.92, + "learning_rate": 1.0917309982663242e-05, + "loss": 0.9535, + "step": 39514 + }, + { + "epoch": 2.92, + "learning_rate": 1.0916912792026606e-05, + "loss": 0.989, + "step": 39515 + }, + { + "epoch": 2.92, + "learning_rate": 1.0916515599931172e-05, + "loss": 1.0364, + "step": 39516 + }, + { + "epoch": 2.92, + "learning_rate": 1.0916118406377573e-05, + "loss": 0.9941, + "step": 39517 + }, + { + "epoch": 2.92, + "learning_rate": 1.0915721211366445e-05, + "loss": 1.0216, + "step": 39518 + }, + { + "epoch": 2.92, + "learning_rate": 1.0915324014898418e-05, + "loss": 0.9889, + "step": 39519 + }, + { + "epoch": 2.92, + "learning_rate": 1.0914926816974124e-05, + "loss": 0.9403, + "step": 39520 + }, + { + "epoch": 2.92, + "learning_rate": 1.0914529617594192e-05, + "loss": 0.934, + "step": 39521 + }, + { + "epoch": 2.92, + "learning_rate": 1.0914132416759258e-05, + "loss": 1.0609, + "step": 39522 + }, + { + "epoch": 2.92, + "learning_rate": 1.0913735214469954e-05, + "loss": 0.9767, + "step": 39523 + }, + { + "epoch": 2.92, + "learning_rate": 1.0913338010726908e-05, + "loss": 1.085, + "step": 39524 + }, + { + "epoch": 2.92, + "learning_rate": 1.0912940805530757e-05, + "loss": 1.0147, + "step": 39525 + }, + { + "epoch": 2.92, + "learning_rate": 1.0912543598882126e-05, + "loss": 1.0431, + "step": 39526 + }, + { + "epoch": 2.92, + "learning_rate": 1.0912146390781653e-05, + "loss": 1.1689, + "step": 39527 + }, + { + "epoch": 2.92, + "learning_rate": 1.0911749181229967e-05, + "loss": 1.0048, + "step": 39528 + }, + { + "epoch": 2.92, + "learning_rate": 1.0911351970227704e-05, + "loss": 1.0106, + "step": 39529 + }, + { + "epoch": 2.92, + "learning_rate": 1.0910954757775492e-05, + "loss": 1.1014, + "step": 39530 + }, + { + "epoch": 2.92, + "learning_rate": 1.0910557543873963e-05, + "loss": 1.0212, + "step": 39531 + }, + { + "epoch": 2.92, + "learning_rate": 1.0910160328523751e-05, + "loss": 0.9707, + "step": 39532 + }, + { + "epoch": 2.92, + "learning_rate": 1.090976311172549e-05, + "loss": 0.9377, + "step": 39533 + }, + { + "epoch": 2.92, + "learning_rate": 1.0909365893479803e-05, + "loss": 0.9046, + "step": 39534 + }, + { + "epoch": 2.92, + "learning_rate": 1.0908968673787331e-05, + "loss": 0.9677, + "step": 39535 + }, + { + "epoch": 2.92, + "learning_rate": 1.0908571452648701e-05, + "loss": 0.9594, + "step": 39536 + }, + { + "epoch": 2.92, + "learning_rate": 1.090817423006455e-05, + "loss": 0.99, + "step": 39537 + }, + { + "epoch": 2.92, + "learning_rate": 1.0907777006035502e-05, + "loss": 0.9747, + "step": 39538 + }, + { + "epoch": 2.92, + "learning_rate": 1.0907379780562197e-05, + "loss": 0.9991, + "step": 39539 + }, + { + "epoch": 2.92, + "learning_rate": 1.0906982553645263e-05, + "loss": 1.0724, + "step": 39540 + }, + { + "epoch": 2.92, + "learning_rate": 1.0906585325285337e-05, + "loss": 1.0373, + "step": 39541 + }, + { + "epoch": 2.92, + "learning_rate": 1.090618809548304e-05, + "loss": 0.9973, + "step": 39542 + }, + { + "epoch": 2.92, + "learning_rate": 1.0905790864239015e-05, + "loss": 0.9452, + "step": 39543 + }, + { + "epoch": 2.92, + "learning_rate": 1.0905393631553886e-05, + "loss": 1.0603, + "step": 39544 + }, + { + "epoch": 2.92, + "learning_rate": 1.0904996397428293e-05, + "loss": 1.0314, + "step": 39545 + }, + { + "epoch": 2.92, + "learning_rate": 1.0904599161862861e-05, + "loss": 0.9823, + "step": 39546 + }, + { + "epoch": 2.92, + "learning_rate": 1.0904201924858227e-05, + "loss": 0.9922, + "step": 39547 + }, + { + "epoch": 2.92, + "learning_rate": 1.0903804686415016e-05, + "loss": 0.9999, + "step": 39548 + }, + { + "epoch": 2.92, + "learning_rate": 1.090340744653387e-05, + "loss": 0.9883, + "step": 39549 + }, + { + "epoch": 2.92, + "learning_rate": 1.0903010205215412e-05, + "loss": 0.9646, + "step": 39550 + }, + { + "epoch": 2.92, + "learning_rate": 1.0902612962460281e-05, + "loss": 0.9358, + "step": 39551 + }, + { + "epoch": 2.92, + "learning_rate": 1.0902215718269102e-05, + "loss": 1.0486, + "step": 39552 + }, + { + "epoch": 2.92, + "learning_rate": 1.0901818472642513e-05, + "loss": 1.0284, + "step": 39553 + }, + { + "epoch": 2.92, + "learning_rate": 1.0901421225581141e-05, + "loss": 1.0062, + "step": 39554 + }, + { + "epoch": 2.92, + "learning_rate": 1.0901023977085624e-05, + "loss": 1.072, + "step": 39555 + }, + { + "epoch": 2.92, + "learning_rate": 1.0900626727156589e-05, + "loss": 0.8706, + "step": 39556 + }, + { + "epoch": 2.92, + "learning_rate": 1.0900229475794673e-05, + "loss": 0.9561, + "step": 39557 + }, + { + "epoch": 2.92, + "learning_rate": 1.08998322230005e-05, + "loss": 1.0989, + "step": 39558 + }, + { + "epoch": 2.92, + "learning_rate": 1.089943496877471e-05, + "loss": 0.943, + "step": 39559 + }, + { + "epoch": 2.92, + "learning_rate": 1.089903771311793e-05, + "loss": 1.0175, + "step": 39560 + }, + { + "epoch": 2.92, + "learning_rate": 1.0898640456030796e-05, + "loss": 0.9565, + "step": 39561 + }, + { + "epoch": 2.92, + "learning_rate": 1.0898243197513936e-05, + "loss": 1.0306, + "step": 39562 + }, + { + "epoch": 2.92, + "learning_rate": 1.0897845937567985e-05, + "loss": 0.8767, + "step": 39563 + }, + { + "epoch": 2.92, + "learning_rate": 1.089744867619357e-05, + "loss": 0.9668, + "step": 39564 + }, + { + "epoch": 2.92, + "learning_rate": 1.0897051413391334e-05, + "loss": 1.1818, + "step": 39565 + }, + { + "epoch": 2.92, + "learning_rate": 1.0896654149161897e-05, + "loss": 0.9976, + "step": 39566 + }, + { + "epoch": 2.92, + "learning_rate": 1.0896256883505899e-05, + "loss": 0.924, + "step": 39567 + }, + { + "epoch": 2.92, + "learning_rate": 1.0895859616423965e-05, + "loss": 1.0037, + "step": 39568 + }, + { + "epoch": 2.92, + "learning_rate": 1.0895462347916734e-05, + "loss": 1.0392, + "step": 39569 + }, + { + "epoch": 2.92, + "learning_rate": 1.0895065077984837e-05, + "loss": 0.9513, + "step": 39570 + }, + { + "epoch": 2.92, + "learning_rate": 1.08946678066289e-05, + "loss": 0.9328, + "step": 39571 + }, + { + "epoch": 2.92, + "learning_rate": 1.0894270533849562e-05, + "loss": 1.0769, + "step": 39572 + }, + { + "epoch": 2.92, + "learning_rate": 1.0893873259647452e-05, + "loss": 0.9698, + "step": 39573 + }, + { + "epoch": 2.92, + "learning_rate": 1.0893475984023203e-05, + "loss": 0.9605, + "step": 39574 + }, + { + "epoch": 2.92, + "learning_rate": 1.0893078706977444e-05, + "loss": 1.1147, + "step": 39575 + }, + { + "epoch": 2.92, + "learning_rate": 1.0892681428510813e-05, + "loss": 0.9599, + "step": 39576 + }, + { + "epoch": 2.92, + "learning_rate": 1.089228414862394e-05, + "loss": 1.0425, + "step": 39577 + }, + { + "epoch": 2.92, + "learning_rate": 1.089188686731745e-05, + "loss": 0.9444, + "step": 39578 + }, + { + "epoch": 2.92, + "learning_rate": 1.0891489584591985e-05, + "loss": 0.8969, + "step": 39579 + }, + { + "epoch": 2.92, + "learning_rate": 1.0891092300448171e-05, + "loss": 0.9497, + "step": 39580 + }, + { + "epoch": 2.92, + "learning_rate": 1.0890695014886642e-05, + "loss": 0.9434, + "step": 39581 + }, + { + "epoch": 2.92, + "learning_rate": 1.089029772790803e-05, + "loss": 1.1322, + "step": 39582 + }, + { + "epoch": 2.92, + "learning_rate": 1.0889900439512966e-05, + "loss": 1.0542, + "step": 39583 + }, + { + "epoch": 2.92, + "learning_rate": 1.0889503149702085e-05, + "loss": 1.0125, + "step": 39584 + }, + { + "epoch": 2.92, + "learning_rate": 1.0889105858476016e-05, + "loss": 1.0014, + "step": 39585 + }, + { + "epoch": 2.92, + "learning_rate": 1.0888708565835393e-05, + "loss": 1.1596, + "step": 39586 + }, + { + "epoch": 2.93, + "learning_rate": 1.0888311271780847e-05, + "loss": 0.9789, + "step": 39587 + }, + { + "epoch": 2.93, + "learning_rate": 1.0887913976313015e-05, + "loss": 1.1007, + "step": 39588 + }, + { + "epoch": 2.93, + "learning_rate": 1.0887516679432517e-05, + "loss": 0.9684, + "step": 39589 + }, + { + "epoch": 2.93, + "learning_rate": 1.088711938114e-05, + "loss": 1.0482, + "step": 39590 + }, + { + "epoch": 2.93, + "learning_rate": 1.088672208143608e-05, + "loss": 1.0531, + "step": 39591 + }, + { + "epoch": 2.93, + "learning_rate": 1.0886324780321405e-05, + "loss": 0.9759, + "step": 39592 + }, + { + "epoch": 2.93, + "learning_rate": 1.0885927477796595e-05, + "loss": 1.0566, + "step": 39593 + }, + { + "epoch": 2.93, + "learning_rate": 1.0885530173862291e-05, + "loss": 0.9778, + "step": 39594 + }, + { + "epoch": 2.93, + "learning_rate": 1.088513286851912e-05, + "loss": 0.9582, + "step": 39595 + }, + { + "epoch": 2.93, + "learning_rate": 1.0884735561767716e-05, + "loss": 0.9519, + "step": 39596 + }, + { + "epoch": 2.93, + "learning_rate": 1.088433825360871e-05, + "loss": 0.9886, + "step": 39597 + }, + { + "epoch": 2.93, + "learning_rate": 1.0883940944042735e-05, + "loss": 0.9559, + "step": 39598 + }, + { + "epoch": 2.93, + "learning_rate": 1.0883543633070418e-05, + "loss": 1.1134, + "step": 39599 + }, + { + "epoch": 2.93, + "learning_rate": 1.0883146320692402e-05, + "loss": 0.9605, + "step": 39600 + }, + { + "epoch": 2.93, + "learning_rate": 1.088274900690931e-05, + "loss": 0.9976, + "step": 39601 + }, + { + "epoch": 2.93, + "learning_rate": 1.0882351691721778e-05, + "loss": 1.0958, + "step": 39602 + }, + { + "epoch": 2.93, + "learning_rate": 1.0881954375130434e-05, + "loss": 0.9899, + "step": 39603 + }, + { + "epoch": 2.93, + "learning_rate": 1.0881557057135916e-05, + "loss": 1.0791, + "step": 39604 + }, + { + "epoch": 2.93, + "learning_rate": 1.088115973773885e-05, + "loss": 1.0071, + "step": 39605 + }, + { + "epoch": 2.93, + "learning_rate": 1.0880762416939876e-05, + "loss": 0.9535, + "step": 39606 + }, + { + "epoch": 2.93, + "learning_rate": 1.088036509473962e-05, + "loss": 0.9726, + "step": 39607 + }, + { + "epoch": 2.93, + "learning_rate": 1.0879967771138718e-05, + "loss": 1.0514, + "step": 39608 + }, + { + "epoch": 2.93, + "learning_rate": 1.0879570446137794e-05, + "loss": 0.904, + "step": 39609 + }, + { + "epoch": 2.93, + "learning_rate": 1.0879173119737492e-05, + "loss": 1.0419, + "step": 39610 + }, + { + "epoch": 2.93, + "learning_rate": 1.0878775791938434e-05, + "loss": 0.9945, + "step": 39611 + }, + { + "epoch": 2.93, + "learning_rate": 1.0878378462741259e-05, + "loss": 0.9743, + "step": 39612 + }, + { + "epoch": 2.93, + "learning_rate": 1.0877981132146594e-05, + "loss": 1.0337, + "step": 39613 + }, + { + "epoch": 2.93, + "learning_rate": 1.0877583800155075e-05, + "loss": 0.8775, + "step": 39614 + }, + { + "epoch": 2.93, + "learning_rate": 1.0877186466767331e-05, + "loss": 1.0749, + "step": 39615 + }, + { + "epoch": 2.93, + "learning_rate": 1.0876789131983998e-05, + "loss": 0.9735, + "step": 39616 + }, + { + "epoch": 2.93, + "learning_rate": 1.0876391795805708e-05, + "loss": 0.9922, + "step": 39617 + }, + { + "epoch": 2.93, + "learning_rate": 1.0875994458233085e-05, + "loss": 1.0484, + "step": 39618 + }, + { + "epoch": 2.93, + "learning_rate": 1.0875597119266774e-05, + "loss": 1.1165, + "step": 39619 + }, + { + "epoch": 2.93, + "learning_rate": 1.0875199778907398e-05, + "loss": 0.983, + "step": 39620 + }, + { + "epoch": 2.93, + "learning_rate": 1.087480243715559e-05, + "loss": 1.0101, + "step": 39621 + }, + { + "epoch": 2.93, + "learning_rate": 1.0874405094011986e-05, + "loss": 1.051, + "step": 39622 + }, + { + "epoch": 2.93, + "learning_rate": 1.0874007749477214e-05, + "loss": 0.9466, + "step": 39623 + }, + { + "epoch": 2.93, + "learning_rate": 1.0873610403551908e-05, + "loss": 1.0361, + "step": 39624 + }, + { + "epoch": 2.93, + "learning_rate": 1.0873213056236703e-05, + "loss": 0.9631, + "step": 39625 + }, + { + "epoch": 2.93, + "learning_rate": 1.0872815707532228e-05, + "loss": 0.9506, + "step": 39626 + }, + { + "epoch": 2.93, + "learning_rate": 1.0872418357439114e-05, + "loss": 0.8993, + "step": 39627 + }, + { + "epoch": 2.93, + "learning_rate": 1.0872021005957995e-05, + "loss": 0.9808, + "step": 39628 + }, + { + "epoch": 2.93, + "learning_rate": 1.0871623653089505e-05, + "loss": 0.8969, + "step": 39629 + }, + { + "epoch": 2.93, + "learning_rate": 1.0871226298834272e-05, + "loss": 1.0365, + "step": 39630 + }, + { + "epoch": 2.93, + "learning_rate": 1.0870828943192931e-05, + "loss": 1.0648, + "step": 39631 + }, + { + "epoch": 2.93, + "learning_rate": 1.0870431586166112e-05, + "loss": 0.9822, + "step": 39632 + }, + { + "epoch": 2.93, + "learning_rate": 1.0870034227754453e-05, + "loss": 1.0831, + "step": 39633 + }, + { + "epoch": 2.93, + "learning_rate": 1.0869636867958578e-05, + "loss": 0.898, + "step": 39634 + }, + { + "epoch": 2.93, + "learning_rate": 1.0869239506779127e-05, + "loss": 0.9653, + "step": 39635 + }, + { + "epoch": 2.93, + "learning_rate": 1.0868842144216727e-05, + "loss": 1.0416, + "step": 39636 + }, + { + "epoch": 2.93, + "learning_rate": 1.0868444780272012e-05, + "loss": 1.0167, + "step": 39637 + }, + { + "epoch": 2.93, + "learning_rate": 1.086804741494561e-05, + "loss": 1.0778, + "step": 39638 + }, + { + "epoch": 2.93, + "learning_rate": 1.0867650048238161e-05, + "loss": 1.0829, + "step": 39639 + }, + { + "epoch": 2.93, + "learning_rate": 1.0867252680150291e-05, + "loss": 1.1136, + "step": 39640 + }, + { + "epoch": 2.93, + "learning_rate": 1.0866855310682637e-05, + "loss": 1.0942, + "step": 39641 + }, + { + "epoch": 2.93, + "learning_rate": 1.0866457939835824e-05, + "loss": 1.0012, + "step": 39642 + }, + { + "epoch": 2.93, + "learning_rate": 1.0866060567610493e-05, + "loss": 0.9152, + "step": 39643 + }, + { + "epoch": 2.93, + "learning_rate": 1.0865663194007273e-05, + "loss": 0.9502, + "step": 39644 + }, + { + "epoch": 2.93, + "learning_rate": 1.0865265819026794e-05, + "loss": 0.9962, + "step": 39645 + }, + { + "epoch": 2.93, + "learning_rate": 1.0864868442669685e-05, + "loss": 0.9751, + "step": 39646 + }, + { + "epoch": 2.93, + "learning_rate": 1.0864471064936588e-05, + "loss": 0.8557, + "step": 39647 + }, + { + "epoch": 2.93, + "learning_rate": 1.0864073685828126e-05, + "loss": 0.9805, + "step": 39648 + }, + { + "epoch": 2.93, + "learning_rate": 1.0863676305344939e-05, + "loss": 1.014, + "step": 39649 + }, + { + "epoch": 2.93, + "learning_rate": 1.0863278923487652e-05, + "loss": 1.0094, + "step": 39650 + }, + { + "epoch": 2.93, + "learning_rate": 1.0862881540256903e-05, + "loss": 0.9913, + "step": 39651 + }, + { + "epoch": 2.93, + "learning_rate": 1.0862484155653318e-05, + "loss": 1.0469, + "step": 39652 + }, + { + "epoch": 2.93, + "learning_rate": 1.0862086769677539e-05, + "loss": 1.0391, + "step": 39653 + }, + { + "epoch": 2.93, + "learning_rate": 1.0861689382330189e-05, + "loss": 0.9289, + "step": 39654 + }, + { + "epoch": 2.93, + "learning_rate": 1.0861291993611905e-05, + "loss": 0.915, + "step": 39655 + }, + { + "epoch": 2.93, + "learning_rate": 1.0860894603523316e-05, + "loss": 1.0551, + "step": 39656 + }, + { + "epoch": 2.93, + "learning_rate": 1.0860497212065055e-05, + "loss": 0.8203, + "step": 39657 + }, + { + "epoch": 2.93, + "learning_rate": 1.0860099819237758e-05, + "loss": 1.0636, + "step": 39658 + }, + { + "epoch": 2.93, + "learning_rate": 1.0859702425042054e-05, + "loss": 0.9592, + "step": 39659 + }, + { + "epoch": 2.93, + "learning_rate": 1.0859305029478573e-05, + "loss": 0.9831, + "step": 39660 + }, + { + "epoch": 2.93, + "learning_rate": 1.0858907632547953e-05, + "loss": 0.8867, + "step": 39661 + }, + { + "epoch": 2.93, + "learning_rate": 1.085851023425082e-05, + "loss": 0.9892, + "step": 39662 + }, + { + "epoch": 2.93, + "learning_rate": 1.0858112834587812e-05, + "loss": 1.118, + "step": 39663 + }, + { + "epoch": 2.93, + "learning_rate": 1.085771543355956e-05, + "loss": 1.0652, + "step": 39664 + }, + { + "epoch": 2.93, + "learning_rate": 1.0857318031166695e-05, + "loss": 1.0673, + "step": 39665 + }, + { + "epoch": 2.93, + "learning_rate": 1.0856920627409846e-05, + "loss": 0.9981, + "step": 39666 + }, + { + "epoch": 2.93, + "learning_rate": 1.0856523222289653e-05, + "loss": 0.9421, + "step": 39667 + }, + { + "epoch": 2.93, + "learning_rate": 1.0856125815806736e-05, + "loss": 0.9671, + "step": 39668 + }, + { + "epoch": 2.93, + "learning_rate": 1.0855728407961744e-05, + "loss": 0.9387, + "step": 39669 + }, + { + "epoch": 2.93, + "learning_rate": 1.0855330998755296e-05, + "loss": 1.0165, + "step": 39670 + }, + { + "epoch": 2.93, + "learning_rate": 1.0854933588188027e-05, + "loss": 1.0658, + "step": 39671 + }, + { + "epoch": 2.93, + "learning_rate": 1.0854536176260574e-05, + "loss": 0.9344, + "step": 39672 + }, + { + "epoch": 2.93, + "learning_rate": 1.0854138762973566e-05, + "loss": 0.8898, + "step": 39673 + }, + { + "epoch": 2.93, + "learning_rate": 1.0853741348327635e-05, + "loss": 1.0439, + "step": 39674 + }, + { + "epoch": 2.93, + "learning_rate": 1.0853343932323413e-05, + "loss": 1.0424, + "step": 39675 + }, + { + "epoch": 2.93, + "learning_rate": 1.0852946514961534e-05, + "loss": 1.0273, + "step": 39676 + }, + { + "epoch": 2.93, + "learning_rate": 1.0852549096242627e-05, + "loss": 1.0086, + "step": 39677 + }, + { + "epoch": 2.93, + "learning_rate": 1.0852151676167328e-05, + "loss": 0.9231, + "step": 39678 + }, + { + "epoch": 2.93, + "learning_rate": 1.085175425473627e-05, + "loss": 1.0633, + "step": 39679 + }, + { + "epoch": 2.93, + "learning_rate": 1.085135683195008e-05, + "loss": 0.94, + "step": 39680 + }, + { + "epoch": 2.93, + "learning_rate": 1.0850959407809396e-05, + "loss": 0.9484, + "step": 39681 + }, + { + "epoch": 2.93, + "learning_rate": 1.0850561982314847e-05, + "loss": 1.0033, + "step": 39682 + }, + { + "epoch": 2.93, + "learning_rate": 1.0850164555467064e-05, + "loss": 0.8984, + "step": 39683 + }, + { + "epoch": 2.93, + "learning_rate": 1.0849767127266684e-05, + "loss": 0.8899, + "step": 39684 + }, + { + "epoch": 2.93, + "learning_rate": 1.0849369697714335e-05, + "loss": 0.9862, + "step": 39685 + }, + { + "epoch": 2.93, + "learning_rate": 1.0848972266810653e-05, + "loss": 0.9148, + "step": 39686 + }, + { + "epoch": 2.93, + "learning_rate": 1.0848574834556264e-05, + "loss": 1.0785, + "step": 39687 + }, + { + "epoch": 2.93, + "learning_rate": 1.084817740095181e-05, + "loss": 0.9801, + "step": 39688 + }, + { + "epoch": 2.93, + "learning_rate": 1.0847779965997914e-05, + "loss": 1.0142, + "step": 39689 + }, + { + "epoch": 2.93, + "learning_rate": 1.0847382529695214e-05, + "loss": 0.9608, + "step": 39690 + }, + { + "epoch": 2.93, + "learning_rate": 1.0846985092044342e-05, + "loss": 0.8761, + "step": 39691 + }, + { + "epoch": 2.93, + "learning_rate": 1.0846587653045926e-05, + "loss": 1.0096, + "step": 39692 + }, + { + "epoch": 2.93, + "learning_rate": 1.0846190212700602e-05, + "loss": 0.9428, + "step": 39693 + }, + { + "epoch": 2.93, + "learning_rate": 1.0845792771009003e-05, + "loss": 0.8165, + "step": 39694 + }, + { + "epoch": 2.93, + "learning_rate": 1.0845395327971754e-05, + "loss": 0.9934, + "step": 39695 + }, + { + "epoch": 2.93, + "learning_rate": 1.08449978835895e-05, + "loss": 0.9623, + "step": 39696 + }, + { + "epoch": 2.93, + "learning_rate": 1.0844600437862862e-05, + "loss": 0.9035, + "step": 39697 + }, + { + "epoch": 2.93, + "learning_rate": 1.0844202990792481e-05, + "loss": 1.0627, + "step": 39698 + }, + { + "epoch": 2.93, + "learning_rate": 1.0843805542378982e-05, + "loss": 0.9771, + "step": 39699 + }, + { + "epoch": 2.93, + "learning_rate": 1.0843408092623004e-05, + "loss": 1.062, + "step": 39700 + }, + { + "epoch": 2.93, + "learning_rate": 1.0843010641525172e-05, + "loss": 1.0201, + "step": 39701 + }, + { + "epoch": 2.93, + "learning_rate": 1.0842613189086123e-05, + "loss": 1.1069, + "step": 39702 + }, + { + "epoch": 2.93, + "learning_rate": 1.0842215735306488e-05, + "loss": 0.8728, + "step": 39703 + }, + { + "epoch": 2.93, + "learning_rate": 1.0841818280186902e-05, + "loss": 1.0153, + "step": 39704 + }, + { + "epoch": 2.93, + "learning_rate": 1.084142082372799e-05, + "loss": 1.0934, + "step": 39705 + }, + { + "epoch": 2.93, + "learning_rate": 1.0841023365930396e-05, + "loss": 0.9825, + "step": 39706 + }, + { + "epoch": 2.93, + "learning_rate": 1.0840625906794741e-05, + "loss": 1.0465, + "step": 39707 + }, + { + "epoch": 2.93, + "learning_rate": 1.0840228446321665e-05, + "loss": 0.9108, + "step": 39708 + }, + { + "epoch": 2.93, + "learning_rate": 1.0839830984511796e-05, + "loss": 0.9921, + "step": 39709 + }, + { + "epoch": 2.93, + "learning_rate": 1.0839433521365772e-05, + "loss": 1.0054, + "step": 39710 + }, + { + "epoch": 2.93, + "learning_rate": 1.0839036056884216e-05, + "loss": 1.1804, + "step": 39711 + }, + { + "epoch": 2.93, + "learning_rate": 1.083863859106777e-05, + "loss": 1.0258, + "step": 39712 + }, + { + "epoch": 2.93, + "learning_rate": 1.0838241123917057e-05, + "loss": 0.9382, + "step": 39713 + }, + { + "epoch": 2.93, + "learning_rate": 1.0837843655432718e-05, + "loss": 1.0452, + "step": 39714 + }, + { + "epoch": 2.93, + "learning_rate": 1.083744618561538e-05, + "loss": 1.0413, + "step": 39715 + }, + { + "epoch": 2.93, + "learning_rate": 1.0837048714465678e-05, + "loss": 0.9231, + "step": 39716 + }, + { + "epoch": 2.93, + "learning_rate": 1.0836651241984242e-05, + "loss": 1.0111, + "step": 39717 + }, + { + "epoch": 2.93, + "learning_rate": 1.083625376817171e-05, + "loss": 1.0609, + "step": 39718 + }, + { + "epoch": 2.93, + "learning_rate": 1.0835856293028706e-05, + "loss": 1.0234, + "step": 39719 + }, + { + "epoch": 2.93, + "learning_rate": 1.0835458816555868e-05, + "loss": 1.062, + "step": 39720 + }, + { + "epoch": 2.93, + "learning_rate": 1.0835061338753828e-05, + "loss": 0.8893, + "step": 39721 + }, + { + "epoch": 2.94, + "learning_rate": 1.0834663859623213e-05, + "loss": 1.0136, + "step": 39722 + }, + { + "epoch": 2.94, + "learning_rate": 1.0834266379164666e-05, + "loss": 1.1231, + "step": 39723 + }, + { + "epoch": 2.94, + "learning_rate": 1.0833868897378807e-05, + "loss": 1.0502, + "step": 39724 + }, + { + "epoch": 2.94, + "learning_rate": 1.0833471414266279e-05, + "loss": 0.9576, + "step": 39725 + }, + { + "epoch": 2.94, + "learning_rate": 1.0833073929827712e-05, + "loss": 1.1822, + "step": 39726 + }, + { + "epoch": 2.94, + "learning_rate": 1.083267644406373e-05, + "loss": 1.0097, + "step": 39727 + }, + { + "epoch": 2.94, + "learning_rate": 1.0832278956974976e-05, + "loss": 0.9604, + "step": 39728 + }, + { + "epoch": 2.94, + "learning_rate": 1.0831881468562077e-05, + "loss": 1.0477, + "step": 39729 + }, + { + "epoch": 2.94, + "learning_rate": 1.0831483978825666e-05, + "loss": 1.055, + "step": 39730 + }, + { + "epoch": 2.94, + "learning_rate": 1.0831086487766376e-05, + "loss": 1.1188, + "step": 39731 + }, + { + "epoch": 2.94, + "learning_rate": 1.0830688995384837e-05, + "loss": 0.9812, + "step": 39732 + }, + { + "epoch": 2.94, + "learning_rate": 1.0830291501681686e-05, + "loss": 1.0359, + "step": 39733 + }, + { + "epoch": 2.94, + "learning_rate": 1.0829894006657553e-05, + "loss": 1.0137, + "step": 39734 + }, + { + "epoch": 2.94, + "learning_rate": 1.082949651031307e-05, + "loss": 1.0421, + "step": 39735 + }, + { + "epoch": 2.94, + "learning_rate": 1.0829099012648872e-05, + "loss": 0.9627, + "step": 39736 + }, + { + "epoch": 2.94, + "learning_rate": 1.0828701513665587e-05, + "loss": 0.9899, + "step": 39737 + }, + { + "epoch": 2.94, + "learning_rate": 1.0828304013363853e-05, + "loss": 0.9481, + "step": 39738 + }, + { + "epoch": 2.94, + "learning_rate": 1.0827906511744298e-05, + "loss": 0.9664, + "step": 39739 + }, + { + "epoch": 2.94, + "learning_rate": 1.0827509008807552e-05, + "loss": 1.057, + "step": 39740 + }, + { + "epoch": 2.94, + "learning_rate": 1.0827111504554255e-05, + "loss": 0.9162, + "step": 39741 + }, + { + "epoch": 2.94, + "learning_rate": 1.0826713998985032e-05, + "loss": 0.9613, + "step": 39742 + }, + { + "epoch": 2.94, + "learning_rate": 1.0826316492100522e-05, + "loss": 1.0808, + "step": 39743 + }, + { + "epoch": 2.94, + "learning_rate": 1.082591898390135e-05, + "loss": 1.0258, + "step": 39744 + }, + { + "epoch": 2.94, + "learning_rate": 1.0825521474388158e-05, + "loss": 0.9291, + "step": 39745 + }, + { + "epoch": 2.94, + "learning_rate": 1.082512396356157e-05, + "loss": 1.032, + "step": 39746 + }, + { + "epoch": 2.94, + "learning_rate": 1.0824726451422226e-05, + "loss": 1.0255, + "step": 39747 + }, + { + "epoch": 2.94, + "learning_rate": 1.0824328937970747e-05, + "loss": 1.0001, + "step": 39748 + }, + { + "epoch": 2.94, + "learning_rate": 1.0823931423207779e-05, + "loss": 0.9367, + "step": 39749 + }, + { + "epoch": 2.94, + "learning_rate": 1.0823533907133942e-05, + "loss": 1.0259, + "step": 39750 + }, + { + "epoch": 2.94, + "learning_rate": 1.0823136389749878e-05, + "loss": 0.9799, + "step": 39751 + }, + { + "epoch": 2.94, + "learning_rate": 1.0822738871056212e-05, + "loss": 0.8923, + "step": 39752 + }, + { + "epoch": 2.94, + "learning_rate": 1.0822341351053586e-05, + "loss": 1.0662, + "step": 39753 + }, + { + "epoch": 2.94, + "learning_rate": 1.0821943829742623e-05, + "loss": 0.9942, + "step": 39754 + }, + { + "epoch": 2.94, + "learning_rate": 1.082154630712396e-05, + "loss": 0.9487, + "step": 39755 + }, + { + "epoch": 2.94, + "learning_rate": 1.082114878319823e-05, + "loss": 1.0659, + "step": 39756 + }, + { + "epoch": 2.94, + "learning_rate": 1.0820751257966062e-05, + "loss": 1.0157, + "step": 39757 + }, + { + "epoch": 2.94, + "learning_rate": 1.082035373142809e-05, + "loss": 1.1113, + "step": 39758 + }, + { + "epoch": 2.94, + "learning_rate": 1.081995620358495e-05, + "loss": 1.0394, + "step": 39759 + }, + { + "epoch": 2.94, + "learning_rate": 1.0819558674437268e-05, + "loss": 1.1176, + "step": 39760 + }, + { + "epoch": 2.94, + "learning_rate": 1.0819161143985682e-05, + "loss": 0.9747, + "step": 39761 + }, + { + "epoch": 2.94, + "learning_rate": 1.0818763612230821e-05, + "loss": 1.0949, + "step": 39762 + }, + { + "epoch": 2.94, + "learning_rate": 1.081836607917332e-05, + "loss": 0.9935, + "step": 39763 + }, + { + "epoch": 2.94, + "learning_rate": 1.0817968544813809e-05, + "loss": 0.988, + "step": 39764 + }, + { + "epoch": 2.94, + "learning_rate": 1.0817571009152923e-05, + "loss": 0.9602, + "step": 39765 + }, + { + "epoch": 2.94, + "learning_rate": 1.0817173472191293e-05, + "loss": 0.9698, + "step": 39766 + }, + { + "epoch": 2.94, + "learning_rate": 1.0816775933929549e-05, + "loss": 0.8727, + "step": 39767 + }, + { + "epoch": 2.94, + "learning_rate": 1.081637839436833e-05, + "loss": 1.0147, + "step": 39768 + }, + { + "epoch": 2.94, + "learning_rate": 1.0815980853508261e-05, + "loss": 1.0028, + "step": 39769 + }, + { + "epoch": 2.94, + "learning_rate": 1.0815583311349983e-05, + "loss": 0.9657, + "step": 39770 + }, + { + "epoch": 2.94, + "learning_rate": 1.081518576789412e-05, + "loss": 0.9618, + "step": 39771 + }, + { + "epoch": 2.94, + "learning_rate": 1.0814788223141308e-05, + "loss": 1.0663, + "step": 39772 + }, + { + "epoch": 2.94, + "learning_rate": 1.0814390677092182e-05, + "loss": 0.9864, + "step": 39773 + }, + { + "epoch": 2.94, + "learning_rate": 1.081399312974737e-05, + "loss": 0.9279, + "step": 39774 + }, + { + "epoch": 2.94, + "learning_rate": 1.0813595581107508e-05, + "loss": 1.0498, + "step": 39775 + }, + { + "epoch": 2.94, + "learning_rate": 1.0813198031173226e-05, + "loss": 0.9853, + "step": 39776 + }, + { + "epoch": 2.94, + "learning_rate": 1.0812800479945157e-05, + "loss": 1.0691, + "step": 39777 + }, + { + "epoch": 2.94, + "learning_rate": 1.0812402927423937e-05, + "loss": 1.1367, + "step": 39778 + }, + { + "epoch": 2.94, + "learning_rate": 1.0812005373610193e-05, + "loss": 1.0011, + "step": 39779 + }, + { + "epoch": 2.94, + "learning_rate": 1.0811607818504562e-05, + "loss": 0.8926, + "step": 39780 + }, + { + "epoch": 2.94, + "learning_rate": 1.0811210262107669e-05, + "loss": 1.0699, + "step": 39781 + }, + { + "epoch": 2.94, + "learning_rate": 1.0810812704420159e-05, + "loss": 0.959, + "step": 39782 + }, + { + "epoch": 2.94, + "learning_rate": 1.0810415145442654e-05, + "loss": 0.8885, + "step": 39783 + }, + { + "epoch": 2.94, + "learning_rate": 1.0810017585175794e-05, + "loss": 1.0262, + "step": 39784 + }, + { + "epoch": 2.94, + "learning_rate": 1.0809620023620202e-05, + "loss": 0.972, + "step": 39785 + }, + { + "epoch": 2.94, + "learning_rate": 1.0809222460776521e-05, + "loss": 0.9712, + "step": 39786 + }, + { + "epoch": 2.94, + "learning_rate": 1.0808824896645374e-05, + "loss": 0.9513, + "step": 39787 + }, + { + "epoch": 2.94, + "learning_rate": 1.0808427331227403e-05, + "loss": 1.1303, + "step": 39788 + }, + { + "epoch": 2.94, + "learning_rate": 1.080802976452323e-05, + "loss": 1.0452, + "step": 39789 + }, + { + "epoch": 2.94, + "learning_rate": 1.0807632196533498e-05, + "loss": 1.0715, + "step": 39790 + }, + { + "epoch": 2.94, + "learning_rate": 1.0807234627258831e-05, + "loss": 0.9056, + "step": 39791 + }, + { + "epoch": 2.94, + "learning_rate": 1.0806837056699868e-05, + "loss": 1.1264, + "step": 39792 + }, + { + "epoch": 2.94, + "learning_rate": 1.0806439484857238e-05, + "loss": 1.0015, + "step": 39793 + }, + { + "epoch": 2.94, + "learning_rate": 1.0806041911731577e-05, + "loss": 1.0893, + "step": 39794 + }, + { + "epoch": 2.94, + "learning_rate": 1.0805644337323511e-05, + "loss": 0.9808, + "step": 39795 + }, + { + "epoch": 2.94, + "learning_rate": 1.0805246761633678e-05, + "loss": 0.9313, + "step": 39796 + }, + { + "epoch": 2.94, + "learning_rate": 1.080484918466271e-05, + "loss": 1.0274, + "step": 39797 + }, + { + "epoch": 2.94, + "learning_rate": 1.0804451606411237e-05, + "loss": 1.0691, + "step": 39798 + }, + { + "epoch": 2.94, + "learning_rate": 1.080405402687989e-05, + "loss": 0.9699, + "step": 39799 + }, + { + "epoch": 2.94, + "learning_rate": 1.0803656446069311e-05, + "loss": 0.9885, + "step": 39800 + }, + { + "epoch": 2.94, + "learning_rate": 1.0803258863980119e-05, + "loss": 0.8708, + "step": 39801 + }, + { + "epoch": 2.94, + "learning_rate": 1.080286128061296e-05, + "loss": 1.0515, + "step": 39802 + }, + { + "epoch": 2.94, + "learning_rate": 1.0802463695968459e-05, + "loss": 0.9554, + "step": 39803 + }, + { + "epoch": 2.94, + "learning_rate": 1.0802066110047249e-05, + "loss": 0.955, + "step": 39804 + }, + { + "epoch": 2.94, + "learning_rate": 1.080166852284996e-05, + "loss": 0.9191, + "step": 39805 + }, + { + "epoch": 2.94, + "learning_rate": 1.0801270934377232e-05, + "loss": 0.9263, + "step": 39806 + }, + { + "epoch": 2.94, + "learning_rate": 1.080087334462969e-05, + "loss": 1.1034, + "step": 39807 + }, + { + "epoch": 2.94, + "learning_rate": 1.0800475753607974e-05, + "loss": 1.0984, + "step": 39808 + }, + { + "epoch": 2.94, + "learning_rate": 1.0800078161312708e-05, + "loss": 1.0136, + "step": 39809 + }, + { + "epoch": 2.94, + "learning_rate": 1.0799680567744532e-05, + "loss": 1.0451, + "step": 39810 + }, + { + "epoch": 2.94, + "learning_rate": 1.0799282972904075e-05, + "loss": 0.9392, + "step": 39811 + }, + { + "epoch": 2.94, + "learning_rate": 1.079888537679197e-05, + "loss": 0.8952, + "step": 39812 + }, + { + "epoch": 2.94, + "learning_rate": 1.0798487779408853e-05, + "loss": 1.0452, + "step": 39813 + }, + { + "epoch": 2.94, + "learning_rate": 1.0798090180755349e-05, + "loss": 1.0381, + "step": 39814 + }, + { + "epoch": 2.94, + "learning_rate": 1.0797692580832098e-05, + "loss": 1.0788, + "step": 39815 + }, + { + "epoch": 2.94, + "learning_rate": 1.0797294979639727e-05, + "loss": 1.0168, + "step": 39816 + }, + { + "epoch": 2.94, + "learning_rate": 1.0796897377178872e-05, + "loss": 0.9048, + "step": 39817 + }, + { + "epoch": 2.94, + "learning_rate": 1.0796499773450165e-05, + "loss": 1.0883, + "step": 39818 + }, + { + "epoch": 2.94, + "learning_rate": 1.0796102168454237e-05, + "loss": 0.9798, + "step": 39819 + }, + { + "epoch": 2.94, + "learning_rate": 1.0795704562191722e-05, + "loss": 0.9525, + "step": 39820 + }, + { + "epoch": 2.94, + "learning_rate": 1.0795306954663253e-05, + "loss": 1.0872, + "step": 39821 + }, + { + "epoch": 2.94, + "learning_rate": 1.0794909345869465e-05, + "loss": 0.9897, + "step": 39822 + }, + { + "epoch": 2.94, + "learning_rate": 1.0794511735810984e-05, + "loss": 0.975, + "step": 39823 + }, + { + "epoch": 2.94, + "learning_rate": 1.0794114124488446e-05, + "loss": 0.8652, + "step": 39824 + }, + { + "epoch": 2.94, + "learning_rate": 1.0793716511902485e-05, + "loss": 1.1246, + "step": 39825 + }, + { + "epoch": 2.94, + "learning_rate": 1.0793318898053733e-05, + "loss": 1.0362, + "step": 39826 + }, + { + "epoch": 2.94, + "learning_rate": 1.0792921282942822e-05, + "loss": 0.9369, + "step": 39827 + }, + { + "epoch": 2.94, + "learning_rate": 1.079252366657038e-05, + "loss": 0.9997, + "step": 39828 + }, + { + "epoch": 2.94, + "learning_rate": 1.079212604893705e-05, + "loss": 1.0509, + "step": 39829 + }, + { + "epoch": 2.94, + "learning_rate": 1.0791728430043456e-05, + "loss": 0.9334, + "step": 39830 + }, + { + "epoch": 2.94, + "learning_rate": 1.0791330809890237e-05, + "loss": 0.8968, + "step": 39831 + }, + { + "epoch": 2.94, + "learning_rate": 1.079093318847802e-05, + "loss": 1.0895, + "step": 39832 + }, + { + "epoch": 2.94, + "learning_rate": 1.0790535565807438e-05, + "loss": 1.0622, + "step": 39833 + }, + { + "epoch": 2.94, + "learning_rate": 1.0790137941879124e-05, + "loss": 0.9573, + "step": 39834 + }, + { + "epoch": 2.94, + "learning_rate": 1.0789740316693715e-05, + "loss": 0.9272, + "step": 39835 + }, + { + "epoch": 2.94, + "learning_rate": 1.0789342690251838e-05, + "loss": 1.007, + "step": 39836 + }, + { + "epoch": 2.94, + "learning_rate": 1.0788945062554131e-05, + "loss": 0.9805, + "step": 39837 + }, + { + "epoch": 2.94, + "learning_rate": 1.078854743360122e-05, + "loss": 0.9772, + "step": 39838 + }, + { + "epoch": 2.94, + "learning_rate": 1.0788149803393745e-05, + "loss": 1.0122, + "step": 39839 + }, + { + "epoch": 2.94, + "learning_rate": 1.078775217193233e-05, + "loss": 0.9783, + "step": 39840 + }, + { + "epoch": 2.94, + "learning_rate": 1.078735453921762e-05, + "loss": 1.0012, + "step": 39841 + }, + { + "epoch": 2.94, + "learning_rate": 1.0786956905250236e-05, + "loss": 1.0258, + "step": 39842 + }, + { + "epoch": 2.94, + "learning_rate": 1.0786559270030816e-05, + "loss": 1.0134, + "step": 39843 + }, + { + "epoch": 2.94, + "learning_rate": 1.0786161633559989e-05, + "loss": 1.0922, + "step": 39844 + }, + { + "epoch": 2.94, + "learning_rate": 1.0785763995838392e-05, + "loss": 0.9976, + "step": 39845 + }, + { + "epoch": 2.94, + "learning_rate": 1.0785366356866653e-05, + "loss": 1.119, + "step": 39846 + }, + { + "epoch": 2.94, + "learning_rate": 1.0784968716645412e-05, + "loss": 0.9912, + "step": 39847 + }, + { + "epoch": 2.94, + "learning_rate": 1.0784571075175291e-05, + "loss": 1.1193, + "step": 39848 + }, + { + "epoch": 2.94, + "learning_rate": 1.0784173432456936e-05, + "loss": 0.966, + "step": 39849 + }, + { + "epoch": 2.94, + "learning_rate": 1.0783775788490968e-05, + "loss": 1.0173, + "step": 39850 + }, + { + "epoch": 2.94, + "learning_rate": 1.0783378143278024e-05, + "loss": 1.0183, + "step": 39851 + }, + { + "epoch": 2.94, + "learning_rate": 1.0782980496818737e-05, + "loss": 0.9995, + "step": 39852 + }, + { + "epoch": 2.94, + "learning_rate": 1.0782582849113741e-05, + "loss": 0.9753, + "step": 39853 + }, + { + "epoch": 2.94, + "learning_rate": 1.0782185200163664e-05, + "loss": 0.9461, + "step": 39854 + }, + { + "epoch": 2.94, + "learning_rate": 1.0781787549969143e-05, + "loss": 0.978, + "step": 39855 + }, + { + "epoch": 2.94, + "learning_rate": 1.0781389898530807e-05, + "loss": 0.9993, + "step": 39856 + }, + { + "epoch": 2.95, + "learning_rate": 1.0780992245849294e-05, + "loss": 1.0634, + "step": 39857 + }, + { + "epoch": 2.95, + "learning_rate": 1.0780594591925231e-05, + "loss": 0.9499, + "step": 39858 + }, + { + "epoch": 2.95, + "learning_rate": 1.0780196936759257e-05, + "loss": 1.0298, + "step": 39859 + }, + { + "epoch": 2.95, + "learning_rate": 1.0779799280351997e-05, + "loss": 0.907, + "step": 39860 + }, + { + "epoch": 2.95, + "learning_rate": 1.0779401622704091e-05, + "loss": 0.9593, + "step": 39861 + }, + { + "epoch": 2.95, + "learning_rate": 1.0779003963816164e-05, + "loss": 0.9102, + "step": 39862 + }, + { + "epoch": 2.95, + "learning_rate": 1.0778606303688854e-05, + "loss": 1.0318, + "step": 39863 + }, + { + "epoch": 2.95, + "learning_rate": 1.0778208642322793e-05, + "loss": 1.0427, + "step": 39864 + }, + { + "epoch": 2.95, + "learning_rate": 1.0777810979718615e-05, + "loss": 0.9879, + "step": 39865 + }, + { + "epoch": 2.95, + "learning_rate": 1.0777413315876948e-05, + "loss": 1.0833, + "step": 39866 + }, + { + "epoch": 2.95, + "learning_rate": 1.077701565079843e-05, + "loss": 0.9375, + "step": 39867 + }, + { + "epoch": 2.95, + "learning_rate": 1.0776617984483688e-05, + "loss": 0.9623, + "step": 39868 + }, + { + "epoch": 2.95, + "learning_rate": 1.077622031693336e-05, + "loss": 1.0551, + "step": 39869 + }, + { + "epoch": 2.95, + "learning_rate": 1.0775822648148078e-05, + "loss": 1.0319, + "step": 39870 + }, + { + "epoch": 2.95, + "learning_rate": 1.0775424978128471e-05, + "loss": 0.9679, + "step": 39871 + }, + { + "epoch": 2.95, + "learning_rate": 1.0775027306875176e-05, + "loss": 0.9603, + "step": 39872 + }, + { + "epoch": 2.95, + "learning_rate": 1.077462963438882e-05, + "loss": 1.0557, + "step": 39873 + }, + { + "epoch": 2.95, + "learning_rate": 1.0774231960670043e-05, + "loss": 0.9654, + "step": 39874 + }, + { + "epoch": 2.95, + "learning_rate": 1.0773834285719473e-05, + "loss": 0.9512, + "step": 39875 + }, + { + "epoch": 2.95, + "learning_rate": 1.0773436609537744e-05, + "loss": 0.9653, + "step": 39876 + }, + { + "epoch": 2.95, + "learning_rate": 1.0773038932125488e-05, + "loss": 1.0193, + "step": 39877 + }, + { + "epoch": 2.95, + "learning_rate": 1.077264125348334e-05, + "loss": 1.0856, + "step": 39878 + }, + { + "epoch": 2.95, + "learning_rate": 1.0772243573611927e-05, + "loss": 1.0556, + "step": 39879 + }, + { + "epoch": 2.95, + "learning_rate": 1.0771845892511888e-05, + "loss": 1.0574, + "step": 39880 + }, + { + "epoch": 2.95, + "learning_rate": 1.0771448210183853e-05, + "loss": 1.0364, + "step": 39881 + }, + { + "epoch": 2.95, + "learning_rate": 1.0771050526628455e-05, + "loss": 1.0393, + "step": 39882 + }, + { + "epoch": 2.95, + "learning_rate": 1.0770652841846323e-05, + "loss": 0.9368, + "step": 39883 + }, + { + "epoch": 2.95, + "learning_rate": 1.0770255155838099e-05, + "loss": 0.9754, + "step": 39884 + }, + { + "epoch": 2.95, + "learning_rate": 1.0769857468604406e-05, + "loss": 0.925, + "step": 39885 + }, + { + "epoch": 2.95, + "learning_rate": 1.0769459780145883e-05, + "loss": 1.0143, + "step": 39886 + }, + { + "epoch": 2.95, + "learning_rate": 1.076906209046316e-05, + "loss": 1.0951, + "step": 39887 + }, + { + "epoch": 2.95, + "learning_rate": 1.0768664399556872e-05, + "loss": 0.978, + "step": 39888 + }, + { + "epoch": 2.95, + "learning_rate": 1.0768266707427645e-05, + "loss": 0.9917, + "step": 39889 + }, + { + "epoch": 2.95, + "learning_rate": 1.0767869014076121e-05, + "loss": 0.9672, + "step": 39890 + }, + { + "epoch": 2.95, + "learning_rate": 1.0767471319502927e-05, + "loss": 0.9611, + "step": 39891 + }, + { + "epoch": 2.95, + "learning_rate": 1.0767073623708698e-05, + "loss": 1.0352, + "step": 39892 + }, + { + "epoch": 2.95, + "learning_rate": 1.0766675926694063e-05, + "loss": 0.8959, + "step": 39893 + }, + { + "epoch": 2.95, + "learning_rate": 1.0766278228459662e-05, + "loss": 0.988, + "step": 39894 + }, + { + "epoch": 2.95, + "learning_rate": 1.0765880529006118e-05, + "loss": 0.8996, + "step": 39895 + }, + { + "epoch": 2.95, + "learning_rate": 1.0765482828334073e-05, + "loss": 1.0345, + "step": 39896 + }, + { + "epoch": 2.95, + "learning_rate": 1.0765085126444157e-05, + "loss": 1.0842, + "step": 39897 + }, + { + "epoch": 2.95, + "learning_rate": 1.0764687423336999e-05, + "loss": 0.9969, + "step": 39898 + }, + { + "epoch": 2.95, + "learning_rate": 1.0764289719013234e-05, + "loss": 0.8983, + "step": 39899 + }, + { + "epoch": 2.95, + "learning_rate": 1.0763892013473496e-05, + "loss": 1.0374, + "step": 39900 + }, + { + "epoch": 2.95, + "learning_rate": 1.0763494306718414e-05, + "loss": 0.9736, + "step": 39901 + }, + { + "epoch": 2.95, + "learning_rate": 1.0763096598748628e-05, + "loss": 1.1022, + "step": 39902 + }, + { + "epoch": 2.95, + "learning_rate": 1.0762698889564763e-05, + "loss": 1.0529, + "step": 39903 + }, + { + "epoch": 2.95, + "learning_rate": 1.0762301179167456e-05, + "loss": 1.0167, + "step": 39904 + }, + { + "epoch": 2.95, + "learning_rate": 1.0761903467557338e-05, + "loss": 0.9733, + "step": 39905 + }, + { + "epoch": 2.95, + "learning_rate": 1.0761505754735047e-05, + "loss": 1.1582, + "step": 39906 + }, + { + "epoch": 2.95, + "learning_rate": 1.0761108040701205e-05, + "loss": 1.0781, + "step": 39907 + }, + { + "epoch": 2.95, + "learning_rate": 1.0760710325456454e-05, + "loss": 1.0875, + "step": 39908 + }, + { + "epoch": 2.95, + "learning_rate": 1.0760312609001423e-05, + "loss": 1.1225, + "step": 39909 + }, + { + "epoch": 2.95, + "learning_rate": 1.0759914891336745e-05, + "loss": 0.9929, + "step": 39910 + }, + { + "epoch": 2.95, + "learning_rate": 1.0759517172463053e-05, + "loss": 1.0032, + "step": 39911 + }, + { + "epoch": 2.95, + "learning_rate": 1.0759119452380982e-05, + "loss": 1.0324, + "step": 39912 + }, + { + "epoch": 2.95, + "learning_rate": 1.075872173109116e-05, + "loss": 1.0977, + "step": 39913 + }, + { + "epoch": 2.95, + "learning_rate": 1.0758324008594226e-05, + "loss": 0.9354, + "step": 39914 + }, + { + "epoch": 2.95, + "learning_rate": 1.0757926284890808e-05, + "loss": 1.1154, + "step": 39915 + }, + { + "epoch": 2.95, + "learning_rate": 1.0757528559981538e-05, + "loss": 0.922, + "step": 39916 + }, + { + "epoch": 2.95, + "learning_rate": 1.0757130833867053e-05, + "loss": 0.996, + "step": 39917 + }, + { + "epoch": 2.95, + "learning_rate": 1.0756733106547983e-05, + "loss": 0.9733, + "step": 39918 + }, + { + "epoch": 2.95, + "learning_rate": 1.0756335378024963e-05, + "loss": 1.0008, + "step": 39919 + }, + { + "epoch": 2.95, + "learning_rate": 1.0755937648298624e-05, + "loss": 0.916, + "step": 39920 + }, + { + "epoch": 2.95, + "learning_rate": 1.0755539917369597e-05, + "loss": 0.9567, + "step": 39921 + }, + { + "epoch": 2.95, + "learning_rate": 1.075514218523852e-05, + "loss": 0.9049, + "step": 39922 + }, + { + "epoch": 2.95, + "learning_rate": 1.0754744451906019e-05, + "loss": 0.9678, + "step": 39923 + }, + { + "epoch": 2.95, + "learning_rate": 1.0754346717372732e-05, + "loss": 1.0196, + "step": 39924 + }, + { + "epoch": 2.95, + "learning_rate": 1.075394898163929e-05, + "loss": 0.9107, + "step": 39925 + }, + { + "epoch": 2.95, + "learning_rate": 1.0753551244706327e-05, + "loss": 0.9648, + "step": 39926 + }, + { + "epoch": 2.95, + "learning_rate": 1.0753153506574472e-05, + "loss": 1.1737, + "step": 39927 + }, + { + "epoch": 2.95, + "learning_rate": 1.0752755767244363e-05, + "loss": 1.1038, + "step": 39928 + }, + { + "epoch": 2.95, + "learning_rate": 1.075235802671663e-05, + "loss": 1.0182, + "step": 39929 + }, + { + "epoch": 2.95, + "learning_rate": 1.0751960284991904e-05, + "loss": 0.9684, + "step": 39930 + }, + { + "epoch": 2.95, + "learning_rate": 1.0751562542070823e-05, + "loss": 0.9104, + "step": 39931 + }, + { + "epoch": 2.95, + "learning_rate": 1.0751164797954015e-05, + "loss": 0.9823, + "step": 39932 + }, + { + "epoch": 2.95, + "learning_rate": 1.0750767052642115e-05, + "loss": 0.9588, + "step": 39933 + }, + { + "epoch": 2.95, + "learning_rate": 1.0750369306135756e-05, + "loss": 1.0389, + "step": 39934 + }, + { + "epoch": 2.95, + "learning_rate": 1.074997155843557e-05, + "loss": 0.9991, + "step": 39935 + }, + { + "epoch": 2.95, + "learning_rate": 1.0749573809542189e-05, + "loss": 1.0238, + "step": 39936 + }, + { + "epoch": 2.95, + "learning_rate": 1.0749176059456248e-05, + "loss": 0.9269, + "step": 39937 + }, + { + "epoch": 2.95, + "learning_rate": 1.0748778308178377e-05, + "loss": 1.0817, + "step": 39938 + }, + { + "epoch": 2.95, + "learning_rate": 1.0748380555709213e-05, + "loss": 0.9849, + "step": 39939 + }, + { + "epoch": 2.95, + "learning_rate": 1.074798280204938e-05, + "loss": 0.9176, + "step": 39940 + }, + { + "epoch": 2.95, + "learning_rate": 1.0747585047199524e-05, + "loss": 0.9788, + "step": 39941 + }, + { + "epoch": 2.95, + "learning_rate": 1.0747187291160268e-05, + "loss": 1.1309, + "step": 39942 + }, + { + "epoch": 2.95, + "learning_rate": 1.0746789533932251e-05, + "loss": 1.0036, + "step": 39943 + }, + { + "epoch": 2.95, + "learning_rate": 1.0746391775516099e-05, + "loss": 1.1279, + "step": 39944 + }, + { + "epoch": 2.95, + "learning_rate": 1.074599401591245e-05, + "loss": 1.0539, + "step": 39945 + }, + { + "epoch": 2.95, + "learning_rate": 1.0745596255121933e-05, + "loss": 0.9868, + "step": 39946 + }, + { + "epoch": 2.95, + "learning_rate": 1.0745198493145185e-05, + "loss": 1.06, + "step": 39947 + }, + { + "epoch": 2.95, + "learning_rate": 1.0744800729982836e-05, + "loss": 1.0235, + "step": 39948 + }, + { + "epoch": 2.95, + "learning_rate": 1.074440296563552e-05, + "loss": 0.9642, + "step": 39949 + }, + { + "epoch": 2.95, + "learning_rate": 1.074400520010387e-05, + "loss": 1.0329, + "step": 39950 + }, + { + "epoch": 2.95, + "learning_rate": 1.0743607433388516e-05, + "loss": 0.9911, + "step": 39951 + }, + { + "epoch": 2.95, + "learning_rate": 1.0743209665490096e-05, + "loss": 1.0827, + "step": 39952 + }, + { + "epoch": 2.95, + "learning_rate": 1.074281189640924e-05, + "loss": 0.9517, + "step": 39953 + }, + { + "epoch": 2.95, + "learning_rate": 1.0742414126146578e-05, + "loss": 0.9858, + "step": 39954 + }, + { + "epoch": 2.95, + "learning_rate": 1.0742016354702752e-05, + "loss": 0.9821, + "step": 39955 + }, + { + "epoch": 2.95, + "learning_rate": 1.0741618582078382e-05, + "loss": 1.0016, + "step": 39956 + }, + { + "epoch": 2.95, + "learning_rate": 1.0741220808274112e-05, + "loss": 0.8939, + "step": 39957 + }, + { + "epoch": 2.95, + "learning_rate": 1.0740823033290563e-05, + "loss": 1.0765, + "step": 39958 + }, + { + "epoch": 2.95, + "learning_rate": 1.0740425257128384e-05, + "loss": 1.0022, + "step": 39959 + }, + { + "epoch": 2.95, + "learning_rate": 1.0740027479788194e-05, + "loss": 0.8785, + "step": 39960 + }, + { + "epoch": 2.95, + "learning_rate": 1.0739629701270633e-05, + "loss": 0.8628, + "step": 39961 + }, + { + "epoch": 2.95, + "learning_rate": 1.0739231921576331e-05, + "loss": 0.979, + "step": 39962 + }, + { + "epoch": 2.95, + "learning_rate": 1.073883414070592e-05, + "loss": 1.0506, + "step": 39963 + }, + { + "epoch": 2.95, + "learning_rate": 1.0738436358660036e-05, + "loss": 1.0435, + "step": 39964 + }, + { + "epoch": 2.95, + "learning_rate": 1.073803857543931e-05, + "loss": 0.9929, + "step": 39965 + }, + { + "epoch": 2.95, + "learning_rate": 1.0737640791044377e-05, + "loss": 1.0065, + "step": 39966 + }, + { + "epoch": 2.95, + "learning_rate": 1.0737243005475865e-05, + "loss": 0.9907, + "step": 39967 + }, + { + "epoch": 2.95, + "learning_rate": 1.073684521873441e-05, + "loss": 1.0193, + "step": 39968 + }, + { + "epoch": 2.95, + "learning_rate": 1.0736447430820648e-05, + "loss": 1.005, + "step": 39969 + }, + { + "epoch": 2.95, + "learning_rate": 1.0736049641735206e-05, + "loss": 1.0262, + "step": 39970 + }, + { + "epoch": 2.95, + "learning_rate": 1.073565185147872e-05, + "loss": 1.0269, + "step": 39971 + }, + { + "epoch": 2.95, + "learning_rate": 1.0735254060051825e-05, + "loss": 1.1549, + "step": 39972 + }, + { + "epoch": 2.95, + "learning_rate": 1.073485626745515e-05, + "loss": 0.9771, + "step": 39973 + }, + { + "epoch": 2.95, + "learning_rate": 1.0734458473689327e-05, + "loss": 0.9961, + "step": 39974 + }, + { + "epoch": 2.95, + "learning_rate": 1.0734060678754992e-05, + "loss": 0.9809, + "step": 39975 + }, + { + "epoch": 2.95, + "learning_rate": 1.0733662882652778e-05, + "loss": 0.9671, + "step": 39976 + }, + { + "epoch": 2.95, + "learning_rate": 1.0733265085383315e-05, + "loss": 0.9743, + "step": 39977 + }, + { + "epoch": 2.95, + "learning_rate": 1.0732867286947239e-05, + "loss": 1.0055, + "step": 39978 + }, + { + "epoch": 2.95, + "learning_rate": 1.0732469487345181e-05, + "loss": 0.9482, + "step": 39979 + }, + { + "epoch": 2.95, + "learning_rate": 1.0732071686577776e-05, + "loss": 0.9438, + "step": 39980 + }, + { + "epoch": 2.95, + "learning_rate": 1.0731673884645654e-05, + "loss": 0.8767, + "step": 39981 + }, + { + "epoch": 2.95, + "learning_rate": 1.0731276081549453e-05, + "loss": 0.9253, + "step": 39982 + }, + { + "epoch": 2.95, + "learning_rate": 1.0730878277289799e-05, + "loss": 1.0508, + "step": 39983 + }, + { + "epoch": 2.95, + "learning_rate": 1.0730480471867329e-05, + "loss": 1.043, + "step": 39984 + }, + { + "epoch": 2.95, + "learning_rate": 1.073008266528267e-05, + "loss": 0.9111, + "step": 39985 + }, + { + "epoch": 2.95, + "learning_rate": 1.0729684857536467e-05, + "loss": 1.0732, + "step": 39986 + }, + { + "epoch": 2.95, + "learning_rate": 1.0729287048629339e-05, + "loss": 1.0262, + "step": 39987 + }, + { + "epoch": 2.95, + "learning_rate": 1.0728889238561932e-05, + "loss": 1.0882, + "step": 39988 + }, + { + "epoch": 2.95, + "learning_rate": 1.072849142733487e-05, + "loss": 0.9334, + "step": 39989 + }, + { + "epoch": 2.95, + "learning_rate": 1.0728093614948792e-05, + "loss": 0.9622, + "step": 39990 + }, + { + "epoch": 2.95, + "learning_rate": 1.0727695801404323e-05, + "loss": 1.0128, + "step": 39991 + }, + { + "epoch": 2.95, + "learning_rate": 1.0727297986702104e-05, + "loss": 0.9711, + "step": 39992 + }, + { + "epoch": 2.96, + "learning_rate": 1.072690017084276e-05, + "loss": 1.0283, + "step": 39993 + }, + { + "epoch": 2.96, + "learning_rate": 1.0726502353826931e-05, + "loss": 1.124, + "step": 39994 + }, + { + "epoch": 2.96, + "learning_rate": 1.0726104535655243e-05, + "loss": 1.0752, + "step": 39995 + }, + { + "epoch": 2.96, + "learning_rate": 1.072570671632834e-05, + "loss": 1.0938, + "step": 39996 + }, + { + "epoch": 2.96, + "learning_rate": 1.0725308895846842e-05, + "loss": 0.853, + "step": 39997 + }, + { + "epoch": 2.96, + "learning_rate": 1.0724911074211392e-05, + "loss": 0.9853, + "step": 39998 + }, + { + "epoch": 2.96, + "learning_rate": 1.0724513251422619e-05, + "loss": 1.0708, + "step": 39999 + }, + { + "epoch": 2.96, + "learning_rate": 1.0724115427481154e-05, + "loss": 1.0346, + "step": 40000 + }, + { + "epoch": 2.96, + "learning_rate": 1.0723717602387632e-05, + "loss": 0.9607, + "step": 40001 + }, + { + "epoch": 2.96, + "learning_rate": 1.0723319776142684e-05, + "loss": 1.0246, + "step": 40002 + }, + { + "epoch": 2.96, + "learning_rate": 1.0722921948746947e-05, + "loss": 1.0032, + "step": 40003 + }, + { + "epoch": 2.96, + "learning_rate": 1.072252412020105e-05, + "loss": 1.1155, + "step": 40004 + }, + { + "epoch": 2.96, + "learning_rate": 1.0722126290505628e-05, + "loss": 1.1117, + "step": 40005 + }, + { + "epoch": 2.96, + "learning_rate": 1.0721728459661314e-05, + "loss": 1.0855, + "step": 40006 + }, + { + "epoch": 2.96, + "learning_rate": 1.0721330627668738e-05, + "loss": 1.0707, + "step": 40007 + }, + { + "epoch": 2.96, + "learning_rate": 1.0720932794528538e-05, + "loss": 0.9128, + "step": 40008 + }, + { + "epoch": 2.96, + "learning_rate": 1.0720534960241345e-05, + "loss": 0.9182, + "step": 40009 + }, + { + "epoch": 2.96, + "learning_rate": 1.072013712480779e-05, + "loss": 0.9619, + "step": 40010 + }, + { + "epoch": 2.96, + "learning_rate": 1.0719739288228505e-05, + "loss": 1.0643, + "step": 40011 + }, + { + "epoch": 2.96, + "learning_rate": 1.0719341450504129e-05, + "loss": 1.0983, + "step": 40012 + }, + { + "epoch": 2.96, + "learning_rate": 1.0718943611635286e-05, + "loss": 1.0821, + "step": 40013 + }, + { + "epoch": 2.96, + "learning_rate": 1.0718545771622618e-05, + "loss": 0.9268, + "step": 40014 + }, + { + "epoch": 2.96, + "learning_rate": 1.0718147930466752e-05, + "loss": 1.0409, + "step": 40015 + }, + { + "epoch": 2.96, + "learning_rate": 1.0717750088168324e-05, + "loss": 0.9989, + "step": 40016 + }, + { + "epoch": 2.96, + "learning_rate": 1.0717352244727965e-05, + "loss": 0.9471, + "step": 40017 + }, + { + "epoch": 2.96, + "learning_rate": 1.0716954400146311e-05, + "loss": 0.9986, + "step": 40018 + }, + { + "epoch": 2.96, + "learning_rate": 1.071655655442399e-05, + "loss": 1.0463, + "step": 40019 + }, + { + "epoch": 2.96, + "learning_rate": 1.071615870756164e-05, + "loss": 1.1927, + "step": 40020 + }, + { + "epoch": 2.96, + "learning_rate": 1.071576085955989e-05, + "loss": 0.9858, + "step": 40021 + }, + { + "epoch": 2.96, + "learning_rate": 1.0715363010419376e-05, + "loss": 1.0512, + "step": 40022 + }, + { + "epoch": 2.96, + "learning_rate": 1.071496516014073e-05, + "loss": 0.9328, + "step": 40023 + }, + { + "epoch": 2.96, + "learning_rate": 1.0714567308724584e-05, + "loss": 1.0175, + "step": 40024 + }, + { + "epoch": 2.96, + "learning_rate": 1.071416945617157e-05, + "loss": 0.9131, + "step": 40025 + }, + { + "epoch": 2.96, + "learning_rate": 1.0713771602482324e-05, + "loss": 1.0905, + "step": 40026 + }, + { + "epoch": 2.96, + "learning_rate": 1.071337374765748e-05, + "loss": 1.0263, + "step": 40027 + }, + { + "epoch": 2.96, + "learning_rate": 1.0712975891697663e-05, + "loss": 1.0249, + "step": 40028 + }, + { + "epoch": 2.96, + "learning_rate": 1.0712578034603516e-05, + "loss": 0.9446, + "step": 40029 + }, + { + "epoch": 2.96, + "learning_rate": 1.0712180176375666e-05, + "loss": 0.9791, + "step": 40030 + }, + { + "epoch": 2.96, + "learning_rate": 1.0711782317014747e-05, + "loss": 1.0282, + "step": 40031 + }, + { + "epoch": 2.96, + "learning_rate": 1.0711384456521392e-05, + "loss": 0.9743, + "step": 40032 + }, + { + "epoch": 2.96, + "learning_rate": 1.0710986594896238e-05, + "loss": 0.9814, + "step": 40033 + }, + { + "epoch": 2.96, + "learning_rate": 1.071058873213991e-05, + "loss": 0.9786, + "step": 40034 + }, + { + "epoch": 2.96, + "learning_rate": 1.071019086825305e-05, + "loss": 0.9613, + "step": 40035 + }, + { + "epoch": 2.96, + "learning_rate": 1.0709793003236282e-05, + "loss": 1.0108, + "step": 40036 + }, + { + "epoch": 2.96, + "learning_rate": 1.070939513709025e-05, + "loss": 1.0854, + "step": 40037 + }, + { + "epoch": 2.96, + "learning_rate": 1.0708997269815575e-05, + "loss": 0.992, + "step": 40038 + }, + { + "epoch": 2.96, + "learning_rate": 1.0708599401412898e-05, + "loss": 0.9858, + "step": 40039 + }, + { + "epoch": 2.96, + "learning_rate": 1.0708201531882846e-05, + "loss": 1.0838, + "step": 40040 + }, + { + "epoch": 2.96, + "learning_rate": 1.070780366122606e-05, + "loss": 1.008, + "step": 40041 + }, + { + "epoch": 2.96, + "learning_rate": 1.0707405789443164e-05, + "loss": 0.9953, + "step": 40042 + }, + { + "epoch": 2.96, + "learning_rate": 1.07070079165348e-05, + "loss": 0.9569, + "step": 40043 + }, + { + "epoch": 2.96, + "learning_rate": 1.0706610042501593e-05, + "loss": 0.8762, + "step": 40044 + }, + { + "epoch": 2.96, + "learning_rate": 1.0706212167344182e-05, + "loss": 1.067, + "step": 40045 + }, + { + "epoch": 2.96, + "learning_rate": 1.0705814291063197e-05, + "loss": 1.0935, + "step": 40046 + }, + { + "epoch": 2.96, + "learning_rate": 1.0705416413659271e-05, + "loss": 0.9951, + "step": 40047 + }, + { + "epoch": 2.96, + "learning_rate": 1.0705018535133037e-05, + "loss": 0.9429, + "step": 40048 + }, + { + "epoch": 2.96, + "learning_rate": 1.0704620655485129e-05, + "loss": 0.9829, + "step": 40049 + }, + { + "epoch": 2.96, + "learning_rate": 1.0704222774716178e-05, + "loss": 1.0437, + "step": 40050 + }, + { + "epoch": 2.96, + "learning_rate": 1.0703824892826821e-05, + "loss": 1.0128, + "step": 40051 + }, + { + "epoch": 2.96, + "learning_rate": 1.0703427009817688e-05, + "loss": 1.0519, + "step": 40052 + }, + { + "epoch": 2.96, + "learning_rate": 1.0703029125689413e-05, + "loss": 1.1127, + "step": 40053 + }, + { + "epoch": 2.96, + "learning_rate": 1.0702631240442628e-05, + "loss": 1.0689, + "step": 40054 + }, + { + "epoch": 2.96, + "learning_rate": 1.0702233354077969e-05, + "loss": 0.9896, + "step": 40055 + }, + { + "epoch": 2.96, + "learning_rate": 1.0701835466596065e-05, + "loss": 1.0025, + "step": 40056 + }, + { + "epoch": 2.96, + "learning_rate": 1.0701437577997552e-05, + "loss": 1.0367, + "step": 40057 + }, + { + "epoch": 2.96, + "learning_rate": 1.0701039688283059e-05, + "loss": 0.9263, + "step": 40058 + }, + { + "epoch": 2.96, + "learning_rate": 1.0700641797453223e-05, + "loss": 0.965, + "step": 40059 + }, + { + "epoch": 2.96, + "learning_rate": 1.0700243905508675e-05, + "loss": 0.9643, + "step": 40060 + }, + { + "epoch": 2.96, + "learning_rate": 1.0699846012450052e-05, + "loss": 0.9122, + "step": 40061 + }, + { + "epoch": 2.96, + "learning_rate": 1.0699448118277983e-05, + "loss": 1.0429, + "step": 40062 + }, + { + "epoch": 2.96, + "learning_rate": 1.0699050222993101e-05, + "loss": 0.9775, + "step": 40063 + }, + { + "epoch": 2.96, + "learning_rate": 1.069865232659604e-05, + "loss": 0.9643, + "step": 40064 + }, + { + "epoch": 2.96, + "learning_rate": 1.0698254429087433e-05, + "loss": 1.0203, + "step": 40065 + }, + { + "epoch": 2.96, + "learning_rate": 1.0697856530467916e-05, + "loss": 1.0685, + "step": 40066 + }, + { + "epoch": 2.96, + "learning_rate": 1.0697458630738116e-05, + "loss": 0.928, + "step": 40067 + }, + { + "epoch": 2.96, + "learning_rate": 1.0697060729898672e-05, + "loss": 0.9657, + "step": 40068 + }, + { + "epoch": 2.96, + "learning_rate": 1.0696662827950213e-05, + "loss": 0.9546, + "step": 40069 + }, + { + "epoch": 2.96, + "learning_rate": 1.0696264924893372e-05, + "loss": 0.949, + "step": 40070 + }, + { + "epoch": 2.96, + "learning_rate": 1.0695867020728785e-05, + "loss": 1.0534, + "step": 40071 + }, + { + "epoch": 2.96, + "learning_rate": 1.0695469115457082e-05, + "loss": 1.0519, + "step": 40072 + }, + { + "epoch": 2.96, + "learning_rate": 1.06950712090789e-05, + "loss": 0.9632, + "step": 40073 + }, + { + "epoch": 2.96, + "learning_rate": 1.0694673301594869e-05, + "loss": 1.0295, + "step": 40074 + }, + { + "epoch": 2.96, + "learning_rate": 1.0694275393005622e-05, + "loss": 1.0378, + "step": 40075 + }, + { + "epoch": 2.96, + "learning_rate": 1.0693877483311794e-05, + "loss": 1.0902, + "step": 40076 + }, + { + "epoch": 2.96, + "learning_rate": 1.0693479572514014e-05, + "loss": 0.9365, + "step": 40077 + }, + { + "epoch": 2.96, + "learning_rate": 1.069308166061292e-05, + "loss": 1.0558, + "step": 40078 + }, + { + "epoch": 2.96, + "learning_rate": 1.0692683747609143e-05, + "loss": 1.0102, + "step": 40079 + }, + { + "epoch": 2.96, + "learning_rate": 1.0692285833503315e-05, + "loss": 1.0151, + "step": 40080 + }, + { + "epoch": 2.96, + "learning_rate": 1.0691887918296072e-05, + "loss": 1.0135, + "step": 40081 + }, + { + "epoch": 2.96, + "learning_rate": 1.0691490001988042e-05, + "loss": 1.0544, + "step": 40082 + }, + { + "epoch": 2.96, + "learning_rate": 1.0691092084579863e-05, + "loss": 1.0253, + "step": 40083 + }, + { + "epoch": 2.96, + "learning_rate": 1.069069416607217e-05, + "loss": 0.9764, + "step": 40084 + }, + { + "epoch": 2.96, + "learning_rate": 1.0690296246465589e-05, + "loss": 0.9654, + "step": 40085 + }, + { + "epoch": 2.96, + "learning_rate": 1.0689898325760757e-05, + "loss": 1.0904, + "step": 40086 + }, + { + "epoch": 2.96, + "learning_rate": 1.0689500403958307e-05, + "loss": 1.0055, + "step": 40087 + }, + { + "epoch": 2.96, + "learning_rate": 1.068910248105887e-05, + "loss": 1.0509, + "step": 40088 + }, + { + "epoch": 2.96, + "learning_rate": 1.068870455706308e-05, + "loss": 0.9675, + "step": 40089 + }, + { + "epoch": 2.96, + "learning_rate": 1.0688306631971574e-05, + "loss": 1.0024, + "step": 40090 + }, + { + "epoch": 2.96, + "learning_rate": 1.0687908705784982e-05, + "loss": 0.9981, + "step": 40091 + }, + { + "epoch": 2.96, + "learning_rate": 1.0687510778503936e-05, + "loss": 1.0235, + "step": 40092 + }, + { + "epoch": 2.96, + "learning_rate": 1.068711285012907e-05, + "loss": 0.9669, + "step": 40093 + }, + { + "epoch": 2.96, + "learning_rate": 1.0686714920661019e-05, + "loss": 0.9495, + "step": 40094 + }, + { + "epoch": 2.96, + "learning_rate": 1.0686316990100412e-05, + "loss": 1.0155, + "step": 40095 + }, + { + "epoch": 2.96, + "learning_rate": 1.0685919058447886e-05, + "loss": 1.0025, + "step": 40096 + }, + { + "epoch": 2.96, + "learning_rate": 1.068552112570407e-05, + "loss": 1.046, + "step": 40097 + }, + { + "epoch": 2.96, + "learning_rate": 1.0685123191869603e-05, + "loss": 0.8039, + "step": 40098 + }, + { + "epoch": 2.96, + "learning_rate": 1.0684725256945111e-05, + "loss": 0.9299, + "step": 40099 + }, + { + "epoch": 2.96, + "learning_rate": 1.0684327320931235e-05, + "loss": 1.0366, + "step": 40100 + }, + { + "epoch": 2.96, + "learning_rate": 1.0683929383828601e-05, + "loss": 0.9102, + "step": 40101 + }, + { + "epoch": 2.96, + "learning_rate": 1.0683531445637852e-05, + "loss": 0.9703, + "step": 40102 + }, + { + "epoch": 2.96, + "learning_rate": 1.0683133506359606e-05, + "loss": 1.0306, + "step": 40103 + }, + { + "epoch": 2.96, + "learning_rate": 1.068273556599451e-05, + "loss": 0.973, + "step": 40104 + }, + { + "epoch": 2.96, + "learning_rate": 1.0682337624543186e-05, + "loss": 0.9747, + "step": 40105 + }, + { + "epoch": 2.96, + "learning_rate": 1.0681939682006276e-05, + "loss": 1.0232, + "step": 40106 + }, + { + "epoch": 2.96, + "learning_rate": 1.0681541738384407e-05, + "loss": 1.0689, + "step": 40107 + }, + { + "epoch": 2.96, + "learning_rate": 1.068114379367822e-05, + "loss": 0.9373, + "step": 40108 + }, + { + "epoch": 2.96, + "learning_rate": 1.0680745847888336e-05, + "loss": 1.0211, + "step": 40109 + }, + { + "epoch": 2.96, + "learning_rate": 1.0680347901015399e-05, + "loss": 0.918, + "step": 40110 + }, + { + "epoch": 2.96, + "learning_rate": 1.0679949953060039e-05, + "loss": 1.0323, + "step": 40111 + }, + { + "epoch": 2.96, + "learning_rate": 1.0679552004022889e-05, + "loss": 0.9772, + "step": 40112 + }, + { + "epoch": 2.96, + "learning_rate": 1.0679154053904579e-05, + "loss": 0.9914, + "step": 40113 + }, + { + "epoch": 2.96, + "learning_rate": 1.0678756102705747e-05, + "loss": 0.9409, + "step": 40114 + }, + { + "epoch": 2.96, + "learning_rate": 1.0678358150427022e-05, + "loss": 0.9636, + "step": 40115 + }, + { + "epoch": 2.96, + "learning_rate": 1.067796019706904e-05, + "loss": 0.9835, + "step": 40116 + }, + { + "epoch": 2.96, + "learning_rate": 1.0677562242632429e-05, + "loss": 1.0046, + "step": 40117 + }, + { + "epoch": 2.96, + "learning_rate": 1.0677164287117831e-05, + "loss": 1.0196, + "step": 40118 + }, + { + "epoch": 2.96, + "learning_rate": 1.0676766330525872e-05, + "loss": 0.9793, + "step": 40119 + }, + { + "epoch": 2.96, + "learning_rate": 1.0676368372857189e-05, + "loss": 0.9816, + "step": 40120 + }, + { + "epoch": 2.96, + "learning_rate": 1.0675970414112413e-05, + "loss": 0.9658, + "step": 40121 + }, + { + "epoch": 2.96, + "learning_rate": 1.0675572454292176e-05, + "loss": 1.1378, + "step": 40122 + }, + { + "epoch": 2.96, + "learning_rate": 1.0675174493397117e-05, + "loss": 1.0145, + "step": 40123 + }, + { + "epoch": 2.96, + "learning_rate": 1.0674776531427859e-05, + "loss": 0.984, + "step": 40124 + }, + { + "epoch": 2.96, + "learning_rate": 1.0674378568385046e-05, + "loss": 0.9951, + "step": 40125 + }, + { + "epoch": 2.96, + "learning_rate": 1.0673980604269305e-05, + "loss": 1.0965, + "step": 40126 + }, + { + "epoch": 2.96, + "learning_rate": 1.067358263908127e-05, + "loss": 0.9477, + "step": 40127 + }, + { + "epoch": 2.97, + "learning_rate": 1.0673184672821575e-05, + "loss": 0.995, + "step": 40128 + }, + { + "epoch": 2.97, + "learning_rate": 1.0672786705490854e-05, + "loss": 1.0534, + "step": 40129 + }, + { + "epoch": 2.97, + "learning_rate": 1.0672388737089738e-05, + "loss": 0.8432, + "step": 40130 + }, + { + "epoch": 2.97, + "learning_rate": 1.0671990767618863e-05, + "loss": 0.9842, + "step": 40131 + }, + { + "epoch": 2.97, + "learning_rate": 1.0671592797078855e-05, + "loss": 0.9905, + "step": 40132 + }, + { + "epoch": 2.97, + "learning_rate": 1.0671194825470357e-05, + "loss": 1.0103, + "step": 40133 + }, + { + "epoch": 2.97, + "learning_rate": 1.0670796852793996e-05, + "loss": 0.9758, + "step": 40134 + }, + { + "epoch": 2.97, + "learning_rate": 1.0670398879050408e-05, + "loss": 1.0713, + "step": 40135 + }, + { + "epoch": 2.97, + "learning_rate": 1.0670000904240224e-05, + "loss": 1.1838, + "step": 40136 + }, + { + "epoch": 2.97, + "learning_rate": 1.0669602928364078e-05, + "loss": 1.0438, + "step": 40137 + }, + { + "epoch": 2.97, + "learning_rate": 1.0669204951422603e-05, + "loss": 1.0113, + "step": 40138 + }, + { + "epoch": 2.97, + "learning_rate": 1.0668806973416434e-05, + "loss": 1.0221, + "step": 40139 + }, + { + "epoch": 2.97, + "learning_rate": 1.0668408994346201e-05, + "loss": 1.0034, + "step": 40140 + }, + { + "epoch": 2.97, + "learning_rate": 1.0668011014212542e-05, + "loss": 0.9403, + "step": 40141 + }, + { + "epoch": 2.97, + "learning_rate": 1.0667613033016084e-05, + "loss": 0.9836, + "step": 40142 + }, + { + "epoch": 2.97, + "learning_rate": 1.0667215050757465e-05, + "loss": 1.0546, + "step": 40143 + }, + { + "epoch": 2.97, + "learning_rate": 1.0666817067437313e-05, + "loss": 1.0104, + "step": 40144 + }, + { + "epoch": 2.97, + "learning_rate": 1.066641908305627e-05, + "loss": 1.002, + "step": 40145 + }, + { + "epoch": 2.97, + "learning_rate": 1.0666021097614958e-05, + "loss": 0.8965, + "step": 40146 + }, + { + "epoch": 2.97, + "learning_rate": 1.0665623111114019e-05, + "loss": 1.1156, + "step": 40147 + }, + { + "epoch": 2.97, + "learning_rate": 1.0665225123554081e-05, + "loss": 1.0255, + "step": 40148 + }, + { + "epoch": 2.97, + "learning_rate": 1.0664827134935785e-05, + "loss": 0.9797, + "step": 40149 + }, + { + "epoch": 2.97, + "learning_rate": 1.0664429145259754e-05, + "loss": 1.0286, + "step": 40150 + }, + { + "epoch": 2.97, + "learning_rate": 1.066403115452663e-05, + "loss": 0.8713, + "step": 40151 + }, + { + "epoch": 2.97, + "learning_rate": 1.0663633162737034e-05, + "loss": 0.9738, + "step": 40152 + }, + { + "epoch": 2.97, + "learning_rate": 1.0663235169891613e-05, + "loss": 0.9891, + "step": 40153 + }, + { + "epoch": 2.97, + "learning_rate": 1.0662837175990992e-05, + "loss": 1.0069, + "step": 40154 + }, + { + "epoch": 2.97, + "learning_rate": 1.0662439181035808e-05, + "loss": 0.9941, + "step": 40155 + }, + { + "epoch": 2.97, + "learning_rate": 1.066204118502669e-05, + "loss": 1.0108, + "step": 40156 + }, + { + "epoch": 2.97, + "learning_rate": 1.0661643187964277e-05, + "loss": 0.9674, + "step": 40157 + }, + { + "epoch": 2.97, + "learning_rate": 1.0661245189849201e-05, + "loss": 0.9462, + "step": 40158 + }, + { + "epoch": 2.97, + "learning_rate": 1.066084719068209e-05, + "loss": 1.0346, + "step": 40159 + }, + { + "epoch": 2.97, + "learning_rate": 1.066044919046358e-05, + "loss": 1.0745, + "step": 40160 + }, + { + "epoch": 2.97, + "learning_rate": 1.0660051189194306e-05, + "loss": 0.919, + "step": 40161 + }, + { + "epoch": 2.97, + "learning_rate": 1.0659653186874901e-05, + "loss": 0.9906, + "step": 40162 + }, + { + "epoch": 2.97, + "learning_rate": 1.0659255183505997e-05, + "loss": 0.9179, + "step": 40163 + }, + { + "epoch": 2.97, + "learning_rate": 1.0658857179088225e-05, + "loss": 0.9869, + "step": 40164 + }, + { + "epoch": 2.97, + "learning_rate": 1.0658459173622222e-05, + "loss": 1.049, + "step": 40165 + }, + { + "epoch": 2.97, + "learning_rate": 1.065806116710862e-05, + "loss": 0.927, + "step": 40166 + }, + { + "epoch": 2.97, + "learning_rate": 1.0657663159548053e-05, + "loss": 0.9819, + "step": 40167 + }, + { + "epoch": 2.97, + "learning_rate": 1.0657265150941154e-05, + "loss": 1.0804, + "step": 40168 + }, + { + "epoch": 2.97, + "learning_rate": 1.065686714128855e-05, + "loss": 1.0873, + "step": 40169 + }, + { + "epoch": 2.97, + "learning_rate": 1.0656469130590887e-05, + "loss": 1.0759, + "step": 40170 + }, + { + "epoch": 2.97, + "learning_rate": 1.0656071118848787e-05, + "loss": 1.0542, + "step": 40171 + }, + { + "epoch": 2.97, + "learning_rate": 1.0655673106062889e-05, + "loss": 1.1268, + "step": 40172 + }, + { + "epoch": 2.97, + "learning_rate": 1.0655275092233822e-05, + "loss": 0.9399, + "step": 40173 + }, + { + "epoch": 2.97, + "learning_rate": 1.0654877077362225e-05, + "loss": 1.0967, + "step": 40174 + }, + { + "epoch": 2.97, + "learning_rate": 1.0654479061448725e-05, + "loss": 0.9369, + "step": 40175 + }, + { + "epoch": 2.97, + "learning_rate": 1.065408104449396e-05, + "loss": 1.0048, + "step": 40176 + }, + { + "epoch": 2.97, + "learning_rate": 1.0653683026498563e-05, + "loss": 1.02, + "step": 40177 + }, + { + "epoch": 2.97, + "learning_rate": 1.0653285007463164e-05, + "loss": 1.0086, + "step": 40178 + }, + { + "epoch": 2.97, + "learning_rate": 1.0652886987388396e-05, + "loss": 0.9393, + "step": 40179 + }, + { + "epoch": 2.97, + "learning_rate": 1.0652488966274897e-05, + "loss": 1.0537, + "step": 40180 + }, + { + "epoch": 2.97, + "learning_rate": 1.0652090944123293e-05, + "loss": 1.0257, + "step": 40181 + }, + { + "epoch": 2.97, + "learning_rate": 1.0651692920934225e-05, + "loss": 1.0305, + "step": 40182 + }, + { + "epoch": 2.97, + "learning_rate": 1.0651294896708322e-05, + "loss": 0.9761, + "step": 40183 + }, + { + "epoch": 2.97, + "learning_rate": 1.065089687144622e-05, + "loss": 0.9688, + "step": 40184 + }, + { + "epoch": 2.97, + "learning_rate": 1.0650498845148547e-05, + "loss": 1.0934, + "step": 40185 + }, + { + "epoch": 2.97, + "learning_rate": 1.0650100817815945e-05, + "loss": 0.9212, + "step": 40186 + }, + { + "epoch": 2.97, + "learning_rate": 1.0649702789449036e-05, + "loss": 0.9643, + "step": 40187 + }, + { + "epoch": 2.97, + "learning_rate": 1.0649304760048465e-05, + "loss": 1.0892, + "step": 40188 + }, + { + "epoch": 2.97, + "learning_rate": 1.0648906729614854e-05, + "loss": 1.1023, + "step": 40189 + }, + { + "epoch": 2.97, + "learning_rate": 1.0648508698148845e-05, + "loss": 1.0458, + "step": 40190 + }, + { + "epoch": 2.97, + "learning_rate": 1.0648110665651067e-05, + "loss": 1.04, + "step": 40191 + }, + { + "epoch": 2.97, + "learning_rate": 1.0647712632122151e-05, + "loss": 1.0101, + "step": 40192 + }, + { + "epoch": 2.97, + "learning_rate": 1.0647314597562737e-05, + "loss": 0.9706, + "step": 40193 + }, + { + "epoch": 2.97, + "learning_rate": 1.0646916561973455e-05, + "loss": 0.9756, + "step": 40194 + }, + { + "epoch": 2.97, + "learning_rate": 1.0646518525354936e-05, + "loss": 1.0154, + "step": 40195 + }, + { + "epoch": 2.97, + "learning_rate": 1.0646120487707818e-05, + "loss": 1.0024, + "step": 40196 + }, + { + "epoch": 2.97, + "learning_rate": 1.0645722449032729e-05, + "loss": 0.955, + "step": 40197 + }, + { + "epoch": 2.97, + "learning_rate": 1.0645324409330307e-05, + "loss": 1.0658, + "step": 40198 + }, + { + "epoch": 2.97, + "learning_rate": 1.0644926368601181e-05, + "loss": 0.9826, + "step": 40199 + }, + { + "epoch": 2.97, + "learning_rate": 1.0644528326845989e-05, + "loss": 1.0651, + "step": 40200 + }, + { + "epoch": 2.97, + "learning_rate": 1.0644130284065358e-05, + "loss": 0.9427, + "step": 40201 + }, + { + "epoch": 2.97, + "learning_rate": 1.0643732240259927e-05, + "loss": 0.9899, + "step": 40202 + }, + { + "epoch": 2.97, + "learning_rate": 1.0643334195430326e-05, + "loss": 1.0839, + "step": 40203 + }, + { + "epoch": 2.97, + "learning_rate": 1.0642936149577192e-05, + "loss": 0.8756, + "step": 40204 + }, + { + "epoch": 2.97, + "learning_rate": 1.0642538102701155e-05, + "loss": 0.9859, + "step": 40205 + }, + { + "epoch": 2.97, + "learning_rate": 1.0642140054802849e-05, + "loss": 1.0802, + "step": 40206 + }, + { + "epoch": 2.97, + "learning_rate": 1.0641742005882906e-05, + "loss": 0.9345, + "step": 40207 + }, + { + "epoch": 2.97, + "learning_rate": 1.0641343955941964e-05, + "loss": 1.0235, + "step": 40208 + }, + { + "epoch": 2.97, + "learning_rate": 1.0640945904980648e-05, + "loss": 1.0702, + "step": 40209 + }, + { + "epoch": 2.97, + "learning_rate": 1.06405478529996e-05, + "loss": 1.118, + "step": 40210 + }, + { + "epoch": 2.97, + "learning_rate": 1.0640149799999446e-05, + "loss": 1.0652, + "step": 40211 + }, + { + "epoch": 2.97, + "learning_rate": 1.0639751745980826e-05, + "loss": 1.0041, + "step": 40212 + }, + { + "epoch": 2.97, + "learning_rate": 1.0639353690944367e-05, + "loss": 0.9107, + "step": 40213 + }, + { + "epoch": 2.97, + "learning_rate": 1.063895563489071e-05, + "loss": 1.0846, + "step": 40214 + }, + { + "epoch": 2.97, + "learning_rate": 1.063855757782048e-05, + "loss": 1.0254, + "step": 40215 + }, + { + "epoch": 2.97, + "learning_rate": 1.0638159519734314e-05, + "loss": 1.028, + "step": 40216 + }, + { + "epoch": 2.97, + "learning_rate": 1.063776146063285e-05, + "loss": 0.8939, + "step": 40217 + }, + { + "epoch": 2.97, + "learning_rate": 1.0637363400516713e-05, + "loss": 0.9305, + "step": 40218 + }, + { + "epoch": 2.97, + "learning_rate": 1.063696533938654e-05, + "loss": 1.0385, + "step": 40219 + }, + { + "epoch": 2.97, + "learning_rate": 1.0636567277242964e-05, + "loss": 0.9654, + "step": 40220 + }, + { + "epoch": 2.97, + "learning_rate": 1.0636169214086619e-05, + "loss": 0.9458, + "step": 40221 + }, + { + "epoch": 2.97, + "learning_rate": 1.063577114991814e-05, + "loss": 1.0181, + "step": 40222 + }, + { + "epoch": 2.97, + "learning_rate": 1.0635373084738153e-05, + "loss": 0.8958, + "step": 40223 + }, + { + "epoch": 2.97, + "learning_rate": 1.06349750185473e-05, + "loss": 0.9419, + "step": 40224 + }, + { + "epoch": 2.97, + "learning_rate": 1.0634576951346211e-05, + "loss": 0.9791, + "step": 40225 + }, + { + "epoch": 2.97, + "learning_rate": 1.0634178883135518e-05, + "loss": 0.9458, + "step": 40226 + }, + { + "epoch": 2.97, + "learning_rate": 1.0633780813915856e-05, + "loss": 0.9831, + "step": 40227 + }, + { + "epoch": 2.97, + "learning_rate": 1.063338274368786e-05, + "loss": 0.9958, + "step": 40228 + }, + { + "epoch": 2.97, + "learning_rate": 1.0632984672452158e-05, + "loss": 0.9655, + "step": 40229 + }, + { + "epoch": 2.97, + "learning_rate": 1.0632586600209387e-05, + "loss": 1.0599, + "step": 40230 + }, + { + "epoch": 2.97, + "learning_rate": 1.0632188526960181e-05, + "loss": 1.0382, + "step": 40231 + }, + { + "epoch": 2.97, + "learning_rate": 1.0631790452705168e-05, + "loss": 1.019, + "step": 40232 + }, + { + "epoch": 2.97, + "learning_rate": 1.0631392377444994e-05, + "loss": 0.9599, + "step": 40233 + }, + { + "epoch": 2.97, + "learning_rate": 1.0630994301180275e-05, + "loss": 0.9822, + "step": 40234 + }, + { + "epoch": 2.97, + "learning_rate": 1.0630596223911657e-05, + "loss": 0.9712, + "step": 40235 + }, + { + "epoch": 2.97, + "learning_rate": 1.063019814563977e-05, + "loss": 1.0394, + "step": 40236 + }, + { + "epoch": 2.97, + "learning_rate": 1.0629800066365246e-05, + "loss": 0.963, + "step": 40237 + }, + { + "epoch": 2.97, + "learning_rate": 1.062940198608872e-05, + "loss": 0.969, + "step": 40238 + }, + { + "epoch": 2.97, + "learning_rate": 1.0629003904810822e-05, + "loss": 1.0268, + "step": 40239 + }, + { + "epoch": 2.97, + "learning_rate": 1.0628605822532188e-05, + "loss": 0.9266, + "step": 40240 + }, + { + "epoch": 2.97, + "learning_rate": 1.0628207739253453e-05, + "loss": 1.1161, + "step": 40241 + }, + { + "epoch": 2.97, + "learning_rate": 1.062780965497525e-05, + "loss": 0.9237, + "step": 40242 + }, + { + "epoch": 2.97, + "learning_rate": 1.0627411569698207e-05, + "loss": 0.9238, + "step": 40243 + }, + { + "epoch": 2.97, + "learning_rate": 1.0627013483422962e-05, + "loss": 1.0553, + "step": 40244 + }, + { + "epoch": 2.97, + "learning_rate": 1.062661539615015e-05, + "loss": 1.0053, + "step": 40245 + }, + { + "epoch": 2.97, + "learning_rate": 1.0626217307880398e-05, + "loss": 1.041, + "step": 40246 + }, + { + "epoch": 2.97, + "learning_rate": 1.0625819218614346e-05, + "loss": 1.0014, + "step": 40247 + }, + { + "epoch": 2.97, + "learning_rate": 1.0625421128352624e-05, + "loss": 1.0374, + "step": 40248 + }, + { + "epoch": 2.97, + "learning_rate": 1.0625023037095868e-05, + "loss": 0.8928, + "step": 40249 + }, + { + "epoch": 2.97, + "learning_rate": 1.0624624944844704e-05, + "loss": 1.0311, + "step": 40250 + }, + { + "epoch": 2.97, + "learning_rate": 1.0624226851599775e-05, + "loss": 0.974, + "step": 40251 + }, + { + "epoch": 2.97, + "learning_rate": 1.062382875736171e-05, + "loss": 1.0428, + "step": 40252 + }, + { + "epoch": 2.97, + "learning_rate": 1.0623430662131143e-05, + "loss": 1.0466, + "step": 40253 + }, + { + "epoch": 2.97, + "learning_rate": 1.0623032565908701e-05, + "loss": 0.986, + "step": 40254 + }, + { + "epoch": 2.97, + "learning_rate": 1.062263446869503e-05, + "loss": 0.9503, + "step": 40255 + }, + { + "epoch": 2.97, + "learning_rate": 1.062223637049075e-05, + "loss": 1.0218, + "step": 40256 + }, + { + "epoch": 2.97, + "learning_rate": 1.0621838271296507e-05, + "loss": 0.9215, + "step": 40257 + }, + { + "epoch": 2.97, + "learning_rate": 1.0621440171112925e-05, + "loss": 0.9948, + "step": 40258 + }, + { + "epoch": 2.97, + "learning_rate": 1.062104206994064e-05, + "loss": 0.9265, + "step": 40259 + }, + { + "epoch": 2.97, + "learning_rate": 1.0620643967780286e-05, + "loss": 0.8532, + "step": 40260 + }, + { + "epoch": 2.97, + "learning_rate": 1.06202458646325e-05, + "loss": 0.9823, + "step": 40261 + }, + { + "epoch": 2.97, + "learning_rate": 1.061984776049791e-05, + "loss": 0.8479, + "step": 40262 + }, + { + "epoch": 2.98, + "learning_rate": 1.061944965537715e-05, + "loss": 0.9181, + "step": 40263 + }, + { + "epoch": 2.98, + "learning_rate": 1.0619051549270853e-05, + "loss": 1.0591, + "step": 40264 + }, + { + "epoch": 2.98, + "learning_rate": 1.0618653442179656e-05, + "loss": 0.9297, + "step": 40265 + }, + { + "epoch": 2.98, + "learning_rate": 1.0618255334104189e-05, + "loss": 0.9162, + "step": 40266 + }, + { + "epoch": 2.98, + "learning_rate": 1.0617857225045086e-05, + "loss": 0.8897, + "step": 40267 + }, + { + "epoch": 2.98, + "learning_rate": 1.0617459115002984e-05, + "loss": 1.035, + "step": 40268 + }, + { + "epoch": 2.98, + "learning_rate": 1.0617061003978512e-05, + "loss": 0.9788, + "step": 40269 + }, + { + "epoch": 2.98, + "learning_rate": 1.0616662891972305e-05, + "loss": 0.9583, + "step": 40270 + }, + { + "epoch": 2.98, + "learning_rate": 1.0616264778984995e-05, + "loss": 0.8846, + "step": 40271 + }, + { + "epoch": 2.98, + "learning_rate": 1.0615866665017218e-05, + "loss": 0.9891, + "step": 40272 + }, + { + "epoch": 2.98, + "learning_rate": 1.0615468550069602e-05, + "loss": 0.9561, + "step": 40273 + }, + { + "epoch": 2.98, + "learning_rate": 1.0615070434142789e-05, + "loss": 1.0593, + "step": 40274 + }, + { + "epoch": 2.98, + "learning_rate": 1.0614672317237407e-05, + "loss": 1.1305, + "step": 40275 + }, + { + "epoch": 2.98, + "learning_rate": 1.0614274199354088e-05, + "loss": 1.0564, + "step": 40276 + }, + { + "epoch": 2.98, + "learning_rate": 1.0613876080493468e-05, + "loss": 0.9847, + "step": 40277 + }, + { + "epoch": 2.98, + "learning_rate": 1.0613477960656183e-05, + "loss": 1.0508, + "step": 40278 + }, + { + "epoch": 2.98, + "learning_rate": 1.0613079839842862e-05, + "loss": 0.9295, + "step": 40279 + }, + { + "epoch": 2.98, + "learning_rate": 1.0612681718054138e-05, + "loss": 1.0093, + "step": 40280 + }, + { + "epoch": 2.98, + "learning_rate": 1.0612283595290647e-05, + "loss": 0.9107, + "step": 40281 + }, + { + "epoch": 2.98, + "learning_rate": 1.0611885471553023e-05, + "loss": 1.0353, + "step": 40282 + }, + { + "epoch": 2.98, + "learning_rate": 1.0611487346841895e-05, + "loss": 0.9498, + "step": 40283 + }, + { + "epoch": 2.98, + "learning_rate": 1.0611089221157903e-05, + "loss": 1.0729, + "step": 40284 + }, + { + "epoch": 2.98, + "learning_rate": 1.0610691094501672e-05, + "loss": 0.9694, + "step": 40285 + }, + { + "epoch": 2.98, + "learning_rate": 1.0610292966873846e-05, + "loss": 0.8902, + "step": 40286 + }, + { + "epoch": 2.98, + "learning_rate": 1.0609894838275045e-05, + "loss": 1.0989, + "step": 40287 + }, + { + "epoch": 2.98, + "learning_rate": 1.0609496708705917e-05, + "loss": 1.1494, + "step": 40288 + }, + { + "epoch": 2.98, + "learning_rate": 1.0609098578167088e-05, + "loss": 1.0063, + "step": 40289 + }, + { + "epoch": 2.98, + "learning_rate": 1.060870044665919e-05, + "loss": 1.0579, + "step": 40290 + }, + { + "epoch": 2.98, + "learning_rate": 1.0608302314182857e-05, + "loss": 1.0528, + "step": 40291 + }, + { + "epoch": 2.98, + "learning_rate": 1.0607904180738726e-05, + "loss": 1.0711, + "step": 40292 + }, + { + "epoch": 2.98, + "learning_rate": 1.0607506046327425e-05, + "loss": 0.9295, + "step": 40293 + }, + { + "epoch": 2.98, + "learning_rate": 1.0607107910949594e-05, + "loss": 0.9161, + "step": 40294 + }, + { + "epoch": 2.98, + "learning_rate": 1.0606709774605861e-05, + "loss": 0.9772, + "step": 40295 + }, + { + "epoch": 2.98, + "learning_rate": 1.0606311637296862e-05, + "loss": 1.0701, + "step": 40296 + }, + { + "epoch": 2.98, + "learning_rate": 1.060591349902323e-05, + "loss": 0.9751, + "step": 40297 + }, + { + "epoch": 2.98, + "learning_rate": 1.06055153597856e-05, + "loss": 1.0192, + "step": 40298 + }, + { + "epoch": 2.98, + "learning_rate": 1.0605117219584602e-05, + "loss": 1.039, + "step": 40299 + }, + { + "epoch": 2.98, + "learning_rate": 1.060471907842087e-05, + "loss": 0.96, + "step": 40300 + }, + { + "epoch": 2.98, + "learning_rate": 1.060432093629504e-05, + "loss": 1.0738, + "step": 40301 + }, + { + "epoch": 2.98, + "learning_rate": 1.0603922793207744e-05, + "loss": 1.0332, + "step": 40302 + }, + { + "epoch": 2.98, + "learning_rate": 1.0603524649159615e-05, + "loss": 1.0849, + "step": 40303 + }, + { + "epoch": 2.98, + "learning_rate": 1.060312650415129e-05, + "loss": 1.0467, + "step": 40304 + }, + { + "epoch": 2.98, + "learning_rate": 1.0602728358183394e-05, + "loss": 1.0199, + "step": 40305 + }, + { + "epoch": 2.98, + "learning_rate": 1.0602330211256569e-05, + "loss": 0.9358, + "step": 40306 + }, + { + "epoch": 2.98, + "learning_rate": 1.0601932063371443e-05, + "loss": 0.9872, + "step": 40307 + }, + { + "epoch": 2.98, + "learning_rate": 1.0601533914528656e-05, + "loss": 1.0641, + "step": 40308 + }, + { + "epoch": 2.98, + "learning_rate": 1.0601135764728834e-05, + "loss": 1.102, + "step": 40309 + }, + { + "epoch": 2.98, + "learning_rate": 1.0600737613972614e-05, + "loss": 0.9414, + "step": 40310 + }, + { + "epoch": 2.98, + "learning_rate": 1.0600339462260627e-05, + "loss": 0.9959, + "step": 40311 + }, + { + "epoch": 2.98, + "learning_rate": 1.0599941309593514e-05, + "loss": 1.0689, + "step": 40312 + }, + { + "epoch": 2.98, + "learning_rate": 1.0599543155971896e-05, + "loss": 0.8149, + "step": 40313 + }, + { + "epoch": 2.98, + "learning_rate": 1.0599145001396418e-05, + "loss": 0.9477, + "step": 40314 + }, + { + "epoch": 2.98, + "learning_rate": 1.0598746845867708e-05, + "loss": 1.0322, + "step": 40315 + }, + { + "epoch": 2.98, + "learning_rate": 1.05983486893864e-05, + "loss": 1.0167, + "step": 40316 + }, + { + "epoch": 2.98, + "learning_rate": 1.0597950531953129e-05, + "loss": 0.972, + "step": 40317 + }, + { + "epoch": 2.98, + "learning_rate": 1.0597552373568525e-05, + "loss": 0.8949, + "step": 40318 + }, + { + "epoch": 2.98, + "learning_rate": 1.0597154214233228e-05, + "loss": 0.9898, + "step": 40319 + }, + { + "epoch": 2.98, + "learning_rate": 1.0596756053947862e-05, + "loss": 1.0599, + "step": 40320 + }, + { + "epoch": 2.98, + "learning_rate": 1.0596357892713068e-05, + "loss": 1.1109, + "step": 40321 + }, + { + "epoch": 2.98, + "learning_rate": 1.0595959730529475e-05, + "loss": 1.0766, + "step": 40322 + }, + { + "epoch": 2.98, + "learning_rate": 1.0595561567397722e-05, + "loss": 1.0422, + "step": 40323 + }, + { + "epoch": 2.98, + "learning_rate": 1.0595163403318438e-05, + "loss": 1.0281, + "step": 40324 + }, + { + "epoch": 2.98, + "learning_rate": 1.0594765238292257e-05, + "loss": 1.0043, + "step": 40325 + }, + { + "epoch": 2.98, + "learning_rate": 1.0594367072319813e-05, + "loss": 1.003, + "step": 40326 + }, + { + "epoch": 2.98, + "learning_rate": 1.0593968905401742e-05, + "loss": 0.9771, + "step": 40327 + }, + { + "epoch": 2.98, + "learning_rate": 1.059357073753867e-05, + "loss": 1.0282, + "step": 40328 + }, + { + "epoch": 2.98, + "learning_rate": 1.059317256873124e-05, + "loss": 1.0094, + "step": 40329 + }, + { + "epoch": 2.98, + "learning_rate": 1.0592774398980077e-05, + "loss": 0.9006, + "step": 40330 + }, + { + "epoch": 2.98, + "learning_rate": 1.0592376228285821e-05, + "loss": 1.1073, + "step": 40331 + }, + { + "epoch": 2.98, + "learning_rate": 1.0591978056649102e-05, + "loss": 0.9511, + "step": 40332 + }, + { + "epoch": 2.98, + "learning_rate": 1.0591579884070555e-05, + "loss": 1.0818, + "step": 40333 + }, + { + "epoch": 2.98, + "learning_rate": 1.059118171055081e-05, + "loss": 1.0213, + "step": 40334 + }, + { + "epoch": 2.98, + "learning_rate": 1.0590783536090507e-05, + "loss": 0.8791, + "step": 40335 + }, + { + "epoch": 2.98, + "learning_rate": 1.0590385360690276e-05, + "loss": 1.0824, + "step": 40336 + }, + { + "epoch": 2.98, + "learning_rate": 1.0589987184350752e-05, + "loss": 1.0492, + "step": 40337 + }, + { + "epoch": 2.98, + "learning_rate": 1.0589589007072563e-05, + "loss": 1.0593, + "step": 40338 + }, + { + "epoch": 2.98, + "learning_rate": 1.0589190828856346e-05, + "loss": 1.0741, + "step": 40339 + }, + { + "epoch": 2.98, + "learning_rate": 1.0588792649702735e-05, + "loss": 0.8403, + "step": 40340 + }, + { + "epoch": 2.98, + "learning_rate": 1.0588394469612366e-05, + "loss": 0.9334, + "step": 40341 + }, + { + "epoch": 2.98, + "learning_rate": 1.0587996288585867e-05, + "loss": 1.0219, + "step": 40342 + }, + { + "epoch": 2.98, + "learning_rate": 1.0587598106623877e-05, + "loss": 0.9972, + "step": 40343 + }, + { + "epoch": 2.98, + "learning_rate": 1.0587199923727023e-05, + "loss": 0.9919, + "step": 40344 + }, + { + "epoch": 2.98, + "learning_rate": 1.0586801739895949e-05, + "loss": 1.0043, + "step": 40345 + }, + { + "epoch": 2.98, + "learning_rate": 1.0586403555131274e-05, + "loss": 0.9691, + "step": 40346 + }, + { + "epoch": 2.98, + "learning_rate": 1.0586005369433646e-05, + "loss": 0.9545, + "step": 40347 + }, + { + "epoch": 2.98, + "learning_rate": 1.0585607182803688e-05, + "loss": 1.0336, + "step": 40348 + }, + { + "epoch": 2.98, + "learning_rate": 1.0585208995242038e-05, + "loss": 1.0331, + "step": 40349 + }, + { + "epoch": 2.98, + "learning_rate": 1.0584810806749329e-05, + "loss": 1.0735, + "step": 40350 + }, + { + "epoch": 2.98, + "learning_rate": 1.0584412617326193e-05, + "loss": 1.0068, + "step": 40351 + }, + { + "epoch": 2.98, + "learning_rate": 1.0584014426973267e-05, + "loss": 0.9748, + "step": 40352 + }, + { + "epoch": 2.98, + "learning_rate": 1.0583616235691182e-05, + "loss": 0.9533, + "step": 40353 + }, + { + "epoch": 2.98, + "learning_rate": 1.058321804348057e-05, + "loss": 1.0244, + "step": 40354 + }, + { + "epoch": 2.98, + "learning_rate": 1.058281985034207e-05, + "loss": 1.1115, + "step": 40355 + }, + { + "epoch": 2.98, + "learning_rate": 1.058242165627631e-05, + "loss": 1.0103, + "step": 40356 + }, + { + "epoch": 2.98, + "learning_rate": 1.0582023461283925e-05, + "loss": 0.9909, + "step": 40357 + }, + { + "epoch": 2.98, + "learning_rate": 1.0581625265365546e-05, + "loss": 0.8972, + "step": 40358 + }, + { + "epoch": 2.98, + "learning_rate": 1.0581227068521816e-05, + "loss": 0.9814, + "step": 40359 + }, + { + "epoch": 2.98, + "learning_rate": 1.0580828870753356e-05, + "loss": 0.9673, + "step": 40360 + }, + { + "epoch": 2.98, + "learning_rate": 1.0580430672060811e-05, + "loss": 1.0059, + "step": 40361 + }, + { + "epoch": 2.98, + "learning_rate": 1.0580032472444806e-05, + "loss": 1.0391, + "step": 40362 + }, + { + "epoch": 2.98, + "learning_rate": 1.0579634271905978e-05, + "loss": 0.9111, + "step": 40363 + }, + { + "epoch": 2.98, + "learning_rate": 1.0579236070444963e-05, + "loss": 1.0242, + "step": 40364 + }, + { + "epoch": 2.98, + "learning_rate": 1.0578837868062386e-05, + "loss": 1.0247, + "step": 40365 + }, + { + "epoch": 2.98, + "learning_rate": 1.057843966475889e-05, + "loss": 0.9303, + "step": 40366 + }, + { + "epoch": 2.98, + "learning_rate": 1.0578041460535105e-05, + "loss": 0.9798, + "step": 40367 + }, + { + "epoch": 2.98, + "learning_rate": 1.0577643255391663e-05, + "loss": 1.15, + "step": 40368 + }, + { + "epoch": 2.98, + "learning_rate": 1.05772450493292e-05, + "loss": 0.9687, + "step": 40369 + }, + { + "epoch": 2.98, + "learning_rate": 1.0576846842348348e-05, + "loss": 0.9803, + "step": 40370 + }, + { + "epoch": 2.98, + "learning_rate": 1.0576448634449742e-05, + "loss": 0.9443, + "step": 40371 + }, + { + "epoch": 2.98, + "learning_rate": 1.057605042563401e-05, + "loss": 0.9234, + "step": 40372 + }, + { + "epoch": 2.98, + "learning_rate": 1.0575652215901795e-05, + "loss": 1.1127, + "step": 40373 + }, + { + "epoch": 2.98, + "learning_rate": 1.0575254005253723e-05, + "loss": 1.0663, + "step": 40374 + }, + { + "epoch": 2.98, + "learning_rate": 1.057485579369043e-05, + "loss": 1.0037, + "step": 40375 + }, + { + "epoch": 2.98, + "learning_rate": 1.0574457581212552e-05, + "loss": 0.9451, + "step": 40376 + }, + { + "epoch": 2.98, + "learning_rate": 1.0574059367820717e-05, + "loss": 0.9785, + "step": 40377 + }, + { + "epoch": 2.98, + "learning_rate": 1.0573661153515564e-05, + "loss": 1.0627, + "step": 40378 + }, + { + "epoch": 2.98, + "learning_rate": 1.0573262938297722e-05, + "loss": 0.92, + "step": 40379 + }, + { + "epoch": 2.98, + "learning_rate": 1.0572864722167831e-05, + "loss": 0.9646, + "step": 40380 + }, + { + "epoch": 2.98, + "learning_rate": 1.0572466505126515e-05, + "loss": 1.0238, + "step": 40381 + }, + { + "epoch": 2.98, + "learning_rate": 1.0572068287174417e-05, + "loss": 0.9453, + "step": 40382 + }, + { + "epoch": 2.98, + "learning_rate": 1.0571670068312167e-05, + "loss": 1.0325, + "step": 40383 + }, + { + "epoch": 2.98, + "learning_rate": 1.0571271848540397e-05, + "loss": 0.979, + "step": 40384 + }, + { + "epoch": 2.98, + "learning_rate": 1.0570873627859738e-05, + "loss": 1.1159, + "step": 40385 + }, + { + "epoch": 2.98, + "learning_rate": 1.0570475406270832e-05, + "loss": 1.0202, + "step": 40386 + }, + { + "epoch": 2.98, + "learning_rate": 1.0570077183774306e-05, + "loss": 0.9127, + "step": 40387 + }, + { + "epoch": 2.98, + "learning_rate": 1.0569678960370797e-05, + "loss": 0.882, + "step": 40388 + }, + { + "epoch": 2.98, + "learning_rate": 1.0569280736060934e-05, + "loss": 0.9837, + "step": 40389 + }, + { + "epoch": 2.98, + "learning_rate": 1.0568882510845356e-05, + "loss": 0.9376, + "step": 40390 + }, + { + "epoch": 2.98, + "learning_rate": 1.056848428472469e-05, + "loss": 0.9213, + "step": 40391 + }, + { + "epoch": 2.98, + "learning_rate": 1.0568086057699583e-05, + "loss": 0.8728, + "step": 40392 + }, + { + "epoch": 2.98, + "learning_rate": 1.056768782977065e-05, + "loss": 0.9773, + "step": 40393 + }, + { + "epoch": 2.98, + "learning_rate": 1.0567289600938537e-05, + "loss": 0.9991, + "step": 40394 + }, + { + "epoch": 2.98, + "learning_rate": 1.0566891371203874e-05, + "loss": 0.9389, + "step": 40395 + }, + { + "epoch": 2.98, + "learning_rate": 1.0566493140567295e-05, + "loss": 0.9465, + "step": 40396 + }, + { + "epoch": 2.98, + "learning_rate": 1.0566094909029434e-05, + "loss": 1.0876, + "step": 40397 + }, + { + "epoch": 2.98, + "learning_rate": 1.0565696676590925e-05, + "loss": 0.9672, + "step": 40398 + }, + { + "epoch": 2.99, + "learning_rate": 1.0565298443252397e-05, + "loss": 1.0218, + "step": 40399 + }, + { + "epoch": 2.99, + "learning_rate": 1.056490020901449e-05, + "loss": 1.0788, + "step": 40400 + }, + { + "epoch": 2.99, + "learning_rate": 1.0564501973877837e-05, + "loss": 1.1064, + "step": 40401 + }, + { + "epoch": 2.99, + "learning_rate": 1.0564103737843068e-05, + "loss": 1.0089, + "step": 40402 + }, + { + "epoch": 2.99, + "learning_rate": 1.0563705500910815e-05, + "loss": 1.1538, + "step": 40403 + }, + { + "epoch": 2.99, + "learning_rate": 1.0563307263081717e-05, + "loss": 0.9479, + "step": 40404 + }, + { + "epoch": 2.99, + "learning_rate": 1.0562909024356403e-05, + "loss": 0.9289, + "step": 40405 + }, + { + "epoch": 2.99, + "learning_rate": 1.0562510784735512e-05, + "loss": 1.088, + "step": 40406 + }, + { + "epoch": 2.99, + "learning_rate": 1.0562112544219671e-05, + "loss": 1.0759, + "step": 40407 + }, + { + "epoch": 2.99, + "learning_rate": 1.0561714302809521e-05, + "loss": 1.1241, + "step": 40408 + }, + { + "epoch": 2.99, + "learning_rate": 1.0561316060505687e-05, + "loss": 1.0565, + "step": 40409 + }, + { + "epoch": 2.99, + "learning_rate": 1.0560917817308812e-05, + "loss": 1.0305, + "step": 40410 + }, + { + "epoch": 2.99, + "learning_rate": 1.0560519573219524e-05, + "loss": 0.8547, + "step": 40411 + }, + { + "epoch": 2.99, + "learning_rate": 1.0560121328238455e-05, + "loss": 1.0856, + "step": 40412 + }, + { + "epoch": 2.99, + "learning_rate": 1.055972308236624e-05, + "loss": 0.9837, + "step": 40413 + }, + { + "epoch": 2.99, + "learning_rate": 1.0559324835603518e-05, + "loss": 1.0615, + "step": 40414 + }, + { + "epoch": 2.99, + "learning_rate": 1.0558926587950913e-05, + "loss": 0.9372, + "step": 40415 + }, + { + "epoch": 2.99, + "learning_rate": 1.0558528339409067e-05, + "loss": 1.086, + "step": 40416 + }, + { + "epoch": 2.99, + "learning_rate": 1.055813008997861e-05, + "loss": 1.0088, + "step": 40417 + }, + { + "epoch": 2.99, + "learning_rate": 1.0557731839660175e-05, + "loss": 0.9365, + "step": 40418 + }, + { + "epoch": 2.99, + "learning_rate": 1.0557333588454397e-05, + "loss": 0.9364, + "step": 40419 + }, + { + "epoch": 2.99, + "learning_rate": 1.0556935336361911e-05, + "loss": 0.9997, + "step": 40420 + }, + { + "epoch": 2.99, + "learning_rate": 1.0556537083383347e-05, + "loss": 1.0136, + "step": 40421 + }, + { + "epoch": 2.99, + "learning_rate": 1.055613882951934e-05, + "loss": 0.9919, + "step": 40422 + }, + { + "epoch": 2.99, + "learning_rate": 1.0555740574770526e-05, + "loss": 1.0758, + "step": 40423 + }, + { + "epoch": 2.99, + "learning_rate": 1.0555342319137534e-05, + "loss": 0.9911, + "step": 40424 + }, + { + "epoch": 2.99, + "learning_rate": 1.0554944062621005e-05, + "loss": 1.0518, + "step": 40425 + }, + { + "epoch": 2.99, + "learning_rate": 1.0554545805221563e-05, + "loss": 1.0349, + "step": 40426 + }, + { + "epoch": 2.99, + "learning_rate": 1.0554147546939848e-05, + "loss": 1.033, + "step": 40427 + }, + { + "epoch": 2.99, + "learning_rate": 1.0553749287776494e-05, + "loss": 1.145, + "step": 40428 + }, + { + "epoch": 2.99, + "learning_rate": 1.0553351027732132e-05, + "loss": 0.9593, + "step": 40429 + }, + { + "epoch": 2.99, + "learning_rate": 1.0552952766807396e-05, + "loss": 0.9625, + "step": 40430 + }, + { + "epoch": 2.99, + "learning_rate": 1.055255450500292e-05, + "loss": 1.0059, + "step": 40431 + }, + { + "epoch": 2.99, + "learning_rate": 1.0552156242319336e-05, + "loss": 1.0424, + "step": 40432 + }, + { + "epoch": 2.99, + "learning_rate": 1.0551757978757282e-05, + "loss": 1.0578, + "step": 40433 + }, + { + "epoch": 2.99, + "learning_rate": 1.0551359714317389e-05, + "loss": 1.005, + "step": 40434 + }, + { + "epoch": 2.99, + "learning_rate": 1.055096144900029e-05, + "loss": 1.0894, + "step": 40435 + }, + { + "epoch": 2.99, + "learning_rate": 1.0550563182806618e-05, + "loss": 1.0884, + "step": 40436 + }, + { + "epoch": 2.99, + "learning_rate": 1.0550164915737009e-05, + "loss": 0.9227, + "step": 40437 + }, + { + "epoch": 2.99, + "learning_rate": 1.0549766647792093e-05, + "loss": 1.0027, + "step": 40438 + }, + { + "epoch": 2.99, + "learning_rate": 1.054936837897251e-05, + "loss": 1.0229, + "step": 40439 + }, + { + "epoch": 2.99, + "learning_rate": 1.0548970109278888e-05, + "loss": 1.0652, + "step": 40440 + }, + { + "epoch": 2.99, + "learning_rate": 1.0548571838711863e-05, + "loss": 0.9034, + "step": 40441 + }, + { + "epoch": 2.99, + "learning_rate": 1.0548173567272066e-05, + "loss": 1.001, + "step": 40442 + }, + { + "epoch": 2.99, + "learning_rate": 1.0547775294960136e-05, + "loss": 0.9881, + "step": 40443 + }, + { + "epoch": 2.99, + "learning_rate": 1.0547377021776699e-05, + "loss": 1.0299, + "step": 40444 + }, + { + "epoch": 2.99, + "learning_rate": 1.0546978747722396e-05, + "loss": 0.8686, + "step": 40445 + }, + { + "epoch": 2.99, + "learning_rate": 1.0546580472797855e-05, + "loss": 0.97, + "step": 40446 + }, + { + "epoch": 2.99, + "learning_rate": 1.0546182197003719e-05, + "loss": 0.9951, + "step": 40447 + }, + { + "epoch": 2.99, + "learning_rate": 1.0545783920340608e-05, + "loss": 0.9976, + "step": 40448 + }, + { + "epoch": 2.99, + "learning_rate": 1.0545385642809166e-05, + "loss": 0.9758, + "step": 40449 + }, + { + "epoch": 2.99, + "learning_rate": 1.0544987364410023e-05, + "loss": 0.9778, + "step": 40450 + }, + { + "epoch": 2.99, + "learning_rate": 1.0544589085143813e-05, + "loss": 1.0112, + "step": 40451 + }, + { + "epoch": 2.99, + "learning_rate": 1.0544190805011167e-05, + "loss": 0.9633, + "step": 40452 + }, + { + "epoch": 2.99, + "learning_rate": 1.0543792524012723e-05, + "loss": 0.9286, + "step": 40453 + }, + { + "epoch": 2.99, + "learning_rate": 1.0543394242149111e-05, + "loss": 1.0363, + "step": 40454 + }, + { + "epoch": 2.99, + "learning_rate": 1.054299595942097e-05, + "loss": 0.9923, + "step": 40455 + }, + { + "epoch": 2.99, + "learning_rate": 1.0542597675828927e-05, + "loss": 1.1379, + "step": 40456 + }, + { + "epoch": 2.99, + "learning_rate": 1.0542199391373624e-05, + "loss": 1.1029, + "step": 40457 + }, + { + "epoch": 2.99, + "learning_rate": 1.0541801106055685e-05, + "loss": 1.0442, + "step": 40458 + }, + { + "epoch": 2.99, + "learning_rate": 1.054140281987575e-05, + "loss": 0.8957, + "step": 40459 + }, + { + "epoch": 2.99, + "learning_rate": 1.054100453283445e-05, + "loss": 1.0656, + "step": 40460 + }, + { + "epoch": 2.99, + "learning_rate": 1.054060624493242e-05, + "loss": 1.0491, + "step": 40461 + }, + { + "epoch": 2.99, + "learning_rate": 1.054020795617029e-05, + "loss": 0.8614, + "step": 40462 + }, + { + "epoch": 2.99, + "learning_rate": 1.0539809666548702e-05, + "loss": 1.0488, + "step": 40463 + }, + { + "epoch": 2.99, + "learning_rate": 1.0539411376068281e-05, + "loss": 1.0275, + "step": 40464 + }, + { + "epoch": 2.99, + "learning_rate": 1.0539013084729665e-05, + "loss": 0.9981, + "step": 40465 + }, + { + "epoch": 2.99, + "learning_rate": 1.0538614792533486e-05, + "loss": 1.0981, + "step": 40466 + }, + { + "epoch": 2.99, + "learning_rate": 1.0538216499480381e-05, + "loss": 1.0702, + "step": 40467 + }, + { + "epoch": 2.99, + "learning_rate": 1.053781820557098e-05, + "loss": 0.9287, + "step": 40468 + }, + { + "epoch": 2.99, + "learning_rate": 1.0537419910805918e-05, + "loss": 1.0651, + "step": 40469 + }, + { + "epoch": 2.99, + "learning_rate": 1.0537021615185827e-05, + "loss": 1.0924, + "step": 40470 + }, + { + "epoch": 2.99, + "learning_rate": 1.0536623318711343e-05, + "loss": 0.9657, + "step": 40471 + }, + { + "epoch": 2.99, + "learning_rate": 1.0536225021383102e-05, + "loss": 1.0585, + "step": 40472 + }, + { + "epoch": 2.99, + "learning_rate": 1.0535826723201732e-05, + "loss": 0.9529, + "step": 40473 + }, + { + "epoch": 2.99, + "learning_rate": 1.0535428424167868e-05, + "loss": 1.037, + "step": 40474 + }, + { + "epoch": 2.99, + "learning_rate": 1.0535030124282148e-05, + "loss": 1.0941, + "step": 40475 + }, + { + "epoch": 2.99, + "learning_rate": 1.0534631823545204e-05, + "loss": 1.0036, + "step": 40476 + }, + { + "epoch": 2.99, + "learning_rate": 1.0534233521957664e-05, + "loss": 0.9726, + "step": 40477 + }, + { + "epoch": 2.99, + "learning_rate": 1.053383521952017e-05, + "loss": 1.072, + "step": 40478 + }, + { + "epoch": 2.99, + "learning_rate": 1.0533436916233349e-05, + "loss": 0.9389, + "step": 40479 + }, + { + "epoch": 2.99, + "learning_rate": 1.0533038612097839e-05, + "loss": 1.0861, + "step": 40480 + }, + { + "epoch": 2.99, + "learning_rate": 1.0532640307114271e-05, + "loss": 1.0781, + "step": 40481 + }, + { + "epoch": 2.99, + "learning_rate": 1.0532242001283279e-05, + "loss": 0.9638, + "step": 40482 + }, + { + "epoch": 2.99, + "learning_rate": 1.05318436946055e-05, + "loss": 1.0105, + "step": 40483 + }, + { + "epoch": 2.99, + "learning_rate": 1.0531445387081565e-05, + "loss": 0.9593, + "step": 40484 + }, + { + "epoch": 2.99, + "learning_rate": 1.0531047078712109e-05, + "loss": 1.0208, + "step": 40485 + }, + { + "epoch": 2.99, + "learning_rate": 1.0530648769497762e-05, + "loss": 1.0273, + "step": 40486 + }, + { + "epoch": 2.99, + "learning_rate": 1.0530250459439163e-05, + "loss": 1.049, + "step": 40487 + }, + { + "epoch": 2.99, + "learning_rate": 1.052985214853694e-05, + "loss": 0.9384, + "step": 40488 + }, + { + "epoch": 2.99, + "learning_rate": 1.052945383679173e-05, + "loss": 0.9825, + "step": 40489 + }, + { + "epoch": 2.99, + "learning_rate": 1.052905552420417e-05, + "loss": 0.9829, + "step": 40490 + }, + { + "epoch": 2.99, + "learning_rate": 1.0528657210774886e-05, + "loss": 0.9107, + "step": 40491 + }, + { + "epoch": 2.99, + "learning_rate": 1.052825889650452e-05, + "loss": 0.932, + "step": 40492 + }, + { + "epoch": 2.99, + "learning_rate": 1.0527860581393698e-05, + "loss": 0.9169, + "step": 40493 + }, + { + "epoch": 2.99, + "learning_rate": 1.052746226544306e-05, + "loss": 0.9035, + "step": 40494 + }, + { + "epoch": 2.99, + "learning_rate": 1.0527063948653238e-05, + "loss": 0.8556, + "step": 40495 + }, + { + "epoch": 2.99, + "learning_rate": 1.0526665631024865e-05, + "loss": 0.9999, + "step": 40496 + }, + { + "epoch": 2.99, + "learning_rate": 1.0526267312558569e-05, + "loss": 1.0541, + "step": 40497 + }, + { + "epoch": 2.99, + "learning_rate": 1.0525868993254992e-05, + "loss": 0.8923, + "step": 40498 + }, + { + "epoch": 2.99, + "learning_rate": 1.0525470673114766e-05, + "loss": 0.9048, + "step": 40499 + }, + { + "epoch": 2.99, + "learning_rate": 1.0525072352138525e-05, + "loss": 1.0253, + "step": 40500 + }, + { + "epoch": 2.99, + "learning_rate": 1.0524674030326898e-05, + "loss": 0.8672, + "step": 40501 + }, + { + "epoch": 2.99, + "learning_rate": 1.0524275707680525e-05, + "loss": 1.1477, + "step": 40502 + }, + { + "epoch": 2.99, + "learning_rate": 1.0523877384200035e-05, + "loss": 0.984, + "step": 40503 + }, + { + "epoch": 2.99, + "learning_rate": 1.0523479059886066e-05, + "loss": 0.9793, + "step": 40504 + }, + { + "epoch": 2.99, + "learning_rate": 1.0523080734739248e-05, + "loss": 1.017, + "step": 40505 + }, + { + "epoch": 2.99, + "learning_rate": 1.0522682408760217e-05, + "loss": 0.9921, + "step": 40506 + }, + { + "epoch": 2.99, + "learning_rate": 1.0522284081949603e-05, + "loss": 1.0736, + "step": 40507 + }, + { + "epoch": 2.99, + "learning_rate": 1.0521885754308043e-05, + "loss": 0.9854, + "step": 40508 + }, + { + "epoch": 2.99, + "learning_rate": 1.052148742583617e-05, + "loss": 1.1321, + "step": 40509 + }, + { + "epoch": 2.99, + "learning_rate": 1.052108909653462e-05, + "loss": 0.9748, + "step": 40510 + }, + { + "epoch": 2.99, + "learning_rate": 1.0520690766404023e-05, + "loss": 1.0205, + "step": 40511 + }, + { + "epoch": 2.99, + "learning_rate": 1.0520292435445016e-05, + "loss": 0.8735, + "step": 40512 + }, + { + "epoch": 2.99, + "learning_rate": 1.051989410365823e-05, + "loss": 1.0171, + "step": 40513 + }, + { + "epoch": 2.99, + "learning_rate": 1.0519495771044301e-05, + "loss": 0.9882, + "step": 40514 + }, + { + "epoch": 2.99, + "learning_rate": 1.0519097437603862e-05, + "loss": 1.0232, + "step": 40515 + }, + { + "epoch": 2.99, + "learning_rate": 1.0518699103337544e-05, + "loss": 1.0529, + "step": 40516 + }, + { + "epoch": 2.99, + "learning_rate": 1.0518300768245985e-05, + "loss": 0.9437, + "step": 40517 + }, + { + "epoch": 2.99, + "learning_rate": 1.0517902432329818e-05, + "loss": 1.1132, + "step": 40518 + }, + { + "epoch": 2.99, + "learning_rate": 1.0517504095589672e-05, + "loss": 0.9948, + "step": 40519 + }, + { + "epoch": 2.99, + "learning_rate": 1.051710575802619e-05, + "loss": 0.9931, + "step": 40520 + }, + { + "epoch": 2.99, + "learning_rate": 1.0516707419639993e-05, + "loss": 0.8524, + "step": 40521 + }, + { + "epoch": 2.99, + "learning_rate": 1.0516309080431729e-05, + "loss": 1.0568, + "step": 40522 + }, + { + "epoch": 2.99, + "learning_rate": 1.051591074040202e-05, + "loss": 1.0107, + "step": 40523 + }, + { + "epoch": 2.99, + "learning_rate": 1.0515512399551503e-05, + "loss": 0.9327, + "step": 40524 + }, + { + "epoch": 2.99, + "learning_rate": 1.0515114057880818e-05, + "loss": 0.9721, + "step": 40525 + }, + { + "epoch": 2.99, + "learning_rate": 1.051471571539059e-05, + "loss": 1.0391, + "step": 40526 + }, + { + "epoch": 2.99, + "learning_rate": 1.051431737208146e-05, + "loss": 1.0709, + "step": 40527 + }, + { + "epoch": 2.99, + "learning_rate": 1.0513919027954055e-05, + "loss": 0.9427, + "step": 40528 + }, + { + "epoch": 2.99, + "learning_rate": 1.0513520683009015e-05, + "loss": 0.9714, + "step": 40529 + }, + { + "epoch": 2.99, + "learning_rate": 1.0513122337246969e-05, + "loss": 1.0534, + "step": 40530 + }, + { + "epoch": 2.99, + "learning_rate": 1.0512723990668554e-05, + "loss": 1.0635, + "step": 40531 + }, + { + "epoch": 2.99, + "learning_rate": 1.0512325643274404e-05, + "loss": 0.993, + "step": 40532 + }, + { + "epoch": 2.99, + "learning_rate": 1.051192729506515e-05, + "loss": 0.9282, + "step": 40533 + }, + { + "epoch": 3.0, + "learning_rate": 1.0511528946041425e-05, + "loss": 0.9834, + "step": 40534 + }, + { + "epoch": 3.0, + "learning_rate": 1.0511130596203866e-05, + "loss": 0.9642, + "step": 40535 + }, + { + "epoch": 3.0, + "learning_rate": 1.0510732245553103e-05, + "loss": 0.9612, + "step": 40536 + }, + { + "epoch": 3.0, + "learning_rate": 1.0510333894089776e-05, + "loss": 1.0529, + "step": 40537 + }, + { + "epoch": 3.0, + "learning_rate": 1.0509935541814513e-05, + "loss": 0.9139, + "step": 40538 + }, + { + "epoch": 3.0, + "learning_rate": 1.0509537188727953e-05, + "loss": 0.9841, + "step": 40539 + }, + { + "epoch": 3.0, + "learning_rate": 1.0509138834830724e-05, + "loss": 1.018, + "step": 40540 + }, + { + "epoch": 3.0, + "learning_rate": 1.0508740480123463e-05, + "loss": 1.0454, + "step": 40541 + }, + { + "epoch": 3.0, + "learning_rate": 1.0508342124606802e-05, + "loss": 0.9755, + "step": 40542 + }, + { + "epoch": 3.0, + "learning_rate": 1.050794376828138e-05, + "loss": 1.0027, + "step": 40543 + }, + { + "epoch": 3.0, + "learning_rate": 1.0507545411147821e-05, + "loss": 0.9647, + "step": 40544 + }, + { + "epoch": 3.0, + "learning_rate": 1.0507147053206768e-05, + "loss": 1.0249, + "step": 40545 + }, + { + "epoch": 3.0, + "learning_rate": 1.050674869445885e-05, + "loss": 1.0759, + "step": 40546 + }, + { + "epoch": 3.0, + "learning_rate": 1.0506350334904703e-05, + "loss": 1.006, + "step": 40547 + }, + { + "epoch": 3.0, + "learning_rate": 1.0505951974544958e-05, + "loss": 0.9447, + "step": 40548 + }, + { + "epoch": 3.0, + "learning_rate": 1.0505553613380253e-05, + "loss": 1.0778, + "step": 40549 + }, + { + "epoch": 3.0, + "learning_rate": 1.0505155251411216e-05, + "loss": 1.0719, + "step": 40550 + }, + { + "epoch": 3.0, + "learning_rate": 1.0504756888638492e-05, + "loss": 0.9364, + "step": 40551 + }, + { + "epoch": 3.0, + "learning_rate": 1.0504358525062699e-05, + "loss": 0.9569, + "step": 40552 + }, + { + "epoch": 3.0, + "learning_rate": 1.0503960160684483e-05, + "loss": 0.8691, + "step": 40553 + }, + { + "epoch": 3.0, + "learning_rate": 1.0503561795504471e-05, + "loss": 1.1082, + "step": 40554 + }, + { + "epoch": 3.0, + "learning_rate": 1.0503163429523303e-05, + "loss": 0.9183, + "step": 40555 + }, + { + "epoch": 3.0, + "learning_rate": 1.0502765062741604e-05, + "loss": 1.0354, + "step": 40556 + }, + { + "epoch": 3.0, + "learning_rate": 1.0502366695160016e-05, + "loss": 0.9884, + "step": 40557 + }, + { + "epoch": 3.0, + "learning_rate": 1.0501968326779168e-05, + "loss": 1.1404, + "step": 40558 + }, + { + "epoch": 3.0, + "learning_rate": 1.0501569957599698e-05, + "loss": 1.026, + "step": 40559 + }, + { + "epoch": 3.0, + "learning_rate": 1.0501171587622237e-05, + "loss": 1.072, + "step": 40560 + }, + { + "epoch": 3.0, + "learning_rate": 1.050077321684742e-05, + "loss": 0.951, + "step": 40561 + }, + { + "epoch": 3.0, + "learning_rate": 1.0500374845275878e-05, + "loss": 0.9928, + "step": 40562 + }, + { + "epoch": 3.0, + "learning_rate": 1.0499976472908246e-05, + "loss": 0.895, + "step": 40563 + }, + { + "epoch": 3.0, + "learning_rate": 1.049957809974516e-05, + "loss": 1.0904, + "step": 40564 + }, + { + "epoch": 3.0, + "learning_rate": 1.0499179725787254e-05, + "loss": 0.9081, + "step": 40565 + }, + { + "epoch": 3.0, + "learning_rate": 1.0498781351035156e-05, + "loss": 1.0283, + "step": 40566 + }, + { + "epoch": 3.0, + "learning_rate": 1.0498382975489507e-05, + "loss": 0.991, + "step": 40567 + }, + { + "epoch": 3.0, + "learning_rate": 1.0497984599150937e-05, + "loss": 1.0031, + "step": 40568 + }, + { + "epoch": 3.0, + "learning_rate": 1.0497586222020083e-05, + "loss": 0.917, + "step": 40569 + }, + { + "epoch": 3.0, + "learning_rate": 1.0497187844097574e-05, + "loss": 0.9346, + "step": 40570 + }, + { + "epoch": 3.0, + "learning_rate": 1.0496789465384046e-05, + "loss": 1.0967, + "step": 40571 + }, + { + "epoch": 3.0, + "learning_rate": 1.0496391085880134e-05, + "loss": 1.0152, + "step": 40572 + }, + { + "epoch": 3.0, + "learning_rate": 1.0495992705586469e-05, + "loss": 0.9183, + "step": 40573 + }, + { + "epoch": 3.0, + "learning_rate": 1.0495594324503689e-05, + "loss": 0.982, + "step": 40574 + }, + { + "epoch": 3.0, + "learning_rate": 1.0495195942632423e-05, + "loss": 1.0465, + "step": 40575 + }, + { + "epoch": 3.0, + "learning_rate": 1.049479755997331e-05, + "loss": 1.0103, + "step": 40576 + }, + { + "epoch": 3.0, + "learning_rate": 1.049439917652698e-05, + "loss": 1.0329, + "step": 40577 + }, + { + "epoch": 3.0, + "learning_rate": 1.049400079229407e-05, + "loss": 1.0319, + "step": 40578 + }, + { + "epoch": 3.0, + "learning_rate": 1.049360240727521e-05, + "loss": 0.9486, + "step": 40579 + }, + { + "epoch": 3.0, + "learning_rate": 1.0493204021471037e-05, + "loss": 0.9718, + "step": 40580 + }, + { + "epoch": 3.0, + "learning_rate": 1.0492805634882181e-05, + "loss": 1.1483, + "step": 40581 + }, + { + "epoch": 3.0, + "learning_rate": 1.049240724750928e-05, + "loss": 1.0244, + "step": 40582 + }, + { + "epoch": 3.0, + "learning_rate": 1.0492008859352964e-05, + "loss": 1.0615, + "step": 40583 + }, + { + "epoch": 3.0, + "learning_rate": 1.0491610470413871e-05, + "loss": 0.9639, + "step": 40584 + }, + { + "epoch": 3.0, + "learning_rate": 1.0491212080692633e-05, + "loss": 1.006, + "step": 40585 + }, + { + "epoch": 3.0, + "learning_rate": 1.0490813690189883e-05, + "loss": 1.0216, + "step": 40586 + }, + { + "epoch": 3.0, + "learning_rate": 1.0490415298906253e-05, + "loss": 1.0989, + "step": 40587 + }, + { + "epoch": 3.0, + "learning_rate": 1.0490016906842386e-05, + "loss": 0.9742, + "step": 40588 + }, + { + "epoch": 3.0, + "learning_rate": 1.0489618513998903e-05, + "loss": 1.0391, + "step": 40589 + }, + { + "epoch": 3.0, + "learning_rate": 1.0489220120376447e-05, + "loss": 0.9412, + "step": 40590 + }, + { + "epoch": 3.0, + "learning_rate": 1.0488821725975647e-05, + "loss": 1.0897, + "step": 40591 + }, + { + "epoch": 3.0, + "learning_rate": 1.048842333079714e-05, + "loss": 0.9487, + "step": 40592 + }, + { + "epoch": 3.0, + "learning_rate": 1.0488024934841554e-05, + "loss": 0.9881, + "step": 40593 + }, + { + "epoch": 3.0, + "learning_rate": 1.0487626538109532e-05, + "loss": 0.9839, + "step": 40594 + }, + { + "epoch": 3.0, + "learning_rate": 1.0487228140601701e-05, + "loss": 1.1215, + "step": 40595 + }, + { + "epoch": 3.0, + "learning_rate": 1.0486829742318701e-05, + "loss": 0.8442, + "step": 40596 + }, + { + "epoch": 3.0, + "learning_rate": 1.0486431343261157e-05, + "loss": 1.0892, + "step": 40597 + }, + { + "epoch": 3.0, + "learning_rate": 1.0486032943429711e-05, + "loss": 0.9773, + "step": 40598 + }, + { + "epoch": 3.0, + "learning_rate": 1.0485634542824992e-05, + "loss": 0.9166, + "step": 40599 + }, + { + "epoch": 3.0, + "learning_rate": 1.0485236141447637e-05, + "loss": 1.0023, + "step": 40600 + }, + { + "epoch": 3.0, + "learning_rate": 1.0484837739298274e-05, + "loss": 1.0056, + "step": 40601 + }, + { + "epoch": 3.0, + "learning_rate": 1.0484439336377545e-05, + "loss": 1.0441, + "step": 40602 + }, + { + "epoch": 3.0, + "learning_rate": 1.0484040932686077e-05, + "loss": 1.1136, + "step": 40603 + }, + { + "epoch": 3.0, + "learning_rate": 1.0483642528224512e-05, + "loss": 1.0495, + "step": 40604 + }, + { + "epoch": 3.0, + "learning_rate": 1.0483244122993472e-05, + "loss": 0.9518, + "step": 40605 + }, + { + "epoch": 3.0, + "learning_rate": 1.0482845716993602e-05, + "loss": 0.9372, + "step": 40606 + }, + { + "epoch": 3.0, + "learning_rate": 1.0482447310225532e-05, + "loss": 0.9973, + "step": 40607 + }, + { + "epoch": 3.0, + "learning_rate": 1.0482048902689893e-05, + "loss": 1.0748, + "step": 40608 + }, + { + "epoch": 3.0, + "learning_rate": 1.048165049438732e-05, + "loss": 1.0097, + "step": 40609 + }, + { + "epoch": 3.0, + "learning_rate": 1.0481252085318452e-05, + "loss": 0.9693, + "step": 40610 + }, + { + "epoch": 3.0, + "learning_rate": 1.0480853675483913e-05, + "loss": 1.0673, + "step": 40611 + }, + { + "epoch": 3.0, + "learning_rate": 1.0480455264884348e-05, + "loss": 0.9148, + "step": 40612 + }, + { + "epoch": 3.0, + "learning_rate": 1.0480056853520382e-05, + "loss": 1.0506, + "step": 40613 + }, + { + "epoch": 3.0, + "learning_rate": 1.0479658441392656e-05, + "loss": 0.9125, + "step": 40614 + }, + { + "epoch": 3.0, + "learning_rate": 1.0479260028501797e-05, + "loss": 0.9848, + "step": 40615 + }, + { + "epoch": 3.0, + "learning_rate": 1.0478861614848444e-05, + "loss": 1.0364, + "step": 40616 + }, + { + "epoch": 3.0, + "learning_rate": 1.0478463200433229e-05, + "loss": 0.9927, + "step": 40617 + }, + { + "epoch": 3.0, + "learning_rate": 1.0478064785256783e-05, + "loss": 0.9845, + "step": 40618 + }, + { + "epoch": 3.0, + "learning_rate": 1.0477666369319746e-05, + "loss": 1.0732, + "step": 40619 + }, + { + "epoch": 3.0, + "learning_rate": 1.0477267952622747e-05, + "loss": 0.9856, + "step": 40620 + }, + { + "epoch": 3.0, + "learning_rate": 1.0476869535166422e-05, + "loss": 0.9754, + "step": 40621 + }, + { + "epoch": 3.0, + "learning_rate": 1.0476471116951405e-05, + "loss": 1.0155, + "step": 40622 + }, + { + "epoch": 3.0, + "learning_rate": 1.0476072697978328e-05, + "loss": 0.9477, + "step": 40623 + }, + { + "epoch": 3.0, + "learning_rate": 1.0475674278247829e-05, + "loss": 0.9043, + "step": 40624 + }, + { + "epoch": 3.0, + "learning_rate": 1.0475275857760536e-05, + "loss": 1.0687, + "step": 40625 + }, + { + "epoch": 3.0, + "learning_rate": 1.0474877436517088e-05, + "loss": 1.161, + "step": 40626 + }, + { + "epoch": 3.0, + "learning_rate": 1.0474479014518115e-05, + "loss": 0.9146, + "step": 40627 + }, + { + "epoch": 3.0, + "learning_rate": 1.0474080591764253e-05, + "loss": 1.0778, + "step": 40628 + }, + { + "epoch": 3.0, + "learning_rate": 1.0473682168256136e-05, + "loss": 1.0997, + "step": 40629 + }, + { + "epoch": 3.0, + "learning_rate": 1.0473283743994397e-05, + "loss": 0.9617, + "step": 40630 + }, + { + "epoch": 3.0, + "learning_rate": 1.047288531897967e-05, + "loss": 1.0443, + "step": 40631 + }, + { + "epoch": 3.0, + "learning_rate": 1.047248689321259e-05, + "loss": 1.0173, + "step": 40632 + }, + { + "epoch": 3.0, + "learning_rate": 1.047208846669379e-05, + "loss": 1.0083, + "step": 40633 + }, + { + "epoch": 3.0, + "learning_rate": 1.0471690039423903e-05, + "loss": 1.0408, + "step": 40634 + }, + { + "epoch": 3.0, + "learning_rate": 1.0471291611403568e-05, + "loss": 0.9794, + "step": 40635 + }, + { + "epoch": 3.0, + "learning_rate": 1.047089318263341e-05, + "loss": 0.9586, + "step": 40636 + }, + { + "epoch": 3.0, + "learning_rate": 1.047049475311407e-05, + "loss": 1.0685, + "step": 40637 + }, + { + "epoch": 3.0, + "learning_rate": 1.047009632284618e-05, + "loss": 1.1807, + "step": 40638 + }, + { + "epoch": 3.0, + "learning_rate": 1.0469697891830373e-05, + "loss": 1.0568, + "step": 40639 + }, + { + "epoch": 3.0, + "learning_rate": 1.046929946006728e-05, + "loss": 0.9616, + "step": 40640 + }, + { + "epoch": 3.0, + "learning_rate": 1.0468901027557543e-05, + "loss": 0.9962, + "step": 40641 + }, + { + "epoch": 3.0, + "learning_rate": 1.0468502594301787e-05, + "loss": 1.0282, + "step": 40642 + }, + { + "epoch": 3.0, + "learning_rate": 1.0468104160300653e-05, + "loss": 0.9796, + "step": 40643 + }, + { + "epoch": 3.0, + "learning_rate": 1.0467705725554774e-05, + "loss": 0.9757, + "step": 40644 + }, + { + "epoch": 3.0, + "learning_rate": 1.046730729006478e-05, + "loss": 0.9224, + "step": 40645 + }, + { + "epoch": 3.0, + "learning_rate": 1.0466908853831306e-05, + "loss": 1.1144, + "step": 40646 + }, + { + "epoch": 3.0, + "learning_rate": 1.0466510416854988e-05, + "loss": 1.0487, + "step": 40647 + }, + { + "epoch": 3.0, + "learning_rate": 1.0466111979136454e-05, + "loss": 0.9263, + "step": 40648 + }, + { + "epoch": 3.0, + "learning_rate": 1.0465713540676347e-05, + "loss": 1.0842, + "step": 40649 + }, + { + "epoch": 3.0, + "learning_rate": 1.0465315101475295e-05, + "loss": 1.0509, + "step": 40650 + }, + { + "epoch": 3.0, + "learning_rate": 1.0464916661533935e-05, + "loss": 1.1172, + "step": 40651 + }, + { + "epoch": 3.0, + "learning_rate": 1.0464518220852894e-05, + "loss": 1.0147, + "step": 40652 + }, + { + "epoch": 3.0, + "learning_rate": 1.0464119779432817e-05, + "loss": 0.8726, + "step": 40653 + }, + { + "epoch": 3.0, + "learning_rate": 1.046372133727433e-05, + "loss": 0.9345, + "step": 40654 + }, + { + "epoch": 3.0, + "learning_rate": 1.0463322894378071e-05, + "loss": 1.0617, + "step": 40655 + }, + { + "epoch": 3.0, + "learning_rate": 1.046292445074467e-05, + "loss": 0.9495, + "step": 40656 + }, + { + "epoch": 3.0, + "learning_rate": 1.0462526006374763e-05, + "loss": 1.0344, + "step": 40657 + }, + { + "epoch": 3.0, + "learning_rate": 1.0462127561268983e-05, + "loss": 1.027, + "step": 40658 + }, + { + "epoch": 3.0, + "learning_rate": 1.0461729115427966e-05, + "loss": 0.8615, + "step": 40659 + }, + { + "epoch": 3.0, + "learning_rate": 1.0461330668852341e-05, + "loss": 1.07, + "step": 40660 + }, + { + "epoch": 3.0, + "learning_rate": 1.046093222154275e-05, + "loss": 1.0395, + "step": 40661 + }, + { + "epoch": 3.0, + "learning_rate": 1.046053377349982e-05, + "loss": 0.9589, + "step": 40662 + }, + { + "epoch": 3.0, + "learning_rate": 1.0460135324724188e-05, + "loss": 0.9235, + "step": 40663 + }, + { + "epoch": 3.0, + "learning_rate": 1.0459736875216489e-05, + "loss": 0.9551, + "step": 40664 + }, + { + "epoch": 3.0, + "learning_rate": 1.0459338424977352e-05, + "loss": 1.1035, + "step": 40665 + }, + { + "epoch": 3.0, + "learning_rate": 1.0458939974007415e-05, + "loss": 0.9433, + "step": 40666 + }, + { + "epoch": 3.0, + "learning_rate": 1.0458541522307316e-05, + "loss": 1.0315, + "step": 40667 + }, + { + "epoch": 3.0, + "learning_rate": 1.0458143069877675e-05, + "loss": 0.9877, + "step": 40668 + }, + { + "epoch": 3.01, + "learning_rate": 1.0457744616719141e-05, + "loss": 0.958, + "step": 40669 + }, + { + "epoch": 3.01, + "learning_rate": 1.045734616283234e-05, + "loss": 1.1214, + "step": 40670 + }, + { + "epoch": 3.01, + "learning_rate": 1.045694770821791e-05, + "loss": 0.9934, + "step": 40671 + }, + { + "epoch": 3.01, + "learning_rate": 1.0456549252876482e-05, + "loss": 1.047, + "step": 40672 + }, + { + "epoch": 3.01, + "learning_rate": 1.045615079680869e-05, + "loss": 1.0279, + "step": 40673 + }, + { + "epoch": 3.01, + "learning_rate": 1.0455752340015166e-05, + "loss": 1.0084, + "step": 40674 + }, + { + "epoch": 3.01, + "learning_rate": 1.0455353882496548e-05, + "loss": 0.9922, + "step": 40675 + }, + { + "epoch": 3.01, + "learning_rate": 1.045495542425347e-05, + "loss": 0.942, + "step": 40676 + }, + { + "epoch": 3.01, + "learning_rate": 1.0454556965286564e-05, + "loss": 0.9923, + "step": 40677 + }, + { + "epoch": 3.01, + "learning_rate": 1.0454158505596466e-05, + "loss": 0.9231, + "step": 40678 + }, + { + "epoch": 3.01, + "learning_rate": 1.0453760045183806e-05, + "loss": 0.928, + "step": 40679 + }, + { + "epoch": 3.01, + "learning_rate": 1.045336158404922e-05, + "loss": 1.1326, + "step": 40680 + }, + { + "epoch": 3.01, + "learning_rate": 1.045296312219334e-05, + "loss": 1.0568, + "step": 40681 + }, + { + "epoch": 3.01, + "learning_rate": 1.0452564659616811e-05, + "loss": 1.0282, + "step": 40682 + }, + { + "epoch": 3.01, + "learning_rate": 1.045216619632025e-05, + "loss": 1.0703, + "step": 40683 + }, + { + "epoch": 3.01, + "learning_rate": 1.0451767732304302e-05, + "loss": 1.0345, + "step": 40684 + }, + { + "epoch": 3.01, + "learning_rate": 1.0451369267569595e-05, + "loss": 0.977, + "step": 40685 + }, + { + "epoch": 3.01, + "learning_rate": 1.045097080211677e-05, + "loss": 0.9354, + "step": 40686 + }, + { + "epoch": 3.01, + "learning_rate": 1.0450572335946453e-05, + "loss": 1.0233, + "step": 40687 + }, + { + "epoch": 3.01, + "learning_rate": 1.0450173869059285e-05, + "loss": 0.9484, + "step": 40688 + }, + { + "epoch": 3.01, + "learning_rate": 1.0449775401455895e-05, + "loss": 0.9585, + "step": 40689 + }, + { + "epoch": 3.01, + "learning_rate": 1.044937693313692e-05, + "loss": 0.9194, + "step": 40690 + }, + { + "epoch": 3.01, + "learning_rate": 1.0448978464102995e-05, + "loss": 0.9779, + "step": 40691 + }, + { + "epoch": 3.01, + "learning_rate": 1.0448579994354748e-05, + "loss": 1.007, + "step": 40692 + }, + { + "epoch": 3.01, + "learning_rate": 1.0448181523892818e-05, + "loss": 1.0187, + "step": 40693 + }, + { + "epoch": 3.01, + "learning_rate": 1.0447783052717837e-05, + "loss": 1.0192, + "step": 40694 + }, + { + "epoch": 3.01, + "learning_rate": 1.044738458083044e-05, + "loss": 0.9801, + "step": 40695 + }, + { + "epoch": 3.01, + "learning_rate": 1.0446986108231261e-05, + "loss": 0.9841, + "step": 40696 + }, + { + "epoch": 3.01, + "learning_rate": 1.044658763492093e-05, + "loss": 1.1262, + "step": 40697 + }, + { + "epoch": 3.01, + "learning_rate": 1.0446189160900088e-05, + "loss": 0.9828, + "step": 40698 + }, + { + "epoch": 3.01, + "learning_rate": 1.0445790686169363e-05, + "loss": 1.0497, + "step": 40699 + }, + { + "epoch": 3.01, + "learning_rate": 1.0445392210729396e-05, + "loss": 0.9425, + "step": 40700 + }, + { + "epoch": 3.01, + "learning_rate": 1.0444993734580814e-05, + "loss": 0.9773, + "step": 40701 + }, + { + "epoch": 3.01, + "learning_rate": 1.0444595257724255e-05, + "loss": 0.9627, + "step": 40702 + }, + { + "epoch": 3.01, + "learning_rate": 1.0444196780160348e-05, + "loss": 0.9588, + "step": 40703 + }, + { + "epoch": 3.01, + "learning_rate": 1.0443798301889731e-05, + "loss": 1.0796, + "step": 40704 + }, + { + "epoch": 3.01, + "learning_rate": 1.0443399822913037e-05, + "loss": 1.0217, + "step": 40705 + }, + { + "epoch": 3.01, + "learning_rate": 1.04430013432309e-05, + "loss": 0.9435, + "step": 40706 + }, + { + "epoch": 3.01, + "learning_rate": 1.0442602862843955e-05, + "loss": 0.9596, + "step": 40707 + }, + { + "epoch": 3.01, + "learning_rate": 1.0442204381752836e-05, + "loss": 1.0585, + "step": 40708 + }, + { + "epoch": 3.01, + "learning_rate": 1.0441805899958176e-05, + "loss": 1.055, + "step": 40709 + }, + { + "epoch": 3.01, + "learning_rate": 1.0441407417460607e-05, + "loss": 1.002, + "step": 40710 + }, + { + "epoch": 3.01, + "learning_rate": 1.0441008934260768e-05, + "loss": 1.13, + "step": 40711 + }, + { + "epoch": 3.01, + "learning_rate": 1.044061045035929e-05, + "loss": 0.9105, + "step": 40712 + }, + { + "epoch": 3.01, + "learning_rate": 1.0440211965756806e-05, + "loss": 0.9054, + "step": 40713 + }, + { + "epoch": 3.01, + "learning_rate": 1.043981348045395e-05, + "loss": 1.0093, + "step": 40714 + }, + { + "epoch": 3.01, + "learning_rate": 1.0439414994451358e-05, + "loss": 0.9149, + "step": 40715 + }, + { + "epoch": 3.01, + "learning_rate": 1.0439016507749663e-05, + "loss": 1.0445, + "step": 40716 + }, + { + "epoch": 3.01, + "learning_rate": 1.0438618020349499e-05, + "loss": 0.9741, + "step": 40717 + }, + { + "epoch": 3.01, + "learning_rate": 1.0438219532251499e-05, + "loss": 1.0085, + "step": 40718 + }, + { + "epoch": 3.01, + "learning_rate": 1.0437821043456301e-05, + "loss": 1.0576, + "step": 40719 + }, + { + "epoch": 3.01, + "learning_rate": 1.0437422553964533e-05, + "loss": 1.0627, + "step": 40720 + }, + { + "epoch": 3.01, + "learning_rate": 1.0437024063776833e-05, + "loss": 0.9881, + "step": 40721 + }, + { + "epoch": 3.01, + "learning_rate": 1.0436625572893833e-05, + "loss": 1.1071, + "step": 40722 + }, + { + "epoch": 3.01, + "learning_rate": 1.043622708131617e-05, + "loss": 0.8843, + "step": 40723 + }, + { + "epoch": 3.01, + "learning_rate": 1.0435828589044475e-05, + "loss": 1.0731, + "step": 40724 + }, + { + "epoch": 3.01, + "learning_rate": 1.0435430096079381e-05, + "loss": 1.0536, + "step": 40725 + }, + { + "epoch": 3.01, + "learning_rate": 1.0435031602421525e-05, + "loss": 0.9791, + "step": 40726 + }, + { + "epoch": 3.01, + "learning_rate": 1.0434633108071543e-05, + "loss": 1.05, + "step": 40727 + }, + { + "epoch": 3.01, + "learning_rate": 1.0434234613030064e-05, + "loss": 1.0468, + "step": 40728 + }, + { + "epoch": 3.01, + "learning_rate": 1.0433836117297724e-05, + "loss": 1.0946, + "step": 40729 + }, + { + "epoch": 3.01, + "learning_rate": 1.0433437620875156e-05, + "loss": 0.9767, + "step": 40730 + }, + { + "epoch": 3.01, + "learning_rate": 1.0433039123762996e-05, + "loss": 0.9719, + "step": 40731 + }, + { + "epoch": 3.01, + "learning_rate": 1.0432640625961876e-05, + "loss": 1.0256, + "step": 40732 + }, + { + "epoch": 3.01, + "learning_rate": 1.0432242127472432e-05, + "loss": 1.0584, + "step": 40733 + }, + { + "epoch": 3.01, + "learning_rate": 1.0431843628295294e-05, + "loss": 1.0499, + "step": 40734 + }, + { + "epoch": 3.01, + "learning_rate": 1.0431445128431104e-05, + "loss": 1.0816, + "step": 40735 + }, + { + "epoch": 3.01, + "learning_rate": 1.0431046627880485e-05, + "loss": 1.1619, + "step": 40736 + }, + { + "epoch": 3.01, + "learning_rate": 1.0430648126644083e-05, + "loss": 0.9162, + "step": 40737 + }, + { + "epoch": 3.01, + "learning_rate": 1.0430249624722524e-05, + "loss": 0.9833, + "step": 40738 + }, + { + "epoch": 3.01, + "learning_rate": 1.0429851122116443e-05, + "loss": 1.0611, + "step": 40739 + }, + { + "epoch": 3.01, + "learning_rate": 1.0429452618826475e-05, + "loss": 1.0526, + "step": 40740 + }, + { + "epoch": 3.01, + "learning_rate": 1.0429054114853254e-05, + "loss": 1.0476, + "step": 40741 + }, + { + "epoch": 3.01, + "learning_rate": 1.0428655610197413e-05, + "loss": 0.9554, + "step": 40742 + }, + { + "epoch": 3.01, + "learning_rate": 1.042825710485959e-05, + "loss": 0.9889, + "step": 40743 + }, + { + "epoch": 3.01, + "learning_rate": 1.0427858598840412e-05, + "loss": 0.9618, + "step": 40744 + }, + { + "epoch": 3.01, + "learning_rate": 1.0427460092140522e-05, + "loss": 0.9678, + "step": 40745 + }, + { + "epoch": 3.01, + "learning_rate": 1.0427061584760544e-05, + "loss": 0.9815, + "step": 40746 + }, + { + "epoch": 3.01, + "learning_rate": 1.0426663076701124e-05, + "loss": 0.8851, + "step": 40747 + }, + { + "epoch": 3.01, + "learning_rate": 1.0426264567962885e-05, + "loss": 0.9721, + "step": 40748 + }, + { + "epoch": 3.01, + "learning_rate": 1.0425866058546465e-05, + "loss": 1.0056, + "step": 40749 + }, + { + "epoch": 3.01, + "learning_rate": 1.0425467548452497e-05, + "loss": 0.9592, + "step": 40750 + }, + { + "epoch": 3.01, + "learning_rate": 1.0425069037681619e-05, + "loss": 1.029, + "step": 40751 + }, + { + "epoch": 3.01, + "learning_rate": 1.042467052623446e-05, + "loss": 0.9847, + "step": 40752 + }, + { + "epoch": 3.01, + "learning_rate": 1.0424272014111657e-05, + "loss": 0.873, + "step": 40753 + }, + { + "epoch": 3.01, + "learning_rate": 1.042387350131384e-05, + "loss": 0.9956, + "step": 40754 + }, + { + "epoch": 3.01, + "learning_rate": 1.0423474987841653e-05, + "loss": 0.8785, + "step": 40755 + }, + { + "epoch": 3.01, + "learning_rate": 1.042307647369572e-05, + "loss": 1.0081, + "step": 40756 + }, + { + "epoch": 3.01, + "learning_rate": 1.0422677958876678e-05, + "loss": 1.0681, + "step": 40757 + }, + { + "epoch": 3.01, + "learning_rate": 1.0422279443385162e-05, + "loss": 0.9175, + "step": 40758 + }, + { + "epoch": 3.01, + "learning_rate": 1.0421880927221805e-05, + "loss": 1.0368, + "step": 40759 + }, + { + "epoch": 3.01, + "learning_rate": 1.0421482410387242e-05, + "loss": 0.9647, + "step": 40760 + }, + { + "epoch": 3.01, + "learning_rate": 1.0421083892882107e-05, + "loss": 1.0676, + "step": 40761 + }, + { + "epoch": 3.01, + "learning_rate": 1.042068537470703e-05, + "loss": 0.854, + "step": 40762 + }, + { + "epoch": 3.01, + "learning_rate": 1.0420286855862653e-05, + "loss": 1.0884, + "step": 40763 + }, + { + "epoch": 3.01, + "learning_rate": 1.0419888336349604e-05, + "loss": 1.065, + "step": 40764 + }, + { + "epoch": 3.01, + "learning_rate": 1.041948981616852e-05, + "loss": 1.0745, + "step": 40765 + }, + { + "epoch": 3.01, + "learning_rate": 1.0419091295320031e-05, + "loss": 1.0596, + "step": 40766 + }, + { + "epoch": 3.01, + "learning_rate": 1.0418692773804774e-05, + "loss": 0.9398, + "step": 40767 + }, + { + "epoch": 3.01, + "learning_rate": 1.0418294251623387e-05, + "loss": 0.9863, + "step": 40768 + }, + { + "epoch": 3.01, + "learning_rate": 1.0417895728776494e-05, + "loss": 1.0066, + "step": 40769 + }, + { + "epoch": 3.01, + "learning_rate": 1.041749720526474e-05, + "loss": 1.0393, + "step": 40770 + }, + { + "epoch": 3.01, + "learning_rate": 1.041709868108875e-05, + "loss": 1.0364, + "step": 40771 + }, + { + "epoch": 3.01, + "learning_rate": 1.0416700156249164e-05, + "loss": 0.9079, + "step": 40772 + }, + { + "epoch": 3.01, + "learning_rate": 1.0416301630746615e-05, + "loss": 0.9393, + "step": 40773 + }, + { + "epoch": 3.01, + "learning_rate": 1.0415903104581733e-05, + "loss": 0.9219, + "step": 40774 + }, + { + "epoch": 3.01, + "learning_rate": 1.0415504577755157e-05, + "loss": 0.9005, + "step": 40775 + }, + { + "epoch": 3.01, + "learning_rate": 1.041510605026752e-05, + "loss": 1.0011, + "step": 40776 + }, + { + "epoch": 3.01, + "learning_rate": 1.0414707522119453e-05, + "loss": 0.9774, + "step": 40777 + }, + { + "epoch": 3.01, + "learning_rate": 1.0414308993311595e-05, + "loss": 1.1319, + "step": 40778 + }, + { + "epoch": 3.01, + "learning_rate": 1.0413910463844574e-05, + "loss": 1.0224, + "step": 40779 + }, + { + "epoch": 3.01, + "learning_rate": 1.0413511933719029e-05, + "loss": 1.0246, + "step": 40780 + }, + { + "epoch": 3.01, + "learning_rate": 1.0413113402935593e-05, + "loss": 1.0748, + "step": 40781 + }, + { + "epoch": 3.01, + "learning_rate": 1.04127148714949e-05, + "loss": 1.0483, + "step": 40782 + }, + { + "epoch": 3.01, + "learning_rate": 1.0412316339397581e-05, + "loss": 1.0872, + "step": 40783 + }, + { + "epoch": 3.01, + "learning_rate": 1.0411917806644274e-05, + "loss": 0.9709, + "step": 40784 + }, + { + "epoch": 3.01, + "learning_rate": 1.0411519273235612e-05, + "loss": 1.1049, + "step": 40785 + }, + { + "epoch": 3.01, + "learning_rate": 1.0411120739172232e-05, + "loss": 0.8665, + "step": 40786 + }, + { + "epoch": 3.01, + "learning_rate": 1.0410722204454757e-05, + "loss": 0.9697, + "step": 40787 + }, + { + "epoch": 3.01, + "learning_rate": 1.0410323669083835e-05, + "loss": 1.1367, + "step": 40788 + }, + { + "epoch": 3.01, + "learning_rate": 1.040992513306009e-05, + "loss": 0.9769, + "step": 40789 + }, + { + "epoch": 3.01, + "learning_rate": 1.0409526596384165e-05, + "loss": 0.9547, + "step": 40790 + }, + { + "epoch": 3.01, + "learning_rate": 1.0409128059056684e-05, + "loss": 1.0942, + "step": 40791 + }, + { + "epoch": 3.01, + "learning_rate": 1.040872952107829e-05, + "loss": 1.0067, + "step": 40792 + }, + { + "epoch": 3.01, + "learning_rate": 1.0408330982449608e-05, + "loss": 0.9751, + "step": 40793 + }, + { + "epoch": 3.01, + "learning_rate": 1.0407932443171285e-05, + "loss": 0.9258, + "step": 40794 + }, + { + "epoch": 3.01, + "learning_rate": 1.0407533903243943e-05, + "loss": 1.077, + "step": 40795 + }, + { + "epoch": 3.01, + "learning_rate": 1.0407135362668218e-05, + "loss": 1.035, + "step": 40796 + }, + { + "epoch": 3.01, + "learning_rate": 1.040673682144475e-05, + "loss": 1.0367, + "step": 40797 + }, + { + "epoch": 3.01, + "learning_rate": 1.040633827957417e-05, + "loss": 0.8966, + "step": 40798 + }, + { + "epoch": 3.01, + "learning_rate": 1.0405939737057106e-05, + "loss": 1.067, + "step": 40799 + }, + { + "epoch": 3.01, + "learning_rate": 1.0405541193894205e-05, + "loss": 0.9658, + "step": 40800 + }, + { + "epoch": 3.01, + "learning_rate": 1.0405142650086089e-05, + "loss": 0.9712, + "step": 40801 + }, + { + "epoch": 3.01, + "learning_rate": 1.0404744105633397e-05, + "loss": 1.0013, + "step": 40802 + }, + { + "epoch": 3.01, + "learning_rate": 1.0404345560536766e-05, + "loss": 0.9775, + "step": 40803 + }, + { + "epoch": 3.01, + "learning_rate": 1.0403947014796828e-05, + "loss": 0.9337, + "step": 40804 + }, + { + "epoch": 3.02, + "learning_rate": 1.0403548468414211e-05, + "loss": 0.9814, + "step": 40805 + }, + { + "epoch": 3.02, + "learning_rate": 1.0403149921389558e-05, + "loss": 1.0412, + "step": 40806 + }, + { + "epoch": 3.02, + "learning_rate": 1.0402751373723493e-05, + "loss": 1.0035, + "step": 40807 + }, + { + "epoch": 3.02, + "learning_rate": 1.0402352825416664e-05, + "loss": 1.1446, + "step": 40808 + }, + { + "epoch": 3.02, + "learning_rate": 1.0401954276469693e-05, + "loss": 0.9973, + "step": 40809 + }, + { + "epoch": 3.02, + "learning_rate": 1.0401555726883222e-05, + "loss": 1.078, + "step": 40810 + }, + { + "epoch": 3.02, + "learning_rate": 1.0401157176657876e-05, + "loss": 1.0745, + "step": 40811 + }, + { + "epoch": 3.02, + "learning_rate": 1.04007586257943e-05, + "loss": 0.9777, + "step": 40812 + }, + { + "epoch": 3.02, + "learning_rate": 1.0400360074293121e-05, + "loss": 0.9891, + "step": 40813 + }, + { + "epoch": 3.02, + "learning_rate": 1.0399961522154973e-05, + "loss": 1.0279, + "step": 40814 + }, + { + "epoch": 3.02, + "learning_rate": 1.0399562969380494e-05, + "loss": 1.0607, + "step": 40815 + }, + { + "epoch": 3.02, + "learning_rate": 1.0399164415970318e-05, + "loss": 1.0831, + "step": 40816 + }, + { + "epoch": 3.02, + "learning_rate": 1.0398765861925073e-05, + "loss": 0.9585, + "step": 40817 + }, + { + "epoch": 3.02, + "learning_rate": 1.0398367307245401e-05, + "loss": 1.0056, + "step": 40818 + }, + { + "epoch": 3.02, + "learning_rate": 1.0397968751931929e-05, + "loss": 0.9053, + "step": 40819 + }, + { + "epoch": 3.02, + "learning_rate": 1.0397570195985296e-05, + "loss": 1.0167, + "step": 40820 + }, + { + "epoch": 3.02, + "learning_rate": 1.0397171639406134e-05, + "loss": 1.0473, + "step": 40821 + }, + { + "epoch": 3.02, + "learning_rate": 1.0396773082195077e-05, + "loss": 0.9455, + "step": 40822 + }, + { + "epoch": 3.02, + "learning_rate": 1.0396374524352761e-05, + "loss": 1.0392, + "step": 40823 + }, + { + "epoch": 3.02, + "learning_rate": 1.0395975965879818e-05, + "loss": 0.8962, + "step": 40824 + }, + { + "epoch": 3.02, + "learning_rate": 1.0395577406776885e-05, + "loss": 1.0431, + "step": 40825 + }, + { + "epoch": 3.02, + "learning_rate": 1.0395178847044592e-05, + "loss": 1.1235, + "step": 40826 + }, + { + "epoch": 3.02, + "learning_rate": 1.0394780286683574e-05, + "loss": 1.0449, + "step": 40827 + }, + { + "epoch": 3.02, + "learning_rate": 1.0394381725694468e-05, + "loss": 1.0173, + "step": 40828 + }, + { + "epoch": 3.02, + "learning_rate": 1.0393983164077906e-05, + "loss": 1.0476, + "step": 40829 + }, + { + "epoch": 3.02, + "learning_rate": 1.0393584601834522e-05, + "loss": 0.8369, + "step": 40830 + }, + { + "epoch": 3.02, + "learning_rate": 1.0393186038964953e-05, + "loss": 0.9605, + "step": 40831 + }, + { + "epoch": 3.02, + "learning_rate": 1.0392787475469828e-05, + "loss": 0.894, + "step": 40832 + }, + { + "epoch": 3.02, + "learning_rate": 1.0392388911349785e-05, + "loss": 0.9574, + "step": 40833 + }, + { + "epoch": 3.02, + "learning_rate": 1.0391990346605454e-05, + "loss": 0.9579, + "step": 40834 + }, + { + "epoch": 3.02, + "learning_rate": 1.0391591781237475e-05, + "loss": 0.9323, + "step": 40835 + }, + { + "epoch": 3.02, + "learning_rate": 1.0391193215246478e-05, + "loss": 1.022, + "step": 40836 + }, + { + "epoch": 3.02, + "learning_rate": 1.0390794648633098e-05, + "loss": 1.031, + "step": 40837 + }, + { + "epoch": 3.02, + "learning_rate": 1.039039608139797e-05, + "loss": 0.9568, + "step": 40838 + }, + { + "epoch": 3.02, + "learning_rate": 1.0389997513541728e-05, + "loss": 1.0634, + "step": 40839 + }, + { + "epoch": 3.02, + "learning_rate": 1.0389598945065002e-05, + "loss": 1.0529, + "step": 40840 + }, + { + "epoch": 3.02, + "learning_rate": 1.0389200375968435e-05, + "loss": 0.9546, + "step": 40841 + }, + { + "epoch": 3.02, + "learning_rate": 1.0388801806252653e-05, + "loss": 0.9895, + "step": 40842 + }, + { + "epoch": 3.02, + "learning_rate": 1.0388403235918293e-05, + "loss": 1.0073, + "step": 40843 + }, + { + "epoch": 3.02, + "learning_rate": 1.0388004664965987e-05, + "loss": 0.9666, + "step": 40844 + }, + { + "epoch": 3.02, + "learning_rate": 1.0387606093396374e-05, + "loss": 0.9273, + "step": 40845 + }, + { + "epoch": 3.02, + "learning_rate": 1.0387207521210083e-05, + "loss": 0.8717, + "step": 40846 + }, + { + "epoch": 3.02, + "learning_rate": 1.0386808948407753e-05, + "loss": 0.9718, + "step": 40847 + }, + { + "epoch": 3.02, + "learning_rate": 1.0386410374990015e-05, + "loss": 1.0205, + "step": 40848 + }, + { + "epoch": 3.02, + "learning_rate": 1.0386011800957503e-05, + "loss": 0.9627, + "step": 40849 + }, + { + "epoch": 3.02, + "learning_rate": 1.038561322631085e-05, + "loss": 1.0486, + "step": 40850 + }, + { + "epoch": 3.02, + "learning_rate": 1.0385214651050697e-05, + "loss": 1.0282, + "step": 40851 + }, + { + "epoch": 3.02, + "learning_rate": 1.0384816075177669e-05, + "loss": 1.0272, + "step": 40852 + }, + { + "epoch": 3.02, + "learning_rate": 1.0384417498692403e-05, + "loss": 1.065, + "step": 40853 + }, + { + "epoch": 3.02, + "learning_rate": 1.0384018921595535e-05, + "loss": 0.8753, + "step": 40854 + }, + { + "epoch": 3.02, + "learning_rate": 1.0383620343887702e-05, + "loss": 0.9248, + "step": 40855 + }, + { + "epoch": 3.02, + "learning_rate": 1.0383221765569528e-05, + "loss": 0.9461, + "step": 40856 + }, + { + "epoch": 3.02, + "learning_rate": 1.038282318664166e-05, + "loss": 1.0545, + "step": 40857 + }, + { + "epoch": 3.02, + "learning_rate": 1.0382424607104722e-05, + "loss": 1.0962, + "step": 40858 + }, + { + "epoch": 3.02, + "learning_rate": 1.0382026026959354e-05, + "loss": 1.003, + "step": 40859 + }, + { + "epoch": 3.02, + "learning_rate": 1.0381627446206189e-05, + "loss": 1.0058, + "step": 40860 + }, + { + "epoch": 3.02, + "learning_rate": 1.0381228864845858e-05, + "loss": 0.9896, + "step": 40861 + }, + { + "epoch": 3.02, + "learning_rate": 1.0380830282878995e-05, + "loss": 1.019, + "step": 40862 + }, + { + "epoch": 3.02, + "learning_rate": 1.0380431700306242e-05, + "loss": 0.8768, + "step": 40863 + }, + { + "epoch": 3.02, + "learning_rate": 1.0380033117128223e-05, + "loss": 1.0067, + "step": 40864 + }, + { + "epoch": 3.02, + "learning_rate": 1.0379634533345579e-05, + "loss": 0.987, + "step": 40865 + }, + { + "epoch": 3.02, + "learning_rate": 1.037923594895894e-05, + "loss": 0.933, + "step": 40866 + }, + { + "epoch": 3.02, + "learning_rate": 1.0378837363968945e-05, + "loss": 0.8783, + "step": 40867 + }, + { + "epoch": 3.02, + "learning_rate": 1.037843877837622e-05, + "loss": 0.9287, + "step": 40868 + }, + { + "epoch": 3.02, + "learning_rate": 1.037804019218141e-05, + "loss": 1.0049, + "step": 40869 + }, + { + "epoch": 3.02, + "learning_rate": 1.0377641605385143e-05, + "loss": 1.0193, + "step": 40870 + }, + { + "epoch": 3.02, + "learning_rate": 1.037724301798805e-05, + "loss": 1.0297, + "step": 40871 + }, + { + "epoch": 3.02, + "learning_rate": 1.0376844429990774e-05, + "loss": 1.0174, + "step": 40872 + }, + { + "epoch": 3.02, + "learning_rate": 1.0376445841393938e-05, + "loss": 0.9972, + "step": 40873 + }, + { + "epoch": 3.02, + "learning_rate": 1.0376047252198186e-05, + "loss": 0.9469, + "step": 40874 + }, + { + "epoch": 3.02, + "learning_rate": 1.0375648662404145e-05, + "loss": 0.9397, + "step": 40875 + }, + { + "epoch": 3.02, + "learning_rate": 1.0375250072012455e-05, + "loss": 1.0861, + "step": 40876 + }, + { + "epoch": 3.02, + "learning_rate": 1.0374851481023748e-05, + "loss": 1.0121, + "step": 40877 + }, + { + "epoch": 3.02, + "learning_rate": 1.0374452889438658e-05, + "loss": 1.0457, + "step": 40878 + }, + { + "epoch": 3.02, + "learning_rate": 1.0374054297257814e-05, + "loss": 1.154, + "step": 40879 + }, + { + "epoch": 3.02, + "learning_rate": 1.037365570448186e-05, + "loss": 1.0431, + "step": 40880 + }, + { + "epoch": 3.02, + "learning_rate": 1.037325711111142e-05, + "loss": 1.0157, + "step": 40881 + }, + { + "epoch": 3.02, + "learning_rate": 1.037285851714714e-05, + "loss": 1.0831, + "step": 40882 + }, + { + "epoch": 3.02, + "learning_rate": 1.0372459922589642e-05, + "loss": 0.9528, + "step": 40883 + }, + { + "epoch": 3.02, + "learning_rate": 1.0372061327439568e-05, + "loss": 0.9542, + "step": 40884 + }, + { + "epoch": 3.02, + "learning_rate": 1.037166273169755e-05, + "loss": 1.0387, + "step": 40885 + }, + { + "epoch": 3.02, + "learning_rate": 1.037126413536422e-05, + "loss": 0.9883, + "step": 40886 + }, + { + "epoch": 3.02, + "learning_rate": 1.0370865538440219e-05, + "loss": 1.1135, + "step": 40887 + }, + { + "epoch": 3.02, + "learning_rate": 1.0370466940926172e-05, + "loss": 0.9029, + "step": 40888 + }, + { + "epoch": 3.02, + "learning_rate": 1.0370068342822716e-05, + "loss": 0.8287, + "step": 40889 + }, + { + "epoch": 3.02, + "learning_rate": 1.0369669744130491e-05, + "loss": 0.9753, + "step": 40890 + }, + { + "epoch": 3.02, + "learning_rate": 1.0369271144850122e-05, + "loss": 0.9379, + "step": 40891 + }, + { + "epoch": 3.02, + "learning_rate": 1.0368872544982252e-05, + "loss": 0.968, + "step": 40892 + }, + { + "epoch": 3.02, + "learning_rate": 1.0368473944527507e-05, + "loss": 1.0169, + "step": 40893 + }, + { + "epoch": 3.02, + "learning_rate": 1.0368075343486527e-05, + "loss": 1.0206, + "step": 40894 + }, + { + "epoch": 3.02, + "learning_rate": 1.0367676741859944e-05, + "loss": 0.889, + "step": 40895 + }, + { + "epoch": 3.02, + "learning_rate": 1.0367278139648396e-05, + "loss": 0.9633, + "step": 40896 + }, + { + "epoch": 3.02, + "learning_rate": 1.0366879536852513e-05, + "loss": 1.0132, + "step": 40897 + }, + { + "epoch": 3.02, + "learning_rate": 1.0366480933472927e-05, + "loss": 0.9052, + "step": 40898 + }, + { + "epoch": 3.02, + "learning_rate": 1.0366082329510277e-05, + "loss": 1.0331, + "step": 40899 + }, + { + "epoch": 3.02, + "learning_rate": 1.0365683724965194e-05, + "loss": 1.0966, + "step": 40900 + }, + { + "epoch": 3.02, + "learning_rate": 1.0365285119838312e-05, + "loss": 0.9945, + "step": 40901 + }, + { + "epoch": 3.02, + "learning_rate": 1.036488651413027e-05, + "loss": 0.9317, + "step": 40902 + }, + { + "epoch": 3.02, + "learning_rate": 1.0364487907841697e-05, + "loss": 0.9568, + "step": 40903 + }, + { + "epoch": 3.02, + "learning_rate": 1.0364089300973228e-05, + "loss": 1.1406, + "step": 40904 + }, + { + "epoch": 3.02, + "learning_rate": 1.03636906935255e-05, + "loss": 0.9024, + "step": 40905 + }, + { + "epoch": 3.02, + "learning_rate": 1.0363292085499146e-05, + "loss": 1.0453, + "step": 40906 + }, + { + "epoch": 3.02, + "learning_rate": 1.0362893476894797e-05, + "loss": 0.9961, + "step": 40907 + }, + { + "epoch": 3.02, + "learning_rate": 1.036249486771309e-05, + "loss": 1.0145, + "step": 40908 + }, + { + "epoch": 3.02, + "learning_rate": 1.036209625795466e-05, + "loss": 1.0059, + "step": 40909 + }, + { + "epoch": 3.02, + "learning_rate": 1.036169764762014e-05, + "loss": 1.003, + "step": 40910 + }, + { + "epoch": 3.02, + "learning_rate": 1.0361299036710165e-05, + "loss": 1.0365, + "step": 40911 + }, + { + "epoch": 3.02, + "learning_rate": 1.0360900425225367e-05, + "loss": 1.0445, + "step": 40912 + }, + { + "epoch": 3.02, + "learning_rate": 1.0360501813166379e-05, + "loss": 0.9621, + "step": 40913 + }, + { + "epoch": 3.02, + "learning_rate": 1.0360103200533842e-05, + "loss": 0.9199, + "step": 40914 + }, + { + "epoch": 3.02, + "learning_rate": 1.0359704587328384e-05, + "loss": 0.9181, + "step": 40915 + }, + { + "epoch": 3.02, + "learning_rate": 1.0359305973550641e-05, + "loss": 1.036, + "step": 40916 + }, + { + "epoch": 3.02, + "learning_rate": 1.0358907359201251e-05, + "loss": 1.0121, + "step": 40917 + }, + { + "epoch": 3.02, + "learning_rate": 1.035850874428084e-05, + "loss": 0.9502, + "step": 40918 + }, + { + "epoch": 3.02, + "learning_rate": 1.0358110128790049e-05, + "loss": 1.031, + "step": 40919 + }, + { + "epoch": 3.02, + "learning_rate": 1.0357711512729506e-05, + "loss": 1.0319, + "step": 40920 + }, + { + "epoch": 3.02, + "learning_rate": 1.0357312896099856e-05, + "loss": 0.9878, + "step": 40921 + }, + { + "epoch": 3.02, + "learning_rate": 1.0356914278901721e-05, + "loss": 0.9193, + "step": 40922 + }, + { + "epoch": 3.02, + "learning_rate": 1.0356515661135742e-05, + "loss": 1.0015, + "step": 40923 + }, + { + "epoch": 3.02, + "learning_rate": 1.0356117042802554e-05, + "loss": 0.9179, + "step": 40924 + }, + { + "epoch": 3.02, + "learning_rate": 1.0355718423902787e-05, + "loss": 0.9876, + "step": 40925 + }, + { + "epoch": 3.02, + "learning_rate": 1.0355319804437074e-05, + "loss": 1.1162, + "step": 40926 + }, + { + "epoch": 3.02, + "learning_rate": 1.0354921184406058e-05, + "loss": 0.914, + "step": 40927 + }, + { + "epoch": 3.02, + "learning_rate": 1.0354522563810364e-05, + "loss": 1.0682, + "step": 40928 + }, + { + "epoch": 3.02, + "learning_rate": 1.0354123942650628e-05, + "loss": 1.0427, + "step": 40929 + }, + { + "epoch": 3.02, + "learning_rate": 1.0353725320927488e-05, + "loss": 0.9482, + "step": 40930 + }, + { + "epoch": 3.02, + "learning_rate": 1.0353326698641576e-05, + "loss": 0.9907, + "step": 40931 + }, + { + "epoch": 3.02, + "learning_rate": 1.0352928075793527e-05, + "loss": 1.0542, + "step": 40932 + }, + { + "epoch": 3.02, + "learning_rate": 1.0352529452383972e-05, + "loss": 0.9765, + "step": 40933 + }, + { + "epoch": 3.02, + "learning_rate": 1.0352130828413551e-05, + "loss": 0.9006, + "step": 40934 + }, + { + "epoch": 3.02, + "learning_rate": 1.0351732203882896e-05, + "loss": 0.9976, + "step": 40935 + }, + { + "epoch": 3.02, + "learning_rate": 1.0351333578792635e-05, + "loss": 1.0818, + "step": 40936 + }, + { + "epoch": 3.02, + "learning_rate": 1.0350934953143411e-05, + "loss": 0.9715, + "step": 40937 + }, + { + "epoch": 3.02, + "learning_rate": 1.0350536326935851e-05, + "loss": 0.9493, + "step": 40938 + }, + { + "epoch": 3.02, + "learning_rate": 1.0350137700170598e-05, + "loss": 1.0364, + "step": 40939 + }, + { + "epoch": 3.03, + "learning_rate": 1.0349739072848278e-05, + "loss": 1.0214, + "step": 40940 + }, + { + "epoch": 3.03, + "learning_rate": 1.0349340444969527e-05, + "loss": 0.9969, + "step": 40941 + }, + { + "epoch": 3.03, + "learning_rate": 1.0348941816534981e-05, + "loss": 1.0766, + "step": 40942 + }, + { + "epoch": 3.03, + "learning_rate": 1.0348543187545275e-05, + "loss": 0.8842, + "step": 40943 + }, + { + "epoch": 3.03, + "learning_rate": 1.034814455800104e-05, + "loss": 0.9777, + "step": 40944 + }, + { + "epoch": 3.03, + "learning_rate": 1.0347745927902915e-05, + "loss": 0.9029, + "step": 40945 + }, + { + "epoch": 3.03, + "learning_rate": 1.0347347297251527e-05, + "loss": 0.9357, + "step": 40946 + }, + { + "epoch": 3.03, + "learning_rate": 1.0346948666047518e-05, + "loss": 0.9524, + "step": 40947 + }, + { + "epoch": 3.03, + "learning_rate": 1.0346550034291516e-05, + "loss": 1.0132, + "step": 40948 + }, + { + "epoch": 3.03, + "learning_rate": 1.034615140198416e-05, + "loss": 0.9524, + "step": 40949 + }, + { + "epoch": 3.03, + "learning_rate": 1.0345752769126078e-05, + "loss": 0.8916, + "step": 40950 + }, + { + "epoch": 3.03, + "learning_rate": 1.0345354135717914e-05, + "loss": 0.873, + "step": 40951 + }, + { + "epoch": 3.03, + "learning_rate": 1.0344955501760292e-05, + "loss": 1.0037, + "step": 40952 + }, + { + "epoch": 3.03, + "learning_rate": 1.0344556867253857e-05, + "loss": 1.0133, + "step": 40953 + }, + { + "epoch": 3.03, + "learning_rate": 1.034415823219923e-05, + "loss": 1.025, + "step": 40954 + }, + { + "epoch": 3.03, + "learning_rate": 1.0343759596597055e-05, + "loss": 0.9273, + "step": 40955 + }, + { + "epoch": 3.03, + "learning_rate": 1.034336096044796e-05, + "loss": 0.9695, + "step": 40956 + }, + { + "epoch": 3.03, + "learning_rate": 1.0342962323752588e-05, + "loss": 1.003, + "step": 40957 + }, + { + "epoch": 3.03, + "learning_rate": 1.0342563686511564e-05, + "loss": 0.9588, + "step": 40958 + }, + { + "epoch": 3.03, + "learning_rate": 1.034216504872553e-05, + "loss": 0.9856, + "step": 40959 + }, + { + "epoch": 3.03, + "learning_rate": 1.0341766410395113e-05, + "loss": 0.8821, + "step": 40960 + }, + { + "epoch": 3.03, + "learning_rate": 1.0341367771520951e-05, + "loss": 0.9721, + "step": 40961 + }, + { + "epoch": 3.03, + "learning_rate": 1.034096913210368e-05, + "loss": 1.0208, + "step": 40962 + }, + { + "epoch": 3.03, + "learning_rate": 1.0340570492143929e-05, + "loss": 1.0013, + "step": 40963 + }, + { + "epoch": 3.03, + "learning_rate": 1.0340171851642338e-05, + "loss": 1.0038, + "step": 40964 + }, + { + "epoch": 3.03, + "learning_rate": 1.0339773210599538e-05, + "loss": 1.0412, + "step": 40965 + }, + { + "epoch": 3.03, + "learning_rate": 1.0339374569016161e-05, + "loss": 1.0823, + "step": 40966 + }, + { + "epoch": 3.03, + "learning_rate": 1.0338975926892845e-05, + "loss": 1.0007, + "step": 40967 + }, + { + "epoch": 3.03, + "learning_rate": 1.0338577284230222e-05, + "loss": 0.9653, + "step": 40968 + }, + { + "epoch": 3.03, + "learning_rate": 1.033817864102893e-05, + "loss": 0.9924, + "step": 40969 + }, + { + "epoch": 3.03, + "learning_rate": 1.03377799972896e-05, + "loss": 1.0296, + "step": 40970 + }, + { + "epoch": 3.03, + "learning_rate": 1.0337381353012868e-05, + "loss": 0.9559, + "step": 40971 + }, + { + "epoch": 3.03, + "learning_rate": 1.0336982708199365e-05, + "loss": 1.0897, + "step": 40972 + }, + { + "epoch": 3.03, + "learning_rate": 1.0336584062849726e-05, + "loss": 1.0766, + "step": 40973 + }, + { + "epoch": 3.03, + "learning_rate": 1.033618541696459e-05, + "loss": 1.0663, + "step": 40974 + }, + { + "epoch": 3.03, + "learning_rate": 1.0335786770544585e-05, + "loss": 1.0672, + "step": 40975 + }, + { + "epoch": 3.03, + "learning_rate": 1.033538812359035e-05, + "loss": 0.9201, + "step": 40976 + }, + { + "epoch": 3.03, + "learning_rate": 1.0334989476102515e-05, + "loss": 0.9273, + "step": 40977 + }, + { + "epoch": 3.03, + "learning_rate": 1.0334590828081719e-05, + "loss": 1.0333, + "step": 40978 + }, + { + "epoch": 3.03, + "learning_rate": 1.0334192179528592e-05, + "loss": 1.0324, + "step": 40979 + }, + { + "epoch": 3.03, + "learning_rate": 1.0333793530443772e-05, + "loss": 1.0142, + "step": 40980 + }, + { + "epoch": 3.03, + "learning_rate": 1.0333394880827892e-05, + "loss": 1.0197, + "step": 40981 + }, + { + "epoch": 3.03, + "learning_rate": 1.0332996230681585e-05, + "loss": 0.9399, + "step": 40982 + }, + { + "epoch": 3.03, + "learning_rate": 1.033259758000548e-05, + "loss": 1.0346, + "step": 40983 + }, + { + "epoch": 3.03, + "learning_rate": 1.0332198928800226e-05, + "loss": 1.0001, + "step": 40984 + }, + { + "epoch": 3.03, + "learning_rate": 1.0331800277066442e-05, + "loss": 1.0388, + "step": 40985 + }, + { + "epoch": 3.03, + "learning_rate": 1.0331401624804772e-05, + "loss": 0.9744, + "step": 40986 + }, + { + "epoch": 3.03, + "learning_rate": 1.0331002972015842e-05, + "loss": 1.0685, + "step": 40987 + }, + { + "epoch": 3.03, + "learning_rate": 1.0330604318700295e-05, + "loss": 0.9452, + "step": 40988 + }, + { + "epoch": 3.03, + "learning_rate": 1.033020566485876e-05, + "loss": 1.0211, + "step": 40989 + }, + { + "epoch": 3.03, + "learning_rate": 1.0329807010491875e-05, + "loss": 1.0349, + "step": 40990 + }, + { + "epoch": 3.03, + "learning_rate": 1.0329408355600267e-05, + "loss": 1.0752, + "step": 40991 + }, + { + "epoch": 3.03, + "learning_rate": 1.032900970018458e-05, + "loss": 0.9811, + "step": 40992 + }, + { + "epoch": 3.03, + "learning_rate": 1.032861104424544e-05, + "loss": 1.0369, + "step": 40993 + }, + { + "epoch": 3.03, + "learning_rate": 1.0328212387783486e-05, + "loss": 0.9806, + "step": 40994 + }, + { + "epoch": 3.03, + "learning_rate": 1.032781373079935e-05, + "loss": 0.9828, + "step": 40995 + }, + { + "epoch": 3.03, + "learning_rate": 1.0327415073293668e-05, + "loss": 0.9689, + "step": 40996 + }, + { + "epoch": 3.03, + "learning_rate": 1.032701641526707e-05, + "loss": 1.0355, + "step": 40997 + }, + { + "epoch": 3.03, + "learning_rate": 1.0326617756720196e-05, + "loss": 1.1563, + "step": 40998 + }, + { + "epoch": 3.03, + "learning_rate": 1.0326219097653678e-05, + "loss": 0.948, + "step": 40999 + }, + { + "epoch": 3.03, + "learning_rate": 1.0325820438068152e-05, + "loss": 1.0371, + "step": 41000 + }, + { + "epoch": 3.03, + "learning_rate": 1.0325421777964248e-05, + "loss": 0.9837, + "step": 41001 + }, + { + "epoch": 3.03, + "learning_rate": 1.0325023117342603e-05, + "loss": 1.0124, + "step": 41002 + }, + { + "epoch": 3.03, + "learning_rate": 1.0324624456203851e-05, + "loss": 1.1412, + "step": 41003 + }, + { + "epoch": 3.03, + "learning_rate": 1.0324225794548626e-05, + "loss": 1.1289, + "step": 41004 + }, + { + "epoch": 3.03, + "learning_rate": 1.0323827132377561e-05, + "loss": 1.0439, + "step": 41005 + }, + { + "epoch": 3.03, + "learning_rate": 1.0323428469691294e-05, + "loss": 1.0087, + "step": 41006 + }, + { + "epoch": 3.03, + "learning_rate": 1.0323029806490454e-05, + "loss": 0.9488, + "step": 41007 + }, + { + "epoch": 3.03, + "learning_rate": 1.0322631142775684e-05, + "loss": 1.0136, + "step": 41008 + }, + { + "epoch": 3.03, + "learning_rate": 1.0322232478547606e-05, + "loss": 1.1003, + "step": 41009 + }, + { + "epoch": 3.03, + "learning_rate": 1.0321833813806866e-05, + "loss": 1.1518, + "step": 41010 + }, + { + "epoch": 3.03, + "learning_rate": 1.0321435148554088e-05, + "loss": 0.9577, + "step": 41011 + }, + { + "epoch": 3.03, + "learning_rate": 1.0321036482789917e-05, + "loss": 0.9568, + "step": 41012 + }, + { + "epoch": 3.03, + "learning_rate": 1.0320637816514976e-05, + "loss": 0.9709, + "step": 41013 + }, + { + "epoch": 3.03, + "learning_rate": 1.0320239149729907e-05, + "loss": 1.0026, + "step": 41014 + }, + { + "epoch": 3.03, + "learning_rate": 1.0319840482435342e-05, + "loss": 1.0318, + "step": 41015 + }, + { + "epoch": 3.03, + "learning_rate": 1.0319441814631916e-05, + "loss": 0.9827, + "step": 41016 + }, + { + "epoch": 3.03, + "learning_rate": 1.0319043146320261e-05, + "loss": 0.9566, + "step": 41017 + }, + { + "epoch": 3.03, + "learning_rate": 1.0318644477501014e-05, + "loss": 1.0149, + "step": 41018 + }, + { + "epoch": 3.03, + "learning_rate": 1.031824580817481e-05, + "loss": 0.9289, + "step": 41019 + }, + { + "epoch": 3.03, + "learning_rate": 1.0317847138342278e-05, + "loss": 0.9019, + "step": 41020 + }, + { + "epoch": 3.03, + "learning_rate": 1.0317448468004059e-05, + "loss": 0.8691, + "step": 41021 + }, + { + "epoch": 3.03, + "learning_rate": 1.0317049797160781e-05, + "loss": 0.9944, + "step": 41022 + }, + { + "epoch": 3.03, + "learning_rate": 1.0316651125813083e-05, + "loss": 1.0342, + "step": 41023 + }, + { + "epoch": 3.03, + "learning_rate": 1.0316252453961598e-05, + "loss": 0.9467, + "step": 41024 + }, + { + "epoch": 3.03, + "learning_rate": 1.0315853781606959e-05, + "loss": 0.974, + "step": 41025 + }, + { + "epoch": 3.03, + "learning_rate": 1.0315455108749803e-05, + "loss": 1.0554, + "step": 41026 + }, + { + "epoch": 3.03, + "learning_rate": 1.031505643539076e-05, + "loss": 1.0257, + "step": 41027 + }, + { + "epoch": 3.03, + "learning_rate": 1.0314657761530467e-05, + "loss": 1.0061, + "step": 41028 + }, + { + "epoch": 3.03, + "learning_rate": 1.031425908716956e-05, + "loss": 1.0128, + "step": 41029 + }, + { + "epoch": 3.03, + "learning_rate": 1.031386041230867e-05, + "loss": 0.9364, + "step": 41030 + }, + { + "epoch": 3.03, + "learning_rate": 1.0313461736948431e-05, + "loss": 0.968, + "step": 41031 + }, + { + "epoch": 3.03, + "learning_rate": 1.0313063061089482e-05, + "loss": 1.009, + "step": 41032 + }, + { + "epoch": 3.03, + "learning_rate": 1.0312664384732451e-05, + "loss": 1.0312, + "step": 41033 + }, + { + "epoch": 3.03, + "learning_rate": 1.0312265707877976e-05, + "loss": 1.0053, + "step": 41034 + }, + { + "epoch": 3.03, + "learning_rate": 1.0311867030526691e-05, + "loss": 1.0001, + "step": 41035 + }, + { + "epoch": 3.03, + "learning_rate": 1.031146835267923e-05, + "loss": 1.0029, + "step": 41036 + }, + { + "epoch": 3.03, + "learning_rate": 1.0311069674336232e-05, + "loss": 0.9549, + "step": 41037 + }, + { + "epoch": 3.03, + "learning_rate": 1.0310670995498321e-05, + "loss": 1.0323, + "step": 41038 + }, + { + "epoch": 3.03, + "learning_rate": 1.0310272316166139e-05, + "loss": 0.9419, + "step": 41039 + }, + { + "epoch": 3.03, + "learning_rate": 1.0309873636340317e-05, + "loss": 0.9327, + "step": 41040 + }, + { + "epoch": 3.03, + "learning_rate": 1.0309474956021492e-05, + "loss": 1.005, + "step": 41041 + }, + { + "epoch": 3.03, + "learning_rate": 1.0309076275210295e-05, + "loss": 1.0158, + "step": 41042 + }, + { + "epoch": 3.03, + "learning_rate": 1.0308677593907363e-05, + "loss": 0.9899, + "step": 41043 + }, + { + "epoch": 3.03, + "learning_rate": 1.0308278912113327e-05, + "loss": 0.9263, + "step": 41044 + }, + { + "epoch": 3.03, + "learning_rate": 1.0307880229828828e-05, + "loss": 0.9052, + "step": 41045 + }, + { + "epoch": 3.03, + "learning_rate": 1.0307481547054495e-05, + "loss": 0.9326, + "step": 41046 + }, + { + "epoch": 3.03, + "learning_rate": 1.0307082863790963e-05, + "loss": 1.0387, + "step": 41047 + }, + { + "epoch": 3.03, + "learning_rate": 1.0306684180038864e-05, + "loss": 1.0811, + "step": 41048 + }, + { + "epoch": 3.03, + "learning_rate": 1.0306285495798837e-05, + "loss": 0.8993, + "step": 41049 + }, + { + "epoch": 3.03, + "learning_rate": 1.0305886811071512e-05, + "loss": 0.9514, + "step": 41050 + }, + { + "epoch": 3.03, + "learning_rate": 1.0305488125857531e-05, + "loss": 1.0011, + "step": 41051 + }, + { + "epoch": 3.03, + "learning_rate": 1.0305089440157516e-05, + "loss": 1.0082, + "step": 41052 + }, + { + "epoch": 3.03, + "learning_rate": 1.0304690753972113e-05, + "loss": 0.94, + "step": 41053 + }, + { + "epoch": 3.03, + "learning_rate": 1.0304292067301948e-05, + "loss": 0.9942, + "step": 41054 + }, + { + "epoch": 3.03, + "learning_rate": 1.030389338014766e-05, + "loss": 1.0601, + "step": 41055 + }, + { + "epoch": 3.03, + "learning_rate": 1.0303494692509883e-05, + "loss": 0.9596, + "step": 41056 + }, + { + "epoch": 3.03, + "learning_rate": 1.030309600438925e-05, + "loss": 1.0789, + "step": 41057 + }, + { + "epoch": 3.03, + "learning_rate": 1.0302697315786394e-05, + "loss": 0.9924, + "step": 41058 + }, + { + "epoch": 3.03, + "learning_rate": 1.0302298626701953e-05, + "loss": 0.9473, + "step": 41059 + }, + { + "epoch": 3.03, + "learning_rate": 1.0301899937136557e-05, + "loss": 0.8957, + "step": 41060 + }, + { + "epoch": 3.03, + "learning_rate": 1.0301501247090843e-05, + "loss": 0.9311, + "step": 41061 + }, + { + "epoch": 3.03, + "learning_rate": 1.0301102556565445e-05, + "loss": 1.0989, + "step": 41062 + }, + { + "epoch": 3.03, + "learning_rate": 1.0300703865561e-05, + "loss": 0.9909, + "step": 41063 + }, + { + "epoch": 3.03, + "learning_rate": 1.0300305174078136e-05, + "loss": 0.9912, + "step": 41064 + }, + { + "epoch": 3.03, + "learning_rate": 1.0299906482117492e-05, + "loss": 0.9623, + "step": 41065 + }, + { + "epoch": 3.03, + "learning_rate": 1.0299507789679702e-05, + "loss": 0.9897, + "step": 41066 + }, + { + "epoch": 3.03, + "learning_rate": 1.0299109096765397e-05, + "loss": 0.9852, + "step": 41067 + }, + { + "epoch": 3.03, + "learning_rate": 1.0298710403375217e-05, + "loss": 0.8965, + "step": 41068 + }, + { + "epoch": 3.03, + "learning_rate": 1.0298311709509787e-05, + "loss": 0.9769, + "step": 41069 + }, + { + "epoch": 3.03, + "learning_rate": 1.0297913015169754e-05, + "loss": 1.0437, + "step": 41070 + }, + { + "epoch": 3.03, + "learning_rate": 1.0297514320355744e-05, + "loss": 0.9428, + "step": 41071 + }, + { + "epoch": 3.03, + "learning_rate": 1.029711562506839e-05, + "loss": 0.9938, + "step": 41072 + }, + { + "epoch": 3.03, + "learning_rate": 1.0296716929308332e-05, + "loss": 0.9709, + "step": 41073 + }, + { + "epoch": 3.03, + "learning_rate": 1.0296318233076202e-05, + "loss": 0.9568, + "step": 41074 + }, + { + "epoch": 3.04, + "learning_rate": 1.0295919536372632e-05, + "loss": 1.0046, + "step": 41075 + }, + { + "epoch": 3.04, + "learning_rate": 1.0295520839198258e-05, + "loss": 1.0056, + "step": 41076 + }, + { + "epoch": 3.04, + "learning_rate": 1.0295122141553714e-05, + "loss": 1.0366, + "step": 41077 + }, + { + "epoch": 3.04, + "learning_rate": 1.0294723443439638e-05, + "loss": 1.0053, + "step": 41078 + }, + { + "epoch": 3.04, + "learning_rate": 1.0294324744856658e-05, + "loss": 0.9555, + "step": 41079 + }, + { + "epoch": 3.04, + "learning_rate": 1.0293926045805415e-05, + "loss": 0.9957, + "step": 41080 + }, + { + "epoch": 3.04, + "learning_rate": 1.0293527346286538e-05, + "loss": 1.0441, + "step": 41081 + }, + { + "epoch": 3.04, + "learning_rate": 1.0293128646300663e-05, + "loss": 0.9448, + "step": 41082 + }, + { + "epoch": 3.04, + "learning_rate": 1.0292729945848422e-05, + "loss": 1.0772, + "step": 41083 + }, + { + "epoch": 3.04, + "learning_rate": 1.029233124493046e-05, + "loss": 1.0212, + "step": 41084 + }, + { + "epoch": 3.04, + "learning_rate": 1.0291932543547396e-05, + "loss": 0.9951, + "step": 41085 + }, + { + "epoch": 3.04, + "learning_rate": 1.0291533841699875e-05, + "loss": 1.0743, + "step": 41086 + }, + { + "epoch": 3.04, + "learning_rate": 1.0291135139388524e-05, + "loss": 0.9647, + "step": 41087 + }, + { + "epoch": 3.04, + "learning_rate": 1.0290736436613985e-05, + "loss": 1.07, + "step": 41088 + }, + { + "epoch": 3.04, + "learning_rate": 1.0290337733376886e-05, + "loss": 0.9775, + "step": 41089 + }, + { + "epoch": 3.04, + "learning_rate": 1.0289939029677864e-05, + "loss": 0.9841, + "step": 41090 + }, + { + "epoch": 3.04, + "learning_rate": 1.0289540325517553e-05, + "loss": 1.0176, + "step": 41091 + }, + { + "epoch": 3.04, + "learning_rate": 1.028914162089659e-05, + "loss": 0.9794, + "step": 41092 + }, + { + "epoch": 3.04, + "learning_rate": 1.0288742915815606e-05, + "loss": 1.0067, + "step": 41093 + }, + { + "epoch": 3.04, + "learning_rate": 1.0288344210275236e-05, + "loss": 1.094, + "step": 41094 + }, + { + "epoch": 3.04, + "learning_rate": 1.0287945504276112e-05, + "loss": 1.0645, + "step": 41095 + }, + { + "epoch": 3.04, + "learning_rate": 1.0287546797818873e-05, + "loss": 0.895, + "step": 41096 + }, + { + "epoch": 3.04, + "learning_rate": 1.0287148090904148e-05, + "loss": 0.9397, + "step": 41097 + }, + { + "epoch": 3.04, + "learning_rate": 1.028674938353258e-05, + "loss": 1.0243, + "step": 41098 + }, + { + "epoch": 3.04, + "learning_rate": 1.0286350675704794e-05, + "loss": 0.9297, + "step": 41099 + }, + { + "epoch": 3.04, + "learning_rate": 1.028595196742143e-05, + "loss": 0.9795, + "step": 41100 + }, + { + "epoch": 3.04, + "learning_rate": 1.0285553258683118e-05, + "loss": 1.1839, + "step": 41101 + }, + { + "epoch": 3.04, + "learning_rate": 1.0285154549490499e-05, + "loss": 0.885, + "step": 41102 + }, + { + "epoch": 3.04, + "learning_rate": 1.0284755839844201e-05, + "loss": 0.9929, + "step": 41103 + }, + { + "epoch": 3.04, + "learning_rate": 1.028435712974486e-05, + "loss": 1.0897, + "step": 41104 + }, + { + "epoch": 3.04, + "learning_rate": 1.028395841919311e-05, + "loss": 1.0474, + "step": 41105 + }, + { + "epoch": 3.04, + "learning_rate": 1.028355970818959e-05, + "loss": 1.1439, + "step": 41106 + }, + { + "epoch": 3.04, + "learning_rate": 1.0283160996734927e-05, + "loss": 1.0772, + "step": 41107 + }, + { + "epoch": 3.04, + "learning_rate": 1.0282762284829759e-05, + "loss": 0.9896, + "step": 41108 + }, + { + "epoch": 3.04, + "learning_rate": 1.0282363572474721e-05, + "loss": 1.0186, + "step": 41109 + }, + { + "epoch": 3.04, + "learning_rate": 1.0281964859670446e-05, + "loss": 1.0625, + "step": 41110 + }, + { + "epoch": 3.04, + "learning_rate": 1.028156614641757e-05, + "loss": 0.9036, + "step": 41111 + }, + { + "epoch": 3.04, + "learning_rate": 1.0281167432716726e-05, + "loss": 1.0272, + "step": 41112 + }, + { + "epoch": 3.04, + "learning_rate": 1.028076871856855e-05, + "loss": 1.1195, + "step": 41113 + }, + { + "epoch": 3.04, + "learning_rate": 1.0280370003973672e-05, + "loss": 1.0467, + "step": 41114 + }, + { + "epoch": 3.04, + "learning_rate": 1.027997128893273e-05, + "loss": 1.0245, + "step": 41115 + }, + { + "epoch": 3.04, + "learning_rate": 1.0279572573446361e-05, + "loss": 0.9275, + "step": 41116 + }, + { + "epoch": 3.04, + "learning_rate": 1.0279173857515193e-05, + "loss": 0.9361, + "step": 41117 + }, + { + "epoch": 3.04, + "learning_rate": 1.0278775141139866e-05, + "loss": 0.9231, + "step": 41118 + }, + { + "epoch": 3.04, + "learning_rate": 1.0278376424321007e-05, + "loss": 1.0048, + "step": 41119 + }, + { + "epoch": 3.04, + "learning_rate": 1.027797770705926e-05, + "loss": 1.0708, + "step": 41120 + }, + { + "epoch": 3.04, + "learning_rate": 1.0277578989355253e-05, + "loss": 1.0024, + "step": 41121 + }, + { + "epoch": 3.04, + "learning_rate": 1.0277180271209618e-05, + "loss": 0.9921, + "step": 41122 + }, + { + "epoch": 3.04, + "learning_rate": 1.0276781552622998e-05, + "loss": 0.9381, + "step": 41123 + }, + { + "epoch": 3.04, + "learning_rate": 1.027638283359602e-05, + "loss": 0.9176, + "step": 41124 + }, + { + "epoch": 3.04, + "learning_rate": 1.0275984114129323e-05, + "loss": 0.9604, + "step": 41125 + }, + { + "epoch": 3.04, + "learning_rate": 1.0275585394223536e-05, + "loss": 1.0747, + "step": 41126 + }, + { + "epoch": 3.04, + "learning_rate": 1.0275186673879299e-05, + "loss": 0.9834, + "step": 41127 + }, + { + "epoch": 3.04, + "learning_rate": 1.0274787953097242e-05, + "loss": 0.9872, + "step": 41128 + }, + { + "epoch": 3.04, + "learning_rate": 1.0274389231878003e-05, + "loss": 0.9757, + "step": 41129 + }, + { + "epoch": 3.04, + "learning_rate": 1.0273990510222216e-05, + "loss": 0.9389, + "step": 41130 + }, + { + "epoch": 3.04, + "learning_rate": 1.0273591788130513e-05, + "loss": 0.9519, + "step": 41131 + }, + { + "epoch": 3.04, + "learning_rate": 1.0273193065603526e-05, + "loss": 1.0249, + "step": 41132 + }, + { + "epoch": 3.04, + "learning_rate": 1.0272794342641898e-05, + "loss": 0.9007, + "step": 41133 + }, + { + "epoch": 3.04, + "learning_rate": 1.0272395619246253e-05, + "loss": 0.9352, + "step": 41134 + }, + { + "epoch": 3.04, + "learning_rate": 1.0271996895417236e-05, + "loss": 1.0275, + "step": 41135 + }, + { + "epoch": 3.04, + "learning_rate": 1.027159817115547e-05, + "loss": 0.9401, + "step": 41136 + }, + { + "epoch": 3.04, + "learning_rate": 1.02711994464616e-05, + "loss": 0.8958, + "step": 41137 + }, + { + "epoch": 3.04, + "learning_rate": 1.0270800721336251e-05, + "loss": 1.0172, + "step": 41138 + }, + { + "epoch": 3.04, + "learning_rate": 1.0270401995780066e-05, + "loss": 0.9722, + "step": 41139 + }, + { + "epoch": 3.04, + "learning_rate": 1.0270003269793676e-05, + "loss": 0.9362, + "step": 41140 + }, + { + "epoch": 3.04, + "learning_rate": 1.026960454337771e-05, + "loss": 1.0413, + "step": 41141 + }, + { + "epoch": 3.04, + "learning_rate": 1.026920581653281e-05, + "loss": 0.8545, + "step": 41142 + }, + { + "epoch": 3.04, + "learning_rate": 1.0268807089259608e-05, + "loss": 0.9904, + "step": 41143 + }, + { + "epoch": 3.04, + "learning_rate": 1.0268408361558734e-05, + "loss": 0.9658, + "step": 41144 + }, + { + "epoch": 3.04, + "learning_rate": 1.0268009633430832e-05, + "loss": 0.9934, + "step": 41145 + }, + { + "epoch": 3.04, + "learning_rate": 1.0267610904876526e-05, + "loss": 1.0978, + "step": 41146 + }, + { + "epoch": 3.04, + "learning_rate": 1.0267212175896458e-05, + "loss": 1.0402, + "step": 41147 + }, + { + "epoch": 3.04, + "learning_rate": 1.0266813446491255e-05, + "loss": 0.9599, + "step": 41148 + }, + { + "epoch": 3.04, + "learning_rate": 1.0266414716661562e-05, + "loss": 0.958, + "step": 41149 + }, + { + "epoch": 3.04, + "learning_rate": 1.0266015986408e-05, + "loss": 1.0366, + "step": 41150 + }, + { + "epoch": 3.04, + "learning_rate": 1.0265617255731216e-05, + "loss": 1.0605, + "step": 41151 + }, + { + "epoch": 3.04, + "learning_rate": 1.0265218524631835e-05, + "loss": 1.004, + "step": 41152 + }, + { + "epoch": 3.04, + "learning_rate": 1.0264819793110497e-05, + "loss": 0.9706, + "step": 41153 + }, + { + "epoch": 3.04, + "learning_rate": 1.026442106116783e-05, + "loss": 1.1133, + "step": 41154 + }, + { + "epoch": 3.04, + "learning_rate": 1.0264022328804479e-05, + "loss": 1.0592, + "step": 41155 + }, + { + "epoch": 3.04, + "learning_rate": 1.0263623596021068e-05, + "loss": 1.0471, + "step": 41156 + }, + { + "epoch": 3.04, + "learning_rate": 1.0263224862818239e-05, + "loss": 0.9874, + "step": 41157 + }, + { + "epoch": 3.04, + "learning_rate": 1.0262826129196621e-05, + "loss": 1.0204, + "step": 41158 + }, + { + "epoch": 3.04, + "learning_rate": 1.0262427395156852e-05, + "loss": 0.9178, + "step": 41159 + }, + { + "epoch": 3.04, + "learning_rate": 1.0262028660699564e-05, + "loss": 1.0286, + "step": 41160 + }, + { + "epoch": 3.04, + "learning_rate": 1.0261629925825393e-05, + "loss": 0.9774, + "step": 41161 + }, + { + "epoch": 3.04, + "learning_rate": 1.026123119053497e-05, + "loss": 1.0176, + "step": 41162 + }, + { + "epoch": 3.04, + "learning_rate": 1.0260832454828935e-05, + "loss": 0.9998, + "step": 41163 + }, + { + "epoch": 3.04, + "learning_rate": 1.0260433718707916e-05, + "loss": 1.0085, + "step": 41164 + }, + { + "epoch": 3.04, + "learning_rate": 1.0260034982172553e-05, + "loss": 1.1198, + "step": 41165 + }, + { + "epoch": 3.04, + "learning_rate": 1.0259636245223476e-05, + "loss": 0.92, + "step": 41166 + }, + { + "epoch": 3.04, + "learning_rate": 1.0259237507861324e-05, + "loss": 1.0159, + "step": 41167 + }, + { + "epoch": 3.04, + "learning_rate": 1.0258838770086727e-05, + "loss": 0.9806, + "step": 41168 + }, + { + "epoch": 3.04, + "learning_rate": 1.0258440031900322e-05, + "loss": 0.9809, + "step": 41169 + }, + { + "epoch": 3.04, + "learning_rate": 1.0258041293302744e-05, + "loss": 1.0589, + "step": 41170 + }, + { + "epoch": 3.04, + "learning_rate": 1.0257642554294622e-05, + "loss": 0.9514, + "step": 41171 + }, + { + "epoch": 3.04, + "learning_rate": 1.0257243814876598e-05, + "loss": 1.0684, + "step": 41172 + }, + { + "epoch": 3.04, + "learning_rate": 1.0256845075049302e-05, + "loss": 1.0036, + "step": 41173 + }, + { + "epoch": 3.04, + "learning_rate": 1.0256446334813369e-05, + "loss": 0.9744, + "step": 41174 + }, + { + "epoch": 3.04, + "learning_rate": 1.0256047594169431e-05, + "loss": 0.9463, + "step": 41175 + }, + { + "epoch": 3.04, + "learning_rate": 1.025564885311813e-05, + "loss": 1.106, + "step": 41176 + }, + { + "epoch": 3.04, + "learning_rate": 1.0255250111660093e-05, + "loss": 0.9904, + "step": 41177 + }, + { + "epoch": 3.04, + "learning_rate": 1.0254851369795957e-05, + "loss": 1.0795, + "step": 41178 + }, + { + "epoch": 3.04, + "learning_rate": 1.0254452627526355e-05, + "loss": 0.9179, + "step": 41179 + }, + { + "epoch": 3.04, + "learning_rate": 1.0254053884851922e-05, + "loss": 0.9063, + "step": 41180 + }, + { + "epoch": 3.04, + "learning_rate": 1.0253655141773293e-05, + "loss": 1.1212, + "step": 41181 + }, + { + "epoch": 3.04, + "learning_rate": 1.0253256398291104e-05, + "loss": 1.0444, + "step": 41182 + }, + { + "epoch": 3.04, + "learning_rate": 1.0252857654405985e-05, + "loss": 0.927, + "step": 41183 + }, + { + "epoch": 3.04, + "learning_rate": 1.0252458910118576e-05, + "loss": 0.9577, + "step": 41184 + }, + { + "epoch": 3.04, + "learning_rate": 1.0252060165429507e-05, + "loss": 0.971, + "step": 41185 + }, + { + "epoch": 3.04, + "learning_rate": 1.0251661420339416e-05, + "loss": 1.058, + "step": 41186 + }, + { + "epoch": 3.04, + "learning_rate": 1.0251262674848934e-05, + "loss": 0.9121, + "step": 41187 + }, + { + "epoch": 3.04, + "learning_rate": 1.0250863928958696e-05, + "loss": 0.9575, + "step": 41188 + }, + { + "epoch": 3.04, + "learning_rate": 1.0250465182669339e-05, + "loss": 1.0413, + "step": 41189 + }, + { + "epoch": 3.04, + "learning_rate": 1.0250066435981494e-05, + "loss": 1.0437, + "step": 41190 + }, + { + "epoch": 3.04, + "learning_rate": 1.0249667688895795e-05, + "loss": 1.1301, + "step": 41191 + }, + { + "epoch": 3.04, + "learning_rate": 1.024926894141288e-05, + "loss": 0.923, + "step": 41192 + }, + { + "epoch": 3.04, + "learning_rate": 1.024887019353338e-05, + "loss": 0.8973, + "step": 41193 + }, + { + "epoch": 3.04, + "learning_rate": 1.0248471445257935e-05, + "loss": 1.0658, + "step": 41194 + }, + { + "epoch": 3.04, + "learning_rate": 1.0248072696587173e-05, + "loss": 0.9715, + "step": 41195 + }, + { + "epoch": 3.04, + "learning_rate": 1.0247673947521732e-05, + "loss": 1.0589, + "step": 41196 + }, + { + "epoch": 3.04, + "learning_rate": 1.0247275198062244e-05, + "loss": 0.9973, + "step": 41197 + }, + { + "epoch": 3.04, + "learning_rate": 1.0246876448209347e-05, + "loss": 0.9331, + "step": 41198 + }, + { + "epoch": 3.04, + "learning_rate": 1.0246477697963668e-05, + "loss": 1.0816, + "step": 41199 + }, + { + "epoch": 3.04, + "learning_rate": 1.0246078947325851e-05, + "loss": 1.0277, + "step": 41200 + }, + { + "epoch": 3.04, + "learning_rate": 1.0245680196296524e-05, + "loss": 1.0425, + "step": 41201 + }, + { + "epoch": 3.04, + "learning_rate": 1.0245281444876325e-05, + "loss": 0.9236, + "step": 41202 + }, + { + "epoch": 3.04, + "learning_rate": 1.0244882693065885e-05, + "loss": 0.8833, + "step": 41203 + }, + { + "epoch": 3.04, + "learning_rate": 1.0244483940865841e-05, + "loss": 1.0363, + "step": 41204 + }, + { + "epoch": 3.04, + "learning_rate": 1.0244085188276826e-05, + "loss": 0.9244, + "step": 41205 + }, + { + "epoch": 3.04, + "learning_rate": 1.0243686435299478e-05, + "loss": 0.9173, + "step": 41206 + }, + { + "epoch": 3.04, + "learning_rate": 1.0243287681934423e-05, + "loss": 1.0112, + "step": 41207 + }, + { + "epoch": 3.04, + "learning_rate": 1.0242888928182305e-05, + "loss": 0.9056, + "step": 41208 + }, + { + "epoch": 3.04, + "learning_rate": 1.0242490174043751e-05, + "loss": 1.0975, + "step": 41209 + }, + { + "epoch": 3.04, + "learning_rate": 1.0242091419519402e-05, + "loss": 0.9747, + "step": 41210 + }, + { + "epoch": 3.05, + "learning_rate": 1.0241692664609886e-05, + "loss": 1.0203, + "step": 41211 + }, + { + "epoch": 3.05, + "learning_rate": 1.0241293909315843e-05, + "loss": 1.0091, + "step": 41212 + }, + { + "epoch": 3.05, + "learning_rate": 1.0240895153637902e-05, + "loss": 0.9973, + "step": 41213 + }, + { + "epoch": 3.05, + "learning_rate": 1.0240496397576703e-05, + "loss": 0.8934, + "step": 41214 + }, + { + "epoch": 3.05, + "learning_rate": 1.0240097641132877e-05, + "loss": 1.1362, + "step": 41215 + }, + { + "epoch": 3.05, + "learning_rate": 1.0239698884307056e-05, + "loss": 0.9436, + "step": 41216 + }, + { + "epoch": 3.05, + "learning_rate": 1.0239300127099883e-05, + "loss": 0.9474, + "step": 41217 + }, + { + "epoch": 3.05, + "learning_rate": 1.0238901369511983e-05, + "loss": 0.9772, + "step": 41218 + }, + { + "epoch": 3.05, + "learning_rate": 1.0238502611543994e-05, + "loss": 0.9721, + "step": 41219 + }, + { + "epoch": 3.05, + "learning_rate": 1.0238103853196555e-05, + "loss": 1.0766, + "step": 41220 + }, + { + "epoch": 3.05, + "learning_rate": 1.0237705094470292e-05, + "loss": 1.1039, + "step": 41221 + }, + { + "epoch": 3.05, + "learning_rate": 1.0237306335365847e-05, + "loss": 0.8815, + "step": 41222 + }, + { + "epoch": 3.05, + "learning_rate": 1.0236907575883846e-05, + "loss": 0.9351, + "step": 41223 + }, + { + "epoch": 3.05, + "learning_rate": 1.0236508816024932e-05, + "loss": 1.0049, + "step": 41224 + }, + { + "epoch": 3.05, + "learning_rate": 1.0236110055789736e-05, + "loss": 0.9958, + "step": 41225 + }, + { + "epoch": 3.05, + "learning_rate": 1.023571129517889e-05, + "loss": 0.988, + "step": 41226 + }, + { + "epoch": 3.05, + "learning_rate": 1.0235312534193034e-05, + "loss": 1.0822, + "step": 41227 + }, + { + "epoch": 3.05, + "learning_rate": 1.0234913772832795e-05, + "loss": 0.8798, + "step": 41228 + }, + { + "epoch": 3.05, + "learning_rate": 1.0234515011098816e-05, + "loss": 1.0475, + "step": 41229 + }, + { + "epoch": 3.05, + "learning_rate": 1.0234116248991723e-05, + "loss": 0.9865, + "step": 41230 + }, + { + "epoch": 3.05, + "learning_rate": 1.0233717486512159e-05, + "loss": 0.9316, + "step": 41231 + }, + { + "epoch": 3.05, + "learning_rate": 1.0233318723660749e-05, + "loss": 0.8423, + "step": 41232 + }, + { + "epoch": 3.05, + "learning_rate": 1.0232919960438137e-05, + "loss": 1.016, + "step": 41233 + }, + { + "epoch": 3.05, + "learning_rate": 1.023252119684495e-05, + "loss": 1.0348, + "step": 41234 + }, + { + "epoch": 3.05, + "learning_rate": 1.0232122432881824e-05, + "loss": 1.0786, + "step": 41235 + }, + { + "epoch": 3.05, + "learning_rate": 1.0231723668549394e-05, + "loss": 0.7754, + "step": 41236 + }, + { + "epoch": 3.05, + "learning_rate": 1.02313249038483e-05, + "loss": 0.9388, + "step": 41237 + }, + { + "epoch": 3.05, + "learning_rate": 1.0230926138779164e-05, + "loss": 0.9632, + "step": 41238 + }, + { + "epoch": 3.05, + "learning_rate": 1.0230527373342633e-05, + "loss": 1.0279, + "step": 41239 + }, + { + "epoch": 3.05, + "learning_rate": 1.0230128607539334e-05, + "loss": 1.08, + "step": 41240 + }, + { + "epoch": 3.05, + "learning_rate": 1.0229729841369906e-05, + "loss": 0.9954, + "step": 41241 + }, + { + "epoch": 3.05, + "learning_rate": 1.0229331074834978e-05, + "loss": 1.0007, + "step": 41242 + }, + { + "epoch": 3.05, + "learning_rate": 1.0228932307935193e-05, + "loss": 1.0929, + "step": 41243 + }, + { + "epoch": 3.05, + "learning_rate": 1.0228533540671175e-05, + "loss": 1.0008, + "step": 41244 + }, + { + "epoch": 3.05, + "learning_rate": 1.0228134773043565e-05, + "loss": 0.9377, + "step": 41245 + }, + { + "epoch": 3.05, + "learning_rate": 1.0227736005052994e-05, + "loss": 0.963, + "step": 41246 + }, + { + "epoch": 3.05, + "learning_rate": 1.0227337236700102e-05, + "loss": 0.955, + "step": 41247 + }, + { + "epoch": 3.05, + "learning_rate": 1.0226938467985516e-05, + "loss": 1.0693, + "step": 41248 + }, + { + "epoch": 3.05, + "learning_rate": 1.0226539698909876e-05, + "loss": 0.9832, + "step": 41249 + }, + { + "epoch": 3.05, + "learning_rate": 1.0226140929473813e-05, + "loss": 1.0045, + "step": 41250 + }, + { + "epoch": 3.05, + "learning_rate": 1.0225742159677965e-05, + "loss": 1.1058, + "step": 41251 + }, + { + "epoch": 3.05, + "learning_rate": 1.0225343389522966e-05, + "loss": 0.9581, + "step": 41252 + }, + { + "epoch": 3.05, + "learning_rate": 1.0224944619009446e-05, + "loss": 0.9936, + "step": 41253 + }, + { + "epoch": 3.05, + "learning_rate": 1.0224545848138044e-05, + "loss": 0.9892, + "step": 41254 + }, + { + "epoch": 3.05, + "learning_rate": 1.0224147076909392e-05, + "loss": 1.0274, + "step": 41255 + }, + { + "epoch": 3.05, + "learning_rate": 1.0223748305324122e-05, + "loss": 0.99, + "step": 41256 + }, + { + "epoch": 3.05, + "learning_rate": 1.0223349533382876e-05, + "loss": 0.9042, + "step": 41257 + }, + { + "epoch": 3.05, + "learning_rate": 1.0222950761086281e-05, + "loss": 0.8985, + "step": 41258 + }, + { + "epoch": 3.05, + "learning_rate": 1.0222551988434976e-05, + "loss": 0.9073, + "step": 41259 + }, + { + "epoch": 3.05, + "learning_rate": 1.0222153215429596e-05, + "loss": 0.9742, + "step": 41260 + }, + { + "epoch": 3.05, + "learning_rate": 1.0221754442070771e-05, + "loss": 0.9757, + "step": 41261 + }, + { + "epoch": 3.05, + "learning_rate": 1.0221355668359138e-05, + "loss": 0.9412, + "step": 41262 + }, + { + "epoch": 3.05, + "learning_rate": 1.0220956894295331e-05, + "loss": 1.0232, + "step": 41263 + }, + { + "epoch": 3.05, + "learning_rate": 1.0220558119879987e-05, + "loss": 1.06, + "step": 41264 + }, + { + "epoch": 3.05, + "learning_rate": 1.0220159345113739e-05, + "loss": 1.022, + "step": 41265 + }, + { + "epoch": 3.05, + "learning_rate": 1.0219760569997217e-05, + "loss": 0.9963, + "step": 41266 + }, + { + "epoch": 3.05, + "learning_rate": 1.021936179453106e-05, + "loss": 0.9898, + "step": 41267 + }, + { + "epoch": 3.05, + "learning_rate": 1.0218963018715901e-05, + "loss": 0.9012, + "step": 41268 + }, + { + "epoch": 3.05, + "learning_rate": 1.0218564242552376e-05, + "loss": 0.9609, + "step": 41269 + }, + { + "epoch": 3.05, + "learning_rate": 1.0218165466041116e-05, + "loss": 1.0845, + "step": 41270 + }, + { + "epoch": 3.05, + "learning_rate": 1.0217766689182761e-05, + "loss": 1.0936, + "step": 41271 + }, + { + "epoch": 3.05, + "learning_rate": 1.0217367911977944e-05, + "loss": 1.0664, + "step": 41272 + }, + { + "epoch": 3.05, + "learning_rate": 1.0216969134427294e-05, + "loss": 1.0979, + "step": 41273 + }, + { + "epoch": 3.05, + "learning_rate": 1.0216570356531452e-05, + "loss": 1.0379, + "step": 41274 + }, + { + "epoch": 3.05, + "learning_rate": 1.0216171578291046e-05, + "loss": 0.9095, + "step": 41275 + }, + { + "epoch": 3.05, + "learning_rate": 1.0215772799706717e-05, + "loss": 1.0502, + "step": 41276 + }, + { + "epoch": 3.05, + "learning_rate": 1.0215374020779095e-05, + "loss": 1.0074, + "step": 41277 + }, + { + "epoch": 3.05, + "learning_rate": 1.0214975241508819e-05, + "loss": 1.0398, + "step": 41278 + }, + { + "epoch": 3.05, + "learning_rate": 1.0214576461896515e-05, + "loss": 1.0532, + "step": 41279 + }, + { + "epoch": 3.05, + "learning_rate": 1.021417768194283e-05, + "loss": 1.0175, + "step": 41280 + }, + { + "epoch": 3.05, + "learning_rate": 1.0213778901648385e-05, + "loss": 1.0402, + "step": 41281 + }, + { + "epoch": 3.05, + "learning_rate": 1.0213380121013825e-05, + "loss": 0.9266, + "step": 41282 + }, + { + "epoch": 3.05, + "learning_rate": 1.0212981340039777e-05, + "loss": 1.0258, + "step": 41283 + }, + { + "epoch": 3.05, + "learning_rate": 1.0212582558726883e-05, + "loss": 1.0878, + "step": 41284 + }, + { + "epoch": 3.05, + "learning_rate": 1.0212183777075768e-05, + "loss": 1.0702, + "step": 41285 + }, + { + "epoch": 3.05, + "learning_rate": 1.0211784995087075e-05, + "loss": 0.9964, + "step": 41286 + }, + { + "epoch": 3.05, + "learning_rate": 1.0211386212761434e-05, + "loss": 1.0743, + "step": 41287 + }, + { + "epoch": 3.05, + "learning_rate": 1.0210987430099482e-05, + "loss": 1.0623, + "step": 41288 + }, + { + "epoch": 3.05, + "learning_rate": 1.021058864710185e-05, + "loss": 1.0775, + "step": 41289 + }, + { + "epoch": 3.05, + "learning_rate": 1.0210189863769177e-05, + "loss": 0.9454, + "step": 41290 + }, + { + "epoch": 3.05, + "learning_rate": 1.0209791080102092e-05, + "loss": 0.9876, + "step": 41291 + }, + { + "epoch": 3.05, + "learning_rate": 1.0209392296101235e-05, + "loss": 0.9904, + "step": 41292 + }, + { + "epoch": 3.05, + "learning_rate": 1.0208993511767236e-05, + "loss": 1.0337, + "step": 41293 + }, + { + "epoch": 3.05, + "learning_rate": 1.0208594727100732e-05, + "loss": 0.9823, + "step": 41294 + }, + { + "epoch": 3.05, + "learning_rate": 1.0208195942102356e-05, + "loss": 0.931, + "step": 41295 + }, + { + "epoch": 3.05, + "learning_rate": 1.0207797156772745e-05, + "loss": 0.9628, + "step": 41296 + }, + { + "epoch": 3.05, + "learning_rate": 1.0207398371112528e-05, + "loss": 1.0793, + "step": 41297 + }, + { + "epoch": 3.05, + "learning_rate": 1.0206999585122348e-05, + "loss": 1.0695, + "step": 41298 + }, + { + "epoch": 3.05, + "learning_rate": 1.0206600798802832e-05, + "loss": 0.9362, + "step": 41299 + }, + { + "epoch": 3.05, + "learning_rate": 1.020620201215462e-05, + "loss": 1.0976, + "step": 41300 + }, + { + "epoch": 3.05, + "learning_rate": 1.0205803225178338e-05, + "loss": 1.0234, + "step": 41301 + }, + { + "epoch": 3.05, + "learning_rate": 1.020540443787463e-05, + "loss": 0.9797, + "step": 41302 + }, + { + "epoch": 3.05, + "learning_rate": 1.0205005650244124e-05, + "loss": 1.087, + "step": 41303 + }, + { + "epoch": 3.05, + "learning_rate": 1.020460686228746e-05, + "loss": 1.0059, + "step": 41304 + }, + { + "epoch": 3.05, + "learning_rate": 1.0204208074005267e-05, + "loss": 1.0012, + "step": 41305 + }, + { + "epoch": 3.05, + "learning_rate": 1.0203809285398184e-05, + "loss": 1.0369, + "step": 41306 + }, + { + "epoch": 3.05, + "learning_rate": 1.0203410496466841e-05, + "loss": 1.0478, + "step": 41307 + }, + { + "epoch": 3.05, + "learning_rate": 1.0203011707211879e-05, + "loss": 0.9858, + "step": 41308 + }, + { + "epoch": 3.05, + "learning_rate": 1.0202612917633923e-05, + "loss": 0.965, + "step": 41309 + }, + { + "epoch": 3.05, + "learning_rate": 1.0202214127733616e-05, + "loss": 1.0866, + "step": 41310 + }, + { + "epoch": 3.05, + "learning_rate": 1.0201815337511586e-05, + "loss": 0.9969, + "step": 41311 + }, + { + "epoch": 3.05, + "learning_rate": 1.0201416546968474e-05, + "loss": 0.9231, + "step": 41312 + }, + { + "epoch": 3.05, + "learning_rate": 1.020101775610491e-05, + "loss": 1.0899, + "step": 41313 + }, + { + "epoch": 3.05, + "learning_rate": 1.0200618964921529e-05, + "loss": 0.9013, + "step": 41314 + }, + { + "epoch": 3.05, + "learning_rate": 1.0200220173418967e-05, + "loss": 1.1212, + "step": 41315 + }, + { + "epoch": 3.05, + "learning_rate": 1.0199821381597856e-05, + "loss": 0.9485, + "step": 41316 + }, + { + "epoch": 3.05, + "learning_rate": 1.0199422589458837e-05, + "loss": 0.972, + "step": 41317 + }, + { + "epoch": 3.05, + "learning_rate": 1.0199023797002534e-05, + "loss": 0.9964, + "step": 41318 + }, + { + "epoch": 3.05, + "learning_rate": 1.019862500422959e-05, + "loss": 0.9531, + "step": 41319 + }, + { + "epoch": 3.05, + "learning_rate": 1.0198226211140634e-05, + "loss": 1.039, + "step": 41320 + }, + { + "epoch": 3.05, + "learning_rate": 1.0197827417736304e-05, + "loss": 0.9727, + "step": 41321 + }, + { + "epoch": 3.05, + "learning_rate": 1.019742862401723e-05, + "loss": 1.0137, + "step": 41322 + }, + { + "epoch": 3.05, + "learning_rate": 1.0197029829984058e-05, + "loss": 1.0095, + "step": 41323 + }, + { + "epoch": 3.05, + "learning_rate": 1.0196631035637408e-05, + "loss": 1.0355, + "step": 41324 + }, + { + "epoch": 3.05, + "learning_rate": 1.0196232240977924e-05, + "loss": 0.9939, + "step": 41325 + }, + { + "epoch": 3.05, + "learning_rate": 1.0195833446006235e-05, + "loss": 1.0124, + "step": 41326 + }, + { + "epoch": 3.05, + "learning_rate": 1.0195434650722979e-05, + "loss": 1.0687, + "step": 41327 + }, + { + "epoch": 3.05, + "learning_rate": 1.019503585512879e-05, + "loss": 1.0714, + "step": 41328 + }, + { + "epoch": 3.05, + "learning_rate": 1.0194637059224302e-05, + "loss": 1.0599, + "step": 41329 + }, + { + "epoch": 3.05, + "learning_rate": 1.0194238263010146e-05, + "loss": 0.9975, + "step": 41330 + }, + { + "epoch": 3.05, + "learning_rate": 1.0193839466486964e-05, + "loss": 1.0253, + "step": 41331 + }, + { + "epoch": 3.05, + "learning_rate": 1.019344066965538e-05, + "loss": 0.9471, + "step": 41332 + }, + { + "epoch": 3.05, + "learning_rate": 1.0193041872516041e-05, + "loss": 0.9284, + "step": 41333 + }, + { + "epoch": 3.05, + "learning_rate": 1.0192643075069571e-05, + "loss": 1.0676, + "step": 41334 + }, + { + "epoch": 3.05, + "learning_rate": 1.0192244277316612e-05, + "loss": 1.0575, + "step": 41335 + }, + { + "epoch": 3.05, + "learning_rate": 1.0191845479257794e-05, + "loss": 0.881, + "step": 41336 + }, + { + "epoch": 3.05, + "learning_rate": 1.0191446680893753e-05, + "loss": 0.9787, + "step": 41337 + }, + { + "epoch": 3.05, + "learning_rate": 1.019104788222512e-05, + "loss": 1.0225, + "step": 41338 + }, + { + "epoch": 3.05, + "learning_rate": 1.0190649083252535e-05, + "loss": 0.9509, + "step": 41339 + }, + { + "epoch": 3.05, + "learning_rate": 1.0190250283976628e-05, + "loss": 1.0392, + "step": 41340 + }, + { + "epoch": 3.05, + "learning_rate": 1.0189851484398039e-05, + "loss": 0.982, + "step": 41341 + }, + { + "epoch": 3.05, + "learning_rate": 1.0189452684517396e-05, + "loss": 0.9662, + "step": 41342 + }, + { + "epoch": 3.05, + "learning_rate": 1.0189053884335338e-05, + "loss": 0.9662, + "step": 41343 + }, + { + "epoch": 3.05, + "learning_rate": 1.0188655083852497e-05, + "loss": 1.0304, + "step": 41344 + }, + { + "epoch": 3.05, + "learning_rate": 1.0188256283069514e-05, + "loss": 0.9442, + "step": 41345 + }, + { + "epoch": 3.06, + "learning_rate": 1.0187857481987009e-05, + "loss": 1.0294, + "step": 41346 + }, + { + "epoch": 3.06, + "learning_rate": 1.018745868060563e-05, + "loss": 1.0308, + "step": 41347 + }, + { + "epoch": 3.06, + "learning_rate": 1.0187059878926007e-05, + "loss": 0.8687, + "step": 41348 + }, + { + "epoch": 3.06, + "learning_rate": 1.0186661076948774e-05, + "loss": 1.1394, + "step": 41349 + }, + { + "epoch": 3.06, + "learning_rate": 1.0186262274674566e-05, + "loss": 1.0047, + "step": 41350 + }, + { + "epoch": 3.06, + "learning_rate": 1.0185863472104015e-05, + "loss": 1.0831, + "step": 41351 + }, + { + "epoch": 3.06, + "learning_rate": 1.0185464669237761e-05, + "loss": 1.1165, + "step": 41352 + }, + { + "epoch": 3.06, + "learning_rate": 1.0185065866076436e-05, + "loss": 0.9698, + "step": 41353 + }, + { + "epoch": 3.06, + "learning_rate": 1.018466706262067e-05, + "loss": 0.9433, + "step": 41354 + }, + { + "epoch": 3.06, + "learning_rate": 1.0184268258871106e-05, + "loss": 1.0177, + "step": 41355 + }, + { + "epoch": 3.06, + "learning_rate": 1.018386945482837e-05, + "loss": 1.0405, + "step": 41356 + }, + { + "epoch": 3.06, + "learning_rate": 1.0183470650493104e-05, + "loss": 0.9518, + "step": 41357 + }, + { + "epoch": 3.06, + "learning_rate": 1.0183071845865936e-05, + "loss": 1.0094, + "step": 41358 + }, + { + "epoch": 3.06, + "learning_rate": 1.0182673040947503e-05, + "loss": 0.9206, + "step": 41359 + }, + { + "epoch": 3.06, + "learning_rate": 1.018227423573844e-05, + "loss": 1.0316, + "step": 41360 + }, + { + "epoch": 3.06, + "learning_rate": 1.0181875430239382e-05, + "loss": 0.9637, + "step": 41361 + }, + { + "epoch": 3.06, + "learning_rate": 1.0181476624450961e-05, + "loss": 0.9487, + "step": 41362 + }, + { + "epoch": 3.06, + "learning_rate": 1.0181077818373816e-05, + "loss": 0.9758, + "step": 41363 + }, + { + "epoch": 3.06, + "learning_rate": 1.018067901200858e-05, + "loss": 0.9143, + "step": 41364 + }, + { + "epoch": 3.06, + "learning_rate": 1.0180280205355884e-05, + "loss": 0.9928, + "step": 41365 + }, + { + "epoch": 3.06, + "learning_rate": 1.0179881398416365e-05, + "loss": 1.006, + "step": 41366 + }, + { + "epoch": 3.06, + "learning_rate": 1.0179482591190654e-05, + "loss": 1.009, + "step": 41367 + }, + { + "epoch": 3.06, + "learning_rate": 1.0179083783679393e-05, + "loss": 0.9774, + "step": 41368 + }, + { + "epoch": 3.06, + "learning_rate": 1.0178684975883212e-05, + "loss": 0.9366, + "step": 41369 + }, + { + "epoch": 3.06, + "learning_rate": 1.0178286167802745e-05, + "loss": 0.9285, + "step": 41370 + }, + { + "epoch": 3.06, + "learning_rate": 1.0177887359438627e-05, + "loss": 0.9699, + "step": 41371 + }, + { + "epoch": 3.06, + "learning_rate": 1.0177488550791493e-05, + "loss": 0.9937, + "step": 41372 + }, + { + "epoch": 3.06, + "learning_rate": 1.0177089741861978e-05, + "loss": 1.0182, + "step": 41373 + }, + { + "epoch": 3.06, + "learning_rate": 1.0176690932650715e-05, + "loss": 0.9514, + "step": 41374 + }, + { + "epoch": 3.06, + "learning_rate": 1.0176292123158337e-05, + "loss": 1.0221, + "step": 41375 + }, + { + "epoch": 3.06, + "learning_rate": 1.0175893313385485e-05, + "loss": 1.0119, + "step": 41376 + }, + { + "epoch": 3.06, + "learning_rate": 1.0175494503332786e-05, + "loss": 0.9515, + "step": 41377 + }, + { + "epoch": 3.06, + "learning_rate": 1.017509569300088e-05, + "loss": 0.987, + "step": 41378 + }, + { + "epoch": 3.06, + "learning_rate": 1.0174696882390398e-05, + "loss": 0.9598, + "step": 41379 + }, + { + "epoch": 3.06, + "learning_rate": 1.0174298071501976e-05, + "loss": 1.0276, + "step": 41380 + }, + { + "epoch": 3.06, + "learning_rate": 1.0173899260336247e-05, + "loss": 0.8644, + "step": 41381 + }, + { + "epoch": 3.06, + "learning_rate": 1.017350044889385e-05, + "loss": 1.0105, + "step": 41382 + }, + { + "epoch": 3.06, + "learning_rate": 1.0173101637175414e-05, + "loss": 1.0264, + "step": 41383 + }, + { + "epoch": 3.06, + "learning_rate": 1.0172702825181578e-05, + "loss": 1.0151, + "step": 41384 + }, + { + "epoch": 3.06, + "learning_rate": 1.017230401291297e-05, + "loss": 1.1593, + "step": 41385 + }, + { + "epoch": 3.06, + "learning_rate": 1.0171905200370233e-05, + "loss": 0.9257, + "step": 41386 + }, + { + "epoch": 3.06, + "learning_rate": 1.0171506387553993e-05, + "loss": 1.0887, + "step": 41387 + }, + { + "epoch": 3.06, + "learning_rate": 1.0171107574464893e-05, + "loss": 1.012, + "step": 41388 + }, + { + "epoch": 3.06, + "learning_rate": 1.017070876110356e-05, + "loss": 1.0654, + "step": 41389 + }, + { + "epoch": 3.06, + "learning_rate": 1.0170309947470634e-05, + "loss": 0.949, + "step": 41390 + }, + { + "epoch": 3.06, + "learning_rate": 1.0169911133566746e-05, + "loss": 1.0636, + "step": 41391 + }, + { + "epoch": 3.06, + "learning_rate": 1.0169512319392534e-05, + "loss": 1.0474, + "step": 41392 + }, + { + "epoch": 3.06, + "learning_rate": 1.0169113504948628e-05, + "loss": 0.977, + "step": 41393 + }, + { + "epoch": 3.06, + "learning_rate": 1.0168714690235666e-05, + "loss": 0.9704, + "step": 41394 + }, + { + "epoch": 3.06, + "learning_rate": 1.0168315875254279e-05, + "loss": 0.9888, + "step": 41395 + }, + { + "epoch": 3.06, + "learning_rate": 1.0167917060005108e-05, + "loss": 0.9232, + "step": 41396 + }, + { + "epoch": 3.06, + "learning_rate": 1.016751824448878e-05, + "loss": 0.9903, + "step": 41397 + }, + { + "epoch": 3.06, + "learning_rate": 1.0167119428705936e-05, + "loss": 0.9224, + "step": 41398 + }, + { + "epoch": 3.06, + "learning_rate": 1.0166720612657205e-05, + "loss": 1.0472, + "step": 41399 + }, + { + "epoch": 3.06, + "learning_rate": 1.0166321796343223e-05, + "loss": 1.1271, + "step": 41400 + }, + { + "epoch": 3.06, + "learning_rate": 1.0165922979764627e-05, + "loss": 1.0878, + "step": 41401 + }, + { + "epoch": 3.06, + "learning_rate": 1.0165524162922051e-05, + "loss": 0.9201, + "step": 41402 + }, + { + "epoch": 3.06, + "learning_rate": 1.0165125345816127e-05, + "loss": 1.0784, + "step": 41403 + }, + { + "epoch": 3.06, + "learning_rate": 1.016472652844749e-05, + "loss": 1.1287, + "step": 41404 + }, + { + "epoch": 3.06, + "learning_rate": 1.0164327710816777e-05, + "loss": 0.9505, + "step": 41405 + }, + { + "epoch": 3.06, + "learning_rate": 1.0163928892924619e-05, + "loss": 0.9306, + "step": 41406 + }, + { + "epoch": 3.06, + "learning_rate": 1.0163530074771654e-05, + "loss": 1.0094, + "step": 41407 + }, + { + "epoch": 3.06, + "learning_rate": 1.0163131256358517e-05, + "loss": 0.9789, + "step": 41408 + }, + { + "epoch": 3.06, + "learning_rate": 1.0162732437685837e-05, + "loss": 1.0952, + "step": 41409 + }, + { + "epoch": 3.06, + "learning_rate": 1.0162333618754252e-05, + "loss": 1.0235, + "step": 41410 + }, + { + "epoch": 3.06, + "learning_rate": 1.0161934799564399e-05, + "loss": 0.9529, + "step": 41411 + }, + { + "epoch": 3.06, + "learning_rate": 1.0161535980116907e-05, + "loss": 0.9906, + "step": 41412 + }, + { + "epoch": 3.06, + "learning_rate": 1.0161137160412417e-05, + "loss": 0.9931, + "step": 41413 + }, + { + "epoch": 3.06, + "learning_rate": 1.016073834045156e-05, + "loss": 1.0091, + "step": 41414 + }, + { + "epoch": 3.06, + "learning_rate": 1.0160339520234968e-05, + "loss": 1.0265, + "step": 41415 + }, + { + "epoch": 3.06, + "learning_rate": 1.015994069976328e-05, + "loss": 1.0223, + "step": 41416 + }, + { + "epoch": 3.06, + "learning_rate": 1.0159541879037127e-05, + "loss": 0.9844, + "step": 41417 + }, + { + "epoch": 3.06, + "learning_rate": 1.0159143058057148e-05, + "loss": 1.0514, + "step": 41418 + }, + { + "epoch": 3.06, + "learning_rate": 1.015874423682397e-05, + "loss": 0.9382, + "step": 41419 + }, + { + "epoch": 3.06, + "learning_rate": 1.0158345415338235e-05, + "loss": 0.9707, + "step": 41420 + }, + { + "epoch": 3.06, + "learning_rate": 1.0157946593600577e-05, + "loss": 0.9136, + "step": 41421 + }, + { + "epoch": 3.06, + "learning_rate": 1.0157547771611625e-05, + "loss": 0.951, + "step": 41422 + }, + { + "epoch": 3.06, + "learning_rate": 1.0157148949372018e-05, + "loss": 1.0294, + "step": 41423 + }, + { + "epoch": 3.06, + "learning_rate": 1.0156750126882389e-05, + "loss": 0.9111, + "step": 41424 + }, + { + "epoch": 3.06, + "learning_rate": 1.0156351304143372e-05, + "loss": 0.9917, + "step": 41425 + }, + { + "epoch": 3.06, + "learning_rate": 1.0155952481155601e-05, + "loss": 1.0818, + "step": 41426 + }, + { + "epoch": 3.06, + "learning_rate": 1.0155553657919717e-05, + "loss": 1.0197, + "step": 41427 + }, + { + "epoch": 3.06, + "learning_rate": 1.0155154834436346e-05, + "loss": 1.0393, + "step": 41428 + }, + { + "epoch": 3.06, + "learning_rate": 1.0154756010706127e-05, + "loss": 0.9771, + "step": 41429 + }, + { + "epoch": 3.06, + "learning_rate": 1.015435718672969e-05, + "loss": 1.0432, + "step": 41430 + }, + { + "epoch": 3.06, + "learning_rate": 1.0153958362507679e-05, + "loss": 1.0019, + "step": 41431 + }, + { + "epoch": 3.06, + "learning_rate": 1.0153559538040715e-05, + "loss": 0.9914, + "step": 41432 + }, + { + "epoch": 3.06, + "learning_rate": 1.0153160713329446e-05, + "loss": 0.9911, + "step": 41433 + }, + { + "epoch": 3.06, + "learning_rate": 1.0152761888374498e-05, + "loss": 1.0416, + "step": 41434 + }, + { + "epoch": 3.06, + "learning_rate": 1.015236306317651e-05, + "loss": 0.9132, + "step": 41435 + }, + { + "epoch": 3.06, + "learning_rate": 1.0151964237736109e-05, + "loss": 1.0201, + "step": 41436 + }, + { + "epoch": 3.06, + "learning_rate": 1.0151565412053941e-05, + "loss": 0.9915, + "step": 41437 + }, + { + "epoch": 3.06, + "learning_rate": 1.0151166586130632e-05, + "loss": 0.9773, + "step": 41438 + }, + { + "epoch": 3.06, + "learning_rate": 1.0150767759966823e-05, + "loss": 0.922, + "step": 41439 + }, + { + "epoch": 3.06, + "learning_rate": 1.0150368933563139e-05, + "loss": 0.919, + "step": 41440 + }, + { + "epoch": 3.06, + "learning_rate": 1.0149970106920224e-05, + "loss": 1.0638, + "step": 41441 + }, + { + "epoch": 3.06, + "learning_rate": 1.0149571280038705e-05, + "loss": 0.9136, + "step": 41442 + }, + { + "epoch": 3.06, + "learning_rate": 1.0149172452919224e-05, + "loss": 1.0989, + "step": 41443 + }, + { + "epoch": 3.06, + "learning_rate": 1.0148773625562408e-05, + "loss": 0.9652, + "step": 41444 + }, + { + "epoch": 3.06, + "learning_rate": 1.0148374797968899e-05, + "loss": 1.0394, + "step": 41445 + }, + { + "epoch": 3.06, + "learning_rate": 1.0147975970139325e-05, + "loss": 0.9219, + "step": 41446 + }, + { + "epoch": 3.06, + "learning_rate": 1.0147577142074325e-05, + "loss": 1.0154, + "step": 41447 + }, + { + "epoch": 3.06, + "learning_rate": 1.0147178313774532e-05, + "loss": 0.8895, + "step": 41448 + }, + { + "epoch": 3.06, + "learning_rate": 1.0146779485240583e-05, + "loss": 1.0044, + "step": 41449 + }, + { + "epoch": 3.06, + "learning_rate": 1.0146380656473104e-05, + "loss": 0.9813, + "step": 41450 + }, + { + "epoch": 3.06, + "learning_rate": 1.014598182747274e-05, + "loss": 0.9721, + "step": 41451 + }, + { + "epoch": 3.06, + "learning_rate": 1.0145582998240118e-05, + "loss": 1.0339, + "step": 41452 + }, + { + "epoch": 3.06, + "learning_rate": 1.0145184168775879e-05, + "loss": 0.9851, + "step": 41453 + }, + { + "epoch": 3.06, + "learning_rate": 1.0144785339080649e-05, + "loss": 0.9658, + "step": 41454 + }, + { + "epoch": 3.06, + "learning_rate": 1.0144386509155073e-05, + "loss": 1.1225, + "step": 41455 + }, + { + "epoch": 3.06, + "learning_rate": 1.0143987678999775e-05, + "loss": 0.9858, + "step": 41456 + }, + { + "epoch": 3.06, + "learning_rate": 1.01435888486154e-05, + "loss": 1.0452, + "step": 41457 + }, + { + "epoch": 3.06, + "learning_rate": 1.0143190018002575e-05, + "loss": 0.9926, + "step": 41458 + }, + { + "epoch": 3.06, + "learning_rate": 1.0142791187161937e-05, + "loss": 0.9528, + "step": 41459 + }, + { + "epoch": 3.06, + "learning_rate": 1.0142392356094118e-05, + "loss": 1.0114, + "step": 41460 + }, + { + "epoch": 3.06, + "learning_rate": 1.0141993524799759e-05, + "loss": 0.9889, + "step": 41461 + }, + { + "epoch": 3.06, + "learning_rate": 1.0141594693279484e-05, + "loss": 1.0726, + "step": 41462 + }, + { + "epoch": 3.06, + "learning_rate": 1.014119586153394e-05, + "loss": 1.1076, + "step": 41463 + }, + { + "epoch": 3.06, + "learning_rate": 1.0140797029563752e-05, + "loss": 0.9644, + "step": 41464 + }, + { + "epoch": 3.06, + "learning_rate": 1.014039819736956e-05, + "loss": 0.9599, + "step": 41465 + }, + { + "epoch": 3.06, + "learning_rate": 1.0139999364951994e-05, + "loss": 0.9028, + "step": 41466 + }, + { + "epoch": 3.06, + "learning_rate": 1.0139600532311693e-05, + "loss": 0.9599, + "step": 41467 + }, + { + "epoch": 3.06, + "learning_rate": 1.0139201699449288e-05, + "loss": 0.9928, + "step": 41468 + }, + { + "epoch": 3.06, + "learning_rate": 1.0138802866365416e-05, + "loss": 0.9777, + "step": 41469 + }, + { + "epoch": 3.06, + "learning_rate": 1.0138404033060711e-05, + "loss": 0.9628, + "step": 41470 + }, + { + "epoch": 3.06, + "learning_rate": 1.0138005199535806e-05, + "loss": 1.0253, + "step": 41471 + }, + { + "epoch": 3.06, + "learning_rate": 1.013760636579134e-05, + "loss": 0.991, + "step": 41472 + }, + { + "epoch": 3.06, + "learning_rate": 1.013720753182794e-05, + "loss": 0.9582, + "step": 41473 + }, + { + "epoch": 3.06, + "learning_rate": 1.0136808697646245e-05, + "loss": 1.1115, + "step": 41474 + }, + { + "epoch": 3.06, + "learning_rate": 1.0136409863246892e-05, + "loss": 0.9951, + "step": 41475 + }, + { + "epoch": 3.06, + "learning_rate": 1.0136011028630514e-05, + "loss": 1.0524, + "step": 41476 + }, + { + "epoch": 3.06, + "learning_rate": 1.0135612193797741e-05, + "loss": 0.952, + "step": 41477 + }, + { + "epoch": 3.06, + "learning_rate": 1.0135213358749212e-05, + "loss": 1.0426, + "step": 41478 + }, + { + "epoch": 3.06, + "learning_rate": 1.0134814523485558e-05, + "loss": 0.9447, + "step": 41479 + }, + { + "epoch": 3.06, + "learning_rate": 1.0134415688007423e-05, + "loss": 1.0249, + "step": 41480 + }, + { + "epoch": 3.07, + "learning_rate": 1.0134016852315426e-05, + "loss": 1.0603, + "step": 41481 + }, + { + "epoch": 3.07, + "learning_rate": 1.0133618016410216e-05, + "loss": 1.0013, + "step": 41482 + }, + { + "epoch": 3.07, + "learning_rate": 1.013321918029242e-05, + "loss": 0.9514, + "step": 41483 + }, + { + "epoch": 3.07, + "learning_rate": 1.0132820343962675e-05, + "loss": 0.9812, + "step": 41484 + }, + { + "epoch": 3.07, + "learning_rate": 1.013242150742161e-05, + "loss": 0.9356, + "step": 41485 + }, + { + "epoch": 3.07, + "learning_rate": 1.0132022670669872e-05, + "loss": 1.1137, + "step": 41486 + }, + { + "epoch": 3.07, + "learning_rate": 1.0131623833708085e-05, + "loss": 1.0821, + "step": 41487 + }, + { + "epoch": 3.07, + "learning_rate": 1.0131224996536884e-05, + "loss": 0.9633, + "step": 41488 + }, + { + "epoch": 3.07, + "learning_rate": 1.0130826159156905e-05, + "loss": 0.9344, + "step": 41489 + }, + { + "epoch": 3.07, + "learning_rate": 1.0130427321568787e-05, + "loss": 1.0385, + "step": 41490 + }, + { + "epoch": 3.07, + "learning_rate": 1.013002848377316e-05, + "loss": 0.9305, + "step": 41491 + }, + { + "epoch": 3.07, + "learning_rate": 1.012962964577066e-05, + "loss": 0.942, + "step": 41492 + }, + { + "epoch": 3.07, + "learning_rate": 1.0129230807561918e-05, + "loss": 0.9702, + "step": 41493 + }, + { + "epoch": 3.07, + "learning_rate": 1.0128831969147575e-05, + "loss": 0.9047, + "step": 41494 + }, + { + "epoch": 3.07, + "learning_rate": 1.0128433130528261e-05, + "loss": 0.9387, + "step": 41495 + }, + { + "epoch": 3.07, + "learning_rate": 1.0128034291704613e-05, + "loss": 0.9652, + "step": 41496 + }, + { + "epoch": 3.07, + "learning_rate": 1.0127635452677262e-05, + "loss": 0.9775, + "step": 41497 + }, + { + "epoch": 3.07, + "learning_rate": 1.0127236613446846e-05, + "loss": 1.0264, + "step": 41498 + }, + { + "epoch": 3.07, + "learning_rate": 1.0126837774013998e-05, + "loss": 1.0838, + "step": 41499 + }, + { + "epoch": 3.07, + "learning_rate": 1.0126438934379355e-05, + "loss": 1.015, + "step": 41500 + }, + { + "epoch": 3.07, + "learning_rate": 1.0126040094543546e-05, + "loss": 1.0013, + "step": 41501 + }, + { + "epoch": 3.07, + "learning_rate": 1.0125641254507211e-05, + "loss": 1.0058, + "step": 41502 + }, + { + "epoch": 3.07, + "learning_rate": 1.0125242414270981e-05, + "loss": 0.9683, + "step": 41503 + }, + { + "epoch": 3.07, + "learning_rate": 1.0124843573835497e-05, + "loss": 1.0795, + "step": 41504 + }, + { + "epoch": 3.07, + "learning_rate": 1.0124444733201383e-05, + "loss": 0.9718, + "step": 41505 + }, + { + "epoch": 3.07, + "learning_rate": 1.0124045892369283e-05, + "loss": 1.0488, + "step": 41506 + }, + { + "epoch": 3.07, + "learning_rate": 1.0123647051339825e-05, + "loss": 1.0817, + "step": 41507 + }, + { + "epoch": 3.07, + "learning_rate": 1.0123248210113647e-05, + "loss": 1.0064, + "step": 41508 + }, + { + "epoch": 3.07, + "learning_rate": 1.0122849368691383e-05, + "loss": 0.9561, + "step": 41509 + }, + { + "epoch": 3.07, + "learning_rate": 1.0122450527073668e-05, + "loss": 1.0159, + "step": 41510 + }, + { + "epoch": 3.07, + "learning_rate": 1.0122051685261133e-05, + "loss": 1.0201, + "step": 41511 + }, + { + "epoch": 3.07, + "learning_rate": 1.012165284325442e-05, + "loss": 0.9633, + "step": 41512 + }, + { + "epoch": 3.07, + "learning_rate": 1.0121254001054155e-05, + "loss": 0.9534, + "step": 41513 + }, + { + "epoch": 3.07, + "learning_rate": 1.012085515866098e-05, + "loss": 1.0738, + "step": 41514 + }, + { + "epoch": 3.07, + "learning_rate": 1.0120456316075527e-05, + "loss": 1.02, + "step": 41515 + }, + { + "epoch": 3.07, + "learning_rate": 1.0120057473298425e-05, + "loss": 1.0226, + "step": 41516 + }, + { + "epoch": 3.07, + "learning_rate": 1.0119658630330317e-05, + "loss": 1.051, + "step": 41517 + }, + { + "epoch": 3.07, + "learning_rate": 1.011925978717183e-05, + "loss": 1.1397, + "step": 41518 + }, + { + "epoch": 3.07, + "learning_rate": 1.0118860943823607e-05, + "loss": 1.0528, + "step": 41519 + }, + { + "epoch": 3.07, + "learning_rate": 1.0118462100286278e-05, + "loss": 1.1052, + "step": 41520 + }, + { + "epoch": 3.07, + "learning_rate": 1.0118063256560474e-05, + "loss": 1.0645, + "step": 41521 + }, + { + "epoch": 3.07, + "learning_rate": 1.0117664412646835e-05, + "loss": 1.072, + "step": 41522 + }, + { + "epoch": 3.07, + "learning_rate": 1.0117265568545995e-05, + "loss": 1.0559, + "step": 41523 + }, + { + "epoch": 3.07, + "learning_rate": 1.0116866724258583e-05, + "loss": 1.0065, + "step": 41524 + }, + { + "epoch": 3.07, + "learning_rate": 1.0116467879785242e-05, + "loss": 1.0865, + "step": 41525 + }, + { + "epoch": 3.07, + "learning_rate": 1.0116069035126598e-05, + "loss": 0.9648, + "step": 41526 + }, + { + "epoch": 3.07, + "learning_rate": 1.0115670190283294e-05, + "loss": 1.0339, + "step": 41527 + }, + { + "epoch": 3.07, + "learning_rate": 1.0115271345255957e-05, + "loss": 1.0188, + "step": 41528 + }, + { + "epoch": 3.07, + "learning_rate": 1.0114872500045228e-05, + "loss": 1.0682, + "step": 41529 + }, + { + "epoch": 3.07, + "learning_rate": 1.0114473654651735e-05, + "loss": 0.9293, + "step": 41530 + }, + { + "epoch": 3.07, + "learning_rate": 1.0114074809076118e-05, + "loss": 1.0667, + "step": 41531 + }, + { + "epoch": 3.07, + "learning_rate": 1.0113675963319012e-05, + "loss": 0.9294, + "step": 41532 + }, + { + "epoch": 3.07, + "learning_rate": 1.0113277117381046e-05, + "loss": 0.9394, + "step": 41533 + }, + { + "epoch": 3.07, + "learning_rate": 1.0112878271262858e-05, + "loss": 1.0321, + "step": 41534 + }, + { + "epoch": 3.07, + "learning_rate": 1.0112479424965084e-05, + "loss": 1.083, + "step": 41535 + }, + { + "epoch": 3.07, + "learning_rate": 1.0112080578488354e-05, + "loss": 0.97, + "step": 41536 + }, + { + "epoch": 3.07, + "learning_rate": 1.011168173183331e-05, + "loss": 0.9956, + "step": 41537 + }, + { + "epoch": 3.07, + "learning_rate": 1.0111282885000576e-05, + "loss": 0.9435, + "step": 41538 + }, + { + "epoch": 3.07, + "learning_rate": 1.0110884037990796e-05, + "loss": 0.9776, + "step": 41539 + }, + { + "epoch": 3.07, + "learning_rate": 1.01104851908046e-05, + "loss": 1.0182, + "step": 41540 + }, + { + "epoch": 3.07, + "learning_rate": 1.0110086343442624e-05, + "loss": 0.9296, + "step": 41541 + }, + { + "epoch": 3.07, + "learning_rate": 1.0109687495905505e-05, + "loss": 0.9978, + "step": 41542 + }, + { + "epoch": 3.07, + "learning_rate": 1.0109288648193874e-05, + "loss": 0.9846, + "step": 41543 + }, + { + "epoch": 3.07, + "learning_rate": 1.010888980030836e-05, + "loss": 1.0013, + "step": 41544 + }, + { + "epoch": 3.07, + "learning_rate": 1.0108490952249612e-05, + "loss": 0.9776, + "step": 41545 + }, + { + "epoch": 3.07, + "learning_rate": 1.010809210401825e-05, + "loss": 1.0244, + "step": 41546 + }, + { + "epoch": 3.07, + "learning_rate": 1.0107693255614918e-05, + "loss": 0.9915, + "step": 41547 + }, + { + "epoch": 3.07, + "learning_rate": 1.0107294407040247e-05, + "loss": 0.8504, + "step": 41548 + }, + { + "epoch": 3.07, + "learning_rate": 1.0106895558294873e-05, + "loss": 0.8526, + "step": 41549 + }, + { + "epoch": 3.07, + "learning_rate": 1.010649670937943e-05, + "loss": 0.9301, + "step": 41550 + }, + { + "epoch": 3.07, + "learning_rate": 1.0106097860294554e-05, + "loss": 1.0202, + "step": 41551 + }, + { + "epoch": 3.07, + "learning_rate": 1.0105699011040874e-05, + "loss": 0.9974, + "step": 41552 + }, + { + "epoch": 3.07, + "learning_rate": 1.010530016161903e-05, + "loss": 1.0513, + "step": 41553 + }, + { + "epoch": 3.07, + "learning_rate": 1.0104901312029653e-05, + "loss": 1.0032, + "step": 41554 + }, + { + "epoch": 3.07, + "learning_rate": 1.0104502462273383e-05, + "loss": 1.0209, + "step": 41555 + }, + { + "epoch": 3.07, + "learning_rate": 1.0104103612350848e-05, + "loss": 0.9852, + "step": 41556 + }, + { + "epoch": 3.07, + "learning_rate": 1.0103704762262687e-05, + "loss": 0.8801, + "step": 41557 + }, + { + "epoch": 3.07, + "learning_rate": 1.0103305912009533e-05, + "loss": 0.9894, + "step": 41558 + }, + { + "epoch": 3.07, + "learning_rate": 1.0102907061592021e-05, + "loss": 0.9895, + "step": 41559 + }, + { + "epoch": 3.07, + "learning_rate": 1.0102508211010787e-05, + "loss": 1.0445, + "step": 41560 + }, + { + "epoch": 3.07, + "learning_rate": 1.010210936026646e-05, + "loss": 1.0311, + "step": 41561 + }, + { + "epoch": 3.07, + "learning_rate": 1.0101710509359684e-05, + "loss": 1.0099, + "step": 41562 + }, + { + "epoch": 3.07, + "learning_rate": 1.0101311658291084e-05, + "loss": 0.9348, + "step": 41563 + }, + { + "epoch": 3.07, + "learning_rate": 1.0100912807061298e-05, + "loss": 1.0973, + "step": 41564 + }, + { + "epoch": 3.07, + "learning_rate": 1.0100513955670964e-05, + "loss": 0.8891, + "step": 41565 + }, + { + "epoch": 3.07, + "learning_rate": 1.0100115104120711e-05, + "loss": 1.009, + "step": 41566 + }, + { + "epoch": 3.07, + "learning_rate": 1.0099716252411178e-05, + "loss": 1.0195, + "step": 41567 + }, + { + "epoch": 3.07, + "learning_rate": 1.0099317400542997e-05, + "loss": 1.0017, + "step": 41568 + }, + { + "epoch": 3.07, + "learning_rate": 1.0098918548516804e-05, + "loss": 0.9499, + "step": 41569 + }, + { + "epoch": 3.07, + "learning_rate": 1.0098519696333233e-05, + "loss": 0.9122, + "step": 41570 + }, + { + "epoch": 3.07, + "learning_rate": 1.0098120843992918e-05, + "loss": 1.0417, + "step": 41571 + }, + { + "epoch": 3.07, + "learning_rate": 1.0097721991496494e-05, + "loss": 1.0529, + "step": 41572 + }, + { + "epoch": 3.07, + "learning_rate": 1.0097323138844594e-05, + "loss": 0.9642, + "step": 41573 + }, + { + "epoch": 3.07, + "learning_rate": 1.009692428603786e-05, + "loss": 1.1516, + "step": 41574 + }, + { + "epoch": 3.07, + "learning_rate": 1.0096525433076917e-05, + "loss": 0.96, + "step": 41575 + }, + { + "epoch": 3.07, + "learning_rate": 1.0096126579962403e-05, + "loss": 0.9504, + "step": 41576 + }, + { + "epoch": 3.07, + "learning_rate": 1.0095727726694952e-05, + "loss": 0.9509, + "step": 41577 + }, + { + "epoch": 3.07, + "learning_rate": 1.0095328873275203e-05, + "loss": 0.891, + "step": 41578 + }, + { + "epoch": 3.07, + "learning_rate": 1.0094930019703786e-05, + "loss": 1.0382, + "step": 41579 + }, + { + "epoch": 3.07, + "learning_rate": 1.0094531165981337e-05, + "loss": 0.9488, + "step": 41580 + }, + { + "epoch": 3.07, + "learning_rate": 1.009413231210849e-05, + "loss": 1.0042, + "step": 41581 + }, + { + "epoch": 3.07, + "learning_rate": 1.0093733458085877e-05, + "loss": 0.9067, + "step": 41582 + }, + { + "epoch": 3.07, + "learning_rate": 1.0093334603914137e-05, + "loss": 1.0342, + "step": 41583 + }, + { + "epoch": 3.07, + "learning_rate": 1.0092935749593904e-05, + "loss": 1.0609, + "step": 41584 + }, + { + "epoch": 3.07, + "learning_rate": 1.0092536895125812e-05, + "loss": 0.9976, + "step": 41585 + }, + { + "epoch": 3.07, + "learning_rate": 1.0092138040510493e-05, + "loss": 1.0822, + "step": 41586 + }, + { + "epoch": 3.07, + "learning_rate": 1.0091739185748586e-05, + "loss": 1.0536, + "step": 41587 + }, + { + "epoch": 3.07, + "learning_rate": 1.0091340330840724e-05, + "loss": 0.9945, + "step": 41588 + }, + { + "epoch": 3.07, + "learning_rate": 1.0090941475787538e-05, + "loss": 1.0128, + "step": 41589 + }, + { + "epoch": 3.07, + "learning_rate": 1.0090542620589667e-05, + "loss": 1.0105, + "step": 41590 + }, + { + "epoch": 3.07, + "learning_rate": 1.0090143765247741e-05, + "loss": 1.1457, + "step": 41591 + }, + { + "epoch": 3.07, + "learning_rate": 1.0089744909762403e-05, + "loss": 1.0489, + "step": 41592 + }, + { + "epoch": 3.07, + "learning_rate": 1.0089346054134276e-05, + "loss": 0.9663, + "step": 41593 + }, + { + "epoch": 3.07, + "learning_rate": 1.0088947198364006e-05, + "loss": 0.9987, + "step": 41594 + }, + { + "epoch": 3.07, + "learning_rate": 1.008854834245222e-05, + "loss": 1.0247, + "step": 41595 + }, + { + "epoch": 3.07, + "learning_rate": 1.0088149486399554e-05, + "loss": 1.0779, + "step": 41596 + }, + { + "epoch": 3.07, + "learning_rate": 1.0087750630206645e-05, + "loss": 1.0793, + "step": 41597 + }, + { + "epoch": 3.07, + "learning_rate": 1.0087351773874127e-05, + "loss": 0.9537, + "step": 41598 + }, + { + "epoch": 3.07, + "learning_rate": 1.0086952917402632e-05, + "loss": 0.9535, + "step": 41599 + }, + { + "epoch": 3.07, + "learning_rate": 1.0086554060792796e-05, + "loss": 0.9994, + "step": 41600 + }, + { + "epoch": 3.07, + "learning_rate": 1.0086155204045253e-05, + "loss": 1.0139, + "step": 41601 + }, + { + "epoch": 3.07, + "learning_rate": 1.008575634716064e-05, + "loss": 1.0095, + "step": 41602 + }, + { + "epoch": 3.07, + "learning_rate": 1.0085357490139588e-05, + "loss": 1.0169, + "step": 41603 + }, + { + "epoch": 3.07, + "learning_rate": 1.0084958632982735e-05, + "loss": 1.0699, + "step": 41604 + }, + { + "epoch": 3.07, + "learning_rate": 1.0084559775690712e-05, + "loss": 1.0108, + "step": 41605 + }, + { + "epoch": 3.07, + "learning_rate": 1.0084160918264159e-05, + "loss": 1.0645, + "step": 41606 + }, + { + "epoch": 3.07, + "learning_rate": 1.0083762060703706e-05, + "loss": 1.0199, + "step": 41607 + }, + { + "epoch": 3.07, + "learning_rate": 1.008336320300999e-05, + "loss": 1.0548, + "step": 41608 + }, + { + "epoch": 3.07, + "learning_rate": 1.008296434518364e-05, + "loss": 1.1867, + "step": 41609 + }, + { + "epoch": 3.07, + "learning_rate": 1.00825654872253e-05, + "loss": 0.9326, + "step": 41610 + }, + { + "epoch": 3.07, + "learning_rate": 1.0082166629135595e-05, + "loss": 1.0599, + "step": 41611 + }, + { + "epoch": 3.07, + "learning_rate": 1.0081767770915167e-05, + "loss": 1.0356, + "step": 41612 + }, + { + "epoch": 3.07, + "learning_rate": 1.0081368912564644e-05, + "loss": 0.9674, + "step": 41613 + }, + { + "epoch": 3.07, + "learning_rate": 1.0080970054084668e-05, + "loss": 1.0159, + "step": 41614 + }, + { + "epoch": 3.07, + "learning_rate": 1.0080571195475869e-05, + "loss": 1.0179, + "step": 41615 + }, + { + "epoch": 3.07, + "learning_rate": 1.0080172336738882e-05, + "loss": 1.0761, + "step": 41616 + }, + { + "epoch": 3.08, + "learning_rate": 1.0079773477874341e-05, + "loss": 0.9779, + "step": 41617 + }, + { + "epoch": 3.08, + "learning_rate": 1.0079374618882882e-05, + "loss": 1.046, + "step": 41618 + }, + { + "epoch": 3.08, + "learning_rate": 1.0078975759765141e-05, + "loss": 1.0084, + "step": 41619 + }, + { + "epoch": 3.08, + "learning_rate": 1.0078576900521748e-05, + "loss": 1.0406, + "step": 41620 + }, + { + "epoch": 3.08, + "learning_rate": 1.0078178041153344e-05, + "loss": 0.9248, + "step": 41621 + }, + { + "epoch": 3.08, + "learning_rate": 1.0077779181660554e-05, + "loss": 1.023, + "step": 41622 + }, + { + "epoch": 3.08, + "learning_rate": 1.0077380322044024e-05, + "loss": 1.0406, + "step": 41623 + }, + { + "epoch": 3.08, + "learning_rate": 1.0076981462304382e-05, + "loss": 1.0431, + "step": 41624 + }, + { + "epoch": 3.08, + "learning_rate": 1.007658260244226e-05, + "loss": 0.9745, + "step": 41625 + }, + { + "epoch": 3.08, + "learning_rate": 1.00761837424583e-05, + "loss": 0.9729, + "step": 41626 + }, + { + "epoch": 3.08, + "learning_rate": 1.0075784882353132e-05, + "loss": 1.1258, + "step": 41627 + }, + { + "epoch": 3.08, + "learning_rate": 1.0075386022127389e-05, + "loss": 1.0355, + "step": 41628 + }, + { + "epoch": 3.08, + "learning_rate": 1.0074987161781711e-05, + "loss": 1.0675, + "step": 41629 + }, + { + "epoch": 3.08, + "learning_rate": 1.0074588301316728e-05, + "loss": 1.0189, + "step": 41630 + }, + { + "epoch": 3.08, + "learning_rate": 1.0074189440733078e-05, + "loss": 0.9613, + "step": 41631 + }, + { + "epoch": 3.08, + "learning_rate": 1.0073790580031389e-05, + "loss": 1.0037, + "step": 41632 + }, + { + "epoch": 3.08, + "learning_rate": 1.0073391719212304e-05, + "loss": 1.0534, + "step": 41633 + }, + { + "epoch": 3.08, + "learning_rate": 1.0072992858276452e-05, + "loss": 0.9398, + "step": 41634 + }, + { + "epoch": 3.08, + "learning_rate": 1.0072593997224474e-05, + "loss": 0.8931, + "step": 41635 + }, + { + "epoch": 3.08, + "learning_rate": 1.0072195136056994e-05, + "loss": 0.9761, + "step": 41636 + }, + { + "epoch": 3.08, + "learning_rate": 1.0071796274774657e-05, + "loss": 1.0586, + "step": 41637 + }, + { + "epoch": 3.08, + "learning_rate": 1.007139741337809e-05, + "loss": 1.0093, + "step": 41638 + }, + { + "epoch": 3.08, + "learning_rate": 1.0070998551867932e-05, + "loss": 1.063, + "step": 41639 + }, + { + "epoch": 3.08, + "learning_rate": 1.0070599690244816e-05, + "loss": 1.0122, + "step": 41640 + }, + { + "epoch": 3.08, + "learning_rate": 1.0070200828509377e-05, + "loss": 0.9998, + "step": 41641 + }, + { + "epoch": 3.08, + "learning_rate": 1.0069801966662249e-05, + "loss": 0.9607, + "step": 41642 + }, + { + "epoch": 3.08, + "learning_rate": 1.0069403104704067e-05, + "loss": 0.9556, + "step": 41643 + }, + { + "epoch": 3.08, + "learning_rate": 1.0069004242635465e-05, + "loss": 0.968, + "step": 41644 + }, + { + "epoch": 3.08, + "learning_rate": 1.0068605380457084e-05, + "loss": 0.9808, + "step": 41645 + }, + { + "epoch": 3.08, + "learning_rate": 1.0068206518169545e-05, + "loss": 1.0886, + "step": 41646 + }, + { + "epoch": 3.08, + "learning_rate": 1.0067807655773497e-05, + "loss": 1.0785, + "step": 41647 + }, + { + "epoch": 3.08, + "learning_rate": 1.0067408793269562e-05, + "loss": 1.0345, + "step": 41648 + }, + { + "epoch": 3.08, + "learning_rate": 1.0067009930658384e-05, + "loss": 1.0903, + "step": 41649 + }, + { + "epoch": 3.08, + "learning_rate": 1.0066611067940592e-05, + "loss": 1.0491, + "step": 41650 + }, + { + "epoch": 3.08, + "learning_rate": 1.0066212205116826e-05, + "loss": 1.1174, + "step": 41651 + }, + { + "epoch": 3.08, + "learning_rate": 1.0065813342187714e-05, + "loss": 0.9638, + "step": 41652 + }, + { + "epoch": 3.08, + "learning_rate": 1.0065414479153897e-05, + "loss": 1.0358, + "step": 41653 + }, + { + "epoch": 3.08, + "learning_rate": 1.0065015616016006e-05, + "loss": 0.9196, + "step": 41654 + }, + { + "epoch": 3.08, + "learning_rate": 1.0064616752774676e-05, + "loss": 0.983, + "step": 41655 + }, + { + "epoch": 3.08, + "learning_rate": 1.0064217889430539e-05, + "loss": 1.1049, + "step": 41656 + }, + { + "epoch": 3.08, + "learning_rate": 1.0063819025984234e-05, + "loss": 1.0086, + "step": 41657 + }, + { + "epoch": 3.08, + "learning_rate": 1.0063420162436393e-05, + "loss": 0.9989, + "step": 41658 + }, + { + "epoch": 3.08, + "learning_rate": 1.0063021298787654e-05, + "loss": 0.9666, + "step": 41659 + }, + { + "epoch": 3.08, + "learning_rate": 1.0062622435038645e-05, + "loss": 1.0587, + "step": 41660 + }, + { + "epoch": 3.08, + "learning_rate": 1.0062223571190008e-05, + "loss": 1.0506, + "step": 41661 + }, + { + "epoch": 3.08, + "learning_rate": 1.0061824707242372e-05, + "loss": 0.9479, + "step": 41662 + }, + { + "epoch": 3.08, + "learning_rate": 1.0061425843196376e-05, + "loss": 0.9304, + "step": 41663 + }, + { + "epoch": 3.08, + "learning_rate": 1.0061026979052653e-05, + "loss": 1.0716, + "step": 41664 + }, + { + "epoch": 3.08, + "learning_rate": 1.0060628114811833e-05, + "loss": 1.0779, + "step": 41665 + }, + { + "epoch": 3.08, + "learning_rate": 1.0060229250474558e-05, + "loss": 0.9685, + "step": 41666 + }, + { + "epoch": 3.08, + "learning_rate": 1.0059830386041456e-05, + "loss": 0.9229, + "step": 41667 + }, + { + "epoch": 3.08, + "learning_rate": 1.0059431521513168e-05, + "loss": 1.0899, + "step": 41668 + }, + { + "epoch": 3.08, + "learning_rate": 1.0059032656890324e-05, + "loss": 1.0477, + "step": 41669 + }, + { + "epoch": 3.08, + "learning_rate": 1.0058633792173558e-05, + "loss": 0.9747, + "step": 41670 + }, + { + "epoch": 3.08, + "learning_rate": 1.0058234927363508e-05, + "loss": 0.9131, + "step": 41671 + }, + { + "epoch": 3.08, + "learning_rate": 1.0057836062460806e-05, + "loss": 1.1756, + "step": 41672 + }, + { + "epoch": 3.08, + "learning_rate": 1.0057437197466092e-05, + "loss": 0.9716, + "step": 41673 + }, + { + "epoch": 3.08, + "learning_rate": 1.0057038332379994e-05, + "loss": 0.9971, + "step": 41674 + }, + { + "epoch": 3.08, + "learning_rate": 1.0056639467203146e-05, + "loss": 1.0663, + "step": 41675 + }, + { + "epoch": 3.08, + "learning_rate": 1.0056240601936186e-05, + "loss": 0.9688, + "step": 41676 + }, + { + "epoch": 3.08, + "learning_rate": 1.005584173657975e-05, + "loss": 0.9521, + "step": 41677 + }, + { + "epoch": 3.08, + "learning_rate": 1.005544287113447e-05, + "loss": 0.9989, + "step": 41678 + }, + { + "epoch": 3.08, + "learning_rate": 1.005504400560098e-05, + "loss": 0.9856, + "step": 41679 + }, + { + "epoch": 3.08, + "learning_rate": 1.005464513997992e-05, + "loss": 0.9676, + "step": 41680 + }, + { + "epoch": 3.08, + "learning_rate": 1.0054246274271915e-05, + "loss": 0.9865, + "step": 41681 + }, + { + "epoch": 3.08, + "learning_rate": 1.005384740847761e-05, + "loss": 1.0922, + "step": 41682 + }, + { + "epoch": 3.08, + "learning_rate": 1.005344854259763e-05, + "loss": 0.98, + "step": 41683 + }, + { + "epoch": 3.08, + "learning_rate": 1.0053049676632618e-05, + "loss": 0.9692, + "step": 41684 + }, + { + "epoch": 3.08, + "learning_rate": 1.00526508105832e-05, + "loss": 0.9593, + "step": 41685 + }, + { + "epoch": 3.08, + "learning_rate": 1.005225194445002e-05, + "loss": 0.8905, + "step": 41686 + }, + { + "epoch": 3.08, + "learning_rate": 1.0051853078233704e-05, + "loss": 0.8468, + "step": 41687 + }, + { + "epoch": 3.08, + "learning_rate": 1.0051454211934892e-05, + "loss": 0.9689, + "step": 41688 + }, + { + "epoch": 3.08, + "learning_rate": 1.0051055345554217e-05, + "loss": 1.0712, + "step": 41689 + }, + { + "epoch": 3.08, + "learning_rate": 1.0050656479092316e-05, + "loss": 1.0703, + "step": 41690 + }, + { + "epoch": 3.08, + "learning_rate": 1.0050257612549821e-05, + "loss": 0.9786, + "step": 41691 + }, + { + "epoch": 3.08, + "learning_rate": 1.0049858745927366e-05, + "loss": 0.9996, + "step": 41692 + }, + { + "epoch": 3.08, + "learning_rate": 1.0049459879225583e-05, + "loss": 0.9814, + "step": 41693 + }, + { + "epoch": 3.08, + "learning_rate": 1.0049061012445115e-05, + "loss": 0.9576, + "step": 41694 + }, + { + "epoch": 3.08, + "learning_rate": 1.0048662145586587e-05, + "loss": 1.0512, + "step": 41695 + }, + { + "epoch": 3.08, + "learning_rate": 1.004826327865064e-05, + "loss": 1.0281, + "step": 41696 + }, + { + "epoch": 3.08, + "learning_rate": 1.0047864411637906e-05, + "loss": 0.9414, + "step": 41697 + }, + { + "epoch": 3.08, + "learning_rate": 1.0047465544549024e-05, + "loss": 0.9982, + "step": 41698 + }, + { + "epoch": 3.08, + "learning_rate": 1.0047066677384622e-05, + "loss": 0.9817, + "step": 41699 + }, + { + "epoch": 3.08, + "learning_rate": 1.0046667810145338e-05, + "loss": 1.0035, + "step": 41700 + }, + { + "epoch": 3.08, + "learning_rate": 1.004626894283181e-05, + "loss": 0.9183, + "step": 41701 + }, + { + "epoch": 3.08, + "learning_rate": 1.0045870075444667e-05, + "loss": 1.0528, + "step": 41702 + }, + { + "epoch": 3.08, + "learning_rate": 1.004547120798454e-05, + "loss": 0.9206, + "step": 41703 + }, + { + "epoch": 3.08, + "learning_rate": 1.0045072340452074e-05, + "loss": 1.0813, + "step": 41704 + }, + { + "epoch": 3.08, + "learning_rate": 1.0044673472847896e-05, + "loss": 0.9488, + "step": 41705 + }, + { + "epoch": 3.08, + "learning_rate": 1.0044274605172645e-05, + "loss": 1.0053, + "step": 41706 + }, + { + "epoch": 3.08, + "learning_rate": 1.0043875737426952e-05, + "loss": 0.9249, + "step": 41707 + }, + { + "epoch": 3.08, + "learning_rate": 1.0043476869611455e-05, + "loss": 0.9827, + "step": 41708 + }, + { + "epoch": 3.08, + "learning_rate": 1.0043078001726785e-05, + "loss": 1.0362, + "step": 41709 + }, + { + "epoch": 3.08, + "learning_rate": 1.004267913377358e-05, + "loss": 1.1113, + "step": 41710 + }, + { + "epoch": 3.08, + "learning_rate": 1.0042280265752475e-05, + "loss": 0.9714, + "step": 41711 + }, + { + "epoch": 3.08, + "learning_rate": 1.0041881397664102e-05, + "loss": 1.0397, + "step": 41712 + }, + { + "epoch": 3.08, + "learning_rate": 1.0041482529509092e-05, + "loss": 0.915, + "step": 41713 + }, + { + "epoch": 3.08, + "learning_rate": 1.004108366128809e-05, + "loss": 0.9701, + "step": 41714 + }, + { + "epoch": 3.08, + "learning_rate": 1.0040684793001718e-05, + "loss": 1.019, + "step": 41715 + }, + { + "epoch": 3.08, + "learning_rate": 1.0040285924650623e-05, + "loss": 0.9658, + "step": 41716 + }, + { + "epoch": 3.08, + "learning_rate": 1.0039887056235428e-05, + "loss": 1.0276, + "step": 41717 + }, + { + "epoch": 3.08, + "learning_rate": 1.0039488187756779e-05, + "loss": 0.9013, + "step": 41718 + }, + { + "epoch": 3.08, + "learning_rate": 1.0039089319215303e-05, + "loss": 1.0147, + "step": 41719 + }, + { + "epoch": 3.08, + "learning_rate": 1.0038690450611635e-05, + "loss": 0.9192, + "step": 41720 + }, + { + "epoch": 3.08, + "learning_rate": 1.0038291581946411e-05, + "loss": 0.9446, + "step": 41721 + }, + { + "epoch": 3.08, + "learning_rate": 1.0037892713220265e-05, + "loss": 1.0012, + "step": 41722 + }, + { + "epoch": 3.08, + "learning_rate": 1.0037493844433834e-05, + "loss": 1.0438, + "step": 41723 + }, + { + "epoch": 3.08, + "learning_rate": 1.003709497558775e-05, + "loss": 0.9766, + "step": 41724 + }, + { + "epoch": 3.08, + "learning_rate": 1.003669610668265e-05, + "loss": 1.0993, + "step": 41725 + }, + { + "epoch": 3.08, + "learning_rate": 1.0036297237719164e-05, + "loss": 1.0591, + "step": 41726 + }, + { + "epoch": 3.08, + "learning_rate": 1.0035898368697934e-05, + "loss": 0.962, + "step": 41727 + }, + { + "epoch": 3.08, + "learning_rate": 1.0035499499619585e-05, + "loss": 1.0906, + "step": 41728 + }, + { + "epoch": 3.08, + "learning_rate": 1.0035100630484763e-05, + "loss": 0.9646, + "step": 41729 + }, + { + "epoch": 3.08, + "learning_rate": 1.0034701761294092e-05, + "loss": 1.0189, + "step": 41730 + }, + { + "epoch": 3.08, + "learning_rate": 1.0034302892048214e-05, + "loss": 0.9378, + "step": 41731 + }, + { + "epoch": 3.08, + "learning_rate": 1.0033904022747756e-05, + "loss": 1.0183, + "step": 41732 + }, + { + "epoch": 3.08, + "learning_rate": 1.0033505153393359e-05, + "loss": 0.9888, + "step": 41733 + }, + { + "epoch": 3.08, + "learning_rate": 1.0033106283985656e-05, + "loss": 0.9534, + "step": 41734 + }, + { + "epoch": 3.08, + "learning_rate": 1.0032707414525282e-05, + "loss": 0.9759, + "step": 41735 + }, + { + "epoch": 3.08, + "learning_rate": 1.003230854501287e-05, + "loss": 1.0881, + "step": 41736 + }, + { + "epoch": 3.08, + "learning_rate": 1.0031909675449059e-05, + "loss": 0.9612, + "step": 41737 + }, + { + "epoch": 3.08, + "learning_rate": 1.0031510805834477e-05, + "loss": 0.8869, + "step": 41738 + }, + { + "epoch": 3.08, + "learning_rate": 1.0031111936169765e-05, + "loss": 1.0181, + "step": 41739 + }, + { + "epoch": 3.08, + "learning_rate": 1.0030713066455548e-05, + "loss": 0.9939, + "step": 41740 + }, + { + "epoch": 3.08, + "learning_rate": 1.0030314196692473e-05, + "loss": 0.9617, + "step": 41741 + }, + { + "epoch": 3.08, + "learning_rate": 1.0029915326881163e-05, + "loss": 0.9834, + "step": 41742 + }, + { + "epoch": 3.08, + "learning_rate": 1.0029516457022264e-05, + "loss": 0.9709, + "step": 41743 + }, + { + "epoch": 3.08, + "learning_rate": 1.00291175871164e-05, + "loss": 1.0661, + "step": 41744 + }, + { + "epoch": 3.08, + "learning_rate": 1.0028718717164215e-05, + "loss": 1.0571, + "step": 41745 + }, + { + "epoch": 3.08, + "learning_rate": 1.0028319847166332e-05, + "loss": 0.9531, + "step": 41746 + }, + { + "epoch": 3.08, + "learning_rate": 1.0027920977123402e-05, + "loss": 1.0452, + "step": 41747 + }, + { + "epoch": 3.08, + "learning_rate": 1.0027522107036042e-05, + "loss": 0.8971, + "step": 41748 + }, + { + "epoch": 3.08, + "learning_rate": 1.0027123236904899e-05, + "loss": 0.9923, + "step": 41749 + }, + { + "epoch": 3.08, + "learning_rate": 1.00267243667306e-05, + "loss": 1.046, + "step": 41750 + }, + { + "epoch": 3.08, + "learning_rate": 1.0026325496513788e-05, + "loss": 1.1208, + "step": 41751 + }, + { + "epoch": 3.09, + "learning_rate": 1.0025926626255086e-05, + "loss": 1.0723, + "step": 41752 + }, + { + "epoch": 3.09, + "learning_rate": 1.002552775595514e-05, + "loss": 1.0619, + "step": 41753 + }, + { + "epoch": 3.09, + "learning_rate": 1.0025128885614576e-05, + "loss": 0.9511, + "step": 41754 + }, + { + "epoch": 3.09, + "learning_rate": 1.0024730015234036e-05, + "loss": 0.9379, + "step": 41755 + }, + { + "epoch": 3.09, + "learning_rate": 1.0024331144814148e-05, + "loss": 1.0226, + "step": 41756 + }, + { + "epoch": 3.09, + "learning_rate": 1.0023932274355554e-05, + "loss": 0.9865, + "step": 41757 + }, + { + "epoch": 3.09, + "learning_rate": 1.002353340385888e-05, + "loss": 0.9175, + "step": 41758 + }, + { + "epoch": 3.09, + "learning_rate": 1.0023134533324768e-05, + "loss": 1.0815, + "step": 41759 + }, + { + "epoch": 3.09, + "learning_rate": 1.0022735662753844e-05, + "loss": 1.0604, + "step": 41760 + }, + { + "epoch": 3.09, + "learning_rate": 1.0022336792146751e-05, + "loss": 1.1309, + "step": 41761 + }, + { + "epoch": 3.09, + "learning_rate": 1.002193792150412e-05, + "loss": 1.0232, + "step": 41762 + }, + { + "epoch": 3.09, + "learning_rate": 1.0021539050826587e-05, + "loss": 0.9798, + "step": 41763 + }, + { + "epoch": 3.09, + "learning_rate": 1.0021140180114786e-05, + "loss": 1.0161, + "step": 41764 + }, + { + "epoch": 3.09, + "learning_rate": 1.002074130936935e-05, + "loss": 1.0267, + "step": 41765 + }, + { + "epoch": 3.09, + "learning_rate": 1.0020342438590915e-05, + "loss": 1.0699, + "step": 41766 + }, + { + "epoch": 3.09, + "learning_rate": 1.0019943567780115e-05, + "loss": 0.9701, + "step": 41767 + }, + { + "epoch": 3.09, + "learning_rate": 1.0019544696937589e-05, + "loss": 1.0335, + "step": 41768 + }, + { + "epoch": 3.09, + "learning_rate": 1.0019145826063962e-05, + "loss": 0.9972, + "step": 41769 + }, + { + "epoch": 3.09, + "learning_rate": 1.0018746955159876e-05, + "loss": 1.1286, + "step": 41770 + }, + { + "epoch": 3.09, + "learning_rate": 1.0018348084225964e-05, + "loss": 0.985, + "step": 41771 + }, + { + "epoch": 3.09, + "learning_rate": 1.0017949213262864e-05, + "loss": 1.0347, + "step": 41772 + }, + { + "epoch": 3.09, + "learning_rate": 1.0017550342271203e-05, + "loss": 1.084, + "step": 41773 + }, + { + "epoch": 3.09, + "learning_rate": 1.001715147125162e-05, + "loss": 1.062, + "step": 41774 + }, + { + "epoch": 3.09, + "learning_rate": 1.001675260020475e-05, + "loss": 0.9999, + "step": 41775 + }, + { + "epoch": 3.09, + "learning_rate": 1.0016353729131229e-05, + "loss": 1.0285, + "step": 41776 + }, + { + "epoch": 3.09, + "learning_rate": 1.0015954858031686e-05, + "loss": 0.9872, + "step": 41777 + }, + { + "epoch": 3.09, + "learning_rate": 1.0015555986906762e-05, + "loss": 0.9849, + "step": 41778 + }, + { + "epoch": 3.09, + "learning_rate": 1.0015157115757085e-05, + "loss": 0.9492, + "step": 41779 + }, + { + "epoch": 3.09, + "learning_rate": 1.0014758244583296e-05, + "loss": 1.0881, + "step": 41780 + }, + { + "epoch": 3.09, + "learning_rate": 1.0014359373386024e-05, + "loss": 0.9337, + "step": 41781 + }, + { + "epoch": 3.09, + "learning_rate": 1.0013960502165912e-05, + "loss": 0.9451, + "step": 41782 + }, + { + "epoch": 3.09, + "learning_rate": 1.0013561630923584e-05, + "loss": 0.9999, + "step": 41783 + }, + { + "epoch": 3.09, + "learning_rate": 1.001316275965968e-05, + "loss": 0.9079, + "step": 41784 + }, + { + "epoch": 3.09, + "learning_rate": 1.0012763888374837e-05, + "loss": 0.9552, + "step": 41785 + }, + { + "epoch": 3.09, + "learning_rate": 1.0012365017069688e-05, + "loss": 0.985, + "step": 41786 + }, + { + "epoch": 3.09, + "learning_rate": 1.001196614574486e-05, + "loss": 0.9757, + "step": 41787 + }, + { + "epoch": 3.09, + "learning_rate": 1.0011567274401e-05, + "loss": 0.9661, + "step": 41788 + }, + { + "epoch": 3.09, + "learning_rate": 1.0011168403038733e-05, + "loss": 1.0675, + "step": 41789 + }, + { + "epoch": 3.09, + "learning_rate": 1.0010769531658699e-05, + "loss": 1.0049, + "step": 41790 + }, + { + "epoch": 3.09, + "learning_rate": 1.001037066026153e-05, + "loss": 1.0158, + "step": 41791 + }, + { + "epoch": 3.09, + "learning_rate": 1.0009971788847863e-05, + "loss": 1.1097, + "step": 41792 + }, + { + "epoch": 3.09, + "learning_rate": 1.0009572917418329e-05, + "loss": 1.1192, + "step": 41793 + }, + { + "epoch": 3.09, + "learning_rate": 1.0009174045973568e-05, + "loss": 1.0622, + "step": 41794 + }, + { + "epoch": 3.09, + "learning_rate": 1.0008775174514206e-05, + "loss": 1.1528, + "step": 41795 + }, + { + "epoch": 3.09, + "learning_rate": 1.0008376303040887e-05, + "loss": 1.1011, + "step": 41796 + }, + { + "epoch": 3.09, + "learning_rate": 1.0007977431554239e-05, + "loss": 0.9982, + "step": 41797 + }, + { + "epoch": 3.09, + "learning_rate": 1.0007578560054902e-05, + "loss": 1.0269, + "step": 41798 + }, + { + "epoch": 3.09, + "learning_rate": 1.0007179688543505e-05, + "loss": 1.202, + "step": 41799 + }, + { + "epoch": 3.09, + "learning_rate": 1.0006780817020685e-05, + "loss": 0.9595, + "step": 41800 + }, + { + "epoch": 3.09, + "learning_rate": 1.0006381945487077e-05, + "loss": 0.9663, + "step": 41801 + }, + { + "epoch": 3.09, + "learning_rate": 1.0005983073943317e-05, + "loss": 0.9299, + "step": 41802 + }, + { + "epoch": 3.09, + "learning_rate": 1.0005584202390038e-05, + "loss": 1.0099, + "step": 41803 + }, + { + "epoch": 3.09, + "learning_rate": 1.0005185330827875e-05, + "loss": 1.0549, + "step": 41804 + }, + { + "epoch": 3.09, + "learning_rate": 1.0004786459257459e-05, + "loss": 1.0312, + "step": 41805 + }, + { + "epoch": 3.09, + "learning_rate": 1.000438758767943e-05, + "loss": 0.9274, + "step": 41806 + }, + { + "epoch": 3.09, + "learning_rate": 1.000398871609442e-05, + "loss": 1.0198, + "step": 41807 + }, + { + "epoch": 3.09, + "learning_rate": 1.0003589844503065e-05, + "loss": 1.0045, + "step": 41808 + }, + { + "epoch": 3.09, + "learning_rate": 1.0003190972905996e-05, + "loss": 1.0408, + "step": 41809 + }, + { + "epoch": 3.09, + "learning_rate": 1.0002792101303853e-05, + "loss": 1.0058, + "step": 41810 + }, + { + "epoch": 3.09, + "learning_rate": 1.0002393229697267e-05, + "loss": 1.0166, + "step": 41811 + }, + { + "epoch": 3.09, + "learning_rate": 1.0001994358086874e-05, + "loss": 1.0386, + "step": 41812 + }, + { + "epoch": 3.09, + "learning_rate": 1.0001595486473308e-05, + "loss": 0.9568, + "step": 41813 + }, + { + "epoch": 3.09, + "learning_rate": 1.00011966148572e-05, + "loss": 1.0047, + "step": 41814 + }, + { + "epoch": 3.09, + "learning_rate": 1.0000797743239193e-05, + "loss": 1.036, + "step": 41815 + }, + { + "epoch": 3.09, + "learning_rate": 1.0000398871619914e-05, + "loss": 0.9614, + "step": 41816 + }, + { + "epoch": 3.09, + "learning_rate": 1e-05, + "loss": 0.9828, + "step": 41817 + }, + { + "epoch": 3.09, + "learning_rate": 9.99960112838009e-06, + "loss": 0.9459, + "step": 41818 + }, + { + "epoch": 3.09, + "learning_rate": 9.99920225676081e-06, + "loss": 0.9836, + "step": 41819 + }, + { + "epoch": 3.09, + "learning_rate": 9.998803385142803e-06, + "loss": 0.9751, + "step": 41820 + }, + { + "epoch": 3.09, + "learning_rate": 9.998404513526696e-06, + "loss": 1.079, + "step": 41821 + }, + { + "epoch": 3.09, + "learning_rate": 9.99800564191313e-06, + "loss": 0.9987, + "step": 41822 + }, + { + "epoch": 3.09, + "learning_rate": 9.997606770302735e-06, + "loss": 1.0193, + "step": 41823 + }, + { + "epoch": 3.09, + "learning_rate": 9.99720789869615e-06, + "loss": 1.0181, + "step": 41824 + }, + { + "epoch": 3.09, + "learning_rate": 9.996809027094004e-06, + "loss": 0.9954, + "step": 41825 + }, + { + "epoch": 3.09, + "learning_rate": 9.996410155496937e-06, + "loss": 1.0351, + "step": 41826 + }, + { + "epoch": 3.09, + "learning_rate": 9.996011283905585e-06, + "loss": 1.0307, + "step": 41827 + }, + { + "epoch": 3.09, + "learning_rate": 9.995612412320573e-06, + "loss": 1.044, + "step": 41828 + }, + { + "epoch": 3.09, + "learning_rate": 9.995213540742545e-06, + "loss": 1.0206, + "step": 41829 + }, + { + "epoch": 3.09, + "learning_rate": 9.994814669172128e-06, + "loss": 0.9448, + "step": 41830 + }, + { + "epoch": 3.09, + "learning_rate": 9.994415797609967e-06, + "loss": 1.0145, + "step": 41831 + }, + { + "epoch": 3.09, + "learning_rate": 9.994016926056683e-06, + "loss": 0.9801, + "step": 41832 + }, + { + "epoch": 3.09, + "learning_rate": 9.993618054512926e-06, + "loss": 0.9816, + "step": 41833 + }, + { + "epoch": 3.09, + "learning_rate": 9.993219182979316e-06, + "loss": 1.0045, + "step": 41834 + }, + { + "epoch": 3.09, + "learning_rate": 9.9928203114565e-06, + "loss": 1.0092, + "step": 41835 + }, + { + "epoch": 3.09, + "learning_rate": 9.992421439945103e-06, + "loss": 0.9626, + "step": 41836 + }, + { + "epoch": 3.09, + "learning_rate": 9.992022568445765e-06, + "loss": 0.866, + "step": 41837 + }, + { + "epoch": 3.09, + "learning_rate": 9.991623696959118e-06, + "loss": 0.965, + "step": 41838 + }, + { + "epoch": 3.09, + "learning_rate": 9.991224825485795e-06, + "loss": 1.0597, + "step": 41839 + }, + { + "epoch": 3.09, + "learning_rate": 9.990825954026437e-06, + "loss": 1.0414, + "step": 41840 + }, + { + "epoch": 3.09, + "learning_rate": 9.990427082581673e-06, + "loss": 1.0464, + "step": 41841 + }, + { + "epoch": 3.09, + "learning_rate": 9.990028211152142e-06, + "loss": 1.0317, + "step": 41842 + }, + { + "epoch": 3.09, + "learning_rate": 9.989629339738472e-06, + "loss": 1.0438, + "step": 41843 + }, + { + "epoch": 3.09, + "learning_rate": 9.989230468341303e-06, + "loss": 0.9427, + "step": 41844 + }, + { + "epoch": 3.09, + "learning_rate": 9.988831596961269e-06, + "loss": 0.9431, + "step": 41845 + }, + { + "epoch": 3.09, + "learning_rate": 9.988432725599004e-06, + "loss": 1.0846, + "step": 41846 + }, + { + "epoch": 3.09, + "learning_rate": 9.988033854255143e-06, + "loss": 1.0271, + "step": 41847 + }, + { + "epoch": 3.09, + "learning_rate": 9.987634982930316e-06, + "loss": 1.0615, + "step": 41848 + }, + { + "epoch": 3.09, + "learning_rate": 9.987236111625168e-06, + "loss": 0.9169, + "step": 41849 + }, + { + "epoch": 3.09, + "learning_rate": 9.986837240340319e-06, + "loss": 0.9008, + "step": 41850 + }, + { + "epoch": 3.09, + "learning_rate": 9.98643836907642e-06, + "loss": 1.0891, + "step": 41851 + }, + { + "epoch": 3.09, + "learning_rate": 9.98603949783409e-06, + "loss": 0.9576, + "step": 41852 + }, + { + "epoch": 3.09, + "learning_rate": 9.985640626613979e-06, + "loss": 1.1177, + "step": 41853 + }, + { + "epoch": 3.09, + "learning_rate": 9.985241755416704e-06, + "loss": 1.09, + "step": 41854 + }, + { + "epoch": 3.09, + "learning_rate": 9.984842884242918e-06, + "loss": 1.018, + "step": 41855 + }, + { + "epoch": 3.09, + "learning_rate": 9.984444013093245e-06, + "loss": 0.9506, + "step": 41856 + }, + { + "epoch": 3.09, + "learning_rate": 9.984045141968317e-06, + "loss": 1.0504, + "step": 41857 + }, + { + "epoch": 3.09, + "learning_rate": 9.983646270868776e-06, + "loss": 0.9446, + "step": 41858 + }, + { + "epoch": 3.09, + "learning_rate": 9.983247399795251e-06, + "loss": 1.0553, + "step": 41859 + }, + { + "epoch": 3.09, + "learning_rate": 9.982848528748383e-06, + "loss": 1.0392, + "step": 41860 + }, + { + "epoch": 3.09, + "learning_rate": 9.982449657728799e-06, + "loss": 0.9995, + "step": 41861 + }, + { + "epoch": 3.09, + "learning_rate": 9.98205078673714e-06, + "loss": 0.8497, + "step": 41862 + }, + { + "epoch": 3.09, + "learning_rate": 9.981651915774035e-06, + "loss": 1.011, + "step": 41863 + }, + { + "epoch": 3.09, + "learning_rate": 9.981253044840125e-06, + "loss": 0.9909, + "step": 41864 + }, + { + "epoch": 3.09, + "learning_rate": 9.980854173936041e-06, + "loss": 1.1105, + "step": 41865 + }, + { + "epoch": 3.09, + "learning_rate": 9.980455303062416e-06, + "loss": 1.076, + "step": 41866 + }, + { + "epoch": 3.09, + "learning_rate": 9.980056432219887e-06, + "loss": 0.9459, + "step": 41867 + }, + { + "epoch": 3.09, + "learning_rate": 9.979657561409086e-06, + "loss": 1.1092, + "step": 41868 + }, + { + "epoch": 3.09, + "learning_rate": 9.979258690630653e-06, + "loss": 1.0374, + "step": 41869 + }, + { + "epoch": 3.09, + "learning_rate": 9.978859819885216e-06, + "loss": 0.8882, + "step": 41870 + }, + { + "epoch": 3.09, + "learning_rate": 9.978460949173415e-06, + "loss": 0.9748, + "step": 41871 + }, + { + "epoch": 3.09, + "learning_rate": 9.978062078495881e-06, + "loss": 0.9851, + "step": 41872 + }, + { + "epoch": 3.09, + "learning_rate": 9.97766320785325e-06, + "loss": 1.0897, + "step": 41873 + }, + { + "epoch": 3.09, + "learning_rate": 9.97726433724616e-06, + "loss": 0.9638, + "step": 41874 + }, + { + "epoch": 3.09, + "learning_rate": 9.976865466675235e-06, + "loss": 1.062, + "step": 41875 + }, + { + "epoch": 3.09, + "learning_rate": 9.976466596141125e-06, + "loss": 1.0271, + "step": 41876 + }, + { + "epoch": 3.09, + "learning_rate": 9.976067725644448e-06, + "loss": 0.975, + "step": 41877 + }, + { + "epoch": 3.09, + "learning_rate": 9.975668855185854e-06, + "loss": 1.0644, + "step": 41878 + }, + { + "epoch": 3.09, + "learning_rate": 9.975269984765966e-06, + "loss": 1.107, + "step": 41879 + }, + { + "epoch": 3.09, + "learning_rate": 9.974871114385425e-06, + "loss": 1.0161, + "step": 41880 + }, + { + "epoch": 3.09, + "learning_rate": 9.974472244044862e-06, + "loss": 0.9826, + "step": 41881 + }, + { + "epoch": 3.09, + "learning_rate": 9.974073373744915e-06, + "loss": 0.9263, + "step": 41882 + }, + { + "epoch": 3.09, + "learning_rate": 9.973674503486218e-06, + "loss": 0.9023, + "step": 41883 + }, + { + "epoch": 3.09, + "learning_rate": 9.973275633269402e-06, + "loss": 1.1361, + "step": 41884 + }, + { + "epoch": 3.09, + "learning_rate": 9.972876763095104e-06, + "loss": 0.9873, + "step": 41885 + }, + { + "epoch": 3.09, + "learning_rate": 9.97247789296396e-06, + "loss": 1.0492, + "step": 41886 + }, + { + "epoch": 3.1, + "learning_rate": 9.972079022876605e-06, + "loss": 1.0087, + "step": 41887 + }, + { + "epoch": 3.1, + "learning_rate": 9.971680152833667e-06, + "loss": 0.9665, + "step": 41888 + }, + { + "epoch": 3.1, + "learning_rate": 9.97128128283579e-06, + "loss": 1.0248, + "step": 41889 + }, + { + "epoch": 3.1, + "learning_rate": 9.9708824128836e-06, + "loss": 0.9834, + "step": 41890 + }, + { + "epoch": 3.1, + "learning_rate": 9.970483542977739e-06, + "loss": 0.9668, + "step": 41891 + }, + { + "epoch": 3.1, + "learning_rate": 9.97008467311884e-06, + "loss": 0.987, + "step": 41892 + }, + { + "epoch": 3.1, + "learning_rate": 9.96968580330753e-06, + "loss": 1.0241, + "step": 41893 + }, + { + "epoch": 3.1, + "learning_rate": 9.969286933544457e-06, + "loss": 0.9249, + "step": 41894 + }, + { + "epoch": 3.1, + "learning_rate": 9.96888806383024e-06, + "loss": 1.0736, + "step": 41895 + }, + { + "epoch": 3.1, + "learning_rate": 9.968489194165528e-06, + "loss": 0.9663, + "step": 41896 + }, + { + "epoch": 3.1, + "learning_rate": 9.968090324550943e-06, + "loss": 0.9348, + "step": 41897 + }, + { + "epoch": 3.1, + "learning_rate": 9.967691454987132e-06, + "loss": 1.0402, + "step": 41898 + }, + { + "epoch": 3.1, + "learning_rate": 9.967292585474718e-06, + "loss": 1.0969, + "step": 41899 + }, + { + "epoch": 3.1, + "learning_rate": 9.966893716014347e-06, + "loss": 0.8717, + "step": 41900 + }, + { + "epoch": 3.1, + "learning_rate": 9.96649484660664e-06, + "loss": 0.9872, + "step": 41901 + }, + { + "epoch": 3.1, + "learning_rate": 9.966095977252248e-06, + "loss": 1.0001, + "step": 41902 + }, + { + "epoch": 3.1, + "learning_rate": 9.965697107951793e-06, + "loss": 0.9792, + "step": 41903 + }, + { + "epoch": 3.1, + "learning_rate": 9.965298238705911e-06, + "loss": 1.0653, + "step": 41904 + }, + { + "epoch": 3.1, + "learning_rate": 9.964899369515242e-06, + "loss": 1.0818, + "step": 41905 + }, + { + "epoch": 3.1, + "learning_rate": 9.964500500380415e-06, + "loss": 0.9849, + "step": 41906 + }, + { + "epoch": 3.1, + "learning_rate": 9.96410163130207e-06, + "loss": 1.0204, + "step": 41907 + }, + { + "epoch": 3.1, + "learning_rate": 9.963702762280836e-06, + "loss": 0.957, + "step": 41908 + }, + { + "epoch": 3.1, + "learning_rate": 9.963303893317353e-06, + "loss": 1.054, + "step": 41909 + }, + { + "epoch": 3.1, + "learning_rate": 9.962905024412251e-06, + "loss": 1.0402, + "step": 41910 + }, + { + "epoch": 3.1, + "learning_rate": 9.962506155566167e-06, + "loss": 1.0049, + "step": 41911 + }, + { + "epoch": 3.1, + "learning_rate": 9.962107286779738e-06, + "loss": 1.024, + "step": 41912 + }, + { + "epoch": 3.1, + "learning_rate": 9.961708418053592e-06, + "loss": 1.0081, + "step": 41913 + }, + { + "epoch": 3.1, + "learning_rate": 9.96130954938837e-06, + "loss": 1.1288, + "step": 41914 + }, + { + "epoch": 3.1, + "learning_rate": 9.9609106807847e-06, + "loss": 0.9875, + "step": 41915 + }, + { + "epoch": 3.1, + "learning_rate": 9.960511812243224e-06, + "loss": 0.9797, + "step": 41916 + }, + { + "epoch": 3.1, + "learning_rate": 9.960112943764572e-06, + "loss": 0.9701, + "step": 41917 + }, + { + "epoch": 3.1, + "learning_rate": 9.95971407534938e-06, + "loss": 1.0122, + "step": 41918 + }, + { + "epoch": 3.1, + "learning_rate": 9.959315206998282e-06, + "loss": 0.9571, + "step": 41919 + }, + { + "epoch": 3.1, + "learning_rate": 9.958916338711914e-06, + "loss": 1.003, + "step": 41920 + }, + { + "epoch": 3.1, + "learning_rate": 9.958517470490911e-06, + "loss": 0.974, + "step": 41921 + }, + { + "epoch": 3.1, + "learning_rate": 9.958118602335901e-06, + "loss": 1.0085, + "step": 41922 + }, + { + "epoch": 3.1, + "learning_rate": 9.95771973424753e-06, + "loss": 1.073, + "step": 41923 + }, + { + "epoch": 3.1, + "learning_rate": 9.957320866226423e-06, + "loss": 1.0297, + "step": 41924 + }, + { + "epoch": 3.1, + "learning_rate": 9.956921998273218e-06, + "loss": 1.0053, + "step": 41925 + }, + { + "epoch": 3.1, + "learning_rate": 9.956523130388549e-06, + "loss": 1.041, + "step": 41926 + }, + { + "epoch": 3.1, + "learning_rate": 9.956124262573051e-06, + "loss": 0.9274, + "step": 41927 + }, + { + "epoch": 3.1, + "learning_rate": 9.955725394827358e-06, + "loss": 0.9821, + "step": 41928 + }, + { + "epoch": 3.1, + "learning_rate": 9.955326527152107e-06, + "loss": 1.0746, + "step": 41929 + }, + { + "epoch": 3.1, + "learning_rate": 9.954927659547931e-06, + "loss": 0.984, + "step": 41930 + }, + { + "epoch": 3.1, + "learning_rate": 9.954528792015461e-06, + "loss": 0.9423, + "step": 41931 + }, + { + "epoch": 3.1, + "learning_rate": 9.95412992455534e-06, + "loss": 1.1229, + "step": 41932 + }, + { + "epoch": 3.1, + "learning_rate": 9.953731057168194e-06, + "loss": 1.0527, + "step": 41933 + }, + { + "epoch": 3.1, + "learning_rate": 9.953332189854664e-06, + "loss": 1.01, + "step": 41934 + }, + { + "epoch": 3.1, + "learning_rate": 9.95293332261538e-06, + "loss": 1.034, + "step": 41935 + }, + { + "epoch": 3.1, + "learning_rate": 9.952534455450979e-06, + "loss": 0.9748, + "step": 41936 + }, + { + "epoch": 3.1, + "learning_rate": 9.952135588362092e-06, + "loss": 1.0416, + "step": 41937 + }, + { + "epoch": 3.1, + "learning_rate": 9.951736721349362e-06, + "loss": 1.1069, + "step": 41938 + }, + { + "epoch": 3.1, + "learning_rate": 9.951337854413418e-06, + "loss": 1.0264, + "step": 41939 + }, + { + "epoch": 3.1, + "learning_rate": 9.950938987554889e-06, + "loss": 0.9096, + "step": 41940 + }, + { + "epoch": 3.1, + "learning_rate": 9.950540120774422e-06, + "loss": 1.0019, + "step": 41941 + }, + { + "epoch": 3.1, + "learning_rate": 9.950141254072639e-06, + "loss": 1.0879, + "step": 41942 + }, + { + "epoch": 3.1, + "learning_rate": 9.949742387450186e-06, + "loss": 1.0577, + "step": 41943 + }, + { + "epoch": 3.1, + "learning_rate": 9.949343520907686e-06, + "loss": 1.0449, + "step": 41944 + }, + { + "epoch": 3.1, + "learning_rate": 9.948944654445786e-06, + "loss": 0.993, + "step": 41945 + }, + { + "epoch": 3.1, + "learning_rate": 9.948545788065107e-06, + "loss": 1.0017, + "step": 41946 + }, + { + "epoch": 3.1, + "learning_rate": 9.948146921766298e-06, + "loss": 1.0616, + "step": 41947 + }, + { + "epoch": 3.1, + "learning_rate": 9.947748055549987e-06, + "loss": 0.8363, + "step": 41948 + }, + { + "epoch": 3.1, + "learning_rate": 9.947349189416802e-06, + "loss": 0.9905, + "step": 41949 + }, + { + "epoch": 3.1, + "learning_rate": 9.946950323367389e-06, + "loss": 1.0138, + "step": 41950 + }, + { + "epoch": 3.1, + "learning_rate": 9.946551457402372e-06, + "loss": 1.0591, + "step": 41951 + }, + { + "epoch": 3.1, + "learning_rate": 9.946152591522396e-06, + "loss": 1.0606, + "step": 41952 + }, + { + "epoch": 3.1, + "learning_rate": 9.945753725728087e-06, + "loss": 1.0046, + "step": 41953 + }, + { + "epoch": 3.1, + "learning_rate": 9.945354860020084e-06, + "loss": 1.0566, + "step": 41954 + }, + { + "epoch": 3.1, + "learning_rate": 9.94495599439902e-06, + "loss": 1.0416, + "step": 41955 + }, + { + "epoch": 3.1, + "learning_rate": 9.944557128865531e-06, + "loss": 1.037, + "step": 41956 + }, + { + "epoch": 3.1, + "learning_rate": 9.94415826342025e-06, + "loss": 1.0586, + "step": 41957 + }, + { + "epoch": 3.1, + "learning_rate": 9.943759398063815e-06, + "loss": 1.0244, + "step": 41958 + }, + { + "epoch": 3.1, + "learning_rate": 9.943360532796857e-06, + "loss": 1.0611, + "step": 41959 + }, + { + "epoch": 3.1, + "learning_rate": 9.94296166762001e-06, + "loss": 1.0089, + "step": 41960 + }, + { + "epoch": 3.1, + "learning_rate": 9.942562802533912e-06, + "loss": 0.952, + "step": 41961 + }, + { + "epoch": 3.1, + "learning_rate": 9.942163937539193e-06, + "loss": 0.9557, + "step": 41962 + }, + { + "epoch": 3.1, + "learning_rate": 9.941765072636493e-06, + "loss": 0.935, + "step": 41963 + }, + { + "epoch": 3.1, + "learning_rate": 9.941366207826443e-06, + "loss": 1.0124, + "step": 41964 + }, + { + "epoch": 3.1, + "learning_rate": 9.940967343109679e-06, + "loss": 1.0618, + "step": 41965 + }, + { + "epoch": 3.1, + "learning_rate": 9.940568478486833e-06, + "loss": 1.0911, + "step": 41966 + }, + { + "epoch": 3.1, + "learning_rate": 9.940169613958547e-06, + "loss": 1.0479, + "step": 41967 + }, + { + "epoch": 3.1, + "learning_rate": 9.939770749525447e-06, + "loss": 0.949, + "step": 41968 + }, + { + "epoch": 3.1, + "learning_rate": 9.93937188518817e-06, + "loss": 0.943, + "step": 41969 + }, + { + "epoch": 3.1, + "learning_rate": 9.938973020947354e-06, + "loss": 0.9882, + "step": 41970 + }, + { + "epoch": 3.1, + "learning_rate": 9.938574156803627e-06, + "loss": 1.0229, + "step": 41971 + }, + { + "epoch": 3.1, + "learning_rate": 9.938175292757632e-06, + "loss": 1.0612, + "step": 41972 + }, + { + "epoch": 3.1, + "learning_rate": 9.937776428809993e-06, + "loss": 1.0263, + "step": 41973 + }, + { + "epoch": 3.1, + "learning_rate": 9.937377564961358e-06, + "loss": 0.8855, + "step": 41974 + }, + { + "epoch": 3.1, + "learning_rate": 9.936978701212349e-06, + "loss": 1.0462, + "step": 41975 + }, + { + "epoch": 3.1, + "learning_rate": 9.936579837563608e-06, + "loss": 0.9776, + "step": 41976 + }, + { + "epoch": 3.1, + "learning_rate": 9.93618097401577e-06, + "loss": 1.0817, + "step": 41977 + }, + { + "epoch": 3.1, + "learning_rate": 9.935782110569465e-06, + "loss": 0.9728, + "step": 41978 + }, + { + "epoch": 3.1, + "learning_rate": 9.935383247225329e-06, + "loss": 0.9452, + "step": 41979 + }, + { + "epoch": 3.1, + "learning_rate": 9.934984383983998e-06, + "loss": 0.9729, + "step": 41980 + }, + { + "epoch": 3.1, + "learning_rate": 9.934585520846106e-06, + "loss": 1.014, + "step": 41981 + }, + { + "epoch": 3.1, + "learning_rate": 9.934186657812286e-06, + "loss": 1.0929, + "step": 41982 + }, + { + "epoch": 3.1, + "learning_rate": 9.933787794883176e-06, + "loss": 0.9736, + "step": 41983 + }, + { + "epoch": 3.1, + "learning_rate": 9.933388932059408e-06, + "loss": 1.0114, + "step": 41984 + }, + { + "epoch": 3.1, + "learning_rate": 9.932990069341618e-06, + "loss": 1.0793, + "step": 41985 + }, + { + "epoch": 3.1, + "learning_rate": 9.932591206730441e-06, + "loss": 1.01, + "step": 41986 + }, + { + "epoch": 3.1, + "learning_rate": 9.932192344226506e-06, + "loss": 0.937, + "step": 41987 + }, + { + "epoch": 3.1, + "learning_rate": 9.931793481830458e-06, + "loss": 0.8332, + "step": 41988 + }, + { + "epoch": 3.1, + "learning_rate": 9.93139461954292e-06, + "loss": 1.0086, + "step": 41989 + }, + { + "epoch": 3.1, + "learning_rate": 9.930995757364537e-06, + "loss": 1.0141, + "step": 41990 + }, + { + "epoch": 3.1, + "learning_rate": 9.930596895295933e-06, + "loss": 0.9908, + "step": 41991 + }, + { + "epoch": 3.1, + "learning_rate": 9.930198033337755e-06, + "loss": 0.8984, + "step": 41992 + }, + { + "epoch": 3.1, + "learning_rate": 9.929799171490624e-06, + "loss": 0.9171, + "step": 41993 + }, + { + "epoch": 3.1, + "learning_rate": 9.929400309755189e-06, + "loss": 1.0285, + "step": 41994 + }, + { + "epoch": 3.1, + "learning_rate": 9.929001448132073e-06, + "loss": 0.8602, + "step": 41995 + }, + { + "epoch": 3.1, + "learning_rate": 9.928602586621914e-06, + "loss": 0.9414, + "step": 41996 + }, + { + "epoch": 3.1, + "learning_rate": 9.928203725225348e-06, + "loss": 0.9774, + "step": 41997 + }, + { + "epoch": 3.1, + "learning_rate": 9.927804863943007e-06, + "loss": 1.0535, + "step": 41998 + }, + { + "epoch": 3.1, + "learning_rate": 9.927406002775531e-06, + "loss": 1.0071, + "step": 41999 + }, + { + "epoch": 3.1, + "learning_rate": 9.927007141723548e-06, + "loss": 0.8983, + "step": 42000 + }, + { + "epoch": 3.1, + "learning_rate": 9.926608280787698e-06, + "loss": 0.9273, + "step": 42001 + }, + { + "epoch": 3.1, + "learning_rate": 9.926209419968611e-06, + "loss": 1.0541, + "step": 42002 + }, + { + "epoch": 3.1, + "learning_rate": 9.925810559266927e-06, + "loss": 0.9877, + "step": 42003 + }, + { + "epoch": 3.1, + "learning_rate": 9.925411698683275e-06, + "loss": 1.151, + "step": 42004 + }, + { + "epoch": 3.1, + "learning_rate": 9.92501283821829e-06, + "loss": 1.0041, + "step": 42005 + }, + { + "epoch": 3.1, + "learning_rate": 9.924613977872613e-06, + "loss": 1.0422, + "step": 42006 + }, + { + "epoch": 3.1, + "learning_rate": 9.924215117646871e-06, + "loss": 1.0633, + "step": 42007 + }, + { + "epoch": 3.1, + "learning_rate": 9.923816257541703e-06, + "loss": 1.0538, + "step": 42008 + }, + { + "epoch": 3.1, + "learning_rate": 9.92341739755774e-06, + "loss": 0.9943, + "step": 42009 + }, + { + "epoch": 3.1, + "learning_rate": 9.923018537695621e-06, + "loss": 0.9438, + "step": 42010 + }, + { + "epoch": 3.1, + "learning_rate": 9.922619677955976e-06, + "loss": 0.9992, + "step": 42011 + }, + { + "epoch": 3.1, + "learning_rate": 9.922220818339448e-06, + "loss": 0.9936, + "step": 42012 + }, + { + "epoch": 3.1, + "learning_rate": 9.921821958846658e-06, + "loss": 1.0081, + "step": 42013 + }, + { + "epoch": 3.1, + "learning_rate": 9.921423099478254e-06, + "loss": 0.8862, + "step": 42014 + }, + { + "epoch": 3.1, + "learning_rate": 9.921024240234864e-06, + "loss": 1.026, + "step": 42015 + }, + { + "epoch": 3.1, + "learning_rate": 9.920625381117121e-06, + "loss": 0.9215, + "step": 42016 + }, + { + "epoch": 3.1, + "learning_rate": 9.920226522125662e-06, + "loss": 1.0137, + "step": 42017 + }, + { + "epoch": 3.1, + "learning_rate": 9.919827663261121e-06, + "loss": 1.0229, + "step": 42018 + }, + { + "epoch": 3.1, + "learning_rate": 9.919428804524135e-06, + "loss": 1.0112, + "step": 42019 + }, + { + "epoch": 3.1, + "learning_rate": 9.919029945915333e-06, + "loss": 1.0484, + "step": 42020 + }, + { + "epoch": 3.1, + "learning_rate": 9.918631087435359e-06, + "loss": 1.0546, + "step": 42021 + }, + { + "epoch": 3.1, + "learning_rate": 9.918232229084836e-06, + "loss": 1.0325, + "step": 42022 + }, + { + "epoch": 3.11, + "learning_rate": 9.917833370864408e-06, + "loss": 1.0675, + "step": 42023 + }, + { + "epoch": 3.11, + "learning_rate": 9.917434512774706e-06, + "loss": 0.886, + "step": 42024 + }, + { + "epoch": 3.11, + "learning_rate": 9.917035654816363e-06, + "loss": 0.9739, + "step": 42025 + }, + { + "epoch": 3.11, + "learning_rate": 9.916636796990016e-06, + "loss": 1.0052, + "step": 42026 + }, + { + "epoch": 3.11, + "learning_rate": 9.916237939296295e-06, + "loss": 1.0271, + "step": 42027 + }, + { + "epoch": 3.11, + "learning_rate": 9.915839081735844e-06, + "loss": 0.9675, + "step": 42028 + }, + { + "epoch": 3.11, + "learning_rate": 9.915440224309287e-06, + "loss": 0.8861, + "step": 42029 + }, + { + "epoch": 3.11, + "learning_rate": 9.915041367017268e-06, + "loss": 1.0055, + "step": 42030 + }, + { + "epoch": 3.11, + "learning_rate": 9.914642509860413e-06, + "loss": 0.9838, + "step": 42031 + }, + { + "epoch": 3.11, + "learning_rate": 9.914243652839363e-06, + "loss": 0.9796, + "step": 42032 + }, + { + "epoch": 3.11, + "learning_rate": 9.913844795954752e-06, + "loss": 1.0163, + "step": 42033 + }, + { + "epoch": 3.11, + "learning_rate": 9.913445939207205e-06, + "loss": 0.9815, + "step": 42034 + }, + { + "epoch": 3.11, + "learning_rate": 9.913047082597373e-06, + "loss": 0.9845, + "step": 42035 + }, + { + "epoch": 3.11, + "learning_rate": 9.912648226125874e-06, + "loss": 0.9505, + "step": 42036 + }, + { + "epoch": 3.11, + "learning_rate": 9.912249369793359e-06, + "loss": 0.969, + "step": 42037 + }, + { + "epoch": 3.11, + "learning_rate": 9.911850513600446e-06, + "loss": 0.9763, + "step": 42038 + }, + { + "epoch": 3.11, + "learning_rate": 9.911451657547783e-06, + "loss": 0.9914, + "step": 42039 + }, + { + "epoch": 3.11, + "learning_rate": 9.911052801635994e-06, + "loss": 0.9961, + "step": 42040 + }, + { + "epoch": 3.11, + "learning_rate": 9.910653945865725e-06, + "loss": 1.0491, + "step": 42041 + }, + { + "epoch": 3.11, + "learning_rate": 9.910255090237602e-06, + "loss": 0.984, + "step": 42042 + }, + { + "epoch": 3.11, + "learning_rate": 9.90985623475226e-06, + "loss": 1.0247, + "step": 42043 + }, + { + "epoch": 3.11, + "learning_rate": 9.909457379410338e-06, + "loss": 1.0801, + "step": 42044 + }, + { + "epoch": 3.11, + "learning_rate": 9.909058524212465e-06, + "loss": 0.9332, + "step": 42045 + }, + { + "epoch": 3.11, + "learning_rate": 9.908659669159281e-06, + "loss": 1.0476, + "step": 42046 + }, + { + "epoch": 3.11, + "learning_rate": 9.908260814251415e-06, + "loss": 0.8453, + "step": 42047 + }, + { + "epoch": 3.11, + "learning_rate": 9.907861959489509e-06, + "loss": 0.9878, + "step": 42048 + }, + { + "epoch": 3.11, + "learning_rate": 9.90746310487419e-06, + "loss": 1.0429, + "step": 42049 + }, + { + "epoch": 3.11, + "learning_rate": 9.907064250406097e-06, + "loss": 0.9981, + "step": 42050 + }, + { + "epoch": 3.11, + "learning_rate": 9.906665396085864e-06, + "loss": 0.992, + "step": 42051 + }, + { + "epoch": 3.11, + "learning_rate": 9.906266541914125e-06, + "loss": 1.0149, + "step": 42052 + }, + { + "epoch": 3.11, + "learning_rate": 9.905867687891515e-06, + "loss": 1.0846, + "step": 42053 + }, + { + "epoch": 3.11, + "learning_rate": 9.905468834018665e-06, + "loss": 1.0614, + "step": 42054 + }, + { + "epoch": 3.11, + "learning_rate": 9.905069980296219e-06, + "loss": 1.0637, + "step": 42055 + }, + { + "epoch": 3.11, + "learning_rate": 9.904671126724798e-06, + "loss": 0.9601, + "step": 42056 + }, + { + "epoch": 3.11, + "learning_rate": 9.904272273305051e-06, + "loss": 0.9964, + "step": 42057 + }, + { + "epoch": 3.11, + "learning_rate": 9.903873420037597e-06, + "loss": 0.9919, + "step": 42058 + }, + { + "epoch": 3.11, + "learning_rate": 9.903474566923088e-06, + "loss": 0.9966, + "step": 42059 + }, + { + "epoch": 3.11, + "learning_rate": 9.903075713962142e-06, + "loss": 1.0378, + "step": 42060 + }, + { + "epoch": 3.11, + "learning_rate": 9.902676861155408e-06, + "loss": 0.9856, + "step": 42061 + }, + { + "epoch": 3.11, + "learning_rate": 9.902278008503511e-06, + "loss": 0.9687, + "step": 42062 + }, + { + "epoch": 3.11, + "learning_rate": 9.901879156007086e-06, + "loss": 0.9545, + "step": 42063 + }, + { + "epoch": 3.11, + "learning_rate": 9.901480303666772e-06, + "loss": 0.9744, + "step": 42064 + }, + { + "epoch": 3.11, + "learning_rate": 9.901081451483198e-06, + "loss": 1.023, + "step": 42065 + }, + { + "epoch": 3.11, + "learning_rate": 9.900682599457006e-06, + "loss": 0.993, + "step": 42066 + }, + { + "epoch": 3.11, + "learning_rate": 9.900283747588825e-06, + "loss": 1.0546, + "step": 42067 + }, + { + "epoch": 3.11, + "learning_rate": 9.899884895879292e-06, + "loss": 0.9782, + "step": 42068 + }, + { + "epoch": 3.11, + "learning_rate": 9.89948604432904e-06, + "loss": 1.0509, + "step": 42069 + }, + { + "epoch": 3.11, + "learning_rate": 9.899087192938704e-06, + "loss": 1.0005, + "step": 42070 + }, + { + "epoch": 3.11, + "learning_rate": 9.898688341708921e-06, + "loss": 1.0329, + "step": 42071 + }, + { + "epoch": 3.11, + "learning_rate": 9.898289490640321e-06, + "loss": 1.0354, + "step": 42072 + }, + { + "epoch": 3.11, + "learning_rate": 9.897890639733541e-06, + "loss": 1.0836, + "step": 42073 + }, + { + "epoch": 3.11, + "learning_rate": 9.897491788989215e-06, + "loss": 0.9383, + "step": 42074 + }, + { + "epoch": 3.11, + "learning_rate": 9.89709293840798e-06, + "loss": 1.0299, + "step": 42075 + }, + { + "epoch": 3.11, + "learning_rate": 9.896694087990467e-06, + "loss": 1.1251, + "step": 42076 + }, + { + "epoch": 3.11, + "learning_rate": 9.896295237737315e-06, + "loss": 1.0175, + "step": 42077 + }, + { + "epoch": 3.11, + "learning_rate": 9.895896387649152e-06, + "loss": 0.9454, + "step": 42078 + }, + { + "epoch": 3.11, + "learning_rate": 9.895497537726619e-06, + "loss": 1.0421, + "step": 42079 + }, + { + "epoch": 3.11, + "learning_rate": 9.89509868797035e-06, + "loss": 0.9928, + "step": 42080 + }, + { + "epoch": 3.11, + "learning_rate": 9.894699838380972e-06, + "loss": 0.9652, + "step": 42081 + }, + { + "epoch": 3.11, + "learning_rate": 9.894300988959131e-06, + "loss": 0.8944, + "step": 42082 + }, + { + "epoch": 3.11, + "learning_rate": 9.89390213970545e-06, + "loss": 0.9172, + "step": 42083 + }, + { + "epoch": 3.11, + "learning_rate": 9.893503290620575e-06, + "loss": 0.9122, + "step": 42084 + }, + { + "epoch": 3.11, + "learning_rate": 9.893104441705127e-06, + "loss": 0.942, + "step": 42085 + }, + { + "epoch": 3.11, + "learning_rate": 9.892705592959755e-06, + "loss": 0.9956, + "step": 42086 + }, + { + "epoch": 3.11, + "learning_rate": 9.89230674438508e-06, + "loss": 1.0054, + "step": 42087 + }, + { + "epoch": 3.11, + "learning_rate": 9.891907895981751e-06, + "loss": 1.07, + "step": 42088 + }, + { + "epoch": 3.11, + "learning_rate": 9.891509047750395e-06, + "loss": 1.0137, + "step": 42089 + }, + { + "epoch": 3.11, + "learning_rate": 9.89111019969164e-06, + "loss": 1.0549, + "step": 42090 + }, + { + "epoch": 3.11, + "learning_rate": 9.890711351806132e-06, + "loss": 0.9963, + "step": 42091 + }, + { + "epoch": 3.11, + "learning_rate": 9.890312504094498e-06, + "loss": 1.0214, + "step": 42092 + }, + { + "epoch": 3.11, + "learning_rate": 9.889913656557378e-06, + "loss": 0.8791, + "step": 42093 + }, + { + "epoch": 3.11, + "learning_rate": 9.8895148091954e-06, + "loss": 0.9621, + "step": 42094 + }, + { + "epoch": 3.11, + "learning_rate": 9.889115962009206e-06, + "loss": 1.0257, + "step": 42095 + }, + { + "epoch": 3.11, + "learning_rate": 9.888717114999424e-06, + "loss": 0.9119, + "step": 42096 + }, + { + "epoch": 3.11, + "learning_rate": 9.888318268166695e-06, + "loss": 1.0611, + "step": 42097 + }, + { + "epoch": 3.11, + "learning_rate": 9.887919421511648e-06, + "loss": 1.0843, + "step": 42098 + }, + { + "epoch": 3.11, + "learning_rate": 9.88752057503492e-06, + "loss": 1.1773, + "step": 42099 + }, + { + "epoch": 3.11, + "learning_rate": 9.887121728737147e-06, + "loss": 1.0016, + "step": 42100 + }, + { + "epoch": 3.11, + "learning_rate": 9.886722882618956e-06, + "loss": 1.0407, + "step": 42101 + }, + { + "epoch": 3.11, + "learning_rate": 9.886324036680995e-06, + "loss": 0.9967, + "step": 42102 + }, + { + "epoch": 3.11, + "learning_rate": 9.885925190923882e-06, + "loss": 0.9969, + "step": 42103 + }, + { + "epoch": 3.11, + "learning_rate": 9.885526345348268e-06, + "loss": 0.9725, + "step": 42104 + }, + { + "epoch": 3.11, + "learning_rate": 9.885127499954773e-06, + "loss": 0.9785, + "step": 42105 + }, + { + "epoch": 3.11, + "learning_rate": 9.884728654744046e-06, + "loss": 0.9214, + "step": 42106 + }, + { + "epoch": 3.11, + "learning_rate": 9.884329809716711e-06, + "loss": 1.1506, + "step": 42107 + }, + { + "epoch": 3.11, + "learning_rate": 9.883930964873405e-06, + "loss": 0.9665, + "step": 42108 + }, + { + "epoch": 3.11, + "learning_rate": 9.883532120214763e-06, + "loss": 1.0621, + "step": 42109 + }, + { + "epoch": 3.11, + "learning_rate": 9.883133275741419e-06, + "loss": 0.9948, + "step": 42110 + }, + { + "epoch": 3.11, + "learning_rate": 9.88273443145401e-06, + "loss": 0.9281, + "step": 42111 + }, + { + "epoch": 3.11, + "learning_rate": 9.882335587353169e-06, + "loss": 0.9833, + "step": 42112 + }, + { + "epoch": 3.11, + "learning_rate": 9.881936743439528e-06, + "loss": 1.0191, + "step": 42113 + }, + { + "epoch": 3.11, + "learning_rate": 9.881537899713726e-06, + "loss": 0.9708, + "step": 42114 + }, + { + "epoch": 3.11, + "learning_rate": 9.881139056176395e-06, + "loss": 1.0379, + "step": 42115 + }, + { + "epoch": 3.11, + "learning_rate": 9.88074021282817e-06, + "loss": 1.0014, + "step": 42116 + }, + { + "epoch": 3.11, + "learning_rate": 9.880341369669686e-06, + "loss": 1.1564, + "step": 42117 + }, + { + "epoch": 3.11, + "learning_rate": 9.879942526701578e-06, + "loss": 0.9764, + "step": 42118 + }, + { + "epoch": 3.11, + "learning_rate": 9.879543683924478e-06, + "loss": 0.9273, + "step": 42119 + }, + { + "epoch": 3.11, + "learning_rate": 9.879144841339023e-06, + "loss": 1.0135, + "step": 42120 + }, + { + "epoch": 3.11, + "learning_rate": 9.878745998945844e-06, + "loss": 1.0967, + "step": 42121 + }, + { + "epoch": 3.11, + "learning_rate": 9.878347156745583e-06, + "loss": 1.0751, + "step": 42122 + }, + { + "epoch": 3.11, + "learning_rate": 9.877948314738866e-06, + "loss": 0.9473, + "step": 42123 + }, + { + "epoch": 3.11, + "learning_rate": 9.877549472926335e-06, + "loss": 0.9872, + "step": 42124 + }, + { + "epoch": 3.11, + "learning_rate": 9.877150631308618e-06, + "loss": 1.0759, + "step": 42125 + }, + { + "epoch": 3.11, + "learning_rate": 9.876751789886356e-06, + "loss": 0.9969, + "step": 42126 + }, + { + "epoch": 3.11, + "learning_rate": 9.87635294866018e-06, + "loss": 1.1018, + "step": 42127 + }, + { + "epoch": 3.11, + "learning_rate": 9.87595410763072e-06, + "loss": 0.9758, + "step": 42128 + }, + { + "epoch": 3.11, + "learning_rate": 9.875555266798622e-06, + "loss": 1.0042, + "step": 42129 + }, + { + "epoch": 3.11, + "learning_rate": 9.875156426164507e-06, + "loss": 0.8996, + "step": 42130 + }, + { + "epoch": 3.11, + "learning_rate": 9.87475758572902e-06, + "loss": 0.9027, + "step": 42131 + }, + { + "epoch": 3.11, + "learning_rate": 9.874358745492789e-06, + "loss": 1.002, + "step": 42132 + }, + { + "epoch": 3.11, + "learning_rate": 9.873959905456457e-06, + "loss": 0.9937, + "step": 42133 + }, + { + "epoch": 3.11, + "learning_rate": 9.873561065620648e-06, + "loss": 0.9711, + "step": 42134 + }, + { + "epoch": 3.11, + "learning_rate": 9.873162225986005e-06, + "loss": 1.0015, + "step": 42135 + }, + { + "epoch": 3.11, + "learning_rate": 9.872763386553159e-06, + "loss": 0.9552, + "step": 42136 + }, + { + "epoch": 3.11, + "learning_rate": 9.87236454732274e-06, + "loss": 1.0789, + "step": 42137 + }, + { + "epoch": 3.11, + "learning_rate": 9.871965708295392e-06, + "loss": 1.0344, + "step": 42138 + }, + { + "epoch": 3.11, + "learning_rate": 9.87156686947174e-06, + "loss": 1.0345, + "step": 42139 + }, + { + "epoch": 3.11, + "learning_rate": 9.871168030852428e-06, + "loss": 1.0829, + "step": 42140 + }, + { + "epoch": 3.11, + "learning_rate": 9.870769192438082e-06, + "loss": 0.9677, + "step": 42141 + }, + { + "epoch": 3.11, + "learning_rate": 9.870370354229344e-06, + "loss": 0.9993, + "step": 42142 + }, + { + "epoch": 3.11, + "learning_rate": 9.869971516226841e-06, + "loss": 0.9888, + "step": 42143 + }, + { + "epoch": 3.11, + "learning_rate": 9.869572678431215e-06, + "loss": 1.0424, + "step": 42144 + }, + { + "epoch": 3.11, + "learning_rate": 9.8691738408431e-06, + "loss": 0.9682, + "step": 42145 + }, + { + "epoch": 3.11, + "learning_rate": 9.868775003463117e-06, + "loss": 0.9532, + "step": 42146 + }, + { + "epoch": 3.11, + "learning_rate": 9.868376166291922e-06, + "loss": 1.0215, + "step": 42147 + }, + { + "epoch": 3.11, + "learning_rate": 9.86797732933013e-06, + "loss": 1.0312, + "step": 42148 + }, + { + "epoch": 3.11, + "learning_rate": 9.867578492578392e-06, + "loss": 1.0258, + "step": 42149 + }, + { + "epoch": 3.11, + "learning_rate": 9.867179656037326e-06, + "loss": 1.0659, + "step": 42150 + }, + { + "epoch": 3.11, + "learning_rate": 9.866780819707583e-06, + "loss": 0.955, + "step": 42151 + }, + { + "epoch": 3.11, + "learning_rate": 9.866381983589785e-06, + "loss": 1.016, + "step": 42152 + }, + { + "epoch": 3.11, + "learning_rate": 9.865983147684575e-06, + "loss": 1.0201, + "step": 42153 + }, + { + "epoch": 3.11, + "learning_rate": 9.865584311992584e-06, + "loss": 0.9662, + "step": 42154 + }, + { + "epoch": 3.11, + "learning_rate": 9.865185476514443e-06, + "loss": 0.9313, + "step": 42155 + }, + { + "epoch": 3.11, + "learning_rate": 9.864786641250793e-06, + "loss": 1.033, + "step": 42156 + }, + { + "epoch": 3.11, + "learning_rate": 9.864387806202262e-06, + "loss": 0.9996, + "step": 42157 + }, + { + "epoch": 3.12, + "learning_rate": 9.86398897136949e-06, + "loss": 1.0471, + "step": 42158 + }, + { + "epoch": 3.12, + "learning_rate": 9.86359013675311e-06, + "loss": 1.0293, + "step": 42159 + }, + { + "epoch": 3.12, + "learning_rate": 9.863191302353756e-06, + "loss": 1.0301, + "step": 42160 + }, + { + "epoch": 3.12, + "learning_rate": 9.862792468172062e-06, + "loss": 0.9417, + "step": 42161 + }, + { + "epoch": 3.12, + "learning_rate": 9.862393634208665e-06, + "loss": 1.0211, + "step": 42162 + }, + { + "epoch": 3.12, + "learning_rate": 9.861994800464197e-06, + "loss": 1.0177, + "step": 42163 + }, + { + "epoch": 3.12, + "learning_rate": 9.86159596693929e-06, + "loss": 0.9792, + "step": 42164 + }, + { + "epoch": 3.12, + "learning_rate": 9.861197133634587e-06, + "loss": 1.031, + "step": 42165 + }, + { + "epoch": 3.12, + "learning_rate": 9.860798300550714e-06, + "loss": 0.9574, + "step": 42166 + }, + { + "epoch": 3.12, + "learning_rate": 9.86039946768831e-06, + "loss": 1.0363, + "step": 42167 + }, + { + "epoch": 3.12, + "learning_rate": 9.860000635048008e-06, + "loss": 1.0973, + "step": 42168 + }, + { + "epoch": 3.12, + "learning_rate": 9.859601802630444e-06, + "loss": 1.0437, + "step": 42169 + }, + { + "epoch": 3.12, + "learning_rate": 9.85920297043625e-06, + "loss": 1.0708, + "step": 42170 + }, + { + "epoch": 3.12, + "learning_rate": 9.858804138466064e-06, + "loss": 1.0756, + "step": 42171 + }, + { + "epoch": 3.12, + "learning_rate": 9.858405306720516e-06, + "loss": 0.9886, + "step": 42172 + }, + { + "epoch": 3.12, + "learning_rate": 9.858006475200246e-06, + "loss": 0.8889, + "step": 42173 + }, + { + "epoch": 3.12, + "learning_rate": 9.857607643905886e-06, + "loss": 0.9675, + "step": 42174 + }, + { + "epoch": 3.12, + "learning_rate": 9.857208812838066e-06, + "loss": 0.9574, + "step": 42175 + }, + { + "epoch": 3.12, + "learning_rate": 9.85680998199743e-06, + "loss": 0.9789, + "step": 42176 + }, + { + "epoch": 3.12, + "learning_rate": 9.856411151384604e-06, + "loss": 1.048, + "step": 42177 + }, + { + "epoch": 3.12, + "learning_rate": 9.856012321000226e-06, + "loss": 0.9903, + "step": 42178 + }, + { + "epoch": 3.12, + "learning_rate": 9.85561349084493e-06, + "loss": 0.9162, + "step": 42179 + }, + { + "epoch": 3.12, + "learning_rate": 9.855214660919353e-06, + "loss": 0.9829, + "step": 42180 + }, + { + "epoch": 3.12, + "learning_rate": 9.854815831224125e-06, + "loss": 0.9678, + "step": 42181 + }, + { + "epoch": 3.12, + "learning_rate": 9.854417001759885e-06, + "loss": 0.8034, + "step": 42182 + }, + { + "epoch": 3.12, + "learning_rate": 9.854018172527264e-06, + "loss": 1.0468, + "step": 42183 + }, + { + "epoch": 3.12, + "learning_rate": 9.853619343526897e-06, + "loss": 1.0564, + "step": 42184 + }, + { + "epoch": 3.12, + "learning_rate": 9.853220514759422e-06, + "loss": 1.1708, + "step": 42185 + }, + { + "epoch": 3.12, + "learning_rate": 9.85282168622547e-06, + "loss": 0.8885, + "step": 42186 + }, + { + "epoch": 3.12, + "learning_rate": 9.852422857925677e-06, + "loss": 1.0013, + "step": 42187 + }, + { + "epoch": 3.12, + "learning_rate": 9.852024029860675e-06, + "loss": 0.936, + "step": 42188 + }, + { + "epoch": 3.12, + "learning_rate": 9.851625202031103e-06, + "loss": 1.1019, + "step": 42189 + }, + { + "epoch": 3.12, + "learning_rate": 9.851226374437592e-06, + "loss": 0.9564, + "step": 42190 + }, + { + "epoch": 3.12, + "learning_rate": 9.850827547080779e-06, + "loss": 0.9356, + "step": 42191 + }, + { + "epoch": 3.12, + "learning_rate": 9.8504287199613e-06, + "loss": 0.9686, + "step": 42192 + }, + { + "epoch": 3.12, + "learning_rate": 9.850029893079779e-06, + "loss": 0.9831, + "step": 42193 + }, + { + "epoch": 3.12, + "learning_rate": 9.849631066436866e-06, + "loss": 0.996, + "step": 42194 + }, + { + "epoch": 3.12, + "learning_rate": 9.84923224003318e-06, + "loss": 0.9674, + "step": 42195 + }, + { + "epoch": 3.12, + "learning_rate": 9.848833413869371e-06, + "loss": 1.0127, + "step": 42196 + }, + { + "epoch": 3.12, + "learning_rate": 9.84843458794606e-06, + "loss": 0.8889, + "step": 42197 + }, + { + "epoch": 3.12, + "learning_rate": 9.848035762263893e-06, + "loss": 0.9861, + "step": 42198 + }, + { + "epoch": 3.12, + "learning_rate": 9.847636936823492e-06, + "loss": 1.0079, + "step": 42199 + }, + { + "epoch": 3.12, + "learning_rate": 9.847238111625505e-06, + "loss": 1.02, + "step": 42200 + }, + { + "epoch": 3.12, + "learning_rate": 9.84683928667056e-06, + "loss": 0.9221, + "step": 42201 + }, + { + "epoch": 3.12, + "learning_rate": 9.846440461959286e-06, + "loss": 1.0523, + "step": 42202 + }, + { + "epoch": 3.12, + "learning_rate": 9.846041637492328e-06, + "loss": 1.0254, + "step": 42203 + }, + { + "epoch": 3.12, + "learning_rate": 9.845642813270312e-06, + "loss": 1.0332, + "step": 42204 + }, + { + "epoch": 3.12, + "learning_rate": 9.845243989293879e-06, + "loss": 1.1299, + "step": 42205 + }, + { + "epoch": 3.12, + "learning_rate": 9.844845165563656e-06, + "loss": 0.9539, + "step": 42206 + }, + { + "epoch": 3.12, + "learning_rate": 9.844446342080288e-06, + "loss": 0.8939, + "step": 42207 + }, + { + "epoch": 3.12, + "learning_rate": 9.844047518844399e-06, + "loss": 0.9416, + "step": 42208 + }, + { + "epoch": 3.12, + "learning_rate": 9.843648695856631e-06, + "loss": 1.0141, + "step": 42209 + }, + { + "epoch": 3.12, + "learning_rate": 9.843249873117615e-06, + "loss": 0.9444, + "step": 42210 + }, + { + "epoch": 3.12, + "learning_rate": 9.842851050627984e-06, + "loss": 1.1025, + "step": 42211 + }, + { + "epoch": 3.12, + "learning_rate": 9.842452228388378e-06, + "loss": 1.101, + "step": 42212 + }, + { + "epoch": 3.12, + "learning_rate": 9.842053406399425e-06, + "loss": 0.9966, + "step": 42213 + }, + { + "epoch": 3.12, + "learning_rate": 9.841654584661766e-06, + "loss": 1.0806, + "step": 42214 + }, + { + "epoch": 3.12, + "learning_rate": 9.84125576317603e-06, + "loss": 1.0243, + "step": 42215 + }, + { + "epoch": 3.12, + "learning_rate": 9.840856941942856e-06, + "loss": 0.9983, + "step": 42216 + }, + { + "epoch": 3.12, + "learning_rate": 9.840458120962873e-06, + "loss": 1.0272, + "step": 42217 + }, + { + "epoch": 3.12, + "learning_rate": 9.840059300236722e-06, + "loss": 0.9461, + "step": 42218 + }, + { + "epoch": 3.12, + "learning_rate": 9.839660479765037e-06, + "loss": 0.9616, + "step": 42219 + }, + { + "epoch": 3.12, + "learning_rate": 9.839261659548443e-06, + "loss": 0.9202, + "step": 42220 + }, + { + "epoch": 3.12, + "learning_rate": 9.838862839587588e-06, + "loss": 1.0697, + "step": 42221 + }, + { + "epoch": 3.12, + "learning_rate": 9.838464019883094e-06, + "loss": 1.0415, + "step": 42222 + }, + { + "epoch": 3.12, + "learning_rate": 9.838065200435604e-06, + "loss": 0.9727, + "step": 42223 + }, + { + "epoch": 3.12, + "learning_rate": 9.83766638124575e-06, + "loss": 0.9633, + "step": 42224 + }, + { + "epoch": 3.12, + "learning_rate": 9.837267562314167e-06, + "loss": 1.0684, + "step": 42225 + }, + { + "epoch": 3.12, + "learning_rate": 9.836868743641486e-06, + "loss": 0.944, + "step": 42226 + }, + { + "epoch": 3.12, + "learning_rate": 9.836469925228349e-06, + "loss": 1.0267, + "step": 42227 + }, + { + "epoch": 3.12, + "learning_rate": 9.836071107075383e-06, + "loss": 1.1243, + "step": 42228 + }, + { + "epoch": 3.12, + "learning_rate": 9.835672289183226e-06, + "loss": 1.0494, + "step": 42229 + }, + { + "epoch": 3.12, + "learning_rate": 9.835273471552515e-06, + "loss": 0.9684, + "step": 42230 + }, + { + "epoch": 3.12, + "learning_rate": 9.834874654183878e-06, + "loss": 1.0116, + "step": 42231 + }, + { + "epoch": 3.12, + "learning_rate": 9.834475837077954e-06, + "loss": 0.9668, + "step": 42232 + }, + { + "epoch": 3.12, + "learning_rate": 9.834077020235375e-06, + "loss": 0.995, + "step": 42233 + }, + { + "epoch": 3.12, + "learning_rate": 9.83367820365678e-06, + "loss": 0.9951, + "step": 42234 + }, + { + "epoch": 3.12, + "learning_rate": 9.833279387342797e-06, + "loss": 0.9917, + "step": 42235 + }, + { + "epoch": 3.12, + "learning_rate": 9.832880571294067e-06, + "loss": 1.015, + "step": 42236 + }, + { + "epoch": 3.12, + "learning_rate": 9.832481755511221e-06, + "loss": 1.1114, + "step": 42237 + }, + { + "epoch": 3.12, + "learning_rate": 9.832082939994894e-06, + "loss": 1.1069, + "step": 42238 + }, + { + "epoch": 3.12, + "learning_rate": 9.831684124745724e-06, + "loss": 0.9097, + "step": 42239 + }, + { + "epoch": 3.12, + "learning_rate": 9.831285309764336e-06, + "loss": 0.8722, + "step": 42240 + }, + { + "epoch": 3.12, + "learning_rate": 9.830886495051377e-06, + "loss": 1.065, + "step": 42241 + }, + { + "epoch": 3.12, + "learning_rate": 9.830487680607468e-06, + "loss": 1.0255, + "step": 42242 + }, + { + "epoch": 3.12, + "learning_rate": 9.830088866433258e-06, + "loss": 1.0053, + "step": 42243 + }, + { + "epoch": 3.12, + "learning_rate": 9.829690052529368e-06, + "loss": 0.9539, + "step": 42244 + }, + { + "epoch": 3.12, + "learning_rate": 9.829291238896444e-06, + "loss": 1.0089, + "step": 42245 + }, + { + "epoch": 3.12, + "learning_rate": 9.828892425535109e-06, + "loss": 1.0304, + "step": 42246 + }, + { + "epoch": 3.12, + "learning_rate": 9.828493612446009e-06, + "loss": 1.0563, + "step": 42247 + }, + { + "epoch": 3.12, + "learning_rate": 9.828094799629772e-06, + "loss": 0.8726, + "step": 42248 + }, + { + "epoch": 3.12, + "learning_rate": 9.827695987087033e-06, + "loss": 1.0648, + "step": 42249 + }, + { + "epoch": 3.12, + "learning_rate": 9.827297174818427e-06, + "loss": 0.9343, + "step": 42250 + }, + { + "epoch": 3.12, + "learning_rate": 9.826898362824588e-06, + "loss": 0.9545, + "step": 42251 + }, + { + "epoch": 3.12, + "learning_rate": 9.826499551106153e-06, + "loss": 1.0013, + "step": 42252 + }, + { + "epoch": 3.12, + "learning_rate": 9.826100739663753e-06, + "loss": 0.9398, + "step": 42253 + }, + { + "epoch": 3.12, + "learning_rate": 9.825701928498027e-06, + "loss": 1.0252, + "step": 42254 + }, + { + "epoch": 3.12, + "learning_rate": 9.825303117609604e-06, + "loss": 0.9989, + "step": 42255 + }, + { + "epoch": 3.12, + "learning_rate": 9.824904306999122e-06, + "loss": 1.0178, + "step": 42256 + }, + { + "epoch": 3.12, + "learning_rate": 9.824505496667216e-06, + "loss": 1.1099, + "step": 42257 + }, + { + "epoch": 3.12, + "learning_rate": 9.824106686614517e-06, + "loss": 0.9867, + "step": 42258 + }, + { + "epoch": 3.12, + "learning_rate": 9.823707876841664e-06, + "loss": 0.9093, + "step": 42259 + }, + { + "epoch": 3.12, + "learning_rate": 9.823309067349286e-06, + "loss": 1.0338, + "step": 42260 + }, + { + "epoch": 3.12, + "learning_rate": 9.822910258138025e-06, + "loss": 0.9617, + "step": 42261 + }, + { + "epoch": 3.12, + "learning_rate": 9.822511449208509e-06, + "loss": 1.0829, + "step": 42262 + }, + { + "epoch": 3.12, + "learning_rate": 9.822112640561375e-06, + "loss": 1.0381, + "step": 42263 + }, + { + "epoch": 3.12, + "learning_rate": 9.821713832197257e-06, + "loss": 0.8482, + "step": 42264 + }, + { + "epoch": 3.12, + "learning_rate": 9.82131502411679e-06, + "loss": 1.0316, + "step": 42265 + }, + { + "epoch": 3.12, + "learning_rate": 9.82091621632061e-06, + "loss": 1.0667, + "step": 42266 + }, + { + "epoch": 3.12, + "learning_rate": 9.820517408809347e-06, + "loss": 0.919, + "step": 42267 + }, + { + "epoch": 3.12, + "learning_rate": 9.82011860158364e-06, + "loss": 0.9761, + "step": 42268 + }, + { + "epoch": 3.12, + "learning_rate": 9.819719794644121e-06, + "loss": 1.0058, + "step": 42269 + }, + { + "epoch": 3.12, + "learning_rate": 9.819320987991426e-06, + "loss": 1.0042, + "step": 42270 + }, + { + "epoch": 3.12, + "learning_rate": 9.818922181626185e-06, + "loss": 0.9602, + "step": 42271 + }, + { + "epoch": 3.12, + "learning_rate": 9.81852337554904e-06, + "loss": 0.9381, + "step": 42272 + }, + { + "epoch": 3.12, + "learning_rate": 9.81812456976062e-06, + "loss": 1.1164, + "step": 42273 + }, + { + "epoch": 3.12, + "learning_rate": 9.817725764261563e-06, + "loss": 0.8527, + "step": 42274 + }, + { + "epoch": 3.12, + "learning_rate": 9.8173269590525e-06, + "loss": 1.0769, + "step": 42275 + }, + { + "epoch": 3.12, + "learning_rate": 9.816928154134068e-06, + "loss": 0.8838, + "step": 42276 + }, + { + "epoch": 3.12, + "learning_rate": 9.816529349506901e-06, + "loss": 0.9796, + "step": 42277 + }, + { + "epoch": 3.12, + "learning_rate": 9.816130545171633e-06, + "loss": 1.0488, + "step": 42278 + }, + { + "epoch": 3.12, + "learning_rate": 9.815731741128899e-06, + "loss": 0.9617, + "step": 42279 + }, + { + "epoch": 3.12, + "learning_rate": 9.815332937379331e-06, + "loss": 0.9147, + "step": 42280 + }, + { + "epoch": 3.12, + "learning_rate": 9.814934133923569e-06, + "loss": 0.9861, + "step": 42281 + }, + { + "epoch": 3.12, + "learning_rate": 9.81453533076224e-06, + "loss": 1.0591, + "step": 42282 + }, + { + "epoch": 3.12, + "learning_rate": 9.814136527895986e-06, + "loss": 1.0189, + "step": 42283 + }, + { + "epoch": 3.12, + "learning_rate": 9.813737725325436e-06, + "loss": 0.9991, + "step": 42284 + }, + { + "epoch": 3.12, + "learning_rate": 9.813338923051229e-06, + "loss": 1.0596, + "step": 42285 + }, + { + "epoch": 3.12, + "learning_rate": 9.812940121073998e-06, + "loss": 1.0037, + "step": 42286 + }, + { + "epoch": 3.12, + "learning_rate": 9.812541319394371e-06, + "loss": 0.9404, + "step": 42287 + }, + { + "epoch": 3.12, + "learning_rate": 9.812142518012994e-06, + "loss": 0.9465, + "step": 42288 + }, + { + "epoch": 3.12, + "learning_rate": 9.811743716930491e-06, + "loss": 0.9971, + "step": 42289 + }, + { + "epoch": 3.12, + "learning_rate": 9.811344916147506e-06, + "loss": 1.0246, + "step": 42290 + }, + { + "epoch": 3.12, + "learning_rate": 9.810946115664662e-06, + "loss": 1.0541, + "step": 42291 + }, + { + "epoch": 3.12, + "learning_rate": 9.810547315482607e-06, + "loss": 0.9691, + "step": 42292 + }, + { + "epoch": 3.13, + "learning_rate": 9.810148515601961e-06, + "loss": 0.9903, + "step": 42293 + }, + { + "epoch": 3.13, + "learning_rate": 9.809749716023374e-06, + "loss": 0.9101, + "step": 42294 + }, + { + "epoch": 3.13, + "learning_rate": 9.80935091674747e-06, + "loss": 1.0956, + "step": 42295 + }, + { + "epoch": 3.13, + "learning_rate": 9.808952117774882e-06, + "loss": 0.9758, + "step": 42296 + }, + { + "epoch": 3.13, + "learning_rate": 9.808553319106254e-06, + "loss": 0.9708, + "step": 42297 + }, + { + "epoch": 3.13, + "learning_rate": 9.80815452074221e-06, + "loss": 0.9659, + "step": 42298 + }, + { + "epoch": 3.13, + "learning_rate": 9.807755722683393e-06, + "loss": 0.9017, + "step": 42299 + }, + { + "epoch": 3.13, + "learning_rate": 9.80735692493043e-06, + "loss": 1.0458, + "step": 42300 + }, + { + "epoch": 3.13, + "learning_rate": 9.806958127483962e-06, + "loss": 1.0864, + "step": 42301 + }, + { + "epoch": 3.13, + "learning_rate": 9.80655933034462e-06, + "loss": 1.0447, + "step": 42302 + }, + { + "epoch": 3.13, + "learning_rate": 9.806160533513041e-06, + "loss": 0.9754, + "step": 42303 + }, + { + "epoch": 3.13, + "learning_rate": 9.805761736989857e-06, + "loss": 1.011, + "step": 42304 + }, + { + "epoch": 3.13, + "learning_rate": 9.805362940775702e-06, + "loss": 0.9388, + "step": 42305 + }, + { + "epoch": 3.13, + "learning_rate": 9.804964144871215e-06, + "loss": 0.9089, + "step": 42306 + }, + { + "epoch": 3.13, + "learning_rate": 9.804565349277023e-06, + "loss": 0.9945, + "step": 42307 + }, + { + "epoch": 3.13, + "learning_rate": 9.804166553993766e-06, + "loss": 0.939, + "step": 42308 + }, + { + "epoch": 3.13, + "learning_rate": 9.803767759022078e-06, + "loss": 1.0129, + "step": 42309 + }, + { + "epoch": 3.13, + "learning_rate": 9.803368964362595e-06, + "loss": 1.003, + "step": 42310 + }, + { + "epoch": 3.13, + "learning_rate": 9.802970170015944e-06, + "loss": 0.9993, + "step": 42311 + }, + { + "epoch": 3.13, + "learning_rate": 9.80257137598277e-06, + "loss": 0.9487, + "step": 42312 + }, + { + "epoch": 3.13, + "learning_rate": 9.802172582263701e-06, + "loss": 0.9049, + "step": 42313 + }, + { + "epoch": 3.13, + "learning_rate": 9.80177378885937e-06, + "loss": 1.1105, + "step": 42314 + }, + { + "epoch": 3.13, + "learning_rate": 9.801374995770416e-06, + "loss": 1.0371, + "step": 42315 + }, + { + "epoch": 3.13, + "learning_rate": 9.800976202997469e-06, + "loss": 1.1254, + "step": 42316 + }, + { + "epoch": 3.13, + "learning_rate": 9.80057741054117e-06, + "loss": 0.9655, + "step": 42317 + }, + { + "epoch": 3.13, + "learning_rate": 9.800178618402146e-06, + "loss": 1.0095, + "step": 42318 + }, + { + "epoch": 3.13, + "learning_rate": 9.799779826581037e-06, + "loss": 0.9086, + "step": 42319 + }, + { + "epoch": 3.13, + "learning_rate": 9.799381035078473e-06, + "loss": 1.0018, + "step": 42320 + }, + { + "epoch": 3.13, + "learning_rate": 9.798982243895093e-06, + "loss": 0.9432, + "step": 42321 + }, + { + "epoch": 3.13, + "learning_rate": 9.79858345303153e-06, + "loss": 1.0184, + "step": 42322 + }, + { + "epoch": 3.13, + "learning_rate": 9.798184662488415e-06, + "loss": 0.9135, + "step": 42323 + }, + { + "epoch": 3.13, + "learning_rate": 9.79778587226639e-06, + "loss": 1.0026, + "step": 42324 + }, + { + "epoch": 3.13, + "learning_rate": 9.79738708236608e-06, + "loss": 0.9093, + "step": 42325 + }, + { + "epoch": 3.13, + "learning_rate": 9.796988292788126e-06, + "loss": 0.9944, + "step": 42326 + }, + { + "epoch": 3.13, + "learning_rate": 9.79658950353316e-06, + "loss": 1.0465, + "step": 42327 + }, + { + "epoch": 3.13, + "learning_rate": 9.796190714601818e-06, + "loss": 1.1098, + "step": 42328 + }, + { + "epoch": 3.13, + "learning_rate": 9.795791925994733e-06, + "loss": 1.0709, + "step": 42329 + }, + { + "epoch": 3.13, + "learning_rate": 9.795393137712543e-06, + "loss": 1.0017, + "step": 42330 + }, + { + "epoch": 3.13, + "learning_rate": 9.794994349755876e-06, + "loss": 0.9376, + "step": 42331 + }, + { + "epoch": 3.13, + "learning_rate": 9.794595562125373e-06, + "loss": 0.9276, + "step": 42332 + }, + { + "epoch": 3.13, + "learning_rate": 9.794196774821667e-06, + "loss": 0.9961, + "step": 42333 + }, + { + "epoch": 3.13, + "learning_rate": 9.793797987845384e-06, + "loss": 0.9798, + "step": 42334 + }, + { + "epoch": 3.13, + "learning_rate": 9.793399201197172e-06, + "loss": 0.9515, + "step": 42335 + }, + { + "epoch": 3.13, + "learning_rate": 9.793000414877653e-06, + "loss": 0.9221, + "step": 42336 + }, + { + "epoch": 3.13, + "learning_rate": 9.792601628887473e-06, + "loss": 1.0725, + "step": 42337 + }, + { + "epoch": 3.13, + "learning_rate": 9.792202843227257e-06, + "loss": 1.0291, + "step": 42338 + }, + { + "epoch": 3.13, + "learning_rate": 9.791804057897647e-06, + "loss": 1.0168, + "step": 42339 + }, + { + "epoch": 3.13, + "learning_rate": 9.791405272899268e-06, + "loss": 0.9898, + "step": 42340 + }, + { + "epoch": 3.13, + "learning_rate": 9.791006488232767e-06, + "loss": 1.0445, + "step": 42341 + }, + { + "epoch": 3.13, + "learning_rate": 9.79060770389877e-06, + "loss": 0.9554, + "step": 42342 + }, + { + "epoch": 3.13, + "learning_rate": 9.79020891989791e-06, + "loss": 0.92, + "step": 42343 + }, + { + "epoch": 3.13, + "learning_rate": 9.789810136230828e-06, + "loss": 1.0237, + "step": 42344 + }, + { + "epoch": 3.13, + "learning_rate": 9.789411352898151e-06, + "loss": 1.0835, + "step": 42345 + }, + { + "epoch": 3.13, + "learning_rate": 9.789012569900521e-06, + "loss": 1.0016, + "step": 42346 + }, + { + "epoch": 3.13, + "learning_rate": 9.788613787238566e-06, + "loss": 0.8624, + "step": 42347 + }, + { + "epoch": 3.13, + "learning_rate": 9.788215004912928e-06, + "loss": 1.0687, + "step": 42348 + }, + { + "epoch": 3.13, + "learning_rate": 9.787816222924232e-06, + "loss": 0.9692, + "step": 42349 + }, + { + "epoch": 3.13, + "learning_rate": 9.78741744127312e-06, + "loss": 0.9975, + "step": 42350 + }, + { + "epoch": 3.13, + "learning_rate": 9.787018659960224e-06, + "loss": 1.0663, + "step": 42351 + }, + { + "epoch": 3.13, + "learning_rate": 9.786619878986178e-06, + "loss": 0.982, + "step": 42352 + }, + { + "epoch": 3.13, + "learning_rate": 9.786221098351617e-06, + "loss": 0.932, + "step": 42353 + }, + { + "epoch": 3.13, + "learning_rate": 9.785822318057172e-06, + "loss": 0.9332, + "step": 42354 + }, + { + "epoch": 3.13, + "learning_rate": 9.785423538103486e-06, + "loss": 1.0261, + "step": 42355 + }, + { + "epoch": 3.13, + "learning_rate": 9.785024758491183e-06, + "loss": 0.985, + "step": 42356 + }, + { + "epoch": 3.13, + "learning_rate": 9.784625979220907e-06, + "loss": 0.9692, + "step": 42357 + }, + { + "epoch": 3.13, + "learning_rate": 9.784227200293284e-06, + "loss": 0.9942, + "step": 42358 + }, + { + "epoch": 3.13, + "learning_rate": 9.783828421708955e-06, + "loss": 0.9488, + "step": 42359 + }, + { + "epoch": 3.13, + "learning_rate": 9.783429643468554e-06, + "loss": 0.9947, + "step": 42360 + }, + { + "epoch": 3.13, + "learning_rate": 9.78303086557271e-06, + "loss": 0.9364, + "step": 42361 + }, + { + "epoch": 3.13, + "learning_rate": 9.782632088022062e-06, + "loss": 1.0205, + "step": 42362 + }, + { + "epoch": 3.13, + "learning_rate": 9.78223331081724e-06, + "loss": 0.9452, + "step": 42363 + }, + { + "epoch": 3.13, + "learning_rate": 9.781834533958885e-06, + "loss": 1.0919, + "step": 42364 + }, + { + "epoch": 3.13, + "learning_rate": 9.781435757447626e-06, + "loss": 0.8587, + "step": 42365 + }, + { + "epoch": 3.13, + "learning_rate": 9.781036981284102e-06, + "loss": 1.0, + "step": 42366 + }, + { + "epoch": 3.13, + "learning_rate": 9.780638205468942e-06, + "loss": 1.006, + "step": 42367 + }, + { + "epoch": 3.13, + "learning_rate": 9.780239430002788e-06, + "loss": 0.9793, + "step": 42368 + }, + { + "epoch": 3.13, + "learning_rate": 9.779840654886266e-06, + "loss": 1.0305, + "step": 42369 + }, + { + "epoch": 3.13, + "learning_rate": 9.779441880120015e-06, + "loss": 0.9568, + "step": 42370 + }, + { + "epoch": 3.13, + "learning_rate": 9.77904310570467e-06, + "loss": 1.0103, + "step": 42371 + }, + { + "epoch": 3.13, + "learning_rate": 9.778644331640863e-06, + "loss": 1.0417, + "step": 42372 + }, + { + "epoch": 3.13, + "learning_rate": 9.778245557929232e-06, + "loss": 0.9443, + "step": 42373 + }, + { + "epoch": 3.13, + "learning_rate": 9.777846784570406e-06, + "loss": 1.0085, + "step": 42374 + }, + { + "epoch": 3.13, + "learning_rate": 9.777448011565026e-06, + "loss": 0.9847, + "step": 42375 + }, + { + "epoch": 3.13, + "learning_rate": 9.777049238913719e-06, + "loss": 0.9483, + "step": 42376 + }, + { + "epoch": 3.13, + "learning_rate": 9.776650466617127e-06, + "loss": 0.9982, + "step": 42377 + }, + { + "epoch": 3.13, + "learning_rate": 9.776251694675882e-06, + "loss": 1.0025, + "step": 42378 + }, + { + "epoch": 3.13, + "learning_rate": 9.775852923090612e-06, + "loss": 0.9954, + "step": 42379 + }, + { + "epoch": 3.13, + "learning_rate": 9.775454151861963e-06, + "loss": 1.0369, + "step": 42380 + }, + { + "epoch": 3.13, + "learning_rate": 9.775055380990555e-06, + "loss": 0.9299, + "step": 42381 + }, + { + "epoch": 3.13, + "learning_rate": 9.77465661047704e-06, + "loss": 1.0887, + "step": 42382 + }, + { + "epoch": 3.13, + "learning_rate": 9.774257840322035e-06, + "loss": 0.9949, + "step": 42383 + }, + { + "epoch": 3.13, + "learning_rate": 9.77385907052619e-06, + "loss": 1.0259, + "step": 42384 + }, + { + "epoch": 3.13, + "learning_rate": 9.773460301090124e-06, + "loss": 1.0217, + "step": 42385 + }, + { + "epoch": 3.13, + "learning_rate": 9.773061532014487e-06, + "loss": 1.0169, + "step": 42386 + }, + { + "epoch": 3.13, + "learning_rate": 9.772662763299902e-06, + "loss": 0.9936, + "step": 42387 + }, + { + "epoch": 3.13, + "learning_rate": 9.772263994947009e-06, + "loss": 1.0265, + "step": 42388 + }, + { + "epoch": 3.13, + "learning_rate": 9.77186522695644e-06, + "loss": 0.9638, + "step": 42389 + }, + { + "epoch": 3.13, + "learning_rate": 9.771466459328828e-06, + "loss": 0.8759, + "step": 42390 + }, + { + "epoch": 3.13, + "learning_rate": 9.771067692064812e-06, + "loss": 1.0781, + "step": 42391 + }, + { + "epoch": 3.13, + "learning_rate": 9.770668925165022e-06, + "loss": 0.9291, + "step": 42392 + }, + { + "epoch": 3.13, + "learning_rate": 9.770270158630097e-06, + "loss": 1.0278, + "step": 42393 + }, + { + "epoch": 3.13, + "learning_rate": 9.769871392460666e-06, + "loss": 0.9222, + "step": 42394 + }, + { + "epoch": 3.13, + "learning_rate": 9.769472626657369e-06, + "loss": 1.0586, + "step": 42395 + }, + { + "epoch": 3.13, + "learning_rate": 9.769073861220836e-06, + "loss": 0.932, + "step": 42396 + }, + { + "epoch": 3.13, + "learning_rate": 9.768675096151706e-06, + "loss": 1.077, + "step": 42397 + }, + { + "epoch": 3.13, + "learning_rate": 9.76827633145061e-06, + "loss": 1.0174, + "step": 42398 + }, + { + "epoch": 3.13, + "learning_rate": 9.76787756711818e-06, + "loss": 1.0511, + "step": 42399 + }, + { + "epoch": 3.13, + "learning_rate": 9.767478803155057e-06, + "loss": 1.0426, + "step": 42400 + }, + { + "epoch": 3.13, + "learning_rate": 9.767080039561867e-06, + "loss": 0.994, + "step": 42401 + }, + { + "epoch": 3.13, + "learning_rate": 9.766681276339255e-06, + "loss": 0.9043, + "step": 42402 + }, + { + "epoch": 3.13, + "learning_rate": 9.766282513487843e-06, + "loss": 1.0396, + "step": 42403 + }, + { + "epoch": 3.13, + "learning_rate": 9.765883751008279e-06, + "loss": 0.8576, + "step": 42404 + }, + { + "epoch": 3.13, + "learning_rate": 9.765484988901186e-06, + "loss": 1.0134, + "step": 42405 + }, + { + "epoch": 3.13, + "learning_rate": 9.765086227167207e-06, + "loss": 0.9562, + "step": 42406 + }, + { + "epoch": 3.13, + "learning_rate": 9.764687465806971e-06, + "loss": 0.9079, + "step": 42407 + }, + { + "epoch": 3.13, + "learning_rate": 9.764288704821112e-06, + "loss": 1.0323, + "step": 42408 + }, + { + "epoch": 3.13, + "learning_rate": 9.763889944210269e-06, + "loss": 0.9734, + "step": 42409 + }, + { + "epoch": 3.13, + "learning_rate": 9.763491183975071e-06, + "loss": 1.0519, + "step": 42410 + }, + { + "epoch": 3.13, + "learning_rate": 9.763092424116157e-06, + "loss": 0.949, + "step": 42411 + }, + { + "epoch": 3.13, + "learning_rate": 9.762693664634158e-06, + "loss": 1.0368, + "step": 42412 + }, + { + "epoch": 3.13, + "learning_rate": 9.762294905529712e-06, + "loss": 1.0127, + "step": 42413 + }, + { + "epoch": 3.13, + "learning_rate": 9.761896146803447e-06, + "loss": 0.9966, + "step": 42414 + }, + { + "epoch": 3.13, + "learning_rate": 9.761497388456008e-06, + "loss": 1.0422, + "step": 42415 + }, + { + "epoch": 3.13, + "learning_rate": 9.76109863048802e-06, + "loss": 1.0556, + "step": 42416 + }, + { + "epoch": 3.13, + "learning_rate": 9.76069987290012e-06, + "loss": 1.0378, + "step": 42417 + }, + { + "epoch": 3.13, + "learning_rate": 9.760301115692945e-06, + "loss": 0.9419, + "step": 42418 + }, + { + "epoch": 3.13, + "learning_rate": 9.759902358867126e-06, + "loss": 0.9276, + "step": 42419 + }, + { + "epoch": 3.13, + "learning_rate": 9.7595036024233e-06, + "loss": 1.074, + "step": 42420 + }, + { + "epoch": 3.13, + "learning_rate": 9.7591048463621e-06, + "loss": 1.0002, + "step": 42421 + }, + { + "epoch": 3.13, + "learning_rate": 9.75870609068416e-06, + "loss": 1.0422, + "step": 42422 + }, + { + "epoch": 3.13, + "learning_rate": 9.758307335390114e-06, + "loss": 1.0213, + "step": 42423 + }, + { + "epoch": 3.13, + "learning_rate": 9.7579085804806e-06, + "loss": 1.0503, + "step": 42424 + }, + { + "epoch": 3.13, + "learning_rate": 9.757509825956252e-06, + "loss": 1.0086, + "step": 42425 + }, + { + "epoch": 3.13, + "learning_rate": 9.757111071817698e-06, + "loss": 1.0377, + "step": 42426 + }, + { + "epoch": 3.13, + "learning_rate": 9.756712318065582e-06, + "loss": 0.9744, + "step": 42427 + }, + { + "epoch": 3.13, + "learning_rate": 9.756313564700526e-06, + "loss": 1.0239, + "step": 42428 + }, + { + "epoch": 3.14, + "learning_rate": 9.755914811723179e-06, + "loss": 1.0328, + "step": 42429 + }, + { + "epoch": 3.14, + "learning_rate": 9.75551605913416e-06, + "loss": 0.9322, + "step": 42430 + }, + { + "epoch": 3.14, + "learning_rate": 9.755117306934119e-06, + "loss": 1.0539, + "step": 42431 + }, + { + "epoch": 3.14, + "learning_rate": 9.754718555123678e-06, + "loss": 0.954, + "step": 42432 + }, + { + "epoch": 3.14, + "learning_rate": 9.75431980370348e-06, + "loss": 0.9078, + "step": 42433 + }, + { + "epoch": 3.14, + "learning_rate": 9.753921052674154e-06, + "loss": 1.0626, + "step": 42434 + }, + { + "epoch": 3.14, + "learning_rate": 9.753522302036334e-06, + "loss": 0.9484, + "step": 42435 + }, + { + "epoch": 3.14, + "learning_rate": 9.75312355179066e-06, + "loss": 0.9565, + "step": 42436 + }, + { + "epoch": 3.14, + "learning_rate": 9.75272480193776e-06, + "loss": 0.9106, + "step": 42437 + }, + { + "epoch": 3.14, + "learning_rate": 9.752326052478273e-06, + "loss": 1.0384, + "step": 42438 + }, + { + "epoch": 3.14, + "learning_rate": 9.751927303412829e-06, + "loss": 1.0718, + "step": 42439 + }, + { + "epoch": 3.14, + "learning_rate": 9.751528554742068e-06, + "loss": 1.0014, + "step": 42440 + }, + { + "epoch": 3.14, + "learning_rate": 9.75112980646662e-06, + "loss": 1.0282, + "step": 42441 + }, + { + "epoch": 3.14, + "learning_rate": 9.750731058587122e-06, + "loss": 1.0172, + "step": 42442 + }, + { + "epoch": 3.14, + "learning_rate": 9.750332311104205e-06, + "loss": 1.0003, + "step": 42443 + }, + { + "epoch": 3.14, + "learning_rate": 9.74993356401851e-06, + "loss": 1.0695, + "step": 42444 + }, + { + "epoch": 3.14, + "learning_rate": 9.749534817330668e-06, + "loss": 0.8706, + "step": 42445 + }, + { + "epoch": 3.14, + "learning_rate": 9.749136071041305e-06, + "loss": 0.9473, + "step": 42446 + }, + { + "epoch": 3.14, + "learning_rate": 9.748737325151071e-06, + "loss": 0.9471, + "step": 42447 + }, + { + "epoch": 3.14, + "learning_rate": 9.748338579660586e-06, + "loss": 1.0487, + "step": 42448 + }, + { + "epoch": 3.14, + "learning_rate": 9.747939834570496e-06, + "loss": 0.9685, + "step": 42449 + }, + { + "epoch": 3.14, + "learning_rate": 9.747541089881424e-06, + "loss": 0.9838, + "step": 42450 + }, + { + "epoch": 3.14, + "learning_rate": 9.747142345594017e-06, + "loss": 1.0417, + "step": 42451 + }, + { + "epoch": 3.14, + "learning_rate": 9.746743601708896e-06, + "loss": 0.8739, + "step": 42452 + }, + { + "epoch": 3.14, + "learning_rate": 9.74634485822671e-06, + "loss": 0.9464, + "step": 42453 + }, + { + "epoch": 3.14, + "learning_rate": 9.745946115148083e-06, + "loss": 1.0212, + "step": 42454 + }, + { + "epoch": 3.14, + "learning_rate": 9.745547372473649e-06, + "loss": 0.856, + "step": 42455 + }, + { + "epoch": 3.14, + "learning_rate": 9.745148630204048e-06, + "loss": 1.0213, + "step": 42456 + }, + { + "epoch": 3.14, + "learning_rate": 9.74474988833991e-06, + "loss": 1.0312, + "step": 42457 + }, + { + "epoch": 3.14, + "learning_rate": 9.744351146881875e-06, + "loss": 0.9763, + "step": 42458 + }, + { + "epoch": 3.14, + "learning_rate": 9.743952405830569e-06, + "loss": 1.0095, + "step": 42459 + }, + { + "epoch": 3.14, + "learning_rate": 9.743553665186635e-06, + "loss": 1.0578, + "step": 42460 + }, + { + "epoch": 3.14, + "learning_rate": 9.7431549249507e-06, + "loss": 1.016, + "step": 42461 + }, + { + "epoch": 3.14, + "learning_rate": 9.742756185123405e-06, + "loss": 1.0521, + "step": 42462 + }, + { + "epoch": 3.14, + "learning_rate": 9.74235744570538e-06, + "loss": 0.9494, + "step": 42463 + }, + { + "epoch": 3.14, + "learning_rate": 9.741958706697258e-06, + "loss": 1.0168, + "step": 42464 + }, + { + "epoch": 3.14, + "learning_rate": 9.741559968099681e-06, + "loss": 0.9733, + "step": 42465 + }, + { + "epoch": 3.14, + "learning_rate": 9.741161229913274e-06, + "loss": 1.0449, + "step": 42466 + }, + { + "epoch": 3.14, + "learning_rate": 9.74076249213868e-06, + "loss": 0.991, + "step": 42467 + }, + { + "epoch": 3.14, + "learning_rate": 9.740363754776524e-06, + "loss": 1.0267, + "step": 42468 + }, + { + "epoch": 3.14, + "learning_rate": 9.739965017827449e-06, + "loss": 0.919, + "step": 42469 + }, + { + "epoch": 3.14, + "learning_rate": 9.739566281292084e-06, + "loss": 1.0243, + "step": 42470 + }, + { + "epoch": 3.14, + "learning_rate": 9.739167545171069e-06, + "loss": 1.0199, + "step": 42471 + }, + { + "epoch": 3.14, + "learning_rate": 9.738768809465035e-06, + "loss": 0.9022, + "step": 42472 + }, + { + "epoch": 3.14, + "learning_rate": 9.73837007417461e-06, + "loss": 0.9638, + "step": 42473 + }, + { + "epoch": 3.14, + "learning_rate": 9.737971339300441e-06, + "loss": 1.1095, + "step": 42474 + }, + { + "epoch": 3.14, + "learning_rate": 9.73757260484315e-06, + "loss": 1.0992, + "step": 42475 + }, + { + "epoch": 3.14, + "learning_rate": 9.737173870803382e-06, + "loss": 1.0808, + "step": 42476 + }, + { + "epoch": 3.14, + "learning_rate": 9.736775137181763e-06, + "loss": 1.0861, + "step": 42477 + }, + { + "epoch": 3.14, + "learning_rate": 9.736376403978933e-06, + "loss": 0.9264, + "step": 42478 + }, + { + "epoch": 3.14, + "learning_rate": 9.735977671195523e-06, + "loss": 1.1327, + "step": 42479 + }, + { + "epoch": 3.14, + "learning_rate": 9.735578938832171e-06, + "loss": 1.0154, + "step": 42480 + }, + { + "epoch": 3.14, + "learning_rate": 9.735180206889508e-06, + "loss": 0.9747, + "step": 42481 + }, + { + "epoch": 3.14, + "learning_rate": 9.73478147536817e-06, + "loss": 0.9676, + "step": 42482 + }, + { + "epoch": 3.14, + "learning_rate": 9.73438274426879e-06, + "loss": 1.053, + "step": 42483 + }, + { + "epoch": 3.14, + "learning_rate": 9.733984013592002e-06, + "loss": 1.0784, + "step": 42484 + }, + { + "epoch": 3.14, + "learning_rate": 9.733585283338443e-06, + "loss": 1.1037, + "step": 42485 + }, + { + "epoch": 3.14, + "learning_rate": 9.733186553508745e-06, + "loss": 0.8922, + "step": 42486 + }, + { + "epoch": 3.14, + "learning_rate": 9.732787824103547e-06, + "loss": 1.0024, + "step": 42487 + }, + { + "epoch": 3.14, + "learning_rate": 9.732389095123476e-06, + "loss": 0.9409, + "step": 42488 + }, + { + "epoch": 3.14, + "learning_rate": 9.731990366569172e-06, + "loss": 0.9707, + "step": 42489 + }, + { + "epoch": 3.14, + "learning_rate": 9.731591638441264e-06, + "loss": 0.9883, + "step": 42490 + }, + { + "epoch": 3.14, + "learning_rate": 9.731192910740395e-06, + "loss": 1.0408, + "step": 42491 + }, + { + "epoch": 3.14, + "learning_rate": 9.730794183467194e-06, + "loss": 1.0447, + "step": 42492 + }, + { + "epoch": 3.14, + "learning_rate": 9.730395456622291e-06, + "loss": 0.97, + "step": 42493 + }, + { + "epoch": 3.14, + "learning_rate": 9.72999673020633e-06, + "loss": 0.9581, + "step": 42494 + }, + { + "epoch": 3.14, + "learning_rate": 9.729598004219934e-06, + "loss": 1.0228, + "step": 42495 + }, + { + "epoch": 3.14, + "learning_rate": 9.729199278663752e-06, + "loss": 0.9983, + "step": 42496 + }, + { + "epoch": 3.14, + "learning_rate": 9.728800553538402e-06, + "loss": 0.9527, + "step": 42497 + }, + { + "epoch": 3.14, + "learning_rate": 9.728401828844532e-06, + "loss": 1.0455, + "step": 42498 + }, + { + "epoch": 3.14, + "learning_rate": 9.728003104582766e-06, + "loss": 0.9225, + "step": 42499 + }, + { + "epoch": 3.14, + "learning_rate": 9.727604380753748e-06, + "loss": 0.9768, + "step": 42500 + }, + { + "epoch": 3.14, + "learning_rate": 9.727205657358107e-06, + "loss": 0.9473, + "step": 42501 + }, + { + "epoch": 3.14, + "learning_rate": 9.726806934396476e-06, + "loss": 0.9708, + "step": 42502 + }, + { + "epoch": 3.14, + "learning_rate": 9.726408211869492e-06, + "loss": 0.9725, + "step": 42503 + }, + { + "epoch": 3.14, + "learning_rate": 9.726009489777789e-06, + "loss": 1.0411, + "step": 42504 + }, + { + "epoch": 3.14, + "learning_rate": 9.725610768121999e-06, + "loss": 1.0655, + "step": 42505 + }, + { + "epoch": 3.14, + "learning_rate": 9.72521204690276e-06, + "loss": 0.9123, + "step": 42506 + }, + { + "epoch": 3.14, + "learning_rate": 9.724813326120705e-06, + "loss": 1.0692, + "step": 42507 + }, + { + "epoch": 3.14, + "learning_rate": 9.724414605776466e-06, + "loss": 1.0555, + "step": 42508 + }, + { + "epoch": 3.14, + "learning_rate": 9.724015885870682e-06, + "loss": 1.0226, + "step": 42509 + }, + { + "epoch": 3.14, + "learning_rate": 9.723617166403984e-06, + "loss": 1.0091, + "step": 42510 + }, + { + "epoch": 3.14, + "learning_rate": 9.723218447377005e-06, + "loss": 1.1086, + "step": 42511 + }, + { + "epoch": 3.14, + "learning_rate": 9.722819728790385e-06, + "loss": 1.0777, + "step": 42512 + }, + { + "epoch": 3.14, + "learning_rate": 9.722421010644752e-06, + "loss": 0.984, + "step": 42513 + }, + { + "epoch": 3.14, + "learning_rate": 9.722022292940744e-06, + "loss": 1.1341, + "step": 42514 + }, + { + "epoch": 3.14, + "learning_rate": 9.721623575678993e-06, + "loss": 1.0251, + "step": 42515 + }, + { + "epoch": 3.14, + "learning_rate": 9.72122485886014e-06, + "loss": 1.0003, + "step": 42516 + }, + { + "epoch": 3.14, + "learning_rate": 9.720826142484809e-06, + "loss": 0.8745, + "step": 42517 + }, + { + "epoch": 3.14, + "learning_rate": 9.72042742655364e-06, + "loss": 1.0469, + "step": 42518 + }, + { + "epoch": 3.14, + "learning_rate": 9.720028711067273e-06, + "loss": 1.1162, + "step": 42519 + }, + { + "epoch": 3.14, + "learning_rate": 9.719629996026331e-06, + "loss": 0.9544, + "step": 42520 + }, + { + "epoch": 3.14, + "learning_rate": 9.719231281431455e-06, + "loss": 0.9433, + "step": 42521 + }, + { + "epoch": 3.14, + "learning_rate": 9.718832567283278e-06, + "loss": 0.931, + "step": 42522 + }, + { + "epoch": 3.14, + "learning_rate": 9.718433853582433e-06, + "loss": 0.9388, + "step": 42523 + }, + { + "epoch": 3.14, + "learning_rate": 9.718035140329556e-06, + "loss": 0.9214, + "step": 42524 + }, + { + "epoch": 3.14, + "learning_rate": 9.717636427525282e-06, + "loss": 1.1351, + "step": 42525 + }, + { + "epoch": 3.14, + "learning_rate": 9.717237715170243e-06, + "loss": 1.0428, + "step": 42526 + }, + { + "epoch": 3.14, + "learning_rate": 9.716839003265076e-06, + "loss": 0.9317, + "step": 42527 + }, + { + "epoch": 3.14, + "learning_rate": 9.716440291810416e-06, + "loss": 1.0366, + "step": 42528 + }, + { + "epoch": 3.14, + "learning_rate": 9.716041580806892e-06, + "loss": 1.0256, + "step": 42529 + }, + { + "epoch": 3.14, + "learning_rate": 9.715642870255144e-06, + "loss": 1.0714, + "step": 42530 + }, + { + "epoch": 3.14, + "learning_rate": 9.715244160155802e-06, + "loss": 1.0422, + "step": 42531 + }, + { + "epoch": 3.14, + "learning_rate": 9.714845450509505e-06, + "loss": 0.9842, + "step": 42532 + }, + { + "epoch": 3.14, + "learning_rate": 9.714446741316882e-06, + "loss": 1.067, + "step": 42533 + }, + { + "epoch": 3.14, + "learning_rate": 9.714048032578573e-06, + "loss": 0.9108, + "step": 42534 + }, + { + "epoch": 3.14, + "learning_rate": 9.713649324295207e-06, + "loss": 1.0371, + "step": 42535 + }, + { + "epoch": 3.14, + "learning_rate": 9.713250616467422e-06, + "loss": 1.0104, + "step": 42536 + }, + { + "epoch": 3.14, + "learning_rate": 9.712851909095855e-06, + "loss": 1.0326, + "step": 42537 + }, + { + "epoch": 3.14, + "learning_rate": 9.712453202181129e-06, + "loss": 1.0273, + "step": 42538 + }, + { + "epoch": 3.14, + "learning_rate": 9.712054495723893e-06, + "loss": 1.0418, + "step": 42539 + }, + { + "epoch": 3.14, + "learning_rate": 9.711655789724767e-06, + "loss": 1.0517, + "step": 42540 + }, + { + "epoch": 3.14, + "learning_rate": 9.7112570841844e-06, + "loss": 1.0618, + "step": 42541 + }, + { + "epoch": 3.14, + "learning_rate": 9.710858379103411e-06, + "loss": 0.9791, + "step": 42542 + }, + { + "epoch": 3.14, + "learning_rate": 9.71045967448245e-06, + "loss": 0.9744, + "step": 42543 + }, + { + "epoch": 3.14, + "learning_rate": 9.710060970322136e-06, + "loss": 0.993, + "step": 42544 + }, + { + "epoch": 3.14, + "learning_rate": 9.709662266623118e-06, + "loss": 0.8964, + "step": 42545 + }, + { + "epoch": 3.14, + "learning_rate": 9.709263563386015e-06, + "loss": 0.9652, + "step": 42546 + }, + { + "epoch": 3.14, + "learning_rate": 9.70886486061148e-06, + "loss": 1.0467, + "step": 42547 + }, + { + "epoch": 3.14, + "learning_rate": 9.70846615830013e-06, + "loss": 0.9161, + "step": 42548 + }, + { + "epoch": 3.14, + "learning_rate": 9.708067456452607e-06, + "loss": 1.0631, + "step": 42549 + }, + { + "epoch": 3.14, + "learning_rate": 9.707668755069546e-06, + "loss": 0.9585, + "step": 42550 + }, + { + "epoch": 3.14, + "learning_rate": 9.707270054151576e-06, + "loss": 0.933, + "step": 42551 + }, + { + "epoch": 3.14, + "learning_rate": 9.70687135369934e-06, + "loss": 1.027, + "step": 42552 + }, + { + "epoch": 3.14, + "learning_rate": 9.706472653713463e-06, + "loss": 0.9615, + "step": 42553 + }, + { + "epoch": 3.14, + "learning_rate": 9.706073954194588e-06, + "loss": 1.0285, + "step": 42554 + }, + { + "epoch": 3.14, + "learning_rate": 9.705675255143341e-06, + "loss": 1.0096, + "step": 42555 + }, + { + "epoch": 3.14, + "learning_rate": 9.705276556560365e-06, + "loss": 0.9198, + "step": 42556 + }, + { + "epoch": 3.14, + "learning_rate": 9.704877858446287e-06, + "loss": 1.0482, + "step": 42557 + }, + { + "epoch": 3.14, + "learning_rate": 9.704479160801744e-06, + "loss": 0.9409, + "step": 42558 + }, + { + "epoch": 3.14, + "learning_rate": 9.704080463627373e-06, + "loss": 1.0713, + "step": 42559 + }, + { + "epoch": 3.14, + "learning_rate": 9.703681766923801e-06, + "loss": 0.9608, + "step": 42560 + }, + { + "epoch": 3.14, + "learning_rate": 9.703283070691672e-06, + "loss": 0.9395, + "step": 42561 + }, + { + "epoch": 3.14, + "learning_rate": 9.702884374931612e-06, + "loss": 1.0629, + "step": 42562 + }, + { + "epoch": 3.14, + "learning_rate": 9.70248567964426e-06, + "loss": 1.0102, + "step": 42563 + }, + { + "epoch": 3.15, + "learning_rate": 9.702086984830247e-06, + "loss": 0.9956, + "step": 42564 + }, + { + "epoch": 3.15, + "learning_rate": 9.701688290490215e-06, + "loss": 0.8685, + "step": 42565 + }, + { + "epoch": 3.15, + "learning_rate": 9.70128959662479e-06, + "loss": 1.0279, + "step": 42566 + }, + { + "epoch": 3.15, + "learning_rate": 9.700890903234606e-06, + "loss": 1.057, + "step": 42567 + }, + { + "epoch": 3.15, + "learning_rate": 9.700492210320303e-06, + "loss": 1.0273, + "step": 42568 + }, + { + "epoch": 3.15, + "learning_rate": 9.700093517882512e-06, + "loss": 0.9782, + "step": 42569 + }, + { + "epoch": 3.15, + "learning_rate": 9.699694825921869e-06, + "loss": 1.0005, + "step": 42570 + }, + { + "epoch": 3.15, + "learning_rate": 9.699296134439004e-06, + "loss": 0.9459, + "step": 42571 + }, + { + "epoch": 3.15, + "learning_rate": 9.698897443434558e-06, + "loss": 0.9835, + "step": 42572 + }, + { + "epoch": 3.15, + "learning_rate": 9.698498752909159e-06, + "loss": 1.0767, + "step": 42573 + }, + { + "epoch": 3.15, + "learning_rate": 9.698100062863446e-06, + "loss": 0.9878, + "step": 42574 + }, + { + "epoch": 3.15, + "learning_rate": 9.697701373298052e-06, + "loss": 1.0159, + "step": 42575 + }, + { + "epoch": 3.15, + "learning_rate": 9.69730268421361e-06, + "loss": 1.1103, + "step": 42576 + }, + { + "epoch": 3.15, + "learning_rate": 9.696903995610754e-06, + "loss": 1.0717, + "step": 42577 + }, + { + "epoch": 3.15, + "learning_rate": 9.696505307490119e-06, + "loss": 0.9945, + "step": 42578 + }, + { + "epoch": 3.15, + "learning_rate": 9.696106619852343e-06, + "loss": 1.0195, + "step": 42579 + }, + { + "epoch": 3.15, + "learning_rate": 9.695707932698053e-06, + "loss": 0.9319, + "step": 42580 + }, + { + "epoch": 3.15, + "learning_rate": 9.69530924602789e-06, + "loss": 1.024, + "step": 42581 + }, + { + "epoch": 3.15, + "learning_rate": 9.694910559842484e-06, + "loss": 0.9421, + "step": 42582 + }, + { + "epoch": 3.15, + "learning_rate": 9.694511874142472e-06, + "loss": 0.9981, + "step": 42583 + }, + { + "epoch": 3.15, + "learning_rate": 9.694113188928491e-06, + "loss": 0.9508, + "step": 42584 + }, + { + "epoch": 3.15, + "learning_rate": 9.693714504201165e-06, + "loss": 0.9707, + "step": 42585 + }, + { + "epoch": 3.15, + "learning_rate": 9.69331581996114e-06, + "loss": 0.9237, + "step": 42586 + }, + { + "epoch": 3.15, + "learning_rate": 9.692917136209039e-06, + "loss": 1.029, + "step": 42587 + }, + { + "epoch": 3.15, + "learning_rate": 9.69251845294551e-06, + "loss": 1.0746, + "step": 42588 + }, + { + "epoch": 3.15, + "learning_rate": 9.692119770171172e-06, + "loss": 1.0618, + "step": 42589 + }, + { + "epoch": 3.15, + "learning_rate": 9.691721087886675e-06, + "loss": 0.9191, + "step": 42590 + }, + { + "epoch": 3.15, + "learning_rate": 9.691322406092637e-06, + "loss": 0.9469, + "step": 42591 + }, + { + "epoch": 3.15, + "learning_rate": 9.690923724789708e-06, + "loss": 0.9808, + "step": 42592 + }, + { + "epoch": 3.15, + "learning_rate": 9.690525043978513e-06, + "loss": 1.0511, + "step": 42593 + }, + { + "epoch": 3.15, + "learning_rate": 9.690126363659686e-06, + "loss": 1.0598, + "step": 42594 + }, + { + "epoch": 3.15, + "learning_rate": 9.689727683833864e-06, + "loss": 1.0963, + "step": 42595 + }, + { + "epoch": 3.15, + "learning_rate": 9.689329004501682e-06, + "loss": 0.9586, + "step": 42596 + }, + { + "epoch": 3.15, + "learning_rate": 9.688930325663773e-06, + "loss": 1.0525, + "step": 42597 + }, + { + "epoch": 3.15, + "learning_rate": 9.68853164732077e-06, + "loss": 0.9472, + "step": 42598 + }, + { + "epoch": 3.15, + "learning_rate": 9.68813296947331e-06, + "loss": 1.0701, + "step": 42599 + }, + { + "epoch": 3.15, + "learning_rate": 9.687734292122024e-06, + "loss": 1.1253, + "step": 42600 + }, + { + "epoch": 3.15, + "learning_rate": 9.687335615267552e-06, + "loss": 0.9971, + "step": 42601 + }, + { + "epoch": 3.15, + "learning_rate": 9.68693693891052e-06, + "loss": 0.9994, + "step": 42602 + }, + { + "epoch": 3.15, + "learning_rate": 9.68653826305157e-06, + "loss": 1.1416, + "step": 42603 + }, + { + "epoch": 3.15, + "learning_rate": 9.686139587691334e-06, + "loss": 0.869, + "step": 42604 + }, + { + "epoch": 3.15, + "learning_rate": 9.685740912830442e-06, + "loss": 1.0618, + "step": 42605 + }, + { + "epoch": 3.15, + "learning_rate": 9.685342238469534e-06, + "loss": 1.0017, + "step": 42606 + }, + { + "epoch": 3.15, + "learning_rate": 9.684943564609242e-06, + "loss": 0.9422, + "step": 42607 + }, + { + "epoch": 3.15, + "learning_rate": 9.6845448912502e-06, + "loss": 0.9349, + "step": 42608 + }, + { + "epoch": 3.15, + "learning_rate": 9.684146218393041e-06, + "loss": 0.9794, + "step": 42609 + }, + { + "epoch": 3.15, + "learning_rate": 9.683747546038406e-06, + "loss": 1.0703, + "step": 42610 + }, + { + "epoch": 3.15, + "learning_rate": 9.683348874186917e-06, + "loss": 0.9918, + "step": 42611 + }, + { + "epoch": 3.15, + "learning_rate": 9.682950202839222e-06, + "loss": 0.967, + "step": 42612 + }, + { + "epoch": 3.15, + "learning_rate": 9.682551531995946e-06, + "loss": 1.1046, + "step": 42613 + }, + { + "epoch": 3.15, + "learning_rate": 9.682152861657724e-06, + "loss": 0.9907, + "step": 42614 + }, + { + "epoch": 3.15, + "learning_rate": 9.681754191825195e-06, + "loss": 0.971, + "step": 42615 + }, + { + "epoch": 3.15, + "learning_rate": 9.681355522498988e-06, + "loss": 0.9844, + "step": 42616 + }, + { + "epoch": 3.15, + "learning_rate": 9.68095685367974e-06, + "loss": 1.0216, + "step": 42617 + }, + { + "epoch": 3.15, + "learning_rate": 9.680558185368087e-06, + "loss": 0.9874, + "step": 42618 + }, + { + "epoch": 3.15, + "learning_rate": 9.680159517564661e-06, + "loss": 0.9821, + "step": 42619 + }, + { + "epoch": 3.15, + "learning_rate": 9.679760850270094e-06, + "loss": 0.8838, + "step": 42620 + }, + { + "epoch": 3.15, + "learning_rate": 9.679362183485028e-06, + "loss": 1.0234, + "step": 42621 + }, + { + "epoch": 3.15, + "learning_rate": 9.67896351721009e-06, + "loss": 1.0215, + "step": 42622 + }, + { + "epoch": 3.15, + "learning_rate": 9.678564851445913e-06, + "loss": 0.9549, + "step": 42623 + }, + { + "epoch": 3.15, + "learning_rate": 9.678166186193138e-06, + "loss": 0.968, + "step": 42624 + }, + { + "epoch": 3.15, + "learning_rate": 9.677767521452395e-06, + "loss": 1.0212, + "step": 42625 + }, + { + "epoch": 3.15, + "learning_rate": 9.677368857224321e-06, + "loss": 0.9266, + "step": 42626 + }, + { + "epoch": 3.15, + "learning_rate": 9.676970193509546e-06, + "loss": 1.0045, + "step": 42627 + }, + { + "epoch": 3.15, + "learning_rate": 9.676571530308708e-06, + "loss": 0.9881, + "step": 42628 + }, + { + "epoch": 3.15, + "learning_rate": 9.676172867622438e-06, + "loss": 0.9851, + "step": 42629 + }, + { + "epoch": 3.15, + "learning_rate": 9.675774205451375e-06, + "loss": 0.9073, + "step": 42630 + }, + { + "epoch": 3.15, + "learning_rate": 9.675375543796156e-06, + "loss": 1.0739, + "step": 42631 + }, + { + "epoch": 3.15, + "learning_rate": 9.674976882657399e-06, + "loss": 0.9681, + "step": 42632 + }, + { + "epoch": 3.15, + "learning_rate": 9.674578222035757e-06, + "loss": 1.0201, + "step": 42633 + }, + { + "epoch": 3.15, + "learning_rate": 9.674179561931851e-06, + "loss": 1.0275, + "step": 42634 + }, + { + "epoch": 3.15, + "learning_rate": 9.673780902346326e-06, + "loss": 0.9316, + "step": 42635 + }, + { + "epoch": 3.15, + "learning_rate": 9.673382243279803e-06, + "loss": 1.0778, + "step": 42636 + }, + { + "epoch": 3.15, + "learning_rate": 9.672983584732933e-06, + "loss": 1.0722, + "step": 42637 + }, + { + "epoch": 3.15, + "learning_rate": 9.672584926706332e-06, + "loss": 1.0273, + "step": 42638 + }, + { + "epoch": 3.15, + "learning_rate": 9.672186269200654e-06, + "loss": 1.0241, + "step": 42639 + }, + { + "epoch": 3.15, + "learning_rate": 9.67178761221652e-06, + "loss": 1.0115, + "step": 42640 + }, + { + "epoch": 3.15, + "learning_rate": 9.671388955754563e-06, + "loss": 1.049, + "step": 42641 + }, + { + "epoch": 3.15, + "learning_rate": 9.670990299815425e-06, + "loss": 1.0953, + "step": 42642 + }, + { + "epoch": 3.15, + "learning_rate": 9.670591644399734e-06, + "loss": 1.076, + "step": 42643 + }, + { + "epoch": 3.15, + "learning_rate": 9.67019298950813e-06, + "loss": 0.8731, + "step": 42644 + }, + { + "epoch": 3.15, + "learning_rate": 9.669794335141242e-06, + "loss": 0.9222, + "step": 42645 + }, + { + "epoch": 3.15, + "learning_rate": 9.669395681299708e-06, + "loss": 1.0763, + "step": 42646 + }, + { + "epoch": 3.15, + "learning_rate": 9.668997027984158e-06, + "loss": 1.0948, + "step": 42647 + }, + { + "epoch": 3.15, + "learning_rate": 9.668598375195233e-06, + "loss": 1.0022, + "step": 42648 + }, + { + "epoch": 3.15, + "learning_rate": 9.668199722933563e-06, + "loss": 0.9613, + "step": 42649 + }, + { + "epoch": 3.15, + "learning_rate": 9.667801071199779e-06, + "loss": 1.0801, + "step": 42650 + }, + { + "epoch": 3.15, + "learning_rate": 9.667402419994521e-06, + "loss": 1.0014, + "step": 42651 + }, + { + "epoch": 3.15, + "learning_rate": 9.667003769318419e-06, + "loss": 0.9196, + "step": 42652 + }, + { + "epoch": 3.15, + "learning_rate": 9.666605119172113e-06, + "loss": 1.0537, + "step": 42653 + }, + { + "epoch": 3.15, + "learning_rate": 9.666206469556228e-06, + "loss": 1.0266, + "step": 42654 + }, + { + "epoch": 3.15, + "learning_rate": 9.66580782047141e-06, + "loss": 0.9667, + "step": 42655 + }, + { + "epoch": 3.15, + "learning_rate": 9.665409171918281e-06, + "loss": 1.0346, + "step": 42656 + }, + { + "epoch": 3.15, + "learning_rate": 9.665010523897488e-06, + "loss": 0.9752, + "step": 42657 + }, + { + "epoch": 3.15, + "learning_rate": 9.664611876409651e-06, + "loss": 1.0651, + "step": 42658 + }, + { + "epoch": 3.15, + "learning_rate": 9.664213229455418e-06, + "loss": 0.9952, + "step": 42659 + }, + { + "epoch": 3.15, + "learning_rate": 9.663814583035414e-06, + "loss": 0.9196, + "step": 42660 + }, + { + "epoch": 3.15, + "learning_rate": 9.663415937150277e-06, + "loss": 0.9282, + "step": 42661 + }, + { + "epoch": 3.15, + "learning_rate": 9.66301729180064e-06, + "loss": 0.9386, + "step": 42662 + }, + { + "epoch": 3.15, + "learning_rate": 9.662618646987137e-06, + "loss": 0.9668, + "step": 42663 + }, + { + "epoch": 3.15, + "learning_rate": 9.662220002710404e-06, + "loss": 0.9917, + "step": 42664 + }, + { + "epoch": 3.15, + "learning_rate": 9.661821358971073e-06, + "loss": 1.1015, + "step": 42665 + }, + { + "epoch": 3.15, + "learning_rate": 9.66142271576978e-06, + "loss": 0.9941, + "step": 42666 + }, + { + "epoch": 3.15, + "learning_rate": 9.661024073107156e-06, + "loss": 0.9718, + "step": 42667 + }, + { + "epoch": 3.15, + "learning_rate": 9.660625430983844e-06, + "loss": 0.8818, + "step": 42668 + }, + { + "epoch": 3.15, + "learning_rate": 9.660226789400469e-06, + "loss": 0.9235, + "step": 42669 + }, + { + "epoch": 3.15, + "learning_rate": 9.659828148357667e-06, + "loss": 1.1013, + "step": 42670 + }, + { + "epoch": 3.15, + "learning_rate": 9.659429507856075e-06, + "loss": 1.0507, + "step": 42671 + }, + { + "epoch": 3.15, + "learning_rate": 9.659030867896323e-06, + "loss": 0.9433, + "step": 42672 + }, + { + "epoch": 3.15, + "learning_rate": 9.658632228479052e-06, + "loss": 1.1222, + "step": 42673 + }, + { + "epoch": 3.15, + "learning_rate": 9.658233589604888e-06, + "loss": 1.0325, + "step": 42674 + }, + { + "epoch": 3.15, + "learning_rate": 9.657834951274472e-06, + "loss": 0.9976, + "step": 42675 + }, + { + "epoch": 3.15, + "learning_rate": 9.657436313488436e-06, + "loss": 1.0238, + "step": 42676 + }, + { + "epoch": 3.15, + "learning_rate": 9.657037676247414e-06, + "loss": 0.963, + "step": 42677 + }, + { + "epoch": 3.15, + "learning_rate": 9.656639039552043e-06, + "loss": 0.9253, + "step": 42678 + }, + { + "epoch": 3.15, + "learning_rate": 9.656240403402948e-06, + "loss": 0.9842, + "step": 42679 + }, + { + "epoch": 3.15, + "learning_rate": 9.655841767800776e-06, + "loss": 1.0107, + "step": 42680 + }, + { + "epoch": 3.15, + "learning_rate": 9.655443132746148e-06, + "loss": 1.0094, + "step": 42681 + }, + { + "epoch": 3.15, + "learning_rate": 9.655044498239711e-06, + "loss": 1.0529, + "step": 42682 + }, + { + "epoch": 3.15, + "learning_rate": 9.654645864282088e-06, + "loss": 1.0705, + "step": 42683 + }, + { + "epoch": 3.15, + "learning_rate": 9.654247230873924e-06, + "loss": 1.0848, + "step": 42684 + }, + { + "epoch": 3.15, + "learning_rate": 9.653848598015842e-06, + "loss": 1.0671, + "step": 42685 + }, + { + "epoch": 3.15, + "learning_rate": 9.653449965708488e-06, + "loss": 0.972, + "step": 42686 + }, + { + "epoch": 3.15, + "learning_rate": 9.653051333952487e-06, + "loss": 1.0042, + "step": 42687 + }, + { + "epoch": 3.15, + "learning_rate": 9.652652702748474e-06, + "loss": 0.9857, + "step": 42688 + }, + { + "epoch": 3.15, + "learning_rate": 9.65225407209709e-06, + "loss": 1.0397, + "step": 42689 + }, + { + "epoch": 3.15, + "learning_rate": 9.651855441998961e-06, + "loss": 0.9244, + "step": 42690 + }, + { + "epoch": 3.15, + "learning_rate": 9.651456812454728e-06, + "loss": 0.9781, + "step": 42691 + }, + { + "epoch": 3.15, + "learning_rate": 9.65105818346502e-06, + "loss": 1.0287, + "step": 42692 + }, + { + "epoch": 3.15, + "learning_rate": 9.650659555030477e-06, + "loss": 1.0135, + "step": 42693 + }, + { + "epoch": 3.15, + "learning_rate": 9.650260927151724e-06, + "loss": 1.1352, + "step": 42694 + }, + { + "epoch": 3.15, + "learning_rate": 9.649862299829406e-06, + "loss": 0.8963, + "step": 42695 + }, + { + "epoch": 3.15, + "learning_rate": 9.64946367306415e-06, + "loss": 0.9443, + "step": 42696 + }, + { + "epoch": 3.15, + "learning_rate": 9.649065046856592e-06, + "loss": 1.0933, + "step": 42697 + }, + { + "epoch": 3.15, + "learning_rate": 9.648666421207369e-06, + "loss": 0.9577, + "step": 42698 + }, + { + "epoch": 3.16, + "learning_rate": 9.648267796117107e-06, + "loss": 0.9689, + "step": 42699 + }, + { + "epoch": 3.16, + "learning_rate": 9.647869171586452e-06, + "loss": 1.0294, + "step": 42700 + }, + { + "epoch": 3.16, + "learning_rate": 9.647470547616026e-06, + "loss": 0.9441, + "step": 42701 + }, + { + "epoch": 3.16, + "learning_rate": 9.647071924206477e-06, + "loss": 1.0353, + "step": 42702 + }, + { + "epoch": 3.16, + "learning_rate": 9.646673301358424e-06, + "loss": 0.9962, + "step": 42703 + }, + { + "epoch": 3.16, + "learning_rate": 9.646274679072514e-06, + "loss": 1.0621, + "step": 42704 + }, + { + "epoch": 3.16, + "learning_rate": 9.645876057349372e-06, + "loss": 1.0318, + "step": 42705 + }, + { + "epoch": 3.16, + "learning_rate": 9.645477436189641e-06, + "loss": 1.0235, + "step": 42706 + }, + { + "epoch": 3.16, + "learning_rate": 9.645078815593947e-06, + "loss": 0.9578, + "step": 42707 + }, + { + "epoch": 3.16, + "learning_rate": 9.644680195562928e-06, + "loss": 0.9967, + "step": 42708 + }, + { + "epoch": 3.16, + "learning_rate": 9.64428157609722e-06, + "loss": 1.0934, + "step": 42709 + }, + { + "epoch": 3.16, + "learning_rate": 9.64388295719745e-06, + "loss": 0.9956, + "step": 42710 + }, + { + "epoch": 3.16, + "learning_rate": 9.643484338864261e-06, + "loss": 1.0172, + "step": 42711 + }, + { + "epoch": 3.16, + "learning_rate": 9.64308572109828e-06, + "loss": 0.9118, + "step": 42712 + }, + { + "epoch": 3.16, + "learning_rate": 9.642687103900147e-06, + "loss": 0.9948, + "step": 42713 + }, + { + "epoch": 3.16, + "learning_rate": 9.642288487270492e-06, + "loss": 1.0267, + "step": 42714 + }, + { + "epoch": 3.16, + "learning_rate": 9.641889871209954e-06, + "loss": 1.0064, + "step": 42715 + }, + { + "epoch": 3.16, + "learning_rate": 9.641491255719163e-06, + "loss": 0.9749, + "step": 42716 + }, + { + "epoch": 3.16, + "learning_rate": 9.641092640798752e-06, + "loss": 0.9314, + "step": 42717 + }, + { + "epoch": 3.16, + "learning_rate": 9.64069402644936e-06, + "loss": 0.9505, + "step": 42718 + }, + { + "epoch": 3.16, + "learning_rate": 9.640295412671617e-06, + "loss": 1.0058, + "step": 42719 + }, + { + "epoch": 3.16, + "learning_rate": 9.63989679946616e-06, + "loss": 1.106, + "step": 42720 + }, + { + "epoch": 3.16, + "learning_rate": 9.63949818683362e-06, + "loss": 0.9853, + "step": 42721 + }, + { + "epoch": 3.16, + "learning_rate": 9.639099574774636e-06, + "loss": 1.0813, + "step": 42722 + }, + { + "epoch": 3.16, + "learning_rate": 9.638700963289837e-06, + "loss": 0.9815, + "step": 42723 + }, + { + "epoch": 3.16, + "learning_rate": 9.638302352379863e-06, + "loss": 0.9476, + "step": 42724 + }, + { + "epoch": 3.16, + "learning_rate": 9.637903742045344e-06, + "loss": 0.9638, + "step": 42725 + }, + { + "epoch": 3.16, + "learning_rate": 9.637505132286911e-06, + "loss": 0.9359, + "step": 42726 + }, + { + "epoch": 3.16, + "learning_rate": 9.637106523105208e-06, + "loss": 0.9783, + "step": 42727 + }, + { + "epoch": 3.16, + "learning_rate": 9.636707914500857e-06, + "loss": 0.9443, + "step": 42728 + }, + { + "epoch": 3.16, + "learning_rate": 9.636309306474504e-06, + "loss": 0.9317, + "step": 42729 + }, + { + "epoch": 3.16, + "learning_rate": 9.635910699026773e-06, + "loss": 0.9572, + "step": 42730 + }, + { + "epoch": 3.16, + "learning_rate": 9.635512092158308e-06, + "loss": 0.9217, + "step": 42731 + }, + { + "epoch": 3.16, + "learning_rate": 9.635113485869734e-06, + "loss": 0.9573, + "step": 42732 + }, + { + "epoch": 3.16, + "learning_rate": 9.634714880161691e-06, + "loss": 0.927, + "step": 42733 + }, + { + "epoch": 3.16, + "learning_rate": 9.634316275034812e-06, + "loss": 1.0985, + "step": 42734 + }, + { + "epoch": 3.16, + "learning_rate": 9.633917670489727e-06, + "loss": 0.9674, + "step": 42735 + }, + { + "epoch": 3.16, + "learning_rate": 9.633519066527078e-06, + "loss": 1.0893, + "step": 42736 + }, + { + "epoch": 3.16, + "learning_rate": 9.633120463147493e-06, + "loss": 1.0201, + "step": 42737 + }, + { + "epoch": 3.16, + "learning_rate": 9.632721860351608e-06, + "loss": 1.063, + "step": 42738 + }, + { + "epoch": 3.16, + "learning_rate": 9.632323258140056e-06, + "loss": 1.0056, + "step": 42739 + }, + { + "epoch": 3.16, + "learning_rate": 9.631924656513474e-06, + "loss": 1.0274, + "step": 42740 + }, + { + "epoch": 3.16, + "learning_rate": 9.631526055472493e-06, + "loss": 0.9224, + "step": 42741 + }, + { + "epoch": 3.16, + "learning_rate": 9.631127455017751e-06, + "loss": 0.98, + "step": 42742 + }, + { + "epoch": 3.16, + "learning_rate": 9.630728855149882e-06, + "loss": 1.0928, + "step": 42743 + }, + { + "epoch": 3.16, + "learning_rate": 9.630330255869512e-06, + "loss": 0.9995, + "step": 42744 + }, + { + "epoch": 3.16, + "learning_rate": 9.629931657177289e-06, + "loss": 1.0031, + "step": 42745 + }, + { + "epoch": 3.16, + "learning_rate": 9.629533059073831e-06, + "loss": 1.0493, + "step": 42746 + }, + { + "epoch": 3.16, + "learning_rate": 9.629134461559787e-06, + "loss": 0.8501, + "step": 42747 + }, + { + "epoch": 3.16, + "learning_rate": 9.62873586463578e-06, + "loss": 1.0553, + "step": 42748 + }, + { + "epoch": 3.16, + "learning_rate": 9.628337268302453e-06, + "loss": 0.9397, + "step": 42749 + }, + { + "epoch": 3.16, + "learning_rate": 9.627938672560432e-06, + "loss": 1.0127, + "step": 42750 + }, + { + "epoch": 3.16, + "learning_rate": 9.62754007741036e-06, + "loss": 0.959, + "step": 42751 + }, + { + "epoch": 3.16, + "learning_rate": 9.627141482852865e-06, + "loss": 1.0529, + "step": 42752 + }, + { + "epoch": 3.16, + "learning_rate": 9.626742888888581e-06, + "loss": 1.0368, + "step": 42753 + }, + { + "epoch": 3.16, + "learning_rate": 9.626344295518145e-06, + "loss": 1.1173, + "step": 42754 + }, + { + "epoch": 3.16, + "learning_rate": 9.625945702742188e-06, + "loss": 1.0589, + "step": 42755 + }, + { + "epoch": 3.16, + "learning_rate": 9.625547110561349e-06, + "loss": 0.9819, + "step": 42756 + }, + { + "epoch": 3.16, + "learning_rate": 9.625148518976255e-06, + "loss": 0.9118, + "step": 42757 + }, + { + "epoch": 3.16, + "learning_rate": 9.624749927987547e-06, + "loss": 0.9923, + "step": 42758 + }, + { + "epoch": 3.16, + "learning_rate": 9.624351337595855e-06, + "loss": 1.0219, + "step": 42759 + }, + { + "epoch": 3.16, + "learning_rate": 9.623952747801818e-06, + "loss": 0.9608, + "step": 42760 + }, + { + "epoch": 3.16, + "learning_rate": 9.623554158606064e-06, + "loss": 1.0113, + "step": 42761 + }, + { + "epoch": 3.16, + "learning_rate": 9.623155570009231e-06, + "loss": 0.9752, + "step": 42762 + }, + { + "epoch": 3.16, + "learning_rate": 9.622756982011952e-06, + "loss": 1.0017, + "step": 42763 + }, + { + "epoch": 3.16, + "learning_rate": 9.62235839461486e-06, + "loss": 0.9135, + "step": 42764 + }, + { + "epoch": 3.16, + "learning_rate": 9.621959807818593e-06, + "loss": 1.0907, + "step": 42765 + }, + { + "epoch": 3.16, + "learning_rate": 9.621561221623778e-06, + "loss": 0.9243, + "step": 42766 + }, + { + "epoch": 3.16, + "learning_rate": 9.621162636031058e-06, + "loss": 0.9396, + "step": 42767 + }, + { + "epoch": 3.16, + "learning_rate": 9.620764051041061e-06, + "loss": 0.9885, + "step": 42768 + }, + { + "epoch": 3.16, + "learning_rate": 9.620365466654425e-06, + "loss": 0.9356, + "step": 42769 + }, + { + "epoch": 3.16, + "learning_rate": 9.619966882871778e-06, + "loss": 1.1486, + "step": 42770 + }, + { + "epoch": 3.16, + "learning_rate": 9.619568299693763e-06, + "loss": 1.0448, + "step": 42771 + }, + { + "epoch": 3.16, + "learning_rate": 9.619169717121008e-06, + "loss": 1.1547, + "step": 42772 + }, + { + "epoch": 3.16, + "learning_rate": 9.618771135154145e-06, + "loss": 1.0488, + "step": 42773 + }, + { + "epoch": 3.16, + "learning_rate": 9.618372553793818e-06, + "loss": 0.9326, + "step": 42774 + }, + { + "epoch": 3.16, + "learning_rate": 9.61797397304065e-06, + "loss": 0.9255, + "step": 42775 + }, + { + "epoch": 3.16, + "learning_rate": 9.617575392895281e-06, + "loss": 1.0014, + "step": 42776 + }, + { + "epoch": 3.16, + "learning_rate": 9.617176813358344e-06, + "loss": 0.923, + "step": 42777 + }, + { + "epoch": 3.16, + "learning_rate": 9.616778234430474e-06, + "loss": 1.1273, + "step": 42778 + }, + { + "epoch": 3.16, + "learning_rate": 9.616379656112303e-06, + "loss": 1.0001, + "step": 42779 + }, + { + "epoch": 3.16, + "learning_rate": 9.615981078404467e-06, + "loss": 0.9621, + "step": 42780 + }, + { + "epoch": 3.16, + "learning_rate": 9.6155825013076e-06, + "loss": 0.9676, + "step": 42781 + }, + { + "epoch": 3.16, + "learning_rate": 9.615183924822335e-06, + "loss": 0.9713, + "step": 42782 + }, + { + "epoch": 3.16, + "learning_rate": 9.61478534894931e-06, + "loss": 1.0594, + "step": 42783 + }, + { + "epoch": 3.16, + "learning_rate": 9.614386773689151e-06, + "loss": 0.9506, + "step": 42784 + }, + { + "epoch": 3.16, + "learning_rate": 9.6139881990425e-06, + "loss": 1.0042, + "step": 42785 + }, + { + "epoch": 3.16, + "learning_rate": 9.613589625009986e-06, + "loss": 0.9812, + "step": 42786 + }, + { + "epoch": 3.16, + "learning_rate": 9.613191051592248e-06, + "loss": 0.93, + "step": 42787 + }, + { + "epoch": 3.16, + "learning_rate": 9.612792478789916e-06, + "loss": 1.0458, + "step": 42788 + }, + { + "epoch": 3.16, + "learning_rate": 9.612393906603628e-06, + "loss": 1.0374, + "step": 42789 + }, + { + "epoch": 3.16, + "learning_rate": 9.611995335034016e-06, + "loss": 1.0379, + "step": 42790 + }, + { + "epoch": 3.16, + "learning_rate": 9.61159676408171e-06, + "loss": 1.005, + "step": 42791 + }, + { + "epoch": 3.16, + "learning_rate": 9.611198193747352e-06, + "loss": 0.8993, + "step": 42792 + }, + { + "epoch": 3.16, + "learning_rate": 9.610799624031566e-06, + "loss": 0.9328, + "step": 42793 + }, + { + "epoch": 3.16, + "learning_rate": 9.610401054935001e-06, + "loss": 0.9632, + "step": 42794 + }, + { + "epoch": 3.16, + "learning_rate": 9.610002486458274e-06, + "loss": 1.0673, + "step": 42795 + }, + { + "epoch": 3.16, + "learning_rate": 9.609603918602034e-06, + "loss": 0.8787, + "step": 42796 + }, + { + "epoch": 3.16, + "learning_rate": 9.609205351366902e-06, + "loss": 1.1137, + "step": 42797 + }, + { + "epoch": 3.16, + "learning_rate": 9.608806784753525e-06, + "loss": 1.0331, + "step": 42798 + }, + { + "epoch": 3.16, + "learning_rate": 9.60840821876253e-06, + "loss": 0.8697, + "step": 42799 + }, + { + "epoch": 3.16, + "learning_rate": 9.608009653394547e-06, + "loss": 1.0231, + "step": 42800 + }, + { + "epoch": 3.16, + "learning_rate": 9.60761108865022e-06, + "loss": 1.0004, + "step": 42801 + }, + { + "epoch": 3.16, + "learning_rate": 9.607212524530175e-06, + "loss": 1.0544, + "step": 42802 + }, + { + "epoch": 3.16, + "learning_rate": 9.606813961035052e-06, + "loss": 0.9968, + "step": 42803 + }, + { + "epoch": 3.16, + "learning_rate": 9.60641539816548e-06, + "loss": 1.1049, + "step": 42804 + }, + { + "epoch": 3.16, + "learning_rate": 9.606016835922096e-06, + "loss": 0.9772, + "step": 42805 + }, + { + "epoch": 3.16, + "learning_rate": 9.605618274305534e-06, + "loss": 0.9865, + "step": 42806 + }, + { + "epoch": 3.16, + "learning_rate": 9.605219713316428e-06, + "loss": 1.0674, + "step": 42807 + }, + { + "epoch": 3.16, + "learning_rate": 9.604821152955413e-06, + "loss": 0.982, + "step": 42808 + }, + { + "epoch": 3.16, + "learning_rate": 9.604422593223119e-06, + "loss": 0.9334, + "step": 42809 + }, + { + "epoch": 3.16, + "learning_rate": 9.604024034120183e-06, + "loss": 1.0226, + "step": 42810 + }, + { + "epoch": 3.16, + "learning_rate": 9.60362547564724e-06, + "loss": 0.9464, + "step": 42811 + }, + { + "epoch": 3.16, + "learning_rate": 9.603226917804926e-06, + "loss": 1.0096, + "step": 42812 + }, + { + "epoch": 3.16, + "learning_rate": 9.602828360593868e-06, + "loss": 1.1514, + "step": 42813 + }, + { + "epoch": 3.16, + "learning_rate": 9.602429804014706e-06, + "loss": 1.0217, + "step": 42814 + }, + { + "epoch": 3.16, + "learning_rate": 9.602031248068071e-06, + "loss": 0.963, + "step": 42815 + }, + { + "epoch": 3.16, + "learning_rate": 9.601632692754602e-06, + "loss": 1.0108, + "step": 42816 + }, + { + "epoch": 3.16, + "learning_rate": 9.601234138074927e-06, + "loss": 0.9496, + "step": 42817 + }, + { + "epoch": 3.16, + "learning_rate": 9.600835584029685e-06, + "loss": 0.9033, + "step": 42818 + }, + { + "epoch": 3.16, + "learning_rate": 9.60043703061951e-06, + "loss": 1.0562, + "step": 42819 + }, + { + "epoch": 3.16, + "learning_rate": 9.60003847784503e-06, + "loss": 0.9966, + "step": 42820 + }, + { + "epoch": 3.16, + "learning_rate": 9.599639925706884e-06, + "loss": 1.012, + "step": 42821 + }, + { + "epoch": 3.16, + "learning_rate": 9.599241374205704e-06, + "loss": 1.0238, + "step": 42822 + }, + { + "epoch": 3.16, + "learning_rate": 9.598842823342125e-06, + "loss": 0.9944, + "step": 42823 + }, + { + "epoch": 3.16, + "learning_rate": 9.598444273116783e-06, + "loss": 0.9918, + "step": 42824 + }, + { + "epoch": 3.16, + "learning_rate": 9.59804572353031e-06, + "loss": 1.0654, + "step": 42825 + }, + { + "epoch": 3.16, + "learning_rate": 9.59764717458334e-06, + "loss": 0.9499, + "step": 42826 + }, + { + "epoch": 3.16, + "learning_rate": 9.597248626276508e-06, + "loss": 0.9536, + "step": 42827 + }, + { + "epoch": 3.16, + "learning_rate": 9.596850078610449e-06, + "loss": 1.0031, + "step": 42828 + }, + { + "epoch": 3.16, + "learning_rate": 9.596451531585794e-06, + "loss": 0.9235, + "step": 42829 + }, + { + "epoch": 3.16, + "learning_rate": 9.596052985203179e-06, + "loss": 0.9397, + "step": 42830 + }, + { + "epoch": 3.16, + "learning_rate": 9.595654439463237e-06, + "loss": 0.992, + "step": 42831 + }, + { + "epoch": 3.16, + "learning_rate": 9.595255894366605e-06, + "loss": 1.1089, + "step": 42832 + }, + { + "epoch": 3.16, + "learning_rate": 9.594857349913913e-06, + "loss": 0.9357, + "step": 42833 + }, + { + "epoch": 3.16, + "learning_rate": 9.5944588061058e-06, + "loss": 1.0734, + "step": 42834 + }, + { + "epoch": 3.17, + "learning_rate": 9.594060262942894e-06, + "loss": 1.0323, + "step": 42835 + }, + { + "epoch": 3.17, + "learning_rate": 9.593661720425834e-06, + "loss": 0.9975, + "step": 42836 + }, + { + "epoch": 3.17, + "learning_rate": 9.593263178555256e-06, + "loss": 1.031, + "step": 42837 + }, + { + "epoch": 3.17, + "learning_rate": 9.592864637331783e-06, + "loss": 0.9588, + "step": 42838 + }, + { + "epoch": 3.17, + "learning_rate": 9.592466096756064e-06, + "loss": 0.9312, + "step": 42839 + }, + { + "epoch": 3.17, + "learning_rate": 9.592067556828718e-06, + "loss": 1.0647, + "step": 42840 + }, + { + "epoch": 3.17, + "learning_rate": 9.591669017550394e-06, + "loss": 0.9507, + "step": 42841 + }, + { + "epoch": 3.17, + "learning_rate": 9.591270478921712e-06, + "loss": 0.9751, + "step": 42842 + }, + { + "epoch": 3.17, + "learning_rate": 9.59087194094332e-06, + "loss": 1.0721, + "step": 42843 + }, + { + "epoch": 3.17, + "learning_rate": 9.590473403615836e-06, + "loss": 0.9381, + "step": 42844 + }, + { + "epoch": 3.17, + "learning_rate": 9.590074866939911e-06, + "loss": 1.0036, + "step": 42845 + }, + { + "epoch": 3.17, + "learning_rate": 9.58967633091617e-06, + "loss": 1.0102, + "step": 42846 + }, + { + "epoch": 3.17, + "learning_rate": 9.589277795545245e-06, + "loss": 1.0416, + "step": 42847 + }, + { + "epoch": 3.17, + "learning_rate": 9.588879260827775e-06, + "loss": 1.011, + "step": 42848 + }, + { + "epoch": 3.17, + "learning_rate": 9.588480726764389e-06, + "loss": 1.0277, + "step": 42849 + }, + { + "epoch": 3.17, + "learning_rate": 9.588082193355729e-06, + "loss": 1.0072, + "step": 42850 + }, + { + "epoch": 3.17, + "learning_rate": 9.58768366060242e-06, + "loss": 1.0023, + "step": 42851 + }, + { + "epoch": 3.17, + "learning_rate": 9.587285128505104e-06, + "loss": 0.982, + "step": 42852 + }, + { + "epoch": 3.17, + "learning_rate": 9.586886597064408e-06, + "loss": 1.0557, + "step": 42853 + }, + { + "epoch": 3.17, + "learning_rate": 9.586488066280973e-06, + "loss": 1.0319, + "step": 42854 + }, + { + "epoch": 3.17, + "learning_rate": 9.586089536155428e-06, + "loss": 1.0793, + "step": 42855 + }, + { + "epoch": 3.17, + "learning_rate": 9.585691006688408e-06, + "loss": 1.0131, + "step": 42856 + }, + { + "epoch": 3.17, + "learning_rate": 9.58529247788055e-06, + "loss": 1.0075, + "step": 42857 + }, + { + "epoch": 3.17, + "learning_rate": 9.584893949732483e-06, + "loss": 0.9502, + "step": 42858 + }, + { + "epoch": 3.17, + "learning_rate": 9.584495422244844e-06, + "loss": 1.0694, + "step": 42859 + }, + { + "epoch": 3.17, + "learning_rate": 9.584096895418267e-06, + "loss": 1.0018, + "step": 42860 + }, + { + "epoch": 3.17, + "learning_rate": 9.583698369253389e-06, + "loss": 0.9569, + "step": 42861 + }, + { + "epoch": 3.17, + "learning_rate": 9.583299843750836e-06, + "loss": 1.0606, + "step": 42862 + }, + { + "epoch": 3.17, + "learning_rate": 9.582901318911253e-06, + "loss": 1.0258, + "step": 42863 + }, + { + "epoch": 3.17, + "learning_rate": 9.582502794735265e-06, + "loss": 0.9952, + "step": 42864 + }, + { + "epoch": 3.17, + "learning_rate": 9.582104271223507e-06, + "loss": 0.9536, + "step": 42865 + }, + { + "epoch": 3.17, + "learning_rate": 9.581705748376618e-06, + "loss": 1.1403, + "step": 42866 + }, + { + "epoch": 3.17, + "learning_rate": 9.581307226195227e-06, + "loss": 0.9132, + "step": 42867 + }, + { + "epoch": 3.17, + "learning_rate": 9.580908704679972e-06, + "loss": 0.9808, + "step": 42868 + }, + { + "epoch": 3.17, + "learning_rate": 9.580510183831482e-06, + "loss": 0.896, + "step": 42869 + }, + { + "epoch": 3.17, + "learning_rate": 9.580111663650399e-06, + "loss": 0.9635, + "step": 42870 + }, + { + "epoch": 3.17, + "learning_rate": 9.57971314413735e-06, + "loss": 1.0433, + "step": 42871 + }, + { + "epoch": 3.17, + "learning_rate": 9.57931462529297e-06, + "loss": 1.0508, + "step": 42872 + }, + { + "epoch": 3.17, + "learning_rate": 9.578916107117897e-06, + "loss": 1.0501, + "step": 42873 + }, + { + "epoch": 3.17, + "learning_rate": 9.578517589612761e-06, + "loss": 0.943, + "step": 42874 + }, + { + "epoch": 3.17, + "learning_rate": 9.578119072778198e-06, + "loss": 0.9582, + "step": 42875 + }, + { + "epoch": 3.17, + "learning_rate": 9.57772055661484e-06, + "loss": 0.9437, + "step": 42876 + }, + { + "epoch": 3.17, + "learning_rate": 9.577322041123325e-06, + "loss": 1.0366, + "step": 42877 + }, + { + "epoch": 3.17, + "learning_rate": 9.576923526304282e-06, + "loss": 1.0076, + "step": 42878 + }, + { + "epoch": 3.17, + "learning_rate": 9.57652501215835e-06, + "loss": 1.073, + "step": 42879 + }, + { + "epoch": 3.17, + "learning_rate": 9.57612649868616e-06, + "loss": 0.9068, + "step": 42880 + }, + { + "epoch": 3.17, + "learning_rate": 9.575727985888345e-06, + "loss": 1.0387, + "step": 42881 + }, + { + "epoch": 3.17, + "learning_rate": 9.575329473765542e-06, + "loss": 1.0524, + "step": 42882 + }, + { + "epoch": 3.17, + "learning_rate": 9.574930962318384e-06, + "loss": 1.0403, + "step": 42883 + }, + { + "epoch": 3.17, + "learning_rate": 9.574532451547508e-06, + "loss": 0.9871, + "step": 42884 + }, + { + "epoch": 3.17, + "learning_rate": 9.574133941453538e-06, + "loss": 1.0601, + "step": 42885 + }, + { + "epoch": 3.17, + "learning_rate": 9.57373543203712e-06, + "loss": 1.0721, + "step": 42886 + }, + { + "epoch": 3.17, + "learning_rate": 9.573336923298878e-06, + "loss": 0.8777, + "step": 42887 + }, + { + "epoch": 3.17, + "learning_rate": 9.572938415239457e-06, + "loss": 0.9755, + "step": 42888 + }, + { + "epoch": 3.17, + "learning_rate": 9.572539907859478e-06, + "loss": 0.8952, + "step": 42889 + }, + { + "epoch": 3.17, + "learning_rate": 9.57214140115959e-06, + "loss": 1.0771, + "step": 42890 + }, + { + "epoch": 3.17, + "learning_rate": 9.57174289514041e-06, + "loss": 0.9607, + "step": 42891 + }, + { + "epoch": 3.17, + "learning_rate": 9.57134438980259e-06, + "loss": 0.9754, + "step": 42892 + }, + { + "epoch": 3.17, + "learning_rate": 9.570945885146751e-06, + "loss": 1.0508, + "step": 42893 + }, + { + "epoch": 3.17, + "learning_rate": 9.570547381173528e-06, + "loss": 0.9746, + "step": 42894 + }, + { + "epoch": 3.17, + "learning_rate": 9.570148877883562e-06, + "loss": 0.9057, + "step": 42895 + }, + { + "epoch": 3.17, + "learning_rate": 9.569750375277479e-06, + "loss": 0.9618, + "step": 42896 + }, + { + "epoch": 3.17, + "learning_rate": 9.56935187335592e-06, + "loss": 1.1036, + "step": 42897 + }, + { + "epoch": 3.17, + "learning_rate": 9.568953372119515e-06, + "loss": 1.0215, + "step": 42898 + }, + { + "epoch": 3.17, + "learning_rate": 9.5685548715689e-06, + "loss": 1.1147, + "step": 42899 + }, + { + "epoch": 3.17, + "learning_rate": 9.568156371704706e-06, + "loss": 0.9286, + "step": 42900 + }, + { + "epoch": 3.17, + "learning_rate": 9.56775787252757e-06, + "loss": 0.9954, + "step": 42901 + }, + { + "epoch": 3.17, + "learning_rate": 9.567359374038127e-06, + "loss": 1.0716, + "step": 42902 + }, + { + "epoch": 3.17, + "learning_rate": 9.566960876237007e-06, + "loss": 1.0944, + "step": 42903 + }, + { + "epoch": 3.17, + "learning_rate": 9.566562379124848e-06, + "loss": 1.0103, + "step": 42904 + }, + { + "epoch": 3.17, + "learning_rate": 9.56616388270228e-06, + "loss": 0.9397, + "step": 42905 + }, + { + "epoch": 3.17, + "learning_rate": 9.565765386969938e-06, + "loss": 0.9347, + "step": 42906 + }, + { + "epoch": 3.17, + "learning_rate": 9.565366891928458e-06, + "loss": 0.9887, + "step": 42907 + }, + { + "epoch": 3.17, + "learning_rate": 9.564968397578476e-06, + "loss": 0.9594, + "step": 42908 + }, + { + "epoch": 3.17, + "learning_rate": 9.564569903920619e-06, + "loss": 0.935, + "step": 42909 + }, + { + "epoch": 3.17, + "learning_rate": 9.564171410955529e-06, + "loss": 0.971, + "step": 42910 + }, + { + "epoch": 3.17, + "learning_rate": 9.563772918683835e-06, + "loss": 0.942, + "step": 42911 + }, + { + "epoch": 3.17, + "learning_rate": 9.563374427106168e-06, + "loss": 0.8388, + "step": 42912 + }, + { + "epoch": 3.17, + "learning_rate": 9.562975936223172e-06, + "loss": 0.9728, + "step": 42913 + }, + { + "epoch": 3.17, + "learning_rate": 9.56257744603547e-06, + "loss": 0.9733, + "step": 42914 + }, + { + "epoch": 3.17, + "learning_rate": 9.562178956543704e-06, + "loss": 1.0353, + "step": 42915 + }, + { + "epoch": 3.17, + "learning_rate": 9.561780467748503e-06, + "loss": 0.9291, + "step": 42916 + }, + { + "epoch": 3.17, + "learning_rate": 9.561381979650505e-06, + "loss": 0.998, + "step": 42917 + }, + { + "epoch": 3.17, + "learning_rate": 9.56098349225034e-06, + "loss": 0.9299, + "step": 42918 + }, + { + "epoch": 3.17, + "learning_rate": 9.560585005548645e-06, + "loss": 0.991, + "step": 42919 + }, + { + "epoch": 3.17, + "learning_rate": 9.560186519546051e-06, + "loss": 0.9728, + "step": 42920 + }, + { + "epoch": 3.17, + "learning_rate": 9.559788034243197e-06, + "loss": 0.8614, + "step": 42921 + }, + { + "epoch": 3.17, + "learning_rate": 9.559389549640715e-06, + "loss": 1.0667, + "step": 42922 + }, + { + "epoch": 3.17, + "learning_rate": 9.558991065739233e-06, + "loss": 0.9574, + "step": 42923 + }, + { + "epoch": 3.17, + "learning_rate": 9.558592582539395e-06, + "loss": 0.9755, + "step": 42924 + }, + { + "epoch": 3.17, + "learning_rate": 9.558194100041826e-06, + "loss": 0.9695, + "step": 42925 + }, + { + "epoch": 3.17, + "learning_rate": 9.557795618247166e-06, + "loss": 0.9687, + "step": 42926 + }, + { + "epoch": 3.17, + "learning_rate": 9.557397137156047e-06, + "loss": 0.7902, + "step": 42927 + }, + { + "epoch": 3.17, + "learning_rate": 9.556998656769101e-06, + "loss": 1.0677, + "step": 42928 + }, + { + "epoch": 3.17, + "learning_rate": 9.556600177086965e-06, + "loss": 1.0169, + "step": 42929 + }, + { + "epoch": 3.17, + "learning_rate": 9.556201698110272e-06, + "loss": 1.0482, + "step": 42930 + }, + { + "epoch": 3.17, + "learning_rate": 9.555803219839659e-06, + "loss": 1.0145, + "step": 42931 + }, + { + "epoch": 3.17, + "learning_rate": 9.555404742275749e-06, + "loss": 0.9277, + "step": 42932 + }, + { + "epoch": 3.17, + "learning_rate": 9.55500626541919e-06, + "loss": 1.0851, + "step": 42933 + }, + { + "epoch": 3.17, + "learning_rate": 9.554607789270605e-06, + "loss": 1.0833, + "step": 42934 + }, + { + "epoch": 3.17, + "learning_rate": 9.554209313830639e-06, + "loss": 1.024, + "step": 42935 + }, + { + "epoch": 3.17, + "learning_rate": 9.553810839099912e-06, + "loss": 1.0575, + "step": 42936 + }, + { + "epoch": 3.17, + "learning_rate": 9.553412365079073e-06, + "loss": 0.9589, + "step": 42937 + }, + { + "epoch": 3.17, + "learning_rate": 9.55301389176874e-06, + "loss": 1.0043, + "step": 42938 + }, + { + "epoch": 3.17, + "learning_rate": 9.552615419169565e-06, + "loss": 1.0343, + "step": 42939 + }, + { + "epoch": 3.17, + "learning_rate": 9.552216947282168e-06, + "loss": 1.0057, + "step": 42940 + }, + { + "epoch": 3.17, + "learning_rate": 9.551818476107186e-06, + "loss": 0.9857, + "step": 42941 + }, + { + "epoch": 3.17, + "learning_rate": 9.551420005645257e-06, + "loss": 0.9119, + "step": 42942 + }, + { + "epoch": 3.17, + "learning_rate": 9.55102153589701e-06, + "loss": 0.9775, + "step": 42943 + }, + { + "epoch": 3.17, + "learning_rate": 9.550623066863081e-06, + "loss": 1.1484, + "step": 42944 + }, + { + "epoch": 3.17, + "learning_rate": 9.550224598544105e-06, + "loss": 1.1041, + "step": 42945 + }, + { + "epoch": 3.17, + "learning_rate": 9.549826130940717e-06, + "loss": 0.9906, + "step": 42946 + }, + { + "epoch": 3.17, + "learning_rate": 9.549427664053547e-06, + "loss": 0.9514, + "step": 42947 + }, + { + "epoch": 3.17, + "learning_rate": 9.549029197883233e-06, + "loss": 1.1191, + "step": 42948 + }, + { + "epoch": 3.17, + "learning_rate": 9.548630732430407e-06, + "loss": 1.1082, + "step": 42949 + }, + { + "epoch": 3.17, + "learning_rate": 9.5482322676957e-06, + "loss": 1.0274, + "step": 42950 + }, + { + "epoch": 3.17, + "learning_rate": 9.547833803679753e-06, + "loss": 1.0277, + "step": 42951 + }, + { + "epoch": 3.17, + "learning_rate": 9.547435340383194e-06, + "loss": 1.0074, + "step": 42952 + }, + { + "epoch": 3.17, + "learning_rate": 9.547036877806661e-06, + "loss": 0.9803, + "step": 42953 + }, + { + "epoch": 3.17, + "learning_rate": 9.54663841595078e-06, + "loss": 0.9673, + "step": 42954 + }, + { + "epoch": 3.17, + "learning_rate": 9.5462399548162e-06, + "loss": 0.935, + "step": 42955 + }, + { + "epoch": 3.17, + "learning_rate": 9.545841494403536e-06, + "loss": 1.046, + "step": 42956 + }, + { + "epoch": 3.17, + "learning_rate": 9.545443034713439e-06, + "loss": 1.0338, + "step": 42957 + }, + { + "epoch": 3.17, + "learning_rate": 9.545044575746534e-06, + "loss": 1.0454, + "step": 42958 + }, + { + "epoch": 3.17, + "learning_rate": 9.544646117503454e-06, + "loss": 1.0916, + "step": 42959 + }, + { + "epoch": 3.17, + "learning_rate": 9.544247659984838e-06, + "loss": 0.8752, + "step": 42960 + }, + { + "epoch": 3.17, + "learning_rate": 9.543849203191316e-06, + "loss": 1.0095, + "step": 42961 + }, + { + "epoch": 3.17, + "learning_rate": 9.543450747123523e-06, + "loss": 0.8708, + "step": 42962 + }, + { + "epoch": 3.17, + "learning_rate": 9.543052291782093e-06, + "loss": 0.8673, + "step": 42963 + }, + { + "epoch": 3.17, + "learning_rate": 9.542653837167662e-06, + "loss": 0.9662, + "step": 42964 + }, + { + "epoch": 3.17, + "learning_rate": 9.54225538328086e-06, + "loss": 0.9171, + "step": 42965 + }, + { + "epoch": 3.17, + "learning_rate": 9.541856930122326e-06, + "loss": 1.0356, + "step": 42966 + }, + { + "epoch": 3.17, + "learning_rate": 9.541458477692691e-06, + "loss": 1.0019, + "step": 42967 + }, + { + "epoch": 3.17, + "learning_rate": 9.541060025992586e-06, + "loss": 1.0364, + "step": 42968 + }, + { + "epoch": 3.17, + "learning_rate": 9.540661575022651e-06, + "loss": 1.0761, + "step": 42969 + }, + { + "epoch": 3.18, + "learning_rate": 9.540263124783514e-06, + "loss": 1.1117, + "step": 42970 + }, + { + "epoch": 3.18, + "learning_rate": 9.539864675275816e-06, + "loss": 0.8922, + "step": 42971 + }, + { + "epoch": 3.18, + "learning_rate": 9.53946622650018e-06, + "loss": 1.0192, + "step": 42972 + }, + { + "epoch": 3.18, + "learning_rate": 9.539067778457253e-06, + "loss": 1.0147, + "step": 42973 + }, + { + "epoch": 3.18, + "learning_rate": 9.538669331147659e-06, + "loss": 1.0718, + "step": 42974 + }, + { + "epoch": 3.18, + "learning_rate": 9.538270884572037e-06, + "loss": 0.8632, + "step": 42975 + }, + { + "epoch": 3.18, + "learning_rate": 9.537872438731019e-06, + "loss": 0.893, + "step": 42976 + }, + { + "epoch": 3.18, + "learning_rate": 9.53747399362524e-06, + "loss": 0.8935, + "step": 42977 + }, + { + "epoch": 3.18, + "learning_rate": 9.537075549255335e-06, + "loss": 1.0404, + "step": 42978 + }, + { + "epoch": 3.18, + "learning_rate": 9.536677105621932e-06, + "loss": 0.9646, + "step": 42979 + }, + { + "epoch": 3.18, + "learning_rate": 9.536278662725673e-06, + "loss": 1.012, + "step": 42980 + }, + { + "epoch": 3.18, + "learning_rate": 9.535880220567183e-06, + "loss": 0.8168, + "step": 42981 + }, + { + "epoch": 3.18, + "learning_rate": 9.535481779147107e-06, + "loss": 1.0614, + "step": 42982 + }, + { + "epoch": 3.18, + "learning_rate": 9.535083338466066e-06, + "loss": 0.9875, + "step": 42983 + }, + { + "epoch": 3.18, + "learning_rate": 9.534684898524709e-06, + "loss": 1.0312, + "step": 42984 + }, + { + "epoch": 3.18, + "learning_rate": 9.534286459323654e-06, + "loss": 0.8931, + "step": 42985 + }, + { + "epoch": 3.18, + "learning_rate": 9.533888020863549e-06, + "loss": 1.0472, + "step": 42986 + }, + { + "epoch": 3.18, + "learning_rate": 9.533489583145017e-06, + "loss": 1.0991, + "step": 42987 + }, + { + "epoch": 3.18, + "learning_rate": 9.533091146168699e-06, + "loss": 1.1129, + "step": 42988 + }, + { + "epoch": 3.18, + "learning_rate": 9.532692709935226e-06, + "loss": 1.0689, + "step": 42989 + }, + { + "epoch": 3.18, + "learning_rate": 9.532294274445228e-06, + "loss": 1.0177, + "step": 42990 + }, + { + "epoch": 3.18, + "learning_rate": 9.531895839699349e-06, + "loss": 1.0431, + "step": 42991 + }, + { + "epoch": 3.18, + "learning_rate": 9.531497405698213e-06, + "loss": 0.9669, + "step": 42992 + }, + { + "epoch": 3.18, + "learning_rate": 9.531098972442459e-06, + "loss": 1.0183, + "step": 42993 + }, + { + "epoch": 3.18, + "learning_rate": 9.53070053993272e-06, + "loss": 0.9596, + "step": 42994 + }, + { + "epoch": 3.18, + "learning_rate": 9.53030210816963e-06, + "loss": 1.04, + "step": 42995 + }, + { + "epoch": 3.18, + "learning_rate": 9.529903677153825e-06, + "loss": 1.1047, + "step": 42996 + }, + { + "epoch": 3.18, + "learning_rate": 9.529505246885931e-06, + "loss": 1.0046, + "step": 42997 + }, + { + "epoch": 3.18, + "learning_rate": 9.529106817366594e-06, + "loss": 0.9939, + "step": 42998 + }, + { + "epoch": 3.18, + "learning_rate": 9.528708388596433e-06, + "loss": 0.9941, + "step": 42999 + }, + { + "epoch": 3.18, + "learning_rate": 9.528309960576099e-06, + "loss": 1.009, + "step": 43000 + }, + { + "epoch": 3.18, + "learning_rate": 9.52791153330621e-06, + "loss": 1.0387, + "step": 43001 + }, + { + "epoch": 3.18, + "learning_rate": 9.527513106787413e-06, + "loss": 1.0544, + "step": 43002 + }, + { + "epoch": 3.18, + "learning_rate": 9.52711468102033e-06, + "loss": 0.9294, + "step": 43003 + }, + { + "epoch": 3.18, + "learning_rate": 9.526716256005605e-06, + "loss": 0.8947, + "step": 43004 + }, + { + "epoch": 3.18, + "learning_rate": 9.526317831743869e-06, + "loss": 1.003, + "step": 43005 + }, + { + "epoch": 3.18, + "learning_rate": 9.52591940823575e-06, + "loss": 0.9931, + "step": 43006 + }, + { + "epoch": 3.18, + "learning_rate": 9.52552098548189e-06, + "loss": 0.9396, + "step": 43007 + }, + { + "epoch": 3.18, + "learning_rate": 9.525122563482917e-06, + "loss": 1.1051, + "step": 43008 + }, + { + "epoch": 3.18, + "learning_rate": 9.524724142239467e-06, + "loss": 0.9712, + "step": 43009 + }, + { + "epoch": 3.18, + "learning_rate": 9.524325721752175e-06, + "loss": 0.9658, + "step": 43010 + }, + { + "epoch": 3.18, + "learning_rate": 9.523927302021674e-06, + "loss": 1.0771, + "step": 43011 + }, + { + "epoch": 3.18, + "learning_rate": 9.523528883048598e-06, + "loss": 0.8518, + "step": 43012 + }, + { + "epoch": 3.18, + "learning_rate": 9.52313046483358e-06, + "loss": 0.8965, + "step": 43013 + }, + { + "epoch": 3.18, + "learning_rate": 9.522732047377257e-06, + "loss": 1.0937, + "step": 43014 + }, + { + "epoch": 3.18, + "learning_rate": 9.522333630680257e-06, + "loss": 0.9465, + "step": 43015 + }, + { + "epoch": 3.18, + "learning_rate": 9.52193521474322e-06, + "loss": 1.0451, + "step": 43016 + }, + { + "epoch": 3.18, + "learning_rate": 9.521536799566775e-06, + "loss": 0.9782, + "step": 43017 + }, + { + "epoch": 3.18, + "learning_rate": 9.52113838515156e-06, + "loss": 1.0222, + "step": 43018 + }, + { + "epoch": 3.18, + "learning_rate": 9.520739971498205e-06, + "loss": 1.0265, + "step": 43019 + }, + { + "epoch": 3.18, + "learning_rate": 9.520341558607348e-06, + "loss": 0.9885, + "step": 43020 + }, + { + "epoch": 3.18, + "learning_rate": 9.51994314647962e-06, + "loss": 1.0525, + "step": 43021 + }, + { + "epoch": 3.18, + "learning_rate": 9.519544735115654e-06, + "loss": 1.0229, + "step": 43022 + }, + { + "epoch": 3.18, + "learning_rate": 9.51914632451609e-06, + "loss": 0.9414, + "step": 43023 + }, + { + "epoch": 3.18, + "learning_rate": 9.518747914681551e-06, + "loss": 0.9644, + "step": 43024 + }, + { + "epoch": 3.18, + "learning_rate": 9.518349505612685e-06, + "loss": 1.0385, + "step": 43025 + }, + { + "epoch": 3.18, + "learning_rate": 9.51795109731011e-06, + "loss": 1.0003, + "step": 43026 + }, + { + "epoch": 3.18, + "learning_rate": 9.517552689774474e-06, + "loss": 0.9813, + "step": 43027 + }, + { + "epoch": 3.18, + "learning_rate": 9.517154283006398e-06, + "loss": 0.9186, + "step": 43028 + }, + { + "epoch": 3.18, + "learning_rate": 9.51675587700653e-06, + "loss": 1.098, + "step": 43029 + }, + { + "epoch": 3.18, + "learning_rate": 9.516357471775493e-06, + "loss": 1.1327, + "step": 43030 + }, + { + "epoch": 3.18, + "learning_rate": 9.515959067313925e-06, + "loss": 1.0468, + "step": 43031 + }, + { + "epoch": 3.18, + "learning_rate": 9.515560663622457e-06, + "loss": 1.0052, + "step": 43032 + }, + { + "epoch": 3.18, + "learning_rate": 9.515162260701728e-06, + "loss": 1.0042, + "step": 43033 + }, + { + "epoch": 3.18, + "learning_rate": 9.514763858552368e-06, + "loss": 1.077, + "step": 43034 + }, + { + "epoch": 3.18, + "learning_rate": 9.51436545717501e-06, + "loss": 0.9757, + "step": 43035 + }, + { + "epoch": 3.18, + "learning_rate": 9.513967056570292e-06, + "loss": 1.0362, + "step": 43036 + }, + { + "epoch": 3.18, + "learning_rate": 9.513568656738844e-06, + "loss": 0.8844, + "step": 43037 + }, + { + "epoch": 3.18, + "learning_rate": 9.513170257681304e-06, + "loss": 1.0469, + "step": 43038 + }, + { + "epoch": 3.18, + "learning_rate": 9.512771859398299e-06, + "loss": 0.8842, + "step": 43039 + }, + { + "epoch": 3.18, + "learning_rate": 9.51237346189047e-06, + "loss": 0.9162, + "step": 43040 + }, + { + "epoch": 3.18, + "learning_rate": 9.511975065158445e-06, + "loss": 1.0602, + "step": 43041 + }, + { + "epoch": 3.18, + "learning_rate": 9.511576669202864e-06, + "loss": 0.9562, + "step": 43042 + }, + { + "epoch": 3.18, + "learning_rate": 9.51117827402436e-06, + "loss": 0.9976, + "step": 43043 + }, + { + "epoch": 3.18, + "learning_rate": 9.510779879623556e-06, + "loss": 0.9799, + "step": 43044 + }, + { + "epoch": 3.18, + "learning_rate": 9.510381486001102e-06, + "loss": 1.0111, + "step": 43045 + }, + { + "epoch": 3.18, + "learning_rate": 9.509983093157618e-06, + "loss": 1.0862, + "step": 43046 + }, + { + "epoch": 3.18, + "learning_rate": 9.509584701093749e-06, + "loss": 0.9306, + "step": 43047 + }, + { + "epoch": 3.18, + "learning_rate": 9.509186309810117e-06, + "loss": 1.0178, + "step": 43048 + }, + { + "epoch": 3.18, + "learning_rate": 9.50878791930737e-06, + "loss": 0.9368, + "step": 43049 + }, + { + "epoch": 3.18, + "learning_rate": 9.508389529586129e-06, + "loss": 0.9664, + "step": 43050 + }, + { + "epoch": 3.18, + "learning_rate": 9.507991140647039e-06, + "loss": 1.0331, + "step": 43051 + }, + { + "epoch": 3.18, + "learning_rate": 9.507592752490725e-06, + "loss": 1.0529, + "step": 43052 + }, + { + "epoch": 3.18, + "learning_rate": 9.507194365117822e-06, + "loss": 1.0747, + "step": 43053 + }, + { + "epoch": 3.18, + "learning_rate": 9.50679597852897e-06, + "loss": 1.0845, + "step": 43054 + }, + { + "epoch": 3.18, + "learning_rate": 9.506397592724792e-06, + "loss": 1.0032, + "step": 43055 + }, + { + "epoch": 3.18, + "learning_rate": 9.505999207705934e-06, + "loss": 1.0521, + "step": 43056 + }, + { + "epoch": 3.18, + "learning_rate": 9.50560082347302e-06, + "loss": 1.0055, + "step": 43057 + }, + { + "epoch": 3.18, + "learning_rate": 9.505202440026691e-06, + "loss": 1.0668, + "step": 43058 + }, + { + "epoch": 3.18, + "learning_rate": 9.504804057367577e-06, + "loss": 1.011, + "step": 43059 + }, + { + "epoch": 3.18, + "learning_rate": 9.504405675496313e-06, + "loss": 0.9862, + "step": 43060 + }, + { + "epoch": 3.18, + "learning_rate": 9.504007294413533e-06, + "loss": 0.9326, + "step": 43061 + }, + { + "epoch": 3.18, + "learning_rate": 9.503608914119868e-06, + "loss": 1.084, + "step": 43062 + }, + { + "epoch": 3.18, + "learning_rate": 9.503210534615958e-06, + "loss": 1.0939, + "step": 43063 + }, + { + "epoch": 3.18, + "learning_rate": 9.502812155902429e-06, + "loss": 0.9825, + "step": 43064 + }, + { + "epoch": 3.18, + "learning_rate": 9.50241377797992e-06, + "loss": 1.0344, + "step": 43065 + }, + { + "epoch": 3.18, + "learning_rate": 9.502015400849065e-06, + "loss": 1.0088, + "step": 43066 + }, + { + "epoch": 3.18, + "learning_rate": 9.501617024510496e-06, + "loss": 0.8791, + "step": 43067 + }, + { + "epoch": 3.18, + "learning_rate": 9.501218648964844e-06, + "loss": 1.0363, + "step": 43068 + }, + { + "epoch": 3.18, + "learning_rate": 9.50082027421275e-06, + "loss": 1.0693, + "step": 43069 + }, + { + "epoch": 3.18, + "learning_rate": 9.500421900254845e-06, + "loss": 0.9529, + "step": 43070 + }, + { + "epoch": 3.18, + "learning_rate": 9.500023527091756e-06, + "loss": 0.9805, + "step": 43071 + }, + { + "epoch": 3.18, + "learning_rate": 9.499625154724129e-06, + "loss": 1.0022, + "step": 43072 + }, + { + "epoch": 3.18, + "learning_rate": 9.499226783152584e-06, + "loss": 1.0178, + "step": 43073 + }, + { + "epoch": 3.18, + "learning_rate": 9.498828412377768e-06, + "loss": 0.9829, + "step": 43074 + }, + { + "epoch": 3.18, + "learning_rate": 9.498430042400305e-06, + "loss": 0.8475, + "step": 43075 + }, + { + "epoch": 3.18, + "learning_rate": 9.498031673220834e-06, + "loss": 0.997, + "step": 43076 + }, + { + "epoch": 3.18, + "learning_rate": 9.497633304839987e-06, + "loss": 0.9948, + "step": 43077 + }, + { + "epoch": 3.18, + "learning_rate": 9.4972349372584e-06, + "loss": 0.9544, + "step": 43078 + }, + { + "epoch": 3.18, + "learning_rate": 9.496836570476704e-06, + "loss": 1.0276, + "step": 43079 + }, + { + "epoch": 3.18, + "learning_rate": 9.496438204495532e-06, + "loss": 1.004, + "step": 43080 + }, + { + "epoch": 3.18, + "learning_rate": 9.496039839315522e-06, + "loss": 0.9685, + "step": 43081 + }, + { + "epoch": 3.18, + "learning_rate": 9.495641474937304e-06, + "loss": 1.0388, + "step": 43082 + }, + { + "epoch": 3.18, + "learning_rate": 9.495243111361513e-06, + "loss": 1.1402, + "step": 43083 + }, + { + "epoch": 3.18, + "learning_rate": 9.494844748588782e-06, + "loss": 1.0305, + "step": 43084 + }, + { + "epoch": 3.18, + "learning_rate": 9.49444638661975e-06, + "loss": 1.0486, + "step": 43085 + }, + { + "epoch": 3.18, + "learning_rate": 9.494048025455043e-06, + "loss": 0.9822, + "step": 43086 + }, + { + "epoch": 3.18, + "learning_rate": 9.4936496650953e-06, + "loss": 1.0164, + "step": 43087 + }, + { + "epoch": 3.18, + "learning_rate": 9.49325130554115e-06, + "loss": 0.9442, + "step": 43088 + }, + { + "epoch": 3.18, + "learning_rate": 9.492852946793234e-06, + "loss": 0.9415, + "step": 43089 + }, + { + "epoch": 3.18, + "learning_rate": 9.492454588852184e-06, + "loss": 1.0706, + "step": 43090 + }, + { + "epoch": 3.18, + "learning_rate": 9.492056231718624e-06, + "loss": 1.0011, + "step": 43091 + }, + { + "epoch": 3.18, + "learning_rate": 9.491657875393203e-06, + "loss": 1.0245, + "step": 43092 + }, + { + "epoch": 3.18, + "learning_rate": 9.491259519876539e-06, + "loss": 1.0109, + "step": 43093 + }, + { + "epoch": 3.18, + "learning_rate": 9.490861165169281e-06, + "loss": 1.0525, + "step": 43094 + }, + { + "epoch": 3.18, + "learning_rate": 9.490462811272049e-06, + "loss": 0.9627, + "step": 43095 + }, + { + "epoch": 3.18, + "learning_rate": 9.490064458185489e-06, + "loss": 1.0264, + "step": 43096 + }, + { + "epoch": 3.18, + "learning_rate": 9.489666105910223e-06, + "loss": 0.9506, + "step": 43097 + }, + { + "epoch": 3.18, + "learning_rate": 9.489267754446898e-06, + "loss": 0.9615, + "step": 43098 + }, + { + "epoch": 3.18, + "learning_rate": 9.488869403796139e-06, + "loss": 1.0757, + "step": 43099 + }, + { + "epoch": 3.18, + "learning_rate": 9.488471053958579e-06, + "loss": 0.9372, + "step": 43100 + }, + { + "epoch": 3.18, + "learning_rate": 9.488072704934856e-06, + "loss": 1.0488, + "step": 43101 + }, + { + "epoch": 3.18, + "learning_rate": 9.487674356725601e-06, + "loss": 1.0515, + "step": 43102 + }, + { + "epoch": 3.18, + "learning_rate": 9.48727600933145e-06, + "loss": 0.9083, + "step": 43103 + }, + { + "epoch": 3.18, + "learning_rate": 9.486877662753033e-06, + "loss": 0.9594, + "step": 43104 + }, + { + "epoch": 3.19, + "learning_rate": 9.486479316990987e-06, + "loss": 0.9736, + "step": 43105 + }, + { + "epoch": 3.19, + "learning_rate": 9.486080972045946e-06, + "loss": 0.9458, + "step": 43106 + }, + { + "epoch": 3.19, + "learning_rate": 9.485682627918543e-06, + "loss": 1.0715, + "step": 43107 + }, + { + "epoch": 3.19, + "learning_rate": 9.485284284609411e-06, + "loss": 0.9557, + "step": 43108 + }, + { + "epoch": 3.19, + "learning_rate": 9.484885942119184e-06, + "loss": 1.0483, + "step": 43109 + }, + { + "epoch": 3.19, + "learning_rate": 9.484487600448498e-06, + "loss": 0.9424, + "step": 43110 + }, + { + "epoch": 3.19, + "learning_rate": 9.484089259597983e-06, + "loss": 0.9464, + "step": 43111 + }, + { + "epoch": 3.19, + "learning_rate": 9.483690919568276e-06, + "loss": 0.9533, + "step": 43112 + }, + { + "epoch": 3.19, + "learning_rate": 9.483292580360007e-06, + "loss": 1.1119, + "step": 43113 + }, + { + "epoch": 3.19, + "learning_rate": 9.482894241973815e-06, + "loss": 1.0268, + "step": 43114 + }, + { + "epoch": 3.19, + "learning_rate": 9.482495904410326e-06, + "loss": 0.8495, + "step": 43115 + }, + { + "epoch": 3.19, + "learning_rate": 9.482097567670184e-06, + "loss": 0.9826, + "step": 43116 + }, + { + "epoch": 3.19, + "learning_rate": 9.48169923175402e-06, + "loss": 1.0253, + "step": 43117 + }, + { + "epoch": 3.19, + "learning_rate": 9.48130089666246e-06, + "loss": 1.0525, + "step": 43118 + }, + { + "epoch": 3.19, + "learning_rate": 9.480902562396143e-06, + "loss": 1.0122, + "step": 43119 + }, + { + "epoch": 3.19, + "learning_rate": 9.480504228955702e-06, + "loss": 0.8833, + "step": 43120 + }, + { + "epoch": 3.19, + "learning_rate": 9.480105896341774e-06, + "loss": 0.9811, + "step": 43121 + }, + { + "epoch": 3.19, + "learning_rate": 9.479707564554987e-06, + "loss": 1.0211, + "step": 43122 + }, + { + "epoch": 3.19, + "learning_rate": 9.479309233595978e-06, + "loss": 0.9323, + "step": 43123 + }, + { + "epoch": 3.19, + "learning_rate": 9.478910903465381e-06, + "loss": 0.9789, + "step": 43124 + }, + { + "epoch": 3.19, + "learning_rate": 9.478512574163831e-06, + "loss": 1.0161, + "step": 43125 + }, + { + "epoch": 3.19, + "learning_rate": 9.478114245691962e-06, + "loss": 0.9987, + "step": 43126 + }, + { + "epoch": 3.19, + "learning_rate": 9.4777159180504e-06, + "loss": 0.9094, + "step": 43127 + }, + { + "epoch": 3.19, + "learning_rate": 9.47731759123979e-06, + "loss": 0.8962, + "step": 43128 + }, + { + "epoch": 3.19, + "learning_rate": 9.476919265260755e-06, + "loss": 0.9235, + "step": 43129 + }, + { + "epoch": 3.19, + "learning_rate": 9.476520940113938e-06, + "loss": 0.9364, + "step": 43130 + }, + { + "epoch": 3.19, + "learning_rate": 9.476122615799967e-06, + "loss": 1.0697, + "step": 43131 + }, + { + "epoch": 3.19, + "learning_rate": 9.475724292319478e-06, + "loss": 1.0435, + "step": 43132 + }, + { + "epoch": 3.19, + "learning_rate": 9.475325969673103e-06, + "loss": 0.8703, + "step": 43133 + }, + { + "epoch": 3.19, + "learning_rate": 9.474927647861478e-06, + "loss": 1.1583, + "step": 43134 + }, + { + "epoch": 3.19, + "learning_rate": 9.474529326885239e-06, + "loss": 0.9602, + "step": 43135 + }, + { + "epoch": 3.19, + "learning_rate": 9.47413100674501e-06, + "loss": 1.0353, + "step": 43136 + }, + { + "epoch": 3.19, + "learning_rate": 9.473732687441436e-06, + "loss": 1.0014, + "step": 43137 + }, + { + "epoch": 3.19, + "learning_rate": 9.47333436897514e-06, + "loss": 0.9113, + "step": 43138 + }, + { + "epoch": 3.19, + "learning_rate": 9.472936051346768e-06, + "loss": 1.0693, + "step": 43139 + }, + { + "epoch": 3.19, + "learning_rate": 9.47253773455694e-06, + "loss": 1.0367, + "step": 43140 + }, + { + "epoch": 3.19, + "learning_rate": 9.472139418606305e-06, + "loss": 1.0817, + "step": 43141 + }, + { + "epoch": 3.19, + "learning_rate": 9.471741103495481e-06, + "loss": 1.0168, + "step": 43142 + }, + { + "epoch": 3.19, + "learning_rate": 9.471342789225117e-06, + "loss": 0.9706, + "step": 43143 + }, + { + "epoch": 3.19, + "learning_rate": 9.470944475795831e-06, + "loss": 1.0177, + "step": 43144 + }, + { + "epoch": 3.19, + "learning_rate": 9.470546163208271e-06, + "loss": 1.0254, + "step": 43145 + }, + { + "epoch": 3.19, + "learning_rate": 9.470147851463065e-06, + "loss": 1.0451, + "step": 43146 + }, + { + "epoch": 3.19, + "learning_rate": 9.469749540560842e-06, + "loss": 1.0212, + "step": 43147 + }, + { + "epoch": 3.19, + "learning_rate": 9.469351230502241e-06, + "loss": 0.9035, + "step": 43148 + }, + { + "epoch": 3.19, + "learning_rate": 9.468952921287895e-06, + "loss": 1.0547, + "step": 43149 + }, + { + "epoch": 3.19, + "learning_rate": 9.468554612918438e-06, + "loss": 1.0296, + "step": 43150 + }, + { + "epoch": 3.19, + "learning_rate": 9.468156305394502e-06, + "loss": 1.0037, + "step": 43151 + }, + { + "epoch": 3.19, + "learning_rate": 9.467757998716723e-06, + "loss": 1.1214, + "step": 43152 + }, + { + "epoch": 3.19, + "learning_rate": 9.46735969288573e-06, + "loss": 0.9569, + "step": 43153 + }, + { + "epoch": 3.19, + "learning_rate": 9.466961387902165e-06, + "loss": 1.0129, + "step": 43154 + }, + { + "epoch": 3.19, + "learning_rate": 9.466563083766655e-06, + "loss": 0.9617, + "step": 43155 + }, + { + "epoch": 3.19, + "learning_rate": 9.466164780479833e-06, + "loss": 0.9758, + "step": 43156 + }, + { + "epoch": 3.19, + "learning_rate": 9.465766478042338e-06, + "loss": 0.8933, + "step": 43157 + }, + { + "epoch": 3.19, + "learning_rate": 9.465368176454798e-06, + "loss": 1.0633, + "step": 43158 + }, + { + "epoch": 3.19, + "learning_rate": 9.464969875717854e-06, + "loss": 0.984, + "step": 43159 + }, + { + "epoch": 3.19, + "learning_rate": 9.464571575832132e-06, + "loss": 0.9905, + "step": 43160 + }, + { + "epoch": 3.19, + "learning_rate": 9.46417327679827e-06, + "loss": 1.0911, + "step": 43161 + }, + { + "epoch": 3.19, + "learning_rate": 9.463774978616898e-06, + "loss": 1.0227, + "step": 43162 + }, + { + "epoch": 3.19, + "learning_rate": 9.463376681288658e-06, + "loss": 0.9239, + "step": 43163 + }, + { + "epoch": 3.19, + "learning_rate": 9.462978384814177e-06, + "loss": 1.0947, + "step": 43164 + }, + { + "epoch": 3.19, + "learning_rate": 9.462580089194087e-06, + "loss": 0.8824, + "step": 43165 + }, + { + "epoch": 3.19, + "learning_rate": 9.462181794429025e-06, + "loss": 1.021, + "step": 43166 + }, + { + "epoch": 3.19, + "learning_rate": 9.461783500519622e-06, + "loss": 1.0549, + "step": 43167 + }, + { + "epoch": 3.19, + "learning_rate": 9.461385207466517e-06, + "loss": 0.9802, + "step": 43168 + }, + { + "epoch": 3.19, + "learning_rate": 9.460986915270338e-06, + "loss": 1.1178, + "step": 43169 + }, + { + "epoch": 3.19, + "learning_rate": 9.460588623931724e-06, + "loss": 0.9615, + "step": 43170 + }, + { + "epoch": 3.19, + "learning_rate": 9.460190333451302e-06, + "loss": 1.0455, + "step": 43171 + }, + { + "epoch": 3.19, + "learning_rate": 9.459792043829713e-06, + "loss": 1.0293, + "step": 43172 + }, + { + "epoch": 3.19, + "learning_rate": 9.459393755067586e-06, + "loss": 1.0126, + "step": 43173 + }, + { + "epoch": 3.19, + "learning_rate": 9.458995467165555e-06, + "loss": 0.943, + "step": 43174 + }, + { + "epoch": 3.19, + "learning_rate": 9.458597180124255e-06, + "loss": 0.9905, + "step": 43175 + }, + { + "epoch": 3.19, + "learning_rate": 9.458198893944319e-06, + "loss": 1.1287, + "step": 43176 + }, + { + "epoch": 3.19, + "learning_rate": 9.45780060862638e-06, + "loss": 1.0124, + "step": 43177 + }, + { + "epoch": 3.19, + "learning_rate": 9.457402324171073e-06, + "loss": 0.9275, + "step": 43178 + }, + { + "epoch": 3.19, + "learning_rate": 9.457004040579031e-06, + "loss": 0.9647, + "step": 43179 + }, + { + "epoch": 3.19, + "learning_rate": 9.456605757850888e-06, + "loss": 1.0714, + "step": 43180 + }, + { + "epoch": 3.19, + "learning_rate": 9.45620747598728e-06, + "loss": 1.0656, + "step": 43181 + }, + { + "epoch": 3.19, + "learning_rate": 9.45580919498884e-06, + "loss": 0.9712, + "step": 43182 + }, + { + "epoch": 3.19, + "learning_rate": 9.45541091485619e-06, + "loss": 1.0459, + "step": 43183 + }, + { + "epoch": 3.19, + "learning_rate": 9.455012635589982e-06, + "loss": 0.9017, + "step": 43184 + }, + { + "epoch": 3.19, + "learning_rate": 9.454614357190835e-06, + "loss": 1.0293, + "step": 43185 + }, + { + "epoch": 3.19, + "learning_rate": 9.454216079659395e-06, + "loss": 0.9584, + "step": 43186 + }, + { + "epoch": 3.19, + "learning_rate": 9.453817802996283e-06, + "loss": 0.9673, + "step": 43187 + }, + { + "epoch": 3.19, + "learning_rate": 9.453419527202146e-06, + "loss": 0.9975, + "step": 43188 + }, + { + "epoch": 3.19, + "learning_rate": 9.453021252277603e-06, + "loss": 1.0735, + "step": 43189 + }, + { + "epoch": 3.19, + "learning_rate": 9.452622978223303e-06, + "loss": 1.003, + "step": 43190 + }, + { + "epoch": 3.19, + "learning_rate": 9.452224705039866e-06, + "loss": 0.9885, + "step": 43191 + }, + { + "epoch": 3.19, + "learning_rate": 9.451826432727936e-06, + "loss": 1.0444, + "step": 43192 + }, + { + "epoch": 3.19, + "learning_rate": 9.451428161288142e-06, + "loss": 0.9024, + "step": 43193 + }, + { + "epoch": 3.19, + "learning_rate": 9.451029890721115e-06, + "loss": 1.0259, + "step": 43194 + }, + { + "epoch": 3.19, + "learning_rate": 9.450631621027495e-06, + "loss": 1.0304, + "step": 43195 + }, + { + "epoch": 3.19, + "learning_rate": 9.450233352207909e-06, + "loss": 1.0045, + "step": 43196 + }, + { + "epoch": 3.19, + "learning_rate": 9.449835084262995e-06, + "loss": 0.9385, + "step": 43197 + }, + { + "epoch": 3.19, + "learning_rate": 9.449436817193384e-06, + "loss": 0.9357, + "step": 43198 + }, + { + "epoch": 3.19, + "learning_rate": 9.449038550999714e-06, + "loss": 0.9753, + "step": 43199 + }, + { + "epoch": 3.19, + "learning_rate": 9.448640285682613e-06, + "loss": 1.0078, + "step": 43200 + }, + { + "epoch": 3.19, + "learning_rate": 9.448242021242721e-06, + "loss": 0.9593, + "step": 43201 + }, + { + "epoch": 3.19, + "learning_rate": 9.447843757680665e-06, + "loss": 1.0545, + "step": 43202 + }, + { + "epoch": 3.19, + "learning_rate": 9.447445494997083e-06, + "loss": 0.9898, + "step": 43203 + }, + { + "epoch": 3.19, + "learning_rate": 9.447047233192609e-06, + "loss": 1.0398, + "step": 43204 + }, + { + "epoch": 3.19, + "learning_rate": 9.44664897226787e-06, + "loss": 0.9391, + "step": 43205 + }, + { + "epoch": 3.19, + "learning_rate": 9.446250712223508e-06, + "loss": 1.0221, + "step": 43206 + }, + { + "epoch": 3.19, + "learning_rate": 9.445852453060152e-06, + "loss": 0.9737, + "step": 43207 + }, + { + "epoch": 3.19, + "learning_rate": 9.44545419477844e-06, + "loss": 0.9745, + "step": 43208 + }, + { + "epoch": 3.19, + "learning_rate": 9.445055937378997e-06, + "loss": 1.0072, + "step": 43209 + }, + { + "epoch": 3.19, + "learning_rate": 9.444657680862468e-06, + "loss": 1.0552, + "step": 43210 + }, + { + "epoch": 3.19, + "learning_rate": 9.444259425229479e-06, + "loss": 1.0695, + "step": 43211 + }, + { + "epoch": 3.19, + "learning_rate": 9.443861170480662e-06, + "loss": 0.9841, + "step": 43212 + }, + { + "epoch": 3.19, + "learning_rate": 9.443462916616658e-06, + "loss": 0.9072, + "step": 43213 + }, + { + "epoch": 3.19, + "learning_rate": 9.443064663638092e-06, + "loss": 0.992, + "step": 43214 + }, + { + "epoch": 3.19, + "learning_rate": 9.442666411545607e-06, + "loss": 1.0403, + "step": 43215 + }, + { + "epoch": 3.19, + "learning_rate": 9.442268160339827e-06, + "loss": 1.0796, + "step": 43216 + }, + { + "epoch": 3.19, + "learning_rate": 9.441869910021394e-06, + "loss": 1.0766, + "step": 43217 + }, + { + "epoch": 3.19, + "learning_rate": 9.441471660590934e-06, + "loss": 0.9288, + "step": 43218 + }, + { + "epoch": 3.19, + "learning_rate": 9.44107341204909e-06, + "loss": 0.9542, + "step": 43219 + }, + { + "epoch": 3.19, + "learning_rate": 9.440675164396489e-06, + "loss": 1.0912, + "step": 43220 + }, + { + "epoch": 3.19, + "learning_rate": 9.440276917633763e-06, + "loss": 0.9991, + "step": 43221 + }, + { + "epoch": 3.19, + "learning_rate": 9.43987867176155e-06, + "loss": 0.9215, + "step": 43222 + }, + { + "epoch": 3.19, + "learning_rate": 9.439480426780482e-06, + "loss": 1.0645, + "step": 43223 + }, + { + "epoch": 3.19, + "learning_rate": 9.439082182691192e-06, + "loss": 0.918, + "step": 43224 + }, + { + "epoch": 3.19, + "learning_rate": 9.438683939494313e-06, + "loss": 1.0383, + "step": 43225 + }, + { + "epoch": 3.19, + "learning_rate": 9.438285697190482e-06, + "loss": 0.9523, + "step": 43226 + }, + { + "epoch": 3.19, + "learning_rate": 9.437887455780329e-06, + "loss": 0.9863, + "step": 43227 + }, + { + "epoch": 3.19, + "learning_rate": 9.437489215264491e-06, + "loss": 1.0333, + "step": 43228 + }, + { + "epoch": 3.19, + "learning_rate": 9.437090975643602e-06, + "loss": 0.8809, + "step": 43229 + }, + { + "epoch": 3.19, + "learning_rate": 9.436692736918287e-06, + "loss": 0.9771, + "step": 43230 + }, + { + "epoch": 3.19, + "learning_rate": 9.436294499089192e-06, + "loss": 0.9381, + "step": 43231 + }, + { + "epoch": 3.19, + "learning_rate": 9.435896262156936e-06, + "loss": 0.9754, + "step": 43232 + }, + { + "epoch": 3.19, + "learning_rate": 9.43549802612217e-06, + "loss": 0.931, + "step": 43233 + }, + { + "epoch": 3.19, + "learning_rate": 9.43509979098551e-06, + "loss": 1.0377, + "step": 43234 + }, + { + "epoch": 3.19, + "learning_rate": 9.434701556747606e-06, + "loss": 1.0945, + "step": 43235 + }, + { + "epoch": 3.19, + "learning_rate": 9.434303323409077e-06, + "loss": 0.942, + "step": 43236 + }, + { + "epoch": 3.19, + "learning_rate": 9.43390509097057e-06, + "loss": 1.0308, + "step": 43237 + }, + { + "epoch": 3.19, + "learning_rate": 9.43350685943271e-06, + "loss": 0.8691, + "step": 43238 + }, + { + "epoch": 3.19, + "learning_rate": 9.433108628796128e-06, + "loss": 0.9909, + "step": 43239 + }, + { + "epoch": 3.19, + "learning_rate": 9.432710399061466e-06, + "loss": 1.068, + "step": 43240 + }, + { + "epoch": 3.2, + "learning_rate": 9.432312170229352e-06, + "loss": 1.178, + "step": 43241 + }, + { + "epoch": 3.2, + "learning_rate": 9.431913942300424e-06, + "loss": 0.9885, + "step": 43242 + }, + { + "epoch": 3.2, + "learning_rate": 9.43151571527531e-06, + "loss": 0.9652, + "step": 43243 + }, + { + "epoch": 3.2, + "learning_rate": 9.431117489154646e-06, + "loss": 0.9998, + "step": 43244 + }, + { + "epoch": 3.2, + "learning_rate": 9.430719263939066e-06, + "loss": 0.9086, + "step": 43245 + }, + { + "epoch": 3.2, + "learning_rate": 9.430321039629205e-06, + "loss": 1.0478, + "step": 43246 + }, + { + "epoch": 3.2, + "learning_rate": 9.429922816225694e-06, + "loss": 0.9952, + "step": 43247 + }, + { + "epoch": 3.2, + "learning_rate": 9.42952459372917e-06, + "loss": 0.9911, + "step": 43248 + }, + { + "epoch": 3.2, + "learning_rate": 9.429126372140263e-06, + "loss": 1.0114, + "step": 43249 + }, + { + "epoch": 3.2, + "learning_rate": 9.428728151459605e-06, + "loss": 0.9212, + "step": 43250 + }, + { + "epoch": 3.2, + "learning_rate": 9.42832993168784e-06, + "loss": 0.9781, + "step": 43251 + }, + { + "epoch": 3.2, + "learning_rate": 9.427931712825583e-06, + "loss": 1.0438, + "step": 43252 + }, + { + "epoch": 3.2, + "learning_rate": 9.427533494873488e-06, + "loss": 1.1466, + "step": 43253 + }, + { + "epoch": 3.2, + "learning_rate": 9.42713527783217e-06, + "loss": 0.9643, + "step": 43254 + }, + { + "epoch": 3.2, + "learning_rate": 9.42673706170228e-06, + "loss": 0.8789, + "step": 43255 + }, + { + "epoch": 3.2, + "learning_rate": 9.426338846484436e-06, + "loss": 1.0335, + "step": 43256 + }, + { + "epoch": 3.2, + "learning_rate": 9.425940632179286e-06, + "loss": 0.9668, + "step": 43257 + }, + { + "epoch": 3.2, + "learning_rate": 9.425542418787453e-06, + "loss": 0.9065, + "step": 43258 + }, + { + "epoch": 3.2, + "learning_rate": 9.425144206309573e-06, + "loss": 1.082, + "step": 43259 + }, + { + "epoch": 3.2, + "learning_rate": 9.424745994746282e-06, + "loss": 0.9905, + "step": 43260 + }, + { + "epoch": 3.2, + "learning_rate": 9.424347784098209e-06, + "loss": 0.9554, + "step": 43261 + }, + { + "epoch": 3.2, + "learning_rate": 9.423949574365993e-06, + "loss": 0.9507, + "step": 43262 + }, + { + "epoch": 3.2, + "learning_rate": 9.423551365550263e-06, + "loss": 0.9465, + "step": 43263 + }, + { + "epoch": 3.2, + "learning_rate": 9.423153157651657e-06, + "loss": 0.9814, + "step": 43264 + }, + { + "epoch": 3.2, + "learning_rate": 9.422754950670803e-06, + "loss": 1.034, + "step": 43265 + }, + { + "epoch": 3.2, + "learning_rate": 9.42235674460834e-06, + "loss": 0.9697, + "step": 43266 + }, + { + "epoch": 3.2, + "learning_rate": 9.4219585394649e-06, + "loss": 0.9774, + "step": 43267 + }, + { + "epoch": 3.2, + "learning_rate": 9.421560335241113e-06, + "loss": 0.9808, + "step": 43268 + }, + { + "epoch": 3.2, + "learning_rate": 9.421162131937616e-06, + "loss": 0.9987, + "step": 43269 + }, + { + "epoch": 3.2, + "learning_rate": 9.42076392955504e-06, + "loss": 0.9015, + "step": 43270 + }, + { + "epoch": 3.2, + "learning_rate": 9.420365728094024e-06, + "loss": 1.0289, + "step": 43271 + }, + { + "epoch": 3.2, + "learning_rate": 9.419967527555195e-06, + "loss": 0.9787, + "step": 43272 + }, + { + "epoch": 3.2, + "learning_rate": 9.419569327939192e-06, + "loss": 0.9771, + "step": 43273 + }, + { + "epoch": 3.2, + "learning_rate": 9.419171129246642e-06, + "loss": 1.0961, + "step": 43274 + }, + { + "epoch": 3.2, + "learning_rate": 9.418772931478187e-06, + "loss": 1.0418, + "step": 43275 + }, + { + "epoch": 3.2, + "learning_rate": 9.418374734634457e-06, + "loss": 1.0948, + "step": 43276 + }, + { + "epoch": 3.2, + "learning_rate": 9.417976538716078e-06, + "loss": 1.0829, + "step": 43277 + }, + { + "epoch": 3.2, + "learning_rate": 9.417578343723697e-06, + "loss": 1.1135, + "step": 43278 + }, + { + "epoch": 3.2, + "learning_rate": 9.417180149657932e-06, + "loss": 1.0533, + "step": 43279 + }, + { + "epoch": 3.2, + "learning_rate": 9.416781956519434e-06, + "loss": 0.9979, + "step": 43280 + }, + { + "epoch": 3.2, + "learning_rate": 9.41638376430882e-06, + "loss": 1.059, + "step": 43281 + }, + { + "epoch": 3.2, + "learning_rate": 9.415985573026737e-06, + "loss": 0.9254, + "step": 43282 + }, + { + "epoch": 3.2, + "learning_rate": 9.415587382673808e-06, + "loss": 0.9112, + "step": 43283 + }, + { + "epoch": 3.2, + "learning_rate": 9.415189193250676e-06, + "loss": 1.0075, + "step": 43284 + }, + { + "epoch": 3.2, + "learning_rate": 9.414791004757967e-06, + "loss": 1.0726, + "step": 43285 + }, + { + "epoch": 3.2, + "learning_rate": 9.414392817196316e-06, + "loss": 1.0385, + "step": 43286 + }, + { + "epoch": 3.2, + "learning_rate": 9.41399463056636e-06, + "loss": 1.0743, + "step": 43287 + }, + { + "epoch": 3.2, + "learning_rate": 9.413596444868727e-06, + "loss": 0.9478, + "step": 43288 + }, + { + "epoch": 3.2, + "learning_rate": 9.413198260104056e-06, + "loss": 1.0049, + "step": 43289 + }, + { + "epoch": 3.2, + "learning_rate": 9.412800076272976e-06, + "loss": 1.0221, + "step": 43290 + }, + { + "epoch": 3.2, + "learning_rate": 9.412401893376127e-06, + "loss": 1.0142, + "step": 43291 + }, + { + "epoch": 3.2, + "learning_rate": 9.412003711414133e-06, + "loss": 0.8471, + "step": 43292 + }, + { + "epoch": 3.2, + "learning_rate": 9.411605530387637e-06, + "loss": 1.0101, + "step": 43293 + }, + { + "epoch": 3.2, + "learning_rate": 9.411207350297267e-06, + "loss": 0.9948, + "step": 43294 + }, + { + "epoch": 3.2, + "learning_rate": 9.410809171143655e-06, + "loss": 1.0704, + "step": 43295 + }, + { + "epoch": 3.2, + "learning_rate": 9.410410992927443e-06, + "loss": 0.9994, + "step": 43296 + }, + { + "epoch": 3.2, + "learning_rate": 9.410012815649253e-06, + "loss": 1.1383, + "step": 43297 + }, + { + "epoch": 3.2, + "learning_rate": 9.409614639309727e-06, + "loss": 1.0075, + "step": 43298 + }, + { + "epoch": 3.2, + "learning_rate": 9.409216463909491e-06, + "loss": 1.059, + "step": 43299 + }, + { + "epoch": 3.2, + "learning_rate": 9.408818289449191e-06, + "loss": 0.9154, + "step": 43300 + }, + { + "epoch": 3.2, + "learning_rate": 9.408420115929445e-06, + "loss": 0.9497, + "step": 43301 + }, + { + "epoch": 3.2, + "learning_rate": 9.408021943350901e-06, + "loss": 0.9351, + "step": 43302 + }, + { + "epoch": 3.2, + "learning_rate": 9.407623771714179e-06, + "loss": 0.9731, + "step": 43303 + }, + { + "epoch": 3.2, + "learning_rate": 9.407225601019926e-06, + "loss": 1.0703, + "step": 43304 + }, + { + "epoch": 3.2, + "learning_rate": 9.406827431268766e-06, + "loss": 0.9614, + "step": 43305 + }, + { + "epoch": 3.2, + "learning_rate": 9.406429262461332e-06, + "loss": 0.8834, + "step": 43306 + }, + { + "epoch": 3.2, + "learning_rate": 9.406031094598263e-06, + "loss": 0.9997, + "step": 43307 + }, + { + "epoch": 3.2, + "learning_rate": 9.40563292768019e-06, + "loss": 1.0083, + "step": 43308 + }, + { + "epoch": 3.2, + "learning_rate": 9.405234761707746e-06, + "loss": 0.8951, + "step": 43309 + }, + { + "epoch": 3.2, + "learning_rate": 9.404836596681565e-06, + "loss": 1.0179, + "step": 43310 + }, + { + "epoch": 3.2, + "learning_rate": 9.404438432602281e-06, + "loss": 0.8796, + "step": 43311 + }, + { + "epoch": 3.2, + "learning_rate": 9.404040269470525e-06, + "loss": 0.9816, + "step": 43312 + }, + { + "epoch": 3.2, + "learning_rate": 9.403642107286934e-06, + "loss": 1.0625, + "step": 43313 + }, + { + "epoch": 3.2, + "learning_rate": 9.403243946052142e-06, + "loss": 0.8933, + "step": 43314 + }, + { + "epoch": 3.2, + "learning_rate": 9.402845785766777e-06, + "loss": 0.9768, + "step": 43315 + }, + { + "epoch": 3.2, + "learning_rate": 9.402447626431477e-06, + "loss": 0.9794, + "step": 43316 + }, + { + "epoch": 3.2, + "learning_rate": 9.402049468046872e-06, + "loss": 0.9631, + "step": 43317 + }, + { + "epoch": 3.2, + "learning_rate": 9.401651310613602e-06, + "loss": 1.0882, + "step": 43318 + }, + { + "epoch": 3.2, + "learning_rate": 9.401253154132292e-06, + "loss": 1.0307, + "step": 43319 + }, + { + "epoch": 3.2, + "learning_rate": 9.400854998603583e-06, + "loss": 1.0936, + "step": 43320 + }, + { + "epoch": 3.2, + "learning_rate": 9.400456844028103e-06, + "loss": 1.0223, + "step": 43321 + }, + { + "epoch": 3.2, + "learning_rate": 9.400058690406491e-06, + "loss": 1.0852, + "step": 43322 + }, + { + "epoch": 3.2, + "learning_rate": 9.399660537739376e-06, + "loss": 1.0433, + "step": 43323 + }, + { + "epoch": 3.2, + "learning_rate": 9.399262386027388e-06, + "loss": 1.0994, + "step": 43324 + }, + { + "epoch": 3.2, + "learning_rate": 9.39886423527117e-06, + "loss": 0.8292, + "step": 43325 + }, + { + "epoch": 3.2, + "learning_rate": 9.398466085471347e-06, + "loss": 1.0642, + "step": 43326 + }, + { + "epoch": 3.2, + "learning_rate": 9.39806793662856e-06, + "loss": 0.9998, + "step": 43327 + }, + { + "epoch": 3.2, + "learning_rate": 9.397669788743435e-06, + "loss": 0.9648, + "step": 43328 + }, + { + "epoch": 3.2, + "learning_rate": 9.39727164181661e-06, + "loss": 1.1131, + "step": 43329 + }, + { + "epoch": 3.2, + "learning_rate": 9.396873495848716e-06, + "loss": 1.1002, + "step": 43330 + }, + { + "epoch": 3.2, + "learning_rate": 9.396475350840388e-06, + "loss": 0.9336, + "step": 43331 + }, + { + "epoch": 3.2, + "learning_rate": 9.39607720679226e-06, + "loss": 0.9221, + "step": 43332 + }, + { + "epoch": 3.2, + "learning_rate": 9.395679063704963e-06, + "loss": 1.0636, + "step": 43333 + }, + { + "epoch": 3.2, + "learning_rate": 9.395280921579133e-06, + "loss": 0.9458, + "step": 43334 + }, + { + "epoch": 3.2, + "learning_rate": 9.394882780415401e-06, + "loss": 0.9462, + "step": 43335 + }, + { + "epoch": 3.2, + "learning_rate": 9.394484640214404e-06, + "loss": 0.9515, + "step": 43336 + }, + { + "epoch": 3.2, + "learning_rate": 9.394086500976771e-06, + "loss": 1.022, + "step": 43337 + }, + { + "epoch": 3.2, + "learning_rate": 9.39368836270314e-06, + "loss": 0.9475, + "step": 43338 + }, + { + "epoch": 3.2, + "learning_rate": 9.39329022539414e-06, + "loss": 1.0468, + "step": 43339 + }, + { + "epoch": 3.2, + "learning_rate": 9.392892089050408e-06, + "loss": 0.9001, + "step": 43340 + }, + { + "epoch": 3.2, + "learning_rate": 9.392493953672579e-06, + "loss": 0.9644, + "step": 43341 + }, + { + "epoch": 3.2, + "learning_rate": 9.392095819261277e-06, + "loss": 0.8933, + "step": 43342 + }, + { + "epoch": 3.2, + "learning_rate": 9.391697685817148e-06, + "loss": 1.0943, + "step": 43343 + }, + { + "epoch": 3.2, + "learning_rate": 9.391299553340812e-06, + "loss": 1.031, + "step": 43344 + }, + { + "epoch": 3.2, + "learning_rate": 9.390901421832917e-06, + "loss": 0.9532, + "step": 43345 + }, + { + "epoch": 3.2, + "learning_rate": 9.390503291294084e-06, + "loss": 0.9946, + "step": 43346 + }, + { + "epoch": 3.2, + "learning_rate": 9.390105161724957e-06, + "loss": 1.0282, + "step": 43347 + }, + { + "epoch": 3.2, + "learning_rate": 9.389707033126156e-06, + "loss": 1.0376, + "step": 43348 + }, + { + "epoch": 3.2, + "learning_rate": 9.389308905498331e-06, + "loss": 1.0058, + "step": 43349 + }, + { + "epoch": 3.2, + "learning_rate": 9.388910778842102e-06, + "loss": 1.0172, + "step": 43350 + }, + { + "epoch": 3.2, + "learning_rate": 9.388512653158108e-06, + "loss": 1.1012, + "step": 43351 + }, + { + "epoch": 3.2, + "learning_rate": 9.388114528446982e-06, + "loss": 1.107, + "step": 43352 + }, + { + "epoch": 3.2, + "learning_rate": 9.387716404709356e-06, + "loss": 0.9666, + "step": 43353 + }, + { + "epoch": 3.2, + "learning_rate": 9.387318281945865e-06, + "loss": 1.012, + "step": 43354 + }, + { + "epoch": 3.2, + "learning_rate": 9.386920160157142e-06, + "loss": 1.0643, + "step": 43355 + }, + { + "epoch": 3.2, + "learning_rate": 9.38652203934382e-06, + "loss": 0.9396, + "step": 43356 + }, + { + "epoch": 3.2, + "learning_rate": 9.386123919506532e-06, + "loss": 1.0355, + "step": 43357 + }, + { + "epoch": 3.2, + "learning_rate": 9.385725800645914e-06, + "loss": 1.0129, + "step": 43358 + }, + { + "epoch": 3.2, + "learning_rate": 9.385327682762595e-06, + "loss": 0.9951, + "step": 43359 + }, + { + "epoch": 3.2, + "learning_rate": 9.384929565857213e-06, + "loss": 0.9852, + "step": 43360 + }, + { + "epoch": 3.2, + "learning_rate": 9.3845314499304e-06, + "loss": 0.9922, + "step": 43361 + }, + { + "epoch": 3.2, + "learning_rate": 9.384133334982785e-06, + "loss": 0.9773, + "step": 43362 + }, + { + "epoch": 3.2, + "learning_rate": 9.383735221015008e-06, + "loss": 0.9933, + "step": 43363 + }, + { + "epoch": 3.2, + "learning_rate": 9.383337108027697e-06, + "loss": 1.0413, + "step": 43364 + }, + { + "epoch": 3.2, + "learning_rate": 9.382938996021491e-06, + "loss": 0.9752, + "step": 43365 + }, + { + "epoch": 3.2, + "learning_rate": 9.382540884997016e-06, + "loss": 1.0852, + "step": 43366 + }, + { + "epoch": 3.2, + "learning_rate": 9.382142774954915e-06, + "loss": 0.9289, + "step": 43367 + }, + { + "epoch": 3.2, + "learning_rate": 9.381744665895811e-06, + "loss": 1.0298, + "step": 43368 + }, + { + "epoch": 3.2, + "learning_rate": 9.381346557820346e-06, + "loss": 1.0005, + "step": 43369 + }, + { + "epoch": 3.2, + "learning_rate": 9.380948450729152e-06, + "loss": 0.9257, + "step": 43370 + }, + { + "epoch": 3.2, + "learning_rate": 9.380550344622854e-06, + "loss": 1.1237, + "step": 43371 + }, + { + "epoch": 3.2, + "learning_rate": 9.380152239502095e-06, + "loss": 0.9827, + "step": 43372 + }, + { + "epoch": 3.2, + "learning_rate": 9.379754135367504e-06, + "loss": 1.0216, + "step": 43373 + }, + { + "epoch": 3.2, + "learning_rate": 9.379356032219717e-06, + "loss": 1.054, + "step": 43374 + }, + { + "epoch": 3.2, + "learning_rate": 9.378957930059363e-06, + "loss": 1.0969, + "step": 43375 + }, + { + "epoch": 3.21, + "learning_rate": 9.378559828887079e-06, + "loss": 0.964, + "step": 43376 + }, + { + "epoch": 3.21, + "learning_rate": 9.378161728703496e-06, + "loss": 1.0797, + "step": 43377 + }, + { + "epoch": 3.21, + "learning_rate": 9.377763629509251e-06, + "loss": 1.035, + "step": 43378 + }, + { + "epoch": 3.21, + "learning_rate": 9.377365531304976e-06, + "loss": 0.9693, + "step": 43379 + }, + { + "epoch": 3.21, + "learning_rate": 9.3769674340913e-06, + "loss": 0.8755, + "step": 43380 + }, + { + "epoch": 3.21, + "learning_rate": 9.376569337868864e-06, + "loss": 0.8328, + "step": 43381 + }, + { + "epoch": 3.21, + "learning_rate": 9.376171242638294e-06, + "loss": 1.0067, + "step": 43382 + }, + { + "epoch": 3.21, + "learning_rate": 9.375773148400229e-06, + "loss": 1.0151, + "step": 43383 + }, + { + "epoch": 3.21, + "learning_rate": 9.375375055155297e-06, + "loss": 0.9888, + "step": 43384 + }, + { + "epoch": 3.21, + "learning_rate": 9.374976962904136e-06, + "loss": 0.9907, + "step": 43385 + }, + { + "epoch": 3.21, + "learning_rate": 9.374578871647377e-06, + "loss": 0.9694, + "step": 43386 + }, + { + "epoch": 3.21, + "learning_rate": 9.374180781385655e-06, + "loss": 0.9789, + "step": 43387 + }, + { + "epoch": 3.21, + "learning_rate": 9.373782692119606e-06, + "loss": 1.0798, + "step": 43388 + }, + { + "epoch": 3.21, + "learning_rate": 9.373384603849853e-06, + "loss": 0.9996, + "step": 43389 + }, + { + "epoch": 3.21, + "learning_rate": 9.372986516577043e-06, + "loss": 0.9687, + "step": 43390 + }, + { + "epoch": 3.21, + "learning_rate": 9.372588430301795e-06, + "loss": 0.9225, + "step": 43391 + }, + { + "epoch": 3.21, + "learning_rate": 9.372190345024757e-06, + "loss": 0.964, + "step": 43392 + }, + { + "epoch": 3.21, + "learning_rate": 9.371792260746547e-06, + "loss": 0.9829, + "step": 43393 + }, + { + "epoch": 3.21, + "learning_rate": 9.371394177467816e-06, + "loss": 1.0966, + "step": 43394 + }, + { + "epoch": 3.21, + "learning_rate": 9.37099609518918e-06, + "loss": 1.0381, + "step": 43395 + }, + { + "epoch": 3.21, + "learning_rate": 9.370598013911285e-06, + "loss": 0.9983, + "step": 43396 + }, + { + "epoch": 3.21, + "learning_rate": 9.370199933634759e-06, + "loss": 0.9598, + "step": 43397 + }, + { + "epoch": 3.21, + "learning_rate": 9.369801854360234e-06, + "loss": 0.9285, + "step": 43398 + }, + { + "epoch": 3.21, + "learning_rate": 9.369403776088346e-06, + "loss": 1.0058, + "step": 43399 + }, + { + "epoch": 3.21, + "learning_rate": 9.369005698819727e-06, + "loss": 1.0695, + "step": 43400 + }, + { + "epoch": 3.21, + "learning_rate": 9.368607622555013e-06, + "loss": 0.9984, + "step": 43401 + }, + { + "epoch": 3.21, + "learning_rate": 9.368209547294832e-06, + "loss": 1.0089, + "step": 43402 + }, + { + "epoch": 3.21, + "learning_rate": 9.367811473039822e-06, + "loss": 0.8826, + "step": 43403 + }, + { + "epoch": 3.21, + "learning_rate": 9.367413399790615e-06, + "loss": 1.0438, + "step": 43404 + }, + { + "epoch": 3.21, + "learning_rate": 9.367015327547845e-06, + "loss": 1.0459, + "step": 43405 + }, + { + "epoch": 3.21, + "learning_rate": 9.366617256312142e-06, + "loss": 1.0908, + "step": 43406 + }, + { + "epoch": 3.21, + "learning_rate": 9.366219186084145e-06, + "loss": 1.0029, + "step": 43407 + }, + { + "epoch": 3.21, + "learning_rate": 9.365821116864484e-06, + "loss": 1.0069, + "step": 43408 + }, + { + "epoch": 3.21, + "learning_rate": 9.36542304865379e-06, + "loss": 1.0077, + "step": 43409 + }, + { + "epoch": 3.21, + "learning_rate": 9.365024981452702e-06, + "loss": 1.0403, + "step": 43410 + }, + { + "epoch": 3.21, + "learning_rate": 9.364626915261847e-06, + "loss": 1.0542, + "step": 43411 + }, + { + "epoch": 3.21, + "learning_rate": 9.364228850081865e-06, + "loss": 0.9561, + "step": 43412 + }, + { + "epoch": 3.21, + "learning_rate": 9.363830785913383e-06, + "loss": 1.0382, + "step": 43413 + }, + { + "epoch": 3.21, + "learning_rate": 9.363432722757039e-06, + "loss": 1.0309, + "step": 43414 + }, + { + "epoch": 3.21, + "learning_rate": 9.363034660613462e-06, + "loss": 1.0096, + "step": 43415 + }, + { + "epoch": 3.21, + "learning_rate": 9.36263659948329e-06, + "loss": 0.9149, + "step": 43416 + }, + { + "epoch": 3.21, + "learning_rate": 9.362238539367156e-06, + "loss": 1.0041, + "step": 43417 + }, + { + "epoch": 3.21, + "learning_rate": 9.361840480265687e-06, + "loss": 1.0044, + "step": 43418 + }, + { + "epoch": 3.21, + "learning_rate": 9.361442422179523e-06, + "loss": 0.9932, + "step": 43419 + }, + { + "epoch": 3.21, + "learning_rate": 9.361044365109292e-06, + "loss": 1.0249, + "step": 43420 + }, + { + "epoch": 3.21, + "learning_rate": 9.360646309055635e-06, + "loss": 1.0058, + "step": 43421 + }, + { + "epoch": 3.21, + "learning_rate": 9.360248254019176e-06, + "loss": 1.035, + "step": 43422 + }, + { + "epoch": 3.21, + "learning_rate": 9.359850200000557e-06, + "loss": 0.9786, + "step": 43423 + }, + { + "epoch": 3.21, + "learning_rate": 9.359452147000404e-06, + "loss": 0.9689, + "step": 43424 + }, + { + "epoch": 3.21, + "learning_rate": 9.359054095019355e-06, + "loss": 1.0672, + "step": 43425 + }, + { + "epoch": 3.21, + "learning_rate": 9.358656044058041e-06, + "loss": 1.0126, + "step": 43426 + }, + { + "epoch": 3.21, + "learning_rate": 9.358257994117097e-06, + "loss": 1.0476, + "step": 43427 + }, + { + "epoch": 3.21, + "learning_rate": 9.357859945197156e-06, + "loss": 1.0177, + "step": 43428 + }, + { + "epoch": 3.21, + "learning_rate": 9.357461897298847e-06, + "loss": 1.0625, + "step": 43429 + }, + { + "epoch": 3.21, + "learning_rate": 9.357063850422811e-06, + "loss": 0.9702, + "step": 43430 + }, + { + "epoch": 3.21, + "learning_rate": 9.356665804569673e-06, + "loss": 0.9137, + "step": 43431 + }, + { + "epoch": 3.21, + "learning_rate": 9.356267759740075e-06, + "loss": 0.965, + "step": 43432 + }, + { + "epoch": 3.21, + "learning_rate": 9.355869715934642e-06, + "loss": 1.0143, + "step": 43433 + }, + { + "epoch": 3.21, + "learning_rate": 9.355471673154015e-06, + "loss": 1.0082, + "step": 43434 + }, + { + "epoch": 3.21, + "learning_rate": 9.355073631398824e-06, + "loss": 1.0964, + "step": 43435 + }, + { + "epoch": 3.21, + "learning_rate": 9.354675590669695e-06, + "loss": 0.9764, + "step": 43436 + }, + { + "epoch": 3.21, + "learning_rate": 9.354277550967275e-06, + "loss": 0.8719, + "step": 43437 + }, + { + "epoch": 3.21, + "learning_rate": 9.353879512292184e-06, + "loss": 1.0258, + "step": 43438 + }, + { + "epoch": 3.21, + "learning_rate": 9.353481474645067e-06, + "loss": 0.9562, + "step": 43439 + }, + { + "epoch": 3.21, + "learning_rate": 9.353083438026545e-06, + "loss": 0.9418, + "step": 43440 + }, + { + "epoch": 3.21, + "learning_rate": 9.352685402437266e-06, + "loss": 1.05, + "step": 43441 + }, + { + "epoch": 3.21, + "learning_rate": 9.352287367877849e-06, + "loss": 0.985, + "step": 43442 + }, + { + "epoch": 3.21, + "learning_rate": 9.351889334348938e-06, + "loss": 0.9911, + "step": 43443 + }, + { + "epoch": 3.21, + "learning_rate": 9.35149130185116e-06, + "loss": 1.1377, + "step": 43444 + }, + { + "epoch": 3.21, + "learning_rate": 9.35109327038515e-06, + "loss": 0.9513, + "step": 43445 + }, + { + "epoch": 3.21, + "learning_rate": 9.350695239951542e-06, + "loss": 0.9782, + "step": 43446 + }, + { + "epoch": 3.21, + "learning_rate": 9.350297210550965e-06, + "loss": 0.946, + "step": 43447 + }, + { + "epoch": 3.21, + "learning_rate": 9.34989918218406e-06, + "loss": 1.1606, + "step": 43448 + }, + { + "epoch": 3.21, + "learning_rate": 9.349501154851453e-06, + "loss": 0.9244, + "step": 43449 + }, + { + "epoch": 3.21, + "learning_rate": 9.349103128553784e-06, + "loss": 0.9461, + "step": 43450 + }, + { + "epoch": 3.21, + "learning_rate": 9.348705103291678e-06, + "loss": 1.0362, + "step": 43451 + }, + { + "epoch": 3.21, + "learning_rate": 9.348307079065776e-06, + "loss": 0.974, + "step": 43452 + }, + { + "epoch": 3.21, + "learning_rate": 9.347909055876709e-06, + "loss": 1.0753, + "step": 43453 + }, + { + "epoch": 3.21, + "learning_rate": 9.347511033725108e-06, + "loss": 0.9723, + "step": 43454 + }, + { + "epoch": 3.21, + "learning_rate": 9.347113012611609e-06, + "loss": 0.9954, + "step": 43455 + }, + { + "epoch": 3.21, + "learning_rate": 9.346714992536841e-06, + "loss": 0.8499, + "step": 43456 + }, + { + "epoch": 3.21, + "learning_rate": 9.346316973501442e-06, + "loss": 1.0208, + "step": 43457 + }, + { + "epoch": 3.21, + "learning_rate": 9.345918955506042e-06, + "loss": 1.0781, + "step": 43458 + }, + { + "epoch": 3.21, + "learning_rate": 9.345520938551276e-06, + "loss": 1.0498, + "step": 43459 + }, + { + "epoch": 3.21, + "learning_rate": 9.345122922637776e-06, + "loss": 0.9191, + "step": 43460 + }, + { + "epoch": 3.21, + "learning_rate": 9.34472490776618e-06, + "loss": 0.8799, + "step": 43461 + }, + { + "epoch": 3.21, + "learning_rate": 9.344326893937113e-06, + "loss": 0.8913, + "step": 43462 + }, + { + "epoch": 3.21, + "learning_rate": 9.343928881151216e-06, + "loss": 0.9743, + "step": 43463 + }, + { + "epoch": 3.21, + "learning_rate": 9.343530869409118e-06, + "loss": 0.9745, + "step": 43464 + }, + { + "epoch": 3.21, + "learning_rate": 9.343132858711451e-06, + "loss": 1.014, + "step": 43465 + }, + { + "epoch": 3.21, + "learning_rate": 9.342734849058853e-06, + "loss": 1.024, + "step": 43466 + }, + { + "epoch": 3.21, + "learning_rate": 9.34233684045195e-06, + "loss": 1.1343, + "step": 43467 + }, + { + "epoch": 3.21, + "learning_rate": 9.341938832891383e-06, + "loss": 0.883, + "step": 43468 + }, + { + "epoch": 3.21, + "learning_rate": 9.34154082637778e-06, + "loss": 0.9916, + "step": 43469 + }, + { + "epoch": 3.21, + "learning_rate": 9.341142820911779e-06, + "loss": 0.8864, + "step": 43470 + }, + { + "epoch": 3.21, + "learning_rate": 9.340744816494007e-06, + "loss": 1.0943, + "step": 43471 + }, + { + "epoch": 3.21, + "learning_rate": 9.340346813125102e-06, + "loss": 1.0073, + "step": 43472 + }, + { + "epoch": 3.21, + "learning_rate": 9.339948810805699e-06, + "loss": 1.0122, + "step": 43473 + }, + { + "epoch": 3.21, + "learning_rate": 9.339550809536423e-06, + "loss": 0.897, + "step": 43474 + }, + { + "epoch": 3.21, + "learning_rate": 9.339152809317915e-06, + "loss": 1.0951, + "step": 43475 + }, + { + "epoch": 3.21, + "learning_rate": 9.338754810150804e-06, + "loss": 0.9354, + "step": 43476 + }, + { + "epoch": 3.21, + "learning_rate": 9.338356812035724e-06, + "loss": 0.9807, + "step": 43477 + }, + { + "epoch": 3.21, + "learning_rate": 9.33795881497331e-06, + "loss": 0.9485, + "step": 43478 + }, + { + "epoch": 3.21, + "learning_rate": 9.337560818964196e-06, + "loss": 1.0021, + "step": 43479 + }, + { + "epoch": 3.21, + "learning_rate": 9.33716282400901e-06, + "loss": 0.9913, + "step": 43480 + }, + { + "epoch": 3.21, + "learning_rate": 9.33676483010839e-06, + "loss": 0.9983, + "step": 43481 + }, + { + "epoch": 3.21, + "learning_rate": 9.33636683726297e-06, + "loss": 0.9927, + "step": 43482 + }, + { + "epoch": 3.21, + "learning_rate": 9.335968845473376e-06, + "loss": 1.0687, + "step": 43483 + }, + { + "epoch": 3.21, + "learning_rate": 9.335570854740251e-06, + "loss": 0.8937, + "step": 43484 + }, + { + "epoch": 3.21, + "learning_rate": 9.335172865064217e-06, + "loss": 1.2043, + "step": 43485 + }, + { + "epoch": 3.21, + "learning_rate": 9.33477487644592e-06, + "loss": 1.022, + "step": 43486 + }, + { + "epoch": 3.21, + "learning_rate": 9.334376888885981e-06, + "loss": 1.1323, + "step": 43487 + }, + { + "epoch": 3.21, + "learning_rate": 9.333978902385045e-06, + "loss": 0.9327, + "step": 43488 + }, + { + "epoch": 3.21, + "learning_rate": 9.333580916943732e-06, + "loss": 0.9389, + "step": 43489 + }, + { + "epoch": 3.21, + "learning_rate": 9.333182932562688e-06, + "loss": 1.0412, + "step": 43490 + }, + { + "epoch": 3.21, + "learning_rate": 9.33278494924254e-06, + "loss": 0.9821, + "step": 43491 + }, + { + "epoch": 3.21, + "learning_rate": 9.33238696698392e-06, + "loss": 0.985, + "step": 43492 + }, + { + "epoch": 3.21, + "learning_rate": 9.331988985787463e-06, + "loss": 0.9972, + "step": 43493 + }, + { + "epoch": 3.21, + "learning_rate": 9.3315910056538e-06, + "loss": 0.9608, + "step": 43494 + }, + { + "epoch": 3.21, + "learning_rate": 9.331193026583569e-06, + "loss": 1.0302, + "step": 43495 + }, + { + "epoch": 3.21, + "learning_rate": 9.330795048577397e-06, + "loss": 0.9954, + "step": 43496 + }, + { + "epoch": 3.21, + "learning_rate": 9.330397071635924e-06, + "loss": 1.0282, + "step": 43497 + }, + { + "epoch": 3.21, + "learning_rate": 9.329999095759778e-06, + "loss": 1.0305, + "step": 43498 + }, + { + "epoch": 3.21, + "learning_rate": 9.329601120949595e-06, + "loss": 1.0721, + "step": 43499 + }, + { + "epoch": 3.21, + "learning_rate": 9.329203147206007e-06, + "loss": 1.106, + "step": 43500 + }, + { + "epoch": 3.21, + "learning_rate": 9.328805174529644e-06, + "loss": 0.9519, + "step": 43501 + }, + { + "epoch": 3.21, + "learning_rate": 9.328407202921147e-06, + "loss": 0.9638, + "step": 43502 + }, + { + "epoch": 3.21, + "learning_rate": 9.328009232381142e-06, + "loss": 0.9964, + "step": 43503 + }, + { + "epoch": 3.21, + "learning_rate": 9.327611262910264e-06, + "loss": 0.9229, + "step": 43504 + }, + { + "epoch": 3.21, + "learning_rate": 9.327213294509148e-06, + "loss": 0.9396, + "step": 43505 + }, + { + "epoch": 3.21, + "learning_rate": 9.326815327178429e-06, + "loss": 0.9802, + "step": 43506 + }, + { + "epoch": 3.21, + "learning_rate": 9.32641736091873e-06, + "loss": 1.0717, + "step": 43507 + }, + { + "epoch": 3.21, + "learning_rate": 9.326019395730698e-06, + "loss": 1.1284, + "step": 43508 + }, + { + "epoch": 3.21, + "learning_rate": 9.325621431614959e-06, + "loss": 0.9453, + "step": 43509 + }, + { + "epoch": 3.21, + "learning_rate": 9.325223468572143e-06, + "loss": 0.9605, + "step": 43510 + }, + { + "epoch": 3.21, + "learning_rate": 9.324825506602888e-06, + "loss": 0.9818, + "step": 43511 + }, + { + "epoch": 3.22, + "learning_rate": 9.324427545707826e-06, + "loss": 0.9941, + "step": 43512 + }, + { + "epoch": 3.22, + "learning_rate": 9.324029585887592e-06, + "loss": 0.9666, + "step": 43513 + }, + { + "epoch": 3.22, + "learning_rate": 9.323631627142814e-06, + "loss": 0.9697, + "step": 43514 + }, + { + "epoch": 3.22, + "learning_rate": 9.323233669474131e-06, + "loss": 1.1532, + "step": 43515 + }, + { + "epoch": 3.22, + "learning_rate": 9.322835712882172e-06, + "loss": 0.9978, + "step": 43516 + }, + { + "epoch": 3.22, + "learning_rate": 9.322437757367573e-06, + "loss": 1.1224, + "step": 43517 + }, + { + "epoch": 3.22, + "learning_rate": 9.322039802930964e-06, + "loss": 0.9911, + "step": 43518 + }, + { + "epoch": 3.22, + "learning_rate": 9.321641849572982e-06, + "loss": 1.0735, + "step": 43519 + }, + { + "epoch": 3.22, + "learning_rate": 9.321243897294259e-06, + "loss": 0.9588, + "step": 43520 + }, + { + "epoch": 3.22, + "learning_rate": 9.320845946095424e-06, + "loss": 1.1022, + "step": 43521 + }, + { + "epoch": 3.22, + "learning_rate": 9.320447995977115e-06, + "loss": 0.972, + "step": 43522 + }, + { + "epoch": 3.22, + "learning_rate": 9.320050046939963e-06, + "loss": 0.9682, + "step": 43523 + }, + { + "epoch": 3.22, + "learning_rate": 9.319652098984603e-06, + "loss": 1.0334, + "step": 43524 + }, + { + "epoch": 3.22, + "learning_rate": 9.319254152111664e-06, + "loss": 0.9982, + "step": 43525 + }, + { + "epoch": 3.22, + "learning_rate": 9.318856206321784e-06, + "loss": 1.0727, + "step": 43526 + }, + { + "epoch": 3.22, + "learning_rate": 9.318458261615593e-06, + "loss": 0.8995, + "step": 43527 + }, + { + "epoch": 3.22, + "learning_rate": 9.318060317993726e-06, + "loss": 0.9936, + "step": 43528 + }, + { + "epoch": 3.22, + "learning_rate": 9.31766237545682e-06, + "loss": 1.0511, + "step": 43529 + }, + { + "epoch": 3.22, + "learning_rate": 9.317264434005496e-06, + "loss": 0.9933, + "step": 43530 + }, + { + "epoch": 3.22, + "learning_rate": 9.3168664936404e-06, + "loss": 1.0048, + "step": 43531 + }, + { + "epoch": 3.22, + "learning_rate": 9.316468554362153e-06, + "loss": 0.9394, + "step": 43532 + }, + { + "epoch": 3.22, + "learning_rate": 9.3160706161714e-06, + "loss": 0.9556, + "step": 43533 + }, + { + "epoch": 3.22, + "learning_rate": 9.315672679068765e-06, + "loss": 1.0635, + "step": 43534 + }, + { + "epoch": 3.22, + "learning_rate": 9.31527474305489e-06, + "loss": 0.938, + "step": 43535 + }, + { + "epoch": 3.22, + "learning_rate": 9.314876808130398e-06, + "loss": 0.9599, + "step": 43536 + }, + { + "epoch": 3.22, + "learning_rate": 9.314478874295931e-06, + "loss": 1.0689, + "step": 43537 + }, + { + "epoch": 3.22, + "learning_rate": 9.314080941552119e-06, + "loss": 0.9951, + "step": 43538 + }, + { + "epoch": 3.22, + "learning_rate": 9.313683009899591e-06, + "loss": 1.0721, + "step": 43539 + }, + { + "epoch": 3.22, + "learning_rate": 9.313285079338986e-06, + "loss": 1.0459, + "step": 43540 + }, + { + "epoch": 3.22, + "learning_rate": 9.312887149870931e-06, + "loss": 0.9824, + "step": 43541 + }, + { + "epoch": 3.22, + "learning_rate": 9.312489221496067e-06, + "loss": 0.9963, + "step": 43542 + }, + { + "epoch": 3.22, + "learning_rate": 9.31209129421502e-06, + "loss": 1.1267, + "step": 43543 + }, + { + "epoch": 3.22, + "learning_rate": 9.311693368028428e-06, + "loss": 0.9842, + "step": 43544 + }, + { + "epoch": 3.22, + "learning_rate": 9.31129544293692e-06, + "loss": 0.9956, + "step": 43545 + }, + { + "epoch": 3.22, + "learning_rate": 9.310897518941133e-06, + "loss": 0.8216, + "step": 43546 + }, + { + "epoch": 3.22, + "learning_rate": 9.310499596041698e-06, + "loss": 1.1447, + "step": 43547 + }, + { + "epoch": 3.22, + "learning_rate": 9.310101674239246e-06, + "loss": 0.9811, + "step": 43548 + }, + { + "epoch": 3.22, + "learning_rate": 9.309703753534413e-06, + "loss": 0.9721, + "step": 43549 + }, + { + "epoch": 3.22, + "learning_rate": 9.309305833927833e-06, + "loss": 1.0684, + "step": 43550 + }, + { + "epoch": 3.22, + "learning_rate": 9.308907915420138e-06, + "loss": 1.0372, + "step": 43551 + }, + { + "epoch": 3.22, + "learning_rate": 9.308509998011958e-06, + "loss": 0.9621, + "step": 43552 + }, + { + "epoch": 3.22, + "learning_rate": 9.308112081703933e-06, + "loss": 1.0187, + "step": 43553 + }, + { + "epoch": 3.22, + "learning_rate": 9.307714166496685e-06, + "loss": 0.9495, + "step": 43554 + }, + { + "epoch": 3.22, + "learning_rate": 9.307316252390862e-06, + "loss": 1.1277, + "step": 43555 + }, + { + "epoch": 3.22, + "learning_rate": 9.306918339387084e-06, + "loss": 1.051, + "step": 43556 + }, + { + "epoch": 3.22, + "learning_rate": 9.30652042748599e-06, + "loss": 0.9505, + "step": 43557 + }, + { + "epoch": 3.22, + "learning_rate": 9.306122516688211e-06, + "loss": 1.0577, + "step": 43558 + }, + { + "epoch": 3.22, + "learning_rate": 9.305724606994381e-06, + "loss": 0.9609, + "step": 43559 + }, + { + "epoch": 3.22, + "learning_rate": 9.305326698405135e-06, + "loss": 1.0283, + "step": 43560 + }, + { + "epoch": 3.22, + "learning_rate": 9.304928790921104e-06, + "loss": 1.0985, + "step": 43561 + }, + { + "epoch": 3.22, + "learning_rate": 9.304530884542921e-06, + "loss": 1.0225, + "step": 43562 + }, + { + "epoch": 3.22, + "learning_rate": 9.304132979271218e-06, + "loss": 1.0849, + "step": 43563 + }, + { + "epoch": 3.22, + "learning_rate": 9.303735075106631e-06, + "loss": 0.8842, + "step": 43564 + }, + { + "epoch": 3.22, + "learning_rate": 9.303337172049792e-06, + "loss": 1.0008, + "step": 43565 + }, + { + "epoch": 3.22, + "learning_rate": 9.302939270101333e-06, + "loss": 1.0711, + "step": 43566 + }, + { + "epoch": 3.22, + "learning_rate": 9.302541369261887e-06, + "loss": 0.9447, + "step": 43567 + }, + { + "epoch": 3.22, + "learning_rate": 9.302143469532087e-06, + "loss": 0.9991, + "step": 43568 + }, + { + "epoch": 3.22, + "learning_rate": 9.301745570912569e-06, + "loss": 1.1884, + "step": 43569 + }, + { + "epoch": 3.22, + "learning_rate": 9.301347673403962e-06, + "loss": 0.9846, + "step": 43570 + }, + { + "epoch": 3.22, + "learning_rate": 9.3009497770069e-06, + "loss": 0.9402, + "step": 43571 + }, + { + "epoch": 3.22, + "learning_rate": 9.300551881722019e-06, + "loss": 0.9813, + "step": 43572 + }, + { + "epoch": 3.22, + "learning_rate": 9.300153987549951e-06, + "loss": 1.0775, + "step": 43573 + }, + { + "epoch": 3.22, + "learning_rate": 9.299756094491324e-06, + "loss": 0.9961, + "step": 43574 + }, + { + "epoch": 3.22, + "learning_rate": 9.299358202546778e-06, + "loss": 0.9035, + "step": 43575 + }, + { + "epoch": 3.22, + "learning_rate": 9.298960311716946e-06, + "loss": 0.9739, + "step": 43576 + }, + { + "epoch": 3.22, + "learning_rate": 9.298562422002452e-06, + "loss": 1.0284, + "step": 43577 + }, + { + "epoch": 3.22, + "learning_rate": 9.29816453340394e-06, + "loss": 0.9813, + "step": 43578 + }, + { + "epoch": 3.22, + "learning_rate": 9.297766645922033e-06, + "loss": 0.9472, + "step": 43579 + }, + { + "epoch": 3.22, + "learning_rate": 9.297368759557376e-06, + "loss": 1.0927, + "step": 43580 + }, + { + "epoch": 3.22, + "learning_rate": 9.296970874310587e-06, + "loss": 0.99, + "step": 43581 + }, + { + "epoch": 3.22, + "learning_rate": 9.296572990182315e-06, + "loss": 0.9762, + "step": 43582 + }, + { + "epoch": 3.22, + "learning_rate": 9.29617510717318e-06, + "loss": 0.9988, + "step": 43583 + }, + { + "epoch": 3.22, + "learning_rate": 9.295777225283824e-06, + "loss": 0.9823, + "step": 43584 + }, + { + "epoch": 3.22, + "learning_rate": 9.295379344514876e-06, + "loss": 0.941, + "step": 43585 + }, + { + "epoch": 3.22, + "learning_rate": 9.294981464866967e-06, + "loss": 0.9983, + "step": 43586 + }, + { + "epoch": 3.22, + "learning_rate": 9.294583586340734e-06, + "loss": 0.952, + "step": 43587 + }, + { + "epoch": 3.22, + "learning_rate": 9.294185708936808e-06, + "loss": 1.1243, + "step": 43588 + }, + { + "epoch": 3.22, + "learning_rate": 9.293787832655823e-06, + "loss": 1.0062, + "step": 43589 + }, + { + "epoch": 3.22, + "learning_rate": 9.293389957498408e-06, + "loss": 0.9911, + "step": 43590 + }, + { + "epoch": 3.22, + "learning_rate": 9.292992083465204e-06, + "loss": 1.0124, + "step": 43591 + }, + { + "epoch": 3.22, + "learning_rate": 9.292594210556837e-06, + "loss": 0.9838, + "step": 43592 + }, + { + "epoch": 3.22, + "learning_rate": 9.292196338773944e-06, + "loss": 1.0084, + "step": 43593 + }, + { + "epoch": 3.22, + "learning_rate": 9.291798468117159e-06, + "loss": 0.9838, + "step": 43594 + }, + { + "epoch": 3.22, + "learning_rate": 9.291400598587105e-06, + "loss": 0.8488, + "step": 43595 + }, + { + "epoch": 3.22, + "learning_rate": 9.29100273018443e-06, + "loss": 1.1188, + "step": 43596 + }, + { + "epoch": 3.22, + "learning_rate": 9.290604862909753e-06, + "loss": 1.0196, + "step": 43597 + }, + { + "epoch": 3.22, + "learning_rate": 9.29020699676372e-06, + "loss": 0.914, + "step": 43598 + }, + { + "epoch": 3.22, + "learning_rate": 9.28980913174695e-06, + "loss": 0.9574, + "step": 43599 + }, + { + "epoch": 3.22, + "learning_rate": 9.289411267860091e-06, + "loss": 0.8689, + "step": 43600 + }, + { + "epoch": 3.22, + "learning_rate": 9.289013405103762e-06, + "loss": 1.0446, + "step": 43601 + }, + { + "epoch": 3.22, + "learning_rate": 9.28861554347861e-06, + "loss": 1.0545, + "step": 43602 + }, + { + "epoch": 3.22, + "learning_rate": 9.288217682985256e-06, + "loss": 0.9894, + "step": 43603 + }, + { + "epoch": 3.22, + "learning_rate": 9.287819823624338e-06, + "loss": 1.0677, + "step": 43604 + }, + { + "epoch": 3.22, + "learning_rate": 9.287421965396489e-06, + "loss": 1.1191, + "step": 43605 + }, + { + "epoch": 3.22, + "learning_rate": 9.287024108302339e-06, + "loss": 0.9199, + "step": 43606 + }, + { + "epoch": 3.22, + "learning_rate": 9.286626252342526e-06, + "loss": 0.9141, + "step": 43607 + }, + { + "epoch": 3.22, + "learning_rate": 9.286228397517677e-06, + "loss": 1.0626, + "step": 43608 + }, + { + "epoch": 3.22, + "learning_rate": 9.285830543828434e-06, + "loss": 0.9522, + "step": 43609 + }, + { + "epoch": 3.22, + "learning_rate": 9.28543269127542e-06, + "loss": 0.9219, + "step": 43610 + }, + { + "epoch": 3.22, + "learning_rate": 9.285034839859274e-06, + "loss": 1.0225, + "step": 43611 + }, + { + "epoch": 3.22, + "learning_rate": 9.28463698958063e-06, + "loss": 0.873, + "step": 43612 + }, + { + "epoch": 3.22, + "learning_rate": 9.284239140440112e-06, + "loss": 0.9363, + "step": 43613 + }, + { + "epoch": 3.22, + "learning_rate": 9.283841292438365e-06, + "loss": 0.8987, + "step": 43614 + }, + { + "epoch": 3.22, + "learning_rate": 9.283443445576011e-06, + "loss": 0.9823, + "step": 43615 + }, + { + "epoch": 3.22, + "learning_rate": 9.283045599853692e-06, + "loss": 1.0551, + "step": 43616 + }, + { + "epoch": 3.22, + "learning_rate": 9.282647755272037e-06, + "loss": 0.8864, + "step": 43617 + }, + { + "epoch": 3.22, + "learning_rate": 9.282249911831679e-06, + "loss": 0.9619, + "step": 43618 + }, + { + "epoch": 3.22, + "learning_rate": 9.281852069533248e-06, + "loss": 0.8981, + "step": 43619 + }, + { + "epoch": 3.22, + "learning_rate": 9.281454228377384e-06, + "loss": 0.9959, + "step": 43620 + }, + { + "epoch": 3.22, + "learning_rate": 9.281056388364714e-06, + "loss": 0.981, + "step": 43621 + }, + { + "epoch": 3.22, + "learning_rate": 9.280658549495874e-06, + "loss": 0.9911, + "step": 43622 + }, + { + "epoch": 3.22, + "learning_rate": 9.2802607117715e-06, + "loss": 0.955, + "step": 43623 + }, + { + "epoch": 3.22, + "learning_rate": 9.279862875192212e-06, + "loss": 1.0691, + "step": 43624 + }, + { + "epoch": 3.22, + "learning_rate": 9.27946503975866e-06, + "loss": 1.1309, + "step": 43625 + }, + { + "epoch": 3.22, + "learning_rate": 9.279067205471462e-06, + "loss": 1.033, + "step": 43626 + }, + { + "epoch": 3.22, + "learning_rate": 9.278669372331265e-06, + "loss": 1.0588, + "step": 43627 + }, + { + "epoch": 3.22, + "learning_rate": 9.27827154033869e-06, + "loss": 1.0474, + "step": 43628 + }, + { + "epoch": 3.22, + "learning_rate": 9.277873709494376e-06, + "loss": 0.9106, + "step": 43629 + }, + { + "epoch": 3.22, + "learning_rate": 9.277475879798952e-06, + "loss": 0.9367, + "step": 43630 + }, + { + "epoch": 3.22, + "learning_rate": 9.277078051253057e-06, + "loss": 1.0328, + "step": 43631 + }, + { + "epoch": 3.22, + "learning_rate": 9.276680223857319e-06, + "loss": 1.0667, + "step": 43632 + }, + { + "epoch": 3.22, + "learning_rate": 9.276282397612372e-06, + "loss": 1.1377, + "step": 43633 + }, + { + "epoch": 3.22, + "learning_rate": 9.27588457251885e-06, + "loss": 1.0115, + "step": 43634 + }, + { + "epoch": 3.22, + "learning_rate": 9.275486748577384e-06, + "loss": 1.0108, + "step": 43635 + }, + { + "epoch": 3.22, + "learning_rate": 9.275088925788611e-06, + "loss": 0.9659, + "step": 43636 + }, + { + "epoch": 3.22, + "learning_rate": 9.274691104153158e-06, + "loss": 1.0147, + "step": 43637 + }, + { + "epoch": 3.22, + "learning_rate": 9.274293283671664e-06, + "loss": 0.9884, + "step": 43638 + }, + { + "epoch": 3.22, + "learning_rate": 9.273895464344755e-06, + "loss": 1.0495, + "step": 43639 + }, + { + "epoch": 3.22, + "learning_rate": 9.273497646173072e-06, + "loss": 0.9765, + "step": 43640 + }, + { + "epoch": 3.22, + "learning_rate": 9.273099829157245e-06, + "loss": 1.0044, + "step": 43641 + }, + { + "epoch": 3.22, + "learning_rate": 9.272702013297901e-06, + "loss": 1.0352, + "step": 43642 + }, + { + "epoch": 3.22, + "learning_rate": 9.272304198595682e-06, + "loss": 0.9649, + "step": 43643 + }, + { + "epoch": 3.22, + "learning_rate": 9.271906385051211e-06, + "loss": 1.0105, + "step": 43644 + }, + { + "epoch": 3.22, + "learning_rate": 9.271508572665133e-06, + "loss": 0.951, + "step": 43645 + }, + { + "epoch": 3.22, + "learning_rate": 9.271110761438068e-06, + "loss": 0.9698, + "step": 43646 + }, + { + "epoch": 3.23, + "learning_rate": 9.270712951370663e-06, + "loss": 1.1069, + "step": 43647 + }, + { + "epoch": 3.23, + "learning_rate": 9.270315142463535e-06, + "loss": 0.9306, + "step": 43648 + }, + { + "epoch": 3.23, + "learning_rate": 9.269917334717331e-06, + "loss": 0.999, + "step": 43649 + }, + { + "epoch": 3.23, + "learning_rate": 9.269519528132678e-06, + "loss": 0.9786, + "step": 43650 + }, + { + "epoch": 3.23, + "learning_rate": 9.269121722710206e-06, + "loss": 0.9599, + "step": 43651 + }, + { + "epoch": 3.23, + "learning_rate": 9.268723918450553e-06, + "loss": 0.9642, + "step": 43652 + }, + { + "epoch": 3.23, + "learning_rate": 9.268326115354348e-06, + "loss": 1.057, + "step": 43653 + }, + { + "epoch": 3.23, + "learning_rate": 9.267928313422226e-06, + "loss": 0.8946, + "step": 43654 + }, + { + "epoch": 3.23, + "learning_rate": 9.26753051265482e-06, + "loss": 1.0251, + "step": 43655 + }, + { + "epoch": 3.23, + "learning_rate": 9.267132713052763e-06, + "loss": 0.9658, + "step": 43656 + }, + { + "epoch": 3.23, + "learning_rate": 9.266734914616687e-06, + "loss": 0.9646, + "step": 43657 + }, + { + "epoch": 3.23, + "learning_rate": 9.266337117347225e-06, + "loss": 1.0698, + "step": 43658 + }, + { + "epoch": 3.23, + "learning_rate": 9.265939321245011e-06, + "loss": 0.9136, + "step": 43659 + }, + { + "epoch": 3.23, + "learning_rate": 9.265541526310674e-06, + "loss": 0.9027, + "step": 43660 + }, + { + "epoch": 3.23, + "learning_rate": 9.265143732544855e-06, + "loss": 1.1019, + "step": 43661 + }, + { + "epoch": 3.23, + "learning_rate": 9.264745939948178e-06, + "loss": 0.8736, + "step": 43662 + }, + { + "epoch": 3.23, + "learning_rate": 9.264348148521281e-06, + "loss": 1.1035, + "step": 43663 + }, + { + "epoch": 3.23, + "learning_rate": 9.263950358264795e-06, + "loss": 1.0062, + "step": 43664 + }, + { + "epoch": 3.23, + "learning_rate": 9.263552569179353e-06, + "loss": 1.1122, + "step": 43665 + }, + { + "epoch": 3.23, + "learning_rate": 9.263154781265588e-06, + "loss": 1.0221, + "step": 43666 + }, + { + "epoch": 3.23, + "learning_rate": 9.262756994524136e-06, + "loss": 1.0353, + "step": 43667 + }, + { + "epoch": 3.23, + "learning_rate": 9.26235920895563e-06, + "loss": 1.0687, + "step": 43668 + }, + { + "epoch": 3.23, + "learning_rate": 9.261961424560691e-06, + "loss": 1.0402, + "step": 43669 + }, + { + "epoch": 3.23, + "learning_rate": 9.261563641339968e-06, + "loss": 1.059, + "step": 43670 + }, + { + "epoch": 3.23, + "learning_rate": 9.261165859294083e-06, + "loss": 1.04, + "step": 43671 + }, + { + "epoch": 3.23, + "learning_rate": 9.260768078423674e-06, + "loss": 1.0413, + "step": 43672 + }, + { + "epoch": 3.23, + "learning_rate": 9.26037029872937e-06, + "loss": 0.8554, + "step": 43673 + }, + { + "epoch": 3.23, + "learning_rate": 9.25997252021181e-06, + "loss": 0.9674, + "step": 43674 + }, + { + "epoch": 3.23, + "learning_rate": 9.25957474287162e-06, + "loss": 1.0553, + "step": 43675 + }, + { + "epoch": 3.23, + "learning_rate": 9.259176966709438e-06, + "loss": 0.8628, + "step": 43676 + }, + { + "epoch": 3.23, + "learning_rate": 9.258779191725893e-06, + "loss": 0.8995, + "step": 43677 + }, + { + "epoch": 3.23, + "learning_rate": 9.258381417921622e-06, + "loss": 0.8857, + "step": 43678 + }, + { + "epoch": 3.23, + "learning_rate": 9.257983645297255e-06, + "loss": 1.0089, + "step": 43679 + }, + { + "epoch": 3.23, + "learning_rate": 9.257585873853424e-06, + "loss": 0.9835, + "step": 43680 + }, + { + "epoch": 3.23, + "learning_rate": 9.257188103590765e-06, + "loss": 1.0471, + "step": 43681 + }, + { + "epoch": 3.23, + "learning_rate": 9.256790334509906e-06, + "loss": 0.957, + "step": 43682 + }, + { + "epoch": 3.23, + "learning_rate": 9.256392566611487e-06, + "loss": 1.1136, + "step": 43683 + }, + { + "epoch": 3.23, + "learning_rate": 9.255994799896132e-06, + "loss": 1.149, + "step": 43684 + }, + { + "epoch": 3.23, + "learning_rate": 9.255597034364483e-06, + "loss": 0.9878, + "step": 43685 + }, + { + "epoch": 3.23, + "learning_rate": 9.255199270017166e-06, + "loss": 0.992, + "step": 43686 + }, + { + "epoch": 3.23, + "learning_rate": 9.254801506854818e-06, + "loss": 1.0173, + "step": 43687 + }, + { + "epoch": 3.23, + "learning_rate": 9.254403744878072e-06, + "loss": 0.9635, + "step": 43688 + }, + { + "epoch": 3.23, + "learning_rate": 9.254005984087552e-06, + "loss": 1.0037, + "step": 43689 + }, + { + "epoch": 3.23, + "learning_rate": 9.253608224483907e-06, + "loss": 1.0286, + "step": 43690 + }, + { + "epoch": 3.23, + "learning_rate": 9.253210466067752e-06, + "loss": 1.042, + "step": 43691 + }, + { + "epoch": 3.23, + "learning_rate": 9.252812708839735e-06, + "loss": 0.9695, + "step": 43692 + }, + { + "epoch": 3.23, + "learning_rate": 9.252414952800476e-06, + "loss": 1.0721, + "step": 43693 + }, + { + "epoch": 3.23, + "learning_rate": 9.252017197950621e-06, + "loss": 0.9204, + "step": 43694 + }, + { + "epoch": 3.23, + "learning_rate": 9.251619444290788e-06, + "loss": 0.9501, + "step": 43695 + }, + { + "epoch": 3.23, + "learning_rate": 9.251221691821625e-06, + "loss": 0.9329, + "step": 43696 + }, + { + "epoch": 3.23, + "learning_rate": 9.250823940543757e-06, + "loss": 0.9678, + "step": 43697 + }, + { + "epoch": 3.23, + "learning_rate": 9.250426190457814e-06, + "loss": 1.0348, + "step": 43698 + }, + { + "epoch": 3.23, + "learning_rate": 9.250028441564434e-06, + "loss": 0.9688, + "step": 43699 + }, + { + "epoch": 3.23, + "learning_rate": 9.249630693864246e-06, + "loss": 1.075, + "step": 43700 + }, + { + "epoch": 3.23, + "learning_rate": 9.249232947357888e-06, + "loss": 1.0756, + "step": 43701 + }, + { + "epoch": 3.23, + "learning_rate": 9.248835202045987e-06, + "loss": 1.0023, + "step": 43702 + }, + { + "epoch": 3.23, + "learning_rate": 9.24843745792918e-06, + "loss": 1.1001, + "step": 43703 + }, + { + "epoch": 3.23, + "learning_rate": 9.248039715008097e-06, + "loss": 0.8964, + "step": 43704 + }, + { + "epoch": 3.23, + "learning_rate": 9.247641973283374e-06, + "loss": 1.0114, + "step": 43705 + }, + { + "epoch": 3.23, + "learning_rate": 9.24724423275564e-06, + "loss": 1.021, + "step": 43706 + }, + { + "epoch": 3.23, + "learning_rate": 9.24684649342553e-06, + "loss": 0.9842, + "step": 43707 + }, + { + "epoch": 3.23, + "learning_rate": 9.246448755293678e-06, + "loss": 1.0251, + "step": 43708 + }, + { + "epoch": 3.23, + "learning_rate": 9.246051018360712e-06, + "loss": 0.9507, + "step": 43709 + }, + { + "epoch": 3.23, + "learning_rate": 9.245653282627271e-06, + "loss": 0.9718, + "step": 43710 + }, + { + "epoch": 3.23, + "learning_rate": 9.245255548093983e-06, + "loss": 0.9549, + "step": 43711 + }, + { + "epoch": 3.23, + "learning_rate": 9.244857814761484e-06, + "loss": 0.9606, + "step": 43712 + }, + { + "epoch": 3.23, + "learning_rate": 9.244460082630405e-06, + "loss": 1.0156, + "step": 43713 + }, + { + "epoch": 3.23, + "learning_rate": 9.24406235170138e-06, + "loss": 0.9476, + "step": 43714 + }, + { + "epoch": 3.23, + "learning_rate": 9.243664621975042e-06, + "loss": 0.9213, + "step": 43715 + }, + { + "epoch": 3.23, + "learning_rate": 9.24326689345202e-06, + "loss": 1.0591, + "step": 43716 + }, + { + "epoch": 3.23, + "learning_rate": 9.24286916613295e-06, + "loss": 1.0009, + "step": 43717 + }, + { + "epoch": 3.23, + "learning_rate": 9.242471440018465e-06, + "loss": 1.0019, + "step": 43718 + }, + { + "epoch": 3.23, + "learning_rate": 9.242073715109197e-06, + "loss": 1.087, + "step": 43719 + }, + { + "epoch": 3.23, + "learning_rate": 9.241675991405778e-06, + "loss": 0.9457, + "step": 43720 + }, + { + "epoch": 3.23, + "learning_rate": 9.241278268908843e-06, + "loss": 0.9406, + "step": 43721 + }, + { + "epoch": 3.23, + "learning_rate": 9.240880547619022e-06, + "loss": 1.0507, + "step": 43722 + }, + { + "epoch": 3.23, + "learning_rate": 9.240482827536949e-06, + "loss": 1.0697, + "step": 43723 + }, + { + "epoch": 3.23, + "learning_rate": 9.24008510866326e-06, + "loss": 0.978, + "step": 43724 + }, + { + "epoch": 3.23, + "learning_rate": 9.239687390998582e-06, + "loss": 1.1526, + "step": 43725 + }, + { + "epoch": 3.23, + "learning_rate": 9.23928967454355e-06, + "loss": 1.0125, + "step": 43726 + }, + { + "epoch": 3.23, + "learning_rate": 9.238891959298798e-06, + "loss": 1.1354, + "step": 43727 + }, + { + "epoch": 3.23, + "learning_rate": 9.23849424526496e-06, + "loss": 0.9806, + "step": 43728 + }, + { + "epoch": 3.23, + "learning_rate": 9.238096532442662e-06, + "loss": 1.0744, + "step": 43729 + }, + { + "epoch": 3.23, + "learning_rate": 9.237698820832547e-06, + "loss": 1.01, + "step": 43730 + }, + { + "epoch": 3.23, + "learning_rate": 9.237301110435239e-06, + "loss": 0.9749, + "step": 43731 + }, + { + "epoch": 3.23, + "learning_rate": 9.236903401251375e-06, + "loss": 0.9194, + "step": 43732 + }, + { + "epoch": 3.23, + "learning_rate": 9.236505693281586e-06, + "loss": 0.9517, + "step": 43733 + }, + { + "epoch": 3.23, + "learning_rate": 9.236107986526508e-06, + "loss": 0.9949, + "step": 43734 + }, + { + "epoch": 3.23, + "learning_rate": 9.235710280986771e-06, + "loss": 1.0364, + "step": 43735 + }, + { + "epoch": 3.23, + "learning_rate": 9.235312576663004e-06, + "loss": 1.0736, + "step": 43736 + }, + { + "epoch": 3.23, + "learning_rate": 9.23491487355585e-06, + "loss": 0.9552, + "step": 43737 + }, + { + "epoch": 3.23, + "learning_rate": 9.234517171665927e-06, + "loss": 1.026, + "step": 43738 + }, + { + "epoch": 3.23, + "learning_rate": 9.234119470993884e-06, + "loss": 1.0807, + "step": 43739 + }, + { + "epoch": 3.23, + "learning_rate": 9.23372177154034e-06, + "loss": 1.0838, + "step": 43740 + }, + { + "epoch": 3.23, + "learning_rate": 9.233324073305939e-06, + "loss": 0.9156, + "step": 43741 + }, + { + "epoch": 3.23, + "learning_rate": 9.232926376291302e-06, + "loss": 1.0406, + "step": 43742 + }, + { + "epoch": 3.23, + "learning_rate": 9.232528680497076e-06, + "loss": 0.9883, + "step": 43743 + }, + { + "epoch": 3.23, + "learning_rate": 9.232130985923884e-06, + "loss": 1.0251, + "step": 43744 + }, + { + "epoch": 3.23, + "learning_rate": 9.231733292572356e-06, + "loss": 0.9626, + "step": 43745 + }, + { + "epoch": 3.23, + "learning_rate": 9.231335600443133e-06, + "loss": 1.0641, + "step": 43746 + }, + { + "epoch": 3.23, + "learning_rate": 9.230937909536843e-06, + "loss": 0.9427, + "step": 43747 + }, + { + "epoch": 3.23, + "learning_rate": 9.23054021985412e-06, + "loss": 1.0145, + "step": 43748 + }, + { + "epoch": 3.23, + "learning_rate": 9.230142531395594e-06, + "loss": 0.8781, + "step": 43749 + }, + { + "epoch": 3.23, + "learning_rate": 9.229744844161905e-06, + "loss": 0.988, + "step": 43750 + }, + { + "epoch": 3.23, + "learning_rate": 9.229347158153677e-06, + "loss": 1.0219, + "step": 43751 + }, + { + "epoch": 3.23, + "learning_rate": 9.228949473371549e-06, + "loss": 1.0472, + "step": 43752 + }, + { + "epoch": 3.23, + "learning_rate": 9.228551789816152e-06, + "loss": 1.1109, + "step": 43753 + }, + { + "epoch": 3.23, + "learning_rate": 9.228154107488113e-06, + "loss": 1.0912, + "step": 43754 + }, + { + "epoch": 3.23, + "learning_rate": 9.227756426388076e-06, + "loss": 1.0622, + "step": 43755 + }, + { + "epoch": 3.23, + "learning_rate": 9.227358746516663e-06, + "loss": 1.082, + "step": 43756 + }, + { + "epoch": 3.23, + "learning_rate": 9.226961067874515e-06, + "loss": 1.0041, + "step": 43757 + }, + { + "epoch": 3.23, + "learning_rate": 9.226563390462258e-06, + "loss": 1.0392, + "step": 43758 + }, + { + "epoch": 3.23, + "learning_rate": 9.226165714280529e-06, + "loss": 1.0881, + "step": 43759 + }, + { + "epoch": 3.23, + "learning_rate": 9.225768039329957e-06, + "loss": 1.0619, + "step": 43760 + }, + { + "epoch": 3.23, + "learning_rate": 9.225370365611181e-06, + "loss": 1.039, + "step": 43761 + }, + { + "epoch": 3.23, + "learning_rate": 9.22497269312483e-06, + "loss": 1.0644, + "step": 43762 + }, + { + "epoch": 3.23, + "learning_rate": 9.22457502187153e-06, + "loss": 1.0527, + "step": 43763 + }, + { + "epoch": 3.23, + "learning_rate": 9.224177351851926e-06, + "loss": 1.0142, + "step": 43764 + }, + { + "epoch": 3.23, + "learning_rate": 9.223779683066641e-06, + "loss": 1.0434, + "step": 43765 + }, + { + "epoch": 3.23, + "learning_rate": 9.223382015516315e-06, + "loss": 1.0097, + "step": 43766 + }, + { + "epoch": 3.23, + "learning_rate": 9.222984349201573e-06, + "loss": 0.9514, + "step": 43767 + }, + { + "epoch": 3.23, + "learning_rate": 9.222586684123055e-06, + "loss": 0.9175, + "step": 43768 + }, + { + "epoch": 3.23, + "learning_rate": 9.222189020281389e-06, + "loss": 1.0107, + "step": 43769 + }, + { + "epoch": 3.23, + "learning_rate": 9.221791357677208e-06, + "loss": 0.9283, + "step": 43770 + }, + { + "epoch": 3.23, + "learning_rate": 9.22139369631115e-06, + "loss": 0.9941, + "step": 43771 + }, + { + "epoch": 3.23, + "learning_rate": 9.220996036183839e-06, + "loss": 0.8865, + "step": 43772 + }, + { + "epoch": 3.23, + "learning_rate": 9.220598377295916e-06, + "loss": 1.0382, + "step": 43773 + }, + { + "epoch": 3.23, + "learning_rate": 9.220200719648005e-06, + "loss": 0.9797, + "step": 43774 + }, + { + "epoch": 3.23, + "learning_rate": 9.219803063240748e-06, + "loss": 0.9899, + "step": 43775 + }, + { + "epoch": 3.23, + "learning_rate": 9.21940540807477e-06, + "loss": 0.9481, + "step": 43776 + }, + { + "epoch": 3.23, + "learning_rate": 9.219007754150708e-06, + "loss": 1.0976, + "step": 43777 + }, + { + "epoch": 3.23, + "learning_rate": 9.218610101469193e-06, + "loss": 0.9825, + "step": 43778 + }, + { + "epoch": 3.23, + "learning_rate": 9.21821245003086e-06, + "loss": 0.9596, + "step": 43779 + }, + { + "epoch": 3.23, + "learning_rate": 9.217814799836343e-06, + "loss": 1.0922, + "step": 43780 + }, + { + "epoch": 3.23, + "learning_rate": 9.217417150886262e-06, + "loss": 1.0466, + "step": 43781 + }, + { + "epoch": 3.24, + "learning_rate": 9.217019503181268e-06, + "loss": 1.0939, + "step": 43782 + }, + { + "epoch": 3.24, + "learning_rate": 9.216621856721977e-06, + "loss": 1.0845, + "step": 43783 + }, + { + "epoch": 3.24, + "learning_rate": 9.216224211509037e-06, + "loss": 1.0642, + "step": 43784 + }, + { + "epoch": 3.24, + "learning_rate": 9.215826567543066e-06, + "loss": 0.9554, + "step": 43785 + }, + { + "epoch": 3.24, + "learning_rate": 9.21542892482471e-06, + "loss": 1.0809, + "step": 43786 + }, + { + "epoch": 3.24, + "learning_rate": 9.21503128335459e-06, + "loss": 1.0614, + "step": 43787 + }, + { + "epoch": 3.24, + "learning_rate": 9.214633643133348e-06, + "loss": 0.8659, + "step": 43788 + }, + { + "epoch": 3.24, + "learning_rate": 9.21423600416161e-06, + "loss": 0.9976, + "step": 43789 + }, + { + "epoch": 3.24, + "learning_rate": 9.213838366440014e-06, + "loss": 1.0947, + "step": 43790 + }, + { + "epoch": 3.24, + "learning_rate": 9.213440729969189e-06, + "loss": 1.0926, + "step": 43791 + }, + { + "epoch": 3.24, + "learning_rate": 9.213043094749767e-06, + "loss": 0.9533, + "step": 43792 + }, + { + "epoch": 3.24, + "learning_rate": 9.212645460782385e-06, + "loss": 0.9095, + "step": 43793 + }, + { + "epoch": 3.24, + "learning_rate": 9.21224782806767e-06, + "loss": 1.0618, + "step": 43794 + }, + { + "epoch": 3.24, + "learning_rate": 9.211850196606258e-06, + "loss": 1.0562, + "step": 43795 + }, + { + "epoch": 3.24, + "learning_rate": 9.21145256639878e-06, + "loss": 0.9533, + "step": 43796 + }, + { + "epoch": 3.24, + "learning_rate": 9.211054937445872e-06, + "loss": 0.9257, + "step": 43797 + }, + { + "epoch": 3.24, + "learning_rate": 9.210657309748162e-06, + "loss": 0.9769, + "step": 43798 + }, + { + "epoch": 3.24, + "learning_rate": 9.210259683306288e-06, + "loss": 0.9613, + "step": 43799 + }, + { + "epoch": 3.24, + "learning_rate": 9.209862058120879e-06, + "loss": 1.0736, + "step": 43800 + }, + { + "epoch": 3.24, + "learning_rate": 9.209464434192565e-06, + "loss": 0.9371, + "step": 43801 + }, + { + "epoch": 3.24, + "learning_rate": 9.209066811521985e-06, + "loss": 1.0206, + "step": 43802 + }, + { + "epoch": 3.24, + "learning_rate": 9.208669190109766e-06, + "loss": 0.9987, + "step": 43803 + }, + { + "epoch": 3.24, + "learning_rate": 9.208271569956548e-06, + "loss": 0.8492, + "step": 43804 + }, + { + "epoch": 3.24, + "learning_rate": 9.20787395106295e-06, + "loss": 1.038, + "step": 43805 + }, + { + "epoch": 3.24, + "learning_rate": 9.207476333429621e-06, + "loss": 1.0163, + "step": 43806 + }, + { + "epoch": 3.24, + "learning_rate": 9.20707871705718e-06, + "loss": 1.094, + "step": 43807 + }, + { + "epoch": 3.24, + "learning_rate": 9.20668110194627e-06, + "loss": 0.9687, + "step": 43808 + }, + { + "epoch": 3.24, + "learning_rate": 9.206283488097519e-06, + "loss": 1.0529, + "step": 43809 + }, + { + "epoch": 3.24, + "learning_rate": 9.205885875511557e-06, + "loss": 0.955, + "step": 43810 + }, + { + "epoch": 3.24, + "learning_rate": 9.205488264189021e-06, + "loss": 0.859, + "step": 43811 + }, + { + "epoch": 3.24, + "learning_rate": 9.20509065413054e-06, + "loss": 1.1243, + "step": 43812 + }, + { + "epoch": 3.24, + "learning_rate": 9.20469304533675e-06, + "loss": 0.9616, + "step": 43813 + }, + { + "epoch": 3.24, + "learning_rate": 9.20429543780828e-06, + "loss": 1.0655, + "step": 43814 + }, + { + "epoch": 3.24, + "learning_rate": 9.203897831545767e-06, + "loss": 1.0718, + "step": 43815 + }, + { + "epoch": 3.24, + "learning_rate": 9.203500226549837e-06, + "loss": 0.8931, + "step": 43816 + }, + { + "epoch": 3.24, + "learning_rate": 9.203102622821131e-06, + "loss": 1.0017, + "step": 43817 + }, + { + "epoch": 3.24, + "learning_rate": 9.202705020360278e-06, + "loss": 0.976, + "step": 43818 + }, + { + "epoch": 3.24, + "learning_rate": 9.202307419167906e-06, + "loss": 0.9357, + "step": 43819 + }, + { + "epoch": 3.24, + "learning_rate": 9.201909819244655e-06, + "loss": 1.0768, + "step": 43820 + }, + { + "epoch": 3.24, + "learning_rate": 9.201512220591152e-06, + "loss": 1.1365, + "step": 43821 + }, + { + "epoch": 3.24, + "learning_rate": 9.201114623208032e-06, + "loss": 1.0317, + "step": 43822 + }, + { + "epoch": 3.24, + "learning_rate": 9.200717027095926e-06, + "loss": 1.1078, + "step": 43823 + }, + { + "epoch": 3.24, + "learning_rate": 9.20031943225547e-06, + "loss": 1.0385, + "step": 43824 + }, + { + "epoch": 3.24, + "learning_rate": 9.199921838687293e-06, + "loss": 0.9661, + "step": 43825 + }, + { + "epoch": 3.24, + "learning_rate": 9.19952424639203e-06, + "loss": 0.9959, + "step": 43826 + }, + { + "epoch": 3.24, + "learning_rate": 9.199126655370314e-06, + "loss": 1.0043, + "step": 43827 + }, + { + "epoch": 3.24, + "learning_rate": 9.198729065622771e-06, + "loss": 1.0492, + "step": 43828 + }, + { + "epoch": 3.24, + "learning_rate": 9.198331477150044e-06, + "loss": 0.8838, + "step": 43829 + }, + { + "epoch": 3.24, + "learning_rate": 9.197933889952755e-06, + "loss": 1.1693, + "step": 43830 + }, + { + "epoch": 3.24, + "learning_rate": 9.197536304031546e-06, + "loss": 0.9841, + "step": 43831 + }, + { + "epoch": 3.24, + "learning_rate": 9.197138719387041e-06, + "loss": 0.9998, + "step": 43832 + }, + { + "epoch": 3.24, + "learning_rate": 9.196741136019883e-06, + "loss": 0.8744, + "step": 43833 + }, + { + "epoch": 3.24, + "learning_rate": 9.19634355393069e-06, + "loss": 0.9955, + "step": 43834 + }, + { + "epoch": 3.24, + "learning_rate": 9.195945973120111e-06, + "loss": 1.0135, + "step": 43835 + }, + { + "epoch": 3.24, + "learning_rate": 9.195548393588765e-06, + "loss": 1.1031, + "step": 43836 + }, + { + "epoch": 3.24, + "learning_rate": 9.195150815337294e-06, + "loss": 0.9602, + "step": 43837 + }, + { + "epoch": 3.24, + "learning_rate": 9.194753238366327e-06, + "loss": 1.0375, + "step": 43838 + }, + { + "epoch": 3.24, + "learning_rate": 9.194355662676492e-06, + "loss": 0.9708, + "step": 43839 + }, + { + "epoch": 3.24, + "learning_rate": 9.193958088268428e-06, + "loss": 1.1017, + "step": 43840 + }, + { + "epoch": 3.24, + "learning_rate": 9.193560515142764e-06, + "loss": 0.9155, + "step": 43841 + }, + { + "epoch": 3.24, + "learning_rate": 9.193162943300134e-06, + "loss": 1.0022, + "step": 43842 + }, + { + "epoch": 3.24, + "learning_rate": 9.192765372741169e-06, + "loss": 1.0769, + "step": 43843 + }, + { + "epoch": 3.24, + "learning_rate": 9.192367803466506e-06, + "loss": 1.1126, + "step": 43844 + }, + { + "epoch": 3.24, + "learning_rate": 9.191970235476771e-06, + "loss": 1.0866, + "step": 43845 + }, + { + "epoch": 3.24, + "learning_rate": 9.191572668772602e-06, + "loss": 1.073, + "step": 43846 + }, + { + "epoch": 3.24, + "learning_rate": 9.191175103354627e-06, + "loss": 1.0496, + "step": 43847 + }, + { + "epoch": 3.24, + "learning_rate": 9.190777539223484e-06, + "loss": 1.0937, + "step": 43848 + }, + { + "epoch": 3.24, + "learning_rate": 9.190379976379803e-06, + "loss": 1.1173, + "step": 43849 + }, + { + "epoch": 3.24, + "learning_rate": 9.18998241482421e-06, + "loss": 1.0277, + "step": 43850 + }, + { + "epoch": 3.24, + "learning_rate": 9.18958485455735e-06, + "loss": 0.9776, + "step": 43851 + }, + { + "epoch": 3.24, + "learning_rate": 9.189187295579843e-06, + "loss": 1.0405, + "step": 43852 + }, + { + "epoch": 3.24, + "learning_rate": 9.188789737892333e-06, + "loss": 0.9753, + "step": 43853 + }, + { + "epoch": 3.24, + "learning_rate": 9.18839218149544e-06, + "loss": 0.9462, + "step": 43854 + }, + { + "epoch": 3.24, + "learning_rate": 9.187994626389812e-06, + "loss": 0.9488, + "step": 43855 + }, + { + "epoch": 3.24, + "learning_rate": 9.187597072576068e-06, + "loss": 1.0509, + "step": 43856 + }, + { + "epoch": 3.24, + "learning_rate": 9.187199520054844e-06, + "loss": 1.0463, + "step": 43857 + }, + { + "epoch": 3.24, + "learning_rate": 9.186801968826779e-06, + "loss": 1.0308, + "step": 43858 + }, + { + "epoch": 3.24, + "learning_rate": 9.186404418892495e-06, + "loss": 0.951, + "step": 43859 + }, + { + "epoch": 3.24, + "learning_rate": 9.186006870252632e-06, + "loss": 1.075, + "step": 43860 + }, + { + "epoch": 3.24, + "learning_rate": 9.18560932290782e-06, + "loss": 1.0048, + "step": 43861 + }, + { + "epoch": 3.24, + "learning_rate": 9.185211776858695e-06, + "loss": 0.9886, + "step": 43862 + }, + { + "epoch": 3.24, + "learning_rate": 9.184814232105881e-06, + "loss": 0.9474, + "step": 43863 + }, + { + "epoch": 3.24, + "learning_rate": 9.18441668865002e-06, + "loss": 0.982, + "step": 43864 + }, + { + "epoch": 3.24, + "learning_rate": 9.18401914649174e-06, + "loss": 0.9971, + "step": 43865 + }, + { + "epoch": 3.24, + "learning_rate": 9.183621605631673e-06, + "loss": 0.9315, + "step": 43866 + }, + { + "epoch": 3.24, + "learning_rate": 9.183224066070453e-06, + "loss": 0.989, + "step": 43867 + }, + { + "epoch": 3.24, + "learning_rate": 9.18282652780871e-06, + "loss": 1.0103, + "step": 43868 + }, + { + "epoch": 3.24, + "learning_rate": 9.18242899084708e-06, + "loss": 0.9597, + "step": 43869 + }, + { + "epoch": 3.24, + "learning_rate": 9.182031455186193e-06, + "loss": 1.0425, + "step": 43870 + }, + { + "epoch": 3.24, + "learning_rate": 9.181633920826684e-06, + "loss": 1.0112, + "step": 43871 + }, + { + "epoch": 3.24, + "learning_rate": 9.18123638776918e-06, + "loss": 0.896, + "step": 43872 + }, + { + "epoch": 3.24, + "learning_rate": 9.180838856014321e-06, + "loss": 0.9961, + "step": 43873 + }, + { + "epoch": 3.24, + "learning_rate": 9.180441325562737e-06, + "loss": 1.0214, + "step": 43874 + }, + { + "epoch": 3.24, + "learning_rate": 9.180043796415054e-06, + "loss": 1.0067, + "step": 43875 + }, + { + "epoch": 3.24, + "learning_rate": 9.179646268571915e-06, + "loss": 0.9949, + "step": 43876 + }, + { + "epoch": 3.24, + "learning_rate": 9.17924874203394e-06, + "loss": 0.9628, + "step": 43877 + }, + { + "epoch": 3.24, + "learning_rate": 9.178851216801775e-06, + "loss": 1.0728, + "step": 43878 + }, + { + "epoch": 3.24, + "learning_rate": 9.178453692876041e-06, + "loss": 1.0242, + "step": 43879 + }, + { + "epoch": 3.24, + "learning_rate": 9.17805617025738e-06, + "loss": 0.943, + "step": 43880 + }, + { + "epoch": 3.24, + "learning_rate": 9.177658648946415e-06, + "loss": 1.0705, + "step": 43881 + }, + { + "epoch": 3.24, + "learning_rate": 9.17726112894379e-06, + "loss": 0.9701, + "step": 43882 + }, + { + "epoch": 3.24, + "learning_rate": 9.176863610250126e-06, + "loss": 1.0034, + "step": 43883 + }, + { + "epoch": 3.24, + "learning_rate": 9.176466092866061e-06, + "loss": 1.0464, + "step": 43884 + }, + { + "epoch": 3.24, + "learning_rate": 9.176068576792228e-06, + "loss": 1.0724, + "step": 43885 + }, + { + "epoch": 3.24, + "learning_rate": 9.175671062029254e-06, + "loss": 1.0154, + "step": 43886 + }, + { + "epoch": 3.24, + "learning_rate": 9.17527354857778e-06, + "loss": 0.93, + "step": 43887 + }, + { + "epoch": 3.24, + "learning_rate": 9.174876036438431e-06, + "loss": 1.0697, + "step": 43888 + }, + { + "epoch": 3.24, + "learning_rate": 9.174478525611845e-06, + "loss": 0.9052, + "step": 43889 + }, + { + "epoch": 3.24, + "learning_rate": 9.17408101609865e-06, + "loss": 0.918, + "step": 43890 + }, + { + "epoch": 3.24, + "learning_rate": 9.173683507899482e-06, + "loss": 1.0033, + "step": 43891 + }, + { + "epoch": 3.24, + "learning_rate": 9.173286001014968e-06, + "loss": 0.9443, + "step": 43892 + }, + { + "epoch": 3.24, + "learning_rate": 9.172888495445748e-06, + "loss": 0.9757, + "step": 43893 + }, + { + "epoch": 3.24, + "learning_rate": 9.172490991192453e-06, + "loss": 1.0746, + "step": 43894 + }, + { + "epoch": 3.24, + "learning_rate": 9.172093488255706e-06, + "loss": 1.0942, + "step": 43895 + }, + { + "epoch": 3.24, + "learning_rate": 9.171695986636152e-06, + "loss": 0.9602, + "step": 43896 + }, + { + "epoch": 3.24, + "learning_rate": 9.171298486334412e-06, + "loss": 1.0678, + "step": 43897 + }, + { + "epoch": 3.24, + "learning_rate": 9.170900987351132e-06, + "loss": 1.1511, + "step": 43898 + }, + { + "epoch": 3.24, + "learning_rate": 9.170503489686928e-06, + "loss": 1.1161, + "step": 43899 + }, + { + "epoch": 3.24, + "learning_rate": 9.17010599334245e-06, + "loss": 1.0129, + "step": 43900 + }, + { + "epoch": 3.24, + "learning_rate": 9.169708498318312e-06, + "loss": 1.0887, + "step": 43901 + }, + { + "epoch": 3.24, + "learning_rate": 9.169311004615166e-06, + "loss": 0.9866, + "step": 43902 + }, + { + "epoch": 3.24, + "learning_rate": 9.168913512233629e-06, + "loss": 1.0198, + "step": 43903 + }, + { + "epoch": 3.24, + "learning_rate": 9.168516021174339e-06, + "loss": 1.0584, + "step": 43904 + }, + { + "epoch": 3.24, + "learning_rate": 9.168118531437928e-06, + "loss": 0.9583, + "step": 43905 + }, + { + "epoch": 3.24, + "learning_rate": 9.167721043025028e-06, + "loss": 0.9944, + "step": 43906 + }, + { + "epoch": 3.24, + "learning_rate": 9.167323555936272e-06, + "loss": 0.9986, + "step": 43907 + }, + { + "epoch": 3.24, + "learning_rate": 9.166926070172293e-06, + "loss": 1.0054, + "step": 43908 + }, + { + "epoch": 3.24, + "learning_rate": 9.166528585733723e-06, + "loss": 0.9918, + "step": 43909 + }, + { + "epoch": 3.24, + "learning_rate": 9.166131102621193e-06, + "loss": 0.986, + "step": 43910 + }, + { + "epoch": 3.24, + "learning_rate": 9.165733620835338e-06, + "loss": 0.9602, + "step": 43911 + }, + { + "epoch": 3.24, + "learning_rate": 9.165336140376789e-06, + "loss": 1.058, + "step": 43912 + }, + { + "epoch": 3.24, + "learning_rate": 9.164938661246175e-06, + "loss": 0.9887, + "step": 43913 + }, + { + "epoch": 3.24, + "learning_rate": 9.164541183444135e-06, + "loss": 1.0181, + "step": 43914 + }, + { + "epoch": 3.24, + "learning_rate": 9.164143706971296e-06, + "loss": 1.0813, + "step": 43915 + }, + { + "epoch": 3.24, + "learning_rate": 9.163746231828294e-06, + "loss": 1.0011, + "step": 43916 + }, + { + "epoch": 3.24, + "learning_rate": 9.163348758015758e-06, + "loss": 1.1028, + "step": 43917 + }, + { + "epoch": 3.25, + "learning_rate": 9.162951285534325e-06, + "loss": 0.9748, + "step": 43918 + }, + { + "epoch": 3.25, + "learning_rate": 9.16255381438462e-06, + "loss": 1.0691, + "step": 43919 + }, + { + "epoch": 3.25, + "learning_rate": 9.162156344567285e-06, + "loss": 1.114, + "step": 43920 + }, + { + "epoch": 3.25, + "learning_rate": 9.161758876082948e-06, + "loss": 1.035, + "step": 43921 + }, + { + "epoch": 3.25, + "learning_rate": 9.161361408932233e-06, + "loss": 1.0012, + "step": 43922 + }, + { + "epoch": 3.25, + "learning_rate": 9.160963943115789e-06, + "loss": 1.0025, + "step": 43923 + }, + { + "epoch": 3.25, + "learning_rate": 9.160566478634231e-06, + "loss": 1.0643, + "step": 43924 + }, + { + "epoch": 3.25, + "learning_rate": 9.160169015488207e-06, + "loss": 0.9307, + "step": 43925 + }, + { + "epoch": 3.25, + "learning_rate": 9.159771553678337e-06, + "loss": 0.9723, + "step": 43926 + }, + { + "epoch": 3.25, + "learning_rate": 9.15937409320526e-06, + "loss": 1.0645, + "step": 43927 + }, + { + "epoch": 3.25, + "learning_rate": 9.158976634069607e-06, + "loss": 1.0701, + "step": 43928 + }, + { + "epoch": 3.25, + "learning_rate": 9.158579176272012e-06, + "loss": 0.9532, + "step": 43929 + }, + { + "epoch": 3.25, + "learning_rate": 9.158181719813104e-06, + "loss": 1.0856, + "step": 43930 + }, + { + "epoch": 3.25, + "learning_rate": 9.157784264693514e-06, + "loss": 0.9522, + "step": 43931 + }, + { + "epoch": 3.25, + "learning_rate": 9.157386810913882e-06, + "loss": 0.9917, + "step": 43932 + }, + { + "epoch": 3.25, + "learning_rate": 9.156989358474832e-06, + "loss": 1.143, + "step": 43933 + }, + { + "epoch": 3.25, + "learning_rate": 9.156591907377001e-06, + "loss": 0.9188, + "step": 43934 + }, + { + "epoch": 3.25, + "learning_rate": 9.15619445762102e-06, + "loss": 0.9937, + "step": 43935 + }, + { + "epoch": 3.25, + "learning_rate": 9.155797009207522e-06, + "loss": 1.0742, + "step": 43936 + }, + { + "epoch": 3.25, + "learning_rate": 9.155399562137137e-06, + "loss": 1.0391, + "step": 43937 + }, + { + "epoch": 3.25, + "learning_rate": 9.155002116410503e-06, + "loss": 0.9845, + "step": 43938 + }, + { + "epoch": 3.25, + "learning_rate": 9.154604672028249e-06, + "loss": 0.9583, + "step": 43939 + }, + { + "epoch": 3.25, + "learning_rate": 9.154207228991e-06, + "loss": 1.033, + "step": 43940 + }, + { + "epoch": 3.25, + "learning_rate": 9.153809787299403e-06, + "loss": 0.9917, + "step": 43941 + }, + { + "epoch": 3.25, + "learning_rate": 9.153412346954076e-06, + "loss": 1.1146, + "step": 43942 + }, + { + "epoch": 3.25, + "learning_rate": 9.153014907955663e-06, + "loss": 0.9236, + "step": 43943 + }, + { + "epoch": 3.25, + "learning_rate": 9.152617470304787e-06, + "loss": 0.995, + "step": 43944 + }, + { + "epoch": 3.25, + "learning_rate": 9.15222003400209e-06, + "loss": 0.9683, + "step": 43945 + }, + { + "epoch": 3.25, + "learning_rate": 9.151822599048192e-06, + "loss": 0.911, + "step": 43946 + }, + { + "epoch": 3.25, + "learning_rate": 9.151425165443738e-06, + "loss": 0.9817, + "step": 43947 + }, + { + "epoch": 3.25, + "learning_rate": 9.151027733189349e-06, + "loss": 0.8892, + "step": 43948 + }, + { + "epoch": 3.25, + "learning_rate": 9.150630302285668e-06, + "loss": 1.0768, + "step": 43949 + }, + { + "epoch": 3.25, + "learning_rate": 9.15023287273332e-06, + "loss": 0.9992, + "step": 43950 + }, + { + "epoch": 3.25, + "learning_rate": 9.149835444532939e-06, + "loss": 1.0623, + "step": 43951 + }, + { + "epoch": 3.25, + "learning_rate": 9.149438017685157e-06, + "loss": 1.0415, + "step": 43952 + }, + { + "epoch": 3.25, + "learning_rate": 9.149040592190606e-06, + "loss": 0.9514, + "step": 43953 + }, + { + "epoch": 3.25, + "learning_rate": 9.148643168049922e-06, + "loss": 0.9703, + "step": 43954 + }, + { + "epoch": 3.25, + "learning_rate": 9.148245745263732e-06, + "loss": 0.948, + "step": 43955 + }, + { + "epoch": 3.25, + "learning_rate": 9.147848323832673e-06, + "loss": 1.0679, + "step": 43956 + }, + { + "epoch": 3.25, + "learning_rate": 9.147450903757373e-06, + "loss": 1.0296, + "step": 43957 + }, + { + "epoch": 3.25, + "learning_rate": 9.14705348503847e-06, + "loss": 1.0213, + "step": 43958 + }, + { + "epoch": 3.25, + "learning_rate": 9.14665606767659e-06, + "loss": 0.8794, + "step": 43959 + }, + { + "epoch": 3.25, + "learning_rate": 9.146258651672368e-06, + "loss": 0.9451, + "step": 43960 + }, + { + "epoch": 3.25, + "learning_rate": 9.145861237026437e-06, + "loss": 0.8683, + "step": 43961 + }, + { + "epoch": 3.25, + "learning_rate": 9.145463823739427e-06, + "loss": 1.037, + "step": 43962 + }, + { + "epoch": 3.25, + "learning_rate": 9.145066411811974e-06, + "loss": 0.8984, + "step": 43963 + }, + { + "epoch": 3.25, + "learning_rate": 9.144669001244706e-06, + "loss": 1.0009, + "step": 43964 + }, + { + "epoch": 3.25, + "learning_rate": 9.14427159203826e-06, + "loss": 0.9867, + "step": 43965 + }, + { + "epoch": 3.25, + "learning_rate": 9.143874184193262e-06, + "loss": 1.0391, + "step": 43966 + }, + { + "epoch": 3.25, + "learning_rate": 9.143476777710352e-06, + "loss": 1.0008, + "step": 43967 + }, + { + "epoch": 3.25, + "learning_rate": 9.143079372590159e-06, + "loss": 1.0909, + "step": 43968 + }, + { + "epoch": 3.25, + "learning_rate": 9.142681968833308e-06, + "loss": 0.8823, + "step": 43969 + }, + { + "epoch": 3.25, + "learning_rate": 9.142284566440444e-06, + "loss": 1.0358, + "step": 43970 + }, + { + "epoch": 3.25, + "learning_rate": 9.14188716541219e-06, + "loss": 1.053, + "step": 43971 + }, + { + "epoch": 3.25, + "learning_rate": 9.141489765749183e-06, + "loss": 0.964, + "step": 43972 + }, + { + "epoch": 3.25, + "learning_rate": 9.14109236745205e-06, + "loss": 1.1055, + "step": 43973 + }, + { + "epoch": 3.25, + "learning_rate": 9.14069497052143e-06, + "loss": 1.033, + "step": 43974 + }, + { + "epoch": 3.25, + "learning_rate": 9.14029757495795e-06, + "loss": 1.0966, + "step": 43975 + }, + { + "epoch": 3.25, + "learning_rate": 9.139900180762247e-06, + "loss": 0.9672, + "step": 43976 + }, + { + "epoch": 3.25, + "learning_rate": 9.139502787934949e-06, + "loss": 1.0477, + "step": 43977 + }, + { + "epoch": 3.25, + "learning_rate": 9.139105396476689e-06, + "loss": 1.0492, + "step": 43978 + }, + { + "epoch": 3.25, + "learning_rate": 9.1387080063881e-06, + "loss": 0.9767, + "step": 43979 + }, + { + "epoch": 3.25, + "learning_rate": 9.138310617669814e-06, + "loss": 0.9887, + "step": 43980 + }, + { + "epoch": 3.25, + "learning_rate": 9.137913230322465e-06, + "loss": 1.0082, + "step": 43981 + }, + { + "epoch": 3.25, + "learning_rate": 9.137515844346682e-06, + "loss": 1.041, + "step": 43982 + }, + { + "epoch": 3.25, + "learning_rate": 9.1371184597431e-06, + "loss": 0.986, + "step": 43983 + }, + { + "epoch": 3.25, + "learning_rate": 9.13672107651235e-06, + "loss": 0.9186, + "step": 43984 + }, + { + "epoch": 3.25, + "learning_rate": 9.136323694655065e-06, + "loss": 1.0277, + "step": 43985 + }, + { + "epoch": 3.25, + "learning_rate": 9.135926314171879e-06, + "loss": 1.0183, + "step": 43986 + }, + { + "epoch": 3.25, + "learning_rate": 9.135528935063415e-06, + "loss": 0.9693, + "step": 43987 + }, + { + "epoch": 3.25, + "learning_rate": 9.135131557330318e-06, + "loss": 0.9815, + "step": 43988 + }, + { + "epoch": 3.25, + "learning_rate": 9.134734180973211e-06, + "loss": 0.9328, + "step": 43989 + }, + { + "epoch": 3.25, + "learning_rate": 9.134336805992734e-06, + "loss": 0.9656, + "step": 43990 + }, + { + "epoch": 3.25, + "learning_rate": 9.133939432389509e-06, + "loss": 1.0918, + "step": 43991 + }, + { + "epoch": 3.25, + "learning_rate": 9.133542060164179e-06, + "loss": 1.0367, + "step": 43992 + }, + { + "epoch": 3.25, + "learning_rate": 9.133144689317365e-06, + "loss": 1.0752, + "step": 43993 + }, + { + "epoch": 3.25, + "learning_rate": 9.132747319849712e-06, + "loss": 0.9979, + "step": 43994 + }, + { + "epoch": 3.25, + "learning_rate": 9.132349951761844e-06, + "loss": 0.9843, + "step": 43995 + }, + { + "epoch": 3.25, + "learning_rate": 9.131952585054392e-06, + "loss": 1.0441, + "step": 43996 + }, + { + "epoch": 3.25, + "learning_rate": 9.131555219727995e-06, + "loss": 1.0271, + "step": 43997 + }, + { + "epoch": 3.25, + "learning_rate": 9.131157855783278e-06, + "loss": 0.9803, + "step": 43998 + }, + { + "epoch": 3.25, + "learning_rate": 9.130760493220878e-06, + "loss": 1.0641, + "step": 43999 + }, + { + "epoch": 3.25, + "learning_rate": 9.130363132041422e-06, + "loss": 1.0229, + "step": 44000 + }, + { + "epoch": 3.25, + "learning_rate": 9.12996577224555e-06, + "loss": 1.0709, + "step": 44001 + }, + { + "epoch": 3.25, + "learning_rate": 9.129568413833886e-06, + "loss": 1.0236, + "step": 44002 + }, + { + "epoch": 3.25, + "learning_rate": 9.12917105680707e-06, + "loss": 0.9766, + "step": 44003 + }, + { + "epoch": 3.25, + "learning_rate": 9.12877370116573e-06, + "loss": 1.0512, + "step": 44004 + }, + { + "epoch": 3.25, + "learning_rate": 9.128376346910499e-06, + "loss": 0.9428, + "step": 44005 + }, + { + "epoch": 3.25, + "learning_rate": 9.127978994042008e-06, + "loss": 1.0295, + "step": 44006 + }, + { + "epoch": 3.25, + "learning_rate": 9.127581642560888e-06, + "loss": 0.9669, + "step": 44007 + }, + { + "epoch": 3.25, + "learning_rate": 9.127184292467775e-06, + "loss": 0.9147, + "step": 44008 + }, + { + "epoch": 3.25, + "learning_rate": 9.126786943763299e-06, + "loss": 1.0145, + "step": 44009 + }, + { + "epoch": 3.25, + "learning_rate": 9.126389596448093e-06, + "loss": 0.9176, + "step": 44010 + }, + { + "epoch": 3.25, + "learning_rate": 9.125992250522788e-06, + "loss": 1.0962, + "step": 44011 + }, + { + "epoch": 3.25, + "learning_rate": 9.125594905988017e-06, + "loss": 1.092, + "step": 44012 + }, + { + "epoch": 3.25, + "learning_rate": 9.125197562844412e-06, + "loss": 0.9355, + "step": 44013 + }, + { + "epoch": 3.25, + "learning_rate": 9.124800221092604e-06, + "loss": 1.1105, + "step": 44014 + }, + { + "epoch": 3.25, + "learning_rate": 9.124402880733231e-06, + "loss": 0.9048, + "step": 44015 + }, + { + "epoch": 3.25, + "learning_rate": 9.124005541766916e-06, + "loss": 1.0069, + "step": 44016 + }, + { + "epoch": 3.25, + "learning_rate": 9.123608204194297e-06, + "loss": 1.0208, + "step": 44017 + }, + { + "epoch": 3.25, + "learning_rate": 9.123210868016003e-06, + "loss": 0.9804, + "step": 44018 + }, + { + "epoch": 3.25, + "learning_rate": 9.12281353323267e-06, + "loss": 1.0567, + "step": 44019 + }, + { + "epoch": 3.25, + "learning_rate": 9.122416199844928e-06, + "loss": 1.0692, + "step": 44020 + }, + { + "epoch": 3.25, + "learning_rate": 9.12201886785341e-06, + "loss": 1.0025, + "step": 44021 + }, + { + "epoch": 3.25, + "learning_rate": 9.121621537258745e-06, + "loss": 1.0258, + "step": 44022 + }, + { + "epoch": 3.25, + "learning_rate": 9.121224208061569e-06, + "loss": 0.9537, + "step": 44023 + }, + { + "epoch": 3.25, + "learning_rate": 9.120826880262513e-06, + "loss": 1.028, + "step": 44024 + }, + { + "epoch": 3.25, + "learning_rate": 9.120429553862207e-06, + "loss": 0.9641, + "step": 44025 + }, + { + "epoch": 3.25, + "learning_rate": 9.120032228861287e-06, + "loss": 1.025, + "step": 44026 + }, + { + "epoch": 3.25, + "learning_rate": 9.119634905260382e-06, + "loss": 1.0355, + "step": 44027 + }, + { + "epoch": 3.25, + "learning_rate": 9.119237583060127e-06, + "loss": 0.9964, + "step": 44028 + }, + { + "epoch": 3.25, + "learning_rate": 9.118840262261149e-06, + "loss": 1.0238, + "step": 44029 + }, + { + "epoch": 3.25, + "learning_rate": 9.118442942864087e-06, + "loss": 1.0722, + "step": 44030 + }, + { + "epoch": 3.25, + "learning_rate": 9.118045624869568e-06, + "loss": 1.0914, + "step": 44031 + }, + { + "epoch": 3.25, + "learning_rate": 9.117648308278226e-06, + "loss": 0.9611, + "step": 44032 + }, + { + "epoch": 3.25, + "learning_rate": 9.117250993090696e-06, + "loss": 0.9337, + "step": 44033 + }, + { + "epoch": 3.25, + "learning_rate": 9.116853679307602e-06, + "loss": 1.0523, + "step": 44034 + }, + { + "epoch": 3.25, + "learning_rate": 9.116456366929585e-06, + "loss": 1.0926, + "step": 44035 + }, + { + "epoch": 3.25, + "learning_rate": 9.116059055957268e-06, + "loss": 1.0567, + "step": 44036 + }, + { + "epoch": 3.25, + "learning_rate": 9.115661746391296e-06, + "loss": 1.0355, + "step": 44037 + }, + { + "epoch": 3.25, + "learning_rate": 9.115264438232286e-06, + "loss": 1.0094, + "step": 44038 + }, + { + "epoch": 3.25, + "learning_rate": 9.114867131480884e-06, + "loss": 1.0668, + "step": 44039 + }, + { + "epoch": 3.25, + "learning_rate": 9.114469826137709e-06, + "loss": 0.9731, + "step": 44040 + }, + { + "epoch": 3.25, + "learning_rate": 9.114072522203406e-06, + "loss": 0.9885, + "step": 44041 + }, + { + "epoch": 3.25, + "learning_rate": 9.1136752196786e-06, + "loss": 0.846, + "step": 44042 + }, + { + "epoch": 3.25, + "learning_rate": 9.113277918563921e-06, + "loss": 1.0037, + "step": 44043 + }, + { + "epoch": 3.25, + "learning_rate": 9.112880618860007e-06, + "loss": 1.008, + "step": 44044 + }, + { + "epoch": 3.25, + "learning_rate": 9.112483320567484e-06, + "loss": 1.1588, + "step": 44045 + }, + { + "epoch": 3.25, + "learning_rate": 9.11208602368699e-06, + "loss": 0.9572, + "step": 44046 + }, + { + "epoch": 3.25, + "learning_rate": 9.111688728219153e-06, + "loss": 0.8985, + "step": 44047 + }, + { + "epoch": 3.25, + "learning_rate": 9.111291434164609e-06, + "loss": 1.0773, + "step": 44048 + }, + { + "epoch": 3.25, + "learning_rate": 9.110894141523986e-06, + "loss": 1.059, + "step": 44049 + }, + { + "epoch": 3.25, + "learning_rate": 9.110496850297916e-06, + "loss": 1.0032, + "step": 44050 + }, + { + "epoch": 3.25, + "learning_rate": 9.110099560487034e-06, + "loss": 1.031, + "step": 44051 + }, + { + "epoch": 3.25, + "learning_rate": 9.109702272091973e-06, + "loss": 0.9633, + "step": 44052 + }, + { + "epoch": 3.26, + "learning_rate": 9.10930498511336e-06, + "loss": 0.912, + "step": 44053 + }, + { + "epoch": 3.26, + "learning_rate": 9.10890769955183e-06, + "loss": 0.9479, + "step": 44054 + }, + { + "epoch": 3.26, + "learning_rate": 9.108510415408019e-06, + "loss": 1.0223, + "step": 44055 + }, + { + "epoch": 3.26, + "learning_rate": 9.10811313268255e-06, + "loss": 0.8791, + "step": 44056 + }, + { + "epoch": 3.26, + "learning_rate": 9.107715851376065e-06, + "loss": 1.0724, + "step": 44057 + }, + { + "epoch": 3.26, + "learning_rate": 9.107318571489188e-06, + "loss": 0.8705, + "step": 44058 + }, + { + "epoch": 3.26, + "learning_rate": 9.106921293022557e-06, + "loss": 1.0895, + "step": 44059 + }, + { + "epoch": 3.26, + "learning_rate": 9.106524015976797e-06, + "loss": 1.0039, + "step": 44060 + }, + { + "epoch": 3.26, + "learning_rate": 9.106126740352551e-06, + "loss": 0.9893, + "step": 44061 + }, + { + "epoch": 3.26, + "learning_rate": 9.105729466150441e-06, + "loss": 1.036, + "step": 44062 + }, + { + "epoch": 3.26, + "learning_rate": 9.105332193371101e-06, + "loss": 1.041, + "step": 44063 + }, + { + "epoch": 3.26, + "learning_rate": 9.104934922015168e-06, + "loss": 1.0211, + "step": 44064 + }, + { + "epoch": 3.26, + "learning_rate": 9.104537652083268e-06, + "loss": 1.0423, + "step": 44065 + }, + { + "epoch": 3.26, + "learning_rate": 9.104140383576039e-06, + "loss": 0.9332, + "step": 44066 + }, + { + "epoch": 3.26, + "learning_rate": 9.103743116494105e-06, + "loss": 1.0569, + "step": 44067 + }, + { + "epoch": 3.26, + "learning_rate": 9.103345850838105e-06, + "loss": 1.053, + "step": 44068 + }, + { + "epoch": 3.26, + "learning_rate": 9.10294858660867e-06, + "loss": 0.9785, + "step": 44069 + }, + { + "epoch": 3.26, + "learning_rate": 9.102551323806431e-06, + "loss": 1.0162, + "step": 44070 + }, + { + "epoch": 3.26, + "learning_rate": 9.10215406243202e-06, + "loss": 0.9879, + "step": 44071 + }, + { + "epoch": 3.26, + "learning_rate": 9.101756802486068e-06, + "loss": 1.0678, + "step": 44072 + }, + { + "epoch": 3.26, + "learning_rate": 9.101359543969207e-06, + "loss": 1.0492, + "step": 44073 + }, + { + "epoch": 3.26, + "learning_rate": 9.100962286882071e-06, + "loss": 0.9125, + "step": 44074 + }, + { + "epoch": 3.26, + "learning_rate": 9.100565031225293e-06, + "loss": 1.0426, + "step": 44075 + }, + { + "epoch": 3.26, + "learning_rate": 9.100167776999502e-06, + "loss": 1.1634, + "step": 44076 + }, + { + "epoch": 3.26, + "learning_rate": 9.099770524205332e-06, + "loss": 1.1102, + "step": 44077 + }, + { + "epoch": 3.26, + "learning_rate": 9.099373272843411e-06, + "loss": 0.9405, + "step": 44078 + }, + { + "epoch": 3.26, + "learning_rate": 9.098976022914377e-06, + "loss": 1.0165, + "step": 44079 + }, + { + "epoch": 3.26, + "learning_rate": 9.098578774418862e-06, + "loss": 0.9835, + "step": 44080 + }, + { + "epoch": 3.26, + "learning_rate": 9.098181527357489e-06, + "loss": 1.0479, + "step": 44081 + }, + { + "epoch": 3.26, + "learning_rate": 9.097784281730903e-06, + "loss": 1.067, + "step": 44082 + }, + { + "epoch": 3.26, + "learning_rate": 9.097387037539722e-06, + "loss": 0.941, + "step": 44083 + }, + { + "epoch": 3.26, + "learning_rate": 9.096989794784592e-06, + "loss": 1.0335, + "step": 44084 + }, + { + "epoch": 3.26, + "learning_rate": 9.096592553466132e-06, + "loss": 1.0855, + "step": 44085 + }, + { + "epoch": 3.26, + "learning_rate": 9.096195313584986e-06, + "loss": 0.9994, + "step": 44086 + }, + { + "epoch": 3.26, + "learning_rate": 9.095798075141775e-06, + "loss": 0.9549, + "step": 44087 + }, + { + "epoch": 3.26, + "learning_rate": 9.095400838137142e-06, + "loss": 0.9848, + "step": 44088 + }, + { + "epoch": 3.26, + "learning_rate": 9.095003602571712e-06, + "loss": 1.124, + "step": 44089 + }, + { + "epoch": 3.26, + "learning_rate": 9.094606368446116e-06, + "loss": 1.0619, + "step": 44090 + }, + { + "epoch": 3.26, + "learning_rate": 9.09420913576099e-06, + "loss": 0.9126, + "step": 44091 + }, + { + "epoch": 3.26, + "learning_rate": 9.093811904516961e-06, + "loss": 0.9034, + "step": 44092 + }, + { + "epoch": 3.26, + "learning_rate": 9.093414674714669e-06, + "loss": 0.9908, + "step": 44093 + }, + { + "epoch": 3.26, + "learning_rate": 9.093017446354737e-06, + "loss": 1.0236, + "step": 44094 + }, + { + "epoch": 3.26, + "learning_rate": 9.092620219437805e-06, + "loss": 0.9841, + "step": 44095 + }, + { + "epoch": 3.26, + "learning_rate": 9.092222993964498e-06, + "loss": 0.8915, + "step": 44096 + }, + { + "epoch": 3.26, + "learning_rate": 9.091825769935454e-06, + "loss": 1.0592, + "step": 44097 + }, + { + "epoch": 3.26, + "learning_rate": 9.091428547351302e-06, + "loss": 1.0333, + "step": 44098 + }, + { + "epoch": 3.26, + "learning_rate": 9.09103132621267e-06, + "loss": 1.0084, + "step": 44099 + }, + { + "epoch": 3.26, + "learning_rate": 9.090634106520199e-06, + "loss": 0.9594, + "step": 44100 + }, + { + "epoch": 3.26, + "learning_rate": 9.090236888274514e-06, + "loss": 0.9238, + "step": 44101 + }, + { + "epoch": 3.26, + "learning_rate": 9.08983967147625e-06, + "loss": 1.1314, + "step": 44102 + }, + { + "epoch": 3.26, + "learning_rate": 9.089442456126037e-06, + "loss": 0.8826, + "step": 44103 + }, + { + "epoch": 3.26, + "learning_rate": 9.089045242224513e-06, + "loss": 1.059, + "step": 44104 + }, + { + "epoch": 3.26, + "learning_rate": 9.088648029772296e-06, + "loss": 0.9078, + "step": 44105 + }, + { + "epoch": 3.26, + "learning_rate": 9.088250818770035e-06, + "loss": 0.9411, + "step": 44106 + }, + { + "epoch": 3.26, + "learning_rate": 9.087853609218347e-06, + "loss": 1.0799, + "step": 44107 + }, + { + "epoch": 3.26, + "learning_rate": 9.087456401117879e-06, + "loss": 0.9902, + "step": 44108 + }, + { + "epoch": 3.26, + "learning_rate": 9.08705919446925e-06, + "loss": 1.0023, + "step": 44109 + }, + { + "epoch": 3.26, + "learning_rate": 9.086661989273095e-06, + "loss": 0.9865, + "step": 44110 + }, + { + "epoch": 3.26, + "learning_rate": 9.08626478553005e-06, + "loss": 0.9304, + "step": 44111 + }, + { + "epoch": 3.26, + "learning_rate": 9.085867583240743e-06, + "loss": 1.0126, + "step": 44112 + }, + { + "epoch": 3.26, + "learning_rate": 9.085470382405812e-06, + "loss": 1.0277, + "step": 44113 + }, + { + "epoch": 3.26, + "learning_rate": 9.085073183025878e-06, + "loss": 1.0218, + "step": 44114 + }, + { + "epoch": 3.26, + "learning_rate": 9.084675985101586e-06, + "loss": 1.0345, + "step": 44115 + }, + { + "epoch": 3.26, + "learning_rate": 9.084278788633556e-06, + "loss": 1.001, + "step": 44116 + }, + { + "epoch": 3.26, + "learning_rate": 9.083881593622429e-06, + "loss": 0.9219, + "step": 44117 + }, + { + "epoch": 3.26, + "learning_rate": 9.083484400068833e-06, + "loss": 1.0314, + "step": 44118 + }, + { + "epoch": 3.26, + "learning_rate": 9.083087207973398e-06, + "loss": 0.9858, + "step": 44119 + }, + { + "epoch": 3.26, + "learning_rate": 9.08269001733676e-06, + "loss": 1.0787, + "step": 44120 + }, + { + "epoch": 3.26, + "learning_rate": 9.082292828159546e-06, + "loss": 0.9314, + "step": 44121 + }, + { + "epoch": 3.26, + "learning_rate": 9.081895640442396e-06, + "loss": 0.9943, + "step": 44122 + }, + { + "epoch": 3.26, + "learning_rate": 9.081498454185931e-06, + "loss": 1.0221, + "step": 44123 + }, + { + "epoch": 3.26, + "learning_rate": 9.081101269390793e-06, + "loss": 0.9676, + "step": 44124 + }, + { + "epoch": 3.26, + "learning_rate": 9.080704086057607e-06, + "loss": 1.0581, + "step": 44125 + }, + { + "epoch": 3.26, + "learning_rate": 9.080306904187012e-06, + "loss": 1.0397, + "step": 44126 + }, + { + "epoch": 3.26, + "learning_rate": 9.079909723779637e-06, + "loss": 0.9157, + "step": 44127 + }, + { + "epoch": 3.26, + "learning_rate": 9.079512544836104e-06, + "loss": 1.0396, + "step": 44128 + }, + { + "epoch": 3.26, + "learning_rate": 9.079115367357061e-06, + "loss": 1.0533, + "step": 44129 + }, + { + "epoch": 3.26, + "learning_rate": 9.078718191343126e-06, + "loss": 1.0673, + "step": 44130 + }, + { + "epoch": 3.26, + "learning_rate": 9.078321016794944e-06, + "loss": 0.9446, + "step": 44131 + }, + { + "epoch": 3.26, + "learning_rate": 9.077923843713132e-06, + "loss": 0.9494, + "step": 44132 + }, + { + "epoch": 3.26, + "learning_rate": 9.077526672098336e-06, + "loss": 0.9123, + "step": 44133 + }, + { + "epoch": 3.26, + "learning_rate": 9.077129501951178e-06, + "loss": 1.0436, + "step": 44134 + }, + { + "epoch": 3.26, + "learning_rate": 9.0767323332723e-06, + "loss": 0.9387, + "step": 44135 + }, + { + "epoch": 3.26, + "learning_rate": 9.076335166062323e-06, + "loss": 0.9937, + "step": 44136 + }, + { + "epoch": 3.26, + "learning_rate": 9.075938000321884e-06, + "loss": 0.9895, + "step": 44137 + }, + { + "epoch": 3.26, + "learning_rate": 9.075540836051615e-06, + "loss": 0.966, + "step": 44138 + }, + { + "epoch": 3.26, + "learning_rate": 9.075143673252146e-06, + "loss": 1.0349, + "step": 44139 + }, + { + "epoch": 3.26, + "learning_rate": 9.074746511924112e-06, + "loss": 0.9641, + "step": 44140 + }, + { + "epoch": 3.26, + "learning_rate": 9.07434935206814e-06, + "loss": 1.0257, + "step": 44141 + }, + { + "epoch": 3.26, + "learning_rate": 9.073952193684868e-06, + "loss": 1.006, + "step": 44142 + }, + { + "epoch": 3.26, + "learning_rate": 9.073555036774922e-06, + "loss": 0.9809, + "step": 44143 + }, + { + "epoch": 3.26, + "learning_rate": 9.07315788133894e-06, + "loss": 0.9562, + "step": 44144 + }, + { + "epoch": 3.26, + "learning_rate": 9.072760727377549e-06, + "loss": 0.9918, + "step": 44145 + }, + { + "epoch": 3.26, + "learning_rate": 9.072363574891384e-06, + "loss": 0.9663, + "step": 44146 + }, + { + "epoch": 3.26, + "learning_rate": 9.071966423881076e-06, + "loss": 1.0164, + "step": 44147 + }, + { + "epoch": 3.26, + "learning_rate": 9.07156927434725e-06, + "loss": 1.0698, + "step": 44148 + }, + { + "epoch": 3.26, + "learning_rate": 9.071172126290551e-06, + "loss": 0.9677, + "step": 44149 + }, + { + "epoch": 3.26, + "learning_rate": 9.070774979711597e-06, + "loss": 0.9146, + "step": 44150 + }, + { + "epoch": 3.26, + "learning_rate": 9.070377834611034e-06, + "loss": 0.8622, + "step": 44151 + }, + { + "epoch": 3.26, + "learning_rate": 9.06998069098948e-06, + "loss": 1.0017, + "step": 44152 + }, + { + "epoch": 3.26, + "learning_rate": 9.069583548847579e-06, + "loss": 0.9637, + "step": 44153 + }, + { + "epoch": 3.26, + "learning_rate": 9.069186408185956e-06, + "loss": 0.9886, + "step": 44154 + }, + { + "epoch": 3.26, + "learning_rate": 9.068789269005242e-06, + "loss": 0.9847, + "step": 44155 + }, + { + "epoch": 3.26, + "learning_rate": 9.068392131306073e-06, + "loss": 1.0225, + "step": 44156 + }, + { + "epoch": 3.26, + "learning_rate": 9.067994995089076e-06, + "loss": 0.9785, + "step": 44157 + }, + { + "epoch": 3.26, + "learning_rate": 9.067597860354888e-06, + "loss": 1.0459, + "step": 44158 + }, + { + "epoch": 3.26, + "learning_rate": 9.067200727104136e-06, + "loss": 0.9256, + "step": 44159 + }, + { + "epoch": 3.26, + "learning_rate": 9.066803595337457e-06, + "loss": 1.0024, + "step": 44160 + }, + { + "epoch": 3.26, + "learning_rate": 9.066406465055477e-06, + "loss": 0.9203, + "step": 44161 + }, + { + "epoch": 3.26, + "learning_rate": 9.066009336258835e-06, + "loss": 1.0334, + "step": 44162 + }, + { + "epoch": 3.26, + "learning_rate": 9.065612208948155e-06, + "loss": 0.9534, + "step": 44163 + }, + { + "epoch": 3.26, + "learning_rate": 9.065215083124075e-06, + "loss": 1.0796, + "step": 44164 + }, + { + "epoch": 3.26, + "learning_rate": 9.064817958787225e-06, + "loss": 1.1436, + "step": 44165 + }, + { + "epoch": 3.26, + "learning_rate": 9.064420835938234e-06, + "loss": 0.9331, + "step": 44166 + }, + { + "epoch": 3.26, + "learning_rate": 9.064023714577738e-06, + "loss": 0.929, + "step": 44167 + }, + { + "epoch": 3.26, + "learning_rate": 9.063626594706366e-06, + "loss": 0.984, + "step": 44168 + }, + { + "epoch": 3.26, + "learning_rate": 9.06322947632475e-06, + "loss": 1.0078, + "step": 44169 + }, + { + "epoch": 3.26, + "learning_rate": 9.062832359433522e-06, + "loss": 0.9845, + "step": 44170 + }, + { + "epoch": 3.26, + "learning_rate": 9.062435244033317e-06, + "loss": 0.962, + "step": 44171 + }, + { + "epoch": 3.26, + "learning_rate": 9.062038130124762e-06, + "loss": 0.9876, + "step": 44172 + }, + { + "epoch": 3.26, + "learning_rate": 9.061641017708491e-06, + "loss": 0.9345, + "step": 44173 + }, + { + "epoch": 3.26, + "learning_rate": 9.06124390678514e-06, + "loss": 0.9825, + "step": 44174 + }, + { + "epoch": 3.26, + "learning_rate": 9.06084679735533e-06, + "loss": 1.0081, + "step": 44175 + }, + { + "epoch": 3.26, + "learning_rate": 9.060449689419706e-06, + "loss": 0.9446, + "step": 44176 + }, + { + "epoch": 3.26, + "learning_rate": 9.060052582978885e-06, + "loss": 1.1143, + "step": 44177 + }, + { + "epoch": 3.26, + "learning_rate": 9.059655478033514e-06, + "loss": 0.9779, + "step": 44178 + }, + { + "epoch": 3.26, + "learning_rate": 9.05925837458421e-06, + "loss": 1.0788, + "step": 44179 + }, + { + "epoch": 3.26, + "learning_rate": 9.058861272631622e-06, + "loss": 1.0025, + "step": 44180 + }, + { + "epoch": 3.26, + "learning_rate": 9.058464172176365e-06, + "loss": 0.94, + "step": 44181 + }, + { + "epoch": 3.26, + "learning_rate": 9.058067073219082e-06, + "loss": 1.0462, + "step": 44182 + }, + { + "epoch": 3.26, + "learning_rate": 9.0576699757604e-06, + "loss": 1.0013, + "step": 44183 + }, + { + "epoch": 3.26, + "learning_rate": 9.05727287980095e-06, + "loss": 1.0019, + "step": 44184 + }, + { + "epoch": 3.26, + "learning_rate": 9.056875785341367e-06, + "loss": 0.99, + "step": 44185 + }, + { + "epoch": 3.26, + "learning_rate": 9.056478692382278e-06, + "loss": 1.0335, + "step": 44186 + }, + { + "epoch": 3.26, + "learning_rate": 9.056081600924322e-06, + "loss": 1.0295, + "step": 44187 + }, + { + "epoch": 3.27, + "learning_rate": 9.055684510968124e-06, + "loss": 1.0184, + "step": 44188 + }, + { + "epoch": 3.27, + "learning_rate": 9.05528742251432e-06, + "loss": 1.0006, + "step": 44189 + }, + { + "epoch": 3.27, + "learning_rate": 9.054890335563538e-06, + "loss": 1.0037, + "step": 44190 + }, + { + "epoch": 3.27, + "learning_rate": 9.054493250116416e-06, + "loss": 1.0443, + "step": 44191 + }, + { + "epoch": 3.27, + "learning_rate": 9.054096166173581e-06, + "loss": 0.9893, + "step": 44192 + }, + { + "epoch": 3.27, + "learning_rate": 9.053699083735661e-06, + "loss": 0.9735, + "step": 44193 + }, + { + "epoch": 3.27, + "learning_rate": 9.053302002803299e-06, + "loss": 1.0091, + "step": 44194 + }, + { + "epoch": 3.27, + "learning_rate": 9.052904923377111e-06, + "loss": 0.9792, + "step": 44195 + }, + { + "epoch": 3.27, + "learning_rate": 9.052507845457746e-06, + "loss": 1.0263, + "step": 44196 + }, + { + "epoch": 3.27, + "learning_rate": 9.05211076904582e-06, + "loss": 0.9895, + "step": 44197 + }, + { + "epoch": 3.27, + "learning_rate": 9.05171369414198e-06, + "loss": 0.9755, + "step": 44198 + }, + { + "epoch": 3.27, + "learning_rate": 9.051316620746843e-06, + "loss": 0.9509, + "step": 44199 + }, + { + "epoch": 3.27, + "learning_rate": 9.050919548861055e-06, + "loss": 0.9567, + "step": 44200 + }, + { + "epoch": 3.27, + "learning_rate": 9.050522478485237e-06, + "loss": 1.0565, + "step": 44201 + }, + { + "epoch": 3.27, + "learning_rate": 9.050125409620022e-06, + "loss": 1.0509, + "step": 44202 + }, + { + "epoch": 3.27, + "learning_rate": 9.049728342266047e-06, + "loss": 1.0094, + "step": 44203 + }, + { + "epoch": 3.27, + "learning_rate": 9.049331276423937e-06, + "loss": 0.9565, + "step": 44204 + }, + { + "epoch": 3.27, + "learning_rate": 9.04893421209433e-06, + "loss": 1.0096, + "step": 44205 + }, + { + "epoch": 3.27, + "learning_rate": 9.048537149277854e-06, + "loss": 1.0, + "step": 44206 + }, + { + "epoch": 3.27, + "learning_rate": 9.048140087975143e-06, + "loss": 1.0338, + "step": 44207 + }, + { + "epoch": 3.27, + "learning_rate": 9.047743028186825e-06, + "loss": 1.0495, + "step": 44208 + }, + { + "epoch": 3.27, + "learning_rate": 9.047345969913537e-06, + "loss": 1.065, + "step": 44209 + }, + { + "epoch": 3.27, + "learning_rate": 9.046948913155908e-06, + "loss": 1.0145, + "step": 44210 + }, + { + "epoch": 3.27, + "learning_rate": 9.046551857914568e-06, + "loss": 1.0411, + "step": 44211 + }, + { + "epoch": 3.27, + "learning_rate": 9.04615480419015e-06, + "loss": 1.0102, + "step": 44212 + }, + { + "epoch": 3.27, + "learning_rate": 9.045757751983285e-06, + "loss": 0.937, + "step": 44213 + }, + { + "epoch": 3.27, + "learning_rate": 9.04536070129461e-06, + "loss": 1.0089, + "step": 44214 + }, + { + "epoch": 3.27, + "learning_rate": 9.044963652124748e-06, + "loss": 0.8683, + "step": 44215 + }, + { + "epoch": 3.27, + "learning_rate": 9.044566604474338e-06, + "loss": 1.0161, + "step": 44216 + }, + { + "epoch": 3.27, + "learning_rate": 9.044169558344008e-06, + "loss": 1.0496, + "step": 44217 + }, + { + "epoch": 3.27, + "learning_rate": 9.043772513734391e-06, + "loss": 1.0729, + "step": 44218 + }, + { + "epoch": 3.27, + "learning_rate": 9.043375470646117e-06, + "loss": 0.9928, + "step": 44219 + }, + { + "epoch": 3.27, + "learning_rate": 9.042978429079821e-06, + "loss": 1.0616, + "step": 44220 + }, + { + "epoch": 3.27, + "learning_rate": 9.042581389036134e-06, + "loss": 0.9145, + "step": 44221 + }, + { + "epoch": 3.27, + "learning_rate": 9.042184350515682e-06, + "loss": 0.9833, + "step": 44222 + }, + { + "epoch": 3.27, + "learning_rate": 9.041787313519105e-06, + "loss": 1.0272, + "step": 44223 + }, + { + "epoch": 3.27, + "learning_rate": 9.041390278047026e-06, + "loss": 0.9063, + "step": 44224 + }, + { + "epoch": 3.27, + "learning_rate": 9.040993244100085e-06, + "loss": 1.0038, + "step": 44225 + }, + { + "epoch": 3.27, + "learning_rate": 9.040596211678907e-06, + "loss": 0.9427, + "step": 44226 + }, + { + "epoch": 3.27, + "learning_rate": 9.04019918078413e-06, + "loss": 1.0485, + "step": 44227 + }, + { + "epoch": 3.27, + "learning_rate": 9.03980215141638e-06, + "loss": 0.9376, + "step": 44228 + }, + { + "epoch": 3.27, + "learning_rate": 9.039405123576294e-06, + "loss": 0.9892, + "step": 44229 + }, + { + "epoch": 3.27, + "learning_rate": 9.039008097264498e-06, + "loss": 1.0767, + "step": 44230 + }, + { + "epoch": 3.27, + "learning_rate": 9.038611072481626e-06, + "loss": 0.9937, + "step": 44231 + }, + { + "epoch": 3.27, + "learning_rate": 9.038214049228314e-06, + "loss": 0.9695, + "step": 44232 + }, + { + "epoch": 3.27, + "learning_rate": 9.037817027505184e-06, + "loss": 1.0155, + "step": 44233 + }, + { + "epoch": 3.27, + "learning_rate": 9.037420007312878e-06, + "loss": 1.0273, + "step": 44234 + }, + { + "epoch": 3.27, + "learning_rate": 9.03702298865202e-06, + "loss": 1.0344, + "step": 44235 + }, + { + "epoch": 3.27, + "learning_rate": 9.036625971523246e-06, + "loss": 0.9439, + "step": 44236 + }, + { + "epoch": 3.27, + "learning_rate": 9.036228955927184e-06, + "loss": 0.9543, + "step": 44237 + }, + { + "epoch": 3.27, + "learning_rate": 9.035831941864473e-06, + "loss": 0.955, + "step": 44238 + }, + { + "epoch": 3.27, + "learning_rate": 9.035434929335738e-06, + "loss": 0.9688, + "step": 44239 + }, + { + "epoch": 3.27, + "learning_rate": 9.035037918341609e-06, + "loss": 1.0022, + "step": 44240 + }, + { + "epoch": 3.27, + "learning_rate": 9.034640908882725e-06, + "loss": 1.0149, + "step": 44241 + }, + { + "epoch": 3.27, + "learning_rate": 9.034243900959706e-06, + "loss": 0.9752, + "step": 44242 + }, + { + "epoch": 3.27, + "learning_rate": 9.0338468945732e-06, + "loss": 1.0597, + "step": 44243 + }, + { + "epoch": 3.27, + "learning_rate": 9.033449889723822e-06, + "loss": 1.0466, + "step": 44244 + }, + { + "epoch": 3.27, + "learning_rate": 9.03305288641222e-06, + "loss": 1.0032, + "step": 44245 + }, + { + "epoch": 3.27, + "learning_rate": 9.032655884639009e-06, + "loss": 1.0283, + "step": 44246 + }, + { + "epoch": 3.27, + "learning_rate": 9.032258884404835e-06, + "loss": 1.0256, + "step": 44247 + }, + { + "epoch": 3.27, + "learning_rate": 9.03186188571032e-06, + "loss": 1.055, + "step": 44248 + }, + { + "epoch": 3.27, + "learning_rate": 9.0314648885561e-06, + "loss": 1.0311, + "step": 44249 + }, + { + "epoch": 3.27, + "learning_rate": 9.031067892942806e-06, + "loss": 1.0601, + "step": 44250 + }, + { + "epoch": 3.27, + "learning_rate": 9.030670898871065e-06, + "loss": 0.9725, + "step": 44251 + }, + { + "epoch": 3.27, + "learning_rate": 9.030273906341518e-06, + "loss": 1.1076, + "step": 44252 + }, + { + "epoch": 3.27, + "learning_rate": 9.029876915354786e-06, + "loss": 0.9841, + "step": 44253 + }, + { + "epoch": 3.27, + "learning_rate": 9.029479925911511e-06, + "loss": 0.9634, + "step": 44254 + }, + { + "epoch": 3.27, + "learning_rate": 9.029082938012316e-06, + "loss": 0.9566, + "step": 44255 + }, + { + "epoch": 3.27, + "learning_rate": 9.02868595165784e-06, + "loss": 1.0479, + "step": 44256 + }, + { + "epoch": 3.27, + "learning_rate": 9.028288966848709e-06, + "loss": 0.9467, + "step": 44257 + }, + { + "epoch": 3.27, + "learning_rate": 9.027891983585554e-06, + "loss": 1.1306, + "step": 44258 + }, + { + "epoch": 3.27, + "learning_rate": 9.027495001869013e-06, + "loss": 0.9617, + "step": 44259 + }, + { + "epoch": 3.27, + "learning_rate": 9.027098021699708e-06, + "loss": 0.9978, + "step": 44260 + }, + { + "epoch": 3.27, + "learning_rate": 9.02670104307828e-06, + "loss": 0.9631, + "step": 44261 + }, + { + "epoch": 3.27, + "learning_rate": 9.026304066005355e-06, + "loss": 0.9533, + "step": 44262 + }, + { + "epoch": 3.27, + "learning_rate": 9.025907090481569e-06, + "loss": 0.9536, + "step": 44263 + }, + { + "epoch": 3.27, + "learning_rate": 9.025510116507548e-06, + "loss": 0.9472, + "step": 44264 + }, + { + "epoch": 3.27, + "learning_rate": 9.02511314408393e-06, + "loss": 0.8451, + "step": 44265 + }, + { + "epoch": 3.27, + "learning_rate": 9.024716173211338e-06, + "loss": 0.9856, + "step": 44266 + }, + { + "epoch": 3.27, + "learning_rate": 9.024319203890413e-06, + "loss": 1.0221, + "step": 44267 + }, + { + "epoch": 3.27, + "learning_rate": 9.023922236121785e-06, + "loss": 0.9046, + "step": 44268 + }, + { + "epoch": 3.27, + "learning_rate": 9.023525269906076e-06, + "loss": 1.0095, + "step": 44269 + }, + { + "epoch": 3.27, + "learning_rate": 9.023128305243927e-06, + "loss": 0.9473, + "step": 44270 + }, + { + "epoch": 3.27, + "learning_rate": 9.022731342135964e-06, + "loss": 0.9818, + "step": 44271 + }, + { + "epoch": 3.27, + "learning_rate": 9.022334380582826e-06, + "loss": 1.064, + "step": 44272 + }, + { + "epoch": 3.27, + "learning_rate": 9.021937420585137e-06, + "loss": 0.9676, + "step": 44273 + }, + { + "epoch": 3.27, + "learning_rate": 9.021540462143534e-06, + "loss": 0.9856, + "step": 44274 + }, + { + "epoch": 3.27, + "learning_rate": 9.021143505258643e-06, + "loss": 0.9684, + "step": 44275 + }, + { + "epoch": 3.27, + "learning_rate": 9.0207465499311e-06, + "loss": 0.9628, + "step": 44276 + }, + { + "epoch": 3.27, + "learning_rate": 9.020349596161537e-06, + "loss": 0.9769, + "step": 44277 + }, + { + "epoch": 3.27, + "learning_rate": 9.019952643950581e-06, + "loss": 0.9064, + "step": 44278 + }, + { + "epoch": 3.27, + "learning_rate": 9.019555693298869e-06, + "loss": 0.9787, + "step": 44279 + }, + { + "epoch": 3.27, + "learning_rate": 9.019158744207029e-06, + "loss": 0.918, + "step": 44280 + }, + { + "epoch": 3.27, + "learning_rate": 9.018761796675692e-06, + "loss": 1.0276, + "step": 44281 + }, + { + "epoch": 3.27, + "learning_rate": 9.01836485070549e-06, + "loss": 1.0501, + "step": 44282 + }, + { + "epoch": 3.27, + "learning_rate": 9.01796790629706e-06, + "loss": 1.0407, + "step": 44283 + }, + { + "epoch": 3.27, + "learning_rate": 9.017570963451025e-06, + "loss": 0.8946, + "step": 44284 + }, + { + "epoch": 3.27, + "learning_rate": 9.017174022168022e-06, + "loss": 0.8673, + "step": 44285 + }, + { + "epoch": 3.27, + "learning_rate": 9.016777082448685e-06, + "loss": 0.9526, + "step": 44286 + }, + { + "epoch": 3.27, + "learning_rate": 9.016380144293634e-06, + "loss": 0.8357, + "step": 44287 + }, + { + "epoch": 3.27, + "learning_rate": 9.015983207703515e-06, + "loss": 1.0194, + "step": 44288 + }, + { + "epoch": 3.27, + "learning_rate": 9.015586272678946e-06, + "loss": 0.9666, + "step": 44289 + }, + { + "epoch": 3.27, + "learning_rate": 9.01518933922057e-06, + "loss": 0.9695, + "step": 44290 + }, + { + "epoch": 3.27, + "learning_rate": 9.01479240732901e-06, + "loss": 1.06, + "step": 44291 + }, + { + "epoch": 3.27, + "learning_rate": 9.014395477004908e-06, + "loss": 1.0743, + "step": 44292 + }, + { + "epoch": 3.27, + "learning_rate": 9.013998548248878e-06, + "loss": 1.038, + "step": 44293 + }, + { + "epoch": 3.27, + "learning_rate": 9.013601621061573e-06, + "loss": 1.1326, + "step": 44294 + }, + { + "epoch": 3.27, + "learning_rate": 9.013204695443609e-06, + "loss": 1.0903, + "step": 44295 + }, + { + "epoch": 3.27, + "learning_rate": 9.012807771395623e-06, + "loss": 0.9504, + "step": 44296 + }, + { + "epoch": 3.27, + "learning_rate": 9.012410848918244e-06, + "loss": 0.9693, + "step": 44297 + }, + { + "epoch": 3.27, + "learning_rate": 9.012013928012106e-06, + "loss": 0.9942, + "step": 44298 + }, + { + "epoch": 3.27, + "learning_rate": 9.011617008677839e-06, + "loss": 0.9544, + "step": 44299 + }, + { + "epoch": 3.27, + "learning_rate": 9.011220090916075e-06, + "loss": 0.9563, + "step": 44300 + }, + { + "epoch": 3.27, + "learning_rate": 9.010823174727447e-06, + "loss": 0.9476, + "step": 44301 + }, + { + "epoch": 3.27, + "learning_rate": 9.010426260112585e-06, + "loss": 0.9188, + "step": 44302 + }, + { + "epoch": 3.27, + "learning_rate": 9.01002934707212e-06, + "loss": 1.1088, + "step": 44303 + }, + { + "epoch": 3.27, + "learning_rate": 9.009632435606687e-06, + "loss": 0.9451, + "step": 44304 + }, + { + "epoch": 3.27, + "learning_rate": 9.00923552571691e-06, + "loss": 1.1284, + "step": 44305 + }, + { + "epoch": 3.27, + "learning_rate": 9.00883861740343e-06, + "loss": 1.0485, + "step": 44306 + }, + { + "epoch": 3.27, + "learning_rate": 9.008441710666869e-06, + "loss": 0.9491, + "step": 44307 + }, + { + "epoch": 3.27, + "learning_rate": 9.008044805507866e-06, + "loss": 0.9814, + "step": 44308 + }, + { + "epoch": 3.27, + "learning_rate": 9.007647901927047e-06, + "loss": 0.9555, + "step": 44309 + }, + { + "epoch": 3.27, + "learning_rate": 9.00725099992505e-06, + "loss": 1.0758, + "step": 44310 + }, + { + "epoch": 3.27, + "learning_rate": 9.006854099502498e-06, + "loss": 1.0829, + "step": 44311 + }, + { + "epoch": 3.27, + "learning_rate": 9.00645720066003e-06, + "loss": 1.0188, + "step": 44312 + }, + { + "epoch": 3.27, + "learning_rate": 9.006060303398276e-06, + "loss": 1.0358, + "step": 44313 + }, + { + "epoch": 3.27, + "learning_rate": 9.005663407717862e-06, + "loss": 1.1136, + "step": 44314 + }, + { + "epoch": 3.27, + "learning_rate": 9.005266513619426e-06, + "loss": 1.0796, + "step": 44315 + }, + { + "epoch": 3.27, + "learning_rate": 9.004869621103593e-06, + "loss": 1.0676, + "step": 44316 + }, + { + "epoch": 3.27, + "learning_rate": 9.004472730171001e-06, + "loss": 1.0395, + "step": 44317 + }, + { + "epoch": 3.27, + "learning_rate": 9.004075840822277e-06, + "loss": 1.049, + "step": 44318 + }, + { + "epoch": 3.27, + "learning_rate": 9.003678953058057e-06, + "loss": 1.0213, + "step": 44319 + }, + { + "epoch": 3.27, + "learning_rate": 9.003282066878965e-06, + "loss": 1.0934, + "step": 44320 + }, + { + "epoch": 3.27, + "learning_rate": 9.002885182285641e-06, + "loss": 1.0451, + "step": 44321 + }, + { + "epoch": 3.27, + "learning_rate": 9.00248829927871e-06, + "loss": 0.9781, + "step": 44322 + }, + { + "epoch": 3.27, + "learning_rate": 9.002091417858807e-06, + "loss": 1.023, + "step": 44323 + }, + { + "epoch": 3.28, + "learning_rate": 9.001694538026564e-06, + "loss": 1.0214, + "step": 44324 + }, + { + "epoch": 3.28, + "learning_rate": 9.001297659782607e-06, + "loss": 1.0549, + "step": 44325 + }, + { + "epoch": 3.28, + "learning_rate": 9.000900783127575e-06, + "loss": 0.9725, + "step": 44326 + }, + { + "epoch": 3.28, + "learning_rate": 9.00050390806209e-06, + "loss": 0.912, + "step": 44327 + }, + { + "epoch": 3.28, + "learning_rate": 9.000107034586794e-06, + "loss": 0.9919, + "step": 44328 + }, + { + "epoch": 3.28, + "learning_rate": 8.99971016270231e-06, + "loss": 1.0672, + "step": 44329 + }, + { + "epoch": 3.28, + "learning_rate": 8.999313292409278e-06, + "loss": 0.9106, + "step": 44330 + }, + { + "epoch": 3.28, + "learning_rate": 8.998916423708318e-06, + "loss": 1.0436, + "step": 44331 + }, + { + "epoch": 3.28, + "learning_rate": 8.998519556600072e-06, + "loss": 1.0225, + "step": 44332 + }, + { + "epoch": 3.28, + "learning_rate": 8.998122691085169e-06, + "loss": 1.0854, + "step": 44333 + }, + { + "epoch": 3.28, + "learning_rate": 8.997725827164231e-06, + "loss": 1.0636, + "step": 44334 + }, + { + "epoch": 3.28, + "learning_rate": 8.997328964837904e-06, + "loss": 1.0245, + "step": 44335 + }, + { + "epoch": 3.28, + "learning_rate": 8.996932104106806e-06, + "loss": 1.031, + "step": 44336 + }, + { + "epoch": 3.28, + "learning_rate": 8.996535244971581e-06, + "loss": 1.0815, + "step": 44337 + }, + { + "epoch": 3.28, + "learning_rate": 8.996138387432847e-06, + "loss": 1.0427, + "step": 44338 + }, + { + "epoch": 3.28, + "learning_rate": 8.995741531491251e-06, + "loss": 0.9581, + "step": 44339 + }, + { + "epoch": 3.28, + "learning_rate": 8.995344677147407e-06, + "loss": 0.9288, + "step": 44340 + }, + { + "epoch": 3.28, + "learning_rate": 8.994947824401964e-06, + "loss": 0.974, + "step": 44341 + }, + { + "epoch": 3.28, + "learning_rate": 8.99455097325554e-06, + "loss": 1.0712, + "step": 44342 + }, + { + "epoch": 3.28, + "learning_rate": 8.99415412370877e-06, + "loss": 0.8537, + "step": 44343 + }, + { + "epoch": 3.28, + "learning_rate": 8.993757275762288e-06, + "loss": 1.1024, + "step": 44344 + }, + { + "epoch": 3.28, + "learning_rate": 8.993360429416724e-06, + "loss": 1.0195, + "step": 44345 + }, + { + "epoch": 3.28, + "learning_rate": 8.992963584672709e-06, + "loss": 1.0441, + "step": 44346 + }, + { + "epoch": 3.28, + "learning_rate": 8.992566741530873e-06, + "loss": 0.9743, + "step": 44347 + }, + { + "epoch": 3.28, + "learning_rate": 8.992169899991851e-06, + "loss": 0.9659, + "step": 44348 + }, + { + "epoch": 3.28, + "learning_rate": 8.99177306005627e-06, + "loss": 1.0324, + "step": 44349 + }, + { + "epoch": 3.28, + "learning_rate": 8.991376221724766e-06, + "loss": 1.0151, + "step": 44350 + }, + { + "epoch": 3.28, + "learning_rate": 8.990979384997967e-06, + "loss": 0.8761, + "step": 44351 + }, + { + "epoch": 3.28, + "learning_rate": 8.990582549876504e-06, + "loss": 1.0002, + "step": 44352 + }, + { + "epoch": 3.28, + "learning_rate": 8.990185716361012e-06, + "loss": 1.0637, + "step": 44353 + }, + { + "epoch": 3.28, + "learning_rate": 8.989788884452117e-06, + "loss": 1.0126, + "step": 44354 + }, + { + "epoch": 3.28, + "learning_rate": 8.989392054150458e-06, + "loss": 1.1008, + "step": 44355 + }, + { + "epoch": 3.28, + "learning_rate": 8.988995225456656e-06, + "loss": 1.0481, + "step": 44356 + }, + { + "epoch": 3.28, + "learning_rate": 8.988598398371354e-06, + "loss": 0.8984, + "step": 44357 + }, + { + "epoch": 3.28, + "learning_rate": 8.988201572895173e-06, + "loss": 1.0118, + "step": 44358 + }, + { + "epoch": 3.28, + "learning_rate": 8.987804749028753e-06, + "loss": 0.9964, + "step": 44359 + }, + { + "epoch": 3.28, + "learning_rate": 8.987407926772721e-06, + "loss": 1.0331, + "step": 44360 + }, + { + "epoch": 3.28, + "learning_rate": 8.987011106127705e-06, + "loss": 1.1103, + "step": 44361 + }, + { + "epoch": 3.28, + "learning_rate": 8.986614287094341e-06, + "loss": 1.0487, + "step": 44362 + }, + { + "epoch": 3.28, + "learning_rate": 8.98621746967326e-06, + "loss": 1.0376, + "step": 44363 + }, + { + "epoch": 3.28, + "learning_rate": 8.985820653865092e-06, + "loss": 1.0524, + "step": 44364 + }, + { + "epoch": 3.28, + "learning_rate": 8.985423839670467e-06, + "loss": 1.0679, + "step": 44365 + }, + { + "epoch": 3.28, + "learning_rate": 8.985027027090021e-06, + "loss": 1.0088, + "step": 44366 + }, + { + "epoch": 3.28, + "learning_rate": 8.98463021612438e-06, + "loss": 1.0132, + "step": 44367 + }, + { + "epoch": 3.28, + "learning_rate": 8.98423340677418e-06, + "loss": 1.1364, + "step": 44368 + }, + { + "epoch": 3.28, + "learning_rate": 8.983836599040051e-06, + "loss": 0.9573, + "step": 44369 + }, + { + "epoch": 3.28, + "learning_rate": 8.983439792922619e-06, + "loss": 1.0718, + "step": 44370 + }, + { + "epoch": 3.28, + "learning_rate": 8.983042988422523e-06, + "loss": 0.9918, + "step": 44371 + }, + { + "epoch": 3.28, + "learning_rate": 8.98264618554039e-06, + "loss": 1.0954, + "step": 44372 + }, + { + "epoch": 3.28, + "learning_rate": 8.982249384276855e-06, + "loss": 1.0379, + "step": 44373 + }, + { + "epoch": 3.28, + "learning_rate": 8.981852584632542e-06, + "loss": 1.0662, + "step": 44374 + }, + { + "epoch": 3.28, + "learning_rate": 8.98145578660809e-06, + "loss": 0.9778, + "step": 44375 + }, + { + "epoch": 3.28, + "learning_rate": 8.981058990204126e-06, + "loss": 0.9453, + "step": 44376 + }, + { + "epoch": 3.28, + "learning_rate": 8.980662195421285e-06, + "loss": 1.037, + "step": 44377 + }, + { + "epoch": 3.28, + "learning_rate": 8.980265402260192e-06, + "loss": 1.065, + "step": 44378 + }, + { + "epoch": 3.28, + "learning_rate": 8.979868610721485e-06, + "loss": 1.0296, + "step": 44379 + }, + { + "epoch": 3.28, + "learning_rate": 8.979471820805795e-06, + "loss": 0.9251, + "step": 44380 + }, + { + "epoch": 3.28, + "learning_rate": 8.979075032513744e-06, + "loss": 1.0498, + "step": 44381 + }, + { + "epoch": 3.28, + "learning_rate": 8.978678245845977e-06, + "loss": 1.1246, + "step": 44382 + }, + { + "epoch": 3.28, + "learning_rate": 8.978281460803111e-06, + "loss": 0.9848, + "step": 44383 + }, + { + "epoch": 3.28, + "learning_rate": 8.977884677385792e-06, + "loss": 0.9466, + "step": 44384 + }, + { + "epoch": 3.28, + "learning_rate": 8.977487895594636e-06, + "loss": 1.0372, + "step": 44385 + }, + { + "epoch": 3.28, + "learning_rate": 8.97709111543029e-06, + "loss": 1.0368, + "step": 44386 + }, + { + "epoch": 3.28, + "learning_rate": 8.97669433689337e-06, + "loss": 0.9804, + "step": 44387 + }, + { + "epoch": 3.28, + "learning_rate": 8.976297559984523e-06, + "loss": 0.9674, + "step": 44388 + }, + { + "epoch": 3.28, + "learning_rate": 8.975900784704368e-06, + "loss": 1.0554, + "step": 44389 + }, + { + "epoch": 3.28, + "learning_rate": 8.975504011053538e-06, + "loss": 1.1688, + "step": 44390 + }, + { + "epoch": 3.28, + "learning_rate": 8.97510723903267e-06, + "loss": 1.06, + "step": 44391 + }, + { + "epoch": 3.28, + "learning_rate": 8.974710468642387e-06, + "loss": 0.9797, + "step": 44392 + }, + { + "epoch": 3.28, + "learning_rate": 8.97431369988333e-06, + "loss": 1.0446, + "step": 44393 + }, + { + "epoch": 3.28, + "learning_rate": 8.97391693275612e-06, + "loss": 0.9755, + "step": 44394 + }, + { + "epoch": 3.28, + "learning_rate": 8.973520167261398e-06, + "loss": 1.0002, + "step": 44395 + }, + { + "epoch": 3.28, + "learning_rate": 8.973123403399789e-06, + "loss": 1.0158, + "step": 44396 + }, + { + "epoch": 3.28, + "learning_rate": 8.972726641171926e-06, + "loss": 1.0339, + "step": 44397 + }, + { + "epoch": 3.28, + "learning_rate": 8.97232988057844e-06, + "loss": 1.1333, + "step": 44398 + }, + { + "epoch": 3.28, + "learning_rate": 8.971933121619962e-06, + "loss": 1.0192, + "step": 44399 + }, + { + "epoch": 3.28, + "learning_rate": 8.971536364297126e-06, + "loss": 0.9907, + "step": 44400 + }, + { + "epoch": 3.28, + "learning_rate": 8.971139608610558e-06, + "loss": 0.9788, + "step": 44401 + }, + { + "epoch": 3.28, + "learning_rate": 8.970742854560897e-06, + "loss": 1.0658, + "step": 44402 + }, + { + "epoch": 3.28, + "learning_rate": 8.970346102148762e-06, + "loss": 0.9528, + "step": 44403 + }, + { + "epoch": 3.28, + "learning_rate": 8.969949351374799e-06, + "loss": 1.0254, + "step": 44404 + }, + { + "epoch": 3.28, + "learning_rate": 8.969552602239626e-06, + "loss": 0.9474, + "step": 44405 + }, + { + "epoch": 3.28, + "learning_rate": 8.969155854743885e-06, + "loss": 0.9827, + "step": 44406 + }, + { + "epoch": 3.28, + "learning_rate": 8.9687591088882e-06, + "loss": 0.9267, + "step": 44407 + }, + { + "epoch": 3.28, + "learning_rate": 8.968362364673205e-06, + "loss": 0.9265, + "step": 44408 + }, + { + "epoch": 3.28, + "learning_rate": 8.96796562209953e-06, + "loss": 0.9507, + "step": 44409 + }, + { + "epoch": 3.28, + "learning_rate": 8.967568881167805e-06, + "loss": 0.9164, + "step": 44410 + }, + { + "epoch": 3.28, + "learning_rate": 8.967172141878665e-06, + "loss": 1.0143, + "step": 44411 + }, + { + "epoch": 3.28, + "learning_rate": 8.966775404232739e-06, + "loss": 0.9567, + "step": 44412 + }, + { + "epoch": 3.28, + "learning_rate": 8.966378668230661e-06, + "loss": 1.0335, + "step": 44413 + }, + { + "epoch": 3.28, + "learning_rate": 8.965981933873055e-06, + "loss": 1.0484, + "step": 44414 + }, + { + "epoch": 3.28, + "learning_rate": 8.965585201160562e-06, + "loss": 0.9659, + "step": 44415 + }, + { + "epoch": 3.28, + "learning_rate": 8.965188470093806e-06, + "loss": 0.9852, + "step": 44416 + }, + { + "epoch": 3.28, + "learning_rate": 8.96479174067342e-06, + "loss": 1.021, + "step": 44417 + }, + { + "epoch": 3.28, + "learning_rate": 8.964395012900035e-06, + "loss": 1.0776, + "step": 44418 + }, + { + "epoch": 3.28, + "learning_rate": 8.963998286774284e-06, + "loss": 0.8621, + "step": 44419 + }, + { + "epoch": 3.28, + "learning_rate": 8.963601562296796e-06, + "loss": 1.0355, + "step": 44420 + }, + { + "epoch": 3.28, + "learning_rate": 8.963204839468203e-06, + "loss": 1.1649, + "step": 44421 + }, + { + "epoch": 3.28, + "learning_rate": 8.962808118289137e-06, + "loss": 1.0372, + "step": 44422 + }, + { + "epoch": 3.28, + "learning_rate": 8.962411398760227e-06, + "loss": 0.9645, + "step": 44423 + }, + { + "epoch": 3.28, + "learning_rate": 8.962014680882107e-06, + "loss": 1.1221, + "step": 44424 + }, + { + "epoch": 3.28, + "learning_rate": 8.96161796465541e-06, + "loss": 0.9897, + "step": 44425 + }, + { + "epoch": 3.28, + "learning_rate": 8.961221250080757e-06, + "loss": 1.049, + "step": 44426 + }, + { + "epoch": 3.28, + "learning_rate": 8.960824537158793e-06, + "loss": 1.0596, + "step": 44427 + }, + { + "epoch": 3.28, + "learning_rate": 8.960427825890136e-06, + "loss": 1.0618, + "step": 44428 + }, + { + "epoch": 3.28, + "learning_rate": 8.96003111627543e-06, + "loss": 0.9434, + "step": 44429 + }, + { + "epoch": 3.28, + "learning_rate": 8.959634408315293e-06, + "loss": 1.0995, + "step": 44430 + }, + { + "epoch": 3.28, + "learning_rate": 8.959237702010368e-06, + "loss": 0.9858, + "step": 44431 + }, + { + "epoch": 3.28, + "learning_rate": 8.958840997361274e-06, + "loss": 1.0103, + "step": 44432 + }, + { + "epoch": 3.28, + "learning_rate": 8.958444294368658e-06, + "loss": 0.9718, + "step": 44433 + }, + { + "epoch": 3.28, + "learning_rate": 8.958047593033135e-06, + "loss": 0.9854, + "step": 44434 + }, + { + "epoch": 3.28, + "learning_rate": 8.957650893355348e-06, + "loss": 0.9846, + "step": 44435 + }, + { + "epoch": 3.28, + "learning_rate": 8.957254195335924e-06, + "loss": 0.938, + "step": 44436 + }, + { + "epoch": 3.28, + "learning_rate": 8.95685749897549e-06, + "loss": 0.9905, + "step": 44437 + }, + { + "epoch": 3.28, + "learning_rate": 8.956460804274682e-06, + "loss": 1.0614, + "step": 44438 + }, + { + "epoch": 3.28, + "learning_rate": 8.956064111234129e-06, + "loss": 0.9856, + "step": 44439 + }, + { + "epoch": 3.28, + "learning_rate": 8.955667419854465e-06, + "loss": 0.986, + "step": 44440 + }, + { + "epoch": 3.28, + "learning_rate": 8.955270730136319e-06, + "loss": 0.9631, + "step": 44441 + }, + { + "epoch": 3.28, + "learning_rate": 8.954874042080323e-06, + "loss": 0.8811, + "step": 44442 + }, + { + "epoch": 3.28, + "learning_rate": 8.954477355687104e-06, + "loss": 1.0727, + "step": 44443 + }, + { + "epoch": 3.28, + "learning_rate": 8.9540806709573e-06, + "loss": 1.01, + "step": 44444 + }, + { + "epoch": 3.28, + "learning_rate": 8.953683987891539e-06, + "loss": 0.9756, + "step": 44445 + }, + { + "epoch": 3.28, + "learning_rate": 8.953287306490448e-06, + "loss": 0.9182, + "step": 44446 + }, + { + "epoch": 3.28, + "learning_rate": 8.952890626754668e-06, + "loss": 1.1223, + "step": 44447 + }, + { + "epoch": 3.28, + "learning_rate": 8.952493948684818e-06, + "loss": 1.0121, + "step": 44448 + }, + { + "epoch": 3.28, + "learning_rate": 8.95209727228154e-06, + "loss": 1.055, + "step": 44449 + }, + { + "epoch": 3.28, + "learning_rate": 8.951700597545453e-06, + "loss": 0.9747, + "step": 44450 + }, + { + "epoch": 3.28, + "learning_rate": 8.951303924477204e-06, + "loss": 0.9585, + "step": 44451 + }, + { + "epoch": 3.28, + "learning_rate": 8.950907253077407e-06, + "loss": 0.9049, + "step": 44452 + }, + { + "epoch": 3.28, + "learning_rate": 8.95051058334671e-06, + "loss": 0.9899, + "step": 44453 + }, + { + "epoch": 3.28, + "learning_rate": 8.950113915285732e-06, + "loss": 1.0628, + "step": 44454 + }, + { + "epoch": 3.28, + "learning_rate": 8.949717248895107e-06, + "loss": 1.0071, + "step": 44455 + }, + { + "epoch": 3.28, + "learning_rate": 8.949320584175468e-06, + "loss": 1.0134, + "step": 44456 + }, + { + "epoch": 3.28, + "learning_rate": 8.948923921127442e-06, + "loss": 0.9224, + "step": 44457 + }, + { + "epoch": 3.28, + "learning_rate": 8.948527259751667e-06, + "loss": 1.0091, + "step": 44458 + }, + { + "epoch": 3.29, + "learning_rate": 8.948130600048767e-06, + "loss": 0.9853, + "step": 44459 + }, + { + "epoch": 3.29, + "learning_rate": 8.947733942019377e-06, + "loss": 0.9921, + "step": 44460 + }, + { + "epoch": 3.29, + "learning_rate": 8.947337285664128e-06, + "loss": 1.0619, + "step": 44461 + }, + { + "epoch": 3.29, + "learning_rate": 8.94694063098365e-06, + "loss": 0.933, + "step": 44462 + }, + { + "epoch": 3.29, + "learning_rate": 8.946543977978575e-06, + "loss": 0.9215, + "step": 44463 + }, + { + "epoch": 3.29, + "learning_rate": 8.946147326649533e-06, + "loss": 0.9696, + "step": 44464 + }, + { + "epoch": 3.29, + "learning_rate": 8.945750676997156e-06, + "loss": 0.9111, + "step": 44465 + }, + { + "epoch": 3.29, + "learning_rate": 8.945354029022073e-06, + "loss": 1.0887, + "step": 44466 + }, + { + "epoch": 3.29, + "learning_rate": 8.944957382724918e-06, + "loss": 0.9119, + "step": 44467 + }, + { + "epoch": 3.29, + "learning_rate": 8.944560738106317e-06, + "loss": 0.9907, + "step": 44468 + }, + { + "epoch": 3.29, + "learning_rate": 8.944164095166911e-06, + "loss": 0.9888, + "step": 44469 + }, + { + "epoch": 3.29, + "learning_rate": 8.94376745390732e-06, + "loss": 1.0388, + "step": 44470 + }, + { + "epoch": 3.29, + "learning_rate": 8.943370814328183e-06, + "loss": 1.1374, + "step": 44471 + }, + { + "epoch": 3.29, + "learning_rate": 8.94297417643013e-06, + "loss": 1.0816, + "step": 44472 + }, + { + "epoch": 3.29, + "learning_rate": 8.942577540213783e-06, + "loss": 1.0532, + "step": 44473 + }, + { + "epoch": 3.29, + "learning_rate": 8.942180905679786e-06, + "loss": 1.0509, + "step": 44474 + }, + { + "epoch": 3.29, + "learning_rate": 8.94178427282876e-06, + "loss": 0.9257, + "step": 44475 + }, + { + "epoch": 3.29, + "learning_rate": 8.941387641661344e-06, + "loss": 1.0711, + "step": 44476 + }, + { + "epoch": 3.29, + "learning_rate": 8.94099101217816e-06, + "loss": 1.0654, + "step": 44477 + }, + { + "epoch": 3.29, + "learning_rate": 8.940594384379849e-06, + "loss": 0.9901, + "step": 44478 + }, + { + "epoch": 3.29, + "learning_rate": 8.940197758267035e-06, + "loss": 1.0809, + "step": 44479 + }, + { + "epoch": 3.29, + "learning_rate": 8.939801133840352e-06, + "loss": 1.1233, + "step": 44480 + }, + { + "epoch": 3.29, + "learning_rate": 8.939404511100428e-06, + "loss": 0.9701, + "step": 44481 + }, + { + "epoch": 3.29, + "learning_rate": 8.939007890047899e-06, + "loss": 1.0319, + "step": 44482 + }, + { + "epoch": 3.29, + "learning_rate": 8.938611270683393e-06, + "loss": 0.9455, + "step": 44483 + }, + { + "epoch": 3.29, + "learning_rate": 8.938214653007539e-06, + "loss": 0.9818, + "step": 44484 + }, + { + "epoch": 3.29, + "learning_rate": 8.937818037020973e-06, + "loss": 1.0579, + "step": 44485 + }, + { + "epoch": 3.29, + "learning_rate": 8.937421422724321e-06, + "loss": 0.9486, + "step": 44486 + }, + { + "epoch": 3.29, + "learning_rate": 8.937024810118218e-06, + "loss": 0.8875, + "step": 44487 + }, + { + "epoch": 3.29, + "learning_rate": 8.936628199203291e-06, + "loss": 1.0227, + "step": 44488 + }, + { + "epoch": 3.29, + "learning_rate": 8.936231589980175e-06, + "loss": 1.0252, + "step": 44489 + }, + { + "epoch": 3.29, + "learning_rate": 8.9358349824495e-06, + "loss": 1.0404, + "step": 44490 + }, + { + "epoch": 3.29, + "learning_rate": 8.935438376611895e-06, + "loss": 0.982, + "step": 44491 + }, + { + "epoch": 3.29, + "learning_rate": 8.935041772467995e-06, + "loss": 0.9306, + "step": 44492 + }, + { + "epoch": 3.29, + "learning_rate": 8.934645170018424e-06, + "loss": 0.885, + "step": 44493 + }, + { + "epoch": 3.29, + "learning_rate": 8.934248569263823e-06, + "loss": 1.0375, + "step": 44494 + }, + { + "epoch": 3.29, + "learning_rate": 8.93385197020481e-06, + "loss": 1.0583, + "step": 44495 + }, + { + "epoch": 3.29, + "learning_rate": 8.93345537284203e-06, + "loss": 1.0362, + "step": 44496 + }, + { + "epoch": 3.29, + "learning_rate": 8.933058777176101e-06, + "loss": 1.0013, + "step": 44497 + }, + { + "epoch": 3.29, + "learning_rate": 8.932662183207666e-06, + "loss": 0.9003, + "step": 44498 + }, + { + "epoch": 3.29, + "learning_rate": 8.932265590937345e-06, + "loss": 1.0942, + "step": 44499 + }, + { + "epoch": 3.29, + "learning_rate": 8.931869000365781e-06, + "loss": 1.0785, + "step": 44500 + }, + { + "epoch": 3.29, + "learning_rate": 8.931472411493594e-06, + "loss": 0.9911, + "step": 44501 + }, + { + "epoch": 3.29, + "learning_rate": 8.931075824321417e-06, + "loss": 1.0067, + "step": 44502 + }, + { + "epoch": 3.29, + "learning_rate": 8.930679238849888e-06, + "loss": 1.0801, + "step": 44503 + }, + { + "epoch": 3.29, + "learning_rate": 8.930282655079629e-06, + "loss": 0.9479, + "step": 44504 + }, + { + "epoch": 3.29, + "learning_rate": 8.929886073011276e-06, + "loss": 0.978, + "step": 44505 + }, + { + "epoch": 3.29, + "learning_rate": 8.929489492645458e-06, + "loss": 1.0641, + "step": 44506 + }, + { + "epoch": 3.29, + "learning_rate": 8.929092913982808e-06, + "loss": 1.054, + "step": 44507 + }, + { + "epoch": 3.29, + "learning_rate": 8.928696337023956e-06, + "loss": 1.0, + "step": 44508 + }, + { + "epoch": 3.29, + "learning_rate": 8.928299761769533e-06, + "loss": 0.9931, + "step": 44509 + }, + { + "epoch": 3.29, + "learning_rate": 8.92790318822017e-06, + "loss": 0.9047, + "step": 44510 + }, + { + "epoch": 3.29, + "learning_rate": 8.927506616376496e-06, + "loss": 1.0413, + "step": 44511 + }, + { + "epoch": 3.29, + "learning_rate": 8.927110046239146e-06, + "loss": 1.0134, + "step": 44512 + }, + { + "epoch": 3.29, + "learning_rate": 8.926713477808745e-06, + "loss": 1.0297, + "step": 44513 + }, + { + "epoch": 3.29, + "learning_rate": 8.926316911085933e-06, + "loss": 1.0594, + "step": 44514 + }, + { + "epoch": 3.29, + "learning_rate": 8.925920346071331e-06, + "loss": 0.9017, + "step": 44515 + }, + { + "epoch": 3.29, + "learning_rate": 8.925523782765577e-06, + "loss": 0.9337, + "step": 44516 + }, + { + "epoch": 3.29, + "learning_rate": 8.925127221169295e-06, + "loss": 0.9524, + "step": 44517 + }, + { + "epoch": 3.29, + "learning_rate": 8.924730661283125e-06, + "loss": 0.8889, + "step": 44518 + }, + { + "epoch": 3.29, + "learning_rate": 8.924334103107694e-06, + "loss": 1.0355, + "step": 44519 + }, + { + "epoch": 3.29, + "learning_rate": 8.923937546643626e-06, + "loss": 0.7756, + "step": 44520 + }, + { + "epoch": 3.29, + "learning_rate": 8.923540991891564e-06, + "loss": 0.9508, + "step": 44521 + }, + { + "epoch": 3.29, + "learning_rate": 8.923144438852127e-06, + "loss": 0.8107, + "step": 44522 + }, + { + "epoch": 3.29, + "learning_rate": 8.922747887525957e-06, + "loss": 1.0893, + "step": 44523 + }, + { + "epoch": 3.29, + "learning_rate": 8.922351337913678e-06, + "loss": 0.9252, + "step": 44524 + }, + { + "epoch": 3.29, + "learning_rate": 8.921954790015923e-06, + "loss": 1.0309, + "step": 44525 + }, + { + "epoch": 3.29, + "learning_rate": 8.921558243833321e-06, + "loss": 1.1295, + "step": 44526 + }, + { + "epoch": 3.29, + "learning_rate": 8.921161699366506e-06, + "loss": 0.9745, + "step": 44527 + }, + { + "epoch": 3.29, + "learning_rate": 8.920765156616108e-06, + "loss": 0.9865, + "step": 44528 + }, + { + "epoch": 3.29, + "learning_rate": 8.920368615582752e-06, + "loss": 1.0566, + "step": 44529 + }, + { + "epoch": 3.29, + "learning_rate": 8.91997207626708e-06, + "loss": 1.0715, + "step": 44530 + }, + { + "epoch": 3.29, + "learning_rate": 8.919575538669713e-06, + "loss": 0.9841, + "step": 44531 + }, + { + "epoch": 3.29, + "learning_rate": 8.919179002791289e-06, + "loss": 1.0769, + "step": 44532 + }, + { + "epoch": 3.29, + "learning_rate": 8.918782468632432e-06, + "loss": 0.9623, + "step": 44533 + }, + { + "epoch": 3.29, + "learning_rate": 8.918385936193781e-06, + "loss": 0.9839, + "step": 44534 + }, + { + "epoch": 3.29, + "learning_rate": 8.91798940547596e-06, + "loss": 1.0312, + "step": 44535 + }, + { + "epoch": 3.29, + "learning_rate": 8.917592876479603e-06, + "loss": 0.9414, + "step": 44536 + }, + { + "epoch": 3.29, + "learning_rate": 8.917196349205339e-06, + "loss": 1.0051, + "step": 44537 + }, + { + "epoch": 3.29, + "learning_rate": 8.9167998236538e-06, + "loss": 0.9948, + "step": 44538 + }, + { + "epoch": 3.29, + "learning_rate": 8.916403299825623e-06, + "loss": 1.0626, + "step": 44539 + }, + { + "epoch": 3.29, + "learning_rate": 8.916006777721424e-06, + "loss": 0.9841, + "step": 44540 + }, + { + "epoch": 3.29, + "learning_rate": 8.91561025734185e-06, + "loss": 1.0042, + "step": 44541 + }, + { + "epoch": 3.29, + "learning_rate": 8.915213738687517e-06, + "loss": 0.9654, + "step": 44542 + }, + { + "epoch": 3.29, + "learning_rate": 8.914817221759073e-06, + "loss": 0.9188, + "step": 44543 + }, + { + "epoch": 3.29, + "learning_rate": 8.91442070655713e-06, + "loss": 0.9626, + "step": 44544 + }, + { + "epoch": 3.29, + "learning_rate": 8.914024193082335e-06, + "loss": 0.9779, + "step": 44545 + }, + { + "epoch": 3.29, + "learning_rate": 8.913627681335304e-06, + "loss": 1.015, + "step": 44546 + }, + { + "epoch": 3.29, + "learning_rate": 8.913231171316685e-06, + "loss": 1.1272, + "step": 44547 + }, + { + "epoch": 3.29, + "learning_rate": 8.912834663027096e-06, + "loss": 1.0062, + "step": 44548 + }, + { + "epoch": 3.29, + "learning_rate": 8.912438156467172e-06, + "loss": 1.0538, + "step": 44549 + }, + { + "epoch": 3.29, + "learning_rate": 8.912041651637541e-06, + "loss": 1.0409, + "step": 44550 + }, + { + "epoch": 3.29, + "learning_rate": 8.911645148538838e-06, + "loss": 0.9148, + "step": 44551 + }, + { + "epoch": 3.29, + "learning_rate": 8.911248647171691e-06, + "loss": 0.9722, + "step": 44552 + }, + { + "epoch": 3.29, + "learning_rate": 8.91085214753673e-06, + "loss": 0.9874, + "step": 44553 + }, + { + "epoch": 3.29, + "learning_rate": 8.910455649634593e-06, + "loss": 0.9827, + "step": 44554 + }, + { + "epoch": 3.29, + "learning_rate": 8.910059153465901e-06, + "loss": 1.0637, + "step": 44555 + }, + { + "epoch": 3.29, + "learning_rate": 8.909662659031293e-06, + "loss": 1.0288, + "step": 44556 + }, + { + "epoch": 3.29, + "learning_rate": 8.909266166331394e-06, + "loss": 1.0397, + "step": 44557 + }, + { + "epoch": 3.29, + "learning_rate": 8.908869675366835e-06, + "loss": 1.0334, + "step": 44558 + }, + { + "epoch": 3.29, + "learning_rate": 8.90847318613825e-06, + "loss": 0.9183, + "step": 44559 + }, + { + "epoch": 3.29, + "learning_rate": 8.90807669864627e-06, + "loss": 1.029, + "step": 44560 + }, + { + "epoch": 3.29, + "learning_rate": 8.907680212891524e-06, + "loss": 0.9716, + "step": 44561 + }, + { + "epoch": 3.29, + "learning_rate": 8.907283728874642e-06, + "loss": 1.0284, + "step": 44562 + }, + { + "epoch": 3.29, + "learning_rate": 8.906887246596257e-06, + "loss": 0.9193, + "step": 44563 + }, + { + "epoch": 3.29, + "learning_rate": 8.906490766056997e-06, + "loss": 1.0708, + "step": 44564 + }, + { + "epoch": 3.29, + "learning_rate": 8.906094287257497e-06, + "loss": 0.8631, + "step": 44565 + }, + { + "epoch": 3.29, + "learning_rate": 8.905697810198387e-06, + "loss": 1.0635, + "step": 44566 + }, + { + "epoch": 3.29, + "learning_rate": 8.90530133488029e-06, + "loss": 1.0925, + "step": 44567 + }, + { + "epoch": 3.29, + "learning_rate": 8.90490486130385e-06, + "loss": 0.9485, + "step": 44568 + }, + { + "epoch": 3.29, + "learning_rate": 8.904508389469686e-06, + "loss": 1.0452, + "step": 44569 + }, + { + "epoch": 3.29, + "learning_rate": 8.904111919378435e-06, + "loss": 0.944, + "step": 44570 + }, + { + "epoch": 3.29, + "learning_rate": 8.903715451030723e-06, + "loss": 1.0528, + "step": 44571 + }, + { + "epoch": 3.29, + "learning_rate": 8.903318984427188e-06, + "loss": 0.9709, + "step": 44572 + }, + { + "epoch": 3.29, + "learning_rate": 8.902922519568454e-06, + "loss": 0.9181, + "step": 44573 + }, + { + "epoch": 3.29, + "learning_rate": 8.902526056455157e-06, + "loss": 1.0234, + "step": 44574 + }, + { + "epoch": 3.29, + "learning_rate": 8.902129595087927e-06, + "loss": 1.0958, + "step": 44575 + }, + { + "epoch": 3.29, + "learning_rate": 8.901733135467387e-06, + "loss": 0.9276, + "step": 44576 + }, + { + "epoch": 3.29, + "learning_rate": 8.901336677594178e-06, + "loss": 0.9683, + "step": 44577 + }, + { + "epoch": 3.29, + "learning_rate": 8.900940221468925e-06, + "loss": 1.0506, + "step": 44578 + }, + { + "epoch": 3.29, + "learning_rate": 8.900543767092263e-06, + "loss": 1.1019, + "step": 44579 + }, + { + "epoch": 3.29, + "learning_rate": 8.900147314464817e-06, + "loss": 0.9636, + "step": 44580 + }, + { + "epoch": 3.29, + "learning_rate": 8.899750863587224e-06, + "loss": 0.9896, + "step": 44581 + }, + { + "epoch": 3.29, + "learning_rate": 8.899354414460108e-06, + "loss": 0.9877, + "step": 44582 + }, + { + "epoch": 3.29, + "learning_rate": 8.898957967084107e-06, + "loss": 0.998, + "step": 44583 + }, + { + "epoch": 3.29, + "learning_rate": 8.89856152145985e-06, + "loss": 0.9715, + "step": 44584 + }, + { + "epoch": 3.29, + "learning_rate": 8.898165077587959e-06, + "loss": 0.9673, + "step": 44585 + }, + { + "epoch": 3.29, + "learning_rate": 8.897768635469078e-06, + "loss": 1.1091, + "step": 44586 + }, + { + "epoch": 3.29, + "learning_rate": 8.897372195103825e-06, + "loss": 0.8825, + "step": 44587 + }, + { + "epoch": 3.29, + "learning_rate": 8.896975756492845e-06, + "loss": 1.0635, + "step": 44588 + }, + { + "epoch": 3.29, + "learning_rate": 8.89657931963675e-06, + "loss": 0.8215, + "step": 44589 + }, + { + "epoch": 3.29, + "learning_rate": 8.896182884536192e-06, + "loss": 1.0012, + "step": 44590 + }, + { + "epoch": 3.29, + "learning_rate": 8.895786451191783e-06, + "loss": 1.0327, + "step": 44591 + }, + { + "epoch": 3.29, + "learning_rate": 8.89539001960417e-06, + "loss": 0.9687, + "step": 44592 + }, + { + "epoch": 3.29, + "learning_rate": 8.894993589773969e-06, + "loss": 1.0528, + "step": 44593 + }, + { + "epoch": 3.3, + "learning_rate": 8.894597161701822e-06, + "loss": 1.0254, + "step": 44594 + }, + { + "epoch": 3.3, + "learning_rate": 8.894200735388354e-06, + "loss": 1.041, + "step": 44595 + }, + { + "epoch": 3.3, + "learning_rate": 8.893804310834194e-06, + "loss": 0.9995, + "step": 44596 + }, + { + "epoch": 3.3, + "learning_rate": 8.893407888039978e-06, + "loss": 1.082, + "step": 44597 + }, + { + "epoch": 3.3, + "learning_rate": 8.893011467006333e-06, + "loss": 1.076, + "step": 44598 + }, + { + "epoch": 3.3, + "learning_rate": 8.892615047733892e-06, + "loss": 1.0032, + "step": 44599 + }, + { + "epoch": 3.3, + "learning_rate": 8.892218630223284e-06, + "loss": 0.9979, + "step": 44600 + }, + { + "epoch": 3.3, + "learning_rate": 8.891822214475141e-06, + "loss": 0.9936, + "step": 44601 + }, + { + "epoch": 3.3, + "learning_rate": 8.89142580049009e-06, + "loss": 1.0408, + "step": 44602 + }, + { + "epoch": 3.3, + "learning_rate": 8.891029388268769e-06, + "loss": 0.9376, + "step": 44603 + }, + { + "epoch": 3.3, + "learning_rate": 8.890632977811804e-06, + "loss": 0.9198, + "step": 44604 + }, + { + "epoch": 3.3, + "learning_rate": 8.890236569119823e-06, + "loss": 0.9732, + "step": 44605 + }, + { + "epoch": 3.3, + "learning_rate": 8.889840162193462e-06, + "loss": 1.1515, + "step": 44606 + }, + { + "epoch": 3.3, + "learning_rate": 8.889443757033346e-06, + "loss": 0.9006, + "step": 44607 + }, + { + "epoch": 3.3, + "learning_rate": 8.889047353640114e-06, + "loss": 0.9858, + "step": 44608 + }, + { + "epoch": 3.3, + "learning_rate": 8.888650952014389e-06, + "loss": 1.0715, + "step": 44609 + }, + { + "epoch": 3.3, + "learning_rate": 8.888254552156806e-06, + "loss": 0.9711, + "step": 44610 + }, + { + "epoch": 3.3, + "learning_rate": 8.88785815406799e-06, + "loss": 0.9794, + "step": 44611 + }, + { + "epoch": 3.3, + "learning_rate": 8.887461757748584e-06, + "loss": 1.0142, + "step": 44612 + }, + { + "epoch": 3.3, + "learning_rate": 8.887065363199206e-06, + "loss": 0.9572, + "step": 44613 + }, + { + "epoch": 3.3, + "learning_rate": 8.886668970420489e-06, + "loss": 0.9568, + "step": 44614 + }, + { + "epoch": 3.3, + "learning_rate": 8.886272579413067e-06, + "loss": 0.9835, + "step": 44615 + }, + { + "epoch": 3.3, + "learning_rate": 8.885876190177569e-06, + "loss": 0.9355, + "step": 44616 + }, + { + "epoch": 3.3, + "learning_rate": 8.885479802714627e-06, + "loss": 0.9726, + "step": 44617 + }, + { + "epoch": 3.3, + "learning_rate": 8.885083417024868e-06, + "loss": 0.9877, + "step": 44618 + }, + { + "epoch": 3.3, + "learning_rate": 8.88468703310893e-06, + "loss": 0.9617, + "step": 44619 + }, + { + "epoch": 3.3, + "learning_rate": 8.884290650967436e-06, + "loss": 1.1463, + "step": 44620 + }, + { + "epoch": 3.3, + "learning_rate": 8.883894270601021e-06, + "loss": 1.0659, + "step": 44621 + }, + { + "epoch": 3.3, + "learning_rate": 8.883497892010314e-06, + "loss": 0.9859, + "step": 44622 + }, + { + "epoch": 3.3, + "learning_rate": 8.883101515195944e-06, + "loss": 0.919, + "step": 44623 + }, + { + "epoch": 3.3, + "learning_rate": 8.882705140158544e-06, + "loss": 0.9082, + "step": 44624 + }, + { + "epoch": 3.3, + "learning_rate": 8.882308766898745e-06, + "loss": 1.0925, + "step": 44625 + }, + { + "epoch": 3.3, + "learning_rate": 8.881912395417179e-06, + "loss": 1.0679, + "step": 44626 + }, + { + "epoch": 3.3, + "learning_rate": 8.88151602571447e-06, + "loss": 1.0192, + "step": 44627 + }, + { + "epoch": 3.3, + "learning_rate": 8.881119657791256e-06, + "loss": 0.9648, + "step": 44628 + }, + { + "epoch": 3.3, + "learning_rate": 8.880723291648163e-06, + "loss": 0.9894, + "step": 44629 + }, + { + "epoch": 3.3, + "learning_rate": 8.880326927285824e-06, + "loss": 0.945, + "step": 44630 + }, + { + "epoch": 3.3, + "learning_rate": 8.879930564704872e-06, + "loss": 0.9654, + "step": 44631 + }, + { + "epoch": 3.3, + "learning_rate": 8.879534203905928e-06, + "loss": 0.9998, + "step": 44632 + }, + { + "epoch": 3.3, + "learning_rate": 8.879137844889634e-06, + "loss": 0.9817, + "step": 44633 + }, + { + "epoch": 3.3, + "learning_rate": 8.878741487656611e-06, + "loss": 0.9483, + "step": 44634 + }, + { + "epoch": 3.3, + "learning_rate": 8.8783451322075e-06, + "loss": 1.0024, + "step": 44635 + }, + { + "epoch": 3.3, + "learning_rate": 8.87794877854292e-06, + "loss": 1.0684, + "step": 44636 + }, + { + "epoch": 3.3, + "learning_rate": 8.877552426663514e-06, + "loss": 1.1363, + "step": 44637 + }, + { + "epoch": 3.3, + "learning_rate": 8.8771560765699e-06, + "loss": 0.9586, + "step": 44638 + }, + { + "epoch": 3.3, + "learning_rate": 8.87675972826272e-06, + "loss": 0.8969, + "step": 44639 + }, + { + "epoch": 3.3, + "learning_rate": 8.876363381742597e-06, + "loss": 1.0671, + "step": 44640 + }, + { + "epoch": 3.3, + "learning_rate": 8.875967037010162e-06, + "loss": 1.0075, + "step": 44641 + }, + { + "epoch": 3.3, + "learning_rate": 8.875570694066049e-06, + "loss": 1.0218, + "step": 44642 + }, + { + "epoch": 3.3, + "learning_rate": 8.875174352910885e-06, + "loss": 1.0142, + "step": 44643 + }, + { + "epoch": 3.3, + "learning_rate": 8.874778013545304e-06, + "loss": 0.9594, + "step": 44644 + }, + { + "epoch": 3.3, + "learning_rate": 8.874381675969935e-06, + "loss": 0.8386, + "step": 44645 + }, + { + "epoch": 3.3, + "learning_rate": 8.873985340185409e-06, + "loss": 0.957, + "step": 44646 + }, + { + "epoch": 3.3, + "learning_rate": 8.873589006192355e-06, + "loss": 0.892, + "step": 44647 + }, + { + "epoch": 3.3, + "learning_rate": 8.873192673991407e-06, + "loss": 0.8986, + "step": 44648 + }, + { + "epoch": 3.3, + "learning_rate": 8.87279634358319e-06, + "loss": 0.9668, + "step": 44649 + }, + { + "epoch": 3.3, + "learning_rate": 8.872400014968341e-06, + "loss": 1.0084, + "step": 44650 + }, + { + "epoch": 3.3, + "learning_rate": 8.872003688147487e-06, + "loss": 0.8901, + "step": 44651 + }, + { + "epoch": 3.3, + "learning_rate": 8.871607363121257e-06, + "loss": 0.9616, + "step": 44652 + }, + { + "epoch": 3.3, + "learning_rate": 8.871211039890286e-06, + "loss": 1.0491, + "step": 44653 + }, + { + "epoch": 3.3, + "learning_rate": 8.870814718455199e-06, + "loss": 0.938, + "step": 44654 + }, + { + "epoch": 3.3, + "learning_rate": 8.870418398816632e-06, + "loss": 1.128, + "step": 44655 + }, + { + "epoch": 3.3, + "learning_rate": 8.870022080975213e-06, + "loss": 0.9415, + "step": 44656 + }, + { + "epoch": 3.3, + "learning_rate": 8.869625764931576e-06, + "loss": 1.0279, + "step": 44657 + }, + { + "epoch": 3.3, + "learning_rate": 8.86922945068634e-06, + "loss": 0.9417, + "step": 44658 + }, + { + "epoch": 3.3, + "learning_rate": 8.868833138240152e-06, + "loss": 1.0021, + "step": 44659 + }, + { + "epoch": 3.3, + "learning_rate": 8.868436827593632e-06, + "loss": 1.0425, + "step": 44660 + }, + { + "epoch": 3.3, + "learning_rate": 8.86804051874741e-06, + "loss": 0.9882, + "step": 44661 + }, + { + "epoch": 3.3, + "learning_rate": 8.86764421170212e-06, + "loss": 1.0401, + "step": 44662 + }, + { + "epoch": 3.3, + "learning_rate": 8.867247906458393e-06, + "loss": 0.9738, + "step": 44663 + }, + { + "epoch": 3.3, + "learning_rate": 8.866851603016859e-06, + "loss": 0.9281, + "step": 44664 + }, + { + "epoch": 3.3, + "learning_rate": 8.866455301378145e-06, + "loss": 1.1535, + "step": 44665 + }, + { + "epoch": 3.3, + "learning_rate": 8.866059001542887e-06, + "loss": 1.0369, + "step": 44666 + }, + { + "epoch": 3.3, + "learning_rate": 8.865662703511712e-06, + "loss": 0.9356, + "step": 44667 + }, + { + "epoch": 3.3, + "learning_rate": 8.865266407285252e-06, + "loss": 1.0068, + "step": 44668 + }, + { + "epoch": 3.3, + "learning_rate": 8.864870112864138e-06, + "loss": 0.9239, + "step": 44669 + }, + { + "epoch": 3.3, + "learning_rate": 8.864473820248997e-06, + "loss": 1.0309, + "step": 44670 + }, + { + "epoch": 3.3, + "learning_rate": 8.864077529440464e-06, + "loss": 0.9394, + "step": 44671 + }, + { + "epoch": 3.3, + "learning_rate": 8.863681240439165e-06, + "loss": 1.1435, + "step": 44672 + }, + { + "epoch": 3.3, + "learning_rate": 8.863284953245734e-06, + "loss": 0.933, + "step": 44673 + }, + { + "epoch": 3.3, + "learning_rate": 8.8628886678608e-06, + "loss": 0.9341, + "step": 44674 + }, + { + "epoch": 3.3, + "learning_rate": 8.862492384284996e-06, + "loss": 1.0233, + "step": 44675 + }, + { + "epoch": 3.3, + "learning_rate": 8.862096102518947e-06, + "loss": 1.0685, + "step": 44676 + }, + { + "epoch": 3.3, + "learning_rate": 8.86169982256329e-06, + "loss": 0.9827, + "step": 44677 + }, + { + "epoch": 3.3, + "learning_rate": 8.861303544418654e-06, + "loss": 0.9989, + "step": 44678 + }, + { + "epoch": 3.3, + "learning_rate": 8.86090726808566e-06, + "loss": 0.9563, + "step": 44679 + }, + { + "epoch": 3.3, + "learning_rate": 8.860510993564955e-06, + "loss": 1.091, + "step": 44680 + }, + { + "epoch": 3.3, + "learning_rate": 8.860114720857153e-06, + "loss": 0.8954, + "step": 44681 + }, + { + "epoch": 3.3, + "learning_rate": 8.859718449962899e-06, + "loss": 1.1127, + "step": 44682 + }, + { + "epoch": 3.3, + "learning_rate": 8.859322180882809e-06, + "loss": 0.9719, + "step": 44683 + }, + { + "epoch": 3.3, + "learning_rate": 8.858925913617529e-06, + "loss": 0.9948, + "step": 44684 + }, + { + "epoch": 3.3, + "learning_rate": 8.858529648167674e-06, + "loss": 0.9943, + "step": 44685 + }, + { + "epoch": 3.3, + "learning_rate": 8.85813338453389e-06, + "loss": 0.9757, + "step": 44686 + }, + { + "epoch": 3.3, + "learning_rate": 8.857737122716796e-06, + "loss": 0.9896, + "step": 44687 + }, + { + "epoch": 3.3, + "learning_rate": 8.857340862717024e-06, + "loss": 0.9963, + "step": 44688 + }, + { + "epoch": 3.3, + "learning_rate": 8.856944604535209e-06, + "loss": 1.0482, + "step": 44689 + }, + { + "epoch": 3.3, + "learning_rate": 8.856548348171977e-06, + "loss": 1.0361, + "step": 44690 + }, + { + "epoch": 3.3, + "learning_rate": 8.856152093627961e-06, + "loss": 1.1006, + "step": 44691 + }, + { + "epoch": 3.3, + "learning_rate": 8.855755840903789e-06, + "loss": 1.0698, + "step": 44692 + }, + { + "epoch": 3.3, + "learning_rate": 8.855359590000096e-06, + "loss": 1.0722, + "step": 44693 + }, + { + "epoch": 3.3, + "learning_rate": 8.854963340917506e-06, + "loss": 1.0188, + "step": 44694 + }, + { + "epoch": 3.3, + "learning_rate": 8.854567093656656e-06, + "loss": 0.9946, + "step": 44695 + }, + { + "epoch": 3.3, + "learning_rate": 8.85417084821817e-06, + "loss": 1.055, + "step": 44696 + }, + { + "epoch": 3.3, + "learning_rate": 8.853774604602686e-06, + "loss": 0.9268, + "step": 44697 + }, + { + "epoch": 3.3, + "learning_rate": 8.853378362810829e-06, + "loss": 1.0359, + "step": 44698 + }, + { + "epoch": 3.3, + "learning_rate": 8.852982122843229e-06, + "loss": 0.9549, + "step": 44699 + }, + { + "epoch": 3.3, + "learning_rate": 8.852585884700519e-06, + "loss": 0.956, + "step": 44700 + }, + { + "epoch": 3.3, + "learning_rate": 8.852189648383328e-06, + "loss": 0.8468, + "step": 44701 + }, + { + "epoch": 3.3, + "learning_rate": 8.85179341389229e-06, + "loss": 1.0319, + "step": 44702 + }, + { + "epoch": 3.3, + "learning_rate": 8.851397181228025e-06, + "loss": 0.8964, + "step": 44703 + }, + { + "epoch": 3.3, + "learning_rate": 8.85100095039118e-06, + "loss": 1.0475, + "step": 44704 + }, + { + "epoch": 3.3, + "learning_rate": 8.850604721382366e-06, + "loss": 0.9932, + "step": 44705 + }, + { + "epoch": 3.3, + "learning_rate": 8.850208494202233e-06, + "loss": 1.0192, + "step": 44706 + }, + { + "epoch": 3.3, + "learning_rate": 8.849812268851397e-06, + "loss": 0.9684, + "step": 44707 + }, + { + "epoch": 3.3, + "learning_rate": 8.849416045330494e-06, + "loss": 1.0071, + "step": 44708 + }, + { + "epoch": 3.3, + "learning_rate": 8.849019823640154e-06, + "loss": 0.9855, + "step": 44709 + }, + { + "epoch": 3.3, + "learning_rate": 8.848623603781005e-06, + "loss": 0.9459, + "step": 44710 + }, + { + "epoch": 3.3, + "learning_rate": 8.84822738575368e-06, + "loss": 1.0454, + "step": 44711 + }, + { + "epoch": 3.3, + "learning_rate": 8.84783116955881e-06, + "loss": 0.9388, + "step": 44712 + }, + { + "epoch": 3.3, + "learning_rate": 8.847434955197022e-06, + "loss": 0.9064, + "step": 44713 + }, + { + "epoch": 3.3, + "learning_rate": 8.84703874266895e-06, + "loss": 0.9543, + "step": 44714 + }, + { + "epoch": 3.3, + "learning_rate": 8.846642531975223e-06, + "loss": 0.9309, + "step": 44715 + }, + { + "epoch": 3.3, + "learning_rate": 8.846246323116473e-06, + "loss": 0.9667, + "step": 44716 + }, + { + "epoch": 3.3, + "learning_rate": 8.845850116093325e-06, + "loss": 1.0313, + "step": 44717 + }, + { + "epoch": 3.3, + "learning_rate": 8.845453910906416e-06, + "loss": 1.061, + "step": 44718 + }, + { + "epoch": 3.3, + "learning_rate": 8.84505770755637e-06, + "loss": 1.021, + "step": 44719 + }, + { + "epoch": 3.3, + "learning_rate": 8.844661506043824e-06, + "loss": 1.0599, + "step": 44720 + }, + { + "epoch": 3.3, + "learning_rate": 8.844265306369403e-06, + "loss": 0.9731, + "step": 44721 + }, + { + "epoch": 3.3, + "learning_rate": 8.84386910853374e-06, + "loss": 1.0469, + "step": 44722 + }, + { + "epoch": 3.3, + "learning_rate": 8.843472912537464e-06, + "loss": 0.9245, + "step": 44723 + }, + { + "epoch": 3.3, + "learning_rate": 8.843076718381207e-06, + "loss": 0.9738, + "step": 44724 + }, + { + "epoch": 3.3, + "learning_rate": 8.842680526065602e-06, + "loss": 1.0705, + "step": 44725 + }, + { + "epoch": 3.3, + "learning_rate": 8.842284335591268e-06, + "loss": 1.0132, + "step": 44726 + }, + { + "epoch": 3.3, + "learning_rate": 8.84188814695885e-06, + "loss": 1.0215, + "step": 44727 + }, + { + "epoch": 3.3, + "learning_rate": 8.841491960168965e-06, + "loss": 1.1285, + "step": 44728 + }, + { + "epoch": 3.3, + "learning_rate": 8.841095775222257e-06, + "loss": 0.9758, + "step": 44729 + }, + { + "epoch": 3.31, + "learning_rate": 8.840699592119341e-06, + "loss": 0.9268, + "step": 44730 + }, + { + "epoch": 3.31, + "learning_rate": 8.840303410860862e-06, + "loss": 1.0385, + "step": 44731 + }, + { + "epoch": 3.31, + "learning_rate": 8.839907231447437e-06, + "loss": 1.1813, + "step": 44732 + }, + { + "epoch": 3.31, + "learning_rate": 8.83951105387971e-06, + "loss": 1.034, + "step": 44733 + }, + { + "epoch": 3.31, + "learning_rate": 8.839114878158304e-06, + "loss": 1.0346, + "step": 44734 + }, + { + "epoch": 3.31, + "learning_rate": 8.838718704283845e-06, + "loss": 1.0894, + "step": 44735 + }, + { + "epoch": 3.31, + "learning_rate": 8.838322532256971e-06, + "loss": 0.977, + "step": 44736 + }, + { + "epoch": 3.31, + "learning_rate": 8.837926362078308e-06, + "loss": 1.1125, + "step": 44737 + }, + { + "epoch": 3.31, + "learning_rate": 8.837530193748489e-06, + "loss": 0.9883, + "step": 44738 + }, + { + "epoch": 3.31, + "learning_rate": 8.837134027268139e-06, + "loss": 0.9554, + "step": 44739 + }, + { + "epoch": 3.31, + "learning_rate": 8.836737862637895e-06, + "loss": 0.9686, + "step": 44740 + }, + { + "epoch": 3.31, + "learning_rate": 8.836341699858383e-06, + "loss": 1.091, + "step": 44741 + }, + { + "epoch": 3.31, + "learning_rate": 8.835945538930238e-06, + "loss": 1.0432, + "step": 44742 + }, + { + "epoch": 3.31, + "learning_rate": 8.835549379854085e-06, + "loss": 0.9895, + "step": 44743 + }, + { + "epoch": 3.31, + "learning_rate": 8.835153222630555e-06, + "loss": 0.9792, + "step": 44744 + }, + { + "epoch": 3.31, + "learning_rate": 8.834757067260284e-06, + "loss": 1.0101, + "step": 44745 + }, + { + "epoch": 3.31, + "learning_rate": 8.83436091374389e-06, + "loss": 1.0219, + "step": 44746 + }, + { + "epoch": 3.31, + "learning_rate": 8.83396476208202e-06, + "loss": 0.9766, + "step": 44747 + }, + { + "epoch": 3.31, + "learning_rate": 8.833568612275286e-06, + "loss": 0.9735, + "step": 44748 + }, + { + "epoch": 3.31, + "learning_rate": 8.833172464324336e-06, + "loss": 0.9146, + "step": 44749 + }, + { + "epoch": 3.31, + "learning_rate": 8.832776318229786e-06, + "loss": 1.0031, + "step": 44750 + }, + { + "epoch": 3.31, + "learning_rate": 8.832380173992277e-06, + "loss": 1.0658, + "step": 44751 + }, + { + "epoch": 3.31, + "learning_rate": 8.831984031612429e-06, + "loss": 0.9599, + "step": 44752 + }, + { + "epoch": 3.31, + "learning_rate": 8.831587891090883e-06, + "loss": 0.9305, + "step": 44753 + }, + { + "epoch": 3.31, + "learning_rate": 8.831191752428262e-06, + "loss": 1.0043, + "step": 44754 + }, + { + "epoch": 3.31, + "learning_rate": 8.830795615625198e-06, + "loss": 1.0356, + "step": 44755 + }, + { + "epoch": 3.31, + "learning_rate": 8.830399480682322e-06, + "loss": 0.8478, + "step": 44756 + }, + { + "epoch": 3.31, + "learning_rate": 8.830003347600261e-06, + "loss": 0.9792, + "step": 44757 + }, + { + "epoch": 3.31, + "learning_rate": 8.829607216379652e-06, + "loss": 0.9702, + "step": 44758 + }, + { + "epoch": 3.31, + "learning_rate": 8.829211087021118e-06, + "loss": 1.0374, + "step": 44759 + }, + { + "epoch": 3.31, + "learning_rate": 8.828814959525295e-06, + "loss": 0.9244, + "step": 44760 + }, + { + "epoch": 3.31, + "learning_rate": 8.828418833892808e-06, + "loss": 0.994, + "step": 44761 + }, + { + "epoch": 3.31, + "learning_rate": 8.828022710124292e-06, + "loss": 0.9921, + "step": 44762 + }, + { + "epoch": 3.31, + "learning_rate": 8.827626588220375e-06, + "loss": 0.9476, + "step": 44763 + }, + { + "epoch": 3.31, + "learning_rate": 8.827230468181686e-06, + "loss": 0.9917, + "step": 44764 + }, + { + "epoch": 3.31, + "learning_rate": 8.826834350008857e-06, + "loss": 0.8941, + "step": 44765 + }, + { + "epoch": 3.31, + "learning_rate": 8.826438233702517e-06, + "loss": 1.0592, + "step": 44766 + }, + { + "epoch": 3.31, + "learning_rate": 8.826042119263298e-06, + "loss": 1.0838, + "step": 44767 + }, + { + "epoch": 3.31, + "learning_rate": 8.825646006691828e-06, + "loss": 0.9268, + "step": 44768 + }, + { + "epoch": 3.31, + "learning_rate": 8.825249895988742e-06, + "loss": 1.0384, + "step": 44769 + }, + { + "epoch": 3.31, + "learning_rate": 8.82485378715466e-06, + "loss": 1.0046, + "step": 44770 + }, + { + "epoch": 3.31, + "learning_rate": 8.824457680190226e-06, + "loss": 0.9293, + "step": 44771 + }, + { + "epoch": 3.31, + "learning_rate": 8.824061575096063e-06, + "loss": 1.141, + "step": 44772 + }, + { + "epoch": 3.31, + "learning_rate": 8.823665471872794e-06, + "loss": 0.9286, + "step": 44773 + }, + { + "epoch": 3.31, + "learning_rate": 8.823269370521064e-06, + "loss": 0.9335, + "step": 44774 + }, + { + "epoch": 3.31, + "learning_rate": 8.822873271041488e-06, + "loss": 0.9366, + "step": 44775 + }, + { + "epoch": 3.31, + "learning_rate": 8.822477173434712e-06, + "loss": 0.9598, + "step": 44776 + }, + { + "epoch": 3.31, + "learning_rate": 8.82208107770135e-06, + "loss": 1.0804, + "step": 44777 + }, + { + "epoch": 3.31, + "learning_rate": 8.821684983842046e-06, + "loss": 1.1216, + "step": 44778 + }, + { + "epoch": 3.31, + "learning_rate": 8.821288891857421e-06, + "loss": 1.0106, + "step": 44779 + }, + { + "epoch": 3.31, + "learning_rate": 8.820892801748112e-06, + "loss": 0.9417, + "step": 44780 + }, + { + "epoch": 3.31, + "learning_rate": 8.820496713514743e-06, + "loss": 0.948, + "step": 44781 + }, + { + "epoch": 3.31, + "learning_rate": 8.820100627157947e-06, + "loss": 1.0444, + "step": 44782 + }, + { + "epoch": 3.31, + "learning_rate": 8.819704542678355e-06, + "loss": 0.8592, + "step": 44783 + }, + { + "epoch": 3.31, + "learning_rate": 8.819308460076595e-06, + "loss": 0.9425, + "step": 44784 + }, + { + "epoch": 3.31, + "learning_rate": 8.8189123793533e-06, + "loss": 0.9495, + "step": 44785 + }, + { + "epoch": 3.31, + "learning_rate": 8.818516300509096e-06, + "loss": 1.0394, + "step": 44786 + }, + { + "epoch": 3.31, + "learning_rate": 8.818120223544616e-06, + "loss": 0.9377, + "step": 44787 + }, + { + "epoch": 3.31, + "learning_rate": 8.817724148460492e-06, + "loss": 0.9794, + "step": 44788 + }, + { + "epoch": 3.31, + "learning_rate": 8.81732807525735e-06, + "loss": 0.9619, + "step": 44789 + }, + { + "epoch": 3.31, + "learning_rate": 8.816932003935828e-06, + "loss": 1.0549, + "step": 44790 + }, + { + "epoch": 3.31, + "learning_rate": 8.816535934496541e-06, + "loss": 0.9483, + "step": 44791 + }, + { + "epoch": 3.31, + "learning_rate": 8.816139866940136e-06, + "loss": 0.982, + "step": 44792 + }, + { + "epoch": 3.31, + "learning_rate": 8.815743801267229e-06, + "loss": 0.9441, + "step": 44793 + }, + { + "epoch": 3.31, + "learning_rate": 8.815347737478462e-06, + "loss": 1.0377, + "step": 44794 + }, + { + "epoch": 3.31, + "learning_rate": 8.814951675574455e-06, + "loss": 0.9557, + "step": 44795 + }, + { + "epoch": 3.31, + "learning_rate": 8.814555615555848e-06, + "loss": 1.0878, + "step": 44796 + }, + { + "epoch": 3.31, + "learning_rate": 8.814159557423259e-06, + "loss": 0.9614, + "step": 44797 + }, + { + "epoch": 3.31, + "learning_rate": 8.813763501177334e-06, + "loss": 1.1327, + "step": 44798 + }, + { + "epoch": 3.31, + "learning_rate": 8.813367446818691e-06, + "loss": 0.9967, + "step": 44799 + }, + { + "epoch": 3.31, + "learning_rate": 8.812971394347961e-06, + "loss": 1.1241, + "step": 44800 + }, + { + "epoch": 3.31, + "learning_rate": 8.81257534376578e-06, + "loss": 1.0487, + "step": 44801 + }, + { + "epoch": 3.31, + "learning_rate": 8.81217929507277e-06, + "loss": 0.9749, + "step": 44802 + }, + { + "epoch": 3.31, + "learning_rate": 8.81178324826957e-06, + "loss": 1.0592, + "step": 44803 + }, + { + "epoch": 3.31, + "learning_rate": 8.811387203356803e-06, + "loss": 1.0057, + "step": 44804 + }, + { + "epoch": 3.31, + "learning_rate": 8.810991160335103e-06, + "loss": 0.9828, + "step": 44805 + }, + { + "epoch": 3.31, + "learning_rate": 8.8105951192051e-06, + "loss": 0.956, + "step": 44806 + }, + { + "epoch": 3.31, + "learning_rate": 8.810199079967421e-06, + "loss": 1.1007, + "step": 44807 + }, + { + "epoch": 3.31, + "learning_rate": 8.8098030426227e-06, + "loss": 1.0193, + "step": 44808 + }, + { + "epoch": 3.31, + "learning_rate": 8.809407007171565e-06, + "loss": 0.925, + "step": 44809 + }, + { + "epoch": 3.31, + "learning_rate": 8.809010973614648e-06, + "loss": 0.9661, + "step": 44810 + }, + { + "epoch": 3.31, + "learning_rate": 8.808614941952575e-06, + "loss": 1.0966, + "step": 44811 + }, + { + "epoch": 3.31, + "learning_rate": 8.80821891218598e-06, + "loss": 0.8762, + "step": 44812 + }, + { + "epoch": 3.31, + "learning_rate": 8.80782288431549e-06, + "loss": 0.9162, + "step": 44813 + }, + { + "epoch": 3.31, + "learning_rate": 8.80742685834174e-06, + "loss": 0.9574, + "step": 44814 + }, + { + "epoch": 3.31, + "learning_rate": 8.807030834265353e-06, + "loss": 1.0603, + "step": 44815 + }, + { + "epoch": 3.31, + "learning_rate": 8.806634812086964e-06, + "loss": 1.0587, + "step": 44816 + }, + { + "epoch": 3.31, + "learning_rate": 8.806238791807202e-06, + "loss": 0.9823, + "step": 44817 + }, + { + "epoch": 3.31, + "learning_rate": 8.805842773426698e-06, + "loss": 0.9202, + "step": 44818 + }, + { + "epoch": 3.31, + "learning_rate": 8.805446756946085e-06, + "loss": 1.0477, + "step": 44819 + }, + { + "epoch": 3.31, + "learning_rate": 8.80505074236598e-06, + "loss": 1.0253, + "step": 44820 + }, + { + "epoch": 3.31, + "learning_rate": 8.80465472968703e-06, + "loss": 0.9514, + "step": 44821 + }, + { + "epoch": 3.31, + "learning_rate": 8.80425871890985e-06, + "loss": 1.0363, + "step": 44822 + }, + { + "epoch": 3.31, + "learning_rate": 8.803862710035085e-06, + "loss": 1.0617, + "step": 44823 + }, + { + "epoch": 3.31, + "learning_rate": 8.803466703063352e-06, + "loss": 1.0585, + "step": 44824 + }, + { + "epoch": 3.31, + "learning_rate": 8.80307069799529e-06, + "loss": 1.0243, + "step": 44825 + }, + { + "epoch": 3.31, + "learning_rate": 8.802674694831522e-06, + "loss": 0.9575, + "step": 44826 + }, + { + "epoch": 3.31, + "learning_rate": 8.802278693572683e-06, + "loss": 1.0734, + "step": 44827 + }, + { + "epoch": 3.31, + "learning_rate": 8.801882694219403e-06, + "loss": 1.0561, + "step": 44828 + }, + { + "epoch": 3.31, + "learning_rate": 8.801486696772306e-06, + "loss": 0.8931, + "step": 44829 + }, + { + "epoch": 3.31, + "learning_rate": 8.801090701232031e-06, + "loss": 1.0424, + "step": 44830 + }, + { + "epoch": 3.31, + "learning_rate": 8.8006947075992e-06, + "loss": 1.0006, + "step": 44831 + }, + { + "epoch": 3.31, + "learning_rate": 8.80029871587445e-06, + "loss": 1.0746, + "step": 44832 + }, + { + "epoch": 3.31, + "learning_rate": 8.799902726058405e-06, + "loss": 0.9524, + "step": 44833 + }, + { + "epoch": 3.31, + "learning_rate": 8.7995067381517e-06, + "loss": 0.9488, + "step": 44834 + }, + { + "epoch": 3.31, + "learning_rate": 8.79911075215496e-06, + "loss": 1.0829, + "step": 44835 + }, + { + "epoch": 3.31, + "learning_rate": 8.798714768068818e-06, + "loss": 1.0365, + "step": 44836 + }, + { + "epoch": 3.31, + "learning_rate": 8.798318785893909e-06, + "loss": 0.9788, + "step": 44837 + }, + { + "epoch": 3.31, + "learning_rate": 8.79792280563085e-06, + "loss": 1.0179, + "step": 44838 + }, + { + "epoch": 3.31, + "learning_rate": 8.797526827280284e-06, + "loss": 0.9564, + "step": 44839 + }, + { + "epoch": 3.31, + "learning_rate": 8.79713085084283e-06, + "loss": 1.0104, + "step": 44840 + }, + { + "epoch": 3.31, + "learning_rate": 8.79673487631913e-06, + "loss": 0.9247, + "step": 44841 + }, + { + "epoch": 3.31, + "learning_rate": 8.7963389037098e-06, + "loss": 0.9903, + "step": 44842 + }, + { + "epoch": 3.31, + "learning_rate": 8.795942933015485e-06, + "loss": 1.162, + "step": 44843 + }, + { + "epoch": 3.31, + "learning_rate": 8.795546964236801e-06, + "loss": 1.0527, + "step": 44844 + }, + { + "epoch": 3.31, + "learning_rate": 8.795150997374392e-06, + "loss": 1.0744, + "step": 44845 + }, + { + "epoch": 3.31, + "learning_rate": 8.794755032428878e-06, + "loss": 0.984, + "step": 44846 + }, + { + "epoch": 3.31, + "learning_rate": 8.794359069400889e-06, + "loss": 1.1059, + "step": 44847 + }, + { + "epoch": 3.31, + "learning_rate": 8.79396310829106e-06, + "loss": 1.0784, + "step": 44848 + }, + { + "epoch": 3.31, + "learning_rate": 8.793567149100014e-06, + "loss": 1.0311, + "step": 44849 + }, + { + "epoch": 3.31, + "learning_rate": 8.79317119182839e-06, + "loss": 0.8558, + "step": 44850 + }, + { + "epoch": 3.31, + "learning_rate": 8.792775236476811e-06, + "loss": 0.9887, + "step": 44851 + }, + { + "epoch": 3.31, + "learning_rate": 8.79237928304591e-06, + "loss": 0.9694, + "step": 44852 + }, + { + "epoch": 3.31, + "learning_rate": 8.791983331536317e-06, + "loss": 0.9739, + "step": 44853 + }, + { + "epoch": 3.31, + "learning_rate": 8.79158738194866e-06, + "loss": 1.0313, + "step": 44854 + }, + { + "epoch": 3.31, + "learning_rate": 8.791191434283572e-06, + "loss": 1.0578, + "step": 44855 + }, + { + "epoch": 3.31, + "learning_rate": 8.79079548854168e-06, + "loss": 0.9759, + "step": 44856 + }, + { + "epoch": 3.31, + "learning_rate": 8.790399544723616e-06, + "loss": 0.9606, + "step": 44857 + }, + { + "epoch": 3.31, + "learning_rate": 8.790003602830007e-06, + "loss": 1.0058, + "step": 44858 + }, + { + "epoch": 3.31, + "learning_rate": 8.789607662861486e-06, + "loss": 0.9676, + "step": 44859 + }, + { + "epoch": 3.31, + "learning_rate": 8.789211724818682e-06, + "loss": 0.9438, + "step": 44860 + }, + { + "epoch": 3.31, + "learning_rate": 8.788815788702226e-06, + "loss": 1.1025, + "step": 44861 + }, + { + "epoch": 3.31, + "learning_rate": 8.788419854512744e-06, + "loss": 1.0974, + "step": 44862 + }, + { + "epoch": 3.31, + "learning_rate": 8.78802392225087e-06, + "loss": 1.0058, + "step": 44863 + }, + { + "epoch": 3.31, + "learning_rate": 8.78762799191723e-06, + "loss": 1.0014, + "step": 44864 + }, + { + "epoch": 3.32, + "learning_rate": 8.787232063512462e-06, + "loss": 0.9201, + "step": 44865 + }, + { + "epoch": 3.32, + "learning_rate": 8.786836137037192e-06, + "loss": 0.8508, + "step": 44866 + }, + { + "epoch": 3.32, + "learning_rate": 8.786440212492042e-06, + "loss": 0.8586, + "step": 44867 + }, + { + "epoch": 3.32, + "learning_rate": 8.786044289877652e-06, + "loss": 0.9868, + "step": 44868 + }, + { + "epoch": 3.32, + "learning_rate": 8.785648369194645e-06, + "loss": 1.1221, + "step": 44869 + }, + { + "epoch": 3.32, + "learning_rate": 8.785252450443657e-06, + "loss": 1.0198, + "step": 44870 + }, + { + "epoch": 3.32, + "learning_rate": 8.784856533625311e-06, + "loss": 1.054, + "step": 44871 + }, + { + "epoch": 3.32, + "learning_rate": 8.784460618740245e-06, + "loss": 0.9617, + "step": 44872 + }, + { + "epoch": 3.32, + "learning_rate": 8.784064705789081e-06, + "loss": 0.9624, + "step": 44873 + }, + { + "epoch": 3.32, + "learning_rate": 8.783668794772457e-06, + "loss": 1.0302, + "step": 44874 + }, + { + "epoch": 3.32, + "learning_rate": 8.783272885690997e-06, + "loss": 1.0828, + "step": 44875 + }, + { + "epoch": 3.32, + "learning_rate": 8.78287697854533e-06, + "loss": 1.0138, + "step": 44876 + }, + { + "epoch": 3.32, + "learning_rate": 8.782481073336091e-06, + "loss": 1.0723, + "step": 44877 + }, + { + "epoch": 3.32, + "learning_rate": 8.782085170063905e-06, + "loss": 1.0586, + "step": 44878 + }, + { + "epoch": 3.32, + "learning_rate": 8.781689268729408e-06, + "loss": 1.075, + "step": 44879 + }, + { + "epoch": 3.32, + "learning_rate": 8.781293369333222e-06, + "loss": 1.0042, + "step": 44880 + }, + { + "epoch": 3.32, + "learning_rate": 8.780897471875983e-06, + "loss": 1.0705, + "step": 44881 + }, + { + "epoch": 3.32, + "learning_rate": 8.780501576358315e-06, + "loss": 1.0727, + "step": 44882 + }, + { + "epoch": 3.32, + "learning_rate": 8.780105682780856e-06, + "loss": 1.0018, + "step": 44883 + }, + { + "epoch": 3.32, + "learning_rate": 8.779709791144232e-06, + "loss": 1.0721, + "step": 44884 + }, + { + "epoch": 3.32, + "learning_rate": 8.779313901449068e-06, + "loss": 0.9921, + "step": 44885 + }, + { + "epoch": 3.32, + "learning_rate": 8.778918013696004e-06, + "loss": 0.9073, + "step": 44886 + }, + { + "epoch": 3.32, + "learning_rate": 8.778522127885655e-06, + "loss": 0.8959, + "step": 44887 + }, + { + "epoch": 3.32, + "learning_rate": 8.778126244018668e-06, + "loss": 1.0598, + "step": 44888 + }, + { + "epoch": 3.32, + "learning_rate": 8.77773036209566e-06, + "loss": 0.9693, + "step": 44889 + }, + { + "epoch": 3.32, + "learning_rate": 8.777334482117268e-06, + "loss": 1.0029, + "step": 44890 + }, + { + "epoch": 3.32, + "learning_rate": 8.776938604084116e-06, + "loss": 0.9141, + "step": 44891 + }, + { + "epoch": 3.32, + "learning_rate": 8.776542727996844e-06, + "loss": 1.0196, + "step": 44892 + }, + { + "epoch": 3.32, + "learning_rate": 8.77614685385607e-06, + "loss": 0.9408, + "step": 44893 + }, + { + "epoch": 3.32, + "learning_rate": 8.775750981662429e-06, + "loss": 0.9844, + "step": 44894 + }, + { + "epoch": 3.32, + "learning_rate": 8.775355111416553e-06, + "loss": 1.0179, + "step": 44895 + }, + { + "epoch": 3.32, + "learning_rate": 8.774959243119065e-06, + "loss": 0.9529, + "step": 44896 + }, + { + "epoch": 3.32, + "learning_rate": 8.774563376770603e-06, + "loss": 1.0797, + "step": 44897 + }, + { + "epoch": 3.32, + "learning_rate": 8.774167512371791e-06, + "loss": 1.0764, + "step": 44898 + }, + { + "epoch": 3.32, + "learning_rate": 8.773771649923262e-06, + "loss": 0.9203, + "step": 44899 + }, + { + "epoch": 3.32, + "learning_rate": 8.773375789425645e-06, + "loss": 0.9231, + "step": 44900 + }, + { + "epoch": 3.32, + "learning_rate": 8.772979930879571e-06, + "loss": 1.1114, + "step": 44901 + }, + { + "epoch": 3.32, + "learning_rate": 8.772584074285668e-06, + "loss": 1.0195, + "step": 44902 + }, + { + "epoch": 3.32, + "learning_rate": 8.772188219644563e-06, + "loss": 1.0208, + "step": 44903 + }, + { + "epoch": 3.32, + "learning_rate": 8.771792366956891e-06, + "loss": 1.1081, + "step": 44904 + }, + { + "epoch": 3.32, + "learning_rate": 8.771396516223279e-06, + "loss": 1.0148, + "step": 44905 + }, + { + "epoch": 3.32, + "learning_rate": 8.771000667444359e-06, + "loss": 1.0198, + "step": 44906 + }, + { + "epoch": 3.32, + "learning_rate": 8.770604820620757e-06, + "loss": 0.9415, + "step": 44907 + }, + { + "epoch": 3.32, + "learning_rate": 8.770208975753108e-06, + "loss": 1.0929, + "step": 44908 + }, + { + "epoch": 3.32, + "learning_rate": 8.769813132842036e-06, + "loss": 1.0497, + "step": 44909 + }, + { + "epoch": 3.32, + "learning_rate": 8.769417291888176e-06, + "loss": 0.9025, + "step": 44910 + }, + { + "epoch": 3.32, + "learning_rate": 8.769021452892155e-06, + "loss": 1.0396, + "step": 44911 + }, + { + "epoch": 3.32, + "learning_rate": 8.768625615854606e-06, + "loss": 1.0029, + "step": 44912 + }, + { + "epoch": 3.32, + "learning_rate": 8.768229780776154e-06, + "loss": 1.0484, + "step": 44913 + }, + { + "epoch": 3.32, + "learning_rate": 8.767833947657429e-06, + "loss": 0.9547, + "step": 44914 + }, + { + "epoch": 3.32, + "learning_rate": 8.767438116499064e-06, + "loss": 1.0327, + "step": 44915 + }, + { + "epoch": 3.32, + "learning_rate": 8.767042287301686e-06, + "loss": 0.9465, + "step": 44916 + }, + { + "epoch": 3.32, + "learning_rate": 8.766646460065928e-06, + "loss": 0.9911, + "step": 44917 + }, + { + "epoch": 3.32, + "learning_rate": 8.766250634792414e-06, + "loss": 0.9772, + "step": 44918 + }, + { + "epoch": 3.32, + "learning_rate": 8.765854811481783e-06, + "loss": 1.0572, + "step": 44919 + }, + { + "epoch": 3.32, + "learning_rate": 8.765458990134654e-06, + "loss": 1.0227, + "step": 44920 + }, + { + "epoch": 3.32, + "learning_rate": 8.765063170751667e-06, + "loss": 0.9732, + "step": 44921 + }, + { + "epoch": 3.32, + "learning_rate": 8.764667353333445e-06, + "loss": 0.9733, + "step": 44922 + }, + { + "epoch": 3.32, + "learning_rate": 8.764271537880618e-06, + "loss": 0.9811, + "step": 44923 + }, + { + "epoch": 3.32, + "learning_rate": 8.76387572439382e-06, + "loss": 0.9433, + "step": 44924 + }, + { + "epoch": 3.32, + "learning_rate": 8.763479912873675e-06, + "loss": 1.0195, + "step": 44925 + }, + { + "epoch": 3.32, + "learning_rate": 8.763084103320818e-06, + "loss": 0.9988, + "step": 44926 + }, + { + "epoch": 3.32, + "learning_rate": 8.762688295735874e-06, + "loss": 1.0306, + "step": 44927 + }, + { + "epoch": 3.32, + "learning_rate": 8.762292490119478e-06, + "loss": 0.9986, + "step": 44928 + }, + { + "epoch": 3.32, + "learning_rate": 8.761896686472254e-06, + "loss": 1.0857, + "step": 44929 + }, + { + "epoch": 3.32, + "learning_rate": 8.761500884794837e-06, + "loss": 0.9856, + "step": 44930 + }, + { + "epoch": 3.32, + "learning_rate": 8.761105085087858e-06, + "loss": 1.0625, + "step": 44931 + }, + { + "epoch": 3.32, + "learning_rate": 8.760709287351936e-06, + "loss": 1.1059, + "step": 44932 + }, + { + "epoch": 3.32, + "learning_rate": 8.760313491587714e-06, + "loss": 1.0138, + "step": 44933 + }, + { + "epoch": 3.32, + "learning_rate": 8.759917697795809e-06, + "loss": 1.0051, + "step": 44934 + }, + { + "epoch": 3.32, + "learning_rate": 8.759521905976863e-06, + "loss": 1.0128, + "step": 44935 + }, + { + "epoch": 3.32, + "learning_rate": 8.759126116131495e-06, + "loss": 1.1168, + "step": 44936 + }, + { + "epoch": 3.32, + "learning_rate": 8.758730328260344e-06, + "loss": 1.0541, + "step": 44937 + }, + { + "epoch": 3.32, + "learning_rate": 8.75833454236403e-06, + "loss": 0.9582, + "step": 44938 + }, + { + "epoch": 3.32, + "learning_rate": 8.757938758443195e-06, + "loss": 0.9436, + "step": 44939 + }, + { + "epoch": 3.32, + "learning_rate": 8.757542976498459e-06, + "loss": 0.9534, + "step": 44940 + }, + { + "epoch": 3.32, + "learning_rate": 8.757147196530451e-06, + "loss": 0.9074, + "step": 44941 + }, + { + "epoch": 3.32, + "learning_rate": 8.756751418539808e-06, + "loss": 0.9128, + "step": 44942 + }, + { + "epoch": 3.32, + "learning_rate": 8.756355642527152e-06, + "loss": 0.8821, + "step": 44943 + }, + { + "epoch": 3.32, + "learning_rate": 8.75595986849312e-06, + "loss": 0.894, + "step": 44944 + }, + { + "epoch": 3.32, + "learning_rate": 8.755564096438334e-06, + "loss": 1.0095, + "step": 44945 + }, + { + "epoch": 3.32, + "learning_rate": 8.755168326363432e-06, + "loss": 1.0161, + "step": 44946 + }, + { + "epoch": 3.32, + "learning_rate": 8.754772558269037e-06, + "loss": 1.0232, + "step": 44947 + }, + { + "epoch": 3.32, + "learning_rate": 8.754376792155781e-06, + "loss": 1.07, + "step": 44948 + }, + { + "epoch": 3.32, + "learning_rate": 8.753981028024296e-06, + "loss": 0.9067, + "step": 44949 + }, + { + "epoch": 3.32, + "learning_rate": 8.753585265875207e-06, + "loss": 1.0703, + "step": 44950 + }, + { + "epoch": 3.32, + "learning_rate": 8.753189505709147e-06, + "loss": 1.0338, + "step": 44951 + }, + { + "epoch": 3.32, + "learning_rate": 8.752793747526743e-06, + "loss": 0.9928, + "step": 44952 + }, + { + "epoch": 3.32, + "learning_rate": 8.75239799132863e-06, + "loss": 1.0507, + "step": 44953 + }, + { + "epoch": 3.32, + "learning_rate": 8.75200223711543e-06, + "loss": 1.0139, + "step": 44954 + }, + { + "epoch": 3.32, + "learning_rate": 8.75160648488778e-06, + "loss": 0.8817, + "step": 44955 + }, + { + "epoch": 3.32, + "learning_rate": 8.751210734646301e-06, + "loss": 1.003, + "step": 44956 + }, + { + "epoch": 3.32, + "learning_rate": 8.750814986391635e-06, + "loss": 0.9589, + "step": 44957 + }, + { + "epoch": 3.32, + "learning_rate": 8.750419240124402e-06, + "loss": 0.9277, + "step": 44958 + }, + { + "epoch": 3.32, + "learning_rate": 8.75002349584523e-06, + "loss": 0.9639, + "step": 44959 + }, + { + "epoch": 3.32, + "learning_rate": 8.749627753554758e-06, + "loss": 1.0802, + "step": 44960 + }, + { + "epoch": 3.32, + "learning_rate": 8.749232013253606e-06, + "loss": 0.9724, + "step": 44961 + }, + { + "epoch": 3.32, + "learning_rate": 8.748836274942412e-06, + "loss": 0.9759, + "step": 44962 + }, + { + "epoch": 3.32, + "learning_rate": 8.748440538621797e-06, + "loss": 0.9379, + "step": 44963 + }, + { + "epoch": 3.32, + "learning_rate": 8.748044804292399e-06, + "loss": 0.9651, + "step": 44964 + }, + { + "epoch": 3.32, + "learning_rate": 8.74764907195484e-06, + "loss": 0.9427, + "step": 44965 + }, + { + "epoch": 3.32, + "learning_rate": 8.747253341609758e-06, + "loss": 1.0839, + "step": 44966 + }, + { + "epoch": 3.32, + "learning_rate": 8.746857613257775e-06, + "loss": 1.0055, + "step": 44967 + }, + { + "epoch": 3.32, + "learning_rate": 8.746461886899525e-06, + "loss": 1.0684, + "step": 44968 + }, + { + "epoch": 3.32, + "learning_rate": 8.746066162535637e-06, + "loss": 1.1063, + "step": 44969 + }, + { + "epoch": 3.32, + "learning_rate": 8.745670440166737e-06, + "loss": 1.0106, + "step": 44970 + }, + { + "epoch": 3.32, + "learning_rate": 8.74527471979346e-06, + "loss": 0.9595, + "step": 44971 + }, + { + "epoch": 3.32, + "learning_rate": 8.744879001416431e-06, + "loss": 1.0117, + "step": 44972 + }, + { + "epoch": 3.32, + "learning_rate": 8.744483285036283e-06, + "loss": 1.0795, + "step": 44973 + }, + { + "epoch": 3.32, + "learning_rate": 8.744087570653643e-06, + "loss": 0.9439, + "step": 44974 + }, + { + "epoch": 3.32, + "learning_rate": 8.743691858269142e-06, + "loss": 1.0699, + "step": 44975 + }, + { + "epoch": 3.32, + "learning_rate": 8.743296147883408e-06, + "loss": 1.037, + "step": 44976 + }, + { + "epoch": 3.32, + "learning_rate": 8.742900439497076e-06, + "loss": 0.9643, + "step": 44977 + }, + { + "epoch": 3.32, + "learning_rate": 8.74250473311077e-06, + "loss": 1.0239, + "step": 44978 + }, + { + "epoch": 3.32, + "learning_rate": 8.742109028725117e-06, + "loss": 0.9388, + "step": 44979 + }, + { + "epoch": 3.32, + "learning_rate": 8.741713326340756e-06, + "loss": 0.9603, + "step": 44980 + }, + { + "epoch": 3.32, + "learning_rate": 8.741317625958305e-06, + "loss": 1.1153, + "step": 44981 + }, + { + "epoch": 3.32, + "learning_rate": 8.740921927578407e-06, + "loss": 1.0301, + "step": 44982 + }, + { + "epoch": 3.32, + "learning_rate": 8.740526231201676e-06, + "loss": 1.0007, + "step": 44983 + }, + { + "epoch": 3.32, + "learning_rate": 8.740130536828757e-06, + "loss": 0.9607, + "step": 44984 + }, + { + "epoch": 3.32, + "learning_rate": 8.739734844460266e-06, + "loss": 1.0805, + "step": 44985 + }, + { + "epoch": 3.32, + "learning_rate": 8.739339154096845e-06, + "loss": 1.0717, + "step": 44986 + }, + { + "epoch": 3.32, + "learning_rate": 8.738943465739116e-06, + "loss": 1.0496, + "step": 44987 + }, + { + "epoch": 3.32, + "learning_rate": 8.738547779387707e-06, + "loss": 0.8503, + "step": 44988 + }, + { + "epoch": 3.32, + "learning_rate": 8.738152095043254e-06, + "loss": 1.0907, + "step": 44989 + }, + { + "epoch": 3.32, + "learning_rate": 8.737756412706377e-06, + "loss": 1.061, + "step": 44990 + }, + { + "epoch": 3.32, + "learning_rate": 8.737360732377718e-06, + "loss": 0.9331, + "step": 44991 + }, + { + "epoch": 3.32, + "learning_rate": 8.736965054057896e-06, + "loss": 1.0614, + "step": 44992 + }, + { + "epoch": 3.32, + "learning_rate": 8.736569377747545e-06, + "loss": 1.096, + "step": 44993 + }, + { + "epoch": 3.32, + "learning_rate": 8.736173703447296e-06, + "loss": 0.9972, + "step": 44994 + }, + { + "epoch": 3.32, + "learning_rate": 8.735778031157775e-06, + "loss": 0.7998, + "step": 44995 + }, + { + "epoch": 3.32, + "learning_rate": 8.735382360879614e-06, + "loss": 1.0231, + "step": 44996 + }, + { + "epoch": 3.32, + "learning_rate": 8.73498669261344e-06, + "loss": 1.0632, + "step": 44997 + }, + { + "epoch": 3.32, + "learning_rate": 8.734591026359887e-06, + "loss": 0.9957, + "step": 44998 + }, + { + "epoch": 3.32, + "learning_rate": 8.734195362119578e-06, + "loss": 0.8385, + "step": 44999 + }, + { + "epoch": 3.33, + "learning_rate": 8.733799699893151e-06, + "loss": 0.9412, + "step": 45000 + }, + { + "epoch": 3.33, + "learning_rate": 8.733404039681223e-06, + "loss": 0.9885, + "step": 45001 + }, + { + "epoch": 3.33, + "learning_rate": 8.733008381484438e-06, + "loss": 0.9389, + "step": 45002 + }, + { + "epoch": 3.33, + "learning_rate": 8.732612725303413e-06, + "loss": 1.0047, + "step": 45003 + }, + { + "epoch": 3.33, + "learning_rate": 8.73221707113879e-06, + "loss": 1.0337, + "step": 45004 + }, + { + "epoch": 3.33, + "learning_rate": 8.731821418991187e-06, + "loss": 0.9775, + "step": 45005 + }, + { + "epoch": 3.33, + "learning_rate": 8.731425768861238e-06, + "loss": 1.0043, + "step": 45006 + }, + { + "epoch": 3.33, + "learning_rate": 8.731030120749573e-06, + "loss": 1.0044, + "step": 45007 + }, + { + "epoch": 3.33, + "learning_rate": 8.73063447465682e-06, + "loss": 0.9424, + "step": 45008 + }, + { + "epoch": 3.33, + "learning_rate": 8.73023883058361e-06, + "loss": 1.1005, + "step": 45009 + }, + { + "epoch": 3.33, + "learning_rate": 8.72984318853057e-06, + "loss": 1.0106, + "step": 45010 + }, + { + "epoch": 3.33, + "learning_rate": 8.729447548498335e-06, + "loss": 0.9581, + "step": 45011 + }, + { + "epoch": 3.33, + "learning_rate": 8.72905191048753e-06, + "loss": 1.0472, + "step": 45012 + }, + { + "epoch": 3.33, + "learning_rate": 8.728656274498783e-06, + "loss": 0.9885, + "step": 45013 + }, + { + "epoch": 3.33, + "learning_rate": 8.728260640532728e-06, + "loss": 1.0488, + "step": 45014 + }, + { + "epoch": 3.33, + "learning_rate": 8.72786500858999e-06, + "loss": 1.0798, + "step": 45015 + }, + { + "epoch": 3.33, + "learning_rate": 8.727469378671205e-06, + "loss": 1.0296, + "step": 45016 + }, + { + "epoch": 3.33, + "learning_rate": 8.727073750776991e-06, + "loss": 0.8828, + "step": 45017 + }, + { + "epoch": 3.33, + "learning_rate": 8.72667812490799e-06, + "loss": 0.9693, + "step": 45018 + }, + { + "epoch": 3.33, + "learning_rate": 8.726282501064822e-06, + "loss": 0.9811, + "step": 45019 + }, + { + "epoch": 3.33, + "learning_rate": 8.725886879248125e-06, + "loss": 1.1378, + "step": 45020 + }, + { + "epoch": 3.33, + "learning_rate": 8.725491259458518e-06, + "loss": 0.9487, + "step": 45021 + }, + { + "epoch": 3.33, + "learning_rate": 8.725095641696642e-06, + "loss": 1.084, + "step": 45022 + }, + { + "epoch": 3.33, + "learning_rate": 8.724700025963115e-06, + "loss": 0.9514, + "step": 45023 + }, + { + "epoch": 3.33, + "learning_rate": 8.724304412258577e-06, + "loss": 1.0156, + "step": 45024 + }, + { + "epoch": 3.33, + "learning_rate": 8.723908800583654e-06, + "loss": 0.9553, + "step": 45025 + }, + { + "epoch": 3.33, + "learning_rate": 8.723513190938966e-06, + "loss": 1.0984, + "step": 45026 + }, + { + "epoch": 3.33, + "learning_rate": 8.723117583325158e-06, + "loss": 0.9751, + "step": 45027 + }, + { + "epoch": 3.33, + "learning_rate": 8.722721977742844e-06, + "loss": 1.0766, + "step": 45028 + }, + { + "epoch": 3.33, + "learning_rate": 8.722326374192668e-06, + "loss": 1.0133, + "step": 45029 + }, + { + "epoch": 3.33, + "learning_rate": 8.721930772675246e-06, + "loss": 1.0076, + "step": 45030 + }, + { + "epoch": 3.33, + "learning_rate": 8.721535173191221e-06, + "loss": 1.0354, + "step": 45031 + }, + { + "epoch": 3.33, + "learning_rate": 8.721139575741208e-06, + "loss": 1.0736, + "step": 45032 + }, + { + "epoch": 3.33, + "learning_rate": 8.72074398032585e-06, + "loss": 0.9407, + "step": 45033 + }, + { + "epoch": 3.33, + "learning_rate": 8.720348386945768e-06, + "loss": 0.9363, + "step": 45034 + }, + { + "epoch": 3.33, + "learning_rate": 8.719952795601591e-06, + "loss": 1.0182, + "step": 45035 + }, + { + "epoch": 3.33, + "learning_rate": 8.719557206293955e-06, + "loss": 1.0413, + "step": 45036 + }, + { + "epoch": 3.33, + "learning_rate": 8.71916161902348e-06, + "loss": 1.0328, + "step": 45037 + }, + { + "epoch": 3.33, + "learning_rate": 8.718766033790805e-06, + "loss": 1.0366, + "step": 45038 + }, + { + "epoch": 3.33, + "learning_rate": 8.718370450596551e-06, + "loss": 0.984, + "step": 45039 + }, + { + "epoch": 3.33, + "learning_rate": 8.717974869441355e-06, + "loss": 0.9948, + "step": 45040 + }, + { + "epoch": 3.33, + "learning_rate": 8.71757929032584e-06, + "loss": 0.9614, + "step": 45041 + }, + { + "epoch": 3.33, + "learning_rate": 8.71718371325064e-06, + "loss": 0.9878, + "step": 45042 + }, + { + "epoch": 3.33, + "learning_rate": 8.716788138216381e-06, + "loss": 1.0866, + "step": 45043 + }, + { + "epoch": 3.33, + "learning_rate": 8.716392565223695e-06, + "loss": 0.838, + "step": 45044 + }, + { + "epoch": 3.33, + "learning_rate": 8.715996994273211e-06, + "loss": 1.0179, + "step": 45045 + }, + { + "epoch": 3.33, + "learning_rate": 8.715601425365553e-06, + "loss": 0.9694, + "step": 45046 + }, + { + "epoch": 3.33, + "learning_rate": 8.715205858501361e-06, + "loss": 1.0295, + "step": 45047 + }, + { + "epoch": 3.33, + "learning_rate": 8.71481029368125e-06, + "loss": 0.9665, + "step": 45048 + }, + { + "epoch": 3.33, + "learning_rate": 8.714414730905866e-06, + "loss": 1.0575, + "step": 45049 + }, + { + "epoch": 3.33, + "learning_rate": 8.714019170175821e-06, + "loss": 0.9734, + "step": 45050 + }, + { + "epoch": 3.33, + "learning_rate": 8.713623611491762e-06, + "loss": 1.053, + "step": 45051 + }, + { + "epoch": 3.33, + "learning_rate": 8.713228054854306e-06, + "loss": 1.0596, + "step": 45052 + }, + { + "epoch": 3.33, + "learning_rate": 8.712832500264084e-06, + "loss": 1.0211, + "step": 45053 + }, + { + "epoch": 3.33, + "learning_rate": 8.712436947721729e-06, + "loss": 0.949, + "step": 45054 + }, + { + "epoch": 3.33, + "learning_rate": 8.712041397227865e-06, + "loss": 0.9648, + "step": 45055 + }, + { + "epoch": 3.33, + "learning_rate": 8.711645848783129e-06, + "loss": 1.0146, + "step": 45056 + }, + { + "epoch": 3.33, + "learning_rate": 8.711250302388142e-06, + "loss": 1.0748, + "step": 45057 + }, + { + "epoch": 3.33, + "learning_rate": 8.710854758043541e-06, + "loss": 1.0591, + "step": 45058 + }, + { + "epoch": 3.33, + "learning_rate": 8.71045921574995e-06, + "loss": 1.0173, + "step": 45059 + }, + { + "epoch": 3.33, + "learning_rate": 8.710063675508e-06, + "loss": 0.9552, + "step": 45060 + }, + { + "epoch": 3.33, + "learning_rate": 8.709668137318321e-06, + "loss": 1.0708, + "step": 45061 + }, + { + "epoch": 3.33, + "learning_rate": 8.70927260118154e-06, + "loss": 0.9269, + "step": 45062 + }, + { + "epoch": 3.33, + "learning_rate": 8.708877067098289e-06, + "loss": 0.9859, + "step": 45063 + }, + { + "epoch": 3.33, + "learning_rate": 8.708481535069194e-06, + "loss": 1.0435, + "step": 45064 + }, + { + "epoch": 3.33, + "learning_rate": 8.708086005094889e-06, + "loss": 0.9658, + "step": 45065 + }, + { + "epoch": 3.33, + "learning_rate": 8.707690477175998e-06, + "loss": 0.9189, + "step": 45066 + }, + { + "epoch": 3.33, + "learning_rate": 8.707294951313156e-06, + "loss": 1.0119, + "step": 45067 + }, + { + "epoch": 3.33, + "learning_rate": 8.706899427506987e-06, + "loss": 0.9965, + "step": 45068 + }, + { + "epoch": 3.33, + "learning_rate": 8.706503905758125e-06, + "loss": 0.9736, + "step": 45069 + }, + { + "epoch": 3.33, + "learning_rate": 8.706108386067199e-06, + "loss": 1.0717, + "step": 45070 + }, + { + "epoch": 3.33, + "learning_rate": 8.705712868434829e-06, + "loss": 0.8534, + "step": 45071 + }, + { + "epoch": 3.33, + "learning_rate": 8.705317352861657e-06, + "loss": 0.9646, + "step": 45072 + }, + { + "epoch": 3.33, + "learning_rate": 8.704921839348301e-06, + "loss": 0.9643, + "step": 45073 + }, + { + "epoch": 3.33, + "learning_rate": 8.704526327895402e-06, + "loss": 1.0466, + "step": 45074 + }, + { + "epoch": 3.33, + "learning_rate": 8.704130818503577e-06, + "loss": 1.0225, + "step": 45075 + }, + { + "epoch": 3.33, + "learning_rate": 8.703735311173467e-06, + "loss": 0.9514, + "step": 45076 + }, + { + "epoch": 3.33, + "learning_rate": 8.703339805905691e-06, + "loss": 1.01, + "step": 45077 + }, + { + "epoch": 3.33, + "learning_rate": 8.702944302700886e-06, + "loss": 0.9833, + "step": 45078 + }, + { + "epoch": 3.33, + "learning_rate": 8.702548801559677e-06, + "loss": 1.0201, + "step": 45079 + }, + { + "epoch": 3.33, + "learning_rate": 8.702153302482695e-06, + "loss": 1.0283, + "step": 45080 + }, + { + "epoch": 3.33, + "learning_rate": 8.701757805470568e-06, + "loss": 1.0559, + "step": 45081 + }, + { + "epoch": 3.33, + "learning_rate": 8.701362310523924e-06, + "loss": 1.0212, + "step": 45082 + }, + { + "epoch": 3.33, + "learning_rate": 8.700966817643396e-06, + "loss": 0.9222, + "step": 45083 + }, + { + "epoch": 3.33, + "learning_rate": 8.70057132682961e-06, + "loss": 1.0104, + "step": 45084 + }, + { + "epoch": 3.33, + "learning_rate": 8.700175838083197e-06, + "loss": 1.033, + "step": 45085 + }, + { + "epoch": 3.33, + "learning_rate": 8.699780351404786e-06, + "loss": 1.0309, + "step": 45086 + }, + { + "epoch": 3.33, + "learning_rate": 8.699384866795007e-06, + "loss": 0.9907, + "step": 45087 + }, + { + "epoch": 3.33, + "learning_rate": 8.698989384254485e-06, + "loss": 1.0184, + "step": 45088 + }, + { + "epoch": 3.33, + "learning_rate": 8.698593903783856e-06, + "loss": 1.0511, + "step": 45089 + }, + { + "epoch": 3.33, + "learning_rate": 8.698198425383746e-06, + "loss": 0.9885, + "step": 45090 + }, + { + "epoch": 3.33, + "learning_rate": 8.697802949054776e-06, + "loss": 0.9679, + "step": 45091 + }, + { + "epoch": 3.33, + "learning_rate": 8.697407474797593e-06, + "loss": 1.1134, + "step": 45092 + }, + { + "epoch": 3.33, + "learning_rate": 8.697012002612806e-06, + "loss": 1.0693, + "step": 45093 + }, + { + "epoch": 3.33, + "learning_rate": 8.696616532501063e-06, + "loss": 0.9681, + "step": 45094 + }, + { + "epoch": 3.33, + "learning_rate": 8.696221064462977e-06, + "loss": 0.9102, + "step": 45095 + }, + { + "epoch": 3.33, + "learning_rate": 8.695825598499192e-06, + "loss": 1.132, + "step": 45096 + }, + { + "epoch": 3.33, + "learning_rate": 8.695430134610324e-06, + "loss": 1.0523, + "step": 45097 + }, + { + "epoch": 3.33, + "learning_rate": 8.695034672797012e-06, + "loss": 1.0076, + "step": 45098 + }, + { + "epoch": 3.33, + "learning_rate": 8.69463921305988e-06, + "loss": 1.1453, + "step": 45099 + }, + { + "epoch": 3.33, + "learning_rate": 8.694243755399556e-06, + "loss": 1.0538, + "step": 45100 + }, + { + "epoch": 3.33, + "learning_rate": 8.693848299816676e-06, + "loss": 1.0876, + "step": 45101 + }, + { + "epoch": 3.33, + "learning_rate": 8.693452846311859e-06, + "loss": 0.9737, + "step": 45102 + }, + { + "epoch": 3.33, + "learning_rate": 8.693057394885744e-06, + "loss": 0.9651, + "step": 45103 + }, + { + "epoch": 3.33, + "learning_rate": 8.692661945538954e-06, + "loss": 0.9667, + "step": 45104 + }, + { + "epoch": 3.33, + "learning_rate": 8.692266498272122e-06, + "loss": 1.0052, + "step": 45105 + }, + { + "epoch": 3.33, + "learning_rate": 8.691871053085872e-06, + "loss": 1.064, + "step": 45106 + }, + { + "epoch": 3.33, + "learning_rate": 8.69147560998084e-06, + "loss": 0.9662, + "step": 45107 + }, + { + "epoch": 3.33, + "learning_rate": 8.691080168957651e-06, + "loss": 0.9564, + "step": 45108 + }, + { + "epoch": 3.33, + "learning_rate": 8.690684730016932e-06, + "loss": 1.0462, + "step": 45109 + }, + { + "epoch": 3.33, + "learning_rate": 8.690289293159318e-06, + "loss": 0.9644, + "step": 45110 + }, + { + "epoch": 3.33, + "learning_rate": 8.689893858385432e-06, + "loss": 1.0435, + "step": 45111 + }, + { + "epoch": 3.33, + "learning_rate": 8.68949842569591e-06, + "loss": 0.8388, + "step": 45112 + }, + { + "epoch": 3.33, + "learning_rate": 8.689102995091372e-06, + "loss": 1.0548, + "step": 45113 + }, + { + "epoch": 3.33, + "learning_rate": 8.688707566572457e-06, + "loss": 0.9456, + "step": 45114 + }, + { + "epoch": 3.33, + "learning_rate": 8.688312140139787e-06, + "loss": 0.9929, + "step": 45115 + }, + { + "epoch": 3.33, + "learning_rate": 8.687916715793995e-06, + "loss": 0.9756, + "step": 45116 + }, + { + "epoch": 3.33, + "learning_rate": 8.687521293535712e-06, + "loss": 1.0234, + "step": 45117 + }, + { + "epoch": 3.33, + "learning_rate": 8.687125873365557e-06, + "loss": 1.0007, + "step": 45118 + }, + { + "epoch": 3.33, + "learning_rate": 8.686730455284173e-06, + "loss": 0.9454, + "step": 45119 + }, + { + "epoch": 3.33, + "learning_rate": 8.686335039292174e-06, + "loss": 0.9467, + "step": 45120 + }, + { + "epoch": 3.33, + "learning_rate": 8.685939625390205e-06, + "loss": 0.9113, + "step": 45121 + }, + { + "epoch": 3.33, + "learning_rate": 8.685544213578882e-06, + "loss": 1.0689, + "step": 45122 + }, + { + "epoch": 3.33, + "learning_rate": 8.685148803858842e-06, + "loss": 0.987, + "step": 45123 + }, + { + "epoch": 3.33, + "learning_rate": 8.684753396230708e-06, + "loss": 0.8875, + "step": 45124 + }, + { + "epoch": 3.33, + "learning_rate": 8.684357990695117e-06, + "loss": 1.0305, + "step": 45125 + }, + { + "epoch": 3.33, + "learning_rate": 8.68396258725269e-06, + "loss": 0.9414, + "step": 45126 + }, + { + "epoch": 3.33, + "learning_rate": 8.683567185904063e-06, + "loss": 1.0024, + "step": 45127 + }, + { + "epoch": 3.33, + "learning_rate": 8.68317178664986e-06, + "loss": 0.9321, + "step": 45128 + }, + { + "epoch": 3.33, + "learning_rate": 8.68277638949071e-06, + "loss": 1.0107, + "step": 45129 + }, + { + "epoch": 3.33, + "learning_rate": 8.682380994427248e-06, + "loss": 0.9831, + "step": 45130 + }, + { + "epoch": 3.33, + "learning_rate": 8.681985601460095e-06, + "loss": 0.9299, + "step": 45131 + }, + { + "epoch": 3.33, + "learning_rate": 8.681590210589886e-06, + "loss": 1.0553, + "step": 45132 + }, + { + "epoch": 3.33, + "learning_rate": 8.681194821817247e-06, + "loss": 0.9832, + "step": 45133 + }, + { + "epoch": 3.33, + "learning_rate": 8.680799435142811e-06, + "loss": 1.0303, + "step": 45134 + }, + { + "epoch": 3.33, + "learning_rate": 8.6804040505672e-06, + "loss": 0.9072, + "step": 45135 + }, + { + "epoch": 3.34, + "learning_rate": 8.68000866809105e-06, + "loss": 0.9126, + "step": 45136 + }, + { + "epoch": 3.34, + "learning_rate": 8.679613287714991e-06, + "loss": 1.0776, + "step": 45137 + }, + { + "epoch": 3.34, + "learning_rate": 8.67921790943964e-06, + "loss": 1.0672, + "step": 45138 + }, + { + "epoch": 3.34, + "learning_rate": 8.678822533265642e-06, + "loss": 0.9458, + "step": 45139 + }, + { + "epoch": 3.34, + "learning_rate": 8.678427159193612e-06, + "loss": 1.1095, + "step": 45140 + }, + { + "epoch": 3.34, + "learning_rate": 8.678031787224192e-06, + "loss": 1.009, + "step": 45141 + }, + { + "epoch": 3.34, + "learning_rate": 8.677636417357997e-06, + "loss": 0.9783, + "step": 45142 + }, + { + "epoch": 3.34, + "learning_rate": 8.67724104959567e-06, + "loss": 0.9913, + "step": 45143 + }, + { + "epoch": 3.34, + "learning_rate": 8.676845683937828e-06, + "loss": 1.0343, + "step": 45144 + }, + { + "epoch": 3.34, + "learning_rate": 8.676450320385112e-06, + "loss": 1.0598, + "step": 45145 + }, + { + "epoch": 3.34, + "learning_rate": 8.676054958938143e-06, + "loss": 0.8713, + "step": 45146 + }, + { + "epoch": 3.34, + "learning_rate": 8.675659599597549e-06, + "loss": 1.0112, + "step": 45147 + }, + { + "epoch": 3.34, + "learning_rate": 8.675264242363963e-06, + "loss": 1.0612, + "step": 45148 + }, + { + "epoch": 3.34, + "learning_rate": 8.674868887238011e-06, + "loss": 1.0553, + "step": 45149 + }, + { + "epoch": 3.34, + "learning_rate": 8.674473534220326e-06, + "loss": 1.0936, + "step": 45150 + }, + { + "epoch": 3.34, + "learning_rate": 8.674078183311533e-06, + "loss": 0.9491, + "step": 45151 + }, + { + "epoch": 3.34, + "learning_rate": 8.673682834512264e-06, + "loss": 1.0323, + "step": 45152 + }, + { + "epoch": 3.34, + "learning_rate": 8.673287487823144e-06, + "loss": 1.0545, + "step": 45153 + }, + { + "epoch": 3.34, + "learning_rate": 8.672892143244807e-06, + "loss": 1.0797, + "step": 45154 + }, + { + "epoch": 3.34, + "learning_rate": 8.67249680077788e-06, + "loss": 0.8917, + "step": 45155 + }, + { + "epoch": 3.34, + "learning_rate": 8.67210146042299e-06, + "loss": 1.0073, + "step": 45156 + }, + { + "epoch": 3.34, + "learning_rate": 8.671706122180769e-06, + "loss": 0.8806, + "step": 45157 + }, + { + "epoch": 3.34, + "learning_rate": 8.671310786051844e-06, + "loss": 1.0109, + "step": 45158 + }, + { + "epoch": 3.34, + "learning_rate": 8.670915452036845e-06, + "loss": 0.8919, + "step": 45159 + }, + { + "epoch": 3.34, + "learning_rate": 8.670520120136398e-06, + "loss": 1.0519, + "step": 45160 + }, + { + "epoch": 3.34, + "learning_rate": 8.670124790351138e-06, + "loss": 0.9663, + "step": 45161 + }, + { + "epoch": 3.34, + "learning_rate": 8.669729462681687e-06, + "loss": 1.009, + "step": 45162 + }, + { + "epoch": 3.34, + "learning_rate": 8.66933413712868e-06, + "loss": 1.0253, + "step": 45163 + }, + { + "epoch": 3.34, + "learning_rate": 8.668938813692746e-06, + "loss": 1.0069, + "step": 45164 + }, + { + "epoch": 3.34, + "learning_rate": 8.668543492374504e-06, + "loss": 0.9165, + "step": 45165 + }, + { + "epoch": 3.34, + "learning_rate": 8.668148173174596e-06, + "loss": 0.9144, + "step": 45166 + }, + { + "epoch": 3.34, + "learning_rate": 8.667752856093642e-06, + "loss": 1.0692, + "step": 45167 + }, + { + "epoch": 3.34, + "learning_rate": 8.667357541132277e-06, + "loss": 1.0399, + "step": 45168 + }, + { + "epoch": 3.34, + "learning_rate": 8.666962228291125e-06, + "loss": 0.9944, + "step": 45169 + }, + { + "epoch": 3.34, + "learning_rate": 8.666566917570818e-06, + "loss": 1.1595, + "step": 45170 + }, + { + "epoch": 3.34, + "learning_rate": 8.666171608971982e-06, + "loss": 0.9616, + "step": 45171 + }, + { + "epoch": 3.34, + "learning_rate": 8.66577630249525e-06, + "loss": 0.9485, + "step": 45172 + }, + { + "epoch": 3.34, + "learning_rate": 8.665380998141249e-06, + "loss": 1.0378, + "step": 45173 + }, + { + "epoch": 3.34, + "learning_rate": 8.664985695910606e-06, + "loss": 0.9965, + "step": 45174 + }, + { + "epoch": 3.34, + "learning_rate": 8.664590395803954e-06, + "loss": 0.9803, + "step": 45175 + }, + { + "epoch": 3.34, + "learning_rate": 8.664195097821916e-06, + "loss": 1.0117, + "step": 45176 + }, + { + "epoch": 3.34, + "learning_rate": 8.663799801965127e-06, + "loss": 0.8912, + "step": 45177 + }, + { + "epoch": 3.34, + "learning_rate": 8.663404508234212e-06, + "loss": 0.9541, + "step": 45178 + }, + { + "epoch": 3.34, + "learning_rate": 8.663009216629802e-06, + "loss": 1.0109, + "step": 45179 + }, + { + "epoch": 3.34, + "learning_rate": 8.662613927152526e-06, + "loss": 0.9471, + "step": 45180 + }, + { + "epoch": 3.34, + "learning_rate": 8.662218639803012e-06, + "loss": 0.9692, + "step": 45181 + }, + { + "epoch": 3.34, + "learning_rate": 8.661823354581887e-06, + "loss": 1.0615, + "step": 45182 + }, + { + "epoch": 3.34, + "learning_rate": 8.661428071489785e-06, + "loss": 1.0063, + "step": 45183 + }, + { + "epoch": 3.34, + "learning_rate": 8.661032790527333e-06, + "loss": 1.0129, + "step": 45184 + }, + { + "epoch": 3.34, + "learning_rate": 8.660637511695154e-06, + "loss": 0.9996, + "step": 45185 + }, + { + "epoch": 3.34, + "learning_rate": 8.660242234993886e-06, + "loss": 0.9656, + "step": 45186 + }, + { + "epoch": 3.34, + "learning_rate": 8.659846960424147e-06, + "loss": 0.9775, + "step": 45187 + }, + { + "epoch": 3.34, + "learning_rate": 8.65945168798658e-06, + "loss": 1.0952, + "step": 45188 + }, + { + "epoch": 3.34, + "learning_rate": 8.659056417681798e-06, + "loss": 1.0004, + "step": 45189 + }, + { + "epoch": 3.34, + "learning_rate": 8.658661149510445e-06, + "loss": 1.05, + "step": 45190 + }, + { + "epoch": 3.34, + "learning_rate": 8.658265883473139e-06, + "loss": 1.0269, + "step": 45191 + }, + { + "epoch": 3.34, + "learning_rate": 8.657870619570516e-06, + "loss": 0.9446, + "step": 45192 + }, + { + "epoch": 3.34, + "learning_rate": 8.657475357803202e-06, + "loss": 1.0029, + "step": 45193 + }, + { + "epoch": 3.34, + "learning_rate": 8.657080098171822e-06, + "loss": 0.8886, + "step": 45194 + }, + { + "epoch": 3.34, + "learning_rate": 8.65668484067701e-06, + "loss": 1.0319, + "step": 45195 + }, + { + "epoch": 3.34, + "learning_rate": 8.656289585319391e-06, + "loss": 1.1164, + "step": 45196 + }, + { + "epoch": 3.34, + "learning_rate": 8.655894332099599e-06, + "loss": 0.9655, + "step": 45197 + }, + { + "epoch": 3.34, + "learning_rate": 8.655499081018259e-06, + "loss": 1.0475, + "step": 45198 + }, + { + "epoch": 3.34, + "learning_rate": 8.655103832076e-06, + "loss": 1.0142, + "step": 45199 + }, + { + "epoch": 3.34, + "learning_rate": 8.654708585273451e-06, + "loss": 1.0334, + "step": 45200 + }, + { + "epoch": 3.34, + "learning_rate": 8.654313340611245e-06, + "loss": 0.8459, + "step": 45201 + }, + { + "epoch": 3.34, + "learning_rate": 8.653918098090005e-06, + "loss": 0.9306, + "step": 45202 + }, + { + "epoch": 3.34, + "learning_rate": 8.65352285771036e-06, + "loss": 0.9793, + "step": 45203 + }, + { + "epoch": 3.34, + "learning_rate": 8.653127619472942e-06, + "loss": 0.9946, + "step": 45204 + }, + { + "epoch": 3.34, + "learning_rate": 8.65273238337838e-06, + "loss": 0.9407, + "step": 45205 + }, + { + "epoch": 3.34, + "learning_rate": 8.652337149427301e-06, + "loss": 0.9978, + "step": 45206 + }, + { + "epoch": 3.34, + "learning_rate": 8.651941917620333e-06, + "loss": 1.1683, + "step": 45207 + }, + { + "epoch": 3.34, + "learning_rate": 8.65154668795811e-06, + "loss": 0.926, + "step": 45208 + }, + { + "epoch": 3.34, + "learning_rate": 8.651151460441252e-06, + "loss": 1.1503, + "step": 45209 + }, + { + "epoch": 3.34, + "learning_rate": 8.650756235070398e-06, + "loss": 1.0398, + "step": 45210 + }, + { + "epoch": 3.34, + "learning_rate": 8.650361011846168e-06, + "loss": 1.1083, + "step": 45211 + }, + { + "epoch": 3.34, + "learning_rate": 8.649965790769195e-06, + "loss": 0.9829, + "step": 45212 + }, + { + "epoch": 3.34, + "learning_rate": 8.649570571840107e-06, + "loss": 1.0836, + "step": 45213 + }, + { + "epoch": 3.34, + "learning_rate": 8.649175355059532e-06, + "loss": 1.017, + "step": 45214 + }, + { + "epoch": 3.34, + "learning_rate": 8.648780140428101e-06, + "loss": 1.0154, + "step": 45215 + }, + { + "epoch": 3.34, + "learning_rate": 8.64838492794644e-06, + "loss": 1.0017, + "step": 45216 + }, + { + "epoch": 3.34, + "learning_rate": 8.64798971761518e-06, + "loss": 1.0481, + "step": 45217 + }, + { + "epoch": 3.34, + "learning_rate": 8.64759450943495e-06, + "loss": 0.9468, + "step": 45218 + }, + { + "epoch": 3.34, + "learning_rate": 8.647199303406376e-06, + "loss": 1.0238, + "step": 45219 + }, + { + "epoch": 3.34, + "learning_rate": 8.64680409953009e-06, + "loss": 1.0474, + "step": 45220 + }, + { + "epoch": 3.34, + "learning_rate": 8.646408897806719e-06, + "loss": 0.9402, + "step": 45221 + }, + { + "epoch": 3.34, + "learning_rate": 8.646013698236892e-06, + "loss": 0.9068, + "step": 45222 + }, + { + "epoch": 3.34, + "learning_rate": 8.645618500821237e-06, + "loss": 0.9588, + "step": 45223 + }, + { + "epoch": 3.34, + "learning_rate": 8.645223305560386e-06, + "loss": 1.0253, + "step": 45224 + }, + { + "epoch": 3.34, + "learning_rate": 8.644828112454961e-06, + "loss": 1.0371, + "step": 45225 + }, + { + "epoch": 3.34, + "learning_rate": 8.644432921505599e-06, + "loss": 0.9154, + "step": 45226 + }, + { + "epoch": 3.34, + "learning_rate": 8.644037732712924e-06, + "loss": 0.9871, + "step": 45227 + }, + { + "epoch": 3.34, + "learning_rate": 8.643642546077565e-06, + "loss": 0.9971, + "step": 45228 + }, + { + "epoch": 3.34, + "learning_rate": 8.643247361600155e-06, + "loss": 1.0712, + "step": 45229 + }, + { + "epoch": 3.34, + "learning_rate": 8.642852179281312e-06, + "loss": 1.0418, + "step": 45230 + }, + { + "epoch": 3.34, + "learning_rate": 8.642456999121677e-06, + "loss": 1.0073, + "step": 45231 + }, + { + "epoch": 3.34, + "learning_rate": 8.64206182112187e-06, + "loss": 0.9457, + "step": 45232 + }, + { + "epoch": 3.34, + "learning_rate": 8.641666645282528e-06, + "loss": 0.9512, + "step": 45233 + }, + { + "epoch": 3.34, + "learning_rate": 8.641271471604268e-06, + "loss": 1.2464, + "step": 45234 + }, + { + "epoch": 3.34, + "learning_rate": 8.640876300087732e-06, + "loss": 1.1568, + "step": 45235 + }, + { + "epoch": 3.34, + "learning_rate": 8.640481130733536e-06, + "loss": 0.9487, + "step": 45236 + }, + { + "epoch": 3.34, + "learning_rate": 8.640085963542324e-06, + "loss": 0.9599, + "step": 45237 + }, + { + "epoch": 3.34, + "learning_rate": 8.639690798514706e-06, + "loss": 0.9615, + "step": 45238 + }, + { + "epoch": 3.34, + "learning_rate": 8.639295635651328e-06, + "loss": 1.0207, + "step": 45239 + }, + { + "epoch": 3.34, + "learning_rate": 8.63890047495281e-06, + "loss": 0.961, + "step": 45240 + }, + { + "epoch": 3.34, + "learning_rate": 8.638505316419778e-06, + "loss": 0.9304, + "step": 45241 + }, + { + "epoch": 3.34, + "learning_rate": 8.638110160052868e-06, + "loss": 1.09, + "step": 45242 + }, + { + "epoch": 3.34, + "learning_rate": 8.637715005852704e-06, + "loss": 1.058, + "step": 45243 + }, + { + "epoch": 3.34, + "learning_rate": 8.637319853819917e-06, + "loss": 0.9907, + "step": 45244 + }, + { + "epoch": 3.34, + "learning_rate": 8.636924703955131e-06, + "loss": 1.0694, + "step": 45245 + }, + { + "epoch": 3.34, + "learning_rate": 8.636529556258983e-06, + "loss": 0.8811, + "step": 45246 + }, + { + "epoch": 3.34, + "learning_rate": 8.636134410732093e-06, + "loss": 0.9828, + "step": 45247 + }, + { + "epoch": 3.34, + "learning_rate": 8.635739267375095e-06, + "loss": 0.9544, + "step": 45248 + }, + { + "epoch": 3.34, + "learning_rate": 8.63534412618862e-06, + "loss": 0.908, + "step": 45249 + }, + { + "epoch": 3.34, + "learning_rate": 8.634948987173289e-06, + "loss": 1.1078, + "step": 45250 + }, + { + "epoch": 3.34, + "learning_rate": 8.634553850329735e-06, + "loss": 0.9427, + "step": 45251 + }, + { + "epoch": 3.34, + "learning_rate": 8.634158715658586e-06, + "loss": 1.0395, + "step": 45252 + }, + { + "epoch": 3.34, + "learning_rate": 8.633763583160472e-06, + "loss": 1.0384, + "step": 45253 + }, + { + "epoch": 3.34, + "learning_rate": 8.63336845283602e-06, + "loss": 0.9591, + "step": 45254 + }, + { + "epoch": 3.34, + "learning_rate": 8.632973324685863e-06, + "loss": 0.973, + "step": 45255 + }, + { + "epoch": 3.34, + "learning_rate": 8.632578198710618e-06, + "loss": 0.9685, + "step": 45256 + }, + { + "epoch": 3.34, + "learning_rate": 8.632183074910928e-06, + "loss": 1.0181, + "step": 45257 + }, + { + "epoch": 3.34, + "learning_rate": 8.631787953287415e-06, + "loss": 1.1397, + "step": 45258 + }, + { + "epoch": 3.34, + "learning_rate": 8.631392833840705e-06, + "loss": 0.9786, + "step": 45259 + }, + { + "epoch": 3.34, + "learning_rate": 8.630997716571431e-06, + "loss": 1.0749, + "step": 45260 + }, + { + "epoch": 3.34, + "learning_rate": 8.630602601480217e-06, + "loss": 0.9786, + "step": 45261 + }, + { + "epoch": 3.34, + "learning_rate": 8.6302074885677e-06, + "loss": 1.0073, + "step": 45262 + }, + { + "epoch": 3.34, + "learning_rate": 8.629812377834497e-06, + "loss": 0.9537, + "step": 45263 + }, + { + "epoch": 3.34, + "learning_rate": 8.629417269281248e-06, + "loss": 0.9688, + "step": 45264 + }, + { + "epoch": 3.34, + "learning_rate": 8.629022162908573e-06, + "loss": 1.0342, + "step": 45265 + }, + { + "epoch": 3.34, + "learning_rate": 8.628627058717108e-06, + "loss": 1.0249, + "step": 45266 + }, + { + "epoch": 3.34, + "learning_rate": 8.628231956707476e-06, + "loss": 1.0269, + "step": 45267 + }, + { + "epoch": 3.34, + "learning_rate": 8.627836856880306e-06, + "loss": 1.0033, + "step": 45268 + }, + { + "epoch": 3.34, + "learning_rate": 8.62744175923623e-06, + "loss": 1.066, + "step": 45269 + }, + { + "epoch": 3.34, + "learning_rate": 8.627046663775871e-06, + "loss": 0.968, + "step": 45270 + }, + { + "epoch": 3.35, + "learning_rate": 8.626651570499864e-06, + "loss": 1.0381, + "step": 45271 + }, + { + "epoch": 3.35, + "learning_rate": 8.626256479408834e-06, + "loss": 0.9879, + "step": 45272 + }, + { + "epoch": 3.35, + "learning_rate": 8.62586139050341e-06, + "loss": 1.0093, + "step": 45273 + }, + { + "epoch": 3.35, + "learning_rate": 8.62546630378422e-06, + "loss": 0.9948, + "step": 45274 + }, + { + "epoch": 3.35, + "learning_rate": 8.625071219251895e-06, + "loss": 1.056, + "step": 45275 + }, + { + "epoch": 3.35, + "learning_rate": 8.624676136907067e-06, + "loss": 1.0207, + "step": 45276 + }, + { + "epoch": 3.35, + "learning_rate": 8.62428105675035e-06, + "loss": 0.9459, + "step": 45277 + }, + { + "epoch": 3.35, + "learning_rate": 8.62388597878239e-06, + "loss": 1.0364, + "step": 45278 + }, + { + "epoch": 3.35, + "learning_rate": 8.623490903003801e-06, + "loss": 0.9386, + "step": 45279 + }, + { + "epoch": 3.35, + "learning_rate": 8.623095829415225e-06, + "loss": 1.047, + "step": 45280 + }, + { + "epoch": 3.35, + "learning_rate": 8.622700758017277e-06, + "loss": 0.9877, + "step": 45281 + }, + { + "epoch": 3.35, + "learning_rate": 8.622305688810597e-06, + "loss": 0.9649, + "step": 45282 + }, + { + "epoch": 3.35, + "learning_rate": 8.621910621795804e-06, + "loss": 0.9571, + "step": 45283 + }, + { + "epoch": 3.35, + "learning_rate": 8.62151555697354e-06, + "loss": 0.8782, + "step": 45284 + }, + { + "epoch": 3.35, + "learning_rate": 8.621120494344421e-06, + "loss": 0.9902, + "step": 45285 + }, + { + "epoch": 3.35, + "learning_rate": 8.620725433909077e-06, + "loss": 0.9263, + "step": 45286 + }, + { + "epoch": 3.35, + "learning_rate": 8.620330375668143e-06, + "loss": 1.0688, + "step": 45287 + }, + { + "epoch": 3.35, + "learning_rate": 8.619935319622239e-06, + "loss": 1.0305, + "step": 45288 + }, + { + "epoch": 3.35, + "learning_rate": 8.619540265772002e-06, + "loss": 1.0818, + "step": 45289 + }, + { + "epoch": 3.35, + "learning_rate": 8.619145214118054e-06, + "loss": 0.935, + "step": 45290 + }, + { + "epoch": 3.35, + "learning_rate": 8.618750164661028e-06, + "loss": 0.9441, + "step": 45291 + }, + { + "epoch": 3.35, + "learning_rate": 8.618355117401549e-06, + "loss": 1.0767, + "step": 45292 + }, + { + "epoch": 3.35, + "learning_rate": 8.61796007234025e-06, + "loss": 1.0574, + "step": 45293 + }, + { + "epoch": 3.35, + "learning_rate": 8.617565029477754e-06, + "loss": 0.9505, + "step": 45294 + }, + { + "epoch": 3.35, + "learning_rate": 8.617169988814693e-06, + "loss": 1.1096, + "step": 45295 + }, + { + "epoch": 3.35, + "learning_rate": 8.616774950351696e-06, + "loss": 1.0109, + "step": 45296 + }, + { + "epoch": 3.35, + "learning_rate": 8.616379914089388e-06, + "loss": 1.1257, + "step": 45297 + }, + { + "epoch": 3.35, + "learning_rate": 8.615984880028406e-06, + "loss": 0.9908, + "step": 45298 + }, + { + "epoch": 3.35, + "learning_rate": 8.615589848169363e-06, + "loss": 0.9744, + "step": 45299 + }, + { + "epoch": 3.35, + "learning_rate": 8.615194818512904e-06, + "loss": 0.864, + "step": 45300 + }, + { + "epoch": 3.35, + "learning_rate": 8.614799791059645e-06, + "loss": 1.0581, + "step": 45301 + }, + { + "epoch": 3.35, + "learning_rate": 8.614404765810225e-06, + "loss": 0.9686, + "step": 45302 + }, + { + "epoch": 3.35, + "learning_rate": 8.614009742765262e-06, + "loss": 1.0369, + "step": 45303 + }, + { + "epoch": 3.35, + "learning_rate": 8.613614721925395e-06, + "loss": 1.0054, + "step": 45304 + }, + { + "epoch": 3.35, + "learning_rate": 8.613219703291244e-06, + "loss": 0.938, + "step": 45305 + }, + { + "epoch": 3.35, + "learning_rate": 8.61282468686344e-06, + "loss": 1.0901, + "step": 45306 + }, + { + "epoch": 3.35, + "learning_rate": 8.612429672642614e-06, + "loss": 1.0784, + "step": 45307 + }, + { + "epoch": 3.35, + "learning_rate": 8.612034660629391e-06, + "loss": 1.0276, + "step": 45308 + }, + { + "epoch": 3.35, + "learning_rate": 8.611639650824401e-06, + "loss": 1.0912, + "step": 45309 + }, + { + "epoch": 3.35, + "learning_rate": 8.611244643228272e-06, + "loss": 1.0286, + "step": 45310 + }, + { + "epoch": 3.35, + "learning_rate": 8.610849637841635e-06, + "loss": 1.0598, + "step": 45311 + }, + { + "epoch": 3.35, + "learning_rate": 8.610454634665114e-06, + "loss": 0.9672, + "step": 45312 + }, + { + "epoch": 3.35, + "learning_rate": 8.610059633699341e-06, + "loss": 1.0239, + "step": 45313 + }, + { + "epoch": 3.35, + "learning_rate": 8.609664634944945e-06, + "loss": 0.9163, + "step": 45314 + }, + { + "epoch": 3.35, + "learning_rate": 8.60926963840255e-06, + "loss": 1.0873, + "step": 45315 + }, + { + "epoch": 3.35, + "learning_rate": 8.60887464407279e-06, + "loss": 0.9156, + "step": 45316 + }, + { + "epoch": 3.35, + "learning_rate": 8.608479651956286e-06, + "loss": 1.0206, + "step": 45317 + }, + { + "epoch": 3.35, + "learning_rate": 8.608084662053673e-06, + "loss": 0.9719, + "step": 45318 + }, + { + "epoch": 3.35, + "learning_rate": 8.607689674365578e-06, + "loss": 0.9881, + "step": 45319 + }, + { + "epoch": 3.35, + "learning_rate": 8.60729468889263e-06, + "loss": 1.0375, + "step": 45320 + }, + { + "epoch": 3.35, + "learning_rate": 8.606899705635453e-06, + "loss": 0.9286, + "step": 45321 + }, + { + "epoch": 3.35, + "learning_rate": 8.606504724594681e-06, + "loss": 0.9776, + "step": 45322 + }, + { + "epoch": 3.35, + "learning_rate": 8.606109745770943e-06, + "loss": 1.0991, + "step": 45323 + }, + { + "epoch": 3.35, + "learning_rate": 8.605714769164857e-06, + "loss": 1.0621, + "step": 45324 + }, + { + "epoch": 3.35, + "learning_rate": 8.605319794777066e-06, + "loss": 0.9796, + "step": 45325 + }, + { + "epoch": 3.35, + "learning_rate": 8.604924822608185e-06, + "loss": 0.9853, + "step": 45326 + }, + { + "epoch": 3.35, + "learning_rate": 8.604529852658855e-06, + "loss": 0.9867, + "step": 45327 + }, + { + "epoch": 3.35, + "learning_rate": 8.604134884929692e-06, + "loss": 1.0548, + "step": 45328 + }, + { + "epoch": 3.35, + "learning_rate": 8.603739919421335e-06, + "loss": 0.9777, + "step": 45329 + }, + { + "epoch": 3.35, + "learning_rate": 8.603344956134402e-06, + "loss": 1.0045, + "step": 45330 + }, + { + "epoch": 3.35, + "learning_rate": 8.602949995069537e-06, + "loss": 0.9387, + "step": 45331 + }, + { + "epoch": 3.35, + "learning_rate": 8.602555036227352e-06, + "loss": 0.9879, + "step": 45332 + }, + { + "epoch": 3.35, + "learning_rate": 8.60216007960848e-06, + "loss": 0.9935, + "step": 45333 + }, + { + "epoch": 3.35, + "learning_rate": 8.601765125213557e-06, + "loss": 0.9871, + "step": 45334 + }, + { + "epoch": 3.35, + "learning_rate": 8.601370173043202e-06, + "loss": 0.9361, + "step": 45335 + }, + { + "epoch": 3.35, + "learning_rate": 8.600975223098047e-06, + "loss": 1.018, + "step": 45336 + }, + { + "epoch": 3.35, + "learning_rate": 8.600580275378721e-06, + "loss": 1.1398, + "step": 45337 + }, + { + "epoch": 3.35, + "learning_rate": 8.600185329885853e-06, + "loss": 0.9871, + "step": 45338 + }, + { + "epoch": 3.35, + "learning_rate": 8.599790386620069e-06, + "loss": 1.0045, + "step": 45339 + }, + { + "epoch": 3.35, + "learning_rate": 8.599395445581998e-06, + "loss": 0.9016, + "step": 45340 + }, + { + "epoch": 3.35, + "learning_rate": 8.599000506772269e-06, + "loss": 0.8424, + "step": 45341 + }, + { + "epoch": 3.35, + "learning_rate": 8.59860557019151e-06, + "loss": 1.0012, + "step": 45342 + }, + { + "epoch": 3.35, + "learning_rate": 8.598210635840353e-06, + "loss": 1.0393, + "step": 45343 + }, + { + "epoch": 3.35, + "learning_rate": 8.597815703719416e-06, + "loss": 1.054, + "step": 45344 + }, + { + "epoch": 3.35, + "learning_rate": 8.597420773829343e-06, + "loss": 0.9952, + "step": 45345 + }, + { + "epoch": 3.35, + "learning_rate": 8.597025846170743e-06, + "loss": 0.9939, + "step": 45346 + }, + { + "epoch": 3.35, + "learning_rate": 8.596630920744264e-06, + "loss": 1.0796, + "step": 45347 + }, + { + "epoch": 3.35, + "learning_rate": 8.596235997550519e-06, + "loss": 1.0075, + "step": 45348 + }, + { + "epoch": 3.35, + "learning_rate": 8.595841076590148e-06, + "loss": 0.9146, + "step": 45349 + }, + { + "epoch": 3.35, + "learning_rate": 8.595446157863766e-06, + "loss": 0.9353, + "step": 45350 + }, + { + "epoch": 3.35, + "learning_rate": 8.595051241372017e-06, + "loss": 0.9691, + "step": 45351 + }, + { + "epoch": 3.35, + "learning_rate": 8.594656327115519e-06, + "loss": 1.0014, + "step": 45352 + }, + { + "epoch": 3.35, + "learning_rate": 8.5942614150949e-06, + "loss": 1.0088, + "step": 45353 + }, + { + "epoch": 3.35, + "learning_rate": 8.593866505310794e-06, + "loss": 0.9669, + "step": 45354 + }, + { + "epoch": 3.35, + "learning_rate": 8.593471597763823e-06, + "loss": 1.0079, + "step": 45355 + }, + { + "epoch": 3.35, + "learning_rate": 8.593076692454623e-06, + "loss": 1.159, + "step": 45356 + }, + { + "epoch": 3.35, + "learning_rate": 8.592681789383813e-06, + "loss": 1.0353, + "step": 45357 + }, + { + "epoch": 3.35, + "learning_rate": 8.59228688855203e-06, + "loss": 1.0001, + "step": 45358 + }, + { + "epoch": 3.35, + "learning_rate": 8.591891989959894e-06, + "loss": 1.0026, + "step": 45359 + }, + { + "epoch": 3.35, + "learning_rate": 8.591497093608043e-06, + "loss": 1.0045, + "step": 45360 + }, + { + "epoch": 3.35, + "learning_rate": 8.591102199497098e-06, + "loss": 1.0267, + "step": 45361 + }, + { + "epoch": 3.35, + "learning_rate": 8.590707307627689e-06, + "loss": 1.0389, + "step": 45362 + }, + { + "epoch": 3.35, + "learning_rate": 8.590312418000444e-06, + "loss": 0.9888, + "step": 45363 + }, + { + "epoch": 3.35, + "learning_rate": 8.58991753061599e-06, + "loss": 1.0244, + "step": 45364 + }, + { + "epoch": 3.35, + "learning_rate": 8.58952264547496e-06, + "loss": 0.9769, + "step": 45365 + }, + { + "epoch": 3.35, + "learning_rate": 8.589127762577978e-06, + "loss": 0.9206, + "step": 45366 + }, + { + "epoch": 3.35, + "learning_rate": 8.588732881925674e-06, + "loss": 0.9202, + "step": 45367 + }, + { + "epoch": 3.35, + "learning_rate": 8.588338003518674e-06, + "loss": 0.9749, + "step": 45368 + }, + { + "epoch": 3.35, + "learning_rate": 8.587943127357611e-06, + "loss": 0.9683, + "step": 45369 + }, + { + "epoch": 3.35, + "learning_rate": 8.587548253443113e-06, + "loss": 1.103, + "step": 45370 + }, + { + "epoch": 3.35, + "learning_rate": 8.587153381775799e-06, + "loss": 0.9861, + "step": 45371 + }, + { + "epoch": 3.35, + "learning_rate": 8.586758512356309e-06, + "loss": 0.9829, + "step": 45372 + }, + { + "epoch": 3.35, + "learning_rate": 8.586363645185259e-06, + "loss": 1.0908, + "step": 45373 + }, + { + "epoch": 3.35, + "learning_rate": 8.585968780263291e-06, + "loss": 1.0918, + "step": 45374 + }, + { + "epoch": 3.35, + "learning_rate": 8.585573917591021e-06, + "loss": 1.039, + "step": 45375 + }, + { + "epoch": 3.35, + "learning_rate": 8.585179057169089e-06, + "loss": 0.972, + "step": 45376 + }, + { + "epoch": 3.35, + "learning_rate": 8.584784198998113e-06, + "loss": 0.9553, + "step": 45377 + }, + { + "epoch": 3.35, + "learning_rate": 8.584389343078727e-06, + "loss": 0.9973, + "step": 45378 + }, + { + "epoch": 3.35, + "learning_rate": 8.583994489411556e-06, + "loss": 1.0746, + "step": 45379 + }, + { + "epoch": 3.35, + "learning_rate": 8.58359963799723e-06, + "loss": 1.1, + "step": 45380 + }, + { + "epoch": 3.35, + "learning_rate": 8.583204788836375e-06, + "loss": 0.9839, + "step": 45381 + }, + { + "epoch": 3.35, + "learning_rate": 8.582809941929621e-06, + "loss": 1.0514, + "step": 45382 + }, + { + "epoch": 3.35, + "learning_rate": 8.5824150972776e-06, + "loss": 1.0879, + "step": 45383 + }, + { + "epoch": 3.35, + "learning_rate": 8.582020254880931e-06, + "loss": 0.9734, + "step": 45384 + }, + { + "epoch": 3.35, + "learning_rate": 8.58162541474025e-06, + "loss": 0.9532, + "step": 45385 + }, + { + "epoch": 3.35, + "learning_rate": 8.581230576856182e-06, + "loss": 1.0791, + "step": 45386 + }, + { + "epoch": 3.35, + "learning_rate": 8.580835741229357e-06, + "loss": 0.946, + "step": 45387 + }, + { + "epoch": 3.35, + "learning_rate": 8.580440907860404e-06, + "loss": 1.0495, + "step": 45388 + }, + { + "epoch": 3.35, + "learning_rate": 8.580046076749943e-06, + "loss": 0.9828, + "step": 45389 + }, + { + "epoch": 3.35, + "learning_rate": 8.579651247898616e-06, + "loss": 1.0373, + "step": 45390 + }, + { + "epoch": 3.35, + "learning_rate": 8.579256421307036e-06, + "loss": 0.9062, + "step": 45391 + }, + { + "epoch": 3.35, + "learning_rate": 8.578861596975846e-06, + "loss": 1.0341, + "step": 45392 + }, + { + "epoch": 3.35, + "learning_rate": 8.578466774905659e-06, + "loss": 0.9515, + "step": 45393 + }, + { + "epoch": 3.35, + "learning_rate": 8.57807195509712e-06, + "loss": 1.0158, + "step": 45394 + }, + { + "epoch": 3.35, + "learning_rate": 8.577677137550839e-06, + "loss": 1.0846, + "step": 45395 + }, + { + "epoch": 3.35, + "learning_rate": 8.57728232226746e-06, + "loss": 1.0564, + "step": 45396 + }, + { + "epoch": 3.35, + "learning_rate": 8.576887509247599e-06, + "loss": 0.9322, + "step": 45397 + }, + { + "epoch": 3.35, + "learning_rate": 8.576492698491896e-06, + "loss": 1.0863, + "step": 45398 + }, + { + "epoch": 3.35, + "learning_rate": 8.576097890000971e-06, + "loss": 0.9858, + "step": 45399 + }, + { + "epoch": 3.35, + "learning_rate": 8.575703083775451e-06, + "loss": 0.9947, + "step": 45400 + }, + { + "epoch": 3.35, + "learning_rate": 8.57530827981597e-06, + "loss": 1.0322, + "step": 45401 + }, + { + "epoch": 3.35, + "learning_rate": 8.57491347812315e-06, + "loss": 0.9548, + "step": 45402 + }, + { + "epoch": 3.35, + "learning_rate": 8.574518678697624e-06, + "loss": 1.0837, + "step": 45403 + }, + { + "epoch": 3.35, + "learning_rate": 8.574123881540017e-06, + "loss": 0.9876, + "step": 45404 + }, + { + "epoch": 3.35, + "learning_rate": 8.573729086650962e-06, + "loss": 1.0401, + "step": 45405 + }, + { + "epoch": 3.36, + "learning_rate": 8.57333429403108e-06, + "loss": 0.9185, + "step": 45406 + }, + { + "epoch": 3.36, + "learning_rate": 8.572939503681004e-06, + "loss": 0.9272, + "step": 45407 + }, + { + "epoch": 3.36, + "learning_rate": 8.572544715601362e-06, + "loss": 1.0276, + "step": 45408 + }, + { + "epoch": 3.36, + "learning_rate": 8.572149929792779e-06, + "loss": 1.1209, + "step": 45409 + }, + { + "epoch": 3.36, + "learning_rate": 8.571755146255886e-06, + "loss": 1.056, + "step": 45410 + }, + { + "epoch": 3.36, + "learning_rate": 8.57136036499131e-06, + "loss": 0.9494, + "step": 45411 + }, + { + "epoch": 3.36, + "learning_rate": 8.570965585999679e-06, + "loss": 1.0952, + "step": 45412 + }, + { + "epoch": 3.36, + "learning_rate": 8.570570809281619e-06, + "loss": 0.9859, + "step": 45413 + }, + { + "epoch": 3.36, + "learning_rate": 8.570176034837765e-06, + "loss": 1.077, + "step": 45414 + }, + { + "epoch": 3.36, + "learning_rate": 8.569781262668737e-06, + "loss": 1.119, + "step": 45415 + }, + { + "epoch": 3.36, + "learning_rate": 8.56938649277517e-06, + "loss": 1.0488, + "step": 45416 + }, + { + "epoch": 3.36, + "learning_rate": 8.568991725157689e-06, + "loss": 0.933, + "step": 45417 + }, + { + "epoch": 3.36, + "learning_rate": 8.568596959816914e-06, + "loss": 0.9459, + "step": 45418 + }, + { + "epoch": 3.36, + "learning_rate": 8.568202196753491e-06, + "loss": 1.0634, + "step": 45419 + }, + { + "epoch": 3.36, + "learning_rate": 8.567807435968031e-06, + "loss": 0.9645, + "step": 45420 + }, + { + "epoch": 3.36, + "learning_rate": 8.567412677461173e-06, + "loss": 1.0457, + "step": 45421 + }, + { + "epoch": 3.36, + "learning_rate": 8.567017921233536e-06, + "loss": 0.9775, + "step": 45422 + }, + { + "epoch": 3.36, + "learning_rate": 8.566623167285758e-06, + "loss": 1.0951, + "step": 45423 + }, + { + "epoch": 3.36, + "learning_rate": 8.566228415618457e-06, + "loss": 1.0042, + "step": 45424 + }, + { + "epoch": 3.36, + "learning_rate": 8.565833666232271e-06, + "loss": 0.9808, + "step": 45425 + }, + { + "epoch": 3.36, + "learning_rate": 8.565438919127821e-06, + "loss": 0.9907, + "step": 45426 + }, + { + "epoch": 3.36, + "learning_rate": 8.565044174305735e-06, + "loss": 1.0619, + "step": 45427 + }, + { + "epoch": 3.36, + "learning_rate": 8.564649431766646e-06, + "loss": 1.0055, + "step": 45428 + }, + { + "epoch": 3.36, + "learning_rate": 8.564254691511177e-06, + "loss": 1.0734, + "step": 45429 + }, + { + "epoch": 3.36, + "learning_rate": 8.563859953539962e-06, + "loss": 0.8505, + "step": 45430 + }, + { + "epoch": 3.36, + "learning_rate": 8.563465217853621e-06, + "loss": 1.0105, + "step": 45431 + }, + { + "epoch": 3.36, + "learning_rate": 8.563070484452789e-06, + "loss": 1.0058, + "step": 45432 + }, + { + "epoch": 3.36, + "learning_rate": 8.56267575333809e-06, + "loss": 0.9666, + "step": 45433 + }, + { + "epoch": 3.36, + "learning_rate": 8.562281024510154e-06, + "loss": 0.9674, + "step": 45434 + }, + { + "epoch": 3.36, + "learning_rate": 8.561886297969611e-06, + "loss": 0.9358, + "step": 45435 + }, + { + "epoch": 3.36, + "learning_rate": 8.56149157371708e-06, + "loss": 0.9838, + "step": 45436 + }, + { + "epoch": 3.36, + "learning_rate": 8.561096851753202e-06, + "loss": 1.0646, + "step": 45437 + }, + { + "epoch": 3.36, + "learning_rate": 8.560702132078593e-06, + "loss": 0.9458, + "step": 45438 + }, + { + "epoch": 3.36, + "learning_rate": 8.560307414693891e-06, + "loss": 1.0212, + "step": 45439 + }, + { + "epoch": 3.36, + "learning_rate": 8.559912699599712e-06, + "loss": 1.0203, + "step": 45440 + }, + { + "epoch": 3.36, + "learning_rate": 8.559517986796699e-06, + "loss": 0.8587, + "step": 45441 + }, + { + "epoch": 3.36, + "learning_rate": 8.559123276285466e-06, + "loss": 1.0284, + "step": 45442 + }, + { + "epoch": 3.36, + "learning_rate": 8.558728568066654e-06, + "loss": 1.1394, + "step": 45443 + }, + { + "epoch": 3.36, + "learning_rate": 8.558333862140882e-06, + "loss": 1.0419, + "step": 45444 + }, + { + "epoch": 3.36, + "learning_rate": 8.557939158508779e-06, + "loss": 1.005, + "step": 45445 + }, + { + "epoch": 3.36, + "learning_rate": 8.557544457170974e-06, + "loss": 1.0638, + "step": 45446 + }, + { + "epoch": 3.36, + "learning_rate": 8.557149758128093e-06, + "loss": 0.9821, + "step": 45447 + }, + { + "epoch": 3.36, + "learning_rate": 8.55675506138077e-06, + "loss": 0.9737, + "step": 45448 + }, + { + "epoch": 3.36, + "learning_rate": 8.556360366929625e-06, + "loss": 1.0046, + "step": 45449 + }, + { + "epoch": 3.36, + "learning_rate": 8.555965674775296e-06, + "loss": 0.9997, + "step": 45450 + }, + { + "epoch": 3.36, + "learning_rate": 8.5555709849184e-06, + "loss": 1.011, + "step": 45451 + }, + { + "epoch": 3.36, + "learning_rate": 8.555176297359571e-06, + "loss": 0.9148, + "step": 45452 + }, + { + "epoch": 3.36, + "learning_rate": 8.554781612099434e-06, + "loss": 1.0399, + "step": 45453 + }, + { + "epoch": 3.36, + "learning_rate": 8.554386929138623e-06, + "loss": 0.9608, + "step": 45454 + }, + { + "epoch": 3.36, + "learning_rate": 8.55399224847776e-06, + "loss": 1.127, + "step": 45455 + }, + { + "epoch": 3.36, + "learning_rate": 8.553597570117474e-06, + "loss": 0.976, + "step": 45456 + }, + { + "epoch": 3.36, + "learning_rate": 8.553202894058394e-06, + "loss": 0.9444, + "step": 45457 + }, + { + "epoch": 3.36, + "learning_rate": 8.552808220301146e-06, + "loss": 0.9676, + "step": 45458 + }, + { + "epoch": 3.36, + "learning_rate": 8.55241354884636e-06, + "loss": 1.0289, + "step": 45459 + }, + { + "epoch": 3.36, + "learning_rate": 8.552018879694663e-06, + "loss": 0.9392, + "step": 45460 + }, + { + "epoch": 3.36, + "learning_rate": 8.551624212846686e-06, + "loss": 1.0331, + "step": 45461 + }, + { + "epoch": 3.36, + "learning_rate": 8.551229548303051e-06, + "loss": 0.9848, + "step": 45462 + }, + { + "epoch": 3.36, + "learning_rate": 8.550834886064392e-06, + "loss": 1.0139, + "step": 45463 + }, + { + "epoch": 3.36, + "learning_rate": 8.550440226131335e-06, + "loss": 1.0201, + "step": 45464 + }, + { + "epoch": 3.36, + "learning_rate": 8.550045568504503e-06, + "loss": 1.0299, + "step": 45465 + }, + { + "epoch": 3.36, + "learning_rate": 8.54965091318453e-06, + "loss": 0.9817, + "step": 45466 + }, + { + "epoch": 3.36, + "learning_rate": 8.549256260172039e-06, + "loss": 0.9448, + "step": 45467 + }, + { + "epoch": 3.36, + "learning_rate": 8.548861609467665e-06, + "loss": 1.0314, + "step": 45468 + }, + { + "epoch": 3.36, + "learning_rate": 8.548466961072027e-06, + "loss": 1.0072, + "step": 45469 + }, + { + "epoch": 3.36, + "learning_rate": 8.54807231498576e-06, + "loss": 0.9341, + "step": 45470 + }, + { + "epoch": 3.36, + "learning_rate": 8.547677671209487e-06, + "loss": 1.0857, + "step": 45471 + }, + { + "epoch": 3.36, + "learning_rate": 8.547283029743839e-06, + "loss": 0.9487, + "step": 45472 + }, + { + "epoch": 3.36, + "learning_rate": 8.546888390589445e-06, + "loss": 0.9261, + "step": 45473 + }, + { + "epoch": 3.36, + "learning_rate": 8.546493753746928e-06, + "loss": 0.9668, + "step": 45474 + }, + { + "epoch": 3.36, + "learning_rate": 8.54609911921692e-06, + "loss": 0.9335, + "step": 45475 + }, + { + "epoch": 3.36, + "learning_rate": 8.545704487000046e-06, + "loss": 1.0148, + "step": 45476 + }, + { + "epoch": 3.36, + "learning_rate": 8.545309857096938e-06, + "loss": 0.9914, + "step": 45477 + }, + { + "epoch": 3.36, + "learning_rate": 8.544915229508217e-06, + "loss": 0.9083, + "step": 45478 + }, + { + "epoch": 3.36, + "learning_rate": 8.544520604234518e-06, + "loss": 1.0569, + "step": 45479 + }, + { + "epoch": 3.36, + "learning_rate": 8.544125981276465e-06, + "loss": 1.0572, + "step": 45480 + }, + { + "epoch": 3.36, + "learning_rate": 8.543731360634688e-06, + "loss": 0.9228, + "step": 45481 + }, + { + "epoch": 3.36, + "learning_rate": 8.543336742309816e-06, + "loss": 0.98, + "step": 45482 + }, + { + "epoch": 3.36, + "learning_rate": 8.542942126302468e-06, + "loss": 1.0148, + "step": 45483 + }, + { + "epoch": 3.36, + "learning_rate": 8.542547512613284e-06, + "loss": 1.0828, + "step": 45484 + }, + { + "epoch": 3.36, + "learning_rate": 8.542152901242881e-06, + "loss": 0.9809, + "step": 45485 + }, + { + "epoch": 3.36, + "learning_rate": 8.541758292191898e-06, + "loss": 0.9951, + "step": 45486 + }, + { + "epoch": 3.36, + "learning_rate": 8.541363685460948e-06, + "loss": 1.0095, + "step": 45487 + }, + { + "epoch": 3.36, + "learning_rate": 8.540969081050675e-06, + "loss": 0.9422, + "step": 45488 + }, + { + "epoch": 3.36, + "learning_rate": 8.540574478961693e-06, + "loss": 0.9943, + "step": 45489 + }, + { + "epoch": 3.36, + "learning_rate": 8.540179879194645e-06, + "loss": 0.9429, + "step": 45490 + }, + { + "epoch": 3.36, + "learning_rate": 8.539785281750146e-06, + "loss": 1.0342, + "step": 45491 + }, + { + "epoch": 3.36, + "learning_rate": 8.539390686628826e-06, + "loss": 1.0111, + "step": 45492 + }, + { + "epoch": 3.36, + "learning_rate": 8.538996093831316e-06, + "loss": 1.019, + "step": 45493 + }, + { + "epoch": 3.36, + "learning_rate": 8.538601503358239e-06, + "loss": 1.0294, + "step": 45494 + }, + { + "epoch": 3.36, + "learning_rate": 8.53820691521023e-06, + "loss": 1.0675, + "step": 45495 + }, + { + "epoch": 3.36, + "learning_rate": 8.53781232938791e-06, + "loss": 1.0247, + "step": 45496 + }, + { + "epoch": 3.36, + "learning_rate": 8.537417745891911e-06, + "loss": 1.1149, + "step": 45497 + }, + { + "epoch": 3.36, + "learning_rate": 8.537023164722859e-06, + "loss": 1.0337, + "step": 45498 + }, + { + "epoch": 3.36, + "learning_rate": 8.536628585881384e-06, + "loss": 0.8996, + "step": 45499 + }, + { + "epoch": 3.36, + "learning_rate": 8.536234009368111e-06, + "loss": 1.0639, + "step": 45500 + }, + { + "epoch": 3.36, + "learning_rate": 8.535839435183669e-06, + "loss": 1.0106, + "step": 45501 + }, + { + "epoch": 3.36, + "learning_rate": 8.535444863328685e-06, + "loss": 1.0112, + "step": 45502 + }, + { + "epoch": 3.36, + "learning_rate": 8.535050293803787e-06, + "loss": 0.8904, + "step": 45503 + }, + { + "epoch": 3.36, + "learning_rate": 8.534655726609604e-06, + "loss": 1.045, + "step": 45504 + }, + { + "epoch": 3.36, + "learning_rate": 8.534261161746761e-06, + "loss": 0.8431, + "step": 45505 + }, + { + "epoch": 3.36, + "learning_rate": 8.53386659921589e-06, + "loss": 0.9834, + "step": 45506 + }, + { + "epoch": 3.36, + "learning_rate": 8.533472039017613e-06, + "loss": 1.0772, + "step": 45507 + }, + { + "epoch": 3.36, + "learning_rate": 8.533077481152563e-06, + "loss": 0.9975, + "step": 45508 + }, + { + "epoch": 3.36, + "learning_rate": 8.532682925621365e-06, + "loss": 0.897, + "step": 45509 + }, + { + "epoch": 3.36, + "learning_rate": 8.532288372424651e-06, + "loss": 0.8843, + "step": 45510 + }, + { + "epoch": 3.36, + "learning_rate": 8.531893821563044e-06, + "loss": 0.981, + "step": 45511 + }, + { + "epoch": 3.36, + "learning_rate": 8.53149927303717e-06, + "loss": 1.0038, + "step": 45512 + }, + { + "epoch": 3.36, + "learning_rate": 8.53110472684766e-06, + "loss": 1.1202, + "step": 45513 + }, + { + "epoch": 3.36, + "learning_rate": 8.530710182995143e-06, + "loss": 0.9951, + "step": 45514 + }, + { + "epoch": 3.36, + "learning_rate": 8.530315641480245e-06, + "loss": 0.9793, + "step": 45515 + }, + { + "epoch": 3.36, + "learning_rate": 8.529921102303591e-06, + "loss": 1.091, + "step": 45516 + }, + { + "epoch": 3.36, + "learning_rate": 8.529526565465814e-06, + "loss": 1.0061, + "step": 45517 + }, + { + "epoch": 3.36, + "learning_rate": 8.529132030967538e-06, + "loss": 1.0807, + "step": 45518 + }, + { + "epoch": 3.36, + "learning_rate": 8.528737498809395e-06, + "loss": 0.9951, + "step": 45519 + }, + { + "epoch": 3.36, + "learning_rate": 8.528342968992008e-06, + "loss": 0.9013, + "step": 45520 + }, + { + "epoch": 3.36, + "learning_rate": 8.527948441516003e-06, + "loss": 0.9292, + "step": 45521 + }, + { + "epoch": 3.36, + "learning_rate": 8.527553916382014e-06, + "loss": 0.9323, + "step": 45522 + }, + { + "epoch": 3.36, + "learning_rate": 8.527159393590666e-06, + "loss": 0.9495, + "step": 45523 + }, + { + "epoch": 3.36, + "learning_rate": 8.526764873142586e-06, + "loss": 0.8813, + "step": 45524 + }, + { + "epoch": 3.36, + "learning_rate": 8.5263703550384e-06, + "loss": 0.9446, + "step": 45525 + }, + { + "epoch": 3.36, + "learning_rate": 8.52597583927874e-06, + "loss": 0.9377, + "step": 45526 + }, + { + "epoch": 3.36, + "learning_rate": 8.525581325864229e-06, + "loss": 0.9899, + "step": 45527 + }, + { + "epoch": 3.36, + "learning_rate": 8.5251868147955e-06, + "loss": 0.9863, + "step": 45528 + }, + { + "epoch": 3.36, + "learning_rate": 8.524792306073178e-06, + "loss": 1.0156, + "step": 45529 + }, + { + "epoch": 3.36, + "learning_rate": 8.524397799697886e-06, + "loss": 1.0662, + "step": 45530 + }, + { + "epoch": 3.36, + "learning_rate": 8.52400329567026e-06, + "loss": 0.9811, + "step": 45531 + }, + { + "epoch": 3.36, + "learning_rate": 8.52360879399092e-06, + "loss": 1.0432, + "step": 45532 + }, + { + "epoch": 3.36, + "learning_rate": 8.523214294660502e-06, + "loss": 1.0147, + "step": 45533 + }, + { + "epoch": 3.36, + "learning_rate": 8.522819797679625e-06, + "loss": 0.9374, + "step": 45534 + }, + { + "epoch": 3.36, + "learning_rate": 8.522425303048925e-06, + "loss": 1.0895, + "step": 45535 + }, + { + "epoch": 3.36, + "learning_rate": 8.522030810769018e-06, + "loss": 0.9058, + "step": 45536 + }, + { + "epoch": 3.36, + "learning_rate": 8.521636320840546e-06, + "loss": 0.9953, + "step": 45537 + }, + { + "epoch": 3.36, + "learning_rate": 8.521241833264126e-06, + "loss": 1.0454, + "step": 45538 + }, + { + "epoch": 3.36, + "learning_rate": 8.52084734804039e-06, + "loss": 1.006, + "step": 45539 + }, + { + "epoch": 3.36, + "learning_rate": 8.520452865169965e-06, + "loss": 1.0333, + "step": 45540 + }, + { + "epoch": 3.36, + "learning_rate": 8.520058384653476e-06, + "loss": 1.0752, + "step": 45541 + }, + { + "epoch": 3.37, + "learning_rate": 8.519663906491556e-06, + "loss": 1.0814, + "step": 45542 + }, + { + "epoch": 3.37, + "learning_rate": 8.519269430684827e-06, + "loss": 1.0176, + "step": 45543 + }, + { + "epoch": 3.37, + "learning_rate": 8.51887495723392e-06, + "loss": 0.9482, + "step": 45544 + }, + { + "epoch": 3.37, + "learning_rate": 8.518480486139461e-06, + "loss": 1.096, + "step": 45545 + }, + { + "epoch": 3.37, + "learning_rate": 8.51808601740208e-06, + "loss": 1.0792, + "step": 45546 + }, + { + "epoch": 3.37, + "learning_rate": 8.517691551022403e-06, + "loss": 1.0818, + "step": 45547 + }, + { + "epoch": 3.37, + "learning_rate": 8.517297087001057e-06, + "loss": 1.0104, + "step": 45548 + }, + { + "epoch": 3.37, + "learning_rate": 8.51690262533867e-06, + "loss": 1.0725, + "step": 45549 + }, + { + "epoch": 3.37, + "learning_rate": 8.516508166035868e-06, + "loss": 0.978, + "step": 45550 + }, + { + "epoch": 3.37, + "learning_rate": 8.516113709093283e-06, + "loss": 0.9426, + "step": 45551 + }, + { + "epoch": 3.37, + "learning_rate": 8.515719254511538e-06, + "loss": 0.9663, + "step": 45552 + }, + { + "epoch": 3.37, + "learning_rate": 8.515324802291266e-06, + "loss": 1.1045, + "step": 45553 + }, + { + "epoch": 3.37, + "learning_rate": 8.514930352433085e-06, + "loss": 0.9824, + "step": 45554 + }, + { + "epoch": 3.37, + "learning_rate": 8.514535904937636e-06, + "loss": 1.0978, + "step": 45555 + }, + { + "epoch": 3.37, + "learning_rate": 8.514141459805534e-06, + "loss": 0.9709, + "step": 45556 + }, + { + "epoch": 3.37, + "learning_rate": 8.513747017037412e-06, + "loss": 1.0871, + "step": 45557 + }, + { + "epoch": 3.37, + "learning_rate": 8.513352576633899e-06, + "loss": 1.0914, + "step": 45558 + }, + { + "epoch": 3.37, + "learning_rate": 8.512958138595617e-06, + "loss": 1.0573, + "step": 45559 + }, + { + "epoch": 3.37, + "learning_rate": 8.512563702923203e-06, + "loss": 0.9779, + "step": 45560 + }, + { + "epoch": 3.37, + "learning_rate": 8.512169269617273e-06, + "loss": 0.8708, + "step": 45561 + }, + { + "epoch": 3.37, + "learning_rate": 8.511774838678464e-06, + "loss": 0.9514, + "step": 45562 + }, + { + "epoch": 3.37, + "learning_rate": 8.511380410107398e-06, + "loss": 1.0534, + "step": 45563 + }, + { + "epoch": 3.37, + "learning_rate": 8.510985983904707e-06, + "loss": 0.9409, + "step": 45564 + }, + { + "epoch": 3.37, + "learning_rate": 8.510591560071014e-06, + "loss": 1.0617, + "step": 45565 + }, + { + "epoch": 3.37, + "learning_rate": 8.510197138606949e-06, + "loss": 0.8555, + "step": 45566 + }, + { + "epoch": 3.37, + "learning_rate": 8.50980271951314e-06, + "loss": 1.0024, + "step": 45567 + }, + { + "epoch": 3.37, + "learning_rate": 8.509408302790211e-06, + "loss": 1.0025, + "step": 45568 + }, + { + "epoch": 3.37, + "learning_rate": 8.509013888438796e-06, + "loss": 0.9781, + "step": 45569 + }, + { + "epoch": 3.37, + "learning_rate": 8.508619476459515e-06, + "loss": 1.0159, + "step": 45570 + }, + { + "epoch": 3.37, + "learning_rate": 8.508225066853001e-06, + "loss": 0.9678, + "step": 45571 + }, + { + "epoch": 3.37, + "learning_rate": 8.507830659619876e-06, + "loss": 1.0277, + "step": 45572 + }, + { + "epoch": 3.37, + "learning_rate": 8.507436254760776e-06, + "loss": 1.0261, + "step": 45573 + }, + { + "epoch": 3.37, + "learning_rate": 8.50704185227632e-06, + "loss": 1.018, + "step": 45574 + }, + { + "epoch": 3.37, + "learning_rate": 8.506647452167142e-06, + "loss": 1.0933, + "step": 45575 + }, + { + "epoch": 3.37, + "learning_rate": 8.506253054433868e-06, + "loss": 1.044, + "step": 45576 + }, + { + "epoch": 3.37, + "learning_rate": 8.505858659077117e-06, + "loss": 1.0205, + "step": 45577 + }, + { + "epoch": 3.37, + "learning_rate": 8.505464266097532e-06, + "loss": 1.0995, + "step": 45578 + }, + { + "epoch": 3.37, + "learning_rate": 8.505069875495724e-06, + "loss": 1.0697, + "step": 45579 + }, + { + "epoch": 3.37, + "learning_rate": 8.504675487272335e-06, + "loss": 1.0144, + "step": 45580 + }, + { + "epoch": 3.37, + "learning_rate": 8.50428110142798e-06, + "loss": 1.0547, + "step": 45581 + }, + { + "epoch": 3.37, + "learning_rate": 8.503886717963298e-06, + "loss": 1.0098, + "step": 45582 + }, + { + "epoch": 3.37, + "learning_rate": 8.503492336878905e-06, + "loss": 0.9786, + "step": 45583 + }, + { + "epoch": 3.37, + "learning_rate": 8.503097958175442e-06, + "loss": 0.9472, + "step": 45584 + }, + { + "epoch": 3.37, + "learning_rate": 8.502703581853524e-06, + "loss": 0.9728, + "step": 45585 + }, + { + "epoch": 3.37, + "learning_rate": 8.50230920791378e-06, + "loss": 1.0307, + "step": 45586 + }, + { + "epoch": 3.37, + "learning_rate": 8.501914836356846e-06, + "loss": 1.0156, + "step": 45587 + }, + { + "epoch": 3.37, + "learning_rate": 8.501520467183342e-06, + "loss": 0.9312, + "step": 45588 + }, + { + "epoch": 3.37, + "learning_rate": 8.501126100393897e-06, + "loss": 0.9801, + "step": 45589 + }, + { + "epoch": 3.37, + "learning_rate": 8.50073173598914e-06, + "loss": 0.9251, + "step": 45590 + }, + { + "epoch": 3.37, + "learning_rate": 8.500337373969697e-06, + "loss": 1.1047, + "step": 45591 + }, + { + "epoch": 3.37, + "learning_rate": 8.499943014336194e-06, + "loss": 1.0582, + "step": 45592 + }, + { + "epoch": 3.37, + "learning_rate": 8.499548657089263e-06, + "loss": 1.0293, + "step": 45593 + }, + { + "epoch": 3.37, + "learning_rate": 8.499154302229528e-06, + "loss": 0.9571, + "step": 45594 + }, + { + "epoch": 3.37, + "learning_rate": 8.498759949757615e-06, + "loss": 1.1033, + "step": 45595 + }, + { + "epoch": 3.37, + "learning_rate": 8.498365599674157e-06, + "loss": 1.0859, + "step": 45596 + }, + { + "epoch": 3.37, + "learning_rate": 8.497971251979774e-06, + "loss": 1.0279, + "step": 45597 + }, + { + "epoch": 3.37, + "learning_rate": 8.497576906675103e-06, + "loss": 0.9994, + "step": 45598 + }, + { + "epoch": 3.37, + "learning_rate": 8.497182563760757e-06, + "loss": 1.0857, + "step": 45599 + }, + { + "epoch": 3.37, + "learning_rate": 8.496788223237382e-06, + "loss": 1.0158, + "step": 45600 + }, + { + "epoch": 3.37, + "learning_rate": 8.496393885105585e-06, + "loss": 0.982, + "step": 45601 + }, + { + "epoch": 3.37, + "learning_rate": 8.495999549366012e-06, + "loss": 0.9507, + "step": 45602 + }, + { + "epoch": 3.37, + "learning_rate": 8.495605216019281e-06, + "loss": 1.0912, + "step": 45603 + }, + { + "epoch": 3.37, + "learning_rate": 8.495210885066016e-06, + "loss": 1.1586, + "step": 45604 + }, + { + "epoch": 3.37, + "learning_rate": 8.494816556506853e-06, + "loss": 1.0575, + "step": 45605 + }, + { + "epoch": 3.37, + "learning_rate": 8.494422230342414e-06, + "loss": 1.0976, + "step": 45606 + }, + { + "epoch": 3.37, + "learning_rate": 8.494027906573328e-06, + "loss": 1.0454, + "step": 45607 + }, + { + "epoch": 3.37, + "learning_rate": 8.493633585200221e-06, + "loss": 0.9654, + "step": 45608 + }, + { + "epoch": 3.37, + "learning_rate": 8.493239266223723e-06, + "loss": 1.0342, + "step": 45609 + }, + { + "epoch": 3.37, + "learning_rate": 8.492844949644458e-06, + "loss": 1.0319, + "step": 45610 + }, + { + "epoch": 3.37, + "learning_rate": 8.492450635463057e-06, + "loss": 0.8979, + "step": 45611 + }, + { + "epoch": 3.37, + "learning_rate": 8.492056323680142e-06, + "loss": 0.9155, + "step": 45612 + }, + { + "epoch": 3.37, + "learning_rate": 8.491662014296347e-06, + "loss": 1.0383, + "step": 45613 + }, + { + "epoch": 3.37, + "learning_rate": 8.491267707312297e-06, + "loss": 1.044, + "step": 45614 + }, + { + "epoch": 3.37, + "learning_rate": 8.490873402728616e-06, + "loss": 1.0311, + "step": 45615 + }, + { + "epoch": 3.37, + "learning_rate": 8.490479100545935e-06, + "loss": 1.0093, + "step": 45616 + }, + { + "epoch": 3.37, + "learning_rate": 8.49008480076488e-06, + "loss": 1.0194, + "step": 45617 + }, + { + "epoch": 3.37, + "learning_rate": 8.489690503386079e-06, + "loss": 0.9869, + "step": 45618 + }, + { + "epoch": 3.37, + "learning_rate": 8.489296208410157e-06, + "loss": 0.9781, + "step": 45619 + }, + { + "epoch": 3.37, + "learning_rate": 8.488901915837746e-06, + "loss": 1.0698, + "step": 45620 + }, + { + "epoch": 3.37, + "learning_rate": 8.488507625669466e-06, + "loss": 1.1387, + "step": 45621 + }, + { + "epoch": 3.37, + "learning_rate": 8.488113337905954e-06, + "loss": 1.0333, + "step": 45622 + }, + { + "epoch": 3.37, + "learning_rate": 8.487719052547832e-06, + "loss": 1.0175, + "step": 45623 + }, + { + "epoch": 3.37, + "learning_rate": 8.487324769595721e-06, + "loss": 0.9863, + "step": 45624 + }, + { + "epoch": 3.37, + "learning_rate": 8.486930489050263e-06, + "loss": 1.0379, + "step": 45625 + }, + { + "epoch": 3.37, + "learning_rate": 8.486536210912069e-06, + "loss": 1.0851, + "step": 45626 + }, + { + "epoch": 3.37, + "learning_rate": 8.48614193518178e-06, + "loss": 1.022, + "step": 45627 + }, + { + "epoch": 3.37, + "learning_rate": 8.485747661860011e-06, + "loss": 0.9431, + "step": 45628 + }, + { + "epoch": 3.37, + "learning_rate": 8.485353390947404e-06, + "loss": 0.9623, + "step": 45629 + }, + { + "epoch": 3.37, + "learning_rate": 8.484959122444571e-06, + "loss": 1.0532, + "step": 45630 + }, + { + "epoch": 3.37, + "learning_rate": 8.484564856352153e-06, + "loss": 0.9846, + "step": 45631 + }, + { + "epoch": 3.37, + "learning_rate": 8.484170592670768e-06, + "loss": 0.9306, + "step": 45632 + }, + { + "epoch": 3.37, + "learning_rate": 8.483776331401044e-06, + "loss": 1.0396, + "step": 45633 + }, + { + "epoch": 3.37, + "learning_rate": 8.483382072543611e-06, + "loss": 0.9045, + "step": 45634 + }, + { + "epoch": 3.37, + "learning_rate": 8.482987816099095e-06, + "loss": 1.0342, + "step": 45635 + }, + { + "epoch": 3.37, + "learning_rate": 8.482593562068124e-06, + "loss": 0.8291, + "step": 45636 + }, + { + "epoch": 3.37, + "learning_rate": 8.482199310451325e-06, + "loss": 0.8643, + "step": 45637 + }, + { + "epoch": 3.37, + "learning_rate": 8.481805061249326e-06, + "loss": 0.9789, + "step": 45638 + }, + { + "epoch": 3.37, + "learning_rate": 8.481410814462752e-06, + "loss": 1.0268, + "step": 45639 + }, + { + "epoch": 3.37, + "learning_rate": 8.481016570092233e-06, + "loss": 1.0449, + "step": 45640 + }, + { + "epoch": 3.37, + "learning_rate": 8.480622328138395e-06, + "loss": 0.9743, + "step": 45641 + }, + { + "epoch": 3.37, + "learning_rate": 8.480228088601864e-06, + "loss": 0.9876, + "step": 45642 + }, + { + "epoch": 3.37, + "learning_rate": 8.479833851483272e-06, + "loss": 1.0831, + "step": 45643 + }, + { + "epoch": 3.37, + "learning_rate": 8.479439616783236e-06, + "loss": 1.0493, + "step": 45644 + }, + { + "epoch": 3.37, + "learning_rate": 8.479045384502396e-06, + "loss": 0.9645, + "step": 45645 + }, + { + "epoch": 3.37, + "learning_rate": 8.478651154641367e-06, + "loss": 0.9942, + "step": 45646 + }, + { + "epoch": 3.37, + "learning_rate": 8.47825692720079e-06, + "loss": 0.974, + "step": 45647 + }, + { + "epoch": 3.37, + "learning_rate": 8.477862702181277e-06, + "loss": 0.9831, + "step": 45648 + }, + { + "epoch": 3.37, + "learning_rate": 8.477468479583469e-06, + "loss": 0.9489, + "step": 45649 + }, + { + "epoch": 3.37, + "learning_rate": 8.477074259407983e-06, + "loss": 0.9847, + "step": 45650 + }, + { + "epoch": 3.37, + "learning_rate": 8.476680041655449e-06, + "loss": 0.8346, + "step": 45651 + }, + { + "epoch": 3.37, + "learning_rate": 8.476285826326498e-06, + "loss": 1.0261, + "step": 45652 + }, + { + "epoch": 3.37, + "learning_rate": 8.475891613421753e-06, + "loss": 1.0101, + "step": 45653 + }, + { + "epoch": 3.37, + "learning_rate": 8.475497402941844e-06, + "loss": 1.0634, + "step": 45654 + }, + { + "epoch": 3.37, + "learning_rate": 8.475103194887395e-06, + "loss": 1.0258, + "step": 45655 + }, + { + "epoch": 3.37, + "learning_rate": 8.474708989259036e-06, + "loss": 1.05, + "step": 45656 + }, + { + "epoch": 3.37, + "learning_rate": 8.474314786057391e-06, + "loss": 0.922, + "step": 45657 + }, + { + "epoch": 3.37, + "learning_rate": 8.473920585283094e-06, + "loss": 1.0282, + "step": 45658 + }, + { + "epoch": 3.37, + "learning_rate": 8.473526386936765e-06, + "loss": 1.0725, + "step": 45659 + }, + { + "epoch": 3.37, + "learning_rate": 8.473132191019033e-06, + "loss": 0.9856, + "step": 45660 + }, + { + "epoch": 3.37, + "learning_rate": 8.472737997530526e-06, + "loss": 1.0102, + "step": 45661 + }, + { + "epoch": 3.37, + "learning_rate": 8.47234380647187e-06, + "loss": 1.0538, + "step": 45662 + }, + { + "epoch": 3.37, + "learning_rate": 8.471949617843696e-06, + "loss": 0.9697, + "step": 45663 + }, + { + "epoch": 3.37, + "learning_rate": 8.471555431646625e-06, + "loss": 0.9967, + "step": 45664 + }, + { + "epoch": 3.37, + "learning_rate": 8.471161247881291e-06, + "loss": 0.949, + "step": 45665 + }, + { + "epoch": 3.37, + "learning_rate": 8.470767066548314e-06, + "loss": 0.9804, + "step": 45666 + }, + { + "epoch": 3.37, + "learning_rate": 8.470372887648329e-06, + "loss": 1.0749, + "step": 45667 + }, + { + "epoch": 3.37, + "learning_rate": 8.469978711181954e-06, + "loss": 0.9134, + "step": 45668 + }, + { + "epoch": 3.37, + "learning_rate": 8.469584537149824e-06, + "loss": 1.0965, + "step": 45669 + }, + { + "epoch": 3.37, + "learning_rate": 8.469190365552567e-06, + "loss": 1.0582, + "step": 45670 + }, + { + "epoch": 3.37, + "learning_rate": 8.468796196390798e-06, + "loss": 1.0581, + "step": 45671 + }, + { + "epoch": 3.37, + "learning_rate": 8.468402029665158e-06, + "loss": 0.9953, + "step": 45672 + }, + { + "epoch": 3.37, + "learning_rate": 8.468007865376264e-06, + "loss": 0.9399, + "step": 45673 + }, + { + "epoch": 3.37, + "learning_rate": 8.467613703524753e-06, + "loss": 1.047, + "step": 45674 + }, + { + "epoch": 3.37, + "learning_rate": 8.467219544111242e-06, + "loss": 1.0003, + "step": 45675 + }, + { + "epoch": 3.37, + "learning_rate": 8.466825387136366e-06, + "loss": 0.9875, + "step": 45676 + }, + { + "epoch": 3.38, + "learning_rate": 8.466431232600745e-06, + "loss": 1.0151, + "step": 45677 + }, + { + "epoch": 3.38, + "learning_rate": 8.466037080505014e-06, + "loss": 1.0867, + "step": 45678 + }, + { + "epoch": 3.38, + "learning_rate": 8.465642930849796e-06, + "loss": 1.1439, + "step": 45679 + }, + { + "epoch": 3.38, + "learning_rate": 8.465248783635714e-06, + "loss": 1.053, + "step": 45680 + }, + { + "epoch": 3.38, + "learning_rate": 8.464854638863403e-06, + "loss": 0.9847, + "step": 45681 + }, + { + "epoch": 3.38, + "learning_rate": 8.464460496533484e-06, + "loss": 0.9843, + "step": 45682 + }, + { + "epoch": 3.38, + "learning_rate": 8.464066356646588e-06, + "loss": 0.9357, + "step": 45683 + }, + { + "epoch": 3.38, + "learning_rate": 8.46367221920334e-06, + "loss": 0.995, + "step": 45684 + }, + { + "epoch": 3.38, + "learning_rate": 8.463278084204368e-06, + "loss": 1.0389, + "step": 45685 + }, + { + "epoch": 3.38, + "learning_rate": 8.462883951650297e-06, + "loss": 0.9249, + "step": 45686 + }, + { + "epoch": 3.38, + "learning_rate": 8.462489821541757e-06, + "loss": 0.9128, + "step": 45687 + }, + { + "epoch": 3.38, + "learning_rate": 8.462095693879376e-06, + "loss": 1.0263, + "step": 45688 + }, + { + "epoch": 3.38, + "learning_rate": 8.461701568663774e-06, + "loss": 1.029, + "step": 45689 + }, + { + "epoch": 3.38, + "learning_rate": 8.461307445895588e-06, + "loss": 0.9597, + "step": 45690 + }, + { + "epoch": 3.38, + "learning_rate": 8.460913325575432e-06, + "loss": 0.9023, + "step": 45691 + }, + { + "epoch": 3.38, + "learning_rate": 8.46051920770395e-06, + "loss": 0.9804, + "step": 45692 + }, + { + "epoch": 3.38, + "learning_rate": 8.460125092281752e-06, + "loss": 0.9716, + "step": 45693 + }, + { + "epoch": 3.38, + "learning_rate": 8.45973097930948e-06, + "loss": 0.9953, + "step": 45694 + }, + { + "epoch": 3.38, + "learning_rate": 8.459336868787746e-06, + "loss": 1.0643, + "step": 45695 + }, + { + "epoch": 3.38, + "learning_rate": 8.458942760717193e-06, + "loss": 0.9567, + "step": 45696 + }, + { + "epoch": 3.38, + "learning_rate": 8.458548655098437e-06, + "loss": 1.0102, + "step": 45697 + }, + { + "epoch": 3.38, + "learning_rate": 8.458154551932107e-06, + "loss": 1.0648, + "step": 45698 + }, + { + "epoch": 3.38, + "learning_rate": 8.457760451218831e-06, + "loss": 1.0834, + "step": 45699 + }, + { + "epoch": 3.38, + "learning_rate": 8.457366352959236e-06, + "loss": 1.0307, + "step": 45700 + }, + { + "epoch": 3.38, + "learning_rate": 8.45697225715395e-06, + "loss": 0.9423, + "step": 45701 + }, + { + "epoch": 3.38, + "learning_rate": 8.456578163803598e-06, + "loss": 0.9208, + "step": 45702 + }, + { + "epoch": 3.38, + "learning_rate": 8.45618407290881e-06, + "loss": 0.916, + "step": 45703 + }, + { + "epoch": 3.38, + "learning_rate": 8.455789984470209e-06, + "loss": 0.9851, + "step": 45704 + }, + { + "epoch": 3.38, + "learning_rate": 8.455395898488425e-06, + "loss": 0.9426, + "step": 45705 + }, + { + "epoch": 3.38, + "learning_rate": 8.455001814964086e-06, + "loss": 0.9976, + "step": 45706 + }, + { + "epoch": 3.38, + "learning_rate": 8.454607733897812e-06, + "loss": 1.1173, + "step": 45707 + }, + { + "epoch": 3.38, + "learning_rate": 8.45421365529024e-06, + "loss": 0.9767, + "step": 45708 + }, + { + "epoch": 3.38, + "learning_rate": 8.453819579141988e-06, + "loss": 0.9321, + "step": 45709 + }, + { + "epoch": 3.38, + "learning_rate": 8.45342550545369e-06, + "loss": 0.8014, + "step": 45710 + }, + { + "epoch": 3.38, + "learning_rate": 8.453031434225969e-06, + "loss": 1.0832, + "step": 45711 + }, + { + "epoch": 3.38, + "learning_rate": 8.452637365459452e-06, + "loss": 1.0484, + "step": 45712 + }, + { + "epoch": 3.38, + "learning_rate": 8.452243299154767e-06, + "loss": 0.9798, + "step": 45713 + }, + { + "epoch": 3.38, + "learning_rate": 8.451849235312541e-06, + "loss": 0.9604, + "step": 45714 + }, + { + "epoch": 3.38, + "learning_rate": 8.451455173933403e-06, + "loss": 0.9202, + "step": 45715 + }, + { + "epoch": 3.38, + "learning_rate": 8.451061115017973e-06, + "loss": 0.9145, + "step": 45716 + }, + { + "epoch": 3.38, + "learning_rate": 8.450667058566889e-06, + "loss": 1.0504, + "step": 45717 + }, + { + "epoch": 3.38, + "learning_rate": 8.450273004580763e-06, + "loss": 0.8754, + "step": 45718 + }, + { + "epoch": 3.38, + "learning_rate": 8.449878953060239e-06, + "loss": 1.0059, + "step": 45719 + }, + { + "epoch": 3.38, + "learning_rate": 8.44948490400593e-06, + "loss": 1.1018, + "step": 45720 + }, + { + "epoch": 3.38, + "learning_rate": 8.449090857418472e-06, + "loss": 1.0375, + "step": 45721 + }, + { + "epoch": 3.38, + "learning_rate": 8.448696813298483e-06, + "loss": 1.0132, + "step": 45722 + }, + { + "epoch": 3.38, + "learning_rate": 8.4483027716466e-06, + "loss": 0.9643, + "step": 45723 + }, + { + "epoch": 3.38, + "learning_rate": 8.447908732463443e-06, + "loss": 0.9838, + "step": 45724 + }, + { + "epoch": 3.38, + "learning_rate": 8.44751469574964e-06, + "loss": 0.9946, + "step": 45725 + }, + { + "epoch": 3.38, + "learning_rate": 8.447120661505824e-06, + "loss": 1.0969, + "step": 45726 + }, + { + "epoch": 3.38, + "learning_rate": 8.44672662973261e-06, + "loss": 0.9149, + "step": 45727 + }, + { + "epoch": 3.38, + "learning_rate": 8.446332600430636e-06, + "loss": 1.0069, + "step": 45728 + }, + { + "epoch": 3.38, + "learning_rate": 8.445938573600523e-06, + "loss": 0.9798, + "step": 45729 + }, + { + "epoch": 3.38, + "learning_rate": 8.4455445492429e-06, + "loss": 1.0407, + "step": 45730 + }, + { + "epoch": 3.38, + "learning_rate": 8.445150527358393e-06, + "loss": 0.9127, + "step": 45731 + }, + { + "epoch": 3.38, + "learning_rate": 8.44475650794763e-06, + "loss": 0.9889, + "step": 45732 + }, + { + "epoch": 3.38, + "learning_rate": 8.444362491011235e-06, + "loss": 0.9194, + "step": 45733 + }, + { + "epoch": 3.38, + "learning_rate": 8.44396847654984e-06, + "loss": 0.9331, + "step": 45734 + }, + { + "epoch": 3.38, + "learning_rate": 8.44357446456407e-06, + "loss": 1.0136, + "step": 45735 + }, + { + "epoch": 3.38, + "learning_rate": 8.443180455054545e-06, + "loss": 1.0148, + "step": 45736 + }, + { + "epoch": 3.38, + "learning_rate": 8.442786448021904e-06, + "loss": 0.9455, + "step": 45737 + }, + { + "epoch": 3.38, + "learning_rate": 8.442392443466761e-06, + "loss": 0.8983, + "step": 45738 + }, + { + "epoch": 3.38, + "learning_rate": 8.441998441389756e-06, + "loss": 1.0415, + "step": 45739 + }, + { + "epoch": 3.38, + "learning_rate": 8.441604441791504e-06, + "loss": 0.9925, + "step": 45740 + }, + { + "epoch": 3.38, + "learning_rate": 8.441210444672642e-06, + "loss": 1.0361, + "step": 45741 + }, + { + "epoch": 3.38, + "learning_rate": 8.440816450033785e-06, + "loss": 1.0177, + "step": 45742 + }, + { + "epoch": 3.38, + "learning_rate": 8.440422457875574e-06, + "loss": 1.0755, + "step": 45743 + }, + { + "epoch": 3.38, + "learning_rate": 8.440028468198626e-06, + "loss": 0.9588, + "step": 45744 + }, + { + "epoch": 3.38, + "learning_rate": 8.439634481003567e-06, + "loss": 1.0593, + "step": 45745 + }, + { + "epoch": 3.38, + "learning_rate": 8.439240496291032e-06, + "loss": 0.9915, + "step": 45746 + }, + { + "epoch": 3.38, + "learning_rate": 8.43884651406164e-06, + "loss": 1.0318, + "step": 45747 + }, + { + "epoch": 3.38, + "learning_rate": 8.438452534316021e-06, + "loss": 0.9509, + "step": 45748 + }, + { + "epoch": 3.38, + "learning_rate": 8.438058557054802e-06, + "loss": 0.999, + "step": 45749 + }, + { + "epoch": 3.38, + "learning_rate": 8.43766458227861e-06, + "loss": 0.9663, + "step": 45750 + }, + { + "epoch": 3.38, + "learning_rate": 8.43727060998807e-06, + "loss": 0.9444, + "step": 45751 + }, + { + "epoch": 3.38, + "learning_rate": 8.436876640183813e-06, + "loss": 1.1344, + "step": 45752 + }, + { + "epoch": 3.38, + "learning_rate": 8.436482672866462e-06, + "loss": 1.0402, + "step": 45753 + }, + { + "epoch": 3.38, + "learning_rate": 8.436088708036642e-06, + "loss": 1.1023, + "step": 45754 + }, + { + "epoch": 3.38, + "learning_rate": 8.435694745694986e-06, + "loss": 1.0397, + "step": 45755 + }, + { + "epoch": 3.38, + "learning_rate": 8.435300785842114e-06, + "loss": 1.006, + "step": 45756 + }, + { + "epoch": 3.38, + "learning_rate": 8.434906828478658e-06, + "loss": 1.0685, + "step": 45757 + }, + { + "epoch": 3.38, + "learning_rate": 8.434512873605241e-06, + "loss": 1.0393, + "step": 45758 + }, + { + "epoch": 3.38, + "learning_rate": 8.434118921222495e-06, + "loss": 1.0341, + "step": 45759 + }, + { + "epoch": 3.38, + "learning_rate": 8.43372497133104e-06, + "loss": 1.0718, + "step": 45760 + }, + { + "epoch": 3.38, + "learning_rate": 8.433331023931509e-06, + "loss": 1.0665, + "step": 45761 + }, + { + "epoch": 3.38, + "learning_rate": 8.432937079024528e-06, + "loss": 1.0033, + "step": 45762 + }, + { + "epoch": 3.38, + "learning_rate": 8.432543136610717e-06, + "loss": 1.0301, + "step": 45763 + }, + { + "epoch": 3.38, + "learning_rate": 8.432149196690709e-06, + "loss": 0.9673, + "step": 45764 + }, + { + "epoch": 3.38, + "learning_rate": 8.431755259265128e-06, + "loss": 0.9325, + "step": 45765 + }, + { + "epoch": 3.38, + "learning_rate": 8.431361324334604e-06, + "loss": 1.1352, + "step": 45766 + }, + { + "epoch": 3.38, + "learning_rate": 8.430967391899758e-06, + "loss": 1.0463, + "step": 45767 + }, + { + "epoch": 3.38, + "learning_rate": 8.430573461961225e-06, + "loss": 1.0048, + "step": 45768 + }, + { + "epoch": 3.38, + "learning_rate": 8.430179534519625e-06, + "loss": 1.0844, + "step": 45769 + }, + { + "epoch": 3.38, + "learning_rate": 8.429785609575588e-06, + "loss": 0.8698, + "step": 45770 + }, + { + "epoch": 3.38, + "learning_rate": 8.429391687129738e-06, + "loss": 0.9825, + "step": 45771 + }, + { + "epoch": 3.38, + "learning_rate": 8.428997767182704e-06, + "loss": 1.05, + "step": 45772 + }, + { + "epoch": 3.38, + "learning_rate": 8.428603849735114e-06, + "loss": 0.9024, + "step": 45773 + }, + { + "epoch": 3.38, + "learning_rate": 8.428209934787588e-06, + "loss": 1.068, + "step": 45774 + }, + { + "epoch": 3.38, + "learning_rate": 8.427816022340763e-06, + "loss": 0.9284, + "step": 45775 + }, + { + "epoch": 3.38, + "learning_rate": 8.427422112395255e-06, + "loss": 1.0881, + "step": 45776 + }, + { + "epoch": 3.38, + "learning_rate": 8.4270282049517e-06, + "loss": 1.0958, + "step": 45777 + }, + { + "epoch": 3.38, + "learning_rate": 8.426634300010718e-06, + "loss": 1.0075, + "step": 45778 + }, + { + "epoch": 3.38, + "learning_rate": 8.42624039757294e-06, + "loss": 1.03, + "step": 45779 + }, + { + "epoch": 3.38, + "learning_rate": 8.42584649763899e-06, + "loss": 0.9612, + "step": 45780 + }, + { + "epoch": 3.38, + "learning_rate": 8.425452600209496e-06, + "loss": 0.9432, + "step": 45781 + }, + { + "epoch": 3.38, + "learning_rate": 8.425058705285087e-06, + "loss": 0.9311, + "step": 45782 + }, + { + "epoch": 3.38, + "learning_rate": 8.424664812866382e-06, + "loss": 0.9755, + "step": 45783 + }, + { + "epoch": 3.38, + "learning_rate": 8.42427092295402e-06, + "loss": 1.0228, + "step": 45784 + }, + { + "epoch": 3.38, + "learning_rate": 8.42387703554861e-06, + "loss": 1.0982, + "step": 45785 + }, + { + "epoch": 3.38, + "learning_rate": 8.4234831506508e-06, + "loss": 1.0223, + "step": 45786 + }, + { + "epoch": 3.38, + "learning_rate": 8.423089268261196e-06, + "loss": 1.0569, + "step": 45787 + }, + { + "epoch": 3.38, + "learning_rate": 8.422695388380444e-06, + "loss": 1.0037, + "step": 45788 + }, + { + "epoch": 3.38, + "learning_rate": 8.422301511009151e-06, + "loss": 0.9099, + "step": 45789 + }, + { + "epoch": 3.38, + "learning_rate": 8.421907636147963e-06, + "loss": 0.9888, + "step": 45790 + }, + { + "epoch": 3.38, + "learning_rate": 8.421513763797492e-06, + "loss": 1.0007, + "step": 45791 + }, + { + "epoch": 3.38, + "learning_rate": 8.421119893958371e-06, + "loss": 1.0738, + "step": 45792 + }, + { + "epoch": 3.38, + "learning_rate": 8.420726026631227e-06, + "loss": 1.0318, + "step": 45793 + }, + { + "epoch": 3.38, + "learning_rate": 8.420332161816684e-06, + "loss": 0.8935, + "step": 45794 + }, + { + "epoch": 3.38, + "learning_rate": 8.41993829951537e-06, + "loss": 0.922, + "step": 45795 + }, + { + "epoch": 3.38, + "learning_rate": 8.419544439727912e-06, + "loss": 0.9205, + "step": 45796 + }, + { + "epoch": 3.38, + "learning_rate": 8.419150582454936e-06, + "loss": 0.934, + "step": 45797 + }, + { + "epoch": 3.38, + "learning_rate": 8.418756727697067e-06, + "loss": 0.9474, + "step": 45798 + }, + { + "epoch": 3.38, + "learning_rate": 8.418362875454936e-06, + "loss": 0.9031, + "step": 45799 + }, + { + "epoch": 3.38, + "learning_rate": 8.417969025729166e-06, + "loss": 0.9555, + "step": 45800 + }, + { + "epoch": 3.38, + "learning_rate": 8.417575178520385e-06, + "loss": 1.05, + "step": 45801 + }, + { + "epoch": 3.38, + "learning_rate": 8.41718133382922e-06, + "loss": 1.0728, + "step": 45802 + }, + { + "epoch": 3.38, + "learning_rate": 8.416787491656295e-06, + "loss": 0.9909, + "step": 45803 + }, + { + "epoch": 3.38, + "learning_rate": 8.41639365200224e-06, + "loss": 1.0328, + "step": 45804 + }, + { + "epoch": 3.38, + "learning_rate": 8.415999814867678e-06, + "loss": 0.9075, + "step": 45805 + }, + { + "epoch": 3.38, + "learning_rate": 8.41560598025324e-06, + "loss": 1.0565, + "step": 45806 + }, + { + "epoch": 3.38, + "learning_rate": 8.415212148159547e-06, + "loss": 1.0559, + "step": 45807 + }, + { + "epoch": 3.38, + "learning_rate": 8.414818318587235e-06, + "loss": 0.9359, + "step": 45808 + }, + { + "epoch": 3.38, + "learning_rate": 8.414424491536922e-06, + "loss": 1.0116, + "step": 45809 + }, + { + "epoch": 3.38, + "learning_rate": 8.414030667009233e-06, + "loss": 0.9721, + "step": 45810 + }, + { + "epoch": 3.38, + "learning_rate": 8.413636845004802e-06, + "loss": 0.9358, + "step": 45811 + }, + { + "epoch": 3.39, + "learning_rate": 8.413243025524249e-06, + "loss": 0.9816, + "step": 45812 + }, + { + "epoch": 3.39, + "learning_rate": 8.412849208568208e-06, + "loss": 1.0365, + "step": 45813 + }, + { + "epoch": 3.39, + "learning_rate": 8.412455394137297e-06, + "loss": 0.9525, + "step": 45814 + }, + { + "epoch": 3.39, + "learning_rate": 8.412061582232149e-06, + "loss": 1.0362, + "step": 45815 + }, + { + "epoch": 3.39, + "learning_rate": 8.411667772853387e-06, + "loss": 1.02, + "step": 45816 + }, + { + "epoch": 3.39, + "learning_rate": 8.41127396600164e-06, + "loss": 0.9452, + "step": 45817 + }, + { + "epoch": 3.39, + "learning_rate": 8.410880161677535e-06, + "loss": 0.9496, + "step": 45818 + }, + { + "epoch": 3.39, + "learning_rate": 8.410486359881694e-06, + "loss": 0.9276, + "step": 45819 + }, + { + "epoch": 3.39, + "learning_rate": 8.410092560614747e-06, + "loss": 0.983, + "step": 45820 + }, + { + "epoch": 3.39, + "learning_rate": 8.40969876387732e-06, + "loss": 1.0525, + "step": 45821 + }, + { + "epoch": 3.39, + "learning_rate": 8.40930496967004e-06, + "loss": 0.9751, + "step": 45822 + }, + { + "epoch": 3.39, + "learning_rate": 8.408911177993534e-06, + "loss": 1.0474, + "step": 45823 + }, + { + "epoch": 3.39, + "learning_rate": 8.408517388848425e-06, + "loss": 0.9253, + "step": 45824 + }, + { + "epoch": 3.39, + "learning_rate": 8.408123602235343e-06, + "loss": 0.9804, + "step": 45825 + }, + { + "epoch": 3.39, + "learning_rate": 8.407729818154914e-06, + "loss": 1.1113, + "step": 45826 + }, + { + "epoch": 3.39, + "learning_rate": 8.407336036607765e-06, + "loss": 1.0792, + "step": 45827 + }, + { + "epoch": 3.39, + "learning_rate": 8.40694225759452e-06, + "loss": 1.0277, + "step": 45828 + }, + { + "epoch": 3.39, + "learning_rate": 8.406548481115813e-06, + "loss": 0.9051, + "step": 45829 + }, + { + "epoch": 3.39, + "learning_rate": 8.406154707172256e-06, + "loss": 0.9499, + "step": 45830 + }, + { + "epoch": 3.39, + "learning_rate": 8.405760935764489e-06, + "loss": 0.9565, + "step": 45831 + }, + { + "epoch": 3.39, + "learning_rate": 8.40536716689313e-06, + "loss": 0.9998, + "step": 45832 + }, + { + "epoch": 3.39, + "learning_rate": 8.404973400558814e-06, + "loss": 1.0158, + "step": 45833 + }, + { + "epoch": 3.39, + "learning_rate": 8.404579636762156e-06, + "loss": 1.0253, + "step": 45834 + }, + { + "epoch": 3.39, + "learning_rate": 8.404185875503797e-06, + "loss": 1.0484, + "step": 45835 + }, + { + "epoch": 3.39, + "learning_rate": 8.403792116784346e-06, + "loss": 1.069, + "step": 45836 + }, + { + "epoch": 3.39, + "learning_rate": 8.403398360604448e-06, + "loss": 0.9927, + "step": 45837 + }, + { + "epoch": 3.39, + "learning_rate": 8.403004606964716e-06, + "loss": 1.0427, + "step": 45838 + }, + { + "epoch": 3.39, + "learning_rate": 8.40261085586578e-06, + "loss": 0.9923, + "step": 45839 + }, + { + "epoch": 3.39, + "learning_rate": 8.40221710730827e-06, + "loss": 0.9736, + "step": 45840 + }, + { + "epoch": 3.39, + "learning_rate": 8.401823361292808e-06, + "loss": 1.0345, + "step": 45841 + }, + { + "epoch": 3.39, + "learning_rate": 8.401429617820021e-06, + "loss": 1.0256, + "step": 45842 + }, + { + "epoch": 3.39, + "learning_rate": 8.401035876890537e-06, + "loss": 1.0468, + "step": 45843 + }, + { + "epoch": 3.39, + "learning_rate": 8.400642138504985e-06, + "loss": 1.1427, + "step": 45844 + }, + { + "epoch": 3.39, + "learning_rate": 8.400248402663984e-06, + "loss": 1.0861, + "step": 45845 + }, + { + "epoch": 3.39, + "learning_rate": 8.399854669368168e-06, + "loss": 0.9945, + "step": 45846 + }, + { + "epoch": 3.39, + "learning_rate": 8.399460938618161e-06, + "loss": 0.91, + "step": 45847 + }, + { + "epoch": 3.39, + "learning_rate": 8.399067210414587e-06, + "loss": 1.0827, + "step": 45848 + }, + { + "epoch": 3.39, + "learning_rate": 8.398673484758075e-06, + "loss": 1.0798, + "step": 45849 + }, + { + "epoch": 3.39, + "learning_rate": 8.398279761649248e-06, + "loss": 1.054, + "step": 45850 + }, + { + "epoch": 3.39, + "learning_rate": 8.397886041088738e-06, + "loss": 0.8901, + "step": 45851 + }, + { + "epoch": 3.39, + "learning_rate": 8.397492323077167e-06, + "loss": 0.9969, + "step": 45852 + }, + { + "epoch": 3.39, + "learning_rate": 8.397098607615165e-06, + "loss": 0.9734, + "step": 45853 + }, + { + "epoch": 3.39, + "learning_rate": 8.396704894703353e-06, + "loss": 0.9609, + "step": 45854 + }, + { + "epoch": 3.39, + "learning_rate": 8.396311184342365e-06, + "loss": 1.0346, + "step": 45855 + }, + { + "epoch": 3.39, + "learning_rate": 8.39591747653282e-06, + "loss": 0.9212, + "step": 45856 + }, + { + "epoch": 3.39, + "learning_rate": 8.395523771275346e-06, + "loss": 0.9952, + "step": 45857 + }, + { + "epoch": 3.39, + "learning_rate": 8.395130068570574e-06, + "loss": 0.9632, + "step": 45858 + }, + { + "epoch": 3.39, + "learning_rate": 8.394736368419123e-06, + "loss": 1.0265, + "step": 45859 + }, + { + "epoch": 3.39, + "learning_rate": 8.394342670821627e-06, + "loss": 0.8617, + "step": 45860 + }, + { + "epoch": 3.39, + "learning_rate": 8.393948975778707e-06, + "loss": 0.893, + "step": 45861 + }, + { + "epoch": 3.39, + "learning_rate": 8.393555283290991e-06, + "loss": 1.0456, + "step": 45862 + }, + { + "epoch": 3.39, + "learning_rate": 8.393161593359107e-06, + "loss": 0.9874, + "step": 45863 + }, + { + "epoch": 3.39, + "learning_rate": 8.392767905983679e-06, + "loss": 0.9812, + "step": 45864 + }, + { + "epoch": 3.39, + "learning_rate": 8.392374221165336e-06, + "loss": 1.0288, + "step": 45865 + }, + { + "epoch": 3.39, + "learning_rate": 8.3919805389047e-06, + "loss": 1.0385, + "step": 45866 + }, + { + "epoch": 3.39, + "learning_rate": 8.391586859202401e-06, + "loss": 1.0517, + "step": 45867 + }, + { + "epoch": 3.39, + "learning_rate": 8.391193182059064e-06, + "loss": 0.9267, + "step": 45868 + }, + { + "epoch": 3.39, + "learning_rate": 8.390799507475316e-06, + "loss": 1.0475, + "step": 45869 + }, + { + "epoch": 3.39, + "learning_rate": 8.390405835451781e-06, + "loss": 0.9894, + "step": 45870 + }, + { + "epoch": 3.39, + "learning_rate": 8.39001216598909e-06, + "loss": 0.8332, + "step": 45871 + }, + { + "epoch": 3.39, + "learning_rate": 8.389618499087865e-06, + "loss": 1.0683, + "step": 45872 + }, + { + "epoch": 3.39, + "learning_rate": 8.389224834748735e-06, + "loss": 1.0523, + "step": 45873 + }, + { + "epoch": 3.39, + "learning_rate": 8.388831172972329e-06, + "loss": 0.9521, + "step": 45874 + }, + { + "epoch": 3.39, + "learning_rate": 8.388437513759261e-06, + "loss": 1.1104, + "step": 45875 + }, + { + "epoch": 3.39, + "learning_rate": 8.388043857110174e-06, + "loss": 0.9622, + "step": 45876 + }, + { + "epoch": 3.39, + "learning_rate": 8.387650203025679e-06, + "loss": 1.1215, + "step": 45877 + }, + { + "epoch": 3.39, + "learning_rate": 8.387256551506416e-06, + "loss": 1.1385, + "step": 45878 + }, + { + "epoch": 3.39, + "learning_rate": 8.386862902552998e-06, + "loss": 1.0418, + "step": 45879 + }, + { + "epoch": 3.39, + "learning_rate": 8.386469256166065e-06, + "loss": 0.9466, + "step": 45880 + }, + { + "epoch": 3.39, + "learning_rate": 8.386075612346229e-06, + "loss": 1.012, + "step": 45881 + }, + { + "epoch": 3.39, + "learning_rate": 8.385681971094132e-06, + "loss": 1.0088, + "step": 45882 + }, + { + "epoch": 3.39, + "learning_rate": 8.385288332410384e-06, + "loss": 0.9447, + "step": 45883 + }, + { + "epoch": 3.39, + "learning_rate": 8.384894696295625e-06, + "loss": 0.9787, + "step": 45884 + }, + { + "epoch": 3.39, + "learning_rate": 8.384501062750474e-06, + "loss": 1.0443, + "step": 45885 + }, + { + "epoch": 3.39, + "learning_rate": 8.384107431775557e-06, + "loss": 0.9954, + "step": 45886 + }, + { + "epoch": 3.39, + "learning_rate": 8.383713803371504e-06, + "loss": 0.9336, + "step": 45887 + }, + { + "epoch": 3.39, + "learning_rate": 8.383320177538935e-06, + "loss": 1.0025, + "step": 45888 + }, + { + "epoch": 3.39, + "learning_rate": 8.382926554278486e-06, + "loss": 1.0631, + "step": 45889 + }, + { + "epoch": 3.39, + "learning_rate": 8.382532933590774e-06, + "loss": 0.9984, + "step": 45890 + }, + { + "epoch": 3.39, + "learning_rate": 8.38213931547643e-06, + "loss": 1.0908, + "step": 45891 + }, + { + "epoch": 3.39, + "learning_rate": 8.381745699936079e-06, + "loss": 1.0043, + "step": 45892 + }, + { + "epoch": 3.39, + "learning_rate": 8.381352086970349e-06, + "loss": 1.0105, + "step": 45893 + }, + { + "epoch": 3.39, + "learning_rate": 8.380958476579864e-06, + "loss": 1.0337, + "step": 45894 + }, + { + "epoch": 3.39, + "learning_rate": 8.38056486876525e-06, + "loss": 0.9001, + "step": 45895 + }, + { + "epoch": 3.39, + "learning_rate": 8.380171263527138e-06, + "loss": 1.0053, + "step": 45896 + }, + { + "epoch": 3.39, + "learning_rate": 8.379777660866142e-06, + "loss": 1.0142, + "step": 45897 + }, + { + "epoch": 3.39, + "learning_rate": 8.379384060782904e-06, + "loss": 1.0134, + "step": 45898 + }, + { + "epoch": 3.39, + "learning_rate": 8.378990463278038e-06, + "loss": 1.0469, + "step": 45899 + }, + { + "epoch": 3.39, + "learning_rate": 8.37859686835218e-06, + "loss": 1.0546, + "step": 45900 + }, + { + "epoch": 3.39, + "learning_rate": 8.378203276005945e-06, + "loss": 1.0004, + "step": 45901 + }, + { + "epoch": 3.39, + "learning_rate": 8.377809686239972e-06, + "loss": 0.9046, + "step": 45902 + }, + { + "epoch": 3.39, + "learning_rate": 8.377416099054878e-06, + "loss": 1.0012, + "step": 45903 + }, + { + "epoch": 3.39, + "learning_rate": 8.37702251445129e-06, + "loss": 1.0113, + "step": 45904 + }, + { + "epoch": 3.39, + "learning_rate": 8.376628932429837e-06, + "loss": 0.9682, + "step": 45905 + }, + { + "epoch": 3.39, + "learning_rate": 8.376235352991144e-06, + "loss": 0.9406, + "step": 45906 + }, + { + "epoch": 3.39, + "learning_rate": 8.375841776135838e-06, + "loss": 0.9981, + "step": 45907 + }, + { + "epoch": 3.39, + "learning_rate": 8.375448201864542e-06, + "loss": 1.0547, + "step": 45908 + }, + { + "epoch": 3.39, + "learning_rate": 8.375054630177888e-06, + "loss": 0.9998, + "step": 45909 + }, + { + "epoch": 3.39, + "learning_rate": 8.374661061076496e-06, + "loss": 1.0466, + "step": 45910 + }, + { + "epoch": 3.39, + "learning_rate": 8.374267494560998e-06, + "loss": 0.955, + "step": 45911 + }, + { + "epoch": 3.39, + "learning_rate": 8.373873930632017e-06, + "loss": 0.9555, + "step": 45912 + }, + { + "epoch": 3.39, + "learning_rate": 8.373480369290177e-06, + "loss": 0.9032, + "step": 45913 + }, + { + "epoch": 3.39, + "learning_rate": 8.373086810536108e-06, + "loss": 1.0059, + "step": 45914 + }, + { + "epoch": 3.39, + "learning_rate": 8.372693254370433e-06, + "loss": 1.078, + "step": 45915 + }, + { + "epoch": 3.39, + "learning_rate": 8.372299700793781e-06, + "loss": 1.0495, + "step": 45916 + }, + { + "epoch": 3.39, + "learning_rate": 8.371906149806776e-06, + "loss": 0.9766, + "step": 45917 + }, + { + "epoch": 3.39, + "learning_rate": 8.371512601410046e-06, + "loss": 1.1205, + "step": 45918 + }, + { + "epoch": 3.39, + "learning_rate": 8.371119055604214e-06, + "loss": 1.0315, + "step": 45919 + }, + { + "epoch": 3.39, + "learning_rate": 8.370725512389913e-06, + "loss": 0.9861, + "step": 45920 + }, + { + "epoch": 3.39, + "learning_rate": 8.370331971767763e-06, + "loss": 1.1098, + "step": 45921 + }, + { + "epoch": 3.39, + "learning_rate": 8.369938433738388e-06, + "loss": 1.0327, + "step": 45922 + }, + { + "epoch": 3.39, + "learning_rate": 8.369544898302422e-06, + "loss": 0.8407, + "step": 45923 + }, + { + "epoch": 3.39, + "learning_rate": 8.36915136546048e-06, + "loss": 0.9709, + "step": 45924 + }, + { + "epoch": 3.39, + "learning_rate": 8.368757835213203e-06, + "loss": 1.0302, + "step": 45925 + }, + { + "epoch": 3.39, + "learning_rate": 8.368364307561201e-06, + "loss": 0.9685, + "step": 45926 + }, + { + "epoch": 3.39, + "learning_rate": 8.367970782505117e-06, + "loss": 0.9222, + "step": 45927 + }, + { + "epoch": 3.39, + "learning_rate": 8.367577260045559e-06, + "loss": 1.0286, + "step": 45928 + }, + { + "epoch": 3.39, + "learning_rate": 8.367183740183168e-06, + "loss": 1.0753, + "step": 45929 + }, + { + "epoch": 3.39, + "learning_rate": 8.366790222918565e-06, + "loss": 0.9441, + "step": 45930 + }, + { + "epoch": 3.39, + "learning_rate": 8.366396708252373e-06, + "loss": 0.9576, + "step": 45931 + }, + { + "epoch": 3.39, + "learning_rate": 8.366003196185222e-06, + "loss": 0.9438, + "step": 45932 + }, + { + "epoch": 3.39, + "learning_rate": 8.365609686717733e-06, + "loss": 0.9366, + "step": 45933 + }, + { + "epoch": 3.39, + "learning_rate": 8.365216179850539e-06, + "loss": 0.8847, + "step": 45934 + }, + { + "epoch": 3.39, + "learning_rate": 8.36482267558426e-06, + "loss": 1.0096, + "step": 45935 + }, + { + "epoch": 3.39, + "learning_rate": 8.364429173919527e-06, + "loss": 0.9024, + "step": 45936 + }, + { + "epoch": 3.39, + "learning_rate": 8.364035674856962e-06, + "loss": 1.0831, + "step": 45937 + }, + { + "epoch": 3.39, + "learning_rate": 8.363642178397196e-06, + "loss": 0.9697, + "step": 45938 + }, + { + "epoch": 3.39, + "learning_rate": 8.363248684540848e-06, + "loss": 1.0197, + "step": 45939 + }, + { + "epoch": 3.39, + "learning_rate": 8.36285519328855e-06, + "loss": 1.0136, + "step": 45940 + }, + { + "epoch": 3.39, + "learning_rate": 8.36246170464093e-06, + "loss": 0.9745, + "step": 45941 + }, + { + "epoch": 3.39, + "learning_rate": 8.362068218598602e-06, + "loss": 1.049, + "step": 45942 + }, + { + "epoch": 3.39, + "learning_rate": 8.361674735162206e-06, + "loss": 1.1049, + "step": 45943 + }, + { + "epoch": 3.39, + "learning_rate": 8.361281254332356e-06, + "loss": 1.0218, + "step": 45944 + }, + { + "epoch": 3.39, + "learning_rate": 8.360887776109692e-06, + "loss": 1.0012, + "step": 45945 + }, + { + "epoch": 3.39, + "learning_rate": 8.360494300494824e-06, + "loss": 1.0299, + "step": 45946 + }, + { + "epoch": 3.39, + "learning_rate": 8.360100827488394e-06, + "loss": 0.9936, + "step": 45947 + }, + { + "epoch": 3.4, + "learning_rate": 8.359707357091011e-06, + "loss": 0.9984, + "step": 45948 + }, + { + "epoch": 3.4, + "learning_rate": 8.359313889303319e-06, + "loss": 1.0233, + "step": 45949 + }, + { + "epoch": 3.4, + "learning_rate": 8.358920424125933e-06, + "loss": 1.0149, + "step": 45950 + }, + { + "epoch": 3.4, + "learning_rate": 8.358526961559478e-06, + "loss": 0.9665, + "step": 45951 + }, + { + "epoch": 3.4, + "learning_rate": 8.358133501604586e-06, + "loss": 0.8981, + "step": 45952 + }, + { + "epoch": 3.4, + "learning_rate": 8.357740044261877e-06, + "loss": 1.0835, + "step": 45953 + }, + { + "epoch": 3.4, + "learning_rate": 8.357346589531983e-06, + "loss": 1.1014, + "step": 45954 + }, + { + "epoch": 3.4, + "learning_rate": 8.356953137415525e-06, + "loss": 1.0865, + "step": 45955 + }, + { + "epoch": 3.4, + "learning_rate": 8.356559687913131e-06, + "loss": 1.0421, + "step": 45956 + }, + { + "epoch": 3.4, + "learning_rate": 8.356166241025427e-06, + "loss": 1.0933, + "step": 45957 + }, + { + "epoch": 3.4, + "learning_rate": 8.355772796753042e-06, + "loss": 0.9949, + "step": 45958 + }, + { + "epoch": 3.4, + "learning_rate": 8.355379355096598e-06, + "loss": 0.965, + "step": 45959 + }, + { + "epoch": 3.4, + "learning_rate": 8.35498591605672e-06, + "loss": 1.0678, + "step": 45960 + }, + { + "epoch": 3.4, + "learning_rate": 8.354592479634035e-06, + "loss": 0.9435, + "step": 45961 + }, + { + "epoch": 3.4, + "learning_rate": 8.354199045829173e-06, + "loss": 0.9952, + "step": 45962 + }, + { + "epoch": 3.4, + "learning_rate": 8.353805614642755e-06, + "loss": 0.9336, + "step": 45963 + }, + { + "epoch": 3.4, + "learning_rate": 8.353412186075408e-06, + "loss": 0.9764, + "step": 45964 + }, + { + "epoch": 3.4, + "learning_rate": 8.35301876012776e-06, + "loss": 0.8591, + "step": 45965 + }, + { + "epoch": 3.4, + "learning_rate": 8.352625336800436e-06, + "loss": 1.0423, + "step": 45966 + }, + { + "epoch": 3.4, + "learning_rate": 8.35223191609406e-06, + "loss": 0.9505, + "step": 45967 + }, + { + "epoch": 3.4, + "learning_rate": 8.351838498009263e-06, + "loss": 0.9021, + "step": 45968 + }, + { + "epoch": 3.4, + "learning_rate": 8.351445082546662e-06, + "loss": 0.9778, + "step": 45969 + }, + { + "epoch": 3.4, + "learning_rate": 8.351051669706892e-06, + "loss": 1.0295, + "step": 45970 + }, + { + "epoch": 3.4, + "learning_rate": 8.350658259490573e-06, + "loss": 1.0147, + "step": 45971 + }, + { + "epoch": 3.4, + "learning_rate": 8.350264851898337e-06, + "loss": 1.0067, + "step": 45972 + }, + { + "epoch": 3.4, + "learning_rate": 8.349871446930799e-06, + "loss": 1.0974, + "step": 45973 + }, + { + "epoch": 3.4, + "learning_rate": 8.349478044588599e-06, + "loss": 0.978, + "step": 45974 + }, + { + "epoch": 3.4, + "learning_rate": 8.349084644872353e-06, + "loss": 0.9714, + "step": 45975 + }, + { + "epoch": 3.4, + "learning_rate": 8.34869124778269e-06, + "loss": 0.947, + "step": 45976 + }, + { + "epoch": 3.4, + "learning_rate": 8.348297853320237e-06, + "loss": 1.0298, + "step": 45977 + }, + { + "epoch": 3.4, + "learning_rate": 8.347904461485614e-06, + "loss": 1.0154, + "step": 45978 + }, + { + "epoch": 3.4, + "learning_rate": 8.347511072279455e-06, + "loss": 1.0385, + "step": 45979 + }, + { + "epoch": 3.4, + "learning_rate": 8.34711768570238e-06, + "loss": 0.9781, + "step": 45980 + }, + { + "epoch": 3.4, + "learning_rate": 8.34672430175502e-06, + "loss": 0.9827, + "step": 45981 + }, + { + "epoch": 3.4, + "learning_rate": 8.346330920437995e-06, + "loss": 1.023, + "step": 45982 + }, + { + "epoch": 3.4, + "learning_rate": 8.345937541751936e-06, + "loss": 1.1141, + "step": 45983 + }, + { + "epoch": 3.4, + "learning_rate": 8.345544165697466e-06, + "loss": 1.0212, + "step": 45984 + }, + { + "epoch": 3.4, + "learning_rate": 8.345150792275212e-06, + "loss": 1.015, + "step": 45985 + }, + { + "epoch": 3.4, + "learning_rate": 8.344757421485801e-06, + "loss": 0.927, + "step": 45986 + }, + { + "epoch": 3.4, + "learning_rate": 8.344364053329852e-06, + "loss": 0.9815, + "step": 45987 + }, + { + "epoch": 3.4, + "learning_rate": 8.343970687808004e-06, + "loss": 0.8986, + "step": 45988 + }, + { + "epoch": 3.4, + "learning_rate": 8.343577324920866e-06, + "loss": 0.9529, + "step": 45989 + }, + { + "epoch": 3.4, + "learning_rate": 8.34318396466908e-06, + "loss": 1.0081, + "step": 45990 + }, + { + "epoch": 3.4, + "learning_rate": 8.342790607053258e-06, + "loss": 0.9622, + "step": 45991 + }, + { + "epoch": 3.4, + "learning_rate": 8.342397252074039e-06, + "loss": 0.987, + "step": 45992 + }, + { + "epoch": 3.4, + "learning_rate": 8.342003899732036e-06, + "loss": 1.0088, + "step": 45993 + }, + { + "epoch": 3.4, + "learning_rate": 8.341610550027887e-06, + "loss": 1.1556, + "step": 45994 + }, + { + "epoch": 3.4, + "learning_rate": 8.341217202962207e-06, + "loss": 1.0274, + "step": 45995 + }, + { + "epoch": 3.4, + "learning_rate": 8.340823858535632e-06, + "loss": 1.0297, + "step": 45996 + }, + { + "epoch": 3.4, + "learning_rate": 8.340430516748781e-06, + "loss": 1.0026, + "step": 45997 + }, + { + "epoch": 3.4, + "learning_rate": 8.340037177602278e-06, + "loss": 0.9846, + "step": 45998 + }, + { + "epoch": 3.4, + "learning_rate": 8.339643841096756e-06, + "loss": 1.0268, + "step": 45999 + }, + { + "epoch": 3.4, + "learning_rate": 8.339250507232832e-06, + "loss": 0.9139, + "step": 46000 + }, + { + "epoch": 3.4, + "learning_rate": 8.33885717601114e-06, + "loss": 0.9517, + "step": 46001 + }, + { + "epoch": 3.4, + "learning_rate": 8.338463847432302e-06, + "loss": 1.095, + "step": 46002 + }, + { + "epoch": 3.4, + "learning_rate": 8.338070521496944e-06, + "loss": 0.9532, + "step": 46003 + }, + { + "epoch": 3.4, + "learning_rate": 8.337677198205691e-06, + "loss": 0.9967, + "step": 46004 + }, + { + "epoch": 3.4, + "learning_rate": 8.337283877559172e-06, + "loss": 1.1105, + "step": 46005 + }, + { + "epoch": 3.4, + "learning_rate": 8.33689055955801e-06, + "loss": 1.0229, + "step": 46006 + }, + { + "epoch": 3.4, + "learning_rate": 8.33649724420283e-06, + "loss": 0.9542, + "step": 46007 + }, + { + "epoch": 3.4, + "learning_rate": 8.336103931494261e-06, + "loss": 0.9929, + "step": 46008 + }, + { + "epoch": 3.4, + "learning_rate": 8.335710621432926e-06, + "loss": 1.0644, + "step": 46009 + }, + { + "epoch": 3.4, + "learning_rate": 8.335317314019452e-06, + "loss": 0.9605, + "step": 46010 + }, + { + "epoch": 3.4, + "learning_rate": 8.334924009254463e-06, + "loss": 0.9819, + "step": 46011 + }, + { + "epoch": 3.4, + "learning_rate": 8.334530707138587e-06, + "loss": 0.9861, + "step": 46012 + }, + { + "epoch": 3.4, + "learning_rate": 8.334137407672446e-06, + "loss": 0.936, + "step": 46013 + }, + { + "epoch": 3.4, + "learning_rate": 8.333744110856674e-06, + "loss": 1.0388, + "step": 46014 + }, + { + "epoch": 3.4, + "learning_rate": 8.33335081669189e-06, + "loss": 0.9413, + "step": 46015 + }, + { + "epoch": 3.4, + "learning_rate": 8.332957525178717e-06, + "loss": 1.0291, + "step": 46016 + }, + { + "epoch": 3.4, + "learning_rate": 8.33256423631779e-06, + "loss": 0.9841, + "step": 46017 + }, + { + "epoch": 3.4, + "learning_rate": 8.332170950109726e-06, + "loss": 1.1205, + "step": 46018 + }, + { + "epoch": 3.4, + "learning_rate": 8.331777666555155e-06, + "loss": 1.0718, + "step": 46019 + }, + { + "epoch": 3.4, + "learning_rate": 8.3313843856547e-06, + "loss": 1.0426, + "step": 46020 + }, + { + "epoch": 3.4, + "learning_rate": 8.330991107408993e-06, + "loss": 0.9635, + "step": 46021 + }, + { + "epoch": 3.4, + "learning_rate": 8.33059783181865e-06, + "loss": 1.0051, + "step": 46022 + }, + { + "epoch": 3.4, + "learning_rate": 8.330204558884307e-06, + "loss": 0.9285, + "step": 46023 + }, + { + "epoch": 3.4, + "learning_rate": 8.329811288606581e-06, + "loss": 0.9613, + "step": 46024 + }, + { + "epoch": 3.4, + "learning_rate": 8.329418020986101e-06, + "loss": 1.0358, + "step": 46025 + }, + { + "epoch": 3.4, + "learning_rate": 8.329024756023496e-06, + "loss": 0.9477, + "step": 46026 + }, + { + "epoch": 3.4, + "learning_rate": 8.328631493719386e-06, + "loss": 0.8886, + "step": 46027 + }, + { + "epoch": 3.4, + "learning_rate": 8.328238234074402e-06, + "loss": 0.984, + "step": 46028 + }, + { + "epoch": 3.4, + "learning_rate": 8.327844977089165e-06, + "loss": 1.046, + "step": 46029 + }, + { + "epoch": 3.4, + "learning_rate": 8.327451722764304e-06, + "loss": 0.9819, + "step": 46030 + }, + { + "epoch": 3.4, + "learning_rate": 8.32705847110044e-06, + "loss": 1.044, + "step": 46031 + }, + { + "epoch": 3.4, + "learning_rate": 8.326665222098207e-06, + "loss": 0.8386, + "step": 46032 + }, + { + "epoch": 3.4, + "learning_rate": 8.326271975758227e-06, + "loss": 1.1156, + "step": 46033 + }, + { + "epoch": 3.4, + "learning_rate": 8.325878732081116e-06, + "loss": 1.0092, + "step": 46034 + }, + { + "epoch": 3.4, + "learning_rate": 8.325485491067516e-06, + "loss": 0.9374, + "step": 46035 + }, + { + "epoch": 3.4, + "learning_rate": 8.325092252718039e-06, + "loss": 1.1067, + "step": 46036 + }, + { + "epoch": 3.4, + "learning_rate": 8.324699017033322e-06, + "loss": 1.0868, + "step": 46037 + }, + { + "epoch": 3.4, + "learning_rate": 8.324305784013977e-06, + "loss": 1.1323, + "step": 46038 + }, + { + "epoch": 3.4, + "learning_rate": 8.323912553660646e-06, + "loss": 0.9885, + "step": 46039 + }, + { + "epoch": 3.4, + "learning_rate": 8.32351932597394e-06, + "loss": 0.9409, + "step": 46040 + }, + { + "epoch": 3.4, + "learning_rate": 8.323126100954495e-06, + "loss": 1.1106, + "step": 46041 + }, + { + "epoch": 3.4, + "learning_rate": 8.322732878602926e-06, + "loss": 0.9564, + "step": 46042 + }, + { + "epoch": 3.4, + "learning_rate": 8.322339658919874e-06, + "loss": 0.9657, + "step": 46043 + }, + { + "epoch": 3.4, + "learning_rate": 8.32194644190595e-06, + "loss": 0.9919, + "step": 46044 + }, + { + "epoch": 3.4, + "learning_rate": 8.321553227561784e-06, + "loss": 1.0585, + "step": 46045 + }, + { + "epoch": 3.4, + "learning_rate": 8.321160015888009e-06, + "loss": 1.0166, + "step": 46046 + }, + { + "epoch": 3.4, + "learning_rate": 8.320766806885238e-06, + "loss": 0.9885, + "step": 46047 + }, + { + "epoch": 3.4, + "learning_rate": 8.320373600554105e-06, + "loss": 1.0166, + "step": 46048 + }, + { + "epoch": 3.4, + "learning_rate": 8.319980396895233e-06, + "loss": 1.0702, + "step": 46049 + }, + { + "epoch": 3.4, + "learning_rate": 8.319587195909251e-06, + "loss": 0.906, + "step": 46050 + }, + { + "epoch": 3.4, + "learning_rate": 8.319193997596778e-06, + "loss": 1.0444, + "step": 46051 + }, + { + "epoch": 3.4, + "learning_rate": 8.318800801958447e-06, + "loss": 1.0155, + "step": 46052 + }, + { + "epoch": 3.4, + "learning_rate": 8.318407608994879e-06, + "loss": 0.9749, + "step": 46053 + }, + { + "epoch": 3.4, + "learning_rate": 8.318014418706698e-06, + "loss": 1.0449, + "step": 46054 + }, + { + "epoch": 3.4, + "learning_rate": 8.317621231094534e-06, + "loss": 0.9585, + "step": 46055 + }, + { + "epoch": 3.4, + "learning_rate": 8.317228046159008e-06, + "loss": 0.928, + "step": 46056 + }, + { + "epoch": 3.4, + "learning_rate": 8.31683486390075e-06, + "loss": 0.9702, + "step": 46057 + }, + { + "epoch": 3.4, + "learning_rate": 8.316441684320384e-06, + "loss": 0.9203, + "step": 46058 + }, + { + "epoch": 3.4, + "learning_rate": 8.316048507418535e-06, + "loss": 0.8919, + "step": 46059 + }, + { + "epoch": 3.4, + "learning_rate": 8.315655333195828e-06, + "loss": 0.976, + "step": 46060 + }, + { + "epoch": 3.4, + "learning_rate": 8.315262161652891e-06, + "loss": 1.0658, + "step": 46061 + }, + { + "epoch": 3.4, + "learning_rate": 8.31486899279035e-06, + "loss": 0.9374, + "step": 46062 + }, + { + "epoch": 3.4, + "learning_rate": 8.314475826608823e-06, + "loss": 0.998, + "step": 46063 + }, + { + "epoch": 3.4, + "learning_rate": 8.314082663108945e-06, + "loss": 1.0132, + "step": 46064 + }, + { + "epoch": 3.4, + "learning_rate": 8.313689502291334e-06, + "loss": 1.0046, + "step": 46065 + }, + { + "epoch": 3.4, + "learning_rate": 8.31329634415662e-06, + "loss": 0.8838, + "step": 46066 + }, + { + "epoch": 3.4, + "learning_rate": 8.312903188705425e-06, + "loss": 1.0966, + "step": 46067 + }, + { + "epoch": 3.4, + "learning_rate": 8.312510035938381e-06, + "loss": 0.883, + "step": 46068 + }, + { + "epoch": 3.4, + "learning_rate": 8.312116885856108e-06, + "loss": 1.0061, + "step": 46069 + }, + { + "epoch": 3.4, + "learning_rate": 8.311723738459232e-06, + "loss": 0.9725, + "step": 46070 + }, + { + "epoch": 3.4, + "learning_rate": 8.311330593748381e-06, + "loss": 1.0778, + "step": 46071 + }, + { + "epoch": 3.4, + "learning_rate": 8.310937451724177e-06, + "loss": 0.9853, + "step": 46072 + }, + { + "epoch": 3.4, + "learning_rate": 8.31054431238725e-06, + "loss": 1.0044, + "step": 46073 + }, + { + "epoch": 3.4, + "learning_rate": 8.31015117573822e-06, + "loss": 1.0393, + "step": 46074 + }, + { + "epoch": 3.4, + "learning_rate": 8.309758041777716e-06, + "loss": 0.9931, + "step": 46075 + }, + { + "epoch": 3.4, + "learning_rate": 8.309364910506363e-06, + "loss": 0.9716, + "step": 46076 + }, + { + "epoch": 3.4, + "learning_rate": 8.308971781924786e-06, + "loss": 1.0663, + "step": 46077 + }, + { + "epoch": 3.4, + "learning_rate": 8.30857865603361e-06, + "loss": 1.0316, + "step": 46078 + }, + { + "epoch": 3.4, + "learning_rate": 8.308185532833463e-06, + "loss": 0.9331, + "step": 46079 + }, + { + "epoch": 3.4, + "learning_rate": 8.307792412324971e-06, + "loss": 0.9827, + "step": 46080 + }, + { + "epoch": 3.4, + "learning_rate": 8.307399294508751e-06, + "loss": 0.9727, + "step": 46081 + }, + { + "epoch": 3.4, + "learning_rate": 8.30700617938544e-06, + "loss": 1.0928, + "step": 46082 + }, + { + "epoch": 3.41, + "learning_rate": 8.306613066955652e-06, + "loss": 0.9166, + "step": 46083 + }, + { + "epoch": 3.41, + "learning_rate": 8.306219957220025e-06, + "loss": 0.9673, + "step": 46084 + }, + { + "epoch": 3.41, + "learning_rate": 8.30582685017917e-06, + "loss": 0.9561, + "step": 46085 + }, + { + "epoch": 3.41, + "learning_rate": 8.305433745833729e-06, + "loss": 1.0484, + "step": 46086 + }, + { + "epoch": 3.41, + "learning_rate": 8.305040644184312e-06, + "loss": 0.8696, + "step": 46087 + }, + { + "epoch": 3.41, + "learning_rate": 8.304647545231556e-06, + "loss": 0.9636, + "step": 46088 + }, + { + "epoch": 3.41, + "learning_rate": 8.304254448976079e-06, + "loss": 1.0075, + "step": 46089 + }, + { + "epoch": 3.41, + "learning_rate": 8.303861355418508e-06, + "loss": 0.9367, + "step": 46090 + }, + { + "epoch": 3.41, + "learning_rate": 8.30346826455947e-06, + "loss": 0.994, + "step": 46091 + }, + { + "epoch": 3.41, + "learning_rate": 8.30307517639959e-06, + "loss": 0.966, + "step": 46092 + }, + { + "epoch": 3.41, + "learning_rate": 8.302682090939495e-06, + "loss": 0.9524, + "step": 46093 + }, + { + "epoch": 3.41, + "learning_rate": 8.302289008179804e-06, + "loss": 1.0145, + "step": 46094 + }, + { + "epoch": 3.41, + "learning_rate": 8.301895928121152e-06, + "loss": 1.055, + "step": 46095 + }, + { + "epoch": 3.41, + "learning_rate": 8.301502850764153e-06, + "loss": 0.9605, + "step": 46096 + }, + { + "epoch": 3.41, + "learning_rate": 8.301109776109444e-06, + "loss": 0.9678, + "step": 46097 + }, + { + "epoch": 3.41, + "learning_rate": 8.300716704157643e-06, + "loss": 0.9436, + "step": 46098 + }, + { + "epoch": 3.41, + "learning_rate": 8.300323634909379e-06, + "loss": 0.942, + "step": 46099 + }, + { + "epoch": 3.41, + "learning_rate": 8.299930568365276e-06, + "loss": 1.0298, + "step": 46100 + }, + { + "epoch": 3.41, + "learning_rate": 8.299537504525956e-06, + "loss": 0.9241, + "step": 46101 + }, + { + "epoch": 3.41, + "learning_rate": 8.29914444339205e-06, + "loss": 0.8791, + "step": 46102 + }, + { + "epoch": 3.41, + "learning_rate": 8.29875138496418e-06, + "loss": 0.883, + "step": 46103 + }, + { + "epoch": 3.41, + "learning_rate": 8.298358329242972e-06, + "loss": 1.0398, + "step": 46104 + }, + { + "epoch": 3.41, + "learning_rate": 8.29796527622905e-06, + "loss": 0.9209, + "step": 46105 + }, + { + "epoch": 3.41, + "learning_rate": 8.297572225923047e-06, + "loss": 0.984, + "step": 46106 + }, + { + "epoch": 3.41, + "learning_rate": 8.297179178325574e-06, + "loss": 1.0556, + "step": 46107 + }, + { + "epoch": 3.41, + "learning_rate": 8.296786133437273e-06, + "loss": 1.0373, + "step": 46108 + }, + { + "epoch": 3.41, + "learning_rate": 8.296393091258757e-06, + "loss": 0.9448, + "step": 46109 + }, + { + "epoch": 3.41, + "learning_rate": 8.296000051790654e-06, + "loss": 0.9371, + "step": 46110 + }, + { + "epoch": 3.41, + "learning_rate": 8.295607015033591e-06, + "loss": 0.9722, + "step": 46111 + }, + { + "epoch": 3.41, + "learning_rate": 8.295213980988193e-06, + "loss": 0.8682, + "step": 46112 + }, + { + "epoch": 3.41, + "learning_rate": 8.294820949655085e-06, + "loss": 1.0349, + "step": 46113 + }, + { + "epoch": 3.41, + "learning_rate": 8.294427921034892e-06, + "loss": 1.0767, + "step": 46114 + }, + { + "epoch": 3.41, + "learning_rate": 8.29403489512824e-06, + "loss": 0.9731, + "step": 46115 + }, + { + "epoch": 3.41, + "learning_rate": 8.293641871935755e-06, + "loss": 0.983, + "step": 46116 + }, + { + "epoch": 3.41, + "learning_rate": 8.293248851458063e-06, + "loss": 1.0322, + "step": 46117 + }, + { + "epoch": 3.41, + "learning_rate": 8.292855833695785e-06, + "loss": 0.9609, + "step": 46118 + }, + { + "epoch": 3.41, + "learning_rate": 8.292462818649547e-06, + "loss": 1.0286, + "step": 46119 + }, + { + "epoch": 3.41, + "learning_rate": 8.292069806319982e-06, + "loss": 1.0669, + "step": 46120 + }, + { + "epoch": 3.41, + "learning_rate": 8.291676796707704e-06, + "loss": 1.0002, + "step": 46121 + }, + { + "epoch": 3.41, + "learning_rate": 8.291283789813347e-06, + "loss": 0.9838, + "step": 46122 + }, + { + "epoch": 3.41, + "learning_rate": 8.29089078563753e-06, + "loss": 0.9976, + "step": 46123 + }, + { + "epoch": 3.41, + "learning_rate": 8.290497784180885e-06, + "loss": 0.9825, + "step": 46124 + }, + { + "epoch": 3.41, + "learning_rate": 8.290104785444031e-06, + "loss": 0.9896, + "step": 46125 + }, + { + "epoch": 3.41, + "learning_rate": 8.289711789427598e-06, + "loss": 0.9704, + "step": 46126 + }, + { + "epoch": 3.41, + "learning_rate": 8.289318796132211e-06, + "loss": 0.9764, + "step": 46127 + }, + { + "epoch": 3.41, + "learning_rate": 8.288925805558487e-06, + "loss": 0.8747, + "step": 46128 + }, + { + "epoch": 3.41, + "learning_rate": 8.288532817707064e-06, + "loss": 0.9969, + "step": 46129 + }, + { + "epoch": 3.41, + "learning_rate": 8.288139832578553e-06, + "loss": 1.0528, + "step": 46130 + }, + { + "epoch": 3.41, + "learning_rate": 8.287746850173594e-06, + "loss": 1.0755, + "step": 46131 + }, + { + "epoch": 3.41, + "learning_rate": 8.2873538704928e-06, + "loss": 1.0176, + "step": 46132 + }, + { + "epoch": 3.41, + "learning_rate": 8.286960893536809e-06, + "loss": 0.9466, + "step": 46133 + }, + { + "epoch": 3.41, + "learning_rate": 8.28656791930623e-06, + "loss": 0.9866, + "step": 46134 + }, + { + "epoch": 3.41, + "learning_rate": 8.286174947801703e-06, + "loss": 0.9457, + "step": 46135 + }, + { + "epoch": 3.41, + "learning_rate": 8.285781979023848e-06, + "loss": 0.8586, + "step": 46136 + }, + { + "epoch": 3.41, + "learning_rate": 8.285389012973284e-06, + "loss": 1.1017, + "step": 46137 + }, + { + "epoch": 3.41, + "learning_rate": 8.284996049650646e-06, + "loss": 1.0451, + "step": 46138 + }, + { + "epoch": 3.41, + "learning_rate": 8.284603089056551e-06, + "loss": 0.9473, + "step": 46139 + }, + { + "epoch": 3.41, + "learning_rate": 8.284210131191631e-06, + "loss": 0.9706, + "step": 46140 + }, + { + "epoch": 3.41, + "learning_rate": 8.283817176056505e-06, + "loss": 0.9795, + "step": 46141 + }, + { + "epoch": 3.41, + "learning_rate": 8.283424223651804e-06, + "loss": 1.0154, + "step": 46142 + }, + { + "epoch": 3.41, + "learning_rate": 8.283031273978148e-06, + "loss": 1.0637, + "step": 46143 + }, + { + "epoch": 3.41, + "learning_rate": 8.282638327036168e-06, + "loss": 0.8253, + "step": 46144 + }, + { + "epoch": 3.41, + "learning_rate": 8.282245382826485e-06, + "loss": 0.9267, + "step": 46145 + }, + { + "epoch": 3.41, + "learning_rate": 8.281852441349722e-06, + "loss": 0.9757, + "step": 46146 + }, + { + "epoch": 3.41, + "learning_rate": 8.28145950260651e-06, + "loss": 1.01, + "step": 46147 + }, + { + "epoch": 3.41, + "learning_rate": 8.28106656659747e-06, + "loss": 1.0187, + "step": 46148 + }, + { + "epoch": 3.41, + "learning_rate": 8.280673633323231e-06, + "loss": 1.1183, + "step": 46149 + }, + { + "epoch": 3.41, + "learning_rate": 8.280280702784412e-06, + "loss": 1.0462, + "step": 46150 + }, + { + "epoch": 3.41, + "learning_rate": 8.279887774981646e-06, + "loss": 1.0638, + "step": 46151 + }, + { + "epoch": 3.41, + "learning_rate": 8.279494849915549e-06, + "loss": 1.005, + "step": 46152 + }, + { + "epoch": 3.41, + "learning_rate": 8.279101927586757e-06, + "loss": 0.984, + "step": 46153 + }, + { + "epoch": 3.41, + "learning_rate": 8.278709007995881e-06, + "loss": 1.0376, + "step": 46154 + }, + { + "epoch": 3.41, + "learning_rate": 8.278316091143563e-06, + "loss": 0.8857, + "step": 46155 + }, + { + "epoch": 3.41, + "learning_rate": 8.277923177030417e-06, + "loss": 1.0075, + "step": 46156 + }, + { + "epoch": 3.41, + "learning_rate": 8.277530265657068e-06, + "loss": 0.9554, + "step": 46157 + }, + { + "epoch": 3.41, + "learning_rate": 8.277137357024147e-06, + "loss": 1.0496, + "step": 46158 + }, + { + "epoch": 3.41, + "learning_rate": 8.276744451132272e-06, + "loss": 0.944, + "step": 46159 + }, + { + "epoch": 3.41, + "learning_rate": 8.276351547982075e-06, + "loss": 1.1593, + "step": 46160 + }, + { + "epoch": 3.41, + "learning_rate": 8.275958647574174e-06, + "loss": 1.0364, + "step": 46161 + }, + { + "epoch": 3.41, + "learning_rate": 8.275565749909202e-06, + "loss": 1.0826, + "step": 46162 + }, + { + "epoch": 3.41, + "learning_rate": 8.27517285498778e-06, + "loss": 1.0631, + "step": 46163 + }, + { + "epoch": 3.41, + "learning_rate": 8.274779962810532e-06, + "loss": 0.9888, + "step": 46164 + }, + { + "epoch": 3.41, + "learning_rate": 8.274387073378086e-06, + "loss": 1.0464, + "step": 46165 + }, + { + "epoch": 3.41, + "learning_rate": 8.273994186691062e-06, + "loss": 0.956, + "step": 46166 + }, + { + "epoch": 3.41, + "learning_rate": 8.273601302750091e-06, + "loss": 0.9542, + "step": 46167 + }, + { + "epoch": 3.41, + "learning_rate": 8.273208421555797e-06, + "loss": 1.0782, + "step": 46168 + }, + { + "epoch": 3.41, + "learning_rate": 8.272815543108802e-06, + "loss": 1.01, + "step": 46169 + }, + { + "epoch": 3.41, + "learning_rate": 8.272422667409732e-06, + "loss": 0.8432, + "step": 46170 + }, + { + "epoch": 3.41, + "learning_rate": 8.272029794459214e-06, + "loss": 0.9377, + "step": 46171 + }, + { + "epoch": 3.41, + "learning_rate": 8.27163692425787e-06, + "loss": 1.0196, + "step": 46172 + }, + { + "epoch": 3.41, + "learning_rate": 8.271244056806329e-06, + "loss": 0.9932, + "step": 46173 + }, + { + "epoch": 3.41, + "learning_rate": 8.270851192105217e-06, + "loss": 1.0365, + "step": 46174 + }, + { + "epoch": 3.41, + "learning_rate": 8.270458330155149e-06, + "loss": 0.9738, + "step": 46175 + }, + { + "epoch": 3.41, + "learning_rate": 8.270065470956765e-06, + "loss": 1.0057, + "step": 46176 + }, + { + "epoch": 3.41, + "learning_rate": 8.269672614510673e-06, + "loss": 1.0367, + "step": 46177 + }, + { + "epoch": 3.41, + "learning_rate": 8.269279760817514e-06, + "loss": 1.0655, + "step": 46178 + }, + { + "epoch": 3.41, + "learning_rate": 8.268886909877901e-06, + "loss": 1.0519, + "step": 46179 + }, + { + "epoch": 3.41, + "learning_rate": 8.268494061692472e-06, + "loss": 1.0058, + "step": 46180 + }, + { + "epoch": 3.41, + "learning_rate": 8.268101216261834e-06, + "loss": 1.0029, + "step": 46181 + }, + { + "epoch": 3.41, + "learning_rate": 8.267708373586632e-06, + "loss": 0.9027, + "step": 46182 + }, + { + "epoch": 3.41, + "learning_rate": 8.267315533667476e-06, + "loss": 1.0739, + "step": 46183 + }, + { + "epoch": 3.41, + "learning_rate": 8.266922696504997e-06, + "loss": 0.9975, + "step": 46184 + }, + { + "epoch": 3.41, + "learning_rate": 8.266529862099819e-06, + "loss": 0.9531, + "step": 46185 + }, + { + "epoch": 3.41, + "learning_rate": 8.266137030452566e-06, + "loss": 0.9637, + "step": 46186 + }, + { + "epoch": 3.41, + "learning_rate": 8.265744201563866e-06, + "loss": 0.9518, + "step": 46187 + }, + { + "epoch": 3.41, + "learning_rate": 8.26535137543434e-06, + "loss": 0.9886, + "step": 46188 + }, + { + "epoch": 3.41, + "learning_rate": 8.264958552064618e-06, + "loss": 0.9472, + "step": 46189 + }, + { + "epoch": 3.41, + "learning_rate": 8.26456573145532e-06, + "loss": 0.9725, + "step": 46190 + }, + { + "epoch": 3.41, + "learning_rate": 8.264172913607075e-06, + "loss": 1.0988, + "step": 46191 + }, + { + "epoch": 3.41, + "learning_rate": 8.263780098520505e-06, + "loss": 0.9883, + "step": 46192 + }, + { + "epoch": 3.41, + "learning_rate": 8.263387286196235e-06, + "loss": 0.986, + "step": 46193 + }, + { + "epoch": 3.41, + "learning_rate": 8.262994476634893e-06, + "loss": 1.017, + "step": 46194 + }, + { + "epoch": 3.41, + "learning_rate": 8.2626016698371e-06, + "loss": 0.9877, + "step": 46195 + }, + { + "epoch": 3.41, + "learning_rate": 8.262208865803488e-06, + "loss": 1.0043, + "step": 46196 + }, + { + "epoch": 3.41, + "learning_rate": 8.261816064534668e-06, + "loss": 1.0304, + "step": 46197 + }, + { + "epoch": 3.41, + "learning_rate": 8.261423266031281e-06, + "loss": 1.0454, + "step": 46198 + }, + { + "epoch": 3.41, + "learning_rate": 8.261030470293937e-06, + "loss": 1.0957, + "step": 46199 + }, + { + "epoch": 3.41, + "learning_rate": 8.260637677323279e-06, + "loss": 1.0211, + "step": 46200 + }, + { + "epoch": 3.41, + "learning_rate": 8.260244887119918e-06, + "loss": 1.0439, + "step": 46201 + }, + { + "epoch": 3.41, + "learning_rate": 8.259852099684477e-06, + "loss": 0.9263, + "step": 46202 + }, + { + "epoch": 3.41, + "learning_rate": 8.259459315017593e-06, + "loss": 1.0913, + "step": 46203 + }, + { + "epoch": 3.41, + "learning_rate": 8.25906653311988e-06, + "loss": 1.0416, + "step": 46204 + }, + { + "epoch": 3.41, + "learning_rate": 8.258673753991969e-06, + "loss": 0.9649, + "step": 46205 + }, + { + "epoch": 3.41, + "learning_rate": 8.258280977634481e-06, + "loss": 0.9502, + "step": 46206 + }, + { + "epoch": 3.41, + "learning_rate": 8.257888204048046e-06, + "loss": 0.9199, + "step": 46207 + }, + { + "epoch": 3.41, + "learning_rate": 8.257495433233285e-06, + "loss": 1.0635, + "step": 46208 + }, + { + "epoch": 3.41, + "learning_rate": 8.257102665190823e-06, + "loss": 1.0254, + "step": 46209 + }, + { + "epoch": 3.41, + "learning_rate": 8.256709899921286e-06, + "loss": 1.0186, + "step": 46210 + }, + { + "epoch": 3.41, + "learning_rate": 8.256317137425301e-06, + "loss": 1.0477, + "step": 46211 + }, + { + "epoch": 3.41, + "learning_rate": 8.255924377703488e-06, + "loss": 0.9917, + "step": 46212 + }, + { + "epoch": 3.41, + "learning_rate": 8.255531620756473e-06, + "loss": 0.9241, + "step": 46213 + }, + { + "epoch": 3.41, + "learning_rate": 8.255138866584886e-06, + "loss": 0.9989, + "step": 46214 + }, + { + "epoch": 3.41, + "learning_rate": 8.254746115189345e-06, + "loss": 1.0501, + "step": 46215 + }, + { + "epoch": 3.41, + "learning_rate": 8.25435336657048e-06, + "loss": 0.979, + "step": 46216 + }, + { + "epoch": 3.41, + "learning_rate": 8.253960620728912e-06, + "loss": 0.9776, + "step": 46217 + }, + { + "epoch": 3.42, + "learning_rate": 8.25356787766527e-06, + "loss": 1.0002, + "step": 46218 + }, + { + "epoch": 3.42, + "learning_rate": 8.253175137380175e-06, + "loss": 1.0298, + "step": 46219 + }, + { + "epoch": 3.42, + "learning_rate": 8.252782399874253e-06, + "loss": 1.0834, + "step": 46220 + }, + { + "epoch": 3.42, + "learning_rate": 8.252389665148133e-06, + "loss": 1.1475, + "step": 46221 + }, + { + "epoch": 3.42, + "learning_rate": 8.25199693320243e-06, + "loss": 1.1266, + "step": 46222 + }, + { + "epoch": 3.42, + "learning_rate": 8.25160420403778e-06, + "loss": 0.9423, + "step": 46223 + }, + { + "epoch": 3.42, + "learning_rate": 8.251211477654796e-06, + "loss": 0.973, + "step": 46224 + }, + { + "epoch": 3.42, + "learning_rate": 8.250818754054119e-06, + "loss": 0.9727, + "step": 46225 + }, + { + "epoch": 3.42, + "learning_rate": 8.250426033236356e-06, + "loss": 1.0295, + "step": 46226 + }, + { + "epoch": 3.42, + "learning_rate": 8.250033315202147e-06, + "loss": 1.065, + "step": 46227 + }, + { + "epoch": 3.42, + "learning_rate": 8.249640599952104e-06, + "loss": 1.0162, + "step": 46228 + }, + { + "epoch": 3.42, + "learning_rate": 8.24924788748686e-06, + "loss": 1.0397, + "step": 46229 + }, + { + "epoch": 3.42, + "learning_rate": 8.248855177807039e-06, + "loss": 1.1457, + "step": 46230 + }, + { + "epoch": 3.42, + "learning_rate": 8.248462470913263e-06, + "loss": 0.8489, + "step": 46231 + }, + { + "epoch": 3.42, + "learning_rate": 8.248069766806159e-06, + "loss": 1.0058, + "step": 46232 + }, + { + "epoch": 3.42, + "learning_rate": 8.247677065486348e-06, + "loss": 1.0165, + "step": 46233 + }, + { + "epoch": 3.42, + "learning_rate": 8.24728436695446e-06, + "loss": 0.9873, + "step": 46234 + }, + { + "epoch": 3.42, + "learning_rate": 8.246891671211118e-06, + "loss": 0.9573, + "step": 46235 + }, + { + "epoch": 3.42, + "learning_rate": 8.246498978256944e-06, + "loss": 1.0162, + "step": 46236 + }, + { + "epoch": 3.42, + "learning_rate": 8.246106288092565e-06, + "loss": 0.9425, + "step": 46237 + }, + { + "epoch": 3.42, + "learning_rate": 8.24571360071861e-06, + "loss": 0.9603, + "step": 46238 + }, + { + "epoch": 3.42, + "learning_rate": 8.2453209161357e-06, + "loss": 0.9716, + "step": 46239 + }, + { + "epoch": 3.42, + "learning_rate": 8.244928234344453e-06, + "loss": 1.0085, + "step": 46240 + }, + { + "epoch": 3.42, + "learning_rate": 8.244535555345506e-06, + "loss": 1.0897, + "step": 46241 + }, + { + "epoch": 3.42, + "learning_rate": 8.244142879139471e-06, + "loss": 0.857, + "step": 46242 + }, + { + "epoch": 3.42, + "learning_rate": 8.243750205726988e-06, + "loss": 1.0062, + "step": 46243 + }, + { + "epoch": 3.42, + "learning_rate": 8.243357535108665e-06, + "loss": 0.9995, + "step": 46244 + }, + { + "epoch": 3.42, + "learning_rate": 8.242964867285143e-06, + "loss": 0.9743, + "step": 46245 + }, + { + "epoch": 3.42, + "learning_rate": 8.24257220225703e-06, + "loss": 1.0519, + "step": 46246 + }, + { + "epoch": 3.42, + "learning_rate": 8.242179540024968e-06, + "loss": 0.9524, + "step": 46247 + }, + { + "epoch": 3.42, + "learning_rate": 8.24178688058957e-06, + "loss": 0.906, + "step": 46248 + }, + { + "epoch": 3.42, + "learning_rate": 8.241394223951461e-06, + "loss": 0.9967, + "step": 46249 + }, + { + "epoch": 3.42, + "learning_rate": 8.241001570111274e-06, + "loss": 0.8978, + "step": 46250 + }, + { + "epoch": 3.42, + "learning_rate": 8.240608919069624e-06, + "loss": 1.0096, + "step": 46251 + }, + { + "epoch": 3.42, + "learning_rate": 8.240216270827141e-06, + "loss": 1.0043, + "step": 46252 + }, + { + "epoch": 3.42, + "learning_rate": 8.239823625384449e-06, + "loss": 0.9949, + "step": 46253 + }, + { + "epoch": 3.42, + "learning_rate": 8.239430982742174e-06, + "loss": 0.9511, + "step": 46254 + }, + { + "epoch": 3.42, + "learning_rate": 8.239038342900935e-06, + "loss": 1.0847, + "step": 46255 + }, + { + "epoch": 3.42, + "learning_rate": 8.238645705861365e-06, + "loss": 0.8338, + "step": 46256 + }, + { + "epoch": 3.42, + "learning_rate": 8.238253071624083e-06, + "loss": 1.0214, + "step": 46257 + }, + { + "epoch": 3.42, + "learning_rate": 8.237860440189716e-06, + "loss": 1.022, + "step": 46258 + }, + { + "epoch": 3.42, + "learning_rate": 8.237467811558887e-06, + "loss": 1.106, + "step": 46259 + }, + { + "epoch": 3.42, + "learning_rate": 8.237075185732219e-06, + "loss": 0.9615, + "step": 46260 + }, + { + "epoch": 3.42, + "learning_rate": 8.236682562710342e-06, + "loss": 0.9849, + "step": 46261 + }, + { + "epoch": 3.42, + "learning_rate": 8.236289942493878e-06, + "loss": 0.8972, + "step": 46262 + }, + { + "epoch": 3.42, + "learning_rate": 8.235897325083451e-06, + "loss": 0.9949, + "step": 46263 + }, + { + "epoch": 3.42, + "learning_rate": 8.235504710479685e-06, + "loss": 1.057, + "step": 46264 + }, + { + "epoch": 3.42, + "learning_rate": 8.235112098683207e-06, + "loss": 1.0179, + "step": 46265 + }, + { + "epoch": 3.42, + "learning_rate": 8.23471948969464e-06, + "loss": 1.1445, + "step": 46266 + }, + { + "epoch": 3.42, + "learning_rate": 8.23432688351461e-06, + "loss": 0.9568, + "step": 46267 + }, + { + "epoch": 3.42, + "learning_rate": 8.233934280143742e-06, + "loss": 1.0389, + "step": 46268 + }, + { + "epoch": 3.42, + "learning_rate": 8.233541679582655e-06, + "loss": 0.9558, + "step": 46269 + }, + { + "epoch": 3.42, + "learning_rate": 8.233149081831983e-06, + "loss": 0.8771, + "step": 46270 + }, + { + "epoch": 3.42, + "learning_rate": 8.232756486892339e-06, + "loss": 1.0066, + "step": 46271 + }, + { + "epoch": 3.42, + "learning_rate": 8.232363894764362e-06, + "loss": 1.0129, + "step": 46272 + }, + { + "epoch": 3.42, + "learning_rate": 8.231971305448662e-06, + "loss": 1.0222, + "step": 46273 + }, + { + "epoch": 3.42, + "learning_rate": 8.231578718945874e-06, + "loss": 0.9668, + "step": 46274 + }, + { + "epoch": 3.42, + "learning_rate": 8.231186135256617e-06, + "loss": 1.0598, + "step": 46275 + }, + { + "epoch": 3.42, + "learning_rate": 8.230793554381519e-06, + "loss": 1.1348, + "step": 46276 + }, + { + "epoch": 3.42, + "learning_rate": 8.230400976321204e-06, + "loss": 0.9298, + "step": 46277 + }, + { + "epoch": 3.42, + "learning_rate": 8.230008401076293e-06, + "loss": 0.9575, + "step": 46278 + }, + { + "epoch": 3.42, + "learning_rate": 8.229615828647415e-06, + "loss": 0.9474, + "step": 46279 + }, + { + "epoch": 3.42, + "learning_rate": 8.229223259035191e-06, + "loss": 0.9925, + "step": 46280 + }, + { + "epoch": 3.42, + "learning_rate": 8.22883069224025e-06, + "loss": 0.8296, + "step": 46281 + }, + { + "epoch": 3.42, + "learning_rate": 8.228438128263212e-06, + "loss": 1.06, + "step": 46282 + }, + { + "epoch": 3.42, + "learning_rate": 8.228045567104704e-06, + "loss": 1.034, + "step": 46283 + }, + { + "epoch": 3.42, + "learning_rate": 8.22765300876535e-06, + "loss": 1.0224, + "step": 46284 + }, + { + "epoch": 3.42, + "learning_rate": 8.227260453245777e-06, + "loss": 0.9394, + "step": 46285 + }, + { + "epoch": 3.42, + "learning_rate": 8.226867900546607e-06, + "loss": 0.9808, + "step": 46286 + }, + { + "epoch": 3.42, + "learning_rate": 8.226475350668462e-06, + "loss": 0.9875, + "step": 46287 + }, + { + "epoch": 3.42, + "learning_rate": 8.226082803611973e-06, + "loss": 1.0273, + "step": 46288 + }, + { + "epoch": 3.42, + "learning_rate": 8.225690259377756e-06, + "loss": 1.0088, + "step": 46289 + }, + { + "epoch": 3.42, + "learning_rate": 8.225297717966445e-06, + "loss": 0.8718, + "step": 46290 + }, + { + "epoch": 3.42, + "learning_rate": 8.224905179378654e-06, + "loss": 0.999, + "step": 46291 + }, + { + "epoch": 3.42, + "learning_rate": 8.224512643615022e-06, + "loss": 1.0074, + "step": 46292 + }, + { + "epoch": 3.42, + "learning_rate": 8.224120110676157e-06, + "loss": 0.8776, + "step": 46293 + }, + { + "epoch": 3.42, + "learning_rate": 8.223727580562697e-06, + "loss": 0.9133, + "step": 46294 + }, + { + "epoch": 3.42, + "learning_rate": 8.22333505327526e-06, + "loss": 0.9419, + "step": 46295 + }, + { + "epoch": 3.42, + "learning_rate": 8.222942528814469e-06, + "loss": 0.9578, + "step": 46296 + }, + { + "epoch": 3.42, + "learning_rate": 8.222550007180954e-06, + "loss": 0.9441, + "step": 46297 + }, + { + "epoch": 3.42, + "learning_rate": 8.222157488375333e-06, + "loss": 1.1521, + "step": 46298 + }, + { + "epoch": 3.42, + "learning_rate": 8.221764972398238e-06, + "loss": 0.9972, + "step": 46299 + }, + { + "epoch": 3.42, + "learning_rate": 8.221372459250285e-06, + "loss": 1.0596, + "step": 46300 + }, + { + "epoch": 3.42, + "learning_rate": 8.220979948932107e-06, + "loss": 1.0535, + "step": 46301 + }, + { + "epoch": 3.42, + "learning_rate": 8.22058744144432e-06, + "loss": 1.0122, + "step": 46302 + }, + { + "epoch": 3.42, + "learning_rate": 8.220194936787558e-06, + "loss": 0.999, + "step": 46303 + }, + { + "epoch": 3.42, + "learning_rate": 8.21980243496244e-06, + "loss": 0.9748, + "step": 46304 + }, + { + "epoch": 3.42, + "learning_rate": 8.219409935969587e-06, + "loss": 1.0186, + "step": 46305 + }, + { + "epoch": 3.42, + "learning_rate": 8.21901743980963e-06, + "loss": 0.9531, + "step": 46306 + }, + { + "epoch": 3.42, + "learning_rate": 8.218624946483189e-06, + "loss": 1.0079, + "step": 46307 + }, + { + "epoch": 3.42, + "learning_rate": 8.218232455990891e-06, + "loss": 0.9299, + "step": 46308 + }, + { + "epoch": 3.42, + "learning_rate": 8.21783996833336e-06, + "loss": 0.9617, + "step": 46309 + }, + { + "epoch": 3.42, + "learning_rate": 8.217447483511222e-06, + "loss": 0.9798, + "step": 46310 + }, + { + "epoch": 3.42, + "learning_rate": 8.217055001525096e-06, + "loss": 1.0359, + "step": 46311 + }, + { + "epoch": 3.42, + "learning_rate": 8.216662522375613e-06, + "loss": 0.9352, + "step": 46312 + }, + { + "epoch": 3.42, + "learning_rate": 8.216270046063392e-06, + "loss": 1.0828, + "step": 46313 + }, + { + "epoch": 3.42, + "learning_rate": 8.215877572589062e-06, + "loss": 0.9512, + "step": 46314 + }, + { + "epoch": 3.42, + "learning_rate": 8.215485101953246e-06, + "loss": 1.0519, + "step": 46315 + }, + { + "epoch": 3.42, + "learning_rate": 8.215092634156563e-06, + "loss": 1.0418, + "step": 46316 + }, + { + "epoch": 3.42, + "learning_rate": 8.214700169199647e-06, + "loss": 1.0014, + "step": 46317 + }, + { + "epoch": 3.42, + "learning_rate": 8.214307707083114e-06, + "loss": 0.949, + "step": 46318 + }, + { + "epoch": 3.42, + "learning_rate": 8.213915247807595e-06, + "loss": 0.9391, + "step": 46319 + }, + { + "epoch": 3.42, + "learning_rate": 8.213522791373708e-06, + "loss": 1.0751, + "step": 46320 + }, + { + "epoch": 3.42, + "learning_rate": 8.213130337782084e-06, + "loss": 0.9706, + "step": 46321 + }, + { + "epoch": 3.42, + "learning_rate": 8.212737887033341e-06, + "loss": 0.9697, + "step": 46322 + }, + { + "epoch": 3.42, + "learning_rate": 8.212345439128108e-06, + "loss": 0.9154, + "step": 46323 + }, + { + "epoch": 3.42, + "learning_rate": 8.211952994067009e-06, + "loss": 0.9328, + "step": 46324 + }, + { + "epoch": 3.42, + "learning_rate": 8.211560551850664e-06, + "loss": 0.9637, + "step": 46325 + }, + { + "epoch": 3.42, + "learning_rate": 8.211168112479703e-06, + "loss": 1.0049, + "step": 46326 + }, + { + "epoch": 3.42, + "learning_rate": 8.210775675954746e-06, + "loss": 1.0574, + "step": 46327 + }, + { + "epoch": 3.42, + "learning_rate": 8.210383242276423e-06, + "loss": 0.87, + "step": 46328 + }, + { + "epoch": 3.42, + "learning_rate": 8.20999081144535e-06, + "loss": 1.0211, + "step": 46329 + }, + { + "epoch": 3.42, + "learning_rate": 8.209598383462159e-06, + "loss": 0.892, + "step": 46330 + }, + { + "epoch": 3.42, + "learning_rate": 8.20920595832747e-06, + "loss": 1.0528, + "step": 46331 + }, + { + "epoch": 3.42, + "learning_rate": 8.20881353604191e-06, + "loss": 0.9194, + "step": 46332 + }, + { + "epoch": 3.42, + "learning_rate": 8.208421116606105e-06, + "loss": 0.8844, + "step": 46333 + }, + { + "epoch": 3.42, + "learning_rate": 8.208028700020669e-06, + "loss": 1.0777, + "step": 46334 + }, + { + "epoch": 3.42, + "learning_rate": 8.207636286286241e-06, + "loss": 0.9453, + "step": 46335 + }, + { + "epoch": 3.42, + "learning_rate": 8.20724387540343e-06, + "loss": 1.0411, + "step": 46336 + }, + { + "epoch": 3.42, + "learning_rate": 8.206851467372875e-06, + "loss": 1.1217, + "step": 46337 + }, + { + "epoch": 3.42, + "learning_rate": 8.20645906219519e-06, + "loss": 1.028, + "step": 46338 + }, + { + "epoch": 3.42, + "learning_rate": 8.206066659871007e-06, + "loss": 0.9063, + "step": 46339 + }, + { + "epoch": 3.42, + "learning_rate": 8.205674260400939e-06, + "loss": 1.0648, + "step": 46340 + }, + { + "epoch": 3.42, + "learning_rate": 8.205281863785626e-06, + "loss": 1.0122, + "step": 46341 + }, + { + "epoch": 3.42, + "learning_rate": 8.20488947002568e-06, + "loss": 1.0167, + "step": 46342 + }, + { + "epoch": 3.42, + "learning_rate": 8.20449707912173e-06, + "loss": 1.0054, + "step": 46343 + }, + { + "epoch": 3.42, + "learning_rate": 8.2041046910744e-06, + "loss": 1.0767, + "step": 46344 + }, + { + "epoch": 3.42, + "learning_rate": 8.20371230588431e-06, + "loss": 0.9133, + "step": 46345 + }, + { + "epoch": 3.42, + "learning_rate": 8.203319923552093e-06, + "loss": 0.9635, + "step": 46346 + }, + { + "epoch": 3.42, + "learning_rate": 8.202927544078364e-06, + "loss": 1.0195, + "step": 46347 + }, + { + "epoch": 3.42, + "learning_rate": 8.202535167463754e-06, + "loss": 1.0237, + "step": 46348 + }, + { + "epoch": 3.42, + "learning_rate": 8.202142793708883e-06, + "loss": 1.1224, + "step": 46349 + }, + { + "epoch": 3.42, + "learning_rate": 8.20175042281438e-06, + "loss": 0.9088, + "step": 46350 + }, + { + "epoch": 3.42, + "learning_rate": 8.201358054780864e-06, + "loss": 0.9396, + "step": 46351 + }, + { + "epoch": 3.42, + "learning_rate": 8.200965689608961e-06, + "loss": 1.0623, + "step": 46352 + }, + { + "epoch": 3.42, + "learning_rate": 8.2005733272993e-06, + "loss": 0.937, + "step": 46353 + }, + { + "epoch": 3.43, + "learning_rate": 8.200180967852495e-06, + "loss": 1.0022, + "step": 46354 + }, + { + "epoch": 3.43, + "learning_rate": 8.199788611269181e-06, + "loss": 1.0207, + "step": 46355 + }, + { + "epoch": 3.43, + "learning_rate": 8.199396257549975e-06, + "loss": 1.0625, + "step": 46356 + }, + { + "epoch": 3.43, + "learning_rate": 8.199003906695506e-06, + "loss": 1.0171, + "step": 46357 + }, + { + "epoch": 3.43, + "learning_rate": 8.198611558706395e-06, + "loss": 0.9625, + "step": 46358 + }, + { + "epoch": 3.43, + "learning_rate": 8.198219213583267e-06, + "loss": 0.9929, + "step": 46359 + }, + { + "epoch": 3.43, + "learning_rate": 8.19782687132675e-06, + "loss": 1.0727, + "step": 46360 + }, + { + "epoch": 3.43, + "learning_rate": 8.19743453193746e-06, + "loss": 1.0532, + "step": 46361 + }, + { + "epoch": 3.43, + "learning_rate": 8.197042195416026e-06, + "loss": 1.0819, + "step": 46362 + }, + { + "epoch": 3.43, + "learning_rate": 8.196649861763073e-06, + "loss": 0.9965, + "step": 46363 + }, + { + "epoch": 3.43, + "learning_rate": 8.196257530979226e-06, + "loss": 1.0265, + "step": 46364 + }, + { + "epoch": 3.43, + "learning_rate": 8.195865203065105e-06, + "loss": 1.1088, + "step": 46365 + }, + { + "epoch": 3.43, + "learning_rate": 8.195472878021337e-06, + "loss": 1.0586, + "step": 46366 + }, + { + "epoch": 3.43, + "learning_rate": 8.195080555848547e-06, + "loss": 0.9569, + "step": 46367 + }, + { + "epoch": 3.43, + "learning_rate": 8.194688236547358e-06, + "loss": 0.9769, + "step": 46368 + }, + { + "epoch": 3.43, + "learning_rate": 8.194295920118392e-06, + "loss": 0.9749, + "step": 46369 + }, + { + "epoch": 3.43, + "learning_rate": 8.193903606562278e-06, + "loss": 0.9476, + "step": 46370 + }, + { + "epoch": 3.43, + "learning_rate": 8.193511295879637e-06, + "loss": 0.9963, + "step": 46371 + }, + { + "epoch": 3.43, + "learning_rate": 8.193118988071093e-06, + "loss": 0.9749, + "step": 46372 + }, + { + "epoch": 3.43, + "learning_rate": 8.19272668313727e-06, + "loss": 0.9495, + "step": 46373 + }, + { + "epoch": 3.43, + "learning_rate": 8.192334381078794e-06, + "loss": 1.0812, + "step": 46374 + }, + { + "epoch": 3.43, + "learning_rate": 8.191942081896289e-06, + "loss": 0.9414, + "step": 46375 + }, + { + "epoch": 3.43, + "learning_rate": 8.191549785590375e-06, + "loss": 1.0711, + "step": 46376 + }, + { + "epoch": 3.43, + "learning_rate": 8.191157492161685e-06, + "loss": 0.9517, + "step": 46377 + }, + { + "epoch": 3.43, + "learning_rate": 8.190765201610832e-06, + "loss": 1.0245, + "step": 46378 + }, + { + "epoch": 3.43, + "learning_rate": 8.190372913938449e-06, + "loss": 0.9698, + "step": 46379 + }, + { + "epoch": 3.43, + "learning_rate": 8.18998062914516e-06, + "loss": 1.0501, + "step": 46380 + }, + { + "epoch": 3.43, + "learning_rate": 8.189588347231578e-06, + "loss": 0.9086, + "step": 46381 + }, + { + "epoch": 3.43, + "learning_rate": 8.189196068198341e-06, + "loss": 1.0311, + "step": 46382 + }, + { + "epoch": 3.43, + "learning_rate": 8.188803792046062e-06, + "loss": 1.0354, + "step": 46383 + }, + { + "epoch": 3.43, + "learning_rate": 8.188411518775377e-06, + "loss": 0.9907, + "step": 46384 + }, + { + "epoch": 3.43, + "learning_rate": 8.188019248386896e-06, + "loss": 0.8833, + "step": 46385 + }, + { + "epoch": 3.43, + "learning_rate": 8.187626980881257e-06, + "loss": 0.9356, + "step": 46386 + }, + { + "epoch": 3.43, + "learning_rate": 8.187234716259071e-06, + "loss": 0.9129, + "step": 46387 + }, + { + "epoch": 3.43, + "learning_rate": 8.186842454520975e-06, + "loss": 0.9503, + "step": 46388 + }, + { + "epoch": 3.43, + "learning_rate": 8.186450195667584e-06, + "loss": 1.0735, + "step": 46389 + }, + { + "epoch": 3.43, + "learning_rate": 8.186057939699524e-06, + "loss": 1.1322, + "step": 46390 + }, + { + "epoch": 3.43, + "learning_rate": 8.185665686617422e-06, + "loss": 0.9558, + "step": 46391 + }, + { + "epoch": 3.43, + "learning_rate": 8.185273436421897e-06, + "loss": 0.9724, + "step": 46392 + }, + { + "epoch": 3.43, + "learning_rate": 8.184881189113578e-06, + "loss": 1.0112, + "step": 46393 + }, + { + "epoch": 3.43, + "learning_rate": 8.184488944693085e-06, + "loss": 1.05, + "step": 46394 + }, + { + "epoch": 3.43, + "learning_rate": 8.184096703161048e-06, + "loss": 1.0542, + "step": 46395 + }, + { + "epoch": 3.43, + "learning_rate": 8.183704464518083e-06, + "loss": 1.0957, + "step": 46396 + }, + { + "epoch": 3.43, + "learning_rate": 8.183312228764822e-06, + "loss": 1.0608, + "step": 46397 + }, + { + "epoch": 3.43, + "learning_rate": 8.182919995901884e-06, + "loss": 0.9605, + "step": 46398 + }, + { + "epoch": 3.43, + "learning_rate": 8.182527765929892e-06, + "loss": 1.0096, + "step": 46399 + }, + { + "epoch": 3.43, + "learning_rate": 8.182135538849474e-06, + "loss": 1.1568, + "step": 46400 + }, + { + "epoch": 3.43, + "learning_rate": 8.181743314661252e-06, + "loss": 1.0393, + "step": 46401 + }, + { + "epoch": 3.43, + "learning_rate": 8.181351093365852e-06, + "loss": 1.0106, + "step": 46402 + }, + { + "epoch": 3.43, + "learning_rate": 8.180958874963893e-06, + "loss": 0.9465, + "step": 46403 + }, + { + "epoch": 3.43, + "learning_rate": 8.180566659456005e-06, + "loss": 1.0969, + "step": 46404 + }, + { + "epoch": 3.43, + "learning_rate": 8.180174446842808e-06, + "loss": 1.0195, + "step": 46405 + }, + { + "epoch": 3.43, + "learning_rate": 8.179782237124932e-06, + "loss": 0.9358, + "step": 46406 + }, + { + "epoch": 3.43, + "learning_rate": 8.179390030302993e-06, + "loss": 0.9168, + "step": 46407 + }, + { + "epoch": 3.43, + "learning_rate": 8.178997826377618e-06, + "loss": 1.0114, + "step": 46408 + }, + { + "epoch": 3.43, + "learning_rate": 8.178605625349431e-06, + "loss": 1.0325, + "step": 46409 + }, + { + "epoch": 3.43, + "learning_rate": 8.178213427219054e-06, + "loss": 0.9484, + "step": 46410 + }, + { + "epoch": 3.43, + "learning_rate": 8.177821231987119e-06, + "loss": 0.981, + "step": 46411 + }, + { + "epoch": 3.43, + "learning_rate": 8.17742903965424e-06, + "loss": 1.0068, + "step": 46412 + }, + { + "epoch": 3.43, + "learning_rate": 8.177036850221049e-06, + "loss": 1.1161, + "step": 46413 + }, + { + "epoch": 3.43, + "learning_rate": 8.176644663688162e-06, + "loss": 1.0459, + "step": 46414 + }, + { + "epoch": 3.43, + "learning_rate": 8.176252480056211e-06, + "loss": 0.9692, + "step": 46415 + }, + { + "epoch": 3.43, + "learning_rate": 8.175860299325815e-06, + "loss": 1.0086, + "step": 46416 + }, + { + "epoch": 3.43, + "learning_rate": 8.175468121497597e-06, + "loss": 1.014, + "step": 46417 + }, + { + "epoch": 3.43, + "learning_rate": 8.175075946572186e-06, + "loss": 1.0054, + "step": 46418 + }, + { + "epoch": 3.43, + "learning_rate": 8.1746837745502e-06, + "loss": 0.8951, + "step": 46419 + }, + { + "epoch": 3.43, + "learning_rate": 8.17429160543227e-06, + "loss": 0.9521, + "step": 46420 + }, + { + "epoch": 3.43, + "learning_rate": 8.173899439219013e-06, + "loss": 1.0565, + "step": 46421 + }, + { + "epoch": 3.43, + "learning_rate": 8.173507275911058e-06, + "loss": 1.0905, + "step": 46422 + }, + { + "epoch": 3.43, + "learning_rate": 8.173115115509023e-06, + "loss": 0.9232, + "step": 46423 + }, + { + "epoch": 3.43, + "learning_rate": 8.17272295801354e-06, + "loss": 1.0046, + "step": 46424 + }, + { + "epoch": 3.43, + "learning_rate": 8.172330803425226e-06, + "loss": 1.0381, + "step": 46425 + }, + { + "epoch": 3.43, + "learning_rate": 8.171938651744708e-06, + "loss": 1.0278, + "step": 46426 + }, + { + "epoch": 3.43, + "learning_rate": 8.171546502972613e-06, + "loss": 1.0035, + "step": 46427 + }, + { + "epoch": 3.43, + "learning_rate": 8.171154357109556e-06, + "loss": 1.0136, + "step": 46428 + }, + { + "epoch": 3.43, + "learning_rate": 8.170762214156171e-06, + "loss": 0.9334, + "step": 46429 + }, + { + "epoch": 3.43, + "learning_rate": 8.170370074113072e-06, + "loss": 0.9916, + "step": 46430 + }, + { + "epoch": 3.43, + "learning_rate": 8.169977936980894e-06, + "loss": 0.9456, + "step": 46431 + }, + { + "epoch": 3.43, + "learning_rate": 8.169585802760249e-06, + "loss": 0.937, + "step": 46432 + }, + { + "epoch": 3.43, + "learning_rate": 8.169193671451771e-06, + "loss": 1.054, + "step": 46433 + }, + { + "epoch": 3.43, + "learning_rate": 8.168801543056075e-06, + "loss": 0.9108, + "step": 46434 + }, + { + "epoch": 3.43, + "learning_rate": 8.168409417573796e-06, + "loss": 0.9171, + "step": 46435 + }, + { + "epoch": 3.43, + "learning_rate": 8.16801729500555e-06, + "loss": 1.0182, + "step": 46436 + }, + { + "epoch": 3.43, + "learning_rate": 8.167625175351959e-06, + "loss": 1.0501, + "step": 46437 + }, + { + "epoch": 3.43, + "learning_rate": 8.167233058613652e-06, + "loss": 0.9831, + "step": 46438 + }, + { + "epoch": 3.43, + "learning_rate": 8.16684094479125e-06, + "loss": 1.1017, + "step": 46439 + }, + { + "epoch": 3.43, + "learning_rate": 8.166448833885378e-06, + "loss": 1.0388, + "step": 46440 + }, + { + "epoch": 3.43, + "learning_rate": 8.166056725896659e-06, + "loss": 1.0216, + "step": 46441 + }, + { + "epoch": 3.43, + "learning_rate": 8.16566462082572e-06, + "loss": 1.084, + "step": 46442 + }, + { + "epoch": 3.43, + "learning_rate": 8.165272518673178e-06, + "loss": 0.9592, + "step": 46443 + }, + { + "epoch": 3.43, + "learning_rate": 8.164880419439666e-06, + "loss": 0.9975, + "step": 46444 + }, + { + "epoch": 3.43, + "learning_rate": 8.164488323125802e-06, + "loss": 0.9208, + "step": 46445 + }, + { + "epoch": 3.43, + "learning_rate": 8.164096229732208e-06, + "loss": 0.9545, + "step": 46446 + }, + { + "epoch": 3.43, + "learning_rate": 8.163704139259513e-06, + "loss": 0.9125, + "step": 46447 + }, + { + "epoch": 3.43, + "learning_rate": 8.163312051708336e-06, + "loss": 0.9343, + "step": 46448 + }, + { + "epoch": 3.43, + "learning_rate": 8.162919967079306e-06, + "loss": 0.9979, + "step": 46449 + }, + { + "epoch": 3.43, + "learning_rate": 8.162527885373042e-06, + "loss": 1.031, + "step": 46450 + }, + { + "epoch": 3.43, + "learning_rate": 8.162135806590175e-06, + "loss": 0.9549, + "step": 46451 + }, + { + "epoch": 3.43, + "learning_rate": 8.161743730731317e-06, + "loss": 1.0225, + "step": 46452 + }, + { + "epoch": 3.43, + "learning_rate": 8.161351657797104e-06, + "loss": 0.9838, + "step": 46453 + }, + { + "epoch": 3.43, + "learning_rate": 8.160959587788151e-06, + "loss": 1.0284, + "step": 46454 + }, + { + "epoch": 3.43, + "learning_rate": 8.160567520705086e-06, + "loss": 0.9388, + "step": 46455 + }, + { + "epoch": 3.43, + "learning_rate": 8.160175456548531e-06, + "loss": 0.9692, + "step": 46456 + }, + { + "epoch": 3.43, + "learning_rate": 8.159783395319109e-06, + "loss": 0.9027, + "step": 46457 + }, + { + "epoch": 3.43, + "learning_rate": 8.159391337017448e-06, + "loss": 0.939, + "step": 46458 + }, + { + "epoch": 3.43, + "learning_rate": 8.158999281644167e-06, + "loss": 1.098, + "step": 46459 + }, + { + "epoch": 3.43, + "learning_rate": 8.158607229199893e-06, + "loss": 1.0021, + "step": 46460 + }, + { + "epoch": 3.43, + "learning_rate": 8.158215179685246e-06, + "loss": 0.9499, + "step": 46461 + }, + { + "epoch": 3.43, + "learning_rate": 8.157823133100856e-06, + "loss": 0.8552, + "step": 46462 + }, + { + "epoch": 3.43, + "learning_rate": 8.157431089447342e-06, + "loss": 1.0446, + "step": 46463 + }, + { + "epoch": 3.43, + "learning_rate": 8.157039048725328e-06, + "loss": 1.0725, + "step": 46464 + }, + { + "epoch": 3.43, + "learning_rate": 8.156647010935439e-06, + "loss": 0.9211, + "step": 46465 + }, + { + "epoch": 3.43, + "learning_rate": 8.156254976078297e-06, + "loss": 1.0447, + "step": 46466 + }, + { + "epoch": 3.43, + "learning_rate": 8.15586294415453e-06, + "loss": 0.9703, + "step": 46467 + }, + { + "epoch": 3.43, + "learning_rate": 8.155470915164755e-06, + "loss": 0.9673, + "step": 46468 + }, + { + "epoch": 3.43, + "learning_rate": 8.155078889109603e-06, + "loss": 0.9323, + "step": 46469 + }, + { + "epoch": 3.43, + "learning_rate": 8.154686865989689e-06, + "loss": 1.0449, + "step": 46470 + }, + { + "epoch": 3.43, + "learning_rate": 8.154294845805645e-06, + "loss": 1.0722, + "step": 46471 + }, + { + "epoch": 3.43, + "learning_rate": 8.153902828558092e-06, + "loss": 0.93, + "step": 46472 + }, + { + "epoch": 3.43, + "learning_rate": 8.153510814247654e-06, + "loss": 0.9907, + "step": 46473 + }, + { + "epoch": 3.43, + "learning_rate": 8.153118802874956e-06, + "loss": 1.0595, + "step": 46474 + }, + { + "epoch": 3.43, + "learning_rate": 8.15272679444061e-06, + "loss": 1.0104, + "step": 46475 + }, + { + "epoch": 3.43, + "learning_rate": 8.15233478894526e-06, + "loss": 0.9593, + "step": 46476 + }, + { + "epoch": 3.43, + "learning_rate": 8.15194278638951e-06, + "loss": 1.1192, + "step": 46477 + }, + { + "epoch": 3.43, + "learning_rate": 8.151550786774e-06, + "loss": 1.0336, + "step": 46478 + }, + { + "epoch": 3.43, + "learning_rate": 8.15115879009934e-06, + "loss": 1.1207, + "step": 46479 + }, + { + "epoch": 3.43, + "learning_rate": 8.150766796366166e-06, + "loss": 1.061, + "step": 46480 + }, + { + "epoch": 3.43, + "learning_rate": 8.15037480557509e-06, + "loss": 1.1586, + "step": 46481 + }, + { + "epoch": 3.43, + "learning_rate": 8.149982817726746e-06, + "loss": 1.0045, + "step": 46482 + }, + { + "epoch": 3.43, + "learning_rate": 8.149590832821752e-06, + "loss": 0.9971, + "step": 46483 + }, + { + "epoch": 3.43, + "learning_rate": 8.149198850860731e-06, + "loss": 1.013, + "step": 46484 + }, + { + "epoch": 3.43, + "learning_rate": 8.148806871844308e-06, + "loss": 0.994, + "step": 46485 + }, + { + "epoch": 3.43, + "learning_rate": 8.148414895773107e-06, + "loss": 0.9599, + "step": 46486 + }, + { + "epoch": 3.43, + "learning_rate": 8.148022922647752e-06, + "loss": 0.9624, + "step": 46487 + }, + { + "epoch": 3.43, + "learning_rate": 8.147630952468864e-06, + "loss": 0.9819, + "step": 46488 + }, + { + "epoch": 3.44, + "learning_rate": 8.14723898523707e-06, + "loss": 1.0363, + "step": 46489 + }, + { + "epoch": 3.44, + "learning_rate": 8.146847020952991e-06, + "loss": 0.9507, + "step": 46490 + }, + { + "epoch": 3.44, + "learning_rate": 8.146455059617256e-06, + "loss": 1.1666, + "step": 46491 + }, + { + "epoch": 3.44, + "learning_rate": 8.146063101230482e-06, + "loss": 0.9847, + "step": 46492 + }, + { + "epoch": 3.44, + "learning_rate": 8.145671145793294e-06, + "loss": 1.0258, + "step": 46493 + }, + { + "epoch": 3.44, + "learning_rate": 8.145279193306321e-06, + "loss": 1.0298, + "step": 46494 + }, + { + "epoch": 3.44, + "learning_rate": 8.144887243770175e-06, + "loss": 1.0676, + "step": 46495 + }, + { + "epoch": 3.44, + "learning_rate": 8.144495297185493e-06, + "loss": 1.0197, + "step": 46496 + }, + { + "epoch": 3.44, + "learning_rate": 8.144103353552887e-06, + "loss": 1.0135, + "step": 46497 + }, + { + "epoch": 3.44, + "learning_rate": 8.143711412872993e-06, + "loss": 0.9921, + "step": 46498 + }, + { + "epoch": 3.44, + "learning_rate": 8.143319475146419e-06, + "loss": 0.9766, + "step": 46499 + }, + { + "epoch": 3.44, + "learning_rate": 8.142927540373805e-06, + "loss": 1.0666, + "step": 46500 + }, + { + "epoch": 3.44, + "learning_rate": 8.142535608555765e-06, + "loss": 1.0401, + "step": 46501 + }, + { + "epoch": 3.44, + "learning_rate": 8.14214367969292e-06, + "loss": 1.0103, + "step": 46502 + }, + { + "epoch": 3.44, + "learning_rate": 8.141751753785902e-06, + "loss": 0.9205, + "step": 46503 + }, + { + "epoch": 3.44, + "learning_rate": 8.141359830835329e-06, + "loss": 1.0104, + "step": 46504 + }, + { + "epoch": 3.44, + "learning_rate": 8.140967910841827e-06, + "loss": 0.9263, + "step": 46505 + }, + { + "epoch": 3.44, + "learning_rate": 8.140575993806016e-06, + "loss": 1.0138, + "step": 46506 + }, + { + "epoch": 3.44, + "learning_rate": 8.140184079728523e-06, + "loss": 1.072, + "step": 46507 + }, + { + "epoch": 3.44, + "learning_rate": 8.13979216860997e-06, + "loss": 0.9987, + "step": 46508 + }, + { + "epoch": 3.44, + "learning_rate": 8.139400260450983e-06, + "loss": 0.9224, + "step": 46509 + }, + { + "epoch": 3.44, + "learning_rate": 8.139008355252184e-06, + "loss": 1.0374, + "step": 46510 + }, + { + "epoch": 3.44, + "learning_rate": 8.138616453014193e-06, + "loss": 1.0727, + "step": 46511 + }, + { + "epoch": 3.44, + "learning_rate": 8.138224553737639e-06, + "loss": 1.0416, + "step": 46512 + }, + { + "epoch": 3.44, + "learning_rate": 8.13783265742314e-06, + "loss": 0.95, + "step": 46513 + }, + { + "epoch": 3.44, + "learning_rate": 8.137440764071327e-06, + "loss": 1.0428, + "step": 46514 + }, + { + "epoch": 3.44, + "learning_rate": 8.137048873682816e-06, + "loss": 0.9801, + "step": 46515 + }, + { + "epoch": 3.44, + "learning_rate": 8.136656986258236e-06, + "loss": 0.9776, + "step": 46516 + }, + { + "epoch": 3.44, + "learning_rate": 8.136265101798205e-06, + "loss": 0.9831, + "step": 46517 + }, + { + "epoch": 3.44, + "learning_rate": 8.135873220303351e-06, + "loss": 1.0599, + "step": 46518 + }, + { + "epoch": 3.44, + "learning_rate": 8.135481341774301e-06, + "loss": 1.0901, + "step": 46519 + }, + { + "epoch": 3.44, + "learning_rate": 8.135089466211665e-06, + "loss": 1.023, + "step": 46520 + }, + { + "epoch": 3.44, + "learning_rate": 8.13469759361608e-06, + "loss": 1.0006, + "step": 46521 + }, + { + "epoch": 3.44, + "learning_rate": 8.13430572398816e-06, + "loss": 0.9039, + "step": 46522 + }, + { + "epoch": 3.44, + "learning_rate": 8.13391385732854e-06, + "loss": 1.1245, + "step": 46523 + }, + { + "epoch": 3.44, + "learning_rate": 8.13352199363783e-06, + "loss": 1.0544, + "step": 46524 + }, + { + "epoch": 3.44, + "learning_rate": 8.133130132916665e-06, + "loss": 0.8208, + "step": 46525 + }, + { + "epoch": 3.44, + "learning_rate": 8.132738275165658e-06, + "loss": 0.8784, + "step": 46526 + }, + { + "epoch": 3.44, + "learning_rate": 8.132346420385443e-06, + "loss": 0.9508, + "step": 46527 + }, + { + "epoch": 3.44, + "learning_rate": 8.131954568576633e-06, + "loss": 1.035, + "step": 46528 + }, + { + "epoch": 3.44, + "learning_rate": 8.13156271973986e-06, + "loss": 1.0428, + "step": 46529 + }, + { + "epoch": 3.44, + "learning_rate": 8.131170873875744e-06, + "loss": 0.9816, + "step": 46530 + }, + { + "epoch": 3.44, + "learning_rate": 8.130779030984905e-06, + "loss": 1.0541, + "step": 46531 + }, + { + "epoch": 3.44, + "learning_rate": 8.130387191067973e-06, + "loss": 0.9688, + "step": 46532 + }, + { + "epoch": 3.44, + "learning_rate": 8.129995354125566e-06, + "loss": 1.0266, + "step": 46533 + }, + { + "epoch": 3.44, + "learning_rate": 8.129603520158312e-06, + "loss": 0.964, + "step": 46534 + }, + { + "epoch": 3.44, + "learning_rate": 8.12921168916683e-06, + "loss": 1.1077, + "step": 46535 + }, + { + "epoch": 3.44, + "learning_rate": 8.128819861151746e-06, + "loss": 1.0436, + "step": 46536 + }, + { + "epoch": 3.44, + "learning_rate": 8.128428036113683e-06, + "loss": 1.0436, + "step": 46537 + }, + { + "epoch": 3.44, + "learning_rate": 8.128036214053265e-06, + "loss": 0.9549, + "step": 46538 + }, + { + "epoch": 3.44, + "learning_rate": 8.127644394971116e-06, + "loss": 1.0802, + "step": 46539 + }, + { + "epoch": 3.44, + "learning_rate": 8.127252578867854e-06, + "loss": 1.0256, + "step": 46540 + }, + { + "epoch": 3.44, + "learning_rate": 8.12686076574411e-06, + "loss": 1.0591, + "step": 46541 + }, + { + "epoch": 3.44, + "learning_rate": 8.126468955600499e-06, + "loss": 1.0733, + "step": 46542 + }, + { + "epoch": 3.44, + "learning_rate": 8.126077148437657e-06, + "loss": 0.9841, + "step": 46543 + }, + { + "epoch": 3.44, + "learning_rate": 8.12568534425619e-06, + "loss": 1.0301, + "step": 46544 + }, + { + "epoch": 3.44, + "learning_rate": 8.125293543056738e-06, + "loss": 1.0028, + "step": 46545 + }, + { + "epoch": 3.44, + "learning_rate": 8.124901744839913e-06, + "loss": 1.0583, + "step": 46546 + }, + { + "epoch": 3.44, + "learning_rate": 8.124509949606346e-06, + "loss": 0.916, + "step": 46547 + }, + { + "epoch": 3.44, + "learning_rate": 8.124118157356656e-06, + "loss": 0.9513, + "step": 46548 + }, + { + "epoch": 3.44, + "learning_rate": 8.123726368091465e-06, + "loss": 1.0362, + "step": 46549 + }, + { + "epoch": 3.44, + "learning_rate": 8.1233345818114e-06, + "loss": 0.9926, + "step": 46550 + }, + { + "epoch": 3.44, + "learning_rate": 8.122942798517082e-06, + "loss": 1.0146, + "step": 46551 + }, + { + "epoch": 3.44, + "learning_rate": 8.122551018209137e-06, + "loss": 1.0522, + "step": 46552 + }, + { + "epoch": 3.44, + "learning_rate": 8.122159240888184e-06, + "loss": 0.9872, + "step": 46553 + }, + { + "epoch": 3.44, + "learning_rate": 8.12176746655485e-06, + "loss": 0.9412, + "step": 46554 + }, + { + "epoch": 3.44, + "learning_rate": 8.121375695209754e-06, + "loss": 0.931, + "step": 46555 + }, + { + "epoch": 3.44, + "learning_rate": 8.120983926853526e-06, + "loss": 0.992, + "step": 46556 + }, + { + "epoch": 3.44, + "learning_rate": 8.120592161486787e-06, + "loss": 0.8602, + "step": 46557 + }, + { + "epoch": 3.44, + "learning_rate": 8.120200399110157e-06, + "loss": 0.9721, + "step": 46558 + }, + { + "epoch": 3.44, + "learning_rate": 8.119808639724261e-06, + "loss": 0.9686, + "step": 46559 + }, + { + "epoch": 3.44, + "learning_rate": 8.119416883329721e-06, + "loss": 1.0341, + "step": 46560 + }, + { + "epoch": 3.44, + "learning_rate": 8.119025129927164e-06, + "loss": 1.1132, + "step": 46561 + }, + { + "epoch": 3.44, + "learning_rate": 8.11863337951721e-06, + "loss": 0.9328, + "step": 46562 + }, + { + "epoch": 3.44, + "learning_rate": 8.118241632100485e-06, + "loss": 1.0189, + "step": 46563 + }, + { + "epoch": 3.44, + "learning_rate": 8.11784988767761e-06, + "loss": 0.9579, + "step": 46564 + }, + { + "epoch": 3.44, + "learning_rate": 8.117458146249208e-06, + "loss": 0.9192, + "step": 46565 + }, + { + "epoch": 3.44, + "learning_rate": 8.117066407815907e-06, + "loss": 0.9521, + "step": 46566 + }, + { + "epoch": 3.44, + "learning_rate": 8.11667467237832e-06, + "loss": 1.0209, + "step": 46567 + }, + { + "epoch": 3.44, + "learning_rate": 8.116282939937084e-06, + "loss": 1.0061, + "step": 46568 + }, + { + "epoch": 3.44, + "learning_rate": 8.115891210492805e-06, + "loss": 1.0662, + "step": 46569 + }, + { + "epoch": 3.44, + "learning_rate": 8.115499484046125e-06, + "loss": 1.0372, + "step": 46570 + }, + { + "epoch": 3.44, + "learning_rate": 8.115107760597655e-06, + "loss": 1.0229, + "step": 46571 + }, + { + "epoch": 3.44, + "learning_rate": 8.114716040148023e-06, + "loss": 1.0093, + "step": 46572 + }, + { + "epoch": 3.44, + "learning_rate": 8.114324322697849e-06, + "loss": 0.9912, + "step": 46573 + }, + { + "epoch": 3.44, + "learning_rate": 8.113932608247759e-06, + "loss": 0.9822, + "step": 46574 + }, + { + "epoch": 3.44, + "learning_rate": 8.113540896798375e-06, + "loss": 1.0542, + "step": 46575 + }, + { + "epoch": 3.44, + "learning_rate": 8.11314918835032e-06, + "loss": 1.0232, + "step": 46576 + }, + { + "epoch": 3.44, + "learning_rate": 8.112757482904219e-06, + "loss": 0.9936, + "step": 46577 + }, + { + "epoch": 3.44, + "learning_rate": 8.11236578046069e-06, + "loss": 0.8781, + "step": 46578 + }, + { + "epoch": 3.44, + "learning_rate": 8.111974081020363e-06, + "loss": 0.9483, + "step": 46579 + }, + { + "epoch": 3.44, + "learning_rate": 8.111582384583859e-06, + "loss": 1.0292, + "step": 46580 + }, + { + "epoch": 3.44, + "learning_rate": 8.1111906911518e-06, + "loss": 0.8892, + "step": 46581 + }, + { + "epoch": 3.44, + "learning_rate": 8.110799000724807e-06, + "loss": 1.0367, + "step": 46582 + }, + { + "epoch": 3.44, + "learning_rate": 8.110407313303508e-06, + "loss": 1.0713, + "step": 46583 + }, + { + "epoch": 3.44, + "learning_rate": 8.110015628888524e-06, + "loss": 1.061, + "step": 46584 + }, + { + "epoch": 3.44, + "learning_rate": 8.109623947480478e-06, + "loss": 0.9198, + "step": 46585 + }, + { + "epoch": 3.44, + "learning_rate": 8.109232269079997e-06, + "loss": 1.0653, + "step": 46586 + }, + { + "epoch": 3.44, + "learning_rate": 8.108840593687692e-06, + "loss": 0.9139, + "step": 46587 + }, + { + "epoch": 3.44, + "learning_rate": 8.108448921304203e-06, + "loss": 1.0213, + "step": 46588 + }, + { + "epoch": 3.44, + "learning_rate": 8.108057251930137e-06, + "loss": 1.0476, + "step": 46589 + }, + { + "epoch": 3.44, + "learning_rate": 8.107665585566131e-06, + "loss": 0.8861, + "step": 46590 + }, + { + "epoch": 3.44, + "learning_rate": 8.107273922212797e-06, + "loss": 0.974, + "step": 46591 + }, + { + "epoch": 3.44, + "learning_rate": 8.106882261870769e-06, + "loss": 1.0341, + "step": 46592 + }, + { + "epoch": 3.44, + "learning_rate": 8.106490604540657e-06, + "loss": 1.1379, + "step": 46593 + }, + { + "epoch": 3.44, + "learning_rate": 8.106098950223098e-06, + "loss": 0.8825, + "step": 46594 + }, + { + "epoch": 3.44, + "learning_rate": 8.105707298918707e-06, + "loss": 1.0118, + "step": 46595 + }, + { + "epoch": 3.44, + "learning_rate": 8.105315650628105e-06, + "loss": 1.0455, + "step": 46596 + }, + { + "epoch": 3.44, + "learning_rate": 8.104924005351923e-06, + "loss": 1.071, + "step": 46597 + }, + { + "epoch": 3.44, + "learning_rate": 8.104532363090776e-06, + "loss": 1.0268, + "step": 46598 + }, + { + "epoch": 3.44, + "learning_rate": 8.104140723845294e-06, + "loss": 1.1057, + "step": 46599 + }, + { + "epoch": 3.44, + "learning_rate": 8.103749087616094e-06, + "loss": 1.0886, + "step": 46600 + }, + { + "epoch": 3.44, + "learning_rate": 8.103357454403805e-06, + "loss": 1.0416, + "step": 46601 + }, + { + "epoch": 3.44, + "learning_rate": 8.102965824209044e-06, + "loss": 0.8707, + "step": 46602 + }, + { + "epoch": 3.44, + "learning_rate": 8.10257419703244e-06, + "loss": 1.0484, + "step": 46603 + }, + { + "epoch": 3.44, + "learning_rate": 8.102182572874614e-06, + "loss": 1.0014, + "step": 46604 + }, + { + "epoch": 3.44, + "learning_rate": 8.101790951736186e-06, + "loss": 0.9922, + "step": 46605 + }, + { + "epoch": 3.44, + "learning_rate": 8.101399333617783e-06, + "loss": 1.0422, + "step": 46606 + }, + { + "epoch": 3.44, + "learning_rate": 8.101007718520024e-06, + "loss": 1.0616, + "step": 46607 + }, + { + "epoch": 3.44, + "learning_rate": 8.100616106443538e-06, + "loss": 1.11, + "step": 46608 + }, + { + "epoch": 3.44, + "learning_rate": 8.10022449738894e-06, + "loss": 1.0943, + "step": 46609 + }, + { + "epoch": 3.44, + "learning_rate": 8.099832891356863e-06, + "loss": 1.0337, + "step": 46610 + }, + { + "epoch": 3.44, + "learning_rate": 8.09944128834792e-06, + "loss": 0.9683, + "step": 46611 + }, + { + "epoch": 3.44, + "learning_rate": 8.099049688362742e-06, + "loss": 0.8618, + "step": 46612 + }, + { + "epoch": 3.44, + "learning_rate": 8.09865809140195e-06, + "loss": 1.033, + "step": 46613 + }, + { + "epoch": 3.44, + "learning_rate": 8.09826649746616e-06, + "loss": 0.9818, + "step": 46614 + }, + { + "epoch": 3.44, + "learning_rate": 8.097874906556009e-06, + "loss": 0.9183, + "step": 46615 + }, + { + "epoch": 3.44, + "learning_rate": 8.097483318672106e-06, + "loss": 0.9627, + "step": 46616 + }, + { + "epoch": 3.44, + "learning_rate": 8.09709173381508e-06, + "loss": 1.0035, + "step": 46617 + }, + { + "epoch": 3.44, + "learning_rate": 8.096700151985554e-06, + "loss": 1.0808, + "step": 46618 + }, + { + "epoch": 3.44, + "learning_rate": 8.096308573184152e-06, + "loss": 0.9972, + "step": 46619 + }, + { + "epoch": 3.44, + "learning_rate": 8.095916997411494e-06, + "loss": 0.967, + "step": 46620 + }, + { + "epoch": 3.44, + "learning_rate": 8.095525424668209e-06, + "loss": 1.0692, + "step": 46621 + }, + { + "epoch": 3.44, + "learning_rate": 8.095133854954913e-06, + "loss": 0.9864, + "step": 46622 + }, + { + "epoch": 3.44, + "learning_rate": 8.094742288272231e-06, + "loss": 0.9523, + "step": 46623 + }, + { + "epoch": 3.45, + "learning_rate": 8.094350724620787e-06, + "loss": 0.9838, + "step": 46624 + }, + { + "epoch": 3.45, + "learning_rate": 8.093959164001203e-06, + "loss": 0.9685, + "step": 46625 + }, + { + "epoch": 3.45, + "learning_rate": 8.093567606414105e-06, + "loss": 0.9844, + "step": 46626 + }, + { + "epoch": 3.45, + "learning_rate": 8.093176051860112e-06, + "loss": 0.9592, + "step": 46627 + }, + { + "epoch": 3.45, + "learning_rate": 8.092784500339849e-06, + "loss": 0.9024, + "step": 46628 + }, + { + "epoch": 3.45, + "learning_rate": 8.092392951853939e-06, + "loss": 0.9698, + "step": 46629 + }, + { + "epoch": 3.45, + "learning_rate": 8.092001406403006e-06, + "loss": 0.7653, + "step": 46630 + }, + { + "epoch": 3.45, + "learning_rate": 8.091609863987672e-06, + "loss": 1.0198, + "step": 46631 + }, + { + "epoch": 3.45, + "learning_rate": 8.091218324608554e-06, + "loss": 1.0386, + "step": 46632 + }, + { + "epoch": 3.45, + "learning_rate": 8.090826788266286e-06, + "loss": 1.1522, + "step": 46633 + }, + { + "epoch": 3.45, + "learning_rate": 8.09043525496148e-06, + "loss": 1.137, + "step": 46634 + }, + { + "epoch": 3.45, + "learning_rate": 8.090043724694769e-06, + "loss": 0.9873, + "step": 46635 + }, + { + "epoch": 3.45, + "learning_rate": 8.089652197466767e-06, + "loss": 0.8958, + "step": 46636 + }, + { + "epoch": 3.45, + "learning_rate": 8.089260673278105e-06, + "loss": 0.9701, + "step": 46637 + }, + { + "epoch": 3.45, + "learning_rate": 8.088869152129397e-06, + "loss": 1.1448, + "step": 46638 + }, + { + "epoch": 3.45, + "learning_rate": 8.088477634021277e-06, + "loss": 1.0714, + "step": 46639 + }, + { + "epoch": 3.45, + "learning_rate": 8.088086118954354e-06, + "loss": 0.9991, + "step": 46640 + }, + { + "epoch": 3.45, + "learning_rate": 8.087694606929267e-06, + "loss": 1.0229, + "step": 46641 + }, + { + "epoch": 3.45, + "learning_rate": 8.087303097946629e-06, + "loss": 0.9795, + "step": 46642 + }, + { + "epoch": 3.45, + "learning_rate": 8.08691159200706e-06, + "loss": 0.9981, + "step": 46643 + }, + { + "epoch": 3.45, + "learning_rate": 8.08652008911119e-06, + "loss": 0.921, + "step": 46644 + }, + { + "epoch": 3.45, + "learning_rate": 8.086128589259637e-06, + "loss": 1.0729, + "step": 46645 + }, + { + "epoch": 3.45, + "learning_rate": 8.085737092453029e-06, + "loss": 0.9695, + "step": 46646 + }, + { + "epoch": 3.45, + "learning_rate": 8.085345598691983e-06, + "loss": 1.0352, + "step": 46647 + }, + { + "epoch": 3.45, + "learning_rate": 8.084954107977128e-06, + "loss": 1.0772, + "step": 46648 + }, + { + "epoch": 3.45, + "learning_rate": 8.08456262030908e-06, + "loss": 1.0184, + "step": 46649 + }, + { + "epoch": 3.45, + "learning_rate": 8.084171135688468e-06, + "loss": 0.9698, + "step": 46650 + }, + { + "epoch": 3.45, + "learning_rate": 8.083779654115913e-06, + "loss": 1.0824, + "step": 46651 + }, + { + "epoch": 3.45, + "learning_rate": 8.083388175592036e-06, + "loss": 1.0035, + "step": 46652 + }, + { + "epoch": 3.45, + "learning_rate": 8.08299670011746e-06, + "loss": 0.9179, + "step": 46653 + }, + { + "epoch": 3.45, + "learning_rate": 8.082605227692808e-06, + "loss": 0.9861, + "step": 46654 + }, + { + "epoch": 3.45, + "learning_rate": 8.082213758318707e-06, + "loss": 0.9612, + "step": 46655 + }, + { + "epoch": 3.45, + "learning_rate": 8.081822291995774e-06, + "loss": 1.0509, + "step": 46656 + }, + { + "epoch": 3.45, + "learning_rate": 8.081430828724636e-06, + "loss": 1.0218, + "step": 46657 + }, + { + "epoch": 3.45, + "learning_rate": 8.081039368505911e-06, + "loss": 0.9996, + "step": 46658 + }, + { + "epoch": 3.45, + "learning_rate": 8.080647911340228e-06, + "loss": 0.9712, + "step": 46659 + }, + { + "epoch": 3.45, + "learning_rate": 8.08025645722821e-06, + "loss": 0.8818, + "step": 46660 + }, + { + "epoch": 3.45, + "learning_rate": 8.07986500617047e-06, + "loss": 0.996, + "step": 46661 + }, + { + "epoch": 3.45, + "learning_rate": 8.079473558167642e-06, + "loss": 0.9726, + "step": 46662 + }, + { + "epoch": 3.45, + "learning_rate": 8.079082113220339e-06, + "loss": 0.9613, + "step": 46663 + }, + { + "epoch": 3.45, + "learning_rate": 8.078690671329192e-06, + "loss": 1.1385, + "step": 46664 + }, + { + "epoch": 3.45, + "learning_rate": 8.078299232494819e-06, + "loss": 0.9001, + "step": 46665 + }, + { + "epoch": 3.45, + "learning_rate": 8.077907796717847e-06, + "loss": 0.9778, + "step": 46666 + }, + { + "epoch": 3.45, + "learning_rate": 8.077516363998894e-06, + "loss": 1.0839, + "step": 46667 + }, + { + "epoch": 3.45, + "learning_rate": 8.077124934338586e-06, + "loss": 1.1233, + "step": 46668 + }, + { + "epoch": 3.45, + "learning_rate": 8.076733507737546e-06, + "loss": 0.9766, + "step": 46669 + }, + { + "epoch": 3.45, + "learning_rate": 8.076342084196392e-06, + "loss": 0.9799, + "step": 46670 + }, + { + "epoch": 3.45, + "learning_rate": 8.075950663715754e-06, + "loss": 0.9971, + "step": 46671 + }, + { + "epoch": 3.45, + "learning_rate": 8.075559246296248e-06, + "loss": 0.9948, + "step": 46672 + }, + { + "epoch": 3.45, + "learning_rate": 8.075167831938502e-06, + "loss": 0.9963, + "step": 46673 + }, + { + "epoch": 3.45, + "learning_rate": 8.074776420643133e-06, + "loss": 0.936, + "step": 46674 + }, + { + "epoch": 3.45, + "learning_rate": 8.07438501241077e-06, + "loss": 1.0847, + "step": 46675 + }, + { + "epoch": 3.45, + "learning_rate": 8.073993607242032e-06, + "loss": 1.0333, + "step": 46676 + }, + { + "epoch": 3.45, + "learning_rate": 8.073602205137543e-06, + "loss": 0.9951, + "step": 46677 + }, + { + "epoch": 3.45, + "learning_rate": 8.073210806097929e-06, + "loss": 1.0555, + "step": 46678 + }, + { + "epoch": 3.45, + "learning_rate": 8.072819410123802e-06, + "loss": 0.9556, + "step": 46679 + }, + { + "epoch": 3.45, + "learning_rate": 8.072428017215798e-06, + "loss": 1.087, + "step": 46680 + }, + { + "epoch": 3.45, + "learning_rate": 8.072036627374526e-06, + "loss": 0.9976, + "step": 46681 + }, + { + "epoch": 3.45, + "learning_rate": 8.071645240600626e-06, + "loss": 0.9611, + "step": 46682 + }, + { + "epoch": 3.45, + "learning_rate": 8.071253856894701e-06, + "loss": 1.0468, + "step": 46683 + }, + { + "epoch": 3.45, + "learning_rate": 8.07086247625739e-06, + "loss": 0.9414, + "step": 46684 + }, + { + "epoch": 3.45, + "learning_rate": 8.070471098689305e-06, + "loss": 0.9239, + "step": 46685 + }, + { + "epoch": 3.45, + "learning_rate": 8.070079724191078e-06, + "loss": 1.0464, + "step": 46686 + }, + { + "epoch": 3.45, + "learning_rate": 8.06968835276332e-06, + "loss": 1.0267, + "step": 46687 + }, + { + "epoch": 3.45, + "learning_rate": 8.069296984406666e-06, + "loss": 0.927, + "step": 46688 + }, + { + "epoch": 3.45, + "learning_rate": 8.068905619121733e-06, + "loss": 1.0302, + "step": 46689 + }, + { + "epoch": 3.45, + "learning_rate": 8.06851425690914e-06, + "loss": 0.9832, + "step": 46690 + }, + { + "epoch": 3.45, + "learning_rate": 8.068122897769514e-06, + "loss": 1.1211, + "step": 46691 + }, + { + "epoch": 3.45, + "learning_rate": 8.067731541703478e-06, + "loss": 0.9572, + "step": 46692 + }, + { + "epoch": 3.45, + "learning_rate": 8.067340188711654e-06, + "loss": 0.8702, + "step": 46693 + }, + { + "epoch": 3.45, + "learning_rate": 8.066948838794661e-06, + "loss": 0.9612, + "step": 46694 + }, + { + "epoch": 3.45, + "learning_rate": 8.066557491953126e-06, + "loss": 0.9954, + "step": 46695 + }, + { + "epoch": 3.45, + "learning_rate": 8.066166148187672e-06, + "loss": 0.9294, + "step": 46696 + }, + { + "epoch": 3.45, + "learning_rate": 8.065774807498918e-06, + "loss": 1.0712, + "step": 46697 + }, + { + "epoch": 3.45, + "learning_rate": 8.065383469887491e-06, + "loss": 0.9965, + "step": 46698 + }, + { + "epoch": 3.45, + "learning_rate": 8.064992135354008e-06, + "loss": 0.9357, + "step": 46699 + }, + { + "epoch": 3.45, + "learning_rate": 8.064600803899098e-06, + "loss": 0.9212, + "step": 46700 + }, + { + "epoch": 3.45, + "learning_rate": 8.064209475523378e-06, + "loss": 0.9853, + "step": 46701 + }, + { + "epoch": 3.45, + "learning_rate": 8.063818150227475e-06, + "loss": 1.06, + "step": 46702 + }, + { + "epoch": 3.45, + "learning_rate": 8.063426828012007e-06, + "loss": 1.0173, + "step": 46703 + }, + { + "epoch": 3.45, + "learning_rate": 8.063035508877604e-06, + "loss": 0.964, + "step": 46704 + }, + { + "epoch": 3.45, + "learning_rate": 8.062644192824877e-06, + "loss": 1.0361, + "step": 46705 + }, + { + "epoch": 3.45, + "learning_rate": 8.062252879854464e-06, + "loss": 0.929, + "step": 46706 + }, + { + "epoch": 3.45, + "learning_rate": 8.061861569966974e-06, + "loss": 1.0955, + "step": 46707 + }, + { + "epoch": 3.45, + "learning_rate": 8.061470263163032e-06, + "loss": 1.024, + "step": 46708 + }, + { + "epoch": 3.45, + "learning_rate": 8.061078959443267e-06, + "loss": 1.0512, + "step": 46709 + }, + { + "epoch": 3.45, + "learning_rate": 8.060687658808294e-06, + "loss": 0.9218, + "step": 46710 + }, + { + "epoch": 3.45, + "learning_rate": 8.060296361258744e-06, + "loss": 0.9125, + "step": 46711 + }, + { + "epoch": 3.45, + "learning_rate": 8.05990506679523e-06, + "loss": 1.0556, + "step": 46712 + }, + { + "epoch": 3.45, + "learning_rate": 8.059513775418382e-06, + "loss": 1.0112, + "step": 46713 + }, + { + "epoch": 3.45, + "learning_rate": 8.059122487128818e-06, + "loss": 1.0598, + "step": 46714 + }, + { + "epoch": 3.45, + "learning_rate": 8.058731201927165e-06, + "loss": 1.0392, + "step": 46715 + }, + { + "epoch": 3.45, + "learning_rate": 8.058339919814042e-06, + "loss": 0.9803, + "step": 46716 + }, + { + "epoch": 3.45, + "learning_rate": 8.057948640790071e-06, + "loss": 1.0492, + "step": 46717 + }, + { + "epoch": 3.45, + "learning_rate": 8.057557364855877e-06, + "loss": 1.0189, + "step": 46718 + }, + { + "epoch": 3.45, + "learning_rate": 8.057166092012081e-06, + "loss": 0.9362, + "step": 46719 + }, + { + "epoch": 3.45, + "learning_rate": 8.056774822259308e-06, + "loss": 1.0649, + "step": 46720 + }, + { + "epoch": 3.45, + "learning_rate": 8.056383555598175e-06, + "loss": 0.9409, + "step": 46721 + }, + { + "epoch": 3.45, + "learning_rate": 8.05599229202931e-06, + "loss": 1.0337, + "step": 46722 + }, + { + "epoch": 3.45, + "learning_rate": 8.055601031553332e-06, + "loss": 1.0258, + "step": 46723 + }, + { + "epoch": 3.45, + "learning_rate": 8.055209774170867e-06, + "loss": 1.0172, + "step": 46724 + }, + { + "epoch": 3.45, + "learning_rate": 8.054818519882537e-06, + "loss": 0.9784, + "step": 46725 + }, + { + "epoch": 3.45, + "learning_rate": 8.054427268688957e-06, + "loss": 0.9923, + "step": 46726 + }, + { + "epoch": 3.45, + "learning_rate": 8.054036020590762e-06, + "loss": 0.9793, + "step": 46727 + }, + { + "epoch": 3.45, + "learning_rate": 8.05364477558856e-06, + "loss": 1.0684, + "step": 46728 + }, + { + "epoch": 3.45, + "learning_rate": 8.05325353368299e-06, + "loss": 0.9884, + "step": 46729 + }, + { + "epoch": 3.45, + "learning_rate": 8.052862294874657e-06, + "loss": 0.9811, + "step": 46730 + }, + { + "epoch": 3.45, + "learning_rate": 8.052471059164201e-06, + "loss": 1.0198, + "step": 46731 + }, + { + "epoch": 3.45, + "learning_rate": 8.052079826552229e-06, + "loss": 0.9332, + "step": 46732 + }, + { + "epoch": 3.45, + "learning_rate": 8.051688597039376e-06, + "loss": 1.023, + "step": 46733 + }, + { + "epoch": 3.45, + "learning_rate": 8.051297370626255e-06, + "loss": 0.9388, + "step": 46734 + }, + { + "epoch": 3.45, + "learning_rate": 8.050906147313491e-06, + "loss": 1.0214, + "step": 46735 + }, + { + "epoch": 3.45, + "learning_rate": 8.05051492710171e-06, + "loss": 0.9971, + "step": 46736 + }, + { + "epoch": 3.45, + "learning_rate": 8.050123709991528e-06, + "loss": 0.9921, + "step": 46737 + }, + { + "epoch": 3.45, + "learning_rate": 8.049732495983576e-06, + "loss": 1.0643, + "step": 46738 + }, + { + "epoch": 3.45, + "learning_rate": 8.049341285078467e-06, + "loss": 1.0644, + "step": 46739 + }, + { + "epoch": 3.45, + "learning_rate": 8.048950077276832e-06, + "loss": 1.0601, + "step": 46740 + }, + { + "epoch": 3.45, + "learning_rate": 8.048558872579288e-06, + "loss": 0.9144, + "step": 46741 + }, + { + "epoch": 3.45, + "learning_rate": 8.04816767098646e-06, + "loss": 0.9534, + "step": 46742 + }, + { + "epoch": 3.45, + "learning_rate": 8.047776472498966e-06, + "loss": 0.9879, + "step": 46743 + }, + { + "epoch": 3.45, + "learning_rate": 8.047385277117436e-06, + "loss": 0.8753, + "step": 46744 + }, + { + "epoch": 3.45, + "learning_rate": 8.046994084842487e-06, + "loss": 0.875, + "step": 46745 + }, + { + "epoch": 3.45, + "learning_rate": 8.04660289567474e-06, + "loss": 1.0262, + "step": 46746 + }, + { + "epoch": 3.45, + "learning_rate": 8.046211709614822e-06, + "loss": 0.9919, + "step": 46747 + }, + { + "epoch": 3.45, + "learning_rate": 8.045820526663352e-06, + "loss": 0.9433, + "step": 46748 + }, + { + "epoch": 3.45, + "learning_rate": 8.045429346820956e-06, + "loss": 0.941, + "step": 46749 + }, + { + "epoch": 3.45, + "learning_rate": 8.04503817008825e-06, + "loss": 0.9282, + "step": 46750 + }, + { + "epoch": 3.45, + "learning_rate": 8.044646996465866e-06, + "loss": 1.0446, + "step": 46751 + }, + { + "epoch": 3.45, + "learning_rate": 8.044255825954414e-06, + "loss": 0.9455, + "step": 46752 + }, + { + "epoch": 3.45, + "learning_rate": 8.04386465855453e-06, + "loss": 0.8983, + "step": 46753 + }, + { + "epoch": 3.45, + "learning_rate": 8.043473494266825e-06, + "loss": 1.0152, + "step": 46754 + }, + { + "epoch": 3.45, + "learning_rate": 8.043082333091925e-06, + "loss": 0.9211, + "step": 46755 + }, + { + "epoch": 3.45, + "learning_rate": 8.042691175030455e-06, + "loss": 1.0377, + "step": 46756 + }, + { + "epoch": 3.45, + "learning_rate": 8.042300020083034e-06, + "loss": 1.0357, + "step": 46757 + }, + { + "epoch": 3.45, + "learning_rate": 8.041908868250286e-06, + "loss": 0.9745, + "step": 46758 + }, + { + "epoch": 3.45, + "learning_rate": 8.041517719532832e-06, + "loss": 1.0372, + "step": 46759 + }, + { + "epoch": 3.46, + "learning_rate": 8.041126573931296e-06, + "loss": 0.989, + "step": 46760 + }, + { + "epoch": 3.46, + "learning_rate": 8.040735431446301e-06, + "loss": 0.9213, + "step": 46761 + }, + { + "epoch": 3.46, + "learning_rate": 8.040344292078467e-06, + "loss": 1.059, + "step": 46762 + }, + { + "epoch": 3.46, + "learning_rate": 8.039953155828418e-06, + "loss": 0.9771, + "step": 46763 + }, + { + "epoch": 3.46, + "learning_rate": 8.039562022696774e-06, + "loss": 1.0737, + "step": 46764 + }, + { + "epoch": 3.46, + "learning_rate": 8.03917089268416e-06, + "loss": 1.0408, + "step": 46765 + }, + { + "epoch": 3.46, + "learning_rate": 8.038779765791194e-06, + "loss": 1.0066, + "step": 46766 + }, + { + "epoch": 3.46, + "learning_rate": 8.038388642018505e-06, + "loss": 1.0097, + "step": 46767 + }, + { + "epoch": 3.46, + "learning_rate": 8.037997521366709e-06, + "loss": 0.9916, + "step": 46768 + }, + { + "epoch": 3.46, + "learning_rate": 8.037606403836434e-06, + "loss": 1.1053, + "step": 46769 + }, + { + "epoch": 3.46, + "learning_rate": 8.037215289428296e-06, + "loss": 0.9885, + "step": 46770 + }, + { + "epoch": 3.46, + "learning_rate": 8.036824178142923e-06, + "loss": 0.9914, + "step": 46771 + }, + { + "epoch": 3.46, + "learning_rate": 8.036433069980936e-06, + "loss": 1.0042, + "step": 46772 + }, + { + "epoch": 3.46, + "learning_rate": 8.03604196494295e-06, + "loss": 0.9599, + "step": 46773 + }, + { + "epoch": 3.46, + "learning_rate": 8.0356508630296e-06, + "loss": 0.8368, + "step": 46774 + }, + { + "epoch": 3.46, + "learning_rate": 8.035259764241494e-06, + "loss": 0.9873, + "step": 46775 + }, + { + "epoch": 3.46, + "learning_rate": 8.034868668579269e-06, + "loss": 1.0095, + "step": 46776 + }, + { + "epoch": 3.46, + "learning_rate": 8.034477576043532e-06, + "loss": 0.898, + "step": 46777 + }, + { + "epoch": 3.46, + "learning_rate": 8.034086486634919e-06, + "loss": 1.0326, + "step": 46778 + }, + { + "epoch": 3.46, + "learning_rate": 8.033695400354043e-06, + "loss": 0.9773, + "step": 46779 + }, + { + "epoch": 3.46, + "learning_rate": 8.033304317201534e-06, + "loss": 1.0359, + "step": 46780 + }, + { + "epoch": 3.46, + "learning_rate": 8.032913237178007e-06, + "loss": 0.9678, + "step": 46781 + }, + { + "epoch": 3.46, + "learning_rate": 8.032522160284087e-06, + "loss": 0.9493, + "step": 46782 + }, + { + "epoch": 3.46, + "learning_rate": 8.032131086520396e-06, + "loss": 1.0307, + "step": 46783 + }, + { + "epoch": 3.46, + "learning_rate": 8.031740015887556e-06, + "loss": 1.0357, + "step": 46784 + }, + { + "epoch": 3.46, + "learning_rate": 8.03134894838619e-06, + "loss": 0.9666, + "step": 46785 + }, + { + "epoch": 3.46, + "learning_rate": 8.030957884016917e-06, + "loss": 0.9767, + "step": 46786 + }, + { + "epoch": 3.46, + "learning_rate": 8.030566822780364e-06, + "loss": 0.9656, + "step": 46787 + }, + { + "epoch": 3.46, + "learning_rate": 8.03017576467715e-06, + "loss": 1.0747, + "step": 46788 + }, + { + "epoch": 3.46, + "learning_rate": 8.0297847097079e-06, + "loss": 0.994, + "step": 46789 + }, + { + "epoch": 3.46, + "learning_rate": 8.029393657873235e-06, + "loss": 1.0289, + "step": 46790 + }, + { + "epoch": 3.46, + "learning_rate": 8.029002609173774e-06, + "loss": 1.1588, + "step": 46791 + }, + { + "epoch": 3.46, + "learning_rate": 8.028611563610143e-06, + "loss": 0.9348, + "step": 46792 + }, + { + "epoch": 3.46, + "learning_rate": 8.02822052118296e-06, + "loss": 0.9778, + "step": 46793 + }, + { + "epoch": 3.46, + "learning_rate": 8.027829481892857e-06, + "loss": 0.9818, + "step": 46794 + }, + { + "epoch": 3.46, + "learning_rate": 8.02743844574044e-06, + "loss": 1.1359, + "step": 46795 + }, + { + "epoch": 3.46, + "learning_rate": 8.02704741272635e-06, + "loss": 1.1207, + "step": 46796 + }, + { + "epoch": 3.46, + "learning_rate": 8.02665638285119e-06, + "loss": 0.9216, + "step": 46797 + }, + { + "epoch": 3.46, + "learning_rate": 8.026265356115598e-06, + "loss": 0.9484, + "step": 46798 + }, + { + "epoch": 3.46, + "learning_rate": 8.025874332520183e-06, + "loss": 0.9906, + "step": 46799 + }, + { + "epoch": 3.46, + "learning_rate": 8.02548331206558e-06, + "loss": 0.9708, + "step": 46800 + }, + { + "epoch": 3.46, + "learning_rate": 8.025092294752404e-06, + "loss": 0.9789, + "step": 46801 + }, + { + "epoch": 3.46, + "learning_rate": 8.024701280581276e-06, + "loss": 1.089, + "step": 46802 + }, + { + "epoch": 3.46, + "learning_rate": 8.02431026955282e-06, + "loss": 1.0764, + "step": 46803 + }, + { + "epoch": 3.46, + "learning_rate": 8.02391926166766e-06, + "loss": 0.9413, + "step": 46804 + }, + { + "epoch": 3.46, + "learning_rate": 8.023528256926415e-06, + "loss": 1.0197, + "step": 46805 + }, + { + "epoch": 3.46, + "learning_rate": 8.023137255329705e-06, + "loss": 1.1022, + "step": 46806 + }, + { + "epoch": 3.46, + "learning_rate": 8.02274625687816e-06, + "loss": 1.0682, + "step": 46807 + }, + { + "epoch": 3.46, + "learning_rate": 8.022355261572395e-06, + "loss": 1.0217, + "step": 46808 + }, + { + "epoch": 3.46, + "learning_rate": 8.021964269413036e-06, + "loss": 0.9441, + "step": 46809 + }, + { + "epoch": 3.46, + "learning_rate": 8.021573280400703e-06, + "loss": 0.8982, + "step": 46810 + }, + { + "epoch": 3.46, + "learning_rate": 8.021182294536017e-06, + "loss": 1.0249, + "step": 46811 + }, + { + "epoch": 3.46, + "learning_rate": 8.020791311819604e-06, + "loss": 0.9683, + "step": 46812 + }, + { + "epoch": 3.46, + "learning_rate": 8.020400332252082e-06, + "loss": 0.8706, + "step": 46813 + }, + { + "epoch": 3.46, + "learning_rate": 8.020009355834076e-06, + "loss": 1.0576, + "step": 46814 + }, + { + "epoch": 3.46, + "learning_rate": 8.019618382566206e-06, + "loss": 0.7971, + "step": 46815 + }, + { + "epoch": 3.46, + "learning_rate": 8.019227412449095e-06, + "loss": 0.9823, + "step": 46816 + }, + { + "epoch": 3.46, + "learning_rate": 8.018836445483364e-06, + "loss": 0.9474, + "step": 46817 + }, + { + "epoch": 3.46, + "learning_rate": 8.018445481669638e-06, + "loss": 1.0722, + "step": 46818 + }, + { + "epoch": 3.46, + "learning_rate": 8.018054521008537e-06, + "loss": 1.0126, + "step": 46819 + }, + { + "epoch": 3.46, + "learning_rate": 8.017663563500679e-06, + "loss": 0.93, + "step": 46820 + }, + { + "epoch": 3.46, + "learning_rate": 8.017272609146695e-06, + "loss": 1.0219, + "step": 46821 + }, + { + "epoch": 3.46, + "learning_rate": 8.016881657947198e-06, + "loss": 0.926, + "step": 46822 + }, + { + "epoch": 3.46, + "learning_rate": 8.016490709902817e-06, + "loss": 0.936, + "step": 46823 + }, + { + "epoch": 3.46, + "learning_rate": 8.016099765014168e-06, + "loss": 0.9849, + "step": 46824 + }, + { + "epoch": 3.46, + "learning_rate": 8.015708823281878e-06, + "loss": 0.9776, + "step": 46825 + }, + { + "epoch": 3.46, + "learning_rate": 8.015317884706566e-06, + "loss": 1.0812, + "step": 46826 + }, + { + "epoch": 3.46, + "learning_rate": 8.014926949288855e-06, + "loss": 0.9381, + "step": 46827 + }, + { + "epoch": 3.46, + "learning_rate": 8.014536017029368e-06, + "loss": 1.0093, + "step": 46828 + }, + { + "epoch": 3.46, + "learning_rate": 8.014145087928723e-06, + "loss": 1.0128, + "step": 46829 + }, + { + "epoch": 3.46, + "learning_rate": 8.013754161987547e-06, + "loss": 1.0056, + "step": 46830 + }, + { + "epoch": 3.46, + "learning_rate": 8.013363239206457e-06, + "loss": 0.8645, + "step": 46831 + }, + { + "epoch": 3.46, + "learning_rate": 8.012972319586081e-06, + "loss": 1.0192, + "step": 46832 + }, + { + "epoch": 3.46, + "learning_rate": 8.012581403127035e-06, + "loss": 1.0799, + "step": 46833 + }, + { + "epoch": 3.46, + "learning_rate": 8.012190489829945e-06, + "loss": 1.049, + "step": 46834 + }, + { + "epoch": 3.46, + "learning_rate": 8.011799579695431e-06, + "loss": 0.9193, + "step": 46835 + }, + { + "epoch": 3.46, + "learning_rate": 8.011408672724115e-06, + "loss": 0.9183, + "step": 46836 + }, + { + "epoch": 3.46, + "learning_rate": 8.011017768916624e-06, + "loss": 0.9486, + "step": 46837 + }, + { + "epoch": 3.46, + "learning_rate": 8.010626868273568e-06, + "loss": 0.9461, + "step": 46838 + }, + { + "epoch": 3.46, + "learning_rate": 8.010235970795583e-06, + "loss": 0.9339, + "step": 46839 + }, + { + "epoch": 3.46, + "learning_rate": 8.009845076483278e-06, + "loss": 1.0774, + "step": 46840 + }, + { + "epoch": 3.46, + "learning_rate": 8.009454185337285e-06, + "loss": 1.0056, + "step": 46841 + }, + { + "epoch": 3.46, + "learning_rate": 8.009063297358217e-06, + "loss": 0.9981, + "step": 46842 + }, + { + "epoch": 3.46, + "learning_rate": 8.008672412546708e-06, + "loss": 1.111, + "step": 46843 + }, + { + "epoch": 3.46, + "learning_rate": 8.008281530903365e-06, + "loss": 0.9576, + "step": 46844 + }, + { + "epoch": 3.46, + "learning_rate": 8.007890652428824e-06, + "loss": 1.0451, + "step": 46845 + }, + { + "epoch": 3.46, + "learning_rate": 8.0074997771237e-06, + "loss": 1.0448, + "step": 46846 + }, + { + "epoch": 3.46, + "learning_rate": 8.007108904988612e-06, + "loss": 0.8699, + "step": 46847 + }, + { + "epoch": 3.46, + "learning_rate": 8.006718036024188e-06, + "loss": 0.9188, + "step": 46848 + }, + { + "epoch": 3.46, + "learning_rate": 8.006327170231044e-06, + "loss": 0.9549, + "step": 46849 + }, + { + "epoch": 3.46, + "learning_rate": 8.005936307609807e-06, + "loss": 1.0025, + "step": 46850 + }, + { + "epoch": 3.46, + "learning_rate": 8.005545448161097e-06, + "loss": 0.9789, + "step": 46851 + }, + { + "epoch": 3.46, + "learning_rate": 8.005154591885534e-06, + "loss": 1.044, + "step": 46852 + }, + { + "epoch": 3.46, + "learning_rate": 8.004763738783741e-06, + "loss": 0.998, + "step": 46853 + }, + { + "epoch": 3.46, + "learning_rate": 8.004372888856345e-06, + "loss": 0.9163, + "step": 46854 + }, + { + "epoch": 3.46, + "learning_rate": 8.003982042103959e-06, + "loss": 1.0041, + "step": 46855 + }, + { + "epoch": 3.46, + "learning_rate": 8.003591198527209e-06, + "loss": 1.0095, + "step": 46856 + }, + { + "epoch": 3.46, + "learning_rate": 8.00320035812672e-06, + "loss": 1.0304, + "step": 46857 + }, + { + "epoch": 3.46, + "learning_rate": 8.002809520903108e-06, + "loss": 0.9889, + "step": 46858 + }, + { + "epoch": 3.46, + "learning_rate": 8.002418686856998e-06, + "loss": 0.9893, + "step": 46859 + }, + { + "epoch": 3.46, + "learning_rate": 8.00202785598901e-06, + "loss": 0.8632, + "step": 46860 + }, + { + "epoch": 3.46, + "learning_rate": 8.00163702829977e-06, + "loss": 0.9834, + "step": 46861 + }, + { + "epoch": 3.46, + "learning_rate": 8.001246203789894e-06, + "loss": 1.0462, + "step": 46862 + }, + { + "epoch": 3.46, + "learning_rate": 8.00085538246001e-06, + "loss": 0.999, + "step": 46863 + }, + { + "epoch": 3.46, + "learning_rate": 8.000464564310733e-06, + "loss": 0.9554, + "step": 46864 + }, + { + "epoch": 3.46, + "learning_rate": 8.000073749342689e-06, + "loss": 1.1041, + "step": 46865 + }, + { + "epoch": 3.46, + "learning_rate": 7.999682937556505e-06, + "loss": 0.9789, + "step": 46866 + }, + { + "epoch": 3.46, + "learning_rate": 7.999292128952787e-06, + "loss": 1.0545, + "step": 46867 + }, + { + "epoch": 3.46, + "learning_rate": 7.998901323532174e-06, + "loss": 0.9581, + "step": 46868 + }, + { + "epoch": 3.46, + "learning_rate": 7.998510521295276e-06, + "loss": 0.9469, + "step": 46869 + }, + { + "epoch": 3.46, + "learning_rate": 7.998119722242723e-06, + "loss": 1.0452, + "step": 46870 + }, + { + "epoch": 3.46, + "learning_rate": 7.99772892637513e-06, + "loss": 0.9496, + "step": 46871 + }, + { + "epoch": 3.46, + "learning_rate": 7.997338133693122e-06, + "loss": 0.9371, + "step": 46872 + }, + { + "epoch": 3.46, + "learning_rate": 7.99694734419732e-06, + "loss": 0.9766, + "step": 46873 + }, + { + "epoch": 3.46, + "learning_rate": 7.996556557888347e-06, + "loss": 1.0563, + "step": 46874 + }, + { + "epoch": 3.46, + "learning_rate": 7.996165774766826e-06, + "loss": 1.0476, + "step": 46875 + }, + { + "epoch": 3.46, + "learning_rate": 7.995774994833372e-06, + "loss": 1.0321, + "step": 46876 + }, + { + "epoch": 3.46, + "learning_rate": 7.995384218088615e-06, + "loss": 1.0245, + "step": 46877 + }, + { + "epoch": 3.46, + "learning_rate": 7.99499344453317e-06, + "loss": 0.9584, + "step": 46878 + }, + { + "epoch": 3.46, + "learning_rate": 7.994602674167665e-06, + "loss": 1.0768, + "step": 46879 + }, + { + "epoch": 3.46, + "learning_rate": 7.994211906992714e-06, + "loss": 1.0406, + "step": 46880 + }, + { + "epoch": 3.46, + "learning_rate": 7.993821143008947e-06, + "loss": 0.9745, + "step": 46881 + }, + { + "epoch": 3.46, + "learning_rate": 7.993430382216981e-06, + "loss": 0.8933, + "step": 46882 + }, + { + "epoch": 3.46, + "learning_rate": 7.993039624617438e-06, + "loss": 1.0752, + "step": 46883 + }, + { + "epoch": 3.46, + "learning_rate": 7.992648870210943e-06, + "loss": 1.0895, + "step": 46884 + }, + { + "epoch": 3.46, + "learning_rate": 7.99225811899811e-06, + "loss": 0.8888, + "step": 46885 + }, + { + "epoch": 3.46, + "learning_rate": 7.991867370979572e-06, + "loss": 1.0725, + "step": 46886 + }, + { + "epoch": 3.46, + "learning_rate": 7.991476626155936e-06, + "loss": 0.9861, + "step": 46887 + }, + { + "epoch": 3.46, + "learning_rate": 7.99108588452784e-06, + "loss": 1.0969, + "step": 46888 + }, + { + "epoch": 3.46, + "learning_rate": 7.99069514609589e-06, + "loss": 0.8875, + "step": 46889 + }, + { + "epoch": 3.46, + "learning_rate": 7.990304410860724e-06, + "loss": 0.9497, + "step": 46890 + }, + { + "epoch": 3.46, + "learning_rate": 7.989913678822944e-06, + "loss": 1.0055, + "step": 46891 + }, + { + "epoch": 3.46, + "learning_rate": 7.989522949983192e-06, + "loss": 0.9762, + "step": 46892 + }, + { + "epoch": 3.46, + "learning_rate": 7.989132224342079e-06, + "loss": 0.8841, + "step": 46893 + }, + { + "epoch": 3.46, + "learning_rate": 7.988741501900224e-06, + "loss": 0.9824, + "step": 46894 + }, + { + "epoch": 3.47, + "learning_rate": 7.988350782658256e-06, + "loss": 1.0714, + "step": 46895 + }, + { + "epoch": 3.47, + "learning_rate": 7.98796006661679e-06, + "loss": 1.0168, + "step": 46896 + }, + { + "epoch": 3.47, + "learning_rate": 7.987569353776453e-06, + "loss": 1.0185, + "step": 46897 + }, + { + "epoch": 3.47, + "learning_rate": 7.987178644137863e-06, + "loss": 0.9131, + "step": 46898 + }, + { + "epoch": 3.47, + "learning_rate": 7.986787937701643e-06, + "loss": 1.119, + "step": 46899 + }, + { + "epoch": 3.47, + "learning_rate": 7.986397234468415e-06, + "loss": 0.986, + "step": 46900 + }, + { + "epoch": 3.47, + "learning_rate": 7.9860065344388e-06, + "loss": 0.9606, + "step": 46901 + }, + { + "epoch": 3.47, + "learning_rate": 7.98561583761342e-06, + "loss": 1.1105, + "step": 46902 + }, + { + "epoch": 3.47, + "learning_rate": 7.985225143992897e-06, + "loss": 1.0092, + "step": 46903 + }, + { + "epoch": 3.47, + "learning_rate": 7.984834453577852e-06, + "loss": 1.039, + "step": 46904 + }, + { + "epoch": 3.47, + "learning_rate": 7.984443766368906e-06, + "loss": 0.9861, + "step": 46905 + }, + { + "epoch": 3.47, + "learning_rate": 7.98405308236668e-06, + "loss": 1.1859, + "step": 46906 + }, + { + "epoch": 3.47, + "learning_rate": 7.983662401571799e-06, + "loss": 0.9799, + "step": 46907 + }, + { + "epoch": 3.47, + "learning_rate": 7.983271723984882e-06, + "loss": 0.9728, + "step": 46908 + }, + { + "epoch": 3.47, + "learning_rate": 7.982881049606548e-06, + "loss": 1.0532, + "step": 46909 + }, + { + "epoch": 3.47, + "learning_rate": 7.982490378437424e-06, + "loss": 1.0109, + "step": 46910 + }, + { + "epoch": 3.47, + "learning_rate": 7.982099710478126e-06, + "loss": 0.9321, + "step": 46911 + }, + { + "epoch": 3.47, + "learning_rate": 7.981709045729283e-06, + "loss": 0.8907, + "step": 46912 + }, + { + "epoch": 3.47, + "learning_rate": 7.981318384191512e-06, + "loss": 1.0055, + "step": 46913 + }, + { + "epoch": 3.47, + "learning_rate": 7.980927725865432e-06, + "loss": 1.0006, + "step": 46914 + }, + { + "epoch": 3.47, + "learning_rate": 7.980537070751668e-06, + "loss": 1.0019, + "step": 46915 + }, + { + "epoch": 3.47, + "learning_rate": 7.980146418850837e-06, + "loss": 0.8651, + "step": 46916 + }, + { + "epoch": 3.47, + "learning_rate": 7.97975577016357e-06, + "loss": 0.8563, + "step": 46917 + }, + { + "epoch": 3.47, + "learning_rate": 7.979365124690478e-06, + "loss": 0.9256, + "step": 46918 + }, + { + "epoch": 3.47, + "learning_rate": 7.97897448243219e-06, + "loss": 1.0339, + "step": 46919 + }, + { + "epoch": 3.47, + "learning_rate": 7.978583843389322e-06, + "loss": 1.0855, + "step": 46920 + }, + { + "epoch": 3.47, + "learning_rate": 7.978193207562502e-06, + "loss": 1.0279, + "step": 46921 + }, + { + "epoch": 3.47, + "learning_rate": 7.977802574952347e-06, + "loss": 1.065, + "step": 46922 + }, + { + "epoch": 3.47, + "learning_rate": 7.977411945559475e-06, + "loss": 1.0023, + "step": 46923 + }, + { + "epoch": 3.47, + "learning_rate": 7.977021319384516e-06, + "loss": 0.961, + "step": 46924 + }, + { + "epoch": 3.47, + "learning_rate": 7.976630696428085e-06, + "loss": 0.9819, + "step": 46925 + }, + { + "epoch": 3.47, + "learning_rate": 7.976240076690806e-06, + "loss": 1.0914, + "step": 46926 + }, + { + "epoch": 3.47, + "learning_rate": 7.9758494601733e-06, + "loss": 1.0589, + "step": 46927 + }, + { + "epoch": 3.47, + "learning_rate": 7.975458846876189e-06, + "loss": 0.9318, + "step": 46928 + }, + { + "epoch": 3.47, + "learning_rate": 7.975068236800093e-06, + "loss": 0.9876, + "step": 46929 + }, + { + "epoch": 3.47, + "learning_rate": 7.974677629945637e-06, + "loss": 0.8864, + "step": 46930 + }, + { + "epoch": 3.47, + "learning_rate": 7.974287026313441e-06, + "loss": 0.8884, + "step": 46931 + }, + { + "epoch": 3.47, + "learning_rate": 7.973896425904118e-06, + "loss": 0.9887, + "step": 46932 + }, + { + "epoch": 3.47, + "learning_rate": 7.973505828718304e-06, + "loss": 0.9455, + "step": 46933 + }, + { + "epoch": 3.47, + "learning_rate": 7.973115234756607e-06, + "loss": 0.9981, + "step": 46934 + }, + { + "epoch": 3.47, + "learning_rate": 7.972724644019663e-06, + "loss": 0.9993, + "step": 46935 + }, + { + "epoch": 3.47, + "learning_rate": 7.972334056508076e-06, + "loss": 0.9072, + "step": 46936 + }, + { + "epoch": 3.47, + "learning_rate": 7.971943472222484e-06, + "loss": 0.9171, + "step": 46937 + }, + { + "epoch": 3.47, + "learning_rate": 7.971552891163495e-06, + "loss": 1.0306, + "step": 46938 + }, + { + "epoch": 3.47, + "learning_rate": 7.971162313331741e-06, + "loss": 1.0403, + "step": 46939 + }, + { + "epoch": 3.47, + "learning_rate": 7.970771738727839e-06, + "loss": 1.1019, + "step": 46940 + }, + { + "epoch": 3.47, + "learning_rate": 7.970381167352406e-06, + "loss": 1.0026, + "step": 46941 + }, + { + "epoch": 3.47, + "learning_rate": 7.96999059920607e-06, + "loss": 1.0451, + "step": 46942 + }, + { + "epoch": 3.47, + "learning_rate": 7.969600034289448e-06, + "loss": 0.9786, + "step": 46943 + }, + { + "epoch": 3.47, + "learning_rate": 7.969209472603166e-06, + "loss": 0.9924, + "step": 46944 + }, + { + "epoch": 3.47, + "learning_rate": 7.96881891414784e-06, + "loss": 0.9439, + "step": 46945 + }, + { + "epoch": 3.47, + "learning_rate": 7.968428358924096e-06, + "loss": 1.0421, + "step": 46946 + }, + { + "epoch": 3.47, + "learning_rate": 7.968037806932553e-06, + "loss": 1.0496, + "step": 46947 + }, + { + "epoch": 3.47, + "learning_rate": 7.967647258173833e-06, + "loss": 1.0614, + "step": 46948 + }, + { + "epoch": 3.47, + "learning_rate": 7.967256712648557e-06, + "loss": 0.9656, + "step": 46949 + }, + { + "epoch": 3.47, + "learning_rate": 7.966866170357345e-06, + "loss": 0.9542, + "step": 46950 + }, + { + "epoch": 3.47, + "learning_rate": 7.966475631300822e-06, + "loss": 1.1005, + "step": 46951 + }, + { + "epoch": 3.47, + "learning_rate": 7.966085095479606e-06, + "loss": 1.0094, + "step": 46952 + }, + { + "epoch": 3.47, + "learning_rate": 7.96569456289432e-06, + "loss": 0.9813, + "step": 46953 + }, + { + "epoch": 3.47, + "learning_rate": 7.965304033545584e-06, + "loss": 0.9513, + "step": 46954 + }, + { + "epoch": 3.47, + "learning_rate": 7.964913507434021e-06, + "loss": 1.024, + "step": 46955 + }, + { + "epoch": 3.47, + "learning_rate": 7.96452298456025e-06, + "loss": 0.8986, + "step": 46956 + }, + { + "epoch": 3.47, + "learning_rate": 7.964132464924897e-06, + "loss": 1.0223, + "step": 46957 + }, + { + "epoch": 3.47, + "learning_rate": 7.963741948528575e-06, + "loss": 1.0562, + "step": 46958 + }, + { + "epoch": 3.47, + "learning_rate": 7.963351435371918e-06, + "loss": 0.9264, + "step": 46959 + }, + { + "epoch": 3.47, + "learning_rate": 7.962960925455535e-06, + "loss": 0.9802, + "step": 46960 + }, + { + "epoch": 3.47, + "learning_rate": 7.962570418780052e-06, + "loss": 1.0035, + "step": 46961 + }, + { + "epoch": 3.47, + "learning_rate": 7.96217991534609e-06, + "loss": 0.9706, + "step": 46962 + }, + { + "epoch": 3.47, + "learning_rate": 7.96178941515427e-06, + "loss": 1.0429, + "step": 46963 + }, + { + "epoch": 3.47, + "learning_rate": 7.961398918205217e-06, + "loss": 1.0623, + "step": 46964 + }, + { + "epoch": 3.47, + "learning_rate": 7.961008424499546e-06, + "loss": 1.0123, + "step": 46965 + }, + { + "epoch": 3.47, + "learning_rate": 7.960617934037884e-06, + "loss": 1.1313, + "step": 46966 + }, + { + "epoch": 3.47, + "learning_rate": 7.960227446820848e-06, + "loss": 0.9588, + "step": 46967 + }, + { + "epoch": 3.47, + "learning_rate": 7.959836962849061e-06, + "loss": 0.8825, + "step": 46968 + }, + { + "epoch": 3.47, + "learning_rate": 7.959446482123145e-06, + "loss": 0.9927, + "step": 46969 + }, + { + "epoch": 3.47, + "learning_rate": 7.959056004643719e-06, + "loss": 0.948, + "step": 46970 + }, + { + "epoch": 3.47, + "learning_rate": 7.958665530411408e-06, + "loss": 0.9342, + "step": 46971 + }, + { + "epoch": 3.47, + "learning_rate": 7.958275059426828e-06, + "loss": 0.9274, + "step": 46972 + }, + { + "epoch": 3.47, + "learning_rate": 7.957884591690606e-06, + "loss": 1.0128, + "step": 46973 + }, + { + "epoch": 3.47, + "learning_rate": 7.957494127203357e-06, + "loss": 0.8954, + "step": 46974 + }, + { + "epoch": 3.47, + "learning_rate": 7.95710366596571e-06, + "loss": 1.0057, + "step": 46975 + }, + { + "epoch": 3.47, + "learning_rate": 7.956713207978278e-06, + "loss": 1.0817, + "step": 46976 + }, + { + "epoch": 3.47, + "learning_rate": 7.95632275324169e-06, + "loss": 0.9383, + "step": 46977 + }, + { + "epoch": 3.47, + "learning_rate": 7.955932301756564e-06, + "loss": 0.995, + "step": 46978 + }, + { + "epoch": 3.47, + "learning_rate": 7.955541853523513e-06, + "loss": 0.9769, + "step": 46979 + }, + { + "epoch": 3.47, + "learning_rate": 7.955151408543173e-06, + "loss": 0.9037, + "step": 46980 + }, + { + "epoch": 3.47, + "learning_rate": 7.954760966816152e-06, + "loss": 0.9092, + "step": 46981 + }, + { + "epoch": 3.47, + "learning_rate": 7.954370528343083e-06, + "loss": 0.9289, + "step": 46982 + }, + { + "epoch": 3.47, + "learning_rate": 7.953980093124575e-06, + "loss": 1.0884, + "step": 46983 + }, + { + "epoch": 3.47, + "learning_rate": 7.953589661161262e-06, + "loss": 0.9749, + "step": 46984 + }, + { + "epoch": 3.47, + "learning_rate": 7.953199232453751e-06, + "loss": 0.8752, + "step": 46985 + }, + { + "epoch": 3.47, + "learning_rate": 7.952808807002678e-06, + "loss": 0.947, + "step": 46986 + }, + { + "epoch": 3.47, + "learning_rate": 7.952418384808655e-06, + "loss": 1.0357, + "step": 46987 + }, + { + "epoch": 3.47, + "learning_rate": 7.952027965872302e-06, + "loss": 0.9961, + "step": 46988 + }, + { + "epoch": 3.47, + "learning_rate": 7.951637550194247e-06, + "loss": 0.8973, + "step": 46989 + }, + { + "epoch": 3.47, + "learning_rate": 7.951247137775102e-06, + "loss": 1.0406, + "step": 46990 + }, + { + "epoch": 3.47, + "learning_rate": 7.9508567286155e-06, + "loss": 0.8753, + "step": 46991 + }, + { + "epoch": 3.47, + "learning_rate": 7.950466322716051e-06, + "loss": 0.9166, + "step": 46992 + }, + { + "epoch": 3.47, + "learning_rate": 7.950075920077383e-06, + "loss": 1.0059, + "step": 46993 + }, + { + "epoch": 3.47, + "learning_rate": 7.949685520700113e-06, + "loss": 1.0233, + "step": 46994 + }, + { + "epoch": 3.47, + "learning_rate": 7.949295124584865e-06, + "loss": 1.0482, + "step": 46995 + }, + { + "epoch": 3.47, + "learning_rate": 7.948904731732261e-06, + "loss": 0.9423, + "step": 46996 + }, + { + "epoch": 3.47, + "learning_rate": 7.948514342142916e-06, + "loss": 0.9586, + "step": 46997 + }, + { + "epoch": 3.47, + "learning_rate": 7.948123955817459e-06, + "loss": 0.9751, + "step": 46998 + }, + { + "epoch": 3.47, + "learning_rate": 7.947733572756507e-06, + "loss": 1.0357, + "step": 46999 + }, + { + "epoch": 3.47, + "learning_rate": 7.94734319296068e-06, + "loss": 1.0629, + "step": 47000 + }, + { + "epoch": 3.47, + "learning_rate": 7.946952816430601e-06, + "loss": 0.991, + "step": 47001 + }, + { + "epoch": 3.47, + "learning_rate": 7.946562443166891e-06, + "loss": 0.9424, + "step": 47002 + }, + { + "epoch": 3.47, + "learning_rate": 7.94617207317017e-06, + "loss": 1.0136, + "step": 47003 + }, + { + "epoch": 3.47, + "learning_rate": 7.945781706441064e-06, + "loss": 1.041, + "step": 47004 + }, + { + "epoch": 3.47, + "learning_rate": 7.945391342980187e-06, + "loss": 0.9406, + "step": 47005 + }, + { + "epoch": 3.47, + "learning_rate": 7.945000982788162e-06, + "loss": 1.0109, + "step": 47006 + }, + { + "epoch": 3.47, + "learning_rate": 7.944610625865613e-06, + "loss": 1.0554, + "step": 47007 + }, + { + "epoch": 3.47, + "learning_rate": 7.944220272213157e-06, + "loss": 0.9193, + "step": 47008 + }, + { + "epoch": 3.47, + "learning_rate": 7.94382992183142e-06, + "loss": 0.9564, + "step": 47009 + }, + { + "epoch": 3.47, + "learning_rate": 7.943439574721015e-06, + "loss": 1.0208, + "step": 47010 + }, + { + "epoch": 3.47, + "learning_rate": 7.943049230882571e-06, + "loss": 0.9611, + "step": 47011 + }, + { + "epoch": 3.47, + "learning_rate": 7.942658890316706e-06, + "loss": 0.9579, + "step": 47012 + }, + { + "epoch": 3.47, + "learning_rate": 7.942268553024043e-06, + "loss": 1.0, + "step": 47013 + }, + { + "epoch": 3.47, + "learning_rate": 7.941878219005199e-06, + "loss": 1.0054, + "step": 47014 + }, + { + "epoch": 3.47, + "learning_rate": 7.9414878882608e-06, + "loss": 1.0028, + "step": 47015 + }, + { + "epoch": 3.47, + "learning_rate": 7.941097560791462e-06, + "loss": 1.0653, + "step": 47016 + }, + { + "epoch": 3.47, + "learning_rate": 7.94070723659781e-06, + "loss": 1.0089, + "step": 47017 + }, + { + "epoch": 3.47, + "learning_rate": 7.940316915680461e-06, + "loss": 1.0917, + "step": 47018 + }, + { + "epoch": 3.47, + "learning_rate": 7.93992659804004e-06, + "loss": 0.9972, + "step": 47019 + }, + { + "epoch": 3.47, + "learning_rate": 7.939536283677166e-06, + "loss": 0.9748, + "step": 47020 + }, + { + "epoch": 3.47, + "learning_rate": 7.939145972592459e-06, + "loss": 0.9933, + "step": 47021 + }, + { + "epoch": 3.47, + "learning_rate": 7.938755664786544e-06, + "loss": 1.0776, + "step": 47022 + }, + { + "epoch": 3.47, + "learning_rate": 7.938365360260036e-06, + "loss": 0.9829, + "step": 47023 + }, + { + "epoch": 3.47, + "learning_rate": 7.937975059013562e-06, + "loss": 0.9694, + "step": 47024 + }, + { + "epoch": 3.47, + "learning_rate": 7.937584761047742e-06, + "loss": 1.0143, + "step": 47025 + }, + { + "epoch": 3.47, + "learning_rate": 7.937194466363189e-06, + "loss": 0.9417, + "step": 47026 + }, + { + "epoch": 3.47, + "learning_rate": 7.936804174960536e-06, + "loss": 0.9178, + "step": 47027 + }, + { + "epoch": 3.47, + "learning_rate": 7.936413886840392e-06, + "loss": 0.995, + "step": 47028 + }, + { + "epoch": 3.47, + "learning_rate": 7.936023602003391e-06, + "loss": 0.9983, + "step": 47029 + }, + { + "epoch": 3.48, + "learning_rate": 7.935633320450141e-06, + "loss": 0.9164, + "step": 47030 + }, + { + "epoch": 3.48, + "learning_rate": 7.935243042181273e-06, + "loss": 0.9419, + "step": 47031 + }, + { + "epoch": 3.48, + "learning_rate": 7.934852767197398e-06, + "loss": 1.0472, + "step": 47032 + }, + { + "epoch": 3.48, + "learning_rate": 7.93446249549915e-06, + "loss": 1.0731, + "step": 47033 + }, + { + "epoch": 3.48, + "learning_rate": 7.93407222708714e-06, + "loss": 0.9611, + "step": 47034 + }, + { + "epoch": 3.48, + "learning_rate": 7.93368196196199e-06, + "loss": 0.9458, + "step": 47035 + }, + { + "epoch": 3.48, + "learning_rate": 7.933291700124323e-06, + "loss": 1.064, + "step": 47036 + }, + { + "epoch": 3.48, + "learning_rate": 7.932901441574758e-06, + "loss": 1.0417, + "step": 47037 + }, + { + "epoch": 3.48, + "learning_rate": 7.93251118631392e-06, + "loss": 0.9602, + "step": 47038 + }, + { + "epoch": 3.48, + "learning_rate": 7.932120934342425e-06, + "loss": 1.0002, + "step": 47039 + }, + { + "epoch": 3.48, + "learning_rate": 7.931730685660898e-06, + "loss": 1.0091, + "step": 47040 + }, + { + "epoch": 3.48, + "learning_rate": 7.931340440269956e-06, + "loss": 0.9897, + "step": 47041 + }, + { + "epoch": 3.48, + "learning_rate": 7.930950198170223e-06, + "loss": 0.8115, + "step": 47042 + }, + { + "epoch": 3.48, + "learning_rate": 7.93055995936232e-06, + "loss": 1.0222, + "step": 47043 + }, + { + "epoch": 3.48, + "learning_rate": 7.930169723846863e-06, + "loss": 0.9759, + "step": 47044 + }, + { + "epoch": 3.48, + "learning_rate": 7.92977949162448e-06, + "loss": 1.1442, + "step": 47045 + }, + { + "epoch": 3.48, + "learning_rate": 7.929389262695783e-06, + "loss": 0.9797, + "step": 47046 + }, + { + "epoch": 3.48, + "learning_rate": 7.928999037061405e-06, + "loss": 1.0946, + "step": 47047 + }, + { + "epoch": 3.48, + "learning_rate": 7.928608814721953e-06, + "loss": 0.8648, + "step": 47048 + }, + { + "epoch": 3.48, + "learning_rate": 7.92821859567806e-06, + "loss": 0.8383, + "step": 47049 + }, + { + "epoch": 3.48, + "learning_rate": 7.927828379930337e-06, + "loss": 1.0276, + "step": 47050 + }, + { + "epoch": 3.48, + "learning_rate": 7.927438167479415e-06, + "loss": 1.0211, + "step": 47051 + }, + { + "epoch": 3.48, + "learning_rate": 7.927047958325908e-06, + "loss": 0.9, + "step": 47052 + }, + { + "epoch": 3.48, + "learning_rate": 7.926657752470435e-06, + "loss": 1.0044, + "step": 47053 + }, + { + "epoch": 3.48, + "learning_rate": 7.92626754991362e-06, + "loss": 1.0778, + "step": 47054 + }, + { + "epoch": 3.48, + "learning_rate": 7.925877350656085e-06, + "loss": 1.0121, + "step": 47055 + }, + { + "epoch": 3.48, + "learning_rate": 7.92548715469845e-06, + "loss": 0.996, + "step": 47056 + }, + { + "epoch": 3.48, + "learning_rate": 7.925096962041333e-06, + "loss": 0.9886, + "step": 47057 + }, + { + "epoch": 3.48, + "learning_rate": 7.924706772685359e-06, + "loss": 1.0423, + "step": 47058 + }, + { + "epoch": 3.48, + "learning_rate": 7.924316586631145e-06, + "loss": 1.0244, + "step": 47059 + }, + { + "epoch": 3.48, + "learning_rate": 7.923926403879316e-06, + "loss": 1.0388, + "step": 47060 + }, + { + "epoch": 3.48, + "learning_rate": 7.92353622443049e-06, + "loss": 0.9682, + "step": 47061 + }, + { + "epoch": 3.48, + "learning_rate": 7.923146048285286e-06, + "loss": 0.9326, + "step": 47062 + }, + { + "epoch": 3.48, + "learning_rate": 7.92275587544433e-06, + "loss": 1.0269, + "step": 47063 + }, + { + "epoch": 3.48, + "learning_rate": 7.922365705908237e-06, + "loss": 1.0389, + "step": 47064 + }, + { + "epoch": 3.48, + "learning_rate": 7.921975539677631e-06, + "loss": 0.9542, + "step": 47065 + }, + { + "epoch": 3.48, + "learning_rate": 7.921585376753133e-06, + "loss": 1.0173, + "step": 47066 + }, + { + "epoch": 3.48, + "learning_rate": 7.921195217135362e-06, + "loss": 0.9865, + "step": 47067 + }, + { + "epoch": 3.48, + "learning_rate": 7.92080506082494e-06, + "loss": 1.0445, + "step": 47068 + }, + { + "epoch": 3.48, + "learning_rate": 7.920414907822489e-06, + "loss": 1.1115, + "step": 47069 + }, + { + "epoch": 3.48, + "learning_rate": 7.920024758128624e-06, + "loss": 0.9725, + "step": 47070 + }, + { + "epoch": 3.48, + "learning_rate": 7.919634611743975e-06, + "loss": 1.0005, + "step": 47071 + }, + { + "epoch": 3.48, + "learning_rate": 7.919244468669158e-06, + "loss": 1.0535, + "step": 47072 + }, + { + "epoch": 3.48, + "learning_rate": 7.918854328904786e-06, + "loss": 1.0603, + "step": 47073 + }, + { + "epoch": 3.48, + "learning_rate": 7.918464192451497e-06, + "loss": 1.0442, + "step": 47074 + }, + { + "epoch": 3.48, + "learning_rate": 7.918074059309891e-06, + "loss": 1.0369, + "step": 47075 + }, + { + "epoch": 3.48, + "learning_rate": 7.917683929480609e-06, + "loss": 0.9877, + "step": 47076 + }, + { + "epoch": 3.48, + "learning_rate": 7.917293802964253e-06, + "loss": 1.0222, + "step": 47077 + }, + { + "epoch": 3.48, + "learning_rate": 7.91690367976146e-06, + "loss": 1.0186, + "step": 47078 + }, + { + "epoch": 3.48, + "learning_rate": 7.916513559872839e-06, + "loss": 1.0449, + "step": 47079 + }, + { + "epoch": 3.48, + "learning_rate": 7.916123443299018e-06, + "loss": 0.9399, + "step": 47080 + }, + { + "epoch": 3.48, + "learning_rate": 7.915733330040616e-06, + "loss": 0.9916, + "step": 47081 + }, + { + "epoch": 3.48, + "learning_rate": 7.915343220098248e-06, + "loss": 1.1806, + "step": 47082 + }, + { + "epoch": 3.48, + "learning_rate": 7.914953113472543e-06, + "loss": 0.9609, + "step": 47083 + }, + { + "epoch": 3.48, + "learning_rate": 7.914563010164115e-06, + "loss": 0.862, + "step": 47084 + }, + { + "epoch": 3.48, + "learning_rate": 7.914172910173588e-06, + "loss": 1.0672, + "step": 47085 + }, + { + "epoch": 3.48, + "learning_rate": 7.913782813501582e-06, + "loss": 0.8902, + "step": 47086 + }, + { + "epoch": 3.48, + "learning_rate": 7.913392720148719e-06, + "loss": 1.0186, + "step": 47087 + }, + { + "epoch": 3.48, + "learning_rate": 7.913002630115616e-06, + "loss": 1.1293, + "step": 47088 + }, + { + "epoch": 3.48, + "learning_rate": 7.912612543402899e-06, + "loss": 0.9702, + "step": 47089 + }, + { + "epoch": 3.48, + "learning_rate": 7.912222460011185e-06, + "loss": 0.999, + "step": 47090 + }, + { + "epoch": 3.48, + "learning_rate": 7.911832379941093e-06, + "loss": 1.048, + "step": 47091 + }, + { + "epoch": 3.48, + "learning_rate": 7.911442303193251e-06, + "loss": 0.9734, + "step": 47092 + }, + { + "epoch": 3.48, + "learning_rate": 7.911052229768266e-06, + "loss": 0.9574, + "step": 47093 + }, + { + "epoch": 3.48, + "learning_rate": 7.910662159666776e-06, + "loss": 1.054, + "step": 47094 + }, + { + "epoch": 3.48, + "learning_rate": 7.910272092889383e-06, + "loss": 0.925, + "step": 47095 + }, + { + "epoch": 3.48, + "learning_rate": 7.909882029436727e-06, + "loss": 1.0022, + "step": 47096 + }, + { + "epoch": 3.48, + "learning_rate": 7.90949196930941e-06, + "loss": 0.9751, + "step": 47097 + }, + { + "epoch": 3.48, + "learning_rate": 7.909101912508069e-06, + "loss": 1.0462, + "step": 47098 + }, + { + "epoch": 3.48, + "learning_rate": 7.908711859033313e-06, + "loss": 0.9686, + "step": 47099 + }, + { + "epoch": 3.48, + "learning_rate": 7.908321808885766e-06, + "loss": 1.0851, + "step": 47100 + }, + { + "epoch": 3.48, + "learning_rate": 7.907931762066052e-06, + "loss": 1.0065, + "step": 47101 + }, + { + "epoch": 3.48, + "learning_rate": 7.907541718574786e-06, + "loss": 1.0361, + "step": 47102 + }, + { + "epoch": 3.48, + "learning_rate": 7.907151678412591e-06, + "loss": 1.0542, + "step": 47103 + }, + { + "epoch": 3.48, + "learning_rate": 7.906761641580087e-06, + "loss": 1.0538, + "step": 47104 + }, + { + "epoch": 3.48, + "learning_rate": 7.906371608077899e-06, + "loss": 1.0656, + "step": 47105 + }, + { + "epoch": 3.48, + "learning_rate": 7.90598157790664e-06, + "loss": 1.0284, + "step": 47106 + }, + { + "epoch": 3.48, + "learning_rate": 7.905591551066936e-06, + "loss": 0.9209, + "step": 47107 + }, + { + "epoch": 3.48, + "learning_rate": 7.905201527559407e-06, + "loss": 0.9775, + "step": 47108 + }, + { + "epoch": 3.48, + "learning_rate": 7.904811507384671e-06, + "loss": 0.9811, + "step": 47109 + }, + { + "epoch": 3.48, + "learning_rate": 7.90442149054335e-06, + "loss": 1.0064, + "step": 47110 + }, + { + "epoch": 3.48, + "learning_rate": 7.904031477036063e-06, + "loss": 1.0085, + "step": 47111 + }, + { + "epoch": 3.48, + "learning_rate": 7.903641466863436e-06, + "loss": 0.9763, + "step": 47112 + }, + { + "epoch": 3.48, + "learning_rate": 7.903251460026081e-06, + "loss": 0.9855, + "step": 47113 + }, + { + "epoch": 3.48, + "learning_rate": 7.902861456524626e-06, + "loss": 1.0601, + "step": 47114 + }, + { + "epoch": 3.48, + "learning_rate": 7.902471456359687e-06, + "loss": 1.0144, + "step": 47115 + }, + { + "epoch": 3.48, + "learning_rate": 7.902081459531886e-06, + "loss": 1.0226, + "step": 47116 + }, + { + "epoch": 3.48, + "learning_rate": 7.901691466041844e-06, + "loss": 0.9552, + "step": 47117 + }, + { + "epoch": 3.48, + "learning_rate": 7.90130147589018e-06, + "loss": 1.0299, + "step": 47118 + }, + { + "epoch": 3.48, + "learning_rate": 7.90091148907752e-06, + "loss": 0.9629, + "step": 47119 + }, + { + "epoch": 3.48, + "learning_rate": 7.90052150560447e-06, + "loss": 0.9799, + "step": 47120 + }, + { + "epoch": 3.48, + "learning_rate": 7.90013152547167e-06, + "loss": 1.0873, + "step": 47121 + }, + { + "epoch": 3.48, + "learning_rate": 7.899741548679722e-06, + "loss": 0.9301, + "step": 47122 + }, + { + "epoch": 3.48, + "learning_rate": 7.899351575229263e-06, + "loss": 1.009, + "step": 47123 + }, + { + "epoch": 3.48, + "learning_rate": 7.8989616051209e-06, + "loss": 0.9626, + "step": 47124 + }, + { + "epoch": 3.48, + "learning_rate": 7.898571638355264e-06, + "loss": 1.1262, + "step": 47125 + }, + { + "epoch": 3.48, + "learning_rate": 7.898181674932967e-06, + "loss": 1.0433, + "step": 47126 + }, + { + "epoch": 3.48, + "learning_rate": 7.897791714854634e-06, + "loss": 0.9361, + "step": 47127 + }, + { + "epoch": 3.48, + "learning_rate": 7.897401758120886e-06, + "loss": 1.0158, + "step": 47128 + }, + { + "epoch": 3.48, + "learning_rate": 7.897011804732338e-06, + "loss": 1.0084, + "step": 47129 + }, + { + "epoch": 3.48, + "learning_rate": 7.896621854689616e-06, + "loss": 0.8949, + "step": 47130 + }, + { + "epoch": 3.48, + "learning_rate": 7.896231907993336e-06, + "loss": 0.9762, + "step": 47131 + }, + { + "epoch": 3.48, + "learning_rate": 7.895841964644125e-06, + "loss": 1.0036, + "step": 47132 + }, + { + "epoch": 3.48, + "learning_rate": 7.895452024642595e-06, + "loss": 0.9387, + "step": 47133 + }, + { + "epoch": 3.48, + "learning_rate": 7.895062087989375e-06, + "loss": 1.0045, + "step": 47134 + }, + { + "epoch": 3.48, + "learning_rate": 7.89467215468508e-06, + "loss": 0.959, + "step": 47135 + }, + { + "epoch": 3.48, + "learning_rate": 7.89428222473033e-06, + "loss": 1.0047, + "step": 47136 + }, + { + "epoch": 3.48, + "learning_rate": 7.893892298125752e-06, + "loss": 1.0165, + "step": 47137 + }, + { + "epoch": 3.48, + "learning_rate": 7.893502374871953e-06, + "loss": 1.1637, + "step": 47138 + }, + { + "epoch": 3.48, + "learning_rate": 7.893112454969568e-06, + "loss": 0.977, + "step": 47139 + }, + { + "epoch": 3.48, + "learning_rate": 7.892722538419207e-06, + "loss": 0.9205, + "step": 47140 + }, + { + "epoch": 3.48, + "learning_rate": 7.892332625221496e-06, + "loss": 1.1415, + "step": 47141 + }, + { + "epoch": 3.48, + "learning_rate": 7.89194271537705e-06, + "loss": 0.9957, + "step": 47142 + }, + { + "epoch": 3.48, + "learning_rate": 7.891552808886501e-06, + "loss": 0.9253, + "step": 47143 + }, + { + "epoch": 3.48, + "learning_rate": 7.891162905750453e-06, + "loss": 0.95, + "step": 47144 + }, + { + "epoch": 3.48, + "learning_rate": 7.890773005969541e-06, + "loss": 1.0561, + "step": 47145 + }, + { + "epoch": 3.48, + "learning_rate": 7.890383109544377e-06, + "loss": 1.0315, + "step": 47146 + }, + { + "epoch": 3.48, + "learning_rate": 7.889993216475582e-06, + "loss": 0.9428, + "step": 47147 + }, + { + "epoch": 3.48, + "learning_rate": 7.88960332676378e-06, + "loss": 1.0874, + "step": 47148 + }, + { + "epoch": 3.48, + "learning_rate": 7.889213440409585e-06, + "loss": 1.0776, + "step": 47149 + }, + { + "epoch": 3.48, + "learning_rate": 7.888823557413624e-06, + "loss": 0.9848, + "step": 47150 + }, + { + "epoch": 3.48, + "learning_rate": 7.888433677776512e-06, + "loss": 1.0147, + "step": 47151 + }, + { + "epoch": 3.48, + "learning_rate": 7.888043801498876e-06, + "loss": 0.9919, + "step": 47152 + }, + { + "epoch": 3.48, + "learning_rate": 7.887653928581328e-06, + "loss": 0.9581, + "step": 47153 + }, + { + "epoch": 3.48, + "learning_rate": 7.887264059024495e-06, + "loss": 1.0138, + "step": 47154 + }, + { + "epoch": 3.48, + "learning_rate": 7.886874192828995e-06, + "loss": 1.0093, + "step": 47155 + }, + { + "epoch": 3.48, + "learning_rate": 7.886484329995445e-06, + "loss": 1.0093, + "step": 47156 + }, + { + "epoch": 3.48, + "learning_rate": 7.886094470524472e-06, + "loss": 0.9525, + "step": 47157 + }, + { + "epoch": 3.48, + "learning_rate": 7.885704614416687e-06, + "loss": 0.9596, + "step": 47158 + }, + { + "epoch": 3.48, + "learning_rate": 7.885314761672721e-06, + "loss": 1.0484, + "step": 47159 + }, + { + "epoch": 3.48, + "learning_rate": 7.884924912293185e-06, + "loss": 1.0487, + "step": 47160 + }, + { + "epoch": 3.48, + "learning_rate": 7.884535066278706e-06, + "loss": 0.9289, + "step": 47161 + }, + { + "epoch": 3.48, + "learning_rate": 7.8841452236299e-06, + "loss": 1.0695, + "step": 47162 + }, + { + "epoch": 3.48, + "learning_rate": 7.88375538434739e-06, + "loss": 1.0248, + "step": 47163 + }, + { + "epoch": 3.48, + "learning_rate": 7.883365548431798e-06, + "loss": 0.9721, + "step": 47164 + }, + { + "epoch": 3.48, + "learning_rate": 7.882975715883735e-06, + "loss": 0.9245, + "step": 47165 + }, + { + "epoch": 3.49, + "learning_rate": 7.882585886703832e-06, + "loss": 1.0856, + "step": 47166 + }, + { + "epoch": 3.49, + "learning_rate": 7.882196060892698e-06, + "loss": 0.9349, + "step": 47167 + }, + { + "epoch": 3.49, + "learning_rate": 7.881806238450967e-06, + "loss": 1.0436, + "step": 47168 + }, + { + "epoch": 3.49, + "learning_rate": 7.881416419379249e-06, + "loss": 0.9336, + "step": 47169 + }, + { + "epoch": 3.49, + "learning_rate": 7.881026603678168e-06, + "loss": 1.0852, + "step": 47170 + }, + { + "epoch": 3.49, + "learning_rate": 7.880636791348341e-06, + "loss": 0.9493, + "step": 47171 + }, + { + "epoch": 3.49, + "learning_rate": 7.880246982390393e-06, + "loss": 0.9628, + "step": 47172 + }, + { + "epoch": 3.49, + "learning_rate": 7.879857176804939e-06, + "loss": 1.0473, + "step": 47173 + }, + { + "epoch": 3.49, + "learning_rate": 7.879467374592604e-06, + "loss": 0.9728, + "step": 47174 + }, + { + "epoch": 3.49, + "learning_rate": 7.879077575754008e-06, + "loss": 0.9192, + "step": 47175 + }, + { + "epoch": 3.49, + "learning_rate": 7.878687780289766e-06, + "loss": 1.0448, + "step": 47176 + }, + { + "epoch": 3.49, + "learning_rate": 7.878297988200502e-06, + "loss": 0.8948, + "step": 47177 + }, + { + "epoch": 3.49, + "learning_rate": 7.877908199486835e-06, + "loss": 1.0712, + "step": 47178 + }, + { + "epoch": 3.49, + "learning_rate": 7.877518414149388e-06, + "loss": 0.9635, + "step": 47179 + }, + { + "epoch": 3.49, + "learning_rate": 7.877128632188776e-06, + "loss": 1.0443, + "step": 47180 + }, + { + "epoch": 3.49, + "learning_rate": 7.876738853605624e-06, + "loss": 0.9669, + "step": 47181 + }, + { + "epoch": 3.49, + "learning_rate": 7.87634907840055e-06, + "loss": 0.9212, + "step": 47182 + }, + { + "epoch": 3.49, + "learning_rate": 7.875959306574173e-06, + "loss": 1.0048, + "step": 47183 + }, + { + "epoch": 3.49, + "learning_rate": 7.875569538127117e-06, + "loss": 1.0546, + "step": 47184 + }, + { + "epoch": 3.49, + "learning_rate": 7.875179773059996e-06, + "loss": 1.0467, + "step": 47185 + }, + { + "epoch": 3.49, + "learning_rate": 7.874790011373437e-06, + "loss": 1.0412, + "step": 47186 + }, + { + "epoch": 3.49, + "learning_rate": 7.874400253068053e-06, + "loss": 1.0463, + "step": 47187 + }, + { + "epoch": 3.49, + "learning_rate": 7.874010498144473e-06, + "loss": 0.9729, + "step": 47188 + }, + { + "epoch": 3.49, + "learning_rate": 7.873620746603305e-06, + "loss": 1.0056, + "step": 47189 + }, + { + "epoch": 3.49, + "learning_rate": 7.873230998445182e-06, + "loss": 1.0375, + "step": 47190 + }, + { + "epoch": 3.49, + "learning_rate": 7.872841253670712e-06, + "loss": 0.9759, + "step": 47191 + }, + { + "epoch": 3.49, + "learning_rate": 7.872451512280528e-06, + "loss": 0.9125, + "step": 47192 + }, + { + "epoch": 3.49, + "learning_rate": 7.87206177427524e-06, + "loss": 1.043, + "step": 47193 + }, + { + "epoch": 3.49, + "learning_rate": 7.87167203965547e-06, + "loss": 1.1026, + "step": 47194 + }, + { + "epoch": 3.49, + "learning_rate": 7.871282308421841e-06, + "loss": 1.033, + "step": 47195 + }, + { + "epoch": 3.49, + "learning_rate": 7.87089258057497e-06, + "loss": 0.9193, + "step": 47196 + }, + { + "epoch": 3.49, + "learning_rate": 7.87050285611548e-06, + "loss": 1.0016, + "step": 47197 + }, + { + "epoch": 3.49, + "learning_rate": 7.870113135043987e-06, + "loss": 0.9763, + "step": 47198 + }, + { + "epoch": 3.49, + "learning_rate": 7.869723417361117e-06, + "loss": 1.1062, + "step": 47199 + }, + { + "epoch": 3.49, + "learning_rate": 7.869333703067483e-06, + "loss": 0.8929, + "step": 47200 + }, + { + "epoch": 3.49, + "learning_rate": 7.868943992163712e-06, + "loss": 1.0655, + "step": 47201 + }, + { + "epoch": 3.49, + "learning_rate": 7.86855428465042e-06, + "loss": 1.0536, + "step": 47202 + }, + { + "epoch": 3.49, + "learning_rate": 7.868164580528225e-06, + "loss": 0.9876, + "step": 47203 + }, + { + "epoch": 3.49, + "learning_rate": 7.867774879797752e-06, + "loss": 1.0477, + "step": 47204 + }, + { + "epoch": 3.49, + "learning_rate": 7.867385182459617e-06, + "loss": 0.9375, + "step": 47205 + }, + { + "epoch": 3.49, + "learning_rate": 7.866995488514444e-06, + "loss": 1.0436, + "step": 47206 + }, + { + "epoch": 3.49, + "learning_rate": 7.866605797962849e-06, + "loss": 0.8514, + "step": 47207 + }, + { + "epoch": 3.49, + "learning_rate": 7.866216110805455e-06, + "loss": 1.0139, + "step": 47208 + }, + { + "epoch": 3.49, + "learning_rate": 7.865826427042878e-06, + "loss": 0.9546, + "step": 47209 + }, + { + "epoch": 3.49, + "learning_rate": 7.865436746675743e-06, + "loss": 1.1083, + "step": 47210 + }, + { + "epoch": 3.49, + "learning_rate": 7.86504706970467e-06, + "loss": 1.0556, + "step": 47211 + }, + { + "epoch": 3.49, + "learning_rate": 7.864657396130269e-06, + "loss": 1.0621, + "step": 47212 + }, + { + "epoch": 3.49, + "learning_rate": 7.864267725953177e-06, + "loss": 1.0406, + "step": 47213 + }, + { + "epoch": 3.49, + "learning_rate": 7.863878059173997e-06, + "loss": 1.0302, + "step": 47214 + }, + { + "epoch": 3.49, + "learning_rate": 7.863488395793359e-06, + "loss": 1.141, + "step": 47215 + }, + { + "epoch": 3.49, + "learning_rate": 7.86309873581188e-06, + "loss": 1.0051, + "step": 47216 + }, + { + "epoch": 3.49, + "learning_rate": 7.862709079230183e-06, + "loss": 0.9819, + "step": 47217 + }, + { + "epoch": 3.49, + "learning_rate": 7.86231942604888e-06, + "loss": 1.05, + "step": 47218 + }, + { + "epoch": 3.49, + "learning_rate": 7.8619297762686e-06, + "loss": 1.0424, + "step": 47219 + }, + { + "epoch": 3.49, + "learning_rate": 7.86154012988996e-06, + "loss": 0.9349, + "step": 47220 + }, + { + "epoch": 3.49, + "learning_rate": 7.861150486913576e-06, + "loss": 1.0346, + "step": 47221 + }, + { + "epoch": 3.49, + "learning_rate": 7.860760847340073e-06, + "loss": 1.0075, + "step": 47222 + }, + { + "epoch": 3.49, + "learning_rate": 7.860371211170066e-06, + "loss": 1.0171, + "step": 47223 + }, + { + "epoch": 3.49, + "learning_rate": 7.85998157840418e-06, + "loss": 1.0843, + "step": 47224 + }, + { + "epoch": 3.49, + "learning_rate": 7.859591949043033e-06, + "loss": 1.0321, + "step": 47225 + }, + { + "epoch": 3.49, + "learning_rate": 7.859202323087245e-06, + "loss": 1.0729, + "step": 47226 + }, + { + "epoch": 3.49, + "learning_rate": 7.858812700537433e-06, + "loss": 0.912, + "step": 47227 + }, + { + "epoch": 3.49, + "learning_rate": 7.858423081394222e-06, + "loss": 1.0036, + "step": 47228 + }, + { + "epoch": 3.49, + "learning_rate": 7.858033465658228e-06, + "loss": 1.0193, + "step": 47229 + }, + { + "epoch": 3.49, + "learning_rate": 7.857643853330073e-06, + "loss": 1.0286, + "step": 47230 + }, + { + "epoch": 3.49, + "learning_rate": 7.857254244410379e-06, + "loss": 1.0012, + "step": 47231 + }, + { + "epoch": 3.49, + "learning_rate": 7.856864638899755e-06, + "loss": 0.9326, + "step": 47232 + }, + { + "epoch": 3.49, + "learning_rate": 7.856475036798837e-06, + "loss": 1.0701, + "step": 47233 + }, + { + "epoch": 3.49, + "learning_rate": 7.856085438108227e-06, + "loss": 1.0248, + "step": 47234 + }, + { + "epoch": 3.49, + "learning_rate": 7.855695842828563e-06, + "loss": 1.0458, + "step": 47235 + }, + { + "epoch": 3.49, + "learning_rate": 7.85530625096045e-06, + "loss": 1.009, + "step": 47236 + }, + { + "epoch": 3.49, + "learning_rate": 7.854916662504521e-06, + "loss": 0.9554, + "step": 47237 + }, + { + "epoch": 3.49, + "learning_rate": 7.854527077461381e-06, + "loss": 1.0675, + "step": 47238 + }, + { + "epoch": 3.49, + "learning_rate": 7.854137495831665e-06, + "loss": 1.0321, + "step": 47239 + }, + { + "epoch": 3.49, + "learning_rate": 7.853747917615983e-06, + "loss": 1.0097, + "step": 47240 + }, + { + "epoch": 3.49, + "learning_rate": 7.853358342814956e-06, + "loss": 0.9655, + "step": 47241 + }, + { + "epoch": 3.49, + "learning_rate": 7.852968771429207e-06, + "loss": 0.8923, + "step": 47242 + }, + { + "epoch": 3.49, + "learning_rate": 7.852579203459351e-06, + "loss": 0.9362, + "step": 47243 + }, + { + "epoch": 3.49, + "learning_rate": 7.852189638906014e-06, + "loss": 0.9729, + "step": 47244 + }, + { + "epoch": 3.49, + "learning_rate": 7.85180007776981e-06, + "loss": 0.9824, + "step": 47245 + }, + { + "epoch": 3.49, + "learning_rate": 7.851410520051363e-06, + "loss": 1.057, + "step": 47246 + }, + { + "epoch": 3.49, + "learning_rate": 7.851020965751289e-06, + "loss": 0.9956, + "step": 47247 + }, + { + "epoch": 3.49, + "learning_rate": 7.850631414870214e-06, + "loss": 0.967, + "step": 47248 + }, + { + "epoch": 3.49, + "learning_rate": 7.85024186740875e-06, + "loss": 0.8959, + "step": 47249 + }, + { + "epoch": 3.49, + "learning_rate": 7.849852323367522e-06, + "loss": 1.0244, + "step": 47250 + }, + { + "epoch": 3.49, + "learning_rate": 7.849462782747147e-06, + "loss": 1.0687, + "step": 47251 + }, + { + "epoch": 3.49, + "learning_rate": 7.849073245548247e-06, + "loss": 0.9633, + "step": 47252 + }, + { + "epoch": 3.49, + "learning_rate": 7.84868371177144e-06, + "loss": 1.0895, + "step": 47253 + }, + { + "epoch": 3.49, + "learning_rate": 7.848294181417347e-06, + "loss": 1.0553, + "step": 47254 + }, + { + "epoch": 3.49, + "learning_rate": 7.847904654486587e-06, + "loss": 1.0141, + "step": 47255 + }, + { + "epoch": 3.49, + "learning_rate": 7.84751513097978e-06, + "loss": 1.0347, + "step": 47256 + }, + { + "epoch": 3.49, + "learning_rate": 7.847125610897546e-06, + "loss": 0.8449, + "step": 47257 + }, + { + "epoch": 3.49, + "learning_rate": 7.846736094240505e-06, + "loss": 0.923, + "step": 47258 + }, + { + "epoch": 3.49, + "learning_rate": 7.846346581009276e-06, + "loss": 0.9506, + "step": 47259 + }, + { + "epoch": 3.49, + "learning_rate": 7.845957071204478e-06, + "loss": 0.9731, + "step": 47260 + }, + { + "epoch": 3.49, + "learning_rate": 7.845567564826729e-06, + "loss": 1.0478, + "step": 47261 + }, + { + "epoch": 3.49, + "learning_rate": 7.845178061876655e-06, + "loss": 1.1049, + "step": 47262 + }, + { + "epoch": 3.49, + "learning_rate": 7.84478856235487e-06, + "loss": 0.9369, + "step": 47263 + }, + { + "epoch": 3.49, + "learning_rate": 7.844399066261996e-06, + "loss": 0.9568, + "step": 47264 + }, + { + "epoch": 3.49, + "learning_rate": 7.844009573598651e-06, + "loss": 1.0146, + "step": 47265 + }, + { + "epoch": 3.49, + "learning_rate": 7.843620084365458e-06, + "loss": 0.9205, + "step": 47266 + }, + { + "epoch": 3.49, + "learning_rate": 7.843230598563034e-06, + "loss": 0.9707, + "step": 47267 + }, + { + "epoch": 3.49, + "learning_rate": 7.842841116191996e-06, + "loss": 1.0184, + "step": 47268 + }, + { + "epoch": 3.49, + "learning_rate": 7.842451637252972e-06, + "loss": 1.0817, + "step": 47269 + }, + { + "epoch": 3.49, + "learning_rate": 7.842062161746572e-06, + "loss": 0.9603, + "step": 47270 + }, + { + "epoch": 3.49, + "learning_rate": 7.841672689673424e-06, + "loss": 1.0307, + "step": 47271 + }, + { + "epoch": 3.49, + "learning_rate": 7.841283221034141e-06, + "loss": 0.9951, + "step": 47272 + }, + { + "epoch": 3.49, + "learning_rate": 7.840893755829348e-06, + "loss": 1.0273, + "step": 47273 + }, + { + "epoch": 3.49, + "learning_rate": 7.840504294059661e-06, + "loss": 0.9749, + "step": 47274 + }, + { + "epoch": 3.49, + "learning_rate": 7.840114835725703e-06, + "loss": 1.0167, + "step": 47275 + }, + { + "epoch": 3.49, + "learning_rate": 7.839725380828092e-06, + "loss": 0.9489, + "step": 47276 + }, + { + "epoch": 3.49, + "learning_rate": 7.83933592936744e-06, + "loss": 0.9823, + "step": 47277 + }, + { + "epoch": 3.49, + "learning_rate": 7.83894648134438e-06, + "loss": 0.9731, + "step": 47278 + }, + { + "epoch": 3.49, + "learning_rate": 7.83855703675952e-06, + "loss": 1.0206, + "step": 47279 + }, + { + "epoch": 3.49, + "learning_rate": 7.838167595613492e-06, + "loss": 1.0006, + "step": 47280 + }, + { + "epoch": 3.49, + "learning_rate": 7.8377781579069e-06, + "loss": 1.0082, + "step": 47281 + }, + { + "epoch": 3.49, + "learning_rate": 7.83738872364038e-06, + "loss": 1.0013, + "step": 47282 + }, + { + "epoch": 3.49, + "learning_rate": 7.836999292814537e-06, + "loss": 0.9551, + "step": 47283 + }, + { + "epoch": 3.49, + "learning_rate": 7.836609865430003e-06, + "loss": 0.9803, + "step": 47284 + }, + { + "epoch": 3.49, + "learning_rate": 7.836220441487385e-06, + "loss": 0.9839, + "step": 47285 + }, + { + "epoch": 3.49, + "learning_rate": 7.835831020987317e-06, + "loss": 0.891, + "step": 47286 + }, + { + "epoch": 3.49, + "learning_rate": 7.835441603930407e-06, + "loss": 0.9615, + "step": 47287 + }, + { + "epoch": 3.49, + "learning_rate": 7.835052190317278e-06, + "loss": 0.9964, + "step": 47288 + }, + { + "epoch": 3.49, + "learning_rate": 7.834662780148552e-06, + "loss": 0.9431, + "step": 47289 + }, + { + "epoch": 3.49, + "learning_rate": 7.834273373424842e-06, + "loss": 1.048, + "step": 47290 + }, + { + "epoch": 3.49, + "learning_rate": 7.833883970146777e-06, + "loss": 0.9475, + "step": 47291 + }, + { + "epoch": 3.49, + "learning_rate": 7.833494570314968e-06, + "loss": 1.0657, + "step": 47292 + }, + { + "epoch": 3.49, + "learning_rate": 7.833105173930042e-06, + "loss": 0.8714, + "step": 47293 + }, + { + "epoch": 3.49, + "learning_rate": 7.832715780992612e-06, + "loss": 1.0392, + "step": 47294 + }, + { + "epoch": 3.49, + "learning_rate": 7.832326391503301e-06, + "loss": 0.9896, + "step": 47295 + }, + { + "epoch": 3.49, + "learning_rate": 7.831937005462729e-06, + "loss": 1.1301, + "step": 47296 + }, + { + "epoch": 3.49, + "learning_rate": 7.83154762287151e-06, + "loss": 0.9898, + "step": 47297 + }, + { + "epoch": 3.49, + "learning_rate": 7.831158243730272e-06, + "loss": 0.9905, + "step": 47298 + }, + { + "epoch": 3.49, + "learning_rate": 7.830768868039627e-06, + "loss": 1.0026, + "step": 47299 + }, + { + "epoch": 3.49, + "learning_rate": 7.8303794958002e-06, + "loss": 0.9741, + "step": 47300 + }, + { + "epoch": 3.5, + "learning_rate": 7.829990127012606e-06, + "loss": 0.9478, + "step": 47301 + }, + { + "epoch": 3.5, + "learning_rate": 7.829600761677471e-06, + "loss": 0.943, + "step": 47302 + }, + { + "epoch": 3.5, + "learning_rate": 7.829211399795404e-06, + "loss": 0.9526, + "step": 47303 + }, + { + "epoch": 3.5, + "learning_rate": 7.828822041367036e-06, + "loss": 1.0424, + "step": 47304 + }, + { + "epoch": 3.5, + "learning_rate": 7.828432686392981e-06, + "loss": 1.0317, + "step": 47305 + }, + { + "epoch": 3.5, + "learning_rate": 7.828043334873855e-06, + "loss": 0.9888, + "step": 47306 + }, + { + "epoch": 3.5, + "learning_rate": 7.827653986810285e-06, + "loss": 0.9063, + "step": 47307 + }, + { + "epoch": 3.5, + "learning_rate": 7.827264642202881e-06, + "loss": 0.9921, + "step": 47308 + }, + { + "epoch": 3.5, + "learning_rate": 7.826875301052272e-06, + "loss": 1.0428, + "step": 47309 + }, + { + "epoch": 3.5, + "learning_rate": 7.826485963359071e-06, + "loss": 1.1138, + "step": 47310 + }, + { + "epoch": 3.5, + "learning_rate": 7.826096629123901e-06, + "loss": 0.989, + "step": 47311 + }, + { + "epoch": 3.5, + "learning_rate": 7.825707298347379e-06, + "loss": 1.0618, + "step": 47312 + }, + { + "epoch": 3.5, + "learning_rate": 7.825317971030127e-06, + "loss": 0.9452, + "step": 47313 + }, + { + "epoch": 3.5, + "learning_rate": 7.824928647172764e-06, + "loss": 1.0169, + "step": 47314 + }, + { + "epoch": 3.5, + "learning_rate": 7.824539326775905e-06, + "loss": 1.0758, + "step": 47315 + }, + { + "epoch": 3.5, + "learning_rate": 7.824150009840175e-06, + "loss": 1.0498, + "step": 47316 + }, + { + "epoch": 3.5, + "learning_rate": 7.823760696366191e-06, + "loss": 1.0362, + "step": 47317 + }, + { + "epoch": 3.5, + "learning_rate": 7.823371386354573e-06, + "loss": 0.9817, + "step": 47318 + }, + { + "epoch": 3.5, + "learning_rate": 7.822982079805939e-06, + "loss": 1.0603, + "step": 47319 + }, + { + "epoch": 3.5, + "learning_rate": 7.82259277672091e-06, + "loss": 1.0009, + "step": 47320 + }, + { + "epoch": 3.5, + "learning_rate": 7.822203477100102e-06, + "loss": 0.8929, + "step": 47321 + }, + { + "epoch": 3.5, + "learning_rate": 7.821814180944142e-06, + "loss": 0.9059, + "step": 47322 + }, + { + "epoch": 3.5, + "learning_rate": 7.821424888253646e-06, + "loss": 1.0507, + "step": 47323 + }, + { + "epoch": 3.5, + "learning_rate": 7.821035599029225e-06, + "loss": 0.9404, + "step": 47324 + }, + { + "epoch": 3.5, + "learning_rate": 7.820646313271509e-06, + "loss": 1.0422, + "step": 47325 + }, + { + "epoch": 3.5, + "learning_rate": 7.82025703098111e-06, + "loss": 0.9682, + "step": 47326 + }, + { + "epoch": 3.5, + "learning_rate": 7.819867752158656e-06, + "loss": 0.9903, + "step": 47327 + }, + { + "epoch": 3.5, + "learning_rate": 7.819478476804756e-06, + "loss": 0.9823, + "step": 47328 + }, + { + "epoch": 3.5, + "learning_rate": 7.819089204920041e-06, + "loss": 1.0858, + "step": 47329 + }, + { + "epoch": 3.5, + "learning_rate": 7.818699936505116e-06, + "loss": 0.9485, + "step": 47330 + }, + { + "epoch": 3.5, + "learning_rate": 7.818310671560616e-06, + "loss": 0.9711, + "step": 47331 + }, + { + "epoch": 3.5, + "learning_rate": 7.817921410087147e-06, + "loss": 0.927, + "step": 47332 + }, + { + "epoch": 3.5, + "learning_rate": 7.817532152085337e-06, + "loss": 0.9411, + "step": 47333 + }, + { + "epoch": 3.5, + "learning_rate": 7.817142897555803e-06, + "loss": 1.0834, + "step": 47334 + }, + { + "epoch": 3.5, + "learning_rate": 7.816753646499158e-06, + "loss": 1.0493, + "step": 47335 + }, + { + "epoch": 3.5, + "learning_rate": 7.816364398916031e-06, + "loss": 1.0351, + "step": 47336 + }, + { + "epoch": 3.5, + "learning_rate": 7.815975154807036e-06, + "loss": 0.9923, + "step": 47337 + }, + { + "epoch": 3.5, + "learning_rate": 7.815585914172794e-06, + "loss": 0.9483, + "step": 47338 + }, + { + "epoch": 3.5, + "learning_rate": 7.81519667701392e-06, + "loss": 1.074, + "step": 47339 + }, + { + "epoch": 3.5, + "learning_rate": 7.814807443331042e-06, + "loss": 1.1308, + "step": 47340 + }, + { + "epoch": 3.5, + "learning_rate": 7.81441821312477e-06, + "loss": 0.9929, + "step": 47341 + }, + { + "epoch": 3.5, + "learning_rate": 7.81402898639573e-06, + "loss": 1.0042, + "step": 47342 + }, + { + "epoch": 3.5, + "learning_rate": 7.81363976314454e-06, + "loss": 1.0097, + "step": 47343 + }, + { + "epoch": 3.5, + "learning_rate": 7.813250543371814e-06, + "loss": 1.0182, + "step": 47344 + }, + { + "epoch": 3.5, + "learning_rate": 7.812861327078177e-06, + "loss": 1.0407, + "step": 47345 + }, + { + "epoch": 3.5, + "learning_rate": 7.812472114264245e-06, + "loss": 1.0178, + "step": 47346 + }, + { + "epoch": 3.5, + "learning_rate": 7.812082904930643e-06, + "loss": 0.9525, + "step": 47347 + }, + { + "epoch": 3.5, + "learning_rate": 7.811693699077978e-06, + "loss": 1.0503, + "step": 47348 + }, + { + "epoch": 3.5, + "learning_rate": 7.811304496706885e-06, + "loss": 1.0373, + "step": 47349 + }, + { + "epoch": 3.5, + "learning_rate": 7.810915297817968e-06, + "loss": 1.0787, + "step": 47350 + }, + { + "epoch": 3.5, + "learning_rate": 7.810526102411859e-06, + "loss": 0.9512, + "step": 47351 + }, + { + "epoch": 3.5, + "learning_rate": 7.810136910489173e-06, + "loss": 0.9018, + "step": 47352 + }, + { + "epoch": 3.5, + "learning_rate": 7.809747722050523e-06, + "loss": 0.8203, + "step": 47353 + }, + { + "epoch": 3.5, + "learning_rate": 7.809358537096535e-06, + "loss": 0.946, + "step": 47354 + }, + { + "epoch": 3.5, + "learning_rate": 7.808969355627824e-06, + "loss": 1.1033, + "step": 47355 + }, + { + "epoch": 3.5, + "learning_rate": 7.808580177645014e-06, + "loss": 1.0591, + "step": 47356 + }, + { + "epoch": 3.5, + "learning_rate": 7.80819100314872e-06, + "loss": 1.0459, + "step": 47357 + }, + { + "epoch": 3.5, + "learning_rate": 7.807801832139563e-06, + "loss": 1.05, + "step": 47358 + }, + { + "epoch": 3.5, + "learning_rate": 7.807412664618163e-06, + "loss": 1.064, + "step": 47359 + }, + { + "epoch": 3.5, + "learning_rate": 7.807023500585138e-06, + "loss": 1.0254, + "step": 47360 + }, + { + "epoch": 3.5, + "learning_rate": 7.806634340041108e-06, + "loss": 0.923, + "step": 47361 + }, + { + "epoch": 3.5, + "learning_rate": 7.806245182986689e-06, + "loss": 0.9928, + "step": 47362 + }, + { + "epoch": 3.5, + "learning_rate": 7.805856029422503e-06, + "loss": 0.9956, + "step": 47363 + }, + { + "epoch": 3.5, + "learning_rate": 7.805466879349169e-06, + "loss": 1.035, + "step": 47364 + }, + { + "epoch": 3.5, + "learning_rate": 7.805077732767306e-06, + "loss": 1.0131, + "step": 47365 + }, + { + "epoch": 3.5, + "learning_rate": 7.804688589677533e-06, + "loss": 0.9477, + "step": 47366 + }, + { + "epoch": 3.5, + "learning_rate": 7.80429945008047e-06, + "loss": 1.0147, + "step": 47367 + }, + { + "epoch": 3.5, + "learning_rate": 7.803910313976732e-06, + "loss": 1.0749, + "step": 47368 + }, + { + "epoch": 3.5, + "learning_rate": 7.803521181366945e-06, + "loss": 1.0972, + "step": 47369 + }, + { + "epoch": 3.5, + "learning_rate": 7.803132052251727e-06, + "loss": 1.012, + "step": 47370 + }, + { + "epoch": 3.5, + "learning_rate": 7.802742926631685e-06, + "loss": 1.0067, + "step": 47371 + }, + { + "epoch": 3.5, + "learning_rate": 7.802353804507456e-06, + "loss": 0.9643, + "step": 47372 + }, + { + "epoch": 3.5, + "learning_rate": 7.801964685879644e-06, + "loss": 1.0445, + "step": 47373 + }, + { + "epoch": 3.5, + "learning_rate": 7.801575570748882e-06, + "loss": 0.9208, + "step": 47374 + }, + { + "epoch": 3.5, + "learning_rate": 7.801186459115775e-06, + "loss": 0.9798, + "step": 47375 + }, + { + "epoch": 3.5, + "learning_rate": 7.800797350980955e-06, + "loss": 1.018, + "step": 47376 + }, + { + "epoch": 3.5, + "learning_rate": 7.800408246345029e-06, + "loss": 1.1174, + "step": 47377 + }, + { + "epoch": 3.5, + "learning_rate": 7.800019145208629e-06, + "loss": 1.0029, + "step": 47378 + }, + { + "epoch": 3.5, + "learning_rate": 7.799630047572363e-06, + "loss": 1.0709, + "step": 47379 + }, + { + "epoch": 3.5, + "learning_rate": 7.799240953436853e-06, + "loss": 0.9978, + "step": 47380 + }, + { + "epoch": 3.5, + "learning_rate": 7.79885186280272e-06, + "loss": 0.9949, + "step": 47381 + }, + { + "epoch": 3.5, + "learning_rate": 7.798462775670582e-06, + "loss": 0.9072, + "step": 47382 + }, + { + "epoch": 3.5, + "learning_rate": 7.79807369204106e-06, + "loss": 1.0291, + "step": 47383 + }, + { + "epoch": 3.5, + "learning_rate": 7.79768461191477e-06, + "loss": 1.0736, + "step": 47384 + }, + { + "epoch": 3.5, + "learning_rate": 7.797295535292333e-06, + "loss": 1.0165, + "step": 47385 + }, + { + "epoch": 3.5, + "learning_rate": 7.796906462174364e-06, + "loss": 0.8973, + "step": 47386 + }, + { + "epoch": 3.5, + "learning_rate": 7.796517392561489e-06, + "loss": 1.0551, + "step": 47387 + }, + { + "epoch": 3.5, + "learning_rate": 7.79612832645432e-06, + "loss": 0.9591, + "step": 47388 + }, + { + "epoch": 3.5, + "learning_rate": 7.795739263853483e-06, + "loss": 1.0338, + "step": 47389 + }, + { + "epoch": 3.5, + "learning_rate": 7.795350204759595e-06, + "loss": 0.9464, + "step": 47390 + }, + { + "epoch": 3.5, + "learning_rate": 7.794961149173267e-06, + "loss": 1.0846, + "step": 47391 + }, + { + "epoch": 3.5, + "learning_rate": 7.79457209709513e-06, + "loss": 1.0834, + "step": 47392 + }, + { + "epoch": 3.5, + "learning_rate": 7.79418304852579e-06, + "loss": 0.9438, + "step": 47393 + }, + { + "epoch": 3.5, + "learning_rate": 7.793794003465883e-06, + "loss": 1.0237, + "step": 47394 + }, + { + "epoch": 3.5, + "learning_rate": 7.793404961916009e-06, + "loss": 0.8902, + "step": 47395 + }, + { + "epoch": 3.5, + "learning_rate": 7.793015923876802e-06, + "loss": 0.9751, + "step": 47396 + }, + { + "epoch": 3.5, + "learning_rate": 7.79262688934887e-06, + "loss": 0.9026, + "step": 47397 + }, + { + "epoch": 3.5, + "learning_rate": 7.792237858332842e-06, + "loss": 1.0532, + "step": 47398 + }, + { + "epoch": 3.5, + "learning_rate": 7.791848830829332e-06, + "loss": 1.0349, + "step": 47399 + }, + { + "epoch": 3.5, + "learning_rate": 7.791459806838956e-06, + "loss": 0.8461, + "step": 47400 + }, + { + "epoch": 3.5, + "learning_rate": 7.791070786362337e-06, + "loss": 1.0185, + "step": 47401 + }, + { + "epoch": 3.5, + "learning_rate": 7.790681769400092e-06, + "loss": 0.9625, + "step": 47402 + }, + { + "epoch": 3.5, + "learning_rate": 7.790292755952843e-06, + "loss": 0.9758, + "step": 47403 + }, + { + "epoch": 3.5, + "learning_rate": 7.789903746021203e-06, + "loss": 1.004, + "step": 47404 + }, + { + "epoch": 3.5, + "learning_rate": 7.789514739605799e-06, + "loss": 0.9518, + "step": 47405 + }, + { + "epoch": 3.5, + "learning_rate": 7.789125736707241e-06, + "loss": 1.0989, + "step": 47406 + }, + { + "epoch": 3.5, + "learning_rate": 7.788736737326155e-06, + "loss": 0.7944, + "step": 47407 + }, + { + "epoch": 3.5, + "learning_rate": 7.788347741463157e-06, + "loss": 0.9232, + "step": 47408 + }, + { + "epoch": 3.5, + "learning_rate": 7.787958749118865e-06, + "loss": 1.0808, + "step": 47409 + }, + { + "epoch": 3.5, + "learning_rate": 7.7875697602939e-06, + "loss": 1.022, + "step": 47410 + }, + { + "epoch": 3.5, + "learning_rate": 7.787180774988878e-06, + "loss": 1.0156, + "step": 47411 + }, + { + "epoch": 3.5, + "learning_rate": 7.786791793204422e-06, + "loss": 0.9772, + "step": 47412 + }, + { + "epoch": 3.5, + "learning_rate": 7.786402814941147e-06, + "loss": 0.9566, + "step": 47413 + }, + { + "epoch": 3.5, + "learning_rate": 7.786013840199674e-06, + "loss": 1.0005, + "step": 47414 + }, + { + "epoch": 3.5, + "learning_rate": 7.78562486898062e-06, + "loss": 1.0621, + "step": 47415 + }, + { + "epoch": 3.5, + "learning_rate": 7.785235901284608e-06, + "loss": 0.9539, + "step": 47416 + }, + { + "epoch": 3.5, + "learning_rate": 7.784846937112256e-06, + "loss": 1.0886, + "step": 47417 + }, + { + "epoch": 3.5, + "learning_rate": 7.784457976464175e-06, + "loss": 1.0514, + "step": 47418 + }, + { + "epoch": 3.5, + "learning_rate": 7.784069019340994e-06, + "loss": 1.0402, + "step": 47419 + }, + { + "epoch": 3.5, + "learning_rate": 7.783680065743323e-06, + "loss": 1.0724, + "step": 47420 + }, + { + "epoch": 3.5, + "learning_rate": 7.783291115671791e-06, + "loss": 0.9532, + "step": 47421 + }, + { + "epoch": 3.5, + "learning_rate": 7.782902169127004e-06, + "loss": 0.9998, + "step": 47422 + }, + { + "epoch": 3.5, + "learning_rate": 7.782513226109595e-06, + "loss": 1.0416, + "step": 47423 + }, + { + "epoch": 3.5, + "learning_rate": 7.782124286620171e-06, + "loss": 1.1185, + "step": 47424 + }, + { + "epoch": 3.5, + "learning_rate": 7.781735350659358e-06, + "loss": 0.9579, + "step": 47425 + }, + { + "epoch": 3.5, + "learning_rate": 7.781346418227773e-06, + "loss": 0.942, + "step": 47426 + }, + { + "epoch": 3.5, + "learning_rate": 7.78095748932603e-06, + "loss": 1.0722, + "step": 47427 + }, + { + "epoch": 3.5, + "learning_rate": 7.780568563954757e-06, + "loss": 0.8714, + "step": 47428 + }, + { + "epoch": 3.5, + "learning_rate": 7.780179642114564e-06, + "loss": 1.0635, + "step": 47429 + }, + { + "epoch": 3.5, + "learning_rate": 7.779790723806076e-06, + "loss": 0.9984, + "step": 47430 + }, + { + "epoch": 3.5, + "learning_rate": 7.779401809029906e-06, + "loss": 1.0715, + "step": 47431 + }, + { + "epoch": 3.5, + "learning_rate": 7.77901289778668e-06, + "loss": 0.9788, + "step": 47432 + }, + { + "epoch": 3.5, + "learning_rate": 7.778623990077008e-06, + "loss": 1.0594, + "step": 47433 + }, + { + "epoch": 3.5, + "learning_rate": 7.778235085901516e-06, + "loss": 0.9729, + "step": 47434 + }, + { + "epoch": 3.5, + "learning_rate": 7.777846185260823e-06, + "loss": 0.9987, + "step": 47435 + }, + { + "epoch": 3.51, + "learning_rate": 7.777457288155539e-06, + "loss": 0.9408, + "step": 47436 + }, + { + "epoch": 3.51, + "learning_rate": 7.777068394586295e-06, + "loss": 0.9591, + "step": 47437 + }, + { + "epoch": 3.51, + "learning_rate": 7.776679504553697e-06, + "loss": 1.0102, + "step": 47438 + }, + { + "epoch": 3.51, + "learning_rate": 7.776290618058376e-06, + "loss": 1.0259, + "step": 47439 + }, + { + "epoch": 3.51, + "learning_rate": 7.775901735100939e-06, + "loss": 0.9843, + "step": 47440 + }, + { + "epoch": 3.51, + "learning_rate": 7.775512855682016e-06, + "loss": 1.0043, + "step": 47441 + }, + { + "epoch": 3.51, + "learning_rate": 7.775123979802213e-06, + "loss": 0.9752, + "step": 47442 + }, + { + "epoch": 3.51, + "learning_rate": 7.774735107462165e-06, + "loss": 0.9448, + "step": 47443 + }, + { + "epoch": 3.51, + "learning_rate": 7.774346238662473e-06, + "loss": 0.9302, + "step": 47444 + }, + { + "epoch": 3.51, + "learning_rate": 7.773957373403773e-06, + "loss": 0.8535, + "step": 47445 + }, + { + "epoch": 3.51, + "learning_rate": 7.773568511686671e-06, + "loss": 0.9649, + "step": 47446 + }, + { + "epoch": 3.51, + "learning_rate": 7.773179653511787e-06, + "loss": 1.0804, + "step": 47447 + }, + { + "epoch": 3.51, + "learning_rate": 7.772790798879746e-06, + "loss": 1.0636, + "step": 47448 + }, + { + "epoch": 3.51, + "learning_rate": 7.77240194779116e-06, + "loss": 1.0051, + "step": 47449 + }, + { + "epoch": 3.51, + "learning_rate": 7.772013100246653e-06, + "loss": 1.0071, + "step": 47450 + }, + { + "epoch": 3.51, + "learning_rate": 7.771624256246839e-06, + "loss": 0.9258, + "step": 47451 + }, + { + "epoch": 3.51, + "learning_rate": 7.771235415792341e-06, + "loss": 1.0481, + "step": 47452 + }, + { + "epoch": 3.51, + "learning_rate": 7.770846578883774e-06, + "loss": 1.0334, + "step": 47453 + }, + { + "epoch": 3.51, + "learning_rate": 7.77045774552176e-06, + "loss": 0.9925, + "step": 47454 + }, + { + "epoch": 3.51, + "learning_rate": 7.770068915706915e-06, + "loss": 0.9593, + "step": 47455 + }, + { + "epoch": 3.51, + "learning_rate": 7.769680089439857e-06, + "loss": 0.9831, + "step": 47456 + }, + { + "epoch": 3.51, + "learning_rate": 7.769291266721208e-06, + "loss": 1.0165, + "step": 47457 + }, + { + "epoch": 3.51, + "learning_rate": 7.768902447551582e-06, + "loss": 1.0405, + "step": 47458 + }, + { + "epoch": 3.51, + "learning_rate": 7.768513631931604e-06, + "loss": 1.0224, + "step": 47459 + }, + { + "epoch": 3.51, + "learning_rate": 7.768124819861887e-06, + "loss": 1.0051, + "step": 47460 + }, + { + "epoch": 3.51, + "learning_rate": 7.767736011343052e-06, + "loss": 0.9951, + "step": 47461 + }, + { + "epoch": 3.51, + "learning_rate": 7.767347206375715e-06, + "loss": 1.06, + "step": 47462 + }, + { + "epoch": 3.51, + "learning_rate": 7.7669584049605e-06, + "loss": 1.0019, + "step": 47463 + }, + { + "epoch": 3.51, + "learning_rate": 7.766569607098024e-06, + "loss": 1.0601, + "step": 47464 + }, + { + "epoch": 3.51, + "learning_rate": 7.766180812788897e-06, + "loss": 1.0006, + "step": 47465 + }, + { + "epoch": 3.51, + "learning_rate": 7.76579202203375e-06, + "loss": 1.0156, + "step": 47466 + }, + { + "epoch": 3.51, + "learning_rate": 7.765403234833191e-06, + "loss": 1.0618, + "step": 47467 + }, + { + "epoch": 3.51, + "learning_rate": 7.765014451187848e-06, + "loss": 1.0751, + "step": 47468 + }, + { + "epoch": 3.51, + "learning_rate": 7.764625671098333e-06, + "loss": 1.0822, + "step": 47469 + }, + { + "epoch": 3.51, + "learning_rate": 7.764236894565266e-06, + "loss": 0.8841, + "step": 47470 + }, + { + "epoch": 3.51, + "learning_rate": 7.763848121589267e-06, + "loss": 0.9956, + "step": 47471 + }, + { + "epoch": 3.51, + "learning_rate": 7.763459352170955e-06, + "loss": 0.9195, + "step": 47472 + }, + { + "epoch": 3.51, + "learning_rate": 7.763070586310946e-06, + "loss": 1.0724, + "step": 47473 + }, + { + "epoch": 3.51, + "learning_rate": 7.762681824009859e-06, + "loss": 1.129, + "step": 47474 + }, + { + "epoch": 3.51, + "learning_rate": 7.762293065268315e-06, + "loss": 0.9863, + "step": 47475 + }, + { + "epoch": 3.51, + "learning_rate": 7.761904310086928e-06, + "loss": 0.876, + "step": 47476 + }, + { + "epoch": 3.51, + "learning_rate": 7.76151555846632e-06, + "loss": 0.9687, + "step": 47477 + }, + { + "epoch": 3.51, + "learning_rate": 7.76112681040711e-06, + "loss": 0.9827, + "step": 47478 + }, + { + "epoch": 3.51, + "learning_rate": 7.760738065909915e-06, + "loss": 1.135, + "step": 47479 + }, + { + "epoch": 3.51, + "learning_rate": 7.76034932497535e-06, + "loss": 0.9092, + "step": 47480 + }, + { + "epoch": 3.51, + "learning_rate": 7.759960587604043e-06, + "loss": 0.9066, + "step": 47481 + }, + { + "epoch": 3.51, + "learning_rate": 7.759571853796605e-06, + "loss": 0.9971, + "step": 47482 + }, + { + "epoch": 3.51, + "learning_rate": 7.759183123553654e-06, + "loss": 1.0389, + "step": 47483 + }, + { + "epoch": 3.51, + "learning_rate": 7.758794396875813e-06, + "loss": 1.0774, + "step": 47484 + }, + { + "epoch": 3.51, + "learning_rate": 7.758405673763693e-06, + "loss": 1.1149, + "step": 47485 + }, + { + "epoch": 3.51, + "learning_rate": 7.758016954217926e-06, + "loss": 1.105, + "step": 47486 + }, + { + "epoch": 3.51, + "learning_rate": 7.757628238239113e-06, + "loss": 1.1068, + "step": 47487 + }, + { + "epoch": 3.51, + "learning_rate": 7.757239525827888e-06, + "loss": 0.972, + "step": 47488 + }, + { + "epoch": 3.51, + "learning_rate": 7.756850816984858e-06, + "loss": 0.9228, + "step": 47489 + }, + { + "epoch": 3.51, + "learning_rate": 7.756462111710653e-06, + "loss": 1.031, + "step": 47490 + }, + { + "epoch": 3.51, + "learning_rate": 7.75607341000588e-06, + "loss": 0.8742, + "step": 47491 + }, + { + "epoch": 3.51, + "learning_rate": 7.75568471187116e-06, + "loss": 1.0153, + "step": 47492 + }, + { + "epoch": 3.51, + "learning_rate": 7.755296017307118e-06, + "loss": 0.8769, + "step": 47493 + }, + { + "epoch": 3.51, + "learning_rate": 7.754907326314364e-06, + "loss": 1.0695, + "step": 47494 + }, + { + "epoch": 3.51, + "learning_rate": 7.754518638893524e-06, + "loss": 0.9636, + "step": 47495 + }, + { + "epoch": 3.51, + "learning_rate": 7.75412995504521e-06, + "loss": 0.8573, + "step": 47496 + }, + { + "epoch": 3.51, + "learning_rate": 7.753741274770045e-06, + "loss": 0.9825, + "step": 47497 + }, + { + "epoch": 3.51, + "learning_rate": 7.753352598068642e-06, + "loss": 1.0462, + "step": 47498 + }, + { + "epoch": 3.51, + "learning_rate": 7.752963924941627e-06, + "loss": 0.8879, + "step": 47499 + }, + { + "epoch": 3.51, + "learning_rate": 7.752575255389611e-06, + "loss": 0.9882, + "step": 47500 + }, + { + "epoch": 3.51, + "learning_rate": 7.752186589413218e-06, + "loss": 1.0032, + "step": 47501 + }, + { + "epoch": 3.51, + "learning_rate": 7.751797927013065e-06, + "loss": 0.9415, + "step": 47502 + }, + { + "epoch": 3.51, + "learning_rate": 7.751409268189767e-06, + "loss": 0.9747, + "step": 47503 + }, + { + "epoch": 3.51, + "learning_rate": 7.751020612943946e-06, + "loss": 1.0733, + "step": 47504 + }, + { + "epoch": 3.51, + "learning_rate": 7.750631961276218e-06, + "loss": 0.9681, + "step": 47505 + }, + { + "epoch": 3.51, + "learning_rate": 7.750243313187202e-06, + "loss": 0.9976, + "step": 47506 + }, + { + "epoch": 3.51, + "learning_rate": 7.749854668677517e-06, + "loss": 0.8892, + "step": 47507 + }, + { + "epoch": 3.51, + "learning_rate": 7.749466027747782e-06, + "loss": 1.0841, + "step": 47508 + }, + { + "epoch": 3.51, + "learning_rate": 7.749077390398612e-06, + "loss": 1.0965, + "step": 47509 + }, + { + "epoch": 3.51, + "learning_rate": 7.748688756630632e-06, + "loss": 0.9448, + "step": 47510 + }, + { + "epoch": 3.51, + "learning_rate": 7.748300126444455e-06, + "loss": 1.0442, + "step": 47511 + }, + { + "epoch": 3.51, + "learning_rate": 7.747911499840697e-06, + "loss": 0.9606, + "step": 47512 + }, + { + "epoch": 3.51, + "learning_rate": 7.747522876819982e-06, + "loss": 1.0143, + "step": 47513 + }, + { + "epoch": 3.51, + "learning_rate": 7.747134257382925e-06, + "loss": 1.0809, + "step": 47514 + }, + { + "epoch": 3.51, + "learning_rate": 7.746745641530145e-06, + "loss": 0.9274, + "step": 47515 + }, + { + "epoch": 3.51, + "learning_rate": 7.74635702926226e-06, + "loss": 0.9289, + "step": 47516 + }, + { + "epoch": 3.51, + "learning_rate": 7.745968420579891e-06, + "loss": 1.0376, + "step": 47517 + }, + { + "epoch": 3.51, + "learning_rate": 7.74557981548365e-06, + "loss": 1.0362, + "step": 47518 + }, + { + "epoch": 3.51, + "learning_rate": 7.745191213974163e-06, + "loss": 1.0167, + "step": 47519 + }, + { + "epoch": 3.51, + "learning_rate": 7.744802616052044e-06, + "loss": 1.0412, + "step": 47520 + }, + { + "epoch": 3.51, + "learning_rate": 7.74441402171791e-06, + "loss": 0.9778, + "step": 47521 + }, + { + "epoch": 3.51, + "learning_rate": 7.744025430972384e-06, + "loss": 0.9952, + "step": 47522 + }, + { + "epoch": 3.51, + "learning_rate": 7.743636843816077e-06, + "loss": 0.9288, + "step": 47523 + }, + { + "epoch": 3.51, + "learning_rate": 7.743248260249615e-06, + "loss": 0.9842, + "step": 47524 + }, + { + "epoch": 3.51, + "learning_rate": 7.74285968027361e-06, + "loss": 0.9191, + "step": 47525 + }, + { + "epoch": 3.51, + "learning_rate": 7.742471103888686e-06, + "loss": 1.0382, + "step": 47526 + }, + { + "epoch": 3.51, + "learning_rate": 7.742082531095456e-06, + "loss": 0.8485, + "step": 47527 + }, + { + "epoch": 3.51, + "learning_rate": 7.74169396189454e-06, + "loss": 1.032, + "step": 47528 + }, + { + "epoch": 3.51, + "learning_rate": 7.741305396286562e-06, + "loss": 0.9502, + "step": 47529 + }, + { + "epoch": 3.51, + "learning_rate": 7.740916834272126e-06, + "loss": 0.9791, + "step": 47530 + }, + { + "epoch": 3.51, + "learning_rate": 7.740528275851868e-06, + "loss": 1.0124, + "step": 47531 + }, + { + "epoch": 3.51, + "learning_rate": 7.740139721026389e-06, + "loss": 1.0251, + "step": 47532 + }, + { + "epoch": 3.51, + "learning_rate": 7.73975116979632e-06, + "loss": 0.9545, + "step": 47533 + }, + { + "epoch": 3.51, + "learning_rate": 7.739362622162271e-06, + "loss": 0.9773, + "step": 47534 + }, + { + "epoch": 3.51, + "learning_rate": 7.73897407812487e-06, + "loss": 0.9739, + "step": 47535 + }, + { + "epoch": 3.51, + "learning_rate": 7.738585537684722e-06, + "loss": 0.9453, + "step": 47536 + }, + { + "epoch": 3.51, + "learning_rate": 7.738197000842457e-06, + "loss": 0.9883, + "step": 47537 + }, + { + "epoch": 3.51, + "learning_rate": 7.737808467598688e-06, + "loss": 1.0224, + "step": 47538 + }, + { + "epoch": 3.51, + "learning_rate": 7.73741993795403e-06, + "loss": 1.0181, + "step": 47539 + }, + { + "epoch": 3.51, + "learning_rate": 7.737031411909109e-06, + "loss": 1.0039, + "step": 47540 + }, + { + "epoch": 3.51, + "learning_rate": 7.736642889464534e-06, + "loss": 1.0867, + "step": 47541 + }, + { + "epoch": 3.51, + "learning_rate": 7.73625437062093e-06, + "loss": 1.0015, + "step": 47542 + }, + { + "epoch": 3.51, + "learning_rate": 7.735865855378911e-06, + "loss": 1.0549, + "step": 47543 + }, + { + "epoch": 3.51, + "learning_rate": 7.7354773437391e-06, + "loss": 0.939, + "step": 47544 + }, + { + "epoch": 3.51, + "learning_rate": 7.73508883570211e-06, + "loss": 0.9653, + "step": 47545 + }, + { + "epoch": 3.51, + "learning_rate": 7.734700331268564e-06, + "loss": 0.9828, + "step": 47546 + }, + { + "epoch": 3.51, + "learning_rate": 7.734311830439074e-06, + "loss": 0.9515, + "step": 47547 + }, + { + "epoch": 3.51, + "learning_rate": 7.733923333214263e-06, + "loss": 1.029, + "step": 47548 + }, + { + "epoch": 3.51, + "learning_rate": 7.733534839594747e-06, + "loss": 0.9387, + "step": 47549 + }, + { + "epoch": 3.51, + "learning_rate": 7.733146349581144e-06, + "loss": 0.9813, + "step": 47550 + }, + { + "epoch": 3.51, + "learning_rate": 7.732757863174075e-06, + "loss": 0.9234, + "step": 47551 + }, + { + "epoch": 3.51, + "learning_rate": 7.732369380374154e-06, + "loss": 0.9937, + "step": 47552 + }, + { + "epoch": 3.51, + "learning_rate": 7.731980901182002e-06, + "loss": 0.992, + "step": 47553 + }, + { + "epoch": 3.51, + "learning_rate": 7.731592425598234e-06, + "loss": 1.0045, + "step": 47554 + }, + { + "epoch": 3.51, + "learning_rate": 7.731203953623471e-06, + "loss": 1.0237, + "step": 47555 + }, + { + "epoch": 3.51, + "learning_rate": 7.730815485258331e-06, + "loss": 0.993, + "step": 47556 + }, + { + "epoch": 3.51, + "learning_rate": 7.730427020503432e-06, + "loss": 0.9979, + "step": 47557 + }, + { + "epoch": 3.51, + "learning_rate": 7.730038559359392e-06, + "loss": 0.959, + "step": 47558 + }, + { + "epoch": 3.51, + "learning_rate": 7.729650101826824e-06, + "loss": 1.0272, + "step": 47559 + }, + { + "epoch": 3.51, + "learning_rate": 7.729261647906353e-06, + "loss": 0.9258, + "step": 47560 + }, + { + "epoch": 3.51, + "learning_rate": 7.728873197598591e-06, + "loss": 0.996, + "step": 47561 + }, + { + "epoch": 3.51, + "learning_rate": 7.728484750904164e-06, + "loss": 1.0233, + "step": 47562 + }, + { + "epoch": 3.51, + "learning_rate": 7.728096307823681e-06, + "loss": 1.0526, + "step": 47563 + }, + { + "epoch": 3.51, + "learning_rate": 7.727707868357767e-06, + "loss": 1.118, + "step": 47564 + }, + { + "epoch": 3.51, + "learning_rate": 7.727319432507035e-06, + "loss": 1.0504, + "step": 47565 + }, + { + "epoch": 3.51, + "learning_rate": 7.726931000272108e-06, + "loss": 1.0973, + "step": 47566 + }, + { + "epoch": 3.51, + "learning_rate": 7.7265425716536e-06, + "loss": 0.978, + "step": 47567 + }, + { + "epoch": 3.51, + "learning_rate": 7.72615414665213e-06, + "loss": 1.0459, + "step": 47568 + }, + { + "epoch": 3.51, + "learning_rate": 7.725765725268318e-06, + "loss": 0.9538, + "step": 47569 + }, + { + "epoch": 3.51, + "learning_rate": 7.725377307502776e-06, + "loss": 0.9577, + "step": 47570 + }, + { + "epoch": 3.51, + "learning_rate": 7.724988893356129e-06, + "loss": 1.0085, + "step": 47571 + }, + { + "epoch": 3.52, + "learning_rate": 7.72460048282899e-06, + "loss": 1.0023, + "step": 47572 + }, + { + "epoch": 3.52, + "learning_rate": 7.724212075921983e-06, + "loss": 0.9883, + "step": 47573 + }, + { + "epoch": 3.52, + "learning_rate": 7.72382367263572e-06, + "loss": 1.1143, + "step": 47574 + }, + { + "epoch": 3.52, + "learning_rate": 7.723435272970821e-06, + "loss": 0.9601, + "step": 47575 + }, + { + "epoch": 3.52, + "learning_rate": 7.723046876927906e-06, + "loss": 1.0414, + "step": 47576 + }, + { + "epoch": 3.52, + "learning_rate": 7.722658484507584e-06, + "loss": 0.9819, + "step": 47577 + }, + { + "epoch": 3.52, + "learning_rate": 7.72227009571049e-06, + "loss": 1.0129, + "step": 47578 + }, + { + "epoch": 3.52, + "learning_rate": 7.721881710537221e-06, + "loss": 1.0112, + "step": 47579 + }, + { + "epoch": 3.52, + "learning_rate": 7.721493328988415e-06, + "loss": 1.0219, + "step": 47580 + }, + { + "epoch": 3.52, + "learning_rate": 7.721104951064672e-06, + "loss": 1.0248, + "step": 47581 + }, + { + "epoch": 3.52, + "learning_rate": 7.720716576766625e-06, + "loss": 0.9376, + "step": 47582 + }, + { + "epoch": 3.52, + "learning_rate": 7.720328206094881e-06, + "loss": 1.0155, + "step": 47583 + }, + { + "epoch": 3.52, + "learning_rate": 7.719939839050065e-06, + "loss": 1.0872, + "step": 47584 + }, + { + "epoch": 3.52, + "learning_rate": 7.719551475632792e-06, + "loss": 1.0416, + "step": 47585 + }, + { + "epoch": 3.52, + "learning_rate": 7.71916311584368e-06, + "loss": 0.9689, + "step": 47586 + }, + { + "epoch": 3.52, + "learning_rate": 7.718774759683346e-06, + "loss": 1.081, + "step": 47587 + }, + { + "epoch": 3.52, + "learning_rate": 7.718386407152407e-06, + "loss": 0.9635, + "step": 47588 + }, + { + "epoch": 3.52, + "learning_rate": 7.717998058251485e-06, + "loss": 1.0077, + "step": 47589 + }, + { + "epoch": 3.52, + "learning_rate": 7.717609712981195e-06, + "loss": 0.9515, + "step": 47590 + }, + { + "epoch": 3.52, + "learning_rate": 7.717221371342155e-06, + "loss": 0.9874, + "step": 47591 + }, + { + "epoch": 3.52, + "learning_rate": 7.716833033334982e-06, + "loss": 1.0279, + "step": 47592 + }, + { + "epoch": 3.52, + "learning_rate": 7.716444698960297e-06, + "loss": 1.1487, + "step": 47593 + }, + { + "epoch": 3.52, + "learning_rate": 7.716056368218716e-06, + "loss": 0.9379, + "step": 47594 + }, + { + "epoch": 3.52, + "learning_rate": 7.715668041110855e-06, + "loss": 1.0208, + "step": 47595 + }, + { + "epoch": 3.52, + "learning_rate": 7.715279717637334e-06, + "loss": 1.0222, + "step": 47596 + }, + { + "epoch": 3.52, + "learning_rate": 7.71489139779877e-06, + "loss": 0.8816, + "step": 47597 + }, + { + "epoch": 3.52, + "learning_rate": 7.714503081595782e-06, + "loss": 1.0119, + "step": 47598 + }, + { + "epoch": 3.52, + "learning_rate": 7.714114769028985e-06, + "loss": 1.0656, + "step": 47599 + }, + { + "epoch": 3.52, + "learning_rate": 7.713726460099001e-06, + "loss": 0.9729, + "step": 47600 + }, + { + "epoch": 3.52, + "learning_rate": 7.713338154806444e-06, + "loss": 0.8939, + "step": 47601 + }, + { + "epoch": 3.52, + "learning_rate": 7.712949853151936e-06, + "loss": 1.0276, + "step": 47602 + }, + { + "epoch": 3.52, + "learning_rate": 7.712561555136086e-06, + "loss": 1.0699, + "step": 47603 + }, + { + "epoch": 3.52, + "learning_rate": 7.712173260759526e-06, + "loss": 0.9054, + "step": 47604 + }, + { + "epoch": 3.52, + "learning_rate": 7.711784970022862e-06, + "loss": 1.0127, + "step": 47605 + }, + { + "epoch": 3.52, + "learning_rate": 7.711396682926713e-06, + "loss": 1.006, + "step": 47606 + }, + { + "epoch": 3.52, + "learning_rate": 7.711008399471702e-06, + "loss": 1.013, + "step": 47607 + }, + { + "epoch": 3.52, + "learning_rate": 7.71062011965844e-06, + "loss": 0.9868, + "step": 47608 + }, + { + "epoch": 3.52, + "learning_rate": 7.710231843487553e-06, + "loss": 1.0298, + "step": 47609 + }, + { + "epoch": 3.52, + "learning_rate": 7.709843570959652e-06, + "loss": 1.0212, + "step": 47610 + }, + { + "epoch": 3.52, + "learning_rate": 7.709455302075359e-06, + "loss": 0.9157, + "step": 47611 + }, + { + "epoch": 3.52, + "learning_rate": 7.709067036835288e-06, + "loss": 0.946, + "step": 47612 + }, + { + "epoch": 3.52, + "learning_rate": 7.708678775240059e-06, + "loss": 1.0781, + "step": 47613 + }, + { + "epoch": 3.52, + "learning_rate": 7.708290517290292e-06, + "loss": 1.1814, + "step": 47614 + }, + { + "epoch": 3.52, + "learning_rate": 7.707902262986598e-06, + "loss": 1.0506, + "step": 47615 + }, + { + "epoch": 3.52, + "learning_rate": 7.7075140123296e-06, + "loss": 0.9494, + "step": 47616 + }, + { + "epoch": 3.52, + "learning_rate": 7.707125765319914e-06, + "loss": 1.0133, + "step": 47617 + }, + { + "epoch": 3.52, + "learning_rate": 7.70673752195816e-06, + "loss": 0.9709, + "step": 47618 + }, + { + "epoch": 3.52, + "learning_rate": 7.706349282244951e-06, + "loss": 0.9774, + "step": 47619 + }, + { + "epoch": 3.52, + "learning_rate": 7.70596104618091e-06, + "loss": 1.0744, + "step": 47620 + }, + { + "epoch": 3.52, + "learning_rate": 7.70557281376665e-06, + "loss": 0.959, + "step": 47621 + }, + { + "epoch": 3.52, + "learning_rate": 7.705184585002793e-06, + "loss": 0.9566, + "step": 47622 + }, + { + "epoch": 3.52, + "learning_rate": 7.704796359889956e-06, + "loss": 0.9182, + "step": 47623 + }, + { + "epoch": 3.52, + "learning_rate": 7.70440813842875e-06, + "loss": 1.0182, + "step": 47624 + }, + { + "epoch": 3.52, + "learning_rate": 7.704019920619803e-06, + "loss": 1.037, + "step": 47625 + }, + { + "epoch": 3.52, + "learning_rate": 7.703631706463722e-06, + "loss": 1.0742, + "step": 47626 + }, + { + "epoch": 3.52, + "learning_rate": 7.703243495961135e-06, + "loss": 1.1092, + "step": 47627 + }, + { + "epoch": 3.52, + "learning_rate": 7.702855289112649e-06, + "loss": 0.9575, + "step": 47628 + }, + { + "epoch": 3.52, + "learning_rate": 7.702467085918895e-06, + "loss": 0.9625, + "step": 47629 + }, + { + "epoch": 3.52, + "learning_rate": 7.702078886380476e-06, + "loss": 1.1223, + "step": 47630 + }, + { + "epoch": 3.52, + "learning_rate": 7.701690690498023e-06, + "loss": 0.9457, + "step": 47631 + }, + { + "epoch": 3.52, + "learning_rate": 7.701302498272145e-06, + "loss": 0.9996, + "step": 47632 + }, + { + "epoch": 3.52, + "learning_rate": 7.70091430970346e-06, + "loss": 1.0654, + "step": 47633 + }, + { + "epoch": 3.52, + "learning_rate": 7.700526124792588e-06, + "loss": 0.8924, + "step": 47634 + }, + { + "epoch": 3.52, + "learning_rate": 7.700137943540145e-06, + "loss": 1.0217, + "step": 47635 + }, + { + "epoch": 3.52, + "learning_rate": 7.699749765946752e-06, + "loss": 1.0086, + "step": 47636 + }, + { + "epoch": 3.52, + "learning_rate": 7.699361592013022e-06, + "loss": 1.0163, + "step": 47637 + }, + { + "epoch": 3.52, + "learning_rate": 7.698973421739576e-06, + "loss": 0.8804, + "step": 47638 + }, + { + "epoch": 3.52, + "learning_rate": 7.698585255127029e-06, + "loss": 1.0654, + "step": 47639 + }, + { + "epoch": 3.52, + "learning_rate": 7.698197092176003e-06, + "loss": 1.1315, + "step": 47640 + }, + { + "epoch": 3.52, + "learning_rate": 7.69780893288711e-06, + "loss": 1.0206, + "step": 47641 + }, + { + "epoch": 3.52, + "learning_rate": 7.697420777260971e-06, + "loss": 0.8211, + "step": 47642 + }, + { + "epoch": 3.52, + "learning_rate": 7.697032625298202e-06, + "loss": 0.9964, + "step": 47643 + }, + { + "epoch": 3.52, + "learning_rate": 7.69664447699942e-06, + "loss": 1.0035, + "step": 47644 + }, + { + "epoch": 3.52, + "learning_rate": 7.69625633236525e-06, + "loss": 0.9609, + "step": 47645 + }, + { + "epoch": 3.52, + "learning_rate": 7.695868191396294e-06, + "loss": 0.8607, + "step": 47646 + }, + { + "epoch": 3.52, + "learning_rate": 7.695480054093186e-06, + "loss": 1.0135, + "step": 47647 + }, + { + "epoch": 3.52, + "learning_rate": 7.69509192045653e-06, + "loss": 1.0291, + "step": 47648 + }, + { + "epoch": 3.52, + "learning_rate": 7.694703790486956e-06, + "loss": 1.1478, + "step": 47649 + }, + { + "epoch": 3.52, + "learning_rate": 7.694315664185074e-06, + "loss": 1.0475, + "step": 47650 + }, + { + "epoch": 3.52, + "learning_rate": 7.6939275415515e-06, + "loss": 1.0431, + "step": 47651 + }, + { + "epoch": 3.52, + "learning_rate": 7.693539422586857e-06, + "loss": 0.952, + "step": 47652 + }, + { + "epoch": 3.52, + "learning_rate": 7.693151307291756e-06, + "loss": 1.0585, + "step": 47653 + }, + { + "epoch": 3.52, + "learning_rate": 7.69276319566682e-06, + "loss": 0.9878, + "step": 47654 + }, + { + "epoch": 3.52, + "learning_rate": 7.692375087712665e-06, + "loss": 1.0494, + "step": 47655 + }, + { + "epoch": 3.52, + "learning_rate": 7.691986983429909e-06, + "loss": 0.919, + "step": 47656 + }, + { + "epoch": 3.52, + "learning_rate": 7.691598882819168e-06, + "loss": 0.8518, + "step": 47657 + }, + { + "epoch": 3.52, + "learning_rate": 7.69121078588106e-06, + "loss": 1.0101, + "step": 47658 + }, + { + "epoch": 3.52, + "learning_rate": 7.690822692616203e-06, + "loss": 0.9858, + "step": 47659 + }, + { + "epoch": 3.52, + "learning_rate": 7.690434603025215e-06, + "loss": 1.063, + "step": 47660 + }, + { + "epoch": 3.52, + "learning_rate": 7.690046517108712e-06, + "loss": 0.9599, + "step": 47661 + }, + { + "epoch": 3.52, + "learning_rate": 7.68965843486731e-06, + "loss": 0.9204, + "step": 47662 + }, + { + "epoch": 3.52, + "learning_rate": 7.689270356301632e-06, + "loss": 1.0591, + "step": 47663 + }, + { + "epoch": 3.52, + "learning_rate": 7.688882281412288e-06, + "loss": 1.0337, + "step": 47664 + }, + { + "epoch": 3.52, + "learning_rate": 7.688494210199902e-06, + "loss": 1.0448, + "step": 47665 + }, + { + "epoch": 3.52, + "learning_rate": 7.688106142665087e-06, + "loss": 0.8825, + "step": 47666 + }, + { + "epoch": 3.52, + "learning_rate": 7.687718078808462e-06, + "loss": 0.9128, + "step": 47667 + }, + { + "epoch": 3.52, + "learning_rate": 7.687330018630643e-06, + "loss": 1.0069, + "step": 47668 + }, + { + "epoch": 3.52, + "learning_rate": 7.686941962132253e-06, + "loss": 1.0651, + "step": 47669 + }, + { + "epoch": 3.52, + "learning_rate": 7.686553909313906e-06, + "loss": 0.9786, + "step": 47670 + }, + { + "epoch": 3.52, + "learning_rate": 7.686165860176212e-06, + "loss": 0.9169, + "step": 47671 + }, + { + "epoch": 3.52, + "learning_rate": 7.685777814719801e-06, + "loss": 1.0533, + "step": 47672 + }, + { + "epoch": 3.52, + "learning_rate": 7.68538977294528e-06, + "loss": 0.9954, + "step": 47673 + }, + { + "epoch": 3.52, + "learning_rate": 7.685001734853275e-06, + "loss": 1.0231, + "step": 47674 + }, + { + "epoch": 3.52, + "learning_rate": 7.684613700444394e-06, + "loss": 1.1424, + "step": 47675 + }, + { + "epoch": 3.52, + "learning_rate": 7.684225669719265e-06, + "loss": 1.0135, + "step": 47676 + }, + { + "epoch": 3.52, + "learning_rate": 7.683837642678494e-06, + "loss": 1.0432, + "step": 47677 + }, + { + "epoch": 3.52, + "learning_rate": 7.68344961932271e-06, + "loss": 1.1399, + "step": 47678 + }, + { + "epoch": 3.52, + "learning_rate": 7.683061599652524e-06, + "loss": 1.0546, + "step": 47679 + }, + { + "epoch": 3.52, + "learning_rate": 7.68267358366855e-06, + "loss": 1.1171, + "step": 47680 + }, + { + "epoch": 3.52, + "learning_rate": 7.682285571371412e-06, + "loss": 0.9738, + "step": 47681 + }, + { + "epoch": 3.52, + "learning_rate": 7.681897562761723e-06, + "loss": 0.9684, + "step": 47682 + }, + { + "epoch": 3.52, + "learning_rate": 7.681509557840104e-06, + "loss": 0.9933, + "step": 47683 + }, + { + "epoch": 3.52, + "learning_rate": 7.681121556607167e-06, + "loss": 0.9085, + "step": 47684 + }, + { + "epoch": 3.52, + "learning_rate": 7.680733559063534e-06, + "loss": 1.0267, + "step": 47685 + }, + { + "epoch": 3.52, + "learning_rate": 7.680345565209819e-06, + "loss": 0.9886, + "step": 47686 + }, + { + "epoch": 3.52, + "learning_rate": 7.679957575046645e-06, + "loss": 0.9935, + "step": 47687 + }, + { + "epoch": 3.52, + "learning_rate": 7.679569588574623e-06, + "loss": 0.8097, + "step": 47688 + }, + { + "epoch": 3.52, + "learning_rate": 7.67918160579437e-06, + "loss": 1.0589, + "step": 47689 + }, + { + "epoch": 3.52, + "learning_rate": 7.678793626706512e-06, + "loss": 0.9342, + "step": 47690 + }, + { + "epoch": 3.52, + "learning_rate": 7.678405651311654e-06, + "loss": 1.0459, + "step": 47691 + }, + { + "epoch": 3.52, + "learning_rate": 7.678017679610424e-06, + "loss": 1.0558, + "step": 47692 + }, + { + "epoch": 3.52, + "learning_rate": 7.67762971160343e-06, + "loss": 0.9688, + "step": 47693 + }, + { + "epoch": 3.52, + "learning_rate": 7.677241747291298e-06, + "loss": 1.0799, + "step": 47694 + }, + { + "epoch": 3.52, + "learning_rate": 7.676853786674635e-06, + "loss": 1.1151, + "step": 47695 + }, + { + "epoch": 3.52, + "learning_rate": 7.676465829754074e-06, + "loss": 1.1012, + "step": 47696 + }, + { + "epoch": 3.52, + "learning_rate": 7.676077876530219e-06, + "loss": 1.0157, + "step": 47697 + }, + { + "epoch": 3.52, + "learning_rate": 7.675689927003686e-06, + "loss": 1.0007, + "step": 47698 + }, + { + "epoch": 3.52, + "learning_rate": 7.675301981175102e-06, + "loss": 1.0566, + "step": 47699 + }, + { + "epoch": 3.52, + "learning_rate": 7.674914039045076e-06, + "loss": 1.004, + "step": 47700 + }, + { + "epoch": 3.52, + "learning_rate": 7.674526100614233e-06, + "loss": 1.1125, + "step": 47701 + }, + { + "epoch": 3.52, + "learning_rate": 7.67413816588318e-06, + "loss": 0.9595, + "step": 47702 + }, + { + "epoch": 3.52, + "learning_rate": 7.673750234852545e-06, + "loss": 0.9565, + "step": 47703 + }, + { + "epoch": 3.52, + "learning_rate": 7.673362307522936e-06, + "loss": 1.0323, + "step": 47704 + }, + { + "epoch": 3.52, + "learning_rate": 7.672974383894978e-06, + "loss": 1.0652, + "step": 47705 + }, + { + "epoch": 3.52, + "learning_rate": 7.672586463969283e-06, + "loss": 1.0355, + "step": 47706 + }, + { + "epoch": 3.53, + "learning_rate": 7.672198547746468e-06, + "loss": 0.9525, + "step": 47707 + }, + { + "epoch": 3.53, + "learning_rate": 7.671810635227155e-06, + "loss": 1.0256, + "step": 47708 + }, + { + "epoch": 3.53, + "learning_rate": 7.671422726411955e-06, + "loss": 1.0134, + "step": 47709 + }, + { + "epoch": 3.53, + "learning_rate": 7.67103482130149e-06, + "loss": 0.9798, + "step": 47710 + }, + { + "epoch": 3.53, + "learning_rate": 7.670646919896373e-06, + "loss": 0.9233, + "step": 47711 + }, + { + "epoch": 3.53, + "learning_rate": 7.670259022197225e-06, + "loss": 0.8654, + "step": 47712 + }, + { + "epoch": 3.53, + "learning_rate": 7.66987112820466e-06, + "loss": 0.9018, + "step": 47713 + }, + { + "epoch": 3.53, + "learning_rate": 7.6694832379193e-06, + "loss": 1.0014, + "step": 47714 + }, + { + "epoch": 3.53, + "learning_rate": 7.669095351341756e-06, + "loss": 0.9874, + "step": 47715 + }, + { + "epoch": 3.53, + "learning_rate": 7.66870746847265e-06, + "loss": 0.971, + "step": 47716 + }, + { + "epoch": 3.53, + "learning_rate": 7.668319589312598e-06, + "loss": 1.0329, + "step": 47717 + }, + { + "epoch": 3.53, + "learning_rate": 7.667931713862212e-06, + "loss": 0.993, + "step": 47718 + }, + { + "epoch": 3.53, + "learning_rate": 7.667543842122118e-06, + "loss": 0.958, + "step": 47719 + }, + { + "epoch": 3.53, + "learning_rate": 7.667155974092923e-06, + "loss": 1.0502, + "step": 47720 + }, + { + "epoch": 3.53, + "learning_rate": 7.666768109775255e-06, + "loss": 1.0316, + "step": 47721 + }, + { + "epoch": 3.53, + "learning_rate": 7.66638024916972e-06, + "loss": 1.0021, + "step": 47722 + }, + { + "epoch": 3.53, + "learning_rate": 7.665992392276947e-06, + "loss": 0.9201, + "step": 47723 + }, + { + "epoch": 3.53, + "learning_rate": 7.66560453909754e-06, + "loss": 0.9699, + "step": 47724 + }, + { + "epoch": 3.53, + "learning_rate": 7.665216689632128e-06, + "loss": 1.0034, + "step": 47725 + }, + { + "epoch": 3.53, + "learning_rate": 7.66482884388132e-06, + "loss": 0.9975, + "step": 47726 + }, + { + "epoch": 3.53, + "learning_rate": 7.664441001845737e-06, + "loss": 1.0042, + "step": 47727 + }, + { + "epoch": 3.53, + "learning_rate": 7.664053163525995e-06, + "loss": 1.0463, + "step": 47728 + }, + { + "epoch": 3.53, + "learning_rate": 7.663665328922709e-06, + "loss": 1.0227, + "step": 47729 + }, + { + "epoch": 3.53, + "learning_rate": 7.663277498036501e-06, + "loss": 0.9142, + "step": 47730 + }, + { + "epoch": 3.53, + "learning_rate": 7.662889670867984e-06, + "loss": 1.0383, + "step": 47731 + }, + { + "epoch": 3.53, + "learning_rate": 7.662501847417776e-06, + "loss": 0.9941, + "step": 47732 + }, + { + "epoch": 3.53, + "learning_rate": 7.662114027686492e-06, + "loss": 0.9125, + "step": 47733 + }, + { + "epoch": 3.53, + "learning_rate": 7.661726211674754e-06, + "loss": 0.887, + "step": 47734 + }, + { + "epoch": 3.53, + "learning_rate": 7.66133839938318e-06, + "loss": 0.9594, + "step": 47735 + }, + { + "epoch": 3.53, + "learning_rate": 7.660950590812374e-06, + "loss": 0.9718, + "step": 47736 + }, + { + "epoch": 3.53, + "learning_rate": 7.66056278596297e-06, + "loss": 1.1067, + "step": 47737 + }, + { + "epoch": 3.53, + "learning_rate": 7.66017498483557e-06, + "loss": 1.0396, + "step": 47738 + }, + { + "epoch": 3.53, + "learning_rate": 7.659787187430805e-06, + "loss": 0.9738, + "step": 47739 + }, + { + "epoch": 3.53, + "learning_rate": 7.659399393749278e-06, + "loss": 0.9587, + "step": 47740 + }, + { + "epoch": 3.53, + "learning_rate": 7.65901160379162e-06, + "loss": 1.0418, + "step": 47741 + }, + { + "epoch": 3.53, + "learning_rate": 7.658623817558435e-06, + "loss": 0.9547, + "step": 47742 + }, + { + "epoch": 3.53, + "learning_rate": 7.658236035050353e-06, + "loss": 0.9229, + "step": 47743 + }, + { + "epoch": 3.53, + "learning_rate": 7.65784825626798e-06, + "loss": 0.9833, + "step": 47744 + }, + { + "epoch": 3.53, + "learning_rate": 7.657460481211936e-06, + "loss": 0.9461, + "step": 47745 + }, + { + "epoch": 3.53, + "learning_rate": 7.65707270988284e-06, + "loss": 1.0236, + "step": 47746 + }, + { + "epoch": 3.53, + "learning_rate": 7.656684942281308e-06, + "loss": 1.0973, + "step": 47747 + }, + { + "epoch": 3.53, + "learning_rate": 7.656297178407956e-06, + "loss": 1.0446, + "step": 47748 + }, + { + "epoch": 3.53, + "learning_rate": 7.655909418263401e-06, + "loss": 1.1052, + "step": 47749 + }, + { + "epoch": 3.53, + "learning_rate": 7.655521661848262e-06, + "loss": 0.9567, + "step": 47750 + }, + { + "epoch": 3.53, + "learning_rate": 7.655133909163152e-06, + "loss": 0.9338, + "step": 47751 + }, + { + "epoch": 3.53, + "learning_rate": 7.654746160208693e-06, + "loss": 1.0319, + "step": 47752 + }, + { + "epoch": 3.53, + "learning_rate": 7.654358414985501e-06, + "loss": 1.0444, + "step": 47753 + }, + { + "epoch": 3.53, + "learning_rate": 7.653970673494188e-06, + "loss": 1.0107, + "step": 47754 + }, + { + "epoch": 3.53, + "learning_rate": 7.653582935735376e-06, + "loss": 1.0316, + "step": 47755 + }, + { + "epoch": 3.53, + "learning_rate": 7.653195201709676e-06, + "loss": 1.0578, + "step": 47756 + }, + { + "epoch": 3.53, + "learning_rate": 7.652807471417713e-06, + "loss": 0.9288, + "step": 47757 + }, + { + "epoch": 3.53, + "learning_rate": 7.652419744860098e-06, + "loss": 1.0248, + "step": 47758 + }, + { + "epoch": 3.53, + "learning_rate": 7.65203202203745e-06, + "loss": 1.0755, + "step": 47759 + }, + { + "epoch": 3.53, + "learning_rate": 7.651644302950385e-06, + "loss": 0.9764, + "step": 47760 + }, + { + "epoch": 3.53, + "learning_rate": 7.651256587599522e-06, + "loss": 1.0578, + "step": 47761 + }, + { + "epoch": 3.53, + "learning_rate": 7.650868875985474e-06, + "loss": 1.1007, + "step": 47762 + }, + { + "epoch": 3.53, + "learning_rate": 7.650481168108863e-06, + "loss": 1.1098, + "step": 47763 + }, + { + "epoch": 3.53, + "learning_rate": 7.650093463970302e-06, + "loss": 1.0586, + "step": 47764 + }, + { + "epoch": 3.53, + "learning_rate": 7.649705763570406e-06, + "loss": 1.135, + "step": 47765 + }, + { + "epoch": 3.53, + "learning_rate": 7.6493180669098e-06, + "loss": 1.0495, + "step": 47766 + }, + { + "epoch": 3.53, + "learning_rate": 7.64893037398909e-06, + "loss": 0.9857, + "step": 47767 + }, + { + "epoch": 3.53, + "learning_rate": 7.6485426848089e-06, + "loss": 1.0316, + "step": 47768 + }, + { + "epoch": 3.53, + "learning_rate": 7.648154999369844e-06, + "loss": 0.9754, + "step": 47769 + }, + { + "epoch": 3.53, + "learning_rate": 7.64776731767254e-06, + "loss": 1.0466, + "step": 47770 + }, + { + "epoch": 3.53, + "learning_rate": 7.647379639717603e-06, + "loss": 1.1032, + "step": 47771 + }, + { + "epoch": 3.53, + "learning_rate": 7.646991965505655e-06, + "loss": 1.0367, + "step": 47772 + }, + { + "epoch": 3.53, + "learning_rate": 7.646604295037308e-06, + "loss": 1.0295, + "step": 47773 + }, + { + "epoch": 3.53, + "learning_rate": 7.646216628313177e-06, + "loss": 0.9122, + "step": 47774 + }, + { + "epoch": 3.53, + "learning_rate": 7.645828965333886e-06, + "loss": 1.0195, + "step": 47775 + }, + { + "epoch": 3.53, + "learning_rate": 7.645441306100045e-06, + "loss": 0.8958, + "step": 47776 + }, + { + "epoch": 3.53, + "learning_rate": 7.645053650612274e-06, + "loss": 1.1653, + "step": 47777 + }, + { + "epoch": 3.53, + "learning_rate": 7.644665998871186e-06, + "loss": 1.0253, + "step": 47778 + }, + { + "epoch": 3.53, + "learning_rate": 7.644278350877404e-06, + "loss": 0.9144, + "step": 47779 + }, + { + "epoch": 3.53, + "learning_rate": 7.64389070663154e-06, + "loss": 0.9945, + "step": 47780 + }, + { + "epoch": 3.53, + "learning_rate": 7.643503066134214e-06, + "loss": 0.9271, + "step": 47781 + }, + { + "epoch": 3.53, + "learning_rate": 7.643115429386042e-06, + "loss": 1.0034, + "step": 47782 + }, + { + "epoch": 3.53, + "learning_rate": 7.642727796387634e-06, + "loss": 1.1255, + "step": 47783 + }, + { + "epoch": 3.53, + "learning_rate": 7.642340167139618e-06, + "loss": 1.0926, + "step": 47784 + }, + { + "epoch": 3.53, + "learning_rate": 7.6419525416426e-06, + "loss": 1.0636, + "step": 47785 + }, + { + "epoch": 3.53, + "learning_rate": 7.641564919897208e-06, + "loss": 1.1271, + "step": 47786 + }, + { + "epoch": 3.53, + "learning_rate": 7.641177301904046e-06, + "loss": 0.9356, + "step": 47787 + }, + { + "epoch": 3.53, + "learning_rate": 7.640789687663742e-06, + "loss": 1.0851, + "step": 47788 + }, + { + "epoch": 3.53, + "learning_rate": 7.640402077176903e-06, + "loss": 0.9646, + "step": 47789 + }, + { + "epoch": 3.53, + "learning_rate": 7.640014470444157e-06, + "loss": 0.9348, + "step": 47790 + }, + { + "epoch": 3.53, + "learning_rate": 7.639626867466111e-06, + "loss": 0.9927, + "step": 47791 + }, + { + "epoch": 3.53, + "learning_rate": 7.639239268243384e-06, + "loss": 1.0006, + "step": 47792 + }, + { + "epoch": 3.53, + "learning_rate": 7.638851672776594e-06, + "loss": 0.9364, + "step": 47793 + }, + { + "epoch": 3.53, + "learning_rate": 7.638464081066357e-06, + "loss": 0.992, + "step": 47794 + }, + { + "epoch": 3.53, + "learning_rate": 7.638076493113291e-06, + "loss": 0.9658, + "step": 47795 + }, + { + "epoch": 3.53, + "learning_rate": 7.63768890891801e-06, + "loss": 1.0267, + "step": 47796 + }, + { + "epoch": 3.53, + "learning_rate": 7.637301328481132e-06, + "loss": 0.9545, + "step": 47797 + }, + { + "epoch": 3.53, + "learning_rate": 7.636913751803274e-06, + "loss": 1.0815, + "step": 47798 + }, + { + "epoch": 3.53, + "learning_rate": 7.636526178885054e-06, + "loss": 0.9075, + "step": 47799 + }, + { + "epoch": 3.53, + "learning_rate": 7.636138609727087e-06, + "loss": 0.9203, + "step": 47800 + }, + { + "epoch": 3.53, + "learning_rate": 7.635751044329988e-06, + "loss": 0.8503, + "step": 47801 + }, + { + "epoch": 3.53, + "learning_rate": 7.635363482694377e-06, + "loss": 1.0534, + "step": 47802 + }, + { + "epoch": 3.53, + "learning_rate": 7.634975924820867e-06, + "loss": 1.0083, + "step": 47803 + }, + { + "epoch": 3.53, + "learning_rate": 7.634588370710078e-06, + "loss": 0.8926, + "step": 47804 + }, + { + "epoch": 3.53, + "learning_rate": 7.634200820362623e-06, + "loss": 1.0898, + "step": 47805 + }, + { + "epoch": 3.53, + "learning_rate": 7.633813273779123e-06, + "loss": 0.8737, + "step": 47806 + }, + { + "epoch": 3.53, + "learning_rate": 7.63342573096019e-06, + "loss": 1.0051, + "step": 47807 + }, + { + "epoch": 3.53, + "learning_rate": 7.633038191906444e-06, + "loss": 0.9834, + "step": 47808 + }, + { + "epoch": 3.53, + "learning_rate": 7.632650656618503e-06, + "loss": 0.9912, + "step": 47809 + }, + { + "epoch": 3.53, + "learning_rate": 7.632263125096977e-06, + "loss": 1.0147, + "step": 47810 + }, + { + "epoch": 3.53, + "learning_rate": 7.63187559734249e-06, + "loss": 0.9857, + "step": 47811 + }, + { + "epoch": 3.53, + "learning_rate": 7.631488073355652e-06, + "loss": 1.0257, + "step": 47812 + }, + { + "epoch": 3.53, + "learning_rate": 7.631100553137084e-06, + "loss": 1.0152, + "step": 47813 + }, + { + "epoch": 3.53, + "learning_rate": 7.6307130366874e-06, + "loss": 1.1142, + "step": 47814 + }, + { + "epoch": 3.53, + "learning_rate": 7.630325524007218e-06, + "loss": 1.0086, + "step": 47815 + }, + { + "epoch": 3.53, + "learning_rate": 7.629938015097153e-06, + "loss": 1.1168, + "step": 47816 + }, + { + "epoch": 3.53, + "learning_rate": 7.629550509957826e-06, + "loss": 1.0461, + "step": 47817 + }, + { + "epoch": 3.53, + "learning_rate": 7.629163008589846e-06, + "loss": 1.0469, + "step": 47818 + }, + { + "epoch": 3.53, + "learning_rate": 7.628775510993837e-06, + "loss": 1.0331, + "step": 47819 + }, + { + "epoch": 3.53, + "learning_rate": 7.628388017170413e-06, + "loss": 1.0511, + "step": 47820 + }, + { + "epoch": 3.53, + "learning_rate": 7.628000527120188e-06, + "loss": 1.0231, + "step": 47821 + }, + { + "epoch": 3.53, + "learning_rate": 7.62761304084378e-06, + "loss": 1.0936, + "step": 47822 + }, + { + "epoch": 3.53, + "learning_rate": 7.627225558341804e-06, + "loss": 0.9845, + "step": 47823 + }, + { + "epoch": 3.53, + "learning_rate": 7.6268380796148825e-06, + "loss": 0.9808, + "step": 47824 + }, + { + "epoch": 3.53, + "learning_rate": 7.626450604663623e-06, + "loss": 1.0297, + "step": 47825 + }, + { + "epoch": 3.53, + "learning_rate": 7.6260631334886505e-06, + "loss": 1.0683, + "step": 47826 + }, + { + "epoch": 3.53, + "learning_rate": 7.625675666090575e-06, + "loss": 1.0002, + "step": 47827 + }, + { + "epoch": 3.53, + "learning_rate": 7.625288202470017e-06, + "loss": 1.1095, + "step": 47828 + }, + { + "epoch": 3.53, + "learning_rate": 7.624900742627595e-06, + "loss": 1.0141, + "step": 47829 + }, + { + "epoch": 3.53, + "learning_rate": 7.624513286563914e-06, + "loss": 0.9665, + "step": 47830 + }, + { + "epoch": 3.53, + "learning_rate": 7.6241258342796055e-06, + "loss": 0.9482, + "step": 47831 + }, + { + "epoch": 3.53, + "learning_rate": 7.6237383857752725e-06, + "loss": 1.0999, + "step": 47832 + }, + { + "epoch": 3.53, + "learning_rate": 7.623350941051544e-06, + "loss": 0.9655, + "step": 47833 + }, + { + "epoch": 3.53, + "learning_rate": 7.622963500109024e-06, + "loss": 0.9841, + "step": 47834 + }, + { + "epoch": 3.53, + "learning_rate": 7.622576062948342e-06, + "loss": 1.0527, + "step": 47835 + }, + { + "epoch": 3.53, + "learning_rate": 7.6221886295700995e-06, + "loss": 0.8314, + "step": 47836 + }, + { + "epoch": 3.53, + "learning_rate": 7.6218011999749275e-06, + "loss": 1.0712, + "step": 47837 + }, + { + "epoch": 3.53, + "learning_rate": 7.621413774163434e-06, + "loss": 0.9706, + "step": 47838 + }, + { + "epoch": 3.53, + "learning_rate": 7.621026352136236e-06, + "loss": 0.9424, + "step": 47839 + }, + { + "epoch": 3.53, + "learning_rate": 7.620638933893951e-06, + "loss": 1.149, + "step": 47840 + }, + { + "epoch": 3.53, + "learning_rate": 7.620251519437197e-06, + "loss": 1.0292, + "step": 47841 + }, + { + "epoch": 3.54, + "learning_rate": 7.619864108766587e-06, + "loss": 1.0598, + "step": 47842 + }, + { + "epoch": 3.54, + "learning_rate": 7.6194767018827385e-06, + "loss": 0.9722, + "step": 47843 + }, + { + "epoch": 3.54, + "learning_rate": 7.619089298786271e-06, + "loss": 0.9264, + "step": 47844 + }, + { + "epoch": 3.54, + "learning_rate": 7.618701899477796e-06, + "loss": 0.9176, + "step": 47845 + }, + { + "epoch": 3.54, + "learning_rate": 7.618314503957934e-06, + "loss": 0.9379, + "step": 47846 + }, + { + "epoch": 3.54, + "learning_rate": 7.6179271122273e-06, + "loss": 1.121, + "step": 47847 + }, + { + "epoch": 3.54, + "learning_rate": 7.617539724286508e-06, + "loss": 0.9309, + "step": 47848 + }, + { + "epoch": 3.54, + "learning_rate": 7.617152340136178e-06, + "loss": 0.9573, + "step": 47849 + }, + { + "epoch": 3.54, + "learning_rate": 7.616764959776921e-06, + "loss": 0.9677, + "step": 47850 + }, + { + "epoch": 3.54, + "learning_rate": 7.616377583209361e-06, + "loss": 0.9526, + "step": 47851 + }, + { + "epoch": 3.54, + "learning_rate": 7.615990210434108e-06, + "loss": 0.8489, + "step": 47852 + }, + { + "epoch": 3.54, + "learning_rate": 7.615602841451781e-06, + "loss": 0.9594, + "step": 47853 + }, + { + "epoch": 3.54, + "learning_rate": 7.615215476262995e-06, + "loss": 1.0514, + "step": 47854 + }, + { + "epoch": 3.54, + "learning_rate": 7.614828114868371e-06, + "loss": 1.0054, + "step": 47855 + }, + { + "epoch": 3.54, + "learning_rate": 7.614440757268519e-06, + "loss": 1.0943, + "step": 47856 + }, + { + "epoch": 3.54, + "learning_rate": 7.614053403464056e-06, + "loss": 0.8561, + "step": 47857 + }, + { + "epoch": 3.54, + "learning_rate": 7.613666053455602e-06, + "loss": 1.1189, + "step": 47858 + }, + { + "epoch": 3.54, + "learning_rate": 7.613278707243767e-06, + "loss": 1.0155, + "step": 47859 + }, + { + "epoch": 3.54, + "learning_rate": 7.612891364829176e-06, + "loss": 1.0113, + "step": 47860 + }, + { + "epoch": 3.54, + "learning_rate": 7.612504026212437e-06, + "loss": 0.9915, + "step": 47861 + }, + { + "epoch": 3.54, + "learning_rate": 7.612116691394174e-06, + "loss": 0.911, + "step": 47862 + }, + { + "epoch": 3.54, + "learning_rate": 7.611729360374995e-06, + "loss": 1.0703, + "step": 47863 + }, + { + "epoch": 3.54, + "learning_rate": 7.611342033155524e-06, + "loss": 1.0663, + "step": 47864 + }, + { + "epoch": 3.54, + "learning_rate": 7.610954709736374e-06, + "loss": 0.9482, + "step": 47865 + }, + { + "epoch": 3.54, + "learning_rate": 7.610567390118156e-06, + "loss": 1.048, + "step": 47866 + }, + { + "epoch": 3.54, + "learning_rate": 7.610180074301496e-06, + "loss": 1.0, + "step": 47867 + }, + { + "epoch": 3.54, + "learning_rate": 7.609792762287003e-06, + "loss": 0.9945, + "step": 47868 + }, + { + "epoch": 3.54, + "learning_rate": 7.609405454075296e-06, + "loss": 1.116, + "step": 47869 + }, + { + "epoch": 3.54, + "learning_rate": 7.609018149666989e-06, + "loss": 1.0582, + "step": 47870 + }, + { + "epoch": 3.54, + "learning_rate": 7.608630849062702e-06, + "loss": 1.037, + "step": 47871 + }, + { + "epoch": 3.54, + "learning_rate": 7.6082435522630485e-06, + "loss": 1.0498, + "step": 47872 + }, + { + "epoch": 3.54, + "learning_rate": 7.607856259268647e-06, + "loss": 1.0447, + "step": 47873 + }, + { + "epoch": 3.54, + "learning_rate": 7.6074689700801085e-06, + "loss": 1.022, + "step": 47874 + }, + { + "epoch": 3.54, + "learning_rate": 7.6070816846980556e-06, + "loss": 0.9381, + "step": 47875 + }, + { + "epoch": 3.54, + "learning_rate": 7.606694403123104e-06, + "loss": 1.0274, + "step": 47876 + }, + { + "epoch": 3.54, + "learning_rate": 7.6063071253558595e-06, + "loss": 0.9948, + "step": 47877 + }, + { + "epoch": 3.54, + "learning_rate": 7.605919851396955e-06, + "loss": 1.0778, + "step": 47878 + }, + { + "epoch": 3.54, + "learning_rate": 7.605532581246989e-06, + "loss": 1.0914, + "step": 47879 + }, + { + "epoch": 3.54, + "learning_rate": 7.605145314906594e-06, + "loss": 1.0625, + "step": 47880 + }, + { + "epoch": 3.54, + "learning_rate": 7.604758052376371e-06, + "loss": 1.0473, + "step": 47881 + }, + { + "epoch": 3.54, + "learning_rate": 7.604370793656951e-06, + "loss": 1.0318, + "step": 47882 + }, + { + "epoch": 3.54, + "learning_rate": 7.603983538748937e-06, + "loss": 0.9025, + "step": 47883 + }, + { + "epoch": 3.54, + "learning_rate": 7.603596287652957e-06, + "loss": 1.0053, + "step": 47884 + }, + { + "epoch": 3.54, + "learning_rate": 7.603209040369618e-06, + "loss": 0.9593, + "step": 47885 + }, + { + "epoch": 3.54, + "learning_rate": 7.602821796899537e-06, + "loss": 1.0233, + "step": 47886 + }, + { + "epoch": 3.54, + "learning_rate": 7.602434557243334e-06, + "loss": 1.0373, + "step": 47887 + }, + { + "epoch": 3.54, + "learning_rate": 7.6020473214016235e-06, + "loss": 0.9605, + "step": 47888 + }, + { + "epoch": 3.54, + "learning_rate": 7.601660089375021e-06, + "loss": 1.0261, + "step": 47889 + }, + { + "epoch": 3.54, + "learning_rate": 7.6012728611641415e-06, + "loss": 1.0714, + "step": 47890 + }, + { + "epoch": 3.54, + "learning_rate": 7.600885636769606e-06, + "loss": 1.0541, + "step": 47891 + }, + { + "epoch": 3.54, + "learning_rate": 7.600498416192024e-06, + "loss": 1.0605, + "step": 47892 + }, + { + "epoch": 3.54, + "learning_rate": 7.600111199432016e-06, + "loss": 0.9772, + "step": 47893 + }, + { + "epoch": 3.54, + "learning_rate": 7.599723986490198e-06, + "loss": 0.9394, + "step": 47894 + }, + { + "epoch": 3.54, + "learning_rate": 7.599336777367183e-06, + "loss": 0.999, + "step": 47895 + }, + { + "epoch": 3.54, + "learning_rate": 7.598949572063589e-06, + "loss": 1.0589, + "step": 47896 + }, + { + "epoch": 3.54, + "learning_rate": 7.598562370580031e-06, + "loss": 1.0726, + "step": 47897 + }, + { + "epoch": 3.54, + "learning_rate": 7.59817517291713e-06, + "loss": 1.0358, + "step": 47898 + }, + { + "epoch": 3.54, + "learning_rate": 7.597787979075493e-06, + "loss": 0.8924, + "step": 47899 + }, + { + "epoch": 3.54, + "learning_rate": 7.597400789055746e-06, + "loss": 1.0074, + "step": 47900 + }, + { + "epoch": 3.54, + "learning_rate": 7.597013602858494e-06, + "loss": 1.0363, + "step": 47901 + }, + { + "epoch": 3.54, + "learning_rate": 7.5966264204843656e-06, + "loss": 0.9548, + "step": 47902 + }, + { + "epoch": 3.54, + "learning_rate": 7.596239241933967e-06, + "loss": 0.9307, + "step": 47903 + }, + { + "epoch": 3.54, + "learning_rate": 7.595852067207916e-06, + "loss": 0.9671, + "step": 47904 + }, + { + "epoch": 3.54, + "learning_rate": 7.595464896306831e-06, + "loss": 0.9718, + "step": 47905 + }, + { + "epoch": 3.54, + "learning_rate": 7.595077729231326e-06, + "loss": 1.05, + "step": 47906 + }, + { + "epoch": 3.54, + "learning_rate": 7.594690565982019e-06, + "loss": 0.9551, + "step": 47907 + }, + { + "epoch": 3.54, + "learning_rate": 7.594303406559525e-06, + "loss": 0.9241, + "step": 47908 + }, + { + "epoch": 3.54, + "learning_rate": 7.593916250964459e-06, + "loss": 0.9985, + "step": 47909 + }, + { + "epoch": 3.54, + "learning_rate": 7.5935290991974375e-06, + "loss": 1.0218, + "step": 47910 + }, + { + "epoch": 3.54, + "learning_rate": 7.593141951259079e-06, + "loss": 0.9419, + "step": 47911 + }, + { + "epoch": 3.54, + "learning_rate": 7.5927548071499964e-06, + "loss": 1.0033, + "step": 47912 + }, + { + "epoch": 3.54, + "learning_rate": 7.592367666870805e-06, + "loss": 0.8916, + "step": 47913 + }, + { + "epoch": 3.54, + "learning_rate": 7.591980530422124e-06, + "loss": 1.0511, + "step": 47914 + }, + { + "epoch": 3.54, + "learning_rate": 7.591593397804565e-06, + "loss": 0.9936, + "step": 47915 + }, + { + "epoch": 3.54, + "learning_rate": 7.591206269018749e-06, + "loss": 0.987, + "step": 47916 + }, + { + "epoch": 3.54, + "learning_rate": 7.590819144065287e-06, + "loss": 0.9991, + "step": 47917 + }, + { + "epoch": 3.54, + "learning_rate": 7.5904320229448006e-06, + "loss": 1.0972, + "step": 47918 + }, + { + "epoch": 3.54, + "learning_rate": 7.590044905657898e-06, + "loss": 1.0699, + "step": 47919 + }, + { + "epoch": 3.54, + "learning_rate": 7.589657792205203e-06, + "loss": 1.0134, + "step": 47920 + }, + { + "epoch": 3.54, + "learning_rate": 7.589270682587331e-06, + "loss": 0.9727, + "step": 47921 + }, + { + "epoch": 3.54, + "learning_rate": 7.588883576804886e-06, + "loss": 1.0408, + "step": 47922 + }, + { + "epoch": 3.54, + "learning_rate": 7.588496474858501e-06, + "loss": 1.0167, + "step": 47923 + }, + { + "epoch": 3.54, + "learning_rate": 7.588109376748777e-06, + "loss": 1.0574, + "step": 47924 + }, + { + "epoch": 3.54, + "learning_rate": 7.587722282476343e-06, + "loss": 0.9968, + "step": 47925 + }, + { + "epoch": 3.54, + "learning_rate": 7.587335192041802e-06, + "loss": 0.9473, + "step": 47926 + }, + { + "epoch": 3.54, + "learning_rate": 7.586948105445781e-06, + "loss": 1.0532, + "step": 47927 + }, + { + "epoch": 3.54, + "learning_rate": 7.586561022688887e-06, + "loss": 1.0434, + "step": 47928 + }, + { + "epoch": 3.54, + "learning_rate": 7.586173943771744e-06, + "loss": 1.0961, + "step": 47929 + }, + { + "epoch": 3.54, + "learning_rate": 7.585786868694959e-06, + "loss": 1.049, + "step": 47930 + }, + { + "epoch": 3.54, + "learning_rate": 7.585399797459159e-06, + "loss": 1.0837, + "step": 47931 + }, + { + "epoch": 3.54, + "learning_rate": 7.585012730064949e-06, + "loss": 0.9689, + "step": 47932 + }, + { + "epoch": 3.54, + "learning_rate": 7.5846256665129495e-06, + "loss": 1.0357, + "step": 47933 + }, + { + "epoch": 3.54, + "learning_rate": 7.584238606803777e-06, + "loss": 0.9329, + "step": 47934 + }, + { + "epoch": 3.54, + "learning_rate": 7.583851550938044e-06, + "loss": 1.0125, + "step": 47935 + }, + { + "epoch": 3.54, + "learning_rate": 7.583464498916372e-06, + "loss": 0.9571, + "step": 47936 + }, + { + "epoch": 3.54, + "learning_rate": 7.58307745073937e-06, + "loss": 0.9042, + "step": 47937 + }, + { + "epoch": 3.54, + "learning_rate": 7.58269040640766e-06, + "loss": 1.0674, + "step": 47938 + }, + { + "epoch": 3.54, + "learning_rate": 7.582303365921853e-06, + "loss": 0.8866, + "step": 47939 + }, + { + "epoch": 3.54, + "learning_rate": 7.581916329282568e-06, + "loss": 0.9546, + "step": 47940 + }, + { + "epoch": 3.54, + "learning_rate": 7.58152929649042e-06, + "loss": 0.9508, + "step": 47941 + }, + { + "epoch": 3.54, + "learning_rate": 7.581142267546022e-06, + "loss": 0.9188, + "step": 47942 + }, + { + "epoch": 3.54, + "learning_rate": 7.580755242449995e-06, + "loss": 0.9952, + "step": 47943 + }, + { + "epoch": 3.54, + "learning_rate": 7.580368221202949e-06, + "loss": 0.9839, + "step": 47944 + }, + { + "epoch": 3.54, + "learning_rate": 7.579981203805506e-06, + "loss": 1.1107, + "step": 47945 + }, + { + "epoch": 3.54, + "learning_rate": 7.579594190258274e-06, + "loss": 1.0513, + "step": 47946 + }, + { + "epoch": 3.54, + "learning_rate": 7.579207180561878e-06, + "loss": 0.933, + "step": 47947 + }, + { + "epoch": 3.54, + "learning_rate": 7.578820174716922e-06, + "loss": 0.9255, + "step": 47948 + }, + { + "epoch": 3.54, + "learning_rate": 7.578433172724036e-06, + "loss": 1.0057, + "step": 47949 + }, + { + "epoch": 3.54, + "learning_rate": 7.578046174583825e-06, + "loss": 1.0054, + "step": 47950 + }, + { + "epoch": 3.54, + "learning_rate": 7.577659180296905e-06, + "loss": 1.0394, + "step": 47951 + }, + { + "epoch": 3.54, + "learning_rate": 7.577272189863899e-06, + "loss": 0.9293, + "step": 47952 + }, + { + "epoch": 3.54, + "learning_rate": 7.576885203285415e-06, + "loss": 0.9851, + "step": 47953 + }, + { + "epoch": 3.54, + "learning_rate": 7.576498220562074e-06, + "loss": 1.0861, + "step": 47954 + }, + { + "epoch": 3.54, + "learning_rate": 7.576111241694488e-06, + "loss": 1.0974, + "step": 47955 + }, + { + "epoch": 3.54, + "learning_rate": 7.575724266683275e-06, + "loss": 1.068, + "step": 47956 + }, + { + "epoch": 3.54, + "learning_rate": 7.5753372955290504e-06, + "loss": 0.9262, + "step": 47957 + }, + { + "epoch": 3.54, + "learning_rate": 7.57495032823243e-06, + "loss": 1.0206, + "step": 47958 + }, + { + "epoch": 3.54, + "learning_rate": 7.574563364794029e-06, + "loss": 0.9504, + "step": 47959 + }, + { + "epoch": 3.54, + "learning_rate": 7.57417640521446e-06, + "loss": 1.076, + "step": 47960 + }, + { + "epoch": 3.54, + "learning_rate": 7.573789449494346e-06, + "loss": 0.9909, + "step": 47961 + }, + { + "epoch": 3.54, + "learning_rate": 7.573402497634295e-06, + "loss": 1.0211, + "step": 47962 + }, + { + "epoch": 3.54, + "learning_rate": 7.573015549634928e-06, + "loss": 0.9975, + "step": 47963 + }, + { + "epoch": 3.54, + "learning_rate": 7.572628605496856e-06, + "loss": 1.0351, + "step": 47964 + }, + { + "epoch": 3.54, + "learning_rate": 7.572241665220699e-06, + "loss": 0.947, + "step": 47965 + }, + { + "epoch": 3.54, + "learning_rate": 7.5718547288070695e-06, + "loss": 1.0162, + "step": 47966 + }, + { + "epoch": 3.54, + "learning_rate": 7.5714677962565865e-06, + "loss": 1.0371, + "step": 47967 + }, + { + "epoch": 3.54, + "learning_rate": 7.5710808675698655e-06, + "loss": 1.0793, + "step": 47968 + }, + { + "epoch": 3.54, + "learning_rate": 7.570693942747514e-06, + "loss": 0.9527, + "step": 47969 + }, + { + "epoch": 3.54, + "learning_rate": 7.570307021790159e-06, + "loss": 1.0211, + "step": 47970 + }, + { + "epoch": 3.54, + "learning_rate": 7.569920104698405e-06, + "loss": 0.8413, + "step": 47971 + }, + { + "epoch": 3.54, + "learning_rate": 7.56953319147288e-06, + "loss": 0.931, + "step": 47972 + }, + { + "epoch": 3.54, + "learning_rate": 7.569146282114186e-06, + "loss": 1.0387, + "step": 47973 + }, + { + "epoch": 3.54, + "learning_rate": 7.5687593766229515e-06, + "loss": 0.9993, + "step": 47974 + }, + { + "epoch": 3.54, + "learning_rate": 7.568372474999781e-06, + "loss": 0.9731, + "step": 47975 + }, + { + "epoch": 3.54, + "learning_rate": 7.567985577245301e-06, + "loss": 0.9225, + "step": 47976 + }, + { + "epoch": 3.54, + "learning_rate": 7.567598683360119e-06, + "loss": 0.9255, + "step": 47977 + }, + { + "epoch": 3.55, + "learning_rate": 7.56721179334485e-06, + "loss": 0.9858, + "step": 47978 + }, + { + "epoch": 3.55, + "learning_rate": 7.566824907200114e-06, + "loss": 0.8777, + "step": 47979 + }, + { + "epoch": 3.55, + "learning_rate": 7.566438024926524e-06, + "loss": 1.118, + "step": 47980 + }, + { + "epoch": 3.55, + "learning_rate": 7.566051146524698e-06, + "loss": 0.9249, + "step": 47981 + }, + { + "epoch": 3.55, + "learning_rate": 7.5656642719952475e-06, + "loss": 1.0782, + "step": 47982 + }, + { + "epoch": 3.55, + "learning_rate": 7.565277401338793e-06, + "loss": 0.9416, + "step": 47983 + }, + { + "epoch": 3.55, + "learning_rate": 7.564890534555944e-06, + "loss": 1.0017, + "step": 47984 + }, + { + "epoch": 3.55, + "learning_rate": 7.564503671647322e-06, + "loss": 1.0046, + "step": 47985 + }, + { + "epoch": 3.55, + "learning_rate": 7.564116812613538e-06, + "loss": 0.9097, + "step": 47986 + }, + { + "epoch": 3.55, + "learning_rate": 7.56372995745521e-06, + "loss": 0.9508, + "step": 47987 + }, + { + "epoch": 3.55, + "learning_rate": 7.563343106172957e-06, + "loss": 1.072, + "step": 47988 + }, + { + "epoch": 3.55, + "learning_rate": 7.562956258767384e-06, + "loss": 1.0725, + "step": 47989 + }, + { + "epoch": 3.55, + "learning_rate": 7.562569415239117e-06, + "loss": 0.9716, + "step": 47990 + }, + { + "epoch": 3.55, + "learning_rate": 7.562182575588763e-06, + "loss": 1.0941, + "step": 47991 + }, + { + "epoch": 3.55, + "learning_rate": 7.561795739816946e-06, + "loss": 1.0611, + "step": 47992 + }, + { + "epoch": 3.55, + "learning_rate": 7.561408907924272e-06, + "loss": 1.0221, + "step": 47993 + }, + { + "epoch": 3.55, + "learning_rate": 7.561022079911366e-06, + "loss": 1.0154, + "step": 47994 + }, + { + "epoch": 3.55, + "learning_rate": 7.5606352557788355e-06, + "loss": 0.8982, + "step": 47995 + }, + { + "epoch": 3.55, + "learning_rate": 7.560248435527304e-06, + "loss": 1.0452, + "step": 47996 + }, + { + "epoch": 3.55, + "learning_rate": 7.5598616191573805e-06, + "loss": 1.0186, + "step": 47997 + }, + { + "epoch": 3.55, + "learning_rate": 7.55947480666968e-06, + "loss": 0.8349, + "step": 47998 + }, + { + "epoch": 3.55, + "learning_rate": 7.559087998064823e-06, + "loss": 1.0136, + "step": 47999 + }, + { + "epoch": 3.55, + "learning_rate": 7.55870119334342e-06, + "loss": 0.9529, + "step": 48000 + }, + { + "epoch": 3.55, + "learning_rate": 7.55831439250609e-06, + "loss": 1.0349, + "step": 48001 + }, + { + "epoch": 3.55, + "learning_rate": 7.557927595553445e-06, + "loss": 1.1373, + "step": 48002 + }, + { + "epoch": 3.55, + "learning_rate": 7.557540802486103e-06, + "loss": 1.0544, + "step": 48003 + }, + { + "epoch": 3.55, + "learning_rate": 7.557154013304678e-06, + "loss": 1.0397, + "step": 48004 + }, + { + "epoch": 3.55, + "learning_rate": 7.556767228009788e-06, + "loss": 0.9386, + "step": 48005 + }, + { + "epoch": 3.55, + "learning_rate": 7.556380446602046e-06, + "loss": 1.0973, + "step": 48006 + }, + { + "epoch": 3.55, + "learning_rate": 7.555993669082066e-06, + "loss": 0.836, + "step": 48007 + }, + { + "epoch": 3.55, + "learning_rate": 7.555606895450467e-06, + "loss": 1.0275, + "step": 48008 + }, + { + "epoch": 3.55, + "learning_rate": 7.5552201257078605e-06, + "loss": 0.9417, + "step": 48009 + }, + { + "epoch": 3.55, + "learning_rate": 7.554833359854865e-06, + "loss": 0.983, + "step": 48010 + }, + { + "epoch": 3.55, + "learning_rate": 7.554446597892093e-06, + "loss": 0.9862, + "step": 48011 + }, + { + "epoch": 3.55, + "learning_rate": 7.554059839820165e-06, + "loss": 0.9421, + "step": 48012 + }, + { + "epoch": 3.55, + "learning_rate": 7.55367308563969e-06, + "loss": 1.0483, + "step": 48013 + }, + { + "epoch": 3.55, + "learning_rate": 7.553286335351288e-06, + "loss": 1.0018, + "step": 48014 + }, + { + "epoch": 3.55, + "learning_rate": 7.552899588955573e-06, + "loss": 0.9591, + "step": 48015 + }, + { + "epoch": 3.55, + "learning_rate": 7.552512846453155e-06, + "loss": 0.9164, + "step": 48016 + }, + { + "epoch": 3.55, + "learning_rate": 7.5521261078446604e-06, + "loss": 0.9833, + "step": 48017 + }, + { + "epoch": 3.55, + "learning_rate": 7.551739373130692e-06, + "loss": 0.8822, + "step": 48018 + }, + { + "epoch": 3.55, + "learning_rate": 7.551352642311878e-06, + "loss": 0.942, + "step": 48019 + }, + { + "epoch": 3.55, + "learning_rate": 7.550965915388819e-06, + "loss": 0.9379, + "step": 48020 + }, + { + "epoch": 3.55, + "learning_rate": 7.550579192362145e-06, + "loss": 1.0113, + "step": 48021 + }, + { + "epoch": 3.55, + "learning_rate": 7.550192473232461e-06, + "loss": 0.9589, + "step": 48022 + }, + { + "epoch": 3.55, + "learning_rate": 7.549805758000387e-06, + "loss": 1.1578, + "step": 48023 + }, + { + "epoch": 3.55, + "learning_rate": 7.549419046666537e-06, + "loss": 1.0467, + "step": 48024 + }, + { + "epoch": 3.55, + "learning_rate": 7.549032339231524e-06, + "loss": 0.9043, + "step": 48025 + }, + { + "epoch": 3.55, + "learning_rate": 7.548645635695968e-06, + "loss": 0.9775, + "step": 48026 + }, + { + "epoch": 3.55, + "learning_rate": 7.548258936060479e-06, + "loss": 0.9883, + "step": 48027 + }, + { + "epoch": 3.55, + "learning_rate": 7.5478722403256785e-06, + "loss": 0.9609, + "step": 48028 + }, + { + "epoch": 3.55, + "learning_rate": 7.547485548492176e-06, + "loss": 1.1023, + "step": 48029 + }, + { + "epoch": 3.55, + "learning_rate": 7.54709886056059e-06, + "loss": 0.8489, + "step": 48030 + }, + { + "epoch": 3.55, + "learning_rate": 7.546712176531533e-06, + "loss": 1.0261, + "step": 48031 + }, + { + "epoch": 3.55, + "learning_rate": 7.546325496405623e-06, + "loss": 0.9769, + "step": 48032 + }, + { + "epoch": 3.55, + "learning_rate": 7.545938820183474e-06, + "loss": 1.1111, + "step": 48033 + }, + { + "epoch": 3.55, + "learning_rate": 7.545552147865701e-06, + "loss": 0.9983, + "step": 48034 + }, + { + "epoch": 3.55, + "learning_rate": 7.545165479452923e-06, + "loss": 1.0742, + "step": 48035 + }, + { + "epoch": 3.55, + "learning_rate": 7.544778814945744e-06, + "loss": 1.0457, + "step": 48036 + }, + { + "epoch": 3.55, + "learning_rate": 7.544392154344794e-06, + "loss": 1.034, + "step": 48037 + }, + { + "epoch": 3.55, + "learning_rate": 7.544005497650675e-06, + "loss": 1.028, + "step": 48038 + }, + { + "epoch": 3.55, + "learning_rate": 7.543618844864014e-06, + "loss": 1.0082, + "step": 48039 + }, + { + "epoch": 3.55, + "learning_rate": 7.543232195985414e-06, + "loss": 1.1251, + "step": 48040 + }, + { + "epoch": 3.55, + "learning_rate": 7.542845551015503e-06, + "loss": 1.0887, + "step": 48041 + }, + { + "epoch": 3.55, + "learning_rate": 7.5424589099548815e-06, + "loss": 1.0353, + "step": 48042 + }, + { + "epoch": 3.55, + "learning_rate": 7.542072272804182e-06, + "loss": 0.9672, + "step": 48043 + }, + { + "epoch": 3.55, + "learning_rate": 7.541685639564006e-06, + "loss": 0.9705, + "step": 48044 + }, + { + "epoch": 3.55, + "learning_rate": 7.541299010234972e-06, + "loss": 1.0323, + "step": 48045 + }, + { + "epoch": 3.55, + "learning_rate": 7.540912384817697e-06, + "loss": 0.9616, + "step": 48046 + }, + { + "epoch": 3.55, + "learning_rate": 7.540525763312794e-06, + "loss": 0.9767, + "step": 48047 + }, + { + "epoch": 3.55, + "learning_rate": 7.540139145720882e-06, + "loss": 1.0702, + "step": 48048 + }, + { + "epoch": 3.55, + "learning_rate": 7.539752532042571e-06, + "loss": 1.089, + "step": 48049 + }, + { + "epoch": 3.55, + "learning_rate": 7.539365922278479e-06, + "loss": 1.0193, + "step": 48050 + }, + { + "epoch": 3.55, + "learning_rate": 7.5389793164292205e-06, + "loss": 1.0271, + "step": 48051 + }, + { + "epoch": 3.55, + "learning_rate": 7.538592714495411e-06, + "loss": 0.913, + "step": 48052 + }, + { + "epoch": 3.55, + "learning_rate": 7.538206116477666e-06, + "loss": 0.9352, + "step": 48053 + }, + { + "epoch": 3.55, + "learning_rate": 7.537819522376598e-06, + "loss": 0.9654, + "step": 48054 + }, + { + "epoch": 3.55, + "learning_rate": 7.537432932192827e-06, + "loss": 0.8825, + "step": 48055 + }, + { + "epoch": 3.55, + "learning_rate": 7.537046345926961e-06, + "loss": 1.1411, + "step": 48056 + }, + { + "epoch": 3.55, + "learning_rate": 7.536659763579622e-06, + "loss": 0.9837, + "step": 48057 + }, + { + "epoch": 3.55, + "learning_rate": 7.5362731851514205e-06, + "loss": 1.0201, + "step": 48058 + }, + { + "epoch": 3.55, + "learning_rate": 7.535886610642975e-06, + "loss": 1.0212, + "step": 48059 + }, + { + "epoch": 3.55, + "learning_rate": 7.5355000400548965e-06, + "loss": 0.9615, + "step": 48060 + }, + { + "epoch": 3.55, + "learning_rate": 7.535113473387803e-06, + "loss": 0.9055, + "step": 48061 + }, + { + "epoch": 3.55, + "learning_rate": 7.534726910642312e-06, + "loss": 0.9783, + "step": 48062 + }, + { + "epoch": 3.55, + "learning_rate": 7.5343403518190296e-06, + "loss": 0.9366, + "step": 48063 + }, + { + "epoch": 3.55, + "learning_rate": 7.533953796918582e-06, + "loss": 0.946, + "step": 48064 + }, + { + "epoch": 3.55, + "learning_rate": 7.533567245941571e-06, + "loss": 1.0445, + "step": 48065 + }, + { + "epoch": 3.55, + "learning_rate": 7.533180698888627e-06, + "loss": 1.0007, + "step": 48066 + }, + { + "epoch": 3.55, + "learning_rate": 7.532794155760354e-06, + "loss": 1.0128, + "step": 48067 + }, + { + "epoch": 3.55, + "learning_rate": 7.53240761655737e-06, + "loss": 0.9743, + "step": 48068 + }, + { + "epoch": 3.55, + "learning_rate": 7.53202108128029e-06, + "loss": 0.9099, + "step": 48069 + }, + { + "epoch": 3.55, + "learning_rate": 7.5316345499297295e-06, + "loss": 0.8863, + "step": 48070 + }, + { + "epoch": 3.55, + "learning_rate": 7.5312480225063045e-06, + "loss": 1.0425, + "step": 48071 + }, + { + "epoch": 3.55, + "learning_rate": 7.5308614990106256e-06, + "loss": 0.9973, + "step": 48072 + }, + { + "epoch": 3.55, + "learning_rate": 7.530474979443313e-06, + "loss": 1.0628, + "step": 48073 + }, + { + "epoch": 3.55, + "learning_rate": 7.530088463804977e-06, + "loss": 0.9504, + "step": 48074 + }, + { + "epoch": 3.55, + "learning_rate": 7.529701952096238e-06, + "loss": 1.1471, + "step": 48075 + }, + { + "epoch": 3.55, + "learning_rate": 7.529315444317705e-06, + "loss": 0.9597, + "step": 48076 + }, + { + "epoch": 3.55, + "learning_rate": 7.528928940469999e-06, + "loss": 1.0626, + "step": 48077 + }, + { + "epoch": 3.55, + "learning_rate": 7.528542440553728e-06, + "loss": 0.8764, + "step": 48078 + }, + { + "epoch": 3.55, + "learning_rate": 7.528155944569513e-06, + "loss": 1.0629, + "step": 48079 + }, + { + "epoch": 3.55, + "learning_rate": 7.52776945251797e-06, + "loss": 1.0479, + "step": 48080 + }, + { + "epoch": 3.55, + "learning_rate": 7.527382964399704e-06, + "loss": 1.0101, + "step": 48081 + }, + { + "epoch": 3.55, + "learning_rate": 7.52699648021534e-06, + "loss": 0.9543, + "step": 48082 + }, + { + "epoch": 3.55, + "learning_rate": 7.526609999965485e-06, + "loss": 0.8427, + "step": 48083 + }, + { + "epoch": 3.55, + "learning_rate": 7.526223523650764e-06, + "loss": 1.1312, + "step": 48084 + }, + { + "epoch": 3.55, + "learning_rate": 7.5258370512717805e-06, + "loss": 0.9484, + "step": 48085 + }, + { + "epoch": 3.55, + "learning_rate": 7.5254505828291604e-06, + "loss": 1.0234, + "step": 48086 + }, + { + "epoch": 3.55, + "learning_rate": 7.525064118323506e-06, + "loss": 1.0463, + "step": 48087 + }, + { + "epoch": 3.55, + "learning_rate": 7.524677657755447e-06, + "loss": 0.8609, + "step": 48088 + }, + { + "epoch": 3.55, + "learning_rate": 7.5242912011255835e-06, + "loss": 0.9315, + "step": 48089 + }, + { + "epoch": 3.55, + "learning_rate": 7.523904748434543e-06, + "loss": 1.003, + "step": 48090 + }, + { + "epoch": 3.55, + "learning_rate": 7.523518299682933e-06, + "loss": 0.959, + "step": 48091 + }, + { + "epoch": 3.55, + "learning_rate": 7.523131854871368e-06, + "loss": 1.056, + "step": 48092 + }, + { + "epoch": 3.55, + "learning_rate": 7.5227454140004665e-06, + "loss": 1.0225, + "step": 48093 + }, + { + "epoch": 3.55, + "learning_rate": 7.522358977070839e-06, + "loss": 0.9543, + "step": 48094 + }, + { + "epoch": 3.55, + "learning_rate": 7.521972544083107e-06, + "loss": 1.0822, + "step": 48095 + }, + { + "epoch": 3.55, + "learning_rate": 7.5215861150378775e-06, + "loss": 1.0124, + "step": 48096 + }, + { + "epoch": 3.55, + "learning_rate": 7.521199689935771e-06, + "loss": 0.9576, + "step": 48097 + }, + { + "epoch": 3.55, + "learning_rate": 7.5208132687773995e-06, + "loss": 1.0487, + "step": 48098 + }, + { + "epoch": 3.55, + "learning_rate": 7.5204268515633795e-06, + "loss": 0.9815, + "step": 48099 + }, + { + "epoch": 3.55, + "learning_rate": 7.520040438294325e-06, + "loss": 0.9939, + "step": 48100 + }, + { + "epoch": 3.55, + "learning_rate": 7.519654028970849e-06, + "loss": 0.9111, + "step": 48101 + }, + { + "epoch": 3.55, + "learning_rate": 7.51926762359357e-06, + "loss": 1.0432, + "step": 48102 + }, + { + "epoch": 3.55, + "learning_rate": 7.518881222163099e-06, + "loss": 0.9994, + "step": 48103 + }, + { + "epoch": 3.55, + "learning_rate": 7.518494824680054e-06, + "loss": 0.901, + "step": 48104 + }, + { + "epoch": 3.55, + "learning_rate": 7.518108431145046e-06, + "loss": 0.9214, + "step": 48105 + }, + { + "epoch": 3.55, + "learning_rate": 7.517722041558695e-06, + "loss": 1.0088, + "step": 48106 + }, + { + "epoch": 3.55, + "learning_rate": 7.517335655921609e-06, + "loss": 0.9416, + "step": 48107 + }, + { + "epoch": 3.55, + "learning_rate": 7.516949274234409e-06, + "loss": 0.977, + "step": 48108 + }, + { + "epoch": 3.55, + "learning_rate": 7.516562896497711e-06, + "loss": 1.0848, + "step": 48109 + }, + { + "epoch": 3.55, + "learning_rate": 7.5161765227121195e-06, + "loss": 1.0313, + "step": 48110 + }, + { + "epoch": 3.55, + "learning_rate": 7.51579015287826e-06, + "loss": 1.0005, + "step": 48111 + }, + { + "epoch": 3.55, + "learning_rate": 7.515403786996739e-06, + "loss": 1.0309, + "step": 48112 + }, + { + "epoch": 3.56, + "learning_rate": 7.515017425068178e-06, + "loss": 0.966, + "step": 48113 + }, + { + "epoch": 3.56, + "learning_rate": 7.514631067093186e-06, + "loss": 0.9777, + "step": 48114 + }, + { + "epoch": 3.56, + "learning_rate": 7.514244713072383e-06, + "loss": 1.0127, + "step": 48115 + }, + { + "epoch": 3.56, + "learning_rate": 7.513858363006379e-06, + "loss": 0.9145, + "step": 48116 + }, + { + "epoch": 3.56, + "learning_rate": 7.513472016895792e-06, + "loss": 0.9714, + "step": 48117 + }, + { + "epoch": 3.56, + "learning_rate": 7.5130856747412364e-06, + "loss": 1.0315, + "step": 48118 + }, + { + "epoch": 3.56, + "learning_rate": 7.5126993365433236e-06, + "loss": 0.9305, + "step": 48119 + }, + { + "epoch": 3.56, + "learning_rate": 7.5123130023026716e-06, + "loss": 1.0351, + "step": 48120 + }, + { + "epoch": 3.56, + "learning_rate": 7.511926672019894e-06, + "loss": 1.0314, + "step": 48121 + }, + { + "epoch": 3.56, + "learning_rate": 7.511540345695605e-06, + "loss": 1.0166, + "step": 48122 + }, + { + "epoch": 3.56, + "learning_rate": 7.511154023330419e-06, + "loss": 0.9936, + "step": 48123 + }, + { + "epoch": 3.56, + "learning_rate": 7.510767704924954e-06, + "loss": 1.0431, + "step": 48124 + }, + { + "epoch": 3.56, + "learning_rate": 7.510381390479819e-06, + "loss": 0.9504, + "step": 48125 + }, + { + "epoch": 3.56, + "learning_rate": 7.509995079995634e-06, + "loss": 0.935, + "step": 48126 + }, + { + "epoch": 3.56, + "learning_rate": 7.509608773473014e-06, + "loss": 1.0281, + "step": 48127 + }, + { + "epoch": 3.56, + "learning_rate": 7.509222470912562e-06, + "loss": 1.0634, + "step": 48128 + }, + { + "epoch": 3.56, + "learning_rate": 7.50883617231491e-06, + "loss": 1.0457, + "step": 48129 + }, + { + "epoch": 3.56, + "learning_rate": 7.508449877680658e-06, + "loss": 0.9873, + "step": 48130 + }, + { + "epoch": 3.56, + "learning_rate": 7.508063587010432e-06, + "loss": 1.0414, + "step": 48131 + }, + { + "epoch": 3.56, + "learning_rate": 7.507677300304834e-06, + "loss": 0.9682, + "step": 48132 + }, + { + "epoch": 3.56, + "learning_rate": 7.507291017564494e-06, + "loss": 0.9326, + "step": 48133 + }, + { + "epoch": 3.56, + "learning_rate": 7.506904738790011e-06, + "loss": 0.9764, + "step": 48134 + }, + { + "epoch": 3.56, + "learning_rate": 7.506518463982013e-06, + "loss": 1.0844, + "step": 48135 + }, + { + "epoch": 3.56, + "learning_rate": 7.506132193141107e-06, + "loss": 1.0554, + "step": 48136 + }, + { + "epoch": 3.56, + "learning_rate": 7.505745926267906e-06, + "loss": 1.0733, + "step": 48137 + }, + { + "epoch": 3.56, + "learning_rate": 7.50535966336303e-06, + "loss": 0.9797, + "step": 48138 + }, + { + "epoch": 3.56, + "learning_rate": 7.504973404427089e-06, + "loss": 0.9306, + "step": 48139 + }, + { + "epoch": 3.56, + "learning_rate": 7.504587149460703e-06, + "loss": 1.015, + "step": 48140 + }, + { + "epoch": 3.56, + "learning_rate": 7.5042008984644796e-06, + "loss": 0.9639, + "step": 48141 + }, + { + "epoch": 3.56, + "learning_rate": 7.503814651439038e-06, + "loss": 0.96, + "step": 48142 + }, + { + "epoch": 3.56, + "learning_rate": 7.5034284083849915e-06, + "loss": 1.0113, + "step": 48143 + }, + { + "epoch": 3.56, + "learning_rate": 7.5030421693029565e-06, + "loss": 1.0798, + "step": 48144 + }, + { + "epoch": 3.56, + "learning_rate": 7.5026559341935434e-06, + "loss": 0.993, + "step": 48145 + }, + { + "epoch": 3.56, + "learning_rate": 7.50226970305737e-06, + "loss": 0.9857, + "step": 48146 + }, + { + "epoch": 3.56, + "learning_rate": 7.501883475895051e-06, + "loss": 0.953, + "step": 48147 + }, + { + "epoch": 3.56, + "learning_rate": 7.501497252707197e-06, + "loss": 0.9454, + "step": 48148 + }, + { + "epoch": 3.56, + "learning_rate": 7.501111033494427e-06, + "loss": 1.0465, + "step": 48149 + }, + { + "epoch": 3.56, + "learning_rate": 7.500724818257352e-06, + "loss": 0.9909, + "step": 48150 + }, + { + "epoch": 3.56, + "learning_rate": 7.50033860699659e-06, + "loss": 0.9962, + "step": 48151 + }, + { + "epoch": 3.56, + "learning_rate": 7.499952399712751e-06, + "loss": 1.0339, + "step": 48152 + }, + { + "epoch": 3.56, + "learning_rate": 7.499566196406455e-06, + "loss": 1.0104, + "step": 48153 + }, + { + "epoch": 3.56, + "learning_rate": 7.499179997078309e-06, + "loss": 0.8649, + "step": 48154 + }, + { + "epoch": 3.56, + "learning_rate": 7.498793801728939e-06, + "loss": 0.9512, + "step": 48155 + }, + { + "epoch": 3.56, + "learning_rate": 7.498407610358948e-06, + "loss": 0.999, + "step": 48156 + }, + { + "epoch": 3.56, + "learning_rate": 7.498021422968955e-06, + "loss": 0.9122, + "step": 48157 + }, + { + "epoch": 3.56, + "learning_rate": 7.497635239559575e-06, + "loss": 1.0599, + "step": 48158 + }, + { + "epoch": 3.56, + "learning_rate": 7.497249060131419e-06, + "loss": 1.0036, + "step": 48159 + }, + { + "epoch": 3.56, + "learning_rate": 7.4968628846851054e-06, + "loss": 1.0444, + "step": 48160 + }, + { + "epoch": 3.56, + "learning_rate": 7.4964767132212465e-06, + "loss": 0.9404, + "step": 48161 + }, + { + "epoch": 3.56, + "learning_rate": 7.4960905457404584e-06, + "loss": 1.0749, + "step": 48162 + }, + { + "epoch": 3.56, + "learning_rate": 7.495704382243354e-06, + "loss": 0.9826, + "step": 48163 + }, + { + "epoch": 3.56, + "learning_rate": 7.4953182227305476e-06, + "loss": 1.044, + "step": 48164 + }, + { + "epoch": 3.56, + "learning_rate": 7.494932067202656e-06, + "loss": 1.034, + "step": 48165 + }, + { + "epoch": 3.56, + "learning_rate": 7.49454591566029e-06, + "loss": 1.0708, + "step": 48166 + }, + { + "epoch": 3.56, + "learning_rate": 7.4941597681040655e-06, + "loss": 1.0345, + "step": 48167 + }, + { + "epoch": 3.56, + "learning_rate": 7.493773624534596e-06, + "loss": 1.0892, + "step": 48168 + }, + { + "epoch": 3.56, + "learning_rate": 7.493387484952499e-06, + "loss": 0.9137, + "step": 48169 + }, + { + "epoch": 3.56, + "learning_rate": 7.4930013493583845e-06, + "loss": 1.0489, + "step": 48170 + }, + { + "epoch": 3.56, + "learning_rate": 7.492615217752871e-06, + "loss": 1.0816, + "step": 48171 + }, + { + "epoch": 3.56, + "learning_rate": 7.492229090136569e-06, + "loss": 1.0165, + "step": 48172 + }, + { + "epoch": 3.56, + "learning_rate": 7.491842966510097e-06, + "loss": 1.0205, + "step": 48173 + }, + { + "epoch": 3.56, + "learning_rate": 7.491456846874067e-06, + "loss": 0.9366, + "step": 48174 + }, + { + "epoch": 3.56, + "learning_rate": 7.4910707312290885e-06, + "loss": 1.0111, + "step": 48175 + }, + { + "epoch": 3.56, + "learning_rate": 7.490684619575788e-06, + "loss": 0.9718, + "step": 48176 + }, + { + "epoch": 3.56, + "learning_rate": 7.4902985119147645e-06, + "loss": 0.9103, + "step": 48177 + }, + { + "epoch": 3.56, + "learning_rate": 7.489912408246647e-06, + "loss": 1.0644, + "step": 48178 + }, + { + "epoch": 3.56, + "learning_rate": 7.489526308572037e-06, + "loss": 1.0021, + "step": 48179 + }, + { + "epoch": 3.56, + "learning_rate": 7.4891402128915595e-06, + "loss": 0.9569, + "step": 48180 + }, + { + "epoch": 3.56, + "learning_rate": 7.488754121205819e-06, + "loss": 1.1189, + "step": 48181 + }, + { + "epoch": 3.56, + "learning_rate": 7.4883680335154404e-06, + "loss": 0.9738, + "step": 48182 + }, + { + "epoch": 3.56, + "learning_rate": 7.487981949821031e-06, + "loss": 1.0695, + "step": 48183 + }, + { + "epoch": 3.56, + "learning_rate": 7.487595870123204e-06, + "loss": 1.032, + "step": 48184 + }, + { + "epoch": 3.56, + "learning_rate": 7.487209794422579e-06, + "loss": 1.1513, + "step": 48185 + }, + { + "epoch": 3.56, + "learning_rate": 7.486823722719762e-06, + "loss": 0.9149, + "step": 48186 + }, + { + "epoch": 3.56, + "learning_rate": 7.486437655015379e-06, + "loss": 0.9437, + "step": 48187 + }, + { + "epoch": 3.56, + "learning_rate": 7.486051591310033e-06, + "loss": 1.0256, + "step": 48188 + }, + { + "epoch": 3.56, + "learning_rate": 7.485665531604345e-06, + "loss": 1.0142, + "step": 48189 + }, + { + "epoch": 3.56, + "learning_rate": 7.485279475898926e-06, + "loss": 0.9379, + "step": 48190 + }, + { + "epoch": 3.56, + "learning_rate": 7.484893424194392e-06, + "loss": 0.9081, + "step": 48191 + }, + { + "epoch": 3.56, + "learning_rate": 7.484507376491359e-06, + "loss": 1.0846, + "step": 48192 + }, + { + "epoch": 3.56, + "learning_rate": 7.484121332790435e-06, + "loss": 0.9416, + "step": 48193 + }, + { + "epoch": 3.56, + "learning_rate": 7.48373529309224e-06, + "loss": 1.0191, + "step": 48194 + }, + { + "epoch": 3.56, + "learning_rate": 7.483349257397384e-06, + "loss": 1.0788, + "step": 48195 + }, + { + "epoch": 3.56, + "learning_rate": 7.482963225706486e-06, + "loss": 0.9835, + "step": 48196 + }, + { + "epoch": 3.56, + "learning_rate": 7.482577198020156e-06, + "loss": 0.9735, + "step": 48197 + }, + { + "epoch": 3.56, + "learning_rate": 7.482191174339013e-06, + "loss": 1.0106, + "step": 48198 + }, + { + "epoch": 3.56, + "learning_rate": 7.481805154663661e-06, + "loss": 1.03, + "step": 48199 + }, + { + "epoch": 3.56, + "learning_rate": 7.481419138994727e-06, + "loss": 1.0649, + "step": 48200 + }, + { + "epoch": 3.56, + "learning_rate": 7.481033127332815e-06, + "loss": 0.9009, + "step": 48201 + }, + { + "epoch": 3.56, + "learning_rate": 7.480647119678548e-06, + "loss": 0.9457, + "step": 48202 + }, + { + "epoch": 3.56, + "learning_rate": 7.480261116032533e-06, + "loss": 1.0491, + "step": 48203 + }, + { + "epoch": 3.56, + "learning_rate": 7.479875116395385e-06, + "loss": 0.9612, + "step": 48204 + }, + { + "epoch": 3.56, + "learning_rate": 7.479489120767722e-06, + "loss": 0.9559, + "step": 48205 + }, + { + "epoch": 3.56, + "learning_rate": 7.4791031291501534e-06, + "loss": 1.0163, + "step": 48206 + }, + { + "epoch": 3.56, + "learning_rate": 7.4787171415432966e-06, + "loss": 0.9202, + "step": 48207 + }, + { + "epoch": 3.56, + "learning_rate": 7.478331157947765e-06, + "loss": 0.8187, + "step": 48208 + }, + { + "epoch": 3.56, + "learning_rate": 7.477945178364172e-06, + "loss": 0.9532, + "step": 48209 + }, + { + "epoch": 3.56, + "learning_rate": 7.477559202793132e-06, + "loss": 0.9966, + "step": 48210 + }, + { + "epoch": 3.56, + "learning_rate": 7.477173231235261e-06, + "loss": 1.0969, + "step": 48211 + }, + { + "epoch": 3.56, + "learning_rate": 7.476787263691171e-06, + "loss": 0.922, + "step": 48212 + }, + { + "epoch": 3.56, + "learning_rate": 7.4764013001614734e-06, + "loss": 1.0012, + "step": 48213 + }, + { + "epoch": 3.56, + "learning_rate": 7.476015340646789e-06, + "loss": 1.0445, + "step": 48214 + }, + { + "epoch": 3.56, + "learning_rate": 7.4756293851477245e-06, + "loss": 0.9164, + "step": 48215 + }, + { + "epoch": 3.56, + "learning_rate": 7.475243433664901e-06, + "loss": 0.892, + "step": 48216 + }, + { + "epoch": 3.56, + "learning_rate": 7.474857486198925e-06, + "loss": 1.0298, + "step": 48217 + }, + { + "epoch": 3.56, + "learning_rate": 7.474471542750419e-06, + "loss": 0.9732, + "step": 48218 + }, + { + "epoch": 3.56, + "learning_rate": 7.474085603319989e-06, + "loss": 0.9574, + "step": 48219 + }, + { + "epoch": 3.56, + "learning_rate": 7.4736996679082555e-06, + "loss": 0.9684, + "step": 48220 + }, + { + "epoch": 3.56, + "learning_rate": 7.47331373651583e-06, + "loss": 1.0343, + "step": 48221 + }, + { + "epoch": 3.56, + "learning_rate": 7.472927809143322e-06, + "loss": 1.0242, + "step": 48222 + }, + { + "epoch": 3.56, + "learning_rate": 7.472541885791356e-06, + "loss": 0.8933, + "step": 48223 + }, + { + "epoch": 3.56, + "learning_rate": 7.472155966460533e-06, + "loss": 0.9752, + "step": 48224 + }, + { + "epoch": 3.56, + "learning_rate": 7.471770051151479e-06, + "loss": 0.942, + "step": 48225 + }, + { + "epoch": 3.56, + "learning_rate": 7.471384139864796e-06, + "loss": 1.0403, + "step": 48226 + }, + { + "epoch": 3.56, + "learning_rate": 7.470998232601114e-06, + "loss": 1.0391, + "step": 48227 + }, + { + "epoch": 3.56, + "learning_rate": 7.4706123293610286e-06, + "loss": 1.0023, + "step": 48228 + }, + { + "epoch": 3.56, + "learning_rate": 7.470226430145171e-06, + "loss": 1.0513, + "step": 48229 + }, + { + "epoch": 3.56, + "learning_rate": 7.469840534954143e-06, + "loss": 1.0987, + "step": 48230 + }, + { + "epoch": 3.56, + "learning_rate": 7.469454643788561e-06, + "loss": 0.9261, + "step": 48231 + }, + { + "epoch": 3.56, + "learning_rate": 7.469068756649042e-06, + "loss": 1.0772, + "step": 48232 + }, + { + "epoch": 3.56, + "learning_rate": 7.4686828735361975e-06, + "loss": 0.9706, + "step": 48233 + }, + { + "epoch": 3.56, + "learning_rate": 7.4682969944506444e-06, + "loss": 0.911, + "step": 48234 + }, + { + "epoch": 3.56, + "learning_rate": 7.467911119392992e-06, + "loss": 0.9627, + "step": 48235 + }, + { + "epoch": 3.56, + "learning_rate": 7.4675252483638584e-06, + "loss": 0.9677, + "step": 48236 + }, + { + "epoch": 3.56, + "learning_rate": 7.4671393813638545e-06, + "loss": 0.9792, + "step": 48237 + }, + { + "epoch": 3.56, + "learning_rate": 7.466753518393598e-06, + "loss": 1.0082, + "step": 48238 + }, + { + "epoch": 3.56, + "learning_rate": 7.4663676594536995e-06, + "loss": 1.0588, + "step": 48239 + }, + { + "epoch": 3.56, + "learning_rate": 7.465981804544773e-06, + "loss": 0.9184, + "step": 48240 + }, + { + "epoch": 3.56, + "learning_rate": 7.465595953667434e-06, + "loss": 1.039, + "step": 48241 + }, + { + "epoch": 3.56, + "learning_rate": 7.465210106822294e-06, + "loss": 0.9767, + "step": 48242 + }, + { + "epoch": 3.56, + "learning_rate": 7.464824264009972e-06, + "loss": 0.9564, + "step": 48243 + }, + { + "epoch": 3.56, + "learning_rate": 7.464438425231073e-06, + "loss": 1.0504, + "step": 48244 + }, + { + "epoch": 3.56, + "learning_rate": 7.464052590486222e-06, + "loss": 0.9914, + "step": 48245 + }, + { + "epoch": 3.56, + "learning_rate": 7.463666759776021e-06, + "loss": 1.0052, + "step": 48246 + }, + { + "epoch": 3.56, + "learning_rate": 7.4632809331010954e-06, + "loss": 0.9444, + "step": 48247 + }, + { + "epoch": 3.57, + "learning_rate": 7.4628951104620476e-06, + "loss": 0.9925, + "step": 48248 + }, + { + "epoch": 3.57, + "learning_rate": 7.462509291859505e-06, + "loss": 0.9488, + "step": 48249 + }, + { + "epoch": 3.57, + "learning_rate": 7.46212347729407e-06, + "loss": 0.9768, + "step": 48250 + }, + { + "epoch": 3.57, + "learning_rate": 7.461737666766359e-06, + "loss": 0.9575, + "step": 48251 + }, + { + "epoch": 3.57, + "learning_rate": 7.461351860276989e-06, + "loss": 0.9937, + "step": 48252 + }, + { + "epoch": 3.57, + "learning_rate": 7.460966057826569e-06, + "loss": 1.0143, + "step": 48253 + }, + { + "epoch": 3.57, + "learning_rate": 7.460580259415719e-06, + "loss": 0.9887, + "step": 48254 + }, + { + "epoch": 3.57, + "learning_rate": 7.460194465045047e-06, + "loss": 0.8651, + "step": 48255 + }, + { + "epoch": 3.57, + "learning_rate": 7.459808674715171e-06, + "loss": 0.9681, + "step": 48256 + }, + { + "epoch": 3.57, + "learning_rate": 7.4594228884267014e-06, + "loss": 1.0179, + "step": 48257 + }, + { + "epoch": 3.57, + "learning_rate": 7.459037106180256e-06, + "loss": 0.9423, + "step": 48258 + }, + { + "epoch": 3.57, + "learning_rate": 7.458651327976445e-06, + "loss": 1.0289, + "step": 48259 + }, + { + "epoch": 3.57, + "learning_rate": 7.458265553815882e-06, + "loss": 0.9001, + "step": 48260 + }, + { + "epoch": 3.57, + "learning_rate": 7.4578797836991844e-06, + "loss": 1.0356, + "step": 48261 + }, + { + "epoch": 3.57, + "learning_rate": 7.45749401762696e-06, + "loss": 1.0239, + "step": 48262 + }, + { + "epoch": 3.57, + "learning_rate": 7.45710825559983e-06, + "loss": 0.9737, + "step": 48263 + }, + { + "epoch": 3.57, + "learning_rate": 7.4567224976184025e-06, + "loss": 1.0131, + "step": 48264 + }, + { + "epoch": 3.57, + "learning_rate": 7.456336743683295e-06, + "loss": 1.021, + "step": 48265 + }, + { + "epoch": 3.57, + "learning_rate": 7.455950993795117e-06, + "loss": 0.9841, + "step": 48266 + }, + { + "epoch": 3.57, + "learning_rate": 7.455565247954486e-06, + "loss": 0.9861, + "step": 48267 + }, + { + "epoch": 3.57, + "learning_rate": 7.455179506162018e-06, + "loss": 0.9636, + "step": 48268 + }, + { + "epoch": 3.57, + "learning_rate": 7.454793768418316e-06, + "loss": 0.9846, + "step": 48269 + }, + { + "epoch": 3.57, + "learning_rate": 7.454408034724006e-06, + "loss": 1.0569, + "step": 48270 + }, + { + "epoch": 3.57, + "learning_rate": 7.4540223050796914e-06, + "loss": 1.046, + "step": 48271 + }, + { + "epoch": 3.57, + "learning_rate": 7.453636579485997e-06, + "loss": 1.0207, + "step": 48272 + }, + { + "epoch": 3.57, + "learning_rate": 7.453250857943523e-06, + "loss": 0.9291, + "step": 48273 + }, + { + "epoch": 3.57, + "learning_rate": 7.452865140452897e-06, + "loss": 1.0241, + "step": 48274 + }, + { + "epoch": 3.57, + "learning_rate": 7.452479427014721e-06, + "loss": 1.05, + "step": 48275 + }, + { + "epoch": 3.57, + "learning_rate": 7.452093717629618e-06, + "loss": 1.0876, + "step": 48276 + }, + { + "epoch": 3.57, + "learning_rate": 7.451708012298198e-06, + "loss": 1.0373, + "step": 48277 + }, + { + "epoch": 3.57, + "learning_rate": 7.451322311021069e-06, + "loss": 0.9866, + "step": 48278 + }, + { + "epoch": 3.57, + "learning_rate": 7.450936613798854e-06, + "loss": 0.9103, + "step": 48279 + }, + { + "epoch": 3.57, + "learning_rate": 7.45055092063216e-06, + "loss": 0.8622, + "step": 48280 + }, + { + "epoch": 3.57, + "learning_rate": 7.4501652315216055e-06, + "loss": 1.0662, + "step": 48281 + }, + { + "epoch": 3.57, + "learning_rate": 7.449779546467799e-06, + "loss": 0.9743, + "step": 48282 + }, + { + "epoch": 3.57, + "learning_rate": 7.449393865471358e-06, + "loss": 0.9381, + "step": 48283 + }, + { + "epoch": 3.57, + "learning_rate": 7.449008188532894e-06, + "loss": 0.8891, + "step": 48284 + }, + { + "epoch": 3.57, + "learning_rate": 7.448622515653023e-06, + "loss": 1.0387, + "step": 48285 + }, + { + "epoch": 3.57, + "learning_rate": 7.448236846832357e-06, + "loss": 0.9658, + "step": 48286 + }, + { + "epoch": 3.57, + "learning_rate": 7.447851182071509e-06, + "loss": 0.9207, + "step": 48287 + }, + { + "epoch": 3.57, + "learning_rate": 7.447465521371095e-06, + "loss": 1.039, + "step": 48288 + }, + { + "epoch": 3.57, + "learning_rate": 7.447079864731722e-06, + "loss": 1.0119, + "step": 48289 + }, + { + "epoch": 3.57, + "learning_rate": 7.446694212154015e-06, + "loss": 1.036, + "step": 48290 + }, + { + "epoch": 3.57, + "learning_rate": 7.446308563638573e-06, + "loss": 1.0469, + "step": 48291 + }, + { + "epoch": 3.57, + "learning_rate": 7.445922919186026e-06, + "loss": 1.003, + "step": 48292 + }, + { + "epoch": 3.57, + "learning_rate": 7.445537278796972e-06, + "loss": 0.9812, + "step": 48293 + }, + { + "epoch": 3.57, + "learning_rate": 7.445151642472037e-06, + "loss": 0.9573, + "step": 48294 + }, + { + "epoch": 3.57, + "learning_rate": 7.444766010211828e-06, + "loss": 1.1022, + "step": 48295 + }, + { + "epoch": 3.57, + "learning_rate": 7.444380382016958e-06, + "loss": 0.9624, + "step": 48296 + }, + { + "epoch": 3.57, + "learning_rate": 7.443994757888043e-06, + "loss": 0.9269, + "step": 48297 + }, + { + "epoch": 3.57, + "learning_rate": 7.443609137825695e-06, + "loss": 0.9215, + "step": 48298 + }, + { + "epoch": 3.57, + "learning_rate": 7.44322352183053e-06, + "loss": 0.9359, + "step": 48299 + }, + { + "epoch": 3.57, + "learning_rate": 7.442837909903157e-06, + "loss": 1.0915, + "step": 48300 + }, + { + "epoch": 3.57, + "learning_rate": 7.442452302044194e-06, + "loss": 0.9588, + "step": 48301 + }, + { + "epoch": 3.57, + "learning_rate": 7.442066698254253e-06, + "loss": 1.0795, + "step": 48302 + }, + { + "epoch": 3.57, + "learning_rate": 7.441681098533947e-06, + "loss": 0.8821, + "step": 48303 + }, + { + "epoch": 3.57, + "learning_rate": 7.441295502883889e-06, + "loss": 0.9535, + "step": 48304 + }, + { + "epoch": 3.57, + "learning_rate": 7.440909911304695e-06, + "loss": 0.977, + "step": 48305 + }, + { + "epoch": 3.57, + "learning_rate": 7.440524323796976e-06, + "loss": 1.0577, + "step": 48306 + }, + { + "epoch": 3.57, + "learning_rate": 7.440138740361346e-06, + "loss": 1.0843, + "step": 48307 + }, + { + "epoch": 3.57, + "learning_rate": 7.43975316099842e-06, + "loss": 1.0246, + "step": 48308 + }, + { + "epoch": 3.57, + "learning_rate": 7.439367585708806e-06, + "loss": 1.0424, + "step": 48309 + }, + { + "epoch": 3.57, + "learning_rate": 7.438982014493126e-06, + "loss": 1.0732, + "step": 48310 + }, + { + "epoch": 3.57, + "learning_rate": 7.438596447351986e-06, + "loss": 0.9794, + "step": 48311 + }, + { + "epoch": 3.57, + "learning_rate": 7.4382108842860035e-06, + "loss": 1.0186, + "step": 48312 + }, + { + "epoch": 3.57, + "learning_rate": 7.437825325295791e-06, + "loss": 1.018, + "step": 48313 + }, + { + "epoch": 3.57, + "learning_rate": 7.437439770381962e-06, + "loss": 0.8748, + "step": 48314 + }, + { + "epoch": 3.57, + "learning_rate": 7.437054219545132e-06, + "loss": 1.1262, + "step": 48315 + }, + { + "epoch": 3.57, + "learning_rate": 7.436668672785907e-06, + "loss": 0.9892, + "step": 48316 + }, + { + "epoch": 3.57, + "learning_rate": 7.436283130104911e-06, + "loss": 0.9904, + "step": 48317 + }, + { + "epoch": 3.57, + "learning_rate": 7.435897591502745e-06, + "loss": 0.96, + "step": 48318 + }, + { + "epoch": 3.57, + "learning_rate": 7.4355120569800365e-06, + "loss": 1.0395, + "step": 48319 + }, + { + "epoch": 3.57, + "learning_rate": 7.4351265265373866e-06, + "loss": 0.9481, + "step": 48320 + }, + { + "epoch": 3.57, + "learning_rate": 7.434741000175416e-06, + "loss": 1.1031, + "step": 48321 + }, + { + "epoch": 3.57, + "learning_rate": 7.434355477894733e-06, + "loss": 1.1325, + "step": 48322 + }, + { + "epoch": 3.57, + "learning_rate": 7.433969959695957e-06, + "loss": 0.9563, + "step": 48323 + }, + { + "epoch": 3.57, + "learning_rate": 7.433584445579698e-06, + "loss": 1.0848, + "step": 48324 + }, + { + "epoch": 3.57, + "learning_rate": 7.433198935546567e-06, + "loss": 0.9998, + "step": 48325 + }, + { + "epoch": 3.57, + "learning_rate": 7.4328134295971825e-06, + "loss": 1.1102, + "step": 48326 + }, + { + "epoch": 3.57, + "learning_rate": 7.432427927732152e-06, + "loss": 1.0815, + "step": 48327 + }, + { + "epoch": 3.57, + "learning_rate": 7.4320424299520945e-06, + "loss": 0.989, + "step": 48328 + }, + { + "epoch": 3.57, + "learning_rate": 7.431656936257618e-06, + "loss": 0.9908, + "step": 48329 + }, + { + "epoch": 3.57, + "learning_rate": 7.431271446649342e-06, + "loss": 0.9181, + "step": 48330 + }, + { + "epoch": 3.57, + "learning_rate": 7.430885961127874e-06, + "loss": 1.0552, + "step": 48331 + }, + { + "epoch": 3.57, + "learning_rate": 7.430500479693831e-06, + "loss": 1.1857, + "step": 48332 + }, + { + "epoch": 3.57, + "learning_rate": 7.430115002347827e-06, + "loss": 1.0332, + "step": 48333 + }, + { + "epoch": 3.57, + "learning_rate": 7.429729529090468e-06, + "loss": 1.1182, + "step": 48334 + }, + { + "epoch": 3.57, + "learning_rate": 7.429344059922377e-06, + "loss": 0.9894, + "step": 48335 + }, + { + "epoch": 3.57, + "learning_rate": 7.428958594844158e-06, + "loss": 1.0056, + "step": 48336 + }, + { + "epoch": 3.57, + "learning_rate": 7.428573133856436e-06, + "loss": 1.0718, + "step": 48337 + }, + { + "epoch": 3.57, + "learning_rate": 7.4281876769598095e-06, + "loss": 1.0568, + "step": 48338 + }, + { + "epoch": 3.57, + "learning_rate": 7.427802224154906e-06, + "loss": 0.8896, + "step": 48339 + }, + { + "epoch": 3.57, + "learning_rate": 7.427416775442326e-06, + "loss": 0.9875, + "step": 48340 + }, + { + "epoch": 3.57, + "learning_rate": 7.427031330822697e-06, + "loss": 1.0354, + "step": 48341 + }, + { + "epoch": 3.57, + "learning_rate": 7.426645890296622e-06, + "loss": 0.9767, + "step": 48342 + }, + { + "epoch": 3.57, + "learning_rate": 7.426260453864713e-06, + "loss": 0.9694, + "step": 48343 + }, + { + "epoch": 3.57, + "learning_rate": 7.42587502152759e-06, + "loss": 0.9409, + "step": 48344 + }, + { + "epoch": 3.57, + "learning_rate": 7.42548959328586e-06, + "loss": 1.0498, + "step": 48345 + }, + { + "epoch": 3.57, + "learning_rate": 7.425104169140143e-06, + "loss": 0.9933, + "step": 48346 + }, + { + "epoch": 3.57, + "learning_rate": 7.424718749091045e-06, + "loss": 1.1089, + "step": 48347 + }, + { + "epoch": 3.57, + "learning_rate": 7.424333333139184e-06, + "loss": 1.0004, + "step": 48348 + }, + { + "epoch": 3.57, + "learning_rate": 7.4239479212851704e-06, + "loss": 1.1037, + "step": 48349 + }, + { + "epoch": 3.57, + "learning_rate": 7.423562513529622e-06, + "loss": 1.0792, + "step": 48350 + }, + { + "epoch": 3.57, + "learning_rate": 7.423177109873147e-06, + "loss": 0.9126, + "step": 48351 + }, + { + "epoch": 3.57, + "learning_rate": 7.422791710316359e-06, + "loss": 0.9723, + "step": 48352 + }, + { + "epoch": 3.57, + "learning_rate": 7.422406314859876e-06, + "loss": 0.8932, + "step": 48353 + }, + { + "epoch": 3.57, + "learning_rate": 7.422020923504304e-06, + "loss": 1.0294, + "step": 48354 + }, + { + "epoch": 3.57, + "learning_rate": 7.421635536250262e-06, + "loss": 1.0633, + "step": 48355 + }, + { + "epoch": 3.57, + "learning_rate": 7.4212501530983595e-06, + "loss": 0.9241, + "step": 48356 + }, + { + "epoch": 3.57, + "learning_rate": 7.4208647740492125e-06, + "loss": 0.9023, + "step": 48357 + }, + { + "epoch": 3.57, + "learning_rate": 7.420479399103431e-06, + "loss": 1.1793, + "step": 48358 + }, + { + "epoch": 3.57, + "learning_rate": 7.420094028261633e-06, + "loss": 1.0097, + "step": 48359 + }, + { + "epoch": 3.57, + "learning_rate": 7.4197086615244255e-06, + "loss": 1.0693, + "step": 48360 + }, + { + "epoch": 3.57, + "learning_rate": 7.419323298892426e-06, + "loss": 1.1533, + "step": 48361 + }, + { + "epoch": 3.57, + "learning_rate": 7.418937940366249e-06, + "loss": 1.0461, + "step": 48362 + }, + { + "epoch": 3.57, + "learning_rate": 7.418552585946498e-06, + "loss": 0.9616, + "step": 48363 + }, + { + "epoch": 3.57, + "learning_rate": 7.418167235633799e-06, + "loss": 1.0156, + "step": 48364 + }, + { + "epoch": 3.57, + "learning_rate": 7.417781889428756e-06, + "loss": 0.9558, + "step": 48365 + }, + { + "epoch": 3.57, + "learning_rate": 7.417396547331987e-06, + "loss": 0.9918, + "step": 48366 + }, + { + "epoch": 3.57, + "learning_rate": 7.4170112093441e-06, + "loss": 1.0232, + "step": 48367 + }, + { + "epoch": 3.57, + "learning_rate": 7.416625875465714e-06, + "loss": 1.0182, + "step": 48368 + }, + { + "epoch": 3.57, + "learning_rate": 7.416240545697437e-06, + "loss": 1.0241, + "step": 48369 + }, + { + "epoch": 3.57, + "learning_rate": 7.415855220039887e-06, + "loss": 0.896, + "step": 48370 + }, + { + "epoch": 3.57, + "learning_rate": 7.415469898493672e-06, + "loss": 0.9742, + "step": 48371 + }, + { + "epoch": 3.57, + "learning_rate": 7.415084581059409e-06, + "loss": 0.9747, + "step": 48372 + }, + { + "epoch": 3.57, + "learning_rate": 7.4146992677377084e-06, + "loss": 1.0476, + "step": 48373 + }, + { + "epoch": 3.57, + "learning_rate": 7.414313958529184e-06, + "loss": 0.9676, + "step": 48374 + }, + { + "epoch": 3.57, + "learning_rate": 7.41392865343445e-06, + "loss": 1.0056, + "step": 48375 + }, + { + "epoch": 3.57, + "learning_rate": 7.413543352454118e-06, + "loss": 0.9402, + "step": 48376 + }, + { + "epoch": 3.57, + "learning_rate": 7.413158055588803e-06, + "loss": 0.8625, + "step": 48377 + }, + { + "epoch": 3.57, + "learning_rate": 7.412772762839114e-06, + "loss": 1.0205, + "step": 48378 + }, + { + "epoch": 3.57, + "learning_rate": 7.4123874742056685e-06, + "loss": 1.0035, + "step": 48379 + }, + { + "epoch": 3.57, + "learning_rate": 7.4120021896890805e-06, + "loss": 0.9926, + "step": 48380 + }, + { + "epoch": 3.57, + "learning_rate": 7.411616909289953e-06, + "loss": 1.1255, + "step": 48381 + }, + { + "epoch": 3.57, + "learning_rate": 7.411231633008912e-06, + "loss": 0.9715, + "step": 48382 + }, + { + "epoch": 3.57, + "learning_rate": 7.410846360846558e-06, + "loss": 0.9714, + "step": 48383 + }, + { + "epoch": 3.58, + "learning_rate": 7.410461092803518e-06, + "loss": 1.0679, + "step": 48384 + }, + { + "epoch": 3.58, + "learning_rate": 7.4100758288803895e-06, + "loss": 1.0262, + "step": 48385 + }, + { + "epoch": 3.58, + "learning_rate": 7.409690569077801e-06, + "loss": 1.0013, + "step": 48386 + }, + { + "epoch": 3.58, + "learning_rate": 7.40930531339635e-06, + "loss": 0.9606, + "step": 48387 + }, + { + "epoch": 3.58, + "learning_rate": 7.408920061836664e-06, + "loss": 1.0582, + "step": 48388 + }, + { + "epoch": 3.58, + "learning_rate": 7.408534814399348e-06, + "loss": 0.9836, + "step": 48389 + }, + { + "epoch": 3.58, + "learning_rate": 7.408149571085012e-06, + "loss": 0.8589, + "step": 48390 + }, + { + "epoch": 3.58, + "learning_rate": 7.407764331894276e-06, + "loss": 0.9605, + "step": 48391 + }, + { + "epoch": 3.58, + "learning_rate": 7.407379096827749e-06, + "loss": 0.9272, + "step": 48392 + }, + { + "epoch": 3.58, + "learning_rate": 7.406993865886045e-06, + "loss": 0.9893, + "step": 48393 + }, + { + "epoch": 3.58, + "learning_rate": 7.406608639069777e-06, + "loss": 0.9324, + "step": 48394 + }, + { + "epoch": 3.58, + "learning_rate": 7.406223416379557e-06, + "loss": 0.9815, + "step": 48395 + }, + { + "epoch": 3.58, + "learning_rate": 7.4058381978159975e-06, + "loss": 1.0363, + "step": 48396 + }, + { + "epoch": 3.58, + "learning_rate": 7.405452983379714e-06, + "loss": 0.9563, + "step": 48397 + }, + { + "epoch": 3.58, + "learning_rate": 7.405067773071318e-06, + "loss": 1.0433, + "step": 48398 + }, + { + "epoch": 3.58, + "learning_rate": 7.40468256689142e-06, + "loss": 0.9924, + "step": 48399 + }, + { + "epoch": 3.58, + "learning_rate": 7.404297364840637e-06, + "loss": 0.9556, + "step": 48400 + }, + { + "epoch": 3.58, + "learning_rate": 7.403912166919578e-06, + "loss": 1.0746, + "step": 48401 + }, + { + "epoch": 3.58, + "learning_rate": 7.403526973128858e-06, + "loss": 0.9977, + "step": 48402 + }, + { + "epoch": 3.58, + "learning_rate": 7.40314178346909e-06, + "loss": 1.0055, + "step": 48403 + }, + { + "epoch": 3.58, + "learning_rate": 7.4027565979408856e-06, + "loss": 1.1206, + "step": 48404 + }, + { + "epoch": 3.58, + "learning_rate": 7.402371416544858e-06, + "loss": 1.0341, + "step": 48405 + }, + { + "epoch": 3.58, + "learning_rate": 7.4019862392816226e-06, + "loss": 1.0643, + "step": 48406 + }, + { + "epoch": 3.58, + "learning_rate": 7.4016010661517915e-06, + "loss": 0.9607, + "step": 48407 + }, + { + "epoch": 3.58, + "learning_rate": 7.401215897155969e-06, + "loss": 0.9615, + "step": 48408 + }, + { + "epoch": 3.58, + "learning_rate": 7.4008307322947815e-06, + "loss": 0.9581, + "step": 48409 + }, + { + "epoch": 3.58, + "learning_rate": 7.400445571568831e-06, + "loss": 1.0394, + "step": 48410 + }, + { + "epoch": 3.58, + "learning_rate": 7.400060414978737e-06, + "loss": 0.8782, + "step": 48411 + }, + { + "epoch": 3.58, + "learning_rate": 7.399675262525108e-06, + "loss": 1.0276, + "step": 48412 + }, + { + "epoch": 3.58, + "learning_rate": 7.399290114208559e-06, + "loss": 0.9033, + "step": 48413 + }, + { + "epoch": 3.58, + "learning_rate": 7.398904970029701e-06, + "loss": 1.0169, + "step": 48414 + }, + { + "epoch": 3.58, + "learning_rate": 7.398519829989149e-06, + "loss": 0.9973, + "step": 48415 + }, + { + "epoch": 3.58, + "learning_rate": 7.398134694087513e-06, + "loss": 1.0313, + "step": 48416 + }, + { + "epoch": 3.58, + "learning_rate": 7.397749562325412e-06, + "loss": 1.0265, + "step": 48417 + }, + { + "epoch": 3.58, + "learning_rate": 7.397364434703452e-06, + "loss": 0.9531, + "step": 48418 + }, + { + "epoch": 3.58, + "learning_rate": 7.396979311222245e-06, + "loss": 0.9908, + "step": 48419 + }, + { + "epoch": 3.58, + "learning_rate": 7.39659419188241e-06, + "loss": 1.0301, + "step": 48420 + }, + { + "epoch": 3.58, + "learning_rate": 7.396209076684555e-06, + "loss": 0.9911, + "step": 48421 + }, + { + "epoch": 3.58, + "learning_rate": 7.3958239656292955e-06, + "loss": 0.9382, + "step": 48422 + }, + { + "epoch": 3.58, + "learning_rate": 7.395438858717241e-06, + "loss": 0.9752, + "step": 48423 + }, + { + "epoch": 3.58, + "learning_rate": 7.395053755949006e-06, + "loss": 1.0296, + "step": 48424 + }, + { + "epoch": 3.58, + "learning_rate": 7.394668657325204e-06, + "loss": 0.8993, + "step": 48425 + }, + { + "epoch": 3.58, + "learning_rate": 7.394283562846447e-06, + "loss": 0.9481, + "step": 48426 + }, + { + "epoch": 3.58, + "learning_rate": 7.393898472513351e-06, + "loss": 0.8856, + "step": 48427 + }, + { + "epoch": 3.58, + "learning_rate": 7.393513386326518e-06, + "loss": 1.0397, + "step": 48428 + }, + { + "epoch": 3.58, + "learning_rate": 7.3931283042865756e-06, + "loss": 0.9917, + "step": 48429 + }, + { + "epoch": 3.58, + "learning_rate": 7.392743226394121e-06, + "loss": 1.0291, + "step": 48430 + }, + { + "epoch": 3.58, + "learning_rate": 7.392358152649781e-06, + "loss": 0.9559, + "step": 48431 + }, + { + "epoch": 3.58, + "learning_rate": 7.3919730830541555e-06, + "loss": 1.0105, + "step": 48432 + }, + { + "epoch": 3.58, + "learning_rate": 7.391588017607871e-06, + "loss": 0.9669, + "step": 48433 + }, + { + "epoch": 3.58, + "learning_rate": 7.391202956311526e-06, + "loss": 0.9357, + "step": 48434 + }, + { + "epoch": 3.58, + "learning_rate": 7.390817899165745e-06, + "loss": 0.9754, + "step": 48435 + }, + { + "epoch": 3.58, + "learning_rate": 7.390432846171135e-06, + "loss": 1.0774, + "step": 48436 + }, + { + "epoch": 3.58, + "learning_rate": 7.390047797328306e-06, + "loss": 1.061, + "step": 48437 + }, + { + "epoch": 3.58, + "learning_rate": 7.389662752637876e-06, + "loss": 0.9248, + "step": 48438 + }, + { + "epoch": 3.58, + "learning_rate": 7.389277712100452e-06, + "loss": 1.0042, + "step": 48439 + }, + { + "epoch": 3.58, + "learning_rate": 7.388892675716652e-06, + "loss": 1.0543, + "step": 48440 + }, + { + "epoch": 3.58, + "learning_rate": 7.388507643487086e-06, + "loss": 1.0018, + "step": 48441 + }, + { + "epoch": 3.58, + "learning_rate": 7.388122615412368e-06, + "loss": 1.0063, + "step": 48442 + }, + { + "epoch": 3.58, + "learning_rate": 7.387737591493107e-06, + "loss": 1.1146, + "step": 48443 + }, + { + "epoch": 3.58, + "learning_rate": 7.387352571729922e-06, + "loss": 0.8682, + "step": 48444 + }, + { + "epoch": 3.58, + "learning_rate": 7.38696755612342e-06, + "loss": 1.0215, + "step": 48445 + }, + { + "epoch": 3.58, + "learning_rate": 7.386582544674214e-06, + "loss": 0.9352, + "step": 48446 + }, + { + "epoch": 3.58, + "learning_rate": 7.3861975373829186e-06, + "loss": 1.0286, + "step": 48447 + }, + { + "epoch": 3.58, + "learning_rate": 7.3858125342501445e-06, + "loss": 1.0279, + "step": 48448 + }, + { + "epoch": 3.58, + "learning_rate": 7.385427535276507e-06, + "loss": 1.0585, + "step": 48449 + }, + { + "epoch": 3.58, + "learning_rate": 7.385042540462615e-06, + "loss": 1.0371, + "step": 48450 + }, + { + "epoch": 3.58, + "learning_rate": 7.384657549809084e-06, + "loss": 0.9905, + "step": 48451 + }, + { + "epoch": 3.58, + "learning_rate": 7.384272563316525e-06, + "loss": 1.0516, + "step": 48452 + }, + { + "epoch": 3.58, + "learning_rate": 7.383887580985553e-06, + "loss": 1.0204, + "step": 48453 + }, + { + "epoch": 3.58, + "learning_rate": 7.383502602816779e-06, + "loss": 1.0565, + "step": 48454 + }, + { + "epoch": 3.58, + "learning_rate": 7.383117628810812e-06, + "loss": 0.9785, + "step": 48455 + }, + { + "epoch": 3.58, + "learning_rate": 7.382732658968268e-06, + "loss": 1.1898, + "step": 48456 + }, + { + "epoch": 3.58, + "learning_rate": 7.382347693289758e-06, + "loss": 1.0485, + "step": 48457 + }, + { + "epoch": 3.58, + "learning_rate": 7.381962731775898e-06, + "loss": 0.979, + "step": 48458 + }, + { + "epoch": 3.58, + "learning_rate": 7.381577774427294e-06, + "loss": 0.9832, + "step": 48459 + }, + { + "epoch": 3.58, + "learning_rate": 7.3811928212445645e-06, + "loss": 0.9614, + "step": 48460 + }, + { + "epoch": 3.58, + "learning_rate": 7.38080787222832e-06, + "loss": 1.0058, + "step": 48461 + }, + { + "epoch": 3.58, + "learning_rate": 7.380422927379173e-06, + "loss": 1.0751, + "step": 48462 + }, + { + "epoch": 3.58, + "learning_rate": 7.380037986697735e-06, + "loss": 1.0737, + "step": 48463 + }, + { + "epoch": 3.58, + "learning_rate": 7.379653050184619e-06, + "loss": 1.0465, + "step": 48464 + }, + { + "epoch": 3.58, + "learning_rate": 7.379268117840439e-06, + "loss": 1.0132, + "step": 48465 + }, + { + "epoch": 3.58, + "learning_rate": 7.378883189665802e-06, + "loss": 1.1471, + "step": 48466 + }, + { + "epoch": 3.58, + "learning_rate": 7.378498265661329e-06, + "loss": 1.0078, + "step": 48467 + }, + { + "epoch": 3.58, + "learning_rate": 7.378113345827624e-06, + "loss": 0.9641, + "step": 48468 + }, + { + "epoch": 3.58, + "learning_rate": 7.377728430165305e-06, + "loss": 1.0381, + "step": 48469 + }, + { + "epoch": 3.58, + "learning_rate": 7.377343518674982e-06, + "loss": 0.904, + "step": 48470 + }, + { + "epoch": 3.58, + "learning_rate": 7.376958611357269e-06, + "loss": 0.9417, + "step": 48471 + }, + { + "epoch": 3.58, + "learning_rate": 7.376573708212775e-06, + "loss": 1.0219, + "step": 48472 + }, + { + "epoch": 3.58, + "learning_rate": 7.376188809242118e-06, + "loss": 0.9577, + "step": 48473 + }, + { + "epoch": 3.58, + "learning_rate": 7.375803914445908e-06, + "loss": 0.9337, + "step": 48474 + }, + { + "epoch": 3.58, + "learning_rate": 7.375419023824751e-06, + "loss": 1.0246, + "step": 48475 + }, + { + "epoch": 3.58, + "learning_rate": 7.375034137379271e-06, + "loss": 0.8867, + "step": 48476 + }, + { + "epoch": 3.58, + "learning_rate": 7.374649255110067e-06, + "loss": 0.9483, + "step": 48477 + }, + { + "epoch": 3.58, + "learning_rate": 7.3742643770177645e-06, + "loss": 1.0948, + "step": 48478 + }, + { + "epoch": 3.58, + "learning_rate": 7.373879503102965e-06, + "loss": 0.9835, + "step": 48479 + }, + { + "epoch": 3.58, + "learning_rate": 7.373494633366292e-06, + "loss": 0.958, + "step": 48480 + }, + { + "epoch": 3.58, + "learning_rate": 7.373109767808343e-06, + "loss": 0.9886, + "step": 48481 + }, + { + "epoch": 3.58, + "learning_rate": 7.3727249064297465e-06, + "loss": 1.0356, + "step": 48482 + }, + { + "epoch": 3.58, + "learning_rate": 7.372340049231105e-06, + "loss": 1.073, + "step": 48483 + }, + { + "epoch": 3.58, + "learning_rate": 7.37195519621303e-06, + "loss": 0.9502, + "step": 48484 + }, + { + "epoch": 3.58, + "learning_rate": 7.37157034737614e-06, + "loss": 0.928, + "step": 48485 + }, + { + "epoch": 3.58, + "learning_rate": 7.371185502721041e-06, + "loss": 1.0439, + "step": 48486 + }, + { + "epoch": 3.58, + "learning_rate": 7.370800662248351e-06, + "loss": 1.0861, + "step": 48487 + }, + { + "epoch": 3.58, + "learning_rate": 7.370415825958677e-06, + "loss": 1.1366, + "step": 48488 + }, + { + "epoch": 3.58, + "learning_rate": 7.370030993852636e-06, + "loss": 0.9579, + "step": 48489 + }, + { + "epoch": 3.58, + "learning_rate": 7.369646165930836e-06, + "loss": 0.9566, + "step": 48490 + }, + { + "epoch": 3.58, + "learning_rate": 7.369261342193893e-06, + "loss": 0.9394, + "step": 48491 + }, + { + "epoch": 3.58, + "learning_rate": 7.368876522642419e-06, + "loss": 0.8719, + "step": 48492 + }, + { + "epoch": 3.58, + "learning_rate": 7.368491707277022e-06, + "loss": 1.062, + "step": 48493 + }, + { + "epoch": 3.58, + "learning_rate": 7.368106896098317e-06, + "loss": 1.0069, + "step": 48494 + }, + { + "epoch": 3.58, + "learning_rate": 7.3677220891069165e-06, + "loss": 1.0074, + "step": 48495 + }, + { + "epoch": 3.58, + "learning_rate": 7.367337286303434e-06, + "loss": 0.9939, + "step": 48496 + }, + { + "epoch": 3.58, + "learning_rate": 7.366952487688478e-06, + "loss": 0.9982, + "step": 48497 + }, + { + "epoch": 3.58, + "learning_rate": 7.366567693262667e-06, + "loss": 0.8968, + "step": 48498 + }, + { + "epoch": 3.58, + "learning_rate": 7.366182903026604e-06, + "loss": 1.0064, + "step": 48499 + }, + { + "epoch": 3.58, + "learning_rate": 7.365798116980911e-06, + "loss": 0.9698, + "step": 48500 + }, + { + "epoch": 3.58, + "learning_rate": 7.365413335126195e-06, + "loss": 1.0835, + "step": 48501 + }, + { + "epoch": 3.58, + "learning_rate": 7.365028557463065e-06, + "loss": 1.0642, + "step": 48502 + }, + { + "epoch": 3.58, + "learning_rate": 7.3646437839921406e-06, + "loss": 1.0459, + "step": 48503 + }, + { + "epoch": 3.58, + "learning_rate": 7.364259014714026e-06, + "loss": 1.0785, + "step": 48504 + }, + { + "epoch": 3.58, + "learning_rate": 7.363874249629341e-06, + "loss": 0.8724, + "step": 48505 + }, + { + "epoch": 3.58, + "learning_rate": 7.363489488738692e-06, + "loss": 1.0788, + "step": 48506 + }, + { + "epoch": 3.58, + "learning_rate": 7.363104732042696e-06, + "loss": 1.0687, + "step": 48507 + }, + { + "epoch": 3.58, + "learning_rate": 7.36271997954196e-06, + "loss": 1.0332, + "step": 48508 + }, + { + "epoch": 3.58, + "learning_rate": 7.362335231237101e-06, + "loss": 1.0551, + "step": 48509 + }, + { + "epoch": 3.58, + "learning_rate": 7.361950487128729e-06, + "loss": 1.0806, + "step": 48510 + }, + { + "epoch": 3.58, + "learning_rate": 7.361565747217454e-06, + "loss": 1.0093, + "step": 48511 + }, + { + "epoch": 3.58, + "learning_rate": 7.361181011503893e-06, + "loss": 0.9918, + "step": 48512 + }, + { + "epoch": 3.58, + "learning_rate": 7.360796279988652e-06, + "loss": 0.961, + "step": 48513 + }, + { + "epoch": 3.58, + "learning_rate": 7.360411552672349e-06, + "loss": 0.9725, + "step": 48514 + }, + { + "epoch": 3.58, + "learning_rate": 7.3600268295555914e-06, + "loss": 1.095, + "step": 48515 + }, + { + "epoch": 3.58, + "learning_rate": 7.359642110638996e-06, + "loss": 1.021, + "step": 48516 + }, + { + "epoch": 3.58, + "learning_rate": 7.359257395923169e-06, + "loss": 0.9735, + "step": 48517 + }, + { + "epoch": 3.58, + "learning_rate": 7.358872685408728e-06, + "loss": 0.9724, + "step": 48518 + }, + { + "epoch": 3.59, + "learning_rate": 7.35848797909628e-06, + "loss": 1.0494, + "step": 48519 + }, + { + "epoch": 3.59, + "learning_rate": 7.358103276986444e-06, + "loss": 0.9582, + "step": 48520 + }, + { + "epoch": 3.59, + "learning_rate": 7.357718579079829e-06, + "loss": 0.9746, + "step": 48521 + }, + { + "epoch": 3.59, + "learning_rate": 7.357333885377038e-06, + "loss": 0.9741, + "step": 48522 + }, + { + "epoch": 3.59, + "learning_rate": 7.3569491958787e-06, + "loss": 0.9779, + "step": 48523 + }, + { + "epoch": 3.59, + "learning_rate": 7.35656451058541e-06, + "loss": 0.9622, + "step": 48524 + }, + { + "epoch": 3.59, + "learning_rate": 7.356179829497795e-06, + "loss": 1.0265, + "step": 48525 + }, + { + "epoch": 3.59, + "learning_rate": 7.355795152616453e-06, + "loss": 0.8899, + "step": 48526 + }, + { + "epoch": 3.59, + "learning_rate": 7.35541047994201e-06, + "loss": 1.0105, + "step": 48527 + }, + { + "epoch": 3.59, + "learning_rate": 7.355025811475065e-06, + "loss": 1.0583, + "step": 48528 + }, + { + "epoch": 3.59, + "learning_rate": 7.354641147216242e-06, + "loss": 0.9814, + "step": 48529 + }, + { + "epoch": 3.59, + "learning_rate": 7.354256487166145e-06, + "loss": 1.0393, + "step": 48530 + }, + { + "epoch": 3.59, + "learning_rate": 7.353871831325385e-06, + "loss": 1.0198, + "step": 48531 + }, + { + "epoch": 3.59, + "learning_rate": 7.353487179694581e-06, + "loss": 1.0606, + "step": 48532 + }, + { + "epoch": 3.59, + "learning_rate": 7.353102532274339e-06, + "loss": 0.9469, + "step": 48533 + }, + { + "epoch": 3.59, + "learning_rate": 7.352717889065273e-06, + "loss": 1.0287, + "step": 48534 + }, + { + "epoch": 3.59, + "learning_rate": 7.352333250067995e-06, + "loss": 0.997, + "step": 48535 + }, + { + "epoch": 3.59, + "learning_rate": 7.351948615283118e-06, + "loss": 1.0899, + "step": 48536 + }, + { + "epoch": 3.59, + "learning_rate": 7.35156398471125e-06, + "loss": 1.009, + "step": 48537 + }, + { + "epoch": 3.59, + "learning_rate": 7.3511793583530085e-06, + "loss": 1.0901, + "step": 48538 + }, + { + "epoch": 3.59, + "learning_rate": 7.350794736209004e-06, + "loss": 0.9959, + "step": 48539 + }, + { + "epoch": 3.59, + "learning_rate": 7.350410118279844e-06, + "loss": 1.0988, + "step": 48540 + }, + { + "epoch": 3.59, + "learning_rate": 7.350025504566145e-06, + "loss": 1.1487, + "step": 48541 + }, + { + "epoch": 3.59, + "learning_rate": 7.349640895068517e-06, + "loss": 1.1408, + "step": 48542 + }, + { + "epoch": 3.59, + "learning_rate": 7.3492562897875766e-06, + "loss": 0.9428, + "step": 48543 + }, + { + "epoch": 3.59, + "learning_rate": 7.348871688723924e-06, + "loss": 1.0225, + "step": 48544 + }, + { + "epoch": 3.59, + "learning_rate": 7.348487091878185e-06, + "loss": 1.0333, + "step": 48545 + }, + { + "epoch": 3.59, + "learning_rate": 7.348102499250959e-06, + "loss": 0.9285, + "step": 48546 + }, + { + "epoch": 3.59, + "learning_rate": 7.3477179108428706e-06, + "loss": 1.0545, + "step": 48547 + }, + { + "epoch": 3.59, + "learning_rate": 7.347333326654522e-06, + "loss": 0.9812, + "step": 48548 + }, + { + "epoch": 3.59, + "learning_rate": 7.346948746686527e-06, + "loss": 0.974, + "step": 48549 + }, + { + "epoch": 3.59, + "learning_rate": 7.3465641709395005e-06, + "loss": 0.9953, + "step": 48550 + }, + { + "epoch": 3.59, + "learning_rate": 7.34617959941405e-06, + "loss": 0.9289, + "step": 48551 + }, + { + "epoch": 3.59, + "learning_rate": 7.3457950321107926e-06, + "loss": 0.9174, + "step": 48552 + }, + { + "epoch": 3.59, + "learning_rate": 7.345410469030335e-06, + "loss": 0.9648, + "step": 48553 + }, + { + "epoch": 3.59, + "learning_rate": 7.345025910173294e-06, + "loss": 1.0713, + "step": 48554 + }, + { + "epoch": 3.59, + "learning_rate": 7.344641355540275e-06, + "loss": 1.0111, + "step": 48555 + }, + { + "epoch": 3.59, + "learning_rate": 7.3442568051318975e-06, + "loss": 0.9798, + "step": 48556 + }, + { + "epoch": 3.59, + "learning_rate": 7.343872258948768e-06, + "loss": 1.027, + "step": 48557 + }, + { + "epoch": 3.59, + "learning_rate": 7.343487716991499e-06, + "loss": 1.0507, + "step": 48558 + }, + { + "epoch": 3.59, + "learning_rate": 7.343103179260705e-06, + "loss": 1.0902, + "step": 48559 + }, + { + "epoch": 3.59, + "learning_rate": 7.342718645756992e-06, + "loss": 1.0836, + "step": 48560 + }, + { + "epoch": 3.59, + "learning_rate": 7.342334116480978e-06, + "loss": 1.0198, + "step": 48561 + }, + { + "epoch": 3.59, + "learning_rate": 7.341949591433272e-06, + "loss": 0.9885, + "step": 48562 + }, + { + "epoch": 3.59, + "learning_rate": 7.341565070614487e-06, + "loss": 0.9078, + "step": 48563 + }, + { + "epoch": 3.59, + "learning_rate": 7.341180554025232e-06, + "loss": 0.9701, + "step": 48564 + }, + { + "epoch": 3.59, + "learning_rate": 7.340796041666122e-06, + "loss": 0.9923, + "step": 48565 + }, + { + "epoch": 3.59, + "learning_rate": 7.340411533537772e-06, + "loss": 1.0778, + "step": 48566 + }, + { + "epoch": 3.59, + "learning_rate": 7.340027029640782e-06, + "loss": 1.0746, + "step": 48567 + }, + { + "epoch": 3.59, + "learning_rate": 7.3396425299757765e-06, + "loss": 0.9093, + "step": 48568 + }, + { + "epoch": 3.59, + "learning_rate": 7.339258034543355e-06, + "loss": 1.0755, + "step": 48569 + }, + { + "epoch": 3.59, + "learning_rate": 7.338873543344142e-06, + "loss": 0.9299, + "step": 48570 + }, + { + "epoch": 3.59, + "learning_rate": 7.338489056378739e-06, + "loss": 1.01, + "step": 48571 + }, + { + "epoch": 3.59, + "learning_rate": 7.338104573647766e-06, + "loss": 1.0017, + "step": 48572 + }, + { + "epoch": 3.59, + "learning_rate": 7.337720095151825e-06, + "loss": 1.036, + "step": 48573 + }, + { + "epoch": 3.59, + "learning_rate": 7.33733562089154e-06, + "loss": 1.0367, + "step": 48574 + }, + { + "epoch": 3.59, + "learning_rate": 7.33695115086751e-06, + "loss": 0.8916, + "step": 48575 + }, + { + "epoch": 3.59, + "learning_rate": 7.3365666850803575e-06, + "loss": 0.8985, + "step": 48576 + }, + { + "epoch": 3.59, + "learning_rate": 7.336182223530688e-06, + "loss": 0.9655, + "step": 48577 + }, + { + "epoch": 3.59, + "learning_rate": 7.335797766219111e-06, + "loss": 1.0577, + "step": 48578 + }, + { + "epoch": 3.59, + "learning_rate": 7.335413313146246e-06, + "loss": 0.8501, + "step": 48579 + }, + { + "epoch": 3.59, + "learning_rate": 7.335028864312696e-06, + "loss": 1.0453, + "step": 48580 + }, + { + "epoch": 3.59, + "learning_rate": 7.334644419719081e-06, + "loss": 1.1085, + "step": 48581 + }, + { + "epoch": 3.59, + "learning_rate": 7.3342599793660055e-06, + "loss": 1.0061, + "step": 48582 + }, + { + "epoch": 3.59, + "learning_rate": 7.333875543254086e-06, + "loss": 1.1039, + "step": 48583 + }, + { + "epoch": 3.59, + "learning_rate": 7.33349111138393e-06, + "loss": 1.0488, + "step": 48584 + }, + { + "epoch": 3.59, + "learning_rate": 7.333106683756155e-06, + "loss": 1.0777, + "step": 48585 + }, + { + "epoch": 3.59, + "learning_rate": 7.332722260371371e-06, + "loss": 1.0149, + "step": 48586 + }, + { + "epoch": 3.59, + "learning_rate": 7.332337841230181e-06, + "loss": 0.9568, + "step": 48587 + }, + { + "epoch": 3.59, + "learning_rate": 7.331953426333209e-06, + "loss": 0.9975, + "step": 48588 + }, + { + "epoch": 3.59, + "learning_rate": 7.3315690156810545e-06, + "loss": 0.9597, + "step": 48589 + }, + { + "epoch": 3.59, + "learning_rate": 7.3311846092743436e-06, + "loss": 0.9868, + "step": 48590 + }, + { + "epoch": 3.59, + "learning_rate": 7.330800207113672e-06, + "loss": 1.0916, + "step": 48591 + }, + { + "epoch": 3.59, + "learning_rate": 7.330415809199665e-06, + "loss": 1.0713, + "step": 48592 + }, + { + "epoch": 3.59, + "learning_rate": 7.330031415532923e-06, + "loss": 0.947, + "step": 48593 + }, + { + "epoch": 3.59, + "learning_rate": 7.329647026114069e-06, + "loss": 0.9461, + "step": 48594 + }, + { + "epoch": 3.59, + "learning_rate": 7.329262640943706e-06, + "loss": 0.935, + "step": 48595 + }, + { + "epoch": 3.59, + "learning_rate": 7.328878260022447e-06, + "loss": 1.0211, + "step": 48596 + }, + { + "epoch": 3.59, + "learning_rate": 7.328493883350905e-06, + "loss": 1.0394, + "step": 48597 + }, + { + "epoch": 3.59, + "learning_rate": 7.32810951092969e-06, + "loss": 0.9547, + "step": 48598 + }, + { + "epoch": 3.59, + "learning_rate": 7.327725142759415e-06, + "loss": 0.8864, + "step": 48599 + }, + { + "epoch": 3.59, + "learning_rate": 7.32734077884069e-06, + "loss": 0.9823, + "step": 48600 + }, + { + "epoch": 3.59, + "learning_rate": 7.32695641917413e-06, + "loss": 1.095, + "step": 48601 + }, + { + "epoch": 3.59, + "learning_rate": 7.326572063760342e-06, + "loss": 1.0502, + "step": 48602 + }, + { + "epoch": 3.59, + "learning_rate": 7.326187712599941e-06, + "loss": 0.9406, + "step": 48603 + }, + { + "epoch": 3.59, + "learning_rate": 7.3258033656935375e-06, + "loss": 0.9791, + "step": 48604 + }, + { + "epoch": 3.59, + "learning_rate": 7.325419023041741e-06, + "loss": 0.9532, + "step": 48605 + }, + { + "epoch": 3.59, + "learning_rate": 7.325034684645166e-06, + "loss": 1.0171, + "step": 48606 + }, + { + "epoch": 3.59, + "learning_rate": 7.324650350504421e-06, + "loss": 1.0561, + "step": 48607 + }, + { + "epoch": 3.59, + "learning_rate": 7.32426602062012e-06, + "loss": 1.0261, + "step": 48608 + }, + { + "epoch": 3.59, + "learning_rate": 7.323881694992872e-06, + "loss": 1.0418, + "step": 48609 + }, + { + "epoch": 3.59, + "learning_rate": 7.323497373623292e-06, + "loss": 1.1058, + "step": 48610 + }, + { + "epoch": 3.59, + "learning_rate": 7.3231130565119875e-06, + "loss": 0.9111, + "step": 48611 + }, + { + "epoch": 3.59, + "learning_rate": 7.322728743659574e-06, + "loss": 0.8846, + "step": 48612 + }, + { + "epoch": 3.59, + "learning_rate": 7.322344435066663e-06, + "loss": 0.9301, + "step": 48613 + }, + { + "epoch": 3.59, + "learning_rate": 7.321960130733856e-06, + "loss": 0.9489, + "step": 48614 + }, + { + "epoch": 3.59, + "learning_rate": 7.32157583066178e-06, + "loss": 0.9877, + "step": 48615 + }, + { + "epoch": 3.59, + "learning_rate": 7.321191534851032e-06, + "loss": 0.9736, + "step": 48616 + }, + { + "epoch": 3.59, + "learning_rate": 7.320807243302236e-06, + "loss": 1.1296, + "step": 48617 + }, + { + "epoch": 3.59, + "learning_rate": 7.320422956015991e-06, + "loss": 1.0068, + "step": 48618 + }, + { + "epoch": 3.59, + "learning_rate": 7.3200386729929216e-06, + "loss": 0.9173, + "step": 48619 + }, + { + "epoch": 3.59, + "learning_rate": 7.319654394233626e-06, + "loss": 1.0092, + "step": 48620 + }, + { + "epoch": 3.59, + "learning_rate": 7.319270119738726e-06, + "loss": 0.9699, + "step": 48621 + }, + { + "epoch": 3.59, + "learning_rate": 7.318885849508829e-06, + "loss": 0.9579, + "step": 48622 + }, + { + "epoch": 3.59, + "learning_rate": 7.318501583544544e-06, + "loss": 1.1967, + "step": 48623 + }, + { + "epoch": 3.59, + "learning_rate": 7.3181173218464865e-06, + "loss": 1.0532, + "step": 48624 + }, + { + "epoch": 3.59, + "learning_rate": 7.317733064415263e-06, + "loss": 1.0225, + "step": 48625 + }, + { + "epoch": 3.59, + "learning_rate": 7.31734881125149e-06, + "loss": 0.9541, + "step": 48626 + }, + { + "epoch": 3.59, + "learning_rate": 7.316964562355776e-06, + "loss": 1.0996, + "step": 48627 + }, + { + "epoch": 3.59, + "learning_rate": 7.316580317728734e-06, + "loss": 1.0519, + "step": 48628 + }, + { + "epoch": 3.59, + "learning_rate": 7.316196077370973e-06, + "loss": 0.9424, + "step": 48629 + }, + { + "epoch": 3.59, + "learning_rate": 7.315811841283107e-06, + "loss": 0.9493, + "step": 48630 + }, + { + "epoch": 3.59, + "learning_rate": 7.315427609465744e-06, + "loss": 1.0613, + "step": 48631 + }, + { + "epoch": 3.59, + "learning_rate": 7.3150433819195e-06, + "loss": 1.043, + "step": 48632 + }, + { + "epoch": 3.59, + "learning_rate": 7.314659158644986e-06, + "loss": 0.9387, + "step": 48633 + }, + { + "epoch": 3.59, + "learning_rate": 7.3142749396428025e-06, + "loss": 0.8412, + "step": 48634 + }, + { + "epoch": 3.59, + "learning_rate": 7.313890724913577e-06, + "loss": 1.049, + "step": 48635 + }, + { + "epoch": 3.59, + "learning_rate": 7.313506514457906e-06, + "loss": 0.9107, + "step": 48636 + }, + { + "epoch": 3.59, + "learning_rate": 7.313122308276414e-06, + "loss": 1.0201, + "step": 48637 + }, + { + "epoch": 3.59, + "learning_rate": 7.3127381063697e-06, + "loss": 0.9618, + "step": 48638 + }, + { + "epoch": 3.59, + "learning_rate": 7.312353908738388e-06, + "loss": 0.999, + "step": 48639 + }, + { + "epoch": 3.59, + "learning_rate": 7.311969715383075e-06, + "loss": 1.0993, + "step": 48640 + }, + { + "epoch": 3.59, + "learning_rate": 7.311585526304387e-06, + "loss": 1.0038, + "step": 48641 + }, + { + "epoch": 3.59, + "learning_rate": 7.311201341502925e-06, + "loss": 0.9689, + "step": 48642 + }, + { + "epoch": 3.59, + "learning_rate": 7.310817160979302e-06, + "loss": 0.952, + "step": 48643 + }, + { + "epoch": 3.59, + "learning_rate": 7.310432984734132e-06, + "loss": 1.0385, + "step": 48644 + }, + { + "epoch": 3.59, + "learning_rate": 7.3100488127680225e-06, + "loss": 1.0075, + "step": 48645 + }, + { + "epoch": 3.59, + "learning_rate": 7.309664645081589e-06, + "loss": 0.9597, + "step": 48646 + }, + { + "epoch": 3.59, + "learning_rate": 7.309280481675438e-06, + "loss": 0.9948, + "step": 48647 + }, + { + "epoch": 3.59, + "learning_rate": 7.3088963225501865e-06, + "loss": 1.0244, + "step": 48648 + }, + { + "epoch": 3.59, + "learning_rate": 7.30851216770644e-06, + "loss": 0.999, + "step": 48649 + }, + { + "epoch": 3.59, + "learning_rate": 7.308128017144813e-06, + "loss": 1.0383, + "step": 48650 + }, + { + "epoch": 3.59, + "learning_rate": 7.307743870865918e-06, + "loss": 1.028, + "step": 48651 + }, + { + "epoch": 3.59, + "learning_rate": 7.30735972887036e-06, + "loss": 1.0571, + "step": 48652 + }, + { + "epoch": 3.59, + "learning_rate": 7.306975591158757e-06, + "loss": 0.9513, + "step": 48653 + }, + { + "epoch": 3.6, + "learning_rate": 7.306591457731716e-06, + "loss": 0.9561, + "step": 48654 + }, + { + "epoch": 3.6, + "learning_rate": 7.306207328589849e-06, + "loss": 0.9815, + "step": 48655 + }, + { + "epoch": 3.6, + "learning_rate": 7.3058232037337685e-06, + "loss": 0.9938, + "step": 48656 + }, + { + "epoch": 3.6, + "learning_rate": 7.3054390831640855e-06, + "loss": 0.9755, + "step": 48657 + }, + { + "epoch": 3.6, + "learning_rate": 7.305054966881409e-06, + "loss": 1.0345, + "step": 48658 + }, + { + "epoch": 3.6, + "learning_rate": 7.304670854886352e-06, + "loss": 0.9979, + "step": 48659 + }, + { + "epoch": 3.6, + "learning_rate": 7.304286747179528e-06, + "loss": 0.9982, + "step": 48660 + }, + { + "epoch": 3.6, + "learning_rate": 7.303902643761539e-06, + "loss": 0.9304, + "step": 48661 + }, + { + "epoch": 3.6, + "learning_rate": 7.303518544633008e-06, + "loss": 0.8896, + "step": 48662 + }, + { + "epoch": 3.6, + "learning_rate": 7.303134449794538e-06, + "loss": 0.9596, + "step": 48663 + }, + { + "epoch": 3.6, + "learning_rate": 7.302750359246743e-06, + "loss": 0.8435, + "step": 48664 + }, + { + "epoch": 3.6, + "learning_rate": 7.302366272990231e-06, + "loss": 0.9518, + "step": 48665 + }, + { + "epoch": 3.6, + "learning_rate": 7.301982191025619e-06, + "loss": 1.1038, + "step": 48666 + }, + { + "epoch": 3.6, + "learning_rate": 7.301598113353513e-06, + "loss": 0.9799, + "step": 48667 + }, + { + "epoch": 3.6, + "learning_rate": 7.301214039974527e-06, + "loss": 1.0952, + "step": 48668 + }, + { + "epoch": 3.6, + "learning_rate": 7.300829970889272e-06, + "loss": 0.902, + "step": 48669 + }, + { + "epoch": 3.6, + "learning_rate": 7.300445906098355e-06, + "loss": 0.9491, + "step": 48670 + }, + { + "epoch": 3.6, + "learning_rate": 7.3000618456023905e-06, + "loss": 0.9005, + "step": 48671 + }, + { + "epoch": 3.6, + "learning_rate": 7.299677789401989e-06, + "loss": 0.9624, + "step": 48672 + }, + { + "epoch": 3.6, + "learning_rate": 7.2992937374977625e-06, + "loss": 1.0394, + "step": 48673 + }, + { + "epoch": 3.6, + "learning_rate": 7.298909689890319e-06, + "loss": 0.9489, + "step": 48674 + }, + { + "epoch": 3.6, + "learning_rate": 7.298525646580275e-06, + "loss": 0.9864, + "step": 48675 + }, + { + "epoch": 3.6, + "learning_rate": 7.298141607568234e-06, + "loss": 1.0174, + "step": 48676 + }, + { + "epoch": 3.6, + "learning_rate": 7.297757572854814e-06, + "loss": 1.0256, + "step": 48677 + }, + { + "epoch": 3.6, + "learning_rate": 7.297373542440621e-06, + "loss": 0.9673, + "step": 48678 + }, + { + "epoch": 3.6, + "learning_rate": 7.296989516326271e-06, + "loss": 0.9992, + "step": 48679 + }, + { + "epoch": 3.6, + "learning_rate": 7.296605494512372e-06, + "loss": 0.9741, + "step": 48680 + }, + { + "epoch": 3.6, + "learning_rate": 7.2962214769995295e-06, + "loss": 1.1177, + "step": 48681 + }, + { + "epoch": 3.6, + "learning_rate": 7.295837463788368e-06, + "loss": 1.0276, + "step": 48682 + }, + { + "epoch": 3.6, + "learning_rate": 7.295453454879482e-06, + "loss": 1.0696, + "step": 48683 + }, + { + "epoch": 3.6, + "learning_rate": 7.295069450273497e-06, + "loss": 0.9608, + "step": 48684 + }, + { + "epoch": 3.6, + "learning_rate": 7.294685449971013e-06, + "loss": 1.0823, + "step": 48685 + }, + { + "epoch": 3.6, + "learning_rate": 7.294301453972652e-06, + "loss": 1.0198, + "step": 48686 + }, + { + "epoch": 3.6, + "learning_rate": 7.293917462279012e-06, + "loss": 1.0031, + "step": 48687 + }, + { + "epoch": 3.6, + "learning_rate": 7.293533474890717e-06, + "loss": 0.9381, + "step": 48688 + }, + { + "epoch": 3.6, + "learning_rate": 7.293149491808369e-06, + "loss": 0.9905, + "step": 48689 + }, + { + "epoch": 3.6, + "learning_rate": 7.292765513032579e-06, + "loss": 0.9663, + "step": 48690 + }, + { + "epoch": 3.6, + "learning_rate": 7.292381538563963e-06, + "loss": 1.0043, + "step": 48691 + }, + { + "epoch": 3.6, + "learning_rate": 7.291997568403126e-06, + "loss": 1.0828, + "step": 48692 + }, + { + "epoch": 3.6, + "learning_rate": 7.291613602550686e-06, + "loss": 0.9057, + "step": 48693 + }, + { + "epoch": 3.6, + "learning_rate": 7.291229641007247e-06, + "loss": 1.0207, + "step": 48694 + }, + { + "epoch": 3.6, + "learning_rate": 7.290845683773425e-06, + "loss": 1.0412, + "step": 48695 + }, + { + "epoch": 3.6, + "learning_rate": 7.290461730849827e-06, + "loss": 1.0658, + "step": 48696 + }, + { + "epoch": 3.6, + "learning_rate": 7.290077782237068e-06, + "loss": 1.0261, + "step": 48697 + }, + { + "epoch": 3.6, + "learning_rate": 7.2896938379357565e-06, + "loss": 1.0573, + "step": 48698 + }, + { + "epoch": 3.6, + "learning_rate": 7.2893098979465e-06, + "loss": 0.9995, + "step": 48699 + }, + { + "epoch": 3.6, + "learning_rate": 7.288925962269915e-06, + "loss": 0.9743, + "step": 48700 + }, + { + "epoch": 3.6, + "learning_rate": 7.288542030906609e-06, + "loss": 1.034, + "step": 48701 + }, + { + "epoch": 3.6, + "learning_rate": 7.288158103857195e-06, + "loss": 0.9741, + "step": 48702 + }, + { + "epoch": 3.6, + "learning_rate": 7.287774181122281e-06, + "loss": 0.9128, + "step": 48703 + }, + { + "epoch": 3.6, + "learning_rate": 7.287390262702482e-06, + "loss": 1.0962, + "step": 48704 + }, + { + "epoch": 3.6, + "learning_rate": 7.287006348598403e-06, + "loss": 1.0715, + "step": 48705 + }, + { + "epoch": 3.6, + "learning_rate": 7.286622438810661e-06, + "loss": 0.9763, + "step": 48706 + }, + { + "epoch": 3.6, + "learning_rate": 7.286238533339867e-06, + "loss": 1.0149, + "step": 48707 + }, + { + "epoch": 3.6, + "learning_rate": 7.285854632186622e-06, + "loss": 1.0354, + "step": 48708 + }, + { + "epoch": 3.6, + "learning_rate": 7.285470735351548e-06, + "loss": 0.9655, + "step": 48709 + }, + { + "epoch": 3.6, + "learning_rate": 7.285086842835248e-06, + "loss": 0.9792, + "step": 48710 + }, + { + "epoch": 3.6, + "learning_rate": 7.284702954638336e-06, + "loss": 1.0563, + "step": 48711 + }, + { + "epoch": 3.6, + "learning_rate": 7.284319070761423e-06, + "loss": 1.047, + "step": 48712 + }, + { + "epoch": 3.6, + "learning_rate": 7.283935191205122e-06, + "loss": 0.9981, + "step": 48713 + }, + { + "epoch": 3.6, + "learning_rate": 7.283551315970039e-06, + "loss": 0.9452, + "step": 48714 + }, + { + "epoch": 3.6, + "learning_rate": 7.283167445056787e-06, + "loss": 0.9579, + "step": 48715 + }, + { + "epoch": 3.6, + "learning_rate": 7.282783578465979e-06, + "loss": 0.9844, + "step": 48716 + }, + { + "epoch": 3.6, + "learning_rate": 7.282399716198221e-06, + "loss": 0.9208, + "step": 48717 + }, + { + "epoch": 3.6, + "learning_rate": 7.282015858254128e-06, + "loss": 0.9681, + "step": 48718 + }, + { + "epoch": 3.6, + "learning_rate": 7.281632004634306e-06, + "loss": 1.0672, + "step": 48719 + }, + { + "epoch": 3.6, + "learning_rate": 7.281248155339371e-06, + "loss": 1.0468, + "step": 48720 + }, + { + "epoch": 3.6, + "learning_rate": 7.28086431036993e-06, + "loss": 1.0538, + "step": 48721 + }, + { + "epoch": 3.6, + "learning_rate": 7.280480469726596e-06, + "loss": 1.0845, + "step": 48722 + }, + { + "epoch": 3.6, + "learning_rate": 7.2800966334099774e-06, + "loss": 1.0962, + "step": 48723 + }, + { + "epoch": 3.6, + "learning_rate": 7.279712801420688e-06, + "loss": 1.0221, + "step": 48724 + }, + { + "epoch": 3.6, + "learning_rate": 7.279328973759339e-06, + "loss": 1.1706, + "step": 48725 + }, + { + "epoch": 3.6, + "learning_rate": 7.2789451504265315e-06, + "loss": 1.0369, + "step": 48726 + }, + { + "epoch": 3.6, + "learning_rate": 7.278561331422891e-06, + "loss": 1.0284, + "step": 48727 + }, + { + "epoch": 3.6, + "learning_rate": 7.278177516749013e-06, + "loss": 0.9615, + "step": 48728 + }, + { + "epoch": 3.6, + "learning_rate": 7.277793706405523e-06, + "loss": 0.9592, + "step": 48729 + }, + { + "epoch": 3.6, + "learning_rate": 7.277409900393016e-06, + "loss": 1.0858, + "step": 48730 + }, + { + "epoch": 3.6, + "learning_rate": 7.277026098712119e-06, + "loss": 1.0476, + "step": 48731 + }, + { + "epoch": 3.6, + "learning_rate": 7.276642301363428e-06, + "loss": 1.0725, + "step": 48732 + }, + { + "epoch": 3.6, + "learning_rate": 7.276258508347566e-06, + "loss": 1.0466, + "step": 48733 + }, + { + "epoch": 3.6, + "learning_rate": 7.275874719665131e-06, + "loss": 0.9377, + "step": 48734 + }, + { + "epoch": 3.6, + "learning_rate": 7.275490935316748e-06, + "loss": 0.974, + "step": 48735 + }, + { + "epoch": 3.6, + "learning_rate": 7.2751071553030165e-06, + "loss": 1.0229, + "step": 48736 + }, + { + "epoch": 3.6, + "learning_rate": 7.274723379624549e-06, + "loss": 0.9868, + "step": 48737 + }, + { + "epoch": 3.6, + "learning_rate": 7.274339608281959e-06, + "loss": 0.9765, + "step": 48738 + }, + { + "epoch": 3.6, + "learning_rate": 7.2739558412758544e-06, + "loss": 0.9484, + "step": 48739 + }, + { + "epoch": 3.6, + "learning_rate": 7.27357207860685e-06, + "loss": 0.9242, + "step": 48740 + }, + { + "epoch": 3.6, + "learning_rate": 7.27318832027555e-06, + "loss": 1.0285, + "step": 48741 + }, + { + "epoch": 3.6, + "learning_rate": 7.272804566282572e-06, + "loss": 1.1063, + "step": 48742 + }, + { + "epoch": 3.6, + "learning_rate": 7.272420816628518e-06, + "loss": 0.9855, + "step": 48743 + }, + { + "epoch": 3.6, + "learning_rate": 7.272037071314009e-06, + "loss": 0.9853, + "step": 48744 + }, + { + "epoch": 3.6, + "learning_rate": 7.271653330339648e-06, + "loss": 0.9532, + "step": 48745 + }, + { + "epoch": 3.6, + "learning_rate": 7.2712695937060455e-06, + "loss": 0.9427, + "step": 48746 + }, + { + "epoch": 3.6, + "learning_rate": 7.270885861413816e-06, + "loss": 1.0674, + "step": 48747 + }, + { + "epoch": 3.6, + "learning_rate": 7.270502133463566e-06, + "loss": 1.0573, + "step": 48748 + }, + { + "epoch": 3.6, + "learning_rate": 7.270118409855911e-06, + "loss": 1.0582, + "step": 48749 + }, + { + "epoch": 3.6, + "learning_rate": 7.269734690591455e-06, + "loss": 0.9718, + "step": 48750 + }, + { + "epoch": 3.6, + "learning_rate": 7.2693509756708145e-06, + "loss": 0.9045, + "step": 48751 + }, + { + "epoch": 3.6, + "learning_rate": 7.268967265094595e-06, + "loss": 1.0292, + "step": 48752 + }, + { + "epoch": 3.6, + "learning_rate": 7.2685835588634154e-06, + "loss": 1.0352, + "step": 48753 + }, + { + "epoch": 3.6, + "learning_rate": 7.268199856977878e-06, + "loss": 0.9543, + "step": 48754 + }, + { + "epoch": 3.6, + "learning_rate": 7.267816159438592e-06, + "loss": 0.9594, + "step": 48755 + }, + { + "epoch": 3.6, + "learning_rate": 7.267432466246175e-06, + "loss": 1.1055, + "step": 48756 + }, + { + "epoch": 3.6, + "learning_rate": 7.267048777401231e-06, + "loss": 1.0084, + "step": 48757 + }, + { + "epoch": 3.6, + "learning_rate": 7.266665092904375e-06, + "loss": 0.9546, + "step": 48758 + }, + { + "epoch": 3.6, + "learning_rate": 7.266281412756214e-06, + "loss": 1.0729, + "step": 48759 + }, + { + "epoch": 3.6, + "learning_rate": 7.265897736957363e-06, + "loss": 0.9314, + "step": 48760 + }, + { + "epoch": 3.6, + "learning_rate": 7.265514065508425e-06, + "loss": 1.0262, + "step": 48761 + }, + { + "epoch": 3.6, + "learning_rate": 7.265130398410021e-06, + "loss": 0.9651, + "step": 48762 + }, + { + "epoch": 3.6, + "learning_rate": 7.264746735662753e-06, + "loss": 0.9821, + "step": 48763 + }, + { + "epoch": 3.6, + "learning_rate": 7.264363077267231e-06, + "loss": 1.031, + "step": 48764 + }, + { + "epoch": 3.6, + "learning_rate": 7.263979423224072e-06, + "loss": 0.9716, + "step": 48765 + }, + { + "epoch": 3.6, + "learning_rate": 7.263595773533881e-06, + "loss": 1.0717, + "step": 48766 + }, + { + "epoch": 3.6, + "learning_rate": 7.26321212819727e-06, + "loss": 0.9379, + "step": 48767 + }, + { + "epoch": 3.6, + "learning_rate": 7.262828487214848e-06, + "loss": 1.0189, + "step": 48768 + }, + { + "epoch": 3.6, + "learning_rate": 7.262444850587229e-06, + "loss": 0.97, + "step": 48769 + }, + { + "epoch": 3.6, + "learning_rate": 7.262061218315018e-06, + "loss": 1.0143, + "step": 48770 + }, + { + "epoch": 3.6, + "learning_rate": 7.261677590398833e-06, + "loss": 0.8742, + "step": 48771 + }, + { + "epoch": 3.6, + "learning_rate": 7.261293966839279e-06, + "loss": 0.9617, + "step": 48772 + }, + { + "epoch": 3.6, + "learning_rate": 7.260910347636963e-06, + "loss": 0.8342, + "step": 48773 + }, + { + "epoch": 3.6, + "learning_rate": 7.2605267327925045e-06, + "loss": 0.9723, + "step": 48774 + }, + { + "epoch": 3.6, + "learning_rate": 7.260143122306502e-06, + "loss": 1.0207, + "step": 48775 + }, + { + "epoch": 3.6, + "learning_rate": 7.259759516179581e-06, + "loss": 1.0881, + "step": 48776 + }, + { + "epoch": 3.6, + "learning_rate": 7.259375914412335e-06, + "loss": 0.9756, + "step": 48777 + }, + { + "epoch": 3.6, + "learning_rate": 7.25899231700539e-06, + "loss": 0.9146, + "step": 48778 + }, + { + "epoch": 3.6, + "learning_rate": 7.258608723959342e-06, + "loss": 0.9331, + "step": 48779 + }, + { + "epoch": 3.6, + "learning_rate": 7.258225135274816e-06, + "loss": 1.0394, + "step": 48780 + }, + { + "epoch": 3.6, + "learning_rate": 7.257841550952411e-06, + "loss": 0.982, + "step": 48781 + }, + { + "epoch": 3.6, + "learning_rate": 7.25745797099274e-06, + "loss": 1.0609, + "step": 48782 + }, + { + "epoch": 3.6, + "learning_rate": 7.257074395396415e-06, + "loss": 1.0391, + "step": 48783 + }, + { + "epoch": 3.6, + "learning_rate": 7.256690824164044e-06, + "loss": 0.94, + "step": 48784 + }, + { + "epoch": 3.6, + "learning_rate": 7.256307257296241e-06, + "loss": 1.0393, + "step": 48785 + }, + { + "epoch": 3.6, + "learning_rate": 7.2559236947936115e-06, + "loss": 0.9072, + "step": 48786 + }, + { + "epoch": 3.6, + "learning_rate": 7.25554013665677e-06, + "loss": 0.9112, + "step": 48787 + }, + { + "epoch": 3.6, + "learning_rate": 7.255156582886322e-06, + "loss": 0.9794, + "step": 48788 + }, + { + "epoch": 3.6, + "learning_rate": 7.2547730334828845e-06, + "loss": 1.0437, + "step": 48789 + }, + { + "epoch": 3.61, + "learning_rate": 7.254389488447061e-06, + "loss": 1.0098, + "step": 48790 + }, + { + "epoch": 3.61, + "learning_rate": 7.254005947779467e-06, + "loss": 1.0046, + "step": 48791 + }, + { + "epoch": 3.61, + "learning_rate": 7.253622411480709e-06, + "loss": 1.0699, + "step": 48792 + }, + { + "epoch": 3.61, + "learning_rate": 7.253238879551398e-06, + "loss": 1.1049, + "step": 48793 + }, + { + "epoch": 3.61, + "learning_rate": 7.252855351992145e-06, + "loss": 0.9687, + "step": 48794 + }, + { + "epoch": 3.61, + "learning_rate": 7.252471828803559e-06, + "loss": 0.9049, + "step": 48795 + }, + { + "epoch": 3.61, + "learning_rate": 7.252088309986255e-06, + "loss": 1.0108, + "step": 48796 + }, + { + "epoch": 3.61, + "learning_rate": 7.251704795540832e-06, + "loss": 0.9907, + "step": 48797 + }, + { + "epoch": 3.61, + "learning_rate": 7.251321285467915e-06, + "loss": 1.0835, + "step": 48798 + }, + { + "epoch": 3.61, + "learning_rate": 7.250937779768099e-06, + "loss": 1.0104, + "step": 48799 + }, + { + "epoch": 3.61, + "learning_rate": 7.250554278442009e-06, + "loss": 0.9998, + "step": 48800 + }, + { + "epoch": 3.61, + "learning_rate": 7.250170781490245e-06, + "loss": 1.0025, + "step": 48801 + }, + { + "epoch": 3.61, + "learning_rate": 7.249787288913419e-06, + "loss": 0.9718, + "step": 48802 + }, + { + "epoch": 3.61, + "learning_rate": 7.249403800712143e-06, + "loss": 0.9589, + "step": 48803 + }, + { + "epoch": 3.61, + "learning_rate": 7.249020316887025e-06, + "loss": 0.9974, + "step": 48804 + }, + { + "epoch": 3.61, + "learning_rate": 7.248636837438676e-06, + "loss": 1.0556, + "step": 48805 + }, + { + "epoch": 3.61, + "learning_rate": 7.248253362367708e-06, + "loss": 1.0824, + "step": 48806 + }, + { + "epoch": 3.61, + "learning_rate": 7.2478698916747285e-06, + "loss": 0.9348, + "step": 48807 + }, + { + "epoch": 3.61, + "learning_rate": 7.247486425360348e-06, + "loss": 1.0819, + "step": 48808 + }, + { + "epoch": 3.61, + "learning_rate": 7.247102963425179e-06, + "loss": 1.0244, + "step": 48809 + }, + { + "epoch": 3.61, + "learning_rate": 7.24671950586983e-06, + "loss": 1.0951, + "step": 48810 + }, + { + "epoch": 3.61, + "learning_rate": 7.246336052694909e-06, + "loss": 0.9081, + "step": 48811 + }, + { + "epoch": 3.61, + "learning_rate": 7.2459526039010295e-06, + "loss": 1.0244, + "step": 48812 + }, + { + "epoch": 3.61, + "learning_rate": 7.2455691594887975e-06, + "loss": 1.0442, + "step": 48813 + }, + { + "epoch": 3.61, + "learning_rate": 7.2451857194588285e-06, + "loss": 1.1624, + "step": 48814 + }, + { + "epoch": 3.61, + "learning_rate": 7.244802283811727e-06, + "loss": 1.0956, + "step": 48815 + }, + { + "epoch": 3.61, + "learning_rate": 7.244418852548107e-06, + "loss": 1.0036, + "step": 48816 + }, + { + "epoch": 3.61, + "learning_rate": 7.244035425668577e-06, + "loss": 1.0704, + "step": 48817 + }, + { + "epoch": 3.61, + "learning_rate": 7.243652003173749e-06, + "loss": 1.0181, + "step": 48818 + }, + { + "epoch": 3.61, + "learning_rate": 7.243268585064233e-06, + "loss": 0.9763, + "step": 48819 + }, + { + "epoch": 3.61, + "learning_rate": 7.24288517134063e-06, + "loss": 1.1583, + "step": 48820 + }, + { + "epoch": 3.61, + "learning_rate": 7.242501762003564e-06, + "loss": 1.0546, + "step": 48821 + }, + { + "epoch": 3.61, + "learning_rate": 7.2421183570536325e-06, + "loss": 0.9551, + "step": 48822 + }, + { + "epoch": 3.61, + "learning_rate": 7.241734956491458e-06, + "loss": 0.8472, + "step": 48823 + }, + { + "epoch": 3.61, + "learning_rate": 7.241351560317635e-06, + "loss": 1.0015, + "step": 48824 + }, + { + "epoch": 3.61, + "learning_rate": 7.24096816853279e-06, + "loss": 0.9991, + "step": 48825 + }, + { + "epoch": 3.61, + "learning_rate": 7.2405847811375195e-06, + "loss": 1.0272, + "step": 48826 + }, + { + "epoch": 3.61, + "learning_rate": 7.2402013981324445e-06, + "loss": 0.9835, + "step": 48827 + }, + { + "epoch": 3.61, + "learning_rate": 7.239818019518168e-06, + "loss": 1.0645, + "step": 48828 + }, + { + "epoch": 3.61, + "learning_rate": 7.2394346452953e-06, + "loss": 1.0939, + "step": 48829 + }, + { + "epoch": 3.61, + "learning_rate": 7.239051275464453e-06, + "loss": 0.9236, + "step": 48830 + }, + { + "epoch": 3.61, + "learning_rate": 7.238667910026234e-06, + "loss": 1.0334, + "step": 48831 + }, + { + "epoch": 3.61, + "learning_rate": 7.238284548981258e-06, + "loss": 1.0723, + "step": 48832 + }, + { + "epoch": 3.61, + "learning_rate": 7.2379011923301276e-06, + "loss": 0.9424, + "step": 48833 + }, + { + "epoch": 3.61, + "learning_rate": 7.23751784007346e-06, + "loss": 1.0174, + "step": 48834 + }, + { + "epoch": 3.61, + "learning_rate": 7.2371344922118615e-06, + "loss": 1.0293, + "step": 48835 + }, + { + "epoch": 3.61, + "learning_rate": 7.236751148745942e-06, + "loss": 0.9577, + "step": 48836 + }, + { + "epoch": 3.61, + "learning_rate": 7.236367809676314e-06, + "loss": 1.0047, + "step": 48837 + }, + { + "epoch": 3.61, + "learning_rate": 7.235984475003582e-06, + "loss": 0.9755, + "step": 48838 + }, + { + "epoch": 3.61, + "learning_rate": 7.235601144728362e-06, + "loss": 0.9179, + "step": 48839 + }, + { + "epoch": 3.61, + "learning_rate": 7.235217818851258e-06, + "loss": 1.0419, + "step": 48840 + }, + { + "epoch": 3.61, + "learning_rate": 7.234834497372887e-06, + "loss": 1.0205, + "step": 48841 + }, + { + "epoch": 3.61, + "learning_rate": 7.234451180293849e-06, + "loss": 1.0808, + "step": 48842 + }, + { + "epoch": 3.61, + "learning_rate": 7.234067867614767e-06, + "loss": 1.0031, + "step": 48843 + }, + { + "epoch": 3.61, + "learning_rate": 7.233684559336236e-06, + "loss": 0.984, + "step": 48844 + }, + { + "epoch": 3.61, + "learning_rate": 7.233301255458879e-06, + "loss": 1.0253, + "step": 48845 + }, + { + "epoch": 3.61, + "learning_rate": 7.232917955983294e-06, + "loss": 1.0, + "step": 48846 + }, + { + "epoch": 3.61, + "learning_rate": 7.2325346609101044e-06, + "loss": 0.9776, + "step": 48847 + }, + { + "epoch": 3.61, + "learning_rate": 7.232151370239909e-06, + "loss": 1.0279, + "step": 48848 + }, + { + "epoch": 3.61, + "learning_rate": 7.231768083973321e-06, + "loss": 0.9985, + "step": 48849 + }, + { + "epoch": 3.61, + "learning_rate": 7.23138480211095e-06, + "loss": 0.9029, + "step": 48850 + }, + { + "epoch": 3.61, + "learning_rate": 7.231001524653406e-06, + "loss": 1.0816, + "step": 48851 + }, + { + "epoch": 3.61, + "learning_rate": 7.2306182516013e-06, + "loss": 1.0184, + "step": 48852 + }, + { + "epoch": 3.61, + "learning_rate": 7.23023498295524e-06, + "loss": 0.9398, + "step": 48853 + }, + { + "epoch": 3.61, + "learning_rate": 7.2298517187158364e-06, + "loss": 1.0389, + "step": 48854 + }, + { + "epoch": 3.61, + "learning_rate": 7.229468458883699e-06, + "loss": 1.0351, + "step": 48855 + }, + { + "epoch": 3.61, + "learning_rate": 7.229085203459438e-06, + "loss": 1.0264, + "step": 48856 + }, + { + "epoch": 3.61, + "learning_rate": 7.228701952443664e-06, + "loss": 0.9985, + "step": 48857 + }, + { + "epoch": 3.61, + "learning_rate": 7.228318705836983e-06, + "loss": 0.9517, + "step": 48858 + }, + { + "epoch": 3.61, + "learning_rate": 7.22793546364001e-06, + "loss": 0.9504, + "step": 48859 + }, + { + "epoch": 3.61, + "learning_rate": 7.227552225853349e-06, + "loss": 0.9415, + "step": 48860 + }, + { + "epoch": 3.61, + "learning_rate": 7.227168992477616e-06, + "loss": 1.0958, + "step": 48861 + }, + { + "epoch": 3.61, + "learning_rate": 7.226785763513416e-06, + "loss": 0.8496, + "step": 48862 + }, + { + "epoch": 3.61, + "learning_rate": 7.226402538961359e-06, + "loss": 0.9753, + "step": 48863 + }, + { + "epoch": 3.61, + "learning_rate": 7.226019318822057e-06, + "loss": 1.0181, + "step": 48864 + }, + { + "epoch": 3.61, + "learning_rate": 7.225636103096119e-06, + "loss": 0.9692, + "step": 48865 + }, + { + "epoch": 3.61, + "learning_rate": 7.225252891784157e-06, + "loss": 1.0678, + "step": 48866 + }, + { + "epoch": 3.61, + "learning_rate": 7.2248696848867704e-06, + "loss": 0.9704, + "step": 48867 + }, + { + "epoch": 3.61, + "learning_rate": 7.224486482404584e-06, + "loss": 0.9655, + "step": 48868 + }, + { + "epoch": 3.61, + "learning_rate": 7.2241032843381945e-06, + "loss": 1.0149, + "step": 48869 + }, + { + "epoch": 3.61, + "learning_rate": 7.2237200906882225e-06, + "loss": 0.9205, + "step": 48870 + }, + { + "epoch": 3.61, + "learning_rate": 7.223336901455265e-06, + "loss": 0.9662, + "step": 48871 + }, + { + "epoch": 3.61, + "learning_rate": 7.222953716639945e-06, + "loss": 1.0144, + "step": 48872 + }, + { + "epoch": 3.61, + "learning_rate": 7.22257053624286e-06, + "loss": 1.1343, + "step": 48873 + }, + { + "epoch": 3.61, + "learning_rate": 7.222187360264632e-06, + "loss": 1.0248, + "step": 48874 + }, + { + "epoch": 3.61, + "learning_rate": 7.221804188705862e-06, + "loss": 0.936, + "step": 48875 + }, + { + "epoch": 3.61, + "learning_rate": 7.22142102156716e-06, + "loss": 1.0087, + "step": 48876 + }, + { + "epoch": 3.61, + "learning_rate": 7.221037858849139e-06, + "loss": 0.909, + "step": 48877 + }, + { + "epoch": 3.61, + "learning_rate": 7.220654700552405e-06, + "loss": 1.0112, + "step": 48878 + }, + { + "epoch": 3.61, + "learning_rate": 7.220271546677572e-06, + "loss": 1.0736, + "step": 48879 + }, + { + "epoch": 3.61, + "learning_rate": 7.219888397225245e-06, + "loss": 0.9436, + "step": 48880 + }, + { + "epoch": 3.61, + "learning_rate": 7.219505252196037e-06, + "loss": 1.0247, + "step": 48881 + }, + { + "epoch": 3.61, + "learning_rate": 7.219122111590555e-06, + "loss": 1.1583, + "step": 48882 + }, + { + "epoch": 3.61, + "learning_rate": 7.218738975409413e-06, + "loss": 1.0172, + "step": 48883 + }, + { + "epoch": 3.61, + "learning_rate": 7.218355843653217e-06, + "loss": 1.0242, + "step": 48884 + }, + { + "epoch": 3.61, + "learning_rate": 7.217972716322575e-06, + "loss": 1.05, + "step": 48885 + }, + { + "epoch": 3.61, + "learning_rate": 7.217589593418101e-06, + "loss": 1.084, + "step": 48886 + }, + { + "epoch": 3.61, + "learning_rate": 7.217206474940397e-06, + "loss": 1.0333, + "step": 48887 + }, + { + "epoch": 3.61, + "learning_rate": 7.216823360890084e-06, + "loss": 0.9255, + "step": 48888 + }, + { + "epoch": 3.61, + "learning_rate": 7.21644025126776e-06, + "loss": 0.92, + "step": 48889 + }, + { + "epoch": 3.61, + "learning_rate": 7.216057146074045e-06, + "loss": 1.0024, + "step": 48890 + }, + { + "epoch": 3.61, + "learning_rate": 7.215674045309537e-06, + "loss": 0.8965, + "step": 48891 + }, + { + "epoch": 3.61, + "learning_rate": 7.215290948974858e-06, + "loss": 0.9864, + "step": 48892 + }, + { + "epoch": 3.61, + "learning_rate": 7.2149078570706055e-06, + "loss": 0.9746, + "step": 48893 + }, + { + "epoch": 3.61, + "learning_rate": 7.214524769597399e-06, + "loss": 0.9372, + "step": 48894 + }, + { + "epoch": 3.61, + "learning_rate": 7.214141686555844e-06, + "loss": 0.9254, + "step": 48895 + }, + { + "epoch": 3.61, + "learning_rate": 7.213758607946548e-06, + "loss": 1.0298, + "step": 48896 + }, + { + "epoch": 3.61, + "learning_rate": 7.213375533770123e-06, + "loss": 0.875, + "step": 48897 + }, + { + "epoch": 3.61, + "learning_rate": 7.212992464027175e-06, + "loss": 0.899, + "step": 48898 + }, + { + "epoch": 3.61, + "learning_rate": 7.21260939871832e-06, + "loss": 1.1054, + "step": 48899 + }, + { + "epoch": 3.61, + "learning_rate": 7.212226337844159e-06, + "loss": 0.9611, + "step": 48900 + }, + { + "epoch": 3.61, + "learning_rate": 7.211843281405311e-06, + "loss": 0.9906, + "step": 48901 + }, + { + "epoch": 3.61, + "learning_rate": 7.211460229402376e-06, + "loss": 1.1112, + "step": 48902 + }, + { + "epoch": 3.61, + "learning_rate": 7.21107718183597e-06, + "loss": 0.9526, + "step": 48903 + }, + { + "epoch": 3.61, + "learning_rate": 7.210694138706702e-06, + "loss": 0.8728, + "step": 48904 + }, + { + "epoch": 3.61, + "learning_rate": 7.210311100015177e-06, + "loss": 1.0274, + "step": 48905 + }, + { + "epoch": 3.61, + "learning_rate": 7.2099280657620085e-06, + "loss": 1.0279, + "step": 48906 + }, + { + "epoch": 3.61, + "learning_rate": 7.209545035947803e-06, + "loss": 1.0032, + "step": 48907 + }, + { + "epoch": 3.61, + "learning_rate": 7.209162010573174e-06, + "loss": 1.0692, + "step": 48908 + }, + { + "epoch": 3.61, + "learning_rate": 7.208778989638725e-06, + "loss": 0.9862, + "step": 48909 + }, + { + "epoch": 3.61, + "learning_rate": 7.208395973145072e-06, + "loss": 1.1666, + "step": 48910 + }, + { + "epoch": 3.61, + "learning_rate": 7.208012961092818e-06, + "loss": 1.007, + "step": 48911 + }, + { + "epoch": 3.61, + "learning_rate": 7.2076299534825775e-06, + "loss": 0.9192, + "step": 48912 + }, + { + "epoch": 3.61, + "learning_rate": 7.207246950314961e-06, + "loss": 1.0939, + "step": 48913 + }, + { + "epoch": 3.61, + "learning_rate": 7.206863951590568e-06, + "loss": 0.9513, + "step": 48914 + }, + { + "epoch": 3.61, + "learning_rate": 7.206480957310021e-06, + "loss": 1.0113, + "step": 48915 + }, + { + "epoch": 3.61, + "learning_rate": 7.206097967473915e-06, + "loss": 1.0043, + "step": 48916 + }, + { + "epoch": 3.61, + "learning_rate": 7.205714982082874e-06, + "loss": 1.0558, + "step": 48917 + }, + { + "epoch": 3.61, + "learning_rate": 7.205332001137497e-06, + "loss": 1.061, + "step": 48918 + }, + { + "epoch": 3.61, + "learning_rate": 7.204949024638399e-06, + "loss": 0.9214, + "step": 48919 + }, + { + "epoch": 3.61, + "learning_rate": 7.204566052586185e-06, + "loss": 1.1533, + "step": 48920 + }, + { + "epoch": 3.61, + "learning_rate": 7.204183084981468e-06, + "loss": 1.0082, + "step": 48921 + }, + { + "epoch": 3.61, + "learning_rate": 7.203800121824855e-06, + "loss": 1.0113, + "step": 48922 + }, + { + "epoch": 3.61, + "learning_rate": 7.203417163116955e-06, + "loss": 0.9662, + "step": 48923 + }, + { + "epoch": 3.61, + "learning_rate": 7.2030342088583805e-06, + "loss": 1.0902, + "step": 48924 + }, + { + "epoch": 3.62, + "learning_rate": 7.202651259049735e-06, + "loss": 1.0442, + "step": 48925 + }, + { + "epoch": 3.62, + "learning_rate": 7.202268313691636e-06, + "loss": 1.0087, + "step": 48926 + }, + { + "epoch": 3.62, + "learning_rate": 7.201885372784685e-06, + "loss": 0.9697, + "step": 48927 + }, + { + "epoch": 3.62, + "learning_rate": 7.201502436329495e-06, + "loss": 1.0265, + "step": 48928 + }, + { + "epoch": 3.62, + "learning_rate": 7.2011195043266735e-06, + "loss": 1.0648, + "step": 48929 + }, + { + "epoch": 3.62, + "learning_rate": 7.2007365767768335e-06, + "loss": 0.9308, + "step": 48930 + }, + { + "epoch": 3.62, + "learning_rate": 7.2003536536805825e-06, + "loss": 0.9389, + "step": 48931 + }, + { + "epoch": 3.62, + "learning_rate": 7.199970735038523e-06, + "loss": 0.9672, + "step": 48932 + }, + { + "epoch": 3.62, + "learning_rate": 7.199587820851277e-06, + "loss": 0.9647, + "step": 48933 + }, + { + "epoch": 3.62, + "learning_rate": 7.1992049111194394e-06, + "loss": 0.9147, + "step": 48934 + }, + { + "epoch": 3.62, + "learning_rate": 7.1988220058436335e-06, + "loss": 0.8974, + "step": 48935 + }, + { + "epoch": 3.62, + "learning_rate": 7.198439105024455e-06, + "loss": 1.0107, + "step": 48936 + }, + { + "epoch": 3.62, + "learning_rate": 7.1980562086625264e-06, + "loss": 1.1022, + "step": 48937 + }, + { + "epoch": 3.62, + "learning_rate": 7.197673316758443e-06, + "loss": 0.9671, + "step": 48938 + }, + { + "epoch": 3.62, + "learning_rate": 7.197290429312828e-06, + "loss": 1.0796, + "step": 48939 + }, + { + "epoch": 3.62, + "learning_rate": 7.196907546326284e-06, + "loss": 1.0554, + "step": 48940 + }, + { + "epoch": 3.62, + "learning_rate": 7.196524667799416e-06, + "loss": 1.0186, + "step": 48941 + }, + { + "epoch": 3.62, + "learning_rate": 7.196141793732838e-06, + "loss": 1.0281, + "step": 48942 + }, + { + "epoch": 3.62, + "learning_rate": 7.1957589241271586e-06, + "loss": 1.0128, + "step": 48943 + }, + { + "epoch": 3.62, + "learning_rate": 7.195376058982988e-06, + "loss": 1.041, + "step": 48944 + }, + { + "epoch": 3.62, + "learning_rate": 7.19499319830093e-06, + "loss": 1.087, + "step": 48945 + }, + { + "epoch": 3.62, + "learning_rate": 7.1946103420816025e-06, + "loss": 0.9676, + "step": 48946 + }, + { + "epoch": 3.62, + "learning_rate": 7.194227490325606e-06, + "loss": 1.0342, + "step": 48947 + }, + { + "epoch": 3.62, + "learning_rate": 7.193844643033556e-06, + "loss": 0.987, + "step": 48948 + }, + { + "epoch": 3.62, + "learning_rate": 7.193461800206056e-06, + "loss": 1.0036, + "step": 48949 + }, + { + "epoch": 3.62, + "learning_rate": 7.193078961843723e-06, + "loss": 0.9763, + "step": 48950 + }, + { + "epoch": 3.62, + "learning_rate": 7.192696127947159e-06, + "loss": 1.1499, + "step": 48951 + }, + { + "epoch": 3.62, + "learning_rate": 7.192313298516973e-06, + "loss": 0.9709, + "step": 48952 + }, + { + "epoch": 3.62, + "learning_rate": 7.191930473553779e-06, + "loss": 1.0459, + "step": 48953 + }, + { + "epoch": 3.62, + "learning_rate": 7.1915476530581815e-06, + "loss": 1.0178, + "step": 48954 + }, + { + "epoch": 3.62, + "learning_rate": 7.191164837030793e-06, + "loss": 1.0261, + "step": 48955 + }, + { + "epoch": 3.62, + "learning_rate": 7.19078202547222e-06, + "loss": 0.9559, + "step": 48956 + }, + { + "epoch": 3.62, + "learning_rate": 7.190399218383075e-06, + "loss": 1.0045, + "step": 48957 + }, + { + "epoch": 3.62, + "learning_rate": 7.190016415763961e-06, + "loss": 1.0265, + "step": 48958 + }, + { + "epoch": 3.62, + "learning_rate": 7.189633617615493e-06, + "loss": 0.9579, + "step": 48959 + }, + { + "epoch": 3.62, + "learning_rate": 7.18925082393828e-06, + "loss": 0.9496, + "step": 48960 + }, + { + "epoch": 3.62, + "learning_rate": 7.188868034732923e-06, + "loss": 0.9811, + "step": 48961 + }, + { + "epoch": 3.62, + "learning_rate": 7.1884852500000436e-06, + "loss": 1.0761, + "step": 48962 + }, + { + "epoch": 3.62, + "learning_rate": 7.188102469740239e-06, + "loss": 1.0775, + "step": 48963 + }, + { + "epoch": 3.62, + "learning_rate": 7.187719693954125e-06, + "loss": 1.0176, + "step": 48964 + }, + { + "epoch": 3.62, + "learning_rate": 7.1873369226423064e-06, + "loss": 1.0273, + "step": 48965 + }, + { + "epoch": 3.62, + "learning_rate": 7.186954155805397e-06, + "loss": 1.0694, + "step": 48966 + }, + { + "epoch": 3.62, + "learning_rate": 7.186571393444002e-06, + "loss": 1.0486, + "step": 48967 + }, + { + "epoch": 3.62, + "learning_rate": 7.186188635558732e-06, + "loss": 0.9892, + "step": 48968 + }, + { + "epoch": 3.62, + "learning_rate": 7.1858058821501966e-06, + "loss": 0.9289, + "step": 48969 + }, + { + "epoch": 3.62, + "learning_rate": 7.185423133219002e-06, + "loss": 1.0119, + "step": 48970 + }, + { + "epoch": 3.62, + "learning_rate": 7.18504038876576e-06, + "loss": 1.049, + "step": 48971 + }, + { + "epoch": 3.62, + "learning_rate": 7.184657648791078e-06, + "loss": 1.0031, + "step": 48972 + }, + { + "epoch": 3.62, + "learning_rate": 7.184274913295566e-06, + "loss": 0.9505, + "step": 48973 + }, + { + "epoch": 3.62, + "learning_rate": 7.18389218227983e-06, + "loss": 1.0023, + "step": 48974 + }, + { + "epoch": 3.62, + "learning_rate": 7.183509455744484e-06, + "loss": 1.0491, + "step": 48975 + }, + { + "epoch": 3.62, + "learning_rate": 7.183126733690133e-06, + "loss": 1.0722, + "step": 48976 + }, + { + "epoch": 3.62, + "learning_rate": 7.182744016117388e-06, + "loss": 1.0442, + "step": 48977 + }, + { + "epoch": 3.62, + "learning_rate": 7.182361303026859e-06, + "loss": 0.948, + "step": 48978 + }, + { + "epoch": 3.62, + "learning_rate": 7.1819785944191476e-06, + "loss": 0.9429, + "step": 48979 + }, + { + "epoch": 3.62, + "learning_rate": 7.181595890294873e-06, + "loss": 0.963, + "step": 48980 + }, + { + "epoch": 3.62, + "learning_rate": 7.181213190654633e-06, + "loss": 1.0123, + "step": 48981 + }, + { + "epoch": 3.62, + "learning_rate": 7.180830495499049e-06, + "loss": 0.8631, + "step": 48982 + }, + { + "epoch": 3.62, + "learning_rate": 7.180447804828718e-06, + "loss": 1.0048, + "step": 48983 + }, + { + "epoch": 3.62, + "learning_rate": 7.180065118644259e-06, + "loss": 0.951, + "step": 48984 + }, + { + "epoch": 3.62, + "learning_rate": 7.17968243694627e-06, + "loss": 1.1221, + "step": 48985 + }, + { + "epoch": 3.62, + "learning_rate": 7.179299759735373e-06, + "loss": 0.954, + "step": 48986 + }, + { + "epoch": 3.62, + "learning_rate": 7.1789170870121675e-06, + "loss": 0.9998, + "step": 48987 + }, + { + "epoch": 3.62, + "learning_rate": 7.178534418777261e-06, + "loss": 1.0378, + "step": 48988 + }, + { + "epoch": 3.62, + "learning_rate": 7.17815175503127e-06, + "loss": 1.0999, + "step": 48989 + }, + { + "epoch": 3.62, + "learning_rate": 7.177769095774796e-06, + "loss": 0.9731, + "step": 48990 + }, + { + "epoch": 3.62, + "learning_rate": 7.177386441008454e-06, + "loss": 1.0284, + "step": 48991 + }, + { + "epoch": 3.62, + "learning_rate": 7.177003790732846e-06, + "loss": 1.0187, + "step": 48992 + }, + { + "epoch": 3.62, + "learning_rate": 7.176621144948588e-06, + "loss": 1.0653, + "step": 48993 + }, + { + "epoch": 3.62, + "learning_rate": 7.176238503656283e-06, + "loss": 0.9876, + "step": 48994 + }, + { + "epoch": 3.62, + "learning_rate": 7.175855866856543e-06, + "loss": 0.9328, + "step": 48995 + }, + { + "epoch": 3.62, + "learning_rate": 7.175473234549978e-06, + "loss": 0.9884, + "step": 48996 + }, + { + "epoch": 3.62, + "learning_rate": 7.1750906067371916e-06, + "loss": 1.0658, + "step": 48997 + }, + { + "epoch": 3.62, + "learning_rate": 7.1747079834187984e-06, + "loss": 0.9422, + "step": 48998 + }, + { + "epoch": 3.62, + "learning_rate": 7.1743253645954e-06, + "loss": 0.9907, + "step": 48999 + }, + { + "epoch": 3.62, + "learning_rate": 7.173942750267614e-06, + "loss": 1.0293, + "step": 49000 + }, + { + "epoch": 3.62, + "learning_rate": 7.1735601404360425e-06, + "loss": 1.1294, + "step": 49001 + }, + { + "epoch": 3.62, + "learning_rate": 7.173177535101296e-06, + "loss": 0.9682, + "step": 49002 + }, + { + "epoch": 3.62, + "learning_rate": 7.172794934263984e-06, + "loss": 0.9716, + "step": 49003 + }, + { + "epoch": 3.62, + "learning_rate": 7.172412337924717e-06, + "loss": 0.9389, + "step": 49004 + }, + { + "epoch": 3.62, + "learning_rate": 7.172029746084098e-06, + "loss": 1.0211, + "step": 49005 + }, + { + "epoch": 3.62, + "learning_rate": 7.171647158742744e-06, + "loss": 1.0504, + "step": 49006 + }, + { + "epoch": 3.62, + "learning_rate": 7.171264575901257e-06, + "loss": 0.9705, + "step": 49007 + }, + { + "epoch": 3.62, + "learning_rate": 7.1708819975602454e-06, + "loss": 0.9758, + "step": 49008 + }, + { + "epoch": 3.62, + "learning_rate": 7.170499423720321e-06, + "loss": 1.0739, + "step": 49009 + }, + { + "epoch": 3.62, + "learning_rate": 7.170116854382091e-06, + "loss": 1.0009, + "step": 49010 + }, + { + "epoch": 3.62, + "learning_rate": 7.169734289546166e-06, + "loss": 1.0571, + "step": 49011 + }, + { + "epoch": 3.62, + "learning_rate": 7.169351729213152e-06, + "loss": 1.1008, + "step": 49012 + }, + { + "epoch": 3.62, + "learning_rate": 7.1689691733836595e-06, + "loss": 0.9516, + "step": 49013 + }, + { + "epoch": 3.62, + "learning_rate": 7.168586622058296e-06, + "loss": 0.9695, + "step": 49014 + }, + { + "epoch": 3.62, + "learning_rate": 7.168204075237672e-06, + "loss": 1.0837, + "step": 49015 + }, + { + "epoch": 3.62, + "learning_rate": 7.167821532922394e-06, + "loss": 1.0064, + "step": 49016 + }, + { + "epoch": 3.62, + "learning_rate": 7.167438995113071e-06, + "loss": 1.0569, + "step": 49017 + }, + { + "epoch": 3.62, + "learning_rate": 7.167056461810313e-06, + "loss": 0.9956, + "step": 49018 + }, + { + "epoch": 3.62, + "learning_rate": 7.166673933014725e-06, + "loss": 1.0368, + "step": 49019 + }, + { + "epoch": 3.62, + "learning_rate": 7.166291408726922e-06, + "loss": 1.0168, + "step": 49020 + }, + { + "epoch": 3.62, + "learning_rate": 7.165908888947506e-06, + "loss": 0.9614, + "step": 49021 + }, + { + "epoch": 3.62, + "learning_rate": 7.1655263736770894e-06, + "loss": 1.0187, + "step": 49022 + }, + { + "epoch": 3.62, + "learning_rate": 7.16514386291628e-06, + "loss": 1.0274, + "step": 49023 + }, + { + "epoch": 3.62, + "learning_rate": 7.164761356665686e-06, + "loss": 1.0741, + "step": 49024 + }, + { + "epoch": 3.62, + "learning_rate": 7.16437885492592e-06, + "loss": 0.9782, + "step": 49025 + }, + { + "epoch": 3.62, + "learning_rate": 7.163996357697579e-06, + "loss": 0.9412, + "step": 49026 + }, + { + "epoch": 3.62, + "learning_rate": 7.163613864981286e-06, + "loss": 0.9886, + "step": 49027 + }, + { + "epoch": 3.62, + "learning_rate": 7.163231376777636e-06, + "loss": 1.1961, + "step": 49028 + }, + { + "epoch": 3.62, + "learning_rate": 7.162848893087251e-06, + "loss": 1.0544, + "step": 49029 + }, + { + "epoch": 3.62, + "learning_rate": 7.162466413910727e-06, + "loss": 1.0107, + "step": 49030 + }, + { + "epoch": 3.62, + "learning_rate": 7.162083939248685e-06, + "loss": 0.9706, + "step": 49031 + }, + { + "epoch": 3.62, + "learning_rate": 7.161701469101718e-06, + "loss": 1.0204, + "step": 49032 + }, + { + "epoch": 3.62, + "learning_rate": 7.161319003470453e-06, + "loss": 0.8597, + "step": 49033 + }, + { + "epoch": 3.62, + "learning_rate": 7.1609365423554854e-06, + "loss": 1.0222, + "step": 49034 + }, + { + "epoch": 3.62, + "learning_rate": 7.160554085757424e-06, + "loss": 0.9487, + "step": 49035 + }, + { + "epoch": 3.62, + "learning_rate": 7.160171633676883e-06, + "loss": 1.0718, + "step": 49036 + }, + { + "epoch": 3.62, + "learning_rate": 7.159789186114467e-06, + "loss": 1.0425, + "step": 49037 + }, + { + "epoch": 3.62, + "learning_rate": 7.1594067430707866e-06, + "loss": 0.9728, + "step": 49038 + }, + { + "epoch": 3.62, + "learning_rate": 7.159024304546449e-06, + "loss": 1.0479, + "step": 49039 + }, + { + "epoch": 3.62, + "learning_rate": 7.158641870542065e-06, + "loss": 1.085, + "step": 49040 + }, + { + "epoch": 3.62, + "learning_rate": 7.158259441058237e-06, + "loss": 1.032, + "step": 49041 + }, + { + "epoch": 3.62, + "learning_rate": 7.157877016095583e-06, + "loss": 0.9852, + "step": 49042 + }, + { + "epoch": 3.62, + "learning_rate": 7.157494595654703e-06, + "loss": 0.9809, + "step": 49043 + }, + { + "epoch": 3.62, + "learning_rate": 7.157112179736208e-06, + "loss": 0.9248, + "step": 49044 + }, + { + "epoch": 3.62, + "learning_rate": 7.156729768340708e-06, + "loss": 0.8932, + "step": 49045 + }, + { + "epoch": 3.62, + "learning_rate": 7.156347361468807e-06, + "loss": 0.9492, + "step": 49046 + }, + { + "epoch": 3.62, + "learning_rate": 7.155964959121121e-06, + "loss": 1.0204, + "step": 49047 + }, + { + "epoch": 3.62, + "learning_rate": 7.155582561298251e-06, + "loss": 0.8934, + "step": 49048 + }, + { + "epoch": 3.62, + "learning_rate": 7.155200168000811e-06, + "loss": 0.9498, + "step": 49049 + }, + { + "epoch": 3.62, + "learning_rate": 7.1548177792294035e-06, + "loss": 1.0253, + "step": 49050 + }, + { + "epoch": 3.62, + "learning_rate": 7.154435394984647e-06, + "loss": 1.0772, + "step": 49051 + }, + { + "epoch": 3.62, + "learning_rate": 7.154053015267139e-06, + "loss": 1.0776, + "step": 49052 + }, + { + "epoch": 3.62, + "learning_rate": 7.15367064007749e-06, + "loss": 1.0444, + "step": 49053 + }, + { + "epoch": 3.62, + "learning_rate": 7.153288269416312e-06, + "loss": 1.064, + "step": 49054 + }, + { + "epoch": 3.62, + "learning_rate": 7.1529059032842105e-06, + "loss": 1.0168, + "step": 49055 + }, + { + "epoch": 3.62, + "learning_rate": 7.1525235416817974e-06, + "loss": 0.9889, + "step": 49056 + }, + { + "epoch": 3.62, + "learning_rate": 7.152141184609676e-06, + "loss": 1.1167, + "step": 49057 + }, + { + "epoch": 3.62, + "learning_rate": 7.1517588320684585e-06, + "loss": 0.8752, + "step": 49058 + }, + { + "epoch": 3.62, + "learning_rate": 7.15137648405875e-06, + "loss": 1.0029, + "step": 49059 + }, + { + "epoch": 3.63, + "learning_rate": 7.150994140581163e-06, + "loss": 1.0627, + "step": 49060 + }, + { + "epoch": 3.63, + "learning_rate": 7.150611801636303e-06, + "loss": 1.1244, + "step": 49061 + }, + { + "epoch": 3.63, + "learning_rate": 7.150229467224779e-06, + "loss": 0.9279, + "step": 49062 + }, + { + "epoch": 3.63, + "learning_rate": 7.1498471373472e-06, + "loss": 1.0593, + "step": 49063 + }, + { + "epoch": 3.63, + "learning_rate": 7.149464812004171e-06, + "loss": 0.9974, + "step": 49064 + }, + { + "epoch": 3.63, + "learning_rate": 7.149082491196305e-06, + "loss": 1.0521, + "step": 49065 + }, + { + "epoch": 3.63, + "learning_rate": 7.148700174924206e-06, + "loss": 1.0213, + "step": 49066 + }, + { + "epoch": 3.63, + "learning_rate": 7.148317863188487e-06, + "loss": 0.9795, + "step": 49067 + }, + { + "epoch": 3.63, + "learning_rate": 7.147935555989749e-06, + "loss": 1.0577, + "step": 49068 + }, + { + "epoch": 3.63, + "learning_rate": 7.1475532533286095e-06, + "loss": 1.1147, + "step": 49069 + }, + { + "epoch": 3.63, + "learning_rate": 7.147170955205669e-06, + "loss": 1.0525, + "step": 49070 + }, + { + "epoch": 3.63, + "learning_rate": 7.14678866162154e-06, + "loss": 1.0382, + "step": 49071 + }, + { + "epoch": 3.63, + "learning_rate": 7.1464063725768325e-06, + "loss": 0.9653, + "step": 49072 + }, + { + "epoch": 3.63, + "learning_rate": 7.146024088072144e-06, + "loss": 0.8927, + "step": 49073 + }, + { + "epoch": 3.63, + "learning_rate": 7.145641808108098e-06, + "loss": 0.9847, + "step": 49074 + }, + { + "epoch": 3.63, + "learning_rate": 7.145259532685288e-06, + "loss": 0.9061, + "step": 49075 + }, + { + "epoch": 3.63, + "learning_rate": 7.144877261804336e-06, + "loss": 0.9822, + "step": 49076 + }, + { + "epoch": 3.63, + "learning_rate": 7.144494995465837e-06, + "loss": 1.0182, + "step": 49077 + }, + { + "epoch": 3.63, + "learning_rate": 7.144112733670412e-06, + "loss": 0.949, + "step": 49078 + }, + { + "epoch": 3.63, + "learning_rate": 7.143730476418655e-06, + "loss": 1.079, + "step": 49079 + }, + { + "epoch": 3.63, + "learning_rate": 7.143348223711191e-06, + "loss": 1.0502, + "step": 49080 + }, + { + "epoch": 3.63, + "learning_rate": 7.142965975548615e-06, + "loss": 1.0087, + "step": 49081 + }, + { + "epoch": 3.63, + "learning_rate": 7.1425837319315384e-06, + "loss": 0.9813, + "step": 49082 + }, + { + "epoch": 3.63, + "learning_rate": 7.142201492860572e-06, + "loss": 1.0234, + "step": 49083 + }, + { + "epoch": 3.63, + "learning_rate": 7.141819258336319e-06, + "loss": 0.9904, + "step": 49084 + }, + { + "epoch": 3.63, + "learning_rate": 7.1414370283593935e-06, + "loss": 0.953, + "step": 49085 + }, + { + "epoch": 3.63, + "learning_rate": 7.141054802930398e-06, + "loss": 1.0852, + "step": 49086 + }, + { + "epoch": 3.63, + "learning_rate": 7.140672582049947e-06, + "loss": 1.042, + "step": 49087 + }, + { + "epoch": 3.63, + "learning_rate": 7.140290365718642e-06, + "loss": 1.0217, + "step": 49088 + }, + { + "epoch": 3.63, + "learning_rate": 7.1399081539370965e-06, + "loss": 1.0181, + "step": 49089 + }, + { + "epoch": 3.63, + "learning_rate": 7.1395259467059155e-06, + "loss": 0.983, + "step": 49090 + }, + { + "epoch": 3.63, + "learning_rate": 7.139143744025706e-06, + "loss": 0.9621, + "step": 49091 + }, + { + "epoch": 3.63, + "learning_rate": 7.138761545897081e-06, + "loss": 0.9998, + "step": 49092 + }, + { + "epoch": 3.63, + "learning_rate": 7.1383793523206415e-06, + "loss": 1.0423, + "step": 49093 + }, + { + "epoch": 3.63, + "learning_rate": 7.137997163297004e-06, + "loss": 1.0145, + "step": 49094 + }, + { + "epoch": 3.63, + "learning_rate": 7.137614978826768e-06, + "loss": 1.0792, + "step": 49095 + }, + { + "epoch": 3.63, + "learning_rate": 7.137232798910549e-06, + "loss": 1.0522, + "step": 49096 + }, + { + "epoch": 3.63, + "learning_rate": 7.136850623548948e-06, + "loss": 0.8729, + "step": 49097 + }, + { + "epoch": 3.63, + "learning_rate": 7.136468452742581e-06, + "loss": 1.0297, + "step": 49098 + }, + { + "epoch": 3.63, + "learning_rate": 7.13608628649205e-06, + "loss": 0.965, + "step": 49099 + }, + { + "epoch": 3.63, + "learning_rate": 7.135704124797964e-06, + "loss": 0.9831, + "step": 49100 + }, + { + "epoch": 3.63, + "learning_rate": 7.135321967660932e-06, + "loss": 0.9822, + "step": 49101 + }, + { + "epoch": 3.63, + "learning_rate": 7.134939815081561e-06, + "loss": 1.0298, + "step": 49102 + }, + { + "epoch": 3.63, + "learning_rate": 7.134557667060461e-06, + "loss": 0.9769, + "step": 49103 + }, + { + "epoch": 3.63, + "learning_rate": 7.134175523598237e-06, + "loss": 1.0598, + "step": 49104 + }, + { + "epoch": 3.63, + "learning_rate": 7.133793384695501e-06, + "loss": 0.9846, + "step": 49105 + }, + { + "epoch": 3.63, + "learning_rate": 7.133411250352856e-06, + "loss": 0.9069, + "step": 49106 + }, + { + "epoch": 3.63, + "learning_rate": 7.133029120570915e-06, + "loss": 1.0053, + "step": 49107 + }, + { + "epoch": 3.63, + "learning_rate": 7.132646995350283e-06, + "loss": 0.9595, + "step": 49108 + }, + { + "epoch": 3.63, + "learning_rate": 7.132264874691569e-06, + "loss": 1.0989, + "step": 49109 + }, + { + "epoch": 3.63, + "learning_rate": 7.131882758595381e-06, + "loss": 0.8925, + "step": 49110 + }, + { + "epoch": 3.63, + "learning_rate": 7.1315006470623236e-06, + "loss": 0.9965, + "step": 49111 + }, + { + "epoch": 3.63, + "learning_rate": 7.131118540093012e-06, + "loss": 0.9959, + "step": 49112 + }, + { + "epoch": 3.63, + "learning_rate": 7.130736437688046e-06, + "loss": 1.0138, + "step": 49113 + }, + { + "epoch": 3.63, + "learning_rate": 7.130354339848039e-06, + "loss": 0.9677, + "step": 49114 + }, + { + "epoch": 3.63, + "learning_rate": 7.129972246573595e-06, + "loss": 1.018, + "step": 49115 + }, + { + "epoch": 3.63, + "learning_rate": 7.129590157865327e-06, + "loss": 0.9701, + "step": 49116 + }, + { + "epoch": 3.63, + "learning_rate": 7.129208073723839e-06, + "loss": 1.0029, + "step": 49117 + }, + { + "epoch": 3.63, + "learning_rate": 7.1288259941497405e-06, + "loss": 1.059, + "step": 49118 + }, + { + "epoch": 3.63, + "learning_rate": 7.1284439191436415e-06, + "loss": 1.0505, + "step": 49119 + }, + { + "epoch": 3.63, + "learning_rate": 7.12806184870614e-06, + "loss": 0.984, + "step": 49120 + }, + { + "epoch": 3.63, + "learning_rate": 7.127679782837857e-06, + "loss": 1.079, + "step": 49121 + }, + { + "epoch": 3.63, + "learning_rate": 7.12729772153939e-06, + "loss": 0.9225, + "step": 49122 + }, + { + "epoch": 3.63, + "learning_rate": 7.1269156648113555e-06, + "loss": 1.0635, + "step": 49123 + }, + { + "epoch": 3.63, + "learning_rate": 7.126533612654351e-06, + "loss": 1.0744, + "step": 49124 + }, + { + "epoch": 3.63, + "learning_rate": 7.1261515650689985e-06, + "loss": 0.9813, + "step": 49125 + }, + { + "epoch": 3.63, + "learning_rate": 7.1257695220558896e-06, + "loss": 0.9599, + "step": 49126 + }, + { + "epoch": 3.63, + "learning_rate": 7.125387483615648e-06, + "loss": 0.9778, + "step": 49127 + }, + { + "epoch": 3.63, + "learning_rate": 7.1250054497488704e-06, + "loss": 0.9456, + "step": 49128 + }, + { + "epoch": 3.63, + "learning_rate": 7.124623420456167e-06, + "loss": 0.9874, + "step": 49129 + }, + { + "epoch": 3.63, + "learning_rate": 7.1242413957381475e-06, + "loss": 1.0255, + "step": 49130 + }, + { + "epoch": 3.63, + "learning_rate": 7.123859375595418e-06, + "loss": 1.0513, + "step": 49131 + }, + { + "epoch": 3.63, + "learning_rate": 7.123477360028588e-06, + "loss": 1.0291, + "step": 49132 + }, + { + "epoch": 3.63, + "learning_rate": 7.123095349038263e-06, + "loss": 1.1051, + "step": 49133 + }, + { + "epoch": 3.63, + "learning_rate": 7.1227133426250535e-06, + "loss": 1.1013, + "step": 49134 + }, + { + "epoch": 3.63, + "learning_rate": 7.122331340789565e-06, + "loss": 1.0616, + "step": 49135 + }, + { + "epoch": 3.63, + "learning_rate": 7.121949343532407e-06, + "loss": 1.0666, + "step": 49136 + }, + { + "epoch": 3.63, + "learning_rate": 7.121567350854188e-06, + "loss": 0.9572, + "step": 49137 + }, + { + "epoch": 3.63, + "learning_rate": 7.121185362755511e-06, + "loss": 1.1058, + "step": 49138 + }, + { + "epoch": 3.63, + "learning_rate": 7.1208033792369915e-06, + "loss": 1.0476, + "step": 49139 + }, + { + "epoch": 3.63, + "learning_rate": 7.120421400299225e-06, + "loss": 1.0361, + "step": 49140 + }, + { + "epoch": 3.63, + "learning_rate": 7.120039425942834e-06, + "loss": 1.0657, + "step": 49141 + }, + { + "epoch": 3.63, + "learning_rate": 7.119657456168414e-06, + "loss": 1.0303, + "step": 49142 + }, + { + "epoch": 3.63, + "learning_rate": 7.119275490976583e-06, + "loss": 1.0148, + "step": 49143 + }, + { + "epoch": 3.63, + "learning_rate": 7.1188935303679365e-06, + "loss": 1.0486, + "step": 49144 + }, + { + "epoch": 3.63, + "learning_rate": 7.118511574343097e-06, + "loss": 1.0605, + "step": 49145 + }, + { + "epoch": 3.63, + "learning_rate": 7.118129622902661e-06, + "loss": 1.0671, + "step": 49146 + }, + { + "epoch": 3.63, + "learning_rate": 7.1177476760472375e-06, + "loss": 1.0497, + "step": 49147 + }, + { + "epoch": 3.63, + "learning_rate": 7.117365733777438e-06, + "loss": 1.1068, + "step": 49148 + }, + { + "epoch": 3.63, + "learning_rate": 7.116983796093867e-06, + "loss": 0.9774, + "step": 49149 + }, + { + "epoch": 3.63, + "learning_rate": 7.116601862997136e-06, + "loss": 1.0006, + "step": 49150 + }, + { + "epoch": 3.63, + "learning_rate": 7.116219934487847e-06, + "loss": 1.0429, + "step": 49151 + }, + { + "epoch": 3.63, + "learning_rate": 7.1158380105666135e-06, + "loss": 1.0219, + "step": 49152 + }, + { + "epoch": 3.63, + "learning_rate": 7.1154560912340386e-06, + "loss": 1.0409, + "step": 49153 + }, + { + "epoch": 3.63, + "learning_rate": 7.115074176490734e-06, + "loss": 0.9562, + "step": 49154 + }, + { + "epoch": 3.63, + "learning_rate": 7.114692266337304e-06, + "loss": 1.031, + "step": 49155 + }, + { + "epoch": 3.63, + "learning_rate": 7.114310360774357e-06, + "loss": 0.9796, + "step": 49156 + }, + { + "epoch": 3.63, + "learning_rate": 7.113928459802502e-06, + "loss": 1.009, + "step": 49157 + }, + { + "epoch": 3.63, + "learning_rate": 7.1135465634223435e-06, + "loss": 0.944, + "step": 49158 + }, + { + "epoch": 3.63, + "learning_rate": 7.113164671634494e-06, + "loss": 1.0004, + "step": 49159 + }, + { + "epoch": 3.63, + "learning_rate": 7.1127827844395535e-06, + "loss": 0.8595, + "step": 49160 + }, + { + "epoch": 3.63, + "learning_rate": 7.112400901838139e-06, + "loss": 0.9388, + "step": 49161 + }, + { + "epoch": 3.63, + "learning_rate": 7.1120190238308495e-06, + "loss": 0.9165, + "step": 49162 + }, + { + "epoch": 3.63, + "learning_rate": 7.1116371504183005e-06, + "loss": 0.9576, + "step": 49163 + }, + { + "epoch": 3.63, + "learning_rate": 7.111255281601093e-06, + "loss": 0.9717, + "step": 49164 + }, + { + "epoch": 3.63, + "learning_rate": 7.110873417379838e-06, + "loss": 1.1601, + "step": 49165 + }, + { + "epoch": 3.63, + "learning_rate": 7.110491557755144e-06, + "loss": 0.9936, + "step": 49166 + }, + { + "epoch": 3.63, + "learning_rate": 7.110109702727611e-06, + "loss": 1.1002, + "step": 49167 + }, + { + "epoch": 3.63, + "learning_rate": 7.109727852297859e-06, + "loss": 0.9656, + "step": 49168 + }, + { + "epoch": 3.63, + "learning_rate": 7.109346006466481e-06, + "loss": 1.0114, + "step": 49169 + }, + { + "epoch": 3.63, + "learning_rate": 7.108964165234099e-06, + "loss": 0.9905, + "step": 49170 + }, + { + "epoch": 3.63, + "learning_rate": 7.108582328601308e-06, + "loss": 1.0232, + "step": 49171 + }, + { + "epoch": 3.63, + "learning_rate": 7.108200496568727e-06, + "loss": 1.0755, + "step": 49172 + }, + { + "epoch": 3.63, + "learning_rate": 7.107818669136952e-06, + "loss": 0.9783, + "step": 49173 + }, + { + "epoch": 3.63, + "learning_rate": 7.1074368463066005e-06, + "loss": 0.9954, + "step": 49174 + }, + { + "epoch": 3.63, + "learning_rate": 7.1070550280782735e-06, + "loss": 1.0808, + "step": 49175 + }, + { + "epoch": 3.63, + "learning_rate": 7.1066732144525796e-06, + "loss": 0.9411, + "step": 49176 + }, + { + "epoch": 3.63, + "learning_rate": 7.10629140543013e-06, + "loss": 1.0925, + "step": 49177 + }, + { + "epoch": 3.63, + "learning_rate": 7.105909601011527e-06, + "loss": 1.0162, + "step": 49178 + }, + { + "epoch": 3.63, + "learning_rate": 7.105527801197382e-06, + "loss": 0.9237, + "step": 49179 + }, + { + "epoch": 3.63, + "learning_rate": 7.1051460059882984e-06, + "loss": 0.9457, + "step": 49180 + }, + { + "epoch": 3.63, + "learning_rate": 7.104764215384888e-06, + "loss": 0.9859, + "step": 49181 + }, + { + "epoch": 3.63, + "learning_rate": 7.104382429387755e-06, + "loss": 0.8549, + "step": 49182 + }, + { + "epoch": 3.63, + "learning_rate": 7.104000647997511e-06, + "loss": 0.8682, + "step": 49183 + }, + { + "epoch": 3.63, + "learning_rate": 7.103618871214761e-06, + "loss": 1.0713, + "step": 49184 + }, + { + "epoch": 3.63, + "learning_rate": 7.103237099040107e-06, + "loss": 0.9968, + "step": 49185 + }, + { + "epoch": 3.63, + "learning_rate": 7.1028553314741666e-06, + "loss": 0.9589, + "step": 49186 + }, + { + "epoch": 3.63, + "learning_rate": 7.102473568517536e-06, + "loss": 1.0412, + "step": 49187 + }, + { + "epoch": 3.63, + "learning_rate": 7.1020918101708346e-06, + "loss": 0.9774, + "step": 49188 + }, + { + "epoch": 3.63, + "learning_rate": 7.101710056434658e-06, + "loss": 1.0165, + "step": 49189 + }, + { + "epoch": 3.63, + "learning_rate": 7.101328307309625e-06, + "loss": 0.9122, + "step": 49190 + }, + { + "epoch": 3.63, + "learning_rate": 7.100946562796331e-06, + "loss": 1.069, + "step": 49191 + }, + { + "epoch": 3.63, + "learning_rate": 7.100564822895395e-06, + "loss": 0.9534, + "step": 49192 + }, + { + "epoch": 3.63, + "learning_rate": 7.100183087607417e-06, + "loss": 1.0937, + "step": 49193 + }, + { + "epoch": 3.63, + "learning_rate": 7.099801356933005e-06, + "loss": 1.0954, + "step": 49194 + }, + { + "epoch": 3.63, + "learning_rate": 7.0994196308727705e-06, + "loss": 1.0226, + "step": 49195 + }, + { + "epoch": 3.64, + "learning_rate": 7.099037909427313e-06, + "loss": 1.0215, + "step": 49196 + }, + { + "epoch": 3.64, + "learning_rate": 7.098656192597249e-06, + "loss": 1.0363, + "step": 49197 + }, + { + "epoch": 3.64, + "learning_rate": 7.098274480383179e-06, + "loss": 0.9716, + "step": 49198 + }, + { + "epoch": 3.64, + "learning_rate": 7.097892772785715e-06, + "loss": 0.995, + "step": 49199 + }, + { + "epoch": 3.64, + "learning_rate": 7.097511069805459e-06, + "loss": 1.0541, + "step": 49200 + }, + { + "epoch": 3.64, + "learning_rate": 7.097129371443025e-06, + "loss": 1.0301, + "step": 49201 + }, + { + "epoch": 3.64, + "learning_rate": 7.096747677699015e-06, + "loss": 1.0286, + "step": 49202 + }, + { + "epoch": 3.64, + "learning_rate": 7.096365988574037e-06, + "loss": 1.0781, + "step": 49203 + }, + { + "epoch": 3.64, + "learning_rate": 7.0959843040687004e-06, + "loss": 0.9635, + "step": 49204 + }, + { + "epoch": 3.64, + "learning_rate": 7.095602624183609e-06, + "loss": 0.9612, + "step": 49205 + }, + { + "epoch": 3.64, + "learning_rate": 7.095220948919376e-06, + "loss": 0.8932, + "step": 49206 + }, + { + "epoch": 3.64, + "learning_rate": 7.094839278276602e-06, + "loss": 0.9697, + "step": 49207 + }, + { + "epoch": 3.64, + "learning_rate": 7.094457612255898e-06, + "loss": 0.9606, + "step": 49208 + }, + { + "epoch": 3.64, + "learning_rate": 7.09407595085787e-06, + "loss": 1.163, + "step": 49209 + }, + { + "epoch": 3.64, + "learning_rate": 7.093694294083126e-06, + "loss": 0.9288, + "step": 49210 + }, + { + "epoch": 3.64, + "learning_rate": 7.0933126419322766e-06, + "loss": 1.1163, + "step": 49211 + }, + { + "epoch": 3.64, + "learning_rate": 7.0929309944059175e-06, + "loss": 1.0199, + "step": 49212 + }, + { + "epoch": 3.64, + "learning_rate": 7.092549351504671e-06, + "loss": 1.025, + "step": 49213 + }, + { + "epoch": 3.64, + "learning_rate": 7.09216771322913e-06, + "loss": 0.9078, + "step": 49214 + }, + { + "epoch": 3.64, + "learning_rate": 7.091786079579914e-06, + "loss": 0.9979, + "step": 49215 + }, + { + "epoch": 3.64, + "learning_rate": 7.09140445055762e-06, + "loss": 1.0518, + "step": 49216 + }, + { + "epoch": 3.64, + "learning_rate": 7.091022826162866e-06, + "loss": 0.9874, + "step": 49217 + }, + { + "epoch": 3.64, + "learning_rate": 7.090641206396248e-06, + "loss": 0.9242, + "step": 49218 + }, + { + "epoch": 3.64, + "learning_rate": 7.090259591258382e-06, + "loss": 1.0391, + "step": 49219 + }, + { + "epoch": 3.64, + "learning_rate": 7.089877980749866e-06, + "loss": 0.9342, + "step": 49220 + }, + { + "epoch": 3.64, + "learning_rate": 7.089496374871318e-06, + "loss": 1.0528, + "step": 49221 + }, + { + "epoch": 3.64, + "learning_rate": 7.089114773623338e-06, + "loss": 1.0266, + "step": 49222 + }, + { + "epoch": 3.64, + "learning_rate": 7.088733177006532e-06, + "loss": 1.0375, + "step": 49223 + }, + { + "epoch": 3.64, + "learning_rate": 7.088351585021514e-06, + "loss": 0.9997, + "step": 49224 + }, + { + "epoch": 3.64, + "learning_rate": 7.087969997668883e-06, + "loss": 0.9277, + "step": 49225 + }, + { + "epoch": 3.64, + "learning_rate": 7.087588414949253e-06, + "loss": 0.9258, + "step": 49226 + }, + { + "epoch": 3.64, + "learning_rate": 7.087206836863225e-06, + "loss": 1.0649, + "step": 49227 + }, + { + "epoch": 3.64, + "learning_rate": 7.0868252634114136e-06, + "loss": 1.0456, + "step": 49228 + }, + { + "epoch": 3.64, + "learning_rate": 7.086443694594418e-06, + "loss": 0.9781, + "step": 49229 + }, + { + "epoch": 3.64, + "learning_rate": 7.086062130412851e-06, + "loss": 0.9631, + "step": 49230 + }, + { + "epoch": 3.64, + "learning_rate": 7.08568057086732e-06, + "loss": 1.0998, + "step": 49231 + }, + { + "epoch": 3.64, + "learning_rate": 7.085299015958422e-06, + "loss": 0.9558, + "step": 49232 + }, + { + "epoch": 3.64, + "learning_rate": 7.084917465686778e-06, + "loss": 1.1068, + "step": 49233 + }, + { + "epoch": 3.64, + "learning_rate": 7.084535920052984e-06, + "loss": 0.9445, + "step": 49234 + }, + { + "epoch": 3.64, + "learning_rate": 7.084154379057657e-06, + "loss": 1.0151, + "step": 49235 + }, + { + "epoch": 3.64, + "learning_rate": 7.083772842701392e-06, + "loss": 0.9643, + "step": 49236 + }, + { + "epoch": 3.64, + "learning_rate": 7.083391310984809e-06, + "loss": 1.0748, + "step": 49237 + }, + { + "epoch": 3.64, + "learning_rate": 7.083009783908503e-06, + "loss": 0.9531, + "step": 49238 + }, + { + "epoch": 3.64, + "learning_rate": 7.082628261473092e-06, + "loss": 1.0158, + "step": 49239 + }, + { + "epoch": 3.64, + "learning_rate": 7.082246743679177e-06, + "loss": 0.9682, + "step": 49240 + }, + { + "epoch": 3.64, + "learning_rate": 7.081865230527363e-06, + "loss": 1.0308, + "step": 49241 + }, + { + "epoch": 3.64, + "learning_rate": 7.081483722018261e-06, + "loss": 1.0007, + "step": 49242 + }, + { + "epoch": 3.64, + "learning_rate": 7.081102218152476e-06, + "loss": 0.9581, + "step": 49243 + }, + { + "epoch": 3.64, + "learning_rate": 7.080720718930616e-06, + "loss": 0.9844, + "step": 49244 + }, + { + "epoch": 3.64, + "learning_rate": 7.080339224353286e-06, + "loss": 1.0097, + "step": 49245 + }, + { + "epoch": 3.64, + "learning_rate": 7.079957734421097e-06, + "loss": 0.987, + "step": 49246 + }, + { + "epoch": 3.64, + "learning_rate": 7.079576249134652e-06, + "loss": 1.1103, + "step": 49247 + }, + { + "epoch": 3.64, + "learning_rate": 7.079194768494561e-06, + "loss": 0.9284, + "step": 49248 + }, + { + "epoch": 3.64, + "learning_rate": 7.0788132925014285e-06, + "loss": 1.0491, + "step": 49249 + }, + { + "epoch": 3.64, + "learning_rate": 7.07843182115586e-06, + "loss": 0.9841, + "step": 49250 + }, + { + "epoch": 3.64, + "learning_rate": 7.078050354458468e-06, + "loss": 0.9915, + "step": 49251 + }, + { + "epoch": 3.64, + "learning_rate": 7.077668892409853e-06, + "loss": 0.9362, + "step": 49252 + }, + { + "epoch": 3.64, + "learning_rate": 7.077287435010629e-06, + "loss": 0.9992, + "step": 49253 + }, + { + "epoch": 3.64, + "learning_rate": 7.076905982261395e-06, + "loss": 1.1109, + "step": 49254 + }, + { + "epoch": 3.64, + "learning_rate": 7.076524534162764e-06, + "loss": 0.974, + "step": 49255 + }, + { + "epoch": 3.64, + "learning_rate": 7.07614309071534e-06, + "loss": 1.0255, + "step": 49256 + }, + { + "epoch": 3.64, + "learning_rate": 7.07576165191973e-06, + "loss": 0.987, + "step": 49257 + }, + { + "epoch": 3.64, + "learning_rate": 7.075380217776545e-06, + "loss": 0.9059, + "step": 49258 + }, + { + "epoch": 3.64, + "learning_rate": 7.074998788286382e-06, + "loss": 0.9111, + "step": 49259 + }, + { + "epoch": 3.64, + "learning_rate": 7.0746173634498604e-06, + "loss": 1.0282, + "step": 49260 + }, + { + "epoch": 3.64, + "learning_rate": 7.074235943267576e-06, + "loss": 1.0619, + "step": 49261 + }, + { + "epoch": 3.64, + "learning_rate": 7.0738545277401425e-06, + "loss": 1.0538, + "step": 49262 + }, + { + "epoch": 3.64, + "learning_rate": 7.0734731168681635e-06, + "loss": 1.1188, + "step": 49263 + }, + { + "epoch": 3.64, + "learning_rate": 7.073091710652247e-06, + "loss": 0.988, + "step": 49264 + }, + { + "epoch": 3.64, + "learning_rate": 7.072710309092999e-06, + "loss": 0.9466, + "step": 49265 + }, + { + "epoch": 3.64, + "learning_rate": 7.072328912191029e-06, + "loss": 0.9771, + "step": 49266 + }, + { + "epoch": 3.64, + "learning_rate": 7.071947519946942e-06, + "loss": 0.9633, + "step": 49267 + }, + { + "epoch": 3.64, + "learning_rate": 7.071566132361342e-06, + "loss": 0.9221, + "step": 49268 + }, + { + "epoch": 3.64, + "learning_rate": 7.071184749434841e-06, + "loss": 1.0164, + "step": 49269 + }, + { + "epoch": 3.64, + "learning_rate": 7.07080337116804e-06, + "loss": 0.9786, + "step": 49270 + }, + { + "epoch": 3.64, + "learning_rate": 7.0704219975615525e-06, + "loss": 1.0402, + "step": 49271 + }, + { + "epoch": 3.64, + "learning_rate": 7.07004062861598e-06, + "loss": 0.9882, + "step": 49272 + }, + { + "epoch": 3.64, + "learning_rate": 7.069659264331932e-06, + "loss": 1.0605, + "step": 49273 + }, + { + "epoch": 3.64, + "learning_rate": 7.0692779047100115e-06, + "loss": 1.1139, + "step": 49274 + }, + { + "epoch": 3.64, + "learning_rate": 7.06889654975083e-06, + "loss": 1.0402, + "step": 49275 + }, + { + "epoch": 3.64, + "learning_rate": 7.0685151994549905e-06, + "loss": 1.0857, + "step": 49276 + }, + { + "epoch": 3.64, + "learning_rate": 7.0681338538231034e-06, + "loss": 0.9711, + "step": 49277 + }, + { + "epoch": 3.64, + "learning_rate": 7.067752512855777e-06, + "loss": 1.023, + "step": 49278 + }, + { + "epoch": 3.64, + "learning_rate": 7.067371176553605e-06, + "loss": 0.9834, + "step": 49279 + }, + { + "epoch": 3.64, + "learning_rate": 7.0669898449172124e-06, + "loss": 0.9398, + "step": 49280 + }, + { + "epoch": 3.64, + "learning_rate": 7.06660851794719e-06, + "loss": 1.0158, + "step": 49281 + }, + { + "epoch": 3.64, + "learning_rate": 7.066227195644157e-06, + "loss": 1.1008, + "step": 49282 + }, + { + "epoch": 3.64, + "learning_rate": 7.06584587800871e-06, + "loss": 1.0393, + "step": 49283 + }, + { + "epoch": 3.64, + "learning_rate": 7.0654645650414644e-06, + "loss": 1.0524, + "step": 49284 + }, + { + "epoch": 3.64, + "learning_rate": 7.065083256743017e-06, + "loss": 0.987, + "step": 49285 + }, + { + "epoch": 3.64, + "learning_rate": 7.064701953113988e-06, + "loss": 1.0753, + "step": 49286 + }, + { + "epoch": 3.64, + "learning_rate": 7.064320654154972e-06, + "loss": 0.9126, + "step": 49287 + }, + { + "epoch": 3.64, + "learning_rate": 7.0639393598665795e-06, + "loss": 1.0113, + "step": 49288 + }, + { + "epoch": 3.64, + "learning_rate": 7.063558070249419e-06, + "loss": 1.0249, + "step": 49289 + }, + { + "epoch": 3.64, + "learning_rate": 7.063176785304094e-06, + "loss": 0.9889, + "step": 49290 + }, + { + "epoch": 3.64, + "learning_rate": 7.062795505031213e-06, + "loss": 0.8822, + "step": 49291 + }, + { + "epoch": 3.64, + "learning_rate": 7.062414229431381e-06, + "loss": 1.0782, + "step": 49292 + }, + { + "epoch": 3.64, + "learning_rate": 7.062032958505209e-06, + "loss": 0.9908, + "step": 49293 + }, + { + "epoch": 3.64, + "learning_rate": 7.061651692253298e-06, + "loss": 1.0797, + "step": 49294 + }, + { + "epoch": 3.64, + "learning_rate": 7.061270430676259e-06, + "loss": 0.9715, + "step": 49295 + }, + { + "epoch": 3.64, + "learning_rate": 7.060889173774696e-06, + "loss": 0.9265, + "step": 49296 + }, + { + "epoch": 3.64, + "learning_rate": 7.060507921549216e-06, + "loss": 0.9423, + "step": 49297 + }, + { + "epoch": 3.64, + "learning_rate": 7.060126674000426e-06, + "loss": 1.0528, + "step": 49298 + }, + { + "epoch": 3.64, + "learning_rate": 7.059745431128931e-06, + "loss": 1.0629, + "step": 49299 + }, + { + "epoch": 3.64, + "learning_rate": 7.059364192935341e-06, + "loss": 1.0354, + "step": 49300 + }, + { + "epoch": 3.64, + "learning_rate": 7.058982959420258e-06, + "loss": 1.0157, + "step": 49301 + }, + { + "epoch": 3.64, + "learning_rate": 7.058601730584295e-06, + "loss": 1.0281, + "step": 49302 + }, + { + "epoch": 3.64, + "learning_rate": 7.058220506428051e-06, + "loss": 1.039, + "step": 49303 + }, + { + "epoch": 3.64, + "learning_rate": 7.057839286952137e-06, + "loss": 0.968, + "step": 49304 + }, + { + "epoch": 3.64, + "learning_rate": 7.057458072157163e-06, + "loss": 1.0021, + "step": 49305 + }, + { + "epoch": 3.64, + "learning_rate": 7.057076862043725e-06, + "loss": 0.9356, + "step": 49306 + }, + { + "epoch": 3.64, + "learning_rate": 7.0566956566124404e-06, + "loss": 0.886, + "step": 49307 + }, + { + "epoch": 3.64, + "learning_rate": 7.056314455863906e-06, + "loss": 0.9229, + "step": 49308 + }, + { + "epoch": 3.64, + "learning_rate": 7.055933259798736e-06, + "loss": 1.0292, + "step": 49309 + }, + { + "epoch": 3.64, + "learning_rate": 7.055552068417533e-06, + "loss": 0.9034, + "step": 49310 + }, + { + "epoch": 3.64, + "learning_rate": 7.055170881720905e-06, + "loss": 1.0887, + "step": 49311 + }, + { + "epoch": 3.64, + "learning_rate": 7.054789699709457e-06, + "loss": 0.9687, + "step": 49312 + }, + { + "epoch": 3.64, + "learning_rate": 7.054408522383799e-06, + "loss": 0.9659, + "step": 49313 + }, + { + "epoch": 3.64, + "learning_rate": 7.054027349744534e-06, + "loss": 1.0174, + "step": 49314 + }, + { + "epoch": 3.64, + "learning_rate": 7.053646181792267e-06, + "loss": 1.0222, + "step": 49315 + }, + { + "epoch": 3.64, + "learning_rate": 7.05326501852761e-06, + "loss": 0.9157, + "step": 49316 + }, + { + "epoch": 3.64, + "learning_rate": 7.052883859951163e-06, + "loss": 1.0633, + "step": 49317 + }, + { + "epoch": 3.64, + "learning_rate": 7.052502706063539e-06, + "loss": 0.9435, + "step": 49318 + }, + { + "epoch": 3.64, + "learning_rate": 7.052121556865337e-06, + "loss": 1.0355, + "step": 49319 + }, + { + "epoch": 3.64, + "learning_rate": 7.051740412357171e-06, + "loss": 1.0157, + "step": 49320 + }, + { + "epoch": 3.64, + "learning_rate": 7.051359272539642e-06, + "loss": 1.037, + "step": 49321 + }, + { + "epoch": 3.64, + "learning_rate": 7.050978137413359e-06, + "loss": 0.9116, + "step": 49322 + }, + { + "epoch": 3.64, + "learning_rate": 7.050597006978926e-06, + "loss": 0.9534, + "step": 49323 + }, + { + "epoch": 3.64, + "learning_rate": 7.050215881236953e-06, + "loss": 1.0913, + "step": 49324 + }, + { + "epoch": 3.64, + "learning_rate": 7.049834760188047e-06, + "loss": 1.0145, + "step": 49325 + }, + { + "epoch": 3.64, + "learning_rate": 7.049453643832804e-06, + "loss": 1.0093, + "step": 49326 + }, + { + "epoch": 3.64, + "learning_rate": 7.049072532171846e-06, + "loss": 1.0121, + "step": 49327 + }, + { + "epoch": 3.64, + "learning_rate": 7.048691425205765e-06, + "loss": 0.9662, + "step": 49328 + }, + { + "epoch": 3.64, + "learning_rate": 7.048310322935179e-06, + "loss": 0.9388, + "step": 49329 + }, + { + "epoch": 3.64, + "learning_rate": 7.047929225360683e-06, + "loss": 1.0232, + "step": 49330 + }, + { + "epoch": 3.65, + "learning_rate": 7.047548132482896e-06, + "loss": 1.0084, + "step": 49331 + }, + { + "epoch": 3.65, + "learning_rate": 7.047167044302412e-06, + "loss": 1.0679, + "step": 49332 + }, + { + "epoch": 3.65, + "learning_rate": 7.046785960819848e-06, + "loss": 1.0358, + "step": 49333 + }, + { + "epoch": 3.65, + "learning_rate": 7.046404882035805e-06, + "loss": 0.8802, + "step": 49334 + }, + { + "epoch": 3.65, + "learning_rate": 7.046023807950885e-06, + "loss": 1.0452, + "step": 49335 + }, + { + "epoch": 3.65, + "learning_rate": 7.0456427385657026e-06, + "loss": 0.9282, + "step": 49336 + }, + { + "epoch": 3.65, + "learning_rate": 7.045261673880857e-06, + "loss": 1.0154, + "step": 49337 + }, + { + "epoch": 3.65, + "learning_rate": 7.04488061389696e-06, + "loss": 0.9363, + "step": 49338 + }, + { + "epoch": 3.65, + "learning_rate": 7.044499558614615e-06, + "loss": 1.032, + "step": 49339 + }, + { + "epoch": 3.65, + "learning_rate": 7.04411850803443e-06, + "loss": 1.0623, + "step": 49340 + }, + { + "epoch": 3.65, + "learning_rate": 7.043737462157008e-06, + "loss": 1.0094, + "step": 49341 + }, + { + "epoch": 3.65, + "learning_rate": 7.043356420982961e-06, + "loss": 0.9382, + "step": 49342 + }, + { + "epoch": 3.65, + "learning_rate": 7.04297538451289e-06, + "loss": 1.0404, + "step": 49343 + }, + { + "epoch": 3.65, + "learning_rate": 7.042594352747401e-06, + "loss": 1.0582, + "step": 49344 + }, + { + "epoch": 3.65, + "learning_rate": 7.042213325687104e-06, + "loss": 0.9293, + "step": 49345 + }, + { + "epoch": 3.65, + "learning_rate": 7.041832303332602e-06, + "loss": 1.0775, + "step": 49346 + }, + { + "epoch": 3.65, + "learning_rate": 7.041451285684504e-06, + "loss": 1.0197, + "step": 49347 + }, + { + "epoch": 3.65, + "learning_rate": 7.041070272743414e-06, + "loss": 0.9981, + "step": 49348 + }, + { + "epoch": 3.65, + "learning_rate": 7.040689264509941e-06, + "loss": 1.0489, + "step": 49349 + }, + { + "epoch": 3.65, + "learning_rate": 7.040308260984683e-06, + "loss": 0.984, + "step": 49350 + }, + { + "epoch": 3.65, + "learning_rate": 7.039927262168259e-06, + "loss": 0.9104, + "step": 49351 + }, + { + "epoch": 3.65, + "learning_rate": 7.039546268061266e-06, + "loss": 0.9856, + "step": 49352 + }, + { + "epoch": 3.65, + "learning_rate": 7.039165278664311e-06, + "loss": 0.9687, + "step": 49353 + }, + { + "epoch": 3.65, + "learning_rate": 7.038784293978004e-06, + "loss": 1.0207, + "step": 49354 + }, + { + "epoch": 3.65, + "learning_rate": 7.038403314002946e-06, + "loss": 1.0678, + "step": 49355 + }, + { + "epoch": 3.65, + "learning_rate": 7.0380223387397475e-06, + "loss": 1.0635, + "step": 49356 + }, + { + "epoch": 3.65, + "learning_rate": 7.037641368189012e-06, + "loss": 1.0187, + "step": 49357 + }, + { + "epoch": 3.65, + "learning_rate": 7.037260402351349e-06, + "loss": 1.0256, + "step": 49358 + }, + { + "epoch": 3.65, + "learning_rate": 7.03687944122736e-06, + "loss": 0.8951, + "step": 49359 + }, + { + "epoch": 3.65, + "learning_rate": 7.036498484817656e-06, + "loss": 1.0286, + "step": 49360 + }, + { + "epoch": 3.65, + "learning_rate": 7.0361175331228395e-06, + "loss": 0.9835, + "step": 49361 + }, + { + "epoch": 3.65, + "learning_rate": 7.035736586143516e-06, + "loss": 0.9915, + "step": 49362 + }, + { + "epoch": 3.65, + "learning_rate": 7.035355643880294e-06, + "loss": 1.0408, + "step": 49363 + }, + { + "epoch": 3.65, + "learning_rate": 7.034974706333777e-06, + "loss": 1.0024, + "step": 49364 + }, + { + "epoch": 3.65, + "learning_rate": 7.034593773504576e-06, + "loss": 1.0852, + "step": 49365 + }, + { + "epoch": 3.65, + "learning_rate": 7.034212845393291e-06, + "loss": 0.9494, + "step": 49366 + }, + { + "epoch": 3.65, + "learning_rate": 7.0338319220005335e-06, + "loss": 0.8631, + "step": 49367 + }, + { + "epoch": 3.65, + "learning_rate": 7.0334510033269055e-06, + "loss": 0.9518, + "step": 49368 + }, + { + "epoch": 3.65, + "learning_rate": 7.033070089373015e-06, + "loss": 1.0756, + "step": 49369 + }, + { + "epoch": 3.65, + "learning_rate": 7.032689180139471e-06, + "loss": 1.1202, + "step": 49370 + }, + { + "epoch": 3.65, + "learning_rate": 7.032308275626869e-06, + "loss": 1.0595, + "step": 49371 + }, + { + "epoch": 3.65, + "learning_rate": 7.0319273758358295e-06, + "loss": 0.9654, + "step": 49372 + }, + { + "epoch": 3.65, + "learning_rate": 7.031546480766944e-06, + "loss": 1.0823, + "step": 49373 + }, + { + "epoch": 3.65, + "learning_rate": 7.031165590420832e-06, + "loss": 1.1693, + "step": 49374 + }, + { + "epoch": 3.65, + "learning_rate": 7.0307847047980855e-06, + "loss": 1.0423, + "step": 49375 + }, + { + "epoch": 3.65, + "learning_rate": 7.030403823899326e-06, + "loss": 0.9232, + "step": 49376 + }, + { + "epoch": 3.65, + "learning_rate": 7.030022947725145e-06, + "loss": 0.9377, + "step": 49377 + }, + { + "epoch": 3.65, + "learning_rate": 7.029642076276161e-06, + "loss": 0.9461, + "step": 49378 + }, + { + "epoch": 3.65, + "learning_rate": 7.029261209552968e-06, + "loss": 1.042, + "step": 49379 + }, + { + "epoch": 3.65, + "learning_rate": 7.0288803475561836e-06, + "loss": 0.9482, + "step": 49380 + }, + { + "epoch": 3.65, + "learning_rate": 7.028499490286407e-06, + "loss": 1.0648, + "step": 49381 + }, + { + "epoch": 3.65, + "learning_rate": 7.0281186377442415e-06, + "loss": 0.9741, + "step": 49382 + }, + { + "epoch": 3.65, + "learning_rate": 7.0277377899303e-06, + "loss": 0.9928, + "step": 49383 + }, + { + "epoch": 3.65, + "learning_rate": 7.027356946845183e-06, + "loss": 0.9005, + "step": 49384 + }, + { + "epoch": 3.65, + "learning_rate": 7.0269761084894995e-06, + "loss": 0.9391, + "step": 49385 + }, + { + "epoch": 3.65, + "learning_rate": 7.0265952748638535e-06, + "loss": 0.8896, + "step": 49386 + }, + { + "epoch": 3.65, + "learning_rate": 7.026214445968853e-06, + "loss": 1.1033, + "step": 49387 + }, + { + "epoch": 3.65, + "learning_rate": 7.025833621805103e-06, + "loss": 0.9514, + "step": 49388 + }, + { + "epoch": 3.65, + "learning_rate": 7.025452802373208e-06, + "loss": 1.0397, + "step": 49389 + }, + { + "epoch": 3.65, + "learning_rate": 7.0250719876737774e-06, + "loss": 0.9569, + "step": 49390 + }, + { + "epoch": 3.65, + "learning_rate": 7.024691177707413e-06, + "loss": 0.9898, + "step": 49391 + }, + { + "epoch": 3.65, + "learning_rate": 7.024310372474722e-06, + "loss": 1.0312, + "step": 49392 + }, + { + "epoch": 3.65, + "learning_rate": 7.02392957197631e-06, + "loss": 1.0722, + "step": 49393 + }, + { + "epoch": 3.65, + "learning_rate": 7.023548776212787e-06, + "loss": 0.9701, + "step": 49394 + }, + { + "epoch": 3.65, + "learning_rate": 7.02316798518475e-06, + "loss": 0.9716, + "step": 49395 + }, + { + "epoch": 3.65, + "learning_rate": 7.022787198892815e-06, + "loss": 1.0964, + "step": 49396 + }, + { + "epoch": 3.65, + "learning_rate": 7.022406417337578e-06, + "loss": 0.977, + "step": 49397 + }, + { + "epoch": 3.65, + "learning_rate": 7.022025640519656e-06, + "loss": 1.0327, + "step": 49398 + }, + { + "epoch": 3.65, + "learning_rate": 7.021644868439647e-06, + "loss": 1.0715, + "step": 49399 + }, + { + "epoch": 3.65, + "learning_rate": 7.021264101098155e-06, + "loss": 0.9302, + "step": 49400 + }, + { + "epoch": 3.65, + "learning_rate": 7.0208833384957916e-06, + "loss": 1.0522, + "step": 49401 + }, + { + "epoch": 3.65, + "learning_rate": 7.020502580633158e-06, + "loss": 0.9875, + "step": 49402 + }, + { + "epoch": 3.65, + "learning_rate": 7.0201218275108655e-06, + "loss": 1.0192, + "step": 49403 + }, + { + "epoch": 3.65, + "learning_rate": 7.019741079129514e-06, + "loss": 0.9738, + "step": 49404 + }, + { + "epoch": 3.65, + "learning_rate": 7.019360335489714e-06, + "loss": 1.0818, + "step": 49405 + }, + { + "epoch": 3.65, + "learning_rate": 7.018979596592065e-06, + "loss": 0.8817, + "step": 49406 + }, + { + "epoch": 3.65, + "learning_rate": 7.01859886243718e-06, + "loss": 0.9227, + "step": 49407 + }, + { + "epoch": 3.65, + "learning_rate": 7.018218133025663e-06, + "loss": 1.0735, + "step": 49408 + }, + { + "epoch": 3.65, + "learning_rate": 7.017837408358115e-06, + "loss": 1.0058, + "step": 49409 + }, + { + "epoch": 3.65, + "learning_rate": 7.017456688435147e-06, + "loss": 0.9896, + "step": 49410 + }, + { + "epoch": 3.65, + "learning_rate": 7.01707597325736e-06, + "loss": 1.0233, + "step": 49411 + }, + { + "epoch": 3.65, + "learning_rate": 7.016695262825366e-06, + "loss": 1.1117, + "step": 49412 + }, + { + "epoch": 3.65, + "learning_rate": 7.016314557139764e-06, + "loss": 0.9568, + "step": 49413 + }, + { + "epoch": 3.65, + "learning_rate": 7.015933856201164e-06, + "loss": 0.8925, + "step": 49414 + }, + { + "epoch": 3.65, + "learning_rate": 7.015553160010171e-06, + "loss": 1.0823, + "step": 49415 + }, + { + "epoch": 3.65, + "learning_rate": 7.015172468567391e-06, + "loss": 1.0119, + "step": 49416 + }, + { + "epoch": 3.65, + "learning_rate": 7.0147917818734315e-06, + "loss": 0.9976, + "step": 49417 + }, + { + "epoch": 3.65, + "learning_rate": 7.014411099928888e-06, + "loss": 0.8886, + "step": 49418 + }, + { + "epoch": 3.65, + "learning_rate": 7.0140304227343815e-06, + "loss": 1.0137, + "step": 49419 + }, + { + "epoch": 3.65, + "learning_rate": 7.013649750290503e-06, + "loss": 0.7889, + "step": 49420 + }, + { + "epoch": 3.65, + "learning_rate": 7.013269082597871e-06, + "loss": 0.9568, + "step": 49421 + }, + { + "epoch": 3.65, + "learning_rate": 7.012888419657079e-06, + "loss": 0.934, + "step": 49422 + }, + { + "epoch": 3.65, + "learning_rate": 7.012507761468745e-06, + "loss": 1.1069, + "step": 49423 + }, + { + "epoch": 3.65, + "learning_rate": 7.012127108033463e-06, + "loss": 0.9581, + "step": 49424 + }, + { + "epoch": 3.65, + "learning_rate": 7.011746459351849e-06, + "loss": 1.0221, + "step": 49425 + }, + { + "epoch": 3.65, + "learning_rate": 7.011365815424502e-06, + "loss": 0.9631, + "step": 49426 + }, + { + "epoch": 3.65, + "learning_rate": 7.010985176252028e-06, + "loss": 0.9894, + "step": 49427 + }, + { + "epoch": 3.65, + "learning_rate": 7.010604541835035e-06, + "loss": 0.9521, + "step": 49428 + }, + { + "epoch": 3.65, + "learning_rate": 7.010223912174125e-06, + "loss": 0.9716, + "step": 49429 + }, + { + "epoch": 3.65, + "learning_rate": 7.009843287269908e-06, + "loss": 0.9643, + "step": 49430 + }, + { + "epoch": 3.65, + "learning_rate": 7.009462667122987e-06, + "loss": 1.0072, + "step": 49431 + }, + { + "epoch": 3.65, + "learning_rate": 7.00908205173397e-06, + "loss": 1.0271, + "step": 49432 + }, + { + "epoch": 3.65, + "learning_rate": 7.008701441103458e-06, + "loss": 0.9943, + "step": 49433 + }, + { + "epoch": 3.65, + "learning_rate": 7.0083208352320605e-06, + "loss": 0.9455, + "step": 49434 + }, + { + "epoch": 3.65, + "learning_rate": 7.007940234120381e-06, + "loss": 1.0719, + "step": 49435 + }, + { + "epoch": 3.65, + "learning_rate": 7.007559637769026e-06, + "loss": 1.0988, + "step": 49436 + }, + { + "epoch": 3.65, + "learning_rate": 7.007179046178603e-06, + "loss": 0.9973, + "step": 49437 + }, + { + "epoch": 3.65, + "learning_rate": 7.006798459349712e-06, + "loss": 1.0184, + "step": 49438 + }, + { + "epoch": 3.65, + "learning_rate": 7.006417877282965e-06, + "loss": 0.9353, + "step": 49439 + }, + { + "epoch": 3.65, + "learning_rate": 7.00603729997896e-06, + "loss": 0.9158, + "step": 49440 + }, + { + "epoch": 3.65, + "learning_rate": 7.005656727438312e-06, + "loss": 1.0575, + "step": 49441 + }, + { + "epoch": 3.65, + "learning_rate": 7.005276159661616e-06, + "loss": 1.0255, + "step": 49442 + }, + { + "epoch": 3.65, + "learning_rate": 7.004895596649488e-06, + "loss": 1.1365, + "step": 49443 + }, + { + "epoch": 3.65, + "learning_rate": 7.004515038402522e-06, + "loss": 1.0551, + "step": 49444 + }, + { + "epoch": 3.65, + "learning_rate": 7.004134484921337e-06, + "loss": 0.9924, + "step": 49445 + }, + { + "epoch": 3.65, + "learning_rate": 7.003753936206527e-06, + "loss": 0.9873, + "step": 49446 + }, + { + "epoch": 3.65, + "learning_rate": 7.003373392258702e-06, + "loss": 0.8931, + "step": 49447 + }, + { + "epoch": 3.65, + "learning_rate": 7.002992853078468e-06, + "loss": 1.021, + "step": 49448 + }, + { + "epoch": 3.65, + "learning_rate": 7.002612318666428e-06, + "loss": 1.028, + "step": 49449 + }, + { + "epoch": 3.65, + "learning_rate": 7.0022317890231906e-06, + "loss": 1.0154, + "step": 49450 + }, + { + "epoch": 3.65, + "learning_rate": 7.001851264149357e-06, + "loss": 1.0404, + "step": 49451 + }, + { + "epoch": 3.65, + "learning_rate": 7.001470744045538e-06, + "loss": 1.0727, + "step": 49452 + }, + { + "epoch": 3.65, + "learning_rate": 7.001090228712334e-06, + "loss": 0.9594, + "step": 49453 + }, + { + "epoch": 3.65, + "learning_rate": 7.000709718150354e-06, + "loss": 0.9217, + "step": 49454 + }, + { + "epoch": 3.65, + "learning_rate": 7.000329212360202e-06, + "loss": 0.986, + "step": 49455 + }, + { + "epoch": 3.65, + "learning_rate": 6.9999487113424816e-06, + "loss": 1.0724, + "step": 49456 + }, + { + "epoch": 3.65, + "learning_rate": 6.999568215097802e-06, + "loss": 1.0618, + "step": 49457 + }, + { + "epoch": 3.65, + "learning_rate": 6.999187723626763e-06, + "loss": 0.9873, + "step": 49458 + }, + { + "epoch": 3.65, + "learning_rate": 6.998807236929977e-06, + "loss": 0.9735, + "step": 49459 + }, + { + "epoch": 3.65, + "learning_rate": 6.998426755008045e-06, + "loss": 0.977, + "step": 49460 + }, + { + "epoch": 3.65, + "learning_rate": 6.998046277861572e-06, + "loss": 1.1159, + "step": 49461 + }, + { + "epoch": 3.65, + "learning_rate": 6.997665805491165e-06, + "loss": 0.9807, + "step": 49462 + }, + { + "epoch": 3.65, + "learning_rate": 6.997285337897429e-06, + "loss": 1.0733, + "step": 49463 + }, + { + "epoch": 3.65, + "learning_rate": 6.996904875080972e-06, + "loss": 1.0833, + "step": 49464 + }, + { + "epoch": 3.65, + "learning_rate": 6.9965244170423896e-06, + "loss": 1.0763, + "step": 49465 + }, + { + "epoch": 3.66, + "learning_rate": 6.9961439637823e-06, + "loss": 0.9299, + "step": 49466 + }, + { + "epoch": 3.66, + "learning_rate": 6.995763515301297e-06, + "loss": 0.9976, + "step": 49467 + }, + { + "epoch": 3.66, + "learning_rate": 6.995383071599997e-06, + "loss": 1.0731, + "step": 49468 + }, + { + "epoch": 3.66, + "learning_rate": 6.995002632678994e-06, + "loss": 0.9208, + "step": 49469 + }, + { + "epoch": 3.66, + "learning_rate": 6.994622198538904e-06, + "loss": 0.8979, + "step": 49470 + }, + { + "epoch": 3.66, + "learning_rate": 6.994241769180323e-06, + "loss": 1.0351, + "step": 49471 + }, + { + "epoch": 3.66, + "learning_rate": 6.993861344603863e-06, + "loss": 0.9424, + "step": 49472 + }, + { + "epoch": 3.66, + "learning_rate": 6.993480924810127e-06, + "loss": 0.9351, + "step": 49473 + }, + { + "epoch": 3.66, + "learning_rate": 6.993100509799718e-06, + "loss": 0.9857, + "step": 49474 + }, + { + "epoch": 3.66, + "learning_rate": 6.992720099573242e-06, + "loss": 0.9857, + "step": 49475 + }, + { + "epoch": 3.66, + "learning_rate": 6.992339694131307e-06, + "loss": 1.0487, + "step": 49476 + }, + { + "epoch": 3.66, + "learning_rate": 6.991959293474517e-06, + "loss": 1.0751, + "step": 49477 + }, + { + "epoch": 3.66, + "learning_rate": 6.991578897603475e-06, + "loss": 1.0341, + "step": 49478 + }, + { + "epoch": 3.66, + "learning_rate": 6.9911985065187894e-06, + "loss": 0.9017, + "step": 49479 + }, + { + "epoch": 3.66, + "learning_rate": 6.990818120221061e-06, + "loss": 1.0175, + "step": 49480 + }, + { + "epoch": 3.66, + "learning_rate": 6.990437738710902e-06, + "loss": 1.0784, + "step": 49481 + }, + { + "epoch": 3.66, + "learning_rate": 6.990057361988915e-06, + "loss": 0.929, + "step": 49482 + }, + { + "epoch": 3.66, + "learning_rate": 6.9896769900556974e-06, + "loss": 0.8206, + "step": 49483 + }, + { + "epoch": 3.66, + "learning_rate": 6.9892966229118665e-06, + "loss": 1.0497, + "step": 49484 + }, + { + "epoch": 3.66, + "learning_rate": 6.988916260558016e-06, + "loss": 0.9468, + "step": 49485 + }, + { + "epoch": 3.66, + "learning_rate": 6.9885359029947615e-06, + "loss": 0.97, + "step": 49486 + }, + { + "epoch": 3.66, + "learning_rate": 6.988155550222699e-06, + "loss": 0.9681, + "step": 49487 + }, + { + "epoch": 3.66, + "learning_rate": 6.987775202242443e-06, + "loss": 0.9643, + "step": 49488 + }, + { + "epoch": 3.66, + "learning_rate": 6.987394859054587e-06, + "loss": 1.0687, + "step": 49489 + }, + { + "epoch": 3.66, + "learning_rate": 6.987014520659749e-06, + "loss": 1.016, + "step": 49490 + }, + { + "epoch": 3.66, + "learning_rate": 6.986634187058523e-06, + "loss": 0.931, + "step": 49491 + }, + { + "epoch": 3.66, + "learning_rate": 6.9862538582515235e-06, + "loss": 1.017, + "step": 49492 + }, + { + "epoch": 3.66, + "learning_rate": 6.985873534239349e-06, + "loss": 0.9585, + "step": 49493 + }, + { + "epoch": 3.66, + "learning_rate": 6.9854932150226055e-06, + "loss": 0.9986, + "step": 49494 + }, + { + "epoch": 3.66, + "learning_rate": 6.9851129006019e-06, + "loss": 0.9133, + "step": 49495 + }, + { + "epoch": 3.66, + "learning_rate": 6.984732590977836e-06, + "loss": 1.0419, + "step": 49496 + }, + { + "epoch": 3.66, + "learning_rate": 6.98435228615102e-06, + "loss": 1.0378, + "step": 49497 + }, + { + "epoch": 3.66, + "learning_rate": 6.983971986122056e-06, + "loss": 0.9657, + "step": 49498 + }, + { + "epoch": 3.66, + "learning_rate": 6.98359169089155e-06, + "loss": 0.8712, + "step": 49499 + }, + { + "epoch": 3.66, + "learning_rate": 6.983211400460106e-06, + "loss": 1.0669, + "step": 49500 + }, + { + "epoch": 3.66, + "learning_rate": 6.98283111482833e-06, + "loss": 0.9488, + "step": 49501 + }, + { + "epoch": 3.66, + "learning_rate": 6.982450833996827e-06, + "loss": 1.0684, + "step": 49502 + }, + { + "epoch": 3.66, + "learning_rate": 6.982070557966199e-06, + "loss": 1.025, + "step": 49503 + }, + { + "epoch": 3.66, + "learning_rate": 6.981690286737056e-06, + "loss": 1.0127, + "step": 49504 + }, + { + "epoch": 3.66, + "learning_rate": 6.9813100203099996e-06, + "loss": 0.9551, + "step": 49505 + }, + { + "epoch": 3.66, + "learning_rate": 6.980929758685636e-06, + "loss": 0.9364, + "step": 49506 + }, + { + "epoch": 3.66, + "learning_rate": 6.9805495018645684e-06, + "loss": 0.9488, + "step": 49507 + }, + { + "epoch": 3.66, + "learning_rate": 6.980169249847406e-06, + "loss": 0.8987, + "step": 49508 + }, + { + "epoch": 3.66, + "learning_rate": 6.979789002634749e-06, + "loss": 0.9569, + "step": 49509 + }, + { + "epoch": 3.66, + "learning_rate": 6.9794087602272065e-06, + "loss": 0.9695, + "step": 49510 + }, + { + "epoch": 3.66, + "learning_rate": 6.979028522625384e-06, + "loss": 0.9489, + "step": 49511 + }, + { + "epoch": 3.66, + "learning_rate": 6.978648289829877e-06, + "loss": 0.9627, + "step": 49512 + }, + { + "epoch": 3.66, + "learning_rate": 6.9782680618413034e-06, + "loss": 1.0062, + "step": 49513 + }, + { + "epoch": 3.66, + "learning_rate": 6.977887838660255e-06, + "loss": 0.949, + "step": 49514 + }, + { + "epoch": 3.66, + "learning_rate": 6.9775076202873515e-06, + "loss": 1.0557, + "step": 49515 + }, + { + "epoch": 3.66, + "learning_rate": 6.977127406723187e-06, + "loss": 1.0236, + "step": 49516 + }, + { + "epoch": 3.66, + "learning_rate": 6.97674719796837e-06, + "loss": 0.9154, + "step": 49517 + }, + { + "epoch": 3.66, + "learning_rate": 6.976366994023503e-06, + "loss": 1.024, + "step": 49518 + }, + { + "epoch": 3.66, + "learning_rate": 6.975986794889195e-06, + "loss": 0.9826, + "step": 49519 + }, + { + "epoch": 3.66, + "learning_rate": 6.975606600566048e-06, + "loss": 0.9791, + "step": 49520 + }, + { + "epoch": 3.66, + "learning_rate": 6.9752264110546674e-06, + "loss": 0.8944, + "step": 49521 + }, + { + "epoch": 3.66, + "learning_rate": 6.9748462263556584e-06, + "loss": 0.9972, + "step": 49522 + }, + { + "epoch": 3.66, + "learning_rate": 6.974466046469625e-06, + "loss": 1.0165, + "step": 49523 + }, + { + "epoch": 3.66, + "learning_rate": 6.974085871397174e-06, + "loss": 1.0169, + "step": 49524 + }, + { + "epoch": 3.66, + "learning_rate": 6.973705701138908e-06, + "loss": 1.0163, + "step": 49525 + }, + { + "epoch": 3.66, + "learning_rate": 6.973325535695434e-06, + "loss": 0.9196, + "step": 49526 + }, + { + "epoch": 3.66, + "learning_rate": 6.972945375067355e-06, + "loss": 1.0125, + "step": 49527 + }, + { + "epoch": 3.66, + "learning_rate": 6.972565219255277e-06, + "loss": 0.9973, + "step": 49528 + }, + { + "epoch": 3.66, + "learning_rate": 6.972185068259808e-06, + "loss": 1.006, + "step": 49529 + }, + { + "epoch": 3.66, + "learning_rate": 6.971804922081542e-06, + "loss": 1.0497, + "step": 49530 + }, + { + "epoch": 3.66, + "learning_rate": 6.971424780721099e-06, + "loss": 1.0054, + "step": 49531 + }, + { + "epoch": 3.66, + "learning_rate": 6.971044644179068e-06, + "loss": 0.9767, + "step": 49532 + }, + { + "epoch": 3.66, + "learning_rate": 6.970664512456067e-06, + "loss": 1.1389, + "step": 49533 + }, + { + "epoch": 3.66, + "learning_rate": 6.97028438555269e-06, + "loss": 0.9376, + "step": 49534 + }, + { + "epoch": 3.66, + "learning_rate": 6.969904263469554e-06, + "loss": 1.0649, + "step": 49535 + }, + { + "epoch": 3.66, + "learning_rate": 6.969524146207251e-06, + "loss": 0.9549, + "step": 49536 + }, + { + "epoch": 3.66, + "learning_rate": 6.969144033766397e-06, + "loss": 1.0056, + "step": 49537 + }, + { + "epoch": 3.66, + "learning_rate": 6.968763926147586e-06, + "loss": 0.9961, + "step": 49538 + }, + { + "epoch": 3.66, + "learning_rate": 6.968383823351434e-06, + "loss": 1.0133, + "step": 49539 + }, + { + "epoch": 3.66, + "learning_rate": 6.968003725378538e-06, + "loss": 1.1167, + "step": 49540 + }, + { + "epoch": 3.66, + "learning_rate": 6.967623632229503e-06, + "loss": 0.9387, + "step": 49541 + }, + { + "epoch": 3.66, + "learning_rate": 6.967243543904938e-06, + "loss": 1.0693, + "step": 49542 + }, + { + "epoch": 3.66, + "learning_rate": 6.966863460405442e-06, + "loss": 0.9781, + "step": 49543 + }, + { + "epoch": 3.66, + "learning_rate": 6.966483381731625e-06, + "loss": 1.0277, + "step": 49544 + }, + { + "epoch": 3.66, + "learning_rate": 6.966103307884087e-06, + "loss": 1.0042, + "step": 49545 + }, + { + "epoch": 3.66, + "learning_rate": 6.965723238863437e-06, + "loss": 0.8994, + "step": 49546 + }, + { + "epoch": 3.66, + "learning_rate": 6.9653431746702785e-06, + "loss": 1.0741, + "step": 49547 + }, + { + "epoch": 3.66, + "learning_rate": 6.964963115305214e-06, + "loss": 1.1445, + "step": 49548 + }, + { + "epoch": 3.66, + "learning_rate": 6.964583060768853e-06, + "loss": 0.9084, + "step": 49549 + }, + { + "epoch": 3.66, + "learning_rate": 6.964203011061793e-06, + "loss": 0.9162, + "step": 49550 + }, + { + "epoch": 3.66, + "learning_rate": 6.963822966184645e-06, + "loss": 1.1016, + "step": 49551 + }, + { + "epoch": 3.66, + "learning_rate": 6.963442926138009e-06, + "loss": 1.0435, + "step": 49552 + }, + { + "epoch": 3.66, + "learning_rate": 6.9630628909224936e-06, + "loss": 1.0225, + "step": 49553 + }, + { + "epoch": 3.66, + "learning_rate": 6.962682860538701e-06, + "loss": 0.9118, + "step": 49554 + }, + { + "epoch": 3.66, + "learning_rate": 6.962302834987238e-06, + "loss": 1.0305, + "step": 49555 + }, + { + "epoch": 3.66, + "learning_rate": 6.961922814268705e-06, + "loss": 1.0075, + "step": 49556 + }, + { + "epoch": 3.66, + "learning_rate": 6.9615427983837125e-06, + "loss": 1.0109, + "step": 49557 + }, + { + "epoch": 3.66, + "learning_rate": 6.961162787332864e-06, + "loss": 1.0275, + "step": 49558 + }, + { + "epoch": 3.66, + "learning_rate": 6.960782781116755e-06, + "loss": 1.0218, + "step": 49559 + }, + { + "epoch": 3.66, + "learning_rate": 6.960402779736003e-06, + "loss": 0.9881, + "step": 49560 + }, + { + "epoch": 3.66, + "learning_rate": 6.960022783191203e-06, + "loss": 0.9884, + "step": 49561 + }, + { + "epoch": 3.66, + "learning_rate": 6.959642791482966e-06, + "loss": 1.0782, + "step": 49562 + }, + { + "epoch": 3.66, + "learning_rate": 6.9592628046118904e-06, + "loss": 1.0341, + "step": 49563 + }, + { + "epoch": 3.66, + "learning_rate": 6.958882822578588e-06, + "loss": 0.9437, + "step": 49564 + }, + { + "epoch": 3.66, + "learning_rate": 6.9585028453836555e-06, + "loss": 1.0884, + "step": 49565 + }, + { + "epoch": 3.66, + "learning_rate": 6.958122873027704e-06, + "loss": 0.847, + "step": 49566 + }, + { + "epoch": 3.66, + "learning_rate": 6.957742905511336e-06, + "loss": 1.009, + "step": 49567 + }, + { + "epoch": 3.66, + "learning_rate": 6.957362942835154e-06, + "loss": 0.9363, + "step": 49568 + }, + { + "epoch": 3.66, + "learning_rate": 6.956982984999765e-06, + "loss": 1.0088, + "step": 49569 + }, + { + "epoch": 3.66, + "learning_rate": 6.956603032005771e-06, + "loss": 1.0478, + "step": 49570 + }, + { + "epoch": 3.66, + "learning_rate": 6.95622308385378e-06, + "loss": 1.0505, + "step": 49571 + }, + { + "epoch": 3.66, + "learning_rate": 6.955843140544393e-06, + "loss": 0.9814, + "step": 49572 + }, + { + "epoch": 3.66, + "learning_rate": 6.9554632020782174e-06, + "loss": 0.947, + "step": 49573 + }, + { + "epoch": 3.66, + "learning_rate": 6.955083268455854e-06, + "loss": 0.9437, + "step": 49574 + }, + { + "epoch": 3.66, + "learning_rate": 6.954703339677913e-06, + "loss": 0.8601, + "step": 49575 + }, + { + "epoch": 3.66, + "learning_rate": 6.954323415744997e-06, + "loss": 0.972, + "step": 49576 + }, + { + "epoch": 3.66, + "learning_rate": 6.9539434966577014e-06, + "loss": 0.9424, + "step": 49577 + }, + { + "epoch": 3.66, + "learning_rate": 6.9535635824166445e-06, + "loss": 1.0239, + "step": 49578 + }, + { + "epoch": 3.66, + "learning_rate": 6.9531836730224186e-06, + "loss": 1.0522, + "step": 49579 + }, + { + "epoch": 3.66, + "learning_rate": 6.95280376847564e-06, + "loss": 0.9293, + "step": 49580 + }, + { + "epoch": 3.66, + "learning_rate": 6.952423868776901e-06, + "loss": 0.8481, + "step": 49581 + }, + { + "epoch": 3.66, + "learning_rate": 6.952043973926818e-06, + "loss": 1.0887, + "step": 49582 + }, + { + "epoch": 3.66, + "learning_rate": 6.951664083925984e-06, + "loss": 1.031, + "step": 49583 + }, + { + "epoch": 3.66, + "learning_rate": 6.951284198775014e-06, + "loss": 1.0252, + "step": 49584 + }, + { + "epoch": 3.66, + "learning_rate": 6.9509043184745074e-06, + "loss": 0.984, + "step": 49585 + }, + { + "epoch": 3.66, + "learning_rate": 6.950524443025064e-06, + "loss": 1.117, + "step": 49586 + }, + { + "epoch": 3.66, + "learning_rate": 6.950144572427295e-06, + "loss": 0.9789, + "step": 49587 + }, + { + "epoch": 3.66, + "learning_rate": 6.949764706681801e-06, + "loss": 0.9244, + "step": 49588 + }, + { + "epoch": 3.66, + "learning_rate": 6.949384845789189e-06, + "loss": 0.958, + "step": 49589 + }, + { + "epoch": 3.66, + "learning_rate": 6.94900498975006e-06, + "loss": 0.9468, + "step": 49590 + }, + { + "epoch": 3.66, + "learning_rate": 6.948625138565023e-06, + "loss": 0.9347, + "step": 49591 + }, + { + "epoch": 3.66, + "learning_rate": 6.948245292234677e-06, + "loss": 0.9933, + "step": 49592 + }, + { + "epoch": 3.66, + "learning_rate": 6.9478654507596325e-06, + "loss": 0.9608, + "step": 49593 + }, + { + "epoch": 3.66, + "learning_rate": 6.947485614140487e-06, + "loss": 0.958, + "step": 49594 + }, + { + "epoch": 3.66, + "learning_rate": 6.9471057823778495e-06, + "loss": 0.9563, + "step": 49595 + }, + { + "epoch": 3.66, + "learning_rate": 6.946725955472325e-06, + "loss": 0.9411, + "step": 49596 + }, + { + "epoch": 3.66, + "learning_rate": 6.946346133424513e-06, + "loss": 0.9921, + "step": 49597 + }, + { + "epoch": 3.66, + "learning_rate": 6.945966316235022e-06, + "loss": 1.0817, + "step": 49598 + }, + { + "epoch": 3.66, + "learning_rate": 6.945586503904453e-06, + "loss": 0.9877, + "step": 49599 + }, + { + "epoch": 3.66, + "learning_rate": 6.9452066964334155e-06, + "loss": 1.0229, + "step": 49600 + }, + { + "epoch": 3.66, + "learning_rate": 6.944826893822508e-06, + "loss": 0.9949, + "step": 49601 + }, + { + "epoch": 3.67, + "learning_rate": 6.944447096072339e-06, + "loss": 0.9488, + "step": 49602 + }, + { + "epoch": 3.67, + "learning_rate": 6.944067303183508e-06, + "loss": 0.9386, + "step": 49603 + }, + { + "epoch": 3.67, + "learning_rate": 6.943687515156627e-06, + "loss": 0.8864, + "step": 49604 + }, + { + "epoch": 3.67, + "learning_rate": 6.943307731992295e-06, + "loss": 1.0166, + "step": 49605 + }, + { + "epoch": 3.67, + "learning_rate": 6.942927953691112e-06, + "loss": 0.9665, + "step": 49606 + }, + { + "epoch": 3.67, + "learning_rate": 6.94254818025369e-06, + "loss": 1.0234, + "step": 49607 + }, + { + "epoch": 3.67, + "learning_rate": 6.9421684116806275e-06, + "loss": 0.9613, + "step": 49608 + }, + { + "epoch": 3.67, + "learning_rate": 6.941788647972535e-06, + "loss": 1.0367, + "step": 49609 + }, + { + "epoch": 3.67, + "learning_rate": 6.94140888913001e-06, + "loss": 1.0608, + "step": 49610 + }, + { + "epoch": 3.67, + "learning_rate": 6.941029135153661e-06, + "loss": 0.9957, + "step": 49611 + }, + { + "epoch": 3.67, + "learning_rate": 6.940649386044089e-06, + "loss": 1.029, + "step": 49612 + }, + { + "epoch": 3.67, + "learning_rate": 6.940269641801902e-06, + "loss": 0.9014, + "step": 49613 + }, + { + "epoch": 3.67, + "learning_rate": 6.939889902427704e-06, + "loss": 0.9569, + "step": 49614 + }, + { + "epoch": 3.67, + "learning_rate": 6.939510167922095e-06, + "loss": 0.9844, + "step": 49615 + }, + { + "epoch": 3.67, + "learning_rate": 6.939130438285682e-06, + "loss": 0.9355, + "step": 49616 + }, + { + "epoch": 3.67, + "learning_rate": 6.938750713519067e-06, + "loss": 0.9523, + "step": 49617 + }, + { + "epoch": 3.67, + "learning_rate": 6.9383709936228585e-06, + "loss": 1.0094, + "step": 49618 + }, + { + "epoch": 3.67, + "learning_rate": 6.937991278597656e-06, + "loss": 0.9989, + "step": 49619 + }, + { + "epoch": 3.67, + "learning_rate": 6.937611568444067e-06, + "loss": 0.9644, + "step": 49620 + }, + { + "epoch": 3.67, + "learning_rate": 6.9372318631626925e-06, + "loss": 0.9379, + "step": 49621 + }, + { + "epoch": 3.67, + "learning_rate": 6.93685216275414e-06, + "loss": 0.9853, + "step": 49622 + }, + { + "epoch": 3.67, + "learning_rate": 6.936472467219013e-06, + "loss": 0.9545, + "step": 49623 + }, + { + "epoch": 3.67, + "learning_rate": 6.93609277655791e-06, + "loss": 1.0105, + "step": 49624 + }, + { + "epoch": 3.67, + "learning_rate": 6.935713090771446e-06, + "loss": 0.9527, + "step": 49625 + }, + { + "epoch": 3.67, + "learning_rate": 6.93533340986021e-06, + "loss": 1.0042, + "step": 49626 + }, + { + "epoch": 3.67, + "learning_rate": 6.934953733824823e-06, + "loss": 0.9243, + "step": 49627 + }, + { + "epoch": 3.67, + "learning_rate": 6.934574062665873e-06, + "loss": 1.0203, + "step": 49628 + }, + { + "epoch": 3.67, + "learning_rate": 6.9341943963839796e-06, + "loss": 1.0644, + "step": 49629 + }, + { + "epoch": 3.67, + "learning_rate": 6.933814734979732e-06, + "loss": 1.0087, + "step": 49630 + }, + { + "epoch": 3.67, + "learning_rate": 6.933435078453747e-06, + "loss": 0.998, + "step": 49631 + }, + { + "epoch": 3.67, + "learning_rate": 6.933055426806623e-06, + "loss": 0.9871, + "step": 49632 + }, + { + "epoch": 3.67, + "learning_rate": 6.93267578003896e-06, + "loss": 0.9269, + "step": 49633 + }, + { + "epoch": 3.67, + "learning_rate": 6.932296138151369e-06, + "loss": 0.9491, + "step": 49634 + }, + { + "epoch": 3.67, + "learning_rate": 6.931916501144448e-06, + "loss": 0.9923, + "step": 49635 + }, + { + "epoch": 3.67, + "learning_rate": 6.931536869018807e-06, + "loss": 1.0851, + "step": 49636 + }, + { + "epoch": 3.67, + "learning_rate": 6.931157241775044e-06, + "loss": 1.0915, + "step": 49637 + }, + { + "epoch": 3.67, + "learning_rate": 6.930777619413768e-06, + "loss": 0.9784, + "step": 49638 + }, + { + "epoch": 3.67, + "learning_rate": 6.930398001935579e-06, + "loss": 1.0376, + "step": 49639 + }, + { + "epoch": 3.67, + "learning_rate": 6.930018389341084e-06, + "loss": 0.9766, + "step": 49640 + }, + { + "epoch": 3.67, + "learning_rate": 6.929638781630888e-06, + "loss": 1.0328, + "step": 49641 + }, + { + "epoch": 3.67, + "learning_rate": 6.92925917880559e-06, + "loss": 1.0998, + "step": 49642 + }, + { + "epoch": 3.67, + "learning_rate": 6.928879580865797e-06, + "loss": 0.9511, + "step": 49643 + }, + { + "epoch": 3.67, + "learning_rate": 6.928499987812111e-06, + "loss": 1.0277, + "step": 49644 + }, + { + "epoch": 3.67, + "learning_rate": 6.9281203996451416e-06, + "loss": 0.9863, + "step": 49645 + }, + { + "epoch": 3.67, + "learning_rate": 6.927740816365486e-06, + "loss": 0.9626, + "step": 49646 + }, + { + "epoch": 3.67, + "learning_rate": 6.927361237973752e-06, + "loss": 0.9863, + "step": 49647 + }, + { + "epoch": 3.67, + "learning_rate": 6.92698166447054e-06, + "loss": 0.9231, + "step": 49648 + }, + { + "epoch": 3.67, + "learning_rate": 6.926602095856462e-06, + "loss": 1.0859, + "step": 49649 + }, + { + "epoch": 3.67, + "learning_rate": 6.926222532132108e-06, + "loss": 0.9268, + "step": 49650 + }, + { + "epoch": 3.67, + "learning_rate": 6.925842973298098e-06, + "loss": 0.9978, + "step": 49651 + }, + { + "epoch": 3.67, + "learning_rate": 6.925463419355025e-06, + "loss": 1.17, + "step": 49652 + }, + { + "epoch": 3.67, + "learning_rate": 6.925083870303493e-06, + "loss": 1.0608, + "step": 49653 + }, + { + "epoch": 3.67, + "learning_rate": 6.924704326144111e-06, + "loss": 1.0878, + "step": 49654 + }, + { + "epoch": 3.67, + "learning_rate": 6.9243247868774785e-06, + "loss": 0.9647, + "step": 49655 + }, + { + "epoch": 3.67, + "learning_rate": 6.9239452525042024e-06, + "loss": 0.9433, + "step": 49656 + }, + { + "epoch": 3.67, + "learning_rate": 6.923565723024885e-06, + "loss": 1.0353, + "step": 49657 + }, + { + "epoch": 3.67, + "learning_rate": 6.923186198440133e-06, + "loss": 1.0011, + "step": 49658 + }, + { + "epoch": 3.67, + "learning_rate": 6.922806678750544e-06, + "loss": 0.9672, + "step": 49659 + }, + { + "epoch": 3.67, + "learning_rate": 6.9224271639567275e-06, + "loss": 1.0702, + "step": 49660 + }, + { + "epoch": 3.67, + "learning_rate": 6.922047654059287e-06, + "loss": 0.9716, + "step": 49661 + }, + { + "epoch": 3.67, + "learning_rate": 6.92166814905882e-06, + "loss": 1.0185, + "step": 49662 + }, + { + "epoch": 3.67, + "learning_rate": 6.92128864895594e-06, + "loss": 1.0278, + "step": 49663 + }, + { + "epoch": 3.67, + "learning_rate": 6.920909153751241e-06, + "loss": 0.9513, + "step": 49664 + }, + { + "epoch": 3.67, + "learning_rate": 6.920529663445334e-06, + "loss": 1.057, + "step": 49665 + }, + { + "epoch": 3.67, + "learning_rate": 6.92015017803882e-06, + "loss": 0.902, + "step": 49666 + }, + { + "epoch": 3.67, + "learning_rate": 6.9197706975323044e-06, + "loss": 0.9064, + "step": 49667 + }, + { + "epoch": 3.67, + "learning_rate": 6.919391221926386e-06, + "loss": 0.9778, + "step": 49668 + }, + { + "epoch": 3.67, + "learning_rate": 6.919011751221676e-06, + "loss": 0.9723, + "step": 49669 + }, + { + "epoch": 3.67, + "learning_rate": 6.918632285418775e-06, + "loss": 0.9412, + "step": 49670 + }, + { + "epoch": 3.67, + "learning_rate": 6.91825282451828e-06, + "loss": 0.9867, + "step": 49671 + }, + { + "epoch": 3.67, + "learning_rate": 6.917873368520807e-06, + "loss": 1.0194, + "step": 49672 + }, + { + "epoch": 3.67, + "learning_rate": 6.917493917426947e-06, + "loss": 1.0538, + "step": 49673 + }, + { + "epoch": 3.67, + "learning_rate": 6.917114471237317e-06, + "loss": 1.0552, + "step": 49674 + }, + { + "epoch": 3.67, + "learning_rate": 6.916735029952507e-06, + "loss": 1.0282, + "step": 49675 + }, + { + "epoch": 3.67, + "learning_rate": 6.916355593573134e-06, + "loss": 1.0263, + "step": 49676 + }, + { + "epoch": 3.67, + "learning_rate": 6.915976162099789e-06, + "loss": 1.0467, + "step": 49677 + }, + { + "epoch": 3.67, + "learning_rate": 6.915596735533088e-06, + "loss": 1.0198, + "step": 49678 + }, + { + "epoch": 3.67, + "learning_rate": 6.9152173138736254e-06, + "loss": 1.0622, + "step": 49679 + }, + { + "epoch": 3.67, + "learning_rate": 6.914837897122007e-06, + "loss": 1.0552, + "step": 49680 + }, + { + "epoch": 3.67, + "learning_rate": 6.914458485278839e-06, + "loss": 1.0677, + "step": 49681 + }, + { + "epoch": 3.67, + "learning_rate": 6.9140790783447205e-06, + "loss": 1.0215, + "step": 49682 + }, + { + "epoch": 3.67, + "learning_rate": 6.913699676320261e-06, + "loss": 1.0155, + "step": 49683 + }, + { + "epoch": 3.67, + "learning_rate": 6.913320279206059e-06, + "loss": 0.9969, + "step": 49684 + }, + { + "epoch": 3.67, + "learning_rate": 6.9129408870027225e-06, + "loss": 1.1031, + "step": 49685 + }, + { + "epoch": 3.67, + "learning_rate": 6.91256149971085e-06, + "loss": 0.9283, + "step": 49686 + }, + { + "epoch": 3.67, + "learning_rate": 6.912182117331051e-06, + "loss": 0.9685, + "step": 49687 + }, + { + "epoch": 3.67, + "learning_rate": 6.9118027398639265e-06, + "loss": 1.0317, + "step": 49688 + }, + { + "epoch": 3.67, + "learning_rate": 6.911423367310076e-06, + "loss": 1.0066, + "step": 49689 + }, + { + "epoch": 3.67, + "learning_rate": 6.91104399967011e-06, + "loss": 1.088, + "step": 49690 + }, + { + "epoch": 3.67, + "learning_rate": 6.910664636944626e-06, + "loss": 1.02, + "step": 49691 + }, + { + "epoch": 3.67, + "learning_rate": 6.910285279134234e-06, + "loss": 1.0399, + "step": 49692 + }, + { + "epoch": 3.67, + "learning_rate": 6.909905926239529e-06, + "loss": 0.9294, + "step": 49693 + }, + { + "epoch": 3.67, + "learning_rate": 6.909526578261125e-06, + "loss": 1.0624, + "step": 49694 + }, + { + "epoch": 3.67, + "learning_rate": 6.909147235199616e-06, + "loss": 1.1011, + "step": 49695 + }, + { + "epoch": 3.67, + "learning_rate": 6.9087678970556125e-06, + "loss": 1.0741, + "step": 49696 + }, + { + "epoch": 3.67, + "learning_rate": 6.908388563829715e-06, + "loss": 0.971, + "step": 49697 + }, + { + "epoch": 3.67, + "learning_rate": 6.908009235522524e-06, + "loss": 0.8838, + "step": 49698 + }, + { + "epoch": 3.67, + "learning_rate": 6.9076299121346476e-06, + "loss": 1.0244, + "step": 49699 + }, + { + "epoch": 3.67, + "learning_rate": 6.9072505936666875e-06, + "loss": 0.9927, + "step": 49700 + }, + { + "epoch": 3.67, + "learning_rate": 6.906871280119249e-06, + "loss": 1.0394, + "step": 49701 + }, + { + "epoch": 3.67, + "learning_rate": 6.9064919714929305e-06, + "loss": 0.9925, + "step": 49702 + }, + { + "epoch": 3.67, + "learning_rate": 6.906112667788343e-06, + "loss": 0.9987, + "step": 49703 + }, + { + "epoch": 3.67, + "learning_rate": 6.905733369006083e-06, + "loss": 0.9956, + "step": 49704 + }, + { + "epoch": 3.67, + "learning_rate": 6.905354075146757e-06, + "loss": 1.0318, + "step": 49705 + }, + { + "epoch": 3.67, + "learning_rate": 6.904974786210969e-06, + "loss": 1.1215, + "step": 49706 + }, + { + "epoch": 3.67, + "learning_rate": 6.904595502199324e-06, + "loss": 0.9756, + "step": 49707 + }, + { + "epoch": 3.67, + "learning_rate": 6.9042162231124224e-06, + "loss": 1.0402, + "step": 49708 + }, + { + "epoch": 3.67, + "learning_rate": 6.903836948950866e-06, + "loss": 1.0491, + "step": 49709 + }, + { + "epoch": 3.67, + "learning_rate": 6.903457679715264e-06, + "loss": 1.0616, + "step": 49710 + }, + { + "epoch": 3.67, + "learning_rate": 6.903078415406213e-06, + "loss": 1.0093, + "step": 49711 + }, + { + "epoch": 3.67, + "learning_rate": 6.9026991560243225e-06, + "loss": 1.0756, + "step": 49712 + }, + { + "epoch": 3.67, + "learning_rate": 6.902319901570191e-06, + "loss": 1.1282, + "step": 49713 + }, + { + "epoch": 3.67, + "learning_rate": 6.901940652044427e-06, + "loss": 1.1114, + "step": 49714 + }, + { + "epoch": 3.67, + "learning_rate": 6.90156140744763e-06, + "loss": 0.9667, + "step": 49715 + }, + { + "epoch": 3.67, + "learning_rate": 6.901182167780405e-06, + "loss": 1.0575, + "step": 49716 + }, + { + "epoch": 3.67, + "learning_rate": 6.9008029330433575e-06, + "loss": 1.0245, + "step": 49717 + }, + { + "epoch": 3.67, + "learning_rate": 6.900423703237081e-06, + "loss": 1.1473, + "step": 49718 + }, + { + "epoch": 3.67, + "learning_rate": 6.900044478362194e-06, + "loss": 0.9537, + "step": 49719 + }, + { + "epoch": 3.67, + "learning_rate": 6.899665258419283e-06, + "loss": 1.0892, + "step": 49720 + }, + { + "epoch": 3.67, + "learning_rate": 6.899286043408969e-06, + "loss": 1.033, + "step": 49721 + }, + { + "epoch": 3.67, + "learning_rate": 6.89890683333184e-06, + "loss": 1.0215, + "step": 49722 + }, + { + "epoch": 3.67, + "learning_rate": 6.898527628188511e-06, + "loss": 1.0762, + "step": 49723 + }, + { + "epoch": 3.67, + "learning_rate": 6.898148427979574e-06, + "loss": 0.9879, + "step": 49724 + }, + { + "epoch": 3.67, + "learning_rate": 6.897769232705646e-06, + "loss": 0.9715, + "step": 49725 + }, + { + "epoch": 3.67, + "learning_rate": 6.897390042367319e-06, + "loss": 0.9933, + "step": 49726 + }, + { + "epoch": 3.67, + "learning_rate": 6.897010856965198e-06, + "loss": 0.9225, + "step": 49727 + }, + { + "epoch": 3.67, + "learning_rate": 6.896631676499892e-06, + "loss": 1.1022, + "step": 49728 + }, + { + "epoch": 3.67, + "learning_rate": 6.896252500971999e-06, + "loss": 0.8922, + "step": 49729 + }, + { + "epoch": 3.67, + "learning_rate": 6.895873330382124e-06, + "loss": 1.1869, + "step": 49730 + }, + { + "epoch": 3.67, + "learning_rate": 6.895494164730868e-06, + "loss": 1.0283, + "step": 49731 + }, + { + "epoch": 3.67, + "learning_rate": 6.89511500401884e-06, + "loss": 0.9849, + "step": 49732 + }, + { + "epoch": 3.67, + "learning_rate": 6.894735848246636e-06, + "loss": 0.9743, + "step": 49733 + }, + { + "epoch": 3.67, + "learning_rate": 6.894356697414866e-06, + "loss": 1.0406, + "step": 49734 + }, + { + "epoch": 3.67, + "learning_rate": 6.8939775515241296e-06, + "loss": 1.0389, + "step": 49735 + }, + { + "epoch": 3.67, + "learning_rate": 6.893598410575029e-06, + "loss": 0.9514, + "step": 49736 + }, + { + "epoch": 3.68, + "learning_rate": 6.893219274568172e-06, + "loss": 0.9968, + "step": 49737 + }, + { + "epoch": 3.68, + "learning_rate": 6.892840143504152e-06, + "loss": 1.0784, + "step": 49738 + }, + { + "epoch": 3.68, + "learning_rate": 6.892461017383587e-06, + "loss": 1.0349, + "step": 49739 + }, + { + "epoch": 3.68, + "learning_rate": 6.892081896207065e-06, + "loss": 1.0078, + "step": 49740 + }, + { + "epoch": 3.68, + "learning_rate": 6.891702779975202e-06, + "loss": 1.0277, + "step": 49741 + }, + { + "epoch": 3.68, + "learning_rate": 6.89132366868859e-06, + "loss": 1.0413, + "step": 49742 + }, + { + "epoch": 3.68, + "learning_rate": 6.8909445623478434e-06, + "loss": 1.0156, + "step": 49743 + }, + { + "epoch": 3.68, + "learning_rate": 6.890565460953557e-06, + "loss": 1.0346, + "step": 49744 + }, + { + "epoch": 3.68, + "learning_rate": 6.890186364506335e-06, + "loss": 0.9491, + "step": 49745 + }, + { + "epoch": 3.68, + "learning_rate": 6.889807273006783e-06, + "loss": 0.9608, + "step": 49746 + }, + { + "epoch": 3.68, + "learning_rate": 6.889428186455502e-06, + "loss": 0.9929, + "step": 49747 + }, + { + "epoch": 3.68, + "learning_rate": 6.889049104853098e-06, + "loss": 0.8502, + "step": 49748 + }, + { + "epoch": 3.68, + "learning_rate": 6.888670028200171e-06, + "loss": 0.8836, + "step": 49749 + }, + { + "epoch": 3.68, + "learning_rate": 6.8882909564973265e-06, + "loss": 1.0601, + "step": 49750 + }, + { + "epoch": 3.68, + "learning_rate": 6.8879118897451645e-06, + "loss": 0.952, + "step": 49751 + }, + { + "epoch": 3.68, + "learning_rate": 6.887532827944293e-06, + "loss": 1.0208, + "step": 49752 + }, + { + "epoch": 3.68, + "learning_rate": 6.88715377109531e-06, + "loss": 0.8869, + "step": 49753 + }, + { + "epoch": 3.68, + "learning_rate": 6.886774719198823e-06, + "loss": 1.0541, + "step": 49754 + }, + { + "epoch": 3.68, + "learning_rate": 6.886395672255432e-06, + "loss": 0.9698, + "step": 49755 + }, + { + "epoch": 3.68, + "learning_rate": 6.886016630265741e-06, + "loss": 0.9478, + "step": 49756 + }, + { + "epoch": 3.68, + "learning_rate": 6.885637593230352e-06, + "loss": 1.12, + "step": 49757 + }, + { + "epoch": 3.68, + "learning_rate": 6.8852585611498705e-06, + "loss": 1.0178, + "step": 49758 + }, + { + "epoch": 3.68, + "learning_rate": 6.884879534024897e-06, + "loss": 1.1048, + "step": 49759 + }, + { + "epoch": 3.68, + "learning_rate": 6.884500511856035e-06, + "loss": 0.9699, + "step": 49760 + }, + { + "epoch": 3.68, + "learning_rate": 6.884121494643889e-06, + "loss": 1.1677, + "step": 49761 + }, + { + "epoch": 3.68, + "learning_rate": 6.8837424823890604e-06, + "loss": 1.0114, + "step": 49762 + }, + { + "epoch": 3.68, + "learning_rate": 6.883363475092155e-06, + "loss": 0.9878, + "step": 49763 + }, + { + "epoch": 3.68, + "learning_rate": 6.882984472753775e-06, + "loss": 1.0368, + "step": 49764 + }, + { + "epoch": 3.68, + "learning_rate": 6.882605475374516e-06, + "loss": 1.1386, + "step": 49765 + }, + { + "epoch": 3.68, + "learning_rate": 6.882226482954994e-06, + "loss": 1.0104, + "step": 49766 + }, + { + "epoch": 3.68, + "learning_rate": 6.881847495495798e-06, + "loss": 1.012, + "step": 49767 + }, + { + "epoch": 3.68, + "learning_rate": 6.881468512997544e-06, + "loss": 1.0394, + "step": 49768 + }, + { + "epoch": 3.68, + "learning_rate": 6.881089535460824e-06, + "loss": 0.9984, + "step": 49769 + }, + { + "epoch": 3.68, + "learning_rate": 6.880710562886252e-06, + "loss": 1.0189, + "step": 49770 + }, + { + "epoch": 3.68, + "learning_rate": 6.8803315952744196e-06, + "loss": 0.9772, + "step": 49771 + }, + { + "epoch": 3.68, + "learning_rate": 6.879952632625938e-06, + "loss": 1.0259, + "step": 49772 + }, + { + "epoch": 3.68, + "learning_rate": 6.879573674941407e-06, + "loss": 0.9417, + "step": 49773 + }, + { + "epoch": 3.68, + "learning_rate": 6.879194722221427e-06, + "loss": 1.081, + "step": 49774 + }, + { + "epoch": 3.68, + "learning_rate": 6.878815774466606e-06, + "loss": 1.0134, + "step": 49775 + }, + { + "epoch": 3.68, + "learning_rate": 6.878436831677542e-06, + "loss": 0.9981, + "step": 49776 + }, + { + "epoch": 3.68, + "learning_rate": 6.878057893854843e-06, + "loss": 1.0627, + "step": 49777 + }, + { + "epoch": 3.68, + "learning_rate": 6.877678960999108e-06, + "loss": 0.9392, + "step": 49778 + }, + { + "epoch": 3.68, + "learning_rate": 6.877300033110941e-06, + "loss": 1.1237, + "step": 49779 + }, + { + "epoch": 3.68, + "learning_rate": 6.876921110190944e-06, + "loss": 0.9555, + "step": 49780 + }, + { + "epoch": 3.68, + "learning_rate": 6.876542192239724e-06, + "loss": 0.9894, + "step": 49781 + }, + { + "epoch": 3.68, + "learning_rate": 6.8761632792578815e-06, + "loss": 1.0729, + "step": 49782 + }, + { + "epoch": 3.68, + "learning_rate": 6.875784371246013e-06, + "loss": 1.0948, + "step": 49783 + }, + { + "epoch": 3.68, + "learning_rate": 6.875405468204732e-06, + "loss": 1.1353, + "step": 49784 + }, + { + "epoch": 3.68, + "learning_rate": 6.87502657013463e-06, + "loss": 1.0521, + "step": 49785 + }, + { + "epoch": 3.68, + "learning_rate": 6.8746476770363236e-06, + "loss": 1.0137, + "step": 49786 + }, + { + "epoch": 3.68, + "learning_rate": 6.8742687889104e-06, + "loss": 1.0024, + "step": 49787 + }, + { + "epoch": 3.68, + "learning_rate": 6.873889905757477e-06, + "loss": 1.0572, + "step": 49788 + }, + { + "epoch": 3.68, + "learning_rate": 6.873511027578145e-06, + "loss": 0.9789, + "step": 49789 + }, + { + "epoch": 3.68, + "learning_rate": 6.873132154373018e-06, + "loss": 1.0138, + "step": 49790 + }, + { + "epoch": 3.68, + "learning_rate": 6.87275328614269e-06, + "loss": 1.0226, + "step": 49791 + }, + { + "epoch": 3.68, + "learning_rate": 6.872374422887766e-06, + "loss": 0.9121, + "step": 49792 + }, + { + "epoch": 3.68, + "learning_rate": 6.871995564608851e-06, + "loss": 1.0943, + "step": 49793 + }, + { + "epoch": 3.68, + "learning_rate": 6.871616711306544e-06, + "loss": 0.9258, + "step": 49794 + }, + { + "epoch": 3.68, + "learning_rate": 6.871237862981453e-06, + "loss": 0.9927, + "step": 49795 + }, + { + "epoch": 3.68, + "learning_rate": 6.8708590196341755e-06, + "loss": 1.0118, + "step": 49796 + }, + { + "epoch": 3.68, + "learning_rate": 6.8704801812653175e-06, + "loss": 0.9724, + "step": 49797 + }, + { + "epoch": 3.68, + "learning_rate": 6.870101347875479e-06, + "loss": 1.0815, + "step": 49798 + }, + { + "epoch": 3.68, + "learning_rate": 6.869722519465267e-06, + "loss": 1.0672, + "step": 49799 + }, + { + "epoch": 3.68, + "learning_rate": 6.869343696035282e-06, + "loss": 0.9615, + "step": 49800 + }, + { + "epoch": 3.68, + "learning_rate": 6.868964877586124e-06, + "loss": 1.0491, + "step": 49801 + }, + { + "epoch": 3.68, + "learning_rate": 6.868586064118401e-06, + "loss": 0.9647, + "step": 49802 + }, + { + "epoch": 3.68, + "learning_rate": 6.86820725563271e-06, + "loss": 0.8724, + "step": 49803 + }, + { + "epoch": 3.68, + "learning_rate": 6.867828452129659e-06, + "loss": 0.9638, + "step": 49804 + }, + { + "epoch": 3.68, + "learning_rate": 6.867449653609844e-06, + "loss": 1.0445, + "step": 49805 + }, + { + "epoch": 3.68, + "learning_rate": 6.867070860073877e-06, + "loss": 1.0096, + "step": 49806 + }, + { + "epoch": 3.68, + "learning_rate": 6.86669207152235e-06, + "loss": 0.9887, + "step": 49807 + }, + { + "epoch": 3.68, + "learning_rate": 6.866313287955874e-06, + "loss": 0.8965, + "step": 49808 + }, + { + "epoch": 3.68, + "learning_rate": 6.86593450937505e-06, + "loss": 0.9945, + "step": 49809 + }, + { + "epoch": 3.68, + "learning_rate": 6.865555735780478e-06, + "loss": 1.1285, + "step": 49810 + }, + { + "epoch": 3.68, + "learning_rate": 6.865176967172764e-06, + "loss": 1.0215, + "step": 49811 + }, + { + "epoch": 3.68, + "learning_rate": 6.864798203552504e-06, + "loss": 0.9926, + "step": 49812 + }, + { + "epoch": 3.68, + "learning_rate": 6.864419444920312e-06, + "loss": 1.0153, + "step": 49813 + }, + { + "epoch": 3.68, + "learning_rate": 6.8640406912767775e-06, + "loss": 0.9265, + "step": 49814 + }, + { + "epoch": 3.68, + "learning_rate": 6.863661942622513e-06, + "loss": 1.0524, + "step": 49815 + }, + { + "epoch": 3.68, + "learning_rate": 6.863283198958114e-06, + "loss": 0.9703, + "step": 49816 + }, + { + "epoch": 3.68, + "learning_rate": 6.86290446028419e-06, + "loss": 0.933, + "step": 49817 + }, + { + "epoch": 3.68, + "learning_rate": 6.862525726601337e-06, + "loss": 1.0815, + "step": 49818 + }, + { + "epoch": 3.68, + "learning_rate": 6.862146997910163e-06, + "loss": 0.9934, + "step": 49819 + }, + { + "epoch": 3.68, + "learning_rate": 6.861768274211269e-06, + "loss": 1.0416, + "step": 49820 + }, + { + "epoch": 3.68, + "learning_rate": 6.861389555505253e-06, + "loss": 1.0031, + "step": 49821 + }, + { + "epoch": 3.68, + "learning_rate": 6.861010841792725e-06, + "loss": 0.9557, + "step": 49822 + }, + { + "epoch": 3.68, + "learning_rate": 6.860632133074281e-06, + "loss": 0.8996, + "step": 49823 + }, + { + "epoch": 3.68, + "learning_rate": 6.86025342935053e-06, + "loss": 0.9286, + "step": 49824 + }, + { + "epoch": 3.68, + "learning_rate": 6.859874730622067e-06, + "loss": 1.0746, + "step": 49825 + }, + { + "epoch": 3.68, + "learning_rate": 6.859496036889501e-06, + "loss": 0.9247, + "step": 49826 + }, + { + "epoch": 3.68, + "learning_rate": 6.85911734815343e-06, + "loss": 0.8835, + "step": 49827 + }, + { + "epoch": 3.68, + "learning_rate": 6.85873866441446e-06, + "loss": 0.9806, + "step": 49828 + }, + { + "epoch": 3.68, + "learning_rate": 6.858359985673193e-06, + "loss": 1.0841, + "step": 49829 + }, + { + "epoch": 3.68, + "learning_rate": 6.857981311930227e-06, + "loss": 0.9854, + "step": 49830 + }, + { + "epoch": 3.68, + "learning_rate": 6.857602643186172e-06, + "loss": 1.0096, + "step": 49831 + }, + { + "epoch": 3.68, + "learning_rate": 6.857223979441619e-06, + "loss": 0.8873, + "step": 49832 + }, + { + "epoch": 3.68, + "learning_rate": 6.856845320697185e-06, + "loss": 0.9844, + "step": 49833 + }, + { + "epoch": 3.68, + "learning_rate": 6.85646666695346e-06, + "loss": 0.9956, + "step": 49834 + }, + { + "epoch": 3.68, + "learning_rate": 6.856088018211056e-06, + "loss": 0.938, + "step": 49835 + }, + { + "epoch": 3.68, + "learning_rate": 6.855709374470566e-06, + "loss": 0.9554, + "step": 49836 + }, + { + "epoch": 3.68, + "learning_rate": 6.8553307357326035e-06, + "loss": 1.0982, + "step": 49837 + }, + { + "epoch": 3.68, + "learning_rate": 6.854952101997763e-06, + "loss": 0.9843, + "step": 49838 + }, + { + "epoch": 3.68, + "learning_rate": 6.854573473266646e-06, + "loss": 1.1159, + "step": 49839 + }, + { + "epoch": 3.68, + "learning_rate": 6.854194849539859e-06, + "loss": 0.8998, + "step": 49840 + }, + { + "epoch": 3.68, + "learning_rate": 6.853816230818001e-06, + "loss": 0.9515, + "step": 49841 + }, + { + "epoch": 3.68, + "learning_rate": 6.853437617101679e-06, + "loss": 0.9883, + "step": 49842 + }, + { + "epoch": 3.68, + "learning_rate": 6.853059008391492e-06, + "loss": 1.1118, + "step": 49843 + }, + { + "epoch": 3.68, + "learning_rate": 6.852680404688043e-06, + "loss": 1.121, + "step": 49844 + }, + { + "epoch": 3.68, + "learning_rate": 6.852301805991933e-06, + "loss": 1.0316, + "step": 49845 + }, + { + "epoch": 3.68, + "learning_rate": 6.851923212303768e-06, + "loss": 0.9038, + "step": 49846 + }, + { + "epoch": 3.68, + "learning_rate": 6.851544623624149e-06, + "loss": 1.0352, + "step": 49847 + }, + { + "epoch": 3.68, + "learning_rate": 6.851166039953674e-06, + "loss": 1.0172, + "step": 49848 + }, + { + "epoch": 3.68, + "learning_rate": 6.850787461292951e-06, + "loss": 1.0394, + "step": 49849 + }, + { + "epoch": 3.68, + "learning_rate": 6.850408887642579e-06, + "loss": 0.9292, + "step": 49850 + }, + { + "epoch": 3.68, + "learning_rate": 6.850030319003163e-06, + "loss": 1.0076, + "step": 49851 + }, + { + "epoch": 3.68, + "learning_rate": 6.849651755375302e-06, + "loss": 0.8952, + "step": 49852 + }, + { + "epoch": 3.68, + "learning_rate": 6.849273196759602e-06, + "loss": 1.0885, + "step": 49853 + }, + { + "epoch": 3.68, + "learning_rate": 6.84889464315666e-06, + "loss": 1.0242, + "step": 49854 + }, + { + "epoch": 3.68, + "learning_rate": 6.848516094567085e-06, + "loss": 1.0188, + "step": 49855 + }, + { + "epoch": 3.68, + "learning_rate": 6.848137550991477e-06, + "loss": 1.204, + "step": 49856 + }, + { + "epoch": 3.68, + "learning_rate": 6.847759012430431e-06, + "loss": 0.985, + "step": 49857 + }, + { + "epoch": 3.68, + "learning_rate": 6.847380478884562e-06, + "loss": 0.9932, + "step": 49858 + }, + { + "epoch": 3.68, + "learning_rate": 6.847001950354462e-06, + "loss": 1.1385, + "step": 49859 + }, + { + "epoch": 3.68, + "learning_rate": 6.846623426840737e-06, + "loss": 1.1139, + "step": 49860 + }, + { + "epoch": 3.68, + "learning_rate": 6.846244908343988e-06, + "loss": 0.9649, + "step": 49861 + }, + { + "epoch": 3.68, + "learning_rate": 6.84586639486482e-06, + "loss": 0.9935, + "step": 49862 + }, + { + "epoch": 3.68, + "learning_rate": 6.8454878864038325e-06, + "loss": 1.1522, + "step": 49863 + }, + { + "epoch": 3.68, + "learning_rate": 6.845109382961629e-06, + "loss": 0.9621, + "step": 49864 + }, + { + "epoch": 3.68, + "learning_rate": 6.844730884538811e-06, + "loss": 1.0225, + "step": 49865 + }, + { + "epoch": 3.68, + "learning_rate": 6.844352391135981e-06, + "loss": 0.8938, + "step": 49866 + }, + { + "epoch": 3.68, + "learning_rate": 6.843973902753742e-06, + "loss": 1.0958, + "step": 49867 + }, + { + "epoch": 3.68, + "learning_rate": 6.843595419392694e-06, + "loss": 0.9612, + "step": 49868 + }, + { + "epoch": 3.68, + "learning_rate": 6.843216941053443e-06, + "loss": 1.0036, + "step": 49869 + }, + { + "epoch": 3.68, + "learning_rate": 6.842838467736586e-06, + "loss": 1.0045, + "step": 49870 + }, + { + "epoch": 3.68, + "learning_rate": 6.8424599994427284e-06, + "loss": 1.022, + "step": 49871 + }, + { + "epoch": 3.69, + "learning_rate": 6.842081536172471e-06, + "loss": 0.8999, + "step": 49872 + }, + { + "epoch": 3.69, + "learning_rate": 6.8417030779264195e-06, + "loss": 1.0705, + "step": 49873 + }, + { + "epoch": 3.69, + "learning_rate": 6.84132462470517e-06, + "loss": 0.9628, + "step": 49874 + }, + { + "epoch": 3.69, + "learning_rate": 6.840946176509331e-06, + "loss": 1.0321, + "step": 49875 + }, + { + "epoch": 3.69, + "learning_rate": 6.840567733339503e-06, + "loss": 0.9987, + "step": 49876 + }, + { + "epoch": 3.69, + "learning_rate": 6.84018929519628e-06, + "loss": 1.0212, + "step": 49877 + }, + { + "epoch": 3.69, + "learning_rate": 6.839810862080276e-06, + "loss": 1.0971, + "step": 49878 + }, + { + "epoch": 3.69, + "learning_rate": 6.8394324339920834e-06, + "loss": 0.874, + "step": 49879 + }, + { + "epoch": 3.69, + "learning_rate": 6.839054010932313e-06, + "loss": 1.0378, + "step": 49880 + }, + { + "epoch": 3.69, + "learning_rate": 6.838675592901558e-06, + "loss": 1.0846, + "step": 49881 + }, + { + "epoch": 3.69, + "learning_rate": 6.83829717990043e-06, + "loss": 0.9081, + "step": 49882 + }, + { + "epoch": 3.69, + "learning_rate": 6.837918771929519e-06, + "loss": 0.9813, + "step": 49883 + }, + { + "epoch": 3.69, + "learning_rate": 6.837540368989441e-06, + "loss": 0.8454, + "step": 49884 + }, + { + "epoch": 3.69, + "learning_rate": 6.83716197108079e-06, + "loss": 0.958, + "step": 49885 + }, + { + "epoch": 3.69, + "learning_rate": 6.836783578204167e-06, + "loss": 1.0382, + "step": 49886 + }, + { + "epoch": 3.69, + "learning_rate": 6.836405190360176e-06, + "loss": 1.1184, + "step": 49887 + }, + { + "epoch": 3.69, + "learning_rate": 6.836026807549418e-06, + "loss": 0.8727, + "step": 49888 + }, + { + "epoch": 3.69, + "learning_rate": 6.835648429772499e-06, + "loss": 0.9975, + "step": 49889 + }, + { + "epoch": 3.69, + "learning_rate": 6.835270057030014e-06, + "loss": 1.055, + "step": 49890 + }, + { + "epoch": 3.69, + "learning_rate": 6.834891689322573e-06, + "loss": 0.9695, + "step": 49891 + }, + { + "epoch": 3.69, + "learning_rate": 6.834513326650773e-06, + "loss": 0.9498, + "step": 49892 + }, + { + "epoch": 3.69, + "learning_rate": 6.834134969015217e-06, + "loss": 0.9869, + "step": 49893 + }, + { + "epoch": 3.69, + "learning_rate": 6.833756616416507e-06, + "loss": 0.9214, + "step": 49894 + }, + { + "epoch": 3.69, + "learning_rate": 6.833378268855244e-06, + "loss": 0.9416, + "step": 49895 + }, + { + "epoch": 3.69, + "learning_rate": 6.832999926332032e-06, + "loss": 1.1278, + "step": 49896 + }, + { + "epoch": 3.69, + "learning_rate": 6.83262158884747e-06, + "loss": 1.0561, + "step": 49897 + }, + { + "epoch": 3.69, + "learning_rate": 6.832243256402164e-06, + "loss": 1.1101, + "step": 49898 + }, + { + "epoch": 3.69, + "learning_rate": 6.831864928996712e-06, + "loss": 0.9754, + "step": 49899 + }, + { + "epoch": 3.69, + "learning_rate": 6.8314866066317185e-06, + "loss": 1.0436, + "step": 49900 + }, + { + "epoch": 3.69, + "learning_rate": 6.8311082893077826e-06, + "loss": 0.9721, + "step": 49901 + }, + { + "epoch": 3.69, + "learning_rate": 6.830729977025511e-06, + "loss": 0.9554, + "step": 49902 + }, + { + "epoch": 3.69, + "learning_rate": 6.830351669785504e-06, + "loss": 1.0515, + "step": 49903 + }, + { + "epoch": 3.69, + "learning_rate": 6.829973367588359e-06, + "loss": 0.9812, + "step": 49904 + }, + { + "epoch": 3.69, + "learning_rate": 6.829595070434683e-06, + "loss": 1.0197, + "step": 49905 + }, + { + "epoch": 3.69, + "learning_rate": 6.829216778325072e-06, + "loss": 1.028, + "step": 49906 + }, + { + "epoch": 3.69, + "learning_rate": 6.8288384912601345e-06, + "loss": 0.9847, + "step": 49907 + }, + { + "epoch": 3.69, + "learning_rate": 6.828460209240468e-06, + "loss": 1.0245, + "step": 49908 + }, + { + "epoch": 3.69, + "learning_rate": 6.828081932266679e-06, + "loss": 0.9818, + "step": 49909 + }, + { + "epoch": 3.69, + "learning_rate": 6.827703660339362e-06, + "loss": 0.997, + "step": 49910 + }, + { + "epoch": 3.69, + "learning_rate": 6.827325393459127e-06, + "loss": 0.9306, + "step": 49911 + }, + { + "epoch": 3.69, + "learning_rate": 6.826947131626571e-06, + "loss": 0.9602, + "step": 49912 + }, + { + "epoch": 3.69, + "learning_rate": 6.826568874842295e-06, + "loss": 1.0279, + "step": 49913 + }, + { + "epoch": 3.69, + "learning_rate": 6.826190623106905e-06, + "loss": 0.9519, + "step": 49914 + }, + { + "epoch": 3.69, + "learning_rate": 6.825812376420997e-06, + "loss": 0.9058, + "step": 49915 + }, + { + "epoch": 3.69, + "learning_rate": 6.8254341347851795e-06, + "loss": 0.9657, + "step": 49916 + }, + { + "epoch": 3.69, + "learning_rate": 6.825055898200048e-06, + "loss": 1.1231, + "step": 49917 + }, + { + "epoch": 3.69, + "learning_rate": 6.824677666666209e-06, + "loss": 1.0989, + "step": 49918 + }, + { + "epoch": 3.69, + "learning_rate": 6.824299440184261e-06, + "loss": 1.0774, + "step": 49919 + }, + { + "epoch": 3.69, + "learning_rate": 6.82392121875481e-06, + "loss": 0.9698, + "step": 49920 + }, + { + "epoch": 3.69, + "learning_rate": 6.823543002378451e-06, + "loss": 1.018, + "step": 49921 + }, + { + "epoch": 3.69, + "learning_rate": 6.8231647910557915e-06, + "loss": 1.0598, + "step": 49922 + }, + { + "epoch": 3.69, + "learning_rate": 6.8227865847874355e-06, + "loss": 0.974, + "step": 49923 + }, + { + "epoch": 3.69, + "learning_rate": 6.822408383573973e-06, + "loss": 1.0606, + "step": 49924 + }, + { + "epoch": 3.69, + "learning_rate": 6.82203018741602e-06, + "loss": 0.9827, + "step": 49925 + }, + { + "epoch": 3.69, + "learning_rate": 6.821651996314165e-06, + "loss": 0.9612, + "step": 49926 + }, + { + "epoch": 3.69, + "learning_rate": 6.821273810269023e-06, + "loss": 0.8739, + "step": 49927 + }, + { + "epoch": 3.69, + "learning_rate": 6.820895629281181e-06, + "loss": 1.0029, + "step": 49928 + }, + { + "epoch": 3.69, + "learning_rate": 6.820517453351257e-06, + "loss": 1.0063, + "step": 49929 + }, + { + "epoch": 3.69, + "learning_rate": 6.820139282479837e-06, + "loss": 0.9469, + "step": 49930 + }, + { + "epoch": 3.69, + "learning_rate": 6.819761116667536e-06, + "loss": 0.9706, + "step": 49931 + }, + { + "epoch": 3.69, + "learning_rate": 6.819382955914947e-06, + "loss": 1.0094, + "step": 49932 + }, + { + "epoch": 3.69, + "learning_rate": 6.819004800222673e-06, + "loss": 0.9054, + "step": 49933 + }, + { + "epoch": 3.69, + "learning_rate": 6.818626649591319e-06, + "loss": 1.0628, + "step": 49934 + }, + { + "epoch": 3.69, + "learning_rate": 6.818248504021482e-06, + "loss": 1.0025, + "step": 49935 + }, + { + "epoch": 3.69, + "learning_rate": 6.8178703635137676e-06, + "loss": 1.109, + "step": 49936 + }, + { + "epoch": 3.69, + "learning_rate": 6.817492228068775e-06, + "loss": 1.0425, + "step": 49937 + }, + { + "epoch": 3.69, + "learning_rate": 6.817114097687107e-06, + "loss": 0.9696, + "step": 49938 + }, + { + "epoch": 3.69, + "learning_rate": 6.8167359723693664e-06, + "loss": 1.0188, + "step": 49939 + }, + { + "epoch": 3.69, + "learning_rate": 6.8163578521161524e-06, + "loss": 0.9555, + "step": 49940 + }, + { + "epoch": 3.69, + "learning_rate": 6.8159797369280676e-06, + "loss": 1.14, + "step": 49941 + }, + { + "epoch": 3.69, + "learning_rate": 6.815601626805712e-06, + "loss": 0.914, + "step": 49942 + }, + { + "epoch": 3.69, + "learning_rate": 6.815223521749692e-06, + "loss": 0.9911, + "step": 49943 + }, + { + "epoch": 3.69, + "learning_rate": 6.8148454217606016e-06, + "loss": 0.999, + "step": 49944 + }, + { + "epoch": 3.69, + "learning_rate": 6.81446732683905e-06, + "loss": 1.0902, + "step": 49945 + }, + { + "epoch": 3.69, + "learning_rate": 6.814089236985633e-06, + "loss": 0.9514, + "step": 49946 + }, + { + "epoch": 3.69, + "learning_rate": 6.813711152200958e-06, + "loss": 0.9953, + "step": 49947 + }, + { + "epoch": 3.69, + "learning_rate": 6.813333072485617e-06, + "loss": 0.8934, + "step": 49948 + }, + { + "epoch": 3.69, + "learning_rate": 6.8129549978402244e-06, + "loss": 1.0333, + "step": 49949 + }, + { + "epoch": 3.69, + "learning_rate": 6.812576928265372e-06, + "loss": 1.0252, + "step": 49950 + }, + { + "epoch": 3.69, + "learning_rate": 6.812198863761662e-06, + "loss": 1.0111, + "step": 49951 + }, + { + "epoch": 3.69, + "learning_rate": 6.811820804329701e-06, + "loss": 1.0051, + "step": 49952 + }, + { + "epoch": 3.69, + "learning_rate": 6.811442749970083e-06, + "loss": 0.9784, + "step": 49953 + }, + { + "epoch": 3.69, + "learning_rate": 6.8110647006834185e-06, + "loss": 1.0161, + "step": 49954 + }, + { + "epoch": 3.69, + "learning_rate": 6.810686656470302e-06, + "loss": 0.9964, + "step": 49955 + }, + { + "epoch": 3.69, + "learning_rate": 6.810308617331338e-06, + "loss": 1.0641, + "step": 49956 + }, + { + "epoch": 3.69, + "learning_rate": 6.8099305832671265e-06, + "loss": 0.8846, + "step": 49957 + }, + { + "epoch": 3.69, + "learning_rate": 6.809552554278271e-06, + "loss": 0.9703, + "step": 49958 + }, + { + "epoch": 3.69, + "learning_rate": 6.809174530365374e-06, + "loss": 0.9675, + "step": 49959 + }, + { + "epoch": 3.69, + "learning_rate": 6.80879651152903e-06, + "loss": 0.9843, + "step": 49960 + }, + { + "epoch": 3.69, + "learning_rate": 6.808418497769848e-06, + "loss": 0.9914, + "step": 49961 + }, + { + "epoch": 3.69, + "learning_rate": 6.808040489088426e-06, + "loss": 1.0677, + "step": 49962 + }, + { + "epoch": 3.69, + "learning_rate": 6.807662485485367e-06, + "loss": 0.9882, + "step": 49963 + }, + { + "epoch": 3.69, + "learning_rate": 6.807284486961268e-06, + "loss": 1.0259, + "step": 49964 + }, + { + "epoch": 3.69, + "learning_rate": 6.806906493516736e-06, + "loss": 1.0047, + "step": 49965 + }, + { + "epoch": 3.69, + "learning_rate": 6.80652850515237e-06, + "loss": 0.9106, + "step": 49966 + }, + { + "epoch": 3.69, + "learning_rate": 6.806150521868771e-06, + "loss": 0.9879, + "step": 49967 + }, + { + "epoch": 3.69, + "learning_rate": 6.80577254366654e-06, + "loss": 1.0025, + "step": 49968 + }, + { + "epoch": 3.69, + "learning_rate": 6.805394570546282e-06, + "loss": 0.8966, + "step": 49969 + }, + { + "epoch": 3.69, + "learning_rate": 6.805016602508596e-06, + "loss": 0.98, + "step": 49970 + }, + { + "epoch": 3.69, + "learning_rate": 6.804638639554077e-06, + "loss": 0.9936, + "step": 49971 + }, + { + "epoch": 3.69, + "learning_rate": 6.804260681683339e-06, + "loss": 1.0728, + "step": 49972 + }, + { + "epoch": 3.69, + "learning_rate": 6.803882728896971e-06, + "loss": 0.9745, + "step": 49973 + }, + { + "epoch": 3.69, + "learning_rate": 6.803504781195585e-06, + "loss": 0.9818, + "step": 49974 + }, + { + "epoch": 3.69, + "learning_rate": 6.8031268385797715e-06, + "loss": 1.0769, + "step": 49975 + }, + { + "epoch": 3.69, + "learning_rate": 6.802748901050142e-06, + "loss": 1.0766, + "step": 49976 + }, + { + "epoch": 3.69, + "learning_rate": 6.8023709686072905e-06, + "loss": 1.0479, + "step": 49977 + }, + { + "epoch": 3.69, + "learning_rate": 6.8019930412518245e-06, + "loss": 1.0041, + "step": 49978 + }, + { + "epoch": 3.69, + "learning_rate": 6.80161511898434e-06, + "loss": 1.0348, + "step": 49979 + }, + { + "epoch": 3.69, + "learning_rate": 6.8012372018054385e-06, + "loss": 1.0703, + "step": 49980 + }, + { + "epoch": 3.69, + "learning_rate": 6.800859289715725e-06, + "loss": 0.9336, + "step": 49981 + }, + { + "epoch": 3.69, + "learning_rate": 6.800481382715795e-06, + "loss": 1.0433, + "step": 49982 + }, + { + "epoch": 3.69, + "learning_rate": 6.800103480806257e-06, + "loss": 1.0435, + "step": 49983 + }, + { + "epoch": 3.69, + "learning_rate": 6.799725583987708e-06, + "loss": 1.0138, + "step": 49984 + }, + { + "epoch": 3.69, + "learning_rate": 6.799347692260749e-06, + "loss": 1.0028, + "step": 49985 + }, + { + "epoch": 3.69, + "learning_rate": 6.798969805625981e-06, + "loss": 1.0247, + "step": 49986 + }, + { + "epoch": 3.69, + "learning_rate": 6.798591924084008e-06, + "loss": 0.9338, + "step": 49987 + }, + { + "epoch": 3.69, + "learning_rate": 6.798214047635431e-06, + "loss": 0.9806, + "step": 49988 + }, + { + "epoch": 3.69, + "learning_rate": 6.797836176280845e-06, + "loss": 1.1028, + "step": 49989 + }, + { + "epoch": 3.69, + "learning_rate": 6.797458310020861e-06, + "loss": 1.039, + "step": 49990 + }, + { + "epoch": 3.69, + "learning_rate": 6.797080448856071e-06, + "loss": 0.8957, + "step": 49991 + }, + { + "epoch": 3.69, + "learning_rate": 6.796702592787085e-06, + "loss": 0.942, + "step": 49992 + }, + { + "epoch": 3.69, + "learning_rate": 6.796324741814493e-06, + "loss": 0.9973, + "step": 49993 + }, + { + "epoch": 3.69, + "learning_rate": 6.795946895938907e-06, + "loss": 0.9936, + "step": 49994 + }, + { + "epoch": 3.69, + "learning_rate": 6.79556905516092e-06, + "loss": 0.9092, + "step": 49995 + }, + { + "epoch": 3.69, + "learning_rate": 6.795191219481141e-06, + "loss": 1.0129, + "step": 49996 + }, + { + "epoch": 3.69, + "learning_rate": 6.794813388900165e-06, + "loss": 0.9624, + "step": 49997 + }, + { + "epoch": 3.69, + "learning_rate": 6.794435563418595e-06, + "loss": 1.0071, + "step": 49998 + }, + { + "epoch": 3.69, + "learning_rate": 6.794057743037033e-06, + "loss": 0.9748, + "step": 49999 + }, + { + "epoch": 3.69, + "learning_rate": 6.793679927756076e-06, + "loss": 0.9435, + "step": 50000 + }, + { + "epoch": 3.69, + "learning_rate": 6.793302117576333e-06, + "loss": 1.0104, + "step": 50001 + }, + { + "epoch": 3.69, + "learning_rate": 6.792924312498397e-06, + "loss": 0.9046, + "step": 50002 + }, + { + "epoch": 3.69, + "learning_rate": 6.7925465125228725e-06, + "loss": 0.9583, + "step": 50003 + }, + { + "epoch": 3.69, + "learning_rate": 6.792168717650361e-06, + "loss": 1.0092, + "step": 50004 + }, + { + "epoch": 3.69, + "learning_rate": 6.791790927881465e-06, + "loss": 1.033, + "step": 50005 + }, + { + "epoch": 3.69, + "learning_rate": 6.791413143216784e-06, + "loss": 1.0398, + "step": 50006 + }, + { + "epoch": 3.69, + "learning_rate": 6.791035363656916e-06, + "loss": 0.9901, + "step": 50007 + }, + { + "epoch": 3.7, + "learning_rate": 6.790657589202467e-06, + "loss": 1.0127, + "step": 50008 + }, + { + "epoch": 3.7, + "learning_rate": 6.790279819854034e-06, + "loss": 0.9741, + "step": 50009 + }, + { + "epoch": 3.7, + "learning_rate": 6.789902055612222e-06, + "loss": 1.085, + "step": 50010 + }, + { + "epoch": 3.7, + "learning_rate": 6.7895242964776275e-06, + "loss": 0.9874, + "step": 50011 + }, + { + "epoch": 3.7, + "learning_rate": 6.7891465424508566e-06, + "loss": 1.038, + "step": 50012 + }, + { + "epoch": 3.7, + "learning_rate": 6.788768793532505e-06, + "loss": 0.9485, + "step": 50013 + }, + { + "epoch": 3.7, + "learning_rate": 6.788391049723178e-06, + "loss": 0.9912, + "step": 50014 + }, + { + "epoch": 3.7, + "learning_rate": 6.788013311023478e-06, + "loss": 1.0237, + "step": 50015 + }, + { + "epoch": 3.7, + "learning_rate": 6.787635577433995e-06, + "loss": 1.0555, + "step": 50016 + }, + { + "epoch": 3.7, + "learning_rate": 6.787257848955346e-06, + "loss": 0.96, + "step": 50017 + }, + { + "epoch": 3.7, + "learning_rate": 6.786880125588117e-06, + "loss": 1.0697, + "step": 50018 + }, + { + "epoch": 3.7, + "learning_rate": 6.786502407332922e-06, + "loss": 1.0787, + "step": 50019 + }, + { + "epoch": 3.7, + "learning_rate": 6.786124694190348e-06, + "loss": 0.9175, + "step": 50020 + }, + { + "epoch": 3.7, + "learning_rate": 6.7857469861610126e-06, + "loss": 1.051, + "step": 50021 + }, + { + "epoch": 3.7, + "learning_rate": 6.7853692832455e-06, + "loss": 0.984, + "step": 50022 + }, + { + "epoch": 3.7, + "learning_rate": 6.784991585444425e-06, + "loss": 1.0872, + "step": 50023 + }, + { + "epoch": 3.7, + "learning_rate": 6.784613892758376e-06, + "loss": 0.9469, + "step": 50024 + }, + { + "epoch": 3.7, + "learning_rate": 6.784236205187966e-06, + "loss": 1.0807, + "step": 50025 + }, + { + "epoch": 3.7, + "learning_rate": 6.7838585227337885e-06, + "loss": 1.0064, + "step": 50026 + }, + { + "epoch": 3.7, + "learning_rate": 6.7834808453964455e-06, + "loss": 0.9776, + "step": 50027 + }, + { + "epoch": 3.7, + "learning_rate": 6.783103173176539e-06, + "loss": 1.0947, + "step": 50028 + }, + { + "epoch": 3.7, + "learning_rate": 6.782725506074668e-06, + "loss": 0.9063, + "step": 50029 + }, + { + "epoch": 3.7, + "learning_rate": 6.782347844091436e-06, + "loss": 0.9621, + "step": 50030 + }, + { + "epoch": 3.7, + "learning_rate": 6.78197018722744e-06, + "loss": 0.9493, + "step": 50031 + }, + { + "epoch": 3.7, + "learning_rate": 6.781592535483287e-06, + "loss": 1.0467, + "step": 50032 + }, + { + "epoch": 3.7, + "learning_rate": 6.781214888859571e-06, + "loss": 0.9614, + "step": 50033 + }, + { + "epoch": 3.7, + "learning_rate": 6.7808372473568995e-06, + "loss": 0.9824, + "step": 50034 + }, + { + "epoch": 3.7, + "learning_rate": 6.780459610975869e-06, + "loss": 1.1105, + "step": 50035 + }, + { + "epoch": 3.7, + "learning_rate": 6.780081979717079e-06, + "loss": 1.0465, + "step": 50036 + }, + { + "epoch": 3.7, + "learning_rate": 6.779704353581135e-06, + "loss": 1.0628, + "step": 50037 + }, + { + "epoch": 3.7, + "learning_rate": 6.7793267325686315e-06, + "loss": 0.9973, + "step": 50038 + }, + { + "epoch": 3.7, + "learning_rate": 6.7789491166801805e-06, + "loss": 1.0049, + "step": 50039 + }, + { + "epoch": 3.7, + "learning_rate": 6.778571505916367e-06, + "loss": 0.9686, + "step": 50040 + }, + { + "epoch": 3.7, + "learning_rate": 6.778193900277805e-06, + "loss": 1.0233, + "step": 50041 + }, + { + "epoch": 3.7, + "learning_rate": 6.777816299765086e-06, + "loss": 1.0095, + "step": 50042 + }, + { + "epoch": 3.7, + "learning_rate": 6.777438704378822e-06, + "loss": 1.0112, + "step": 50043 + }, + { + "epoch": 3.7, + "learning_rate": 6.7770611141196045e-06, + "loss": 0.9303, + "step": 50044 + }, + { + "epoch": 3.7, + "learning_rate": 6.776683528988034e-06, + "loss": 1.0726, + "step": 50045 + }, + { + "epoch": 3.7, + "learning_rate": 6.776305948984716e-06, + "loss": 1.0139, + "step": 50046 + }, + { + "epoch": 3.7, + "learning_rate": 6.775928374110248e-06, + "loss": 0.9354, + "step": 50047 + }, + { + "epoch": 3.7, + "learning_rate": 6.775550804365232e-06, + "loss": 1.1202, + "step": 50048 + }, + { + "epoch": 3.7, + "learning_rate": 6.775173239750268e-06, + "loss": 1.0271, + "step": 50049 + }, + { + "epoch": 3.7, + "learning_rate": 6.774795680265958e-06, + "loss": 1.0804, + "step": 50050 + }, + { + "epoch": 3.7, + "learning_rate": 6.7744181259129e-06, + "loss": 0.9222, + "step": 50051 + }, + { + "epoch": 3.7, + "learning_rate": 6.7740405766917e-06, + "loss": 1.0198, + "step": 50052 + }, + { + "epoch": 3.7, + "learning_rate": 6.773663032602955e-06, + "loss": 0.9793, + "step": 50053 + }, + { + "epoch": 3.7, + "learning_rate": 6.773285493647263e-06, + "loss": 1.0067, + "step": 50054 + }, + { + "epoch": 3.7, + "learning_rate": 6.77290795982523e-06, + "loss": 0.9766, + "step": 50055 + }, + { + "epoch": 3.7, + "learning_rate": 6.772530431137454e-06, + "loss": 0.9334, + "step": 50056 + }, + { + "epoch": 3.7, + "learning_rate": 6.7721529075845346e-06, + "loss": 1.0369, + "step": 50057 + }, + { + "epoch": 3.7, + "learning_rate": 6.771775389167074e-06, + "loss": 1.0491, + "step": 50058 + }, + { + "epoch": 3.7, + "learning_rate": 6.771397875885675e-06, + "loss": 0.9256, + "step": 50059 + }, + { + "epoch": 3.7, + "learning_rate": 6.771020367740933e-06, + "loss": 0.9729, + "step": 50060 + }, + { + "epoch": 3.7, + "learning_rate": 6.770642864733452e-06, + "loss": 1.0261, + "step": 50061 + }, + { + "epoch": 3.7, + "learning_rate": 6.7702653668638365e-06, + "loss": 1.0434, + "step": 50062 + }, + { + "epoch": 3.7, + "learning_rate": 6.769887874132674e-06, + "loss": 0.9014, + "step": 50063 + }, + { + "epoch": 3.7, + "learning_rate": 6.769510386540582e-06, + "loss": 1.0356, + "step": 50064 + }, + { + "epoch": 3.7, + "learning_rate": 6.769132904088145e-06, + "loss": 0.9327, + "step": 50065 + }, + { + "epoch": 3.7, + "learning_rate": 6.768755426775979e-06, + "loss": 0.9984, + "step": 50066 + }, + { + "epoch": 3.7, + "learning_rate": 6.76837795460467e-06, + "loss": 1.0693, + "step": 50067 + }, + { + "epoch": 3.7, + "learning_rate": 6.7680004875748304e-06, + "loss": 0.9752, + "step": 50068 + }, + { + "epoch": 3.7, + "learning_rate": 6.767623025687053e-06, + "loss": 0.9775, + "step": 50069 + }, + { + "epoch": 3.7, + "learning_rate": 6.767245568941943e-06, + "loss": 0.974, + "step": 50070 + }, + { + "epoch": 3.7, + "learning_rate": 6.7668681173401e-06, + "loss": 0.9504, + "step": 50071 + }, + { + "epoch": 3.7, + "learning_rate": 6.76649067088212e-06, + "loss": 0.9927, + "step": 50072 + }, + { + "epoch": 3.7, + "learning_rate": 6.76611322956861e-06, + "loss": 1.0255, + "step": 50073 + }, + { + "epoch": 3.7, + "learning_rate": 6.7657357934001655e-06, + "loss": 1.0465, + "step": 50074 + }, + { + "epoch": 3.7, + "learning_rate": 6.76535836237739e-06, + "loss": 0.8891, + "step": 50075 + }, + { + "epoch": 3.7, + "learning_rate": 6.764980936500883e-06, + "loss": 1.0378, + "step": 50076 + }, + { + "epoch": 3.7, + "learning_rate": 6.764603515771244e-06, + "loss": 0.9397, + "step": 50077 + }, + { + "epoch": 3.7, + "learning_rate": 6.764226100189075e-06, + "loss": 1.0566, + "step": 50078 + }, + { + "epoch": 3.7, + "learning_rate": 6.763848689754977e-06, + "loss": 1.0286, + "step": 50079 + }, + { + "epoch": 3.7, + "learning_rate": 6.763471284469549e-06, + "loss": 1.0104, + "step": 50080 + }, + { + "epoch": 3.7, + "learning_rate": 6.763093884333391e-06, + "loss": 1.0116, + "step": 50081 + }, + { + "epoch": 3.7, + "learning_rate": 6.7627164893471095e-06, + "loss": 1.0278, + "step": 50082 + }, + { + "epoch": 3.7, + "learning_rate": 6.762339099511291e-06, + "loss": 1.0606, + "step": 50083 + }, + { + "epoch": 3.7, + "learning_rate": 6.761961714826553e-06, + "loss": 0.9339, + "step": 50084 + }, + { + "epoch": 3.7, + "learning_rate": 6.761584335293481e-06, + "loss": 0.9842, + "step": 50085 + }, + { + "epoch": 3.7, + "learning_rate": 6.761206960912688e-06, + "loss": 1.0482, + "step": 50086 + }, + { + "epoch": 3.7, + "learning_rate": 6.760829591684761e-06, + "loss": 1.0073, + "step": 50087 + }, + { + "epoch": 3.7, + "learning_rate": 6.760452227610315e-06, + "loss": 1.0346, + "step": 50088 + }, + { + "epoch": 3.7, + "learning_rate": 6.760074868689936e-06, + "loss": 0.8754, + "step": 50089 + }, + { + "epoch": 3.7, + "learning_rate": 6.759697514924238e-06, + "loss": 1.1138, + "step": 50090 + }, + { + "epoch": 3.7, + "learning_rate": 6.759320166313814e-06, + "loss": 0.9664, + "step": 50091 + }, + { + "epoch": 3.7, + "learning_rate": 6.758942822859263e-06, + "loss": 1.0538, + "step": 50092 + }, + { + "epoch": 3.7, + "learning_rate": 6.758565484561188e-06, + "loss": 1.006, + "step": 50093 + }, + { + "epoch": 3.7, + "learning_rate": 6.758188151420188e-06, + "loss": 1.1425, + "step": 50094 + }, + { + "epoch": 3.7, + "learning_rate": 6.7578108234368664e-06, + "loss": 0.9478, + "step": 50095 + }, + { + "epoch": 3.7, + "learning_rate": 6.7574335006118185e-06, + "loss": 0.9508, + "step": 50096 + }, + { + "epoch": 3.7, + "learning_rate": 6.757056182945652e-06, + "loss": 1.0175, + "step": 50097 + }, + { + "epoch": 3.7, + "learning_rate": 6.756678870438958e-06, + "loss": 1.028, + "step": 50098 + }, + { + "epoch": 3.7, + "learning_rate": 6.756301563092344e-06, + "loss": 1.128, + "step": 50099 + }, + { + "epoch": 3.7, + "learning_rate": 6.7559242609064094e-06, + "loss": 1.0727, + "step": 50100 + }, + { + "epoch": 3.7, + "learning_rate": 6.75554696388175e-06, + "loss": 0.9835, + "step": 50101 + }, + { + "epoch": 3.7, + "learning_rate": 6.75516967201897e-06, + "loss": 1.0436, + "step": 50102 + }, + { + "epoch": 3.7, + "learning_rate": 6.754792385318667e-06, + "loss": 0.9681, + "step": 50103 + }, + { + "epoch": 3.7, + "learning_rate": 6.754415103781445e-06, + "loss": 1.0021, + "step": 50104 + }, + { + "epoch": 3.7, + "learning_rate": 6.754037827407901e-06, + "loss": 0.9409, + "step": 50105 + }, + { + "epoch": 3.7, + "learning_rate": 6.7536605561986376e-06, + "loss": 0.9061, + "step": 50106 + }, + { + "epoch": 3.7, + "learning_rate": 6.753283290154252e-06, + "loss": 0.9443, + "step": 50107 + }, + { + "epoch": 3.7, + "learning_rate": 6.752906029275348e-06, + "loss": 1.0086, + "step": 50108 + }, + { + "epoch": 3.7, + "learning_rate": 6.752528773562527e-06, + "loss": 1.0128, + "step": 50109 + }, + { + "epoch": 3.7, + "learning_rate": 6.752151523016379e-06, + "loss": 0.9183, + "step": 50110 + }, + { + "epoch": 3.7, + "learning_rate": 6.7517742776375186e-06, + "loss": 1.0385, + "step": 50111 + }, + { + "epoch": 3.7, + "learning_rate": 6.751397037426532e-06, + "loss": 1.0556, + "step": 50112 + }, + { + "epoch": 3.7, + "learning_rate": 6.751019802384032e-06, + "loss": 0.9741, + "step": 50113 + }, + { + "epoch": 3.7, + "learning_rate": 6.750642572510609e-06, + "loss": 1.0434, + "step": 50114 + }, + { + "epoch": 3.7, + "learning_rate": 6.75026534780687e-06, + "loss": 0.9566, + "step": 50115 + }, + { + "epoch": 3.7, + "learning_rate": 6.74988812827341e-06, + "loss": 1.0098, + "step": 50116 + }, + { + "epoch": 3.7, + "learning_rate": 6.749510913910834e-06, + "loss": 0.9601, + "step": 50117 + }, + { + "epoch": 3.7, + "learning_rate": 6.749133704719739e-06, + "loss": 1.0157, + "step": 50118 + }, + { + "epoch": 3.7, + "learning_rate": 6.748756500700724e-06, + "loss": 0.9146, + "step": 50119 + }, + { + "epoch": 3.7, + "learning_rate": 6.748379301854393e-06, + "loss": 0.8839, + "step": 50120 + }, + { + "epoch": 3.7, + "learning_rate": 6.748002108181342e-06, + "loss": 0.9221, + "step": 50121 + }, + { + "epoch": 3.7, + "learning_rate": 6.747624919682174e-06, + "loss": 0.9969, + "step": 50122 + }, + { + "epoch": 3.7, + "learning_rate": 6.747247736357488e-06, + "loss": 1.0392, + "step": 50123 + }, + { + "epoch": 3.7, + "learning_rate": 6.7468705582078855e-06, + "loss": 0.983, + "step": 50124 + }, + { + "epoch": 3.7, + "learning_rate": 6.746493385233963e-06, + "loss": 0.9664, + "step": 50125 + }, + { + "epoch": 3.7, + "learning_rate": 6.746116217436325e-06, + "loss": 0.9892, + "step": 50126 + }, + { + "epoch": 3.7, + "learning_rate": 6.7457390548155725e-06, + "loss": 0.9818, + "step": 50127 + }, + { + "epoch": 3.7, + "learning_rate": 6.7453618973722955e-06, + "loss": 1.061, + "step": 50128 + }, + { + "epoch": 3.7, + "learning_rate": 6.744984745107107e-06, + "loss": 1.1186, + "step": 50129 + }, + { + "epoch": 3.7, + "learning_rate": 6.744607598020596e-06, + "loss": 0.9499, + "step": 50130 + }, + { + "epoch": 3.7, + "learning_rate": 6.744230456113375e-06, + "loss": 0.8995, + "step": 50131 + }, + { + "epoch": 3.7, + "learning_rate": 6.743853319386028e-06, + "loss": 1.0762, + "step": 50132 + }, + { + "epoch": 3.7, + "learning_rate": 6.743476187839171e-06, + "loss": 1.0186, + "step": 50133 + }, + { + "epoch": 3.7, + "learning_rate": 6.74309906147339e-06, + "loss": 1.0763, + "step": 50134 + }, + { + "epoch": 3.7, + "learning_rate": 6.742721940289298e-06, + "loss": 1.0121, + "step": 50135 + }, + { + "epoch": 3.7, + "learning_rate": 6.7423448242874844e-06, + "loss": 1.0149, + "step": 50136 + }, + { + "epoch": 3.7, + "learning_rate": 6.741967713468559e-06, + "loss": 1.0814, + "step": 50137 + }, + { + "epoch": 3.7, + "learning_rate": 6.741590607833114e-06, + "loss": 1.0119, + "step": 50138 + }, + { + "epoch": 3.7, + "learning_rate": 6.74121350738175e-06, + "loss": 0.88, + "step": 50139 + }, + { + "epoch": 3.7, + "learning_rate": 6.7408364121150705e-06, + "loss": 0.9725, + "step": 50140 + }, + { + "epoch": 3.7, + "learning_rate": 6.74045932203367e-06, + "loss": 1.0574, + "step": 50141 + }, + { + "epoch": 3.7, + "learning_rate": 6.740082237138157e-06, + "loss": 1.0504, + "step": 50142 + }, + { + "epoch": 3.71, + "learning_rate": 6.739705157429121e-06, + "loss": 1.0019, + "step": 50143 + }, + { + "epoch": 3.71, + "learning_rate": 6.739328082907173e-06, + "loss": 0.9666, + "step": 50144 + }, + { + "epoch": 3.71, + "learning_rate": 6.7389510135729045e-06, + "loss": 1.0544, + "step": 50145 + }, + { + "epoch": 3.71, + "learning_rate": 6.738573949426919e-06, + "loss": 0.9472, + "step": 50146 + }, + { + "epoch": 3.71, + "learning_rate": 6.738196890469817e-06, + "loss": 0.983, + "step": 50147 + }, + { + "epoch": 3.71, + "learning_rate": 6.737819836702194e-06, + "loss": 1.0438, + "step": 50148 + }, + { + "epoch": 3.71, + "learning_rate": 6.737442788124655e-06, + "loss": 0.9623, + "step": 50149 + }, + { + "epoch": 3.71, + "learning_rate": 6.737065744737797e-06, + "loss": 1.0577, + "step": 50150 + }, + { + "epoch": 3.71, + "learning_rate": 6.736688706542222e-06, + "loss": 0.919, + "step": 50151 + }, + { + "epoch": 3.71, + "learning_rate": 6.736311673538527e-06, + "loss": 0.977, + "step": 50152 + }, + { + "epoch": 3.71, + "learning_rate": 6.7359346457273145e-06, + "loss": 1.0056, + "step": 50153 + }, + { + "epoch": 3.71, + "learning_rate": 6.735557623109181e-06, + "loss": 0.9445, + "step": 50154 + }, + { + "epoch": 3.71, + "learning_rate": 6.735180605684732e-06, + "loss": 0.9898, + "step": 50155 + }, + { + "epoch": 3.71, + "learning_rate": 6.734803593454565e-06, + "loss": 1.0642, + "step": 50156 + }, + { + "epoch": 3.71, + "learning_rate": 6.734426586419276e-06, + "loss": 1.0173, + "step": 50157 + }, + { + "epoch": 3.71, + "learning_rate": 6.734049584579468e-06, + "loss": 1.0102, + "step": 50158 + }, + { + "epoch": 3.71, + "learning_rate": 6.733672587935739e-06, + "loss": 0.9281, + "step": 50159 + }, + { + "epoch": 3.71, + "learning_rate": 6.733295596488692e-06, + "loss": 0.9836, + "step": 50160 + }, + { + "epoch": 3.71, + "learning_rate": 6.732918610238923e-06, + "loss": 1.0774, + "step": 50161 + }, + { + "epoch": 3.71, + "learning_rate": 6.732541629187037e-06, + "loss": 1.0762, + "step": 50162 + }, + { + "epoch": 3.71, + "learning_rate": 6.732164653333627e-06, + "loss": 1.0168, + "step": 50163 + }, + { + "epoch": 3.71, + "learning_rate": 6.731787682679298e-06, + "loss": 1.044, + "step": 50164 + }, + { + "epoch": 3.71, + "learning_rate": 6.731410717224649e-06, + "loss": 1.0137, + "step": 50165 + }, + { + "epoch": 3.71, + "learning_rate": 6.731033756970278e-06, + "loss": 0.9142, + "step": 50166 + }, + { + "epoch": 3.71, + "learning_rate": 6.730656801916785e-06, + "loss": 0.9147, + "step": 50167 + }, + { + "epoch": 3.71, + "learning_rate": 6.730279852064769e-06, + "loss": 0.9995, + "step": 50168 + }, + { + "epoch": 3.71, + "learning_rate": 6.729902907414834e-06, + "loss": 0.9973, + "step": 50169 + }, + { + "epoch": 3.71, + "learning_rate": 6.729525967967574e-06, + "loss": 1.0614, + "step": 50170 + }, + { + "epoch": 3.71, + "learning_rate": 6.729149033723593e-06, + "loss": 0.9899, + "step": 50171 + }, + { + "epoch": 3.71, + "learning_rate": 6.728772104683487e-06, + "loss": 1.1065, + "step": 50172 + }, + { + "epoch": 3.71, + "learning_rate": 6.728395180847861e-06, + "loss": 0.9921, + "step": 50173 + }, + { + "epoch": 3.71, + "learning_rate": 6.728018262217313e-06, + "loss": 0.9215, + "step": 50174 + }, + { + "epoch": 3.71, + "learning_rate": 6.727641348792435e-06, + "loss": 0.9973, + "step": 50175 + }, + { + "epoch": 3.71, + "learning_rate": 6.727264440573838e-06, + "loss": 0.9335, + "step": 50176 + }, + { + "epoch": 3.71, + "learning_rate": 6.726887537562111e-06, + "loss": 0.9645, + "step": 50177 + }, + { + "epoch": 3.71, + "learning_rate": 6.726510639757866e-06, + "loss": 0.9422, + "step": 50178 + }, + { + "epoch": 3.71, + "learning_rate": 6.726133747161688e-06, + "loss": 1.0147, + "step": 50179 + }, + { + "epoch": 3.71, + "learning_rate": 6.725756859774192e-06, + "loss": 1.0493, + "step": 50180 + }, + { + "epoch": 3.71, + "learning_rate": 6.725379977595963e-06, + "loss": 1.0232, + "step": 50181 + }, + { + "epoch": 3.71, + "learning_rate": 6.725003100627614e-06, + "loss": 0.9271, + "step": 50182 + }, + { + "epoch": 3.71, + "learning_rate": 6.7246262288697315e-06, + "loss": 1.0785, + "step": 50183 + }, + { + "epoch": 3.71, + "learning_rate": 6.724249362322929e-06, + "loss": 0.9843, + "step": 50184 + }, + { + "epoch": 3.71, + "learning_rate": 6.723872500987796e-06, + "loss": 1.0209, + "step": 50185 + }, + { + "epoch": 3.71, + "learning_rate": 6.723495644864933e-06, + "loss": 1.0908, + "step": 50186 + }, + { + "epoch": 3.71, + "learning_rate": 6.723118793954943e-06, + "loss": 0.9848, + "step": 50187 + }, + { + "epoch": 3.71, + "learning_rate": 6.722741948258425e-06, + "loss": 1.0307, + "step": 50188 + }, + { + "epoch": 3.71, + "learning_rate": 6.722365107775977e-06, + "loss": 1.0141, + "step": 50189 + }, + { + "epoch": 3.71, + "learning_rate": 6.721988272508198e-06, + "loss": 0.9718, + "step": 50190 + }, + { + "epoch": 3.71, + "learning_rate": 6.721611442455691e-06, + "loss": 1.0393, + "step": 50191 + }, + { + "epoch": 3.71, + "learning_rate": 6.7212346176190504e-06, + "loss": 0.9377, + "step": 50192 + }, + { + "epoch": 3.71, + "learning_rate": 6.720857797998881e-06, + "loss": 0.9692, + "step": 50193 + }, + { + "epoch": 3.71, + "learning_rate": 6.7204809835957805e-06, + "loss": 0.8979, + "step": 50194 + }, + { + "epoch": 3.71, + "learning_rate": 6.720104174410345e-06, + "loss": 0.9958, + "step": 50195 + }, + { + "epoch": 3.71, + "learning_rate": 6.719727370443179e-06, + "loss": 1.0644, + "step": 50196 + }, + { + "epoch": 3.71, + "learning_rate": 6.719350571694878e-06, + "loss": 1.1645, + "step": 50197 + }, + { + "epoch": 3.71, + "learning_rate": 6.718973778166045e-06, + "loss": 0.975, + "step": 50198 + }, + { + "epoch": 3.71, + "learning_rate": 6.718596989857276e-06, + "loss": 1.0675, + "step": 50199 + }, + { + "epoch": 3.71, + "learning_rate": 6.718220206769175e-06, + "loss": 1.0731, + "step": 50200 + }, + { + "epoch": 3.71, + "learning_rate": 6.717843428902336e-06, + "loss": 1.0187, + "step": 50201 + }, + { + "epoch": 3.71, + "learning_rate": 6.717466656257366e-06, + "loss": 0.9484, + "step": 50202 + }, + { + "epoch": 3.71, + "learning_rate": 6.717089888834855e-06, + "loss": 1.0712, + "step": 50203 + }, + { + "epoch": 3.71, + "learning_rate": 6.716713126635408e-06, + "loss": 0.8928, + "step": 50204 + }, + { + "epoch": 3.71, + "learning_rate": 6.716336369659624e-06, + "loss": 0.988, + "step": 50205 + }, + { + "epoch": 3.71, + "learning_rate": 6.715959617908099e-06, + "loss": 0.9703, + "step": 50206 + }, + { + "epoch": 3.71, + "learning_rate": 6.715582871381438e-06, + "loss": 1.0012, + "step": 50207 + }, + { + "epoch": 3.71, + "learning_rate": 6.7152061300802366e-06, + "loss": 0.9449, + "step": 50208 + }, + { + "epoch": 3.71, + "learning_rate": 6.714829394005097e-06, + "loss": 0.9637, + "step": 50209 + }, + { + "epoch": 3.71, + "learning_rate": 6.7144526631566144e-06, + "loss": 0.9561, + "step": 50210 + }, + { + "epoch": 3.71, + "learning_rate": 6.714075937535393e-06, + "loss": 0.9946, + "step": 50211 + }, + { + "epoch": 3.71, + "learning_rate": 6.713699217142029e-06, + "loss": 0.9796, + "step": 50212 + }, + { + "epoch": 3.71, + "learning_rate": 6.713322501977121e-06, + "loss": 0.9672, + "step": 50213 + }, + { + "epoch": 3.71, + "learning_rate": 6.712945792041272e-06, + "loss": 0.9573, + "step": 50214 + }, + { + "epoch": 3.71, + "learning_rate": 6.712569087335077e-06, + "loss": 0.9543, + "step": 50215 + }, + { + "epoch": 3.71, + "learning_rate": 6.71219238785914e-06, + "loss": 1.1826, + "step": 50216 + }, + { + "epoch": 3.71, + "learning_rate": 6.711815693614055e-06, + "loss": 1.0107, + "step": 50217 + }, + { + "epoch": 3.71, + "learning_rate": 6.711439004600426e-06, + "loss": 0.9895, + "step": 50218 + }, + { + "epoch": 3.71, + "learning_rate": 6.711062320818849e-06, + "loss": 1.0923, + "step": 50219 + }, + { + "epoch": 3.71, + "learning_rate": 6.7106856422699276e-06, + "loss": 0.995, + "step": 50220 + }, + { + "epoch": 3.71, + "learning_rate": 6.71030896895426e-06, + "loss": 1.083, + "step": 50221 + }, + { + "epoch": 3.71, + "learning_rate": 6.709932300872437e-06, + "loss": 0.9044, + "step": 50222 + }, + { + "epoch": 3.71, + "learning_rate": 6.709555638025072e-06, + "loss": 1.1043, + "step": 50223 + }, + { + "epoch": 3.71, + "learning_rate": 6.7091789804127495e-06, + "loss": 1.1247, + "step": 50224 + }, + { + "epoch": 3.71, + "learning_rate": 6.708802328036081e-06, + "loss": 0.8908, + "step": 50225 + }, + { + "epoch": 3.71, + "learning_rate": 6.708425680895657e-06, + "loss": 0.9195, + "step": 50226 + }, + { + "epoch": 3.71, + "learning_rate": 6.708049038992086e-06, + "loss": 1.0108, + "step": 50227 + }, + { + "epoch": 3.71, + "learning_rate": 6.707672402325957e-06, + "loss": 0.8933, + "step": 50228 + }, + { + "epoch": 3.71, + "learning_rate": 6.707295770897879e-06, + "loss": 0.9866, + "step": 50229 + }, + { + "epoch": 3.71, + "learning_rate": 6.706919144708443e-06, + "loss": 0.98, + "step": 50230 + }, + { + "epoch": 3.71, + "learning_rate": 6.706542523758252e-06, + "loss": 0.9082, + "step": 50231 + }, + { + "epoch": 3.71, + "learning_rate": 6.7061659080479055e-06, + "loss": 0.9594, + "step": 50232 + }, + { + "epoch": 3.71, + "learning_rate": 6.705789297578001e-06, + "loss": 1.0527, + "step": 50233 + }, + { + "epoch": 3.71, + "learning_rate": 6.7054126923491394e-06, + "loss": 0.9963, + "step": 50234 + }, + { + "epoch": 3.71, + "learning_rate": 6.705036092361916e-06, + "loss": 1.0229, + "step": 50235 + }, + { + "epoch": 3.71, + "learning_rate": 6.704659497616937e-06, + "loss": 1.0368, + "step": 50236 + }, + { + "epoch": 3.71, + "learning_rate": 6.704282908114796e-06, + "loss": 0.9599, + "step": 50237 + }, + { + "epoch": 3.71, + "learning_rate": 6.703906323856094e-06, + "loss": 1.0806, + "step": 50238 + }, + { + "epoch": 3.71, + "learning_rate": 6.703529744841428e-06, + "loss": 1.0058, + "step": 50239 + }, + { + "epoch": 3.71, + "learning_rate": 6.703153171071402e-06, + "loss": 0.9973, + "step": 50240 + }, + { + "epoch": 3.71, + "learning_rate": 6.7027766025466124e-06, + "loss": 0.9629, + "step": 50241 + }, + { + "epoch": 3.71, + "learning_rate": 6.702400039267655e-06, + "loss": 0.9601, + "step": 50242 + }, + { + "epoch": 3.71, + "learning_rate": 6.702023481235133e-06, + "loss": 0.9651, + "step": 50243 + }, + { + "epoch": 3.71, + "learning_rate": 6.701646928449645e-06, + "loss": 1.126, + "step": 50244 + }, + { + "epoch": 3.71, + "learning_rate": 6.7012703809117905e-06, + "loss": 0.9387, + "step": 50245 + }, + { + "epoch": 3.71, + "learning_rate": 6.7008938386221656e-06, + "loss": 0.9707, + "step": 50246 + }, + { + "epoch": 3.71, + "learning_rate": 6.700517301581375e-06, + "loss": 1.1286, + "step": 50247 + }, + { + "epoch": 3.71, + "learning_rate": 6.7001407697900086e-06, + "loss": 1.0095, + "step": 50248 + }, + { + "epoch": 3.71, + "learning_rate": 6.699764243248677e-06, + "loss": 0.9082, + "step": 50249 + }, + { + "epoch": 3.71, + "learning_rate": 6.699387721957971e-06, + "loss": 1.0107, + "step": 50250 + }, + { + "epoch": 3.71, + "learning_rate": 6.69901120591849e-06, + "loss": 1.1437, + "step": 50251 + }, + { + "epoch": 3.71, + "learning_rate": 6.698634695130837e-06, + "loss": 1.0578, + "step": 50252 + }, + { + "epoch": 3.71, + "learning_rate": 6.698258189595607e-06, + "loss": 0.9045, + "step": 50253 + }, + { + "epoch": 3.71, + "learning_rate": 6.697881689313404e-06, + "loss": 0.9828, + "step": 50254 + }, + { + "epoch": 3.71, + "learning_rate": 6.697505194284819e-06, + "loss": 0.9233, + "step": 50255 + }, + { + "epoch": 3.71, + "learning_rate": 6.697128704510461e-06, + "loss": 1.0601, + "step": 50256 + }, + { + "epoch": 3.71, + "learning_rate": 6.6967522199909205e-06, + "loss": 1.0296, + "step": 50257 + }, + { + "epoch": 3.71, + "learning_rate": 6.696375740726801e-06, + "loss": 0.9772, + "step": 50258 + }, + { + "epoch": 3.71, + "learning_rate": 6.695999266718702e-06, + "loss": 1.0279, + "step": 50259 + }, + { + "epoch": 3.71, + "learning_rate": 6.69562279796722e-06, + "loss": 1.0476, + "step": 50260 + }, + { + "epoch": 3.71, + "learning_rate": 6.695246334472956e-06, + "loss": 0.9233, + "step": 50261 + }, + { + "epoch": 3.71, + "learning_rate": 6.694869876236505e-06, + "loss": 1.002, + "step": 50262 + }, + { + "epoch": 3.71, + "learning_rate": 6.69449342325847e-06, + "loss": 0.9412, + "step": 50263 + }, + { + "epoch": 3.71, + "learning_rate": 6.6941169755394484e-06, + "loss": 1.0204, + "step": 50264 + }, + { + "epoch": 3.71, + "learning_rate": 6.69374053308004e-06, + "loss": 0.9906, + "step": 50265 + }, + { + "epoch": 3.71, + "learning_rate": 6.693364095880842e-06, + "loss": 1.0638, + "step": 50266 + }, + { + "epoch": 3.71, + "learning_rate": 6.692987663942456e-06, + "loss": 0.9733, + "step": 50267 + }, + { + "epoch": 3.71, + "learning_rate": 6.692611237265483e-06, + "loss": 0.9144, + "step": 50268 + }, + { + "epoch": 3.71, + "learning_rate": 6.692234815850509e-06, + "loss": 0.9744, + "step": 50269 + }, + { + "epoch": 3.71, + "learning_rate": 6.69185839969815e-06, + "loss": 1.0347, + "step": 50270 + }, + { + "epoch": 3.71, + "learning_rate": 6.691481988808992e-06, + "loss": 0.8805, + "step": 50271 + }, + { + "epoch": 3.71, + "learning_rate": 6.691105583183643e-06, + "loss": 1.0326, + "step": 50272 + }, + { + "epoch": 3.71, + "learning_rate": 6.690729182822693e-06, + "loss": 0.9042, + "step": 50273 + }, + { + "epoch": 3.71, + "learning_rate": 6.69035278772675e-06, + "loss": 1.0226, + "step": 50274 + }, + { + "epoch": 3.71, + "learning_rate": 6.689976397896402e-06, + "loss": 1.137, + "step": 50275 + }, + { + "epoch": 3.71, + "learning_rate": 6.6896000133322625e-06, + "loss": 1.0543, + "step": 50276 + }, + { + "epoch": 3.71, + "learning_rate": 6.689223634034919e-06, + "loss": 1.0015, + "step": 50277 + }, + { + "epoch": 3.72, + "learning_rate": 6.688847260004971e-06, + "loss": 1.0119, + "step": 50278 + }, + { + "epoch": 3.72, + "learning_rate": 6.688470891243022e-06, + "loss": 0.9937, + "step": 50279 + }, + { + "epoch": 3.72, + "learning_rate": 6.688094527749667e-06, + "loss": 0.9514, + "step": 50280 + }, + { + "epoch": 3.72, + "learning_rate": 6.6877181695255065e-06, + "loss": 0.9491, + "step": 50281 + }, + { + "epoch": 3.72, + "learning_rate": 6.687341816571138e-06, + "loss": 0.969, + "step": 50282 + }, + { + "epoch": 3.72, + "learning_rate": 6.686965468887163e-06, + "loss": 1.0084, + "step": 50283 + }, + { + "epoch": 3.72, + "learning_rate": 6.686589126474177e-06, + "loss": 1.0613, + "step": 50284 + }, + { + "epoch": 3.72, + "learning_rate": 6.6862127893327825e-06, + "loss": 0.9169, + "step": 50285 + }, + { + "epoch": 3.72, + "learning_rate": 6.685836457463575e-06, + "loss": 1.0306, + "step": 50286 + }, + { + "epoch": 3.72, + "learning_rate": 6.685460130867154e-06, + "loss": 0.9277, + "step": 50287 + }, + { + "epoch": 3.72, + "learning_rate": 6.685083809544119e-06, + "loss": 1.0134, + "step": 50288 + }, + { + "epoch": 3.72, + "learning_rate": 6.684707493495066e-06, + "loss": 1.0119, + "step": 50289 + }, + { + "epoch": 3.72, + "learning_rate": 6.6843311827206e-06, + "loss": 1.0193, + "step": 50290 + }, + { + "epoch": 3.72, + "learning_rate": 6.683954877221311e-06, + "loss": 1.003, + "step": 50291 + }, + { + "epoch": 3.72, + "learning_rate": 6.683578576997808e-06, + "loss": 0.8861, + "step": 50292 + }, + { + "epoch": 3.72, + "learning_rate": 6.683202282050677e-06, + "loss": 1.0801, + "step": 50293 + }, + { + "epoch": 3.72, + "learning_rate": 6.6828259923805304e-06, + "loss": 1.0342, + "step": 50294 + }, + { + "epoch": 3.72, + "learning_rate": 6.682449707987954e-06, + "loss": 1.0184, + "step": 50295 + }, + { + "epoch": 3.72, + "learning_rate": 6.682073428873559e-06, + "loss": 0.9246, + "step": 50296 + }, + { + "epoch": 3.72, + "learning_rate": 6.681697155037936e-06, + "loss": 0.9775, + "step": 50297 + }, + { + "epoch": 3.72, + "learning_rate": 6.681320886481683e-06, + "loss": 1.1728, + "step": 50298 + }, + { + "epoch": 3.72, + "learning_rate": 6.680944623205402e-06, + "loss": 0.9595, + "step": 50299 + }, + { + "epoch": 3.72, + "learning_rate": 6.68056836520969e-06, + "loss": 0.8963, + "step": 50300 + }, + { + "epoch": 3.72, + "learning_rate": 6.680192112495148e-06, + "loss": 1.022, + "step": 50301 + }, + { + "epoch": 3.72, + "learning_rate": 6.679815865062369e-06, + "loss": 1.0825, + "step": 50302 + }, + { + "epoch": 3.72, + "learning_rate": 6.67943962291196e-06, + "loss": 0.8268, + "step": 50303 + }, + { + "epoch": 3.72, + "learning_rate": 6.679063386044511e-06, + "loss": 0.9996, + "step": 50304 + }, + { + "epoch": 3.72, + "learning_rate": 6.678687154460628e-06, + "loss": 1.0813, + "step": 50305 + }, + { + "epoch": 3.72, + "learning_rate": 6.678310928160906e-06, + "loss": 1.0455, + "step": 50306 + }, + { + "epoch": 3.72, + "learning_rate": 6.677934707145941e-06, + "loss": 0.967, + "step": 50307 + }, + { + "epoch": 3.72, + "learning_rate": 6.677558491416338e-06, + "loss": 1.0096, + "step": 50308 + }, + { + "epoch": 3.72, + "learning_rate": 6.67718228097269e-06, + "loss": 1.0967, + "step": 50309 + }, + { + "epoch": 3.72, + "learning_rate": 6.6768060758155965e-06, + "loss": 1.0329, + "step": 50310 + }, + { + "epoch": 3.72, + "learning_rate": 6.676429875945658e-06, + "loss": 1.0335, + "step": 50311 + }, + { + "epoch": 3.72, + "learning_rate": 6.676053681363472e-06, + "loss": 1.0577, + "step": 50312 + }, + { + "epoch": 3.72, + "learning_rate": 6.675677492069637e-06, + "loss": 1.0502, + "step": 50313 + }, + { + "epoch": 3.72, + "learning_rate": 6.675301308064752e-06, + "loss": 1.0345, + "step": 50314 + }, + { + "epoch": 3.72, + "learning_rate": 6.674925129349417e-06, + "loss": 0.9659, + "step": 50315 + }, + { + "epoch": 3.72, + "learning_rate": 6.674548955924223e-06, + "loss": 0.9306, + "step": 50316 + }, + { + "epoch": 3.72, + "learning_rate": 6.67417278778978e-06, + "loss": 0.9163, + "step": 50317 + }, + { + "epoch": 3.72, + "learning_rate": 6.6737966249466745e-06, + "loss": 0.9898, + "step": 50318 + }, + { + "epoch": 3.72, + "learning_rate": 6.673420467395517e-06, + "loss": 0.9591, + "step": 50319 + }, + { + "epoch": 3.72, + "learning_rate": 6.673044315136894e-06, + "loss": 0.9589, + "step": 50320 + }, + { + "epoch": 3.72, + "learning_rate": 6.672668168171417e-06, + "loss": 0.9766, + "step": 50321 + }, + { + "epoch": 3.72, + "learning_rate": 6.672292026499669e-06, + "loss": 1.1328, + "step": 50322 + }, + { + "epoch": 3.72, + "learning_rate": 6.671915890122264e-06, + "loss": 1.0416, + "step": 50323 + }, + { + "epoch": 3.72, + "learning_rate": 6.671539759039791e-06, + "loss": 0.9943, + "step": 50324 + }, + { + "epoch": 3.72, + "learning_rate": 6.671163633252848e-06, + "loss": 1.079, + "step": 50325 + }, + { + "epoch": 3.72, + "learning_rate": 6.67078751276204e-06, + "loss": 1.0369, + "step": 50326 + }, + { + "epoch": 3.72, + "learning_rate": 6.670411397567959e-06, + "loss": 0.9762, + "step": 50327 + }, + { + "epoch": 3.72, + "learning_rate": 6.670035287671206e-06, + "loss": 1.1088, + "step": 50328 + }, + { + "epoch": 3.72, + "learning_rate": 6.669659183072379e-06, + "loss": 1.0616, + "step": 50329 + }, + { + "epoch": 3.72, + "learning_rate": 6.669283083772078e-06, + "loss": 0.9579, + "step": 50330 + }, + { + "epoch": 3.72, + "learning_rate": 6.668906989770898e-06, + "loss": 1.0163, + "step": 50331 + }, + { + "epoch": 3.72, + "learning_rate": 6.668530901069441e-06, + "loss": 0.9473, + "step": 50332 + }, + { + "epoch": 3.72, + "learning_rate": 6.668154817668305e-06, + "loss": 1.0184, + "step": 50333 + }, + { + "epoch": 3.72, + "learning_rate": 6.667778739568083e-06, + "loss": 1.0668, + "step": 50334 + }, + { + "epoch": 3.72, + "learning_rate": 6.667402666769383e-06, + "loss": 1.0339, + "step": 50335 + }, + { + "epoch": 3.72, + "learning_rate": 6.66702659927279e-06, + "loss": 0.9835, + "step": 50336 + }, + { + "epoch": 3.72, + "learning_rate": 6.666650537078916e-06, + "loss": 1.0477, + "step": 50337 + }, + { + "epoch": 3.72, + "learning_rate": 6.6662744801883486e-06, + "loss": 1.0175, + "step": 50338 + }, + { + "epoch": 3.72, + "learning_rate": 6.665898428601696e-06, + "loss": 0.9403, + "step": 50339 + }, + { + "epoch": 3.72, + "learning_rate": 6.665522382319546e-06, + "loss": 0.9889, + "step": 50340 + }, + { + "epoch": 3.72, + "learning_rate": 6.665146341342507e-06, + "loss": 0.9813, + "step": 50341 + }, + { + "epoch": 3.72, + "learning_rate": 6.664770305671173e-06, + "loss": 0.9175, + "step": 50342 + }, + { + "epoch": 3.72, + "learning_rate": 6.664394275306138e-06, + "loss": 0.9916, + "step": 50343 + }, + { + "epoch": 3.72, + "learning_rate": 6.664018250248005e-06, + "loss": 0.8878, + "step": 50344 + }, + { + "epoch": 3.72, + "learning_rate": 6.663642230497371e-06, + "loss": 1.0015, + "step": 50345 + }, + { + "epoch": 3.72, + "learning_rate": 6.663266216054836e-06, + "loss": 1.0045, + "step": 50346 + }, + { + "epoch": 3.72, + "learning_rate": 6.662890206920993e-06, + "loss": 0.9258, + "step": 50347 + }, + { + "epoch": 3.72, + "learning_rate": 6.662514203096448e-06, + "loss": 0.9434, + "step": 50348 + }, + { + "epoch": 3.72, + "learning_rate": 6.6621382045817924e-06, + "loss": 0.9796, + "step": 50349 + }, + { + "epoch": 3.72, + "learning_rate": 6.661762211377629e-06, + "loss": 0.9681, + "step": 50350 + }, + { + "epoch": 3.72, + "learning_rate": 6.661386223484554e-06, + "loss": 1.087, + "step": 50351 + }, + { + "epoch": 3.72, + "learning_rate": 6.661010240903165e-06, + "loss": 0.9479, + "step": 50352 + }, + { + "epoch": 3.72, + "learning_rate": 6.660634263634062e-06, + "loss": 0.9929, + "step": 50353 + }, + { + "epoch": 3.72, + "learning_rate": 6.6602582916778415e-06, + "loss": 1.0216, + "step": 50354 + }, + { + "epoch": 3.72, + "learning_rate": 6.659882325035103e-06, + "loss": 1.0243, + "step": 50355 + }, + { + "epoch": 3.72, + "learning_rate": 6.659506363706441e-06, + "loss": 0.8384, + "step": 50356 + }, + { + "epoch": 3.72, + "learning_rate": 6.65913040769246e-06, + "loss": 1.0673, + "step": 50357 + }, + { + "epoch": 3.72, + "learning_rate": 6.658754456993752e-06, + "loss": 0.913, + "step": 50358 + }, + { + "epoch": 3.72, + "learning_rate": 6.658378511610921e-06, + "loss": 1.0569, + "step": 50359 + }, + { + "epoch": 3.72, + "learning_rate": 6.658002571544558e-06, + "loss": 1.0618, + "step": 50360 + }, + { + "epoch": 3.72, + "learning_rate": 6.657626636795269e-06, + "loss": 1.0145, + "step": 50361 + }, + { + "epoch": 3.72, + "learning_rate": 6.65725070736365e-06, + "loss": 0.964, + "step": 50362 + }, + { + "epoch": 3.72, + "learning_rate": 6.656874783250289e-06, + "loss": 1.0135, + "step": 50363 + }, + { + "epoch": 3.72, + "learning_rate": 6.6564988644558e-06, + "loss": 0.9969, + "step": 50364 + }, + { + "epoch": 3.72, + "learning_rate": 6.656122950980768e-06, + "loss": 0.9657, + "step": 50365 + }, + { + "epoch": 3.72, + "learning_rate": 6.655747042825802e-06, + "loss": 0.9034, + "step": 50366 + }, + { + "epoch": 3.72, + "learning_rate": 6.655371139991489e-06, + "loss": 0.9487, + "step": 50367 + }, + { + "epoch": 3.72, + "learning_rate": 6.654995242478439e-06, + "loss": 1.0223, + "step": 50368 + }, + { + "epoch": 3.72, + "learning_rate": 6.654619350287239e-06, + "loss": 0.9866, + "step": 50369 + }, + { + "epoch": 3.72, + "learning_rate": 6.654243463418493e-06, + "loss": 0.9559, + "step": 50370 + }, + { + "epoch": 3.72, + "learning_rate": 6.6538675818728e-06, + "loss": 1.0014, + "step": 50371 + }, + { + "epoch": 3.72, + "learning_rate": 6.653491705650752e-06, + "loss": 0.9707, + "step": 50372 + }, + { + "epoch": 3.72, + "learning_rate": 6.6531158347529525e-06, + "loss": 0.9024, + "step": 50373 + }, + { + "epoch": 3.72, + "learning_rate": 6.6527399691799965e-06, + "loss": 1.0638, + "step": 50374 + }, + { + "epoch": 3.72, + "learning_rate": 6.652364108932485e-06, + "loss": 0.9892, + "step": 50375 + }, + { + "epoch": 3.72, + "learning_rate": 6.651988254011012e-06, + "loss": 1.0772, + "step": 50376 + }, + { + "epoch": 3.72, + "learning_rate": 6.651612404416181e-06, + "loss": 0.9609, + "step": 50377 + }, + { + "epoch": 3.72, + "learning_rate": 6.651236560148584e-06, + "loss": 0.9355, + "step": 50378 + }, + { + "epoch": 3.72, + "learning_rate": 6.650860721208825e-06, + "loss": 1.0652, + "step": 50379 + }, + { + "epoch": 3.72, + "learning_rate": 6.6504848875975e-06, + "loss": 0.9224, + "step": 50380 + }, + { + "epoch": 3.72, + "learning_rate": 6.650109059315199e-06, + "loss": 0.9996, + "step": 50381 + }, + { + "epoch": 3.72, + "learning_rate": 6.6497332363625326e-06, + "loss": 0.9993, + "step": 50382 + }, + { + "epoch": 3.72, + "learning_rate": 6.649357418740086e-06, + "loss": 0.9544, + "step": 50383 + }, + { + "epoch": 3.72, + "learning_rate": 6.6489816064484715e-06, + "loss": 0.9902, + "step": 50384 + }, + { + "epoch": 3.72, + "learning_rate": 6.648605799488273e-06, + "loss": 0.8761, + "step": 50385 + }, + { + "epoch": 3.72, + "learning_rate": 6.6482299978601e-06, + "loss": 0.9859, + "step": 50386 + }, + { + "epoch": 3.72, + "learning_rate": 6.647854201564539e-06, + "loss": 0.9945, + "step": 50387 + }, + { + "epoch": 3.72, + "learning_rate": 6.6474784106021995e-06, + "loss": 0.958, + "step": 50388 + }, + { + "epoch": 3.72, + "learning_rate": 6.647102624973673e-06, + "loss": 1.1048, + "step": 50389 + }, + { + "epoch": 3.72, + "learning_rate": 6.6467268446795565e-06, + "loss": 1.0611, + "step": 50390 + }, + { + "epoch": 3.72, + "learning_rate": 6.646351069720451e-06, + "loss": 1.0189, + "step": 50391 + }, + { + "epoch": 3.72, + "learning_rate": 6.645975300096951e-06, + "loss": 0.9766, + "step": 50392 + }, + { + "epoch": 3.72, + "learning_rate": 6.6455995358096585e-06, + "loss": 1.007, + "step": 50393 + }, + { + "epoch": 3.72, + "learning_rate": 6.645223776859167e-06, + "loss": 0.944, + "step": 50394 + }, + { + "epoch": 3.72, + "learning_rate": 6.644848023246078e-06, + "loss": 0.9325, + "step": 50395 + }, + { + "epoch": 3.72, + "learning_rate": 6.644472274970986e-06, + "loss": 1.0136, + "step": 50396 + }, + { + "epoch": 3.72, + "learning_rate": 6.644096532034493e-06, + "loss": 0.987, + "step": 50397 + }, + { + "epoch": 3.72, + "learning_rate": 6.643720794437192e-06, + "loss": 1.1082, + "step": 50398 + }, + { + "epoch": 3.72, + "learning_rate": 6.643345062179686e-06, + "loss": 0.9838, + "step": 50399 + }, + { + "epoch": 3.72, + "learning_rate": 6.64296933526257e-06, + "loss": 0.9877, + "step": 50400 + }, + { + "epoch": 3.72, + "learning_rate": 6.642593613686439e-06, + "loss": 1.0216, + "step": 50401 + }, + { + "epoch": 3.72, + "learning_rate": 6.642217897451894e-06, + "loss": 1.0609, + "step": 50402 + }, + { + "epoch": 3.72, + "learning_rate": 6.641842186559533e-06, + "loss": 1.0593, + "step": 50403 + }, + { + "epoch": 3.72, + "learning_rate": 6.641466481009955e-06, + "loss": 1.1505, + "step": 50404 + }, + { + "epoch": 3.72, + "learning_rate": 6.641090780803752e-06, + "loss": 1.0114, + "step": 50405 + }, + { + "epoch": 3.72, + "learning_rate": 6.640715085941529e-06, + "loss": 1.0167, + "step": 50406 + }, + { + "epoch": 3.72, + "learning_rate": 6.640339396423877e-06, + "loss": 1.0338, + "step": 50407 + }, + { + "epoch": 3.72, + "learning_rate": 6.639963712251399e-06, + "loss": 0.8193, + "step": 50408 + }, + { + "epoch": 3.72, + "learning_rate": 6.639588033424693e-06, + "loss": 1.0169, + "step": 50409 + }, + { + "epoch": 3.72, + "learning_rate": 6.639212359944349e-06, + "loss": 1.0303, + "step": 50410 + }, + { + "epoch": 3.72, + "learning_rate": 6.638836691810976e-06, + "loss": 1.0351, + "step": 50411 + }, + { + "epoch": 3.72, + "learning_rate": 6.63846102902516e-06, + "loss": 0.9172, + "step": 50412 + }, + { + "epoch": 3.72, + "learning_rate": 6.638085371587508e-06, + "loss": 0.9221, + "step": 50413 + }, + { + "epoch": 3.73, + "learning_rate": 6.637709719498613e-06, + "loss": 1.0225, + "step": 50414 + }, + { + "epoch": 3.73, + "learning_rate": 6.637334072759074e-06, + "loss": 1.0114, + "step": 50415 + }, + { + "epoch": 3.73, + "learning_rate": 6.636958431369487e-06, + "loss": 0.8824, + "step": 50416 + }, + { + "epoch": 3.73, + "learning_rate": 6.636582795330453e-06, + "loss": 0.9752, + "step": 50417 + }, + { + "epoch": 3.73, + "learning_rate": 6.636207164642567e-06, + "loss": 1.0132, + "step": 50418 + }, + { + "epoch": 3.73, + "learning_rate": 6.635831539306426e-06, + "loss": 0.947, + "step": 50419 + }, + { + "epoch": 3.73, + "learning_rate": 6.635455919322631e-06, + "loss": 0.9713, + "step": 50420 + }, + { + "epoch": 3.73, + "learning_rate": 6.635080304691774e-06, + "loss": 0.9266, + "step": 50421 + }, + { + "epoch": 3.73, + "learning_rate": 6.63470469541446e-06, + "loss": 1.0545, + "step": 50422 + }, + { + "epoch": 3.73, + "learning_rate": 6.634329091491281e-06, + "loss": 1.0702, + "step": 50423 + }, + { + "epoch": 3.73, + "learning_rate": 6.633953492922836e-06, + "loss": 0.9769, + "step": 50424 + }, + { + "epoch": 3.73, + "learning_rate": 6.633577899709722e-06, + "loss": 0.9757, + "step": 50425 + }, + { + "epoch": 3.73, + "learning_rate": 6.63320231185254e-06, + "loss": 0.9615, + "step": 50426 + }, + { + "epoch": 3.73, + "learning_rate": 6.632826729351886e-06, + "loss": 0.9913, + "step": 50427 + }, + { + "epoch": 3.73, + "learning_rate": 6.632451152208353e-06, + "loss": 0.9868, + "step": 50428 + }, + { + "epoch": 3.73, + "learning_rate": 6.632075580422545e-06, + "loss": 1.1037, + "step": 50429 + }, + { + "epoch": 3.73, + "learning_rate": 6.631700013995051e-06, + "loss": 0.9052, + "step": 50430 + }, + { + "epoch": 3.73, + "learning_rate": 6.631324452926481e-06, + "loss": 1.0926, + "step": 50431 + }, + { + "epoch": 3.73, + "learning_rate": 6.63094889721742e-06, + "loss": 0.9256, + "step": 50432 + }, + { + "epoch": 3.73, + "learning_rate": 6.630573346868475e-06, + "loss": 0.9969, + "step": 50433 + }, + { + "epoch": 3.73, + "learning_rate": 6.630197801880236e-06, + "loss": 1.0614, + "step": 50434 + }, + { + "epoch": 3.73, + "learning_rate": 6.629822262253308e-06, + "loss": 0.9597, + "step": 50435 + }, + { + "epoch": 3.73, + "learning_rate": 6.629446727988284e-06, + "loss": 0.9045, + "step": 50436 + }, + { + "epoch": 3.73, + "learning_rate": 6.6290711990857596e-06, + "loss": 1.0526, + "step": 50437 + }, + { + "epoch": 3.73, + "learning_rate": 6.628695675546337e-06, + "loss": 1.0666, + "step": 50438 + }, + { + "epoch": 3.73, + "learning_rate": 6.628320157370609e-06, + "loss": 0.9567, + "step": 50439 + }, + { + "epoch": 3.73, + "learning_rate": 6.627944644559177e-06, + "loss": 0.9202, + "step": 50440 + }, + { + "epoch": 3.73, + "learning_rate": 6.627569137112635e-06, + "loss": 0.9269, + "step": 50441 + }, + { + "epoch": 3.73, + "learning_rate": 6.627193635031585e-06, + "loss": 0.9634, + "step": 50442 + }, + { + "epoch": 3.73, + "learning_rate": 6.6268181383166185e-06, + "loss": 1.027, + "step": 50443 + }, + { + "epoch": 3.73, + "learning_rate": 6.626442646968339e-06, + "loss": 0.966, + "step": 50444 + }, + { + "epoch": 3.73, + "learning_rate": 6.626067160987341e-06, + "loss": 0.9386, + "step": 50445 + }, + { + "epoch": 3.73, + "learning_rate": 6.62569168037422e-06, + "loss": 1.0456, + "step": 50446 + }, + { + "epoch": 3.73, + "learning_rate": 6.625316205129576e-06, + "loss": 0.9698, + "step": 50447 + }, + { + "epoch": 3.73, + "learning_rate": 6.624940735254004e-06, + "loss": 0.9464, + "step": 50448 + }, + { + "epoch": 3.73, + "learning_rate": 6.624565270748107e-06, + "loss": 1.0182, + "step": 50449 + }, + { + "epoch": 3.73, + "learning_rate": 6.624189811612474e-06, + "loss": 0.9952, + "step": 50450 + }, + { + "epoch": 3.73, + "learning_rate": 6.623814357847708e-06, + "loss": 0.986, + "step": 50451 + }, + { + "epoch": 3.73, + "learning_rate": 6.623438909454406e-06, + "loss": 0.9338, + "step": 50452 + }, + { + "epoch": 3.73, + "learning_rate": 6.623063466433165e-06, + "loss": 0.973, + "step": 50453 + }, + { + "epoch": 3.73, + "learning_rate": 6.62268802878458e-06, + "loss": 1.0521, + "step": 50454 + }, + { + "epoch": 3.73, + "learning_rate": 6.622312596509252e-06, + "loss": 0.9729, + "step": 50455 + }, + { + "epoch": 3.73, + "learning_rate": 6.621937169607777e-06, + "loss": 1.0033, + "step": 50456 + }, + { + "epoch": 3.73, + "learning_rate": 6.621561748080749e-06, + "loss": 1.0456, + "step": 50457 + }, + { + "epoch": 3.73, + "learning_rate": 6.621186331928771e-06, + "loss": 0.9742, + "step": 50458 + }, + { + "epoch": 3.73, + "learning_rate": 6.6208109211524315e-06, + "loss": 0.9919, + "step": 50459 + }, + { + "epoch": 3.73, + "learning_rate": 6.620435515752339e-06, + "loss": 1.0123, + "step": 50460 + }, + { + "epoch": 3.73, + "learning_rate": 6.620060115729081e-06, + "loss": 0.9483, + "step": 50461 + }, + { + "epoch": 3.73, + "learning_rate": 6.619684721083264e-06, + "loss": 0.886, + "step": 50462 + }, + { + "epoch": 3.73, + "learning_rate": 6.619309331815475e-06, + "loss": 1.0823, + "step": 50463 + }, + { + "epoch": 3.73, + "learning_rate": 6.61893394792632e-06, + "loss": 0.987, + "step": 50464 + }, + { + "epoch": 3.73, + "learning_rate": 6.6185585694163925e-06, + "loss": 1.0436, + "step": 50465 + }, + { + "epoch": 3.73, + "learning_rate": 6.618183196286288e-06, + "loss": 1.0946, + "step": 50466 + }, + { + "epoch": 3.73, + "learning_rate": 6.617807828536608e-06, + "loss": 1.0294, + "step": 50467 + }, + { + "epoch": 3.73, + "learning_rate": 6.617432466167945e-06, + "loss": 0.8439, + "step": 50468 + }, + { + "epoch": 3.73, + "learning_rate": 6.6170571091809e-06, + "loss": 0.9828, + "step": 50469 + }, + { + "epoch": 3.73, + "learning_rate": 6.616681757576067e-06, + "loss": 1.0986, + "step": 50470 + }, + { + "epoch": 3.73, + "learning_rate": 6.616306411354047e-06, + "loss": 0.9623, + "step": 50471 + }, + { + "epoch": 3.73, + "learning_rate": 6.615931070515433e-06, + "loss": 1.027, + "step": 50472 + }, + { + "epoch": 3.73, + "learning_rate": 6.615555735060827e-06, + "loss": 1.01, + "step": 50473 + }, + { + "epoch": 3.73, + "learning_rate": 6.615180404990824e-06, + "loss": 1.0106, + "step": 50474 + }, + { + "epoch": 3.73, + "learning_rate": 6.614805080306016e-06, + "loss": 0.8918, + "step": 50475 + }, + { + "epoch": 3.73, + "learning_rate": 6.6144297610070105e-06, + "loss": 0.9688, + "step": 50476 + }, + { + "epoch": 3.73, + "learning_rate": 6.614054447094391e-06, + "loss": 1.015, + "step": 50477 + }, + { + "epoch": 3.73, + "learning_rate": 6.613679138568772e-06, + "loss": 0.9146, + "step": 50478 + }, + { + "epoch": 3.73, + "learning_rate": 6.613303835430731e-06, + "loss": 0.891, + "step": 50479 + }, + { + "epoch": 3.73, + "learning_rate": 6.6129285376808835e-06, + "loss": 0.9385, + "step": 50480 + }, + { + "epoch": 3.73, + "learning_rate": 6.612553245319811e-06, + "loss": 0.9694, + "step": 50481 + }, + { + "epoch": 3.73, + "learning_rate": 6.612177958348125e-06, + "loss": 1.0107, + "step": 50482 + }, + { + "epoch": 3.73, + "learning_rate": 6.611802676766413e-06, + "loss": 0.9402, + "step": 50483 + }, + { + "epoch": 3.73, + "learning_rate": 6.611427400575272e-06, + "loss": 0.9024, + "step": 50484 + }, + { + "epoch": 3.73, + "learning_rate": 6.611052129775304e-06, + "loss": 0.9611, + "step": 50485 + }, + { + "epoch": 3.73, + "learning_rate": 6.610676864367102e-06, + "loss": 0.952, + "step": 50486 + }, + { + "epoch": 3.73, + "learning_rate": 6.610301604351266e-06, + "loss": 1.0981, + "step": 50487 + }, + { + "epoch": 3.73, + "learning_rate": 6.6099263497283905e-06, + "loss": 0.9798, + "step": 50488 + }, + { + "epoch": 3.73, + "learning_rate": 6.609551100499075e-06, + "loss": 1.0487, + "step": 50489 + }, + { + "epoch": 3.73, + "learning_rate": 6.609175856663912e-06, + "loss": 0.9925, + "step": 50490 + }, + { + "epoch": 3.73, + "learning_rate": 6.608800618223505e-06, + "loss": 1.008, + "step": 50491 + }, + { + "epoch": 3.73, + "learning_rate": 6.608425385178447e-06, + "loss": 1.0442, + "step": 50492 + }, + { + "epoch": 3.73, + "learning_rate": 6.608050157529335e-06, + "loss": 1.0597, + "step": 50493 + }, + { + "epoch": 3.73, + "learning_rate": 6.6076749352767685e-06, + "loss": 1.0332, + "step": 50494 + }, + { + "epoch": 3.73, + "learning_rate": 6.607299718421339e-06, + "loss": 1.0545, + "step": 50495 + }, + { + "epoch": 3.73, + "learning_rate": 6.60692450696365e-06, + "loss": 1.0537, + "step": 50496 + }, + { + "epoch": 3.73, + "learning_rate": 6.606549300904295e-06, + "loss": 0.952, + "step": 50497 + }, + { + "epoch": 3.73, + "learning_rate": 6.606174100243872e-06, + "loss": 1.0105, + "step": 50498 + }, + { + "epoch": 3.73, + "learning_rate": 6.6057989049829756e-06, + "loss": 0.9658, + "step": 50499 + }, + { + "epoch": 3.73, + "learning_rate": 6.605423715122206e-06, + "loss": 1.0016, + "step": 50500 + }, + { + "epoch": 3.73, + "learning_rate": 6.605048530662161e-06, + "loss": 0.9626, + "step": 50501 + }, + { + "epoch": 3.73, + "learning_rate": 6.604673351603432e-06, + "loss": 1.1291, + "step": 50502 + }, + { + "epoch": 3.73, + "learning_rate": 6.604298177946621e-06, + "loss": 0.9985, + "step": 50503 + }, + { + "epoch": 3.73, + "learning_rate": 6.603923009692321e-06, + "loss": 0.9389, + "step": 50504 + }, + { + "epoch": 3.73, + "learning_rate": 6.603547846841132e-06, + "loss": 1.0699, + "step": 50505 + }, + { + "epoch": 3.73, + "learning_rate": 6.6031726893936485e-06, + "loss": 1.0104, + "step": 50506 + }, + { + "epoch": 3.73, + "learning_rate": 6.6027975373504704e-06, + "loss": 0.9618, + "step": 50507 + }, + { + "epoch": 3.73, + "learning_rate": 6.602422390712191e-06, + "loss": 1.0156, + "step": 50508 + }, + { + "epoch": 3.73, + "learning_rate": 6.602047249479411e-06, + "loss": 1.0979, + "step": 50509 + }, + { + "epoch": 3.73, + "learning_rate": 6.6016721136527225e-06, + "loss": 0.9807, + "step": 50510 + }, + { + "epoch": 3.73, + "learning_rate": 6.601296983232728e-06, + "loss": 0.9701, + "step": 50511 + }, + { + "epoch": 3.73, + "learning_rate": 6.600921858220022e-06, + "loss": 0.9422, + "step": 50512 + }, + { + "epoch": 3.73, + "learning_rate": 6.600546738615196e-06, + "loss": 0.9958, + "step": 50513 + }, + { + "epoch": 3.73, + "learning_rate": 6.6001716244188564e-06, + "loss": 1.0321, + "step": 50514 + }, + { + "epoch": 3.73, + "learning_rate": 6.599796515631591e-06, + "loss": 0.9044, + "step": 50515 + }, + { + "epoch": 3.73, + "learning_rate": 6.599421412254003e-06, + "loss": 1.0577, + "step": 50516 + }, + { + "epoch": 3.73, + "learning_rate": 6.599046314286687e-06, + "loss": 1.0387, + "step": 50517 + }, + { + "epoch": 3.73, + "learning_rate": 6.5986712217302395e-06, + "loss": 0.9831, + "step": 50518 + }, + { + "epoch": 3.73, + "learning_rate": 6.5982961345852555e-06, + "loss": 0.9038, + "step": 50519 + }, + { + "epoch": 3.73, + "learning_rate": 6.597921052852336e-06, + "loss": 0.9252, + "step": 50520 + }, + { + "epoch": 3.73, + "learning_rate": 6.597545976532078e-06, + "loss": 1.0387, + "step": 50521 + }, + { + "epoch": 3.73, + "learning_rate": 6.59717090562507e-06, + "loss": 1.1629, + "step": 50522 + }, + { + "epoch": 3.73, + "learning_rate": 6.59679584013192e-06, + "loss": 0.9892, + "step": 50523 + }, + { + "epoch": 3.73, + "learning_rate": 6.596420780053212e-06, + "loss": 0.9359, + "step": 50524 + }, + { + "epoch": 3.73, + "learning_rate": 6.596045725389557e-06, + "loss": 0.9707, + "step": 50525 + }, + { + "epoch": 3.73, + "learning_rate": 6.595670676141539e-06, + "loss": 1.0441, + "step": 50526 + }, + { + "epoch": 3.73, + "learning_rate": 6.595295632309766e-06, + "loss": 0.9883, + "step": 50527 + }, + { + "epoch": 3.73, + "learning_rate": 6.5949205938948215e-06, + "loss": 0.8746, + "step": 50528 + }, + { + "epoch": 3.73, + "learning_rate": 6.5945455608973164e-06, + "loss": 0.9726, + "step": 50529 + }, + { + "epoch": 3.73, + "learning_rate": 6.594170533317838e-06, + "loss": 0.8954, + "step": 50530 + }, + { + "epoch": 3.73, + "learning_rate": 6.5937955111569855e-06, + "loss": 0.9697, + "step": 50531 + }, + { + "epoch": 3.73, + "learning_rate": 6.5934204944153564e-06, + "loss": 1.0417, + "step": 50532 + }, + { + "epoch": 3.73, + "learning_rate": 6.593045483093544e-06, + "loss": 1.0488, + "step": 50533 + }, + { + "epoch": 3.73, + "learning_rate": 6.592670477192151e-06, + "loss": 1.0532, + "step": 50534 + }, + { + "epoch": 3.73, + "learning_rate": 6.592295476711767e-06, + "loss": 0.9267, + "step": 50535 + }, + { + "epoch": 3.73, + "learning_rate": 6.591920481652995e-06, + "loss": 1.0084, + "step": 50536 + }, + { + "epoch": 3.73, + "learning_rate": 6.591545492016427e-06, + "loss": 1.0165, + "step": 50537 + }, + { + "epoch": 3.73, + "learning_rate": 6.591170507802662e-06, + "loss": 0.9919, + "step": 50538 + }, + { + "epoch": 3.73, + "learning_rate": 6.590795529012297e-06, + "loss": 1.0075, + "step": 50539 + }, + { + "epoch": 3.73, + "learning_rate": 6.590420555645924e-06, + "loss": 0.9486, + "step": 50540 + }, + { + "epoch": 3.73, + "learning_rate": 6.590045587704148e-06, + "loss": 0.9579, + "step": 50541 + }, + { + "epoch": 3.73, + "learning_rate": 6.589670625187556e-06, + "loss": 0.9666, + "step": 50542 + }, + { + "epoch": 3.73, + "learning_rate": 6.589295668096754e-06, + "loss": 0.947, + "step": 50543 + }, + { + "epoch": 3.73, + "learning_rate": 6.588920716432329e-06, + "loss": 0.9264, + "step": 50544 + }, + { + "epoch": 3.73, + "learning_rate": 6.588545770194887e-06, + "loss": 0.9165, + "step": 50545 + }, + { + "epoch": 3.73, + "learning_rate": 6.588170829385014e-06, + "loss": 0.9035, + "step": 50546 + }, + { + "epoch": 3.73, + "learning_rate": 6.587795894003319e-06, + "loss": 1.0409, + "step": 50547 + }, + { + "epoch": 3.73, + "learning_rate": 6.587420964050391e-06, + "loss": 0.8944, + "step": 50548 + }, + { + "epoch": 3.74, + "learning_rate": 6.587046039526824e-06, + "loss": 1.0736, + "step": 50549 + }, + { + "epoch": 3.74, + "learning_rate": 6.586671120433219e-06, + "loss": 0.9403, + "step": 50550 + }, + { + "epoch": 3.74, + "learning_rate": 6.586296206770171e-06, + "loss": 0.934, + "step": 50551 + }, + { + "epoch": 3.74, + "learning_rate": 6.585921298538278e-06, + "loss": 0.9537, + "step": 50552 + }, + { + "epoch": 3.74, + "learning_rate": 6.585546395738134e-06, + "loss": 1.1076, + "step": 50553 + }, + { + "epoch": 3.74, + "learning_rate": 6.585171498370337e-06, + "loss": 1.0706, + "step": 50554 + }, + { + "epoch": 3.74, + "learning_rate": 6.584796606435484e-06, + "loss": 1.0068, + "step": 50555 + }, + { + "epoch": 3.74, + "learning_rate": 6.584421719934171e-06, + "loss": 0.9598, + "step": 50556 + }, + { + "epoch": 3.74, + "learning_rate": 6.584046838866996e-06, + "loss": 0.8537, + "step": 50557 + }, + { + "epoch": 3.74, + "learning_rate": 6.5836719632345496e-06, + "loss": 1.0296, + "step": 50558 + }, + { + "epoch": 3.74, + "learning_rate": 6.583297093037435e-06, + "loss": 0.9725, + "step": 50559 + }, + { + "epoch": 3.74, + "learning_rate": 6.582922228276243e-06, + "loss": 0.9958, + "step": 50560 + }, + { + "epoch": 3.74, + "learning_rate": 6.582547368951576e-06, + "loss": 1.0182, + "step": 50561 + }, + { + "epoch": 3.74, + "learning_rate": 6.5821725150640245e-06, + "loss": 1.065, + "step": 50562 + }, + { + "epoch": 3.74, + "learning_rate": 6.58179766661419e-06, + "loss": 0.9359, + "step": 50563 + }, + { + "epoch": 3.74, + "learning_rate": 6.5814228236026645e-06, + "loss": 0.9724, + "step": 50564 + }, + { + "epoch": 3.74, + "learning_rate": 6.581047986030049e-06, + "loss": 0.9624, + "step": 50565 + }, + { + "epoch": 3.74, + "learning_rate": 6.580673153896935e-06, + "loss": 0.975, + "step": 50566 + }, + { + "epoch": 3.74, + "learning_rate": 6.580298327203922e-06, + "loss": 0.9695, + "step": 50567 + }, + { + "epoch": 3.74, + "learning_rate": 6.579923505951609e-06, + "loss": 0.9875, + "step": 50568 + }, + { + "epoch": 3.74, + "learning_rate": 6.57954869014058e-06, + "loss": 1.1082, + "step": 50569 + }, + { + "epoch": 3.74, + "learning_rate": 6.579173879771449e-06, + "loss": 1.0656, + "step": 50570 + }, + { + "epoch": 3.74, + "learning_rate": 6.578799074844796e-06, + "loss": 0.9525, + "step": 50571 + }, + { + "epoch": 3.74, + "learning_rate": 6.578424275361231e-06, + "loss": 0.9213, + "step": 50572 + }, + { + "epoch": 3.74, + "learning_rate": 6.578049481321338e-06, + "loss": 0.9836, + "step": 50573 + }, + { + "epoch": 3.74, + "learning_rate": 6.577674692725727e-06, + "loss": 1.0001, + "step": 50574 + }, + { + "epoch": 3.74, + "learning_rate": 6.577299909574978e-06, + "loss": 1.0276, + "step": 50575 + }, + { + "epoch": 3.74, + "learning_rate": 6.576925131869702e-06, + "loss": 1.0028, + "step": 50576 + }, + { + "epoch": 3.74, + "learning_rate": 6.5765503596104896e-06, + "loss": 1.026, + "step": 50577 + }, + { + "epoch": 3.74, + "learning_rate": 6.576175592797933e-06, + "loss": 1.0219, + "step": 50578 + }, + { + "epoch": 3.74, + "learning_rate": 6.575800831432634e-06, + "loss": 1.0633, + "step": 50579 + }, + { + "epoch": 3.74, + "learning_rate": 6.5754260755151845e-06, + "loss": 0.9942, + "step": 50580 + }, + { + "epoch": 3.74, + "learning_rate": 6.575051325046186e-06, + "loss": 1.0008, + "step": 50581 + }, + { + "epoch": 3.74, + "learning_rate": 6.574676580026229e-06, + "loss": 0.8928, + "step": 50582 + }, + { + "epoch": 3.74, + "learning_rate": 6.574301840455914e-06, + "loss": 0.9742, + "step": 50583 + }, + { + "epoch": 3.74, + "learning_rate": 6.5739271063358356e-06, + "loss": 1.0731, + "step": 50584 + }, + { + "epoch": 3.74, + "learning_rate": 6.5735523776665925e-06, + "loss": 1.0785, + "step": 50585 + }, + { + "epoch": 3.74, + "learning_rate": 6.573177654448777e-06, + "loss": 1.1141, + "step": 50586 + }, + { + "epoch": 3.74, + "learning_rate": 6.572802936682985e-06, + "loss": 1.0032, + "step": 50587 + }, + { + "epoch": 3.74, + "learning_rate": 6.572428224369815e-06, + "loss": 0.9853, + "step": 50588 + }, + { + "epoch": 3.74, + "learning_rate": 6.572053517509863e-06, + "loss": 0.9437, + "step": 50589 + }, + { + "epoch": 3.74, + "learning_rate": 6.571678816103728e-06, + "loss": 1.064, + "step": 50590 + }, + { + "epoch": 3.74, + "learning_rate": 6.571304120151998e-06, + "loss": 1.0291, + "step": 50591 + }, + { + "epoch": 3.74, + "learning_rate": 6.570929429655279e-06, + "loss": 1.0666, + "step": 50592 + }, + { + "epoch": 3.74, + "learning_rate": 6.570554744614156e-06, + "loss": 1.0958, + "step": 50593 + }, + { + "epoch": 3.74, + "learning_rate": 6.570180065029238e-06, + "loss": 1.0487, + "step": 50594 + }, + { + "epoch": 3.74, + "learning_rate": 6.569805390901111e-06, + "loss": 0.9024, + "step": 50595 + }, + { + "epoch": 3.74, + "learning_rate": 6.569430722230375e-06, + "loss": 0.8795, + "step": 50596 + }, + { + "epoch": 3.74, + "learning_rate": 6.569056059017627e-06, + "loss": 0.9796, + "step": 50597 + }, + { + "epoch": 3.74, + "learning_rate": 6.568681401263458e-06, + "loss": 1.0397, + "step": 50598 + }, + { + "epoch": 3.74, + "learning_rate": 6.568306748968472e-06, + "loss": 1.1094, + "step": 50599 + }, + { + "epoch": 3.74, + "learning_rate": 6.567932102133256e-06, + "loss": 0.9712, + "step": 50600 + }, + { + "epoch": 3.74, + "learning_rate": 6.567557460758416e-06, + "loss": 0.9277, + "step": 50601 + }, + { + "epoch": 3.74, + "learning_rate": 6.567182824844539e-06, + "loss": 0.9655, + "step": 50602 + }, + { + "epoch": 3.74, + "learning_rate": 6.566808194392228e-06, + "loss": 0.9898, + "step": 50603 + }, + { + "epoch": 3.74, + "learning_rate": 6.566433569402075e-06, + "loss": 1.0746, + "step": 50604 + }, + { + "epoch": 3.74, + "learning_rate": 6.5660589498746765e-06, + "loss": 0.9948, + "step": 50605 + }, + { + "epoch": 3.74, + "learning_rate": 6.565684335810629e-06, + "loss": 1.06, + "step": 50606 + }, + { + "epoch": 3.74, + "learning_rate": 6.565309727210529e-06, + "loss": 1.0615, + "step": 50607 + }, + { + "epoch": 3.74, + "learning_rate": 6.564935124074973e-06, + "loss": 1.0513, + "step": 50608 + }, + { + "epoch": 3.74, + "learning_rate": 6.564560526404553e-06, + "loss": 0.9282, + "step": 50609 + }, + { + "epoch": 3.74, + "learning_rate": 6.564185934199872e-06, + "loss": 0.943, + "step": 50610 + }, + { + "epoch": 3.74, + "learning_rate": 6.56381134746152e-06, + "loss": 1.0267, + "step": 50611 + }, + { + "epoch": 3.74, + "learning_rate": 6.563436766190095e-06, + "loss": 1.0189, + "step": 50612 + }, + { + "epoch": 3.74, + "learning_rate": 6.5630621903861955e-06, + "loss": 1.0011, + "step": 50613 + }, + { + "epoch": 3.74, + "learning_rate": 6.56268762005041e-06, + "loss": 1.0666, + "step": 50614 + }, + { + "epoch": 3.74, + "learning_rate": 6.562313055183345e-06, + "loss": 0.9026, + "step": 50615 + }, + { + "epoch": 3.74, + "learning_rate": 6.561938495785584e-06, + "loss": 1.0147, + "step": 50616 + }, + { + "epoch": 3.74, + "learning_rate": 6.561563941857736e-06, + "loss": 0.9771, + "step": 50617 + }, + { + "epoch": 3.74, + "learning_rate": 6.561189393400385e-06, + "loss": 0.9788, + "step": 50618 + }, + { + "epoch": 3.74, + "learning_rate": 6.560814850414139e-06, + "loss": 1.0085, + "step": 50619 + }, + { + "epoch": 3.74, + "learning_rate": 6.56044031289958e-06, + "loss": 0.9493, + "step": 50620 + }, + { + "epoch": 3.74, + "learning_rate": 6.560065780857318e-06, + "loss": 1.0164, + "step": 50621 + }, + { + "epoch": 3.74, + "learning_rate": 6.559691254287938e-06, + "loss": 1.0269, + "step": 50622 + }, + { + "epoch": 3.74, + "learning_rate": 6.5593167331920415e-06, + "loss": 1.0593, + "step": 50623 + }, + { + "epoch": 3.74, + "learning_rate": 6.5589422175702236e-06, + "loss": 1.0036, + "step": 50624 + }, + { + "epoch": 3.74, + "learning_rate": 6.558567707423077e-06, + "loss": 0.9585, + "step": 50625 + }, + { + "epoch": 3.74, + "learning_rate": 6.558193202751202e-06, + "loss": 0.9161, + "step": 50626 + }, + { + "epoch": 3.74, + "learning_rate": 6.55781870355519e-06, + "loss": 0.9489, + "step": 50627 + }, + { + "epoch": 3.74, + "learning_rate": 6.557444209835642e-06, + "loss": 1.0006, + "step": 50628 + }, + { + "epoch": 3.74, + "learning_rate": 6.5570697215931476e-06, + "loss": 0.8801, + "step": 50629 + }, + { + "epoch": 3.74, + "learning_rate": 6.55669523882831e-06, + "loss": 1.0755, + "step": 50630 + }, + { + "epoch": 3.74, + "learning_rate": 6.556320761541717e-06, + "loss": 1.0212, + "step": 50631 + }, + { + "epoch": 3.74, + "learning_rate": 6.555946289733972e-06, + "loss": 1.0278, + "step": 50632 + }, + { + "epoch": 3.74, + "learning_rate": 6.555571823405669e-06, + "loss": 1.0304, + "step": 50633 + }, + { + "epoch": 3.74, + "learning_rate": 6.555197362557395e-06, + "loss": 0.9005, + "step": 50634 + }, + { + "epoch": 3.74, + "learning_rate": 6.55482290718976e-06, + "loss": 0.8729, + "step": 50635 + }, + { + "epoch": 3.74, + "learning_rate": 6.554448457303345e-06, + "loss": 1.0128, + "step": 50636 + }, + { + "epoch": 3.74, + "learning_rate": 6.55407401289876e-06, + "loss": 1.0117, + "step": 50637 + }, + { + "epoch": 3.74, + "learning_rate": 6.553699573976587e-06, + "loss": 0.8539, + "step": 50638 + }, + { + "epoch": 3.74, + "learning_rate": 6.553325140537436e-06, + "loss": 0.9342, + "step": 50639 + }, + { + "epoch": 3.74, + "learning_rate": 6.552950712581889e-06, + "loss": 1.0249, + "step": 50640 + }, + { + "epoch": 3.74, + "learning_rate": 6.552576290110555e-06, + "loss": 0.8907, + "step": 50641 + }, + { + "epoch": 3.74, + "learning_rate": 6.552201873124019e-06, + "loss": 1.0071, + "step": 50642 + }, + { + "epoch": 3.74, + "learning_rate": 6.551827461622879e-06, + "loss": 0.9435, + "step": 50643 + }, + { + "epoch": 3.74, + "learning_rate": 6.551453055607735e-06, + "loss": 0.9936, + "step": 50644 + }, + { + "epoch": 3.74, + "learning_rate": 6.551078655079177e-06, + "loss": 1.0153, + "step": 50645 + }, + { + "epoch": 3.74, + "learning_rate": 6.550704260037806e-06, + "loss": 1.0035, + "step": 50646 + }, + { + "epoch": 3.74, + "learning_rate": 6.550329870484213e-06, + "loss": 1.0301, + "step": 50647 + }, + { + "epoch": 3.74, + "learning_rate": 6.549955486418997e-06, + "loss": 0.9952, + "step": 50648 + }, + { + "epoch": 3.74, + "learning_rate": 6.549581107842752e-06, + "loss": 1.0301, + "step": 50649 + }, + { + "epoch": 3.74, + "learning_rate": 6.549206734756075e-06, + "loss": 0.9566, + "step": 50650 + }, + { + "epoch": 3.74, + "learning_rate": 6.548832367159561e-06, + "loss": 0.9979, + "step": 50651 + }, + { + "epoch": 3.74, + "learning_rate": 6.548458005053804e-06, + "loss": 1.0018, + "step": 50652 + }, + { + "epoch": 3.74, + "learning_rate": 6.548083648439403e-06, + "loss": 0.995, + "step": 50653 + }, + { + "epoch": 3.74, + "learning_rate": 6.547709297316947e-06, + "loss": 1.0208, + "step": 50654 + }, + { + "epoch": 3.74, + "learning_rate": 6.547334951687041e-06, + "loss": 0.8965, + "step": 50655 + }, + { + "epoch": 3.74, + "learning_rate": 6.546960611550273e-06, + "loss": 0.9952, + "step": 50656 + }, + { + "epoch": 3.74, + "learning_rate": 6.546586276907243e-06, + "loss": 0.9969, + "step": 50657 + }, + { + "epoch": 3.74, + "learning_rate": 6.546211947758543e-06, + "loss": 0.9433, + "step": 50658 + }, + { + "epoch": 3.74, + "learning_rate": 6.545837624104771e-06, + "loss": 1.0552, + "step": 50659 + }, + { + "epoch": 3.74, + "learning_rate": 6.545463305946526e-06, + "loss": 0.9739, + "step": 50660 + }, + { + "epoch": 3.74, + "learning_rate": 6.545088993284392e-06, + "loss": 1.0227, + "step": 50661 + }, + { + "epoch": 3.74, + "learning_rate": 6.544714686118978e-06, + "loss": 0.9818, + "step": 50662 + }, + { + "epoch": 3.74, + "learning_rate": 6.544340384450868e-06, + "loss": 1.0089, + "step": 50663 + }, + { + "epoch": 3.74, + "learning_rate": 6.543966088280669e-06, + "loss": 1.0244, + "step": 50664 + }, + { + "epoch": 3.74, + "learning_rate": 6.543591797608964e-06, + "loss": 0.9424, + "step": 50665 + }, + { + "epoch": 3.74, + "learning_rate": 6.543217512436362e-06, + "loss": 1.0528, + "step": 50666 + }, + { + "epoch": 3.74, + "learning_rate": 6.542843232763447e-06, + "loss": 1.0492, + "step": 50667 + }, + { + "epoch": 3.74, + "learning_rate": 6.542468958590821e-06, + "loss": 0.8635, + "step": 50668 + }, + { + "epoch": 3.74, + "learning_rate": 6.542094689919075e-06, + "loss": 1.1007, + "step": 50669 + }, + { + "epoch": 3.74, + "learning_rate": 6.54172042674881e-06, + "loss": 1.0368, + "step": 50670 + }, + { + "epoch": 3.74, + "learning_rate": 6.541346169080617e-06, + "loss": 0.9663, + "step": 50671 + }, + { + "epoch": 3.74, + "learning_rate": 6.540971916915091e-06, + "loss": 1.0653, + "step": 50672 + }, + { + "epoch": 3.74, + "learning_rate": 6.540597670252832e-06, + "loss": 0.9684, + "step": 50673 + }, + { + "epoch": 3.74, + "learning_rate": 6.54022342909443e-06, + "loss": 0.9303, + "step": 50674 + }, + { + "epoch": 3.74, + "learning_rate": 6.539849193440485e-06, + "loss": 0.9918, + "step": 50675 + }, + { + "epoch": 3.74, + "learning_rate": 6.5394749632915874e-06, + "loss": 1.0137, + "step": 50676 + }, + { + "epoch": 3.74, + "learning_rate": 6.53910073864834e-06, + "loss": 1.0275, + "step": 50677 + }, + { + "epoch": 3.74, + "learning_rate": 6.53872651951133e-06, + "loss": 1.049, + "step": 50678 + }, + { + "epoch": 3.74, + "learning_rate": 6.538352305881159e-06, + "loss": 0.9886, + "step": 50679 + }, + { + "epoch": 3.74, + "learning_rate": 6.537978097758423e-06, + "loss": 1.09, + "step": 50680 + }, + { + "epoch": 3.74, + "learning_rate": 6.537603895143707e-06, + "loss": 0.9441, + "step": 50681 + }, + { + "epoch": 3.74, + "learning_rate": 6.537229698037621e-06, + "loss": 1.0434, + "step": 50682 + }, + { + "epoch": 3.74, + "learning_rate": 6.5368555064407445e-06, + "loss": 1.0417, + "step": 50683 + }, + { + "epoch": 3.75, + "learning_rate": 6.536481320353689e-06, + "loss": 1.0046, + "step": 50684 + }, + { + "epoch": 3.75, + "learning_rate": 6.536107139777037e-06, + "loss": 0.9537, + "step": 50685 + }, + { + "epoch": 3.75, + "learning_rate": 6.535732964711395e-06, + "loss": 1.0438, + "step": 50686 + }, + { + "epoch": 3.75, + "learning_rate": 6.535358795157344e-06, + "loss": 0.9782, + "step": 50687 + }, + { + "epoch": 3.75, + "learning_rate": 6.5349846311154954e-06, + "loss": 1.0351, + "step": 50688 + }, + { + "epoch": 3.75, + "learning_rate": 6.534610472586433e-06, + "loss": 0.9482, + "step": 50689 + }, + { + "epoch": 3.75, + "learning_rate": 6.5342363195707556e-06, + "loss": 0.9226, + "step": 50690 + }, + { + "epoch": 3.75, + "learning_rate": 6.533862172069061e-06, + "loss": 0.9999, + "step": 50691 + }, + { + "epoch": 3.75, + "learning_rate": 6.533488030081938e-06, + "loss": 1.1635, + "step": 50692 + }, + { + "epoch": 3.75, + "learning_rate": 6.533113893609989e-06, + "loss": 0.9286, + "step": 50693 + }, + { + "epoch": 3.75, + "learning_rate": 6.532739762653805e-06, + "loss": 0.9061, + "step": 50694 + }, + { + "epoch": 3.75, + "learning_rate": 6.5323656372139835e-06, + "loss": 1.094, + "step": 50695 + }, + { + "epoch": 3.75, + "learning_rate": 6.531991517291117e-06, + "loss": 0.9588, + "step": 50696 + }, + { + "epoch": 3.75, + "learning_rate": 6.531617402885804e-06, + "loss": 1.0515, + "step": 50697 + }, + { + "epoch": 3.75, + "learning_rate": 6.531243293998638e-06, + "loss": 1.187, + "step": 50698 + }, + { + "epoch": 3.75, + "learning_rate": 6.530869190630213e-06, + "loss": 1.0213, + "step": 50699 + }, + { + "epoch": 3.75, + "learning_rate": 6.530495092781126e-06, + "loss": 0.9937, + "step": 50700 + }, + { + "epoch": 3.75, + "learning_rate": 6.530121000451972e-06, + "loss": 1.0089, + "step": 50701 + }, + { + "epoch": 3.75, + "learning_rate": 6.529746913643347e-06, + "loss": 0.9882, + "step": 50702 + }, + { + "epoch": 3.75, + "learning_rate": 6.529372832355841e-06, + "loss": 1.0607, + "step": 50703 + }, + { + "epoch": 3.75, + "learning_rate": 6.528998756590057e-06, + "loss": 1.2066, + "step": 50704 + }, + { + "epoch": 3.75, + "learning_rate": 6.528624686346584e-06, + "loss": 1.0476, + "step": 50705 + }, + { + "epoch": 3.75, + "learning_rate": 6.528250621626022e-06, + "loss": 0.964, + "step": 50706 + }, + { + "epoch": 3.75, + "learning_rate": 6.5278765624289655e-06, + "loss": 1.0107, + "step": 50707 + }, + { + "epoch": 3.75, + "learning_rate": 6.527502508756001e-06, + "loss": 0.967, + "step": 50708 + }, + { + "epoch": 3.75, + "learning_rate": 6.527128460607737e-06, + "loss": 1.133, + "step": 50709 + }, + { + "epoch": 3.75, + "learning_rate": 6.5267544179847555e-06, + "loss": 1.0041, + "step": 50710 + }, + { + "epoch": 3.75, + "learning_rate": 6.526380380887664e-06, + "loss": 0.9048, + "step": 50711 + }, + { + "epoch": 3.75, + "learning_rate": 6.526006349317047e-06, + "loss": 1.074, + "step": 50712 + }, + { + "epoch": 3.75, + "learning_rate": 6.525632323273507e-06, + "loss": 0.9882, + "step": 50713 + }, + { + "epoch": 3.75, + "learning_rate": 6.525258302757634e-06, + "loss": 0.9049, + "step": 50714 + }, + { + "epoch": 3.75, + "learning_rate": 6.524884287770028e-06, + "loss": 0.9656, + "step": 50715 + }, + { + "epoch": 3.75, + "learning_rate": 6.5245102783112806e-06, + "loss": 1.0622, + "step": 50716 + }, + { + "epoch": 3.75, + "learning_rate": 6.524136274381986e-06, + "loss": 0.9872, + "step": 50717 + }, + { + "epoch": 3.75, + "learning_rate": 6.5237622759827435e-06, + "loss": 0.9917, + "step": 50718 + }, + { + "epoch": 3.75, + "learning_rate": 6.523388283114143e-06, + "loss": 0.9947, + "step": 50719 + }, + { + "epoch": 3.75, + "learning_rate": 6.5230142957767825e-06, + "loss": 1.0763, + "step": 50720 + }, + { + "epoch": 3.75, + "learning_rate": 6.5226403139712555e-06, + "loss": 0.9933, + "step": 50721 + }, + { + "epoch": 3.75, + "learning_rate": 6.52226633769816e-06, + "loss": 1.0855, + "step": 50722 + }, + { + "epoch": 3.75, + "learning_rate": 6.521892366958088e-06, + "loss": 0.9259, + "step": 50723 + }, + { + "epoch": 3.75, + "learning_rate": 6.521518401751636e-06, + "loss": 0.9242, + "step": 50724 + }, + { + "epoch": 3.75, + "learning_rate": 6.521144442079398e-06, + "loss": 1.0317, + "step": 50725 + }, + { + "epoch": 3.75, + "learning_rate": 6.52077048794197e-06, + "loss": 1.0738, + "step": 50726 + }, + { + "epoch": 3.75, + "learning_rate": 6.52039653933995e-06, + "loss": 0.9548, + "step": 50727 + }, + { + "epoch": 3.75, + "learning_rate": 6.520022596273922e-06, + "loss": 1.031, + "step": 50728 + }, + { + "epoch": 3.75, + "learning_rate": 6.519648658744494e-06, + "loss": 1.0816, + "step": 50729 + }, + { + "epoch": 3.75, + "learning_rate": 6.5192747267522495e-06, + "loss": 1.0968, + "step": 50730 + }, + { + "epoch": 3.75, + "learning_rate": 6.518900800297796e-06, + "loss": 0.9858, + "step": 50731 + }, + { + "epoch": 3.75, + "learning_rate": 6.518526879381715e-06, + "loss": 1.0312, + "step": 50732 + }, + { + "epoch": 3.75, + "learning_rate": 6.5181529640046135e-06, + "loss": 1.0687, + "step": 50733 + }, + { + "epoch": 3.75, + "learning_rate": 6.517779054167075e-06, + "loss": 0.9633, + "step": 50734 + }, + { + "epoch": 3.75, + "learning_rate": 6.517405149869707e-06, + "loss": 0.9583, + "step": 50735 + }, + { + "epoch": 3.75, + "learning_rate": 6.517031251113095e-06, + "loss": 1.0174, + "step": 50736 + }, + { + "epoch": 3.75, + "learning_rate": 6.516657357897833e-06, + "loss": 1.0742, + "step": 50737 + }, + { + "epoch": 3.75, + "learning_rate": 6.516283470224522e-06, + "loss": 1.1224, + "step": 50738 + }, + { + "epoch": 3.75, + "learning_rate": 6.515909588093754e-06, + "loss": 0.9747, + "step": 50739 + }, + { + "epoch": 3.75, + "learning_rate": 6.515535711506124e-06, + "loss": 0.9773, + "step": 50740 + }, + { + "epoch": 3.75, + "learning_rate": 6.515161840462226e-06, + "loss": 1.0353, + "step": 50741 + }, + { + "epoch": 3.75, + "learning_rate": 6.514787974962658e-06, + "loss": 1.0419, + "step": 50742 + }, + { + "epoch": 3.75, + "learning_rate": 6.51441411500801e-06, + "loss": 0.9998, + "step": 50743 + }, + { + "epoch": 3.75, + "learning_rate": 6.514040260598881e-06, + "loss": 1.0635, + "step": 50744 + }, + { + "epoch": 3.75, + "learning_rate": 6.5136664117358635e-06, + "loss": 1.0107, + "step": 50745 + }, + { + "epoch": 3.75, + "learning_rate": 6.513292568419552e-06, + "loss": 0.9531, + "step": 50746 + }, + { + "epoch": 3.75, + "learning_rate": 6.512918730650544e-06, + "loss": 0.928, + "step": 50747 + }, + { + "epoch": 3.75, + "learning_rate": 6.5125448984294295e-06, + "loss": 1.0663, + "step": 50748 + }, + { + "epoch": 3.75, + "learning_rate": 6.512171071756808e-06, + "loss": 0.9977, + "step": 50749 + }, + { + "epoch": 3.75, + "learning_rate": 6.5117972506332715e-06, + "loss": 0.9548, + "step": 50750 + }, + { + "epoch": 3.75, + "learning_rate": 6.511423435059416e-06, + "loss": 0.9716, + "step": 50751 + }, + { + "epoch": 3.75, + "learning_rate": 6.511049625035837e-06, + "loss": 0.838, + "step": 50752 + }, + { + "epoch": 3.75, + "learning_rate": 6.510675820563128e-06, + "loss": 0.9449, + "step": 50753 + }, + { + "epoch": 3.75, + "learning_rate": 6.5103020216418856e-06, + "loss": 1.0091, + "step": 50754 + }, + { + "epoch": 3.75, + "learning_rate": 6.509928228272699e-06, + "loss": 0.9522, + "step": 50755 + }, + { + "epoch": 3.75, + "learning_rate": 6.50955444045617e-06, + "loss": 1.127, + "step": 50756 + }, + { + "epoch": 3.75, + "learning_rate": 6.5091806581928866e-06, + "loss": 1.0312, + "step": 50757 + }, + { + "epoch": 3.75, + "learning_rate": 6.508806881483449e-06, + "loss": 0.9773, + "step": 50758 + }, + { + "epoch": 3.75, + "learning_rate": 6.5084331103284475e-06, + "loss": 1.089, + "step": 50759 + }, + { + "epoch": 3.75, + "learning_rate": 6.508059344728481e-06, + "loss": 1.0761, + "step": 50760 + }, + { + "epoch": 3.75, + "learning_rate": 6.507685584684141e-06, + "loss": 0.9421, + "step": 50761 + }, + { + "epoch": 3.75, + "learning_rate": 6.507311830196024e-06, + "loss": 1.0442, + "step": 50762 + }, + { + "epoch": 3.75, + "learning_rate": 6.506938081264724e-06, + "loss": 0.9187, + "step": 50763 + }, + { + "epoch": 3.75, + "learning_rate": 6.506564337890835e-06, + "loss": 0.9891, + "step": 50764 + }, + { + "epoch": 3.75, + "learning_rate": 6.506190600074953e-06, + "loss": 1.0748, + "step": 50765 + }, + { + "epoch": 3.75, + "learning_rate": 6.505816867817669e-06, + "loss": 1.0361, + "step": 50766 + }, + { + "epoch": 3.75, + "learning_rate": 6.505443141119585e-06, + "loss": 0.9655, + "step": 50767 + }, + { + "epoch": 3.75, + "learning_rate": 6.505069419981287e-06, + "loss": 1.0531, + "step": 50768 + }, + { + "epoch": 3.75, + "learning_rate": 6.504695704403375e-06, + "loss": 1.0054, + "step": 50769 + }, + { + "epoch": 3.75, + "learning_rate": 6.504321994386442e-06, + "loss": 0.9611, + "step": 50770 + }, + { + "epoch": 3.75, + "learning_rate": 6.503948289931084e-06, + "loss": 1.0501, + "step": 50771 + }, + { + "epoch": 3.75, + "learning_rate": 6.503574591037896e-06, + "loss": 0.9811, + "step": 50772 + }, + { + "epoch": 3.75, + "learning_rate": 6.503200897707465e-06, + "loss": 0.9886, + "step": 50773 + }, + { + "epoch": 3.75, + "learning_rate": 6.502827209940397e-06, + "loss": 1.0525, + "step": 50774 + }, + { + "epoch": 3.75, + "learning_rate": 6.502453527737273e-06, + "loss": 1.0372, + "step": 50775 + }, + { + "epoch": 3.75, + "learning_rate": 6.502079851098705e-06, + "loss": 1.0932, + "step": 50776 + }, + { + "epoch": 3.75, + "learning_rate": 6.501706180025269e-06, + "loss": 0.9872, + "step": 50777 + }, + { + "epoch": 3.75, + "learning_rate": 6.501332514517577e-06, + "loss": 1.0009, + "step": 50778 + }, + { + "epoch": 3.75, + "learning_rate": 6.500958854576207e-06, + "loss": 1.0474, + "step": 50779 + }, + { + "epoch": 3.75, + "learning_rate": 6.500585200201768e-06, + "loss": 0.9534, + "step": 50780 + }, + { + "epoch": 3.75, + "learning_rate": 6.500211551394842e-06, + "loss": 1.0066, + "step": 50781 + }, + { + "epoch": 3.75, + "learning_rate": 6.499837908156034e-06, + "loss": 0.9512, + "step": 50782 + }, + { + "epoch": 3.75, + "learning_rate": 6.499464270485935e-06, + "loss": 0.982, + "step": 50783 + }, + { + "epoch": 3.75, + "learning_rate": 6.4990906383851335e-06, + "loss": 1.0818, + "step": 50784 + }, + { + "epoch": 3.75, + "learning_rate": 6.498717011854232e-06, + "loss": 0.976, + "step": 50785 + }, + { + "epoch": 3.75, + "learning_rate": 6.498343390893819e-06, + "loss": 1.0029, + "step": 50786 + }, + { + "epoch": 3.75, + "learning_rate": 6.497969775504493e-06, + "loss": 1.0489, + "step": 50787 + }, + { + "epoch": 3.75, + "learning_rate": 6.497596165686846e-06, + "loss": 0.9811, + "step": 50788 + }, + { + "epoch": 3.75, + "learning_rate": 6.4972225614414755e-06, + "loss": 0.9158, + "step": 50789 + }, + { + "epoch": 3.75, + "learning_rate": 6.496848962768971e-06, + "loss": 0.9834, + "step": 50790 + }, + { + "epoch": 3.75, + "learning_rate": 6.496475369669932e-06, + "loss": 1.094, + "step": 50791 + }, + { + "epoch": 3.75, + "learning_rate": 6.4961017821449505e-06, + "loss": 1.0371, + "step": 50792 + }, + { + "epoch": 3.75, + "learning_rate": 6.4957282001946195e-06, + "loss": 1.1247, + "step": 50793 + }, + { + "epoch": 3.75, + "learning_rate": 6.4953546238195365e-06, + "loss": 1.0214, + "step": 50794 + }, + { + "epoch": 3.75, + "learning_rate": 6.494981053020292e-06, + "loss": 0.9433, + "step": 50795 + }, + { + "epoch": 3.75, + "learning_rate": 6.494607487797484e-06, + "loss": 0.9198, + "step": 50796 + }, + { + "epoch": 3.75, + "learning_rate": 6.494233928151704e-06, + "loss": 1.0546, + "step": 50797 + }, + { + "epoch": 3.75, + "learning_rate": 6.493860374083549e-06, + "loss": 0.8653, + "step": 50798 + }, + { + "epoch": 3.75, + "learning_rate": 6.493486825593611e-06, + "loss": 1.0461, + "step": 50799 + }, + { + "epoch": 3.75, + "learning_rate": 6.493113282682489e-06, + "loss": 1.0088, + "step": 50800 + }, + { + "epoch": 3.75, + "learning_rate": 6.492739745350771e-06, + "loss": 0.9802, + "step": 50801 + }, + { + "epoch": 3.75, + "learning_rate": 6.492366213599053e-06, + "loss": 1.0199, + "step": 50802 + }, + { + "epoch": 3.75, + "learning_rate": 6.4919926874279305e-06, + "loss": 1.1285, + "step": 50803 + }, + { + "epoch": 3.75, + "learning_rate": 6.491619166837997e-06, + "loss": 1.0701, + "step": 50804 + }, + { + "epoch": 3.75, + "learning_rate": 6.491245651829849e-06, + "loss": 0.9081, + "step": 50805 + }, + { + "epoch": 3.75, + "learning_rate": 6.490872142404075e-06, + "loss": 1.0127, + "step": 50806 + }, + { + "epoch": 3.75, + "learning_rate": 6.490498638561278e-06, + "loss": 0.9846, + "step": 50807 + }, + { + "epoch": 3.75, + "learning_rate": 6.4901251403020435e-06, + "loss": 1.0065, + "step": 50808 + }, + { + "epoch": 3.75, + "learning_rate": 6.489751647626972e-06, + "loss": 0.98, + "step": 50809 + }, + { + "epoch": 3.75, + "learning_rate": 6.489378160536657e-06, + "loss": 0.8129, + "step": 50810 + }, + { + "epoch": 3.75, + "learning_rate": 6.489004679031687e-06, + "loss": 0.9511, + "step": 50811 + }, + { + "epoch": 3.75, + "learning_rate": 6.488631203112663e-06, + "loss": 0.9356, + "step": 50812 + }, + { + "epoch": 3.75, + "learning_rate": 6.488257732780175e-06, + "loss": 1.0788, + "step": 50813 + }, + { + "epoch": 3.75, + "learning_rate": 6.487884268034821e-06, + "loss": 0.9324, + "step": 50814 + }, + { + "epoch": 3.75, + "learning_rate": 6.48751080887719e-06, + "loss": 1.0069, + "step": 50815 + }, + { + "epoch": 3.75, + "learning_rate": 6.487137355307881e-06, + "loss": 0.9591, + "step": 50816 + }, + { + "epoch": 3.75, + "learning_rate": 6.486763907327485e-06, + "loss": 0.9987, + "step": 50817 + }, + { + "epoch": 3.75, + "learning_rate": 6.486390464936599e-06, + "loss": 0.8694, + "step": 50818 + }, + { + "epoch": 3.75, + "learning_rate": 6.486017028135817e-06, + "loss": 0.9923, + "step": 50819 + }, + { + "epoch": 3.76, + "learning_rate": 6.485643596925726e-06, + "loss": 1.0296, + "step": 50820 + }, + { + "epoch": 3.76, + "learning_rate": 6.4852701713069324e-06, + "loss": 0.9671, + "step": 50821 + }, + { + "epoch": 3.76, + "learning_rate": 6.4848967512800165e-06, + "loss": 0.9639, + "step": 50822 + }, + { + "epoch": 3.76, + "learning_rate": 6.4845233368455865e-06, + "loss": 1.0144, + "step": 50823 + }, + { + "epoch": 3.76, + "learning_rate": 6.484149928004222e-06, + "loss": 1.0325, + "step": 50824 + }, + { + "epoch": 3.76, + "learning_rate": 6.483776524756533e-06, + "loss": 0.9911, + "step": 50825 + }, + { + "epoch": 3.76, + "learning_rate": 6.483403127103099e-06, + "loss": 1.0256, + "step": 50826 + }, + { + "epoch": 3.76, + "learning_rate": 6.483029735044524e-06, + "loss": 0.9976, + "step": 50827 + }, + { + "epoch": 3.76, + "learning_rate": 6.482656348581398e-06, + "loss": 0.9604, + "step": 50828 + }, + { + "epoch": 3.76, + "learning_rate": 6.4822829677143125e-06, + "loss": 1.0522, + "step": 50829 + }, + { + "epoch": 3.76, + "learning_rate": 6.481909592443868e-06, + "loss": 0.9129, + "step": 50830 + }, + { + "epoch": 3.76, + "learning_rate": 6.481536222770653e-06, + "loss": 1.0638, + "step": 50831 + }, + { + "epoch": 3.76, + "learning_rate": 6.481162858695265e-06, + "loss": 0.9429, + "step": 50832 + }, + { + "epoch": 3.76, + "learning_rate": 6.480789500218293e-06, + "loss": 0.955, + "step": 50833 + }, + { + "epoch": 3.76, + "learning_rate": 6.480416147340338e-06, + "loss": 1.0572, + "step": 50834 + }, + { + "epoch": 3.76, + "learning_rate": 6.4800428000619885e-06, + "loss": 0.827, + "step": 50835 + }, + { + "epoch": 3.76, + "learning_rate": 6.479669458383841e-06, + "loss": 0.9645, + "step": 50836 + }, + { + "epoch": 3.76, + "learning_rate": 6.47929612230649e-06, + "loss": 1.1337, + "step": 50837 + }, + { + "epoch": 3.76, + "learning_rate": 6.478922791830529e-06, + "loss": 1.1477, + "step": 50838 + }, + { + "epoch": 3.76, + "learning_rate": 6.478549466956552e-06, + "loss": 0.9327, + "step": 50839 + }, + { + "epoch": 3.76, + "learning_rate": 6.478176147685149e-06, + "loss": 0.9823, + "step": 50840 + }, + { + "epoch": 3.76, + "learning_rate": 6.4778028340169194e-06, + "loss": 1.052, + "step": 50841 + }, + { + "epoch": 3.76, + "learning_rate": 6.477429525952454e-06, + "loss": 0.9645, + "step": 50842 + }, + { + "epoch": 3.76, + "learning_rate": 6.477056223492351e-06, + "loss": 0.9703, + "step": 50843 + }, + { + "epoch": 3.76, + "learning_rate": 6.476682926637197e-06, + "loss": 1.0864, + "step": 50844 + }, + { + "epoch": 3.76, + "learning_rate": 6.476309635387594e-06, + "loss": 0.9929, + "step": 50845 + }, + { + "epoch": 3.76, + "learning_rate": 6.4759363497441275e-06, + "loss": 0.9344, + "step": 50846 + }, + { + "epoch": 3.76, + "learning_rate": 6.475563069707402e-06, + "loss": 1.0305, + "step": 50847 + }, + { + "epoch": 3.76, + "learning_rate": 6.475189795278002e-06, + "loss": 0.9759, + "step": 50848 + }, + { + "epoch": 3.76, + "learning_rate": 6.474816526456523e-06, + "loss": 0.9708, + "step": 50849 + }, + { + "epoch": 3.76, + "learning_rate": 6.474443263243563e-06, + "loss": 1.0433, + "step": 50850 + }, + { + "epoch": 3.76, + "learning_rate": 6.47407000563971e-06, + "loss": 1.0392, + "step": 50851 + }, + { + "epoch": 3.76, + "learning_rate": 6.473696753645564e-06, + "loss": 0.9964, + "step": 50852 + }, + { + "epoch": 3.76, + "learning_rate": 6.473323507261715e-06, + "loss": 1.0088, + "step": 50853 + }, + { + "epoch": 3.76, + "learning_rate": 6.472950266488758e-06, + "loss": 0.9136, + "step": 50854 + }, + { + "epoch": 3.76, + "learning_rate": 6.472577031327285e-06, + "loss": 0.9719, + "step": 50855 + }, + { + "epoch": 3.76, + "learning_rate": 6.472203801777894e-06, + "loss": 1.0046, + "step": 50856 + }, + { + "epoch": 3.76, + "learning_rate": 6.471830577841177e-06, + "loss": 1.0013, + "step": 50857 + }, + { + "epoch": 3.76, + "learning_rate": 6.471457359517724e-06, + "loss": 1.0154, + "step": 50858 + }, + { + "epoch": 3.76, + "learning_rate": 6.471084146808135e-06, + "loss": 0.9679, + "step": 50859 + }, + { + "epoch": 3.76, + "learning_rate": 6.470710939712997e-06, + "loss": 1.029, + "step": 50860 + }, + { + "epoch": 3.76, + "learning_rate": 6.47033773823291e-06, + "loss": 1.0449, + "step": 50861 + }, + { + "epoch": 3.76, + "learning_rate": 6.469964542368463e-06, + "loss": 0.9505, + "step": 50862 + }, + { + "epoch": 3.76, + "learning_rate": 6.469591352120255e-06, + "loss": 1.0731, + "step": 50863 + }, + { + "epoch": 3.76, + "learning_rate": 6.4692181674888746e-06, + "loss": 1.0346, + "step": 50864 + }, + { + "epoch": 3.76, + "learning_rate": 6.468844988474919e-06, + "loss": 0.956, + "step": 50865 + }, + { + "epoch": 3.76, + "learning_rate": 6.468471815078982e-06, + "loss": 0.9841, + "step": 50866 + }, + { + "epoch": 3.76, + "learning_rate": 6.46809864730165e-06, + "loss": 0.9531, + "step": 50867 + }, + { + "epoch": 3.76, + "learning_rate": 6.467725485143529e-06, + "loss": 0.9945, + "step": 50868 + }, + { + "epoch": 3.76, + "learning_rate": 6.4673523286052e-06, + "loss": 1.0204, + "step": 50869 + }, + { + "epoch": 3.76, + "learning_rate": 6.466979177687269e-06, + "loss": 0.9682, + "step": 50870 + }, + { + "epoch": 3.76, + "learning_rate": 6.466606032390317e-06, + "loss": 0.9228, + "step": 50871 + }, + { + "epoch": 3.76, + "learning_rate": 6.466232892714951e-06, + "loss": 0.993, + "step": 50872 + }, + { + "epoch": 3.76, + "learning_rate": 6.465859758661751e-06, + "loss": 0.9923, + "step": 50873 + }, + { + "epoch": 3.76, + "learning_rate": 6.465486630231325e-06, + "loss": 1.0161, + "step": 50874 + }, + { + "epoch": 3.76, + "learning_rate": 6.465113507424257e-06, + "loss": 0.9939, + "step": 50875 + }, + { + "epoch": 3.76, + "learning_rate": 6.46474039024114e-06, + "loss": 0.9591, + "step": 50876 + }, + { + "epoch": 3.76, + "learning_rate": 6.464367278682572e-06, + "loss": 1.0027, + "step": 50877 + }, + { + "epoch": 3.76, + "learning_rate": 6.463994172749144e-06, + "loss": 0.998, + "step": 50878 + }, + { + "epoch": 3.76, + "learning_rate": 6.463621072441452e-06, + "loss": 1.0478, + "step": 50879 + }, + { + "epoch": 3.76, + "learning_rate": 6.463247977760087e-06, + "loss": 0.9275, + "step": 50880 + }, + { + "epoch": 3.76, + "learning_rate": 6.462874888705646e-06, + "loss": 1.071, + "step": 50881 + }, + { + "epoch": 3.76, + "learning_rate": 6.462501805278717e-06, + "loss": 0.9327, + "step": 50882 + }, + { + "epoch": 3.76, + "learning_rate": 6.462128727479899e-06, + "loss": 0.9498, + "step": 50883 + }, + { + "epoch": 3.76, + "learning_rate": 6.461755655309784e-06, + "loss": 1.0129, + "step": 50884 + }, + { + "epoch": 3.76, + "learning_rate": 6.461382588768965e-06, + "loss": 1.0071, + "step": 50885 + }, + { + "epoch": 3.76, + "learning_rate": 6.461009527858038e-06, + "loss": 1.0895, + "step": 50886 + }, + { + "epoch": 3.76, + "learning_rate": 6.460636472577589e-06, + "loss": 0.9195, + "step": 50887 + }, + { + "epoch": 3.76, + "learning_rate": 6.460263422928224e-06, + "loss": 0.9373, + "step": 50888 + }, + { + "epoch": 3.76, + "learning_rate": 6.459890378910521e-06, + "loss": 1.006, + "step": 50889 + }, + { + "epoch": 3.76, + "learning_rate": 6.459517340525088e-06, + "loss": 1.0139, + "step": 50890 + }, + { + "epoch": 3.76, + "learning_rate": 6.4591443077725055e-06, + "loss": 1.0333, + "step": 50891 + }, + { + "epoch": 3.76, + "learning_rate": 6.458771280653382e-06, + "loss": 0.8772, + "step": 50892 + }, + { + "epoch": 3.76, + "learning_rate": 6.458398259168295e-06, + "loss": 1.0401, + "step": 50893 + }, + { + "epoch": 3.76, + "learning_rate": 6.4580252433178525e-06, + "loss": 1.0575, + "step": 50894 + }, + { + "epoch": 3.76, + "learning_rate": 6.457652233102638e-06, + "loss": 0.9195, + "step": 50895 + }, + { + "epoch": 3.76, + "learning_rate": 6.4572792285232474e-06, + "loss": 0.9344, + "step": 50896 + }, + { + "epoch": 3.76, + "learning_rate": 6.456906229580277e-06, + "loss": 0.9299, + "step": 50897 + }, + { + "epoch": 3.76, + "learning_rate": 6.456533236274316e-06, + "loss": 0.9465, + "step": 50898 + }, + { + "epoch": 3.76, + "learning_rate": 6.4561602486059605e-06, + "loss": 0.9643, + "step": 50899 + }, + { + "epoch": 3.76, + "learning_rate": 6.455787266575802e-06, + "loss": 0.9752, + "step": 50900 + }, + { + "epoch": 3.76, + "learning_rate": 6.455414290184439e-06, + "loss": 1.0615, + "step": 50901 + }, + { + "epoch": 3.76, + "learning_rate": 6.455041319432457e-06, + "loss": 0.9907, + "step": 50902 + }, + { + "epoch": 3.76, + "learning_rate": 6.454668354320457e-06, + "loss": 0.9574, + "step": 50903 + }, + { + "epoch": 3.76, + "learning_rate": 6.454295394849027e-06, + "loss": 0.9553, + "step": 50904 + }, + { + "epoch": 3.76, + "learning_rate": 6.4539224410187615e-06, + "loss": 0.9371, + "step": 50905 + }, + { + "epoch": 3.76, + "learning_rate": 6.453549492830258e-06, + "loss": 0.914, + "step": 50906 + }, + { + "epoch": 3.76, + "learning_rate": 6.453176550284102e-06, + "loss": 1.0654, + "step": 50907 + }, + { + "epoch": 3.76, + "learning_rate": 6.452803613380894e-06, + "loss": 0.9186, + "step": 50908 + }, + { + "epoch": 3.76, + "learning_rate": 6.452430682121224e-06, + "loss": 1.044, + "step": 50909 + }, + { + "epoch": 3.76, + "learning_rate": 6.452057756505687e-06, + "loss": 0.9166, + "step": 50910 + }, + { + "epoch": 3.76, + "learning_rate": 6.451684836534874e-06, + "loss": 1.0462, + "step": 50911 + }, + { + "epoch": 3.76, + "learning_rate": 6.451311922209381e-06, + "loss": 0.9654, + "step": 50912 + }, + { + "epoch": 3.76, + "learning_rate": 6.4509390135298025e-06, + "loss": 0.9225, + "step": 50913 + }, + { + "epoch": 3.76, + "learning_rate": 6.450566110496724e-06, + "loss": 0.9907, + "step": 50914 + }, + { + "epoch": 3.76, + "learning_rate": 6.450193213110748e-06, + "loss": 1.0443, + "step": 50915 + }, + { + "epoch": 3.76, + "learning_rate": 6.449820321372459e-06, + "loss": 1.0234, + "step": 50916 + }, + { + "epoch": 3.76, + "learning_rate": 6.449447435282461e-06, + "loss": 1.0141, + "step": 50917 + }, + { + "epoch": 3.76, + "learning_rate": 6.4490745548413346e-06, + "loss": 1.0468, + "step": 50918 + }, + { + "epoch": 3.76, + "learning_rate": 6.448701680049687e-06, + "loss": 1.0721, + "step": 50919 + }, + { + "epoch": 3.76, + "learning_rate": 6.448328810908098e-06, + "loss": 1.0072, + "step": 50920 + }, + { + "epoch": 3.76, + "learning_rate": 6.4479559474171725e-06, + "loss": 0.9028, + "step": 50921 + }, + { + "epoch": 3.76, + "learning_rate": 6.447583089577496e-06, + "loss": 0.8984, + "step": 50922 + }, + { + "epoch": 3.76, + "learning_rate": 6.447210237389663e-06, + "loss": 1.0074, + "step": 50923 + }, + { + "epoch": 3.76, + "learning_rate": 6.4468373908542685e-06, + "loss": 0.9132, + "step": 50924 + }, + { + "epoch": 3.76, + "learning_rate": 6.446464549971903e-06, + "loss": 1.0654, + "step": 50925 + }, + { + "epoch": 3.76, + "learning_rate": 6.446091714743165e-06, + "loss": 1.0647, + "step": 50926 + }, + { + "epoch": 3.76, + "learning_rate": 6.445718885168641e-06, + "loss": 0.9701, + "step": 50927 + }, + { + "epoch": 3.76, + "learning_rate": 6.4453460612489295e-06, + "loss": 0.9846, + "step": 50928 + }, + { + "epoch": 3.76, + "learning_rate": 6.44497324298462e-06, + "loss": 1.0338, + "step": 50929 + }, + { + "epoch": 3.76, + "learning_rate": 6.444600430376309e-06, + "loss": 1.1038, + "step": 50930 + }, + { + "epoch": 3.76, + "learning_rate": 6.444227623424587e-06, + "loss": 1.1573, + "step": 50931 + }, + { + "epoch": 3.76, + "learning_rate": 6.443854822130047e-06, + "loss": 0.92, + "step": 50932 + }, + { + "epoch": 3.76, + "learning_rate": 6.443482026493287e-06, + "loss": 1.0871, + "step": 50933 + }, + { + "epoch": 3.76, + "learning_rate": 6.4431092365148895e-06, + "loss": 0.9999, + "step": 50934 + }, + { + "epoch": 3.76, + "learning_rate": 6.442736452195462e-06, + "loss": 1.0871, + "step": 50935 + }, + { + "epoch": 3.76, + "learning_rate": 6.442363673535581e-06, + "loss": 0.9732, + "step": 50936 + }, + { + "epoch": 3.76, + "learning_rate": 6.4419909005358575e-06, + "loss": 0.9726, + "step": 50937 + }, + { + "epoch": 3.76, + "learning_rate": 6.441618133196867e-06, + "loss": 1.0488, + "step": 50938 + }, + { + "epoch": 3.76, + "learning_rate": 6.44124537151922e-06, + "loss": 1.0501, + "step": 50939 + }, + { + "epoch": 3.76, + "learning_rate": 6.440872615503493e-06, + "loss": 0.9395, + "step": 50940 + }, + { + "epoch": 3.76, + "learning_rate": 6.440499865150293e-06, + "loss": 1.1026, + "step": 50941 + }, + { + "epoch": 3.76, + "learning_rate": 6.4401271204602045e-06, + "loss": 0.9184, + "step": 50942 + }, + { + "epoch": 3.76, + "learning_rate": 6.43975438143382e-06, + "loss": 1.0473, + "step": 50943 + }, + { + "epoch": 3.76, + "learning_rate": 6.439381648071739e-06, + "loss": 1.0019, + "step": 50944 + }, + { + "epoch": 3.76, + "learning_rate": 6.439008920374548e-06, + "loss": 1.0802, + "step": 50945 + }, + { + "epoch": 3.76, + "learning_rate": 6.438636198342845e-06, + "loss": 1.1178, + "step": 50946 + }, + { + "epoch": 3.76, + "learning_rate": 6.438263481977219e-06, + "loss": 0.9588, + "step": 50947 + }, + { + "epoch": 3.76, + "learning_rate": 6.437890771278267e-06, + "loss": 0.926, + "step": 50948 + }, + { + "epoch": 3.76, + "learning_rate": 6.437518066246577e-06, + "loss": 1.0991, + "step": 50949 + }, + { + "epoch": 3.76, + "learning_rate": 6.4371453668827474e-06, + "loss": 0.9674, + "step": 50950 + }, + { + "epoch": 3.76, + "learning_rate": 6.436772673187369e-06, + "loss": 1.0148, + "step": 50951 + }, + { + "epoch": 3.76, + "learning_rate": 6.436399985161033e-06, + "loss": 0.9861, + "step": 50952 + }, + { + "epoch": 3.76, + "learning_rate": 6.436027302804333e-06, + "loss": 0.9411, + "step": 50953 + }, + { + "epoch": 3.76, + "learning_rate": 6.435654626117862e-06, + "loss": 0.9316, + "step": 50954 + }, + { + "epoch": 3.77, + "learning_rate": 6.4352819551022165e-06, + "loss": 1.049, + "step": 50955 + }, + { + "epoch": 3.77, + "learning_rate": 6.434909289757982e-06, + "loss": 1.0118, + "step": 50956 + }, + { + "epoch": 3.77, + "learning_rate": 6.43453663008576e-06, + "loss": 1.066, + "step": 50957 + }, + { + "epoch": 3.77, + "learning_rate": 6.434163976086136e-06, + "loss": 1.0334, + "step": 50958 + }, + { + "epoch": 3.77, + "learning_rate": 6.433791327759709e-06, + "loss": 0.9374, + "step": 50959 + }, + { + "epoch": 3.77, + "learning_rate": 6.433418685107071e-06, + "loss": 1.0789, + "step": 50960 + }, + { + "epoch": 3.77, + "learning_rate": 6.433046048128806e-06, + "loss": 0.9972, + "step": 50961 + }, + { + "epoch": 3.77, + "learning_rate": 6.43267341682552e-06, + "loss": 0.9519, + "step": 50962 + }, + { + "epoch": 3.77, + "learning_rate": 6.432300791197794e-06, + "loss": 0.9957, + "step": 50963 + }, + { + "epoch": 3.77, + "learning_rate": 6.431928171246231e-06, + "loss": 1.0115, + "step": 50964 + }, + { + "epoch": 3.77, + "learning_rate": 6.431555556971416e-06, + "loss": 0.9497, + "step": 50965 + }, + { + "epoch": 3.77, + "learning_rate": 6.431182948373949e-06, + "loss": 1.0264, + "step": 50966 + }, + { + "epoch": 3.77, + "learning_rate": 6.430810345454415e-06, + "loss": 0.9379, + "step": 50967 + }, + { + "epoch": 3.77, + "learning_rate": 6.430437748213412e-06, + "loss": 0.9536, + "step": 50968 + }, + { + "epoch": 3.77, + "learning_rate": 6.4300651566515325e-06, + "loss": 1.051, + "step": 50969 + }, + { + "epoch": 3.77, + "learning_rate": 6.4296925707693655e-06, + "loss": 1.1016, + "step": 50970 + }, + { + "epoch": 3.77, + "learning_rate": 6.429319990567508e-06, + "loss": 1.1133, + "step": 50971 + }, + { + "epoch": 3.77, + "learning_rate": 6.428947416046552e-06, + "loss": 1.1349, + "step": 50972 + }, + { + "epoch": 3.77, + "learning_rate": 6.42857484720709e-06, + "loss": 1.1046, + "step": 50973 + }, + { + "epoch": 3.77, + "learning_rate": 6.4282022840497115e-06, + "loss": 1.0816, + "step": 50974 + }, + { + "epoch": 3.77, + "learning_rate": 6.427829726575015e-06, + "loss": 1.0236, + "step": 50975 + }, + { + "epoch": 3.77, + "learning_rate": 6.427457174783589e-06, + "loss": 1.0162, + "step": 50976 + }, + { + "epoch": 3.77, + "learning_rate": 6.427084628676028e-06, + "loss": 1.0336, + "step": 50977 + }, + { + "epoch": 3.77, + "learning_rate": 6.426712088252926e-06, + "loss": 1.0496, + "step": 50978 + }, + { + "epoch": 3.77, + "learning_rate": 6.426339553514869e-06, + "loss": 1.05, + "step": 50979 + }, + { + "epoch": 3.77, + "learning_rate": 6.42596702446246e-06, + "loss": 1.0147, + "step": 50980 + }, + { + "epoch": 3.77, + "learning_rate": 6.425594501096282e-06, + "loss": 1.0679, + "step": 50981 + }, + { + "epoch": 3.77, + "learning_rate": 6.425221983416937e-06, + "loss": 1.0672, + "step": 50982 + }, + { + "epoch": 3.77, + "learning_rate": 6.424849471425007e-06, + "loss": 1.0588, + "step": 50983 + }, + { + "epoch": 3.77, + "learning_rate": 6.4244769651210955e-06, + "loss": 1.0397, + "step": 50984 + }, + { + "epoch": 3.77, + "learning_rate": 6.424104464505784e-06, + "loss": 0.9133, + "step": 50985 + }, + { + "epoch": 3.77, + "learning_rate": 6.423731969579677e-06, + "loss": 1.05, + "step": 50986 + }, + { + "epoch": 3.77, + "learning_rate": 6.423359480343361e-06, + "loss": 0.8761, + "step": 50987 + }, + { + "epoch": 3.77, + "learning_rate": 6.422986996797425e-06, + "loss": 0.9863, + "step": 50988 + }, + { + "epoch": 3.77, + "learning_rate": 6.422614518942469e-06, + "loss": 0.9977, + "step": 50989 + }, + { + "epoch": 3.77, + "learning_rate": 6.422242046779078e-06, + "loss": 0.9212, + "step": 50990 + }, + { + "epoch": 3.77, + "learning_rate": 6.421869580307852e-06, + "loss": 0.8329, + "step": 50991 + }, + { + "epoch": 3.77, + "learning_rate": 6.421497119529378e-06, + "loss": 1.0355, + "step": 50992 + }, + { + "epoch": 3.77, + "learning_rate": 6.4211246644442535e-06, + "loss": 0.9889, + "step": 50993 + }, + { + "epoch": 3.77, + "learning_rate": 6.420752215053065e-06, + "loss": 0.9497, + "step": 50994 + }, + { + "epoch": 3.77, + "learning_rate": 6.420379771356413e-06, + "loss": 0.9901, + "step": 50995 + }, + { + "epoch": 3.77, + "learning_rate": 6.420007333354882e-06, + "loss": 1.0352, + "step": 50996 + }, + { + "epoch": 3.77, + "learning_rate": 6.41963490104907e-06, + "loss": 1.0398, + "step": 50997 + }, + { + "epoch": 3.77, + "learning_rate": 6.419262474439568e-06, + "loss": 1.0496, + "step": 50998 + }, + { + "epoch": 3.77, + "learning_rate": 6.418890053526966e-06, + "loss": 0.9582, + "step": 50999 + }, + { + "epoch": 3.77, + "learning_rate": 6.418517638311861e-06, + "loss": 1.0465, + "step": 51000 + }, + { + "epoch": 3.77, + "learning_rate": 6.418145228794841e-06, + "loss": 1.0165, + "step": 51001 + }, + { + "epoch": 3.77, + "learning_rate": 6.417772824976502e-06, + "loss": 0.9301, + "step": 51002 + }, + { + "epoch": 3.77, + "learning_rate": 6.417400426857434e-06, + "loss": 1.0257, + "step": 51003 + }, + { + "epoch": 3.77, + "learning_rate": 6.417028034438233e-06, + "loss": 0.9356, + "step": 51004 + }, + { + "epoch": 3.77, + "learning_rate": 6.416655647719487e-06, + "loss": 0.9849, + "step": 51005 + }, + { + "epoch": 3.77, + "learning_rate": 6.416283266701793e-06, + "loss": 0.9652, + "step": 51006 + }, + { + "epoch": 3.77, + "learning_rate": 6.415910891385742e-06, + "loss": 0.9809, + "step": 51007 + }, + { + "epoch": 3.77, + "learning_rate": 6.415538521771919e-06, + "loss": 0.9309, + "step": 51008 + }, + { + "epoch": 3.77, + "learning_rate": 6.41516615786093e-06, + "loss": 0.9758, + "step": 51009 + }, + { + "epoch": 3.77, + "learning_rate": 6.414793799653356e-06, + "loss": 1.0624, + "step": 51010 + }, + { + "epoch": 3.77, + "learning_rate": 6.414421447149796e-06, + "loss": 0.9827, + "step": 51011 + }, + { + "epoch": 3.77, + "learning_rate": 6.414049100350837e-06, + "loss": 0.9587, + "step": 51012 + }, + { + "epoch": 3.77, + "learning_rate": 6.413676759257079e-06, + "loss": 1.0536, + "step": 51013 + }, + { + "epoch": 3.77, + "learning_rate": 6.413304423869106e-06, + "loss": 0.9712, + "step": 51014 + }, + { + "epoch": 3.77, + "learning_rate": 6.412932094187516e-06, + "loss": 0.9418, + "step": 51015 + }, + { + "epoch": 3.77, + "learning_rate": 6.4125597702129015e-06, + "loss": 1.0275, + "step": 51016 + }, + { + "epoch": 3.77, + "learning_rate": 6.412187451945851e-06, + "loss": 1.0308, + "step": 51017 + }, + { + "epoch": 3.77, + "learning_rate": 6.411815139386959e-06, + "loss": 0.9824, + "step": 51018 + }, + { + "epoch": 3.77, + "learning_rate": 6.411442832536817e-06, + "loss": 0.9262, + "step": 51019 + }, + { + "epoch": 3.77, + "learning_rate": 6.4110705313960195e-06, + "loss": 1.1039, + "step": 51020 + }, + { + "epoch": 3.77, + "learning_rate": 6.410698235965156e-06, + "loss": 0.9269, + "step": 51021 + }, + { + "epoch": 3.77, + "learning_rate": 6.410325946244823e-06, + "loss": 0.9253, + "step": 51022 + }, + { + "epoch": 3.77, + "learning_rate": 6.409953662235608e-06, + "loss": 1.0654, + "step": 51023 + }, + { + "epoch": 3.77, + "learning_rate": 6.409581383938106e-06, + "loss": 1.0296, + "step": 51024 + }, + { + "epoch": 3.77, + "learning_rate": 6.409209111352911e-06, + "loss": 0.9413, + "step": 51025 + }, + { + "epoch": 3.77, + "learning_rate": 6.408836844480607e-06, + "loss": 1.0705, + "step": 51026 + }, + { + "epoch": 3.77, + "learning_rate": 6.4084645833217985e-06, + "loss": 0.8862, + "step": 51027 + }, + { + "epoch": 3.77, + "learning_rate": 6.408092327877065e-06, + "loss": 0.9966, + "step": 51028 + }, + { + "epoch": 3.77, + "learning_rate": 6.407720078147013e-06, + "loss": 1.0016, + "step": 51029 + }, + { + "epoch": 3.77, + "learning_rate": 6.407347834132219e-06, + "loss": 1.0619, + "step": 51030 + }, + { + "epoch": 3.77, + "learning_rate": 6.406975595833289e-06, + "loss": 1.0303, + "step": 51031 + }, + { + "epoch": 3.77, + "learning_rate": 6.406603363250805e-06, + "loss": 0.9469, + "step": 51032 + }, + { + "epoch": 3.77, + "learning_rate": 6.406231136385368e-06, + "loss": 1.0665, + "step": 51033 + }, + { + "epoch": 3.77, + "learning_rate": 6.405858915237565e-06, + "loss": 0.9875, + "step": 51034 + }, + { + "epoch": 3.77, + "learning_rate": 6.405486699807986e-06, + "loss": 0.8892, + "step": 51035 + }, + { + "epoch": 3.77, + "learning_rate": 6.40511449009723e-06, + "loss": 0.9331, + "step": 51036 + }, + { + "epoch": 3.77, + "learning_rate": 6.404742286105882e-06, + "loss": 1.1206, + "step": 51037 + }, + { + "epoch": 3.77, + "learning_rate": 6.40437008783454e-06, + "loss": 0.9897, + "step": 51038 + }, + { + "epoch": 3.77, + "learning_rate": 6.403997895283791e-06, + "loss": 1.0499, + "step": 51039 + }, + { + "epoch": 3.77, + "learning_rate": 6.403625708454233e-06, + "loss": 1.0202, + "step": 51040 + }, + { + "epoch": 3.77, + "learning_rate": 6.403253527346452e-06, + "loss": 0.9937, + "step": 51041 + }, + { + "epoch": 3.77, + "learning_rate": 6.402881351961045e-06, + "loss": 0.9574, + "step": 51042 + }, + { + "epoch": 3.77, + "learning_rate": 6.402509182298604e-06, + "loss": 1.0886, + "step": 51043 + }, + { + "epoch": 3.77, + "learning_rate": 6.402137018359717e-06, + "loss": 1.0359, + "step": 51044 + }, + { + "epoch": 3.77, + "learning_rate": 6.401764860144979e-06, + "loss": 1.1051, + "step": 51045 + }, + { + "epoch": 3.77, + "learning_rate": 6.401392707654981e-06, + "loss": 0.9963, + "step": 51046 + }, + { + "epoch": 3.77, + "learning_rate": 6.4010205608903174e-06, + "loss": 1.031, + "step": 51047 + }, + { + "epoch": 3.77, + "learning_rate": 6.400648419851576e-06, + "loss": 1.0207, + "step": 51048 + }, + { + "epoch": 3.77, + "learning_rate": 6.400276284539355e-06, + "loss": 1.0185, + "step": 51049 + }, + { + "epoch": 3.77, + "learning_rate": 6.399904154954239e-06, + "loss": 1.0127, + "step": 51050 + }, + { + "epoch": 3.77, + "learning_rate": 6.399532031096827e-06, + "loss": 0.9808, + "step": 51051 + }, + { + "epoch": 3.77, + "learning_rate": 6.399159912967706e-06, + "loss": 0.9992, + "step": 51052 + }, + { + "epoch": 3.77, + "learning_rate": 6.398787800567472e-06, + "loss": 1.0467, + "step": 51053 + }, + { + "epoch": 3.77, + "learning_rate": 6.398415693896718e-06, + "loss": 1.0141, + "step": 51054 + }, + { + "epoch": 3.77, + "learning_rate": 6.398043592956027e-06, + "loss": 1.0038, + "step": 51055 + }, + { + "epoch": 3.77, + "learning_rate": 6.3976714977460006e-06, + "loss": 1.0396, + "step": 51056 + }, + { + "epoch": 3.77, + "learning_rate": 6.3972994082672255e-06, + "loss": 1.0287, + "step": 51057 + }, + { + "epoch": 3.77, + "learning_rate": 6.396927324520297e-06, + "loss": 1.0729, + "step": 51058 + }, + { + "epoch": 3.77, + "learning_rate": 6.3965552465058036e-06, + "loss": 1.0018, + "step": 51059 + }, + { + "epoch": 3.77, + "learning_rate": 6.396183174224341e-06, + "loss": 1.067, + "step": 51060 + }, + { + "epoch": 3.77, + "learning_rate": 6.395811107676498e-06, + "loss": 0.9443, + "step": 51061 + }, + { + "epoch": 3.77, + "learning_rate": 6.395439046862869e-06, + "loss": 0.9382, + "step": 51062 + }, + { + "epoch": 3.77, + "learning_rate": 6.395066991784046e-06, + "loss": 0.9416, + "step": 51063 + }, + { + "epoch": 3.77, + "learning_rate": 6.394694942440617e-06, + "loss": 0.9731, + "step": 51064 + }, + { + "epoch": 3.77, + "learning_rate": 6.394322898833179e-06, + "loss": 1.053, + "step": 51065 + }, + { + "epoch": 3.77, + "learning_rate": 6.393950860962319e-06, + "loss": 0.9137, + "step": 51066 + }, + { + "epoch": 3.77, + "learning_rate": 6.3935788288286345e-06, + "loss": 1.0562, + "step": 51067 + }, + { + "epoch": 3.77, + "learning_rate": 6.3932068024327135e-06, + "loss": 1.0214, + "step": 51068 + }, + { + "epoch": 3.77, + "learning_rate": 6.392834781775149e-06, + "loss": 1.0255, + "step": 51069 + }, + { + "epoch": 3.77, + "learning_rate": 6.392462766856531e-06, + "loss": 0.944, + "step": 51070 + }, + { + "epoch": 3.77, + "learning_rate": 6.392090757677456e-06, + "loss": 1.0197, + "step": 51071 + }, + { + "epoch": 3.77, + "learning_rate": 6.391718754238515e-06, + "loss": 0.9772, + "step": 51072 + }, + { + "epoch": 3.77, + "learning_rate": 6.391346756540291e-06, + "loss": 1.0198, + "step": 51073 + }, + { + "epoch": 3.77, + "learning_rate": 6.390974764583389e-06, + "loss": 1.0565, + "step": 51074 + }, + { + "epoch": 3.77, + "learning_rate": 6.390602778368389e-06, + "loss": 1.0482, + "step": 51075 + }, + { + "epoch": 3.77, + "learning_rate": 6.390230797895893e-06, + "loss": 1.1272, + "step": 51076 + }, + { + "epoch": 3.77, + "learning_rate": 6.389858823166484e-06, + "loss": 0.9661, + "step": 51077 + }, + { + "epoch": 3.77, + "learning_rate": 6.389486854180763e-06, + "loss": 0.992, + "step": 51078 + }, + { + "epoch": 3.77, + "learning_rate": 6.389114890939311e-06, + "loss": 0.9347, + "step": 51079 + }, + { + "epoch": 3.77, + "learning_rate": 6.388742933442731e-06, + "loss": 0.9158, + "step": 51080 + }, + { + "epoch": 3.77, + "learning_rate": 6.388370981691608e-06, + "loss": 0.9177, + "step": 51081 + }, + { + "epoch": 3.77, + "learning_rate": 6.387999035686532e-06, + "loss": 0.9588, + "step": 51082 + }, + { + "epoch": 3.77, + "learning_rate": 6.387627095428102e-06, + "loss": 0.8924, + "step": 51083 + }, + { + "epoch": 3.77, + "learning_rate": 6.387255160916901e-06, + "loss": 0.9695, + "step": 51084 + }, + { + "epoch": 3.77, + "learning_rate": 6.386883232153529e-06, + "loss": 1.0647, + "step": 51085 + }, + { + "epoch": 3.77, + "learning_rate": 6.386511309138572e-06, + "loss": 1.0201, + "step": 51086 + }, + { + "epoch": 3.77, + "learning_rate": 6.386139391872625e-06, + "loss": 1.0863, + "step": 51087 + }, + { + "epoch": 3.77, + "learning_rate": 6.385767480356278e-06, + "loss": 0.8842, + "step": 51088 + }, + { + "epoch": 3.77, + "learning_rate": 6.385395574590123e-06, + "loss": 0.9246, + "step": 51089 + }, + { + "epoch": 3.78, + "learning_rate": 6.385023674574753e-06, + "loss": 0.9355, + "step": 51090 + }, + { + "epoch": 3.78, + "learning_rate": 6.384651780310758e-06, + "loss": 1.0356, + "step": 51091 + }, + { + "epoch": 3.78, + "learning_rate": 6.38427989179873e-06, + "loss": 0.9668, + "step": 51092 + }, + { + "epoch": 3.78, + "learning_rate": 6.3839080090392605e-06, + "loss": 0.9756, + "step": 51093 + }, + { + "epoch": 3.78, + "learning_rate": 6.383536132032942e-06, + "loss": 1.0795, + "step": 51094 + }, + { + "epoch": 3.78, + "learning_rate": 6.383164260780366e-06, + "loss": 0.9528, + "step": 51095 + }, + { + "epoch": 3.78, + "learning_rate": 6.382792395282124e-06, + "loss": 1.0365, + "step": 51096 + }, + { + "epoch": 3.78, + "learning_rate": 6.382420535538807e-06, + "loss": 0.849, + "step": 51097 + }, + { + "epoch": 3.78, + "learning_rate": 6.382048681551009e-06, + "loss": 0.9788, + "step": 51098 + }, + { + "epoch": 3.78, + "learning_rate": 6.3816768333193166e-06, + "loss": 0.9834, + "step": 51099 + }, + { + "epoch": 3.78, + "learning_rate": 6.381304990844329e-06, + "loss": 0.9217, + "step": 51100 + }, + { + "epoch": 3.78, + "learning_rate": 6.380933154126631e-06, + "loss": 0.8925, + "step": 51101 + }, + { + "epoch": 3.78, + "learning_rate": 6.380561323166816e-06, + "loss": 1.0136, + "step": 51102 + }, + { + "epoch": 3.78, + "learning_rate": 6.380189497965478e-06, + "loss": 1.0718, + "step": 51103 + }, + { + "epoch": 3.78, + "learning_rate": 6.379817678523203e-06, + "loss": 1.0612, + "step": 51104 + }, + { + "epoch": 3.78, + "learning_rate": 6.37944586484059e-06, + "loss": 0.9579, + "step": 51105 + }, + { + "epoch": 3.78, + "learning_rate": 6.379074056918224e-06, + "loss": 1.0144, + "step": 51106 + }, + { + "epoch": 3.78, + "learning_rate": 6.3787022547567015e-06, + "loss": 0.9616, + "step": 51107 + }, + { + "epoch": 3.78, + "learning_rate": 6.37833045835661e-06, + "loss": 1.0364, + "step": 51108 + }, + { + "epoch": 3.78, + "learning_rate": 6.377958667718545e-06, + "loss": 1.0874, + "step": 51109 + }, + { + "epoch": 3.78, + "learning_rate": 6.377586882843096e-06, + "loss": 0.9882, + "step": 51110 + }, + { + "epoch": 3.78, + "learning_rate": 6.377215103730851e-06, + "loss": 1.0106, + "step": 51111 + }, + { + "epoch": 3.78, + "learning_rate": 6.37684333038241e-06, + "loss": 1.077, + "step": 51112 + }, + { + "epoch": 3.78, + "learning_rate": 6.3764715627983555e-06, + "loss": 1.089, + "step": 51113 + }, + { + "epoch": 3.78, + "learning_rate": 6.376099800979285e-06, + "loss": 0.8869, + "step": 51114 + }, + { + "epoch": 3.78, + "learning_rate": 6.375728044925786e-06, + "loss": 1.1457, + "step": 51115 + }, + { + "epoch": 3.78, + "learning_rate": 6.375356294638453e-06, + "loss": 0.9843, + "step": 51116 + }, + { + "epoch": 3.78, + "learning_rate": 6.374984550117876e-06, + "loss": 1.0877, + "step": 51117 + }, + { + "epoch": 3.78, + "learning_rate": 6.374612811364647e-06, + "loss": 1.0079, + "step": 51118 + }, + { + "epoch": 3.78, + "learning_rate": 6.37424107837936e-06, + "loss": 0.9959, + "step": 51119 + }, + { + "epoch": 3.78, + "learning_rate": 6.373869351162597e-06, + "loss": 1.052, + "step": 51120 + }, + { + "epoch": 3.78, + "learning_rate": 6.373497629714963e-06, + "loss": 1.0015, + "step": 51121 + }, + { + "epoch": 3.78, + "learning_rate": 6.373125914037035e-06, + "loss": 1.0131, + "step": 51122 + }, + { + "epoch": 3.78, + "learning_rate": 6.372754204129419e-06, + "loss": 1.0528, + "step": 51123 + }, + { + "epoch": 3.78, + "learning_rate": 6.372382499992693e-06, + "loss": 1.0632, + "step": 51124 + }, + { + "epoch": 3.78, + "learning_rate": 6.37201080162746e-06, + "loss": 0.974, + "step": 51125 + }, + { + "epoch": 3.78, + "learning_rate": 6.3716391090343e-06, + "loss": 0.9201, + "step": 51126 + }, + { + "epoch": 3.78, + "learning_rate": 6.3712674222138175e-06, + "loss": 1.0084, + "step": 51127 + }, + { + "epoch": 3.78, + "learning_rate": 6.370895741166593e-06, + "loss": 0.9439, + "step": 51128 + }, + { + "epoch": 3.78, + "learning_rate": 6.37052406589322e-06, + "loss": 1.1007, + "step": 51129 + }, + { + "epoch": 3.78, + "learning_rate": 6.370152396394293e-06, + "loss": 1.0806, + "step": 51130 + }, + { + "epoch": 3.78, + "learning_rate": 6.369780732670401e-06, + "loss": 1.027, + "step": 51131 + }, + { + "epoch": 3.78, + "learning_rate": 6.369409074722136e-06, + "loss": 1.0271, + "step": 51132 + }, + { + "epoch": 3.78, + "learning_rate": 6.3690374225500885e-06, + "loss": 0.9903, + "step": 51133 + }, + { + "epoch": 3.78, + "learning_rate": 6.368665776154852e-06, + "loss": 0.9409, + "step": 51134 + }, + { + "epoch": 3.78, + "learning_rate": 6.368294135537014e-06, + "loss": 0.9579, + "step": 51135 + }, + { + "epoch": 3.78, + "learning_rate": 6.36792250069717e-06, + "loss": 1.0997, + "step": 51136 + }, + { + "epoch": 3.78, + "learning_rate": 6.36755087163591e-06, + "loss": 1.0286, + "step": 51137 + }, + { + "epoch": 3.78, + "learning_rate": 6.367179248353823e-06, + "loss": 0.9705, + "step": 51138 + }, + { + "epoch": 3.78, + "learning_rate": 6.3668076308515024e-06, + "loss": 1.0487, + "step": 51139 + }, + { + "epoch": 3.78, + "learning_rate": 6.3664360191295385e-06, + "loss": 1.0039, + "step": 51140 + }, + { + "epoch": 3.78, + "learning_rate": 6.3660644131885244e-06, + "loss": 1.0835, + "step": 51141 + }, + { + "epoch": 3.78, + "learning_rate": 6.365692813029046e-06, + "loss": 0.9118, + "step": 51142 + }, + { + "epoch": 3.78, + "learning_rate": 6.365321218651704e-06, + "loss": 1.0544, + "step": 51143 + }, + { + "epoch": 3.78, + "learning_rate": 6.364949630057078e-06, + "loss": 1.0408, + "step": 51144 + }, + { + "epoch": 3.78, + "learning_rate": 6.3645780472457705e-06, + "loss": 1.0214, + "step": 51145 + }, + { + "epoch": 3.78, + "learning_rate": 6.364206470218366e-06, + "loss": 0.8596, + "step": 51146 + }, + { + "epoch": 3.78, + "learning_rate": 6.363834898975455e-06, + "loss": 1.0986, + "step": 51147 + }, + { + "epoch": 3.78, + "learning_rate": 6.363463333517632e-06, + "loss": 1.0727, + "step": 51148 + }, + { + "epoch": 3.78, + "learning_rate": 6.363091773845484e-06, + "loss": 1.0987, + "step": 51149 + }, + { + "epoch": 3.78, + "learning_rate": 6.362720219959609e-06, + "loss": 1.0688, + "step": 51150 + }, + { + "epoch": 3.78, + "learning_rate": 6.362348671860592e-06, + "loss": 0.9997, + "step": 51151 + }, + { + "epoch": 3.78, + "learning_rate": 6.361977129549026e-06, + "loss": 1.0015, + "step": 51152 + }, + { + "epoch": 3.78, + "learning_rate": 6.361605593025503e-06, + "loss": 1.0605, + "step": 51153 + }, + { + "epoch": 3.78, + "learning_rate": 6.361234062290613e-06, + "loss": 1.023, + "step": 51154 + }, + { + "epoch": 3.78, + "learning_rate": 6.360862537344948e-06, + "loss": 0.9583, + "step": 51155 + }, + { + "epoch": 3.78, + "learning_rate": 6.360491018189099e-06, + "loss": 1.0164, + "step": 51156 + }, + { + "epoch": 3.78, + "learning_rate": 6.3601195048236564e-06, + "loss": 1.0189, + "step": 51157 + }, + { + "epoch": 3.78, + "learning_rate": 6.3597479972492104e-06, + "loss": 1.0179, + "step": 51158 + }, + { + "epoch": 3.78, + "learning_rate": 6.3593764954663565e-06, + "loss": 1.0079, + "step": 51159 + }, + { + "epoch": 3.78, + "learning_rate": 6.359004999475677e-06, + "loss": 1.0575, + "step": 51160 + }, + { + "epoch": 3.78, + "learning_rate": 6.358633509277774e-06, + "loss": 1.0468, + "step": 51161 + }, + { + "epoch": 3.78, + "learning_rate": 6.35826202487323e-06, + "loss": 1.0169, + "step": 51162 + }, + { + "epoch": 3.78, + "learning_rate": 6.3578905462626395e-06, + "loss": 0.9045, + "step": 51163 + }, + { + "epoch": 3.78, + "learning_rate": 6.357519073446592e-06, + "loss": 0.997, + "step": 51164 + }, + { + "epoch": 3.78, + "learning_rate": 6.357147606425682e-06, + "loss": 1.0567, + "step": 51165 + }, + { + "epoch": 3.78, + "learning_rate": 6.3567761452005e-06, + "loss": 1.0237, + "step": 51166 + }, + { + "epoch": 3.78, + "learning_rate": 6.3564046897716294e-06, + "loss": 0.9413, + "step": 51167 + }, + { + "epoch": 3.78, + "learning_rate": 6.356033240139672e-06, + "loss": 0.982, + "step": 51168 + }, + { + "epoch": 3.78, + "learning_rate": 6.355661796305207e-06, + "loss": 1.0322, + "step": 51169 + }, + { + "epoch": 3.78, + "learning_rate": 6.355290358268839e-06, + "loss": 1.0103, + "step": 51170 + }, + { + "epoch": 3.78, + "learning_rate": 6.354918926031145e-06, + "loss": 0.9707, + "step": 51171 + }, + { + "epoch": 3.78, + "learning_rate": 6.3545474995927294e-06, + "loss": 1.0315, + "step": 51172 + }, + { + "epoch": 3.78, + "learning_rate": 6.354176078954171e-06, + "loss": 1.1209, + "step": 51173 + }, + { + "epoch": 3.78, + "learning_rate": 6.353804664116072e-06, + "loss": 1.0212, + "step": 51174 + }, + { + "epoch": 3.78, + "learning_rate": 6.3534332550790155e-06, + "loss": 0.9465, + "step": 51175 + }, + { + "epoch": 3.78, + "learning_rate": 6.3530618518435914e-06, + "loss": 1.0117, + "step": 51176 + }, + { + "epoch": 3.78, + "learning_rate": 6.352690454410397e-06, + "loss": 1.0048, + "step": 51177 + }, + { + "epoch": 3.78, + "learning_rate": 6.352319062780018e-06, + "loss": 1.0598, + "step": 51178 + }, + { + "epoch": 3.78, + "learning_rate": 6.35194767695305e-06, + "loss": 0.9998, + "step": 51179 + }, + { + "epoch": 3.78, + "learning_rate": 6.351576296930078e-06, + "loss": 0.9206, + "step": 51180 + }, + { + "epoch": 3.78, + "learning_rate": 6.351204922711697e-06, + "loss": 1.0602, + "step": 51181 + }, + { + "epoch": 3.78, + "learning_rate": 6.350833554298494e-06, + "loss": 0.9043, + "step": 51182 + }, + { + "epoch": 3.78, + "learning_rate": 6.350462191691067e-06, + "loss": 0.9914, + "step": 51183 + }, + { + "epoch": 3.78, + "learning_rate": 6.350090834890001e-06, + "loss": 0.9735, + "step": 51184 + }, + { + "epoch": 3.78, + "learning_rate": 6.349719483895889e-06, + "loss": 1.0537, + "step": 51185 + }, + { + "epoch": 3.78, + "learning_rate": 6.349348138709319e-06, + "loss": 0.9294, + "step": 51186 + }, + { + "epoch": 3.78, + "learning_rate": 6.348976799330885e-06, + "loss": 1.1025, + "step": 51187 + }, + { + "epoch": 3.78, + "learning_rate": 6.34860546576118e-06, + "loss": 0.839, + "step": 51188 + }, + { + "epoch": 3.78, + "learning_rate": 6.348234138000784e-06, + "loss": 1.0286, + "step": 51189 + }, + { + "epoch": 3.78, + "learning_rate": 6.347862816050303e-06, + "loss": 0.9151, + "step": 51190 + }, + { + "epoch": 3.78, + "learning_rate": 6.347491499910312e-06, + "loss": 0.9724, + "step": 51191 + }, + { + "epoch": 3.78, + "learning_rate": 6.347120189581417e-06, + "loss": 1.0083, + "step": 51192 + }, + { + "epoch": 3.78, + "learning_rate": 6.3467488850642e-06, + "loss": 0.8252, + "step": 51193 + }, + { + "epoch": 3.78, + "learning_rate": 6.346377586359252e-06, + "loss": 1.0644, + "step": 51194 + }, + { + "epoch": 3.78, + "learning_rate": 6.346006293467164e-06, + "loss": 1.0737, + "step": 51195 + }, + { + "epoch": 3.78, + "learning_rate": 6.345635006388529e-06, + "loss": 1.0384, + "step": 51196 + }, + { + "epoch": 3.78, + "learning_rate": 6.345263725123936e-06, + "loss": 0.9715, + "step": 51197 + }, + { + "epoch": 3.78, + "learning_rate": 6.344892449673975e-06, + "loss": 1.0012, + "step": 51198 + }, + { + "epoch": 3.78, + "learning_rate": 6.344521180039241e-06, + "loss": 1.0879, + "step": 51199 + }, + { + "epoch": 3.78, + "learning_rate": 6.344149916220317e-06, + "loss": 1.0484, + "step": 51200 + }, + { + "epoch": 3.78, + "learning_rate": 6.343778658217801e-06, + "loss": 1.0343, + "step": 51201 + }, + { + "epoch": 3.78, + "learning_rate": 6.343407406032282e-06, + "loss": 0.8997, + "step": 51202 + }, + { + "epoch": 3.78, + "learning_rate": 6.343036159664346e-06, + "loss": 1.0293, + "step": 51203 + }, + { + "epoch": 3.78, + "learning_rate": 6.342664919114591e-06, + "loss": 1.04, + "step": 51204 + }, + { + "epoch": 3.78, + "learning_rate": 6.342293684383601e-06, + "loss": 1.0669, + "step": 51205 + }, + { + "epoch": 3.78, + "learning_rate": 6.34192245547197e-06, + "loss": 1.0065, + "step": 51206 + }, + { + "epoch": 3.78, + "learning_rate": 6.341551232380288e-06, + "loss": 1.0112, + "step": 51207 + }, + { + "epoch": 3.78, + "learning_rate": 6.341180015109147e-06, + "loss": 1.137, + "step": 51208 + }, + { + "epoch": 3.78, + "learning_rate": 6.340808803659133e-06, + "loss": 1.031, + "step": 51209 + }, + { + "epoch": 3.78, + "learning_rate": 6.340437598030844e-06, + "loss": 0.9689, + "step": 51210 + }, + { + "epoch": 3.78, + "learning_rate": 6.340066398224863e-06, + "loss": 0.8697, + "step": 51211 + }, + { + "epoch": 3.78, + "learning_rate": 6.339695204241787e-06, + "loss": 1.0386, + "step": 51212 + }, + { + "epoch": 3.78, + "learning_rate": 6.339324016082205e-06, + "loss": 0.9596, + "step": 51213 + }, + { + "epoch": 3.78, + "learning_rate": 6.338952833746699e-06, + "loss": 0.9753, + "step": 51214 + }, + { + "epoch": 3.78, + "learning_rate": 6.338581657235874e-06, + "loss": 0.9694, + "step": 51215 + }, + { + "epoch": 3.78, + "learning_rate": 6.338210486550307e-06, + "loss": 0.9908, + "step": 51216 + }, + { + "epoch": 3.78, + "learning_rate": 6.3378393216906e-06, + "loss": 0.9257, + "step": 51217 + }, + { + "epoch": 3.78, + "learning_rate": 6.337468162657333e-06, + "loss": 0.9966, + "step": 51218 + }, + { + "epoch": 3.78, + "learning_rate": 6.337097009451107e-06, + "loss": 0.9331, + "step": 51219 + }, + { + "epoch": 3.78, + "learning_rate": 6.336725862072505e-06, + "loss": 0.9472, + "step": 51220 + }, + { + "epoch": 3.78, + "learning_rate": 6.33635472052212e-06, + "loss": 1.0678, + "step": 51221 + }, + { + "epoch": 3.78, + "learning_rate": 6.335983584800543e-06, + "loss": 0.8228, + "step": 51222 + }, + { + "epoch": 3.78, + "learning_rate": 6.335612454908362e-06, + "loss": 0.9885, + "step": 51223 + }, + { + "epoch": 3.78, + "learning_rate": 6.335241330846171e-06, + "loss": 0.923, + "step": 51224 + }, + { + "epoch": 3.78, + "learning_rate": 6.3348702126145555e-06, + "loss": 1.0848, + "step": 51225 + }, + { + "epoch": 3.79, + "learning_rate": 6.334499100214112e-06, + "loss": 0.9014, + "step": 51226 + }, + { + "epoch": 3.79, + "learning_rate": 6.334127993645426e-06, + "loss": 1.0958, + "step": 51227 + }, + { + "epoch": 3.79, + "learning_rate": 6.333756892909091e-06, + "loss": 0.9655, + "step": 51228 + }, + { + "epoch": 3.79, + "learning_rate": 6.333385798005696e-06, + "loss": 0.8936, + "step": 51229 + }, + { + "epoch": 3.79, + "learning_rate": 6.333014708935831e-06, + "loss": 1.0706, + "step": 51230 + }, + { + "epoch": 3.79, + "learning_rate": 6.3326436257000925e-06, + "loss": 1.0126, + "step": 51231 + }, + { + "epoch": 3.79, + "learning_rate": 6.332272548299058e-06, + "loss": 1.0155, + "step": 51232 + }, + { + "epoch": 3.79, + "learning_rate": 6.331901476733331e-06, + "loss": 1.066, + "step": 51233 + }, + { + "epoch": 3.79, + "learning_rate": 6.33153041100349e-06, + "loss": 0.9287, + "step": 51234 + }, + { + "epoch": 3.79, + "learning_rate": 6.331159351110139e-06, + "loss": 0.945, + "step": 51235 + }, + { + "epoch": 3.79, + "learning_rate": 6.330788297053853e-06, + "loss": 1.1319, + "step": 51236 + }, + { + "epoch": 3.79, + "learning_rate": 6.330417248835239e-06, + "loss": 0.9739, + "step": 51237 + }, + { + "epoch": 3.79, + "learning_rate": 6.330046206454871e-06, + "loss": 0.9753, + "step": 51238 + }, + { + "epoch": 3.79, + "learning_rate": 6.329675169913353e-06, + "loss": 0.9316, + "step": 51239 + }, + { + "epoch": 3.79, + "learning_rate": 6.329304139211267e-06, + "loss": 1.0046, + "step": 51240 + }, + { + "epoch": 3.79, + "learning_rate": 6.328933114349205e-06, + "loss": 0.9484, + "step": 51241 + }, + { + "epoch": 3.79, + "learning_rate": 6.3285620953277595e-06, + "loss": 1.0073, + "step": 51242 + }, + { + "epoch": 3.79, + "learning_rate": 6.328191082147515e-06, + "loss": 0.8416, + "step": 51243 + }, + { + "epoch": 3.79, + "learning_rate": 6.32782007480907e-06, + "loss": 0.999, + "step": 51244 + }, + { + "epoch": 3.79, + "learning_rate": 6.327449073313008e-06, + "loss": 0.997, + "step": 51245 + }, + { + "epoch": 3.79, + "learning_rate": 6.327078077659925e-06, + "loss": 0.9696, + "step": 51246 + }, + { + "epoch": 3.79, + "learning_rate": 6.326707087850406e-06, + "loss": 0.9659, + "step": 51247 + }, + { + "epoch": 3.79, + "learning_rate": 6.3263361038850445e-06, + "loss": 1.0125, + "step": 51248 + }, + { + "epoch": 3.79, + "learning_rate": 6.325965125764431e-06, + "loss": 0.8912, + "step": 51249 + }, + { + "epoch": 3.79, + "learning_rate": 6.325594153489151e-06, + "loss": 0.9438, + "step": 51250 + }, + { + "epoch": 3.79, + "learning_rate": 6.3252231870598e-06, + "loss": 0.9496, + "step": 51251 + }, + { + "epoch": 3.79, + "learning_rate": 6.3248522264769654e-06, + "loss": 1.0315, + "step": 51252 + }, + { + "epoch": 3.79, + "learning_rate": 6.324481271741241e-06, + "loss": 0.9481, + "step": 51253 + }, + { + "epoch": 3.79, + "learning_rate": 6.32411032285321e-06, + "loss": 0.946, + "step": 51254 + }, + { + "epoch": 3.79, + "learning_rate": 6.323739379813471e-06, + "loss": 1.0508, + "step": 51255 + }, + { + "epoch": 3.79, + "learning_rate": 6.3233684426226065e-06, + "loss": 0.992, + "step": 51256 + }, + { + "epoch": 3.79, + "learning_rate": 6.322997511281213e-06, + "loss": 1.0102, + "step": 51257 + }, + { + "epoch": 3.79, + "learning_rate": 6.32262658578988e-06, + "loss": 1.0033, + "step": 51258 + }, + { + "epoch": 3.79, + "learning_rate": 6.322255666149188e-06, + "loss": 0.9923, + "step": 51259 + }, + { + "epoch": 3.79, + "learning_rate": 6.3218847523597435e-06, + "loss": 1.0119, + "step": 51260 + }, + { + "epoch": 3.79, + "learning_rate": 6.32151384442212e-06, + "loss": 0.972, + "step": 51261 + }, + { + "epoch": 3.79, + "learning_rate": 6.3211429423369205e-06, + "loss": 1.0882, + "step": 51262 + }, + { + "epoch": 3.79, + "learning_rate": 6.3207720461047264e-06, + "loss": 0.942, + "step": 51263 + }, + { + "epoch": 3.79, + "learning_rate": 6.320401155726135e-06, + "loss": 1.0444, + "step": 51264 + }, + { + "epoch": 3.79, + "learning_rate": 6.320030271201729e-06, + "loss": 0.9237, + "step": 51265 + }, + { + "epoch": 3.79, + "learning_rate": 6.319659392532105e-06, + "loss": 0.9679, + "step": 51266 + }, + { + "epoch": 3.79, + "learning_rate": 6.319288519717848e-06, + "loss": 0.9514, + "step": 51267 + }, + { + "epoch": 3.79, + "learning_rate": 6.318917652759553e-06, + "loss": 0.9429, + "step": 51268 + }, + { + "epoch": 3.79, + "learning_rate": 6.3185467916578055e-06, + "loss": 1.0558, + "step": 51269 + }, + { + "epoch": 3.79, + "learning_rate": 6.318175936413198e-06, + "loss": 0.9991, + "step": 51270 + }, + { + "epoch": 3.79, + "learning_rate": 6.31780508702632e-06, + "loss": 0.9255, + "step": 51271 + }, + { + "epoch": 3.79, + "learning_rate": 6.3174342434977594e-06, + "loss": 1.0625, + "step": 51272 + }, + { + "epoch": 3.79, + "learning_rate": 6.3170634058281096e-06, + "loss": 0.9748, + "step": 51273 + }, + { + "epoch": 3.79, + "learning_rate": 6.31669257401796e-06, + "loss": 0.9234, + "step": 51274 + }, + { + "epoch": 3.79, + "learning_rate": 6.3163217480678995e-06, + "loss": 0.9879, + "step": 51275 + }, + { + "epoch": 3.79, + "learning_rate": 6.315950927978517e-06, + "loss": 1.0092, + "step": 51276 + }, + { + "epoch": 3.79, + "learning_rate": 6.3155801137504054e-06, + "loss": 1.0581, + "step": 51277 + }, + { + "epoch": 3.79, + "learning_rate": 6.315209305384156e-06, + "loss": 1.1074, + "step": 51278 + }, + { + "epoch": 3.79, + "learning_rate": 6.3148385028803495e-06, + "loss": 0.9794, + "step": 51279 + }, + { + "epoch": 3.79, + "learning_rate": 6.314467706239589e-06, + "loss": 1.0225, + "step": 51280 + }, + { + "epoch": 3.79, + "learning_rate": 6.31409691546245e-06, + "loss": 1.0816, + "step": 51281 + }, + { + "epoch": 3.79, + "learning_rate": 6.313726130549537e-06, + "loss": 0.9956, + "step": 51282 + }, + { + "epoch": 3.79, + "learning_rate": 6.3133553515014265e-06, + "loss": 1.0449, + "step": 51283 + }, + { + "epoch": 3.79, + "learning_rate": 6.312984578318722e-06, + "loss": 0.9298, + "step": 51284 + }, + { + "epoch": 3.79, + "learning_rate": 6.3126138110019996e-06, + "loss": 1.0868, + "step": 51285 + }, + { + "epoch": 3.79, + "learning_rate": 6.312243049551861e-06, + "loss": 0.9867, + "step": 51286 + }, + { + "epoch": 3.79, + "learning_rate": 6.31187229396889e-06, + "loss": 1.0561, + "step": 51287 + }, + { + "epoch": 3.79, + "learning_rate": 6.3115015442536755e-06, + "loss": 0.9126, + "step": 51288 + }, + { + "epoch": 3.79, + "learning_rate": 6.311130800406811e-06, + "loss": 1.082, + "step": 51289 + }, + { + "epoch": 3.79, + "learning_rate": 6.310760062428882e-06, + "loss": 1.0242, + "step": 51290 + }, + { + "epoch": 3.79, + "learning_rate": 6.310389330320483e-06, + "loss": 1.112, + "step": 51291 + }, + { + "epoch": 3.79, + "learning_rate": 6.310018604082201e-06, + "loss": 0.9743, + "step": 51292 + }, + { + "epoch": 3.79, + "learning_rate": 6.309647883714628e-06, + "loss": 1.0242, + "step": 51293 + }, + { + "epoch": 3.79, + "learning_rate": 6.309277169218351e-06, + "loss": 1.0982, + "step": 51294 + }, + { + "epoch": 3.79, + "learning_rate": 6.308906460593962e-06, + "loss": 0.9821, + "step": 51295 + }, + { + "epoch": 3.79, + "learning_rate": 6.3085357578420505e-06, + "loss": 0.9205, + "step": 51296 + }, + { + "epoch": 3.79, + "learning_rate": 6.308165060963204e-06, + "loss": 0.9518, + "step": 51297 + }, + { + "epoch": 3.79, + "learning_rate": 6.307794369958016e-06, + "loss": 0.9754, + "step": 51298 + }, + { + "epoch": 3.79, + "learning_rate": 6.307423684827072e-06, + "loss": 1.0309, + "step": 51299 + }, + { + "epoch": 3.79, + "learning_rate": 6.307053005570967e-06, + "loss": 0.9666, + "step": 51300 + }, + { + "epoch": 3.79, + "learning_rate": 6.306682332190283e-06, + "loss": 1.0111, + "step": 51301 + }, + { + "epoch": 3.79, + "learning_rate": 6.306311664685621e-06, + "loss": 1.0276, + "step": 51302 + }, + { + "epoch": 3.79, + "learning_rate": 6.305941003057558e-06, + "loss": 0.9211, + "step": 51303 + }, + { + "epoch": 3.79, + "learning_rate": 6.305570347306696e-06, + "loss": 1.101, + "step": 51304 + }, + { + "epoch": 3.79, + "learning_rate": 6.3051996974336185e-06, + "loss": 0.9133, + "step": 51305 + }, + { + "epoch": 3.79, + "learning_rate": 6.304829053438909e-06, + "loss": 0.9964, + "step": 51306 + }, + { + "epoch": 3.79, + "learning_rate": 6.30445841532317e-06, + "loss": 0.9644, + "step": 51307 + }, + { + "epoch": 3.79, + "learning_rate": 6.30408778308698e-06, + "loss": 1.0941, + "step": 51308 + }, + { + "epoch": 3.79, + "learning_rate": 6.303717156730937e-06, + "loss": 1.0447, + "step": 51309 + }, + { + "epoch": 3.79, + "learning_rate": 6.303346536255626e-06, + "loss": 0.9791, + "step": 51310 + }, + { + "epoch": 3.79, + "learning_rate": 6.3029759216616384e-06, + "loss": 0.9013, + "step": 51311 + }, + { + "epoch": 3.79, + "learning_rate": 6.3026053129495614e-06, + "loss": 0.9792, + "step": 51312 + }, + { + "epoch": 3.79, + "learning_rate": 6.302234710119988e-06, + "loss": 0.9908, + "step": 51313 + }, + { + "epoch": 3.79, + "learning_rate": 6.301864113173504e-06, + "loss": 0.9608, + "step": 51314 + }, + { + "epoch": 3.79, + "learning_rate": 6.301493522110704e-06, + "loss": 0.9785, + "step": 51315 + }, + { + "epoch": 3.79, + "learning_rate": 6.301122936932175e-06, + "loss": 1.0633, + "step": 51316 + }, + { + "epoch": 3.79, + "learning_rate": 6.300752357638504e-06, + "loss": 1.0091, + "step": 51317 + }, + { + "epoch": 3.79, + "learning_rate": 6.300381784230285e-06, + "loss": 0.982, + "step": 51318 + }, + { + "epoch": 3.79, + "learning_rate": 6.300011216708104e-06, + "loss": 1.0534, + "step": 51319 + }, + { + "epoch": 3.79, + "learning_rate": 6.299640655072554e-06, + "loss": 1.0714, + "step": 51320 + }, + { + "epoch": 3.79, + "learning_rate": 6.299270099324221e-06, + "loss": 0.9729, + "step": 51321 + }, + { + "epoch": 3.79, + "learning_rate": 6.298899549463699e-06, + "loss": 0.9292, + "step": 51322 + }, + { + "epoch": 3.79, + "learning_rate": 6.298529005491571e-06, + "loss": 1.0657, + "step": 51323 + }, + { + "epoch": 3.79, + "learning_rate": 6.298158467408434e-06, + "loss": 0.9971, + "step": 51324 + }, + { + "epoch": 3.79, + "learning_rate": 6.297787935214876e-06, + "loss": 1.1131, + "step": 51325 + }, + { + "epoch": 3.79, + "learning_rate": 6.2974174089114774e-06, + "loss": 1.1261, + "step": 51326 + }, + { + "epoch": 3.79, + "learning_rate": 6.297046888498841e-06, + "loss": 1.0427, + "step": 51327 + }, + { + "epoch": 3.79, + "learning_rate": 6.296676373977544e-06, + "loss": 1.1875, + "step": 51328 + }, + { + "epoch": 3.79, + "learning_rate": 6.296305865348189e-06, + "loss": 0.9576, + "step": 51329 + }, + { + "epoch": 3.79, + "learning_rate": 6.2959353626113506e-06, + "loss": 0.919, + "step": 51330 + }, + { + "epoch": 3.79, + "learning_rate": 6.295564865767634e-06, + "loss": 1.0963, + "step": 51331 + }, + { + "epoch": 3.79, + "learning_rate": 6.2951943748176125e-06, + "loss": 1.0406, + "step": 51332 + }, + { + "epoch": 3.79, + "learning_rate": 6.294823889761892e-06, + "loss": 0.8853, + "step": 51333 + }, + { + "epoch": 3.79, + "learning_rate": 6.294453410601049e-06, + "loss": 1.0583, + "step": 51334 + }, + { + "epoch": 3.79, + "learning_rate": 6.294082937335677e-06, + "loss": 1.0394, + "step": 51335 + }, + { + "epoch": 3.79, + "learning_rate": 6.293712469966369e-06, + "loss": 0.9542, + "step": 51336 + }, + { + "epoch": 3.79, + "learning_rate": 6.293342008493708e-06, + "loss": 1.0454, + "step": 51337 + }, + { + "epoch": 3.79, + "learning_rate": 6.29297155291829e-06, + "loss": 1.1075, + "step": 51338 + }, + { + "epoch": 3.79, + "learning_rate": 6.292601103240699e-06, + "loss": 1.0344, + "step": 51339 + }, + { + "epoch": 3.79, + "learning_rate": 6.292230659461528e-06, + "loss": 0.9377, + "step": 51340 + }, + { + "epoch": 3.79, + "learning_rate": 6.291860221581363e-06, + "loss": 1.0025, + "step": 51341 + }, + { + "epoch": 3.79, + "learning_rate": 6.291489789600798e-06, + "loss": 0.9475, + "step": 51342 + }, + { + "epoch": 3.79, + "learning_rate": 6.291119363520419e-06, + "loss": 0.9115, + "step": 51343 + }, + { + "epoch": 3.79, + "learning_rate": 6.290748943340814e-06, + "loss": 1.0605, + "step": 51344 + }, + { + "epoch": 3.79, + "learning_rate": 6.290378529062577e-06, + "loss": 0.9983, + "step": 51345 + }, + { + "epoch": 3.79, + "learning_rate": 6.290008120686292e-06, + "loss": 0.8967, + "step": 51346 + }, + { + "epoch": 3.79, + "learning_rate": 6.289637718212553e-06, + "loss": 1.0301, + "step": 51347 + }, + { + "epoch": 3.79, + "learning_rate": 6.2892673216419455e-06, + "loss": 1.0206, + "step": 51348 + }, + { + "epoch": 3.79, + "learning_rate": 6.288896930975063e-06, + "loss": 1.0558, + "step": 51349 + }, + { + "epoch": 3.79, + "learning_rate": 6.2885265462124905e-06, + "loss": 1.0744, + "step": 51350 + }, + { + "epoch": 3.79, + "learning_rate": 6.28815616735482e-06, + "loss": 0.9518, + "step": 51351 + }, + { + "epoch": 3.79, + "learning_rate": 6.287785794402643e-06, + "loss": 0.9868, + "step": 51352 + }, + { + "epoch": 3.79, + "learning_rate": 6.287415427356541e-06, + "loss": 0.8548, + "step": 51353 + }, + { + "epoch": 3.79, + "learning_rate": 6.28704506621711e-06, + "loss": 0.9528, + "step": 51354 + }, + { + "epoch": 3.79, + "learning_rate": 6.286674710984937e-06, + "loss": 1.0195, + "step": 51355 + }, + { + "epoch": 3.79, + "learning_rate": 6.2863043616606115e-06, + "loss": 0.9784, + "step": 51356 + }, + { + "epoch": 3.79, + "learning_rate": 6.28593401824472e-06, + "loss": 1.0217, + "step": 51357 + }, + { + "epoch": 3.79, + "learning_rate": 6.285563680737858e-06, + "loss": 0.963, + "step": 51358 + }, + { + "epoch": 3.79, + "learning_rate": 6.285193349140608e-06, + "loss": 0.9436, + "step": 51359 + }, + { + "epoch": 3.79, + "learning_rate": 6.2848230234535655e-06, + "loss": 0.8786, + "step": 51360 + }, + { + "epoch": 3.8, + "learning_rate": 6.284452703677315e-06, + "loss": 1.0896, + "step": 51361 + }, + { + "epoch": 3.8, + "learning_rate": 6.2840823898124444e-06, + "loss": 0.9541, + "step": 51362 + }, + { + "epoch": 3.8, + "learning_rate": 6.28371208185955e-06, + "loss": 1.0867, + "step": 51363 + }, + { + "epoch": 3.8, + "learning_rate": 6.2833417798192135e-06, + "loss": 1.0352, + "step": 51364 + }, + { + "epoch": 3.8, + "learning_rate": 6.282971483692028e-06, + "loss": 0.972, + "step": 51365 + }, + { + "epoch": 3.8, + "learning_rate": 6.2826011934785814e-06, + "loss": 0.9735, + "step": 51366 + }, + { + "epoch": 3.8, + "learning_rate": 6.282230909179463e-06, + "loss": 1.0869, + "step": 51367 + }, + { + "epoch": 3.8, + "learning_rate": 6.281860630795262e-06, + "loss": 1.0624, + "step": 51368 + }, + { + "epoch": 3.8, + "learning_rate": 6.281490358326569e-06, + "loss": 0.8625, + "step": 51369 + }, + { + "epoch": 3.8, + "learning_rate": 6.281120091773969e-06, + "loss": 1.089, + "step": 51370 + }, + { + "epoch": 3.8, + "learning_rate": 6.280749831138057e-06, + "loss": 0.8985, + "step": 51371 + }, + { + "epoch": 3.8, + "learning_rate": 6.28037957641942e-06, + "loss": 1.0502, + "step": 51372 + }, + { + "epoch": 3.8, + "learning_rate": 6.2800093276186395e-06, + "loss": 1.044, + "step": 51373 + }, + { + "epoch": 3.8, + "learning_rate": 6.279639084736317e-06, + "loss": 1.1203, + "step": 51374 + }, + { + "epoch": 3.8, + "learning_rate": 6.27926884777303e-06, + "loss": 1.0209, + "step": 51375 + }, + { + "epoch": 3.8, + "learning_rate": 6.278898616729379e-06, + "loss": 1.0892, + "step": 51376 + }, + { + "epoch": 3.8, + "learning_rate": 6.2785283916059406e-06, + "loss": 1.0107, + "step": 51377 + }, + { + "epoch": 3.8, + "learning_rate": 6.278158172403317e-06, + "loss": 1.0572, + "step": 51378 + }, + { + "epoch": 3.8, + "learning_rate": 6.277787959122084e-06, + "loss": 1.0305, + "step": 51379 + }, + { + "epoch": 3.8, + "learning_rate": 6.277417751762842e-06, + "loss": 1.0268, + "step": 51380 + }, + { + "epoch": 3.8, + "learning_rate": 6.277047550326174e-06, + "loss": 0.9491, + "step": 51381 + }, + { + "epoch": 3.8, + "learning_rate": 6.276677354812669e-06, + "loss": 1.0939, + "step": 51382 + }, + { + "epoch": 3.8, + "learning_rate": 6.276307165222919e-06, + "loss": 0.9205, + "step": 51383 + }, + { + "epoch": 3.8, + "learning_rate": 6.2759369815575075e-06, + "loss": 0.9619, + "step": 51384 + }, + { + "epoch": 3.8, + "learning_rate": 6.275566803817031e-06, + "loss": 1.0603, + "step": 51385 + }, + { + "epoch": 3.8, + "learning_rate": 6.2751966320020694e-06, + "loss": 1.0014, + "step": 51386 + }, + { + "epoch": 3.8, + "learning_rate": 6.274826466113222e-06, + "loss": 0.9724, + "step": 51387 + }, + { + "epoch": 3.8, + "learning_rate": 6.274456306151069e-06, + "loss": 1.0368, + "step": 51388 + }, + { + "epoch": 3.8, + "learning_rate": 6.274086152116205e-06, + "loss": 1.0895, + "step": 51389 + }, + { + "epoch": 3.8, + "learning_rate": 6.273716004009216e-06, + "loss": 0.896, + "step": 51390 + }, + { + "epoch": 3.8, + "learning_rate": 6.2733458618306886e-06, + "loss": 0.994, + "step": 51391 + }, + { + "epoch": 3.8, + "learning_rate": 6.272975725581219e-06, + "loss": 1.014, + "step": 51392 + }, + { + "epoch": 3.8, + "learning_rate": 6.2726055952613875e-06, + "loss": 0.9957, + "step": 51393 + }, + { + "epoch": 3.8, + "learning_rate": 6.27223547087179e-06, + "loss": 0.873, + "step": 51394 + }, + { + "epoch": 3.8, + "learning_rate": 6.271865352413011e-06, + "loss": 1.038, + "step": 51395 + }, + { + "epoch": 3.8, + "learning_rate": 6.2714952398856435e-06, + "loss": 0.9873, + "step": 51396 + }, + { + "epoch": 3.8, + "learning_rate": 6.27112513329027e-06, + "loss": 1.1753, + "step": 51397 + }, + { + "epoch": 3.8, + "learning_rate": 6.270755032627489e-06, + "loss": 1.0006, + "step": 51398 + }, + { + "epoch": 3.8, + "learning_rate": 6.2703849378978795e-06, + "loss": 0.9683, + "step": 51399 + }, + { + "epoch": 3.8, + "learning_rate": 6.270014849102033e-06, + "loss": 0.9786, + "step": 51400 + }, + { + "epoch": 3.8, + "learning_rate": 6.269644766240541e-06, + "loss": 0.9762, + "step": 51401 + }, + { + "epoch": 3.8, + "learning_rate": 6.26927468931399e-06, + "loss": 0.9373, + "step": 51402 + }, + { + "epoch": 3.8, + "learning_rate": 6.268904618322971e-06, + "loss": 1.0056, + "step": 51403 + }, + { + "epoch": 3.8, + "learning_rate": 6.268534553268069e-06, + "loss": 1.0398, + "step": 51404 + }, + { + "epoch": 3.8, + "learning_rate": 6.268164494149878e-06, + "loss": 1.0587, + "step": 51405 + }, + { + "epoch": 3.8, + "learning_rate": 6.267794440968981e-06, + "loss": 1.0096, + "step": 51406 + }, + { + "epoch": 3.8, + "learning_rate": 6.267424393725972e-06, + "loss": 1.0395, + "step": 51407 + }, + { + "epoch": 3.8, + "learning_rate": 6.267054352421438e-06, + "loss": 1.0059, + "step": 51408 + }, + { + "epoch": 3.8, + "learning_rate": 6.266684317055963e-06, + "loss": 0.9712, + "step": 51409 + }, + { + "epoch": 3.8, + "learning_rate": 6.2663142876301445e-06, + "loss": 1.0578, + "step": 51410 + }, + { + "epoch": 3.8, + "learning_rate": 6.265944264144562e-06, + "loss": 0.9254, + "step": 51411 + }, + { + "epoch": 3.8, + "learning_rate": 6.2655742465998126e-06, + "loss": 0.938, + "step": 51412 + }, + { + "epoch": 3.8, + "learning_rate": 6.265204234996478e-06, + "loss": 0.9274, + "step": 51413 + }, + { + "epoch": 3.8, + "learning_rate": 6.264834229335152e-06, + "loss": 0.9656, + "step": 51414 + }, + { + "epoch": 3.8, + "learning_rate": 6.26446422961642e-06, + "loss": 1.111, + "step": 51415 + }, + { + "epoch": 3.8, + "learning_rate": 6.264094235840875e-06, + "loss": 0.9846, + "step": 51416 + }, + { + "epoch": 3.8, + "learning_rate": 6.2637242480091045e-06, + "loss": 0.9525, + "step": 51417 + }, + { + "epoch": 3.8, + "learning_rate": 6.263354266121687e-06, + "loss": 1.1077, + "step": 51418 + }, + { + "epoch": 3.8, + "learning_rate": 6.262984290179227e-06, + "loss": 0.9695, + "step": 51419 + }, + { + "epoch": 3.8, + "learning_rate": 6.262614320182299e-06, + "loss": 0.961, + "step": 51420 + }, + { + "epoch": 3.8, + "learning_rate": 6.262244356131504e-06, + "loss": 1.1707, + "step": 51421 + }, + { + "epoch": 3.8, + "learning_rate": 6.261874398027419e-06, + "loss": 1.0252, + "step": 51422 + }, + { + "epoch": 3.8, + "learning_rate": 6.261504445870646e-06, + "loss": 1.1026, + "step": 51423 + }, + { + "epoch": 3.8, + "learning_rate": 6.261134499661757e-06, + "loss": 0.9442, + "step": 51424 + }, + { + "epoch": 3.8, + "learning_rate": 6.260764559401357e-06, + "loss": 1.0096, + "step": 51425 + }, + { + "epoch": 3.8, + "learning_rate": 6.260394625090021e-06, + "loss": 1.1138, + "step": 51426 + }, + { + "epoch": 3.8, + "learning_rate": 6.260024696728351e-06, + "loss": 1.0633, + "step": 51427 + }, + { + "epoch": 3.8, + "learning_rate": 6.259654774316923e-06, + "loss": 0.9943, + "step": 51428 + }, + { + "epoch": 3.8, + "learning_rate": 6.259284857856332e-06, + "loss": 1.0124, + "step": 51429 + }, + { + "epoch": 3.8, + "learning_rate": 6.258914947347165e-06, + "loss": 1.0051, + "step": 51430 + }, + { + "epoch": 3.8, + "learning_rate": 6.25854504279001e-06, + "loss": 1.0721, + "step": 51431 + }, + { + "epoch": 3.8, + "learning_rate": 6.258175144185458e-06, + "loss": 1.0325, + "step": 51432 + }, + { + "epoch": 3.8, + "learning_rate": 6.257805251534093e-06, + "loss": 0.9276, + "step": 51433 + }, + { + "epoch": 3.8, + "learning_rate": 6.257435364836511e-06, + "loss": 1.105, + "step": 51434 + }, + { + "epoch": 3.8, + "learning_rate": 6.25706548409329e-06, + "loss": 0.9828, + "step": 51435 + }, + { + "epoch": 3.8, + "learning_rate": 6.256695609305028e-06, + "loss": 1.0303, + "step": 51436 + }, + { + "epoch": 3.8, + "learning_rate": 6.256325740472311e-06, + "loss": 1.0603, + "step": 51437 + }, + { + "epoch": 3.8, + "learning_rate": 6.255955877595723e-06, + "loss": 0.9748, + "step": 51438 + }, + { + "epoch": 3.8, + "learning_rate": 6.2555860206758565e-06, + "loss": 0.9423, + "step": 51439 + }, + { + "epoch": 3.8, + "learning_rate": 6.255216169713299e-06, + "loss": 0.8715, + "step": 51440 + }, + { + "epoch": 3.8, + "learning_rate": 6.254846324708642e-06, + "loss": 0.9745, + "step": 51441 + }, + { + "epoch": 3.8, + "learning_rate": 6.2544764856624655e-06, + "loss": 1.0829, + "step": 51442 + }, + { + "epoch": 3.8, + "learning_rate": 6.254106652575369e-06, + "loss": 0.9631, + "step": 51443 + }, + { + "epoch": 3.8, + "learning_rate": 6.253736825447929e-06, + "loss": 0.9991, + "step": 51444 + }, + { + "epoch": 3.8, + "learning_rate": 6.253367004280746e-06, + "loss": 1.0772, + "step": 51445 + }, + { + "epoch": 3.8, + "learning_rate": 6.252997189074402e-06, + "loss": 1.0365, + "step": 51446 + }, + { + "epoch": 3.8, + "learning_rate": 6.252627379829481e-06, + "loss": 0.9953, + "step": 51447 + }, + { + "epoch": 3.8, + "learning_rate": 6.252257576546581e-06, + "loss": 0.9548, + "step": 51448 + }, + { + "epoch": 3.8, + "learning_rate": 6.251887779226282e-06, + "loss": 0.9335, + "step": 51449 + }, + { + "epoch": 3.8, + "learning_rate": 6.251517987869178e-06, + "loss": 0.8596, + "step": 51450 + }, + { + "epoch": 3.8, + "learning_rate": 6.251148202475854e-06, + "loss": 1.0415, + "step": 51451 + }, + { + "epoch": 3.8, + "learning_rate": 6.250778423046902e-06, + "loss": 0.9939, + "step": 51452 + }, + { + "epoch": 3.8, + "learning_rate": 6.250408649582904e-06, + "loss": 1.0608, + "step": 51453 + }, + { + "epoch": 3.8, + "learning_rate": 6.250038882084455e-06, + "loss": 1.0165, + "step": 51454 + }, + { + "epoch": 3.8, + "learning_rate": 6.249669120552141e-06, + "loss": 0.9008, + "step": 51455 + }, + { + "epoch": 3.8, + "learning_rate": 6.249299364986547e-06, + "loss": 0.8409, + "step": 51456 + }, + { + "epoch": 3.8, + "learning_rate": 6.248929615388266e-06, + "loss": 1.0011, + "step": 51457 + }, + { + "epoch": 3.8, + "learning_rate": 6.248559871757884e-06, + "loss": 1.1195, + "step": 51458 + }, + { + "epoch": 3.8, + "learning_rate": 6.248190134095991e-06, + "loss": 1.0471, + "step": 51459 + }, + { + "epoch": 3.8, + "learning_rate": 6.2478204024031706e-06, + "loss": 1.0985, + "step": 51460 + }, + { + "epoch": 3.8, + "learning_rate": 6.247450676680017e-06, + "loss": 0.9719, + "step": 51461 + }, + { + "epoch": 3.8, + "learning_rate": 6.247080956927114e-06, + "loss": 1.0079, + "step": 51462 + }, + { + "epoch": 3.8, + "learning_rate": 6.246711243145052e-06, + "loss": 0.955, + "step": 51463 + }, + { + "epoch": 3.8, + "learning_rate": 6.2463415353344215e-06, + "loss": 1.0317, + "step": 51464 + }, + { + "epoch": 3.8, + "learning_rate": 6.245971833495803e-06, + "loss": 1.0545, + "step": 51465 + }, + { + "epoch": 3.8, + "learning_rate": 6.245602137629795e-06, + "loss": 0.8925, + "step": 51466 + }, + { + "epoch": 3.8, + "learning_rate": 6.245232447736973e-06, + "loss": 1.0118, + "step": 51467 + }, + { + "epoch": 3.8, + "learning_rate": 6.244862763817941e-06, + "loss": 1.1397, + "step": 51468 + }, + { + "epoch": 3.8, + "learning_rate": 6.244493085873271e-06, + "loss": 1.0321, + "step": 51469 + }, + { + "epoch": 3.8, + "learning_rate": 6.244123413903564e-06, + "loss": 0.9955, + "step": 51470 + }, + { + "epoch": 3.8, + "learning_rate": 6.243753747909398e-06, + "loss": 1.0164, + "step": 51471 + }, + { + "epoch": 3.8, + "learning_rate": 6.243384087891372e-06, + "loss": 0.9929, + "step": 51472 + }, + { + "epoch": 3.8, + "learning_rate": 6.2430144338500665e-06, + "loss": 1.0247, + "step": 51473 + }, + { + "epoch": 3.8, + "learning_rate": 6.242644785786068e-06, + "loss": 0.9963, + "step": 51474 + }, + { + "epoch": 3.8, + "learning_rate": 6.242275143699972e-06, + "loss": 0.9789, + "step": 51475 + }, + { + "epoch": 3.8, + "learning_rate": 6.241905507592358e-06, + "loss": 0.881, + "step": 51476 + }, + { + "epoch": 3.8, + "learning_rate": 6.2415358774638224e-06, + "loss": 0.9709, + "step": 51477 + }, + { + "epoch": 3.8, + "learning_rate": 6.241166253314946e-06, + "loss": 0.9454, + "step": 51478 + }, + { + "epoch": 3.8, + "learning_rate": 6.240796635146324e-06, + "loss": 1.0646, + "step": 51479 + }, + { + "epoch": 3.8, + "learning_rate": 6.240427022958538e-06, + "loss": 0.994, + "step": 51480 + }, + { + "epoch": 3.8, + "learning_rate": 6.240057416752181e-06, + "loss": 1.0064, + "step": 51481 + }, + { + "epoch": 3.8, + "learning_rate": 6.239687816527835e-06, + "loss": 1.1136, + "step": 51482 + }, + { + "epoch": 3.8, + "learning_rate": 6.239318222286097e-06, + "loss": 1.0986, + "step": 51483 + }, + { + "epoch": 3.8, + "learning_rate": 6.238948634027549e-06, + "loss": 0.9012, + "step": 51484 + }, + { + "epoch": 3.8, + "learning_rate": 6.238579051752775e-06, + "loss": 0.9345, + "step": 51485 + }, + { + "epoch": 3.8, + "learning_rate": 6.238209475462375e-06, + "loss": 1.0741, + "step": 51486 + }, + { + "epoch": 3.8, + "learning_rate": 6.237839905156924e-06, + "loss": 0.9231, + "step": 51487 + }, + { + "epoch": 3.8, + "learning_rate": 6.23747034083702e-06, + "loss": 0.964, + "step": 51488 + }, + { + "epoch": 3.8, + "learning_rate": 6.237100782503242e-06, + "loss": 0.9459, + "step": 51489 + }, + { + "epoch": 3.8, + "learning_rate": 6.236731230156188e-06, + "loss": 1.0041, + "step": 51490 + }, + { + "epoch": 3.8, + "learning_rate": 6.236361683796436e-06, + "loss": 0.9672, + "step": 51491 + }, + { + "epoch": 3.8, + "learning_rate": 6.235992143424584e-06, + "loss": 0.9765, + "step": 51492 + }, + { + "epoch": 3.8, + "learning_rate": 6.2356226090412135e-06, + "loss": 1.0105, + "step": 51493 + }, + { + "epoch": 3.8, + "learning_rate": 6.23525308064691e-06, + "loss": 0.9506, + "step": 51494 + }, + { + "epoch": 3.8, + "learning_rate": 6.2348835582422684e-06, + "loss": 0.9871, + "step": 51495 + }, + { + "epoch": 3.81, + "learning_rate": 6.23451404182787e-06, + "loss": 0.9792, + "step": 51496 + }, + { + "epoch": 3.81, + "learning_rate": 6.234144531404309e-06, + "loss": 1.0148, + "step": 51497 + }, + { + "epoch": 3.81, + "learning_rate": 6.233775026972166e-06, + "loss": 0.9115, + "step": 51498 + }, + { + "epoch": 3.81, + "learning_rate": 6.2334055285320375e-06, + "loss": 0.9569, + "step": 51499 + }, + { + "epoch": 3.81, + "learning_rate": 6.233036036084504e-06, + "loss": 0.9839, + "step": 51500 + }, + { + "epoch": 3.81, + "learning_rate": 6.23266654963016e-06, + "loss": 0.8841, + "step": 51501 + }, + { + "epoch": 3.81, + "learning_rate": 6.232297069169586e-06, + "loss": 0.98, + "step": 51502 + }, + { + "epoch": 3.81, + "learning_rate": 6.231927594703374e-06, + "loss": 0.9958, + "step": 51503 + }, + { + "epoch": 3.81, + "learning_rate": 6.231558126232114e-06, + "loss": 1.0809, + "step": 51504 + }, + { + "epoch": 3.81, + "learning_rate": 6.231188663756387e-06, + "loss": 1.0045, + "step": 51505 + }, + { + "epoch": 3.81, + "learning_rate": 6.230819207276788e-06, + "loss": 1.1765, + "step": 51506 + }, + { + "epoch": 3.81, + "learning_rate": 6.2304497567939e-06, + "loss": 1.0129, + "step": 51507 + }, + { + "epoch": 3.81, + "learning_rate": 6.230080312308313e-06, + "loss": 1.0305, + "step": 51508 + }, + { + "epoch": 3.81, + "learning_rate": 6.229710873820614e-06, + "loss": 1.0901, + "step": 51509 + }, + { + "epoch": 3.81, + "learning_rate": 6.229341441331391e-06, + "loss": 0.9834, + "step": 51510 + }, + { + "epoch": 3.81, + "learning_rate": 6.228972014841236e-06, + "loss": 0.9117, + "step": 51511 + }, + { + "epoch": 3.81, + "learning_rate": 6.228602594350726e-06, + "loss": 1.0432, + "step": 51512 + }, + { + "epoch": 3.81, + "learning_rate": 6.228233179860461e-06, + "loss": 1.0356, + "step": 51513 + }, + { + "epoch": 3.81, + "learning_rate": 6.227863771371017e-06, + "loss": 0.9846, + "step": 51514 + }, + { + "epoch": 3.81, + "learning_rate": 6.227494368882994e-06, + "loss": 0.9948, + "step": 51515 + }, + { + "epoch": 3.81, + "learning_rate": 6.2271249723969675e-06, + "loss": 1.0247, + "step": 51516 + }, + { + "epoch": 3.81, + "learning_rate": 6.226755581913536e-06, + "loss": 1.0518, + "step": 51517 + }, + { + "epoch": 3.81, + "learning_rate": 6.226386197433278e-06, + "loss": 1.0011, + "step": 51518 + }, + { + "epoch": 3.81, + "learning_rate": 6.226016818956791e-06, + "loss": 1.0468, + "step": 51519 + }, + { + "epoch": 3.81, + "learning_rate": 6.225647446484655e-06, + "loss": 1.1169, + "step": 51520 + }, + { + "epoch": 3.81, + "learning_rate": 6.225278080017459e-06, + "loss": 1.0153, + "step": 51521 + }, + { + "epoch": 3.81, + "learning_rate": 6.224908719555791e-06, + "loss": 1.0436, + "step": 51522 + }, + { + "epoch": 3.81, + "learning_rate": 6.224539365100239e-06, + "loss": 0.8933, + "step": 51523 + }, + { + "epoch": 3.81, + "learning_rate": 6.224170016651393e-06, + "loss": 0.9945, + "step": 51524 + }, + { + "epoch": 3.81, + "learning_rate": 6.2238006742098365e-06, + "loss": 1.0805, + "step": 51525 + }, + { + "epoch": 3.81, + "learning_rate": 6.22343133777616e-06, + "loss": 0.9916, + "step": 51526 + }, + { + "epoch": 3.81, + "learning_rate": 6.223062007350949e-06, + "loss": 1.0768, + "step": 51527 + }, + { + "epoch": 3.81, + "learning_rate": 6.222692682934794e-06, + "loss": 0.9733, + "step": 51528 + }, + { + "epoch": 3.81, + "learning_rate": 6.22232336452828e-06, + "loss": 1.0851, + "step": 51529 + }, + { + "epoch": 3.81, + "learning_rate": 6.221954052131998e-06, + "loss": 0.9646, + "step": 51530 + }, + { + "epoch": 3.81, + "learning_rate": 6.221584745746533e-06, + "loss": 1.0384, + "step": 51531 + }, + { + "epoch": 3.81, + "learning_rate": 6.221215445372467e-06, + "loss": 0.9026, + "step": 51532 + }, + { + "epoch": 3.81, + "learning_rate": 6.2208461510104e-06, + "loss": 0.883, + "step": 51533 + }, + { + "epoch": 3.81, + "learning_rate": 6.220476862660906e-06, + "loss": 1.0191, + "step": 51534 + }, + { + "epoch": 3.81, + "learning_rate": 6.2201075803245855e-06, + "loss": 0.9321, + "step": 51535 + }, + { + "epoch": 3.81, + "learning_rate": 6.219738304002014e-06, + "loss": 0.9465, + "step": 51536 + }, + { + "epoch": 3.81, + "learning_rate": 6.219369033693789e-06, + "loss": 0.9672, + "step": 51537 + }, + { + "epoch": 3.81, + "learning_rate": 6.21899976940049e-06, + "loss": 0.9431, + "step": 51538 + }, + { + "epoch": 3.81, + "learning_rate": 6.218630511122714e-06, + "loss": 1.0176, + "step": 51539 + }, + { + "epoch": 3.81, + "learning_rate": 6.2182612588610404e-06, + "loss": 1.0114, + "step": 51540 + }, + { + "epoch": 3.81, + "learning_rate": 6.217892012616056e-06, + "loss": 0.9637, + "step": 51541 + }, + { + "epoch": 3.81, + "learning_rate": 6.217522772388355e-06, + "loss": 1.0442, + "step": 51542 + }, + { + "epoch": 3.81, + "learning_rate": 6.2171535381785175e-06, + "loss": 1.0856, + "step": 51543 + }, + { + "epoch": 3.81, + "learning_rate": 6.216784309987138e-06, + "loss": 0.9836, + "step": 51544 + }, + { + "epoch": 3.81, + "learning_rate": 6.216415087814798e-06, + "loss": 0.9163, + "step": 51545 + }, + { + "epoch": 3.81, + "learning_rate": 6.21604587166209e-06, + "loss": 1.0444, + "step": 51546 + }, + { + "epoch": 3.81, + "learning_rate": 6.215676661529596e-06, + "loss": 1.0004, + "step": 51547 + }, + { + "epoch": 3.81, + "learning_rate": 6.215307457417908e-06, + "loss": 0.989, + "step": 51548 + }, + { + "epoch": 3.81, + "learning_rate": 6.214938259327613e-06, + "loss": 1.2013, + "step": 51549 + }, + { + "epoch": 3.81, + "learning_rate": 6.214569067259294e-06, + "loss": 1.1179, + "step": 51550 + }, + { + "epoch": 3.81, + "learning_rate": 6.214199881213544e-06, + "loss": 0.9924, + "step": 51551 + }, + { + "epoch": 3.81, + "learning_rate": 6.213830701190946e-06, + "loss": 1.0138, + "step": 51552 + }, + { + "epoch": 3.81, + "learning_rate": 6.213461527192091e-06, + "loss": 0.8805, + "step": 51553 + }, + { + "epoch": 3.81, + "learning_rate": 6.213092359217562e-06, + "loss": 0.9933, + "step": 51554 + }, + { + "epoch": 3.81, + "learning_rate": 6.212723197267951e-06, + "loss": 1.0227, + "step": 51555 + }, + { + "epoch": 3.81, + "learning_rate": 6.2123540413438425e-06, + "loss": 1.0476, + "step": 51556 + }, + { + "epoch": 3.81, + "learning_rate": 6.211984891445825e-06, + "loss": 1.0066, + "step": 51557 + }, + { + "epoch": 3.81, + "learning_rate": 6.211615747574488e-06, + "loss": 0.9129, + "step": 51558 + }, + { + "epoch": 3.81, + "learning_rate": 6.21124660973041e-06, + "loss": 1.056, + "step": 51559 + }, + { + "epoch": 3.81, + "learning_rate": 6.2108774779141904e-06, + "loss": 0.8872, + "step": 51560 + }, + { + "epoch": 3.81, + "learning_rate": 6.2105083521264035e-06, + "loss": 0.9833, + "step": 51561 + }, + { + "epoch": 3.81, + "learning_rate": 6.210139232367651e-06, + "loss": 0.9396, + "step": 51562 + }, + { + "epoch": 3.81, + "learning_rate": 6.209770118638508e-06, + "loss": 0.9555, + "step": 51563 + }, + { + "epoch": 3.81, + "learning_rate": 6.209401010939568e-06, + "loss": 1.0751, + "step": 51564 + }, + { + "epoch": 3.81, + "learning_rate": 6.209031909271417e-06, + "loss": 1.0495, + "step": 51565 + }, + { + "epoch": 3.81, + "learning_rate": 6.208662813634643e-06, + "loss": 0.9961, + "step": 51566 + }, + { + "epoch": 3.81, + "learning_rate": 6.208293724029832e-06, + "loss": 1.0742, + "step": 51567 + }, + { + "epoch": 3.81, + "learning_rate": 6.207924640457569e-06, + "loss": 0.9842, + "step": 51568 + }, + { + "epoch": 3.81, + "learning_rate": 6.2075555629184455e-06, + "loss": 1.0398, + "step": 51569 + }, + { + "epoch": 3.81, + "learning_rate": 6.207186491413045e-06, + "loss": 0.9475, + "step": 51570 + }, + { + "epoch": 3.81, + "learning_rate": 6.20681742594196e-06, + "loss": 0.9229, + "step": 51571 + }, + { + "epoch": 3.81, + "learning_rate": 6.206448366505771e-06, + "loss": 1.0533, + "step": 51572 + }, + { + "epoch": 3.81, + "learning_rate": 6.206079313105069e-06, + "loss": 0.9287, + "step": 51573 + }, + { + "epoch": 3.81, + "learning_rate": 6.2057102657404415e-06, + "loss": 0.9916, + "step": 51574 + }, + { + "epoch": 3.81, + "learning_rate": 6.205341224412473e-06, + "loss": 1.0345, + "step": 51575 + }, + { + "epoch": 3.81, + "learning_rate": 6.204972189121757e-06, + "loss": 0.9965, + "step": 51576 + }, + { + "epoch": 3.81, + "learning_rate": 6.2046031598688694e-06, + "loss": 0.9618, + "step": 51577 + }, + { + "epoch": 3.81, + "learning_rate": 6.20423413665441e-06, + "loss": 0.9835, + "step": 51578 + }, + { + "epoch": 3.81, + "learning_rate": 6.203865119478953e-06, + "loss": 1.0094, + "step": 51579 + }, + { + "epoch": 3.81, + "learning_rate": 6.2034961083431e-06, + "loss": 1.1165, + "step": 51580 + }, + { + "epoch": 3.81, + "learning_rate": 6.203127103247423e-06, + "loss": 1.0619, + "step": 51581 + }, + { + "epoch": 3.81, + "learning_rate": 6.202758104192524e-06, + "loss": 0.9567, + "step": 51582 + }, + { + "epoch": 3.81, + "learning_rate": 6.202389111178975e-06, + "loss": 1.0318, + "step": 51583 + }, + { + "epoch": 3.81, + "learning_rate": 6.2020201242073755e-06, + "loss": 0.9068, + "step": 51584 + }, + { + "epoch": 3.81, + "learning_rate": 6.2016511432783034e-06, + "loss": 0.9426, + "step": 51585 + }, + { + "epoch": 3.81, + "learning_rate": 6.201282168392355e-06, + "loss": 0.9929, + "step": 51586 + }, + { + "epoch": 3.81, + "learning_rate": 6.200913199550112e-06, + "loss": 1.0203, + "step": 51587 + }, + { + "epoch": 3.81, + "learning_rate": 6.200544236752157e-06, + "loss": 1.0729, + "step": 51588 + }, + { + "epoch": 3.81, + "learning_rate": 6.200175279999086e-06, + "loss": 1.0815, + "step": 51589 + }, + { + "epoch": 3.81, + "learning_rate": 6.19980632929148e-06, + "loss": 1.0354, + "step": 51590 + }, + { + "epoch": 3.81, + "learning_rate": 6.199437384629929e-06, + "loss": 1.089, + "step": 51591 + }, + { + "epoch": 3.81, + "learning_rate": 6.199068446015015e-06, + "loss": 0.9992, + "step": 51592 + }, + { + "epoch": 3.81, + "learning_rate": 6.198699513447332e-06, + "loss": 0.9928, + "step": 51593 + }, + { + "epoch": 3.81, + "learning_rate": 6.198330586927463e-06, + "loss": 1.0527, + "step": 51594 + }, + { + "epoch": 3.81, + "learning_rate": 6.197961666455997e-06, + "loss": 1.0071, + "step": 51595 + }, + { + "epoch": 3.81, + "learning_rate": 6.1975927520335185e-06, + "loss": 0.9692, + "step": 51596 + }, + { + "epoch": 3.81, + "learning_rate": 6.197223843660616e-06, + "loss": 0.9794, + "step": 51597 + }, + { + "epoch": 3.81, + "learning_rate": 6.196854941337875e-06, + "loss": 0.9918, + "step": 51598 + }, + { + "epoch": 3.81, + "learning_rate": 6.196486045065884e-06, + "loss": 0.8401, + "step": 51599 + }, + { + "epoch": 3.81, + "learning_rate": 6.19611715484523e-06, + "loss": 0.9947, + "step": 51600 + }, + { + "epoch": 3.81, + "learning_rate": 6.195748270676496e-06, + "loss": 1.0022, + "step": 51601 + }, + { + "epoch": 3.81, + "learning_rate": 6.195379392560275e-06, + "loss": 0.8845, + "step": 51602 + }, + { + "epoch": 3.81, + "learning_rate": 6.195010520497149e-06, + "loss": 1.1071, + "step": 51603 + }, + { + "epoch": 3.81, + "learning_rate": 6.194641654487709e-06, + "loss": 0.931, + "step": 51604 + }, + { + "epoch": 3.81, + "learning_rate": 6.194272794532542e-06, + "loss": 0.925, + "step": 51605 + }, + { + "epoch": 3.81, + "learning_rate": 6.193903940632227e-06, + "loss": 0.9467, + "step": 51606 + }, + { + "epoch": 3.81, + "learning_rate": 6.19353509278736e-06, + "loss": 1.0355, + "step": 51607 + }, + { + "epoch": 3.81, + "learning_rate": 6.193166250998523e-06, + "loss": 0.9296, + "step": 51608 + }, + { + "epoch": 3.81, + "learning_rate": 6.192797415266304e-06, + "loss": 1.0468, + "step": 51609 + }, + { + "epoch": 3.81, + "learning_rate": 6.192428585591288e-06, + "loss": 1.0004, + "step": 51610 + }, + { + "epoch": 3.81, + "learning_rate": 6.192059761974067e-06, + "loss": 1.0224, + "step": 51611 + }, + { + "epoch": 3.81, + "learning_rate": 6.191690944415222e-06, + "loss": 0.9556, + "step": 51612 + }, + { + "epoch": 3.81, + "learning_rate": 6.1913221329153425e-06, + "loss": 1.0836, + "step": 51613 + }, + { + "epoch": 3.81, + "learning_rate": 6.190953327475018e-06, + "loss": 0.9254, + "step": 51614 + }, + { + "epoch": 3.81, + "learning_rate": 6.190584528094827e-06, + "loss": 1.0592, + "step": 51615 + }, + { + "epoch": 3.81, + "learning_rate": 6.1902157347753665e-06, + "loss": 1.0412, + "step": 51616 + }, + { + "epoch": 3.81, + "learning_rate": 6.189846947517215e-06, + "loss": 1.1185, + "step": 51617 + }, + { + "epoch": 3.81, + "learning_rate": 6.189478166320963e-06, + "loss": 1.0373, + "step": 51618 + }, + { + "epoch": 3.81, + "learning_rate": 6.189109391187198e-06, + "loss": 1.0605, + "step": 51619 + }, + { + "epoch": 3.81, + "learning_rate": 6.188740622116505e-06, + "loss": 0.9347, + "step": 51620 + }, + { + "epoch": 3.81, + "learning_rate": 6.18837185910947e-06, + "loss": 0.9786, + "step": 51621 + }, + { + "epoch": 3.81, + "learning_rate": 6.188003102166682e-06, + "loss": 0.9984, + "step": 51622 + }, + { + "epoch": 3.81, + "learning_rate": 6.1876343512887294e-06, + "loss": 0.9268, + "step": 51623 + }, + { + "epoch": 3.81, + "learning_rate": 6.187265606476189e-06, + "loss": 1.0563, + "step": 51624 + }, + { + "epoch": 3.81, + "learning_rate": 6.186896867729662e-06, + "loss": 0.9167, + "step": 51625 + }, + { + "epoch": 3.81, + "learning_rate": 6.186528135049721e-06, + "loss": 0.9541, + "step": 51626 + }, + { + "epoch": 3.81, + "learning_rate": 6.186159408436965e-06, + "loss": 1.092, + "step": 51627 + }, + { + "epoch": 3.81, + "learning_rate": 6.185790687891968e-06, + "loss": 0.8364, + "step": 51628 + }, + { + "epoch": 3.81, + "learning_rate": 6.18542197341533e-06, + "loss": 0.8837, + "step": 51629 + }, + { + "epoch": 3.81, + "learning_rate": 6.1850532650076255e-06, + "loss": 0.8752, + "step": 51630 + }, + { + "epoch": 3.81, + "learning_rate": 6.184684562669452e-06, + "loss": 0.9101, + "step": 51631 + }, + { + "epoch": 3.82, + "learning_rate": 6.184315866401389e-06, + "loss": 0.9645, + "step": 51632 + }, + { + "epoch": 3.82, + "learning_rate": 6.1839471762040225e-06, + "loss": 0.9978, + "step": 51633 + }, + { + "epoch": 3.82, + "learning_rate": 6.183578492077944e-06, + "loss": 0.9864, + "step": 51634 + }, + { + "epoch": 3.82, + "learning_rate": 6.183209814023736e-06, + "loss": 1.0811, + "step": 51635 + }, + { + "epoch": 3.82, + "learning_rate": 6.182841142041987e-06, + "loss": 1.0079, + "step": 51636 + }, + { + "epoch": 3.82, + "learning_rate": 6.182472476133283e-06, + "loss": 0.9885, + "step": 51637 + }, + { + "epoch": 3.82, + "learning_rate": 6.1821038162982114e-06, + "loss": 0.9069, + "step": 51638 + }, + { + "epoch": 3.82, + "learning_rate": 6.181735162537357e-06, + "loss": 0.9168, + "step": 51639 + }, + { + "epoch": 3.82, + "learning_rate": 6.181366514851308e-06, + "loss": 1.0612, + "step": 51640 + }, + { + "epoch": 3.82, + "learning_rate": 6.18099787324065e-06, + "loss": 1.0207, + "step": 51641 + }, + { + "epoch": 3.82, + "learning_rate": 6.1806292377059704e-06, + "loss": 0.9771, + "step": 51642 + }, + { + "epoch": 3.82, + "learning_rate": 6.180260608247855e-06, + "loss": 1.0033, + "step": 51643 + }, + { + "epoch": 3.82, + "learning_rate": 6.179891984866888e-06, + "loss": 1.0374, + "step": 51644 + }, + { + "epoch": 3.82, + "learning_rate": 6.179523367563662e-06, + "loss": 0.9433, + "step": 51645 + }, + { + "epoch": 3.82, + "learning_rate": 6.179154756338756e-06, + "loss": 1.082, + "step": 51646 + }, + { + "epoch": 3.82, + "learning_rate": 6.178786151192763e-06, + "loss": 0.8781, + "step": 51647 + }, + { + "epoch": 3.82, + "learning_rate": 6.178417552126263e-06, + "loss": 0.9482, + "step": 51648 + }, + { + "epoch": 3.82, + "learning_rate": 6.1780489591398504e-06, + "loss": 1.0011, + "step": 51649 + }, + { + "epoch": 3.82, + "learning_rate": 6.1776803722341035e-06, + "loss": 1.0595, + "step": 51650 + }, + { + "epoch": 3.82, + "learning_rate": 6.177311791409615e-06, + "loss": 1.017, + "step": 51651 + }, + { + "epoch": 3.82, + "learning_rate": 6.176943216666971e-06, + "loss": 0.9567, + "step": 51652 + }, + { + "epoch": 3.82, + "learning_rate": 6.1765746480067515e-06, + "loss": 0.8705, + "step": 51653 + }, + { + "epoch": 3.82, + "learning_rate": 6.176206085429548e-06, + "loss": 1.0983, + "step": 51654 + }, + { + "epoch": 3.82, + "learning_rate": 6.175837528935946e-06, + "loss": 0.909, + "step": 51655 + }, + { + "epoch": 3.82, + "learning_rate": 6.175468978526532e-06, + "loss": 0.9919, + "step": 51656 + }, + { + "epoch": 3.82, + "learning_rate": 6.17510043420189e-06, + "loss": 1.0209, + "step": 51657 + }, + { + "epoch": 3.82, + "learning_rate": 6.174731895962611e-06, + "loss": 1.0981, + "step": 51658 + }, + { + "epoch": 3.82, + "learning_rate": 6.174363363809278e-06, + "loss": 0.9601, + "step": 51659 + }, + { + "epoch": 3.82, + "learning_rate": 6.173994837742479e-06, + "loss": 1.1265, + "step": 51660 + }, + { + "epoch": 3.82, + "learning_rate": 6.1736263177628e-06, + "loss": 0.963, + "step": 51661 + }, + { + "epoch": 3.82, + "learning_rate": 6.173257803870825e-06, + "loss": 0.9818, + "step": 51662 + }, + { + "epoch": 3.82, + "learning_rate": 6.1728892960671425e-06, + "loss": 1.0104, + "step": 51663 + }, + { + "epoch": 3.82, + "learning_rate": 6.1725207943523365e-06, + "loss": 1.2046, + "step": 51664 + }, + { + "epoch": 3.82, + "learning_rate": 6.1721522987269986e-06, + "loss": 0.9875, + "step": 51665 + }, + { + "epoch": 3.82, + "learning_rate": 6.171783809191709e-06, + "loss": 0.9395, + "step": 51666 + }, + { + "epoch": 3.82, + "learning_rate": 6.171415325747059e-06, + "loss": 1.0709, + "step": 51667 + }, + { + "epoch": 3.82, + "learning_rate": 6.17104684839363e-06, + "loss": 1.0557, + "step": 51668 + }, + { + "epoch": 3.82, + "learning_rate": 6.170678377132013e-06, + "loss": 0.8922, + "step": 51669 + }, + { + "epoch": 3.82, + "learning_rate": 6.170309911962793e-06, + "loss": 0.9894, + "step": 51670 + }, + { + "epoch": 3.82, + "learning_rate": 6.16994145288655e-06, + "loss": 1.1152, + "step": 51671 + }, + { + "epoch": 3.82, + "learning_rate": 6.169572999903881e-06, + "loss": 0.9263, + "step": 51672 + }, + { + "epoch": 3.82, + "learning_rate": 6.169204553015361e-06, + "loss": 0.9438, + "step": 51673 + }, + { + "epoch": 3.82, + "learning_rate": 6.168836112221588e-06, + "loss": 1.0298, + "step": 51674 + }, + { + "epoch": 3.82, + "learning_rate": 6.168467677523136e-06, + "loss": 1.0352, + "step": 51675 + }, + { + "epoch": 3.82, + "learning_rate": 6.168099248920601e-06, + "loss": 0.938, + "step": 51676 + }, + { + "epoch": 3.82, + "learning_rate": 6.167730826414562e-06, + "loss": 1.0492, + "step": 51677 + }, + { + "epoch": 3.82, + "learning_rate": 6.167362410005613e-06, + "loss": 0.9823, + "step": 51678 + }, + { + "epoch": 3.82, + "learning_rate": 6.166993999694333e-06, + "loss": 1.0047, + "step": 51679 + }, + { + "epoch": 3.82, + "learning_rate": 6.1666255954813105e-06, + "loss": 1.0361, + "step": 51680 + }, + { + "epoch": 3.82, + "learning_rate": 6.166257197367131e-06, + "loss": 0.9468, + "step": 51681 + }, + { + "epoch": 3.82, + "learning_rate": 6.165888805352382e-06, + "loss": 0.9876, + "step": 51682 + }, + { + "epoch": 3.82, + "learning_rate": 6.16552041943765e-06, + "loss": 1.026, + "step": 51683 + }, + { + "epoch": 3.82, + "learning_rate": 6.165152039623519e-06, + "loss": 1.0069, + "step": 51684 + }, + { + "epoch": 3.82, + "learning_rate": 6.1647836659105775e-06, + "loss": 1.0822, + "step": 51685 + }, + { + "epoch": 3.82, + "learning_rate": 6.164415298299408e-06, + "loss": 0.9131, + "step": 51686 + }, + { + "epoch": 3.82, + "learning_rate": 6.164046936790602e-06, + "loss": 0.9956, + "step": 51687 + }, + { + "epoch": 3.82, + "learning_rate": 6.163678581384742e-06, + "loss": 0.8704, + "step": 51688 + }, + { + "epoch": 3.82, + "learning_rate": 6.163310232082413e-06, + "loss": 0.8877, + "step": 51689 + }, + { + "epoch": 3.82, + "learning_rate": 6.162941888884203e-06, + "loss": 0.9929, + "step": 51690 + }, + { + "epoch": 3.82, + "learning_rate": 6.162573551790696e-06, + "loss": 1.0249, + "step": 51691 + }, + { + "epoch": 3.82, + "learning_rate": 6.1622052208024815e-06, + "loss": 1.0066, + "step": 51692 + }, + { + "epoch": 3.82, + "learning_rate": 6.1618368959201435e-06, + "loss": 1.0769, + "step": 51693 + }, + { + "epoch": 3.82, + "learning_rate": 6.1614685771442675e-06, + "loss": 0.9722, + "step": 51694 + }, + { + "epoch": 3.82, + "learning_rate": 6.161100264475439e-06, + "loss": 0.9755, + "step": 51695 + }, + { + "epoch": 3.82, + "learning_rate": 6.16073195791425e-06, + "loss": 1.0347, + "step": 51696 + }, + { + "epoch": 3.82, + "learning_rate": 6.160363657461275e-06, + "loss": 1.0106, + "step": 51697 + }, + { + "epoch": 3.82, + "learning_rate": 6.159995363117113e-06, + "loss": 0.9832, + "step": 51698 + }, + { + "epoch": 3.82, + "learning_rate": 6.15962707488234e-06, + "loss": 0.964, + "step": 51699 + }, + { + "epoch": 3.82, + "learning_rate": 6.159258792757543e-06, + "loss": 1.0356, + "step": 51700 + }, + { + "epoch": 3.82, + "learning_rate": 6.158890516743314e-06, + "loss": 1.0268, + "step": 51701 + }, + { + "epoch": 3.82, + "learning_rate": 6.158522246840233e-06, + "loss": 0.9895, + "step": 51702 + }, + { + "epoch": 3.82, + "learning_rate": 6.158153983048889e-06, + "loss": 0.9642, + "step": 51703 + }, + { + "epoch": 3.82, + "learning_rate": 6.157785725369867e-06, + "loss": 1.0544, + "step": 51704 + }, + { + "epoch": 3.82, + "learning_rate": 6.157417473803753e-06, + "loss": 1.1007, + "step": 51705 + }, + { + "epoch": 3.82, + "learning_rate": 6.1570492283511315e-06, + "loss": 1.0731, + "step": 51706 + }, + { + "epoch": 3.82, + "learning_rate": 6.156680989012592e-06, + "loss": 1.0175, + "step": 51707 + }, + { + "epoch": 3.82, + "learning_rate": 6.156312755788718e-06, + "loss": 0.9595, + "step": 51708 + }, + { + "epoch": 3.82, + "learning_rate": 6.155944528680093e-06, + "loss": 0.8879, + "step": 51709 + }, + { + "epoch": 3.82, + "learning_rate": 6.155576307687308e-06, + "loss": 0.8993, + "step": 51710 + }, + { + "epoch": 3.82, + "learning_rate": 6.155208092810943e-06, + "loss": 1.0411, + "step": 51711 + }, + { + "epoch": 3.82, + "learning_rate": 6.154839884051589e-06, + "loss": 1.0286, + "step": 51712 + }, + { + "epoch": 3.82, + "learning_rate": 6.154471681409829e-06, + "loss": 1.0251, + "step": 51713 + }, + { + "epoch": 3.82, + "learning_rate": 6.154103484886251e-06, + "loss": 0.9954, + "step": 51714 + }, + { + "epoch": 3.82, + "learning_rate": 6.153735294481437e-06, + "loss": 1.0478, + "step": 51715 + }, + { + "epoch": 3.82, + "learning_rate": 6.153367110195977e-06, + "loss": 0.9955, + "step": 51716 + }, + { + "epoch": 3.82, + "learning_rate": 6.152998932030456e-06, + "loss": 1.0644, + "step": 51717 + }, + { + "epoch": 3.82, + "learning_rate": 6.152630759985454e-06, + "loss": 0.9755, + "step": 51718 + }, + { + "epoch": 3.82, + "learning_rate": 6.1522625940615674e-06, + "loss": 0.9475, + "step": 51719 + }, + { + "epoch": 3.82, + "learning_rate": 6.15189443425937e-06, + "loss": 1.1001, + "step": 51720 + }, + { + "epoch": 3.82, + "learning_rate": 6.151526280579459e-06, + "loss": 1.1125, + "step": 51721 + }, + { + "epoch": 3.82, + "learning_rate": 6.151158133022409e-06, + "loss": 0.9888, + "step": 51722 + }, + { + "epoch": 3.82, + "learning_rate": 6.150789991588817e-06, + "loss": 0.9618, + "step": 51723 + }, + { + "epoch": 3.82, + "learning_rate": 6.150421856279258e-06, + "loss": 0.9956, + "step": 51724 + }, + { + "epoch": 3.82, + "learning_rate": 6.1500537270943275e-06, + "loss": 1.0295, + "step": 51725 + }, + { + "epoch": 3.82, + "learning_rate": 6.149685604034605e-06, + "loss": 1.0066, + "step": 51726 + }, + { + "epoch": 3.82, + "learning_rate": 6.149317487100676e-06, + "loss": 1.0329, + "step": 51727 + }, + { + "epoch": 3.82, + "learning_rate": 6.148949376293129e-06, + "loss": 0.867, + "step": 51728 + }, + { + "epoch": 3.82, + "learning_rate": 6.148581271612546e-06, + "loss": 1.0492, + "step": 51729 + }, + { + "epoch": 3.82, + "learning_rate": 6.1482131730595186e-06, + "loss": 1.0135, + "step": 51730 + }, + { + "epoch": 3.82, + "learning_rate": 6.147845080634627e-06, + "loss": 1.0387, + "step": 51731 + }, + { + "epoch": 3.82, + "learning_rate": 6.147476994338459e-06, + "loss": 0.9523, + "step": 51732 + }, + { + "epoch": 3.82, + "learning_rate": 6.147108914171599e-06, + "loss": 1.085, + "step": 51733 + }, + { + "epoch": 3.82, + "learning_rate": 6.146740840134635e-06, + "loss": 1.1453, + "step": 51734 + }, + { + "epoch": 3.82, + "learning_rate": 6.146372772228153e-06, + "loss": 1.0436, + "step": 51735 + }, + { + "epoch": 3.82, + "learning_rate": 6.146004710452733e-06, + "loss": 0.9879, + "step": 51736 + }, + { + "epoch": 3.82, + "learning_rate": 6.145636654808967e-06, + "loss": 1.0068, + "step": 51737 + }, + { + "epoch": 3.82, + "learning_rate": 6.145268605297435e-06, + "loss": 1.0053, + "step": 51738 + }, + { + "epoch": 3.82, + "learning_rate": 6.144900561918728e-06, + "loss": 0.8789, + "step": 51739 + }, + { + "epoch": 3.82, + "learning_rate": 6.144532524673428e-06, + "loss": 0.9808, + "step": 51740 + }, + { + "epoch": 3.82, + "learning_rate": 6.144164493562124e-06, + "loss": 0.9765, + "step": 51741 + }, + { + "epoch": 3.82, + "learning_rate": 6.1437964685853945e-06, + "loss": 0.9951, + "step": 51742 + }, + { + "epoch": 3.82, + "learning_rate": 6.143428449743836e-06, + "loss": 1.0463, + "step": 51743 + }, + { + "epoch": 3.82, + "learning_rate": 6.14306043703802e-06, + "loss": 1.0478, + "step": 51744 + }, + { + "epoch": 3.82, + "learning_rate": 6.142692430468547e-06, + "loss": 0.9905, + "step": 51745 + }, + { + "epoch": 3.82, + "learning_rate": 6.142324430035993e-06, + "loss": 1.0053, + "step": 51746 + }, + { + "epoch": 3.82, + "learning_rate": 6.141956435740943e-06, + "loss": 0.9712, + "step": 51747 + }, + { + "epoch": 3.82, + "learning_rate": 6.1415884475839885e-06, + "loss": 1.0058, + "step": 51748 + }, + { + "epoch": 3.82, + "learning_rate": 6.141220465565707e-06, + "loss": 0.8601, + "step": 51749 + }, + { + "epoch": 3.82, + "learning_rate": 6.140852489686693e-06, + "loss": 1.0076, + "step": 51750 + }, + { + "epoch": 3.82, + "learning_rate": 6.1404845199475234e-06, + "loss": 0.9256, + "step": 51751 + }, + { + "epoch": 3.82, + "learning_rate": 6.140116556348792e-06, + "loss": 1.0513, + "step": 51752 + }, + { + "epoch": 3.82, + "learning_rate": 6.1397485988910754e-06, + "loss": 0.9536, + "step": 51753 + }, + { + "epoch": 3.82, + "learning_rate": 6.1393806475749675e-06, + "loss": 0.9475, + "step": 51754 + }, + { + "epoch": 3.82, + "learning_rate": 6.1390127024010495e-06, + "loss": 0.8469, + "step": 51755 + }, + { + "epoch": 3.82, + "learning_rate": 6.138644763369904e-06, + "loss": 1.0524, + "step": 51756 + }, + { + "epoch": 3.82, + "learning_rate": 6.138276830482123e-06, + "loss": 1.044, + "step": 51757 + }, + { + "epoch": 3.82, + "learning_rate": 6.137908903738286e-06, + "loss": 0.9183, + "step": 51758 + }, + { + "epoch": 3.82, + "learning_rate": 6.1375409831389805e-06, + "loss": 1.0947, + "step": 51759 + }, + { + "epoch": 3.82, + "learning_rate": 6.137173068684792e-06, + "loss": 0.9937, + "step": 51760 + }, + { + "epoch": 3.82, + "learning_rate": 6.136805160376308e-06, + "loss": 1.0184, + "step": 51761 + }, + { + "epoch": 3.82, + "learning_rate": 6.136437258214109e-06, + "loss": 1.0818, + "step": 51762 + }, + { + "epoch": 3.82, + "learning_rate": 6.136069362198785e-06, + "loss": 1.0671, + "step": 51763 + }, + { + "epoch": 3.82, + "learning_rate": 6.135701472330922e-06, + "loss": 0.9774, + "step": 51764 + }, + { + "epoch": 3.82, + "learning_rate": 6.135333588611097e-06, + "loss": 0.9331, + "step": 51765 + }, + { + "epoch": 3.82, + "learning_rate": 6.134965711039907e-06, + "loss": 0.963, + "step": 51766 + }, + { + "epoch": 3.83, + "learning_rate": 6.134597839617923e-06, + "loss": 0.9586, + "step": 51767 + }, + { + "epoch": 3.83, + "learning_rate": 6.134229974345747e-06, + "loss": 1.0567, + "step": 51768 + }, + { + "epoch": 3.83, + "learning_rate": 6.133862115223948e-06, + "loss": 0.8851, + "step": 51769 + }, + { + "epoch": 3.83, + "learning_rate": 6.133494262253126e-06, + "loss": 1.0583, + "step": 51770 + }, + { + "epoch": 3.83, + "learning_rate": 6.133126415433853e-06, + "loss": 0.9397, + "step": 51771 + }, + { + "epoch": 3.83, + "learning_rate": 6.1327585747667264e-06, + "loss": 0.9551, + "step": 51772 + }, + { + "epoch": 3.83, + "learning_rate": 6.132390740252325e-06, + "loss": 1.0212, + "step": 51773 + }, + { + "epoch": 3.83, + "learning_rate": 6.132022911891231e-06, + "loss": 1.0242, + "step": 51774 + }, + { + "epoch": 3.83, + "learning_rate": 6.131655089684037e-06, + "loss": 1.0053, + "step": 51775 + }, + { + "epoch": 3.83, + "learning_rate": 6.13128727363132e-06, + "loss": 0.9812, + "step": 51776 + }, + { + "epoch": 3.83, + "learning_rate": 6.130919463733674e-06, + "loss": 0.9613, + "step": 51777 + }, + { + "epoch": 3.83, + "learning_rate": 6.130551659991675e-06, + "loss": 1.0002, + "step": 51778 + }, + { + "epoch": 3.83, + "learning_rate": 6.130183862405916e-06, + "loss": 0.9883, + "step": 51779 + }, + { + "epoch": 3.83, + "learning_rate": 6.129816070976978e-06, + "loss": 0.9987, + "step": 51780 + }, + { + "epoch": 3.83, + "learning_rate": 6.1294482857054495e-06, + "loss": 0.9802, + "step": 51781 + }, + { + "epoch": 3.83, + "learning_rate": 6.129080506591912e-06, + "loss": 1.0863, + "step": 51782 + }, + { + "epoch": 3.83, + "learning_rate": 6.128712733636951e-06, + "loss": 1.0471, + "step": 51783 + }, + { + "epoch": 3.83, + "learning_rate": 6.1283449668411555e-06, + "loss": 0.9719, + "step": 51784 + }, + { + "epoch": 3.83, + "learning_rate": 6.127977206205105e-06, + "loss": 1.0813, + "step": 51785 + }, + { + "epoch": 3.83, + "learning_rate": 6.12760945172939e-06, + "loss": 1.0651, + "step": 51786 + }, + { + "epoch": 3.83, + "learning_rate": 6.1272417034145885e-06, + "loss": 0.89, + "step": 51787 + }, + { + "epoch": 3.83, + "learning_rate": 6.126873961261295e-06, + "loss": 1.021, + "step": 51788 + }, + { + "epoch": 3.83, + "learning_rate": 6.126506225270084e-06, + "loss": 0.9612, + "step": 51789 + }, + { + "epoch": 3.83, + "learning_rate": 6.1261384954415524e-06, + "loss": 0.9517, + "step": 51790 + }, + { + "epoch": 3.83, + "learning_rate": 6.125770771776277e-06, + "loss": 0.9868, + "step": 51791 + }, + { + "epoch": 3.83, + "learning_rate": 6.125403054274844e-06, + "loss": 0.9293, + "step": 51792 + }, + { + "epoch": 3.83, + "learning_rate": 6.1250353429378394e-06, + "loss": 1.0617, + "step": 51793 + }, + { + "epoch": 3.83, + "learning_rate": 6.124667637765847e-06, + "loss": 1.0632, + "step": 51794 + }, + { + "epoch": 3.83, + "learning_rate": 6.124299938759455e-06, + "loss": 1.016, + "step": 51795 + }, + { + "epoch": 3.83, + "learning_rate": 6.123932245919244e-06, + "loss": 1.08, + "step": 51796 + }, + { + "epoch": 3.83, + "learning_rate": 6.123564559245803e-06, + "loss": 1.0052, + "step": 51797 + }, + { + "epoch": 3.83, + "learning_rate": 6.123196878739715e-06, + "loss": 0.947, + "step": 51798 + }, + { + "epoch": 3.83, + "learning_rate": 6.1228292044015645e-06, + "loss": 0.971, + "step": 51799 + }, + { + "epoch": 3.83, + "learning_rate": 6.122461536231938e-06, + "loss": 1.0494, + "step": 51800 + }, + { + "epoch": 3.83, + "learning_rate": 6.122093874231421e-06, + "loss": 0.9651, + "step": 51801 + }, + { + "epoch": 3.83, + "learning_rate": 6.121726218400595e-06, + "loss": 0.9339, + "step": 51802 + }, + { + "epoch": 3.83, + "learning_rate": 6.121358568740048e-06, + "loss": 1.0668, + "step": 51803 + }, + { + "epoch": 3.83, + "learning_rate": 6.1209909252503655e-06, + "loss": 1.1003, + "step": 51804 + }, + { + "epoch": 3.83, + "learning_rate": 6.120623287932127e-06, + "loss": 1.005, + "step": 51805 + }, + { + "epoch": 3.83, + "learning_rate": 6.120255656785925e-06, + "loss": 1.0518, + "step": 51806 + }, + { + "epoch": 3.83, + "learning_rate": 6.1198880318123376e-06, + "loss": 1.0568, + "step": 51807 + }, + { + "epoch": 3.83, + "learning_rate": 6.1195204130119566e-06, + "loss": 0.9941, + "step": 51808 + }, + { + "epoch": 3.83, + "learning_rate": 6.11915280038536e-06, + "loss": 0.9995, + "step": 51809 + }, + { + "epoch": 3.83, + "learning_rate": 6.118785193933138e-06, + "loss": 1.0418, + "step": 51810 + }, + { + "epoch": 3.83, + "learning_rate": 6.118417593655874e-06, + "loss": 0.9609, + "step": 51811 + }, + { + "epoch": 3.83, + "learning_rate": 6.118049999554148e-06, + "loss": 1.0247, + "step": 51812 + }, + { + "epoch": 3.83, + "learning_rate": 6.117682411628554e-06, + "loss": 0.9557, + "step": 51813 + }, + { + "epoch": 3.83, + "learning_rate": 6.117314829879666e-06, + "loss": 1.0404, + "step": 51814 + }, + { + "epoch": 3.83, + "learning_rate": 6.116947254308081e-06, + "loss": 0.9506, + "step": 51815 + }, + { + "epoch": 3.83, + "learning_rate": 6.11657968491437e-06, + "loss": 0.9803, + "step": 51816 + }, + { + "epoch": 3.83, + "learning_rate": 6.116212121699132e-06, + "loss": 0.9756, + "step": 51817 + }, + { + "epoch": 3.83, + "learning_rate": 6.115844564662941e-06, + "loss": 1.02, + "step": 51818 + }, + { + "epoch": 3.83, + "learning_rate": 6.115477013806388e-06, + "loss": 0.9787, + "step": 51819 + }, + { + "epoch": 3.83, + "learning_rate": 6.115109469130054e-06, + "loss": 0.9929, + "step": 51820 + }, + { + "epoch": 3.83, + "learning_rate": 6.114741930634525e-06, + "loss": 1.0554, + "step": 51821 + }, + { + "epoch": 3.83, + "learning_rate": 6.114374398320388e-06, + "loss": 1.0757, + "step": 51822 + }, + { + "epoch": 3.83, + "learning_rate": 6.114006872188223e-06, + "loss": 0.9259, + "step": 51823 + }, + { + "epoch": 3.83, + "learning_rate": 6.113639352238618e-06, + "loss": 1.0058, + "step": 51824 + }, + { + "epoch": 3.83, + "learning_rate": 6.113271838472158e-06, + "loss": 1.0452, + "step": 51825 + }, + { + "epoch": 3.83, + "learning_rate": 6.112904330889427e-06, + "loss": 1.0351, + "step": 51826 + }, + { + "epoch": 3.83, + "learning_rate": 6.112536829491009e-06, + "loss": 0.9384, + "step": 51827 + }, + { + "epoch": 3.83, + "learning_rate": 6.11216933427749e-06, + "loss": 1.0777, + "step": 51828 + }, + { + "epoch": 3.83, + "learning_rate": 6.111801845249457e-06, + "loss": 1.1037, + "step": 51829 + }, + { + "epoch": 3.83, + "learning_rate": 6.111434362407484e-06, + "loss": 0.9604, + "step": 51830 + }, + { + "epoch": 3.83, + "learning_rate": 6.11106688575217e-06, + "loss": 0.9185, + "step": 51831 + }, + { + "epoch": 3.83, + "learning_rate": 6.110699415284088e-06, + "loss": 0.9064, + "step": 51832 + }, + { + "epoch": 3.83, + "learning_rate": 6.110331951003834e-06, + "loss": 1.0505, + "step": 51833 + }, + { + "epoch": 3.83, + "learning_rate": 6.109964492911979e-06, + "loss": 1.0101, + "step": 51834 + }, + { + "epoch": 3.83, + "learning_rate": 6.109597041009121e-06, + "loss": 1.027, + "step": 51835 + }, + { + "epoch": 3.83, + "learning_rate": 6.109229595295831e-06, + "loss": 1.0006, + "step": 51836 + }, + { + "epoch": 3.83, + "learning_rate": 6.108862155772709e-06, + "loss": 1.0029, + "step": 51837 + }, + { + "epoch": 3.83, + "learning_rate": 6.108494722440329e-06, + "loss": 0.9928, + "step": 51838 + }, + { + "epoch": 3.83, + "learning_rate": 6.108127295299277e-06, + "loss": 0.9078, + "step": 51839 + }, + { + "epoch": 3.83, + "learning_rate": 6.10775987435014e-06, + "loss": 0.8972, + "step": 51840 + }, + { + "epoch": 3.83, + "learning_rate": 6.1073924595935e-06, + "loss": 0.9481, + "step": 51841 + }, + { + "epoch": 3.83, + "learning_rate": 6.107025051029943e-06, + "loss": 0.9708, + "step": 51842 + }, + { + "epoch": 3.83, + "learning_rate": 6.106657648660053e-06, + "loss": 1.0258, + "step": 51843 + }, + { + "epoch": 3.83, + "learning_rate": 6.106290252484417e-06, + "loss": 1.0723, + "step": 51844 + }, + { + "epoch": 3.83, + "learning_rate": 6.105922862503615e-06, + "loss": 0.9671, + "step": 51845 + }, + { + "epoch": 3.83, + "learning_rate": 6.105555478718237e-06, + "loss": 0.9949, + "step": 51846 + }, + { + "epoch": 3.83, + "learning_rate": 6.105188101128863e-06, + "loss": 1.0106, + "step": 51847 + }, + { + "epoch": 3.83, + "learning_rate": 6.104820729736078e-06, + "loss": 0.9992, + "step": 51848 + }, + { + "epoch": 3.83, + "learning_rate": 6.10445336454047e-06, + "loss": 1.0394, + "step": 51849 + }, + { + "epoch": 3.83, + "learning_rate": 6.1040860055426174e-06, + "loss": 0.9359, + "step": 51850 + }, + { + "epoch": 3.83, + "learning_rate": 6.103718652743111e-06, + "loss": 1.0663, + "step": 51851 + }, + { + "epoch": 3.83, + "learning_rate": 6.103351306142531e-06, + "loss": 1.0791, + "step": 51852 + }, + { + "epoch": 3.83, + "learning_rate": 6.102983965741465e-06, + "loss": 1.0698, + "step": 51853 + }, + { + "epoch": 3.83, + "learning_rate": 6.102616631540493e-06, + "loss": 0.9232, + "step": 51854 + }, + { + "epoch": 3.83, + "learning_rate": 6.102249303540205e-06, + "loss": 1.0163, + "step": 51855 + }, + { + "epoch": 3.83, + "learning_rate": 6.10188198174118e-06, + "loss": 0.931, + "step": 51856 + }, + { + "epoch": 3.83, + "learning_rate": 6.101514666144009e-06, + "loss": 1.0027, + "step": 51857 + }, + { + "epoch": 3.83, + "learning_rate": 6.101147356749272e-06, + "loss": 0.9421, + "step": 51858 + }, + { + "epoch": 3.83, + "learning_rate": 6.100780053557549e-06, + "loss": 0.9655, + "step": 51859 + }, + { + "epoch": 3.83, + "learning_rate": 6.100412756569434e-06, + "loss": 0.9776, + "step": 51860 + }, + { + "epoch": 3.83, + "learning_rate": 6.100045465785501e-06, + "loss": 0.946, + "step": 51861 + }, + { + "epoch": 3.83, + "learning_rate": 6.099678181206346e-06, + "loss": 0.9698, + "step": 51862 + }, + { + "epoch": 3.83, + "learning_rate": 6.099310902832544e-06, + "loss": 0.9875, + "step": 51863 + }, + { + "epoch": 3.83, + "learning_rate": 6.098943630664683e-06, + "loss": 1.1147, + "step": 51864 + }, + { + "epoch": 3.83, + "learning_rate": 6.098576364703345e-06, + "loss": 1.0584, + "step": 51865 + }, + { + "epoch": 3.83, + "learning_rate": 6.09820910494912e-06, + "loss": 0.9932, + "step": 51866 + }, + { + "epoch": 3.83, + "learning_rate": 6.097841851402587e-06, + "loss": 1.0229, + "step": 51867 + }, + { + "epoch": 3.83, + "learning_rate": 6.097474604064331e-06, + "loss": 0.9865, + "step": 51868 + }, + { + "epoch": 3.83, + "learning_rate": 6.097107362934938e-06, + "loss": 0.9152, + "step": 51869 + }, + { + "epoch": 3.83, + "learning_rate": 6.0967401280149886e-06, + "loss": 0.978, + "step": 51870 + }, + { + "epoch": 3.83, + "learning_rate": 6.096372899305074e-06, + "loss": 1.0723, + "step": 51871 + }, + { + "epoch": 3.83, + "learning_rate": 6.096005676805769e-06, + "loss": 0.9661, + "step": 51872 + }, + { + "epoch": 3.83, + "learning_rate": 6.095638460517669e-06, + "loss": 1.0936, + "step": 51873 + }, + { + "epoch": 3.83, + "learning_rate": 6.095271250441348e-06, + "loss": 0.9728, + "step": 51874 + }, + { + "epoch": 3.83, + "learning_rate": 6.0949040465773965e-06, + "loss": 1.1284, + "step": 51875 + }, + { + "epoch": 3.83, + "learning_rate": 6.0945368489264005e-06, + "loss": 0.9509, + "step": 51876 + }, + { + "epoch": 3.83, + "learning_rate": 6.094169657488933e-06, + "loss": 1.0216, + "step": 51877 + }, + { + "epoch": 3.83, + "learning_rate": 6.093802472265592e-06, + "loss": 1.0204, + "step": 51878 + }, + { + "epoch": 3.83, + "learning_rate": 6.093435293256948e-06, + "loss": 1.0021, + "step": 51879 + }, + { + "epoch": 3.83, + "learning_rate": 6.093068120463601e-06, + "loss": 1.0592, + "step": 51880 + }, + { + "epoch": 3.83, + "learning_rate": 6.092700953886119e-06, + "loss": 0.926, + "step": 51881 + }, + { + "epoch": 3.83, + "learning_rate": 6.092333793525101e-06, + "loss": 0.9122, + "step": 51882 + }, + { + "epoch": 3.83, + "learning_rate": 6.091966639381118e-06, + "loss": 1.0319, + "step": 51883 + }, + { + "epoch": 3.83, + "learning_rate": 6.091599491454766e-06, + "loss": 1.0311, + "step": 51884 + }, + { + "epoch": 3.83, + "learning_rate": 6.091232349746621e-06, + "loss": 0.8684, + "step": 51885 + }, + { + "epoch": 3.83, + "learning_rate": 6.090865214257267e-06, + "loss": 1.0189, + "step": 51886 + }, + { + "epoch": 3.83, + "learning_rate": 6.090498084987293e-06, + "loss": 0.9204, + "step": 51887 + }, + { + "epoch": 3.83, + "learning_rate": 6.0901309619372796e-06, + "loss": 0.8747, + "step": 51888 + }, + { + "epoch": 3.83, + "learning_rate": 6.089763845107813e-06, + "loss": 1.0057, + "step": 51889 + }, + { + "epoch": 3.83, + "learning_rate": 6.089396734499473e-06, + "loss": 1.059, + "step": 51890 + }, + { + "epoch": 3.83, + "learning_rate": 6.089029630112851e-06, + "loss": 1.132, + "step": 51891 + }, + { + "epoch": 3.83, + "learning_rate": 6.088662531948524e-06, + "loss": 0.952, + "step": 51892 + }, + { + "epoch": 3.83, + "learning_rate": 6.08829544000708e-06, + "loss": 0.9546, + "step": 51893 + }, + { + "epoch": 3.83, + "learning_rate": 6.087928354289103e-06, + "loss": 0.9277, + "step": 51894 + }, + { + "epoch": 3.83, + "learning_rate": 6.087561274795173e-06, + "loss": 1.0412, + "step": 51895 + }, + { + "epoch": 3.83, + "learning_rate": 6.08719420152588e-06, + "loss": 0.9739, + "step": 51896 + }, + { + "epoch": 3.83, + "learning_rate": 6.086827134481803e-06, + "loss": 0.9212, + "step": 51897 + }, + { + "epoch": 3.83, + "learning_rate": 6.08646007366353e-06, + "loss": 0.9759, + "step": 51898 + }, + { + "epoch": 3.83, + "learning_rate": 6.086093019071641e-06, + "loss": 1.0708, + "step": 51899 + }, + { + "epoch": 3.83, + "learning_rate": 6.085725970706724e-06, + "loss": 1.003, + "step": 51900 + }, + { + "epoch": 3.83, + "learning_rate": 6.085358928569358e-06, + "loss": 1.0391, + "step": 51901 + }, + { + "epoch": 3.84, + "learning_rate": 6.084991892660133e-06, + "loss": 0.9577, + "step": 51902 + }, + { + "epoch": 3.84, + "learning_rate": 6.084624862979631e-06, + "loss": 0.9774, + "step": 51903 + }, + { + "epoch": 3.84, + "learning_rate": 6.08425783952843e-06, + "loss": 0.9625, + "step": 51904 + }, + { + "epoch": 3.84, + "learning_rate": 6.083890822307125e-06, + "loss": 1.0641, + "step": 51905 + }, + { + "epoch": 3.84, + "learning_rate": 6.083523811316288e-06, + "loss": 1.1186, + "step": 51906 + }, + { + "epoch": 3.84, + "learning_rate": 6.0831568065565115e-06, + "loss": 1.0349, + "step": 51907 + }, + { + "epoch": 3.84, + "learning_rate": 6.0827898080283755e-06, + "loss": 0.9022, + "step": 51908 + }, + { + "epoch": 3.84, + "learning_rate": 6.082422815732465e-06, + "loss": 1.0117, + "step": 51909 + }, + { + "epoch": 3.84, + "learning_rate": 6.082055829669363e-06, + "loss": 0.925, + "step": 51910 + }, + { + "epoch": 3.84, + "learning_rate": 6.0816888498396555e-06, + "loss": 1.0152, + "step": 51911 + }, + { + "epoch": 3.84, + "learning_rate": 6.081321876243924e-06, + "loss": 1.021, + "step": 51912 + }, + { + "epoch": 3.84, + "learning_rate": 6.080954908882754e-06, + "loss": 1.049, + "step": 51913 + }, + { + "epoch": 3.84, + "learning_rate": 6.08058794775673e-06, + "loss": 0.9641, + "step": 51914 + }, + { + "epoch": 3.84, + "learning_rate": 6.0802209928664314e-06, + "loss": 0.9038, + "step": 51915 + }, + { + "epoch": 3.84, + "learning_rate": 6.0798540442124475e-06, + "loss": 1.1325, + "step": 51916 + }, + { + "epoch": 3.84, + "learning_rate": 6.079487101795357e-06, + "loss": 0.9474, + "step": 51917 + }, + { + "epoch": 3.84, + "learning_rate": 6.079120165615749e-06, + "loss": 0.9482, + "step": 51918 + }, + { + "epoch": 3.84, + "learning_rate": 6.078753235674204e-06, + "loss": 0.8865, + "step": 51919 + }, + { + "epoch": 3.84, + "learning_rate": 6.078386311971307e-06, + "loss": 1.0829, + "step": 51920 + }, + { + "epoch": 3.84, + "learning_rate": 6.0780193945076395e-06, + "loss": 0.9419, + "step": 51921 + }, + { + "epoch": 3.84, + "learning_rate": 6.077652483283789e-06, + "loss": 1.1139, + "step": 51922 + }, + { + "epoch": 3.84, + "learning_rate": 6.0772855783003395e-06, + "loss": 1.034, + "step": 51923 + }, + { + "epoch": 3.84, + "learning_rate": 6.076918679557866e-06, + "loss": 1.0062, + "step": 51924 + }, + { + "epoch": 3.84, + "learning_rate": 6.076551787056966e-06, + "loss": 0.9962, + "step": 51925 + }, + { + "epoch": 3.84, + "learning_rate": 6.076184900798208e-06, + "loss": 1.0028, + "step": 51926 + }, + { + "epoch": 3.84, + "learning_rate": 6.075818020782192e-06, + "loss": 1.0018, + "step": 51927 + }, + { + "epoch": 3.84, + "learning_rate": 6.0754511470094855e-06, + "loss": 1.0034, + "step": 51928 + }, + { + "epoch": 3.84, + "learning_rate": 6.075084279480688e-06, + "loss": 1.1015, + "step": 51929 + }, + { + "epoch": 3.84, + "learning_rate": 6.074717418196367e-06, + "loss": 1.1041, + "step": 51930 + }, + { + "epoch": 3.84, + "learning_rate": 6.0743505631571205e-06, + "loss": 0.9478, + "step": 51931 + }, + { + "epoch": 3.84, + "learning_rate": 6.073983714363525e-06, + "loss": 0.937, + "step": 51932 + }, + { + "epoch": 3.84, + "learning_rate": 6.073616871816163e-06, + "loss": 0.9508, + "step": 51933 + }, + { + "epoch": 3.84, + "learning_rate": 6.073250035515623e-06, + "loss": 0.8953, + "step": 51934 + }, + { + "epoch": 3.84, + "learning_rate": 6.072883205462484e-06, + "loss": 1.0635, + "step": 51935 + }, + { + "epoch": 3.84, + "learning_rate": 6.072516381657334e-06, + "loss": 0.9565, + "step": 51936 + }, + { + "epoch": 3.84, + "learning_rate": 6.072149564100751e-06, + "loss": 1.044, + "step": 51937 + }, + { + "epoch": 3.84, + "learning_rate": 6.0717827527933244e-06, + "loss": 1.0025, + "step": 51938 + }, + { + "epoch": 3.84, + "learning_rate": 6.071415947735632e-06, + "loss": 0.9947, + "step": 51939 + }, + { + "epoch": 3.84, + "learning_rate": 6.071049148928265e-06, + "loss": 0.9175, + "step": 51940 + }, + { + "epoch": 3.84, + "learning_rate": 6.070682356371802e-06, + "loss": 1.0115, + "step": 51941 + }, + { + "epoch": 3.84, + "learning_rate": 6.070315570066825e-06, + "loss": 0.8082, + "step": 51942 + }, + { + "epoch": 3.84, + "learning_rate": 6.0699487900139196e-06, + "loss": 0.9659, + "step": 51943 + }, + { + "epoch": 3.84, + "learning_rate": 6.0695820162136694e-06, + "loss": 0.9203, + "step": 51944 + }, + { + "epoch": 3.84, + "learning_rate": 6.06921524866666e-06, + "loss": 1.0134, + "step": 51945 + }, + { + "epoch": 3.84, + "learning_rate": 6.068848487373471e-06, + "loss": 0.9146, + "step": 51946 + }, + { + "epoch": 3.84, + "learning_rate": 6.0684817323346905e-06, + "loss": 1.0187, + "step": 51947 + }, + { + "epoch": 3.84, + "learning_rate": 6.068114983550894e-06, + "loss": 0.9228, + "step": 51948 + }, + { + "epoch": 3.84, + "learning_rate": 6.067748241022675e-06, + "loss": 1.1598, + "step": 51949 + }, + { + "epoch": 3.84, + "learning_rate": 6.067381504750615e-06, + "loss": 0.9231, + "step": 51950 + }, + { + "epoch": 3.84, + "learning_rate": 6.067014774735289e-06, + "loss": 0.9749, + "step": 51951 + }, + { + "epoch": 3.84, + "learning_rate": 6.066648050977288e-06, + "loss": 0.9515, + "step": 51952 + }, + { + "epoch": 3.84, + "learning_rate": 6.066281333477193e-06, + "loss": 0.9841, + "step": 51953 + }, + { + "epoch": 3.84, + "learning_rate": 6.065914622235589e-06, + "loss": 1.0335, + "step": 51954 + }, + { + "epoch": 3.84, + "learning_rate": 6.0655479172530586e-06, + "loss": 1.0798, + "step": 51955 + }, + { + "epoch": 3.84, + "learning_rate": 6.065181218530185e-06, + "loss": 1.0166, + "step": 51956 + }, + { + "epoch": 3.84, + "learning_rate": 6.06481452606755e-06, + "loss": 1.0045, + "step": 51957 + }, + { + "epoch": 3.84, + "learning_rate": 6.064447839865742e-06, + "loss": 0.9502, + "step": 51958 + }, + { + "epoch": 3.84, + "learning_rate": 6.0640811599253395e-06, + "loss": 1.076, + "step": 51959 + }, + { + "epoch": 3.84, + "learning_rate": 6.063714486246927e-06, + "loss": 0.9872, + "step": 51960 + }, + { + "epoch": 3.84, + "learning_rate": 6.06334781883109e-06, + "loss": 1.0372, + "step": 51961 + }, + { + "epoch": 3.84, + "learning_rate": 6.062981157678409e-06, + "loss": 1.0458, + "step": 51962 + }, + { + "epoch": 3.84, + "learning_rate": 6.06261450278947e-06, + "loss": 0.9776, + "step": 51963 + }, + { + "epoch": 3.84, + "learning_rate": 6.062247854164852e-06, + "loss": 0.989, + "step": 51964 + }, + { + "epoch": 3.84, + "learning_rate": 6.061881211805145e-06, + "loss": 1.0495, + "step": 51965 + }, + { + "epoch": 3.84, + "learning_rate": 6.061514575710925e-06, + "loss": 0.8991, + "step": 51966 + }, + { + "epoch": 3.84, + "learning_rate": 6.061147945882781e-06, + "loss": 1.0975, + "step": 51967 + }, + { + "epoch": 3.84, + "learning_rate": 6.060781322321294e-06, + "loss": 1.0164, + "step": 51968 + }, + { + "epoch": 3.84, + "learning_rate": 6.060414705027049e-06, + "loss": 1.0032, + "step": 51969 + }, + { + "epoch": 3.84, + "learning_rate": 6.06004809400063e-06, + "loss": 0.9687, + "step": 51970 + }, + { + "epoch": 3.84, + "learning_rate": 6.05968148924261e-06, + "loss": 0.9673, + "step": 51971 + }, + { + "epoch": 3.84, + "learning_rate": 6.059314890753589e-06, + "loss": 0.973, + "step": 51972 + }, + { + "epoch": 3.84, + "learning_rate": 6.058948298534135e-06, + "loss": 1.1255, + "step": 51973 + }, + { + "epoch": 3.84, + "learning_rate": 6.058581712584844e-06, + "loss": 1.0018, + "step": 51974 + }, + { + "epoch": 3.84, + "learning_rate": 6.0582151329062856e-06, + "loss": 0.989, + "step": 51975 + }, + { + "epoch": 3.84, + "learning_rate": 6.057848559499058e-06, + "loss": 0.959, + "step": 51976 + }, + { + "epoch": 3.84, + "learning_rate": 6.057481992363729e-06, + "loss": 0.9585, + "step": 51977 + }, + { + "epoch": 3.84, + "learning_rate": 6.057115431500897e-06, + "loss": 0.9906, + "step": 51978 + }, + { + "epoch": 3.84, + "learning_rate": 6.056748876911137e-06, + "loss": 0.9848, + "step": 51979 + }, + { + "epoch": 3.84, + "learning_rate": 6.056382328595029e-06, + "loss": 1.0052, + "step": 51980 + }, + { + "epoch": 3.84, + "learning_rate": 6.056015786553165e-06, + "loss": 1.1382, + "step": 51981 + }, + { + "epoch": 3.84, + "learning_rate": 6.05564925078612e-06, + "loss": 1.0033, + "step": 51982 + }, + { + "epoch": 3.84, + "learning_rate": 6.055282721294482e-06, + "loss": 1.0901, + "step": 51983 + }, + { + "epoch": 3.84, + "learning_rate": 6.054916198078832e-06, + "loss": 0.9936, + "step": 51984 + }, + { + "epoch": 3.84, + "learning_rate": 6.054549681139756e-06, + "loss": 1.0263, + "step": 51985 + }, + { + "epoch": 3.84, + "learning_rate": 6.054183170477832e-06, + "loss": 1.0835, + "step": 51986 + }, + { + "epoch": 3.84, + "learning_rate": 6.0538166660936484e-06, + "loss": 0.9317, + "step": 51987 + }, + { + "epoch": 3.84, + "learning_rate": 6.053450167987784e-06, + "loss": 1.0056, + "step": 51988 + }, + { + "epoch": 3.84, + "learning_rate": 6.053083676160825e-06, + "loss": 1.0653, + "step": 51989 + }, + { + "epoch": 3.84, + "learning_rate": 6.052717190613354e-06, + "loss": 0.9673, + "step": 51990 + }, + { + "epoch": 3.84, + "learning_rate": 6.0523507113459514e-06, + "loss": 0.9733, + "step": 51991 + }, + { + "epoch": 3.84, + "learning_rate": 6.051984238359205e-06, + "loss": 0.9782, + "step": 51992 + }, + { + "epoch": 3.84, + "learning_rate": 6.051617771653693e-06, + "loss": 1.0017, + "step": 51993 + }, + { + "epoch": 3.84, + "learning_rate": 6.051251311230002e-06, + "loss": 0.9786, + "step": 51994 + }, + { + "epoch": 3.84, + "learning_rate": 6.050884857088713e-06, + "loss": 0.8714, + "step": 51995 + }, + { + "epoch": 3.84, + "learning_rate": 6.0505184092304125e-06, + "loss": 0.9299, + "step": 51996 + }, + { + "epoch": 3.84, + "learning_rate": 6.05015196765568e-06, + "loss": 1.0233, + "step": 51997 + }, + { + "epoch": 3.84, + "learning_rate": 6.0497855323650955e-06, + "loss": 1.0099, + "step": 51998 + }, + { + "epoch": 3.84, + "learning_rate": 6.049419103359248e-06, + "loss": 0.9587, + "step": 51999 + }, + { + "epoch": 3.84, + "learning_rate": 6.049052680638718e-06, + "loss": 1.1239, + "step": 52000 + }, + { + "epoch": 3.84, + "learning_rate": 6.048686264204089e-06, + "loss": 1.0906, + "step": 52001 + }, + { + "epoch": 3.84, + "learning_rate": 6.048319854055942e-06, + "loss": 1.0121, + "step": 52002 + }, + { + "epoch": 3.84, + "learning_rate": 6.047953450194863e-06, + "loss": 1.054, + "step": 52003 + }, + { + "epoch": 3.84, + "learning_rate": 6.047587052621432e-06, + "loss": 0.9922, + "step": 52004 + }, + { + "epoch": 3.84, + "learning_rate": 6.0472206613362346e-06, + "loss": 1.1008, + "step": 52005 + }, + { + "epoch": 3.84, + "learning_rate": 6.046854276339855e-06, + "loss": 1.1369, + "step": 52006 + }, + { + "epoch": 3.84, + "learning_rate": 6.046487897632869e-06, + "loss": 0.9764, + "step": 52007 + }, + { + "epoch": 3.84, + "learning_rate": 6.046121525215868e-06, + "loss": 0.977, + "step": 52008 + }, + { + "epoch": 3.84, + "learning_rate": 6.045755159089428e-06, + "loss": 0.9701, + "step": 52009 + }, + { + "epoch": 3.84, + "learning_rate": 6.045388799254137e-06, + "loss": 0.9744, + "step": 52010 + }, + { + "epoch": 3.84, + "learning_rate": 6.045022445710576e-06, + "loss": 1.0628, + "step": 52011 + }, + { + "epoch": 3.84, + "learning_rate": 6.044656098459327e-06, + "loss": 0.976, + "step": 52012 + }, + { + "epoch": 3.84, + "learning_rate": 6.044289757500972e-06, + "loss": 1.0737, + "step": 52013 + }, + { + "epoch": 3.84, + "learning_rate": 6.043923422836098e-06, + "loss": 0.9529, + "step": 52014 + }, + { + "epoch": 3.84, + "learning_rate": 6.043557094465283e-06, + "loss": 0.978, + "step": 52015 + }, + { + "epoch": 3.84, + "learning_rate": 6.043190772389113e-06, + "loss": 1.113, + "step": 52016 + }, + { + "epoch": 3.84, + "learning_rate": 6.042824456608174e-06, + "loss": 0.9981, + "step": 52017 + }, + { + "epoch": 3.84, + "learning_rate": 6.042458147123038e-06, + "loss": 1.1142, + "step": 52018 + }, + { + "epoch": 3.84, + "learning_rate": 6.0420918439343005e-06, + "loss": 1.0627, + "step": 52019 + }, + { + "epoch": 3.84, + "learning_rate": 6.041725547042532e-06, + "loss": 1.0119, + "step": 52020 + }, + { + "epoch": 3.84, + "learning_rate": 6.041359256448327e-06, + "loss": 1.0555, + "step": 52021 + }, + { + "epoch": 3.84, + "learning_rate": 6.040992972152257e-06, + "loss": 1.1556, + "step": 52022 + }, + { + "epoch": 3.84, + "learning_rate": 6.040626694154916e-06, + "loss": 0.8264, + "step": 52023 + }, + { + "epoch": 3.84, + "learning_rate": 6.040260422456877e-06, + "loss": 1.1319, + "step": 52024 + }, + { + "epoch": 3.84, + "learning_rate": 6.039894157058732e-06, + "loss": 1.0879, + "step": 52025 + }, + { + "epoch": 3.84, + "learning_rate": 6.039527897961057e-06, + "loss": 1.0335, + "step": 52026 + }, + { + "epoch": 3.84, + "learning_rate": 6.039161645164434e-06, + "loss": 0.9337, + "step": 52027 + }, + { + "epoch": 3.84, + "learning_rate": 6.03879539866945e-06, + "loss": 0.9636, + "step": 52028 + }, + { + "epoch": 3.84, + "learning_rate": 6.0384291584766845e-06, + "loss": 0.9598, + "step": 52029 + }, + { + "epoch": 3.84, + "learning_rate": 6.038062924586723e-06, + "loss": 0.9778, + "step": 52030 + }, + { + "epoch": 3.84, + "learning_rate": 6.037696697000146e-06, + "loss": 1.1522, + "step": 52031 + }, + { + "epoch": 3.84, + "learning_rate": 6.037330475717538e-06, + "loss": 0.9255, + "step": 52032 + }, + { + "epoch": 3.84, + "learning_rate": 6.036964260739477e-06, + "loss": 1.0898, + "step": 52033 + }, + { + "epoch": 3.84, + "learning_rate": 6.036598052066553e-06, + "loss": 1.0906, + "step": 52034 + }, + { + "epoch": 3.84, + "learning_rate": 6.036231849699344e-06, + "loss": 1.0188, + "step": 52035 + }, + { + "epoch": 3.84, + "learning_rate": 6.0358656536384315e-06, + "loss": 0.9511, + "step": 52036 + }, + { + "epoch": 3.84, + "learning_rate": 6.035499463884402e-06, + "loss": 1.0021, + "step": 52037 + }, + { + "epoch": 3.85, + "learning_rate": 6.035133280437835e-06, + "loss": 0.9491, + "step": 52038 + }, + { + "epoch": 3.85, + "learning_rate": 6.034767103299317e-06, + "loss": 0.8695, + "step": 52039 + }, + { + "epoch": 3.85, + "learning_rate": 6.034400932469423e-06, + "loss": 0.9526, + "step": 52040 + }, + { + "epoch": 3.85, + "learning_rate": 6.034034767948747e-06, + "loss": 1.0054, + "step": 52041 + }, + { + "epoch": 3.85, + "learning_rate": 6.0336686097378575e-06, + "loss": 1.0486, + "step": 52042 + }, + { + "epoch": 3.85, + "learning_rate": 6.03330245783735e-06, + "loss": 1.0081, + "step": 52043 + }, + { + "epoch": 3.85, + "learning_rate": 6.0329363122478e-06, + "loss": 1.0207, + "step": 52044 + }, + { + "epoch": 3.85, + "learning_rate": 6.0325701729697895e-06, + "loss": 1.036, + "step": 52045 + }, + { + "epoch": 3.85, + "learning_rate": 6.032204040003906e-06, + "loss": 1.0666, + "step": 52046 + }, + { + "epoch": 3.85, + "learning_rate": 6.031837913350727e-06, + "loss": 1.0678, + "step": 52047 + }, + { + "epoch": 3.85, + "learning_rate": 6.031471793010838e-06, + "loss": 1.0696, + "step": 52048 + }, + { + "epoch": 3.85, + "learning_rate": 6.031105678984819e-06, + "loss": 0.9574, + "step": 52049 + }, + { + "epoch": 3.85, + "learning_rate": 6.030739571273256e-06, + "loss": 1.0818, + "step": 52050 + }, + { + "epoch": 3.85, + "learning_rate": 6.030373469876728e-06, + "loss": 0.9367, + "step": 52051 + }, + { + "epoch": 3.85, + "learning_rate": 6.030007374795821e-06, + "loss": 0.9653, + "step": 52052 + }, + { + "epoch": 3.85, + "learning_rate": 6.029641286031116e-06, + "loss": 1.0338, + "step": 52053 + }, + { + "epoch": 3.85, + "learning_rate": 6.0292752035831915e-06, + "loss": 0.9067, + "step": 52054 + }, + { + "epoch": 3.85, + "learning_rate": 6.0289091274526355e-06, + "loss": 0.9736, + "step": 52055 + }, + { + "epoch": 3.85, + "learning_rate": 6.028543057640027e-06, + "loss": 1.0497, + "step": 52056 + }, + { + "epoch": 3.85, + "learning_rate": 6.028176994145951e-06, + "loss": 0.9919, + "step": 52057 + }, + { + "epoch": 3.85, + "learning_rate": 6.027810936970988e-06, + "loss": 0.9221, + "step": 52058 + }, + { + "epoch": 3.85, + "learning_rate": 6.027444886115722e-06, + "loss": 0.9377, + "step": 52059 + }, + { + "epoch": 3.85, + "learning_rate": 6.027078841580733e-06, + "loss": 0.9185, + "step": 52060 + }, + { + "epoch": 3.85, + "learning_rate": 6.026712803366605e-06, + "loss": 0.9893, + "step": 52061 + }, + { + "epoch": 3.85, + "learning_rate": 6.026346771473923e-06, + "loss": 1.0972, + "step": 52062 + }, + { + "epoch": 3.85, + "learning_rate": 6.025980745903262e-06, + "loss": 0.9866, + "step": 52063 + }, + { + "epoch": 3.85, + "learning_rate": 6.025614726655214e-06, + "loss": 0.9708, + "step": 52064 + }, + { + "epoch": 3.85, + "learning_rate": 6.0252487137303486e-06, + "loss": 0.9969, + "step": 52065 + }, + { + "epoch": 3.85, + "learning_rate": 6.024882707129263e-06, + "loss": 0.9873, + "step": 52066 + }, + { + "epoch": 3.85, + "learning_rate": 6.024516706852524e-06, + "loss": 1.0594, + "step": 52067 + }, + { + "epoch": 3.85, + "learning_rate": 6.02415071290073e-06, + "loss": 0.9834, + "step": 52068 + }, + { + "epoch": 3.85, + "learning_rate": 6.02378472527445e-06, + "loss": 1.0383, + "step": 52069 + }, + { + "epoch": 3.85, + "learning_rate": 6.0234187439742765e-06, + "loss": 0.9571, + "step": 52070 + }, + { + "epoch": 3.85, + "learning_rate": 6.023052769000781e-06, + "loss": 0.8455, + "step": 52071 + }, + { + "epoch": 3.85, + "learning_rate": 6.022686800354558e-06, + "loss": 1.0384, + "step": 52072 + }, + { + "epoch": 3.85, + "learning_rate": 6.02232083803618e-06, + "loss": 1.0379, + "step": 52073 + }, + { + "epoch": 3.85, + "learning_rate": 6.0219548820462325e-06, + "loss": 1.0255, + "step": 52074 + }, + { + "epoch": 3.85, + "learning_rate": 6.0215889323852985e-06, + "loss": 1.0472, + "step": 52075 + }, + { + "epoch": 3.85, + "learning_rate": 6.021222989053959e-06, + "loss": 1.1464, + "step": 52076 + }, + { + "epoch": 3.85, + "learning_rate": 6.020857052052798e-06, + "loss": 1.0068, + "step": 52077 + }, + { + "epoch": 3.85, + "learning_rate": 6.0204911213823935e-06, + "loss": 0.9417, + "step": 52078 + }, + { + "epoch": 3.85, + "learning_rate": 6.020125197043334e-06, + "loss": 0.9411, + "step": 52079 + }, + { + "epoch": 3.85, + "learning_rate": 6.019759279036196e-06, + "loss": 0.9694, + "step": 52080 + }, + { + "epoch": 3.85, + "learning_rate": 6.019393367361566e-06, + "loss": 1.0668, + "step": 52081 + }, + { + "epoch": 3.85, + "learning_rate": 6.019027462020024e-06, + "loss": 0.9512, + "step": 52082 + }, + { + "epoch": 3.85, + "learning_rate": 6.018661563012152e-06, + "loss": 0.9972, + "step": 52083 + }, + { + "epoch": 3.85, + "learning_rate": 6.018295670338534e-06, + "loss": 1.0752, + "step": 52084 + }, + { + "epoch": 3.85, + "learning_rate": 6.017929783999745e-06, + "loss": 0.9646, + "step": 52085 + }, + { + "epoch": 3.85, + "learning_rate": 6.01756390399638e-06, + "loss": 1.0775, + "step": 52086 + }, + { + "epoch": 3.85, + "learning_rate": 6.0171980303290075e-06, + "loss": 1.0314, + "step": 52087 + }, + { + "epoch": 3.85, + "learning_rate": 6.016832162998221e-06, + "loss": 0.9773, + "step": 52088 + }, + { + "epoch": 3.85, + "learning_rate": 6.016466302004593e-06, + "loss": 1.0239, + "step": 52089 + }, + { + "epoch": 3.85, + "learning_rate": 6.016100447348715e-06, + "loss": 0.9034, + "step": 52090 + }, + { + "epoch": 3.85, + "learning_rate": 6.015734599031162e-06, + "loss": 0.9419, + "step": 52091 + }, + { + "epoch": 3.85, + "learning_rate": 6.015368757052518e-06, + "loss": 0.923, + "step": 52092 + }, + { + "epoch": 3.85, + "learning_rate": 6.015002921413366e-06, + "loss": 0.9936, + "step": 52093 + }, + { + "epoch": 3.85, + "learning_rate": 6.014637092114285e-06, + "loss": 0.971, + "step": 52094 + }, + { + "epoch": 3.85, + "learning_rate": 6.0142712691558626e-06, + "loss": 0.9382, + "step": 52095 + }, + { + "epoch": 3.85, + "learning_rate": 6.013905452538674e-06, + "loss": 1.0126, + "step": 52096 + }, + { + "epoch": 3.85, + "learning_rate": 6.013539642263308e-06, + "loss": 1.0277, + "step": 52097 + }, + { + "epoch": 3.85, + "learning_rate": 6.013173838330341e-06, + "loss": 1.0171, + "step": 52098 + }, + { + "epoch": 3.85, + "learning_rate": 6.012808040740359e-06, + "loss": 1.1116, + "step": 52099 + }, + { + "epoch": 3.85, + "learning_rate": 6.012442249493943e-06, + "loss": 1.0303, + "step": 52100 + }, + { + "epoch": 3.85, + "learning_rate": 6.012076464591672e-06, + "loss": 1.0005, + "step": 52101 + }, + { + "epoch": 3.85, + "learning_rate": 6.0117106860341324e-06, + "loss": 1.0437, + "step": 52102 + }, + { + "epoch": 3.85, + "learning_rate": 6.011344913821902e-06, + "loss": 0.9872, + "step": 52103 + }, + { + "epoch": 3.85, + "learning_rate": 6.010979147955567e-06, + "loss": 1.0043, + "step": 52104 + }, + { + "epoch": 3.85, + "learning_rate": 6.010613388435704e-06, + "loss": 0.9757, + "step": 52105 + }, + { + "epoch": 3.85, + "learning_rate": 6.010247635262901e-06, + "loss": 1.0799, + "step": 52106 + }, + { + "epoch": 3.85, + "learning_rate": 6.009881888437735e-06, + "loss": 1.0855, + "step": 52107 + }, + { + "epoch": 3.85, + "learning_rate": 6.009516147960792e-06, + "loss": 0.9299, + "step": 52108 + }, + { + "epoch": 3.85, + "learning_rate": 6.009150413832654e-06, + "loss": 0.9772, + "step": 52109 + }, + { + "epoch": 3.85, + "learning_rate": 6.008784686053893e-06, + "loss": 0.9433, + "step": 52110 + }, + { + "epoch": 3.85, + "learning_rate": 6.008418964625106e-06, + "loss": 1.0436, + "step": 52111 + }, + { + "epoch": 3.85, + "learning_rate": 6.00805324954686e-06, + "loss": 0.9585, + "step": 52112 + }, + { + "epoch": 3.85, + "learning_rate": 6.007687540819751e-06, + "loss": 0.9855, + "step": 52113 + }, + { + "epoch": 3.85, + "learning_rate": 6.007321838444349e-06, + "loss": 1.0325, + "step": 52114 + }, + { + "epoch": 3.85, + "learning_rate": 6.006956142421245e-06, + "loss": 0.9552, + "step": 52115 + }, + { + "epoch": 3.85, + "learning_rate": 6.006590452751011e-06, + "loss": 1.204, + "step": 52116 + }, + { + "epoch": 3.85, + "learning_rate": 6.006224769434241e-06, + "loss": 0.9652, + "step": 52117 + }, + { + "epoch": 3.85, + "learning_rate": 6.005859092471509e-06, + "loss": 1.0669, + "step": 52118 + }, + { + "epoch": 3.85, + "learning_rate": 6.005493421863394e-06, + "loss": 0.9174, + "step": 52119 + }, + { + "epoch": 3.85, + "learning_rate": 6.005127757610485e-06, + "loss": 1.0322, + "step": 52120 + }, + { + "epoch": 3.85, + "learning_rate": 6.00476209971336e-06, + "loss": 0.9992, + "step": 52121 + }, + { + "epoch": 3.85, + "learning_rate": 6.0043964481726e-06, + "loss": 1.0222, + "step": 52122 + }, + { + "epoch": 3.85, + "learning_rate": 6.004030802988788e-06, + "loss": 1.0091, + "step": 52123 + }, + { + "epoch": 3.85, + "learning_rate": 6.003665164162508e-06, + "loss": 1.0561, + "step": 52124 + }, + { + "epoch": 3.85, + "learning_rate": 6.003299531694337e-06, + "loss": 1.1056, + "step": 52125 + }, + { + "epoch": 3.85, + "learning_rate": 6.002933905584862e-06, + "loss": 1.0918, + "step": 52126 + }, + { + "epoch": 3.85, + "learning_rate": 6.00256828583466e-06, + "loss": 1.0158, + "step": 52127 + }, + { + "epoch": 3.85, + "learning_rate": 6.002202672444315e-06, + "loss": 1.0841, + "step": 52128 + }, + { + "epoch": 3.85, + "learning_rate": 6.001837065414413e-06, + "loss": 1.0379, + "step": 52129 + }, + { + "epoch": 3.85, + "learning_rate": 6.001471464745523e-06, + "loss": 1.1494, + "step": 52130 + }, + { + "epoch": 3.85, + "learning_rate": 6.001105870438242e-06, + "loss": 0.971, + "step": 52131 + }, + { + "epoch": 3.85, + "learning_rate": 6.000740282493136e-06, + "loss": 0.9776, + "step": 52132 + }, + { + "epoch": 3.85, + "learning_rate": 6.000374700910803e-06, + "loss": 0.9402, + "step": 52133 + }, + { + "epoch": 3.85, + "learning_rate": 6.00000912569181e-06, + "loss": 1.1277, + "step": 52134 + }, + { + "epoch": 3.85, + "learning_rate": 5.999643556836752e-06, + "loss": 0.9305, + "step": 52135 + }, + { + "epoch": 3.85, + "learning_rate": 5.9992779943461975e-06, + "loss": 0.9196, + "step": 52136 + }, + { + "epoch": 3.85, + "learning_rate": 5.998912438220741e-06, + "loss": 0.994, + "step": 52137 + }, + { + "epoch": 3.85, + "learning_rate": 5.9985468884609546e-06, + "loss": 1.1561, + "step": 52138 + }, + { + "epoch": 3.85, + "learning_rate": 5.99818134506742e-06, + "loss": 0.9859, + "step": 52139 + }, + { + "epoch": 3.85, + "learning_rate": 5.9978158080407255e-06, + "loss": 1.0025, + "step": 52140 + }, + { + "epoch": 3.85, + "learning_rate": 5.997450277381446e-06, + "loss": 0.9572, + "step": 52141 + }, + { + "epoch": 3.85, + "learning_rate": 5.997084753090167e-06, + "loss": 1.0125, + "step": 52142 + }, + { + "epoch": 3.85, + "learning_rate": 5.996719235167468e-06, + "loss": 1.0848, + "step": 52143 + }, + { + "epoch": 3.85, + "learning_rate": 5.9963537236139335e-06, + "loss": 0.9957, + "step": 52144 + }, + { + "epoch": 3.85, + "learning_rate": 5.9959882184301395e-06, + "loss": 0.9591, + "step": 52145 + }, + { + "epoch": 3.85, + "learning_rate": 5.995622719616675e-06, + "loss": 1.0509, + "step": 52146 + }, + { + "epoch": 3.85, + "learning_rate": 5.995257227174116e-06, + "loss": 1.1035, + "step": 52147 + }, + { + "epoch": 3.85, + "learning_rate": 5.994891741103044e-06, + "loss": 0.9328, + "step": 52148 + }, + { + "epoch": 3.85, + "learning_rate": 5.994526261404044e-06, + "loss": 0.9809, + "step": 52149 + }, + { + "epoch": 3.85, + "learning_rate": 5.994160788077694e-06, + "loss": 1.0538, + "step": 52150 + }, + { + "epoch": 3.85, + "learning_rate": 5.993795321124578e-06, + "loss": 0.9674, + "step": 52151 + }, + { + "epoch": 3.85, + "learning_rate": 5.9934298605452745e-06, + "loss": 1.0442, + "step": 52152 + }, + { + "epoch": 3.85, + "learning_rate": 5.9930644063403684e-06, + "loss": 1.0645, + "step": 52153 + }, + { + "epoch": 3.85, + "learning_rate": 5.992698958510439e-06, + "loss": 0.9178, + "step": 52154 + }, + { + "epoch": 3.85, + "learning_rate": 5.992333517056068e-06, + "loss": 0.9679, + "step": 52155 + }, + { + "epoch": 3.85, + "learning_rate": 5.991968081977841e-06, + "loss": 0.9296, + "step": 52156 + }, + { + "epoch": 3.85, + "learning_rate": 5.991602653276328e-06, + "loss": 1.0022, + "step": 52157 + }, + { + "epoch": 3.85, + "learning_rate": 5.991237230952125e-06, + "loss": 1.0283, + "step": 52158 + }, + { + "epoch": 3.85, + "learning_rate": 5.990871815005798e-06, + "loss": 0.8769, + "step": 52159 + }, + { + "epoch": 3.85, + "learning_rate": 5.990506405437946e-06, + "loss": 0.91, + "step": 52160 + }, + { + "epoch": 3.85, + "learning_rate": 5.990141002249134e-06, + "loss": 0.9771, + "step": 52161 + }, + { + "epoch": 3.85, + "learning_rate": 5.989775605439953e-06, + "loss": 1.0255, + "step": 52162 + }, + { + "epoch": 3.85, + "learning_rate": 5.989410215010981e-06, + "loss": 1.0246, + "step": 52163 + }, + { + "epoch": 3.85, + "learning_rate": 5.9890448309628e-06, + "loss": 1.0612, + "step": 52164 + }, + { + "epoch": 3.85, + "learning_rate": 5.988679453295994e-06, + "loss": 0.9876, + "step": 52165 + }, + { + "epoch": 3.85, + "learning_rate": 5.988314082011138e-06, + "loss": 0.9889, + "step": 52166 + }, + { + "epoch": 3.85, + "learning_rate": 5.987948717108818e-06, + "loss": 1.1406, + "step": 52167 + }, + { + "epoch": 3.85, + "learning_rate": 5.987583358589613e-06, + "loss": 1.0476, + "step": 52168 + }, + { + "epoch": 3.85, + "learning_rate": 5.987218006454107e-06, + "loss": 1.1248, + "step": 52169 + }, + { + "epoch": 3.85, + "learning_rate": 5.986852660702879e-06, + "loss": 0.9509, + "step": 52170 + }, + { + "epoch": 3.85, + "learning_rate": 5.986487321336511e-06, + "loss": 0.968, + "step": 52171 + }, + { + "epoch": 3.85, + "learning_rate": 5.986121988355584e-06, + "loss": 1.1606, + "step": 52172 + }, + { + "epoch": 3.86, + "learning_rate": 5.985756661760682e-06, + "loss": 1.0327, + "step": 52173 + }, + { + "epoch": 3.86, + "learning_rate": 5.98539134155238e-06, + "loss": 1.0228, + "step": 52174 + }, + { + "epoch": 3.86, + "learning_rate": 5.9850260277312655e-06, + "loss": 1.0074, + "step": 52175 + }, + { + "epoch": 3.86, + "learning_rate": 5.984660720297919e-06, + "loss": 0.9654, + "step": 52176 + }, + { + "epoch": 3.86, + "learning_rate": 5.984295419252915e-06, + "loss": 0.9443, + "step": 52177 + }, + { + "epoch": 3.86, + "learning_rate": 5.983930124596844e-06, + "loss": 0.9532, + "step": 52178 + }, + { + "epoch": 3.86, + "learning_rate": 5.983564836330278e-06, + "loss": 1.0718, + "step": 52179 + }, + { + "epoch": 3.86, + "learning_rate": 5.983199554453807e-06, + "loss": 1.0057, + "step": 52180 + }, + { + "epoch": 3.86, + "learning_rate": 5.982834278968004e-06, + "loss": 0.8792, + "step": 52181 + }, + { + "epoch": 3.86, + "learning_rate": 5.982469009873459e-06, + "loss": 1.0809, + "step": 52182 + }, + { + "epoch": 3.86, + "learning_rate": 5.982103747170743e-06, + "loss": 0.9809, + "step": 52183 + }, + { + "epoch": 3.86, + "learning_rate": 5.981738490860448e-06, + "loss": 0.9205, + "step": 52184 + }, + { + "epoch": 3.86, + "learning_rate": 5.981373240943148e-06, + "loss": 1.0795, + "step": 52185 + }, + { + "epoch": 3.86, + "learning_rate": 5.981007997419424e-06, + "loss": 1.0037, + "step": 52186 + }, + { + "epoch": 3.86, + "learning_rate": 5.98064276028986e-06, + "loss": 1.06, + "step": 52187 + }, + { + "epoch": 3.86, + "learning_rate": 5.980277529555034e-06, + "loss": 1.057, + "step": 52188 + }, + { + "epoch": 3.86, + "learning_rate": 5.97991230521553e-06, + "loss": 1.0953, + "step": 52189 + }, + { + "epoch": 3.86, + "learning_rate": 5.979547087271928e-06, + "loss": 1.0935, + "step": 52190 + }, + { + "epoch": 3.86, + "learning_rate": 5.97918187572481e-06, + "loss": 1.0026, + "step": 52191 + }, + { + "epoch": 3.86, + "learning_rate": 5.978816670574754e-06, + "loss": 1.0562, + "step": 52192 + }, + { + "epoch": 3.86, + "learning_rate": 5.978451471822346e-06, + "loss": 1.0701, + "step": 52193 + }, + { + "epoch": 3.86, + "learning_rate": 5.978086279468163e-06, + "loss": 1.0231, + "step": 52194 + }, + { + "epoch": 3.86, + "learning_rate": 5.977721093512785e-06, + "loss": 1.0348, + "step": 52195 + }, + { + "epoch": 3.86, + "learning_rate": 5.977355913956799e-06, + "loss": 1.039, + "step": 52196 + }, + { + "epoch": 3.86, + "learning_rate": 5.976990740800779e-06, + "loss": 1.0399, + "step": 52197 + }, + { + "epoch": 3.86, + "learning_rate": 5.9766255740453115e-06, + "loss": 0.9718, + "step": 52198 + }, + { + "epoch": 3.86, + "learning_rate": 5.976260413690973e-06, + "loss": 1.0471, + "step": 52199 + }, + { + "epoch": 3.86, + "learning_rate": 5.975895259738348e-06, + "loss": 1.1732, + "step": 52200 + }, + { + "epoch": 3.86, + "learning_rate": 5.975530112188015e-06, + "loss": 0.9834, + "step": 52201 + }, + { + "epoch": 3.86, + "learning_rate": 5.975164971040557e-06, + "loss": 1.0429, + "step": 52202 + }, + { + "epoch": 3.86, + "learning_rate": 5.974799836296556e-06, + "loss": 1.0453, + "step": 52203 + }, + { + "epoch": 3.86, + "learning_rate": 5.974434707956585e-06, + "loss": 1.0256, + "step": 52204 + }, + { + "epoch": 3.86, + "learning_rate": 5.974069586021238e-06, + "loss": 1.0291, + "step": 52205 + }, + { + "epoch": 3.86, + "learning_rate": 5.973704470491084e-06, + "loss": 0.9655, + "step": 52206 + }, + { + "epoch": 3.86, + "learning_rate": 5.97333936136671e-06, + "loss": 0.965, + "step": 52207 + }, + { + "epoch": 3.86, + "learning_rate": 5.972974258648694e-06, + "loss": 1.0416, + "step": 52208 + }, + { + "epoch": 3.86, + "learning_rate": 5.972609162337621e-06, + "loss": 0.9738, + "step": 52209 + }, + { + "epoch": 3.86, + "learning_rate": 5.972244072434065e-06, + "loss": 1.0344, + "step": 52210 + }, + { + "epoch": 3.86, + "learning_rate": 5.971878988938615e-06, + "loss": 0.8732, + "step": 52211 + }, + { + "epoch": 3.86, + "learning_rate": 5.971513911851847e-06, + "loss": 1.0086, + "step": 52212 + }, + { + "epoch": 3.86, + "learning_rate": 5.971148841174341e-06, + "loss": 1.0505, + "step": 52213 + }, + { + "epoch": 3.86, + "learning_rate": 5.970783776906684e-06, + "loss": 0.943, + "step": 52214 + }, + { + "epoch": 3.86, + "learning_rate": 5.970418719049447e-06, + "loss": 0.9816, + "step": 52215 + }, + { + "epoch": 3.86, + "learning_rate": 5.970053667603219e-06, + "loss": 1.072, + "step": 52216 + }, + { + "epoch": 3.86, + "learning_rate": 5.969688622568577e-06, + "loss": 1.0895, + "step": 52217 + }, + { + "epoch": 3.86, + "learning_rate": 5.969323583946103e-06, + "loss": 0.9647, + "step": 52218 + }, + { + "epoch": 3.86, + "learning_rate": 5.968958551736377e-06, + "loss": 0.825, + "step": 52219 + }, + { + "epoch": 3.86, + "learning_rate": 5.9685935259399804e-06, + "loss": 0.9856, + "step": 52220 + }, + { + "epoch": 3.86, + "learning_rate": 5.968228506557498e-06, + "loss": 1.0552, + "step": 52221 + }, + { + "epoch": 3.86, + "learning_rate": 5.967863493589498e-06, + "loss": 1.0169, + "step": 52222 + }, + { + "epoch": 3.86, + "learning_rate": 5.9674984870365784e-06, + "loss": 1.0536, + "step": 52223 + }, + { + "epoch": 3.86, + "learning_rate": 5.967133486899304e-06, + "loss": 1.0612, + "step": 52224 + }, + { + "epoch": 3.86, + "learning_rate": 5.9667684931782676e-06, + "loss": 1.0159, + "step": 52225 + }, + { + "epoch": 3.86, + "learning_rate": 5.9664035058740385e-06, + "loss": 1.0553, + "step": 52226 + }, + { + "epoch": 3.86, + "learning_rate": 5.966038524987211e-06, + "loss": 1.0565, + "step": 52227 + }, + { + "epoch": 3.86, + "learning_rate": 5.9656735505183514e-06, + "loss": 0.9912, + "step": 52228 + }, + { + "epoch": 3.86, + "learning_rate": 5.965308582468055e-06, + "loss": 1.0123, + "step": 52229 + }, + { + "epoch": 3.86, + "learning_rate": 5.964943620836887e-06, + "loss": 1.0353, + "step": 52230 + }, + { + "epoch": 3.86, + "learning_rate": 5.964578665625444e-06, + "loss": 0.9787, + "step": 52231 + }, + { + "epoch": 3.86, + "learning_rate": 5.964213716834295e-06, + "loss": 0.9482, + "step": 52232 + }, + { + "epoch": 3.86, + "learning_rate": 5.963848774464023e-06, + "loss": 0.9593, + "step": 52233 + }, + { + "epoch": 3.86, + "learning_rate": 5.963483838515212e-06, + "loss": 0.9516, + "step": 52234 + }, + { + "epoch": 3.86, + "learning_rate": 5.963118908988438e-06, + "loss": 1.0018, + "step": 52235 + }, + { + "epoch": 3.86, + "learning_rate": 5.9627539858842865e-06, + "loss": 0.9201, + "step": 52236 + }, + { + "epoch": 3.86, + "learning_rate": 5.962389069203333e-06, + "loss": 0.9974, + "step": 52237 + }, + { + "epoch": 3.86, + "learning_rate": 5.9620241589461625e-06, + "loss": 0.9695, + "step": 52238 + }, + { + "epoch": 3.86, + "learning_rate": 5.961659255113353e-06, + "loss": 1.0591, + "step": 52239 + }, + { + "epoch": 3.86, + "learning_rate": 5.9612943577054876e-06, + "loss": 0.9335, + "step": 52240 + }, + { + "epoch": 3.86, + "learning_rate": 5.960929466723144e-06, + "loss": 1.0252, + "step": 52241 + }, + { + "epoch": 3.86, + "learning_rate": 5.9605645821669026e-06, + "loss": 1.0459, + "step": 52242 + }, + { + "epoch": 3.86, + "learning_rate": 5.9601997040373464e-06, + "loss": 0.9127, + "step": 52243 + }, + { + "epoch": 3.86, + "learning_rate": 5.959834832335054e-06, + "loss": 1.0812, + "step": 52244 + }, + { + "epoch": 3.86, + "learning_rate": 5.959469967060607e-06, + "loss": 1.04, + "step": 52245 + }, + { + "epoch": 3.86, + "learning_rate": 5.9591051082145845e-06, + "loss": 1.0062, + "step": 52246 + }, + { + "epoch": 3.86, + "learning_rate": 5.958740255797571e-06, + "loss": 0.8978, + "step": 52247 + }, + { + "epoch": 3.86, + "learning_rate": 5.95837540981014e-06, + "loss": 1.0198, + "step": 52248 + }, + { + "epoch": 3.86, + "learning_rate": 5.958010570252881e-06, + "loss": 0.9872, + "step": 52249 + }, + { + "epoch": 3.86, + "learning_rate": 5.957645737126366e-06, + "loss": 0.9997, + "step": 52250 + }, + { + "epoch": 3.86, + "learning_rate": 5.957280910431178e-06, + "loss": 1.0217, + "step": 52251 + }, + { + "epoch": 3.86, + "learning_rate": 5.9569160901679e-06, + "loss": 0.9418, + "step": 52252 + }, + { + "epoch": 3.86, + "learning_rate": 5.956551276337108e-06, + "loss": 0.9716, + "step": 52253 + }, + { + "epoch": 3.86, + "learning_rate": 5.956186468939389e-06, + "loss": 0.9605, + "step": 52254 + }, + { + "epoch": 3.86, + "learning_rate": 5.955821667975316e-06, + "loss": 0.9185, + "step": 52255 + }, + { + "epoch": 3.86, + "learning_rate": 5.9554568734454755e-06, + "loss": 1.0167, + "step": 52256 + }, + { + "epoch": 3.86, + "learning_rate": 5.955092085350442e-06, + "loss": 0.9784, + "step": 52257 + }, + { + "epoch": 3.86, + "learning_rate": 5.9547273036908015e-06, + "loss": 1.0121, + "step": 52258 + }, + { + "epoch": 3.86, + "learning_rate": 5.9543625284671325e-06, + "loss": 1.0503, + "step": 52259 + }, + { + "epoch": 3.86, + "learning_rate": 5.9539977596800126e-06, + "loss": 1.0113, + "step": 52260 + }, + { + "epoch": 3.86, + "learning_rate": 5.953632997330028e-06, + "loss": 1.0327, + "step": 52261 + }, + { + "epoch": 3.86, + "learning_rate": 5.953268241417751e-06, + "loss": 1.0683, + "step": 52262 + }, + { + "epoch": 3.86, + "learning_rate": 5.952903491943769e-06, + "loss": 0.9209, + "step": 52263 + }, + { + "epoch": 3.86, + "learning_rate": 5.9525387489086585e-06, + "loss": 0.9693, + "step": 52264 + }, + { + "epoch": 3.86, + "learning_rate": 5.952174012313002e-06, + "loss": 0.9426, + "step": 52265 + }, + { + "epoch": 3.86, + "learning_rate": 5.951809282157377e-06, + "loss": 1.0757, + "step": 52266 + }, + { + "epoch": 3.86, + "learning_rate": 5.9514445584423676e-06, + "loss": 0.8941, + "step": 52267 + }, + { + "epoch": 3.86, + "learning_rate": 5.951079841168554e-06, + "loss": 1.049, + "step": 52268 + }, + { + "epoch": 3.86, + "learning_rate": 5.9507151303365085e-06, + "loss": 1.0703, + "step": 52269 + }, + { + "epoch": 3.86, + "learning_rate": 5.950350425946823e-06, + "loss": 1.0817, + "step": 52270 + }, + { + "epoch": 3.86, + "learning_rate": 5.9499857280000674e-06, + "loss": 1.0094, + "step": 52271 + }, + { + "epoch": 3.86, + "learning_rate": 5.949621036496831e-06, + "loss": 0.9434, + "step": 52272 + }, + { + "epoch": 3.86, + "learning_rate": 5.949256351437684e-06, + "loss": 0.9546, + "step": 52273 + }, + { + "epoch": 3.86, + "learning_rate": 5.948891672823218e-06, + "loss": 1.1174, + "step": 52274 + }, + { + "epoch": 3.86, + "learning_rate": 5.9485270006540005e-06, + "loss": 1.0758, + "step": 52275 + }, + { + "epoch": 3.86, + "learning_rate": 5.948162334930626e-06, + "loss": 0.8158, + "step": 52276 + }, + { + "epoch": 3.86, + "learning_rate": 5.947797675653666e-06, + "loss": 0.9654, + "step": 52277 + }, + { + "epoch": 3.86, + "learning_rate": 5.947433022823699e-06, + "loss": 0.9741, + "step": 52278 + }, + { + "epoch": 3.86, + "learning_rate": 5.94706837644131e-06, + "loss": 0.9891, + "step": 52279 + }, + { + "epoch": 3.86, + "learning_rate": 5.946703736507073e-06, + "loss": 1.0219, + "step": 52280 + }, + { + "epoch": 3.86, + "learning_rate": 5.946339103021577e-06, + "loss": 1.0044, + "step": 52281 + }, + { + "epoch": 3.86, + "learning_rate": 5.945974475985395e-06, + "loss": 0.9946, + "step": 52282 + }, + { + "epoch": 3.86, + "learning_rate": 5.945609855399113e-06, + "loss": 0.9054, + "step": 52283 + }, + { + "epoch": 3.86, + "learning_rate": 5.945245241263303e-06, + "loss": 0.9937, + "step": 52284 + }, + { + "epoch": 3.86, + "learning_rate": 5.944880633578554e-06, + "loss": 0.9725, + "step": 52285 + }, + { + "epoch": 3.86, + "learning_rate": 5.944516032345438e-06, + "loss": 1.0006, + "step": 52286 + }, + { + "epoch": 3.86, + "learning_rate": 5.944151437564543e-06, + "loss": 1.0085, + "step": 52287 + }, + { + "epoch": 3.86, + "learning_rate": 5.943786849236444e-06, + "loss": 1.1239, + "step": 52288 + }, + { + "epoch": 3.86, + "learning_rate": 5.94342226736172e-06, + "loss": 0.9527, + "step": 52289 + }, + { + "epoch": 3.86, + "learning_rate": 5.943057691940954e-06, + "loss": 0.9295, + "step": 52290 + }, + { + "epoch": 3.86, + "learning_rate": 5.942693122974725e-06, + "loss": 1.0922, + "step": 52291 + }, + { + "epoch": 3.86, + "learning_rate": 5.942328560463615e-06, + "loss": 1.066, + "step": 52292 + }, + { + "epoch": 3.86, + "learning_rate": 5.941964004408198e-06, + "loss": 1.1202, + "step": 52293 + }, + { + "epoch": 3.86, + "learning_rate": 5.941599454809065e-06, + "loss": 0.9965, + "step": 52294 + }, + { + "epoch": 3.86, + "learning_rate": 5.941234911666782e-06, + "loss": 1.1509, + "step": 52295 + }, + { + "epoch": 3.86, + "learning_rate": 5.940870374981942e-06, + "loss": 0.997, + "step": 52296 + }, + { + "epoch": 3.86, + "learning_rate": 5.940505844755118e-06, + "loss": 0.9442, + "step": 52297 + }, + { + "epoch": 3.86, + "learning_rate": 5.940141320986888e-06, + "loss": 0.9302, + "step": 52298 + }, + { + "epoch": 3.86, + "learning_rate": 5.939776803677838e-06, + "loss": 0.9872, + "step": 52299 + }, + { + "epoch": 3.86, + "learning_rate": 5.9394122928285434e-06, + "loss": 0.8285, + "step": 52300 + }, + { + "epoch": 3.86, + "learning_rate": 5.939047788439587e-06, + "loss": 0.8993, + "step": 52301 + }, + { + "epoch": 3.86, + "learning_rate": 5.938683290511544e-06, + "loss": 1.0752, + "step": 52302 + }, + { + "epoch": 3.86, + "learning_rate": 5.938318799045002e-06, + "loss": 0.8811, + "step": 52303 + }, + { + "epoch": 3.86, + "learning_rate": 5.9379543140405346e-06, + "loss": 1.0822, + "step": 52304 + }, + { + "epoch": 3.86, + "learning_rate": 5.937589835498725e-06, + "loss": 0.9193, + "step": 52305 + }, + { + "epoch": 3.86, + "learning_rate": 5.9372253634201515e-06, + "loss": 1.0275, + "step": 52306 + }, + { + "epoch": 3.86, + "learning_rate": 5.9368608978053934e-06, + "loss": 1.0363, + "step": 52307 + }, + { + "epoch": 3.87, + "learning_rate": 5.9364964386550324e-06, + "loss": 1.053, + "step": 52308 + }, + { + "epoch": 3.87, + "learning_rate": 5.936131985969646e-06, + "loss": 1.0151, + "step": 52309 + }, + { + "epoch": 3.87, + "learning_rate": 5.935767539749817e-06, + "loss": 1.2025, + "step": 52310 + }, + { + "epoch": 3.87, + "learning_rate": 5.935403099996122e-06, + "loss": 1.1044, + "step": 52311 + }, + { + "epoch": 3.87, + "learning_rate": 5.9350386667091455e-06, + "loss": 0.9126, + "step": 52312 + }, + { + "epoch": 3.87, + "learning_rate": 5.934674239889462e-06, + "loss": 1.0485, + "step": 52313 + }, + { + "epoch": 3.87, + "learning_rate": 5.9343098195376536e-06, + "loss": 0.9905, + "step": 52314 + }, + { + "epoch": 3.87, + "learning_rate": 5.9339454056543045e-06, + "loss": 0.9946, + "step": 52315 + }, + { + "epoch": 3.87, + "learning_rate": 5.933580998239983e-06, + "loss": 0.9965, + "step": 52316 + }, + { + "epoch": 3.87, + "learning_rate": 5.933216597295282e-06, + "loss": 0.992, + "step": 52317 + }, + { + "epoch": 3.87, + "learning_rate": 5.9328522028207694e-06, + "loss": 0.9773, + "step": 52318 + }, + { + "epoch": 3.87, + "learning_rate": 5.9324878148170366e-06, + "loss": 0.9641, + "step": 52319 + }, + { + "epoch": 3.87, + "learning_rate": 5.9321234332846515e-06, + "loss": 1.0878, + "step": 52320 + }, + { + "epoch": 3.87, + "learning_rate": 5.931759058224205e-06, + "loss": 0.9848, + "step": 52321 + }, + { + "epoch": 3.87, + "learning_rate": 5.931394689636268e-06, + "loss": 0.963, + "step": 52322 + }, + { + "epoch": 3.87, + "learning_rate": 5.931030327521428e-06, + "loss": 0.9969, + "step": 52323 + }, + { + "epoch": 3.87, + "learning_rate": 5.930665971880259e-06, + "loss": 0.9716, + "step": 52324 + }, + { + "epoch": 3.87, + "learning_rate": 5.930301622713339e-06, + "loss": 0.968, + "step": 52325 + }, + { + "epoch": 3.87, + "learning_rate": 5.929937280021254e-06, + "loss": 0.9704, + "step": 52326 + }, + { + "epoch": 3.87, + "learning_rate": 5.929572943804578e-06, + "loss": 0.9924, + "step": 52327 + }, + { + "epoch": 3.87, + "learning_rate": 5.929208614063896e-06, + "loss": 1.0879, + "step": 52328 + }, + { + "epoch": 3.87, + "learning_rate": 5.928844290799782e-06, + "loss": 1.0747, + "step": 52329 + }, + { + "epoch": 3.87, + "learning_rate": 5.928479974012822e-06, + "loss": 1.0342, + "step": 52330 + }, + { + "epoch": 3.87, + "learning_rate": 5.9281156637035886e-06, + "loss": 0.9654, + "step": 52331 + }, + { + "epoch": 3.87, + "learning_rate": 5.927751359872668e-06, + "loss": 0.9405, + "step": 52332 + }, + { + "epoch": 3.87, + "learning_rate": 5.927387062520636e-06, + "loss": 0.8944, + "step": 52333 + }, + { + "epoch": 3.87, + "learning_rate": 5.927022771648071e-06, + "loss": 1.0757, + "step": 52334 + }, + { + "epoch": 3.87, + "learning_rate": 5.926658487255557e-06, + "loss": 0.9739, + "step": 52335 + }, + { + "epoch": 3.87, + "learning_rate": 5.926294209343668e-06, + "loss": 1.1004, + "step": 52336 + }, + { + "epoch": 3.87, + "learning_rate": 5.92592993791299e-06, + "loss": 0.9575, + "step": 52337 + }, + { + "epoch": 3.87, + "learning_rate": 5.925565672964095e-06, + "loss": 1.0509, + "step": 52338 + }, + { + "epoch": 3.87, + "learning_rate": 5.925201414497573e-06, + "loss": 0.899, + "step": 52339 + }, + { + "epoch": 3.87, + "learning_rate": 5.92483716251399e-06, + "loss": 1.0187, + "step": 52340 + }, + { + "epoch": 3.87, + "learning_rate": 5.924472917013939e-06, + "loss": 1.0167, + "step": 52341 + }, + { + "epoch": 3.87, + "learning_rate": 5.924108677997988e-06, + "loss": 0.9721, + "step": 52342 + }, + { + "epoch": 3.87, + "learning_rate": 5.923744445466728e-06, + "loss": 1.0798, + "step": 52343 + }, + { + "epoch": 3.87, + "learning_rate": 5.923380219420729e-06, + "loss": 1.0497, + "step": 52344 + }, + { + "epoch": 3.87, + "learning_rate": 5.923015999860573e-06, + "loss": 0.9526, + "step": 52345 + }, + { + "epoch": 3.87, + "learning_rate": 5.922651786786841e-06, + "loss": 0.9569, + "step": 52346 + }, + { + "epoch": 3.87, + "learning_rate": 5.92228758020011e-06, + "loss": 0.9623, + "step": 52347 + }, + { + "epoch": 3.87, + "learning_rate": 5.9219233801009635e-06, + "loss": 1.0559, + "step": 52348 + }, + { + "epoch": 3.87, + "learning_rate": 5.921559186489976e-06, + "loss": 0.8345, + "step": 52349 + }, + { + "epoch": 3.87, + "learning_rate": 5.921194999367732e-06, + "loss": 0.9924, + "step": 52350 + }, + { + "epoch": 3.87, + "learning_rate": 5.920830818734806e-06, + "loss": 1.0452, + "step": 52351 + }, + { + "epoch": 3.87, + "learning_rate": 5.920466644591782e-06, + "loss": 0.9808, + "step": 52352 + }, + { + "epoch": 3.87, + "learning_rate": 5.920102476939238e-06, + "loss": 1.0451, + "step": 52353 + }, + { + "epoch": 3.87, + "learning_rate": 5.91973831577775e-06, + "loss": 0.977, + "step": 52354 + }, + { + "epoch": 3.87, + "learning_rate": 5.9193741611079e-06, + "loss": 0.9988, + "step": 52355 + }, + { + "epoch": 3.87, + "learning_rate": 5.919010012930268e-06, + "loss": 1.0002, + "step": 52356 + }, + { + "epoch": 3.87, + "learning_rate": 5.918645871245434e-06, + "loss": 0.9766, + "step": 52357 + }, + { + "epoch": 3.87, + "learning_rate": 5.918281736053973e-06, + "loss": 1.0179, + "step": 52358 + }, + { + "epoch": 3.87, + "learning_rate": 5.917917607356472e-06, + "loss": 0.9979, + "step": 52359 + }, + { + "epoch": 3.87, + "learning_rate": 5.9175534851535e-06, + "loss": 0.9405, + "step": 52360 + }, + { + "epoch": 3.87, + "learning_rate": 5.917189369445646e-06, + "loss": 0.9217, + "step": 52361 + }, + { + "epoch": 3.87, + "learning_rate": 5.916825260233488e-06, + "loss": 1.0032, + "step": 52362 + }, + { + "epoch": 3.87, + "learning_rate": 5.916461157517594e-06, + "loss": 1.0914, + "step": 52363 + }, + { + "epoch": 3.87, + "learning_rate": 5.91609706129856e-06, + "loss": 1.0649, + "step": 52364 + }, + { + "epoch": 3.87, + "learning_rate": 5.915732971576951e-06, + "loss": 0.9939, + "step": 52365 + }, + { + "epoch": 3.87, + "learning_rate": 5.915368888353359e-06, + "loss": 1.0086, + "step": 52366 + }, + { + "epoch": 3.87, + "learning_rate": 5.915004811628348e-06, + "loss": 0.9487, + "step": 52367 + }, + { + "epoch": 3.87, + "learning_rate": 5.9146407414025155e-06, + "loss": 1.0713, + "step": 52368 + }, + { + "epoch": 3.87, + "learning_rate": 5.914276677676423e-06, + "loss": 1.0004, + "step": 52369 + }, + { + "epoch": 3.87, + "learning_rate": 5.913912620450664e-06, + "loss": 1.0728, + "step": 52370 + }, + { + "epoch": 3.87, + "learning_rate": 5.9135485697258095e-06, + "loss": 0.9737, + "step": 52371 + }, + { + "epoch": 3.87, + "learning_rate": 5.91318452550244e-06, + "loss": 1.0712, + "step": 52372 + }, + { + "epoch": 3.87, + "learning_rate": 5.912820487781135e-06, + "loss": 1.154, + "step": 52373 + }, + { + "epoch": 3.87, + "learning_rate": 5.912456456562474e-06, + "loss": 0.9821, + "step": 52374 + }, + { + "epoch": 3.87, + "learning_rate": 5.912092431847039e-06, + "loss": 1.0568, + "step": 52375 + }, + { + "epoch": 3.87, + "learning_rate": 5.911728413635403e-06, + "loss": 1.0647, + "step": 52376 + }, + { + "epoch": 3.87, + "learning_rate": 5.9113644019281505e-06, + "loss": 1.0259, + "step": 52377 + }, + { + "epoch": 3.87, + "learning_rate": 5.911000396725856e-06, + "loss": 1.0499, + "step": 52378 + }, + { + "epoch": 3.87, + "learning_rate": 5.910636398029106e-06, + "loss": 1.1323, + "step": 52379 + }, + { + "epoch": 3.87, + "learning_rate": 5.910272405838473e-06, + "loss": 1.0563, + "step": 52380 + }, + { + "epoch": 3.87, + "learning_rate": 5.9099084201545365e-06, + "loss": 0.9629, + "step": 52381 + }, + { + "epoch": 3.87, + "learning_rate": 5.909544440977881e-06, + "loss": 0.9358, + "step": 52382 + }, + { + "epoch": 3.87, + "learning_rate": 5.909180468309077e-06, + "loss": 0.9966, + "step": 52383 + }, + { + "epoch": 3.87, + "learning_rate": 5.908816502148714e-06, + "loss": 1.0218, + "step": 52384 + }, + { + "epoch": 3.87, + "learning_rate": 5.908452542497357e-06, + "loss": 1.0479, + "step": 52385 + }, + { + "epoch": 3.87, + "learning_rate": 5.908088589355602e-06, + "loss": 0.9471, + "step": 52386 + }, + { + "epoch": 3.87, + "learning_rate": 5.907724642724013e-06, + "loss": 1.0491, + "step": 52387 + }, + { + "epoch": 3.87, + "learning_rate": 5.907360702603181e-06, + "loss": 0.9686, + "step": 52388 + }, + { + "epoch": 3.87, + "learning_rate": 5.9069967689936735e-06, + "loss": 1.0291, + "step": 52389 + }, + { + "epoch": 3.87, + "learning_rate": 5.9066328418960824e-06, + "loss": 1.0555, + "step": 52390 + }, + { + "epoch": 3.87, + "learning_rate": 5.906268921310977e-06, + "loss": 0.9822, + "step": 52391 + }, + { + "epoch": 3.87, + "learning_rate": 5.905905007238937e-06, + "loss": 1.06, + "step": 52392 + }, + { + "epoch": 3.87, + "learning_rate": 5.9055410996805475e-06, + "loss": 1.0854, + "step": 52393 + }, + { + "epoch": 3.87, + "learning_rate": 5.9051771986363796e-06, + "loss": 0.9194, + "step": 52394 + }, + { + "epoch": 3.87, + "learning_rate": 5.904813304107019e-06, + "loss": 0.9493, + "step": 52395 + }, + { + "epoch": 3.87, + "learning_rate": 5.904449416093038e-06, + "loss": 0.9664, + "step": 52396 + }, + { + "epoch": 3.87, + "learning_rate": 5.904085534595024e-06, + "loss": 1.0684, + "step": 52397 + }, + { + "epoch": 3.87, + "learning_rate": 5.903721659613548e-06, + "loss": 1.0024, + "step": 52398 + }, + { + "epoch": 3.87, + "learning_rate": 5.9033577911491955e-06, + "loss": 1.0025, + "step": 52399 + }, + { + "epoch": 3.87, + "learning_rate": 5.90299392920254e-06, + "loss": 0.9228, + "step": 52400 + }, + { + "epoch": 3.87, + "learning_rate": 5.902630073774163e-06, + "loss": 1.0309, + "step": 52401 + }, + { + "epoch": 3.87, + "learning_rate": 5.902266224864643e-06, + "loss": 0.9588, + "step": 52402 + }, + { + "epoch": 3.87, + "learning_rate": 5.901902382474558e-06, + "loss": 1.0018, + "step": 52403 + }, + { + "epoch": 3.87, + "learning_rate": 5.9015385466044885e-06, + "loss": 1.0009, + "step": 52404 + }, + { + "epoch": 3.87, + "learning_rate": 5.901174717255012e-06, + "loss": 1.0061, + "step": 52405 + }, + { + "epoch": 3.87, + "learning_rate": 5.900810894426708e-06, + "loss": 0.9773, + "step": 52406 + }, + { + "epoch": 3.87, + "learning_rate": 5.900447078120155e-06, + "loss": 0.9838, + "step": 52407 + }, + { + "epoch": 3.87, + "learning_rate": 5.900083268335932e-06, + "loss": 1.0194, + "step": 52408 + }, + { + "epoch": 3.87, + "learning_rate": 5.899719465074623e-06, + "loss": 0.9767, + "step": 52409 + }, + { + "epoch": 3.87, + "learning_rate": 5.8993556683367925e-06, + "loss": 0.8987, + "step": 52410 + }, + { + "epoch": 3.87, + "learning_rate": 5.898991878123036e-06, + "loss": 1.014, + "step": 52411 + }, + { + "epoch": 3.87, + "learning_rate": 5.898628094433918e-06, + "loss": 1.0536, + "step": 52412 + }, + { + "epoch": 3.87, + "learning_rate": 5.898264317270031e-06, + "loss": 0.9648, + "step": 52413 + }, + { + "epoch": 3.87, + "learning_rate": 5.897900546631941e-06, + "loss": 1.0544, + "step": 52414 + }, + { + "epoch": 3.87, + "learning_rate": 5.897536782520237e-06, + "loss": 1.1039, + "step": 52415 + }, + { + "epoch": 3.87, + "learning_rate": 5.897173024935491e-06, + "loss": 1.03, + "step": 52416 + }, + { + "epoch": 3.87, + "learning_rate": 5.896809273878286e-06, + "loss": 0.9273, + "step": 52417 + }, + { + "epoch": 3.87, + "learning_rate": 5.896445529349197e-06, + "loss": 1.0263, + "step": 52418 + }, + { + "epoch": 3.87, + "learning_rate": 5.896081791348803e-06, + "loss": 1.0243, + "step": 52419 + }, + { + "epoch": 3.87, + "learning_rate": 5.895718059877687e-06, + "loss": 1.0791, + "step": 52420 + }, + { + "epoch": 3.87, + "learning_rate": 5.895354334936422e-06, + "loss": 1.0637, + "step": 52421 + }, + { + "epoch": 3.87, + "learning_rate": 5.894990616525592e-06, + "loss": 0.9678, + "step": 52422 + }, + { + "epoch": 3.87, + "learning_rate": 5.894626904645771e-06, + "loss": 1.0328, + "step": 52423 + }, + { + "epoch": 3.87, + "learning_rate": 5.894263199297542e-06, + "loss": 1.0523, + "step": 52424 + }, + { + "epoch": 3.87, + "learning_rate": 5.893899500481479e-06, + "loss": 1.0024, + "step": 52425 + }, + { + "epoch": 3.87, + "learning_rate": 5.893535808198165e-06, + "loss": 1.0184, + "step": 52426 + }, + { + "epoch": 3.87, + "learning_rate": 5.8931721224481804e-06, + "loss": 0.9928, + "step": 52427 + }, + { + "epoch": 3.87, + "learning_rate": 5.892808443232093e-06, + "loss": 0.9032, + "step": 52428 + }, + { + "epoch": 3.87, + "learning_rate": 5.892444770550495e-06, + "loss": 1.0042, + "step": 52429 + }, + { + "epoch": 3.87, + "learning_rate": 5.8920811044039525e-06, + "loss": 1.0034, + "step": 52430 + }, + { + "epoch": 3.87, + "learning_rate": 5.891717444793056e-06, + "loss": 0.8381, + "step": 52431 + }, + { + "epoch": 3.87, + "learning_rate": 5.891353791718373e-06, + "loss": 1.029, + "step": 52432 + }, + { + "epoch": 3.87, + "learning_rate": 5.890990145180492e-06, + "loss": 1.0993, + "step": 52433 + }, + { + "epoch": 3.87, + "learning_rate": 5.890626505179982e-06, + "loss": 0.9678, + "step": 52434 + }, + { + "epoch": 3.87, + "learning_rate": 5.890262871717432e-06, + "loss": 1.1345, + "step": 52435 + }, + { + "epoch": 3.87, + "learning_rate": 5.889899244793412e-06, + "loss": 0.9561, + "step": 52436 + }, + { + "epoch": 3.87, + "learning_rate": 5.8895356244085026e-06, + "loss": 1.0214, + "step": 52437 + }, + { + "epoch": 3.87, + "learning_rate": 5.889172010563285e-06, + "loss": 0.8534, + "step": 52438 + }, + { + "epoch": 3.87, + "learning_rate": 5.888808403258334e-06, + "loss": 1.0053, + "step": 52439 + }, + { + "epoch": 3.87, + "learning_rate": 5.888444802494232e-06, + "loss": 1.0311, + "step": 52440 + }, + { + "epoch": 3.87, + "learning_rate": 5.888081208271553e-06, + "loss": 0.9924, + "step": 52441 + }, + { + "epoch": 3.87, + "learning_rate": 5.88771762059088e-06, + "loss": 0.9381, + "step": 52442 + }, + { + "epoch": 3.87, + "learning_rate": 5.887354039452789e-06, + "loss": 1.0097, + "step": 52443 + }, + { + "epoch": 3.88, + "learning_rate": 5.88699046485786e-06, + "loss": 1.0762, + "step": 52444 + }, + { + "epoch": 3.88, + "learning_rate": 5.8866268968066674e-06, + "loss": 1.0315, + "step": 52445 + }, + { + "epoch": 3.88, + "learning_rate": 5.886263335299796e-06, + "loss": 1.1002, + "step": 52446 + }, + { + "epoch": 3.88, + "learning_rate": 5.885899780337819e-06, + "loss": 1.0515, + "step": 52447 + }, + { + "epoch": 3.88, + "learning_rate": 5.885536231921316e-06, + "loss": 0.9087, + "step": 52448 + }, + { + "epoch": 3.88, + "learning_rate": 5.8851726900508675e-06, + "loss": 1.1123, + "step": 52449 + }, + { + "epoch": 3.88, + "learning_rate": 5.884809154727049e-06, + "loss": 0.9929, + "step": 52450 + }, + { + "epoch": 3.88, + "learning_rate": 5.884445625950441e-06, + "loss": 1.0287, + "step": 52451 + }, + { + "epoch": 3.88, + "learning_rate": 5.88408210372162e-06, + "loss": 0.9739, + "step": 52452 + }, + { + "epoch": 3.88, + "learning_rate": 5.883718588041168e-06, + "loss": 0.9283, + "step": 52453 + }, + { + "epoch": 3.88, + "learning_rate": 5.8833550789096585e-06, + "loss": 1.1104, + "step": 52454 + }, + { + "epoch": 3.88, + "learning_rate": 5.882991576327672e-06, + "loss": 1.016, + "step": 52455 + }, + { + "epoch": 3.88, + "learning_rate": 5.882628080295791e-06, + "loss": 0.951, + "step": 52456 + }, + { + "epoch": 3.88, + "learning_rate": 5.882264590814584e-06, + "loss": 0.909, + "step": 52457 + }, + { + "epoch": 3.88, + "learning_rate": 5.8819011078846415e-06, + "loss": 1.0303, + "step": 52458 + }, + { + "epoch": 3.88, + "learning_rate": 5.881537631506528e-06, + "loss": 1.0557, + "step": 52459 + }, + { + "epoch": 3.88, + "learning_rate": 5.881174161680836e-06, + "loss": 0.9436, + "step": 52460 + }, + { + "epoch": 3.88, + "learning_rate": 5.880810698408133e-06, + "loss": 0.9708, + "step": 52461 + }, + { + "epoch": 3.88, + "learning_rate": 5.880447241689002e-06, + "loss": 0.9608, + "step": 52462 + }, + { + "epoch": 3.88, + "learning_rate": 5.88008379152402e-06, + "loss": 0.9214, + "step": 52463 + }, + { + "epoch": 3.88, + "learning_rate": 5.879720347913767e-06, + "loss": 0.9061, + "step": 52464 + }, + { + "epoch": 3.88, + "learning_rate": 5.879356910858821e-06, + "loss": 1.038, + "step": 52465 + }, + { + "epoch": 3.88, + "learning_rate": 5.878993480359756e-06, + "loss": 0.951, + "step": 52466 + }, + { + "epoch": 3.88, + "learning_rate": 5.878630056417156e-06, + "loss": 1.1025, + "step": 52467 + }, + { + "epoch": 3.88, + "learning_rate": 5.878266639031596e-06, + "loss": 1.0329, + "step": 52468 + }, + { + "epoch": 3.88, + "learning_rate": 5.877903228203654e-06, + "loss": 0.986, + "step": 52469 + }, + { + "epoch": 3.88, + "learning_rate": 5.877539823933909e-06, + "loss": 1.0189, + "step": 52470 + }, + { + "epoch": 3.88, + "learning_rate": 5.87717642622294e-06, + "loss": 1.0501, + "step": 52471 + }, + { + "epoch": 3.88, + "learning_rate": 5.876813035071323e-06, + "loss": 0.9822, + "step": 52472 + }, + { + "epoch": 3.88, + "learning_rate": 5.876449650479639e-06, + "loss": 0.9141, + "step": 52473 + }, + { + "epoch": 3.88, + "learning_rate": 5.876086272448467e-06, + "loss": 1.0192, + "step": 52474 + }, + { + "epoch": 3.88, + "learning_rate": 5.875722900978376e-06, + "loss": 0.969, + "step": 52475 + }, + { + "epoch": 3.88, + "learning_rate": 5.875359536069958e-06, + "loss": 0.8479, + "step": 52476 + }, + { + "epoch": 3.88, + "learning_rate": 5.8749961777237775e-06, + "loss": 1.0827, + "step": 52477 + }, + { + "epoch": 3.88, + "learning_rate": 5.874632825940425e-06, + "loss": 1.0688, + "step": 52478 + }, + { + "epoch": 3.88, + "learning_rate": 5.874269480720467e-06, + "loss": 1.1573, + "step": 52479 + }, + { + "epoch": 3.88, + "learning_rate": 5.8739061420644925e-06, + "loss": 0.9041, + "step": 52480 + }, + { + "epoch": 3.88, + "learning_rate": 5.873542809973068e-06, + "loss": 1.0138, + "step": 52481 + }, + { + "epoch": 3.88, + "learning_rate": 5.873179484446785e-06, + "loss": 0.9615, + "step": 52482 + }, + { + "epoch": 3.88, + "learning_rate": 5.872816165486213e-06, + "loss": 1.0527, + "step": 52483 + }, + { + "epoch": 3.88, + "learning_rate": 5.872452853091928e-06, + "loss": 0.8593, + "step": 52484 + }, + { + "epoch": 3.88, + "learning_rate": 5.8720895472645135e-06, + "loss": 1.0144, + "step": 52485 + }, + { + "epoch": 3.88, + "learning_rate": 5.871726248004543e-06, + "loss": 1.0396, + "step": 52486 + }, + { + "epoch": 3.88, + "learning_rate": 5.871362955312601e-06, + "loss": 1.0078, + "step": 52487 + }, + { + "epoch": 3.88, + "learning_rate": 5.8709996691892565e-06, + "loss": 1.0101, + "step": 52488 + }, + { + "epoch": 3.88, + "learning_rate": 5.870636389635097e-06, + "loss": 1.0782, + "step": 52489 + }, + { + "epoch": 3.88, + "learning_rate": 5.870273116650693e-06, + "loss": 0.9979, + "step": 52490 + }, + { + "epoch": 3.88, + "learning_rate": 5.869909850236628e-06, + "loss": 1.0499, + "step": 52491 + }, + { + "epoch": 3.88, + "learning_rate": 5.869546590393476e-06, + "loss": 0.9941, + "step": 52492 + }, + { + "epoch": 3.88, + "learning_rate": 5.869183337121816e-06, + "loss": 1.0157, + "step": 52493 + }, + { + "epoch": 3.88, + "learning_rate": 5.868820090422227e-06, + "loss": 1.0689, + "step": 52494 + }, + { + "epoch": 3.88, + "learning_rate": 5.868456850295284e-06, + "loss": 1.0551, + "step": 52495 + }, + { + "epoch": 3.88, + "learning_rate": 5.86809361674157e-06, + "loss": 1.1115, + "step": 52496 + }, + { + "epoch": 3.88, + "learning_rate": 5.867730389761657e-06, + "loss": 0.9812, + "step": 52497 + }, + { + "epoch": 3.88, + "learning_rate": 5.867367169356129e-06, + "loss": 0.9123, + "step": 52498 + }, + { + "epoch": 3.88, + "learning_rate": 5.867003955525557e-06, + "loss": 1.0318, + "step": 52499 + }, + { + "epoch": 3.88, + "learning_rate": 5.866640748270525e-06, + "loss": 1.0305, + "step": 52500 + }, + { + "epoch": 3.88, + "learning_rate": 5.866277547591608e-06, + "loss": 1.0999, + "step": 52501 + }, + { + "epoch": 3.88, + "learning_rate": 5.865914353489385e-06, + "loss": 0.9595, + "step": 52502 + }, + { + "epoch": 3.88, + "learning_rate": 5.865551165964435e-06, + "loss": 1.0622, + "step": 52503 + }, + { + "epoch": 3.88, + "learning_rate": 5.865187985017331e-06, + "loss": 1.035, + "step": 52504 + }, + { + "epoch": 3.88, + "learning_rate": 5.864824810648655e-06, + "loss": 0.9436, + "step": 52505 + }, + { + "epoch": 3.88, + "learning_rate": 5.86446164285898e-06, + "loss": 1.0701, + "step": 52506 + }, + { + "epoch": 3.88, + "learning_rate": 5.864098481648892e-06, + "loss": 0.9943, + "step": 52507 + }, + { + "epoch": 3.88, + "learning_rate": 5.86373532701896e-06, + "loss": 1.0099, + "step": 52508 + }, + { + "epoch": 3.88, + "learning_rate": 5.86337217896977e-06, + "loss": 0.9827, + "step": 52509 + }, + { + "epoch": 3.88, + "learning_rate": 5.8630090375018925e-06, + "loss": 0.9858, + "step": 52510 + }, + { + "epoch": 3.88, + "learning_rate": 5.862645902615911e-06, + "loss": 0.9411, + "step": 52511 + }, + { + "epoch": 3.88, + "learning_rate": 5.8622827743124e-06, + "loss": 1.0138, + "step": 52512 + }, + { + "epoch": 3.88, + "learning_rate": 5.8619196525919366e-06, + "loss": 1.1132, + "step": 52513 + }, + { + "epoch": 3.88, + "learning_rate": 5.861556537455101e-06, + "loss": 0.9752, + "step": 52514 + }, + { + "epoch": 3.88, + "learning_rate": 5.861193428902467e-06, + "loss": 1.0677, + "step": 52515 + }, + { + "epoch": 3.88, + "learning_rate": 5.860830326934619e-06, + "loss": 0.9864, + "step": 52516 + }, + { + "epoch": 3.88, + "learning_rate": 5.860467231552127e-06, + "loss": 0.94, + "step": 52517 + }, + { + "epoch": 3.88, + "learning_rate": 5.860104142755576e-06, + "loss": 1.1382, + "step": 52518 + }, + { + "epoch": 3.88, + "learning_rate": 5.859741060545538e-06, + "loss": 0.9714, + "step": 52519 + }, + { + "epoch": 3.88, + "learning_rate": 5.859377984922594e-06, + "loss": 0.961, + "step": 52520 + }, + { + "epoch": 3.88, + "learning_rate": 5.859014915887322e-06, + "loss": 0.9999, + "step": 52521 + }, + { + "epoch": 3.88, + "learning_rate": 5.858651853440293e-06, + "loss": 1.0035, + "step": 52522 + }, + { + "epoch": 3.88, + "learning_rate": 5.858288797582095e-06, + "loss": 0.9906, + "step": 52523 + }, + { + "epoch": 3.88, + "learning_rate": 5.857925748313295e-06, + "loss": 1.0824, + "step": 52524 + }, + { + "epoch": 3.88, + "learning_rate": 5.857562705634481e-06, + "loss": 0.9396, + "step": 52525 + }, + { + "epoch": 3.88, + "learning_rate": 5.8571996695462185e-06, + "loss": 1.1287, + "step": 52526 + }, + { + "epoch": 3.88, + "learning_rate": 5.8568366400490994e-06, + "loss": 1.0323, + "step": 52527 + }, + { + "epoch": 3.88, + "learning_rate": 5.856473617143689e-06, + "loss": 0.8903, + "step": 52528 + }, + { + "epoch": 3.88, + "learning_rate": 5.8561106008305735e-06, + "loss": 0.8614, + "step": 52529 + }, + { + "epoch": 3.88, + "learning_rate": 5.855747591110327e-06, + "loss": 1.0744, + "step": 52530 + }, + { + "epoch": 3.88, + "learning_rate": 5.855384587983524e-06, + "loss": 0.9727, + "step": 52531 + }, + { + "epoch": 3.88, + "learning_rate": 5.855021591450746e-06, + "loss": 1.0193, + "step": 52532 + }, + { + "epoch": 3.88, + "learning_rate": 5.8546586015125685e-06, + "loss": 1.0102, + "step": 52533 + }, + { + "epoch": 3.88, + "learning_rate": 5.85429561816957e-06, + "loss": 1.0044, + "step": 52534 + }, + { + "epoch": 3.88, + "learning_rate": 5.853932641422327e-06, + "loss": 1.031, + "step": 52535 + }, + { + "epoch": 3.88, + "learning_rate": 5.85356967127142e-06, + "loss": 1.0775, + "step": 52536 + }, + { + "epoch": 3.88, + "learning_rate": 5.8532067077174225e-06, + "loss": 1.0525, + "step": 52537 + }, + { + "epoch": 3.88, + "learning_rate": 5.852843750760915e-06, + "loss": 1.0335, + "step": 52538 + }, + { + "epoch": 3.88, + "learning_rate": 5.852480800402475e-06, + "loss": 1.0364, + "step": 52539 + }, + { + "epoch": 3.88, + "learning_rate": 5.852117856642676e-06, + "loss": 1.0371, + "step": 52540 + }, + { + "epoch": 3.88, + "learning_rate": 5.8517549194821e-06, + "loss": 1.0569, + "step": 52541 + }, + { + "epoch": 3.88, + "learning_rate": 5.85139198892132e-06, + "loss": 0.9982, + "step": 52542 + }, + { + "epoch": 3.88, + "learning_rate": 5.851029064960919e-06, + "loss": 1.0091, + "step": 52543 + }, + { + "epoch": 3.88, + "learning_rate": 5.8506661476014695e-06, + "loss": 0.8567, + "step": 52544 + }, + { + "epoch": 3.88, + "learning_rate": 5.850303236843553e-06, + "loss": 0.9833, + "step": 52545 + }, + { + "epoch": 3.88, + "learning_rate": 5.8499403326877415e-06, + "loss": 0.9988, + "step": 52546 + }, + { + "epoch": 3.88, + "learning_rate": 5.849577435134618e-06, + "loss": 1.0044, + "step": 52547 + }, + { + "epoch": 3.88, + "learning_rate": 5.849214544184759e-06, + "loss": 1.0693, + "step": 52548 + }, + { + "epoch": 3.88, + "learning_rate": 5.848851659838737e-06, + "loss": 0.9852, + "step": 52549 + }, + { + "epoch": 3.88, + "learning_rate": 5.8484887820971345e-06, + "loss": 1.0198, + "step": 52550 + }, + { + "epoch": 3.88, + "learning_rate": 5.8481259109605245e-06, + "loss": 1.064, + "step": 52551 + }, + { + "epoch": 3.88, + "learning_rate": 5.84776304642949e-06, + "loss": 1.0373, + "step": 52552 + }, + { + "epoch": 3.88, + "learning_rate": 5.847400188504602e-06, + "loss": 0.9734, + "step": 52553 + }, + { + "epoch": 3.88, + "learning_rate": 5.847037337186443e-06, + "loss": 1.0174, + "step": 52554 + }, + { + "epoch": 3.88, + "learning_rate": 5.846674492475586e-06, + "loss": 1.1874, + "step": 52555 + }, + { + "epoch": 3.88, + "learning_rate": 5.846311654372613e-06, + "loss": 1.004, + "step": 52556 + }, + { + "epoch": 3.88, + "learning_rate": 5.845948822878096e-06, + "loss": 1.0231, + "step": 52557 + }, + { + "epoch": 3.88, + "learning_rate": 5.845585997992618e-06, + "loss": 1.0247, + "step": 52558 + }, + { + "epoch": 3.88, + "learning_rate": 5.845223179716752e-06, + "loss": 1.0562, + "step": 52559 + }, + { + "epoch": 3.88, + "learning_rate": 5.844860368051075e-06, + "loss": 1.0341, + "step": 52560 + }, + { + "epoch": 3.88, + "learning_rate": 5.844497562996168e-06, + "loss": 0.9704, + "step": 52561 + }, + { + "epoch": 3.88, + "learning_rate": 5.844134764552605e-06, + "loss": 0.9986, + "step": 52562 + }, + { + "epoch": 3.88, + "learning_rate": 5.843771972720964e-06, + "loss": 1.0115, + "step": 52563 + }, + { + "epoch": 3.88, + "learning_rate": 5.8434091875018215e-06, + "loss": 0.9177, + "step": 52564 + }, + { + "epoch": 3.88, + "learning_rate": 5.843046408895757e-06, + "loss": 0.9909, + "step": 52565 + }, + { + "epoch": 3.88, + "learning_rate": 5.842683636903345e-06, + "loss": 1.0224, + "step": 52566 + }, + { + "epoch": 3.88, + "learning_rate": 5.842320871525165e-06, + "loss": 1.025, + "step": 52567 + }, + { + "epoch": 3.88, + "learning_rate": 5.841958112761795e-06, + "loss": 0.9373, + "step": 52568 + }, + { + "epoch": 3.88, + "learning_rate": 5.841595360613805e-06, + "loss": 0.9785, + "step": 52569 + }, + { + "epoch": 3.88, + "learning_rate": 5.841232615081782e-06, + "loss": 0.8885, + "step": 52570 + }, + { + "epoch": 3.88, + "learning_rate": 5.840869876166293e-06, + "loss": 0.9637, + "step": 52571 + }, + { + "epoch": 3.88, + "learning_rate": 5.840507143867927e-06, + "loss": 0.9753, + "step": 52572 + }, + { + "epoch": 3.88, + "learning_rate": 5.840144418187248e-06, + "loss": 1.0722, + "step": 52573 + }, + { + "epoch": 3.88, + "learning_rate": 5.839781699124847e-06, + "loss": 1.0641, + "step": 52574 + }, + { + "epoch": 3.88, + "learning_rate": 5.839418986681287e-06, + "loss": 0.9249, + "step": 52575 + }, + { + "epoch": 3.88, + "learning_rate": 5.839056280857157e-06, + "loss": 0.993, + "step": 52576 + }, + { + "epoch": 3.88, + "learning_rate": 5.838693581653028e-06, + "loss": 1.0864, + "step": 52577 + }, + { + "epoch": 3.88, + "learning_rate": 5.838330889069474e-06, + "loss": 0.8455, + "step": 52578 + }, + { + "epoch": 3.89, + "learning_rate": 5.8379682031070826e-06, + "loss": 0.9048, + "step": 52579 + }, + { + "epoch": 3.89, + "learning_rate": 5.837605523766418e-06, + "loss": 0.971, + "step": 52580 + }, + { + "epoch": 3.89, + "learning_rate": 5.837242851048066e-06, + "loss": 0.9243, + "step": 52581 + }, + { + "epoch": 3.89, + "learning_rate": 5.836880184952598e-06, + "loss": 1.078, + "step": 52582 + }, + { + "epoch": 3.89, + "learning_rate": 5.836517525480601e-06, + "loss": 0.9503, + "step": 52583 + }, + { + "epoch": 3.89, + "learning_rate": 5.836154872632638e-06, + "loss": 1.0089, + "step": 52584 + }, + { + "epoch": 3.89, + "learning_rate": 5.835792226409298e-06, + "loss": 0.9631, + "step": 52585 + }, + { + "epoch": 3.89, + "learning_rate": 5.835429586811151e-06, + "loss": 0.9777, + "step": 52586 + }, + { + "epoch": 3.89, + "learning_rate": 5.835066953838777e-06, + "loss": 0.9216, + "step": 52587 + }, + { + "epoch": 3.89, + "learning_rate": 5.834704327492752e-06, + "loss": 0.9598, + "step": 52588 + }, + { + "epoch": 3.89, + "learning_rate": 5.834341707773648e-06, + "loss": 1.0801, + "step": 52589 + }, + { + "epoch": 3.89, + "learning_rate": 5.833979094682051e-06, + "loss": 0.9124, + "step": 52590 + }, + { + "epoch": 3.89, + "learning_rate": 5.8336164882185345e-06, + "loss": 1.0279, + "step": 52591 + }, + { + "epoch": 3.89, + "learning_rate": 5.833253888383674e-06, + "loss": 0.9318, + "step": 52592 + }, + { + "epoch": 3.89, + "learning_rate": 5.832891295178043e-06, + "loss": 0.9229, + "step": 52593 + }, + { + "epoch": 3.89, + "learning_rate": 5.832528708602225e-06, + "loss": 0.9775, + "step": 52594 + }, + { + "epoch": 3.89, + "learning_rate": 5.832166128656795e-06, + "loss": 0.9778, + "step": 52595 + }, + { + "epoch": 3.89, + "learning_rate": 5.83180355534233e-06, + "loss": 0.9589, + "step": 52596 + }, + { + "epoch": 3.89, + "learning_rate": 5.831440988659406e-06, + "loss": 1.0302, + "step": 52597 + }, + { + "epoch": 3.89, + "learning_rate": 5.831078428608595e-06, + "loss": 0.9607, + "step": 52598 + }, + { + "epoch": 3.89, + "learning_rate": 5.8307158751904825e-06, + "loss": 0.829, + "step": 52599 + }, + { + "epoch": 3.89, + "learning_rate": 5.830353328405643e-06, + "loss": 0.9886, + "step": 52600 + }, + { + "epoch": 3.89, + "learning_rate": 5.829990788254649e-06, + "loss": 1.0385, + "step": 52601 + }, + { + "epoch": 3.89, + "learning_rate": 5.829628254738078e-06, + "loss": 0.933, + "step": 52602 + }, + { + "epoch": 3.89, + "learning_rate": 5.829265727856512e-06, + "loss": 1.0389, + "step": 52603 + }, + { + "epoch": 3.89, + "learning_rate": 5.828903207610525e-06, + "loss": 0.9988, + "step": 52604 + }, + { + "epoch": 3.89, + "learning_rate": 5.828540694000693e-06, + "loss": 0.9159, + "step": 52605 + }, + { + "epoch": 3.89, + "learning_rate": 5.828178187027592e-06, + "loss": 0.9229, + "step": 52606 + }, + { + "epoch": 3.89, + "learning_rate": 5.827815686691798e-06, + "loss": 1.0472, + "step": 52607 + }, + { + "epoch": 3.89, + "learning_rate": 5.827453192993895e-06, + "loss": 0.8976, + "step": 52608 + }, + { + "epoch": 3.89, + "learning_rate": 5.827090705934449e-06, + "loss": 1.0716, + "step": 52609 + }, + { + "epoch": 3.89, + "learning_rate": 5.826728225514045e-06, + "loss": 1.0578, + "step": 52610 + }, + { + "epoch": 3.89, + "learning_rate": 5.826365751733255e-06, + "loss": 0.9271, + "step": 52611 + }, + { + "epoch": 3.89, + "learning_rate": 5.826003284592658e-06, + "loss": 0.9558, + "step": 52612 + }, + { + "epoch": 3.89, + "learning_rate": 5.825640824092828e-06, + "loss": 1.1571, + "step": 52613 + }, + { + "epoch": 3.89, + "learning_rate": 5.825278370234346e-06, + "loss": 0.986, + "step": 52614 + }, + { + "epoch": 3.89, + "learning_rate": 5.824915923017786e-06, + "loss": 0.9827, + "step": 52615 + }, + { + "epoch": 3.89, + "learning_rate": 5.824553482443726e-06, + "loss": 0.9494, + "step": 52616 + }, + { + "epoch": 3.89, + "learning_rate": 5.824191048512742e-06, + "loss": 1.0107, + "step": 52617 + }, + { + "epoch": 3.89, + "learning_rate": 5.823828621225406e-06, + "loss": 1.1315, + "step": 52618 + }, + { + "epoch": 3.89, + "learning_rate": 5.8234662005823026e-06, + "loss": 1.1259, + "step": 52619 + }, + { + "epoch": 3.89, + "learning_rate": 5.823103786584006e-06, + "loss": 1.0564, + "step": 52620 + }, + { + "epoch": 3.89, + "learning_rate": 5.82274137923109e-06, + "loss": 0.9947, + "step": 52621 + }, + { + "epoch": 3.89, + "learning_rate": 5.822378978524128e-06, + "loss": 1.0296, + "step": 52622 + }, + { + "epoch": 3.89, + "learning_rate": 5.822016584463707e-06, + "loss": 1.1041, + "step": 52623 + }, + { + "epoch": 3.89, + "learning_rate": 5.821654197050401e-06, + "loss": 1.103, + "step": 52624 + }, + { + "epoch": 3.89, + "learning_rate": 5.8212918162847755e-06, + "loss": 0.9228, + "step": 52625 + }, + { + "epoch": 3.89, + "learning_rate": 5.820929442167419e-06, + "loss": 0.9503, + "step": 52626 + }, + { + "epoch": 3.89, + "learning_rate": 5.820567074698899e-06, + "loss": 0.8608, + "step": 52627 + }, + { + "epoch": 3.89, + "learning_rate": 5.820204713879807e-06, + "loss": 1.1028, + "step": 52628 + }, + { + "epoch": 3.89, + "learning_rate": 5.819842359710701e-06, + "loss": 0.9516, + "step": 52629 + }, + { + "epoch": 3.89, + "learning_rate": 5.819480012192169e-06, + "loss": 1.0614, + "step": 52630 + }, + { + "epoch": 3.89, + "learning_rate": 5.819117671324781e-06, + "loss": 1.0775, + "step": 52631 + }, + { + "epoch": 3.89, + "learning_rate": 5.818755337109125e-06, + "loss": 1.0394, + "step": 52632 + }, + { + "epoch": 3.89, + "learning_rate": 5.818393009545765e-06, + "loss": 1.0193, + "step": 52633 + }, + { + "epoch": 3.89, + "learning_rate": 5.8180306886352785e-06, + "loss": 1.0227, + "step": 52634 + }, + { + "epoch": 3.89, + "learning_rate": 5.817668374378249e-06, + "loss": 0.8999, + "step": 52635 + }, + { + "epoch": 3.89, + "learning_rate": 5.81730606677525e-06, + "loss": 1.0466, + "step": 52636 + }, + { + "epoch": 3.89, + "learning_rate": 5.816943765826855e-06, + "loss": 0.9704, + "step": 52637 + }, + { + "epoch": 3.89, + "learning_rate": 5.816581471533641e-06, + "loss": 1.0071, + "step": 52638 + }, + { + "epoch": 3.89, + "learning_rate": 5.816219183896189e-06, + "loss": 1.039, + "step": 52639 + }, + { + "epoch": 3.89, + "learning_rate": 5.815856902915074e-06, + "loss": 0.9769, + "step": 52640 + }, + { + "epoch": 3.89, + "learning_rate": 5.815494628590867e-06, + "loss": 0.9735, + "step": 52641 + }, + { + "epoch": 3.89, + "learning_rate": 5.815132360924152e-06, + "loss": 1.0245, + "step": 52642 + }, + { + "epoch": 3.89, + "learning_rate": 5.814770099915497e-06, + "loss": 1.0667, + "step": 52643 + }, + { + "epoch": 3.89, + "learning_rate": 5.814407845565485e-06, + "loss": 1.0281, + "step": 52644 + }, + { + "epoch": 3.89, + "learning_rate": 5.814045597874691e-06, + "loss": 0.9802, + "step": 52645 + }, + { + "epoch": 3.89, + "learning_rate": 5.8136833568436915e-06, + "loss": 0.9854, + "step": 52646 + }, + { + "epoch": 3.89, + "learning_rate": 5.813321122473057e-06, + "loss": 0.9829, + "step": 52647 + }, + { + "epoch": 3.89, + "learning_rate": 5.812958894763372e-06, + "loss": 1.0995, + "step": 52648 + }, + { + "epoch": 3.89, + "learning_rate": 5.812596673715212e-06, + "loss": 1.0202, + "step": 52649 + }, + { + "epoch": 3.89, + "learning_rate": 5.812234459329149e-06, + "loss": 1.0229, + "step": 52650 + }, + { + "epoch": 3.89, + "learning_rate": 5.811872251605761e-06, + "loss": 1.0151, + "step": 52651 + }, + { + "epoch": 3.89, + "learning_rate": 5.8115100505456205e-06, + "loss": 1.0608, + "step": 52652 + }, + { + "epoch": 3.89, + "learning_rate": 5.811147856149315e-06, + "loss": 0.9401, + "step": 52653 + }, + { + "epoch": 3.89, + "learning_rate": 5.810785668417407e-06, + "loss": 0.9724, + "step": 52654 + }, + { + "epoch": 3.89, + "learning_rate": 5.810423487350481e-06, + "loss": 1.0366, + "step": 52655 + }, + { + "epoch": 3.89, + "learning_rate": 5.810061312949112e-06, + "loss": 1.0648, + "step": 52656 + }, + { + "epoch": 3.89, + "learning_rate": 5.809699145213876e-06, + "loss": 0.8839, + "step": 52657 + }, + { + "epoch": 3.89, + "learning_rate": 5.809336984145345e-06, + "loss": 1.0724, + "step": 52658 + }, + { + "epoch": 3.89, + "learning_rate": 5.808974829744103e-06, + "loss": 0.9269, + "step": 52659 + }, + { + "epoch": 3.89, + "learning_rate": 5.8086126820107215e-06, + "loss": 0.9638, + "step": 52660 + }, + { + "epoch": 3.89, + "learning_rate": 5.808250540945777e-06, + "loss": 1.1169, + "step": 52661 + }, + { + "epoch": 3.89, + "learning_rate": 5.807888406549847e-06, + "loss": 1.0489, + "step": 52662 + }, + { + "epoch": 3.89, + "learning_rate": 5.807526278823501e-06, + "loss": 1.1481, + "step": 52663 + }, + { + "epoch": 3.89, + "learning_rate": 5.807164157767324e-06, + "loss": 1.0038, + "step": 52664 + }, + { + "epoch": 3.89, + "learning_rate": 5.8068020433818915e-06, + "loss": 1.0524, + "step": 52665 + }, + { + "epoch": 3.89, + "learning_rate": 5.806439935667776e-06, + "loss": 0.8815, + "step": 52666 + }, + { + "epoch": 3.89, + "learning_rate": 5.80607783462555e-06, + "loss": 0.96, + "step": 52667 + }, + { + "epoch": 3.89, + "learning_rate": 5.805715740255798e-06, + "loss": 0.9884, + "step": 52668 + }, + { + "epoch": 3.89, + "learning_rate": 5.805353652559094e-06, + "loss": 1.0921, + "step": 52669 + }, + { + "epoch": 3.89, + "learning_rate": 5.8049915715360096e-06, + "loss": 0.9207, + "step": 52670 + }, + { + "epoch": 3.89, + "learning_rate": 5.804629497187126e-06, + "loss": 1.0098, + "step": 52671 + }, + { + "epoch": 3.89, + "learning_rate": 5.8042674295130106e-06, + "loss": 1.0129, + "step": 52672 + }, + { + "epoch": 3.89, + "learning_rate": 5.803905368514255e-06, + "loss": 1.0583, + "step": 52673 + }, + { + "epoch": 3.89, + "learning_rate": 5.803543314191417e-06, + "loss": 1.0873, + "step": 52674 + }, + { + "epoch": 3.89, + "learning_rate": 5.803181266545088e-06, + "loss": 0.9647, + "step": 52675 + }, + { + "epoch": 3.89, + "learning_rate": 5.802819225575832e-06, + "loss": 1.0025, + "step": 52676 + }, + { + "epoch": 3.89, + "learning_rate": 5.802457191284239e-06, + "loss": 1.0273, + "step": 52677 + }, + { + "epoch": 3.89, + "learning_rate": 5.8020951636708664e-06, + "loss": 0.8632, + "step": 52678 + }, + { + "epoch": 3.89, + "learning_rate": 5.801733142736308e-06, + "loss": 1.0005, + "step": 52679 + }, + { + "epoch": 3.89, + "learning_rate": 5.8013711284811295e-06, + "loss": 0.9012, + "step": 52680 + }, + { + "epoch": 3.89, + "learning_rate": 5.8010091209059094e-06, + "loss": 0.905, + "step": 52681 + }, + { + "epoch": 3.89, + "learning_rate": 5.800647120011225e-06, + "loss": 1.031, + "step": 52682 + }, + { + "epoch": 3.89, + "learning_rate": 5.800285125797647e-06, + "loss": 0.9742, + "step": 52683 + }, + { + "epoch": 3.89, + "learning_rate": 5.7999231382657575e-06, + "loss": 1.0231, + "step": 52684 + }, + { + "epoch": 3.89, + "learning_rate": 5.799561157416131e-06, + "loss": 1.068, + "step": 52685 + }, + { + "epoch": 3.89, + "learning_rate": 5.799199183249343e-06, + "loss": 1.0286, + "step": 52686 + }, + { + "epoch": 3.89, + "learning_rate": 5.798837215765965e-06, + "loss": 1.0064, + "step": 52687 + }, + { + "epoch": 3.89, + "learning_rate": 5.798475254966581e-06, + "loss": 0.9663, + "step": 52688 + }, + { + "epoch": 3.89, + "learning_rate": 5.798113300851762e-06, + "loss": 0.9907, + "step": 52689 + }, + { + "epoch": 3.89, + "learning_rate": 5.797751353422086e-06, + "loss": 1.0214, + "step": 52690 + }, + { + "epoch": 3.89, + "learning_rate": 5.797389412678127e-06, + "loss": 0.904, + "step": 52691 + }, + { + "epoch": 3.89, + "learning_rate": 5.797027478620456e-06, + "loss": 0.9105, + "step": 52692 + }, + { + "epoch": 3.89, + "learning_rate": 5.7966655512496596e-06, + "loss": 1.0635, + "step": 52693 + }, + { + "epoch": 3.89, + "learning_rate": 5.796303630566308e-06, + "loss": 0.9404, + "step": 52694 + }, + { + "epoch": 3.89, + "learning_rate": 5.795941716570976e-06, + "loss": 1.0329, + "step": 52695 + }, + { + "epoch": 3.89, + "learning_rate": 5.795579809264238e-06, + "loss": 1.0305, + "step": 52696 + }, + { + "epoch": 3.89, + "learning_rate": 5.795217908646675e-06, + "loss": 0.929, + "step": 52697 + }, + { + "epoch": 3.89, + "learning_rate": 5.794856014718866e-06, + "loss": 0.8753, + "step": 52698 + }, + { + "epoch": 3.89, + "learning_rate": 5.794494127481372e-06, + "loss": 1.0894, + "step": 52699 + }, + { + "epoch": 3.89, + "learning_rate": 5.794132246934781e-06, + "loss": 0.9322, + "step": 52700 + }, + { + "epoch": 3.89, + "learning_rate": 5.7937703730796655e-06, + "loss": 0.9708, + "step": 52701 + }, + { + "epoch": 3.89, + "learning_rate": 5.7934085059166024e-06, + "loss": 1.0403, + "step": 52702 + }, + { + "epoch": 3.89, + "learning_rate": 5.793046645446161e-06, + "loss": 0.8221, + "step": 52703 + }, + { + "epoch": 3.89, + "learning_rate": 5.7926847916689265e-06, + "loss": 1.0027, + "step": 52704 + }, + { + "epoch": 3.89, + "learning_rate": 5.7923229445854715e-06, + "loss": 0.878, + "step": 52705 + }, + { + "epoch": 3.89, + "learning_rate": 5.791961104196369e-06, + "loss": 0.8293, + "step": 52706 + }, + { + "epoch": 3.89, + "learning_rate": 5.791599270502196e-06, + "loss": 1.0028, + "step": 52707 + }, + { + "epoch": 3.89, + "learning_rate": 5.791237443503526e-06, + "loss": 1.0839, + "step": 52708 + }, + { + "epoch": 3.89, + "learning_rate": 5.790875623200939e-06, + "loss": 0.8933, + "step": 52709 + }, + { + "epoch": 3.89, + "learning_rate": 5.790513809595009e-06, + "loss": 0.9733, + "step": 52710 + }, + { + "epoch": 3.89, + "learning_rate": 5.790152002686312e-06, + "loss": 0.9301, + "step": 52711 + }, + { + "epoch": 3.89, + "learning_rate": 5.789790202475419e-06, + "loss": 0.879, + "step": 52712 + }, + { + "epoch": 3.89, + "learning_rate": 5.7894284089629135e-06, + "loss": 0.9862, + "step": 52713 + }, + { + "epoch": 3.9, + "learning_rate": 5.789066622149366e-06, + "loss": 0.986, + "step": 52714 + }, + { + "epoch": 3.9, + "learning_rate": 5.788704842035356e-06, + "loss": 1.0664, + "step": 52715 + }, + { + "epoch": 3.9, + "learning_rate": 5.7883430686214495e-06, + "loss": 1.0767, + "step": 52716 + }, + { + "epoch": 3.9, + "learning_rate": 5.787981301908235e-06, + "loss": 1.0297, + "step": 52717 + }, + { + "epoch": 3.9, + "learning_rate": 5.7876195418962844e-06, + "loss": 1.0131, + "step": 52718 + }, + { + "epoch": 3.9, + "learning_rate": 5.787257788586164e-06, + "loss": 1.0394, + "step": 52719 + }, + { + "epoch": 3.9, + "learning_rate": 5.786896041978459e-06, + "loss": 1.0015, + "step": 52720 + }, + { + "epoch": 3.9, + "learning_rate": 5.786534302073738e-06, + "loss": 1.1689, + "step": 52721 + }, + { + "epoch": 3.9, + "learning_rate": 5.78617256887259e-06, + "loss": 0.9241, + "step": 52722 + }, + { + "epoch": 3.9, + "learning_rate": 5.785810842375572e-06, + "loss": 0.9334, + "step": 52723 + }, + { + "epoch": 3.9, + "learning_rate": 5.7854491225832735e-06, + "loss": 1.0357, + "step": 52724 + }, + { + "epoch": 3.9, + "learning_rate": 5.785087409496262e-06, + "loss": 0.8712, + "step": 52725 + }, + { + "epoch": 3.9, + "learning_rate": 5.7847257031151215e-06, + "loss": 1.0244, + "step": 52726 + }, + { + "epoch": 3.9, + "learning_rate": 5.78436400344042e-06, + "loss": 0.9303, + "step": 52727 + }, + { + "epoch": 3.9, + "learning_rate": 5.784002310472732e-06, + "loss": 1.1422, + "step": 52728 + }, + { + "epoch": 3.9, + "learning_rate": 5.7836406242126374e-06, + "loss": 0.8964, + "step": 52729 + }, + { + "epoch": 3.9, + "learning_rate": 5.783278944660712e-06, + "loss": 0.9315, + "step": 52730 + }, + { + "epoch": 3.9, + "learning_rate": 5.782917271817529e-06, + "loss": 0.9267, + "step": 52731 + }, + { + "epoch": 3.9, + "learning_rate": 5.78255560568366e-06, + "loss": 0.8741, + "step": 52732 + }, + { + "epoch": 3.9, + "learning_rate": 5.782193946259689e-06, + "loss": 1.1136, + "step": 52733 + }, + { + "epoch": 3.9, + "learning_rate": 5.781832293546187e-06, + "loss": 0.9706, + "step": 52734 + }, + { + "epoch": 3.9, + "learning_rate": 5.7814706475437295e-06, + "loss": 0.9584, + "step": 52735 + }, + { + "epoch": 3.9, + "learning_rate": 5.781109008252892e-06, + "loss": 1.0322, + "step": 52736 + }, + { + "epoch": 3.9, + "learning_rate": 5.780747375674245e-06, + "loss": 1.0639, + "step": 52737 + }, + { + "epoch": 3.9, + "learning_rate": 5.780385749808372e-06, + "loss": 1.0375, + "step": 52738 + }, + { + "epoch": 3.9, + "learning_rate": 5.780024130655847e-06, + "loss": 0.9961, + "step": 52739 + }, + { + "epoch": 3.9, + "learning_rate": 5.779662518217242e-06, + "loss": 0.9844, + "step": 52740 + }, + { + "epoch": 3.9, + "learning_rate": 5.77930091249313e-06, + "loss": 1.0288, + "step": 52741 + }, + { + "epoch": 3.9, + "learning_rate": 5.778939313484098e-06, + "loss": 1.1096, + "step": 52742 + }, + { + "epoch": 3.9, + "learning_rate": 5.7785777211907035e-06, + "loss": 1.0246, + "step": 52743 + }, + { + "epoch": 3.9, + "learning_rate": 5.778216135613537e-06, + "loss": 1.0027, + "step": 52744 + }, + { + "epoch": 3.9, + "learning_rate": 5.777854556753168e-06, + "loss": 0.9514, + "step": 52745 + }, + { + "epoch": 3.9, + "learning_rate": 5.777492984610172e-06, + "loss": 0.9515, + "step": 52746 + }, + { + "epoch": 3.9, + "learning_rate": 5.777131419185125e-06, + "loss": 0.992, + "step": 52747 + }, + { + "epoch": 3.9, + "learning_rate": 5.776769860478596e-06, + "loss": 1.0458, + "step": 52748 + }, + { + "epoch": 3.9, + "learning_rate": 5.77640830849117e-06, + "loss": 0.9949, + "step": 52749 + }, + { + "epoch": 3.9, + "learning_rate": 5.77604676322342e-06, + "loss": 1.0184, + "step": 52750 + }, + { + "epoch": 3.9, + "learning_rate": 5.775685224675916e-06, + "loss": 1.0052, + "step": 52751 + }, + { + "epoch": 3.9, + "learning_rate": 5.775323692849235e-06, + "loss": 1.035, + "step": 52752 + }, + { + "epoch": 3.9, + "learning_rate": 5.774962167743956e-06, + "loss": 1.0468, + "step": 52753 + }, + { + "epoch": 3.9, + "learning_rate": 5.774600649360652e-06, + "loss": 0.9323, + "step": 52754 + }, + { + "epoch": 3.9, + "learning_rate": 5.774239137699897e-06, + "loss": 0.9985, + "step": 52755 + }, + { + "epoch": 3.9, + "learning_rate": 5.77387763276227e-06, + "loss": 1.0545, + "step": 52756 + }, + { + "epoch": 3.9, + "learning_rate": 5.773516134548336e-06, + "loss": 0.9645, + "step": 52757 + }, + { + "epoch": 3.9, + "learning_rate": 5.773154643058682e-06, + "loss": 1.0738, + "step": 52758 + }, + { + "epoch": 3.9, + "learning_rate": 5.7727931582938814e-06, + "loss": 1.2091, + "step": 52759 + }, + { + "epoch": 3.9, + "learning_rate": 5.7724316802545035e-06, + "loss": 1.1368, + "step": 52760 + }, + { + "epoch": 3.9, + "learning_rate": 5.772070208941123e-06, + "loss": 0.9434, + "step": 52761 + }, + { + "epoch": 3.9, + "learning_rate": 5.771708744354323e-06, + "loss": 1.0556, + "step": 52762 + }, + { + "epoch": 3.9, + "learning_rate": 5.771347286494677e-06, + "loss": 1.0214, + "step": 52763 + }, + { + "epoch": 3.9, + "learning_rate": 5.770985835362748e-06, + "loss": 0.9869, + "step": 52764 + }, + { + "epoch": 3.9, + "learning_rate": 5.770624390959125e-06, + "loss": 0.9664, + "step": 52765 + }, + { + "epoch": 3.9, + "learning_rate": 5.7702629532843755e-06, + "loss": 0.9364, + "step": 52766 + }, + { + "epoch": 3.9, + "learning_rate": 5.769901522339083e-06, + "loss": 1.0179, + "step": 52767 + }, + { + "epoch": 3.9, + "learning_rate": 5.76954009812381e-06, + "loss": 0.8931, + "step": 52768 + }, + { + "epoch": 3.9, + "learning_rate": 5.769178680639143e-06, + "loss": 1.0097, + "step": 52769 + }, + { + "epoch": 3.9, + "learning_rate": 5.768817269885646e-06, + "loss": 0.9834, + "step": 52770 + }, + { + "epoch": 3.9, + "learning_rate": 5.768455865863909e-06, + "loss": 0.9224, + "step": 52771 + }, + { + "epoch": 3.9, + "learning_rate": 5.76809446857449e-06, + "loss": 1.0194, + "step": 52772 + }, + { + "epoch": 3.9, + "learning_rate": 5.767733078017976e-06, + "loss": 1.0632, + "step": 52773 + }, + { + "epoch": 3.9, + "learning_rate": 5.767371694194939e-06, + "loss": 0.9718, + "step": 52774 + }, + { + "epoch": 3.9, + "learning_rate": 5.767010317105953e-06, + "loss": 0.9353, + "step": 52775 + }, + { + "epoch": 3.9, + "learning_rate": 5.766648946751593e-06, + "loss": 0.9653, + "step": 52776 + }, + { + "epoch": 3.9, + "learning_rate": 5.766287583132429e-06, + "loss": 1.0609, + "step": 52777 + }, + { + "epoch": 3.9, + "learning_rate": 5.765926226249046e-06, + "loss": 0.9802, + "step": 52778 + }, + { + "epoch": 3.9, + "learning_rate": 5.765564876102014e-06, + "loss": 0.9472, + "step": 52779 + }, + { + "epoch": 3.9, + "learning_rate": 5.7652035326919065e-06, + "loss": 1.0117, + "step": 52780 + }, + { + "epoch": 3.9, + "learning_rate": 5.764842196019296e-06, + "loss": 0.9759, + "step": 52781 + }, + { + "epoch": 3.9, + "learning_rate": 5.764480866084765e-06, + "loss": 1.1223, + "step": 52782 + }, + { + "epoch": 3.9, + "learning_rate": 5.764119542888884e-06, + "loss": 1.0304, + "step": 52783 + }, + { + "epoch": 3.9, + "learning_rate": 5.763758226432229e-06, + "loss": 0.9638, + "step": 52784 + }, + { + "epoch": 3.9, + "learning_rate": 5.7633969167153735e-06, + "loss": 0.9757, + "step": 52785 + }, + { + "epoch": 3.9, + "learning_rate": 5.76303561373889e-06, + "loss": 0.9681, + "step": 52786 + }, + { + "epoch": 3.9, + "learning_rate": 5.762674317503363e-06, + "loss": 0.9655, + "step": 52787 + }, + { + "epoch": 3.9, + "learning_rate": 5.762313028009352e-06, + "loss": 0.9697, + "step": 52788 + }, + { + "epoch": 3.9, + "learning_rate": 5.761951745257446e-06, + "loss": 0.996, + "step": 52789 + }, + { + "epoch": 3.9, + "learning_rate": 5.761590469248209e-06, + "loss": 1.026, + "step": 52790 + }, + { + "epoch": 3.9, + "learning_rate": 5.761229199982229e-06, + "loss": 1.034, + "step": 52791 + }, + { + "epoch": 3.9, + "learning_rate": 5.760867937460069e-06, + "loss": 1.0277, + "step": 52792 + }, + { + "epoch": 3.9, + "learning_rate": 5.760506681682303e-06, + "loss": 0.9879, + "step": 52793 + }, + { + "epoch": 3.9, + "learning_rate": 5.760145432649516e-06, + "loss": 0.9925, + "step": 52794 + }, + { + "epoch": 3.9, + "learning_rate": 5.759784190362274e-06, + "loss": 0.9635, + "step": 52795 + }, + { + "epoch": 3.9, + "learning_rate": 5.759422954821156e-06, + "loss": 1.1001, + "step": 52796 + }, + { + "epoch": 3.9, + "learning_rate": 5.759061726026732e-06, + "loss": 0.9857, + "step": 52797 + }, + { + "epoch": 3.9, + "learning_rate": 5.758700503979584e-06, + "loss": 1.0593, + "step": 52798 + }, + { + "epoch": 3.9, + "learning_rate": 5.7583392886802824e-06, + "loss": 1.0712, + "step": 52799 + }, + { + "epoch": 3.9, + "learning_rate": 5.757978080129403e-06, + "loss": 1.0924, + "step": 52800 + }, + { + "epoch": 3.9, + "learning_rate": 5.757616878327519e-06, + "loss": 1.0482, + "step": 52801 + }, + { + "epoch": 3.9, + "learning_rate": 5.757255683275202e-06, + "loss": 0.9782, + "step": 52802 + }, + { + "epoch": 3.9, + "learning_rate": 5.756894494973035e-06, + "loss": 1.041, + "step": 52803 + }, + { + "epoch": 3.9, + "learning_rate": 5.756533313421588e-06, + "loss": 1.0512, + "step": 52804 + }, + { + "epoch": 3.9, + "learning_rate": 5.756172138621435e-06, + "loss": 0.9869, + "step": 52805 + }, + { + "epoch": 3.9, + "learning_rate": 5.755810970573148e-06, + "loss": 0.9976, + "step": 52806 + }, + { + "epoch": 3.9, + "learning_rate": 5.755449809277309e-06, + "loss": 0.9419, + "step": 52807 + }, + { + "epoch": 3.9, + "learning_rate": 5.755088654734488e-06, + "loss": 1.0078, + "step": 52808 + }, + { + "epoch": 3.9, + "learning_rate": 5.7547275069452614e-06, + "loss": 1.0547, + "step": 52809 + }, + { + "epoch": 3.9, + "learning_rate": 5.754366365910202e-06, + "loss": 0.9841, + "step": 52810 + }, + { + "epoch": 3.9, + "learning_rate": 5.75400523162988e-06, + "loss": 0.9931, + "step": 52811 + }, + { + "epoch": 3.9, + "learning_rate": 5.753644104104885e-06, + "loss": 0.9522, + "step": 52812 + }, + { + "epoch": 3.9, + "learning_rate": 5.753282983335772e-06, + "loss": 1.0869, + "step": 52813 + }, + { + "epoch": 3.9, + "learning_rate": 5.7529218693231296e-06, + "loss": 0.9682, + "step": 52814 + }, + { + "epoch": 3.9, + "learning_rate": 5.7525607620675225e-06, + "loss": 1.1637, + "step": 52815 + }, + { + "epoch": 3.9, + "learning_rate": 5.75219966156954e-06, + "loss": 0.9203, + "step": 52816 + }, + { + "epoch": 3.9, + "learning_rate": 5.751838567829739e-06, + "loss": 0.9334, + "step": 52817 + }, + { + "epoch": 3.9, + "learning_rate": 5.751477480848705e-06, + "loss": 1.0663, + "step": 52818 + }, + { + "epoch": 3.9, + "learning_rate": 5.75111640062701e-06, + "loss": 0.9516, + "step": 52819 + }, + { + "epoch": 3.9, + "learning_rate": 5.7507553271652285e-06, + "loss": 1.0001, + "step": 52820 + }, + { + "epoch": 3.9, + "learning_rate": 5.750394260463934e-06, + "loss": 0.9678, + "step": 52821 + }, + { + "epoch": 3.9, + "learning_rate": 5.750033200523697e-06, + "loss": 0.9616, + "step": 52822 + }, + { + "epoch": 3.9, + "learning_rate": 5.7496721473451025e-06, + "loss": 1.0558, + "step": 52823 + }, + { + "epoch": 3.9, + "learning_rate": 5.749311100928716e-06, + "loss": 0.9356, + "step": 52824 + }, + { + "epoch": 3.9, + "learning_rate": 5.748950061275116e-06, + "loss": 0.8851, + "step": 52825 + }, + { + "epoch": 3.9, + "learning_rate": 5.748589028384872e-06, + "loss": 1.0669, + "step": 52826 + }, + { + "epoch": 3.9, + "learning_rate": 5.7482280022585644e-06, + "loss": 1.0227, + "step": 52827 + }, + { + "epoch": 3.9, + "learning_rate": 5.7478669828967685e-06, + "loss": 1.1107, + "step": 52828 + }, + { + "epoch": 3.9, + "learning_rate": 5.747505970300051e-06, + "loss": 1.0279, + "step": 52829 + }, + { + "epoch": 3.9, + "learning_rate": 5.747144964468993e-06, + "loss": 1.0171, + "step": 52830 + }, + { + "epoch": 3.9, + "learning_rate": 5.746783965404163e-06, + "loss": 0.9729, + "step": 52831 + }, + { + "epoch": 3.9, + "learning_rate": 5.746422973106148e-06, + "loss": 0.977, + "step": 52832 + }, + { + "epoch": 3.9, + "learning_rate": 5.746061987575502e-06, + "loss": 0.9552, + "step": 52833 + }, + { + "epoch": 3.9, + "learning_rate": 5.745701008812816e-06, + "loss": 0.9753, + "step": 52834 + }, + { + "epoch": 3.9, + "learning_rate": 5.7453400368186555e-06, + "loss": 1.0249, + "step": 52835 + }, + { + "epoch": 3.9, + "learning_rate": 5.744979071593606e-06, + "loss": 0.9414, + "step": 52836 + }, + { + "epoch": 3.9, + "learning_rate": 5.744618113138225e-06, + "loss": 0.8887, + "step": 52837 + }, + { + "epoch": 3.9, + "learning_rate": 5.744257161453099e-06, + "loss": 0.9317, + "step": 52838 + }, + { + "epoch": 3.9, + "learning_rate": 5.7438962165388e-06, + "loss": 0.9235, + "step": 52839 + }, + { + "epoch": 3.9, + "learning_rate": 5.743535278395902e-06, + "loss": 1.0393, + "step": 52840 + }, + { + "epoch": 3.9, + "learning_rate": 5.743174347024978e-06, + "loss": 1.0151, + "step": 52841 + }, + { + "epoch": 3.9, + "learning_rate": 5.742813422426599e-06, + "loss": 1.0411, + "step": 52842 + }, + { + "epoch": 3.9, + "learning_rate": 5.742452504601346e-06, + "loss": 1.0587, + "step": 52843 + }, + { + "epoch": 3.9, + "learning_rate": 5.74209159354979e-06, + "loss": 0.9518, + "step": 52844 + }, + { + "epoch": 3.9, + "learning_rate": 5.741730689272507e-06, + "loss": 1.0301, + "step": 52845 + }, + { + "epoch": 3.9, + "learning_rate": 5.741369791770064e-06, + "loss": 0.9379, + "step": 52846 + }, + { + "epoch": 3.9, + "learning_rate": 5.741008901043046e-06, + "loss": 1.03, + "step": 52847 + }, + { + "epoch": 3.9, + "learning_rate": 5.7406480170920205e-06, + "loss": 1.0493, + "step": 52848 + }, + { + "epoch": 3.9, + "learning_rate": 5.7402871399175645e-06, + "loss": 1.014, + "step": 52849 + }, + { + "epoch": 3.91, + "learning_rate": 5.739926269520249e-06, + "loss": 1.0282, + "step": 52850 + }, + { + "epoch": 3.91, + "learning_rate": 5.739565405900647e-06, + "loss": 0.9814, + "step": 52851 + }, + { + "epoch": 3.91, + "learning_rate": 5.739204549059338e-06, + "loss": 1.0531, + "step": 52852 + }, + { + "epoch": 3.91, + "learning_rate": 5.738843698996895e-06, + "loss": 1.0016, + "step": 52853 + }, + { + "epoch": 3.91, + "learning_rate": 5.738482855713892e-06, + "loss": 1.048, + "step": 52854 + }, + { + "epoch": 3.91, + "learning_rate": 5.738122019210896e-06, + "loss": 1.011, + "step": 52855 + }, + { + "epoch": 3.91, + "learning_rate": 5.737761189488491e-06, + "loss": 0.9521, + "step": 52856 + }, + { + "epoch": 3.91, + "learning_rate": 5.73740036654725e-06, + "loss": 1.0101, + "step": 52857 + }, + { + "epoch": 3.91, + "learning_rate": 5.737039550387738e-06, + "loss": 1.0509, + "step": 52858 + }, + { + "epoch": 3.91, + "learning_rate": 5.736678741010538e-06, + "loss": 0.9346, + "step": 52859 + }, + { + "epoch": 3.91, + "learning_rate": 5.7363179384162185e-06, + "loss": 1.0344, + "step": 52860 + }, + { + "epoch": 3.91, + "learning_rate": 5.735957142605362e-06, + "loss": 0.9566, + "step": 52861 + }, + { + "epoch": 3.91, + "learning_rate": 5.73559635357853e-06, + "loss": 1.0188, + "step": 52862 + }, + { + "epoch": 3.91, + "learning_rate": 5.735235571336306e-06, + "loss": 0.9579, + "step": 52863 + }, + { + "epoch": 3.91, + "learning_rate": 5.7348747958792635e-06, + "loss": 0.9115, + "step": 52864 + }, + { + "epoch": 3.91, + "learning_rate": 5.734514027207972e-06, + "loss": 0.9042, + "step": 52865 + }, + { + "epoch": 3.91, + "learning_rate": 5.734153265323008e-06, + "loss": 0.9971, + "step": 52866 + }, + { + "epoch": 3.91, + "learning_rate": 5.7337925102249425e-06, + "loss": 1.0459, + "step": 52867 + }, + { + "epoch": 3.91, + "learning_rate": 5.7334317619143555e-06, + "loss": 1.0246, + "step": 52868 + }, + { + "epoch": 3.91, + "learning_rate": 5.733071020391816e-06, + "loss": 0.9249, + "step": 52869 + }, + { + "epoch": 3.91, + "learning_rate": 5.732710285657901e-06, + "loss": 1.0497, + "step": 52870 + }, + { + "epoch": 3.91, + "learning_rate": 5.732349557713179e-06, + "loss": 1.0362, + "step": 52871 + }, + { + "epoch": 3.91, + "learning_rate": 5.731988836558231e-06, + "loss": 1.1245, + "step": 52872 + }, + { + "epoch": 3.91, + "learning_rate": 5.731628122193629e-06, + "loss": 0.9404, + "step": 52873 + }, + { + "epoch": 3.91, + "learning_rate": 5.731267414619943e-06, + "loss": 0.9985, + "step": 52874 + }, + { + "epoch": 3.91, + "learning_rate": 5.7309067138377475e-06, + "loss": 1.0544, + "step": 52875 + }, + { + "epoch": 3.91, + "learning_rate": 5.730546019847622e-06, + "loss": 0.9523, + "step": 52876 + }, + { + "epoch": 3.91, + "learning_rate": 5.730185332650141e-06, + "loss": 1.0289, + "step": 52877 + }, + { + "epoch": 3.91, + "learning_rate": 5.7298246522458655e-06, + "loss": 0.9206, + "step": 52878 + }, + { + "epoch": 3.91, + "learning_rate": 5.729463978635383e-06, + "loss": 0.9426, + "step": 52879 + }, + { + "epoch": 3.91, + "learning_rate": 5.729103311819256e-06, + "loss": 0.9764, + "step": 52880 + }, + { + "epoch": 3.91, + "learning_rate": 5.728742651798074e-06, + "loss": 0.9719, + "step": 52881 + }, + { + "epoch": 3.91, + "learning_rate": 5.728381998572393e-06, + "loss": 0.9007, + "step": 52882 + }, + { + "epoch": 3.91, + "learning_rate": 5.728021352142798e-06, + "loss": 1.0325, + "step": 52883 + }, + { + "epoch": 3.91, + "learning_rate": 5.727660712509858e-06, + "loss": 1.014, + "step": 52884 + }, + { + "epoch": 3.91, + "learning_rate": 5.727300079674156e-06, + "loss": 0.917, + "step": 52885 + }, + { + "epoch": 3.91, + "learning_rate": 5.726939453636255e-06, + "loss": 1.0353, + "step": 52886 + }, + { + "epoch": 3.91, + "learning_rate": 5.726578834396728e-06, + "loss": 1.0118, + "step": 52887 + }, + { + "epoch": 3.91, + "learning_rate": 5.726218221956155e-06, + "loss": 1.0365, + "step": 52888 + }, + { + "epoch": 3.91, + "learning_rate": 5.725857616315109e-06, + "loss": 1.0766, + "step": 52889 + }, + { + "epoch": 3.91, + "learning_rate": 5.7254970174741616e-06, + "loss": 1.0953, + "step": 52890 + }, + { + "epoch": 3.91, + "learning_rate": 5.725136425433886e-06, + "loss": 1.0686, + "step": 52891 + }, + { + "epoch": 3.91, + "learning_rate": 5.7247758401948585e-06, + "loss": 1.0091, + "step": 52892 + }, + { + "epoch": 3.91, + "learning_rate": 5.724415261757652e-06, + "loss": 1.0139, + "step": 52893 + }, + { + "epoch": 3.91, + "learning_rate": 5.72405469012284e-06, + "loss": 0.934, + "step": 52894 + }, + { + "epoch": 3.91, + "learning_rate": 5.723694125290995e-06, + "loss": 1.0297, + "step": 52895 + }, + { + "epoch": 3.91, + "learning_rate": 5.723333567262689e-06, + "loss": 1.0157, + "step": 52896 + }, + { + "epoch": 3.91, + "learning_rate": 5.7229730160385015e-06, + "loss": 0.9512, + "step": 52897 + }, + { + "epoch": 3.91, + "learning_rate": 5.7226124716190025e-06, + "loss": 0.9055, + "step": 52898 + }, + { + "epoch": 3.91, + "learning_rate": 5.722251934004765e-06, + "loss": 1.0832, + "step": 52899 + }, + { + "epoch": 3.91, + "learning_rate": 5.7218914031963605e-06, + "loss": 1.0528, + "step": 52900 + }, + { + "epoch": 3.91, + "learning_rate": 5.7215308791943704e-06, + "loss": 0.9985, + "step": 52901 + }, + { + "epoch": 3.91, + "learning_rate": 5.7211703619993616e-06, + "loss": 1.0544, + "step": 52902 + }, + { + "epoch": 3.91, + "learning_rate": 5.720809851611911e-06, + "loss": 0.8942, + "step": 52903 + }, + { + "epoch": 3.91, + "learning_rate": 5.72044934803259e-06, + "loss": 0.9198, + "step": 52904 + }, + { + "epoch": 3.91, + "learning_rate": 5.720088851261969e-06, + "loss": 1.0054, + "step": 52905 + }, + { + "epoch": 3.91, + "learning_rate": 5.719728361300633e-06, + "loss": 0.9164, + "step": 52906 + }, + { + "epoch": 3.91, + "learning_rate": 5.71936787814914e-06, + "loss": 1.0095, + "step": 52907 + }, + { + "epoch": 3.91, + "learning_rate": 5.719007401808077e-06, + "loss": 1.0803, + "step": 52908 + }, + { + "epoch": 3.91, + "learning_rate": 5.718646932278009e-06, + "loss": 0.9502, + "step": 52909 + }, + { + "epoch": 3.91, + "learning_rate": 5.718286469559515e-06, + "loss": 1.0509, + "step": 52910 + }, + { + "epoch": 3.91, + "learning_rate": 5.717926013653161e-06, + "loss": 1.0245, + "step": 52911 + }, + { + "epoch": 3.91, + "learning_rate": 5.71756556455953e-06, + "loss": 0.948, + "step": 52912 + }, + { + "epoch": 3.91, + "learning_rate": 5.71720512227919e-06, + "loss": 1.0056, + "step": 52913 + }, + { + "epoch": 3.91, + "learning_rate": 5.716844686812716e-06, + "loss": 1.0314, + "step": 52914 + }, + { + "epoch": 3.91, + "learning_rate": 5.71648425816068e-06, + "loss": 0.95, + "step": 52915 + }, + { + "epoch": 3.91, + "learning_rate": 5.716123836323654e-06, + "loss": 1.0246, + "step": 52916 + }, + { + "epoch": 3.91, + "learning_rate": 5.715763421302216e-06, + "loss": 1.055, + "step": 52917 + }, + { + "epoch": 3.91, + "learning_rate": 5.715403013096939e-06, + "loss": 0.9917, + "step": 52918 + }, + { + "epoch": 3.91, + "learning_rate": 5.715042611708392e-06, + "loss": 1.0639, + "step": 52919 + }, + { + "epoch": 3.91, + "learning_rate": 5.714682217137149e-06, + "loss": 1.0057, + "step": 52920 + }, + { + "epoch": 3.91, + "learning_rate": 5.714321829383788e-06, + "loss": 0.8952, + "step": 52921 + }, + { + "epoch": 3.91, + "learning_rate": 5.713961448448884e-06, + "loss": 1.0534, + "step": 52922 + }, + { + "epoch": 3.91, + "learning_rate": 5.713601074332998e-06, + "loss": 1.016, + "step": 52923 + }, + { + "epoch": 3.91, + "learning_rate": 5.713240707036716e-06, + "loss": 0.9976, + "step": 52924 + }, + { + "epoch": 3.91, + "learning_rate": 5.712880346560602e-06, + "loss": 1.0474, + "step": 52925 + }, + { + "epoch": 3.91, + "learning_rate": 5.712519992905242e-06, + "loss": 1.0472, + "step": 52926 + }, + { + "epoch": 3.91, + "learning_rate": 5.712159646071194e-06, + "loss": 0.9789, + "step": 52927 + }, + { + "epoch": 3.91, + "learning_rate": 5.7117993060590415e-06, + "loss": 1.0874, + "step": 52928 + }, + { + "epoch": 3.91, + "learning_rate": 5.711438972869352e-06, + "loss": 0.9769, + "step": 52929 + }, + { + "epoch": 3.91, + "learning_rate": 5.711078646502709e-06, + "loss": 1.1126, + "step": 52930 + }, + { + "epoch": 3.91, + "learning_rate": 5.710718326959671e-06, + "loss": 0.9345, + "step": 52931 + }, + { + "epoch": 3.91, + "learning_rate": 5.710358014240822e-06, + "loss": 0.9761, + "step": 52932 + }, + { + "epoch": 3.91, + "learning_rate": 5.709997708346733e-06, + "loss": 1.0612, + "step": 52933 + }, + { + "epoch": 3.91, + "learning_rate": 5.709637409277978e-06, + "loss": 1.0318, + "step": 52934 + }, + { + "epoch": 3.91, + "learning_rate": 5.709277117035124e-06, + "loss": 1.0113, + "step": 52935 + }, + { + "epoch": 3.91, + "learning_rate": 5.7089168316187485e-06, + "loss": 1.0279, + "step": 52936 + }, + { + "epoch": 3.91, + "learning_rate": 5.708556553029428e-06, + "loss": 0.9672, + "step": 52937 + }, + { + "epoch": 3.91, + "learning_rate": 5.70819628126773e-06, + "loss": 0.972, + "step": 52938 + }, + { + "epoch": 3.91, + "learning_rate": 5.707836016334234e-06, + "loss": 0.9712, + "step": 52939 + }, + { + "epoch": 3.91, + "learning_rate": 5.707475758229504e-06, + "loss": 1.0726, + "step": 52940 + }, + { + "epoch": 3.91, + "learning_rate": 5.707115506954122e-06, + "loss": 0.9808, + "step": 52941 + }, + { + "epoch": 3.91, + "learning_rate": 5.706755262508658e-06, + "loss": 1.1029, + "step": 52942 + }, + { + "epoch": 3.91, + "learning_rate": 5.7063950248936865e-06, + "loss": 1.0423, + "step": 52943 + }, + { + "epoch": 3.91, + "learning_rate": 5.706034794109778e-06, + "loss": 0.8872, + "step": 52944 + }, + { + "epoch": 3.91, + "learning_rate": 5.7056745701575025e-06, + "loss": 0.9138, + "step": 52945 + }, + { + "epoch": 3.91, + "learning_rate": 5.705314353037441e-06, + "loss": 0.9056, + "step": 52946 + }, + { + "epoch": 3.91, + "learning_rate": 5.704954142750164e-06, + "loss": 0.9965, + "step": 52947 + }, + { + "epoch": 3.91, + "learning_rate": 5.7045939392962435e-06, + "loss": 0.9953, + "step": 52948 + }, + { + "epoch": 3.91, + "learning_rate": 5.7042337426762476e-06, + "loss": 0.9349, + "step": 52949 + }, + { + "epoch": 3.91, + "learning_rate": 5.703873552890759e-06, + "loss": 1.0546, + "step": 52950 + }, + { + "epoch": 3.91, + "learning_rate": 5.70351336994035e-06, + "loss": 1.0031, + "step": 52951 + }, + { + "epoch": 3.91, + "learning_rate": 5.703153193825582e-06, + "loss": 1.1111, + "step": 52952 + }, + { + "epoch": 3.91, + "learning_rate": 5.702793024547038e-06, + "loss": 0.8929, + "step": 52953 + }, + { + "epoch": 3.91, + "learning_rate": 5.702432862105292e-06, + "loss": 1.095, + "step": 52954 + }, + { + "epoch": 3.91, + "learning_rate": 5.7020727065009115e-06, + "loss": 0.9445, + "step": 52955 + }, + { + "epoch": 3.91, + "learning_rate": 5.701712557734468e-06, + "loss": 0.9784, + "step": 52956 + }, + { + "epoch": 3.91, + "learning_rate": 5.701352415806542e-06, + "loss": 0.991, + "step": 52957 + }, + { + "epoch": 3.91, + "learning_rate": 5.700992280717705e-06, + "loss": 0.973, + "step": 52958 + }, + { + "epoch": 3.91, + "learning_rate": 5.700632152468526e-06, + "loss": 0.9515, + "step": 52959 + }, + { + "epoch": 3.91, + "learning_rate": 5.70027203105958e-06, + "loss": 1.0611, + "step": 52960 + }, + { + "epoch": 3.91, + "learning_rate": 5.699911916491435e-06, + "loss": 1.0386, + "step": 52961 + }, + { + "epoch": 3.91, + "learning_rate": 5.699551808764674e-06, + "loss": 0.9062, + "step": 52962 + }, + { + "epoch": 3.91, + "learning_rate": 5.699191707879863e-06, + "loss": 0.9206, + "step": 52963 + }, + { + "epoch": 3.91, + "learning_rate": 5.698831613837579e-06, + "loss": 1.0401, + "step": 52964 + }, + { + "epoch": 3.91, + "learning_rate": 5.698471526638385e-06, + "loss": 1.0466, + "step": 52965 + }, + { + "epoch": 3.91, + "learning_rate": 5.698111446282868e-06, + "loss": 0.961, + "step": 52966 + }, + { + "epoch": 3.91, + "learning_rate": 5.697751372771593e-06, + "loss": 0.9833, + "step": 52967 + }, + { + "epoch": 3.91, + "learning_rate": 5.697391306105133e-06, + "loss": 1.0105, + "step": 52968 + }, + { + "epoch": 3.91, + "learning_rate": 5.697031246284063e-06, + "loss": 0.936, + "step": 52969 + }, + { + "epoch": 3.91, + "learning_rate": 5.696671193308951e-06, + "loss": 0.8667, + "step": 52970 + }, + { + "epoch": 3.91, + "learning_rate": 5.696311147180382e-06, + "loss": 1.1733, + "step": 52971 + }, + { + "epoch": 3.91, + "learning_rate": 5.695951107898912e-06, + "loss": 0.9381, + "step": 52972 + }, + { + "epoch": 3.91, + "learning_rate": 5.695591075465125e-06, + "loss": 1.0351, + "step": 52973 + }, + { + "epoch": 3.91, + "learning_rate": 5.695231049879588e-06, + "loss": 0.9791, + "step": 52974 + }, + { + "epoch": 3.91, + "learning_rate": 5.694871031142886e-06, + "loss": 0.9673, + "step": 52975 + }, + { + "epoch": 3.91, + "learning_rate": 5.694511019255573e-06, + "loss": 1.0818, + "step": 52976 + }, + { + "epoch": 3.91, + "learning_rate": 5.694151014218235e-06, + "loss": 0.9889, + "step": 52977 + }, + { + "epoch": 3.91, + "learning_rate": 5.693791016031443e-06, + "loss": 0.9349, + "step": 52978 + }, + { + "epoch": 3.91, + "learning_rate": 5.693431024695766e-06, + "loss": 1.0404, + "step": 52979 + }, + { + "epoch": 3.91, + "learning_rate": 5.693071040211778e-06, + "loss": 1.0315, + "step": 52980 + }, + { + "epoch": 3.91, + "learning_rate": 5.69271106258005e-06, + "loss": 1.0261, + "step": 52981 + }, + { + "epoch": 3.91, + "learning_rate": 5.692351091801161e-06, + "loss": 0.9448, + "step": 52982 + }, + { + "epoch": 3.91, + "learning_rate": 5.69199112787568e-06, + "loss": 0.9928, + "step": 52983 + }, + { + "epoch": 3.91, + "learning_rate": 5.691631170804178e-06, + "loss": 0.9752, + "step": 52984 + }, + { + "epoch": 3.92, + "learning_rate": 5.691271220587227e-06, + "loss": 1.0572, + "step": 52985 + }, + { + "epoch": 3.92, + "learning_rate": 5.690911277225404e-06, + "loss": 1.0416, + "step": 52986 + }, + { + "epoch": 3.92, + "learning_rate": 5.690551340719281e-06, + "loss": 0.9392, + "step": 52987 + }, + { + "epoch": 3.92, + "learning_rate": 5.6901914110694275e-06, + "loss": 0.9236, + "step": 52988 + }, + { + "epoch": 3.92, + "learning_rate": 5.689831488276421e-06, + "loss": 0.8578, + "step": 52989 + }, + { + "epoch": 3.92, + "learning_rate": 5.689471572340823e-06, + "loss": 0.9145, + "step": 52990 + }, + { + "epoch": 3.92, + "learning_rate": 5.6891116632632205e-06, + "loss": 1.0498, + "step": 52991 + }, + { + "epoch": 3.92, + "learning_rate": 5.68875176104418e-06, + "loss": 1.0434, + "step": 52992 + }, + { + "epoch": 3.92, + "learning_rate": 5.688391865684273e-06, + "loss": 0.9525, + "step": 52993 + }, + { + "epoch": 3.92, + "learning_rate": 5.688031977184068e-06, + "loss": 0.9893, + "step": 52994 + }, + { + "epoch": 3.92, + "learning_rate": 5.687672095544147e-06, + "loss": 0.9805, + "step": 52995 + }, + { + "epoch": 3.92, + "learning_rate": 5.687312220765078e-06, + "loss": 0.9921, + "step": 52996 + }, + { + "epoch": 3.92, + "learning_rate": 5.686952352847434e-06, + "loss": 0.9683, + "step": 52997 + }, + { + "epoch": 3.92, + "learning_rate": 5.686592491791787e-06, + "loss": 1.0013, + "step": 52998 + }, + { + "epoch": 3.92, + "learning_rate": 5.6862326375987096e-06, + "loss": 1.0573, + "step": 52999 + }, + { + "epoch": 3.92, + "learning_rate": 5.6858727902687735e-06, + "loss": 0.931, + "step": 53000 + }, + { + "epoch": 3.92, + "learning_rate": 5.68551294980255e-06, + "loss": 1.0726, + "step": 53001 + }, + { + "epoch": 3.92, + "learning_rate": 5.685153116200615e-06, + "loss": 1.0508, + "step": 53002 + }, + { + "epoch": 3.92, + "learning_rate": 5.684793289463541e-06, + "loss": 1.0345, + "step": 53003 + }, + { + "epoch": 3.92, + "learning_rate": 5.684433469591899e-06, + "loss": 0.8794, + "step": 53004 + }, + { + "epoch": 3.92, + "learning_rate": 5.684073656586258e-06, + "loss": 1.0065, + "step": 53005 + }, + { + "epoch": 3.92, + "learning_rate": 5.683713850447199e-06, + "loss": 1.033, + "step": 53006 + }, + { + "epoch": 3.92, + "learning_rate": 5.6833540511752875e-06, + "loss": 0.9927, + "step": 53007 + }, + { + "epoch": 3.92, + "learning_rate": 5.6829942587711e-06, + "loss": 1.0265, + "step": 53008 + }, + { + "epoch": 3.92, + "learning_rate": 5.6826344732352046e-06, + "loss": 0.9018, + "step": 53009 + }, + { + "epoch": 3.92, + "learning_rate": 5.6822746945681724e-06, + "loss": 1.0377, + "step": 53010 + }, + { + "epoch": 3.92, + "learning_rate": 5.681914922770585e-06, + "loss": 0.9172, + "step": 53011 + }, + { + "epoch": 3.92, + "learning_rate": 5.6815551578430085e-06, + "loss": 0.9874, + "step": 53012 + }, + { + "epoch": 3.92, + "learning_rate": 5.681195399786015e-06, + "loss": 0.9958, + "step": 53013 + }, + { + "epoch": 3.92, + "learning_rate": 5.680835648600175e-06, + "loss": 1.0132, + "step": 53014 + }, + { + "epoch": 3.92, + "learning_rate": 5.680475904286067e-06, + "loss": 1.0199, + "step": 53015 + }, + { + "epoch": 3.92, + "learning_rate": 5.680116166844264e-06, + "loss": 1.055, + "step": 53016 + }, + { + "epoch": 3.92, + "learning_rate": 5.6797564362753265e-06, + "loss": 0.9091, + "step": 53017 + }, + { + "epoch": 3.92, + "learning_rate": 5.679396712579839e-06, + "loss": 1.0474, + "step": 53018 + }, + { + "epoch": 3.92, + "learning_rate": 5.6790369957583645e-06, + "loss": 0.9858, + "step": 53019 + }, + { + "epoch": 3.92, + "learning_rate": 5.678677285811489e-06, + "loss": 1.0023, + "step": 53020 + }, + { + "epoch": 3.92, + "learning_rate": 5.67831758273977e-06, + "loss": 1.1022, + "step": 53021 + }, + { + "epoch": 3.92, + "learning_rate": 5.677957886543788e-06, + "loss": 1.1598, + "step": 53022 + }, + { + "epoch": 3.92, + "learning_rate": 5.677598197224108e-06, + "loss": 0.9237, + "step": 53023 + }, + { + "epoch": 3.92, + "learning_rate": 5.677238514781318e-06, + "loss": 0.9701, + "step": 53024 + }, + { + "epoch": 3.92, + "learning_rate": 5.676878839215974e-06, + "loss": 1.0721, + "step": 53025 + }, + { + "epoch": 3.92, + "learning_rate": 5.676519170528649e-06, + "loss": 1.0159, + "step": 53026 + }, + { + "epoch": 3.92, + "learning_rate": 5.676159508719927e-06, + "loss": 1.0266, + "step": 53027 + }, + { + "epoch": 3.92, + "learning_rate": 5.675799853790372e-06, + "loss": 1.0035, + "step": 53028 + }, + { + "epoch": 3.92, + "learning_rate": 5.675440205740557e-06, + "loss": 0.8741, + "step": 53029 + }, + { + "epoch": 3.92, + "learning_rate": 5.6750805645710504e-06, + "loss": 1.0323, + "step": 53030 + }, + { + "epoch": 3.92, + "learning_rate": 5.674720930282433e-06, + "loss": 1.0252, + "step": 53031 + }, + { + "epoch": 3.92, + "learning_rate": 5.6743613028752746e-06, + "loss": 0.9592, + "step": 53032 + }, + { + "epoch": 3.92, + "learning_rate": 5.674001682350143e-06, + "loss": 1.0233, + "step": 53033 + }, + { + "epoch": 3.92, + "learning_rate": 5.673642068707615e-06, + "loss": 0.8868, + "step": 53034 + }, + { + "epoch": 3.92, + "learning_rate": 5.673282461948255e-06, + "loss": 0.9386, + "step": 53035 + }, + { + "epoch": 3.92, + "learning_rate": 5.672922862072645e-06, + "loss": 1.0395, + "step": 53036 + }, + { + "epoch": 3.92, + "learning_rate": 5.672563269081353e-06, + "loss": 0.9657, + "step": 53037 + }, + { + "epoch": 3.92, + "learning_rate": 5.67220368297495e-06, + "loss": 1.0314, + "step": 53038 + }, + { + "epoch": 3.92, + "learning_rate": 5.671844103754008e-06, + "loss": 0.9387, + "step": 53039 + }, + { + "epoch": 3.92, + "learning_rate": 5.671484531419101e-06, + "loss": 1.095, + "step": 53040 + }, + { + "epoch": 3.92, + "learning_rate": 5.671124965970801e-06, + "loss": 0.9798, + "step": 53041 + }, + { + "epoch": 3.92, + "learning_rate": 5.67076540740968e-06, + "loss": 0.9862, + "step": 53042 + }, + { + "epoch": 3.92, + "learning_rate": 5.670405855736304e-06, + "loss": 0.9768, + "step": 53043 + }, + { + "epoch": 3.92, + "learning_rate": 5.670046310951259e-06, + "loss": 1.0569, + "step": 53044 + }, + { + "epoch": 3.92, + "learning_rate": 5.669686773055105e-06, + "loss": 1.0029, + "step": 53045 + }, + { + "epoch": 3.92, + "learning_rate": 5.669327242048413e-06, + "loss": 0.9474, + "step": 53046 + }, + { + "epoch": 3.92, + "learning_rate": 5.6689677179317625e-06, + "loss": 1.0921, + "step": 53047 + }, + { + "epoch": 3.92, + "learning_rate": 5.668608200705723e-06, + "loss": 0.9676, + "step": 53048 + }, + { + "epoch": 3.92, + "learning_rate": 5.668248690370865e-06, + "loss": 0.9837, + "step": 53049 + }, + { + "epoch": 3.92, + "learning_rate": 5.667889186927759e-06, + "loss": 0.9473, + "step": 53050 + }, + { + "epoch": 3.92, + "learning_rate": 5.6675296903769825e-06, + "loss": 1.1012, + "step": 53051 + }, + { + "epoch": 3.92, + "learning_rate": 5.667170200719106e-06, + "loss": 0.9592, + "step": 53052 + }, + { + "epoch": 3.92, + "learning_rate": 5.666810717954698e-06, + "loss": 0.9487, + "step": 53053 + }, + { + "epoch": 3.92, + "learning_rate": 5.666451242084332e-06, + "loss": 0.8764, + "step": 53054 + }, + { + "epoch": 3.92, + "learning_rate": 5.6660917731085765e-06, + "loss": 0.9904, + "step": 53055 + }, + { + "epoch": 3.92, + "learning_rate": 5.66573231102801e-06, + "loss": 1.0628, + "step": 53056 + }, + { + "epoch": 3.92, + "learning_rate": 5.665372855843203e-06, + "loss": 1.0076, + "step": 53057 + }, + { + "epoch": 3.92, + "learning_rate": 5.6650134075547244e-06, + "loss": 0.965, + "step": 53058 + }, + { + "epoch": 3.92, + "learning_rate": 5.664653966163145e-06, + "loss": 0.907, + "step": 53059 + }, + { + "epoch": 3.92, + "learning_rate": 5.6642945316690416e-06, + "loss": 0.9113, + "step": 53060 + }, + { + "epoch": 3.92, + "learning_rate": 5.663935104072984e-06, + "loss": 1.0193, + "step": 53061 + }, + { + "epoch": 3.92, + "learning_rate": 5.663575683375544e-06, + "loss": 0.9837, + "step": 53062 + }, + { + "epoch": 3.92, + "learning_rate": 5.663216269577293e-06, + "loss": 0.9687, + "step": 53063 + }, + { + "epoch": 3.92, + "learning_rate": 5.662856862678799e-06, + "loss": 1.038, + "step": 53064 + }, + { + "epoch": 3.92, + "learning_rate": 5.662497462680645e-06, + "loss": 0.9938, + "step": 53065 + }, + { + "epoch": 3.92, + "learning_rate": 5.6621380695833875e-06, + "loss": 0.9622, + "step": 53066 + }, + { + "epoch": 3.92, + "learning_rate": 5.661778683387611e-06, + "loss": 0.9716, + "step": 53067 + }, + { + "epoch": 3.92, + "learning_rate": 5.661419304093878e-06, + "loss": 1.0036, + "step": 53068 + }, + { + "epoch": 3.92, + "learning_rate": 5.6610599317027726e-06, + "loss": 1.0277, + "step": 53069 + }, + { + "epoch": 3.92, + "learning_rate": 5.660700566214852e-06, + "loss": 0.9803, + "step": 53070 + }, + { + "epoch": 3.92, + "learning_rate": 5.660341207630697e-06, + "loss": 1.0686, + "step": 53071 + }, + { + "epoch": 3.92, + "learning_rate": 5.659981855950878e-06, + "loss": 1.0328, + "step": 53072 + }, + { + "epoch": 3.92, + "learning_rate": 5.659622511175966e-06, + "loss": 0.9672, + "step": 53073 + }, + { + "epoch": 3.92, + "learning_rate": 5.659263173306531e-06, + "loss": 0.9255, + "step": 53074 + }, + { + "epoch": 3.92, + "learning_rate": 5.658903842343144e-06, + "loss": 1.0856, + "step": 53075 + }, + { + "epoch": 3.92, + "learning_rate": 5.658544518286382e-06, + "loss": 1.0878, + "step": 53076 + }, + { + "epoch": 3.92, + "learning_rate": 5.658185201136815e-06, + "loss": 0.9891, + "step": 53077 + }, + { + "epoch": 3.92, + "learning_rate": 5.657825890895011e-06, + "loss": 0.8695, + "step": 53078 + }, + { + "epoch": 3.92, + "learning_rate": 5.657466587561541e-06, + "loss": 0.9183, + "step": 53079 + }, + { + "epoch": 3.92, + "learning_rate": 5.657107291136983e-06, + "loss": 0.9856, + "step": 53080 + }, + { + "epoch": 3.92, + "learning_rate": 5.6567480016219055e-06, + "loss": 0.9719, + "step": 53081 + }, + { + "epoch": 3.92, + "learning_rate": 5.6563887190168805e-06, + "loss": 1.0095, + "step": 53082 + }, + { + "epoch": 3.92, + "learning_rate": 5.656029443322477e-06, + "loss": 0.9807, + "step": 53083 + }, + { + "epoch": 3.92, + "learning_rate": 5.655670174539267e-06, + "loss": 0.9823, + "step": 53084 + }, + { + "epoch": 3.92, + "learning_rate": 5.655310912667831e-06, + "loss": 0.9746, + "step": 53085 + }, + { + "epoch": 3.92, + "learning_rate": 5.654951657708723e-06, + "loss": 1.0755, + "step": 53086 + }, + { + "epoch": 3.92, + "learning_rate": 5.654592409662531e-06, + "loss": 0.984, + "step": 53087 + }, + { + "epoch": 3.92, + "learning_rate": 5.654233168529815e-06, + "loss": 0.9953, + "step": 53088 + }, + { + "epoch": 3.92, + "learning_rate": 5.65387393431116e-06, + "loss": 1.0286, + "step": 53089 + }, + { + "epoch": 3.92, + "learning_rate": 5.653514707007121e-06, + "loss": 0.9748, + "step": 53090 + }, + { + "epoch": 3.92, + "learning_rate": 5.653155486618285e-06, + "loss": 1.0112, + "step": 53091 + }, + { + "epoch": 3.92, + "learning_rate": 5.6527962731452135e-06, + "loss": 1.0339, + "step": 53092 + }, + { + "epoch": 3.92, + "learning_rate": 5.652437066588481e-06, + "loss": 1.0014, + "step": 53093 + }, + { + "epoch": 3.92, + "learning_rate": 5.6520778669486596e-06, + "loss": 1.026, + "step": 53094 + }, + { + "epoch": 3.92, + "learning_rate": 5.651718674226316e-06, + "loss": 1.0382, + "step": 53095 + }, + { + "epoch": 3.92, + "learning_rate": 5.65135948842203e-06, + "loss": 1.0671, + "step": 53096 + }, + { + "epoch": 3.92, + "learning_rate": 5.651000309536368e-06, + "loss": 0.9906, + "step": 53097 + }, + { + "epoch": 3.92, + "learning_rate": 5.6506411375699034e-06, + "loss": 1.0928, + "step": 53098 + }, + { + "epoch": 3.92, + "learning_rate": 5.650281972523201e-06, + "loss": 0.9343, + "step": 53099 + }, + { + "epoch": 3.92, + "learning_rate": 5.649922814396844e-06, + "loss": 0.9325, + "step": 53100 + }, + { + "epoch": 3.92, + "learning_rate": 5.6495636631913965e-06, + "loss": 1.0655, + "step": 53101 + }, + { + "epoch": 3.92, + "learning_rate": 5.64920451890743e-06, + "loss": 1.0305, + "step": 53102 + }, + { + "epoch": 3.92, + "learning_rate": 5.648845381545518e-06, + "loss": 0.9439, + "step": 53103 + }, + { + "epoch": 3.92, + "learning_rate": 5.648486251106227e-06, + "loss": 0.8746, + "step": 53104 + }, + { + "epoch": 3.92, + "learning_rate": 5.648127127590135e-06, + "loss": 0.9606, + "step": 53105 + }, + { + "epoch": 3.92, + "learning_rate": 5.647768010997812e-06, + "loss": 0.9419, + "step": 53106 + }, + { + "epoch": 3.92, + "learning_rate": 5.6474089013298265e-06, + "loss": 1.0153, + "step": 53107 + }, + { + "epoch": 3.92, + "learning_rate": 5.647049798586746e-06, + "loss": 1.0285, + "step": 53108 + }, + { + "epoch": 3.92, + "learning_rate": 5.646690702769153e-06, + "loss": 1.0008, + "step": 53109 + }, + { + "epoch": 3.92, + "learning_rate": 5.646331613877616e-06, + "loss": 0.9472, + "step": 53110 + }, + { + "epoch": 3.92, + "learning_rate": 5.645972531912696e-06, + "loss": 1.0317, + "step": 53111 + }, + { + "epoch": 3.92, + "learning_rate": 5.645613456874974e-06, + "loss": 1.0248, + "step": 53112 + }, + { + "epoch": 3.92, + "learning_rate": 5.645254388765016e-06, + "loss": 0.9733, + "step": 53113 + }, + { + "epoch": 3.92, + "learning_rate": 5.644895327583404e-06, + "loss": 0.9595, + "step": 53114 + }, + { + "epoch": 3.92, + "learning_rate": 5.6445362733306915e-06, + "loss": 0.9861, + "step": 53115 + }, + { + "epoch": 3.92, + "learning_rate": 5.644177226007466e-06, + "loss": 0.9516, + "step": 53116 + }, + { + "epoch": 3.92, + "learning_rate": 5.643818185614286e-06, + "loss": 0.919, + "step": 53117 + }, + { + "epoch": 3.92, + "learning_rate": 5.643459152151738e-06, + "loss": 1.004, + "step": 53118 + }, + { + "epoch": 3.92, + "learning_rate": 5.64310012562038e-06, + "loss": 0.8323, + "step": 53119 + }, + { + "epoch": 3.93, + "learning_rate": 5.642741106020784e-06, + "loss": 1.0556, + "step": 53120 + }, + { + "epoch": 3.93, + "learning_rate": 5.642382093353527e-06, + "loss": 1.0018, + "step": 53121 + }, + { + "epoch": 3.93, + "learning_rate": 5.642023087619179e-06, + "loss": 0.9436, + "step": 53122 + }, + { + "epoch": 3.93, + "learning_rate": 5.641664088818309e-06, + "loss": 1.0758, + "step": 53123 + }, + { + "epoch": 3.93, + "learning_rate": 5.641305096951486e-06, + "loss": 1.0293, + "step": 53124 + }, + { + "epoch": 3.93, + "learning_rate": 5.640946112019288e-06, + "loss": 0.9969, + "step": 53125 + }, + { + "epoch": 3.93, + "learning_rate": 5.640587134022282e-06, + "loss": 1.1137, + "step": 53126 + }, + { + "epoch": 3.93, + "learning_rate": 5.640228162961039e-06, + "loss": 1.1008, + "step": 53127 + }, + { + "epoch": 3.93, + "learning_rate": 5.639869198836132e-06, + "loss": 1.0792, + "step": 53128 + }, + { + "epoch": 3.93, + "learning_rate": 5.639510241648125e-06, + "loss": 1.0153, + "step": 53129 + }, + { + "epoch": 3.93, + "learning_rate": 5.639151291397605e-06, + "loss": 0.9478, + "step": 53130 + }, + { + "epoch": 3.93, + "learning_rate": 5.638792348085123e-06, + "loss": 0.9488, + "step": 53131 + }, + { + "epoch": 3.93, + "learning_rate": 5.638433411711265e-06, + "loss": 1.0983, + "step": 53132 + }, + { + "epoch": 3.93, + "learning_rate": 5.638074482276592e-06, + "loss": 0.955, + "step": 53133 + }, + { + "epoch": 3.93, + "learning_rate": 5.637715559781689e-06, + "loss": 1.0367, + "step": 53134 + }, + { + "epoch": 3.93, + "learning_rate": 5.63735664422711e-06, + "loss": 0.8883, + "step": 53135 + }, + { + "epoch": 3.93, + "learning_rate": 5.6369977356134385e-06, + "loss": 0.9862, + "step": 53136 + }, + { + "epoch": 3.93, + "learning_rate": 5.636638833941242e-06, + "loss": 1.0251, + "step": 53137 + }, + { + "epoch": 3.93, + "learning_rate": 5.636279939211089e-06, + "loss": 0.9153, + "step": 53138 + }, + { + "epoch": 3.93, + "learning_rate": 5.635921051423552e-06, + "loss": 0.938, + "step": 53139 + }, + { + "epoch": 3.93, + "learning_rate": 5.635562170579198e-06, + "loss": 1.0238, + "step": 53140 + }, + { + "epoch": 3.93, + "learning_rate": 5.635203296678609e-06, + "loss": 1.1825, + "step": 53141 + }, + { + "epoch": 3.93, + "learning_rate": 5.634844429722346e-06, + "loss": 1.0013, + "step": 53142 + }, + { + "epoch": 3.93, + "learning_rate": 5.634485569710985e-06, + "loss": 1.0398, + "step": 53143 + }, + { + "epoch": 3.93, + "learning_rate": 5.63412671664509e-06, + "loss": 1.0833, + "step": 53144 + }, + { + "epoch": 3.93, + "learning_rate": 5.633767870525241e-06, + "loss": 0.9099, + "step": 53145 + }, + { + "epoch": 3.93, + "learning_rate": 5.633409031352004e-06, + "loss": 0.9933, + "step": 53146 + }, + { + "epoch": 3.93, + "learning_rate": 5.633050199125952e-06, + "loss": 1.0994, + "step": 53147 + }, + { + "epoch": 3.93, + "learning_rate": 5.632691373847654e-06, + "loss": 0.9284, + "step": 53148 + }, + { + "epoch": 3.93, + "learning_rate": 5.6323325555176785e-06, + "loss": 1.074, + "step": 53149 + }, + { + "epoch": 3.93, + "learning_rate": 5.631973744136603e-06, + "loss": 0.9124, + "step": 53150 + }, + { + "epoch": 3.93, + "learning_rate": 5.6316149397049945e-06, + "loss": 1.0613, + "step": 53151 + }, + { + "epoch": 3.93, + "learning_rate": 5.6312561422234245e-06, + "loss": 0.9987, + "step": 53152 + }, + { + "epoch": 3.93, + "learning_rate": 5.6308973516924596e-06, + "loss": 1.041, + "step": 53153 + }, + { + "epoch": 3.93, + "learning_rate": 5.630538568112678e-06, + "loss": 0.9677, + "step": 53154 + }, + { + "epoch": 3.93, + "learning_rate": 5.630179791484647e-06, + "loss": 0.966, + "step": 53155 + }, + { + "epoch": 3.93, + "learning_rate": 5.6298210218089366e-06, + "loss": 1.0223, + "step": 53156 + }, + { + "epoch": 3.93, + "learning_rate": 5.629462259086119e-06, + "loss": 1.158, + "step": 53157 + }, + { + "epoch": 3.93, + "learning_rate": 5.629103503316762e-06, + "loss": 1.0034, + "step": 53158 + }, + { + "epoch": 3.93, + "learning_rate": 5.628744754501445e-06, + "loss": 0.9434, + "step": 53159 + }, + { + "epoch": 3.93, + "learning_rate": 5.628386012640725e-06, + "loss": 1.0066, + "step": 53160 + }, + { + "epoch": 3.93, + "learning_rate": 5.628027277735184e-06, + "loss": 0.9793, + "step": 53161 + }, + { + "epoch": 3.93, + "learning_rate": 5.62766854978539e-06, + "loss": 1.1357, + "step": 53162 + }, + { + "epoch": 3.93, + "learning_rate": 5.627309828791911e-06, + "loss": 1.0195, + "step": 53163 + }, + { + "epoch": 3.93, + "learning_rate": 5.626951114755318e-06, + "loss": 1.0565, + "step": 53164 + }, + { + "epoch": 3.93, + "learning_rate": 5.626592407676186e-06, + "loss": 0.8691, + "step": 53165 + }, + { + "epoch": 3.93, + "learning_rate": 5.626233707555082e-06, + "loss": 0.8709, + "step": 53166 + }, + { + "epoch": 3.93, + "learning_rate": 5.625875014392579e-06, + "loss": 1.1058, + "step": 53167 + }, + { + "epoch": 3.93, + "learning_rate": 5.6255163281892455e-06, + "loss": 1.0538, + "step": 53168 + }, + { + "epoch": 3.93, + "learning_rate": 5.625157648945649e-06, + "loss": 0.9953, + "step": 53169 + }, + { + "epoch": 3.93, + "learning_rate": 5.624798976662369e-06, + "loss": 0.9103, + "step": 53170 + }, + { + "epoch": 3.93, + "learning_rate": 5.6244403113399695e-06, + "loss": 1.066, + "step": 53171 + }, + { + "epoch": 3.93, + "learning_rate": 5.624081652979024e-06, + "loss": 1.0788, + "step": 53172 + }, + { + "epoch": 3.93, + "learning_rate": 5.623723001580098e-06, + "loss": 1.1191, + "step": 53173 + }, + { + "epoch": 3.93, + "learning_rate": 5.62336435714377e-06, + "loss": 1.0063, + "step": 53174 + }, + { + "epoch": 3.93, + "learning_rate": 5.6230057196706115e-06, + "loss": 1.0384, + "step": 53175 + }, + { + "epoch": 3.93, + "learning_rate": 5.622647089161179e-06, + "loss": 1.0275, + "step": 53176 + }, + { + "epoch": 3.93, + "learning_rate": 5.622288465616056e-06, + "loss": 0.9329, + "step": 53177 + }, + { + "epoch": 3.93, + "learning_rate": 5.621929849035806e-06, + "loss": 1.0217, + "step": 53178 + }, + { + "epoch": 3.93, + "learning_rate": 5.621571239421011e-06, + "loss": 1.0015, + "step": 53179 + }, + { + "epoch": 3.93, + "learning_rate": 5.6212126367722266e-06, + "loss": 1.0701, + "step": 53180 + }, + { + "epoch": 3.93, + "learning_rate": 5.620854041090032e-06, + "loss": 1.0681, + "step": 53181 + }, + { + "epoch": 3.93, + "learning_rate": 5.620495452374994e-06, + "loss": 1.0204, + "step": 53182 + }, + { + "epoch": 3.93, + "learning_rate": 5.620136870627693e-06, + "loss": 0.9899, + "step": 53183 + }, + { + "epoch": 3.93, + "learning_rate": 5.619778295848686e-06, + "loss": 0.9805, + "step": 53184 + }, + { + "epoch": 3.93, + "learning_rate": 5.619419728038546e-06, + "loss": 0.8409, + "step": 53185 + }, + { + "epoch": 3.93, + "learning_rate": 5.61906116719785e-06, + "loss": 1.0769, + "step": 53186 + }, + { + "epoch": 3.93, + "learning_rate": 5.618702613327166e-06, + "loss": 0.8995, + "step": 53187 + }, + { + "epoch": 3.93, + "learning_rate": 5.618344066427063e-06, + "loss": 0.9476, + "step": 53188 + }, + { + "epoch": 3.93, + "learning_rate": 5.6179855264981075e-06, + "loss": 1.0138, + "step": 53189 + }, + { + "epoch": 3.93, + "learning_rate": 5.617626993540879e-06, + "loss": 1.0442, + "step": 53190 + }, + { + "epoch": 3.93, + "learning_rate": 5.617268467555943e-06, + "loss": 1.0286, + "step": 53191 + }, + { + "epoch": 3.93, + "learning_rate": 5.616909948543869e-06, + "loss": 1.0057, + "step": 53192 + }, + { + "epoch": 3.93, + "learning_rate": 5.616551436505231e-06, + "loss": 1.0578, + "step": 53193 + }, + { + "epoch": 3.93, + "learning_rate": 5.616192931440591e-06, + "loss": 1.0498, + "step": 53194 + }, + { + "epoch": 3.93, + "learning_rate": 5.61583443335053e-06, + "loss": 0.9335, + "step": 53195 + }, + { + "epoch": 3.93, + "learning_rate": 5.615475942235613e-06, + "loss": 1.0683, + "step": 53196 + }, + { + "epoch": 3.93, + "learning_rate": 5.615117458096412e-06, + "loss": 1.0696, + "step": 53197 + }, + { + "epoch": 3.93, + "learning_rate": 5.614758980933492e-06, + "loss": 0.993, + "step": 53198 + }, + { + "epoch": 3.93, + "learning_rate": 5.614400510747432e-06, + "loss": 0.9577, + "step": 53199 + }, + { + "epoch": 3.93, + "learning_rate": 5.6140420475387965e-06, + "loss": 0.8835, + "step": 53200 + }, + { + "epoch": 3.93, + "learning_rate": 5.613683591308159e-06, + "loss": 1.0823, + "step": 53201 + }, + { + "epoch": 3.93, + "learning_rate": 5.613325142056084e-06, + "loss": 1.0383, + "step": 53202 + }, + { + "epoch": 3.93, + "learning_rate": 5.61296669978315e-06, + "loss": 0.9711, + "step": 53203 + }, + { + "epoch": 3.93, + "learning_rate": 5.612608264489927e-06, + "loss": 0.9518, + "step": 53204 + }, + { + "epoch": 3.93, + "learning_rate": 5.6122498361769725e-06, + "loss": 1.0185, + "step": 53205 + }, + { + "epoch": 3.93, + "learning_rate": 5.611891414844872e-06, + "loss": 0.9739, + "step": 53206 + }, + { + "epoch": 3.93, + "learning_rate": 5.611533000494187e-06, + "loss": 1.0204, + "step": 53207 + }, + { + "epoch": 3.93, + "learning_rate": 5.611174593125492e-06, + "loss": 1.0751, + "step": 53208 + }, + { + "epoch": 3.93, + "learning_rate": 5.6108161927393505e-06, + "loss": 0.952, + "step": 53209 + }, + { + "epoch": 3.93, + "learning_rate": 5.610457799336342e-06, + "loss": 1.015, + "step": 53210 + }, + { + "epoch": 3.93, + "learning_rate": 5.610099412917032e-06, + "loss": 1.1109, + "step": 53211 + }, + { + "epoch": 3.93, + "learning_rate": 5.609741033481991e-06, + "loss": 1.0776, + "step": 53212 + }, + { + "epoch": 3.93, + "learning_rate": 5.6093826610317905e-06, + "loss": 0.9924, + "step": 53213 + }, + { + "epoch": 3.93, + "learning_rate": 5.609024295566993e-06, + "loss": 0.9563, + "step": 53214 + }, + { + "epoch": 3.93, + "learning_rate": 5.608665937088181e-06, + "loss": 0.9322, + "step": 53215 + }, + { + "epoch": 3.93, + "learning_rate": 5.608307585595918e-06, + "loss": 0.9447, + "step": 53216 + }, + { + "epoch": 3.93, + "learning_rate": 5.607949241090775e-06, + "loss": 0.9862, + "step": 53217 + }, + { + "epoch": 3.93, + "learning_rate": 5.607590903573318e-06, + "loss": 0.9164, + "step": 53218 + }, + { + "epoch": 3.93, + "learning_rate": 5.607232573044124e-06, + "loss": 0.9791, + "step": 53219 + }, + { + "epoch": 3.93, + "learning_rate": 5.606874249503762e-06, + "loss": 1.0937, + "step": 53220 + }, + { + "epoch": 3.93, + "learning_rate": 5.606515932952799e-06, + "loss": 0.9421, + "step": 53221 + }, + { + "epoch": 3.93, + "learning_rate": 5.606157623391806e-06, + "loss": 0.8723, + "step": 53222 + }, + { + "epoch": 3.93, + "learning_rate": 5.60579932082135e-06, + "loss": 0.8997, + "step": 53223 + }, + { + "epoch": 3.93, + "learning_rate": 5.605441025242011e-06, + "loss": 0.9635, + "step": 53224 + }, + { + "epoch": 3.93, + "learning_rate": 5.605082736654346e-06, + "loss": 1.0174, + "step": 53225 + }, + { + "epoch": 3.93, + "learning_rate": 5.604724455058935e-06, + "loss": 0.933, + "step": 53226 + }, + { + "epoch": 3.93, + "learning_rate": 5.604366180456341e-06, + "loss": 0.9782, + "step": 53227 + }, + { + "epoch": 3.93, + "learning_rate": 5.604007912847145e-06, + "loss": 1.0907, + "step": 53228 + }, + { + "epoch": 3.93, + "learning_rate": 5.603649652231902e-06, + "loss": 1.0427, + "step": 53229 + }, + { + "epoch": 3.93, + "learning_rate": 5.6032913986111924e-06, + "loss": 0.9527, + "step": 53230 + }, + { + "epoch": 3.93, + "learning_rate": 5.602933151985584e-06, + "loss": 0.9968, + "step": 53231 + }, + { + "epoch": 3.93, + "learning_rate": 5.602574912355646e-06, + "loss": 1.0044, + "step": 53232 + }, + { + "epoch": 3.93, + "learning_rate": 5.602216679721948e-06, + "loss": 1.0591, + "step": 53233 + }, + { + "epoch": 3.93, + "learning_rate": 5.601858454085057e-06, + "loss": 1.0162, + "step": 53234 + }, + { + "epoch": 3.93, + "learning_rate": 5.601500235445551e-06, + "loss": 1.0045, + "step": 53235 + }, + { + "epoch": 3.93, + "learning_rate": 5.601142023803994e-06, + "loss": 0.9479, + "step": 53236 + }, + { + "epoch": 3.93, + "learning_rate": 5.600783819160958e-06, + "loss": 0.9689, + "step": 53237 + }, + { + "epoch": 3.93, + "learning_rate": 5.600425621517007e-06, + "loss": 1.0912, + "step": 53238 + }, + { + "epoch": 3.93, + "learning_rate": 5.60006743087272e-06, + "loss": 0.914, + "step": 53239 + }, + { + "epoch": 3.93, + "learning_rate": 5.599709247228662e-06, + "loss": 0.9589, + "step": 53240 + }, + { + "epoch": 3.93, + "learning_rate": 5.599351070585406e-06, + "loss": 0.9609, + "step": 53241 + }, + { + "epoch": 3.93, + "learning_rate": 5.598992900943517e-06, + "loss": 1.0326, + "step": 53242 + }, + { + "epoch": 3.93, + "learning_rate": 5.598634738303565e-06, + "loss": 1.0152, + "step": 53243 + }, + { + "epoch": 3.93, + "learning_rate": 5.5982765826661256e-06, + "loss": 0.9412, + "step": 53244 + }, + { + "epoch": 3.93, + "learning_rate": 5.5979184340317645e-06, + "loss": 0.9584, + "step": 53245 + }, + { + "epoch": 3.93, + "learning_rate": 5.597560292401053e-06, + "loss": 0.9967, + "step": 53246 + }, + { + "epoch": 3.93, + "learning_rate": 5.597202157774556e-06, + "loss": 0.9838, + "step": 53247 + }, + { + "epoch": 3.93, + "learning_rate": 5.5968440301528515e-06, + "loss": 1.0425, + "step": 53248 + }, + { + "epoch": 3.93, + "learning_rate": 5.596485909536509e-06, + "loss": 1.0641, + "step": 53249 + }, + { + "epoch": 3.93, + "learning_rate": 5.596127795926086e-06, + "loss": 1.0013, + "step": 53250 + }, + { + "epoch": 3.93, + "learning_rate": 5.595769689322166e-06, + "loss": 1.0114, + "step": 53251 + }, + { + "epoch": 3.93, + "learning_rate": 5.595411589725313e-06, + "loss": 1.0492, + "step": 53252 + }, + { + "epoch": 3.93, + "learning_rate": 5.5950534971360955e-06, + "loss": 1.0843, + "step": 53253 + }, + { + "epoch": 3.93, + "learning_rate": 5.594695411555082e-06, + "loss": 0.9981, + "step": 53254 + }, + { + "epoch": 3.93, + "learning_rate": 5.594337332982849e-06, + "loss": 0.9058, + "step": 53255 + }, + { + "epoch": 3.94, + "learning_rate": 5.593979261419962e-06, + "loss": 1.0376, + "step": 53256 + }, + { + "epoch": 3.94, + "learning_rate": 5.593621196866992e-06, + "loss": 0.9326, + "step": 53257 + }, + { + "epoch": 3.94, + "learning_rate": 5.593263139324503e-06, + "loss": 1.0, + "step": 53258 + }, + { + "epoch": 3.94, + "learning_rate": 5.592905088793073e-06, + "loss": 0.9563, + "step": 53259 + }, + { + "epoch": 3.94, + "learning_rate": 5.592547045273268e-06, + "loss": 0.8958, + "step": 53260 + }, + { + "epoch": 3.94, + "learning_rate": 5.592189008765658e-06, + "loss": 0.9536, + "step": 53261 + }, + { + "epoch": 3.94, + "learning_rate": 5.591830979270811e-06, + "loss": 0.9476, + "step": 53262 + }, + { + "epoch": 3.94, + "learning_rate": 5.591472956789294e-06, + "loss": 0.9933, + "step": 53263 + }, + { + "epoch": 3.94, + "learning_rate": 5.591114941321687e-06, + "loss": 1.0599, + "step": 53264 + }, + { + "epoch": 3.94, + "learning_rate": 5.590756932868551e-06, + "loss": 1.1317, + "step": 53265 + }, + { + "epoch": 3.94, + "learning_rate": 5.590398931430457e-06, + "loss": 1.0269, + "step": 53266 + }, + { + "epoch": 3.94, + "learning_rate": 5.590040937007972e-06, + "loss": 1.0009, + "step": 53267 + }, + { + "epoch": 3.94, + "learning_rate": 5.589682949601674e-06, + "loss": 0.9568, + "step": 53268 + }, + { + "epoch": 3.94, + "learning_rate": 5.5893249692121295e-06, + "loss": 0.9276, + "step": 53269 + }, + { + "epoch": 3.94, + "learning_rate": 5.588966995839899e-06, + "loss": 1.0429, + "step": 53270 + }, + { + "epoch": 3.94, + "learning_rate": 5.588609029485563e-06, + "loss": 1.1424, + "step": 53271 + }, + { + "epoch": 3.94, + "learning_rate": 5.588251070149682e-06, + "loss": 0.9576, + "step": 53272 + }, + { + "epoch": 3.94, + "learning_rate": 5.587893117832839e-06, + "loss": 1.0105, + "step": 53273 + }, + { + "epoch": 3.94, + "learning_rate": 5.587535172535586e-06, + "loss": 0.9836, + "step": 53274 + }, + { + "epoch": 3.94, + "learning_rate": 5.5871772342585075e-06, + "loss": 1.0373, + "step": 53275 + }, + { + "epoch": 3.94, + "learning_rate": 5.58681930300216e-06, + "loss": 0.9907, + "step": 53276 + }, + { + "epoch": 3.94, + "learning_rate": 5.58646137876713e-06, + "loss": 1.0607, + "step": 53277 + }, + { + "epoch": 3.94, + "learning_rate": 5.586103461553973e-06, + "loss": 1.0715, + "step": 53278 + }, + { + "epoch": 3.94, + "learning_rate": 5.585745551363257e-06, + "loss": 0.9727, + "step": 53279 + }, + { + "epoch": 3.94, + "learning_rate": 5.585387648195561e-06, + "loss": 0.9694, + "step": 53280 + }, + { + "epoch": 3.94, + "learning_rate": 5.58502975205145e-06, + "loss": 0.9972, + "step": 53281 + }, + { + "epoch": 3.94, + "learning_rate": 5.584671862931495e-06, + "loss": 0.9459, + "step": 53282 + }, + { + "epoch": 3.94, + "learning_rate": 5.584313980836258e-06, + "loss": 1.067, + "step": 53283 + }, + { + "epoch": 3.94, + "learning_rate": 5.583956105766318e-06, + "loss": 0.946, + "step": 53284 + }, + { + "epoch": 3.94, + "learning_rate": 5.583598237722241e-06, + "loss": 0.9968, + "step": 53285 + }, + { + "epoch": 3.94, + "learning_rate": 5.583240376704598e-06, + "loss": 1.0482, + "step": 53286 + }, + { + "epoch": 3.94, + "learning_rate": 5.582882522713954e-06, + "loss": 1.0227, + "step": 53287 + }, + { + "epoch": 3.94, + "learning_rate": 5.582524675750878e-06, + "loss": 1.0637, + "step": 53288 + }, + { + "epoch": 3.94, + "learning_rate": 5.582166835815945e-06, + "loss": 0.9656, + "step": 53289 + }, + { + "epoch": 3.94, + "learning_rate": 5.581809002909722e-06, + "loss": 0.9725, + "step": 53290 + }, + { + "epoch": 3.94, + "learning_rate": 5.581451177032777e-06, + "loss": 1.1431, + "step": 53291 + }, + { + "epoch": 3.94, + "learning_rate": 5.581093358185678e-06, + "loss": 0.9377, + "step": 53292 + }, + { + "epoch": 3.94, + "learning_rate": 5.580735546368999e-06, + "loss": 1.1095, + "step": 53293 + }, + { + "epoch": 3.94, + "learning_rate": 5.580377741583306e-06, + "loss": 1.0087, + "step": 53294 + }, + { + "epoch": 3.94, + "learning_rate": 5.580019943829169e-06, + "loss": 1.0269, + "step": 53295 + }, + { + "epoch": 3.94, + "learning_rate": 5.579662153107157e-06, + "loss": 0.9817, + "step": 53296 + }, + { + "epoch": 3.94, + "learning_rate": 5.57930436941784e-06, + "loss": 0.8882, + "step": 53297 + }, + { + "epoch": 3.94, + "learning_rate": 5.578946592761785e-06, + "loss": 1.0186, + "step": 53298 + }, + { + "epoch": 3.94, + "learning_rate": 5.578588823139561e-06, + "loss": 1.1083, + "step": 53299 + }, + { + "epoch": 3.94, + "learning_rate": 5.578231060551741e-06, + "loss": 1.0242, + "step": 53300 + }, + { + "epoch": 3.94, + "learning_rate": 5.577873304998893e-06, + "loss": 0.978, + "step": 53301 + }, + { + "epoch": 3.94, + "learning_rate": 5.577515556481585e-06, + "loss": 1.1025, + "step": 53302 + }, + { + "epoch": 3.94, + "learning_rate": 5.5771578150003826e-06, + "loss": 0.9038, + "step": 53303 + }, + { + "epoch": 3.94, + "learning_rate": 5.576800080555862e-06, + "loss": 1.027, + "step": 53304 + }, + { + "epoch": 3.94, + "learning_rate": 5.57644235314859e-06, + "loss": 0.9529, + "step": 53305 + }, + { + "epoch": 3.94, + "learning_rate": 5.576084632779135e-06, + "loss": 0.9925, + "step": 53306 + }, + { + "epoch": 3.94, + "learning_rate": 5.575726919448066e-06, + "loss": 0.9294, + "step": 53307 + }, + { + "epoch": 3.94, + "learning_rate": 5.575369213155948e-06, + "loss": 1.0684, + "step": 53308 + }, + { + "epoch": 3.94, + "learning_rate": 5.575011513903359e-06, + "loss": 1.0159, + "step": 53309 + }, + { + "epoch": 3.94, + "learning_rate": 5.5746538216908615e-06, + "loss": 1.0119, + "step": 53310 + }, + { + "epoch": 3.94, + "learning_rate": 5.5742961365190284e-06, + "loss": 0.9992, + "step": 53311 + }, + { + "epoch": 3.94, + "learning_rate": 5.573938458388422e-06, + "loss": 0.9865, + "step": 53312 + }, + { + "epoch": 3.94, + "learning_rate": 5.5735807872996194e-06, + "loss": 0.9943, + "step": 53313 + }, + { + "epoch": 3.94, + "learning_rate": 5.573223123253189e-06, + "loss": 1.0001, + "step": 53314 + }, + { + "epoch": 3.94, + "learning_rate": 5.572865466249695e-06, + "loss": 0.9925, + "step": 53315 + }, + { + "epoch": 3.94, + "learning_rate": 5.57250781628971e-06, + "loss": 1.0673, + "step": 53316 + }, + { + "epoch": 3.94, + "learning_rate": 5.572150173373796e-06, + "loss": 0.9178, + "step": 53317 + }, + { + "epoch": 3.94, + "learning_rate": 5.571792537502536e-06, + "loss": 1.1838, + "step": 53318 + }, + { + "epoch": 3.94, + "learning_rate": 5.571434908676484e-06, + "loss": 0.962, + "step": 53319 + }, + { + "epoch": 3.94, + "learning_rate": 5.571077286896218e-06, + "loss": 0.9238, + "step": 53320 + }, + { + "epoch": 3.94, + "learning_rate": 5.5707196721623035e-06, + "loss": 0.9812, + "step": 53321 + }, + { + "epoch": 3.94, + "learning_rate": 5.5703620644753164e-06, + "loss": 1.0307, + "step": 53322 + }, + { + "epoch": 3.94, + "learning_rate": 5.570004463835812e-06, + "loss": 0.8986, + "step": 53323 + }, + { + "epoch": 3.94, + "learning_rate": 5.5696468702443715e-06, + "loss": 1.0109, + "step": 53324 + }, + { + "epoch": 3.94, + "learning_rate": 5.56928928370156e-06, + "loss": 0.8589, + "step": 53325 + }, + { + "epoch": 3.94, + "learning_rate": 5.568931704207945e-06, + "loss": 1.1126, + "step": 53326 + }, + { + "epoch": 3.94, + "learning_rate": 5.568574131764097e-06, + "loss": 1.0373, + "step": 53327 + }, + { + "epoch": 3.94, + "learning_rate": 5.568216566370579e-06, + "loss": 0.9289, + "step": 53328 + }, + { + "epoch": 3.94, + "learning_rate": 5.567859008027971e-06, + "loss": 1.0319, + "step": 53329 + }, + { + "epoch": 3.94, + "learning_rate": 5.567501456736835e-06, + "loss": 1.0533, + "step": 53330 + }, + { + "epoch": 3.94, + "learning_rate": 5.567143912497739e-06, + "loss": 0.9956, + "step": 53331 + }, + { + "epoch": 3.94, + "learning_rate": 5.566786375311251e-06, + "loss": 0.9814, + "step": 53332 + }, + { + "epoch": 3.94, + "learning_rate": 5.5664288451779465e-06, + "loss": 1.0861, + "step": 53333 + }, + { + "epoch": 3.94, + "learning_rate": 5.5660713220983895e-06, + "loss": 1.0354, + "step": 53334 + }, + { + "epoch": 3.94, + "learning_rate": 5.565713806073152e-06, + "loss": 1.0614, + "step": 53335 + }, + { + "epoch": 3.94, + "learning_rate": 5.565356297102797e-06, + "loss": 1.0459, + "step": 53336 + }, + { + "epoch": 3.94, + "learning_rate": 5.564998795187896e-06, + "loss": 0.8723, + "step": 53337 + }, + { + "epoch": 3.94, + "learning_rate": 5.564641300329021e-06, + "loss": 1.0166, + "step": 53338 + }, + { + "epoch": 3.94, + "learning_rate": 5.564283812526739e-06, + "loss": 0.88, + "step": 53339 + }, + { + "epoch": 3.94, + "learning_rate": 5.563926331781615e-06, + "loss": 1.0063, + "step": 53340 + }, + { + "epoch": 3.94, + "learning_rate": 5.56356885809422e-06, + "loss": 0.8781, + "step": 53341 + }, + { + "epoch": 3.94, + "learning_rate": 5.563211391465129e-06, + "loss": 0.9215, + "step": 53342 + }, + { + "epoch": 3.94, + "learning_rate": 5.562853931894903e-06, + "loss": 1.0837, + "step": 53343 + }, + { + "epoch": 3.94, + "learning_rate": 5.562496479384109e-06, + "loss": 0.9037, + "step": 53344 + }, + { + "epoch": 3.94, + "learning_rate": 5.562139033933323e-06, + "loss": 1.0052, + "step": 53345 + }, + { + "epoch": 3.94, + "learning_rate": 5.56178159554311e-06, + "loss": 1.0405, + "step": 53346 + }, + { + "epoch": 3.94, + "learning_rate": 5.56142416421404e-06, + "loss": 1.0071, + "step": 53347 + }, + { + "epoch": 3.94, + "learning_rate": 5.561066739946676e-06, + "loss": 1.0492, + "step": 53348 + }, + { + "epoch": 3.94, + "learning_rate": 5.560709322741596e-06, + "loss": 0.9741, + "step": 53349 + }, + { + "epoch": 3.94, + "learning_rate": 5.560351912599362e-06, + "loss": 1.0662, + "step": 53350 + }, + { + "epoch": 3.94, + "learning_rate": 5.559994509520545e-06, + "loss": 0.9478, + "step": 53351 + }, + { + "epoch": 3.94, + "learning_rate": 5.559637113505715e-06, + "loss": 0.9842, + "step": 53352 + }, + { + "epoch": 3.94, + "learning_rate": 5.559279724555433e-06, + "loss": 0.9476, + "step": 53353 + }, + { + "epoch": 3.94, + "learning_rate": 5.5589223426702785e-06, + "loss": 0.9426, + "step": 53354 + }, + { + "epoch": 3.94, + "learning_rate": 5.558564967850815e-06, + "loss": 1.0361, + "step": 53355 + }, + { + "epoch": 3.94, + "learning_rate": 5.558207600097611e-06, + "loss": 1.0169, + "step": 53356 + }, + { + "epoch": 3.94, + "learning_rate": 5.557850239411231e-06, + "loss": 1.0898, + "step": 53357 + }, + { + "epoch": 3.94, + "learning_rate": 5.557492885792252e-06, + "loss": 1.1513, + "step": 53358 + }, + { + "epoch": 3.94, + "learning_rate": 5.557135539241238e-06, + "loss": 0.9008, + "step": 53359 + }, + { + "epoch": 3.94, + "learning_rate": 5.556778199758758e-06, + "loss": 1.0471, + "step": 53360 + }, + { + "epoch": 3.94, + "learning_rate": 5.556420867345375e-06, + "loss": 1.017, + "step": 53361 + }, + { + "epoch": 3.94, + "learning_rate": 5.556063542001669e-06, + "loss": 1.0188, + "step": 53362 + }, + { + "epoch": 3.94, + "learning_rate": 5.5557062237282055e-06, + "loss": 1.0801, + "step": 53363 + }, + { + "epoch": 3.94, + "learning_rate": 5.555348912525541e-06, + "loss": 1.0573, + "step": 53364 + }, + { + "epoch": 3.94, + "learning_rate": 5.554991608394258e-06, + "loss": 0.9635, + "step": 53365 + }, + { + "epoch": 3.94, + "learning_rate": 5.554634311334914e-06, + "loss": 1.0358, + "step": 53366 + }, + { + "epoch": 3.94, + "learning_rate": 5.554277021348093e-06, + "loss": 1.054, + "step": 53367 + }, + { + "epoch": 3.94, + "learning_rate": 5.553919738434345e-06, + "loss": 1.0169, + "step": 53368 + }, + { + "epoch": 3.94, + "learning_rate": 5.5535624625942504e-06, + "loss": 1.0965, + "step": 53369 + }, + { + "epoch": 3.94, + "learning_rate": 5.55320519382837e-06, + "loss": 1.0476, + "step": 53370 + }, + { + "epoch": 3.94, + "learning_rate": 5.5528479321372865e-06, + "loss": 1.0117, + "step": 53371 + }, + { + "epoch": 3.94, + "learning_rate": 5.5524906775215535e-06, + "loss": 1.0167, + "step": 53372 + }, + { + "epoch": 3.94, + "learning_rate": 5.55213342998174e-06, + "loss": 0.9192, + "step": 53373 + }, + { + "epoch": 3.94, + "learning_rate": 5.551776189518423e-06, + "loss": 0.8981, + "step": 53374 + }, + { + "epoch": 3.94, + "learning_rate": 5.5514189561321656e-06, + "loss": 1.0233, + "step": 53375 + }, + { + "epoch": 3.94, + "learning_rate": 5.551061729823537e-06, + "loss": 0.9802, + "step": 53376 + }, + { + "epoch": 3.94, + "learning_rate": 5.5507045105931015e-06, + "loss": 0.9536, + "step": 53377 + }, + { + "epoch": 3.94, + "learning_rate": 5.550347298441437e-06, + "loss": 0.9881, + "step": 53378 + }, + { + "epoch": 3.94, + "learning_rate": 5.5499900933691045e-06, + "loss": 0.9783, + "step": 53379 + }, + { + "epoch": 3.94, + "learning_rate": 5.549632895376675e-06, + "loss": 0.9706, + "step": 53380 + }, + { + "epoch": 3.94, + "learning_rate": 5.5492757044647165e-06, + "loss": 0.9724, + "step": 53381 + }, + { + "epoch": 3.94, + "learning_rate": 5.5489185206337905e-06, + "loss": 1.0112, + "step": 53382 + }, + { + "epoch": 3.94, + "learning_rate": 5.5485613438844775e-06, + "loss": 0.9612, + "step": 53383 + }, + { + "epoch": 3.94, + "learning_rate": 5.548204174217339e-06, + "loss": 1.0552, + "step": 53384 + }, + { + "epoch": 3.94, + "learning_rate": 5.547847011632943e-06, + "loss": 0.9886, + "step": 53385 + }, + { + "epoch": 3.94, + "learning_rate": 5.547489856131855e-06, + "loss": 1.0361, + "step": 53386 + }, + { + "epoch": 3.94, + "learning_rate": 5.547132707714656e-06, + "loss": 0.8641, + "step": 53387 + }, + { + "epoch": 3.94, + "learning_rate": 5.546775566381896e-06, + "loss": 1.0064, + "step": 53388 + }, + { + "epoch": 3.94, + "learning_rate": 5.546418432134159e-06, + "loss": 1.0349, + "step": 53389 + }, + { + "epoch": 3.94, + "learning_rate": 5.546061304972004e-06, + "loss": 1.0623, + "step": 53390 + }, + { + "epoch": 3.95, + "learning_rate": 5.5457041848960025e-06, + "loss": 1.1066, + "step": 53391 + }, + { + "epoch": 3.95, + "learning_rate": 5.545347071906721e-06, + "loss": 1.0727, + "step": 53392 + }, + { + "epoch": 3.95, + "learning_rate": 5.544989966004726e-06, + "loss": 0.9992, + "step": 53393 + }, + { + "epoch": 3.95, + "learning_rate": 5.544632867190591e-06, + "loss": 0.952, + "step": 53394 + }, + { + "epoch": 3.95, + "learning_rate": 5.5442757754648825e-06, + "loss": 1.0013, + "step": 53395 + }, + { + "epoch": 3.95, + "learning_rate": 5.543918690828167e-06, + "loss": 1.0265, + "step": 53396 + }, + { + "epoch": 3.95, + "learning_rate": 5.543561613281008e-06, + "loss": 0.952, + "step": 53397 + }, + { + "epoch": 3.95, + "learning_rate": 5.5432045428239855e-06, + "loss": 0.9747, + "step": 53398 + }, + { + "epoch": 3.95, + "learning_rate": 5.542847479457658e-06, + "loss": 1.0623, + "step": 53399 + }, + { + "epoch": 3.95, + "learning_rate": 5.542490423182598e-06, + "loss": 1.0895, + "step": 53400 + }, + { + "epoch": 3.95, + "learning_rate": 5.542133373999371e-06, + "loss": 0.9163, + "step": 53401 + }, + { + "epoch": 3.95, + "learning_rate": 5.541776331908543e-06, + "loss": 0.9332, + "step": 53402 + }, + { + "epoch": 3.95, + "learning_rate": 5.541419296910688e-06, + "loss": 1.0425, + "step": 53403 + }, + { + "epoch": 3.95, + "learning_rate": 5.541062269006372e-06, + "loss": 0.9993, + "step": 53404 + }, + { + "epoch": 3.95, + "learning_rate": 5.540705248196161e-06, + "loss": 1.0064, + "step": 53405 + }, + { + "epoch": 3.95, + "learning_rate": 5.540348234480621e-06, + "loss": 1.0755, + "step": 53406 + }, + { + "epoch": 3.95, + "learning_rate": 5.539991227860327e-06, + "loss": 1.0659, + "step": 53407 + }, + { + "epoch": 3.95, + "learning_rate": 5.539634228335848e-06, + "loss": 1.0757, + "step": 53408 + }, + { + "epoch": 3.95, + "learning_rate": 5.539277235907738e-06, + "loss": 0.9395, + "step": 53409 + }, + { + "epoch": 3.95, + "learning_rate": 5.538920250576578e-06, + "loss": 0.9074, + "step": 53410 + }, + { + "epoch": 3.95, + "learning_rate": 5.538563272342929e-06, + "loss": 0.908, + "step": 53411 + }, + { + "epoch": 3.95, + "learning_rate": 5.538206301207369e-06, + "loss": 0.9994, + "step": 53412 + }, + { + "epoch": 3.95, + "learning_rate": 5.537849337170451e-06, + "loss": 0.9598, + "step": 53413 + }, + { + "epoch": 3.95, + "learning_rate": 5.537492380232756e-06, + "loss": 0.9118, + "step": 53414 + }, + { + "epoch": 3.95, + "learning_rate": 5.537135430394842e-06, + "loss": 0.9871, + "step": 53415 + }, + { + "epoch": 3.95, + "learning_rate": 5.53677848765729e-06, + "loss": 0.9571, + "step": 53416 + }, + { + "epoch": 3.95, + "learning_rate": 5.536421552020651e-06, + "loss": 0.9862, + "step": 53417 + }, + { + "epoch": 3.95, + "learning_rate": 5.536064623485506e-06, + "loss": 0.977, + "step": 53418 + }, + { + "epoch": 3.95, + "learning_rate": 5.535707702052417e-06, + "loss": 0.9941, + "step": 53419 + }, + { + "epoch": 3.95, + "learning_rate": 5.535350787721955e-06, + "loss": 1.0449, + "step": 53420 + }, + { + "epoch": 3.95, + "learning_rate": 5.534993880494686e-06, + "loss": 1.1292, + "step": 53421 + }, + { + "epoch": 3.95, + "learning_rate": 5.534636980371172e-06, + "loss": 1.0096, + "step": 53422 + }, + { + "epoch": 3.95, + "learning_rate": 5.534280087351992e-06, + "loss": 0.9512, + "step": 53423 + }, + { + "epoch": 3.95, + "learning_rate": 5.533923201437708e-06, + "loss": 1.0654, + "step": 53424 + }, + { + "epoch": 3.95, + "learning_rate": 5.533566322628889e-06, + "loss": 0.9941, + "step": 53425 + }, + { + "epoch": 3.95, + "learning_rate": 5.533209450926098e-06, + "loss": 0.9501, + "step": 53426 + }, + { + "epoch": 3.95, + "learning_rate": 5.53285258632991e-06, + "loss": 0.9541, + "step": 53427 + }, + { + "epoch": 3.95, + "learning_rate": 5.532495728840894e-06, + "loss": 1.0096, + "step": 53428 + }, + { + "epoch": 3.95, + "learning_rate": 5.532138878459605e-06, + "loss": 1.0016, + "step": 53429 + }, + { + "epoch": 3.95, + "learning_rate": 5.531782035186624e-06, + "loss": 0.9931, + "step": 53430 + }, + { + "epoch": 3.95, + "learning_rate": 5.531425199022509e-06, + "loss": 1.0191, + "step": 53431 + }, + { + "epoch": 3.95, + "learning_rate": 5.531068369967841e-06, + "loss": 1.1228, + "step": 53432 + }, + { + "epoch": 3.95, + "learning_rate": 5.530711548023171e-06, + "loss": 0.9913, + "step": 53433 + }, + { + "epoch": 3.95, + "learning_rate": 5.53035473318908e-06, + "loss": 0.9752, + "step": 53434 + }, + { + "epoch": 3.95, + "learning_rate": 5.529997925466126e-06, + "loss": 0.9719, + "step": 53435 + }, + { + "epoch": 3.95, + "learning_rate": 5.52964112485489e-06, + "loss": 1.0578, + "step": 53436 + }, + { + "epoch": 3.95, + "learning_rate": 5.5292843313559255e-06, + "loss": 0.9679, + "step": 53437 + }, + { + "epoch": 3.95, + "learning_rate": 5.528927544969802e-06, + "loss": 1.0394, + "step": 53438 + }, + { + "epoch": 3.95, + "learning_rate": 5.528570765697096e-06, + "loss": 1.0447, + "step": 53439 + }, + { + "epoch": 3.95, + "learning_rate": 5.528213993538369e-06, + "loss": 0.9634, + "step": 53440 + }, + { + "epoch": 3.95, + "learning_rate": 5.527857228494189e-06, + "loss": 0.9815, + "step": 53441 + }, + { + "epoch": 3.95, + "learning_rate": 5.527500470565121e-06, + "loss": 0.9558, + "step": 53442 + }, + { + "epoch": 3.95, + "learning_rate": 5.52714371975174e-06, + "loss": 1.014, + "step": 53443 + }, + { + "epoch": 3.95, + "learning_rate": 5.52678697605461e-06, + "loss": 1.0316, + "step": 53444 + }, + { + "epoch": 3.95, + "learning_rate": 5.526430239474296e-06, + "loss": 1.1229, + "step": 53445 + }, + { + "epoch": 3.95, + "learning_rate": 5.526073510011369e-06, + "loss": 0.989, + "step": 53446 + }, + { + "epoch": 3.95, + "learning_rate": 5.525716787666391e-06, + "loss": 0.8635, + "step": 53447 + }, + { + "epoch": 3.95, + "learning_rate": 5.525360072439936e-06, + "loss": 0.9997, + "step": 53448 + }, + { + "epoch": 3.95, + "learning_rate": 5.525003364332571e-06, + "loss": 1.0099, + "step": 53449 + }, + { + "epoch": 3.95, + "learning_rate": 5.524646663344859e-06, + "loss": 0.9855, + "step": 53450 + }, + { + "epoch": 3.95, + "learning_rate": 5.524289969477369e-06, + "loss": 1.0081, + "step": 53451 + }, + { + "epoch": 3.95, + "learning_rate": 5.523933282730673e-06, + "loss": 1.0111, + "step": 53452 + }, + { + "epoch": 3.95, + "learning_rate": 5.523576603105334e-06, + "loss": 0.9707, + "step": 53453 + }, + { + "epoch": 3.95, + "learning_rate": 5.523219930601922e-06, + "loss": 1.0199, + "step": 53454 + }, + { + "epoch": 3.95, + "learning_rate": 5.522863265221001e-06, + "loss": 0.9712, + "step": 53455 + }, + { + "epoch": 3.95, + "learning_rate": 5.522506606963138e-06, + "loss": 0.9673, + "step": 53456 + }, + { + "epoch": 3.95, + "learning_rate": 5.522149955828911e-06, + "loss": 1.0348, + "step": 53457 + }, + { + "epoch": 3.95, + "learning_rate": 5.52179331181887e-06, + "loss": 0.9961, + "step": 53458 + }, + { + "epoch": 3.95, + "learning_rate": 5.521436674933598e-06, + "loss": 0.8349, + "step": 53459 + }, + { + "epoch": 3.95, + "learning_rate": 5.52108004517365e-06, + "loss": 0.9563, + "step": 53460 + }, + { + "epoch": 3.95, + "learning_rate": 5.520723422539609e-06, + "loss": 0.9698, + "step": 53461 + }, + { + "epoch": 3.95, + "learning_rate": 5.520366807032025e-06, + "loss": 1.0008, + "step": 53462 + }, + { + "epoch": 3.95, + "learning_rate": 5.520010198651476e-06, + "loss": 1.0028, + "step": 53463 + }, + { + "epoch": 3.95, + "learning_rate": 5.519653597398527e-06, + "loss": 1.0184, + "step": 53464 + }, + { + "epoch": 3.95, + "learning_rate": 5.519297003273746e-06, + "loss": 1.0321, + "step": 53465 + }, + { + "epoch": 3.95, + "learning_rate": 5.518940416277698e-06, + "loss": 1.0076, + "step": 53466 + }, + { + "epoch": 3.95, + "learning_rate": 5.518583836410949e-06, + "loss": 1.005, + "step": 53467 + }, + { + "epoch": 3.95, + "learning_rate": 5.518227263674073e-06, + "loss": 1.0464, + "step": 53468 + }, + { + "epoch": 3.95, + "learning_rate": 5.517870698067633e-06, + "loss": 1.0037, + "step": 53469 + }, + { + "epoch": 3.95, + "learning_rate": 5.517514139592196e-06, + "loss": 1.0592, + "step": 53470 + }, + { + "epoch": 3.95, + "learning_rate": 5.5171575882483265e-06, + "loss": 0.8716, + "step": 53471 + }, + { + "epoch": 3.95, + "learning_rate": 5.5168010440366e-06, + "loss": 0.9202, + "step": 53472 + }, + { + "epoch": 3.95, + "learning_rate": 5.516444506957576e-06, + "loss": 1.0032, + "step": 53473 + }, + { + "epoch": 3.95, + "learning_rate": 5.516087977011827e-06, + "loss": 0.8881, + "step": 53474 + }, + { + "epoch": 3.95, + "learning_rate": 5.515731454199917e-06, + "loss": 1.0744, + "step": 53475 + }, + { + "epoch": 3.95, + "learning_rate": 5.515374938522411e-06, + "loss": 1.013, + "step": 53476 + }, + { + "epoch": 3.95, + "learning_rate": 5.515018429979886e-06, + "loss": 0.9435, + "step": 53477 + }, + { + "epoch": 3.95, + "learning_rate": 5.514661928572896e-06, + "loss": 1.0616, + "step": 53478 + }, + { + "epoch": 3.95, + "learning_rate": 5.514305434302017e-06, + "loss": 1.0299, + "step": 53479 + }, + { + "epoch": 3.95, + "learning_rate": 5.513948947167811e-06, + "loss": 1.0122, + "step": 53480 + }, + { + "epoch": 3.95, + "learning_rate": 5.513592467170857e-06, + "loss": 0.9631, + "step": 53481 + }, + { + "epoch": 3.95, + "learning_rate": 5.513235994311703e-06, + "loss": 0.9768, + "step": 53482 + }, + { + "epoch": 3.95, + "learning_rate": 5.512879528590932e-06, + "loss": 0.9266, + "step": 53483 + }, + { + "epoch": 3.95, + "learning_rate": 5.512523070009105e-06, + "loss": 0.9658, + "step": 53484 + }, + { + "epoch": 3.95, + "learning_rate": 5.512166618566789e-06, + "loss": 0.9938, + "step": 53485 + }, + { + "epoch": 3.95, + "learning_rate": 5.5118101742645514e-06, + "loss": 0.9775, + "step": 53486 + }, + { + "epoch": 3.95, + "learning_rate": 5.511453737102957e-06, + "loss": 0.9714, + "step": 53487 + }, + { + "epoch": 3.95, + "learning_rate": 5.511097307082579e-06, + "loss": 1.1168, + "step": 53488 + }, + { + "epoch": 3.95, + "learning_rate": 5.51074088420398e-06, + "loss": 0.915, + "step": 53489 + }, + { + "epoch": 3.95, + "learning_rate": 5.510384468467729e-06, + "loss": 0.9863, + "step": 53490 + }, + { + "epoch": 3.95, + "learning_rate": 5.510028059874388e-06, + "loss": 1.0304, + "step": 53491 + }, + { + "epoch": 3.95, + "learning_rate": 5.509671658424532e-06, + "loss": 1.1505, + "step": 53492 + }, + { + "epoch": 3.95, + "learning_rate": 5.509315264118725e-06, + "loss": 1.0191, + "step": 53493 + }, + { + "epoch": 3.95, + "learning_rate": 5.508958876957532e-06, + "loss": 1.0508, + "step": 53494 + }, + { + "epoch": 3.95, + "learning_rate": 5.508602496941522e-06, + "loss": 1.0439, + "step": 53495 + }, + { + "epoch": 3.95, + "learning_rate": 5.508246124071256e-06, + "loss": 1.0378, + "step": 53496 + }, + { + "epoch": 3.95, + "learning_rate": 5.507889758347311e-06, + "loss": 0.9714, + "step": 53497 + }, + { + "epoch": 3.95, + "learning_rate": 5.5075333997702485e-06, + "loss": 0.9354, + "step": 53498 + }, + { + "epoch": 3.95, + "learning_rate": 5.507177048340637e-06, + "loss": 0.9822, + "step": 53499 + }, + { + "epoch": 3.95, + "learning_rate": 5.5068207040590374e-06, + "loss": 1.064, + "step": 53500 + }, + { + "epoch": 3.95, + "learning_rate": 5.506464366926027e-06, + "loss": 0.9388, + "step": 53501 + }, + { + "epoch": 3.95, + "learning_rate": 5.5061080369421704e-06, + "loss": 0.9459, + "step": 53502 + }, + { + "epoch": 3.95, + "learning_rate": 5.5057517141080245e-06, + "loss": 1.0969, + "step": 53503 + }, + { + "epoch": 3.95, + "learning_rate": 5.505395398424167e-06, + "loss": 0.9084, + "step": 53504 + }, + { + "epoch": 3.95, + "learning_rate": 5.505039089891156e-06, + "loss": 1.0601, + "step": 53505 + }, + { + "epoch": 3.95, + "learning_rate": 5.504682788509573e-06, + "loss": 0.9877, + "step": 53506 + }, + { + "epoch": 3.95, + "learning_rate": 5.504326494279968e-06, + "loss": 0.9788, + "step": 53507 + }, + { + "epoch": 3.95, + "learning_rate": 5.503970207202918e-06, + "loss": 0.9777, + "step": 53508 + }, + { + "epoch": 3.95, + "learning_rate": 5.503613927278986e-06, + "loss": 1.0128, + "step": 53509 + }, + { + "epoch": 3.95, + "learning_rate": 5.503257654508741e-06, + "loss": 0.9385, + "step": 53510 + }, + { + "epoch": 3.95, + "learning_rate": 5.50290138889275e-06, + "loss": 1.0536, + "step": 53511 + }, + { + "epoch": 3.95, + "learning_rate": 5.502545130431574e-06, + "loss": 0.9226, + "step": 53512 + }, + { + "epoch": 3.95, + "learning_rate": 5.502188879125787e-06, + "loss": 0.9312, + "step": 53513 + }, + { + "epoch": 3.95, + "learning_rate": 5.501832634975953e-06, + "loss": 0.9239, + "step": 53514 + }, + { + "epoch": 3.95, + "learning_rate": 5.50147639798264e-06, + "loss": 0.9256, + "step": 53515 + }, + { + "epoch": 3.95, + "learning_rate": 5.50112016814641e-06, + "loss": 0.9763, + "step": 53516 + }, + { + "epoch": 3.95, + "learning_rate": 5.500763945467837e-06, + "loss": 1.0867, + "step": 53517 + }, + { + "epoch": 3.95, + "learning_rate": 5.500407729947484e-06, + "loss": 1.1367, + "step": 53518 + }, + { + "epoch": 3.95, + "learning_rate": 5.500051521585919e-06, + "loss": 1.0688, + "step": 53519 + }, + { + "epoch": 3.95, + "learning_rate": 5.4996953203837025e-06, + "loss": 0.9429, + "step": 53520 + }, + { + "epoch": 3.95, + "learning_rate": 5.49933912634141e-06, + "loss": 1.0585, + "step": 53521 + }, + { + "epoch": 3.95, + "learning_rate": 5.49898293945961e-06, + "loss": 0.9368, + "step": 53522 + }, + { + "epoch": 3.95, + "learning_rate": 5.4986267597388545e-06, + "loss": 0.933, + "step": 53523 + }, + { + "epoch": 3.95, + "learning_rate": 5.498270587179722e-06, + "loss": 0.9707, + "step": 53524 + }, + { + "epoch": 3.95, + "learning_rate": 5.497914421782775e-06, + "loss": 1.0942, + "step": 53525 + }, + { + "epoch": 3.96, + "learning_rate": 5.497558263548589e-06, + "loss": 1.0341, + "step": 53526 + }, + { + "epoch": 3.96, + "learning_rate": 5.497202112477715e-06, + "loss": 1.0741, + "step": 53527 + }, + { + "epoch": 3.96, + "learning_rate": 5.496845968570733e-06, + "loss": 1.0205, + "step": 53528 + }, + { + "epoch": 3.96, + "learning_rate": 5.4964898318282e-06, + "loss": 0.989, + "step": 53529 + }, + { + "epoch": 3.96, + "learning_rate": 5.496133702250696e-06, + "loss": 0.9624, + "step": 53530 + }, + { + "epoch": 3.96, + "learning_rate": 5.4957775798387744e-06, + "loss": 1.0997, + "step": 53531 + }, + { + "epoch": 3.96, + "learning_rate": 5.495421464593002e-06, + "loss": 0.9721, + "step": 53532 + }, + { + "epoch": 3.96, + "learning_rate": 5.4950653565139536e-06, + "loss": 1.0106, + "step": 53533 + }, + { + "epoch": 3.96, + "learning_rate": 5.494709255602191e-06, + "loss": 1.0206, + "step": 53534 + }, + { + "epoch": 3.96, + "learning_rate": 5.494353161858283e-06, + "loss": 0.9885, + "step": 53535 + }, + { + "epoch": 3.96, + "learning_rate": 5.4939970752827895e-06, + "loss": 1.029, + "step": 53536 + }, + { + "epoch": 3.96, + "learning_rate": 5.493640995876287e-06, + "loss": 1.0544, + "step": 53537 + }, + { + "epoch": 3.96, + "learning_rate": 5.493284923639337e-06, + "loss": 0.9993, + "step": 53538 + }, + { + "epoch": 3.96, + "learning_rate": 5.492928858572506e-06, + "loss": 0.9782, + "step": 53539 + }, + { + "epoch": 3.96, + "learning_rate": 5.492572800676361e-06, + "loss": 0.9941, + "step": 53540 + }, + { + "epoch": 3.96, + "learning_rate": 5.492216749951463e-06, + "loss": 1.0189, + "step": 53541 + }, + { + "epoch": 3.96, + "learning_rate": 5.491860706398389e-06, + "loss": 0.9386, + "step": 53542 + }, + { + "epoch": 3.96, + "learning_rate": 5.4915046700177e-06, + "loss": 0.8765, + "step": 53543 + }, + { + "epoch": 3.96, + "learning_rate": 5.491148640809963e-06, + "loss": 0.9237, + "step": 53544 + }, + { + "epoch": 3.96, + "learning_rate": 5.4907926187757375e-06, + "loss": 1.0104, + "step": 53545 + }, + { + "epoch": 3.96, + "learning_rate": 5.4904366039156035e-06, + "loss": 0.9404, + "step": 53546 + }, + { + "epoch": 3.96, + "learning_rate": 5.490080596230118e-06, + "loss": 0.8884, + "step": 53547 + }, + { + "epoch": 3.96, + "learning_rate": 5.489724595719852e-06, + "loss": 1.0363, + "step": 53548 + }, + { + "epoch": 3.96, + "learning_rate": 5.4893686023853675e-06, + "loss": 1.0605, + "step": 53549 + }, + { + "epoch": 3.96, + "learning_rate": 5.489012616227234e-06, + "loss": 0.9675, + "step": 53550 + }, + { + "epoch": 3.96, + "learning_rate": 5.488656637246017e-06, + "loss": 0.9973, + "step": 53551 + }, + { + "epoch": 3.96, + "learning_rate": 5.488300665442279e-06, + "loss": 1.0525, + "step": 53552 + }, + { + "epoch": 3.96, + "learning_rate": 5.487944700816594e-06, + "loss": 1.0027, + "step": 53553 + }, + { + "epoch": 3.96, + "learning_rate": 5.487588743369523e-06, + "loss": 0.9249, + "step": 53554 + }, + { + "epoch": 3.96, + "learning_rate": 5.487232793101636e-06, + "loss": 1.0032, + "step": 53555 + }, + { + "epoch": 3.96, + "learning_rate": 5.486876850013492e-06, + "loss": 1.0026, + "step": 53556 + }, + { + "epoch": 3.96, + "learning_rate": 5.486520914105667e-06, + "loss": 1.0641, + "step": 53557 + }, + { + "epoch": 3.96, + "learning_rate": 5.48616498537872e-06, + "loss": 1.052, + "step": 53558 + }, + { + "epoch": 3.96, + "learning_rate": 5.4858090638332225e-06, + "loss": 1.0782, + "step": 53559 + }, + { + "epoch": 3.96, + "learning_rate": 5.485453149469737e-06, + "loss": 1.0008, + "step": 53560 + }, + { + "epoch": 3.96, + "learning_rate": 5.485097242288828e-06, + "loss": 0.9661, + "step": 53561 + }, + { + "epoch": 3.96, + "learning_rate": 5.484741342291068e-06, + "loss": 1.0754, + "step": 53562 + }, + { + "epoch": 3.96, + "learning_rate": 5.48438544947702e-06, + "loss": 1.0398, + "step": 53563 + }, + { + "epoch": 3.96, + "learning_rate": 5.484029563847251e-06, + "loss": 1.1641, + "step": 53564 + }, + { + "epoch": 3.96, + "learning_rate": 5.483673685402321e-06, + "loss": 0.9876, + "step": 53565 + }, + { + "epoch": 3.96, + "learning_rate": 5.483317814142806e-06, + "loss": 0.9341, + "step": 53566 + }, + { + "epoch": 3.96, + "learning_rate": 5.482961950069272e-06, + "loss": 1.0064, + "step": 53567 + }, + { + "epoch": 3.96, + "learning_rate": 5.482606093182272e-06, + "loss": 1.0053, + "step": 53568 + }, + { + "epoch": 3.96, + "learning_rate": 5.482250243482386e-06, + "loss": 0.9915, + "step": 53569 + }, + { + "epoch": 3.96, + "learning_rate": 5.481894400970171e-06, + "loss": 1.0221, + "step": 53570 + }, + { + "epoch": 3.96, + "learning_rate": 5.4815385656462065e-06, + "loss": 0.8739, + "step": 53571 + }, + { + "epoch": 3.96, + "learning_rate": 5.481182737511039e-06, + "loss": 0.9972, + "step": 53572 + }, + { + "epoch": 3.96, + "learning_rate": 5.480826916565251e-06, + "loss": 0.9793, + "step": 53573 + }, + { + "epoch": 3.96, + "learning_rate": 5.480471102809399e-06, + "loss": 1.0798, + "step": 53574 + }, + { + "epoch": 3.96, + "learning_rate": 5.48011529624406e-06, + "loss": 1.0928, + "step": 53575 + }, + { + "epoch": 3.96, + "learning_rate": 5.479759496869785e-06, + "loss": 1.0331, + "step": 53576 + }, + { + "epoch": 3.96, + "learning_rate": 5.4794037046871516e-06, + "loss": 0.9485, + "step": 53577 + }, + { + "epoch": 3.96, + "learning_rate": 5.479047919696723e-06, + "loss": 1.0162, + "step": 53578 + }, + { + "epoch": 3.96, + "learning_rate": 5.4786921418990636e-06, + "loss": 1.0452, + "step": 53579 + }, + { + "epoch": 3.96, + "learning_rate": 5.478336371294742e-06, + "loss": 1.1279, + "step": 53580 + }, + { + "epoch": 3.96, + "learning_rate": 5.477980607884316e-06, + "loss": 0.9266, + "step": 53581 + }, + { + "epoch": 3.96, + "learning_rate": 5.4776248516683635e-06, + "loss": 0.9491, + "step": 53582 + }, + { + "epoch": 3.96, + "learning_rate": 5.477269102647445e-06, + "loss": 1.0636, + "step": 53583 + }, + { + "epoch": 3.96, + "learning_rate": 5.476913360822127e-06, + "loss": 0.9325, + "step": 53584 + }, + { + "epoch": 3.96, + "learning_rate": 5.476557626192971e-06, + "loss": 0.9232, + "step": 53585 + }, + { + "epoch": 3.96, + "learning_rate": 5.476201898760552e-06, + "loss": 1.0143, + "step": 53586 + }, + { + "epoch": 3.96, + "learning_rate": 5.475846178525428e-06, + "loss": 1.0082, + "step": 53587 + }, + { + "epoch": 3.96, + "learning_rate": 5.475490465488171e-06, + "loss": 1.064, + "step": 53588 + }, + { + "epoch": 3.96, + "learning_rate": 5.475134759649343e-06, + "loss": 0.9148, + "step": 53589 + }, + { + "epoch": 3.96, + "learning_rate": 5.474779061009506e-06, + "loss": 0.9826, + "step": 53590 + }, + { + "epoch": 3.96, + "learning_rate": 5.474423369569236e-06, + "loss": 1.0163, + "step": 53591 + }, + { + "epoch": 3.96, + "learning_rate": 5.474067685329093e-06, + "loss": 0.9751, + "step": 53592 + }, + { + "epoch": 3.96, + "learning_rate": 5.473712008289645e-06, + "loss": 1.0691, + "step": 53593 + }, + { + "epoch": 3.96, + "learning_rate": 5.47335633845145e-06, + "loss": 0.9553, + "step": 53594 + }, + { + "epoch": 3.96, + "learning_rate": 5.47300067581509e-06, + "loss": 1.0834, + "step": 53595 + }, + { + "epoch": 3.96, + "learning_rate": 5.472645020381117e-06, + "loss": 1.0296, + "step": 53596 + }, + { + "epoch": 3.96, + "learning_rate": 5.472289372150097e-06, + "loss": 0.9977, + "step": 53597 + }, + { + "epoch": 3.96, + "learning_rate": 5.471933731122603e-06, + "loss": 0.9625, + "step": 53598 + }, + { + "epoch": 3.96, + "learning_rate": 5.471578097299197e-06, + "loss": 0.8887, + "step": 53599 + }, + { + "epoch": 3.96, + "learning_rate": 5.471222470680448e-06, + "loss": 1.0698, + "step": 53600 + }, + { + "epoch": 3.96, + "learning_rate": 5.470866851266914e-06, + "loss": 1.0275, + "step": 53601 + }, + { + "epoch": 3.96, + "learning_rate": 5.470511239059169e-06, + "loss": 0.9705, + "step": 53602 + }, + { + "epoch": 3.96, + "learning_rate": 5.4701556340577765e-06, + "loss": 0.9796, + "step": 53603 + }, + { + "epoch": 3.96, + "learning_rate": 5.469800036263302e-06, + "loss": 1.008, + "step": 53604 + }, + { + "epoch": 3.96, + "learning_rate": 5.469444445676311e-06, + "loss": 1.1128, + "step": 53605 + }, + { + "epoch": 3.96, + "learning_rate": 5.469088862297365e-06, + "loss": 1.0056, + "step": 53606 + }, + { + "epoch": 3.96, + "learning_rate": 5.468733286127037e-06, + "loss": 0.9308, + "step": 53607 + }, + { + "epoch": 3.96, + "learning_rate": 5.46837771716589e-06, + "loss": 0.9727, + "step": 53608 + }, + { + "epoch": 3.96, + "learning_rate": 5.46802215541449e-06, + "loss": 0.976, + "step": 53609 + }, + { + "epoch": 3.96, + "learning_rate": 5.467666600873396e-06, + "loss": 1.0914, + "step": 53610 + }, + { + "epoch": 3.96, + "learning_rate": 5.467311053543185e-06, + "loss": 1.0947, + "step": 53611 + }, + { + "epoch": 3.96, + "learning_rate": 5.466955513424417e-06, + "loss": 0.9682, + "step": 53612 + }, + { + "epoch": 3.96, + "learning_rate": 5.466599980517657e-06, + "loss": 0.896, + "step": 53613 + }, + { + "epoch": 3.96, + "learning_rate": 5.466244454823473e-06, + "loss": 1.0717, + "step": 53614 + }, + { + "epoch": 3.96, + "learning_rate": 5.465888936342425e-06, + "loss": 1.0939, + "step": 53615 + }, + { + "epoch": 3.96, + "learning_rate": 5.465533425075091e-06, + "loss": 0.9086, + "step": 53616 + }, + { + "epoch": 3.96, + "learning_rate": 5.46517792102202e-06, + "loss": 1.0672, + "step": 53617 + }, + { + "epoch": 3.96, + "learning_rate": 5.464822424183789e-06, + "loss": 1.0985, + "step": 53618 + }, + { + "epoch": 3.96, + "learning_rate": 5.464466934560958e-06, + "loss": 0.9903, + "step": 53619 + }, + { + "epoch": 3.96, + "learning_rate": 5.464111452154103e-06, + "loss": 0.9328, + "step": 53620 + }, + { + "epoch": 3.96, + "learning_rate": 5.463755976963774e-06, + "loss": 0.9506, + "step": 53621 + }, + { + "epoch": 3.96, + "learning_rate": 5.463400508990547e-06, + "loss": 0.9423, + "step": 53622 + }, + { + "epoch": 3.96, + "learning_rate": 5.463045048234986e-06, + "loss": 0.9049, + "step": 53623 + }, + { + "epoch": 3.96, + "learning_rate": 5.462689594697656e-06, + "loss": 1.0157, + "step": 53624 + }, + { + "epoch": 3.96, + "learning_rate": 5.46233414837912e-06, + "loss": 0.9782, + "step": 53625 + }, + { + "epoch": 3.96, + "learning_rate": 5.461978709279943e-06, + "loss": 1.1314, + "step": 53626 + }, + { + "epoch": 3.96, + "learning_rate": 5.461623277400697e-06, + "loss": 1.1001, + "step": 53627 + }, + { + "epoch": 3.96, + "learning_rate": 5.461267852741943e-06, + "loss": 1.1016, + "step": 53628 + }, + { + "epoch": 3.96, + "learning_rate": 5.460912435304247e-06, + "loss": 1.0279, + "step": 53629 + }, + { + "epoch": 3.96, + "learning_rate": 5.46055702508817e-06, + "loss": 1.0364, + "step": 53630 + }, + { + "epoch": 3.96, + "learning_rate": 5.460201622094286e-06, + "loss": 1.0417, + "step": 53631 + }, + { + "epoch": 3.96, + "learning_rate": 5.459846226323157e-06, + "loss": 0.9715, + "step": 53632 + }, + { + "epoch": 3.96, + "learning_rate": 5.459490837775347e-06, + "loss": 1.009, + "step": 53633 + }, + { + "epoch": 3.96, + "learning_rate": 5.459135456451422e-06, + "loss": 0.9832, + "step": 53634 + }, + { + "epoch": 3.96, + "learning_rate": 5.458780082351944e-06, + "loss": 1.1086, + "step": 53635 + }, + { + "epoch": 3.96, + "learning_rate": 5.458424715477485e-06, + "loss": 0.9683, + "step": 53636 + }, + { + "epoch": 3.96, + "learning_rate": 5.458069355828608e-06, + "loss": 0.9096, + "step": 53637 + }, + { + "epoch": 3.96, + "learning_rate": 5.457714003405878e-06, + "loss": 0.9379, + "step": 53638 + }, + { + "epoch": 3.96, + "learning_rate": 5.457358658209855e-06, + "loss": 0.9301, + "step": 53639 + }, + { + "epoch": 3.96, + "learning_rate": 5.457003320241119e-06, + "loss": 0.9282, + "step": 53640 + }, + { + "epoch": 3.96, + "learning_rate": 5.456647989500217e-06, + "loss": 1.046, + "step": 53641 + }, + { + "epoch": 3.96, + "learning_rate": 5.456292665987726e-06, + "loss": 1.0358, + "step": 53642 + }, + { + "epoch": 3.96, + "learning_rate": 5.455937349704209e-06, + "loss": 0.9523, + "step": 53643 + }, + { + "epoch": 3.96, + "learning_rate": 5.455582040650232e-06, + "loss": 0.8959, + "step": 53644 + }, + { + "epoch": 3.96, + "learning_rate": 5.455226738826356e-06, + "loss": 1.1137, + "step": 53645 + }, + { + "epoch": 3.96, + "learning_rate": 5.454871444233148e-06, + "loss": 0.9866, + "step": 53646 + }, + { + "epoch": 3.96, + "learning_rate": 5.4545161568711765e-06, + "loss": 0.9859, + "step": 53647 + }, + { + "epoch": 3.96, + "learning_rate": 5.454160876741006e-06, + "loss": 0.9075, + "step": 53648 + }, + { + "epoch": 3.96, + "learning_rate": 5.4538056038432e-06, + "loss": 1.0508, + "step": 53649 + }, + { + "epoch": 3.96, + "learning_rate": 5.453450338178321e-06, + "loss": 0.9897, + "step": 53650 + }, + { + "epoch": 3.96, + "learning_rate": 5.453095079746941e-06, + "loss": 1.0422, + "step": 53651 + }, + { + "epoch": 3.96, + "learning_rate": 5.4527398285496214e-06, + "loss": 1.1379, + "step": 53652 + }, + { + "epoch": 3.96, + "learning_rate": 5.452384584586927e-06, + "loss": 0.9942, + "step": 53653 + }, + { + "epoch": 3.96, + "learning_rate": 5.452029347859425e-06, + "loss": 0.9518, + "step": 53654 + }, + { + "epoch": 3.96, + "learning_rate": 5.451674118367676e-06, + "loss": 0.9577, + "step": 53655 + }, + { + "epoch": 3.96, + "learning_rate": 5.45131889611225e-06, + "loss": 0.979, + "step": 53656 + }, + { + "epoch": 3.96, + "learning_rate": 5.450963681093713e-06, + "loss": 1.0004, + "step": 53657 + }, + { + "epoch": 3.96, + "learning_rate": 5.450608473312626e-06, + "loss": 1.0431, + "step": 53658 + }, + { + "epoch": 3.96, + "learning_rate": 5.450253272769552e-06, + "loss": 1.0015, + "step": 53659 + }, + { + "epoch": 3.96, + "learning_rate": 5.449898079465064e-06, + "loss": 0.972, + "step": 53660 + }, + { + "epoch": 3.96, + "learning_rate": 5.449542893399727e-06, + "loss": 0.9417, + "step": 53661 + }, + { + "epoch": 3.97, + "learning_rate": 5.449187714574095e-06, + "loss": 0.9802, + "step": 53662 + }, + { + "epoch": 3.97, + "learning_rate": 5.448832542988743e-06, + "loss": 1.1213, + "step": 53663 + }, + { + "epoch": 3.97, + "learning_rate": 5.44847737864423e-06, + "loss": 0.9284, + "step": 53664 + }, + { + "epoch": 3.97, + "learning_rate": 5.4481222215411325e-06, + "loss": 1.1222, + "step": 53665 + }, + { + "epoch": 3.97, + "learning_rate": 5.447767071679999e-06, + "loss": 1.0154, + "step": 53666 + }, + { + "epoch": 3.97, + "learning_rate": 5.447411929061408e-06, + "loss": 0.9994, + "step": 53667 + }, + { + "epoch": 3.97, + "learning_rate": 5.447056793685914e-06, + "loss": 1.1175, + "step": 53668 + }, + { + "epoch": 3.97, + "learning_rate": 5.446701665554097e-06, + "loss": 1.0592, + "step": 53669 + }, + { + "epoch": 3.97, + "learning_rate": 5.4463465446665075e-06, + "loss": 0.9123, + "step": 53670 + }, + { + "epoch": 3.97, + "learning_rate": 5.445991431023712e-06, + "loss": 0.9056, + "step": 53671 + }, + { + "epoch": 3.97, + "learning_rate": 5.445636324626282e-06, + "loss": 0.8802, + "step": 53672 + }, + { + "epoch": 3.97, + "learning_rate": 5.44528122547478e-06, + "loss": 1.0912, + "step": 53673 + }, + { + "epoch": 3.97, + "learning_rate": 5.44492613356977e-06, + "loss": 0.9308, + "step": 53674 + }, + { + "epoch": 3.97, + "learning_rate": 5.444571048911814e-06, + "loss": 0.9847, + "step": 53675 + }, + { + "epoch": 3.97, + "learning_rate": 5.444215971501485e-06, + "loss": 0.922, + "step": 53676 + }, + { + "epoch": 3.97, + "learning_rate": 5.443860901339342e-06, + "loss": 1.0048, + "step": 53677 + }, + { + "epoch": 3.97, + "learning_rate": 5.443505838425951e-06, + "loss": 1.0453, + "step": 53678 + }, + { + "epoch": 3.97, + "learning_rate": 5.4431507827618776e-06, + "loss": 0.9495, + "step": 53679 + }, + { + "epoch": 3.97, + "learning_rate": 5.442795734347682e-06, + "loss": 0.9821, + "step": 53680 + }, + { + "epoch": 3.97, + "learning_rate": 5.4424406931839365e-06, + "loss": 1.0666, + "step": 53681 + }, + { + "epoch": 3.97, + "learning_rate": 5.4420856592712036e-06, + "loss": 0.9842, + "step": 53682 + }, + { + "epoch": 3.97, + "learning_rate": 5.441730632610048e-06, + "loss": 1.0179, + "step": 53683 + }, + { + "epoch": 3.97, + "learning_rate": 5.441375613201028e-06, + "loss": 0.9664, + "step": 53684 + }, + { + "epoch": 3.97, + "learning_rate": 5.441020601044723e-06, + "loss": 0.9501, + "step": 53685 + }, + { + "epoch": 3.97, + "learning_rate": 5.440665596141681e-06, + "loss": 0.9177, + "step": 53686 + }, + { + "epoch": 3.97, + "learning_rate": 5.4403105984924784e-06, + "loss": 1.0615, + "step": 53687 + }, + { + "epoch": 3.97, + "learning_rate": 5.439955608097673e-06, + "loss": 0.9566, + "step": 53688 + }, + { + "epoch": 3.97, + "learning_rate": 5.43960062495784e-06, + "loss": 0.8608, + "step": 53689 + }, + { + "epoch": 3.97, + "learning_rate": 5.439245649073535e-06, + "loss": 0.9279, + "step": 53690 + }, + { + "epoch": 3.97, + "learning_rate": 5.438890680445319e-06, + "loss": 0.9843, + "step": 53691 + }, + { + "epoch": 3.97, + "learning_rate": 5.4385357190737674e-06, + "loss": 1.077, + "step": 53692 + }, + { + "epoch": 3.97, + "learning_rate": 5.43818076495944e-06, + "loss": 0.9939, + "step": 53693 + }, + { + "epoch": 3.97, + "learning_rate": 5.437825818102902e-06, + "loss": 0.9302, + "step": 53694 + }, + { + "epoch": 3.97, + "learning_rate": 5.4374708785047135e-06, + "loss": 0.9026, + "step": 53695 + }, + { + "epoch": 3.97, + "learning_rate": 5.437115946165448e-06, + "loss": 0.9736, + "step": 53696 + }, + { + "epoch": 3.97, + "learning_rate": 5.436761021085666e-06, + "loss": 0.9019, + "step": 53697 + }, + { + "epoch": 3.97, + "learning_rate": 5.436406103265931e-06, + "loss": 0.9415, + "step": 53698 + }, + { + "epoch": 3.97, + "learning_rate": 5.4360511927068095e-06, + "loss": 1.0739, + "step": 53699 + }, + { + "epoch": 3.97, + "learning_rate": 5.43569628940886e-06, + "loss": 1.0169, + "step": 53700 + }, + { + "epoch": 3.97, + "learning_rate": 5.435341393372657e-06, + "loss": 0.9209, + "step": 53701 + }, + { + "epoch": 3.97, + "learning_rate": 5.4349865045987615e-06, + "loss": 0.9736, + "step": 53702 + }, + { + "epoch": 3.97, + "learning_rate": 5.434631623087736e-06, + "loss": 1.0026, + "step": 53703 + }, + { + "epoch": 3.97, + "learning_rate": 5.434276748840144e-06, + "loss": 1.0392, + "step": 53704 + }, + { + "epoch": 3.97, + "learning_rate": 5.433921881856554e-06, + "loss": 0.9273, + "step": 53705 + }, + { + "epoch": 3.97, + "learning_rate": 5.43356702213753e-06, + "loss": 0.999, + "step": 53706 + }, + { + "epoch": 3.97, + "learning_rate": 5.4332121696836355e-06, + "loss": 1.093, + "step": 53707 + }, + { + "epoch": 3.97, + "learning_rate": 5.432857324495435e-06, + "loss": 0.9419, + "step": 53708 + }, + { + "epoch": 3.97, + "learning_rate": 5.43250248657349e-06, + "loss": 1.031, + "step": 53709 + }, + { + "epoch": 3.97, + "learning_rate": 5.432147655918375e-06, + "loss": 1.1295, + "step": 53710 + }, + { + "epoch": 3.97, + "learning_rate": 5.43179283253064e-06, + "loss": 0.9528, + "step": 53711 + }, + { + "epoch": 3.97, + "learning_rate": 5.431438016410862e-06, + "loss": 0.9259, + "step": 53712 + }, + { + "epoch": 3.97, + "learning_rate": 5.431083207559595e-06, + "loss": 0.9137, + "step": 53713 + }, + { + "epoch": 3.97, + "learning_rate": 5.430728405977419e-06, + "loss": 1.032, + "step": 53714 + }, + { + "epoch": 3.97, + "learning_rate": 5.430373611664881e-06, + "loss": 0.9386, + "step": 53715 + }, + { + "epoch": 3.97, + "learning_rate": 5.430018824622556e-06, + "loss": 0.8675, + "step": 53716 + }, + { + "epoch": 3.97, + "learning_rate": 5.429664044851007e-06, + "loss": 1.0635, + "step": 53717 + }, + { + "epoch": 3.97, + "learning_rate": 5.429309272350796e-06, + "loss": 1.0689, + "step": 53718 + }, + { + "epoch": 3.97, + "learning_rate": 5.428954507122488e-06, + "loss": 1.017, + "step": 53719 + }, + { + "epoch": 3.97, + "learning_rate": 5.428599749166645e-06, + "loss": 1.0336, + "step": 53720 + }, + { + "epoch": 3.97, + "learning_rate": 5.428244998483838e-06, + "loss": 0.9396, + "step": 53721 + }, + { + "epoch": 3.97, + "learning_rate": 5.427890255074628e-06, + "loss": 1.0066, + "step": 53722 + }, + { + "epoch": 3.97, + "learning_rate": 5.427535518939579e-06, + "loss": 0.9831, + "step": 53723 + }, + { + "epoch": 3.97, + "learning_rate": 5.427180790079252e-06, + "loss": 0.9507, + "step": 53724 + }, + { + "epoch": 3.97, + "learning_rate": 5.426826068494218e-06, + "loss": 0.9495, + "step": 53725 + }, + { + "epoch": 3.97, + "learning_rate": 5.426471354185039e-06, + "loss": 1.0515, + "step": 53726 + }, + { + "epoch": 3.97, + "learning_rate": 5.426116647152278e-06, + "loss": 1.0877, + "step": 53727 + }, + { + "epoch": 3.97, + "learning_rate": 5.4257619473965e-06, + "loss": 1.0899, + "step": 53728 + }, + { + "epoch": 3.97, + "learning_rate": 5.425407254918267e-06, + "loss": 0.9497, + "step": 53729 + }, + { + "epoch": 3.97, + "learning_rate": 5.425052569718152e-06, + "loss": 0.9791, + "step": 53730 + }, + { + "epoch": 3.97, + "learning_rate": 5.424697891796706e-06, + "loss": 1.0844, + "step": 53731 + }, + { + "epoch": 3.97, + "learning_rate": 5.424343221154502e-06, + "loss": 0.9585, + "step": 53732 + }, + { + "epoch": 3.97, + "learning_rate": 5.423988557792101e-06, + "loss": 0.9019, + "step": 53733 + }, + { + "epoch": 3.97, + "learning_rate": 5.423633901710075e-06, + "loss": 1.0467, + "step": 53734 + }, + { + "epoch": 3.97, + "learning_rate": 5.423279252908975e-06, + "loss": 1.0827, + "step": 53735 + }, + { + "epoch": 3.97, + "learning_rate": 5.422924611389377e-06, + "loss": 1.0137, + "step": 53736 + }, + { + "epoch": 3.97, + "learning_rate": 5.422569977151839e-06, + "loss": 0.9419, + "step": 53737 + }, + { + "epoch": 3.97, + "learning_rate": 5.422215350196926e-06, + "loss": 0.9961, + "step": 53738 + }, + { + "epoch": 3.97, + "learning_rate": 5.421860730525204e-06, + "loss": 1.0318, + "step": 53739 + }, + { + "epoch": 3.97, + "learning_rate": 5.421506118137232e-06, + "loss": 0.9567, + "step": 53740 + }, + { + "epoch": 3.97, + "learning_rate": 5.421151513033582e-06, + "loss": 1.0221, + "step": 53741 + }, + { + "epoch": 3.97, + "learning_rate": 5.420796915214816e-06, + "loss": 1.0408, + "step": 53742 + }, + { + "epoch": 3.97, + "learning_rate": 5.420442324681494e-06, + "loss": 1.0271, + "step": 53743 + }, + { + "epoch": 3.97, + "learning_rate": 5.420087741434181e-06, + "loss": 0.9333, + "step": 53744 + }, + { + "epoch": 3.97, + "learning_rate": 5.4197331654734456e-06, + "loss": 1.0257, + "step": 53745 + }, + { + "epoch": 3.97, + "learning_rate": 5.419378596799849e-06, + "loss": 1.0704, + "step": 53746 + }, + { + "epoch": 3.97, + "learning_rate": 5.419024035413955e-06, + "loss": 1.0086, + "step": 53747 + }, + { + "epoch": 3.97, + "learning_rate": 5.4186694813163296e-06, + "loss": 1.045, + "step": 53748 + }, + { + "epoch": 3.97, + "learning_rate": 5.418314934507532e-06, + "loss": 1.0533, + "step": 53749 + }, + { + "epoch": 3.97, + "learning_rate": 5.417960394988133e-06, + "loss": 1.0066, + "step": 53750 + }, + { + "epoch": 3.97, + "learning_rate": 5.417605862758692e-06, + "loss": 1.0115, + "step": 53751 + }, + { + "epoch": 3.97, + "learning_rate": 5.417251337819778e-06, + "loss": 1.0373, + "step": 53752 + }, + { + "epoch": 3.97, + "learning_rate": 5.416896820171944e-06, + "loss": 1.0148, + "step": 53753 + }, + { + "epoch": 3.97, + "learning_rate": 5.416542309815768e-06, + "loss": 0.9948, + "step": 53754 + }, + { + "epoch": 3.97, + "learning_rate": 5.41618780675181e-06, + "loss": 0.9996, + "step": 53755 + }, + { + "epoch": 3.97, + "learning_rate": 5.415833310980625e-06, + "loss": 1.0088, + "step": 53756 + }, + { + "epoch": 3.97, + "learning_rate": 5.415478822502787e-06, + "loss": 0.9401, + "step": 53757 + }, + { + "epoch": 3.97, + "learning_rate": 5.415124341318851e-06, + "loss": 1.0582, + "step": 53758 + }, + { + "epoch": 3.97, + "learning_rate": 5.414769867429397e-06, + "loss": 0.9359, + "step": 53759 + }, + { + "epoch": 3.97, + "learning_rate": 5.414415400834969e-06, + "loss": 0.9134, + "step": 53760 + }, + { + "epoch": 3.97, + "learning_rate": 5.414060941536146e-06, + "loss": 1.0917, + "step": 53761 + }, + { + "epoch": 3.97, + "learning_rate": 5.413706489533485e-06, + "loss": 1.0976, + "step": 53762 + }, + { + "epoch": 3.97, + "learning_rate": 5.413352044827551e-06, + "loss": 1.0325, + "step": 53763 + }, + { + "epoch": 3.97, + "learning_rate": 5.4129976074189095e-06, + "loss": 1.0221, + "step": 53764 + }, + { + "epoch": 3.97, + "learning_rate": 5.412643177308119e-06, + "loss": 0.9716, + "step": 53765 + }, + { + "epoch": 3.97, + "learning_rate": 5.412288754495752e-06, + "loss": 1.0642, + "step": 53766 + }, + { + "epoch": 3.97, + "learning_rate": 5.4119343389823664e-06, + "loss": 0.9151, + "step": 53767 + }, + { + "epoch": 3.97, + "learning_rate": 5.411579930768529e-06, + "loss": 0.9697, + "step": 53768 + }, + { + "epoch": 3.97, + "learning_rate": 5.411225529854797e-06, + "loss": 0.94, + "step": 53769 + }, + { + "epoch": 3.97, + "learning_rate": 5.4108711362417445e-06, + "loss": 0.9799, + "step": 53770 + }, + { + "epoch": 3.97, + "learning_rate": 5.41051674992993e-06, + "loss": 0.978, + "step": 53771 + }, + { + "epoch": 3.97, + "learning_rate": 5.410162370919918e-06, + "loss": 0.9859, + "step": 53772 + }, + { + "epoch": 3.97, + "learning_rate": 5.409807999212272e-06, + "loss": 0.8983, + "step": 53773 + }, + { + "epoch": 3.97, + "learning_rate": 5.4094536348075525e-06, + "loss": 0.9499, + "step": 53774 + }, + { + "epoch": 3.97, + "learning_rate": 5.409099277706333e-06, + "loss": 1.0177, + "step": 53775 + }, + { + "epoch": 3.97, + "learning_rate": 5.408744927909164e-06, + "loss": 0.9398, + "step": 53776 + }, + { + "epoch": 3.97, + "learning_rate": 5.40839058541662e-06, + "loss": 1.0052, + "step": 53777 + }, + { + "epoch": 3.97, + "learning_rate": 5.408036250229256e-06, + "loss": 1.0989, + "step": 53778 + }, + { + "epoch": 3.97, + "learning_rate": 5.40768192234765e-06, + "loss": 1.1155, + "step": 53779 + }, + { + "epoch": 3.97, + "learning_rate": 5.407327601772348e-06, + "loss": 0.9629, + "step": 53780 + }, + { + "epoch": 3.97, + "learning_rate": 5.406973288503925e-06, + "loss": 0.882, + "step": 53781 + }, + { + "epoch": 3.97, + "learning_rate": 5.406618982542942e-06, + "loss": 1.0005, + "step": 53782 + }, + { + "epoch": 3.97, + "learning_rate": 5.406264683889963e-06, + "loss": 1.0262, + "step": 53783 + }, + { + "epoch": 3.97, + "learning_rate": 5.405910392545552e-06, + "loss": 1.089, + "step": 53784 + }, + { + "epoch": 3.97, + "learning_rate": 5.405556108510266e-06, + "loss": 0.9289, + "step": 53785 + }, + { + "epoch": 3.97, + "learning_rate": 5.40520183178468e-06, + "loss": 0.9987, + "step": 53786 + }, + { + "epoch": 3.97, + "learning_rate": 5.404847562369351e-06, + "loss": 0.9898, + "step": 53787 + }, + { + "epoch": 3.97, + "learning_rate": 5.404493300264845e-06, + "loss": 1.0771, + "step": 53788 + }, + { + "epoch": 3.97, + "learning_rate": 5.404139045471719e-06, + "loss": 0.9441, + "step": 53789 + }, + { + "epoch": 3.97, + "learning_rate": 5.403784797990548e-06, + "loss": 1.0272, + "step": 53790 + }, + { + "epoch": 3.97, + "learning_rate": 5.403430557821887e-06, + "loss": 1.0177, + "step": 53791 + }, + { + "epoch": 3.97, + "learning_rate": 5.403076324966305e-06, + "loss": 0.9831, + "step": 53792 + }, + { + "epoch": 3.97, + "learning_rate": 5.40272209942436e-06, + "loss": 1.0526, + "step": 53793 + }, + { + "epoch": 3.97, + "learning_rate": 5.4023678811966175e-06, + "loss": 0.9098, + "step": 53794 + }, + { + "epoch": 3.97, + "learning_rate": 5.402013670283643e-06, + "loss": 1.1095, + "step": 53795 + }, + { + "epoch": 3.97, + "learning_rate": 5.401659466686e-06, + "loss": 0.9973, + "step": 53796 + }, + { + "epoch": 3.98, + "learning_rate": 5.401305270404251e-06, + "loss": 1.0016, + "step": 53797 + }, + { + "epoch": 3.98, + "learning_rate": 5.4009510814389566e-06, + "loss": 0.9874, + "step": 53798 + }, + { + "epoch": 3.98, + "learning_rate": 5.400596899790687e-06, + "loss": 1.0045, + "step": 53799 + }, + { + "epoch": 3.98, + "learning_rate": 5.400242725460001e-06, + "loss": 0.8847, + "step": 53800 + }, + { + "epoch": 3.98, + "learning_rate": 5.399888558447463e-06, + "loss": 0.8574, + "step": 53801 + }, + { + "epoch": 3.98, + "learning_rate": 5.399534398753637e-06, + "loss": 0.9211, + "step": 53802 + }, + { + "epoch": 3.98, + "learning_rate": 5.399180246379081e-06, + "loss": 0.9718, + "step": 53803 + }, + { + "epoch": 3.98, + "learning_rate": 5.3988261013243725e-06, + "loss": 1.0734, + "step": 53804 + }, + { + "epoch": 3.98, + "learning_rate": 5.398471963590058e-06, + "loss": 0.9822, + "step": 53805 + }, + { + "epoch": 3.98, + "learning_rate": 5.398117833176711e-06, + "loss": 0.9409, + "step": 53806 + }, + { + "epoch": 3.98, + "learning_rate": 5.397763710084895e-06, + "loss": 1.1025, + "step": 53807 + }, + { + "epoch": 3.98, + "learning_rate": 5.397409594315169e-06, + "loss": 1.0195, + "step": 53808 + }, + { + "epoch": 3.98, + "learning_rate": 5.397055485868096e-06, + "loss": 1.0131, + "step": 53809 + }, + { + "epoch": 3.98, + "learning_rate": 5.396701384744245e-06, + "loss": 0.9834, + "step": 53810 + }, + { + "epoch": 3.98, + "learning_rate": 5.396347290944177e-06, + "loss": 0.9484, + "step": 53811 + }, + { + "epoch": 3.98, + "learning_rate": 5.395993204468454e-06, + "loss": 0.9261, + "step": 53812 + }, + { + "epoch": 3.98, + "learning_rate": 5.39563912531764e-06, + "loss": 1.0139, + "step": 53813 + }, + { + "epoch": 3.98, + "learning_rate": 5.395285053492294e-06, + "loss": 0.9425, + "step": 53814 + }, + { + "epoch": 3.98, + "learning_rate": 5.394930988992988e-06, + "loss": 0.979, + "step": 53815 + }, + { + "epoch": 3.98, + "learning_rate": 5.394576931820281e-06, + "loss": 0.9307, + "step": 53816 + }, + { + "epoch": 3.98, + "learning_rate": 5.394222881974735e-06, + "loss": 1.0373, + "step": 53817 + }, + { + "epoch": 3.98, + "learning_rate": 5.393868839456913e-06, + "loss": 1.0094, + "step": 53818 + }, + { + "epoch": 3.98, + "learning_rate": 5.393514804267382e-06, + "loss": 0.8308, + "step": 53819 + }, + { + "epoch": 3.98, + "learning_rate": 5.3931607764067074e-06, + "loss": 1.048, + "step": 53820 + }, + { + "epoch": 3.98, + "learning_rate": 5.39280675587544e-06, + "loss": 0.9577, + "step": 53821 + }, + { + "epoch": 3.98, + "learning_rate": 5.3924527426741556e-06, + "loss": 0.9837, + "step": 53822 + }, + { + "epoch": 3.98, + "learning_rate": 5.392098736803408e-06, + "loss": 1.0059, + "step": 53823 + }, + { + "epoch": 3.98, + "learning_rate": 5.391744738263774e-06, + "loss": 1.02, + "step": 53824 + }, + { + "epoch": 3.98, + "learning_rate": 5.3913907470558e-06, + "loss": 0.9958, + "step": 53825 + }, + { + "epoch": 3.98, + "learning_rate": 5.391036763180062e-06, + "loss": 0.9498, + "step": 53826 + }, + { + "epoch": 3.98, + "learning_rate": 5.390682786637114e-06, + "loss": 0.8732, + "step": 53827 + }, + { + "epoch": 3.98, + "learning_rate": 5.390328817427533e-06, + "loss": 1.1254, + "step": 53828 + }, + { + "epoch": 3.98, + "learning_rate": 5.389974855551867e-06, + "loss": 1.06, + "step": 53829 + }, + { + "epoch": 3.98, + "learning_rate": 5.3896209010106814e-06, + "loss": 0.9996, + "step": 53830 + }, + { + "epoch": 3.98, + "learning_rate": 5.389266953804548e-06, + "loss": 1.0478, + "step": 53831 + }, + { + "epoch": 3.98, + "learning_rate": 5.388913013934024e-06, + "loss": 0.9638, + "step": 53832 + }, + { + "epoch": 3.98, + "learning_rate": 5.388559081399674e-06, + "loss": 0.9797, + "step": 53833 + }, + { + "epoch": 3.98, + "learning_rate": 5.3882051562020555e-06, + "loss": 0.9883, + "step": 53834 + }, + { + "epoch": 3.98, + "learning_rate": 5.387851238341741e-06, + "loss": 1.0529, + "step": 53835 + }, + { + "epoch": 3.98, + "learning_rate": 5.387497327819291e-06, + "loss": 0.994, + "step": 53836 + }, + { + "epoch": 3.98, + "learning_rate": 5.387143424635264e-06, + "loss": 0.9502, + "step": 53837 + }, + { + "epoch": 3.98, + "learning_rate": 5.386789528790228e-06, + "loss": 0.8992, + "step": 53838 + }, + { + "epoch": 3.98, + "learning_rate": 5.38643564028474e-06, + "loss": 1.0711, + "step": 53839 + }, + { + "epoch": 3.98, + "learning_rate": 5.3860817591193685e-06, + "loss": 1.0321, + "step": 53840 + }, + { + "epoch": 3.98, + "learning_rate": 5.385727885294676e-06, + "loss": 1.1074, + "step": 53841 + }, + { + "epoch": 3.98, + "learning_rate": 5.385374018811224e-06, + "loss": 1.0237, + "step": 53842 + }, + { + "epoch": 3.98, + "learning_rate": 5.385020159669574e-06, + "loss": 1.0464, + "step": 53843 + }, + { + "epoch": 3.98, + "learning_rate": 5.3846663078702935e-06, + "loss": 1.1215, + "step": 53844 + }, + { + "epoch": 3.98, + "learning_rate": 5.384312463413944e-06, + "loss": 0.9877, + "step": 53845 + }, + { + "epoch": 3.98, + "learning_rate": 5.383958626301085e-06, + "loss": 0.9587, + "step": 53846 + }, + { + "epoch": 3.98, + "learning_rate": 5.383604796532278e-06, + "loss": 1.0105, + "step": 53847 + }, + { + "epoch": 3.98, + "learning_rate": 5.383250974108095e-06, + "loss": 1.0147, + "step": 53848 + }, + { + "epoch": 3.98, + "learning_rate": 5.3828971590290966e-06, + "loss": 1.0041, + "step": 53849 + }, + { + "epoch": 3.98, + "learning_rate": 5.382543351295837e-06, + "loss": 1.0521, + "step": 53850 + }, + { + "epoch": 3.98, + "learning_rate": 5.382189550908887e-06, + "loss": 0.9939, + "step": 53851 + }, + { + "epoch": 3.98, + "learning_rate": 5.3818357578688075e-06, + "loss": 0.9834, + "step": 53852 + }, + { + "epoch": 3.98, + "learning_rate": 5.381481972176161e-06, + "loss": 0.9421, + "step": 53853 + }, + { + "epoch": 3.98, + "learning_rate": 5.381128193831508e-06, + "loss": 0.9245, + "step": 53854 + }, + { + "epoch": 3.98, + "learning_rate": 5.380774422835417e-06, + "loss": 0.9446, + "step": 53855 + }, + { + "epoch": 3.98, + "learning_rate": 5.380420659188448e-06, + "loss": 0.9753, + "step": 53856 + }, + { + "epoch": 3.98, + "learning_rate": 5.380066902891163e-06, + "loss": 0.9694, + "step": 53857 + }, + { + "epoch": 3.98, + "learning_rate": 5.379713153944126e-06, + "loss": 0.9688, + "step": 53858 + }, + { + "epoch": 3.98, + "learning_rate": 5.379359412347895e-06, + "loss": 1.1281, + "step": 53859 + }, + { + "epoch": 3.98, + "learning_rate": 5.379005678103041e-06, + "loss": 1.0303, + "step": 53860 + }, + { + "epoch": 3.98, + "learning_rate": 5.378651951210123e-06, + "loss": 1.0113, + "step": 53861 + }, + { + "epoch": 3.98, + "learning_rate": 5.378298231669703e-06, + "loss": 1.0396, + "step": 53862 + }, + { + "epoch": 3.98, + "learning_rate": 5.377944519482342e-06, + "loss": 1.1048, + "step": 53863 + }, + { + "epoch": 3.98, + "learning_rate": 5.377590814648608e-06, + "loss": 1.0788, + "step": 53864 + }, + { + "epoch": 3.98, + "learning_rate": 5.377237117169061e-06, + "loss": 0.9814, + "step": 53865 + }, + { + "epoch": 3.98, + "learning_rate": 5.376883427044264e-06, + "loss": 0.9396, + "step": 53866 + }, + { + "epoch": 3.98, + "learning_rate": 5.376529744274779e-06, + "loss": 0.9241, + "step": 53867 + }, + { + "epoch": 3.98, + "learning_rate": 5.376176068861165e-06, + "loss": 1.0763, + "step": 53868 + }, + { + "epoch": 3.98, + "learning_rate": 5.375822400803996e-06, + "loss": 0.9279, + "step": 53869 + }, + { + "epoch": 3.98, + "learning_rate": 5.375468740103821e-06, + "loss": 0.9584, + "step": 53870 + }, + { + "epoch": 3.98, + "learning_rate": 5.375115086761212e-06, + "loss": 0.8651, + "step": 53871 + }, + { + "epoch": 3.98, + "learning_rate": 5.3747614407767255e-06, + "loss": 1.1225, + "step": 53872 + }, + { + "epoch": 3.98, + "learning_rate": 5.3744078021509335e-06, + "loss": 0.9596, + "step": 53873 + }, + { + "epoch": 3.98, + "learning_rate": 5.374054170884387e-06, + "loss": 0.9946, + "step": 53874 + }, + { + "epoch": 3.98, + "learning_rate": 5.3737005469776556e-06, + "loss": 0.9845, + "step": 53875 + }, + { + "epoch": 3.98, + "learning_rate": 5.373346930431302e-06, + "loss": 0.8951, + "step": 53876 + }, + { + "epoch": 3.98, + "learning_rate": 5.372993321245888e-06, + "loss": 1.0261, + "step": 53877 + }, + { + "epoch": 3.98, + "learning_rate": 5.372639719421975e-06, + "loss": 0.9686, + "step": 53878 + }, + { + "epoch": 3.98, + "learning_rate": 5.3722861249601214e-06, + "loss": 0.9814, + "step": 53879 + }, + { + "epoch": 3.98, + "learning_rate": 5.371932537860898e-06, + "loss": 1.0404, + "step": 53880 + }, + { + "epoch": 3.98, + "learning_rate": 5.371578958124863e-06, + "loss": 1.0542, + "step": 53881 + }, + { + "epoch": 3.98, + "learning_rate": 5.371225385752582e-06, + "loss": 0.8923, + "step": 53882 + }, + { + "epoch": 3.98, + "learning_rate": 5.37087182074461e-06, + "loss": 1.0203, + "step": 53883 + }, + { + "epoch": 3.98, + "learning_rate": 5.370518263101518e-06, + "loss": 0.8741, + "step": 53884 + }, + { + "epoch": 3.98, + "learning_rate": 5.370164712823866e-06, + "loss": 1.1564, + "step": 53885 + }, + { + "epoch": 3.98, + "learning_rate": 5.3698111699122155e-06, + "loss": 1.0978, + "step": 53886 + }, + { + "epoch": 3.98, + "learning_rate": 5.36945763436713e-06, + "loss": 0.9232, + "step": 53887 + }, + { + "epoch": 3.98, + "learning_rate": 5.369104106189167e-06, + "loss": 0.9595, + "step": 53888 + }, + { + "epoch": 3.98, + "learning_rate": 5.368750585378896e-06, + "loss": 0.851, + "step": 53889 + }, + { + "epoch": 3.98, + "learning_rate": 5.368397071936877e-06, + "loss": 0.958, + "step": 53890 + }, + { + "epoch": 3.98, + "learning_rate": 5.368043565863672e-06, + "loss": 0.9358, + "step": 53891 + }, + { + "epoch": 3.98, + "learning_rate": 5.36769006715984e-06, + "loss": 1.0592, + "step": 53892 + }, + { + "epoch": 3.98, + "learning_rate": 5.367336575825954e-06, + "loss": 0.94, + "step": 53893 + }, + { + "epoch": 3.98, + "learning_rate": 5.366983091862565e-06, + "loss": 0.9278, + "step": 53894 + }, + { + "epoch": 3.98, + "learning_rate": 5.366629615270236e-06, + "loss": 1.0522, + "step": 53895 + }, + { + "epoch": 3.98, + "learning_rate": 5.366276146049538e-06, + "loss": 1.0266, + "step": 53896 + }, + { + "epoch": 3.98, + "learning_rate": 5.365922684201027e-06, + "loss": 1.0, + "step": 53897 + }, + { + "epoch": 3.98, + "learning_rate": 5.365569229725267e-06, + "loss": 1.0278, + "step": 53898 + }, + { + "epoch": 3.98, + "learning_rate": 5.365215782622817e-06, + "loss": 1.0957, + "step": 53899 + }, + { + "epoch": 3.98, + "learning_rate": 5.364862342894246e-06, + "loss": 0.9118, + "step": 53900 + }, + { + "epoch": 3.98, + "learning_rate": 5.364508910540112e-06, + "loss": 0.9893, + "step": 53901 + }, + { + "epoch": 3.98, + "learning_rate": 5.364155485560978e-06, + "loss": 0.9328, + "step": 53902 + }, + { + "epoch": 3.98, + "learning_rate": 5.363802067957402e-06, + "loss": 0.9396, + "step": 53903 + }, + { + "epoch": 3.98, + "learning_rate": 5.363448657729956e-06, + "loss": 0.9385, + "step": 53904 + }, + { + "epoch": 3.98, + "learning_rate": 5.363095254879196e-06, + "loss": 1.0353, + "step": 53905 + }, + { + "epoch": 3.98, + "learning_rate": 5.3627418594056846e-06, + "loss": 0.9301, + "step": 53906 + }, + { + "epoch": 3.98, + "learning_rate": 5.362388471309985e-06, + "loss": 0.9269, + "step": 53907 + }, + { + "epoch": 3.98, + "learning_rate": 5.362035090592655e-06, + "loss": 0.9981, + "step": 53908 + }, + { + "epoch": 3.98, + "learning_rate": 5.361681717254266e-06, + "loss": 1.1122, + "step": 53909 + }, + { + "epoch": 3.98, + "learning_rate": 5.361328351295374e-06, + "loss": 1.0911, + "step": 53910 + }, + { + "epoch": 3.98, + "learning_rate": 5.3609749927165415e-06, + "loss": 0.982, + "step": 53911 + }, + { + "epoch": 3.98, + "learning_rate": 5.360621641518329e-06, + "loss": 0.8967, + "step": 53912 + }, + { + "epoch": 3.98, + "learning_rate": 5.360268297701303e-06, + "loss": 0.9683, + "step": 53913 + }, + { + "epoch": 3.98, + "learning_rate": 5.35991496126603e-06, + "loss": 1.054, + "step": 53914 + }, + { + "epoch": 3.98, + "learning_rate": 5.359561632213056e-06, + "loss": 0.9808, + "step": 53915 + }, + { + "epoch": 3.98, + "learning_rate": 5.3592083105429584e-06, + "loss": 0.9749, + "step": 53916 + }, + { + "epoch": 3.98, + "learning_rate": 5.358854996256289e-06, + "loss": 1.0383, + "step": 53917 + }, + { + "epoch": 3.98, + "learning_rate": 5.358501689353624e-06, + "loss": 1.0525, + "step": 53918 + }, + { + "epoch": 3.98, + "learning_rate": 5.358148389835507e-06, + "loss": 1.0596, + "step": 53919 + }, + { + "epoch": 3.98, + "learning_rate": 5.357795097702514e-06, + "loss": 1.1183, + "step": 53920 + }, + { + "epoch": 3.98, + "learning_rate": 5.3574418129552e-06, + "loss": 0.9747, + "step": 53921 + }, + { + "epoch": 3.98, + "learning_rate": 5.357088535594136e-06, + "loss": 1.0848, + "step": 53922 + }, + { + "epoch": 3.98, + "learning_rate": 5.356735265619874e-06, + "loss": 0.9646, + "step": 53923 + }, + { + "epoch": 3.98, + "learning_rate": 5.356382003032976e-06, + "loss": 1.0698, + "step": 53924 + }, + { + "epoch": 3.98, + "learning_rate": 5.356028747834013e-06, + "loss": 0.9847, + "step": 53925 + }, + { + "epoch": 3.98, + "learning_rate": 5.35567550002354e-06, + "loss": 0.9628, + "step": 53926 + }, + { + "epoch": 3.98, + "learning_rate": 5.355322259602121e-06, + "loss": 1.0518, + "step": 53927 + }, + { + "epoch": 3.98, + "learning_rate": 5.354969026570314e-06, + "loss": 0.9621, + "step": 53928 + }, + { + "epoch": 3.98, + "learning_rate": 5.354615800928689e-06, + "loss": 0.9307, + "step": 53929 + }, + { + "epoch": 3.98, + "learning_rate": 5.354262582677804e-06, + "loss": 1.0501, + "step": 53930 + }, + { + "epoch": 3.98, + "learning_rate": 5.35390937181822e-06, + "loss": 1.0561, + "step": 53931 + }, + { + "epoch": 3.99, + "learning_rate": 5.3535561683505e-06, + "loss": 1.0541, + "step": 53932 + }, + { + "epoch": 3.99, + "learning_rate": 5.353202972275202e-06, + "loss": 0.9411, + "step": 53933 + }, + { + "epoch": 3.99, + "learning_rate": 5.3528497835928965e-06, + "loss": 0.9536, + "step": 53934 + }, + { + "epoch": 3.99, + "learning_rate": 5.352496602304139e-06, + "loss": 1.0002, + "step": 53935 + }, + { + "epoch": 3.99, + "learning_rate": 5.352143428409493e-06, + "loss": 1.0257, + "step": 53936 + }, + { + "epoch": 3.99, + "learning_rate": 5.351790261909517e-06, + "loss": 1.07, + "step": 53937 + }, + { + "epoch": 3.99, + "learning_rate": 5.351437102804784e-06, + "loss": 1.0015, + "step": 53938 + }, + { + "epoch": 3.99, + "learning_rate": 5.351083951095841e-06, + "loss": 0.9274, + "step": 53939 + }, + { + "epoch": 3.99, + "learning_rate": 5.3507308067832594e-06, + "loss": 0.9125, + "step": 53940 + }, + { + "epoch": 3.99, + "learning_rate": 5.350377669867599e-06, + "loss": 1.0019, + "step": 53941 + }, + { + "epoch": 3.99, + "learning_rate": 5.350024540349421e-06, + "loss": 0.9664, + "step": 53942 + }, + { + "epoch": 3.99, + "learning_rate": 5.349671418229288e-06, + "loss": 0.972, + "step": 53943 + }, + { + "epoch": 3.99, + "learning_rate": 5.349318303507758e-06, + "loss": 1.086, + "step": 53944 + }, + { + "epoch": 3.99, + "learning_rate": 5.348965196185398e-06, + "loss": 1.0927, + "step": 53945 + }, + { + "epoch": 3.99, + "learning_rate": 5.348612096262769e-06, + "loss": 0.9181, + "step": 53946 + }, + { + "epoch": 3.99, + "learning_rate": 5.348259003740431e-06, + "loss": 0.9522, + "step": 53947 + }, + { + "epoch": 3.99, + "learning_rate": 5.347905918618943e-06, + "loss": 1.0427, + "step": 53948 + }, + { + "epoch": 3.99, + "learning_rate": 5.347552840898874e-06, + "loss": 1.0255, + "step": 53949 + }, + { + "epoch": 3.99, + "learning_rate": 5.347199770580782e-06, + "loss": 0.9945, + "step": 53950 + }, + { + "epoch": 3.99, + "learning_rate": 5.346846707665229e-06, + "loss": 1.0219, + "step": 53951 + }, + { + "epoch": 3.99, + "learning_rate": 5.346493652152776e-06, + "loss": 1.044, + "step": 53952 + }, + { + "epoch": 3.99, + "learning_rate": 5.34614060404398e-06, + "loss": 0.9587, + "step": 53953 + }, + { + "epoch": 3.99, + "learning_rate": 5.345787563339413e-06, + "loss": 1.0176, + "step": 53954 + }, + { + "epoch": 3.99, + "learning_rate": 5.34543453003963e-06, + "loss": 1.0049, + "step": 53955 + }, + { + "epoch": 3.99, + "learning_rate": 5.3450815041451955e-06, + "loss": 0.9806, + "step": 53956 + }, + { + "epoch": 3.99, + "learning_rate": 5.344728485656665e-06, + "loss": 0.9537, + "step": 53957 + }, + { + "epoch": 3.99, + "learning_rate": 5.344375474574609e-06, + "loss": 0.9365, + "step": 53958 + }, + { + "epoch": 3.99, + "learning_rate": 5.344022470899586e-06, + "loss": 1.0131, + "step": 53959 + }, + { + "epoch": 3.99, + "learning_rate": 5.343669474632155e-06, + "loss": 0.9419, + "step": 53960 + }, + { + "epoch": 3.99, + "learning_rate": 5.34331648577288e-06, + "loss": 1.117, + "step": 53961 + }, + { + "epoch": 3.99, + "learning_rate": 5.342963504322317e-06, + "loss": 1.064, + "step": 53962 + }, + { + "epoch": 3.99, + "learning_rate": 5.3426105302810415e-06, + "loss": 0.9842, + "step": 53963 + }, + { + "epoch": 3.99, + "learning_rate": 5.342257563649598e-06, + "loss": 1.0721, + "step": 53964 + }, + { + "epoch": 3.99, + "learning_rate": 5.3419046044285595e-06, + "loss": 1.0579, + "step": 53965 + }, + { + "epoch": 3.99, + "learning_rate": 5.34155165261848e-06, + "loss": 0.9589, + "step": 53966 + }, + { + "epoch": 3.99, + "learning_rate": 5.341198708219934e-06, + "loss": 0.9753, + "step": 53967 + }, + { + "epoch": 3.99, + "learning_rate": 5.340845771233465e-06, + "loss": 1.0081, + "step": 53968 + }, + { + "epoch": 3.99, + "learning_rate": 5.3404928416596495e-06, + "loss": 0.9492, + "step": 53969 + }, + { + "epoch": 3.99, + "learning_rate": 5.340139919499041e-06, + "loss": 0.9304, + "step": 53970 + }, + { + "epoch": 3.99, + "learning_rate": 5.339787004752205e-06, + "loss": 1.0116, + "step": 53971 + }, + { + "epoch": 3.99, + "learning_rate": 5.3394340974197e-06, + "loss": 0.919, + "step": 53972 + }, + { + "epoch": 3.99, + "learning_rate": 5.339081197502086e-06, + "loss": 1.0701, + "step": 53973 + }, + { + "epoch": 3.99, + "learning_rate": 5.3387283049999304e-06, + "loss": 0.9813, + "step": 53974 + }, + { + "epoch": 3.99, + "learning_rate": 5.338375419913791e-06, + "loss": 1.0561, + "step": 53975 + }, + { + "epoch": 3.99, + "learning_rate": 5.33802254224423e-06, + "loss": 1.0197, + "step": 53976 + }, + { + "epoch": 3.99, + "learning_rate": 5.337669671991806e-06, + "loss": 1.0109, + "step": 53977 + }, + { + "epoch": 3.99, + "learning_rate": 5.337316809157085e-06, + "loss": 0.9435, + "step": 53978 + }, + { + "epoch": 3.99, + "learning_rate": 5.336963953740626e-06, + "loss": 1.1342, + "step": 53979 + }, + { + "epoch": 3.99, + "learning_rate": 5.33661110574299e-06, + "loss": 0.9886, + "step": 53980 + }, + { + "epoch": 3.99, + "learning_rate": 5.336258265164741e-06, + "loss": 1.0135, + "step": 53981 + }, + { + "epoch": 3.99, + "learning_rate": 5.335905432006434e-06, + "loss": 1.1032, + "step": 53982 + }, + { + "epoch": 3.99, + "learning_rate": 5.335552606268641e-06, + "loss": 1.058, + "step": 53983 + }, + { + "epoch": 3.99, + "learning_rate": 5.335199787951911e-06, + "loss": 1.0535, + "step": 53984 + }, + { + "epoch": 3.99, + "learning_rate": 5.334846977056815e-06, + "loss": 1.0304, + "step": 53985 + }, + { + "epoch": 3.99, + "learning_rate": 5.334494173583906e-06, + "loss": 1.0301, + "step": 53986 + }, + { + "epoch": 3.99, + "learning_rate": 5.3341413775337595e-06, + "loss": 0.959, + "step": 53987 + }, + { + "epoch": 3.99, + "learning_rate": 5.333788588906921e-06, + "loss": 0.9834, + "step": 53988 + }, + { + "epoch": 3.99, + "learning_rate": 5.333435807703956e-06, + "loss": 0.9647, + "step": 53989 + }, + { + "epoch": 3.99, + "learning_rate": 5.3330830339254305e-06, + "loss": 1.1085, + "step": 53990 + }, + { + "epoch": 3.99, + "learning_rate": 5.3327302675719035e-06, + "loss": 0.9522, + "step": 53991 + }, + { + "epoch": 3.99, + "learning_rate": 5.3323775086439374e-06, + "loss": 0.9964, + "step": 53992 + }, + { + "epoch": 3.99, + "learning_rate": 5.332024757142087e-06, + "loss": 0.9911, + "step": 53993 + }, + { + "epoch": 3.99, + "learning_rate": 5.331672013066922e-06, + "loss": 1.0441, + "step": 53994 + }, + { + "epoch": 3.99, + "learning_rate": 5.331319276419e-06, + "loss": 1.0067, + "step": 53995 + }, + { + "epoch": 3.99, + "learning_rate": 5.330966547198882e-06, + "loss": 1.0493, + "step": 53996 + }, + { + "epoch": 3.99, + "learning_rate": 5.33061382540713e-06, + "loss": 0.9967, + "step": 53997 + }, + { + "epoch": 3.99, + "learning_rate": 5.3302611110443e-06, + "loss": 1.0118, + "step": 53998 + }, + { + "epoch": 3.99, + "learning_rate": 5.329908404110961e-06, + "loss": 1.0152, + "step": 53999 + }, + { + "epoch": 3.99, + "learning_rate": 5.329555704607674e-06, + "loss": 1.0734, + "step": 54000 + }, + { + "epoch": 3.99, + "learning_rate": 5.3292030125349945e-06, + "loss": 0.929, + "step": 54001 + }, + { + "epoch": 3.99, + "learning_rate": 5.328850327893482e-06, + "loss": 1.0489, + "step": 54002 + }, + { + "epoch": 3.99, + "learning_rate": 5.3284976506837085e-06, + "loss": 0.9138, + "step": 54003 + }, + { + "epoch": 3.99, + "learning_rate": 5.328144980906226e-06, + "loss": 1.0608, + "step": 54004 + }, + { + "epoch": 3.99, + "learning_rate": 5.327792318561598e-06, + "loss": 0.9725, + "step": 54005 + }, + { + "epoch": 3.99, + "learning_rate": 5.327439663650382e-06, + "loss": 0.9726, + "step": 54006 + }, + { + "epoch": 3.99, + "learning_rate": 5.327087016173147e-06, + "loss": 1.0213, + "step": 54007 + }, + { + "epoch": 3.99, + "learning_rate": 5.326734376130454e-06, + "loss": 1.0163, + "step": 54008 + }, + { + "epoch": 3.99, + "learning_rate": 5.3263817435228505e-06, + "loss": 1.0579, + "step": 54009 + }, + { + "epoch": 3.99, + "learning_rate": 5.326029118350911e-06, + "loss": 1.0741, + "step": 54010 + }, + { + "epoch": 3.99, + "learning_rate": 5.325676500615189e-06, + "loss": 0.9307, + "step": 54011 + }, + { + "epoch": 3.99, + "learning_rate": 5.325323890316256e-06, + "loss": 0.9849, + "step": 54012 + }, + { + "epoch": 3.99, + "learning_rate": 5.324971287454659e-06, + "loss": 1.0099, + "step": 54013 + }, + { + "epoch": 3.99, + "learning_rate": 5.3246186920309695e-06, + "loss": 0.9358, + "step": 54014 + }, + { + "epoch": 3.99, + "learning_rate": 5.324266104045743e-06, + "loss": 1.0651, + "step": 54015 + }, + { + "epoch": 3.99, + "learning_rate": 5.3239135234995445e-06, + "loss": 1.0034, + "step": 54016 + }, + { + "epoch": 3.99, + "learning_rate": 5.323560950392932e-06, + "loss": 0.9527, + "step": 54017 + }, + { + "epoch": 3.99, + "learning_rate": 5.323208384726463e-06, + "loss": 0.9857, + "step": 54018 + }, + { + "epoch": 3.99, + "learning_rate": 5.322855826500706e-06, + "loss": 0.9541, + "step": 54019 + }, + { + "epoch": 3.99, + "learning_rate": 5.32250327571622e-06, + "loss": 1.0206, + "step": 54020 + }, + { + "epoch": 3.99, + "learning_rate": 5.322150732373564e-06, + "loss": 0.9961, + "step": 54021 + }, + { + "epoch": 3.99, + "learning_rate": 5.3217981964732935e-06, + "loss": 0.9928, + "step": 54022 + }, + { + "epoch": 3.99, + "learning_rate": 5.32144566801598e-06, + "loss": 1.0682, + "step": 54023 + }, + { + "epoch": 3.99, + "learning_rate": 5.32109314700218e-06, + "loss": 1.1177, + "step": 54024 + }, + { + "epoch": 3.99, + "learning_rate": 5.320740633432453e-06, + "loss": 1.0302, + "step": 54025 + }, + { + "epoch": 3.99, + "learning_rate": 5.320388127307361e-06, + "loss": 1.0343, + "step": 54026 + }, + { + "epoch": 3.99, + "learning_rate": 5.320035628627462e-06, + "loss": 1.0269, + "step": 54027 + }, + { + "epoch": 3.99, + "learning_rate": 5.319683137393326e-06, + "loss": 0.8931, + "step": 54028 + }, + { + "epoch": 3.99, + "learning_rate": 5.319330653605499e-06, + "loss": 1.0518, + "step": 54029 + }, + { + "epoch": 3.99, + "learning_rate": 5.318978177264553e-06, + "loss": 1.033, + "step": 54030 + }, + { + "epoch": 3.99, + "learning_rate": 5.3186257083710435e-06, + "loss": 1.0931, + "step": 54031 + }, + { + "epoch": 3.99, + "learning_rate": 5.318273246925542e-06, + "loss": 1.0025, + "step": 54032 + }, + { + "epoch": 3.99, + "learning_rate": 5.317920792928592e-06, + "loss": 0.9127, + "step": 54033 + }, + { + "epoch": 3.99, + "learning_rate": 5.317568346380766e-06, + "loss": 1.0596, + "step": 54034 + }, + { + "epoch": 3.99, + "learning_rate": 5.317215907282623e-06, + "loss": 0.946, + "step": 54035 + }, + { + "epoch": 3.99, + "learning_rate": 5.316863475634721e-06, + "loss": 0.9603, + "step": 54036 + }, + { + "epoch": 3.99, + "learning_rate": 5.316511051437624e-06, + "loss": 0.9588, + "step": 54037 + }, + { + "epoch": 3.99, + "learning_rate": 5.316158634691886e-06, + "loss": 1.0226, + "step": 54038 + }, + { + "epoch": 3.99, + "learning_rate": 5.315806225398077e-06, + "loss": 0.9871, + "step": 54039 + }, + { + "epoch": 3.99, + "learning_rate": 5.315453823556753e-06, + "loss": 1.147, + "step": 54040 + }, + { + "epoch": 3.99, + "learning_rate": 5.315101429168476e-06, + "loss": 0.9767, + "step": 54041 + }, + { + "epoch": 3.99, + "learning_rate": 5.3147490422338e-06, + "loss": 0.8982, + "step": 54042 + }, + { + "epoch": 3.99, + "learning_rate": 5.314396662753296e-06, + "loss": 1.0, + "step": 54043 + }, + { + "epoch": 3.99, + "learning_rate": 5.314044290727521e-06, + "loss": 1.04, + "step": 54044 + }, + { + "epoch": 3.99, + "learning_rate": 5.313691926157032e-06, + "loss": 1.0277, + "step": 54045 + }, + { + "epoch": 3.99, + "learning_rate": 5.313339569042393e-06, + "loss": 1.0606, + "step": 54046 + }, + { + "epoch": 3.99, + "learning_rate": 5.31298721938416e-06, + "loss": 0.8995, + "step": 54047 + }, + { + "epoch": 3.99, + "learning_rate": 5.3126348771829015e-06, + "loss": 0.8601, + "step": 54048 + }, + { + "epoch": 3.99, + "learning_rate": 5.312282542439173e-06, + "loss": 0.9626, + "step": 54049 + }, + { + "epoch": 3.99, + "learning_rate": 5.311930215153538e-06, + "loss": 0.943, + "step": 54050 + }, + { + "epoch": 3.99, + "learning_rate": 5.31157789532655e-06, + "loss": 1.083, + "step": 54051 + }, + { + "epoch": 3.99, + "learning_rate": 5.311225582958777e-06, + "loss": 0.9851, + "step": 54052 + }, + { + "epoch": 3.99, + "learning_rate": 5.310873278050781e-06, + "loss": 1.0228, + "step": 54053 + }, + { + "epoch": 3.99, + "learning_rate": 5.3105209806031125e-06, + "loss": 1.0648, + "step": 54054 + }, + { + "epoch": 3.99, + "learning_rate": 5.310168690616341e-06, + "loss": 1.0303, + "step": 54055 + }, + { + "epoch": 3.99, + "learning_rate": 5.309816408091019e-06, + "loss": 1.0369, + "step": 54056 + }, + { + "epoch": 3.99, + "learning_rate": 5.309464133027722e-06, + "loss": 0.9559, + "step": 54057 + }, + { + "epoch": 3.99, + "learning_rate": 5.309111865426992e-06, + "loss": 0.9459, + "step": 54058 + }, + { + "epoch": 3.99, + "learning_rate": 5.3087596052894005e-06, + "loss": 1.0289, + "step": 54059 + }, + { + "epoch": 3.99, + "learning_rate": 5.308407352615507e-06, + "loss": 1.0496, + "step": 54060 + }, + { + "epoch": 3.99, + "learning_rate": 5.308055107405869e-06, + "loss": 1.1611, + "step": 54061 + }, + { + "epoch": 3.99, + "learning_rate": 5.307702869661044e-06, + "loss": 0.9949, + "step": 54062 + }, + { + "epoch": 3.99, + "learning_rate": 5.307350639381602e-06, + "loss": 0.9453, + "step": 54063 + }, + { + "epoch": 3.99, + "learning_rate": 5.306998416568097e-06, + "loss": 0.8736, + "step": 54064 + }, + { + "epoch": 3.99, + "learning_rate": 5.30664620122109e-06, + "loss": 0.9595, + "step": 54065 + }, + { + "epoch": 3.99, + "learning_rate": 5.3062939933411426e-06, + "loss": 1.0526, + "step": 54066 + }, + { + "epoch": 3.99, + "learning_rate": 5.30594179292881e-06, + "loss": 0.9811, + "step": 54067 + }, + { + "epoch": 4.0, + "learning_rate": 5.305589599984662e-06, + "loss": 1.0066, + "step": 54068 + }, + { + "epoch": 4.0, + "learning_rate": 5.305237414509252e-06, + "loss": 1.0162, + "step": 54069 + }, + { + "epoch": 4.0, + "learning_rate": 5.3048852365031435e-06, + "loss": 1.0746, + "step": 54070 + }, + { + "epoch": 4.0, + "learning_rate": 5.30453306596689e-06, + "loss": 0.939, + "step": 54071 + }, + { + "epoch": 4.0, + "learning_rate": 5.304180902901063e-06, + "loss": 0.9336, + "step": 54072 + }, + { + "epoch": 4.0, + "learning_rate": 5.30382874730622e-06, + "loss": 1.0186, + "step": 54073 + }, + { + "epoch": 4.0, + "learning_rate": 5.3034765991829105e-06, + "loss": 0.9515, + "step": 54074 + }, + { + "epoch": 4.0, + "learning_rate": 5.303124458531705e-06, + "loss": 1.0067, + "step": 54075 + }, + { + "epoch": 4.0, + "learning_rate": 5.302772325353159e-06, + "loss": 0.9299, + "step": 54076 + }, + { + "epoch": 4.0, + "learning_rate": 5.302420199647843e-06, + "loss": 0.9802, + "step": 54077 + }, + { + "epoch": 4.0, + "learning_rate": 5.302068081416301e-06, + "loss": 0.9934, + "step": 54078 + }, + { + "epoch": 4.0, + "learning_rate": 5.301715970659105e-06, + "loss": 0.9593, + "step": 54079 + }, + { + "epoch": 4.0, + "learning_rate": 5.301363867376808e-06, + "loss": 1.0152, + "step": 54080 + }, + { + "epoch": 4.0, + "learning_rate": 5.301011771569981e-06, + "loss": 1.0471, + "step": 54081 + }, + { + "epoch": 4.0, + "learning_rate": 5.300659683239175e-06, + "loss": 0.9693, + "step": 54082 + }, + { + "epoch": 4.0, + "learning_rate": 5.300307602384946e-06, + "loss": 0.9743, + "step": 54083 + }, + { + "epoch": 4.0, + "learning_rate": 5.299955529007866e-06, + "loss": 0.9541, + "step": 54084 + }, + { + "epoch": 4.0, + "learning_rate": 5.299603463108489e-06, + "loss": 1.0817, + "step": 54085 + }, + { + "epoch": 4.0, + "learning_rate": 5.299251404687375e-06, + "loss": 0.9014, + "step": 54086 + }, + { + "epoch": 4.0, + "learning_rate": 5.298899353745081e-06, + "loss": 1.0743, + "step": 54087 + }, + { + "epoch": 4.0, + "learning_rate": 5.298547310282176e-06, + "loss": 0.914, + "step": 54088 + }, + { + "epoch": 4.0, + "learning_rate": 5.298195274299215e-06, + "loss": 1.0143, + "step": 54089 + }, + { + "epoch": 4.0, + "learning_rate": 5.297843245796757e-06, + "loss": 1.1035, + "step": 54090 + }, + { + "epoch": 4.0, + "learning_rate": 5.297491224775364e-06, + "loss": 1.0925, + "step": 54091 + }, + { + "epoch": 4.0, + "learning_rate": 5.29713921123559e-06, + "loss": 0.9486, + "step": 54092 + }, + { + "epoch": 4.0, + "learning_rate": 5.296787205178005e-06, + "loss": 1.0433, + "step": 54093 + }, + { + "epoch": 4.0, + "learning_rate": 5.2964352066031635e-06, + "loss": 0.949, + "step": 54094 + }, + { + "epoch": 4.0, + "learning_rate": 5.296083215511627e-06, + "loss": 0.994, + "step": 54095 + }, + { + "epoch": 4.0, + "learning_rate": 5.295731231903951e-06, + "loss": 0.9925, + "step": 54096 + }, + { + "epoch": 4.0, + "learning_rate": 5.295379255780703e-06, + "loss": 1.0024, + "step": 54097 + }, + { + "epoch": 4.0, + "learning_rate": 5.295027287142438e-06, + "loss": 1.0734, + "step": 54098 + }, + { + "epoch": 4.0, + "learning_rate": 5.2946753259897195e-06, + "loss": 1.1114, + "step": 54099 + }, + { + "epoch": 4.0, + "learning_rate": 5.294323372323105e-06, + "loss": 0.9638, + "step": 54100 + }, + { + "epoch": 4.0, + "learning_rate": 5.293971426143149e-06, + "loss": 0.9665, + "step": 54101 + }, + { + "epoch": 4.0, + "learning_rate": 5.2936194874504255e-06, + "loss": 1.0073, + "step": 54102 + }, + { + "epoch": 4.0, + "learning_rate": 5.293267556245478e-06, + "loss": 0.959, + "step": 54103 + }, + { + "epoch": 4.0, + "learning_rate": 5.29291563252888e-06, + "loss": 1.0462, + "step": 54104 + }, + { + "epoch": 4.0, + "learning_rate": 5.292563716301185e-06, + "loss": 1.057, + "step": 54105 + }, + { + "epoch": 4.0, + "learning_rate": 5.292211807562953e-06, + "loss": 1.0516, + "step": 54106 + }, + { + "epoch": 4.0, + "learning_rate": 5.291859906314742e-06, + "loss": 1.022, + "step": 54107 + }, + { + "epoch": 4.0, + "learning_rate": 5.291508012557118e-06, + "loss": 1.0709, + "step": 54108 + }, + { + "epoch": 4.0, + "learning_rate": 5.291156126290637e-06, + "loss": 1.0379, + "step": 54109 + }, + { + "epoch": 4.0, + "learning_rate": 5.290804247515859e-06, + "loss": 0.9328, + "step": 54110 + }, + { + "epoch": 4.0, + "learning_rate": 5.290452376233346e-06, + "loss": 0.9528, + "step": 54111 + }, + { + "epoch": 4.0, + "learning_rate": 5.29010051244365e-06, + "loss": 0.9595, + "step": 54112 + }, + { + "epoch": 4.0, + "learning_rate": 5.289748656147341e-06, + "loss": 1.1035, + "step": 54113 + }, + { + "epoch": 4.0, + "learning_rate": 5.289396807344975e-06, + "loss": 1.0263, + "step": 54114 + }, + { + "epoch": 4.0, + "learning_rate": 5.2890449660371114e-06, + "loss": 1.0053, + "step": 54115 + }, + { + "epoch": 4.0, + "learning_rate": 5.288693132224305e-06, + "loss": 1.0378, + "step": 54116 + }, + { + "epoch": 4.0, + "learning_rate": 5.288341305907124e-06, + "loss": 1.053, + "step": 54117 + }, + { + "epoch": 4.0, + "learning_rate": 5.287989487086126e-06, + "loss": 1.0164, + "step": 54118 + }, + { + "epoch": 4.0, + "learning_rate": 5.287637675761868e-06, + "loss": 1.0282, + "step": 54119 + }, + { + "epoch": 4.0, + "learning_rate": 5.287285871934913e-06, + "loss": 0.9322, + "step": 54120 + }, + { + "epoch": 4.0, + "learning_rate": 5.2869340756058126e-06, + "loss": 1.005, + "step": 54121 + }, + { + "epoch": 4.0, + "learning_rate": 5.286582286775142e-06, + "loss": 1.0053, + "step": 54122 + }, + { + "epoch": 4.0, + "learning_rate": 5.286230505443442e-06, + "loss": 1.1103, + "step": 54123 + }, + { + "epoch": 4.0, + "learning_rate": 5.285878731611287e-06, + "loss": 0.9481, + "step": 54124 + }, + { + "epoch": 4.0, + "learning_rate": 5.2855269652792286e-06, + "loss": 1.0162, + "step": 54125 + }, + { + "epoch": 4.0, + "learning_rate": 5.2851752064478355e-06, + "loss": 0.9839, + "step": 54126 + }, + { + "epoch": 4.0, + "learning_rate": 5.2848234551176536e-06, + "loss": 1.0314, + "step": 54127 + }, + { + "epoch": 4.0, + "learning_rate": 5.284471711289254e-06, + "loss": 0.9787, + "step": 54128 + }, + { + "epoch": 4.0, + "learning_rate": 5.284119974963193e-06, + "loss": 0.934, + "step": 54129 + }, + { + "epoch": 4.0, + "learning_rate": 5.283768246140029e-06, + "loss": 0.9305, + "step": 54130 + }, + { + "epoch": 4.0, + "learning_rate": 5.283416524820324e-06, + "loss": 1.0045, + "step": 54131 + }, + { + "epoch": 4.0, + "learning_rate": 5.28306481100463e-06, + "loss": 0.9894, + "step": 54132 + }, + { + "epoch": 4.0, + "learning_rate": 5.282713104693518e-06, + "loss": 1.0567, + "step": 54133 + }, + { + "epoch": 4.0, + "learning_rate": 5.282361405887539e-06, + "loss": 0.9378, + "step": 54134 + }, + { + "epoch": 4.0, + "learning_rate": 5.282009714587257e-06, + "loss": 1.0969, + "step": 54135 + }, + { + "epoch": 4.0, + "learning_rate": 5.281658030793227e-06, + "loss": 0.991, + "step": 54136 + }, + { + "epoch": 4.0, + "learning_rate": 5.281306354506015e-06, + "loss": 0.9519, + "step": 54137 + }, + { + "epoch": 4.0, + "learning_rate": 5.280954685726177e-06, + "loss": 0.9285, + "step": 54138 + }, + { + "epoch": 4.0, + "learning_rate": 5.280603024454274e-06, + "loss": 0.9064, + "step": 54139 + }, + { + "epoch": 4.0, + "learning_rate": 5.280251370690862e-06, + "loss": 1.0691, + "step": 54140 + }, + { + "epoch": 4.0, + "learning_rate": 5.279899724436499e-06, + "loss": 0.9348, + "step": 54141 + }, + { + "epoch": 4.0, + "learning_rate": 5.279548085691752e-06, + "loss": 1.0078, + "step": 54142 + }, + { + "epoch": 4.0, + "learning_rate": 5.279196454457177e-06, + "loss": 0.9446, + "step": 54143 + }, + { + "epoch": 4.0, + "learning_rate": 5.278844830733332e-06, + "loss": 0.9941, + "step": 54144 + }, + { + "epoch": 4.0, + "learning_rate": 5.278493214520774e-06, + "loss": 1.0202, + "step": 54145 + }, + { + "epoch": 4.0, + "learning_rate": 5.278141605820071e-06, + "loss": 1.0194, + "step": 54146 + }, + { + "epoch": 4.0, + "learning_rate": 5.2777900046317784e-06, + "loss": 0.9637, + "step": 54147 + }, + { + "epoch": 4.0, + "learning_rate": 5.277438410956448e-06, + "loss": 0.9827, + "step": 54148 + }, + { + "epoch": 4.0, + "learning_rate": 5.277086824794649e-06, + "loss": 0.9303, + "step": 54149 + }, + { + "epoch": 4.0, + "learning_rate": 5.276735246146938e-06, + "loss": 1.0674, + "step": 54150 + }, + { + "epoch": 4.0, + "learning_rate": 5.276383675013873e-06, + "loss": 1.0943, + "step": 54151 + }, + { + "epoch": 4.0, + "learning_rate": 5.27603211139601e-06, + "loss": 1.0371, + "step": 54152 + }, + { + "epoch": 4.0, + "learning_rate": 5.275680555293916e-06, + "loss": 1.0009, + "step": 54153 + }, + { + "epoch": 4.0, + "learning_rate": 5.275329006708147e-06, + "loss": 1.0133, + "step": 54154 + }, + { + "epoch": 4.0, + "learning_rate": 5.274977465639264e-06, + "loss": 1.023, + "step": 54155 + }, + { + "epoch": 4.0, + "learning_rate": 5.274625932087821e-06, + "loss": 0.9454, + "step": 54156 + }, + { + "epoch": 4.0, + "learning_rate": 5.274274406054378e-06, + "loss": 1.0052, + "step": 54157 + }, + { + "epoch": 4.0, + "learning_rate": 5.273922887539501e-06, + "loss": 0.8881, + "step": 54158 + }, + { + "epoch": 4.0, + "learning_rate": 5.2735713765437445e-06, + "loss": 0.8924, + "step": 54159 + }, + { + "epoch": 4.0, + "learning_rate": 5.273219873067669e-06, + "loss": 0.9191, + "step": 54160 + }, + { + "epoch": 4.0, + "learning_rate": 5.272868377111829e-06, + "loss": 1.012, + "step": 54161 + }, + { + "epoch": 4.0, + "learning_rate": 5.2725168886767905e-06, + "loss": 0.9527, + "step": 54162 + }, + { + "epoch": 4.0, + "learning_rate": 5.272165407763111e-06, + "loss": 0.9837, + "step": 54163 + }, + { + "epoch": 4.0, + "learning_rate": 5.271813934371348e-06, + "loss": 1.0071, + "step": 54164 + }, + { + "epoch": 4.0, + "learning_rate": 5.2714624685020585e-06, + "loss": 0.9795, + "step": 54165 + }, + { + "epoch": 4.0, + "learning_rate": 5.2711110101558085e-06, + "loss": 1.1022, + "step": 54166 + }, + { + "epoch": 4.0, + "learning_rate": 5.270759559333156e-06, + "loss": 1.0297, + "step": 54167 + }, + { + "epoch": 4.0, + "learning_rate": 5.270408116034651e-06, + "loss": 1.0421, + "step": 54168 + }, + { + "epoch": 4.0, + "learning_rate": 5.270056680260862e-06, + "loss": 1.0448, + "step": 54169 + }, + { + "epoch": 4.0, + "learning_rate": 5.26970525201234e-06, + "loss": 1.0423, + "step": 54170 + }, + { + "epoch": 4.0, + "learning_rate": 5.26935383128966e-06, + "loss": 0.9884, + "step": 54171 + }, + { + "epoch": 4.0, + "learning_rate": 5.269002418093361e-06, + "loss": 0.8977, + "step": 54172 + }, + { + "epoch": 4.0, + "learning_rate": 5.268651012424015e-06, + "loss": 0.9789, + "step": 54173 + }, + { + "epoch": 4.0, + "learning_rate": 5.268299614282174e-06, + "loss": 0.9809, + "step": 54174 + }, + { + "epoch": 4.0, + "learning_rate": 5.267948223668409e-06, + "loss": 0.998, + "step": 54175 + }, + { + "epoch": 4.0, + "learning_rate": 5.267596840583266e-06, + "loss": 1.0426, + "step": 54176 + }, + { + "epoch": 4.0, + "learning_rate": 5.267245465027304e-06, + "loss": 0.9972, + "step": 54177 + }, + { + "epoch": 4.0, + "learning_rate": 5.266894097001091e-06, + "loss": 1.0134, + "step": 54178 + }, + { + "epoch": 4.0, + "learning_rate": 5.2665427365051835e-06, + "loss": 1.0266, + "step": 54179 + }, + { + "epoch": 4.0, + "learning_rate": 5.266191383540138e-06, + "loss": 1.0391, + "step": 54180 + }, + { + "epoch": 4.0, + "learning_rate": 5.26584003810651e-06, + "loss": 1.0647, + "step": 54181 + }, + { + "epoch": 4.0, + "learning_rate": 5.265488700204866e-06, + "loss": 1.1475, + "step": 54182 + }, + { + "epoch": 4.0, + "learning_rate": 5.265137369835763e-06, + "loss": 1.0399, + "step": 54183 + }, + { + "epoch": 4.0, + "learning_rate": 5.264786046999759e-06, + "loss": 0.9707, + "step": 54184 + }, + { + "epoch": 4.0, + "learning_rate": 5.264434731697411e-06, + "loss": 0.9505, + "step": 54185 + }, + { + "epoch": 4.0, + "learning_rate": 5.264083423929276e-06, + "loss": 0.9218, + "step": 54186 + }, + { + "epoch": 4.0, + "learning_rate": 5.263732123695921e-06, + "loss": 0.9211, + "step": 54187 + }, + { + "epoch": 4.0, + "learning_rate": 5.263380830997902e-06, + "loss": 0.9606, + "step": 54188 + }, + { + "epoch": 4.0, + "learning_rate": 5.2630295458357735e-06, + "loss": 0.9552, + "step": 54189 + }, + { + "epoch": 4.0, + "learning_rate": 5.2626782682100955e-06, + "loss": 0.9892, + "step": 54190 + }, + { + "epoch": 4.0, + "learning_rate": 5.262326998121431e-06, + "loss": 1.0699, + "step": 54191 + }, + { + "epoch": 4.0, + "learning_rate": 5.261975735570337e-06, + "loss": 0.9309, + "step": 54192 + }, + { + "epoch": 4.0, + "learning_rate": 5.261624480557371e-06, + "loss": 0.9881, + "step": 54193 + }, + { + "epoch": 4.0, + "learning_rate": 5.261273233083094e-06, + "loss": 0.9039, + "step": 54194 + }, + { + "epoch": 4.0, + "learning_rate": 5.2609219931480626e-06, + "loss": 1.0628, + "step": 54195 + }, + { + "epoch": 4.0, + "learning_rate": 5.260570760752837e-06, + "loss": 0.9829, + "step": 54196 + }, + { + "epoch": 4.0, + "learning_rate": 5.260219535897971e-06, + "loss": 1.0445, + "step": 54197 + }, + { + "epoch": 4.0, + "learning_rate": 5.2598683185840325e-06, + "loss": 1.031, + "step": 54198 + }, + { + "epoch": 4.0, + "learning_rate": 5.259517108811576e-06, + "loss": 0.9288, + "step": 54199 + }, + { + "epoch": 4.0, + "learning_rate": 5.259165906581159e-06, + "loss": 0.9537, + "step": 54200 + }, + { + "epoch": 4.0, + "learning_rate": 5.258814711893339e-06, + "loss": 1.0331, + "step": 54201 + }, + { + "epoch": 4.0, + "learning_rate": 5.258463524748678e-06, + "loss": 0.9455, + "step": 54202 + }, + { + "epoch": 4.01, + "learning_rate": 5.258112345147736e-06, + "loss": 1.0264, + "step": 54203 + }, + { + "epoch": 4.01, + "learning_rate": 5.257761173091068e-06, + "loss": 1.0236, + "step": 54204 + }, + { + "epoch": 4.01, + "learning_rate": 5.257410008579236e-06, + "loss": 1.02, + "step": 54205 + }, + { + "epoch": 4.01, + "learning_rate": 5.257058851612791e-06, + "loss": 1.0084, + "step": 54206 + }, + { + "epoch": 4.01, + "learning_rate": 5.256707702192302e-06, + "loss": 1.0621, + "step": 54207 + }, + { + "epoch": 4.01, + "learning_rate": 5.256356560318323e-06, + "loss": 0.984, + "step": 54208 + }, + { + "epoch": 4.01, + "learning_rate": 5.256005425991413e-06, + "loss": 1.0653, + "step": 54209 + }, + { + "epoch": 4.01, + "learning_rate": 5.255654299212127e-06, + "loss": 0.9889, + "step": 54210 + }, + { + "epoch": 4.01, + "learning_rate": 5.25530317998103e-06, + "loss": 0.9664, + "step": 54211 + }, + { + "epoch": 4.01, + "learning_rate": 5.254952068298683e-06, + "loss": 1.0893, + "step": 54212 + }, + { + "epoch": 4.01, + "learning_rate": 5.254600964165631e-06, + "loss": 1.0785, + "step": 54213 + }, + { + "epoch": 4.01, + "learning_rate": 5.254249867582445e-06, + "loss": 0.9185, + "step": 54214 + }, + { + "epoch": 4.01, + "learning_rate": 5.2538987785496755e-06, + "loss": 0.9992, + "step": 54215 + }, + { + "epoch": 4.01, + "learning_rate": 5.253547697067893e-06, + "loss": 1.0602, + "step": 54216 + }, + { + "epoch": 4.01, + "learning_rate": 5.2531966231376396e-06, + "loss": 0.981, + "step": 54217 + }, + { + "epoch": 4.01, + "learning_rate": 5.2528455567594874e-06, + "loss": 0.992, + "step": 54218 + }, + { + "epoch": 4.01, + "learning_rate": 5.252494497933987e-06, + "loss": 1.0871, + "step": 54219 + }, + { + "epoch": 4.01, + "learning_rate": 5.252143446661706e-06, + "loss": 0.969, + "step": 54220 + }, + { + "epoch": 4.01, + "learning_rate": 5.25179240294319e-06, + "loss": 0.9256, + "step": 54221 + }, + { + "epoch": 4.01, + "learning_rate": 5.251441366779007e-06, + "loss": 1.0403, + "step": 54222 + }, + { + "epoch": 4.01, + "learning_rate": 5.251090338169713e-06, + "loss": 1.191, + "step": 54223 + }, + { + "epoch": 4.01, + "learning_rate": 5.250739317115867e-06, + "loss": 1.0649, + "step": 54224 + }, + { + "epoch": 4.01, + "learning_rate": 5.2503883036180285e-06, + "loss": 1.0442, + "step": 54225 + }, + { + "epoch": 4.01, + "learning_rate": 5.250037297676748e-06, + "loss": 0.9857, + "step": 54226 + }, + { + "epoch": 4.01, + "learning_rate": 5.249686299292594e-06, + "loss": 0.9572, + "step": 54227 + }, + { + "epoch": 4.01, + "learning_rate": 5.249335308466123e-06, + "loss": 1.0142, + "step": 54228 + }, + { + "epoch": 4.01, + "learning_rate": 5.24898432519789e-06, + "loss": 1.0257, + "step": 54229 + }, + { + "epoch": 4.01, + "learning_rate": 5.2486333494884515e-06, + "loss": 1.1067, + "step": 54230 + }, + { + "epoch": 4.01, + "learning_rate": 5.248282381338373e-06, + "loss": 1.049, + "step": 54231 + }, + { + "epoch": 4.01, + "learning_rate": 5.247931420748209e-06, + "loss": 0.9686, + "step": 54232 + }, + { + "epoch": 4.01, + "learning_rate": 5.247580467718519e-06, + "loss": 0.997, + "step": 54233 + }, + { + "epoch": 4.01, + "learning_rate": 5.24722952224986e-06, + "loss": 0.9292, + "step": 54234 + }, + { + "epoch": 4.01, + "learning_rate": 5.246878584342787e-06, + "loss": 0.8709, + "step": 54235 + }, + { + "epoch": 4.01, + "learning_rate": 5.246527653997871e-06, + "loss": 0.9825, + "step": 54236 + }, + { + "epoch": 4.01, + "learning_rate": 5.246176731215652e-06, + "loss": 0.9227, + "step": 54237 + }, + { + "epoch": 4.01, + "learning_rate": 5.245825815996703e-06, + "loss": 1.0372, + "step": 54238 + }, + { + "epoch": 4.01, + "learning_rate": 5.245474908341572e-06, + "loss": 1.1086, + "step": 54239 + }, + { + "epoch": 4.01, + "learning_rate": 5.24512400825083e-06, + "loss": 1.0356, + "step": 54240 + }, + { + "epoch": 4.01, + "learning_rate": 5.244773115725026e-06, + "loss": 0.9884, + "step": 54241 + }, + { + "epoch": 4.01, + "learning_rate": 5.244422230764714e-06, + "loss": 1.0496, + "step": 54242 + }, + { + "epoch": 4.01, + "learning_rate": 5.2440713533704635e-06, + "loss": 0.965, + "step": 54243 + }, + { + "epoch": 4.01, + "learning_rate": 5.2437204835428255e-06, + "loss": 0.9767, + "step": 54244 + }, + { + "epoch": 4.01, + "learning_rate": 5.243369621282361e-06, + "loss": 0.8866, + "step": 54245 + }, + { + "epoch": 4.01, + "learning_rate": 5.243018766589625e-06, + "loss": 0.8621, + "step": 54246 + }, + { + "epoch": 4.01, + "learning_rate": 5.242667919465181e-06, + "loss": 0.9977, + "step": 54247 + }, + { + "epoch": 4.01, + "learning_rate": 5.2423170799095825e-06, + "loss": 1.0835, + "step": 54248 + }, + { + "epoch": 4.01, + "learning_rate": 5.241966247923391e-06, + "loss": 1.0033, + "step": 54249 + }, + { + "epoch": 4.01, + "learning_rate": 5.241615423507164e-06, + "loss": 0.9726, + "step": 54250 + }, + { + "epoch": 4.01, + "learning_rate": 5.241264606661454e-06, + "loss": 1.0253, + "step": 54251 + }, + { + "epoch": 4.01, + "learning_rate": 5.240913797386827e-06, + "loss": 1.0367, + "step": 54252 + }, + { + "epoch": 4.01, + "learning_rate": 5.24056299568384e-06, + "loss": 0.9437, + "step": 54253 + }, + { + "epoch": 4.01, + "learning_rate": 5.240212201553048e-06, + "loss": 0.893, + "step": 54254 + }, + { + "epoch": 4.01, + "learning_rate": 5.239861414995006e-06, + "loss": 1.1013, + "step": 54255 + }, + { + "epoch": 4.01, + "learning_rate": 5.239510636010281e-06, + "loss": 0.9655, + "step": 54256 + }, + { + "epoch": 4.01, + "learning_rate": 5.239159864599426e-06, + "loss": 0.9854, + "step": 54257 + }, + { + "epoch": 4.01, + "learning_rate": 5.238809100762999e-06, + "loss": 1.0067, + "step": 54258 + }, + { + "epoch": 4.01, + "learning_rate": 5.23845834450156e-06, + "loss": 1.1233, + "step": 54259 + }, + { + "epoch": 4.01, + "learning_rate": 5.23810759581566e-06, + "loss": 1.0555, + "step": 54260 + }, + { + "epoch": 4.01, + "learning_rate": 5.237756854705872e-06, + "loss": 0.997, + "step": 54261 + }, + { + "epoch": 4.01, + "learning_rate": 5.237406121172737e-06, + "loss": 0.971, + "step": 54262 + }, + { + "epoch": 4.01, + "learning_rate": 5.237055395216824e-06, + "loss": 0.9521, + "step": 54263 + }, + { + "epoch": 4.01, + "learning_rate": 5.236704676838684e-06, + "loss": 0.9781, + "step": 54264 + }, + { + "epoch": 4.01, + "learning_rate": 5.236353966038886e-06, + "loss": 0.987, + "step": 54265 + }, + { + "epoch": 4.01, + "learning_rate": 5.236003262817973e-06, + "loss": 1.0053, + "step": 54266 + }, + { + "epoch": 4.01, + "learning_rate": 5.235652567176515e-06, + "loss": 1.1161, + "step": 54267 + }, + { + "epoch": 4.01, + "learning_rate": 5.2353018791150654e-06, + "loss": 1.0721, + "step": 54268 + }, + { + "epoch": 4.01, + "learning_rate": 5.234951198634181e-06, + "loss": 1.0106, + "step": 54269 + }, + { + "epoch": 4.01, + "learning_rate": 5.234600525734423e-06, + "loss": 1.0291, + "step": 54270 + }, + { + "epoch": 4.01, + "learning_rate": 5.234249860416344e-06, + "loss": 1.0522, + "step": 54271 + }, + { + "epoch": 4.01, + "learning_rate": 5.233899202680508e-06, + "loss": 1.0347, + "step": 54272 + }, + { + "epoch": 4.01, + "learning_rate": 5.23354855252747e-06, + "loss": 1.0863, + "step": 54273 + }, + { + "epoch": 4.01, + "learning_rate": 5.233197909957788e-06, + "loss": 0.9414, + "step": 54274 + }, + { + "epoch": 4.01, + "learning_rate": 5.232847274972017e-06, + "loss": 0.9175, + "step": 54275 + }, + { + "epoch": 4.01, + "learning_rate": 5.23249664757072e-06, + "loss": 1.0068, + "step": 54276 + }, + { + "epoch": 4.01, + "learning_rate": 5.232146027754454e-06, + "loss": 1.0994, + "step": 54277 + }, + { + "epoch": 4.01, + "learning_rate": 5.231795415523775e-06, + "loss": 0.9346, + "step": 54278 + }, + { + "epoch": 4.01, + "learning_rate": 5.231444810879242e-06, + "loss": 1.0792, + "step": 54279 + }, + { + "epoch": 4.01, + "learning_rate": 5.231094213821408e-06, + "loss": 0.9459, + "step": 54280 + }, + { + "epoch": 4.01, + "learning_rate": 5.230743624350842e-06, + "loss": 1.0774, + "step": 54281 + }, + { + "epoch": 4.01, + "learning_rate": 5.230393042468087e-06, + "loss": 1.1366, + "step": 54282 + }, + { + "epoch": 4.01, + "learning_rate": 5.230042468173713e-06, + "loss": 0.9481, + "step": 54283 + }, + { + "epoch": 4.01, + "learning_rate": 5.229691901468268e-06, + "loss": 0.9337, + "step": 54284 + }, + { + "epoch": 4.01, + "learning_rate": 5.229341342352323e-06, + "loss": 0.9929, + "step": 54285 + }, + { + "epoch": 4.01, + "learning_rate": 5.2289907908264205e-06, + "loss": 0.9087, + "step": 54286 + }, + { + "epoch": 4.01, + "learning_rate": 5.228640246891129e-06, + "loss": 0.8645, + "step": 54287 + }, + { + "epoch": 4.01, + "learning_rate": 5.2282897105470034e-06, + "loss": 0.974, + "step": 54288 + }, + { + "epoch": 4.01, + "learning_rate": 5.227939181794599e-06, + "loss": 0.9293, + "step": 54289 + }, + { + "epoch": 4.01, + "learning_rate": 5.227588660634477e-06, + "loss": 0.9891, + "step": 54290 + }, + { + "epoch": 4.01, + "learning_rate": 5.2272381470671885e-06, + "loss": 0.962, + "step": 54291 + }, + { + "epoch": 4.01, + "learning_rate": 5.226887641093299e-06, + "loss": 1.0252, + "step": 54292 + }, + { + "epoch": 4.01, + "learning_rate": 5.226537142713364e-06, + "loss": 0.9838, + "step": 54293 + }, + { + "epoch": 4.01, + "learning_rate": 5.226186651927938e-06, + "loss": 0.9677, + "step": 54294 + }, + { + "epoch": 4.01, + "learning_rate": 5.225836168737579e-06, + "loss": 1.0085, + "step": 54295 + }, + { + "epoch": 4.01, + "learning_rate": 5.225485693142849e-06, + "loss": 0.9421, + "step": 54296 + }, + { + "epoch": 4.01, + "learning_rate": 5.225135225144304e-06, + "loss": 0.9958, + "step": 54297 + }, + { + "epoch": 4.01, + "learning_rate": 5.224784764742501e-06, + "loss": 1.0491, + "step": 54298 + }, + { + "epoch": 4.01, + "learning_rate": 5.224434311937996e-06, + "loss": 0.8239, + "step": 54299 + }, + { + "epoch": 4.01, + "learning_rate": 5.224083866731343e-06, + "loss": 0.9846, + "step": 54300 + }, + { + "epoch": 4.01, + "learning_rate": 5.22373342912311e-06, + "loss": 0.9156, + "step": 54301 + }, + { + "epoch": 4.01, + "learning_rate": 5.223382999113848e-06, + "loss": 1.105, + "step": 54302 + }, + { + "epoch": 4.01, + "learning_rate": 5.223032576704114e-06, + "loss": 1.004, + "step": 54303 + }, + { + "epoch": 4.01, + "learning_rate": 5.222682161894466e-06, + "loss": 0.9824, + "step": 54304 + }, + { + "epoch": 4.01, + "learning_rate": 5.222331754685464e-06, + "loss": 0.9014, + "step": 54305 + }, + { + "epoch": 4.01, + "learning_rate": 5.221981355077667e-06, + "loss": 1.0432, + "step": 54306 + }, + { + "epoch": 4.01, + "learning_rate": 5.2216309630716225e-06, + "loss": 1.0845, + "step": 54307 + }, + { + "epoch": 4.01, + "learning_rate": 5.2212805786679e-06, + "loss": 1.048, + "step": 54308 + }, + { + "epoch": 4.01, + "learning_rate": 5.220930201867046e-06, + "loss": 1.082, + "step": 54309 + }, + { + "epoch": 4.01, + "learning_rate": 5.220579832669632e-06, + "loss": 1.0705, + "step": 54310 + }, + { + "epoch": 4.01, + "learning_rate": 5.220229471076199e-06, + "loss": 1.0689, + "step": 54311 + }, + { + "epoch": 4.01, + "learning_rate": 5.2198791170873165e-06, + "loss": 1.0396, + "step": 54312 + }, + { + "epoch": 4.01, + "learning_rate": 5.219528770703535e-06, + "loss": 1.04, + "step": 54313 + }, + { + "epoch": 4.01, + "learning_rate": 5.219178431925423e-06, + "loss": 1.0654, + "step": 54314 + }, + { + "epoch": 4.01, + "learning_rate": 5.218828100753524e-06, + "loss": 0.9921, + "step": 54315 + }, + { + "epoch": 4.01, + "learning_rate": 5.218477777188397e-06, + "loss": 0.9671, + "step": 54316 + }, + { + "epoch": 4.01, + "learning_rate": 5.218127461230609e-06, + "loss": 0.925, + "step": 54317 + }, + { + "epoch": 4.01, + "learning_rate": 5.2177771528807105e-06, + "loss": 0.94, + "step": 54318 + }, + { + "epoch": 4.01, + "learning_rate": 5.217426852139261e-06, + "loss": 0.9576, + "step": 54319 + }, + { + "epoch": 4.01, + "learning_rate": 5.2170765590068125e-06, + "loss": 0.9649, + "step": 54320 + }, + { + "epoch": 4.01, + "learning_rate": 5.216726273483932e-06, + "loss": 0.9914, + "step": 54321 + }, + { + "epoch": 4.01, + "learning_rate": 5.21637599557117e-06, + "loss": 0.9153, + "step": 54322 + }, + { + "epoch": 4.01, + "learning_rate": 5.216025725269085e-06, + "loss": 0.878, + "step": 54323 + }, + { + "epoch": 4.01, + "learning_rate": 5.2156754625782355e-06, + "loss": 1.0301, + "step": 54324 + }, + { + "epoch": 4.01, + "learning_rate": 5.215325207499175e-06, + "loss": 0.9402, + "step": 54325 + }, + { + "epoch": 4.01, + "learning_rate": 5.214974960032471e-06, + "loss": 0.9715, + "step": 54326 + }, + { + "epoch": 4.01, + "learning_rate": 5.214624720178664e-06, + "loss": 1.0175, + "step": 54327 + }, + { + "epoch": 4.01, + "learning_rate": 5.214274487938326e-06, + "loss": 0.9047, + "step": 54328 + }, + { + "epoch": 4.01, + "learning_rate": 5.2139242633120035e-06, + "loss": 1.0018, + "step": 54329 + }, + { + "epoch": 4.01, + "learning_rate": 5.2135740463002685e-06, + "loss": 1.1705, + "step": 54330 + }, + { + "epoch": 4.01, + "learning_rate": 5.2132238369036605e-06, + "loss": 0.9146, + "step": 54331 + }, + { + "epoch": 4.01, + "learning_rate": 5.2128736351227485e-06, + "loss": 1.0093, + "step": 54332 + }, + { + "epoch": 4.01, + "learning_rate": 5.2125234409580815e-06, + "loss": 1.0168, + "step": 54333 + }, + { + "epoch": 4.01, + "learning_rate": 5.212173254410229e-06, + "loss": 0.9613, + "step": 54334 + }, + { + "epoch": 4.01, + "learning_rate": 5.211823075479737e-06, + "loss": 1.0262, + "step": 54335 + }, + { + "epoch": 4.01, + "learning_rate": 5.211472904167163e-06, + "loss": 1.0323, + "step": 54336 + }, + { + "epoch": 4.01, + "learning_rate": 5.2111227404730695e-06, + "loss": 1.0034, + "step": 54337 + }, + { + "epoch": 4.01, + "learning_rate": 5.210772584398012e-06, + "loss": 1.0649, + "step": 54338 + }, + { + "epoch": 4.02, + "learning_rate": 5.2104224359425455e-06, + "loss": 0.9324, + "step": 54339 + }, + { + "epoch": 4.02, + "learning_rate": 5.210072295107227e-06, + "loss": 1.0196, + "step": 54340 + }, + { + "epoch": 4.02, + "learning_rate": 5.2097221618926165e-06, + "loss": 1.096, + "step": 54341 + }, + { + "epoch": 4.02, + "learning_rate": 5.2093720362992705e-06, + "loss": 0.947, + "step": 54342 + }, + { + "epoch": 4.02, + "learning_rate": 5.2090219183277455e-06, + "loss": 0.9666, + "step": 54343 + }, + { + "epoch": 4.02, + "learning_rate": 5.208671807978597e-06, + "loss": 0.9699, + "step": 54344 + }, + { + "epoch": 4.02, + "learning_rate": 5.2083217052523794e-06, + "loss": 1.042, + "step": 54345 + }, + { + "epoch": 4.02, + "learning_rate": 5.207971610149658e-06, + "loss": 1.0149, + "step": 54346 + }, + { + "epoch": 4.02, + "learning_rate": 5.207621522670985e-06, + "loss": 0.9471, + "step": 54347 + }, + { + "epoch": 4.02, + "learning_rate": 5.207271442816917e-06, + "loss": 0.9869, + "step": 54348 + }, + { + "epoch": 4.02, + "learning_rate": 5.206921370588008e-06, + "loss": 0.9477, + "step": 54349 + }, + { + "epoch": 4.02, + "learning_rate": 5.206571305984822e-06, + "loss": 1.0287, + "step": 54350 + }, + { + "epoch": 4.02, + "learning_rate": 5.206221249007914e-06, + "loss": 0.9991, + "step": 54351 + }, + { + "epoch": 4.02, + "learning_rate": 5.205871199657838e-06, + "loss": 0.9493, + "step": 54352 + }, + { + "epoch": 4.02, + "learning_rate": 5.205521157935154e-06, + "loss": 1.0336, + "step": 54353 + }, + { + "epoch": 4.02, + "learning_rate": 5.205171123840411e-06, + "loss": 1.0066, + "step": 54354 + }, + { + "epoch": 4.02, + "learning_rate": 5.204821097374182e-06, + "loss": 0.9291, + "step": 54355 + }, + { + "epoch": 4.02, + "learning_rate": 5.204471078537005e-06, + "loss": 0.9245, + "step": 54356 + }, + { + "epoch": 4.02, + "learning_rate": 5.204121067329449e-06, + "loss": 1.0006, + "step": 54357 + }, + { + "epoch": 4.02, + "learning_rate": 5.20377106375207e-06, + "loss": 1.102, + "step": 54358 + }, + { + "epoch": 4.02, + "learning_rate": 5.203421067805421e-06, + "loss": 0.9584, + "step": 54359 + }, + { + "epoch": 4.02, + "learning_rate": 5.203071079490056e-06, + "loss": 0.9159, + "step": 54360 + }, + { + "epoch": 4.02, + "learning_rate": 5.2027210988065425e-06, + "loss": 1.0557, + "step": 54361 + }, + { + "epoch": 4.02, + "learning_rate": 5.20237112575543e-06, + "loss": 1.0379, + "step": 54362 + }, + { + "epoch": 4.02, + "learning_rate": 5.202021160337275e-06, + "loss": 1.0421, + "step": 54363 + }, + { + "epoch": 4.02, + "learning_rate": 5.201671202552637e-06, + "loss": 1.0028, + "step": 54364 + }, + { + "epoch": 4.02, + "learning_rate": 5.201321252402067e-06, + "loss": 0.9979, + "step": 54365 + }, + { + "epoch": 4.02, + "learning_rate": 5.200971309886132e-06, + "loss": 1.0349, + "step": 54366 + }, + { + "epoch": 4.02, + "learning_rate": 5.20062137500538e-06, + "loss": 1.0227, + "step": 54367 + }, + { + "epoch": 4.02, + "learning_rate": 5.2002714477603724e-06, + "loss": 0.8996, + "step": 54368 + }, + { + "epoch": 4.02, + "learning_rate": 5.199921528151661e-06, + "loss": 1.0217, + "step": 54369 + }, + { + "epoch": 4.02, + "learning_rate": 5.199571616179808e-06, + "loss": 0.976, + "step": 54370 + }, + { + "epoch": 4.02, + "learning_rate": 5.1992217118453726e-06, + "loss": 0.9584, + "step": 54371 + }, + { + "epoch": 4.02, + "learning_rate": 5.198871815148899e-06, + "loss": 0.9931, + "step": 54372 + }, + { + "epoch": 4.02, + "learning_rate": 5.198521926090956e-06, + "loss": 0.8223, + "step": 54373 + }, + { + "epoch": 4.02, + "learning_rate": 5.198172044672092e-06, + "loss": 0.9856, + "step": 54374 + }, + { + "epoch": 4.02, + "learning_rate": 5.1978221708928746e-06, + "loss": 1.0327, + "step": 54375 + }, + { + "epoch": 4.02, + "learning_rate": 5.197472304753846e-06, + "loss": 1.0493, + "step": 54376 + }, + { + "epoch": 4.02, + "learning_rate": 5.197122446255574e-06, + "loss": 1.0855, + "step": 54377 + }, + { + "epoch": 4.02, + "learning_rate": 5.196772595398607e-06, + "loss": 0.993, + "step": 54378 + }, + { + "epoch": 4.02, + "learning_rate": 5.196422752183515e-06, + "loss": 1.0494, + "step": 54379 + }, + { + "epoch": 4.02, + "learning_rate": 5.196072916610837e-06, + "loss": 0.9003, + "step": 54380 + }, + { + "epoch": 4.02, + "learning_rate": 5.195723088681144e-06, + "loss": 0.857, + "step": 54381 + }, + { + "epoch": 4.02, + "learning_rate": 5.195373268394985e-06, + "loss": 0.9111, + "step": 54382 + }, + { + "epoch": 4.02, + "learning_rate": 5.1950234557529186e-06, + "loss": 0.99, + "step": 54383 + }, + { + "epoch": 4.02, + "learning_rate": 5.194673650755502e-06, + "loss": 0.9735, + "step": 54384 + }, + { + "epoch": 4.02, + "learning_rate": 5.194323853403285e-06, + "loss": 0.9494, + "step": 54385 + }, + { + "epoch": 4.02, + "learning_rate": 5.193974063696836e-06, + "loss": 1.0344, + "step": 54386 + }, + { + "epoch": 4.02, + "learning_rate": 5.1936242816367045e-06, + "loss": 0.8792, + "step": 54387 + }, + { + "epoch": 4.02, + "learning_rate": 5.193274507223448e-06, + "loss": 1.0316, + "step": 54388 + }, + { + "epoch": 4.02, + "learning_rate": 5.192924740457619e-06, + "loss": 1.0079, + "step": 54389 + }, + { + "epoch": 4.02, + "learning_rate": 5.192574981339782e-06, + "loss": 0.983, + "step": 54390 + }, + { + "epoch": 4.02, + "learning_rate": 5.19222522987049e-06, + "loss": 0.958, + "step": 54391 + }, + { + "epoch": 4.02, + "learning_rate": 5.191875486050298e-06, + "loss": 1.0704, + "step": 54392 + }, + { + "epoch": 4.02, + "learning_rate": 5.191525749879764e-06, + "loss": 0.9729, + "step": 54393 + }, + { + "epoch": 4.02, + "learning_rate": 5.19117602135944e-06, + "loss": 1.0103, + "step": 54394 + }, + { + "epoch": 4.02, + "learning_rate": 5.19082630048989e-06, + "loss": 1.0869, + "step": 54395 + }, + { + "epoch": 4.02, + "learning_rate": 5.190476587271666e-06, + "loss": 0.9534, + "step": 54396 + }, + { + "epoch": 4.02, + "learning_rate": 5.190126881705324e-06, + "loss": 1.0556, + "step": 54397 + }, + { + "epoch": 4.02, + "learning_rate": 5.189777183791419e-06, + "loss": 0.9719, + "step": 54398 + }, + { + "epoch": 4.02, + "learning_rate": 5.189427493530513e-06, + "loss": 1.0075, + "step": 54399 + }, + { + "epoch": 4.02, + "learning_rate": 5.189077810923163e-06, + "loss": 1.0415, + "step": 54400 + }, + { + "epoch": 4.02, + "learning_rate": 5.188728135969914e-06, + "loss": 0.9564, + "step": 54401 + }, + { + "epoch": 4.02, + "learning_rate": 5.188378468671333e-06, + "loss": 0.9705, + "step": 54402 + }, + { + "epoch": 4.02, + "learning_rate": 5.188028809027974e-06, + "loss": 0.8989, + "step": 54403 + }, + { + "epoch": 4.02, + "learning_rate": 5.187679157040391e-06, + "loss": 0.9677, + "step": 54404 + }, + { + "epoch": 4.02, + "learning_rate": 5.187329512709138e-06, + "loss": 1.0238, + "step": 54405 + }, + { + "epoch": 4.02, + "learning_rate": 5.186979876034779e-06, + "loss": 1.0201, + "step": 54406 + }, + { + "epoch": 4.02, + "learning_rate": 5.186630247017866e-06, + "loss": 0.9024, + "step": 54407 + }, + { + "epoch": 4.02, + "learning_rate": 5.186280625658956e-06, + "loss": 1.0029, + "step": 54408 + }, + { + "epoch": 4.02, + "learning_rate": 5.185931011958605e-06, + "loss": 0.9429, + "step": 54409 + }, + { + "epoch": 4.02, + "learning_rate": 5.185581405917364e-06, + "loss": 1.0122, + "step": 54410 + }, + { + "epoch": 4.02, + "learning_rate": 5.185231807535798e-06, + "loss": 1.0698, + "step": 54411 + }, + { + "epoch": 4.02, + "learning_rate": 5.1848822168144594e-06, + "loss": 0.9898, + "step": 54412 + }, + { + "epoch": 4.02, + "learning_rate": 5.184532633753904e-06, + "loss": 1.0218, + "step": 54413 + }, + { + "epoch": 4.02, + "learning_rate": 5.1841830583546845e-06, + "loss": 1.0001, + "step": 54414 + }, + { + "epoch": 4.02, + "learning_rate": 5.183833490617365e-06, + "loss": 0.9924, + "step": 54415 + }, + { + "epoch": 4.02, + "learning_rate": 5.183483930542498e-06, + "loss": 1.0091, + "step": 54416 + }, + { + "epoch": 4.02, + "learning_rate": 5.183134378130638e-06, + "loss": 0.906, + "step": 54417 + }, + { + "epoch": 4.02, + "learning_rate": 5.182784833382344e-06, + "loss": 0.8614, + "step": 54418 + }, + { + "epoch": 4.02, + "learning_rate": 5.182435296298165e-06, + "loss": 0.9578, + "step": 54419 + }, + { + "epoch": 4.02, + "learning_rate": 5.182085766878671e-06, + "loss": 0.9917, + "step": 54420 + }, + { + "epoch": 4.02, + "learning_rate": 5.181736245124403e-06, + "loss": 0.9748, + "step": 54421 + }, + { + "epoch": 4.02, + "learning_rate": 5.181386731035925e-06, + "loss": 1.1061, + "step": 54422 + }, + { + "epoch": 4.02, + "learning_rate": 5.181037224613789e-06, + "loss": 1.0208, + "step": 54423 + }, + { + "epoch": 4.02, + "learning_rate": 5.1806877258585615e-06, + "loss": 0.95, + "step": 54424 + }, + { + "epoch": 4.02, + "learning_rate": 5.180338234770783e-06, + "loss": 0.9609, + "step": 54425 + }, + { + "epoch": 4.02, + "learning_rate": 5.179988751351021e-06, + "loss": 1.0108, + "step": 54426 + }, + { + "epoch": 4.02, + "learning_rate": 5.17963927559983e-06, + "loss": 0.8357, + "step": 54427 + }, + { + "epoch": 4.02, + "learning_rate": 5.179289807517762e-06, + "loss": 0.9675, + "step": 54428 + }, + { + "epoch": 4.02, + "learning_rate": 5.1789403471053756e-06, + "loss": 0.9465, + "step": 54429 + }, + { + "epoch": 4.02, + "learning_rate": 5.178590894363222e-06, + "loss": 1.1564, + "step": 54430 + }, + { + "epoch": 4.02, + "learning_rate": 5.178241449291865e-06, + "loss": 1.0024, + "step": 54431 + }, + { + "epoch": 4.02, + "learning_rate": 5.177892011891857e-06, + "loss": 1.0009, + "step": 54432 + }, + { + "epoch": 4.02, + "learning_rate": 5.177542582163755e-06, + "loss": 1.0396, + "step": 54433 + }, + { + "epoch": 4.02, + "learning_rate": 5.1771931601081084e-06, + "loss": 1.0299, + "step": 54434 + }, + { + "epoch": 4.02, + "learning_rate": 5.176843745725483e-06, + "loss": 1.0353, + "step": 54435 + }, + { + "epoch": 4.02, + "learning_rate": 5.176494339016429e-06, + "loss": 0.9524, + "step": 54436 + }, + { + "epoch": 4.02, + "learning_rate": 5.1761449399815045e-06, + "loss": 1.0104, + "step": 54437 + }, + { + "epoch": 4.02, + "learning_rate": 5.175795548621265e-06, + "loss": 0.9743, + "step": 54438 + }, + { + "epoch": 4.02, + "learning_rate": 5.175446164936262e-06, + "loss": 1.0328, + "step": 54439 + }, + { + "epoch": 4.02, + "learning_rate": 5.175096788927058e-06, + "loss": 1.057, + "step": 54440 + }, + { + "epoch": 4.02, + "learning_rate": 5.174747420594206e-06, + "loss": 1.0168, + "step": 54441 + }, + { + "epoch": 4.02, + "learning_rate": 5.174398059938262e-06, + "loss": 1.0535, + "step": 54442 + }, + { + "epoch": 4.02, + "learning_rate": 5.1740487069597775e-06, + "loss": 1.0436, + "step": 54443 + }, + { + "epoch": 4.02, + "learning_rate": 5.173699361659317e-06, + "loss": 0.9999, + "step": 54444 + }, + { + "epoch": 4.02, + "learning_rate": 5.1733500240374315e-06, + "loss": 0.9446, + "step": 54445 + }, + { + "epoch": 4.02, + "learning_rate": 5.173000694094677e-06, + "loss": 1.0955, + "step": 54446 + }, + { + "epoch": 4.02, + "learning_rate": 5.17265137183161e-06, + "loss": 0.9114, + "step": 54447 + }, + { + "epoch": 4.02, + "learning_rate": 5.1723020572487835e-06, + "loss": 1.0574, + "step": 54448 + }, + { + "epoch": 4.02, + "learning_rate": 5.171952750346757e-06, + "loss": 1.0155, + "step": 54449 + }, + { + "epoch": 4.02, + "learning_rate": 5.1716034511260805e-06, + "loss": 1.0566, + "step": 54450 + }, + { + "epoch": 4.02, + "learning_rate": 5.171254159587318e-06, + "loss": 0.9076, + "step": 54451 + }, + { + "epoch": 4.02, + "learning_rate": 5.17090487573102e-06, + "loss": 0.9912, + "step": 54452 + }, + { + "epoch": 4.02, + "learning_rate": 5.170555599557745e-06, + "loss": 1.0667, + "step": 54453 + }, + { + "epoch": 4.02, + "learning_rate": 5.170206331068042e-06, + "loss": 1.0481, + "step": 54454 + }, + { + "epoch": 4.02, + "learning_rate": 5.169857070262474e-06, + "loss": 0.9958, + "step": 54455 + }, + { + "epoch": 4.02, + "learning_rate": 5.169507817141597e-06, + "loss": 1.1283, + "step": 54456 + }, + { + "epoch": 4.02, + "learning_rate": 5.169158571705962e-06, + "loss": 0.9938, + "step": 54457 + }, + { + "epoch": 4.02, + "learning_rate": 5.1688093339561285e-06, + "loss": 1.0463, + "step": 54458 + }, + { + "epoch": 4.02, + "learning_rate": 5.168460103892645e-06, + "loss": 0.952, + "step": 54459 + }, + { + "epoch": 4.02, + "learning_rate": 5.168110881516075e-06, + "loss": 1.0087, + "step": 54460 + }, + { + "epoch": 4.02, + "learning_rate": 5.167761666826974e-06, + "loss": 1.0016, + "step": 54461 + }, + { + "epoch": 4.02, + "learning_rate": 5.167412459825894e-06, + "loss": 0.9975, + "step": 54462 + }, + { + "epoch": 4.02, + "learning_rate": 5.167063260513388e-06, + "loss": 0.9185, + "step": 54463 + }, + { + "epoch": 4.02, + "learning_rate": 5.166714068890017e-06, + "loss": 0.9253, + "step": 54464 + }, + { + "epoch": 4.02, + "learning_rate": 5.166364884956341e-06, + "loss": 1.0704, + "step": 54465 + }, + { + "epoch": 4.02, + "learning_rate": 5.166015708712902e-06, + "loss": 1.0487, + "step": 54466 + }, + { + "epoch": 4.02, + "learning_rate": 5.165666540160266e-06, + "loss": 0.9998, + "step": 54467 + }, + { + "epoch": 4.02, + "learning_rate": 5.16531737929898e-06, + "loss": 1.0881, + "step": 54468 + }, + { + "epoch": 4.02, + "learning_rate": 5.164968226129614e-06, + "loss": 0.902, + "step": 54469 + }, + { + "epoch": 4.02, + "learning_rate": 5.164619080652706e-06, + "loss": 1.1196, + "step": 54470 + }, + { + "epoch": 4.02, + "learning_rate": 5.164269942868825e-06, + "loss": 1.0146, + "step": 54471 + }, + { + "epoch": 4.02, + "learning_rate": 5.163920812778516e-06, + "loss": 1.0171, + "step": 54472 + }, + { + "epoch": 4.02, + "learning_rate": 5.1635716903823486e-06, + "loss": 0.9613, + "step": 54473 + }, + { + "epoch": 4.03, + "learning_rate": 5.163222575680865e-06, + "loss": 0.827, + "step": 54474 + }, + { + "epoch": 4.03, + "learning_rate": 5.16287346867462e-06, + "loss": 0.9509, + "step": 54475 + }, + { + "epoch": 4.03, + "learning_rate": 5.162524369364182e-06, + "loss": 1.0755, + "step": 54476 + }, + { + "epoch": 4.03, + "learning_rate": 5.1621752777500946e-06, + "loss": 1.0761, + "step": 54477 + }, + { + "epoch": 4.03, + "learning_rate": 5.161826193832919e-06, + "loss": 0.9692, + "step": 54478 + }, + { + "epoch": 4.03, + "learning_rate": 5.161477117613203e-06, + "loss": 1.0475, + "step": 54479 + }, + { + "epoch": 4.03, + "learning_rate": 5.161128049091513e-06, + "loss": 1.0004, + "step": 54480 + }, + { + "epoch": 4.03, + "learning_rate": 5.160778988268399e-06, + "loss": 1.013, + "step": 54481 + }, + { + "epoch": 4.03, + "learning_rate": 5.160429935144415e-06, + "loss": 0.9799, + "step": 54482 + }, + { + "epoch": 4.03, + "learning_rate": 5.160080889720121e-06, + "loss": 0.9352, + "step": 54483 + }, + { + "epoch": 4.03, + "learning_rate": 5.159731851996061e-06, + "loss": 0.9559, + "step": 54484 + }, + { + "epoch": 4.03, + "learning_rate": 5.159382821972805e-06, + "loss": 1.0423, + "step": 54485 + }, + { + "epoch": 4.03, + "learning_rate": 5.159033799650901e-06, + "loss": 0.9215, + "step": 54486 + }, + { + "epoch": 4.03, + "learning_rate": 5.158684785030905e-06, + "loss": 1.0653, + "step": 54487 + }, + { + "epoch": 4.03, + "learning_rate": 5.158335778113368e-06, + "loss": 1.0778, + "step": 54488 + }, + { + "epoch": 4.03, + "learning_rate": 5.157986778898853e-06, + "loss": 1.0669, + "step": 54489 + }, + { + "epoch": 4.03, + "learning_rate": 5.157637787387914e-06, + "loss": 1.0376, + "step": 54490 + }, + { + "epoch": 4.03, + "learning_rate": 5.157288803581102e-06, + "loss": 1.0283, + "step": 54491 + }, + { + "epoch": 4.03, + "learning_rate": 5.15693982747897e-06, + "loss": 1.021, + "step": 54492 + }, + { + "epoch": 4.03, + "learning_rate": 5.156590859082087e-06, + "loss": 1.0343, + "step": 54493 + }, + { + "epoch": 4.03, + "learning_rate": 5.156241898390995e-06, + "loss": 0.9515, + "step": 54494 + }, + { + "epoch": 4.03, + "learning_rate": 5.1558929454062465e-06, + "loss": 0.9709, + "step": 54495 + }, + { + "epoch": 4.03, + "learning_rate": 5.155544000128407e-06, + "loss": 1.0175, + "step": 54496 + }, + { + "epoch": 4.03, + "learning_rate": 5.155195062558029e-06, + "loss": 0.9697, + "step": 54497 + }, + { + "epoch": 4.03, + "learning_rate": 5.1548461326956675e-06, + "loss": 1.0413, + "step": 54498 + }, + { + "epoch": 4.03, + "learning_rate": 5.154497210541871e-06, + "loss": 1.0559, + "step": 54499 + }, + { + "epoch": 4.03, + "learning_rate": 5.154148296097203e-06, + "loss": 0.9771, + "step": 54500 + }, + { + "epoch": 4.03, + "learning_rate": 5.1537993893622174e-06, + "loss": 1.0351, + "step": 54501 + }, + { + "epoch": 4.03, + "learning_rate": 5.1534504903374686e-06, + "loss": 1.1131, + "step": 54502 + }, + { + "epoch": 4.03, + "learning_rate": 5.153101599023509e-06, + "loss": 0.9722, + "step": 54503 + }, + { + "epoch": 4.03, + "learning_rate": 5.152752715420891e-06, + "loss": 1.058, + "step": 54504 + }, + { + "epoch": 4.03, + "learning_rate": 5.152403839530179e-06, + "loss": 1.0151, + "step": 54505 + }, + { + "epoch": 4.03, + "learning_rate": 5.152054971351923e-06, + "loss": 0.9781, + "step": 54506 + }, + { + "epoch": 4.03, + "learning_rate": 5.151706110886678e-06, + "loss": 0.9829, + "step": 54507 + }, + { + "epoch": 4.03, + "learning_rate": 5.151357258134996e-06, + "loss": 0.9793, + "step": 54508 + }, + { + "epoch": 4.03, + "learning_rate": 5.151008413097438e-06, + "loss": 0.9718, + "step": 54509 + }, + { + "epoch": 4.03, + "learning_rate": 5.150659575774557e-06, + "loss": 0.9746, + "step": 54510 + }, + { + "epoch": 4.03, + "learning_rate": 5.1503107461669064e-06, + "loss": 0.9084, + "step": 54511 + }, + { + "epoch": 4.03, + "learning_rate": 5.149961924275043e-06, + "loss": 0.9894, + "step": 54512 + }, + { + "epoch": 4.03, + "learning_rate": 5.149613110099516e-06, + "loss": 1.0614, + "step": 54513 + }, + { + "epoch": 4.03, + "learning_rate": 5.149264303640895e-06, + "loss": 1.0264, + "step": 54514 + }, + { + "epoch": 4.03, + "learning_rate": 5.148915504899715e-06, + "loss": 1.0282, + "step": 54515 + }, + { + "epoch": 4.03, + "learning_rate": 5.148566713876546e-06, + "loss": 0.9675, + "step": 54516 + }, + { + "epoch": 4.03, + "learning_rate": 5.148217930571934e-06, + "loss": 0.9133, + "step": 54517 + }, + { + "epoch": 4.03, + "learning_rate": 5.147869154986444e-06, + "loss": 0.9694, + "step": 54518 + }, + { + "epoch": 4.03, + "learning_rate": 5.147520387120619e-06, + "loss": 0.8875, + "step": 54519 + }, + { + "epoch": 4.03, + "learning_rate": 5.147171626975023e-06, + "loss": 0.9952, + "step": 54520 + }, + { + "epoch": 4.03, + "learning_rate": 5.146822874550208e-06, + "loss": 0.9781, + "step": 54521 + }, + { + "epoch": 4.03, + "learning_rate": 5.146474129846728e-06, + "loss": 1.0087, + "step": 54522 + }, + { + "epoch": 4.03, + "learning_rate": 5.146125392865138e-06, + "loss": 1.0549, + "step": 54523 + }, + { + "epoch": 4.03, + "learning_rate": 5.145776663605988e-06, + "loss": 0.9031, + "step": 54524 + }, + { + "epoch": 4.03, + "learning_rate": 5.145427942069843e-06, + "loss": 1.0241, + "step": 54525 + }, + { + "epoch": 4.03, + "learning_rate": 5.145079228257253e-06, + "loss": 0.9289, + "step": 54526 + }, + { + "epoch": 4.03, + "learning_rate": 5.144730522168773e-06, + "loss": 0.9697, + "step": 54527 + }, + { + "epoch": 4.03, + "learning_rate": 5.1443818238049516e-06, + "loss": 0.9533, + "step": 54528 + }, + { + "epoch": 4.03, + "learning_rate": 5.144033133166354e-06, + "loss": 1.0418, + "step": 54529 + }, + { + "epoch": 4.03, + "learning_rate": 5.14368445025353e-06, + "loss": 1.0249, + "step": 54530 + }, + { + "epoch": 4.03, + "learning_rate": 5.143335775067035e-06, + "loss": 1.0012, + "step": 54531 + }, + { + "epoch": 4.03, + "learning_rate": 5.142987107607422e-06, + "loss": 0.9814, + "step": 54532 + }, + { + "epoch": 4.03, + "learning_rate": 5.142638447875246e-06, + "loss": 0.9737, + "step": 54533 + }, + { + "epoch": 4.03, + "learning_rate": 5.1422897958710635e-06, + "loss": 0.9542, + "step": 54534 + }, + { + "epoch": 4.03, + "learning_rate": 5.14194115159543e-06, + "loss": 0.9622, + "step": 54535 + }, + { + "epoch": 4.03, + "learning_rate": 5.1415925150488985e-06, + "loss": 1.0315, + "step": 54536 + }, + { + "epoch": 4.03, + "learning_rate": 5.1412438862320194e-06, + "loss": 1.0809, + "step": 54537 + }, + { + "epoch": 4.03, + "learning_rate": 5.1408952651453604e-06, + "loss": 1.0183, + "step": 54538 + }, + { + "epoch": 4.03, + "learning_rate": 5.140546651789462e-06, + "loss": 1.0357, + "step": 54539 + }, + { + "epoch": 4.03, + "learning_rate": 5.140198046164883e-06, + "loss": 1.0366, + "step": 54540 + }, + { + "epoch": 4.03, + "learning_rate": 5.139849448272181e-06, + "loss": 1.0351, + "step": 54541 + }, + { + "epoch": 4.03, + "learning_rate": 5.13950085811191e-06, + "loss": 0.9077, + "step": 54542 + }, + { + "epoch": 4.03, + "learning_rate": 5.139152275684623e-06, + "loss": 0.9995, + "step": 54543 + }, + { + "epoch": 4.03, + "learning_rate": 5.138803700990872e-06, + "loss": 0.9906, + "step": 54544 + }, + { + "epoch": 4.03, + "learning_rate": 5.138455134031218e-06, + "loss": 1.0251, + "step": 54545 + }, + { + "epoch": 4.03, + "learning_rate": 5.138106574806213e-06, + "loss": 1.0549, + "step": 54546 + }, + { + "epoch": 4.03, + "learning_rate": 5.13775802331641e-06, + "loss": 1.0401, + "step": 54547 + }, + { + "epoch": 4.03, + "learning_rate": 5.137409479562362e-06, + "loss": 1.0073, + "step": 54548 + }, + { + "epoch": 4.03, + "learning_rate": 5.137060943544629e-06, + "loss": 0.9612, + "step": 54549 + }, + { + "epoch": 4.03, + "learning_rate": 5.136712415263763e-06, + "loss": 1.0799, + "step": 54550 + }, + { + "epoch": 4.03, + "learning_rate": 5.136363894720316e-06, + "loss": 1.0575, + "step": 54551 + }, + { + "epoch": 4.03, + "learning_rate": 5.136015381914846e-06, + "loss": 1.0123, + "step": 54552 + }, + { + "epoch": 4.03, + "learning_rate": 5.135666876847902e-06, + "loss": 1.0926, + "step": 54553 + }, + { + "epoch": 4.03, + "learning_rate": 5.1353183795200465e-06, + "loss": 0.9654, + "step": 54554 + }, + { + "epoch": 4.03, + "learning_rate": 5.134969889931829e-06, + "loss": 1.0484, + "step": 54555 + }, + { + "epoch": 4.03, + "learning_rate": 5.134621408083807e-06, + "loss": 0.9313, + "step": 54556 + }, + { + "epoch": 4.03, + "learning_rate": 5.134272933976525e-06, + "loss": 1.0297, + "step": 54557 + }, + { + "epoch": 4.03, + "learning_rate": 5.1339244676105515e-06, + "loss": 1.0887, + "step": 54558 + }, + { + "epoch": 4.03, + "learning_rate": 5.133576008986438e-06, + "loss": 1.014, + "step": 54559 + }, + { + "epoch": 4.03, + "learning_rate": 5.1332275581047285e-06, + "loss": 0.9203, + "step": 54560 + }, + { + "epoch": 4.03, + "learning_rate": 5.1328791149659876e-06, + "loss": 0.9658, + "step": 54561 + }, + { + "epoch": 4.03, + "learning_rate": 5.132530679570762e-06, + "loss": 0.9233, + "step": 54562 + }, + { + "epoch": 4.03, + "learning_rate": 5.132182251919619e-06, + "loss": 0.9335, + "step": 54563 + }, + { + "epoch": 4.03, + "learning_rate": 5.131833832013096e-06, + "loss": 1.0414, + "step": 54564 + }, + { + "epoch": 4.03, + "learning_rate": 5.131485419851761e-06, + "loss": 1.0633, + "step": 54565 + }, + { + "epoch": 4.03, + "learning_rate": 5.131137015436156e-06, + "loss": 0.9463, + "step": 54566 + }, + { + "epoch": 4.03, + "learning_rate": 5.130788618766853e-06, + "loss": 1.0281, + "step": 54567 + }, + { + "epoch": 4.03, + "learning_rate": 5.130440229844391e-06, + "loss": 1.0089, + "step": 54568 + }, + { + "epoch": 4.03, + "learning_rate": 5.130091848669325e-06, + "loss": 0.9534, + "step": 54569 + }, + { + "epoch": 4.03, + "learning_rate": 5.129743475242217e-06, + "loss": 0.8844, + "step": 54570 + }, + { + "epoch": 4.03, + "learning_rate": 5.129395109563618e-06, + "loss": 1.0226, + "step": 54571 + }, + { + "epoch": 4.03, + "learning_rate": 5.129046751634081e-06, + "loss": 0.965, + "step": 54572 + }, + { + "epoch": 4.03, + "learning_rate": 5.128698401454156e-06, + "loss": 0.9317, + "step": 54573 + }, + { + "epoch": 4.03, + "learning_rate": 5.128350059024407e-06, + "loss": 1.1707, + "step": 54574 + }, + { + "epoch": 4.03, + "learning_rate": 5.128001724345384e-06, + "loss": 0.9638, + "step": 54575 + }, + { + "epoch": 4.03, + "learning_rate": 5.12765339741764e-06, + "loss": 0.9816, + "step": 54576 + }, + { + "epoch": 4.03, + "learning_rate": 5.12730507824173e-06, + "loss": 1.0318, + "step": 54577 + }, + { + "epoch": 4.03, + "learning_rate": 5.1269567668182045e-06, + "loss": 1.0611, + "step": 54578 + }, + { + "epoch": 4.03, + "learning_rate": 5.126608463147628e-06, + "loss": 0.8974, + "step": 54579 + }, + { + "epoch": 4.03, + "learning_rate": 5.12626016723054e-06, + "loss": 1.0081, + "step": 54580 + }, + { + "epoch": 4.03, + "learning_rate": 5.125911879067507e-06, + "loss": 1.0315, + "step": 54581 + }, + { + "epoch": 4.03, + "learning_rate": 5.125563598659074e-06, + "loss": 0.9093, + "step": 54582 + }, + { + "epoch": 4.03, + "learning_rate": 5.125215326005806e-06, + "loss": 0.9896, + "step": 54583 + }, + { + "epoch": 4.03, + "learning_rate": 5.124867061108245e-06, + "loss": 1.052, + "step": 54584 + }, + { + "epoch": 4.03, + "learning_rate": 5.124518803966954e-06, + "loss": 0.9035, + "step": 54585 + }, + { + "epoch": 4.03, + "learning_rate": 5.124170554582484e-06, + "loss": 0.929, + "step": 54586 + }, + { + "epoch": 4.03, + "learning_rate": 5.1238223129553875e-06, + "loss": 0.945, + "step": 54587 + }, + { + "epoch": 4.03, + "learning_rate": 5.123474079086221e-06, + "loss": 0.94, + "step": 54588 + }, + { + "epoch": 4.03, + "learning_rate": 5.123125852975533e-06, + "loss": 0.9368, + "step": 54589 + }, + { + "epoch": 4.03, + "learning_rate": 5.1227776346238875e-06, + "loss": 1.0293, + "step": 54590 + }, + { + "epoch": 4.03, + "learning_rate": 5.122429424031832e-06, + "loss": 1.0226, + "step": 54591 + }, + { + "epoch": 4.03, + "learning_rate": 5.122081221199922e-06, + "loss": 0.9642, + "step": 54592 + }, + { + "epoch": 4.03, + "learning_rate": 5.1217330261287056e-06, + "loss": 1.0601, + "step": 54593 + }, + { + "epoch": 4.03, + "learning_rate": 5.121384838818746e-06, + "loss": 0.9635, + "step": 54594 + }, + { + "epoch": 4.03, + "learning_rate": 5.121036659270595e-06, + "loss": 0.9583, + "step": 54595 + }, + { + "epoch": 4.03, + "learning_rate": 5.120688487484804e-06, + "loss": 1.116, + "step": 54596 + }, + { + "epoch": 4.03, + "learning_rate": 5.1203403234619275e-06, + "loss": 0.9566, + "step": 54597 + }, + { + "epoch": 4.03, + "learning_rate": 5.1199921672025164e-06, + "loss": 1.0055, + "step": 54598 + }, + { + "epoch": 4.03, + "learning_rate": 5.119644018707133e-06, + "loss": 0.9274, + "step": 54599 + }, + { + "epoch": 4.03, + "learning_rate": 5.1192958779763245e-06, + "loss": 1.0444, + "step": 54600 + }, + { + "epoch": 4.03, + "learning_rate": 5.1189477450106465e-06, + "loss": 1.0357, + "step": 54601 + }, + { + "epoch": 4.03, + "learning_rate": 5.1185996198106495e-06, + "loss": 0.9673, + "step": 54602 + }, + { + "epoch": 4.03, + "learning_rate": 5.118251502376894e-06, + "loss": 0.9499, + "step": 54603 + }, + { + "epoch": 4.03, + "learning_rate": 5.117903392709931e-06, + "loss": 0.9901, + "step": 54604 + }, + { + "epoch": 4.03, + "learning_rate": 5.117555290810313e-06, + "loss": 1.0588, + "step": 54605 + }, + { + "epoch": 4.03, + "learning_rate": 5.1172071966785955e-06, + "loss": 0.9626, + "step": 54606 + }, + { + "epoch": 4.03, + "learning_rate": 5.116859110315328e-06, + "loss": 0.8835, + "step": 54607 + }, + { + "epoch": 4.03, + "learning_rate": 5.116511031721074e-06, + "loss": 0.9469, + "step": 54608 + }, + { + "epoch": 4.04, + "learning_rate": 5.116162960896375e-06, + "loss": 1.0829, + "step": 54609 + }, + { + "epoch": 4.04, + "learning_rate": 5.1158148978417936e-06, + "loss": 0.8906, + "step": 54610 + }, + { + "epoch": 4.04, + "learning_rate": 5.115466842557877e-06, + "loss": 0.9314, + "step": 54611 + }, + { + "epoch": 4.04, + "learning_rate": 5.115118795045191e-06, + "loss": 0.9663, + "step": 54612 + }, + { + "epoch": 4.04, + "learning_rate": 5.114770755304273e-06, + "loss": 1.093, + "step": 54613 + }, + { + "epoch": 4.04, + "learning_rate": 5.11442272333569e-06, + "loss": 0.994, + "step": 54614 + }, + { + "epoch": 4.04, + "learning_rate": 5.114074699139989e-06, + "loss": 0.9112, + "step": 54615 + }, + { + "epoch": 4.04, + "learning_rate": 5.113726682717726e-06, + "loss": 1.0182, + "step": 54616 + }, + { + "epoch": 4.04, + "learning_rate": 5.113378674069454e-06, + "loss": 1.0294, + "step": 54617 + }, + { + "epoch": 4.04, + "learning_rate": 5.113030673195723e-06, + "loss": 1.0619, + "step": 54618 + }, + { + "epoch": 4.04, + "learning_rate": 5.112682680097092e-06, + "loss": 0.9788, + "step": 54619 + }, + { + "epoch": 4.04, + "learning_rate": 5.1123346947741144e-06, + "loss": 0.9963, + "step": 54620 + }, + { + "epoch": 4.04, + "learning_rate": 5.111986717227343e-06, + "loss": 1.0246, + "step": 54621 + }, + { + "epoch": 4.04, + "learning_rate": 5.111638747457326e-06, + "loss": 1.0084, + "step": 54622 + }, + { + "epoch": 4.04, + "learning_rate": 5.111290785464627e-06, + "loss": 1.0466, + "step": 54623 + }, + { + "epoch": 4.04, + "learning_rate": 5.1109428312497965e-06, + "loss": 1.1128, + "step": 54624 + }, + { + "epoch": 4.04, + "learning_rate": 5.110594884813379e-06, + "loss": 0.9814, + "step": 54625 + }, + { + "epoch": 4.04, + "learning_rate": 5.110246946155939e-06, + "loss": 1.0143, + "step": 54626 + }, + { + "epoch": 4.04, + "learning_rate": 5.109899015278023e-06, + "loss": 1.0443, + "step": 54627 + }, + { + "epoch": 4.04, + "learning_rate": 5.109551092180194e-06, + "loss": 1.0019, + "step": 54628 + }, + { + "epoch": 4.04, + "learning_rate": 5.1092031768629925e-06, + "loss": 0.9514, + "step": 54629 + }, + { + "epoch": 4.04, + "learning_rate": 5.108855269326982e-06, + "loss": 0.9852, + "step": 54630 + }, + { + "epoch": 4.04, + "learning_rate": 5.1085073695727106e-06, + "loss": 1.1468, + "step": 54631 + }, + { + "epoch": 4.04, + "learning_rate": 5.10815947760074e-06, + "loss": 0.9482, + "step": 54632 + }, + { + "epoch": 4.04, + "learning_rate": 5.107811593411614e-06, + "loss": 1.0478, + "step": 54633 + }, + { + "epoch": 4.04, + "learning_rate": 5.107463717005886e-06, + "loss": 0.9902, + "step": 54634 + }, + { + "epoch": 4.04, + "learning_rate": 5.107115848384117e-06, + "loss": 0.9581, + "step": 54635 + }, + { + "epoch": 4.04, + "learning_rate": 5.106767987546857e-06, + "loss": 0.966, + "step": 54636 + }, + { + "epoch": 4.04, + "learning_rate": 5.10642013449466e-06, + "loss": 1.0346, + "step": 54637 + }, + { + "epoch": 4.04, + "learning_rate": 5.106072289228074e-06, + "loss": 0.9976, + "step": 54638 + }, + { + "epoch": 4.04, + "learning_rate": 5.10572445174766e-06, + "loss": 1.1672, + "step": 54639 + }, + { + "epoch": 4.04, + "learning_rate": 5.105376622053968e-06, + "loss": 0.9685, + "step": 54640 + }, + { + "epoch": 4.04, + "learning_rate": 5.105028800147554e-06, + "loss": 1.079, + "step": 54641 + }, + { + "epoch": 4.04, + "learning_rate": 5.104680986028967e-06, + "loss": 1.0202, + "step": 54642 + }, + { + "epoch": 4.04, + "learning_rate": 5.10433317969876e-06, + "loss": 0.9588, + "step": 54643 + }, + { + "epoch": 4.04, + "learning_rate": 5.103985381157492e-06, + "loss": 1.1577, + "step": 54644 + }, + { + "epoch": 4.04, + "learning_rate": 5.1036375904057135e-06, + "loss": 1.014, + "step": 54645 + }, + { + "epoch": 4.04, + "learning_rate": 5.103289807443977e-06, + "loss": 0.9456, + "step": 54646 + }, + { + "epoch": 4.04, + "learning_rate": 5.102942032272833e-06, + "loss": 1.0618, + "step": 54647 + }, + { + "epoch": 4.04, + "learning_rate": 5.102594264892842e-06, + "loss": 0.9862, + "step": 54648 + }, + { + "epoch": 4.04, + "learning_rate": 5.102246505304554e-06, + "loss": 1.0921, + "step": 54649 + }, + { + "epoch": 4.04, + "learning_rate": 5.101898753508521e-06, + "loss": 0.9227, + "step": 54650 + }, + { + "epoch": 4.04, + "learning_rate": 5.101551009505293e-06, + "loss": 1.038, + "step": 54651 + }, + { + "epoch": 4.04, + "learning_rate": 5.101203273295432e-06, + "loss": 0.94, + "step": 54652 + }, + { + "epoch": 4.04, + "learning_rate": 5.10085554487949e-06, + "loss": 1.0272, + "step": 54653 + }, + { + "epoch": 4.04, + "learning_rate": 5.100507824258009e-06, + "loss": 1.0168, + "step": 54654 + }, + { + "epoch": 4.04, + "learning_rate": 5.100160111431554e-06, + "loss": 0.8886, + "step": 54655 + }, + { + "epoch": 4.04, + "learning_rate": 5.099812406400668e-06, + "loss": 1.0814, + "step": 54656 + }, + { + "epoch": 4.04, + "learning_rate": 5.099464709165921e-06, + "loss": 1.0079, + "step": 54657 + }, + { + "epoch": 4.04, + "learning_rate": 5.099117019727846e-06, + "loss": 0.9364, + "step": 54658 + }, + { + "epoch": 4.04, + "learning_rate": 5.09876933808701e-06, + "loss": 0.971, + "step": 54659 + }, + { + "epoch": 4.04, + "learning_rate": 5.098421664243963e-06, + "loss": 0.8788, + "step": 54660 + }, + { + "epoch": 4.04, + "learning_rate": 5.098073998199256e-06, + "loss": 1.0314, + "step": 54661 + }, + { + "epoch": 4.04, + "learning_rate": 5.0977263399534434e-06, + "loss": 0.9507, + "step": 54662 + }, + { + "epoch": 4.04, + "learning_rate": 5.097378689507075e-06, + "loss": 0.9631, + "step": 54663 + }, + { + "epoch": 4.04, + "learning_rate": 5.09703104686071e-06, + "loss": 0.9827, + "step": 54664 + }, + { + "epoch": 4.04, + "learning_rate": 5.096683412014898e-06, + "loss": 0.9521, + "step": 54665 + }, + { + "epoch": 4.04, + "learning_rate": 5.096335784970193e-06, + "loss": 0.9741, + "step": 54666 + }, + { + "epoch": 4.04, + "learning_rate": 5.095988165727144e-06, + "loss": 0.8783, + "step": 54667 + }, + { + "epoch": 4.04, + "learning_rate": 5.095640554286311e-06, + "loss": 0.9479, + "step": 54668 + }, + { + "epoch": 4.04, + "learning_rate": 5.095292950648244e-06, + "loss": 1.0274, + "step": 54669 + }, + { + "epoch": 4.04, + "learning_rate": 5.0949453548134965e-06, + "loss": 0.9353, + "step": 54670 + }, + { + "epoch": 4.04, + "learning_rate": 5.094597766782619e-06, + "loss": 1.021, + "step": 54671 + }, + { + "epoch": 4.04, + "learning_rate": 5.0942501865561635e-06, + "loss": 0.8806, + "step": 54672 + }, + { + "epoch": 4.04, + "learning_rate": 5.0939026141346935e-06, + "loss": 0.9699, + "step": 54673 + }, + { + "epoch": 4.04, + "learning_rate": 5.0935550495187456e-06, + "loss": 0.9804, + "step": 54674 + }, + { + "epoch": 4.04, + "learning_rate": 5.093207492708887e-06, + "loss": 0.9792, + "step": 54675 + }, + { + "epoch": 4.04, + "learning_rate": 5.092859943705659e-06, + "loss": 0.9015, + "step": 54676 + }, + { + "epoch": 4.04, + "learning_rate": 5.092512402509631e-06, + "loss": 1.0537, + "step": 54677 + }, + { + "epoch": 4.04, + "learning_rate": 5.092164869121336e-06, + "loss": 0.9843, + "step": 54678 + }, + { + "epoch": 4.04, + "learning_rate": 5.091817343541341e-06, + "loss": 1.0241, + "step": 54679 + }, + { + "epoch": 4.04, + "learning_rate": 5.091469825770195e-06, + "loss": 1.0526, + "step": 54680 + }, + { + "epoch": 4.04, + "learning_rate": 5.091122315808448e-06, + "loss": 1.0584, + "step": 54681 + }, + { + "epoch": 4.04, + "learning_rate": 5.090774813656657e-06, + "loss": 0.9089, + "step": 54682 + }, + { + "epoch": 4.04, + "learning_rate": 5.090427319315368e-06, + "loss": 1.0801, + "step": 54683 + }, + { + "epoch": 4.04, + "learning_rate": 5.090079832785144e-06, + "loss": 0.8762, + "step": 54684 + }, + { + "epoch": 4.04, + "learning_rate": 5.089732354066532e-06, + "loss": 1.057, + "step": 54685 + }, + { + "epoch": 4.04, + "learning_rate": 5.089384883160087e-06, + "loss": 1.0109, + "step": 54686 + }, + { + "epoch": 4.04, + "learning_rate": 5.089037420066354e-06, + "loss": 0.9848, + "step": 54687 + }, + { + "epoch": 4.04, + "learning_rate": 5.088689964785898e-06, + "loss": 1.0586, + "step": 54688 + }, + { + "epoch": 4.04, + "learning_rate": 5.0883425173192645e-06, + "loss": 1.0232, + "step": 54689 + }, + { + "epoch": 4.04, + "learning_rate": 5.08799507766701e-06, + "loss": 0.9914, + "step": 54690 + }, + { + "epoch": 4.04, + "learning_rate": 5.087647645829683e-06, + "loss": 0.9445, + "step": 54691 + }, + { + "epoch": 4.04, + "learning_rate": 5.087300221807836e-06, + "loss": 0.9958, + "step": 54692 + }, + { + "epoch": 4.04, + "learning_rate": 5.0869528056020265e-06, + "loss": 1.0797, + "step": 54693 + }, + { + "epoch": 4.04, + "learning_rate": 5.086605397212805e-06, + "loss": 1.0202, + "step": 54694 + }, + { + "epoch": 4.04, + "learning_rate": 5.086257996640724e-06, + "loss": 1.0379, + "step": 54695 + }, + { + "epoch": 4.04, + "learning_rate": 5.085910603886334e-06, + "loss": 0.9615, + "step": 54696 + }, + { + "epoch": 4.04, + "learning_rate": 5.085563218950192e-06, + "loss": 0.9532, + "step": 54697 + }, + { + "epoch": 4.04, + "learning_rate": 5.085215841832854e-06, + "loss": 0.9571, + "step": 54698 + }, + { + "epoch": 4.04, + "learning_rate": 5.084868472534858e-06, + "loss": 1.0949, + "step": 54699 + }, + { + "epoch": 4.04, + "learning_rate": 5.08452111105677e-06, + "loss": 0.9722, + "step": 54700 + }, + { + "epoch": 4.04, + "learning_rate": 5.084173757399139e-06, + "loss": 0.9181, + "step": 54701 + }, + { + "epoch": 4.04, + "learning_rate": 5.083826411562518e-06, + "loss": 0.921, + "step": 54702 + }, + { + "epoch": 4.04, + "learning_rate": 5.083479073547453e-06, + "loss": 1.0488, + "step": 54703 + }, + { + "epoch": 4.04, + "learning_rate": 5.083131743354508e-06, + "loss": 0.9329, + "step": 54704 + }, + { + "epoch": 4.04, + "learning_rate": 5.082784420984229e-06, + "loss": 1.0517, + "step": 54705 + }, + { + "epoch": 4.04, + "learning_rate": 5.08243710643717e-06, + "loss": 0.9688, + "step": 54706 + }, + { + "epoch": 4.04, + "learning_rate": 5.082089799713881e-06, + "loss": 1.0135, + "step": 54707 + }, + { + "epoch": 4.04, + "learning_rate": 5.081742500814918e-06, + "loss": 1.1697, + "step": 54708 + }, + { + "epoch": 4.04, + "learning_rate": 5.081395209740834e-06, + "loss": 0.9471, + "step": 54709 + }, + { + "epoch": 4.04, + "learning_rate": 5.081047926492178e-06, + "loss": 1.0273, + "step": 54710 + }, + { + "epoch": 4.04, + "learning_rate": 5.080700651069507e-06, + "loss": 0.9389, + "step": 54711 + }, + { + "epoch": 4.04, + "learning_rate": 5.080353383473366e-06, + "loss": 0.987, + "step": 54712 + }, + { + "epoch": 4.04, + "learning_rate": 5.0800061237043155e-06, + "loss": 1.043, + "step": 54713 + }, + { + "epoch": 4.04, + "learning_rate": 5.079658871762905e-06, + "loss": 1.0855, + "step": 54714 + }, + { + "epoch": 4.04, + "learning_rate": 5.079311627649687e-06, + "loss": 1.1429, + "step": 54715 + }, + { + "epoch": 4.04, + "learning_rate": 5.07896439136521e-06, + "loss": 1.0503, + "step": 54716 + }, + { + "epoch": 4.04, + "learning_rate": 5.078617162910034e-06, + "loss": 0.9027, + "step": 54717 + }, + { + "epoch": 4.04, + "learning_rate": 5.078269942284713e-06, + "loss": 0.9711, + "step": 54718 + }, + { + "epoch": 4.04, + "learning_rate": 5.077922729489784e-06, + "loss": 1.0381, + "step": 54719 + }, + { + "epoch": 4.04, + "learning_rate": 5.077575524525814e-06, + "loss": 0.9909, + "step": 54720 + }, + { + "epoch": 4.04, + "learning_rate": 5.077228327393348e-06, + "loss": 0.9811, + "step": 54721 + }, + { + "epoch": 4.04, + "learning_rate": 5.076881138092949e-06, + "loss": 0.9987, + "step": 54722 + }, + { + "epoch": 4.04, + "learning_rate": 5.076533956625152e-06, + "loss": 1.0029, + "step": 54723 + }, + { + "epoch": 4.04, + "learning_rate": 5.076186782990526e-06, + "loss": 1.0108, + "step": 54724 + }, + { + "epoch": 4.04, + "learning_rate": 5.075839617189611e-06, + "loss": 0.996, + "step": 54725 + }, + { + "epoch": 4.04, + "learning_rate": 5.075492459222972e-06, + "loss": 1.0244, + "step": 54726 + }, + { + "epoch": 4.04, + "learning_rate": 5.07514530909115e-06, + "loss": 0.9258, + "step": 54727 + }, + { + "epoch": 4.04, + "learning_rate": 5.074798166794698e-06, + "loss": 0.9107, + "step": 54728 + }, + { + "epoch": 4.04, + "learning_rate": 5.074451032334175e-06, + "loss": 1.0696, + "step": 54729 + }, + { + "epoch": 4.04, + "learning_rate": 5.074103905710131e-06, + "loss": 1.0526, + "step": 54730 + }, + { + "epoch": 4.04, + "learning_rate": 5.073756786923116e-06, + "loss": 0.984, + "step": 54731 + }, + { + "epoch": 4.04, + "learning_rate": 5.0734096759736805e-06, + "loss": 1.0886, + "step": 54732 + }, + { + "epoch": 4.04, + "learning_rate": 5.073062572862384e-06, + "loss": 0.9529, + "step": 54733 + }, + { + "epoch": 4.04, + "learning_rate": 5.072715477589774e-06, + "loss": 0.9299, + "step": 54734 + }, + { + "epoch": 4.04, + "learning_rate": 5.072368390156401e-06, + "loss": 0.968, + "step": 54735 + }, + { + "epoch": 4.04, + "learning_rate": 5.072021310562822e-06, + "loss": 0.8781, + "step": 54736 + }, + { + "epoch": 4.04, + "learning_rate": 5.071674238809583e-06, + "loss": 1.0549, + "step": 54737 + }, + { + "epoch": 4.04, + "learning_rate": 5.071327174897242e-06, + "loss": 0.9938, + "step": 54738 + }, + { + "epoch": 4.04, + "learning_rate": 5.0709801188263496e-06, + "loss": 1.0156, + "step": 54739 + }, + { + "epoch": 4.04, + "learning_rate": 5.070633070597457e-06, + "loss": 1.1543, + "step": 54740 + }, + { + "epoch": 4.04, + "learning_rate": 5.0702860302111136e-06, + "loss": 1.0258, + "step": 54741 + }, + { + "epoch": 4.04, + "learning_rate": 5.069938997667878e-06, + "loss": 0.933, + "step": 54742 + }, + { + "epoch": 4.04, + "learning_rate": 5.0695919729683e-06, + "loss": 0.9709, + "step": 54743 + }, + { + "epoch": 4.04, + "learning_rate": 5.069244956112929e-06, + "loss": 0.9798, + "step": 54744 + }, + { + "epoch": 4.05, + "learning_rate": 5.068897947102318e-06, + "loss": 1.0541, + "step": 54745 + }, + { + "epoch": 4.05, + "learning_rate": 5.068550945937022e-06, + "loss": 0.9175, + "step": 54746 + }, + { + "epoch": 4.05, + "learning_rate": 5.0682039526175895e-06, + "loss": 0.9726, + "step": 54747 + }, + { + "epoch": 4.05, + "learning_rate": 5.067856967144571e-06, + "loss": 0.9571, + "step": 54748 + }, + { + "epoch": 4.05, + "learning_rate": 5.067509989518525e-06, + "loss": 1.1199, + "step": 54749 + }, + { + "epoch": 4.05, + "learning_rate": 5.06716301974e-06, + "loss": 0.9206, + "step": 54750 + }, + { + "epoch": 4.05, + "learning_rate": 5.066816057809548e-06, + "loss": 1.0079, + "step": 54751 + }, + { + "epoch": 4.05, + "learning_rate": 5.066469103727717e-06, + "loss": 0.9938, + "step": 54752 + }, + { + "epoch": 4.05, + "learning_rate": 5.066122157495068e-06, + "loss": 1.0604, + "step": 54753 + }, + { + "epoch": 4.05, + "learning_rate": 5.065775219112147e-06, + "loss": 0.986, + "step": 54754 + }, + { + "epoch": 4.05, + "learning_rate": 5.065428288579507e-06, + "loss": 1.0062, + "step": 54755 + }, + { + "epoch": 4.05, + "learning_rate": 5.065081365897699e-06, + "loss": 0.9675, + "step": 54756 + }, + { + "epoch": 4.05, + "learning_rate": 5.0647344510672744e-06, + "loss": 1.004, + "step": 54757 + }, + { + "epoch": 4.05, + "learning_rate": 5.064387544088789e-06, + "loss": 0.9586, + "step": 54758 + }, + { + "epoch": 4.05, + "learning_rate": 5.064040644962793e-06, + "loss": 1.1559, + "step": 54759 + }, + { + "epoch": 4.05, + "learning_rate": 5.063693753689838e-06, + "loss": 1.0464, + "step": 54760 + }, + { + "epoch": 4.05, + "learning_rate": 5.0633468702704714e-06, + "loss": 0.9526, + "step": 54761 + }, + { + "epoch": 4.05, + "learning_rate": 5.0629999947052534e-06, + "loss": 0.989, + "step": 54762 + }, + { + "epoch": 4.05, + "learning_rate": 5.062653126994731e-06, + "loss": 1.0476, + "step": 54763 + }, + { + "epoch": 4.05, + "learning_rate": 5.062306267139456e-06, + "loss": 1.0513, + "step": 54764 + }, + { + "epoch": 4.05, + "learning_rate": 5.061959415139983e-06, + "loss": 1.0217, + "step": 54765 + }, + { + "epoch": 4.05, + "learning_rate": 5.0616125709968574e-06, + "loss": 0.9387, + "step": 54766 + }, + { + "epoch": 4.05, + "learning_rate": 5.0612657347106434e-06, + "loss": 1.0272, + "step": 54767 + }, + { + "epoch": 4.05, + "learning_rate": 5.060918906281877e-06, + "loss": 0.9855, + "step": 54768 + }, + { + "epoch": 4.05, + "learning_rate": 5.060572085711122e-06, + "loss": 0.9669, + "step": 54769 + }, + { + "epoch": 4.05, + "learning_rate": 5.060225272998922e-06, + "loss": 0.9098, + "step": 54770 + }, + { + "epoch": 4.05, + "learning_rate": 5.059878468145841e-06, + "loss": 1.0092, + "step": 54771 + }, + { + "epoch": 4.05, + "learning_rate": 5.059531671152416e-06, + "loss": 0.9314, + "step": 54772 + }, + { + "epoch": 4.05, + "learning_rate": 5.059184882019208e-06, + "loss": 1.0495, + "step": 54773 + }, + { + "epoch": 4.05, + "learning_rate": 5.058838100746766e-06, + "loss": 0.9827, + "step": 54774 + }, + { + "epoch": 4.05, + "learning_rate": 5.058491327335641e-06, + "loss": 0.9589, + "step": 54775 + }, + { + "epoch": 4.05, + "learning_rate": 5.058144561786386e-06, + "loss": 0.9253, + "step": 54776 + }, + { + "epoch": 4.05, + "learning_rate": 5.057797804099551e-06, + "loss": 0.9786, + "step": 54777 + }, + { + "epoch": 4.05, + "learning_rate": 5.0574510542756906e-06, + "loss": 1.0254, + "step": 54778 + }, + { + "epoch": 4.05, + "learning_rate": 5.057104312315355e-06, + "loss": 0.9412, + "step": 54779 + }, + { + "epoch": 4.05, + "learning_rate": 5.056757578219096e-06, + "loss": 1.0425, + "step": 54780 + }, + { + "epoch": 4.05, + "learning_rate": 5.056410851987461e-06, + "loss": 1.1285, + "step": 54781 + }, + { + "epoch": 4.05, + "learning_rate": 5.056064133621011e-06, + "loss": 0.9938, + "step": 54782 + }, + { + "epoch": 4.05, + "learning_rate": 5.05571742312029e-06, + "loss": 1.0888, + "step": 54783 + }, + { + "epoch": 4.05, + "learning_rate": 5.055370720485853e-06, + "loss": 1.0655, + "step": 54784 + }, + { + "epoch": 4.05, + "learning_rate": 5.055024025718249e-06, + "loss": 1.0644, + "step": 54785 + }, + { + "epoch": 4.05, + "learning_rate": 5.054677338818028e-06, + "loss": 1.0445, + "step": 54786 + }, + { + "epoch": 4.05, + "learning_rate": 5.054330659785749e-06, + "loss": 0.8949, + "step": 54787 + }, + { + "epoch": 4.05, + "learning_rate": 5.053983988621959e-06, + "loss": 1.0649, + "step": 54788 + }, + { + "epoch": 4.05, + "learning_rate": 5.05363732532721e-06, + "loss": 0.9489, + "step": 54789 + }, + { + "epoch": 4.05, + "learning_rate": 5.053290669902049e-06, + "loss": 0.9957, + "step": 54790 + }, + { + "epoch": 4.05, + "learning_rate": 5.05294402234704e-06, + "loss": 0.9011, + "step": 54791 + }, + { + "epoch": 4.05, + "learning_rate": 5.052597382662721e-06, + "loss": 1.0069, + "step": 54792 + }, + { + "epoch": 4.05, + "learning_rate": 5.052250750849645e-06, + "loss": 0.9903, + "step": 54793 + }, + { + "epoch": 4.05, + "learning_rate": 5.051904126908372e-06, + "loss": 1.0049, + "step": 54794 + }, + { + "epoch": 4.05, + "learning_rate": 5.051557510839449e-06, + "loss": 0.8687, + "step": 54795 + }, + { + "epoch": 4.05, + "learning_rate": 5.0512109026434265e-06, + "loss": 1.0256, + "step": 54796 + }, + { + "epoch": 4.05, + "learning_rate": 5.050864302320852e-06, + "loss": 0.9158, + "step": 54797 + }, + { + "epoch": 4.05, + "learning_rate": 5.050517709872288e-06, + "loss": 0.9944, + "step": 54798 + }, + { + "epoch": 4.05, + "learning_rate": 5.050171125298277e-06, + "loss": 0.9771, + "step": 54799 + }, + { + "epoch": 4.05, + "learning_rate": 5.049824548599375e-06, + "loss": 0.9389, + "step": 54800 + }, + { + "epoch": 4.05, + "learning_rate": 5.04947797977613e-06, + "loss": 0.9439, + "step": 54801 + }, + { + "epoch": 4.05, + "learning_rate": 5.049131418829092e-06, + "loss": 1.1481, + "step": 54802 + }, + { + "epoch": 4.05, + "learning_rate": 5.048784865758818e-06, + "loss": 1.0739, + "step": 54803 + }, + { + "epoch": 4.05, + "learning_rate": 5.048438320565856e-06, + "loss": 1.024, + "step": 54804 + }, + { + "epoch": 4.05, + "learning_rate": 5.048091783250758e-06, + "loss": 0.9572, + "step": 54805 + }, + { + "epoch": 4.05, + "learning_rate": 5.047745253814071e-06, + "loss": 0.9635, + "step": 54806 + }, + { + "epoch": 4.05, + "learning_rate": 5.047398732256356e-06, + "loss": 1.0522, + "step": 54807 + }, + { + "epoch": 4.05, + "learning_rate": 5.047052218578156e-06, + "loss": 0.8644, + "step": 54808 + }, + { + "epoch": 4.05, + "learning_rate": 5.046705712780028e-06, + "loss": 0.9774, + "step": 54809 + }, + { + "epoch": 4.05, + "learning_rate": 5.046359214862515e-06, + "loss": 1.1415, + "step": 54810 + }, + { + "epoch": 4.05, + "learning_rate": 5.046012724826177e-06, + "loss": 1.1024, + "step": 54811 + }, + { + "epoch": 4.05, + "learning_rate": 5.0456662426715675e-06, + "loss": 1.0323, + "step": 54812 + }, + { + "epoch": 4.05, + "learning_rate": 5.045319768399223e-06, + "loss": 0.9394, + "step": 54813 + }, + { + "epoch": 4.05, + "learning_rate": 5.044973302009708e-06, + "loss": 0.9354, + "step": 54814 + }, + { + "epoch": 4.05, + "learning_rate": 5.044626843503566e-06, + "loss": 1.07, + "step": 54815 + }, + { + "epoch": 4.05, + "learning_rate": 5.044280392881359e-06, + "loss": 0.9131, + "step": 54816 + }, + { + "epoch": 4.05, + "learning_rate": 5.043933950143623e-06, + "loss": 1.064, + "step": 54817 + }, + { + "epoch": 4.05, + "learning_rate": 5.043587515290922e-06, + "loss": 0.9247, + "step": 54818 + }, + { + "epoch": 4.05, + "learning_rate": 5.043241088323798e-06, + "loss": 1.0649, + "step": 54819 + }, + { + "epoch": 4.05, + "learning_rate": 5.0428946692428135e-06, + "loss": 0.9243, + "step": 54820 + }, + { + "epoch": 4.05, + "learning_rate": 5.0425482580485095e-06, + "loss": 1.0642, + "step": 54821 + }, + { + "epoch": 4.05, + "learning_rate": 5.0422018547414375e-06, + "loss": 0.9434, + "step": 54822 + }, + { + "epoch": 4.05, + "learning_rate": 5.041855459322154e-06, + "loss": 1.1038, + "step": 54823 + }, + { + "epoch": 4.05, + "learning_rate": 5.041509071791207e-06, + "loss": 0.9772, + "step": 54824 + }, + { + "epoch": 4.05, + "learning_rate": 5.04116269214915e-06, + "loss": 1.0765, + "step": 54825 + }, + { + "epoch": 4.05, + "learning_rate": 5.040816320396527e-06, + "loss": 0.8876, + "step": 54826 + }, + { + "epoch": 4.05, + "learning_rate": 5.040469956533898e-06, + "loss": 1.0076, + "step": 54827 + }, + { + "epoch": 4.05, + "learning_rate": 5.04012360056181e-06, + "loss": 1.0555, + "step": 54828 + }, + { + "epoch": 4.05, + "learning_rate": 5.039777252480815e-06, + "loss": 1.0511, + "step": 54829 + }, + { + "epoch": 4.05, + "learning_rate": 5.039430912291466e-06, + "loss": 1.038, + "step": 54830 + }, + { + "epoch": 4.05, + "learning_rate": 5.039084579994303e-06, + "loss": 1.0962, + "step": 54831 + }, + { + "epoch": 4.05, + "learning_rate": 5.0387382555898925e-06, + "loss": 0.977, + "step": 54832 + }, + { + "epoch": 4.05, + "learning_rate": 5.038391939078777e-06, + "loss": 1.0226, + "step": 54833 + }, + { + "epoch": 4.05, + "learning_rate": 5.03804563046151e-06, + "loss": 0.9638, + "step": 54834 + }, + { + "epoch": 4.05, + "learning_rate": 5.037699329738637e-06, + "loss": 1.0299, + "step": 54835 + }, + { + "epoch": 4.05, + "learning_rate": 5.037353036910719e-06, + "loss": 1.0877, + "step": 54836 + }, + { + "epoch": 4.05, + "learning_rate": 5.0370067519782975e-06, + "loss": 1.0133, + "step": 54837 + }, + { + "epoch": 4.05, + "learning_rate": 5.036660474941928e-06, + "loss": 1.071, + "step": 54838 + }, + { + "epoch": 4.05, + "learning_rate": 5.0363142058021615e-06, + "loss": 1.0812, + "step": 54839 + }, + { + "epoch": 4.05, + "learning_rate": 5.035967944559549e-06, + "loss": 1.0266, + "step": 54840 + }, + { + "epoch": 4.05, + "learning_rate": 5.03562169121464e-06, + "loss": 1.0376, + "step": 54841 + }, + { + "epoch": 4.05, + "learning_rate": 5.035275445767982e-06, + "loss": 1.0583, + "step": 54842 + }, + { + "epoch": 4.05, + "learning_rate": 5.034929208220134e-06, + "loss": 0.9885, + "step": 54843 + }, + { + "epoch": 4.05, + "learning_rate": 5.034582978571641e-06, + "loss": 0.8899, + "step": 54844 + }, + { + "epoch": 4.05, + "learning_rate": 5.034236756823057e-06, + "loss": 0.9802, + "step": 54845 + }, + { + "epoch": 4.05, + "learning_rate": 5.0338905429749264e-06, + "loss": 1.0153, + "step": 54846 + }, + { + "epoch": 4.05, + "learning_rate": 5.033544337027809e-06, + "loss": 0.9931, + "step": 54847 + }, + { + "epoch": 4.05, + "learning_rate": 5.033198138982253e-06, + "loss": 0.958, + "step": 54848 + }, + { + "epoch": 4.05, + "learning_rate": 5.032851948838806e-06, + "loss": 0.9783, + "step": 54849 + }, + { + "epoch": 4.05, + "learning_rate": 5.032505766598022e-06, + "loss": 1.0017, + "step": 54850 + }, + { + "epoch": 4.05, + "learning_rate": 5.032159592260444e-06, + "loss": 0.9478, + "step": 54851 + }, + { + "epoch": 4.05, + "learning_rate": 5.031813425826635e-06, + "loss": 0.9612, + "step": 54852 + }, + { + "epoch": 4.05, + "learning_rate": 5.031467267297139e-06, + "loss": 1.0302, + "step": 54853 + }, + { + "epoch": 4.05, + "learning_rate": 5.031121116672506e-06, + "loss": 0.9436, + "step": 54854 + }, + { + "epoch": 4.05, + "learning_rate": 5.030774973953287e-06, + "loss": 1.1044, + "step": 54855 + }, + { + "epoch": 4.05, + "learning_rate": 5.030428839140036e-06, + "loss": 0.9146, + "step": 54856 + }, + { + "epoch": 4.05, + "learning_rate": 5.030082712233306e-06, + "loss": 1.0002, + "step": 54857 + }, + { + "epoch": 4.05, + "learning_rate": 5.029736593233636e-06, + "loss": 1.0035, + "step": 54858 + }, + { + "epoch": 4.05, + "learning_rate": 5.029390482141587e-06, + "loss": 1.0577, + "step": 54859 + }, + { + "epoch": 4.05, + "learning_rate": 5.029044378957704e-06, + "loss": 1.0477, + "step": 54860 + }, + { + "epoch": 4.05, + "learning_rate": 5.028698283682547e-06, + "loss": 1.0838, + "step": 54861 + }, + { + "epoch": 4.05, + "learning_rate": 5.028352196316653e-06, + "loss": 1.0689, + "step": 54862 + }, + { + "epoch": 4.05, + "learning_rate": 5.028006116860582e-06, + "loss": 1.003, + "step": 54863 + }, + { + "epoch": 4.05, + "learning_rate": 5.027660045314879e-06, + "loss": 1.0169, + "step": 54864 + }, + { + "epoch": 4.05, + "learning_rate": 5.027313981680106e-06, + "loss": 0.9637, + "step": 54865 + }, + { + "epoch": 4.05, + "learning_rate": 5.0269679259567965e-06, + "loss": 1.0456, + "step": 54866 + }, + { + "epoch": 4.05, + "learning_rate": 5.0266218781455146e-06, + "loss": 1.0567, + "step": 54867 + }, + { + "epoch": 4.05, + "learning_rate": 5.026275838246805e-06, + "loss": 0.9274, + "step": 54868 + }, + { + "epoch": 4.05, + "learning_rate": 5.02592980626122e-06, + "loss": 0.9622, + "step": 54869 + }, + { + "epoch": 4.05, + "learning_rate": 5.0255837821893115e-06, + "loss": 0.8676, + "step": 54870 + }, + { + "epoch": 4.05, + "learning_rate": 5.025237766031621e-06, + "loss": 1.0635, + "step": 54871 + }, + { + "epoch": 4.05, + "learning_rate": 5.024891757788713e-06, + "loss": 1.0273, + "step": 54872 + }, + { + "epoch": 4.05, + "learning_rate": 5.024545757461128e-06, + "loss": 1.033, + "step": 54873 + }, + { + "epoch": 4.05, + "learning_rate": 5.024199765049421e-06, + "loss": 1.0311, + "step": 54874 + }, + { + "epoch": 4.05, + "learning_rate": 5.023853780554137e-06, + "loss": 1.1056, + "step": 54875 + }, + { + "epoch": 4.05, + "learning_rate": 5.023507803975834e-06, + "loss": 1.0029, + "step": 54876 + }, + { + "epoch": 4.05, + "learning_rate": 5.02316183531506e-06, + "loss": 0.9998, + "step": 54877 + }, + { + "epoch": 4.05, + "learning_rate": 5.022815874572363e-06, + "loss": 1.0197, + "step": 54878 + }, + { + "epoch": 4.05, + "learning_rate": 5.022469921748295e-06, + "loss": 1.0738, + "step": 54879 + }, + { + "epoch": 4.06, + "learning_rate": 5.022123976843403e-06, + "loss": 1.0522, + "step": 54880 + }, + { + "epoch": 4.06, + "learning_rate": 5.021778039858247e-06, + "loss": 0.9692, + "step": 54881 + }, + { + "epoch": 4.06, + "learning_rate": 5.021432110793365e-06, + "loss": 0.9265, + "step": 54882 + }, + { + "epoch": 4.06, + "learning_rate": 5.021086189649316e-06, + "loss": 0.9885, + "step": 54883 + }, + { + "epoch": 4.06, + "learning_rate": 5.020740276426643e-06, + "loss": 1.0159, + "step": 54884 + }, + { + "epoch": 4.06, + "learning_rate": 5.020394371125909e-06, + "loss": 0.9577, + "step": 54885 + }, + { + "epoch": 4.06, + "learning_rate": 5.020048473747654e-06, + "loss": 0.9794, + "step": 54886 + }, + { + "epoch": 4.06, + "learning_rate": 5.019702584292425e-06, + "loss": 1.1633, + "step": 54887 + }, + { + "epoch": 4.06, + "learning_rate": 5.019356702760783e-06, + "loss": 0.9898, + "step": 54888 + }, + { + "epoch": 4.06, + "learning_rate": 5.019010829153272e-06, + "loss": 1.0132, + "step": 54889 + }, + { + "epoch": 4.06, + "learning_rate": 5.0186649634704455e-06, + "loss": 1.0042, + "step": 54890 + }, + { + "epoch": 4.06, + "learning_rate": 5.018319105712847e-06, + "loss": 0.954, + "step": 54891 + }, + { + "epoch": 4.06, + "learning_rate": 5.017973255881035e-06, + "loss": 1.0613, + "step": 54892 + }, + { + "epoch": 4.06, + "learning_rate": 5.017627413975557e-06, + "loss": 0.9635, + "step": 54893 + }, + { + "epoch": 4.06, + "learning_rate": 5.017281579996961e-06, + "loss": 1.057, + "step": 54894 + }, + { + "epoch": 4.06, + "learning_rate": 5.016935753945799e-06, + "loss": 1.0795, + "step": 54895 + }, + { + "epoch": 4.06, + "learning_rate": 5.016589935822618e-06, + "loss": 0.9502, + "step": 54896 + }, + { + "epoch": 4.06, + "learning_rate": 5.016244125627975e-06, + "loss": 0.9596, + "step": 54897 + }, + { + "epoch": 4.06, + "learning_rate": 5.015898323362416e-06, + "loss": 0.9809, + "step": 54898 + }, + { + "epoch": 4.06, + "learning_rate": 5.015552529026491e-06, + "loss": 1.0187, + "step": 54899 + }, + { + "epoch": 4.06, + "learning_rate": 5.015206742620747e-06, + "loss": 0.8585, + "step": 54900 + }, + { + "epoch": 4.06, + "learning_rate": 5.0148609641457425e-06, + "loss": 1.0902, + "step": 54901 + }, + { + "epoch": 4.06, + "learning_rate": 5.014515193602021e-06, + "loss": 0.9729, + "step": 54902 + }, + { + "epoch": 4.06, + "learning_rate": 5.014169430990137e-06, + "loss": 1.0304, + "step": 54903 + }, + { + "epoch": 4.06, + "learning_rate": 5.013823676310636e-06, + "loss": 1.0209, + "step": 54904 + }, + { + "epoch": 4.06, + "learning_rate": 5.013477929564067e-06, + "loss": 1.0157, + "step": 54905 + }, + { + "epoch": 4.06, + "learning_rate": 5.013132190750991e-06, + "loss": 0.943, + "step": 54906 + }, + { + "epoch": 4.06, + "learning_rate": 5.012786459871942e-06, + "loss": 0.9041, + "step": 54907 + }, + { + "epoch": 4.06, + "learning_rate": 5.012440736927483e-06, + "loss": 0.9834, + "step": 54908 + }, + { + "epoch": 4.06, + "learning_rate": 5.012095021918155e-06, + "loss": 0.9329, + "step": 54909 + }, + { + "epoch": 4.06, + "learning_rate": 5.01174931484452e-06, + "loss": 1.0466, + "step": 54910 + }, + { + "epoch": 4.06, + "learning_rate": 5.011403615707114e-06, + "loss": 1.0844, + "step": 54911 + }, + { + "epoch": 4.06, + "learning_rate": 5.011057924506497e-06, + "loss": 0.9543, + "step": 54912 + }, + { + "epoch": 4.06, + "learning_rate": 5.010712241243215e-06, + "loss": 0.9294, + "step": 54913 + }, + { + "epoch": 4.06, + "learning_rate": 5.010366565917819e-06, + "loss": 1.0311, + "step": 54914 + }, + { + "epoch": 4.06, + "learning_rate": 5.010020898530858e-06, + "loss": 1.0605, + "step": 54915 + }, + { + "epoch": 4.06, + "learning_rate": 5.009675239082878e-06, + "loss": 1.0335, + "step": 54916 + }, + { + "epoch": 4.06, + "learning_rate": 5.009329587574439e-06, + "loss": 1.0932, + "step": 54917 + }, + { + "epoch": 4.06, + "learning_rate": 5.0089839440060826e-06, + "loss": 0.9537, + "step": 54918 + }, + { + "epoch": 4.06, + "learning_rate": 5.008638308378363e-06, + "loss": 1.1425, + "step": 54919 + }, + { + "epoch": 4.06, + "learning_rate": 5.008292680691824e-06, + "loss": 1.0632, + "step": 54920 + }, + { + "epoch": 4.06, + "learning_rate": 5.007947060947025e-06, + "loss": 1.0734, + "step": 54921 + }, + { + "epoch": 4.06, + "learning_rate": 5.007601449144509e-06, + "loss": 1.0099, + "step": 54922 + }, + { + "epoch": 4.06, + "learning_rate": 5.0072558452848285e-06, + "loss": 1.0681, + "step": 54923 + }, + { + "epoch": 4.06, + "learning_rate": 5.006910249368533e-06, + "loss": 0.9029, + "step": 54924 + }, + { + "epoch": 4.06, + "learning_rate": 5.0065646613961676e-06, + "loss": 1.0129, + "step": 54925 + }, + { + "epoch": 4.06, + "learning_rate": 5.006219081368293e-06, + "loss": 0.8276, + "step": 54926 + }, + { + "epoch": 4.06, + "learning_rate": 5.005873509285446e-06, + "loss": 1.0573, + "step": 54927 + }, + { + "epoch": 4.06, + "learning_rate": 5.005527945148186e-06, + "loss": 1.0261, + "step": 54928 + }, + { + "epoch": 4.06, + "learning_rate": 5.005182388957056e-06, + "loss": 1.0206, + "step": 54929 + }, + { + "epoch": 4.06, + "learning_rate": 5.004836840712617e-06, + "loss": 1.031, + "step": 54930 + }, + { + "epoch": 4.06, + "learning_rate": 5.004491300415404e-06, + "loss": 1.0785, + "step": 54931 + }, + { + "epoch": 4.06, + "learning_rate": 5.004145768065976e-06, + "loss": 0.9956, + "step": 54932 + }, + { + "epoch": 4.06, + "learning_rate": 5.003800243664882e-06, + "loss": 1.0868, + "step": 54933 + }, + { + "epoch": 4.06, + "learning_rate": 5.003454727212671e-06, + "loss": 1.0587, + "step": 54934 + }, + { + "epoch": 4.06, + "learning_rate": 5.0031092187098905e-06, + "loss": 1.086, + "step": 54935 + }, + { + "epoch": 4.06, + "learning_rate": 5.002763718157089e-06, + "loss": 0.9634, + "step": 54936 + }, + { + "epoch": 4.06, + "learning_rate": 5.002418225554821e-06, + "loss": 0.9684, + "step": 54937 + }, + { + "epoch": 4.06, + "learning_rate": 5.0020727409036365e-06, + "loss": 0.9081, + "step": 54938 + }, + { + "epoch": 4.06, + "learning_rate": 5.001727264204081e-06, + "loss": 0.9968, + "step": 54939 + }, + { + "epoch": 4.06, + "learning_rate": 5.0013817954567035e-06, + "loss": 1.0823, + "step": 54940 + }, + { + "epoch": 4.06, + "learning_rate": 5.0010363346620595e-06, + "loss": 0.9483, + "step": 54941 + }, + { + "epoch": 4.06, + "learning_rate": 5.000690881820694e-06, + "loss": 0.9068, + "step": 54942 + }, + { + "epoch": 4.06, + "learning_rate": 5.000345436933159e-06, + "loss": 1.0014, + "step": 54943 + }, + { + "epoch": 4.06, + "learning_rate": 5.000000000000003e-06, + "loss": 0.9527, + "step": 54944 + }, + { + "epoch": 4.06, + "learning_rate": 4.999654571021771e-06, + "loss": 0.9769, + "step": 54945 + }, + { + "epoch": 4.06, + "learning_rate": 4.999309149999021e-06, + "loss": 0.9616, + "step": 54946 + }, + { + "epoch": 4.06, + "learning_rate": 4.9989637369322975e-06, + "loss": 0.9775, + "step": 54947 + }, + { + "epoch": 4.06, + "learning_rate": 4.998618331822153e-06, + "loss": 1.0918, + "step": 54948 + }, + { + "epoch": 4.06, + "learning_rate": 4.998272934669131e-06, + "loss": 0.9307, + "step": 54949 + }, + { + "epoch": 4.06, + "learning_rate": 4.997927545473787e-06, + "loss": 0.9965, + "step": 54950 + }, + { + "epoch": 4.06, + "learning_rate": 4.997582164236675e-06, + "loss": 0.9719, + "step": 54951 + }, + { + "epoch": 4.06, + "learning_rate": 4.99723679095833e-06, + "loss": 1.0002, + "step": 54952 + }, + { + "epoch": 4.06, + "learning_rate": 4.9968914256393135e-06, + "loss": 0.9906, + "step": 54953 + }, + { + "epoch": 4.06, + "learning_rate": 4.996546068280166e-06, + "loss": 0.9765, + "step": 54954 + }, + { + "epoch": 4.06, + "learning_rate": 4.996200718881451e-06, + "loss": 1.0087, + "step": 54955 + }, + { + "epoch": 4.06, + "learning_rate": 4.995855377443701e-06, + "loss": 1.0878, + "step": 54956 + }, + { + "epoch": 4.06, + "learning_rate": 4.995510043967478e-06, + "loss": 0.9762, + "step": 54957 + }, + { + "epoch": 4.06, + "learning_rate": 4.9951647184533254e-06, + "loss": 0.9223, + "step": 54958 + }, + { + "epoch": 4.06, + "learning_rate": 4.994819400901796e-06, + "loss": 1.0403, + "step": 54959 + }, + { + "epoch": 4.06, + "learning_rate": 4.9944740913134356e-06, + "loss": 1.0284, + "step": 54960 + }, + { + "epoch": 4.06, + "learning_rate": 4.994128789688792e-06, + "loss": 0.9935, + "step": 54961 + }, + { + "epoch": 4.06, + "learning_rate": 4.993783496028423e-06, + "loss": 0.9704, + "step": 54962 + }, + { + "epoch": 4.06, + "learning_rate": 4.993438210332872e-06, + "loss": 0.9642, + "step": 54963 + }, + { + "epoch": 4.06, + "learning_rate": 4.993092932602689e-06, + "loss": 1.0394, + "step": 54964 + }, + { + "epoch": 4.06, + "learning_rate": 4.9927476628384194e-06, + "loss": 1.0723, + "step": 54965 + }, + { + "epoch": 4.06, + "learning_rate": 4.9924024010406216e-06, + "loss": 0.9573, + "step": 54966 + }, + { + "epoch": 4.06, + "learning_rate": 4.9920571472098386e-06, + "loss": 0.961, + "step": 54967 + }, + { + "epoch": 4.06, + "learning_rate": 4.991711901346621e-06, + "loss": 0.9359, + "step": 54968 + }, + { + "epoch": 4.06, + "learning_rate": 4.991366663451519e-06, + "loss": 1.0457, + "step": 54969 + }, + { + "epoch": 4.06, + "learning_rate": 4.991021433525077e-06, + "loss": 1.0903, + "step": 54970 + }, + { + "epoch": 4.06, + "learning_rate": 4.990676211567856e-06, + "loss": 1.0603, + "step": 54971 + }, + { + "epoch": 4.06, + "learning_rate": 4.9903309975803895e-06, + "loss": 0.9299, + "step": 54972 + }, + { + "epoch": 4.06, + "learning_rate": 4.989985791563237e-06, + "loss": 0.9272, + "step": 54973 + }, + { + "epoch": 4.06, + "learning_rate": 4.989640593516945e-06, + "loss": 1.0333, + "step": 54974 + }, + { + "epoch": 4.06, + "learning_rate": 4.989295403442068e-06, + "loss": 1.0325, + "step": 54975 + }, + { + "epoch": 4.06, + "learning_rate": 4.988950221339144e-06, + "loss": 0.95, + "step": 54976 + }, + { + "epoch": 4.06, + "learning_rate": 4.988605047208731e-06, + "loss": 1.0037, + "step": 54977 + }, + { + "epoch": 4.06, + "learning_rate": 4.988259881051374e-06, + "loss": 0.9445, + "step": 54978 + }, + { + "epoch": 4.06, + "learning_rate": 4.987914722867629e-06, + "loss": 1.0403, + "step": 54979 + }, + { + "epoch": 4.06, + "learning_rate": 4.9875695726580375e-06, + "loss": 1.0515, + "step": 54980 + }, + { + "epoch": 4.06, + "learning_rate": 4.987224430423147e-06, + "loss": 1.0107, + "step": 54981 + }, + { + "epoch": 4.06, + "learning_rate": 4.9868792961635145e-06, + "loss": 0.9547, + "step": 54982 + }, + { + "epoch": 4.06, + "learning_rate": 4.986534169879686e-06, + "loss": 1.0324, + "step": 54983 + }, + { + "epoch": 4.06, + "learning_rate": 4.98618905157221e-06, + "loss": 0.9078, + "step": 54984 + }, + { + "epoch": 4.06, + "learning_rate": 4.98584394124163e-06, + "loss": 0.8937, + "step": 54985 + }, + { + "epoch": 4.06, + "learning_rate": 4.985498838888506e-06, + "loss": 0.9649, + "step": 54986 + }, + { + "epoch": 4.06, + "learning_rate": 4.985153744513382e-06, + "loss": 0.9226, + "step": 54987 + }, + { + "epoch": 4.06, + "learning_rate": 4.984808658116804e-06, + "loss": 0.9566, + "step": 54988 + }, + { + "epoch": 4.06, + "learning_rate": 4.984463579699326e-06, + "loss": 0.9625, + "step": 54989 + }, + { + "epoch": 4.06, + "learning_rate": 4.9841185092614905e-06, + "loss": 1.0386, + "step": 54990 + }, + { + "epoch": 4.06, + "learning_rate": 4.983773446803854e-06, + "loss": 0.9857, + "step": 54991 + }, + { + "epoch": 4.06, + "learning_rate": 4.9834283923269624e-06, + "loss": 1.0593, + "step": 54992 + }, + { + "epoch": 4.06, + "learning_rate": 4.983083345831364e-06, + "loss": 1.0149, + "step": 54993 + }, + { + "epoch": 4.06, + "learning_rate": 4.982738307317605e-06, + "loss": 0.9431, + "step": 54994 + }, + { + "epoch": 4.06, + "learning_rate": 4.982393276786242e-06, + "loss": 0.9323, + "step": 54995 + }, + { + "epoch": 4.06, + "learning_rate": 4.982048254237818e-06, + "loss": 0.9909, + "step": 54996 + }, + { + "epoch": 4.06, + "learning_rate": 4.981703239672885e-06, + "loss": 1.0185, + "step": 54997 + }, + { + "epoch": 4.06, + "learning_rate": 4.9813582330919885e-06, + "loss": 0.983, + "step": 54998 + }, + { + "epoch": 4.06, + "learning_rate": 4.981013234495677e-06, + "loss": 1.0607, + "step": 54999 + }, + { + "epoch": 4.06, + "learning_rate": 4.980668243884509e-06, + "loss": 0.8817, + "step": 55000 + }, + { + "epoch": 4.06, + "learning_rate": 4.9803232612590176e-06, + "loss": 0.9748, + "step": 55001 + }, + { + "epoch": 4.06, + "learning_rate": 4.979978286619765e-06, + "loss": 0.9612, + "step": 55002 + }, + { + "epoch": 4.06, + "learning_rate": 4.979633319967294e-06, + "loss": 1.0033, + "step": 55003 + }, + { + "epoch": 4.06, + "learning_rate": 4.979288361302155e-06, + "loss": 0.9238, + "step": 55004 + }, + { + "epoch": 4.06, + "learning_rate": 4.978943410624892e-06, + "loss": 1.0015, + "step": 55005 + }, + { + "epoch": 4.06, + "learning_rate": 4.9785984679360634e-06, + "loss": 1.0365, + "step": 55006 + }, + { + "epoch": 4.06, + "learning_rate": 4.978253533236211e-06, + "loss": 0.8321, + "step": 55007 + }, + { + "epoch": 4.06, + "learning_rate": 4.9779086065258875e-06, + "loss": 1.0263, + "step": 55008 + }, + { + "epoch": 4.06, + "learning_rate": 4.977563687805638e-06, + "loss": 0.9938, + "step": 55009 + }, + { + "epoch": 4.06, + "learning_rate": 4.9772187770760096e-06, + "loss": 1.0146, + "step": 55010 + }, + { + "epoch": 4.06, + "learning_rate": 4.9768738743375575e-06, + "loss": 0.9781, + "step": 55011 + }, + { + "epoch": 4.06, + "learning_rate": 4.976528979590827e-06, + "loss": 1.0034, + "step": 55012 + }, + { + "epoch": 4.06, + "learning_rate": 4.976184092836368e-06, + "loss": 0.9675, + "step": 55013 + }, + { + "epoch": 4.06, + "learning_rate": 4.975839214074724e-06, + "loss": 0.9217, + "step": 55014 + }, + { + "epoch": 4.07, + "learning_rate": 4.975494343306453e-06, + "loss": 0.9435, + "step": 55015 + }, + { + "epoch": 4.07, + "learning_rate": 4.975149480532101e-06, + "loss": 1.0843, + "step": 55016 + }, + { + "epoch": 4.07, + "learning_rate": 4.974804625752206e-06, + "loss": 0.9485, + "step": 55017 + }, + { + "epoch": 4.07, + "learning_rate": 4.97445977896733e-06, + "loss": 1.0133, + "step": 55018 + }, + { + "epoch": 4.07, + "learning_rate": 4.974114940178012e-06, + "loss": 1.0624, + "step": 55019 + }, + { + "epoch": 4.07, + "learning_rate": 4.9737701093848134e-06, + "loss": 1.0582, + "step": 55020 + }, + { + "epoch": 4.07, + "learning_rate": 4.973425286588267e-06, + "loss": 1.1038, + "step": 55021 + }, + { + "epoch": 4.07, + "learning_rate": 4.973080471788932e-06, + "loss": 1.0513, + "step": 55022 + }, + { + "epoch": 4.07, + "learning_rate": 4.972735664987351e-06, + "loss": 1.0453, + "step": 55023 + }, + { + "epoch": 4.07, + "learning_rate": 4.972390866184084e-06, + "loss": 0.9915, + "step": 55024 + }, + { + "epoch": 4.07, + "learning_rate": 4.9720460753796626e-06, + "loss": 1.036, + "step": 55025 + }, + { + "epoch": 4.07, + "learning_rate": 4.971701292574648e-06, + "loss": 0.9635, + "step": 55026 + }, + { + "epoch": 4.07, + "learning_rate": 4.9713565177695846e-06, + "loss": 1.0617, + "step": 55027 + }, + { + "epoch": 4.07, + "learning_rate": 4.971011750965021e-06, + "loss": 0.953, + "step": 55028 + }, + { + "epoch": 4.07, + "learning_rate": 4.970666992161506e-06, + "loss": 0.8978, + "step": 55029 + }, + { + "epoch": 4.07, + "learning_rate": 4.970322241359584e-06, + "loss": 0.9535, + "step": 55030 + }, + { + "epoch": 4.07, + "learning_rate": 4.969977498559812e-06, + "loss": 1.0548, + "step": 55031 + }, + { + "epoch": 4.07, + "learning_rate": 4.969632763762731e-06, + "loss": 1.0433, + "step": 55032 + }, + { + "epoch": 4.07, + "learning_rate": 4.969288036968896e-06, + "loss": 0.9822, + "step": 55033 + }, + { + "epoch": 4.07, + "learning_rate": 4.968943318178845e-06, + "loss": 1.0064, + "step": 55034 + }, + { + "epoch": 4.07, + "learning_rate": 4.968598607393138e-06, + "loss": 0.9674, + "step": 55035 + }, + { + "epoch": 4.07, + "learning_rate": 4.968253904612319e-06, + "loss": 0.96, + "step": 55036 + }, + { + "epoch": 4.07, + "learning_rate": 4.9679092098369365e-06, + "loss": 0.9673, + "step": 55037 + }, + { + "epoch": 4.07, + "learning_rate": 4.967564523067537e-06, + "loss": 1.1206, + "step": 55038 + }, + { + "epoch": 4.07, + "learning_rate": 4.967219844304667e-06, + "loss": 1.0794, + "step": 55039 + }, + { + "epoch": 4.07, + "learning_rate": 4.966875173548883e-06, + "loss": 0.9819, + "step": 55040 + }, + { + "epoch": 4.07, + "learning_rate": 4.966530510800728e-06, + "loss": 1.0332, + "step": 55041 + }, + { + "epoch": 4.07, + "learning_rate": 4.96618585606075e-06, + "loss": 1.022, + "step": 55042 + }, + { + "epoch": 4.07, + "learning_rate": 4.9658412093294965e-06, + "loss": 1.0445, + "step": 55043 + }, + { + "epoch": 4.07, + "learning_rate": 4.965496570607523e-06, + "loss": 1.0635, + "step": 55044 + }, + { + "epoch": 4.07, + "learning_rate": 4.9651519398953695e-06, + "loss": 1.0644, + "step": 55045 + }, + { + "epoch": 4.07, + "learning_rate": 4.964807317193584e-06, + "loss": 1.0857, + "step": 55046 + }, + { + "epoch": 4.07, + "learning_rate": 4.964462702502722e-06, + "loss": 1.0277, + "step": 55047 + }, + { + "epoch": 4.07, + "learning_rate": 4.964118095823327e-06, + "loss": 1.0252, + "step": 55048 + }, + { + "epoch": 4.07, + "learning_rate": 4.963773497155948e-06, + "loss": 0.8391, + "step": 55049 + }, + { + "epoch": 4.07, + "learning_rate": 4.9634289065011295e-06, + "loss": 0.9043, + "step": 55050 + }, + { + "epoch": 4.07, + "learning_rate": 4.963084323859428e-06, + "loss": 1.0381, + "step": 55051 + }, + { + "epoch": 4.07, + "learning_rate": 4.9627397492313866e-06, + "loss": 0.9983, + "step": 55052 + }, + { + "epoch": 4.07, + "learning_rate": 4.962395182617555e-06, + "loss": 1.1117, + "step": 55053 + }, + { + "epoch": 4.07, + "learning_rate": 4.96205062401848e-06, + "loss": 0.9972, + "step": 55054 + }, + { + "epoch": 4.07, + "learning_rate": 4.961706073434706e-06, + "loss": 1.0168, + "step": 55055 + }, + { + "epoch": 4.07, + "learning_rate": 4.9613615308667905e-06, + "loss": 0.9668, + "step": 55056 + }, + { + "epoch": 4.07, + "learning_rate": 4.9610169963152765e-06, + "loss": 1.0082, + "step": 55057 + }, + { + "epoch": 4.07, + "learning_rate": 4.960672469780712e-06, + "loss": 1.0226, + "step": 55058 + }, + { + "epoch": 4.07, + "learning_rate": 4.960327951263642e-06, + "loss": 1.0952, + "step": 55059 + }, + { + "epoch": 4.07, + "learning_rate": 4.959983440764621e-06, + "loss": 1.0397, + "step": 55060 + }, + { + "epoch": 4.07, + "learning_rate": 4.959638938284195e-06, + "loss": 0.9898, + "step": 55061 + }, + { + "epoch": 4.07, + "learning_rate": 4.959294443822911e-06, + "loss": 0.9305, + "step": 55062 + }, + { + "epoch": 4.07, + "learning_rate": 4.958949957381318e-06, + "loss": 0.9999, + "step": 55063 + }, + { + "epoch": 4.07, + "learning_rate": 4.958605478959959e-06, + "loss": 1.0357, + "step": 55064 + }, + { + "epoch": 4.07, + "learning_rate": 4.958261008559395e-06, + "loss": 1.0246, + "step": 55065 + }, + { + "epoch": 4.07, + "learning_rate": 4.957916546180157e-06, + "loss": 1.0378, + "step": 55066 + }, + { + "epoch": 4.07, + "learning_rate": 4.957572091822806e-06, + "loss": 1.0937, + "step": 55067 + }, + { + "epoch": 4.07, + "learning_rate": 4.9572276454878825e-06, + "loss": 0.9525, + "step": 55068 + }, + { + "epoch": 4.07, + "learning_rate": 4.9568832071759445e-06, + "loss": 1.0316, + "step": 55069 + }, + { + "epoch": 4.07, + "learning_rate": 4.956538776887525e-06, + "loss": 0.9715, + "step": 55070 + }, + { + "epoch": 4.07, + "learning_rate": 4.956194354623185e-06, + "loss": 1.139, + "step": 55071 + }, + { + "epoch": 4.07, + "learning_rate": 4.955849940383467e-06, + "loss": 0.9841, + "step": 55072 + }, + { + "epoch": 4.07, + "learning_rate": 4.95550553416892e-06, + "loss": 0.8876, + "step": 55073 + }, + { + "epoch": 4.07, + "learning_rate": 4.955161135980091e-06, + "loss": 1.072, + "step": 55074 + }, + { + "epoch": 4.07, + "learning_rate": 4.954816745817525e-06, + "loss": 0.9945, + "step": 55075 + }, + { + "epoch": 4.07, + "learning_rate": 4.954472363681777e-06, + "loss": 0.9378, + "step": 55076 + }, + { + "epoch": 4.07, + "learning_rate": 4.954127989573392e-06, + "loss": 0.9597, + "step": 55077 + }, + { + "epoch": 4.07, + "learning_rate": 4.953783623492916e-06, + "loss": 1.0077, + "step": 55078 + }, + { + "epoch": 4.07, + "learning_rate": 4.9534392654408945e-06, + "loss": 0.9528, + "step": 55079 + }, + { + "epoch": 4.07, + "learning_rate": 4.953094915417883e-06, + "loss": 1.0049, + "step": 55080 + }, + { + "epoch": 4.07, + "learning_rate": 4.952750573424427e-06, + "loss": 1.0635, + "step": 55081 + }, + { + "epoch": 4.07, + "learning_rate": 4.952406239461069e-06, + "loss": 0.9575, + "step": 55082 + }, + { + "epoch": 4.07, + "learning_rate": 4.952061913528363e-06, + "loss": 1.0486, + "step": 55083 + }, + { + "epoch": 4.07, + "learning_rate": 4.95171759562685e-06, + "loss": 1.0947, + "step": 55084 + }, + { + "epoch": 4.07, + "learning_rate": 4.951373285757086e-06, + "loss": 0.8967, + "step": 55085 + }, + { + "epoch": 4.07, + "learning_rate": 4.951028983919615e-06, + "loss": 1.0409, + "step": 55086 + }, + { + "epoch": 4.07, + "learning_rate": 4.950684690114984e-06, + "loss": 0.9788, + "step": 55087 + }, + { + "epoch": 4.07, + "learning_rate": 4.950340404343737e-06, + "loss": 1.0296, + "step": 55088 + }, + { + "epoch": 4.07, + "learning_rate": 4.9499961266064355e-06, + "loss": 1.0562, + "step": 55089 + }, + { + "epoch": 4.07, + "learning_rate": 4.949651856903609e-06, + "loss": 1.0891, + "step": 55090 + }, + { + "epoch": 4.07, + "learning_rate": 4.949307595235818e-06, + "loss": 1.0096, + "step": 55091 + }, + { + "epoch": 4.07, + "learning_rate": 4.948963341603608e-06, + "loss": 0.9358, + "step": 55092 + }, + { + "epoch": 4.07, + "learning_rate": 4.9486190960075235e-06, + "loss": 0.9225, + "step": 55093 + }, + { + "epoch": 4.07, + "learning_rate": 4.948274858448114e-06, + "loss": 1.004, + "step": 55094 + }, + { + "epoch": 4.07, + "learning_rate": 4.947930628925923e-06, + "loss": 0.9402, + "step": 55095 + }, + { + "epoch": 4.07, + "learning_rate": 4.947586407441506e-06, + "loss": 1.0018, + "step": 55096 + }, + { + "epoch": 4.07, + "learning_rate": 4.947242193995408e-06, + "loss": 1.0618, + "step": 55097 + }, + { + "epoch": 4.07, + "learning_rate": 4.946897988588173e-06, + "loss": 1.0115, + "step": 55098 + }, + { + "epoch": 4.07, + "learning_rate": 4.946553791220349e-06, + "loss": 0.8932, + "step": 55099 + }, + { + "epoch": 4.07, + "learning_rate": 4.946209601892489e-06, + "loss": 0.9934, + "step": 55100 + }, + { + "epoch": 4.07, + "learning_rate": 4.945865420605136e-06, + "loss": 1.1049, + "step": 55101 + }, + { + "epoch": 4.07, + "learning_rate": 4.94552124735884e-06, + "loss": 1.0609, + "step": 55102 + }, + { + "epoch": 4.07, + "learning_rate": 4.9451770821541475e-06, + "loss": 0.9632, + "step": 55103 + }, + { + "epoch": 4.07, + "learning_rate": 4.944832924991601e-06, + "loss": 1.0007, + "step": 55104 + }, + { + "epoch": 4.07, + "learning_rate": 4.944488775871759e-06, + "loss": 1.0698, + "step": 55105 + }, + { + "epoch": 4.07, + "learning_rate": 4.94414463479516e-06, + "loss": 0.9508, + "step": 55106 + }, + { + "epoch": 4.07, + "learning_rate": 4.943800501762357e-06, + "loss": 0.9704, + "step": 55107 + }, + { + "epoch": 4.07, + "learning_rate": 4.943456376773891e-06, + "loss": 1.044, + "step": 55108 + }, + { + "epoch": 4.07, + "learning_rate": 4.943112259830316e-06, + "loss": 1.0249, + "step": 55109 + }, + { + "epoch": 4.07, + "learning_rate": 4.942768150932182e-06, + "loss": 1.0287, + "step": 55110 + }, + { + "epoch": 4.07, + "learning_rate": 4.9424240500800235e-06, + "loss": 0.9235, + "step": 55111 + }, + { + "epoch": 4.07, + "learning_rate": 4.9420799572744e-06, + "loss": 1.0072, + "step": 55112 + }, + { + "epoch": 4.07, + "learning_rate": 4.94173587251585e-06, + "loss": 1.0455, + "step": 55113 + }, + { + "epoch": 4.07, + "learning_rate": 4.941391795804935e-06, + "loss": 1.0868, + "step": 55114 + }, + { + "epoch": 4.07, + "learning_rate": 4.941047727142185e-06, + "loss": 1.0115, + "step": 55115 + }, + { + "epoch": 4.07, + "learning_rate": 4.94070366652816e-06, + "loss": 0.9803, + "step": 55116 + }, + { + "epoch": 4.07, + "learning_rate": 4.940359613963399e-06, + "loss": 0.9887, + "step": 55117 + }, + { + "epoch": 4.07, + "learning_rate": 4.940015569448461e-06, + "loss": 0.9031, + "step": 55118 + }, + { + "epoch": 4.07, + "learning_rate": 4.939671532983882e-06, + "loss": 0.9766, + "step": 55119 + }, + { + "epoch": 4.07, + "learning_rate": 4.939327504570209e-06, + "loss": 1.0165, + "step": 55120 + }, + { + "epoch": 4.07, + "learning_rate": 4.938983484207998e-06, + "loss": 0.9452, + "step": 55121 + }, + { + "epoch": 4.07, + "learning_rate": 4.9386394718977916e-06, + "loss": 0.9459, + "step": 55122 + }, + { + "epoch": 4.07, + "learning_rate": 4.938295467640137e-06, + "loss": 0.9098, + "step": 55123 + }, + { + "epoch": 4.07, + "learning_rate": 4.937951471435578e-06, + "loss": 0.9967, + "step": 55124 + }, + { + "epoch": 4.07, + "learning_rate": 4.93760748328467e-06, + "loss": 0.9408, + "step": 55125 + }, + { + "epoch": 4.07, + "learning_rate": 4.937263503187957e-06, + "loss": 1.0198, + "step": 55126 + }, + { + "epoch": 4.07, + "learning_rate": 4.9369195311459835e-06, + "loss": 0.9203, + "step": 55127 + }, + { + "epoch": 4.07, + "learning_rate": 4.936575567159299e-06, + "loss": 1.0035, + "step": 55128 + }, + { + "epoch": 4.07, + "learning_rate": 4.936231611228448e-06, + "loss": 0.9168, + "step": 55129 + }, + { + "epoch": 4.07, + "learning_rate": 4.935887663353983e-06, + "loss": 0.9912, + "step": 55130 + }, + { + "epoch": 4.07, + "learning_rate": 4.935543723536448e-06, + "loss": 1.0201, + "step": 55131 + }, + { + "epoch": 4.07, + "learning_rate": 4.935199791776391e-06, + "loss": 0.9434, + "step": 55132 + }, + { + "epoch": 4.07, + "learning_rate": 4.934855868074354e-06, + "loss": 1.0758, + "step": 55133 + }, + { + "epoch": 4.07, + "learning_rate": 4.934511952430897e-06, + "loss": 1.0351, + "step": 55134 + }, + { + "epoch": 4.07, + "learning_rate": 4.934168044846551e-06, + "loss": 1.0005, + "step": 55135 + }, + { + "epoch": 4.07, + "learning_rate": 4.933824145321876e-06, + "loss": 1.0055, + "step": 55136 + }, + { + "epoch": 4.07, + "learning_rate": 4.933480253857409e-06, + "loss": 0.8717, + "step": 55137 + }, + { + "epoch": 4.07, + "learning_rate": 4.933136370453711e-06, + "loss": 1.0428, + "step": 55138 + }, + { + "epoch": 4.07, + "learning_rate": 4.932792495111317e-06, + "loss": 0.9869, + "step": 55139 + }, + { + "epoch": 4.07, + "learning_rate": 4.932448627830772e-06, + "loss": 1.0512, + "step": 55140 + }, + { + "epoch": 4.07, + "learning_rate": 4.9321047686126345e-06, + "loss": 1.0443, + "step": 55141 + }, + { + "epoch": 4.07, + "learning_rate": 4.931760917457446e-06, + "loss": 1.1847, + "step": 55142 + }, + { + "epoch": 4.07, + "learning_rate": 4.931417074365752e-06, + "loss": 1.0185, + "step": 55143 + }, + { + "epoch": 4.07, + "learning_rate": 4.9310732393380976e-06, + "loss": 1.0038, + "step": 55144 + }, + { + "epoch": 4.07, + "learning_rate": 4.930729412375037e-06, + "loss": 0.9709, + "step": 55145 + }, + { + "epoch": 4.07, + "learning_rate": 4.930385593477114e-06, + "loss": 0.9767, + "step": 55146 + }, + { + "epoch": 4.07, + "learning_rate": 4.9300417826448745e-06, + "loss": 0.9923, + "step": 55147 + }, + { + "epoch": 4.07, + "learning_rate": 4.929697979878866e-06, + "loss": 0.9024, + "step": 55148 + }, + { + "epoch": 4.07, + "learning_rate": 4.929354185179631e-06, + "loss": 1.0092, + "step": 55149 + }, + { + "epoch": 4.07, + "learning_rate": 4.929010398547726e-06, + "loss": 1.0279, + "step": 55150 + }, + { + "epoch": 4.08, + "learning_rate": 4.928666619983692e-06, + "loss": 1.033, + "step": 55151 + }, + { + "epoch": 4.08, + "learning_rate": 4.9283228494880766e-06, + "loss": 1.0151, + "step": 55152 + }, + { + "epoch": 4.08, + "learning_rate": 4.9279790870614244e-06, + "loss": 0.9356, + "step": 55153 + }, + { + "epoch": 4.08, + "learning_rate": 4.927635332704287e-06, + "loss": 0.9519, + "step": 55154 + }, + { + "epoch": 4.08, + "learning_rate": 4.927291586417211e-06, + "loss": 0.9814, + "step": 55155 + }, + { + "epoch": 4.08, + "learning_rate": 4.92694784820074e-06, + "loss": 1.0635, + "step": 55156 + }, + { + "epoch": 4.08, + "learning_rate": 4.926604118055423e-06, + "loss": 1.0206, + "step": 55157 + }, + { + "epoch": 4.08, + "learning_rate": 4.926260395981804e-06, + "loss": 1.0391, + "step": 55158 + }, + { + "epoch": 4.08, + "learning_rate": 4.925916681980437e-06, + "loss": 0.9506, + "step": 55159 + }, + { + "epoch": 4.08, + "learning_rate": 4.925572976051858e-06, + "loss": 1.0139, + "step": 55160 + }, + { + "epoch": 4.08, + "learning_rate": 4.925229278196623e-06, + "loss": 1.2013, + "step": 55161 + }, + { + "epoch": 4.08, + "learning_rate": 4.924885588415272e-06, + "loss": 0.9064, + "step": 55162 + }, + { + "epoch": 4.08, + "learning_rate": 4.924541906708364e-06, + "loss": 0.9737, + "step": 55163 + }, + { + "epoch": 4.08, + "learning_rate": 4.924198233076427e-06, + "loss": 0.9988, + "step": 55164 + }, + { + "epoch": 4.08, + "learning_rate": 4.9238545675200245e-06, + "loss": 0.997, + "step": 55165 + }, + { + "epoch": 4.08, + "learning_rate": 4.923510910039695e-06, + "loss": 1.0802, + "step": 55166 + }, + { + "epoch": 4.08, + "learning_rate": 4.923167260635988e-06, + "loss": 0.9938, + "step": 55167 + }, + { + "epoch": 4.08, + "learning_rate": 4.922823619309449e-06, + "loss": 0.9674, + "step": 55168 + }, + { + "epoch": 4.08, + "learning_rate": 4.922479986060621e-06, + "loss": 1.0438, + "step": 55169 + }, + { + "epoch": 4.08, + "learning_rate": 4.9221363608900584e-06, + "loss": 0.9383, + "step": 55170 + }, + { + "epoch": 4.08, + "learning_rate": 4.9217927437983045e-06, + "loss": 1.1049, + "step": 55171 + }, + { + "epoch": 4.08, + "learning_rate": 4.9214491347859065e-06, + "loss": 0.9702, + "step": 55172 + }, + { + "epoch": 4.08, + "learning_rate": 4.921105533853404e-06, + "loss": 0.9994, + "step": 55173 + }, + { + "epoch": 4.08, + "learning_rate": 4.920761941001356e-06, + "loss": 1.049, + "step": 55174 + }, + { + "epoch": 4.08, + "learning_rate": 4.920418356230302e-06, + "loss": 1.0994, + "step": 55175 + }, + { + "epoch": 4.08, + "learning_rate": 4.92007477954079e-06, + "loss": 1.1105, + "step": 55176 + }, + { + "epoch": 4.08, + "learning_rate": 4.919731210933368e-06, + "loss": 1.04, + "step": 55177 + }, + { + "epoch": 4.08, + "learning_rate": 4.9193876504085746e-06, + "loss": 0.8941, + "step": 55178 + }, + { + "epoch": 4.08, + "learning_rate": 4.91904409796697e-06, + "loss": 1.0219, + "step": 55179 + }, + { + "epoch": 4.08, + "learning_rate": 4.9187005536090885e-06, + "loss": 0.9389, + "step": 55180 + }, + { + "epoch": 4.08, + "learning_rate": 4.918357017335485e-06, + "loss": 0.9973, + "step": 55181 + }, + { + "epoch": 4.08, + "learning_rate": 4.918013489146698e-06, + "loss": 0.9061, + "step": 55182 + }, + { + "epoch": 4.08, + "learning_rate": 4.917669969043286e-06, + "loss": 1.0143, + "step": 55183 + }, + { + "epoch": 4.08, + "learning_rate": 4.9173264570257854e-06, + "loss": 0.9704, + "step": 55184 + }, + { + "epoch": 4.08, + "learning_rate": 4.916982953094742e-06, + "loss": 1.0759, + "step": 55185 + }, + { + "epoch": 4.08, + "learning_rate": 4.916639457250709e-06, + "loss": 1.0199, + "step": 55186 + }, + { + "epoch": 4.08, + "learning_rate": 4.916295969494232e-06, + "loss": 0.9977, + "step": 55187 + }, + { + "epoch": 4.08, + "learning_rate": 4.915952489825853e-06, + "loss": 1.0107, + "step": 55188 + }, + { + "epoch": 4.08, + "learning_rate": 4.9156090182461184e-06, + "loss": 0.9802, + "step": 55189 + }, + { + "epoch": 4.08, + "learning_rate": 4.915265554755579e-06, + "loss": 0.9881, + "step": 55190 + }, + { + "epoch": 4.08, + "learning_rate": 4.9149220993547806e-06, + "loss": 1.0121, + "step": 55191 + }, + { + "epoch": 4.08, + "learning_rate": 4.91457865204427e-06, + "loss": 1.0209, + "step": 55192 + }, + { + "epoch": 4.08, + "learning_rate": 4.914235212824586e-06, + "loss": 0.9047, + "step": 55193 + }, + { + "epoch": 4.08, + "learning_rate": 4.913891781696285e-06, + "loss": 1.086, + "step": 55194 + }, + { + "epoch": 4.08, + "learning_rate": 4.913548358659911e-06, + "loss": 1.055, + "step": 55195 + }, + { + "epoch": 4.08, + "learning_rate": 4.913204943716007e-06, + "loss": 0.8314, + "step": 55196 + }, + { + "epoch": 4.08, + "learning_rate": 4.912861536865122e-06, + "loss": 0.9792, + "step": 55197 + }, + { + "epoch": 4.08, + "learning_rate": 4.912518138107798e-06, + "loss": 0.9775, + "step": 55198 + }, + { + "epoch": 4.08, + "learning_rate": 4.912174747444588e-06, + "loss": 0.9893, + "step": 55199 + }, + { + "epoch": 4.08, + "learning_rate": 4.911831364876035e-06, + "loss": 1.0277, + "step": 55200 + }, + { + "epoch": 4.08, + "learning_rate": 4.9114879904026856e-06, + "loss": 1.0817, + "step": 55201 + }, + { + "epoch": 4.08, + "learning_rate": 4.911144624025083e-06, + "loss": 0.9681, + "step": 55202 + }, + { + "epoch": 4.08, + "learning_rate": 4.91080126574378e-06, + "loss": 0.9036, + "step": 55203 + }, + { + "epoch": 4.08, + "learning_rate": 4.910457915559323e-06, + "loss": 0.979, + "step": 55204 + }, + { + "epoch": 4.08, + "learning_rate": 4.910114573472246e-06, + "loss": 1.028, + "step": 55205 + }, + { + "epoch": 4.08, + "learning_rate": 4.90977123948311e-06, + "loss": 0.9652, + "step": 55206 + }, + { + "epoch": 4.08, + "learning_rate": 4.9094279135924495e-06, + "loss": 1.0309, + "step": 55207 + }, + { + "epoch": 4.08, + "learning_rate": 4.909084595800825e-06, + "loss": 1.0232, + "step": 55208 + }, + { + "epoch": 4.08, + "learning_rate": 4.908741286108766e-06, + "loss": 1.0942, + "step": 55209 + }, + { + "epoch": 4.08, + "learning_rate": 4.9083979845168305e-06, + "loss": 0.9868, + "step": 55210 + }, + { + "epoch": 4.08, + "learning_rate": 4.9080546910255615e-06, + "loss": 0.983, + "step": 55211 + }, + { + "epoch": 4.08, + "learning_rate": 4.907711405635504e-06, + "loss": 0.9893, + "step": 55212 + }, + { + "epoch": 4.08, + "learning_rate": 4.907368128347204e-06, + "loss": 1.0196, + "step": 55213 + }, + { + "epoch": 4.08, + "learning_rate": 4.907024859161206e-06, + "loss": 0.9768, + "step": 55214 + }, + { + "epoch": 4.08, + "learning_rate": 4.906681598078061e-06, + "loss": 0.9135, + "step": 55215 + }, + { + "epoch": 4.08, + "learning_rate": 4.906338345098314e-06, + "loss": 0.9172, + "step": 55216 + }, + { + "epoch": 4.08, + "learning_rate": 4.905995100222509e-06, + "loss": 1.0395, + "step": 55217 + }, + { + "epoch": 4.08, + "learning_rate": 4.905651863451191e-06, + "loss": 1.1236, + "step": 55218 + }, + { + "epoch": 4.08, + "learning_rate": 4.90530863478491e-06, + "loss": 0.9616, + "step": 55219 + }, + { + "epoch": 4.08, + "learning_rate": 4.90496541422421e-06, + "loss": 1.0208, + "step": 55220 + }, + { + "epoch": 4.08, + "learning_rate": 4.904622201769638e-06, + "loss": 1.0475, + "step": 55221 + }, + { + "epoch": 4.08, + "learning_rate": 4.904278997421738e-06, + "loss": 0.9698, + "step": 55222 + }, + { + "epoch": 4.08, + "learning_rate": 4.9039358011810545e-06, + "loss": 1.08, + "step": 55223 + }, + { + "epoch": 4.08, + "learning_rate": 4.903592613048144e-06, + "loss": 1.0913, + "step": 55224 + }, + { + "epoch": 4.08, + "learning_rate": 4.903249433023536e-06, + "loss": 0.9786, + "step": 55225 + }, + { + "epoch": 4.08, + "learning_rate": 4.902906261107789e-06, + "loss": 0.9258, + "step": 55226 + }, + { + "epoch": 4.08, + "learning_rate": 4.902563097301443e-06, + "loss": 1.034, + "step": 55227 + }, + { + "epoch": 4.08, + "learning_rate": 4.902219941605053e-06, + "loss": 1.0347, + "step": 55228 + }, + { + "epoch": 4.08, + "learning_rate": 4.9018767940191505e-06, + "loss": 0.9815, + "step": 55229 + }, + { + "epoch": 4.08, + "learning_rate": 4.901533654544292e-06, + "loss": 0.9854, + "step": 55230 + }, + { + "epoch": 4.08, + "learning_rate": 4.901190523181022e-06, + "loss": 0.9947, + "step": 55231 + }, + { + "epoch": 4.08, + "learning_rate": 4.900847399929883e-06, + "loss": 1.0004, + "step": 55232 + }, + { + "epoch": 4.08, + "learning_rate": 4.900504284791425e-06, + "loss": 1.0245, + "step": 55233 + }, + { + "epoch": 4.08, + "learning_rate": 4.900161177766186e-06, + "loss": 0.9728, + "step": 55234 + }, + { + "epoch": 4.08, + "learning_rate": 4.899818078854723e-06, + "loss": 0.9918, + "step": 55235 + }, + { + "epoch": 4.08, + "learning_rate": 4.899474988057575e-06, + "loss": 0.8694, + "step": 55236 + }, + { + "epoch": 4.08, + "learning_rate": 4.899131905375291e-06, + "loss": 1.0107, + "step": 55237 + }, + { + "epoch": 4.08, + "learning_rate": 4.898788830808411e-06, + "loss": 0.978, + "step": 55238 + }, + { + "epoch": 4.08, + "learning_rate": 4.898445764357488e-06, + "loss": 1.0217, + "step": 55239 + }, + { + "epoch": 4.08, + "learning_rate": 4.898102706023065e-06, + "loss": 0.9179, + "step": 55240 + }, + { + "epoch": 4.08, + "learning_rate": 4.897759655805687e-06, + "loss": 0.9851, + "step": 55241 + }, + { + "epoch": 4.08, + "learning_rate": 4.897416613705901e-06, + "loss": 1.0826, + "step": 55242 + }, + { + "epoch": 4.08, + "learning_rate": 4.897073579724249e-06, + "loss": 0.9733, + "step": 55243 + }, + { + "epoch": 4.08, + "learning_rate": 4.896730553861283e-06, + "loss": 0.8691, + "step": 55244 + }, + { + "epoch": 4.08, + "learning_rate": 4.896387536117546e-06, + "loss": 0.9985, + "step": 55245 + }, + { + "epoch": 4.08, + "learning_rate": 4.896044526493583e-06, + "loss": 1.0614, + "step": 55246 + }, + { + "epoch": 4.08, + "learning_rate": 4.895701524989936e-06, + "loss": 0.9661, + "step": 55247 + }, + { + "epoch": 4.08, + "learning_rate": 4.895358531607159e-06, + "loss": 1.0556, + "step": 55248 + }, + { + "epoch": 4.08, + "learning_rate": 4.8950155463457936e-06, + "loss": 0.9974, + "step": 55249 + }, + { + "epoch": 4.08, + "learning_rate": 4.894672569206387e-06, + "loss": 0.9424, + "step": 55250 + }, + { + "epoch": 4.08, + "learning_rate": 4.894329600189481e-06, + "loss": 1.1621, + "step": 55251 + }, + { + "epoch": 4.08, + "learning_rate": 4.89398663929562e-06, + "loss": 1.0132, + "step": 55252 + }, + { + "epoch": 4.08, + "learning_rate": 4.893643686525362e-06, + "loss": 1.0989, + "step": 55253 + }, + { + "epoch": 4.08, + "learning_rate": 4.893300741879236e-06, + "loss": 0.9842, + "step": 55254 + }, + { + "epoch": 4.08, + "learning_rate": 4.892957805357798e-06, + "loss": 0.9662, + "step": 55255 + }, + { + "epoch": 4.08, + "learning_rate": 4.892614876961592e-06, + "loss": 0.934, + "step": 55256 + }, + { + "epoch": 4.08, + "learning_rate": 4.892271956691164e-06, + "loss": 1.0168, + "step": 55257 + }, + { + "epoch": 4.08, + "learning_rate": 4.8919290445470524e-06, + "loss": 1.1537, + "step": 55258 + }, + { + "epoch": 4.08, + "learning_rate": 4.891586140529812e-06, + "loss": 0.9405, + "step": 55259 + }, + { + "epoch": 4.08, + "learning_rate": 4.891243244639987e-06, + "loss": 1.0468, + "step": 55260 + }, + { + "epoch": 4.08, + "learning_rate": 4.89090035687812e-06, + "loss": 0.9414, + "step": 55261 + }, + { + "epoch": 4.08, + "learning_rate": 4.890557477244757e-06, + "loss": 1.0636, + "step": 55262 + }, + { + "epoch": 4.08, + "learning_rate": 4.890214605740441e-06, + "loss": 0.9904, + "step": 55263 + }, + { + "epoch": 4.08, + "learning_rate": 4.889871742365724e-06, + "loss": 0.9485, + "step": 55264 + }, + { + "epoch": 4.08, + "learning_rate": 4.8895288871211475e-06, + "loss": 0.9122, + "step": 55265 + }, + { + "epoch": 4.08, + "learning_rate": 4.889186040007258e-06, + "loss": 0.9762, + "step": 55266 + }, + { + "epoch": 4.08, + "learning_rate": 4.888843201024595e-06, + "loss": 0.9887, + "step": 55267 + }, + { + "epoch": 4.08, + "learning_rate": 4.888500370173715e-06, + "loss": 0.9235, + "step": 55268 + }, + { + "epoch": 4.08, + "learning_rate": 4.888157547455161e-06, + "loss": 1.0676, + "step": 55269 + }, + { + "epoch": 4.08, + "learning_rate": 4.887814732869467e-06, + "loss": 1.0421, + "step": 55270 + }, + { + "epoch": 4.08, + "learning_rate": 4.887471926417191e-06, + "loss": 1.0487, + "step": 55271 + }, + { + "epoch": 4.08, + "learning_rate": 4.887129128098871e-06, + "loss": 0.9907, + "step": 55272 + }, + { + "epoch": 4.08, + "learning_rate": 4.886786337915062e-06, + "loss": 0.984, + "step": 55273 + }, + { + "epoch": 4.08, + "learning_rate": 4.8864435558662945e-06, + "loss": 1.0053, + "step": 55274 + }, + { + "epoch": 4.08, + "learning_rate": 4.886100781953127e-06, + "loss": 0.9271, + "step": 55275 + }, + { + "epoch": 4.08, + "learning_rate": 4.8857580161760965e-06, + "loss": 0.9191, + "step": 55276 + }, + { + "epoch": 4.08, + "learning_rate": 4.8854152585357584e-06, + "loss": 1.0435, + "step": 55277 + }, + { + "epoch": 4.08, + "learning_rate": 4.885072509032648e-06, + "loss": 0.8855, + "step": 55278 + }, + { + "epoch": 4.08, + "learning_rate": 4.884729767667311e-06, + "loss": 0.953, + "step": 55279 + }, + { + "epoch": 4.08, + "learning_rate": 4.884387034440299e-06, + "loss": 0.9725, + "step": 55280 + }, + { + "epoch": 4.08, + "learning_rate": 4.884044309352154e-06, + "loss": 1.0108, + "step": 55281 + }, + { + "epoch": 4.08, + "learning_rate": 4.883701592403422e-06, + "loss": 1.0313, + "step": 55282 + }, + { + "epoch": 4.08, + "learning_rate": 4.883358883594642e-06, + "loss": 0.9304, + "step": 55283 + }, + { + "epoch": 4.08, + "learning_rate": 4.88301618292637e-06, + "loss": 1.0629, + "step": 55284 + }, + { + "epoch": 4.08, + "learning_rate": 4.882673490399146e-06, + "loss": 1.0015, + "step": 55285 + }, + { + "epoch": 4.09, + "learning_rate": 4.882330806013514e-06, + "loss": 1.126, + "step": 55286 + }, + { + "epoch": 4.09, + "learning_rate": 4.881988129770022e-06, + "loss": 0.8654, + "step": 55287 + }, + { + "epoch": 4.09, + "learning_rate": 4.881645461669209e-06, + "loss": 0.884, + "step": 55288 + }, + { + "epoch": 4.09, + "learning_rate": 4.881302801711629e-06, + "loss": 1.0912, + "step": 55289 + }, + { + "epoch": 4.09, + "learning_rate": 4.880960149897823e-06, + "loss": 1.0299, + "step": 55290 + }, + { + "epoch": 4.09, + "learning_rate": 4.880617506228336e-06, + "loss": 1.0577, + "step": 55291 + }, + { + "epoch": 4.09, + "learning_rate": 4.880274870703709e-06, + "loss": 0.9671, + "step": 55292 + }, + { + "epoch": 4.09, + "learning_rate": 4.8799322433244964e-06, + "loss": 0.9286, + "step": 55293 + }, + { + "epoch": 4.09, + "learning_rate": 4.879589624091238e-06, + "loss": 0.9834, + "step": 55294 + }, + { + "epoch": 4.09, + "learning_rate": 4.879247013004479e-06, + "loss": 1.0905, + "step": 55295 + }, + { + "epoch": 4.09, + "learning_rate": 4.878904410064761e-06, + "loss": 1.0286, + "step": 55296 + }, + { + "epoch": 4.09, + "learning_rate": 4.8785618152726365e-06, + "loss": 1.0178, + "step": 55297 + }, + { + "epoch": 4.09, + "learning_rate": 4.87821922862865e-06, + "loss": 0.8542, + "step": 55298 + }, + { + "epoch": 4.09, + "learning_rate": 4.8778766501333375e-06, + "loss": 1.0749, + "step": 55299 + }, + { + "epoch": 4.09, + "learning_rate": 4.877534079787253e-06, + "loss": 0.8833, + "step": 55300 + }, + { + "epoch": 4.09, + "learning_rate": 4.877191517590938e-06, + "loss": 0.9454, + "step": 55301 + }, + { + "epoch": 4.09, + "learning_rate": 4.876848963544938e-06, + "loss": 1.1005, + "step": 55302 + }, + { + "epoch": 4.09, + "learning_rate": 4.876506417649794e-06, + "loss": 1.0262, + "step": 55303 + }, + { + "epoch": 4.09, + "learning_rate": 4.876163879906059e-06, + "loss": 1.1249, + "step": 55304 + }, + { + "epoch": 4.09, + "learning_rate": 4.875821350314275e-06, + "loss": 0.9719, + "step": 55305 + }, + { + "epoch": 4.09, + "learning_rate": 4.875478828874985e-06, + "loss": 0.9944, + "step": 55306 + }, + { + "epoch": 4.09, + "learning_rate": 4.875136315588735e-06, + "loss": 1.0116, + "step": 55307 + }, + { + "epoch": 4.09, + "learning_rate": 4.874793810456067e-06, + "loss": 0.9813, + "step": 55308 + }, + { + "epoch": 4.09, + "learning_rate": 4.8744513134775305e-06, + "loss": 1.0006, + "step": 55309 + }, + { + "epoch": 4.09, + "learning_rate": 4.874108824653671e-06, + "loss": 0.9068, + "step": 55310 + }, + { + "epoch": 4.09, + "learning_rate": 4.873766343985029e-06, + "loss": 1.0177, + "step": 55311 + }, + { + "epoch": 4.09, + "learning_rate": 4.8734238714721485e-06, + "loss": 1.0892, + "step": 55312 + }, + { + "epoch": 4.09, + "learning_rate": 4.873081407115581e-06, + "loss": 0.9504, + "step": 55313 + }, + { + "epoch": 4.09, + "learning_rate": 4.872738950915867e-06, + "loss": 0.9645, + "step": 55314 + }, + { + "epoch": 4.09, + "learning_rate": 4.872396502873552e-06, + "loss": 1.0796, + "step": 55315 + }, + { + "epoch": 4.09, + "learning_rate": 4.872054062989182e-06, + "loss": 1.1672, + "step": 55316 + }, + { + "epoch": 4.09, + "learning_rate": 4.871711631263296e-06, + "loss": 0.9534, + "step": 55317 + }, + { + "epoch": 4.09, + "learning_rate": 4.871369207696451e-06, + "loss": 1.0499, + "step": 55318 + }, + { + "epoch": 4.09, + "learning_rate": 4.8710267922891764e-06, + "loss": 0.9462, + "step": 55319 + }, + { + "epoch": 4.09, + "learning_rate": 4.870684385042029e-06, + "loss": 0.9124, + "step": 55320 + }, + { + "epoch": 4.09, + "learning_rate": 4.870341985955545e-06, + "loss": 0.9557, + "step": 55321 + }, + { + "epoch": 4.09, + "learning_rate": 4.8699995950302826e-06, + "loss": 0.9938, + "step": 55322 + }, + { + "epoch": 4.09, + "learning_rate": 4.869657212266768e-06, + "loss": 1.0486, + "step": 55323 + }, + { + "epoch": 4.09, + "learning_rate": 4.869314837665561e-06, + "loss": 0.9596, + "step": 55324 + }, + { + "epoch": 4.09, + "learning_rate": 4.868972471227199e-06, + "loss": 1.1054, + "step": 55325 + }, + { + "epoch": 4.09, + "learning_rate": 4.86863011295223e-06, + "loss": 1.1288, + "step": 55326 + }, + { + "epoch": 4.09, + "learning_rate": 4.868287762841195e-06, + "loss": 0.9626, + "step": 55327 + }, + { + "epoch": 4.09, + "learning_rate": 4.867945420894638e-06, + "loss": 0.9821, + "step": 55328 + }, + { + "epoch": 4.09, + "learning_rate": 4.867603087113111e-06, + "loss": 0.9018, + "step": 55329 + }, + { + "epoch": 4.09, + "learning_rate": 4.8672607614971535e-06, + "loss": 1.0492, + "step": 55330 + }, + { + "epoch": 4.09, + "learning_rate": 4.8669184440473105e-06, + "loss": 1.0476, + "step": 55331 + }, + { + "epoch": 4.09, + "learning_rate": 4.866576134764123e-06, + "loss": 0.9892, + "step": 55332 + }, + { + "epoch": 4.09, + "learning_rate": 4.866233833648143e-06, + "loss": 0.8505, + "step": 55333 + }, + { + "epoch": 4.09, + "learning_rate": 4.8658915406999115e-06, + "loss": 1.1224, + "step": 55334 + }, + { + "epoch": 4.09, + "learning_rate": 4.865549255919973e-06, + "loss": 0.9769, + "step": 55335 + }, + { + "epoch": 4.09, + "learning_rate": 4.865206979308873e-06, + "loss": 1.0541, + "step": 55336 + }, + { + "epoch": 4.09, + "learning_rate": 4.864864710867151e-06, + "loss": 0.859, + "step": 55337 + }, + { + "epoch": 4.09, + "learning_rate": 4.864522450595359e-06, + "loss": 1.0192, + "step": 55338 + }, + { + "epoch": 4.09, + "learning_rate": 4.86418019849404e-06, + "loss": 1.0579, + "step": 55339 + }, + { + "epoch": 4.09, + "learning_rate": 4.8638379545637345e-06, + "loss": 1.0802, + "step": 55340 + }, + { + "epoch": 4.09, + "learning_rate": 4.863495718804987e-06, + "loss": 1.125, + "step": 55341 + }, + { + "epoch": 4.09, + "learning_rate": 4.863153491218348e-06, + "loss": 0.8941, + "step": 55342 + }, + { + "epoch": 4.09, + "learning_rate": 4.862811271804362e-06, + "loss": 1.0528, + "step": 55343 + }, + { + "epoch": 4.09, + "learning_rate": 4.862469060563563e-06, + "loss": 1.0415, + "step": 55344 + }, + { + "epoch": 4.09, + "learning_rate": 4.862126857496504e-06, + "loss": 1.0137, + "step": 55345 + }, + { + "epoch": 4.09, + "learning_rate": 4.861784662603729e-06, + "loss": 1.0829, + "step": 55346 + }, + { + "epoch": 4.09, + "learning_rate": 4.861442475885781e-06, + "loss": 0.9419, + "step": 55347 + }, + { + "epoch": 4.09, + "learning_rate": 4.861100297343201e-06, + "loss": 1.0515, + "step": 55348 + }, + { + "epoch": 4.09, + "learning_rate": 4.86075812697654e-06, + "loss": 1.0401, + "step": 55349 + }, + { + "epoch": 4.09, + "learning_rate": 4.860415964786339e-06, + "loss": 0.9241, + "step": 55350 + }, + { + "epoch": 4.09, + "learning_rate": 4.860073810773144e-06, + "loss": 0.9893, + "step": 55351 + }, + { + "epoch": 4.09, + "learning_rate": 4.8597316649374934e-06, + "loss": 0.9429, + "step": 55352 + }, + { + "epoch": 4.09, + "learning_rate": 4.8593895272799395e-06, + "loss": 0.8973, + "step": 55353 + }, + { + "epoch": 4.09, + "learning_rate": 4.859047397801024e-06, + "loss": 0.9634, + "step": 55354 + }, + { + "epoch": 4.09, + "learning_rate": 4.858705276501291e-06, + "loss": 0.9553, + "step": 55355 + }, + { + "epoch": 4.09, + "learning_rate": 4.858363163381284e-06, + "loss": 1.0123, + "step": 55356 + }, + { + "epoch": 4.09, + "learning_rate": 4.858021058441543e-06, + "loss": 1.0988, + "step": 55357 + }, + { + "epoch": 4.09, + "learning_rate": 4.857678961682622e-06, + "loss": 1.0167, + "step": 55358 + }, + { + "epoch": 4.09, + "learning_rate": 4.857336873105059e-06, + "loss": 0.9206, + "step": 55359 + }, + { + "epoch": 4.09, + "learning_rate": 4.856994792709399e-06, + "loss": 1.0522, + "step": 55360 + }, + { + "epoch": 4.09, + "learning_rate": 4.856652720496184e-06, + "loss": 0.9761, + "step": 55361 + }, + { + "epoch": 4.09, + "learning_rate": 4.856310656465965e-06, + "loss": 1.0636, + "step": 55362 + }, + { + "epoch": 4.09, + "learning_rate": 4.8559686006192855e-06, + "loss": 1.0692, + "step": 55363 + }, + { + "epoch": 4.09, + "learning_rate": 4.855626552956679e-06, + "loss": 1.0667, + "step": 55364 + }, + { + "epoch": 4.09, + "learning_rate": 4.855284513478701e-06, + "loss": 1.0044, + "step": 55365 + }, + { + "epoch": 4.09, + "learning_rate": 4.854942482185887e-06, + "loss": 1.0176, + "step": 55366 + }, + { + "epoch": 4.09, + "learning_rate": 4.8546004590787945e-06, + "loss": 0.9844, + "step": 55367 + }, + { + "epoch": 4.09, + "learning_rate": 4.854258444157951e-06, + "loss": 0.9753, + "step": 55368 + }, + { + "epoch": 4.09, + "learning_rate": 4.8539164374239125e-06, + "loss": 1.0483, + "step": 55369 + }, + { + "epoch": 4.09, + "learning_rate": 4.8535744388772155e-06, + "loss": 0.974, + "step": 55370 + }, + { + "epoch": 4.09, + "learning_rate": 4.853232448518416e-06, + "loss": 1.0737, + "step": 55371 + }, + { + "epoch": 4.09, + "learning_rate": 4.852890466348045e-06, + "loss": 0.992, + "step": 55372 + }, + { + "epoch": 4.09, + "learning_rate": 4.852548492366649e-06, + "loss": 0.9953, + "step": 55373 + }, + { + "epoch": 4.09, + "learning_rate": 4.852206526574777e-06, + "loss": 1.0168, + "step": 55374 + }, + { + "epoch": 4.09, + "learning_rate": 4.851864568972973e-06, + "loss": 0.948, + "step": 55375 + }, + { + "epoch": 4.09, + "learning_rate": 4.851522619561777e-06, + "loss": 1.0669, + "step": 55376 + }, + { + "epoch": 4.09, + "learning_rate": 4.851180678341732e-06, + "loss": 0.9938, + "step": 55377 + }, + { + "epoch": 4.09, + "learning_rate": 4.850838745313389e-06, + "loss": 1.0763, + "step": 55378 + }, + { + "epoch": 4.09, + "learning_rate": 4.8504968204772864e-06, + "loss": 1.013, + "step": 55379 + }, + { + "epoch": 4.09, + "learning_rate": 4.850154903833971e-06, + "loss": 1.122, + "step": 55380 + }, + { + "epoch": 4.09, + "learning_rate": 4.849812995383984e-06, + "loss": 0.9429, + "step": 55381 + }, + { + "epoch": 4.09, + "learning_rate": 4.849471095127869e-06, + "loss": 0.9962, + "step": 55382 + }, + { + "epoch": 4.09, + "learning_rate": 4.849129203066175e-06, + "loss": 0.9375, + "step": 55383 + }, + { + "epoch": 4.09, + "learning_rate": 4.848787319199442e-06, + "loss": 0.9856, + "step": 55384 + }, + { + "epoch": 4.09, + "learning_rate": 4.8484454435282155e-06, + "loss": 1.063, + "step": 55385 + }, + { + "epoch": 4.09, + "learning_rate": 4.848103576053034e-06, + "loss": 0.9763, + "step": 55386 + }, + { + "epoch": 4.09, + "learning_rate": 4.847761716774454e-06, + "loss": 0.9849, + "step": 55387 + }, + { + "epoch": 4.09, + "learning_rate": 4.847419865693004e-06, + "loss": 1.0125, + "step": 55388 + }, + { + "epoch": 4.09, + "learning_rate": 4.847078022809238e-06, + "loss": 0.8663, + "step": 55389 + }, + { + "epoch": 4.09, + "learning_rate": 4.846736188123698e-06, + "loss": 0.9632, + "step": 55390 + }, + { + "epoch": 4.09, + "learning_rate": 4.846394361636928e-06, + "loss": 1.109, + "step": 55391 + }, + { + "epoch": 4.09, + "learning_rate": 4.84605254334947e-06, + "loss": 0.9376, + "step": 55392 + }, + { + "epoch": 4.09, + "learning_rate": 4.845710733261864e-06, + "loss": 0.9667, + "step": 55393 + }, + { + "epoch": 4.09, + "learning_rate": 4.845368931374663e-06, + "loss": 1.1837, + "step": 55394 + }, + { + "epoch": 4.09, + "learning_rate": 4.845027137688406e-06, + "loss": 0.9796, + "step": 55395 + }, + { + "epoch": 4.09, + "learning_rate": 4.8446853522036375e-06, + "loss": 1.0169, + "step": 55396 + }, + { + "epoch": 4.09, + "learning_rate": 4.844343574920897e-06, + "loss": 0.9767, + "step": 55397 + }, + { + "epoch": 4.09, + "learning_rate": 4.8440018058407355e-06, + "loss": 0.8878, + "step": 55398 + }, + { + "epoch": 4.09, + "learning_rate": 4.843660044963694e-06, + "loss": 0.9544, + "step": 55399 + }, + { + "epoch": 4.09, + "learning_rate": 4.843318292290315e-06, + "loss": 0.877, + "step": 55400 + }, + { + "epoch": 4.09, + "learning_rate": 4.842976547821143e-06, + "loss": 0.8902, + "step": 55401 + }, + { + "epoch": 4.09, + "learning_rate": 4.842634811556718e-06, + "loss": 1.0472, + "step": 55402 + }, + { + "epoch": 4.09, + "learning_rate": 4.842293083497591e-06, + "loss": 0.9577, + "step": 55403 + }, + { + "epoch": 4.09, + "learning_rate": 4.841951363644302e-06, + "loss": 1.0104, + "step": 55404 + }, + { + "epoch": 4.09, + "learning_rate": 4.841609651997394e-06, + "loss": 0.9783, + "step": 55405 + }, + { + "epoch": 4.09, + "learning_rate": 4.841267948557408e-06, + "loss": 1.0908, + "step": 55406 + }, + { + "epoch": 4.09, + "learning_rate": 4.840926253324894e-06, + "loss": 1.0621, + "step": 55407 + }, + { + "epoch": 4.09, + "learning_rate": 4.840584566300393e-06, + "loss": 0.9716, + "step": 55408 + }, + { + "epoch": 4.09, + "learning_rate": 4.840242887484448e-06, + "loss": 1.0941, + "step": 55409 + }, + { + "epoch": 4.09, + "learning_rate": 4.8399012168776026e-06, + "loss": 0.9538, + "step": 55410 + }, + { + "epoch": 4.09, + "learning_rate": 4.839559554480397e-06, + "loss": 0.9609, + "step": 55411 + }, + { + "epoch": 4.09, + "learning_rate": 4.839217900293386e-06, + "loss": 1.0236, + "step": 55412 + }, + { + "epoch": 4.09, + "learning_rate": 4.838876254317098e-06, + "loss": 0.9276, + "step": 55413 + }, + { + "epoch": 4.09, + "learning_rate": 4.838534616552087e-06, + "loss": 0.9924, + "step": 55414 + }, + { + "epoch": 4.09, + "learning_rate": 4.838192986998891e-06, + "loss": 1.0139, + "step": 55415 + }, + { + "epoch": 4.09, + "learning_rate": 4.837851365658063e-06, + "loss": 0.8354, + "step": 55416 + }, + { + "epoch": 4.09, + "learning_rate": 4.837509752530132e-06, + "loss": 1.1709, + "step": 55417 + }, + { + "epoch": 4.09, + "learning_rate": 4.837168147615653e-06, + "loss": 0.9972, + "step": 55418 + }, + { + "epoch": 4.09, + "learning_rate": 4.836826550915165e-06, + "loss": 1.105, + "step": 55419 + }, + { + "epoch": 4.09, + "learning_rate": 4.836484962429213e-06, + "loss": 1.1505, + "step": 55420 + }, + { + "epoch": 4.1, + "learning_rate": 4.8361433821583395e-06, + "loss": 1.0557, + "step": 55421 + }, + { + "epoch": 4.1, + "learning_rate": 4.835801810103084e-06, + "loss": 0.9993, + "step": 55422 + }, + { + "epoch": 4.1, + "learning_rate": 4.835460246263996e-06, + "loss": 1.0899, + "step": 55423 + }, + { + "epoch": 4.1, + "learning_rate": 4.835118690641619e-06, + "loss": 0.9626, + "step": 55424 + }, + { + "epoch": 4.1, + "learning_rate": 4.834777143236495e-06, + "loss": 0.9583, + "step": 55425 + }, + { + "epoch": 4.1, + "learning_rate": 4.834435604049162e-06, + "loss": 1.1123, + "step": 55426 + }, + { + "epoch": 4.1, + "learning_rate": 4.83409407308017e-06, + "loss": 0.9218, + "step": 55427 + }, + { + "epoch": 4.1, + "learning_rate": 4.833752550330062e-06, + "loss": 0.9778, + "step": 55428 + }, + { + "epoch": 4.1, + "learning_rate": 4.83341103579938e-06, + "loss": 1.0406, + "step": 55429 + }, + { + "epoch": 4.1, + "learning_rate": 4.833069529488666e-06, + "loss": 1.0395, + "step": 55430 + }, + { + "epoch": 4.1, + "learning_rate": 4.832728031398463e-06, + "loss": 0.9574, + "step": 55431 + }, + { + "epoch": 4.1, + "learning_rate": 4.832386541529322e-06, + "loss": 0.9662, + "step": 55432 + }, + { + "epoch": 4.1, + "learning_rate": 4.832045059881771e-06, + "loss": 1.0871, + "step": 55433 + }, + { + "epoch": 4.1, + "learning_rate": 4.831703586456368e-06, + "loss": 0.9318, + "step": 55434 + }, + { + "epoch": 4.1, + "learning_rate": 4.831362121253647e-06, + "loss": 0.9905, + "step": 55435 + }, + { + "epoch": 4.1, + "learning_rate": 4.83102066427416e-06, + "loss": 0.9647, + "step": 55436 + }, + { + "epoch": 4.1, + "learning_rate": 4.830679215518444e-06, + "loss": 0.9608, + "step": 55437 + }, + { + "epoch": 4.1, + "learning_rate": 4.830337774987039e-06, + "loss": 1.0321, + "step": 55438 + }, + { + "epoch": 4.1, + "learning_rate": 4.829996342680496e-06, + "loss": 0.992, + "step": 55439 + }, + { + "epoch": 4.1, + "learning_rate": 4.829654918599356e-06, + "loss": 0.9476, + "step": 55440 + }, + { + "epoch": 4.1, + "learning_rate": 4.82931350274416e-06, + "loss": 1.1027, + "step": 55441 + }, + { + "epoch": 4.1, + "learning_rate": 4.828972095115448e-06, + "loss": 1.0409, + "step": 55442 + }, + { + "epoch": 4.1, + "learning_rate": 4.828630695713772e-06, + "loss": 0.9854, + "step": 55443 + }, + { + "epoch": 4.1, + "learning_rate": 4.82828930453967e-06, + "loss": 1.0373, + "step": 55444 + }, + { + "epoch": 4.1, + "learning_rate": 4.827947921593687e-06, + "loss": 1.0704, + "step": 55445 + }, + { + "epoch": 4.1, + "learning_rate": 4.827606546876363e-06, + "loss": 0.9066, + "step": 55446 + }, + { + "epoch": 4.1, + "learning_rate": 4.82726518038824e-06, + "loss": 1.0048, + "step": 55447 + }, + { + "epoch": 4.1, + "learning_rate": 4.8269238221298685e-06, + "loss": 1.0952, + "step": 55448 + }, + { + "epoch": 4.1, + "learning_rate": 4.826582472101786e-06, + "loss": 0.975, + "step": 55449 + }, + { + "epoch": 4.1, + "learning_rate": 4.8262411303045375e-06, + "loss": 0.9247, + "step": 55450 + }, + { + "epoch": 4.1, + "learning_rate": 4.825899796738662e-06, + "loss": 0.9008, + "step": 55451 + }, + { + "epoch": 4.1, + "learning_rate": 4.825558471404709e-06, + "loss": 1.0026, + "step": 55452 + }, + { + "epoch": 4.1, + "learning_rate": 4.825217154303219e-06, + "loss": 1.1074, + "step": 55453 + }, + { + "epoch": 4.1, + "learning_rate": 4.824875845434733e-06, + "loss": 0.9041, + "step": 55454 + }, + { + "epoch": 4.1, + "learning_rate": 4.824534544799797e-06, + "loss": 0.9957, + "step": 55455 + }, + { + "epoch": 4.1, + "learning_rate": 4.824193252398949e-06, + "loss": 1.1045, + "step": 55456 + }, + { + "epoch": 4.1, + "learning_rate": 4.8238519682327426e-06, + "loss": 1.0127, + "step": 55457 + }, + { + "epoch": 4.1, + "learning_rate": 4.823510692301705e-06, + "loss": 0.9923, + "step": 55458 + }, + { + "epoch": 4.1, + "learning_rate": 4.8231694246063925e-06, + "loss": 1.0374, + "step": 55459 + }, + { + "epoch": 4.1, + "learning_rate": 4.82282816514734e-06, + "loss": 0.9262, + "step": 55460 + }, + { + "epoch": 4.1, + "learning_rate": 4.822486913925101e-06, + "loss": 0.9777, + "step": 55461 + }, + { + "epoch": 4.1, + "learning_rate": 4.822145670940203e-06, + "loss": 0.9764, + "step": 55462 + }, + { + "epoch": 4.1, + "learning_rate": 4.821804436193201e-06, + "loss": 1.1086, + "step": 55463 + }, + { + "epoch": 4.1, + "learning_rate": 4.8214632096846315e-06, + "loss": 0.9592, + "step": 55464 + }, + { + "epoch": 4.1, + "learning_rate": 4.821121991415046e-06, + "loss": 1.0582, + "step": 55465 + }, + { + "epoch": 4.1, + "learning_rate": 4.820780781384979e-06, + "loss": 0.9083, + "step": 55466 + }, + { + "epoch": 4.1, + "learning_rate": 4.820439579594971e-06, + "loss": 1.0002, + "step": 55467 + }, + { + "epoch": 4.1, + "learning_rate": 4.820098386045574e-06, + "loss": 0.9544, + "step": 55468 + }, + { + "epoch": 4.1, + "learning_rate": 4.819757200737325e-06, + "loss": 0.9824, + "step": 55469 + }, + { + "epoch": 4.1, + "learning_rate": 4.81941602367077e-06, + "loss": 0.9962, + "step": 55470 + }, + { + "epoch": 4.1, + "learning_rate": 4.819074854846445e-06, + "loss": 1.0627, + "step": 55471 + }, + { + "epoch": 4.1, + "learning_rate": 4.8187336942649025e-06, + "loss": 0.9325, + "step": 55472 + }, + { + "epoch": 4.1, + "learning_rate": 4.818392541926679e-06, + "loss": 1.0447, + "step": 55473 + }, + { + "epoch": 4.1, + "learning_rate": 4.818051397832319e-06, + "loss": 0.9969, + "step": 55474 + }, + { + "epoch": 4.1, + "learning_rate": 4.817710261982367e-06, + "loss": 0.9333, + "step": 55475 + }, + { + "epoch": 4.1, + "learning_rate": 4.817369134377359e-06, + "loss": 0.8964, + "step": 55476 + }, + { + "epoch": 4.1, + "learning_rate": 4.81702801501785e-06, + "loss": 1.0147, + "step": 55477 + }, + { + "epoch": 4.1, + "learning_rate": 4.816686903904367e-06, + "loss": 0.9822, + "step": 55478 + }, + { + "epoch": 4.1, + "learning_rate": 4.8163458010374655e-06, + "loss": 0.9393, + "step": 55479 + }, + { + "epoch": 4.1, + "learning_rate": 4.81600470641768e-06, + "loss": 0.9376, + "step": 55480 + }, + { + "epoch": 4.1, + "learning_rate": 4.8156636200455645e-06, + "loss": 0.9351, + "step": 55481 + }, + { + "epoch": 4.1, + "learning_rate": 4.815322541921645e-06, + "loss": 0.9695, + "step": 55482 + }, + { + "epoch": 4.1, + "learning_rate": 4.814981472046479e-06, + "loss": 1.0493, + "step": 55483 + }, + { + "epoch": 4.1, + "learning_rate": 4.814640410420602e-06, + "loss": 0.9366, + "step": 55484 + }, + { + "epoch": 4.1, + "learning_rate": 4.814299357044558e-06, + "loss": 0.9169, + "step": 55485 + }, + { + "epoch": 4.1, + "learning_rate": 4.81395831191889e-06, + "loss": 1.0763, + "step": 55486 + }, + { + "epoch": 4.1, + "learning_rate": 4.813617275044136e-06, + "loss": 0.965, + "step": 55487 + }, + { + "epoch": 4.1, + "learning_rate": 4.813276246420847e-06, + "loss": 0.9428, + "step": 55488 + }, + { + "epoch": 4.1, + "learning_rate": 4.812935226049561e-06, + "loss": 1.0076, + "step": 55489 + }, + { + "epoch": 4.1, + "learning_rate": 4.81259421393082e-06, + "loss": 0.9962, + "step": 55490 + }, + { + "epoch": 4.1, + "learning_rate": 4.812253210065166e-06, + "loss": 0.9778, + "step": 55491 + }, + { + "epoch": 4.1, + "learning_rate": 4.811912214453144e-06, + "loss": 0.9083, + "step": 55492 + }, + { + "epoch": 4.1, + "learning_rate": 4.811571227095296e-06, + "loss": 0.9804, + "step": 55493 + }, + { + "epoch": 4.1, + "learning_rate": 4.811230247992164e-06, + "loss": 1.0474, + "step": 55494 + }, + { + "epoch": 4.1, + "learning_rate": 4.810889277144292e-06, + "loss": 1.1274, + "step": 55495 + }, + { + "epoch": 4.1, + "learning_rate": 4.810548314552216e-06, + "loss": 0.9647, + "step": 55496 + }, + { + "epoch": 4.1, + "learning_rate": 4.810207360216487e-06, + "loss": 0.9829, + "step": 55497 + }, + { + "epoch": 4.1, + "learning_rate": 4.8098664141376435e-06, + "loss": 1.1062, + "step": 55498 + }, + { + "epoch": 4.1, + "learning_rate": 4.8095254763162295e-06, + "loss": 0.9725, + "step": 55499 + }, + { + "epoch": 4.1, + "learning_rate": 4.809184546752782e-06, + "loss": 1.1226, + "step": 55500 + }, + { + "epoch": 4.1, + "learning_rate": 4.808843625447851e-06, + "loss": 1.1015, + "step": 55501 + }, + { + "epoch": 4.1, + "learning_rate": 4.808502712401979e-06, + "loss": 0.9661, + "step": 55502 + }, + { + "epoch": 4.1, + "learning_rate": 4.808161807615698e-06, + "loss": 1.0071, + "step": 55503 + }, + { + "epoch": 4.1, + "learning_rate": 4.807820911089561e-06, + "loss": 0.9667, + "step": 55504 + }, + { + "epoch": 4.1, + "learning_rate": 4.807480022824103e-06, + "loss": 1.0081, + "step": 55505 + }, + { + "epoch": 4.1, + "learning_rate": 4.807139142819876e-06, + "loss": 0.9925, + "step": 55506 + }, + { + "epoch": 4.1, + "learning_rate": 4.806798271077411e-06, + "loss": 0.9029, + "step": 55507 + }, + { + "epoch": 4.1, + "learning_rate": 4.806457407597258e-06, + "loss": 0.9312, + "step": 55508 + }, + { + "epoch": 4.1, + "learning_rate": 4.806116552379958e-06, + "loss": 1.052, + "step": 55509 + }, + { + "epoch": 4.1, + "learning_rate": 4.805775705426051e-06, + "loss": 1.055, + "step": 55510 + }, + { + "epoch": 4.1, + "learning_rate": 4.805434866736078e-06, + "loss": 1.0844, + "step": 55511 + }, + { + "epoch": 4.1, + "learning_rate": 4.8050940363105866e-06, + "loss": 0.9785, + "step": 55512 + }, + { + "epoch": 4.1, + "learning_rate": 4.804753214150117e-06, + "loss": 0.9609, + "step": 55513 + }, + { + "epoch": 4.1, + "learning_rate": 4.80441240025521e-06, + "loss": 1.0384, + "step": 55514 + }, + { + "epoch": 4.1, + "learning_rate": 4.804071594626409e-06, + "loss": 1.0483, + "step": 55515 + }, + { + "epoch": 4.1, + "learning_rate": 4.803730797264252e-06, + "loss": 1.1082, + "step": 55516 + }, + { + "epoch": 4.1, + "learning_rate": 4.803390008169289e-06, + "loss": 0.9508, + "step": 55517 + }, + { + "epoch": 4.1, + "learning_rate": 4.803049227342057e-06, + "loss": 1.0769, + "step": 55518 + }, + { + "epoch": 4.1, + "learning_rate": 4.802708454783101e-06, + "loss": 1.0231, + "step": 55519 + }, + { + "epoch": 4.1, + "learning_rate": 4.8023676904929565e-06, + "loss": 0.9441, + "step": 55520 + }, + { + "epoch": 4.1, + "learning_rate": 4.802026934472175e-06, + "loss": 1.0764, + "step": 55521 + }, + { + "epoch": 4.1, + "learning_rate": 4.8016861867212975e-06, + "loss": 0.909, + "step": 55522 + }, + { + "epoch": 4.1, + "learning_rate": 4.801345447240855e-06, + "loss": 0.9703, + "step": 55523 + }, + { + "epoch": 4.1, + "learning_rate": 4.801004716031402e-06, + "loss": 1.0215, + "step": 55524 + }, + { + "epoch": 4.1, + "learning_rate": 4.800663993093473e-06, + "loss": 1.0481, + "step": 55525 + }, + { + "epoch": 4.1, + "learning_rate": 4.80032327842762e-06, + "loss": 1.0688, + "step": 55526 + }, + { + "epoch": 4.1, + "learning_rate": 4.7999825720343706e-06, + "loss": 0.8991, + "step": 55527 + }, + { + "epoch": 4.1, + "learning_rate": 4.799641873914278e-06, + "loss": 0.9888, + "step": 55528 + }, + { + "epoch": 4.1, + "learning_rate": 4.799301184067877e-06, + "loss": 0.9812, + "step": 55529 + }, + { + "epoch": 4.1, + "learning_rate": 4.798960502495722e-06, + "loss": 1.1241, + "step": 55530 + }, + { + "epoch": 4.1, + "learning_rate": 4.798619829198342e-06, + "loss": 1.0477, + "step": 55531 + }, + { + "epoch": 4.1, + "learning_rate": 4.798279164176279e-06, + "loss": 1.0742, + "step": 55532 + }, + { + "epoch": 4.1, + "learning_rate": 4.797938507430085e-06, + "loss": 1.0178, + "step": 55533 + }, + { + "epoch": 4.1, + "learning_rate": 4.797597858960295e-06, + "loss": 0.9745, + "step": 55534 + }, + { + "epoch": 4.1, + "learning_rate": 4.797257218767453e-06, + "loss": 0.907, + "step": 55535 + }, + { + "epoch": 4.1, + "learning_rate": 4.796916586852095e-06, + "loss": 1.0533, + "step": 55536 + }, + { + "epoch": 4.1, + "learning_rate": 4.796575963214774e-06, + "loss": 1.0107, + "step": 55537 + }, + { + "epoch": 4.1, + "learning_rate": 4.796235347856025e-06, + "loss": 1.0895, + "step": 55538 + }, + { + "epoch": 4.1, + "learning_rate": 4.795894740776391e-06, + "loss": 0.979, + "step": 55539 + }, + { + "epoch": 4.1, + "learning_rate": 4.795554141976415e-06, + "loss": 1.0701, + "step": 55540 + }, + { + "epoch": 4.1, + "learning_rate": 4.795213551456634e-06, + "loss": 0.992, + "step": 55541 + }, + { + "epoch": 4.1, + "learning_rate": 4.794872969217597e-06, + "loss": 1.0337, + "step": 55542 + }, + { + "epoch": 4.1, + "learning_rate": 4.794532395259843e-06, + "loss": 0.9788, + "step": 55543 + }, + { + "epoch": 4.1, + "learning_rate": 4.794191829583913e-06, + "loss": 0.9693, + "step": 55544 + }, + { + "epoch": 4.1, + "learning_rate": 4.793851272190346e-06, + "loss": 0.8885, + "step": 55545 + }, + { + "epoch": 4.1, + "learning_rate": 4.7935107230796905e-06, + "loss": 0.9749, + "step": 55546 + }, + { + "epoch": 4.1, + "learning_rate": 4.793170182252486e-06, + "loss": 0.9384, + "step": 55547 + }, + { + "epoch": 4.1, + "learning_rate": 4.792829649709272e-06, + "loss": 0.9034, + "step": 55548 + }, + { + "epoch": 4.1, + "learning_rate": 4.792489125450592e-06, + "loss": 0.9386, + "step": 55549 + }, + { + "epoch": 4.1, + "learning_rate": 4.792148609476983e-06, + "loss": 1.0378, + "step": 55550 + }, + { + "epoch": 4.1, + "learning_rate": 4.791808101788999e-06, + "loss": 0.9328, + "step": 55551 + }, + { + "epoch": 4.1, + "learning_rate": 4.791467602387167e-06, + "loss": 0.9701, + "step": 55552 + }, + { + "epoch": 4.1, + "learning_rate": 4.791127111272038e-06, + "loss": 1.0478, + "step": 55553 + }, + { + "epoch": 4.1, + "learning_rate": 4.790786628444153e-06, + "loss": 1.037, + "step": 55554 + }, + { + "epoch": 4.1, + "learning_rate": 4.79044615390405e-06, + "loss": 0.955, + "step": 55555 + }, + { + "epoch": 4.1, + "learning_rate": 4.79010568765227e-06, + "loss": 1.0815, + "step": 55556 + }, + { + "epoch": 4.11, + "learning_rate": 4.7897652296893606e-06, + "loss": 0.9414, + "step": 55557 + }, + { + "epoch": 4.11, + "learning_rate": 4.7894247800158615e-06, + "loss": 0.9205, + "step": 55558 + }, + { + "epoch": 4.11, + "learning_rate": 4.78908433863231e-06, + "loss": 1.0515, + "step": 55559 + }, + { + "epoch": 4.11, + "learning_rate": 4.788743905539254e-06, + "loss": 0.9864, + "step": 55560 + }, + { + "epoch": 4.11, + "learning_rate": 4.788403480737226e-06, + "loss": 0.9518, + "step": 55561 + }, + { + "epoch": 4.11, + "learning_rate": 4.788063064226778e-06, + "loss": 0.9873, + "step": 55562 + }, + { + "epoch": 4.11, + "learning_rate": 4.787722656008447e-06, + "loss": 1.0074, + "step": 55563 + }, + { + "epoch": 4.11, + "learning_rate": 4.7873822560827756e-06, + "loss": 1.0508, + "step": 55564 + }, + { + "epoch": 4.11, + "learning_rate": 4.787041864450299e-06, + "loss": 0.9969, + "step": 55565 + }, + { + "epoch": 4.11, + "learning_rate": 4.7867014811115695e-06, + "loss": 0.9313, + "step": 55566 + }, + { + "epoch": 4.11, + "learning_rate": 4.786361106067122e-06, + "loss": 1.0282, + "step": 55567 + }, + { + "epoch": 4.11, + "learning_rate": 4.7860207393175005e-06, + "loss": 0.9724, + "step": 55568 + }, + { + "epoch": 4.11, + "learning_rate": 4.7856803808632445e-06, + "loss": 1.1855, + "step": 55569 + }, + { + "epoch": 4.11, + "learning_rate": 4.7853400307048935e-06, + "loss": 1.0195, + "step": 55570 + }, + { + "epoch": 4.11, + "learning_rate": 4.784999688843e-06, + "loss": 0.9253, + "step": 55571 + }, + { + "epoch": 4.11, + "learning_rate": 4.784659355278089e-06, + "loss": 1.0571, + "step": 55572 + }, + { + "epoch": 4.11, + "learning_rate": 4.784319030010713e-06, + "loss": 0.9691, + "step": 55573 + }, + { + "epoch": 4.11, + "learning_rate": 4.7839787130414085e-06, + "loss": 0.9874, + "step": 55574 + }, + { + "epoch": 4.11, + "learning_rate": 4.783638404370726e-06, + "loss": 1.0147, + "step": 55575 + }, + { + "epoch": 4.11, + "learning_rate": 4.783298103999193e-06, + "loss": 0.9163, + "step": 55576 + }, + { + "epoch": 4.11, + "learning_rate": 4.782957811927362e-06, + "loss": 1.1268, + "step": 55577 + }, + { + "epoch": 4.11, + "learning_rate": 4.782617528155771e-06, + "loss": 1.0973, + "step": 55578 + }, + { + "epoch": 4.11, + "learning_rate": 4.782277252684959e-06, + "loss": 0.9546, + "step": 55579 + }, + { + "epoch": 4.11, + "learning_rate": 4.781936985515471e-06, + "loss": 0.8659, + "step": 55580 + }, + { + "epoch": 4.11, + "learning_rate": 4.7815967266478424e-06, + "loss": 1.0798, + "step": 55581 + }, + { + "epoch": 4.11, + "learning_rate": 4.781256476082622e-06, + "loss": 0.9822, + "step": 55582 + }, + { + "epoch": 4.11, + "learning_rate": 4.780916233820348e-06, + "loss": 0.9611, + "step": 55583 + }, + { + "epoch": 4.11, + "learning_rate": 4.7805759998615616e-06, + "loss": 1.0787, + "step": 55584 + }, + { + "epoch": 4.11, + "learning_rate": 4.7802357742068005e-06, + "loss": 0.9634, + "step": 55585 + }, + { + "epoch": 4.11, + "learning_rate": 4.779895556856614e-06, + "loss": 0.9945, + "step": 55586 + }, + { + "epoch": 4.11, + "learning_rate": 4.779555347811539e-06, + "loss": 1.0248, + "step": 55587 + }, + { + "epoch": 4.11, + "learning_rate": 4.779215147072115e-06, + "loss": 1.008, + "step": 55588 + }, + { + "epoch": 4.11, + "learning_rate": 4.778874954638887e-06, + "loss": 0.9115, + "step": 55589 + }, + { + "epoch": 4.11, + "learning_rate": 4.778534770512389e-06, + "loss": 0.9443, + "step": 55590 + }, + { + "epoch": 4.11, + "learning_rate": 4.778194594693172e-06, + "loss": 1.0292, + "step": 55591 + }, + { + "epoch": 4.11, + "learning_rate": 4.777854427181773e-06, + "loss": 1.0467, + "step": 55592 + }, + { + "epoch": 4.11, + "learning_rate": 4.777514267978732e-06, + "loss": 0.9792, + "step": 55593 + }, + { + "epoch": 4.11, + "learning_rate": 4.777174117084588e-06, + "loss": 1.0617, + "step": 55594 + }, + { + "epoch": 4.11, + "learning_rate": 4.776833974499888e-06, + "loss": 1.1317, + "step": 55595 + }, + { + "epoch": 4.11, + "learning_rate": 4.776493840225176e-06, + "loss": 1.0419, + "step": 55596 + }, + { + "epoch": 4.11, + "learning_rate": 4.776153714260978e-06, + "loss": 0.9312, + "step": 55597 + }, + { + "epoch": 4.11, + "learning_rate": 4.77581359660785e-06, + "loss": 0.9041, + "step": 55598 + }, + { + "epoch": 4.11, + "learning_rate": 4.775473487266327e-06, + "loss": 0.9986, + "step": 55599 + }, + { + "epoch": 4.11, + "learning_rate": 4.775133386236951e-06, + "loss": 1.0306, + "step": 55600 + }, + { + "epoch": 4.11, + "learning_rate": 4.77479329352026e-06, + "loss": 1.0114, + "step": 55601 + }, + { + "epoch": 4.11, + "learning_rate": 4.774453209116801e-06, + "loss": 1.0693, + "step": 55602 + }, + { + "epoch": 4.11, + "learning_rate": 4.7741131330271125e-06, + "loss": 0.9149, + "step": 55603 + }, + { + "epoch": 4.11, + "learning_rate": 4.773773065251736e-06, + "loss": 0.8927, + "step": 55604 + }, + { + "epoch": 4.11, + "learning_rate": 4.77343300579121e-06, + "loss": 0.9835, + "step": 55605 + }, + { + "epoch": 4.11, + "learning_rate": 4.773092954646075e-06, + "loss": 1.0697, + "step": 55606 + }, + { + "epoch": 4.11, + "learning_rate": 4.772752911816878e-06, + "loss": 1.0291, + "step": 55607 + }, + { + "epoch": 4.11, + "learning_rate": 4.772412877304155e-06, + "loss": 0.9895, + "step": 55608 + }, + { + "epoch": 4.11, + "learning_rate": 4.77207285110845e-06, + "loss": 0.981, + "step": 55609 + }, + { + "epoch": 4.11, + "learning_rate": 4.7717328332302985e-06, + "loss": 0.9572, + "step": 55610 + }, + { + "epoch": 4.11, + "learning_rate": 4.7713928236702475e-06, + "loss": 0.9573, + "step": 55611 + }, + { + "epoch": 4.11, + "learning_rate": 4.771052822428837e-06, + "loss": 0.9785, + "step": 55612 + }, + { + "epoch": 4.11, + "learning_rate": 4.770712829506606e-06, + "loss": 1.0401, + "step": 55613 + }, + { + "epoch": 4.11, + "learning_rate": 4.770372844904096e-06, + "loss": 0.9745, + "step": 55614 + }, + { + "epoch": 4.11, + "learning_rate": 4.7700328686218444e-06, + "loss": 1.0972, + "step": 55615 + }, + { + "epoch": 4.11, + "learning_rate": 4.769692900660404e-06, + "loss": 0.9654, + "step": 55616 + }, + { + "epoch": 4.11, + "learning_rate": 4.769352941020299e-06, + "loss": 0.9954, + "step": 55617 + }, + { + "epoch": 4.11, + "learning_rate": 4.769012989702082e-06, + "loss": 1.0472, + "step": 55618 + }, + { + "epoch": 4.11, + "learning_rate": 4.768673046706288e-06, + "loss": 1.0304, + "step": 55619 + }, + { + "epoch": 4.11, + "learning_rate": 4.768333112033467e-06, + "loss": 0.9685, + "step": 55620 + }, + { + "epoch": 4.11, + "learning_rate": 4.767993185684145e-06, + "loss": 1.0054, + "step": 55621 + }, + { + "epoch": 4.11, + "learning_rate": 4.767653267658876e-06, + "loss": 1.0827, + "step": 55622 + }, + { + "epoch": 4.11, + "learning_rate": 4.767313357958192e-06, + "loss": 0.9236, + "step": 55623 + }, + { + "epoch": 4.11, + "learning_rate": 4.766973456582644e-06, + "loss": 1.0595, + "step": 55624 + }, + { + "epoch": 4.11, + "learning_rate": 4.7666335635327635e-06, + "loss": 1.0254, + "step": 55625 + }, + { + "epoch": 4.11, + "learning_rate": 4.766293678809091e-06, + "loss": 1.1048, + "step": 55626 + }, + { + "epoch": 4.11, + "learning_rate": 4.765953802412172e-06, + "loss": 0.9454, + "step": 55627 + }, + { + "epoch": 4.11, + "learning_rate": 4.765613934342547e-06, + "loss": 1.071, + "step": 55628 + }, + { + "epoch": 4.11, + "learning_rate": 4.765274074600755e-06, + "loss": 0.8295, + "step": 55629 + }, + { + "epoch": 4.11, + "learning_rate": 4.764934223187334e-06, + "loss": 0.9871, + "step": 55630 + }, + { + "epoch": 4.11, + "learning_rate": 4.76459438010283e-06, + "loss": 0.9013, + "step": 55631 + }, + { + "epoch": 4.11, + "learning_rate": 4.764254545347784e-06, + "loss": 1.0384, + "step": 55632 + }, + { + "epoch": 4.11, + "learning_rate": 4.763914718922732e-06, + "loss": 1.0396, + "step": 55633 + }, + { + "epoch": 4.11, + "learning_rate": 4.763574900828217e-06, + "loss": 0.9477, + "step": 55634 + }, + { + "epoch": 4.11, + "learning_rate": 4.763235091064776e-06, + "loss": 1.0613, + "step": 55635 + }, + { + "epoch": 4.11, + "learning_rate": 4.762895289632957e-06, + "loss": 0.8477, + "step": 55636 + }, + { + "epoch": 4.11, + "learning_rate": 4.7625554965332965e-06, + "loss": 0.9869, + "step": 55637 + }, + { + "epoch": 4.11, + "learning_rate": 4.762215711766335e-06, + "loss": 0.9973, + "step": 55638 + }, + { + "epoch": 4.11, + "learning_rate": 4.761875935332609e-06, + "loss": 0.9884, + "step": 55639 + }, + { + "epoch": 4.11, + "learning_rate": 4.761536167232669e-06, + "loss": 1.0058, + "step": 55640 + }, + { + "epoch": 4.11, + "learning_rate": 4.761196407467049e-06, + "loss": 0.9978, + "step": 55641 + }, + { + "epoch": 4.11, + "learning_rate": 4.76085665603629e-06, + "loss": 0.8602, + "step": 55642 + }, + { + "epoch": 4.11, + "learning_rate": 4.760516912940935e-06, + "loss": 0.9906, + "step": 55643 + }, + { + "epoch": 4.11, + "learning_rate": 4.760177178181521e-06, + "loss": 1.0003, + "step": 55644 + }, + { + "epoch": 4.11, + "learning_rate": 4.75983745175859e-06, + "loss": 0.9007, + "step": 55645 + }, + { + "epoch": 4.11, + "learning_rate": 4.759497733672679e-06, + "loss": 1.0519, + "step": 55646 + }, + { + "epoch": 4.11, + "learning_rate": 4.7591580239243365e-06, + "loss": 1.031, + "step": 55647 + }, + { + "epoch": 4.11, + "learning_rate": 4.7588183225140984e-06, + "loss": 0.9779, + "step": 55648 + }, + { + "epoch": 4.11, + "learning_rate": 4.758478629442504e-06, + "loss": 0.9852, + "step": 55649 + }, + { + "epoch": 4.11, + "learning_rate": 4.758138944710093e-06, + "loss": 0.9675, + "step": 55650 + }, + { + "epoch": 4.11, + "learning_rate": 4.757799268317412e-06, + "loss": 0.934, + "step": 55651 + }, + { + "epoch": 4.11, + "learning_rate": 4.7574596002649966e-06, + "loss": 1.068, + "step": 55652 + }, + { + "epoch": 4.11, + "learning_rate": 4.757119940553388e-06, + "loss": 0.9698, + "step": 55653 + }, + { + "epoch": 4.11, + "learning_rate": 4.7567802891831276e-06, + "loss": 0.9532, + "step": 55654 + }, + { + "epoch": 4.11, + "learning_rate": 4.756440646154749e-06, + "loss": 0.9166, + "step": 55655 + }, + { + "epoch": 4.11, + "learning_rate": 4.7561010114688036e-06, + "loss": 1.0007, + "step": 55656 + }, + { + "epoch": 4.11, + "learning_rate": 4.755761385125825e-06, + "loss": 0.9915, + "step": 55657 + }, + { + "epoch": 4.11, + "learning_rate": 4.755421767126356e-06, + "loss": 0.9847, + "step": 55658 + }, + { + "epoch": 4.11, + "learning_rate": 4.7550821574709325e-06, + "loss": 0.9458, + "step": 55659 + }, + { + "epoch": 4.11, + "learning_rate": 4.754742556160101e-06, + "loss": 0.9772, + "step": 55660 + }, + { + "epoch": 4.11, + "learning_rate": 4.754402963194403e-06, + "loss": 1.141, + "step": 55661 + }, + { + "epoch": 4.11, + "learning_rate": 4.754063378574369e-06, + "loss": 0.9675, + "step": 55662 + }, + { + "epoch": 4.11, + "learning_rate": 4.753723802300547e-06, + "loss": 0.9444, + "step": 55663 + }, + { + "epoch": 4.11, + "learning_rate": 4.753384234373472e-06, + "loss": 0.9167, + "step": 55664 + }, + { + "epoch": 4.11, + "learning_rate": 4.753044674793695e-06, + "loss": 0.9277, + "step": 55665 + }, + { + "epoch": 4.11, + "learning_rate": 4.752705123561742e-06, + "loss": 1.0187, + "step": 55666 + }, + { + "epoch": 4.11, + "learning_rate": 4.752365580678163e-06, + "loss": 0.9949, + "step": 55667 + }, + { + "epoch": 4.11, + "learning_rate": 4.752026046143493e-06, + "loss": 0.9875, + "step": 55668 + }, + { + "epoch": 4.11, + "learning_rate": 4.751686519958281e-06, + "loss": 0.995, + "step": 55669 + }, + { + "epoch": 4.11, + "learning_rate": 4.751347002123057e-06, + "loss": 1.0418, + "step": 55670 + }, + { + "epoch": 4.11, + "learning_rate": 4.751007492638361e-06, + "loss": 1.0543, + "step": 55671 + }, + { + "epoch": 4.11, + "learning_rate": 4.750667991504741e-06, + "loss": 1.1126, + "step": 55672 + }, + { + "epoch": 4.11, + "learning_rate": 4.750328498722732e-06, + "loss": 0.9114, + "step": 55673 + }, + { + "epoch": 4.11, + "learning_rate": 4.7499890142928765e-06, + "loss": 0.9708, + "step": 55674 + }, + { + "epoch": 4.11, + "learning_rate": 4.749649538215709e-06, + "loss": 0.9386, + "step": 55675 + }, + { + "epoch": 4.11, + "learning_rate": 4.749310070491778e-06, + "loss": 1.0233, + "step": 55676 + }, + { + "epoch": 4.11, + "learning_rate": 4.7489706111216195e-06, + "loss": 0.9641, + "step": 55677 + }, + { + "epoch": 4.11, + "learning_rate": 4.748631160105773e-06, + "loss": 0.8771, + "step": 55678 + }, + { + "epoch": 4.11, + "learning_rate": 4.748291717444776e-06, + "loss": 1.0231, + "step": 55679 + }, + { + "epoch": 4.11, + "learning_rate": 4.747952283139176e-06, + "loss": 1.0209, + "step": 55680 + }, + { + "epoch": 4.11, + "learning_rate": 4.747612857189507e-06, + "loss": 0.9081, + "step": 55681 + }, + { + "epoch": 4.11, + "learning_rate": 4.747273439596313e-06, + "loss": 1.1432, + "step": 55682 + }, + { + "epoch": 4.11, + "learning_rate": 4.746934030360131e-06, + "loss": 0.9904, + "step": 55683 + }, + { + "epoch": 4.11, + "learning_rate": 4.746594629481498e-06, + "loss": 0.9808, + "step": 55684 + }, + { + "epoch": 4.11, + "learning_rate": 4.746255236960962e-06, + "loss": 1.0392, + "step": 55685 + }, + { + "epoch": 4.11, + "learning_rate": 4.745915852799058e-06, + "loss": 1.0477, + "step": 55686 + }, + { + "epoch": 4.11, + "learning_rate": 4.7455764769963265e-06, + "loss": 0.9878, + "step": 55687 + }, + { + "epoch": 4.11, + "learning_rate": 4.7452371095533045e-06, + "loss": 0.961, + "step": 55688 + }, + { + "epoch": 4.11, + "learning_rate": 4.744897750470542e-06, + "loss": 1.0321, + "step": 55689 + }, + { + "epoch": 4.11, + "learning_rate": 4.744558399748569e-06, + "loss": 1.0135, + "step": 55690 + }, + { + "epoch": 4.11, + "learning_rate": 4.744219057387924e-06, + "loss": 1.0829, + "step": 55691 + }, + { + "epoch": 4.12, + "learning_rate": 4.7438797233891556e-06, + "loss": 1.0345, + "step": 55692 + }, + { + "epoch": 4.12, + "learning_rate": 4.743540397752798e-06, + "loss": 1.0119, + "step": 55693 + }, + { + "epoch": 4.12, + "learning_rate": 4.743201080479394e-06, + "loss": 0.9398, + "step": 55694 + }, + { + "epoch": 4.12, + "learning_rate": 4.742861771569478e-06, + "loss": 0.951, + "step": 55695 + }, + { + "epoch": 4.12, + "learning_rate": 4.742522471023597e-06, + "loss": 1.0289, + "step": 55696 + }, + { + "epoch": 4.12, + "learning_rate": 4.742183178842288e-06, + "loss": 0.9565, + "step": 55697 + }, + { + "epoch": 4.12, + "learning_rate": 4.741843895026089e-06, + "loss": 1.0226, + "step": 55698 + }, + { + "epoch": 4.12, + "learning_rate": 4.741504619575542e-06, + "loss": 0.977, + "step": 55699 + }, + { + "epoch": 4.12, + "learning_rate": 4.741165352491183e-06, + "loss": 1.0561, + "step": 55700 + }, + { + "epoch": 4.12, + "learning_rate": 4.740826093773557e-06, + "loss": 0.9354, + "step": 55701 + }, + { + "epoch": 4.12, + "learning_rate": 4.740486843423202e-06, + "loss": 1.0149, + "step": 55702 + }, + { + "epoch": 4.12, + "learning_rate": 4.740147601440658e-06, + "loss": 0.9681, + "step": 55703 + }, + { + "epoch": 4.12, + "learning_rate": 4.739808367826458e-06, + "loss": 1.0132, + "step": 55704 + }, + { + "epoch": 4.12, + "learning_rate": 4.739469142581154e-06, + "loss": 1.0489, + "step": 55705 + }, + { + "epoch": 4.12, + "learning_rate": 4.739129925705278e-06, + "loss": 0.9968, + "step": 55706 + }, + { + "epoch": 4.12, + "learning_rate": 4.738790717199371e-06, + "loss": 0.9482, + "step": 55707 + }, + { + "epoch": 4.12, + "learning_rate": 4.738451517063972e-06, + "loss": 0.9708, + "step": 55708 + }, + { + "epoch": 4.12, + "learning_rate": 4.7381123252996186e-06, + "loss": 1.0138, + "step": 55709 + }, + { + "epoch": 4.12, + "learning_rate": 4.737773141906861e-06, + "loss": 0.9597, + "step": 55710 + }, + { + "epoch": 4.12, + "learning_rate": 4.737433966886224e-06, + "loss": 0.9863, + "step": 55711 + }, + { + "epoch": 4.12, + "learning_rate": 4.7370948002382565e-06, + "loss": 1.0882, + "step": 55712 + }, + { + "epoch": 4.12, + "learning_rate": 4.736755641963492e-06, + "loss": 0.9754, + "step": 55713 + }, + { + "epoch": 4.12, + "learning_rate": 4.7364164920624824e-06, + "loss": 0.9723, + "step": 55714 + }, + { + "epoch": 4.12, + "learning_rate": 4.736077350535751e-06, + "loss": 0.942, + "step": 55715 + }, + { + "epoch": 4.12, + "learning_rate": 4.735738217383849e-06, + "loss": 1.0611, + "step": 55716 + }, + { + "epoch": 4.12, + "learning_rate": 4.7353990926073126e-06, + "loss": 0.9569, + "step": 55717 + }, + { + "epoch": 4.12, + "learning_rate": 4.7350599762066806e-06, + "loss": 0.9466, + "step": 55718 + }, + { + "epoch": 4.12, + "learning_rate": 4.734720868182493e-06, + "loss": 1.0493, + "step": 55719 + }, + { + "epoch": 4.12, + "learning_rate": 4.734381768535286e-06, + "loss": 1.066, + "step": 55720 + }, + { + "epoch": 4.12, + "learning_rate": 4.7340426772656046e-06, + "loss": 1.0065, + "step": 55721 + }, + { + "epoch": 4.12, + "learning_rate": 4.7337035943739864e-06, + "loss": 1.0248, + "step": 55722 + }, + { + "epoch": 4.12, + "learning_rate": 4.733364519860971e-06, + "loss": 0.9324, + "step": 55723 + }, + { + "epoch": 4.12, + "learning_rate": 4.733025453727092e-06, + "loss": 1.0675, + "step": 55724 + }, + { + "epoch": 4.12, + "learning_rate": 4.7326863959729e-06, + "loss": 1.1198, + "step": 55725 + }, + { + "epoch": 4.12, + "learning_rate": 4.732347346598928e-06, + "loss": 1.0222, + "step": 55726 + }, + { + "epoch": 4.12, + "learning_rate": 4.7320083056057155e-06, + "loss": 0.96, + "step": 55727 + }, + { + "epoch": 4.12, + "learning_rate": 4.731669272993803e-06, + "loss": 0.9514, + "step": 55728 + }, + { + "epoch": 4.12, + "learning_rate": 4.731330248763725e-06, + "loss": 1.0448, + "step": 55729 + }, + { + "epoch": 4.12, + "learning_rate": 4.730991232916033e-06, + "loss": 0.932, + "step": 55730 + }, + { + "epoch": 4.12, + "learning_rate": 4.730652225451251e-06, + "loss": 0.977, + "step": 55731 + }, + { + "epoch": 4.12, + "learning_rate": 4.7303132263699295e-06, + "loss": 1.0231, + "step": 55732 + }, + { + "epoch": 4.12, + "learning_rate": 4.7299742356725995e-06, + "loss": 1.0108, + "step": 55733 + }, + { + "epoch": 4.12, + "learning_rate": 4.729635253359815e-06, + "loss": 0.9905, + "step": 55734 + }, + { + "epoch": 4.12, + "learning_rate": 4.729296279432096e-06, + "loss": 1.0832, + "step": 55735 + }, + { + "epoch": 4.12, + "learning_rate": 4.728957313889996e-06, + "loss": 1.018, + "step": 55736 + }, + { + "epoch": 4.12, + "learning_rate": 4.728618356734048e-06, + "loss": 0.9004, + "step": 55737 + }, + { + "epoch": 4.12, + "learning_rate": 4.728279407964794e-06, + "loss": 1.0763, + "step": 55738 + }, + { + "epoch": 4.12, + "learning_rate": 4.727940467582773e-06, + "loss": 0.9094, + "step": 55739 + }, + { + "epoch": 4.12, + "learning_rate": 4.727601535588517e-06, + "loss": 1.0096, + "step": 55740 + }, + { + "epoch": 4.12, + "learning_rate": 4.727262611982577e-06, + "loss": 1.0117, + "step": 55741 + }, + { + "epoch": 4.12, + "learning_rate": 4.7269236967654865e-06, + "loss": 0.8944, + "step": 55742 + }, + { + "epoch": 4.12, + "learning_rate": 4.7265847899377845e-06, + "loss": 0.9374, + "step": 55743 + }, + { + "epoch": 4.12, + "learning_rate": 4.726245891500007e-06, + "loss": 1.0274, + "step": 55744 + }, + { + "epoch": 4.12, + "learning_rate": 4.725907001452702e-06, + "loss": 0.9042, + "step": 55745 + }, + { + "epoch": 4.12, + "learning_rate": 4.725568119796402e-06, + "loss": 1.1179, + "step": 55746 + }, + { + "epoch": 4.12, + "learning_rate": 4.725229246531648e-06, + "loss": 1.0247, + "step": 55747 + }, + { + "epoch": 4.12, + "learning_rate": 4.724890381658979e-06, + "loss": 0.9797, + "step": 55748 + }, + { + "epoch": 4.12, + "learning_rate": 4.72455152517893e-06, + "loss": 0.9507, + "step": 55749 + }, + { + "epoch": 4.12, + "learning_rate": 4.724212677092047e-06, + "loss": 1.0923, + "step": 55750 + }, + { + "epoch": 4.12, + "learning_rate": 4.723873837398868e-06, + "loss": 0.9769, + "step": 55751 + }, + { + "epoch": 4.12, + "learning_rate": 4.723535006099929e-06, + "loss": 1.0778, + "step": 55752 + }, + { + "epoch": 4.12, + "learning_rate": 4.7231961831957664e-06, + "loss": 1.1108, + "step": 55753 + }, + { + "epoch": 4.12, + "learning_rate": 4.722857368686927e-06, + "loss": 0.8631, + "step": 55754 + }, + { + "epoch": 4.12, + "learning_rate": 4.72251856257395e-06, + "loss": 0.9431, + "step": 55755 + }, + { + "epoch": 4.12, + "learning_rate": 4.722179764857363e-06, + "loss": 1.2015, + "step": 55756 + }, + { + "epoch": 4.12, + "learning_rate": 4.7218409755377155e-06, + "loss": 0.9623, + "step": 55757 + }, + { + "epoch": 4.12, + "learning_rate": 4.721502194615541e-06, + "loss": 1.1169, + "step": 55758 + }, + { + "epoch": 4.12, + "learning_rate": 4.721163422091388e-06, + "loss": 0.9067, + "step": 55759 + }, + { + "epoch": 4.12, + "learning_rate": 4.720824657965782e-06, + "loss": 0.9441, + "step": 55760 + }, + { + "epoch": 4.12, + "learning_rate": 4.7204859022392706e-06, + "loss": 0.9065, + "step": 55761 + }, + { + "epoch": 4.12, + "learning_rate": 4.720147154912387e-06, + "loss": 0.9906, + "step": 55762 + }, + { + "epoch": 4.12, + "learning_rate": 4.719808415985682e-06, + "loss": 0.8568, + "step": 55763 + }, + { + "epoch": 4.12, + "learning_rate": 4.719469685459681e-06, + "loss": 1.0574, + "step": 55764 + }, + { + "epoch": 4.12, + "learning_rate": 4.719130963334926e-06, + "loss": 1.0628, + "step": 55765 + }, + { + "epoch": 4.12, + "learning_rate": 4.718792249611961e-06, + "loss": 1.0266, + "step": 55766 + }, + { + "epoch": 4.12, + "learning_rate": 4.7184535442913224e-06, + "loss": 0.9182, + "step": 55767 + }, + { + "epoch": 4.12, + "learning_rate": 4.718114847373549e-06, + "loss": 1.081, + "step": 55768 + }, + { + "epoch": 4.12, + "learning_rate": 4.717776158859175e-06, + "loss": 1.0432, + "step": 55769 + }, + { + "epoch": 4.12, + "learning_rate": 4.717437478748747e-06, + "loss": 0.906, + "step": 55770 + }, + { + "epoch": 4.12, + "learning_rate": 4.717098807042801e-06, + "loss": 0.9137, + "step": 55771 + }, + { + "epoch": 4.12, + "learning_rate": 4.716760143741875e-06, + "loss": 0.9127, + "step": 55772 + }, + { + "epoch": 4.12, + "learning_rate": 4.716421488846509e-06, + "loss": 1.0297, + "step": 55773 + }, + { + "epoch": 4.12, + "learning_rate": 4.716082842357236e-06, + "loss": 1.0336, + "step": 55774 + }, + { + "epoch": 4.12, + "learning_rate": 4.715744204274606e-06, + "loss": 0.8966, + "step": 55775 + }, + { + "epoch": 4.12, + "learning_rate": 4.715405574599145e-06, + "loss": 1.033, + "step": 55776 + }, + { + "epoch": 4.12, + "learning_rate": 4.715066953331402e-06, + "loss": 0.9403, + "step": 55777 + }, + { + "epoch": 4.12, + "learning_rate": 4.714728340471909e-06, + "loss": 1.0254, + "step": 55778 + }, + { + "epoch": 4.12, + "learning_rate": 4.714389736021213e-06, + "loss": 0.9335, + "step": 55779 + }, + { + "epoch": 4.12, + "learning_rate": 4.714051139979842e-06, + "loss": 0.9644, + "step": 55780 + }, + { + "epoch": 4.12, + "learning_rate": 4.7137125523483415e-06, + "loss": 0.9695, + "step": 55781 + }, + { + "epoch": 4.12, + "learning_rate": 4.713373973127246e-06, + "loss": 0.9427, + "step": 55782 + }, + { + "epoch": 4.12, + "learning_rate": 4.713035402317105e-06, + "loss": 1.0331, + "step": 55783 + }, + { + "epoch": 4.12, + "learning_rate": 4.7126968399184445e-06, + "loss": 0.9662, + "step": 55784 + }, + { + "epoch": 4.12, + "learning_rate": 4.712358285931805e-06, + "loss": 0.9755, + "step": 55785 + }, + { + "epoch": 4.12, + "learning_rate": 4.712019740357731e-06, + "loss": 1.0295, + "step": 55786 + }, + { + "epoch": 4.12, + "learning_rate": 4.711681203196757e-06, + "loss": 1.0538, + "step": 55787 + }, + { + "epoch": 4.12, + "learning_rate": 4.711342674449424e-06, + "loss": 1.0727, + "step": 55788 + }, + { + "epoch": 4.12, + "learning_rate": 4.7110041541162645e-06, + "loss": 1.1088, + "step": 55789 + }, + { + "epoch": 4.12, + "learning_rate": 4.7106656421978255e-06, + "loss": 0.99, + "step": 55790 + }, + { + "epoch": 4.12, + "learning_rate": 4.710327138694641e-06, + "loss": 0.9479, + "step": 55791 + }, + { + "epoch": 4.12, + "learning_rate": 4.709988643607252e-06, + "loss": 0.9704, + "step": 55792 + }, + { + "epoch": 4.12, + "learning_rate": 4.709650156936195e-06, + "loss": 1.0234, + "step": 55793 + }, + { + "epoch": 4.12, + "learning_rate": 4.709311678682005e-06, + "loss": 1.1261, + "step": 55794 + }, + { + "epoch": 4.12, + "learning_rate": 4.7089732088452274e-06, + "loss": 1.0421, + "step": 55795 + }, + { + "epoch": 4.12, + "learning_rate": 4.708634747426398e-06, + "loss": 1.0425, + "step": 55796 + }, + { + "epoch": 4.12, + "learning_rate": 4.708296294426056e-06, + "loss": 1.0541, + "step": 55797 + }, + { + "epoch": 4.12, + "learning_rate": 4.7079578498447345e-06, + "loss": 1.0261, + "step": 55798 + }, + { + "epoch": 4.12, + "learning_rate": 4.70761941368298e-06, + "loss": 0.9427, + "step": 55799 + }, + { + "epoch": 4.12, + "learning_rate": 4.707280985941327e-06, + "loss": 0.9957, + "step": 55800 + }, + { + "epoch": 4.12, + "learning_rate": 4.706942566620314e-06, + "loss": 1.0483, + "step": 55801 + }, + { + "epoch": 4.12, + "learning_rate": 4.70660415572048e-06, + "loss": 1.0357, + "step": 55802 + }, + { + "epoch": 4.12, + "learning_rate": 4.70626575324236e-06, + "loss": 1.0406, + "step": 55803 + }, + { + "epoch": 4.12, + "learning_rate": 4.705927359186503e-06, + "loss": 1.0131, + "step": 55804 + }, + { + "epoch": 4.12, + "learning_rate": 4.705588973553431e-06, + "loss": 0.9886, + "step": 55805 + }, + { + "epoch": 4.12, + "learning_rate": 4.705250596343696e-06, + "loss": 1.0562, + "step": 55806 + }, + { + "epoch": 4.12, + "learning_rate": 4.704912227557827e-06, + "loss": 1.0319, + "step": 55807 + }, + { + "epoch": 4.12, + "learning_rate": 4.704573867196374e-06, + "loss": 0.9776, + "step": 55808 + }, + { + "epoch": 4.12, + "learning_rate": 4.704235515259863e-06, + "loss": 0.9387, + "step": 55809 + }, + { + "epoch": 4.12, + "learning_rate": 4.703897171748839e-06, + "loss": 1.0506, + "step": 55810 + }, + { + "epoch": 4.12, + "learning_rate": 4.70355883666384e-06, + "loss": 0.9529, + "step": 55811 + }, + { + "epoch": 4.12, + "learning_rate": 4.703220510005403e-06, + "loss": 0.9414, + "step": 55812 + }, + { + "epoch": 4.12, + "learning_rate": 4.702882191774067e-06, + "loss": 1.0303, + "step": 55813 + }, + { + "epoch": 4.12, + "learning_rate": 4.702543881970364e-06, + "loss": 0.9754, + "step": 55814 + }, + { + "epoch": 4.12, + "learning_rate": 4.702205580594843e-06, + "loss": 1.0115, + "step": 55815 + }, + { + "epoch": 4.12, + "learning_rate": 4.701867287648036e-06, + "loss": 0.934, + "step": 55816 + }, + { + "epoch": 4.12, + "learning_rate": 4.701529003130484e-06, + "loss": 0.9632, + "step": 55817 + }, + { + "epoch": 4.12, + "learning_rate": 4.701190727042719e-06, + "loss": 0.9648, + "step": 55818 + }, + { + "epoch": 4.12, + "learning_rate": 4.700852459385288e-06, + "loss": 0.9937, + "step": 55819 + }, + { + "epoch": 4.12, + "learning_rate": 4.700514200158728e-06, + "loss": 1.0916, + "step": 55820 + }, + { + "epoch": 4.12, + "learning_rate": 4.700175949363567e-06, + "loss": 1.1078, + "step": 55821 + }, + { + "epoch": 4.12, + "learning_rate": 4.699837707000353e-06, + "loss": 0.9256, + "step": 55822 + }, + { + "epoch": 4.12, + "learning_rate": 4.699499473069618e-06, + "loss": 0.9363, + "step": 55823 + }, + { + "epoch": 4.12, + "learning_rate": 4.699161247571912e-06, + "loss": 1.0934, + "step": 55824 + }, + { + "epoch": 4.12, + "learning_rate": 4.698823030507756e-06, + "loss": 1.0276, + "step": 55825 + }, + { + "epoch": 4.12, + "learning_rate": 4.698484821877701e-06, + "loss": 0.9964, + "step": 55826 + }, + { + "epoch": 4.13, + "learning_rate": 4.698146621682277e-06, + "loss": 0.9822, + "step": 55827 + }, + { + "epoch": 4.13, + "learning_rate": 4.697808429922033e-06, + "loss": 1.0207, + "step": 55828 + }, + { + "epoch": 4.13, + "learning_rate": 4.697470246597496e-06, + "loss": 0.992, + "step": 55829 + }, + { + "epoch": 4.13, + "learning_rate": 4.697132071709205e-06, + "loss": 0.9122, + "step": 55830 + }, + { + "epoch": 4.13, + "learning_rate": 4.6967939052577046e-06, + "loss": 1.0692, + "step": 55831 + }, + { + "epoch": 4.13, + "learning_rate": 4.696455747243529e-06, + "loss": 0.927, + "step": 55832 + }, + { + "epoch": 4.13, + "learning_rate": 4.696117597667217e-06, + "loss": 1.1132, + "step": 55833 + }, + { + "epoch": 4.13, + "learning_rate": 4.695779456529302e-06, + "loss": 0.9592, + "step": 55834 + }, + { + "epoch": 4.13, + "learning_rate": 4.695441323830329e-06, + "loss": 0.9727, + "step": 55835 + }, + { + "epoch": 4.13, + "learning_rate": 4.695103199570834e-06, + "loss": 0.9103, + "step": 55836 + }, + { + "epoch": 4.13, + "learning_rate": 4.694765083751353e-06, + "loss": 0.9853, + "step": 55837 + }, + { + "epoch": 4.13, + "learning_rate": 4.694426976372422e-06, + "loss": 0.9938, + "step": 55838 + }, + { + "epoch": 4.13, + "learning_rate": 4.694088877434585e-06, + "loss": 1.0321, + "step": 55839 + }, + { + "epoch": 4.13, + "learning_rate": 4.693750786938376e-06, + "loss": 1.0486, + "step": 55840 + }, + { + "epoch": 4.13, + "learning_rate": 4.693412704884335e-06, + "loss": 0.9798, + "step": 55841 + }, + { + "epoch": 4.13, + "learning_rate": 4.693074631272997e-06, + "loss": 1.0568, + "step": 55842 + }, + { + "epoch": 4.13, + "learning_rate": 4.692736566104898e-06, + "loss": 1.0433, + "step": 55843 + }, + { + "epoch": 4.13, + "learning_rate": 4.6923985093805835e-06, + "loss": 1.0336, + "step": 55844 + }, + { + "epoch": 4.13, + "learning_rate": 4.692060461100587e-06, + "loss": 0.9862, + "step": 55845 + }, + { + "epoch": 4.13, + "learning_rate": 4.691722421265445e-06, + "loss": 1.0358, + "step": 55846 + }, + { + "epoch": 4.13, + "learning_rate": 4.691384389875695e-06, + "loss": 0.8817, + "step": 55847 + }, + { + "epoch": 4.13, + "learning_rate": 4.69104636693188e-06, + "loss": 1.045, + "step": 55848 + }, + { + "epoch": 4.13, + "learning_rate": 4.690708352434537e-06, + "loss": 0.9858, + "step": 55849 + }, + { + "epoch": 4.13, + "learning_rate": 4.690370346384193e-06, + "loss": 0.9983, + "step": 55850 + }, + { + "epoch": 4.13, + "learning_rate": 4.690032348781399e-06, + "loss": 0.9926, + "step": 55851 + }, + { + "epoch": 4.13, + "learning_rate": 4.689694359626687e-06, + "loss": 0.9988, + "step": 55852 + }, + { + "epoch": 4.13, + "learning_rate": 4.6893563789205955e-06, + "loss": 0.9913, + "step": 55853 + }, + { + "epoch": 4.13, + "learning_rate": 4.689018406663658e-06, + "loss": 1.0115, + "step": 55854 + }, + { + "epoch": 4.13, + "learning_rate": 4.68868044285642e-06, + "loss": 0.9168, + "step": 55855 + }, + { + "epoch": 4.13, + "learning_rate": 4.688342487499416e-06, + "loss": 1.1137, + "step": 55856 + }, + { + "epoch": 4.13, + "learning_rate": 4.688004540593182e-06, + "loss": 0.9941, + "step": 55857 + }, + { + "epoch": 4.13, + "learning_rate": 4.687666602138258e-06, + "loss": 0.9123, + "step": 55858 + }, + { + "epoch": 4.13, + "learning_rate": 4.687328672135175e-06, + "loss": 0.9165, + "step": 55859 + }, + { + "epoch": 4.13, + "learning_rate": 4.686990750584481e-06, + "loss": 1.0084, + "step": 55860 + }, + { + "epoch": 4.13, + "learning_rate": 4.686652837486708e-06, + "loss": 1.0243, + "step": 55861 + }, + { + "epoch": 4.13, + "learning_rate": 4.686314932842395e-06, + "loss": 0.9549, + "step": 55862 + }, + { + "epoch": 4.13, + "learning_rate": 4.6859770366520754e-06, + "loss": 1.0556, + "step": 55863 + }, + { + "epoch": 4.13, + "learning_rate": 4.685639148916293e-06, + "loss": 1.0434, + "step": 55864 + }, + { + "epoch": 4.13, + "learning_rate": 4.685301269635584e-06, + "loss": 0.9033, + "step": 55865 + }, + { + "epoch": 4.13, + "learning_rate": 4.684963398810484e-06, + "loss": 0.9887, + "step": 55866 + }, + { + "epoch": 4.13, + "learning_rate": 4.6846255364415315e-06, + "loss": 1.0544, + "step": 55867 + }, + { + "epoch": 4.13, + "learning_rate": 4.684287682529259e-06, + "loss": 1.0403, + "step": 55868 + }, + { + "epoch": 4.13, + "learning_rate": 4.6839498370742175e-06, + "loss": 1.0086, + "step": 55869 + }, + { + "epoch": 4.13, + "learning_rate": 4.683612000076927e-06, + "loss": 0.9394, + "step": 55870 + }, + { + "epoch": 4.13, + "learning_rate": 4.683274171537938e-06, + "loss": 0.9837, + "step": 55871 + }, + { + "epoch": 4.13, + "learning_rate": 4.6829363514577806e-06, + "loss": 1.0076, + "step": 55872 + }, + { + "epoch": 4.13, + "learning_rate": 4.682598539837002e-06, + "loss": 0.9058, + "step": 55873 + }, + { + "epoch": 4.13, + "learning_rate": 4.682260736676126e-06, + "loss": 0.9483, + "step": 55874 + }, + { + "epoch": 4.13, + "learning_rate": 4.681922941975701e-06, + "loss": 0.9865, + "step": 55875 + }, + { + "epoch": 4.13, + "learning_rate": 4.68158515573626e-06, + "loss": 1.1578, + "step": 55876 + }, + { + "epoch": 4.13, + "learning_rate": 4.681247377958341e-06, + "loss": 1.0004, + "step": 55877 + }, + { + "epoch": 4.13, + "learning_rate": 4.680909608642482e-06, + "loss": 0.97, + "step": 55878 + }, + { + "epoch": 4.13, + "learning_rate": 4.680571847789216e-06, + "loss": 1.032, + "step": 55879 + }, + { + "epoch": 4.13, + "learning_rate": 4.6802340953990866e-06, + "loss": 0.8685, + "step": 55880 + }, + { + "epoch": 4.13, + "learning_rate": 4.679896351472629e-06, + "loss": 0.9468, + "step": 55881 + }, + { + "epoch": 4.13, + "learning_rate": 4.679558616010381e-06, + "loss": 0.9684, + "step": 55882 + }, + { + "epoch": 4.13, + "learning_rate": 4.679220889012876e-06, + "loss": 0.9735, + "step": 55883 + }, + { + "epoch": 4.13, + "learning_rate": 4.6788831704806556e-06, + "loss": 1.0328, + "step": 55884 + }, + { + "epoch": 4.13, + "learning_rate": 4.678545460414258e-06, + "loss": 1.074, + "step": 55885 + }, + { + "epoch": 4.13, + "learning_rate": 4.678207758814218e-06, + "loss": 0.9881, + "step": 55886 + }, + { + "epoch": 4.13, + "learning_rate": 4.677870065681073e-06, + "loss": 1.0729, + "step": 55887 + }, + { + "epoch": 4.13, + "learning_rate": 4.677532381015357e-06, + "loss": 0.9642, + "step": 55888 + }, + { + "epoch": 4.13, + "learning_rate": 4.677194704817615e-06, + "loss": 0.9592, + "step": 55889 + }, + { + "epoch": 4.13, + "learning_rate": 4.676857037088379e-06, + "loss": 1.0831, + "step": 55890 + }, + { + "epoch": 4.13, + "learning_rate": 4.6765193778281885e-06, + "loss": 0.9908, + "step": 55891 + }, + { + "epoch": 4.13, + "learning_rate": 4.676181727037575e-06, + "loss": 0.8698, + "step": 55892 + }, + { + "epoch": 4.13, + "learning_rate": 4.675844084717084e-06, + "loss": 0.9662, + "step": 55893 + }, + { + "epoch": 4.13, + "learning_rate": 4.675506450867249e-06, + "loss": 0.989, + "step": 55894 + }, + { + "epoch": 4.13, + "learning_rate": 4.675168825488607e-06, + "loss": 1.0774, + "step": 55895 + }, + { + "epoch": 4.13, + "learning_rate": 4.674831208581695e-06, + "loss": 0.999, + "step": 55896 + }, + { + "epoch": 4.13, + "learning_rate": 4.674493600147051e-06, + "loss": 1.0319, + "step": 55897 + }, + { + "epoch": 4.13, + "learning_rate": 4.674156000185212e-06, + "loss": 0.9368, + "step": 55898 + }, + { + "epoch": 4.13, + "learning_rate": 4.67381840869671e-06, + "loss": 1.0375, + "step": 55899 + }, + { + "epoch": 4.13, + "learning_rate": 4.67348082568209e-06, + "loss": 0.9199, + "step": 55900 + }, + { + "epoch": 4.13, + "learning_rate": 4.673143251141886e-06, + "loss": 1.0923, + "step": 55901 + }, + { + "epoch": 4.13, + "learning_rate": 4.672805685076634e-06, + "loss": 0.9711, + "step": 55902 + }, + { + "epoch": 4.13, + "learning_rate": 4.67246812748687e-06, + "loss": 1.0506, + "step": 55903 + }, + { + "epoch": 4.13, + "learning_rate": 4.672130578373136e-06, + "loss": 0.9814, + "step": 55904 + }, + { + "epoch": 4.13, + "learning_rate": 4.6717930377359665e-06, + "loss": 1.0767, + "step": 55905 + }, + { + "epoch": 4.13, + "learning_rate": 4.6714555055758985e-06, + "loss": 0.961, + "step": 55906 + }, + { + "epoch": 4.13, + "learning_rate": 4.671117981893467e-06, + "loss": 1.0726, + "step": 55907 + }, + { + "epoch": 4.13, + "learning_rate": 4.670780466689207e-06, + "loss": 0.9655, + "step": 55908 + }, + { + "epoch": 4.13, + "learning_rate": 4.670442959963663e-06, + "loss": 1.0889, + "step": 55909 + }, + { + "epoch": 4.13, + "learning_rate": 4.670105461717369e-06, + "loss": 0.9753, + "step": 55910 + }, + { + "epoch": 4.13, + "learning_rate": 4.66976797195086e-06, + "loss": 1.069, + "step": 55911 + }, + { + "epoch": 4.13, + "learning_rate": 4.66943049066467e-06, + "loss": 0.9295, + "step": 55912 + }, + { + "epoch": 4.13, + "learning_rate": 4.669093017859344e-06, + "loss": 0.9367, + "step": 55913 + }, + { + "epoch": 4.13, + "learning_rate": 4.668755553535419e-06, + "loss": 1.0471, + "step": 55914 + }, + { + "epoch": 4.13, + "learning_rate": 4.668418097693419e-06, + "loss": 1.0106, + "step": 55915 + }, + { + "epoch": 4.13, + "learning_rate": 4.668080650333894e-06, + "loss": 1.0821, + "step": 55916 + }, + { + "epoch": 4.13, + "learning_rate": 4.667743211457373e-06, + "loss": 1.0391, + "step": 55917 + }, + { + "epoch": 4.13, + "learning_rate": 4.6674057810644026e-06, + "loss": 0.9447, + "step": 55918 + }, + { + "epoch": 4.13, + "learning_rate": 4.667068359155507e-06, + "loss": 1.0446, + "step": 55919 + }, + { + "epoch": 4.13, + "learning_rate": 4.666730945731231e-06, + "loss": 0.9808, + "step": 55920 + }, + { + "epoch": 4.13, + "learning_rate": 4.666393540792108e-06, + "loss": 1.0653, + "step": 55921 + }, + { + "epoch": 4.13, + "learning_rate": 4.666056144338684e-06, + "loss": 1.0887, + "step": 55922 + }, + { + "epoch": 4.13, + "learning_rate": 4.665718756371485e-06, + "loss": 0.9127, + "step": 55923 + }, + { + "epoch": 4.13, + "learning_rate": 4.665381376891046e-06, + "loss": 1.1133, + "step": 55924 + }, + { + "epoch": 4.13, + "learning_rate": 4.665044005897913e-06, + "loss": 0.9929, + "step": 55925 + }, + { + "epoch": 4.13, + "learning_rate": 4.664706643392619e-06, + "loss": 1.0967, + "step": 55926 + }, + { + "epoch": 4.13, + "learning_rate": 4.664369289375699e-06, + "loss": 1.0243, + "step": 55927 + }, + { + "epoch": 4.13, + "learning_rate": 4.66403194384769e-06, + "loss": 0.8845, + "step": 55928 + }, + { + "epoch": 4.13, + "learning_rate": 4.663694606809132e-06, + "loss": 1.0851, + "step": 55929 + }, + { + "epoch": 4.13, + "learning_rate": 4.6633572782605595e-06, + "loss": 0.9153, + "step": 55930 + }, + { + "epoch": 4.13, + "learning_rate": 4.663019958202509e-06, + "loss": 1.1233, + "step": 55931 + }, + { + "epoch": 4.13, + "learning_rate": 4.662682646635519e-06, + "loss": 1.0257, + "step": 55932 + }, + { + "epoch": 4.13, + "learning_rate": 4.662345343560119e-06, + "loss": 0.9886, + "step": 55933 + }, + { + "epoch": 4.13, + "learning_rate": 4.662008048976856e-06, + "loss": 0.9969, + "step": 55934 + }, + { + "epoch": 4.13, + "learning_rate": 4.661670762886261e-06, + "loss": 1.0006, + "step": 55935 + }, + { + "epoch": 4.13, + "learning_rate": 4.661333485288873e-06, + "loss": 1.0015, + "step": 55936 + }, + { + "epoch": 4.13, + "learning_rate": 4.660996216185222e-06, + "loss": 1.0949, + "step": 55937 + }, + { + "epoch": 4.13, + "learning_rate": 4.660658955575853e-06, + "loss": 0.9856, + "step": 55938 + }, + { + "epoch": 4.13, + "learning_rate": 4.660321703461299e-06, + "loss": 1.018, + "step": 55939 + }, + { + "epoch": 4.13, + "learning_rate": 4.659984459842098e-06, + "loss": 0.9145, + "step": 55940 + }, + { + "epoch": 4.13, + "learning_rate": 4.659647224718781e-06, + "loss": 0.9456, + "step": 55941 + }, + { + "epoch": 4.13, + "learning_rate": 4.659309998091897e-06, + "loss": 1.0236, + "step": 55942 + }, + { + "epoch": 4.13, + "learning_rate": 4.65897277996197e-06, + "loss": 1.0084, + "step": 55943 + }, + { + "epoch": 4.13, + "learning_rate": 4.658635570329537e-06, + "loss": 1.0484, + "step": 55944 + }, + { + "epoch": 4.13, + "learning_rate": 4.658298369195142e-06, + "loss": 1.0336, + "step": 55945 + }, + { + "epoch": 4.13, + "learning_rate": 4.657961176559318e-06, + "loss": 1.0451, + "step": 55946 + }, + { + "epoch": 4.13, + "learning_rate": 4.6576239924226016e-06, + "loss": 0.8959, + "step": 55947 + }, + { + "epoch": 4.13, + "learning_rate": 4.6572868167855255e-06, + "loss": 1.0087, + "step": 55948 + }, + { + "epoch": 4.13, + "learning_rate": 4.6569496496486335e-06, + "loss": 1.0373, + "step": 55949 + }, + { + "epoch": 4.13, + "learning_rate": 4.6566124910124574e-06, + "loss": 0.9605, + "step": 55950 + }, + { + "epoch": 4.13, + "learning_rate": 4.656275340877535e-06, + "loss": 1.0236, + "step": 55951 + }, + { + "epoch": 4.13, + "learning_rate": 4.655938199244402e-06, + "loss": 1.1541, + "step": 55952 + }, + { + "epoch": 4.13, + "learning_rate": 4.655601066113591e-06, + "loss": 1.0771, + "step": 55953 + }, + { + "epoch": 4.13, + "learning_rate": 4.655263941485645e-06, + "loss": 1.0416, + "step": 55954 + }, + { + "epoch": 4.13, + "learning_rate": 4.654926825361099e-06, + "loss": 1.0928, + "step": 55955 + }, + { + "epoch": 4.13, + "learning_rate": 4.654589717740487e-06, + "loss": 0.9431, + "step": 55956 + }, + { + "epoch": 4.13, + "learning_rate": 4.654252618624343e-06, + "loss": 1.1781, + "step": 55957 + }, + { + "epoch": 4.13, + "learning_rate": 4.653915528013211e-06, + "loss": 1.0366, + "step": 55958 + }, + { + "epoch": 4.13, + "learning_rate": 4.6535784459076215e-06, + "loss": 1.0264, + "step": 55959 + }, + { + "epoch": 4.13, + "learning_rate": 4.653241372308113e-06, + "loss": 1.0905, + "step": 55960 + }, + { + "epoch": 4.13, + "learning_rate": 4.652904307215221e-06, + "loss": 0.8757, + "step": 55961 + }, + { + "epoch": 4.13, + "learning_rate": 4.652567250629477e-06, + "loss": 0.9793, + "step": 55962 + }, + { + "epoch": 4.14, + "learning_rate": 4.652230202551431e-06, + "loss": 0.9404, + "step": 55963 + }, + { + "epoch": 4.14, + "learning_rate": 4.651893162981603e-06, + "loss": 1.0049, + "step": 55964 + }, + { + "epoch": 4.14, + "learning_rate": 4.651556131920538e-06, + "loss": 0.997, + "step": 55965 + }, + { + "epoch": 4.14, + "learning_rate": 4.651219109368769e-06, + "loss": 0.9425, + "step": 55966 + }, + { + "epoch": 4.14, + "learning_rate": 4.650882095326841e-06, + "loss": 0.9407, + "step": 55967 + }, + { + "epoch": 4.14, + "learning_rate": 4.650545089795274e-06, + "loss": 0.989, + "step": 55968 + }, + { + "epoch": 4.14, + "learning_rate": 4.6502080927746185e-06, + "loss": 1.0136, + "step": 55969 + }, + { + "epoch": 4.14, + "learning_rate": 4.6498711042654054e-06, + "loss": 1.0179, + "step": 55970 + }, + { + "epoch": 4.14, + "learning_rate": 4.6495341242681704e-06, + "loss": 0.9601, + "step": 55971 + }, + { + "epoch": 4.14, + "learning_rate": 4.64919715278345e-06, + "loss": 1.0382, + "step": 55972 + }, + { + "epoch": 4.14, + "learning_rate": 4.648860189811777e-06, + "loss": 0.9523, + "step": 55973 + }, + { + "epoch": 4.14, + "learning_rate": 4.648523235353693e-06, + "loss": 1.0862, + "step": 55974 + }, + { + "epoch": 4.14, + "learning_rate": 4.648186289409734e-06, + "loss": 1.0292, + "step": 55975 + }, + { + "epoch": 4.14, + "learning_rate": 4.647849351980434e-06, + "loss": 0.9423, + "step": 55976 + }, + { + "epoch": 4.14, + "learning_rate": 4.647512423066325e-06, + "loss": 1.0017, + "step": 55977 + }, + { + "epoch": 4.14, + "learning_rate": 4.64717550266795e-06, + "loss": 0.9936, + "step": 55978 + }, + { + "epoch": 4.14, + "learning_rate": 4.646838590785843e-06, + "loss": 0.9896, + "step": 55979 + }, + { + "epoch": 4.14, + "learning_rate": 4.646501687420539e-06, + "loss": 1.0085, + "step": 55980 + }, + { + "epoch": 4.14, + "learning_rate": 4.646164792572574e-06, + "loss": 1.0128, + "step": 55981 + }, + { + "epoch": 4.14, + "learning_rate": 4.645827906242481e-06, + "loss": 0.9554, + "step": 55982 + }, + { + "epoch": 4.14, + "learning_rate": 4.645491028430802e-06, + "loss": 0.9573, + "step": 55983 + }, + { + "epoch": 4.14, + "learning_rate": 4.645154159138071e-06, + "loss": 1.0791, + "step": 55984 + }, + { + "epoch": 4.14, + "learning_rate": 4.644817298364823e-06, + "loss": 0.9947, + "step": 55985 + }, + { + "epoch": 4.14, + "learning_rate": 4.6444804461115896e-06, + "loss": 0.9636, + "step": 55986 + }, + { + "epoch": 4.14, + "learning_rate": 4.64414360237892e-06, + "loss": 0.9218, + "step": 55987 + }, + { + "epoch": 4.14, + "learning_rate": 4.643806767167337e-06, + "loss": 0.9714, + "step": 55988 + }, + { + "epoch": 4.14, + "learning_rate": 4.643469940477376e-06, + "loss": 0.9744, + "step": 55989 + }, + { + "epoch": 4.14, + "learning_rate": 4.643133122309583e-06, + "loss": 0.9953, + "step": 55990 + }, + { + "epoch": 4.14, + "learning_rate": 4.642796312664488e-06, + "loss": 0.9414, + "step": 55991 + }, + { + "epoch": 4.14, + "learning_rate": 4.642459511542627e-06, + "loss": 0.8615, + "step": 55992 + }, + { + "epoch": 4.14, + "learning_rate": 4.642122718944533e-06, + "loss": 1.1301, + "step": 55993 + }, + { + "epoch": 4.14, + "learning_rate": 4.641785934870748e-06, + "loss": 1.0519, + "step": 55994 + }, + { + "epoch": 4.14, + "learning_rate": 4.641449159321807e-06, + "loss": 0.9175, + "step": 55995 + }, + { + "epoch": 4.14, + "learning_rate": 4.6411123922982415e-06, + "loss": 0.9753, + "step": 55996 + }, + { + "epoch": 4.14, + "learning_rate": 4.640775633800587e-06, + "loss": 0.9196, + "step": 55997 + }, + { + "epoch": 4.14, + "learning_rate": 4.640438883829384e-06, + "loss": 1.0908, + "step": 55998 + }, + { + "epoch": 4.14, + "learning_rate": 4.640102142385168e-06, + "loss": 1.0461, + "step": 55999 + }, + { + "epoch": 4.14, + "learning_rate": 4.639765409468472e-06, + "loss": 1.0564, + "step": 56000 + }, + { + "epoch": 4.14, + "learning_rate": 4.639428685079832e-06, + "loss": 0.8908, + "step": 56001 + }, + { + "epoch": 4.14, + "learning_rate": 4.6390919692197815e-06, + "loss": 0.9803, + "step": 56002 + }, + { + "epoch": 4.14, + "learning_rate": 4.638755261888862e-06, + "loss": 0.9931, + "step": 56003 + }, + { + "epoch": 4.14, + "learning_rate": 4.638418563087607e-06, + "loss": 1.0169, + "step": 56004 + }, + { + "epoch": 4.14, + "learning_rate": 4.638081872816551e-06, + "loss": 1.0322, + "step": 56005 + }, + { + "epoch": 4.14, + "learning_rate": 4.637745191076226e-06, + "loss": 0.9216, + "step": 56006 + }, + { + "epoch": 4.14, + "learning_rate": 4.637408517867175e-06, + "loss": 1.0012, + "step": 56007 + }, + { + "epoch": 4.14, + "learning_rate": 4.637071853189935e-06, + "loss": 1.1063, + "step": 56008 + }, + { + "epoch": 4.14, + "learning_rate": 4.636735197045029e-06, + "loss": 1.0018, + "step": 56009 + }, + { + "epoch": 4.14, + "learning_rate": 4.636398549433004e-06, + "loss": 0.8556, + "step": 56010 + }, + { + "epoch": 4.14, + "learning_rate": 4.636061910354389e-06, + "loss": 0.9803, + "step": 56011 + }, + { + "epoch": 4.14, + "learning_rate": 4.63572527980973e-06, + "loss": 0.9733, + "step": 56012 + }, + { + "epoch": 4.14, + "learning_rate": 4.635388657799548e-06, + "loss": 1.1141, + "step": 56013 + }, + { + "epoch": 4.14, + "learning_rate": 4.635052044324388e-06, + "loss": 0.8188, + "step": 56014 + }, + { + "epoch": 4.14, + "learning_rate": 4.634715439384783e-06, + "loss": 0.9832, + "step": 56015 + }, + { + "epoch": 4.14, + "learning_rate": 4.634378842981274e-06, + "loss": 0.9767, + "step": 56016 + }, + { + "epoch": 4.14, + "learning_rate": 4.634042255114388e-06, + "loss": 1.1682, + "step": 56017 + }, + { + "epoch": 4.14, + "learning_rate": 4.633705675784661e-06, + "loss": 1.1015, + "step": 56018 + }, + { + "epoch": 4.14, + "learning_rate": 4.633369104992635e-06, + "loss": 0.986, + "step": 56019 + }, + { + "epoch": 4.14, + "learning_rate": 4.6330325427388415e-06, + "loss": 0.9818, + "step": 56020 + }, + { + "epoch": 4.14, + "learning_rate": 4.632695989023816e-06, + "loss": 0.9979, + "step": 56021 + }, + { + "epoch": 4.14, + "learning_rate": 4.632359443848092e-06, + "loss": 0.9955, + "step": 56022 + }, + { + "epoch": 4.14, + "learning_rate": 4.632022907212211e-06, + "loss": 0.9956, + "step": 56023 + }, + { + "epoch": 4.14, + "learning_rate": 4.631686379116703e-06, + "loss": 0.9854, + "step": 56024 + }, + { + "epoch": 4.14, + "learning_rate": 4.631349859562108e-06, + "loss": 0.9391, + "step": 56025 + }, + { + "epoch": 4.14, + "learning_rate": 4.631013348548956e-06, + "loss": 1.0925, + "step": 56026 + }, + { + "epoch": 4.14, + "learning_rate": 4.630676846077782e-06, + "loss": 1.1158, + "step": 56027 + }, + { + "epoch": 4.14, + "learning_rate": 4.630340352149129e-06, + "loss": 1.0704, + "step": 56028 + }, + { + "epoch": 4.14, + "learning_rate": 4.6300038667635265e-06, + "loss": 1.0851, + "step": 56029 + }, + { + "epoch": 4.14, + "learning_rate": 4.629667389921512e-06, + "loss": 0.9653, + "step": 56030 + }, + { + "epoch": 4.14, + "learning_rate": 4.629330921623615e-06, + "loss": 1.0625, + "step": 56031 + }, + { + "epoch": 4.14, + "learning_rate": 4.628994461870383e-06, + "loss": 1.0402, + "step": 56032 + }, + { + "epoch": 4.14, + "learning_rate": 4.628658010662337e-06, + "loss": 1.0128, + "step": 56033 + }, + { + "epoch": 4.14, + "learning_rate": 4.628321568000023e-06, + "loss": 1.0947, + "step": 56034 + }, + { + "epoch": 4.14, + "learning_rate": 4.6279851338839735e-06, + "loss": 0.9885, + "step": 56035 + }, + { + "epoch": 4.14, + "learning_rate": 4.6276487083147215e-06, + "loss": 0.9727, + "step": 56036 + }, + { + "epoch": 4.14, + "learning_rate": 4.6273122912928036e-06, + "loss": 1.0459, + "step": 56037 + }, + { + "epoch": 4.14, + "learning_rate": 4.626975882818752e-06, + "loss": 1.044, + "step": 56038 + }, + { + "epoch": 4.14, + "learning_rate": 4.626639482893108e-06, + "loss": 0.9626, + "step": 56039 + }, + { + "epoch": 4.14, + "learning_rate": 4.626303091516404e-06, + "loss": 0.9782, + "step": 56040 + }, + { + "epoch": 4.14, + "learning_rate": 4.625966708689175e-06, + "loss": 1.0397, + "step": 56041 + }, + { + "epoch": 4.14, + "learning_rate": 4.625630334411952e-06, + "loss": 1.1761, + "step": 56042 + }, + { + "epoch": 4.14, + "learning_rate": 4.625293968685278e-06, + "loss": 0.9777, + "step": 56043 + }, + { + "epoch": 4.14, + "learning_rate": 4.624957611509685e-06, + "loss": 1.0078, + "step": 56044 + }, + { + "epoch": 4.14, + "learning_rate": 4.624621262885707e-06, + "loss": 1.0783, + "step": 56045 + }, + { + "epoch": 4.14, + "learning_rate": 4.62428492281388e-06, + "loss": 1.068, + "step": 56046 + }, + { + "epoch": 4.14, + "learning_rate": 4.623948591294735e-06, + "loss": 0.9812, + "step": 56047 + }, + { + "epoch": 4.14, + "learning_rate": 4.623612268328814e-06, + "loss": 1.0981, + "step": 56048 + }, + { + "epoch": 4.14, + "learning_rate": 4.6232759539166495e-06, + "loss": 1.0793, + "step": 56049 + }, + { + "epoch": 4.14, + "learning_rate": 4.6229396480587765e-06, + "loss": 1.0067, + "step": 56050 + }, + { + "epoch": 4.14, + "learning_rate": 4.622603350755725e-06, + "loss": 0.9911, + "step": 56051 + }, + { + "epoch": 4.14, + "learning_rate": 4.622267062008038e-06, + "loss": 1.092, + "step": 56052 + }, + { + "epoch": 4.14, + "learning_rate": 4.621930781816249e-06, + "loss": 1.0282, + "step": 56053 + }, + { + "epoch": 4.14, + "learning_rate": 4.621594510180889e-06, + "loss": 0.9363, + "step": 56054 + }, + { + "epoch": 4.14, + "learning_rate": 4.621258247102496e-06, + "loss": 0.9495, + "step": 56055 + }, + { + "epoch": 4.14, + "learning_rate": 4.620921992581601e-06, + "loss": 0.9734, + "step": 56056 + }, + { + "epoch": 4.14, + "learning_rate": 4.620585746618748e-06, + "loss": 1.0623, + "step": 56057 + }, + { + "epoch": 4.14, + "learning_rate": 4.6202495092144604e-06, + "loss": 1.0605, + "step": 56058 + }, + { + "epoch": 4.14, + "learning_rate": 4.619913280369282e-06, + "loss": 0.9359, + "step": 56059 + }, + { + "epoch": 4.14, + "learning_rate": 4.619577060083741e-06, + "loss": 1.0485, + "step": 56060 + }, + { + "epoch": 4.14, + "learning_rate": 4.6192408483583836e-06, + "loss": 0.8822, + "step": 56061 + }, + { + "epoch": 4.14, + "learning_rate": 4.618904645193729e-06, + "loss": 0.9192, + "step": 56062 + }, + { + "epoch": 4.14, + "learning_rate": 4.618568450590324e-06, + "loss": 0.8932, + "step": 56063 + }, + { + "epoch": 4.14, + "learning_rate": 4.6182322645487e-06, + "loss": 1.1042, + "step": 56064 + }, + { + "epoch": 4.14, + "learning_rate": 4.617896087069391e-06, + "loss": 1.0745, + "step": 56065 + }, + { + "epoch": 4.14, + "learning_rate": 4.617559918152933e-06, + "loss": 1.2267, + "step": 56066 + }, + { + "epoch": 4.14, + "learning_rate": 4.617223757799856e-06, + "loss": 0.9424, + "step": 56067 + }, + { + "epoch": 4.14, + "learning_rate": 4.616887606010701e-06, + "loss": 0.8466, + "step": 56068 + }, + { + "epoch": 4.14, + "learning_rate": 4.616551462786003e-06, + "loss": 1.1037, + "step": 56069 + }, + { + "epoch": 4.14, + "learning_rate": 4.616215328126295e-06, + "loss": 1.0369, + "step": 56070 + }, + { + "epoch": 4.14, + "learning_rate": 4.615879202032107e-06, + "loss": 0.9167, + "step": 56071 + }, + { + "epoch": 4.14, + "learning_rate": 4.6155430845039805e-06, + "loss": 0.9196, + "step": 56072 + }, + { + "epoch": 4.14, + "learning_rate": 4.6152069755424525e-06, + "loss": 0.9184, + "step": 56073 + }, + { + "epoch": 4.14, + "learning_rate": 4.614870875148046e-06, + "loss": 0.9693, + "step": 56074 + }, + { + "epoch": 4.14, + "learning_rate": 4.614534783321306e-06, + "loss": 0.9764, + "step": 56075 + }, + { + "epoch": 4.14, + "learning_rate": 4.614198700062762e-06, + "loss": 1.074, + "step": 56076 + }, + { + "epoch": 4.14, + "learning_rate": 4.6138626253729565e-06, + "loss": 1.1436, + "step": 56077 + }, + { + "epoch": 4.14, + "learning_rate": 4.613526559252412e-06, + "loss": 1.022, + "step": 56078 + }, + { + "epoch": 4.14, + "learning_rate": 4.613190501701673e-06, + "loss": 1.0798, + "step": 56079 + }, + { + "epoch": 4.14, + "learning_rate": 4.612854452721267e-06, + "loss": 0.9658, + "step": 56080 + }, + { + "epoch": 4.14, + "learning_rate": 4.612518412311741e-06, + "loss": 1.0233, + "step": 56081 + }, + { + "epoch": 4.14, + "learning_rate": 4.612182380473616e-06, + "loss": 0.9137, + "step": 56082 + }, + { + "epoch": 4.14, + "learning_rate": 4.611846357207428e-06, + "loss": 0.988, + "step": 56083 + }, + { + "epoch": 4.14, + "learning_rate": 4.61151034251372e-06, + "loss": 1.0232, + "step": 56084 + }, + { + "epoch": 4.14, + "learning_rate": 4.611174336393022e-06, + "loss": 1.0036, + "step": 56085 + }, + { + "epoch": 4.14, + "learning_rate": 4.610838338845867e-06, + "loss": 1.0561, + "step": 56086 + }, + { + "epoch": 4.14, + "learning_rate": 4.61050234987279e-06, + "loss": 0.93, + "step": 56087 + }, + { + "epoch": 4.14, + "learning_rate": 4.610166369474328e-06, + "loss": 1.004, + "step": 56088 + }, + { + "epoch": 4.14, + "learning_rate": 4.609830397651015e-06, + "loss": 0.9918, + "step": 56089 + }, + { + "epoch": 4.14, + "learning_rate": 4.6094944344033855e-06, + "loss": 0.9833, + "step": 56090 + }, + { + "epoch": 4.14, + "learning_rate": 4.6091584797319725e-06, + "loss": 0.8943, + "step": 56091 + }, + { + "epoch": 4.14, + "learning_rate": 4.608822533637307e-06, + "loss": 0.9832, + "step": 56092 + }, + { + "epoch": 4.14, + "learning_rate": 4.608486596119932e-06, + "loss": 0.918, + "step": 56093 + }, + { + "epoch": 4.14, + "learning_rate": 4.6081506671803774e-06, + "loss": 0.9309, + "step": 56094 + }, + { + "epoch": 4.14, + "learning_rate": 4.607814746819178e-06, + "loss": 1.0512, + "step": 56095 + }, + { + "epoch": 4.14, + "learning_rate": 4.6074788350368645e-06, + "loss": 1.0656, + "step": 56096 + }, + { + "epoch": 4.14, + "learning_rate": 4.607142931833979e-06, + "loss": 0.896, + "step": 56097 + }, + { + "epoch": 4.15, + "learning_rate": 4.6068070372110515e-06, + "loss": 0.9657, + "step": 56098 + }, + { + "epoch": 4.15, + "learning_rate": 4.606471151168618e-06, + "loss": 1.0215, + "step": 56099 + }, + { + "epoch": 4.15, + "learning_rate": 4.606135273707211e-06, + "loss": 0.9774, + "step": 56100 + }, + { + "epoch": 4.15, + "learning_rate": 4.6057994048273614e-06, + "loss": 0.9976, + "step": 56101 + }, + { + "epoch": 4.15, + "learning_rate": 4.605463544529616e-06, + "loss": 0.9664, + "step": 56102 + }, + { + "epoch": 4.15, + "learning_rate": 4.605127692814493e-06, + "loss": 1.0593, + "step": 56103 + }, + { + "epoch": 4.15, + "learning_rate": 4.604791849682538e-06, + "loss": 0.9785, + "step": 56104 + }, + { + "epoch": 4.15, + "learning_rate": 4.604456015134278e-06, + "loss": 1.0305, + "step": 56105 + }, + { + "epoch": 4.15, + "learning_rate": 4.60412018917026e-06, + "loss": 1.0998, + "step": 56106 + }, + { + "epoch": 4.15, + "learning_rate": 4.603784371791002e-06, + "loss": 0.995, + "step": 56107 + }, + { + "epoch": 4.15, + "learning_rate": 4.60344856299705e-06, + "loss": 1.0311, + "step": 56108 + }, + { + "epoch": 4.15, + "learning_rate": 4.603112762788933e-06, + "loss": 1.0385, + "step": 56109 + }, + { + "epoch": 4.15, + "learning_rate": 4.6027769711671855e-06, + "loss": 1.047, + "step": 56110 + }, + { + "epoch": 4.15, + "learning_rate": 4.602441188132344e-06, + "loss": 0.9993, + "step": 56111 + }, + { + "epoch": 4.15, + "learning_rate": 4.6021054136849375e-06, + "loss": 1.0817, + "step": 56112 + }, + { + "epoch": 4.15, + "learning_rate": 4.601769647825508e-06, + "loss": 1.0694, + "step": 56113 + }, + { + "epoch": 4.15, + "learning_rate": 4.601433890554585e-06, + "loss": 1.0202, + "step": 56114 + }, + { + "epoch": 4.15, + "learning_rate": 4.601098141872704e-06, + "loss": 1.0118, + "step": 56115 + }, + { + "epoch": 4.15, + "learning_rate": 4.600762401780394e-06, + "loss": 0.8521, + "step": 56116 + }, + { + "epoch": 4.15, + "learning_rate": 4.600426670278198e-06, + "loss": 0.8818, + "step": 56117 + }, + { + "epoch": 4.15, + "learning_rate": 4.600090947366646e-06, + "loss": 1.131, + "step": 56118 + }, + { + "epoch": 4.15, + "learning_rate": 4.5997552330462735e-06, + "loss": 0.931, + "step": 56119 + }, + { + "epoch": 4.15, + "learning_rate": 4.599419527317611e-06, + "loss": 0.9647, + "step": 56120 + }, + { + "epoch": 4.15, + "learning_rate": 4.599083830181193e-06, + "loss": 0.8127, + "step": 56121 + }, + { + "epoch": 4.15, + "learning_rate": 4.5987481416375614e-06, + "loss": 0.9376, + "step": 56122 + }, + { + "epoch": 4.15, + "learning_rate": 4.598412461687237e-06, + "loss": 1.0274, + "step": 56123 + }, + { + "epoch": 4.15, + "learning_rate": 4.598076790330765e-06, + "loss": 0.9453, + "step": 56124 + }, + { + "epoch": 4.15, + "learning_rate": 4.5977411275686715e-06, + "loss": 0.9773, + "step": 56125 + }, + { + "epoch": 4.15, + "learning_rate": 4.597405473401501e-06, + "loss": 1.0491, + "step": 56126 + }, + { + "epoch": 4.15, + "learning_rate": 4.597069827829775e-06, + "loss": 0.9944, + "step": 56127 + }, + { + "epoch": 4.15, + "learning_rate": 4.596734190854037e-06, + "loss": 0.9898, + "step": 56128 + }, + { + "epoch": 4.15, + "learning_rate": 4.596398562474819e-06, + "loss": 0.9902, + "step": 56129 + }, + { + "epoch": 4.15, + "learning_rate": 4.596062942692651e-06, + "loss": 0.9257, + "step": 56130 + }, + { + "epoch": 4.15, + "learning_rate": 4.5957273315080716e-06, + "loss": 0.9906, + "step": 56131 + }, + { + "epoch": 4.15, + "learning_rate": 4.595391728921609e-06, + "loss": 0.9, + "step": 56132 + }, + { + "epoch": 4.15, + "learning_rate": 4.595056134933803e-06, + "loss": 1.0346, + "step": 56133 + }, + { + "epoch": 4.15, + "learning_rate": 4.594720549545186e-06, + "loss": 1.0133, + "step": 56134 + }, + { + "epoch": 4.15, + "learning_rate": 4.594384972756292e-06, + "loss": 1.0774, + "step": 56135 + }, + { + "epoch": 4.15, + "learning_rate": 4.59404940456765e-06, + "loss": 0.9917, + "step": 56136 + }, + { + "epoch": 4.15, + "learning_rate": 4.5937138449798015e-06, + "loss": 0.9937, + "step": 56137 + }, + { + "epoch": 4.15, + "learning_rate": 4.593378293993277e-06, + "loss": 0.9331, + "step": 56138 + }, + { + "epoch": 4.15, + "learning_rate": 4.593042751608611e-06, + "loss": 0.9754, + "step": 56139 + }, + { + "epoch": 4.15, + "learning_rate": 4.592707217826335e-06, + "loss": 0.9911, + "step": 56140 + }, + { + "epoch": 4.15, + "learning_rate": 4.592371692646982e-06, + "loss": 0.9243, + "step": 56141 + }, + { + "epoch": 4.15, + "learning_rate": 4.592036176071092e-06, + "loss": 1.0523, + "step": 56142 + }, + { + "epoch": 4.15, + "learning_rate": 4.591700668099195e-06, + "loss": 0.977, + "step": 56143 + }, + { + "epoch": 4.15, + "learning_rate": 4.591365168731824e-06, + "loss": 1.0232, + "step": 56144 + }, + { + "epoch": 4.15, + "learning_rate": 4.591029677969511e-06, + "loss": 0.9859, + "step": 56145 + }, + { + "epoch": 4.15, + "learning_rate": 4.590694195812795e-06, + "loss": 1.0165, + "step": 56146 + }, + { + "epoch": 4.15, + "learning_rate": 4.590358722262212e-06, + "loss": 1.0433, + "step": 56147 + }, + { + "epoch": 4.15, + "learning_rate": 4.590023257318282e-06, + "loss": 1.0175, + "step": 56148 + }, + { + "epoch": 4.15, + "learning_rate": 4.589687800981553e-06, + "loss": 0.9471, + "step": 56149 + }, + { + "epoch": 4.15, + "learning_rate": 4.589352353252549e-06, + "loss": 1.0404, + "step": 56150 + }, + { + "epoch": 4.15, + "learning_rate": 4.589016914131815e-06, + "loss": 0.9781, + "step": 56151 + }, + { + "epoch": 4.15, + "learning_rate": 4.58868148361987e-06, + "loss": 1.0265, + "step": 56152 + }, + { + "epoch": 4.15, + "learning_rate": 4.588346061717261e-06, + "loss": 0.9019, + "step": 56153 + }, + { + "epoch": 4.15, + "learning_rate": 4.588010648424514e-06, + "loss": 1.0107, + "step": 56154 + }, + { + "epoch": 4.15, + "learning_rate": 4.5876752437421646e-06, + "loss": 1.015, + "step": 56155 + }, + { + "epoch": 4.15, + "learning_rate": 4.587339847670743e-06, + "loss": 0.9956, + "step": 56156 + }, + { + "epoch": 4.15, + "learning_rate": 4.587004460210791e-06, + "loss": 1.0401, + "step": 56157 + }, + { + "epoch": 4.15, + "learning_rate": 4.586669081362836e-06, + "loss": 1.1064, + "step": 56158 + }, + { + "epoch": 4.15, + "learning_rate": 4.5863337111274145e-06, + "loss": 0.9241, + "step": 56159 + }, + { + "epoch": 4.15, + "learning_rate": 4.5859983495050574e-06, + "loss": 1.0143, + "step": 56160 + }, + { + "epoch": 4.15, + "learning_rate": 4.585662996496296e-06, + "loss": 0.931, + "step": 56161 + }, + { + "epoch": 4.15, + "learning_rate": 4.585327652101672e-06, + "loss": 1.0002, + "step": 56162 + }, + { + "epoch": 4.15, + "learning_rate": 4.584992316321714e-06, + "loss": 0.9949, + "step": 56163 + }, + { + "epoch": 4.15, + "learning_rate": 4.584656989156954e-06, + "loss": 1.0168, + "step": 56164 + }, + { + "epoch": 4.15, + "learning_rate": 4.584321670607925e-06, + "loss": 0.8583, + "step": 56165 + }, + { + "epoch": 4.15, + "learning_rate": 4.583986360675167e-06, + "loss": 0.8697, + "step": 56166 + }, + { + "epoch": 4.15, + "learning_rate": 4.583651059359211e-06, + "loss": 0.945, + "step": 56167 + }, + { + "epoch": 4.15, + "learning_rate": 4.583315766660583e-06, + "loss": 1.0297, + "step": 56168 + }, + { + "epoch": 4.15, + "learning_rate": 4.5829804825798255e-06, + "loss": 0.9513, + "step": 56169 + }, + { + "epoch": 4.15, + "learning_rate": 4.582645207117464e-06, + "loss": 0.9614, + "step": 56170 + }, + { + "epoch": 4.15, + "learning_rate": 4.582309940274044e-06, + "loss": 0.9967, + "step": 56171 + }, + { + "epoch": 4.15, + "learning_rate": 4.581974682050084e-06, + "loss": 0.9466, + "step": 56172 + }, + { + "epoch": 4.15, + "learning_rate": 4.58163943244613e-06, + "loss": 1.0229, + "step": 56173 + }, + { + "epoch": 4.15, + "learning_rate": 4.5813041914627054e-06, + "loss": 0.9056, + "step": 56174 + }, + { + "epoch": 4.15, + "learning_rate": 4.580968959100354e-06, + "loss": 0.8412, + "step": 56175 + }, + { + "epoch": 4.15, + "learning_rate": 4.580633735359602e-06, + "loss": 1.0148, + "step": 56176 + }, + { + "epoch": 4.15, + "learning_rate": 4.58029852024098e-06, + "loss": 0.8932, + "step": 56177 + }, + { + "epoch": 4.15, + "learning_rate": 4.579963313745029e-06, + "loss": 0.8983, + "step": 56178 + }, + { + "epoch": 4.15, + "learning_rate": 4.5796281158722785e-06, + "loss": 1.0706, + "step": 56179 + }, + { + "epoch": 4.15, + "learning_rate": 4.579292926623263e-06, + "loss": 0.9313, + "step": 56180 + }, + { + "epoch": 4.15, + "learning_rate": 4.578957745998511e-06, + "loss": 0.9466, + "step": 56181 + }, + { + "epoch": 4.15, + "learning_rate": 4.578622573998562e-06, + "loss": 0.9796, + "step": 56182 + }, + { + "epoch": 4.15, + "learning_rate": 4.578287410623949e-06, + "loss": 1.0157, + "step": 56183 + }, + { + "epoch": 4.15, + "learning_rate": 4.577952255875202e-06, + "loss": 0.9495, + "step": 56184 + }, + { + "epoch": 4.15, + "learning_rate": 4.577617109752856e-06, + "loss": 1.1181, + "step": 56185 + }, + { + "epoch": 4.15, + "learning_rate": 4.577281972257439e-06, + "loss": 0.9601, + "step": 56186 + }, + { + "epoch": 4.15, + "learning_rate": 4.576946843389493e-06, + "loss": 0.8221, + "step": 56187 + }, + { + "epoch": 4.15, + "learning_rate": 4.576611723149547e-06, + "loss": 0.9603, + "step": 56188 + }, + { + "epoch": 4.15, + "learning_rate": 4.576276611538134e-06, + "loss": 1.0127, + "step": 56189 + }, + { + "epoch": 4.15, + "learning_rate": 4.575941508555783e-06, + "loss": 1.1053, + "step": 56190 + }, + { + "epoch": 4.15, + "learning_rate": 4.575606414203037e-06, + "loss": 0.9881, + "step": 56191 + }, + { + "epoch": 4.15, + "learning_rate": 4.575271328480422e-06, + "loss": 1.06, + "step": 56192 + }, + { + "epoch": 4.15, + "learning_rate": 4.574936251388472e-06, + "loss": 0.9894, + "step": 56193 + }, + { + "epoch": 4.15, + "learning_rate": 4.574601182927723e-06, + "loss": 0.9439, + "step": 56194 + }, + { + "epoch": 4.15, + "learning_rate": 4.574266123098704e-06, + "loss": 1.0121, + "step": 56195 + }, + { + "epoch": 4.15, + "learning_rate": 4.5739310719019504e-06, + "loss": 0.8394, + "step": 56196 + }, + { + "epoch": 4.15, + "learning_rate": 4.573596029337992e-06, + "loss": 0.9526, + "step": 56197 + }, + { + "epoch": 4.15, + "learning_rate": 4.573260995407368e-06, + "loss": 0.9704, + "step": 56198 + }, + { + "epoch": 4.15, + "learning_rate": 4.572925970110608e-06, + "loss": 1.1081, + "step": 56199 + }, + { + "epoch": 4.15, + "learning_rate": 4.5725909534482445e-06, + "loss": 1.097, + "step": 56200 + }, + { + "epoch": 4.15, + "learning_rate": 4.5722559454208085e-06, + "loss": 0.9124, + "step": 56201 + }, + { + "epoch": 4.15, + "learning_rate": 4.571920946028839e-06, + "loss": 1.0116, + "step": 56202 + }, + { + "epoch": 4.15, + "learning_rate": 4.5715859552728644e-06, + "loss": 1.0247, + "step": 56203 + }, + { + "epoch": 4.15, + "learning_rate": 4.57125097315342e-06, + "loss": 1.0525, + "step": 56204 + }, + { + "epoch": 4.15, + "learning_rate": 4.570915999671038e-06, + "loss": 0.9722, + "step": 56205 + }, + { + "epoch": 4.15, + "learning_rate": 4.570581034826247e-06, + "loss": 1.0115, + "step": 56206 + }, + { + "epoch": 4.15, + "learning_rate": 4.570246078619588e-06, + "loss": 1.0278, + "step": 56207 + }, + { + "epoch": 4.15, + "learning_rate": 4.5699111310515884e-06, + "loss": 0.9943, + "step": 56208 + }, + { + "epoch": 4.15, + "learning_rate": 4.569576192122784e-06, + "loss": 1.1232, + "step": 56209 + }, + { + "epoch": 4.15, + "learning_rate": 4.569241261833701e-06, + "loss": 0.9289, + "step": 56210 + }, + { + "epoch": 4.15, + "learning_rate": 4.568906340184882e-06, + "loss": 1.1975, + "step": 56211 + }, + { + "epoch": 4.15, + "learning_rate": 4.568571427176856e-06, + "loss": 1.0633, + "step": 56212 + }, + { + "epoch": 4.15, + "learning_rate": 4.568236522810155e-06, + "loss": 1.0367, + "step": 56213 + }, + { + "epoch": 4.15, + "learning_rate": 4.567901627085311e-06, + "loss": 1.0452, + "step": 56214 + }, + { + "epoch": 4.15, + "learning_rate": 4.567566740002855e-06, + "loss": 1.0034, + "step": 56215 + }, + { + "epoch": 4.15, + "learning_rate": 4.567231861563328e-06, + "loss": 1.0328, + "step": 56216 + }, + { + "epoch": 4.15, + "learning_rate": 4.566896991767253e-06, + "loss": 1.0447, + "step": 56217 + }, + { + "epoch": 4.15, + "learning_rate": 4.566562130615169e-06, + "loss": 0.9923, + "step": 56218 + }, + { + "epoch": 4.15, + "learning_rate": 4.566227278107603e-06, + "loss": 0.8869, + "step": 56219 + }, + { + "epoch": 4.15, + "learning_rate": 4.565892434245099e-06, + "loss": 0.9621, + "step": 56220 + }, + { + "epoch": 4.15, + "learning_rate": 4.5655575990281765e-06, + "loss": 1.0838, + "step": 56221 + }, + { + "epoch": 4.15, + "learning_rate": 4.565222772457376e-06, + "loss": 0.9158, + "step": 56222 + }, + { + "epoch": 4.15, + "learning_rate": 4.5648879545332295e-06, + "loss": 1.0111, + "step": 56223 + }, + { + "epoch": 4.15, + "learning_rate": 4.564553145256268e-06, + "loss": 1.0953, + "step": 56224 + }, + { + "epoch": 4.15, + "learning_rate": 4.564218344627025e-06, + "loss": 1.0163, + "step": 56225 + }, + { + "epoch": 4.15, + "learning_rate": 4.563883552646029e-06, + "loss": 1.0166, + "step": 56226 + }, + { + "epoch": 4.15, + "learning_rate": 4.56354876931382e-06, + "loss": 1.0212, + "step": 56227 + }, + { + "epoch": 4.15, + "learning_rate": 4.563213994630927e-06, + "loss": 1.0408, + "step": 56228 + }, + { + "epoch": 4.15, + "learning_rate": 4.562879228597884e-06, + "loss": 1.0024, + "step": 56229 + }, + { + "epoch": 4.15, + "learning_rate": 4.5625444712152175e-06, + "loss": 0.9955, + "step": 56230 + }, + { + "epoch": 4.15, + "learning_rate": 4.562209722483469e-06, + "loss": 0.9772, + "step": 56231 + }, + { + "epoch": 4.15, + "learning_rate": 4.561874982403167e-06, + "loss": 0.9802, + "step": 56232 + }, + { + "epoch": 4.16, + "learning_rate": 4.561540250974844e-06, + "loss": 0.958, + "step": 56233 + }, + { + "epoch": 4.16, + "learning_rate": 4.561205528199032e-06, + "loss": 1.0174, + "step": 56234 + }, + { + "epoch": 4.16, + "learning_rate": 4.560870814076262e-06, + "loss": 1.0741, + "step": 56235 + }, + { + "epoch": 4.16, + "learning_rate": 4.560536108607071e-06, + "loss": 0.9903, + "step": 56236 + }, + { + "epoch": 4.16, + "learning_rate": 4.56020141179199e-06, + "loss": 1.0445, + "step": 56237 + }, + { + "epoch": 4.16, + "learning_rate": 4.559866723631551e-06, + "loss": 1.0548, + "step": 56238 + }, + { + "epoch": 4.16, + "learning_rate": 4.5595320441262815e-06, + "loss": 0.8974, + "step": 56239 + }, + { + "epoch": 4.16, + "learning_rate": 4.559197373276727e-06, + "loss": 1.0349, + "step": 56240 + }, + { + "epoch": 4.16, + "learning_rate": 4.558862711083407e-06, + "loss": 1.0213, + "step": 56241 + }, + { + "epoch": 4.16, + "learning_rate": 4.558528057546855e-06, + "loss": 1.0129, + "step": 56242 + }, + { + "epoch": 4.16, + "learning_rate": 4.558193412667612e-06, + "loss": 1.1049, + "step": 56243 + }, + { + "epoch": 4.16, + "learning_rate": 4.557858776446204e-06, + "loss": 0.941, + "step": 56244 + }, + { + "epoch": 4.16, + "learning_rate": 4.557524148883166e-06, + "loss": 0.9627, + "step": 56245 + }, + { + "epoch": 4.16, + "learning_rate": 4.557189529979023e-06, + "loss": 0.9498, + "step": 56246 + }, + { + "epoch": 4.16, + "learning_rate": 4.55685491973432e-06, + "loss": 1.0133, + "step": 56247 + }, + { + "epoch": 4.16, + "learning_rate": 4.556520318149582e-06, + "loss": 1.017, + "step": 56248 + }, + { + "epoch": 4.16, + "learning_rate": 4.556185725225342e-06, + "loss": 0.9227, + "step": 56249 + }, + { + "epoch": 4.16, + "learning_rate": 4.555851140962134e-06, + "loss": 1.0171, + "step": 56250 + }, + { + "epoch": 4.16, + "learning_rate": 4.555516565360484e-06, + "loss": 1.1232, + "step": 56251 + }, + { + "epoch": 4.16, + "learning_rate": 4.555181998420933e-06, + "loss": 0.9774, + "step": 56252 + }, + { + "epoch": 4.16, + "learning_rate": 4.55484744014401e-06, + "loss": 1.0888, + "step": 56253 + }, + { + "epoch": 4.16, + "learning_rate": 4.554512890530246e-06, + "loss": 0.9873, + "step": 56254 + }, + { + "epoch": 4.16, + "learning_rate": 4.5541783495801715e-06, + "loss": 1.0696, + "step": 56255 + }, + { + "epoch": 4.16, + "learning_rate": 4.553843817294325e-06, + "loss": 0.9822, + "step": 56256 + }, + { + "epoch": 4.16, + "learning_rate": 4.553509293673234e-06, + "loss": 1.0113, + "step": 56257 + }, + { + "epoch": 4.16, + "learning_rate": 4.5531747787174315e-06, + "loss": 0.9341, + "step": 56258 + }, + { + "epoch": 4.16, + "learning_rate": 4.552840272427451e-06, + "loss": 0.8587, + "step": 56259 + }, + { + "epoch": 4.16, + "learning_rate": 4.552505774803821e-06, + "loss": 0.9989, + "step": 56260 + }, + { + "epoch": 4.16, + "learning_rate": 4.552171285847081e-06, + "loss": 0.9615, + "step": 56261 + }, + { + "epoch": 4.16, + "learning_rate": 4.551836805557753e-06, + "loss": 0.9105, + "step": 56262 + }, + { + "epoch": 4.16, + "learning_rate": 4.551502333936378e-06, + "loss": 1.0736, + "step": 56263 + }, + { + "epoch": 4.16, + "learning_rate": 4.551167870983482e-06, + "loss": 0.9954, + "step": 56264 + }, + { + "epoch": 4.16, + "learning_rate": 4.550833416699606e-06, + "loss": 1.0176, + "step": 56265 + }, + { + "epoch": 4.16, + "learning_rate": 4.55049897108527e-06, + "loss": 1.1122, + "step": 56266 + }, + { + "epoch": 4.16, + "learning_rate": 4.550164534141015e-06, + "loss": 0.9625, + "step": 56267 + }, + { + "epoch": 4.16, + "learning_rate": 4.549830105867366e-06, + "loss": 1.0445, + "step": 56268 + }, + { + "epoch": 4.16, + "learning_rate": 4.549495686264867e-06, + "loss": 0.996, + "step": 56269 + }, + { + "epoch": 4.16, + "learning_rate": 4.54916127533404e-06, + "loss": 0.977, + "step": 56270 + }, + { + "epoch": 4.16, + "learning_rate": 4.548826873075415e-06, + "loss": 1.0537, + "step": 56271 + }, + { + "epoch": 4.16, + "learning_rate": 4.548492479489531e-06, + "loss": 1.0108, + "step": 56272 + }, + { + "epoch": 4.16, + "learning_rate": 4.548158094576919e-06, + "loss": 0.9344, + "step": 56273 + }, + { + "epoch": 4.16, + "learning_rate": 4.547823718338109e-06, + "loss": 1.071, + "step": 56274 + }, + { + "epoch": 4.16, + "learning_rate": 4.54748935077363e-06, + "loss": 1.0205, + "step": 56275 + }, + { + "epoch": 4.16, + "learning_rate": 4.54715499188402e-06, + "loss": 0.9798, + "step": 56276 + }, + { + "epoch": 4.16, + "learning_rate": 4.546820641669809e-06, + "loss": 0.9942, + "step": 56277 + }, + { + "epoch": 4.16, + "learning_rate": 4.546486300131528e-06, + "loss": 0.8884, + "step": 56278 + }, + { + "epoch": 4.16, + "learning_rate": 4.546151967269711e-06, + "loss": 0.9052, + "step": 56279 + }, + { + "epoch": 4.16, + "learning_rate": 4.545817643084882e-06, + "loss": 1.0111, + "step": 56280 + }, + { + "epoch": 4.16, + "learning_rate": 4.545483327577584e-06, + "loss": 1.0615, + "step": 56281 + }, + { + "epoch": 4.16, + "learning_rate": 4.545149020748345e-06, + "loss": 0.9615, + "step": 56282 + }, + { + "epoch": 4.16, + "learning_rate": 4.544814722597694e-06, + "loss": 0.9952, + "step": 56283 + }, + { + "epoch": 4.16, + "learning_rate": 4.544480433126163e-06, + "loss": 0.8553, + "step": 56284 + }, + { + "epoch": 4.16, + "learning_rate": 4.544146152334291e-06, + "loss": 1.0307, + "step": 56285 + }, + { + "epoch": 4.16, + "learning_rate": 4.543811880222598e-06, + "loss": 0.9829, + "step": 56286 + }, + { + "epoch": 4.16, + "learning_rate": 4.543477616791626e-06, + "loss": 0.9848, + "step": 56287 + }, + { + "epoch": 4.16, + "learning_rate": 4.543143362041903e-06, + "loss": 1.0641, + "step": 56288 + }, + { + "epoch": 4.16, + "learning_rate": 4.542809115973961e-06, + "loss": 0.9646, + "step": 56289 + }, + { + "epoch": 4.16, + "learning_rate": 4.542474878588331e-06, + "loss": 1.0226, + "step": 56290 + }, + { + "epoch": 4.16, + "learning_rate": 4.542140649885542e-06, + "loss": 0.9112, + "step": 56291 + }, + { + "epoch": 4.16, + "learning_rate": 4.5418064298661325e-06, + "loss": 0.9675, + "step": 56292 + }, + { + "epoch": 4.16, + "learning_rate": 4.541472218530632e-06, + "loss": 1.0363, + "step": 56293 + }, + { + "epoch": 4.16, + "learning_rate": 4.54113801587957e-06, + "loss": 1.0139, + "step": 56294 + }, + { + "epoch": 4.16, + "learning_rate": 4.540803821913475e-06, + "loss": 0.9239, + "step": 56295 + }, + { + "epoch": 4.16, + "learning_rate": 4.540469636632888e-06, + "loss": 0.8969, + "step": 56296 + }, + { + "epoch": 4.16, + "learning_rate": 4.540135460038336e-06, + "loss": 0.9029, + "step": 56297 + }, + { + "epoch": 4.16, + "learning_rate": 4.539801292130349e-06, + "loss": 1.0802, + "step": 56298 + }, + { + "epoch": 4.16, + "learning_rate": 4.539467132909461e-06, + "loss": 0.8775, + "step": 56299 + }, + { + "epoch": 4.16, + "learning_rate": 4.5391329823761985e-06, + "loss": 0.9888, + "step": 56300 + }, + { + "epoch": 4.16, + "learning_rate": 4.538798840531101e-06, + "loss": 1.1083, + "step": 56301 + }, + { + "epoch": 4.16, + "learning_rate": 4.538464707374697e-06, + "loss": 0.9772, + "step": 56302 + }, + { + "epoch": 4.16, + "learning_rate": 4.538130582907517e-06, + "loss": 1.0022, + "step": 56303 + }, + { + "epoch": 4.16, + "learning_rate": 4.53779646713009e-06, + "loss": 1.0138, + "step": 56304 + }, + { + "epoch": 4.16, + "learning_rate": 4.537462360042954e-06, + "loss": 1.0296, + "step": 56305 + }, + { + "epoch": 4.16, + "learning_rate": 4.53712826164664e-06, + "loss": 0.9792, + "step": 56306 + }, + { + "epoch": 4.16, + "learning_rate": 4.536794171941671e-06, + "loss": 1.0791, + "step": 56307 + }, + { + "epoch": 4.16, + "learning_rate": 4.536460090928587e-06, + "loss": 0.944, + "step": 56308 + }, + { + "epoch": 4.16, + "learning_rate": 4.536126018607913e-06, + "loss": 0.9629, + "step": 56309 + }, + { + "epoch": 4.16, + "learning_rate": 4.535791954980192e-06, + "loss": 0.9584, + "step": 56310 + }, + { + "epoch": 4.16, + "learning_rate": 4.535457900045941e-06, + "loss": 1.0045, + "step": 56311 + }, + { + "epoch": 4.16, + "learning_rate": 4.535123853805701e-06, + "loss": 1.108, + "step": 56312 + }, + { + "epoch": 4.16, + "learning_rate": 4.534789816259998e-06, + "loss": 0.984, + "step": 56313 + }, + { + "epoch": 4.16, + "learning_rate": 4.534455787409373e-06, + "loss": 1.0237, + "step": 56314 + }, + { + "epoch": 4.16, + "learning_rate": 4.534121767254347e-06, + "loss": 0.9963, + "step": 56315 + }, + { + "epoch": 4.16, + "learning_rate": 4.5337877557954515e-06, + "loss": 0.8896, + "step": 56316 + }, + { + "epoch": 4.16, + "learning_rate": 4.533453753033224e-06, + "loss": 0.9315, + "step": 56317 + }, + { + "epoch": 4.16, + "learning_rate": 4.533119758968195e-06, + "loss": 1.0634, + "step": 56318 + }, + { + "epoch": 4.16, + "learning_rate": 4.532785773600894e-06, + "loss": 1.0216, + "step": 56319 + }, + { + "epoch": 4.16, + "learning_rate": 4.532451796931848e-06, + "loss": 0.966, + "step": 56320 + }, + { + "epoch": 4.16, + "learning_rate": 4.532117828961598e-06, + "loss": 1.0771, + "step": 56321 + }, + { + "epoch": 4.16, + "learning_rate": 4.531783869690669e-06, + "loss": 1.0427, + "step": 56322 + }, + { + "epoch": 4.16, + "learning_rate": 4.531449919119595e-06, + "loss": 1.0829, + "step": 56323 + }, + { + "epoch": 4.16, + "learning_rate": 4.531115977248901e-06, + "loss": 0.9839, + "step": 56324 + }, + { + "epoch": 4.16, + "learning_rate": 4.530782044079128e-06, + "loss": 1.1849, + "step": 56325 + }, + { + "epoch": 4.16, + "learning_rate": 4.530448119610801e-06, + "loss": 1.0417, + "step": 56326 + }, + { + "epoch": 4.16, + "learning_rate": 4.530114203844454e-06, + "loss": 0.9983, + "step": 56327 + }, + { + "epoch": 4.16, + "learning_rate": 4.529780296780617e-06, + "loss": 0.9549, + "step": 56328 + }, + { + "epoch": 4.16, + "learning_rate": 4.529446398419817e-06, + "loss": 1.0464, + "step": 56329 + }, + { + "epoch": 4.16, + "learning_rate": 4.529112508762598e-06, + "loss": 1.0736, + "step": 56330 + }, + { + "epoch": 4.16, + "learning_rate": 4.528778627809474e-06, + "loss": 1.0562, + "step": 56331 + }, + { + "epoch": 4.16, + "learning_rate": 4.528444755560989e-06, + "loss": 1.1201, + "step": 56332 + }, + { + "epoch": 4.16, + "learning_rate": 4.528110892017667e-06, + "loss": 1.0829, + "step": 56333 + }, + { + "epoch": 4.16, + "learning_rate": 4.527777037180049e-06, + "loss": 0.9852, + "step": 56334 + }, + { + "epoch": 4.16, + "learning_rate": 4.527443191048656e-06, + "loss": 0.989, + "step": 56335 + }, + { + "epoch": 4.16, + "learning_rate": 4.527109353624018e-06, + "loss": 1.1204, + "step": 56336 + }, + { + "epoch": 4.16, + "learning_rate": 4.526775524906676e-06, + "loss": 1.0354, + "step": 56337 + }, + { + "epoch": 4.16, + "learning_rate": 4.526441704897155e-06, + "loss": 0.9643, + "step": 56338 + }, + { + "epoch": 4.16, + "learning_rate": 4.526107893595987e-06, + "loss": 1.0036, + "step": 56339 + }, + { + "epoch": 4.16, + "learning_rate": 4.5257740910037e-06, + "loss": 1.0223, + "step": 56340 + }, + { + "epoch": 4.16, + "learning_rate": 4.52544029712083e-06, + "loss": 1.1148, + "step": 56341 + }, + { + "epoch": 4.16, + "learning_rate": 4.5251065119479075e-06, + "loss": 1.0471, + "step": 56342 + }, + { + "epoch": 4.16, + "learning_rate": 4.524772735485462e-06, + "loss": 1.1078, + "step": 56343 + }, + { + "epoch": 4.16, + "learning_rate": 4.524438967734025e-06, + "loss": 0.8979, + "step": 56344 + }, + { + "epoch": 4.16, + "learning_rate": 4.524105208694123e-06, + "loss": 0.9837, + "step": 56345 + }, + { + "epoch": 4.16, + "learning_rate": 4.523771458366296e-06, + "loss": 1.0986, + "step": 56346 + }, + { + "epoch": 4.16, + "learning_rate": 4.523437716751069e-06, + "loss": 0.9437, + "step": 56347 + }, + { + "epoch": 4.16, + "learning_rate": 4.523103983848976e-06, + "loss": 0.9698, + "step": 56348 + }, + { + "epoch": 4.16, + "learning_rate": 4.522770259660541e-06, + "loss": 0.9616, + "step": 56349 + }, + { + "epoch": 4.16, + "learning_rate": 4.522436544186304e-06, + "loss": 1.0062, + "step": 56350 + }, + { + "epoch": 4.16, + "learning_rate": 4.522102837426791e-06, + "loss": 0.8811, + "step": 56351 + }, + { + "epoch": 4.16, + "learning_rate": 4.521769139382536e-06, + "loss": 1.0435, + "step": 56352 + }, + { + "epoch": 4.16, + "learning_rate": 4.521435450054067e-06, + "loss": 0.9934, + "step": 56353 + }, + { + "epoch": 4.16, + "learning_rate": 4.521101769441912e-06, + "loss": 0.9728, + "step": 56354 + }, + { + "epoch": 4.16, + "learning_rate": 4.520768097546614e-06, + "loss": 0.8866, + "step": 56355 + }, + { + "epoch": 4.16, + "learning_rate": 4.5204344343686865e-06, + "loss": 1.0257, + "step": 56356 + }, + { + "epoch": 4.16, + "learning_rate": 4.520100779908675e-06, + "loss": 1.0371, + "step": 56357 + }, + { + "epoch": 4.16, + "learning_rate": 4.5197671341671e-06, + "loss": 0.9607, + "step": 56358 + }, + { + "epoch": 4.16, + "learning_rate": 4.519433497144506e-06, + "loss": 0.9632, + "step": 56359 + }, + { + "epoch": 4.16, + "learning_rate": 4.519099868841405e-06, + "loss": 0.9294, + "step": 56360 + }, + { + "epoch": 4.16, + "learning_rate": 4.518766249258343e-06, + "loss": 0.9203, + "step": 56361 + }, + { + "epoch": 4.16, + "learning_rate": 4.518432638395845e-06, + "loss": 0.9116, + "step": 56362 + }, + { + "epoch": 4.16, + "learning_rate": 4.518099036254442e-06, + "loss": 1.0314, + "step": 56363 + }, + { + "epoch": 4.16, + "learning_rate": 4.517765442834665e-06, + "loss": 0.9942, + "step": 56364 + }, + { + "epoch": 4.16, + "learning_rate": 4.517431858137041e-06, + "loss": 0.9629, + "step": 56365 + }, + { + "epoch": 4.16, + "learning_rate": 4.517098282162109e-06, + "loss": 1.0345, + "step": 56366 + }, + { + "epoch": 4.16, + "learning_rate": 4.516764714910395e-06, + "loss": 0.998, + "step": 56367 + }, + { + "epoch": 4.16, + "learning_rate": 4.516431156382429e-06, + "loss": 1.1073, + "step": 56368 + }, + { + "epoch": 4.17, + "learning_rate": 4.51609760657874e-06, + "loss": 0.9868, + "step": 56369 + }, + { + "epoch": 4.17, + "learning_rate": 4.515764065499863e-06, + "loss": 1.0131, + "step": 56370 + }, + { + "epoch": 4.17, + "learning_rate": 4.515430533146329e-06, + "loss": 1.0576, + "step": 56371 + }, + { + "epoch": 4.17, + "learning_rate": 4.515097009518666e-06, + "loss": 0.9363, + "step": 56372 + }, + { + "epoch": 4.17, + "learning_rate": 4.514763494617404e-06, + "loss": 1.0683, + "step": 56373 + }, + { + "epoch": 4.17, + "learning_rate": 4.514429988443073e-06, + "loss": 1.0682, + "step": 56374 + }, + { + "epoch": 4.17, + "learning_rate": 4.514096490996213e-06, + "loss": 0.9905, + "step": 56375 + }, + { + "epoch": 4.17, + "learning_rate": 4.513763002277338e-06, + "loss": 0.9941, + "step": 56376 + }, + { + "epoch": 4.17, + "learning_rate": 4.513429522286992e-06, + "loss": 1.005, + "step": 56377 + }, + { + "epoch": 4.17, + "learning_rate": 4.5130960510256975e-06, + "loss": 0.9969, + "step": 56378 + }, + { + "epoch": 4.17, + "learning_rate": 4.512762588493995e-06, + "loss": 1.0771, + "step": 56379 + }, + { + "epoch": 4.17, + "learning_rate": 4.512429134692402e-06, + "loss": 1.0336, + "step": 56380 + }, + { + "epoch": 4.17, + "learning_rate": 4.51209568962146e-06, + "loss": 1.0581, + "step": 56381 + }, + { + "epoch": 4.17, + "learning_rate": 4.511762253281694e-06, + "loss": 0.8935, + "step": 56382 + }, + { + "epoch": 4.17, + "learning_rate": 4.511428825673637e-06, + "loss": 0.9691, + "step": 56383 + }, + { + "epoch": 4.17, + "learning_rate": 4.511095406797817e-06, + "loss": 1.0893, + "step": 56384 + }, + { + "epoch": 4.17, + "learning_rate": 4.510761996654763e-06, + "loss": 0.9007, + "step": 56385 + }, + { + "epoch": 4.17, + "learning_rate": 4.5104285952450124e-06, + "loss": 0.9881, + "step": 56386 + }, + { + "epoch": 4.17, + "learning_rate": 4.5100952025690905e-06, + "loss": 0.9636, + "step": 56387 + }, + { + "epoch": 4.17, + "learning_rate": 4.509761818627528e-06, + "loss": 1.0319, + "step": 56388 + }, + { + "epoch": 4.17, + "learning_rate": 4.509428443420852e-06, + "loss": 0.9927, + "step": 56389 + }, + { + "epoch": 4.17, + "learning_rate": 4.509095076949602e-06, + "loss": 0.9348, + "step": 56390 + }, + { + "epoch": 4.17, + "learning_rate": 4.508761719214303e-06, + "loss": 0.9946, + "step": 56391 + }, + { + "epoch": 4.17, + "learning_rate": 4.5084283702154844e-06, + "loss": 1.0653, + "step": 56392 + }, + { + "epoch": 4.17, + "learning_rate": 4.508095029953678e-06, + "loss": 1.0852, + "step": 56393 + }, + { + "epoch": 4.17, + "learning_rate": 4.50776169842941e-06, + "loss": 0.9652, + "step": 56394 + }, + { + "epoch": 4.17, + "learning_rate": 4.507428375643219e-06, + "loss": 0.9919, + "step": 56395 + }, + { + "epoch": 4.17, + "learning_rate": 4.507095061595631e-06, + "loss": 1.0857, + "step": 56396 + }, + { + "epoch": 4.17, + "learning_rate": 4.506761756287176e-06, + "loss": 1.0779, + "step": 56397 + }, + { + "epoch": 4.17, + "learning_rate": 4.506428459718379e-06, + "loss": 0.9726, + "step": 56398 + }, + { + "epoch": 4.17, + "learning_rate": 4.506095171889781e-06, + "loss": 0.9804, + "step": 56399 + }, + { + "epoch": 4.17, + "learning_rate": 4.505761892801911e-06, + "loss": 0.979, + "step": 56400 + }, + { + "epoch": 4.17, + "learning_rate": 4.505428622455288e-06, + "loss": 0.9642, + "step": 56401 + }, + { + "epoch": 4.17, + "learning_rate": 4.505095360850452e-06, + "loss": 1.0082, + "step": 56402 + }, + { + "epoch": 4.17, + "learning_rate": 4.504762107987928e-06, + "loss": 1.0454, + "step": 56403 + }, + { + "epoch": 4.17, + "learning_rate": 4.504428863868255e-06, + "loss": 0.8849, + "step": 56404 + }, + { + "epoch": 4.17, + "learning_rate": 4.50409562849195e-06, + "loss": 0.9987, + "step": 56405 + }, + { + "epoch": 4.17, + "learning_rate": 4.503762401859555e-06, + "loss": 1.0548, + "step": 56406 + }, + { + "epoch": 4.17, + "learning_rate": 4.503429183971595e-06, + "loss": 1.0915, + "step": 56407 + }, + { + "epoch": 4.17, + "learning_rate": 4.5030959748286e-06, + "loss": 0.9944, + "step": 56408 + }, + { + "epoch": 4.17, + "learning_rate": 4.502762774431101e-06, + "loss": 0.9312, + "step": 56409 + }, + { + "epoch": 4.17, + "learning_rate": 4.502429582779624e-06, + "loss": 0.9783, + "step": 56410 + }, + { + "epoch": 4.17, + "learning_rate": 4.502096399874706e-06, + "loss": 0.969, + "step": 56411 + }, + { + "epoch": 4.17, + "learning_rate": 4.501763225716874e-06, + "loss": 0.9421, + "step": 56412 + }, + { + "epoch": 4.17, + "learning_rate": 4.501430060306658e-06, + "loss": 1.0061, + "step": 56413 + }, + { + "epoch": 4.17, + "learning_rate": 4.5010969036445864e-06, + "loss": 0.9572, + "step": 56414 + }, + { + "epoch": 4.17, + "learning_rate": 4.500763755731192e-06, + "loss": 0.9986, + "step": 56415 + }, + { + "epoch": 4.17, + "learning_rate": 4.500430616567005e-06, + "loss": 1.0277, + "step": 56416 + }, + { + "epoch": 4.17, + "learning_rate": 4.500097486152555e-06, + "loss": 1.0896, + "step": 56417 + }, + { + "epoch": 4.17, + "learning_rate": 4.49976436448837e-06, + "loss": 0.9898, + "step": 56418 + }, + { + "epoch": 4.17, + "learning_rate": 4.4994312515749774e-06, + "loss": 0.9175, + "step": 56419 + }, + { + "epoch": 4.17, + "learning_rate": 4.499098147412919e-06, + "loss": 1.0418, + "step": 56420 + }, + { + "epoch": 4.17, + "learning_rate": 4.4987650520027095e-06, + "loss": 0.8772, + "step": 56421 + }, + { + "epoch": 4.17, + "learning_rate": 4.49843196534489e-06, + "loss": 1.0298, + "step": 56422 + }, + { + "epoch": 4.17, + "learning_rate": 4.498098887439981e-06, + "loss": 0.987, + "step": 56423 + }, + { + "epoch": 4.17, + "learning_rate": 4.497765818288527e-06, + "loss": 0.9253, + "step": 56424 + }, + { + "epoch": 4.17, + "learning_rate": 4.497432757891042e-06, + "loss": 1.0511, + "step": 56425 + }, + { + "epoch": 4.17, + "learning_rate": 4.497099706248065e-06, + "loss": 1.0025, + "step": 56426 + }, + { + "epoch": 4.17, + "learning_rate": 4.49676666336012e-06, + "loss": 0.9751, + "step": 56427 + }, + { + "epoch": 4.17, + "learning_rate": 4.496433629227749e-06, + "loss": 1.0073, + "step": 56428 + }, + { + "epoch": 4.17, + "learning_rate": 4.496100603851469e-06, + "loss": 1.0019, + "step": 56429 + }, + { + "epoch": 4.17, + "learning_rate": 4.4957675872318105e-06, + "loss": 1.0559, + "step": 56430 + }, + { + "epoch": 4.17, + "learning_rate": 4.495434579369311e-06, + "loss": 1.0204, + "step": 56431 + }, + { + "epoch": 4.17, + "learning_rate": 4.495101580264496e-06, + "loss": 0.9623, + "step": 56432 + }, + { + "epoch": 4.17, + "learning_rate": 4.494768589917895e-06, + "loss": 0.8836, + "step": 56433 + }, + { + "epoch": 4.17, + "learning_rate": 4.4944356083300355e-06, + "loss": 1.0853, + "step": 56434 + }, + { + "epoch": 4.17, + "learning_rate": 4.494102635501454e-06, + "loss": 0.9652, + "step": 56435 + }, + { + "epoch": 4.17, + "learning_rate": 4.493769671432676e-06, + "loss": 0.8798, + "step": 56436 + }, + { + "epoch": 4.17, + "learning_rate": 4.4934367161242324e-06, + "loss": 0.9901, + "step": 56437 + }, + { + "epoch": 4.17, + "learning_rate": 4.493103769576651e-06, + "loss": 0.9589, + "step": 56438 + }, + { + "epoch": 4.17, + "learning_rate": 4.49277083179046e-06, + "loss": 0.9827, + "step": 56439 + }, + { + "epoch": 4.17, + "learning_rate": 4.4924379027661955e-06, + "loss": 0.9987, + "step": 56440 + }, + { + "epoch": 4.17, + "learning_rate": 4.492104982504383e-06, + "loss": 1.0447, + "step": 56441 + }, + { + "epoch": 4.17, + "learning_rate": 4.491772071005552e-06, + "loss": 0.9408, + "step": 56442 + }, + { + "epoch": 4.17, + "learning_rate": 4.491439168270231e-06, + "loss": 0.9742, + "step": 56443 + }, + { + "epoch": 4.17, + "learning_rate": 4.491106274298954e-06, + "loss": 0.9312, + "step": 56444 + }, + { + "epoch": 4.17, + "learning_rate": 4.490773389092248e-06, + "loss": 0.9699, + "step": 56445 + }, + { + "epoch": 4.17, + "learning_rate": 4.490440512650643e-06, + "loss": 0.9783, + "step": 56446 + }, + { + "epoch": 4.17, + "learning_rate": 4.490107644974669e-06, + "loss": 1.0041, + "step": 56447 + }, + { + "epoch": 4.17, + "learning_rate": 4.48977478606485e-06, + "loss": 0.9763, + "step": 56448 + }, + { + "epoch": 4.17, + "learning_rate": 4.489441935921728e-06, + "loss": 1.0179, + "step": 56449 + }, + { + "epoch": 4.17, + "learning_rate": 4.489109094545818e-06, + "loss": 0.8627, + "step": 56450 + }, + { + "epoch": 4.17, + "learning_rate": 4.488776261937661e-06, + "loss": 1.063, + "step": 56451 + }, + { + "epoch": 4.17, + "learning_rate": 4.488443438097782e-06, + "loss": 0.969, + "step": 56452 + }, + { + "epoch": 4.17, + "learning_rate": 4.488110623026709e-06, + "loss": 0.92, + "step": 56453 + }, + { + "epoch": 4.17, + "learning_rate": 4.487777816724971e-06, + "loss": 1.0731, + "step": 56454 + }, + { + "epoch": 4.17, + "learning_rate": 4.487445019193102e-06, + "loss": 0.9885, + "step": 56455 + }, + { + "epoch": 4.17, + "learning_rate": 4.4871122304316305e-06, + "loss": 0.9811, + "step": 56456 + }, + { + "epoch": 4.17, + "learning_rate": 4.486779450441084e-06, + "loss": 1.0579, + "step": 56457 + }, + { + "epoch": 4.17, + "learning_rate": 4.486446679221992e-06, + "loss": 1.0626, + "step": 56458 + }, + { + "epoch": 4.17, + "learning_rate": 4.486113916774881e-06, + "loss": 0.942, + "step": 56459 + }, + { + "epoch": 4.17, + "learning_rate": 4.485781163100288e-06, + "loss": 1.0281, + "step": 56460 + }, + { + "epoch": 4.17, + "learning_rate": 4.485448418198738e-06, + "loss": 1.09, + "step": 56461 + }, + { + "epoch": 4.17, + "learning_rate": 4.485115682070761e-06, + "loss": 0.9705, + "step": 56462 + }, + { + "epoch": 4.17, + "learning_rate": 4.484782954716882e-06, + "loss": 0.9346, + "step": 56463 + }, + { + "epoch": 4.17, + "learning_rate": 4.484450236137638e-06, + "loss": 0.9995, + "step": 56464 + }, + { + "epoch": 4.17, + "learning_rate": 4.484117526333559e-06, + "loss": 1.0855, + "step": 56465 + }, + { + "epoch": 4.17, + "learning_rate": 4.483784825305162e-06, + "loss": 1.0732, + "step": 56466 + }, + { + "epoch": 4.17, + "learning_rate": 4.483452133052989e-06, + "loss": 0.8827, + "step": 56467 + }, + { + "epoch": 4.17, + "learning_rate": 4.48311944957756e-06, + "loss": 0.9366, + "step": 56468 + }, + { + "epoch": 4.17, + "learning_rate": 4.482786774879418e-06, + "loss": 1.0444, + "step": 56469 + }, + { + "epoch": 4.17, + "learning_rate": 4.482454108959073e-06, + "loss": 0.8707, + "step": 56470 + }, + { + "epoch": 4.17, + "learning_rate": 4.482121451817071e-06, + "loss": 0.8583, + "step": 56471 + }, + { + "epoch": 4.17, + "learning_rate": 4.48178880345393e-06, + "loss": 1.0482, + "step": 56472 + }, + { + "epoch": 4.17, + "learning_rate": 4.481456163870191e-06, + "loss": 1.0288, + "step": 56473 + }, + { + "epoch": 4.17, + "learning_rate": 4.481123533066374e-06, + "loss": 1.0085, + "step": 56474 + }, + { + "epoch": 4.17, + "learning_rate": 4.480790911043006e-06, + "loss": 1.0209, + "step": 56475 + }, + { + "epoch": 4.17, + "learning_rate": 4.480458297800625e-06, + "loss": 0.9436, + "step": 56476 + }, + { + "epoch": 4.17, + "learning_rate": 4.4801256933397555e-06, + "loss": 0.9791, + "step": 56477 + }, + { + "epoch": 4.17, + "learning_rate": 4.479793097660926e-06, + "loss": 0.9432, + "step": 56478 + }, + { + "epoch": 4.17, + "learning_rate": 4.479460510764663e-06, + "loss": 0.9729, + "step": 56479 + }, + { + "epoch": 4.17, + "learning_rate": 4.479127932651505e-06, + "loss": 0.9845, + "step": 56480 + }, + { + "epoch": 4.17, + "learning_rate": 4.478795363321975e-06, + "loss": 1.0386, + "step": 56481 + }, + { + "epoch": 4.17, + "learning_rate": 4.478462802776602e-06, + "loss": 0.8775, + "step": 56482 + }, + { + "epoch": 4.17, + "learning_rate": 4.478130251015911e-06, + "loss": 0.9834, + "step": 56483 + }, + { + "epoch": 4.17, + "learning_rate": 4.477797708040441e-06, + "loss": 0.9877, + "step": 56484 + }, + { + "epoch": 4.17, + "learning_rate": 4.477465173850715e-06, + "loss": 0.9566, + "step": 56485 + }, + { + "epoch": 4.17, + "learning_rate": 4.477132648447264e-06, + "loss": 0.9705, + "step": 56486 + }, + { + "epoch": 4.17, + "learning_rate": 4.476800131830614e-06, + "loss": 1.0101, + "step": 56487 + }, + { + "epoch": 4.17, + "learning_rate": 4.476467624001294e-06, + "loss": 0.9714, + "step": 56488 + }, + { + "epoch": 4.17, + "learning_rate": 4.4761351249598376e-06, + "loss": 1.0801, + "step": 56489 + }, + { + "epoch": 4.17, + "learning_rate": 4.4758026347067715e-06, + "loss": 0.9567, + "step": 56490 + }, + { + "epoch": 4.17, + "learning_rate": 4.475470153242625e-06, + "loss": 0.9815, + "step": 56491 + }, + { + "epoch": 4.17, + "learning_rate": 4.475137680567921e-06, + "loss": 0.8862, + "step": 56492 + }, + { + "epoch": 4.17, + "learning_rate": 4.474805216683199e-06, + "loss": 0.9584, + "step": 56493 + }, + { + "epoch": 4.17, + "learning_rate": 4.474472761588987e-06, + "loss": 0.9779, + "step": 56494 + }, + { + "epoch": 4.17, + "learning_rate": 4.474140315285802e-06, + "loss": 0.97, + "step": 56495 + }, + { + "epoch": 4.17, + "learning_rate": 4.473807877774183e-06, + "loss": 1.0007, + "step": 56496 + }, + { + "epoch": 4.17, + "learning_rate": 4.473475449054658e-06, + "loss": 0.9603, + "step": 56497 + }, + { + "epoch": 4.17, + "learning_rate": 4.473143029127754e-06, + "loss": 0.93, + "step": 56498 + }, + { + "epoch": 4.17, + "learning_rate": 4.472810617993998e-06, + "loss": 1.1012, + "step": 56499 + }, + { + "epoch": 4.17, + "learning_rate": 4.472478215653924e-06, + "loss": 0.9993, + "step": 56500 + }, + { + "epoch": 4.17, + "learning_rate": 4.472145822108057e-06, + "loss": 1.0391, + "step": 56501 + }, + { + "epoch": 4.17, + "learning_rate": 4.4718134373569274e-06, + "loss": 1.0256, + "step": 56502 + }, + { + "epoch": 4.17, + "learning_rate": 4.471481061401065e-06, + "loss": 1.0452, + "step": 56503 + }, + { + "epoch": 4.18, + "learning_rate": 4.471148694240992e-06, + "loss": 0.9966, + "step": 56504 + }, + { + "epoch": 4.18, + "learning_rate": 4.470816335877246e-06, + "loss": 0.9687, + "step": 56505 + }, + { + "epoch": 4.18, + "learning_rate": 4.470483986310354e-06, + "loss": 0.9782, + "step": 56506 + }, + { + "epoch": 4.18, + "learning_rate": 4.47015164554084e-06, + "loss": 1.0301, + "step": 56507 + }, + { + "epoch": 4.18, + "learning_rate": 4.469819313569233e-06, + "loss": 1.0023, + "step": 56508 + }, + { + "epoch": 4.18, + "learning_rate": 4.469486990396069e-06, + "loss": 1.0208, + "step": 56509 + }, + { + "epoch": 4.18, + "learning_rate": 4.469154676021871e-06, + "loss": 1.1203, + "step": 56510 + }, + { + "epoch": 4.18, + "learning_rate": 4.468822370447169e-06, + "loss": 1.0924, + "step": 56511 + }, + { + "epoch": 4.18, + "learning_rate": 4.468490073672492e-06, + "loss": 0.9947, + "step": 56512 + }, + { + "epoch": 4.18, + "learning_rate": 4.468157785698364e-06, + "loss": 0.9247, + "step": 56513 + }, + { + "epoch": 4.18, + "learning_rate": 4.467825506525325e-06, + "loss": 1.0753, + "step": 56514 + }, + { + "epoch": 4.18, + "learning_rate": 4.467493236153889e-06, + "loss": 0.9562, + "step": 56515 + }, + { + "epoch": 4.18, + "learning_rate": 4.467160974584597e-06, + "loss": 0.9925, + "step": 56516 + }, + { + "epoch": 4.18, + "learning_rate": 4.466828721817968e-06, + "loss": 1.0632, + "step": 56517 + }, + { + "epoch": 4.18, + "learning_rate": 4.466496477854543e-06, + "loss": 1.0656, + "step": 56518 + }, + { + "epoch": 4.18, + "learning_rate": 4.466164242694835e-06, + "loss": 1.0588, + "step": 56519 + }, + { + "epoch": 4.18, + "learning_rate": 4.4658320163393855e-06, + "loss": 0.9626, + "step": 56520 + }, + { + "epoch": 4.18, + "learning_rate": 4.4654997987887174e-06, + "loss": 0.9627, + "step": 56521 + }, + { + "epoch": 4.18, + "learning_rate": 4.465167590043359e-06, + "loss": 1.051, + "step": 56522 + }, + { + "epoch": 4.18, + "learning_rate": 4.464835390103841e-06, + "loss": 0.9679, + "step": 56523 + }, + { + "epoch": 4.18, + "learning_rate": 4.464503198970686e-06, + "loss": 0.8416, + "step": 56524 + }, + { + "epoch": 4.18, + "learning_rate": 4.464171016644432e-06, + "loss": 1.0865, + "step": 56525 + }, + { + "epoch": 4.18, + "learning_rate": 4.463838843125602e-06, + "loss": 0.9946, + "step": 56526 + }, + { + "epoch": 4.18, + "learning_rate": 4.4635066784147255e-06, + "loss": 1.0087, + "step": 56527 + }, + { + "epoch": 4.18, + "learning_rate": 4.463174522512327e-06, + "loss": 0.9189, + "step": 56528 + }, + { + "epoch": 4.18, + "learning_rate": 4.4628423754189424e-06, + "loss": 0.9688, + "step": 56529 + }, + { + "epoch": 4.18, + "learning_rate": 4.462510237135096e-06, + "loss": 0.9643, + "step": 56530 + }, + { + "epoch": 4.18, + "learning_rate": 4.462178107661316e-06, + "loss": 0.955, + "step": 56531 + }, + { + "epoch": 4.18, + "learning_rate": 4.4618459869981326e-06, + "loss": 1.0257, + "step": 56532 + }, + { + "epoch": 4.18, + "learning_rate": 4.461513875146068e-06, + "loss": 0.9997, + "step": 56533 + }, + { + "epoch": 4.18, + "learning_rate": 4.461181772105661e-06, + "loss": 1.0351, + "step": 56534 + }, + { + "epoch": 4.18, + "learning_rate": 4.460849677877433e-06, + "loss": 0.9899, + "step": 56535 + }, + { + "epoch": 4.18, + "learning_rate": 4.460517592461915e-06, + "loss": 0.9665, + "step": 56536 + }, + { + "epoch": 4.18, + "learning_rate": 4.460185515859629e-06, + "loss": 0.9061, + "step": 56537 + }, + { + "epoch": 4.18, + "learning_rate": 4.459853448071118e-06, + "loss": 0.9692, + "step": 56538 + }, + { + "epoch": 4.18, + "learning_rate": 4.459521389096893e-06, + "loss": 1.0261, + "step": 56539 + }, + { + "epoch": 4.18, + "learning_rate": 4.459189338937493e-06, + "loss": 0.9548, + "step": 56540 + }, + { + "epoch": 4.18, + "learning_rate": 4.458857297593444e-06, + "loss": 1.0023, + "step": 56541 + }, + { + "epoch": 4.18, + "learning_rate": 4.4585252650652734e-06, + "loss": 1.0896, + "step": 56542 + }, + { + "epoch": 4.18, + "learning_rate": 4.45819324135351e-06, + "loss": 0.9447, + "step": 56543 + }, + { + "epoch": 4.18, + "learning_rate": 4.457861226458678e-06, + "loss": 0.9192, + "step": 56544 + }, + { + "epoch": 4.18, + "learning_rate": 4.457529220381314e-06, + "loss": 1.0033, + "step": 56545 + }, + { + "epoch": 4.18, + "learning_rate": 4.457197223121941e-06, + "loss": 1.124, + "step": 56546 + }, + { + "epoch": 4.18, + "learning_rate": 4.456865234681088e-06, + "loss": 1.0614, + "step": 56547 + }, + { + "epoch": 4.18, + "learning_rate": 4.45653325505928e-06, + "loss": 0.9936, + "step": 56548 + }, + { + "epoch": 4.18, + "learning_rate": 4.456201284257051e-06, + "loss": 1.081, + "step": 56549 + }, + { + "epoch": 4.18, + "learning_rate": 4.455869322274927e-06, + "loss": 1.0129, + "step": 56550 + }, + { + "epoch": 4.18, + "learning_rate": 4.455537369113435e-06, + "loss": 1.0067, + "step": 56551 + }, + { + "epoch": 4.18, + "learning_rate": 4.455205424773105e-06, + "loss": 0.9395, + "step": 56552 + }, + { + "epoch": 4.18, + "learning_rate": 4.454873489254459e-06, + "loss": 1.0203, + "step": 56553 + }, + { + "epoch": 4.18, + "learning_rate": 4.454541562558035e-06, + "loss": 0.967, + "step": 56554 + }, + { + "epoch": 4.18, + "learning_rate": 4.454209644684354e-06, + "loss": 1.106, + "step": 56555 + }, + { + "epoch": 4.18, + "learning_rate": 4.453877735633947e-06, + "loss": 1.0201, + "step": 56556 + }, + { + "epoch": 4.18, + "learning_rate": 4.453545835407338e-06, + "loss": 1.0893, + "step": 56557 + }, + { + "epoch": 4.18, + "learning_rate": 4.453213944005062e-06, + "loss": 0.971, + "step": 56558 + }, + { + "epoch": 4.18, + "learning_rate": 4.452882061427646e-06, + "loss": 1.0134, + "step": 56559 + }, + { + "epoch": 4.18, + "learning_rate": 4.452550187675608e-06, + "loss": 1.0353, + "step": 56560 + }, + { + "epoch": 4.18, + "learning_rate": 4.4522183227494875e-06, + "loss": 0.9305, + "step": 56561 + }, + { + "epoch": 4.18, + "learning_rate": 4.451886466649804e-06, + "loss": 1.1683, + "step": 56562 + }, + { + "epoch": 4.18, + "learning_rate": 4.451554619377098e-06, + "loss": 1.0252, + "step": 56563 + }, + { + "epoch": 4.18, + "learning_rate": 4.451222780931881e-06, + "loss": 1.0413, + "step": 56564 + }, + { + "epoch": 4.18, + "learning_rate": 4.450890951314694e-06, + "loss": 1.0165, + "step": 56565 + }, + { + "epoch": 4.18, + "learning_rate": 4.450559130526056e-06, + "loss": 1.0437, + "step": 56566 + }, + { + "epoch": 4.18, + "learning_rate": 4.450227318566505e-06, + "loss": 0.8574, + "step": 56567 + }, + { + "epoch": 4.18, + "learning_rate": 4.449895515436561e-06, + "loss": 1.0295, + "step": 56568 + }, + { + "epoch": 4.18, + "learning_rate": 4.449563721136749e-06, + "loss": 0.971, + "step": 56569 + }, + { + "epoch": 4.18, + "learning_rate": 4.449231935667606e-06, + "loss": 0.9526, + "step": 56570 + }, + { + "epoch": 4.18, + "learning_rate": 4.448900159029656e-06, + "loss": 0.9752, + "step": 56571 + }, + { + "epoch": 4.18, + "learning_rate": 4.448568391223425e-06, + "loss": 1.0121, + "step": 56572 + }, + { + "epoch": 4.18, + "learning_rate": 4.448236632249439e-06, + "loss": 0.998, + "step": 56573 + }, + { + "epoch": 4.18, + "learning_rate": 4.447904882108234e-06, + "loss": 1.0895, + "step": 56574 + }, + { + "epoch": 4.18, + "learning_rate": 4.447573140800332e-06, + "loss": 1.0434, + "step": 56575 + }, + { + "epoch": 4.18, + "learning_rate": 4.447241408326262e-06, + "loss": 0.921, + "step": 56576 + }, + { + "epoch": 4.18, + "learning_rate": 4.446909684686551e-06, + "loss": 1.0852, + "step": 56577 + }, + { + "epoch": 4.18, + "learning_rate": 4.446577969881723e-06, + "loss": 1.002, + "step": 56578 + }, + { + "epoch": 4.18, + "learning_rate": 4.446246263912315e-06, + "loss": 1.1155, + "step": 56579 + }, + { + "epoch": 4.18, + "learning_rate": 4.445914566778849e-06, + "loss": 1.0402, + "step": 56580 + }, + { + "epoch": 4.18, + "learning_rate": 4.4455828784818535e-06, + "loss": 1.0282, + "step": 56581 + }, + { + "epoch": 4.18, + "learning_rate": 4.445251199021853e-06, + "loss": 1.0086, + "step": 56582 + }, + { + "epoch": 4.18, + "learning_rate": 4.4449195283993855e-06, + "loss": 1.0182, + "step": 56583 + }, + { + "epoch": 4.18, + "learning_rate": 4.444587866614963e-06, + "loss": 1.0026, + "step": 56584 + }, + { + "epoch": 4.18, + "learning_rate": 4.444256213669127e-06, + "loss": 1.0902, + "step": 56585 + }, + { + "epoch": 4.18, + "learning_rate": 4.4439245695623956e-06, + "loss": 1.0126, + "step": 56586 + }, + { + "epoch": 4.18, + "learning_rate": 4.443592934295309e-06, + "loss": 0.9166, + "step": 56587 + }, + { + "epoch": 4.18, + "learning_rate": 4.443261307868381e-06, + "loss": 0.9881, + "step": 56588 + }, + { + "epoch": 4.18, + "learning_rate": 4.442929690282143e-06, + "loss": 0.9978, + "step": 56589 + }, + { + "epoch": 4.18, + "learning_rate": 4.442598081537126e-06, + "loss": 0.9261, + "step": 56590 + }, + { + "epoch": 4.18, + "learning_rate": 4.442266481633858e-06, + "loss": 1.009, + "step": 56591 + }, + { + "epoch": 4.18, + "learning_rate": 4.4419348905728635e-06, + "loss": 1.0345, + "step": 56592 + }, + { + "epoch": 4.18, + "learning_rate": 4.441603308354668e-06, + "loss": 1.0321, + "step": 56593 + }, + { + "epoch": 4.18, + "learning_rate": 4.441271734979804e-06, + "loss": 0.9861, + "step": 56594 + }, + { + "epoch": 4.18, + "learning_rate": 4.440940170448799e-06, + "loss": 1.0856, + "step": 56595 + }, + { + "epoch": 4.18, + "learning_rate": 4.440608614762178e-06, + "loss": 1.0059, + "step": 56596 + }, + { + "epoch": 4.18, + "learning_rate": 4.440277067920469e-06, + "loss": 1.0853, + "step": 56597 + }, + { + "epoch": 4.18, + "learning_rate": 4.439945529924196e-06, + "loss": 0.9649, + "step": 56598 + }, + { + "epoch": 4.18, + "learning_rate": 4.439614000773894e-06, + "loss": 1.0444, + "step": 56599 + }, + { + "epoch": 4.18, + "learning_rate": 4.439282480470086e-06, + "loss": 1.1665, + "step": 56600 + }, + { + "epoch": 4.18, + "learning_rate": 4.438950969013301e-06, + "loss": 0.9186, + "step": 56601 + }, + { + "epoch": 4.18, + "learning_rate": 4.43861946640406e-06, + "loss": 1.1077, + "step": 56602 + }, + { + "epoch": 4.18, + "learning_rate": 4.438287972642902e-06, + "loss": 0.9765, + "step": 56603 + }, + { + "epoch": 4.18, + "learning_rate": 4.437956487730347e-06, + "loss": 1.04, + "step": 56604 + }, + { + "epoch": 4.18, + "learning_rate": 4.437625011666924e-06, + "loss": 1.0148, + "step": 56605 + }, + { + "epoch": 4.18, + "learning_rate": 4.437293544453159e-06, + "loss": 1.017, + "step": 56606 + }, + { + "epoch": 4.18, + "learning_rate": 4.436962086089578e-06, + "loss": 0.9662, + "step": 56607 + }, + { + "epoch": 4.18, + "learning_rate": 4.436630636576717e-06, + "loss": 0.9754, + "step": 56608 + }, + { + "epoch": 4.18, + "learning_rate": 4.43629919591509e-06, + "loss": 0.917, + "step": 56609 + }, + { + "epoch": 4.18, + "learning_rate": 4.435967764105235e-06, + "loss": 1.0313, + "step": 56610 + }, + { + "epoch": 4.18, + "learning_rate": 4.435636341147673e-06, + "loss": 0.9239, + "step": 56611 + }, + { + "epoch": 4.18, + "learning_rate": 4.435304927042941e-06, + "loss": 0.9425, + "step": 56612 + }, + { + "epoch": 4.18, + "learning_rate": 4.434973521791551e-06, + "loss": 1.0088, + "step": 56613 + }, + { + "epoch": 4.18, + "learning_rate": 4.434642125394043e-06, + "loss": 0.9992, + "step": 56614 + }, + { + "epoch": 4.18, + "learning_rate": 4.434310737850939e-06, + "loss": 1.0334, + "step": 56615 + }, + { + "epoch": 4.18, + "learning_rate": 4.433979359162768e-06, + "loss": 0.9306, + "step": 56616 + }, + { + "epoch": 4.18, + "learning_rate": 4.433647989330055e-06, + "loss": 1.0394, + "step": 56617 + }, + { + "epoch": 4.18, + "learning_rate": 4.4333166283533255e-06, + "loss": 1.0767, + "step": 56618 + }, + { + "epoch": 4.18, + "learning_rate": 4.432985276233113e-06, + "loss": 0.9939, + "step": 56619 + }, + { + "epoch": 4.18, + "learning_rate": 4.432653932969942e-06, + "loss": 0.9639, + "step": 56620 + }, + { + "epoch": 4.18, + "learning_rate": 4.432322598564338e-06, + "loss": 1.0537, + "step": 56621 + }, + { + "epoch": 4.18, + "learning_rate": 4.4319912730168256e-06, + "loss": 1.0062, + "step": 56622 + }, + { + "epoch": 4.18, + "learning_rate": 4.431659956327939e-06, + "loss": 0.9913, + "step": 56623 + }, + { + "epoch": 4.18, + "learning_rate": 4.431328648498202e-06, + "loss": 1.0258, + "step": 56624 + }, + { + "epoch": 4.18, + "learning_rate": 4.430997349528141e-06, + "loss": 0.9023, + "step": 56625 + }, + { + "epoch": 4.18, + "learning_rate": 4.430666059418285e-06, + "loss": 1.0651, + "step": 56626 + }, + { + "epoch": 4.18, + "learning_rate": 4.4303347781691545e-06, + "loss": 1.0189, + "step": 56627 + }, + { + "epoch": 4.18, + "learning_rate": 4.43000350578129e-06, + "loss": 0.9545, + "step": 56628 + }, + { + "epoch": 4.18, + "learning_rate": 4.429672242255202e-06, + "loss": 0.9509, + "step": 56629 + }, + { + "epoch": 4.18, + "learning_rate": 4.4293409875914295e-06, + "loss": 1.0131, + "step": 56630 + }, + { + "epoch": 4.18, + "learning_rate": 4.429009741790492e-06, + "loss": 1.0486, + "step": 56631 + }, + { + "epoch": 4.18, + "learning_rate": 4.428678504852928e-06, + "loss": 1.0545, + "step": 56632 + }, + { + "epoch": 4.18, + "learning_rate": 4.428347276779254e-06, + "loss": 1.0158, + "step": 56633 + }, + { + "epoch": 4.18, + "learning_rate": 4.428016057569994e-06, + "loss": 0.9898, + "step": 56634 + }, + { + "epoch": 4.18, + "learning_rate": 4.427684847225686e-06, + "loss": 1.0041, + "step": 56635 + }, + { + "epoch": 4.18, + "learning_rate": 4.4273536457468514e-06, + "loss": 0.984, + "step": 56636 + }, + { + "epoch": 4.18, + "learning_rate": 4.427022453134017e-06, + "loss": 0.9802, + "step": 56637 + }, + { + "epoch": 4.18, + "learning_rate": 4.426691269387707e-06, + "loss": 1.0699, + "step": 56638 + }, + { + "epoch": 4.19, + "learning_rate": 4.426360094508455e-06, + "loss": 1.003, + "step": 56639 + }, + { + "epoch": 4.19, + "learning_rate": 4.4260289284967835e-06, + "loss": 0.9852, + "step": 56640 + }, + { + "epoch": 4.19, + "learning_rate": 4.4256977713532215e-06, + "loss": 1.158, + "step": 56641 + }, + { + "epoch": 4.19, + "learning_rate": 4.42536662307829e-06, + "loss": 1.0128, + "step": 56642 + }, + { + "epoch": 4.19, + "learning_rate": 4.425035483672525e-06, + "loss": 0.9289, + "step": 56643 + }, + { + "epoch": 4.19, + "learning_rate": 4.424704353136448e-06, + "loss": 0.9932, + "step": 56644 + }, + { + "epoch": 4.19, + "learning_rate": 4.4243732314705864e-06, + "loss": 0.9742, + "step": 56645 + }, + { + "epoch": 4.19, + "learning_rate": 4.424042118675468e-06, + "loss": 0.9914, + "step": 56646 + }, + { + "epoch": 4.19, + "learning_rate": 4.423711014751615e-06, + "loss": 1.0046, + "step": 56647 + }, + { + "epoch": 4.19, + "learning_rate": 4.4233799196995615e-06, + "loss": 1.0946, + "step": 56648 + }, + { + "epoch": 4.19, + "learning_rate": 4.42304883351983e-06, + "loss": 0.9887, + "step": 56649 + }, + { + "epoch": 4.19, + "learning_rate": 4.42271775621295e-06, + "loss": 1.022, + "step": 56650 + }, + { + "epoch": 4.19, + "learning_rate": 4.4223866877794404e-06, + "loss": 1.0065, + "step": 56651 + }, + { + "epoch": 4.19, + "learning_rate": 4.422055628219839e-06, + "loss": 0.9486, + "step": 56652 + }, + { + "epoch": 4.19, + "learning_rate": 4.421724577534671e-06, + "loss": 0.9472, + "step": 56653 + }, + { + "epoch": 4.19, + "learning_rate": 4.421393535724451e-06, + "loss": 0.9176, + "step": 56654 + }, + { + "epoch": 4.19, + "learning_rate": 4.421062502789718e-06, + "loss": 1.0366, + "step": 56655 + }, + { + "epoch": 4.19, + "learning_rate": 4.420731478730991e-06, + "loss": 1.0839, + "step": 56656 + }, + { + "epoch": 4.19, + "learning_rate": 4.420400463548808e-06, + "loss": 1.0272, + "step": 56657 + }, + { + "epoch": 4.19, + "learning_rate": 4.420069457243681e-06, + "loss": 0.9924, + "step": 56658 + }, + { + "epoch": 4.19, + "learning_rate": 4.419738459816146e-06, + "loss": 0.8912, + "step": 56659 + }, + { + "epoch": 4.19, + "learning_rate": 4.419407471266729e-06, + "loss": 0.9823, + "step": 56660 + }, + { + "epoch": 4.19, + "learning_rate": 4.419076491595954e-06, + "loss": 1.0211, + "step": 56661 + }, + { + "epoch": 4.19, + "learning_rate": 4.418745520804348e-06, + "loss": 0.86, + "step": 56662 + }, + { + "epoch": 4.19, + "learning_rate": 4.418414558892435e-06, + "loss": 1.1109, + "step": 56663 + }, + { + "epoch": 4.19, + "learning_rate": 4.418083605860748e-06, + "loss": 1.0738, + "step": 56664 + }, + { + "epoch": 4.19, + "learning_rate": 4.41775266170981e-06, + "loss": 0.9059, + "step": 56665 + }, + { + "epoch": 4.19, + "learning_rate": 4.4174217264401485e-06, + "loss": 1.1575, + "step": 56666 + }, + { + "epoch": 4.19, + "learning_rate": 4.417090800052285e-06, + "loss": 1.0019, + "step": 56667 + }, + { + "epoch": 4.19, + "learning_rate": 4.416759882546754e-06, + "loss": 1.0255, + "step": 56668 + }, + { + "epoch": 4.19, + "learning_rate": 4.416428973924077e-06, + "loss": 0.9733, + "step": 56669 + }, + { + "epoch": 4.19, + "learning_rate": 4.416098074184782e-06, + "loss": 0.9345, + "step": 56670 + }, + { + "epoch": 4.19, + "learning_rate": 4.415767183329396e-06, + "loss": 0.9998, + "step": 56671 + }, + { + "epoch": 4.19, + "learning_rate": 4.415436301358441e-06, + "loss": 1.0581, + "step": 56672 + }, + { + "epoch": 4.19, + "learning_rate": 4.415105428272453e-06, + "loss": 1.0715, + "step": 56673 + }, + { + "epoch": 4.19, + "learning_rate": 4.414774564071946e-06, + "loss": 0.9735, + "step": 56674 + }, + { + "epoch": 4.19, + "learning_rate": 4.414443708757457e-06, + "loss": 1.0093, + "step": 56675 + }, + { + "epoch": 4.19, + "learning_rate": 4.414112862329503e-06, + "loss": 0.9641, + "step": 56676 + }, + { + "epoch": 4.19, + "learning_rate": 4.413782024788623e-06, + "loss": 0.9986, + "step": 56677 + }, + { + "epoch": 4.19, + "learning_rate": 4.4134511961353285e-06, + "loss": 0.8678, + "step": 56678 + }, + { + "epoch": 4.19, + "learning_rate": 4.413120376370158e-06, + "loss": 1.0503, + "step": 56679 + }, + { + "epoch": 4.19, + "learning_rate": 4.412789565493633e-06, + "loss": 1.0332, + "step": 56680 + }, + { + "epoch": 4.19, + "learning_rate": 4.4124587635062785e-06, + "loss": 0.9596, + "step": 56681 + }, + { + "epoch": 4.19, + "learning_rate": 4.412127970408623e-06, + "loss": 1.0823, + "step": 56682 + }, + { + "epoch": 4.19, + "learning_rate": 4.4117971862011885e-06, + "loss": 0.9701, + "step": 56683 + }, + { + "epoch": 4.19, + "learning_rate": 4.411466410884508e-06, + "loss": 1.1046, + "step": 56684 + }, + { + "epoch": 4.19, + "learning_rate": 4.411135644459105e-06, + "loss": 1.001, + "step": 56685 + }, + { + "epoch": 4.19, + "learning_rate": 4.410804886925505e-06, + "loss": 1.0502, + "step": 56686 + }, + { + "epoch": 4.19, + "learning_rate": 4.41047413828423e-06, + "loss": 1.0764, + "step": 56687 + }, + { + "epoch": 4.19, + "learning_rate": 4.410143398535814e-06, + "loss": 1.065, + "step": 56688 + }, + { + "epoch": 4.19, + "learning_rate": 4.4098126676807816e-06, + "loss": 0.9979, + "step": 56689 + }, + { + "epoch": 4.19, + "learning_rate": 4.409481945719657e-06, + "loss": 0.9657, + "step": 56690 + }, + { + "epoch": 4.19, + "learning_rate": 4.409151232652966e-06, + "loss": 1.1282, + "step": 56691 + }, + { + "epoch": 4.19, + "learning_rate": 4.4088205284812316e-06, + "loss": 0.9995, + "step": 56692 + }, + { + "epoch": 4.19, + "learning_rate": 4.408489833204987e-06, + "loss": 0.9945, + "step": 56693 + }, + { + "epoch": 4.19, + "learning_rate": 4.408159146824756e-06, + "loss": 1.0838, + "step": 56694 + }, + { + "epoch": 4.19, + "learning_rate": 4.407828469341065e-06, + "loss": 0.9461, + "step": 56695 + }, + { + "epoch": 4.19, + "learning_rate": 4.407497800754432e-06, + "loss": 1.0449, + "step": 56696 + }, + { + "epoch": 4.19, + "learning_rate": 4.407167141065396e-06, + "loss": 1.0942, + "step": 56697 + }, + { + "epoch": 4.19, + "learning_rate": 4.406836490274477e-06, + "loss": 0.9841, + "step": 56698 + }, + { + "epoch": 4.19, + "learning_rate": 4.4065058483822e-06, + "loss": 0.9689, + "step": 56699 + }, + { + "epoch": 4.19, + "learning_rate": 4.406175215389092e-06, + "loss": 1.0646, + "step": 56700 + }, + { + "epoch": 4.19, + "learning_rate": 4.405844591295677e-06, + "loss": 1.0405, + "step": 56701 + }, + { + "epoch": 4.19, + "learning_rate": 4.4055139761024904e-06, + "loss": 1.0649, + "step": 56702 + }, + { + "epoch": 4.19, + "learning_rate": 4.405183369810043e-06, + "loss": 0.8498, + "step": 56703 + }, + { + "epoch": 4.19, + "learning_rate": 4.404852772418872e-06, + "loss": 1.0587, + "step": 56704 + }, + { + "epoch": 4.19, + "learning_rate": 4.404522183929501e-06, + "loss": 1.0055, + "step": 56705 + }, + { + "epoch": 4.19, + "learning_rate": 4.404191604342455e-06, + "loss": 0.9776, + "step": 56706 + }, + { + "epoch": 4.19, + "learning_rate": 4.403861033658256e-06, + "loss": 0.9456, + "step": 56707 + }, + { + "epoch": 4.19, + "learning_rate": 4.403530471877438e-06, + "loss": 1.1709, + "step": 56708 + }, + { + "epoch": 4.19, + "learning_rate": 4.403199919000523e-06, + "loss": 1.0379, + "step": 56709 + }, + { + "epoch": 4.19, + "learning_rate": 4.402869375028037e-06, + "loss": 1.0234, + "step": 56710 + }, + { + "epoch": 4.19, + "learning_rate": 4.402538839960506e-06, + "loss": 0.9687, + "step": 56711 + }, + { + "epoch": 4.19, + "learning_rate": 4.402208313798451e-06, + "loss": 0.9731, + "step": 56712 + }, + { + "epoch": 4.19, + "learning_rate": 4.401877796542407e-06, + "loss": 0.831, + "step": 56713 + }, + { + "epoch": 4.19, + "learning_rate": 4.401547288192895e-06, + "loss": 1.0205, + "step": 56714 + }, + { + "epoch": 4.19, + "learning_rate": 4.4012167887504405e-06, + "loss": 0.9536, + "step": 56715 + }, + { + "epoch": 4.19, + "learning_rate": 4.400886298215568e-06, + "loss": 0.9309, + "step": 56716 + }, + { + "epoch": 4.19, + "learning_rate": 4.400555816588807e-06, + "loss": 1.0257, + "step": 56717 + }, + { + "epoch": 4.19, + "learning_rate": 4.400225343870686e-06, + "loss": 0.9881, + "step": 56718 + }, + { + "epoch": 4.19, + "learning_rate": 4.39989488006172e-06, + "loss": 0.9619, + "step": 56719 + }, + { + "epoch": 4.19, + "learning_rate": 4.399564425162443e-06, + "loss": 1.0209, + "step": 56720 + }, + { + "epoch": 4.19, + "learning_rate": 4.399233979173377e-06, + "loss": 1.0603, + "step": 56721 + }, + { + "epoch": 4.19, + "learning_rate": 4.3989035420950564e-06, + "loss": 0.9543, + "step": 56722 + }, + { + "epoch": 4.19, + "learning_rate": 4.398573113927993e-06, + "loss": 0.8489, + "step": 56723 + }, + { + "epoch": 4.19, + "learning_rate": 4.398242694672723e-06, + "loss": 1.034, + "step": 56724 + }, + { + "epoch": 4.19, + "learning_rate": 4.397912284329764e-06, + "loss": 1.0303, + "step": 56725 + }, + { + "epoch": 4.19, + "learning_rate": 4.397581882899655e-06, + "loss": 1.0331, + "step": 56726 + }, + { + "epoch": 4.19, + "learning_rate": 4.397251490382909e-06, + "loss": 0.905, + "step": 56727 + }, + { + "epoch": 4.19, + "learning_rate": 4.3969211067800524e-06, + "loss": 1.0447, + "step": 56728 + }, + { + "epoch": 4.19, + "learning_rate": 4.396590732091617e-06, + "loss": 0.9259, + "step": 56729 + }, + { + "epoch": 4.19, + "learning_rate": 4.396260366318127e-06, + "loss": 0.956, + "step": 56730 + }, + { + "epoch": 4.19, + "learning_rate": 4.395930009460105e-06, + "loss": 0.9993, + "step": 56731 + }, + { + "epoch": 4.19, + "learning_rate": 4.395599661518075e-06, + "loss": 0.9251, + "step": 56732 + }, + { + "epoch": 4.19, + "learning_rate": 4.39526932249257e-06, + "loss": 0.9991, + "step": 56733 + }, + { + "epoch": 4.19, + "learning_rate": 4.394938992384111e-06, + "loss": 1.084, + "step": 56734 + }, + { + "epoch": 4.19, + "learning_rate": 4.3946086711932235e-06, + "loss": 1.0193, + "step": 56735 + }, + { + "epoch": 4.19, + "learning_rate": 4.394278358920434e-06, + "loss": 0.968, + "step": 56736 + }, + { + "epoch": 4.19, + "learning_rate": 4.393948055566263e-06, + "loss": 0.8987, + "step": 56737 + }, + { + "epoch": 4.19, + "learning_rate": 4.3936177611312445e-06, + "loss": 0.8165, + "step": 56738 + }, + { + "epoch": 4.19, + "learning_rate": 4.393287475615901e-06, + "loss": 1.0374, + "step": 56739 + }, + { + "epoch": 4.19, + "learning_rate": 4.392957199020755e-06, + "loss": 0.8997, + "step": 56740 + }, + { + "epoch": 4.19, + "learning_rate": 4.392626931346331e-06, + "loss": 1.0849, + "step": 56741 + }, + { + "epoch": 4.19, + "learning_rate": 4.392296672593161e-06, + "loss": 0.9967, + "step": 56742 + }, + { + "epoch": 4.19, + "learning_rate": 4.391966422761767e-06, + "loss": 1.0282, + "step": 56743 + }, + { + "epoch": 4.19, + "learning_rate": 4.3916361818526745e-06, + "loss": 0.9248, + "step": 56744 + }, + { + "epoch": 4.19, + "learning_rate": 4.391305949866408e-06, + "loss": 1.061, + "step": 56745 + }, + { + "epoch": 4.19, + "learning_rate": 4.390975726803489e-06, + "loss": 0.9123, + "step": 56746 + }, + { + "epoch": 4.19, + "learning_rate": 4.3906455126644566e-06, + "loss": 1.0903, + "step": 56747 + }, + { + "epoch": 4.19, + "learning_rate": 4.390315307449818e-06, + "loss": 0.9473, + "step": 56748 + }, + { + "epoch": 4.19, + "learning_rate": 4.389985111160112e-06, + "loss": 0.9077, + "step": 56749 + }, + { + "epoch": 4.19, + "learning_rate": 4.38965492379586e-06, + "loss": 1.0044, + "step": 56750 + }, + { + "epoch": 4.19, + "learning_rate": 4.389324745357585e-06, + "loss": 0.9583, + "step": 56751 + }, + { + "epoch": 4.19, + "learning_rate": 4.388994575845812e-06, + "loss": 1.062, + "step": 56752 + }, + { + "epoch": 4.19, + "learning_rate": 4.388664415261071e-06, + "loss": 1.0826, + "step": 56753 + }, + { + "epoch": 4.19, + "learning_rate": 4.388334263603885e-06, + "loss": 0.9641, + "step": 56754 + }, + { + "epoch": 4.19, + "learning_rate": 4.388004120874779e-06, + "loss": 1.0173, + "step": 56755 + }, + { + "epoch": 4.19, + "learning_rate": 4.387673987074278e-06, + "loss": 1.0167, + "step": 56756 + }, + { + "epoch": 4.19, + "learning_rate": 4.387343862202904e-06, + "loss": 1.0331, + "step": 56757 + }, + { + "epoch": 4.19, + "learning_rate": 4.387013746261189e-06, + "loss": 0.8565, + "step": 56758 + }, + { + "epoch": 4.19, + "learning_rate": 4.386683639249655e-06, + "loss": 0.8783, + "step": 56759 + }, + { + "epoch": 4.19, + "learning_rate": 4.386353541168826e-06, + "loss": 1.0332, + "step": 56760 + }, + { + "epoch": 4.19, + "learning_rate": 4.386023452019226e-06, + "loss": 1.0214, + "step": 56761 + }, + { + "epoch": 4.19, + "learning_rate": 4.385693371801385e-06, + "loss": 1.127, + "step": 56762 + }, + { + "epoch": 4.19, + "learning_rate": 4.385363300515826e-06, + "loss": 0.9571, + "step": 56763 + }, + { + "epoch": 4.19, + "learning_rate": 4.385033238163073e-06, + "loss": 1.0192, + "step": 56764 + }, + { + "epoch": 4.19, + "learning_rate": 4.384703184743653e-06, + "loss": 0.9705, + "step": 56765 + }, + { + "epoch": 4.19, + "learning_rate": 4.384373140258086e-06, + "loss": 0.9883, + "step": 56766 + }, + { + "epoch": 4.19, + "learning_rate": 4.384043104706908e-06, + "loss": 0.9081, + "step": 56767 + }, + { + "epoch": 4.19, + "learning_rate": 4.3837130780906286e-06, + "loss": 1.0069, + "step": 56768 + }, + { + "epoch": 4.19, + "learning_rate": 4.383383060409787e-06, + "loss": 1.0622, + "step": 56769 + }, + { + "epoch": 4.19, + "learning_rate": 4.383053051664898e-06, + "loss": 0.94, + "step": 56770 + }, + { + "epoch": 4.19, + "learning_rate": 4.382723051856498e-06, + "loss": 1.0445, + "step": 56771 + }, + { + "epoch": 4.19, + "learning_rate": 4.382393060985098e-06, + "loss": 1.0897, + "step": 56772 + }, + { + "epoch": 4.19, + "learning_rate": 4.382063079051236e-06, + "loss": 0.9974, + "step": 56773 + }, + { + "epoch": 4.19, + "learning_rate": 4.381733106055429e-06, + "loss": 1.1196, + "step": 56774 + }, + { + "epoch": 4.2, + "learning_rate": 4.381403141998205e-06, + "loss": 0.9717, + "step": 56775 + }, + { + "epoch": 4.2, + "learning_rate": 4.381073186880089e-06, + "loss": 0.9627, + "step": 56776 + }, + { + "epoch": 4.2, + "learning_rate": 4.3807432407015995e-06, + "loss": 1.0192, + "step": 56777 + }, + { + "epoch": 4.2, + "learning_rate": 4.380413303463272e-06, + "loss": 0.9481, + "step": 56778 + }, + { + "epoch": 4.2, + "learning_rate": 4.380083375165627e-06, + "loss": 1.0721, + "step": 56779 + }, + { + "epoch": 4.2, + "learning_rate": 4.379753455809188e-06, + "loss": 0.8638, + "step": 56780 + }, + { + "epoch": 4.2, + "learning_rate": 4.379423545394478e-06, + "loss": 0.9071, + "step": 56781 + }, + { + "epoch": 4.2, + "learning_rate": 4.379093643922028e-06, + "loss": 1.0434, + "step": 56782 + }, + { + "epoch": 4.2, + "learning_rate": 4.378763751392359e-06, + "loss": 1.0299, + "step": 56783 + }, + { + "epoch": 4.2, + "learning_rate": 4.378433867805997e-06, + "loss": 0.9523, + "step": 56784 + }, + { + "epoch": 4.2, + "learning_rate": 4.378103993163466e-06, + "loss": 0.9501, + "step": 56785 + }, + { + "epoch": 4.2, + "learning_rate": 4.377774127465288e-06, + "loss": 1.0233, + "step": 56786 + }, + { + "epoch": 4.2, + "learning_rate": 4.3774442707119935e-06, + "loss": 1.013, + "step": 56787 + }, + { + "epoch": 4.2, + "learning_rate": 4.377114422904106e-06, + "loss": 1.1454, + "step": 56788 + }, + { + "epoch": 4.2, + "learning_rate": 4.3767845840421485e-06, + "loss": 1.0233, + "step": 56789 + }, + { + "epoch": 4.2, + "learning_rate": 4.376454754126642e-06, + "loss": 0.911, + "step": 56790 + }, + { + "epoch": 4.2, + "learning_rate": 4.37612493315812e-06, + "loss": 0.974, + "step": 56791 + }, + { + "epoch": 4.2, + "learning_rate": 4.375795121137107e-06, + "loss": 0.9841, + "step": 56792 + }, + { + "epoch": 4.2, + "learning_rate": 4.375465318064115e-06, + "loss": 1.0488, + "step": 56793 + }, + { + "epoch": 4.2, + "learning_rate": 4.375135523939681e-06, + "loss": 1.0102, + "step": 56794 + }, + { + "epoch": 4.2, + "learning_rate": 4.374805738764326e-06, + "loss": 0.9454, + "step": 56795 + }, + { + "epoch": 4.2, + "learning_rate": 4.374475962538576e-06, + "loss": 1.0548, + "step": 56796 + }, + { + "epoch": 4.2, + "learning_rate": 4.374146195262948e-06, + "loss": 0.9555, + "step": 56797 + }, + { + "epoch": 4.2, + "learning_rate": 4.373816436937979e-06, + "loss": 1.0274, + "step": 56798 + }, + { + "epoch": 4.2, + "learning_rate": 4.373486687564186e-06, + "loss": 1.0344, + "step": 56799 + }, + { + "epoch": 4.2, + "learning_rate": 4.373156947142095e-06, + "loss": 0.8873, + "step": 56800 + }, + { + "epoch": 4.2, + "learning_rate": 4.372827215672228e-06, + "loss": 1.053, + "step": 56801 + }, + { + "epoch": 4.2, + "learning_rate": 4.3724974931551165e-06, + "loss": 0.9302, + "step": 56802 + }, + { + "epoch": 4.2, + "learning_rate": 4.37216777959128e-06, + "loss": 1.0459, + "step": 56803 + }, + { + "epoch": 4.2, + "learning_rate": 4.371838074981244e-06, + "loss": 1.058, + "step": 56804 + }, + { + "epoch": 4.2, + "learning_rate": 4.371508379325534e-06, + "loss": 0.9496, + "step": 56805 + }, + { + "epoch": 4.2, + "learning_rate": 4.371178692624669e-06, + "loss": 0.9533, + "step": 56806 + }, + { + "epoch": 4.2, + "learning_rate": 4.370849014879183e-06, + "loss": 0.8856, + "step": 56807 + }, + { + "epoch": 4.2, + "learning_rate": 4.3705193460895945e-06, + "loss": 1.0633, + "step": 56808 + }, + { + "epoch": 4.2, + "learning_rate": 4.37018968625643e-06, + "loss": 0.9973, + "step": 56809 + }, + { + "epoch": 4.2, + "learning_rate": 4.369860035380209e-06, + "loss": 0.9432, + "step": 56810 + }, + { + "epoch": 4.2, + "learning_rate": 4.369530393461464e-06, + "loss": 1.0849, + "step": 56811 + }, + { + "epoch": 4.2, + "learning_rate": 4.369200760500719e-06, + "loss": 0.9026, + "step": 56812 + }, + { + "epoch": 4.2, + "learning_rate": 4.368871136498487e-06, + "loss": 0.9757, + "step": 56813 + }, + { + "epoch": 4.2, + "learning_rate": 4.368541521455305e-06, + "loss": 1.0042, + "step": 56814 + }, + { + "epoch": 4.2, + "learning_rate": 4.368211915371689e-06, + "loss": 1.0312, + "step": 56815 + }, + { + "epoch": 4.2, + "learning_rate": 4.367882318248175e-06, + "loss": 1.0139, + "step": 56816 + }, + { + "epoch": 4.2, + "learning_rate": 4.367552730085273e-06, + "loss": 0.8915, + "step": 56817 + }, + { + "epoch": 4.2, + "learning_rate": 4.367223150883516e-06, + "loss": 1.0038, + "step": 56818 + }, + { + "epoch": 4.2, + "learning_rate": 4.3668935806434245e-06, + "loss": 0.9941, + "step": 56819 + }, + { + "epoch": 4.2, + "learning_rate": 4.36656401936553e-06, + "loss": 0.9985, + "step": 56820 + }, + { + "epoch": 4.2, + "learning_rate": 4.366234467050349e-06, + "loss": 0.9689, + "step": 56821 + }, + { + "epoch": 4.2, + "learning_rate": 4.365904923698405e-06, + "loss": 1.0002, + "step": 56822 + }, + { + "epoch": 4.2, + "learning_rate": 4.365575389310228e-06, + "loss": 0.9941, + "step": 56823 + }, + { + "epoch": 4.2, + "learning_rate": 4.365245863886341e-06, + "loss": 0.8537, + "step": 56824 + }, + { + "epoch": 4.2, + "learning_rate": 4.364916347427267e-06, + "loss": 0.9573, + "step": 56825 + }, + { + "epoch": 4.2, + "learning_rate": 4.364586839933528e-06, + "loss": 1.0286, + "step": 56826 + }, + { + "epoch": 4.2, + "learning_rate": 4.3642573414056526e-06, + "loss": 1.1398, + "step": 56827 + }, + { + "epoch": 4.2, + "learning_rate": 4.363927851844164e-06, + "loss": 1.1147, + "step": 56828 + }, + { + "epoch": 4.2, + "learning_rate": 4.363598371249585e-06, + "loss": 1.0217, + "step": 56829 + }, + { + "epoch": 4.2, + "learning_rate": 4.363268899622441e-06, + "loss": 0.9591, + "step": 56830 + }, + { + "epoch": 4.2, + "learning_rate": 4.3629394369632515e-06, + "loss": 0.9227, + "step": 56831 + }, + { + "epoch": 4.2, + "learning_rate": 4.362609983272548e-06, + "loss": 1.0402, + "step": 56832 + }, + { + "epoch": 4.2, + "learning_rate": 4.362280538550852e-06, + "loss": 0.9679, + "step": 56833 + }, + { + "epoch": 4.2, + "learning_rate": 4.3619511027986875e-06, + "loss": 0.9303, + "step": 56834 + }, + { + "epoch": 4.2, + "learning_rate": 4.361621676016573e-06, + "loss": 1.0499, + "step": 56835 + }, + { + "epoch": 4.2, + "learning_rate": 4.361292258205042e-06, + "loss": 0.9144, + "step": 56836 + }, + { + "epoch": 4.2, + "learning_rate": 4.360962849364615e-06, + "loss": 0.972, + "step": 56837 + }, + { + "epoch": 4.2, + "learning_rate": 4.360633449495815e-06, + "loss": 1.0223, + "step": 56838 + }, + { + "epoch": 4.2, + "learning_rate": 4.360304058599166e-06, + "loss": 0.973, + "step": 56839 + }, + { + "epoch": 4.2, + "learning_rate": 4.359974676675193e-06, + "loss": 0.9757, + "step": 56840 + }, + { + "epoch": 4.2, + "learning_rate": 4.359645303724419e-06, + "loss": 1.1397, + "step": 56841 + }, + { + "epoch": 4.2, + "learning_rate": 4.359315939747366e-06, + "loss": 0.9289, + "step": 56842 + }, + { + "epoch": 4.2, + "learning_rate": 4.358986584744564e-06, + "loss": 1.0111, + "step": 56843 + }, + { + "epoch": 4.2, + "learning_rate": 4.358657238716533e-06, + "loss": 1.1065, + "step": 56844 + }, + { + "epoch": 4.2, + "learning_rate": 4.358327901663798e-06, + "loss": 0.9422, + "step": 56845 + }, + { + "epoch": 4.2, + "learning_rate": 4.357998573586879e-06, + "loss": 1.0216, + "step": 56846 + }, + { + "epoch": 4.2, + "learning_rate": 4.357669254486307e-06, + "loss": 1.0389, + "step": 56847 + }, + { + "epoch": 4.2, + "learning_rate": 4.357339944362603e-06, + "loss": 1.0041, + "step": 56848 + }, + { + "epoch": 4.2, + "learning_rate": 4.3570106432162896e-06, + "loss": 0.9493, + "step": 56849 + }, + { + "epoch": 4.2, + "learning_rate": 4.356681351047892e-06, + "loss": 0.9881, + "step": 56850 + }, + { + "epoch": 4.2, + "learning_rate": 4.356352067857931e-06, + "loss": 1.0983, + "step": 56851 + }, + { + "epoch": 4.2, + "learning_rate": 4.356022793646934e-06, + "loss": 1.0014, + "step": 56852 + }, + { + "epoch": 4.2, + "learning_rate": 4.355693528415426e-06, + "loss": 1.1263, + "step": 56853 + }, + { + "epoch": 4.2, + "learning_rate": 4.355364272163929e-06, + "loss": 1.1592, + "step": 56854 + }, + { + "epoch": 4.2, + "learning_rate": 4.355035024892963e-06, + "loss": 1.0025, + "step": 56855 + }, + { + "epoch": 4.2, + "learning_rate": 4.354705786603058e-06, + "loss": 0.9357, + "step": 56856 + }, + { + "epoch": 4.2, + "learning_rate": 4.3543765572947376e-06, + "loss": 1.0498, + "step": 56857 + }, + { + "epoch": 4.2, + "learning_rate": 4.354047336968521e-06, + "loss": 1.0039, + "step": 56858 + }, + { + "epoch": 4.2, + "learning_rate": 4.353718125624935e-06, + "loss": 0.9903, + "step": 56859 + }, + { + "epoch": 4.2, + "learning_rate": 4.353388923264499e-06, + "loss": 0.9602, + "step": 56860 + }, + { + "epoch": 4.2, + "learning_rate": 4.353059729887749e-06, + "loss": 0.9569, + "step": 56861 + }, + { + "epoch": 4.2, + "learning_rate": 4.352730545495191e-06, + "loss": 1.0018, + "step": 56862 + }, + { + "epoch": 4.2, + "learning_rate": 4.352401370087363e-06, + "loss": 1.0458, + "step": 56863 + }, + { + "epoch": 4.2, + "learning_rate": 4.352072203664779e-06, + "loss": 0.9856, + "step": 56864 + }, + { + "epoch": 4.2, + "learning_rate": 4.3517430462279745e-06, + "loss": 1.0053, + "step": 56865 + }, + { + "epoch": 4.2, + "learning_rate": 4.351413897777459e-06, + "loss": 1.0014, + "step": 56866 + }, + { + "epoch": 4.2, + "learning_rate": 4.351084758313767e-06, + "loss": 0.9607, + "step": 56867 + }, + { + "epoch": 4.2, + "learning_rate": 4.350755627837418e-06, + "loss": 1.0574, + "step": 56868 + }, + { + "epoch": 4.2, + "learning_rate": 4.350426506348936e-06, + "loss": 0.8867, + "step": 56869 + }, + { + "epoch": 4.2, + "learning_rate": 4.350097393848846e-06, + "loss": 0.8679, + "step": 56870 + }, + { + "epoch": 4.2, + "learning_rate": 4.349768290337665e-06, + "loss": 1.0473, + "step": 56871 + }, + { + "epoch": 4.2, + "learning_rate": 4.349439195815925e-06, + "loss": 1.0348, + "step": 56872 + }, + { + "epoch": 4.2, + "learning_rate": 4.349110110284147e-06, + "loss": 1.0304, + "step": 56873 + }, + { + "epoch": 4.2, + "learning_rate": 4.3487810337428546e-06, + "loss": 1.0531, + "step": 56874 + }, + { + "epoch": 4.2, + "learning_rate": 4.3484519661925674e-06, + "loss": 1.026, + "step": 56875 + }, + { + "epoch": 4.2, + "learning_rate": 4.3481229076338145e-06, + "loss": 0.955, + "step": 56876 + }, + { + "epoch": 4.2, + "learning_rate": 4.347793858067118e-06, + "loss": 0.9096, + "step": 56877 + }, + { + "epoch": 4.2, + "learning_rate": 4.347464817493e-06, + "loss": 0.9993, + "step": 56878 + }, + { + "epoch": 4.2, + "learning_rate": 4.347135785911986e-06, + "loss": 1.0071, + "step": 56879 + }, + { + "epoch": 4.2, + "learning_rate": 4.346806763324593e-06, + "loss": 1.0286, + "step": 56880 + }, + { + "epoch": 4.2, + "learning_rate": 4.346477749731357e-06, + "loss": 1.1111, + "step": 56881 + }, + { + "epoch": 4.2, + "learning_rate": 4.3461487451327875e-06, + "loss": 1.1176, + "step": 56882 + }, + { + "epoch": 4.2, + "learning_rate": 4.345819749529417e-06, + "loss": 1.0769, + "step": 56883 + }, + { + "epoch": 4.2, + "learning_rate": 4.345490762921764e-06, + "loss": 0.9149, + "step": 56884 + }, + { + "epoch": 4.2, + "learning_rate": 4.3451617853103614e-06, + "loss": 1.0447, + "step": 56885 + }, + { + "epoch": 4.2, + "learning_rate": 4.344832816695721e-06, + "loss": 0.9356, + "step": 56886 + }, + { + "epoch": 4.2, + "learning_rate": 4.344503857078367e-06, + "loss": 0.9514, + "step": 56887 + }, + { + "epoch": 4.2, + "learning_rate": 4.344174906458831e-06, + "loss": 0.9767, + "step": 56888 + }, + { + "epoch": 4.2, + "learning_rate": 4.343845964837631e-06, + "loss": 0.9987, + "step": 56889 + }, + { + "epoch": 4.2, + "learning_rate": 4.343517032215292e-06, + "loss": 0.9149, + "step": 56890 + }, + { + "epoch": 4.2, + "learning_rate": 4.3431881085923325e-06, + "loss": 0.9336, + "step": 56891 + }, + { + "epoch": 4.2, + "learning_rate": 4.342859193969282e-06, + "loss": 1.0076, + "step": 56892 + }, + { + "epoch": 4.2, + "learning_rate": 4.342530288346664e-06, + "loss": 1.036, + "step": 56893 + }, + { + "epoch": 4.2, + "learning_rate": 4.342201391724997e-06, + "loss": 1.0387, + "step": 56894 + }, + { + "epoch": 4.2, + "learning_rate": 4.3418725041048086e-06, + "loss": 1.0048, + "step": 56895 + }, + { + "epoch": 4.2, + "learning_rate": 4.341543625486615e-06, + "loss": 0.9983, + "step": 56896 + }, + { + "epoch": 4.2, + "learning_rate": 4.341214755870949e-06, + "loss": 1.0436, + "step": 56897 + }, + { + "epoch": 4.2, + "learning_rate": 4.34088589525833e-06, + "loss": 1.0152, + "step": 56898 + }, + { + "epoch": 4.2, + "learning_rate": 4.340557043649279e-06, + "loss": 1.0028, + "step": 56899 + }, + { + "epoch": 4.2, + "learning_rate": 4.340228201044319e-06, + "loss": 0.9952, + "step": 56900 + }, + { + "epoch": 4.2, + "learning_rate": 4.339899367443977e-06, + "loss": 1.0015, + "step": 56901 + }, + { + "epoch": 4.2, + "learning_rate": 4.3395705428487744e-06, + "loss": 1.016, + "step": 56902 + }, + { + "epoch": 4.2, + "learning_rate": 4.339241727259235e-06, + "loss": 1.0301, + "step": 56903 + }, + { + "epoch": 4.2, + "learning_rate": 4.338912920675879e-06, + "loss": 1.1018, + "step": 56904 + }, + { + "epoch": 4.2, + "learning_rate": 4.3385841230992294e-06, + "loss": 1.0867, + "step": 56905 + }, + { + "epoch": 4.2, + "learning_rate": 4.3382553345298195e-06, + "loss": 1.0201, + "step": 56906 + }, + { + "epoch": 4.2, + "learning_rate": 4.337926554968156e-06, + "loss": 1.0293, + "step": 56907 + }, + { + "epoch": 4.2, + "learning_rate": 4.337597784414774e-06, + "loss": 0.9154, + "step": 56908 + }, + { + "epoch": 4.2, + "learning_rate": 4.337269022870189e-06, + "loss": 1.0053, + "step": 56909 + }, + { + "epoch": 4.21, + "learning_rate": 4.336940270334936e-06, + "loss": 0.9328, + "step": 56910 + }, + { + "epoch": 4.21, + "learning_rate": 4.336611526809522e-06, + "loss": 1.1207, + "step": 56911 + }, + { + "epoch": 4.21, + "learning_rate": 4.336282792294482e-06, + "loss": 1.139, + "step": 56912 + }, + { + "epoch": 4.21, + "learning_rate": 4.335954066790331e-06, + "loss": 1.0863, + "step": 56913 + }, + { + "epoch": 4.21, + "learning_rate": 4.3356253502976045e-06, + "loss": 0.9722, + "step": 56914 + }, + { + "epoch": 4.21, + "learning_rate": 4.3352966428168135e-06, + "loss": 0.9815, + "step": 56915 + }, + { + "epoch": 4.21, + "learning_rate": 4.334967944348479e-06, + "loss": 1.0129, + "step": 56916 + }, + { + "epoch": 4.21, + "learning_rate": 4.334639254893134e-06, + "loss": 1.0766, + "step": 56917 + }, + { + "epoch": 4.21, + "learning_rate": 4.334310574451297e-06, + "loss": 1.0912, + "step": 56918 + }, + { + "epoch": 4.21, + "learning_rate": 4.333981903023491e-06, + "loss": 0.9747, + "step": 56919 + }, + { + "epoch": 4.21, + "learning_rate": 4.333653240610235e-06, + "loss": 1.0953, + "step": 56920 + }, + { + "epoch": 4.21, + "learning_rate": 4.333324587212059e-06, + "loss": 1.0845, + "step": 56921 + }, + { + "epoch": 4.21, + "learning_rate": 4.332995942829482e-06, + "loss": 0.968, + "step": 56922 + }, + { + "epoch": 4.21, + "learning_rate": 4.332667307463029e-06, + "loss": 1.0549, + "step": 56923 + }, + { + "epoch": 4.21, + "learning_rate": 4.3323386811132195e-06, + "loss": 0.8902, + "step": 56924 + }, + { + "epoch": 4.21, + "learning_rate": 4.332010063780575e-06, + "loss": 1.0464, + "step": 56925 + }, + { + "epoch": 4.21, + "learning_rate": 4.33168145546563e-06, + "loss": 0.9421, + "step": 56926 + }, + { + "epoch": 4.21, + "learning_rate": 4.33135285616889e-06, + "loss": 0.9938, + "step": 56927 + }, + { + "epoch": 4.21, + "learning_rate": 4.33102426589089e-06, + "loss": 0.933, + "step": 56928 + }, + { + "epoch": 4.21, + "learning_rate": 4.330695684632146e-06, + "loss": 0.9827, + "step": 56929 + }, + { + "epoch": 4.21, + "learning_rate": 4.330367112393191e-06, + "loss": 1.0393, + "step": 56930 + }, + { + "epoch": 4.21, + "learning_rate": 4.3300385491745345e-06, + "loss": 1.0656, + "step": 56931 + }, + { + "epoch": 4.21, + "learning_rate": 4.3297099949767085e-06, + "loss": 0.9631, + "step": 56932 + }, + { + "epoch": 4.21, + "learning_rate": 4.329381449800233e-06, + "loss": 0.9986, + "step": 56933 + }, + { + "epoch": 4.21, + "learning_rate": 4.329052913645631e-06, + "loss": 0.9961, + "step": 56934 + }, + { + "epoch": 4.21, + "learning_rate": 4.328724386513423e-06, + "loss": 1.0277, + "step": 56935 + }, + { + "epoch": 4.21, + "learning_rate": 4.32839586840413e-06, + "loss": 0.9873, + "step": 56936 + }, + { + "epoch": 4.21, + "learning_rate": 4.328067359318283e-06, + "loss": 1.0607, + "step": 56937 + }, + { + "epoch": 4.21, + "learning_rate": 4.3277388592564e-06, + "loss": 1.088, + "step": 56938 + }, + { + "epoch": 4.21, + "learning_rate": 4.327410368219002e-06, + "loss": 0.9575, + "step": 56939 + }, + { + "epoch": 4.21, + "learning_rate": 4.3270818862066085e-06, + "loss": 0.9681, + "step": 56940 + }, + { + "epoch": 4.21, + "learning_rate": 4.326753413219752e-06, + "loss": 1.0542, + "step": 56941 + }, + { + "epoch": 4.21, + "learning_rate": 4.326424949258949e-06, + "loss": 0.9252, + "step": 56942 + }, + { + "epoch": 4.21, + "learning_rate": 4.326096494324723e-06, + "loss": 1.0637, + "step": 56943 + }, + { + "epoch": 4.21, + "learning_rate": 4.325768048417596e-06, + "loss": 1.0581, + "step": 56944 + }, + { + "epoch": 4.21, + "learning_rate": 4.325439611538087e-06, + "loss": 0.8524, + "step": 56945 + }, + { + "epoch": 4.21, + "learning_rate": 4.325111183686727e-06, + "loss": 0.9572, + "step": 56946 + }, + { + "epoch": 4.21, + "learning_rate": 4.324782764864033e-06, + "loss": 0.9445, + "step": 56947 + }, + { + "epoch": 4.21, + "learning_rate": 4.324454355070529e-06, + "loss": 0.9832, + "step": 56948 + }, + { + "epoch": 4.21, + "learning_rate": 4.324125954306732e-06, + "loss": 1.003, + "step": 56949 + }, + { + "epoch": 4.21, + "learning_rate": 4.323797562573174e-06, + "loss": 1.0272, + "step": 56950 + }, + { + "epoch": 4.21, + "learning_rate": 4.323469179870377e-06, + "loss": 1.0062, + "step": 56951 + }, + { + "epoch": 4.21, + "learning_rate": 4.323140806198851e-06, + "loss": 0.9606, + "step": 56952 + }, + { + "epoch": 4.21, + "learning_rate": 4.322812441559131e-06, + "loss": 1.0274, + "step": 56953 + }, + { + "epoch": 4.21, + "learning_rate": 4.322484085951732e-06, + "loss": 1.0615, + "step": 56954 + }, + { + "epoch": 4.21, + "learning_rate": 4.322155739377186e-06, + "loss": 0.8824, + "step": 56955 + }, + { + "epoch": 4.21, + "learning_rate": 4.321827401836002e-06, + "loss": 1.0014, + "step": 56956 + }, + { + "epoch": 4.21, + "learning_rate": 4.3214990733287135e-06, + "loss": 0.961, + "step": 56957 + }, + { + "epoch": 4.21, + "learning_rate": 4.321170753855835e-06, + "loss": 1.0629, + "step": 56958 + }, + { + "epoch": 4.21, + "learning_rate": 4.320842443417898e-06, + "loss": 0.9874, + "step": 56959 + }, + { + "epoch": 4.21, + "learning_rate": 4.320514142015417e-06, + "loss": 0.9895, + "step": 56960 + }, + { + "epoch": 4.21, + "learning_rate": 4.320185849648913e-06, + "loss": 1.063, + "step": 56961 + }, + { + "epoch": 4.21, + "learning_rate": 4.319857566318917e-06, + "loss": 0.8818, + "step": 56962 + }, + { + "epoch": 4.21, + "learning_rate": 4.319529292025945e-06, + "loss": 1.1032, + "step": 56963 + }, + { + "epoch": 4.21, + "learning_rate": 4.319201026770521e-06, + "loss": 1.0411, + "step": 56964 + }, + { + "epoch": 4.21, + "learning_rate": 4.318872770553164e-06, + "loss": 1.0537, + "step": 56965 + }, + { + "epoch": 4.21, + "learning_rate": 4.318544523374401e-06, + "loss": 1.0285, + "step": 56966 + }, + { + "epoch": 4.21, + "learning_rate": 4.3182162852347544e-06, + "loss": 0.9026, + "step": 56967 + }, + { + "epoch": 4.21, + "learning_rate": 4.317888056134742e-06, + "loss": 0.9957, + "step": 56968 + }, + { + "epoch": 4.21, + "learning_rate": 4.317559836074887e-06, + "loss": 0.8645, + "step": 56969 + }, + { + "epoch": 4.21, + "learning_rate": 4.317231625055716e-06, + "loss": 1.0055, + "step": 56970 + }, + { + "epoch": 4.21, + "learning_rate": 4.316903423077752e-06, + "loss": 1.005, + "step": 56971 + }, + { + "epoch": 4.21, + "learning_rate": 4.316575230141506e-06, + "loss": 0.9856, + "step": 56972 + }, + { + "epoch": 4.21, + "learning_rate": 4.316247046247511e-06, + "loss": 1.0445, + "step": 56973 + }, + { + "epoch": 4.21, + "learning_rate": 4.315918871396281e-06, + "loss": 0.9452, + "step": 56974 + }, + { + "epoch": 4.21, + "learning_rate": 4.31559070558835e-06, + "loss": 1.0228, + "step": 56975 + }, + { + "epoch": 4.21, + "learning_rate": 4.3152625488242275e-06, + "loss": 1.0603, + "step": 56976 + }, + { + "epoch": 4.21, + "learning_rate": 4.314934401104442e-06, + "loss": 1.005, + "step": 56977 + }, + { + "epoch": 4.21, + "learning_rate": 4.314606262429514e-06, + "loss": 0.8826, + "step": 56978 + }, + { + "epoch": 4.21, + "learning_rate": 4.31427813279997e-06, + "loss": 1.2035, + "step": 56979 + }, + { + "epoch": 4.21, + "learning_rate": 4.313950012216327e-06, + "loss": 0.9292, + "step": 56980 + }, + { + "epoch": 4.21, + "learning_rate": 4.3136219006791035e-06, + "loss": 0.9314, + "step": 56981 + }, + { + "epoch": 4.21, + "learning_rate": 4.31329379818883e-06, + "loss": 1.0945, + "step": 56982 + }, + { + "epoch": 4.21, + "learning_rate": 4.312965704746025e-06, + "loss": 1.0911, + "step": 56983 + }, + { + "epoch": 4.21, + "learning_rate": 4.31263762035121e-06, + "loss": 1.0063, + "step": 56984 + }, + { + "epoch": 4.21, + "learning_rate": 4.312309545004904e-06, + "loss": 0.9964, + "step": 56985 + }, + { + "epoch": 4.21, + "learning_rate": 4.311981478707635e-06, + "loss": 0.9405, + "step": 56986 + }, + { + "epoch": 4.21, + "learning_rate": 4.311653421459923e-06, + "loss": 0.9335, + "step": 56987 + }, + { + "epoch": 4.21, + "learning_rate": 4.311325373262288e-06, + "loss": 0.9717, + "step": 56988 + }, + { + "epoch": 4.21, + "learning_rate": 4.310997334115253e-06, + "loss": 0.8577, + "step": 56989 + }, + { + "epoch": 4.21, + "learning_rate": 4.3106693040193375e-06, + "loss": 1.009, + "step": 56990 + }, + { + "epoch": 4.21, + "learning_rate": 4.3103412829750676e-06, + "loss": 0.9823, + "step": 56991 + }, + { + "epoch": 4.21, + "learning_rate": 4.310013270982964e-06, + "loss": 0.9137, + "step": 56992 + }, + { + "epoch": 4.21, + "learning_rate": 4.3096852680435484e-06, + "loss": 0.9442, + "step": 56993 + }, + { + "epoch": 4.21, + "learning_rate": 4.3093572741573385e-06, + "loss": 1.1465, + "step": 56994 + }, + { + "epoch": 4.21, + "learning_rate": 4.309029289324863e-06, + "loss": 1.0369, + "step": 56995 + }, + { + "epoch": 4.21, + "learning_rate": 4.30870131354664e-06, + "loss": 0.998, + "step": 56996 + }, + { + "epoch": 4.21, + "learning_rate": 4.308373346823192e-06, + "loss": 0.9457, + "step": 56997 + }, + { + "epoch": 4.21, + "learning_rate": 4.308045389155041e-06, + "loss": 1.0205, + "step": 56998 + }, + { + "epoch": 4.21, + "learning_rate": 4.307717440542705e-06, + "loss": 1.0396, + "step": 56999 + }, + { + "epoch": 4.21, + "learning_rate": 4.307389500986716e-06, + "loss": 1.015, + "step": 57000 + }, + { + "epoch": 4.21, + "learning_rate": 4.307061570487582e-06, + "loss": 1.0501, + "step": 57001 + }, + { + "epoch": 4.21, + "learning_rate": 4.306733649045835e-06, + "loss": 0.8715, + "step": 57002 + }, + { + "epoch": 4.21, + "learning_rate": 4.306405736661992e-06, + "loss": 0.9371, + "step": 57003 + }, + { + "epoch": 4.21, + "learning_rate": 4.3060778333365775e-06, + "loss": 0.9926, + "step": 57004 + }, + { + "epoch": 4.21, + "learning_rate": 4.305749939070107e-06, + "loss": 1.1142, + "step": 57005 + }, + { + "epoch": 4.21, + "learning_rate": 4.305422053863112e-06, + "loss": 1.0808, + "step": 57006 + }, + { + "epoch": 4.21, + "learning_rate": 4.305094177716108e-06, + "loss": 0.9823, + "step": 57007 + }, + { + "epoch": 4.21, + "learning_rate": 4.304766310629618e-06, + "loss": 1.0424, + "step": 57008 + }, + { + "epoch": 4.21, + "learning_rate": 4.304438452604162e-06, + "loss": 0.9832, + "step": 57009 + }, + { + "epoch": 4.21, + "learning_rate": 4.304110603640262e-06, + "loss": 1.0042, + "step": 57010 + }, + { + "epoch": 4.21, + "learning_rate": 4.303782763738442e-06, + "loss": 0.982, + "step": 57011 + }, + { + "epoch": 4.21, + "learning_rate": 4.303454932899223e-06, + "loss": 0.9953, + "step": 57012 + }, + { + "epoch": 4.21, + "learning_rate": 4.303127111123126e-06, + "loss": 1.1008, + "step": 57013 + }, + { + "epoch": 4.21, + "learning_rate": 4.302799298410667e-06, + "loss": 0.9687, + "step": 57014 + }, + { + "epoch": 4.21, + "learning_rate": 4.302471494762378e-06, + "loss": 1.0585, + "step": 57015 + }, + { + "epoch": 4.21, + "learning_rate": 4.302143700178774e-06, + "loss": 0.9996, + "step": 57016 + }, + { + "epoch": 4.21, + "learning_rate": 4.301815914660379e-06, + "loss": 1.037, + "step": 57017 + }, + { + "epoch": 4.21, + "learning_rate": 4.301488138207713e-06, + "loss": 0.9559, + "step": 57018 + }, + { + "epoch": 4.21, + "learning_rate": 4.301160370821295e-06, + "loss": 0.9325, + "step": 57019 + }, + { + "epoch": 4.21, + "learning_rate": 4.300832612501655e-06, + "loss": 0.9899, + "step": 57020 + }, + { + "epoch": 4.21, + "learning_rate": 4.300504863249303e-06, + "loss": 1.0105, + "step": 57021 + }, + { + "epoch": 4.21, + "learning_rate": 4.300177123064769e-06, + "loss": 1.0641, + "step": 57022 + }, + { + "epoch": 4.21, + "learning_rate": 4.2998493919485685e-06, + "loss": 1.1273, + "step": 57023 + }, + { + "epoch": 4.21, + "learning_rate": 4.2995216699012335e-06, + "loss": 0.8934, + "step": 57024 + }, + { + "epoch": 4.21, + "learning_rate": 4.29919395692327e-06, + "loss": 0.9577, + "step": 57025 + }, + { + "epoch": 4.21, + "learning_rate": 4.29886625301521e-06, + "loss": 1.1128, + "step": 57026 + }, + { + "epoch": 4.21, + "learning_rate": 4.2985385581775744e-06, + "loss": 0.9051, + "step": 57027 + }, + { + "epoch": 4.21, + "learning_rate": 4.2982108724108815e-06, + "loss": 1.0317, + "step": 57028 + }, + { + "epoch": 4.21, + "learning_rate": 4.297883195715653e-06, + "loss": 0.9376, + "step": 57029 + }, + { + "epoch": 4.21, + "learning_rate": 4.297555528092408e-06, + "loss": 1.077, + "step": 57030 + }, + { + "epoch": 4.21, + "learning_rate": 4.297227869541674e-06, + "loss": 0.9792, + "step": 57031 + }, + { + "epoch": 4.21, + "learning_rate": 4.296900220063968e-06, + "loss": 0.9421, + "step": 57032 + }, + { + "epoch": 4.21, + "learning_rate": 4.296572579659813e-06, + "loss": 1.063, + "step": 57033 + }, + { + "epoch": 4.21, + "learning_rate": 4.296244948329725e-06, + "loss": 0.9656, + "step": 57034 + }, + { + "epoch": 4.21, + "learning_rate": 4.295917326074235e-06, + "loss": 0.9936, + "step": 57035 + }, + { + "epoch": 4.21, + "learning_rate": 4.295589712893857e-06, + "loss": 1.0747, + "step": 57036 + }, + { + "epoch": 4.21, + "learning_rate": 4.295262108789116e-06, + "loss": 0.9987, + "step": 57037 + }, + { + "epoch": 4.21, + "learning_rate": 4.294934513760529e-06, + "loss": 0.8554, + "step": 57038 + }, + { + "epoch": 4.21, + "learning_rate": 4.294606927808618e-06, + "loss": 1.0309, + "step": 57039 + }, + { + "epoch": 4.21, + "learning_rate": 4.294279350933908e-06, + "loss": 1.0343, + "step": 57040 + }, + { + "epoch": 4.21, + "learning_rate": 4.2939517831369195e-06, + "loss": 1.0566, + "step": 57041 + }, + { + "epoch": 4.21, + "learning_rate": 4.2936242244181705e-06, + "loss": 0.9244, + "step": 57042 + }, + { + "epoch": 4.21, + "learning_rate": 4.29329667477818e-06, + "loss": 1.06, + "step": 57043 + }, + { + "epoch": 4.21, + "learning_rate": 4.292969134217478e-06, + "loss": 1.0992, + "step": 57044 + }, + { + "epoch": 4.22, + "learning_rate": 4.292641602736583e-06, + "loss": 1.0279, + "step": 57045 + }, + { + "epoch": 4.22, + "learning_rate": 4.292314080336007e-06, + "loss": 0.9701, + "step": 57046 + }, + { + "epoch": 4.22, + "learning_rate": 4.291986567016281e-06, + "loss": 1.0863, + "step": 57047 + }, + { + "epoch": 4.22, + "learning_rate": 4.291659062777922e-06, + "loss": 0.9245, + "step": 57048 + }, + { + "epoch": 4.22, + "learning_rate": 4.291331567621452e-06, + "loss": 0.8981, + "step": 57049 + }, + { + "epoch": 4.22, + "learning_rate": 4.291004081547388e-06, + "loss": 0.9755, + "step": 57050 + }, + { + "epoch": 4.22, + "learning_rate": 4.29067660455626e-06, + "loss": 1.1377, + "step": 57051 + }, + { + "epoch": 4.22, + "learning_rate": 4.290349136648583e-06, + "loss": 1.0297, + "step": 57052 + }, + { + "epoch": 4.22, + "learning_rate": 4.290021677824878e-06, + "loss": 0.9631, + "step": 57053 + }, + { + "epoch": 4.22, + "learning_rate": 4.289694228085667e-06, + "loss": 0.9984, + "step": 57054 + }, + { + "epoch": 4.22, + "learning_rate": 4.289366787431468e-06, + "loss": 0.9003, + "step": 57055 + }, + { + "epoch": 4.22, + "learning_rate": 4.289039355862808e-06, + "loss": 1.061, + "step": 57056 + }, + { + "epoch": 4.22, + "learning_rate": 4.288711933380204e-06, + "loss": 0.9675, + "step": 57057 + }, + { + "epoch": 4.22, + "learning_rate": 4.2883845199841775e-06, + "loss": 1.0669, + "step": 57058 + }, + { + "epoch": 4.22, + "learning_rate": 4.288057115675247e-06, + "loss": 1.0095, + "step": 57059 + }, + { + "epoch": 4.22, + "learning_rate": 4.287729720453939e-06, + "loss": 0.9529, + "step": 57060 + }, + { + "epoch": 4.22, + "learning_rate": 4.287402334320771e-06, + "loss": 1.0687, + "step": 57061 + }, + { + "epoch": 4.22, + "learning_rate": 4.2870749572762646e-06, + "loss": 1.0463, + "step": 57062 + }, + { + "epoch": 4.22, + "learning_rate": 4.286747589320941e-06, + "loss": 1.0244, + "step": 57063 + }, + { + "epoch": 4.22, + "learning_rate": 4.286420230455316e-06, + "loss": 1.0324, + "step": 57064 + }, + { + "epoch": 4.22, + "learning_rate": 4.286092880679922e-06, + "loss": 1.0075, + "step": 57065 + }, + { + "epoch": 4.22, + "learning_rate": 4.285765539995265e-06, + "loss": 1.0175, + "step": 57066 + }, + { + "epoch": 4.22, + "learning_rate": 4.285438208401877e-06, + "loss": 0.9323, + "step": 57067 + }, + { + "epoch": 4.22, + "learning_rate": 4.28511088590027e-06, + "loss": 1.0265, + "step": 57068 + }, + { + "epoch": 4.22, + "learning_rate": 4.284783572490978e-06, + "loss": 0.9561, + "step": 57069 + }, + { + "epoch": 4.22, + "learning_rate": 4.284456268174507e-06, + "loss": 1.0134, + "step": 57070 + }, + { + "epoch": 4.22, + "learning_rate": 4.284128972951388e-06, + "loss": 1.0506, + "step": 57071 + }, + { + "epoch": 4.22, + "learning_rate": 4.2838016868221335e-06, + "loss": 1.0045, + "step": 57072 + }, + { + "epoch": 4.22, + "learning_rate": 4.283474409787276e-06, + "loss": 1.0325, + "step": 57073 + }, + { + "epoch": 4.22, + "learning_rate": 4.2831471418473255e-06, + "loss": 1.0019, + "step": 57074 + }, + { + "epoch": 4.22, + "learning_rate": 4.282819883002804e-06, + "loss": 1.0321, + "step": 57075 + }, + { + "epoch": 4.22, + "learning_rate": 4.282492633254236e-06, + "loss": 1.0294, + "step": 57076 + }, + { + "epoch": 4.22, + "learning_rate": 4.28216539260214e-06, + "loss": 1.0739, + "step": 57077 + }, + { + "epoch": 4.22, + "learning_rate": 4.281838161047038e-06, + "loss": 0.9689, + "step": 57078 + }, + { + "epoch": 4.22, + "learning_rate": 4.281510938589447e-06, + "loss": 1.0093, + "step": 57079 + }, + { + "epoch": 4.22, + "learning_rate": 4.281183725229892e-06, + "loss": 1.0568, + "step": 57080 + }, + { + "epoch": 4.22, + "learning_rate": 4.280856520968894e-06, + "loss": 0.9619, + "step": 57081 + }, + { + "epoch": 4.22, + "learning_rate": 4.28052932580697e-06, + "loss": 1.0804, + "step": 57082 + }, + { + "epoch": 4.22, + "learning_rate": 4.280202139744642e-06, + "loss": 0.9978, + "step": 57083 + }, + { + "epoch": 4.22, + "learning_rate": 4.279874962782428e-06, + "loss": 0.9328, + "step": 57084 + }, + { + "epoch": 4.22, + "learning_rate": 4.279547794920853e-06, + "loss": 1.0095, + "step": 57085 + }, + { + "epoch": 4.22, + "learning_rate": 4.279220636160437e-06, + "loss": 1.0057, + "step": 57086 + }, + { + "epoch": 4.22, + "learning_rate": 4.2788934865016986e-06, + "loss": 0.9911, + "step": 57087 + }, + { + "epoch": 4.22, + "learning_rate": 4.278566345945155e-06, + "loss": 0.9426, + "step": 57088 + }, + { + "epoch": 4.22, + "learning_rate": 4.278239214491334e-06, + "loss": 0.9656, + "step": 57089 + }, + { + "epoch": 4.22, + "learning_rate": 4.2779120921407536e-06, + "loss": 1.0972, + "step": 57090 + }, + { + "epoch": 4.22, + "learning_rate": 4.277584978893933e-06, + "loss": 0.9134, + "step": 57091 + }, + { + "epoch": 4.22, + "learning_rate": 4.277257874751392e-06, + "loss": 1.009, + "step": 57092 + }, + { + "epoch": 4.22, + "learning_rate": 4.276930779713652e-06, + "loss": 1.0182, + "step": 57093 + }, + { + "epoch": 4.22, + "learning_rate": 4.2766036937812325e-06, + "loss": 1.0367, + "step": 57094 + }, + { + "epoch": 4.22, + "learning_rate": 4.276276616954652e-06, + "loss": 1.0177, + "step": 57095 + }, + { + "epoch": 4.22, + "learning_rate": 4.275949549234437e-06, + "loss": 0.9446, + "step": 57096 + }, + { + "epoch": 4.22, + "learning_rate": 4.275622490621105e-06, + "loss": 1.1472, + "step": 57097 + }, + { + "epoch": 4.22, + "learning_rate": 4.275295441115175e-06, + "loss": 1.0252, + "step": 57098 + }, + { + "epoch": 4.22, + "learning_rate": 4.2749684007171645e-06, + "loss": 0.9107, + "step": 57099 + }, + { + "epoch": 4.22, + "learning_rate": 4.274641369427601e-06, + "loss": 1.1414, + "step": 57100 + }, + { + "epoch": 4.22, + "learning_rate": 4.274314347247002e-06, + "loss": 1.0725, + "step": 57101 + }, + { + "epoch": 4.22, + "learning_rate": 4.273987334175886e-06, + "loss": 1.0123, + "step": 57102 + }, + { + "epoch": 4.22, + "learning_rate": 4.273660330214774e-06, + "loss": 1.0876, + "step": 57103 + }, + { + "epoch": 4.22, + "learning_rate": 4.273333335364184e-06, + "loss": 1.0196, + "step": 57104 + }, + { + "epoch": 4.22, + "learning_rate": 4.273006349624641e-06, + "loss": 1.0283, + "step": 57105 + }, + { + "epoch": 4.22, + "learning_rate": 4.272679372996663e-06, + "loss": 0.9211, + "step": 57106 + }, + { + "epoch": 4.22, + "learning_rate": 4.272352405480771e-06, + "loss": 0.9846, + "step": 57107 + }, + { + "epoch": 4.22, + "learning_rate": 4.27202544707748e-06, + "loss": 1.0484, + "step": 57108 + }, + { + "epoch": 4.22, + "learning_rate": 4.27169849778732e-06, + "loss": 1.0938, + "step": 57109 + }, + { + "epoch": 4.22, + "learning_rate": 4.271371557610807e-06, + "loss": 0.8277, + "step": 57110 + }, + { + "epoch": 4.22, + "learning_rate": 4.271044626548454e-06, + "loss": 0.9406, + "step": 57111 + }, + { + "epoch": 4.22, + "learning_rate": 4.270717704600791e-06, + "loss": 1.0588, + "step": 57112 + }, + { + "epoch": 4.22, + "learning_rate": 4.27039079176833e-06, + "loss": 1.0824, + "step": 57113 + }, + { + "epoch": 4.22, + "learning_rate": 4.270063888051602e-06, + "loss": 1.0698, + "step": 57114 + }, + { + "epoch": 4.22, + "learning_rate": 4.2697369934511134e-06, + "loss": 1.0468, + "step": 57115 + }, + { + "epoch": 4.22, + "learning_rate": 4.269410107967394e-06, + "loss": 0.9869, + "step": 57116 + }, + { + "epoch": 4.22, + "learning_rate": 4.269083231600959e-06, + "loss": 1.0921, + "step": 57117 + }, + { + "epoch": 4.22, + "learning_rate": 4.268756364352338e-06, + "loss": 1.0344, + "step": 57118 + }, + { + "epoch": 4.22, + "learning_rate": 4.268429506222039e-06, + "loss": 0.9745, + "step": 57119 + }, + { + "epoch": 4.22, + "learning_rate": 4.268102657210583e-06, + "loss": 0.9801, + "step": 57120 + }, + { + "epoch": 4.22, + "learning_rate": 4.267775817318498e-06, + "loss": 0.9474, + "step": 57121 + }, + { + "epoch": 4.22, + "learning_rate": 4.267448986546299e-06, + "loss": 1.0855, + "step": 57122 + }, + { + "epoch": 4.22, + "learning_rate": 4.267122164894507e-06, + "loss": 1.081, + "step": 57123 + }, + { + "epoch": 4.22, + "learning_rate": 4.2667953523636385e-06, + "loss": 0.9984, + "step": 57124 + }, + { + "epoch": 4.22, + "learning_rate": 4.2664685489542194e-06, + "loss": 1.0513, + "step": 57125 + }, + { + "epoch": 4.22, + "learning_rate": 4.2661417546667676e-06, + "loss": 0.9344, + "step": 57126 + }, + { + "epoch": 4.22, + "learning_rate": 4.2658149695018e-06, + "loss": 0.8784, + "step": 57127 + }, + { + "epoch": 4.22, + "learning_rate": 4.265488193459838e-06, + "loss": 1.0654, + "step": 57128 + }, + { + "epoch": 4.22, + "learning_rate": 4.265161426541405e-06, + "loss": 1.0, + "step": 57129 + }, + { + "epoch": 4.22, + "learning_rate": 4.264834668747016e-06, + "loss": 1.0297, + "step": 57130 + }, + { + "epoch": 4.22, + "learning_rate": 4.2645079200771955e-06, + "loss": 0.9415, + "step": 57131 + }, + { + "epoch": 4.22, + "learning_rate": 4.264181180532461e-06, + "loss": 0.9721, + "step": 57132 + }, + { + "epoch": 4.22, + "learning_rate": 4.263854450113328e-06, + "loss": 1.0033, + "step": 57133 + }, + { + "epoch": 4.22, + "learning_rate": 4.263527728820322e-06, + "loss": 0.9177, + "step": 57134 + }, + { + "epoch": 4.22, + "learning_rate": 4.263201016653964e-06, + "loss": 0.9195, + "step": 57135 + }, + { + "epoch": 4.22, + "learning_rate": 4.2628743136147695e-06, + "loss": 1.0721, + "step": 57136 + }, + { + "epoch": 4.22, + "learning_rate": 4.262547619703257e-06, + "loss": 0.9506, + "step": 57137 + }, + { + "epoch": 4.22, + "learning_rate": 4.262220934919956e-06, + "loss": 0.9321, + "step": 57138 + }, + { + "epoch": 4.22, + "learning_rate": 4.2618942592653765e-06, + "loss": 1.0221, + "step": 57139 + }, + { + "epoch": 4.22, + "learning_rate": 4.261567592740037e-06, + "loss": 1.0105, + "step": 57140 + }, + { + "epoch": 4.22, + "learning_rate": 4.261240935344465e-06, + "loss": 1.0018, + "step": 57141 + }, + { + "epoch": 4.22, + "learning_rate": 4.260914287079176e-06, + "loss": 1.0059, + "step": 57142 + }, + { + "epoch": 4.22, + "learning_rate": 4.260587647944692e-06, + "loss": 1.0131, + "step": 57143 + }, + { + "epoch": 4.22, + "learning_rate": 4.260261017941526e-06, + "loss": 0.9616, + "step": 57144 + }, + { + "epoch": 4.22, + "learning_rate": 4.259934397070207e-06, + "loss": 0.9646, + "step": 57145 + }, + { + "epoch": 4.22, + "learning_rate": 4.259607785331249e-06, + "loss": 1.0625, + "step": 57146 + }, + { + "epoch": 4.22, + "learning_rate": 4.2592811827251734e-06, + "loss": 0.9608, + "step": 57147 + }, + { + "epoch": 4.22, + "learning_rate": 4.258954589252499e-06, + "loss": 0.8175, + "step": 57148 + }, + { + "epoch": 4.22, + "learning_rate": 4.258628004913743e-06, + "loss": 0.9486, + "step": 57149 + }, + { + "epoch": 4.22, + "learning_rate": 4.258301429709432e-06, + "loss": 0.9333, + "step": 57150 + }, + { + "epoch": 4.22, + "learning_rate": 4.25797486364008e-06, + "loss": 0.9751, + "step": 57151 + }, + { + "epoch": 4.22, + "learning_rate": 4.257648306706207e-06, + "loss": 1.0954, + "step": 57152 + }, + { + "epoch": 4.22, + "learning_rate": 4.2573217589083304e-06, + "loss": 0.8674, + "step": 57153 + }, + { + "epoch": 4.22, + "learning_rate": 4.256995220246976e-06, + "loss": 0.9679, + "step": 57154 + }, + { + "epoch": 4.22, + "learning_rate": 4.256668690722661e-06, + "loss": 0.7886, + "step": 57155 + }, + { + "epoch": 4.22, + "learning_rate": 4.256342170335904e-06, + "loss": 1.0505, + "step": 57156 + }, + { + "epoch": 4.22, + "learning_rate": 4.256015659087225e-06, + "loss": 1.0048, + "step": 57157 + }, + { + "epoch": 4.22, + "learning_rate": 4.255689156977138e-06, + "loss": 1.0953, + "step": 57158 + }, + { + "epoch": 4.22, + "learning_rate": 4.255362664006175e-06, + "loss": 1.1246, + "step": 57159 + }, + { + "epoch": 4.22, + "learning_rate": 4.255036180174841e-06, + "loss": 0.969, + "step": 57160 + }, + { + "epoch": 4.22, + "learning_rate": 4.254709705483665e-06, + "loss": 1.0155, + "step": 57161 + }, + { + "epoch": 4.22, + "learning_rate": 4.254383239933161e-06, + "loss": 1.0387, + "step": 57162 + }, + { + "epoch": 4.22, + "learning_rate": 4.254056783523857e-06, + "loss": 0.9506, + "step": 57163 + }, + { + "epoch": 4.22, + "learning_rate": 4.25373033625626e-06, + "loss": 0.9471, + "step": 57164 + }, + { + "epoch": 4.22, + "learning_rate": 4.253403898130899e-06, + "loss": 1.0306, + "step": 57165 + }, + { + "epoch": 4.22, + "learning_rate": 4.253077469148291e-06, + "loss": 1.0245, + "step": 57166 + }, + { + "epoch": 4.22, + "learning_rate": 4.252751049308954e-06, + "loss": 0.9959, + "step": 57167 + }, + { + "epoch": 4.22, + "learning_rate": 4.2524246386134085e-06, + "loss": 0.9796, + "step": 57168 + }, + { + "epoch": 4.22, + "learning_rate": 4.252098237062169e-06, + "loss": 0.8592, + "step": 57169 + }, + { + "epoch": 4.22, + "learning_rate": 4.251771844655762e-06, + "loss": 0.9935, + "step": 57170 + }, + { + "epoch": 4.22, + "learning_rate": 4.251445461394704e-06, + "loss": 1.0613, + "step": 57171 + }, + { + "epoch": 4.22, + "learning_rate": 4.251119087279515e-06, + "loss": 0.984, + "step": 57172 + }, + { + "epoch": 4.22, + "learning_rate": 4.250792722310708e-06, + "loss": 1.0055, + "step": 57173 + }, + { + "epoch": 4.22, + "learning_rate": 4.250466366488812e-06, + "loss": 0.9284, + "step": 57174 + }, + { + "epoch": 4.22, + "learning_rate": 4.250140019814343e-06, + "loss": 1.0233, + "step": 57175 + }, + { + "epoch": 4.22, + "learning_rate": 4.249813682287817e-06, + "loss": 0.9776, + "step": 57176 + }, + { + "epoch": 4.22, + "learning_rate": 4.249487353909757e-06, + "loss": 0.9396, + "step": 57177 + }, + { + "epoch": 4.22, + "learning_rate": 4.2491610346806755e-06, + "loss": 1.0896, + "step": 57178 + }, + { + "epoch": 4.22, + "learning_rate": 4.2488347246010995e-06, + "loss": 1.012, + "step": 57179 + }, + { + "epoch": 4.22, + "learning_rate": 4.248508423671547e-06, + "loss": 1.0562, + "step": 57180 + }, + { + "epoch": 4.23, + "learning_rate": 4.248182131892535e-06, + "loss": 0.9949, + "step": 57181 + }, + { + "epoch": 4.23, + "learning_rate": 4.247855849264577e-06, + "loss": 0.8992, + "step": 57182 + }, + { + "epoch": 4.23, + "learning_rate": 4.247529575788207e-06, + "loss": 0.9983, + "step": 57183 + }, + { + "epoch": 4.23, + "learning_rate": 4.247203311463929e-06, + "loss": 1.0459, + "step": 57184 + }, + { + "epoch": 4.23, + "learning_rate": 4.24687705629227e-06, + "loss": 1.0454, + "step": 57185 + }, + { + "epoch": 4.23, + "learning_rate": 4.2465508102737486e-06, + "loss": 1.0731, + "step": 57186 + }, + { + "epoch": 4.23, + "learning_rate": 4.246224573408882e-06, + "loss": 0.9047, + "step": 57187 + }, + { + "epoch": 4.23, + "learning_rate": 4.2458983456981895e-06, + "loss": 1.1337, + "step": 57188 + }, + { + "epoch": 4.23, + "learning_rate": 4.245572127142187e-06, + "loss": 1.0381, + "step": 57189 + }, + { + "epoch": 4.23, + "learning_rate": 4.2452459177413995e-06, + "loss": 1.0658, + "step": 57190 + }, + { + "epoch": 4.23, + "learning_rate": 4.244919717496345e-06, + "loss": 0.9295, + "step": 57191 + }, + { + "epoch": 4.23, + "learning_rate": 4.24459352640754e-06, + "loss": 1.0447, + "step": 57192 + }, + { + "epoch": 4.23, + "learning_rate": 4.244267344475501e-06, + "loss": 0.9313, + "step": 57193 + }, + { + "epoch": 4.23, + "learning_rate": 4.243941171700754e-06, + "loss": 0.9326, + "step": 57194 + }, + { + "epoch": 4.23, + "learning_rate": 4.2436150080838145e-06, + "loss": 0.9894, + "step": 57195 + }, + { + "epoch": 4.23, + "learning_rate": 4.243288853625201e-06, + "loss": 0.9636, + "step": 57196 + }, + { + "epoch": 4.23, + "learning_rate": 4.242962708325432e-06, + "loss": 0.8845, + "step": 57197 + }, + { + "epoch": 4.23, + "learning_rate": 4.242636572185024e-06, + "loss": 0.9343, + "step": 57198 + }, + { + "epoch": 4.23, + "learning_rate": 4.242310445204501e-06, + "loss": 0.9711, + "step": 57199 + }, + { + "epoch": 4.23, + "learning_rate": 4.241984327384382e-06, + "loss": 1.0033, + "step": 57200 + }, + { + "epoch": 4.23, + "learning_rate": 4.241658218725182e-06, + "loss": 1.0082, + "step": 57201 + }, + { + "epoch": 4.23, + "learning_rate": 4.2413321192274196e-06, + "loss": 0.9815, + "step": 57202 + }, + { + "epoch": 4.23, + "learning_rate": 4.2410060288916165e-06, + "loss": 0.9819, + "step": 57203 + }, + { + "epoch": 4.23, + "learning_rate": 4.240679947718296e-06, + "loss": 0.9978, + "step": 57204 + }, + { + "epoch": 4.23, + "learning_rate": 4.2403538757079634e-06, + "loss": 0.8965, + "step": 57205 + }, + { + "epoch": 4.23, + "learning_rate": 4.2400278128611485e-06, + "loss": 1.0662, + "step": 57206 + }, + { + "epoch": 4.23, + "learning_rate": 4.239701759178364e-06, + "loss": 1.0689, + "step": 57207 + }, + { + "epoch": 4.23, + "learning_rate": 4.239375714660138e-06, + "loss": 1.0647, + "step": 57208 + }, + { + "epoch": 4.23, + "learning_rate": 4.239049679306976e-06, + "loss": 1.0395, + "step": 57209 + }, + { + "epoch": 4.23, + "learning_rate": 4.238723653119408e-06, + "loss": 0.98, + "step": 57210 + }, + { + "epoch": 4.23, + "learning_rate": 4.238397636097943e-06, + "loss": 0.9959, + "step": 57211 + }, + { + "epoch": 4.23, + "learning_rate": 4.238071628243113e-06, + "loss": 0.924, + "step": 57212 + }, + { + "epoch": 4.23, + "learning_rate": 4.237745629555425e-06, + "loss": 1.0326, + "step": 57213 + }, + { + "epoch": 4.23, + "learning_rate": 4.237419640035397e-06, + "loss": 1.0121, + "step": 57214 + }, + { + "epoch": 4.23, + "learning_rate": 4.237093659683557e-06, + "loss": 0.978, + "step": 57215 + }, + { + "epoch": 4.23, + "learning_rate": 4.236767688500417e-06, + "loss": 1.0252, + "step": 57216 + }, + { + "epoch": 4.23, + "learning_rate": 4.236441726486499e-06, + "loss": 1.1038, + "step": 57217 + }, + { + "epoch": 4.23, + "learning_rate": 4.236115773642314e-06, + "loss": 1.1204, + "step": 57218 + }, + { + "epoch": 4.23, + "learning_rate": 4.235789829968392e-06, + "loss": 0.9312, + "step": 57219 + }, + { + "epoch": 4.23, + "learning_rate": 4.235463895465244e-06, + "loss": 0.9564, + "step": 57220 + }, + { + "epoch": 4.23, + "learning_rate": 4.235137970133391e-06, + "loss": 1.1162, + "step": 57221 + }, + { + "epoch": 4.23, + "learning_rate": 4.234812053973352e-06, + "loss": 0.9858, + "step": 57222 + }, + { + "epoch": 4.23, + "learning_rate": 4.23448614698564e-06, + "loss": 1.104, + "step": 57223 + }, + { + "epoch": 4.23, + "learning_rate": 4.234160249170782e-06, + "loss": 1.0134, + "step": 57224 + }, + { + "epoch": 4.23, + "learning_rate": 4.2338343605292916e-06, + "loss": 1.1367, + "step": 57225 + }, + { + "epoch": 4.23, + "learning_rate": 4.23350848106169e-06, + "loss": 0.9473, + "step": 57226 + }, + { + "epoch": 4.23, + "learning_rate": 4.233182610768489e-06, + "loss": 1.0339, + "step": 57227 + }, + { + "epoch": 4.23, + "learning_rate": 4.23285674965022e-06, + "loss": 0.9134, + "step": 57228 + }, + { + "epoch": 4.23, + "learning_rate": 4.2325308977073855e-06, + "loss": 0.9615, + "step": 57229 + }, + { + "epoch": 4.23, + "learning_rate": 4.2322050549405155e-06, + "loss": 0.9481, + "step": 57230 + }, + { + "epoch": 4.23, + "learning_rate": 4.231879221350121e-06, + "loss": 1.0213, + "step": 57231 + }, + { + "epoch": 4.23, + "learning_rate": 4.231553396936731e-06, + "loss": 1.0779, + "step": 57232 + }, + { + "epoch": 4.23, + "learning_rate": 4.231227581700856e-06, + "loss": 1.0011, + "step": 57233 + }, + { + "epoch": 4.23, + "learning_rate": 4.23090177564301e-06, + "loss": 0.9549, + "step": 57234 + }, + { + "epoch": 4.23, + "learning_rate": 4.230575978763719e-06, + "loss": 1.0447, + "step": 57235 + }, + { + "epoch": 4.23, + "learning_rate": 4.230250191063501e-06, + "loss": 0.9339, + "step": 57236 + }, + { + "epoch": 4.23, + "learning_rate": 4.229924412542871e-06, + "loss": 0.9964, + "step": 57237 + }, + { + "epoch": 4.23, + "learning_rate": 4.229598643202346e-06, + "loss": 0.9552, + "step": 57238 + }, + { + "epoch": 4.23, + "learning_rate": 4.22927288304245e-06, + "loss": 0.8979, + "step": 57239 + }, + { + "epoch": 4.23, + "learning_rate": 4.228947132063699e-06, + "loss": 1.0329, + "step": 57240 + }, + { + "epoch": 4.23, + "learning_rate": 4.22862139026661e-06, + "loss": 0.9416, + "step": 57241 + }, + { + "epoch": 4.23, + "learning_rate": 4.228295657651702e-06, + "loss": 1.0006, + "step": 57242 + }, + { + "epoch": 4.23, + "learning_rate": 4.227969934219489e-06, + "loss": 0.9621, + "step": 57243 + }, + { + "epoch": 4.23, + "learning_rate": 4.227644219970497e-06, + "loss": 1.0132, + "step": 57244 + }, + { + "epoch": 4.23, + "learning_rate": 4.227318514905241e-06, + "loss": 0.9457, + "step": 57245 + }, + { + "epoch": 4.23, + "learning_rate": 4.2269928190242385e-06, + "loss": 0.9219, + "step": 57246 + }, + { + "epoch": 4.23, + "learning_rate": 4.226667132328004e-06, + "loss": 0.994, + "step": 57247 + }, + { + "epoch": 4.23, + "learning_rate": 4.2263414548170636e-06, + "loss": 0.8834, + "step": 57248 + }, + { + "epoch": 4.23, + "learning_rate": 4.226015786491931e-06, + "loss": 1.0327, + "step": 57249 + }, + { + "epoch": 4.23, + "learning_rate": 4.225690127353123e-06, + "loss": 1.0583, + "step": 57250 + }, + { + "epoch": 4.23, + "learning_rate": 4.225364477401162e-06, + "loss": 1.0426, + "step": 57251 + }, + { + "epoch": 4.23, + "learning_rate": 4.225038836636558e-06, + "loss": 1.0637, + "step": 57252 + }, + { + "epoch": 4.23, + "learning_rate": 4.224713205059843e-06, + "loss": 1.0296, + "step": 57253 + }, + { + "epoch": 4.23, + "learning_rate": 4.224387582671518e-06, + "loss": 1.0342, + "step": 57254 + }, + { + "epoch": 4.23, + "learning_rate": 4.224061969472115e-06, + "loss": 1.1762, + "step": 57255 + }, + { + "epoch": 4.23, + "learning_rate": 4.223736365462143e-06, + "loss": 0.958, + "step": 57256 + }, + { + "epoch": 4.23, + "learning_rate": 4.22341077064213e-06, + "loss": 1.0459, + "step": 57257 + }, + { + "epoch": 4.23, + "learning_rate": 4.223085185012582e-06, + "loss": 0.9043, + "step": 57258 + }, + { + "epoch": 4.23, + "learning_rate": 4.222759608574025e-06, + "loss": 1.1033, + "step": 57259 + }, + { + "epoch": 4.23, + "learning_rate": 4.222434041326976e-06, + "loss": 0.995, + "step": 57260 + }, + { + "epoch": 4.23, + "learning_rate": 4.22210848327195e-06, + "loss": 1.0302, + "step": 57261 + }, + { + "epoch": 4.23, + "learning_rate": 4.2217829344094685e-06, + "loss": 1.0438, + "step": 57262 + }, + { + "epoch": 4.23, + "learning_rate": 4.221457394740044e-06, + "loss": 1.012, + "step": 57263 + }, + { + "epoch": 4.23, + "learning_rate": 4.221131864264201e-06, + "loss": 0.9897, + "step": 57264 + }, + { + "epoch": 4.23, + "learning_rate": 4.220806342982454e-06, + "loss": 0.8551, + "step": 57265 + }, + { + "epoch": 4.23, + "learning_rate": 4.220480830895323e-06, + "loss": 1.0388, + "step": 57266 + }, + { + "epoch": 4.23, + "learning_rate": 4.220155328003319e-06, + "loss": 1.0852, + "step": 57267 + }, + { + "epoch": 4.23, + "learning_rate": 4.21982983430697e-06, + "loss": 0.9823, + "step": 57268 + }, + { + "epoch": 4.23, + "learning_rate": 4.2195043498067925e-06, + "loss": 0.9714, + "step": 57269 + }, + { + "epoch": 4.23, + "learning_rate": 4.219178874503294e-06, + "loss": 1.0291, + "step": 57270 + }, + { + "epoch": 4.23, + "learning_rate": 4.218853408397001e-06, + "loss": 0.9721, + "step": 57271 + }, + { + "epoch": 4.23, + "learning_rate": 4.218527951488428e-06, + "loss": 0.9584, + "step": 57272 + }, + { + "epoch": 4.23, + "learning_rate": 4.2182025037781005e-06, + "loss": 0.9633, + "step": 57273 + }, + { + "epoch": 4.23, + "learning_rate": 4.217877065266522e-06, + "loss": 0.9102, + "step": 57274 + }, + { + "epoch": 4.23, + "learning_rate": 4.217551635954223e-06, + "loss": 1.0159, + "step": 57275 + }, + { + "epoch": 4.23, + "learning_rate": 4.217226215841714e-06, + "loss": 0.9869, + "step": 57276 + }, + { + "epoch": 4.23, + "learning_rate": 4.2169008049295214e-06, + "loss": 1.0726, + "step": 57277 + }, + { + "epoch": 4.23, + "learning_rate": 4.2165754032181536e-06, + "loss": 1.0082, + "step": 57278 + }, + { + "epoch": 4.23, + "learning_rate": 4.2162500107081275e-06, + "loss": 0.9505, + "step": 57279 + }, + { + "epoch": 4.23, + "learning_rate": 4.215924627399969e-06, + "loss": 1.0452, + "step": 57280 + }, + { + "epoch": 4.23, + "learning_rate": 4.215599253294191e-06, + "loss": 0.9972, + "step": 57281 + }, + { + "epoch": 4.23, + "learning_rate": 4.2152738883913125e-06, + "loss": 1.2156, + "step": 57282 + }, + { + "epoch": 4.23, + "learning_rate": 4.214948532691847e-06, + "loss": 1.153, + "step": 57283 + }, + { + "epoch": 4.23, + "learning_rate": 4.21462318619632e-06, + "loss": 1.0731, + "step": 57284 + }, + { + "epoch": 4.23, + "learning_rate": 4.214297848905242e-06, + "loss": 1.0117, + "step": 57285 + }, + { + "epoch": 4.23, + "learning_rate": 4.213972520819135e-06, + "loss": 1.0039, + "step": 57286 + }, + { + "epoch": 4.23, + "learning_rate": 4.213647201938512e-06, + "loss": 0.9821, + "step": 57287 + }, + { + "epoch": 4.23, + "learning_rate": 4.213321892263897e-06, + "loss": 1.0079, + "step": 57288 + }, + { + "epoch": 4.23, + "learning_rate": 4.212996591795803e-06, + "loss": 1.1166, + "step": 57289 + }, + { + "epoch": 4.23, + "learning_rate": 4.212671300534749e-06, + "loss": 0.9737, + "step": 57290 + }, + { + "epoch": 4.23, + "learning_rate": 4.212346018481253e-06, + "loss": 1.0014, + "step": 57291 + }, + { + "epoch": 4.23, + "learning_rate": 4.212020745635827e-06, + "loss": 1.039, + "step": 57292 + }, + { + "epoch": 4.23, + "learning_rate": 4.2116954819989964e-06, + "loss": 1.0032, + "step": 57293 + }, + { + "epoch": 4.23, + "learning_rate": 4.211370227571277e-06, + "loss": 1.0348, + "step": 57294 + }, + { + "epoch": 4.23, + "learning_rate": 4.211044982353183e-06, + "loss": 0.9343, + "step": 57295 + }, + { + "epoch": 4.23, + "learning_rate": 4.210719746345231e-06, + "loss": 0.8688, + "step": 57296 + }, + { + "epoch": 4.23, + "learning_rate": 4.2103945195479444e-06, + "loss": 0.9893, + "step": 57297 + }, + { + "epoch": 4.23, + "learning_rate": 4.210069301961841e-06, + "loss": 1.0538, + "step": 57298 + }, + { + "epoch": 4.23, + "learning_rate": 4.209744093587428e-06, + "loss": 0.9773, + "step": 57299 + }, + { + "epoch": 4.23, + "learning_rate": 4.209418894425233e-06, + "loss": 0.9554, + "step": 57300 + }, + { + "epoch": 4.23, + "learning_rate": 4.209093704475764e-06, + "loss": 1.011, + "step": 57301 + }, + { + "epoch": 4.23, + "learning_rate": 4.2087685237395535e-06, + "loss": 0.888, + "step": 57302 + }, + { + "epoch": 4.23, + "learning_rate": 4.208443352217101e-06, + "loss": 1.0057, + "step": 57303 + }, + { + "epoch": 4.23, + "learning_rate": 4.208118189908936e-06, + "loss": 1.0185, + "step": 57304 + }, + { + "epoch": 4.23, + "learning_rate": 4.207793036815573e-06, + "loss": 1.0845, + "step": 57305 + }, + { + "epoch": 4.23, + "learning_rate": 4.2074678929375265e-06, + "loss": 0.8778, + "step": 57306 + }, + { + "epoch": 4.23, + "learning_rate": 4.207142758275318e-06, + "loss": 1.0469, + "step": 57307 + }, + { + "epoch": 4.23, + "learning_rate": 4.2068176328294576e-06, + "loss": 0.9699, + "step": 57308 + }, + { + "epoch": 4.23, + "learning_rate": 4.206492516600471e-06, + "loss": 0.9904, + "step": 57309 + }, + { + "epoch": 4.23, + "learning_rate": 4.206167409588873e-06, + "loss": 1.0594, + "step": 57310 + }, + { + "epoch": 4.23, + "learning_rate": 4.205842311795179e-06, + "loss": 0.9474, + "step": 57311 + }, + { + "epoch": 4.23, + "learning_rate": 4.205517223219903e-06, + "loss": 1.0817, + "step": 57312 + }, + { + "epoch": 4.23, + "learning_rate": 4.20519214386357e-06, + "loss": 1.0421, + "step": 57313 + }, + { + "epoch": 4.23, + "learning_rate": 4.204867073726693e-06, + "loss": 0.9508, + "step": 57314 + }, + { + "epoch": 4.23, + "learning_rate": 4.20454201280979e-06, + "loss": 0.9151, + "step": 57315 + }, + { + "epoch": 4.24, + "learning_rate": 4.204216961113379e-06, + "loss": 1.0492, + "step": 57316 + }, + { + "epoch": 4.24, + "learning_rate": 4.20389191863797e-06, + "loss": 0.9455, + "step": 57317 + }, + { + "epoch": 4.24, + "learning_rate": 4.203566885384095e-06, + "loss": 0.9186, + "step": 57318 + }, + { + "epoch": 4.24, + "learning_rate": 4.203241861352254e-06, + "loss": 0.9497, + "step": 57319 + }, + { + "epoch": 4.24, + "learning_rate": 4.202916846542976e-06, + "loss": 0.9496, + "step": 57320 + }, + { + "epoch": 4.24, + "learning_rate": 4.202591840956771e-06, + "loss": 1.1402, + "step": 57321 + }, + { + "epoch": 4.24, + "learning_rate": 4.2022668445941665e-06, + "loss": 1.0444, + "step": 57322 + }, + { + "epoch": 4.24, + "learning_rate": 4.201941857455666e-06, + "loss": 1.0972, + "step": 57323 + }, + { + "epoch": 4.24, + "learning_rate": 4.201616879541796e-06, + "loss": 0.931, + "step": 57324 + }, + { + "epoch": 4.24, + "learning_rate": 4.20129191085307e-06, + "loss": 1.0347, + "step": 57325 + }, + { + "epoch": 4.24, + "learning_rate": 4.200966951390006e-06, + "loss": 0.9599, + "step": 57326 + }, + { + "epoch": 4.24, + "learning_rate": 4.200642001153121e-06, + "loss": 0.9332, + "step": 57327 + }, + { + "epoch": 4.24, + "learning_rate": 4.200317060142927e-06, + "loss": 1.0898, + "step": 57328 + }, + { + "epoch": 4.24, + "learning_rate": 4.199992128359951e-06, + "loss": 0.9482, + "step": 57329 + }, + { + "epoch": 4.24, + "learning_rate": 4.199667205804703e-06, + "loss": 1.0098, + "step": 57330 + }, + { + "epoch": 4.24, + "learning_rate": 4.1993422924777015e-06, + "loss": 0.965, + "step": 57331 + }, + { + "epoch": 4.24, + "learning_rate": 4.1990173883794605e-06, + "loss": 1.0499, + "step": 57332 + }, + { + "epoch": 4.24, + "learning_rate": 4.198692493510503e-06, + "loss": 0.9619, + "step": 57333 + }, + { + "epoch": 4.24, + "learning_rate": 4.198367607871343e-06, + "loss": 1.0105, + "step": 57334 + }, + { + "epoch": 4.24, + "learning_rate": 4.198042731462498e-06, + "loss": 0.904, + "step": 57335 + }, + { + "epoch": 4.24, + "learning_rate": 4.197717864284483e-06, + "loss": 1.0143, + "step": 57336 + }, + { + "epoch": 4.24, + "learning_rate": 4.197393006337812e-06, + "loss": 0.9731, + "step": 57337 + }, + { + "epoch": 4.24, + "learning_rate": 4.19706815762301e-06, + "loss": 0.9708, + "step": 57338 + }, + { + "epoch": 4.24, + "learning_rate": 4.1967433181405895e-06, + "loss": 0.9788, + "step": 57339 + }, + { + "epoch": 4.24, + "learning_rate": 4.196418487891066e-06, + "loss": 0.9063, + "step": 57340 + }, + { + "epoch": 4.24, + "learning_rate": 4.196093666874956e-06, + "loss": 0.9277, + "step": 57341 + }, + { + "epoch": 4.24, + "learning_rate": 4.19576885509278e-06, + "loss": 0.9512, + "step": 57342 + }, + { + "epoch": 4.24, + "learning_rate": 4.195444052545054e-06, + "loss": 1.1508, + "step": 57343 + }, + { + "epoch": 4.24, + "learning_rate": 4.195119259232293e-06, + "loss": 0.9976, + "step": 57344 + }, + { + "epoch": 4.24, + "learning_rate": 4.194794475155014e-06, + "loss": 0.8923, + "step": 57345 + }, + { + "epoch": 4.24, + "learning_rate": 4.194469700313733e-06, + "loss": 0.9848, + "step": 57346 + }, + { + "epoch": 4.24, + "learning_rate": 4.194144934708969e-06, + "loss": 0.9495, + "step": 57347 + }, + { + "epoch": 4.24, + "learning_rate": 4.193820178341234e-06, + "loss": 1.0207, + "step": 57348 + }, + { + "epoch": 4.24, + "learning_rate": 4.193495431211051e-06, + "loss": 0.9517, + "step": 57349 + }, + { + "epoch": 4.24, + "learning_rate": 4.193170693318933e-06, + "loss": 0.9687, + "step": 57350 + }, + { + "epoch": 4.24, + "learning_rate": 4.192845964665399e-06, + "loss": 0.9582, + "step": 57351 + }, + { + "epoch": 4.24, + "learning_rate": 4.192521245250959e-06, + "loss": 1.0591, + "step": 57352 + }, + { + "epoch": 4.24, + "learning_rate": 4.192196535076138e-06, + "loss": 1.0874, + "step": 57353 + }, + { + "epoch": 4.24, + "learning_rate": 4.19187183414145e-06, + "loss": 0.971, + "step": 57354 + }, + { + "epoch": 4.24, + "learning_rate": 4.191547142447411e-06, + "loss": 1.0054, + "step": 57355 + }, + { + "epoch": 4.24, + "learning_rate": 4.191222459994536e-06, + "loss": 1.0451, + "step": 57356 + }, + { + "epoch": 4.24, + "learning_rate": 4.190897786783341e-06, + "loss": 1.0203, + "step": 57357 + }, + { + "epoch": 4.24, + "learning_rate": 4.1905731228143475e-06, + "loss": 0.9899, + "step": 57358 + }, + { + "epoch": 4.24, + "learning_rate": 4.190248468088069e-06, + "loss": 1.0773, + "step": 57359 + }, + { + "epoch": 4.24, + "learning_rate": 4.189923822605022e-06, + "loss": 0.9208, + "step": 57360 + }, + { + "epoch": 4.24, + "learning_rate": 4.189599186365719e-06, + "loss": 0.9776, + "step": 57361 + }, + { + "epoch": 4.24, + "learning_rate": 4.189274559370685e-06, + "loss": 1.0508, + "step": 57362 + }, + { + "epoch": 4.24, + "learning_rate": 4.188949941620436e-06, + "loss": 0.973, + "step": 57363 + }, + { + "epoch": 4.24, + "learning_rate": 4.1886253331154755e-06, + "loss": 1.0126, + "step": 57364 + }, + { + "epoch": 4.24, + "learning_rate": 4.1883007338563344e-06, + "loss": 0.9985, + "step": 57365 + }, + { + "epoch": 4.24, + "learning_rate": 4.187976143843519e-06, + "loss": 0.9908, + "step": 57366 + }, + { + "epoch": 4.24, + "learning_rate": 4.187651563077558e-06, + "loss": 1.0398, + "step": 57367 + }, + { + "epoch": 4.24, + "learning_rate": 4.1873269915589535e-06, + "loss": 1.0126, + "step": 57368 + }, + { + "epoch": 4.24, + "learning_rate": 4.187002429288232e-06, + "loss": 0.9274, + "step": 57369 + }, + { + "epoch": 4.24, + "learning_rate": 4.186677876265902e-06, + "loss": 1.0719, + "step": 57370 + }, + { + "epoch": 4.24, + "learning_rate": 4.186353332492492e-06, + "loss": 0.9821, + "step": 57371 + }, + { + "epoch": 4.24, + "learning_rate": 4.186028797968508e-06, + "loss": 0.989, + "step": 57372 + }, + { + "epoch": 4.24, + "learning_rate": 4.185704272694464e-06, + "loss": 0.9756, + "step": 57373 + }, + { + "epoch": 4.24, + "learning_rate": 4.185379756670885e-06, + "loss": 1.0985, + "step": 57374 + }, + { + "epoch": 4.24, + "learning_rate": 4.185055249898284e-06, + "loss": 1.0009, + "step": 57375 + }, + { + "epoch": 4.24, + "learning_rate": 4.1847307523771775e-06, + "loss": 0.9384, + "step": 57376 + }, + { + "epoch": 4.24, + "learning_rate": 4.184406264108076e-06, + "loss": 0.8818, + "step": 57377 + }, + { + "epoch": 4.24, + "learning_rate": 4.1840817850915064e-06, + "loss": 0.9611, + "step": 57378 + }, + { + "epoch": 4.24, + "learning_rate": 4.183757315327979e-06, + "loss": 0.9246, + "step": 57379 + }, + { + "epoch": 4.24, + "learning_rate": 4.183432854818009e-06, + "loss": 0.9497, + "step": 57380 + }, + { + "epoch": 4.24, + "learning_rate": 4.183108403562116e-06, + "loss": 0.9243, + "step": 57381 + }, + { + "epoch": 4.24, + "learning_rate": 4.18278396156081e-06, + "loss": 0.9138, + "step": 57382 + }, + { + "epoch": 4.24, + "learning_rate": 4.182459528814614e-06, + "loss": 0.9745, + "step": 57383 + }, + { + "epoch": 4.24, + "learning_rate": 4.182135105324043e-06, + "loss": 1.0131, + "step": 57384 + }, + { + "epoch": 4.24, + "learning_rate": 4.181810691089611e-06, + "loss": 0.9667, + "step": 57385 + }, + { + "epoch": 4.24, + "learning_rate": 4.181486286111832e-06, + "loss": 1.0869, + "step": 57386 + }, + { + "epoch": 4.24, + "learning_rate": 4.181161890391229e-06, + "loss": 1.0696, + "step": 57387 + }, + { + "epoch": 4.24, + "learning_rate": 4.1808375039283134e-06, + "loss": 0.9545, + "step": 57388 + }, + { + "epoch": 4.24, + "learning_rate": 4.180513126723602e-06, + "loss": 1.0056, + "step": 57389 + }, + { + "epoch": 4.24, + "learning_rate": 4.180188758777612e-06, + "loss": 1.0933, + "step": 57390 + }, + { + "epoch": 4.24, + "learning_rate": 4.179864400090857e-06, + "loss": 1.1344, + "step": 57391 + }, + { + "epoch": 4.24, + "learning_rate": 4.179540050663856e-06, + "loss": 0.9637, + "step": 57392 + }, + { + "epoch": 4.24, + "learning_rate": 4.179215710497119e-06, + "loss": 0.9549, + "step": 57393 + }, + { + "epoch": 4.24, + "learning_rate": 4.17889137959117e-06, + "loss": 1.054, + "step": 57394 + }, + { + "epoch": 4.24, + "learning_rate": 4.178567057946522e-06, + "loss": 0.9031, + "step": 57395 + }, + { + "epoch": 4.24, + "learning_rate": 4.17824274556369e-06, + "loss": 0.8939, + "step": 57396 + }, + { + "epoch": 4.24, + "learning_rate": 4.177918442443187e-06, + "loss": 1.0158, + "step": 57397 + }, + { + "epoch": 4.24, + "learning_rate": 4.177594148585536e-06, + "loss": 1.0088, + "step": 57398 + }, + { + "epoch": 4.24, + "learning_rate": 4.17726986399125e-06, + "loss": 0.9908, + "step": 57399 + }, + { + "epoch": 4.24, + "learning_rate": 4.176945588660844e-06, + "loss": 1.0128, + "step": 57400 + }, + { + "epoch": 4.24, + "learning_rate": 4.176621322594834e-06, + "loss": 0.9597, + "step": 57401 + }, + { + "epoch": 4.24, + "learning_rate": 4.1762970657937325e-06, + "loss": 0.9504, + "step": 57402 + }, + { + "epoch": 4.24, + "learning_rate": 4.175972818258063e-06, + "loss": 1.0955, + "step": 57403 + }, + { + "epoch": 4.24, + "learning_rate": 4.175648579988337e-06, + "loss": 1.0415, + "step": 57404 + }, + { + "epoch": 4.24, + "learning_rate": 4.1753243509850715e-06, + "loss": 0.944, + "step": 57405 + }, + { + "epoch": 4.24, + "learning_rate": 4.175000131248777e-06, + "loss": 0.955, + "step": 57406 + }, + { + "epoch": 4.24, + "learning_rate": 4.174675920779979e-06, + "loss": 0.9225, + "step": 57407 + }, + { + "epoch": 4.24, + "learning_rate": 4.174351719579187e-06, + "loss": 1.0581, + "step": 57408 + }, + { + "epoch": 4.24, + "learning_rate": 4.174027527646919e-06, + "loss": 0.9899, + "step": 57409 + }, + { + "epoch": 4.24, + "learning_rate": 4.17370334498369e-06, + "loss": 0.9637, + "step": 57410 + }, + { + "epoch": 4.24, + "learning_rate": 4.173379171590011e-06, + "loss": 0.9392, + "step": 57411 + }, + { + "epoch": 4.24, + "learning_rate": 4.173055007466409e-06, + "loss": 1.0143, + "step": 57412 + }, + { + "epoch": 4.24, + "learning_rate": 4.172730852613387e-06, + "loss": 1.0319, + "step": 57413 + }, + { + "epoch": 4.24, + "learning_rate": 4.17240670703147e-06, + "loss": 1.0221, + "step": 57414 + }, + { + "epoch": 4.24, + "learning_rate": 4.172082570721168e-06, + "loss": 0.8458, + "step": 57415 + }, + { + "epoch": 4.24, + "learning_rate": 4.171758443683005e-06, + "loss": 1.0469, + "step": 57416 + }, + { + "epoch": 4.24, + "learning_rate": 4.171434325917485e-06, + "loss": 1.0072, + "step": 57417 + }, + { + "epoch": 4.24, + "learning_rate": 4.1711102174251325e-06, + "loss": 1.0655, + "step": 57418 + }, + { + "epoch": 4.24, + "learning_rate": 4.17078611820646e-06, + "loss": 0.953, + "step": 57419 + }, + { + "epoch": 4.24, + "learning_rate": 4.170462028261983e-06, + "loss": 1.0409, + "step": 57420 + }, + { + "epoch": 4.24, + "learning_rate": 4.170137947592219e-06, + "loss": 1.0211, + "step": 57421 + }, + { + "epoch": 4.24, + "learning_rate": 4.169813876197678e-06, + "loss": 1.0503, + "step": 57422 + }, + { + "epoch": 4.24, + "learning_rate": 4.1694898140788835e-06, + "loss": 0.9821, + "step": 57423 + }, + { + "epoch": 4.24, + "learning_rate": 4.169165761236346e-06, + "loss": 0.9336, + "step": 57424 + }, + { + "epoch": 4.24, + "learning_rate": 4.168841717670583e-06, + "loss": 1.0274, + "step": 57425 + }, + { + "epoch": 4.24, + "learning_rate": 4.168517683382106e-06, + "loss": 1.0811, + "step": 57426 + }, + { + "epoch": 4.24, + "learning_rate": 4.168193658371437e-06, + "loss": 0.9623, + "step": 57427 + }, + { + "epoch": 4.24, + "learning_rate": 4.16786964263909e-06, + "loss": 0.8924, + "step": 57428 + }, + { + "epoch": 4.24, + "learning_rate": 4.1675456361855784e-06, + "loss": 0.9188, + "step": 57429 + }, + { + "epoch": 4.24, + "learning_rate": 4.167221639011417e-06, + "loss": 0.913, + "step": 57430 + }, + { + "epoch": 4.24, + "learning_rate": 4.16689765111712e-06, + "loss": 1.0207, + "step": 57431 + }, + { + "epoch": 4.24, + "learning_rate": 4.166573672503208e-06, + "loss": 1.0789, + "step": 57432 + }, + { + "epoch": 4.24, + "learning_rate": 4.166249703170196e-06, + "loss": 0.991, + "step": 57433 + }, + { + "epoch": 4.24, + "learning_rate": 4.165925743118595e-06, + "loss": 1.0238, + "step": 57434 + }, + { + "epoch": 4.24, + "learning_rate": 4.165601792348919e-06, + "loss": 0.915, + "step": 57435 + }, + { + "epoch": 4.24, + "learning_rate": 4.165277850861696e-06, + "loss": 0.9305, + "step": 57436 + }, + { + "epoch": 4.24, + "learning_rate": 4.164953918657427e-06, + "loss": 1.0057, + "step": 57437 + }, + { + "epoch": 4.24, + "learning_rate": 4.164629995736631e-06, + "loss": 0.9462, + "step": 57438 + }, + { + "epoch": 4.24, + "learning_rate": 4.164306082099827e-06, + "loss": 0.9192, + "step": 57439 + }, + { + "epoch": 4.24, + "learning_rate": 4.16398217774753e-06, + "loss": 1.0572, + "step": 57440 + }, + { + "epoch": 4.24, + "learning_rate": 4.163658282680253e-06, + "loss": 0.9388, + "step": 57441 + }, + { + "epoch": 4.24, + "learning_rate": 4.16333439689851e-06, + "loss": 0.9436, + "step": 57442 + }, + { + "epoch": 4.24, + "learning_rate": 4.163010520402821e-06, + "loss": 1.0505, + "step": 57443 + }, + { + "epoch": 4.24, + "learning_rate": 4.162686653193698e-06, + "loss": 1.1095, + "step": 57444 + }, + { + "epoch": 4.24, + "learning_rate": 4.162362795271657e-06, + "loss": 0.9762, + "step": 57445 + }, + { + "epoch": 4.24, + "learning_rate": 4.162038946637212e-06, + "loss": 1.0316, + "step": 57446 + }, + { + "epoch": 4.24, + "learning_rate": 4.16171510729088e-06, + "loss": 0.9648, + "step": 57447 + }, + { + "epoch": 4.24, + "learning_rate": 4.161391277233178e-06, + "loss": 1.0692, + "step": 57448 + }, + { + "epoch": 4.24, + "learning_rate": 4.161067456464618e-06, + "loss": 0.9835, + "step": 57449 + }, + { + "epoch": 4.24, + "learning_rate": 4.160743644985716e-06, + "loss": 0.9945, + "step": 57450 + }, + { + "epoch": 4.25, + "learning_rate": 4.160419842796984e-06, + "loss": 0.9475, + "step": 57451 + }, + { + "epoch": 4.25, + "learning_rate": 4.160096049898944e-06, + "loss": 1.0234, + "step": 57452 + }, + { + "epoch": 4.25, + "learning_rate": 4.159772266292108e-06, + "loss": 0.9894, + "step": 57453 + }, + { + "epoch": 4.25, + "learning_rate": 4.159448491976991e-06, + "loss": 0.947, + "step": 57454 + }, + { + "epoch": 4.25, + "learning_rate": 4.159124726954104e-06, + "loss": 0.9784, + "step": 57455 + }, + { + "epoch": 4.25, + "learning_rate": 4.158800971223968e-06, + "loss": 1.1055, + "step": 57456 + }, + { + "epoch": 4.25, + "learning_rate": 4.158477224787102e-06, + "loss": 0.897, + "step": 57457 + }, + { + "epoch": 4.25, + "learning_rate": 4.1581534876440066e-06, + "loss": 0.9912, + "step": 57458 + }, + { + "epoch": 4.25, + "learning_rate": 4.15782975979521e-06, + "loss": 0.9868, + "step": 57459 + }, + { + "epoch": 4.25, + "learning_rate": 4.157506041241219e-06, + "loss": 0.9844, + "step": 57460 + }, + { + "epoch": 4.25, + "learning_rate": 4.157182331982559e-06, + "loss": 1.0338, + "step": 57461 + }, + { + "epoch": 4.25, + "learning_rate": 4.15685863201973e-06, + "loss": 1.0297, + "step": 57462 + }, + { + "epoch": 4.25, + "learning_rate": 4.1565349413532605e-06, + "loss": 1.0657, + "step": 57463 + }, + { + "epoch": 4.25, + "learning_rate": 4.1562112599836555e-06, + "loss": 0.9656, + "step": 57464 + }, + { + "epoch": 4.25, + "learning_rate": 4.155887587911443e-06, + "loss": 1.0202, + "step": 57465 + }, + { + "epoch": 4.25, + "learning_rate": 4.155563925137126e-06, + "loss": 1.1014, + "step": 57466 + }, + { + "epoch": 4.25, + "learning_rate": 4.15524027166122e-06, + "loss": 1.0496, + "step": 57467 + }, + { + "epoch": 4.25, + "learning_rate": 4.1549166274842454e-06, + "loss": 0.9405, + "step": 57468 + }, + { + "epoch": 4.25, + "learning_rate": 4.154592992606716e-06, + "loss": 1.0536, + "step": 57469 + }, + { + "epoch": 4.25, + "learning_rate": 4.154269367029144e-06, + "loss": 1.0106, + "step": 57470 + }, + { + "epoch": 4.25, + "learning_rate": 4.1539457507520444e-06, + "loss": 1.0009, + "step": 57471 + }, + { + "epoch": 4.25, + "learning_rate": 4.153622143775936e-06, + "loss": 1.058, + "step": 57472 + }, + { + "epoch": 4.25, + "learning_rate": 4.153298546101331e-06, + "loss": 1.07, + "step": 57473 + }, + { + "epoch": 4.25, + "learning_rate": 4.1529749577287435e-06, + "loss": 1.0307, + "step": 57474 + }, + { + "epoch": 4.25, + "learning_rate": 4.152651378658691e-06, + "loss": 1.073, + "step": 57475 + }, + { + "epoch": 4.25, + "learning_rate": 4.152327808891682e-06, + "loss": 1.0054, + "step": 57476 + }, + { + "epoch": 4.25, + "learning_rate": 4.152004248428238e-06, + "loss": 0.9463, + "step": 57477 + }, + { + "epoch": 4.25, + "learning_rate": 4.151680697268874e-06, + "loss": 0.9491, + "step": 57478 + }, + { + "epoch": 4.25, + "learning_rate": 4.151357155414101e-06, + "loss": 1.1014, + "step": 57479 + }, + { + "epoch": 4.25, + "learning_rate": 4.151033622864432e-06, + "loss": 0.9849, + "step": 57480 + }, + { + "epoch": 4.25, + "learning_rate": 4.150710099620392e-06, + "loss": 1.0273, + "step": 57481 + }, + { + "epoch": 4.25, + "learning_rate": 4.150386585682481e-06, + "loss": 0.9932, + "step": 57482 + }, + { + "epoch": 4.25, + "learning_rate": 4.150063081051224e-06, + "loss": 1.0541, + "step": 57483 + }, + { + "epoch": 4.25, + "learning_rate": 4.149739585727135e-06, + "loss": 0.9798, + "step": 57484 + }, + { + "epoch": 4.25, + "learning_rate": 4.1494160997107255e-06, + "loss": 1.1112, + "step": 57485 + }, + { + "epoch": 4.25, + "learning_rate": 4.1490926230025104e-06, + "loss": 0.8778, + "step": 57486 + }, + { + "epoch": 4.25, + "learning_rate": 4.148769155603003e-06, + "loss": 0.9988, + "step": 57487 + }, + { + "epoch": 4.25, + "learning_rate": 4.148445697512723e-06, + "loss": 0.9702, + "step": 57488 + }, + { + "epoch": 4.25, + "learning_rate": 4.1481222487321824e-06, + "loss": 1.0272, + "step": 57489 + }, + { + "epoch": 4.25, + "learning_rate": 4.1477988092618955e-06, + "loss": 0.9195, + "step": 57490 + }, + { + "epoch": 4.25, + "learning_rate": 4.1474753791023735e-06, + "loss": 1.1235, + "step": 57491 + }, + { + "epoch": 4.25, + "learning_rate": 4.147151958254139e-06, + "loss": 1.1164, + "step": 57492 + }, + { + "epoch": 4.25, + "learning_rate": 4.1468285467177e-06, + "loss": 0.949, + "step": 57493 + }, + { + "epoch": 4.25, + "learning_rate": 4.146505144493575e-06, + "loss": 0.9327, + "step": 57494 + }, + { + "epoch": 4.25, + "learning_rate": 4.146181751582276e-06, + "loss": 0.9555, + "step": 57495 + }, + { + "epoch": 4.25, + "learning_rate": 4.145858367984313e-06, + "loss": 1.077, + "step": 57496 + }, + { + "epoch": 4.25, + "learning_rate": 4.145534993700211e-06, + "loss": 0.9896, + "step": 57497 + }, + { + "epoch": 4.25, + "learning_rate": 4.145211628730478e-06, + "loss": 1.0273, + "step": 57498 + }, + { + "epoch": 4.25, + "learning_rate": 4.14488827307563e-06, + "loss": 1.0903, + "step": 57499 + }, + { + "epoch": 4.25, + "learning_rate": 4.144564926736178e-06, + "loss": 1.0013, + "step": 57500 + }, + { + "epoch": 4.25, + "learning_rate": 4.144241589712642e-06, + "loss": 1.0101, + "step": 57501 + }, + { + "epoch": 4.25, + "learning_rate": 4.143918262005534e-06, + "loss": 1.014, + "step": 57502 + }, + { + "epoch": 4.25, + "learning_rate": 4.143594943615368e-06, + "loss": 1.0365, + "step": 57503 + }, + { + "epoch": 4.25, + "learning_rate": 4.143271634542659e-06, + "loss": 1.0623, + "step": 57504 + }, + { + "epoch": 4.25, + "learning_rate": 4.142948334787918e-06, + "loss": 1.0125, + "step": 57505 + }, + { + "epoch": 4.25, + "learning_rate": 4.142625044351668e-06, + "loss": 0.963, + "step": 57506 + }, + { + "epoch": 4.25, + "learning_rate": 4.142301763234412e-06, + "loss": 1.1003, + "step": 57507 + }, + { + "epoch": 4.25, + "learning_rate": 4.141978491436674e-06, + "loss": 0.9943, + "step": 57508 + }, + { + "epoch": 4.25, + "learning_rate": 4.141655228958959e-06, + "loss": 0.9008, + "step": 57509 + }, + { + "epoch": 4.25, + "learning_rate": 4.1413319758017956e-06, + "loss": 1.0309, + "step": 57510 + }, + { + "epoch": 4.25, + "learning_rate": 4.141008731965681e-06, + "loss": 1.0093, + "step": 57511 + }, + { + "epoch": 4.25, + "learning_rate": 4.140685497451143e-06, + "loss": 1.1488, + "step": 57512 + }, + { + "epoch": 4.25, + "learning_rate": 4.140362272258689e-06, + "loss": 0.8844, + "step": 57513 + }, + { + "epoch": 4.25, + "learning_rate": 4.140039056388835e-06, + "loss": 0.9604, + "step": 57514 + }, + { + "epoch": 4.25, + "learning_rate": 4.1397158498420945e-06, + "loss": 1.0628, + "step": 57515 + }, + { + "epoch": 4.25, + "learning_rate": 4.139392652618978e-06, + "loss": 1.0305, + "step": 57516 + }, + { + "epoch": 4.25, + "learning_rate": 4.13906946472001e-06, + "loss": 1.0499, + "step": 57517 + }, + { + "epoch": 4.25, + "learning_rate": 4.1387462861456976e-06, + "loss": 0.9976, + "step": 57518 + }, + { + "epoch": 4.25, + "learning_rate": 4.138423116896555e-06, + "loss": 1.1207, + "step": 57519 + }, + { + "epoch": 4.25, + "learning_rate": 4.138099956973095e-06, + "loss": 0.9275, + "step": 57520 + }, + { + "epoch": 4.25, + "learning_rate": 4.137776806375837e-06, + "loss": 1.027, + "step": 57521 + }, + { + "epoch": 4.25, + "learning_rate": 4.137453665105292e-06, + "loss": 1.0454, + "step": 57522 + }, + { + "epoch": 4.25, + "learning_rate": 4.137130533161975e-06, + "loss": 1.056, + "step": 57523 + }, + { + "epoch": 4.25, + "learning_rate": 4.136807410546399e-06, + "loss": 0.9548, + "step": 57524 + }, + { + "epoch": 4.25, + "learning_rate": 4.136484297259075e-06, + "loss": 1.0293, + "step": 57525 + }, + { + "epoch": 4.25, + "learning_rate": 4.136161193300529e-06, + "loss": 1.0379, + "step": 57526 + }, + { + "epoch": 4.25, + "learning_rate": 4.135838098671258e-06, + "loss": 0.9112, + "step": 57527 + }, + { + "epoch": 4.25, + "learning_rate": 4.135515013371788e-06, + "loss": 0.9715, + "step": 57528 + }, + { + "epoch": 4.25, + "learning_rate": 4.1351919374026275e-06, + "loss": 1.04, + "step": 57529 + }, + { + "epoch": 4.25, + "learning_rate": 4.134868870764299e-06, + "loss": 1.1019, + "step": 57530 + }, + { + "epoch": 4.25, + "learning_rate": 4.134545813457307e-06, + "loss": 0.8804, + "step": 57531 + }, + { + "epoch": 4.25, + "learning_rate": 4.134222765482165e-06, + "loss": 0.9469, + "step": 57532 + }, + { + "epoch": 4.25, + "learning_rate": 4.133899726839393e-06, + "loss": 1.1639, + "step": 57533 + }, + { + "epoch": 4.25, + "learning_rate": 4.133576697529504e-06, + "loss": 1.1053, + "step": 57534 + }, + { + "epoch": 4.25, + "learning_rate": 4.133253677553011e-06, + "loss": 1.0958, + "step": 57535 + }, + { + "epoch": 4.25, + "learning_rate": 4.132930666910423e-06, + "loss": 0.9368, + "step": 57536 + }, + { + "epoch": 4.25, + "learning_rate": 4.132607665602262e-06, + "loss": 0.9142, + "step": 57537 + }, + { + "epoch": 4.25, + "learning_rate": 4.132284673629038e-06, + "loss": 1.0171, + "step": 57538 + }, + { + "epoch": 4.25, + "learning_rate": 4.131961690991265e-06, + "loss": 0.8841, + "step": 57539 + }, + { + "epoch": 4.25, + "learning_rate": 4.131638717689458e-06, + "loss": 0.9855, + "step": 57540 + }, + { + "epoch": 4.25, + "learning_rate": 4.1313157537241255e-06, + "loss": 0.999, + "step": 57541 + }, + { + "epoch": 4.25, + "learning_rate": 4.13099279909579e-06, + "loss": 1.0464, + "step": 57542 + }, + { + "epoch": 4.25, + "learning_rate": 4.130669853804961e-06, + "loss": 0.9284, + "step": 57543 + }, + { + "epoch": 4.25, + "learning_rate": 4.130346917852151e-06, + "loss": 0.8964, + "step": 57544 + }, + { + "epoch": 4.25, + "learning_rate": 4.130023991237873e-06, + "loss": 0.94, + "step": 57545 + }, + { + "epoch": 4.25, + "learning_rate": 4.129701073962645e-06, + "loss": 0.8666, + "step": 57546 + }, + { + "epoch": 4.25, + "learning_rate": 4.129378166026979e-06, + "loss": 1.0497, + "step": 57547 + }, + { + "epoch": 4.25, + "learning_rate": 4.129055267431389e-06, + "loss": 1.0552, + "step": 57548 + }, + { + "epoch": 4.25, + "learning_rate": 4.128732378176388e-06, + "loss": 1.0706, + "step": 57549 + }, + { + "epoch": 4.25, + "learning_rate": 4.128409498262486e-06, + "loss": 1.0035, + "step": 57550 + }, + { + "epoch": 4.25, + "learning_rate": 4.128086627690207e-06, + "loss": 1.0247, + "step": 57551 + }, + { + "epoch": 4.25, + "learning_rate": 4.127763766460051e-06, + "loss": 0.9766, + "step": 57552 + }, + { + "epoch": 4.25, + "learning_rate": 4.127440914572544e-06, + "loss": 0.9983, + "step": 57553 + }, + { + "epoch": 4.25, + "learning_rate": 4.127118072028189e-06, + "loss": 0.9475, + "step": 57554 + }, + { + "epoch": 4.25, + "learning_rate": 4.126795238827514e-06, + "loss": 1.0288, + "step": 57555 + }, + { + "epoch": 4.25, + "learning_rate": 4.126472414971014e-06, + "loss": 1.0912, + "step": 57556 + }, + { + "epoch": 4.25, + "learning_rate": 4.126149600459219e-06, + "loss": 0.9505, + "step": 57557 + }, + { + "epoch": 4.25, + "learning_rate": 4.125826795292635e-06, + "loss": 0.9745, + "step": 57558 + }, + { + "epoch": 4.25, + "learning_rate": 4.125503999471775e-06, + "loss": 0.9405, + "step": 57559 + }, + { + "epoch": 4.25, + "learning_rate": 4.125181212997156e-06, + "loss": 1.049, + "step": 57560 + }, + { + "epoch": 4.25, + "learning_rate": 4.124858435869285e-06, + "loss": 0.9781, + "step": 57561 + }, + { + "epoch": 4.25, + "learning_rate": 4.124535668088684e-06, + "loss": 1.0228, + "step": 57562 + }, + { + "epoch": 4.25, + "learning_rate": 4.124212909655862e-06, + "loss": 0.9083, + "step": 57563 + }, + { + "epoch": 4.25, + "learning_rate": 4.123890160571334e-06, + "loss": 1.004, + "step": 57564 + }, + { + "epoch": 4.25, + "learning_rate": 4.123567420835609e-06, + "loss": 0.9612, + "step": 57565 + }, + { + "epoch": 4.25, + "learning_rate": 4.123244690449208e-06, + "loss": 0.9994, + "step": 57566 + }, + { + "epoch": 4.25, + "learning_rate": 4.12292196941264e-06, + "loss": 1.0879, + "step": 57567 + }, + { + "epoch": 4.25, + "learning_rate": 4.122599257726419e-06, + "loss": 1.0388, + "step": 57568 + }, + { + "epoch": 4.25, + "learning_rate": 4.122276555391059e-06, + "loss": 0.9392, + "step": 57569 + }, + { + "epoch": 4.25, + "learning_rate": 4.12195386240707e-06, + "loss": 0.9671, + "step": 57570 + }, + { + "epoch": 4.25, + "learning_rate": 4.121631178774974e-06, + "loss": 0.8943, + "step": 57571 + }, + { + "epoch": 4.25, + "learning_rate": 4.121308504495273e-06, + "loss": 1.0396, + "step": 57572 + }, + { + "epoch": 4.25, + "learning_rate": 4.120985839568488e-06, + "loss": 0.9808, + "step": 57573 + }, + { + "epoch": 4.25, + "learning_rate": 4.1206631839951275e-06, + "loss": 0.9985, + "step": 57574 + }, + { + "epoch": 4.25, + "learning_rate": 4.1203405377757155e-06, + "loss": 0.9113, + "step": 57575 + }, + { + "epoch": 4.25, + "learning_rate": 4.12001790091075e-06, + "loss": 1.0016, + "step": 57576 + }, + { + "epoch": 4.25, + "learning_rate": 4.119695273400756e-06, + "loss": 0.8996, + "step": 57577 + }, + { + "epoch": 4.25, + "learning_rate": 4.119372655246241e-06, + "loss": 0.9026, + "step": 57578 + }, + { + "epoch": 4.25, + "learning_rate": 4.119050046447722e-06, + "loss": 0.832, + "step": 57579 + }, + { + "epoch": 4.25, + "learning_rate": 4.118727447005709e-06, + "loss": 1.029, + "step": 57580 + }, + { + "epoch": 4.25, + "learning_rate": 4.118404856920714e-06, + "loss": 1.0373, + "step": 57581 + }, + { + "epoch": 4.25, + "learning_rate": 4.118082276193255e-06, + "loss": 0.931, + "step": 57582 + }, + { + "epoch": 4.25, + "learning_rate": 4.117759704823843e-06, + "loss": 1.0476, + "step": 57583 + }, + { + "epoch": 4.25, + "learning_rate": 4.117437142812992e-06, + "loss": 0.987, + "step": 57584 + }, + { + "epoch": 4.25, + "learning_rate": 4.1171145901612096e-06, + "loss": 0.9451, + "step": 57585 + }, + { + "epoch": 4.25, + "learning_rate": 4.116792046869019e-06, + "loss": 0.9636, + "step": 57586 + }, + { + "epoch": 4.26, + "learning_rate": 4.116469512936927e-06, + "loss": 0.9795, + "step": 57587 + }, + { + "epoch": 4.26, + "learning_rate": 4.116146988365448e-06, + "loss": 0.9841, + "step": 57588 + }, + { + "epoch": 4.26, + "learning_rate": 4.115824473155095e-06, + "loss": 0.9441, + "step": 57589 + }, + { + "epoch": 4.26, + "learning_rate": 4.115501967306377e-06, + "loss": 0.97, + "step": 57590 + }, + { + "epoch": 4.26, + "learning_rate": 4.1151794708198155e-06, + "loss": 0.9244, + "step": 57591 + }, + { + "epoch": 4.26, + "learning_rate": 4.1148569836959186e-06, + "loss": 1.0395, + "step": 57592 + }, + { + "epoch": 4.26, + "learning_rate": 4.1145345059352e-06, + "loss": 1.0583, + "step": 57593 + }, + { + "epoch": 4.26, + "learning_rate": 4.11421203753817e-06, + "loss": 0.9748, + "step": 57594 + }, + { + "epoch": 4.26, + "learning_rate": 4.113889578505348e-06, + "loss": 0.9971, + "step": 57595 + }, + { + "epoch": 4.26, + "learning_rate": 4.113567128837247e-06, + "loss": 0.8782, + "step": 57596 + }, + { + "epoch": 4.26, + "learning_rate": 4.113244688534369e-06, + "loss": 0.9727, + "step": 57597 + }, + { + "epoch": 4.26, + "learning_rate": 4.1129222575972385e-06, + "loss": 1.0662, + "step": 57598 + }, + { + "epoch": 4.26, + "learning_rate": 4.11259983602636e-06, + "loss": 1.0275, + "step": 57599 + }, + { + "epoch": 4.26, + "learning_rate": 4.1122774238222584e-06, + "loss": 0.9375, + "step": 57600 + }, + { + "epoch": 4.26, + "learning_rate": 4.111955020985432e-06, + "loss": 0.9605, + "step": 57601 + }, + { + "epoch": 4.26, + "learning_rate": 4.111632627516404e-06, + "loss": 0.9289, + "step": 57602 + }, + { + "epoch": 4.26, + "learning_rate": 4.1113102434156856e-06, + "loss": 1.0187, + "step": 57603 + }, + { + "epoch": 4.26, + "learning_rate": 4.110987868683788e-06, + "loss": 0.9855, + "step": 57604 + }, + { + "epoch": 4.26, + "learning_rate": 4.110665503321224e-06, + "loss": 0.9646, + "step": 57605 + }, + { + "epoch": 4.26, + "learning_rate": 4.1103431473285024e-06, + "loss": 0.9745, + "step": 57606 + }, + { + "epoch": 4.26, + "learning_rate": 4.110020800706146e-06, + "loss": 1.0619, + "step": 57607 + }, + { + "epoch": 4.26, + "learning_rate": 4.109698463454661e-06, + "loss": 0.9326, + "step": 57608 + }, + { + "epoch": 4.26, + "learning_rate": 4.109376135574562e-06, + "loss": 0.9952, + "step": 57609 + }, + { + "epoch": 4.26, + "learning_rate": 4.109053817066357e-06, + "loss": 0.9496, + "step": 57610 + }, + { + "epoch": 4.26, + "learning_rate": 4.1087315079305675e-06, + "loss": 0.9884, + "step": 57611 + }, + { + "epoch": 4.26, + "learning_rate": 4.108409208167702e-06, + "loss": 1.0687, + "step": 57612 + }, + { + "epoch": 4.26, + "learning_rate": 4.108086917778274e-06, + "loss": 0.9507, + "step": 57613 + }, + { + "epoch": 4.26, + "learning_rate": 4.1077646367627895e-06, + "loss": 1.0546, + "step": 57614 + }, + { + "epoch": 4.26, + "learning_rate": 4.107442365121772e-06, + "loss": 0.9989, + "step": 57615 + }, + { + "epoch": 4.26, + "learning_rate": 4.107120102855733e-06, + "loss": 1.0775, + "step": 57616 + }, + { + "epoch": 4.26, + "learning_rate": 4.106797849965174e-06, + "loss": 1.0379, + "step": 57617 + }, + { + "epoch": 4.26, + "learning_rate": 4.106475606450619e-06, + "loss": 0.8468, + "step": 57618 + }, + { + "epoch": 4.26, + "learning_rate": 4.106153372312573e-06, + "loss": 0.9509, + "step": 57619 + }, + { + "epoch": 4.26, + "learning_rate": 4.105831147551561e-06, + "loss": 0.9677, + "step": 57620 + }, + { + "epoch": 4.26, + "learning_rate": 4.10550893216808e-06, + "loss": 0.9043, + "step": 57621 + }, + { + "epoch": 4.26, + "learning_rate": 4.105186726162654e-06, + "loss": 1.0399, + "step": 57622 + }, + { + "epoch": 4.26, + "learning_rate": 4.1048645295357854e-06, + "loss": 0.946, + "step": 57623 + }, + { + "epoch": 4.26, + "learning_rate": 4.104542342288003e-06, + "loss": 0.9972, + "step": 57624 + }, + { + "epoch": 4.26, + "learning_rate": 4.104220164419804e-06, + "loss": 0.9059, + "step": 57625 + }, + { + "epoch": 4.26, + "learning_rate": 4.1038979959317035e-06, + "loss": 0.951, + "step": 57626 + }, + { + "epoch": 4.26, + "learning_rate": 4.103575836824221e-06, + "loss": 1.0025, + "step": 57627 + }, + { + "epoch": 4.26, + "learning_rate": 4.103253687097863e-06, + "loss": 1.0251, + "step": 57628 + }, + { + "epoch": 4.26, + "learning_rate": 4.102931546753146e-06, + "loss": 1.009, + "step": 57629 + }, + { + "epoch": 4.26, + "learning_rate": 4.102609415790576e-06, + "loss": 1.0044, + "step": 57630 + }, + { + "epoch": 4.26, + "learning_rate": 4.1022872942106736e-06, + "loss": 0.9337, + "step": 57631 + }, + { + "epoch": 4.26, + "learning_rate": 4.101965182013948e-06, + "loss": 1.0162, + "step": 57632 + }, + { + "epoch": 4.26, + "learning_rate": 4.101643079200912e-06, + "loss": 0.899, + "step": 57633 + }, + { + "epoch": 4.26, + "learning_rate": 4.101320985772077e-06, + "loss": 0.98, + "step": 57634 + }, + { + "epoch": 4.26, + "learning_rate": 4.100998901727953e-06, + "loss": 1.0507, + "step": 57635 + }, + { + "epoch": 4.26, + "learning_rate": 4.100676827069058e-06, + "loss": 1.0403, + "step": 57636 + }, + { + "epoch": 4.26, + "learning_rate": 4.100354761795901e-06, + "loss": 0.9205, + "step": 57637 + }, + { + "epoch": 4.26, + "learning_rate": 4.100032705908997e-06, + "loss": 1.0093, + "step": 57638 + }, + { + "epoch": 4.26, + "learning_rate": 4.099710659408852e-06, + "loss": 0.9546, + "step": 57639 + }, + { + "epoch": 4.26, + "learning_rate": 4.099388622295987e-06, + "loss": 0.9072, + "step": 57640 + }, + { + "epoch": 4.26, + "learning_rate": 4.0990665945709095e-06, + "loss": 0.9179, + "step": 57641 + }, + { + "epoch": 4.26, + "learning_rate": 4.098744576234134e-06, + "loss": 0.9893, + "step": 57642 + }, + { + "epoch": 4.26, + "learning_rate": 4.09842256728617e-06, + "loss": 0.9956, + "step": 57643 + }, + { + "epoch": 4.26, + "learning_rate": 4.098100567727529e-06, + "loss": 0.8885, + "step": 57644 + }, + { + "epoch": 4.26, + "learning_rate": 4.0977785775587305e-06, + "loss": 0.9506, + "step": 57645 + }, + { + "epoch": 4.26, + "learning_rate": 4.097456596780277e-06, + "loss": 0.8374, + "step": 57646 + }, + { + "epoch": 4.26, + "learning_rate": 4.097134625392688e-06, + "loss": 1.0213, + "step": 57647 + }, + { + "epoch": 4.26, + "learning_rate": 4.0968126633964734e-06, + "loss": 0.8949, + "step": 57648 + }, + { + "epoch": 4.26, + "learning_rate": 4.096490710792146e-06, + "loss": 0.9609, + "step": 57649 + }, + { + "epoch": 4.26, + "learning_rate": 4.096168767580213e-06, + "loss": 0.9298, + "step": 57650 + }, + { + "epoch": 4.26, + "learning_rate": 4.095846833761195e-06, + "loss": 1.0069, + "step": 57651 + }, + { + "epoch": 4.26, + "learning_rate": 4.095524909335601e-06, + "loss": 1.0326, + "step": 57652 + }, + { + "epoch": 4.26, + "learning_rate": 4.095202994303941e-06, + "loss": 1.0262, + "step": 57653 + }, + { + "epoch": 4.26, + "learning_rate": 4.094881088666729e-06, + "loss": 0.9572, + "step": 57654 + }, + { + "epoch": 4.26, + "learning_rate": 4.0945591924244735e-06, + "loss": 1.0502, + "step": 57655 + }, + { + "epoch": 4.26, + "learning_rate": 4.094237305577693e-06, + "loss": 1.0762, + "step": 57656 + }, + { + "epoch": 4.26, + "learning_rate": 4.093915428126897e-06, + "loss": 0.9138, + "step": 57657 + }, + { + "epoch": 4.26, + "learning_rate": 4.093593560072596e-06, + "loss": 1.0515, + "step": 57658 + }, + { + "epoch": 4.26, + "learning_rate": 4.093271701415301e-06, + "loss": 0.9495, + "step": 57659 + }, + { + "epoch": 4.26, + "learning_rate": 4.092949852155529e-06, + "loss": 0.9311, + "step": 57660 + }, + { + "epoch": 4.26, + "learning_rate": 4.092628012293789e-06, + "loss": 0.8974, + "step": 57661 + }, + { + "epoch": 4.26, + "learning_rate": 4.092306181830595e-06, + "loss": 0.9707, + "step": 57662 + }, + { + "epoch": 4.26, + "learning_rate": 4.0919843607664555e-06, + "loss": 0.9662, + "step": 57663 + }, + { + "epoch": 4.26, + "learning_rate": 4.091662549101882e-06, + "loss": 1.0456, + "step": 57664 + }, + { + "epoch": 4.26, + "learning_rate": 4.091340746837395e-06, + "loss": 1.0434, + "step": 57665 + }, + { + "epoch": 4.26, + "learning_rate": 4.091018953973492e-06, + "loss": 1.061, + "step": 57666 + }, + { + "epoch": 4.26, + "learning_rate": 4.090697170510699e-06, + "loss": 0.9535, + "step": 57667 + }, + { + "epoch": 4.26, + "learning_rate": 4.090375396449518e-06, + "loss": 0.9629, + "step": 57668 + }, + { + "epoch": 4.26, + "learning_rate": 4.090053631790472e-06, + "loss": 1.0588, + "step": 57669 + }, + { + "epoch": 4.26, + "learning_rate": 4.08973187653406e-06, + "loss": 1.0087, + "step": 57670 + }, + { + "epoch": 4.26, + "learning_rate": 4.089410130680802e-06, + "loss": 1.0255, + "step": 57671 + }, + { + "epoch": 4.26, + "learning_rate": 4.08908839423121e-06, + "loss": 0.9271, + "step": 57672 + }, + { + "epoch": 4.26, + "learning_rate": 4.088766667185792e-06, + "loss": 0.9741, + "step": 57673 + }, + { + "epoch": 4.26, + "learning_rate": 4.0884449495450615e-06, + "loss": 1.0003, + "step": 57674 + }, + { + "epoch": 4.26, + "learning_rate": 4.0881232413095285e-06, + "loss": 1.0322, + "step": 57675 + }, + { + "epoch": 4.26, + "learning_rate": 4.0878015424797094e-06, + "loss": 1.0641, + "step": 57676 + }, + { + "epoch": 4.26, + "learning_rate": 4.087479853056113e-06, + "loss": 1.045, + "step": 57677 + }, + { + "epoch": 4.26, + "learning_rate": 4.087158173039253e-06, + "loss": 0.9667, + "step": 57678 + }, + { + "epoch": 4.26, + "learning_rate": 4.0868365024296355e-06, + "loss": 0.9974, + "step": 57679 + }, + { + "epoch": 4.26, + "learning_rate": 4.0865148412277796e-06, + "loss": 0.9962, + "step": 57680 + }, + { + "epoch": 4.26, + "learning_rate": 4.086193189434195e-06, + "loss": 0.9904, + "step": 57681 + }, + { + "epoch": 4.26, + "learning_rate": 4.085871547049392e-06, + "loss": 1.0527, + "step": 57682 + }, + { + "epoch": 4.26, + "learning_rate": 4.085549914073882e-06, + "loss": 0.8921, + "step": 57683 + }, + { + "epoch": 4.26, + "learning_rate": 4.085228290508174e-06, + "loss": 0.9867, + "step": 57684 + }, + { + "epoch": 4.26, + "learning_rate": 4.084906676352788e-06, + "loss": 1.0019, + "step": 57685 + }, + { + "epoch": 4.26, + "learning_rate": 4.08458507160823e-06, + "loss": 0.9947, + "step": 57686 + }, + { + "epoch": 4.26, + "learning_rate": 4.084263476275012e-06, + "loss": 0.9889, + "step": 57687 + }, + { + "epoch": 4.26, + "learning_rate": 4.083941890353643e-06, + "loss": 0.8589, + "step": 57688 + }, + { + "epoch": 4.26, + "learning_rate": 4.083620313844645e-06, + "loss": 1.0333, + "step": 57689 + }, + { + "epoch": 4.26, + "learning_rate": 4.0832987467485195e-06, + "loss": 0.9326, + "step": 57690 + }, + { + "epoch": 4.26, + "learning_rate": 4.082977189065776e-06, + "loss": 0.9913, + "step": 57691 + }, + { + "epoch": 4.26, + "learning_rate": 4.082655640796937e-06, + "loss": 1.0846, + "step": 57692 + }, + { + "epoch": 4.26, + "learning_rate": 4.082334101942507e-06, + "loss": 0.9757, + "step": 57693 + }, + { + "epoch": 4.26, + "learning_rate": 4.082012572502998e-06, + "loss": 0.9198, + "step": 57694 + }, + { + "epoch": 4.26, + "learning_rate": 4.081691052478919e-06, + "loss": 0.8958, + "step": 57695 + }, + { + "epoch": 4.26, + "learning_rate": 4.081369541870789e-06, + "loss": 1.0672, + "step": 57696 + }, + { + "epoch": 4.26, + "learning_rate": 4.081048040679116e-06, + "loss": 1.0076, + "step": 57697 + }, + { + "epoch": 4.26, + "learning_rate": 4.080726548904409e-06, + "loss": 1.0154, + "step": 57698 + }, + { + "epoch": 4.26, + "learning_rate": 4.080405066547183e-06, + "loss": 0.9768, + "step": 57699 + }, + { + "epoch": 4.26, + "learning_rate": 4.080083593607943e-06, + "loss": 0.9391, + "step": 57700 + }, + { + "epoch": 4.26, + "learning_rate": 4.0797621300872105e-06, + "loss": 0.9258, + "step": 57701 + }, + { + "epoch": 4.26, + "learning_rate": 4.079440675985491e-06, + "loss": 0.9116, + "step": 57702 + }, + { + "epoch": 4.26, + "learning_rate": 4.079119231303296e-06, + "loss": 0.9872, + "step": 57703 + }, + { + "epoch": 4.26, + "learning_rate": 4.078797796041134e-06, + "loss": 0.9089, + "step": 57704 + }, + { + "epoch": 4.26, + "learning_rate": 4.078476370199524e-06, + "loss": 1.0496, + "step": 57705 + }, + { + "epoch": 4.26, + "learning_rate": 4.0781549537789744e-06, + "loss": 1.0026, + "step": 57706 + }, + { + "epoch": 4.26, + "learning_rate": 4.077833546779994e-06, + "loss": 1.0463, + "step": 57707 + }, + { + "epoch": 4.26, + "learning_rate": 4.077512149203097e-06, + "loss": 0.9811, + "step": 57708 + }, + { + "epoch": 4.26, + "learning_rate": 4.077190761048788e-06, + "loss": 1.014, + "step": 57709 + }, + { + "epoch": 4.26, + "learning_rate": 4.076869382317593e-06, + "loss": 0.9818, + "step": 57710 + }, + { + "epoch": 4.26, + "learning_rate": 4.076548013010005e-06, + "loss": 1.0445, + "step": 57711 + }, + { + "epoch": 4.26, + "learning_rate": 4.076226653126549e-06, + "loss": 0.8916, + "step": 57712 + }, + { + "epoch": 4.26, + "learning_rate": 4.075905302667729e-06, + "loss": 0.9783, + "step": 57713 + }, + { + "epoch": 4.26, + "learning_rate": 4.0755839616340645e-06, + "loss": 0.9307, + "step": 57714 + }, + { + "epoch": 4.26, + "learning_rate": 4.075262630026055e-06, + "loss": 1.0552, + "step": 57715 + }, + { + "epoch": 4.26, + "learning_rate": 4.0749413078442205e-06, + "loss": 0.9048, + "step": 57716 + }, + { + "epoch": 4.26, + "learning_rate": 4.074619995089066e-06, + "loss": 1.0856, + "step": 57717 + }, + { + "epoch": 4.26, + "learning_rate": 4.074298691761115e-06, + "loss": 1.0351, + "step": 57718 + }, + { + "epoch": 4.26, + "learning_rate": 4.0739773978608645e-06, + "loss": 1.0845, + "step": 57719 + }, + { + "epoch": 4.26, + "learning_rate": 4.073656113388829e-06, + "loss": 0.9997, + "step": 57720 + }, + { + "epoch": 4.26, + "learning_rate": 4.073334838345525e-06, + "loss": 0.9441, + "step": 57721 + }, + { + "epoch": 4.27, + "learning_rate": 4.07301357273146e-06, + "loss": 1.0859, + "step": 57722 + }, + { + "epoch": 4.27, + "learning_rate": 4.0726923165471465e-06, + "loss": 0.862, + "step": 57723 + }, + { + "epoch": 4.27, + "learning_rate": 4.072371069793091e-06, + "loss": 1.0704, + "step": 57724 + }, + { + "epoch": 4.27, + "learning_rate": 4.072049832469811e-06, + "loss": 1.0115, + "step": 57725 + }, + { + "epoch": 4.27, + "learning_rate": 4.071728604577815e-06, + "loss": 1.1173, + "step": 57726 + }, + { + "epoch": 4.27, + "learning_rate": 4.071407386117614e-06, + "loss": 1.0089, + "step": 57727 + }, + { + "epoch": 4.27, + "learning_rate": 4.07108617708972e-06, + "loss": 0.926, + "step": 57728 + }, + { + "epoch": 4.27, + "learning_rate": 4.070764977494639e-06, + "loss": 0.9916, + "step": 57729 + }, + { + "epoch": 4.27, + "learning_rate": 4.07044378733289e-06, + "loss": 1.0748, + "step": 57730 + }, + { + "epoch": 4.27, + "learning_rate": 4.070122606604979e-06, + "loss": 1.0694, + "step": 57731 + }, + { + "epoch": 4.27, + "learning_rate": 4.069801435311419e-06, + "loss": 0.998, + "step": 57732 + }, + { + "epoch": 4.27, + "learning_rate": 4.069480273452716e-06, + "loss": 0.8807, + "step": 57733 + }, + { + "epoch": 4.27, + "learning_rate": 4.069159121029392e-06, + "loss": 0.8991, + "step": 57734 + }, + { + "epoch": 4.27, + "learning_rate": 4.068837978041945e-06, + "loss": 1.0295, + "step": 57735 + }, + { + "epoch": 4.27, + "learning_rate": 4.068516844490894e-06, + "loss": 1.082, + "step": 57736 + }, + { + "epoch": 4.27, + "learning_rate": 4.068195720376748e-06, + "loss": 1.0029, + "step": 57737 + }, + { + "epoch": 4.27, + "learning_rate": 4.06787460570002e-06, + "loss": 1.0367, + "step": 57738 + }, + { + "epoch": 4.27, + "learning_rate": 4.067553500461216e-06, + "loss": 0.9952, + "step": 57739 + }, + { + "epoch": 4.27, + "learning_rate": 4.067232404660847e-06, + "loss": 1.0248, + "step": 57740 + }, + { + "epoch": 4.27, + "learning_rate": 4.06691131829943e-06, + "loss": 1.0178, + "step": 57741 + }, + { + "epoch": 4.27, + "learning_rate": 4.0665902413774716e-06, + "loss": 1.0478, + "step": 57742 + }, + { + "epoch": 4.27, + "learning_rate": 4.066269173895484e-06, + "loss": 0.89, + "step": 57743 + }, + { + "epoch": 4.27, + "learning_rate": 4.065948115853974e-06, + "loss": 0.9372, + "step": 57744 + }, + { + "epoch": 4.27, + "learning_rate": 4.065627067253458e-06, + "loss": 1.1287, + "step": 57745 + }, + { + "epoch": 4.27, + "learning_rate": 4.065306028094444e-06, + "loss": 0.8972, + "step": 57746 + }, + { + "epoch": 4.27, + "learning_rate": 4.064984998377445e-06, + "loss": 0.9311, + "step": 57747 + }, + { + "epoch": 4.27, + "learning_rate": 4.064663978102969e-06, + "loss": 0.9025, + "step": 57748 + }, + { + "epoch": 4.27, + "learning_rate": 4.064342967271524e-06, + "loss": 1.0635, + "step": 57749 + }, + { + "epoch": 4.27, + "learning_rate": 4.0640219658836285e-06, + "loss": 0.9786, + "step": 57750 + }, + { + "epoch": 4.27, + "learning_rate": 4.06370097393979e-06, + "loss": 1.025, + "step": 57751 + }, + { + "epoch": 4.27, + "learning_rate": 4.063379991440516e-06, + "loss": 1.0592, + "step": 57752 + }, + { + "epoch": 4.27, + "learning_rate": 4.063059018386318e-06, + "loss": 1.0011, + "step": 57753 + }, + { + "epoch": 4.27, + "learning_rate": 4.0627380547777095e-06, + "loss": 1.032, + "step": 57754 + }, + { + "epoch": 4.27, + "learning_rate": 4.062417100615206e-06, + "loss": 0.9972, + "step": 57755 + }, + { + "epoch": 4.27, + "learning_rate": 4.0620961558993035e-06, + "loss": 0.9904, + "step": 57756 + }, + { + "epoch": 4.27, + "learning_rate": 4.061775220630524e-06, + "loss": 0.995, + "step": 57757 + }, + { + "epoch": 4.27, + "learning_rate": 4.061454294809372e-06, + "loss": 0.9628, + "step": 57758 + }, + { + "epoch": 4.27, + "learning_rate": 4.061133378436368e-06, + "loss": 0.9429, + "step": 57759 + }, + { + "epoch": 4.27, + "learning_rate": 4.06081247151201e-06, + "loss": 1.0189, + "step": 57760 + }, + { + "epoch": 4.27, + "learning_rate": 4.060491574036817e-06, + "loss": 1.08, + "step": 57761 + }, + { + "epoch": 4.27, + "learning_rate": 4.060170686011293e-06, + "loss": 0.982, + "step": 57762 + }, + { + "epoch": 4.27, + "learning_rate": 4.059849807435961e-06, + "loss": 0.9647, + "step": 57763 + }, + { + "epoch": 4.27, + "learning_rate": 4.059528938311319e-06, + "loss": 1.0331, + "step": 57764 + }, + { + "epoch": 4.27, + "learning_rate": 4.0592080786378775e-06, + "loss": 1.0275, + "step": 57765 + }, + { + "epoch": 4.27, + "learning_rate": 4.058887228416154e-06, + "loss": 0.9596, + "step": 57766 + }, + { + "epoch": 4.27, + "learning_rate": 4.058566387646656e-06, + "loss": 0.7861, + "step": 57767 + }, + { + "epoch": 4.27, + "learning_rate": 4.058245556329895e-06, + "loss": 1.0289, + "step": 57768 + }, + { + "epoch": 4.27, + "learning_rate": 4.057924734466375e-06, + "loss": 0.981, + "step": 57769 + }, + { + "epoch": 4.27, + "learning_rate": 4.057603922056617e-06, + "loss": 0.9639, + "step": 57770 + }, + { + "epoch": 4.27, + "learning_rate": 4.057283119101125e-06, + "loss": 0.9759, + "step": 57771 + }, + { + "epoch": 4.27, + "learning_rate": 4.056962325600412e-06, + "loss": 0.9364, + "step": 57772 + }, + { + "epoch": 4.27, + "learning_rate": 4.056641541554982e-06, + "loss": 0.9634, + "step": 57773 + }, + { + "epoch": 4.27, + "learning_rate": 4.056320766965354e-06, + "loss": 0.9266, + "step": 57774 + }, + { + "epoch": 4.27, + "learning_rate": 4.056000001832035e-06, + "loss": 0.9438, + "step": 57775 + }, + { + "epoch": 4.27, + "learning_rate": 4.055679246155535e-06, + "loss": 0.9665, + "step": 57776 + }, + { + "epoch": 4.27, + "learning_rate": 4.055358499936364e-06, + "loss": 1.0443, + "step": 57777 + }, + { + "epoch": 4.27, + "learning_rate": 4.055037763175028e-06, + "loss": 0.9393, + "step": 57778 + }, + { + "epoch": 4.27, + "learning_rate": 4.054717035872051e-06, + "loss": 0.9628, + "step": 57779 + }, + { + "epoch": 4.27, + "learning_rate": 4.054396318027926e-06, + "loss": 1.0223, + "step": 57780 + }, + { + "epoch": 4.27, + "learning_rate": 4.054075609643175e-06, + "loss": 0.9519, + "step": 57781 + }, + { + "epoch": 4.27, + "learning_rate": 4.0537549107183e-06, + "loss": 0.9746, + "step": 57782 + }, + { + "epoch": 4.27, + "learning_rate": 4.053434221253825e-06, + "loss": 0.9356, + "step": 57783 + }, + { + "epoch": 4.27, + "learning_rate": 4.053113541250247e-06, + "loss": 1.0492, + "step": 57784 + }, + { + "epoch": 4.27, + "learning_rate": 4.052792870708076e-06, + "loss": 1.0387, + "step": 57785 + }, + { + "epoch": 4.27, + "learning_rate": 4.05247220962783e-06, + "loss": 1.0127, + "step": 57786 + }, + { + "epoch": 4.27, + "learning_rate": 4.052151558010016e-06, + "loss": 0.8371, + "step": 57787 + }, + { + "epoch": 4.27, + "learning_rate": 4.051830915855143e-06, + "loss": 1.0584, + "step": 57788 + }, + { + "epoch": 4.27, + "learning_rate": 4.051510283163719e-06, + "loss": 1.072, + "step": 57789 + }, + { + "epoch": 4.27, + "learning_rate": 4.051189659936261e-06, + "loss": 0.9244, + "step": 57790 + }, + { + "epoch": 4.27, + "learning_rate": 4.050869046173274e-06, + "loss": 0.9721, + "step": 57791 + }, + { + "epoch": 4.27, + "learning_rate": 4.050548441875271e-06, + "loss": 1.0224, + "step": 57792 + }, + { + "epoch": 4.27, + "learning_rate": 4.05022784704276e-06, + "loss": 0.9755, + "step": 57793 + }, + { + "epoch": 4.27, + "learning_rate": 4.049907261676246e-06, + "loss": 1.0096, + "step": 57794 + }, + { + "epoch": 4.27, + "learning_rate": 4.049586685776249e-06, + "loss": 0.9985, + "step": 57795 + }, + { + "epoch": 4.27, + "learning_rate": 4.0492661193432744e-06, + "loss": 0.9567, + "step": 57796 + }, + { + "epoch": 4.27, + "learning_rate": 4.048945562377832e-06, + "loss": 1.0809, + "step": 57797 + }, + { + "epoch": 4.27, + "learning_rate": 4.04862501488043e-06, + "loss": 0.9665, + "step": 57798 + }, + { + "epoch": 4.27, + "learning_rate": 4.048304476851582e-06, + "loss": 1.0116, + "step": 57799 + }, + { + "epoch": 4.27, + "learning_rate": 4.047983948291797e-06, + "loss": 0.8343, + "step": 57800 + }, + { + "epoch": 4.27, + "learning_rate": 4.047663429201585e-06, + "loss": 1.0138, + "step": 57801 + }, + { + "epoch": 4.27, + "learning_rate": 4.047342919581455e-06, + "loss": 1.0252, + "step": 57802 + }, + { + "epoch": 4.27, + "learning_rate": 4.0470224194319145e-06, + "loss": 0.9462, + "step": 57803 + }, + { + "epoch": 4.27, + "learning_rate": 4.0467019287534815e-06, + "loss": 0.9122, + "step": 57804 + }, + { + "epoch": 4.27, + "learning_rate": 4.046381447546655e-06, + "loss": 1.0149, + "step": 57805 + }, + { + "epoch": 4.27, + "learning_rate": 4.046060975811953e-06, + "loss": 1.0958, + "step": 57806 + }, + { + "epoch": 4.27, + "learning_rate": 4.045740513549879e-06, + "loss": 1.0833, + "step": 57807 + }, + { + "epoch": 4.27, + "learning_rate": 4.045420060760955e-06, + "loss": 1.0058, + "step": 57808 + }, + { + "epoch": 4.27, + "learning_rate": 4.045099617445675e-06, + "loss": 0.9926, + "step": 57809 + }, + { + "epoch": 4.27, + "learning_rate": 4.044779183604559e-06, + "loss": 1.0244, + "step": 57810 + }, + { + "epoch": 4.27, + "learning_rate": 4.0444587592381145e-06, + "loss": 0.9409, + "step": 57811 + }, + { + "epoch": 4.27, + "learning_rate": 4.044138344346852e-06, + "loss": 0.9344, + "step": 57812 + }, + { + "epoch": 4.27, + "learning_rate": 4.0438179389312795e-06, + "loss": 0.9627, + "step": 57813 + }, + { + "epoch": 4.27, + "learning_rate": 4.043497542991904e-06, + "loss": 0.9501, + "step": 57814 + }, + { + "epoch": 4.27, + "learning_rate": 4.043177156529241e-06, + "loss": 0.9712, + "step": 57815 + }, + { + "epoch": 4.27, + "learning_rate": 4.042856779543799e-06, + "loss": 0.9212, + "step": 57816 + }, + { + "epoch": 4.27, + "learning_rate": 4.042536412036086e-06, + "loss": 0.9071, + "step": 57817 + }, + { + "epoch": 4.27, + "learning_rate": 4.04221605400661e-06, + "loss": 1.0095, + "step": 57818 + }, + { + "epoch": 4.27, + "learning_rate": 4.041895705455886e-06, + "loss": 0.9868, + "step": 57819 + }, + { + "epoch": 4.27, + "learning_rate": 4.04157536638442e-06, + "loss": 1.0177, + "step": 57820 + }, + { + "epoch": 4.27, + "learning_rate": 4.041255036792723e-06, + "loss": 0.9163, + "step": 57821 + }, + { + "epoch": 4.27, + "learning_rate": 4.040934716681304e-06, + "loss": 1.0175, + "step": 57822 + }, + { + "epoch": 4.27, + "learning_rate": 4.040614406050669e-06, + "loss": 1.0586, + "step": 57823 + }, + { + "epoch": 4.27, + "learning_rate": 4.040294104901338e-06, + "loss": 0.9766, + "step": 57824 + }, + { + "epoch": 4.27, + "learning_rate": 4.039973813233807e-06, + "loss": 0.9866, + "step": 57825 + }, + { + "epoch": 4.27, + "learning_rate": 4.0396535310485965e-06, + "loss": 0.9763, + "step": 57826 + }, + { + "epoch": 4.27, + "learning_rate": 4.0393332583462075e-06, + "loss": 0.9206, + "step": 57827 + }, + { + "epoch": 4.27, + "learning_rate": 4.03901299512716e-06, + "loss": 1.0364, + "step": 57828 + }, + { + "epoch": 4.27, + "learning_rate": 4.038692741391953e-06, + "loss": 0.9678, + "step": 57829 + }, + { + "epoch": 4.27, + "learning_rate": 4.038372497141103e-06, + "loss": 0.9693, + "step": 57830 + }, + { + "epoch": 4.27, + "learning_rate": 4.038052262375116e-06, + "loss": 1.031, + "step": 57831 + }, + { + "epoch": 4.27, + "learning_rate": 4.037732037094504e-06, + "loss": 0.9592, + "step": 57832 + }, + { + "epoch": 4.27, + "learning_rate": 4.037411821299775e-06, + "loss": 0.9581, + "step": 57833 + }, + { + "epoch": 4.27, + "learning_rate": 4.037091614991434e-06, + "loss": 1.0557, + "step": 57834 + }, + { + "epoch": 4.27, + "learning_rate": 4.036771418169999e-06, + "loss": 0.9358, + "step": 57835 + }, + { + "epoch": 4.27, + "learning_rate": 4.036451230835975e-06, + "loss": 0.9688, + "step": 57836 + }, + { + "epoch": 4.27, + "learning_rate": 4.036131052989872e-06, + "loss": 0.9124, + "step": 57837 + }, + { + "epoch": 4.27, + "learning_rate": 4.035810884632196e-06, + "loss": 1.0106, + "step": 57838 + }, + { + "epoch": 4.27, + "learning_rate": 4.035490725763464e-06, + "loss": 0.9809, + "step": 57839 + }, + { + "epoch": 4.27, + "learning_rate": 4.0351705763841785e-06, + "loss": 0.9551, + "step": 57840 + }, + { + "epoch": 4.27, + "learning_rate": 4.034850436494853e-06, + "loss": 0.9221, + "step": 57841 + }, + { + "epoch": 4.27, + "learning_rate": 4.034530306095994e-06, + "loss": 1.077, + "step": 57842 + }, + { + "epoch": 4.27, + "learning_rate": 4.0342101851881085e-06, + "loss": 1.1131, + "step": 57843 + }, + { + "epoch": 4.27, + "learning_rate": 4.0338900737717134e-06, + "loss": 0.9799, + "step": 57844 + }, + { + "epoch": 4.27, + "learning_rate": 4.033569971847314e-06, + "loss": 0.9707, + "step": 57845 + }, + { + "epoch": 4.27, + "learning_rate": 4.03324987941542e-06, + "loss": 1.1582, + "step": 57846 + }, + { + "epoch": 4.27, + "learning_rate": 4.032929796476535e-06, + "loss": 0.9099, + "step": 57847 + }, + { + "epoch": 4.27, + "learning_rate": 4.0326097230311775e-06, + "loss": 0.9456, + "step": 57848 + }, + { + "epoch": 4.27, + "learning_rate": 4.032289659079857e-06, + "loss": 0.9693, + "step": 57849 + }, + { + "epoch": 4.27, + "learning_rate": 4.03196960462307e-06, + "loss": 0.9892, + "step": 57850 + }, + { + "epoch": 4.27, + "learning_rate": 4.031649559661337e-06, + "loss": 0.9703, + "step": 57851 + }, + { + "epoch": 4.27, + "learning_rate": 4.031329524195161e-06, + "loss": 1.0216, + "step": 57852 + }, + { + "epoch": 4.27, + "learning_rate": 4.031009498225062e-06, + "loss": 0.9635, + "step": 57853 + }, + { + "epoch": 4.27, + "learning_rate": 4.030689481751534e-06, + "loss": 1.009, + "step": 57854 + }, + { + "epoch": 4.27, + "learning_rate": 4.030369474775097e-06, + "loss": 1.0344, + "step": 57855 + }, + { + "epoch": 4.27, + "learning_rate": 4.030049477296258e-06, + "loss": 1.016, + "step": 57856 + }, + { + "epoch": 4.28, + "learning_rate": 4.0297294893155234e-06, + "loss": 1.0513, + "step": 57857 + }, + { + "epoch": 4.28, + "learning_rate": 4.029409510833405e-06, + "loss": 1.1458, + "step": 57858 + }, + { + "epoch": 4.28, + "learning_rate": 4.029089541850406e-06, + "loss": 0.9861, + "step": 57859 + }, + { + "epoch": 4.28, + "learning_rate": 4.028769582367045e-06, + "loss": 0.9798, + "step": 57860 + }, + { + "epoch": 4.28, + "learning_rate": 4.028449632383825e-06, + "loss": 1.0397, + "step": 57861 + }, + { + "epoch": 4.28, + "learning_rate": 4.028129691901257e-06, + "loss": 0.9859, + "step": 57862 + }, + { + "epoch": 4.28, + "learning_rate": 4.027809760919844e-06, + "loss": 1.0237, + "step": 57863 + }, + { + "epoch": 4.28, + "learning_rate": 4.027489839440105e-06, + "loss": 1.0754, + "step": 57864 + }, + { + "epoch": 4.28, + "learning_rate": 4.027169927462544e-06, + "loss": 1.0307, + "step": 57865 + }, + { + "epoch": 4.28, + "learning_rate": 4.02685002498767e-06, + "loss": 0.9464, + "step": 57866 + }, + { + "epoch": 4.28, + "learning_rate": 4.026530132015994e-06, + "loss": 0.9861, + "step": 57867 + }, + { + "epoch": 4.28, + "learning_rate": 4.026210248548017e-06, + "loss": 1.0204, + "step": 57868 + }, + { + "epoch": 4.28, + "learning_rate": 4.025890374584263e-06, + "loss": 0.9393, + "step": 57869 + }, + { + "epoch": 4.28, + "learning_rate": 4.025570510125223e-06, + "loss": 0.9493, + "step": 57870 + }, + { + "epoch": 4.28, + "learning_rate": 4.02525065517142e-06, + "loss": 0.992, + "step": 57871 + }, + { + "epoch": 4.28, + "learning_rate": 4.024930809723353e-06, + "loss": 1.0363, + "step": 57872 + }, + { + "epoch": 4.28, + "learning_rate": 4.024610973781543e-06, + "loss": 0.9808, + "step": 57873 + }, + { + "epoch": 4.28, + "learning_rate": 4.0242911473464855e-06, + "loss": 1.0377, + "step": 57874 + }, + { + "epoch": 4.28, + "learning_rate": 4.023971330418698e-06, + "loss": 1.1054, + "step": 57875 + }, + { + "epoch": 4.28, + "learning_rate": 4.023651522998686e-06, + "loss": 1.0644, + "step": 57876 + }, + { + "epoch": 4.28, + "learning_rate": 4.02333172508696e-06, + "loss": 1.0063, + "step": 57877 + }, + { + "epoch": 4.28, + "learning_rate": 4.023011936684027e-06, + "loss": 0.9331, + "step": 57878 + }, + { + "epoch": 4.28, + "learning_rate": 4.022692157790393e-06, + "loss": 0.9282, + "step": 57879 + }, + { + "epoch": 4.28, + "learning_rate": 4.022372388406573e-06, + "loss": 0.944, + "step": 57880 + }, + { + "epoch": 4.28, + "learning_rate": 4.022052628533074e-06, + "loss": 0.9989, + "step": 57881 + }, + { + "epoch": 4.28, + "learning_rate": 4.021732878170404e-06, + "loss": 1.0952, + "step": 57882 + }, + { + "epoch": 4.28, + "learning_rate": 4.021413137319068e-06, + "loss": 0.9857, + "step": 57883 + }, + { + "epoch": 4.28, + "learning_rate": 4.021093405979582e-06, + "loss": 0.9592, + "step": 57884 + }, + { + "epoch": 4.28, + "learning_rate": 4.020773684152449e-06, + "loss": 0.9474, + "step": 57885 + }, + { + "epoch": 4.28, + "learning_rate": 4.0204539718381805e-06, + "loss": 1.0504, + "step": 57886 + }, + { + "epoch": 4.28, + "learning_rate": 4.020134269037284e-06, + "loss": 0.9558, + "step": 57887 + }, + { + "epoch": 4.28, + "learning_rate": 4.019814575750266e-06, + "loss": 1.0041, + "step": 57888 + }, + { + "epoch": 4.28, + "learning_rate": 4.01949489197764e-06, + "loss": 0.9653, + "step": 57889 + }, + { + "epoch": 4.28, + "learning_rate": 4.019175217719912e-06, + "loss": 1.1061, + "step": 57890 + }, + { + "epoch": 4.28, + "learning_rate": 4.018855552977591e-06, + "loss": 0.9832, + "step": 57891 + }, + { + "epoch": 4.28, + "learning_rate": 4.018535897751182e-06, + "loss": 1.0201, + "step": 57892 + }, + { + "epoch": 4.28, + "learning_rate": 4.0182162520412e-06, + "loss": 1.0357, + "step": 57893 + }, + { + "epoch": 4.28, + "learning_rate": 4.0178966158481496e-06, + "loss": 1.1088, + "step": 57894 + }, + { + "epoch": 4.28, + "learning_rate": 4.01757698917254e-06, + "loss": 1.0603, + "step": 57895 + }, + { + "epoch": 4.28, + "learning_rate": 4.017257372014881e-06, + "loss": 1.0097, + "step": 57896 + }, + { + "epoch": 4.28, + "learning_rate": 4.016937764375676e-06, + "loss": 1.0119, + "step": 57897 + }, + { + "epoch": 4.28, + "learning_rate": 4.016618166255444e-06, + "loss": 1.0612, + "step": 57898 + }, + { + "epoch": 4.28, + "learning_rate": 4.0162985776546805e-06, + "loss": 0.948, + "step": 57899 + }, + { + "epoch": 4.28, + "learning_rate": 4.0159789985739025e-06, + "loss": 0.9178, + "step": 57900 + }, + { + "epoch": 4.28, + "learning_rate": 4.015659429013617e-06, + "loss": 1.0276, + "step": 57901 + }, + { + "epoch": 4.28, + "learning_rate": 4.015339868974333e-06, + "loss": 0.994, + "step": 57902 + }, + { + "epoch": 4.28, + "learning_rate": 4.015020318456553e-06, + "loss": 0.9763, + "step": 57903 + }, + { + "epoch": 4.28, + "learning_rate": 4.014700777460794e-06, + "loss": 1.0136, + "step": 57904 + }, + { + "epoch": 4.28, + "learning_rate": 4.01438124598756e-06, + "loss": 1.1359, + "step": 57905 + }, + { + "epoch": 4.28, + "learning_rate": 4.01406172403736e-06, + "loss": 0.9367, + "step": 57906 + }, + { + "epoch": 4.28, + "learning_rate": 4.013742211610702e-06, + "loss": 1.0057, + "step": 57907 + }, + { + "epoch": 4.28, + "learning_rate": 4.013422708708091e-06, + "loss": 1.0128, + "step": 57908 + }, + { + "epoch": 4.28, + "learning_rate": 4.013103215330043e-06, + "loss": 0.9829, + "step": 57909 + }, + { + "epoch": 4.28, + "learning_rate": 4.012783731477061e-06, + "loss": 0.902, + "step": 57910 + }, + { + "epoch": 4.28, + "learning_rate": 4.012464257149655e-06, + "loss": 0.9593, + "step": 57911 + }, + { + "epoch": 4.28, + "learning_rate": 4.0121447923483295e-06, + "loss": 1.0078, + "step": 57912 + }, + { + "epoch": 4.28, + "learning_rate": 4.0118253370736e-06, + "loss": 1.016, + "step": 57913 + }, + { + "epoch": 4.28, + "learning_rate": 4.011505891325973e-06, + "loss": 1.0174, + "step": 57914 + }, + { + "epoch": 4.28, + "learning_rate": 4.011186455105947e-06, + "loss": 1.1364, + "step": 57915 + }, + { + "epoch": 4.28, + "learning_rate": 4.010867028414043e-06, + "loss": 0.9966, + "step": 57916 + }, + { + "epoch": 4.28, + "learning_rate": 4.010547611250759e-06, + "loss": 1.0874, + "step": 57917 + }, + { + "epoch": 4.28, + "learning_rate": 4.0102282036166155e-06, + "loss": 0.9185, + "step": 57918 + }, + { + "epoch": 4.28, + "learning_rate": 4.009908805512106e-06, + "loss": 1.1149, + "step": 57919 + }, + { + "epoch": 4.28, + "learning_rate": 4.009589416937749e-06, + "loss": 1.0838, + "step": 57920 + }, + { + "epoch": 4.28, + "learning_rate": 4.009270037894048e-06, + "loss": 0.9717, + "step": 57921 + }, + { + "epoch": 4.28, + "learning_rate": 4.0089506683815175e-06, + "loss": 0.9617, + "step": 57922 + }, + { + "epoch": 4.28, + "learning_rate": 4.008631308400658e-06, + "loss": 0.9319, + "step": 57923 + }, + { + "epoch": 4.28, + "learning_rate": 4.008311957951977e-06, + "loss": 1.0699, + "step": 57924 + }, + { + "epoch": 4.28, + "learning_rate": 4.007992617035988e-06, + "loss": 0.9378, + "step": 57925 + }, + { + "epoch": 4.28, + "learning_rate": 4.0076732856532e-06, + "loss": 0.9642, + "step": 57926 + }, + { + "epoch": 4.28, + "learning_rate": 4.007353963804116e-06, + "loss": 0.9764, + "step": 57927 + }, + { + "epoch": 4.28, + "learning_rate": 4.007034651489242e-06, + "loss": 0.9174, + "step": 57928 + }, + { + "epoch": 4.28, + "learning_rate": 4.006715348709095e-06, + "loss": 0.9832, + "step": 57929 + }, + { + "epoch": 4.28, + "learning_rate": 4.0063960554641775e-06, + "loss": 1.0224, + "step": 57930 + }, + { + "epoch": 4.28, + "learning_rate": 4.006076771754999e-06, + "loss": 1.0255, + "step": 57931 + }, + { + "epoch": 4.28, + "learning_rate": 4.0057574975820625e-06, + "loss": 1.0647, + "step": 57932 + }, + { + "epoch": 4.28, + "learning_rate": 4.005438232945883e-06, + "loss": 0.9772, + "step": 57933 + }, + { + "epoch": 4.28, + "learning_rate": 4.005118977846966e-06, + "loss": 1.0574, + "step": 57934 + }, + { + "epoch": 4.28, + "learning_rate": 4.004799732285819e-06, + "loss": 0.9661, + "step": 57935 + }, + { + "epoch": 4.28, + "learning_rate": 4.00448049626295e-06, + "loss": 1.1167, + "step": 57936 + }, + { + "epoch": 4.28, + "learning_rate": 4.004161269778863e-06, + "loss": 0.9303, + "step": 57937 + }, + { + "epoch": 4.28, + "learning_rate": 4.003842052834073e-06, + "loss": 1.0573, + "step": 57938 + }, + { + "epoch": 4.28, + "learning_rate": 4.0035228454290845e-06, + "loss": 1.1258, + "step": 57939 + }, + { + "epoch": 4.28, + "learning_rate": 4.0032036475644054e-06, + "loss": 0.996, + "step": 57940 + }, + { + "epoch": 4.28, + "learning_rate": 4.00288445924054e-06, + "loss": 1.033, + "step": 57941 + }, + { + "epoch": 4.28, + "learning_rate": 4.002565280458004e-06, + "loss": 0.9712, + "step": 57942 + }, + { + "epoch": 4.28, + "learning_rate": 4.002246111217304e-06, + "loss": 1.0486, + "step": 57943 + }, + { + "epoch": 4.28, + "learning_rate": 4.001926951518936e-06, + "loss": 1.0235, + "step": 57944 + }, + { + "epoch": 4.28, + "learning_rate": 4.001607801363422e-06, + "loss": 1.0135, + "step": 57945 + }, + { + "epoch": 4.28, + "learning_rate": 4.001288660751262e-06, + "loss": 0.9251, + "step": 57946 + }, + { + "epoch": 4.28, + "learning_rate": 4.000969529682968e-06, + "loss": 1.0744, + "step": 57947 + }, + { + "epoch": 4.28, + "learning_rate": 4.000650408159042e-06, + "loss": 0.98, + "step": 57948 + }, + { + "epoch": 4.28, + "learning_rate": 4.0003312961799975e-06, + "loss": 0.931, + "step": 57949 + }, + { + "epoch": 4.28, + "learning_rate": 4.000012193746341e-06, + "loss": 1.0009, + "step": 57950 + }, + { + "epoch": 4.28, + "learning_rate": 3.99969310085858e-06, + "loss": 1.1082, + "step": 57951 + }, + { + "epoch": 4.28, + "learning_rate": 3.999374017517219e-06, + "loss": 1.0692, + "step": 57952 + }, + { + "epoch": 4.28, + "learning_rate": 3.9990549437227665e-06, + "loss": 0.9559, + "step": 57953 + }, + { + "epoch": 4.28, + "learning_rate": 3.998735879475735e-06, + "loss": 0.984, + "step": 57954 + }, + { + "epoch": 4.28, + "learning_rate": 3.998416824776629e-06, + "loss": 1.1127, + "step": 57955 + }, + { + "epoch": 4.28, + "learning_rate": 3.998097779625954e-06, + "loss": 0.9662, + "step": 57956 + }, + { + "epoch": 4.28, + "learning_rate": 3.997778744024218e-06, + "loss": 1.0147, + "step": 57957 + }, + { + "epoch": 4.28, + "learning_rate": 3.997459717971934e-06, + "loss": 1.0147, + "step": 57958 + }, + { + "epoch": 4.28, + "learning_rate": 3.997140701469605e-06, + "loss": 1.0965, + "step": 57959 + }, + { + "epoch": 4.28, + "learning_rate": 3.9968216945177384e-06, + "loss": 0.9504, + "step": 57960 + }, + { + "epoch": 4.28, + "learning_rate": 3.9965026971168445e-06, + "loss": 1.0569, + "step": 57961 + }, + { + "epoch": 4.28, + "learning_rate": 3.996183709267424e-06, + "loss": 0.9876, + "step": 57962 + }, + { + "epoch": 4.28, + "learning_rate": 3.995864730969996e-06, + "loss": 1.0108, + "step": 57963 + }, + { + "epoch": 4.28, + "learning_rate": 3.995545762225055e-06, + "loss": 0.8663, + "step": 57964 + }, + { + "epoch": 4.28, + "learning_rate": 3.995226803033118e-06, + "loss": 1.0712, + "step": 57965 + }, + { + "epoch": 4.28, + "learning_rate": 3.994907853394686e-06, + "loss": 0.9761, + "step": 57966 + }, + { + "epoch": 4.28, + "learning_rate": 3.994588913310278e-06, + "loss": 0.9137, + "step": 57967 + }, + { + "epoch": 4.28, + "learning_rate": 3.994269982780385e-06, + "loss": 1.0471, + "step": 57968 + }, + { + "epoch": 4.28, + "learning_rate": 3.9939510618055246e-06, + "loss": 0.947, + "step": 57969 + }, + { + "epoch": 4.28, + "learning_rate": 3.993632150386204e-06, + "loss": 1.0731, + "step": 57970 + }, + { + "epoch": 4.28, + "learning_rate": 3.993313248522928e-06, + "loss": 1.0682, + "step": 57971 + }, + { + "epoch": 4.28, + "learning_rate": 3.992994356216204e-06, + "loss": 0.9607, + "step": 57972 + }, + { + "epoch": 4.28, + "learning_rate": 3.9926754734665375e-06, + "loss": 0.9223, + "step": 57973 + }, + { + "epoch": 4.28, + "learning_rate": 3.9923566002744414e-06, + "loss": 0.9882, + "step": 57974 + }, + { + "epoch": 4.28, + "learning_rate": 3.992037736640421e-06, + "loss": 0.8854, + "step": 57975 + }, + { + "epoch": 4.28, + "learning_rate": 3.991718882564981e-06, + "loss": 0.9551, + "step": 57976 + }, + { + "epoch": 4.28, + "learning_rate": 3.991400038048628e-06, + "loss": 1.1382, + "step": 57977 + }, + { + "epoch": 4.28, + "learning_rate": 3.991081203091875e-06, + "loss": 1.1306, + "step": 57978 + }, + { + "epoch": 4.28, + "learning_rate": 3.990762377695225e-06, + "loss": 0.896, + "step": 57979 + }, + { + "epoch": 4.28, + "learning_rate": 3.990443561859188e-06, + "loss": 1.0004, + "step": 57980 + }, + { + "epoch": 4.28, + "learning_rate": 3.990124755584267e-06, + "loss": 1.0682, + "step": 57981 + }, + { + "epoch": 4.28, + "learning_rate": 3.989805958870969e-06, + "loss": 0.9635, + "step": 57982 + }, + { + "epoch": 4.28, + "learning_rate": 3.989487171719807e-06, + "loss": 1.0348, + "step": 57983 + }, + { + "epoch": 4.28, + "learning_rate": 3.989168394131285e-06, + "loss": 1.0261, + "step": 57984 + }, + { + "epoch": 4.28, + "learning_rate": 3.9888496261059105e-06, + "loss": 0.948, + "step": 57985 + }, + { + "epoch": 4.28, + "learning_rate": 3.988530867644186e-06, + "loss": 1.0018, + "step": 57986 + }, + { + "epoch": 4.28, + "learning_rate": 3.988212118746627e-06, + "loss": 0.9836, + "step": 57987 + }, + { + "epoch": 4.28, + "learning_rate": 3.987893379413737e-06, + "loss": 0.9754, + "step": 57988 + }, + { + "epoch": 4.28, + "learning_rate": 3.9875746496460214e-06, + "loss": 1.0574, + "step": 57989 + }, + { + "epoch": 4.28, + "learning_rate": 3.987255929443988e-06, + "loss": 1.0929, + "step": 57990 + }, + { + "epoch": 4.28, + "learning_rate": 3.986937218808146e-06, + "loss": 0.9994, + "step": 57991 + }, + { + "epoch": 4.28, + "learning_rate": 3.986618517739e-06, + "loss": 0.9697, + "step": 57992 + }, + { + "epoch": 4.29, + "learning_rate": 3.986299826237056e-06, + "loss": 0.8927, + "step": 57993 + }, + { + "epoch": 4.29, + "learning_rate": 3.985981144302824e-06, + "loss": 0.9964, + "step": 57994 + }, + { + "epoch": 4.29, + "learning_rate": 3.9856624719368105e-06, + "loss": 0.952, + "step": 57995 + }, + { + "epoch": 4.29, + "learning_rate": 3.985343809139524e-06, + "loss": 0.923, + "step": 57996 + }, + { + "epoch": 4.29, + "learning_rate": 3.985025155911464e-06, + "loss": 1.081, + "step": 57997 + }, + { + "epoch": 4.29, + "learning_rate": 3.984706512253147e-06, + "loss": 0.879, + "step": 57998 + }, + { + "epoch": 4.29, + "learning_rate": 3.9843878781650755e-06, + "loss": 0.9118, + "step": 57999 + }, + { + "epoch": 4.29, + "learning_rate": 3.9840692536477564e-06, + "loss": 0.9334, + "step": 58000 + }, + { + "epoch": 4.29, + "learning_rate": 3.983750638701698e-06, + "loss": 0.9999, + "step": 58001 + }, + { + "epoch": 4.29, + "learning_rate": 3.983432033327402e-06, + "loss": 1.1093, + "step": 58002 + }, + { + "epoch": 4.29, + "learning_rate": 3.9831134375253844e-06, + "loss": 0.9773, + "step": 58003 + }, + { + "epoch": 4.29, + "learning_rate": 3.982794851296146e-06, + "loss": 0.9812, + "step": 58004 + }, + { + "epoch": 4.29, + "learning_rate": 3.982476274640196e-06, + "loss": 0.9877, + "step": 58005 + }, + { + "epoch": 4.29, + "learning_rate": 3.982157707558035e-06, + "loss": 0.9466, + "step": 58006 + }, + { + "epoch": 4.29, + "learning_rate": 3.9818391500501796e-06, + "loss": 1.0706, + "step": 58007 + }, + { + "epoch": 4.29, + "learning_rate": 3.981520602117135e-06, + "loss": 1.0237, + "step": 58008 + }, + { + "epoch": 4.29, + "learning_rate": 3.9812020637593986e-06, + "loss": 0.9372, + "step": 58009 + }, + { + "epoch": 4.29, + "learning_rate": 3.980883534977486e-06, + "loss": 0.9909, + "step": 58010 + }, + { + "epoch": 4.29, + "learning_rate": 3.9805650157719e-06, + "loss": 1.1024, + "step": 58011 + }, + { + "epoch": 4.29, + "learning_rate": 3.980246506143154e-06, + "loss": 0.9925, + "step": 58012 + }, + { + "epoch": 4.29, + "learning_rate": 3.979928006091743e-06, + "loss": 0.9918, + "step": 58013 + }, + { + "epoch": 4.29, + "learning_rate": 3.979609515618186e-06, + "loss": 0.9859, + "step": 58014 + }, + { + "epoch": 4.29, + "learning_rate": 3.979291034722979e-06, + "loss": 1.0902, + "step": 58015 + }, + { + "epoch": 4.29, + "learning_rate": 3.97897256340664e-06, + "loss": 1.0438, + "step": 58016 + }, + { + "epoch": 4.29, + "learning_rate": 3.978654101669667e-06, + "loss": 0.9959, + "step": 58017 + }, + { + "epoch": 4.29, + "learning_rate": 3.978335649512566e-06, + "loss": 0.9379, + "step": 58018 + }, + { + "epoch": 4.29, + "learning_rate": 3.978017206935849e-06, + "loss": 0.9807, + "step": 58019 + }, + { + "epoch": 4.29, + "learning_rate": 3.977698773940022e-06, + "loss": 0.992, + "step": 58020 + }, + { + "epoch": 4.29, + "learning_rate": 3.977380350525589e-06, + "loss": 1.0065, + "step": 58021 + }, + { + "epoch": 4.29, + "learning_rate": 3.9770619366930555e-06, + "loss": 0.9805, + "step": 58022 + }, + { + "epoch": 4.29, + "learning_rate": 3.976743532442932e-06, + "loss": 0.9522, + "step": 58023 + }, + { + "epoch": 4.29, + "learning_rate": 3.976425137775724e-06, + "loss": 1.0301, + "step": 58024 + }, + { + "epoch": 4.29, + "learning_rate": 3.9761067526919385e-06, + "loss": 1.0373, + "step": 58025 + }, + { + "epoch": 4.29, + "learning_rate": 3.975788377192079e-06, + "loss": 1.1174, + "step": 58026 + }, + { + "epoch": 4.29, + "learning_rate": 3.975470011276652e-06, + "loss": 1.0313, + "step": 58027 + }, + { + "epoch": 4.29, + "learning_rate": 3.975151654946169e-06, + "loss": 0.9574, + "step": 58028 + }, + { + "epoch": 4.29, + "learning_rate": 3.974833308201134e-06, + "loss": 0.9296, + "step": 58029 + }, + { + "epoch": 4.29, + "learning_rate": 3.974514971042053e-06, + "loss": 1.0683, + "step": 58030 + }, + { + "epoch": 4.29, + "learning_rate": 3.974196643469428e-06, + "loss": 0.9805, + "step": 58031 + }, + { + "epoch": 4.29, + "learning_rate": 3.973878325483775e-06, + "loss": 0.9939, + "step": 58032 + }, + { + "epoch": 4.29, + "learning_rate": 3.973560017085594e-06, + "loss": 0.8653, + "step": 58033 + }, + { + "epoch": 4.29, + "learning_rate": 3.973241718275393e-06, + "loss": 1.1104, + "step": 58034 + }, + { + "epoch": 4.29, + "learning_rate": 3.972923429053678e-06, + "loss": 0.9885, + "step": 58035 + }, + { + "epoch": 4.29, + "learning_rate": 3.972605149420956e-06, + "loss": 1.001, + "step": 58036 + }, + { + "epoch": 4.29, + "learning_rate": 3.972286879377734e-06, + "loss": 0.972, + "step": 58037 + }, + { + "epoch": 4.29, + "learning_rate": 3.9719686189245125e-06, + "loss": 0.9528, + "step": 58038 + }, + { + "epoch": 4.29, + "learning_rate": 3.9716503680618054e-06, + "loss": 0.9343, + "step": 58039 + }, + { + "epoch": 4.29, + "learning_rate": 3.9713321267901185e-06, + "loss": 0.976, + "step": 58040 + }, + { + "epoch": 4.29, + "learning_rate": 3.971013895109954e-06, + "loss": 0.9691, + "step": 58041 + }, + { + "epoch": 4.29, + "learning_rate": 3.9706956730218185e-06, + "loss": 0.8506, + "step": 58042 + }, + { + "epoch": 4.29, + "learning_rate": 3.970377460526221e-06, + "loss": 1.0469, + "step": 58043 + }, + { + "epoch": 4.29, + "learning_rate": 3.970059257623669e-06, + "loss": 1.0251, + "step": 58044 + }, + { + "epoch": 4.29, + "learning_rate": 3.969741064314665e-06, + "loss": 1.0447, + "step": 58045 + }, + { + "epoch": 4.29, + "learning_rate": 3.969422880599717e-06, + "loss": 1.0038, + "step": 58046 + }, + { + "epoch": 4.29, + "learning_rate": 3.969104706479327e-06, + "loss": 1.0463, + "step": 58047 + }, + { + "epoch": 4.29, + "learning_rate": 3.968786541954011e-06, + "loss": 1.0986, + "step": 58048 + }, + { + "epoch": 4.29, + "learning_rate": 3.968468387024267e-06, + "loss": 0.8751, + "step": 58049 + }, + { + "epoch": 4.29, + "learning_rate": 3.9681502416906035e-06, + "loss": 0.9641, + "step": 58050 + }, + { + "epoch": 4.29, + "learning_rate": 3.967832105953524e-06, + "loss": 0.9824, + "step": 58051 + }, + { + "epoch": 4.29, + "learning_rate": 3.967513979813541e-06, + "loss": 1.0698, + "step": 58052 + }, + { + "epoch": 4.29, + "learning_rate": 3.967195863271156e-06, + "loss": 0.9033, + "step": 58053 + }, + { + "epoch": 4.29, + "learning_rate": 3.966877756326877e-06, + "loss": 1.0879, + "step": 58054 + }, + { + "epoch": 4.29, + "learning_rate": 3.966559658981208e-06, + "loss": 0.968, + "step": 58055 + }, + { + "epoch": 4.29, + "learning_rate": 3.966241571234654e-06, + "loss": 0.9926, + "step": 58056 + }, + { + "epoch": 4.29, + "learning_rate": 3.96592349308773e-06, + "loss": 1.0652, + "step": 58057 + }, + { + "epoch": 4.29, + "learning_rate": 3.965605424540928e-06, + "loss": 0.9628, + "step": 58058 + }, + { + "epoch": 4.29, + "learning_rate": 3.965287365594765e-06, + "loss": 0.9539, + "step": 58059 + }, + { + "epoch": 4.29, + "learning_rate": 3.96496931624974e-06, + "loss": 0.8589, + "step": 58060 + }, + { + "epoch": 4.29, + "learning_rate": 3.964651276506369e-06, + "loss": 0.9817, + "step": 58061 + }, + { + "epoch": 4.29, + "learning_rate": 3.964333246365145e-06, + "loss": 0.9338, + "step": 58062 + }, + { + "epoch": 4.29, + "learning_rate": 3.964015225826584e-06, + "loss": 1.0896, + "step": 58063 + }, + { + "epoch": 4.29, + "learning_rate": 3.963697214891189e-06, + "loss": 0.9326, + "step": 58064 + }, + { + "epoch": 4.29, + "learning_rate": 3.963379213559465e-06, + "loss": 1.0055, + "step": 58065 + }, + { + "epoch": 4.29, + "learning_rate": 3.963061221831918e-06, + "loss": 1.0015, + "step": 58066 + }, + { + "epoch": 4.29, + "learning_rate": 3.9627432397090505e-06, + "loss": 1.034, + "step": 58067 + }, + { + "epoch": 4.29, + "learning_rate": 3.962425267191375e-06, + "loss": 0.9895, + "step": 58068 + }, + { + "epoch": 4.29, + "learning_rate": 3.962107304279396e-06, + "loss": 0.9999, + "step": 58069 + }, + { + "epoch": 4.29, + "learning_rate": 3.961789350973617e-06, + "loss": 1.034, + "step": 58070 + }, + { + "epoch": 4.29, + "learning_rate": 3.961471407274542e-06, + "loss": 1.0594, + "step": 58071 + }, + { + "epoch": 4.29, + "learning_rate": 3.9611534731826825e-06, + "loss": 1.0095, + "step": 58072 + }, + { + "epoch": 4.29, + "learning_rate": 3.960835548698541e-06, + "loss": 0.962, + "step": 58073 + }, + { + "epoch": 4.29, + "learning_rate": 3.960517633822624e-06, + "loss": 1.0952, + "step": 58074 + }, + { + "epoch": 4.29, + "learning_rate": 3.960199728555438e-06, + "loss": 0.9765, + "step": 58075 + }, + { + "epoch": 4.29, + "learning_rate": 3.959881832897484e-06, + "loss": 0.9422, + "step": 58076 + }, + { + "epoch": 4.29, + "learning_rate": 3.959563946849279e-06, + "loss": 0.9729, + "step": 58077 + }, + { + "epoch": 4.29, + "learning_rate": 3.959246070411313e-06, + "loss": 1.0094, + "step": 58078 + }, + { + "epoch": 4.29, + "learning_rate": 3.958928203584104e-06, + "loss": 0.9778, + "step": 58079 + }, + { + "epoch": 4.29, + "learning_rate": 3.95861034636815e-06, + "loss": 0.9508, + "step": 58080 + }, + { + "epoch": 4.29, + "learning_rate": 3.958292498763968e-06, + "loss": 0.9462, + "step": 58081 + }, + { + "epoch": 4.29, + "learning_rate": 3.957974660772054e-06, + "loss": 0.9836, + "step": 58082 + }, + { + "epoch": 4.29, + "learning_rate": 3.95765683239291e-06, + "loss": 0.9786, + "step": 58083 + }, + { + "epoch": 4.29, + "learning_rate": 3.957339013627052e-06, + "loss": 0.9821, + "step": 58084 + }, + { + "epoch": 4.29, + "learning_rate": 3.957021204474981e-06, + "loss": 0.9615, + "step": 58085 + }, + { + "epoch": 4.29, + "learning_rate": 3.956703404937204e-06, + "loss": 1.0603, + "step": 58086 + }, + { + "epoch": 4.29, + "learning_rate": 3.95638561501422e-06, + "loss": 1.0914, + "step": 58087 + }, + { + "epoch": 4.29, + "learning_rate": 3.956067834706545e-06, + "loss": 0.9942, + "step": 58088 + }, + { + "epoch": 4.29, + "learning_rate": 3.955750064014679e-06, + "loss": 0.9159, + "step": 58089 + }, + { + "epoch": 4.29, + "learning_rate": 3.955432302939128e-06, + "loss": 0.9961, + "step": 58090 + }, + { + "epoch": 4.29, + "learning_rate": 3.955114551480398e-06, + "loss": 0.8739, + "step": 58091 + }, + { + "epoch": 4.29, + "learning_rate": 3.954796809638991e-06, + "loss": 1.0381, + "step": 58092 + }, + { + "epoch": 4.29, + "learning_rate": 3.954479077415418e-06, + "loss": 1.115, + "step": 58093 + }, + { + "epoch": 4.29, + "learning_rate": 3.954161354810184e-06, + "loss": 1.0244, + "step": 58094 + }, + { + "epoch": 4.29, + "learning_rate": 3.9538436418237924e-06, + "loss": 1.0025, + "step": 58095 + }, + { + "epoch": 4.29, + "learning_rate": 3.953525938456745e-06, + "loss": 0.9217, + "step": 58096 + }, + { + "epoch": 4.29, + "learning_rate": 3.953208244709556e-06, + "loss": 0.9546, + "step": 58097 + }, + { + "epoch": 4.29, + "learning_rate": 3.952890560582724e-06, + "loss": 0.9463, + "step": 58098 + }, + { + "epoch": 4.29, + "learning_rate": 3.9525728860767575e-06, + "loss": 0.9108, + "step": 58099 + }, + { + "epoch": 4.29, + "learning_rate": 3.952255221192158e-06, + "loss": 1.0482, + "step": 58100 + }, + { + "epoch": 4.29, + "learning_rate": 3.951937565929438e-06, + "loss": 1.0004, + "step": 58101 + }, + { + "epoch": 4.29, + "learning_rate": 3.9516199202891015e-06, + "loss": 1.094, + "step": 58102 + }, + { + "epoch": 4.29, + "learning_rate": 3.951302284271644e-06, + "loss": 1.0748, + "step": 58103 + }, + { + "epoch": 4.29, + "learning_rate": 3.950984657877581e-06, + "loss": 0.9617, + "step": 58104 + }, + { + "epoch": 4.29, + "learning_rate": 3.950667041107413e-06, + "loss": 1.0253, + "step": 58105 + }, + { + "epoch": 4.29, + "learning_rate": 3.950349433961654e-06, + "loss": 1.0257, + "step": 58106 + }, + { + "epoch": 4.29, + "learning_rate": 3.950031836440793e-06, + "loss": 1.0362, + "step": 58107 + }, + { + "epoch": 4.29, + "learning_rate": 3.949714248545351e-06, + "loss": 1.0138, + "step": 58108 + }, + { + "epoch": 4.29, + "learning_rate": 3.9493966702758225e-06, + "loss": 0.9566, + "step": 58109 + }, + { + "epoch": 4.29, + "learning_rate": 3.949079101632725e-06, + "loss": 1.036, + "step": 58110 + }, + { + "epoch": 4.29, + "learning_rate": 3.948761542616552e-06, + "loss": 0.9475, + "step": 58111 + }, + { + "epoch": 4.29, + "learning_rate": 3.948443993227809e-06, + "loss": 1.0308, + "step": 58112 + }, + { + "epoch": 4.29, + "learning_rate": 3.948126453467008e-06, + "loss": 1.0373, + "step": 58113 + }, + { + "epoch": 4.29, + "learning_rate": 3.9478089233346525e-06, + "loss": 1.0128, + "step": 58114 + }, + { + "epoch": 4.29, + "learning_rate": 3.947491402831246e-06, + "loss": 0.9514, + "step": 58115 + }, + { + "epoch": 4.29, + "learning_rate": 3.94717389195729e-06, + "loss": 1.0042, + "step": 58116 + }, + { + "epoch": 4.29, + "learning_rate": 3.946856390713298e-06, + "loss": 1.0147, + "step": 58117 + }, + { + "epoch": 4.29, + "learning_rate": 3.946538899099771e-06, + "loss": 0.8998, + "step": 58118 + }, + { + "epoch": 4.29, + "learning_rate": 3.946221417117213e-06, + "loss": 0.9757, + "step": 58119 + }, + { + "epoch": 4.29, + "learning_rate": 3.945903944766132e-06, + "loss": 0.9634, + "step": 58120 + }, + { + "epoch": 4.29, + "learning_rate": 3.945586482047025e-06, + "loss": 0.8991, + "step": 58121 + }, + { + "epoch": 4.29, + "learning_rate": 3.945269028960413e-06, + "loss": 1.029, + "step": 58122 + }, + { + "epoch": 4.29, + "learning_rate": 3.9449515855067836e-06, + "loss": 1.1141, + "step": 58123 + }, + { + "epoch": 4.29, + "learning_rate": 3.944634151686652e-06, + "loss": 0.9119, + "step": 58124 + }, + { + "epoch": 4.29, + "learning_rate": 3.9443167275005176e-06, + "loss": 1.0938, + "step": 58125 + }, + { + "epoch": 4.29, + "learning_rate": 3.943999312948896e-06, + "loss": 1.0064, + "step": 58126 + }, + { + "epoch": 4.29, + "learning_rate": 3.943681908032277e-06, + "loss": 1.0389, + "step": 58127 + }, + { + "epoch": 4.3, + "learning_rate": 3.943364512751179e-06, + "loss": 0.9686, + "step": 58128 + }, + { + "epoch": 4.3, + "learning_rate": 3.9430471271061e-06, + "loss": 1.0136, + "step": 58129 + }, + { + "epoch": 4.3, + "learning_rate": 3.942729751097546e-06, + "loss": 0.9682, + "step": 58130 + }, + { + "epoch": 4.3, + "learning_rate": 3.942412384726023e-06, + "loss": 0.9046, + "step": 58131 + }, + { + "epoch": 4.3, + "learning_rate": 3.942095027992031e-06, + "loss": 1.0101, + "step": 58132 + }, + { + "epoch": 4.3, + "learning_rate": 3.941777680896084e-06, + "loss": 1.1142, + "step": 58133 + }, + { + "epoch": 4.3, + "learning_rate": 3.941460343438681e-06, + "loss": 0.9842, + "step": 58134 + }, + { + "epoch": 4.3, + "learning_rate": 3.941143015620328e-06, + "loss": 0.9611, + "step": 58135 + }, + { + "epoch": 4.3, + "learning_rate": 3.940825697441527e-06, + "loss": 1.0381, + "step": 58136 + }, + { + "epoch": 4.3, + "learning_rate": 3.940508388902788e-06, + "loss": 0.8602, + "step": 58137 + }, + { + "epoch": 4.3, + "learning_rate": 3.9401910900046134e-06, + "loss": 0.9728, + "step": 58138 + }, + { + "epoch": 4.3, + "learning_rate": 3.939873800747509e-06, + "loss": 0.9325, + "step": 58139 + }, + { + "epoch": 4.3, + "learning_rate": 3.939556521131978e-06, + "loss": 0.9225, + "step": 58140 + }, + { + "epoch": 4.3, + "learning_rate": 3.939239251158523e-06, + "loss": 0.9122, + "step": 58141 + }, + { + "epoch": 4.3, + "learning_rate": 3.938921990827655e-06, + "loss": 0.9572, + "step": 58142 + }, + { + "epoch": 4.3, + "learning_rate": 3.938604740139875e-06, + "loss": 1.0387, + "step": 58143 + }, + { + "epoch": 4.3, + "learning_rate": 3.938287499095689e-06, + "loss": 0.9593, + "step": 58144 + }, + { + "epoch": 4.3, + "learning_rate": 3.937970267695596e-06, + "loss": 0.9729, + "step": 58145 + }, + { + "epoch": 4.3, + "learning_rate": 3.937653045940111e-06, + "loss": 0.9412, + "step": 58146 + }, + { + "epoch": 4.3, + "learning_rate": 3.937335833829732e-06, + "loss": 1.0023, + "step": 58147 + }, + { + "epoch": 4.3, + "learning_rate": 3.937018631364964e-06, + "loss": 0.9201, + "step": 58148 + }, + { + "epoch": 4.3, + "learning_rate": 3.9367014385463145e-06, + "loss": 0.9386, + "step": 58149 + }, + { + "epoch": 4.3, + "learning_rate": 3.936384255374281e-06, + "loss": 1.0578, + "step": 58150 + }, + { + "epoch": 4.3, + "learning_rate": 3.9360670818493816e-06, + "loss": 0.998, + "step": 58151 + }, + { + "epoch": 4.3, + "learning_rate": 3.935749917972106e-06, + "loss": 1.0805, + "step": 58152 + }, + { + "epoch": 4.3, + "learning_rate": 3.935432763742968e-06, + "loss": 1.0087, + "step": 58153 + }, + { + "epoch": 4.3, + "learning_rate": 3.935115619162469e-06, + "loss": 1.1424, + "step": 58154 + }, + { + "epoch": 4.3, + "learning_rate": 3.934798484231116e-06, + "loss": 0.9256, + "step": 58155 + }, + { + "epoch": 4.3, + "learning_rate": 3.934481358949408e-06, + "loss": 1.0345, + "step": 58156 + }, + { + "epoch": 4.3, + "learning_rate": 3.934164243317856e-06, + "loss": 0.9408, + "step": 58157 + }, + { + "epoch": 4.3, + "learning_rate": 3.933847137336964e-06, + "loss": 1.0272, + "step": 58158 + }, + { + "epoch": 4.3, + "learning_rate": 3.933530041007232e-06, + "loss": 1.032, + "step": 58159 + }, + { + "epoch": 4.3, + "learning_rate": 3.933212954329167e-06, + "loss": 0.9811, + "step": 58160 + }, + { + "epoch": 4.3, + "learning_rate": 3.932895877303271e-06, + "loss": 0.8816, + "step": 58161 + }, + { + "epoch": 4.3, + "learning_rate": 3.932578809930054e-06, + "loss": 0.9981, + "step": 58162 + }, + { + "epoch": 4.3, + "learning_rate": 3.932261752210018e-06, + "loss": 0.937, + "step": 58163 + }, + { + "epoch": 4.3, + "learning_rate": 3.931944704143666e-06, + "loss": 0.9653, + "step": 58164 + }, + { + "epoch": 4.3, + "learning_rate": 3.9316276657315e-06, + "loss": 1.0195, + "step": 58165 + }, + { + "epoch": 4.3, + "learning_rate": 3.93131063697403e-06, + "loss": 1.0487, + "step": 58166 + }, + { + "epoch": 4.3, + "learning_rate": 3.930993617871762e-06, + "loss": 0.9672, + "step": 58167 + }, + { + "epoch": 4.3, + "learning_rate": 3.93067660842519e-06, + "loss": 1.0136, + "step": 58168 + }, + { + "epoch": 4.3, + "learning_rate": 3.930359608634828e-06, + "loss": 0.9688, + "step": 58169 + }, + { + "epoch": 4.3, + "learning_rate": 3.930042618501173e-06, + "loss": 0.9108, + "step": 58170 + }, + { + "epoch": 4.3, + "learning_rate": 3.929725638024741e-06, + "loss": 1.0166, + "step": 58171 + }, + { + "epoch": 4.3, + "learning_rate": 3.9294086672060215e-06, + "loss": 0.881, + "step": 58172 + }, + { + "epoch": 4.3, + "learning_rate": 3.929091706045529e-06, + "loss": 0.8955, + "step": 58173 + }, + { + "epoch": 4.3, + "learning_rate": 3.928774754543761e-06, + "loss": 0.9875, + "step": 58174 + }, + { + "epoch": 4.3, + "learning_rate": 3.928457812701233e-06, + "loss": 1.0601, + "step": 58175 + }, + { + "epoch": 4.3, + "learning_rate": 3.928140880518438e-06, + "loss": 1.0777, + "step": 58176 + }, + { + "epoch": 4.3, + "learning_rate": 3.927823957995881e-06, + "loss": 1.0257, + "step": 58177 + }, + { + "epoch": 4.3, + "learning_rate": 3.927507045134073e-06, + "loss": 0.903, + "step": 58178 + }, + { + "epoch": 4.3, + "learning_rate": 3.927190141933513e-06, + "loss": 1.0638, + "step": 58179 + }, + { + "epoch": 4.3, + "learning_rate": 3.926873248394708e-06, + "loss": 1.014, + "step": 58180 + }, + { + "epoch": 4.3, + "learning_rate": 3.926556364518157e-06, + "loss": 0.9403, + "step": 58181 + }, + { + "epoch": 4.3, + "learning_rate": 3.926239490304371e-06, + "loss": 1.0889, + "step": 58182 + }, + { + "epoch": 4.3, + "learning_rate": 3.925922625753851e-06, + "loss": 0.8799, + "step": 58183 + }, + { + "epoch": 4.3, + "learning_rate": 3.925605770867101e-06, + "loss": 0.9299, + "step": 58184 + }, + { + "epoch": 4.3, + "learning_rate": 3.925288925644626e-06, + "loss": 1.0367, + "step": 58185 + }, + { + "epoch": 4.3, + "learning_rate": 3.924972090086926e-06, + "loss": 0.9952, + "step": 58186 + }, + { + "epoch": 4.3, + "learning_rate": 3.924655264194511e-06, + "loss": 1.0346, + "step": 58187 + }, + { + "epoch": 4.3, + "learning_rate": 3.924338447967883e-06, + "loss": 1.1069, + "step": 58188 + }, + { + "epoch": 4.3, + "learning_rate": 3.924021641407546e-06, + "loss": 1.0874, + "step": 58189 + }, + { + "epoch": 4.3, + "learning_rate": 3.9237048445139995e-06, + "loss": 0.9049, + "step": 58190 + }, + { + "epoch": 4.3, + "learning_rate": 3.9233880572877555e-06, + "loss": 0.986, + "step": 58191 + }, + { + "epoch": 4.3, + "learning_rate": 3.923071279729314e-06, + "loss": 0.9281, + "step": 58192 + }, + { + "epoch": 4.3, + "learning_rate": 3.92275451183918e-06, + "loss": 1.0182, + "step": 58193 + }, + { + "epoch": 4.3, + "learning_rate": 3.922437753617856e-06, + "loss": 1.0189, + "step": 58194 + }, + { + "epoch": 4.3, + "learning_rate": 3.922121005065843e-06, + "loss": 1.0882, + "step": 58195 + }, + { + "epoch": 4.3, + "learning_rate": 3.921804266183656e-06, + "loss": 0.9475, + "step": 58196 + }, + { + "epoch": 4.3, + "learning_rate": 3.921487536971783e-06, + "loss": 0.9939, + "step": 58197 + }, + { + "epoch": 4.3, + "learning_rate": 3.921170817430741e-06, + "loss": 1.0366, + "step": 58198 + }, + { + "epoch": 4.3, + "learning_rate": 3.920854107561029e-06, + "loss": 1.0511, + "step": 58199 + }, + { + "epoch": 4.3, + "learning_rate": 3.920537407363151e-06, + "loss": 0.9806, + "step": 58200 + }, + { + "epoch": 4.3, + "learning_rate": 3.920220716837608e-06, + "loss": 1.1162, + "step": 58201 + }, + { + "epoch": 4.3, + "learning_rate": 3.91990403598491e-06, + "loss": 1.1081, + "step": 58202 + }, + { + "epoch": 4.3, + "learning_rate": 3.919587364805558e-06, + "loss": 0.9606, + "step": 58203 + }, + { + "epoch": 4.3, + "learning_rate": 3.9192707033000555e-06, + "loss": 1.1144, + "step": 58204 + }, + { + "epoch": 4.3, + "learning_rate": 3.918954051468906e-06, + "loss": 1.0307, + "step": 58205 + }, + { + "epoch": 4.3, + "learning_rate": 3.91863740931261e-06, + "loss": 0.8813, + "step": 58206 + }, + { + "epoch": 4.3, + "learning_rate": 3.918320776831679e-06, + "loss": 1.0057, + "step": 58207 + }, + { + "epoch": 4.3, + "learning_rate": 3.918004154026612e-06, + "loss": 1.0147, + "step": 58208 + }, + { + "epoch": 4.3, + "learning_rate": 3.917687540897913e-06, + "loss": 0.9492, + "step": 58209 + }, + { + "epoch": 4.3, + "learning_rate": 3.917370937446083e-06, + "loss": 0.884, + "step": 58210 + }, + { + "epoch": 4.3, + "learning_rate": 3.917054343671632e-06, + "loss": 0.9445, + "step": 58211 + }, + { + "epoch": 4.3, + "learning_rate": 3.916737759575062e-06, + "loss": 0.9562, + "step": 58212 + }, + { + "epoch": 4.3, + "learning_rate": 3.916421185156874e-06, + "loss": 0.9956, + "step": 58213 + }, + { + "epoch": 4.3, + "learning_rate": 3.916104620417574e-06, + "loss": 1.0365, + "step": 58214 + }, + { + "epoch": 4.3, + "learning_rate": 3.91578806535766e-06, + "loss": 1.1067, + "step": 58215 + }, + { + "epoch": 4.3, + "learning_rate": 3.915471519977646e-06, + "loss": 0.9433, + "step": 58216 + }, + { + "epoch": 4.3, + "learning_rate": 3.9151549842780246e-06, + "loss": 1.0416, + "step": 58217 + }, + { + "epoch": 4.3, + "learning_rate": 3.914838458259307e-06, + "loss": 1.0921, + "step": 58218 + }, + { + "epoch": 4.3, + "learning_rate": 3.914521941921991e-06, + "loss": 1.091, + "step": 58219 + }, + { + "epoch": 4.3, + "learning_rate": 3.914205435266592e-06, + "loss": 0.9778, + "step": 58220 + }, + { + "epoch": 4.3, + "learning_rate": 3.913888938293596e-06, + "loss": 0.9874, + "step": 58221 + }, + { + "epoch": 4.3, + "learning_rate": 3.91357245100352e-06, + "loss": 0.9757, + "step": 58222 + }, + { + "epoch": 4.3, + "learning_rate": 3.9132559733968635e-06, + "loss": 1.043, + "step": 58223 + }, + { + "epoch": 4.3, + "learning_rate": 3.912939505474128e-06, + "loss": 0.9624, + "step": 58224 + }, + { + "epoch": 4.3, + "learning_rate": 3.912623047235821e-06, + "loss": 0.9189, + "step": 58225 + }, + { + "epoch": 4.3, + "learning_rate": 3.9123065986824386e-06, + "loss": 0.9284, + "step": 58226 + }, + { + "epoch": 4.3, + "learning_rate": 3.911990159814493e-06, + "loss": 1.0063, + "step": 58227 + }, + { + "epoch": 4.3, + "learning_rate": 3.911673730632485e-06, + "loss": 0.9963, + "step": 58228 + }, + { + "epoch": 4.3, + "learning_rate": 3.9113573111369155e-06, + "loss": 1.0733, + "step": 58229 + }, + { + "epoch": 4.3, + "learning_rate": 3.911040901328287e-06, + "loss": 0.9559, + "step": 58230 + }, + { + "epoch": 4.3, + "learning_rate": 3.910724501207108e-06, + "loss": 0.9766, + "step": 58231 + }, + { + "epoch": 4.3, + "learning_rate": 3.91040811077388e-06, + "loss": 0.9838, + "step": 58232 + }, + { + "epoch": 4.3, + "learning_rate": 3.910091730029105e-06, + "loss": 0.9939, + "step": 58233 + }, + { + "epoch": 4.3, + "learning_rate": 3.909775358973287e-06, + "loss": 0.9865, + "step": 58234 + }, + { + "epoch": 4.3, + "learning_rate": 3.909458997606925e-06, + "loss": 0.9811, + "step": 58235 + }, + { + "epoch": 4.3, + "learning_rate": 3.909142645930532e-06, + "loss": 0.9029, + "step": 58236 + }, + { + "epoch": 4.3, + "learning_rate": 3.908826303944604e-06, + "loss": 1.0223, + "step": 58237 + }, + { + "epoch": 4.3, + "learning_rate": 3.908509971649648e-06, + "loss": 1.0783, + "step": 58238 + }, + { + "epoch": 4.3, + "learning_rate": 3.90819364904616e-06, + "loss": 0.9136, + "step": 58239 + }, + { + "epoch": 4.3, + "learning_rate": 3.907877336134653e-06, + "loss": 0.9168, + "step": 58240 + }, + { + "epoch": 4.3, + "learning_rate": 3.907561032915631e-06, + "loss": 0.9999, + "step": 58241 + }, + { + "epoch": 4.3, + "learning_rate": 3.907244739389584e-06, + "loss": 1.025, + "step": 58242 + }, + { + "epoch": 4.3, + "learning_rate": 3.906928455557026e-06, + "loss": 1.165, + "step": 58243 + }, + { + "epoch": 4.3, + "learning_rate": 3.906612181418459e-06, + "loss": 0.9992, + "step": 58244 + }, + { + "epoch": 4.3, + "learning_rate": 3.9062959169743846e-06, + "loss": 1.0188, + "step": 58245 + }, + { + "epoch": 4.3, + "learning_rate": 3.905979662225302e-06, + "loss": 1.0683, + "step": 58246 + }, + { + "epoch": 4.3, + "learning_rate": 3.9056634171717235e-06, + "loss": 1.1014, + "step": 58247 + }, + { + "epoch": 4.3, + "learning_rate": 3.9053471818141465e-06, + "loss": 0.9951, + "step": 58248 + }, + { + "epoch": 4.3, + "learning_rate": 3.905030956153074e-06, + "loss": 1.0038, + "step": 58249 + }, + { + "epoch": 4.3, + "learning_rate": 3.904714740189012e-06, + "loss": 0.9832, + "step": 58250 + }, + { + "epoch": 4.3, + "learning_rate": 3.904398533922457e-06, + "loss": 0.9791, + "step": 58251 + }, + { + "epoch": 4.3, + "learning_rate": 3.904082337353919e-06, + "loss": 1.0241, + "step": 58252 + }, + { + "epoch": 4.3, + "learning_rate": 3.903766150483901e-06, + "loss": 1.1554, + "step": 58253 + }, + { + "epoch": 4.3, + "learning_rate": 3.903449973312903e-06, + "loss": 0.9996, + "step": 58254 + }, + { + "epoch": 4.3, + "learning_rate": 3.903133805841424e-06, + "loss": 1.0551, + "step": 58255 + }, + { + "epoch": 4.3, + "learning_rate": 3.902817648069977e-06, + "loss": 0.9714, + "step": 58256 + }, + { + "epoch": 4.3, + "learning_rate": 3.902501499999059e-06, + "loss": 0.9439, + "step": 58257 + }, + { + "epoch": 4.3, + "learning_rate": 3.902185361629175e-06, + "loss": 1.0316, + "step": 58258 + }, + { + "epoch": 4.3, + "learning_rate": 3.901869232960823e-06, + "loss": 1.082, + "step": 58259 + }, + { + "epoch": 4.3, + "learning_rate": 3.901553113994513e-06, + "loss": 0.9863, + "step": 58260 + }, + { + "epoch": 4.3, + "learning_rate": 3.901237004730748e-06, + "loss": 1.0373, + "step": 58261 + }, + { + "epoch": 4.3, + "learning_rate": 3.900920905170022e-06, + "loss": 0.9958, + "step": 58262 + }, + { + "epoch": 4.31, + "learning_rate": 3.9006048153128454e-06, + "loss": 0.9762, + "step": 58263 + }, + { + "epoch": 4.31, + "learning_rate": 3.900288735159717e-06, + "loss": 1.0241, + "step": 58264 + }, + { + "epoch": 4.31, + "learning_rate": 3.899972664711147e-06, + "loss": 0.923, + "step": 58265 + }, + { + "epoch": 4.31, + "learning_rate": 3.899656603967628e-06, + "loss": 1.0071, + "step": 58266 + }, + { + "epoch": 4.31, + "learning_rate": 3.899340552929671e-06, + "loss": 0.9738, + "step": 58267 + }, + { + "epoch": 4.31, + "learning_rate": 3.899024511597772e-06, + "loss": 0.9132, + "step": 58268 + }, + { + "epoch": 4.31, + "learning_rate": 3.898708479972445e-06, + "loss": 1.0114, + "step": 58269 + }, + { + "epoch": 4.31, + "learning_rate": 3.898392458054183e-06, + "loss": 0.9676, + "step": 58270 + }, + { + "epoch": 4.31, + "learning_rate": 3.898076445843487e-06, + "loss": 1.0886, + "step": 58271 + }, + { + "epoch": 4.31, + "learning_rate": 3.897760443340868e-06, + "loss": 1.043, + "step": 58272 + }, + { + "epoch": 4.31, + "learning_rate": 3.8974444505468235e-06, + "loss": 1.0411, + "step": 58273 + }, + { + "epoch": 4.31, + "learning_rate": 3.8971284674618595e-06, + "loss": 0.9679, + "step": 58274 + }, + { + "epoch": 4.31, + "learning_rate": 3.896812494086471e-06, + "loss": 0.9995, + "step": 58275 + }, + { + "epoch": 4.31, + "learning_rate": 3.8964965304211725e-06, + "loss": 0.9798, + "step": 58276 + }, + { + "epoch": 4.31, + "learning_rate": 3.896180576466458e-06, + "loss": 1.0698, + "step": 58277 + }, + { + "epoch": 4.31, + "learning_rate": 3.895864632222836e-06, + "loss": 0.9057, + "step": 58278 + }, + { + "epoch": 4.31, + "learning_rate": 3.895548697690803e-06, + "loss": 1.0828, + "step": 58279 + }, + { + "epoch": 4.31, + "learning_rate": 3.895232772870863e-06, + "loss": 0.9982, + "step": 58280 + }, + { + "epoch": 4.31, + "learning_rate": 3.894916857763523e-06, + "loss": 1.0137, + "step": 58281 + }, + { + "epoch": 4.31, + "learning_rate": 3.894600952369284e-06, + "loss": 0.9238, + "step": 58282 + }, + { + "epoch": 4.31, + "learning_rate": 3.894285056688647e-06, + "loss": 0.9956, + "step": 58283 + }, + { + "epoch": 4.31, + "learning_rate": 3.893969170722111e-06, + "loss": 1.0091, + "step": 58284 + }, + { + "epoch": 4.31, + "learning_rate": 3.893653294470187e-06, + "loss": 1.0214, + "step": 58285 + }, + { + "epoch": 4.31, + "learning_rate": 3.893337427933373e-06, + "loss": 0.9503, + "step": 58286 + }, + { + "epoch": 4.31, + "learning_rate": 3.893021571112171e-06, + "loss": 0.9877, + "step": 58287 + }, + { + "epoch": 4.31, + "learning_rate": 3.8927057240070855e-06, + "loss": 0.9214, + "step": 58288 + }, + { + "epoch": 4.31, + "learning_rate": 3.892389886618617e-06, + "loss": 0.9981, + "step": 58289 + }, + { + "epoch": 4.31, + "learning_rate": 3.892074058947268e-06, + "loss": 0.9513, + "step": 58290 + }, + { + "epoch": 4.31, + "learning_rate": 3.891758240993541e-06, + "loss": 1.069, + "step": 58291 + }, + { + "epoch": 4.31, + "learning_rate": 3.891442432757941e-06, + "loss": 0.9193, + "step": 58292 + }, + { + "epoch": 4.31, + "learning_rate": 3.891126634240968e-06, + "loss": 0.9581, + "step": 58293 + }, + { + "epoch": 4.31, + "learning_rate": 3.890810845443126e-06, + "loss": 1.0703, + "step": 58294 + }, + { + "epoch": 4.31, + "learning_rate": 3.890495066364914e-06, + "loss": 1.0456, + "step": 58295 + }, + { + "epoch": 4.31, + "learning_rate": 3.890179297006838e-06, + "loss": 1.0042, + "step": 58296 + }, + { + "epoch": 4.31, + "learning_rate": 3.8898635373694e-06, + "loss": 0.974, + "step": 58297 + }, + { + "epoch": 4.31, + "learning_rate": 3.889547787453103e-06, + "loss": 0.9827, + "step": 58298 + }, + { + "epoch": 4.31, + "learning_rate": 3.889232047258447e-06, + "loss": 1.0083, + "step": 58299 + }, + { + "epoch": 4.31, + "learning_rate": 3.8889163167859335e-06, + "loss": 1.0531, + "step": 58300 + }, + { + "epoch": 4.31, + "learning_rate": 3.888600596036069e-06, + "loss": 0.9325, + "step": 58301 + }, + { + "epoch": 4.31, + "learning_rate": 3.888284885009354e-06, + "loss": 0.9345, + "step": 58302 + }, + { + "epoch": 4.31, + "learning_rate": 3.8879691837062895e-06, + "loss": 0.8989, + "step": 58303 + }, + { + "epoch": 4.31, + "learning_rate": 3.887653492127375e-06, + "loss": 1.0176, + "step": 58304 + }, + { + "epoch": 4.31, + "learning_rate": 3.88733781027312e-06, + "loss": 1.0551, + "step": 58305 + }, + { + "epoch": 4.31, + "learning_rate": 3.887022138144027e-06, + "loss": 0.875, + "step": 58306 + }, + { + "epoch": 4.31, + "learning_rate": 3.886706475740587e-06, + "loss": 1.0387, + "step": 58307 + }, + { + "epoch": 4.31, + "learning_rate": 3.886390823063313e-06, + "loss": 1.0103, + "step": 58308 + }, + { + "epoch": 4.31, + "learning_rate": 3.8860751801127e-06, + "loss": 0.9368, + "step": 58309 + }, + { + "epoch": 4.31, + "learning_rate": 3.8857595468892615e-06, + "loss": 1.1008, + "step": 58310 + }, + { + "epoch": 4.31, + "learning_rate": 3.885443923393485e-06, + "loss": 0.9791, + "step": 58311 + }, + { + "epoch": 4.31, + "learning_rate": 3.885128309625884e-06, + "loss": 1.0496, + "step": 58312 + }, + { + "epoch": 4.31, + "learning_rate": 3.884812705586951e-06, + "loss": 1.0625, + "step": 58313 + }, + { + "epoch": 4.31, + "learning_rate": 3.884497111277203e-06, + "loss": 0.9705, + "step": 58314 + }, + { + "epoch": 4.31, + "learning_rate": 3.884181526697123e-06, + "loss": 1.0897, + "step": 58315 + }, + { + "epoch": 4.31, + "learning_rate": 3.883865951847228e-06, + "loss": 1.0399, + "step": 58316 + }, + { + "epoch": 4.31, + "learning_rate": 3.883550386728014e-06, + "loss": 0.9719, + "step": 58317 + }, + { + "epoch": 4.31, + "learning_rate": 3.883234831339984e-06, + "loss": 0.9488, + "step": 58318 + }, + { + "epoch": 4.31, + "learning_rate": 3.8829192856836395e-06, + "loss": 0.9243, + "step": 58319 + }, + { + "epoch": 4.31, + "learning_rate": 3.882603749759479e-06, + "loss": 1.0142, + "step": 58320 + }, + { + "epoch": 4.31, + "learning_rate": 3.882288223568013e-06, + "loss": 1.0625, + "step": 58321 + }, + { + "epoch": 4.31, + "learning_rate": 3.881972707109738e-06, + "loss": 0.9474, + "step": 58322 + }, + { + "epoch": 4.31, + "learning_rate": 3.881657200385157e-06, + "loss": 1.0207, + "step": 58323 + }, + { + "epoch": 4.31, + "learning_rate": 3.88134170339477e-06, + "loss": 1.053, + "step": 58324 + }, + { + "epoch": 4.31, + "learning_rate": 3.881026216139082e-06, + "loss": 0.9706, + "step": 58325 + }, + { + "epoch": 4.31, + "learning_rate": 3.880710738618595e-06, + "loss": 1.0826, + "step": 58326 + }, + { + "epoch": 4.31, + "learning_rate": 3.88039527083381e-06, + "loss": 1.1034, + "step": 58327 + }, + { + "epoch": 4.31, + "learning_rate": 3.880079812785228e-06, + "loss": 0.9658, + "step": 58328 + }, + { + "epoch": 4.31, + "learning_rate": 3.879764364473347e-06, + "loss": 0.8259, + "step": 58329 + }, + { + "epoch": 4.31, + "learning_rate": 3.879448925898678e-06, + "loss": 1.0454, + "step": 58330 + }, + { + "epoch": 4.31, + "learning_rate": 3.879133497061718e-06, + "loss": 1.046, + "step": 58331 + }, + { + "epoch": 4.31, + "learning_rate": 3.878818077962969e-06, + "loss": 1.0266, + "step": 58332 + }, + { + "epoch": 4.31, + "learning_rate": 3.878502668602929e-06, + "loss": 1.0017, + "step": 58333 + }, + { + "epoch": 4.31, + "learning_rate": 3.878187268982111e-06, + "loss": 1.0098, + "step": 58334 + }, + { + "epoch": 4.31, + "learning_rate": 3.877871879101006e-06, + "loss": 1.0774, + "step": 58335 + }, + { + "epoch": 4.31, + "learning_rate": 3.8775564989601156e-06, + "loss": 0.9605, + "step": 58336 + }, + { + "epoch": 4.31, + "learning_rate": 3.877241128559949e-06, + "loss": 1.0488, + "step": 58337 + }, + { + "epoch": 4.31, + "learning_rate": 3.876925767901004e-06, + "loss": 0.8883, + "step": 58338 + }, + { + "epoch": 4.31, + "learning_rate": 3.876610416983783e-06, + "loss": 0.9615, + "step": 58339 + }, + { + "epoch": 4.31, + "learning_rate": 3.876295075808782e-06, + "loss": 0.9904, + "step": 58340 + }, + { + "epoch": 4.31, + "learning_rate": 3.875979744376514e-06, + "loss": 1.0045, + "step": 58341 + }, + { + "epoch": 4.31, + "learning_rate": 3.875664422687473e-06, + "loss": 0.9337, + "step": 58342 + }, + { + "epoch": 4.31, + "learning_rate": 3.875349110742162e-06, + "loss": 0.8806, + "step": 58343 + }, + { + "epoch": 4.31, + "learning_rate": 3.875033808541083e-06, + "loss": 0.9697, + "step": 58344 + }, + { + "epoch": 4.31, + "learning_rate": 3.874718516084736e-06, + "loss": 0.944, + "step": 58345 + }, + { + "epoch": 4.31, + "learning_rate": 3.874403233373626e-06, + "loss": 0.9304, + "step": 58346 + }, + { + "epoch": 4.31, + "learning_rate": 3.874087960408252e-06, + "loss": 1.065, + "step": 58347 + }, + { + "epoch": 4.31, + "learning_rate": 3.8737726971891175e-06, + "loss": 0.9922, + "step": 58348 + }, + { + "epoch": 4.31, + "learning_rate": 3.87345744371672e-06, + "loss": 1.1087, + "step": 58349 + }, + { + "epoch": 4.31, + "learning_rate": 3.873142199991566e-06, + "loss": 0.9556, + "step": 58350 + }, + { + "epoch": 4.31, + "learning_rate": 3.872826966014157e-06, + "loss": 1.0604, + "step": 58351 + }, + { + "epoch": 4.31, + "learning_rate": 3.872511741784992e-06, + "loss": 1.002, + "step": 58352 + }, + { + "epoch": 4.31, + "learning_rate": 3.8721965273045725e-06, + "loss": 0.9516, + "step": 58353 + }, + { + "epoch": 4.31, + "learning_rate": 3.871881322573397e-06, + "loss": 0.8795, + "step": 58354 + }, + { + "epoch": 4.31, + "learning_rate": 3.8715661275919794e-06, + "loss": 0.8996, + "step": 58355 + }, + { + "epoch": 4.31, + "learning_rate": 3.871250942360804e-06, + "loss": 0.9487, + "step": 58356 + }, + { + "epoch": 4.31, + "learning_rate": 3.870935766880384e-06, + "loss": 0.9633, + "step": 58357 + }, + { + "epoch": 4.31, + "learning_rate": 3.8706206011512135e-06, + "loss": 0.919, + "step": 58358 + }, + { + "epoch": 4.31, + "learning_rate": 3.870305445173806e-06, + "loss": 0.918, + "step": 58359 + }, + { + "epoch": 4.31, + "learning_rate": 3.8699902989486485e-06, + "loss": 0.9595, + "step": 58360 + }, + { + "epoch": 4.31, + "learning_rate": 3.869675162476251e-06, + "loss": 0.9645, + "step": 58361 + }, + { + "epoch": 4.31, + "learning_rate": 3.869360035757109e-06, + "loss": 0.9233, + "step": 58362 + }, + { + "epoch": 4.31, + "learning_rate": 3.869044918791735e-06, + "loss": 1.0714, + "step": 58363 + }, + { + "epoch": 4.31, + "learning_rate": 3.86872981158062e-06, + "loss": 0.9073, + "step": 58364 + }, + { + "epoch": 4.31, + "learning_rate": 3.868414714124263e-06, + "loss": 0.9443, + "step": 58365 + }, + { + "epoch": 4.31, + "learning_rate": 3.8680996264231755e-06, + "loss": 0.9615, + "step": 58366 + }, + { + "epoch": 4.31, + "learning_rate": 3.867784548477852e-06, + "loss": 1.0856, + "step": 58367 + }, + { + "epoch": 4.31, + "learning_rate": 3.867469480288797e-06, + "loss": 1.0554, + "step": 58368 + }, + { + "epoch": 4.31, + "learning_rate": 3.867154421856506e-06, + "loss": 0.9068, + "step": 58369 + }, + { + "epoch": 4.31, + "learning_rate": 3.866839373181489e-06, + "loss": 0.9695, + "step": 58370 + }, + { + "epoch": 4.31, + "learning_rate": 3.866524334264243e-06, + "loss": 0.9123, + "step": 58371 + }, + { + "epoch": 4.31, + "learning_rate": 3.866209305105267e-06, + "loss": 0.935, + "step": 58372 + }, + { + "epoch": 4.31, + "learning_rate": 3.865894285705066e-06, + "loss": 1.0585, + "step": 58373 + }, + { + "epoch": 4.31, + "learning_rate": 3.865579276064135e-06, + "loss": 1.0315, + "step": 58374 + }, + { + "epoch": 4.31, + "learning_rate": 3.865264276182984e-06, + "loss": 0.979, + "step": 58375 + }, + { + "epoch": 4.31, + "learning_rate": 3.86494928606211e-06, + "loss": 0.9778, + "step": 58376 + }, + { + "epoch": 4.31, + "learning_rate": 3.864634305702013e-06, + "loss": 1.0288, + "step": 58377 + }, + { + "epoch": 4.31, + "learning_rate": 3.864319335103191e-06, + "loss": 0.9079, + "step": 58378 + }, + { + "epoch": 4.31, + "learning_rate": 3.864004374266158e-06, + "loss": 1.0475, + "step": 58379 + }, + { + "epoch": 4.31, + "learning_rate": 3.863689423191398e-06, + "loss": 1.1207, + "step": 58380 + }, + { + "epoch": 4.31, + "learning_rate": 3.863374481879424e-06, + "loss": 0.9324, + "step": 58381 + }, + { + "epoch": 4.31, + "learning_rate": 3.863059550330734e-06, + "loss": 0.9881, + "step": 58382 + }, + { + "epoch": 4.31, + "learning_rate": 3.862744628545827e-06, + "loss": 1.0178, + "step": 58383 + }, + { + "epoch": 4.31, + "learning_rate": 3.8624297165252065e-06, + "loss": 0.9786, + "step": 58384 + }, + { + "epoch": 4.31, + "learning_rate": 3.862114814269369e-06, + "loss": 0.9745, + "step": 58385 + }, + { + "epoch": 4.31, + "learning_rate": 3.861799921778822e-06, + "loss": 1.0186, + "step": 58386 + }, + { + "epoch": 4.31, + "learning_rate": 3.861485039054063e-06, + "loss": 1.0469, + "step": 58387 + }, + { + "epoch": 4.31, + "learning_rate": 3.861170166095596e-06, + "loss": 1.0575, + "step": 58388 + }, + { + "epoch": 4.31, + "learning_rate": 3.860855302903914e-06, + "loss": 1.0472, + "step": 58389 + }, + { + "epoch": 4.31, + "learning_rate": 3.860540449479528e-06, + "loss": 0.9965, + "step": 58390 + }, + { + "epoch": 4.31, + "learning_rate": 3.860225605822933e-06, + "loss": 1.0468, + "step": 58391 + }, + { + "epoch": 4.31, + "learning_rate": 3.859910771934631e-06, + "loss": 0.9248, + "step": 58392 + }, + { + "epoch": 4.31, + "learning_rate": 3.8595959478151245e-06, + "loss": 0.9555, + "step": 58393 + }, + { + "epoch": 4.31, + "learning_rate": 3.859281133464908e-06, + "loss": 1.0532, + "step": 58394 + }, + { + "epoch": 4.31, + "learning_rate": 3.858966328884493e-06, + "loss": 1.0023, + "step": 58395 + }, + { + "epoch": 4.31, + "learning_rate": 3.858651534074373e-06, + "loss": 0.9819, + "step": 58396 + }, + { + "epoch": 4.31, + "learning_rate": 3.8583367490350496e-06, + "loss": 0.91, + "step": 58397 + }, + { + "epoch": 4.31, + "learning_rate": 3.858021973767023e-06, + "loss": 0.9958, + "step": 58398 + }, + { + "epoch": 4.32, + "learning_rate": 3.857707208270799e-06, + "loss": 1.035, + "step": 58399 + }, + { + "epoch": 4.32, + "learning_rate": 3.857392452546877e-06, + "loss": 1.0135, + "step": 58400 + }, + { + "epoch": 4.32, + "learning_rate": 3.85707770659575e-06, + "loss": 1.0741, + "step": 58401 + }, + { + "epoch": 4.32, + "learning_rate": 3.856762970417928e-06, + "loss": 0.9821, + "step": 58402 + }, + { + "epoch": 4.32, + "learning_rate": 3.856448244013904e-06, + "loss": 1.0876, + "step": 58403 + }, + { + "epoch": 4.32, + "learning_rate": 3.856133527384191e-06, + "loss": 1.0802, + "step": 58404 + }, + { + "epoch": 4.32, + "learning_rate": 3.8558188205292724e-06, + "loss": 0.9839, + "step": 58405 + }, + { + "epoch": 4.32, + "learning_rate": 3.855504123449664e-06, + "loss": 0.9755, + "step": 58406 + }, + { + "epoch": 4.32, + "learning_rate": 3.855189436145857e-06, + "loss": 0.9458, + "step": 58407 + }, + { + "epoch": 4.32, + "learning_rate": 3.8548747586183614e-06, + "loss": 1.0032, + "step": 58408 + }, + { + "epoch": 4.32, + "learning_rate": 3.8545600908676686e-06, + "loss": 1.0375, + "step": 58409 + }, + { + "epoch": 4.32, + "learning_rate": 3.85424543289428e-06, + "loss": 1.1049, + "step": 58410 + }, + { + "epoch": 4.32, + "learning_rate": 3.853930784698702e-06, + "loss": 1.0281, + "step": 58411 + }, + { + "epoch": 4.32, + "learning_rate": 3.853616146281432e-06, + "loss": 1.0052, + "step": 58412 + }, + { + "epoch": 4.32, + "learning_rate": 3.853301517642972e-06, + "loss": 0.9873, + "step": 58413 + }, + { + "epoch": 4.32, + "learning_rate": 3.852986898783817e-06, + "loss": 0.9085, + "step": 58414 + }, + { + "epoch": 4.32, + "learning_rate": 3.852672289704476e-06, + "loss": 1.0229, + "step": 58415 + }, + { + "epoch": 4.32, + "learning_rate": 3.852357690405445e-06, + "loss": 0.88, + "step": 58416 + }, + { + "epoch": 4.32, + "learning_rate": 3.852043100887225e-06, + "loss": 0.9684, + "step": 58417 + }, + { + "epoch": 4.32, + "learning_rate": 3.851728521150313e-06, + "loss": 1.0398, + "step": 58418 + }, + { + "epoch": 4.32, + "learning_rate": 3.851413951195216e-06, + "loss": 0.9585, + "step": 58419 + }, + { + "epoch": 4.32, + "learning_rate": 3.851099391022435e-06, + "loss": 1.1123, + "step": 58420 + }, + { + "epoch": 4.32, + "learning_rate": 3.85078484063246e-06, + "loss": 1.0327, + "step": 58421 + }, + { + "epoch": 4.32, + "learning_rate": 3.850470300025802e-06, + "loss": 0.9394, + "step": 58422 + }, + { + "epoch": 4.32, + "learning_rate": 3.850155769202954e-06, + "loss": 0.9175, + "step": 58423 + }, + { + "epoch": 4.32, + "learning_rate": 3.849841248164427e-06, + "loss": 1.0148, + "step": 58424 + }, + { + "epoch": 4.32, + "learning_rate": 3.849526736910709e-06, + "loss": 1.032, + "step": 58425 + }, + { + "epoch": 4.32, + "learning_rate": 3.849212235442309e-06, + "loss": 1.1314, + "step": 58426 + }, + { + "epoch": 4.32, + "learning_rate": 3.84889774375972e-06, + "loss": 0.954, + "step": 58427 + }, + { + "epoch": 4.32, + "learning_rate": 3.848583261863454e-06, + "loss": 0.9265, + "step": 58428 + }, + { + "epoch": 4.32, + "learning_rate": 3.848268789754e-06, + "loss": 0.9785, + "step": 58429 + }, + { + "epoch": 4.32, + "learning_rate": 3.84795432743186e-06, + "loss": 0.9586, + "step": 58430 + }, + { + "epoch": 4.32, + "learning_rate": 3.84763987489754e-06, + "loss": 1.0662, + "step": 58431 + }, + { + "epoch": 4.32, + "learning_rate": 3.847325432151536e-06, + "loss": 1.0496, + "step": 58432 + }, + { + "epoch": 4.32, + "learning_rate": 3.8470109991943495e-06, + "loss": 1.0252, + "step": 58433 + }, + { + "epoch": 4.32, + "learning_rate": 3.846696576026478e-06, + "loss": 1.0346, + "step": 58434 + }, + { + "epoch": 4.32, + "learning_rate": 3.846382162648425e-06, + "loss": 1.025, + "step": 58435 + }, + { + "epoch": 4.32, + "learning_rate": 3.846067759060693e-06, + "loss": 0.9189, + "step": 58436 + }, + { + "epoch": 4.32, + "learning_rate": 3.845753365263778e-06, + "loss": 0.9731, + "step": 58437 + }, + { + "epoch": 4.32, + "learning_rate": 3.84543898125818e-06, + "loss": 1.0278, + "step": 58438 + }, + { + "epoch": 4.32, + "learning_rate": 3.8451246070444e-06, + "loss": 0.9605, + "step": 58439 + }, + { + "epoch": 4.32, + "learning_rate": 3.844810242622939e-06, + "loss": 0.9889, + "step": 58440 + }, + { + "epoch": 4.32, + "learning_rate": 3.844495887994298e-06, + "loss": 1.0199, + "step": 58441 + }, + { + "epoch": 4.32, + "learning_rate": 3.844181543158977e-06, + "loss": 0.9054, + "step": 58442 + }, + { + "epoch": 4.32, + "learning_rate": 3.843867208117471e-06, + "loss": 0.9436, + "step": 58443 + }, + { + "epoch": 4.32, + "learning_rate": 3.843552882870287e-06, + "loss": 0.8995, + "step": 58444 + }, + { + "epoch": 4.32, + "learning_rate": 3.843238567417923e-06, + "loss": 1.1216, + "step": 58445 + }, + { + "epoch": 4.32, + "learning_rate": 3.842924261760879e-06, + "loss": 0.9693, + "step": 58446 + }, + { + "epoch": 4.32, + "learning_rate": 3.842609965899653e-06, + "loss": 0.9934, + "step": 58447 + }, + { + "epoch": 4.32, + "learning_rate": 3.842295679834743e-06, + "loss": 0.987, + "step": 58448 + }, + { + "epoch": 4.32, + "learning_rate": 3.84198140356666e-06, + "loss": 0.9118, + "step": 58449 + }, + { + "epoch": 4.32, + "learning_rate": 3.841667137095891e-06, + "loss": 0.994, + "step": 58450 + }, + { + "epoch": 4.32, + "learning_rate": 3.8413528804229425e-06, + "loss": 0.8629, + "step": 58451 + }, + { + "epoch": 4.32, + "learning_rate": 3.841038633548311e-06, + "loss": 0.9498, + "step": 58452 + }, + { + "epoch": 4.32, + "learning_rate": 3.840724396472506e-06, + "loss": 1.1113, + "step": 58453 + }, + { + "epoch": 4.32, + "learning_rate": 3.840410169196012e-06, + "loss": 0.913, + "step": 58454 + }, + { + "epoch": 4.32, + "learning_rate": 3.840095951719343e-06, + "loss": 0.9264, + "step": 58455 + }, + { + "epoch": 4.32, + "learning_rate": 3.839781744042992e-06, + "loss": 0.9886, + "step": 58456 + }, + { + "epoch": 4.32, + "learning_rate": 3.83946754616746e-06, + "loss": 0.9963, + "step": 58457 + }, + { + "epoch": 4.32, + "learning_rate": 3.839153358093248e-06, + "loss": 1.0535, + "step": 58458 + }, + { + "epoch": 4.32, + "learning_rate": 3.83883917982085e-06, + "loss": 1.1361, + "step": 58459 + }, + { + "epoch": 4.32, + "learning_rate": 3.838525011350776e-06, + "loss": 1.0042, + "step": 58460 + }, + { + "epoch": 4.32, + "learning_rate": 3.838210852683518e-06, + "loss": 1.0789, + "step": 58461 + }, + { + "epoch": 4.32, + "learning_rate": 3.83789670381958e-06, + "loss": 0.8685, + "step": 58462 + }, + { + "epoch": 4.32, + "learning_rate": 3.837582564759457e-06, + "loss": 1.0591, + "step": 58463 + }, + { + "epoch": 4.32, + "learning_rate": 3.837268435503655e-06, + "loss": 1.0263, + "step": 58464 + }, + { + "epoch": 4.32, + "learning_rate": 3.836954316052674e-06, + "loss": 1.1106, + "step": 58465 + }, + { + "epoch": 4.32, + "learning_rate": 3.836640206407004e-06, + "loss": 1.007, + "step": 58466 + }, + { + "epoch": 4.32, + "learning_rate": 3.836326106567154e-06, + "loss": 0.9077, + "step": 58467 + }, + { + "epoch": 4.32, + "learning_rate": 3.836012016533618e-06, + "loss": 0.9151, + "step": 58468 + }, + { + "epoch": 4.32, + "learning_rate": 3.835697936306906e-06, + "loss": 1.0235, + "step": 58469 + }, + { + "epoch": 4.32, + "learning_rate": 3.835383865887503e-06, + "loss": 0.9104, + "step": 58470 + }, + { + "epoch": 4.32, + "learning_rate": 3.83506980527592e-06, + "loss": 0.9192, + "step": 58471 + }, + { + "epoch": 4.32, + "learning_rate": 3.834755754472648e-06, + "loss": 0.8826, + "step": 58472 + }, + { + "epoch": 4.32, + "learning_rate": 3.8344417134781985e-06, + "loss": 0.9721, + "step": 58473 + }, + { + "epoch": 4.32, + "learning_rate": 3.834127682293059e-06, + "loss": 1.0412, + "step": 58474 + }, + { + "epoch": 4.32, + "learning_rate": 3.833813660917735e-06, + "loss": 1.0104, + "step": 58475 + }, + { + "epoch": 4.32, + "learning_rate": 3.833499649352728e-06, + "loss": 1.0833, + "step": 58476 + }, + { + "epoch": 4.32, + "learning_rate": 3.833185647598533e-06, + "loss": 0.9228, + "step": 58477 + }, + { + "epoch": 4.32, + "learning_rate": 3.832871655655651e-06, + "loss": 1.0158, + "step": 58478 + }, + { + "epoch": 4.32, + "learning_rate": 3.83255767352458e-06, + "loss": 1.0237, + "step": 58479 + }, + { + "epoch": 4.32, + "learning_rate": 3.832243701205824e-06, + "loss": 0.9567, + "step": 58480 + }, + { + "epoch": 4.32, + "learning_rate": 3.83192973869988e-06, + "loss": 1.0584, + "step": 58481 + }, + { + "epoch": 4.32, + "learning_rate": 3.831615786007248e-06, + "loss": 1.0139, + "step": 58482 + }, + { + "epoch": 4.32, + "learning_rate": 3.831301843128423e-06, + "loss": 0.936, + "step": 58483 + }, + { + "epoch": 4.32, + "learning_rate": 3.830987910063913e-06, + "loss": 0.9156, + "step": 58484 + }, + { + "epoch": 4.32, + "learning_rate": 3.830673986814212e-06, + "loss": 0.9815, + "step": 58485 + }, + { + "epoch": 4.32, + "learning_rate": 3.830360073379821e-06, + "loss": 1.0029, + "step": 58486 + }, + { + "epoch": 4.32, + "learning_rate": 3.830046169761238e-06, + "loss": 1.0277, + "step": 58487 + }, + { + "epoch": 4.32, + "learning_rate": 3.829732275958961e-06, + "loss": 1.0553, + "step": 58488 + }, + { + "epoch": 4.32, + "learning_rate": 3.829418391973494e-06, + "loss": 1.0679, + "step": 58489 + }, + { + "epoch": 4.32, + "learning_rate": 3.829104517805334e-06, + "loss": 1.0703, + "step": 58490 + }, + { + "epoch": 4.32, + "learning_rate": 3.8287906534549824e-06, + "loss": 1.0031, + "step": 58491 + }, + { + "epoch": 4.32, + "learning_rate": 3.828476798922931e-06, + "loss": 1.0919, + "step": 58492 + }, + { + "epoch": 4.32, + "learning_rate": 3.828162954209689e-06, + "loss": 1.0573, + "step": 58493 + }, + { + "epoch": 4.32, + "learning_rate": 3.827849119315755e-06, + "loss": 0.8502, + "step": 58494 + }, + { + "epoch": 4.32, + "learning_rate": 3.8275352942416175e-06, + "loss": 1.0841, + "step": 58495 + }, + { + "epoch": 4.32, + "learning_rate": 3.827221478987786e-06, + "loss": 0.974, + "step": 58496 + }, + { + "epoch": 4.32, + "learning_rate": 3.826907673554758e-06, + "loss": 1.0341, + "step": 58497 + }, + { + "epoch": 4.32, + "learning_rate": 3.826593877943031e-06, + "loss": 0.9833, + "step": 58498 + }, + { + "epoch": 4.32, + "learning_rate": 3.826280092153101e-06, + "loss": 1.1528, + "step": 58499 + }, + { + "epoch": 4.32, + "learning_rate": 3.825966316185475e-06, + "loss": 0.9231, + "step": 58500 + }, + { + "epoch": 4.32, + "learning_rate": 3.825652550040648e-06, + "loss": 1.057, + "step": 58501 + }, + { + "epoch": 4.32, + "learning_rate": 3.82533879371912e-06, + "loss": 0.9573, + "step": 58502 + }, + { + "epoch": 4.32, + "learning_rate": 3.825025047221389e-06, + "loss": 0.914, + "step": 58503 + }, + { + "epoch": 4.32, + "learning_rate": 3.824711310547951e-06, + "loss": 1.0446, + "step": 58504 + }, + { + "epoch": 4.32, + "learning_rate": 3.824397583699313e-06, + "loss": 1.0011, + "step": 58505 + }, + { + "epoch": 4.32, + "learning_rate": 3.8240838666759705e-06, + "loss": 0.906, + "step": 58506 + }, + { + "epoch": 4.32, + "learning_rate": 3.8237701594784216e-06, + "loss": 0.9734, + "step": 58507 + }, + { + "epoch": 4.32, + "learning_rate": 3.823456462107162e-06, + "loss": 0.9117, + "step": 58508 + }, + { + "epoch": 4.32, + "learning_rate": 3.8231427745626995e-06, + "loss": 1.0535, + "step": 58509 + }, + { + "epoch": 4.32, + "learning_rate": 3.8228290968455275e-06, + "loss": 1.0941, + "step": 58510 + }, + { + "epoch": 4.32, + "learning_rate": 3.822515428956146e-06, + "loss": 1.0387, + "step": 58511 + }, + { + "epoch": 4.32, + "learning_rate": 3.822201770895055e-06, + "loss": 0.9556, + "step": 58512 + }, + { + "epoch": 4.32, + "learning_rate": 3.821888122662749e-06, + "loss": 0.8741, + "step": 58513 + }, + { + "epoch": 4.32, + "learning_rate": 3.821574484259737e-06, + "loss": 1.0696, + "step": 58514 + }, + { + "epoch": 4.32, + "learning_rate": 3.821260855686505e-06, + "loss": 0.8069, + "step": 58515 + }, + { + "epoch": 4.32, + "learning_rate": 3.820947236943562e-06, + "loss": 0.8634, + "step": 58516 + }, + { + "epoch": 4.32, + "learning_rate": 3.820633628031399e-06, + "loss": 0.9778, + "step": 58517 + }, + { + "epoch": 4.32, + "learning_rate": 3.820320028950528e-06, + "loss": 1.0367, + "step": 58518 + }, + { + "epoch": 4.32, + "learning_rate": 3.820006439701431e-06, + "loss": 0.9455, + "step": 58519 + }, + { + "epoch": 4.32, + "learning_rate": 3.819692860284621e-06, + "loss": 0.9366, + "step": 58520 + }, + { + "epoch": 4.32, + "learning_rate": 3.819379290700589e-06, + "loss": 1.0181, + "step": 58521 + }, + { + "epoch": 4.32, + "learning_rate": 3.819065730949837e-06, + "loss": 1.0376, + "step": 58522 + }, + { + "epoch": 4.32, + "learning_rate": 3.818752181032863e-06, + "loss": 1.0176, + "step": 58523 + }, + { + "epoch": 4.32, + "learning_rate": 3.818438640950163e-06, + "loss": 1.042, + "step": 58524 + }, + { + "epoch": 4.32, + "learning_rate": 3.818125110702241e-06, + "loss": 1.0339, + "step": 58525 + }, + { + "epoch": 4.32, + "learning_rate": 3.817811590289594e-06, + "loss": 0.9155, + "step": 58526 + }, + { + "epoch": 4.32, + "learning_rate": 3.81749807971272e-06, + "loss": 1.0204, + "step": 58527 + }, + { + "epoch": 4.32, + "learning_rate": 3.817184578972115e-06, + "loss": 0.9671, + "step": 58528 + }, + { + "epoch": 4.32, + "learning_rate": 3.816871088068284e-06, + "loss": 1.0989, + "step": 58529 + }, + { + "epoch": 4.32, + "learning_rate": 3.816557607001723e-06, + "loss": 1.1131, + "step": 58530 + }, + { + "epoch": 4.32, + "learning_rate": 3.816244135772931e-06, + "loss": 1.0593, + "step": 58531 + }, + { + "epoch": 4.32, + "learning_rate": 3.815930674382404e-06, + "loss": 1.0358, + "step": 58532 + }, + { + "epoch": 4.32, + "learning_rate": 3.815617222830642e-06, + "loss": 0.9931, + "step": 58533 + }, + { + "epoch": 4.33, + "learning_rate": 3.815303781118146e-06, + "loss": 0.9391, + "step": 58534 + }, + { + "epoch": 4.33, + "learning_rate": 3.814990349245415e-06, + "loss": 1.0495, + "step": 58535 + }, + { + "epoch": 4.33, + "learning_rate": 3.8146769272129436e-06, + "loss": 1.0241, + "step": 58536 + }, + { + "epoch": 4.33, + "learning_rate": 3.814363515021231e-06, + "loss": 1.0411, + "step": 58537 + }, + { + "epoch": 4.33, + "learning_rate": 3.8140501126707797e-06, + "loss": 1.0398, + "step": 58538 + }, + { + "epoch": 4.33, + "learning_rate": 3.813736720162088e-06, + "loss": 0.9729, + "step": 58539 + }, + { + "epoch": 4.33, + "learning_rate": 3.8134233374956507e-06, + "loss": 0.9461, + "step": 58540 + }, + { + "epoch": 4.33, + "learning_rate": 3.8131099646719696e-06, + "loss": 0.9561, + "step": 58541 + }, + { + "epoch": 4.33, + "learning_rate": 3.812796601691542e-06, + "loss": 1.0343, + "step": 58542 + }, + { + "epoch": 4.33, + "learning_rate": 3.8124832485548656e-06, + "loss": 1.0352, + "step": 58543 + }, + { + "epoch": 4.33, + "learning_rate": 3.812169905262437e-06, + "loss": 1.0042, + "step": 58544 + }, + { + "epoch": 4.33, + "learning_rate": 3.8118565718147606e-06, + "loss": 0.9394, + "step": 58545 + }, + { + "epoch": 4.33, + "learning_rate": 3.8115432482123326e-06, + "loss": 1.0161, + "step": 58546 + }, + { + "epoch": 4.33, + "learning_rate": 3.8112299344556504e-06, + "loss": 1.111, + "step": 58547 + }, + { + "epoch": 4.33, + "learning_rate": 3.810916630545208e-06, + "loss": 0.9637, + "step": 58548 + }, + { + "epoch": 4.33, + "learning_rate": 3.8106033364815143e-06, + "loss": 1.0309, + "step": 58549 + }, + { + "epoch": 4.33, + "learning_rate": 3.810290052265061e-06, + "loss": 0.9859, + "step": 58550 + }, + { + "epoch": 4.33, + "learning_rate": 3.809976777896348e-06, + "loss": 1.0242, + "step": 58551 + }, + { + "epoch": 4.33, + "learning_rate": 3.8096635133758732e-06, + "loss": 1.0589, + "step": 58552 + }, + { + "epoch": 4.33, + "learning_rate": 3.809350258704132e-06, + "loss": 0.893, + "step": 58553 + }, + { + "epoch": 4.33, + "learning_rate": 3.809037013881629e-06, + "loss": 0.9387, + "step": 58554 + }, + { + "epoch": 4.33, + "learning_rate": 3.8087237789088606e-06, + "loss": 1.0482, + "step": 58555 + }, + { + "epoch": 4.33, + "learning_rate": 3.8084105537863226e-06, + "loss": 0.9917, + "step": 58556 + }, + { + "epoch": 4.33, + "learning_rate": 3.8080973385145114e-06, + "loss": 0.7919, + "step": 58557 + }, + { + "epoch": 4.33, + "learning_rate": 3.8077841330939335e-06, + "loss": 1.0961, + "step": 58558 + }, + { + "epoch": 4.33, + "learning_rate": 3.807470937525085e-06, + "loss": 0.9554, + "step": 58559 + }, + { + "epoch": 4.33, + "learning_rate": 3.8071577518084547e-06, + "loss": 1.0252, + "step": 58560 + }, + { + "epoch": 4.33, + "learning_rate": 3.8068445759445516e-06, + "loss": 0.8229, + "step": 58561 + }, + { + "epoch": 4.33, + "learning_rate": 3.8065314099338668e-06, + "loss": 1.051, + "step": 58562 + }, + { + "epoch": 4.33, + "learning_rate": 3.806218253776909e-06, + "loss": 1.0072, + "step": 58563 + }, + { + "epoch": 4.33, + "learning_rate": 3.8059051074741617e-06, + "loss": 0.9378, + "step": 58564 + }, + { + "epoch": 4.33, + "learning_rate": 3.8055919710261346e-06, + "loss": 1.0037, + "step": 58565 + }, + { + "epoch": 4.33, + "learning_rate": 3.805278844433319e-06, + "loss": 0.8783, + "step": 58566 + }, + { + "epoch": 4.33, + "learning_rate": 3.804965727696223e-06, + "loss": 1.0359, + "step": 58567 + }, + { + "epoch": 4.33, + "learning_rate": 3.8046526208153334e-06, + "loss": 0.955, + "step": 58568 + }, + { + "epoch": 4.33, + "learning_rate": 3.8043395237911496e-06, + "loss": 1.1031, + "step": 58569 + }, + { + "epoch": 4.33, + "learning_rate": 3.8040264366241774e-06, + "loss": 0.9453, + "step": 58570 + }, + { + "epoch": 4.33, + "learning_rate": 3.80371335931491e-06, + "loss": 0.9577, + "step": 58571 + }, + { + "epoch": 4.33, + "learning_rate": 3.803400291863846e-06, + "loss": 0.9627, + "step": 58572 + }, + { + "epoch": 4.33, + "learning_rate": 3.8030872342714797e-06, + "loss": 0.9389, + "step": 58573 + }, + { + "epoch": 4.33, + "learning_rate": 3.8027741865383174e-06, + "loss": 0.9626, + "step": 58574 + }, + { + "epoch": 4.33, + "learning_rate": 3.802461148664851e-06, + "loss": 1.0016, + "step": 58575 + }, + { + "epoch": 4.33, + "learning_rate": 3.802148120651582e-06, + "loss": 0.9853, + "step": 58576 + }, + { + "epoch": 4.33, + "learning_rate": 3.8018351024990017e-06, + "loss": 0.872, + "step": 58577 + }, + { + "epoch": 4.33, + "learning_rate": 3.801522094207617e-06, + "loss": 0.9638, + "step": 58578 + }, + { + "epoch": 4.33, + "learning_rate": 3.8012090957779225e-06, + "loss": 1.0241, + "step": 58579 + }, + { + "epoch": 4.33, + "learning_rate": 3.8008961072104155e-06, + "loss": 1.0581, + "step": 58580 + }, + { + "epoch": 4.33, + "learning_rate": 3.800583128505594e-06, + "loss": 1.0355, + "step": 58581 + }, + { + "epoch": 4.33, + "learning_rate": 3.8002701596639524e-06, + "loss": 0.9568, + "step": 58582 + }, + { + "epoch": 4.33, + "learning_rate": 3.7999572006859965e-06, + "loss": 1.0103, + "step": 58583 + }, + { + "epoch": 4.33, + "learning_rate": 3.7996442515722187e-06, + "loss": 1.0065, + "step": 58584 + }, + { + "epoch": 4.33, + "learning_rate": 3.799331312323119e-06, + "loss": 1.0299, + "step": 58585 + }, + { + "epoch": 4.33, + "learning_rate": 3.7990183829391913e-06, + "loss": 0.954, + "step": 58586 + }, + { + "epoch": 4.33, + "learning_rate": 3.798705463420943e-06, + "loss": 1.0849, + "step": 58587 + }, + { + "epoch": 4.33, + "learning_rate": 3.7983925537688615e-06, + "loss": 0.9195, + "step": 58588 + }, + { + "epoch": 4.33, + "learning_rate": 3.7980796539834462e-06, + "loss": 1.0075, + "step": 58589 + }, + { + "epoch": 4.33, + "learning_rate": 3.7977667640652004e-06, + "loss": 1.0847, + "step": 58590 + }, + { + "epoch": 4.33, + "learning_rate": 3.797453884014619e-06, + "loss": 1.0741, + "step": 58591 + }, + { + "epoch": 4.33, + "learning_rate": 3.7971410138322006e-06, + "loss": 1.0253, + "step": 58592 + }, + { + "epoch": 4.33, + "learning_rate": 3.796828153518438e-06, + "loss": 0.8408, + "step": 58593 + }, + { + "epoch": 4.33, + "learning_rate": 3.796515303073837e-06, + "loss": 1.0033, + "step": 58594 + }, + { + "epoch": 4.33, + "learning_rate": 3.79620246249889e-06, + "loss": 0.9849, + "step": 58595 + }, + { + "epoch": 4.33, + "learning_rate": 3.7958896317940973e-06, + "loss": 1.057, + "step": 58596 + }, + { + "epoch": 4.33, + "learning_rate": 3.7955768109599556e-06, + "loss": 0.9829, + "step": 58597 + }, + { + "epoch": 4.33, + "learning_rate": 3.7952639999969576e-06, + "loss": 0.9113, + "step": 58598 + }, + { + "epoch": 4.33, + "learning_rate": 3.794951198905611e-06, + "loss": 1.0246, + "step": 58599 + }, + { + "epoch": 4.33, + "learning_rate": 3.7946384076864073e-06, + "loss": 1.0499, + "step": 58600 + }, + { + "epoch": 4.33, + "learning_rate": 3.7943256263398455e-06, + "loss": 0.9107, + "step": 58601 + }, + { + "epoch": 4.33, + "learning_rate": 3.794012854866418e-06, + "loss": 1.0277, + "step": 58602 + }, + { + "epoch": 4.33, + "learning_rate": 3.7937000932666324e-06, + "loss": 0.9957, + "step": 58603 + }, + { + "epoch": 4.33, + "learning_rate": 3.7933873415409806e-06, + "loss": 1.0513, + "step": 58604 + }, + { + "epoch": 4.33, + "learning_rate": 3.793074599689961e-06, + "loss": 1.0054, + "step": 58605 + }, + { + "epoch": 4.33, + "learning_rate": 3.7927618677140697e-06, + "loss": 1.0026, + "step": 58606 + }, + { + "epoch": 4.33, + "learning_rate": 3.7924491456138035e-06, + "loss": 0.9791, + "step": 58607 + }, + { + "epoch": 4.33, + "learning_rate": 3.792136433389668e-06, + "loss": 1.0628, + "step": 58608 + }, + { + "epoch": 4.33, + "learning_rate": 3.791823731042148e-06, + "loss": 0.9769, + "step": 58609 + }, + { + "epoch": 4.33, + "learning_rate": 3.791511038571751e-06, + "loss": 0.9322, + "step": 58610 + }, + { + "epoch": 4.33, + "learning_rate": 3.791198355978968e-06, + "loss": 0.9434, + "step": 58611 + }, + { + "epoch": 4.33, + "learning_rate": 3.790885683264306e-06, + "loss": 0.9915, + "step": 58612 + }, + { + "epoch": 4.33, + "learning_rate": 3.7905730204282486e-06, + "loss": 0.9933, + "step": 58613 + }, + { + "epoch": 4.33, + "learning_rate": 3.7902603674713055e-06, + "loss": 0.975, + "step": 58614 + }, + { + "epoch": 4.33, + "learning_rate": 3.789947724393969e-06, + "loss": 0.9545, + "step": 58615 + }, + { + "epoch": 4.33, + "learning_rate": 3.789635091196736e-06, + "loss": 1.053, + "step": 58616 + }, + { + "epoch": 4.33, + "learning_rate": 3.7893224678801055e-06, + "loss": 1.0353, + "step": 58617 + }, + { + "epoch": 4.33, + "learning_rate": 3.7890098544445695e-06, + "loss": 0.9863, + "step": 58618 + }, + { + "epoch": 4.33, + "learning_rate": 3.788697250890635e-06, + "loss": 0.8599, + "step": 58619 + }, + { + "epoch": 4.33, + "learning_rate": 3.7883846572187934e-06, + "loss": 1.0401, + "step": 58620 + }, + { + "epoch": 4.33, + "learning_rate": 3.7880720734295427e-06, + "loss": 0.9207, + "step": 58621 + }, + { + "epoch": 4.33, + "learning_rate": 3.7877594995233767e-06, + "loss": 0.9139, + "step": 58622 + }, + { + "epoch": 4.33, + "learning_rate": 3.7874469355008015e-06, + "loss": 1.0341, + "step": 58623 + }, + { + "epoch": 4.33, + "learning_rate": 3.7871343813623084e-06, + "loss": 0.8885, + "step": 58624 + }, + { + "epoch": 4.33, + "learning_rate": 3.7868218371083955e-06, + "loss": 0.9565, + "step": 58625 + }, + { + "epoch": 4.33, + "learning_rate": 3.7865093027395607e-06, + "loss": 0.9666, + "step": 58626 + }, + { + "epoch": 4.33, + "learning_rate": 3.786196778256296e-06, + "loss": 1.0163, + "step": 58627 + }, + { + "epoch": 4.33, + "learning_rate": 3.7858842636591075e-06, + "loss": 1.1339, + "step": 58628 + }, + { + "epoch": 4.33, + "learning_rate": 3.7855717589484885e-06, + "loss": 1.0231, + "step": 58629 + }, + { + "epoch": 4.33, + "learning_rate": 3.7852592641249353e-06, + "loss": 1.0147, + "step": 58630 + }, + { + "epoch": 4.33, + "learning_rate": 3.784946779188943e-06, + "loss": 0.953, + "step": 58631 + }, + { + "epoch": 4.33, + "learning_rate": 3.7846343041410184e-06, + "loss": 1.0789, + "step": 58632 + }, + { + "epoch": 4.33, + "learning_rate": 3.784321838981644e-06, + "loss": 0.8904, + "step": 58633 + }, + { + "epoch": 4.33, + "learning_rate": 3.784009383711329e-06, + "loss": 1.0342, + "step": 58634 + }, + { + "epoch": 4.33, + "learning_rate": 3.7836969383305656e-06, + "loss": 0.973, + "step": 58635 + }, + { + "epoch": 4.33, + "learning_rate": 3.783384502839851e-06, + "loss": 1.0097, + "step": 58636 + }, + { + "epoch": 4.33, + "learning_rate": 3.7830720772396833e-06, + "loss": 0.9155, + "step": 58637 + }, + { + "epoch": 4.33, + "learning_rate": 3.782759661530555e-06, + "loss": 1.027, + "step": 58638 + }, + { + "epoch": 4.33, + "learning_rate": 3.782447255712971e-06, + "loss": 0.9576, + "step": 58639 + }, + { + "epoch": 4.33, + "learning_rate": 3.7821348597874252e-06, + "loss": 1.0048, + "step": 58640 + }, + { + "epoch": 4.33, + "learning_rate": 3.781822473754413e-06, + "loss": 1.0193, + "step": 58641 + }, + { + "epoch": 4.33, + "learning_rate": 3.781510097614428e-06, + "loss": 0.8877, + "step": 58642 + }, + { + "epoch": 4.33, + "learning_rate": 3.781197731367976e-06, + "loss": 1.0715, + "step": 58643 + }, + { + "epoch": 4.33, + "learning_rate": 3.7808853750155495e-06, + "loss": 1.0646, + "step": 58644 + }, + { + "epoch": 4.33, + "learning_rate": 3.780573028557645e-06, + "loss": 1.0138, + "step": 58645 + }, + { + "epoch": 4.33, + "learning_rate": 3.7802606919947604e-06, + "loss": 0.9579, + "step": 58646 + }, + { + "epoch": 4.33, + "learning_rate": 3.7799483653273884e-06, + "loss": 0.9963, + "step": 58647 + }, + { + "epoch": 4.33, + "learning_rate": 3.7796360485560323e-06, + "loss": 0.9247, + "step": 58648 + }, + { + "epoch": 4.33, + "learning_rate": 3.779323741681187e-06, + "loss": 0.9518, + "step": 58649 + }, + { + "epoch": 4.33, + "learning_rate": 3.7790114447033475e-06, + "loss": 1.0231, + "step": 58650 + }, + { + "epoch": 4.33, + "learning_rate": 3.7786991576230102e-06, + "loss": 0.9896, + "step": 58651 + }, + { + "epoch": 4.33, + "learning_rate": 3.778386880440675e-06, + "loss": 0.99, + "step": 58652 + }, + { + "epoch": 4.33, + "learning_rate": 3.7780746131568426e-06, + "loss": 1.0658, + "step": 58653 + }, + { + "epoch": 4.33, + "learning_rate": 3.777762355771997e-06, + "loss": 1.0375, + "step": 58654 + }, + { + "epoch": 4.33, + "learning_rate": 3.7774501082866454e-06, + "loss": 0.9519, + "step": 58655 + }, + { + "epoch": 4.33, + "learning_rate": 3.7771378707012774e-06, + "loss": 0.8446, + "step": 58656 + }, + { + "epoch": 4.33, + "learning_rate": 3.776825643016402e-06, + "loss": 1.0643, + "step": 58657 + }, + { + "epoch": 4.33, + "learning_rate": 3.776513425232501e-06, + "loss": 0.8616, + "step": 58658 + }, + { + "epoch": 4.33, + "learning_rate": 3.7762012173500817e-06, + "loss": 0.9993, + "step": 58659 + }, + { + "epoch": 4.33, + "learning_rate": 3.775889019369634e-06, + "loss": 0.9736, + "step": 58660 + }, + { + "epoch": 4.33, + "learning_rate": 3.775576831291664e-06, + "loss": 1.0391, + "step": 58661 + }, + { + "epoch": 4.33, + "learning_rate": 3.7752646531166594e-06, + "loss": 0.9703, + "step": 58662 + }, + { + "epoch": 4.33, + "learning_rate": 3.774952484845116e-06, + "loss": 0.9652, + "step": 58663 + }, + { + "epoch": 4.33, + "learning_rate": 3.7746403264775366e-06, + "loss": 0.8722, + "step": 58664 + }, + { + "epoch": 4.33, + "learning_rate": 3.774328178014416e-06, + "loss": 1.0197, + "step": 58665 + }, + { + "epoch": 4.33, + "learning_rate": 3.774016039456251e-06, + "loss": 1.0205, + "step": 58666 + }, + { + "epoch": 4.33, + "learning_rate": 3.7737039108035335e-06, + "loss": 1.094, + "step": 58667 + }, + { + "epoch": 4.33, + "learning_rate": 3.7733917920567665e-06, + "loss": 1.0978, + "step": 58668 + }, + { + "epoch": 4.34, + "learning_rate": 3.7730796832164453e-06, + "loss": 0.9577, + "step": 58669 + }, + { + "epoch": 4.34, + "learning_rate": 3.7727675842830647e-06, + "loss": 1.0592, + "step": 58670 + }, + { + "epoch": 4.34, + "learning_rate": 3.7724554952571223e-06, + "loss": 0.9525, + "step": 58671 + }, + { + "epoch": 4.34, + "learning_rate": 3.77214341613911e-06, + "loss": 0.9864, + "step": 58672 + }, + { + "epoch": 4.34, + "learning_rate": 3.771831346929532e-06, + "loss": 0.9382, + "step": 58673 + }, + { + "epoch": 4.34, + "learning_rate": 3.771519287628881e-06, + "loss": 0.9362, + "step": 58674 + }, + { + "epoch": 4.34, + "learning_rate": 3.771207238237654e-06, + "loss": 0.9316, + "step": 58675 + }, + { + "epoch": 4.34, + "learning_rate": 3.7708951987563436e-06, + "loss": 0.9728, + "step": 58676 + }, + { + "epoch": 4.34, + "learning_rate": 3.7705831691854576e-06, + "loss": 1.0576, + "step": 58677 + }, + { + "epoch": 4.34, + "learning_rate": 3.7702711495254764e-06, + "loss": 0.985, + "step": 58678 + }, + { + "epoch": 4.34, + "learning_rate": 3.769959139776909e-06, + "loss": 0.8807, + "step": 58679 + }, + { + "epoch": 4.34, + "learning_rate": 3.7696471399402467e-06, + "loss": 1.0186, + "step": 58680 + }, + { + "epoch": 4.34, + "learning_rate": 3.769335150015988e-06, + "loss": 0.9408, + "step": 58681 + }, + { + "epoch": 4.34, + "learning_rate": 3.7690231700046263e-06, + "loss": 1.0416, + "step": 58682 + }, + { + "epoch": 4.34, + "learning_rate": 3.7687111999066574e-06, + "loss": 1.0706, + "step": 58683 + }, + { + "epoch": 4.34, + "learning_rate": 3.7683992397225823e-06, + "loss": 0.9776, + "step": 58684 + }, + { + "epoch": 4.34, + "learning_rate": 3.7680872894528953e-06, + "loss": 1.0371, + "step": 58685 + }, + { + "epoch": 4.34, + "learning_rate": 3.7677753490980928e-06, + "loss": 0.9501, + "step": 58686 + }, + { + "epoch": 4.34, + "learning_rate": 3.767463418658667e-06, + "loss": 1.0137, + "step": 58687 + }, + { + "epoch": 4.34, + "learning_rate": 3.7671514981351197e-06, + "loss": 1.0748, + "step": 58688 + }, + { + "epoch": 4.34, + "learning_rate": 3.766839587527946e-06, + "loss": 1.0369, + "step": 58689 + }, + { + "epoch": 4.34, + "learning_rate": 3.766527686837642e-06, + "loss": 0.9453, + "step": 58690 + }, + { + "epoch": 4.34, + "learning_rate": 3.766215796064703e-06, + "loss": 0.9656, + "step": 58691 + }, + { + "epoch": 4.34, + "learning_rate": 3.765903915209622e-06, + "loss": 1.0531, + "step": 58692 + }, + { + "epoch": 4.34, + "learning_rate": 3.765592044272902e-06, + "loss": 0.9908, + "step": 58693 + }, + { + "epoch": 4.34, + "learning_rate": 3.7652801832550357e-06, + "loss": 1.0377, + "step": 58694 + }, + { + "epoch": 4.34, + "learning_rate": 3.76496833215652e-06, + "loss": 1.0526, + "step": 58695 + }, + { + "epoch": 4.34, + "learning_rate": 3.764656490977846e-06, + "loss": 0.9575, + "step": 58696 + }, + { + "epoch": 4.34, + "learning_rate": 3.764344659719519e-06, + "loss": 1.0171, + "step": 58697 + }, + { + "epoch": 4.34, + "learning_rate": 3.764032838382029e-06, + "loss": 0.9621, + "step": 58698 + }, + { + "epoch": 4.34, + "learning_rate": 3.7637210269658754e-06, + "loss": 0.8182, + "step": 58699 + }, + { + "epoch": 4.34, + "learning_rate": 3.7634092254715506e-06, + "loss": 1.0596, + "step": 58700 + }, + { + "epoch": 4.34, + "learning_rate": 3.7630974338995498e-06, + "loss": 1.0084, + "step": 58701 + }, + { + "epoch": 4.34, + "learning_rate": 3.7627856522503794e-06, + "loss": 1.0209, + "step": 58702 + }, + { + "epoch": 4.34, + "learning_rate": 3.7624738805245197e-06, + "loss": 0.835, + "step": 58703 + }, + { + "epoch": 4.34, + "learning_rate": 3.7621621187224786e-06, + "loss": 1.0437, + "step": 58704 + }, + { + "epoch": 4.34, + "learning_rate": 3.7618503668447448e-06, + "loss": 1.0158, + "step": 58705 + }, + { + "epoch": 4.34, + "learning_rate": 3.761538624891825e-06, + "loss": 1.0863, + "step": 58706 + }, + { + "epoch": 4.34, + "learning_rate": 3.7612268928642004e-06, + "loss": 0.8674, + "step": 58707 + }, + { + "epoch": 4.34, + "learning_rate": 3.7609151707623792e-06, + "loss": 0.8599, + "step": 58708 + }, + { + "epoch": 4.34, + "learning_rate": 3.7606034585868525e-06, + "loss": 0.9713, + "step": 58709 + }, + { + "epoch": 4.34, + "learning_rate": 3.7602917563381156e-06, + "loss": 0.9606, + "step": 58710 + }, + { + "epoch": 4.34, + "learning_rate": 3.759980064016665e-06, + "loss": 1.0204, + "step": 58711 + }, + { + "epoch": 4.34, + "learning_rate": 3.759668381622994e-06, + "loss": 0.9527, + "step": 58712 + }, + { + "epoch": 4.34, + "learning_rate": 3.7593567091576057e-06, + "loss": 0.9567, + "step": 58713 + }, + { + "epoch": 4.34, + "learning_rate": 3.7590450466209904e-06, + "loss": 0.9606, + "step": 58714 + }, + { + "epoch": 4.34, + "learning_rate": 3.758733394013645e-06, + "loss": 0.9498, + "step": 58715 + }, + { + "epoch": 4.34, + "learning_rate": 3.758421751336062e-06, + "loss": 1.0647, + "step": 58716 + }, + { + "epoch": 4.34, + "learning_rate": 3.7581101185887448e-06, + "loss": 0.943, + "step": 58717 + }, + { + "epoch": 4.34, + "learning_rate": 3.757798495772185e-06, + "loss": 0.8294, + "step": 58718 + }, + { + "epoch": 4.34, + "learning_rate": 3.7574868828868783e-06, + "loss": 0.9494, + "step": 58719 + }, + { + "epoch": 4.34, + "learning_rate": 3.7571752799333206e-06, + "loss": 1.0048, + "step": 58720 + }, + { + "epoch": 4.34, + "learning_rate": 3.7568636869120044e-06, + "loss": 0.9554, + "step": 58721 + }, + { + "epoch": 4.34, + "learning_rate": 3.7565521038234365e-06, + "loss": 1.0525, + "step": 58722 + }, + { + "epoch": 4.34, + "learning_rate": 3.756240530668096e-06, + "loss": 1.0683, + "step": 58723 + }, + { + "epoch": 4.34, + "learning_rate": 3.7559289674464927e-06, + "loss": 0.9676, + "step": 58724 + }, + { + "epoch": 4.34, + "learning_rate": 3.755617414159113e-06, + "loss": 1.0088, + "step": 58725 + }, + { + "epoch": 4.34, + "learning_rate": 3.7553058708064636e-06, + "loss": 0.9187, + "step": 58726 + }, + { + "epoch": 4.34, + "learning_rate": 3.7549943373890297e-06, + "loss": 1.0522, + "step": 58727 + }, + { + "epoch": 4.34, + "learning_rate": 3.7546828139073066e-06, + "loss": 0.8593, + "step": 58728 + }, + { + "epoch": 4.34, + "learning_rate": 3.7543713003617975e-06, + "loss": 0.9766, + "step": 58729 + }, + { + "epoch": 4.34, + "learning_rate": 3.7540597967529945e-06, + "loss": 0.9555, + "step": 58730 + }, + { + "epoch": 4.34, + "learning_rate": 3.7537483030813924e-06, + "loss": 1.0003, + "step": 58731 + }, + { + "epoch": 4.34, + "learning_rate": 3.7534368193474844e-06, + "loss": 0.9027, + "step": 58732 + }, + { + "epoch": 4.34, + "learning_rate": 3.7531253455517734e-06, + "loss": 0.9391, + "step": 58733 + }, + { + "epoch": 4.34, + "learning_rate": 3.752813881694749e-06, + "loss": 0.9924, + "step": 58734 + }, + { + "epoch": 4.34, + "learning_rate": 3.7525024277769097e-06, + "loss": 0.9912, + "step": 58735 + }, + { + "epoch": 4.34, + "learning_rate": 3.752190983798749e-06, + "loss": 0.9688, + "step": 58736 + }, + { + "epoch": 4.34, + "learning_rate": 3.7518795497607595e-06, + "loss": 1.0228, + "step": 58737 + }, + { + "epoch": 4.34, + "learning_rate": 3.751568125663444e-06, + "loss": 0.9922, + "step": 58738 + }, + { + "epoch": 4.34, + "learning_rate": 3.751256711507294e-06, + "loss": 0.994, + "step": 58739 + }, + { + "epoch": 4.34, + "learning_rate": 3.750945307292805e-06, + "loss": 1.0861, + "step": 58740 + }, + { + "epoch": 4.34, + "learning_rate": 3.750633913020469e-06, + "loss": 1.0082, + "step": 58741 + }, + { + "epoch": 4.34, + "learning_rate": 3.7503225286907883e-06, + "loss": 0.909, + "step": 58742 + }, + { + "epoch": 4.34, + "learning_rate": 3.750011154304256e-06, + "loss": 1.0434, + "step": 58743 + }, + { + "epoch": 4.34, + "learning_rate": 3.7496997898613663e-06, + "loss": 1.0193, + "step": 58744 + }, + { + "epoch": 4.34, + "learning_rate": 3.7493884353626107e-06, + "loss": 1.0485, + "step": 58745 + }, + { + "epoch": 4.34, + "learning_rate": 3.749077090808493e-06, + "loss": 1.091, + "step": 58746 + }, + { + "epoch": 4.34, + "learning_rate": 3.748765756199507e-06, + "loss": 1.0325, + "step": 58747 + }, + { + "epoch": 4.34, + "learning_rate": 3.7484544315361394e-06, + "loss": 0.9832, + "step": 58748 + }, + { + "epoch": 4.34, + "learning_rate": 3.748143116818893e-06, + "loss": 0.9384, + "step": 58749 + }, + { + "epoch": 4.34, + "learning_rate": 3.7478318120482594e-06, + "loss": 1.0505, + "step": 58750 + }, + { + "epoch": 4.34, + "learning_rate": 3.747520517224742e-06, + "loss": 1.1364, + "step": 58751 + }, + { + "epoch": 4.34, + "learning_rate": 3.747209232348824e-06, + "loss": 1.0066, + "step": 58752 + }, + { + "epoch": 4.34, + "learning_rate": 3.7468979574210094e-06, + "loss": 1.0685, + "step": 58753 + }, + { + "epoch": 4.34, + "learning_rate": 3.7465866924417915e-06, + "loss": 1.169, + "step": 58754 + }, + { + "epoch": 4.34, + "learning_rate": 3.7462754374116637e-06, + "loss": 1.003, + "step": 58755 + }, + { + "epoch": 4.34, + "learning_rate": 3.745964192331123e-06, + "loss": 1.0186, + "step": 58756 + }, + { + "epoch": 4.34, + "learning_rate": 3.7456529572006595e-06, + "loss": 0.8904, + "step": 58757 + }, + { + "epoch": 4.34, + "learning_rate": 3.745341732020776e-06, + "loss": 0.8708, + "step": 58758 + }, + { + "epoch": 4.34, + "learning_rate": 3.7450305167919655e-06, + "loss": 0.9451, + "step": 58759 + }, + { + "epoch": 4.34, + "learning_rate": 3.7447193115147205e-06, + "loss": 0.9317, + "step": 58760 + }, + { + "epoch": 4.34, + "learning_rate": 3.744408116189535e-06, + "loss": 0.9321, + "step": 58761 + }, + { + "epoch": 4.34, + "learning_rate": 3.7440969308169107e-06, + "loss": 0.9841, + "step": 58762 + }, + { + "epoch": 4.34, + "learning_rate": 3.743785755397338e-06, + "loss": 1.0579, + "step": 58763 + }, + { + "epoch": 4.34, + "learning_rate": 3.743474589931313e-06, + "loss": 1.0195, + "step": 58764 + }, + { + "epoch": 4.34, + "learning_rate": 3.74316343441933e-06, + "loss": 1.1041, + "step": 58765 + }, + { + "epoch": 4.34, + "learning_rate": 3.7428522888618814e-06, + "loss": 0.9708, + "step": 58766 + }, + { + "epoch": 4.34, + "learning_rate": 3.7425411532594714e-06, + "loss": 1.0339, + "step": 58767 + }, + { + "epoch": 4.34, + "learning_rate": 3.7422300276125824e-06, + "loss": 0.9187, + "step": 58768 + }, + { + "epoch": 4.34, + "learning_rate": 3.741918911921719e-06, + "loss": 0.9413, + "step": 58769 + }, + { + "epoch": 4.34, + "learning_rate": 3.7416078061873705e-06, + "loss": 1.0174, + "step": 58770 + }, + { + "epoch": 4.34, + "learning_rate": 3.741296710410041e-06, + "loss": 1.0563, + "step": 58771 + }, + { + "epoch": 4.34, + "learning_rate": 3.740985624590212e-06, + "loss": 1.0236, + "step": 58772 + }, + { + "epoch": 4.34, + "learning_rate": 3.740674548728388e-06, + "loss": 1.1134, + "step": 58773 + }, + { + "epoch": 4.34, + "learning_rate": 3.7403634828250623e-06, + "loss": 0.9042, + "step": 58774 + }, + { + "epoch": 4.34, + "learning_rate": 3.740052426880728e-06, + "loss": 0.9568, + "step": 58775 + }, + { + "epoch": 4.34, + "learning_rate": 3.7397413808958815e-06, + "loss": 1.0063, + "step": 58776 + }, + { + "epoch": 4.34, + "learning_rate": 3.739430344871012e-06, + "loss": 1.0191, + "step": 58777 + }, + { + "epoch": 4.34, + "learning_rate": 3.7391193188066243e-06, + "loss": 1.0913, + "step": 58778 + }, + { + "epoch": 4.34, + "learning_rate": 3.7388083027032075e-06, + "loss": 0.9516, + "step": 58779 + }, + { + "epoch": 4.34, + "learning_rate": 3.738497296561258e-06, + "loss": 0.9234, + "step": 58780 + }, + { + "epoch": 4.34, + "learning_rate": 3.7381863003812647e-06, + "loss": 1.0397, + "step": 58781 + }, + { + "epoch": 4.34, + "learning_rate": 3.7378753141637313e-06, + "loss": 0.9874, + "step": 58782 + }, + { + "epoch": 4.34, + "learning_rate": 3.73756433790915e-06, + "loss": 1.0168, + "step": 58783 + }, + { + "epoch": 4.34, + "learning_rate": 3.7372533716180127e-06, + "loss": 1.017, + "step": 58784 + }, + { + "epoch": 4.34, + "learning_rate": 3.7369424152908165e-06, + "loss": 0.9646, + "step": 58785 + }, + { + "epoch": 4.34, + "learning_rate": 3.7366314689280514e-06, + "loss": 0.9744, + "step": 58786 + }, + { + "epoch": 4.34, + "learning_rate": 3.73632053253022e-06, + "loss": 1.0585, + "step": 58787 + }, + { + "epoch": 4.34, + "learning_rate": 3.736009606097812e-06, + "loss": 0.9003, + "step": 58788 + }, + { + "epoch": 4.34, + "learning_rate": 3.735698689631324e-06, + "loss": 0.9759, + "step": 58789 + }, + { + "epoch": 4.34, + "learning_rate": 3.7353877831312457e-06, + "loss": 1.0222, + "step": 58790 + }, + { + "epoch": 4.34, + "learning_rate": 3.73507688659808e-06, + "loss": 0.8718, + "step": 58791 + }, + { + "epoch": 4.34, + "learning_rate": 3.734766000032317e-06, + "loss": 1.0411, + "step": 58792 + }, + { + "epoch": 4.34, + "learning_rate": 3.7344551234344516e-06, + "loss": 0.9548, + "step": 58793 + }, + { + "epoch": 4.34, + "learning_rate": 3.7341442568049777e-06, + "loss": 0.9607, + "step": 58794 + }, + { + "epoch": 4.34, + "learning_rate": 3.733833400144389e-06, + "loss": 0.9896, + "step": 58795 + }, + { + "epoch": 4.34, + "learning_rate": 3.733522553453187e-06, + "loss": 0.9891, + "step": 58796 + }, + { + "epoch": 4.34, + "learning_rate": 3.733211716731855e-06, + "loss": 0.9633, + "step": 58797 + }, + { + "epoch": 4.34, + "learning_rate": 3.732900889980896e-06, + "loss": 1.0661, + "step": 58798 + }, + { + "epoch": 4.34, + "learning_rate": 3.7325900732008027e-06, + "loss": 0.889, + "step": 58799 + }, + { + "epoch": 4.34, + "learning_rate": 3.732279266392068e-06, + "loss": 0.9975, + "step": 58800 + }, + { + "epoch": 4.34, + "learning_rate": 3.731968469555185e-06, + "loss": 1.0995, + "step": 58801 + }, + { + "epoch": 4.34, + "learning_rate": 3.7316576826906535e-06, + "loss": 1.0123, + "step": 58802 + }, + { + "epoch": 4.34, + "learning_rate": 3.731346905798965e-06, + "loss": 0.9899, + "step": 58803 + }, + { + "epoch": 4.34, + "learning_rate": 3.731036138880614e-06, + "loss": 1.0383, + "step": 58804 + }, + { + "epoch": 4.35, + "learning_rate": 3.7307253819360944e-06, + "loss": 1.0469, + "step": 58805 + }, + { + "epoch": 4.35, + "learning_rate": 3.7304146349658986e-06, + "loss": 0.9636, + "step": 58806 + }, + { + "epoch": 4.35, + "learning_rate": 3.7301038979705263e-06, + "loss": 1.0265, + "step": 58807 + }, + { + "epoch": 4.35, + "learning_rate": 3.7297931709504686e-06, + "loss": 1.014, + "step": 58808 + }, + { + "epoch": 4.35, + "learning_rate": 3.72948245390622e-06, + "loss": 1.1011, + "step": 58809 + }, + { + "epoch": 4.35, + "learning_rate": 3.7291717468382725e-06, + "loss": 1.0001, + "step": 58810 + }, + { + "epoch": 4.35, + "learning_rate": 3.7288610497471255e-06, + "loss": 0.9802, + "step": 58811 + }, + { + "epoch": 4.35, + "learning_rate": 3.728550362633275e-06, + "loss": 0.9187, + "step": 58812 + }, + { + "epoch": 4.35, + "learning_rate": 3.7282396854972045e-06, + "loss": 1.0223, + "step": 58813 + }, + { + "epoch": 4.35, + "learning_rate": 3.727929018339418e-06, + "loss": 0.9706, + "step": 58814 + }, + { + "epoch": 4.35, + "learning_rate": 3.727618361160403e-06, + "loss": 0.8762, + "step": 58815 + }, + { + "epoch": 4.35, + "learning_rate": 3.727307713960665e-06, + "loss": 1.0345, + "step": 58816 + }, + { + "epoch": 4.35, + "learning_rate": 3.7269970767406838e-06, + "loss": 0.953, + "step": 58817 + }, + { + "epoch": 4.35, + "learning_rate": 3.7266864495009638e-06, + "loss": 0.973, + "step": 58818 + }, + { + "epoch": 4.35, + "learning_rate": 3.726375832241992e-06, + "loss": 0.8978, + "step": 58819 + }, + { + "epoch": 4.35, + "learning_rate": 3.7260652249642736e-06, + "loss": 0.9169, + "step": 58820 + }, + { + "epoch": 4.35, + "learning_rate": 3.7257546276682923e-06, + "loss": 0.9697, + "step": 58821 + }, + { + "epoch": 4.35, + "learning_rate": 3.725444040354541e-06, + "loss": 1.0106, + "step": 58822 + }, + { + "epoch": 4.35, + "learning_rate": 3.7251334630235234e-06, + "loss": 0.9543, + "step": 58823 + }, + { + "epoch": 4.35, + "learning_rate": 3.724822895675728e-06, + "loss": 0.9282, + "step": 58824 + }, + { + "epoch": 4.35, + "learning_rate": 3.72451233831165e-06, + "loss": 1.0005, + "step": 58825 + }, + { + "epoch": 4.35, + "learning_rate": 3.7242017909317796e-06, + "loss": 1.0743, + "step": 58826 + }, + { + "epoch": 4.35, + "learning_rate": 3.7238912535366178e-06, + "loss": 0.8739, + "step": 58827 + }, + { + "epoch": 4.35, + "learning_rate": 3.7235807261266555e-06, + "loss": 1.0493, + "step": 58828 + }, + { + "epoch": 4.35, + "learning_rate": 3.7232702087023855e-06, + "loss": 1.0391, + "step": 58829 + }, + { + "epoch": 4.35, + "learning_rate": 3.722959701264304e-06, + "loss": 1.0533, + "step": 58830 + }, + { + "epoch": 4.35, + "learning_rate": 3.7226492038128992e-06, + "loss": 1.0202, + "step": 58831 + }, + { + "epoch": 4.35, + "learning_rate": 3.7223387163486737e-06, + "loss": 0.9316, + "step": 58832 + }, + { + "epoch": 4.35, + "learning_rate": 3.7220282388721172e-06, + "loss": 0.8816, + "step": 58833 + }, + { + "epoch": 4.35, + "learning_rate": 3.7217177713837238e-06, + "loss": 0.8822, + "step": 58834 + }, + { + "epoch": 4.35, + "learning_rate": 3.7214073138839846e-06, + "loss": 1.0045, + "step": 58835 + }, + { + "epoch": 4.35, + "learning_rate": 3.7210968663734006e-06, + "loss": 0.9827, + "step": 58836 + }, + { + "epoch": 4.35, + "learning_rate": 3.720786428852461e-06, + "loss": 1.0579, + "step": 58837 + }, + { + "epoch": 4.35, + "learning_rate": 3.72047600132166e-06, + "loss": 1.0013, + "step": 58838 + }, + { + "epoch": 4.35, + "learning_rate": 3.7201655837814922e-06, + "loss": 0.9161, + "step": 58839 + }, + { + "epoch": 4.35, + "learning_rate": 3.7198551762324476e-06, + "loss": 0.974, + "step": 58840 + }, + { + "epoch": 4.35, + "learning_rate": 3.7195447786750303e-06, + "loss": 1.1356, + "step": 58841 + }, + { + "epoch": 4.35, + "learning_rate": 3.7192343911097207e-06, + "loss": 1.0252, + "step": 58842 + }, + { + "epoch": 4.35, + "learning_rate": 3.718924013537023e-06, + "loss": 1.0458, + "step": 58843 + }, + { + "epoch": 4.35, + "learning_rate": 3.7186136459574273e-06, + "loss": 0.993, + "step": 58844 + }, + { + "epoch": 4.35, + "learning_rate": 3.7183032883714266e-06, + "loss": 0.9176, + "step": 58845 + }, + { + "epoch": 4.35, + "learning_rate": 3.7179929407795134e-06, + "loss": 0.9413, + "step": 58846 + }, + { + "epoch": 4.35, + "learning_rate": 3.7176826031821855e-06, + "loss": 1.0107, + "step": 58847 + }, + { + "epoch": 4.35, + "learning_rate": 3.7173722755799356e-06, + "loss": 1.0739, + "step": 58848 + }, + { + "epoch": 4.35, + "learning_rate": 3.717061957973257e-06, + "loss": 0.9137, + "step": 58849 + }, + { + "epoch": 4.35, + "learning_rate": 3.716751650362642e-06, + "loss": 0.9708, + "step": 58850 + }, + { + "epoch": 4.35, + "learning_rate": 3.716441352748582e-06, + "loss": 0.9653, + "step": 58851 + }, + { + "epoch": 4.35, + "learning_rate": 3.7161310651315775e-06, + "loss": 0.9106, + "step": 58852 + }, + { + "epoch": 4.35, + "learning_rate": 3.7158207875121176e-06, + "loss": 1.0642, + "step": 58853 + }, + { + "epoch": 4.35, + "learning_rate": 3.7155105198906983e-06, + "loss": 1.0358, + "step": 58854 + }, + { + "epoch": 4.35, + "learning_rate": 3.7152002622678084e-06, + "loss": 0.9574, + "step": 58855 + }, + { + "epoch": 4.35, + "learning_rate": 3.7148900146439473e-06, + "loss": 1.0133, + "step": 58856 + }, + { + "epoch": 4.35, + "learning_rate": 3.714579777019607e-06, + "loss": 1.0115, + "step": 58857 + }, + { + "epoch": 4.35, + "learning_rate": 3.7142695493952808e-06, + "loss": 0.9473, + "step": 58858 + }, + { + "epoch": 4.35, + "learning_rate": 3.7139593317714606e-06, + "loss": 0.9401, + "step": 58859 + }, + { + "epoch": 4.35, + "learning_rate": 3.7136491241486384e-06, + "loss": 0.9637, + "step": 58860 + }, + { + "epoch": 4.35, + "learning_rate": 3.713338926527318e-06, + "loss": 0.9338, + "step": 58861 + }, + { + "epoch": 4.35, + "learning_rate": 3.713028738907978e-06, + "loss": 1.0402, + "step": 58862 + }, + { + "epoch": 4.35, + "learning_rate": 3.712718561291122e-06, + "loss": 0.8626, + "step": 58863 + }, + { + "epoch": 4.35, + "learning_rate": 3.712408393677238e-06, + "loss": 0.9541, + "step": 58864 + }, + { + "epoch": 4.35, + "learning_rate": 3.712098236066829e-06, + "loss": 1.025, + "step": 58865 + }, + { + "epoch": 4.35, + "learning_rate": 3.7117880884603753e-06, + "loss": 1.0305, + "step": 58866 + }, + { + "epoch": 4.35, + "learning_rate": 3.7114779508583796e-06, + "loss": 0.888, + "step": 58867 + }, + { + "epoch": 4.35, + "learning_rate": 3.7111678232613334e-06, + "loss": 1.0767, + "step": 58868 + }, + { + "epoch": 4.35, + "learning_rate": 3.7108577056697283e-06, + "loss": 0.9787, + "step": 58869 + }, + { + "epoch": 4.35, + "learning_rate": 3.7105475980840587e-06, + "loss": 1.0816, + "step": 58870 + }, + { + "epoch": 4.35, + "learning_rate": 3.7102375005048153e-06, + "loss": 0.8804, + "step": 58871 + }, + { + "epoch": 4.35, + "learning_rate": 3.709927412932497e-06, + "loss": 0.9788, + "step": 58872 + }, + { + "epoch": 4.35, + "learning_rate": 3.7096173353675936e-06, + "loss": 0.8916, + "step": 58873 + }, + { + "epoch": 4.35, + "learning_rate": 3.7093072678106e-06, + "loss": 1.0696, + "step": 58874 + }, + { + "epoch": 4.35, + "learning_rate": 3.708997210262004e-06, + "loss": 1.0119, + "step": 58875 + }, + { + "epoch": 4.35, + "learning_rate": 3.7086871627223077e-06, + "loss": 1.0453, + "step": 58876 + }, + { + "epoch": 4.35, + "learning_rate": 3.7083771251919998e-06, + "loss": 0.9926, + "step": 58877 + }, + { + "epoch": 4.35, + "learning_rate": 3.708067097671574e-06, + "loss": 0.9697, + "step": 58878 + }, + { + "epoch": 4.35, + "learning_rate": 3.7077570801615237e-06, + "loss": 1.1401, + "step": 58879 + }, + { + "epoch": 4.35, + "learning_rate": 3.7074470726623378e-06, + "loss": 0.9998, + "step": 58880 + }, + { + "epoch": 4.35, + "learning_rate": 3.707137075174517e-06, + "loss": 0.998, + "step": 58881 + }, + { + "epoch": 4.35, + "learning_rate": 3.7068270876985514e-06, + "loss": 1.0244, + "step": 58882 + }, + { + "epoch": 4.35, + "learning_rate": 3.706517110234934e-06, + "loss": 0.9416, + "step": 58883 + }, + { + "epoch": 4.35, + "learning_rate": 3.706207142784154e-06, + "loss": 0.9761, + "step": 58884 + }, + { + "epoch": 4.35, + "learning_rate": 3.7058971853467162e-06, + "loss": 1.0337, + "step": 58885 + }, + { + "epoch": 4.35, + "learning_rate": 3.705587237923102e-06, + "loss": 0.9131, + "step": 58886 + }, + { + "epoch": 4.35, + "learning_rate": 3.7052773005138044e-06, + "loss": 0.9544, + "step": 58887 + }, + { + "epoch": 4.35, + "learning_rate": 3.7049673731193247e-06, + "loss": 1.0322, + "step": 58888 + }, + { + "epoch": 4.35, + "learning_rate": 3.7046574557401516e-06, + "loss": 1.0452, + "step": 58889 + }, + { + "epoch": 4.35, + "learning_rate": 3.7043475483767776e-06, + "loss": 0.9816, + "step": 58890 + }, + { + "epoch": 4.35, + "learning_rate": 3.704037651029694e-06, + "loss": 1.1016, + "step": 58891 + }, + { + "epoch": 4.35, + "learning_rate": 3.7037277636993996e-06, + "loss": 1.0961, + "step": 58892 + }, + { + "epoch": 4.35, + "learning_rate": 3.7034178863863845e-06, + "loss": 0.975, + "step": 58893 + }, + { + "epoch": 4.35, + "learning_rate": 3.7031080190911417e-06, + "loss": 0.9832, + "step": 58894 + }, + { + "epoch": 4.35, + "learning_rate": 3.7027981618141638e-06, + "loss": 1.0687, + "step": 58895 + }, + { + "epoch": 4.35, + "learning_rate": 3.7024883145559398e-06, + "loss": 0.9298, + "step": 58896 + }, + { + "epoch": 4.35, + "learning_rate": 3.7021784773169712e-06, + "loss": 1.0962, + "step": 58897 + }, + { + "epoch": 4.35, + "learning_rate": 3.7018686500977464e-06, + "loss": 1.0371, + "step": 58898 + }, + { + "epoch": 4.35, + "learning_rate": 3.7015588328987574e-06, + "loss": 0.8381, + "step": 58899 + }, + { + "epoch": 4.35, + "learning_rate": 3.7012490257204956e-06, + "loss": 1.0624, + "step": 58900 + }, + { + "epoch": 4.35, + "learning_rate": 3.7009392285634594e-06, + "loss": 0.9661, + "step": 58901 + }, + { + "epoch": 4.35, + "learning_rate": 3.7006294414281397e-06, + "loss": 1.038, + "step": 58902 + }, + { + "epoch": 4.35, + "learning_rate": 3.7003196643150274e-06, + "loss": 0.9597, + "step": 58903 + }, + { + "epoch": 4.35, + "learning_rate": 3.7000098972246136e-06, + "loss": 1.0051, + "step": 58904 + }, + { + "epoch": 4.35, + "learning_rate": 3.6997001401573974e-06, + "loss": 1.0009, + "step": 58905 + }, + { + "epoch": 4.35, + "learning_rate": 3.6993903931138708e-06, + "loss": 0.8946, + "step": 58906 + }, + { + "epoch": 4.35, + "learning_rate": 3.699080656094518e-06, + "loss": 1.0418, + "step": 58907 + }, + { + "epoch": 4.35, + "learning_rate": 3.698770929099841e-06, + "loss": 1.0105, + "step": 58908 + }, + { + "epoch": 4.35, + "learning_rate": 3.6984612121303255e-06, + "loss": 0.9675, + "step": 58909 + }, + { + "epoch": 4.35, + "learning_rate": 3.698151505186475e-06, + "loss": 0.9222, + "step": 58910 + }, + { + "epoch": 4.35, + "learning_rate": 3.6978418082687682e-06, + "loss": 0.9209, + "step": 58911 + }, + { + "epoch": 4.35, + "learning_rate": 3.6975321213777095e-06, + "loss": 0.9562, + "step": 58912 + }, + { + "epoch": 4.35, + "learning_rate": 3.697222444513782e-06, + "loss": 1.0666, + "step": 58913 + }, + { + "epoch": 4.35, + "learning_rate": 3.6969127776774915e-06, + "loss": 1.0486, + "step": 58914 + }, + { + "epoch": 4.35, + "learning_rate": 3.6966031208693186e-06, + "loss": 1.0355, + "step": 58915 + }, + { + "epoch": 4.35, + "learning_rate": 3.696293474089756e-06, + "loss": 0.9501, + "step": 58916 + }, + { + "epoch": 4.35, + "learning_rate": 3.695983837339304e-06, + "loss": 1.0077, + "step": 58917 + }, + { + "epoch": 4.35, + "learning_rate": 3.6956742106184517e-06, + "loss": 1.0046, + "step": 58918 + }, + { + "epoch": 4.35, + "learning_rate": 3.6953645939276917e-06, + "loss": 0.8885, + "step": 58919 + }, + { + "epoch": 4.35, + "learning_rate": 3.6950549872675136e-06, + "loss": 1.1015, + "step": 58920 + }, + { + "epoch": 4.35, + "learning_rate": 3.6947453906384156e-06, + "loss": 0.9971, + "step": 58921 + }, + { + "epoch": 4.35, + "learning_rate": 3.6944358040408867e-06, + "loss": 1.041, + "step": 58922 + }, + { + "epoch": 4.35, + "learning_rate": 3.694126227475422e-06, + "loss": 1.0121, + "step": 58923 + }, + { + "epoch": 4.35, + "learning_rate": 3.693816660942511e-06, + "loss": 0.9639, + "step": 58924 + }, + { + "epoch": 4.35, + "learning_rate": 3.693507104442644e-06, + "loss": 1.0549, + "step": 58925 + }, + { + "epoch": 4.35, + "learning_rate": 3.6931975579763202e-06, + "loss": 1.0491, + "step": 58926 + }, + { + "epoch": 4.35, + "learning_rate": 3.69288802154403e-06, + "loss": 1.055, + "step": 58927 + }, + { + "epoch": 4.35, + "learning_rate": 3.692578495146264e-06, + "loss": 1.0914, + "step": 58928 + }, + { + "epoch": 4.35, + "learning_rate": 3.692268978783512e-06, + "loss": 1.0048, + "step": 58929 + }, + { + "epoch": 4.35, + "learning_rate": 3.691959472456276e-06, + "loss": 0.93, + "step": 58930 + }, + { + "epoch": 4.35, + "learning_rate": 3.691649976165036e-06, + "loss": 1.0784, + "step": 58931 + }, + { + "epoch": 4.35, + "learning_rate": 3.6913404899102944e-06, + "loss": 0.9688, + "step": 58932 + }, + { + "epoch": 4.35, + "learning_rate": 3.691031013692539e-06, + "loss": 1.0198, + "step": 58933 + }, + { + "epoch": 4.35, + "learning_rate": 3.690721547512264e-06, + "loss": 0.9947, + "step": 58934 + }, + { + "epoch": 4.35, + "learning_rate": 3.690412091369959e-06, + "loss": 1.1133, + "step": 58935 + }, + { + "epoch": 4.35, + "learning_rate": 3.6901026452661172e-06, + "loss": 0.9237, + "step": 58936 + }, + { + "epoch": 4.35, + "learning_rate": 3.6897932092012335e-06, + "loss": 0.9834, + "step": 58937 + }, + { + "epoch": 4.35, + "learning_rate": 3.689483783175799e-06, + "loss": 1.0502, + "step": 58938 + }, + { + "epoch": 4.35, + "learning_rate": 3.6891743671903046e-06, + "loss": 0.9931, + "step": 58939 + }, + { + "epoch": 4.36, + "learning_rate": 3.688864961245241e-06, + "loss": 1.0699, + "step": 58940 + }, + { + "epoch": 4.36, + "learning_rate": 3.6885555653411063e-06, + "loss": 0.8321, + "step": 58941 + }, + { + "epoch": 4.36, + "learning_rate": 3.68824617947839e-06, + "loss": 0.9375, + "step": 58942 + }, + { + "epoch": 4.36, + "learning_rate": 3.6879368036575822e-06, + "loss": 1.0254, + "step": 58943 + }, + { + "epoch": 4.36, + "learning_rate": 3.687627437879178e-06, + "loss": 1.1856, + "step": 58944 + }, + { + "epoch": 4.36, + "learning_rate": 3.687318082143664e-06, + "loss": 0.9922, + "step": 58945 + }, + { + "epoch": 4.36, + "learning_rate": 3.6870087364515405e-06, + "loss": 1.0337, + "step": 58946 + }, + { + "epoch": 4.36, + "learning_rate": 3.6866994008032953e-06, + "loss": 0.8922, + "step": 58947 + }, + { + "epoch": 4.36, + "learning_rate": 3.686390075199422e-06, + "loss": 1.0723, + "step": 58948 + }, + { + "epoch": 4.36, + "learning_rate": 3.686080759640408e-06, + "loss": 1.091, + "step": 58949 + }, + { + "epoch": 4.36, + "learning_rate": 3.685771454126752e-06, + "loss": 1.0009, + "step": 58950 + }, + { + "epoch": 4.36, + "learning_rate": 3.6854621586589466e-06, + "loss": 1.0165, + "step": 58951 + }, + { + "epoch": 4.36, + "learning_rate": 3.685152873237474e-06, + "loss": 0.9641, + "step": 58952 + }, + { + "epoch": 4.36, + "learning_rate": 3.684843597862836e-06, + "loss": 0.9417, + "step": 58953 + }, + { + "epoch": 4.36, + "learning_rate": 3.684534332535519e-06, + "loss": 1.0213, + "step": 58954 + }, + { + "epoch": 4.36, + "learning_rate": 3.684225077256024e-06, + "loss": 0.994, + "step": 58955 + }, + { + "epoch": 4.36, + "learning_rate": 3.6839158320248293e-06, + "loss": 0.9777, + "step": 58956 + }, + { + "epoch": 4.36, + "learning_rate": 3.6836065968424373e-06, + "loss": 0.9866, + "step": 58957 + }, + { + "epoch": 4.36, + "learning_rate": 3.6832973717093345e-06, + "loss": 0.9351, + "step": 58958 + }, + { + "epoch": 4.36, + "learning_rate": 3.6829881566260204e-06, + "loss": 1.074, + "step": 58959 + }, + { + "epoch": 4.36, + "learning_rate": 3.6826789515929763e-06, + "loss": 0.9669, + "step": 58960 + }, + { + "epoch": 4.36, + "learning_rate": 3.682369756610703e-06, + "loss": 1.0634, + "step": 58961 + }, + { + "epoch": 4.36, + "learning_rate": 3.6820605716796885e-06, + "loss": 0.9735, + "step": 58962 + }, + { + "epoch": 4.36, + "learning_rate": 3.681751396800425e-06, + "loss": 0.9928, + "step": 58963 + }, + { + "epoch": 4.36, + "learning_rate": 3.6814422319734056e-06, + "loss": 1.0101, + "step": 58964 + }, + { + "epoch": 4.36, + "learning_rate": 3.6811330771991174e-06, + "loss": 0.951, + "step": 58965 + }, + { + "epoch": 4.36, + "learning_rate": 3.6808239324780594e-06, + "loss": 0.9678, + "step": 58966 + }, + { + "epoch": 4.36, + "learning_rate": 3.6805147978107193e-06, + "loss": 1.018, + "step": 58967 + }, + { + "epoch": 4.36, + "learning_rate": 3.680205673197591e-06, + "loss": 0.9612, + "step": 58968 + }, + { + "epoch": 4.36, + "learning_rate": 3.679896558639161e-06, + "loss": 0.9772, + "step": 58969 + }, + { + "epoch": 4.36, + "learning_rate": 3.679587454135928e-06, + "loss": 0.98, + "step": 58970 + }, + { + "epoch": 4.36, + "learning_rate": 3.679278359688382e-06, + "loss": 1.005, + "step": 58971 + }, + { + "epoch": 4.36, + "learning_rate": 3.678969275297013e-06, + "loss": 0.9677, + "step": 58972 + }, + { + "epoch": 4.36, + "learning_rate": 3.678660200962314e-06, + "loss": 0.8711, + "step": 58973 + }, + { + "epoch": 4.36, + "learning_rate": 3.6783511366847723e-06, + "loss": 0.9073, + "step": 58974 + }, + { + "epoch": 4.36, + "learning_rate": 3.67804208246489e-06, + "loss": 1.0044, + "step": 58975 + }, + { + "epoch": 4.36, + "learning_rate": 3.677733038303146e-06, + "loss": 1.0295, + "step": 58976 + }, + { + "epoch": 4.36, + "learning_rate": 3.677424004200042e-06, + "loss": 0.8888, + "step": 58977 + }, + { + "epoch": 4.36, + "learning_rate": 3.6771149801560625e-06, + "loss": 0.9891, + "step": 58978 + }, + { + "epoch": 4.36, + "learning_rate": 3.676805966171709e-06, + "loss": 1.0217, + "step": 58979 + }, + { + "epoch": 4.36, + "learning_rate": 3.6764969622474633e-06, + "loss": 0.864, + "step": 58980 + }, + { + "epoch": 4.36, + "learning_rate": 3.6761879683838173e-06, + "loss": 1.0251, + "step": 58981 + }, + { + "epoch": 4.36, + "learning_rate": 3.6758789845812694e-06, + "loss": 1.0669, + "step": 58982 + }, + { + "epoch": 4.36, + "learning_rate": 3.675570010840307e-06, + "loss": 0.9227, + "step": 58983 + }, + { + "epoch": 4.36, + "learning_rate": 3.6752610471614225e-06, + "loss": 0.9329, + "step": 58984 + }, + { + "epoch": 4.36, + "learning_rate": 3.674952093545103e-06, + "loss": 1.0519, + "step": 58985 + }, + { + "epoch": 4.36, + "learning_rate": 3.674643149991849e-06, + "loss": 1.0803, + "step": 58986 + }, + { + "epoch": 4.36, + "learning_rate": 3.6743342165021467e-06, + "loss": 1.0245, + "step": 58987 + }, + { + "epoch": 4.36, + "learning_rate": 3.6740252930764876e-06, + "loss": 1.1207, + "step": 58988 + }, + { + "epoch": 4.36, + "learning_rate": 3.673716379715364e-06, + "loss": 1.0848, + "step": 58989 + }, + { + "epoch": 4.36, + "learning_rate": 3.673407476419264e-06, + "loss": 1.0696, + "step": 58990 + }, + { + "epoch": 4.36, + "learning_rate": 3.6730985831886856e-06, + "loss": 1.0706, + "step": 58991 + }, + { + "epoch": 4.36, + "learning_rate": 3.6727897000241164e-06, + "loss": 1.0266, + "step": 58992 + }, + { + "epoch": 4.36, + "learning_rate": 3.6724808269260494e-06, + "loss": 1.0331, + "step": 58993 + }, + { + "epoch": 4.36, + "learning_rate": 3.6721719638949706e-06, + "loss": 1.0455, + "step": 58994 + }, + { + "epoch": 4.36, + "learning_rate": 3.6718631109313785e-06, + "loss": 0.8902, + "step": 58995 + }, + { + "epoch": 4.36, + "learning_rate": 3.671554268035762e-06, + "loss": 0.9588, + "step": 58996 + }, + { + "epoch": 4.36, + "learning_rate": 3.6712454352086126e-06, + "loss": 1.0546, + "step": 58997 + }, + { + "epoch": 4.36, + "learning_rate": 3.670936612450421e-06, + "loss": 0.9707, + "step": 58998 + }, + { + "epoch": 4.36, + "learning_rate": 3.6706277997616755e-06, + "loss": 0.9868, + "step": 58999 + }, + { + "epoch": 4.36, + "learning_rate": 3.6703189971428767e-06, + "loss": 1.0388, + "step": 59000 + }, + { + "epoch": 4.36, + "learning_rate": 3.6700102045945042e-06, + "loss": 1.029, + "step": 59001 + }, + { + "epoch": 4.36, + "learning_rate": 3.6697014221170566e-06, + "loss": 0.9924, + "step": 59002 + }, + { + "epoch": 4.36, + "learning_rate": 3.669392649711021e-06, + "loss": 1.0703, + "step": 59003 + }, + { + "epoch": 4.36, + "learning_rate": 3.6690838873768986e-06, + "loss": 0.8359, + "step": 59004 + }, + { + "epoch": 4.36, + "learning_rate": 3.6687751351151656e-06, + "loss": 0.9278, + "step": 59005 + }, + { + "epoch": 4.36, + "learning_rate": 3.6684663929263243e-06, + "loss": 1.0145, + "step": 59006 + }, + { + "epoch": 4.36, + "learning_rate": 3.6681576608108625e-06, + "loss": 0.9458, + "step": 59007 + }, + { + "epoch": 4.36, + "learning_rate": 3.667848938769271e-06, + "loss": 1.0662, + "step": 59008 + }, + { + "epoch": 4.36, + "learning_rate": 3.667540226802041e-06, + "loss": 1.0075, + "step": 59009 + }, + { + "epoch": 4.36, + "learning_rate": 3.667231524909661e-06, + "loss": 1.1482, + "step": 59010 + }, + { + "epoch": 4.36, + "learning_rate": 3.6669228330926287e-06, + "loss": 0.9739, + "step": 59011 + }, + { + "epoch": 4.36, + "learning_rate": 3.666614151351431e-06, + "loss": 1.0933, + "step": 59012 + }, + { + "epoch": 4.36, + "learning_rate": 3.666305479686559e-06, + "loss": 0.9632, + "step": 59013 + }, + { + "epoch": 4.36, + "learning_rate": 3.665996818098503e-06, + "loss": 1.0167, + "step": 59014 + }, + { + "epoch": 4.36, + "learning_rate": 3.6656881665877574e-06, + "loss": 0.9049, + "step": 59015 + }, + { + "epoch": 4.36, + "learning_rate": 3.6653795251548108e-06, + "loss": 0.9804, + "step": 59016 + }, + { + "epoch": 4.36, + "learning_rate": 3.6650708938001556e-06, + "loss": 1.0148, + "step": 59017 + }, + { + "epoch": 4.36, + "learning_rate": 3.6647622725242816e-06, + "loss": 0.9943, + "step": 59018 + }, + { + "epoch": 4.36, + "learning_rate": 3.664453661327677e-06, + "loss": 0.9639, + "step": 59019 + }, + { + "epoch": 4.36, + "learning_rate": 3.664145060210843e-06, + "loss": 0.9692, + "step": 59020 + }, + { + "epoch": 4.36, + "learning_rate": 3.6638364691742567e-06, + "loss": 1.0085, + "step": 59021 + }, + { + "epoch": 4.36, + "learning_rate": 3.663527888218419e-06, + "loss": 1.0502, + "step": 59022 + }, + { + "epoch": 4.36, + "learning_rate": 3.6632193173438145e-06, + "loss": 1.0524, + "step": 59023 + }, + { + "epoch": 4.36, + "learning_rate": 3.6629107565509434e-06, + "loss": 1.104, + "step": 59024 + }, + { + "epoch": 4.36, + "learning_rate": 3.6626022058402856e-06, + "loss": 0.9724, + "step": 59025 + }, + { + "epoch": 4.36, + "learning_rate": 3.6622936652123387e-06, + "loss": 1.0785, + "step": 59026 + }, + { + "epoch": 4.36, + "learning_rate": 3.6619851346675927e-06, + "loss": 1.131, + "step": 59027 + }, + { + "epoch": 4.36, + "learning_rate": 3.6616766142065372e-06, + "loss": 1.0951, + "step": 59028 + }, + { + "epoch": 4.36, + "learning_rate": 3.6613681038296643e-06, + "loss": 0.93, + "step": 59029 + }, + { + "epoch": 4.36, + "learning_rate": 3.6610596035374603e-06, + "loss": 1.0215, + "step": 59030 + }, + { + "epoch": 4.36, + "learning_rate": 3.660751113330423e-06, + "loss": 1.0222, + "step": 59031 + }, + { + "epoch": 4.36, + "learning_rate": 3.66044263320904e-06, + "loss": 0.9005, + "step": 59032 + }, + { + "epoch": 4.36, + "learning_rate": 3.6601341631738017e-06, + "loss": 0.9993, + "step": 59033 + }, + { + "epoch": 4.36, + "learning_rate": 3.659825703225197e-06, + "loss": 0.9674, + "step": 59034 + }, + { + "epoch": 4.36, + "learning_rate": 3.659517253363721e-06, + "loss": 0.9692, + "step": 59035 + }, + { + "epoch": 4.36, + "learning_rate": 3.659208813589863e-06, + "loss": 1.0557, + "step": 59036 + }, + { + "epoch": 4.36, + "learning_rate": 3.6589003839041137e-06, + "loss": 1.0051, + "step": 59037 + }, + { + "epoch": 4.36, + "learning_rate": 3.6585919643069623e-06, + "loss": 1.0184, + "step": 59038 + }, + { + "epoch": 4.36, + "learning_rate": 3.658283554798897e-06, + "loss": 1.0076, + "step": 59039 + }, + { + "epoch": 4.36, + "learning_rate": 3.6579751553804154e-06, + "loss": 0.9478, + "step": 59040 + }, + { + "epoch": 4.36, + "learning_rate": 3.6576667660520037e-06, + "loss": 0.968, + "step": 59041 + }, + { + "epoch": 4.36, + "learning_rate": 3.657358386814155e-06, + "loss": 1.0565, + "step": 59042 + }, + { + "epoch": 4.36, + "learning_rate": 3.657050017667354e-06, + "loss": 1.0188, + "step": 59043 + }, + { + "epoch": 4.36, + "learning_rate": 3.6567416586120997e-06, + "loss": 0.9667, + "step": 59044 + }, + { + "epoch": 4.36, + "learning_rate": 3.6564333096488813e-06, + "loss": 1.0135, + "step": 59045 + }, + { + "epoch": 4.36, + "learning_rate": 3.65612497077818e-06, + "loss": 1.0394, + "step": 59046 + }, + { + "epoch": 4.36, + "learning_rate": 3.655816642000497e-06, + "loss": 1.0293, + "step": 59047 + }, + { + "epoch": 4.36, + "learning_rate": 3.6555083233163158e-06, + "loss": 1.0262, + "step": 59048 + }, + { + "epoch": 4.36, + "learning_rate": 3.6552000147261355e-06, + "loss": 1.0621, + "step": 59049 + }, + { + "epoch": 4.36, + "learning_rate": 3.6548917162304353e-06, + "loss": 0.9695, + "step": 59050 + }, + { + "epoch": 4.36, + "learning_rate": 3.6545834278297156e-06, + "loss": 1.0357, + "step": 59051 + }, + { + "epoch": 4.36, + "learning_rate": 3.6542751495244623e-06, + "loss": 1.0087, + "step": 59052 + }, + { + "epoch": 4.36, + "learning_rate": 3.6539668813151663e-06, + "loss": 1.074, + "step": 59053 + }, + { + "epoch": 4.36, + "learning_rate": 3.6536586232023187e-06, + "loss": 0.9107, + "step": 59054 + }, + { + "epoch": 4.36, + "learning_rate": 3.6533503751864054e-06, + "loss": 0.9597, + "step": 59055 + }, + { + "epoch": 4.36, + "learning_rate": 3.653042137267925e-06, + "loss": 1.0204, + "step": 59056 + }, + { + "epoch": 4.36, + "learning_rate": 3.6527339094473624e-06, + "loss": 0.917, + "step": 59057 + }, + { + "epoch": 4.36, + "learning_rate": 3.652425691725211e-06, + "loss": 0.9607, + "step": 59058 + }, + { + "epoch": 4.36, + "learning_rate": 3.652117484101955e-06, + "loss": 0.9588, + "step": 59059 + }, + { + "epoch": 4.36, + "learning_rate": 3.651809286578093e-06, + "loss": 0.9636, + "step": 59060 + }, + { + "epoch": 4.36, + "learning_rate": 3.6515010991541113e-06, + "loss": 0.9959, + "step": 59061 + }, + { + "epoch": 4.36, + "learning_rate": 3.651192921830501e-06, + "loss": 0.9421, + "step": 59062 + }, + { + "epoch": 4.36, + "learning_rate": 3.6508847546077474e-06, + "loss": 1.0207, + "step": 59063 + }, + { + "epoch": 4.36, + "learning_rate": 3.6505765974863506e-06, + "loss": 1.0535, + "step": 59064 + }, + { + "epoch": 4.36, + "learning_rate": 3.6502684504667974e-06, + "loss": 1.0521, + "step": 59065 + }, + { + "epoch": 4.36, + "learning_rate": 3.64996031354957e-06, + "loss": 1.0353, + "step": 59066 + }, + { + "epoch": 4.36, + "learning_rate": 3.6496521867351687e-06, + "loss": 0.8871, + "step": 59067 + }, + { + "epoch": 4.36, + "learning_rate": 3.6493440700240756e-06, + "loss": 0.9433, + "step": 59068 + }, + { + "epoch": 4.36, + "learning_rate": 3.6490359634167926e-06, + "loss": 1.0005, + "step": 59069 + }, + { + "epoch": 4.36, + "learning_rate": 3.648727866913796e-06, + "loss": 1.0646, + "step": 59070 + }, + { + "epoch": 4.36, + "learning_rate": 3.6484197805155854e-06, + "loss": 1.0333, + "step": 59071 + }, + { + "epoch": 4.36, + "learning_rate": 3.6481117042226455e-06, + "loss": 0.9294, + "step": 59072 + }, + { + "epoch": 4.36, + "learning_rate": 3.6478036380354753e-06, + "loss": 0.9961, + "step": 59073 + }, + { + "epoch": 4.36, + "learning_rate": 3.6474955819545542e-06, + "loss": 0.9872, + "step": 59074 + }, + { + "epoch": 4.37, + "learning_rate": 3.6471875359803743e-06, + "loss": 1.0503, + "step": 59075 + }, + { + "epoch": 4.37, + "learning_rate": 3.6468795001134316e-06, + "loss": 0.9709, + "step": 59076 + }, + { + "epoch": 4.37, + "learning_rate": 3.6465714743542123e-06, + "loss": 1.0941, + "step": 59077 + }, + { + "epoch": 4.37, + "learning_rate": 3.6462634587032075e-06, + "loss": 1.0365, + "step": 59078 + }, + { + "epoch": 4.37, + "learning_rate": 3.6459554531609032e-06, + "loss": 0.9245, + "step": 59079 + }, + { + "epoch": 4.37, + "learning_rate": 3.6456474577277957e-06, + "loss": 0.9878, + "step": 59080 + }, + { + "epoch": 4.37, + "learning_rate": 3.6453394724043724e-06, + "loss": 1.0074, + "step": 59081 + }, + { + "epoch": 4.37, + "learning_rate": 3.645031497191123e-06, + "loss": 1.0552, + "step": 59082 + }, + { + "epoch": 4.37, + "learning_rate": 3.6447235320885376e-06, + "loss": 0.9818, + "step": 59083 + }, + { + "epoch": 4.37, + "learning_rate": 3.6444155770971024e-06, + "loss": 1.0062, + "step": 59084 + }, + { + "epoch": 4.37, + "learning_rate": 3.6441076322173153e-06, + "loss": 1.1222, + "step": 59085 + }, + { + "epoch": 4.37, + "learning_rate": 3.6437996974496616e-06, + "loss": 1.0241, + "step": 59086 + }, + { + "epoch": 4.37, + "learning_rate": 3.643491772794632e-06, + "loss": 0.9309, + "step": 59087 + }, + { + "epoch": 4.37, + "learning_rate": 3.6431838582527112e-06, + "loss": 0.9113, + "step": 59088 + }, + { + "epoch": 4.37, + "learning_rate": 3.6428759538243983e-06, + "loss": 1.018, + "step": 59089 + }, + { + "epoch": 4.37, + "learning_rate": 3.6425680595101797e-06, + "loss": 1.0048, + "step": 59090 + }, + { + "epoch": 4.37, + "learning_rate": 3.6422601753105434e-06, + "loss": 0.9952, + "step": 59091 + }, + { + "epoch": 4.37, + "learning_rate": 3.6419523012259805e-06, + "loss": 1.0542, + "step": 59092 + }, + { + "epoch": 4.37, + "learning_rate": 3.641644437256977e-06, + "loss": 0.9848, + "step": 59093 + }, + { + "epoch": 4.37, + "learning_rate": 3.6413365834040326e-06, + "loss": 0.9179, + "step": 59094 + }, + { + "epoch": 4.37, + "learning_rate": 3.6410287396676247e-06, + "loss": 0.9668, + "step": 59095 + }, + { + "epoch": 4.37, + "learning_rate": 3.6407209060482516e-06, + "loss": 0.859, + "step": 59096 + }, + { + "epoch": 4.37, + "learning_rate": 3.640413082546401e-06, + "loss": 1.0112, + "step": 59097 + }, + { + "epoch": 4.37, + "learning_rate": 3.6401052691625626e-06, + "loss": 1.051, + "step": 59098 + }, + { + "epoch": 4.37, + "learning_rate": 3.6397974658972225e-06, + "loss": 0.9394, + "step": 59099 + }, + { + "epoch": 4.37, + "learning_rate": 3.6394896727508755e-06, + "loss": 0.9774, + "step": 59100 + }, + { + "epoch": 4.37, + "learning_rate": 3.6391818897240104e-06, + "loss": 1.0785, + "step": 59101 + }, + { + "epoch": 4.37, + "learning_rate": 3.638874116817116e-06, + "loss": 1.0292, + "step": 59102 + }, + { + "epoch": 4.37, + "learning_rate": 3.6385663540306824e-06, + "loss": 1.0029, + "step": 59103 + }, + { + "epoch": 4.37, + "learning_rate": 3.6382586013651945e-06, + "loss": 0.9024, + "step": 59104 + }, + { + "epoch": 4.37, + "learning_rate": 3.6379508588211497e-06, + "loss": 0.9774, + "step": 59105 + }, + { + "epoch": 4.37, + "learning_rate": 3.6376431263990343e-06, + "loss": 1.0065, + "step": 59106 + }, + { + "epoch": 4.37, + "learning_rate": 3.637335404099337e-06, + "loss": 1.0014, + "step": 59107 + }, + { + "epoch": 4.37, + "learning_rate": 3.6370276919225458e-06, + "loss": 1.0138, + "step": 59108 + }, + { + "epoch": 4.37, + "learning_rate": 3.636719989869155e-06, + "loss": 0.9436, + "step": 59109 + }, + { + "epoch": 4.37, + "learning_rate": 3.6364122979396556e-06, + "loss": 0.9489, + "step": 59110 + }, + { + "epoch": 4.37, + "learning_rate": 3.636104616134526e-06, + "loss": 1.0498, + "step": 59111 + }, + { + "epoch": 4.37, + "learning_rate": 3.6357969444542662e-06, + "loss": 1.0256, + "step": 59112 + }, + { + "epoch": 4.37, + "learning_rate": 3.6354892828993593e-06, + "loss": 1.0836, + "step": 59113 + }, + { + "epoch": 4.37, + "learning_rate": 3.6351816314703035e-06, + "loss": 1.0891, + "step": 59114 + }, + { + "epoch": 4.37, + "learning_rate": 3.634873990167578e-06, + "loss": 0.9164, + "step": 59115 + }, + { + "epoch": 4.37, + "learning_rate": 3.634566358991679e-06, + "loss": 0.9441, + "step": 59116 + }, + { + "epoch": 4.37, + "learning_rate": 3.6342587379430905e-06, + "loss": 0.9816, + "step": 59117 + }, + { + "epoch": 4.37, + "learning_rate": 3.6339511270223127e-06, + "loss": 1.0942, + "step": 59118 + }, + { + "epoch": 4.37, + "learning_rate": 3.63364352622982e-06, + "loss": 1.0003, + "step": 59119 + }, + { + "epoch": 4.37, + "learning_rate": 3.6333359355661124e-06, + "loss": 0.9312, + "step": 59120 + }, + { + "epoch": 4.37, + "learning_rate": 3.6330283550316767e-06, + "loss": 1.0029, + "step": 59121 + }, + { + "epoch": 4.37, + "learning_rate": 3.632720784627003e-06, + "loss": 0.9591, + "step": 59122 + }, + { + "epoch": 4.37, + "learning_rate": 3.6324132243525777e-06, + "loss": 1.0067, + "step": 59123 + }, + { + "epoch": 4.37, + "learning_rate": 3.63210567420889e-06, + "loss": 0.9564, + "step": 59124 + }, + { + "epoch": 4.37, + "learning_rate": 3.6317981341964336e-06, + "loss": 1.1304, + "step": 59125 + }, + { + "epoch": 4.37, + "learning_rate": 3.631490604315695e-06, + "loss": 1.0518, + "step": 59126 + }, + { + "epoch": 4.37, + "learning_rate": 3.6311830845671636e-06, + "loss": 0.9924, + "step": 59127 + }, + { + "epoch": 4.37, + "learning_rate": 3.630875574951326e-06, + "loss": 0.8987, + "step": 59128 + }, + { + "epoch": 4.37, + "learning_rate": 3.6305680754686777e-06, + "loss": 0.9635, + "step": 59129 + }, + { + "epoch": 4.37, + "learning_rate": 3.6302605861197037e-06, + "loss": 1.0683, + "step": 59130 + }, + { + "epoch": 4.37, + "learning_rate": 3.6299531069048954e-06, + "loss": 0.9171, + "step": 59131 + }, + { + "epoch": 4.37, + "learning_rate": 3.6296456378247393e-06, + "loss": 1.1581, + "step": 59132 + }, + { + "epoch": 4.37, + "learning_rate": 3.6293381788797233e-06, + "loss": 0.992, + "step": 59133 + }, + { + "epoch": 4.37, + "learning_rate": 3.6290307300703408e-06, + "loss": 1.0567, + "step": 59134 + }, + { + "epoch": 4.37, + "learning_rate": 3.6287232913970814e-06, + "loss": 0.8999, + "step": 59135 + }, + { + "epoch": 4.37, + "learning_rate": 3.6284158628604305e-06, + "loss": 1.0177, + "step": 59136 + }, + { + "epoch": 4.37, + "learning_rate": 3.628108444460876e-06, + "loss": 0.9741, + "step": 59137 + }, + { + "epoch": 4.37, + "learning_rate": 3.6278010361989136e-06, + "loss": 1.0405, + "step": 59138 + }, + { + "epoch": 4.37, + "learning_rate": 3.6274936380750314e-06, + "loss": 0.9397, + "step": 59139 + }, + { + "epoch": 4.37, + "learning_rate": 3.6271862500897093e-06, + "loss": 0.9201, + "step": 59140 + }, + { + "epoch": 4.37, + "learning_rate": 3.6268788722434457e-06, + "loss": 1.0121, + "step": 59141 + }, + { + "epoch": 4.37, + "learning_rate": 3.6265715045367268e-06, + "loss": 1.0028, + "step": 59142 + }, + { + "epoch": 4.37, + "learning_rate": 3.626264146970042e-06, + "loss": 1.0897, + "step": 59143 + }, + { + "epoch": 4.37, + "learning_rate": 3.625956799543875e-06, + "loss": 0.9953, + "step": 59144 + }, + { + "epoch": 4.37, + "learning_rate": 3.6256494622587246e-06, + "loss": 1.0089, + "step": 59145 + }, + { + "epoch": 4.37, + "learning_rate": 3.625342135115073e-06, + "loss": 1.0393, + "step": 59146 + }, + { + "epoch": 4.37, + "learning_rate": 3.625034818113412e-06, + "loss": 0.9965, + "step": 59147 + }, + { + "epoch": 4.37, + "learning_rate": 3.624727511254229e-06, + "loss": 0.8964, + "step": 59148 + }, + { + "epoch": 4.37, + "learning_rate": 3.62442021453801e-06, + "loss": 0.9217, + "step": 59149 + }, + { + "epoch": 4.37, + "learning_rate": 3.624112927965251e-06, + "loss": 0.8649, + "step": 59150 + }, + { + "epoch": 4.37, + "learning_rate": 3.6238056515364363e-06, + "loss": 0.878, + "step": 59151 + }, + { + "epoch": 4.37, + "learning_rate": 3.6234983852520568e-06, + "loss": 0.9187, + "step": 59152 + }, + { + "epoch": 4.37, + "learning_rate": 3.6231911291125955e-06, + "loss": 0.9959, + "step": 59153 + }, + { + "epoch": 4.37, + "learning_rate": 3.6228838831185496e-06, + "loss": 1.1466, + "step": 59154 + }, + { + "epoch": 4.37, + "learning_rate": 3.6225766472704038e-06, + "loss": 1.0543, + "step": 59155 + }, + { + "epoch": 4.37, + "learning_rate": 3.6222694215686483e-06, + "loss": 0.9159, + "step": 59156 + }, + { + "epoch": 4.37, + "learning_rate": 3.6219622060137703e-06, + "loss": 1.0859, + "step": 59157 + }, + { + "epoch": 4.37, + "learning_rate": 3.6216550006062556e-06, + "loss": 1.0975, + "step": 59158 + }, + { + "epoch": 4.37, + "learning_rate": 3.6213478053466047e-06, + "loss": 0.9718, + "step": 59159 + }, + { + "epoch": 4.37, + "learning_rate": 3.6210406202352898e-06, + "loss": 1.0289, + "step": 59160 + }, + { + "epoch": 4.37, + "learning_rate": 3.6207334452728115e-06, + "loss": 1.0944, + "step": 59161 + }, + { + "epoch": 4.37, + "learning_rate": 3.6204262804596523e-06, + "loss": 0.9287, + "step": 59162 + }, + { + "epoch": 4.37, + "learning_rate": 3.6201191257963098e-06, + "loss": 0.9491, + "step": 59163 + }, + { + "epoch": 4.37, + "learning_rate": 3.61981198128326e-06, + "loss": 1.0619, + "step": 59164 + }, + { + "epoch": 4.37, + "learning_rate": 3.6195048469210016e-06, + "loss": 1.0958, + "step": 59165 + }, + { + "epoch": 4.37, + "learning_rate": 3.6191977227100194e-06, + "loss": 0.8789, + "step": 59166 + }, + { + "epoch": 4.37, + "learning_rate": 3.6188906086508034e-06, + "loss": 0.9867, + "step": 59167 + }, + { + "epoch": 4.37, + "learning_rate": 3.61858350474384e-06, + "loss": 0.8811, + "step": 59168 + }, + { + "epoch": 4.37, + "learning_rate": 3.618276410989616e-06, + "loss": 1.0636, + "step": 59169 + }, + { + "epoch": 4.37, + "learning_rate": 3.6179693273886265e-06, + "loss": 0.9274, + "step": 59170 + }, + { + "epoch": 4.37, + "learning_rate": 3.6176622539413554e-06, + "loss": 1.0122, + "step": 59171 + }, + { + "epoch": 4.37, + "learning_rate": 3.617355190648293e-06, + "loss": 0.9048, + "step": 59172 + }, + { + "epoch": 4.37, + "learning_rate": 3.6170481375099243e-06, + "loss": 0.9849, + "step": 59173 + }, + { + "epoch": 4.37, + "learning_rate": 3.6167410945267433e-06, + "loss": 1.0248, + "step": 59174 + }, + { + "epoch": 4.37, + "learning_rate": 3.6164340616992367e-06, + "loss": 0.9774, + "step": 59175 + }, + { + "epoch": 4.37, + "learning_rate": 3.6161270390278914e-06, + "loss": 0.98, + "step": 59176 + }, + { + "epoch": 4.37, + "learning_rate": 3.615820026513197e-06, + "loss": 0.9767, + "step": 59177 + }, + { + "epoch": 4.37, + "learning_rate": 3.615513024155638e-06, + "loss": 1.0029, + "step": 59178 + }, + { + "epoch": 4.37, + "learning_rate": 3.615206031955709e-06, + "loss": 0.943, + "step": 59179 + }, + { + "epoch": 4.37, + "learning_rate": 3.614899049913897e-06, + "loss": 0.9793, + "step": 59180 + }, + { + "epoch": 4.37, + "learning_rate": 3.6145920780306896e-06, + "loss": 0.9896, + "step": 59181 + }, + { + "epoch": 4.37, + "learning_rate": 3.6142851163065705e-06, + "loss": 1.0655, + "step": 59182 + }, + { + "epoch": 4.37, + "learning_rate": 3.613978164742037e-06, + "loss": 0.8993, + "step": 59183 + }, + { + "epoch": 4.37, + "learning_rate": 3.6136712233375716e-06, + "loss": 0.9096, + "step": 59184 + }, + { + "epoch": 4.37, + "learning_rate": 3.613364292093664e-06, + "loss": 1.0313, + "step": 59185 + }, + { + "epoch": 4.37, + "learning_rate": 3.613057371010803e-06, + "loss": 0.9227, + "step": 59186 + }, + { + "epoch": 4.37, + "learning_rate": 3.6127504600894757e-06, + "loss": 1.0364, + "step": 59187 + }, + { + "epoch": 4.37, + "learning_rate": 3.6124435593301722e-06, + "loss": 1.0746, + "step": 59188 + }, + { + "epoch": 4.37, + "learning_rate": 3.6121366687333757e-06, + "loss": 1.0291, + "step": 59189 + }, + { + "epoch": 4.37, + "learning_rate": 3.611829788299581e-06, + "loss": 0.9924, + "step": 59190 + }, + { + "epoch": 4.37, + "learning_rate": 3.6115229180292743e-06, + "loss": 0.8987, + "step": 59191 + }, + { + "epoch": 4.37, + "learning_rate": 3.611216057922944e-06, + "loss": 1.0186, + "step": 59192 + }, + { + "epoch": 4.37, + "learning_rate": 3.6109092079810726e-06, + "loss": 1.0321, + "step": 59193 + }, + { + "epoch": 4.37, + "learning_rate": 3.610602368204158e-06, + "loss": 1.0829, + "step": 59194 + }, + { + "epoch": 4.37, + "learning_rate": 3.610295538592683e-06, + "loss": 0.9865, + "step": 59195 + }, + { + "epoch": 4.37, + "learning_rate": 3.609988719147136e-06, + "loss": 0.9513, + "step": 59196 + }, + { + "epoch": 4.37, + "learning_rate": 3.6096819098680058e-06, + "loss": 0.9774, + "step": 59197 + }, + { + "epoch": 4.37, + "learning_rate": 3.6093751107557773e-06, + "loss": 0.9291, + "step": 59198 + }, + { + "epoch": 4.37, + "learning_rate": 3.6090683218109447e-06, + "loss": 1.0287, + "step": 59199 + }, + { + "epoch": 4.37, + "learning_rate": 3.6087615430339927e-06, + "loss": 1.085, + "step": 59200 + }, + { + "epoch": 4.37, + "learning_rate": 3.6084547744254095e-06, + "loss": 0.9741, + "step": 59201 + }, + { + "epoch": 4.37, + "learning_rate": 3.608148015985681e-06, + "loss": 1.0933, + "step": 59202 + }, + { + "epoch": 4.37, + "learning_rate": 3.6078412677152997e-06, + "loss": 1.0415, + "step": 59203 + }, + { + "epoch": 4.37, + "learning_rate": 3.6075345296147557e-06, + "loss": 0.8584, + "step": 59204 + }, + { + "epoch": 4.37, + "learning_rate": 3.6072278016845253e-06, + "loss": 1.0376, + "step": 59205 + }, + { + "epoch": 4.37, + "learning_rate": 3.606921083925108e-06, + "loss": 1.0102, + "step": 59206 + }, + { + "epoch": 4.37, + "learning_rate": 3.6066143763369844e-06, + "loss": 1.0236, + "step": 59207 + }, + { + "epoch": 4.37, + "learning_rate": 3.6063076789206518e-06, + "loss": 0.9351, + "step": 59208 + }, + { + "epoch": 4.37, + "learning_rate": 3.606000991676586e-06, + "loss": 1.022, + "step": 59209 + }, + { + "epoch": 4.37, + "learning_rate": 3.6056943146052846e-06, + "loss": 1.0571, + "step": 59210 + }, + { + "epoch": 4.38, + "learning_rate": 3.605387647707228e-06, + "loss": 0.9658, + "step": 59211 + }, + { + "epoch": 4.38, + "learning_rate": 3.6050809909829164e-06, + "loss": 1.0476, + "step": 59212 + }, + { + "epoch": 4.38, + "learning_rate": 3.6047743444328252e-06, + "loss": 1.1145, + "step": 59213 + }, + { + "epoch": 4.38, + "learning_rate": 3.604467708057443e-06, + "loss": 1.1685, + "step": 59214 + }, + { + "epoch": 4.38, + "learning_rate": 3.6041610818572646e-06, + "loss": 1.1495, + "step": 59215 + }, + { + "epoch": 4.38, + "learning_rate": 3.6038544658327746e-06, + "loss": 1.0468, + "step": 59216 + }, + { + "epoch": 4.38, + "learning_rate": 3.603547859984461e-06, + "loss": 1.0263, + "step": 59217 + }, + { + "epoch": 4.38, + "learning_rate": 3.6032412643128066e-06, + "loss": 0.9078, + "step": 59218 + }, + { + "epoch": 4.38, + "learning_rate": 3.6029346788183084e-06, + "loss": 1.0152, + "step": 59219 + }, + { + "epoch": 4.38, + "learning_rate": 3.6026281035014486e-06, + "loss": 1.0137, + "step": 59220 + }, + { + "epoch": 4.38, + "learning_rate": 3.6023215383627175e-06, + "loss": 0.9712, + "step": 59221 + }, + { + "epoch": 4.38, + "learning_rate": 3.6020149834025963e-06, + "loss": 0.9514, + "step": 59222 + }, + { + "epoch": 4.38, + "learning_rate": 3.6017084386215817e-06, + "loss": 1.0743, + "step": 59223 + }, + { + "epoch": 4.38, + "learning_rate": 3.6014019040201573e-06, + "loss": 0.7867, + "step": 59224 + }, + { + "epoch": 4.38, + "learning_rate": 3.601095379598811e-06, + "loss": 1.0709, + "step": 59225 + }, + { + "epoch": 4.38, + "learning_rate": 3.6007888653580316e-06, + "loss": 0.9306, + "step": 59226 + }, + { + "epoch": 4.38, + "learning_rate": 3.600482361298301e-06, + "loss": 0.9887, + "step": 59227 + }, + { + "epoch": 4.38, + "learning_rate": 3.6001758674201183e-06, + "loss": 1.1596, + "step": 59228 + }, + { + "epoch": 4.38, + "learning_rate": 3.5998693837239574e-06, + "loss": 0.9261, + "step": 59229 + }, + { + "epoch": 4.38, + "learning_rate": 3.599562910210316e-06, + "loss": 0.9859, + "step": 59230 + }, + { + "epoch": 4.38, + "learning_rate": 3.5992564468796752e-06, + "loss": 1.0255, + "step": 59231 + }, + { + "epoch": 4.38, + "learning_rate": 3.5989499937325322e-06, + "loss": 1.0898, + "step": 59232 + }, + { + "epoch": 4.38, + "learning_rate": 3.598643550769365e-06, + "loss": 1.0986, + "step": 59233 + }, + { + "epoch": 4.38, + "learning_rate": 3.598337117990661e-06, + "loss": 0.9936, + "step": 59234 + }, + { + "epoch": 4.38, + "learning_rate": 3.5980306953969147e-06, + "loss": 1.0884, + "step": 59235 + }, + { + "epoch": 4.38, + "learning_rate": 3.5977242829886103e-06, + "loss": 0.9932, + "step": 59236 + }, + { + "epoch": 4.38, + "learning_rate": 3.597417880766234e-06, + "loss": 1.0057, + "step": 59237 + }, + { + "epoch": 4.38, + "learning_rate": 3.5971114887302716e-06, + "loss": 0.992, + "step": 59238 + }, + { + "epoch": 4.38, + "learning_rate": 3.596805106881216e-06, + "loss": 0.9667, + "step": 59239 + }, + { + "epoch": 4.38, + "learning_rate": 3.5964987352195526e-06, + "loss": 0.9648, + "step": 59240 + }, + { + "epoch": 4.38, + "learning_rate": 3.596192373745767e-06, + "loss": 0.8681, + "step": 59241 + }, + { + "epoch": 4.38, + "learning_rate": 3.5958860224603485e-06, + "loss": 0.9807, + "step": 59242 + }, + { + "epoch": 4.38, + "learning_rate": 3.595579681363781e-06, + "loss": 0.8914, + "step": 59243 + }, + { + "epoch": 4.38, + "learning_rate": 3.5952733504565573e-06, + "loss": 0.9288, + "step": 59244 + }, + { + "epoch": 4.38, + "learning_rate": 3.5949670297391616e-06, + "loss": 0.966, + "step": 59245 + }, + { + "epoch": 4.38, + "learning_rate": 3.594660719212082e-06, + "loss": 0.8958, + "step": 59246 + }, + { + "epoch": 4.38, + "learning_rate": 3.5943544188758027e-06, + "loss": 1.0022, + "step": 59247 + }, + { + "epoch": 4.38, + "learning_rate": 3.594048128730816e-06, + "loss": 1.013, + "step": 59248 + }, + { + "epoch": 4.38, + "learning_rate": 3.5937418487776086e-06, + "loss": 0.9636, + "step": 59249 + }, + { + "epoch": 4.38, + "learning_rate": 3.593435579016665e-06, + "loss": 1.0578, + "step": 59250 + }, + { + "epoch": 4.38, + "learning_rate": 3.5931293194484753e-06, + "loss": 1.0034, + "step": 59251 + }, + { + "epoch": 4.38, + "learning_rate": 3.59282307007352e-06, + "loss": 1.0006, + "step": 59252 + }, + { + "epoch": 4.38, + "learning_rate": 3.5925168308922986e-06, + "loss": 1.0547, + "step": 59253 + }, + { + "epoch": 4.38, + "learning_rate": 3.592210601905285e-06, + "loss": 1.0691, + "step": 59254 + }, + { + "epoch": 4.38, + "learning_rate": 3.591904383112976e-06, + "loss": 1.0595, + "step": 59255 + }, + { + "epoch": 4.38, + "learning_rate": 3.5915981745158524e-06, + "loss": 0.9612, + "step": 59256 + }, + { + "epoch": 4.38, + "learning_rate": 3.5912919761144104e-06, + "loss": 0.9066, + "step": 59257 + }, + { + "epoch": 4.38, + "learning_rate": 3.590985787909125e-06, + "loss": 0.8378, + "step": 59258 + }, + { + "epoch": 4.38, + "learning_rate": 3.5906796099004924e-06, + "loss": 1.0563, + "step": 59259 + }, + { + "epoch": 4.38, + "learning_rate": 3.590373442088998e-06, + "loss": 1.0197, + "step": 59260 + }, + { + "epoch": 4.38, + "learning_rate": 3.590067284475127e-06, + "loss": 0.9601, + "step": 59261 + }, + { + "epoch": 4.38, + "learning_rate": 3.5897611370593667e-06, + "loss": 1.1353, + "step": 59262 + }, + { + "epoch": 4.38, + "learning_rate": 3.5894549998422023e-06, + "loss": 1.0557, + "step": 59263 + }, + { + "epoch": 4.38, + "learning_rate": 3.5891488728241253e-06, + "loss": 1.076, + "step": 59264 + }, + { + "epoch": 4.38, + "learning_rate": 3.5888427560056217e-06, + "loss": 0.9075, + "step": 59265 + }, + { + "epoch": 4.38, + "learning_rate": 3.588536649387178e-06, + "loss": 1.0236, + "step": 59266 + }, + { + "epoch": 4.38, + "learning_rate": 3.5882305529692764e-06, + "loss": 1.0201, + "step": 59267 + }, + { + "epoch": 4.38, + "learning_rate": 3.5879244667524118e-06, + "loss": 0.9259, + "step": 59268 + }, + { + "epoch": 4.38, + "learning_rate": 3.5876183907370675e-06, + "loss": 1.0515, + "step": 59269 + }, + { + "epoch": 4.38, + "learning_rate": 3.587312324923732e-06, + "loss": 1.0752, + "step": 59270 + }, + { + "epoch": 4.38, + "learning_rate": 3.587006269312889e-06, + "loss": 1.0225, + "step": 59271 + }, + { + "epoch": 4.38, + "learning_rate": 3.5867002239050254e-06, + "loss": 0.9553, + "step": 59272 + }, + { + "epoch": 4.38, + "learning_rate": 3.5863941887006358e-06, + "loss": 1.0355, + "step": 59273 + }, + { + "epoch": 4.38, + "learning_rate": 3.5860881637001945e-06, + "loss": 0.995, + "step": 59274 + }, + { + "epoch": 4.38, + "learning_rate": 3.585782148904199e-06, + "loss": 0.9839, + "step": 59275 + }, + { + "epoch": 4.38, + "learning_rate": 3.5854761443131283e-06, + "loss": 0.9806, + "step": 59276 + }, + { + "epoch": 4.38, + "learning_rate": 3.5851701499274804e-06, + "loss": 1.188, + "step": 59277 + }, + { + "epoch": 4.38, + "learning_rate": 3.584864165747729e-06, + "loss": 0.873, + "step": 59278 + }, + { + "epoch": 4.38, + "learning_rate": 3.584558191774369e-06, + "loss": 0.9874, + "step": 59279 + }, + { + "epoch": 4.38, + "learning_rate": 3.5842522280078864e-06, + "loss": 1.0411, + "step": 59280 + }, + { + "epoch": 4.38, + "learning_rate": 3.5839462744487653e-06, + "loss": 1.0254, + "step": 59281 + }, + { + "epoch": 4.38, + "learning_rate": 3.5836403310974954e-06, + "loss": 0.9804, + "step": 59282 + }, + { + "epoch": 4.38, + "learning_rate": 3.5833343979545576e-06, + "loss": 0.8994, + "step": 59283 + }, + { + "epoch": 4.38, + "learning_rate": 3.583028475020446e-06, + "loss": 0.9921, + "step": 59284 + }, + { + "epoch": 4.38, + "learning_rate": 3.5827225622956453e-06, + "loss": 1.0098, + "step": 59285 + }, + { + "epoch": 4.38, + "learning_rate": 3.582416659780641e-06, + "loss": 1.0114, + "step": 59286 + }, + { + "epoch": 4.38, + "learning_rate": 3.582110767475916e-06, + "loss": 0.9888, + "step": 59287 + }, + { + "epoch": 4.38, + "learning_rate": 3.5818048853819644e-06, + "loss": 0.9266, + "step": 59288 + }, + { + "epoch": 4.38, + "learning_rate": 3.58149901349927e-06, + "loss": 0.9241, + "step": 59289 + }, + { + "epoch": 4.38, + "learning_rate": 3.581193151828318e-06, + "loss": 0.9616, + "step": 59290 + }, + { + "epoch": 4.38, + "learning_rate": 3.5808873003695965e-06, + "loss": 1.0985, + "step": 59291 + }, + { + "epoch": 4.38, + "learning_rate": 3.580581459123588e-06, + "loss": 1.0246, + "step": 59292 + }, + { + "epoch": 4.38, + "learning_rate": 3.5802756280907858e-06, + "loss": 1.0244, + "step": 59293 + }, + { + "epoch": 4.38, + "learning_rate": 3.5799698072716726e-06, + "loss": 0.9194, + "step": 59294 + }, + { + "epoch": 4.38, + "learning_rate": 3.579663996666737e-06, + "loss": 1.0465, + "step": 59295 + }, + { + "epoch": 4.38, + "learning_rate": 3.57935819627646e-06, + "loss": 1.036, + "step": 59296 + }, + { + "epoch": 4.38, + "learning_rate": 3.5790524061013353e-06, + "loss": 0.8872, + "step": 59297 + }, + { + "epoch": 4.38, + "learning_rate": 3.57874662614185e-06, + "loss": 0.9286, + "step": 59298 + }, + { + "epoch": 4.38, + "learning_rate": 3.5784408563984806e-06, + "loss": 1.0025, + "step": 59299 + }, + { + "epoch": 4.38, + "learning_rate": 3.578135096871722e-06, + "loss": 0.9755, + "step": 59300 + }, + { + "epoch": 4.38, + "learning_rate": 3.577829347562056e-06, + "loss": 1.1683, + "step": 59301 + }, + { + "epoch": 4.38, + "learning_rate": 3.5775236084699793e-06, + "loss": 1.0673, + "step": 59302 + }, + { + "epoch": 4.38, + "learning_rate": 3.5772178795959635e-06, + "loss": 1.0269, + "step": 59303 + }, + { + "epoch": 4.38, + "learning_rate": 3.5769121609405044e-06, + "loss": 1.0149, + "step": 59304 + }, + { + "epoch": 4.38, + "learning_rate": 3.5766064525040866e-06, + "loss": 1.1183, + "step": 59305 + }, + { + "epoch": 4.38, + "learning_rate": 3.576300754287196e-06, + "loss": 0.9336, + "step": 59306 + }, + { + "epoch": 4.38, + "learning_rate": 3.57599506629032e-06, + "loss": 0.9296, + "step": 59307 + }, + { + "epoch": 4.38, + "learning_rate": 3.5756893885139397e-06, + "loss": 0.9927, + "step": 59308 + }, + { + "epoch": 4.38, + "learning_rate": 3.5753837209585494e-06, + "loss": 1.091, + "step": 59309 + }, + { + "epoch": 4.38, + "learning_rate": 3.5750780636246318e-06, + "loss": 1.0112, + "step": 59310 + }, + { + "epoch": 4.38, + "learning_rate": 3.574772416512673e-06, + "loss": 1.0561, + "step": 59311 + }, + { + "epoch": 4.38, + "learning_rate": 3.574466779623156e-06, + "loss": 0.9959, + "step": 59312 + }, + { + "epoch": 4.38, + "learning_rate": 3.5741611529565722e-06, + "loss": 0.8365, + "step": 59313 + }, + { + "epoch": 4.38, + "learning_rate": 3.5738555365134077e-06, + "loss": 0.928, + "step": 59314 + }, + { + "epoch": 4.38, + "learning_rate": 3.573549930294147e-06, + "loss": 1.0121, + "step": 59315 + }, + { + "epoch": 4.38, + "learning_rate": 3.5732443342992763e-06, + "loss": 0.9888, + "step": 59316 + }, + { + "epoch": 4.38, + "learning_rate": 3.5729387485292788e-06, + "loss": 0.9872, + "step": 59317 + }, + { + "epoch": 4.38, + "learning_rate": 3.57263317298465e-06, + "loss": 0.9381, + "step": 59318 + }, + { + "epoch": 4.38, + "learning_rate": 3.572327607665863e-06, + "loss": 0.9553, + "step": 59319 + }, + { + "epoch": 4.38, + "learning_rate": 3.572022052573415e-06, + "loss": 1.0422, + "step": 59320 + }, + { + "epoch": 4.38, + "learning_rate": 3.5717165077077844e-06, + "loss": 1.0561, + "step": 59321 + }, + { + "epoch": 4.38, + "learning_rate": 3.571410973069467e-06, + "loss": 1.0857, + "step": 59322 + }, + { + "epoch": 4.38, + "learning_rate": 3.571105448658936e-06, + "loss": 1.0055, + "step": 59323 + }, + { + "epoch": 4.38, + "learning_rate": 3.5707999344766874e-06, + "loss": 1.0407, + "step": 59324 + }, + { + "epoch": 4.38, + "learning_rate": 3.570494430523205e-06, + "loss": 0.9733, + "step": 59325 + }, + { + "epoch": 4.38, + "learning_rate": 3.570188936798974e-06, + "loss": 1.0232, + "step": 59326 + }, + { + "epoch": 4.38, + "learning_rate": 3.5698834533044803e-06, + "loss": 1.0058, + "step": 59327 + }, + { + "epoch": 4.38, + "learning_rate": 3.569577980040205e-06, + "loss": 0.9618, + "step": 59328 + }, + { + "epoch": 4.38, + "learning_rate": 3.5692725170066446e-06, + "loss": 0.9676, + "step": 59329 + }, + { + "epoch": 4.38, + "learning_rate": 3.5689670642042785e-06, + "loss": 0.9869, + "step": 59330 + }, + { + "epoch": 4.38, + "learning_rate": 3.5686616216335944e-06, + "loss": 0.9898, + "step": 59331 + }, + { + "epoch": 4.38, + "learning_rate": 3.5683561892950747e-06, + "loss": 0.9185, + "step": 59332 + }, + { + "epoch": 4.38, + "learning_rate": 3.5680507671892107e-06, + "loss": 0.9823, + "step": 59333 + }, + { + "epoch": 4.38, + "learning_rate": 3.5677453553164863e-06, + "loss": 0.9265, + "step": 59334 + }, + { + "epoch": 4.38, + "learning_rate": 3.567439953677386e-06, + "loss": 0.9846, + "step": 59335 + }, + { + "epoch": 4.38, + "learning_rate": 3.5671345622723985e-06, + "loss": 0.9121, + "step": 59336 + }, + { + "epoch": 4.38, + "learning_rate": 3.566829181102003e-06, + "loss": 0.9888, + "step": 59337 + }, + { + "epoch": 4.38, + "learning_rate": 3.5665238101666942e-06, + "loss": 0.9429, + "step": 59338 + }, + { + "epoch": 4.38, + "learning_rate": 3.5662184494669538e-06, + "loss": 1.0575, + "step": 59339 + }, + { + "epoch": 4.38, + "learning_rate": 3.5659130990032685e-06, + "loss": 1.0226, + "step": 59340 + }, + { + "epoch": 4.38, + "learning_rate": 3.5656077587761195e-06, + "loss": 0.9274, + "step": 59341 + }, + { + "epoch": 4.38, + "learning_rate": 3.5653024287859995e-06, + "loss": 0.9487, + "step": 59342 + }, + { + "epoch": 4.38, + "learning_rate": 3.5649971090333923e-06, + "loss": 0.9636, + "step": 59343 + }, + { + "epoch": 4.38, + "learning_rate": 3.5646917995187814e-06, + "loss": 0.798, + "step": 59344 + }, + { + "epoch": 4.38, + "learning_rate": 3.564386500242655e-06, + "loss": 0.9986, + "step": 59345 + }, + { + "epoch": 4.39, + "learning_rate": 3.5640812112054936e-06, + "loss": 0.9845, + "step": 59346 + }, + { + "epoch": 4.39, + "learning_rate": 3.5637759324077934e-06, + "loss": 0.9961, + "step": 59347 + }, + { + "epoch": 4.39, + "learning_rate": 3.563470663850027e-06, + "loss": 1.0733, + "step": 59348 + }, + { + "epoch": 4.39, + "learning_rate": 3.5631654055326903e-06, + "loss": 1.1459, + "step": 59349 + }, + { + "epoch": 4.39, + "learning_rate": 3.562860157456265e-06, + "loss": 1.0034, + "step": 59350 + }, + { + "epoch": 4.39, + "learning_rate": 3.5625549196212373e-06, + "loss": 0.9392, + "step": 59351 + }, + { + "epoch": 4.39, + "learning_rate": 3.562249692028089e-06, + "loss": 1.0678, + "step": 59352 + }, + { + "epoch": 4.39, + "learning_rate": 3.5619444746773124e-06, + "loss": 0.9361, + "step": 59353 + }, + { + "epoch": 4.39, + "learning_rate": 3.561639267569391e-06, + "loss": 0.9595, + "step": 59354 + }, + { + "epoch": 4.39, + "learning_rate": 3.5613340707048096e-06, + "loss": 1.0426, + "step": 59355 + }, + { + "epoch": 4.39, + "learning_rate": 3.5610288840840524e-06, + "loss": 0.9401, + "step": 59356 + }, + { + "epoch": 4.39, + "learning_rate": 3.560723707707603e-06, + "loss": 1.044, + "step": 59357 + }, + { + "epoch": 4.39, + "learning_rate": 3.560418541575954e-06, + "loss": 0.9827, + "step": 59358 + }, + { + "epoch": 4.39, + "learning_rate": 3.5601133856895862e-06, + "loss": 1.0944, + "step": 59359 + }, + { + "epoch": 4.39, + "learning_rate": 3.5598082400489865e-06, + "loss": 1.0149, + "step": 59360 + }, + { + "epoch": 4.39, + "learning_rate": 3.559503104654636e-06, + "loss": 0.9929, + "step": 59361 + }, + { + "epoch": 4.39, + "learning_rate": 3.5591979795070277e-06, + "loss": 1.0949, + "step": 59362 + }, + { + "epoch": 4.39, + "learning_rate": 3.5588928646066466e-06, + "loss": 1.1042, + "step": 59363 + }, + { + "epoch": 4.39, + "learning_rate": 3.5585877599539678e-06, + "loss": 1.1302, + "step": 59364 + }, + { + "epoch": 4.39, + "learning_rate": 3.558282665549487e-06, + "loss": 1.0294, + "step": 59365 + }, + { + "epoch": 4.39, + "learning_rate": 3.5579775813936835e-06, + "loss": 0.9483, + "step": 59366 + }, + { + "epoch": 4.39, + "learning_rate": 3.557672507487051e-06, + "loss": 0.9907, + "step": 59367 + }, + { + "epoch": 4.39, + "learning_rate": 3.557367443830064e-06, + "loss": 1.0387, + "step": 59368 + }, + { + "epoch": 4.39, + "learning_rate": 3.557062390423216e-06, + "loss": 0.9808, + "step": 59369 + }, + { + "epoch": 4.39, + "learning_rate": 3.556757347266986e-06, + "loss": 1.0251, + "step": 59370 + }, + { + "epoch": 4.39, + "learning_rate": 3.55645231436187e-06, + "loss": 1.0273, + "step": 59371 + }, + { + "epoch": 4.39, + "learning_rate": 3.556147291708343e-06, + "loss": 0.9997, + "step": 59372 + }, + { + "epoch": 4.39, + "learning_rate": 3.555842279306889e-06, + "loss": 0.9801, + "step": 59373 + }, + { + "epoch": 4.39, + "learning_rate": 3.5555372771580022e-06, + "loss": 0.9299, + "step": 59374 + }, + { + "epoch": 4.39, + "learning_rate": 3.5552322852621635e-06, + "loss": 0.9377, + "step": 59375 + }, + { + "epoch": 4.39, + "learning_rate": 3.554927303619857e-06, + "loss": 0.9677, + "step": 59376 + }, + { + "epoch": 4.39, + "learning_rate": 3.5546223322315665e-06, + "loss": 1.1089, + "step": 59377 + }, + { + "epoch": 4.39, + "learning_rate": 3.5543173710977828e-06, + "loss": 1.1175, + "step": 59378 + }, + { + "epoch": 4.39, + "learning_rate": 3.5540124202189876e-06, + "loss": 1.0168, + "step": 59379 + }, + { + "epoch": 4.39, + "learning_rate": 3.5537074795956663e-06, + "loss": 1.0382, + "step": 59380 + }, + { + "epoch": 4.39, + "learning_rate": 3.553402549228304e-06, + "loss": 1.1061, + "step": 59381 + }, + { + "epoch": 4.39, + "learning_rate": 3.5530976291173836e-06, + "loss": 0.9387, + "step": 59382 + }, + { + "epoch": 4.39, + "learning_rate": 3.552792719263396e-06, + "loss": 0.878, + "step": 59383 + }, + { + "epoch": 4.39, + "learning_rate": 3.5524878196668223e-06, + "loss": 0.9577, + "step": 59384 + }, + { + "epoch": 4.39, + "learning_rate": 3.5521829303281484e-06, + "loss": 0.9694, + "step": 59385 + }, + { + "epoch": 4.39, + "learning_rate": 3.5518780512478557e-06, + "loss": 0.9462, + "step": 59386 + }, + { + "epoch": 4.39, + "learning_rate": 3.5515731824264365e-06, + "loss": 1.0758, + "step": 59387 + }, + { + "epoch": 4.39, + "learning_rate": 3.551268323864372e-06, + "loss": 0.9431, + "step": 59388 + }, + { + "epoch": 4.39, + "learning_rate": 3.550963475562147e-06, + "loss": 1.0145, + "step": 59389 + }, + { + "epoch": 4.39, + "learning_rate": 3.5506586375202433e-06, + "loss": 0.8855, + "step": 59390 + }, + { + "epoch": 4.39, + "learning_rate": 3.550353809739153e-06, + "loss": 0.995, + "step": 59391 + }, + { + "epoch": 4.39, + "learning_rate": 3.5500489922193614e-06, + "loss": 1.0154, + "step": 59392 + }, + { + "epoch": 4.39, + "learning_rate": 3.5497441849613424e-06, + "loss": 1.0291, + "step": 59393 + }, + { + "epoch": 4.39, + "learning_rate": 3.549439387965592e-06, + "loss": 1.0264, + "step": 59394 + }, + { + "epoch": 4.39, + "learning_rate": 3.54913460123259e-06, + "loss": 0.9809, + "step": 59395 + }, + { + "epoch": 4.39, + "learning_rate": 3.548829824762824e-06, + "loss": 1.0317, + "step": 59396 + }, + { + "epoch": 4.39, + "learning_rate": 3.548525058556773e-06, + "loss": 0.9601, + "step": 59397 + }, + { + "epoch": 4.39, + "learning_rate": 3.5482203026149297e-06, + "loss": 0.9661, + "step": 59398 + }, + { + "epoch": 4.39, + "learning_rate": 3.5479155569377765e-06, + "loss": 1.0151, + "step": 59399 + }, + { + "epoch": 4.39, + "learning_rate": 3.5476108215257964e-06, + "loss": 0.8939, + "step": 59400 + }, + { + "epoch": 4.39, + "learning_rate": 3.547306096379476e-06, + "loss": 0.9621, + "step": 59401 + }, + { + "epoch": 4.39, + "learning_rate": 3.5470013814992953e-06, + "loss": 1.004, + "step": 59402 + }, + { + "epoch": 4.39, + "learning_rate": 3.546696676885747e-06, + "loss": 1.0302, + "step": 59403 + }, + { + "epoch": 4.39, + "learning_rate": 3.5463919825393114e-06, + "loss": 1.0216, + "step": 59404 + }, + { + "epoch": 4.39, + "learning_rate": 3.5460872984604743e-06, + "loss": 1.059, + "step": 59405 + }, + { + "epoch": 4.39, + "learning_rate": 3.5457826246497163e-06, + "loss": 0.9988, + "step": 59406 + }, + { + "epoch": 4.39, + "learning_rate": 3.5454779611075285e-06, + "loss": 1.0009, + "step": 59407 + }, + { + "epoch": 4.39, + "learning_rate": 3.545173307834394e-06, + "loss": 0.9763, + "step": 59408 + }, + { + "epoch": 4.39, + "learning_rate": 3.544868664830796e-06, + "loss": 0.9467, + "step": 59409 + }, + { + "epoch": 4.39, + "learning_rate": 3.5445640320972198e-06, + "loss": 0.9461, + "step": 59410 + }, + { + "epoch": 4.39, + "learning_rate": 3.5442594096341466e-06, + "loss": 1.0555, + "step": 59411 + }, + { + "epoch": 4.39, + "learning_rate": 3.543954797442071e-06, + "loss": 0.9439, + "step": 59412 + }, + { + "epoch": 4.39, + "learning_rate": 3.543650195521463e-06, + "loss": 1.0597, + "step": 59413 + }, + { + "epoch": 4.39, + "learning_rate": 3.54334560387282e-06, + "loss": 0.9727, + "step": 59414 + }, + { + "epoch": 4.39, + "learning_rate": 3.543041022496618e-06, + "loss": 0.9523, + "step": 59415 + }, + { + "epoch": 4.39, + "learning_rate": 3.5427364513933517e-06, + "loss": 0.9535, + "step": 59416 + }, + { + "epoch": 4.39, + "learning_rate": 3.5424318905634938e-06, + "loss": 0.9927, + "step": 59417 + }, + { + "epoch": 4.39, + "learning_rate": 3.542127340007536e-06, + "loss": 0.9666, + "step": 59418 + }, + { + "epoch": 4.39, + "learning_rate": 3.5418227997259626e-06, + "loss": 1.1355, + "step": 59419 + }, + { + "epoch": 4.39, + "learning_rate": 3.541518269719255e-06, + "loss": 1.0176, + "step": 59420 + }, + { + "epoch": 4.39, + "learning_rate": 3.5412137499879008e-06, + "loss": 0.9258, + "step": 59421 + }, + { + "epoch": 4.39, + "learning_rate": 3.5409092405323796e-06, + "loss": 1.0082, + "step": 59422 + }, + { + "epoch": 4.39, + "learning_rate": 3.540604741353183e-06, + "loss": 0.9935, + "step": 59423 + }, + { + "epoch": 4.39, + "learning_rate": 3.5403002524507912e-06, + "loss": 1.0136, + "step": 59424 + }, + { + "epoch": 4.39, + "learning_rate": 3.5399957738256896e-06, + "loss": 0.9966, + "step": 59425 + }, + { + "epoch": 4.39, + "learning_rate": 3.5396913054783588e-06, + "loss": 1.0346, + "step": 59426 + }, + { + "epoch": 4.39, + "learning_rate": 3.5393868474092907e-06, + "loss": 0.9755, + "step": 59427 + }, + { + "epoch": 4.39, + "learning_rate": 3.539082399618965e-06, + "loss": 1.0046, + "step": 59428 + }, + { + "epoch": 4.39, + "learning_rate": 3.5387779621078665e-06, + "loss": 0.9005, + "step": 59429 + }, + { + "epoch": 4.39, + "learning_rate": 3.538473534876481e-06, + "loss": 0.9961, + "step": 59430 + }, + { + "epoch": 4.39, + "learning_rate": 3.5381691179252873e-06, + "loss": 0.9695, + "step": 59431 + }, + { + "epoch": 4.39, + "learning_rate": 3.537864711254777e-06, + "loss": 1.0347, + "step": 59432 + }, + { + "epoch": 4.39, + "learning_rate": 3.5375603148654327e-06, + "loss": 0.9865, + "step": 59433 + }, + { + "epoch": 4.39, + "learning_rate": 3.537255928757737e-06, + "loss": 1.0592, + "step": 59434 + }, + { + "epoch": 4.39, + "learning_rate": 3.5369515529321705e-06, + "loss": 1.1124, + "step": 59435 + }, + { + "epoch": 4.39, + "learning_rate": 3.536647187389226e-06, + "loss": 0.9978, + "step": 59436 + }, + { + "epoch": 4.39, + "learning_rate": 3.5363428321293837e-06, + "loss": 1.0242, + "step": 59437 + }, + { + "epoch": 4.39, + "learning_rate": 3.5360384871531263e-06, + "loss": 0.9429, + "step": 59438 + }, + { + "epoch": 4.39, + "learning_rate": 3.53573415246094e-06, + "loss": 0.9492, + "step": 59439 + }, + { + "epoch": 4.39, + "learning_rate": 3.5354298280533083e-06, + "loss": 1.0131, + "step": 59440 + }, + { + "epoch": 4.39, + "learning_rate": 3.535125513930715e-06, + "loss": 0.9376, + "step": 59441 + }, + { + "epoch": 4.39, + "learning_rate": 3.534821210093641e-06, + "loss": 0.9194, + "step": 59442 + }, + { + "epoch": 4.39, + "learning_rate": 3.5345169165425764e-06, + "loss": 0.9061, + "step": 59443 + }, + { + "epoch": 4.39, + "learning_rate": 3.5342126332780034e-06, + "loss": 1.0513, + "step": 59444 + }, + { + "epoch": 4.39, + "learning_rate": 3.5339083603004066e-06, + "loss": 1.0442, + "step": 59445 + }, + { + "epoch": 4.39, + "learning_rate": 3.5336040976102647e-06, + "loss": 1.0061, + "step": 59446 + }, + { + "epoch": 4.39, + "learning_rate": 3.53329984520807e-06, + "loss": 1.0976, + "step": 59447 + }, + { + "epoch": 4.39, + "learning_rate": 3.5329956030943037e-06, + "loss": 1.0176, + "step": 59448 + }, + { + "epoch": 4.39, + "learning_rate": 3.5326913712694465e-06, + "loss": 1.063, + "step": 59449 + }, + { + "epoch": 4.39, + "learning_rate": 3.5323871497339868e-06, + "loss": 0.9079, + "step": 59450 + }, + { + "epoch": 4.39, + "learning_rate": 3.5320829384884028e-06, + "loss": 0.9618, + "step": 59451 + }, + { + "epoch": 4.39, + "learning_rate": 3.531778737533185e-06, + "loss": 0.9877, + "step": 59452 + }, + { + "epoch": 4.39, + "learning_rate": 3.531474546868815e-06, + "loss": 0.9658, + "step": 59453 + }, + { + "epoch": 4.39, + "learning_rate": 3.5311703664957765e-06, + "loss": 1.0858, + "step": 59454 + }, + { + "epoch": 4.39, + "learning_rate": 3.5308661964145508e-06, + "loss": 1.0319, + "step": 59455 + }, + { + "epoch": 4.39, + "learning_rate": 3.5305620366256266e-06, + "loss": 0.9205, + "step": 59456 + }, + { + "epoch": 4.39, + "learning_rate": 3.530257887129489e-06, + "loss": 0.9853, + "step": 59457 + }, + { + "epoch": 4.39, + "learning_rate": 3.5299537479266122e-06, + "loss": 1.0655, + "step": 59458 + }, + { + "epoch": 4.39, + "learning_rate": 3.52964961901749e-06, + "loss": 1.1169, + "step": 59459 + }, + { + "epoch": 4.39, + "learning_rate": 3.5293455004026e-06, + "loss": 0.972, + "step": 59460 + }, + { + "epoch": 4.39, + "learning_rate": 3.529041392082434e-06, + "loss": 0.9578, + "step": 59461 + }, + { + "epoch": 4.39, + "learning_rate": 3.528737294057465e-06, + "loss": 0.9149, + "step": 59462 + }, + { + "epoch": 4.39, + "learning_rate": 3.5284332063281845e-06, + "loss": 1.0142, + "step": 59463 + }, + { + "epoch": 4.39, + "learning_rate": 3.5281291288950715e-06, + "loss": 0.9474, + "step": 59464 + }, + { + "epoch": 4.39, + "learning_rate": 3.5278250617586197e-06, + "loss": 0.9717, + "step": 59465 + }, + { + "epoch": 4.39, + "learning_rate": 3.527521004919302e-06, + "loss": 1.0407, + "step": 59466 + }, + { + "epoch": 4.39, + "learning_rate": 3.5272169583776015e-06, + "loss": 1.0946, + "step": 59467 + }, + { + "epoch": 4.39, + "learning_rate": 3.5269129221340105e-06, + "loss": 0.9817, + "step": 59468 + }, + { + "epoch": 4.39, + "learning_rate": 3.5266088961890077e-06, + "loss": 0.9555, + "step": 59469 + }, + { + "epoch": 4.39, + "learning_rate": 3.526304880543079e-06, + "loss": 0.9732, + "step": 59470 + }, + { + "epoch": 4.39, + "learning_rate": 3.526000875196701e-06, + "loss": 0.9504, + "step": 59471 + }, + { + "epoch": 4.39, + "learning_rate": 3.5256968801503675e-06, + "loss": 0.935, + "step": 59472 + }, + { + "epoch": 4.39, + "learning_rate": 3.525392895404558e-06, + "loss": 1.0585, + "step": 59473 + }, + { + "epoch": 4.39, + "learning_rate": 3.525088920959756e-06, + "loss": 0.9598, + "step": 59474 + }, + { + "epoch": 4.39, + "learning_rate": 3.524784956816444e-06, + "loss": 1.0178, + "step": 59475 + }, + { + "epoch": 4.39, + "learning_rate": 3.524481002975104e-06, + "loss": 0.9861, + "step": 59476 + }, + { + "epoch": 4.39, + "learning_rate": 3.5241770594362245e-06, + "loss": 1.0932, + "step": 59477 + }, + { + "epoch": 4.39, + "learning_rate": 3.5238731262002877e-06, + "loss": 0.9623, + "step": 59478 + }, + { + "epoch": 4.39, + "learning_rate": 3.523569203267775e-06, + "loss": 1.01, + "step": 59479 + }, + { + "epoch": 4.39, + "learning_rate": 3.5232652906391684e-06, + "loss": 0.9877, + "step": 59480 + }, + { + "epoch": 4.4, + "learning_rate": 3.522961388314957e-06, + "loss": 1.0375, + "step": 59481 + }, + { + "epoch": 4.4, + "learning_rate": 3.522657496295622e-06, + "loss": 0.9964, + "step": 59482 + }, + { + "epoch": 4.4, + "learning_rate": 3.522353614581645e-06, + "loss": 0.9778, + "step": 59483 + }, + { + "epoch": 4.4, + "learning_rate": 3.522049743173512e-06, + "loss": 1.0328, + "step": 59484 + }, + { + "epoch": 4.4, + "learning_rate": 3.521745882071704e-06, + "loss": 0.8714, + "step": 59485 + }, + { + "epoch": 4.4, + "learning_rate": 3.5214420312767074e-06, + "loss": 0.9242, + "step": 59486 + }, + { + "epoch": 4.4, + "learning_rate": 3.5211381907889984e-06, + "loss": 0.9654, + "step": 59487 + }, + { + "epoch": 4.4, + "learning_rate": 3.52083436060907e-06, + "loss": 0.9237, + "step": 59488 + }, + { + "epoch": 4.4, + "learning_rate": 3.5205305407374026e-06, + "loss": 1.0047, + "step": 59489 + }, + { + "epoch": 4.4, + "learning_rate": 3.5202267311744764e-06, + "loss": 0.9493, + "step": 59490 + }, + { + "epoch": 4.4, + "learning_rate": 3.519922931920775e-06, + "loss": 1.1298, + "step": 59491 + }, + { + "epoch": 4.4, + "learning_rate": 3.5196191429767856e-06, + "loss": 0.934, + "step": 59492 + }, + { + "epoch": 4.4, + "learning_rate": 3.51931536434299e-06, + "loss": 0.9796, + "step": 59493 + }, + { + "epoch": 4.4, + "learning_rate": 3.519011596019871e-06, + "loss": 0.9358, + "step": 59494 + }, + { + "epoch": 4.4, + "learning_rate": 3.518707838007912e-06, + "loss": 1.0241, + "step": 59495 + }, + { + "epoch": 4.4, + "learning_rate": 3.5184040903075935e-06, + "loss": 0.9429, + "step": 59496 + }, + { + "epoch": 4.4, + "learning_rate": 3.5181003529194037e-06, + "loss": 0.9325, + "step": 59497 + }, + { + "epoch": 4.4, + "learning_rate": 3.5177966258438235e-06, + "loss": 0.979, + "step": 59498 + }, + { + "epoch": 4.4, + "learning_rate": 3.5174929090813357e-06, + "loss": 1.0366, + "step": 59499 + }, + { + "epoch": 4.4, + "learning_rate": 3.517189202632422e-06, + "loss": 1.0822, + "step": 59500 + }, + { + "epoch": 4.4, + "learning_rate": 3.5168855064975695e-06, + "loss": 0.9715, + "step": 59501 + }, + { + "epoch": 4.4, + "learning_rate": 3.5165818206772596e-06, + "loss": 0.9967, + "step": 59502 + }, + { + "epoch": 4.4, + "learning_rate": 3.516278145171975e-06, + "loss": 0.9099, + "step": 59503 + }, + { + "epoch": 4.4, + "learning_rate": 3.5159744799822005e-06, + "loss": 1.0406, + "step": 59504 + }, + { + "epoch": 4.4, + "learning_rate": 3.515670825108414e-06, + "loss": 0.9537, + "step": 59505 + }, + { + "epoch": 4.4, + "learning_rate": 3.5153671805511082e-06, + "loss": 1.0411, + "step": 59506 + }, + { + "epoch": 4.4, + "learning_rate": 3.515063546310754e-06, + "loss": 0.9419, + "step": 59507 + }, + { + "epoch": 4.4, + "learning_rate": 3.514759922387845e-06, + "loss": 1.0198, + "step": 59508 + }, + { + "epoch": 4.4, + "learning_rate": 3.5144563087828565e-06, + "loss": 0.9966, + "step": 59509 + }, + { + "epoch": 4.4, + "learning_rate": 3.514152705496281e-06, + "loss": 0.9667, + "step": 59510 + }, + { + "epoch": 4.4, + "learning_rate": 3.51384911252859e-06, + "loss": 1.0291, + "step": 59511 + }, + { + "epoch": 4.4, + "learning_rate": 3.513545529880276e-06, + "loss": 1.0436, + "step": 59512 + }, + { + "epoch": 4.4, + "learning_rate": 3.513241957551817e-06, + "loss": 0.9227, + "step": 59513 + }, + { + "epoch": 4.4, + "learning_rate": 3.512938395543698e-06, + "loss": 1.0373, + "step": 59514 + }, + { + "epoch": 4.4, + "learning_rate": 3.512634843856402e-06, + "loss": 0.8772, + "step": 59515 + }, + { + "epoch": 4.4, + "learning_rate": 3.512331302490406e-06, + "loss": 0.9856, + "step": 59516 + }, + { + "epoch": 4.4, + "learning_rate": 3.5120277714462026e-06, + "loss": 0.9768, + "step": 59517 + }, + { + "epoch": 4.4, + "learning_rate": 3.5117242507242698e-06, + "loss": 0.9225, + "step": 59518 + }, + { + "epoch": 4.4, + "learning_rate": 3.511420740325091e-06, + "loss": 1.0802, + "step": 59519 + }, + { + "epoch": 4.4, + "learning_rate": 3.5111172402491454e-06, + "loss": 0.9913, + "step": 59520 + }, + { + "epoch": 4.4, + "learning_rate": 3.510813750496923e-06, + "loss": 0.9789, + "step": 59521 + }, + { + "epoch": 4.4, + "learning_rate": 3.5105102710689034e-06, + "loss": 1.0398, + "step": 59522 + }, + { + "epoch": 4.4, + "learning_rate": 3.5102068019655687e-06, + "loss": 0.9549, + "step": 59523 + }, + { + "epoch": 4.4, + "learning_rate": 3.509903343187403e-06, + "loss": 1.05, + "step": 59524 + }, + { + "epoch": 4.4, + "learning_rate": 3.509599894734884e-06, + "loss": 1.0214, + "step": 59525 + }, + { + "epoch": 4.4, + "learning_rate": 3.509296456608502e-06, + "loss": 0.9285, + "step": 59526 + }, + { + "epoch": 4.4, + "learning_rate": 3.5089930288087372e-06, + "loss": 0.8823, + "step": 59527 + }, + { + "epoch": 4.4, + "learning_rate": 3.50868961133607e-06, + "loss": 0.9506, + "step": 59528 + }, + { + "epoch": 4.4, + "learning_rate": 3.508386204190983e-06, + "loss": 0.9461, + "step": 59529 + }, + { + "epoch": 4.4, + "learning_rate": 3.508082807373967e-06, + "loss": 0.9332, + "step": 59530 + }, + { + "epoch": 4.4, + "learning_rate": 3.507779420885494e-06, + "loss": 0.9663, + "step": 59531 + }, + { + "epoch": 4.4, + "learning_rate": 3.5074760447260493e-06, + "loss": 1.0655, + "step": 59532 + }, + { + "epoch": 4.4, + "learning_rate": 3.50717267889612e-06, + "loss": 1.1249, + "step": 59533 + }, + { + "epoch": 4.4, + "learning_rate": 3.5068693233961857e-06, + "loss": 0.9009, + "step": 59534 + }, + { + "epoch": 4.4, + "learning_rate": 3.5065659782267305e-06, + "loss": 0.9408, + "step": 59535 + }, + { + "epoch": 4.4, + "learning_rate": 3.5062626433882307e-06, + "loss": 0.9766, + "step": 59536 + }, + { + "epoch": 4.4, + "learning_rate": 3.505959318881179e-06, + "loss": 0.9729, + "step": 59537 + }, + { + "epoch": 4.4, + "learning_rate": 3.505656004706053e-06, + "loss": 0.9437, + "step": 59538 + }, + { + "epoch": 4.4, + "learning_rate": 3.505352700863335e-06, + "loss": 1.0114, + "step": 59539 + }, + { + "epoch": 4.4, + "learning_rate": 3.505049407353508e-06, + "loss": 0.9819, + "step": 59540 + }, + { + "epoch": 4.4, + "learning_rate": 3.5047461241770508e-06, + "loss": 0.9496, + "step": 59541 + }, + { + "epoch": 4.4, + "learning_rate": 3.504442851334453e-06, + "loss": 0.9595, + "step": 59542 + }, + { + "epoch": 4.4, + "learning_rate": 3.5041395888261943e-06, + "loss": 0.9804, + "step": 59543 + }, + { + "epoch": 4.4, + "learning_rate": 3.503836336652756e-06, + "loss": 0.9893, + "step": 59544 + }, + { + "epoch": 4.4, + "learning_rate": 3.5035330948146174e-06, + "loss": 1.1047, + "step": 59545 + }, + { + "epoch": 4.4, + "learning_rate": 3.5032298633122686e-06, + "loss": 0.9838, + "step": 59546 + }, + { + "epoch": 4.4, + "learning_rate": 3.5029266421461884e-06, + "loss": 0.9289, + "step": 59547 + }, + { + "epoch": 4.4, + "learning_rate": 3.5026234313168595e-06, + "loss": 1.0673, + "step": 59548 + }, + { + "epoch": 4.4, + "learning_rate": 3.5023202308247596e-06, + "loss": 0.9994, + "step": 59549 + }, + { + "epoch": 4.4, + "learning_rate": 3.502017040670378e-06, + "loss": 1.0445, + "step": 59550 + }, + { + "epoch": 4.4, + "learning_rate": 3.5017138608541977e-06, + "loss": 1.0031, + "step": 59551 + }, + { + "epoch": 4.4, + "learning_rate": 3.501410691376692e-06, + "loss": 1.0629, + "step": 59552 + }, + { + "epoch": 4.4, + "learning_rate": 3.5011075322383513e-06, + "loss": 0.9456, + "step": 59553 + }, + { + "epoch": 4.4, + "learning_rate": 3.5008043834396532e-06, + "loss": 1.0132, + "step": 59554 + }, + { + "epoch": 4.4, + "learning_rate": 3.5005012449810882e-06, + "loss": 1.0002, + "step": 59555 + }, + { + "epoch": 4.4, + "learning_rate": 3.5001981168631273e-06, + "loss": 1.0447, + "step": 59556 + }, + { + "epoch": 4.4, + "learning_rate": 3.4998949990862596e-06, + "loss": 1.0034, + "step": 59557 + }, + { + "epoch": 4.4, + "learning_rate": 3.499591891650964e-06, + "loss": 1.0793, + "step": 59558 + }, + { + "epoch": 4.4, + "learning_rate": 3.4992887945577313e-06, + "loss": 0.9877, + "step": 59559 + }, + { + "epoch": 4.4, + "learning_rate": 3.4989857078070333e-06, + "loss": 1.0487, + "step": 59560 + }, + { + "epoch": 4.4, + "learning_rate": 3.4986826313993526e-06, + "loss": 1.0658, + "step": 59561 + }, + { + "epoch": 4.4, + "learning_rate": 3.4983795653351783e-06, + "loss": 0.9656, + "step": 59562 + }, + { + "epoch": 4.4, + "learning_rate": 3.4980765096149894e-06, + "loss": 0.9349, + "step": 59563 + }, + { + "epoch": 4.4, + "learning_rate": 3.497773464239268e-06, + "loss": 1.0278, + "step": 59564 + }, + { + "epoch": 4.4, + "learning_rate": 3.4974704292084914e-06, + "loss": 1.0883, + "step": 59565 + }, + { + "epoch": 4.4, + "learning_rate": 3.497167404523151e-06, + "loss": 1.0085, + "step": 59566 + }, + { + "epoch": 4.4, + "learning_rate": 3.496864390183724e-06, + "loss": 1.156, + "step": 59567 + }, + { + "epoch": 4.4, + "learning_rate": 3.496561386190692e-06, + "loss": 1.0498, + "step": 59568 + }, + { + "epoch": 4.4, + "learning_rate": 3.496258392544538e-06, + "loss": 0.974, + "step": 59569 + }, + { + "epoch": 4.4, + "learning_rate": 3.4959554092457403e-06, + "loss": 1.0157, + "step": 59570 + }, + { + "epoch": 4.4, + "learning_rate": 3.4956524362947917e-06, + "loss": 1.0231, + "step": 59571 + }, + { + "epoch": 4.4, + "learning_rate": 3.49534947369216e-06, + "loss": 1.0398, + "step": 59572 + }, + { + "epoch": 4.4, + "learning_rate": 3.495046521438338e-06, + "loss": 1.1169, + "step": 59573 + }, + { + "epoch": 4.4, + "learning_rate": 3.4947435795338e-06, + "loss": 0.9121, + "step": 59574 + }, + { + "epoch": 4.4, + "learning_rate": 3.4944406479790393e-06, + "loss": 1.03, + "step": 59575 + }, + { + "epoch": 4.4, + "learning_rate": 3.4941377267745235e-06, + "loss": 1.0442, + "step": 59576 + }, + { + "epoch": 4.4, + "learning_rate": 3.493834815920745e-06, + "loss": 1.0988, + "step": 59577 + }, + { + "epoch": 4.4, + "learning_rate": 3.4935319154181814e-06, + "loss": 1.0285, + "step": 59578 + }, + { + "epoch": 4.4, + "learning_rate": 3.4932290252673164e-06, + "loss": 0.8803, + "step": 59579 + }, + { + "epoch": 4.4, + "learning_rate": 3.4929261454686302e-06, + "loss": 1.0176, + "step": 59580 + }, + { + "epoch": 4.4, + "learning_rate": 3.4926232760226018e-06, + "loss": 0.9088, + "step": 59581 + }, + { + "epoch": 4.4, + "learning_rate": 3.492320416929721e-06, + "loss": 0.9809, + "step": 59582 + }, + { + "epoch": 4.4, + "learning_rate": 3.4920175681904646e-06, + "loss": 1.0269, + "step": 59583 + }, + { + "epoch": 4.4, + "learning_rate": 3.4917147298053156e-06, + "loss": 0.9754, + "step": 59584 + }, + { + "epoch": 4.4, + "learning_rate": 3.491411901774752e-06, + "loss": 0.913, + "step": 59585 + }, + { + "epoch": 4.4, + "learning_rate": 3.491109084099261e-06, + "loss": 0.8672, + "step": 59586 + }, + { + "epoch": 4.4, + "learning_rate": 3.4908062767793237e-06, + "loss": 0.9985, + "step": 59587 + }, + { + "epoch": 4.4, + "learning_rate": 3.4905034798154203e-06, + "loss": 0.9885, + "step": 59588 + }, + { + "epoch": 4.4, + "learning_rate": 3.4902006932080325e-06, + "loss": 0.9843, + "step": 59589 + }, + { + "epoch": 4.4, + "learning_rate": 3.4898979169576397e-06, + "loss": 0.9718, + "step": 59590 + }, + { + "epoch": 4.4, + "learning_rate": 3.4895951510647285e-06, + "loss": 0.9465, + "step": 59591 + }, + { + "epoch": 4.4, + "learning_rate": 3.4892923955297787e-06, + "loss": 0.9253, + "step": 59592 + }, + { + "epoch": 4.4, + "learning_rate": 3.4889896503532715e-06, + "loss": 1.0365, + "step": 59593 + }, + { + "epoch": 4.4, + "learning_rate": 3.488686915535685e-06, + "loss": 0.9987, + "step": 59594 + }, + { + "epoch": 4.4, + "learning_rate": 3.488384191077507e-06, + "loss": 0.9864, + "step": 59595 + }, + { + "epoch": 4.4, + "learning_rate": 3.4880814769792216e-06, + "loss": 0.9692, + "step": 59596 + }, + { + "epoch": 4.4, + "learning_rate": 3.4877787732412984e-06, + "loss": 1.0944, + "step": 59597 + }, + { + "epoch": 4.4, + "learning_rate": 3.4874760798642284e-06, + "loss": 0.9931, + "step": 59598 + }, + { + "epoch": 4.4, + "learning_rate": 3.4871733968484876e-06, + "loss": 1.0469, + "step": 59599 + }, + { + "epoch": 4.4, + "learning_rate": 3.4868707241945687e-06, + "loss": 1.0144, + "step": 59600 + }, + { + "epoch": 4.4, + "learning_rate": 3.4865680619029374e-06, + "loss": 0.9365, + "step": 59601 + }, + { + "epoch": 4.4, + "learning_rate": 3.486265409974087e-06, + "loss": 1.0721, + "step": 59602 + }, + { + "epoch": 4.4, + "learning_rate": 3.4859627684084916e-06, + "loss": 0.9352, + "step": 59603 + }, + { + "epoch": 4.4, + "learning_rate": 3.4856601372066422e-06, + "loss": 0.9807, + "step": 59604 + }, + { + "epoch": 4.4, + "learning_rate": 3.485357516369009e-06, + "loss": 1.1314, + "step": 59605 + }, + { + "epoch": 4.4, + "learning_rate": 3.4850549058960813e-06, + "loss": 0.9535, + "step": 59606 + }, + { + "epoch": 4.4, + "learning_rate": 3.484752305788338e-06, + "loss": 1.052, + "step": 59607 + }, + { + "epoch": 4.4, + "learning_rate": 3.484449716046261e-06, + "loss": 0.9558, + "step": 59608 + }, + { + "epoch": 4.4, + "learning_rate": 3.4841471366703306e-06, + "loss": 0.9555, + "step": 59609 + }, + { + "epoch": 4.4, + "learning_rate": 3.4838445676610266e-06, + "loss": 0.9738, + "step": 59610 + }, + { + "epoch": 4.4, + "learning_rate": 3.4835420090188355e-06, + "loss": 1.0056, + "step": 59611 + }, + { + "epoch": 4.4, + "learning_rate": 3.4832394607442353e-06, + "loss": 1.0156, + "step": 59612 + }, + { + "epoch": 4.4, + "learning_rate": 3.4829369228377084e-06, + "loss": 0.998, + "step": 59613 + }, + { + "epoch": 4.4, + "learning_rate": 3.482634395299732e-06, + "loss": 0.9779, + "step": 59614 + }, + { + "epoch": 4.4, + "learning_rate": 3.4823318781307945e-06, + "loss": 0.9925, + "step": 59615 + }, + { + "epoch": 4.4, + "learning_rate": 3.482029371331377e-06, + "loss": 1.0929, + "step": 59616 + }, + { + "epoch": 4.41, + "learning_rate": 3.4817268749019515e-06, + "loss": 1.0094, + "step": 59617 + }, + { + "epoch": 4.41, + "learning_rate": 3.4814243888430076e-06, + "loss": 0.8685, + "step": 59618 + }, + { + "epoch": 4.41, + "learning_rate": 3.481121913155021e-06, + "loss": 1.0197, + "step": 59619 + }, + { + "epoch": 4.41, + "learning_rate": 3.480819447838484e-06, + "loss": 1.0722, + "step": 59620 + }, + { + "epoch": 4.41, + "learning_rate": 3.480516992893862e-06, + "loss": 0.9591, + "step": 59621 + }, + { + "epoch": 4.41, + "learning_rate": 3.480214548321649e-06, + "loss": 1.0539, + "step": 59622 + }, + { + "epoch": 4.41, + "learning_rate": 3.479912114122316e-06, + "loss": 1.159, + "step": 59623 + }, + { + "epoch": 4.41, + "learning_rate": 3.4796096902963572e-06, + "loss": 0.9235, + "step": 59624 + }, + { + "epoch": 4.41, + "learning_rate": 3.479307276844243e-06, + "loss": 1.0553, + "step": 59625 + }, + { + "epoch": 4.41, + "learning_rate": 3.4790048737664537e-06, + "loss": 0.9619, + "step": 59626 + }, + { + "epoch": 4.41, + "learning_rate": 3.4787024810634786e-06, + "loss": 1.064, + "step": 59627 + }, + { + "epoch": 4.41, + "learning_rate": 3.4784000987357935e-06, + "loss": 1.0889, + "step": 59628 + }, + { + "epoch": 4.41, + "learning_rate": 3.478097726783881e-06, + "loss": 1.0409, + "step": 59629 + }, + { + "epoch": 4.41, + "learning_rate": 3.4777953652082185e-06, + "loss": 1.0195, + "step": 59630 + }, + { + "epoch": 4.41, + "learning_rate": 3.4774930140092945e-06, + "loss": 0.9908, + "step": 59631 + }, + { + "epoch": 4.41, + "learning_rate": 3.477190673187585e-06, + "loss": 1.1305, + "step": 59632 + }, + { + "epoch": 4.41, + "learning_rate": 3.476888342743571e-06, + "loss": 1.1059, + "step": 59633 + }, + { + "epoch": 4.41, + "learning_rate": 3.476586022677735e-06, + "loss": 1.0537, + "step": 59634 + }, + { + "epoch": 4.41, + "learning_rate": 3.4762837129905536e-06, + "loss": 1.021, + "step": 59635 + }, + { + "epoch": 4.41, + "learning_rate": 3.4759814136825155e-06, + "loss": 1.0087, + "step": 59636 + }, + { + "epoch": 4.41, + "learning_rate": 3.475679124754098e-06, + "loss": 0.9468, + "step": 59637 + }, + { + "epoch": 4.41, + "learning_rate": 3.47537684620578e-06, + "loss": 0.9017, + "step": 59638 + }, + { + "epoch": 4.41, + "learning_rate": 3.4750745780380414e-06, + "loss": 1.024, + "step": 59639 + }, + { + "epoch": 4.41, + "learning_rate": 3.4747723202513694e-06, + "loss": 1.0212, + "step": 59640 + }, + { + "epoch": 4.41, + "learning_rate": 3.474470072846242e-06, + "loss": 0.9831, + "step": 59641 + }, + { + "epoch": 4.41, + "learning_rate": 3.4741678358231377e-06, + "loss": 1.0781, + "step": 59642 + }, + { + "epoch": 4.41, + "learning_rate": 3.4738656091825407e-06, + "loss": 0.9807, + "step": 59643 + }, + { + "epoch": 4.41, + "learning_rate": 3.4735633929249255e-06, + "loss": 0.9164, + "step": 59644 + }, + { + "epoch": 4.41, + "learning_rate": 3.473261187050785e-06, + "loss": 0.9786, + "step": 59645 + }, + { + "epoch": 4.41, + "learning_rate": 3.4729589915605855e-06, + "loss": 1.0772, + "step": 59646 + }, + { + "epoch": 4.41, + "learning_rate": 3.472656806454818e-06, + "loss": 0.9367, + "step": 59647 + }, + { + "epoch": 4.41, + "learning_rate": 3.4723546317339573e-06, + "loss": 1.0074, + "step": 59648 + }, + { + "epoch": 4.41, + "learning_rate": 3.4720524673984945e-06, + "loss": 0.9651, + "step": 59649 + }, + { + "epoch": 4.41, + "learning_rate": 3.471750313448895e-06, + "loss": 0.9788, + "step": 59650 + }, + { + "epoch": 4.41, + "learning_rate": 3.4714481698856505e-06, + "loss": 0.9197, + "step": 59651 + }, + { + "epoch": 4.41, + "learning_rate": 3.4711460367092387e-06, + "loss": 0.9917, + "step": 59652 + }, + { + "epoch": 4.41, + "learning_rate": 3.470843913920141e-06, + "loss": 0.9662, + "step": 59653 + }, + { + "epoch": 4.41, + "learning_rate": 3.4705418015188364e-06, + "loss": 0.8823, + "step": 59654 + }, + { + "epoch": 4.41, + "learning_rate": 3.470239699505803e-06, + "loss": 0.9963, + "step": 59655 + }, + { + "epoch": 4.41, + "learning_rate": 3.4699376078815285e-06, + "loss": 1.0448, + "step": 59656 + }, + { + "epoch": 4.41, + "learning_rate": 3.46963552664649e-06, + "loss": 0.9482, + "step": 59657 + }, + { + "epoch": 4.41, + "learning_rate": 3.4693334558011673e-06, + "loss": 0.9153, + "step": 59658 + }, + { + "epoch": 4.41, + "learning_rate": 3.4690313953460386e-06, + "loss": 1.0037, + "step": 59659 + }, + { + "epoch": 4.41, + "learning_rate": 3.4687293452815906e-06, + "loss": 0.9737, + "step": 59660 + }, + { + "epoch": 4.41, + "learning_rate": 3.4684273056083005e-06, + "loss": 0.9562, + "step": 59661 + }, + { + "epoch": 4.41, + "learning_rate": 3.4681252763266503e-06, + "loss": 0.9575, + "step": 59662 + }, + { + "epoch": 4.41, + "learning_rate": 3.4678232574371186e-06, + "loss": 1.0064, + "step": 59663 + }, + { + "epoch": 4.41, + "learning_rate": 3.467521248940183e-06, + "loss": 0.9827, + "step": 59664 + }, + { + "epoch": 4.41, + "learning_rate": 3.4672192508363355e-06, + "loss": 0.8937, + "step": 59665 + }, + { + "epoch": 4.41, + "learning_rate": 3.4669172631260405e-06, + "loss": 0.9867, + "step": 59666 + }, + { + "epoch": 4.41, + "learning_rate": 3.4666152858097913e-06, + "loss": 1.0115, + "step": 59667 + }, + { + "epoch": 4.41, + "learning_rate": 3.4663133188880594e-06, + "loss": 1.0691, + "step": 59668 + }, + { + "epoch": 4.41, + "learning_rate": 3.466011362361337e-06, + "loss": 1.0462, + "step": 59669 + }, + { + "epoch": 4.41, + "learning_rate": 3.4657094162300897e-06, + "loss": 1.0133, + "step": 59670 + }, + { + "epoch": 4.41, + "learning_rate": 3.465407480494809e-06, + "loss": 1.051, + "step": 59671 + }, + { + "epoch": 4.41, + "learning_rate": 3.465105555155971e-06, + "loss": 1.0286, + "step": 59672 + }, + { + "epoch": 4.41, + "learning_rate": 3.4648036402140574e-06, + "loss": 1.0398, + "step": 59673 + }, + { + "epoch": 4.41, + "learning_rate": 3.4645017356695466e-06, + "loss": 0.9697, + "step": 59674 + }, + { + "epoch": 4.41, + "learning_rate": 3.464199841522917e-06, + "loss": 1.081, + "step": 59675 + }, + { + "epoch": 4.41, + "learning_rate": 3.4638979577746558e-06, + "loss": 1.0285, + "step": 59676 + }, + { + "epoch": 4.41, + "learning_rate": 3.4635960844252393e-06, + "loss": 1.0176, + "step": 59677 + }, + { + "epoch": 4.41, + "learning_rate": 3.4632942214751474e-06, + "loss": 0.9329, + "step": 59678 + }, + { + "epoch": 4.41, + "learning_rate": 3.462992368924858e-06, + "loss": 0.8861, + "step": 59679 + }, + { + "epoch": 4.41, + "learning_rate": 3.4626905267748577e-06, + "loss": 1.0715, + "step": 59680 + }, + { + "epoch": 4.41, + "learning_rate": 3.4623886950256226e-06, + "loss": 0.9863, + "step": 59681 + }, + { + "epoch": 4.41, + "learning_rate": 3.462086873677635e-06, + "loss": 0.8948, + "step": 59682 + }, + { + "epoch": 4.41, + "learning_rate": 3.461785062731372e-06, + "loss": 0.8925, + "step": 59683 + }, + { + "epoch": 4.41, + "learning_rate": 3.4614832621873127e-06, + "loss": 0.9853, + "step": 59684 + }, + { + "epoch": 4.41, + "learning_rate": 3.4611814720459436e-06, + "loss": 1.1194, + "step": 59685 + }, + { + "epoch": 4.41, + "learning_rate": 3.4608796923077404e-06, + "loss": 0.9513, + "step": 59686 + }, + { + "epoch": 4.41, + "learning_rate": 3.460577922973185e-06, + "loss": 1.0395, + "step": 59687 + }, + { + "epoch": 4.41, + "learning_rate": 3.460276164042752e-06, + "loss": 0.9351, + "step": 59688 + }, + { + "epoch": 4.41, + "learning_rate": 3.45997441551693e-06, + "loss": 0.9623, + "step": 59689 + }, + { + "epoch": 4.41, + "learning_rate": 3.4596726773961996e-06, + "loss": 1.003, + "step": 59690 + }, + { + "epoch": 4.41, + "learning_rate": 3.459370949681028e-06, + "loss": 0.9977, + "step": 59691 + }, + { + "epoch": 4.41, + "learning_rate": 3.459069232371908e-06, + "loss": 0.9701, + "step": 59692 + }, + { + "epoch": 4.41, + "learning_rate": 3.458767525469315e-06, + "loss": 0.9171, + "step": 59693 + }, + { + "epoch": 4.41, + "learning_rate": 3.4584658289737293e-06, + "loss": 1.0639, + "step": 59694 + }, + { + "epoch": 4.41, + "learning_rate": 3.458164142885627e-06, + "loss": 1.0713, + "step": 59695 + }, + { + "epoch": 4.41, + "learning_rate": 3.457862467205496e-06, + "loss": 1.0772, + "step": 59696 + }, + { + "epoch": 4.41, + "learning_rate": 3.4575608019338125e-06, + "loss": 1.0242, + "step": 59697 + }, + { + "epoch": 4.41, + "learning_rate": 3.457259147071056e-06, + "loss": 1.0613, + "step": 59698 + }, + { + "epoch": 4.41, + "learning_rate": 3.456957502617706e-06, + "loss": 0.9171, + "step": 59699 + }, + { + "epoch": 4.41, + "learning_rate": 3.4566558685742404e-06, + "loss": 0.943, + "step": 59700 + }, + { + "epoch": 4.41, + "learning_rate": 3.4563542449411435e-06, + "loss": 0.9866, + "step": 59701 + }, + { + "epoch": 4.41, + "learning_rate": 3.4560526317188947e-06, + "loss": 0.972, + "step": 59702 + }, + { + "epoch": 4.41, + "learning_rate": 3.455751028907972e-06, + "loss": 1.0212, + "step": 59703 + }, + { + "epoch": 4.41, + "learning_rate": 3.455449436508853e-06, + "loss": 0.883, + "step": 59704 + }, + { + "epoch": 4.41, + "learning_rate": 3.455147854522023e-06, + "loss": 0.9757, + "step": 59705 + }, + { + "epoch": 4.41, + "learning_rate": 3.4548462829479602e-06, + "loss": 0.9901, + "step": 59706 + }, + { + "epoch": 4.41, + "learning_rate": 3.454544721787142e-06, + "loss": 1.1029, + "step": 59707 + }, + { + "epoch": 4.41, + "learning_rate": 3.454243171040047e-06, + "loss": 0.95, + "step": 59708 + }, + { + "epoch": 4.41, + "learning_rate": 3.4539416307071604e-06, + "loss": 1.0232, + "step": 59709 + }, + { + "epoch": 4.41, + "learning_rate": 3.453640100788963e-06, + "loss": 0.9156, + "step": 59710 + }, + { + "epoch": 4.41, + "learning_rate": 3.4533385812859233e-06, + "loss": 1.0112, + "step": 59711 + }, + { + "epoch": 4.41, + "learning_rate": 3.453037072198532e-06, + "loss": 0.9777, + "step": 59712 + }, + { + "epoch": 4.41, + "learning_rate": 3.4527355735272604e-06, + "loss": 1.0918, + "step": 59713 + }, + { + "epoch": 4.41, + "learning_rate": 3.452434085272601e-06, + "loss": 1.0952, + "step": 59714 + }, + { + "epoch": 4.41, + "learning_rate": 3.452132607435018e-06, + "loss": 1.0307, + "step": 59715 + }, + { + "epoch": 4.41, + "learning_rate": 3.451831140015002e-06, + "loss": 0.9722, + "step": 59716 + }, + { + "epoch": 4.41, + "learning_rate": 3.4515296830130254e-06, + "loss": 1.108, + "step": 59717 + }, + { + "epoch": 4.41, + "learning_rate": 3.451228236429578e-06, + "loss": 0.975, + "step": 59718 + }, + { + "epoch": 4.41, + "learning_rate": 3.4509268002651286e-06, + "loss": 1.0091, + "step": 59719 + }, + { + "epoch": 4.41, + "learning_rate": 3.450625374520159e-06, + "loss": 1.0041, + "step": 59720 + }, + { + "epoch": 4.41, + "learning_rate": 3.450323959195153e-06, + "loss": 1.0334, + "step": 59721 + }, + { + "epoch": 4.41, + "learning_rate": 3.4500225542905885e-06, + "loss": 0.9922, + "step": 59722 + }, + { + "epoch": 4.41, + "learning_rate": 3.4497211598069445e-06, + "loss": 1.0623, + "step": 59723 + }, + { + "epoch": 4.41, + "learning_rate": 3.4494197757446968e-06, + "loss": 1.0262, + "step": 59724 + }, + { + "epoch": 4.41, + "learning_rate": 3.449118402104332e-06, + "loss": 0.9827, + "step": 59725 + }, + { + "epoch": 4.41, + "learning_rate": 3.448817038886326e-06, + "loss": 1.0328, + "step": 59726 + }, + { + "epoch": 4.41, + "learning_rate": 3.4485156860911586e-06, + "loss": 0.9603, + "step": 59727 + }, + { + "epoch": 4.41, + "learning_rate": 3.4482143437193083e-06, + "loss": 1.0375, + "step": 59728 + }, + { + "epoch": 4.41, + "learning_rate": 3.4479130117712535e-06, + "loss": 0.8916, + "step": 59729 + }, + { + "epoch": 4.41, + "learning_rate": 3.4476116902474764e-06, + "loss": 0.9275, + "step": 59730 + }, + { + "epoch": 4.41, + "learning_rate": 3.4473103791484574e-06, + "loss": 1.0487, + "step": 59731 + }, + { + "epoch": 4.41, + "learning_rate": 3.4470090784746725e-06, + "loss": 1.043, + "step": 59732 + }, + { + "epoch": 4.41, + "learning_rate": 3.4467077882266e-06, + "loss": 1.102, + "step": 59733 + }, + { + "epoch": 4.41, + "learning_rate": 3.446406508404724e-06, + "loss": 0.9753, + "step": 59734 + }, + { + "epoch": 4.41, + "learning_rate": 3.4461052390095227e-06, + "loss": 1.056, + "step": 59735 + }, + { + "epoch": 4.41, + "learning_rate": 3.4458039800414734e-06, + "loss": 0.8909, + "step": 59736 + }, + { + "epoch": 4.41, + "learning_rate": 3.4455027315010557e-06, + "loss": 1.0262, + "step": 59737 + }, + { + "epoch": 4.41, + "learning_rate": 3.4452014933887503e-06, + "loss": 1.0144, + "step": 59738 + }, + { + "epoch": 4.41, + "learning_rate": 3.4449002657050358e-06, + "loss": 1.0111, + "step": 59739 + }, + { + "epoch": 4.41, + "learning_rate": 3.4445990484503864e-06, + "loss": 1.0155, + "step": 59740 + }, + { + "epoch": 4.41, + "learning_rate": 3.44429784162529e-06, + "loss": 1.0031, + "step": 59741 + }, + { + "epoch": 4.41, + "learning_rate": 3.443996645230223e-06, + "loss": 1.0521, + "step": 59742 + }, + { + "epoch": 4.41, + "learning_rate": 3.4436954592656625e-06, + "loss": 1.0567, + "step": 59743 + }, + { + "epoch": 4.41, + "learning_rate": 3.4433942837320855e-06, + "loss": 0.9259, + "step": 59744 + }, + { + "epoch": 4.41, + "learning_rate": 3.4430931186299786e-06, + "loss": 0.9682, + "step": 59745 + }, + { + "epoch": 4.41, + "learning_rate": 3.4427919639598163e-06, + "loss": 0.9824, + "step": 59746 + }, + { + "epoch": 4.41, + "learning_rate": 3.4424908197220773e-06, + "loss": 0.9441, + "step": 59747 + }, + { + "epoch": 4.41, + "learning_rate": 3.4421896859172422e-06, + "loss": 1.0548, + "step": 59748 + }, + { + "epoch": 4.41, + "learning_rate": 3.4418885625457864e-06, + "loss": 0.9557, + "step": 59749 + }, + { + "epoch": 4.41, + "learning_rate": 3.4415874496081945e-06, + "loss": 0.9538, + "step": 59750 + }, + { + "epoch": 4.41, + "learning_rate": 3.441286347104944e-06, + "loss": 1.0149, + "step": 59751 + }, + { + "epoch": 4.42, + "learning_rate": 3.440985255036513e-06, + "loss": 0.9363, + "step": 59752 + }, + { + "epoch": 4.42, + "learning_rate": 3.440684173403377e-06, + "loss": 1.0324, + "step": 59753 + }, + { + "epoch": 4.42, + "learning_rate": 3.4403831022060217e-06, + "loss": 0.9873, + "step": 59754 + }, + { + "epoch": 4.42, + "learning_rate": 3.4400820414449276e-06, + "loss": 0.9486, + "step": 59755 + }, + { + "epoch": 4.42, + "learning_rate": 3.4397809911205614e-06, + "loss": 1.0126, + "step": 59756 + }, + { + "epoch": 4.42, + "learning_rate": 3.439479951233413e-06, + "loss": 1.046, + "step": 59757 + }, + { + "epoch": 4.42, + "learning_rate": 3.4391789217839554e-06, + "loss": 1.0919, + "step": 59758 + }, + { + "epoch": 4.42, + "learning_rate": 3.438877902772676e-06, + "loss": 1.1665, + "step": 59759 + }, + { + "epoch": 4.42, + "learning_rate": 3.438576894200042e-06, + "loss": 1.07, + "step": 59760 + }, + { + "epoch": 4.42, + "learning_rate": 3.438275896066542e-06, + "loss": 0.8875, + "step": 59761 + }, + { + "epoch": 4.42, + "learning_rate": 3.437974908372648e-06, + "loss": 0.8706, + "step": 59762 + }, + { + "epoch": 4.42, + "learning_rate": 3.4376739311188477e-06, + "loss": 1.1039, + "step": 59763 + }, + { + "epoch": 4.42, + "learning_rate": 3.4373729643056075e-06, + "loss": 0.9854, + "step": 59764 + }, + { + "epoch": 4.42, + "learning_rate": 3.4370720079334175e-06, + "loss": 1.0529, + "step": 59765 + }, + { + "epoch": 4.42, + "learning_rate": 3.4367710620027517e-06, + "loss": 1.0876, + "step": 59766 + }, + { + "epoch": 4.42, + "learning_rate": 3.4364701265140887e-06, + "loss": 1.1901, + "step": 59767 + }, + { + "epoch": 4.42, + "learning_rate": 3.4361692014679084e-06, + "loss": 0.9769, + "step": 59768 + }, + { + "epoch": 4.42, + "learning_rate": 3.435868286864684e-06, + "loss": 0.994, + "step": 59769 + }, + { + "epoch": 4.42, + "learning_rate": 3.4355673827049053e-06, + "loss": 0.9928, + "step": 59770 + }, + { + "epoch": 4.42, + "learning_rate": 3.435266488989043e-06, + "loss": 0.9525, + "step": 59771 + }, + { + "epoch": 4.42, + "learning_rate": 3.4349656057175786e-06, + "loss": 0.928, + "step": 59772 + }, + { + "epoch": 4.42, + "learning_rate": 3.4346647328909866e-06, + "loss": 0.9397, + "step": 59773 + }, + { + "epoch": 4.42, + "learning_rate": 3.4343638705097513e-06, + "loss": 1.0477, + "step": 59774 + }, + { + "epoch": 4.42, + "learning_rate": 3.4340630185743508e-06, + "loss": 0.9696, + "step": 59775 + }, + { + "epoch": 4.42, + "learning_rate": 3.4337621770852615e-06, + "loss": 0.9276, + "step": 59776 + }, + { + "epoch": 4.42, + "learning_rate": 3.4334613460429624e-06, + "loss": 1.0365, + "step": 59777 + }, + { + "epoch": 4.42, + "learning_rate": 3.4331605254479282e-06, + "loss": 0.955, + "step": 59778 + }, + { + "epoch": 4.42, + "learning_rate": 3.432859715300646e-06, + "loss": 0.9415, + "step": 59779 + }, + { + "epoch": 4.42, + "learning_rate": 3.4325589156015904e-06, + "loss": 1.0244, + "step": 59780 + }, + { + "epoch": 4.42, + "learning_rate": 3.432258126351239e-06, + "loss": 0.9105, + "step": 59781 + }, + { + "epoch": 4.42, + "learning_rate": 3.4319573475500677e-06, + "loss": 0.9855, + "step": 59782 + }, + { + "epoch": 4.42, + "learning_rate": 3.4316565791985655e-06, + "loss": 0.9319, + "step": 59783 + }, + { + "epoch": 4.42, + "learning_rate": 3.4313558212971997e-06, + "loss": 0.9719, + "step": 59784 + }, + { + "epoch": 4.42, + "learning_rate": 3.43105507384645e-06, + "loss": 1.0182, + "step": 59785 + }, + { + "epoch": 4.42, + "learning_rate": 3.4307543368468e-06, + "loss": 0.9302, + "step": 59786 + }, + { + "epoch": 4.42, + "learning_rate": 3.430453610298726e-06, + "loss": 0.9511, + "step": 59787 + }, + { + "epoch": 4.42, + "learning_rate": 3.4301528942027075e-06, + "loss": 0.9669, + "step": 59788 + }, + { + "epoch": 4.42, + "learning_rate": 3.4298521885592173e-06, + "loss": 1.0232, + "step": 59789 + }, + { + "epoch": 4.42, + "learning_rate": 3.4295514933687425e-06, + "loss": 0.9767, + "step": 59790 + }, + { + "epoch": 4.42, + "learning_rate": 3.4292508086317568e-06, + "loss": 1.0594, + "step": 59791 + }, + { + "epoch": 4.42, + "learning_rate": 3.4289501343487386e-06, + "loss": 1.065, + "step": 59792 + }, + { + "epoch": 4.42, + "learning_rate": 3.4286494705201678e-06, + "loss": 1.0428, + "step": 59793 + }, + { + "epoch": 4.42, + "learning_rate": 3.4283488171465174e-06, + "loss": 0.9426, + "step": 59794 + }, + { + "epoch": 4.42, + "learning_rate": 3.4280481742282734e-06, + "loss": 0.9686, + "step": 59795 + }, + { + "epoch": 4.42, + "learning_rate": 3.427747541765911e-06, + "loss": 1.072, + "step": 59796 + }, + { + "epoch": 4.42, + "learning_rate": 3.4274469197599093e-06, + "loss": 1.0461, + "step": 59797 + }, + { + "epoch": 4.42, + "learning_rate": 3.42714630821074e-06, + "loss": 1.0717, + "step": 59798 + }, + { + "epoch": 4.42, + "learning_rate": 3.426845707118892e-06, + "loss": 0.9713, + "step": 59799 + }, + { + "epoch": 4.42, + "learning_rate": 3.4265451164848373e-06, + "loss": 1.0018, + "step": 59800 + }, + { + "epoch": 4.42, + "learning_rate": 3.4262445363090557e-06, + "loss": 1.0388, + "step": 59801 + }, + { + "epoch": 4.42, + "learning_rate": 3.4259439665920258e-06, + "loss": 1.0397, + "step": 59802 + }, + { + "epoch": 4.42, + "learning_rate": 3.4256434073342204e-06, + "loss": 1.0573, + "step": 59803 + }, + { + "epoch": 4.42, + "learning_rate": 3.4253428585361294e-06, + "loss": 1.0388, + "step": 59804 + }, + { + "epoch": 4.42, + "learning_rate": 3.4250423201982176e-06, + "loss": 0.9554, + "step": 59805 + }, + { + "epoch": 4.42, + "learning_rate": 3.4247417923209713e-06, + "loss": 0.9724, + "step": 59806 + }, + { + "epoch": 4.42, + "learning_rate": 3.424441274904865e-06, + "loss": 1.0595, + "step": 59807 + }, + { + "epoch": 4.42, + "learning_rate": 3.4241407679503847e-06, + "loss": 0.9337, + "step": 59808 + }, + { + "epoch": 4.42, + "learning_rate": 3.4238402714579956e-06, + "loss": 1.0645, + "step": 59809 + }, + { + "epoch": 4.42, + "learning_rate": 3.423539785428186e-06, + "loss": 1.0615, + "step": 59810 + }, + { + "epoch": 4.42, + "learning_rate": 3.4232393098614303e-06, + "loss": 1.119, + "step": 59811 + }, + { + "epoch": 4.42, + "learning_rate": 3.422938844758207e-06, + "loss": 1.029, + "step": 59812 + }, + { + "epoch": 4.42, + "learning_rate": 3.422638390118993e-06, + "loss": 1.0945, + "step": 59813 + }, + { + "epoch": 4.42, + "learning_rate": 3.4223379459442653e-06, + "loss": 1.0413, + "step": 59814 + }, + { + "epoch": 4.42, + "learning_rate": 3.4220375122345063e-06, + "loss": 1.0354, + "step": 59815 + }, + { + "epoch": 4.42, + "learning_rate": 3.4217370889901913e-06, + "loss": 1.1306, + "step": 59816 + }, + { + "epoch": 4.42, + "learning_rate": 3.421436676211799e-06, + "loss": 0.9098, + "step": 59817 + }, + { + "epoch": 4.42, + "learning_rate": 3.4211362738998034e-06, + "loss": 1.0411, + "step": 59818 + }, + { + "epoch": 4.42, + "learning_rate": 3.4208358820546894e-06, + "loss": 0.88, + "step": 59819 + }, + { + "epoch": 4.42, + "learning_rate": 3.4205355006769315e-06, + "loss": 1.0154, + "step": 59820 + }, + { + "epoch": 4.42, + "learning_rate": 3.4202351297670067e-06, + "loss": 0.9913, + "step": 59821 + }, + { + "epoch": 4.42, + "learning_rate": 3.4199347693253935e-06, + "loss": 0.9376, + "step": 59822 + }, + { + "epoch": 4.42, + "learning_rate": 3.4196344193525676e-06, + "loss": 0.9936, + "step": 59823 + }, + { + "epoch": 4.42, + "learning_rate": 3.4193340798490125e-06, + "loss": 0.907, + "step": 59824 + }, + { + "epoch": 4.42, + "learning_rate": 3.4190337508152015e-06, + "loss": 0.9757, + "step": 59825 + }, + { + "epoch": 4.42, + "learning_rate": 3.418733432251615e-06, + "loss": 1.0461, + "step": 59826 + }, + { + "epoch": 4.42, + "learning_rate": 3.4184331241587256e-06, + "loss": 1.0228, + "step": 59827 + }, + { + "epoch": 4.42, + "learning_rate": 3.4181328265370216e-06, + "loss": 0.9276, + "step": 59828 + }, + { + "epoch": 4.42, + "learning_rate": 3.417832539386967e-06, + "loss": 0.9527, + "step": 59829 + }, + { + "epoch": 4.42, + "learning_rate": 3.4175322627090503e-06, + "loss": 0.9471, + "step": 59830 + }, + { + "epoch": 4.42, + "learning_rate": 3.417231996503746e-06, + "loss": 1.0534, + "step": 59831 + }, + { + "epoch": 4.42, + "learning_rate": 3.4169317407715308e-06, + "loss": 0.9454, + "step": 59832 + }, + { + "epoch": 4.42, + "learning_rate": 3.416631495512883e-06, + "loss": 1.0481, + "step": 59833 + }, + { + "epoch": 4.42, + "learning_rate": 3.416331260728278e-06, + "loss": 1.0397, + "step": 59834 + }, + { + "epoch": 4.42, + "learning_rate": 3.4160310364181983e-06, + "loss": 0.9386, + "step": 59835 + }, + { + "epoch": 4.42, + "learning_rate": 3.4157308225831186e-06, + "loss": 0.9838, + "step": 59836 + }, + { + "epoch": 4.42, + "learning_rate": 3.4154306192235178e-06, + "loss": 1.0704, + "step": 59837 + }, + { + "epoch": 4.42, + "learning_rate": 3.4151304263398688e-06, + "loss": 1.0385, + "step": 59838 + }, + { + "epoch": 4.42, + "learning_rate": 3.4148302439326574e-06, + "loss": 1.0202, + "step": 59839 + }, + { + "epoch": 4.42, + "learning_rate": 3.4145300720023555e-06, + "loss": 0.9815, + "step": 59840 + }, + { + "epoch": 4.42, + "learning_rate": 3.4142299105494425e-06, + "loss": 1.1209, + "step": 59841 + }, + { + "epoch": 4.42, + "learning_rate": 3.413929759574396e-06, + "loss": 1.0983, + "step": 59842 + }, + { + "epoch": 4.42, + "learning_rate": 3.4136296190776896e-06, + "loss": 1.0123, + "step": 59843 + }, + { + "epoch": 4.42, + "learning_rate": 3.4133294890598066e-06, + "loss": 1.0524, + "step": 59844 + }, + { + "epoch": 4.42, + "learning_rate": 3.4130293695212225e-06, + "loss": 1.0836, + "step": 59845 + }, + { + "epoch": 4.42, + "learning_rate": 3.412729260462415e-06, + "loss": 1.0497, + "step": 59846 + }, + { + "epoch": 4.42, + "learning_rate": 3.412429161883858e-06, + "loss": 1.045, + "step": 59847 + }, + { + "epoch": 4.42, + "learning_rate": 3.4121290737860345e-06, + "loss": 0.928, + "step": 59848 + }, + { + "epoch": 4.42, + "learning_rate": 3.411828996169423e-06, + "loss": 1.0464, + "step": 59849 + }, + { + "epoch": 4.42, + "learning_rate": 3.41152892903449e-06, + "loss": 1.0988, + "step": 59850 + }, + { + "epoch": 4.42, + "learning_rate": 3.4112288723817245e-06, + "loss": 1.0054, + "step": 59851 + }, + { + "epoch": 4.42, + "learning_rate": 3.410928826211596e-06, + "loss": 0.9664, + "step": 59852 + }, + { + "epoch": 4.42, + "learning_rate": 3.4106287905245917e-06, + "loss": 1.089, + "step": 59853 + }, + { + "epoch": 4.42, + "learning_rate": 3.4103287653211757e-06, + "loss": 0.9666, + "step": 59854 + }, + { + "epoch": 4.42, + "learning_rate": 3.410028750601837e-06, + "loss": 0.9849, + "step": 59855 + }, + { + "epoch": 4.42, + "learning_rate": 3.4097287463670435e-06, + "loss": 1.0585, + "step": 59856 + }, + { + "epoch": 4.42, + "learning_rate": 3.4094287526172855e-06, + "loss": 0.9326, + "step": 59857 + }, + { + "epoch": 4.42, + "learning_rate": 3.4091287693530274e-06, + "loss": 0.9543, + "step": 59858 + }, + { + "epoch": 4.42, + "learning_rate": 3.408828796574748e-06, + "loss": 0.9469, + "step": 59859 + }, + { + "epoch": 4.42, + "learning_rate": 3.4085288342829313e-06, + "loss": 0.9718, + "step": 59860 + }, + { + "epoch": 4.42, + "learning_rate": 3.408228882478052e-06, + "loss": 0.974, + "step": 59861 + }, + { + "epoch": 4.42, + "learning_rate": 3.407928941160584e-06, + "loss": 0.9821, + "step": 59862 + }, + { + "epoch": 4.42, + "learning_rate": 3.407629010331005e-06, + "loss": 1.0243, + "step": 59863 + }, + { + "epoch": 4.42, + "learning_rate": 3.407329089989797e-06, + "loss": 0.9812, + "step": 59864 + }, + { + "epoch": 4.42, + "learning_rate": 3.4070291801374333e-06, + "loss": 1.0077, + "step": 59865 + }, + { + "epoch": 4.42, + "learning_rate": 3.4067292807743934e-06, + "loss": 0.9149, + "step": 59866 + }, + { + "epoch": 4.42, + "learning_rate": 3.4064293919011483e-06, + "loss": 1.1681, + "step": 59867 + }, + { + "epoch": 4.42, + "learning_rate": 3.4061295135181827e-06, + "loss": 0.9231, + "step": 59868 + }, + { + "epoch": 4.42, + "learning_rate": 3.405829645625972e-06, + "loss": 0.9931, + "step": 59869 + }, + { + "epoch": 4.42, + "learning_rate": 3.4055297882249918e-06, + "loss": 1.0101, + "step": 59870 + }, + { + "epoch": 4.42, + "learning_rate": 3.4052299413157185e-06, + "loss": 1.0372, + "step": 59871 + }, + { + "epoch": 4.42, + "learning_rate": 3.404930104898627e-06, + "loss": 0.9523, + "step": 59872 + }, + { + "epoch": 4.42, + "learning_rate": 3.4046302789742034e-06, + "loss": 0.8905, + "step": 59873 + }, + { + "epoch": 4.42, + "learning_rate": 3.404330463542913e-06, + "loss": 1.052, + "step": 59874 + }, + { + "epoch": 4.42, + "learning_rate": 3.404030658605242e-06, + "loss": 1.0346, + "step": 59875 + }, + { + "epoch": 4.42, + "learning_rate": 3.4037308641616595e-06, + "loss": 0.9364, + "step": 59876 + }, + { + "epoch": 4.42, + "learning_rate": 3.4034310802126535e-06, + "loss": 1.0252, + "step": 59877 + }, + { + "epoch": 4.42, + "learning_rate": 3.4031313067586924e-06, + "loss": 0.9442, + "step": 59878 + }, + { + "epoch": 4.42, + "learning_rate": 3.40283154380025e-06, + "loss": 0.9828, + "step": 59879 + }, + { + "epoch": 4.42, + "learning_rate": 3.402531791337812e-06, + "loss": 0.9809, + "step": 59880 + }, + { + "epoch": 4.42, + "learning_rate": 3.402232049371852e-06, + "loss": 1.0479, + "step": 59881 + }, + { + "epoch": 4.42, + "learning_rate": 3.4019323179028463e-06, + "loss": 1.0965, + "step": 59882 + }, + { + "epoch": 4.42, + "learning_rate": 3.401632596931268e-06, + "loss": 0.9299, + "step": 59883 + }, + { + "epoch": 4.42, + "learning_rate": 3.4013328864576024e-06, + "loss": 1.0423, + "step": 59884 + }, + { + "epoch": 4.42, + "learning_rate": 3.401033186482321e-06, + "loss": 0.8733, + "step": 59885 + }, + { + "epoch": 4.42, + "learning_rate": 3.4007334970059e-06, + "loss": 1.0263, + "step": 59886 + }, + { + "epoch": 4.43, + "learning_rate": 3.40043381802882e-06, + "loss": 1.0058, + "step": 59887 + }, + { + "epoch": 4.43, + "learning_rate": 3.4001341495515505e-06, + "loss": 0.9981, + "step": 59888 + }, + { + "epoch": 4.43, + "learning_rate": 3.3998344915745775e-06, + "loss": 0.9927, + "step": 59889 + }, + { + "epoch": 4.43, + "learning_rate": 3.3995348440983723e-06, + "loss": 1.0625, + "step": 59890 + }, + { + "epoch": 4.43, + "learning_rate": 3.3992352071234135e-06, + "loss": 0.9305, + "step": 59891 + }, + { + "epoch": 4.43, + "learning_rate": 3.3989355806501733e-06, + "loss": 0.9461, + "step": 59892 + }, + { + "epoch": 4.43, + "learning_rate": 3.3986359646791357e-06, + "loss": 1.0314, + "step": 59893 + }, + { + "epoch": 4.43, + "learning_rate": 3.398336359210773e-06, + "loss": 0.9915, + "step": 59894 + }, + { + "epoch": 4.43, + "learning_rate": 3.3980367642455636e-06, + "loss": 0.9925, + "step": 59895 + }, + { + "epoch": 4.43, + "learning_rate": 3.397737179783983e-06, + "loss": 0.9549, + "step": 59896 + }, + { + "epoch": 4.43, + "learning_rate": 3.3974376058265057e-06, + "loss": 0.9642, + "step": 59897 + }, + { + "epoch": 4.43, + "learning_rate": 3.3971380423736167e-06, + "loss": 0.985, + "step": 59898 + }, + { + "epoch": 4.43, + "learning_rate": 3.396838489425779e-06, + "loss": 0.9082, + "step": 59899 + }, + { + "epoch": 4.43, + "learning_rate": 3.3965389469834807e-06, + "loss": 0.9803, + "step": 59900 + }, + { + "epoch": 4.43, + "learning_rate": 3.396239415047192e-06, + "loss": 1.0806, + "step": 59901 + }, + { + "epoch": 4.43, + "learning_rate": 3.3959398936173982e-06, + "loss": 0.9783, + "step": 59902 + }, + { + "epoch": 4.43, + "learning_rate": 3.395640382694562e-06, + "loss": 1.1063, + "step": 59903 + }, + { + "epoch": 4.43, + "learning_rate": 3.3953408822791712e-06, + "loss": 1.062, + "step": 59904 + }, + { + "epoch": 4.43, + "learning_rate": 3.3950413923716983e-06, + "loss": 1.0562, + "step": 59905 + }, + { + "epoch": 4.43, + "learning_rate": 3.394741912972621e-06, + "loss": 0.952, + "step": 59906 + }, + { + "epoch": 4.43, + "learning_rate": 3.394442444082414e-06, + "loss": 1.0786, + "step": 59907 + }, + { + "epoch": 4.43, + "learning_rate": 3.394142985701552e-06, + "loss": 1.0025, + "step": 59908 + }, + { + "epoch": 4.43, + "learning_rate": 3.393843537830517e-06, + "loss": 1.0332, + "step": 59909 + }, + { + "epoch": 4.43, + "learning_rate": 3.3935441004697823e-06, + "loss": 0.9167, + "step": 59910 + }, + { + "epoch": 4.43, + "learning_rate": 3.3932446736198233e-06, + "loss": 0.867, + "step": 59911 + }, + { + "epoch": 4.43, + "learning_rate": 3.392945257281116e-06, + "loss": 1.0146, + "step": 59912 + }, + { + "epoch": 4.43, + "learning_rate": 3.3926458514541404e-06, + "loss": 0.9658, + "step": 59913 + }, + { + "epoch": 4.43, + "learning_rate": 3.3923464561393747e-06, + "loss": 1.0546, + "step": 59914 + }, + { + "epoch": 4.43, + "learning_rate": 3.3920470713372843e-06, + "loss": 0.9783, + "step": 59915 + }, + { + "epoch": 4.43, + "learning_rate": 3.391747697048355e-06, + "loss": 0.91, + "step": 59916 + }, + { + "epoch": 4.43, + "learning_rate": 3.3914483332730586e-06, + "loss": 0.9616, + "step": 59917 + }, + { + "epoch": 4.43, + "learning_rate": 3.391148980011879e-06, + "loss": 0.9609, + "step": 59918 + }, + { + "epoch": 4.43, + "learning_rate": 3.390849637265281e-06, + "loss": 1.1895, + "step": 59919 + }, + { + "epoch": 4.43, + "learning_rate": 3.3905503050337494e-06, + "loss": 0.9508, + "step": 59920 + }, + { + "epoch": 4.43, + "learning_rate": 3.3902509833177557e-06, + "loss": 0.9989, + "step": 59921 + }, + { + "epoch": 4.43, + "learning_rate": 3.3899516721177827e-06, + "loss": 1.0763, + "step": 59922 + }, + { + "epoch": 4.43, + "learning_rate": 3.3896523714342967e-06, + "loss": 0.93, + "step": 59923 + }, + { + "epoch": 4.43, + "learning_rate": 3.3893530812677823e-06, + "loss": 1.0113, + "step": 59924 + }, + { + "epoch": 4.43, + "learning_rate": 3.3890538016187126e-06, + "loss": 0.9954, + "step": 59925 + }, + { + "epoch": 4.43, + "learning_rate": 3.3887545324875626e-06, + "loss": 0.9391, + "step": 59926 + }, + { + "epoch": 4.43, + "learning_rate": 3.388455273874811e-06, + "loss": 1.0756, + "step": 59927 + }, + { + "epoch": 4.43, + "learning_rate": 3.388156025780929e-06, + "loss": 1.1222, + "step": 59928 + }, + { + "epoch": 4.43, + "learning_rate": 3.3878567882063984e-06, + "loss": 1.0785, + "step": 59929 + }, + { + "epoch": 4.43, + "learning_rate": 3.3875575611516944e-06, + "loss": 1.0347, + "step": 59930 + }, + { + "epoch": 4.43, + "learning_rate": 3.3872583446172903e-06, + "loss": 0.9791, + "step": 59931 + }, + { + "epoch": 4.43, + "learning_rate": 3.3869591386036615e-06, + "loss": 0.8536, + "step": 59932 + }, + { + "epoch": 4.43, + "learning_rate": 3.386659943111289e-06, + "loss": 0.9893, + "step": 59933 + }, + { + "epoch": 4.43, + "learning_rate": 3.3863607581406456e-06, + "loss": 1.0037, + "step": 59934 + }, + { + "epoch": 4.43, + "learning_rate": 3.386061583692208e-06, + "loss": 1.048, + "step": 59935 + }, + { + "epoch": 4.43, + "learning_rate": 3.3857624197664517e-06, + "loss": 1.0301, + "step": 59936 + }, + { + "epoch": 4.43, + "learning_rate": 3.385463266363849e-06, + "loss": 0.8375, + "step": 59937 + }, + { + "epoch": 4.43, + "learning_rate": 3.385164123484883e-06, + "loss": 0.9599, + "step": 59938 + }, + { + "epoch": 4.43, + "learning_rate": 3.384864991130027e-06, + "loss": 0.9869, + "step": 59939 + }, + { + "epoch": 4.43, + "learning_rate": 3.384565869299755e-06, + "loss": 1.0201, + "step": 59940 + }, + { + "epoch": 4.43, + "learning_rate": 3.384266757994542e-06, + "loss": 0.9625, + "step": 59941 + }, + { + "epoch": 4.43, + "learning_rate": 3.383967657214868e-06, + "loss": 0.9292, + "step": 59942 + }, + { + "epoch": 4.43, + "learning_rate": 3.3836685669612113e-06, + "loss": 0.9572, + "step": 59943 + }, + { + "epoch": 4.43, + "learning_rate": 3.383369487234036e-06, + "loss": 0.9297, + "step": 59944 + }, + { + "epoch": 4.43, + "learning_rate": 3.383070418033829e-06, + "loss": 0.9524, + "step": 59945 + }, + { + "epoch": 4.43, + "learning_rate": 3.382771359361058e-06, + "loss": 1.0924, + "step": 59946 + }, + { + "epoch": 4.43, + "learning_rate": 3.3824723112162105e-06, + "loss": 0.9612, + "step": 59947 + }, + { + "epoch": 4.43, + "learning_rate": 3.382173273599748e-06, + "loss": 0.9816, + "step": 59948 + }, + { + "epoch": 4.43, + "learning_rate": 3.381874246512157e-06, + "loss": 1.081, + "step": 59949 + }, + { + "epoch": 4.43, + "learning_rate": 3.3815752299539084e-06, + "loss": 1.0553, + "step": 59950 + }, + { + "epoch": 4.43, + "learning_rate": 3.381276223925479e-06, + "loss": 1.0601, + "step": 59951 + }, + { + "epoch": 4.43, + "learning_rate": 3.380977228427346e-06, + "loss": 0.9258, + "step": 59952 + }, + { + "epoch": 4.43, + "learning_rate": 3.3806782434599783e-06, + "loss": 0.9497, + "step": 59953 + }, + { + "epoch": 4.43, + "learning_rate": 3.38037926902386e-06, + "loss": 1.0259, + "step": 59954 + }, + { + "epoch": 4.43, + "learning_rate": 3.380080305119465e-06, + "loss": 1.0157, + "step": 59955 + }, + { + "epoch": 4.43, + "learning_rate": 3.3797813517472665e-06, + "loss": 0.9623, + "step": 59956 + }, + { + "epoch": 4.43, + "learning_rate": 3.3794824089077382e-06, + "loss": 0.9783, + "step": 59957 + }, + { + "epoch": 4.43, + "learning_rate": 3.379183476601362e-06, + "loss": 1.0337, + "step": 59958 + }, + { + "epoch": 4.43, + "learning_rate": 3.378884554828611e-06, + "loss": 1.0777, + "step": 59959 + }, + { + "epoch": 4.43, + "learning_rate": 3.3785856435899587e-06, + "loss": 0.9521, + "step": 59960 + }, + { + "epoch": 4.43, + "learning_rate": 3.3782867428858824e-06, + "loss": 0.9551, + "step": 59961 + }, + { + "epoch": 4.43, + "learning_rate": 3.3779878527168542e-06, + "loss": 1.0025, + "step": 59962 + }, + { + "epoch": 4.43, + "learning_rate": 3.3776889730833586e-06, + "loss": 1.085, + "step": 59963 + }, + { + "epoch": 4.43, + "learning_rate": 3.377390103985858e-06, + "loss": 0.9083, + "step": 59964 + }, + { + "epoch": 4.43, + "learning_rate": 3.3770912454248393e-06, + "loss": 0.9897, + "step": 59965 + }, + { + "epoch": 4.43, + "learning_rate": 3.37679239740077e-06, + "loss": 1.0495, + "step": 59966 + }, + { + "epoch": 4.43, + "learning_rate": 3.3764935599141356e-06, + "loss": 0.9739, + "step": 59967 + }, + { + "epoch": 4.43, + "learning_rate": 3.376194732965399e-06, + "loss": 1.0941, + "step": 59968 + }, + { + "epoch": 4.43, + "learning_rate": 3.375895916555045e-06, + "loss": 0.9328, + "step": 59969 + }, + { + "epoch": 4.43, + "learning_rate": 3.375597110683546e-06, + "loss": 1.0167, + "step": 59970 + }, + { + "epoch": 4.43, + "learning_rate": 3.3752983153513772e-06, + "loss": 1.0126, + "step": 59971 + }, + { + "epoch": 4.43, + "learning_rate": 3.3749995305590134e-06, + "loss": 0.9825, + "step": 59972 + }, + { + "epoch": 4.43, + "learning_rate": 3.3747007563069277e-06, + "loss": 0.9564, + "step": 59973 + }, + { + "epoch": 4.43, + "learning_rate": 3.3744019925956017e-06, + "loss": 0.9052, + "step": 59974 + }, + { + "epoch": 4.43, + "learning_rate": 3.374103239425507e-06, + "loss": 0.9117, + "step": 59975 + }, + { + "epoch": 4.43, + "learning_rate": 3.3738044967971194e-06, + "loss": 1.0177, + "step": 59976 + }, + { + "epoch": 4.43, + "learning_rate": 3.3735057647109102e-06, + "loss": 0.8972, + "step": 59977 + }, + { + "epoch": 4.43, + "learning_rate": 3.3732070431673627e-06, + "loss": 1.0679, + "step": 59978 + }, + { + "epoch": 4.43, + "learning_rate": 3.3729083321669477e-06, + "loss": 0.9464, + "step": 59979 + }, + { + "epoch": 4.43, + "learning_rate": 3.372609631710141e-06, + "loss": 1.0914, + "step": 59980 + }, + { + "epoch": 4.43, + "learning_rate": 3.3723109417974176e-06, + "loss": 0.9917, + "step": 59981 + }, + { + "epoch": 4.43, + "learning_rate": 3.3720122624292482e-06, + "loss": 1.0961, + "step": 59982 + }, + { + "epoch": 4.43, + "learning_rate": 3.371713593606117e-06, + "loss": 1.0108, + "step": 59983 + }, + { + "epoch": 4.43, + "learning_rate": 3.3714149353284943e-06, + "loss": 0.998, + "step": 59984 + }, + { + "epoch": 4.43, + "learning_rate": 3.371116287596855e-06, + "loss": 0.9501, + "step": 59985 + }, + { + "epoch": 4.43, + "learning_rate": 3.370817650411672e-06, + "loss": 1.0335, + "step": 59986 + }, + { + "epoch": 4.43, + "learning_rate": 3.3705190237734265e-06, + "loss": 0.9914, + "step": 59987 + }, + { + "epoch": 4.43, + "learning_rate": 3.3702204076825908e-06, + "loss": 1.094, + "step": 59988 + }, + { + "epoch": 4.43, + "learning_rate": 3.3699218021396395e-06, + "loss": 1.0086, + "step": 59989 + }, + { + "epoch": 4.43, + "learning_rate": 3.3696232071450475e-06, + "loss": 1.0653, + "step": 59990 + }, + { + "epoch": 4.43, + "learning_rate": 3.369324622699287e-06, + "loss": 1.1035, + "step": 59991 + }, + { + "epoch": 4.43, + "learning_rate": 3.3690260488028427e-06, + "loss": 1.0568, + "step": 59992 + }, + { + "epoch": 4.43, + "learning_rate": 3.3687274854561756e-06, + "loss": 1.0179, + "step": 59993 + }, + { + "epoch": 4.43, + "learning_rate": 3.368428932659773e-06, + "loss": 1.029, + "step": 59994 + }, + { + "epoch": 4.43, + "learning_rate": 3.3681303904141037e-06, + "loss": 1.0345, + "step": 59995 + }, + { + "epoch": 4.43, + "learning_rate": 3.3678318587196446e-06, + "loss": 1.0082, + "step": 59996 + }, + { + "epoch": 4.43, + "learning_rate": 3.367533337576867e-06, + "loss": 1.0376, + "step": 59997 + }, + { + "epoch": 4.43, + "learning_rate": 3.367234826986252e-06, + "loss": 1.0128, + "step": 59998 + }, + { + "epoch": 4.43, + "learning_rate": 3.3669363269482726e-06, + "loss": 0.9672, + "step": 59999 + }, + { + "epoch": 4.43, + "learning_rate": 3.366637837463401e-06, + "loss": 0.9434, + "step": 60000 + }, + { + "epoch": 4.43, + "learning_rate": 3.3663393585321147e-06, + "loss": 1.0373, + "step": 60001 + }, + { + "epoch": 4.43, + "learning_rate": 3.3660408901548837e-06, + "loss": 1.0092, + "step": 60002 + }, + { + "epoch": 4.43, + "learning_rate": 3.3657424323321906e-06, + "loss": 1.0152, + "step": 60003 + }, + { + "epoch": 4.43, + "learning_rate": 3.365443985064506e-06, + "loss": 1.0457, + "step": 60004 + }, + { + "epoch": 4.43, + "learning_rate": 3.3651455483523054e-06, + "loss": 0.9719, + "step": 60005 + }, + { + "epoch": 4.43, + "learning_rate": 3.3648471221960598e-06, + "loss": 0.982, + "step": 60006 + }, + { + "epoch": 4.43, + "learning_rate": 3.3645487065962502e-06, + "loss": 1.0202, + "step": 60007 + }, + { + "epoch": 4.43, + "learning_rate": 3.3642503015533524e-06, + "loss": 1.0071, + "step": 60008 + }, + { + "epoch": 4.43, + "learning_rate": 3.3639519070678317e-06, + "loss": 1.0235, + "step": 60009 + }, + { + "epoch": 4.43, + "learning_rate": 3.36365352314017e-06, + "loss": 0.9402, + "step": 60010 + }, + { + "epoch": 4.43, + "learning_rate": 3.3633551497708395e-06, + "loss": 1.0041, + "step": 60011 + }, + { + "epoch": 4.43, + "learning_rate": 3.3630567869603213e-06, + "loss": 1.023, + "step": 60012 + }, + { + "epoch": 4.43, + "learning_rate": 3.362758434709078e-06, + "loss": 0.9559, + "step": 60013 + }, + { + "epoch": 4.43, + "learning_rate": 3.3624600930175945e-06, + "loss": 0.9405, + "step": 60014 + }, + { + "epoch": 4.43, + "learning_rate": 3.362161761886339e-06, + "loss": 0.9974, + "step": 60015 + }, + { + "epoch": 4.43, + "learning_rate": 3.3618634413157957e-06, + "loss": 0.9381, + "step": 60016 + }, + { + "epoch": 4.43, + "learning_rate": 3.3615651313064302e-06, + "loss": 0.9924, + "step": 60017 + }, + { + "epoch": 4.43, + "learning_rate": 3.3612668318587148e-06, + "loss": 1.0374, + "step": 60018 + }, + { + "epoch": 4.43, + "learning_rate": 3.360968542973133e-06, + "loss": 1.0455, + "step": 60019 + }, + { + "epoch": 4.43, + "learning_rate": 3.3606702646501554e-06, + "loss": 0.9355, + "step": 60020 + }, + { + "epoch": 4.43, + "learning_rate": 3.360371996890256e-06, + "loss": 0.9415, + "step": 60021 + }, + { + "epoch": 4.43, + "learning_rate": 3.3600737396939063e-06, + "loss": 0.9522, + "step": 60022 + }, + { + "epoch": 4.44, + "learning_rate": 3.359775493061589e-06, + "loss": 1.0132, + "step": 60023 + }, + { + "epoch": 4.44, + "learning_rate": 3.3594772569937728e-06, + "loss": 0.9888, + "step": 60024 + }, + { + "epoch": 4.44, + "learning_rate": 3.3591790314909333e-06, + "loss": 0.9646, + "step": 60025 + }, + { + "epoch": 4.44, + "learning_rate": 3.3588808165535445e-06, + "loss": 1.0369, + "step": 60026 + }, + { + "epoch": 4.44, + "learning_rate": 3.3585826121820786e-06, + "loss": 1.0045, + "step": 60027 + }, + { + "epoch": 4.44, + "learning_rate": 3.3582844183770158e-06, + "loss": 0.9393, + "step": 60028 + }, + { + "epoch": 4.44, + "learning_rate": 3.357986235138827e-06, + "loss": 0.9598, + "step": 60029 + }, + { + "epoch": 4.44, + "learning_rate": 3.3576880624679876e-06, + "loss": 1.0386, + "step": 60030 + }, + { + "epoch": 4.44, + "learning_rate": 3.357389900364969e-06, + "loss": 1.0696, + "step": 60031 + }, + { + "epoch": 4.44, + "learning_rate": 3.35709174883025e-06, + "loss": 0.9422, + "step": 60032 + }, + { + "epoch": 4.44, + "learning_rate": 3.356793607864304e-06, + "loss": 1.0105, + "step": 60033 + }, + { + "epoch": 4.44, + "learning_rate": 3.356495477467603e-06, + "loss": 0.9989, + "step": 60034 + }, + { + "epoch": 4.44, + "learning_rate": 3.3561973576406205e-06, + "loss": 1.0201, + "step": 60035 + }, + { + "epoch": 4.44, + "learning_rate": 3.3558992483838395e-06, + "loss": 1.0174, + "step": 60036 + }, + { + "epoch": 4.44, + "learning_rate": 3.355601149697724e-06, + "loss": 0.9464, + "step": 60037 + }, + { + "epoch": 4.44, + "learning_rate": 3.355303061582749e-06, + "loss": 0.8853, + "step": 60038 + }, + { + "epoch": 4.44, + "learning_rate": 3.355004984039395e-06, + "loss": 0.9946, + "step": 60039 + }, + { + "epoch": 4.44, + "learning_rate": 3.354706917068132e-06, + "loss": 0.8639, + "step": 60040 + }, + { + "epoch": 4.44, + "learning_rate": 3.354408860669437e-06, + "loss": 1.0669, + "step": 60041 + }, + { + "epoch": 4.44, + "learning_rate": 3.3541108148437772e-06, + "loss": 0.9649, + "step": 60042 + }, + { + "epoch": 4.44, + "learning_rate": 3.3538127795916363e-06, + "loss": 0.9257, + "step": 60043 + }, + { + "epoch": 4.44, + "learning_rate": 3.353514754913485e-06, + "loss": 0.9936, + "step": 60044 + }, + { + "epoch": 4.44, + "learning_rate": 3.3532167408097958e-06, + "loss": 1.0494, + "step": 60045 + }, + { + "epoch": 4.44, + "learning_rate": 3.3529187372810436e-06, + "loss": 0.9743, + "step": 60046 + }, + { + "epoch": 4.44, + "learning_rate": 3.3526207443276994e-06, + "loss": 1.0127, + "step": 60047 + }, + { + "epoch": 4.44, + "learning_rate": 3.3523227619502443e-06, + "loss": 0.9957, + "step": 60048 + }, + { + "epoch": 4.44, + "learning_rate": 3.3520247901491474e-06, + "loss": 1.0156, + "step": 60049 + }, + { + "epoch": 4.44, + "learning_rate": 3.351726828924885e-06, + "loss": 0.97, + "step": 60050 + }, + { + "epoch": 4.44, + "learning_rate": 3.3514288782779256e-06, + "loss": 0.9436, + "step": 60051 + }, + { + "epoch": 4.44, + "learning_rate": 3.351130938208752e-06, + "loss": 0.9469, + "step": 60052 + }, + { + "epoch": 4.44, + "learning_rate": 3.3508330087178342e-06, + "loss": 0.9108, + "step": 60053 + }, + { + "epoch": 4.44, + "learning_rate": 3.3505350898056453e-06, + "loss": 0.9596, + "step": 60054 + }, + { + "epoch": 4.44, + "learning_rate": 3.35023718147266e-06, + "loss": 0.9332, + "step": 60055 + }, + { + "epoch": 4.44, + "learning_rate": 3.349939283719349e-06, + "loss": 0.7903, + "step": 60056 + }, + { + "epoch": 4.44, + "learning_rate": 3.349641396546196e-06, + "loss": 0.991, + "step": 60057 + }, + { + "epoch": 4.44, + "learning_rate": 3.3493435199536615e-06, + "loss": 0.957, + "step": 60058 + }, + { + "epoch": 4.44, + "learning_rate": 3.34904565394223e-06, + "loss": 1.0311, + "step": 60059 + }, + { + "epoch": 4.44, + "learning_rate": 3.3487477985123683e-06, + "loss": 0.9506, + "step": 60060 + }, + { + "epoch": 4.44, + "learning_rate": 3.348449953664561e-06, + "loss": 0.9666, + "step": 60061 + }, + { + "epoch": 4.44, + "learning_rate": 3.3481521193992672e-06, + "loss": 1.0036, + "step": 60062 + }, + { + "epoch": 4.44, + "learning_rate": 3.3478542957169724e-06, + "loss": 1.0216, + "step": 60063 + }, + { + "epoch": 4.44, + "learning_rate": 3.347556482618145e-06, + "loss": 0.8365, + "step": 60064 + }, + { + "epoch": 4.44, + "learning_rate": 3.3472586801032603e-06, + "loss": 1.0407, + "step": 60065 + }, + { + "epoch": 4.44, + "learning_rate": 3.346960888172792e-06, + "loss": 1.0225, + "step": 60066 + }, + { + "epoch": 4.44, + "learning_rate": 3.3466631068272114e-06, + "loss": 0.9689, + "step": 60067 + }, + { + "epoch": 4.44, + "learning_rate": 3.346365336066997e-06, + "loss": 0.9497, + "step": 60068 + }, + { + "epoch": 4.44, + "learning_rate": 3.3460675758926197e-06, + "loss": 0.9885, + "step": 60069 + }, + { + "epoch": 4.44, + "learning_rate": 3.3457698263045548e-06, + "loss": 1.0533, + "step": 60070 + }, + { + "epoch": 4.44, + "learning_rate": 3.345472087303271e-06, + "loss": 1.0819, + "step": 60071 + }, + { + "epoch": 4.44, + "learning_rate": 3.345174358889249e-06, + "loss": 0.9957, + "step": 60072 + }, + { + "epoch": 4.44, + "learning_rate": 3.3448766410629598e-06, + "loss": 1.0419, + "step": 60073 + }, + { + "epoch": 4.44, + "learning_rate": 3.344578933824877e-06, + "loss": 1.0974, + "step": 60074 + }, + { + "epoch": 4.44, + "learning_rate": 3.344281237175473e-06, + "loss": 1.1068, + "step": 60075 + }, + { + "epoch": 4.44, + "learning_rate": 3.3439835511152196e-06, + "loss": 1.118, + "step": 60076 + }, + { + "epoch": 4.44, + "learning_rate": 3.3436858756445967e-06, + "loss": 0.9462, + "step": 60077 + }, + { + "epoch": 4.44, + "learning_rate": 3.3433882107640735e-06, + "loss": 0.9325, + "step": 60078 + }, + { + "epoch": 4.44, + "learning_rate": 3.3430905564741244e-06, + "loss": 0.921, + "step": 60079 + }, + { + "epoch": 4.44, + "learning_rate": 3.34279291277522e-06, + "loss": 0.969, + "step": 60080 + }, + { + "epoch": 4.44, + "learning_rate": 3.3424952796678432e-06, + "loss": 1.1748, + "step": 60081 + }, + { + "epoch": 4.44, + "learning_rate": 3.342197657152455e-06, + "loss": 1.0359, + "step": 60082 + }, + { + "epoch": 4.44, + "learning_rate": 3.3419000452295393e-06, + "loss": 1.0096, + "step": 60083 + }, + { + "epoch": 4.44, + "learning_rate": 3.3416024438995642e-06, + "loss": 1.0538, + "step": 60084 + }, + { + "epoch": 4.44, + "learning_rate": 3.3413048531630045e-06, + "loss": 0.9461, + "step": 60085 + }, + { + "epoch": 4.44, + "learning_rate": 3.3410072730203325e-06, + "loss": 0.9463, + "step": 60086 + }, + { + "epoch": 4.44, + "learning_rate": 3.3407097034720203e-06, + "loss": 1.0119, + "step": 60087 + }, + { + "epoch": 4.44, + "learning_rate": 3.340412144518548e-06, + "loss": 1.0358, + "step": 60088 + }, + { + "epoch": 4.44, + "learning_rate": 3.340114596160383e-06, + "loss": 1.0382, + "step": 60089 + }, + { + "epoch": 4.44, + "learning_rate": 3.3398170583980018e-06, + "loss": 0.9217, + "step": 60090 + }, + { + "epoch": 4.44, + "learning_rate": 3.3395195312318716e-06, + "loss": 0.8927, + "step": 60091 + }, + { + "epoch": 4.44, + "learning_rate": 3.3392220146624744e-06, + "loss": 1.0337, + "step": 60092 + }, + { + "epoch": 4.44, + "learning_rate": 3.33892450869028e-06, + "loss": 0.9441, + "step": 60093 + }, + { + "epoch": 4.44, + "learning_rate": 3.3386270133157618e-06, + "loss": 1.0884, + "step": 60094 + }, + { + "epoch": 4.44, + "learning_rate": 3.3383295285393913e-06, + "loss": 1.0274, + "step": 60095 + }, + { + "epoch": 4.44, + "learning_rate": 3.338032054361641e-06, + "loss": 0.9971, + "step": 60096 + }, + { + "epoch": 4.44, + "learning_rate": 3.3377345907829894e-06, + "loss": 0.8967, + "step": 60097 + }, + { + "epoch": 4.44, + "learning_rate": 3.337437137803907e-06, + "loss": 0.9558, + "step": 60098 + }, + { + "epoch": 4.44, + "learning_rate": 3.337139695424867e-06, + "loss": 0.9686, + "step": 60099 + }, + { + "epoch": 4.44, + "learning_rate": 3.3368422636463383e-06, + "loss": 1.0539, + "step": 60100 + }, + { + "epoch": 4.44, + "learning_rate": 3.336544842468802e-06, + "loss": 0.9523, + "step": 60101 + }, + { + "epoch": 4.44, + "learning_rate": 3.3362474318927305e-06, + "loss": 0.9383, + "step": 60102 + }, + { + "epoch": 4.44, + "learning_rate": 3.3359500319185878e-06, + "loss": 0.9231, + "step": 60103 + }, + { + "epoch": 4.44, + "learning_rate": 3.335652642546856e-06, + "loss": 0.9148, + "step": 60104 + }, + { + "epoch": 4.44, + "learning_rate": 3.335355263778003e-06, + "loss": 1.1052, + "step": 60105 + }, + { + "epoch": 4.44, + "learning_rate": 3.33505789561251e-06, + "loss": 0.9673, + "step": 60106 + }, + { + "epoch": 4.44, + "learning_rate": 3.33476053805084e-06, + "loss": 1.002, + "step": 60107 + }, + { + "epoch": 4.44, + "learning_rate": 3.334463191093471e-06, + "loss": 0.9535, + "step": 60108 + }, + { + "epoch": 4.44, + "learning_rate": 3.334165854740874e-06, + "loss": 1.0394, + "step": 60109 + }, + { + "epoch": 4.44, + "learning_rate": 3.33386852899353e-06, + "loss": 1.0241, + "step": 60110 + }, + { + "epoch": 4.44, + "learning_rate": 3.333571213851904e-06, + "loss": 1.075, + "step": 60111 + }, + { + "epoch": 4.44, + "learning_rate": 3.3332739093164655e-06, + "loss": 1.0296, + "step": 60112 + }, + { + "epoch": 4.44, + "learning_rate": 3.3329766153876964e-06, + "loss": 1.0303, + "step": 60113 + }, + { + "epoch": 4.44, + "learning_rate": 3.3326793320660667e-06, + "loss": 1.0414, + "step": 60114 + }, + { + "epoch": 4.44, + "learning_rate": 3.3323820593520484e-06, + "loss": 0.9121, + "step": 60115 + }, + { + "epoch": 4.44, + "learning_rate": 3.3320847972461123e-06, + "loss": 0.9423, + "step": 60116 + }, + { + "epoch": 4.44, + "learning_rate": 3.331787545748737e-06, + "loss": 0.9648, + "step": 60117 + }, + { + "epoch": 4.44, + "learning_rate": 3.331490304860392e-06, + "loss": 1.0482, + "step": 60118 + }, + { + "epoch": 4.44, + "learning_rate": 3.33119307458155e-06, + "loss": 1.009, + "step": 60119 + }, + { + "epoch": 4.44, + "learning_rate": 3.3308958549126845e-06, + "loss": 0.9004, + "step": 60120 + }, + { + "epoch": 4.44, + "learning_rate": 3.330598645854266e-06, + "loss": 1.0243, + "step": 60121 + }, + { + "epoch": 4.44, + "learning_rate": 3.330301447406772e-06, + "loss": 1.0135, + "step": 60122 + }, + { + "epoch": 4.44, + "learning_rate": 3.330004259570674e-06, + "loss": 0.9951, + "step": 60123 + }, + { + "epoch": 4.44, + "learning_rate": 3.329707082346443e-06, + "loss": 0.9569, + "step": 60124 + }, + { + "epoch": 4.44, + "learning_rate": 3.3294099157345494e-06, + "loss": 0.9849, + "step": 60125 + }, + { + "epoch": 4.44, + "learning_rate": 3.3291127597354756e-06, + "loss": 0.9448, + "step": 60126 + }, + { + "epoch": 4.44, + "learning_rate": 3.328815614349681e-06, + "loss": 0.948, + "step": 60127 + }, + { + "epoch": 4.44, + "learning_rate": 3.328518479577649e-06, + "loss": 1.0243, + "step": 60128 + }, + { + "epoch": 4.44, + "learning_rate": 3.3282213554198495e-06, + "loss": 0.9472, + "step": 60129 + }, + { + "epoch": 4.44, + "learning_rate": 3.3279242418767533e-06, + "loss": 1.0766, + "step": 60130 + }, + { + "epoch": 4.44, + "learning_rate": 3.3276271389488345e-06, + "loss": 0.9909, + "step": 60131 + }, + { + "epoch": 4.44, + "learning_rate": 3.327330046636562e-06, + "loss": 0.9595, + "step": 60132 + }, + { + "epoch": 4.44, + "learning_rate": 3.3270329649404156e-06, + "loss": 0.9151, + "step": 60133 + }, + { + "epoch": 4.44, + "learning_rate": 3.326735893860864e-06, + "loss": 0.8924, + "step": 60134 + }, + { + "epoch": 4.44, + "learning_rate": 3.32643883339838e-06, + "loss": 0.9792, + "step": 60135 + }, + { + "epoch": 4.44, + "learning_rate": 3.3261417835534325e-06, + "loss": 1.0244, + "step": 60136 + }, + { + "epoch": 4.44, + "learning_rate": 3.325844744326503e-06, + "loss": 1.033, + "step": 60137 + }, + { + "epoch": 4.44, + "learning_rate": 3.325547715718057e-06, + "loss": 0.9041, + "step": 60138 + }, + { + "epoch": 4.44, + "learning_rate": 3.3252506977285702e-06, + "loss": 0.9983, + "step": 60139 + }, + { + "epoch": 4.44, + "learning_rate": 3.3249536903585146e-06, + "loss": 0.9502, + "step": 60140 + }, + { + "epoch": 4.44, + "learning_rate": 3.324656693608358e-06, + "loss": 0.8449, + "step": 60141 + }, + { + "epoch": 4.44, + "learning_rate": 3.3243597074785803e-06, + "loss": 1.1198, + "step": 60142 + }, + { + "epoch": 4.44, + "learning_rate": 3.3240627319696504e-06, + "loss": 1.0143, + "step": 60143 + }, + { + "epoch": 4.44, + "learning_rate": 3.323765767082042e-06, + "loss": 0.9, + "step": 60144 + }, + { + "epoch": 4.44, + "learning_rate": 3.3234688128162242e-06, + "loss": 0.9362, + "step": 60145 + }, + { + "epoch": 4.44, + "learning_rate": 3.3231718691726746e-06, + "loss": 1.1198, + "step": 60146 + }, + { + "epoch": 4.44, + "learning_rate": 3.322874936151862e-06, + "loss": 0.9854, + "step": 60147 + }, + { + "epoch": 4.44, + "learning_rate": 3.3225780137542607e-06, + "loss": 1.0298, + "step": 60148 + }, + { + "epoch": 4.44, + "learning_rate": 3.3222811019803427e-06, + "loss": 0.9937, + "step": 60149 + }, + { + "epoch": 4.44, + "learning_rate": 3.321984200830576e-06, + "loss": 0.9971, + "step": 60150 + }, + { + "epoch": 4.44, + "learning_rate": 3.3216873103054437e-06, + "loss": 1.0399, + "step": 60151 + }, + { + "epoch": 4.44, + "learning_rate": 3.3213904304054047e-06, + "loss": 0.9281, + "step": 60152 + }, + { + "epoch": 4.44, + "learning_rate": 3.3210935611309404e-06, + "loss": 1.0247, + "step": 60153 + }, + { + "epoch": 4.44, + "learning_rate": 3.3207967024825184e-06, + "loss": 1.0089, + "step": 60154 + }, + { + "epoch": 4.44, + "learning_rate": 3.3204998544606194e-06, + "loss": 0.95, + "step": 60155 + }, + { + "epoch": 4.44, + "learning_rate": 3.320203017065703e-06, + "loss": 1.1636, + "step": 60156 + }, + { + "epoch": 4.44, + "learning_rate": 3.319906190298252e-06, + "loss": 1.0373, + "step": 60157 + }, + { + "epoch": 4.45, + "learning_rate": 3.3196093741587343e-06, + "loss": 1.0959, + "step": 60158 + }, + { + "epoch": 4.45, + "learning_rate": 3.3193125686476223e-06, + "loss": 0.9961, + "step": 60159 + }, + { + "epoch": 4.45, + "learning_rate": 3.319015773765388e-06, + "loss": 1.0295, + "step": 60160 + }, + { + "epoch": 4.45, + "learning_rate": 3.3187189895125018e-06, + "loss": 1.0387, + "step": 60161 + }, + { + "epoch": 4.45, + "learning_rate": 3.318422215889441e-06, + "loss": 1.1033, + "step": 60162 + }, + { + "epoch": 4.45, + "learning_rate": 3.3181254528966746e-06, + "loss": 0.9926, + "step": 60163 + }, + { + "epoch": 4.45, + "learning_rate": 3.3178287005346753e-06, + "loss": 0.9589, + "step": 60164 + }, + { + "epoch": 4.45, + "learning_rate": 3.317531958803911e-06, + "loss": 1.1637, + "step": 60165 + }, + { + "epoch": 4.45, + "learning_rate": 3.317235227704861e-06, + "loss": 0.9453, + "step": 60166 + }, + { + "epoch": 4.45, + "learning_rate": 3.3169385072379955e-06, + "loss": 1.0045, + "step": 60167 + }, + { + "epoch": 4.45, + "learning_rate": 3.3166417974037845e-06, + "loss": 1.1048, + "step": 60168 + }, + { + "epoch": 4.45, + "learning_rate": 3.3163450982027003e-06, + "loss": 1.0411, + "step": 60169 + }, + { + "epoch": 4.45, + "learning_rate": 3.316048409635212e-06, + "loss": 1.0025, + "step": 60170 + }, + { + "epoch": 4.45, + "learning_rate": 3.315751731701803e-06, + "loss": 1.0239, + "step": 60171 + }, + { + "epoch": 4.45, + "learning_rate": 3.3154550644029292e-06, + "loss": 0.9489, + "step": 60172 + }, + { + "epoch": 4.45, + "learning_rate": 3.3151584077390752e-06, + "loss": 0.9692, + "step": 60173 + }, + { + "epoch": 4.45, + "learning_rate": 3.3148617617107037e-06, + "loss": 1.1087, + "step": 60174 + }, + { + "epoch": 4.45, + "learning_rate": 3.3145651263182996e-06, + "loss": 1.0333, + "step": 60175 + }, + { + "epoch": 4.45, + "learning_rate": 3.3142685015623233e-06, + "loss": 1.0387, + "step": 60176 + }, + { + "epoch": 4.45, + "learning_rate": 3.3139718874432457e-06, + "loss": 0.9649, + "step": 60177 + }, + { + "epoch": 4.45, + "learning_rate": 3.3136752839615484e-06, + "loss": 0.997, + "step": 60178 + }, + { + "epoch": 4.45, + "learning_rate": 3.3133786911176967e-06, + "loss": 0.9603, + "step": 60179 + }, + { + "epoch": 4.45, + "learning_rate": 3.3130821089121634e-06, + "loss": 1.0655, + "step": 60180 + }, + { + "epoch": 4.45, + "learning_rate": 3.3127855373454186e-06, + "loss": 0.9541, + "step": 60181 + }, + { + "epoch": 4.45, + "learning_rate": 3.312488976417939e-06, + "loss": 1.0035, + "step": 60182 + }, + { + "epoch": 4.45, + "learning_rate": 3.312192426130194e-06, + "loss": 1.0585, + "step": 60183 + }, + { + "epoch": 4.45, + "learning_rate": 3.3118958864826556e-06, + "loss": 1.0495, + "step": 60184 + }, + { + "epoch": 4.45, + "learning_rate": 3.3115993574757944e-06, + "loss": 0.9637, + "step": 60185 + }, + { + "epoch": 4.45, + "learning_rate": 3.3113028391100798e-06, + "loss": 0.9883, + "step": 60186 + }, + { + "epoch": 4.45, + "learning_rate": 3.31100633138599e-06, + "loss": 1.0001, + "step": 60187 + }, + { + "epoch": 4.45, + "learning_rate": 3.3107098343039936e-06, + "loss": 0.9991, + "step": 60188 + }, + { + "epoch": 4.45, + "learning_rate": 3.310413347864562e-06, + "loss": 0.8626, + "step": 60189 + }, + { + "epoch": 4.45, + "learning_rate": 3.310116872068163e-06, + "loss": 0.9951, + "step": 60190 + }, + { + "epoch": 4.45, + "learning_rate": 3.3098204069152773e-06, + "loss": 0.9316, + "step": 60191 + }, + { + "epoch": 4.45, + "learning_rate": 3.3095239524063703e-06, + "loss": 1.0995, + "step": 60192 + }, + { + "epoch": 4.45, + "learning_rate": 3.309227508541916e-06, + "loss": 1.0526, + "step": 60193 + }, + { + "epoch": 4.45, + "learning_rate": 3.3089310753223803e-06, + "loss": 1.0384, + "step": 60194 + }, + { + "epoch": 4.45, + "learning_rate": 3.308634652748244e-06, + "loss": 1.0134, + "step": 60195 + }, + { + "epoch": 4.45, + "learning_rate": 3.308338240819977e-06, + "loss": 0.9583, + "step": 60196 + }, + { + "epoch": 4.45, + "learning_rate": 3.308041839538042e-06, + "loss": 1.0045, + "step": 60197 + }, + { + "epoch": 4.45, + "learning_rate": 3.307745448902919e-06, + "loss": 1.1224, + "step": 60198 + }, + { + "epoch": 4.45, + "learning_rate": 3.307449068915075e-06, + "loss": 0.896, + "step": 60199 + }, + { + "epoch": 4.45, + "learning_rate": 3.3071526995749904e-06, + "loss": 1.0365, + "step": 60200 + }, + { + "epoch": 4.45, + "learning_rate": 3.306856340883122e-06, + "loss": 1.0968, + "step": 60201 + }, + { + "epoch": 4.45, + "learning_rate": 3.306559992839955e-06, + "loss": 1.009, + "step": 60202 + }, + { + "epoch": 4.45, + "learning_rate": 3.306263655445954e-06, + "loss": 1.0383, + "step": 60203 + }, + { + "epoch": 4.45, + "learning_rate": 3.3059673287015913e-06, + "loss": 1.015, + "step": 60204 + }, + { + "epoch": 4.45, + "learning_rate": 3.30567101260734e-06, + "loss": 0.9786, + "step": 60205 + }, + { + "epoch": 4.45, + "learning_rate": 3.3053747071636654e-06, + "loss": 1.0113, + "step": 60206 + }, + { + "epoch": 4.45, + "learning_rate": 3.3050784123710487e-06, + "loss": 1.004, + "step": 60207 + }, + { + "epoch": 4.45, + "learning_rate": 3.304782128229955e-06, + "loss": 0.9836, + "step": 60208 + }, + { + "epoch": 4.45, + "learning_rate": 3.3044858547408587e-06, + "loss": 1.0423, + "step": 60209 + }, + { + "epoch": 4.45, + "learning_rate": 3.3041895919042256e-06, + "loss": 0.9307, + "step": 60210 + }, + { + "epoch": 4.45, + "learning_rate": 3.303893339720534e-06, + "loss": 1.0079, + "step": 60211 + }, + { + "epoch": 4.45, + "learning_rate": 3.303597098190253e-06, + "loss": 0.8774, + "step": 60212 + }, + { + "epoch": 4.45, + "learning_rate": 3.3033008673138533e-06, + "loss": 1.0307, + "step": 60213 + }, + { + "epoch": 4.45, + "learning_rate": 3.3030046470918052e-06, + "loss": 1.0376, + "step": 60214 + }, + { + "epoch": 4.45, + "learning_rate": 3.302708437524578e-06, + "loss": 1.0104, + "step": 60215 + }, + { + "epoch": 4.45, + "learning_rate": 3.3024122386126522e-06, + "loss": 1.0246, + "step": 60216 + }, + { + "epoch": 4.45, + "learning_rate": 3.3021160503564866e-06, + "loss": 1.0588, + "step": 60217 + }, + { + "epoch": 4.45, + "learning_rate": 3.301819872756562e-06, + "loss": 1.0634, + "step": 60218 + }, + { + "epoch": 4.45, + "learning_rate": 3.301523705813342e-06, + "loss": 1.1155, + "step": 60219 + }, + { + "epoch": 4.45, + "learning_rate": 3.3012275495273095e-06, + "loss": 0.9172, + "step": 60220 + }, + { + "epoch": 4.45, + "learning_rate": 3.3009314038989215e-06, + "loss": 1.0244, + "step": 60221 + }, + { + "epoch": 4.45, + "learning_rate": 3.300635268928658e-06, + "loss": 0.9222, + "step": 60222 + }, + { + "epoch": 4.45, + "learning_rate": 3.3003391446169886e-06, + "loss": 1.0568, + "step": 60223 + }, + { + "epoch": 4.45, + "learning_rate": 3.3000430309643836e-06, + "loss": 1.0466, + "step": 60224 + }, + { + "epoch": 4.45, + "learning_rate": 3.2997469279713147e-06, + "loss": 0.8931, + "step": 60225 + }, + { + "epoch": 4.45, + "learning_rate": 3.2994508356382495e-06, + "loss": 1.0495, + "step": 60226 + }, + { + "epoch": 4.45, + "learning_rate": 3.2991547539656643e-06, + "loss": 1.1127, + "step": 60227 + }, + { + "epoch": 4.45, + "learning_rate": 3.2988586829540294e-06, + "loss": 1.0153, + "step": 60228 + }, + { + "epoch": 4.45, + "learning_rate": 3.2985626226038136e-06, + "loss": 1.0135, + "step": 60229 + }, + { + "epoch": 4.45, + "learning_rate": 3.298266572915485e-06, + "loss": 1.0219, + "step": 60230 + }, + { + "epoch": 4.45, + "learning_rate": 3.297970533889523e-06, + "loss": 0.9192, + "step": 60231 + }, + { + "epoch": 4.45, + "learning_rate": 3.2976745055263924e-06, + "loss": 1.0666, + "step": 60232 + }, + { + "epoch": 4.45, + "learning_rate": 3.297378487826567e-06, + "loss": 1.0545, + "step": 60233 + }, + { + "epoch": 4.45, + "learning_rate": 3.2970824807905165e-06, + "loss": 1.0823, + "step": 60234 + }, + { + "epoch": 4.45, + "learning_rate": 3.296786484418708e-06, + "loss": 0.9641, + "step": 60235 + }, + { + "epoch": 4.45, + "learning_rate": 3.2964904987116196e-06, + "loss": 0.8412, + "step": 60236 + }, + { + "epoch": 4.45, + "learning_rate": 3.29619452366972e-06, + "loss": 1.013, + "step": 60237 + }, + { + "epoch": 4.45, + "learning_rate": 3.295898559293478e-06, + "loss": 0.9699, + "step": 60238 + }, + { + "epoch": 4.45, + "learning_rate": 3.2956026055833625e-06, + "loss": 1.0328, + "step": 60239 + }, + { + "epoch": 4.45, + "learning_rate": 3.29530666253985e-06, + "loss": 1.0214, + "step": 60240 + }, + { + "epoch": 4.45, + "learning_rate": 3.2950107301634125e-06, + "loss": 1.0248, + "step": 60241 + }, + { + "epoch": 4.45, + "learning_rate": 3.29471480845451e-06, + "loss": 1.0718, + "step": 60242 + }, + { + "epoch": 4.45, + "learning_rate": 3.294418897413625e-06, + "loss": 1.053, + "step": 60243 + }, + { + "epoch": 4.45, + "learning_rate": 3.294122997041218e-06, + "loss": 0.867, + "step": 60244 + }, + { + "epoch": 4.45, + "learning_rate": 3.2938271073377738e-06, + "loss": 1.0423, + "step": 60245 + }, + { + "epoch": 4.45, + "learning_rate": 3.293531228303748e-06, + "loss": 0.9765, + "step": 60246 + }, + { + "epoch": 4.45, + "learning_rate": 3.2932353599396206e-06, + "loss": 0.9627, + "step": 60247 + }, + { + "epoch": 4.45, + "learning_rate": 3.29293950224586e-06, + "loss": 1.1256, + "step": 60248 + }, + { + "epoch": 4.45, + "learning_rate": 3.2926436552229366e-06, + "loss": 1.0546, + "step": 60249 + }, + { + "epoch": 4.45, + "learning_rate": 3.2923478188713176e-06, + "loss": 0.9553, + "step": 60250 + }, + { + "epoch": 4.45, + "learning_rate": 3.292051993191482e-06, + "loss": 1.0482, + "step": 60251 + }, + { + "epoch": 4.45, + "learning_rate": 3.2917561781838947e-06, + "loss": 0.9833, + "step": 60252 + }, + { + "epoch": 4.45, + "learning_rate": 3.2914603738490267e-06, + "loss": 0.9626, + "step": 60253 + }, + { + "epoch": 4.45, + "learning_rate": 3.2911645801873503e-06, + "loss": 1.0886, + "step": 60254 + }, + { + "epoch": 4.45, + "learning_rate": 3.2908687971993313e-06, + "loss": 0.9975, + "step": 60255 + }, + { + "epoch": 4.45, + "learning_rate": 3.2905730248854474e-06, + "loss": 1.02, + "step": 60256 + }, + { + "epoch": 4.45, + "learning_rate": 3.2902772632461654e-06, + "loss": 0.9194, + "step": 60257 + }, + { + "epoch": 4.45, + "learning_rate": 3.2899815122819566e-06, + "loss": 0.9384, + "step": 60258 + }, + { + "epoch": 4.45, + "learning_rate": 3.2896857719932886e-06, + "loss": 1.0413, + "step": 60259 + }, + { + "epoch": 4.45, + "learning_rate": 3.289390042380637e-06, + "loss": 0.9852, + "step": 60260 + }, + { + "epoch": 4.45, + "learning_rate": 3.289094323444474e-06, + "loss": 1.1068, + "step": 60261 + }, + { + "epoch": 4.45, + "learning_rate": 3.288798615185258e-06, + "loss": 0.9877, + "step": 60262 + }, + { + "epoch": 4.45, + "learning_rate": 3.288502917603471e-06, + "loss": 0.9644, + "step": 60263 + }, + { + "epoch": 4.45, + "learning_rate": 3.288207230699577e-06, + "loss": 1.07, + "step": 60264 + }, + { + "epoch": 4.45, + "learning_rate": 3.287911554474056e-06, + "loss": 0.9821, + "step": 60265 + }, + { + "epoch": 4.45, + "learning_rate": 3.2876158889273644e-06, + "loss": 0.9512, + "step": 60266 + }, + { + "epoch": 4.45, + "learning_rate": 3.287320234059984e-06, + "loss": 0.9752, + "step": 60267 + }, + { + "epoch": 4.45, + "learning_rate": 3.287024589872377e-06, + "loss": 0.9008, + "step": 60268 + }, + { + "epoch": 4.45, + "learning_rate": 3.2867289563650253e-06, + "loss": 1.0002, + "step": 60269 + }, + { + "epoch": 4.45, + "learning_rate": 3.286433333538388e-06, + "loss": 0.9343, + "step": 60270 + }, + { + "epoch": 4.45, + "learning_rate": 3.2861377213929357e-06, + "loss": 1.0588, + "step": 60271 + }, + { + "epoch": 4.45, + "learning_rate": 3.2858421199291455e-06, + "loss": 1.0688, + "step": 60272 + }, + { + "epoch": 4.45, + "learning_rate": 3.285546529147485e-06, + "loss": 1.0203, + "step": 60273 + }, + { + "epoch": 4.45, + "learning_rate": 3.285250949048424e-06, + "loss": 0.9871, + "step": 60274 + }, + { + "epoch": 4.45, + "learning_rate": 3.2849553796324286e-06, + "loss": 1.036, + "step": 60275 + }, + { + "epoch": 4.45, + "learning_rate": 3.2846598208999768e-06, + "loss": 1.0547, + "step": 60276 + }, + { + "epoch": 4.45, + "learning_rate": 3.2843642728515356e-06, + "loss": 0.963, + "step": 60277 + }, + { + "epoch": 4.45, + "learning_rate": 3.284068735487573e-06, + "loss": 0.9342, + "step": 60278 + }, + { + "epoch": 4.45, + "learning_rate": 3.283773208808563e-06, + "loss": 1.0623, + "step": 60279 + }, + { + "epoch": 4.45, + "learning_rate": 3.2834776928149693e-06, + "loss": 1.0909, + "step": 60280 + }, + { + "epoch": 4.45, + "learning_rate": 3.28318218750727e-06, + "loss": 0.9755, + "step": 60281 + }, + { + "epoch": 4.45, + "learning_rate": 3.2828866928859325e-06, + "loss": 1.0704, + "step": 60282 + }, + { + "epoch": 4.45, + "learning_rate": 3.2825912089514255e-06, + "loss": 1.0961, + "step": 60283 + }, + { + "epoch": 4.45, + "learning_rate": 3.282295735704216e-06, + "loss": 1.0548, + "step": 60284 + }, + { + "epoch": 4.45, + "learning_rate": 3.2820002731447818e-06, + "loss": 0.9991, + "step": 60285 + }, + { + "epoch": 4.45, + "learning_rate": 3.281704821273589e-06, + "loss": 1.0091, + "step": 60286 + }, + { + "epoch": 4.45, + "learning_rate": 3.2814093800911083e-06, + "loss": 0.9309, + "step": 60287 + }, + { + "epoch": 4.45, + "learning_rate": 3.2811139495978082e-06, + "loss": 0.9582, + "step": 60288 + }, + { + "epoch": 4.45, + "learning_rate": 3.2808185297941563e-06, + "loss": 1.0698, + "step": 60289 + }, + { + "epoch": 4.45, + "learning_rate": 3.280523120680633e-06, + "loss": 0.9849, + "step": 60290 + }, + { + "epoch": 4.45, + "learning_rate": 3.280227722257695e-06, + "loss": 1.0074, + "step": 60291 + }, + { + "epoch": 4.45, + "learning_rate": 3.2799323345258214e-06, + "loss": 1.0104, + "step": 60292 + }, + { + "epoch": 4.46, + "learning_rate": 3.279636957485479e-06, + "loss": 0.9865, + "step": 60293 + }, + { + "epoch": 4.46, + "learning_rate": 3.2793415911371383e-06, + "loss": 0.9933, + "step": 60294 + }, + { + "epoch": 4.46, + "learning_rate": 3.279046235481266e-06, + "loss": 0.8833, + "step": 60295 + }, + { + "epoch": 4.46, + "learning_rate": 3.2787508905183386e-06, + "loss": 0.9905, + "step": 60296 + }, + { + "epoch": 4.46, + "learning_rate": 3.278455556248822e-06, + "loss": 1.1052, + "step": 60297 + }, + { + "epoch": 4.46, + "learning_rate": 3.2781602326731853e-06, + "loss": 1.1047, + "step": 60298 + }, + { + "epoch": 4.46, + "learning_rate": 3.277864919791901e-06, + "loss": 0.9365, + "step": 60299 + }, + { + "epoch": 4.46, + "learning_rate": 3.2775696176054337e-06, + "loss": 1.1346, + "step": 60300 + }, + { + "epoch": 4.46, + "learning_rate": 3.2772743261142603e-06, + "loss": 1.0019, + "step": 60301 + }, + { + "epoch": 4.46, + "learning_rate": 3.2769790453188467e-06, + "loss": 1.0077, + "step": 60302 + }, + { + "epoch": 4.46, + "learning_rate": 3.2766837752196636e-06, + "loss": 0.9802, + "step": 60303 + }, + { + "epoch": 4.46, + "learning_rate": 3.276388515817177e-06, + "loss": 0.9906, + "step": 60304 + }, + { + "epoch": 4.46, + "learning_rate": 3.276093267111864e-06, + "loss": 1.0609, + "step": 60305 + }, + { + "epoch": 4.46, + "learning_rate": 3.27579802910419e-06, + "loss": 1.0448, + "step": 60306 + }, + { + "epoch": 4.46, + "learning_rate": 3.2755028017946254e-06, + "loss": 0.975, + "step": 60307 + }, + { + "epoch": 4.46, + "learning_rate": 3.2752075851836396e-06, + "loss": 1.0002, + "step": 60308 + }, + { + "epoch": 4.46, + "learning_rate": 3.2749123792716984e-06, + "loss": 0.9595, + "step": 60309 + }, + { + "epoch": 4.46, + "learning_rate": 3.274617184059282e-06, + "loss": 1.0675, + "step": 60310 + }, + { + "epoch": 4.46, + "learning_rate": 3.274321999546848e-06, + "loss": 0.957, + "step": 60311 + }, + { + "epoch": 4.46, + "learning_rate": 3.2740268257348728e-06, + "loss": 1.0039, + "step": 60312 + }, + { + "epoch": 4.46, + "learning_rate": 3.2737316626238224e-06, + "loss": 0.9047, + "step": 60313 + }, + { + "epoch": 4.46, + "learning_rate": 3.2734365102141765e-06, + "loss": 1.0198, + "step": 60314 + }, + { + "epoch": 4.46, + "learning_rate": 3.2731413685063883e-06, + "loss": 1.0942, + "step": 60315 + }, + { + "epoch": 4.46, + "learning_rate": 3.2728462375009405e-06, + "loss": 0.9392, + "step": 60316 + }, + { + "epoch": 4.46, + "learning_rate": 3.2725511171982983e-06, + "loss": 0.8894, + "step": 60317 + }, + { + "epoch": 4.46, + "learning_rate": 3.2722560075989296e-06, + "loss": 1.0292, + "step": 60318 + }, + { + "epoch": 4.46, + "learning_rate": 3.2719609087033065e-06, + "loss": 0.9827, + "step": 60319 + }, + { + "epoch": 4.46, + "learning_rate": 3.2716658205118934e-06, + "loss": 0.9258, + "step": 60320 + }, + { + "epoch": 4.46, + "learning_rate": 3.2713707430251675e-06, + "loss": 0.9895, + "step": 60321 + }, + { + "epoch": 4.46, + "learning_rate": 3.2710756762435946e-06, + "loss": 0.926, + "step": 60322 + }, + { + "epoch": 4.46, + "learning_rate": 3.2707806201676427e-06, + "loss": 0.9324, + "step": 60323 + }, + { + "epoch": 4.46, + "learning_rate": 3.2704855747977816e-06, + "loss": 1.0731, + "step": 60324 + }, + { + "epoch": 4.46, + "learning_rate": 3.2701905401344826e-06, + "loss": 1.0621, + "step": 60325 + }, + { + "epoch": 4.46, + "learning_rate": 3.2698955161782154e-06, + "loss": 1.0152, + "step": 60326 + }, + { + "epoch": 4.46, + "learning_rate": 3.2696005029294487e-06, + "loss": 1.0164, + "step": 60327 + }, + { + "epoch": 4.46, + "learning_rate": 3.26930550038865e-06, + "loss": 0.9536, + "step": 60328 + }, + { + "epoch": 4.46, + "learning_rate": 3.269010508556287e-06, + "loss": 0.9941, + "step": 60329 + }, + { + "epoch": 4.46, + "learning_rate": 3.2687155274328354e-06, + "loss": 0.9586, + "step": 60330 + }, + { + "epoch": 4.46, + "learning_rate": 3.268420557018761e-06, + "loss": 1.0291, + "step": 60331 + }, + { + "epoch": 4.46, + "learning_rate": 3.268125597314533e-06, + "loss": 0.9941, + "step": 60332 + }, + { + "epoch": 4.46, + "learning_rate": 3.267830648320619e-06, + "loss": 1.0307, + "step": 60333 + }, + { + "epoch": 4.46, + "learning_rate": 3.2675357100374916e-06, + "loss": 0.9425, + "step": 60334 + }, + { + "epoch": 4.46, + "learning_rate": 3.2672407824656226e-06, + "loss": 0.9965, + "step": 60335 + }, + { + "epoch": 4.46, + "learning_rate": 3.266945865605471e-06, + "loss": 1.121, + "step": 60336 + }, + { + "epoch": 4.46, + "learning_rate": 3.2666509594575148e-06, + "loss": 0.9288, + "step": 60337 + }, + { + "epoch": 4.46, + "learning_rate": 3.2663560640222215e-06, + "loss": 1.0287, + "step": 60338 + }, + { + "epoch": 4.46, + "learning_rate": 3.2660611793000583e-06, + "loss": 0.9171, + "step": 60339 + }, + { + "epoch": 4.46, + "learning_rate": 3.265766305291492e-06, + "loss": 1.1156, + "step": 60340 + }, + { + "epoch": 4.46, + "learning_rate": 3.2654714419969992e-06, + "loss": 1.1019, + "step": 60341 + }, + { + "epoch": 4.46, + "learning_rate": 3.265176589417045e-06, + "loss": 0.9805, + "step": 60342 + }, + { + "epoch": 4.46, + "learning_rate": 3.2648817475520976e-06, + "loss": 0.9346, + "step": 60343 + }, + { + "epoch": 4.46, + "learning_rate": 3.264586916402628e-06, + "loss": 0.9801, + "step": 60344 + }, + { + "epoch": 4.46, + "learning_rate": 3.2642920959691006e-06, + "loss": 1.0283, + "step": 60345 + }, + { + "epoch": 4.46, + "learning_rate": 3.2639972862519907e-06, + "loss": 0.9641, + "step": 60346 + }, + { + "epoch": 4.46, + "learning_rate": 3.2637024872517666e-06, + "loss": 0.9933, + "step": 60347 + }, + { + "epoch": 4.46, + "learning_rate": 3.2634076989688936e-06, + "loss": 1.0754, + "step": 60348 + }, + { + "epoch": 4.46, + "learning_rate": 3.2631129214038402e-06, + "loss": 1.0615, + "step": 60349 + }, + { + "epoch": 4.46, + "learning_rate": 3.2628181545570804e-06, + "loss": 0.9982, + "step": 60350 + }, + { + "epoch": 4.46, + "learning_rate": 3.2625233984290815e-06, + "loss": 1.1386, + "step": 60351 + }, + { + "epoch": 4.46, + "learning_rate": 3.2622286530203105e-06, + "loss": 1.0735, + "step": 60352 + }, + { + "epoch": 4.46, + "learning_rate": 3.261933918331235e-06, + "loss": 1.0679, + "step": 60353 + }, + { + "epoch": 4.46, + "learning_rate": 3.2616391943623284e-06, + "loss": 0.9527, + "step": 60354 + }, + { + "epoch": 4.46, + "learning_rate": 3.2613444811140614e-06, + "loss": 1.0107, + "step": 60355 + }, + { + "epoch": 4.46, + "learning_rate": 3.2610497785868923e-06, + "loss": 1.032, + "step": 60356 + }, + { + "epoch": 4.46, + "learning_rate": 3.2607550867813e-06, + "loss": 1.0339, + "step": 60357 + }, + { + "epoch": 4.46, + "learning_rate": 3.260460405697746e-06, + "loss": 0.981, + "step": 60358 + }, + { + "epoch": 4.46, + "learning_rate": 3.26016573533671e-06, + "loss": 0.9988, + "step": 60359 + }, + { + "epoch": 4.46, + "learning_rate": 3.2598710756986473e-06, + "loss": 0.9279, + "step": 60360 + }, + { + "epoch": 4.46, + "learning_rate": 3.259576426784037e-06, + "loss": 1.043, + "step": 60361 + }, + { + "epoch": 4.46, + "learning_rate": 3.2592817885933405e-06, + "loss": 0.943, + "step": 60362 + }, + { + "epoch": 4.46, + "learning_rate": 3.2589871611270363e-06, + "loss": 0.96, + "step": 60363 + }, + { + "epoch": 4.46, + "learning_rate": 3.2586925443855845e-06, + "loss": 1.0406, + "step": 60364 + }, + { + "epoch": 4.46, + "learning_rate": 3.2583979383694533e-06, + "loss": 1.0633, + "step": 60365 + }, + { + "epoch": 4.46, + "learning_rate": 3.2581033430791164e-06, + "loss": 1.1007, + "step": 60366 + }, + { + "epoch": 4.46, + "learning_rate": 3.257808758515042e-06, + "loss": 1.0378, + "step": 60367 + }, + { + "epoch": 4.46, + "learning_rate": 3.257514184677697e-06, + "loss": 0.9858, + "step": 60368 + }, + { + "epoch": 4.46, + "learning_rate": 3.257219621567548e-06, + "loss": 0.9336, + "step": 60369 + }, + { + "epoch": 4.46, + "learning_rate": 3.2569250691850685e-06, + "loss": 0.9432, + "step": 60370 + }, + { + "epoch": 4.46, + "learning_rate": 3.2566305275307243e-06, + "loss": 0.9537, + "step": 60371 + }, + { + "epoch": 4.46, + "learning_rate": 3.2563359966049846e-06, + "loss": 1.0111, + "step": 60372 + }, + { + "epoch": 4.46, + "learning_rate": 3.256041476408318e-06, + "loss": 0.9359, + "step": 60373 + }, + { + "epoch": 4.46, + "learning_rate": 3.25574696694119e-06, + "loss": 1.0457, + "step": 60374 + }, + { + "epoch": 4.46, + "learning_rate": 3.2554524682040756e-06, + "loss": 1.0083, + "step": 60375 + }, + { + "epoch": 4.46, + "learning_rate": 3.25515798019744e-06, + "loss": 1.0179, + "step": 60376 + }, + { + "epoch": 4.46, + "learning_rate": 3.25486350292175e-06, + "loss": 0.9628, + "step": 60377 + }, + { + "epoch": 4.46, + "learning_rate": 3.254569036377474e-06, + "loss": 1.0176, + "step": 60378 + }, + { + "epoch": 4.46, + "learning_rate": 3.254274580565088e-06, + "loss": 0.964, + "step": 60379 + }, + { + "epoch": 4.46, + "learning_rate": 3.2539801354850485e-06, + "loss": 1.0626, + "step": 60380 + }, + { + "epoch": 4.46, + "learning_rate": 3.253685701137833e-06, + "loss": 1.0119, + "step": 60381 + }, + { + "epoch": 4.46, + "learning_rate": 3.2533912775239073e-06, + "loss": 0.9314, + "step": 60382 + }, + { + "epoch": 4.46, + "learning_rate": 3.253096864643739e-06, + "loss": 0.9755, + "step": 60383 + }, + { + "epoch": 4.46, + "learning_rate": 3.2528024624977973e-06, + "loss": 0.9479, + "step": 60384 + }, + { + "epoch": 4.46, + "learning_rate": 3.2525080710865476e-06, + "loss": 0.995, + "step": 60385 + }, + { + "epoch": 4.46, + "learning_rate": 3.252213690410464e-06, + "loss": 1.1829, + "step": 60386 + }, + { + "epoch": 4.46, + "learning_rate": 3.2519193204700117e-06, + "loss": 0.8956, + "step": 60387 + }, + { + "epoch": 4.46, + "learning_rate": 3.2516249612656593e-06, + "loss": 0.9668, + "step": 60388 + }, + { + "epoch": 4.46, + "learning_rate": 3.2513306127978705e-06, + "loss": 1.0581, + "step": 60389 + }, + { + "epoch": 4.46, + "learning_rate": 3.251036275067123e-06, + "loss": 0.9969, + "step": 60390 + }, + { + "epoch": 4.46, + "learning_rate": 3.25074194807388e-06, + "loss": 1.1036, + "step": 60391 + }, + { + "epoch": 4.46, + "learning_rate": 3.2504476318186094e-06, + "loss": 1.0633, + "step": 60392 + }, + { + "epoch": 4.46, + "learning_rate": 3.25015332630178e-06, + "loss": 0.9588, + "step": 60393 + }, + { + "epoch": 4.46, + "learning_rate": 3.2498590315238574e-06, + "loss": 1.0659, + "step": 60394 + }, + { + "epoch": 4.46, + "learning_rate": 3.2495647474853163e-06, + "loss": 1.0195, + "step": 60395 + }, + { + "epoch": 4.46, + "learning_rate": 3.249270474186621e-06, + "loss": 1.0096, + "step": 60396 + }, + { + "epoch": 4.46, + "learning_rate": 3.2489762116282388e-06, + "loss": 1.08, + "step": 60397 + }, + { + "epoch": 4.46, + "learning_rate": 3.2486819598106357e-06, + "loss": 0.968, + "step": 60398 + }, + { + "epoch": 4.46, + "learning_rate": 3.248387718734287e-06, + "loss": 0.8805, + "step": 60399 + }, + { + "epoch": 4.46, + "learning_rate": 3.2480934883996598e-06, + "loss": 0.9899, + "step": 60400 + }, + { + "epoch": 4.46, + "learning_rate": 3.2477992688072126e-06, + "loss": 1.0053, + "step": 60401 + }, + { + "epoch": 4.46, + "learning_rate": 3.247505059957423e-06, + "loss": 0.9214, + "step": 60402 + }, + { + "epoch": 4.46, + "learning_rate": 3.2472108618507535e-06, + "loss": 1.0836, + "step": 60403 + }, + { + "epoch": 4.46, + "learning_rate": 3.2469166744876812e-06, + "loss": 0.9615, + "step": 60404 + }, + { + "epoch": 4.46, + "learning_rate": 3.246622497868662e-06, + "loss": 1.1271, + "step": 60405 + }, + { + "epoch": 4.46, + "learning_rate": 3.246328331994172e-06, + "loss": 1.1733, + "step": 60406 + }, + { + "epoch": 4.46, + "learning_rate": 3.2460341768646743e-06, + "loss": 0.9361, + "step": 60407 + }, + { + "epoch": 4.46, + "learning_rate": 3.2457400324806464e-06, + "loss": 0.9596, + "step": 60408 + }, + { + "epoch": 4.46, + "learning_rate": 3.245445898842543e-06, + "loss": 0.963, + "step": 60409 + }, + { + "epoch": 4.46, + "learning_rate": 3.245151775950841e-06, + "loss": 0.9966, + "step": 60410 + }, + { + "epoch": 4.46, + "learning_rate": 3.244857663806006e-06, + "loss": 1.0358, + "step": 60411 + }, + { + "epoch": 4.46, + "learning_rate": 3.2445635624085058e-06, + "loss": 1.0001, + "step": 60412 + }, + { + "epoch": 4.46, + "learning_rate": 3.2442694717588097e-06, + "loss": 1.0219, + "step": 60413 + }, + { + "epoch": 4.46, + "learning_rate": 3.243975391857379e-06, + "loss": 0.9716, + "step": 60414 + }, + { + "epoch": 4.46, + "learning_rate": 3.2436813227046914e-06, + "loss": 0.9071, + "step": 60415 + }, + { + "epoch": 4.46, + "learning_rate": 3.2433872643012097e-06, + "loss": 1.1163, + "step": 60416 + }, + { + "epoch": 4.46, + "learning_rate": 3.2430932166474016e-06, + "loss": 0.9335, + "step": 60417 + }, + { + "epoch": 4.46, + "learning_rate": 3.2427991797437332e-06, + "loss": 1.0158, + "step": 60418 + }, + { + "epoch": 4.46, + "learning_rate": 3.242505153590677e-06, + "loss": 0.9397, + "step": 60419 + }, + { + "epoch": 4.46, + "learning_rate": 3.242211138188699e-06, + "loss": 1.0866, + "step": 60420 + }, + { + "epoch": 4.46, + "learning_rate": 3.241917133538267e-06, + "loss": 1.049, + "step": 60421 + }, + { + "epoch": 4.46, + "learning_rate": 3.2416231396398468e-06, + "loss": 1.143, + "step": 60422 + }, + { + "epoch": 4.46, + "learning_rate": 3.2413291564939054e-06, + "loss": 1.1401, + "step": 60423 + }, + { + "epoch": 4.46, + "learning_rate": 3.241035184100918e-06, + "loss": 1.0702, + "step": 60424 + }, + { + "epoch": 4.46, + "learning_rate": 3.240741222461341e-06, + "loss": 1.0185, + "step": 60425 + }, + { + "epoch": 4.46, + "learning_rate": 3.240447271575652e-06, + "loss": 0.944, + "step": 60426 + }, + { + "epoch": 4.46, + "learning_rate": 3.24015333144431e-06, + "loss": 1.0482, + "step": 60427 + }, + { + "epoch": 4.46, + "learning_rate": 3.2398594020677952e-06, + "loss": 0.9532, + "step": 60428 + }, + { + "epoch": 4.47, + "learning_rate": 3.2395654834465627e-06, + "loss": 1.0306, + "step": 60429 + }, + { + "epoch": 4.47, + "learning_rate": 3.239271575581083e-06, + "loss": 0.9934, + "step": 60430 + }, + { + "epoch": 4.47, + "learning_rate": 3.2389776784718274e-06, + "loss": 1.047, + "step": 60431 + }, + { + "epoch": 4.47, + "learning_rate": 3.2386837921192627e-06, + "loss": 1.1005, + "step": 60432 + }, + { + "epoch": 4.47, + "learning_rate": 3.238389916523855e-06, + "loss": 1.0768, + "step": 60433 + }, + { + "epoch": 4.47, + "learning_rate": 3.238096051686068e-06, + "loss": 1.029, + "step": 60434 + }, + { + "epoch": 4.47, + "learning_rate": 3.2378021976063777e-06, + "loss": 0.9797, + "step": 60435 + }, + { + "epoch": 4.47, + "learning_rate": 3.237508354285248e-06, + "loss": 1.0464, + "step": 60436 + }, + { + "epoch": 4.47, + "learning_rate": 3.2372145217231444e-06, + "loss": 0.9942, + "step": 60437 + }, + { + "epoch": 4.47, + "learning_rate": 3.236920699920536e-06, + "loss": 1.0373, + "step": 60438 + }, + { + "epoch": 4.47, + "learning_rate": 3.236626888877887e-06, + "loss": 1.0652, + "step": 60439 + }, + { + "epoch": 4.47, + "learning_rate": 3.2363330885956713e-06, + "loss": 0.9941, + "step": 60440 + }, + { + "epoch": 4.47, + "learning_rate": 3.236039299074353e-06, + "loss": 0.8607, + "step": 60441 + }, + { + "epoch": 4.47, + "learning_rate": 3.2357455203143993e-06, + "loss": 1.052, + "step": 60442 + }, + { + "epoch": 4.47, + "learning_rate": 3.235451752316274e-06, + "loss": 0.9524, + "step": 60443 + }, + { + "epoch": 4.47, + "learning_rate": 3.2351579950804513e-06, + "loss": 1.0219, + "step": 60444 + }, + { + "epoch": 4.47, + "learning_rate": 3.2348642486073955e-06, + "loss": 1.0388, + "step": 60445 + }, + { + "epoch": 4.47, + "learning_rate": 3.2345705128975747e-06, + "loss": 0.9981, + "step": 60446 + }, + { + "epoch": 4.47, + "learning_rate": 3.2342767879514546e-06, + "loss": 1.0763, + "step": 60447 + }, + { + "epoch": 4.47, + "learning_rate": 3.2339830737695013e-06, + "loss": 0.911, + "step": 60448 + }, + { + "epoch": 4.47, + "learning_rate": 3.233689370352189e-06, + "loss": 0.9941, + "step": 60449 + }, + { + "epoch": 4.47, + "learning_rate": 3.233395677699974e-06, + "loss": 1.085, + "step": 60450 + }, + { + "epoch": 4.47, + "learning_rate": 3.2331019958133335e-06, + "loss": 0.9388, + "step": 60451 + }, + { + "epoch": 4.47, + "learning_rate": 3.2328083246927276e-06, + "loss": 1.0364, + "step": 60452 + }, + { + "epoch": 4.47, + "learning_rate": 3.232514664338633e-06, + "loss": 1.0396, + "step": 60453 + }, + { + "epoch": 4.47, + "learning_rate": 3.232221014751504e-06, + "loss": 1.0628, + "step": 60454 + }, + { + "epoch": 4.47, + "learning_rate": 3.231927375931818e-06, + "loss": 0.9327, + "step": 60455 + }, + { + "epoch": 4.47, + "learning_rate": 3.2316337478800395e-06, + "loss": 1.0321, + "step": 60456 + }, + { + "epoch": 4.47, + "learning_rate": 3.231340130596634e-06, + "loss": 1.0611, + "step": 60457 + }, + { + "epoch": 4.47, + "learning_rate": 3.2310465240820697e-06, + "loss": 1.0004, + "step": 60458 + }, + { + "epoch": 4.47, + "learning_rate": 3.23075292833681e-06, + "loss": 1.0021, + "step": 60459 + }, + { + "epoch": 4.47, + "learning_rate": 3.2304593433613297e-06, + "loss": 0.986, + "step": 60460 + }, + { + "epoch": 4.47, + "learning_rate": 3.2301657691560906e-06, + "loss": 0.9904, + "step": 60461 + }, + { + "epoch": 4.47, + "learning_rate": 3.2298722057215616e-06, + "loss": 1.061, + "step": 60462 + }, + { + "epoch": 4.47, + "learning_rate": 3.229578653058205e-06, + "loss": 0.9422, + "step": 60463 + }, + { + "epoch": 4.47, + "learning_rate": 3.229285111166497e-06, + "loss": 1.0166, + "step": 60464 + }, + { + "epoch": 4.47, + "learning_rate": 3.2289915800468975e-06, + "loss": 0.8986, + "step": 60465 + }, + { + "epoch": 4.47, + "learning_rate": 3.228698059699876e-06, + "loss": 0.9836, + "step": 60466 + }, + { + "epoch": 4.47, + "learning_rate": 3.2284045501258997e-06, + "loss": 0.9724, + "step": 60467 + }, + { + "epoch": 4.47, + "learning_rate": 3.228111051325431e-06, + "loss": 0.9766, + "step": 60468 + }, + { + "epoch": 4.47, + "learning_rate": 3.2278175632989485e-06, + "loss": 0.9678, + "step": 60469 + }, + { + "epoch": 4.47, + "learning_rate": 3.2275240860469027e-06, + "loss": 0.9911, + "step": 60470 + }, + { + "epoch": 4.47, + "learning_rate": 3.2272306195697745e-06, + "loss": 0.9792, + "step": 60471 + }, + { + "epoch": 4.47, + "learning_rate": 3.22693716386802e-06, + "loss": 0.9111, + "step": 60472 + }, + { + "epoch": 4.47, + "learning_rate": 3.2266437189421206e-06, + "loss": 1.0555, + "step": 60473 + }, + { + "epoch": 4.47, + "learning_rate": 3.2263502847925253e-06, + "loss": 0.9385, + "step": 60474 + }, + { + "epoch": 4.47, + "learning_rate": 3.226056861419715e-06, + "loss": 1.0026, + "step": 60475 + }, + { + "epoch": 4.47, + "learning_rate": 3.225763448824151e-06, + "loss": 1.0134, + "step": 60476 + }, + { + "epoch": 4.47, + "learning_rate": 3.2254700470062995e-06, + "loss": 0.9626, + "step": 60477 + }, + { + "epoch": 4.47, + "learning_rate": 3.2251766559666287e-06, + "loss": 1.0368, + "step": 60478 + }, + { + "epoch": 4.47, + "learning_rate": 3.2248832757056016e-06, + "loss": 0.9251, + "step": 60479 + }, + { + "epoch": 4.47, + "learning_rate": 3.224589906223691e-06, + "loss": 1.0986, + "step": 60480 + }, + { + "epoch": 4.47, + "learning_rate": 3.2242965475213615e-06, + "loss": 0.986, + "step": 60481 + }, + { + "epoch": 4.47, + "learning_rate": 3.2240031995990783e-06, + "loss": 1.0706, + "step": 60482 + }, + { + "epoch": 4.47, + "learning_rate": 3.2237098624573073e-06, + "loss": 1.0073, + "step": 60483 + }, + { + "epoch": 4.47, + "learning_rate": 3.2234165360965196e-06, + "loss": 1.0751, + "step": 60484 + }, + { + "epoch": 4.47, + "learning_rate": 3.2231232205171793e-06, + "loss": 1.0487, + "step": 60485 + }, + { + "epoch": 4.47, + "learning_rate": 3.2228299157197528e-06, + "loss": 1.0211, + "step": 60486 + }, + { + "epoch": 4.47, + "learning_rate": 3.222536621704707e-06, + "loss": 1.0565, + "step": 60487 + }, + { + "epoch": 4.47, + "learning_rate": 3.2222433384725047e-06, + "loss": 0.9742, + "step": 60488 + }, + { + "epoch": 4.47, + "learning_rate": 3.22195006602362e-06, + "loss": 1.078, + "step": 60489 + }, + { + "epoch": 4.47, + "learning_rate": 3.221656804358516e-06, + "loss": 0.9689, + "step": 60490 + }, + { + "epoch": 4.47, + "learning_rate": 3.221363553477659e-06, + "loss": 0.8807, + "step": 60491 + }, + { + "epoch": 4.47, + "learning_rate": 3.2210703133815123e-06, + "loss": 1.0517, + "step": 60492 + }, + { + "epoch": 4.47, + "learning_rate": 3.220777084070549e-06, + "loss": 1.0168, + "step": 60493 + }, + { + "epoch": 4.47, + "learning_rate": 3.2204838655452353e-06, + "loss": 0.8754, + "step": 60494 + }, + { + "epoch": 4.47, + "learning_rate": 3.220190657806028e-06, + "loss": 0.9533, + "step": 60495 + }, + { + "epoch": 4.47, + "learning_rate": 3.219897460853405e-06, + "loss": 0.9975, + "step": 60496 + }, + { + "epoch": 4.47, + "learning_rate": 3.2196042746878243e-06, + "loss": 1.0077, + "step": 60497 + }, + { + "epoch": 4.47, + "learning_rate": 3.219311099309762e-06, + "loss": 0.9819, + "step": 60498 + }, + { + "epoch": 4.47, + "learning_rate": 3.2190179347196727e-06, + "loss": 0.9747, + "step": 60499 + }, + { + "epoch": 4.47, + "learning_rate": 3.2187247809180324e-06, + "loss": 0.9268, + "step": 60500 + }, + { + "epoch": 4.47, + "learning_rate": 3.2184316379053027e-06, + "loss": 0.943, + "step": 60501 + }, + { + "epoch": 4.47, + "learning_rate": 3.2181385056819525e-06, + "loss": 1.0102, + "step": 60502 + }, + { + "epoch": 4.47, + "learning_rate": 3.217845384248446e-06, + "loss": 1.0288, + "step": 60503 + }, + { + "epoch": 4.47, + "learning_rate": 3.2175522736052477e-06, + "loss": 1.0282, + "step": 60504 + }, + { + "epoch": 4.47, + "learning_rate": 3.2172591737528303e-06, + "loss": 1.0043, + "step": 60505 + }, + { + "epoch": 4.47, + "learning_rate": 3.2169660846916563e-06, + "loss": 1.0308, + "step": 60506 + }, + { + "epoch": 4.47, + "learning_rate": 3.216673006422192e-06, + "loss": 1.0234, + "step": 60507 + }, + { + "epoch": 4.47, + "learning_rate": 3.2163799389449e-06, + "loss": 0.9847, + "step": 60508 + }, + { + "epoch": 4.47, + "learning_rate": 3.216086882260255e-06, + "loss": 1.0916, + "step": 60509 + }, + { + "epoch": 4.47, + "learning_rate": 3.215793836368718e-06, + "loss": 1.0731, + "step": 60510 + }, + { + "epoch": 4.47, + "learning_rate": 3.2155008012707556e-06, + "loss": 0.998, + "step": 60511 + }, + { + "epoch": 4.47, + "learning_rate": 3.2152077769668356e-06, + "loss": 0.9296, + "step": 60512 + }, + { + "epoch": 4.47, + "learning_rate": 3.2149147634574183e-06, + "loss": 0.9428, + "step": 60513 + }, + { + "epoch": 4.47, + "learning_rate": 3.214621760742982e-06, + "loss": 0.9365, + "step": 60514 + }, + { + "epoch": 4.47, + "learning_rate": 3.214328768823979e-06, + "loss": 0.9552, + "step": 60515 + }, + { + "epoch": 4.47, + "learning_rate": 3.214035787700884e-06, + "loss": 1.0568, + "step": 60516 + }, + { + "epoch": 4.47, + "learning_rate": 3.213742817374158e-06, + "loss": 0.9842, + "step": 60517 + }, + { + "epoch": 4.47, + "learning_rate": 3.2134498578442775e-06, + "loss": 0.9616, + "step": 60518 + }, + { + "epoch": 4.47, + "learning_rate": 3.2131569091116943e-06, + "loss": 0.9491, + "step": 60519 + }, + { + "epoch": 4.47, + "learning_rate": 3.212863971176884e-06, + "loss": 0.9323, + "step": 60520 + }, + { + "epoch": 4.47, + "learning_rate": 3.212571044040307e-06, + "loss": 1.0617, + "step": 60521 + }, + { + "epoch": 4.47, + "learning_rate": 3.2122781277024394e-06, + "loss": 0.9022, + "step": 60522 + }, + { + "epoch": 4.47, + "learning_rate": 3.211985222163736e-06, + "loss": 1.0442, + "step": 60523 + }, + { + "epoch": 4.47, + "learning_rate": 3.211692327424665e-06, + "loss": 0.9942, + "step": 60524 + }, + { + "epoch": 4.47, + "learning_rate": 3.2113994434856964e-06, + "loss": 0.9823, + "step": 60525 + }, + { + "epoch": 4.47, + "learning_rate": 3.2111065703472953e-06, + "loss": 1.0093, + "step": 60526 + }, + { + "epoch": 4.47, + "learning_rate": 3.2108137080099255e-06, + "loss": 0.9455, + "step": 60527 + }, + { + "epoch": 4.47, + "learning_rate": 3.2105208564740507e-06, + "loss": 1.0077, + "step": 60528 + }, + { + "epoch": 4.47, + "learning_rate": 3.2102280157401444e-06, + "loss": 0.9541, + "step": 60529 + }, + { + "epoch": 4.47, + "learning_rate": 3.209935185808667e-06, + "loss": 1.1298, + "step": 60530 + }, + { + "epoch": 4.47, + "learning_rate": 3.209642366680087e-06, + "loss": 0.9386, + "step": 60531 + }, + { + "epoch": 4.47, + "learning_rate": 3.2093495583548673e-06, + "loss": 0.9393, + "step": 60532 + }, + { + "epoch": 4.47, + "learning_rate": 3.2090567608334734e-06, + "loss": 1.0149, + "step": 60533 + }, + { + "epoch": 4.47, + "learning_rate": 3.2087639741163743e-06, + "loss": 1.016, + "step": 60534 + }, + { + "epoch": 4.47, + "learning_rate": 3.208471198204036e-06, + "loss": 0.9774, + "step": 60535 + }, + { + "epoch": 4.47, + "learning_rate": 3.2081784330969235e-06, + "loss": 1.0709, + "step": 60536 + }, + { + "epoch": 4.47, + "learning_rate": 3.207885678795497e-06, + "loss": 0.9769, + "step": 60537 + }, + { + "epoch": 4.47, + "learning_rate": 3.207592935300232e-06, + "loss": 1.0458, + "step": 60538 + }, + { + "epoch": 4.47, + "learning_rate": 3.2073002026115886e-06, + "loss": 0.9717, + "step": 60539 + }, + { + "epoch": 4.47, + "learning_rate": 3.207007480730033e-06, + "loss": 1.0573, + "step": 60540 + }, + { + "epoch": 4.47, + "learning_rate": 3.206714769656032e-06, + "loss": 1.0015, + "step": 60541 + }, + { + "epoch": 4.47, + "learning_rate": 3.206422069390047e-06, + "loss": 0.9911, + "step": 60542 + }, + { + "epoch": 4.47, + "learning_rate": 3.2061293799325545e-06, + "loss": 1.0236, + "step": 60543 + }, + { + "epoch": 4.47, + "learning_rate": 3.2058367012840062e-06, + "loss": 1.1072, + "step": 60544 + }, + { + "epoch": 4.47, + "learning_rate": 3.2055440334448774e-06, + "loss": 1.0429, + "step": 60545 + }, + { + "epoch": 4.47, + "learning_rate": 3.2052513764156313e-06, + "loss": 1.0426, + "step": 60546 + }, + { + "epoch": 4.47, + "learning_rate": 3.204958730196732e-06, + "loss": 1.0314, + "step": 60547 + }, + { + "epoch": 4.47, + "learning_rate": 3.2046660947886445e-06, + "loss": 1.0229, + "step": 60548 + }, + { + "epoch": 4.47, + "learning_rate": 3.204373470191837e-06, + "loss": 1.0934, + "step": 60549 + }, + { + "epoch": 4.47, + "learning_rate": 3.204080856406776e-06, + "loss": 1.0591, + "step": 60550 + }, + { + "epoch": 4.47, + "learning_rate": 3.203788253433925e-06, + "loss": 1.0421, + "step": 60551 + }, + { + "epoch": 4.47, + "learning_rate": 3.203495661273749e-06, + "loss": 0.9839, + "step": 60552 + }, + { + "epoch": 4.47, + "learning_rate": 3.203203079926711e-06, + "loss": 0.9234, + "step": 60553 + }, + { + "epoch": 4.47, + "learning_rate": 3.202910509393283e-06, + "loss": 0.9243, + "step": 60554 + }, + { + "epoch": 4.47, + "learning_rate": 3.202617949673927e-06, + "loss": 0.9408, + "step": 60555 + }, + { + "epoch": 4.47, + "learning_rate": 3.2023254007691085e-06, + "loss": 0.9279, + "step": 60556 + }, + { + "epoch": 4.47, + "learning_rate": 3.2020328626792908e-06, + "loss": 0.9757, + "step": 60557 + }, + { + "epoch": 4.47, + "learning_rate": 3.2017403354049437e-06, + "loss": 1.0384, + "step": 60558 + }, + { + "epoch": 4.47, + "learning_rate": 3.2014478189465335e-06, + "loss": 0.9527, + "step": 60559 + }, + { + "epoch": 4.47, + "learning_rate": 3.201155313304517e-06, + "loss": 0.9326, + "step": 60560 + }, + { + "epoch": 4.47, + "learning_rate": 3.200862818479368e-06, + "loss": 1.0317, + "step": 60561 + }, + { + "epoch": 4.47, + "learning_rate": 3.2005703344715454e-06, + "loss": 1.0404, + "step": 60562 + }, + { + "epoch": 4.47, + "learning_rate": 3.2002778612815242e-06, + "loss": 1.0217, + "step": 60563 + }, + { + "epoch": 4.48, + "learning_rate": 3.1999853989097583e-06, + "loss": 1.025, + "step": 60564 + }, + { + "epoch": 4.48, + "learning_rate": 3.1996929473567215e-06, + "loss": 1.0483, + "step": 60565 + }, + { + "epoch": 4.48, + "learning_rate": 3.1994005066228718e-06, + "loss": 0.921, + "step": 60566 + }, + { + "epoch": 4.48, + "learning_rate": 3.199108076708686e-06, + "loss": 1.0671, + "step": 60567 + }, + { + "epoch": 4.48, + "learning_rate": 3.1988156576146144e-06, + "loss": 1.0773, + "step": 60568 + }, + { + "epoch": 4.48, + "learning_rate": 3.1985232493411344e-06, + "loss": 1.0233, + "step": 60569 + }, + { + "epoch": 4.48, + "learning_rate": 3.198230851888705e-06, + "loss": 1.0522, + "step": 60570 + }, + { + "epoch": 4.48, + "learning_rate": 3.197938465257795e-06, + "loss": 0.9772, + "step": 60571 + }, + { + "epoch": 4.48, + "learning_rate": 3.197646089448866e-06, + "loss": 1.025, + "step": 60572 + }, + { + "epoch": 4.48, + "learning_rate": 3.197353724462382e-06, + "loss": 1.0567, + "step": 60573 + }, + { + "epoch": 4.48, + "learning_rate": 3.197061370298814e-06, + "loss": 0.9374, + "step": 60574 + }, + { + "epoch": 4.48, + "learning_rate": 3.1967690269586237e-06, + "loss": 0.9529, + "step": 60575 + }, + { + "epoch": 4.48, + "learning_rate": 3.1964766944422776e-06, + "loss": 1.01, + "step": 60576 + }, + { + "epoch": 4.48, + "learning_rate": 3.196184372750235e-06, + "loss": 1.0053, + "step": 60577 + }, + { + "epoch": 4.48, + "learning_rate": 3.19589206188297e-06, + "loss": 1.0714, + "step": 60578 + }, + { + "epoch": 4.48, + "learning_rate": 3.195599761840943e-06, + "loss": 0.9493, + "step": 60579 + }, + { + "epoch": 4.48, + "learning_rate": 3.1953074726246204e-06, + "loss": 0.8754, + "step": 60580 + }, + { + "epoch": 4.48, + "learning_rate": 3.1950151942344644e-06, + "loss": 1.0807, + "step": 60581 + }, + { + "epoch": 4.48, + "learning_rate": 3.19472292667094e-06, + "loss": 1.0439, + "step": 60582 + }, + { + "epoch": 4.48, + "learning_rate": 3.1944306699345175e-06, + "loss": 0.9325, + "step": 60583 + }, + { + "epoch": 4.48, + "learning_rate": 3.194138424025658e-06, + "loss": 0.9618, + "step": 60584 + }, + { + "epoch": 4.48, + "learning_rate": 3.193846188944826e-06, + "loss": 0.9425, + "step": 60585 + }, + { + "epoch": 4.48, + "learning_rate": 3.193553964692485e-06, + "loss": 0.988, + "step": 60586 + }, + { + "epoch": 4.48, + "learning_rate": 3.1932617512691056e-06, + "loss": 1.0627, + "step": 60587 + }, + { + "epoch": 4.48, + "learning_rate": 3.192969548675152e-06, + "loss": 0.9643, + "step": 60588 + }, + { + "epoch": 4.48, + "learning_rate": 3.1926773569110805e-06, + "loss": 1.0092, + "step": 60589 + }, + { + "epoch": 4.48, + "learning_rate": 3.192385175977365e-06, + "loss": 0.9588, + "step": 60590 + }, + { + "epoch": 4.48, + "learning_rate": 3.192093005874467e-06, + "loss": 0.9477, + "step": 60591 + }, + { + "epoch": 4.48, + "learning_rate": 3.1918008466028515e-06, + "loss": 1.0411, + "step": 60592 + }, + { + "epoch": 4.48, + "learning_rate": 3.1915086981629817e-06, + "loss": 1.0878, + "step": 60593 + }, + { + "epoch": 4.48, + "learning_rate": 3.1912165605553258e-06, + "loss": 0.9822, + "step": 60594 + }, + { + "epoch": 4.48, + "learning_rate": 3.190924433780348e-06, + "loss": 0.9495, + "step": 60595 + }, + { + "epoch": 4.48, + "learning_rate": 3.190632317838511e-06, + "loss": 1.0243, + "step": 60596 + }, + { + "epoch": 4.48, + "learning_rate": 3.190340212730281e-06, + "loss": 0.9003, + "step": 60597 + }, + { + "epoch": 4.48, + "learning_rate": 3.1900481184561194e-06, + "loss": 1.0024, + "step": 60598 + }, + { + "epoch": 4.48, + "learning_rate": 3.1897560350164958e-06, + "loss": 0.9936, + "step": 60599 + }, + { + "epoch": 4.48, + "learning_rate": 3.189463962411874e-06, + "loss": 0.9678, + "step": 60600 + }, + { + "epoch": 4.48, + "learning_rate": 3.189171900642718e-06, + "loss": 0.9262, + "step": 60601 + }, + { + "epoch": 4.48, + "learning_rate": 3.1888798497094876e-06, + "loss": 0.9414, + "step": 60602 + }, + { + "epoch": 4.48, + "learning_rate": 3.1885878096126564e-06, + "loss": 0.9769, + "step": 60603 + }, + { + "epoch": 4.48, + "learning_rate": 3.1882957803526837e-06, + "loss": 1.0244, + "step": 60604 + }, + { + "epoch": 4.48, + "learning_rate": 3.188003761930035e-06, + "loss": 0.9915, + "step": 60605 + }, + { + "epoch": 4.48, + "learning_rate": 3.1877117543451753e-06, + "loss": 1.0582, + "step": 60606 + }, + { + "epoch": 4.48, + "learning_rate": 3.1874197575985653e-06, + "loss": 1.0268, + "step": 60607 + }, + { + "epoch": 4.48, + "learning_rate": 3.187127771690679e-06, + "loss": 0.9981, + "step": 60608 + }, + { + "epoch": 4.48, + "learning_rate": 3.1868357966219686e-06, + "loss": 1.0225, + "step": 60609 + }, + { + "epoch": 4.48, + "learning_rate": 3.1865438323929086e-06, + "loss": 1.0315, + "step": 60610 + }, + { + "epoch": 4.48, + "learning_rate": 3.1862518790039564e-06, + "loss": 0.9547, + "step": 60611 + }, + { + "epoch": 4.48, + "learning_rate": 3.1859599364555873e-06, + "loss": 1.052, + "step": 60612 + }, + { + "epoch": 4.48, + "learning_rate": 3.18566800474825e-06, + "loss": 1.0938, + "step": 60613 + }, + { + "epoch": 4.48, + "learning_rate": 3.185376083882422e-06, + "loss": 1.0352, + "step": 60614 + }, + { + "epoch": 4.48, + "learning_rate": 3.1850841738585635e-06, + "loss": 1.0507, + "step": 60615 + }, + { + "epoch": 4.48, + "learning_rate": 3.1847922746771377e-06, + "loss": 1.0698, + "step": 60616 + }, + { + "epoch": 4.48, + "learning_rate": 3.1845003863386104e-06, + "loss": 1.0874, + "step": 60617 + }, + { + "epoch": 4.48, + "learning_rate": 3.1842085088434427e-06, + "loss": 1.083, + "step": 60618 + }, + { + "epoch": 4.48, + "learning_rate": 3.1839166421921043e-06, + "loss": 0.9143, + "step": 60619 + }, + { + "epoch": 4.48, + "learning_rate": 3.1836247863850568e-06, + "loss": 1.0684, + "step": 60620 + }, + { + "epoch": 4.48, + "learning_rate": 3.183332941422765e-06, + "loss": 1.0287, + "step": 60621 + }, + { + "epoch": 4.48, + "learning_rate": 3.18304110730569e-06, + "loss": 0.9349, + "step": 60622 + }, + { + "epoch": 4.48, + "learning_rate": 3.182749284034302e-06, + "loss": 1.0914, + "step": 60623 + }, + { + "epoch": 4.48, + "learning_rate": 3.1824574716090616e-06, + "loss": 0.9338, + "step": 60624 + }, + { + "epoch": 4.48, + "learning_rate": 3.1821656700304347e-06, + "loss": 0.8028, + "step": 60625 + }, + { + "epoch": 4.48, + "learning_rate": 3.1818738792988847e-06, + "loss": 0.986, + "step": 60626 + }, + { + "epoch": 4.48, + "learning_rate": 3.1815820994148726e-06, + "loss": 0.9595, + "step": 60627 + }, + { + "epoch": 4.48, + "learning_rate": 3.1812903303788687e-06, + "loss": 1.1034, + "step": 60628 + }, + { + "epoch": 4.48, + "learning_rate": 3.180998572191335e-06, + "loss": 0.9889, + "step": 60629 + }, + { + "epoch": 4.48, + "learning_rate": 3.180706824852734e-06, + "loss": 1.0705, + "step": 60630 + }, + { + "epoch": 4.48, + "learning_rate": 3.180415088363529e-06, + "loss": 0.9417, + "step": 60631 + }, + { + "epoch": 4.48, + "learning_rate": 3.180123362724188e-06, + "loss": 1.0742, + "step": 60632 + }, + { + "epoch": 4.48, + "learning_rate": 3.1798316479351732e-06, + "loss": 0.9461, + "step": 60633 + }, + { + "epoch": 4.48, + "learning_rate": 3.1795399439969497e-06, + "loss": 0.9579, + "step": 60634 + }, + { + "epoch": 4.48, + "learning_rate": 3.1792482509099797e-06, + "loss": 0.9265, + "step": 60635 + }, + { + "epoch": 4.48, + "learning_rate": 3.178956568674728e-06, + "loss": 0.9982, + "step": 60636 + }, + { + "epoch": 4.48, + "learning_rate": 3.17866489729166e-06, + "loss": 1.083, + "step": 60637 + }, + { + "epoch": 4.48, + "learning_rate": 3.1783732367612343e-06, + "loss": 0.8891, + "step": 60638 + }, + { + "epoch": 4.48, + "learning_rate": 3.1780815870839223e-06, + "loss": 1.0338, + "step": 60639 + }, + { + "epoch": 4.48, + "learning_rate": 3.1777899482601846e-06, + "loss": 0.9814, + "step": 60640 + }, + { + "epoch": 4.48, + "learning_rate": 3.1774983202904865e-06, + "loss": 1.0104, + "step": 60641 + }, + { + "epoch": 4.48, + "learning_rate": 3.1772067031752872e-06, + "loss": 0.9482, + "step": 60642 + }, + { + "epoch": 4.48, + "learning_rate": 3.1769150969150575e-06, + "loss": 0.946, + "step": 60643 + }, + { + "epoch": 4.48, + "learning_rate": 3.176623501510259e-06, + "loss": 0.9804, + "step": 60644 + }, + { + "epoch": 4.48, + "learning_rate": 3.176331916961355e-06, + "loss": 1.0076, + "step": 60645 + }, + { + "epoch": 4.48, + "learning_rate": 3.176040343268808e-06, + "loss": 0.9924, + "step": 60646 + }, + { + "epoch": 4.48, + "learning_rate": 3.175748780433081e-06, + "loss": 1.0617, + "step": 60647 + }, + { + "epoch": 4.48, + "learning_rate": 3.1754572284546426e-06, + "loss": 1.047, + "step": 60648 + }, + { + "epoch": 4.48, + "learning_rate": 3.1751656873339544e-06, + "loss": 0.9558, + "step": 60649 + }, + { + "epoch": 4.48, + "learning_rate": 3.1748741570714803e-06, + "loss": 0.8722, + "step": 60650 + }, + { + "epoch": 4.48, + "learning_rate": 3.17458263766768e-06, + "loss": 1.0187, + "step": 60651 + }, + { + "epoch": 4.48, + "learning_rate": 3.1742911291230236e-06, + "loss": 0.9219, + "step": 60652 + }, + { + "epoch": 4.48, + "learning_rate": 3.1739996314379775e-06, + "loss": 0.9419, + "step": 60653 + }, + { + "epoch": 4.48, + "learning_rate": 3.1737081446129923e-06, + "loss": 1.0455, + "step": 60654 + }, + { + "epoch": 4.48, + "learning_rate": 3.1734166686485434e-06, + "loss": 1.0323, + "step": 60655 + }, + { + "epoch": 4.48, + "learning_rate": 3.1731252035450875e-06, + "loss": 1.0286, + "step": 60656 + }, + { + "epoch": 4.48, + "learning_rate": 3.172833749303099e-06, + "loss": 1.0707, + "step": 60657 + }, + { + "epoch": 4.48, + "learning_rate": 3.1725423059230276e-06, + "loss": 0.9923, + "step": 60658 + }, + { + "epoch": 4.48, + "learning_rate": 3.172250873405347e-06, + "loss": 0.9563, + "step": 60659 + }, + { + "epoch": 4.48, + "learning_rate": 3.171959451750514e-06, + "loss": 1.0104, + "step": 60660 + }, + { + "epoch": 4.48, + "learning_rate": 3.1716680409590016e-06, + "loss": 0.9957, + "step": 60661 + }, + { + "epoch": 4.48, + "learning_rate": 3.1713766410312654e-06, + "loss": 0.9025, + "step": 60662 + }, + { + "epoch": 4.48, + "learning_rate": 3.171085251967768e-06, + "loss": 1.0563, + "step": 60663 + }, + { + "epoch": 4.48, + "learning_rate": 3.170793873768979e-06, + "loss": 0.9293, + "step": 60664 + }, + { + "epoch": 4.48, + "learning_rate": 3.17050250643536e-06, + "loss": 0.9261, + "step": 60665 + }, + { + "epoch": 4.48, + "learning_rate": 3.170211149967373e-06, + "loss": 0.9484, + "step": 60666 + }, + { + "epoch": 4.48, + "learning_rate": 3.16991980436548e-06, + "loss": 0.9704, + "step": 60667 + }, + { + "epoch": 4.48, + "learning_rate": 3.1696284696301493e-06, + "loss": 1.071, + "step": 60668 + }, + { + "epoch": 4.48, + "learning_rate": 3.1693371457618416e-06, + "loss": 0.9593, + "step": 60669 + }, + { + "epoch": 4.48, + "learning_rate": 3.1690458327610217e-06, + "loss": 1.0747, + "step": 60670 + }, + { + "epoch": 4.48, + "learning_rate": 3.168754530628152e-06, + "loss": 0.9581, + "step": 60671 + }, + { + "epoch": 4.48, + "learning_rate": 3.1684632393636926e-06, + "loss": 0.9608, + "step": 60672 + }, + { + "epoch": 4.48, + "learning_rate": 3.1681719589681135e-06, + "loss": 1.0355, + "step": 60673 + }, + { + "epoch": 4.48, + "learning_rate": 3.167880689441876e-06, + "loss": 1.0226, + "step": 60674 + }, + { + "epoch": 4.48, + "learning_rate": 3.1675894307854417e-06, + "loss": 0.9734, + "step": 60675 + }, + { + "epoch": 4.48, + "learning_rate": 3.167298182999271e-06, + "loss": 0.9813, + "step": 60676 + }, + { + "epoch": 4.48, + "learning_rate": 3.167006946083836e-06, + "loss": 1.0339, + "step": 60677 + }, + { + "epoch": 4.48, + "learning_rate": 3.1667157200395937e-06, + "loss": 0.9823, + "step": 60678 + }, + { + "epoch": 4.48, + "learning_rate": 3.1664245048670096e-06, + "loss": 1.0738, + "step": 60679 + }, + { + "epoch": 4.48, + "learning_rate": 3.166133300566543e-06, + "loss": 1.0205, + "step": 60680 + }, + { + "epoch": 4.48, + "learning_rate": 3.1658421071386667e-06, + "loss": 1.0832, + "step": 60681 + }, + { + "epoch": 4.48, + "learning_rate": 3.1655509245838358e-06, + "loss": 1.0079, + "step": 60682 + }, + { + "epoch": 4.48, + "learning_rate": 3.1652597529025108e-06, + "loss": 0.9468, + "step": 60683 + }, + { + "epoch": 4.48, + "learning_rate": 3.1649685920951643e-06, + "loss": 0.9362, + "step": 60684 + }, + { + "epoch": 4.48, + "learning_rate": 3.164677442162254e-06, + "loss": 1.087, + "step": 60685 + }, + { + "epoch": 4.48, + "learning_rate": 3.1643863031042453e-06, + "loss": 0.8896, + "step": 60686 + }, + { + "epoch": 4.48, + "learning_rate": 3.1640951749215955e-06, + "loss": 0.9097, + "step": 60687 + }, + { + "epoch": 4.48, + "learning_rate": 3.1638040576147764e-06, + "loss": 1.0236, + "step": 60688 + }, + { + "epoch": 4.48, + "learning_rate": 3.163512951184248e-06, + "loss": 1.0079, + "step": 60689 + }, + { + "epoch": 4.48, + "learning_rate": 3.163221855630472e-06, + "loss": 1.0211, + "step": 60690 + }, + { + "epoch": 4.48, + "learning_rate": 3.1629307709539114e-06, + "loss": 0.9412, + "step": 60691 + }, + { + "epoch": 4.48, + "learning_rate": 3.1626396971550277e-06, + "loss": 1.0618, + "step": 60692 + }, + { + "epoch": 4.48, + "learning_rate": 3.162348634234289e-06, + "loss": 1.0948, + "step": 60693 + }, + { + "epoch": 4.48, + "learning_rate": 3.162057582192156e-06, + "loss": 1.0881, + "step": 60694 + }, + { + "epoch": 4.48, + "learning_rate": 3.1617665410290922e-06, + "loss": 0.9271, + "step": 60695 + }, + { + "epoch": 4.48, + "learning_rate": 3.1614755107455563e-06, + "loss": 1.0115, + "step": 60696 + }, + { + "epoch": 4.48, + "learning_rate": 3.161184491342019e-06, + "loss": 1.0478, + "step": 60697 + }, + { + "epoch": 4.48, + "learning_rate": 3.1608934828189385e-06, + "loss": 1.0528, + "step": 60698 + }, + { + "epoch": 4.49, + "learning_rate": 3.160602485176779e-06, + "loss": 0.9582, + "step": 60699 + }, + { + "epoch": 4.49, + "learning_rate": 3.1603114984160033e-06, + "loss": 0.9716, + "step": 60700 + }, + { + "epoch": 4.49, + "learning_rate": 3.1600205225370705e-06, + "loss": 0.9741, + "step": 60701 + }, + { + "epoch": 4.49, + "learning_rate": 3.159729557540453e-06, + "loss": 0.9501, + "step": 60702 + }, + { + "epoch": 4.49, + "learning_rate": 3.1594386034266024e-06, + "loss": 1.0015, + "step": 60703 + }, + { + "epoch": 4.49, + "learning_rate": 3.1591476601959903e-06, + "loss": 0.9853, + "step": 60704 + }, + { + "epoch": 4.49, + "learning_rate": 3.1588567278490724e-06, + "loss": 0.9402, + "step": 60705 + }, + { + "epoch": 4.49, + "learning_rate": 3.1585658063863224e-06, + "loss": 1.0792, + "step": 60706 + }, + { + "epoch": 4.49, + "learning_rate": 3.1582748958081897e-06, + "loss": 0.9477, + "step": 60707 + }, + { + "epoch": 4.49, + "learning_rate": 3.1579839961151483e-06, + "loss": 0.9592, + "step": 60708 + }, + { + "epoch": 4.49, + "learning_rate": 3.1576931073076554e-06, + "loss": 0.9249, + "step": 60709 + }, + { + "epoch": 4.49, + "learning_rate": 3.157402229386175e-06, + "loss": 0.9082, + "step": 60710 + }, + { + "epoch": 4.49, + "learning_rate": 3.1571113623511694e-06, + "loss": 1.0253, + "step": 60711 + }, + { + "epoch": 4.49, + "learning_rate": 3.1568205062030987e-06, + "loss": 1.0188, + "step": 60712 + }, + { + "epoch": 4.49, + "learning_rate": 3.1565296609424324e-06, + "loss": 0.9163, + "step": 60713 + }, + { + "epoch": 4.49, + "learning_rate": 3.1562388265696288e-06, + "loss": 1.0126, + "step": 60714 + }, + { + "epoch": 4.49, + "learning_rate": 3.155948003085152e-06, + "loss": 1.0173, + "step": 60715 + }, + { + "epoch": 4.49, + "learning_rate": 3.1556571904894593e-06, + "loss": 0.9391, + "step": 60716 + }, + { + "epoch": 4.49, + "learning_rate": 3.1553663887830235e-06, + "loss": 1.0016, + "step": 60717 + }, + { + "epoch": 4.49, + "learning_rate": 3.1550755979662994e-06, + "loss": 0.9843, + "step": 60718 + }, + { + "epoch": 4.49, + "learning_rate": 3.1547848180397543e-06, + "loss": 1.0664, + "step": 60719 + }, + { + "epoch": 4.49, + "learning_rate": 3.154494049003848e-06, + "loss": 0.9509, + "step": 60720 + }, + { + "epoch": 4.49, + "learning_rate": 3.1542032908590393e-06, + "loss": 1.0378, + "step": 60721 + }, + { + "epoch": 4.49, + "learning_rate": 3.1539125436058025e-06, + "loss": 1.0193, + "step": 60722 + }, + { + "epoch": 4.49, + "learning_rate": 3.153621807244587e-06, + "loss": 1.0035, + "step": 60723 + }, + { + "epoch": 4.49, + "learning_rate": 3.1533310817758635e-06, + "loss": 1.0177, + "step": 60724 + }, + { + "epoch": 4.49, + "learning_rate": 3.1530403672000896e-06, + "loss": 0.9667, + "step": 60725 + }, + { + "epoch": 4.49, + "learning_rate": 3.1527496635177368e-06, + "loss": 1.0386, + "step": 60726 + }, + { + "epoch": 4.49, + "learning_rate": 3.1524589707292584e-06, + "loss": 0.9671, + "step": 60727 + }, + { + "epoch": 4.49, + "learning_rate": 3.1521682888351167e-06, + "loss": 0.9955, + "step": 60728 + }, + { + "epoch": 4.49, + "learning_rate": 3.15187761783578e-06, + "loss": 0.9824, + "step": 60729 + }, + { + "epoch": 4.49, + "learning_rate": 3.151586957731708e-06, + "loss": 0.8972, + "step": 60730 + }, + { + "epoch": 4.49, + "learning_rate": 3.151296308523364e-06, + "loss": 0.985, + "step": 60731 + }, + { + "epoch": 4.49, + "learning_rate": 3.1510056702112056e-06, + "loss": 1.0418, + "step": 60732 + }, + { + "epoch": 4.49, + "learning_rate": 3.1507150427957022e-06, + "loss": 1.0289, + "step": 60733 + }, + { + "epoch": 4.49, + "learning_rate": 3.1504244262773133e-06, + "loss": 1.0178, + "step": 60734 + }, + { + "epoch": 4.49, + "learning_rate": 3.1501338206565013e-06, + "loss": 1.0076, + "step": 60735 + }, + { + "epoch": 4.49, + "learning_rate": 3.1498432259337254e-06, + "loss": 1.0681, + "step": 60736 + }, + { + "epoch": 4.49, + "learning_rate": 3.1495526421094547e-06, + "loss": 0.99, + "step": 60737 + }, + { + "epoch": 4.49, + "learning_rate": 3.149262069184147e-06, + "loss": 1.1226, + "step": 60738 + }, + { + "epoch": 4.49, + "learning_rate": 3.148971507158265e-06, + "loss": 1.0415, + "step": 60739 + }, + { + "epoch": 4.49, + "learning_rate": 3.148680956032272e-06, + "loss": 1.0162, + "step": 60740 + }, + { + "epoch": 4.49, + "learning_rate": 3.148390415806626e-06, + "loss": 0.8958, + "step": 60741 + }, + { + "epoch": 4.49, + "learning_rate": 3.148099886481796e-06, + "loss": 1.0839, + "step": 60742 + }, + { + "epoch": 4.49, + "learning_rate": 3.14780936805824e-06, + "loss": 0.9566, + "step": 60743 + }, + { + "epoch": 4.49, + "learning_rate": 3.1475188605364227e-06, + "loss": 1.0584, + "step": 60744 + }, + { + "epoch": 4.49, + "learning_rate": 3.147228363916801e-06, + "loss": 0.9688, + "step": 60745 + }, + { + "epoch": 4.49, + "learning_rate": 3.146937878199845e-06, + "loss": 1.0141, + "step": 60746 + }, + { + "epoch": 4.49, + "learning_rate": 3.146647403386015e-06, + "loss": 0.9723, + "step": 60747 + }, + { + "epoch": 4.49, + "learning_rate": 3.1463569394757655e-06, + "loss": 0.9822, + "step": 60748 + }, + { + "epoch": 4.49, + "learning_rate": 3.1460664864695667e-06, + "loss": 1.0249, + "step": 60749 + }, + { + "epoch": 4.49, + "learning_rate": 3.1457760443678744e-06, + "loss": 0.9622, + "step": 60750 + }, + { + "epoch": 4.49, + "learning_rate": 3.145485613171161e-06, + "loss": 0.9786, + "step": 60751 + }, + { + "epoch": 4.49, + "learning_rate": 3.145195192879876e-06, + "loss": 1.0493, + "step": 60752 + }, + { + "epoch": 4.49, + "learning_rate": 3.14490478349449e-06, + "loss": 0.8869, + "step": 60753 + }, + { + "epoch": 4.49, + "learning_rate": 3.14461438501546e-06, + "loss": 0.9993, + "step": 60754 + }, + { + "epoch": 4.49, + "learning_rate": 3.144323997443256e-06, + "loss": 1.001, + "step": 60755 + }, + { + "epoch": 4.49, + "learning_rate": 3.1440336207783317e-06, + "loss": 1.053, + "step": 60756 + }, + { + "epoch": 4.49, + "learning_rate": 3.143743255021148e-06, + "loss": 1.0085, + "step": 60757 + }, + { + "epoch": 4.49, + "learning_rate": 3.1434529001721746e-06, + "loss": 0.9149, + "step": 60758 + }, + { + "epoch": 4.49, + "learning_rate": 3.143162556231869e-06, + "loss": 1.0072, + "step": 60759 + }, + { + "epoch": 4.49, + "learning_rate": 3.1428722232006936e-06, + "loss": 0.9687, + "step": 60760 + }, + { + "epoch": 4.49, + "learning_rate": 3.1425819010791083e-06, + "loss": 0.8885, + "step": 60761 + }, + { + "epoch": 4.49, + "learning_rate": 3.142291589867579e-06, + "loss": 0.9551, + "step": 60762 + }, + { + "epoch": 4.49, + "learning_rate": 3.1420012895665664e-06, + "loss": 1.0091, + "step": 60763 + }, + { + "epoch": 4.49, + "learning_rate": 3.141711000176532e-06, + "loss": 0.9755, + "step": 60764 + }, + { + "epoch": 4.49, + "learning_rate": 3.1414207216979355e-06, + "loss": 0.8493, + "step": 60765 + }, + { + "epoch": 4.49, + "learning_rate": 3.141130454131238e-06, + "loss": 1.0702, + "step": 60766 + }, + { + "epoch": 4.49, + "learning_rate": 3.140840197476911e-06, + "loss": 1.0886, + "step": 60767 + }, + { + "epoch": 4.49, + "learning_rate": 3.140549951735403e-06, + "loss": 0.9541, + "step": 60768 + }, + { + "epoch": 4.49, + "learning_rate": 3.140259716907185e-06, + "loss": 0.9255, + "step": 60769 + }, + { + "epoch": 4.49, + "learning_rate": 3.1399694929927115e-06, + "loss": 1.0163, + "step": 60770 + }, + { + "epoch": 4.49, + "learning_rate": 3.1396792799924546e-06, + "loss": 0.923, + "step": 60771 + }, + { + "epoch": 4.49, + "learning_rate": 3.1393890779068636e-06, + "loss": 1.1244, + "step": 60772 + }, + { + "epoch": 4.49, + "learning_rate": 3.13909888673641e-06, + "loss": 0.9585, + "step": 60773 + }, + { + "epoch": 4.49, + "learning_rate": 3.138808706481552e-06, + "loss": 0.953, + "step": 60774 + }, + { + "epoch": 4.49, + "learning_rate": 3.1385185371427516e-06, + "loss": 0.964, + "step": 60775 + }, + { + "epoch": 4.49, + "learning_rate": 3.1382283787204694e-06, + "loss": 1.1521, + "step": 60776 + }, + { + "epoch": 4.49, + "learning_rate": 3.1379382312151654e-06, + "loss": 1.0846, + "step": 60777 + }, + { + "epoch": 4.49, + "learning_rate": 3.137648094627306e-06, + "loss": 1.0455, + "step": 60778 + }, + { + "epoch": 4.49, + "learning_rate": 3.1373579689573495e-06, + "loss": 1.0336, + "step": 60779 + }, + { + "epoch": 4.49, + "learning_rate": 3.1370678542057588e-06, + "loss": 1.1287, + "step": 60780 + }, + { + "epoch": 4.49, + "learning_rate": 3.1367777503729924e-06, + "loss": 1.0304, + "step": 60781 + }, + { + "epoch": 4.49, + "learning_rate": 3.136487657459517e-06, + "loss": 0.9821, + "step": 60782 + }, + { + "epoch": 4.49, + "learning_rate": 3.1361975754657914e-06, + "loss": 0.9773, + "step": 60783 + }, + { + "epoch": 4.49, + "learning_rate": 3.135907504392277e-06, + "loss": 1.0039, + "step": 60784 + }, + { + "epoch": 4.49, + "learning_rate": 3.135617444239437e-06, + "loss": 0.9379, + "step": 60785 + }, + { + "epoch": 4.49, + "learning_rate": 3.135327395007727e-06, + "loss": 0.9722, + "step": 60786 + }, + { + "epoch": 4.49, + "learning_rate": 3.135037356697617e-06, + "loss": 0.9777, + "step": 60787 + }, + { + "epoch": 4.49, + "learning_rate": 3.1347473293095633e-06, + "loss": 0.9508, + "step": 60788 + }, + { + "epoch": 4.49, + "learning_rate": 3.1344573128440293e-06, + "loss": 1.0993, + "step": 60789 + }, + { + "epoch": 4.49, + "learning_rate": 3.134167307301471e-06, + "loss": 0.967, + "step": 60790 + }, + { + "epoch": 4.49, + "learning_rate": 3.1338773126823584e-06, + "loss": 1.1222, + "step": 60791 + }, + { + "epoch": 4.49, + "learning_rate": 3.133587328987149e-06, + "loss": 0.9684, + "step": 60792 + }, + { + "epoch": 4.49, + "learning_rate": 3.1332973562163037e-06, + "loss": 0.949, + "step": 60793 + }, + { + "epoch": 4.49, + "learning_rate": 3.1330073943702832e-06, + "loss": 0.895, + "step": 60794 + }, + { + "epoch": 4.49, + "learning_rate": 3.132717443449548e-06, + "loss": 0.9611, + "step": 60795 + }, + { + "epoch": 4.49, + "learning_rate": 3.132427503454566e-06, + "loss": 0.9458, + "step": 60796 + }, + { + "epoch": 4.49, + "learning_rate": 3.1321375743857885e-06, + "loss": 0.9856, + "step": 60797 + }, + { + "epoch": 4.49, + "learning_rate": 3.131847656243684e-06, + "loss": 0.9776, + "step": 60798 + }, + { + "epoch": 4.49, + "learning_rate": 3.1315577490287087e-06, + "loss": 0.9374, + "step": 60799 + }, + { + "epoch": 4.49, + "learning_rate": 3.1312678527413332e-06, + "loss": 0.967, + "step": 60800 + }, + { + "epoch": 4.49, + "learning_rate": 3.1309779673820063e-06, + "loss": 0.9648, + "step": 60801 + }, + { + "epoch": 4.49, + "learning_rate": 3.1306880929511985e-06, + "loss": 0.9207, + "step": 60802 + }, + { + "epoch": 4.49, + "learning_rate": 3.1303982294493675e-06, + "loss": 1.0026, + "step": 60803 + }, + { + "epoch": 4.49, + "learning_rate": 3.1301083768769737e-06, + "loss": 1.0389, + "step": 60804 + }, + { + "epoch": 4.49, + "learning_rate": 3.12981853523448e-06, + "loss": 1.1151, + "step": 60805 + }, + { + "epoch": 4.49, + "learning_rate": 3.1295287045223434e-06, + "loss": 1.0085, + "step": 60806 + }, + { + "epoch": 4.49, + "learning_rate": 3.1292388847410315e-06, + "loss": 0.9782, + "step": 60807 + }, + { + "epoch": 4.49, + "learning_rate": 3.1289490758910025e-06, + "loss": 0.9947, + "step": 60808 + }, + { + "epoch": 4.49, + "learning_rate": 3.1286592779727163e-06, + "loss": 1.086, + "step": 60809 + }, + { + "epoch": 4.49, + "learning_rate": 3.1283694909866325e-06, + "loss": 1.0892, + "step": 60810 + }, + { + "epoch": 4.49, + "learning_rate": 3.1280797149332178e-06, + "loss": 0.9709, + "step": 60811 + }, + { + "epoch": 4.49, + "learning_rate": 3.127789949812933e-06, + "loss": 0.9458, + "step": 60812 + }, + { + "epoch": 4.49, + "learning_rate": 3.1275001956262285e-06, + "loss": 1.0135, + "step": 60813 + }, + { + "epoch": 4.49, + "learning_rate": 3.127210452373577e-06, + "loss": 0.853, + "step": 60814 + }, + { + "epoch": 4.49, + "learning_rate": 3.126920720055433e-06, + "loss": 1.0404, + "step": 60815 + }, + { + "epoch": 4.49, + "learning_rate": 3.1266309986722653e-06, + "loss": 1.0059, + "step": 60816 + }, + { + "epoch": 4.49, + "learning_rate": 3.1263412882245225e-06, + "loss": 1.0094, + "step": 60817 + }, + { + "epoch": 4.49, + "learning_rate": 3.1260515887126763e-06, + "loss": 1.0713, + "step": 60818 + }, + { + "epoch": 4.49, + "learning_rate": 3.125761900137181e-06, + "loss": 0.9594, + "step": 60819 + }, + { + "epoch": 4.49, + "learning_rate": 3.125472222498506e-06, + "loss": 0.9925, + "step": 60820 + }, + { + "epoch": 4.49, + "learning_rate": 3.1251825557971026e-06, + "loss": 1.0752, + "step": 60821 + }, + { + "epoch": 4.49, + "learning_rate": 3.124892900033433e-06, + "loss": 0.9239, + "step": 60822 + }, + { + "epoch": 4.49, + "learning_rate": 3.124603255207962e-06, + "loss": 0.9875, + "step": 60823 + }, + { + "epoch": 4.49, + "learning_rate": 3.12431362132115e-06, + "loss": 1.0458, + "step": 60824 + }, + { + "epoch": 4.49, + "learning_rate": 3.1240239983734578e-06, + "loss": 1.0709, + "step": 60825 + }, + { + "epoch": 4.49, + "learning_rate": 3.1237343863653404e-06, + "loss": 0.9947, + "step": 60826 + }, + { + "epoch": 4.49, + "learning_rate": 3.123444785297266e-06, + "loss": 1.0525, + "step": 60827 + }, + { + "epoch": 4.49, + "learning_rate": 3.1231551951696935e-06, + "loss": 1.0048, + "step": 60828 + }, + { + "epoch": 4.49, + "learning_rate": 3.1228656159830826e-06, + "loss": 1.0472, + "step": 60829 + }, + { + "epoch": 4.49, + "learning_rate": 3.122576047737894e-06, + "loss": 0.9713, + "step": 60830 + }, + { + "epoch": 4.49, + "learning_rate": 3.1222864904345864e-06, + "loss": 0.9851, + "step": 60831 + }, + { + "epoch": 4.49, + "learning_rate": 3.1219969440736244e-06, + "loss": 1.0385, + "step": 60832 + }, + { + "epoch": 4.49, + "learning_rate": 3.1217074086554666e-06, + "loss": 1.0107, + "step": 60833 + }, + { + "epoch": 4.49, + "learning_rate": 3.1214178841805743e-06, + "loss": 1.0702, + "step": 60834 + }, + { + "epoch": 4.5, + "learning_rate": 3.1211283706494056e-06, + "loss": 1.0503, + "step": 60835 + }, + { + "epoch": 4.5, + "learning_rate": 3.1208388680624244e-06, + "loss": 0.9409, + "step": 60836 + }, + { + "epoch": 4.5, + "learning_rate": 3.1205493764200923e-06, + "loss": 0.9286, + "step": 60837 + }, + { + "epoch": 4.5, + "learning_rate": 3.120259895722866e-06, + "loss": 0.9556, + "step": 60838 + }, + { + "epoch": 4.5, + "learning_rate": 3.119970425971206e-06, + "loss": 1.1018, + "step": 60839 + }, + { + "epoch": 4.5, + "learning_rate": 3.119680967165577e-06, + "loss": 0.9601, + "step": 60840 + }, + { + "epoch": 4.5, + "learning_rate": 3.1193915193064404e-06, + "loss": 0.9274, + "step": 60841 + }, + { + "epoch": 4.5, + "learning_rate": 3.1191020823942465e-06, + "loss": 1.0762, + "step": 60842 + }, + { + "epoch": 4.5, + "learning_rate": 3.1188126564294663e-06, + "loss": 0.9628, + "step": 60843 + }, + { + "epoch": 4.5, + "learning_rate": 3.1185232414125578e-06, + "loss": 1.0895, + "step": 60844 + }, + { + "epoch": 4.5, + "learning_rate": 3.118233837343979e-06, + "loss": 1.0427, + "step": 60845 + }, + { + "epoch": 4.5, + "learning_rate": 3.117944444224189e-06, + "loss": 0.9717, + "step": 60846 + }, + { + "epoch": 4.5, + "learning_rate": 3.1176550620536537e-06, + "loss": 0.9689, + "step": 60847 + }, + { + "epoch": 4.5, + "learning_rate": 3.117365690832832e-06, + "loss": 0.9506, + "step": 60848 + }, + { + "epoch": 4.5, + "learning_rate": 3.117076330562182e-06, + "loss": 0.9571, + "step": 60849 + }, + { + "epoch": 4.5, + "learning_rate": 3.1167869812421647e-06, + "loss": 1.0169, + "step": 60850 + }, + { + "epoch": 4.5, + "learning_rate": 3.1164976428732387e-06, + "loss": 1.0058, + "step": 60851 + }, + { + "epoch": 4.5, + "learning_rate": 3.1162083154558686e-06, + "loss": 1.0988, + "step": 60852 + }, + { + "epoch": 4.5, + "learning_rate": 3.1159189989905137e-06, + "loss": 1.0164, + "step": 60853 + }, + { + "epoch": 4.5, + "learning_rate": 3.115629693477632e-06, + "loss": 0.9574, + "step": 60854 + }, + { + "epoch": 4.5, + "learning_rate": 3.115340398917682e-06, + "loss": 0.9495, + "step": 60855 + }, + { + "epoch": 4.5, + "learning_rate": 3.1150511153111307e-06, + "loss": 1.0377, + "step": 60856 + }, + { + "epoch": 4.5, + "learning_rate": 3.114761842658435e-06, + "loss": 1.0047, + "step": 60857 + }, + { + "epoch": 4.5, + "learning_rate": 3.114472580960054e-06, + "loss": 1.0371, + "step": 60858 + }, + { + "epoch": 4.5, + "learning_rate": 3.1141833302164483e-06, + "loss": 1.0996, + "step": 60859 + }, + { + "epoch": 4.5, + "learning_rate": 3.1138940904280755e-06, + "loss": 0.9227, + "step": 60860 + }, + { + "epoch": 4.5, + "learning_rate": 3.113604861595405e-06, + "loss": 0.9518, + "step": 60861 + }, + { + "epoch": 4.5, + "learning_rate": 3.113315643718885e-06, + "loss": 1.0678, + "step": 60862 + }, + { + "epoch": 4.5, + "learning_rate": 3.1130264367989836e-06, + "loss": 0.9673, + "step": 60863 + }, + { + "epoch": 4.5, + "learning_rate": 3.1127372408361556e-06, + "loss": 1.0156, + "step": 60864 + }, + { + "epoch": 4.5, + "learning_rate": 3.1124480558308703e-06, + "loss": 1.0144, + "step": 60865 + }, + { + "epoch": 4.5, + "learning_rate": 3.1121588817835756e-06, + "loss": 0.908, + "step": 60866 + }, + { + "epoch": 4.5, + "learning_rate": 3.11186971869474e-06, + "loss": 0.947, + "step": 60867 + }, + { + "epoch": 4.5, + "learning_rate": 3.1115805665648223e-06, + "loss": 0.888, + "step": 60868 + }, + { + "epoch": 4.5, + "learning_rate": 3.1112914253942806e-06, + "loss": 0.9596, + "step": 60869 + }, + { + "epoch": 4.5, + "learning_rate": 3.111002295183576e-06, + "loss": 1.0226, + "step": 60870 + }, + { + "epoch": 4.5, + "learning_rate": 3.1107131759331645e-06, + "loss": 0.9433, + "step": 60871 + }, + { + "epoch": 4.5, + "learning_rate": 3.110424067643514e-06, + "loss": 0.9573, + "step": 60872 + }, + { + "epoch": 4.5, + "learning_rate": 3.1101349703150794e-06, + "loss": 1.048, + "step": 60873 + }, + { + "epoch": 4.5, + "learning_rate": 3.1098458839483215e-06, + "loss": 0.9157, + "step": 60874 + }, + { + "epoch": 4.5, + "learning_rate": 3.1095568085436968e-06, + "loss": 0.9869, + "step": 60875 + }, + { + "epoch": 4.5, + "learning_rate": 3.1092677441016727e-06, + "loss": 1.0188, + "step": 60876 + }, + { + "epoch": 4.5, + "learning_rate": 3.108978690622705e-06, + "loss": 1.0391, + "step": 60877 + }, + { + "epoch": 4.5, + "learning_rate": 3.1086896481072527e-06, + "loss": 0.9257, + "step": 60878 + }, + { + "epoch": 4.5, + "learning_rate": 3.1084006165557767e-06, + "loss": 1.0152, + "step": 60879 + }, + { + "epoch": 4.5, + "learning_rate": 3.1081115959687335e-06, + "loss": 0.9784, + "step": 60880 + }, + { + "epoch": 4.5, + "learning_rate": 3.1078225863465903e-06, + "loss": 1.0288, + "step": 60881 + }, + { + "epoch": 4.5, + "learning_rate": 3.1075335876898015e-06, + "loss": 0.9733, + "step": 60882 + }, + { + "epoch": 4.5, + "learning_rate": 3.1072445999988286e-06, + "loss": 0.9864, + "step": 60883 + }, + { + "epoch": 4.5, + "learning_rate": 3.1069556232741284e-06, + "loss": 0.9757, + "step": 60884 + }, + { + "epoch": 4.5, + "learning_rate": 3.1066666575161665e-06, + "loss": 0.958, + "step": 60885 + }, + { + "epoch": 4.5, + "learning_rate": 3.106377702725402e-06, + "loss": 1.0427, + "step": 60886 + }, + { + "epoch": 4.5, + "learning_rate": 3.1060887589022848e-06, + "loss": 0.9337, + "step": 60887 + }, + { + "epoch": 4.5, + "learning_rate": 3.1057998260472856e-06, + "loss": 1.0132, + "step": 60888 + }, + { + "epoch": 4.5, + "learning_rate": 3.10551090416086e-06, + "loss": 0.9945, + "step": 60889 + }, + { + "epoch": 4.5, + "learning_rate": 3.1052219932434677e-06, + "loss": 0.9761, + "step": 60890 + }, + { + "epoch": 4.5, + "learning_rate": 3.104933093295567e-06, + "loss": 0.9412, + "step": 60891 + }, + { + "epoch": 4.5, + "learning_rate": 3.1046442043176206e-06, + "loss": 0.9591, + "step": 60892 + }, + { + "epoch": 4.5, + "learning_rate": 3.1043553263100867e-06, + "loss": 0.9509, + "step": 60893 + }, + { + "epoch": 4.5, + "learning_rate": 3.104066459273426e-06, + "loss": 1.0552, + "step": 60894 + }, + { + "epoch": 4.5, + "learning_rate": 3.1037776032080923e-06, + "loss": 0.9983, + "step": 60895 + }, + { + "epoch": 4.5, + "learning_rate": 3.1034887581145533e-06, + "loss": 1.0399, + "step": 60896 + }, + { + "epoch": 4.5, + "learning_rate": 3.103199923993265e-06, + "loss": 0.986, + "step": 60897 + }, + { + "epoch": 4.5, + "learning_rate": 3.1029111008446875e-06, + "loss": 0.8831, + "step": 60898 + }, + { + "epoch": 4.5, + "learning_rate": 3.1026222886692802e-06, + "loss": 0.9449, + "step": 60899 + }, + { + "epoch": 4.5, + "learning_rate": 3.102333487467498e-06, + "loss": 1.0615, + "step": 60900 + }, + { + "epoch": 4.5, + "learning_rate": 3.1020446972398077e-06, + "loss": 1.0675, + "step": 60901 + }, + { + "epoch": 4.5, + "learning_rate": 3.1017559179866665e-06, + "loss": 0.9831, + "step": 60902 + }, + { + "epoch": 4.5, + "learning_rate": 3.1014671497085325e-06, + "loss": 1.0391, + "step": 60903 + }, + { + "epoch": 4.5, + "learning_rate": 3.1011783924058614e-06, + "loss": 0.9882, + "step": 60904 + }, + { + "epoch": 4.5, + "learning_rate": 3.1008896460791207e-06, + "loss": 0.9857, + "step": 60905 + }, + { + "epoch": 4.5, + "learning_rate": 3.1006009107287695e-06, + "loss": 0.9517, + "step": 60906 + }, + { + "epoch": 4.5, + "learning_rate": 3.1003121863552564e-06, + "loss": 1.0472, + "step": 60907 + }, + { + "epoch": 4.5, + "learning_rate": 3.1000234729590518e-06, + "loss": 0.9139, + "step": 60908 + }, + { + "epoch": 4.5, + "learning_rate": 3.0997347705406076e-06, + "loss": 1.0043, + "step": 60909 + }, + { + "epoch": 4.5, + "learning_rate": 3.099446079100393e-06, + "loss": 0.9555, + "step": 60910 + }, + { + "epoch": 4.5, + "learning_rate": 3.0991573986388544e-06, + "loss": 0.9657, + "step": 60911 + }, + { + "epoch": 4.5, + "learning_rate": 3.098868729156461e-06, + "loss": 0.9897, + "step": 60912 + }, + { + "epoch": 4.5, + "learning_rate": 3.0985800706536653e-06, + "loss": 0.9718, + "step": 60913 + }, + { + "epoch": 4.5, + "learning_rate": 3.098291423130937e-06, + "loss": 1.0882, + "step": 60914 + }, + { + "epoch": 4.5, + "learning_rate": 3.0980027865887243e-06, + "loss": 0.9776, + "step": 60915 + }, + { + "epoch": 4.5, + "learning_rate": 3.0977141610274875e-06, + "loss": 1.1528, + "step": 60916 + }, + { + "epoch": 4.5, + "learning_rate": 3.097425546447692e-06, + "loss": 0.9981, + "step": 60917 + }, + { + "epoch": 4.5, + "learning_rate": 3.097136942849794e-06, + "loss": 0.9784, + "step": 60918 + }, + { + "epoch": 4.5, + "learning_rate": 3.0968483502342516e-06, + "loss": 1.0907, + "step": 60919 + }, + { + "epoch": 4.5, + "learning_rate": 3.0965597686015224e-06, + "loss": 1.1905, + "step": 60920 + }, + { + "epoch": 4.5, + "learning_rate": 3.096271197952071e-06, + "loss": 1.0858, + "step": 60921 + }, + { + "epoch": 4.5, + "learning_rate": 3.095982638286352e-06, + "loss": 1.0824, + "step": 60922 + }, + { + "epoch": 4.5, + "learning_rate": 3.095694089604827e-06, + "loss": 0.944, + "step": 60923 + }, + { + "epoch": 4.5, + "learning_rate": 3.095405551907954e-06, + "loss": 0.9637, + "step": 60924 + }, + { + "epoch": 4.5, + "learning_rate": 3.0951170251961894e-06, + "loss": 1.1003, + "step": 60925 + }, + { + "epoch": 4.5, + "learning_rate": 3.094828509469997e-06, + "loss": 0.949, + "step": 60926 + }, + { + "epoch": 4.5, + "learning_rate": 3.0945400047298337e-06, + "loss": 0.8962, + "step": 60927 + }, + { + "epoch": 4.5, + "learning_rate": 3.0942515109761583e-06, + "loss": 1.0158, + "step": 60928 + }, + { + "epoch": 4.5, + "learning_rate": 3.093963028209427e-06, + "loss": 1.0345, + "step": 60929 + }, + { + "epoch": 4.5, + "learning_rate": 3.0936745564301053e-06, + "loss": 0.9051, + "step": 60930 + }, + { + "epoch": 4.5, + "learning_rate": 3.0933860956386476e-06, + "loss": 1.0619, + "step": 60931 + }, + { + "epoch": 4.5, + "learning_rate": 3.0930976458355145e-06, + "loss": 0.9768, + "step": 60932 + }, + { + "epoch": 4.5, + "learning_rate": 3.0928092070211634e-06, + "loss": 0.9385, + "step": 60933 + }, + { + "epoch": 4.5, + "learning_rate": 3.0925207791960552e-06, + "loss": 0.9403, + "step": 60934 + }, + { + "epoch": 4.5, + "learning_rate": 3.0922323623606477e-06, + "loss": 0.9659, + "step": 60935 + }, + { + "epoch": 4.5, + "learning_rate": 3.091943956515395e-06, + "loss": 0.9694, + "step": 60936 + }, + { + "epoch": 4.5, + "learning_rate": 3.0916555616607645e-06, + "loss": 1.0465, + "step": 60937 + }, + { + "epoch": 4.5, + "learning_rate": 3.091367177797212e-06, + "loss": 1.1263, + "step": 60938 + }, + { + "epoch": 4.5, + "learning_rate": 3.091078804925195e-06, + "loss": 0.9703, + "step": 60939 + }, + { + "epoch": 4.5, + "learning_rate": 3.0907904430451696e-06, + "loss": 0.9431, + "step": 60940 + }, + { + "epoch": 4.5, + "learning_rate": 3.0905020921576e-06, + "loss": 1.0778, + "step": 60941 + }, + { + "epoch": 4.5, + "learning_rate": 3.090213752262944e-06, + "loss": 1.014, + "step": 60942 + }, + { + "epoch": 4.5, + "learning_rate": 3.0899254233616583e-06, + "loss": 0.9968, + "step": 60943 + }, + { + "epoch": 4.5, + "learning_rate": 3.089637105454203e-06, + "loss": 0.999, + "step": 60944 + }, + { + "epoch": 4.5, + "learning_rate": 3.089348798541032e-06, + "loss": 1.0048, + "step": 60945 + }, + { + "epoch": 4.5, + "learning_rate": 3.0890605026226127e-06, + "loss": 1.0434, + "step": 60946 + }, + { + "epoch": 4.5, + "learning_rate": 3.0887722176993984e-06, + "loss": 1.0703, + "step": 60947 + }, + { + "epoch": 4.5, + "learning_rate": 3.0884839437718482e-06, + "loss": 1.0628, + "step": 60948 + }, + { + "epoch": 4.5, + "learning_rate": 3.088195680840419e-06, + "loss": 1.0427, + "step": 60949 + }, + { + "epoch": 4.5, + "learning_rate": 3.0879074289055743e-06, + "loss": 0.9439, + "step": 60950 + }, + { + "epoch": 4.5, + "learning_rate": 3.0876191879677696e-06, + "loss": 0.9627, + "step": 60951 + }, + { + "epoch": 4.5, + "learning_rate": 3.0873309580274645e-06, + "loss": 1.047, + "step": 60952 + }, + { + "epoch": 4.5, + "learning_rate": 3.0870427390851166e-06, + "loss": 0.8911, + "step": 60953 + }, + { + "epoch": 4.5, + "learning_rate": 3.086754531141182e-06, + "loss": 1.0227, + "step": 60954 + }, + { + "epoch": 4.5, + "learning_rate": 3.086466334196129e-06, + "loss": 1.0611, + "step": 60955 + }, + { + "epoch": 4.5, + "learning_rate": 3.0861781482504017e-06, + "loss": 1.0294, + "step": 60956 + }, + { + "epoch": 4.5, + "learning_rate": 3.085889973304469e-06, + "loss": 0.9622, + "step": 60957 + }, + { + "epoch": 4.5, + "learning_rate": 3.0856018093587835e-06, + "loss": 0.9472, + "step": 60958 + }, + { + "epoch": 4.5, + "learning_rate": 3.085313656413813e-06, + "loss": 1.0747, + "step": 60959 + }, + { + "epoch": 4.5, + "learning_rate": 3.0850255144700035e-06, + "loss": 0.9899, + "step": 60960 + }, + { + "epoch": 4.5, + "learning_rate": 3.0847373835278227e-06, + "loss": 0.9586, + "step": 60961 + }, + { + "epoch": 4.5, + "learning_rate": 3.0844492635877254e-06, + "loss": 0.9729, + "step": 60962 + }, + { + "epoch": 4.5, + "learning_rate": 3.0841611546501706e-06, + "loss": 0.9935, + "step": 60963 + }, + { + "epoch": 4.5, + "learning_rate": 3.083873056715616e-06, + "loss": 1.0572, + "step": 60964 + }, + { + "epoch": 4.5, + "learning_rate": 3.0835849697845177e-06, + "loss": 1.0281, + "step": 60965 + }, + { + "epoch": 4.5, + "learning_rate": 3.0832968938573403e-06, + "loss": 1.0831, + "step": 60966 + }, + { + "epoch": 4.5, + "learning_rate": 3.0830088289345384e-06, + "loss": 0.8641, + "step": 60967 + }, + { + "epoch": 4.5, + "learning_rate": 3.08272077501657e-06, + "loss": 1.0416, + "step": 60968 + }, + { + "epoch": 4.5, + "learning_rate": 3.0824327321038903e-06, + "loss": 0.9048, + "step": 60969 + }, + { + "epoch": 4.51, + "learning_rate": 3.0821447001969653e-06, + "loss": 0.8495, + "step": 60970 + }, + { + "epoch": 4.51, + "learning_rate": 3.081856679296249e-06, + "loss": 1.0313, + "step": 60971 + }, + { + "epoch": 4.51, + "learning_rate": 3.0815686694021996e-06, + "loss": 1.0059, + "step": 60972 + }, + { + "epoch": 4.51, + "learning_rate": 3.081280670515275e-06, + "loss": 0.9038, + "step": 60973 + }, + { + "epoch": 4.51, + "learning_rate": 3.0809926826359305e-06, + "loss": 1.032, + "step": 60974 + }, + { + "epoch": 4.51, + "learning_rate": 3.0807047057646323e-06, + "loss": 1.0919, + "step": 60975 + }, + { + "epoch": 4.51, + "learning_rate": 3.0804167399018326e-06, + "loss": 1.0696, + "step": 60976 + }, + { + "epoch": 4.51, + "learning_rate": 3.0801287850479913e-06, + "loss": 1.0325, + "step": 60977 + }, + { + "epoch": 4.51, + "learning_rate": 3.079840841203563e-06, + "loss": 0.9036, + "step": 60978 + }, + { + "epoch": 4.51, + "learning_rate": 3.0795529083690145e-06, + "loss": 0.8842, + "step": 60979 + }, + { + "epoch": 4.51, + "learning_rate": 3.0792649865447965e-06, + "loss": 1.1248, + "step": 60980 + }, + { + "epoch": 4.51, + "learning_rate": 3.0789770757313643e-06, + "loss": 0.8916, + "step": 60981 + }, + { + "epoch": 4.51, + "learning_rate": 3.0786891759291846e-06, + "loss": 1.0455, + "step": 60982 + }, + { + "epoch": 4.51, + "learning_rate": 3.0784012871387115e-06, + "loss": 1.0938, + "step": 60983 + }, + { + "epoch": 4.51, + "learning_rate": 3.0781134093604024e-06, + "loss": 1.1815, + "step": 60984 + }, + { + "epoch": 4.51, + "learning_rate": 3.0778255425947125e-06, + "loss": 0.9986, + "step": 60985 + }, + { + "epoch": 4.51, + "learning_rate": 3.077537686842107e-06, + "loss": 0.9456, + "step": 60986 + }, + { + "epoch": 4.51, + "learning_rate": 3.0772498421030396e-06, + "loss": 1.0369, + "step": 60987 + }, + { + "epoch": 4.51, + "learning_rate": 3.0769620083779683e-06, + "loss": 0.8836, + "step": 60988 + }, + { + "epoch": 4.51, + "learning_rate": 3.0766741856673522e-06, + "loss": 1.0121, + "step": 60989 + }, + { + "epoch": 4.51, + "learning_rate": 3.076386373971645e-06, + "loss": 1.0357, + "step": 60990 + }, + { + "epoch": 4.51, + "learning_rate": 3.0760985732913106e-06, + "loss": 1.046, + "step": 60991 + }, + { + "epoch": 4.51, + "learning_rate": 3.075810783626805e-06, + "loss": 0.9711, + "step": 60992 + }, + { + "epoch": 4.51, + "learning_rate": 3.0755230049785844e-06, + "loss": 0.9419, + "step": 60993 + }, + { + "epoch": 4.51, + "learning_rate": 3.075235237347106e-06, + "loss": 0.9952, + "step": 60994 + }, + { + "epoch": 4.51, + "learning_rate": 3.0749474807328316e-06, + "loss": 1.0004, + "step": 60995 + }, + { + "epoch": 4.51, + "learning_rate": 3.074659735136216e-06, + "loss": 1.0189, + "step": 60996 + }, + { + "epoch": 4.51, + "learning_rate": 3.074372000557718e-06, + "loss": 0.8891, + "step": 60997 + }, + { + "epoch": 4.51, + "learning_rate": 3.074084276997793e-06, + "loss": 1.0952, + "step": 60998 + }, + { + "epoch": 4.51, + "learning_rate": 3.0737965644569035e-06, + "loss": 1.015, + "step": 60999 + }, + { + "epoch": 4.51, + "learning_rate": 3.073508862935507e-06, + "loss": 1.0332, + "step": 61000 + }, + { + "epoch": 4.51, + "learning_rate": 3.0732211724340534e-06, + "loss": 1.0227, + "step": 61001 + }, + { + "epoch": 4.51, + "learning_rate": 3.0729334929530084e-06, + "loss": 1.0682, + "step": 61002 + }, + { + "epoch": 4.51, + "learning_rate": 3.0726458244928235e-06, + "loss": 1.0277, + "step": 61003 + }, + { + "epoch": 4.51, + "learning_rate": 3.072358167053967e-06, + "loss": 0.9301, + "step": 61004 + }, + { + "epoch": 4.51, + "learning_rate": 3.072070520636883e-06, + "loss": 1.0211, + "step": 61005 + }, + { + "epoch": 4.51, + "learning_rate": 3.0717828852420397e-06, + "loss": 0.8913, + "step": 61006 + }, + { + "epoch": 4.51, + "learning_rate": 3.071495260869886e-06, + "loss": 1.0677, + "step": 61007 + }, + { + "epoch": 4.51, + "learning_rate": 3.071207647520892e-06, + "loss": 0.9981, + "step": 61008 + }, + { + "epoch": 4.51, + "learning_rate": 3.0709200451955023e-06, + "loss": 0.9871, + "step": 61009 + }, + { + "epoch": 4.51, + "learning_rate": 3.0706324538941777e-06, + "loss": 0.9837, + "step": 61010 + }, + { + "epoch": 4.51, + "learning_rate": 3.0703448736173812e-06, + "loss": 1.0113, + "step": 61011 + }, + { + "epoch": 4.51, + "learning_rate": 3.070057304365567e-06, + "loss": 1.0212, + "step": 61012 + }, + { + "epoch": 4.51, + "learning_rate": 3.0697697461391916e-06, + "loss": 1.0203, + "step": 61013 + }, + { + "epoch": 4.51, + "learning_rate": 3.069482198938711e-06, + "loss": 1.0169, + "step": 61014 + }, + { + "epoch": 4.51, + "learning_rate": 3.0691946627645884e-06, + "loss": 0.9744, + "step": 61015 + }, + { + "epoch": 4.51, + "learning_rate": 3.0689071376172773e-06, + "loss": 0.9665, + "step": 61016 + }, + { + "epoch": 4.51, + "learning_rate": 3.0686196234972366e-06, + "loss": 0.9335, + "step": 61017 + }, + { + "epoch": 4.51, + "learning_rate": 3.068332120404922e-06, + "loss": 0.9805, + "step": 61018 + }, + { + "epoch": 4.51, + "learning_rate": 3.0680446283407882e-06, + "loss": 0.992, + "step": 61019 + }, + { + "epoch": 4.51, + "learning_rate": 3.0677571473053e-06, + "loss": 0.9808, + "step": 61020 + }, + { + "epoch": 4.51, + "learning_rate": 3.0674696772989123e-06, + "loss": 1.0057, + "step": 61021 + }, + { + "epoch": 4.51, + "learning_rate": 3.067182218322079e-06, + "loss": 0.9562, + "step": 61022 + }, + { + "epoch": 4.51, + "learning_rate": 3.066894770375258e-06, + "loss": 0.969, + "step": 61023 + }, + { + "epoch": 4.51, + "learning_rate": 3.0666073334589142e-06, + "loss": 0.9309, + "step": 61024 + }, + { + "epoch": 4.51, + "learning_rate": 3.0663199075734918e-06, + "loss": 1.0458, + "step": 61025 + }, + { + "epoch": 4.51, + "learning_rate": 3.0660324927194597e-06, + "loss": 0.8831, + "step": 61026 + }, + { + "epoch": 4.51, + "learning_rate": 3.0657450888972694e-06, + "loss": 0.9598, + "step": 61027 + }, + { + "epoch": 4.51, + "learning_rate": 3.06545769610738e-06, + "loss": 0.8658, + "step": 61028 + }, + { + "epoch": 4.51, + "learning_rate": 3.0651703143502485e-06, + "loss": 1.0013, + "step": 61029 + }, + { + "epoch": 4.51, + "learning_rate": 3.0648829436263285e-06, + "loss": 1.0573, + "step": 61030 + }, + { + "epoch": 4.51, + "learning_rate": 3.0645955839360832e-06, + "loss": 1.0575, + "step": 61031 + }, + { + "epoch": 4.51, + "learning_rate": 3.0643082352799668e-06, + "loss": 1.1746, + "step": 61032 + }, + { + "epoch": 4.51, + "learning_rate": 3.0640208976584374e-06, + "loss": 1.0203, + "step": 61033 + }, + { + "epoch": 4.51, + "learning_rate": 3.0637335710719486e-06, + "loss": 0.8822, + "step": 61034 + }, + { + "epoch": 4.51, + "learning_rate": 3.0634462555209633e-06, + "loss": 0.9805, + "step": 61035 + }, + { + "epoch": 4.51, + "learning_rate": 3.0631589510059356e-06, + "loss": 1.0063, + "step": 61036 + }, + { + "epoch": 4.51, + "learning_rate": 3.0628716575273233e-06, + "loss": 1.0304, + "step": 61037 + }, + { + "epoch": 4.51, + "learning_rate": 3.062584375085582e-06, + "loss": 1.0249, + "step": 61038 + }, + { + "epoch": 4.51, + "learning_rate": 3.062297103681168e-06, + "loss": 1.0704, + "step": 61039 + }, + { + "epoch": 4.51, + "learning_rate": 3.0620098433145418e-06, + "loss": 0.9628, + "step": 61040 + }, + { + "epoch": 4.51, + "learning_rate": 3.0617225939861593e-06, + "loss": 1.095, + "step": 61041 + }, + { + "epoch": 4.51, + "learning_rate": 3.061435355696477e-06, + "loss": 0.9433, + "step": 61042 + }, + { + "epoch": 4.51, + "learning_rate": 3.0611481284459475e-06, + "loss": 1.0487, + "step": 61043 + }, + { + "epoch": 4.51, + "learning_rate": 3.0608609122350363e-06, + "loss": 0.979, + "step": 61044 + }, + { + "epoch": 4.51, + "learning_rate": 3.0605737070641996e-06, + "loss": 0.9374, + "step": 61045 + }, + { + "epoch": 4.51, + "learning_rate": 3.0602865129338843e-06, + "loss": 0.966, + "step": 61046 + }, + { + "epoch": 4.51, + "learning_rate": 3.059999329844556e-06, + "loss": 1.0501, + "step": 61047 + }, + { + "epoch": 4.51, + "learning_rate": 3.059712157796667e-06, + "loss": 1.0639, + "step": 61048 + }, + { + "epoch": 4.51, + "learning_rate": 3.059424996790683e-06, + "loss": 1.021, + "step": 61049 + }, + { + "epoch": 4.51, + "learning_rate": 3.0591378468270484e-06, + "loss": 1.1083, + "step": 61050 + }, + { + "epoch": 4.51, + "learning_rate": 3.058850707906229e-06, + "loss": 1.0207, + "step": 61051 + }, + { + "epoch": 4.51, + "learning_rate": 3.058563580028676e-06, + "loss": 1.0042, + "step": 61052 + }, + { + "epoch": 4.51, + "learning_rate": 3.058276463194856e-06, + "loss": 0.8899, + "step": 61053 + }, + { + "epoch": 4.51, + "learning_rate": 3.0579893574052113e-06, + "loss": 1.0855, + "step": 61054 + }, + { + "epoch": 4.51, + "learning_rate": 3.057702262660209e-06, + "loss": 0.9758, + "step": 61055 + }, + { + "epoch": 4.51, + "learning_rate": 3.0574151789603045e-06, + "loss": 1.0347, + "step": 61056 + }, + { + "epoch": 4.51, + "learning_rate": 3.057128106305952e-06, + "loss": 1.0163, + "step": 61057 + }, + { + "epoch": 4.51, + "learning_rate": 3.0568410446976103e-06, + "loss": 1.0165, + "step": 61058 + }, + { + "epoch": 4.51, + "learning_rate": 3.0565539941357314e-06, + "loss": 1.0483, + "step": 61059 + }, + { + "epoch": 4.51, + "learning_rate": 3.056266954620779e-06, + "loss": 1.0152, + "step": 61060 + }, + { + "epoch": 4.51, + "learning_rate": 3.0559799261532065e-06, + "loss": 1.0371, + "step": 61061 + }, + { + "epoch": 4.51, + "learning_rate": 3.0556929087334707e-06, + "loss": 0.8355, + "step": 61062 + }, + { + "epoch": 4.51, + "learning_rate": 3.055405902362024e-06, + "loss": 1.0159, + "step": 61063 + }, + { + "epoch": 4.51, + "learning_rate": 3.055118907039332e-06, + "loss": 0.9603, + "step": 61064 + }, + { + "epoch": 4.51, + "learning_rate": 3.0548319227658486e-06, + "loss": 0.9746, + "step": 61065 + }, + { + "epoch": 4.51, + "learning_rate": 3.054544949542022e-06, + "loss": 0.9597, + "step": 61066 + }, + { + "epoch": 4.51, + "learning_rate": 3.0542579873683186e-06, + "loss": 0.9973, + "step": 61067 + }, + { + "epoch": 4.51, + "learning_rate": 3.053971036245188e-06, + "loss": 0.977, + "step": 61068 + }, + { + "epoch": 4.51, + "learning_rate": 3.0536840961730963e-06, + "loss": 0.99, + "step": 61069 + }, + { + "epoch": 4.51, + "learning_rate": 3.0533971671524874e-06, + "loss": 1.0462, + "step": 61070 + }, + { + "epoch": 4.51, + "learning_rate": 3.053110249183827e-06, + "loss": 0.9532, + "step": 61071 + }, + { + "epoch": 4.51, + "learning_rate": 3.052823342267566e-06, + "loss": 0.9463, + "step": 61072 + }, + { + "epoch": 4.51, + "learning_rate": 3.0525364464041705e-06, + "loss": 1.0408, + "step": 61073 + }, + { + "epoch": 4.51, + "learning_rate": 3.0522495615940863e-06, + "loss": 1.053, + "step": 61074 + }, + { + "epoch": 4.51, + "learning_rate": 3.0519626878377707e-06, + "loss": 1.0817, + "step": 61075 + }, + { + "epoch": 4.51, + "learning_rate": 3.051675825135686e-06, + "loss": 0.9623, + "step": 61076 + }, + { + "epoch": 4.51, + "learning_rate": 3.0513889734882864e-06, + "loss": 1.0104, + "step": 61077 + }, + { + "epoch": 4.51, + "learning_rate": 3.0511021328960266e-06, + "loss": 0.9762, + "step": 61078 + }, + { + "epoch": 4.51, + "learning_rate": 3.050815303359361e-06, + "loss": 0.9614, + "step": 61079 + }, + { + "epoch": 4.51, + "learning_rate": 3.050528484878752e-06, + "loss": 0.9508, + "step": 61080 + }, + { + "epoch": 4.51, + "learning_rate": 3.0502416774546537e-06, + "loss": 0.9551, + "step": 61081 + }, + { + "epoch": 4.51, + "learning_rate": 3.0499548810875202e-06, + "loss": 0.9316, + "step": 61082 + }, + { + "epoch": 4.51, + "learning_rate": 3.049668095777809e-06, + "loss": 0.9939, + "step": 61083 + }, + { + "epoch": 4.51, + "learning_rate": 3.0493813215259747e-06, + "loss": 1.1232, + "step": 61084 + }, + { + "epoch": 4.51, + "learning_rate": 3.049094558332477e-06, + "loss": 1.0818, + "step": 61085 + }, + { + "epoch": 4.51, + "learning_rate": 3.0488078061977713e-06, + "loss": 0.9918, + "step": 61086 + }, + { + "epoch": 4.51, + "learning_rate": 3.0485210651223118e-06, + "loss": 1.0901, + "step": 61087 + }, + { + "epoch": 4.51, + "learning_rate": 3.0482343351065534e-06, + "loss": 0.9778, + "step": 61088 + }, + { + "epoch": 4.51, + "learning_rate": 3.0479476161509582e-06, + "loss": 1.0634, + "step": 61089 + }, + { + "epoch": 4.51, + "learning_rate": 3.047660908255978e-06, + "loss": 0.9047, + "step": 61090 + }, + { + "epoch": 4.51, + "learning_rate": 3.0473742114220706e-06, + "loss": 1.0025, + "step": 61091 + }, + { + "epoch": 4.51, + "learning_rate": 3.047087525649691e-06, + "loss": 0.9952, + "step": 61092 + }, + { + "epoch": 4.51, + "learning_rate": 3.046800850939293e-06, + "loss": 1.0192, + "step": 61093 + }, + { + "epoch": 4.51, + "learning_rate": 3.046514187291342e-06, + "loss": 0.9633, + "step": 61094 + }, + { + "epoch": 4.51, + "learning_rate": 3.0462275347062797e-06, + "loss": 0.9729, + "step": 61095 + }, + { + "epoch": 4.51, + "learning_rate": 3.0459408931845735e-06, + "loss": 0.9846, + "step": 61096 + }, + { + "epoch": 4.51, + "learning_rate": 3.0456542627266728e-06, + "loss": 1.0538, + "step": 61097 + }, + { + "epoch": 4.51, + "learning_rate": 3.0453676433330438e-06, + "loss": 1.0003, + "step": 61098 + }, + { + "epoch": 4.51, + "learning_rate": 3.0450810350041282e-06, + "loss": 0.967, + "step": 61099 + }, + { + "epoch": 4.51, + "learning_rate": 3.0447944377403924e-06, + "loss": 0.9139, + "step": 61100 + }, + { + "epoch": 4.51, + "learning_rate": 3.0445078515422887e-06, + "loss": 0.9848, + "step": 61101 + }, + { + "epoch": 4.51, + "learning_rate": 3.044221276410274e-06, + "loss": 1.0751, + "step": 61102 + }, + { + "epoch": 4.51, + "learning_rate": 3.0439347123448037e-06, + "loss": 1.0009, + "step": 61103 + }, + { + "epoch": 4.51, + "learning_rate": 3.043648159346331e-06, + "loss": 1.0031, + "step": 61104 + }, + { + "epoch": 4.52, + "learning_rate": 3.043361617415318e-06, + "loss": 0.983, + "step": 61105 + }, + { + "epoch": 4.52, + "learning_rate": 3.043075086552216e-06, + "loss": 0.8694, + "step": 61106 + }, + { + "epoch": 4.52, + "learning_rate": 3.042788566757483e-06, + "loss": 1.0111, + "step": 61107 + }, + { + "epoch": 4.52, + "learning_rate": 3.04250205803157e-06, + "loss": 1.0262, + "step": 61108 + }, + { + "epoch": 4.52, + "learning_rate": 3.042215560374939e-06, + "loss": 1.0275, + "step": 61109 + }, + { + "epoch": 4.52, + "learning_rate": 3.041929073788046e-06, + "loss": 0.9915, + "step": 61110 + }, + { + "epoch": 4.52, + "learning_rate": 3.0416425982713415e-06, + "loss": 0.9912, + "step": 61111 + }, + { + "epoch": 4.52, + "learning_rate": 3.041356133825286e-06, + "loss": 0.8862, + "step": 61112 + }, + { + "epoch": 4.52, + "learning_rate": 3.0410696804503294e-06, + "loss": 1.0091, + "step": 61113 + }, + { + "epoch": 4.52, + "learning_rate": 3.0407832381469383e-06, + "loss": 1.0135, + "step": 61114 + }, + { + "epoch": 4.52, + "learning_rate": 3.0404968069155537e-06, + "loss": 1.041, + "step": 61115 + }, + { + "epoch": 4.52, + "learning_rate": 3.0402103867566434e-06, + "loss": 0.9559, + "step": 61116 + }, + { + "epoch": 4.52, + "learning_rate": 3.0399239776706556e-06, + "loss": 1.0977, + "step": 61117 + }, + { + "epoch": 4.52, + "learning_rate": 3.039637579658056e-06, + "loss": 0.9408, + "step": 61118 + }, + { + "epoch": 4.52, + "learning_rate": 3.0393511927192865e-06, + "loss": 1.0653, + "step": 61119 + }, + { + "epoch": 4.52, + "learning_rate": 3.039064816854813e-06, + "loss": 1.0435, + "step": 61120 + }, + { + "epoch": 4.52, + "learning_rate": 3.0387784520650875e-06, + "loss": 1.0373, + "step": 61121 + }, + { + "epoch": 4.52, + "learning_rate": 3.0384920983505663e-06, + "loss": 0.9967, + "step": 61122 + }, + { + "epoch": 4.52, + "learning_rate": 3.038205755711704e-06, + "loss": 0.9372, + "step": 61123 + }, + { + "epoch": 4.52, + "learning_rate": 3.037919424148954e-06, + "loss": 0.9056, + "step": 61124 + }, + { + "epoch": 4.52, + "learning_rate": 3.0376331036627773e-06, + "loss": 0.9928, + "step": 61125 + }, + { + "epoch": 4.52, + "learning_rate": 3.0373467942536273e-06, + "loss": 0.9868, + "step": 61126 + }, + { + "epoch": 4.52, + "learning_rate": 3.0370604959219586e-06, + "loss": 1.024, + "step": 61127 + }, + { + "epoch": 4.52, + "learning_rate": 3.0367742086682238e-06, + "loss": 1.0617, + "step": 61128 + }, + { + "epoch": 4.52, + "learning_rate": 3.036487932492884e-06, + "loss": 1.0835, + "step": 61129 + }, + { + "epoch": 4.52, + "learning_rate": 3.0362016673963933e-06, + "loss": 1.0414, + "step": 61130 + }, + { + "epoch": 4.52, + "learning_rate": 3.035915413379207e-06, + "loss": 0.9605, + "step": 61131 + }, + { + "epoch": 4.52, + "learning_rate": 3.0356291704417784e-06, + "loss": 1.0295, + "step": 61132 + }, + { + "epoch": 4.52, + "learning_rate": 3.0353429385845612e-06, + "loss": 0.9873, + "step": 61133 + }, + { + "epoch": 4.52, + "learning_rate": 3.035056717808017e-06, + "loss": 1.0793, + "step": 61134 + }, + { + "epoch": 4.52, + "learning_rate": 3.0347705081125977e-06, + "loss": 0.9914, + "step": 61135 + }, + { + "epoch": 4.52, + "learning_rate": 3.0344843094987585e-06, + "loss": 0.9302, + "step": 61136 + }, + { + "epoch": 4.52, + "learning_rate": 3.0341981219669524e-06, + "loss": 1.0097, + "step": 61137 + }, + { + "epoch": 4.52, + "learning_rate": 3.0339119455176404e-06, + "loss": 1.0628, + "step": 61138 + }, + { + "epoch": 4.52, + "learning_rate": 3.033625780151278e-06, + "loss": 0.8678, + "step": 61139 + }, + { + "epoch": 4.52, + "learning_rate": 3.033339625868311e-06, + "loss": 1.0337, + "step": 61140 + }, + { + "epoch": 4.52, + "learning_rate": 3.033053482669205e-06, + "loss": 1.0091, + "step": 61141 + }, + { + "epoch": 4.52, + "learning_rate": 3.0327673505544076e-06, + "loss": 0.9805, + "step": 61142 + }, + { + "epoch": 4.52, + "learning_rate": 3.0324812295243835e-06, + "loss": 1.0722, + "step": 61143 + }, + { + "epoch": 4.52, + "learning_rate": 3.032195119579576e-06, + "loss": 1.0077, + "step": 61144 + }, + { + "epoch": 4.52, + "learning_rate": 3.031909020720449e-06, + "loss": 0.9907, + "step": 61145 + }, + { + "epoch": 4.52, + "learning_rate": 3.0316229329474556e-06, + "loss": 1.0373, + "step": 61146 + }, + { + "epoch": 4.52, + "learning_rate": 3.0313368562610513e-06, + "loss": 1.078, + "step": 61147 + }, + { + "epoch": 4.52, + "learning_rate": 3.031050790661689e-06, + "loss": 0.9602, + "step": 61148 + }, + { + "epoch": 4.52, + "learning_rate": 3.0307647361498226e-06, + "loss": 0.9411, + "step": 61149 + }, + { + "epoch": 4.52, + "learning_rate": 3.030478692725912e-06, + "loss": 0.9403, + "step": 61150 + }, + { + "epoch": 4.52, + "learning_rate": 3.0301926603904117e-06, + "loss": 1.065, + "step": 61151 + }, + { + "epoch": 4.52, + "learning_rate": 3.0299066391437737e-06, + "loss": 0.9362, + "step": 61152 + }, + { + "epoch": 4.52, + "learning_rate": 3.0296206289864525e-06, + "loss": 0.9083, + "step": 61153 + }, + { + "epoch": 4.52, + "learning_rate": 3.0293346299189076e-06, + "loss": 1.0338, + "step": 61154 + }, + { + "epoch": 4.52, + "learning_rate": 3.029048641941592e-06, + "loss": 0.9595, + "step": 61155 + }, + { + "epoch": 4.52, + "learning_rate": 3.0287626650549594e-06, + "loss": 0.9553, + "step": 61156 + }, + { + "epoch": 4.52, + "learning_rate": 3.028476699259467e-06, + "loss": 0.9827, + "step": 61157 + }, + { + "epoch": 4.52, + "learning_rate": 3.0281907445555646e-06, + "loss": 1.0229, + "step": 61158 + }, + { + "epoch": 4.52, + "learning_rate": 3.0279048009437174e-06, + "loss": 1.0205, + "step": 61159 + }, + { + "epoch": 4.52, + "learning_rate": 3.0276188684243668e-06, + "loss": 0.9655, + "step": 61160 + }, + { + "epoch": 4.52, + "learning_rate": 3.0273329469979794e-06, + "loss": 0.9583, + "step": 61161 + }, + { + "epoch": 4.52, + "learning_rate": 3.0270470366650017e-06, + "loss": 0.9853, + "step": 61162 + }, + { + "epoch": 4.52, + "learning_rate": 3.0267611374258987e-06, + "loss": 1.0046, + "step": 61163 + }, + { + "epoch": 4.52, + "learning_rate": 3.026475249281112e-06, + "loss": 0.967, + "step": 61164 + }, + { + "epoch": 4.52, + "learning_rate": 3.0261893722311073e-06, + "loss": 1.0302, + "step": 61165 + }, + { + "epoch": 4.52, + "learning_rate": 3.025903506276332e-06, + "loss": 0.9386, + "step": 61166 + }, + { + "epoch": 4.52, + "learning_rate": 3.025617651417251e-06, + "loss": 1.0459, + "step": 61167 + }, + { + "epoch": 4.52, + "learning_rate": 3.025331807654309e-06, + "loss": 1.0558, + "step": 61168 + }, + { + "epoch": 4.52, + "learning_rate": 3.0250459749879622e-06, + "loss": 1.0883, + "step": 61169 + }, + { + "epoch": 4.52, + "learning_rate": 3.0247601534186698e-06, + "loss": 0.9581, + "step": 61170 + }, + { + "epoch": 4.52, + "learning_rate": 3.0244743429468836e-06, + "loss": 1.0075, + "step": 61171 + }, + { + "epoch": 4.52, + "learning_rate": 3.02418854357306e-06, + "loss": 0.9726, + "step": 61172 + }, + { + "epoch": 4.52, + "learning_rate": 3.0239027552976485e-06, + "loss": 0.9909, + "step": 61173 + }, + { + "epoch": 4.52, + "learning_rate": 3.0236169781211123e-06, + "loss": 0.8936, + "step": 61174 + }, + { + "epoch": 4.52, + "learning_rate": 3.0233312120439005e-06, + "loss": 0.9868, + "step": 61175 + }, + { + "epoch": 4.52, + "learning_rate": 3.0230454570664702e-06, + "loss": 0.9423, + "step": 61176 + }, + { + "epoch": 4.52, + "learning_rate": 3.0227597131892737e-06, + "loss": 0.9632, + "step": 61177 + }, + { + "epoch": 4.52, + "learning_rate": 3.0224739804127633e-06, + "loss": 1.0399, + "step": 61178 + }, + { + "epoch": 4.52, + "learning_rate": 3.0221882587374006e-06, + "loss": 0.937, + "step": 61179 + }, + { + "epoch": 4.52, + "learning_rate": 3.0219025481636368e-06, + "loss": 1.0542, + "step": 61180 + }, + { + "epoch": 4.52, + "learning_rate": 3.021616848691926e-06, + "loss": 1.0342, + "step": 61181 + }, + { + "epoch": 4.52, + "learning_rate": 3.02133116032272e-06, + "loss": 1.0384, + "step": 61182 + }, + { + "epoch": 4.52, + "learning_rate": 3.0210454830564796e-06, + "loss": 0.9668, + "step": 61183 + }, + { + "epoch": 4.52, + "learning_rate": 3.0207598168936548e-06, + "loss": 1.0342, + "step": 61184 + }, + { + "epoch": 4.52, + "learning_rate": 3.020474161834702e-06, + "loss": 1.0091, + "step": 61185 + }, + { + "epoch": 4.52, + "learning_rate": 3.020188517880075e-06, + "loss": 1.0672, + "step": 61186 + }, + { + "epoch": 4.52, + "learning_rate": 3.019902885030228e-06, + "loss": 0.9222, + "step": 61187 + }, + { + "epoch": 4.52, + "learning_rate": 3.0196172632856156e-06, + "loss": 1.0172, + "step": 61188 + }, + { + "epoch": 4.52, + "learning_rate": 3.01933165264669e-06, + "loss": 0.9726, + "step": 61189 + }, + { + "epoch": 4.52, + "learning_rate": 3.0190460531139097e-06, + "loss": 1.0931, + "step": 61190 + }, + { + "epoch": 4.52, + "learning_rate": 3.018760464687729e-06, + "loss": 1.0728, + "step": 61191 + }, + { + "epoch": 4.52, + "learning_rate": 3.0184748873685987e-06, + "loss": 0.9433, + "step": 61192 + }, + { + "epoch": 4.52, + "learning_rate": 3.0181893211569724e-06, + "loss": 1.0923, + "step": 61193 + }, + { + "epoch": 4.52, + "learning_rate": 3.0179037660533104e-06, + "loss": 0.9901, + "step": 61194 + }, + { + "epoch": 4.52, + "learning_rate": 3.017618222058064e-06, + "loss": 1.065, + "step": 61195 + }, + { + "epoch": 4.52, + "learning_rate": 3.017332689171687e-06, + "loss": 1.0644, + "step": 61196 + }, + { + "epoch": 4.52, + "learning_rate": 3.0170471673946333e-06, + "loss": 0.9742, + "step": 61197 + }, + { + "epoch": 4.52, + "learning_rate": 3.016761656727354e-06, + "loss": 1.0038, + "step": 61198 + }, + { + "epoch": 4.52, + "learning_rate": 3.0164761571703104e-06, + "loss": 0.9716, + "step": 61199 + }, + { + "epoch": 4.52, + "learning_rate": 3.016190668723954e-06, + "loss": 0.9781, + "step": 61200 + }, + { + "epoch": 4.52, + "learning_rate": 3.0159051913887382e-06, + "loss": 1.0178, + "step": 61201 + }, + { + "epoch": 4.52, + "learning_rate": 3.015619725165114e-06, + "loss": 0.9843, + "step": 61202 + }, + { + "epoch": 4.52, + "learning_rate": 3.0153342700535416e-06, + "loss": 1.1085, + "step": 61203 + }, + { + "epoch": 4.52, + "learning_rate": 3.0150488260544763e-06, + "loss": 0.9295, + "step": 61204 + }, + { + "epoch": 4.52, + "learning_rate": 3.0147633931683618e-06, + "loss": 1.0382, + "step": 61205 + }, + { + "epoch": 4.52, + "learning_rate": 3.0144779713956617e-06, + "loss": 1.0963, + "step": 61206 + }, + { + "epoch": 4.52, + "learning_rate": 3.0141925607368238e-06, + "loss": 1.0223, + "step": 61207 + }, + { + "epoch": 4.52, + "learning_rate": 3.013907161192313e-06, + "loss": 0.9303, + "step": 61208 + }, + { + "epoch": 4.52, + "learning_rate": 3.013621772762568e-06, + "loss": 1.0089, + "step": 61209 + }, + { + "epoch": 4.52, + "learning_rate": 3.0133363954480554e-06, + "loss": 0.969, + "step": 61210 + }, + { + "epoch": 4.52, + "learning_rate": 3.013051029249221e-06, + "loss": 0.9605, + "step": 61211 + }, + { + "epoch": 4.52, + "learning_rate": 3.012765674166529e-06, + "loss": 1.1029, + "step": 61212 + }, + { + "epoch": 4.52, + "learning_rate": 3.0124803302004202e-06, + "loss": 1.0483, + "step": 61213 + }, + { + "epoch": 4.52, + "learning_rate": 3.0121949973513587e-06, + "loss": 0.9871, + "step": 61214 + }, + { + "epoch": 4.52, + "learning_rate": 3.011909675619794e-06, + "loss": 0.9867, + "step": 61215 + }, + { + "epoch": 4.52, + "learning_rate": 3.0116243650061828e-06, + "loss": 0.9872, + "step": 61216 + }, + { + "epoch": 4.52, + "learning_rate": 3.011339065510975e-06, + "loss": 0.978, + "step": 61217 + }, + { + "epoch": 4.52, + "learning_rate": 3.011053777134626e-06, + "loss": 1.0921, + "step": 61218 + }, + { + "epoch": 4.52, + "learning_rate": 3.010768499877592e-06, + "loss": 1.1074, + "step": 61219 + }, + { + "epoch": 4.52, + "learning_rate": 3.0104832337403257e-06, + "loss": 0.9174, + "step": 61220 + }, + { + "epoch": 4.52, + "learning_rate": 3.0101979787232807e-06, + "loss": 1.0748, + "step": 61221 + }, + { + "epoch": 4.52, + "learning_rate": 3.0099127348269074e-06, + "loss": 0.9314, + "step": 61222 + }, + { + "epoch": 4.52, + "learning_rate": 3.0096275020516664e-06, + "loss": 1.1594, + "step": 61223 + }, + { + "epoch": 4.52, + "learning_rate": 3.009342280398008e-06, + "loss": 0.8117, + "step": 61224 + }, + { + "epoch": 4.52, + "learning_rate": 3.0090570698663857e-06, + "loss": 0.9616, + "step": 61225 + }, + { + "epoch": 4.52, + "learning_rate": 3.008771870457253e-06, + "loss": 0.9833, + "step": 61226 + }, + { + "epoch": 4.52, + "learning_rate": 3.0084866821710624e-06, + "loss": 1.0158, + "step": 61227 + }, + { + "epoch": 4.52, + "learning_rate": 3.0082015050082714e-06, + "loss": 1.0287, + "step": 61228 + }, + { + "epoch": 4.52, + "learning_rate": 3.007916338969332e-06, + "loss": 1.0225, + "step": 61229 + }, + { + "epoch": 4.52, + "learning_rate": 3.007631184054699e-06, + "loss": 1.0089, + "step": 61230 + }, + { + "epoch": 4.52, + "learning_rate": 3.0073460402648203e-06, + "loss": 0.983, + "step": 61231 + }, + { + "epoch": 4.52, + "learning_rate": 3.00706090760016e-06, + "loss": 1.025, + "step": 61232 + }, + { + "epoch": 4.52, + "learning_rate": 3.0067757860611636e-06, + "loss": 0.9256, + "step": 61233 + }, + { + "epoch": 4.52, + "learning_rate": 3.0064906756482835e-06, + "loss": 1.0218, + "step": 61234 + }, + { + "epoch": 4.52, + "learning_rate": 3.006205576361979e-06, + "loss": 0.9676, + "step": 61235 + }, + { + "epoch": 4.52, + "learning_rate": 3.0059204882027027e-06, + "loss": 1.0661, + "step": 61236 + }, + { + "epoch": 4.52, + "learning_rate": 3.0056354111709053e-06, + "loss": 0.9047, + "step": 61237 + }, + { + "epoch": 4.52, + "learning_rate": 3.0053503452670397e-06, + "loss": 0.9928, + "step": 61238 + }, + { + "epoch": 4.52, + "learning_rate": 3.0050652904915645e-06, + "loss": 0.9974, + "step": 61239 + }, + { + "epoch": 4.52, + "learning_rate": 3.0047802468449303e-06, + "loss": 0.9929, + "step": 61240 + }, + { + "epoch": 4.53, + "learning_rate": 3.0044952143275917e-06, + "loss": 0.9648, + "step": 61241 + }, + { + "epoch": 4.53, + "learning_rate": 3.0042101929400004e-06, + "loss": 1.0558, + "step": 61242 + }, + { + "epoch": 4.53, + "learning_rate": 3.0039251826826067e-06, + "loss": 0.9502, + "step": 61243 + }, + { + "epoch": 4.53, + "learning_rate": 3.0036401835558714e-06, + "loss": 1.0592, + "step": 61244 + }, + { + "epoch": 4.53, + "learning_rate": 3.0033551955602446e-06, + "loss": 1.0591, + "step": 61245 + }, + { + "epoch": 4.53, + "learning_rate": 3.00307021869618e-06, + "loss": 1.1825, + "step": 61246 + }, + { + "epoch": 4.53, + "learning_rate": 3.0027852529641278e-06, + "loss": 0.9927, + "step": 61247 + }, + { + "epoch": 4.53, + "learning_rate": 3.0025002983645457e-06, + "loss": 1.105, + "step": 61248 + }, + { + "epoch": 4.53, + "learning_rate": 3.0022153548978874e-06, + "loss": 1.1292, + "step": 61249 + }, + { + "epoch": 4.53, + "learning_rate": 3.001930422564603e-06, + "loss": 0.9765, + "step": 61250 + }, + { + "epoch": 4.53, + "learning_rate": 3.001645501365147e-06, + "loss": 1.005, + "step": 61251 + }, + { + "epoch": 4.53, + "learning_rate": 3.001360591299971e-06, + "loss": 0.8876, + "step": 61252 + }, + { + "epoch": 4.53, + "learning_rate": 3.001075692369535e-06, + "loss": 0.928, + "step": 61253 + }, + { + "epoch": 4.53, + "learning_rate": 3.000790804574282e-06, + "loss": 1.0267, + "step": 61254 + }, + { + "epoch": 4.53, + "learning_rate": 3.0005059279146744e-06, + "loss": 0.9939, + "step": 61255 + }, + { + "epoch": 4.53, + "learning_rate": 3.0002210623911577e-06, + "loss": 1.0568, + "step": 61256 + }, + { + "epoch": 4.53, + "learning_rate": 2.999936208004195e-06, + "loss": 0.9037, + "step": 61257 + }, + { + "epoch": 4.53, + "learning_rate": 2.9996513647542282e-06, + "loss": 1.018, + "step": 61258 + }, + { + "epoch": 4.53, + "learning_rate": 2.9993665326417187e-06, + "loss": 1.0264, + "step": 61259 + }, + { + "epoch": 4.53, + "learning_rate": 2.999081711667117e-06, + "loss": 0.9358, + "step": 61260 + }, + { + "epoch": 4.53, + "learning_rate": 2.998796901830876e-06, + "loss": 0.896, + "step": 61261 + }, + { + "epoch": 4.53, + "learning_rate": 2.9985121031334498e-06, + "loss": 1.0237, + "step": 61262 + }, + { + "epoch": 4.53, + "learning_rate": 2.998227315575286e-06, + "loss": 1.0013, + "step": 61263 + }, + { + "epoch": 4.53, + "learning_rate": 2.997942539156847e-06, + "loss": 1.1021, + "step": 61264 + }, + { + "epoch": 4.53, + "learning_rate": 2.9976577738785804e-06, + "loss": 1.0047, + "step": 61265 + }, + { + "epoch": 4.53, + "learning_rate": 2.99737301974094e-06, + "loss": 0.9935, + "step": 61266 + }, + { + "epoch": 4.53, + "learning_rate": 2.9970882767443767e-06, + "loss": 1.0673, + "step": 61267 + }, + { + "epoch": 4.53, + "learning_rate": 2.9968035448893473e-06, + "loss": 0.9633, + "step": 61268 + }, + { + "epoch": 4.53, + "learning_rate": 2.9965188241763045e-06, + "loss": 1.0292, + "step": 61269 + }, + { + "epoch": 4.53, + "learning_rate": 2.9962341146056994e-06, + "loss": 0.9806, + "step": 61270 + }, + { + "epoch": 4.53, + "learning_rate": 2.9959494161779867e-06, + "loss": 0.9919, + "step": 61271 + }, + { + "epoch": 4.53, + "learning_rate": 2.9956647288936146e-06, + "loss": 1.0079, + "step": 61272 + }, + { + "epoch": 4.53, + "learning_rate": 2.9953800527530417e-06, + "loss": 0.9543, + "step": 61273 + }, + { + "epoch": 4.53, + "learning_rate": 2.99509538775672e-06, + "loss": 1.0297, + "step": 61274 + }, + { + "epoch": 4.53, + "learning_rate": 2.9948107339051013e-06, + "loss": 1.0381, + "step": 61275 + }, + { + "epoch": 4.53, + "learning_rate": 2.9945260911986353e-06, + "loss": 0.9203, + "step": 61276 + }, + { + "epoch": 4.53, + "learning_rate": 2.9942414596377843e-06, + "loss": 0.9515, + "step": 61277 + }, + { + "epoch": 4.53, + "learning_rate": 2.993956839222989e-06, + "loss": 1.0, + "step": 61278 + }, + { + "epoch": 4.53, + "learning_rate": 2.9936722299547105e-06, + "loss": 1.013, + "step": 61279 + }, + { + "epoch": 4.53, + "learning_rate": 2.9933876318333997e-06, + "loss": 0.9069, + "step": 61280 + }, + { + "epoch": 4.53, + "learning_rate": 2.9931030448595087e-06, + "loss": 1.0434, + "step": 61281 + }, + { + "epoch": 4.53, + "learning_rate": 2.9928184690334903e-06, + "loss": 1.077, + "step": 61282 + }, + { + "epoch": 4.53, + "learning_rate": 2.9925339043557957e-06, + "loss": 1.0113, + "step": 61283 + }, + { + "epoch": 4.53, + "learning_rate": 2.992249350826881e-06, + "loss": 1.0568, + "step": 61284 + }, + { + "epoch": 4.53, + "learning_rate": 2.9919648084471975e-06, + "loss": 0.9751, + "step": 61285 + }, + { + "epoch": 4.53, + "learning_rate": 2.9916802772171983e-06, + "loss": 0.8958, + "step": 61286 + }, + { + "epoch": 4.53, + "learning_rate": 2.991395757137332e-06, + "loss": 0.9427, + "step": 61287 + }, + { + "epoch": 4.53, + "learning_rate": 2.9911112482080586e-06, + "loss": 1.0874, + "step": 61288 + }, + { + "epoch": 4.53, + "learning_rate": 2.990826750429826e-06, + "loss": 0.9486, + "step": 61289 + }, + { + "epoch": 4.53, + "learning_rate": 2.9905422638030877e-06, + "loss": 1.0084, + "step": 61290 + }, + { + "epoch": 4.53, + "learning_rate": 2.990257788328297e-06, + "loss": 0.8814, + "step": 61291 + }, + { + "epoch": 4.53, + "learning_rate": 2.9899733240059025e-06, + "loss": 1.0553, + "step": 61292 + }, + { + "epoch": 4.53, + "learning_rate": 2.9896888708363637e-06, + "loss": 0.9757, + "step": 61293 + }, + { + "epoch": 4.53, + "learning_rate": 2.9894044288201283e-06, + "loss": 1.0397, + "step": 61294 + }, + { + "epoch": 4.53, + "learning_rate": 2.989119997957651e-06, + "loss": 0.9654, + "step": 61295 + }, + { + "epoch": 4.53, + "learning_rate": 2.9888355782493807e-06, + "loss": 1.0224, + "step": 61296 + }, + { + "epoch": 4.53, + "learning_rate": 2.988551169695776e-06, + "loss": 0.9079, + "step": 61297 + }, + { + "epoch": 4.53, + "learning_rate": 2.988266772297288e-06, + "loss": 0.9916, + "step": 61298 + }, + { + "epoch": 4.53, + "learning_rate": 2.987982386054362e-06, + "loss": 1.0879, + "step": 61299 + }, + { + "epoch": 4.53, + "learning_rate": 2.987698010967458e-06, + "loss": 0.9744, + "step": 61300 + }, + { + "epoch": 4.53, + "learning_rate": 2.9874136470370243e-06, + "loss": 0.9884, + "step": 61301 + }, + { + "epoch": 4.53, + "learning_rate": 2.98712929426352e-06, + "loss": 1.0944, + "step": 61302 + }, + { + "epoch": 4.53, + "learning_rate": 2.9868449526473876e-06, + "loss": 1.0287, + "step": 61303 + }, + { + "epoch": 4.53, + "learning_rate": 2.9865606221890866e-06, + "loss": 1.005, + "step": 61304 + }, + { + "epoch": 4.53, + "learning_rate": 2.986276302889065e-06, + "loss": 0.9866, + "step": 61305 + }, + { + "epoch": 4.53, + "learning_rate": 2.9859919947477834e-06, + "loss": 1.1034, + "step": 61306 + }, + { + "epoch": 4.53, + "learning_rate": 2.985707697765685e-06, + "loss": 1.001, + "step": 61307 + }, + { + "epoch": 4.53, + "learning_rate": 2.9854234119432226e-06, + "loss": 0.9945, + "step": 61308 + }, + { + "epoch": 4.53, + "learning_rate": 2.9851391372808537e-06, + "loss": 1.0474, + "step": 61309 + }, + { + "epoch": 4.53, + "learning_rate": 2.9848548737790286e-06, + "loss": 1.0426, + "step": 61310 + }, + { + "epoch": 4.53, + "learning_rate": 2.9845706214381975e-06, + "loss": 1.0373, + "step": 61311 + }, + { + "epoch": 4.53, + "learning_rate": 2.9842863802588127e-06, + "loss": 0.9647, + "step": 61312 + }, + { + "epoch": 4.53, + "learning_rate": 2.984002150241331e-06, + "loss": 1.0621, + "step": 61313 + }, + { + "epoch": 4.53, + "learning_rate": 2.9837179313862007e-06, + "loss": 1.0497, + "step": 61314 + }, + { + "epoch": 4.53, + "learning_rate": 2.9834337236938747e-06, + "loss": 0.9815, + "step": 61315 + }, + { + "epoch": 4.53, + "learning_rate": 2.983149527164806e-06, + "loss": 1.0099, + "step": 61316 + }, + { + "epoch": 4.53, + "learning_rate": 2.9828653417994426e-06, + "loss": 0.9756, + "step": 61317 + }, + { + "epoch": 4.53, + "learning_rate": 2.982581167598243e-06, + "loss": 1.0557, + "step": 61318 + }, + { + "epoch": 4.53, + "learning_rate": 2.982297004561656e-06, + "loss": 0.9711, + "step": 61319 + }, + { + "epoch": 4.53, + "learning_rate": 2.982012852690135e-06, + "loss": 0.9918, + "step": 61320 + }, + { + "epoch": 4.53, + "learning_rate": 2.9817287119841276e-06, + "loss": 1.0883, + "step": 61321 + }, + { + "epoch": 4.53, + "learning_rate": 2.9814445824440952e-06, + "loss": 1.0862, + "step": 61322 + }, + { + "epoch": 4.53, + "learning_rate": 2.9811604640704783e-06, + "loss": 1.0482, + "step": 61323 + }, + { + "epoch": 4.53, + "learning_rate": 2.9808763568637376e-06, + "loss": 0.9513, + "step": 61324 + }, + { + "epoch": 4.53, + "learning_rate": 2.98059226082432e-06, + "loss": 0.8989, + "step": 61325 + }, + { + "epoch": 4.53, + "learning_rate": 2.9803081759526853e-06, + "loss": 1.0554, + "step": 61326 + }, + { + "epoch": 4.53, + "learning_rate": 2.980024102249276e-06, + "loss": 1.0858, + "step": 61327 + }, + { + "epoch": 4.53, + "learning_rate": 2.979740039714545e-06, + "loss": 0.9408, + "step": 61328 + }, + { + "epoch": 4.53, + "learning_rate": 2.9794559883489513e-06, + "loss": 0.9615, + "step": 61329 + }, + { + "epoch": 4.53, + "learning_rate": 2.979171948152941e-06, + "loss": 0.9466, + "step": 61330 + }, + { + "epoch": 4.53, + "learning_rate": 2.9788879191269693e-06, + "loss": 0.9369, + "step": 61331 + }, + { + "epoch": 4.53, + "learning_rate": 2.9786039012714828e-06, + "loss": 0.9885, + "step": 61332 + }, + { + "epoch": 4.53, + "learning_rate": 2.9783198945869395e-06, + "loss": 0.9647, + "step": 61333 + }, + { + "epoch": 4.53, + "learning_rate": 2.97803589907379e-06, + "loss": 0.999, + "step": 61334 + }, + { + "epoch": 4.53, + "learning_rate": 2.977751914732484e-06, + "loss": 0.9676, + "step": 61335 + }, + { + "epoch": 4.53, + "learning_rate": 2.977467941563474e-06, + "loss": 0.9461, + "step": 61336 + }, + { + "epoch": 4.53, + "learning_rate": 2.9771839795672098e-06, + "loss": 1.1337, + "step": 61337 + }, + { + "epoch": 4.53, + "learning_rate": 2.9769000287441484e-06, + "loss": 1.0521, + "step": 61338 + }, + { + "epoch": 4.53, + "learning_rate": 2.9766160890947383e-06, + "loss": 1.0401, + "step": 61339 + }, + { + "epoch": 4.53, + "learning_rate": 2.9763321606194318e-06, + "loss": 1.0153, + "step": 61340 + }, + { + "epoch": 4.53, + "learning_rate": 2.9760482433186764e-06, + "loss": 1.0539, + "step": 61341 + }, + { + "epoch": 4.53, + "learning_rate": 2.9757643371929313e-06, + "loss": 1.0118, + "step": 61342 + }, + { + "epoch": 4.53, + "learning_rate": 2.9754804422426454e-06, + "loss": 0.9255, + "step": 61343 + }, + { + "epoch": 4.53, + "learning_rate": 2.9751965584682685e-06, + "loss": 1.0378, + "step": 61344 + }, + { + "epoch": 4.53, + "learning_rate": 2.9749126858702538e-06, + "loss": 1.0327, + "step": 61345 + }, + { + "epoch": 4.53, + "learning_rate": 2.9746288244490497e-06, + "loss": 1.017, + "step": 61346 + }, + { + "epoch": 4.53, + "learning_rate": 2.9743449742051166e-06, + "loss": 1.0914, + "step": 61347 + }, + { + "epoch": 4.53, + "learning_rate": 2.9740611351388937e-06, + "loss": 0.9894, + "step": 61348 + }, + { + "epoch": 4.53, + "learning_rate": 2.9737773072508426e-06, + "loss": 0.9581, + "step": 61349 + }, + { + "epoch": 4.53, + "learning_rate": 2.973493490541408e-06, + "loss": 0.9597, + "step": 61350 + }, + { + "epoch": 4.53, + "learning_rate": 2.9732096850110504e-06, + "loss": 0.9013, + "step": 61351 + }, + { + "epoch": 4.53, + "learning_rate": 2.9729258906602097e-06, + "loss": 0.9467, + "step": 61352 + }, + { + "epoch": 4.53, + "learning_rate": 2.9726421074893463e-06, + "loss": 0.9986, + "step": 61353 + }, + { + "epoch": 4.53, + "learning_rate": 2.9723583354989095e-06, + "loss": 1.0531, + "step": 61354 + }, + { + "epoch": 4.53, + "learning_rate": 2.97207457468935e-06, + "loss": 0.9681, + "step": 61355 + }, + { + "epoch": 4.53, + "learning_rate": 2.9717908250611182e-06, + "loss": 1.0122, + "step": 61356 + }, + { + "epoch": 4.53, + "learning_rate": 2.9715070866146646e-06, + "loss": 1.0876, + "step": 61357 + }, + { + "epoch": 4.53, + "learning_rate": 2.9712233593504456e-06, + "loss": 0.9069, + "step": 61358 + }, + { + "epoch": 4.53, + "learning_rate": 2.97093964326891e-06, + "loss": 0.8929, + "step": 61359 + }, + { + "epoch": 4.53, + "learning_rate": 2.970655938370508e-06, + "loss": 1.0403, + "step": 61360 + }, + { + "epoch": 4.53, + "learning_rate": 2.970372244655689e-06, + "loss": 1.0302, + "step": 61361 + }, + { + "epoch": 4.53, + "learning_rate": 2.97008856212491e-06, + "loss": 0.9816, + "step": 61362 + }, + { + "epoch": 4.53, + "learning_rate": 2.9698048907786203e-06, + "loss": 1.0089, + "step": 61363 + }, + { + "epoch": 4.53, + "learning_rate": 2.9695212306172696e-06, + "loss": 1.0532, + "step": 61364 + }, + { + "epoch": 4.53, + "learning_rate": 2.9692375816413112e-06, + "loss": 1.0182, + "step": 61365 + }, + { + "epoch": 4.53, + "learning_rate": 2.9689539438511904e-06, + "loss": 0.9438, + "step": 61366 + }, + { + "epoch": 4.53, + "learning_rate": 2.9686703172473707e-06, + "loss": 0.9529, + "step": 61367 + }, + { + "epoch": 4.53, + "learning_rate": 2.9683867018302894e-06, + "loss": 0.9961, + "step": 61368 + }, + { + "epoch": 4.53, + "learning_rate": 2.9681030976004076e-06, + "loss": 0.9376, + "step": 61369 + }, + { + "epoch": 4.53, + "learning_rate": 2.967819504558169e-06, + "loss": 1.1427, + "step": 61370 + }, + { + "epoch": 4.53, + "learning_rate": 2.967535922704037e-06, + "loss": 0.8832, + "step": 61371 + }, + { + "epoch": 4.53, + "learning_rate": 2.96725235203845e-06, + "loss": 0.9838, + "step": 61372 + }, + { + "epoch": 4.53, + "learning_rate": 2.966968792561862e-06, + "loss": 0.9965, + "step": 61373 + }, + { + "epoch": 4.53, + "learning_rate": 2.9666852442747285e-06, + "loss": 0.9482, + "step": 61374 + }, + { + "epoch": 4.53, + "learning_rate": 2.9664017071774988e-06, + "loss": 1.0226, + "step": 61375 + }, + { + "epoch": 4.54, + "learning_rate": 2.966118181270622e-06, + "loss": 1.0052, + "step": 61376 + }, + { + "epoch": 4.54, + "learning_rate": 2.965834666554548e-06, + "loss": 0.9398, + "step": 61377 + }, + { + "epoch": 4.54, + "learning_rate": 2.965551163029734e-06, + "loss": 0.976, + "step": 61378 + }, + { + "epoch": 4.54, + "learning_rate": 2.9652676706966266e-06, + "loss": 0.9782, + "step": 61379 + }, + { + "epoch": 4.54, + "learning_rate": 2.9649841895556787e-06, + "loss": 1.0303, + "step": 61380 + }, + { + "epoch": 4.54, + "learning_rate": 2.964700719607336e-06, + "loss": 1.0314, + "step": 61381 + }, + { + "epoch": 4.54, + "learning_rate": 2.9644172608520583e-06, + "loss": 0.9037, + "step": 61382 + }, + { + "epoch": 4.54, + "learning_rate": 2.964133813290292e-06, + "loss": 1.1097, + "step": 61383 + }, + { + "epoch": 4.54, + "learning_rate": 2.9638503769224867e-06, + "loss": 0.961, + "step": 61384 + }, + { + "epoch": 4.54, + "learning_rate": 2.963566951749096e-06, + "loss": 1.1082, + "step": 61385 + }, + { + "epoch": 4.54, + "learning_rate": 2.963283537770566e-06, + "loss": 1.0715, + "step": 61386 + }, + { + "epoch": 4.54, + "learning_rate": 2.963000134987355e-06, + "loss": 0.9375, + "step": 61387 + }, + { + "epoch": 4.54, + "learning_rate": 2.962716743399909e-06, + "loss": 0.9455, + "step": 61388 + }, + { + "epoch": 4.54, + "learning_rate": 2.962433363008681e-06, + "loss": 1.0106, + "step": 61389 + }, + { + "epoch": 4.54, + "learning_rate": 2.9621499938141164e-06, + "loss": 1.0103, + "step": 61390 + }, + { + "epoch": 4.54, + "learning_rate": 2.9618666358166746e-06, + "loss": 1.0119, + "step": 61391 + }, + { + "epoch": 4.54, + "learning_rate": 2.961583289016805e-06, + "loss": 1.0383, + "step": 61392 + }, + { + "epoch": 4.54, + "learning_rate": 2.96129995341495e-06, + "loss": 0.937, + "step": 61393 + }, + { + "epoch": 4.54, + "learning_rate": 2.9610166290115694e-06, + "loss": 0.9778, + "step": 61394 + }, + { + "epoch": 4.54, + "learning_rate": 2.960733315807106e-06, + "loss": 0.9936, + "step": 61395 + }, + { + "epoch": 4.54, + "learning_rate": 2.960450013802023e-06, + "loss": 0.9888, + "step": 61396 + }, + { + "epoch": 4.54, + "learning_rate": 2.9601667229967556e-06, + "loss": 0.9626, + "step": 61397 + }, + { + "epoch": 4.54, + "learning_rate": 2.9598834433917667e-06, + "loss": 0.9339, + "step": 61398 + }, + { + "epoch": 4.54, + "learning_rate": 2.9596001749875024e-06, + "loss": 0.9619, + "step": 61399 + }, + { + "epoch": 4.54, + "learning_rate": 2.9593169177844127e-06, + "loss": 0.9679, + "step": 61400 + }, + { + "epoch": 4.54, + "learning_rate": 2.959033671782949e-06, + "loss": 0.9354, + "step": 61401 + }, + { + "epoch": 4.54, + "learning_rate": 2.9587504369835596e-06, + "loss": 0.9105, + "step": 61402 + }, + { + "epoch": 4.54, + "learning_rate": 2.958467213386701e-06, + "loss": 1.1274, + "step": 61403 + }, + { + "epoch": 4.54, + "learning_rate": 2.95818400099282e-06, + "loss": 0.8885, + "step": 61404 + }, + { + "epoch": 4.54, + "learning_rate": 2.957900799802367e-06, + "loss": 1.0274, + "step": 61405 + }, + { + "epoch": 4.54, + "learning_rate": 2.9576176098157894e-06, + "loss": 1.0591, + "step": 61406 + }, + { + "epoch": 4.54, + "learning_rate": 2.957334431033545e-06, + "loss": 0.8923, + "step": 61407 + }, + { + "epoch": 4.54, + "learning_rate": 2.957051263456081e-06, + "loss": 1.0123, + "step": 61408 + }, + { + "epoch": 4.54, + "learning_rate": 2.9567681070838485e-06, + "loss": 0.8936, + "step": 61409 + }, + { + "epoch": 4.54, + "learning_rate": 2.9564849619172964e-06, + "loss": 0.9358, + "step": 61410 + }, + { + "epoch": 4.54, + "learning_rate": 2.956201827956873e-06, + "loss": 0.8855, + "step": 61411 + }, + { + "epoch": 4.54, + "learning_rate": 2.955918705203037e-06, + "loss": 0.9785, + "step": 61412 + }, + { + "epoch": 4.54, + "learning_rate": 2.9556355936562277e-06, + "loss": 0.9502, + "step": 61413 + }, + { + "epoch": 4.54, + "learning_rate": 2.955352493316905e-06, + "loss": 1.0427, + "step": 61414 + }, + { + "epoch": 4.54, + "learning_rate": 2.955069404185512e-06, + "loss": 0.9805, + "step": 61415 + }, + { + "epoch": 4.54, + "learning_rate": 2.9547863262625097e-06, + "loss": 1.0709, + "step": 61416 + }, + { + "epoch": 4.54, + "learning_rate": 2.954503259548336e-06, + "loss": 0.9905, + "step": 61417 + }, + { + "epoch": 4.54, + "learning_rate": 2.9542202040434486e-06, + "loss": 0.9274, + "step": 61418 + }, + { + "epoch": 4.54, + "learning_rate": 2.9539371597482956e-06, + "loss": 1.1346, + "step": 61419 + }, + { + "epoch": 4.54, + "learning_rate": 2.953654126663329e-06, + "loss": 0.9731, + "step": 61420 + }, + { + "epoch": 4.54, + "learning_rate": 2.953371104788998e-06, + "loss": 0.9109, + "step": 61421 + }, + { + "epoch": 4.54, + "learning_rate": 2.9530880941257488e-06, + "loss": 0.9827, + "step": 61422 + }, + { + "epoch": 4.54, + "learning_rate": 2.9528050946740396e-06, + "loss": 0.8846, + "step": 61423 + }, + { + "epoch": 4.54, + "learning_rate": 2.9525221064343158e-06, + "loss": 1.0564, + "step": 61424 + }, + { + "epoch": 4.54, + "learning_rate": 2.9522391294070284e-06, + "loss": 1.0337, + "step": 61425 + }, + { + "epoch": 4.54, + "learning_rate": 2.951956163592625e-06, + "loss": 0.987, + "step": 61426 + }, + { + "epoch": 4.54, + "learning_rate": 2.951673208991561e-06, + "loss": 0.9837, + "step": 61427 + }, + { + "epoch": 4.54, + "learning_rate": 2.9513902656042847e-06, + "loss": 0.9234, + "step": 61428 + }, + { + "epoch": 4.54, + "learning_rate": 2.9511073334312455e-06, + "loss": 0.9703, + "step": 61429 + }, + { + "epoch": 4.54, + "learning_rate": 2.9508244124728945e-06, + "loss": 1.0358, + "step": 61430 + }, + { + "epoch": 4.54, + "learning_rate": 2.950541502729677e-06, + "loss": 1.0964, + "step": 61431 + }, + { + "epoch": 4.54, + "learning_rate": 2.950258604202051e-06, + "loss": 1.0246, + "step": 61432 + }, + { + "epoch": 4.54, + "learning_rate": 2.9499757168904617e-06, + "loss": 1.0119, + "step": 61433 + }, + { + "epoch": 4.54, + "learning_rate": 2.9496928407953607e-06, + "loss": 1.0152, + "step": 61434 + }, + { + "epoch": 4.54, + "learning_rate": 2.9494099759171948e-06, + "loss": 1.0029, + "step": 61435 + }, + { + "epoch": 4.54, + "learning_rate": 2.9491271222564187e-06, + "loss": 1.0446, + "step": 61436 + }, + { + "epoch": 4.54, + "learning_rate": 2.948844279813481e-06, + "loss": 1.0215, + "step": 61437 + }, + { + "epoch": 4.54, + "learning_rate": 2.9485614485888314e-06, + "loss": 0.8963, + "step": 61438 + }, + { + "epoch": 4.54, + "learning_rate": 2.948278628582919e-06, + "loss": 1.0331, + "step": 61439 + }, + { + "epoch": 4.54, + "learning_rate": 2.947995819796192e-06, + "loss": 1.0443, + "step": 61440 + }, + { + "epoch": 4.54, + "learning_rate": 2.9477130222291095e-06, + "loss": 1.1723, + "step": 61441 + }, + { + "epoch": 4.54, + "learning_rate": 2.9474302358821073e-06, + "loss": 0.9426, + "step": 61442 + }, + { + "epoch": 4.54, + "learning_rate": 2.947147460755646e-06, + "loss": 1.0824, + "step": 61443 + }, + { + "epoch": 4.54, + "learning_rate": 2.9468646968501724e-06, + "loss": 0.9484, + "step": 61444 + }, + { + "epoch": 4.54, + "learning_rate": 2.946581944166137e-06, + "loss": 1.0243, + "step": 61445 + }, + { + "epoch": 4.54, + "learning_rate": 2.946299202703984e-06, + "loss": 0.9108, + "step": 61446 + }, + { + "epoch": 4.54, + "learning_rate": 2.9460164724641727e-06, + "loss": 1.0075, + "step": 61447 + }, + { + "epoch": 4.54, + "learning_rate": 2.9457337534471474e-06, + "loss": 1.0086, + "step": 61448 + }, + { + "epoch": 4.54, + "learning_rate": 2.945451045653359e-06, + "loss": 1.0188, + "step": 61449 + }, + { + "epoch": 4.54, + "learning_rate": 2.945168349083257e-06, + "loss": 0.8913, + "step": 61450 + }, + { + "epoch": 4.54, + "learning_rate": 2.944885663737288e-06, + "loss": 1.1234, + "step": 61451 + }, + { + "epoch": 4.54, + "learning_rate": 2.944602989615908e-06, + "loss": 0.9927, + "step": 61452 + }, + { + "epoch": 4.54, + "learning_rate": 2.944320326719564e-06, + "loss": 0.9615, + "step": 61453 + }, + { + "epoch": 4.54, + "learning_rate": 2.9440376750487053e-06, + "loss": 1.0326, + "step": 61454 + }, + { + "epoch": 4.54, + "learning_rate": 2.9437550346037793e-06, + "loss": 0.938, + "step": 61455 + }, + { + "epoch": 4.54, + "learning_rate": 2.9434724053852403e-06, + "loss": 1.0383, + "step": 61456 + }, + { + "epoch": 4.54, + "learning_rate": 2.943189787393539e-06, + "loss": 1.0054, + "step": 61457 + }, + { + "epoch": 4.54, + "learning_rate": 2.9429071806291155e-06, + "loss": 1.0485, + "step": 61458 + }, + { + "epoch": 4.54, + "learning_rate": 2.9426245850924284e-06, + "loss": 0.9857, + "step": 61459 + }, + { + "epoch": 4.54, + "learning_rate": 2.942342000783922e-06, + "loss": 0.9907, + "step": 61460 + }, + { + "epoch": 4.54, + "learning_rate": 2.9420594277040547e-06, + "loss": 1.0353, + "step": 61461 + }, + { + "epoch": 4.54, + "learning_rate": 2.9417768658532643e-06, + "loss": 0.9886, + "step": 61462 + }, + { + "epoch": 4.54, + "learning_rate": 2.9414943152320086e-06, + "loss": 1.0088, + "step": 61463 + }, + { + "epoch": 4.54, + "learning_rate": 2.941211775840731e-06, + "loss": 1.0151, + "step": 61464 + }, + { + "epoch": 4.54, + "learning_rate": 2.9409292476798913e-06, + "loss": 1.0572, + "step": 61465 + }, + { + "epoch": 4.54, + "learning_rate": 2.9406467307499276e-06, + "loss": 0.9926, + "step": 61466 + }, + { + "epoch": 4.54, + "learning_rate": 2.940364225051292e-06, + "loss": 0.9802, + "step": 61467 + }, + { + "epoch": 4.54, + "learning_rate": 2.9400817305844388e-06, + "loss": 1.0627, + "step": 61468 + }, + { + "epoch": 4.54, + "learning_rate": 2.939799247349815e-06, + "loss": 0.9956, + "step": 61469 + }, + { + "epoch": 4.54, + "learning_rate": 2.939516775347869e-06, + "loss": 1.0174, + "step": 61470 + }, + { + "epoch": 4.54, + "learning_rate": 2.939234314579048e-06, + "loss": 0.9553, + "step": 61471 + }, + { + "epoch": 4.54, + "learning_rate": 2.9389518650438076e-06, + "loss": 1.1142, + "step": 61472 + }, + { + "epoch": 4.54, + "learning_rate": 2.938669426742593e-06, + "loss": 0.874, + "step": 61473 + }, + { + "epoch": 4.54, + "learning_rate": 2.9383869996758553e-06, + "loss": 0.876, + "step": 61474 + }, + { + "epoch": 4.54, + "learning_rate": 2.9381045838440413e-06, + "loss": 0.9492, + "step": 61475 + }, + { + "epoch": 4.54, + "learning_rate": 2.9378221792476e-06, + "loss": 1.0056, + "step": 61476 + }, + { + "epoch": 4.54, + "learning_rate": 2.9375397858869846e-06, + "loss": 0.9976, + "step": 61477 + }, + { + "epoch": 4.54, + "learning_rate": 2.9372574037626435e-06, + "loss": 0.9465, + "step": 61478 + }, + { + "epoch": 4.54, + "learning_rate": 2.9369750328750245e-06, + "loss": 1.0973, + "step": 61479 + }, + { + "epoch": 4.54, + "learning_rate": 2.9366926732245736e-06, + "loss": 0.9811, + "step": 61480 + }, + { + "epoch": 4.54, + "learning_rate": 2.9364103248117482e-06, + "loss": 1.0721, + "step": 61481 + }, + { + "epoch": 4.54, + "learning_rate": 2.936127987636991e-06, + "loss": 0.8979, + "step": 61482 + }, + { + "epoch": 4.54, + "learning_rate": 2.9358456617007536e-06, + "loss": 1.0621, + "step": 61483 + }, + { + "epoch": 4.54, + "learning_rate": 2.9355633470034826e-06, + "loss": 0.9203, + "step": 61484 + }, + { + "epoch": 4.54, + "learning_rate": 2.9352810435456316e-06, + "loss": 0.9006, + "step": 61485 + }, + { + "epoch": 4.54, + "learning_rate": 2.9349987513276502e-06, + "loss": 1.0007, + "step": 61486 + }, + { + "epoch": 4.54, + "learning_rate": 2.9347164703499786e-06, + "loss": 1.0627, + "step": 61487 + }, + { + "epoch": 4.54, + "learning_rate": 2.9344342006130754e-06, + "loss": 1.0865, + "step": 61488 + }, + { + "epoch": 4.54, + "learning_rate": 2.934151942117387e-06, + "loss": 0.953, + "step": 61489 + }, + { + "epoch": 4.54, + "learning_rate": 2.9338696948633603e-06, + "loss": 0.9819, + "step": 61490 + }, + { + "epoch": 4.54, + "learning_rate": 2.9335874588514436e-06, + "loss": 1.0546, + "step": 61491 + }, + { + "epoch": 4.54, + "learning_rate": 2.933305234082091e-06, + "loss": 0.9102, + "step": 61492 + }, + { + "epoch": 4.54, + "learning_rate": 2.9330230205557487e-06, + "loss": 1.0385, + "step": 61493 + }, + { + "epoch": 4.54, + "learning_rate": 2.9327408182728657e-06, + "loss": 1.0277, + "step": 61494 + }, + { + "epoch": 4.54, + "learning_rate": 2.9324586272338906e-06, + "loss": 1.0383, + "step": 61495 + }, + { + "epoch": 4.54, + "learning_rate": 2.93217644743927e-06, + "loss": 0.9779, + "step": 61496 + }, + { + "epoch": 4.54, + "learning_rate": 2.9318942788894587e-06, + "loss": 1.109, + "step": 61497 + }, + { + "epoch": 4.54, + "learning_rate": 2.9316121215849023e-06, + "loss": 1.0352, + "step": 61498 + }, + { + "epoch": 4.54, + "learning_rate": 2.931329975526049e-06, + "loss": 1.0624, + "step": 61499 + }, + { + "epoch": 4.54, + "learning_rate": 2.931047840713347e-06, + "loss": 1.0077, + "step": 61500 + }, + { + "epoch": 4.54, + "learning_rate": 2.930765717147249e-06, + "loss": 1.0353, + "step": 61501 + }, + { + "epoch": 4.54, + "learning_rate": 2.9304836048282025e-06, + "loss": 0.9932, + "step": 61502 + }, + { + "epoch": 4.54, + "learning_rate": 2.9302015037566545e-06, + "loss": 1.0147, + "step": 61503 + }, + { + "epoch": 4.54, + "learning_rate": 2.9299194139330556e-06, + "loss": 0.9545, + "step": 61504 + }, + { + "epoch": 4.54, + "learning_rate": 2.92963733535785e-06, + "loss": 0.9439, + "step": 61505 + }, + { + "epoch": 4.54, + "learning_rate": 2.9293552680314963e-06, + "loss": 0.9438, + "step": 61506 + }, + { + "epoch": 4.54, + "learning_rate": 2.9290732119544308e-06, + "loss": 0.8976, + "step": 61507 + }, + { + "epoch": 4.54, + "learning_rate": 2.9287911671271118e-06, + "loss": 0.991, + "step": 61508 + }, + { + "epoch": 4.54, + "learning_rate": 2.9285091335499827e-06, + "loss": 0.9047, + "step": 61509 + }, + { + "epoch": 4.54, + "learning_rate": 2.9282271112235006e-06, + "loss": 1.0236, + "step": 61510 + }, + { + "epoch": 4.55, + "learning_rate": 2.927945100148101e-06, + "loss": 1.0859, + "step": 61511 + }, + { + "epoch": 4.55, + "learning_rate": 2.927663100324243e-06, + "loss": 1.0853, + "step": 61512 + }, + { + "epoch": 4.55, + "learning_rate": 2.927381111752371e-06, + "loss": 0.9129, + "step": 61513 + }, + { + "epoch": 4.55, + "learning_rate": 2.927099134432936e-06, + "loss": 1.1154, + "step": 61514 + }, + { + "epoch": 4.55, + "learning_rate": 2.9268171683663838e-06, + "loss": 0.9729, + "step": 61515 + }, + { + "epoch": 4.55, + "learning_rate": 2.9265352135531612e-06, + "loss": 1.0986, + "step": 61516 + }, + { + "epoch": 4.55, + "learning_rate": 2.926253269993724e-06, + "loss": 1.1012, + "step": 61517 + }, + { + "epoch": 4.55, + "learning_rate": 2.9259713376885155e-06, + "loss": 0.9637, + "step": 61518 + }, + { + "epoch": 4.55, + "learning_rate": 2.9256894166379856e-06, + "loss": 0.9417, + "step": 61519 + }, + { + "epoch": 4.55, + "learning_rate": 2.9254075068425804e-06, + "loss": 0.9806, + "step": 61520 + }, + { + "epoch": 4.55, + "learning_rate": 2.9251256083027536e-06, + "loss": 0.9956, + "step": 61521 + }, + { + "epoch": 4.55, + "learning_rate": 2.9248437210189494e-06, + "loss": 0.9755, + "step": 61522 + }, + { + "epoch": 4.55, + "learning_rate": 2.924561844991619e-06, + "loss": 1.0741, + "step": 61523 + }, + { + "epoch": 4.55, + "learning_rate": 2.9242799802212085e-06, + "loss": 1.0849, + "step": 61524 + }, + { + "epoch": 4.55, + "learning_rate": 2.9239981267081653e-06, + "loss": 1.0111, + "step": 61525 + }, + { + "epoch": 4.55, + "learning_rate": 2.9237162844529422e-06, + "loss": 1.0681, + "step": 61526 + }, + { + "epoch": 4.55, + "learning_rate": 2.9234344534559855e-06, + "loss": 1.0974, + "step": 61527 + }, + { + "epoch": 4.55, + "learning_rate": 2.923152633717743e-06, + "loss": 0.9195, + "step": 61528 + }, + { + "epoch": 4.55, + "learning_rate": 2.922870825238661e-06, + "loss": 1.0622, + "step": 61529 + }, + { + "epoch": 4.55, + "learning_rate": 2.922589028019196e-06, + "loss": 0.8591, + "step": 61530 + }, + { + "epoch": 4.55, + "learning_rate": 2.9223072420597876e-06, + "loss": 0.9768, + "step": 61531 + }, + { + "epoch": 4.55, + "learning_rate": 2.9220254673608838e-06, + "loss": 1.0467, + "step": 61532 + }, + { + "epoch": 4.55, + "learning_rate": 2.921743703922939e-06, + "loss": 1.0135, + "step": 61533 + }, + { + "epoch": 4.55, + "learning_rate": 2.9214619517463995e-06, + "loss": 1.1385, + "step": 61534 + }, + { + "epoch": 4.55, + "learning_rate": 2.9211802108317123e-06, + "loss": 0.9754, + "step": 61535 + }, + { + "epoch": 4.55, + "learning_rate": 2.920898481179323e-06, + "loss": 1.0055, + "step": 61536 + }, + { + "epoch": 4.55, + "learning_rate": 2.9206167627896863e-06, + "loss": 0.9842, + "step": 61537 + }, + { + "epoch": 4.55, + "learning_rate": 2.920335055663247e-06, + "loss": 1.014, + "step": 61538 + }, + { + "epoch": 4.55, + "learning_rate": 2.920053359800453e-06, + "loss": 0.9574, + "step": 61539 + }, + { + "epoch": 4.55, + "learning_rate": 2.9197716752017504e-06, + "loss": 0.9032, + "step": 61540 + }, + { + "epoch": 4.55, + "learning_rate": 2.919490001867592e-06, + "loss": 0.9948, + "step": 61541 + }, + { + "epoch": 4.55, + "learning_rate": 2.919208339798424e-06, + "loss": 1.0403, + "step": 61542 + }, + { + "epoch": 4.55, + "learning_rate": 2.9189266889946945e-06, + "loss": 1.0286, + "step": 61543 + }, + { + "epoch": 4.55, + "learning_rate": 2.9186450494568508e-06, + "loss": 1.0359, + "step": 61544 + }, + { + "epoch": 4.55, + "learning_rate": 2.9183634211853385e-06, + "loss": 0.9496, + "step": 61545 + }, + { + "epoch": 4.55, + "learning_rate": 2.9180818041806125e-06, + "loss": 1.0375, + "step": 61546 + }, + { + "epoch": 4.55, + "learning_rate": 2.917800198443116e-06, + "loss": 0.9311, + "step": 61547 + }, + { + "epoch": 4.55, + "learning_rate": 2.9175186039732983e-06, + "loss": 1.0033, + "step": 61548 + }, + { + "epoch": 4.55, + "learning_rate": 2.917237020771605e-06, + "loss": 0.9963, + "step": 61549 + }, + { + "epoch": 4.55, + "learning_rate": 2.916955448838489e-06, + "loss": 1.0151, + "step": 61550 + }, + { + "epoch": 4.55, + "learning_rate": 2.9166738881743974e-06, + "loss": 1.0446, + "step": 61551 + }, + { + "epoch": 4.55, + "learning_rate": 2.916392338779771e-06, + "loss": 1.0452, + "step": 61552 + }, + { + "epoch": 4.55, + "learning_rate": 2.9161108006550666e-06, + "loss": 0.9024, + "step": 61553 + }, + { + "epoch": 4.55, + "learning_rate": 2.9158292738007243e-06, + "loss": 1.0812, + "step": 61554 + }, + { + "epoch": 4.55, + "learning_rate": 2.9155477582172043e-06, + "loss": 0.9415, + "step": 61555 + }, + { + "epoch": 4.55, + "learning_rate": 2.915266253904938e-06, + "loss": 1.1021, + "step": 61556 + }, + { + "epoch": 4.55, + "learning_rate": 2.9149847608643868e-06, + "loss": 0.9617, + "step": 61557 + }, + { + "epoch": 4.55, + "learning_rate": 2.9147032790959895e-06, + "loss": 1.001, + "step": 61558 + }, + { + "epoch": 4.55, + "learning_rate": 2.914421808600204e-06, + "loss": 1.0259, + "step": 61559 + }, + { + "epoch": 4.55, + "learning_rate": 2.914140349377469e-06, + "loss": 0.9193, + "step": 61560 + }, + { + "epoch": 4.55, + "learning_rate": 2.9138589014282324e-06, + "loss": 0.8569, + "step": 61561 + }, + { + "epoch": 4.55, + "learning_rate": 2.913577464752948e-06, + "loss": 1.0319, + "step": 61562 + }, + { + "epoch": 4.55, + "learning_rate": 2.91329603935206e-06, + "loss": 1.1827, + "step": 61563 + }, + { + "epoch": 4.55, + "learning_rate": 2.913014625226017e-06, + "loss": 0.8867, + "step": 61564 + }, + { + "epoch": 4.55, + "learning_rate": 2.9127332223752637e-06, + "loss": 0.9686, + "step": 61565 + }, + { + "epoch": 4.55, + "learning_rate": 2.9124518308002526e-06, + "loss": 0.9093, + "step": 61566 + }, + { + "epoch": 4.55, + "learning_rate": 2.9121704505014304e-06, + "loss": 0.9314, + "step": 61567 + }, + { + "epoch": 4.55, + "learning_rate": 2.911889081479242e-06, + "loss": 0.9375, + "step": 61568 + }, + { + "epoch": 4.55, + "learning_rate": 2.911607723734138e-06, + "loss": 0.978, + "step": 61569 + }, + { + "epoch": 4.55, + "learning_rate": 2.911326377266561e-06, + "loss": 1.018, + "step": 61570 + }, + { + "epoch": 4.55, + "learning_rate": 2.9110450420769655e-06, + "loss": 0.9844, + "step": 61571 + }, + { + "epoch": 4.55, + "learning_rate": 2.9107637181657966e-06, + "loss": 1.003, + "step": 61572 + }, + { + "epoch": 4.55, + "learning_rate": 2.9104824055335002e-06, + "loss": 0.9977, + "step": 61573 + }, + { + "epoch": 4.55, + "learning_rate": 2.910201104180522e-06, + "loss": 0.9022, + "step": 61574 + }, + { + "epoch": 4.55, + "learning_rate": 2.9099198141073195e-06, + "loss": 1.0869, + "step": 61575 + }, + { + "epoch": 4.55, + "learning_rate": 2.9096385353143252e-06, + "loss": 0.8736, + "step": 61576 + }, + { + "epoch": 4.55, + "learning_rate": 2.909357267801999e-06, + "loss": 0.9335, + "step": 61577 + }, + { + "epoch": 4.55, + "learning_rate": 2.909076011570784e-06, + "loss": 1.0224, + "step": 61578 + }, + { + "epoch": 4.55, + "learning_rate": 2.908794766621127e-06, + "loss": 1.0215, + "step": 61579 + }, + { + "epoch": 4.55, + "learning_rate": 2.908513532953475e-06, + "loss": 0.952, + "step": 61580 + }, + { + "epoch": 4.55, + "learning_rate": 2.9082323105682753e-06, + "loss": 0.8909, + "step": 61581 + }, + { + "epoch": 4.55, + "learning_rate": 2.9079510994659788e-06, + "loss": 1.0075, + "step": 61582 + }, + { + "epoch": 4.55, + "learning_rate": 2.907669899647031e-06, + "loss": 0.9342, + "step": 61583 + }, + { + "epoch": 4.55, + "learning_rate": 2.907388711111879e-06, + "loss": 1.0152, + "step": 61584 + }, + { + "epoch": 4.55, + "learning_rate": 2.907107533860967e-06, + "loss": 1.0474, + "step": 61585 + }, + { + "epoch": 4.55, + "learning_rate": 2.906826367894747e-06, + "loss": 0.8774, + "step": 61586 + }, + { + "epoch": 4.55, + "learning_rate": 2.9065452132136662e-06, + "loss": 1.0483, + "step": 61587 + }, + { + "epoch": 4.55, + "learning_rate": 2.906264069818171e-06, + "loss": 1.0459, + "step": 61588 + }, + { + "epoch": 4.55, + "learning_rate": 2.9059829377087066e-06, + "loss": 1.0024, + "step": 61589 + }, + { + "epoch": 4.55, + "learning_rate": 2.9057018168857197e-06, + "loss": 1.0585, + "step": 61590 + }, + { + "epoch": 4.55, + "learning_rate": 2.9054207073496633e-06, + "loss": 0.9833, + "step": 61591 + }, + { + "epoch": 4.55, + "learning_rate": 2.90513960910098e-06, + "loss": 0.9583, + "step": 61592 + }, + { + "epoch": 4.55, + "learning_rate": 2.9048585221401183e-06, + "loss": 1.0612, + "step": 61593 + }, + { + "epoch": 4.55, + "learning_rate": 2.9045774464675226e-06, + "loss": 1.0626, + "step": 61594 + }, + { + "epoch": 4.55, + "learning_rate": 2.9042963820836454e-06, + "loss": 1.0321, + "step": 61595 + }, + { + "epoch": 4.55, + "learning_rate": 2.9040153289889315e-06, + "loss": 1.0596, + "step": 61596 + }, + { + "epoch": 4.55, + "learning_rate": 2.903734287183827e-06, + "loss": 1.0182, + "step": 61597 + }, + { + "epoch": 4.55, + "learning_rate": 2.90345325666878e-06, + "loss": 0.9587, + "step": 61598 + }, + { + "epoch": 4.55, + "learning_rate": 2.903172237444234e-06, + "loss": 1.071, + "step": 61599 + }, + { + "epoch": 4.55, + "learning_rate": 2.902891229510646e-06, + "loss": 1.0584, + "step": 61600 + }, + { + "epoch": 4.55, + "learning_rate": 2.9026102328684504e-06, + "loss": 1.0753, + "step": 61601 + }, + { + "epoch": 4.55, + "learning_rate": 2.902329247518103e-06, + "loss": 0.9402, + "step": 61602 + }, + { + "epoch": 4.55, + "learning_rate": 2.9020482734600465e-06, + "loss": 1.0594, + "step": 61603 + }, + { + "epoch": 4.55, + "learning_rate": 2.901767310694734e-06, + "loss": 1.1345, + "step": 61604 + }, + { + "epoch": 4.55, + "learning_rate": 2.9014863592226026e-06, + "loss": 1.0483, + "step": 61605 + }, + { + "epoch": 4.55, + "learning_rate": 2.9012054190441085e-06, + "loss": 0.9682, + "step": 61606 + }, + { + "epoch": 4.55, + "learning_rate": 2.9009244901596947e-06, + "loss": 0.9661, + "step": 61607 + }, + { + "epoch": 4.55, + "learning_rate": 2.900643572569808e-06, + "loss": 0.9423, + "step": 61608 + }, + { + "epoch": 4.55, + "learning_rate": 2.900362666274896e-06, + "loss": 1.0727, + "step": 61609 + }, + { + "epoch": 4.55, + "learning_rate": 2.9000817712754026e-06, + "loss": 1.0625, + "step": 61610 + }, + { + "epoch": 4.55, + "learning_rate": 2.899800887571781e-06, + "loss": 0.98, + "step": 61611 + }, + { + "epoch": 4.55, + "learning_rate": 2.8995200151644733e-06, + "loss": 0.9579, + "step": 61612 + }, + { + "epoch": 4.55, + "learning_rate": 2.899239154053929e-06, + "loss": 0.9564, + "step": 61613 + }, + { + "epoch": 4.55, + "learning_rate": 2.89895830424059e-06, + "loss": 1.0609, + "step": 61614 + }, + { + "epoch": 4.55, + "learning_rate": 2.8986774657249096e-06, + "loss": 0.9219, + "step": 61615 + }, + { + "epoch": 4.55, + "learning_rate": 2.8983966385073316e-06, + "loss": 0.9479, + "step": 61616 + }, + { + "epoch": 4.55, + "learning_rate": 2.898115822588303e-06, + "loss": 0.9823, + "step": 61617 + }, + { + "epoch": 4.55, + "learning_rate": 2.897835017968271e-06, + "loss": 1.0823, + "step": 61618 + }, + { + "epoch": 4.55, + "learning_rate": 2.8975542246476786e-06, + "loss": 1.0965, + "step": 61619 + }, + { + "epoch": 4.55, + "learning_rate": 2.897273442626982e-06, + "loss": 0.8349, + "step": 61620 + }, + { + "epoch": 4.55, + "learning_rate": 2.8969926719066155e-06, + "loss": 0.9911, + "step": 61621 + }, + { + "epoch": 4.55, + "learning_rate": 2.8967119124870356e-06, + "loss": 1.1479, + "step": 61622 + }, + { + "epoch": 4.55, + "learning_rate": 2.8964311643686816e-06, + "loss": 0.9497, + "step": 61623 + }, + { + "epoch": 4.55, + "learning_rate": 2.8961504275520103e-06, + "loss": 0.9072, + "step": 61624 + }, + { + "epoch": 4.55, + "learning_rate": 2.8958697020374593e-06, + "loss": 1.044, + "step": 61625 + }, + { + "epoch": 4.55, + "learning_rate": 2.895588987825475e-06, + "loss": 1.0006, + "step": 61626 + }, + { + "epoch": 4.55, + "learning_rate": 2.895308284916509e-06, + "loss": 0.9853, + "step": 61627 + }, + { + "epoch": 4.55, + "learning_rate": 2.895027593311007e-06, + "loss": 1.1094, + "step": 61628 + }, + { + "epoch": 4.55, + "learning_rate": 2.894746913009414e-06, + "loss": 0.9491, + "step": 61629 + }, + { + "epoch": 4.55, + "learning_rate": 2.894466244012174e-06, + "loss": 0.9832, + "step": 61630 + }, + { + "epoch": 4.55, + "learning_rate": 2.89418558631974e-06, + "loss": 1.0185, + "step": 61631 + }, + { + "epoch": 4.55, + "learning_rate": 2.893904939932554e-06, + "loss": 0.95, + "step": 61632 + }, + { + "epoch": 4.55, + "learning_rate": 2.8936243048510638e-06, + "loss": 1.0123, + "step": 61633 + }, + { + "epoch": 4.55, + "learning_rate": 2.8933436810757165e-06, + "loss": 1.0111, + "step": 61634 + }, + { + "epoch": 4.55, + "learning_rate": 2.893063068606954e-06, + "loss": 1.0318, + "step": 61635 + }, + { + "epoch": 4.55, + "learning_rate": 2.89278246744523e-06, + "loss": 1.0038, + "step": 61636 + }, + { + "epoch": 4.55, + "learning_rate": 2.8925018775909874e-06, + "loss": 1.0958, + "step": 61637 + }, + { + "epoch": 4.55, + "learning_rate": 2.892221299044672e-06, + "loss": 1.1209, + "step": 61638 + }, + { + "epoch": 4.55, + "learning_rate": 2.8919407318067284e-06, + "loss": 1.0942, + "step": 61639 + }, + { + "epoch": 4.55, + "learning_rate": 2.8916601758776077e-06, + "loss": 1.0552, + "step": 61640 + }, + { + "epoch": 4.55, + "learning_rate": 2.8913796312577545e-06, + "loss": 1.0187, + "step": 61641 + }, + { + "epoch": 4.55, + "learning_rate": 2.891099097947615e-06, + "loss": 0.9173, + "step": 61642 + }, + { + "epoch": 4.55, + "learning_rate": 2.8908185759476313e-06, + "loss": 1.0436, + "step": 61643 + }, + { + "epoch": 4.55, + "learning_rate": 2.890538065258257e-06, + "loss": 0.9865, + "step": 61644 + }, + { + "epoch": 4.55, + "learning_rate": 2.8902575658799383e-06, + "loss": 0.9186, + "step": 61645 + }, + { + "epoch": 4.55, + "learning_rate": 2.889977077813111e-06, + "loss": 1.0056, + "step": 61646 + }, + { + "epoch": 4.56, + "learning_rate": 2.8896966010582317e-06, + "loss": 0.998, + "step": 61647 + }, + { + "epoch": 4.56, + "learning_rate": 2.8894161356157413e-06, + "loss": 0.9563, + "step": 61648 + }, + { + "epoch": 4.56, + "learning_rate": 2.889135681486093e-06, + "loss": 0.9428, + "step": 61649 + }, + { + "epoch": 4.56, + "learning_rate": 2.8888552386697223e-06, + "loss": 0.9564, + "step": 61650 + }, + { + "epoch": 4.56, + "learning_rate": 2.888574807167085e-06, + "loss": 1.018, + "step": 61651 + }, + { + "epoch": 4.56, + "learning_rate": 2.888294386978623e-06, + "loss": 1.122, + "step": 61652 + }, + { + "epoch": 4.56, + "learning_rate": 2.8880139781047824e-06, + "loss": 0.8991, + "step": 61653 + }, + { + "epoch": 4.56, + "learning_rate": 2.8877335805460103e-06, + "loss": 1.0354, + "step": 61654 + }, + { + "epoch": 4.56, + "learning_rate": 2.8874531943027496e-06, + "loss": 1.0727, + "step": 61655 + }, + { + "epoch": 4.56, + "learning_rate": 2.887172819375451e-06, + "loss": 0.924, + "step": 61656 + }, + { + "epoch": 4.56, + "learning_rate": 2.8868924557645596e-06, + "loss": 0.9971, + "step": 61657 + }, + { + "epoch": 4.56, + "learning_rate": 2.886612103470521e-06, + "loss": 1.0145, + "step": 61658 + }, + { + "epoch": 4.56, + "learning_rate": 2.8863317624937758e-06, + "loss": 1.0089, + "step": 61659 + }, + { + "epoch": 4.56, + "learning_rate": 2.88605143283478e-06, + "loss": 0.9991, + "step": 61660 + }, + { + "epoch": 4.56, + "learning_rate": 2.8857711144939735e-06, + "loss": 0.9246, + "step": 61661 + }, + { + "epoch": 4.56, + "learning_rate": 2.8854908074718037e-06, + "loss": 1.0429, + "step": 61662 + }, + { + "epoch": 4.56, + "learning_rate": 2.8852105117687167e-06, + "loss": 0.9292, + "step": 61663 + }, + { + "epoch": 4.56, + "learning_rate": 2.884930227385154e-06, + "loss": 0.9781, + "step": 61664 + }, + { + "epoch": 4.56, + "learning_rate": 2.8846499543215723e-06, + "loss": 0.9151, + "step": 61665 + }, + { + "epoch": 4.56, + "learning_rate": 2.884369692578405e-06, + "loss": 1.0731, + "step": 61666 + }, + { + "epoch": 4.56, + "learning_rate": 2.884089442156105e-06, + "loss": 1.063, + "step": 61667 + }, + { + "epoch": 4.56, + "learning_rate": 2.883809203055116e-06, + "loss": 0.9943, + "step": 61668 + }, + { + "epoch": 4.56, + "learning_rate": 2.88352897527589e-06, + "loss": 0.9938, + "step": 61669 + }, + { + "epoch": 4.56, + "learning_rate": 2.883248758818861e-06, + "loss": 0.9403, + "step": 61670 + }, + { + "epoch": 4.56, + "learning_rate": 2.8829685536844843e-06, + "loss": 1.0394, + "step": 61671 + }, + { + "epoch": 4.56, + "learning_rate": 2.8826883598732035e-06, + "loss": 1.0577, + "step": 61672 + }, + { + "epoch": 4.56, + "learning_rate": 2.882408177385464e-06, + "loss": 0.9839, + "step": 61673 + }, + { + "epoch": 4.56, + "learning_rate": 2.8821280062217105e-06, + "loss": 1.0758, + "step": 61674 + }, + { + "epoch": 4.56, + "learning_rate": 2.881847846382386e-06, + "loss": 0.8815, + "step": 61675 + }, + { + "epoch": 4.56, + "learning_rate": 2.881567697867944e-06, + "loss": 0.986, + "step": 61676 + }, + { + "epoch": 4.56, + "learning_rate": 2.881287560678825e-06, + "loss": 0.9919, + "step": 61677 + }, + { + "epoch": 4.56, + "learning_rate": 2.881007434815476e-06, + "loss": 1.0085, + "step": 61678 + }, + { + "epoch": 4.56, + "learning_rate": 2.880727320278339e-06, + "loss": 1.0369, + "step": 61679 + }, + { + "epoch": 4.56, + "learning_rate": 2.8804472170678665e-06, + "loss": 1.0795, + "step": 61680 + }, + { + "epoch": 4.56, + "learning_rate": 2.8801671251845008e-06, + "loss": 1.0706, + "step": 61681 + }, + { + "epoch": 4.56, + "learning_rate": 2.8798870446286862e-06, + "loss": 0.9873, + "step": 61682 + }, + { + "epoch": 4.56, + "learning_rate": 2.8796069754008705e-06, + "loss": 0.8897, + "step": 61683 + }, + { + "epoch": 4.56, + "learning_rate": 2.879326917501495e-06, + "loss": 1.005, + "step": 61684 + }, + { + "epoch": 4.56, + "learning_rate": 2.8790468709310115e-06, + "loss": 1.044, + "step": 61685 + }, + { + "epoch": 4.56, + "learning_rate": 2.8787668356898625e-06, + "loss": 0.9487, + "step": 61686 + }, + { + "epoch": 4.56, + "learning_rate": 2.8784868117784938e-06, + "loss": 0.9675, + "step": 61687 + }, + { + "epoch": 4.56, + "learning_rate": 2.8782067991973473e-06, + "loss": 1.066, + "step": 61688 + }, + { + "epoch": 4.56, + "learning_rate": 2.877926797946875e-06, + "loss": 1.1852, + "step": 61689 + }, + { + "epoch": 4.56, + "learning_rate": 2.877646808027522e-06, + "loss": 0.9833, + "step": 61690 + }, + { + "epoch": 4.56, + "learning_rate": 2.877366829439726e-06, + "loss": 0.8769, + "step": 61691 + }, + { + "epoch": 4.56, + "learning_rate": 2.87708686218394e-06, + "loss": 0.9669, + "step": 61692 + }, + { + "epoch": 4.56, + "learning_rate": 2.876806906260603e-06, + "loss": 1.0266, + "step": 61693 + }, + { + "epoch": 4.56, + "learning_rate": 2.876526961670171e-06, + "loss": 0.9498, + "step": 61694 + }, + { + "epoch": 4.56, + "learning_rate": 2.8762470284130772e-06, + "loss": 1.1236, + "step": 61695 + }, + { + "epoch": 4.56, + "learning_rate": 2.8759671064897743e-06, + "loss": 1.1479, + "step": 61696 + }, + { + "epoch": 4.56, + "learning_rate": 2.875687195900707e-06, + "loss": 1.0318, + "step": 61697 + }, + { + "epoch": 4.56, + "learning_rate": 2.8754072966463186e-06, + "loss": 0.9138, + "step": 61698 + }, + { + "epoch": 4.56, + "learning_rate": 2.875127408727052e-06, + "loss": 1.053, + "step": 61699 + }, + { + "epoch": 4.56, + "learning_rate": 2.87484753214336e-06, + "loss": 0.8851, + "step": 61700 + }, + { + "epoch": 4.56, + "learning_rate": 2.8745676668956825e-06, + "loss": 1.0576, + "step": 61701 + }, + { + "epoch": 4.56, + "learning_rate": 2.8742878129844666e-06, + "loss": 1.0313, + "step": 61702 + }, + { + "epoch": 4.56, + "learning_rate": 2.8740079704101565e-06, + "loss": 0.9126, + "step": 61703 + }, + { + "epoch": 4.56, + "learning_rate": 2.8737281391731943e-06, + "loss": 0.9849, + "step": 61704 + }, + { + "epoch": 4.56, + "learning_rate": 2.8734483192740325e-06, + "loss": 1.0087, + "step": 61705 + }, + { + "epoch": 4.56, + "learning_rate": 2.8731685107131126e-06, + "loss": 1.0221, + "step": 61706 + }, + { + "epoch": 4.56, + "learning_rate": 2.872888713490879e-06, + "loss": 1.0196, + "step": 61707 + }, + { + "epoch": 4.56, + "learning_rate": 2.872608927607774e-06, + "loss": 1.0268, + "step": 61708 + }, + { + "epoch": 4.56, + "learning_rate": 2.8723291530642493e-06, + "loss": 0.8434, + "step": 61709 + }, + { + "epoch": 4.56, + "learning_rate": 2.8720493898607516e-06, + "loss": 0.9715, + "step": 61710 + }, + { + "epoch": 4.56, + "learning_rate": 2.8717696379977134e-06, + "loss": 0.9525, + "step": 61711 + }, + { + "epoch": 4.56, + "learning_rate": 2.871489897475592e-06, + "loss": 0.982, + "step": 61712 + }, + { + "epoch": 4.56, + "learning_rate": 2.8712101682948244e-06, + "loss": 0.9723, + "step": 61713 + }, + { + "epoch": 4.56, + "learning_rate": 2.8709304504558655e-06, + "loss": 0.8817, + "step": 61714 + }, + { + "epoch": 4.56, + "learning_rate": 2.8706507439591492e-06, + "loss": 1.0929, + "step": 61715 + }, + { + "epoch": 4.56, + "learning_rate": 2.8703710488051274e-06, + "loss": 1.0266, + "step": 61716 + }, + { + "epoch": 4.56, + "learning_rate": 2.8700913649942408e-06, + "loss": 1.0036, + "step": 61717 + }, + { + "epoch": 4.56, + "learning_rate": 2.8698116925269427e-06, + "loss": 1.1552, + "step": 61718 + }, + { + "epoch": 4.56, + "learning_rate": 2.869532031403669e-06, + "loss": 1.0859, + "step": 61719 + }, + { + "epoch": 4.56, + "learning_rate": 2.8692523816248653e-06, + "loss": 1.0732, + "step": 61720 + }, + { + "epoch": 4.56, + "learning_rate": 2.868972743190982e-06, + "loss": 0.9358, + "step": 61721 + }, + { + "epoch": 4.56, + "learning_rate": 2.868693116102461e-06, + "loss": 0.924, + "step": 61722 + }, + { + "epoch": 4.56, + "learning_rate": 2.868413500359747e-06, + "loss": 0.9158, + "step": 61723 + }, + { + "epoch": 4.56, + "learning_rate": 2.8681338959632822e-06, + "loss": 0.9961, + "step": 61724 + }, + { + "epoch": 4.56, + "learning_rate": 2.8678543029135163e-06, + "loss": 1.0444, + "step": 61725 + }, + { + "epoch": 4.56, + "learning_rate": 2.867574721210894e-06, + "loss": 1.0269, + "step": 61726 + }, + { + "epoch": 4.56, + "learning_rate": 2.8672951508558565e-06, + "loss": 1.012, + "step": 61727 + }, + { + "epoch": 4.56, + "learning_rate": 2.8670155918488517e-06, + "loss": 1.0065, + "step": 61728 + }, + { + "epoch": 4.56, + "learning_rate": 2.8667360441903193e-06, + "loss": 1.0537, + "step": 61729 + }, + { + "epoch": 4.56, + "learning_rate": 2.8664565078807117e-06, + "loss": 0.9976, + "step": 61730 + }, + { + "epoch": 4.56, + "learning_rate": 2.866176982920469e-06, + "loss": 1.0466, + "step": 61731 + }, + { + "epoch": 4.56, + "learning_rate": 2.865897469310036e-06, + "loss": 0.8969, + "step": 61732 + }, + { + "epoch": 4.56, + "learning_rate": 2.865617967049856e-06, + "loss": 1.0333, + "step": 61733 + }, + { + "epoch": 4.56, + "learning_rate": 2.865338476140379e-06, + "loss": 1.0068, + "step": 61734 + }, + { + "epoch": 4.56, + "learning_rate": 2.8650589965820465e-06, + "loss": 1.0368, + "step": 61735 + }, + { + "epoch": 4.56, + "learning_rate": 2.864779528375302e-06, + "loss": 0.906, + "step": 61736 + }, + { + "epoch": 4.56, + "learning_rate": 2.8645000715205917e-06, + "loss": 0.9919, + "step": 61737 + }, + { + "epoch": 4.56, + "learning_rate": 2.864220626018357e-06, + "loss": 1.0224, + "step": 61738 + }, + { + "epoch": 4.56, + "learning_rate": 2.8639411918690505e-06, + "loss": 1.0546, + "step": 61739 + }, + { + "epoch": 4.56, + "learning_rate": 2.863661769073106e-06, + "loss": 0.959, + "step": 61740 + }, + { + "epoch": 4.56, + "learning_rate": 2.863382357630976e-06, + "loss": 1.0675, + "step": 61741 + }, + { + "epoch": 4.56, + "learning_rate": 2.8631029575431033e-06, + "loss": 1.0452, + "step": 61742 + }, + { + "epoch": 4.56, + "learning_rate": 2.8628235688099315e-06, + "loss": 1.0555, + "step": 61743 + }, + { + "epoch": 4.56, + "learning_rate": 2.8625441914319017e-06, + "loss": 1.0134, + "step": 61744 + }, + { + "epoch": 4.56, + "learning_rate": 2.8622648254094653e-06, + "loss": 1.0399, + "step": 61745 + }, + { + "epoch": 4.56, + "learning_rate": 2.8619854707430638e-06, + "loss": 0.8733, + "step": 61746 + }, + { + "epoch": 4.56, + "learning_rate": 2.8617061274331404e-06, + "loss": 1.0447, + "step": 61747 + }, + { + "epoch": 4.56, + "learning_rate": 2.861426795480141e-06, + "loss": 1.0302, + "step": 61748 + }, + { + "epoch": 4.56, + "learning_rate": 2.861147474884507e-06, + "loss": 1.0451, + "step": 61749 + }, + { + "epoch": 4.56, + "learning_rate": 2.8608681656466877e-06, + "loss": 0.8926, + "step": 61750 + }, + { + "epoch": 4.56, + "learning_rate": 2.8605888677671244e-06, + "loss": 1.1118, + "step": 61751 + }, + { + "epoch": 4.56, + "learning_rate": 2.8603095812462633e-06, + "loss": 1.0778, + "step": 61752 + }, + { + "epoch": 4.56, + "learning_rate": 2.8600303060845434e-06, + "loss": 1.0133, + "step": 61753 + }, + { + "epoch": 4.56, + "learning_rate": 2.859751042282416e-06, + "loss": 1.0407, + "step": 61754 + }, + { + "epoch": 4.56, + "learning_rate": 2.8594717898403235e-06, + "loss": 0.945, + "step": 61755 + }, + { + "epoch": 4.56, + "learning_rate": 2.859192548758709e-06, + "loss": 1.0713, + "step": 61756 + }, + { + "epoch": 4.56, + "learning_rate": 2.858913319038017e-06, + "loss": 1.0098, + "step": 61757 + }, + { + "epoch": 4.56, + "learning_rate": 2.8586341006786878e-06, + "loss": 0.9252, + "step": 61758 + }, + { + "epoch": 4.56, + "learning_rate": 2.858354893681177e-06, + "loss": 0.9091, + "step": 61759 + }, + { + "epoch": 4.56, + "learning_rate": 2.8580756980459146e-06, + "loss": 1.0487, + "step": 61760 + }, + { + "epoch": 4.56, + "learning_rate": 2.857796513773354e-06, + "loss": 0.9586, + "step": 61761 + }, + { + "epoch": 4.56, + "learning_rate": 2.8575173408639346e-06, + "loss": 1.0377, + "step": 61762 + }, + { + "epoch": 4.56, + "learning_rate": 2.857238179318109e-06, + "loss": 1.03, + "step": 61763 + }, + { + "epoch": 4.56, + "learning_rate": 2.856959029136309e-06, + "loss": 0.982, + "step": 61764 + }, + { + "epoch": 4.56, + "learning_rate": 2.8566798903189886e-06, + "loss": 0.9622, + "step": 61765 + }, + { + "epoch": 4.56, + "learning_rate": 2.856400762866588e-06, + "loss": 0.9162, + "step": 61766 + }, + { + "epoch": 4.56, + "learning_rate": 2.8561216467795515e-06, + "loss": 1.077, + "step": 61767 + }, + { + "epoch": 4.56, + "learning_rate": 2.8558425420583235e-06, + "loss": 0.9432, + "step": 61768 + }, + { + "epoch": 4.56, + "learning_rate": 2.8555634487033446e-06, + "loss": 0.9286, + "step": 61769 + }, + { + "epoch": 4.56, + "learning_rate": 2.855284366715064e-06, + "loss": 0.9708, + "step": 61770 + }, + { + "epoch": 4.56, + "learning_rate": 2.8550052960939257e-06, + "loss": 0.8918, + "step": 61771 + }, + { + "epoch": 4.56, + "learning_rate": 2.8547262368403704e-06, + "loss": 1.0274, + "step": 61772 + }, + { + "epoch": 4.56, + "learning_rate": 2.854447188954841e-06, + "loss": 1.1304, + "step": 61773 + }, + { + "epoch": 4.56, + "learning_rate": 2.8541681524377862e-06, + "loss": 1.0399, + "step": 61774 + }, + { + "epoch": 4.56, + "learning_rate": 2.8538891272896484e-06, + "loss": 1.0561, + "step": 61775 + }, + { + "epoch": 4.56, + "learning_rate": 2.8536101135108695e-06, + "loss": 1.005, + "step": 61776 + }, + { + "epoch": 4.56, + "learning_rate": 2.8533311111018957e-06, + "loss": 0.908, + "step": 61777 + }, + { + "epoch": 4.56, + "learning_rate": 2.853052120063167e-06, + "loss": 0.9637, + "step": 61778 + }, + { + "epoch": 4.56, + "learning_rate": 2.8527731403951332e-06, + "loss": 0.884, + "step": 61779 + }, + { + "epoch": 4.56, + "learning_rate": 2.8524941720982345e-06, + "loss": 0.8367, + "step": 61780 + }, + { + "epoch": 4.56, + "learning_rate": 2.8522152151729156e-06, + "loss": 1.0485, + "step": 61781 + }, + { + "epoch": 4.57, + "learning_rate": 2.8519362696196172e-06, + "loss": 1.0707, + "step": 61782 + }, + { + "epoch": 4.57, + "learning_rate": 2.851657335438789e-06, + "loss": 0.9522, + "step": 61783 + }, + { + "epoch": 4.57, + "learning_rate": 2.851378412630874e-06, + "loss": 1.0677, + "step": 61784 + }, + { + "epoch": 4.57, + "learning_rate": 2.8510995011963082e-06, + "loss": 0.9935, + "step": 61785 + }, + { + "epoch": 4.57, + "learning_rate": 2.8508206011355434e-06, + "loss": 1.0156, + "step": 61786 + }, + { + "epoch": 4.57, + "learning_rate": 2.850541712449021e-06, + "loss": 0.9433, + "step": 61787 + }, + { + "epoch": 4.57, + "learning_rate": 2.850262835137184e-06, + "loss": 0.8297, + "step": 61788 + }, + { + "epoch": 4.57, + "learning_rate": 2.849983969200475e-06, + "loss": 0.9949, + "step": 61789 + }, + { + "epoch": 4.57, + "learning_rate": 2.8497051146393406e-06, + "loss": 1.0328, + "step": 61790 + }, + { + "epoch": 4.57, + "learning_rate": 2.849426271454223e-06, + "loss": 1.0742, + "step": 61791 + }, + { + "epoch": 4.57, + "learning_rate": 2.8491474396455676e-06, + "loss": 0.9102, + "step": 61792 + }, + { + "epoch": 4.57, + "learning_rate": 2.848868619213815e-06, + "loss": 1.0146, + "step": 61793 + }, + { + "epoch": 4.57, + "learning_rate": 2.8485898101594067e-06, + "loss": 1.088, + "step": 61794 + }, + { + "epoch": 4.57, + "learning_rate": 2.848311012482794e-06, + "loss": 0.9178, + "step": 61795 + }, + { + "epoch": 4.57, + "learning_rate": 2.848032226184415e-06, + "loss": 0.978, + "step": 61796 + }, + { + "epoch": 4.57, + "learning_rate": 2.8477534512647155e-06, + "loss": 1.0635, + "step": 61797 + }, + { + "epoch": 4.57, + "learning_rate": 2.8474746877241344e-06, + "loss": 0.9635, + "step": 61798 + }, + { + "epoch": 4.57, + "learning_rate": 2.8471959355631217e-06, + "loss": 1.0344, + "step": 61799 + }, + { + "epoch": 4.57, + "learning_rate": 2.8469171947821173e-06, + "loss": 1.042, + "step": 61800 + }, + { + "epoch": 4.57, + "learning_rate": 2.846638465381566e-06, + "loss": 0.999, + "step": 61801 + }, + { + "epoch": 4.57, + "learning_rate": 2.846359747361911e-06, + "loss": 0.9972, + "step": 61802 + }, + { + "epoch": 4.57, + "learning_rate": 2.846081040723592e-06, + "loss": 0.8805, + "step": 61803 + }, + { + "epoch": 4.57, + "learning_rate": 2.8458023454670614e-06, + "loss": 0.9638, + "step": 61804 + }, + { + "epoch": 4.57, + "learning_rate": 2.8455236615927506e-06, + "loss": 0.9348, + "step": 61805 + }, + { + "epoch": 4.57, + "learning_rate": 2.8452449891011124e-06, + "loss": 1.0041, + "step": 61806 + }, + { + "epoch": 4.57, + "learning_rate": 2.8449663279925843e-06, + "loss": 0.9938, + "step": 61807 + }, + { + "epoch": 4.57, + "learning_rate": 2.844687678267618e-06, + "loss": 0.8702, + "step": 61808 + }, + { + "epoch": 4.57, + "learning_rate": 2.8444090399266457e-06, + "loss": 0.9661, + "step": 61809 + }, + { + "epoch": 4.57, + "learning_rate": 2.8441304129701198e-06, + "loss": 1.0209, + "step": 61810 + }, + { + "epoch": 4.57, + "learning_rate": 2.843851797398476e-06, + "loss": 0.8808, + "step": 61811 + }, + { + "epoch": 4.57, + "learning_rate": 2.843573193212168e-06, + "loss": 0.8934, + "step": 61812 + }, + { + "epoch": 4.57, + "learning_rate": 2.8432946004116303e-06, + "loss": 1.065, + "step": 61813 + }, + { + "epoch": 4.57, + "learning_rate": 2.8430160189973042e-06, + "loss": 1.0363, + "step": 61814 + }, + { + "epoch": 4.57, + "learning_rate": 2.842737448969641e-06, + "loss": 1.1102, + "step": 61815 + }, + { + "epoch": 4.57, + "learning_rate": 2.8424588903290796e-06, + "loss": 0.9644, + "step": 61816 + }, + { + "epoch": 4.57, + "learning_rate": 2.842180343076064e-06, + "loss": 1.1663, + "step": 61817 + }, + { + "epoch": 4.57, + "learning_rate": 2.841901807211035e-06, + "loss": 0.9788, + "step": 61818 + }, + { + "epoch": 4.57, + "learning_rate": 2.8416232827344394e-06, + "loss": 0.8965, + "step": 61819 + }, + { + "epoch": 4.57, + "learning_rate": 2.841344769646719e-06, + "loss": 0.9421, + "step": 61820 + }, + { + "epoch": 4.57, + "learning_rate": 2.841066267948317e-06, + "loss": 1.074, + "step": 61821 + }, + { + "epoch": 4.57, + "learning_rate": 2.8407877776396763e-06, + "loss": 1.0042, + "step": 61822 + }, + { + "epoch": 4.57, + "learning_rate": 2.840509298721237e-06, + "loss": 0.9514, + "step": 61823 + }, + { + "epoch": 4.57, + "learning_rate": 2.8402308311934467e-06, + "loss": 0.9548, + "step": 61824 + }, + { + "epoch": 4.57, + "learning_rate": 2.8399523750567483e-06, + "loss": 1.0423, + "step": 61825 + }, + { + "epoch": 4.57, + "learning_rate": 2.839673930311583e-06, + "loss": 1.0248, + "step": 61826 + }, + { + "epoch": 4.57, + "learning_rate": 2.839395496958389e-06, + "loss": 0.9297, + "step": 61827 + }, + { + "epoch": 4.57, + "learning_rate": 2.8391170749976193e-06, + "loss": 1.0963, + "step": 61828 + }, + { + "epoch": 4.57, + "learning_rate": 2.8388386644297116e-06, + "loss": 0.975, + "step": 61829 + }, + { + "epoch": 4.57, + "learning_rate": 2.8385602652551094e-06, + "loss": 1.0319, + "step": 61830 + }, + { + "epoch": 4.57, + "learning_rate": 2.838281877474255e-06, + "loss": 0.9663, + "step": 61831 + }, + { + "epoch": 4.57, + "learning_rate": 2.8380035010875918e-06, + "loss": 1.0664, + "step": 61832 + }, + { + "epoch": 4.57, + "learning_rate": 2.8377251360955615e-06, + "loss": 1.0433, + "step": 61833 + }, + { + "epoch": 4.57, + "learning_rate": 2.837446782498606e-06, + "loss": 1.1115, + "step": 61834 + }, + { + "epoch": 4.57, + "learning_rate": 2.8371684402971735e-06, + "loss": 1.0304, + "step": 61835 + }, + { + "epoch": 4.57, + "learning_rate": 2.836890109491702e-06, + "loss": 0.9839, + "step": 61836 + }, + { + "epoch": 4.57, + "learning_rate": 2.8366117900826372e-06, + "loss": 0.8881, + "step": 61837 + }, + { + "epoch": 4.57, + "learning_rate": 2.836333482070418e-06, + "loss": 0.9639, + "step": 61838 + }, + { + "epoch": 4.57, + "learning_rate": 2.8360551854554918e-06, + "loss": 1.0374, + "step": 61839 + }, + { + "epoch": 4.57, + "learning_rate": 2.8357769002382997e-06, + "loss": 0.9509, + "step": 61840 + }, + { + "epoch": 4.57, + "learning_rate": 2.835498626419284e-06, + "loss": 1.071, + "step": 61841 + }, + { + "epoch": 4.57, + "learning_rate": 2.835220363998887e-06, + "loss": 1.0362, + "step": 61842 + }, + { + "epoch": 4.57, + "learning_rate": 2.834942112977549e-06, + "loss": 0.9993, + "step": 61843 + }, + { + "epoch": 4.57, + "learning_rate": 2.8346638733557185e-06, + "loss": 0.9128, + "step": 61844 + }, + { + "epoch": 4.57, + "learning_rate": 2.8343856451338357e-06, + "loss": 1.0156, + "step": 61845 + }, + { + "epoch": 4.57, + "learning_rate": 2.8341074283123425e-06, + "loss": 1.0193, + "step": 61846 + }, + { + "epoch": 4.57, + "learning_rate": 2.83382922289168e-06, + "loss": 1.0724, + "step": 61847 + }, + { + "epoch": 4.57, + "learning_rate": 2.8335510288722945e-06, + "loss": 0.8758, + "step": 61848 + }, + { + "epoch": 4.57, + "learning_rate": 2.83327284625463e-06, + "loss": 1.0413, + "step": 61849 + }, + { + "epoch": 4.57, + "learning_rate": 2.832994675039119e-06, + "loss": 0.9645, + "step": 61850 + }, + { + "epoch": 4.57, + "learning_rate": 2.832716515226215e-06, + "loss": 1.0437, + "step": 61851 + }, + { + "epoch": 4.57, + "learning_rate": 2.8324383668163537e-06, + "loss": 1.0567, + "step": 61852 + }, + { + "epoch": 4.57, + "learning_rate": 2.832160229809986e-06, + "loss": 1.0406, + "step": 61853 + }, + { + "epoch": 4.57, + "learning_rate": 2.831882104207543e-06, + "loss": 1.0312, + "step": 61854 + }, + { + "epoch": 4.57, + "learning_rate": 2.831603990009475e-06, + "loss": 1.0159, + "step": 61855 + }, + { + "epoch": 4.57, + "learning_rate": 2.8313258872162207e-06, + "loss": 1.0571, + "step": 61856 + }, + { + "epoch": 4.57, + "learning_rate": 2.83104779582823e-06, + "loss": 0.9537, + "step": 61857 + }, + { + "epoch": 4.57, + "learning_rate": 2.8307697158459335e-06, + "loss": 1.05, + "step": 61858 + }, + { + "epoch": 4.57, + "learning_rate": 2.830491647269783e-06, + "loss": 0.9933, + "step": 61859 + }, + { + "epoch": 4.57, + "learning_rate": 2.830213590100218e-06, + "loss": 1.1712, + "step": 61860 + }, + { + "epoch": 4.57, + "learning_rate": 2.82993554433768e-06, + "loss": 0.9893, + "step": 61861 + }, + { + "epoch": 4.57, + "learning_rate": 2.8296575099826117e-06, + "loss": 1.0377, + "step": 61862 + }, + { + "epoch": 4.57, + "learning_rate": 2.8293794870354528e-06, + "loss": 1.0409, + "step": 61863 + }, + { + "epoch": 4.57, + "learning_rate": 2.829101475496653e-06, + "loss": 1.0094, + "step": 61864 + }, + { + "epoch": 4.57, + "learning_rate": 2.828823475366649e-06, + "loss": 0.8962, + "step": 61865 + }, + { + "epoch": 4.57, + "learning_rate": 2.828545486645884e-06, + "loss": 0.951, + "step": 61866 + }, + { + "epoch": 4.57, + "learning_rate": 2.8282675093347977e-06, + "loss": 1.0965, + "step": 61867 + }, + { + "epoch": 4.57, + "learning_rate": 2.827989543433839e-06, + "loss": 1.058, + "step": 61868 + }, + { + "epoch": 4.57, + "learning_rate": 2.827711588943445e-06, + "loss": 0.9766, + "step": 61869 + }, + { + "epoch": 4.57, + "learning_rate": 2.8274336458640606e-06, + "loss": 0.9895, + "step": 61870 + }, + { + "epoch": 4.57, + "learning_rate": 2.8271557141961257e-06, + "loss": 1.0116, + "step": 61871 + }, + { + "epoch": 4.57, + "learning_rate": 2.8268777939400806e-06, + "loss": 0.953, + "step": 61872 + }, + { + "epoch": 4.57, + "learning_rate": 2.826599885096376e-06, + "loss": 0.965, + "step": 61873 + }, + { + "epoch": 4.57, + "learning_rate": 2.8263219876654437e-06, + "loss": 1.056, + "step": 61874 + }, + { + "epoch": 4.57, + "learning_rate": 2.8260441016477324e-06, + "loss": 1.0533, + "step": 61875 + }, + { + "epoch": 4.57, + "learning_rate": 2.825766227043679e-06, + "loss": 1.0805, + "step": 61876 + }, + { + "epoch": 4.57, + "learning_rate": 2.8254883638537356e-06, + "loss": 1.1339, + "step": 61877 + }, + { + "epoch": 4.57, + "learning_rate": 2.8252105120783337e-06, + "loss": 0.9114, + "step": 61878 + }, + { + "epoch": 4.57, + "learning_rate": 2.824932671717917e-06, + "loss": 0.9876, + "step": 61879 + }, + { + "epoch": 4.57, + "learning_rate": 2.8246548427729325e-06, + "loss": 1.0664, + "step": 61880 + }, + { + "epoch": 4.57, + "learning_rate": 2.8243770252438197e-06, + "loss": 0.9762, + "step": 61881 + }, + { + "epoch": 4.57, + "learning_rate": 2.8240992191310203e-06, + "loss": 0.9527, + "step": 61882 + }, + { + "epoch": 4.57, + "learning_rate": 2.8238214244349728e-06, + "loss": 0.9886, + "step": 61883 + }, + { + "epoch": 4.57, + "learning_rate": 2.8235436411561268e-06, + "loss": 1.0351, + "step": 61884 + }, + { + "epoch": 4.57, + "learning_rate": 2.823265869294919e-06, + "loss": 0.9551, + "step": 61885 + }, + { + "epoch": 4.57, + "learning_rate": 2.822988108851794e-06, + "loss": 1.0775, + "step": 61886 + }, + { + "epoch": 4.57, + "learning_rate": 2.822710359827191e-06, + "loss": 1.0247, + "step": 61887 + }, + { + "epoch": 4.57, + "learning_rate": 2.82243262222155e-06, + "loss": 0.9822, + "step": 61888 + }, + { + "epoch": 4.57, + "learning_rate": 2.8221548960353197e-06, + "loss": 0.9656, + "step": 61889 + }, + { + "epoch": 4.57, + "learning_rate": 2.821877181268938e-06, + "loss": 1.1394, + "step": 61890 + }, + { + "epoch": 4.57, + "learning_rate": 2.8215994779228474e-06, + "loss": 1.1011, + "step": 61891 + }, + { + "epoch": 4.57, + "learning_rate": 2.821321785997486e-06, + "loss": 0.983, + "step": 61892 + }, + { + "epoch": 4.57, + "learning_rate": 2.8210441054933014e-06, + "loss": 1.0463, + "step": 61893 + }, + { + "epoch": 4.57, + "learning_rate": 2.8207664364107334e-06, + "loss": 1.1373, + "step": 61894 + }, + { + "epoch": 4.57, + "learning_rate": 2.820488778750223e-06, + "loss": 1.0531, + "step": 61895 + }, + { + "epoch": 4.57, + "learning_rate": 2.820211132512214e-06, + "loss": 0.951, + "step": 61896 + }, + { + "epoch": 4.57, + "learning_rate": 2.8199334976971417e-06, + "loss": 1.001, + "step": 61897 + }, + { + "epoch": 4.57, + "learning_rate": 2.819655874305458e-06, + "loss": 1.0094, + "step": 61898 + }, + { + "epoch": 4.57, + "learning_rate": 2.819378262337594e-06, + "loss": 0.9685, + "step": 61899 + }, + { + "epoch": 4.57, + "learning_rate": 2.8191006617939985e-06, + "loss": 1.0623, + "step": 61900 + }, + { + "epoch": 4.57, + "learning_rate": 2.8188230726751097e-06, + "loss": 1.0319, + "step": 61901 + }, + { + "epoch": 4.57, + "learning_rate": 2.818545494981375e-06, + "loss": 0.8614, + "step": 61902 + }, + { + "epoch": 4.57, + "learning_rate": 2.818267928713226e-06, + "loss": 0.981, + "step": 61903 + }, + { + "epoch": 4.57, + "learning_rate": 2.8179903738711146e-06, + "loss": 0.9848, + "step": 61904 + }, + { + "epoch": 4.57, + "learning_rate": 2.817712830455477e-06, + "loss": 1.0896, + "step": 61905 + }, + { + "epoch": 4.57, + "learning_rate": 2.817435298466754e-06, + "loss": 1.0368, + "step": 61906 + }, + { + "epoch": 4.57, + "learning_rate": 2.8171577779053905e-06, + "loss": 0.962, + "step": 61907 + }, + { + "epoch": 4.57, + "learning_rate": 2.816880268771822e-06, + "loss": 0.9723, + "step": 61908 + }, + { + "epoch": 4.57, + "learning_rate": 2.816602771066498e-06, + "loss": 1.0714, + "step": 61909 + }, + { + "epoch": 4.57, + "learning_rate": 2.816325284789856e-06, + "loss": 0.8991, + "step": 61910 + }, + { + "epoch": 4.57, + "learning_rate": 2.816047809942337e-06, + "loss": 1.0171, + "step": 61911 + }, + { + "epoch": 4.57, + "learning_rate": 2.815770346524381e-06, + "loss": 0.9731, + "step": 61912 + }, + { + "epoch": 4.57, + "learning_rate": 2.8154928945364347e-06, + "loss": 1.0297, + "step": 61913 + }, + { + "epoch": 4.57, + "learning_rate": 2.8152154539789357e-06, + "loss": 1.0277, + "step": 61914 + }, + { + "epoch": 4.57, + "learning_rate": 2.8149380248523263e-06, + "loss": 0.8646, + "step": 61915 + }, + { + "epoch": 4.57, + "learning_rate": 2.8146606071570482e-06, + "loss": 1.0015, + "step": 61916 + }, + { + "epoch": 4.58, + "learning_rate": 2.814383200893539e-06, + "loss": 1.0022, + "step": 61917 + }, + { + "epoch": 4.58, + "learning_rate": 2.8141058060622484e-06, + "loss": 0.9887, + "step": 61918 + }, + { + "epoch": 4.58, + "learning_rate": 2.813828422663608e-06, + "loss": 1.0337, + "step": 61919 + }, + { + "epoch": 4.58, + "learning_rate": 2.813551050698066e-06, + "loss": 0.9534, + "step": 61920 + }, + { + "epoch": 4.58, + "learning_rate": 2.8132736901660586e-06, + "loss": 1.0811, + "step": 61921 + }, + { + "epoch": 4.58, + "learning_rate": 2.812996341068036e-06, + "loss": 0.9653, + "step": 61922 + }, + { + "epoch": 4.58, + "learning_rate": 2.812719003404427e-06, + "loss": 0.9739, + "step": 61923 + }, + { + "epoch": 4.58, + "learning_rate": 2.8124416771756834e-06, + "loss": 0.9934, + "step": 61924 + }, + { + "epoch": 4.58, + "learning_rate": 2.8121643623822415e-06, + "loss": 0.938, + "step": 61925 + }, + { + "epoch": 4.58, + "learning_rate": 2.8118870590245427e-06, + "loss": 1.0199, + "step": 61926 + }, + { + "epoch": 4.58, + "learning_rate": 2.8116097671030296e-06, + "loss": 0.963, + "step": 61927 + }, + { + "epoch": 4.58, + "learning_rate": 2.8113324866181392e-06, + "loss": 1.0636, + "step": 61928 + }, + { + "epoch": 4.58, + "learning_rate": 2.811055217570319e-06, + "loss": 1.005, + "step": 61929 + }, + { + "epoch": 4.58, + "learning_rate": 2.8107779599600073e-06, + "loss": 1.0073, + "step": 61930 + }, + { + "epoch": 4.58, + "learning_rate": 2.810500713787645e-06, + "loss": 0.9319, + "step": 61931 + }, + { + "epoch": 4.58, + "learning_rate": 2.8102234790536696e-06, + "loss": 0.9149, + "step": 61932 + }, + { + "epoch": 4.58, + "learning_rate": 2.8099462557585288e-06, + "loss": 1.058, + "step": 61933 + }, + { + "epoch": 4.58, + "learning_rate": 2.809669043902662e-06, + "loss": 0.9936, + "step": 61934 + }, + { + "epoch": 4.58, + "learning_rate": 2.8093918434865077e-06, + "loss": 1.0507, + "step": 61935 + }, + { + "epoch": 4.58, + "learning_rate": 2.8091146545105075e-06, + "loss": 0.9001, + "step": 61936 + }, + { + "epoch": 4.58, + "learning_rate": 2.8088374769751015e-06, + "loss": 0.9043, + "step": 61937 + }, + { + "epoch": 4.58, + "learning_rate": 2.8085603108807346e-06, + "loss": 1.0371, + "step": 61938 + }, + { + "epoch": 4.58, + "learning_rate": 2.8082831562278444e-06, + "loss": 1.1314, + "step": 61939 + }, + { + "epoch": 4.58, + "learning_rate": 2.808006013016874e-06, + "loss": 0.9473, + "step": 61940 + }, + { + "epoch": 4.58, + "learning_rate": 2.8077288812482593e-06, + "loss": 1.0095, + "step": 61941 + }, + { + "epoch": 4.58, + "learning_rate": 2.8074517609224495e-06, + "loss": 0.9304, + "step": 61942 + }, + { + "epoch": 4.58, + "learning_rate": 2.8071746520398825e-06, + "loss": 1.0116, + "step": 61943 + }, + { + "epoch": 4.58, + "learning_rate": 2.8068975546009923e-06, + "loss": 0.9927, + "step": 61944 + }, + { + "epoch": 4.58, + "learning_rate": 2.8066204686062283e-06, + "loss": 0.9737, + "step": 61945 + }, + { + "epoch": 4.58, + "learning_rate": 2.8063433940560247e-06, + "loss": 1.0005, + "step": 61946 + }, + { + "epoch": 4.58, + "learning_rate": 2.806066330950832e-06, + "loss": 0.9848, + "step": 61947 + }, + { + "epoch": 4.58, + "learning_rate": 2.805789279291079e-06, + "loss": 0.9878, + "step": 61948 + }, + { + "epoch": 4.58, + "learning_rate": 2.805512239077216e-06, + "loss": 1.0443, + "step": 61949 + }, + { + "epoch": 4.58, + "learning_rate": 2.805235210309676e-06, + "loss": 0.9122, + "step": 61950 + }, + { + "epoch": 4.58, + "learning_rate": 2.8049581929889103e-06, + "loss": 0.9365, + "step": 61951 + }, + { + "epoch": 4.58, + "learning_rate": 2.8046811871153503e-06, + "loss": 0.9231, + "step": 61952 + }, + { + "epoch": 4.58, + "learning_rate": 2.8044041926894363e-06, + "loss": 0.9654, + "step": 61953 + }, + { + "epoch": 4.58, + "learning_rate": 2.8041272097116156e-06, + "loss": 0.9454, + "step": 61954 + }, + { + "epoch": 4.58, + "learning_rate": 2.803850238182325e-06, + "loss": 0.9293, + "step": 61955 + }, + { + "epoch": 4.58, + "learning_rate": 2.8035732781020065e-06, + "loss": 0.7911, + "step": 61956 + }, + { + "epoch": 4.58, + "learning_rate": 2.8032963294710956e-06, + "loss": 1.0688, + "step": 61957 + }, + { + "epoch": 4.58, + "learning_rate": 2.8030193922900416e-06, + "loss": 1.1835, + "step": 61958 + }, + { + "epoch": 4.58, + "learning_rate": 2.8027424665592806e-06, + "loss": 1.0217, + "step": 61959 + }, + { + "epoch": 4.58, + "learning_rate": 2.802465552279253e-06, + "loss": 0.8631, + "step": 61960 + }, + { + "epoch": 4.58, + "learning_rate": 2.8021886494504003e-06, + "loss": 1.0199, + "step": 61961 + }, + { + "epoch": 4.58, + "learning_rate": 2.801911758073158e-06, + "loss": 1.1383, + "step": 61962 + }, + { + "epoch": 4.58, + "learning_rate": 2.8016348781479784e-06, + "loss": 0.8748, + "step": 61963 + }, + { + "epoch": 4.58, + "learning_rate": 2.801358009675289e-06, + "loss": 0.9353, + "step": 61964 + }, + { + "epoch": 4.58, + "learning_rate": 2.8010811526555383e-06, + "loss": 0.9574, + "step": 61965 + }, + { + "epoch": 4.58, + "learning_rate": 2.8008043070891612e-06, + "loss": 1.0093, + "step": 61966 + }, + { + "epoch": 4.58, + "learning_rate": 2.8005274729766085e-06, + "loss": 0.9574, + "step": 61967 + }, + { + "epoch": 4.58, + "learning_rate": 2.800250650318307e-06, + "loss": 0.9417, + "step": 61968 + }, + { + "epoch": 4.58, + "learning_rate": 2.7999738391147056e-06, + "loss": 0.8953, + "step": 61969 + }, + { + "epoch": 4.58, + "learning_rate": 2.7996970393662414e-06, + "loss": 1.0755, + "step": 61970 + }, + { + "epoch": 4.58, + "learning_rate": 2.7994202510733616e-06, + "loss": 1.1086, + "step": 61971 + }, + { + "epoch": 4.58, + "learning_rate": 2.7991434742364988e-06, + "loss": 1.0487, + "step": 61972 + }, + { + "epoch": 4.58, + "learning_rate": 2.7988667088560915e-06, + "loss": 0.9755, + "step": 61973 + }, + { + "epoch": 4.58, + "learning_rate": 2.798589954932588e-06, + "loss": 1.0398, + "step": 61974 + }, + { + "epoch": 4.58, + "learning_rate": 2.7983132124664246e-06, + "loss": 1.01, + "step": 61975 + }, + { + "epoch": 4.58, + "learning_rate": 2.7980364814580418e-06, + "loss": 1.0871, + "step": 61976 + }, + { + "epoch": 4.58, + "learning_rate": 2.7977597619078776e-06, + "loss": 0.991, + "step": 61977 + }, + { + "epoch": 4.58, + "learning_rate": 2.7974830538163764e-06, + "loss": 0.9855, + "step": 61978 + }, + { + "epoch": 4.58, + "learning_rate": 2.797206357183978e-06, + "loss": 1.0202, + "step": 61979 + }, + { + "epoch": 4.58, + "learning_rate": 2.7969296720111206e-06, + "loss": 1.1315, + "step": 61980 + }, + { + "epoch": 4.58, + "learning_rate": 2.7966529982982448e-06, + "loss": 0.8976, + "step": 61981 + }, + { + "epoch": 4.58, + "learning_rate": 2.7963763360457874e-06, + "loss": 1.0726, + "step": 61982 + }, + { + "epoch": 4.58, + "learning_rate": 2.7960996852541965e-06, + "loss": 0.9332, + "step": 61983 + }, + { + "epoch": 4.58, + "learning_rate": 2.7958230459239067e-06, + "loss": 0.9341, + "step": 61984 + }, + { + "epoch": 4.58, + "learning_rate": 2.7955464180553606e-06, + "loss": 1.0004, + "step": 61985 + }, + { + "epoch": 4.58, + "learning_rate": 2.795269801648993e-06, + "loss": 0.9457, + "step": 61986 + }, + { + "epoch": 4.58, + "learning_rate": 2.7949931967052513e-06, + "loss": 1.0419, + "step": 61987 + }, + { + "epoch": 4.58, + "learning_rate": 2.7947166032245733e-06, + "loss": 1.0156, + "step": 61988 + }, + { + "epoch": 4.58, + "learning_rate": 2.7944400212073974e-06, + "loss": 0.9367, + "step": 61989 + }, + { + "epoch": 4.58, + "learning_rate": 2.7941634506541636e-06, + "loss": 1.013, + "step": 61990 + }, + { + "epoch": 4.58, + "learning_rate": 2.7938868915653107e-06, + "loss": 1.0187, + "step": 61991 + }, + { + "epoch": 4.58, + "learning_rate": 2.7936103439412863e-06, + "loss": 0.9069, + "step": 61992 + }, + { + "epoch": 4.58, + "learning_rate": 2.793333807782519e-06, + "loss": 0.9634, + "step": 61993 + }, + { + "epoch": 4.58, + "learning_rate": 2.7930572830894566e-06, + "loss": 0.9749, + "step": 61994 + }, + { + "epoch": 4.58, + "learning_rate": 2.7927807698625377e-06, + "loss": 0.967, + "step": 61995 + }, + { + "epoch": 4.58, + "learning_rate": 2.792504268102202e-06, + "loss": 0.9849, + "step": 61996 + }, + { + "epoch": 4.58, + "learning_rate": 2.7922277778088847e-06, + "loss": 0.9091, + "step": 61997 + }, + { + "epoch": 4.58, + "learning_rate": 2.7919512989830333e-06, + "loss": 0.9626, + "step": 61998 + }, + { + "epoch": 4.58, + "learning_rate": 2.7916748316250842e-06, + "loss": 0.8921, + "step": 61999 + }, + { + "epoch": 4.58, + "learning_rate": 2.791398375735478e-06, + "loss": 1.0025, + "step": 62000 + }, + { + "epoch": 4.58, + "learning_rate": 2.791121931314653e-06, + "loss": 1.0125, + "step": 62001 + }, + { + "epoch": 4.58, + "learning_rate": 2.7908454983630473e-06, + "loss": 1.0151, + "step": 62002 + }, + { + "epoch": 4.58, + "learning_rate": 2.7905690768811066e-06, + "loss": 1.0263, + "step": 62003 + }, + { + "epoch": 4.58, + "learning_rate": 2.7902926668692664e-06, + "loss": 0.956, + "step": 62004 + }, + { + "epoch": 4.58, + "learning_rate": 2.7900162683279687e-06, + "loss": 0.9747, + "step": 62005 + }, + { + "epoch": 4.58, + "learning_rate": 2.789739881257647e-06, + "loss": 1.0501, + "step": 62006 + }, + { + "epoch": 4.58, + "learning_rate": 2.7894635056587504e-06, + "loss": 1.033, + "step": 62007 + }, + { + "epoch": 4.58, + "learning_rate": 2.7891871415317175e-06, + "loss": 0.9448, + "step": 62008 + }, + { + "epoch": 4.58, + "learning_rate": 2.7889107888769784e-06, + "loss": 1.0672, + "step": 62009 + }, + { + "epoch": 4.58, + "learning_rate": 2.788634447694981e-06, + "loss": 1.007, + "step": 62010 + }, + { + "epoch": 4.58, + "learning_rate": 2.788358117986161e-06, + "loss": 1.1453, + "step": 62011 + }, + { + "epoch": 4.58, + "learning_rate": 2.7880817997509656e-06, + "loss": 0.9655, + "step": 62012 + }, + { + "epoch": 4.58, + "learning_rate": 2.7878054929898235e-06, + "loss": 0.9315, + "step": 62013 + }, + { + "epoch": 4.58, + "learning_rate": 2.7875291977031825e-06, + "loss": 0.9112, + "step": 62014 + }, + { + "epoch": 4.58, + "learning_rate": 2.787252913891476e-06, + "loss": 0.9644, + "step": 62015 + }, + { + "epoch": 4.58, + "learning_rate": 2.7869766415551535e-06, + "loss": 1.0104, + "step": 62016 + }, + { + "epoch": 4.58, + "learning_rate": 2.7867003806946446e-06, + "loss": 1.1131, + "step": 62017 + }, + { + "epoch": 4.58, + "learning_rate": 2.786424131310389e-06, + "loss": 0.819, + "step": 62018 + }, + { + "epoch": 4.58, + "learning_rate": 2.786147893402832e-06, + "loss": 0.956, + "step": 62019 + }, + { + "epoch": 4.58, + "learning_rate": 2.785871666972412e-06, + "loss": 1.033, + "step": 62020 + }, + { + "epoch": 4.58, + "learning_rate": 2.785595452019565e-06, + "loss": 1.0668, + "step": 62021 + }, + { + "epoch": 4.58, + "learning_rate": 2.7853192485447313e-06, + "loss": 1.0732, + "step": 62022 + }, + { + "epoch": 4.58, + "learning_rate": 2.7850430565483534e-06, + "loss": 0.9682, + "step": 62023 + }, + { + "epoch": 4.58, + "learning_rate": 2.784766876030869e-06, + "loss": 0.8743, + "step": 62024 + }, + { + "epoch": 4.58, + "learning_rate": 2.7844907069927164e-06, + "loss": 1.0348, + "step": 62025 + }, + { + "epoch": 4.58, + "learning_rate": 2.784214549434334e-06, + "loss": 1.1366, + "step": 62026 + }, + { + "epoch": 4.58, + "learning_rate": 2.7839384033561657e-06, + "loss": 1.0196, + "step": 62027 + }, + { + "epoch": 4.58, + "learning_rate": 2.783662268758648e-06, + "loss": 0.9717, + "step": 62028 + }, + { + "epoch": 4.58, + "learning_rate": 2.7833861456422207e-06, + "loss": 1.0253, + "step": 62029 + }, + { + "epoch": 4.58, + "learning_rate": 2.7831100340073224e-06, + "loss": 0.9682, + "step": 62030 + }, + { + "epoch": 4.58, + "learning_rate": 2.78283393385439e-06, + "loss": 0.8809, + "step": 62031 + }, + { + "epoch": 4.58, + "learning_rate": 2.7825578451838685e-06, + "loss": 0.9086, + "step": 62032 + }, + { + "epoch": 4.58, + "learning_rate": 2.7822817679961942e-06, + "loss": 0.9383, + "step": 62033 + }, + { + "epoch": 4.58, + "learning_rate": 2.7820057022918066e-06, + "loss": 0.9939, + "step": 62034 + }, + { + "epoch": 4.58, + "learning_rate": 2.781729648071141e-06, + "loss": 1.0448, + "step": 62035 + }, + { + "epoch": 4.58, + "learning_rate": 2.781453605334644e-06, + "loss": 1.15, + "step": 62036 + }, + { + "epoch": 4.58, + "learning_rate": 2.781177574082754e-06, + "loss": 1.1168, + "step": 62037 + }, + { + "epoch": 4.58, + "learning_rate": 2.7809015543159e-06, + "loss": 0.978, + "step": 62038 + }, + { + "epoch": 4.58, + "learning_rate": 2.780625546034532e-06, + "loss": 1.0292, + "step": 62039 + }, + { + "epoch": 4.58, + "learning_rate": 2.7803495492390864e-06, + "loss": 0.9437, + "step": 62040 + }, + { + "epoch": 4.58, + "learning_rate": 2.7800735639300005e-06, + "loss": 0.9706, + "step": 62041 + }, + { + "epoch": 4.58, + "learning_rate": 2.7797975901077112e-06, + "loss": 1.0105, + "step": 62042 + }, + { + "epoch": 4.58, + "learning_rate": 2.7795216277726643e-06, + "loss": 0.9792, + "step": 62043 + }, + { + "epoch": 4.58, + "learning_rate": 2.779245676925295e-06, + "loss": 1.0268, + "step": 62044 + }, + { + "epoch": 4.58, + "learning_rate": 2.778969737566042e-06, + "loss": 0.9723, + "step": 62045 + }, + { + "epoch": 4.58, + "learning_rate": 2.7786938096953455e-06, + "loss": 1.014, + "step": 62046 + }, + { + "epoch": 4.58, + "learning_rate": 2.7784178933136397e-06, + "loss": 1.0053, + "step": 62047 + }, + { + "epoch": 4.58, + "learning_rate": 2.7781419884213713e-06, + "loss": 0.9398, + "step": 62048 + }, + { + "epoch": 4.58, + "learning_rate": 2.777866095018976e-06, + "loss": 1.0352, + "step": 62049 + }, + { + "epoch": 4.58, + "learning_rate": 2.777590213106892e-06, + "loss": 0.9805, + "step": 62050 + }, + { + "epoch": 4.58, + "learning_rate": 2.7773143426855565e-06, + "loss": 0.9367, + "step": 62051 + }, + { + "epoch": 4.58, + "learning_rate": 2.777038483755412e-06, + "loss": 1.0204, + "step": 62052 + }, + { + "epoch": 4.59, + "learning_rate": 2.7767626363168977e-06, + "loss": 1.0347, + "step": 62053 + }, + { + "epoch": 4.59, + "learning_rate": 2.7764868003704494e-06, + "loss": 1.0691, + "step": 62054 + }, + { + "epoch": 4.59, + "learning_rate": 2.7762109759165077e-06, + "loss": 0.9589, + "step": 62055 + }, + { + "epoch": 4.59, + "learning_rate": 2.7759351629555076e-06, + "loss": 1.0127, + "step": 62056 + }, + { + "epoch": 4.59, + "learning_rate": 2.775659361487898e-06, + "loss": 0.9543, + "step": 62057 + }, + { + "epoch": 4.59, + "learning_rate": 2.775383571514104e-06, + "loss": 0.9399, + "step": 62058 + }, + { + "epoch": 4.59, + "learning_rate": 2.7751077930345747e-06, + "loss": 0.9715, + "step": 62059 + }, + { + "epoch": 4.59, + "learning_rate": 2.774832026049743e-06, + "loss": 1.0824, + "step": 62060 + }, + { + "epoch": 4.59, + "learning_rate": 2.774556270560057e-06, + "loss": 1.0164, + "step": 62061 + }, + { + "epoch": 4.59, + "learning_rate": 2.774280526565941e-06, + "loss": 1.0152, + "step": 62062 + }, + { + "epoch": 4.59, + "learning_rate": 2.774004794067845e-06, + "loss": 0.9943, + "step": 62063 + }, + { + "epoch": 4.59, + "learning_rate": 2.773729073066204e-06, + "loss": 0.9619, + "step": 62064 + }, + { + "epoch": 4.59, + "learning_rate": 2.7734533635614568e-06, + "loss": 0.9767, + "step": 62065 + }, + { + "epoch": 4.59, + "learning_rate": 2.773177665554041e-06, + "loss": 1.0156, + "step": 62066 + }, + { + "epoch": 4.59, + "learning_rate": 2.7729019790443945e-06, + "loss": 0.9821, + "step": 62067 + }, + { + "epoch": 4.59, + "learning_rate": 2.7726263040329594e-06, + "loss": 1.0124, + "step": 62068 + }, + { + "epoch": 4.59, + "learning_rate": 2.7723506405201727e-06, + "loss": 1.0146, + "step": 62069 + }, + { + "epoch": 4.59, + "learning_rate": 2.7720749885064736e-06, + "loss": 1.022, + "step": 62070 + }, + { + "epoch": 4.59, + "learning_rate": 2.7717993479922954e-06, + "loss": 0.9888, + "step": 62071 + }, + { + "epoch": 4.59, + "learning_rate": 2.771523718978084e-06, + "loss": 1.0272, + "step": 62072 + }, + { + "epoch": 4.59, + "learning_rate": 2.7712481014642767e-06, + "loss": 1.0119, + "step": 62073 + }, + { + "epoch": 4.59, + "learning_rate": 2.7709724954513086e-06, + "loss": 1.0903, + "step": 62074 + }, + { + "epoch": 4.59, + "learning_rate": 2.77069690093962e-06, + "loss": 0.9535, + "step": 62075 + }, + { + "epoch": 4.59, + "learning_rate": 2.770421317929647e-06, + "loss": 0.9902, + "step": 62076 + }, + { + "epoch": 4.59, + "learning_rate": 2.7701457464218327e-06, + "loss": 1.0343, + "step": 62077 + }, + { + "epoch": 4.59, + "learning_rate": 2.769870186416613e-06, + "loss": 1.1041, + "step": 62078 + }, + { + "epoch": 4.59, + "learning_rate": 2.7695946379144268e-06, + "loss": 1.0475, + "step": 62079 + }, + { + "epoch": 4.59, + "learning_rate": 2.7693191009157093e-06, + "loss": 1.0307, + "step": 62080 + }, + { + "epoch": 4.59, + "learning_rate": 2.7690435754209045e-06, + "loss": 1.0648, + "step": 62081 + }, + { + "epoch": 4.59, + "learning_rate": 2.768768061430447e-06, + "loss": 1.0275, + "step": 62082 + }, + { + "epoch": 4.59, + "learning_rate": 2.7684925589447765e-06, + "loss": 0.9613, + "step": 62083 + }, + { + "epoch": 4.59, + "learning_rate": 2.7682170679643315e-06, + "loss": 0.9812, + "step": 62084 + }, + { + "epoch": 4.59, + "learning_rate": 2.767941588489549e-06, + "loss": 1.0378, + "step": 62085 + }, + { + "epoch": 4.59, + "learning_rate": 2.7676661205208686e-06, + "loss": 0.9385, + "step": 62086 + }, + { + "epoch": 4.59, + "learning_rate": 2.7673906640587244e-06, + "loss": 0.9735, + "step": 62087 + }, + { + "epoch": 4.59, + "learning_rate": 2.767115219103561e-06, + "loss": 0.9022, + "step": 62088 + }, + { + "epoch": 4.59, + "learning_rate": 2.7668397856558137e-06, + "loss": 0.9877, + "step": 62089 + }, + { + "epoch": 4.59, + "learning_rate": 2.7665643637159223e-06, + "loss": 0.9963, + "step": 62090 + }, + { + "epoch": 4.59, + "learning_rate": 2.7662889532843186e-06, + "loss": 0.9439, + "step": 62091 + }, + { + "epoch": 4.59, + "learning_rate": 2.76601355436145e-06, + "loss": 0.9679, + "step": 62092 + }, + { + "epoch": 4.59, + "learning_rate": 2.76573816694775e-06, + "loss": 1.0554, + "step": 62093 + }, + { + "epoch": 4.59, + "learning_rate": 2.765462791043657e-06, + "loss": 1.001, + "step": 62094 + }, + { + "epoch": 4.59, + "learning_rate": 2.7651874266496083e-06, + "loss": 0.919, + "step": 62095 + }, + { + "epoch": 4.59, + "learning_rate": 2.7649120737660406e-06, + "loss": 0.9001, + "step": 62096 + }, + { + "epoch": 4.59, + "learning_rate": 2.7646367323933977e-06, + "loss": 0.9496, + "step": 62097 + }, + { + "epoch": 4.59, + "learning_rate": 2.7643614025321143e-06, + "loss": 1.0097, + "step": 62098 + }, + { + "epoch": 4.59, + "learning_rate": 2.7640860841826278e-06, + "loss": 1.0166, + "step": 62099 + }, + { + "epoch": 4.59, + "learning_rate": 2.7638107773453736e-06, + "loss": 1.0153, + "step": 62100 + }, + { + "epoch": 4.59, + "learning_rate": 2.763535482020796e-06, + "loss": 1.03, + "step": 62101 + }, + { + "epoch": 4.59, + "learning_rate": 2.763260198209334e-06, + "loss": 1.0171, + "step": 62102 + }, + { + "epoch": 4.59, + "learning_rate": 2.762984925911415e-06, + "loss": 0.9733, + "step": 62103 + }, + { + "epoch": 4.59, + "learning_rate": 2.7627096651274853e-06, + "loss": 1.0763, + "step": 62104 + }, + { + "epoch": 4.59, + "learning_rate": 2.7624344158579788e-06, + "loss": 0.9454, + "step": 62105 + }, + { + "epoch": 4.59, + "learning_rate": 2.7621591781033418e-06, + "loss": 0.8858, + "step": 62106 + }, + { + "epoch": 4.59, + "learning_rate": 2.7618839518639995e-06, + "loss": 0.949, + "step": 62107 + }, + { + "epoch": 4.59, + "learning_rate": 2.7616087371403997e-06, + "loss": 1.1025, + "step": 62108 + }, + { + "epoch": 4.59, + "learning_rate": 2.761333533932974e-06, + "loss": 1.1564, + "step": 62109 + }, + { + "epoch": 4.59, + "learning_rate": 2.761058342242169e-06, + "loss": 1.0995, + "step": 62110 + }, + { + "epoch": 4.59, + "learning_rate": 2.760783162068413e-06, + "loss": 0.9749, + "step": 62111 + }, + { + "epoch": 4.59, + "learning_rate": 2.760507993412145e-06, + "loss": 0.969, + "step": 62112 + }, + { + "epoch": 4.59, + "learning_rate": 2.7602328362738084e-06, + "loss": 1.0173, + "step": 62113 + }, + { + "epoch": 4.59, + "learning_rate": 2.7599576906538373e-06, + "loss": 0.9839, + "step": 62114 + }, + { + "epoch": 4.59, + "learning_rate": 2.7596825565526698e-06, + "loss": 1.0237, + "step": 62115 + }, + { + "epoch": 4.59, + "learning_rate": 2.7594074339707412e-06, + "loss": 1.0591, + "step": 62116 + }, + { + "epoch": 4.59, + "learning_rate": 2.759132322908494e-06, + "loss": 0.9248, + "step": 62117 + }, + { + "epoch": 4.59, + "learning_rate": 2.7588572233663645e-06, + "loss": 1.0245, + "step": 62118 + }, + { + "epoch": 4.59, + "learning_rate": 2.7585821353447897e-06, + "loss": 0.9029, + "step": 62119 + }, + { + "epoch": 4.59, + "learning_rate": 2.7583070588442063e-06, + "loss": 1.0817, + "step": 62120 + }, + { + "epoch": 4.59, + "learning_rate": 2.75803199386505e-06, + "loss": 0.9412, + "step": 62121 + }, + { + "epoch": 4.59, + "learning_rate": 2.7577569404077644e-06, + "loss": 1.1377, + "step": 62122 + }, + { + "epoch": 4.59, + "learning_rate": 2.7574818984727838e-06, + "loss": 1.0028, + "step": 62123 + }, + { + "epoch": 4.59, + "learning_rate": 2.757206868060546e-06, + "loss": 1.0524, + "step": 62124 + }, + { + "epoch": 4.59, + "learning_rate": 2.756931849171485e-06, + "loss": 1.0047, + "step": 62125 + }, + { + "epoch": 4.59, + "learning_rate": 2.756656841806045e-06, + "loss": 0.9579, + "step": 62126 + }, + { + "epoch": 4.59, + "learning_rate": 2.756381845964661e-06, + "loss": 1.039, + "step": 62127 + }, + { + "epoch": 4.59, + "learning_rate": 2.7561068616477693e-06, + "loss": 0.9982, + "step": 62128 + }, + { + "epoch": 4.59, + "learning_rate": 2.7558318888558044e-06, + "loss": 1.0119, + "step": 62129 + }, + { + "epoch": 4.59, + "learning_rate": 2.7555569275892135e-06, + "loss": 0.9374, + "step": 62130 + }, + { + "epoch": 4.59, + "learning_rate": 2.755281977848426e-06, + "loss": 1.0366, + "step": 62131 + }, + { + "epoch": 4.59, + "learning_rate": 2.755007039633877e-06, + "loss": 0.914, + "step": 62132 + }, + { + "epoch": 4.59, + "learning_rate": 2.7547321129460114e-06, + "loss": 0.9905, + "step": 62133 + }, + { + "epoch": 4.59, + "learning_rate": 2.754457197785264e-06, + "loss": 1.0122, + "step": 62134 + }, + { + "epoch": 4.59, + "learning_rate": 2.7541822941520714e-06, + "loss": 0.9581, + "step": 62135 + }, + { + "epoch": 4.59, + "learning_rate": 2.753907402046867e-06, + "loss": 1.0082, + "step": 62136 + }, + { + "epoch": 4.59, + "learning_rate": 2.7536325214700964e-06, + "loss": 0.985, + "step": 62137 + }, + { + "epoch": 4.59, + "learning_rate": 2.753357652422193e-06, + "loss": 1.0776, + "step": 62138 + }, + { + "epoch": 4.59, + "learning_rate": 2.753082794903593e-06, + "loss": 1.1055, + "step": 62139 + }, + { + "epoch": 4.59, + "learning_rate": 2.752807948914734e-06, + "loss": 1.0218, + "step": 62140 + }, + { + "epoch": 4.59, + "learning_rate": 2.752533114456052e-06, + "loss": 1.0381, + "step": 62141 + }, + { + "epoch": 4.59, + "learning_rate": 2.7522582915279894e-06, + "loss": 1.0254, + "step": 62142 + }, + { + "epoch": 4.59, + "learning_rate": 2.7519834801309795e-06, + "loss": 0.9379, + "step": 62143 + }, + { + "epoch": 4.59, + "learning_rate": 2.75170868026546e-06, + "loss": 0.9903, + "step": 62144 + }, + { + "epoch": 4.59, + "learning_rate": 2.751433891931866e-06, + "loss": 0.9805, + "step": 62145 + }, + { + "epoch": 4.59, + "learning_rate": 2.751159115130639e-06, + "loss": 0.8801, + "step": 62146 + }, + { + "epoch": 4.59, + "learning_rate": 2.7508843498622152e-06, + "loss": 0.973, + "step": 62147 + }, + { + "epoch": 4.59, + "learning_rate": 2.7506095961270296e-06, + "loss": 0.9271, + "step": 62148 + }, + { + "epoch": 4.59, + "learning_rate": 2.7503348539255214e-06, + "loss": 0.982, + "step": 62149 + }, + { + "epoch": 4.59, + "learning_rate": 2.7500601232581236e-06, + "loss": 1.122, + "step": 62150 + }, + { + "epoch": 4.59, + "learning_rate": 2.7497854041252816e-06, + "loss": 0.9642, + "step": 62151 + }, + { + "epoch": 4.59, + "learning_rate": 2.749510696527422e-06, + "loss": 1.0592, + "step": 62152 + }, + { + "epoch": 4.59, + "learning_rate": 2.7492360004649898e-06, + "loss": 1.004, + "step": 62153 + }, + { + "epoch": 4.59, + "learning_rate": 2.7489613159384166e-06, + "loss": 1.0956, + "step": 62154 + }, + { + "epoch": 4.59, + "learning_rate": 2.7486866429481484e-06, + "loss": 0.9447, + "step": 62155 + }, + { + "epoch": 4.59, + "learning_rate": 2.7484119814946096e-06, + "loss": 1.0267, + "step": 62156 + }, + { + "epoch": 4.59, + "learning_rate": 2.7481373315782465e-06, + "loss": 0.9874, + "step": 62157 + }, + { + "epoch": 4.59, + "learning_rate": 2.7478626931994933e-06, + "loss": 0.9107, + "step": 62158 + }, + { + "epoch": 4.59, + "learning_rate": 2.7475880663587874e-06, + "loss": 0.9894, + "step": 62159 + }, + { + "epoch": 4.59, + "learning_rate": 2.7473134510565646e-06, + "loss": 0.976, + "step": 62160 + }, + { + "epoch": 4.59, + "learning_rate": 2.7470388472932594e-06, + "loss": 1.036, + "step": 62161 + }, + { + "epoch": 4.59, + "learning_rate": 2.746764255069314e-06, + "loss": 0.9812, + "step": 62162 + }, + { + "epoch": 4.59, + "learning_rate": 2.746489674385163e-06, + "loss": 0.8663, + "step": 62163 + }, + { + "epoch": 4.59, + "learning_rate": 2.7462151052412444e-06, + "loss": 0.8828, + "step": 62164 + }, + { + "epoch": 4.59, + "learning_rate": 2.7459405476379895e-06, + "loss": 0.9577, + "step": 62165 + }, + { + "epoch": 4.59, + "learning_rate": 2.745666001575843e-06, + "loss": 0.8914, + "step": 62166 + }, + { + "epoch": 4.59, + "learning_rate": 2.7453914670552385e-06, + "loss": 1.1499, + "step": 62167 + }, + { + "epoch": 4.59, + "learning_rate": 2.7451169440766114e-06, + "loss": 0.9857, + "step": 62168 + }, + { + "epoch": 4.59, + "learning_rate": 2.7448424326404e-06, + "loss": 1.0196, + "step": 62169 + }, + { + "epoch": 4.59, + "learning_rate": 2.7445679327470374e-06, + "loss": 1.0701, + "step": 62170 + }, + { + "epoch": 4.59, + "learning_rate": 2.744293444396966e-06, + "loss": 1.0205, + "step": 62171 + }, + { + "epoch": 4.59, + "learning_rate": 2.74401896759062e-06, + "loss": 0.9777, + "step": 62172 + }, + { + "epoch": 4.59, + "learning_rate": 2.743744502328437e-06, + "loss": 1.0281, + "step": 62173 + }, + { + "epoch": 4.59, + "learning_rate": 2.743470048610849e-06, + "loss": 0.9387, + "step": 62174 + }, + { + "epoch": 4.59, + "learning_rate": 2.7431956064383025e-06, + "loss": 1.0737, + "step": 62175 + }, + { + "epoch": 4.59, + "learning_rate": 2.742921175811225e-06, + "loss": 0.9215, + "step": 62176 + }, + { + "epoch": 4.59, + "learning_rate": 2.7426467567300518e-06, + "loss": 1.0247, + "step": 62177 + }, + { + "epoch": 4.59, + "learning_rate": 2.742372349195228e-06, + "loss": 0.996, + "step": 62178 + }, + { + "epoch": 4.59, + "learning_rate": 2.7420979532071856e-06, + "loss": 1.0146, + "step": 62179 + }, + { + "epoch": 4.59, + "learning_rate": 2.741823568766362e-06, + "loss": 1.0172, + "step": 62180 + }, + { + "epoch": 4.59, + "learning_rate": 2.74154919587319e-06, + "loss": 1.1175, + "step": 62181 + }, + { + "epoch": 4.59, + "learning_rate": 2.7412748345281126e-06, + "loss": 1.0259, + "step": 62182 + }, + { + "epoch": 4.59, + "learning_rate": 2.741000484731563e-06, + "loss": 1.0728, + "step": 62183 + }, + { + "epoch": 4.59, + "learning_rate": 2.7407261464839773e-06, + "loss": 1.0634, + "step": 62184 + }, + { + "epoch": 4.59, + "learning_rate": 2.74045181978579e-06, + "loss": 0.9965, + "step": 62185 + }, + { + "epoch": 4.59, + "learning_rate": 2.7401775046374424e-06, + "loss": 1.1225, + "step": 62186 + }, + { + "epoch": 4.59, + "learning_rate": 2.739903201039369e-06, + "loss": 1.0368, + "step": 62187 + }, + { + "epoch": 4.6, + "learning_rate": 2.739628908992006e-06, + "loss": 0.9692, + "step": 62188 + }, + { + "epoch": 4.6, + "learning_rate": 2.7393546284957884e-06, + "loss": 0.9829, + "step": 62189 + }, + { + "epoch": 4.6, + "learning_rate": 2.739080359551152e-06, + "loss": 0.9538, + "step": 62190 + }, + { + "epoch": 4.6, + "learning_rate": 2.7388061021585365e-06, + "loss": 0.9656, + "step": 62191 + }, + { + "epoch": 4.6, + "learning_rate": 2.7385318563183773e-06, + "loss": 1.0178, + "step": 62192 + }, + { + "epoch": 4.6, + "learning_rate": 2.7382576220311106e-06, + "loss": 0.9841, + "step": 62193 + }, + { + "epoch": 4.6, + "learning_rate": 2.7379833992971683e-06, + "loss": 0.9529, + "step": 62194 + }, + { + "epoch": 4.6, + "learning_rate": 2.737709188116994e-06, + "loss": 0.8764, + "step": 62195 + }, + { + "epoch": 4.6, + "learning_rate": 2.7374349884910234e-06, + "loss": 0.986, + "step": 62196 + }, + { + "epoch": 4.6, + "learning_rate": 2.7371608004196836e-06, + "loss": 1.0429, + "step": 62197 + }, + { + "epoch": 4.6, + "learning_rate": 2.73688662390342e-06, + "loss": 0.9539, + "step": 62198 + }, + { + "epoch": 4.6, + "learning_rate": 2.7366124589426624e-06, + "loss": 1.0285, + "step": 62199 + }, + { + "epoch": 4.6, + "learning_rate": 2.7363383055378578e-06, + "loss": 1.0759, + "step": 62200 + }, + { + "epoch": 4.6, + "learning_rate": 2.736064163689428e-06, + "loss": 1.0633, + "step": 62201 + }, + { + "epoch": 4.6, + "learning_rate": 2.7357900333978195e-06, + "loss": 1.0237, + "step": 62202 + }, + { + "epoch": 4.6, + "learning_rate": 2.7355159146634625e-06, + "loss": 0.9738, + "step": 62203 + }, + { + "epoch": 4.6, + "learning_rate": 2.7352418074868015e-06, + "loss": 1.0736, + "step": 62204 + }, + { + "epoch": 4.6, + "learning_rate": 2.7349677118682625e-06, + "loss": 0.9743, + "step": 62205 + }, + { + "epoch": 4.6, + "learning_rate": 2.7346936278082847e-06, + "loss": 0.9985, + "step": 62206 + }, + { + "epoch": 4.6, + "learning_rate": 2.734419555307308e-06, + "loss": 0.9124, + "step": 62207 + }, + { + "epoch": 4.6, + "learning_rate": 2.734145494365765e-06, + "loss": 0.9398, + "step": 62208 + }, + { + "epoch": 4.6, + "learning_rate": 2.7338714449840943e-06, + "loss": 1.0839, + "step": 62209 + }, + { + "epoch": 4.6, + "learning_rate": 2.733597407162726e-06, + "loss": 0.9872, + "step": 62210 + }, + { + "epoch": 4.6, + "learning_rate": 2.733323380902103e-06, + "loss": 0.9527, + "step": 62211 + }, + { + "epoch": 4.6, + "learning_rate": 2.733049366202659e-06, + "loss": 0.9656, + "step": 62212 + }, + { + "epoch": 4.6, + "learning_rate": 2.732775363064829e-06, + "loss": 1.0486, + "step": 62213 + }, + { + "epoch": 4.6, + "learning_rate": 2.7325013714890502e-06, + "loss": 1.0496, + "step": 62214 + }, + { + "epoch": 4.6, + "learning_rate": 2.7322273914757536e-06, + "loss": 0.9014, + "step": 62215 + }, + { + "epoch": 4.6, + "learning_rate": 2.7319534230253864e-06, + "loss": 1.1123, + "step": 62216 + }, + { + "epoch": 4.6, + "learning_rate": 2.7316794661383703e-06, + "loss": 1.0087, + "step": 62217 + }, + { + "epoch": 4.6, + "learning_rate": 2.7314055208151522e-06, + "loss": 0.9188, + "step": 62218 + }, + { + "epoch": 4.6, + "learning_rate": 2.7311315870561607e-06, + "loss": 1.0841, + "step": 62219 + }, + { + "epoch": 4.6, + "learning_rate": 2.73085766486184e-06, + "loss": 0.9196, + "step": 62220 + }, + { + "epoch": 4.6, + "learning_rate": 2.730583754232615e-06, + "loss": 1.0422, + "step": 62221 + }, + { + "epoch": 4.6, + "learning_rate": 2.730309855168931e-06, + "loss": 0.9931, + "step": 62222 + }, + { + "epoch": 4.6, + "learning_rate": 2.7300359676712197e-06, + "loss": 0.9862, + "step": 62223 + }, + { + "epoch": 4.6, + "learning_rate": 2.729762091739917e-06, + "loss": 1.0114, + "step": 62224 + }, + { + "epoch": 4.6, + "learning_rate": 2.7294882273754584e-06, + "loss": 1.0704, + "step": 62225 + }, + { + "epoch": 4.6, + "learning_rate": 2.7292143745782786e-06, + "loss": 1.0131, + "step": 62226 + }, + { + "epoch": 4.6, + "learning_rate": 2.728940533348816e-06, + "loss": 1.0312, + "step": 62227 + }, + { + "epoch": 4.6, + "learning_rate": 2.728666703687506e-06, + "loss": 0.9942, + "step": 62228 + }, + { + "epoch": 4.6, + "learning_rate": 2.7283928855947826e-06, + "loss": 0.959, + "step": 62229 + }, + { + "epoch": 4.6, + "learning_rate": 2.728119079071079e-06, + "loss": 0.8677, + "step": 62230 + }, + { + "epoch": 4.6, + "learning_rate": 2.7278452841168367e-06, + "loss": 1.1033, + "step": 62231 + }, + { + "epoch": 4.6, + "learning_rate": 2.72757150073249e-06, + "loss": 1.0838, + "step": 62232 + }, + { + "epoch": 4.6, + "learning_rate": 2.727297728918471e-06, + "loss": 1.0551, + "step": 62233 + }, + { + "epoch": 4.6, + "learning_rate": 2.7270239686752187e-06, + "loss": 1.0131, + "step": 62234 + }, + { + "epoch": 4.6, + "learning_rate": 2.726750220003164e-06, + "loss": 0.997, + "step": 62235 + }, + { + "epoch": 4.6, + "learning_rate": 2.726476482902748e-06, + "loss": 0.95, + "step": 62236 + }, + { + "epoch": 4.6, + "learning_rate": 2.7262027573744042e-06, + "loss": 0.9657, + "step": 62237 + }, + { + "epoch": 4.6, + "learning_rate": 2.725929043418568e-06, + "loss": 1.0317, + "step": 62238 + }, + { + "epoch": 4.6, + "learning_rate": 2.725655341035671e-06, + "loss": 1.1072, + "step": 62239 + }, + { + "epoch": 4.6, + "learning_rate": 2.7253816502261556e-06, + "loss": 0.8812, + "step": 62240 + }, + { + "epoch": 4.6, + "learning_rate": 2.7251079709904537e-06, + "loss": 1.1017, + "step": 62241 + }, + { + "epoch": 4.6, + "learning_rate": 2.7248343033290013e-06, + "loss": 1.0554, + "step": 62242 + }, + { + "epoch": 4.6, + "learning_rate": 2.724560647242234e-06, + "loss": 0.9684, + "step": 62243 + }, + { + "epoch": 4.6, + "learning_rate": 2.7242870027305824e-06, + "loss": 0.9423, + "step": 62244 + }, + { + "epoch": 4.6, + "learning_rate": 2.7240133697944925e-06, + "loss": 0.9847, + "step": 62245 + }, + { + "epoch": 4.6, + "learning_rate": 2.7237397484343873e-06, + "loss": 1.0507, + "step": 62246 + }, + { + "epoch": 4.6, + "learning_rate": 2.723466138650711e-06, + "loss": 1.0761, + "step": 62247 + }, + { + "epoch": 4.6, + "learning_rate": 2.723192540443893e-06, + "loss": 0.868, + "step": 62248 + }, + { + "epoch": 4.6, + "learning_rate": 2.7229189538143775e-06, + "loss": 1.0123, + "step": 62249 + }, + { + "epoch": 4.6, + "learning_rate": 2.7226453787625885e-06, + "loss": 1.006, + "step": 62250 + }, + { + "epoch": 4.6, + "learning_rate": 2.722371815288969e-06, + "loss": 1.0718, + "step": 62251 + }, + { + "epoch": 4.6, + "learning_rate": 2.722098263393952e-06, + "loss": 0.9276, + "step": 62252 + }, + { + "epoch": 4.6, + "learning_rate": 2.7218247230779727e-06, + "loss": 1.1953, + "step": 62253 + }, + { + "epoch": 4.6, + "learning_rate": 2.7215511943414653e-06, + "loss": 1.0006, + "step": 62254 + }, + { + "epoch": 4.6, + "learning_rate": 2.7212776771848636e-06, + "loss": 0.9211, + "step": 62255 + }, + { + "epoch": 4.6, + "learning_rate": 2.721004171608608e-06, + "loss": 0.9102, + "step": 62256 + }, + { + "epoch": 4.6, + "learning_rate": 2.7207306776131293e-06, + "loss": 0.9965, + "step": 62257 + }, + { + "epoch": 4.6, + "learning_rate": 2.720457195198866e-06, + "loss": 0.9344, + "step": 62258 + }, + { + "epoch": 4.6, + "learning_rate": 2.7201837243662465e-06, + "loss": 0.9399, + "step": 62259 + }, + { + "epoch": 4.6, + "learning_rate": 2.719910265115714e-06, + "loss": 1.0039, + "step": 62260 + }, + { + "epoch": 4.6, + "learning_rate": 2.719636817447703e-06, + "loss": 1.0533, + "step": 62261 + }, + { + "epoch": 4.6, + "learning_rate": 2.7193633813626396e-06, + "loss": 0.9921, + "step": 62262 + }, + { + "epoch": 4.6, + "learning_rate": 2.7190899568609687e-06, + "loss": 1.0481, + "step": 62263 + }, + { + "epoch": 4.6, + "learning_rate": 2.718816543943118e-06, + "loss": 1.0172, + "step": 62264 + }, + { + "epoch": 4.6, + "learning_rate": 2.718543142609532e-06, + "loss": 1.0398, + "step": 62265 + }, + { + "epoch": 4.6, + "learning_rate": 2.718269752860634e-06, + "loss": 1.0143, + "step": 62266 + }, + { + "epoch": 4.6, + "learning_rate": 2.7179963746968674e-06, + "loss": 0.9583, + "step": 62267 + }, + { + "epoch": 4.6, + "learning_rate": 2.717723008118661e-06, + "loss": 0.9973, + "step": 62268 + }, + { + "epoch": 4.6, + "learning_rate": 2.7174496531264595e-06, + "loss": 0.8783, + "step": 62269 + }, + { + "epoch": 4.6, + "learning_rate": 2.717176309720688e-06, + "loss": 1.0026, + "step": 62270 + }, + { + "epoch": 4.6, + "learning_rate": 2.716902977901782e-06, + "loss": 1.0701, + "step": 62271 + }, + { + "epoch": 4.6, + "learning_rate": 2.716629657670182e-06, + "loss": 0.9986, + "step": 62272 + }, + { + "epoch": 4.6, + "learning_rate": 2.71635634902632e-06, + "loss": 0.9201, + "step": 62273 + }, + { + "epoch": 4.6, + "learning_rate": 2.7160830519706317e-06, + "loss": 1.0122, + "step": 62274 + }, + { + "epoch": 4.6, + "learning_rate": 2.7158097665035475e-06, + "loss": 1.068, + "step": 62275 + }, + { + "epoch": 4.6, + "learning_rate": 2.715536492625509e-06, + "loss": 1.037, + "step": 62276 + }, + { + "epoch": 4.6, + "learning_rate": 2.7152632303369475e-06, + "loss": 0.9613, + "step": 62277 + }, + { + "epoch": 4.6, + "learning_rate": 2.7149899796382974e-06, + "loss": 1.0892, + "step": 62278 + }, + { + "epoch": 4.6, + "learning_rate": 2.7147167405299946e-06, + "loss": 0.8559, + "step": 62279 + }, + { + "epoch": 4.6, + "learning_rate": 2.7144435130124704e-06, + "loss": 0.962, + "step": 62280 + }, + { + "epoch": 4.6, + "learning_rate": 2.7141702970861648e-06, + "loss": 1.0655, + "step": 62281 + }, + { + "epoch": 4.6, + "learning_rate": 2.7138970927515097e-06, + "loss": 1.0026, + "step": 62282 + }, + { + "epoch": 4.6, + "learning_rate": 2.7136239000089413e-06, + "loss": 1.0814, + "step": 62283 + }, + { + "epoch": 4.6, + "learning_rate": 2.713350718858889e-06, + "loss": 0.93, + "step": 62284 + }, + { + "epoch": 4.6, + "learning_rate": 2.7130775493017946e-06, + "loss": 1.0063, + "step": 62285 + }, + { + "epoch": 4.6, + "learning_rate": 2.71280439133809e-06, + "loss": 0.987, + "step": 62286 + }, + { + "epoch": 4.6, + "learning_rate": 2.7125312449682096e-06, + "loss": 0.9802, + "step": 62287 + }, + { + "epoch": 4.6, + "learning_rate": 2.7122581101925837e-06, + "loss": 0.9647, + "step": 62288 + }, + { + "epoch": 4.6, + "learning_rate": 2.7119849870116542e-06, + "loss": 1.0541, + "step": 62289 + }, + { + "epoch": 4.6, + "learning_rate": 2.7117118754258553e-06, + "loss": 0.9945, + "step": 62290 + }, + { + "epoch": 4.6, + "learning_rate": 2.711438775435613e-06, + "loss": 0.9968, + "step": 62291 + }, + { + "epoch": 4.6, + "learning_rate": 2.7111656870413705e-06, + "loss": 0.8831, + "step": 62292 + }, + { + "epoch": 4.6, + "learning_rate": 2.7108926102435553e-06, + "loss": 0.9945, + "step": 62293 + }, + { + "epoch": 4.6, + "learning_rate": 2.710619545042612e-06, + "loss": 1.0099, + "step": 62294 + }, + { + "epoch": 4.6, + "learning_rate": 2.7103464914389633e-06, + "loss": 0.9936, + "step": 62295 + }, + { + "epoch": 4.6, + "learning_rate": 2.7100734494330518e-06, + "loss": 1.0408, + "step": 62296 + }, + { + "epoch": 4.6, + "learning_rate": 2.709800419025308e-06, + "loss": 0.967, + "step": 62297 + }, + { + "epoch": 4.6, + "learning_rate": 2.7095274002161687e-06, + "loss": 1.0956, + "step": 62298 + }, + { + "epoch": 4.6, + "learning_rate": 2.7092543930060665e-06, + "loss": 1.0318, + "step": 62299 + }, + { + "epoch": 4.6, + "learning_rate": 2.708981397395434e-06, + "loss": 1.0009, + "step": 62300 + }, + { + "epoch": 4.6, + "learning_rate": 2.7087084133847096e-06, + "loss": 1.0643, + "step": 62301 + }, + { + "epoch": 4.6, + "learning_rate": 2.7084354409743263e-06, + "loss": 1.027, + "step": 62302 + }, + { + "epoch": 4.6, + "learning_rate": 2.708162480164718e-06, + "loss": 1.0459, + "step": 62303 + }, + { + "epoch": 4.6, + "learning_rate": 2.7078895309563157e-06, + "loss": 0.9261, + "step": 62304 + }, + { + "epoch": 4.6, + "learning_rate": 2.707616593349559e-06, + "loss": 1.0252, + "step": 62305 + }, + { + "epoch": 4.6, + "learning_rate": 2.7073436673448818e-06, + "loss": 0.9907, + "step": 62306 + }, + { + "epoch": 4.6, + "learning_rate": 2.7070707529427156e-06, + "loss": 0.994, + "step": 62307 + }, + { + "epoch": 4.6, + "learning_rate": 2.7067978501434955e-06, + "loss": 1.0138, + "step": 62308 + }, + { + "epoch": 4.6, + "learning_rate": 2.7065249589476515e-06, + "loss": 0.9975, + "step": 62309 + }, + { + "epoch": 4.6, + "learning_rate": 2.706252079355629e-06, + "loss": 1.1208, + "step": 62310 + }, + { + "epoch": 4.6, + "learning_rate": 2.7059792113678496e-06, + "loss": 1.0323, + "step": 62311 + }, + { + "epoch": 4.6, + "learning_rate": 2.705706354984755e-06, + "loss": 1.072, + "step": 62312 + }, + { + "epoch": 4.6, + "learning_rate": 2.7054335102067753e-06, + "loss": 1.0693, + "step": 62313 + }, + { + "epoch": 4.6, + "learning_rate": 2.7051606770343518e-06, + "loss": 1.0017, + "step": 62314 + }, + { + "epoch": 4.6, + "learning_rate": 2.704887855467907e-06, + "loss": 1.0656, + "step": 62315 + }, + { + "epoch": 4.6, + "learning_rate": 2.7046150455078845e-06, + "loss": 1.0599, + "step": 62316 + }, + { + "epoch": 4.6, + "learning_rate": 2.7043422471547154e-06, + "loss": 1.0071, + "step": 62317 + }, + { + "epoch": 4.6, + "learning_rate": 2.7040694604088334e-06, + "loss": 1.1076, + "step": 62318 + }, + { + "epoch": 4.6, + "learning_rate": 2.703796685270672e-06, + "loss": 1.0786, + "step": 62319 + }, + { + "epoch": 4.6, + "learning_rate": 2.703523921740663e-06, + "loss": 0.9772, + "step": 62320 + }, + { + "epoch": 4.6, + "learning_rate": 2.7032511698192467e-06, + "loss": 1.0008, + "step": 62321 + }, + { + "epoch": 4.6, + "learning_rate": 2.7029784295068516e-06, + "loss": 1.0182, + "step": 62322 + }, + { + "epoch": 4.61, + "learning_rate": 2.7027057008039146e-06, + "loss": 0.9356, + "step": 62323 + }, + { + "epoch": 4.61, + "learning_rate": 2.7024329837108653e-06, + "loss": 0.949, + "step": 62324 + }, + { + "epoch": 4.61, + "learning_rate": 2.7021602782281444e-06, + "loss": 0.9067, + "step": 62325 + }, + { + "epoch": 4.61, + "learning_rate": 2.701887584356181e-06, + "loss": 0.8895, + "step": 62326 + }, + { + "epoch": 4.61, + "learning_rate": 2.701614902095411e-06, + "loss": 0.9359, + "step": 62327 + }, + { + "epoch": 4.61, + "learning_rate": 2.7013422314462668e-06, + "loss": 0.933, + "step": 62328 + }, + { + "epoch": 4.61, + "learning_rate": 2.7010695724091796e-06, + "loss": 0.962, + "step": 62329 + }, + { + "epoch": 4.61, + "learning_rate": 2.700796924984589e-06, + "loss": 1.0751, + "step": 62330 + }, + { + "epoch": 4.61, + "learning_rate": 2.7005242891729266e-06, + "loss": 1.0561, + "step": 62331 + }, + { + "epoch": 4.61, + "learning_rate": 2.7002516649746247e-06, + "loss": 1.0945, + "step": 62332 + }, + { + "epoch": 4.61, + "learning_rate": 2.699979052390116e-06, + "loss": 0.9818, + "step": 62333 + }, + { + "epoch": 4.61, + "learning_rate": 2.699706451419839e-06, + "loss": 0.9707, + "step": 62334 + }, + { + "epoch": 4.61, + "learning_rate": 2.699433862064227e-06, + "loss": 1.0639, + "step": 62335 + }, + { + "epoch": 4.61, + "learning_rate": 2.699161284323706e-06, + "loss": 0.9176, + "step": 62336 + }, + { + "epoch": 4.61, + "learning_rate": 2.698888718198718e-06, + "loss": 1.0136, + "step": 62337 + }, + { + "epoch": 4.61, + "learning_rate": 2.6986161636896925e-06, + "loss": 1.0221, + "step": 62338 + }, + { + "epoch": 4.61, + "learning_rate": 2.6983436207970647e-06, + "loss": 0.9138, + "step": 62339 + }, + { + "epoch": 4.61, + "learning_rate": 2.6980710895212646e-06, + "loss": 1.0361, + "step": 62340 + }, + { + "epoch": 4.61, + "learning_rate": 2.6977985698627327e-06, + "loss": 0.9872, + "step": 62341 + }, + { + "epoch": 4.61, + "learning_rate": 2.697526061821899e-06, + "loss": 0.9376, + "step": 62342 + }, + { + "epoch": 4.61, + "learning_rate": 2.697253565399196e-06, + "loss": 0.9418, + "step": 62343 + }, + { + "epoch": 4.61, + "learning_rate": 2.6969810805950548e-06, + "loss": 0.9467, + "step": 62344 + }, + { + "epoch": 4.61, + "learning_rate": 2.6967086074099156e-06, + "loss": 1.0288, + "step": 62345 + }, + { + "epoch": 4.61, + "learning_rate": 2.696436145844209e-06, + "loss": 1.0042, + "step": 62346 + }, + { + "epoch": 4.61, + "learning_rate": 2.696163695898367e-06, + "loss": 1.1358, + "step": 62347 + }, + { + "epoch": 4.61, + "learning_rate": 2.6958912575728258e-06, + "loss": 1.0682, + "step": 62348 + }, + { + "epoch": 4.61, + "learning_rate": 2.695618830868012e-06, + "loss": 1.0502, + "step": 62349 + }, + { + "epoch": 4.61, + "learning_rate": 2.6953464157843688e-06, + "loss": 1.0022, + "step": 62350 + }, + { + "epoch": 4.61, + "learning_rate": 2.695074012322324e-06, + "loss": 0.9353, + "step": 62351 + }, + { + "epoch": 4.61, + "learning_rate": 2.6948016204823124e-06, + "loss": 0.9426, + "step": 62352 + }, + { + "epoch": 4.61, + "learning_rate": 2.694529240264764e-06, + "loss": 0.9235, + "step": 62353 + }, + { + "epoch": 4.61, + "learning_rate": 2.6942568716701167e-06, + "loss": 1.0202, + "step": 62354 + }, + { + "epoch": 4.61, + "learning_rate": 2.6939845146988063e-06, + "loss": 1.0239, + "step": 62355 + }, + { + "epoch": 4.61, + "learning_rate": 2.693712169351257e-06, + "loss": 1.0815, + "step": 62356 + }, + { + "epoch": 4.61, + "learning_rate": 2.693439835627909e-06, + "loss": 1.0194, + "step": 62357 + }, + { + "epoch": 4.61, + "learning_rate": 2.6931675135291903e-06, + "loss": 0.9811, + "step": 62358 + }, + { + "epoch": 4.61, + "learning_rate": 2.6928952030555443e-06, + "loss": 0.9251, + "step": 62359 + }, + { + "epoch": 4.61, + "learning_rate": 2.6926229042073904e-06, + "loss": 0.9653, + "step": 62360 + }, + { + "epoch": 4.61, + "learning_rate": 2.6923506169851733e-06, + "loss": 0.9516, + "step": 62361 + }, + { + "epoch": 4.61, + "learning_rate": 2.692078341389318e-06, + "loss": 1.0413, + "step": 62362 + }, + { + "epoch": 4.61, + "learning_rate": 2.691806077420267e-06, + "loss": 0.9415, + "step": 62363 + }, + { + "epoch": 4.61, + "learning_rate": 2.691533825078446e-06, + "loss": 0.9343, + "step": 62364 + }, + { + "epoch": 4.61, + "learning_rate": 2.6912615843642864e-06, + "loss": 1.0379, + "step": 62365 + }, + { + "epoch": 4.61, + "learning_rate": 2.6909893552782285e-06, + "loss": 0.9462, + "step": 62366 + }, + { + "epoch": 4.61, + "learning_rate": 2.690717137820702e-06, + "loss": 1.0021, + "step": 62367 + }, + { + "epoch": 4.61, + "learning_rate": 2.69044493199214e-06, + "loss": 0.999, + "step": 62368 + }, + { + "epoch": 4.61, + "learning_rate": 2.690172737792972e-06, + "loss": 0.8108, + "step": 62369 + }, + { + "epoch": 4.61, + "learning_rate": 2.6899005552236377e-06, + "loss": 0.9818, + "step": 62370 + }, + { + "epoch": 4.61, + "learning_rate": 2.6896283842845684e-06, + "loss": 1.006, + "step": 62371 + }, + { + "epoch": 4.61, + "learning_rate": 2.689356224976194e-06, + "loss": 1.0163, + "step": 62372 + }, + { + "epoch": 4.61, + "learning_rate": 2.68908407729895e-06, + "loss": 1.1258, + "step": 62373 + }, + { + "epoch": 4.61, + "learning_rate": 2.6888119412532654e-06, + "loss": 1.0426, + "step": 62374 + }, + { + "epoch": 4.61, + "learning_rate": 2.6885398168395792e-06, + "loss": 0.9325, + "step": 62375 + }, + { + "epoch": 4.61, + "learning_rate": 2.6882677040583228e-06, + "loss": 0.8574, + "step": 62376 + }, + { + "epoch": 4.61, + "learning_rate": 2.6879956029099263e-06, + "loss": 0.933, + "step": 62377 + }, + { + "epoch": 4.61, + "learning_rate": 2.687723513394822e-06, + "loss": 1.036, + "step": 62378 + }, + { + "epoch": 4.61, + "learning_rate": 2.6874514355134475e-06, + "loss": 1.0093, + "step": 62379 + }, + { + "epoch": 4.61, + "learning_rate": 2.687179369266233e-06, + "loss": 1.0936, + "step": 62380 + }, + { + "epoch": 4.61, + "learning_rate": 2.686907314653612e-06, + "loss": 1.0961, + "step": 62381 + }, + { + "epoch": 4.61, + "learning_rate": 2.686635271676017e-06, + "loss": 0.9204, + "step": 62382 + }, + { + "epoch": 4.61, + "learning_rate": 2.6863632403338804e-06, + "loss": 0.9646, + "step": 62383 + }, + { + "epoch": 4.61, + "learning_rate": 2.6860912206276357e-06, + "loss": 1.062, + "step": 62384 + }, + { + "epoch": 4.61, + "learning_rate": 2.685819212557711e-06, + "loss": 1.0101, + "step": 62385 + }, + { + "epoch": 4.61, + "learning_rate": 2.685547216124548e-06, + "loss": 1.0026, + "step": 62386 + }, + { + "epoch": 4.61, + "learning_rate": 2.6852752313285734e-06, + "loss": 0.9528, + "step": 62387 + }, + { + "epoch": 4.61, + "learning_rate": 2.6850032581702213e-06, + "loss": 1.0844, + "step": 62388 + }, + { + "epoch": 4.61, + "learning_rate": 2.6847312966499215e-06, + "loss": 1.0238, + "step": 62389 + }, + { + "epoch": 4.61, + "learning_rate": 2.6844593467681123e-06, + "loss": 1.0718, + "step": 62390 + }, + { + "epoch": 4.61, + "learning_rate": 2.684187408525224e-06, + "loss": 1.2037, + "step": 62391 + }, + { + "epoch": 4.61, + "learning_rate": 2.6839154819216896e-06, + "loss": 0.9624, + "step": 62392 + }, + { + "epoch": 4.61, + "learning_rate": 2.6836435669579395e-06, + "loss": 1.0721, + "step": 62393 + }, + { + "epoch": 4.61, + "learning_rate": 2.6833716636344055e-06, + "loss": 1.0929, + "step": 62394 + }, + { + "epoch": 4.61, + "learning_rate": 2.6830997719515262e-06, + "loss": 1.0214, + "step": 62395 + }, + { + "epoch": 4.61, + "learning_rate": 2.6828278919097295e-06, + "loss": 1.0098, + "step": 62396 + }, + { + "epoch": 4.61, + "learning_rate": 2.6825560235094494e-06, + "loss": 1.0417, + "step": 62397 + }, + { + "epoch": 4.61, + "learning_rate": 2.6822841667511146e-06, + "loss": 1.0436, + "step": 62398 + }, + { + "epoch": 4.61, + "learning_rate": 2.6820123216351646e-06, + "loss": 1.0746, + "step": 62399 + }, + { + "epoch": 4.61, + "learning_rate": 2.681740488162028e-06, + "loss": 0.967, + "step": 62400 + }, + { + "epoch": 4.61, + "learning_rate": 2.681468666332138e-06, + "loss": 1.0959, + "step": 62401 + }, + { + "epoch": 4.61, + "learning_rate": 2.6811968561459254e-06, + "loss": 1.0055, + "step": 62402 + }, + { + "epoch": 4.61, + "learning_rate": 2.6809250576038216e-06, + "loss": 1.0665, + "step": 62403 + }, + { + "epoch": 4.61, + "learning_rate": 2.680653270706267e-06, + "loss": 1.0458, + "step": 62404 + }, + { + "epoch": 4.61, + "learning_rate": 2.6803814954536832e-06, + "loss": 0.9161, + "step": 62405 + }, + { + "epoch": 4.61, + "learning_rate": 2.6801097318465097e-06, + "loss": 0.9884, + "step": 62406 + }, + { + "epoch": 4.61, + "learning_rate": 2.679837979885175e-06, + "loss": 1.0848, + "step": 62407 + }, + { + "epoch": 4.61, + "learning_rate": 2.679566239570118e-06, + "loss": 1.0313, + "step": 62408 + }, + { + "epoch": 4.61, + "learning_rate": 2.679294510901761e-06, + "loss": 1.084, + "step": 62409 + }, + { + "epoch": 4.61, + "learning_rate": 2.6790227938805446e-06, + "loss": 0.9893, + "step": 62410 + }, + { + "epoch": 4.61, + "learning_rate": 2.678751088506899e-06, + "loss": 0.9741, + "step": 62411 + }, + { + "epoch": 4.61, + "learning_rate": 2.6784793947812548e-06, + "loss": 1.0596, + "step": 62412 + }, + { + "epoch": 4.61, + "learning_rate": 2.678207712704045e-06, + "loss": 1.0309, + "step": 62413 + }, + { + "epoch": 4.61, + "learning_rate": 2.6779360422756994e-06, + "loss": 1.0392, + "step": 62414 + }, + { + "epoch": 4.61, + "learning_rate": 2.6776643834966564e-06, + "loss": 0.9803, + "step": 62415 + }, + { + "epoch": 4.61, + "learning_rate": 2.6773927363673437e-06, + "loss": 1.0468, + "step": 62416 + }, + { + "epoch": 4.61, + "learning_rate": 2.6771211008881946e-06, + "loss": 0.9776, + "step": 62417 + }, + { + "epoch": 4.61, + "learning_rate": 2.6768494770596376e-06, + "loss": 1.0723, + "step": 62418 + }, + { + "epoch": 4.61, + "learning_rate": 2.6765778648821116e-06, + "loss": 0.9888, + "step": 62419 + }, + { + "epoch": 4.61, + "learning_rate": 2.6763062643560456e-06, + "loss": 1.0178, + "step": 62420 + }, + { + "epoch": 4.61, + "learning_rate": 2.6760346754818724e-06, + "loss": 1.105, + "step": 62421 + }, + { + "epoch": 4.61, + "learning_rate": 2.6757630982600226e-06, + "loss": 0.8629, + "step": 62422 + }, + { + "epoch": 4.61, + "learning_rate": 2.6754915326909246e-06, + "loss": 1.0404, + "step": 62423 + }, + { + "epoch": 4.61, + "learning_rate": 2.6752199787750187e-06, + "loss": 1.0025, + "step": 62424 + }, + { + "epoch": 4.61, + "learning_rate": 2.6749484365127344e-06, + "loss": 0.9773, + "step": 62425 + }, + { + "epoch": 4.61, + "learning_rate": 2.6746769059045007e-06, + "loss": 1.0003, + "step": 62426 + }, + { + "epoch": 4.61, + "learning_rate": 2.674405386950748e-06, + "loss": 1.0532, + "step": 62427 + }, + { + "epoch": 4.61, + "learning_rate": 2.674133879651919e-06, + "loss": 1.0003, + "step": 62428 + }, + { + "epoch": 4.61, + "learning_rate": 2.673862384008434e-06, + "loss": 0.9387, + "step": 62429 + }, + { + "epoch": 4.61, + "learning_rate": 2.673590900020726e-06, + "loss": 1.0046, + "step": 62430 + }, + { + "epoch": 4.61, + "learning_rate": 2.673319427689234e-06, + "loss": 1.0205, + "step": 62431 + }, + { + "epoch": 4.61, + "learning_rate": 2.673047967014386e-06, + "loss": 0.9347, + "step": 62432 + }, + { + "epoch": 4.61, + "learning_rate": 2.6727765179966125e-06, + "loss": 0.8741, + "step": 62433 + }, + { + "epoch": 4.61, + "learning_rate": 2.672505080636345e-06, + "loss": 1.0879, + "step": 62434 + }, + { + "epoch": 4.61, + "learning_rate": 2.67223365493402e-06, + "loss": 1.0523, + "step": 62435 + }, + { + "epoch": 4.61, + "learning_rate": 2.671962240890066e-06, + "loss": 0.8949, + "step": 62436 + }, + { + "epoch": 4.61, + "learning_rate": 2.6716908385049157e-06, + "loss": 1.1487, + "step": 62437 + }, + { + "epoch": 4.61, + "learning_rate": 2.671419447778999e-06, + "loss": 1.0104, + "step": 62438 + }, + { + "epoch": 4.61, + "learning_rate": 2.6711480687127478e-06, + "loss": 0.9725, + "step": 62439 + }, + { + "epoch": 4.61, + "learning_rate": 2.6708767013065973e-06, + "loss": 0.9474, + "step": 62440 + }, + { + "epoch": 4.61, + "learning_rate": 2.670605345560977e-06, + "loss": 0.9522, + "step": 62441 + }, + { + "epoch": 4.61, + "learning_rate": 2.670334001476319e-06, + "loss": 0.9984, + "step": 62442 + }, + { + "epoch": 4.61, + "learning_rate": 2.670062669053052e-06, + "loss": 1.0429, + "step": 62443 + }, + { + "epoch": 4.61, + "learning_rate": 2.6697913482916137e-06, + "loss": 0.9174, + "step": 62444 + }, + { + "epoch": 4.61, + "learning_rate": 2.6695200391924315e-06, + "loss": 1.0772, + "step": 62445 + }, + { + "epoch": 4.61, + "learning_rate": 2.6692487417559387e-06, + "loss": 1.0381, + "step": 62446 + }, + { + "epoch": 4.61, + "learning_rate": 2.6689774559825653e-06, + "loss": 0.9961, + "step": 62447 + }, + { + "epoch": 4.61, + "learning_rate": 2.6687061818727423e-06, + "loss": 0.9647, + "step": 62448 + }, + { + "epoch": 4.61, + "learning_rate": 2.668434919426908e-06, + "loss": 1.0304, + "step": 62449 + }, + { + "epoch": 4.61, + "learning_rate": 2.6681636686454836e-06, + "loss": 0.9619, + "step": 62450 + }, + { + "epoch": 4.61, + "learning_rate": 2.6678924295289076e-06, + "loss": 1.0038, + "step": 62451 + }, + { + "epoch": 4.61, + "learning_rate": 2.6676212020776083e-06, + "loss": 1.0774, + "step": 62452 + }, + { + "epoch": 4.61, + "learning_rate": 2.6673499862920236e-06, + "loss": 0.9997, + "step": 62453 + }, + { + "epoch": 4.61, + "learning_rate": 2.667078782172575e-06, + "loss": 0.8691, + "step": 62454 + }, + { + "epoch": 4.61, + "learning_rate": 2.6668075897197022e-06, + "loss": 1.0644, + "step": 62455 + }, + { + "epoch": 4.61, + "learning_rate": 2.66653640893383e-06, + "loss": 1.1169, + "step": 62456 + }, + { + "epoch": 4.61, + "learning_rate": 2.666265239815399e-06, + "loss": 1.084, + "step": 62457 + }, + { + "epoch": 4.61, + "learning_rate": 2.6659940823648335e-06, + "loss": 0.9437, + "step": 62458 + }, + { + "epoch": 4.62, + "learning_rate": 2.665722936582562e-06, + "loss": 1.1025, + "step": 62459 + }, + { + "epoch": 4.62, + "learning_rate": 2.6654518024690246e-06, + "loss": 0.9972, + "step": 62460 + }, + { + "epoch": 4.62, + "learning_rate": 2.665180680024647e-06, + "loss": 0.8727, + "step": 62461 + }, + { + "epoch": 4.62, + "learning_rate": 2.664909569249863e-06, + "loss": 1.0096, + "step": 62462 + }, + { + "epoch": 4.62, + "learning_rate": 2.6646384701451e-06, + "loss": 1.0504, + "step": 62463 + }, + { + "epoch": 4.62, + "learning_rate": 2.6643673827107952e-06, + "loss": 0.9948, + "step": 62464 + }, + { + "epoch": 4.62, + "learning_rate": 2.6640963069473767e-06, + "loss": 0.9805, + "step": 62465 + }, + { + "epoch": 4.62, + "learning_rate": 2.663825242855276e-06, + "loss": 0.9088, + "step": 62466 + }, + { + "epoch": 4.62, + "learning_rate": 2.663554190434925e-06, + "loss": 1.0101, + "step": 62467 + }, + { + "epoch": 4.62, + "learning_rate": 2.6632831496867508e-06, + "loss": 0.8759, + "step": 62468 + }, + { + "epoch": 4.62, + "learning_rate": 2.6630121206111903e-06, + "loss": 1.0671, + "step": 62469 + }, + { + "epoch": 4.62, + "learning_rate": 2.662741103208674e-06, + "loss": 1.0284, + "step": 62470 + }, + { + "epoch": 4.62, + "learning_rate": 2.66247009747963e-06, + "loss": 1.0575, + "step": 62471 + }, + { + "epoch": 4.62, + "learning_rate": 2.6621991034244896e-06, + "loss": 1.0249, + "step": 62472 + }, + { + "epoch": 4.62, + "learning_rate": 2.661928121043691e-06, + "loss": 1.1078, + "step": 62473 + }, + { + "epoch": 4.62, + "learning_rate": 2.661657150337654e-06, + "loss": 0.8803, + "step": 62474 + }, + { + "epoch": 4.62, + "learning_rate": 2.6613861913068182e-06, + "loss": 1.029, + "step": 62475 + }, + { + "epoch": 4.62, + "learning_rate": 2.661115243951612e-06, + "loss": 0.9942, + "step": 62476 + }, + { + "epoch": 4.62, + "learning_rate": 2.6608443082724666e-06, + "loss": 1.0505, + "step": 62477 + }, + { + "epoch": 4.62, + "learning_rate": 2.660573384269812e-06, + "loss": 1.1175, + "step": 62478 + }, + { + "epoch": 4.62, + "learning_rate": 2.6603024719440783e-06, + "loss": 1.0817, + "step": 62479 + }, + { + "epoch": 4.62, + "learning_rate": 2.6600315712957003e-06, + "loss": 0.95, + "step": 62480 + }, + { + "epoch": 4.62, + "learning_rate": 2.6597606823251076e-06, + "loss": 0.9496, + "step": 62481 + }, + { + "epoch": 4.62, + "learning_rate": 2.659489805032731e-06, + "loss": 0.8982, + "step": 62482 + }, + { + "epoch": 4.62, + "learning_rate": 2.659218939418998e-06, + "loss": 1.0617, + "step": 62483 + }, + { + "epoch": 4.62, + "learning_rate": 2.658948085484345e-06, + "loss": 0.8729, + "step": 62484 + }, + { + "epoch": 4.62, + "learning_rate": 2.658677243229202e-06, + "loss": 1.0489, + "step": 62485 + }, + { + "epoch": 4.62, + "learning_rate": 2.6584064126539965e-06, + "loss": 0.9622, + "step": 62486 + }, + { + "epoch": 4.62, + "learning_rate": 2.6581355937591624e-06, + "loss": 0.8404, + "step": 62487 + }, + { + "epoch": 4.62, + "learning_rate": 2.6578647865451266e-06, + "loss": 1.1444, + "step": 62488 + }, + { + "epoch": 4.62, + "learning_rate": 2.657593991012326e-06, + "loss": 0.9595, + "step": 62489 + }, + { + "epoch": 4.62, + "learning_rate": 2.657323207161188e-06, + "loss": 1.0045, + "step": 62490 + }, + { + "epoch": 4.62, + "learning_rate": 2.6570524349921432e-06, + "loss": 1.0596, + "step": 62491 + }, + { + "epoch": 4.62, + "learning_rate": 2.6567816745056206e-06, + "loss": 1.027, + "step": 62492 + }, + { + "epoch": 4.62, + "learning_rate": 2.656510925702056e-06, + "loss": 0.9727, + "step": 62493 + }, + { + "epoch": 4.62, + "learning_rate": 2.6562401885818813e-06, + "loss": 0.9584, + "step": 62494 + }, + { + "epoch": 4.62, + "learning_rate": 2.6559694631455155e-06, + "loss": 1.0158, + "step": 62495 + }, + { + "epoch": 4.62, + "learning_rate": 2.6556987493934018e-06, + "loss": 0.966, + "step": 62496 + }, + { + "epoch": 4.62, + "learning_rate": 2.655428047325963e-06, + "loss": 0.8922, + "step": 62497 + }, + { + "epoch": 4.62, + "learning_rate": 2.6551573569436395e-06, + "loss": 0.9361, + "step": 62498 + }, + { + "epoch": 4.62, + "learning_rate": 2.6548866782468486e-06, + "loss": 0.9647, + "step": 62499 + }, + { + "epoch": 4.62, + "learning_rate": 2.654616011236031e-06, + "loss": 1.0588, + "step": 62500 + }, + { + "epoch": 4.62, + "learning_rate": 2.6543453559116116e-06, + "loss": 1.0446, + "step": 62501 + }, + { + "epoch": 4.62, + "learning_rate": 2.6540747122740294e-06, + "loss": 0.9622, + "step": 62502 + }, + { + "epoch": 4.62, + "learning_rate": 2.653804080323704e-06, + "loss": 1.0464, + "step": 62503 + }, + { + "epoch": 4.62, + "learning_rate": 2.653533460061074e-06, + "loss": 1.0176, + "step": 62504 + }, + { + "epoch": 4.62, + "learning_rate": 2.6532628514865667e-06, + "loss": 1.0557, + "step": 62505 + }, + { + "epoch": 4.62, + "learning_rate": 2.6529922546006128e-06, + "loss": 1.0211, + "step": 62506 + }, + { + "epoch": 4.62, + "learning_rate": 2.652721669403644e-06, + "loss": 1.0671, + "step": 62507 + }, + { + "epoch": 4.62, + "learning_rate": 2.6524510958960856e-06, + "loss": 1.0133, + "step": 62508 + }, + { + "epoch": 4.62, + "learning_rate": 2.652180534078377e-06, + "loss": 1.0765, + "step": 62509 + }, + { + "epoch": 4.62, + "learning_rate": 2.6519099839509434e-06, + "loss": 0.9832, + "step": 62510 + }, + { + "epoch": 4.62, + "learning_rate": 2.6516394455142146e-06, + "loss": 1.0479, + "step": 62511 + }, + { + "epoch": 4.62, + "learning_rate": 2.651368918768621e-06, + "loss": 0.9842, + "step": 62512 + }, + { + "epoch": 4.62, + "learning_rate": 2.651098403714597e-06, + "loss": 1.0974, + "step": 62513 + }, + { + "epoch": 4.62, + "learning_rate": 2.6508279003525695e-06, + "loss": 1.0114, + "step": 62514 + }, + { + "epoch": 4.62, + "learning_rate": 2.65055740868297e-06, + "loss": 0.9746, + "step": 62515 + }, + { + "epoch": 4.62, + "learning_rate": 2.6502869287062293e-06, + "loss": 0.94, + "step": 62516 + }, + { + "epoch": 4.62, + "learning_rate": 2.650016460422773e-06, + "loss": 0.9832, + "step": 62517 + }, + { + "epoch": 4.62, + "learning_rate": 2.6497460038330416e-06, + "loss": 0.9251, + "step": 62518 + }, + { + "epoch": 4.62, + "learning_rate": 2.649475558937453e-06, + "loss": 1.1463, + "step": 62519 + }, + { + "epoch": 4.62, + "learning_rate": 2.649205125736447e-06, + "loss": 0.9355, + "step": 62520 + }, + { + "epoch": 4.62, + "learning_rate": 2.648934704230447e-06, + "loss": 0.9702, + "step": 62521 + }, + { + "epoch": 4.62, + "learning_rate": 2.6486642944198937e-06, + "loss": 1.0477, + "step": 62522 + }, + { + "epoch": 4.62, + "learning_rate": 2.6483938963052057e-06, + "loss": 1.0519, + "step": 62523 + }, + { + "epoch": 4.62, + "learning_rate": 2.648123509886816e-06, + "loss": 1.0653, + "step": 62524 + }, + { + "epoch": 4.62, + "learning_rate": 2.647853135165159e-06, + "loss": 0.9988, + "step": 62525 + }, + { + "epoch": 4.62, + "learning_rate": 2.6475827721406634e-06, + "loss": 0.9854, + "step": 62526 + }, + { + "epoch": 4.62, + "learning_rate": 2.647312420813758e-06, + "loss": 0.9962, + "step": 62527 + }, + { + "epoch": 4.62, + "learning_rate": 2.6470420811848695e-06, + "loss": 1.0724, + "step": 62528 + }, + { + "epoch": 4.62, + "learning_rate": 2.6467717532544357e-06, + "loss": 1.064, + "step": 62529 + }, + { + "epoch": 4.62, + "learning_rate": 2.6465014370228826e-06, + "loss": 1.1187, + "step": 62530 + }, + { + "epoch": 4.62, + "learning_rate": 2.646231132490641e-06, + "loss": 0.9951, + "step": 62531 + }, + { + "epoch": 4.62, + "learning_rate": 2.6459608396581405e-06, + "loss": 1.042, + "step": 62532 + }, + { + "epoch": 4.62, + "learning_rate": 2.6456905585258073e-06, + "loss": 0.9887, + "step": 62533 + }, + { + "epoch": 4.62, + "learning_rate": 2.645420289094078e-06, + "loss": 0.9286, + "step": 62534 + }, + { + "epoch": 4.62, + "learning_rate": 2.6451500313633803e-06, + "loss": 0.922, + "step": 62535 + }, + { + "epoch": 4.62, + "learning_rate": 2.6448797853341445e-06, + "loss": 0.9768, + "step": 62536 + }, + { + "epoch": 4.62, + "learning_rate": 2.644609551006796e-06, + "loss": 0.9179, + "step": 62537 + }, + { + "epoch": 4.62, + "learning_rate": 2.644339328381771e-06, + "loss": 0.8861, + "step": 62538 + }, + { + "epoch": 4.62, + "learning_rate": 2.6440691174594966e-06, + "loss": 1.0791, + "step": 62539 + }, + { + "epoch": 4.62, + "learning_rate": 2.6437989182404046e-06, + "loss": 0.9937, + "step": 62540 + }, + { + "epoch": 4.62, + "learning_rate": 2.6435287307249213e-06, + "loss": 1.06, + "step": 62541 + }, + { + "epoch": 4.62, + "learning_rate": 2.643258554913477e-06, + "loss": 1.0337, + "step": 62542 + }, + { + "epoch": 4.62, + "learning_rate": 2.642988390806509e-06, + "loss": 1.0336, + "step": 62543 + }, + { + "epoch": 4.62, + "learning_rate": 2.642718238404435e-06, + "loss": 1.0191, + "step": 62544 + }, + { + "epoch": 4.62, + "learning_rate": 2.6424480977076937e-06, + "loss": 1.0536, + "step": 62545 + }, + { + "epoch": 4.62, + "learning_rate": 2.6421779687167093e-06, + "loss": 1.0073, + "step": 62546 + }, + { + "epoch": 4.62, + "learning_rate": 2.6419078514319206e-06, + "loss": 0.9679, + "step": 62547 + }, + { + "epoch": 4.62, + "learning_rate": 2.6416377458537447e-06, + "loss": 0.9, + "step": 62548 + }, + { + "epoch": 4.62, + "learning_rate": 2.6413676519826216e-06, + "loss": 0.9713, + "step": 62549 + }, + { + "epoch": 4.62, + "learning_rate": 2.6410975698189767e-06, + "loss": 0.9652, + "step": 62550 + }, + { + "epoch": 4.62, + "learning_rate": 2.640827499363241e-06, + "loss": 1.1025, + "step": 62551 + }, + { + "epoch": 4.62, + "learning_rate": 2.640557440615843e-06, + "loss": 1.1052, + "step": 62552 + }, + { + "epoch": 4.62, + "learning_rate": 2.64028739357721e-06, + "loss": 1.0144, + "step": 62553 + }, + { + "epoch": 4.62, + "learning_rate": 2.6400173582477773e-06, + "loss": 0.9208, + "step": 62554 + }, + { + "epoch": 4.62, + "learning_rate": 2.639747334627971e-06, + "loss": 1.1008, + "step": 62555 + }, + { + "epoch": 4.62, + "learning_rate": 2.6394773227182225e-06, + "loss": 1.0139, + "step": 62556 + }, + { + "epoch": 4.62, + "learning_rate": 2.639207322518956e-06, + "loss": 1.0409, + "step": 62557 + }, + { + "epoch": 4.62, + "learning_rate": 2.63893733403061e-06, + "loss": 0.9889, + "step": 62558 + }, + { + "epoch": 4.62, + "learning_rate": 2.638667357253608e-06, + "loss": 0.9702, + "step": 62559 + }, + { + "epoch": 4.62, + "learning_rate": 2.638397392188382e-06, + "loss": 1.0805, + "step": 62560 + }, + { + "epoch": 4.62, + "learning_rate": 2.6381274388353596e-06, + "loss": 0.9809, + "step": 62561 + }, + { + "epoch": 4.62, + "learning_rate": 2.637857497194969e-06, + "loss": 1.015, + "step": 62562 + }, + { + "epoch": 4.62, + "learning_rate": 2.6375875672676466e-06, + "loss": 0.9683, + "step": 62563 + }, + { + "epoch": 4.62, + "learning_rate": 2.637317649053812e-06, + "loss": 0.9734, + "step": 62564 + }, + { + "epoch": 4.62, + "learning_rate": 2.637047742553902e-06, + "loss": 1.0411, + "step": 62565 + }, + { + "epoch": 4.62, + "learning_rate": 2.6367778477683416e-06, + "loss": 1.0423, + "step": 62566 + }, + { + "epoch": 4.62, + "learning_rate": 2.636507964697568e-06, + "loss": 0.9726, + "step": 62567 + }, + { + "epoch": 4.62, + "learning_rate": 2.6362380933419983e-06, + "loss": 0.9536, + "step": 62568 + }, + { + "epoch": 4.62, + "learning_rate": 2.635968233702071e-06, + "loss": 0.9863, + "step": 62569 + }, + { + "epoch": 4.62, + "learning_rate": 2.6356983857782136e-06, + "loss": 1.0481, + "step": 62570 + }, + { + "epoch": 4.62, + "learning_rate": 2.6354285495708552e-06, + "loss": 0.984, + "step": 62571 + }, + { + "epoch": 4.62, + "learning_rate": 2.6351587250804244e-06, + "loss": 1.0416, + "step": 62572 + }, + { + "epoch": 4.62, + "learning_rate": 2.634888912307347e-06, + "loss": 0.9046, + "step": 62573 + }, + { + "epoch": 4.62, + "learning_rate": 2.6346191112520593e-06, + "loss": 0.9632, + "step": 62574 + }, + { + "epoch": 4.62, + "learning_rate": 2.6343493219149874e-06, + "loss": 0.9538, + "step": 62575 + }, + { + "epoch": 4.62, + "learning_rate": 2.6340795442965604e-06, + "loss": 1.0126, + "step": 62576 + }, + { + "epoch": 4.62, + "learning_rate": 2.633809778397204e-06, + "loss": 0.9748, + "step": 62577 + }, + { + "epoch": 4.62, + "learning_rate": 2.6335400242173535e-06, + "loss": 1.1221, + "step": 62578 + }, + { + "epoch": 4.62, + "learning_rate": 2.6332702817574363e-06, + "loss": 1.091, + "step": 62579 + }, + { + "epoch": 4.62, + "learning_rate": 2.6330005510178813e-06, + "loss": 1.0271, + "step": 62580 + }, + { + "epoch": 4.62, + "learning_rate": 2.632730831999115e-06, + "loss": 0.9616, + "step": 62581 + }, + { + "epoch": 4.62, + "learning_rate": 2.6324611247015675e-06, + "loss": 1.0474, + "step": 62582 + }, + { + "epoch": 4.62, + "learning_rate": 2.632191429125671e-06, + "loss": 1.0921, + "step": 62583 + }, + { + "epoch": 4.62, + "learning_rate": 2.6319217452718524e-06, + "loss": 0.9856, + "step": 62584 + }, + { + "epoch": 4.62, + "learning_rate": 2.6316520731405417e-06, + "loss": 1.0236, + "step": 62585 + }, + { + "epoch": 4.62, + "learning_rate": 2.631382412732163e-06, + "loss": 0.9682, + "step": 62586 + }, + { + "epoch": 4.62, + "learning_rate": 2.6311127640471533e-06, + "loss": 0.9699, + "step": 62587 + }, + { + "epoch": 4.62, + "learning_rate": 2.63084312708594e-06, + "loss": 1.0092, + "step": 62588 + }, + { + "epoch": 4.62, + "learning_rate": 2.630573501848944e-06, + "loss": 1.0745, + "step": 62589 + }, + { + "epoch": 4.62, + "learning_rate": 2.630303888336604e-06, + "loss": 1.0843, + "step": 62590 + }, + { + "epoch": 4.62, + "learning_rate": 2.6300342865493422e-06, + "loss": 1.017, + "step": 62591 + }, + { + "epoch": 4.62, + "learning_rate": 2.6297646964875954e-06, + "loss": 1.0485, + "step": 62592 + }, + { + "epoch": 4.62, + "learning_rate": 2.6294951181517814e-06, + "loss": 1.0695, + "step": 62593 + }, + { + "epoch": 4.63, + "learning_rate": 2.6292255515423393e-06, + "loss": 1.0008, + "step": 62594 + }, + { + "epoch": 4.63, + "learning_rate": 2.628955996659693e-06, + "loss": 1.07, + "step": 62595 + }, + { + "epoch": 4.63, + "learning_rate": 2.6286864535042724e-06, + "loss": 1.0998, + "step": 62596 + }, + { + "epoch": 4.63, + "learning_rate": 2.6284169220765064e-06, + "loss": 0.9874, + "step": 62597 + }, + { + "epoch": 4.63, + "learning_rate": 2.6281474023768205e-06, + "loss": 1.0054, + "step": 62598 + }, + { + "epoch": 4.63, + "learning_rate": 2.6278778944056495e-06, + "loss": 1.0443, + "step": 62599 + }, + { + "epoch": 4.63, + "learning_rate": 2.627608398163419e-06, + "loss": 1.0153, + "step": 62600 + }, + { + "epoch": 4.63, + "learning_rate": 2.627338913650559e-06, + "loss": 1.0477, + "step": 62601 + }, + { + "epoch": 4.63, + "learning_rate": 2.6270694408674934e-06, + "loss": 0.9807, + "step": 62602 + }, + { + "epoch": 4.63, + "learning_rate": 2.626799979814658e-06, + "loss": 0.9227, + "step": 62603 + }, + { + "epoch": 4.63, + "learning_rate": 2.6265305304924783e-06, + "loss": 0.9739, + "step": 62604 + }, + { + "epoch": 4.63, + "learning_rate": 2.6262610929013833e-06, + "loss": 0.9463, + "step": 62605 + }, + { + "epoch": 4.63, + "learning_rate": 2.625991667041802e-06, + "loss": 1.0004, + "step": 62606 + }, + { + "epoch": 4.63, + "learning_rate": 2.625722252914158e-06, + "loss": 0.8905, + "step": 62607 + }, + { + "epoch": 4.63, + "learning_rate": 2.6254528505188915e-06, + "loss": 0.9554, + "step": 62608 + }, + { + "epoch": 4.63, + "learning_rate": 2.6251834598564173e-06, + "loss": 1.0321, + "step": 62609 + }, + { + "epoch": 4.63, + "learning_rate": 2.6249140809271744e-06, + "loss": 0.9553, + "step": 62610 + }, + { + "epoch": 4.63, + "learning_rate": 2.6246447137315846e-06, + "loss": 1.0836, + "step": 62611 + }, + { + "epoch": 4.63, + "learning_rate": 2.624375358270085e-06, + "loss": 0.895, + "step": 62612 + }, + { + "epoch": 4.63, + "learning_rate": 2.624106014543093e-06, + "loss": 1.0718, + "step": 62613 + }, + { + "epoch": 4.63, + "learning_rate": 2.6238366825510462e-06, + "loss": 1.0621, + "step": 62614 + }, + { + "epoch": 4.63, + "learning_rate": 2.6235673622943657e-06, + "loss": 1.0185, + "step": 62615 + }, + { + "epoch": 4.63, + "learning_rate": 2.623298053773491e-06, + "loss": 0.9993, + "step": 62616 + }, + { + "epoch": 4.63, + "learning_rate": 2.62302875698884e-06, + "loss": 0.9993, + "step": 62617 + }, + { + "epoch": 4.63, + "learning_rate": 2.6227594719408424e-06, + "loss": 1.0847, + "step": 62618 + }, + { + "epoch": 4.63, + "learning_rate": 2.6224901986299313e-06, + "loss": 1.0243, + "step": 62619 + }, + { + "epoch": 4.63, + "learning_rate": 2.622220937056532e-06, + "loss": 0.9584, + "step": 62620 + }, + { + "epoch": 4.63, + "learning_rate": 2.621951687221075e-06, + "loss": 1.1288, + "step": 62621 + }, + { + "epoch": 4.63, + "learning_rate": 2.6216824491239836e-06, + "loss": 1.0561, + "step": 62622 + }, + { + "epoch": 4.63, + "learning_rate": 2.621413222765693e-06, + "loss": 0.9553, + "step": 62623 + }, + { + "epoch": 4.63, + "learning_rate": 2.621144008146628e-06, + "loss": 0.8458, + "step": 62624 + }, + { + "epoch": 4.63, + "learning_rate": 2.6208748052672175e-06, + "loss": 1.1352, + "step": 62625 + }, + { + "epoch": 4.63, + "learning_rate": 2.6206056141278903e-06, + "loss": 1.0218, + "step": 62626 + }, + { + "epoch": 4.63, + "learning_rate": 2.6203364347290704e-06, + "loss": 0.9312, + "step": 62627 + }, + { + "epoch": 4.63, + "learning_rate": 2.6200672670711923e-06, + "loss": 0.9391, + "step": 62628 + }, + { + "epoch": 4.63, + "learning_rate": 2.619798111154682e-06, + "loss": 1.0693, + "step": 62629 + }, + { + "epoch": 4.63, + "learning_rate": 2.6195289669799675e-06, + "loss": 0.98, + "step": 62630 + }, + { + "epoch": 4.63, + "learning_rate": 2.619259834547474e-06, + "loss": 1.0254, + "step": 62631 + }, + { + "epoch": 4.63, + "learning_rate": 2.6189907138576343e-06, + "loss": 0.9183, + "step": 62632 + }, + { + "epoch": 4.63, + "learning_rate": 2.6187216049108755e-06, + "loss": 1.0175, + "step": 62633 + }, + { + "epoch": 4.63, + "learning_rate": 2.618452507707625e-06, + "loss": 1.0379, + "step": 62634 + }, + { + "epoch": 4.63, + "learning_rate": 2.6181834222483115e-06, + "loss": 1.0205, + "step": 62635 + }, + { + "epoch": 4.63, + "learning_rate": 2.617914348533359e-06, + "loss": 0.9611, + "step": 62636 + }, + { + "epoch": 4.63, + "learning_rate": 2.6176452865632052e-06, + "loss": 1.0532, + "step": 62637 + }, + { + "epoch": 4.63, + "learning_rate": 2.617376236338266e-06, + "loss": 1.0168, + "step": 62638 + }, + { + "epoch": 4.63, + "learning_rate": 2.617107197858979e-06, + "loss": 0.9787, + "step": 62639 + }, + { + "epoch": 4.63, + "learning_rate": 2.6168381711257685e-06, + "loss": 1.1105, + "step": 62640 + }, + { + "epoch": 4.63, + "learning_rate": 2.6165691561390637e-06, + "loss": 1.0569, + "step": 62641 + }, + { + "epoch": 4.63, + "learning_rate": 2.6163001528992872e-06, + "loss": 0.9581, + "step": 62642 + }, + { + "epoch": 4.63, + "learning_rate": 2.6160311614068755e-06, + "loss": 0.9469, + "step": 62643 + }, + { + "epoch": 4.63, + "learning_rate": 2.6157621816622526e-06, + "loss": 0.9973, + "step": 62644 + }, + { + "epoch": 4.63, + "learning_rate": 2.6154932136658463e-06, + "loss": 1.0338, + "step": 62645 + }, + { + "epoch": 4.63, + "learning_rate": 2.615224257418084e-06, + "loss": 0.9525, + "step": 62646 + }, + { + "epoch": 4.63, + "learning_rate": 2.6149553129193916e-06, + "loss": 0.9497, + "step": 62647 + }, + { + "epoch": 4.63, + "learning_rate": 2.6146863801702027e-06, + "loss": 0.974, + "step": 62648 + }, + { + "epoch": 4.63, + "learning_rate": 2.6144174591709425e-06, + "loss": 0.9243, + "step": 62649 + }, + { + "epoch": 4.63, + "learning_rate": 2.614148549922039e-06, + "loss": 1.021, + "step": 62650 + }, + { + "epoch": 4.63, + "learning_rate": 2.613879652423915e-06, + "loss": 1.033, + "step": 62651 + }, + { + "epoch": 4.63, + "learning_rate": 2.613610766677006e-06, + "loss": 1.0235, + "step": 62652 + }, + { + "epoch": 4.63, + "learning_rate": 2.61334189268174e-06, + "loss": 1.0419, + "step": 62653 + }, + { + "epoch": 4.63, + "learning_rate": 2.6130730304385354e-06, + "loss": 0.9854, + "step": 62654 + }, + { + "epoch": 4.63, + "learning_rate": 2.612804179947829e-06, + "loss": 1.0508, + "step": 62655 + }, + { + "epoch": 4.63, + "learning_rate": 2.6125353412100427e-06, + "loss": 0.9804, + "step": 62656 + }, + { + "epoch": 4.63, + "learning_rate": 2.612266514225612e-06, + "loss": 1.0057, + "step": 62657 + }, + { + "epoch": 4.63, + "learning_rate": 2.6119976989949546e-06, + "loss": 0.9948, + "step": 62658 + }, + { + "epoch": 4.63, + "learning_rate": 2.611728895518506e-06, + "loss": 1.0402, + "step": 62659 + }, + { + "epoch": 4.63, + "learning_rate": 2.611460103796688e-06, + "loss": 0.9126, + "step": 62660 + }, + { + "epoch": 4.63, + "learning_rate": 2.611191323829937e-06, + "loss": 0.8912, + "step": 62661 + }, + { + "epoch": 4.63, + "learning_rate": 2.610922555618671e-06, + "loss": 0.9696, + "step": 62662 + }, + { + "epoch": 4.63, + "learning_rate": 2.6106537991633195e-06, + "loss": 0.9988, + "step": 62663 + }, + { + "epoch": 4.63, + "learning_rate": 2.610385054464316e-06, + "loss": 0.9475, + "step": 62664 + }, + { + "epoch": 4.63, + "learning_rate": 2.6101163215220828e-06, + "loss": 0.9678, + "step": 62665 + }, + { + "epoch": 4.63, + "learning_rate": 2.6098476003370487e-06, + "loss": 0.9838, + "step": 62666 + }, + { + "epoch": 4.63, + "learning_rate": 2.6095788909096387e-06, + "loss": 0.9808, + "step": 62667 + }, + { + "epoch": 4.63, + "learning_rate": 2.6093101932402865e-06, + "loss": 1.0049, + "step": 62668 + }, + { + "epoch": 4.63, + "learning_rate": 2.609041507329415e-06, + "loss": 1.1831, + "step": 62669 + }, + { + "epoch": 4.63, + "learning_rate": 2.608772833177454e-06, + "loss": 1.0275, + "step": 62670 + }, + { + "epoch": 4.63, + "learning_rate": 2.6085041707848258e-06, + "loss": 0.9813, + "step": 62671 + }, + { + "epoch": 4.63, + "learning_rate": 2.608235520151965e-06, + "loss": 0.9674, + "step": 62672 + }, + { + "epoch": 4.63, + "learning_rate": 2.607966881279296e-06, + "loss": 1.115, + "step": 62673 + }, + { + "epoch": 4.63, + "learning_rate": 2.607698254167245e-06, + "loss": 0.8677, + "step": 62674 + }, + { + "epoch": 4.63, + "learning_rate": 2.607429638816241e-06, + "loss": 1.0714, + "step": 62675 + }, + { + "epoch": 4.63, + "learning_rate": 2.6071610352267075e-06, + "loss": 0.8845, + "step": 62676 + }, + { + "epoch": 4.63, + "learning_rate": 2.6068924433990785e-06, + "loss": 0.9601, + "step": 62677 + }, + { + "epoch": 4.63, + "learning_rate": 2.606623863333777e-06, + "loss": 0.9193, + "step": 62678 + }, + { + "epoch": 4.63, + "learning_rate": 2.606355295031231e-06, + "loss": 0.9312, + "step": 62679 + }, + { + "epoch": 4.63, + "learning_rate": 2.606086738491865e-06, + "loss": 1.0164, + "step": 62680 + }, + { + "epoch": 4.63, + "learning_rate": 2.6058181937161163e-06, + "loss": 1.0638, + "step": 62681 + }, + { + "epoch": 4.63, + "learning_rate": 2.605549660704402e-06, + "loss": 0.9333, + "step": 62682 + }, + { + "epoch": 4.63, + "learning_rate": 2.6052811394571485e-06, + "loss": 0.9738, + "step": 62683 + }, + { + "epoch": 4.63, + "learning_rate": 2.60501262997479e-06, + "loss": 1.0604, + "step": 62684 + }, + { + "epoch": 4.63, + "learning_rate": 2.6047441322577507e-06, + "loss": 1.1615, + "step": 62685 + }, + { + "epoch": 4.63, + "learning_rate": 2.604475646306458e-06, + "loss": 0.9863, + "step": 62686 + }, + { + "epoch": 4.63, + "learning_rate": 2.604207172121336e-06, + "loss": 0.9111, + "step": 62687 + }, + { + "epoch": 4.63, + "learning_rate": 2.603938709702817e-06, + "loss": 1.0349, + "step": 62688 + }, + { + "epoch": 4.63, + "learning_rate": 2.603670259051325e-06, + "loss": 0.9692, + "step": 62689 + }, + { + "epoch": 4.63, + "learning_rate": 2.6034018201672896e-06, + "loss": 0.8982, + "step": 62690 + }, + { + "epoch": 4.63, + "learning_rate": 2.603133393051135e-06, + "loss": 1.0361, + "step": 62691 + }, + { + "epoch": 4.63, + "learning_rate": 2.6028649777032855e-06, + "loss": 0.9899, + "step": 62692 + }, + { + "epoch": 4.63, + "learning_rate": 2.6025965741241767e-06, + "loss": 1.0829, + "step": 62693 + }, + { + "epoch": 4.63, + "learning_rate": 2.6023281823142297e-06, + "loss": 1.088, + "step": 62694 + }, + { + "epoch": 4.63, + "learning_rate": 2.602059802273873e-06, + "loss": 0.9457, + "step": 62695 + }, + { + "epoch": 4.63, + "learning_rate": 2.601791434003531e-06, + "loss": 0.9543, + "step": 62696 + }, + { + "epoch": 4.63, + "learning_rate": 2.601523077503635e-06, + "loss": 0.9965, + "step": 62697 + }, + { + "epoch": 4.63, + "learning_rate": 2.601254732774611e-06, + "loss": 0.9906, + "step": 62698 + }, + { + "epoch": 4.63, + "learning_rate": 2.6009863998168837e-06, + "loss": 0.9208, + "step": 62699 + }, + { + "epoch": 4.63, + "learning_rate": 2.6007180786308817e-06, + "loss": 1.0823, + "step": 62700 + }, + { + "epoch": 4.63, + "learning_rate": 2.600449769217028e-06, + "loss": 1.0083, + "step": 62701 + }, + { + "epoch": 4.63, + "learning_rate": 2.600181471575759e-06, + "loss": 1.032, + "step": 62702 + }, + { + "epoch": 4.63, + "learning_rate": 2.59991318570749e-06, + "loss": 0.9825, + "step": 62703 + }, + { + "epoch": 4.63, + "learning_rate": 2.5996449116126556e-06, + "loss": 0.9245, + "step": 62704 + }, + { + "epoch": 4.63, + "learning_rate": 2.599376649291677e-06, + "loss": 0.9388, + "step": 62705 + }, + { + "epoch": 4.63, + "learning_rate": 2.5991083987449917e-06, + "loss": 0.9838, + "step": 62706 + }, + { + "epoch": 4.63, + "learning_rate": 2.598840159973012e-06, + "loss": 0.9919, + "step": 62707 + }, + { + "epoch": 4.63, + "learning_rate": 2.598571932976175e-06, + "loss": 1.1024, + "step": 62708 + }, + { + "epoch": 4.63, + "learning_rate": 2.598303717754904e-06, + "loss": 0.9578, + "step": 62709 + }, + { + "epoch": 4.63, + "learning_rate": 2.598035514309626e-06, + "loss": 0.9478, + "step": 62710 + }, + { + "epoch": 4.63, + "learning_rate": 2.5977673226407675e-06, + "loss": 0.9826, + "step": 62711 + }, + { + "epoch": 4.63, + "learning_rate": 2.597499142748752e-06, + "loss": 0.9807, + "step": 62712 + }, + { + "epoch": 4.63, + "learning_rate": 2.597230974634013e-06, + "loss": 1.0117, + "step": 62713 + }, + { + "epoch": 4.63, + "learning_rate": 2.596962818296973e-06, + "loss": 0.9381, + "step": 62714 + }, + { + "epoch": 4.63, + "learning_rate": 2.5966946737380593e-06, + "loss": 0.9763, + "step": 62715 + }, + { + "epoch": 4.63, + "learning_rate": 2.596426540957696e-06, + "loss": 0.8935, + "step": 62716 + }, + { + "epoch": 4.63, + "learning_rate": 2.596158419956315e-06, + "loss": 1.0291, + "step": 62717 + }, + { + "epoch": 4.63, + "learning_rate": 2.5958903107343392e-06, + "loss": 0.9851, + "step": 62718 + }, + { + "epoch": 4.63, + "learning_rate": 2.595622213292197e-06, + "loss": 1.0304, + "step": 62719 + }, + { + "epoch": 4.63, + "learning_rate": 2.595354127630314e-06, + "loss": 1.043, + "step": 62720 + }, + { + "epoch": 4.63, + "learning_rate": 2.595086053749113e-06, + "loss": 0.9887, + "step": 62721 + }, + { + "epoch": 4.63, + "learning_rate": 2.594817991649027e-06, + "loss": 1.0023, + "step": 62722 + }, + { + "epoch": 4.63, + "learning_rate": 2.5945499413304798e-06, + "loss": 1.0172, + "step": 62723 + }, + { + "epoch": 4.63, + "learning_rate": 2.5942819027938982e-06, + "loss": 1.0426, + "step": 62724 + }, + { + "epoch": 4.63, + "learning_rate": 2.594013876039705e-06, + "loss": 0.9855, + "step": 62725 + }, + { + "epoch": 4.63, + "learning_rate": 2.5937458610683353e-06, + "loss": 1.0356, + "step": 62726 + }, + { + "epoch": 4.63, + "learning_rate": 2.5934778578802043e-06, + "loss": 1.0504, + "step": 62727 + }, + { + "epoch": 4.63, + "learning_rate": 2.5932098664757467e-06, + "loss": 0.9865, + "step": 62728 + }, + { + "epoch": 4.64, + "learning_rate": 2.5929418868553867e-06, + "loss": 1.0652, + "step": 62729 + }, + { + "epoch": 4.64, + "learning_rate": 2.5926739190195505e-06, + "loss": 1.0259, + "step": 62730 + }, + { + "epoch": 4.64, + "learning_rate": 2.592405962968663e-06, + "loss": 1.0642, + "step": 62731 + }, + { + "epoch": 4.64, + "learning_rate": 2.5921380187031498e-06, + "loss": 1.0365, + "step": 62732 + }, + { + "epoch": 4.64, + "learning_rate": 2.591870086223441e-06, + "loss": 1.0799, + "step": 62733 + }, + { + "epoch": 4.64, + "learning_rate": 2.591602165529962e-06, + "loss": 0.9874, + "step": 62734 + }, + { + "epoch": 4.64, + "learning_rate": 2.591334256623137e-06, + "loss": 1.0124, + "step": 62735 + }, + { + "epoch": 4.64, + "learning_rate": 2.59106635950339e-06, + "loss": 0.9643, + "step": 62736 + }, + { + "epoch": 4.64, + "learning_rate": 2.5907984741711545e-06, + "loss": 0.9877, + "step": 62737 + }, + { + "epoch": 4.64, + "learning_rate": 2.5905306006268514e-06, + "loss": 0.9231, + "step": 62738 + }, + { + "epoch": 4.64, + "learning_rate": 2.590262738870909e-06, + "loss": 1.065, + "step": 62739 + }, + { + "epoch": 4.64, + "learning_rate": 2.589994888903752e-06, + "loss": 0.9576, + "step": 62740 + }, + { + "epoch": 4.64, + "learning_rate": 2.5897270507258044e-06, + "loss": 0.9937, + "step": 62741 + }, + { + "epoch": 4.64, + "learning_rate": 2.5894592243374984e-06, + "loss": 0.9424, + "step": 62742 + }, + { + "epoch": 4.64, + "learning_rate": 2.5891914097392567e-06, + "loss": 0.9579, + "step": 62743 + }, + { + "epoch": 4.64, + "learning_rate": 2.5889236069315048e-06, + "loss": 0.9941, + "step": 62744 + }, + { + "epoch": 4.64, + "learning_rate": 2.588655815914667e-06, + "loss": 0.9958, + "step": 62745 + }, + { + "epoch": 4.64, + "learning_rate": 2.5883880366891747e-06, + "loss": 0.9345, + "step": 62746 + }, + { + "epoch": 4.64, + "learning_rate": 2.588120269255453e-06, + "loss": 0.9878, + "step": 62747 + }, + { + "epoch": 4.64, + "learning_rate": 2.587852513613921e-06, + "loss": 1.0321, + "step": 62748 + }, + { + "epoch": 4.64, + "learning_rate": 2.5875847697650123e-06, + "loss": 1.1679, + "step": 62749 + }, + { + "epoch": 4.64, + "learning_rate": 2.587317037709147e-06, + "loss": 0.9972, + "step": 62750 + }, + { + "epoch": 4.64, + "learning_rate": 2.5870493174467604e-06, + "loss": 0.9635, + "step": 62751 + }, + { + "epoch": 4.64, + "learning_rate": 2.586781608978266e-06, + "loss": 0.9811, + "step": 62752 + }, + { + "epoch": 4.64, + "learning_rate": 2.5865139123040994e-06, + "loss": 1.0341, + "step": 62753 + }, + { + "epoch": 4.64, + "learning_rate": 2.5862462274246803e-06, + "loss": 0.9315, + "step": 62754 + }, + { + "epoch": 4.64, + "learning_rate": 2.5859785543404435e-06, + "loss": 0.9968, + "step": 62755 + }, + { + "epoch": 4.64, + "learning_rate": 2.585710893051805e-06, + "loss": 0.9964, + "step": 62756 + }, + { + "epoch": 4.64, + "learning_rate": 2.5854432435591914e-06, + "loss": 0.9861, + "step": 62757 + }, + { + "epoch": 4.64, + "learning_rate": 2.5851756058630353e-06, + "loss": 1.1087, + "step": 62758 + }, + { + "epoch": 4.64, + "learning_rate": 2.5849079799637577e-06, + "loss": 1.0986, + "step": 62759 + }, + { + "epoch": 4.64, + "learning_rate": 2.5846403658617857e-06, + "loss": 1.0287, + "step": 62760 + }, + { + "epoch": 4.64, + "learning_rate": 2.5843727635575422e-06, + "loss": 0.93, + "step": 62761 + }, + { + "epoch": 4.64, + "learning_rate": 2.584105173051459e-06, + "loss": 0.9957, + "step": 62762 + }, + { + "epoch": 4.64, + "learning_rate": 2.5838375943439576e-06, + "loss": 1.0357, + "step": 62763 + }, + { + "epoch": 4.64, + "learning_rate": 2.5835700274354647e-06, + "loss": 0.9601, + "step": 62764 + }, + { + "epoch": 4.64, + "learning_rate": 2.5833024723264054e-06, + "loss": 1.0501, + "step": 62765 + }, + { + "epoch": 4.64, + "learning_rate": 2.583034929017203e-06, + "loss": 1.0146, + "step": 62766 + }, + { + "epoch": 4.64, + "learning_rate": 2.582767397508289e-06, + "loss": 0.9867, + "step": 62767 + }, + { + "epoch": 4.64, + "learning_rate": 2.5824998778000854e-06, + "loss": 1.0161, + "step": 62768 + }, + { + "epoch": 4.64, + "learning_rate": 2.5822323698930197e-06, + "loss": 0.9478, + "step": 62769 + }, + { + "epoch": 4.64, + "learning_rate": 2.5819648737875115e-06, + "loss": 0.9697, + "step": 62770 + }, + { + "epoch": 4.64, + "learning_rate": 2.581697389483998e-06, + "loss": 1.0668, + "step": 62771 + }, + { + "epoch": 4.64, + "learning_rate": 2.5814299169828916e-06, + "loss": 1.1089, + "step": 62772 + }, + { + "epoch": 4.64, + "learning_rate": 2.5811624562846273e-06, + "loss": 0.9955, + "step": 62773 + }, + { + "epoch": 4.64, + "learning_rate": 2.5808950073896245e-06, + "loss": 1.0321, + "step": 62774 + }, + { + "epoch": 4.64, + "learning_rate": 2.5806275702983176e-06, + "loss": 0.9938, + "step": 62775 + }, + { + "epoch": 4.64, + "learning_rate": 2.5803601450111224e-06, + "loss": 1.0316, + "step": 62776 + }, + { + "epoch": 4.64, + "learning_rate": 2.580092731528465e-06, + "loss": 1.0346, + "step": 62777 + }, + { + "epoch": 4.64, + "learning_rate": 2.5798253298507782e-06, + "loss": 0.9493, + "step": 62778 + }, + { + "epoch": 4.64, + "learning_rate": 2.579557939978482e-06, + "loss": 1.1106, + "step": 62779 + }, + { + "epoch": 4.64, + "learning_rate": 2.5792905619120033e-06, + "loss": 1.0026, + "step": 62780 + }, + { + "epoch": 4.64, + "learning_rate": 2.579023195651764e-06, + "loss": 0.9718, + "step": 62781 + }, + { + "epoch": 4.64, + "learning_rate": 2.578755841198196e-06, + "loss": 1.0047, + "step": 62782 + }, + { + "epoch": 4.64, + "learning_rate": 2.5784884985517212e-06, + "loss": 1.144, + "step": 62783 + }, + { + "epoch": 4.64, + "learning_rate": 2.5782211677127646e-06, + "loss": 1.061, + "step": 62784 + }, + { + "epoch": 4.64, + "learning_rate": 2.5779538486817514e-06, + "loss": 0.9199, + "step": 62785 + }, + { + "epoch": 4.64, + "learning_rate": 2.5776865414591055e-06, + "loss": 0.9252, + "step": 62786 + }, + { + "epoch": 4.64, + "learning_rate": 2.5774192460452564e-06, + "loss": 0.9549, + "step": 62787 + }, + { + "epoch": 4.64, + "learning_rate": 2.5771519624406283e-06, + "loss": 0.9329, + "step": 62788 + }, + { + "epoch": 4.64, + "learning_rate": 2.5768846906456437e-06, + "loss": 0.9258, + "step": 62789 + }, + { + "epoch": 4.64, + "learning_rate": 2.576617430660727e-06, + "loss": 0.9394, + "step": 62790 + }, + { + "epoch": 4.64, + "learning_rate": 2.576350182486309e-06, + "loss": 0.9805, + "step": 62791 + }, + { + "epoch": 4.64, + "learning_rate": 2.576082946122811e-06, + "loss": 1.0191, + "step": 62792 + }, + { + "epoch": 4.64, + "learning_rate": 2.5758157215706602e-06, + "loss": 0.9611, + "step": 62793 + }, + { + "epoch": 4.64, + "learning_rate": 2.575548508830279e-06, + "loss": 1.0915, + "step": 62794 + }, + { + "epoch": 4.64, + "learning_rate": 2.575281307902091e-06, + "loss": 1.0834, + "step": 62795 + }, + { + "epoch": 4.64, + "learning_rate": 2.5750141187865307e-06, + "loss": 0.9935, + "step": 62796 + }, + { + "epoch": 4.64, + "learning_rate": 2.5747469414840107e-06, + "loss": 0.9858, + "step": 62797 + }, + { + "epoch": 4.64, + "learning_rate": 2.574479775994965e-06, + "loss": 1.0282, + "step": 62798 + }, + { + "epoch": 4.64, + "learning_rate": 2.5742126223198126e-06, + "loss": 1.1558, + "step": 62799 + }, + { + "epoch": 4.64, + "learning_rate": 2.573945480458988e-06, + "loss": 0.9478, + "step": 62800 + }, + { + "epoch": 4.64, + "learning_rate": 2.5736783504129036e-06, + "loss": 0.8127, + "step": 62801 + }, + { + "epoch": 4.64, + "learning_rate": 2.573411232181994e-06, + "loss": 1.0539, + "step": 62802 + }, + { + "epoch": 4.64, + "learning_rate": 2.57314412576668e-06, + "loss": 0.9685, + "step": 62803 + }, + { + "epoch": 4.64, + "learning_rate": 2.572877031167388e-06, + "loss": 1.0257, + "step": 62804 + }, + { + "epoch": 4.64, + "learning_rate": 2.572609948384542e-06, + "loss": 0.9202, + "step": 62805 + }, + { + "epoch": 4.64, + "learning_rate": 2.5723428774185642e-06, + "loss": 0.9265, + "step": 62806 + }, + { + "epoch": 4.64, + "learning_rate": 2.5720758182698868e-06, + "loss": 1.0877, + "step": 62807 + }, + { + "epoch": 4.64, + "learning_rate": 2.5718087709389296e-06, + "loss": 0.9801, + "step": 62808 + }, + { + "epoch": 4.64, + "learning_rate": 2.5715417354261187e-06, + "loss": 1.0076, + "step": 62809 + }, + { + "epoch": 4.64, + "learning_rate": 2.571274711731876e-06, + "loss": 0.9753, + "step": 62810 + }, + { + "epoch": 4.64, + "learning_rate": 2.5710076998566312e-06, + "loss": 1.0799, + "step": 62811 + }, + { + "epoch": 4.64, + "learning_rate": 2.5707406998008067e-06, + "loss": 0.9942, + "step": 62812 + }, + { + "epoch": 4.64, + "learning_rate": 2.570473711564828e-06, + "loss": 1.0606, + "step": 62813 + }, + { + "epoch": 4.64, + "learning_rate": 2.570206735149119e-06, + "loss": 1.0232, + "step": 62814 + }, + { + "epoch": 4.64, + "learning_rate": 2.5699397705541016e-06, + "loss": 1.0895, + "step": 62815 + }, + { + "epoch": 4.64, + "learning_rate": 2.5696728177802098e-06, + "loss": 1.0589, + "step": 62816 + }, + { + "epoch": 4.64, + "learning_rate": 2.5694058768278574e-06, + "loss": 0.9249, + "step": 62817 + }, + { + "epoch": 4.64, + "learning_rate": 2.5691389476974758e-06, + "loss": 1.0283, + "step": 62818 + }, + { + "epoch": 4.64, + "learning_rate": 2.5688720303894853e-06, + "loss": 1.0762, + "step": 62819 + }, + { + "epoch": 4.64, + "learning_rate": 2.568605124904319e-06, + "loss": 1.0181, + "step": 62820 + }, + { + "epoch": 4.64, + "learning_rate": 2.568338231242392e-06, + "loss": 0.9717, + "step": 62821 + }, + { + "epoch": 4.64, + "learning_rate": 2.5680713494041298e-06, + "loss": 1.0263, + "step": 62822 + }, + { + "epoch": 4.64, + "learning_rate": 2.5678044793899626e-06, + "loss": 0.9676, + "step": 62823 + }, + { + "epoch": 4.64, + "learning_rate": 2.5675376212003135e-06, + "loss": 1.0554, + "step": 62824 + }, + { + "epoch": 4.64, + "learning_rate": 2.567270774835604e-06, + "loss": 1.0565, + "step": 62825 + }, + { + "epoch": 4.64, + "learning_rate": 2.5670039402962574e-06, + "loss": 1.059, + "step": 62826 + }, + { + "epoch": 4.64, + "learning_rate": 2.566737117582705e-06, + "loss": 0.943, + "step": 62827 + }, + { + "epoch": 4.64, + "learning_rate": 2.566470306695368e-06, + "loss": 1.0349, + "step": 62828 + }, + { + "epoch": 4.64, + "learning_rate": 2.5662035076346693e-06, + "loss": 1.0291, + "step": 62829 + }, + { + "epoch": 4.64, + "learning_rate": 2.565936720401032e-06, + "loss": 0.9006, + "step": 62830 + }, + { + "epoch": 4.64, + "learning_rate": 2.5656699449948864e-06, + "loss": 0.9241, + "step": 62831 + }, + { + "epoch": 4.64, + "learning_rate": 2.5654031814166526e-06, + "loss": 0.9533, + "step": 62832 + }, + { + "epoch": 4.64, + "learning_rate": 2.5651364296667557e-06, + "loss": 0.9994, + "step": 62833 + }, + { + "epoch": 4.64, + "learning_rate": 2.5648696897456218e-06, + "loss": 0.9648, + "step": 62834 + }, + { + "epoch": 4.64, + "learning_rate": 2.5646029616536693e-06, + "loss": 1.0768, + "step": 62835 + }, + { + "epoch": 4.64, + "learning_rate": 2.564336245391331e-06, + "loss": 1.0131, + "step": 62836 + }, + { + "epoch": 4.64, + "learning_rate": 2.564069540959028e-06, + "loss": 0.9301, + "step": 62837 + }, + { + "epoch": 4.64, + "learning_rate": 2.563802848357183e-06, + "loss": 1.0287, + "step": 62838 + }, + { + "epoch": 4.64, + "learning_rate": 2.563536167586218e-06, + "loss": 0.838, + "step": 62839 + }, + { + "epoch": 4.64, + "learning_rate": 2.5632694986465646e-06, + "loss": 0.9359, + "step": 62840 + }, + { + "epoch": 4.64, + "learning_rate": 2.5630028415386455e-06, + "loss": 1.0155, + "step": 62841 + }, + { + "epoch": 4.64, + "learning_rate": 2.562736196262876e-06, + "loss": 1.1199, + "step": 62842 + }, + { + "epoch": 4.64, + "learning_rate": 2.5624695628196907e-06, + "loss": 0.9756, + "step": 62843 + }, + { + "epoch": 4.64, + "learning_rate": 2.562202941209506e-06, + "loss": 0.9509, + "step": 62844 + }, + { + "epoch": 4.64, + "learning_rate": 2.561936331432756e-06, + "loss": 1.1281, + "step": 62845 + }, + { + "epoch": 4.64, + "learning_rate": 2.561669733489854e-06, + "loss": 0.9542, + "step": 62846 + }, + { + "epoch": 4.64, + "learning_rate": 2.5614031473812316e-06, + "loss": 0.9827, + "step": 62847 + }, + { + "epoch": 4.64, + "learning_rate": 2.56113657310731e-06, + "loss": 0.8723, + "step": 62848 + }, + { + "epoch": 4.64, + "learning_rate": 2.5608700106685137e-06, + "loss": 0.9805, + "step": 62849 + }, + { + "epoch": 4.64, + "learning_rate": 2.5606034600652664e-06, + "loss": 0.9333, + "step": 62850 + }, + { + "epoch": 4.64, + "learning_rate": 2.56033692129799e-06, + "loss": 1.1207, + "step": 62851 + }, + { + "epoch": 4.64, + "learning_rate": 2.5600703943671145e-06, + "loss": 1.011, + "step": 62852 + }, + { + "epoch": 4.64, + "learning_rate": 2.559803879273061e-06, + "loss": 0.9595, + "step": 62853 + }, + { + "epoch": 4.64, + "learning_rate": 2.559537376016251e-06, + "loss": 1.0171, + "step": 62854 + }, + { + "epoch": 4.64, + "learning_rate": 2.559270884597109e-06, + "loss": 0.9923, + "step": 62855 + }, + { + "epoch": 4.64, + "learning_rate": 2.5590044050160633e-06, + "loss": 0.9885, + "step": 62856 + }, + { + "epoch": 4.64, + "learning_rate": 2.5587379372735353e-06, + "loss": 0.9607, + "step": 62857 + }, + { + "epoch": 4.64, + "learning_rate": 2.5584714813699496e-06, + "loss": 0.8414, + "step": 62858 + }, + { + "epoch": 4.64, + "learning_rate": 2.5582050373057275e-06, + "loss": 0.931, + "step": 62859 + }, + { + "epoch": 4.64, + "learning_rate": 2.5579386050812926e-06, + "loss": 0.9314, + "step": 62860 + }, + { + "epoch": 4.64, + "learning_rate": 2.5576721846970776e-06, + "loss": 0.9455, + "step": 62861 + }, + { + "epoch": 4.64, + "learning_rate": 2.557405776153493e-06, + "loss": 1.0494, + "step": 62862 + }, + { + "epoch": 4.64, + "learning_rate": 2.5571393794509725e-06, + "loss": 1.0052, + "step": 62863 + }, + { + "epoch": 4.64, + "learning_rate": 2.5568729945899327e-06, + "loss": 0.929, + "step": 62864 + }, + { + "epoch": 4.65, + "learning_rate": 2.5566066215708087e-06, + "loss": 1.0401, + "step": 62865 + }, + { + "epoch": 4.65, + "learning_rate": 2.5563402603940113e-06, + "loss": 0.9489, + "step": 62866 + }, + { + "epoch": 4.65, + "learning_rate": 2.556073911059972e-06, + "loss": 0.9786, + "step": 62867 + }, + { + "epoch": 4.65, + "learning_rate": 2.5558075735691134e-06, + "loss": 0.9435, + "step": 62868 + }, + { + "epoch": 4.65, + "learning_rate": 2.555541247921858e-06, + "loss": 1.0417, + "step": 62869 + }, + { + "epoch": 4.65, + "learning_rate": 2.55527493411863e-06, + "loss": 1.0181, + "step": 62870 + }, + { + "epoch": 4.65, + "learning_rate": 2.55500863215985e-06, + "loss": 0.9816, + "step": 62871 + }, + { + "epoch": 4.65, + "learning_rate": 2.554742342045948e-06, + "loss": 1.0226, + "step": 62872 + }, + { + "epoch": 4.65, + "learning_rate": 2.5544760637773445e-06, + "loss": 1.1631, + "step": 62873 + }, + { + "epoch": 4.65, + "learning_rate": 2.554209797354462e-06, + "loss": 1.1214, + "step": 62874 + }, + { + "epoch": 4.65, + "learning_rate": 2.5539435427777236e-06, + "loss": 1.0118, + "step": 62875 + }, + { + "epoch": 4.65, + "learning_rate": 2.5536773000475567e-06, + "loss": 0.8242, + "step": 62876 + }, + { + "epoch": 4.65, + "learning_rate": 2.553411069164383e-06, + "loss": 0.9162, + "step": 62877 + }, + { + "epoch": 4.65, + "learning_rate": 2.553144850128626e-06, + "loss": 1.097, + "step": 62878 + }, + { + "epoch": 4.65, + "learning_rate": 2.552878642940708e-06, + "loss": 0.9654, + "step": 62879 + }, + { + "epoch": 4.65, + "learning_rate": 2.5526124476010515e-06, + "loss": 0.9776, + "step": 62880 + }, + { + "epoch": 4.65, + "learning_rate": 2.552346264110085e-06, + "loss": 0.9547, + "step": 62881 + }, + { + "epoch": 4.65, + "learning_rate": 2.5520800924682287e-06, + "loss": 0.9644, + "step": 62882 + }, + { + "epoch": 4.65, + "learning_rate": 2.551813932675906e-06, + "loss": 1.0016, + "step": 62883 + }, + { + "epoch": 4.65, + "learning_rate": 2.551547784733539e-06, + "loss": 1.0029, + "step": 62884 + }, + { + "epoch": 4.65, + "learning_rate": 2.551281648641555e-06, + "loss": 1.0239, + "step": 62885 + }, + { + "epoch": 4.65, + "learning_rate": 2.5510155244003753e-06, + "loss": 0.8546, + "step": 62886 + }, + { + "epoch": 4.65, + "learning_rate": 2.5507494120104237e-06, + "loss": 1.0014, + "step": 62887 + }, + { + "epoch": 4.65, + "learning_rate": 2.5504833114721227e-06, + "loss": 0.9794, + "step": 62888 + }, + { + "epoch": 4.65, + "learning_rate": 2.5502172227858934e-06, + "loss": 1.0554, + "step": 62889 + }, + { + "epoch": 4.65, + "learning_rate": 2.5499511459521674e-06, + "loss": 0.9512, + "step": 62890 + }, + { + "epoch": 4.65, + "learning_rate": 2.549685080971358e-06, + "loss": 0.9939, + "step": 62891 + }, + { + "epoch": 4.65, + "learning_rate": 2.5494190278438945e-06, + "loss": 0.9643, + "step": 62892 + }, + { + "epoch": 4.65, + "learning_rate": 2.5491529865701993e-06, + "loss": 1.0155, + "step": 62893 + }, + { + "epoch": 4.65, + "learning_rate": 2.548886957150696e-06, + "loss": 1.0303, + "step": 62894 + }, + { + "epoch": 4.65, + "learning_rate": 2.548620939585802e-06, + "loss": 0.9105, + "step": 62895 + }, + { + "epoch": 4.65, + "learning_rate": 2.5483549338759506e-06, + "loss": 0.9528, + "step": 62896 + }, + { + "epoch": 4.65, + "learning_rate": 2.548088940021558e-06, + "loss": 1.1164, + "step": 62897 + }, + { + "epoch": 4.65, + "learning_rate": 2.5478229580230508e-06, + "loss": 0.8758, + "step": 62898 + }, + { + "epoch": 4.65, + "learning_rate": 2.5475569878808494e-06, + "loss": 1.0198, + "step": 62899 + }, + { + "epoch": 4.65, + "learning_rate": 2.5472910295953758e-06, + "loss": 0.9257, + "step": 62900 + }, + { + "epoch": 4.65, + "learning_rate": 2.5470250831670586e-06, + "loss": 0.9933, + "step": 62901 + }, + { + "epoch": 4.65, + "learning_rate": 2.5467591485963183e-06, + "loss": 0.9639, + "step": 62902 + }, + { + "epoch": 4.65, + "learning_rate": 2.546493225883576e-06, + "loss": 0.9844, + "step": 62903 + }, + { + "epoch": 4.65, + "learning_rate": 2.5462273150292548e-06, + "loss": 0.9096, + "step": 62904 + }, + { + "epoch": 4.65, + "learning_rate": 2.545961416033782e-06, + "loss": 0.9848, + "step": 62905 + }, + { + "epoch": 4.65, + "learning_rate": 2.5456955288975804e-06, + "loss": 0.9351, + "step": 62906 + }, + { + "epoch": 4.65, + "learning_rate": 2.5454296536210654e-06, + "loss": 1.1472, + "step": 62907 + }, + { + "epoch": 4.65, + "learning_rate": 2.5451637902046677e-06, + "loss": 0.8816, + "step": 62908 + }, + { + "epoch": 4.65, + "learning_rate": 2.5448979386488047e-06, + "loss": 0.9461, + "step": 62909 + }, + { + "epoch": 4.65, + "learning_rate": 2.5446320989539085e-06, + "loss": 1.0353, + "step": 62910 + }, + { + "epoch": 4.65, + "learning_rate": 2.54436627112039e-06, + "loss": 1.0952, + "step": 62911 + }, + { + "epoch": 4.65, + "learning_rate": 2.5441004551486814e-06, + "loss": 1.0412, + "step": 62912 + }, + { + "epoch": 4.65, + "learning_rate": 2.5438346510391987e-06, + "loss": 1.0437, + "step": 62913 + }, + { + "epoch": 4.65, + "learning_rate": 2.5435688587923747e-06, + "loss": 1.128, + "step": 62914 + }, + { + "epoch": 4.65, + "learning_rate": 2.5433030784086233e-06, + "loss": 1.123, + "step": 62915 + }, + { + "epoch": 4.65, + "learning_rate": 2.5430373098883666e-06, + "loss": 1.0624, + "step": 62916 + }, + { + "epoch": 4.65, + "learning_rate": 2.542771553232033e-06, + "loss": 0.9964, + "step": 62917 + }, + { + "epoch": 4.65, + "learning_rate": 2.5425058084400443e-06, + "loss": 0.9494, + "step": 62918 + }, + { + "epoch": 4.65, + "learning_rate": 2.542240075512822e-06, + "loss": 0.9648, + "step": 62919 + }, + { + "epoch": 4.65, + "learning_rate": 2.5419743544507858e-06, + "loss": 0.9988, + "step": 62920 + }, + { + "epoch": 4.65, + "learning_rate": 2.541708645254365e-06, + "loss": 1.0249, + "step": 62921 + }, + { + "epoch": 4.65, + "learning_rate": 2.5414429479239787e-06, + "loss": 0.9731, + "step": 62922 + }, + { + "epoch": 4.65, + "learning_rate": 2.5411772624600505e-06, + "loss": 1.0128, + "step": 62923 + }, + { + "epoch": 4.65, + "learning_rate": 2.540911588863002e-06, + "loss": 1.0244, + "step": 62924 + }, + { + "epoch": 4.65, + "learning_rate": 2.540645927133253e-06, + "loss": 0.9358, + "step": 62925 + }, + { + "epoch": 4.65, + "learning_rate": 2.5403802772712337e-06, + "loss": 0.9684, + "step": 62926 + }, + { + "epoch": 4.65, + "learning_rate": 2.5401146392773625e-06, + "loss": 0.9538, + "step": 62927 + }, + { + "epoch": 4.65, + "learning_rate": 2.5398490131520615e-06, + "loss": 1.0089, + "step": 62928 + }, + { + "epoch": 4.65, + "learning_rate": 2.539583398895751e-06, + "loss": 1.1023, + "step": 62929 + }, + { + "epoch": 4.65, + "learning_rate": 2.53931779650886e-06, + "loss": 0.9456, + "step": 62930 + }, + { + "epoch": 4.65, + "learning_rate": 2.5390522059918075e-06, + "loss": 0.9651, + "step": 62931 + }, + { + "epoch": 4.65, + "learning_rate": 2.538786627345017e-06, + "loss": 1.012, + "step": 62932 + }, + { + "epoch": 4.65, + "learning_rate": 2.5385210605689093e-06, + "loss": 1.0582, + "step": 62933 + }, + { + "epoch": 4.65, + "learning_rate": 2.5382555056639046e-06, + "loss": 0.9589, + "step": 62934 + }, + { + "epoch": 4.65, + "learning_rate": 2.537989962630435e-06, + "loss": 1.0384, + "step": 62935 + }, + { + "epoch": 4.65, + "learning_rate": 2.5377244314689107e-06, + "loss": 0.9412, + "step": 62936 + }, + { + "epoch": 4.65, + "learning_rate": 2.537458912179763e-06, + "loss": 1.051, + "step": 62937 + }, + { + "epoch": 4.65, + "learning_rate": 2.5371934047634107e-06, + "loss": 0.9805, + "step": 62938 + }, + { + "epoch": 4.65, + "learning_rate": 2.5369279092202783e-06, + "loss": 0.9967, + "step": 62939 + }, + { + "epoch": 4.65, + "learning_rate": 2.5366624255507833e-06, + "loss": 1.0297, + "step": 62940 + }, + { + "epoch": 4.65, + "learning_rate": 2.5363969537553533e-06, + "loss": 0.9921, + "step": 62941 + }, + { + "epoch": 4.65, + "learning_rate": 2.5361314938344104e-06, + "loss": 0.9449, + "step": 62942 + }, + { + "epoch": 4.65, + "learning_rate": 2.5358660457883756e-06, + "loss": 0.9598, + "step": 62943 + }, + { + "epoch": 4.65, + "learning_rate": 2.535600609617671e-06, + "loss": 1.059, + "step": 62944 + }, + { + "epoch": 4.65, + "learning_rate": 2.5353351853227146e-06, + "loss": 1.0628, + "step": 62945 + }, + { + "epoch": 4.65, + "learning_rate": 2.5350697729039376e-06, + "loss": 0.9615, + "step": 62946 + }, + { + "epoch": 4.65, + "learning_rate": 2.5348043723617577e-06, + "loss": 1.054, + "step": 62947 + }, + { + "epoch": 4.65, + "learning_rate": 2.534538983696597e-06, + "loss": 1.0604, + "step": 62948 + }, + { + "epoch": 4.65, + "learning_rate": 2.5342736069088746e-06, + "loss": 1.07, + "step": 62949 + }, + { + "epoch": 4.65, + "learning_rate": 2.53400824199902e-06, + "loss": 1.0388, + "step": 62950 + }, + { + "epoch": 4.65, + "learning_rate": 2.533742888967451e-06, + "loss": 1.0539, + "step": 62951 + }, + { + "epoch": 4.65, + "learning_rate": 2.5334775478145903e-06, + "loss": 1.0645, + "step": 62952 + }, + { + "epoch": 4.65, + "learning_rate": 2.533212218540859e-06, + "loss": 0.9684, + "step": 62953 + }, + { + "epoch": 4.65, + "learning_rate": 2.532946901146678e-06, + "loss": 1.022, + "step": 62954 + }, + { + "epoch": 4.65, + "learning_rate": 2.532681595632478e-06, + "loss": 0.8822, + "step": 62955 + }, + { + "epoch": 4.65, + "learning_rate": 2.532416301998668e-06, + "loss": 0.9622, + "step": 62956 + }, + { + "epoch": 4.65, + "learning_rate": 2.5321510202456813e-06, + "loss": 1.0205, + "step": 62957 + }, + { + "epoch": 4.65, + "learning_rate": 2.5318857503739304e-06, + "loss": 0.9389, + "step": 62958 + }, + { + "epoch": 4.65, + "learning_rate": 2.5316204923838493e-06, + "loss": 0.9815, + "step": 62959 + }, + { + "epoch": 4.65, + "learning_rate": 2.5313552462758474e-06, + "loss": 0.8964, + "step": 62960 + }, + { + "epoch": 4.65, + "learning_rate": 2.5310900120503557e-06, + "loss": 1.049, + "step": 62961 + }, + { + "epoch": 4.65, + "learning_rate": 2.530824789707792e-06, + "loss": 1.0418, + "step": 62962 + }, + { + "epoch": 4.65, + "learning_rate": 2.5305595792485804e-06, + "loss": 0.9441, + "step": 62963 + }, + { + "epoch": 4.65, + "learning_rate": 2.5302943806731406e-06, + "loss": 0.8921, + "step": 62964 + }, + { + "epoch": 4.65, + "learning_rate": 2.530029193981893e-06, + "loss": 1.0421, + "step": 62965 + }, + { + "epoch": 4.65, + "learning_rate": 2.529764019175265e-06, + "loss": 1.0487, + "step": 62966 + }, + { + "epoch": 4.65, + "learning_rate": 2.529498856253676e-06, + "loss": 1.0024, + "step": 62967 + }, + { + "epoch": 4.65, + "learning_rate": 2.5292337052175465e-06, + "loss": 1.0574, + "step": 62968 + }, + { + "epoch": 4.65, + "learning_rate": 2.5289685660672967e-06, + "loss": 0.9117, + "step": 62969 + }, + { + "epoch": 4.65, + "learning_rate": 2.528703438803355e-06, + "loss": 1.0362, + "step": 62970 + }, + { + "epoch": 4.65, + "learning_rate": 2.5284383234261376e-06, + "loss": 0.9934, + "step": 62971 + }, + { + "epoch": 4.65, + "learning_rate": 2.5281732199360685e-06, + "loss": 1.0543, + "step": 62972 + }, + { + "epoch": 4.65, + "learning_rate": 2.5279081283335694e-06, + "loss": 1.1065, + "step": 62973 + }, + { + "epoch": 4.65, + "learning_rate": 2.5276430486190583e-06, + "loss": 0.9062, + "step": 62974 + }, + { + "epoch": 4.65, + "learning_rate": 2.5273779807929633e-06, + "loss": 0.921, + "step": 62975 + }, + { + "epoch": 4.65, + "learning_rate": 2.5271129248557025e-06, + "loss": 0.9735, + "step": 62976 + }, + { + "epoch": 4.65, + "learning_rate": 2.526847880807699e-06, + "loss": 0.9442, + "step": 62977 + }, + { + "epoch": 4.65, + "learning_rate": 2.5265828486493703e-06, + "loss": 1.0975, + "step": 62978 + }, + { + "epoch": 4.65, + "learning_rate": 2.5263178283811442e-06, + "loss": 1.0311, + "step": 62979 + }, + { + "epoch": 4.65, + "learning_rate": 2.5260528200034417e-06, + "loss": 0.9883, + "step": 62980 + }, + { + "epoch": 4.65, + "learning_rate": 2.525787823516678e-06, + "loss": 0.9733, + "step": 62981 + }, + { + "epoch": 4.65, + "learning_rate": 2.5255228389212803e-06, + "loss": 1.0675, + "step": 62982 + }, + { + "epoch": 4.65, + "learning_rate": 2.5252578662176697e-06, + "loss": 1.1176, + "step": 62983 + }, + { + "epoch": 4.65, + "learning_rate": 2.524992905406266e-06, + "loss": 1.0695, + "step": 62984 + }, + { + "epoch": 4.65, + "learning_rate": 2.5247279564874894e-06, + "loss": 0.9835, + "step": 62985 + }, + { + "epoch": 4.65, + "learning_rate": 2.524463019461767e-06, + "loss": 1.0338, + "step": 62986 + }, + { + "epoch": 4.65, + "learning_rate": 2.5241980943295165e-06, + "loss": 1.0919, + "step": 62987 + }, + { + "epoch": 4.65, + "learning_rate": 2.5239331810911595e-06, + "loss": 1.0404, + "step": 62988 + }, + { + "epoch": 4.65, + "learning_rate": 2.5236682797471156e-06, + "loss": 1.0483, + "step": 62989 + }, + { + "epoch": 4.65, + "learning_rate": 2.5234033902978105e-06, + "loss": 0.9396, + "step": 62990 + }, + { + "epoch": 4.65, + "learning_rate": 2.523138512743665e-06, + "loss": 0.9858, + "step": 62991 + }, + { + "epoch": 4.65, + "learning_rate": 2.522873647085098e-06, + "loss": 1.0895, + "step": 62992 + }, + { + "epoch": 4.65, + "learning_rate": 2.5226087933225317e-06, + "loss": 1.0605, + "step": 62993 + }, + { + "epoch": 4.65, + "learning_rate": 2.5223439514563863e-06, + "loss": 0.9563, + "step": 62994 + }, + { + "epoch": 4.65, + "learning_rate": 2.5220791214870866e-06, + "loss": 1.0549, + "step": 62995 + }, + { + "epoch": 4.65, + "learning_rate": 2.521814303415052e-06, + "loss": 1.1276, + "step": 62996 + }, + { + "epoch": 4.65, + "learning_rate": 2.521549497240705e-06, + "loss": 0.994, + "step": 62997 + }, + { + "epoch": 4.65, + "learning_rate": 2.5212847029644615e-06, + "loss": 1.0242, + "step": 62998 + }, + { + "epoch": 4.65, + "learning_rate": 2.5210199205867515e-06, + "loss": 0.9471, + "step": 62999 + }, + { + "epoch": 4.66, + "learning_rate": 2.520755150107993e-06, + "loss": 1.0334, + "step": 63000 + }, + { + "epoch": 4.66, + "learning_rate": 2.5204903915286005e-06, + "loss": 1.1167, + "step": 63001 + }, + { + "epoch": 4.66, + "learning_rate": 2.5202256448490047e-06, + "loss": 1.0138, + "step": 63002 + }, + { + "epoch": 4.66, + "learning_rate": 2.51996091006962e-06, + "loss": 1.046, + "step": 63003 + }, + { + "epoch": 4.66, + "learning_rate": 2.5196961871908755e-06, + "loss": 0.9001, + "step": 63004 + }, + { + "epoch": 4.66, + "learning_rate": 2.519431476213181e-06, + "loss": 1.0924, + "step": 63005 + }, + { + "epoch": 4.66, + "learning_rate": 2.519166777136968e-06, + "loss": 0.9587, + "step": 63006 + }, + { + "epoch": 4.66, + "learning_rate": 2.5189020899626504e-06, + "loss": 0.9863, + "step": 63007 + }, + { + "epoch": 4.66, + "learning_rate": 2.518637414690658e-06, + "loss": 0.9916, + "step": 63008 + }, + { + "epoch": 4.66, + "learning_rate": 2.518372751321404e-06, + "loss": 0.9514, + "step": 63009 + }, + { + "epoch": 4.66, + "learning_rate": 2.518108099855309e-06, + "loss": 0.9707, + "step": 63010 + }, + { + "epoch": 4.66, + "learning_rate": 2.5178434602927993e-06, + "loss": 1.0172, + "step": 63011 + }, + { + "epoch": 4.66, + "learning_rate": 2.5175788326342944e-06, + "loss": 0.9864, + "step": 63012 + }, + { + "epoch": 4.66, + "learning_rate": 2.5173142168802133e-06, + "loss": 0.9835, + "step": 63013 + }, + { + "epoch": 4.66, + "learning_rate": 2.517049613030976e-06, + "loss": 1.0784, + "step": 63014 + }, + { + "epoch": 4.66, + "learning_rate": 2.5167850210870093e-06, + "loss": 1.0101, + "step": 63015 + }, + { + "epoch": 4.66, + "learning_rate": 2.5165204410487297e-06, + "loss": 0.9729, + "step": 63016 + }, + { + "epoch": 4.66, + "learning_rate": 2.51625587291656e-06, + "loss": 0.9797, + "step": 63017 + }, + { + "epoch": 4.66, + "learning_rate": 2.515991316690919e-06, + "loss": 1.0705, + "step": 63018 + }, + { + "epoch": 4.66, + "learning_rate": 2.5157267723722257e-06, + "loss": 0.9979, + "step": 63019 + }, + { + "epoch": 4.66, + "learning_rate": 2.5154622399609074e-06, + "loss": 1.0367, + "step": 63020 + }, + { + "epoch": 4.66, + "learning_rate": 2.5151977194573817e-06, + "loss": 0.909, + "step": 63021 + }, + { + "epoch": 4.66, + "learning_rate": 2.5149332108620693e-06, + "loss": 1.0928, + "step": 63022 + }, + { + "epoch": 4.66, + "learning_rate": 2.514668714175389e-06, + "loss": 1.0016, + "step": 63023 + }, + { + "epoch": 4.66, + "learning_rate": 2.514404229397769e-06, + "loss": 0.9153, + "step": 63024 + }, + { + "epoch": 4.66, + "learning_rate": 2.5141397565296175e-06, + "loss": 1.0481, + "step": 63025 + }, + { + "epoch": 4.66, + "learning_rate": 2.5138752955713674e-06, + "loss": 1.0035, + "step": 63026 + }, + { + "epoch": 4.66, + "learning_rate": 2.513610846523433e-06, + "loss": 1.0415, + "step": 63027 + }, + { + "epoch": 4.66, + "learning_rate": 2.513346409386238e-06, + "loss": 0.9251, + "step": 63028 + }, + { + "epoch": 4.66, + "learning_rate": 2.5130819841602006e-06, + "loss": 0.9326, + "step": 63029 + }, + { + "epoch": 4.66, + "learning_rate": 2.5128175708457403e-06, + "loss": 1.0793, + "step": 63030 + }, + { + "epoch": 4.66, + "learning_rate": 2.512553169443284e-06, + "loss": 0.9462, + "step": 63031 + }, + { + "epoch": 4.66, + "learning_rate": 2.5122887799532468e-06, + "loss": 1.0155, + "step": 63032 + }, + { + "epoch": 4.66, + "learning_rate": 2.512024402376052e-06, + "loss": 0.9994, + "step": 63033 + }, + { + "epoch": 4.66, + "learning_rate": 2.511760036712115e-06, + "loss": 1.0691, + "step": 63034 + }, + { + "epoch": 4.66, + "learning_rate": 2.5114956829618654e-06, + "loss": 0.9452, + "step": 63035 + }, + { + "epoch": 4.66, + "learning_rate": 2.511231341125718e-06, + "loss": 1.0574, + "step": 63036 + }, + { + "epoch": 4.66, + "learning_rate": 2.510967011204095e-06, + "loss": 1.0361, + "step": 63037 + }, + { + "epoch": 4.66, + "learning_rate": 2.5107026931974153e-06, + "loss": 1.0092, + "step": 63038 + }, + { + "epoch": 4.66, + "learning_rate": 2.510438387106099e-06, + "loss": 1.0141, + "step": 63039 + }, + { + "epoch": 4.66, + "learning_rate": 2.5101740929305696e-06, + "loss": 0.9175, + "step": 63040 + }, + { + "epoch": 4.66, + "learning_rate": 2.509909810671247e-06, + "loss": 1.0684, + "step": 63041 + }, + { + "epoch": 4.66, + "learning_rate": 2.50964554032855e-06, + "loss": 1.0461, + "step": 63042 + }, + { + "epoch": 4.66, + "learning_rate": 2.509381281902897e-06, + "loss": 1.0077, + "step": 63043 + }, + { + "epoch": 4.66, + "learning_rate": 2.5091170353947147e-06, + "loss": 1.0254, + "step": 63044 + }, + { + "epoch": 4.66, + "learning_rate": 2.50885280080442e-06, + "loss": 1.0637, + "step": 63045 + }, + { + "epoch": 4.66, + "learning_rate": 2.5085885781324326e-06, + "loss": 1.0121, + "step": 63046 + }, + { + "epoch": 4.66, + "learning_rate": 2.5083243673791736e-06, + "loss": 0.9868, + "step": 63047 + }, + { + "epoch": 4.66, + "learning_rate": 2.508060168545061e-06, + "loss": 1.0296, + "step": 63048 + }, + { + "epoch": 4.66, + "learning_rate": 2.507795981630523e-06, + "loss": 0.9471, + "step": 63049 + }, + { + "epoch": 4.66, + "learning_rate": 2.507531806635969e-06, + "loss": 1.026, + "step": 63050 + }, + { + "epoch": 4.66, + "learning_rate": 2.507267643561827e-06, + "loss": 0.9977, + "step": 63051 + }, + { + "epoch": 4.66, + "learning_rate": 2.507003492408512e-06, + "loss": 0.936, + "step": 63052 + }, + { + "epoch": 4.66, + "learning_rate": 2.506739353176454e-06, + "loss": 0.9864, + "step": 63053 + }, + { + "epoch": 4.66, + "learning_rate": 2.5064752258660597e-06, + "loss": 1.0535, + "step": 63054 + }, + { + "epoch": 4.66, + "learning_rate": 2.5062111104777597e-06, + "loss": 1.0213, + "step": 63055 + }, + { + "epoch": 4.66, + "learning_rate": 2.5059470070119694e-06, + "loss": 1.0783, + "step": 63056 + }, + { + "epoch": 4.66, + "learning_rate": 2.5056829154691122e-06, + "loss": 1.0809, + "step": 63057 + }, + { + "epoch": 4.66, + "learning_rate": 2.5054188358496045e-06, + "loss": 0.9956, + "step": 63058 + }, + { + "epoch": 4.66, + "learning_rate": 2.5051547681538656e-06, + "loss": 1.0328, + "step": 63059 + }, + { + "epoch": 4.66, + "learning_rate": 2.504890712382322e-06, + "loss": 0.9294, + "step": 63060 + }, + { + "epoch": 4.66, + "learning_rate": 2.504626668535389e-06, + "loss": 0.9021, + "step": 63061 + }, + { + "epoch": 4.66, + "learning_rate": 2.5043626366134887e-06, + "loss": 1.0044, + "step": 63062 + }, + { + "epoch": 4.66, + "learning_rate": 2.5040986166170368e-06, + "loss": 1.0527, + "step": 63063 + }, + { + "epoch": 4.66, + "learning_rate": 2.5038346085464593e-06, + "loss": 1.0105, + "step": 63064 + }, + { + "epoch": 4.66, + "learning_rate": 2.503570612402174e-06, + "loss": 1.0554, + "step": 63065 + }, + { + "epoch": 4.66, + "learning_rate": 2.503306628184601e-06, + "loss": 0.9832, + "step": 63066 + }, + { + "epoch": 4.66, + "learning_rate": 2.503042655894159e-06, + "loss": 1.065, + "step": 63067 + }, + { + "epoch": 4.66, + "learning_rate": 2.502778695531267e-06, + "loss": 1.01, + "step": 63068 + }, + { + "epoch": 4.66, + "learning_rate": 2.5025147470963517e-06, + "loss": 1.0213, + "step": 63069 + }, + { + "epoch": 4.66, + "learning_rate": 2.5022508105898236e-06, + "loss": 0.8864, + "step": 63070 + }, + { + "epoch": 4.66, + "learning_rate": 2.5019868860121088e-06, + "loss": 1.0717, + "step": 63071 + }, + { + "epoch": 4.66, + "learning_rate": 2.5017229733636227e-06, + "loss": 0.9054, + "step": 63072 + }, + { + "epoch": 4.66, + "learning_rate": 2.5014590726447952e-06, + "loss": 0.982, + "step": 63073 + }, + { + "epoch": 4.66, + "learning_rate": 2.5011951838560342e-06, + "loss": 0.9951, + "step": 63074 + }, + { + "epoch": 4.66, + "learning_rate": 2.5009313069977636e-06, + "loss": 1.011, + "step": 63075 + }, + { + "epoch": 4.66, + "learning_rate": 2.5006674420704057e-06, + "loss": 0.9435, + "step": 63076 + }, + { + "epoch": 4.66, + "learning_rate": 2.5004035890743784e-06, + "loss": 0.9938, + "step": 63077 + }, + { + "epoch": 4.66, + "learning_rate": 2.500139748010102e-06, + "loss": 0.9881, + "step": 63078 + }, + { + "epoch": 4.66, + "learning_rate": 2.499875918877993e-06, + "loss": 0.9843, + "step": 63079 + }, + { + "epoch": 4.66, + "learning_rate": 2.4996121016784767e-06, + "loss": 1.0948, + "step": 63080 + }, + { + "epoch": 4.66, + "learning_rate": 2.499348296411971e-06, + "loss": 1.0087, + "step": 63081 + }, + { + "epoch": 4.66, + "learning_rate": 2.4990845030788945e-06, + "loss": 1.0044, + "step": 63082 + }, + { + "epoch": 4.66, + "learning_rate": 2.498820721679668e-06, + "loss": 0.9121, + "step": 63083 + }, + { + "epoch": 4.66, + "learning_rate": 2.4985569522147057e-06, + "loss": 1.0072, + "step": 63084 + }, + { + "epoch": 4.66, + "learning_rate": 2.4982931946844356e-06, + "loss": 1.0126, + "step": 63085 + }, + { + "epoch": 4.66, + "learning_rate": 2.4980294490892744e-06, + "loss": 0.9776, + "step": 63086 + }, + { + "epoch": 4.66, + "learning_rate": 2.497765715429641e-06, + "loss": 0.9735, + "step": 63087 + }, + { + "epoch": 4.66, + "learning_rate": 2.4975019937059517e-06, + "loss": 1.1061, + "step": 63088 + }, + { + "epoch": 4.66, + "learning_rate": 2.4972382839186314e-06, + "loss": 0.9292, + "step": 63089 + }, + { + "epoch": 4.66, + "learning_rate": 2.4969745860680992e-06, + "loss": 0.8809, + "step": 63090 + }, + { + "epoch": 4.66, + "learning_rate": 2.496710900154772e-06, + "loss": 0.9155, + "step": 63091 + }, + { + "epoch": 4.66, + "learning_rate": 2.4964472261790707e-06, + "loss": 0.8272, + "step": 63092 + }, + { + "epoch": 4.66, + "learning_rate": 2.496183564141412e-06, + "loss": 1.0201, + "step": 63093 + }, + { + "epoch": 4.66, + "learning_rate": 2.495919914042223e-06, + "loss": 0.8703, + "step": 63094 + }, + { + "epoch": 4.66, + "learning_rate": 2.4956562758819137e-06, + "loss": 1.0656, + "step": 63095 + }, + { + "epoch": 4.66, + "learning_rate": 2.4953926496609095e-06, + "loss": 0.8705, + "step": 63096 + }, + { + "epoch": 4.66, + "learning_rate": 2.4951290353796254e-06, + "loss": 1.0834, + "step": 63097 + }, + { + "epoch": 4.66, + "learning_rate": 2.494865433038489e-06, + "loss": 0.9731, + "step": 63098 + }, + { + "epoch": 4.66, + "learning_rate": 2.4946018426379093e-06, + "loss": 1.0359, + "step": 63099 + }, + { + "epoch": 4.66, + "learning_rate": 2.494338264178313e-06, + "loss": 1.0912, + "step": 63100 + }, + { + "epoch": 4.66, + "learning_rate": 2.4940746976601148e-06, + "loss": 0.8555, + "step": 63101 + }, + { + "epoch": 4.66, + "learning_rate": 2.493811143083742e-06, + "loss": 1.042, + "step": 63102 + }, + { + "epoch": 4.66, + "learning_rate": 2.493547600449606e-06, + "loss": 1.1436, + "step": 63103 + }, + { + "epoch": 4.66, + "learning_rate": 2.493284069758124e-06, + "loss": 0.9231, + "step": 63104 + }, + { + "epoch": 4.66, + "learning_rate": 2.4930205510097237e-06, + "loss": 0.9896, + "step": 63105 + }, + { + "epoch": 4.66, + "learning_rate": 2.49275704420482e-06, + "loss": 0.9654, + "step": 63106 + }, + { + "epoch": 4.66, + "learning_rate": 2.4924935493438328e-06, + "loss": 1.0016, + "step": 63107 + }, + { + "epoch": 4.66, + "learning_rate": 2.492230066427177e-06, + "loss": 1.0333, + "step": 63108 + }, + { + "epoch": 4.66, + "learning_rate": 2.4919665954552794e-06, + "loss": 1.0629, + "step": 63109 + }, + { + "epoch": 4.66, + "learning_rate": 2.491703136428556e-06, + "loss": 1.1183, + "step": 63110 + }, + { + "epoch": 4.66, + "learning_rate": 2.491439689347425e-06, + "loss": 0.9977, + "step": 63111 + }, + { + "epoch": 4.66, + "learning_rate": 2.4911762542123054e-06, + "loss": 1.0703, + "step": 63112 + }, + { + "epoch": 4.66, + "learning_rate": 2.4909128310236142e-06, + "loss": 1.0471, + "step": 63113 + }, + { + "epoch": 4.66, + "learning_rate": 2.49064941978178e-06, + "loss": 1.0271, + "step": 63114 + }, + { + "epoch": 4.66, + "learning_rate": 2.4903860204872087e-06, + "loss": 0.9832, + "step": 63115 + }, + { + "epoch": 4.66, + "learning_rate": 2.4901226331403283e-06, + "loss": 0.8628, + "step": 63116 + }, + { + "epoch": 4.66, + "learning_rate": 2.4898592577415536e-06, + "loss": 1.0641, + "step": 63117 + }, + { + "epoch": 4.66, + "learning_rate": 2.4895958942913103e-06, + "loss": 1.0225, + "step": 63118 + }, + { + "epoch": 4.66, + "learning_rate": 2.489332542790006e-06, + "loss": 0.944, + "step": 63119 + }, + { + "epoch": 4.66, + "learning_rate": 2.4890692032380703e-06, + "loss": 0.9814, + "step": 63120 + }, + { + "epoch": 4.66, + "learning_rate": 2.488805875635918e-06, + "loss": 0.9829, + "step": 63121 + }, + { + "epoch": 4.66, + "learning_rate": 2.4885425599839676e-06, + "loss": 1.1191, + "step": 63122 + }, + { + "epoch": 4.66, + "learning_rate": 2.4882792562826385e-06, + "loss": 1.0012, + "step": 63123 + }, + { + "epoch": 4.66, + "learning_rate": 2.4880159645323467e-06, + "loss": 0.9457, + "step": 63124 + }, + { + "epoch": 4.66, + "learning_rate": 2.4877526847335174e-06, + "loss": 0.8877, + "step": 63125 + }, + { + "epoch": 4.66, + "learning_rate": 2.487489416886566e-06, + "loss": 0.9709, + "step": 63126 + }, + { + "epoch": 4.66, + "learning_rate": 2.4872261609919113e-06, + "loss": 0.9833, + "step": 63127 + }, + { + "epoch": 4.66, + "learning_rate": 2.4869629170499687e-06, + "loss": 1.0401, + "step": 63128 + }, + { + "epoch": 4.66, + "learning_rate": 2.4866996850611645e-06, + "loss": 0.9487, + "step": 63129 + }, + { + "epoch": 4.66, + "learning_rate": 2.4864364650259133e-06, + "loss": 1.0405, + "step": 63130 + }, + { + "epoch": 4.66, + "learning_rate": 2.486173256944634e-06, + "loss": 0.9073, + "step": 63131 + }, + { + "epoch": 4.66, + "learning_rate": 2.4859100608177454e-06, + "loss": 0.9726, + "step": 63132 + }, + { + "epoch": 4.66, + "learning_rate": 2.485646876645664e-06, + "loss": 1.0793, + "step": 63133 + }, + { + "epoch": 4.66, + "learning_rate": 2.485383704428813e-06, + "loss": 1.0803, + "step": 63134 + }, + { + "epoch": 4.67, + "learning_rate": 2.4851205441676097e-06, + "loss": 1.0881, + "step": 63135 + }, + { + "epoch": 4.67, + "learning_rate": 2.484857395862472e-06, + "loss": 0.9224, + "step": 63136 + }, + { + "epoch": 4.67, + "learning_rate": 2.484594259513816e-06, + "loss": 1.1579, + "step": 63137 + }, + { + "epoch": 4.67, + "learning_rate": 2.4843311351220655e-06, + "loss": 1.0553, + "step": 63138 + }, + { + "epoch": 4.67, + "learning_rate": 2.484068022687639e-06, + "loss": 1.03, + "step": 63139 + }, + { + "epoch": 4.67, + "learning_rate": 2.4838049222109486e-06, + "loss": 0.9354, + "step": 63140 + }, + { + "epoch": 4.67, + "learning_rate": 2.483541833692419e-06, + "loss": 0.9163, + "step": 63141 + }, + { + "epoch": 4.67, + "learning_rate": 2.4832787571324636e-06, + "loss": 0.8921, + "step": 63142 + }, + { + "epoch": 4.67, + "learning_rate": 2.4830156925315097e-06, + "loss": 1.1579, + "step": 63143 + }, + { + "epoch": 4.67, + "learning_rate": 2.482752639889965e-06, + "loss": 0.9506, + "step": 63144 + }, + { + "epoch": 4.67, + "learning_rate": 2.4824895992082566e-06, + "loss": 0.9748, + "step": 63145 + }, + { + "epoch": 4.67, + "learning_rate": 2.482226570486799e-06, + "loss": 0.9881, + "step": 63146 + }, + { + "epoch": 4.67, + "learning_rate": 2.4819635537260114e-06, + "loss": 1.084, + "step": 63147 + }, + { + "epoch": 4.67, + "learning_rate": 2.4817005489263124e-06, + "loss": 1.0327, + "step": 63148 + }, + { + "epoch": 4.67, + "learning_rate": 2.4814375560881176e-06, + "loss": 0.9822, + "step": 63149 + }, + { + "epoch": 4.67, + "learning_rate": 2.481174575211851e-06, + "loss": 1.038, + "step": 63150 + }, + { + "epoch": 4.67, + "learning_rate": 2.4809116062979275e-06, + "loss": 1.0656, + "step": 63151 + }, + { + "epoch": 4.67, + "learning_rate": 2.480648649346766e-06, + "loss": 1.0938, + "step": 63152 + }, + { + "epoch": 4.67, + "learning_rate": 2.480385704358783e-06, + "loss": 1.0972, + "step": 63153 + }, + { + "epoch": 4.67, + "learning_rate": 2.4801227713344013e-06, + "loss": 0.9908, + "step": 63154 + }, + { + "epoch": 4.67, + "learning_rate": 2.4798598502740355e-06, + "loss": 1.0138, + "step": 63155 + }, + { + "epoch": 4.67, + "learning_rate": 2.4795969411781062e-06, + "loss": 0.9728, + "step": 63156 + }, + { + "epoch": 4.67, + "learning_rate": 2.4793340440470272e-06, + "loss": 0.998, + "step": 63157 + }, + { + "epoch": 4.67, + "learning_rate": 2.4790711588812233e-06, + "loss": 0.9634, + "step": 63158 + }, + { + "epoch": 4.67, + "learning_rate": 2.4788082856811125e-06, + "loss": 1.0105, + "step": 63159 + }, + { + "epoch": 4.67, + "learning_rate": 2.478545424447104e-06, + "loss": 0.9534, + "step": 63160 + }, + { + "epoch": 4.67, + "learning_rate": 2.4782825751796257e-06, + "loss": 1.0946, + "step": 63161 + }, + { + "epoch": 4.67, + "learning_rate": 2.4780197378790895e-06, + "loss": 1.0487, + "step": 63162 + }, + { + "epoch": 4.67, + "learning_rate": 2.477756912545921e-06, + "loss": 0.9221, + "step": 63163 + }, + { + "epoch": 4.67, + "learning_rate": 2.4774940991805284e-06, + "loss": 0.9874, + "step": 63164 + }, + { + "epoch": 4.67, + "learning_rate": 2.477231297783338e-06, + "loss": 1.0163, + "step": 63165 + }, + { + "epoch": 4.67, + "learning_rate": 2.4769685083547622e-06, + "loss": 0.9751, + "step": 63166 + }, + { + "epoch": 4.67, + "learning_rate": 2.4767057308952268e-06, + "loss": 0.946, + "step": 63167 + }, + { + "epoch": 4.67, + "learning_rate": 2.4764429654051427e-06, + "loss": 0.997, + "step": 63168 + }, + { + "epoch": 4.67, + "learning_rate": 2.4761802118849275e-06, + "loss": 1.0696, + "step": 63169 + }, + { + "epoch": 4.67, + "learning_rate": 2.475917470335004e-06, + "loss": 0.8849, + "step": 63170 + }, + { + "epoch": 4.67, + "learning_rate": 2.4756547407557897e-06, + "loss": 0.9254, + "step": 63171 + }, + { + "epoch": 4.67, + "learning_rate": 2.4753920231477e-06, + "loss": 1.0576, + "step": 63172 + }, + { + "epoch": 4.67, + "learning_rate": 2.475129317511151e-06, + "loss": 0.9553, + "step": 63173 + }, + { + "epoch": 4.67, + "learning_rate": 2.474866623846567e-06, + "loss": 1.0156, + "step": 63174 + }, + { + "epoch": 4.67, + "learning_rate": 2.4746039421543613e-06, + "loss": 0.9792, + "step": 63175 + }, + { + "epoch": 4.67, + "learning_rate": 2.4743412724349548e-06, + "loss": 0.9251, + "step": 63176 + }, + { + "epoch": 4.67, + "learning_rate": 2.4740786146887618e-06, + "loss": 0.9545, + "step": 63177 + }, + { + "epoch": 4.67, + "learning_rate": 2.4738159689162002e-06, + "loss": 0.9947, + "step": 63178 + }, + { + "epoch": 4.67, + "learning_rate": 2.4735533351176933e-06, + "loss": 1.02, + "step": 63179 + }, + { + "epoch": 4.67, + "learning_rate": 2.4732907132936547e-06, + "loss": 0.9877, + "step": 63180 + }, + { + "epoch": 4.67, + "learning_rate": 2.4730281034445024e-06, + "loss": 0.9298, + "step": 63181 + }, + { + "epoch": 4.67, + "learning_rate": 2.4727655055706524e-06, + "loss": 1.0039, + "step": 63182 + }, + { + "epoch": 4.67, + "learning_rate": 2.472502919672527e-06, + "loss": 1.0137, + "step": 63183 + }, + { + "epoch": 4.67, + "learning_rate": 2.472240345750542e-06, + "loss": 1.0111, + "step": 63184 + }, + { + "epoch": 4.67, + "learning_rate": 2.471977783805115e-06, + "loss": 1.0765, + "step": 63185 + }, + { + "epoch": 4.67, + "learning_rate": 2.4717152338366636e-06, + "loss": 0.9538, + "step": 63186 + }, + { + "epoch": 4.67, + "learning_rate": 2.471452695845603e-06, + "loss": 1.1359, + "step": 63187 + }, + { + "epoch": 4.67, + "learning_rate": 2.4711901698323593e-06, + "loss": 1.0297, + "step": 63188 + }, + { + "epoch": 4.67, + "learning_rate": 2.4709276557973383e-06, + "loss": 0.9712, + "step": 63189 + }, + { + "epoch": 4.67, + "learning_rate": 2.4706651537409667e-06, + "loss": 0.9543, + "step": 63190 + }, + { + "epoch": 4.67, + "learning_rate": 2.470402663663659e-06, + "loss": 1.0409, + "step": 63191 + }, + { + "epoch": 4.67, + "learning_rate": 2.4701401855658327e-06, + "loss": 1.0034, + "step": 63192 + }, + { + "epoch": 4.67, + "learning_rate": 2.469877719447903e-06, + "loss": 1.0007, + "step": 63193 + }, + { + "epoch": 4.67, + "learning_rate": 2.469615265310292e-06, + "loss": 1.1069, + "step": 63194 + }, + { + "epoch": 4.67, + "learning_rate": 2.469352823153417e-06, + "loss": 1.1093, + "step": 63195 + }, + { + "epoch": 4.67, + "learning_rate": 2.4690903929776934e-06, + "loss": 1.0189, + "step": 63196 + }, + { + "epoch": 4.67, + "learning_rate": 2.4688279747835385e-06, + "loss": 1.0033, + "step": 63197 + }, + { + "epoch": 4.67, + "learning_rate": 2.468565568571368e-06, + "loss": 1.0435, + "step": 63198 + }, + { + "epoch": 4.67, + "learning_rate": 2.4683031743416054e-06, + "loss": 1.0522, + "step": 63199 + }, + { + "epoch": 4.67, + "learning_rate": 2.468040792094665e-06, + "loss": 0.9568, + "step": 63200 + }, + { + "epoch": 4.67, + "learning_rate": 2.467778421830963e-06, + "loss": 0.9433, + "step": 63201 + }, + { + "epoch": 4.67, + "learning_rate": 2.4675160635509145e-06, + "loss": 1.1361, + "step": 63202 + }, + { + "epoch": 4.67, + "learning_rate": 2.4672537172549436e-06, + "loss": 0.9344, + "step": 63203 + }, + { + "epoch": 4.67, + "learning_rate": 2.4669913829434644e-06, + "loss": 0.9198, + "step": 63204 + }, + { + "epoch": 4.67, + "learning_rate": 2.4667290606168937e-06, + "loss": 1.0863, + "step": 63205 + }, + { + "epoch": 4.67, + "learning_rate": 2.4664667502756502e-06, + "loss": 0.8632, + "step": 63206 + }, + { + "epoch": 4.67, + "learning_rate": 2.4662044519201467e-06, + "loss": 1.1018, + "step": 63207 + }, + { + "epoch": 4.67, + "learning_rate": 2.4659421655508098e-06, + "loss": 0.9758, + "step": 63208 + }, + { + "epoch": 4.67, + "learning_rate": 2.465679891168046e-06, + "loss": 1.0188, + "step": 63209 + }, + { + "epoch": 4.67, + "learning_rate": 2.46541762877228e-06, + "loss": 1.0536, + "step": 63210 + }, + { + "epoch": 4.67, + "learning_rate": 2.4651553783639237e-06, + "loss": 1.0006, + "step": 63211 + }, + { + "epoch": 4.67, + "learning_rate": 2.464893139943404e-06, + "loss": 1.0406, + "step": 63212 + }, + { + "epoch": 4.67, + "learning_rate": 2.4646309135111246e-06, + "loss": 1.0356, + "step": 63213 + }, + { + "epoch": 4.67, + "learning_rate": 2.464368699067514e-06, + "loss": 0.9801, + "step": 63214 + }, + { + "epoch": 4.67, + "learning_rate": 2.4641064966129835e-06, + "loss": 0.9618, + "step": 63215 + }, + { + "epoch": 4.67, + "learning_rate": 2.463844306147952e-06, + "loss": 1.016, + "step": 63216 + }, + { + "epoch": 4.67, + "learning_rate": 2.4635821276728377e-06, + "loss": 0.941, + "step": 63217 + }, + { + "epoch": 4.67, + "learning_rate": 2.463319961188052e-06, + "loss": 1.0636, + "step": 63218 + }, + { + "epoch": 4.67, + "learning_rate": 2.4630578066940205e-06, + "loss": 1.0738, + "step": 63219 + }, + { + "epoch": 4.67, + "learning_rate": 2.4627956641911553e-06, + "loss": 1.0891, + "step": 63220 + }, + { + "epoch": 4.67, + "learning_rate": 2.462533533679874e-06, + "loss": 0.9912, + "step": 63221 + }, + { + "epoch": 4.67, + "learning_rate": 2.462271415160592e-06, + "loss": 0.9881, + "step": 63222 + }, + { + "epoch": 4.67, + "learning_rate": 2.4620093086337316e-06, + "loss": 0.9116, + "step": 63223 + }, + { + "epoch": 4.67, + "learning_rate": 2.4617472140997056e-06, + "loss": 0.901, + "step": 63224 + }, + { + "epoch": 4.67, + "learning_rate": 2.4614851315589317e-06, + "loss": 0.8907, + "step": 63225 + }, + { + "epoch": 4.67, + "learning_rate": 2.4612230610118283e-06, + "loss": 1.0553, + "step": 63226 + }, + { + "epoch": 4.67, + "learning_rate": 2.4609610024588073e-06, + "loss": 0.9608, + "step": 63227 + }, + { + "epoch": 4.67, + "learning_rate": 2.4606989559002924e-06, + "loss": 1.0097, + "step": 63228 + }, + { + "epoch": 4.67, + "learning_rate": 2.4604369213366975e-06, + "loss": 0.9576, + "step": 63229 + }, + { + "epoch": 4.67, + "learning_rate": 2.4601748987684405e-06, + "loss": 1.0212, + "step": 63230 + }, + { + "epoch": 4.67, + "learning_rate": 2.4599128881959332e-06, + "loss": 1.0136, + "step": 63231 + }, + { + "epoch": 4.67, + "learning_rate": 2.4596508896196003e-06, + "loss": 0.9773, + "step": 63232 + }, + { + "epoch": 4.67, + "learning_rate": 2.4593889030398578e-06, + "loss": 1.1177, + "step": 63233 + }, + { + "epoch": 4.67, + "learning_rate": 2.459126928457114e-06, + "loss": 1.0104, + "step": 63234 + }, + { + "epoch": 4.67, + "learning_rate": 2.4588649658717944e-06, + "loss": 0.9371, + "step": 63235 + }, + { + "epoch": 4.67, + "learning_rate": 2.458603015284312e-06, + "loss": 0.9428, + "step": 63236 + }, + { + "epoch": 4.67, + "learning_rate": 2.4583410766950843e-06, + "loss": 1.0429, + "step": 63237 + }, + { + "epoch": 4.67, + "learning_rate": 2.4580791501045266e-06, + "loss": 0.9049, + "step": 63238 + }, + { + "epoch": 4.67, + "learning_rate": 2.457817235513058e-06, + "loss": 1.1046, + "step": 63239 + }, + { + "epoch": 4.67, + "learning_rate": 2.4575553329210955e-06, + "loss": 1.0094, + "step": 63240 + }, + { + "epoch": 4.67, + "learning_rate": 2.457293442329054e-06, + "loss": 1.135, + "step": 63241 + }, + { + "epoch": 4.67, + "learning_rate": 2.457031563737351e-06, + "loss": 1.0173, + "step": 63242 + }, + { + "epoch": 4.67, + "learning_rate": 2.4567696971464008e-06, + "loss": 0.9486, + "step": 63243 + }, + { + "epoch": 4.67, + "learning_rate": 2.4565078425566234e-06, + "loss": 1.0559, + "step": 63244 + }, + { + "epoch": 4.67, + "learning_rate": 2.456245999968435e-06, + "loss": 1.0266, + "step": 63245 + }, + { + "epoch": 4.67, + "learning_rate": 2.4559841693822507e-06, + "loss": 0.9165, + "step": 63246 + }, + { + "epoch": 4.67, + "learning_rate": 2.4557223507984853e-06, + "loss": 0.9563, + "step": 63247 + }, + { + "epoch": 4.67, + "learning_rate": 2.4554605442175605e-06, + "loss": 0.8419, + "step": 63248 + }, + { + "epoch": 4.67, + "learning_rate": 2.45519874963989e-06, + "loss": 1.001, + "step": 63249 + }, + { + "epoch": 4.67, + "learning_rate": 2.4549369670658906e-06, + "loss": 0.9827, + "step": 63250 + }, + { + "epoch": 4.67, + "learning_rate": 2.4546751964959782e-06, + "loss": 1.0296, + "step": 63251 + }, + { + "epoch": 4.67, + "learning_rate": 2.454413437930566e-06, + "loss": 1.0013, + "step": 63252 + }, + { + "epoch": 4.67, + "learning_rate": 2.4541516913700813e-06, + "loss": 0.9902, + "step": 63253 + }, + { + "epoch": 4.67, + "learning_rate": 2.4538899568149265e-06, + "loss": 0.9978, + "step": 63254 + }, + { + "epoch": 4.67, + "learning_rate": 2.453628234265528e-06, + "loss": 1.0689, + "step": 63255 + }, + { + "epoch": 4.67, + "learning_rate": 2.453366523722296e-06, + "loss": 0.9783, + "step": 63256 + }, + { + "epoch": 4.67, + "learning_rate": 2.453104825185656e-06, + "loss": 0.9802, + "step": 63257 + }, + { + "epoch": 4.67, + "learning_rate": 2.4528431386560127e-06, + "loss": 1.1381, + "step": 63258 + }, + { + "epoch": 4.67, + "learning_rate": 2.45258146413379e-06, + "loss": 0.9535, + "step": 63259 + }, + { + "epoch": 4.67, + "learning_rate": 2.4523198016193996e-06, + "loss": 1.0856, + "step": 63260 + }, + { + "epoch": 4.67, + "learning_rate": 2.452058151113266e-06, + "loss": 0.9216, + "step": 63261 + }, + { + "epoch": 4.67, + "learning_rate": 2.4517965126157972e-06, + "loss": 1.0326, + "step": 63262 + }, + { + "epoch": 4.67, + "learning_rate": 2.45153488612741e-06, + "loss": 0.9583, + "step": 63263 + }, + { + "epoch": 4.67, + "learning_rate": 2.4512732716485243e-06, + "loss": 0.9859, + "step": 63264 + }, + { + "epoch": 4.67, + "learning_rate": 2.451011669179556e-06, + "loss": 1.0321, + "step": 63265 + }, + { + "epoch": 4.67, + "learning_rate": 2.4507500787209193e-06, + "loss": 1.0093, + "step": 63266 + }, + { + "epoch": 4.67, + "learning_rate": 2.4504885002730295e-06, + "loss": 0.8939, + "step": 63267 + }, + { + "epoch": 4.67, + "learning_rate": 2.450226933836306e-06, + "loss": 0.974, + "step": 63268 + }, + { + "epoch": 4.67, + "learning_rate": 2.4499653794111643e-06, + "loss": 0.9988, + "step": 63269 + }, + { + "epoch": 4.67, + "learning_rate": 2.44970383699802e-06, + "loss": 1.0026, + "step": 63270 + }, + { + "epoch": 4.68, + "learning_rate": 2.449442306597287e-06, + "loss": 1.0958, + "step": 63271 + }, + { + "epoch": 4.68, + "learning_rate": 2.4491807882093823e-06, + "loss": 0.9421, + "step": 63272 + }, + { + "epoch": 4.68, + "learning_rate": 2.448919281834725e-06, + "loss": 0.9546, + "step": 63273 + }, + { + "epoch": 4.68, + "learning_rate": 2.4486577874737293e-06, + "loss": 0.954, + "step": 63274 + }, + { + "epoch": 4.68, + "learning_rate": 2.448396305126811e-06, + "loss": 1.0805, + "step": 63275 + }, + { + "epoch": 4.68, + "learning_rate": 2.4481348347943833e-06, + "loss": 0.9599, + "step": 63276 + }, + { + "epoch": 4.68, + "learning_rate": 2.447873376476868e-06, + "loss": 0.9826, + "step": 63277 + }, + { + "epoch": 4.68, + "learning_rate": 2.447611930174678e-06, + "loss": 1.1465, + "step": 63278 + }, + { + "epoch": 4.68, + "learning_rate": 2.447350495888229e-06, + "loss": 1.0277, + "step": 63279 + }, + { + "epoch": 4.68, + "learning_rate": 2.4470890736179375e-06, + "loss": 0.9652, + "step": 63280 + }, + { + "epoch": 4.68, + "learning_rate": 2.4468276633642196e-06, + "loss": 0.9355, + "step": 63281 + }, + { + "epoch": 4.68, + "learning_rate": 2.4465662651274902e-06, + "loss": 1.02, + "step": 63282 + }, + { + "epoch": 4.68, + "learning_rate": 2.446304878908163e-06, + "loss": 0.8977, + "step": 63283 + }, + { + "epoch": 4.68, + "learning_rate": 2.44604350470666e-06, + "loss": 0.9434, + "step": 63284 + }, + { + "epoch": 4.68, + "learning_rate": 2.445782142523394e-06, + "loss": 1.0315, + "step": 63285 + }, + { + "epoch": 4.68, + "learning_rate": 2.4455207923587796e-06, + "loss": 0.9448, + "step": 63286 + }, + { + "epoch": 4.68, + "learning_rate": 2.4452594542132314e-06, + "loss": 0.8607, + "step": 63287 + }, + { + "epoch": 4.68, + "learning_rate": 2.4449981280871693e-06, + "loss": 1.0046, + "step": 63288 + }, + { + "epoch": 4.68, + "learning_rate": 2.444736813981008e-06, + "loss": 0.8962, + "step": 63289 + }, + { + "epoch": 4.68, + "learning_rate": 2.444475511895161e-06, + "loss": 1.0445, + "step": 63290 + }, + { + "epoch": 4.68, + "learning_rate": 2.444214221830047e-06, + "loss": 0.9819, + "step": 63291 + }, + { + "epoch": 4.68, + "learning_rate": 2.4439529437860765e-06, + "loss": 0.9437, + "step": 63292 + }, + { + "epoch": 4.68, + "learning_rate": 2.443691677763672e-06, + "loss": 1.0715, + "step": 63293 + }, + { + "epoch": 4.68, + "learning_rate": 2.4434304237632454e-06, + "loss": 1.0268, + "step": 63294 + }, + { + "epoch": 4.68, + "learning_rate": 2.443169181785213e-06, + "loss": 0.9611, + "step": 63295 + }, + { + "epoch": 4.68, + "learning_rate": 2.442907951829988e-06, + "loss": 1.0392, + "step": 63296 + }, + { + "epoch": 4.68, + "learning_rate": 2.4426467338979897e-06, + "loss": 0.968, + "step": 63297 + }, + { + "epoch": 4.68, + "learning_rate": 2.442385527989637e-06, + "loss": 1.086, + "step": 63298 + }, + { + "epoch": 4.68, + "learning_rate": 2.4421243341053335e-06, + "loss": 1.0625, + "step": 63299 + }, + { + "epoch": 4.68, + "learning_rate": 2.441863152245506e-06, + "loss": 0.9602, + "step": 63300 + }, + { + "epoch": 4.68, + "learning_rate": 2.441601982410563e-06, + "loss": 1.0502, + "step": 63301 + }, + { + "epoch": 4.68, + "learning_rate": 2.4413408246009285e-06, + "loss": 1.0132, + "step": 63302 + }, + { + "epoch": 4.68, + "learning_rate": 2.4410796788170077e-06, + "loss": 1.0397, + "step": 63303 + }, + { + "epoch": 4.68, + "learning_rate": 2.4408185450592227e-06, + "loss": 0.9318, + "step": 63304 + }, + { + "epoch": 4.68, + "learning_rate": 2.440557423327985e-06, + "loss": 0.9186, + "step": 63305 + }, + { + "epoch": 4.68, + "learning_rate": 2.440296313623718e-06, + "loss": 0.9999, + "step": 63306 + }, + { + "epoch": 4.68, + "learning_rate": 2.440035215946829e-06, + "loss": 0.9467, + "step": 63307 + }, + { + "epoch": 4.68, + "learning_rate": 2.439774130297732e-06, + "loss": 0.9493, + "step": 63308 + }, + { + "epoch": 4.68, + "learning_rate": 2.43951305667685e-06, + "loss": 1.0258, + "step": 63309 + }, + { + "epoch": 4.68, + "learning_rate": 2.4392519950845937e-06, + "loss": 0.9945, + "step": 63310 + }, + { + "epoch": 4.68, + "learning_rate": 2.4389909455213788e-06, + "loss": 0.9054, + "step": 63311 + }, + { + "epoch": 4.68, + "learning_rate": 2.438729907987619e-06, + "loss": 1.0774, + "step": 63312 + }, + { + "epoch": 4.68, + "learning_rate": 2.4384688824837355e-06, + "loss": 1.0966, + "step": 63313 + }, + { + "epoch": 4.68, + "learning_rate": 2.4382078690101384e-06, + "loss": 1.0824, + "step": 63314 + }, + { + "epoch": 4.68, + "learning_rate": 2.437946867567246e-06, + "loss": 1.0598, + "step": 63315 + }, + { + "epoch": 4.68, + "learning_rate": 2.4376858781554667e-06, + "loss": 1.0432, + "step": 63316 + }, + { + "epoch": 4.68, + "learning_rate": 2.437424900775226e-06, + "loss": 1.0393, + "step": 63317 + }, + { + "epoch": 4.68, + "learning_rate": 2.437163935426933e-06, + "loss": 1.0376, + "step": 63318 + }, + { + "epoch": 4.68, + "learning_rate": 2.4369029821110044e-06, + "loss": 1.0272, + "step": 63319 + }, + { + "epoch": 4.68, + "learning_rate": 2.4366420408278546e-06, + "loss": 1.09, + "step": 63320 + }, + { + "epoch": 4.68, + "learning_rate": 2.436381111577897e-06, + "loss": 1.097, + "step": 63321 + }, + { + "epoch": 4.68, + "learning_rate": 2.436120194361551e-06, + "loss": 0.9795, + "step": 63322 + }, + { + "epoch": 4.68, + "learning_rate": 2.4358592891792286e-06, + "loss": 1.0241, + "step": 63323 + }, + { + "epoch": 4.68, + "learning_rate": 2.435598396031348e-06, + "loss": 1.059, + "step": 63324 + }, + { + "epoch": 4.68, + "learning_rate": 2.4353375149183177e-06, + "loss": 1.0895, + "step": 63325 + }, + { + "epoch": 4.68, + "learning_rate": 2.435076645840564e-06, + "loss": 0.9515, + "step": 63326 + }, + { + "epoch": 4.68, + "learning_rate": 2.4348157887984912e-06, + "loss": 1.0369, + "step": 63327 + }, + { + "epoch": 4.68, + "learning_rate": 2.4345549437925155e-06, + "loss": 0.9362, + "step": 63328 + }, + { + "epoch": 4.68, + "learning_rate": 2.434294110823058e-06, + "loss": 0.9624, + "step": 63329 + }, + { + "epoch": 4.68, + "learning_rate": 2.4340332898905305e-06, + "loss": 1.0035, + "step": 63330 + }, + { + "epoch": 4.68, + "learning_rate": 2.4337724809953467e-06, + "loss": 0.9146, + "step": 63331 + }, + { + "epoch": 4.68, + "learning_rate": 2.4335116841379204e-06, + "loss": 1.0316, + "step": 63332 + }, + { + "epoch": 4.68, + "learning_rate": 2.433250899318672e-06, + "loss": 0.8907, + "step": 63333 + }, + { + "epoch": 4.68, + "learning_rate": 2.432990126538013e-06, + "loss": 1.0851, + "step": 63334 + }, + { + "epoch": 4.68, + "learning_rate": 2.432729365796358e-06, + "loss": 1.0446, + "step": 63335 + }, + { + "epoch": 4.68, + "learning_rate": 2.4324686170941227e-06, + "loss": 1.0686, + "step": 63336 + }, + { + "epoch": 4.68, + "learning_rate": 2.4322078804317183e-06, + "loss": 0.9593, + "step": 63337 + }, + { + "epoch": 4.68, + "learning_rate": 2.4319471558095666e-06, + "loss": 0.9814, + "step": 63338 + }, + { + "epoch": 4.68, + "learning_rate": 2.4316864432280784e-06, + "loss": 1.0435, + "step": 63339 + }, + { + "epoch": 4.68, + "learning_rate": 2.4314257426876685e-06, + "loss": 1.0788, + "step": 63340 + }, + { + "epoch": 4.68, + "learning_rate": 2.4311650541887488e-06, + "loss": 1.0138, + "step": 63341 + }, + { + "epoch": 4.68, + "learning_rate": 2.43090437773174e-06, + "loss": 0.9906, + "step": 63342 + }, + { + "epoch": 4.68, + "learning_rate": 2.4306437133170547e-06, + "loss": 1.0348, + "step": 63343 + }, + { + "epoch": 4.68, + "learning_rate": 2.4303830609451075e-06, + "loss": 0.9308, + "step": 63344 + }, + { + "epoch": 4.68, + "learning_rate": 2.430122420616311e-06, + "loss": 1.0161, + "step": 63345 + }, + { + "epoch": 4.68, + "learning_rate": 2.4298617923310795e-06, + "loss": 1.081, + "step": 63346 + }, + { + "epoch": 4.68, + "learning_rate": 2.429601176089835e-06, + "loss": 1.0734, + "step": 63347 + }, + { + "epoch": 4.68, + "learning_rate": 2.4293405718929808e-06, + "loss": 0.9946, + "step": 63348 + }, + { + "epoch": 4.68, + "learning_rate": 2.4290799797409404e-06, + "loss": 1.0504, + "step": 63349 + }, + { + "epoch": 4.68, + "learning_rate": 2.4288193996341224e-06, + "loss": 0.9451, + "step": 63350 + }, + { + "epoch": 4.68, + "learning_rate": 2.4285588315729515e-06, + "loss": 1.0165, + "step": 63351 + }, + { + "epoch": 4.68, + "learning_rate": 2.428298275557828e-06, + "loss": 0.9004, + "step": 63352 + }, + { + "epoch": 4.68, + "learning_rate": 2.4280377315891767e-06, + "loss": 0.995, + "step": 63353 + }, + { + "epoch": 4.68, + "learning_rate": 2.4277771996674093e-06, + "loss": 1.0752, + "step": 63354 + }, + { + "epoch": 4.68, + "learning_rate": 2.4275166797929407e-06, + "loss": 0.972, + "step": 63355 + }, + { + "epoch": 4.68, + "learning_rate": 2.4272561719661838e-06, + "loss": 1.059, + "step": 63356 + }, + { + "epoch": 4.68, + "learning_rate": 2.426995676187551e-06, + "loss": 0.9965, + "step": 63357 + }, + { + "epoch": 4.68, + "learning_rate": 2.4267351924574623e-06, + "loss": 1.0567, + "step": 63358 + }, + { + "epoch": 4.68, + "learning_rate": 2.4264747207763304e-06, + "loss": 0.9976, + "step": 63359 + }, + { + "epoch": 4.68, + "learning_rate": 2.426214261144569e-06, + "loss": 0.9483, + "step": 63360 + }, + { + "epoch": 4.68, + "learning_rate": 2.4259538135625893e-06, + "loss": 0.9999, + "step": 63361 + }, + { + "epoch": 4.68, + "learning_rate": 2.425693378030811e-06, + "loss": 0.8995, + "step": 63362 + }, + { + "epoch": 4.68, + "learning_rate": 2.425432954549647e-06, + "loss": 0.9324, + "step": 63363 + }, + { + "epoch": 4.68, + "learning_rate": 2.4251725431195093e-06, + "loss": 0.997, + "step": 63364 + }, + { + "epoch": 4.68, + "learning_rate": 2.424912143740815e-06, + "loss": 0.9446, + "step": 63365 + }, + { + "epoch": 4.68, + "learning_rate": 2.424651756413974e-06, + "loss": 0.9205, + "step": 63366 + }, + { + "epoch": 4.68, + "learning_rate": 2.4243913811394093e-06, + "loss": 1.0424, + "step": 63367 + }, + { + "epoch": 4.68, + "learning_rate": 2.424131017917524e-06, + "loss": 0.9855, + "step": 63368 + }, + { + "epoch": 4.68, + "learning_rate": 2.423870666748741e-06, + "loss": 1.0579, + "step": 63369 + }, + { + "epoch": 4.68, + "learning_rate": 2.423610327633468e-06, + "loss": 1.0235, + "step": 63370 + }, + { + "epoch": 4.68, + "learning_rate": 2.423350000572128e-06, + "loss": 0.8668, + "step": 63371 + }, + { + "epoch": 4.68, + "learning_rate": 2.423089685565124e-06, + "loss": 1.1105, + "step": 63372 + }, + { + "epoch": 4.68, + "learning_rate": 2.42282938261288e-06, + "loss": 0.9901, + "step": 63373 + }, + { + "epoch": 4.68, + "learning_rate": 2.422569091715805e-06, + "loss": 1.0783, + "step": 63374 + }, + { + "epoch": 4.68, + "learning_rate": 2.4223088128743145e-06, + "loss": 0.8855, + "step": 63375 + }, + { + "epoch": 4.68, + "learning_rate": 2.422048546088822e-06, + "loss": 0.9633, + "step": 63376 + }, + { + "epoch": 4.68, + "learning_rate": 2.421788291359739e-06, + "loss": 0.9467, + "step": 63377 + }, + { + "epoch": 4.68, + "learning_rate": 2.421528048687486e-06, + "loss": 1.0248, + "step": 63378 + }, + { + "epoch": 4.68, + "learning_rate": 2.4212678180724725e-06, + "loss": 1.0425, + "step": 63379 + }, + { + "epoch": 4.68, + "learning_rate": 2.4210075995151137e-06, + "loss": 1.0931, + "step": 63380 + }, + { + "epoch": 4.68, + "learning_rate": 2.42074739301582e-06, + "loss": 0.9559, + "step": 63381 + }, + { + "epoch": 4.68, + "learning_rate": 2.4204871985750124e-06, + "loss": 0.8981, + "step": 63382 + }, + { + "epoch": 4.68, + "learning_rate": 2.420227016193101e-06, + "loss": 1.0084, + "step": 63383 + }, + { + "epoch": 4.68, + "learning_rate": 2.4199668458705004e-06, + "loss": 0.8921, + "step": 63384 + }, + { + "epoch": 4.68, + "learning_rate": 2.4197066876076234e-06, + "loss": 0.9023, + "step": 63385 + }, + { + "epoch": 4.68, + "learning_rate": 2.4194465414048816e-06, + "loss": 1.0836, + "step": 63386 + }, + { + "epoch": 4.68, + "learning_rate": 2.419186407262696e-06, + "loss": 0.9807, + "step": 63387 + }, + { + "epoch": 4.68, + "learning_rate": 2.4189262851814753e-06, + "loss": 0.9823, + "step": 63388 + }, + { + "epoch": 4.68, + "learning_rate": 2.4186661751616347e-06, + "loss": 0.8642, + "step": 63389 + }, + { + "epoch": 4.68, + "learning_rate": 2.4184060772035843e-06, + "loss": 1.0363, + "step": 63390 + }, + { + "epoch": 4.68, + "learning_rate": 2.4181459913077455e-06, + "loss": 1.0236, + "step": 63391 + }, + { + "epoch": 4.68, + "learning_rate": 2.4178859174745297e-06, + "loss": 1.0055, + "step": 63392 + }, + { + "epoch": 4.68, + "learning_rate": 2.417625855704344e-06, + "loss": 0.9488, + "step": 63393 + }, + { + "epoch": 4.68, + "learning_rate": 2.417365805997609e-06, + "loss": 0.9695, + "step": 63394 + }, + { + "epoch": 4.68, + "learning_rate": 2.417105768354734e-06, + "loss": 1.0632, + "step": 63395 + }, + { + "epoch": 4.68, + "learning_rate": 2.416845742776142e-06, + "loss": 0.9354, + "step": 63396 + }, + { + "epoch": 4.68, + "learning_rate": 2.416585729262233e-06, + "loss": 0.9987, + "step": 63397 + }, + { + "epoch": 4.68, + "learning_rate": 2.416325727813431e-06, + "loss": 0.973, + "step": 63398 + }, + { + "epoch": 4.68, + "learning_rate": 2.4160657384301433e-06, + "loss": 1.0771, + "step": 63399 + }, + { + "epoch": 4.68, + "learning_rate": 2.4158057611127916e-06, + "loss": 1.0562, + "step": 63400 + }, + { + "epoch": 4.68, + "learning_rate": 2.415545795861782e-06, + "loss": 1.0114, + "step": 63401 + }, + { + "epoch": 4.68, + "learning_rate": 2.415285842677528e-06, + "loss": 0.9726, + "step": 63402 + }, + { + "epoch": 4.68, + "learning_rate": 2.4150259015604482e-06, + "loss": 1.0419, + "step": 63403 + }, + { + "epoch": 4.68, + "learning_rate": 2.414765972510954e-06, + "loss": 0.9676, + "step": 63404 + }, + { + "epoch": 4.68, + "learning_rate": 2.4145060555294587e-06, + "loss": 1.0373, + "step": 63405 + }, + { + "epoch": 4.69, + "learning_rate": 2.4142461506163715e-06, + "loss": 0.8872, + "step": 63406 + }, + { + "epoch": 4.69, + "learning_rate": 2.413986257772114e-06, + "loss": 1.0335, + "step": 63407 + }, + { + "epoch": 4.69, + "learning_rate": 2.4137263769970965e-06, + "loss": 1.083, + "step": 63408 + }, + { + "epoch": 4.69, + "learning_rate": 2.4134665082917308e-06, + "loss": 0.992, + "step": 63409 + }, + { + "epoch": 4.69, + "learning_rate": 2.4132066516564325e-06, + "loss": 1.0459, + "step": 63410 + }, + { + "epoch": 4.69, + "learning_rate": 2.4129468070916094e-06, + "loss": 0.9919, + "step": 63411 + }, + { + "epoch": 4.69, + "learning_rate": 2.4126869745976865e-06, + "loss": 1.0496, + "step": 63412 + }, + { + "epoch": 4.69, + "learning_rate": 2.412427154175063e-06, + "loss": 0.9971, + "step": 63413 + }, + { + "epoch": 4.69, + "learning_rate": 2.4121673458241635e-06, + "loss": 1.0136, + "step": 63414 + }, + { + "epoch": 4.69, + "learning_rate": 2.411907549545394e-06, + "loss": 0.9949, + "step": 63415 + }, + { + "epoch": 4.69, + "learning_rate": 2.411647765339177e-06, + "loss": 0.9355, + "step": 63416 + }, + { + "epoch": 4.69, + "learning_rate": 2.4113879932059127e-06, + "loss": 0.9937, + "step": 63417 + }, + { + "epoch": 4.69, + "learning_rate": 2.411128233146025e-06, + "loss": 0.9818, + "step": 63418 + }, + { + "epoch": 4.69, + "learning_rate": 2.410868485159922e-06, + "loss": 0.9467, + "step": 63419 + }, + { + "epoch": 4.69, + "learning_rate": 2.4106087492480237e-06, + "loss": 1.0352, + "step": 63420 + }, + { + "epoch": 4.69, + "learning_rate": 2.4103490254107355e-06, + "loss": 0.991, + "step": 63421 + }, + { + "epoch": 4.69, + "learning_rate": 2.4100893136484695e-06, + "loss": 0.9315, + "step": 63422 + }, + { + "epoch": 4.69, + "learning_rate": 2.4098296139616475e-06, + "loss": 1.0163, + "step": 63423 + }, + { + "epoch": 4.69, + "learning_rate": 2.4095699263506765e-06, + "loss": 1.0802, + "step": 63424 + }, + { + "epoch": 4.69, + "learning_rate": 2.4093102508159715e-06, + "loss": 0.9899, + "step": 63425 + }, + { + "epoch": 4.69, + "learning_rate": 2.409050587357943e-06, + "loss": 1.0209, + "step": 63426 + }, + { + "epoch": 4.69, + "learning_rate": 2.408790935977008e-06, + "loss": 0.8931, + "step": 63427 + }, + { + "epoch": 4.69, + "learning_rate": 2.4085312966735796e-06, + "loss": 0.9314, + "step": 63428 + }, + { + "epoch": 4.69, + "learning_rate": 2.408271669448068e-06, + "loss": 0.9594, + "step": 63429 + }, + { + "epoch": 4.69, + "learning_rate": 2.4080120543008877e-06, + "loss": 0.9719, + "step": 63430 + }, + { + "epoch": 4.69, + "learning_rate": 2.407752451232449e-06, + "loss": 0.8501, + "step": 63431 + }, + { + "epoch": 4.69, + "learning_rate": 2.4074928602431715e-06, + "loss": 0.913, + "step": 63432 + }, + { + "epoch": 4.69, + "learning_rate": 2.4072332813334632e-06, + "loss": 0.9143, + "step": 63433 + }, + { + "epoch": 4.69, + "learning_rate": 2.4069737145037374e-06, + "loss": 0.9573, + "step": 63434 + }, + { + "epoch": 4.69, + "learning_rate": 2.4067141597544065e-06, + "loss": 1.0138, + "step": 63435 + }, + { + "epoch": 4.69, + "learning_rate": 2.406454617085886e-06, + "loss": 1.0425, + "step": 63436 + }, + { + "epoch": 4.69, + "learning_rate": 2.4061950864985883e-06, + "loss": 1.0551, + "step": 63437 + }, + { + "epoch": 4.69, + "learning_rate": 2.4059355679929264e-06, + "loss": 0.9738, + "step": 63438 + }, + { + "epoch": 4.69, + "learning_rate": 2.4056760615693107e-06, + "loss": 1.0372, + "step": 63439 + }, + { + "epoch": 4.69, + "learning_rate": 2.4054165672281538e-06, + "loss": 0.9926, + "step": 63440 + }, + { + "epoch": 4.69, + "learning_rate": 2.405157084969877e-06, + "loss": 1.0081, + "step": 63441 + }, + { + "epoch": 4.69, + "learning_rate": 2.40489761479488e-06, + "loss": 1.0261, + "step": 63442 + }, + { + "epoch": 4.69, + "learning_rate": 2.4046381567035847e-06, + "loss": 1.0384, + "step": 63443 + }, + { + "epoch": 4.69, + "learning_rate": 2.404378710696399e-06, + "loss": 1.0274, + "step": 63444 + }, + { + "epoch": 4.69, + "learning_rate": 2.404119276773743e-06, + "loss": 0.8956, + "step": 63445 + }, + { + "epoch": 4.69, + "learning_rate": 2.4038598549360204e-06, + "loss": 1.0261, + "step": 63446 + }, + { + "epoch": 4.69, + "learning_rate": 2.4036004451836503e-06, + "loss": 1.0541, + "step": 63447 + }, + { + "epoch": 4.69, + "learning_rate": 2.4033410475170427e-06, + "loss": 1.0167, + "step": 63448 + }, + { + "epoch": 4.69, + "learning_rate": 2.4030816619366115e-06, + "loss": 0.9869, + "step": 63449 + }, + { + "epoch": 4.69, + "learning_rate": 2.4028222884427675e-06, + "loss": 1.0323, + "step": 63450 + }, + { + "epoch": 4.69, + "learning_rate": 2.402562927035923e-06, + "loss": 1.0372, + "step": 63451 + }, + { + "epoch": 4.69, + "learning_rate": 2.402303577716495e-06, + "loss": 1.023, + "step": 63452 + }, + { + "epoch": 4.69, + "learning_rate": 2.402044240484892e-06, + "loss": 1.0342, + "step": 63453 + }, + { + "epoch": 4.69, + "learning_rate": 2.4017849153415297e-06, + "loss": 0.981, + "step": 63454 + }, + { + "epoch": 4.69, + "learning_rate": 2.401525602286815e-06, + "loss": 1.0329, + "step": 63455 + }, + { + "epoch": 4.69, + "learning_rate": 2.4012663013211668e-06, + "loss": 0.9893, + "step": 63456 + }, + { + "epoch": 4.69, + "learning_rate": 2.401007012444999e-06, + "loss": 0.9897, + "step": 63457 + }, + { + "epoch": 4.69, + "learning_rate": 2.400747735658714e-06, + "loss": 1.0539, + "step": 63458 + }, + { + "epoch": 4.69, + "learning_rate": 2.400488470962733e-06, + "loss": 0.9199, + "step": 63459 + }, + { + "epoch": 4.69, + "learning_rate": 2.400229218357464e-06, + "loss": 0.9728, + "step": 63460 + }, + { + "epoch": 4.69, + "learning_rate": 2.3999699778433273e-06, + "loss": 1.0148, + "step": 63461 + }, + { + "epoch": 4.69, + "learning_rate": 2.399710749420724e-06, + "loss": 0.9629, + "step": 63462 + }, + { + "epoch": 4.69, + "learning_rate": 2.3994515330900747e-06, + "loss": 0.9546, + "step": 63463 + }, + { + "epoch": 4.69, + "learning_rate": 2.3991923288517873e-06, + "loss": 0.9695, + "step": 63464 + }, + { + "epoch": 4.69, + "learning_rate": 2.3989331367062806e-06, + "loss": 0.9765, + "step": 63465 + }, + { + "epoch": 4.69, + "learning_rate": 2.39867395665396e-06, + "loss": 0.9201, + "step": 63466 + }, + { + "epoch": 4.69, + "learning_rate": 2.398414788695238e-06, + "loss": 0.94, + "step": 63467 + }, + { + "epoch": 4.69, + "learning_rate": 2.3981556328305312e-06, + "loss": 0.9383, + "step": 63468 + }, + { + "epoch": 4.69, + "learning_rate": 2.3978964890602505e-06, + "loss": 0.9472, + "step": 63469 + }, + { + "epoch": 4.69, + "learning_rate": 2.3976373573848088e-06, + "loss": 1.0378, + "step": 63470 + }, + { + "epoch": 4.69, + "learning_rate": 2.397378237804613e-06, + "loss": 0.9933, + "step": 63471 + }, + { + "epoch": 4.69, + "learning_rate": 2.3971191303200836e-06, + "loss": 1.009, + "step": 63472 + }, + { + "epoch": 4.69, + "learning_rate": 2.3968600349316286e-06, + "loss": 1.0529, + "step": 63473 + }, + { + "epoch": 4.69, + "learning_rate": 2.396600951639659e-06, + "loss": 0.9529, + "step": 63474 + }, + { + "epoch": 4.69, + "learning_rate": 2.396341880444587e-06, + "loss": 0.9276, + "step": 63475 + }, + { + "epoch": 4.69, + "learning_rate": 2.396082821346829e-06, + "loss": 0.8949, + "step": 63476 + }, + { + "epoch": 4.69, + "learning_rate": 2.395823774346795e-06, + "loss": 1.0268, + "step": 63477 + }, + { + "epoch": 4.69, + "learning_rate": 2.3955647394448957e-06, + "loss": 0.9774, + "step": 63478 + }, + { + "epoch": 4.69, + "learning_rate": 2.395305716641545e-06, + "loss": 1.11, + "step": 63479 + }, + { + "epoch": 4.69, + "learning_rate": 2.3950467059371507e-06, + "loss": 0.9773, + "step": 63480 + }, + { + "epoch": 4.69, + "learning_rate": 2.39478770733213e-06, + "loss": 0.9506, + "step": 63481 + }, + { + "epoch": 4.69, + "learning_rate": 2.3945287208268953e-06, + "loss": 1.0271, + "step": 63482 + }, + { + "epoch": 4.69, + "learning_rate": 2.3942697464218557e-06, + "loss": 0.9518, + "step": 63483 + }, + { + "epoch": 4.69, + "learning_rate": 2.394010784117421e-06, + "loss": 0.9853, + "step": 63484 + }, + { + "epoch": 4.69, + "learning_rate": 2.3937518339140086e-06, + "loss": 1.0323, + "step": 63485 + }, + { + "epoch": 4.69, + "learning_rate": 2.393492895812033e-06, + "loss": 0.9475, + "step": 63486 + }, + { + "epoch": 4.69, + "learning_rate": 2.3932339698118935e-06, + "loss": 1.0014, + "step": 63487 + }, + { + "epoch": 4.69, + "learning_rate": 2.3929750559140142e-06, + "loss": 1.065, + "step": 63488 + }, + { + "epoch": 4.69, + "learning_rate": 2.3927161541188026e-06, + "loss": 1.0711, + "step": 63489 + }, + { + "epoch": 4.69, + "learning_rate": 2.3924572644266695e-06, + "loss": 0.9496, + "step": 63490 + }, + { + "epoch": 4.69, + "learning_rate": 2.3921983868380273e-06, + "loss": 0.8177, + "step": 63491 + }, + { + "epoch": 4.69, + "learning_rate": 2.3919395213532903e-06, + "loss": 0.9469, + "step": 63492 + }, + { + "epoch": 4.69, + "learning_rate": 2.3916806679728675e-06, + "loss": 1.0513, + "step": 63493 + }, + { + "epoch": 4.69, + "learning_rate": 2.391421826697173e-06, + "loss": 1.0353, + "step": 63494 + }, + { + "epoch": 4.69, + "learning_rate": 2.391162997526618e-06, + "loss": 1.0275, + "step": 63495 + }, + { + "epoch": 4.69, + "learning_rate": 2.3909041804616107e-06, + "loss": 1.0218, + "step": 63496 + }, + { + "epoch": 4.69, + "learning_rate": 2.390645375502569e-06, + "loss": 1.0578, + "step": 63497 + }, + { + "epoch": 4.69, + "learning_rate": 2.390386582649902e-06, + "loss": 0.9632, + "step": 63498 + }, + { + "epoch": 4.69, + "learning_rate": 2.3901278019040198e-06, + "loss": 0.9704, + "step": 63499 + }, + { + "epoch": 4.69, + "learning_rate": 2.389869033265333e-06, + "loss": 0.942, + "step": 63500 + }, + { + "epoch": 4.69, + "learning_rate": 2.389610276734259e-06, + "loss": 1.0755, + "step": 63501 + }, + { + "epoch": 4.69, + "learning_rate": 2.389351532311206e-06, + "loss": 1.0671, + "step": 63502 + }, + { + "epoch": 4.69, + "learning_rate": 2.389092799996585e-06, + "loss": 0.8491, + "step": 63503 + }, + { + "epoch": 4.69, + "learning_rate": 2.3888340797908094e-06, + "loss": 0.982, + "step": 63504 + }, + { + "epoch": 4.69, + "learning_rate": 2.3885753716942862e-06, + "loss": 0.9217, + "step": 63505 + }, + { + "epoch": 4.69, + "learning_rate": 2.3883166757074374e-06, + "loss": 0.9685, + "step": 63506 + }, + { + "epoch": 4.69, + "learning_rate": 2.3880579918306613e-06, + "loss": 1.0322, + "step": 63507 + }, + { + "epoch": 4.69, + "learning_rate": 2.38779932006438e-06, + "loss": 0.9135, + "step": 63508 + }, + { + "epoch": 4.69, + "learning_rate": 2.3875406604089968e-06, + "loss": 0.9791, + "step": 63509 + }, + { + "epoch": 4.69, + "learning_rate": 2.3872820128649345e-06, + "loss": 1.012, + "step": 63510 + }, + { + "epoch": 4.69, + "learning_rate": 2.3870233774325913e-06, + "loss": 1.0324, + "step": 63511 + }, + { + "epoch": 4.69, + "learning_rate": 2.3867647541123873e-06, + "loss": 0.9798, + "step": 63512 + }, + { + "epoch": 4.69, + "learning_rate": 2.386506142904732e-06, + "loss": 0.9847, + "step": 63513 + }, + { + "epoch": 4.69, + "learning_rate": 2.3862475438100374e-06, + "loss": 0.9939, + "step": 63514 + }, + { + "epoch": 4.69, + "learning_rate": 2.385988956828713e-06, + "loss": 1.0551, + "step": 63515 + }, + { + "epoch": 4.69, + "learning_rate": 2.385730381961169e-06, + "loss": 0.9633, + "step": 63516 + }, + { + "epoch": 4.69, + "learning_rate": 2.3854718192078206e-06, + "loss": 0.9698, + "step": 63517 + }, + { + "epoch": 4.69, + "learning_rate": 2.3852132685690786e-06, + "loss": 0.93, + "step": 63518 + }, + { + "epoch": 4.69, + "learning_rate": 2.3849547300453535e-06, + "loss": 0.865, + "step": 63519 + }, + { + "epoch": 4.69, + "learning_rate": 2.3846962036370536e-06, + "loss": 0.9835, + "step": 63520 + }, + { + "epoch": 4.69, + "learning_rate": 2.3844376893445953e-06, + "loss": 1.0009, + "step": 63521 + }, + { + "epoch": 4.69, + "learning_rate": 2.3841791871683884e-06, + "loss": 0.9952, + "step": 63522 + }, + { + "epoch": 4.69, + "learning_rate": 2.3839206971088435e-06, + "loss": 0.9293, + "step": 63523 + }, + { + "epoch": 4.69, + "learning_rate": 2.3836622191663704e-06, + "loss": 0.9573, + "step": 63524 + }, + { + "epoch": 4.69, + "learning_rate": 2.38340375334138e-06, + "loss": 0.9975, + "step": 63525 + }, + { + "epoch": 4.69, + "learning_rate": 2.3831452996342885e-06, + "loss": 0.9543, + "step": 63526 + }, + { + "epoch": 4.69, + "learning_rate": 2.3828868580455035e-06, + "loss": 0.9489, + "step": 63527 + }, + { + "epoch": 4.69, + "learning_rate": 2.3826284285754366e-06, + "loss": 0.9943, + "step": 63528 + }, + { + "epoch": 4.69, + "learning_rate": 2.382370011224495e-06, + "loss": 0.961, + "step": 63529 + }, + { + "epoch": 4.69, + "learning_rate": 2.3821116059930984e-06, + "loss": 0.9305, + "step": 63530 + }, + { + "epoch": 4.69, + "learning_rate": 2.3818532128816517e-06, + "loss": 1.027, + "step": 63531 + }, + { + "epoch": 4.69, + "learning_rate": 2.381594831890568e-06, + "loss": 1.0086, + "step": 63532 + }, + { + "epoch": 4.69, + "learning_rate": 2.3813364630202583e-06, + "loss": 0.9251, + "step": 63533 + }, + { + "epoch": 4.69, + "learning_rate": 2.381078106271133e-06, + "loss": 1.0448, + "step": 63534 + }, + { + "epoch": 4.69, + "learning_rate": 2.3808197616436033e-06, + "loss": 1.0331, + "step": 63535 + }, + { + "epoch": 4.69, + "learning_rate": 2.3805614291380772e-06, + "loss": 1.0317, + "step": 63536 + }, + { + "epoch": 4.69, + "learning_rate": 2.380303108754972e-06, + "loss": 1.0119, + "step": 63537 + }, + { + "epoch": 4.69, + "learning_rate": 2.3800448004946964e-06, + "loss": 1.1063, + "step": 63538 + }, + { + "epoch": 4.69, + "learning_rate": 2.3797865043576595e-06, + "loss": 0.9839, + "step": 63539 + }, + { + "epoch": 4.69, + "learning_rate": 2.3795282203442703e-06, + "loss": 1.0441, + "step": 63540 + }, + { + "epoch": 4.7, + "learning_rate": 2.3792699484549466e-06, + "loss": 0.9537, + "step": 63541 + }, + { + "epoch": 4.7, + "learning_rate": 2.379011688690094e-06, + "loss": 0.955, + "step": 63542 + }, + { + "epoch": 4.7, + "learning_rate": 2.3787534410501257e-06, + "loss": 1.0271, + "step": 63543 + }, + { + "epoch": 4.7, + "learning_rate": 2.3784952055354515e-06, + "loss": 1.0836, + "step": 63544 + }, + { + "epoch": 4.7, + "learning_rate": 2.378236982146479e-06, + "loss": 1.0102, + "step": 63545 + }, + { + "epoch": 4.7, + "learning_rate": 2.3779787708836254e-06, + "loss": 0.9185, + "step": 63546 + }, + { + "epoch": 4.7, + "learning_rate": 2.3777205717472994e-06, + "loss": 1.0369, + "step": 63547 + }, + { + "epoch": 4.7, + "learning_rate": 2.3774623847379097e-06, + "loss": 0.8419, + "step": 63548 + }, + { + "epoch": 4.7, + "learning_rate": 2.377204209855867e-06, + "loss": 1.0323, + "step": 63549 + }, + { + "epoch": 4.7, + "learning_rate": 2.376946047101585e-06, + "loss": 1.0343, + "step": 63550 + }, + { + "epoch": 4.7, + "learning_rate": 2.3766878964754758e-06, + "loss": 0.9845, + "step": 63551 + }, + { + "epoch": 4.7, + "learning_rate": 2.3764297579779407e-06, + "loss": 0.9884, + "step": 63552 + }, + { + "epoch": 4.7, + "learning_rate": 2.3761716316094007e-06, + "loss": 0.9764, + "step": 63553 + }, + { + "epoch": 4.7, + "learning_rate": 2.3759135173702595e-06, + "loss": 1.1307, + "step": 63554 + }, + { + "epoch": 4.7, + "learning_rate": 2.375655415260936e-06, + "loss": 1.0204, + "step": 63555 + }, + { + "epoch": 4.7, + "learning_rate": 2.375397325281831e-06, + "loss": 1.0236, + "step": 63556 + }, + { + "epoch": 4.7, + "learning_rate": 2.3751392474333612e-06, + "loss": 1.0127, + "step": 63557 + }, + { + "epoch": 4.7, + "learning_rate": 2.3748811817159336e-06, + "loss": 0.987, + "step": 63558 + }, + { + "epoch": 4.7, + "learning_rate": 2.374623128129967e-06, + "loss": 0.9506, + "step": 63559 + }, + { + "epoch": 4.7, + "learning_rate": 2.3743650866758627e-06, + "loss": 0.9905, + "step": 63560 + }, + { + "epoch": 4.7, + "learning_rate": 2.3741070573540313e-06, + "loss": 1.085, + "step": 63561 + }, + { + "epoch": 4.7, + "learning_rate": 2.3738490401648894e-06, + "loss": 0.9726, + "step": 63562 + }, + { + "epoch": 4.7, + "learning_rate": 2.3735910351088452e-06, + "loss": 0.8894, + "step": 63563 + }, + { + "epoch": 4.7, + "learning_rate": 2.373333042186308e-06, + "loss": 1.2054, + "step": 63564 + }, + { + "epoch": 4.7, + "learning_rate": 2.373075061397686e-06, + "loss": 1.0267, + "step": 63565 + }, + { + "epoch": 4.7, + "learning_rate": 2.3728170927433956e-06, + "loss": 0.9177, + "step": 63566 + }, + { + "epoch": 4.7, + "learning_rate": 2.3725591362238432e-06, + "loss": 0.9863, + "step": 63567 + }, + { + "epoch": 4.7, + "learning_rate": 2.3723011918394413e-06, + "loss": 1.0947, + "step": 63568 + }, + { + "epoch": 4.7, + "learning_rate": 2.3720432595905986e-06, + "loss": 0.9805, + "step": 63569 + }, + { + "epoch": 4.7, + "learning_rate": 2.3717853394777225e-06, + "loss": 0.9755, + "step": 63570 + }, + { + "epoch": 4.7, + "learning_rate": 2.3715274315012306e-06, + "loss": 0.9675, + "step": 63571 + }, + { + "epoch": 4.7, + "learning_rate": 2.371269535661529e-06, + "loss": 0.9323, + "step": 63572 + }, + { + "epoch": 4.7, + "learning_rate": 2.3710116519590288e-06, + "loss": 0.9781, + "step": 63573 + }, + { + "epoch": 4.7, + "learning_rate": 2.370753780394136e-06, + "loss": 1.0549, + "step": 63574 + }, + { + "epoch": 4.7, + "learning_rate": 2.3704959209672685e-06, + "loss": 0.9553, + "step": 63575 + }, + { + "epoch": 4.7, + "learning_rate": 2.3702380736788322e-06, + "loss": 1.0448, + "step": 63576 + }, + { + "epoch": 4.7, + "learning_rate": 2.369980238529238e-06, + "loss": 0.9234, + "step": 63577 + }, + { + "epoch": 4.7, + "learning_rate": 2.369722415518897e-06, + "loss": 0.9732, + "step": 63578 + }, + { + "epoch": 4.7, + "learning_rate": 2.3694646046482185e-06, + "loss": 1.0553, + "step": 63579 + }, + { + "epoch": 4.7, + "learning_rate": 2.369206805917611e-06, + "loss": 0.9175, + "step": 63580 + }, + { + "epoch": 4.7, + "learning_rate": 2.3689490193274845e-06, + "loss": 1.0074, + "step": 63581 + }, + { + "epoch": 4.7, + "learning_rate": 2.368691244878254e-06, + "loss": 0.9735, + "step": 63582 + }, + { + "epoch": 4.7, + "learning_rate": 2.368433482570326e-06, + "loss": 0.9354, + "step": 63583 + }, + { + "epoch": 4.7, + "learning_rate": 2.3681757324041123e-06, + "loss": 0.9895, + "step": 63584 + }, + { + "epoch": 4.7, + "learning_rate": 2.367917994380018e-06, + "loss": 0.934, + "step": 63585 + }, + { + "epoch": 4.7, + "learning_rate": 2.3676602684984595e-06, + "loss": 1.0122, + "step": 63586 + }, + { + "epoch": 4.7, + "learning_rate": 2.3674025547598446e-06, + "loss": 0.9173, + "step": 63587 + }, + { + "epoch": 4.7, + "learning_rate": 2.3671448531645837e-06, + "loss": 1.0628, + "step": 63588 + }, + { + "epoch": 4.7, + "learning_rate": 2.366887163713085e-06, + "loss": 0.9868, + "step": 63589 + }, + { + "epoch": 4.7, + "learning_rate": 2.3666294864057584e-06, + "loss": 0.9414, + "step": 63590 + }, + { + "epoch": 4.7, + "learning_rate": 2.3663718212430164e-06, + "loss": 0.9668, + "step": 63591 + }, + { + "epoch": 4.7, + "learning_rate": 2.3661141682252674e-06, + "loss": 0.9112, + "step": 63592 + }, + { + "epoch": 4.7, + "learning_rate": 2.3658565273529222e-06, + "loss": 0.9768, + "step": 63593 + }, + { + "epoch": 4.7, + "learning_rate": 2.365598898626388e-06, + "loss": 1.0053, + "step": 63594 + }, + { + "epoch": 4.7, + "learning_rate": 2.3653412820460787e-06, + "loss": 0.7968, + "step": 63595 + }, + { + "epoch": 4.7, + "learning_rate": 2.365083677612402e-06, + "loss": 1.0148, + "step": 63596 + }, + { + "epoch": 4.7, + "learning_rate": 2.3648260853257687e-06, + "loss": 0.9245, + "step": 63597 + }, + { + "epoch": 4.7, + "learning_rate": 2.3645685051865876e-06, + "loss": 1.0008, + "step": 63598 + }, + { + "epoch": 4.7, + "learning_rate": 2.3643109371952645e-06, + "loss": 0.9571, + "step": 63599 + }, + { + "epoch": 4.7, + "learning_rate": 2.3640533813522206e-06, + "loss": 1.0018, + "step": 63600 + }, + { + "epoch": 4.7, + "learning_rate": 2.363795837657853e-06, + "loss": 0.9822, + "step": 63601 + }, + { + "epoch": 4.7, + "learning_rate": 2.3635383061125793e-06, + "loss": 1.0722, + "step": 63602 + }, + { + "epoch": 4.7, + "learning_rate": 2.363280786716804e-06, + "loss": 0.9225, + "step": 63603 + }, + { + "epoch": 4.7, + "learning_rate": 2.363023279470945e-06, + "loss": 1.1239, + "step": 63604 + }, + { + "epoch": 4.7, + "learning_rate": 2.3627657843754016e-06, + "loss": 0.9685, + "step": 63605 + }, + { + "epoch": 4.7, + "learning_rate": 2.3625083014305915e-06, + "loss": 1.0543, + "step": 63606 + }, + { + "epoch": 4.7, + "learning_rate": 2.3622508306369217e-06, + "loss": 0.8895, + "step": 63607 + }, + { + "epoch": 4.7, + "learning_rate": 2.3619933719948006e-06, + "loss": 1.0358, + "step": 63608 + }, + { + "epoch": 4.7, + "learning_rate": 2.361735925504639e-06, + "loss": 1.0093, + "step": 63609 + }, + { + "epoch": 4.7, + "learning_rate": 2.361478491166844e-06, + "loss": 1.052, + "step": 63610 + }, + { + "epoch": 4.7, + "learning_rate": 2.3612210689818303e-06, + "loss": 1.0679, + "step": 63611 + }, + { + "epoch": 4.7, + "learning_rate": 2.3609636589500042e-06, + "loss": 1.0018, + "step": 63612 + }, + { + "epoch": 4.7, + "learning_rate": 2.3607062610717757e-06, + "loss": 1.0758, + "step": 63613 + }, + { + "epoch": 4.7, + "learning_rate": 2.3604488753475506e-06, + "loss": 1.0625, + "step": 63614 + }, + { + "epoch": 4.7, + "learning_rate": 2.3601915017777455e-06, + "loss": 1.1551, + "step": 63615 + }, + { + "epoch": 4.7, + "learning_rate": 2.3599341403627675e-06, + "loss": 1.033, + "step": 63616 + }, + { + "epoch": 4.7, + "learning_rate": 2.359676791103024e-06, + "loss": 0.9968, + "step": 63617 + }, + { + "epoch": 4.7, + "learning_rate": 2.3594194539989246e-06, + "loss": 1.1012, + "step": 63618 + }, + { + "epoch": 4.7, + "learning_rate": 2.3591621290508783e-06, + "loss": 1.0246, + "step": 63619 + }, + { + "epoch": 4.7, + "learning_rate": 2.3589048162592976e-06, + "loss": 1.0269, + "step": 63620 + }, + { + "epoch": 4.7, + "learning_rate": 2.35864751562459e-06, + "loss": 0.8817, + "step": 63621 + }, + { + "epoch": 4.7, + "learning_rate": 2.3583902271471647e-06, + "loss": 1.0203, + "step": 63622 + }, + { + "epoch": 4.7, + "learning_rate": 2.3581329508274286e-06, + "loss": 0.999, + "step": 63623 + }, + { + "epoch": 4.7, + "learning_rate": 2.3578756866658005e-06, + "loss": 1.0396, + "step": 63624 + }, + { + "epoch": 4.7, + "learning_rate": 2.3576184346626786e-06, + "loss": 0.9519, + "step": 63625 + }, + { + "epoch": 4.7, + "learning_rate": 2.3573611948184737e-06, + "loss": 0.9789, + "step": 63626 + }, + { + "epoch": 4.7, + "learning_rate": 2.3571039671336003e-06, + "loss": 0.9437, + "step": 63627 + }, + { + "epoch": 4.7, + "learning_rate": 2.356846751608466e-06, + "loss": 0.9233, + "step": 63628 + }, + { + "epoch": 4.7, + "learning_rate": 2.3565895482434786e-06, + "loss": 1.0752, + "step": 63629 + }, + { + "epoch": 4.7, + "learning_rate": 2.3563323570390463e-06, + "loss": 0.9416, + "step": 63630 + }, + { + "epoch": 4.7, + "learning_rate": 2.3560751779955816e-06, + "loss": 1.0087, + "step": 63631 + }, + { + "epoch": 4.7, + "learning_rate": 2.355818011113492e-06, + "loss": 1.0617, + "step": 63632 + }, + { + "epoch": 4.7, + "learning_rate": 2.3555608563931866e-06, + "loss": 1.0621, + "step": 63633 + }, + { + "epoch": 4.7, + "learning_rate": 2.355303713835071e-06, + "loss": 1.0066, + "step": 63634 + }, + { + "epoch": 4.7, + "learning_rate": 2.355046583439562e-06, + "loss": 1.1222, + "step": 63635 + }, + { + "epoch": 4.7, + "learning_rate": 2.3547894652070637e-06, + "loss": 1.0431, + "step": 63636 + }, + { + "epoch": 4.7, + "learning_rate": 2.3545323591379853e-06, + "loss": 0.9614, + "step": 63637 + }, + { + "epoch": 4.7, + "learning_rate": 2.3542752652327384e-06, + "loss": 0.9345, + "step": 63638 + }, + { + "epoch": 4.7, + "learning_rate": 2.3540181834917253e-06, + "loss": 0.9664, + "step": 63639 + }, + { + "epoch": 4.7, + "learning_rate": 2.3537611139153636e-06, + "loss": 0.9185, + "step": 63640 + }, + { + "epoch": 4.7, + "learning_rate": 2.353504056504059e-06, + "loss": 0.9962, + "step": 63641 + }, + { + "epoch": 4.7, + "learning_rate": 2.3532470112582194e-06, + "loss": 1.0394, + "step": 63642 + }, + { + "epoch": 4.7, + "learning_rate": 2.3529899781782518e-06, + "loss": 0.9041, + "step": 63643 + }, + { + "epoch": 4.7, + "learning_rate": 2.3527329572645707e-06, + "loss": 0.9665, + "step": 63644 + }, + { + "epoch": 4.7, + "learning_rate": 2.3524759485175853e-06, + "loss": 0.9021, + "step": 63645 + }, + { + "epoch": 4.7, + "learning_rate": 2.3522189519376946e-06, + "loss": 0.9341, + "step": 63646 + }, + { + "epoch": 4.7, + "learning_rate": 2.351961967525318e-06, + "loss": 1.0602, + "step": 63647 + }, + { + "epoch": 4.7, + "learning_rate": 2.351704995280857e-06, + "loss": 0.9128, + "step": 63648 + }, + { + "epoch": 4.7, + "learning_rate": 2.35144803520473e-06, + "loss": 1.0451, + "step": 63649 + }, + { + "epoch": 4.7, + "learning_rate": 2.351191087297333e-06, + "loss": 0.9571, + "step": 63650 + }, + { + "epoch": 4.7, + "learning_rate": 2.3509341515590855e-06, + "loss": 0.9748, + "step": 63651 + }, + { + "epoch": 4.7, + "learning_rate": 2.350677227990389e-06, + "loss": 0.9923, + "step": 63652 + }, + { + "epoch": 4.7, + "learning_rate": 2.3504203165916605e-06, + "loss": 0.9305, + "step": 63653 + }, + { + "epoch": 4.7, + "learning_rate": 2.3501634173633027e-06, + "loss": 1.0504, + "step": 63654 + }, + { + "epoch": 4.7, + "learning_rate": 2.349906530305721e-06, + "loss": 0.8729, + "step": 63655 + }, + { + "epoch": 4.7, + "learning_rate": 2.3496496554193327e-06, + "loss": 1.0544, + "step": 63656 + }, + { + "epoch": 4.7, + "learning_rate": 2.3493927927045425e-06, + "loss": 0.9593, + "step": 63657 + }, + { + "epoch": 4.7, + "learning_rate": 2.3491359421617577e-06, + "loss": 0.8733, + "step": 63658 + }, + { + "epoch": 4.7, + "learning_rate": 2.348879103791386e-06, + "loss": 0.8308, + "step": 63659 + }, + { + "epoch": 4.7, + "learning_rate": 2.3486222775938405e-06, + "loss": 1.0607, + "step": 63660 + }, + { + "epoch": 4.7, + "learning_rate": 2.3483654635695276e-06, + "loss": 1.1083, + "step": 63661 + }, + { + "epoch": 4.7, + "learning_rate": 2.3481086617188565e-06, + "loss": 1.0567, + "step": 63662 + }, + { + "epoch": 4.7, + "learning_rate": 2.347851872042234e-06, + "loss": 1.0919, + "step": 63663 + }, + { + "epoch": 4.7, + "learning_rate": 2.3475950945400673e-06, + "loss": 0.9869, + "step": 63664 + }, + { + "epoch": 4.7, + "learning_rate": 2.3473383292127704e-06, + "loss": 1.0277, + "step": 63665 + }, + { + "epoch": 4.7, + "learning_rate": 2.3470815760607492e-06, + "loss": 0.9493, + "step": 63666 + }, + { + "epoch": 4.7, + "learning_rate": 2.3468248350844103e-06, + "loss": 1.1036, + "step": 63667 + }, + { + "epoch": 4.7, + "learning_rate": 2.3465681062841616e-06, + "loss": 0.977, + "step": 63668 + }, + { + "epoch": 4.7, + "learning_rate": 2.3463113896604193e-06, + "loss": 0.9163, + "step": 63669 + }, + { + "epoch": 4.7, + "learning_rate": 2.3460546852135803e-06, + "loss": 1.1627, + "step": 63670 + }, + { + "epoch": 4.7, + "learning_rate": 2.345797992944061e-06, + "loss": 1.0089, + "step": 63671 + }, + { + "epoch": 4.7, + "learning_rate": 2.3455413128522686e-06, + "loss": 0.9512, + "step": 63672 + }, + { + "epoch": 4.7, + "learning_rate": 2.3452846449386103e-06, + "loss": 0.8848, + "step": 63673 + }, + { + "epoch": 4.7, + "learning_rate": 2.3450279892034944e-06, + "loss": 1.0269, + "step": 63674 + }, + { + "epoch": 4.7, + "learning_rate": 2.3447713456473264e-06, + "loss": 0.828, + "step": 63675 + }, + { + "epoch": 4.7, + "learning_rate": 2.3445147142705203e-06, + "loss": 1.0087, + "step": 63676 + }, + { + "epoch": 4.71, + "learning_rate": 2.3442580950734817e-06, + "loss": 1.0195, + "step": 63677 + }, + { + "epoch": 4.71, + "learning_rate": 2.3440014880566197e-06, + "loss": 0.9744, + "step": 63678 + }, + { + "epoch": 4.71, + "learning_rate": 2.343744893220339e-06, + "loss": 0.9252, + "step": 63679 + }, + { + "epoch": 4.71, + "learning_rate": 2.343488310565053e-06, + "loss": 1.0168, + "step": 63680 + }, + { + "epoch": 4.71, + "learning_rate": 2.3432317400911676e-06, + "loss": 1.0076, + "step": 63681 + }, + { + "epoch": 4.71, + "learning_rate": 2.342975181799091e-06, + "loss": 0.9022, + "step": 63682 + }, + { + "epoch": 4.71, + "learning_rate": 2.3427186356892307e-06, + "loss": 1.1575, + "step": 63683 + }, + { + "epoch": 4.71, + "learning_rate": 2.342462101761993e-06, + "loss": 1.1612, + "step": 63684 + }, + { + "epoch": 4.71, + "learning_rate": 2.3422055800177925e-06, + "loss": 0.9882, + "step": 63685 + }, + { + "epoch": 4.71, + "learning_rate": 2.3419490704570313e-06, + "loss": 0.9406, + "step": 63686 + }, + { + "epoch": 4.71, + "learning_rate": 2.341692573080121e-06, + "loss": 0.9956, + "step": 63687 + }, + { + "epoch": 4.71, + "learning_rate": 2.3414360878874655e-06, + "loss": 1.0099, + "step": 63688 + }, + { + "epoch": 4.71, + "learning_rate": 2.3411796148794774e-06, + "loss": 0.9942, + "step": 63689 + }, + { + "epoch": 4.71, + "learning_rate": 2.3409231540565634e-06, + "loss": 0.9697, + "step": 63690 + }, + { + "epoch": 4.71, + "learning_rate": 2.3406667054191324e-06, + "loss": 1.0747, + "step": 63691 + }, + { + "epoch": 4.71, + "learning_rate": 2.34041026896759e-06, + "loss": 1.0649, + "step": 63692 + }, + { + "epoch": 4.71, + "learning_rate": 2.340153844702342e-06, + "loss": 1.0529, + "step": 63693 + }, + { + "epoch": 4.71, + "learning_rate": 2.3398974326238057e-06, + "loss": 0.9395, + "step": 63694 + }, + { + "epoch": 4.71, + "learning_rate": 2.339641032732377e-06, + "loss": 0.9964, + "step": 63695 + }, + { + "epoch": 4.71, + "learning_rate": 2.339384645028473e-06, + "loss": 1.0031, + "step": 63696 + }, + { + "epoch": 4.71, + "learning_rate": 2.339128269512496e-06, + "loss": 0.988, + "step": 63697 + }, + { + "epoch": 4.71, + "learning_rate": 2.338871906184862e-06, + "loss": 1.0044, + "step": 63698 + }, + { + "epoch": 4.71, + "learning_rate": 2.3386155550459665e-06, + "loss": 1.0254, + "step": 63699 + }, + { + "epoch": 4.71, + "learning_rate": 2.338359216096228e-06, + "loss": 0.944, + "step": 63700 + }, + { + "epoch": 4.71, + "learning_rate": 2.338102889336049e-06, + "loss": 1.0123, + "step": 63701 + }, + { + "epoch": 4.71, + "learning_rate": 2.33784657476584e-06, + "loss": 1.0914, + "step": 63702 + }, + { + "epoch": 4.71, + "learning_rate": 2.3375902723860066e-06, + "loss": 0.9886, + "step": 63703 + }, + { + "epoch": 4.71, + "learning_rate": 2.3373339821969554e-06, + "loss": 1.0739, + "step": 63704 + }, + { + "epoch": 4.71, + "learning_rate": 2.337077704199099e-06, + "loss": 0.8984, + "step": 63705 + }, + { + "epoch": 4.71, + "learning_rate": 2.336821438392841e-06, + "loss": 0.9255, + "step": 63706 + }, + { + "epoch": 4.71, + "learning_rate": 2.336565184778592e-06, + "loss": 0.9319, + "step": 63707 + }, + { + "epoch": 4.71, + "learning_rate": 2.336308943356754e-06, + "loss": 1.0206, + "step": 63708 + }, + { + "epoch": 4.71, + "learning_rate": 2.336052714127742e-06, + "loss": 0.8448, + "step": 63709 + }, + { + "epoch": 4.71, + "learning_rate": 2.335796497091961e-06, + "loss": 0.999, + "step": 63710 + }, + { + "epoch": 4.71, + "learning_rate": 2.3355402922498184e-06, + "loss": 1.0145, + "step": 63711 + }, + { + "epoch": 4.71, + "learning_rate": 2.335284099601721e-06, + "loss": 1.0481, + "step": 63712 + }, + { + "epoch": 4.71, + "learning_rate": 2.335027919148074e-06, + "loss": 1.0485, + "step": 63713 + }, + { + "epoch": 4.71, + "learning_rate": 2.3347717508892943e-06, + "loss": 1.0492, + "step": 63714 + }, + { + "epoch": 4.71, + "learning_rate": 2.334515594825777e-06, + "loss": 0.9762, + "step": 63715 + }, + { + "epoch": 4.71, + "learning_rate": 2.3342594509579387e-06, + "loss": 1.0056, + "step": 63716 + }, + { + "epoch": 4.71, + "learning_rate": 2.33400331928618e-06, + "loss": 0.9346, + "step": 63717 + }, + { + "epoch": 4.71, + "learning_rate": 2.3337471998109197e-06, + "loss": 1.0612, + "step": 63718 + }, + { + "epoch": 4.71, + "learning_rate": 2.333491092532554e-06, + "loss": 1.084, + "step": 63719 + }, + { + "epoch": 4.71, + "learning_rate": 2.333234997451491e-06, + "loss": 1.034, + "step": 63720 + }, + { + "epoch": 4.71, + "learning_rate": 2.3329789145681447e-06, + "loss": 1.0166, + "step": 63721 + }, + { + "epoch": 4.71, + "learning_rate": 2.3327228438829197e-06, + "loss": 0.9953, + "step": 63722 + }, + { + "epoch": 4.71, + "learning_rate": 2.3324667853962236e-06, + "loss": 0.8866, + "step": 63723 + }, + { + "epoch": 4.71, + "learning_rate": 2.332210739108459e-06, + "loss": 0.9187, + "step": 63724 + }, + { + "epoch": 4.71, + "learning_rate": 2.331954705020041e-06, + "loss": 1.0515, + "step": 63725 + }, + { + "epoch": 4.71, + "learning_rate": 2.3316986831313727e-06, + "loss": 0.9712, + "step": 63726 + }, + { + "epoch": 4.71, + "learning_rate": 2.3314426734428618e-06, + "loss": 0.9137, + "step": 63727 + }, + { + "epoch": 4.71, + "learning_rate": 2.3311866759549173e-06, + "loss": 1.0385, + "step": 63728 + }, + { + "epoch": 4.71, + "learning_rate": 2.3309306906679407e-06, + "loss": 0.9605, + "step": 63729 + }, + { + "epoch": 4.71, + "learning_rate": 2.3306747175823475e-06, + "loss": 1.0018, + "step": 63730 + }, + { + "epoch": 4.71, + "learning_rate": 2.3304187566985405e-06, + "loss": 0.9301, + "step": 63731 + }, + { + "epoch": 4.71, + "learning_rate": 2.330162808016928e-06, + "loss": 1.0461, + "step": 63732 + }, + { + "epoch": 4.71, + "learning_rate": 2.3299068715379137e-06, + "loss": 0.9794, + "step": 63733 + }, + { + "epoch": 4.71, + "learning_rate": 2.329650947261911e-06, + "loss": 1.0433, + "step": 63734 + }, + { + "epoch": 4.71, + "learning_rate": 2.3293950351893237e-06, + "loss": 1.1034, + "step": 63735 + }, + { + "epoch": 4.71, + "learning_rate": 2.3291391353205584e-06, + "loss": 1.0017, + "step": 63736 + }, + { + "epoch": 4.71, + "learning_rate": 2.3288832476560242e-06, + "loss": 1.0211, + "step": 63737 + }, + { + "epoch": 4.71, + "learning_rate": 2.328627372196124e-06, + "loss": 1.087, + "step": 63738 + }, + { + "epoch": 4.71, + "learning_rate": 2.328371508941273e-06, + "loss": 1.0084, + "step": 63739 + }, + { + "epoch": 4.71, + "learning_rate": 2.3281156578918674e-06, + "loss": 1.0751, + "step": 63740 + }, + { + "epoch": 4.71, + "learning_rate": 2.3278598190483226e-06, + "loss": 0.9172, + "step": 63741 + }, + { + "epoch": 4.71, + "learning_rate": 2.3276039924110406e-06, + "loss": 0.9493, + "step": 63742 + }, + { + "epoch": 4.71, + "learning_rate": 2.3273481779804376e-06, + "loss": 1.0807, + "step": 63743 + }, + { + "epoch": 4.71, + "learning_rate": 2.327092375756906e-06, + "loss": 0.9202, + "step": 63744 + }, + { + "epoch": 4.71, + "learning_rate": 2.3268365857408658e-06, + "loss": 1.0334, + "step": 63745 + }, + { + "epoch": 4.71, + "learning_rate": 2.3265808079327178e-06, + "loss": 0.9265, + "step": 63746 + }, + { + "epoch": 4.71, + "learning_rate": 2.3263250423328696e-06, + "loss": 0.8766, + "step": 63747 + }, + { + "epoch": 4.71, + "learning_rate": 2.326069288941729e-06, + "loss": 1.0282, + "step": 63748 + }, + { + "epoch": 4.71, + "learning_rate": 2.3258135477596987e-06, + "loss": 1.016, + "step": 63749 + }, + { + "epoch": 4.71, + "learning_rate": 2.325557818787193e-06, + "loss": 1.0101, + "step": 63750 + }, + { + "epoch": 4.71, + "learning_rate": 2.3253021020246146e-06, + "loss": 1.1041, + "step": 63751 + }, + { + "epoch": 4.71, + "learning_rate": 2.3250463974723715e-06, + "loss": 1.0741, + "step": 63752 + }, + { + "epoch": 4.71, + "learning_rate": 2.324790705130866e-06, + "loss": 1.053, + "step": 63753 + }, + { + "epoch": 4.71, + "learning_rate": 2.3245350250005114e-06, + "loss": 1.113, + "step": 63754 + }, + { + "epoch": 4.71, + "learning_rate": 2.3242793570817123e-06, + "loss": 1.0426, + "step": 63755 + }, + { + "epoch": 4.71, + "learning_rate": 2.3240237013748743e-06, + "loss": 0.9478, + "step": 63756 + }, + { + "epoch": 4.71, + "learning_rate": 2.3237680578804058e-06, + "loss": 0.9453, + "step": 63757 + }, + { + "epoch": 4.71, + "learning_rate": 2.3235124265987086e-06, + "loss": 1.0108, + "step": 63758 + }, + { + "epoch": 4.71, + "learning_rate": 2.3232568075302e-06, + "loss": 0.989, + "step": 63759 + }, + { + "epoch": 4.71, + "learning_rate": 2.3230012006752735e-06, + "loss": 0.9579, + "step": 63760 + }, + { + "epoch": 4.71, + "learning_rate": 2.3227456060343458e-06, + "loss": 0.9554, + "step": 63761 + }, + { + "epoch": 4.71, + "learning_rate": 2.322490023607816e-06, + "loss": 0.9949, + "step": 63762 + }, + { + "epoch": 4.71, + "learning_rate": 2.3222344533961017e-06, + "loss": 0.9205, + "step": 63763 + }, + { + "epoch": 4.71, + "learning_rate": 2.321978895399596e-06, + "loss": 1.077, + "step": 63764 + }, + { + "epoch": 4.71, + "learning_rate": 2.3217233496187153e-06, + "loss": 0.9554, + "step": 63765 + }, + { + "epoch": 4.71, + "learning_rate": 2.3214678160538627e-06, + "loss": 1.0172, + "step": 63766 + }, + { + "epoch": 4.71, + "learning_rate": 2.321212294705445e-06, + "loss": 1.0233, + "step": 63767 + }, + { + "epoch": 4.71, + "learning_rate": 2.320956785573868e-06, + "loss": 1.0385, + "step": 63768 + }, + { + "epoch": 4.71, + "learning_rate": 2.320701288659537e-06, + "loss": 1.0488, + "step": 63769 + }, + { + "epoch": 4.71, + "learning_rate": 2.320445803962863e-06, + "loss": 0.9799, + "step": 63770 + }, + { + "epoch": 4.71, + "learning_rate": 2.32019033148425e-06, + "loss": 1.0477, + "step": 63771 + }, + { + "epoch": 4.71, + "learning_rate": 2.319934871224104e-06, + "loss": 0.925, + "step": 63772 + }, + { + "epoch": 4.71, + "learning_rate": 2.319679423182829e-06, + "loss": 0.9516, + "step": 63773 + }, + { + "epoch": 4.71, + "learning_rate": 2.319423987360837e-06, + "loss": 1.0381, + "step": 63774 + }, + { + "epoch": 4.71, + "learning_rate": 2.3191685637585313e-06, + "loss": 1.0896, + "step": 63775 + }, + { + "epoch": 4.71, + "learning_rate": 2.3189131523763187e-06, + "loss": 1.087, + "step": 63776 + }, + { + "epoch": 4.71, + "learning_rate": 2.318657753214605e-06, + "loss": 0.9742, + "step": 63777 + }, + { + "epoch": 4.71, + "learning_rate": 2.318402366273794e-06, + "loss": 0.91, + "step": 63778 + }, + { + "epoch": 4.71, + "learning_rate": 2.3181469915542986e-06, + "loss": 0.9705, + "step": 63779 + }, + { + "epoch": 4.71, + "learning_rate": 2.317891629056521e-06, + "loss": 1.106, + "step": 63780 + }, + { + "epoch": 4.71, + "learning_rate": 2.317636278780868e-06, + "loss": 1.0011, + "step": 63781 + }, + { + "epoch": 4.71, + "learning_rate": 2.3173809407277435e-06, + "loss": 0.9386, + "step": 63782 + }, + { + "epoch": 4.71, + "learning_rate": 2.3171256148975574e-06, + "loss": 1.0279, + "step": 63783 + }, + { + "epoch": 4.71, + "learning_rate": 2.316870301290719e-06, + "loss": 0.9904, + "step": 63784 + }, + { + "epoch": 4.71, + "learning_rate": 2.3166149999076236e-06, + "loss": 0.8892, + "step": 63785 + }, + { + "epoch": 4.71, + "learning_rate": 2.3163597107486858e-06, + "loss": 1.0631, + "step": 63786 + }, + { + "epoch": 4.71, + "learning_rate": 2.316104433814308e-06, + "loss": 1.0135, + "step": 63787 + }, + { + "epoch": 4.71, + "learning_rate": 2.3158491691049036e-06, + "loss": 0.9898, + "step": 63788 + }, + { + "epoch": 4.71, + "learning_rate": 2.3155939166208674e-06, + "loss": 1.0362, + "step": 63789 + }, + { + "epoch": 4.71, + "learning_rate": 2.315338676362614e-06, + "loss": 0.9581, + "step": 63790 + }, + { + "epoch": 4.71, + "learning_rate": 2.315083448330547e-06, + "loss": 0.9949, + "step": 63791 + }, + { + "epoch": 4.71, + "learning_rate": 2.3148282325250725e-06, + "loss": 0.8924, + "step": 63792 + }, + { + "epoch": 4.71, + "learning_rate": 2.314573028946596e-06, + "loss": 1.0391, + "step": 63793 + }, + { + "epoch": 4.71, + "learning_rate": 2.314317837595521e-06, + "loss": 0.9568, + "step": 63794 + }, + { + "epoch": 4.71, + "learning_rate": 2.3140626584722593e-06, + "loss": 1.0288, + "step": 63795 + }, + { + "epoch": 4.71, + "learning_rate": 2.3138074915772148e-06, + "loss": 0.8786, + "step": 63796 + }, + { + "epoch": 4.71, + "learning_rate": 2.3135523369107913e-06, + "loss": 0.9922, + "step": 63797 + }, + { + "epoch": 4.71, + "learning_rate": 2.313297194473394e-06, + "loss": 1.0926, + "step": 63798 + }, + { + "epoch": 4.71, + "learning_rate": 2.313042064265433e-06, + "loss": 0.9286, + "step": 63799 + }, + { + "epoch": 4.71, + "learning_rate": 2.312786946287313e-06, + "loss": 1.0726, + "step": 63800 + }, + { + "epoch": 4.71, + "learning_rate": 2.312531840539438e-06, + "loss": 0.903, + "step": 63801 + }, + { + "epoch": 4.71, + "learning_rate": 2.312276747022213e-06, + "loss": 0.9238, + "step": 63802 + }, + { + "epoch": 4.71, + "learning_rate": 2.3120216657360473e-06, + "loss": 0.9921, + "step": 63803 + }, + { + "epoch": 4.71, + "learning_rate": 2.3117665966813495e-06, + "loss": 1.0681, + "step": 63804 + }, + { + "epoch": 4.71, + "learning_rate": 2.3115115398585143e-06, + "loss": 1.0369, + "step": 63805 + }, + { + "epoch": 4.71, + "learning_rate": 2.3112564952679572e-06, + "loss": 0.944, + "step": 63806 + }, + { + "epoch": 4.71, + "learning_rate": 2.3110014629100784e-06, + "loss": 0.8978, + "step": 63807 + }, + { + "epoch": 4.71, + "learning_rate": 2.310746442785292e-06, + "loss": 1.0381, + "step": 63808 + }, + { + "epoch": 4.71, + "learning_rate": 2.3104914348939922e-06, + "loss": 1.0031, + "step": 63809 + }, + { + "epoch": 4.71, + "learning_rate": 2.3102364392365937e-06, + "loss": 1.0515, + "step": 63810 + }, + { + "epoch": 4.71, + "learning_rate": 2.309981455813496e-06, + "loss": 0.9592, + "step": 63811 + }, + { + "epoch": 4.72, + "learning_rate": 2.3097264846251133e-06, + "loss": 0.9754, + "step": 63812 + }, + { + "epoch": 4.72, + "learning_rate": 2.3094715256718424e-06, + "loss": 1.0504, + "step": 63813 + }, + { + "epoch": 4.72, + "learning_rate": 2.30921657895409e-06, + "loss": 0.9608, + "step": 63814 + }, + { + "epoch": 4.72, + "learning_rate": 2.3089616444722662e-06, + "loss": 0.963, + "step": 63815 + }, + { + "epoch": 4.72, + "learning_rate": 2.3087067222267758e-06, + "loss": 0.9347, + "step": 63816 + }, + { + "epoch": 4.72, + "learning_rate": 2.308451812218021e-06, + "loss": 0.865, + "step": 63817 + }, + { + "epoch": 4.72, + "learning_rate": 2.3081969144464078e-06, + "loss": 0.9891, + "step": 63818 + }, + { + "epoch": 4.72, + "learning_rate": 2.3079420289123455e-06, + "loss": 0.9816, + "step": 63819 + }, + { + "epoch": 4.72, + "learning_rate": 2.3076871556162385e-06, + "loss": 1.0821, + "step": 63820 + }, + { + "epoch": 4.72, + "learning_rate": 2.3074322945584905e-06, + "loss": 0.9635, + "step": 63821 + }, + { + "epoch": 4.72, + "learning_rate": 2.3071774457395067e-06, + "loss": 0.9537, + "step": 63822 + }, + { + "epoch": 4.72, + "learning_rate": 2.306922609159692e-06, + "loss": 0.9389, + "step": 63823 + }, + { + "epoch": 4.72, + "learning_rate": 2.306667784819455e-06, + "loss": 1.0001, + "step": 63824 + }, + { + "epoch": 4.72, + "learning_rate": 2.306412972719201e-06, + "loss": 1.0953, + "step": 63825 + }, + { + "epoch": 4.72, + "learning_rate": 2.306158172859333e-06, + "loss": 0.9345, + "step": 63826 + }, + { + "epoch": 4.72, + "learning_rate": 2.3059033852402547e-06, + "loss": 1.0085, + "step": 63827 + }, + { + "epoch": 4.72, + "learning_rate": 2.3056486098623766e-06, + "loss": 0.9938, + "step": 63828 + }, + { + "epoch": 4.72, + "learning_rate": 2.3053938467261015e-06, + "loss": 1.0623, + "step": 63829 + }, + { + "epoch": 4.72, + "learning_rate": 2.3051390958318344e-06, + "loss": 0.9754, + "step": 63830 + }, + { + "epoch": 4.72, + "learning_rate": 2.304884357179982e-06, + "loss": 0.9976, + "step": 63831 + }, + { + "epoch": 4.72, + "learning_rate": 2.304629630770947e-06, + "loss": 1.082, + "step": 63832 + }, + { + "epoch": 4.72, + "learning_rate": 2.3043749166051377e-06, + "loss": 1.0001, + "step": 63833 + }, + { + "epoch": 4.72, + "learning_rate": 2.3041202146829533e-06, + "loss": 0.9705, + "step": 63834 + }, + { + "epoch": 4.72, + "learning_rate": 2.303865525004808e-06, + "loss": 1.0848, + "step": 63835 + }, + { + "epoch": 4.72, + "learning_rate": 2.3036108475711017e-06, + "loss": 0.855, + "step": 63836 + }, + { + "epoch": 4.72, + "learning_rate": 2.3033561823822405e-06, + "loss": 1.095, + "step": 63837 + }, + { + "epoch": 4.72, + "learning_rate": 2.3031015294386262e-06, + "loss": 0.9931, + "step": 63838 + }, + { + "epoch": 4.72, + "learning_rate": 2.3028468887406706e-06, + "loss": 0.9537, + "step": 63839 + }, + { + "epoch": 4.72, + "learning_rate": 2.302592260288775e-06, + "loss": 0.9673, + "step": 63840 + }, + { + "epoch": 4.72, + "learning_rate": 2.3023376440833456e-06, + "loss": 0.9543, + "step": 63841 + }, + { + "epoch": 4.72, + "learning_rate": 2.3020830401247873e-06, + "loss": 0.9474, + "step": 63842 + }, + { + "epoch": 4.72, + "learning_rate": 2.301828448413501e-06, + "loss": 1.0483, + "step": 63843 + }, + { + "epoch": 4.72, + "learning_rate": 2.3015738689498977e-06, + "loss": 1.014, + "step": 63844 + }, + { + "epoch": 4.72, + "learning_rate": 2.3013193017343817e-06, + "loss": 1.0384, + "step": 63845 + }, + { + "epoch": 4.72, + "learning_rate": 2.3010647467673565e-06, + "loss": 0.947, + "step": 63846 + }, + { + "epoch": 4.72, + "learning_rate": 2.300810204049223e-06, + "loss": 0.9821, + "step": 63847 + }, + { + "epoch": 4.72, + "learning_rate": 2.3005556735803947e-06, + "loss": 0.9757, + "step": 63848 + }, + { + "epoch": 4.72, + "learning_rate": 2.300301155361271e-06, + "loss": 0.9652, + "step": 63849 + }, + { + "epoch": 4.72, + "learning_rate": 2.300046649392258e-06, + "loss": 0.9412, + "step": 63850 + }, + { + "epoch": 4.72, + "learning_rate": 2.2997921556737614e-06, + "loss": 0.9837, + "step": 63851 + }, + { + "epoch": 4.72, + "learning_rate": 2.2995376742061816e-06, + "loss": 1.0635, + "step": 63852 + }, + { + "epoch": 4.72, + "learning_rate": 2.299283204989934e-06, + "loss": 1.0755, + "step": 63853 + }, + { + "epoch": 4.72, + "learning_rate": 2.29902874802541e-06, + "loss": 1.1366, + "step": 63854 + }, + { + "epoch": 4.72, + "learning_rate": 2.298774303313024e-06, + "loss": 1.0145, + "step": 63855 + }, + { + "epoch": 4.72, + "learning_rate": 2.298519870853175e-06, + "loss": 0.9573, + "step": 63856 + }, + { + "epoch": 4.72, + "learning_rate": 2.2982654506462763e-06, + "loss": 1.0428, + "step": 63857 + }, + { + "epoch": 4.72, + "learning_rate": 2.2980110426927225e-06, + "loss": 1.0287, + "step": 63858 + }, + { + "epoch": 4.72, + "learning_rate": 2.2977566469929245e-06, + "loss": 1.0362, + "step": 63859 + }, + { + "epoch": 4.72, + "learning_rate": 2.2975022635472865e-06, + "loss": 1.0424, + "step": 63860 + }, + { + "epoch": 4.72, + "learning_rate": 2.2972478923562115e-06, + "loss": 0.9504, + "step": 63861 + }, + { + "epoch": 4.72, + "learning_rate": 2.2969935334201054e-06, + "loss": 1.0639, + "step": 63862 + }, + { + "epoch": 4.72, + "learning_rate": 2.2967391867393683e-06, + "loss": 0.9921, + "step": 63863 + }, + { + "epoch": 4.72, + "learning_rate": 2.2964848523144136e-06, + "loss": 0.9247, + "step": 63864 + }, + { + "epoch": 4.72, + "learning_rate": 2.29623053014564e-06, + "loss": 1.034, + "step": 63865 + }, + { + "epoch": 4.72, + "learning_rate": 2.2959762202334536e-06, + "loss": 0.9637, + "step": 63866 + }, + { + "epoch": 4.72, + "learning_rate": 2.295721922578256e-06, + "loss": 1.0063, + "step": 63867 + }, + { + "epoch": 4.72, + "learning_rate": 2.2954676371804575e-06, + "loss": 0.9713, + "step": 63868 + }, + { + "epoch": 4.72, + "learning_rate": 2.2952133640404605e-06, + "loss": 0.9693, + "step": 63869 + }, + { + "epoch": 4.72, + "learning_rate": 2.294959103158667e-06, + "loss": 0.8535, + "step": 63870 + }, + { + "epoch": 4.72, + "learning_rate": 2.2947048545354856e-06, + "loss": 1.0129, + "step": 63871 + }, + { + "epoch": 4.72, + "learning_rate": 2.2944506181713146e-06, + "loss": 0.9943, + "step": 63872 + }, + { + "epoch": 4.72, + "learning_rate": 2.2941963940665648e-06, + "loss": 0.8809, + "step": 63873 + }, + { + "epoch": 4.72, + "learning_rate": 2.2939421822216377e-06, + "loss": 0.9608, + "step": 63874 + }, + { + "epoch": 4.72, + "learning_rate": 2.2936879826369395e-06, + "loss": 0.9805, + "step": 63875 + }, + { + "epoch": 4.72, + "learning_rate": 2.2934337953128705e-06, + "loss": 0.9856, + "step": 63876 + }, + { + "epoch": 4.72, + "learning_rate": 2.2931796202498448e-06, + "loss": 1.0814, + "step": 63877 + }, + { + "epoch": 4.72, + "learning_rate": 2.292925457448255e-06, + "loss": 0.9978, + "step": 63878 + }, + { + "epoch": 4.72, + "learning_rate": 2.2926713069085093e-06, + "loss": 1.113, + "step": 63879 + }, + { + "epoch": 4.72, + "learning_rate": 2.292417168631016e-06, + "loss": 0.9491, + "step": 63880 + }, + { + "epoch": 4.72, + "learning_rate": 2.2921630426161756e-06, + "loss": 0.9806, + "step": 63881 + }, + { + "epoch": 4.72, + "learning_rate": 2.291908928864395e-06, + "loss": 0.8796, + "step": 63882 + }, + { + "epoch": 4.72, + "learning_rate": 2.291654827376073e-06, + "loss": 0.9136, + "step": 63883 + }, + { + "epoch": 4.72, + "learning_rate": 2.291400738151621e-06, + "loss": 1.0058, + "step": 63884 + }, + { + "epoch": 4.72, + "learning_rate": 2.29114666119144e-06, + "loss": 1.0881, + "step": 63885 + }, + { + "epoch": 4.72, + "learning_rate": 2.290892596495935e-06, + "loss": 1.0598, + "step": 63886 + }, + { + "epoch": 4.72, + "learning_rate": 2.290638544065509e-06, + "loss": 0.9904, + "step": 63887 + }, + { + "epoch": 4.72, + "learning_rate": 2.2903845039005645e-06, + "loss": 0.9243, + "step": 63888 + }, + { + "epoch": 4.72, + "learning_rate": 2.2901304760015097e-06, + "loss": 1.0555, + "step": 63889 + }, + { + "epoch": 4.72, + "learning_rate": 2.2898764603687486e-06, + "loss": 1.0205, + "step": 63890 + }, + { + "epoch": 4.72, + "learning_rate": 2.2896224570026827e-06, + "loss": 0.9379, + "step": 63891 + }, + { + "epoch": 4.72, + "learning_rate": 2.289368465903714e-06, + "loss": 0.9329, + "step": 63892 + }, + { + "epoch": 4.72, + "learning_rate": 2.2891144870722527e-06, + "loss": 1.0513, + "step": 63893 + }, + { + "epoch": 4.72, + "learning_rate": 2.288860520508701e-06, + "loss": 1.0225, + "step": 63894 + }, + { + "epoch": 4.72, + "learning_rate": 2.2886065662134607e-06, + "loss": 1.0581, + "step": 63895 + }, + { + "epoch": 4.72, + "learning_rate": 2.288352624186938e-06, + "loss": 1.0072, + "step": 63896 + }, + { + "epoch": 4.72, + "learning_rate": 2.2880986944295324e-06, + "loss": 0.8821, + "step": 63897 + }, + { + "epoch": 4.72, + "learning_rate": 2.287844776941657e-06, + "loss": 1.0057, + "step": 63898 + }, + { + "epoch": 4.72, + "learning_rate": 2.287590871723705e-06, + "loss": 1.0322, + "step": 63899 + }, + { + "epoch": 4.72, + "learning_rate": 2.2873369787760878e-06, + "loss": 0.9952, + "step": 63900 + }, + { + "epoch": 4.72, + "learning_rate": 2.2870830980992053e-06, + "loss": 1.0429, + "step": 63901 + }, + { + "epoch": 4.72, + "learning_rate": 2.286829229693468e-06, + "loss": 1.0494, + "step": 63902 + }, + { + "epoch": 4.72, + "learning_rate": 2.2865753735592698e-06, + "loss": 1.0543, + "step": 63903 + }, + { + "epoch": 4.72, + "learning_rate": 2.2863215296970222e-06, + "loss": 0.9878, + "step": 63904 + }, + { + "epoch": 4.72, + "learning_rate": 2.286067698107125e-06, + "loss": 1.0137, + "step": 63905 + }, + { + "epoch": 4.72, + "learning_rate": 2.2858138787899886e-06, + "loss": 0.9544, + "step": 63906 + }, + { + "epoch": 4.72, + "learning_rate": 2.2855600717460093e-06, + "loss": 0.967, + "step": 63907 + }, + { + "epoch": 4.72, + "learning_rate": 2.2853062769755895e-06, + "loss": 1.1008, + "step": 63908 + }, + { + "epoch": 4.72, + "learning_rate": 2.285052494479141e-06, + "loss": 1.1361, + "step": 63909 + }, + { + "epoch": 4.72, + "learning_rate": 2.2847987242570636e-06, + "loss": 0.9446, + "step": 63910 + }, + { + "epoch": 4.72, + "learning_rate": 2.2845449663097605e-06, + "loss": 1.0732, + "step": 63911 + }, + { + "epoch": 4.72, + "learning_rate": 2.2842912206376343e-06, + "loss": 1.0243, + "step": 63912 + }, + { + "epoch": 4.72, + "learning_rate": 2.284037487241092e-06, + "loss": 1.0257, + "step": 63913 + }, + { + "epoch": 4.72, + "learning_rate": 2.2837837661205364e-06, + "loss": 1.0048, + "step": 63914 + }, + { + "epoch": 4.72, + "learning_rate": 2.2835300572763707e-06, + "loss": 0.9899, + "step": 63915 + }, + { + "epoch": 4.72, + "learning_rate": 2.283276360708997e-06, + "loss": 0.8934, + "step": 63916 + }, + { + "epoch": 4.72, + "learning_rate": 2.283022676418819e-06, + "loss": 0.9038, + "step": 63917 + }, + { + "epoch": 4.72, + "learning_rate": 2.2827690044062436e-06, + "loss": 1.0105, + "step": 63918 + }, + { + "epoch": 4.72, + "learning_rate": 2.282515344671672e-06, + "loss": 1.0971, + "step": 63919 + }, + { + "epoch": 4.72, + "learning_rate": 2.2822616972155076e-06, + "loss": 0.9929, + "step": 63920 + }, + { + "epoch": 4.72, + "learning_rate": 2.2820080620381522e-06, + "loss": 1.076, + "step": 63921 + }, + { + "epoch": 4.72, + "learning_rate": 2.281754439140017e-06, + "loss": 0.9736, + "step": 63922 + }, + { + "epoch": 4.72, + "learning_rate": 2.2815008285214946e-06, + "loss": 0.9024, + "step": 63923 + }, + { + "epoch": 4.72, + "learning_rate": 2.2812472301829967e-06, + "loss": 1.0003, + "step": 63924 + }, + { + "epoch": 4.72, + "learning_rate": 2.2809936441249247e-06, + "loss": 1.0041, + "step": 63925 + }, + { + "epoch": 4.72, + "learning_rate": 2.280740070347681e-06, + "loss": 1.0934, + "step": 63926 + }, + { + "epoch": 4.72, + "learning_rate": 2.2804865088516694e-06, + "loss": 1.0647, + "step": 63927 + }, + { + "epoch": 4.72, + "learning_rate": 2.28023295963729e-06, + "loss": 0.9443, + "step": 63928 + }, + { + "epoch": 4.72, + "learning_rate": 2.279979422704952e-06, + "loss": 1.0164, + "step": 63929 + }, + { + "epoch": 4.72, + "learning_rate": 2.2797258980550572e-06, + "loss": 0.9095, + "step": 63930 + }, + { + "epoch": 4.72, + "learning_rate": 2.2794723856880076e-06, + "loss": 1.0131, + "step": 63931 + }, + { + "epoch": 4.72, + "learning_rate": 2.279218885604204e-06, + "loss": 1.0312, + "step": 63932 + }, + { + "epoch": 4.72, + "learning_rate": 2.278965397804056e-06, + "loss": 0.9559, + "step": 63933 + }, + { + "epoch": 4.72, + "learning_rate": 2.278711922287964e-06, + "loss": 0.8834, + "step": 63934 + }, + { + "epoch": 4.72, + "learning_rate": 2.27845845905633e-06, + "loss": 1.005, + "step": 63935 + }, + { + "epoch": 4.72, + "learning_rate": 2.2782050081095586e-06, + "loss": 1.0166, + "step": 63936 + }, + { + "epoch": 4.72, + "learning_rate": 2.277951569448049e-06, + "loss": 1.0531, + "step": 63937 + }, + { + "epoch": 4.72, + "learning_rate": 2.277698143072211e-06, + "loss": 0.9998, + "step": 63938 + }, + { + "epoch": 4.72, + "learning_rate": 2.277444728982445e-06, + "loss": 0.9193, + "step": 63939 + }, + { + "epoch": 4.72, + "learning_rate": 2.2771913271791546e-06, + "loss": 0.95, + "step": 63940 + }, + { + "epoch": 4.72, + "learning_rate": 2.276937937662739e-06, + "loss": 1.017, + "step": 63941 + }, + { + "epoch": 4.72, + "learning_rate": 2.2766845604336072e-06, + "loss": 1.0041, + "step": 63942 + }, + { + "epoch": 4.72, + "learning_rate": 2.2764311954921627e-06, + "loss": 0.968, + "step": 63943 + }, + { + "epoch": 4.72, + "learning_rate": 2.2761778428388004e-06, + "loss": 0.8156, + "step": 63944 + }, + { + "epoch": 4.72, + "learning_rate": 2.275924502473931e-06, + "loss": 1.0498, + "step": 63945 + }, + { + "epoch": 4.72, + "learning_rate": 2.275671174397952e-06, + "loss": 1.0398, + "step": 63946 + }, + { + "epoch": 4.73, + "learning_rate": 2.275417858611276e-06, + "loss": 1.0111, + "step": 63947 + }, + { + "epoch": 4.73, + "learning_rate": 2.2751645551142932e-06, + "loss": 1.0232, + "step": 63948 + }, + { + "epoch": 4.73, + "learning_rate": 2.2749112639074166e-06, + "loss": 0.9539, + "step": 63949 + }, + { + "epoch": 4.73, + "learning_rate": 2.2746579849910434e-06, + "loss": 0.9817, + "step": 63950 + }, + { + "epoch": 4.73, + "learning_rate": 2.274404718365584e-06, + "loss": 1.0757, + "step": 63951 + }, + { + "epoch": 4.73, + "learning_rate": 2.274151464031432e-06, + "loss": 0.9519, + "step": 63952 + }, + { + "epoch": 4.73, + "learning_rate": 2.273898221988993e-06, + "loss": 1.0751, + "step": 63953 + }, + { + "epoch": 4.73, + "learning_rate": 2.273644992238674e-06, + "loss": 1.0006, + "step": 63954 + }, + { + "epoch": 4.73, + "learning_rate": 2.273391774780875e-06, + "loss": 0.9863, + "step": 63955 + }, + { + "epoch": 4.73, + "learning_rate": 2.2731385696160004e-06, + "loss": 1.0765, + "step": 63956 + }, + { + "epoch": 4.73, + "learning_rate": 2.272885376744447e-06, + "loss": 0.9973, + "step": 63957 + }, + { + "epoch": 4.73, + "learning_rate": 2.2726321961666266e-06, + "loss": 1.0109, + "step": 63958 + }, + { + "epoch": 4.73, + "learning_rate": 2.2723790278829373e-06, + "loss": 1.0298, + "step": 63959 + }, + { + "epoch": 4.73, + "learning_rate": 2.272125871893782e-06, + "loss": 0.9427, + "step": 63960 + }, + { + "epoch": 4.73, + "learning_rate": 2.271872728199562e-06, + "loss": 0.9905, + "step": 63961 + }, + { + "epoch": 4.73, + "learning_rate": 2.2716195968006836e-06, + "loss": 0.9792, + "step": 63962 + }, + { + "epoch": 4.73, + "learning_rate": 2.2713664776975484e-06, + "loss": 0.9454, + "step": 63963 + }, + { + "epoch": 4.73, + "learning_rate": 2.2711133708905596e-06, + "loss": 1.0795, + "step": 63964 + }, + { + "epoch": 4.73, + "learning_rate": 2.270860276380118e-06, + "loss": 1.0856, + "step": 63965 + }, + { + "epoch": 4.73, + "learning_rate": 2.270607194166624e-06, + "loss": 1.1602, + "step": 63966 + }, + { + "epoch": 4.73, + "learning_rate": 2.270354124250489e-06, + "loss": 1.0361, + "step": 63967 + }, + { + "epoch": 4.73, + "learning_rate": 2.270101066632104e-06, + "loss": 1.0219, + "step": 63968 + }, + { + "epoch": 4.73, + "learning_rate": 2.2698480213118812e-06, + "loss": 1.0517, + "step": 63969 + }, + { + "epoch": 4.73, + "learning_rate": 2.2695949882902166e-06, + "loss": 0.9203, + "step": 63970 + }, + { + "epoch": 4.73, + "learning_rate": 2.269341967567522e-06, + "loss": 1.0114, + "step": 63971 + }, + { + "epoch": 4.73, + "learning_rate": 2.26908895914419e-06, + "loss": 0.8808, + "step": 63972 + }, + { + "epoch": 4.73, + "learning_rate": 2.2688359630206247e-06, + "loss": 1.0128, + "step": 63973 + }, + { + "epoch": 4.73, + "learning_rate": 2.2685829791972324e-06, + "loss": 1.0538, + "step": 63974 + }, + { + "epoch": 4.73, + "learning_rate": 2.2683300076744153e-06, + "loss": 0.9069, + "step": 63975 + }, + { + "epoch": 4.73, + "learning_rate": 2.2680770484525736e-06, + "loss": 0.8774, + "step": 63976 + }, + { + "epoch": 4.73, + "learning_rate": 2.267824101532109e-06, + "loss": 1.0066, + "step": 63977 + }, + { + "epoch": 4.73, + "learning_rate": 2.267571166913427e-06, + "loss": 1.0957, + "step": 63978 + }, + { + "epoch": 4.73, + "learning_rate": 2.26731824459693e-06, + "loss": 1.0157, + "step": 63979 + }, + { + "epoch": 4.73, + "learning_rate": 2.2670653345830184e-06, + "loss": 0.9479, + "step": 63980 + }, + { + "epoch": 4.73, + "learning_rate": 2.2668124368720944e-06, + "loss": 0.9682, + "step": 63981 + }, + { + "epoch": 4.73, + "learning_rate": 2.26655955146456e-06, + "loss": 1.028, + "step": 63982 + }, + { + "epoch": 4.73, + "learning_rate": 2.26630667836082e-06, + "loss": 1.0278, + "step": 63983 + }, + { + "epoch": 4.73, + "learning_rate": 2.266053817561277e-06, + "loss": 1.0382, + "step": 63984 + }, + { + "epoch": 4.73, + "learning_rate": 2.2658009690663307e-06, + "loss": 0.9521, + "step": 63985 + }, + { + "epoch": 4.73, + "learning_rate": 2.2655481328763816e-06, + "loss": 0.999, + "step": 63986 + }, + { + "epoch": 4.73, + "learning_rate": 2.265295308991837e-06, + "loss": 0.9366, + "step": 63987 + }, + { + "epoch": 4.73, + "learning_rate": 2.265042497413098e-06, + "loss": 0.99, + "step": 63988 + }, + { + "epoch": 4.73, + "learning_rate": 2.264789698140566e-06, + "loss": 1.013, + "step": 63989 + }, + { + "epoch": 4.73, + "learning_rate": 2.264536911174642e-06, + "loss": 1.0521, + "step": 63990 + }, + { + "epoch": 4.73, + "learning_rate": 2.2642841365157263e-06, + "loss": 1.0029, + "step": 63991 + }, + { + "epoch": 4.73, + "learning_rate": 2.2640313741642295e-06, + "loss": 0.9674, + "step": 63992 + }, + { + "epoch": 4.73, + "learning_rate": 2.2637786241205427e-06, + "loss": 0.9009, + "step": 63993 + }, + { + "epoch": 4.73, + "learning_rate": 2.263525886385076e-06, + "loss": 1.0069, + "step": 63994 + }, + { + "epoch": 4.73, + "learning_rate": 2.263273160958227e-06, + "loss": 0.9286, + "step": 63995 + }, + { + "epoch": 4.73, + "learning_rate": 2.263020447840404e-06, + "loss": 1.0498, + "step": 63996 + }, + { + "epoch": 4.73, + "learning_rate": 2.262767747032e-06, + "loss": 1.1095, + "step": 63997 + }, + { + "epoch": 4.73, + "learning_rate": 2.2625150585334242e-06, + "loss": 0.96, + "step": 63998 + }, + { + "epoch": 4.73, + "learning_rate": 2.262262382345076e-06, + "loss": 0.9655, + "step": 63999 + }, + { + "epoch": 4.73, + "learning_rate": 2.2620097184673585e-06, + "loss": 1.0577, + "step": 64000 + }, + { + "epoch": 4.73, + "learning_rate": 2.2617570669006726e-06, + "loss": 1.1095, + "step": 64001 + }, + { + "epoch": 4.73, + "learning_rate": 2.261504427645418e-06, + "loss": 1.0428, + "step": 64002 + }, + { + "epoch": 4.73, + "learning_rate": 2.2612518007020013e-06, + "loss": 0.9626, + "step": 64003 + }, + { + "epoch": 4.73, + "learning_rate": 2.2609991860708224e-06, + "loss": 0.9531, + "step": 64004 + }, + { + "epoch": 4.73, + "learning_rate": 2.2607465837522825e-06, + "loss": 0.9898, + "step": 64005 + }, + { + "epoch": 4.73, + "learning_rate": 2.260493993746782e-06, + "loss": 1.0155, + "step": 64006 + }, + { + "epoch": 4.73, + "learning_rate": 2.260241416054727e-06, + "loss": 1.0148, + "step": 64007 + }, + { + "epoch": 4.73, + "learning_rate": 2.2599888506765176e-06, + "loss": 1.0272, + "step": 64008 + }, + { + "epoch": 4.73, + "learning_rate": 2.259736297612556e-06, + "loss": 1.0664, + "step": 64009 + }, + { + "epoch": 4.73, + "learning_rate": 2.2594837568632423e-06, + "loss": 0.9019, + "step": 64010 + }, + { + "epoch": 4.73, + "learning_rate": 2.2592312284289765e-06, + "loss": 1.0471, + "step": 64011 + }, + { + "epoch": 4.73, + "learning_rate": 2.258978712310168e-06, + "loss": 0.9338, + "step": 64012 + }, + { + "epoch": 4.73, + "learning_rate": 2.258726208507209e-06, + "loss": 1.0107, + "step": 64013 + }, + { + "epoch": 4.73, + "learning_rate": 2.2584737170205083e-06, + "loss": 0.9148, + "step": 64014 + }, + { + "epoch": 4.73, + "learning_rate": 2.258221237850463e-06, + "loss": 1.1318, + "step": 64015 + }, + { + "epoch": 4.73, + "learning_rate": 2.257968770997482e-06, + "loss": 0.9261, + "step": 64016 + }, + { + "epoch": 4.73, + "learning_rate": 2.2577163164619564e-06, + "loss": 0.9794, + "step": 64017 + }, + { + "epoch": 4.73, + "learning_rate": 2.2574638742442958e-06, + "loss": 0.9895, + "step": 64018 + }, + { + "epoch": 4.73, + "learning_rate": 2.257211444344899e-06, + "loss": 1.0082, + "step": 64019 + }, + { + "epoch": 4.73, + "learning_rate": 2.2569590267641694e-06, + "loss": 1.0696, + "step": 64020 + }, + { + "epoch": 4.73, + "learning_rate": 2.256706621502506e-06, + "loss": 0.9726, + "step": 64021 + }, + { + "epoch": 4.73, + "learning_rate": 2.2564542285603096e-06, + "loss": 1.0394, + "step": 64022 + }, + { + "epoch": 4.73, + "learning_rate": 2.2562018479379854e-06, + "loss": 1.0328, + "step": 64023 + }, + { + "epoch": 4.73, + "learning_rate": 2.255949479635934e-06, + "loss": 1.0564, + "step": 64024 + }, + { + "epoch": 4.73, + "learning_rate": 2.2556971236545567e-06, + "loss": 0.9827, + "step": 64025 + }, + { + "epoch": 4.73, + "learning_rate": 2.2554447799942515e-06, + "loss": 0.9421, + "step": 64026 + }, + { + "epoch": 4.73, + "learning_rate": 2.2551924486554245e-06, + "loss": 0.9992, + "step": 64027 + }, + { + "epoch": 4.73, + "learning_rate": 2.2549401296384765e-06, + "loss": 0.9299, + "step": 64028 + }, + { + "epoch": 4.73, + "learning_rate": 2.2546878229438076e-06, + "loss": 0.9219, + "step": 64029 + }, + { + "epoch": 4.73, + "learning_rate": 2.25443552857182e-06, + "loss": 0.9405, + "step": 64030 + }, + { + "epoch": 4.73, + "learning_rate": 2.254183246522912e-06, + "loss": 1.0332, + "step": 64031 + }, + { + "epoch": 4.73, + "learning_rate": 2.2539309767974894e-06, + "loss": 1.0006, + "step": 64032 + }, + { + "epoch": 4.73, + "learning_rate": 2.253678719395952e-06, + "loss": 0.9182, + "step": 64033 + }, + { + "epoch": 4.73, + "learning_rate": 2.2534264743187016e-06, + "loss": 1.0718, + "step": 64034 + }, + { + "epoch": 4.73, + "learning_rate": 2.253174241566136e-06, + "loss": 0.9477, + "step": 64035 + }, + { + "epoch": 4.73, + "learning_rate": 2.2529220211386616e-06, + "loss": 0.983, + "step": 64036 + }, + { + "epoch": 4.73, + "learning_rate": 2.252669813036681e-06, + "loss": 0.953, + "step": 64037 + }, + { + "epoch": 4.73, + "learning_rate": 2.252417617260585e-06, + "loss": 1.0393, + "step": 64038 + }, + { + "epoch": 4.73, + "learning_rate": 2.2521654338107855e-06, + "loss": 1.0306, + "step": 64039 + }, + { + "epoch": 4.73, + "learning_rate": 2.2519132626876773e-06, + "loss": 0.9857, + "step": 64040 + }, + { + "epoch": 4.73, + "learning_rate": 2.251661103891669e-06, + "loss": 0.9143, + "step": 64041 + }, + { + "epoch": 4.73, + "learning_rate": 2.251408957423152e-06, + "loss": 0.9546, + "step": 64042 + }, + { + "epoch": 4.73, + "learning_rate": 2.2511568232825354e-06, + "loss": 1.0098, + "step": 64043 + }, + { + "epoch": 4.73, + "learning_rate": 2.2509047014702167e-06, + "loss": 1.0711, + "step": 64044 + }, + { + "epoch": 4.73, + "learning_rate": 2.2506525919865985e-06, + "loss": 0.9212, + "step": 64045 + }, + { + "epoch": 4.73, + "learning_rate": 2.250400494832081e-06, + "loss": 1.0862, + "step": 64046 + }, + { + "epoch": 4.73, + "learning_rate": 2.2501484100070625e-06, + "loss": 1.029, + "step": 64047 + }, + { + "epoch": 4.73, + "learning_rate": 2.2498963375119497e-06, + "loss": 1.0588, + "step": 64048 + }, + { + "epoch": 4.73, + "learning_rate": 2.2496442773471417e-06, + "loss": 0.9769, + "step": 64049 + }, + { + "epoch": 4.73, + "learning_rate": 2.2493922295130378e-06, + "loss": 0.9761, + "step": 64050 + }, + { + "epoch": 4.73, + "learning_rate": 2.249140194010038e-06, + "loss": 1.1196, + "step": 64051 + }, + { + "epoch": 4.73, + "learning_rate": 2.248888170838548e-06, + "loss": 0.9729, + "step": 64052 + }, + { + "epoch": 4.73, + "learning_rate": 2.2486361599989648e-06, + "loss": 1.0409, + "step": 64053 + }, + { + "epoch": 4.73, + "learning_rate": 2.248384161491691e-06, + "loss": 0.9995, + "step": 64054 + }, + { + "epoch": 4.73, + "learning_rate": 2.248132175317128e-06, + "loss": 1.0893, + "step": 64055 + }, + { + "epoch": 4.73, + "learning_rate": 2.2478802014756727e-06, + "loss": 0.9583, + "step": 64056 + }, + { + "epoch": 4.73, + "learning_rate": 2.2476282399677342e-06, + "loss": 1.0028, + "step": 64057 + }, + { + "epoch": 4.73, + "learning_rate": 2.2473762907937025e-06, + "loss": 0.9029, + "step": 64058 + }, + { + "epoch": 4.73, + "learning_rate": 2.2471243539539876e-06, + "loss": 1.0301, + "step": 64059 + }, + { + "epoch": 4.73, + "learning_rate": 2.246872429448984e-06, + "loss": 0.9836, + "step": 64060 + }, + { + "epoch": 4.73, + "learning_rate": 2.246620517279101e-06, + "loss": 1.0251, + "step": 64061 + }, + { + "epoch": 4.73, + "learning_rate": 2.2463686174447285e-06, + "loss": 0.9536, + "step": 64062 + }, + { + "epoch": 4.73, + "learning_rate": 2.246116729946275e-06, + "loss": 1.0206, + "step": 64063 + }, + { + "epoch": 4.73, + "learning_rate": 2.2458648547841367e-06, + "loss": 1.0396, + "step": 64064 + }, + { + "epoch": 4.73, + "learning_rate": 2.245612991958721e-06, + "loss": 0.9537, + "step": 64065 + }, + { + "epoch": 4.73, + "learning_rate": 2.2453611414704214e-06, + "loss": 1.0069, + "step": 64066 + }, + { + "epoch": 4.73, + "learning_rate": 2.245109303319639e-06, + "loss": 0.95, + "step": 64067 + }, + { + "epoch": 4.73, + "learning_rate": 2.24485747750678e-06, + "loss": 0.9976, + "step": 64068 + }, + { + "epoch": 4.73, + "learning_rate": 2.244605664032241e-06, + "loss": 1.0225, + "step": 64069 + }, + { + "epoch": 4.73, + "learning_rate": 2.2443538628964236e-06, + "loss": 1.0403, + "step": 64070 + }, + { + "epoch": 4.73, + "learning_rate": 2.2441020740997254e-06, + "loss": 0.9091, + "step": 64071 + }, + { + "epoch": 4.73, + "learning_rate": 2.2438502976425527e-06, + "loss": 0.9601, + "step": 64072 + }, + { + "epoch": 4.73, + "learning_rate": 2.243598533525303e-06, + "loss": 1.0541, + "step": 64073 + }, + { + "epoch": 4.73, + "learning_rate": 2.243346781748377e-06, + "loss": 1.002, + "step": 64074 + }, + { + "epoch": 4.73, + "learning_rate": 2.2430950423121755e-06, + "loss": 0.9907, + "step": 64075 + }, + { + "epoch": 4.73, + "learning_rate": 2.242843315217096e-06, + "loss": 1.0277, + "step": 64076 + }, + { + "epoch": 4.73, + "learning_rate": 2.242591600463545e-06, + "loss": 0.9834, + "step": 64077 + }, + { + "epoch": 4.73, + "learning_rate": 2.2423398980519195e-06, + "loss": 0.9602, + "step": 64078 + }, + { + "epoch": 4.73, + "learning_rate": 2.2420882079826203e-06, + "loss": 1.1348, + "step": 64079 + }, + { + "epoch": 4.73, + "learning_rate": 2.241836530256044e-06, + "loss": 0.9419, + "step": 64080 + }, + { + "epoch": 4.73, + "learning_rate": 2.241584864872599e-06, + "loss": 1.0526, + "step": 64081 + }, + { + "epoch": 4.73, + "learning_rate": 2.2413332118326803e-06, + "loss": 0.9994, + "step": 64082 + }, + { + "epoch": 4.74, + "learning_rate": 2.2410815711366906e-06, + "loss": 1.1119, + "step": 64083 + }, + { + "epoch": 4.74, + "learning_rate": 2.240829942785029e-06, + "loss": 0.9254, + "step": 64084 + }, + { + "epoch": 4.74, + "learning_rate": 2.2405783267780922e-06, + "loss": 0.9968, + "step": 64085 + }, + { + "epoch": 4.74, + "learning_rate": 2.2403267231162904e-06, + "loss": 1.009, + "step": 64086 + }, + { + "epoch": 4.74, + "learning_rate": 2.2400751318000125e-06, + "loss": 1.0182, + "step": 64087 + }, + { + "epoch": 4.74, + "learning_rate": 2.2398235528296663e-06, + "loss": 0.9079, + "step": 64088 + }, + { + "epoch": 4.74, + "learning_rate": 2.23957198620565e-06, + "loss": 1.0886, + "step": 64089 + }, + { + "epoch": 4.74, + "learning_rate": 2.2393204319283633e-06, + "loss": 1.0832, + "step": 64090 + }, + { + "epoch": 4.74, + "learning_rate": 2.2390688899982038e-06, + "loss": 1.0339, + "step": 64091 + }, + { + "epoch": 4.74, + "learning_rate": 2.2388173604155784e-06, + "loss": 0.9807, + "step": 64092 + }, + { + "epoch": 4.74, + "learning_rate": 2.2385658431808834e-06, + "loss": 0.9066, + "step": 64093 + }, + { + "epoch": 4.74, + "learning_rate": 2.2383143382945184e-06, + "loss": 1.0962, + "step": 64094 + }, + { + "epoch": 4.74, + "learning_rate": 2.238062845756884e-06, + "loss": 0.8723, + "step": 64095 + }, + { + "epoch": 4.74, + "learning_rate": 2.2378113655683786e-06, + "loss": 1.0582, + "step": 64096 + }, + { + "epoch": 4.74, + "learning_rate": 2.237559897729407e-06, + "loss": 0.9822, + "step": 64097 + }, + { + "epoch": 4.74, + "learning_rate": 2.2373084422403656e-06, + "loss": 1.0102, + "step": 64098 + }, + { + "epoch": 4.74, + "learning_rate": 2.2370569991016568e-06, + "loss": 1.0051, + "step": 64099 + }, + { + "epoch": 4.74, + "learning_rate": 2.236805568313676e-06, + "loss": 0.9782, + "step": 64100 + }, + { + "epoch": 4.74, + "learning_rate": 2.2365541498768285e-06, + "loss": 1.0, + "step": 64101 + }, + { + "epoch": 4.74, + "learning_rate": 2.2363027437915154e-06, + "loss": 0.9567, + "step": 64102 + }, + { + "epoch": 4.74, + "learning_rate": 2.236051350058128e-06, + "loss": 0.9705, + "step": 64103 + }, + { + "epoch": 4.74, + "learning_rate": 2.2357999686770747e-06, + "loss": 1.0001, + "step": 64104 + }, + { + "epoch": 4.74, + "learning_rate": 2.2355485996487504e-06, + "loss": 0.9242, + "step": 64105 + }, + { + "epoch": 4.74, + "learning_rate": 2.235297242973562e-06, + "loss": 1.0683, + "step": 64106 + }, + { + "epoch": 4.74, + "learning_rate": 2.2350458986519e-06, + "loss": 1.0532, + "step": 64107 + }, + { + "epoch": 4.74, + "learning_rate": 2.23479456668417e-06, + "loss": 0.9847, + "step": 64108 + }, + { + "epoch": 4.74, + "learning_rate": 2.2345432470707686e-06, + "loss": 0.9557, + "step": 64109 + }, + { + "epoch": 4.74, + "learning_rate": 2.2342919398121043e-06, + "loss": 0.9628, + "step": 64110 + }, + { + "epoch": 4.74, + "learning_rate": 2.2340406449085663e-06, + "loss": 0.989, + "step": 64111 + }, + { + "epoch": 4.74, + "learning_rate": 2.2337893623605568e-06, + "loss": 0.9658, + "step": 64112 + }, + { + "epoch": 4.74, + "learning_rate": 2.233538092168479e-06, + "loss": 1.052, + "step": 64113 + }, + { + "epoch": 4.74, + "learning_rate": 2.2332868343327317e-06, + "loss": 0.9704, + "step": 64114 + }, + { + "epoch": 4.74, + "learning_rate": 2.2330355888537135e-06, + "loss": 0.9388, + "step": 64115 + }, + { + "epoch": 4.74, + "learning_rate": 2.232784355731823e-06, + "loss": 1.0092, + "step": 64116 + }, + { + "epoch": 4.74, + "learning_rate": 2.2325331349674627e-06, + "loss": 0.9904, + "step": 64117 + }, + { + "epoch": 4.74, + "learning_rate": 2.232281926561032e-06, + "loss": 1.0896, + "step": 64118 + }, + { + "epoch": 4.74, + "learning_rate": 2.2320307305129295e-06, + "loss": 0.8913, + "step": 64119 + }, + { + "epoch": 4.74, + "learning_rate": 2.2317795468235516e-06, + "loss": 1.0291, + "step": 64120 + }, + { + "epoch": 4.74, + "learning_rate": 2.2315283754933046e-06, + "loss": 1.1337, + "step": 64121 + }, + { + "epoch": 4.74, + "learning_rate": 2.231277216522585e-06, + "loss": 1.0697, + "step": 64122 + }, + { + "epoch": 4.74, + "learning_rate": 2.2310260699117915e-06, + "loss": 1.0873, + "step": 64123 + }, + { + "epoch": 4.74, + "learning_rate": 2.230774935661325e-06, + "loss": 0.9888, + "step": 64124 + }, + { + "epoch": 4.74, + "learning_rate": 2.2305238137715813e-06, + "loss": 1.0535, + "step": 64125 + }, + { + "epoch": 4.74, + "learning_rate": 2.230272704242966e-06, + "loss": 0.9579, + "step": 64126 + }, + { + "epoch": 4.74, + "learning_rate": 2.2300216070758763e-06, + "loss": 0.995, + "step": 64127 + }, + { + "epoch": 4.74, + "learning_rate": 2.22977052227071e-06, + "loss": 0.9682, + "step": 64128 + }, + { + "epoch": 4.74, + "learning_rate": 2.229519449827865e-06, + "loss": 1.0402, + "step": 64129 + }, + { + "epoch": 4.74, + "learning_rate": 2.2292683897477463e-06, + "loss": 1.0335, + "step": 64130 + }, + { + "epoch": 4.74, + "learning_rate": 2.229017342030754e-06, + "loss": 1.0492, + "step": 64131 + }, + { + "epoch": 4.74, + "learning_rate": 2.228766306677277e-06, + "loss": 1.0046, + "step": 64132 + }, + { + "epoch": 4.74, + "learning_rate": 2.2285152836877242e-06, + "loss": 1.0588, + "step": 64133 + }, + { + "epoch": 4.74, + "learning_rate": 2.228264273062494e-06, + "loss": 1.0188, + "step": 64134 + }, + { + "epoch": 4.74, + "learning_rate": 2.2280132748019833e-06, + "loss": 1.0056, + "step": 64135 + }, + { + "epoch": 4.74, + "learning_rate": 2.2277622889065896e-06, + "loss": 1.0752, + "step": 64136 + }, + { + "epoch": 4.74, + "learning_rate": 2.227511315376717e-06, + "loss": 0.9027, + "step": 64137 + }, + { + "epoch": 4.74, + "learning_rate": 2.2272603542127635e-06, + "loss": 0.9695, + "step": 64138 + }, + { + "epoch": 4.74, + "learning_rate": 2.227009405415128e-06, + "loss": 0.9715, + "step": 64139 + }, + { + "epoch": 4.74, + "learning_rate": 2.226758468984208e-06, + "loss": 0.9854, + "step": 64140 + }, + { + "epoch": 4.74, + "learning_rate": 2.226507544920402e-06, + "loss": 0.95, + "step": 64141 + }, + { + "epoch": 4.74, + "learning_rate": 2.2262566332241142e-06, + "loss": 1.0438, + "step": 64142 + }, + { + "epoch": 4.74, + "learning_rate": 2.2260057338957406e-06, + "loss": 1.2125, + "step": 64143 + }, + { + "epoch": 4.74, + "learning_rate": 2.2257548469356803e-06, + "loss": 1.0532, + "step": 64144 + }, + { + "epoch": 4.74, + "learning_rate": 2.2255039723443305e-06, + "loss": 1.0391, + "step": 64145 + }, + { + "epoch": 4.74, + "learning_rate": 2.2252531101220965e-06, + "loss": 0.9258, + "step": 64146 + }, + { + "epoch": 4.74, + "learning_rate": 2.225002260269372e-06, + "loss": 0.9755, + "step": 64147 + }, + { + "epoch": 4.74, + "learning_rate": 2.2247514227865586e-06, + "loss": 0.9991, + "step": 64148 + }, + { + "epoch": 4.74, + "learning_rate": 2.224500597674053e-06, + "loss": 1.052, + "step": 64149 + }, + { + "epoch": 4.74, + "learning_rate": 2.224249784932254e-06, + "loss": 0.9994, + "step": 64150 + }, + { + "epoch": 4.74, + "learning_rate": 2.2239989845615684e-06, + "loss": 1.1568, + "step": 64151 + }, + { + "epoch": 4.74, + "learning_rate": 2.223748196562383e-06, + "loss": 1.0023, + "step": 64152 + }, + { + "epoch": 4.74, + "learning_rate": 2.2234974209351057e-06, + "loss": 0.9765, + "step": 64153 + }, + { + "epoch": 4.74, + "learning_rate": 2.2232466576801292e-06, + "loss": 0.9467, + "step": 64154 + }, + { + "epoch": 4.74, + "learning_rate": 2.222995906797862e-06, + "loss": 0.9835, + "step": 64155 + }, + { + "epoch": 4.74, + "learning_rate": 2.2227451682886924e-06, + "loss": 1.0358, + "step": 64156 + }, + { + "epoch": 4.74, + "learning_rate": 2.222494442153027e-06, + "loss": 1.004, + "step": 64157 + }, + { + "epoch": 4.74, + "learning_rate": 2.22224372839126e-06, + "loss": 1.0008, + "step": 64158 + }, + { + "epoch": 4.74, + "learning_rate": 2.2219930270037925e-06, + "loss": 1.0235, + "step": 64159 + }, + { + "epoch": 4.74, + "learning_rate": 2.2217423379910243e-06, + "loss": 1.0142, + "step": 64160 + }, + { + "epoch": 4.74, + "learning_rate": 2.221491661353348e-06, + "loss": 0.9883, + "step": 64161 + }, + { + "epoch": 4.74, + "learning_rate": 2.2212409970911708e-06, + "loss": 1.0402, + "step": 64162 + }, + { + "epoch": 4.74, + "learning_rate": 2.2209903452048875e-06, + "loss": 0.9695, + "step": 64163 + }, + { + "epoch": 4.74, + "learning_rate": 2.2207397056948975e-06, + "loss": 1.0282, + "step": 64164 + }, + { + "epoch": 4.74, + "learning_rate": 2.2204890785615974e-06, + "loss": 0.9459, + "step": 64165 + }, + { + "epoch": 4.74, + "learning_rate": 2.2202384638053896e-06, + "loss": 0.9655, + "step": 64166 + }, + { + "epoch": 4.74, + "learning_rate": 2.219987861426671e-06, + "loss": 1.0238, + "step": 64167 + }, + { + "epoch": 4.74, + "learning_rate": 2.2197372714258415e-06, + "loss": 1.1025, + "step": 64168 + }, + { + "epoch": 4.74, + "learning_rate": 2.2194866938032977e-06, + "loss": 0.9862, + "step": 64169 + }, + { + "epoch": 4.74, + "learning_rate": 2.2192361285594365e-06, + "loss": 0.8743, + "step": 64170 + }, + { + "epoch": 4.74, + "learning_rate": 2.218985575694662e-06, + "loss": 1.0979, + "step": 64171 + }, + { + "epoch": 4.74, + "learning_rate": 2.2187350352093706e-06, + "loss": 0.9742, + "step": 64172 + }, + { + "epoch": 4.74, + "learning_rate": 2.218484507103961e-06, + "loss": 1.0058, + "step": 64173 + }, + { + "epoch": 4.74, + "learning_rate": 2.2182339913788265e-06, + "loss": 0.9941, + "step": 64174 + }, + { + "epoch": 4.74, + "learning_rate": 2.2179834880343777e-06, + "loss": 1.0594, + "step": 64175 + }, + { + "epoch": 4.74, + "learning_rate": 2.2177329970709993e-06, + "loss": 0.992, + "step": 64176 + }, + { + "epoch": 4.74, + "learning_rate": 2.217482518489099e-06, + "loss": 1.0156, + "step": 64177 + }, + { + "epoch": 4.74, + "learning_rate": 2.2172320522890732e-06, + "loss": 0.9885, + "step": 64178 + }, + { + "epoch": 4.74, + "learning_rate": 2.2169815984713205e-06, + "loss": 1.0834, + "step": 64179 + }, + { + "epoch": 4.74, + "learning_rate": 2.2167311570362383e-06, + "loss": 1.1155, + "step": 64180 + }, + { + "epoch": 4.74, + "learning_rate": 2.2164807279842217e-06, + "loss": 0.9752, + "step": 64181 + }, + { + "epoch": 4.74, + "learning_rate": 2.216230311315677e-06, + "loss": 1.0072, + "step": 64182 + }, + { + "epoch": 4.74, + "learning_rate": 2.2159799070309973e-06, + "loss": 0.9657, + "step": 64183 + }, + { + "epoch": 4.74, + "learning_rate": 2.2157295151305835e-06, + "loss": 0.9528, + "step": 64184 + }, + { + "epoch": 4.74, + "learning_rate": 2.2154791356148286e-06, + "loss": 1.0455, + "step": 64185 + }, + { + "epoch": 4.74, + "learning_rate": 2.2152287684841388e-06, + "loss": 0.9232, + "step": 64186 + }, + { + "epoch": 4.74, + "learning_rate": 2.2149784137389087e-06, + "loss": 0.8983, + "step": 64187 + }, + { + "epoch": 4.74, + "learning_rate": 2.2147280713795347e-06, + "loss": 1.0864, + "step": 64188 + }, + { + "epoch": 4.74, + "learning_rate": 2.214477741406419e-06, + "loss": 0.9986, + "step": 64189 + }, + { + "epoch": 4.74, + "learning_rate": 2.2142274238199545e-06, + "loss": 0.974, + "step": 64190 + }, + { + "epoch": 4.74, + "learning_rate": 2.213977118620545e-06, + "loss": 1.0291, + "step": 64191 + }, + { + "epoch": 4.74, + "learning_rate": 2.2137268258085877e-06, + "loss": 1.0188, + "step": 64192 + }, + { + "epoch": 4.74, + "learning_rate": 2.213476545384479e-06, + "loss": 1.0169, + "step": 64193 + }, + { + "epoch": 4.74, + "learning_rate": 2.213226277348615e-06, + "loss": 0.8982, + "step": 64194 + }, + { + "epoch": 4.74, + "learning_rate": 2.2129760217013983e-06, + "loss": 0.9447, + "step": 64195 + }, + { + "epoch": 4.74, + "learning_rate": 2.2127257784432288e-06, + "loss": 0.946, + "step": 64196 + }, + { + "epoch": 4.74, + "learning_rate": 2.212475547574496e-06, + "loss": 0.9883, + "step": 64197 + }, + { + "epoch": 4.74, + "learning_rate": 2.2122253290956064e-06, + "loss": 1.0851, + "step": 64198 + }, + { + "epoch": 4.74, + "learning_rate": 2.211975123006951e-06, + "loss": 1.0297, + "step": 64199 + }, + { + "epoch": 4.74, + "learning_rate": 2.211724929308937e-06, + "loss": 0.9809, + "step": 64200 + }, + { + "epoch": 4.74, + "learning_rate": 2.2114747480019526e-06, + "loss": 0.9452, + "step": 64201 + }, + { + "epoch": 4.74, + "learning_rate": 2.211224579086404e-06, + "loss": 0.9992, + "step": 64202 + }, + { + "epoch": 4.74, + "learning_rate": 2.2109744225626817e-06, + "loss": 0.9419, + "step": 64203 + }, + { + "epoch": 4.74, + "learning_rate": 2.210724278431193e-06, + "loss": 1.0739, + "step": 64204 + }, + { + "epoch": 4.74, + "learning_rate": 2.2104741466923286e-06, + "loss": 0.9859, + "step": 64205 + }, + { + "epoch": 4.74, + "learning_rate": 2.210224027346485e-06, + "loss": 1.106, + "step": 64206 + }, + { + "epoch": 4.74, + "learning_rate": 2.2099739203940663e-06, + "loss": 1.1008, + "step": 64207 + }, + { + "epoch": 4.74, + "learning_rate": 2.209723825835468e-06, + "loss": 1.0663, + "step": 64208 + }, + { + "epoch": 4.74, + "learning_rate": 2.2094737436710868e-06, + "loss": 1.1573, + "step": 64209 + }, + { + "epoch": 4.74, + "learning_rate": 2.2092236739013195e-06, + "loss": 0.9312, + "step": 64210 + }, + { + "epoch": 4.74, + "learning_rate": 2.2089736165265686e-06, + "loss": 0.9304, + "step": 64211 + }, + { + "epoch": 4.74, + "learning_rate": 2.208723571547229e-06, + "loss": 0.9071, + "step": 64212 + }, + { + "epoch": 4.74, + "learning_rate": 2.208473538963699e-06, + "loss": 0.9534, + "step": 64213 + }, + { + "epoch": 4.74, + "learning_rate": 2.2082235187763754e-06, + "loss": 1.0597, + "step": 64214 + }, + { + "epoch": 4.74, + "learning_rate": 2.2079735109856547e-06, + "loss": 0.9933, + "step": 64215 + }, + { + "epoch": 4.74, + "learning_rate": 2.2077235155919387e-06, + "loss": 1.0231, + "step": 64216 + }, + { + "epoch": 4.74, + "learning_rate": 2.207473532595623e-06, + "loss": 0.9602, + "step": 64217 + }, + { + "epoch": 4.75, + "learning_rate": 2.207223561997106e-06, + "loss": 1.0158, + "step": 64218 + }, + { + "epoch": 4.75, + "learning_rate": 2.2069736037967826e-06, + "loss": 1.0368, + "step": 64219 + }, + { + "epoch": 4.75, + "learning_rate": 2.206723657995058e-06, + "loss": 1.0386, + "step": 64220 + }, + { + "epoch": 4.75, + "learning_rate": 2.2064737245923186e-06, + "loss": 1.0431, + "step": 64221 + }, + { + "epoch": 4.75, + "learning_rate": 2.2062238035889706e-06, + "loss": 1.0324, + "step": 64222 + }, + { + "epoch": 4.75, + "learning_rate": 2.2059738949854094e-06, + "loss": 1.0059, + "step": 64223 + }, + { + "epoch": 4.75, + "learning_rate": 2.205723998782031e-06, + "loss": 1.0163, + "step": 64224 + }, + { + "epoch": 4.75, + "learning_rate": 2.205474114979236e-06, + "loss": 1.1112, + "step": 64225 + }, + { + "epoch": 4.75, + "learning_rate": 2.2052242435774165e-06, + "loss": 0.9633, + "step": 64226 + }, + { + "epoch": 4.75, + "learning_rate": 2.2049743845769767e-06, + "loss": 1.0122, + "step": 64227 + }, + { + "epoch": 4.75, + "learning_rate": 2.204724537978311e-06, + "loss": 1.0947, + "step": 64228 + }, + { + "epoch": 4.75, + "learning_rate": 2.204474703781817e-06, + "loss": 0.9994, + "step": 64229 + }, + { + "epoch": 4.75, + "learning_rate": 2.2042248819878885e-06, + "loss": 1.0686, + "step": 64230 + }, + { + "epoch": 4.75, + "learning_rate": 2.203975072596931e-06, + "loss": 0.925, + "step": 64231 + }, + { + "epoch": 4.75, + "learning_rate": 2.2037252756093374e-06, + "loss": 1.0463, + "step": 64232 + }, + { + "epoch": 4.75, + "learning_rate": 2.2034754910255053e-06, + "loss": 0.9701, + "step": 64233 + }, + { + "epoch": 4.75, + "learning_rate": 2.2032257188458317e-06, + "loss": 1.015, + "step": 64234 + }, + { + "epoch": 4.75, + "learning_rate": 2.2029759590707112e-06, + "loss": 1.0191, + "step": 64235 + }, + { + "epoch": 4.75, + "learning_rate": 2.2027262117005475e-06, + "loss": 1.0503, + "step": 64236 + }, + { + "epoch": 4.75, + "learning_rate": 2.2024764767357352e-06, + "loss": 0.9088, + "step": 64237 + }, + { + "epoch": 4.75, + "learning_rate": 2.2022267541766717e-06, + "loss": 1.0234, + "step": 64238 + }, + { + "epoch": 4.75, + "learning_rate": 2.20197704402375e-06, + "loss": 1.0075, + "step": 64239 + }, + { + "epoch": 4.75, + "learning_rate": 2.201727346277375e-06, + "loss": 0.9453, + "step": 64240 + }, + { + "epoch": 4.75, + "learning_rate": 2.2014776609379398e-06, + "loss": 1.0442, + "step": 64241 + }, + { + "epoch": 4.75, + "learning_rate": 2.2012279880058416e-06, + "loss": 0.9392, + "step": 64242 + }, + { + "epoch": 4.75, + "learning_rate": 2.200978327481479e-06, + "loss": 0.9289, + "step": 64243 + }, + { + "epoch": 4.75, + "learning_rate": 2.2007286793652447e-06, + "loss": 1.0388, + "step": 64244 + }, + { + "epoch": 4.75, + "learning_rate": 2.200479043657545e-06, + "loss": 0.9723, + "step": 64245 + }, + { + "epoch": 4.75, + "learning_rate": 2.2002294203587673e-06, + "loss": 1.0232, + "step": 64246 + }, + { + "epoch": 4.75, + "learning_rate": 2.1999798094693147e-06, + "loss": 1.0662, + "step": 64247 + }, + { + "epoch": 4.75, + "learning_rate": 2.1997302109895793e-06, + "loss": 1.0138, + "step": 64248 + }, + { + "epoch": 4.75, + "learning_rate": 2.1994806249199683e-06, + "loss": 0.9594, + "step": 64249 + }, + { + "epoch": 4.75, + "learning_rate": 2.199231051260866e-06, + "loss": 1.0334, + "step": 64250 + }, + { + "epoch": 4.75, + "learning_rate": 2.1989814900126784e-06, + "loss": 0.9509, + "step": 64251 + }, + { + "epoch": 4.75, + "learning_rate": 2.198731941175799e-06, + "loss": 1.0645, + "step": 64252 + }, + { + "epoch": 4.75, + "learning_rate": 2.198482404750626e-06, + "loss": 1.0501, + "step": 64253 + }, + { + "epoch": 4.75, + "learning_rate": 2.1982328807375555e-06, + "loss": 0.9429, + "step": 64254 + }, + { + "epoch": 4.75, + "learning_rate": 2.1979833691369823e-06, + "loss": 0.8576, + "step": 64255 + }, + { + "epoch": 4.75, + "learning_rate": 2.197733869949309e-06, + "loss": 1.0318, + "step": 64256 + }, + { + "epoch": 4.75, + "learning_rate": 2.197484383174929e-06, + "loss": 0.848, + "step": 64257 + }, + { + "epoch": 4.75, + "learning_rate": 2.1972349088142395e-06, + "loss": 1.0229, + "step": 64258 + }, + { + "epoch": 4.75, + "learning_rate": 2.1969854468676356e-06, + "loss": 1.0972, + "step": 64259 + }, + { + "epoch": 4.75, + "learning_rate": 2.1967359973355196e-06, + "loss": 1.0252, + "step": 64260 + }, + { + "epoch": 4.75, + "learning_rate": 2.196486560218284e-06, + "loss": 1.1176, + "step": 64261 + }, + { + "epoch": 4.75, + "learning_rate": 2.196237135516326e-06, + "loss": 0.9895, + "step": 64262 + }, + { + "epoch": 4.75, + "learning_rate": 2.1959877232300443e-06, + "loss": 1.0336, + "step": 64263 + }, + { + "epoch": 4.75, + "learning_rate": 2.195738323359831e-06, + "loss": 0.965, + "step": 64264 + }, + { + "epoch": 4.75, + "learning_rate": 2.1954889359060917e-06, + "loss": 1.077, + "step": 64265 + }, + { + "epoch": 4.75, + "learning_rate": 2.1952395608692135e-06, + "loss": 1.1472, + "step": 64266 + }, + { + "epoch": 4.75, + "learning_rate": 2.194990198249599e-06, + "loss": 1.0357, + "step": 64267 + }, + { + "epoch": 4.75, + "learning_rate": 2.194740848047641e-06, + "loss": 0.9766, + "step": 64268 + }, + { + "epoch": 4.75, + "learning_rate": 2.1944915102637443e-06, + "loss": 1.0406, + "step": 64269 + }, + { + "epoch": 4.75, + "learning_rate": 2.1942421848982977e-06, + "loss": 1.042, + "step": 64270 + }, + { + "epoch": 4.75, + "learning_rate": 2.1939928719516967e-06, + "loss": 0.8828, + "step": 64271 + }, + { + "epoch": 4.75, + "learning_rate": 2.1937435714243436e-06, + "loss": 0.9837, + "step": 64272 + }, + { + "epoch": 4.75, + "learning_rate": 2.193494283316634e-06, + "loss": 0.9013, + "step": 64273 + }, + { + "epoch": 4.75, + "learning_rate": 2.1932450076289625e-06, + "loss": 1.1047, + "step": 64274 + }, + { + "epoch": 4.75, + "learning_rate": 2.1929957443617235e-06, + "loss": 0.9144, + "step": 64275 + }, + { + "epoch": 4.75, + "learning_rate": 2.1927464935153198e-06, + "loss": 1.0632, + "step": 64276 + }, + { + "epoch": 4.75, + "learning_rate": 2.1924972550901447e-06, + "loss": 0.9751, + "step": 64277 + }, + { + "epoch": 4.75, + "learning_rate": 2.1922480290865943e-06, + "loss": 0.948, + "step": 64278 + }, + { + "epoch": 4.75, + "learning_rate": 2.191998815505064e-06, + "loss": 1.0128, + "step": 64279 + }, + { + "epoch": 4.75, + "learning_rate": 2.191749614345954e-06, + "loss": 1.0867, + "step": 64280 + }, + { + "epoch": 4.75, + "learning_rate": 2.191500425609658e-06, + "loss": 0.9678, + "step": 64281 + }, + { + "epoch": 4.75, + "learning_rate": 2.191251249296573e-06, + "loss": 1.1762, + "step": 64282 + }, + { + "epoch": 4.75, + "learning_rate": 2.191002085407097e-06, + "loss": 0.995, + "step": 64283 + }, + { + "epoch": 4.75, + "learning_rate": 2.1907529339416212e-06, + "loss": 1.0349, + "step": 64284 + }, + { + "epoch": 4.75, + "learning_rate": 2.1905037949005494e-06, + "loss": 1.0024, + "step": 64285 + }, + { + "epoch": 4.75, + "learning_rate": 2.190254668284273e-06, + "loss": 1.0978, + "step": 64286 + }, + { + "epoch": 4.75, + "learning_rate": 2.1900055540931895e-06, + "loss": 1.0421, + "step": 64287 + }, + { + "epoch": 4.75, + "learning_rate": 2.189756452327694e-06, + "loss": 1.0398, + "step": 64288 + }, + { + "epoch": 4.75, + "learning_rate": 2.189507362988187e-06, + "loss": 1.0744, + "step": 64289 + }, + { + "epoch": 4.75, + "learning_rate": 2.1892582860750632e-06, + "loss": 0.9795, + "step": 64290 + }, + { + "epoch": 4.75, + "learning_rate": 2.1890092215887138e-06, + "loss": 1.0712, + "step": 64291 + }, + { + "epoch": 4.75, + "learning_rate": 2.1887601695295412e-06, + "loss": 0.9958, + "step": 64292 + }, + { + "epoch": 4.75, + "learning_rate": 2.1885111298979366e-06, + "loss": 1.0233, + "step": 64293 + }, + { + "epoch": 4.75, + "learning_rate": 2.1882621026943043e-06, + "loss": 0.9794, + "step": 64294 + }, + { + "epoch": 4.75, + "learning_rate": 2.1880130879190297e-06, + "loss": 0.9397, + "step": 64295 + }, + { + "epoch": 4.75, + "learning_rate": 2.187764085572518e-06, + "loss": 0.9995, + "step": 64296 + }, + { + "epoch": 4.75, + "learning_rate": 2.187515095655162e-06, + "loss": 0.9554, + "step": 64297 + }, + { + "epoch": 4.75, + "learning_rate": 2.187266118167356e-06, + "loss": 0.9646, + "step": 64298 + }, + { + "epoch": 4.75, + "learning_rate": 2.1870171531094996e-06, + "loss": 1.1035, + "step": 64299 + }, + { + "epoch": 4.75, + "learning_rate": 2.1867682004819836e-06, + "loss": 1.0014, + "step": 64300 + }, + { + "epoch": 4.75, + "learning_rate": 2.1865192602852105e-06, + "loss": 0.9119, + "step": 64301 + }, + { + "epoch": 4.75, + "learning_rate": 2.1862703325195735e-06, + "loss": 0.9818, + "step": 64302 + }, + { + "epoch": 4.75, + "learning_rate": 2.186021417185469e-06, + "loss": 1.0602, + "step": 64303 + }, + { + "epoch": 4.75, + "learning_rate": 2.185772514283289e-06, + "loss": 1.0247, + "step": 64304 + }, + { + "epoch": 4.75, + "learning_rate": 2.1855236238134357e-06, + "loss": 1.0001, + "step": 64305 + }, + { + "epoch": 4.75, + "learning_rate": 2.1852747457763034e-06, + "loss": 0.9783, + "step": 64306 + }, + { + "epoch": 4.75, + "learning_rate": 2.1850258801722866e-06, + "loss": 1.0348, + "step": 64307 + }, + { + "epoch": 4.75, + "learning_rate": 2.184777027001782e-06, + "loss": 0.9445, + "step": 64308 + }, + { + "epoch": 4.75, + "learning_rate": 2.1845281862651823e-06, + "loss": 0.9515, + "step": 64309 + }, + { + "epoch": 4.75, + "learning_rate": 2.184279357962893e-06, + "loss": 1.1034, + "step": 64310 + }, + { + "epoch": 4.75, + "learning_rate": 2.184030542095297e-06, + "loss": 1.0276, + "step": 64311 + }, + { + "epoch": 4.75, + "learning_rate": 2.1837817386627992e-06, + "loss": 0.878, + "step": 64312 + }, + { + "epoch": 4.75, + "learning_rate": 2.183532947665791e-06, + "loss": 0.8817, + "step": 64313 + }, + { + "epoch": 4.75, + "learning_rate": 2.1832841691046746e-06, + "loss": 0.9904, + "step": 64314 + }, + { + "epoch": 4.75, + "learning_rate": 2.1830354029798363e-06, + "loss": 1.056, + "step": 64315 + }, + { + "epoch": 4.75, + "learning_rate": 2.1827866492916794e-06, + "loss": 1.0925, + "step": 64316 + }, + { + "epoch": 4.75, + "learning_rate": 2.182537908040597e-06, + "loss": 1.0773, + "step": 64317 + }, + { + "epoch": 4.75, + "learning_rate": 2.182289179226985e-06, + "loss": 0.9175, + "step": 64318 + }, + { + "epoch": 4.75, + "learning_rate": 2.1820404628512394e-06, + "loss": 1.0292, + "step": 64319 + }, + { + "epoch": 4.75, + "learning_rate": 2.1817917589137516e-06, + "loss": 0.9221, + "step": 64320 + }, + { + "epoch": 4.75, + "learning_rate": 2.1815430674149253e-06, + "loss": 1.0451, + "step": 64321 + }, + { + "epoch": 4.75, + "learning_rate": 2.1812943883551518e-06, + "loss": 0.9881, + "step": 64322 + }, + { + "epoch": 4.75, + "learning_rate": 2.181045721734827e-06, + "loss": 1.0615, + "step": 64323 + }, + { + "epoch": 4.75, + "learning_rate": 2.1807970675543432e-06, + "loss": 1.0852, + "step": 64324 + }, + { + "epoch": 4.75, + "learning_rate": 2.1805484258141028e-06, + "loss": 0.9602, + "step": 64325 + }, + { + "epoch": 4.75, + "learning_rate": 2.1802997965144978e-06, + "loss": 1.029, + "step": 64326 + }, + { + "epoch": 4.75, + "learning_rate": 2.1800511796559242e-06, + "loss": 1.0087, + "step": 64327 + }, + { + "epoch": 4.75, + "learning_rate": 2.1798025752387773e-06, + "loss": 0.9735, + "step": 64328 + }, + { + "epoch": 4.75, + "learning_rate": 2.179553983263449e-06, + "loss": 1.066, + "step": 64329 + }, + { + "epoch": 4.75, + "learning_rate": 2.179305403730343e-06, + "loss": 0.9759, + "step": 64330 + }, + { + "epoch": 4.75, + "learning_rate": 2.179056836639849e-06, + "loss": 0.9887, + "step": 64331 + }, + { + "epoch": 4.75, + "learning_rate": 2.178808281992364e-06, + "loss": 1.0534, + "step": 64332 + }, + { + "epoch": 4.75, + "learning_rate": 2.1785597397882797e-06, + "loss": 0.9832, + "step": 64333 + }, + { + "epoch": 4.75, + "learning_rate": 2.178311210027998e-06, + "loss": 1.0599, + "step": 64334 + }, + { + "epoch": 4.75, + "learning_rate": 2.178062692711912e-06, + "loss": 1.0115, + "step": 64335 + }, + { + "epoch": 4.75, + "learning_rate": 2.177814187840416e-06, + "loss": 0.9861, + "step": 64336 + }, + { + "epoch": 4.75, + "learning_rate": 2.1775656954139046e-06, + "loss": 1.0391, + "step": 64337 + }, + { + "epoch": 4.75, + "learning_rate": 2.1773172154327726e-06, + "loss": 1.0265, + "step": 64338 + }, + { + "epoch": 4.75, + "learning_rate": 2.1770687478974218e-06, + "loss": 0.9751, + "step": 64339 + }, + { + "epoch": 4.75, + "learning_rate": 2.1768202928082384e-06, + "loss": 0.9399, + "step": 64340 + }, + { + "epoch": 4.75, + "learning_rate": 2.176571850165623e-06, + "loss": 1.0905, + "step": 64341 + }, + { + "epoch": 4.75, + "learning_rate": 2.176323419969971e-06, + "loss": 0.9934, + "step": 64342 + }, + { + "epoch": 4.75, + "learning_rate": 2.176075002221677e-06, + "loss": 0.9235, + "step": 64343 + }, + { + "epoch": 4.75, + "learning_rate": 2.175826596921132e-06, + "loss": 0.992, + "step": 64344 + }, + { + "epoch": 4.75, + "learning_rate": 2.1755782040687377e-06, + "loss": 0.9798, + "step": 64345 + }, + { + "epoch": 4.75, + "learning_rate": 2.1753298236648866e-06, + "loss": 0.9415, + "step": 64346 + }, + { + "epoch": 4.75, + "learning_rate": 2.175081455709974e-06, + "loss": 1.0459, + "step": 64347 + }, + { + "epoch": 4.75, + "learning_rate": 2.1748331002043955e-06, + "loss": 0.9583, + "step": 64348 + }, + { + "epoch": 4.75, + "learning_rate": 2.1745847571485425e-06, + "loss": 1.113, + "step": 64349 + }, + { + "epoch": 4.75, + "learning_rate": 2.1743364265428146e-06, + "loss": 1.0406, + "step": 64350 + }, + { + "epoch": 4.75, + "learning_rate": 2.1740881083876062e-06, + "loss": 1.0657, + "step": 64351 + }, + { + "epoch": 4.75, + "learning_rate": 2.173839802683313e-06, + "loss": 1.0292, + "step": 64352 + }, + { + "epoch": 4.76, + "learning_rate": 2.173591509430324e-06, + "loss": 0.9155, + "step": 64353 + }, + { + "epoch": 4.76, + "learning_rate": 2.1733432286290436e-06, + "loss": 1.0104, + "step": 64354 + }, + { + "epoch": 4.76, + "learning_rate": 2.1730949602798634e-06, + "loss": 0.9807, + "step": 64355 + }, + { + "epoch": 4.76, + "learning_rate": 2.1728467043831725e-06, + "loss": 1.0174, + "step": 64356 + }, + { + "epoch": 4.76, + "learning_rate": 2.1725984609393723e-06, + "loss": 1.0413, + "step": 64357 + }, + { + "epoch": 4.76, + "learning_rate": 2.1723502299488543e-06, + "loss": 0.9219, + "step": 64358 + }, + { + "epoch": 4.76, + "learning_rate": 2.1721020114120194e-06, + "loss": 1.0019, + "step": 64359 + }, + { + "epoch": 4.76, + "learning_rate": 2.1718538053292535e-06, + "loss": 1.0665, + "step": 64360 + }, + { + "epoch": 4.76, + "learning_rate": 2.1716056117009585e-06, + "loss": 1.102, + "step": 64361 + }, + { + "epoch": 4.76, + "learning_rate": 2.1713574305275254e-06, + "loss": 0.9945, + "step": 64362 + }, + { + "epoch": 4.76, + "learning_rate": 2.1711092618093543e-06, + "loss": 0.908, + "step": 64363 + }, + { + "epoch": 4.76, + "learning_rate": 2.1708611055468352e-06, + "loss": 1.0522, + "step": 64364 + }, + { + "epoch": 4.76, + "learning_rate": 2.1706129617403605e-06, + "loss": 1.0965, + "step": 64365 + }, + { + "epoch": 4.76, + "learning_rate": 2.170364830390331e-06, + "loss": 0.8794, + "step": 64366 + }, + { + "epoch": 4.76, + "learning_rate": 2.17011671149714e-06, + "loss": 0.9937, + "step": 64367 + }, + { + "epoch": 4.76, + "learning_rate": 2.1698686050611805e-06, + "loss": 1.002, + "step": 64368 + }, + { + "epoch": 4.76, + "learning_rate": 2.169620511082845e-06, + "loss": 0.912, + "step": 64369 + }, + { + "epoch": 4.76, + "learning_rate": 2.169372429562535e-06, + "loss": 0.9827, + "step": 64370 + }, + { + "epoch": 4.76, + "learning_rate": 2.1691243605006405e-06, + "loss": 1.056, + "step": 64371 + }, + { + "epoch": 4.76, + "learning_rate": 2.168876303897558e-06, + "loss": 1.0204, + "step": 64372 + }, + { + "epoch": 4.76, + "learning_rate": 2.168628259753681e-06, + "loss": 0.9111, + "step": 64373 + }, + { + "epoch": 4.76, + "learning_rate": 2.168380228069401e-06, + "loss": 0.9797, + "step": 64374 + }, + { + "epoch": 4.76, + "learning_rate": 2.1681322088451195e-06, + "loss": 0.9171, + "step": 64375 + }, + { + "epoch": 4.76, + "learning_rate": 2.1678842020812276e-06, + "loss": 1.0055, + "step": 64376 + }, + { + "epoch": 4.76, + "learning_rate": 2.1676362077781187e-06, + "loss": 1.0083, + "step": 64377 + }, + { + "epoch": 4.76, + "learning_rate": 2.1673882259361867e-06, + "loss": 0.9651, + "step": 64378 + }, + { + "epoch": 4.76, + "learning_rate": 2.167140256555831e-06, + "loss": 0.844, + "step": 64379 + }, + { + "epoch": 4.76, + "learning_rate": 2.166892299637443e-06, + "loss": 1.0504, + "step": 64380 + }, + { + "epoch": 4.76, + "learning_rate": 2.1666443551814165e-06, + "loss": 0.9928, + "step": 64381 + }, + { + "epoch": 4.76, + "learning_rate": 2.166396423188147e-06, + "loss": 0.953, + "step": 64382 + }, + { + "epoch": 4.76, + "learning_rate": 2.1661485036580266e-06, + "loss": 0.971, + "step": 64383 + }, + { + "epoch": 4.76, + "learning_rate": 2.165900596591457e-06, + "loss": 0.9718, + "step": 64384 + }, + { + "epoch": 4.76, + "learning_rate": 2.1656527019888217e-06, + "loss": 0.9511, + "step": 64385 + }, + { + "epoch": 4.76, + "learning_rate": 2.165404819850524e-06, + "loss": 1.0557, + "step": 64386 + }, + { + "epoch": 4.76, + "learning_rate": 2.1651569501769553e-06, + "loss": 0.9729, + "step": 64387 + }, + { + "epoch": 4.76, + "learning_rate": 2.1649090929685104e-06, + "loss": 0.9266, + "step": 64388 + }, + { + "epoch": 4.76, + "learning_rate": 2.164661248225579e-06, + "loss": 0.9786, + "step": 64389 + }, + { + "epoch": 4.76, + "learning_rate": 2.1644134159485617e-06, + "loss": 0.9834, + "step": 64390 + }, + { + "epoch": 4.76, + "learning_rate": 2.1641655961378517e-06, + "loss": 0.9948, + "step": 64391 + }, + { + "epoch": 4.76, + "learning_rate": 2.163917788793842e-06, + "loss": 1.0163, + "step": 64392 + }, + { + "epoch": 4.76, + "learning_rate": 2.163669993916927e-06, + "loss": 0.9576, + "step": 64393 + }, + { + "epoch": 4.76, + "learning_rate": 2.1634222115074978e-06, + "loss": 0.8873, + "step": 64394 + }, + { + "epoch": 4.76, + "learning_rate": 2.163174441565955e-06, + "loss": 0.9806, + "step": 64395 + }, + { + "epoch": 4.76, + "learning_rate": 2.162926684092689e-06, + "loss": 1.0495, + "step": 64396 + }, + { + "epoch": 4.76, + "learning_rate": 2.1626789390880952e-06, + "loss": 0.919, + "step": 64397 + }, + { + "epoch": 4.76, + "learning_rate": 2.1624312065525634e-06, + "loss": 1.1017, + "step": 64398 + }, + { + "epoch": 4.76, + "learning_rate": 2.1621834864864945e-06, + "loss": 1.0117, + "step": 64399 + }, + { + "epoch": 4.76, + "learning_rate": 2.1619357788902795e-06, + "loss": 1.0712, + "step": 64400 + }, + { + "epoch": 4.76, + "learning_rate": 2.1616880837643126e-06, + "loss": 1.0135, + "step": 64401 + }, + { + "epoch": 4.76, + "learning_rate": 2.1614404011089882e-06, + "loss": 0.9956, + "step": 64402 + }, + { + "epoch": 4.76, + "learning_rate": 2.1611927309246972e-06, + "loss": 0.9737, + "step": 64403 + }, + { + "epoch": 4.76, + "learning_rate": 2.160945073211842e-06, + "loss": 1.0169, + "step": 64404 + }, + { + "epoch": 4.76, + "learning_rate": 2.160697427970805e-06, + "loss": 0.9885, + "step": 64405 + }, + { + "epoch": 4.76, + "learning_rate": 2.1604497952019886e-06, + "loss": 0.9491, + "step": 64406 + }, + { + "epoch": 4.76, + "learning_rate": 2.160202174905782e-06, + "loss": 0.9712, + "step": 64407 + }, + { + "epoch": 4.76, + "learning_rate": 2.159954567082587e-06, + "loss": 1.0964, + "step": 64408 + }, + { + "epoch": 4.76, + "learning_rate": 2.159706971732788e-06, + "loss": 0.93, + "step": 64409 + }, + { + "epoch": 4.76, + "learning_rate": 2.159459388856784e-06, + "loss": 0.966, + "step": 64410 + }, + { + "epoch": 4.76, + "learning_rate": 2.1592118184549692e-06, + "loss": 1.0652, + "step": 64411 + }, + { + "epoch": 4.76, + "learning_rate": 2.158964260527735e-06, + "loss": 0.9933, + "step": 64412 + }, + { + "epoch": 4.76, + "learning_rate": 2.1587167150754774e-06, + "loss": 0.9558, + "step": 64413 + }, + { + "epoch": 4.76, + "learning_rate": 2.1584691820985857e-06, + "loss": 1.0845, + "step": 64414 + }, + { + "epoch": 4.76, + "learning_rate": 2.1582216615974615e-06, + "loss": 1.026, + "step": 64415 + }, + { + "epoch": 4.76, + "learning_rate": 2.157974153572493e-06, + "loss": 0.9992, + "step": 64416 + }, + { + "epoch": 4.76, + "learning_rate": 2.1577266580240765e-06, + "loss": 0.9741, + "step": 64417 + }, + { + "epoch": 4.76, + "learning_rate": 2.1574791749526014e-06, + "loss": 1.0563, + "step": 64418 + }, + { + "epoch": 4.76, + "learning_rate": 2.1572317043584666e-06, + "loss": 0.9805, + "step": 64419 + }, + { + "epoch": 4.76, + "learning_rate": 2.156984246242065e-06, + "loss": 0.9368, + "step": 64420 + }, + { + "epoch": 4.76, + "learning_rate": 2.1567368006037893e-06, + "loss": 0.9709, + "step": 64421 + }, + { + "epoch": 4.76, + "learning_rate": 2.1564893674440334e-06, + "loss": 1.0974, + "step": 64422 + }, + { + "epoch": 4.76, + "learning_rate": 2.156241946763187e-06, + "loss": 0.931, + "step": 64423 + }, + { + "epoch": 4.76, + "learning_rate": 2.155994538561651e-06, + "loss": 1.0783, + "step": 64424 + }, + { + "epoch": 4.76, + "learning_rate": 2.1557471428398146e-06, + "loss": 0.9829, + "step": 64425 + }, + { + "epoch": 4.76, + "learning_rate": 2.155499759598073e-06, + "loss": 1.0705, + "step": 64426 + }, + { + "epoch": 4.76, + "learning_rate": 2.155252388836816e-06, + "loss": 0.957, + "step": 64427 + }, + { + "epoch": 4.76, + "learning_rate": 2.155005030556443e-06, + "loss": 0.9941, + "step": 64428 + }, + { + "epoch": 4.76, + "learning_rate": 2.154757684757348e-06, + "loss": 1.0146, + "step": 64429 + }, + { + "epoch": 4.76, + "learning_rate": 2.1545103514399157e-06, + "loss": 1.0276, + "step": 64430 + }, + { + "epoch": 4.76, + "learning_rate": 2.1542630306045475e-06, + "loss": 1.0231, + "step": 64431 + }, + { + "epoch": 4.76, + "learning_rate": 2.1540157222516345e-06, + "loss": 1.0522, + "step": 64432 + }, + { + "epoch": 4.76, + "learning_rate": 2.1537684263815706e-06, + "loss": 1.0009, + "step": 64433 + }, + { + "epoch": 4.76, + "learning_rate": 2.1535211429947454e-06, + "loss": 0.8956, + "step": 64434 + }, + { + "epoch": 4.76, + "learning_rate": 2.153273872091559e-06, + "loss": 1.1022, + "step": 64435 + }, + { + "epoch": 4.76, + "learning_rate": 2.153026613672402e-06, + "loss": 1.0341, + "step": 64436 + }, + { + "epoch": 4.76, + "learning_rate": 2.1527793677376663e-06, + "loss": 1.0759, + "step": 64437 + }, + { + "epoch": 4.76, + "learning_rate": 2.152532134287747e-06, + "loss": 0.9729, + "step": 64438 + }, + { + "epoch": 4.76, + "learning_rate": 2.1522849133230337e-06, + "loss": 0.9668, + "step": 64439 + }, + { + "epoch": 4.76, + "learning_rate": 2.1520377048439267e-06, + "loss": 0.8729, + "step": 64440 + }, + { + "epoch": 4.76, + "learning_rate": 2.151790508850815e-06, + "loss": 1.037, + "step": 64441 + }, + { + "epoch": 4.76, + "learning_rate": 2.151543325344092e-06, + "loss": 0.9102, + "step": 64442 + }, + { + "epoch": 4.76, + "learning_rate": 2.1512961543241484e-06, + "loss": 1.061, + "step": 64443 + }, + { + "epoch": 4.76, + "learning_rate": 2.1510489957913827e-06, + "loss": 1.0645, + "step": 64444 + }, + { + "epoch": 4.76, + "learning_rate": 2.150801849746187e-06, + "loss": 1.0822, + "step": 64445 + }, + { + "epoch": 4.76, + "learning_rate": 2.1505547161889517e-06, + "loss": 0.9981, + "step": 64446 + }, + { + "epoch": 4.76, + "learning_rate": 2.1503075951200704e-06, + "loss": 0.9756, + "step": 64447 + }, + { + "epoch": 4.76, + "learning_rate": 2.150060486539939e-06, + "loss": 0.9789, + "step": 64448 + }, + { + "epoch": 4.76, + "learning_rate": 2.1498133904489514e-06, + "loss": 1.1737, + "step": 64449 + }, + { + "epoch": 4.76, + "learning_rate": 2.1495663068474947e-06, + "loss": 1.0319, + "step": 64450 + }, + { + "epoch": 4.76, + "learning_rate": 2.149319235735967e-06, + "loss": 0.9377, + "step": 64451 + }, + { + "epoch": 4.76, + "learning_rate": 2.149072177114757e-06, + "loss": 0.9171, + "step": 64452 + }, + { + "epoch": 4.76, + "learning_rate": 2.148825130984268e-06, + "loss": 0.9724, + "step": 64453 + }, + { + "epoch": 4.76, + "learning_rate": 2.148578097344879e-06, + "loss": 0.9623, + "step": 64454 + }, + { + "epoch": 4.76, + "learning_rate": 2.1483310761969935e-06, + "loss": 0.9586, + "step": 64455 + }, + { + "epoch": 4.76, + "learning_rate": 2.1480840675409975e-06, + "loss": 0.9601, + "step": 64456 + }, + { + "epoch": 4.76, + "learning_rate": 2.1478370713772934e-06, + "loss": 0.9516, + "step": 64457 + }, + { + "epoch": 4.76, + "learning_rate": 2.147590087706265e-06, + "loss": 1.0071, + "step": 64458 + }, + { + "epoch": 4.76, + "learning_rate": 2.1473431165283053e-06, + "loss": 1.0228, + "step": 64459 + }, + { + "epoch": 4.76, + "learning_rate": 2.1470961578438143e-06, + "loss": 0.9934, + "step": 64460 + }, + { + "epoch": 4.76, + "learning_rate": 2.1468492116531804e-06, + "loss": 0.9766, + "step": 64461 + }, + { + "epoch": 4.76, + "learning_rate": 2.146602277956796e-06, + "loss": 0.9381, + "step": 64462 + }, + { + "epoch": 4.76, + "learning_rate": 2.1463553567550534e-06, + "loss": 1.0277, + "step": 64463 + }, + { + "epoch": 4.76, + "learning_rate": 2.1461084480483497e-06, + "loss": 0.9612, + "step": 64464 + }, + { + "epoch": 4.76, + "learning_rate": 2.1458615518370753e-06, + "loss": 0.9744, + "step": 64465 + }, + { + "epoch": 4.76, + "learning_rate": 2.1456146681216226e-06, + "loss": 0.8869, + "step": 64466 + }, + { + "epoch": 4.76, + "learning_rate": 2.1453677969023843e-06, + "loss": 0.8208, + "step": 64467 + }, + { + "epoch": 4.76, + "learning_rate": 2.145120938179751e-06, + "loss": 0.9729, + "step": 64468 + }, + { + "epoch": 4.76, + "learning_rate": 2.14487409195412e-06, + "loss": 1.0072, + "step": 64469 + }, + { + "epoch": 4.76, + "learning_rate": 2.144627258225883e-06, + "loss": 1.0071, + "step": 64470 + }, + { + "epoch": 4.76, + "learning_rate": 2.1443804369954314e-06, + "loss": 0.9828, + "step": 64471 + }, + { + "epoch": 4.76, + "learning_rate": 2.144133628263154e-06, + "loss": 1.0108, + "step": 64472 + }, + { + "epoch": 4.76, + "learning_rate": 2.1438868320294526e-06, + "loss": 1.0145, + "step": 64473 + }, + { + "epoch": 4.76, + "learning_rate": 2.143640048294713e-06, + "loss": 1.0405, + "step": 64474 + }, + { + "epoch": 4.76, + "learning_rate": 2.14339327705933e-06, + "loss": 0.989, + "step": 64475 + }, + { + "epoch": 4.76, + "learning_rate": 2.1431465183236964e-06, + "loss": 1.0069, + "step": 64476 + }, + { + "epoch": 4.76, + "learning_rate": 2.142899772088205e-06, + "loss": 0.9956, + "step": 64477 + }, + { + "epoch": 4.76, + "learning_rate": 2.142653038353246e-06, + "loss": 1.0139, + "step": 64478 + }, + { + "epoch": 4.76, + "learning_rate": 2.1424063171192122e-06, + "loss": 0.9837, + "step": 64479 + }, + { + "epoch": 4.76, + "learning_rate": 2.1421596083864995e-06, + "loss": 0.9473, + "step": 64480 + }, + { + "epoch": 4.76, + "learning_rate": 2.1419129121554983e-06, + "loss": 1.0184, + "step": 64481 + }, + { + "epoch": 4.76, + "learning_rate": 2.1416662284266023e-06, + "loss": 1.0252, + "step": 64482 + }, + { + "epoch": 4.76, + "learning_rate": 2.1414195572001993e-06, + "loss": 0.9891, + "step": 64483 + }, + { + "epoch": 4.76, + "learning_rate": 2.1411728984766878e-06, + "loss": 0.9556, + "step": 64484 + }, + { + "epoch": 4.76, + "learning_rate": 2.140926252256458e-06, + "loss": 1.0487, + "step": 64485 + }, + { + "epoch": 4.76, + "learning_rate": 2.1406796185399027e-06, + "loss": 1.0352, + "step": 64486 + }, + { + "epoch": 4.76, + "learning_rate": 2.140432997327413e-06, + "loss": 1.0674, + "step": 64487 + }, + { + "epoch": 4.76, + "learning_rate": 2.1401863886193797e-06, + "loss": 1.1295, + "step": 64488 + }, + { + "epoch": 4.77, + "learning_rate": 2.1399397924162e-06, + "loss": 0.9746, + "step": 64489 + }, + { + "epoch": 4.77, + "learning_rate": 2.139693208718263e-06, + "loss": 1.0454, + "step": 64490 + }, + { + "epoch": 4.77, + "learning_rate": 2.1394466375259624e-06, + "loss": 0.9702, + "step": 64491 + }, + { + "epoch": 4.77, + "learning_rate": 2.1392000788396873e-06, + "loss": 1.1522, + "step": 64492 + }, + { + "epoch": 4.77, + "learning_rate": 2.138953532659834e-06, + "loss": 1.0089, + "step": 64493 + }, + { + "epoch": 4.77, + "learning_rate": 2.1387069989867935e-06, + "loss": 0.8992, + "step": 64494 + }, + { + "epoch": 4.77, + "learning_rate": 2.1384604778209583e-06, + "loss": 0.9182, + "step": 64495 + }, + { + "epoch": 4.77, + "learning_rate": 2.138213969162719e-06, + "loss": 1.0138, + "step": 64496 + }, + { + "epoch": 4.77, + "learning_rate": 2.1379674730124677e-06, + "loss": 1.0811, + "step": 64497 + }, + { + "epoch": 4.77, + "learning_rate": 2.137720989370602e-06, + "loss": 1.0228, + "step": 64498 + }, + { + "epoch": 4.77, + "learning_rate": 2.1374745182375046e-06, + "loss": 0.985, + "step": 64499 + }, + { + "epoch": 4.77, + "learning_rate": 2.137228059613574e-06, + "loss": 1.0905, + "step": 64500 + }, + { + "epoch": 4.77, + "learning_rate": 2.1369816134992007e-06, + "loss": 0.9895, + "step": 64501 + }, + { + "epoch": 4.77, + "learning_rate": 2.136735179894781e-06, + "loss": 0.9375, + "step": 64502 + }, + { + "epoch": 4.77, + "learning_rate": 2.1364887588006976e-06, + "loss": 1.0294, + "step": 64503 + }, + { + "epoch": 4.77, + "learning_rate": 2.1362423502173525e-06, + "loss": 1.0724, + "step": 64504 + }, + { + "epoch": 4.77, + "learning_rate": 2.1359959541451314e-06, + "loss": 1.1453, + "step": 64505 + }, + { + "epoch": 4.77, + "learning_rate": 2.1357495705844288e-06, + "loss": 0.981, + "step": 64506 + }, + { + "epoch": 4.77, + "learning_rate": 2.135503199535637e-06, + "loss": 0.8392, + "step": 64507 + }, + { + "epoch": 4.77, + "learning_rate": 2.135256840999144e-06, + "loss": 0.9714, + "step": 64508 + }, + { + "epoch": 4.77, + "learning_rate": 2.135010494975347e-06, + "loss": 0.9244, + "step": 64509 + }, + { + "epoch": 4.77, + "learning_rate": 2.1347641614646355e-06, + "loss": 0.9832, + "step": 64510 + }, + { + "epoch": 4.77, + "learning_rate": 2.134517840467403e-06, + "loss": 1.0275, + "step": 64511 + }, + { + "epoch": 4.77, + "learning_rate": 2.1342715319840357e-06, + "loss": 1.0506, + "step": 64512 + }, + { + "epoch": 4.77, + "learning_rate": 2.134025236014934e-06, + "loss": 1.1689, + "step": 64513 + }, + { + "epoch": 4.77, + "learning_rate": 2.1337789525604848e-06, + "loss": 1.0289, + "step": 64514 + }, + { + "epoch": 4.77, + "learning_rate": 2.1335326816210812e-06, + "loss": 0.8797, + "step": 64515 + }, + { + "epoch": 4.77, + "learning_rate": 2.133286423197114e-06, + "loss": 0.9869, + "step": 64516 + }, + { + "epoch": 4.77, + "learning_rate": 2.133040177288973e-06, + "loss": 0.9122, + "step": 64517 + }, + { + "epoch": 4.77, + "learning_rate": 2.1327939438970556e-06, + "loss": 1.0846, + "step": 64518 + }, + { + "epoch": 4.77, + "learning_rate": 2.1325477230217504e-06, + "loss": 1.065, + "step": 64519 + }, + { + "epoch": 4.77, + "learning_rate": 2.1323015146634484e-06, + "loss": 1.0538, + "step": 64520 + }, + { + "epoch": 4.77, + "learning_rate": 2.13205531882254e-06, + "loss": 1.0441, + "step": 64521 + }, + { + "epoch": 4.77, + "learning_rate": 2.1318091354994253e-06, + "loss": 1.0721, + "step": 64522 + }, + { + "epoch": 4.77, + "learning_rate": 2.1315629646944857e-06, + "loss": 0.9497, + "step": 64523 + }, + { + "epoch": 4.77, + "learning_rate": 2.1313168064081145e-06, + "loss": 1.0366, + "step": 64524 + }, + { + "epoch": 4.77, + "learning_rate": 2.1310706606407094e-06, + "loss": 1.0118, + "step": 64525 + }, + { + "epoch": 4.77, + "learning_rate": 2.1308245273926575e-06, + "loss": 0.9641, + "step": 64526 + }, + { + "epoch": 4.77, + "learning_rate": 2.130578406664351e-06, + "loss": 1.0091, + "step": 64527 + }, + { + "epoch": 4.77, + "learning_rate": 2.130332298456179e-06, + "loss": 0.8709, + "step": 64528 + }, + { + "epoch": 4.77, + "learning_rate": 2.1300862027685387e-06, + "loss": 1.0685, + "step": 64529 + }, + { + "epoch": 4.77, + "learning_rate": 2.129840119601818e-06, + "loss": 0.9692, + "step": 64530 + }, + { + "epoch": 4.77, + "learning_rate": 2.1295940489564094e-06, + "loss": 0.9826, + "step": 64531 + }, + { + "epoch": 4.77, + "learning_rate": 2.129347990832704e-06, + "loss": 1.1339, + "step": 64532 + }, + { + "epoch": 4.77, + "learning_rate": 2.129101945231091e-06, + "loss": 1.0468, + "step": 64533 + }, + { + "epoch": 4.77, + "learning_rate": 2.128855912151967e-06, + "loss": 1.0998, + "step": 64534 + }, + { + "epoch": 4.77, + "learning_rate": 2.1286098915957198e-06, + "loss": 0.9908, + "step": 64535 + }, + { + "epoch": 4.77, + "learning_rate": 2.128363883562743e-06, + "loss": 1.0408, + "step": 64536 + }, + { + "epoch": 4.77, + "learning_rate": 2.1281178880534225e-06, + "loss": 1.008, + "step": 64537 + }, + { + "epoch": 4.77, + "learning_rate": 2.1278719050681574e-06, + "loss": 0.9232, + "step": 64538 + }, + { + "epoch": 4.77, + "learning_rate": 2.1276259346073345e-06, + "loss": 1.0595, + "step": 64539 + }, + { + "epoch": 4.77, + "learning_rate": 2.1273799766713477e-06, + "loss": 0.9147, + "step": 64540 + }, + { + "epoch": 4.77, + "learning_rate": 2.127134031260585e-06, + "loss": 0.9992, + "step": 64541 + }, + { + "epoch": 4.77, + "learning_rate": 2.126888098375438e-06, + "loss": 1.0021, + "step": 64542 + }, + { + "epoch": 4.77, + "learning_rate": 2.1266421780163037e-06, + "loss": 1.0132, + "step": 64543 + }, + { + "epoch": 4.77, + "learning_rate": 2.126396270183565e-06, + "loss": 1.0238, + "step": 64544 + }, + { + "epoch": 4.77, + "learning_rate": 2.126150374877619e-06, + "loss": 0.9731, + "step": 64545 + }, + { + "epoch": 4.77, + "learning_rate": 2.1259044920988526e-06, + "loss": 1.016, + "step": 64546 + }, + { + "epoch": 4.77, + "learning_rate": 2.125658621847665e-06, + "loss": 1.0609, + "step": 64547 + }, + { + "epoch": 4.77, + "learning_rate": 2.125412764124436e-06, + "loss": 1.0519, + "step": 64548 + }, + { + "epoch": 4.77, + "learning_rate": 2.1251669189295654e-06, + "loss": 0.9088, + "step": 64549 + }, + { + "epoch": 4.77, + "learning_rate": 2.1249210862634393e-06, + "loss": 1.008, + "step": 64550 + }, + { + "epoch": 4.77, + "learning_rate": 2.1246752661264557e-06, + "loss": 0.9551, + "step": 64551 + }, + { + "epoch": 4.77, + "learning_rate": 2.1244294585189994e-06, + "loss": 0.9387, + "step": 64552 + }, + { + "epoch": 4.77, + "learning_rate": 2.1241836634414605e-06, + "loss": 0.9013, + "step": 64553 + }, + { + "epoch": 4.77, + "learning_rate": 2.1239378808942346e-06, + "loss": 1.063, + "step": 64554 + }, + { + "epoch": 4.77, + "learning_rate": 2.123692110877711e-06, + "loss": 1.0928, + "step": 64555 + }, + { + "epoch": 4.77, + "learning_rate": 2.12344635339228e-06, + "loss": 1.0608, + "step": 64556 + }, + { + "epoch": 4.77, + "learning_rate": 2.1232006084383314e-06, + "loss": 0.9395, + "step": 64557 + }, + { + "epoch": 4.77, + "learning_rate": 2.122954876016261e-06, + "loss": 0.9981, + "step": 64558 + }, + { + "epoch": 4.77, + "learning_rate": 2.122709156126456e-06, + "loss": 1.0203, + "step": 64559 + }, + { + "epoch": 4.77, + "learning_rate": 2.1224634487693075e-06, + "loss": 1.082, + "step": 64560 + }, + { + "epoch": 4.77, + "learning_rate": 2.1222177539452084e-06, + "loss": 1.0833, + "step": 64561 + }, + { + "epoch": 4.77, + "learning_rate": 2.1219720716545445e-06, + "loss": 0.9399, + "step": 64562 + }, + { + "epoch": 4.77, + "learning_rate": 2.121726401897716e-06, + "loss": 1.0841, + "step": 64563 + }, + { + "epoch": 4.77, + "learning_rate": 2.1214807446751017e-06, + "loss": 0.9813, + "step": 64564 + }, + { + "epoch": 4.77, + "learning_rate": 2.121235099987102e-06, + "loss": 1.0792, + "step": 64565 + }, + { + "epoch": 4.77, + "learning_rate": 2.1209894678341027e-06, + "loss": 1.0162, + "step": 64566 + }, + { + "epoch": 4.77, + "learning_rate": 2.120743848216501e-06, + "loss": 1.0197, + "step": 64567 + }, + { + "epoch": 4.77, + "learning_rate": 2.1204982411346775e-06, + "loss": 0.9878, + "step": 64568 + }, + { + "epoch": 4.77, + "learning_rate": 2.1202526465890316e-06, + "loss": 1.034, + "step": 64569 + }, + { + "epoch": 4.77, + "learning_rate": 2.120007064579952e-06, + "loss": 1.033, + "step": 64570 + }, + { + "epoch": 4.77, + "learning_rate": 2.119761495107827e-06, + "loss": 0.9676, + "step": 64571 + }, + { + "epoch": 4.77, + "learning_rate": 2.119515938173049e-06, + "loss": 0.9975, + "step": 64572 + }, + { + "epoch": 4.77, + "learning_rate": 2.119270393776006e-06, + "loss": 0.9198, + "step": 64573 + }, + { + "epoch": 4.77, + "learning_rate": 2.119024861917094e-06, + "loss": 1.0961, + "step": 64574 + }, + { + "epoch": 4.77, + "learning_rate": 2.1187793425967006e-06, + "loss": 0.8991, + "step": 64575 + }, + { + "epoch": 4.77, + "learning_rate": 2.1185338358152164e-06, + "loss": 1.044, + "step": 64576 + }, + { + "epoch": 4.77, + "learning_rate": 2.1182883415730293e-06, + "loss": 1.0002, + "step": 64577 + }, + { + "epoch": 4.77, + "learning_rate": 2.1180428598705348e-06, + "loss": 1.0013, + "step": 64578 + }, + { + "epoch": 4.77, + "learning_rate": 2.1177973907081216e-06, + "loss": 0.8512, + "step": 64579 + }, + { + "epoch": 4.77, + "learning_rate": 2.1175519340861807e-06, + "loss": 0.9422, + "step": 64580 + }, + { + "epoch": 4.77, + "learning_rate": 2.117306490005101e-06, + "loss": 1.056, + "step": 64581 + }, + { + "epoch": 4.77, + "learning_rate": 2.117061058465272e-06, + "loss": 1.1677, + "step": 64582 + }, + { + "epoch": 4.77, + "learning_rate": 2.116815639467088e-06, + "loss": 0.9488, + "step": 64583 + }, + { + "epoch": 4.77, + "learning_rate": 2.1165702330109383e-06, + "loss": 1.0069, + "step": 64584 + }, + { + "epoch": 4.77, + "learning_rate": 2.1163248390972122e-06, + "loss": 1.0564, + "step": 64585 + }, + { + "epoch": 4.77, + "learning_rate": 2.1160794577262976e-06, + "loss": 1.0619, + "step": 64586 + }, + { + "epoch": 4.77, + "learning_rate": 2.1158340888985907e-06, + "loss": 1.0482, + "step": 64587 + }, + { + "epoch": 4.77, + "learning_rate": 2.115588732614482e-06, + "loss": 0.8813, + "step": 64588 + }, + { + "epoch": 4.77, + "learning_rate": 2.1153433888743525e-06, + "loss": 1.0065, + "step": 64589 + }, + { + "epoch": 4.77, + "learning_rate": 2.115098057678603e-06, + "loss": 0.9746, + "step": 64590 + }, + { + "epoch": 4.77, + "learning_rate": 2.1148527390276154e-06, + "loss": 1.0276, + "step": 64591 + }, + { + "epoch": 4.77, + "learning_rate": 2.1146074329217903e-06, + "loss": 0.9632, + "step": 64592 + }, + { + "epoch": 4.77, + "learning_rate": 2.1143621393615065e-06, + "loss": 1.037, + "step": 64593 + }, + { + "epoch": 4.77, + "learning_rate": 2.1141168583471626e-06, + "loss": 1.0354, + "step": 64594 + }, + { + "epoch": 4.77, + "learning_rate": 2.113871589879144e-06, + "loss": 1.0219, + "step": 64595 + }, + { + "epoch": 4.77, + "learning_rate": 2.1136263339578466e-06, + "loss": 1.1435, + "step": 64596 + }, + { + "epoch": 4.77, + "learning_rate": 2.113381090583655e-06, + "loss": 1.1881, + "step": 64597 + }, + { + "epoch": 4.77, + "learning_rate": 2.1131358597569585e-06, + "loss": 1.0311, + "step": 64598 + }, + { + "epoch": 4.77, + "learning_rate": 2.1128906414781534e-06, + "loss": 0.9628, + "step": 64599 + }, + { + "epoch": 4.77, + "learning_rate": 2.1126454357476257e-06, + "loss": 1.0792, + "step": 64600 + }, + { + "epoch": 4.77, + "learning_rate": 2.1124002425657665e-06, + "loss": 0.9145, + "step": 64601 + }, + { + "epoch": 4.77, + "learning_rate": 2.112155061932962e-06, + "loss": 0.9963, + "step": 64602 + }, + { + "epoch": 4.77, + "learning_rate": 2.1119098938496095e-06, + "loss": 1.0058, + "step": 64603 + }, + { + "epoch": 4.77, + "learning_rate": 2.111664738316096e-06, + "loss": 1.0138, + "step": 64604 + }, + { + "epoch": 4.77, + "learning_rate": 2.111419595332811e-06, + "loss": 0.9486, + "step": 64605 + }, + { + "epoch": 4.77, + "learning_rate": 2.111174464900141e-06, + "loss": 0.9783, + "step": 64606 + }, + { + "epoch": 4.77, + "learning_rate": 2.110929347018482e-06, + "loss": 1.0711, + "step": 64607 + }, + { + "epoch": 4.77, + "learning_rate": 2.1106842416882256e-06, + "loss": 1.0743, + "step": 64608 + }, + { + "epoch": 4.77, + "learning_rate": 2.110439148909751e-06, + "loss": 1.0346, + "step": 64609 + }, + { + "epoch": 4.77, + "learning_rate": 2.110194068683459e-06, + "loss": 0.913, + "step": 64610 + }, + { + "epoch": 4.77, + "learning_rate": 2.1099490010097312e-06, + "loss": 0.9375, + "step": 64611 + }, + { + "epoch": 4.77, + "learning_rate": 2.109703945888968e-06, + "loss": 0.9886, + "step": 64612 + }, + { + "epoch": 4.77, + "learning_rate": 2.1094589033215464e-06, + "loss": 0.9053, + "step": 64613 + }, + { + "epoch": 4.77, + "learning_rate": 2.109213873307866e-06, + "loss": 1.0485, + "step": 64614 + }, + { + "epoch": 4.77, + "learning_rate": 2.1089688558483114e-06, + "loss": 1.0622, + "step": 64615 + }, + { + "epoch": 4.77, + "learning_rate": 2.1087238509432793e-06, + "loss": 0.9015, + "step": 64616 + }, + { + "epoch": 4.77, + "learning_rate": 2.1084788585931526e-06, + "loss": 0.9319, + "step": 64617 + }, + { + "epoch": 4.77, + "learning_rate": 2.108233878798319e-06, + "loss": 1.0902, + "step": 64618 + }, + { + "epoch": 4.77, + "learning_rate": 2.107988911559177e-06, + "loss": 1.0217, + "step": 64619 + }, + { + "epoch": 4.77, + "learning_rate": 2.1077439568761103e-06, + "loss": 1.049, + "step": 64620 + }, + { + "epoch": 4.77, + "learning_rate": 2.10749901474951e-06, + "loss": 1.0497, + "step": 64621 + }, + { + "epoch": 4.77, + "learning_rate": 2.107254085179764e-06, + "loss": 1.0212, + "step": 64622 + }, + { + "epoch": 4.77, + "learning_rate": 2.1070091681672654e-06, + "loss": 0.9402, + "step": 64623 + }, + { + "epoch": 4.78, + "learning_rate": 2.106764263712403e-06, + "loss": 1.0353, + "step": 64624 + }, + { + "epoch": 4.78, + "learning_rate": 2.1065193718155662e-06, + "loss": 1.0468, + "step": 64625 + }, + { + "epoch": 4.78, + "learning_rate": 2.1062744924771437e-06, + "loss": 1.0349, + "step": 64626 + }, + { + "epoch": 4.78, + "learning_rate": 2.106029625697522e-06, + "loss": 1.1299, + "step": 64627 + }, + { + "epoch": 4.78, + "learning_rate": 2.1057847714770987e-06, + "loss": 0.923, + "step": 64628 + }, + { + "epoch": 4.78, + "learning_rate": 2.105539929816257e-06, + "loss": 1.0527, + "step": 64629 + }, + { + "epoch": 4.78, + "learning_rate": 2.10529510071539e-06, + "loss": 0.9357, + "step": 64630 + }, + { + "epoch": 4.78, + "learning_rate": 2.105050284174881e-06, + "loss": 0.976, + "step": 64631 + }, + { + "epoch": 4.78, + "learning_rate": 2.1048054801951277e-06, + "loss": 1.0687, + "step": 64632 + }, + { + "epoch": 4.78, + "learning_rate": 2.104560688776517e-06, + "loss": 0.9993, + "step": 64633 + }, + { + "epoch": 4.78, + "learning_rate": 2.1043159099194355e-06, + "loss": 0.9257, + "step": 64634 + }, + { + "epoch": 4.78, + "learning_rate": 2.1040711436242755e-06, + "loss": 1.0443, + "step": 64635 + }, + { + "epoch": 4.78, + "learning_rate": 2.1038263898914225e-06, + "loss": 1.0067, + "step": 64636 + }, + { + "epoch": 4.78, + "learning_rate": 2.1035816487212735e-06, + "loss": 1.0154, + "step": 64637 + }, + { + "epoch": 4.78, + "learning_rate": 2.1033369201142083e-06, + "loss": 1.0407, + "step": 64638 + }, + { + "epoch": 4.78, + "learning_rate": 2.103092204070624e-06, + "loss": 1.011, + "step": 64639 + }, + { + "epoch": 4.78, + "learning_rate": 2.1028475005909066e-06, + "loss": 0.9781, + "step": 64640 + }, + { + "epoch": 4.78, + "learning_rate": 2.102602809675446e-06, + "loss": 0.9198, + "step": 64641 + }, + { + "epoch": 4.78, + "learning_rate": 2.102358131324629e-06, + "loss": 1.0292, + "step": 64642 + }, + { + "epoch": 4.78, + "learning_rate": 2.1021134655388487e-06, + "loss": 0.9161, + "step": 64643 + }, + { + "epoch": 4.78, + "learning_rate": 2.101868812318494e-06, + "loss": 1.0062, + "step": 64644 + }, + { + "epoch": 4.78, + "learning_rate": 2.1016241716639527e-06, + "loss": 0.9222, + "step": 64645 + }, + { + "epoch": 4.78, + "learning_rate": 2.1013795435756147e-06, + "loss": 0.9898, + "step": 64646 + }, + { + "epoch": 4.78, + "learning_rate": 2.101134928053865e-06, + "loss": 1.001, + "step": 64647 + }, + { + "epoch": 4.78, + "learning_rate": 2.1008903250991e-06, + "loss": 1.074, + "step": 64648 + }, + { + "epoch": 4.78, + "learning_rate": 2.100645734711706e-06, + "loss": 0.9453, + "step": 64649 + }, + { + "epoch": 4.78, + "learning_rate": 2.100401156892071e-06, + "loss": 1.027, + "step": 64650 + }, + { + "epoch": 4.78, + "learning_rate": 2.100156591640582e-06, + "loss": 1.0052, + "step": 64651 + }, + { + "epoch": 4.78, + "learning_rate": 2.0999120389576333e-06, + "loss": 1.045, + "step": 64652 + }, + { + "epoch": 4.78, + "learning_rate": 2.0996674988436148e-06, + "loss": 0.975, + "step": 64653 + }, + { + "epoch": 4.78, + "learning_rate": 2.0994229712989067e-06, + "loss": 1.0066, + "step": 64654 + }, + { + "epoch": 4.78, + "learning_rate": 2.099178456323907e-06, + "loss": 1.0141, + "step": 64655 + }, + { + "epoch": 4.78, + "learning_rate": 2.098933953918998e-06, + "loss": 0.8958, + "step": 64656 + }, + { + "epoch": 4.78, + "learning_rate": 2.098689464084577e-06, + "loss": 1.1426, + "step": 64657 + }, + { + "epoch": 4.78, + "learning_rate": 2.098444986821023e-06, + "loss": 1.0426, + "step": 64658 + }, + { + "epoch": 4.78, + "learning_rate": 2.0982005221287326e-06, + "loss": 1.1122, + "step": 64659 + }, + { + "epoch": 4.78, + "learning_rate": 2.0979560700080893e-06, + "loss": 0.9285, + "step": 64660 + }, + { + "epoch": 4.78, + "learning_rate": 2.0977116304594903e-06, + "loss": 1.0596, + "step": 64661 + }, + { + "epoch": 4.78, + "learning_rate": 2.0974672034833144e-06, + "loss": 1.0171, + "step": 64662 + }, + { + "epoch": 4.78, + "learning_rate": 2.097222789079957e-06, + "loss": 0.9118, + "step": 64663 + }, + { + "epoch": 4.78, + "learning_rate": 2.096978387249805e-06, + "loss": 1.1183, + "step": 64664 + }, + { + "epoch": 4.78, + "learning_rate": 2.096733997993248e-06, + "loss": 0.9804, + "step": 64665 + }, + { + "epoch": 4.78, + "learning_rate": 2.096489621310673e-06, + "loss": 1.0188, + "step": 64666 + }, + { + "epoch": 4.78, + "learning_rate": 2.0962452572024685e-06, + "loss": 1.0028, + "step": 64667 + }, + { + "epoch": 4.78, + "learning_rate": 2.0960009056690266e-06, + "loss": 1.0399, + "step": 64668 + }, + { + "epoch": 4.78, + "learning_rate": 2.0957565667107336e-06, + "loss": 1.1277, + "step": 64669 + }, + { + "epoch": 4.78, + "learning_rate": 2.0955122403279805e-06, + "loss": 1.0825, + "step": 64670 + }, + { + "epoch": 4.78, + "learning_rate": 2.0952679265211506e-06, + "loss": 1.0237, + "step": 64671 + }, + { + "epoch": 4.78, + "learning_rate": 2.0950236252906385e-06, + "loss": 0.9988, + "step": 64672 + }, + { + "epoch": 4.78, + "learning_rate": 2.094779336636832e-06, + "loss": 1.0733, + "step": 64673 + }, + { + "epoch": 4.78, + "learning_rate": 2.0945350605601176e-06, + "loss": 1.0363, + "step": 64674 + }, + { + "epoch": 4.78, + "learning_rate": 2.094290797060885e-06, + "loss": 0.9018, + "step": 64675 + }, + { + "epoch": 4.78, + "learning_rate": 2.0940465461395197e-06, + "loss": 1.0179, + "step": 64676 + }, + { + "epoch": 4.78, + "learning_rate": 2.0938023077964166e-06, + "loss": 0.9269, + "step": 64677 + }, + { + "epoch": 4.78, + "learning_rate": 2.09355808203196e-06, + "loss": 1.0126, + "step": 64678 + }, + { + "epoch": 4.78, + "learning_rate": 2.0933138688465402e-06, + "loss": 1.0313, + "step": 64679 + }, + { + "epoch": 4.78, + "learning_rate": 2.0930696682405417e-06, + "loss": 1.0871, + "step": 64680 + }, + { + "epoch": 4.78, + "learning_rate": 2.09282548021436e-06, + "loss": 1.0435, + "step": 64681 + }, + { + "epoch": 4.78, + "learning_rate": 2.092581304768381e-06, + "loss": 0.9792, + "step": 64682 + }, + { + "epoch": 4.78, + "learning_rate": 2.092337141902987e-06, + "loss": 1.0504, + "step": 64683 + }, + { + "epoch": 4.78, + "learning_rate": 2.092092991618575e-06, + "loss": 0.9891, + "step": 64684 + }, + { + "epoch": 4.78, + "learning_rate": 2.0918488539155292e-06, + "loss": 1.0797, + "step": 64685 + }, + { + "epoch": 4.78, + "learning_rate": 2.091604728794239e-06, + "loss": 0.8866, + "step": 64686 + }, + { + "epoch": 4.78, + "learning_rate": 2.0913606162550893e-06, + "loss": 0.9523, + "step": 64687 + }, + { + "epoch": 4.78, + "learning_rate": 2.091116516298475e-06, + "loss": 1.0498, + "step": 64688 + }, + { + "epoch": 4.78, + "learning_rate": 2.0908724289247805e-06, + "loss": 1.0216, + "step": 64689 + }, + { + "epoch": 4.78, + "learning_rate": 2.0906283541343954e-06, + "loss": 0.9629, + "step": 64690 + }, + { + "epoch": 4.78, + "learning_rate": 2.090384291927707e-06, + "loss": 0.9683, + "step": 64691 + }, + { + "epoch": 4.78, + "learning_rate": 2.090140242305102e-06, + "loss": 1.0838, + "step": 64692 + }, + { + "epoch": 4.78, + "learning_rate": 2.0898962052669736e-06, + "loss": 1.0312, + "step": 64693 + }, + { + "epoch": 4.78, + "learning_rate": 2.089652180813706e-06, + "loss": 0.8982, + "step": 64694 + }, + { + "epoch": 4.78, + "learning_rate": 2.0894081689456892e-06, + "loss": 1.0113, + "step": 64695 + }, + { + "epoch": 4.78, + "learning_rate": 2.089164169663308e-06, + "loss": 0.9084, + "step": 64696 + }, + { + "epoch": 4.78, + "learning_rate": 2.088920182966957e-06, + "loss": 0.9387, + "step": 64697 + }, + { + "epoch": 4.78, + "learning_rate": 2.088676208857019e-06, + "loss": 1.0783, + "step": 64698 + }, + { + "epoch": 4.78, + "learning_rate": 2.088432247333886e-06, + "loss": 0.9867, + "step": 64699 + }, + { + "epoch": 4.78, + "learning_rate": 2.0881882983979427e-06, + "loss": 1.0275, + "step": 64700 + }, + { + "epoch": 4.78, + "learning_rate": 2.0879443620495766e-06, + "loss": 0.9497, + "step": 64701 + }, + { + "epoch": 4.78, + "learning_rate": 2.0877004382891817e-06, + "loss": 0.9904, + "step": 64702 + }, + { + "epoch": 4.78, + "learning_rate": 2.0874565271171386e-06, + "loss": 1.0536, + "step": 64703 + }, + { + "epoch": 4.78, + "learning_rate": 2.0872126285338414e-06, + "loss": 0.9424, + "step": 64704 + }, + { + "epoch": 4.78, + "learning_rate": 2.0869687425396722e-06, + "loss": 1.0174, + "step": 64705 + }, + { + "epoch": 4.78, + "learning_rate": 2.0867248691350282e-06, + "loss": 0.9409, + "step": 64706 + }, + { + "epoch": 4.78, + "learning_rate": 2.086481008320287e-06, + "loss": 1.0327, + "step": 64707 + }, + { + "epoch": 4.78, + "learning_rate": 2.086237160095843e-06, + "loss": 1.0371, + "step": 64708 + }, + { + "epoch": 4.78, + "learning_rate": 2.08599332446208e-06, + "loss": 0.9338, + "step": 64709 + }, + { + "epoch": 4.78, + "learning_rate": 2.0857495014193953e-06, + "loss": 1.0734, + "step": 64710 + }, + { + "epoch": 4.78, + "learning_rate": 2.085505690968165e-06, + "loss": 1.0578, + "step": 64711 + }, + { + "epoch": 4.78, + "learning_rate": 2.0852618931087797e-06, + "loss": 1.0478, + "step": 64712 + }, + { + "epoch": 4.78, + "learning_rate": 2.085018107841633e-06, + "loss": 1.0227, + "step": 64713 + }, + { + "epoch": 4.78, + "learning_rate": 2.084774335167109e-06, + "loss": 1.0519, + "step": 64714 + }, + { + "epoch": 4.78, + "learning_rate": 2.084530575085596e-06, + "loss": 0.8934, + "step": 64715 + }, + { + "epoch": 4.78, + "learning_rate": 2.0842868275974783e-06, + "loss": 1.0128, + "step": 64716 + }, + { + "epoch": 4.78, + "learning_rate": 2.0840430927031495e-06, + "loss": 1.0144, + "step": 64717 + }, + { + "epoch": 4.78, + "learning_rate": 2.0837993704029957e-06, + "loss": 0.9802, + "step": 64718 + }, + { + "epoch": 4.78, + "learning_rate": 2.083555660697404e-06, + "loss": 1.0649, + "step": 64719 + }, + { + "epoch": 4.78, + "learning_rate": 2.0833119635867617e-06, + "loss": 0.9831, + "step": 64720 + }, + { + "epoch": 4.78, + "learning_rate": 2.083068279071453e-06, + "loss": 0.9256, + "step": 64721 + }, + { + "epoch": 4.78, + "learning_rate": 2.0828246071518732e-06, + "loss": 0.9316, + "step": 64722 + }, + { + "epoch": 4.78, + "learning_rate": 2.082580947828405e-06, + "loss": 1.1186, + "step": 64723 + }, + { + "epoch": 4.78, + "learning_rate": 2.082337301101439e-06, + "loss": 0.7752, + "step": 64724 + }, + { + "epoch": 4.78, + "learning_rate": 2.0820936669713575e-06, + "loss": 1.0479, + "step": 64725 + }, + { + "epoch": 4.78, + "learning_rate": 2.081850045438555e-06, + "loss": 0.9508, + "step": 64726 + }, + { + "epoch": 4.78, + "learning_rate": 2.081606436503415e-06, + "loss": 0.9743, + "step": 64727 + }, + { + "epoch": 4.78, + "learning_rate": 2.0813628401663255e-06, + "loss": 0.9364, + "step": 64728 + }, + { + "epoch": 4.78, + "learning_rate": 2.081119256427675e-06, + "loss": 0.9781, + "step": 64729 + }, + { + "epoch": 4.78, + "learning_rate": 2.080875685287851e-06, + "loss": 1.0015, + "step": 64730 + }, + { + "epoch": 4.78, + "learning_rate": 2.0806321267472397e-06, + "loss": 0.8314, + "step": 64731 + }, + { + "epoch": 4.78, + "learning_rate": 2.0803885808062264e-06, + "loss": 0.9521, + "step": 64732 + }, + { + "epoch": 4.78, + "learning_rate": 2.080145047465205e-06, + "loss": 0.9457, + "step": 64733 + }, + { + "epoch": 4.78, + "learning_rate": 2.079901526724559e-06, + "loss": 0.9621, + "step": 64734 + }, + { + "epoch": 4.78, + "learning_rate": 2.079658018584677e-06, + "loss": 0.975, + "step": 64735 + }, + { + "epoch": 4.78, + "learning_rate": 2.079414523045943e-06, + "loss": 0.9545, + "step": 64736 + }, + { + "epoch": 4.78, + "learning_rate": 2.0791710401087496e-06, + "loss": 0.9519, + "step": 64737 + }, + { + "epoch": 4.78, + "learning_rate": 2.0789275697734813e-06, + "loss": 0.9639, + "step": 64738 + }, + { + "epoch": 4.78, + "learning_rate": 2.0786841120405264e-06, + "loss": 1.0718, + "step": 64739 + }, + { + "epoch": 4.78, + "learning_rate": 2.078440666910272e-06, + "loss": 0.9141, + "step": 64740 + }, + { + "epoch": 4.78, + "learning_rate": 2.0781972343831013e-06, + "loss": 0.915, + "step": 64741 + }, + { + "epoch": 4.78, + "learning_rate": 2.0779538144594093e-06, + "loss": 1.0329, + "step": 64742 + }, + { + "epoch": 4.78, + "learning_rate": 2.0777104071395794e-06, + "loss": 1.0401, + "step": 64743 + }, + { + "epoch": 4.78, + "learning_rate": 2.0774670124239982e-06, + "loss": 1.0563, + "step": 64744 + }, + { + "epoch": 4.78, + "learning_rate": 2.077223630313052e-06, + "loss": 1.042, + "step": 64745 + }, + { + "epoch": 4.78, + "learning_rate": 2.076980260807131e-06, + "loss": 1.0231, + "step": 64746 + }, + { + "epoch": 4.78, + "learning_rate": 2.0767369039066256e-06, + "loss": 1.0437, + "step": 64747 + }, + { + "epoch": 4.78, + "learning_rate": 2.0764935596119106e-06, + "loss": 0.9609, + "step": 64748 + }, + { + "epoch": 4.78, + "learning_rate": 2.0762502279233852e-06, + "loss": 1.0781, + "step": 64749 + }, + { + "epoch": 4.78, + "learning_rate": 2.076006908841429e-06, + "loss": 1.0572, + "step": 64750 + }, + { + "epoch": 4.78, + "learning_rate": 2.0757636023664383e-06, + "loss": 0.9542, + "step": 64751 + }, + { + "epoch": 4.78, + "learning_rate": 2.0755203084987896e-06, + "loss": 1.0024, + "step": 64752 + }, + { + "epoch": 4.78, + "learning_rate": 2.075277027238877e-06, + "loss": 1.0177, + "step": 64753 + }, + { + "epoch": 4.78, + "learning_rate": 2.075033758587082e-06, + "loss": 0.9969, + "step": 64754 + }, + { + "epoch": 4.78, + "learning_rate": 2.0747905025438e-06, + "loss": 0.9336, + "step": 64755 + }, + { + "epoch": 4.78, + "learning_rate": 2.0745472591094106e-06, + "loss": 0.9973, + "step": 64756 + }, + { + "epoch": 4.78, + "learning_rate": 2.0743040282843007e-06, + "loss": 1.02, + "step": 64757 + }, + { + "epoch": 4.78, + "learning_rate": 2.0740608100688618e-06, + "loss": 0.9913, + "step": 64758 + }, + { + "epoch": 4.79, + "learning_rate": 2.0738176044634793e-06, + "loss": 0.9862, + "step": 64759 + }, + { + "epoch": 4.79, + "learning_rate": 2.0735744114685396e-06, + "loss": 0.962, + "step": 64760 + }, + { + "epoch": 4.79, + "learning_rate": 2.0733312310844268e-06, + "loss": 1.0238, + "step": 64761 + }, + { + "epoch": 4.79, + "learning_rate": 2.0730880633115337e-06, + "loss": 1.0139, + "step": 64762 + }, + { + "epoch": 4.79, + "learning_rate": 2.0728449081502435e-06, + "loss": 1.0772, + "step": 64763 + }, + { + "epoch": 4.79, + "learning_rate": 2.0726017656009434e-06, + "loss": 0.9823, + "step": 64764 + }, + { + "epoch": 4.79, + "learning_rate": 2.0723586356640177e-06, + "loss": 0.986, + "step": 64765 + }, + { + "epoch": 4.79, + "learning_rate": 2.07211551833986e-06, + "loss": 1.0444, + "step": 64766 + }, + { + "epoch": 4.79, + "learning_rate": 2.0718724136288527e-06, + "loss": 0.9942, + "step": 64767 + }, + { + "epoch": 4.79, + "learning_rate": 2.071629321531382e-06, + "loss": 0.9859, + "step": 64768 + }, + { + "epoch": 4.79, + "learning_rate": 2.071386242047836e-06, + "loss": 0.9968, + "step": 64769 + }, + { + "epoch": 4.79, + "learning_rate": 2.071143175178598e-06, + "loss": 1.0849, + "step": 64770 + }, + { + "epoch": 4.79, + "learning_rate": 2.070900120924061e-06, + "loss": 0.9502, + "step": 64771 + }, + { + "epoch": 4.79, + "learning_rate": 2.070657079284608e-06, + "loss": 0.9708, + "step": 64772 + }, + { + "epoch": 4.79, + "learning_rate": 2.0704140502606274e-06, + "loss": 0.993, + "step": 64773 + }, + { + "epoch": 4.79, + "learning_rate": 2.0701710338525005e-06, + "loss": 1.0028, + "step": 64774 + }, + { + "epoch": 4.79, + "learning_rate": 2.069928030060624e-06, + "loss": 1.1357, + "step": 64775 + }, + { + "epoch": 4.79, + "learning_rate": 2.069685038885376e-06, + "loss": 1.0603, + "step": 64776 + }, + { + "epoch": 4.79, + "learning_rate": 2.0694420603271425e-06, + "loss": 0.8743, + "step": 64777 + }, + { + "epoch": 4.79, + "learning_rate": 2.069199094386316e-06, + "loss": 1.0188, + "step": 64778 + }, + { + "epoch": 4.79, + "learning_rate": 2.0689561410632806e-06, + "loss": 0.9441, + "step": 64779 + }, + { + "epoch": 4.79, + "learning_rate": 2.0687132003584222e-06, + "loss": 0.97, + "step": 64780 + }, + { + "epoch": 4.79, + "learning_rate": 2.068470272272125e-06, + "loss": 1.0264, + "step": 64781 + }, + { + "epoch": 4.79, + "learning_rate": 2.0682273568047805e-06, + "loss": 0.9655, + "step": 64782 + }, + { + "epoch": 4.79, + "learning_rate": 2.067984453956774e-06, + "loss": 0.9264, + "step": 64783 + }, + { + "epoch": 4.79, + "learning_rate": 2.0677415637284892e-06, + "loss": 1.0285, + "step": 64784 + }, + { + "epoch": 4.79, + "learning_rate": 2.067498686120315e-06, + "loss": 1.0221, + "step": 64785 + }, + { + "epoch": 4.79, + "learning_rate": 2.0672558211326333e-06, + "loss": 1.0215, + "step": 64786 + }, + { + "epoch": 4.79, + "learning_rate": 2.067012968765838e-06, + "loss": 1.0046, + "step": 64787 + }, + { + "epoch": 4.79, + "learning_rate": 2.0667701290203112e-06, + "loss": 0.9432, + "step": 64788 + }, + { + "epoch": 4.79, + "learning_rate": 2.0665273018964394e-06, + "loss": 1.0308, + "step": 64789 + }, + { + "epoch": 4.79, + "learning_rate": 2.0662844873946066e-06, + "loss": 1.0299, + "step": 64790 + }, + { + "epoch": 4.79, + "learning_rate": 2.0660416855152033e-06, + "loss": 0.9997, + "step": 64791 + }, + { + "epoch": 4.79, + "learning_rate": 2.0657988962586153e-06, + "loss": 1.0428, + "step": 64792 + }, + { + "epoch": 4.79, + "learning_rate": 2.0655561196252284e-06, + "loss": 0.97, + "step": 64793 + }, + { + "epoch": 4.79, + "learning_rate": 2.065313355615427e-06, + "loss": 0.9855, + "step": 64794 + }, + { + "epoch": 4.79, + "learning_rate": 2.0650706042295955e-06, + "loss": 0.9647, + "step": 64795 + }, + { + "epoch": 4.79, + "learning_rate": 2.0648278654681296e-06, + "loss": 1.0152, + "step": 64796 + }, + { + "epoch": 4.79, + "learning_rate": 2.064585139331403e-06, + "loss": 1.0226, + "step": 64797 + }, + { + "epoch": 4.79, + "learning_rate": 2.064342425819811e-06, + "loss": 1.0823, + "step": 64798 + }, + { + "epoch": 4.79, + "learning_rate": 2.0640997249337337e-06, + "loss": 0.9818, + "step": 64799 + }, + { + "epoch": 4.79, + "learning_rate": 2.0638570366735655e-06, + "loss": 0.9397, + "step": 64800 + }, + { + "epoch": 4.79, + "learning_rate": 2.0636143610396817e-06, + "loss": 0.9865, + "step": 64801 + }, + { + "epoch": 4.79, + "learning_rate": 2.063371698032477e-06, + "loss": 0.9906, + "step": 64802 + }, + { + "epoch": 4.79, + "learning_rate": 2.0631290476523346e-06, + "loss": 0.9025, + "step": 64803 + }, + { + "epoch": 4.79, + "learning_rate": 2.0628864098996404e-06, + "loss": 1.0253, + "step": 64804 + }, + { + "epoch": 4.79, + "learning_rate": 2.0626437847747793e-06, + "loss": 1.023, + "step": 64805 + }, + { + "epoch": 4.79, + "learning_rate": 2.0624011722781368e-06, + "loss": 1.104, + "step": 64806 + }, + { + "epoch": 4.79, + "learning_rate": 2.0621585724101024e-06, + "loss": 1.0758, + "step": 64807 + }, + { + "epoch": 4.79, + "learning_rate": 2.06191598517106e-06, + "loss": 0.9616, + "step": 64808 + }, + { + "epoch": 4.79, + "learning_rate": 2.0616734105613965e-06, + "loss": 0.841, + "step": 64809 + }, + { + "epoch": 4.79, + "learning_rate": 2.0614308485814936e-06, + "loss": 1.0177, + "step": 64810 + }, + { + "epoch": 4.79, + "learning_rate": 2.0611882992317443e-06, + "loss": 1.0373, + "step": 64811 + }, + { + "epoch": 4.79, + "learning_rate": 2.0609457625125297e-06, + "loss": 1.0554, + "step": 64812 + }, + { + "epoch": 4.79, + "learning_rate": 2.0607032384242374e-06, + "loss": 1.0277, + "step": 64813 + }, + { + "epoch": 4.79, + "learning_rate": 2.060460726967253e-06, + "loss": 1.0592, + "step": 64814 + }, + { + "epoch": 4.79, + "learning_rate": 2.060218228141958e-06, + "loss": 0.9604, + "step": 64815 + }, + { + "epoch": 4.79, + "learning_rate": 2.0599757419487465e-06, + "loss": 1.0191, + "step": 64816 + }, + { + "epoch": 4.79, + "learning_rate": 2.059733268387999e-06, + "loss": 1.071, + "step": 64817 + }, + { + "epoch": 4.79, + "learning_rate": 2.059490807460103e-06, + "loss": 0.9846, + "step": 64818 + }, + { + "epoch": 4.79, + "learning_rate": 2.05924835916544e-06, + "loss": 0.9403, + "step": 64819 + }, + { + "epoch": 4.79, + "learning_rate": 2.059005923504405e-06, + "loss": 1.0291, + "step": 64820 + }, + { + "epoch": 4.79, + "learning_rate": 2.058763500477372e-06, + "loss": 0.939, + "step": 64821 + }, + { + "epoch": 4.79, + "learning_rate": 2.0585210900847362e-06, + "loss": 0.9676, + "step": 64822 + }, + { + "epoch": 4.79, + "learning_rate": 2.0582786923268794e-06, + "loss": 0.9654, + "step": 64823 + }, + { + "epoch": 4.79, + "learning_rate": 2.0580363072041876e-06, + "loss": 1.1148, + "step": 64824 + }, + { + "epoch": 4.79, + "learning_rate": 2.0577939347170462e-06, + "loss": 0.9335, + "step": 64825 + }, + { + "epoch": 4.79, + "learning_rate": 2.0575515748658382e-06, + "loss": 1.0022, + "step": 64826 + }, + { + "epoch": 4.79, + "learning_rate": 2.0573092276509553e-06, + "loss": 0.9147, + "step": 64827 + }, + { + "epoch": 4.79, + "learning_rate": 2.05706689307278e-06, + "loss": 0.9511, + "step": 64828 + }, + { + "epoch": 4.79, + "learning_rate": 2.056824571131697e-06, + "loss": 1.0008, + "step": 64829 + }, + { + "epoch": 4.79, + "learning_rate": 2.0565822618280893e-06, + "loss": 0.8561, + "step": 64830 + }, + { + "epoch": 4.79, + "learning_rate": 2.0563399651623493e-06, + "loss": 0.9443, + "step": 64831 + }, + { + "epoch": 4.79, + "learning_rate": 2.0560976811348574e-06, + "loss": 0.9328, + "step": 64832 + }, + { + "epoch": 4.79, + "learning_rate": 2.0558554097460014e-06, + "loss": 0.9798, + "step": 64833 + }, + { + "epoch": 4.79, + "learning_rate": 2.055613150996166e-06, + "loss": 0.9887, + "step": 64834 + }, + { + "epoch": 4.79, + "learning_rate": 2.055370904885734e-06, + "loss": 1.0994, + "step": 64835 + }, + { + "epoch": 4.79, + "learning_rate": 2.0551286714150952e-06, + "loss": 1.022, + "step": 64836 + }, + { + "epoch": 4.79, + "learning_rate": 2.0548864505846333e-06, + "loss": 1.0041, + "step": 64837 + }, + { + "epoch": 4.79, + "learning_rate": 2.0546442423947334e-06, + "loss": 1.0082, + "step": 64838 + }, + { + "epoch": 4.79, + "learning_rate": 2.054402046845777e-06, + "loss": 1.017, + "step": 64839 + }, + { + "epoch": 4.79, + "learning_rate": 2.0541598639381575e-06, + "loss": 1.0695, + "step": 64840 + }, + { + "epoch": 4.79, + "learning_rate": 2.0539176936722583e-06, + "loss": 1.0715, + "step": 64841 + }, + { + "epoch": 4.79, + "learning_rate": 2.0536755360484563e-06, + "loss": 1.0932, + "step": 64842 + }, + { + "epoch": 4.79, + "learning_rate": 2.0534333910671454e-06, + "loss": 0.925, + "step": 64843 + }, + { + "epoch": 4.79, + "learning_rate": 2.053191258728706e-06, + "loss": 1.0676, + "step": 64844 + }, + { + "epoch": 4.79, + "learning_rate": 2.0529491390335323e-06, + "loss": 0.9096, + "step": 64845 + }, + { + "epoch": 4.79, + "learning_rate": 2.0527070319819963e-06, + "loss": 0.9134, + "step": 64846 + }, + { + "epoch": 4.79, + "learning_rate": 2.052464937574492e-06, + "loss": 0.9352, + "step": 64847 + }, + { + "epoch": 4.79, + "learning_rate": 2.0522228558114e-06, + "loss": 1.073, + "step": 64848 + }, + { + "epoch": 4.79, + "learning_rate": 2.051980786693113e-06, + "loss": 1.0547, + "step": 64849 + }, + { + "epoch": 4.79, + "learning_rate": 2.0517387302200086e-06, + "loss": 1.0742, + "step": 64850 + }, + { + "epoch": 4.79, + "learning_rate": 2.05149668639247e-06, + "loss": 0.9435, + "step": 64851 + }, + { + "epoch": 4.79, + "learning_rate": 2.0512546552108914e-06, + "loss": 0.9682, + "step": 64852 + }, + { + "epoch": 4.79, + "learning_rate": 2.051012636675651e-06, + "loss": 1.0876, + "step": 64853 + }, + { + "epoch": 4.79, + "learning_rate": 2.0507706307871366e-06, + "loss": 0.849, + "step": 64854 + }, + { + "epoch": 4.79, + "learning_rate": 2.05052863754573e-06, + "loss": 1.0138, + "step": 64855 + }, + { + "epoch": 4.79, + "learning_rate": 2.050286656951821e-06, + "loss": 1.0443, + "step": 64856 + }, + { + "epoch": 4.79, + "learning_rate": 2.0500446890057912e-06, + "loss": 1.0915, + "step": 64857 + }, + { + "epoch": 4.79, + "learning_rate": 2.0498027337080284e-06, + "loss": 1.0641, + "step": 64858 + }, + { + "epoch": 4.79, + "learning_rate": 2.0495607910589145e-06, + "loss": 1.0405, + "step": 64859 + }, + { + "epoch": 4.79, + "learning_rate": 2.0493188610588334e-06, + "loss": 1.0238, + "step": 64860 + }, + { + "epoch": 4.79, + "learning_rate": 2.049076943708175e-06, + "loss": 1.0436, + "step": 64861 + }, + { + "epoch": 4.79, + "learning_rate": 2.048835039007321e-06, + "loss": 1.0812, + "step": 64862 + }, + { + "epoch": 4.79, + "learning_rate": 2.048593146956657e-06, + "loss": 0.956, + "step": 64863 + }, + { + "epoch": 4.79, + "learning_rate": 2.0483512675565643e-06, + "loss": 1.1063, + "step": 64864 + }, + { + "epoch": 4.79, + "learning_rate": 2.0481094008074364e-06, + "loss": 1.0107, + "step": 64865 + }, + { + "epoch": 4.79, + "learning_rate": 2.0478675467096477e-06, + "loss": 0.9628, + "step": 64866 + }, + { + "epoch": 4.79, + "learning_rate": 2.0476257052635908e-06, + "loss": 0.9149, + "step": 64867 + }, + { + "epoch": 4.79, + "learning_rate": 2.0473838764696476e-06, + "loss": 1.0154, + "step": 64868 + }, + { + "epoch": 4.79, + "learning_rate": 2.0471420603282024e-06, + "loss": 1.0775, + "step": 64869 + }, + { + "epoch": 4.79, + "learning_rate": 2.04690025683964e-06, + "loss": 1.0453, + "step": 64870 + }, + { + "epoch": 4.79, + "learning_rate": 2.046658466004344e-06, + "loss": 1.0343, + "step": 64871 + }, + { + "epoch": 4.79, + "learning_rate": 2.046416687822702e-06, + "loss": 0.9604, + "step": 64872 + }, + { + "epoch": 4.79, + "learning_rate": 2.046174922295099e-06, + "loss": 0.9698, + "step": 64873 + }, + { + "epoch": 4.79, + "learning_rate": 2.0459331694219165e-06, + "loss": 1.1079, + "step": 64874 + }, + { + "epoch": 4.79, + "learning_rate": 2.045691429203538e-06, + "loss": 0.8374, + "step": 64875 + }, + { + "epoch": 4.79, + "learning_rate": 2.045449701640354e-06, + "loss": 0.9695, + "step": 64876 + }, + { + "epoch": 4.79, + "learning_rate": 2.0452079867327444e-06, + "loss": 0.9669, + "step": 64877 + }, + { + "epoch": 4.79, + "learning_rate": 2.044966284481097e-06, + "loss": 1.0162, + "step": 64878 + }, + { + "epoch": 4.79, + "learning_rate": 2.044724594885793e-06, + "loss": 0.9937, + "step": 64879 + }, + { + "epoch": 4.79, + "learning_rate": 2.044482917947216e-06, + "loss": 1.0781, + "step": 64880 + }, + { + "epoch": 4.79, + "learning_rate": 2.044241253665755e-06, + "loss": 0.9004, + "step": 64881 + }, + { + "epoch": 4.79, + "learning_rate": 2.043999602041793e-06, + "loss": 0.8987, + "step": 64882 + }, + { + "epoch": 4.79, + "learning_rate": 2.0437579630757145e-06, + "loss": 0.9769, + "step": 64883 + }, + { + "epoch": 4.79, + "learning_rate": 2.043516336767899e-06, + "loss": 1.0255, + "step": 64884 + }, + { + "epoch": 4.79, + "learning_rate": 2.043274723118739e-06, + "loss": 1.0249, + "step": 64885 + }, + { + "epoch": 4.79, + "learning_rate": 2.043033122128615e-06, + "loss": 1.1491, + "step": 64886 + }, + { + "epoch": 4.79, + "learning_rate": 2.042791533797911e-06, + "loss": 1.1143, + "step": 64887 + }, + { + "epoch": 4.79, + "learning_rate": 2.0425499581270114e-06, + "loss": 0.9838, + "step": 64888 + }, + { + "epoch": 4.79, + "learning_rate": 2.0423083951162993e-06, + "loss": 1.0598, + "step": 64889 + }, + { + "epoch": 4.79, + "learning_rate": 2.042066844766165e-06, + "loss": 1.0708, + "step": 64890 + }, + { + "epoch": 4.79, + "learning_rate": 2.0418253070769844e-06, + "loss": 0.8614, + "step": 64891 + }, + { + "epoch": 4.79, + "learning_rate": 2.041583782049148e-06, + "loss": 0.9399, + "step": 64892 + }, + { + "epoch": 4.79, + "learning_rate": 2.0413422696830344e-06, + "loss": 1.0983, + "step": 64893 + }, + { + "epoch": 4.79, + "learning_rate": 2.0411007699790375e-06, + "loss": 0.9877, + "step": 64894 + }, + { + "epoch": 4.8, + "learning_rate": 2.0408592829375305e-06, + "loss": 0.9652, + "step": 64895 + }, + { + "epoch": 4.8, + "learning_rate": 2.0406178085589047e-06, + "loss": 0.9899, + "step": 64896 + }, + { + "epoch": 4.8, + "learning_rate": 2.0403763468435423e-06, + "loss": 0.928, + "step": 64897 + }, + { + "epoch": 4.8, + "learning_rate": 2.0401348977918267e-06, + "loss": 0.9899, + "step": 64898 + }, + { + "epoch": 4.8, + "learning_rate": 2.039893461404143e-06, + "loss": 1.0383, + "step": 64899 + }, + { + "epoch": 4.8, + "learning_rate": 2.0396520376808715e-06, + "loss": 0.9466, + "step": 64900 + }, + { + "epoch": 4.8, + "learning_rate": 2.0394106266224033e-06, + "loss": 0.9623, + "step": 64901 + }, + { + "epoch": 4.8, + "learning_rate": 2.0391692282291187e-06, + "loss": 1.1988, + "step": 64902 + }, + { + "epoch": 4.8, + "learning_rate": 2.0389278425014012e-06, + "loss": 0.9252, + "step": 64903 + }, + { + "epoch": 4.8, + "learning_rate": 2.0386864694396347e-06, + "loss": 1.0119, + "step": 64904 + }, + { + "epoch": 4.8, + "learning_rate": 2.038445109044206e-06, + "loss": 0.9909, + "step": 64905 + }, + { + "epoch": 4.8, + "learning_rate": 2.0382037613154992e-06, + "loss": 1.0216, + "step": 64906 + }, + { + "epoch": 4.8, + "learning_rate": 2.037962426253891e-06, + "loss": 0.9779, + "step": 64907 + }, + { + "epoch": 4.8, + "learning_rate": 2.0377211038597744e-06, + "loss": 1.0699, + "step": 64908 + }, + { + "epoch": 4.8, + "learning_rate": 2.0374797941335256e-06, + "loss": 0.969, + "step": 64909 + }, + { + "epoch": 4.8, + "learning_rate": 2.0372384970755388e-06, + "loss": 1.0093, + "step": 64910 + }, + { + "epoch": 4.8, + "learning_rate": 2.036997212686186e-06, + "loss": 1.0301, + "step": 64911 + }, + { + "epoch": 4.8, + "learning_rate": 2.036755940965859e-06, + "loss": 0.963, + "step": 64912 + }, + { + "epoch": 4.8, + "learning_rate": 2.036514681914937e-06, + "loss": 1.0523, + "step": 64913 + }, + { + "epoch": 4.8, + "learning_rate": 2.036273435533812e-06, + "loss": 1.1465, + "step": 64914 + }, + { + "epoch": 4.8, + "learning_rate": 2.0360322018228585e-06, + "loss": 0.9173, + "step": 64915 + }, + { + "epoch": 4.8, + "learning_rate": 2.0357909807824615e-06, + "loss": 1.0048, + "step": 64916 + }, + { + "epoch": 4.8, + "learning_rate": 2.0355497724130103e-06, + "loss": 1.1276, + "step": 64917 + }, + { + "epoch": 4.8, + "learning_rate": 2.0353085767148852e-06, + "loss": 0.9507, + "step": 64918 + }, + { + "epoch": 4.8, + "learning_rate": 2.03506739368847e-06, + "loss": 0.9852, + "step": 64919 + }, + { + "epoch": 4.8, + "learning_rate": 2.0348262233341455e-06, + "loss": 0.9842, + "step": 64920 + }, + { + "epoch": 4.8, + "learning_rate": 2.0345850656523012e-06, + "loss": 1.0769, + "step": 64921 + }, + { + "epoch": 4.8, + "learning_rate": 2.0343439206433192e-06, + "loss": 0.9894, + "step": 64922 + }, + { + "epoch": 4.8, + "learning_rate": 2.0341027883075815e-06, + "loss": 0.9656, + "step": 64923 + }, + { + "epoch": 4.8, + "learning_rate": 2.033861668645469e-06, + "loss": 0.9694, + "step": 64924 + }, + { + "epoch": 4.8, + "learning_rate": 2.0336205616573724e-06, + "loss": 1.1204, + "step": 64925 + }, + { + "epoch": 4.8, + "learning_rate": 2.03337946734367e-06, + "loss": 1.0063, + "step": 64926 + }, + { + "epoch": 4.8, + "learning_rate": 2.0331383857047482e-06, + "loss": 0.9528, + "step": 64927 + }, + { + "epoch": 4.8, + "learning_rate": 2.032897316740989e-06, + "loss": 0.929, + "step": 64928 + }, + { + "epoch": 4.8, + "learning_rate": 2.032656260452773e-06, + "loss": 1.0116, + "step": 64929 + }, + { + "epoch": 4.8, + "learning_rate": 2.03241521684049e-06, + "loss": 0.9098, + "step": 64930 + }, + { + "epoch": 4.8, + "learning_rate": 2.0321741859045207e-06, + "loss": 0.9615, + "step": 64931 + }, + { + "epoch": 4.8, + "learning_rate": 2.0319331676452482e-06, + "loss": 1.0156, + "step": 64932 + }, + { + "epoch": 4.8, + "learning_rate": 2.0316921620630524e-06, + "loss": 0.9352, + "step": 64933 + }, + { + "epoch": 4.8, + "learning_rate": 2.0314511691583237e-06, + "loss": 1.0633, + "step": 64934 + }, + { + "epoch": 4.8, + "learning_rate": 2.031210188931445e-06, + "loss": 0.9965, + "step": 64935 + }, + { + "epoch": 4.8, + "learning_rate": 2.0309692213827913e-06, + "loss": 1.031, + "step": 64936 + }, + { + "epoch": 4.8, + "learning_rate": 2.0307282665127537e-06, + "loss": 1.0201, + "step": 64937 + }, + { + "epoch": 4.8, + "learning_rate": 2.0304873243217116e-06, + "loss": 1.0442, + "step": 64938 + }, + { + "epoch": 4.8, + "learning_rate": 2.030246394810055e-06, + "loss": 1.0258, + "step": 64939 + }, + { + "epoch": 4.8, + "learning_rate": 2.0300054779781574e-06, + "loss": 0.9644, + "step": 64940 + }, + { + "epoch": 4.8, + "learning_rate": 2.0297645738264096e-06, + "loss": 0.9462, + "step": 64941 + }, + { + "epoch": 4.8, + "learning_rate": 2.029523682355192e-06, + "loss": 1.0489, + "step": 64942 + }, + { + "epoch": 4.8, + "learning_rate": 2.029282803564888e-06, + "loss": 0.937, + "step": 64943 + }, + { + "epoch": 4.8, + "learning_rate": 2.0290419374558824e-06, + "loss": 1.0046, + "step": 64944 + }, + { + "epoch": 4.8, + "learning_rate": 2.028801084028553e-06, + "loss": 0.8432, + "step": 64945 + }, + { + "epoch": 4.8, + "learning_rate": 2.028560243283291e-06, + "loss": 1.063, + "step": 64946 + }, + { + "epoch": 4.8, + "learning_rate": 2.0283194152204746e-06, + "loss": 0.9643, + "step": 64947 + }, + { + "epoch": 4.8, + "learning_rate": 2.028078599840487e-06, + "loss": 1.0247, + "step": 64948 + }, + { + "epoch": 4.8, + "learning_rate": 2.0278377971437115e-06, + "loss": 0.9165, + "step": 64949 + }, + { + "epoch": 4.8, + "learning_rate": 2.027597007130533e-06, + "loss": 0.9421, + "step": 64950 + }, + { + "epoch": 4.8, + "learning_rate": 2.027356229801335e-06, + "loss": 0.962, + "step": 64951 + }, + { + "epoch": 4.8, + "learning_rate": 2.0271154651564984e-06, + "loss": 0.9934, + "step": 64952 + }, + { + "epoch": 4.8, + "learning_rate": 2.026874713196407e-06, + "loss": 0.8433, + "step": 64953 + }, + { + "epoch": 4.8, + "learning_rate": 2.0266339739214415e-06, + "loss": 1.003, + "step": 64954 + }, + { + "epoch": 4.8, + "learning_rate": 2.0263932473319924e-06, + "loss": 1.0131, + "step": 64955 + }, + { + "epoch": 4.8, + "learning_rate": 2.0261525334284325e-06, + "loss": 1.0086, + "step": 64956 + }, + { + "epoch": 4.8, + "learning_rate": 2.0259118322111527e-06, + "loss": 0.9595, + "step": 64957 + }, + { + "epoch": 4.8, + "learning_rate": 2.02567114368053e-06, + "loss": 1.0619, + "step": 64958 + }, + { + "epoch": 4.8, + "learning_rate": 2.025430467836955e-06, + "loss": 1.0381, + "step": 64959 + }, + { + "epoch": 4.8, + "learning_rate": 2.0251898046808014e-06, + "loss": 0.9776, + "step": 64960 + }, + { + "epoch": 4.8, + "learning_rate": 2.02494915421246e-06, + "loss": 1.0482, + "step": 64961 + }, + { + "epoch": 4.8, + "learning_rate": 2.0247085164323098e-06, + "loss": 1.03, + "step": 64962 + }, + { + "epoch": 4.8, + "learning_rate": 2.0244678913407346e-06, + "loss": 1.0732, + "step": 64963 + }, + { + "epoch": 4.8, + "learning_rate": 2.0242272789381166e-06, + "loss": 0.9126, + "step": 64964 + }, + { + "epoch": 4.8, + "learning_rate": 2.0239866792248365e-06, + "loss": 0.9024, + "step": 64965 + }, + { + "epoch": 4.8, + "learning_rate": 2.023746092201282e-06, + "loss": 1.0118, + "step": 64966 + }, + { + "epoch": 4.8, + "learning_rate": 2.0235055178678333e-06, + "loss": 1.0275, + "step": 64967 + }, + { + "epoch": 4.8, + "learning_rate": 2.023264956224873e-06, + "loss": 1.0383, + "step": 64968 + }, + { + "epoch": 4.8, + "learning_rate": 2.023024407272781e-06, + "loss": 1.0194, + "step": 64969 + }, + { + "epoch": 4.8, + "learning_rate": 2.022783871011946e-06, + "loss": 1.0108, + "step": 64970 + }, + { + "epoch": 4.8, + "learning_rate": 2.022543347442748e-06, + "loss": 0.9336, + "step": 64971 + }, + { + "epoch": 4.8, + "learning_rate": 2.0223028365655696e-06, + "loss": 1.0222, + "step": 64972 + }, + { + "epoch": 4.8, + "learning_rate": 2.0220623383807926e-06, + "loss": 1.009, + "step": 64973 + }, + { + "epoch": 4.8, + "learning_rate": 2.021821852888798e-06, + "loss": 1.031, + "step": 64974 + }, + { + "epoch": 4.8, + "learning_rate": 2.021581380089973e-06, + "loss": 1.0741, + "step": 64975 + }, + { + "epoch": 4.8, + "learning_rate": 2.021340919984699e-06, + "loss": 1.0597, + "step": 64976 + }, + { + "epoch": 4.8, + "learning_rate": 2.021100472573356e-06, + "loss": 1.0552, + "step": 64977 + }, + { + "epoch": 4.8, + "learning_rate": 2.020860037856326e-06, + "loss": 1.1209, + "step": 64978 + }, + { + "epoch": 4.8, + "learning_rate": 2.0206196158339966e-06, + "loss": 1.0376, + "step": 64979 + }, + { + "epoch": 4.8, + "learning_rate": 2.0203792065067463e-06, + "loss": 1.0309, + "step": 64980 + }, + { + "epoch": 4.8, + "learning_rate": 2.0201388098749588e-06, + "loss": 0.9953, + "step": 64981 + }, + { + "epoch": 4.8, + "learning_rate": 2.0198984259390165e-06, + "loss": 0.9834, + "step": 64982 + }, + { + "epoch": 4.8, + "learning_rate": 2.0196580546993015e-06, + "loss": 0.9699, + "step": 64983 + }, + { + "epoch": 4.8, + "learning_rate": 2.019417696156196e-06, + "loss": 1.003, + "step": 64984 + }, + { + "epoch": 4.8, + "learning_rate": 2.019177350310081e-06, + "loss": 0.9875, + "step": 64985 + }, + { + "epoch": 4.8, + "learning_rate": 2.018937017161343e-06, + "loss": 1.0664, + "step": 64986 + }, + { + "epoch": 4.8, + "learning_rate": 2.0186966967103618e-06, + "loss": 1.066, + "step": 64987 + }, + { + "epoch": 4.8, + "learning_rate": 2.0184563889575204e-06, + "loss": 0.9798, + "step": 64988 + }, + { + "epoch": 4.8, + "learning_rate": 2.018216093903197e-06, + "loss": 0.926, + "step": 64989 + }, + { + "epoch": 4.8, + "learning_rate": 2.017975811547781e-06, + "loss": 1.0919, + "step": 64990 + }, + { + "epoch": 4.8, + "learning_rate": 2.017735541891651e-06, + "loss": 1.0841, + "step": 64991 + }, + { + "epoch": 4.8, + "learning_rate": 2.0174952849351904e-06, + "loss": 0.9951, + "step": 64992 + }, + { + "epoch": 4.8, + "learning_rate": 2.0172550406787793e-06, + "loss": 0.9595, + "step": 64993 + }, + { + "epoch": 4.8, + "learning_rate": 2.017014809122799e-06, + "loss": 0.9291, + "step": 64994 + }, + { + "epoch": 4.8, + "learning_rate": 2.0167745902676373e-06, + "loss": 1.0277, + "step": 64995 + }, + { + "epoch": 4.8, + "learning_rate": 2.016534384113672e-06, + "loss": 1.0114, + "step": 64996 + }, + { + "epoch": 4.8, + "learning_rate": 2.0162941906612866e-06, + "loss": 1.1043, + "step": 64997 + }, + { + "epoch": 4.8, + "learning_rate": 2.0160540099108607e-06, + "loss": 1.1145, + "step": 64998 + }, + { + "epoch": 4.8, + "learning_rate": 2.0158138418627805e-06, + "loss": 1.0016, + "step": 64999 + }, + { + "epoch": 4.8, + "learning_rate": 2.015573686517429e-06, + "loss": 0.901, + "step": 65000 + }, + { + "epoch": 4.8, + "learning_rate": 2.015333543875181e-06, + "loss": 0.962, + "step": 65001 + }, + { + "epoch": 4.8, + "learning_rate": 2.015093413936425e-06, + "loss": 0.9257, + "step": 65002 + }, + { + "epoch": 4.8, + "learning_rate": 2.014853296701539e-06, + "loss": 1.0033, + "step": 65003 + }, + { + "epoch": 4.8, + "learning_rate": 2.0146131921709123e-06, + "loss": 0.9841, + "step": 65004 + }, + { + "epoch": 4.8, + "learning_rate": 2.0143731003449164e-06, + "loss": 0.9745, + "step": 65005 + }, + { + "epoch": 4.8, + "learning_rate": 2.0141330212239417e-06, + "loss": 1.0607, + "step": 65006 + }, + { + "epoch": 4.8, + "learning_rate": 2.013892954808364e-06, + "loss": 1.0017, + "step": 65007 + }, + { + "epoch": 4.8, + "learning_rate": 2.013652901098574e-06, + "loss": 1.0083, + "step": 65008 + }, + { + "epoch": 4.8, + "learning_rate": 2.0134128600949445e-06, + "loss": 1.0995, + "step": 65009 + }, + { + "epoch": 4.8, + "learning_rate": 2.0131728317978596e-06, + "loss": 1.0495, + "step": 65010 + }, + { + "epoch": 4.8, + "learning_rate": 2.0129328162077033e-06, + "loss": 0.979, + "step": 65011 + }, + { + "epoch": 4.8, + "learning_rate": 2.012692813324858e-06, + "loss": 1.0611, + "step": 65012 + }, + { + "epoch": 4.8, + "learning_rate": 2.012452823149703e-06, + "loss": 0.9709, + "step": 65013 + }, + { + "epoch": 4.8, + "learning_rate": 2.0122128456826195e-06, + "loss": 1.0257, + "step": 65014 + }, + { + "epoch": 4.8, + "learning_rate": 2.0119728809239934e-06, + "loss": 1.0386, + "step": 65015 + }, + { + "epoch": 4.8, + "learning_rate": 2.011732928874205e-06, + "loss": 1.0661, + "step": 65016 + }, + { + "epoch": 4.8, + "learning_rate": 2.011492989533634e-06, + "loss": 1.0021, + "step": 65017 + }, + { + "epoch": 4.8, + "learning_rate": 2.011253062902664e-06, + "loss": 1.0846, + "step": 65018 + }, + { + "epoch": 4.8, + "learning_rate": 2.0110131489816732e-06, + "loss": 0.961, + "step": 65019 + }, + { + "epoch": 4.8, + "learning_rate": 2.0107732477710494e-06, + "loss": 0.9337, + "step": 65020 + }, + { + "epoch": 4.8, + "learning_rate": 2.0105333592711705e-06, + "loss": 0.9575, + "step": 65021 + }, + { + "epoch": 4.8, + "learning_rate": 2.0102934834824185e-06, + "loss": 1.0654, + "step": 65022 + }, + { + "epoch": 4.8, + "learning_rate": 2.0100536204051735e-06, + "loss": 1.0155, + "step": 65023 + }, + { + "epoch": 4.8, + "learning_rate": 2.0098137700398203e-06, + "loss": 1.0282, + "step": 65024 + }, + { + "epoch": 4.8, + "learning_rate": 2.00957393238674e-06, + "loss": 1.0454, + "step": 65025 + }, + { + "epoch": 4.8, + "learning_rate": 2.0093341074463137e-06, + "loss": 1.0468, + "step": 65026 + }, + { + "epoch": 4.8, + "learning_rate": 2.009094295218921e-06, + "loss": 0.9318, + "step": 65027 + }, + { + "epoch": 4.8, + "learning_rate": 2.008854495704946e-06, + "loss": 1.0068, + "step": 65028 + }, + { + "epoch": 4.8, + "learning_rate": 2.0086147089047694e-06, + "loss": 0.9737, + "step": 65029 + }, + { + "epoch": 4.81, + "learning_rate": 2.0083749348187688e-06, + "loss": 1.0117, + "step": 65030 + }, + { + "epoch": 4.81, + "learning_rate": 2.008135173447332e-06, + "loss": 0.8985, + "step": 65031 + }, + { + "epoch": 4.81, + "learning_rate": 2.007895424790839e-06, + "loss": 0.9728, + "step": 65032 + }, + { + "epoch": 4.81, + "learning_rate": 2.007655688849669e-06, + "loss": 1.0003, + "step": 65033 + }, + { + "epoch": 4.81, + "learning_rate": 2.007415965624202e-06, + "loss": 1.0321, + "step": 65034 + }, + { + "epoch": 4.81, + "learning_rate": 2.007176255114824e-06, + "loss": 0.9859, + "step": 65035 + }, + { + "epoch": 4.81, + "learning_rate": 2.006936557321916e-06, + "loss": 1.1317, + "step": 65036 + }, + { + "epoch": 4.81, + "learning_rate": 2.006696872245856e-06, + "loss": 0.9628, + "step": 65037 + }, + { + "epoch": 4.81, + "learning_rate": 2.0064571998870273e-06, + "loss": 0.9198, + "step": 65038 + }, + { + "epoch": 4.81, + "learning_rate": 2.0062175402458086e-06, + "loss": 1.0329, + "step": 65039 + }, + { + "epoch": 4.81, + "learning_rate": 2.005977893322585e-06, + "loss": 0.9683, + "step": 65040 + }, + { + "epoch": 4.81, + "learning_rate": 2.0057382591177364e-06, + "loss": 1.1384, + "step": 65041 + }, + { + "epoch": 4.81, + "learning_rate": 2.0054986376316445e-06, + "loss": 0.9249, + "step": 65042 + }, + { + "epoch": 4.81, + "learning_rate": 2.0052590288646878e-06, + "loss": 0.9716, + "step": 65043 + }, + { + "epoch": 4.81, + "learning_rate": 2.0050194328172513e-06, + "loss": 1.0911, + "step": 65044 + }, + { + "epoch": 4.81, + "learning_rate": 2.004779849489714e-06, + "loss": 0.9287, + "step": 65045 + }, + { + "epoch": 4.81, + "learning_rate": 2.0045402788824586e-06, + "loss": 0.9682, + "step": 65046 + }, + { + "epoch": 4.81, + "learning_rate": 2.004300720995865e-06, + "loss": 1.0293, + "step": 65047 + }, + { + "epoch": 4.81, + "learning_rate": 2.004061175830312e-06, + "loss": 1.0062, + "step": 65048 + }, + { + "epoch": 4.81, + "learning_rate": 2.0038216433861878e-06, + "loss": 1.0231, + "step": 65049 + }, + { + "epoch": 4.81, + "learning_rate": 2.003582123663864e-06, + "loss": 0.8783, + "step": 65050 + }, + { + "epoch": 4.81, + "learning_rate": 2.0033426166637295e-06, + "loss": 0.9122, + "step": 65051 + }, + { + "epoch": 4.81, + "learning_rate": 2.00310312238616e-06, + "loss": 1.0318, + "step": 65052 + }, + { + "epoch": 4.81, + "learning_rate": 2.002863640831544e-06, + "loss": 0.9991, + "step": 65053 + }, + { + "epoch": 4.81, + "learning_rate": 2.0026241720002525e-06, + "loss": 1.0109, + "step": 65054 + }, + { + "epoch": 4.81, + "learning_rate": 2.0023847158926747e-06, + "loss": 1.0037, + "step": 65055 + }, + { + "epoch": 4.81, + "learning_rate": 2.0021452725091873e-06, + "loss": 1.0041, + "step": 65056 + }, + { + "epoch": 4.81, + "learning_rate": 2.0019058418501726e-06, + "loss": 1.0935, + "step": 65057 + }, + { + "epoch": 4.81, + "learning_rate": 2.001666423916011e-06, + "loss": 0.9468, + "step": 65058 + }, + { + "epoch": 4.81, + "learning_rate": 2.0014270187070815e-06, + "loss": 1.0545, + "step": 65059 + }, + { + "epoch": 4.81, + "learning_rate": 2.0011876262237707e-06, + "loss": 0.974, + "step": 65060 + }, + { + "epoch": 4.81, + "learning_rate": 2.0009482464664543e-06, + "loss": 1.0741, + "step": 65061 + }, + { + "epoch": 4.81, + "learning_rate": 2.0007088794355154e-06, + "loss": 0.9372, + "step": 65062 + }, + { + "epoch": 4.81, + "learning_rate": 2.000469525131332e-06, + "loss": 1.0168, + "step": 65063 + }, + { + "epoch": 4.81, + "learning_rate": 2.0002301835542893e-06, + "loss": 0.9269, + "step": 65064 + }, + { + "epoch": 4.81, + "learning_rate": 1.9999908547047665e-06, + "loss": 0.9467, + "step": 65065 + }, + { + "epoch": 4.81, + "learning_rate": 1.9997515385831433e-06, + "loss": 0.886, + "step": 65066 + }, + { + "epoch": 4.81, + "learning_rate": 1.999512235189801e-06, + "loss": 0.9982, + "step": 65067 + }, + { + "epoch": 4.81, + "learning_rate": 1.999272944525117e-06, + "loss": 1.074, + "step": 65068 + }, + { + "epoch": 4.81, + "learning_rate": 1.9990336665894784e-06, + "loss": 0.9672, + "step": 65069 + }, + { + "epoch": 4.81, + "learning_rate": 1.998794401383264e-06, + "loss": 1.0072, + "step": 65070 + }, + { + "epoch": 4.81, + "learning_rate": 1.998555148906851e-06, + "loss": 1.1222, + "step": 65071 + }, + { + "epoch": 4.81, + "learning_rate": 1.9983159091606207e-06, + "loss": 1.0175, + "step": 65072 + }, + { + "epoch": 4.81, + "learning_rate": 1.9980766821449615e-06, + "loss": 1.1201, + "step": 65073 + }, + { + "epoch": 4.81, + "learning_rate": 1.997837467860244e-06, + "loss": 1.0626, + "step": 65074 + }, + { + "epoch": 4.81, + "learning_rate": 1.9975982663068504e-06, + "loss": 1.0764, + "step": 65075 + }, + { + "epoch": 4.81, + "learning_rate": 1.997359077485166e-06, + "loss": 0.9632, + "step": 65076 + }, + { + "epoch": 4.81, + "learning_rate": 1.997119901395569e-06, + "loss": 1.0238, + "step": 65077 + }, + { + "epoch": 4.81, + "learning_rate": 1.9968807380384393e-06, + "loss": 1.0273, + "step": 65078 + }, + { + "epoch": 4.81, + "learning_rate": 1.9966415874141553e-06, + "loss": 0.968, + "step": 65079 + }, + { + "epoch": 4.81, + "learning_rate": 1.9964024495231025e-06, + "loss": 0.9878, + "step": 65080 + }, + { + "epoch": 4.81, + "learning_rate": 1.99616332436566e-06, + "loss": 0.9858, + "step": 65081 + }, + { + "epoch": 4.81, + "learning_rate": 1.995924211942206e-06, + "loss": 1.1438, + "step": 65082 + }, + { + "epoch": 4.81, + "learning_rate": 1.995685112253122e-06, + "loss": 1.0065, + "step": 65083 + }, + { + "epoch": 4.81, + "learning_rate": 1.995446025298786e-06, + "loss": 0.9647, + "step": 65084 + }, + { + "epoch": 4.81, + "learning_rate": 1.9952069510795837e-06, + "loss": 0.9245, + "step": 65085 + }, + { + "epoch": 4.81, + "learning_rate": 1.9949678895958934e-06, + "loss": 1.0381, + "step": 65086 + }, + { + "epoch": 4.81, + "learning_rate": 1.9947288408480936e-06, + "loss": 0.9921, + "step": 65087 + }, + { + "epoch": 4.81, + "learning_rate": 1.9944898048365636e-06, + "loss": 0.9118, + "step": 65088 + }, + { + "epoch": 4.81, + "learning_rate": 1.9942507815616872e-06, + "loss": 0.9697, + "step": 65089 + }, + { + "epoch": 4.81, + "learning_rate": 1.9940117710238447e-06, + "loss": 1.0444, + "step": 65090 + }, + { + "epoch": 4.81, + "learning_rate": 1.9937727732234148e-06, + "loss": 0.9167, + "step": 65091 + }, + { + "epoch": 4.81, + "learning_rate": 1.9935337881607743e-06, + "loss": 0.9773, + "step": 65092 + }, + { + "epoch": 4.81, + "learning_rate": 1.9932948158363095e-06, + "loss": 0.9052, + "step": 65093 + }, + { + "epoch": 4.81, + "learning_rate": 1.993055856250402e-06, + "loss": 0.9315, + "step": 65094 + }, + { + "epoch": 4.81, + "learning_rate": 1.9928169094034223e-06, + "loss": 0.8988, + "step": 65095 + }, + { + "epoch": 4.81, + "learning_rate": 1.9925779752957585e-06, + "loss": 1.006, + "step": 65096 + }, + { + "epoch": 4.81, + "learning_rate": 1.9923390539277866e-06, + "loss": 0.868, + "step": 65097 + }, + { + "epoch": 4.81, + "learning_rate": 1.9921001452998932e-06, + "loss": 1.0723, + "step": 65098 + }, + { + "epoch": 4.81, + "learning_rate": 1.991861249412449e-06, + "loss": 0.9813, + "step": 65099 + }, + { + "epoch": 4.81, + "learning_rate": 1.991622366265842e-06, + "loss": 1.0903, + "step": 65100 + }, + { + "epoch": 4.81, + "learning_rate": 1.991383495860446e-06, + "loss": 0.8874, + "step": 65101 + }, + { + "epoch": 4.81, + "learning_rate": 1.9911446381966504e-06, + "loss": 0.9713, + "step": 65102 + }, + { + "epoch": 4.81, + "learning_rate": 1.9909057932748254e-06, + "loss": 0.9434, + "step": 65103 + }, + { + "epoch": 4.81, + "learning_rate": 1.990666961095353e-06, + "loss": 0.9279, + "step": 65104 + }, + { + "epoch": 4.81, + "learning_rate": 1.9904281416586168e-06, + "loss": 1.0289, + "step": 65105 + }, + { + "epoch": 4.81, + "learning_rate": 1.9901893349649946e-06, + "loss": 0.8734, + "step": 65106 + }, + { + "epoch": 4.81, + "learning_rate": 1.989950541014868e-06, + "loss": 1.0527, + "step": 65107 + }, + { + "epoch": 4.81, + "learning_rate": 1.9897117598086113e-06, + "loss": 1.0092, + "step": 65108 + }, + { + "epoch": 4.81, + "learning_rate": 1.9894729913466115e-06, + "loss": 1.106, + "step": 65109 + }, + { + "epoch": 4.81, + "learning_rate": 1.989234235629246e-06, + "loss": 0.9387, + "step": 65110 + }, + { + "epoch": 4.81, + "learning_rate": 1.9889954926568945e-06, + "loss": 0.8573, + "step": 65111 + }, + { + "epoch": 4.81, + "learning_rate": 1.9887567624299365e-06, + "loss": 1.0429, + "step": 65112 + }, + { + "epoch": 4.81, + "learning_rate": 1.9885180449487484e-06, + "loss": 0.9714, + "step": 65113 + }, + { + "epoch": 4.81, + "learning_rate": 1.9882793402137167e-06, + "loss": 1.1149, + "step": 65114 + }, + { + "epoch": 4.81, + "learning_rate": 1.988040648225218e-06, + "loss": 1.0057, + "step": 65115 + }, + { + "epoch": 4.81, + "learning_rate": 1.9878019689836316e-06, + "loss": 0.9726, + "step": 65116 + }, + { + "epoch": 4.81, + "learning_rate": 1.9875633024893346e-06, + "loss": 1.0397, + "step": 65117 + }, + { + "epoch": 4.81, + "learning_rate": 1.9873246487427167e-06, + "loss": 0.9826, + "step": 65118 + }, + { + "epoch": 4.81, + "learning_rate": 1.9870860077441436e-06, + "loss": 0.952, + "step": 65119 + }, + { + "epoch": 4.81, + "learning_rate": 1.986847379494006e-06, + "loss": 0.9991, + "step": 65120 + }, + { + "epoch": 4.81, + "learning_rate": 1.986608763992679e-06, + "loss": 0.9868, + "step": 65121 + }, + { + "epoch": 4.81, + "learning_rate": 1.986370161240542e-06, + "loss": 1.0507, + "step": 65122 + }, + { + "epoch": 4.81, + "learning_rate": 1.986131571237976e-06, + "loss": 1.0828, + "step": 65123 + }, + { + "epoch": 4.81, + "learning_rate": 1.985892993985358e-06, + "loss": 0.9892, + "step": 65124 + }, + { + "epoch": 4.81, + "learning_rate": 1.9856544294830714e-06, + "loss": 0.9033, + "step": 65125 + }, + { + "epoch": 4.81, + "learning_rate": 1.9854158777314947e-06, + "loss": 0.8963, + "step": 65126 + }, + { + "epoch": 4.81, + "learning_rate": 1.9851773387310058e-06, + "loss": 1.1247, + "step": 65127 + }, + { + "epoch": 4.81, + "learning_rate": 1.9849388124819834e-06, + "loss": 1.1233, + "step": 65128 + }, + { + "epoch": 4.81, + "learning_rate": 1.9847002989848107e-06, + "loss": 0.8929, + "step": 65129 + }, + { + "epoch": 4.81, + "learning_rate": 1.9844617982398653e-06, + "loss": 0.9994, + "step": 65130 + }, + { + "epoch": 4.81, + "learning_rate": 1.984223310247526e-06, + "loss": 1.1171, + "step": 65131 + }, + { + "epoch": 4.81, + "learning_rate": 1.9839848350081736e-06, + "loss": 1.0662, + "step": 65132 + }, + { + "epoch": 4.81, + "learning_rate": 1.983746372522183e-06, + "loss": 0.9933, + "step": 65133 + }, + { + "epoch": 4.81, + "learning_rate": 1.9835079227899403e-06, + "loss": 0.9696, + "step": 65134 + }, + { + "epoch": 4.81, + "learning_rate": 1.9832694858118217e-06, + "loss": 0.9722, + "step": 65135 + }, + { + "epoch": 4.81, + "learning_rate": 1.9830310615882074e-06, + "loss": 0.9734, + "step": 65136 + }, + { + "epoch": 4.81, + "learning_rate": 1.982792650119473e-06, + "loss": 1.0964, + "step": 65137 + }, + { + "epoch": 4.81, + "learning_rate": 1.9825542514060024e-06, + "loss": 0.9923, + "step": 65138 + }, + { + "epoch": 4.81, + "learning_rate": 1.9823158654481744e-06, + "loss": 0.8624, + "step": 65139 + }, + { + "epoch": 4.81, + "learning_rate": 1.982077492246367e-06, + "loss": 1.0584, + "step": 65140 + }, + { + "epoch": 4.81, + "learning_rate": 1.98183913180096e-06, + "loss": 0.9577, + "step": 65141 + }, + { + "epoch": 4.81, + "learning_rate": 1.9816007841123285e-06, + "loss": 0.9567, + "step": 65142 + }, + { + "epoch": 4.81, + "learning_rate": 1.9813624491808613e-06, + "loss": 0.9692, + "step": 65143 + }, + { + "epoch": 4.81, + "learning_rate": 1.981124127006926e-06, + "loss": 0.8784, + "step": 65144 + }, + { + "epoch": 4.81, + "learning_rate": 1.9808858175909106e-06, + "loss": 0.971, + "step": 65145 + }, + { + "epoch": 4.81, + "learning_rate": 1.9806475209331886e-06, + "loss": 0.9673, + "step": 65146 + }, + { + "epoch": 4.81, + "learning_rate": 1.9804092370341465e-06, + "loss": 1.0174, + "step": 65147 + }, + { + "epoch": 4.81, + "learning_rate": 1.9801709658941535e-06, + "loss": 0.9944, + "step": 65148 + }, + { + "epoch": 4.81, + "learning_rate": 1.9799327075135964e-06, + "loss": 0.9688, + "step": 65149 + }, + { + "epoch": 4.81, + "learning_rate": 1.9796944618928515e-06, + "loss": 1.0114, + "step": 65150 + }, + { + "epoch": 4.81, + "learning_rate": 1.9794562290322984e-06, + "loss": 1.0082, + "step": 65151 + }, + { + "epoch": 4.81, + "learning_rate": 1.9792180089323153e-06, + "loss": 1.0136, + "step": 65152 + }, + { + "epoch": 4.81, + "learning_rate": 1.978979801593279e-06, + "loss": 1.041, + "step": 65153 + }, + { + "epoch": 4.81, + "learning_rate": 1.978741607015574e-06, + "loss": 0.9412, + "step": 65154 + }, + { + "epoch": 4.81, + "learning_rate": 1.978503425199576e-06, + "loss": 1.0316, + "step": 65155 + }, + { + "epoch": 4.81, + "learning_rate": 1.9782652561456638e-06, + "loss": 0.9995, + "step": 65156 + }, + { + "epoch": 4.81, + "learning_rate": 1.978027099854215e-06, + "loss": 0.9974, + "step": 65157 + }, + { + "epoch": 4.81, + "learning_rate": 1.9777889563256127e-06, + "loss": 0.9732, + "step": 65158 + }, + { + "epoch": 4.81, + "learning_rate": 1.9775508255602337e-06, + "loss": 0.9355, + "step": 65159 + }, + { + "epoch": 4.81, + "learning_rate": 1.9773127075584553e-06, + "loss": 0.862, + "step": 65160 + }, + { + "epoch": 4.81, + "learning_rate": 1.977074602320659e-06, + "loss": 1.0636, + "step": 65161 + }, + { + "epoch": 4.81, + "learning_rate": 1.9768365098472185e-06, + "loss": 1.0203, + "step": 65162 + }, + { + "epoch": 4.81, + "learning_rate": 1.976598430138522e-06, + "loss": 0.9521, + "step": 65163 + }, + { + "epoch": 4.81, + "learning_rate": 1.976360363194938e-06, + "loss": 1.0443, + "step": 65164 + }, + { + "epoch": 4.81, + "learning_rate": 1.9761223090168525e-06, + "loss": 0.9447, + "step": 65165 + }, + { + "epoch": 4.82, + "learning_rate": 1.9758842676046386e-06, + "loss": 0.9187, + "step": 65166 + }, + { + "epoch": 4.82, + "learning_rate": 1.975646238958683e-06, + "loss": 1.0457, + "step": 65167 + }, + { + "epoch": 4.82, + "learning_rate": 1.975408223079357e-06, + "loss": 1.0847, + "step": 65168 + }, + { + "epoch": 4.82, + "learning_rate": 1.9751702199670387e-06, + "loss": 1.0392, + "step": 65169 + }, + { + "epoch": 4.82, + "learning_rate": 1.974932229622113e-06, + "loss": 1.0352, + "step": 65170 + }, + { + "epoch": 4.82, + "learning_rate": 1.974694252044954e-06, + "loss": 1.0645, + "step": 65171 + }, + { + "epoch": 4.82, + "learning_rate": 1.974456287235943e-06, + "loss": 0.932, + "step": 65172 + }, + { + "epoch": 4.82, + "learning_rate": 1.9742183351954548e-06, + "loss": 0.844, + "step": 65173 + }, + { + "epoch": 4.82, + "learning_rate": 1.973980395923872e-06, + "loss": 0.9728, + "step": 65174 + }, + { + "epoch": 4.82, + "learning_rate": 1.973742469421572e-06, + "loss": 0.9639, + "step": 65175 + }, + { + "epoch": 4.82, + "learning_rate": 1.9735045556889333e-06, + "loss": 0.9377, + "step": 65176 + }, + { + "epoch": 4.82, + "learning_rate": 1.9732666547263335e-06, + "loss": 0.9931, + "step": 65177 + }, + { + "epoch": 4.82, + "learning_rate": 1.9730287665341487e-06, + "loss": 1.0107, + "step": 65178 + }, + { + "epoch": 4.82, + "learning_rate": 1.9727908911127635e-06, + "loss": 0.9122, + "step": 65179 + }, + { + "epoch": 4.82, + "learning_rate": 1.9725530284625527e-06, + "loss": 1.0321, + "step": 65180 + }, + { + "epoch": 4.82, + "learning_rate": 1.9723151785838958e-06, + "loss": 1.0447, + "step": 65181 + }, + { + "epoch": 4.82, + "learning_rate": 1.9720773414771675e-06, + "loss": 1.0353, + "step": 65182 + }, + { + "epoch": 4.82, + "learning_rate": 1.9718395171427517e-06, + "loss": 0.9576, + "step": 65183 + }, + { + "epoch": 4.82, + "learning_rate": 1.9716017055810244e-06, + "loss": 1.0201, + "step": 65184 + }, + { + "epoch": 4.82, + "learning_rate": 1.971363906792365e-06, + "loss": 1.054, + "step": 65185 + }, + { + "epoch": 4.82, + "learning_rate": 1.9711261207771493e-06, + "loss": 1.1029, + "step": 65186 + }, + { + "epoch": 4.82, + "learning_rate": 1.9708883475357554e-06, + "loss": 0.9821, + "step": 65187 + }, + { + "epoch": 4.82, + "learning_rate": 1.9706505870685678e-06, + "loss": 1.0749, + "step": 65188 + }, + { + "epoch": 4.82, + "learning_rate": 1.970412839375956e-06, + "loss": 1.0087, + "step": 65189 + }, + { + "epoch": 4.82, + "learning_rate": 1.970175104458304e-06, + "loss": 0.9559, + "step": 65190 + }, + { + "epoch": 4.82, + "learning_rate": 1.9699373823159863e-06, + "loss": 1.1215, + "step": 65191 + }, + { + "epoch": 4.82, + "learning_rate": 1.969699672949389e-06, + "loss": 0.9154, + "step": 65192 + }, + { + "epoch": 4.82, + "learning_rate": 1.9694619763588786e-06, + "loss": 1.0577, + "step": 65193 + }, + { + "epoch": 4.82, + "learning_rate": 1.9692242925448422e-06, + "loss": 0.9722, + "step": 65194 + }, + { + "epoch": 4.82, + "learning_rate": 1.968986621507656e-06, + "loss": 1.0189, + "step": 65195 + }, + { + "epoch": 4.82, + "learning_rate": 1.9687489632476963e-06, + "loss": 0.988, + "step": 65196 + }, + { + "epoch": 4.82, + "learning_rate": 1.9685113177653427e-06, + "loss": 0.9705, + "step": 65197 + }, + { + "epoch": 4.82, + "learning_rate": 1.9682736850609686e-06, + "loss": 1.031, + "step": 65198 + }, + { + "epoch": 4.82, + "learning_rate": 1.9680360651349594e-06, + "loss": 0.9894, + "step": 65199 + }, + { + "epoch": 4.82, + "learning_rate": 1.9677984579876906e-06, + "loss": 0.9994, + "step": 65200 + }, + { + "epoch": 4.82, + "learning_rate": 1.96756086361954e-06, + "loss": 1.0292, + "step": 65201 + }, + { + "epoch": 4.82, + "learning_rate": 1.967323282030881e-06, + "loss": 1.0801, + "step": 65202 + }, + { + "epoch": 4.82, + "learning_rate": 1.9670857132220987e-06, + "loss": 0.9952, + "step": 65203 + }, + { + "epoch": 4.82, + "learning_rate": 1.9668481571935683e-06, + "loss": 1.13, + "step": 65204 + }, + { + "epoch": 4.82, + "learning_rate": 1.9666106139456685e-06, + "loss": 0.9266, + "step": 65205 + }, + { + "epoch": 4.82, + "learning_rate": 1.9663730834787754e-06, + "loss": 1.1218, + "step": 65206 + }, + { + "epoch": 4.82, + "learning_rate": 1.9661355657932644e-06, + "loss": 1.0377, + "step": 65207 + }, + { + "epoch": 4.82, + "learning_rate": 1.965898060889523e-06, + "loss": 0.9815, + "step": 65208 + }, + { + "epoch": 4.82, + "learning_rate": 1.9656605687679177e-06, + "loss": 1.1189, + "step": 65209 + }, + { + "epoch": 4.82, + "learning_rate": 1.965423089428834e-06, + "loss": 0.9439, + "step": 65210 + }, + { + "epoch": 4.82, + "learning_rate": 1.965185622872644e-06, + "loss": 1.0415, + "step": 65211 + }, + { + "epoch": 4.82, + "learning_rate": 1.964948169099734e-06, + "loss": 1.0208, + "step": 65212 + }, + { + "epoch": 4.82, + "learning_rate": 1.9647107281104725e-06, + "loss": 0.9039, + "step": 65213 + }, + { + "epoch": 4.82, + "learning_rate": 1.964473299905243e-06, + "loss": 1.0503, + "step": 65214 + }, + { + "epoch": 4.82, + "learning_rate": 1.964235884484421e-06, + "loss": 1.0812, + "step": 65215 + }, + { + "epoch": 4.82, + "learning_rate": 1.9639984818483857e-06, + "loss": 0.888, + "step": 65216 + }, + { + "epoch": 4.82, + "learning_rate": 1.9637610919975136e-06, + "loss": 0.9278, + "step": 65217 + }, + { + "epoch": 4.82, + "learning_rate": 1.9635237149321794e-06, + "loss": 0.9725, + "step": 65218 + }, + { + "epoch": 4.82, + "learning_rate": 1.9632863506527665e-06, + "loss": 1.0264, + "step": 65219 + }, + { + "epoch": 4.82, + "learning_rate": 1.9630489991596514e-06, + "loss": 1.0473, + "step": 65220 + }, + { + "epoch": 4.82, + "learning_rate": 1.962811660453209e-06, + "loss": 0.9995, + "step": 65221 + }, + { + "epoch": 4.82, + "learning_rate": 1.9625743345338153e-06, + "loss": 1.0145, + "step": 65222 + }, + { + "epoch": 4.82, + "learning_rate": 1.9623370214018555e-06, + "loss": 1.0922, + "step": 65223 + }, + { + "epoch": 4.82, + "learning_rate": 1.9620997210577007e-06, + "loss": 1.0039, + "step": 65224 + }, + { + "epoch": 4.82, + "learning_rate": 1.9618624335017313e-06, + "loss": 1.0568, + "step": 65225 + }, + { + "epoch": 4.82, + "learning_rate": 1.9616251587343225e-06, + "loss": 1.1328, + "step": 65226 + }, + { + "epoch": 4.82, + "learning_rate": 1.961387896755852e-06, + "loss": 0.8558, + "step": 65227 + }, + { + "epoch": 4.82, + "learning_rate": 1.9611506475666995e-06, + "loss": 0.875, + "step": 65228 + }, + { + "epoch": 4.82, + "learning_rate": 1.960913411167242e-06, + "loss": 0.896, + "step": 65229 + }, + { + "epoch": 4.82, + "learning_rate": 1.960676187557856e-06, + "loss": 1.0184, + "step": 65230 + }, + { + "epoch": 4.82, + "learning_rate": 1.9604389767389175e-06, + "loss": 1.0657, + "step": 65231 + }, + { + "epoch": 4.82, + "learning_rate": 1.9602017787108074e-06, + "loss": 1.0006, + "step": 65232 + }, + { + "epoch": 4.82, + "learning_rate": 1.9599645934739044e-06, + "loss": 0.9842, + "step": 65233 + }, + { + "epoch": 4.82, + "learning_rate": 1.9597274210285767e-06, + "loss": 0.953, + "step": 65234 + }, + { + "epoch": 4.82, + "learning_rate": 1.95949026137521e-06, + "loss": 0.994, + "step": 65235 + }, + { + "epoch": 4.82, + "learning_rate": 1.959253114514177e-06, + "loss": 1.1133, + "step": 65236 + }, + { + "epoch": 4.82, + "learning_rate": 1.959015980445863e-06, + "loss": 0.9981, + "step": 65237 + }, + { + "epoch": 4.82, + "learning_rate": 1.958778859170634e-06, + "loss": 1.0691, + "step": 65238 + }, + { + "epoch": 4.82, + "learning_rate": 1.958541750688875e-06, + "loss": 1.0811, + "step": 65239 + }, + { + "epoch": 4.82, + "learning_rate": 1.958304655000961e-06, + "loss": 0.9747, + "step": 65240 + }, + { + "epoch": 4.82, + "learning_rate": 1.9580675721072696e-06, + "loss": 0.9133, + "step": 65241 + }, + { + "epoch": 4.82, + "learning_rate": 1.9578305020081777e-06, + "loss": 0.9278, + "step": 65242 + }, + { + "epoch": 4.82, + "learning_rate": 1.95759344470406e-06, + "loss": 0.9712, + "step": 65243 + }, + { + "epoch": 4.82, + "learning_rate": 1.957356400195298e-06, + "loss": 0.8882, + "step": 65244 + }, + { + "epoch": 4.82, + "learning_rate": 1.9571193684822678e-06, + "loss": 1.078, + "step": 65245 + }, + { + "epoch": 4.82, + "learning_rate": 1.9568823495653455e-06, + "loss": 1.0226, + "step": 65246 + }, + { + "epoch": 4.82, + "learning_rate": 1.956645343444905e-06, + "loss": 0.9411, + "step": 65247 + }, + { + "epoch": 4.82, + "learning_rate": 1.9564083501213295e-06, + "loss": 0.9094, + "step": 65248 + }, + { + "epoch": 4.82, + "learning_rate": 1.9561713695949934e-06, + "loss": 1.0806, + "step": 65249 + }, + { + "epoch": 4.82, + "learning_rate": 1.955934401866273e-06, + "loss": 1.0933, + "step": 65250 + }, + { + "epoch": 4.82, + "learning_rate": 1.9556974469355437e-06, + "loss": 0.9265, + "step": 65251 + }, + { + "epoch": 4.82, + "learning_rate": 1.955460504803187e-06, + "loss": 0.8219, + "step": 65252 + }, + { + "epoch": 4.82, + "learning_rate": 1.9552235754695815e-06, + "loss": 0.9739, + "step": 65253 + }, + { + "epoch": 4.82, + "learning_rate": 1.954986658935093e-06, + "loss": 1.0091, + "step": 65254 + }, + { + "epoch": 4.82, + "learning_rate": 1.9547497552001093e-06, + "loss": 0.9092, + "step": 65255 + }, + { + "epoch": 4.82, + "learning_rate": 1.954512864265001e-06, + "loss": 0.9779, + "step": 65256 + }, + { + "epoch": 4.82, + "learning_rate": 1.954275986130153e-06, + "loss": 0.8851, + "step": 65257 + }, + { + "epoch": 4.82, + "learning_rate": 1.954039120795931e-06, + "loss": 1.0499, + "step": 65258 + }, + { + "epoch": 4.82, + "learning_rate": 1.9538022682627213e-06, + "loss": 0.9784, + "step": 65259 + }, + { + "epoch": 4.82, + "learning_rate": 1.9535654285308938e-06, + "loss": 0.9829, + "step": 65260 + }, + { + "epoch": 4.82, + "learning_rate": 1.9533286016008334e-06, + "loss": 0.9195, + "step": 65261 + }, + { + "epoch": 4.82, + "learning_rate": 1.9530917874729095e-06, + "loss": 0.9693, + "step": 65262 + }, + { + "epoch": 4.82, + "learning_rate": 1.952854986147499e-06, + "loss": 0.9192, + "step": 65263 + }, + { + "epoch": 4.82, + "learning_rate": 1.9526181976249837e-06, + "loss": 0.8551, + "step": 65264 + }, + { + "epoch": 4.82, + "learning_rate": 1.952381421905738e-06, + "loss": 1.0149, + "step": 65265 + }, + { + "epoch": 4.82, + "learning_rate": 1.9521446589901382e-06, + "loss": 1.0641, + "step": 65266 + }, + { + "epoch": 4.82, + "learning_rate": 1.9519079088785576e-06, + "loss": 1.0259, + "step": 65267 + }, + { + "epoch": 4.82, + "learning_rate": 1.95167117157138e-06, + "loss": 1.036, + "step": 65268 + }, + { + "epoch": 4.82, + "learning_rate": 1.9514344470689783e-06, + "loss": 1.0715, + "step": 65269 + }, + { + "epoch": 4.82, + "learning_rate": 1.9511977353717293e-06, + "loss": 0.9008, + "step": 65270 + }, + { + "epoch": 4.82, + "learning_rate": 1.950961036480009e-06, + "loss": 1.1542, + "step": 65271 + }, + { + "epoch": 4.82, + "learning_rate": 1.950724350394192e-06, + "loss": 1.0552, + "step": 65272 + }, + { + "epoch": 4.82, + "learning_rate": 1.9504876771146606e-06, + "loss": 0.9762, + "step": 65273 + }, + { + "epoch": 4.82, + "learning_rate": 1.9502510166417875e-06, + "loss": 1.0739, + "step": 65274 + }, + { + "epoch": 4.82, + "learning_rate": 1.950014368975951e-06, + "loss": 0.9201, + "step": 65275 + }, + { + "epoch": 4.82, + "learning_rate": 1.9497777341175227e-06, + "loss": 1.0261, + "step": 65276 + }, + { + "epoch": 4.82, + "learning_rate": 1.949541112066886e-06, + "loss": 0.8786, + "step": 65277 + }, + { + "epoch": 4.82, + "learning_rate": 1.949304502824414e-06, + "loss": 1.0509, + "step": 65278 + }, + { + "epoch": 4.82, + "learning_rate": 1.9490679063904826e-06, + "loss": 1.0344, + "step": 65279 + }, + { + "epoch": 4.82, + "learning_rate": 1.94883132276547e-06, + "loss": 1.0412, + "step": 65280 + }, + { + "epoch": 4.82, + "learning_rate": 1.9485947519497484e-06, + "loss": 0.957, + "step": 65281 + }, + { + "epoch": 4.82, + "learning_rate": 1.9483581939437037e-06, + "loss": 0.9618, + "step": 65282 + }, + { + "epoch": 4.82, + "learning_rate": 1.9481216487476995e-06, + "loss": 1.0708, + "step": 65283 + }, + { + "epoch": 4.82, + "learning_rate": 1.9478851163621216e-06, + "loss": 1.0004, + "step": 65284 + }, + { + "epoch": 4.82, + "learning_rate": 1.947648596787343e-06, + "loss": 0.9767, + "step": 65285 + }, + { + "epoch": 4.82, + "learning_rate": 1.9474120900237404e-06, + "loss": 1.006, + "step": 65286 + }, + { + "epoch": 4.82, + "learning_rate": 1.9471755960716875e-06, + "loss": 1.0403, + "step": 65287 + }, + { + "epoch": 4.82, + "learning_rate": 1.9469391149315653e-06, + "loss": 0.9085, + "step": 65288 + }, + { + "epoch": 4.82, + "learning_rate": 1.9467026466037477e-06, + "loss": 0.9195, + "step": 65289 + }, + { + "epoch": 4.82, + "learning_rate": 1.9464661910886117e-06, + "loss": 1.0419, + "step": 65290 + }, + { + "epoch": 4.82, + "learning_rate": 1.946229748386532e-06, + "loss": 1.0119, + "step": 65291 + }, + { + "epoch": 4.82, + "learning_rate": 1.9459933184978828e-06, + "loss": 1.1102, + "step": 65292 + }, + { + "epoch": 4.82, + "learning_rate": 1.945756901423046e-06, + "loss": 1.03, + "step": 65293 + }, + { + "epoch": 4.82, + "learning_rate": 1.9455204971623952e-06, + "loss": 0.9129, + "step": 65294 + }, + { + "epoch": 4.82, + "learning_rate": 1.9452841057163054e-06, + "loss": 1.0299, + "step": 65295 + }, + { + "epoch": 4.82, + "learning_rate": 1.945047727085151e-06, + "loss": 0.9673, + "step": 65296 + }, + { + "epoch": 4.82, + "learning_rate": 1.9448113612693133e-06, + "loss": 0.9658, + "step": 65297 + }, + { + "epoch": 4.82, + "learning_rate": 1.944575008269167e-06, + "loss": 1.032, + "step": 65298 + }, + { + "epoch": 4.82, + "learning_rate": 1.944338668085083e-06, + "loss": 0.8696, + "step": 65299 + }, + { + "epoch": 4.82, + "learning_rate": 1.944102340717442e-06, + "loss": 0.9843, + "step": 65300 + }, + { + "epoch": 4.83, + "learning_rate": 1.9438660261666177e-06, + "loss": 1.0149, + "step": 65301 + }, + { + "epoch": 4.83, + "learning_rate": 1.9436297244329915e-06, + "loss": 0.9307, + "step": 65302 + }, + { + "epoch": 4.83, + "learning_rate": 1.9433934355169306e-06, + "loss": 1.0467, + "step": 65303 + }, + { + "epoch": 4.83, + "learning_rate": 1.943157159418817e-06, + "loss": 0.8712, + "step": 65304 + }, + { + "epoch": 4.83, + "learning_rate": 1.942920896139023e-06, + "loss": 0.9289, + "step": 65305 + }, + { + "epoch": 4.83, + "learning_rate": 1.9426846456779323e-06, + "loss": 0.8641, + "step": 65306 + }, + { + "epoch": 4.83, + "learning_rate": 1.9424484080359096e-06, + "loss": 0.8946, + "step": 65307 + }, + { + "epoch": 4.83, + "learning_rate": 1.942212183213339e-06, + "loss": 1.0071, + "step": 65308 + }, + { + "epoch": 4.83, + "learning_rate": 1.9419759712105923e-06, + "loss": 1.0335, + "step": 65309 + }, + { + "epoch": 4.83, + "learning_rate": 1.9417397720280485e-06, + "loss": 1.0595, + "step": 65310 + }, + { + "epoch": 4.83, + "learning_rate": 1.94150358566608e-06, + "loss": 0.966, + "step": 65311 + }, + { + "epoch": 4.83, + "learning_rate": 1.9412674121250618e-06, + "loss": 1.0175, + "step": 65312 + }, + { + "epoch": 4.83, + "learning_rate": 1.941031251405374e-06, + "loss": 0.9283, + "step": 65313 + }, + { + "epoch": 4.83, + "learning_rate": 1.940795103507391e-06, + "loss": 1.1184, + "step": 65314 + }, + { + "epoch": 4.83, + "learning_rate": 1.9405589684314874e-06, + "loss": 1.0182, + "step": 65315 + }, + { + "epoch": 4.83, + "learning_rate": 1.9403228461780365e-06, + "loss": 0.9947, + "step": 65316 + }, + { + "epoch": 4.83, + "learning_rate": 1.9400867367474185e-06, + "loss": 1.076, + "step": 65317 + }, + { + "epoch": 4.83, + "learning_rate": 1.9398506401400086e-06, + "loss": 1.078, + "step": 65318 + }, + { + "epoch": 4.83, + "learning_rate": 1.9396145563561798e-06, + "loss": 0.9486, + "step": 65319 + }, + { + "epoch": 4.83, + "learning_rate": 1.939378485396309e-06, + "loss": 1.0947, + "step": 65320 + }, + { + "epoch": 4.83, + "learning_rate": 1.9391424272607693e-06, + "loss": 1.0931, + "step": 65321 + }, + { + "epoch": 4.83, + "learning_rate": 1.938906381949942e-06, + "loss": 0.998, + "step": 65322 + }, + { + "epoch": 4.83, + "learning_rate": 1.938670349464199e-06, + "loss": 1.0622, + "step": 65323 + }, + { + "epoch": 4.83, + "learning_rate": 1.9384343298039154e-06, + "loss": 0.93, + "step": 65324 + }, + { + "epoch": 4.83, + "learning_rate": 1.9381983229694656e-06, + "loss": 1.0825, + "step": 65325 + }, + { + "epoch": 4.83, + "learning_rate": 1.937962328961229e-06, + "loss": 1.1259, + "step": 65326 + }, + { + "epoch": 4.83, + "learning_rate": 1.9377263477795825e-06, + "loss": 1.0115, + "step": 65327 + }, + { + "epoch": 4.83, + "learning_rate": 1.937490379424892e-06, + "loss": 1.0145, + "step": 65328 + }, + { + "epoch": 4.83, + "learning_rate": 1.937254423897542e-06, + "loss": 1.0233, + "step": 65329 + }, + { + "epoch": 4.83, + "learning_rate": 1.9370184811979044e-06, + "loss": 1.0783, + "step": 65330 + }, + { + "epoch": 4.83, + "learning_rate": 1.9367825513263548e-06, + "loss": 1.0409, + "step": 65331 + }, + { + "epoch": 4.83, + "learning_rate": 1.9365466342832673e-06, + "loss": 1.0329, + "step": 65332 + }, + { + "epoch": 4.83, + "learning_rate": 1.93631073006902e-06, + "loss": 0.9593, + "step": 65333 + }, + { + "epoch": 4.83, + "learning_rate": 1.936074838683988e-06, + "loss": 0.8684, + "step": 65334 + }, + { + "epoch": 4.83, + "learning_rate": 1.9358389601285453e-06, + "loss": 0.9113, + "step": 65335 + }, + { + "epoch": 4.83, + "learning_rate": 1.9356030944030667e-06, + "loss": 1.0109, + "step": 65336 + }, + { + "epoch": 4.83, + "learning_rate": 1.935367241507926e-06, + "loss": 1.0236, + "step": 65337 + }, + { + "epoch": 4.83, + "learning_rate": 1.935131401443504e-06, + "loss": 0.9084, + "step": 65338 + }, + { + "epoch": 4.83, + "learning_rate": 1.9348955742101715e-06, + "loss": 1.0148, + "step": 65339 + }, + { + "epoch": 4.83, + "learning_rate": 1.9346597598083063e-06, + "loss": 0.9929, + "step": 65340 + }, + { + "epoch": 4.83, + "learning_rate": 1.9344239582382775e-06, + "loss": 1.0536, + "step": 65341 + }, + { + "epoch": 4.83, + "learning_rate": 1.9341881695004684e-06, + "loss": 0.9143, + "step": 65342 + }, + { + "epoch": 4.83, + "learning_rate": 1.93395239359525e-06, + "loss": 0.9354, + "step": 65343 + }, + { + "epoch": 4.83, + "learning_rate": 1.933716630522998e-06, + "loss": 1.0162, + "step": 65344 + }, + { + "epoch": 4.83, + "learning_rate": 1.933480880284089e-06, + "loss": 1.0609, + "step": 65345 + }, + { + "epoch": 4.83, + "learning_rate": 1.9332451428788924e-06, + "loss": 1.0851, + "step": 65346 + }, + { + "epoch": 4.83, + "learning_rate": 1.933009418307792e-06, + "loss": 0.9465, + "step": 65347 + }, + { + "epoch": 4.83, + "learning_rate": 1.932773706571154e-06, + "loss": 1.0374, + "step": 65348 + }, + { + "epoch": 4.83, + "learning_rate": 1.93253800766936e-06, + "loss": 0.891, + "step": 65349 + }, + { + "epoch": 4.83, + "learning_rate": 1.9323023216027804e-06, + "loss": 1.0122, + "step": 65350 + }, + { + "epoch": 4.83, + "learning_rate": 1.9320666483717965e-06, + "loss": 1.1034, + "step": 65351 + }, + { + "epoch": 4.83, + "learning_rate": 1.931830987976775e-06, + "loss": 0.9594, + "step": 65352 + }, + { + "epoch": 4.83, + "learning_rate": 1.9315953404180975e-06, + "loss": 1.0275, + "step": 65353 + }, + { + "epoch": 4.83, + "learning_rate": 1.9313597056961354e-06, + "loss": 1.0504, + "step": 65354 + }, + { + "epoch": 4.83, + "learning_rate": 1.9311240838112655e-06, + "loss": 1.0321, + "step": 65355 + }, + { + "epoch": 4.83, + "learning_rate": 1.930888474763861e-06, + "loss": 1.0649, + "step": 65356 + }, + { + "epoch": 4.83, + "learning_rate": 1.9306528785542953e-06, + "loss": 0.9058, + "step": 65357 + }, + { + "epoch": 4.83, + "learning_rate": 1.930417295182948e-06, + "loss": 0.9395, + "step": 65358 + }, + { + "epoch": 4.83, + "learning_rate": 1.9301817246501918e-06, + "loss": 1.0183, + "step": 65359 + }, + { + "epoch": 4.83, + "learning_rate": 1.9299461669564e-06, + "loss": 0.9454, + "step": 65360 + }, + { + "epoch": 4.83, + "learning_rate": 1.929710622101947e-06, + "loss": 0.9427, + "step": 65361 + }, + { + "epoch": 4.83, + "learning_rate": 1.9294750900872105e-06, + "loss": 0.9494, + "step": 65362 + }, + { + "epoch": 4.83, + "learning_rate": 1.9292395709125644e-06, + "loss": 0.9515, + "step": 65363 + }, + { + "epoch": 4.83, + "learning_rate": 1.9290040645783825e-06, + "loss": 0.9957, + "step": 65364 + }, + { + "epoch": 4.83, + "learning_rate": 1.928768571085039e-06, + "loss": 0.9837, + "step": 65365 + }, + { + "epoch": 4.83, + "learning_rate": 1.928533090432907e-06, + "loss": 1.1004, + "step": 65366 + }, + { + "epoch": 4.83, + "learning_rate": 1.928297622622367e-06, + "loss": 1.0163, + "step": 65367 + }, + { + "epoch": 4.83, + "learning_rate": 1.928062167653789e-06, + "loss": 1.0747, + "step": 65368 + }, + { + "epoch": 4.83, + "learning_rate": 1.9278267255275485e-06, + "loss": 0.8964, + "step": 65369 + }, + { + "epoch": 4.83, + "learning_rate": 1.927591296244018e-06, + "loss": 1.0806, + "step": 65370 + }, + { + "epoch": 4.83, + "learning_rate": 1.9273558798035786e-06, + "loss": 1.0431, + "step": 65371 + }, + { + "epoch": 4.83, + "learning_rate": 1.9271204762065963e-06, + "loss": 0.9875, + "step": 65372 + }, + { + "epoch": 4.83, + "learning_rate": 1.9268850854534516e-06, + "loss": 1.006, + "step": 65373 + }, + { + "epoch": 4.83, + "learning_rate": 1.926649707544518e-06, + "loss": 1.0167, + "step": 65374 + }, + { + "epoch": 4.83, + "learning_rate": 1.9264143424801686e-06, + "loss": 1.0057, + "step": 65375 + }, + { + "epoch": 4.83, + "learning_rate": 1.9261789902607787e-06, + "loss": 0.937, + "step": 65376 + }, + { + "epoch": 4.83, + "learning_rate": 1.92594365088672e-06, + "loss": 1.0217, + "step": 65377 + }, + { + "epoch": 4.83, + "learning_rate": 1.9257083243583718e-06, + "loss": 1.0667, + "step": 65378 + }, + { + "epoch": 4.83, + "learning_rate": 1.9254730106761065e-06, + "loss": 1.0462, + "step": 65379 + }, + { + "epoch": 4.83, + "learning_rate": 1.9252377098402974e-06, + "loss": 1.1399, + "step": 65380 + }, + { + "epoch": 4.83, + "learning_rate": 1.9250024218513176e-06, + "loss": 1.0605, + "step": 65381 + }, + { + "epoch": 4.83, + "learning_rate": 1.924767146709546e-06, + "loss": 0.908, + "step": 65382 + }, + { + "epoch": 4.83, + "learning_rate": 1.924531884415355e-06, + "loss": 0.9456, + "step": 65383 + }, + { + "epoch": 4.83, + "learning_rate": 1.9242966349691173e-06, + "loss": 1.1043, + "step": 65384 + }, + { + "epoch": 4.83, + "learning_rate": 1.9240613983712074e-06, + "loss": 1.0596, + "step": 65385 + }, + { + "epoch": 4.83, + "learning_rate": 1.9238261746219987e-06, + "loss": 0.9689, + "step": 65386 + }, + { + "epoch": 4.83, + "learning_rate": 1.92359096372187e-06, + "loss": 1.0616, + "step": 65387 + }, + { + "epoch": 4.83, + "learning_rate": 1.923355765671192e-06, + "loss": 0.9836, + "step": 65388 + }, + { + "epoch": 4.83, + "learning_rate": 1.923120580470339e-06, + "loss": 0.9134, + "step": 65389 + }, + { + "epoch": 4.83, + "learning_rate": 1.9228854081196835e-06, + "loss": 0.9815, + "step": 65390 + }, + { + "epoch": 4.83, + "learning_rate": 1.922650248619604e-06, + "loss": 0.98, + "step": 65391 + }, + { + "epoch": 4.83, + "learning_rate": 1.922415101970475e-06, + "loss": 1.1026, + "step": 65392 + }, + { + "epoch": 4.83, + "learning_rate": 1.922179968172664e-06, + "loss": 0.9483, + "step": 65393 + }, + { + "epoch": 4.83, + "learning_rate": 1.92194484722655e-06, + "loss": 1.0031, + "step": 65394 + }, + { + "epoch": 4.83, + "learning_rate": 1.9217097391325048e-06, + "loss": 0.9321, + "step": 65395 + }, + { + "epoch": 4.83, + "learning_rate": 1.921474643890908e-06, + "loss": 1.0394, + "step": 65396 + }, + { + "epoch": 4.83, + "learning_rate": 1.921239561502125e-06, + "loss": 0.9592, + "step": 65397 + }, + { + "epoch": 4.83, + "learning_rate": 1.921004491966536e-06, + "loss": 1.0077, + "step": 65398 + }, + { + "epoch": 4.83, + "learning_rate": 1.9207694352845117e-06, + "loss": 1.0036, + "step": 65399 + }, + { + "epoch": 4.83, + "learning_rate": 1.920534391456432e-06, + "loss": 1.0433, + "step": 65400 + }, + { + "epoch": 4.83, + "learning_rate": 1.9202993604826646e-06, + "loss": 0.8641, + "step": 65401 + }, + { + "epoch": 4.83, + "learning_rate": 1.920064342363582e-06, + "loss": 0.9625, + "step": 65402 + }, + { + "epoch": 4.83, + "learning_rate": 1.9198293370995647e-06, + "loss": 1.0659, + "step": 65403 + }, + { + "epoch": 4.83, + "learning_rate": 1.9195943446909837e-06, + "loss": 0.9855, + "step": 65404 + }, + { + "epoch": 4.83, + "learning_rate": 1.9193593651382113e-06, + "loss": 0.991, + "step": 65405 + }, + { + "epoch": 4.83, + "learning_rate": 1.9191243984416207e-06, + "loss": 0.9789, + "step": 65406 + }, + { + "epoch": 4.83, + "learning_rate": 1.9188894446015905e-06, + "loss": 0.9601, + "step": 65407 + }, + { + "epoch": 4.83, + "learning_rate": 1.9186545036184914e-06, + "loss": 1.057, + "step": 65408 + }, + { + "epoch": 4.83, + "learning_rate": 1.918419575492696e-06, + "loss": 0.927, + "step": 65409 + }, + { + "epoch": 4.83, + "learning_rate": 1.918184660224578e-06, + "loss": 1.0803, + "step": 65410 + }, + { + "epoch": 4.83, + "learning_rate": 1.9179497578145144e-06, + "loss": 1.1116, + "step": 65411 + }, + { + "epoch": 4.83, + "learning_rate": 1.917714868262879e-06, + "loss": 0.9056, + "step": 65412 + }, + { + "epoch": 4.83, + "learning_rate": 1.9174799915700415e-06, + "loss": 0.9636, + "step": 65413 + }, + { + "epoch": 4.83, + "learning_rate": 1.917245127736379e-06, + "loss": 1.0187, + "step": 65414 + }, + { + "epoch": 4.83, + "learning_rate": 1.9170102767622602e-06, + "loss": 1.051, + "step": 65415 + }, + { + "epoch": 4.83, + "learning_rate": 1.9167754386480688e-06, + "loss": 0.9378, + "step": 65416 + }, + { + "epoch": 4.83, + "learning_rate": 1.9165406133941655e-06, + "loss": 1.1221, + "step": 65417 + }, + { + "epoch": 4.83, + "learning_rate": 1.916305801000934e-06, + "loss": 0.9048, + "step": 65418 + }, + { + "epoch": 4.83, + "learning_rate": 1.916071001468741e-06, + "loss": 1.0647, + "step": 65419 + }, + { + "epoch": 4.83, + "learning_rate": 1.915836214797969e-06, + "loss": 1.0111, + "step": 65420 + }, + { + "epoch": 4.83, + "learning_rate": 1.9156014409889834e-06, + "loss": 0.931, + "step": 65421 + }, + { + "epoch": 4.83, + "learning_rate": 1.915366680042158e-06, + "loss": 0.9084, + "step": 65422 + }, + { + "epoch": 4.83, + "learning_rate": 1.9151319319578697e-06, + "loss": 0.9807, + "step": 65423 + }, + { + "epoch": 4.83, + "learning_rate": 1.9148971967364927e-06, + "loss": 1.0659, + "step": 65424 + }, + { + "epoch": 4.83, + "learning_rate": 1.914662474378397e-06, + "loss": 1.0741, + "step": 65425 + }, + { + "epoch": 4.83, + "learning_rate": 1.914427764883956e-06, + "loss": 0.91, + "step": 65426 + }, + { + "epoch": 4.83, + "learning_rate": 1.914193068253547e-06, + "loss": 1.1091, + "step": 65427 + }, + { + "epoch": 4.83, + "learning_rate": 1.9139583844875408e-06, + "loss": 1.1519, + "step": 65428 + }, + { + "epoch": 4.83, + "learning_rate": 1.913723713586312e-06, + "loss": 1.0402, + "step": 65429 + }, + { + "epoch": 4.83, + "learning_rate": 1.913489055550233e-06, + "loss": 0.9691, + "step": 65430 + }, + { + "epoch": 4.83, + "learning_rate": 1.9132544103796736e-06, + "loss": 0.8312, + "step": 65431 + }, + { + "epoch": 4.83, + "learning_rate": 1.9130197780750138e-06, + "loss": 1.0173, + "step": 65432 + }, + { + "epoch": 4.83, + "learning_rate": 1.9127851586366243e-06, + "loss": 0.9242, + "step": 65433 + }, + { + "epoch": 4.83, + "learning_rate": 1.912550552064878e-06, + "loss": 1.0061, + "step": 65434 + }, + { + "epoch": 4.83, + "learning_rate": 1.912315958360145e-06, + "loss": 1.0438, + "step": 65435 + }, + { + "epoch": 4.84, + "learning_rate": 1.912081377522803e-06, + "loss": 1.0091, + "step": 65436 + }, + { + "epoch": 4.84, + "learning_rate": 1.911846809553226e-06, + "loss": 1.0014, + "step": 65437 + }, + { + "epoch": 4.84, + "learning_rate": 1.9116122544517835e-06, + "loss": 1.0603, + "step": 65438 + }, + { + "epoch": 4.84, + "learning_rate": 1.9113777122188505e-06, + "loss": 1.0725, + "step": 65439 + }, + { + "epoch": 4.84, + "learning_rate": 1.9111431828547976e-06, + "loss": 1.0893, + "step": 65440 + }, + { + "epoch": 4.84, + "learning_rate": 1.910908666360005e-06, + "loss": 1.0407, + "step": 65441 + }, + { + "epoch": 4.84, + "learning_rate": 1.910674162734837e-06, + "loss": 0.962, + "step": 65442 + }, + { + "epoch": 4.84, + "learning_rate": 1.9104396719796715e-06, + "loss": 0.9424, + "step": 65443 + }, + { + "epoch": 4.84, + "learning_rate": 1.9102051940948786e-06, + "loss": 1.0133, + "step": 65444 + }, + { + "epoch": 4.84, + "learning_rate": 1.909970729080839e-06, + "loss": 1.0615, + "step": 65445 + }, + { + "epoch": 4.84, + "learning_rate": 1.9097362769379158e-06, + "loss": 1.0758, + "step": 65446 + }, + { + "epoch": 4.84, + "learning_rate": 1.9095018376664878e-06, + "loss": 1.0318, + "step": 65447 + }, + { + "epoch": 4.84, + "learning_rate": 1.9092674112669273e-06, + "loss": 1.0295, + "step": 65448 + }, + { + "epoch": 4.84, + "learning_rate": 1.909032997739606e-06, + "loss": 0.9729, + "step": 65449 + }, + { + "epoch": 4.84, + "learning_rate": 1.908798597084898e-06, + "loss": 1.0968, + "step": 65450 + }, + { + "epoch": 4.84, + "learning_rate": 1.908564209303172e-06, + "loss": 0.9558, + "step": 65451 + }, + { + "epoch": 4.84, + "learning_rate": 1.908329834394809e-06, + "loss": 1.0309, + "step": 65452 + }, + { + "epoch": 4.84, + "learning_rate": 1.9080954723601763e-06, + "loss": 0.9811, + "step": 65453 + }, + { + "epoch": 4.84, + "learning_rate": 1.907861123199648e-06, + "loss": 1.0279, + "step": 65454 + }, + { + "epoch": 4.84, + "learning_rate": 1.9076267869135946e-06, + "loss": 0.8136, + "step": 65455 + }, + { + "epoch": 4.84, + "learning_rate": 1.9073924635023935e-06, + "loss": 0.9916, + "step": 65456 + }, + { + "epoch": 4.84, + "learning_rate": 1.9071581529664152e-06, + "loss": 0.92, + "step": 65457 + }, + { + "epoch": 4.84, + "learning_rate": 1.906923855306032e-06, + "loss": 1.0129, + "step": 65458 + }, + { + "epoch": 4.84, + "learning_rate": 1.906689570521618e-06, + "loss": 1.0504, + "step": 65459 + }, + { + "epoch": 4.84, + "learning_rate": 1.906455298613542e-06, + "loss": 0.941, + "step": 65460 + }, + { + "epoch": 4.84, + "learning_rate": 1.9062210395821857e-06, + "loss": 1.0286, + "step": 65461 + }, + { + "epoch": 4.84, + "learning_rate": 1.9059867934279097e-06, + "loss": 1.0441, + "step": 65462 + }, + { + "epoch": 4.84, + "learning_rate": 1.9057525601510962e-06, + "loss": 0.9994, + "step": 65463 + }, + { + "epoch": 4.84, + "learning_rate": 1.9055183397521116e-06, + "loss": 0.9504, + "step": 65464 + }, + { + "epoch": 4.84, + "learning_rate": 1.9052841322313376e-06, + "loss": 0.9059, + "step": 65465 + }, + { + "epoch": 4.84, + "learning_rate": 1.905049937589134e-06, + "loss": 0.9965, + "step": 65466 + }, + { + "epoch": 4.84, + "learning_rate": 1.904815755825884e-06, + "loss": 0.9651, + "step": 65467 + }, + { + "epoch": 4.84, + "learning_rate": 1.904581586941956e-06, + "loss": 0.9035, + "step": 65468 + }, + { + "epoch": 4.84, + "learning_rate": 1.904347430937723e-06, + "loss": 0.9984, + "step": 65469 + }, + { + "epoch": 4.84, + "learning_rate": 1.904113287813557e-06, + "loss": 0.9777, + "step": 65470 + }, + { + "epoch": 4.84, + "learning_rate": 1.9038791575698278e-06, + "loss": 1.0291, + "step": 65471 + }, + { + "epoch": 4.84, + "learning_rate": 1.903645040206915e-06, + "loss": 0.9537, + "step": 65472 + }, + { + "epoch": 4.84, + "learning_rate": 1.9034109357251862e-06, + "loss": 1.0152, + "step": 65473 + }, + { + "epoch": 4.84, + "learning_rate": 1.9031768441250154e-06, + "loss": 0.9878, + "step": 65474 + }, + { + "epoch": 4.84, + "learning_rate": 1.902942765406771e-06, + "loss": 0.9763, + "step": 65475 + }, + { + "epoch": 4.84, + "learning_rate": 1.9027086995708321e-06, + "loss": 1.1058, + "step": 65476 + }, + { + "epoch": 4.84, + "learning_rate": 1.902474646617567e-06, + "loss": 1.0912, + "step": 65477 + }, + { + "epoch": 4.84, + "learning_rate": 1.9022406065473498e-06, + "loss": 0.9915, + "step": 65478 + }, + { + "epoch": 4.84, + "learning_rate": 1.9020065793605514e-06, + "loss": 1.1214, + "step": 65479 + }, + { + "epoch": 4.84, + "learning_rate": 1.9017725650575424e-06, + "loss": 1.1133, + "step": 65480 + }, + { + "epoch": 4.84, + "learning_rate": 1.9015385636386996e-06, + "loss": 1.0464, + "step": 65481 + }, + { + "epoch": 4.84, + "learning_rate": 1.9013045751043935e-06, + "loss": 1.0374, + "step": 65482 + }, + { + "epoch": 4.84, + "learning_rate": 1.901070599454995e-06, + "loss": 1.0004, + "step": 65483 + }, + { + "epoch": 4.84, + "learning_rate": 1.9008366366908759e-06, + "loss": 1.0964, + "step": 65484 + }, + { + "epoch": 4.84, + "learning_rate": 1.9006026868124116e-06, + "loss": 0.8912, + "step": 65485 + }, + { + "epoch": 4.84, + "learning_rate": 1.9003687498199752e-06, + "loss": 1.0119, + "step": 65486 + }, + { + "epoch": 4.84, + "learning_rate": 1.9001348257139317e-06, + "loss": 0.965, + "step": 65487 + }, + { + "epoch": 4.84, + "learning_rate": 1.89990091449466e-06, + "loss": 0.8935, + "step": 65488 + }, + { + "epoch": 4.84, + "learning_rate": 1.8996670161625274e-06, + "loss": 1.0436, + "step": 65489 + }, + { + "epoch": 4.84, + "learning_rate": 1.8994331307179137e-06, + "loss": 1.0943, + "step": 65490 + }, + { + "epoch": 4.84, + "learning_rate": 1.8991992581611807e-06, + "loss": 0.9156, + "step": 65491 + }, + { + "epoch": 4.84, + "learning_rate": 1.8989653984927093e-06, + "loss": 1.0966, + "step": 65492 + }, + { + "epoch": 4.84, + "learning_rate": 1.8987315517128669e-06, + "loss": 1.1025, + "step": 65493 + }, + { + "epoch": 4.84, + "learning_rate": 1.8984977178220277e-06, + "loss": 0.957, + "step": 65494 + }, + { + "epoch": 4.84, + "learning_rate": 1.8982638968205625e-06, + "loss": 0.9722, + "step": 65495 + }, + { + "epoch": 4.84, + "learning_rate": 1.89803008870884e-06, + "loss": 1.0028, + "step": 65496 + }, + { + "epoch": 4.84, + "learning_rate": 1.8977962934872397e-06, + "loss": 1.0771, + "step": 65497 + }, + { + "epoch": 4.84, + "learning_rate": 1.897562511156128e-06, + "loss": 0.9317, + "step": 65498 + }, + { + "epoch": 4.84, + "learning_rate": 1.8973287417158792e-06, + "loss": 1.0465, + "step": 65499 + }, + { + "epoch": 4.84, + "learning_rate": 1.8970949851668618e-06, + "loss": 0.9288, + "step": 65500 + }, + { + "epoch": 4.84, + "learning_rate": 1.8968612415094534e-06, + "loss": 1.0533, + "step": 65501 + }, + { + "epoch": 4.84, + "learning_rate": 1.8966275107440223e-06, + "loss": 0.9787, + "step": 65502 + }, + { + "epoch": 4.84, + "learning_rate": 1.896393792870942e-06, + "loss": 1.129, + "step": 65503 + }, + { + "epoch": 4.84, + "learning_rate": 1.8961600878905818e-06, + "loss": 1.0363, + "step": 65504 + }, + { + "epoch": 4.84, + "learning_rate": 1.8959263958033126e-06, + "loss": 1.0317, + "step": 65505 + }, + { + "epoch": 4.84, + "learning_rate": 1.895692716609514e-06, + "loss": 1.1742, + "step": 65506 + }, + { + "epoch": 4.84, + "learning_rate": 1.8954590503095471e-06, + "loss": 1.032, + "step": 65507 + }, + { + "epoch": 4.84, + "learning_rate": 1.8952253969037914e-06, + "loss": 1.0909, + "step": 65508 + }, + { + "epoch": 4.84, + "learning_rate": 1.8949917563926146e-06, + "loss": 1.0593, + "step": 65509 + }, + { + "epoch": 4.84, + "learning_rate": 1.8947581287763938e-06, + "loss": 0.9816, + "step": 65510 + }, + { + "epoch": 4.84, + "learning_rate": 1.8945245140554923e-06, + "loss": 1.0506, + "step": 65511 + }, + { + "epoch": 4.84, + "learning_rate": 1.8942909122302889e-06, + "loss": 1.0126, + "step": 65512 + }, + { + "epoch": 4.84, + "learning_rate": 1.8940573233011516e-06, + "loss": 0.9588, + "step": 65513 + }, + { + "epoch": 4.84, + "learning_rate": 1.893823747268454e-06, + "loss": 0.9917, + "step": 65514 + }, + { + "epoch": 4.84, + "learning_rate": 1.8935901841325676e-06, + "loss": 1.0364, + "step": 65515 + }, + { + "epoch": 4.84, + "learning_rate": 1.8933566338938591e-06, + "loss": 0.9618, + "step": 65516 + }, + { + "epoch": 4.84, + "learning_rate": 1.8931230965527081e-06, + "loss": 1.0053, + "step": 65517 + }, + { + "epoch": 4.84, + "learning_rate": 1.892889572109482e-06, + "loss": 1.0268, + "step": 65518 + }, + { + "epoch": 4.84, + "learning_rate": 1.8926560605645517e-06, + "loss": 0.9597, + "step": 65519 + }, + { + "epoch": 4.84, + "learning_rate": 1.892422561918288e-06, + "loss": 1.0332, + "step": 65520 + }, + { + "epoch": 4.84, + "learning_rate": 1.892189076171066e-06, + "loss": 0.8956, + "step": 65521 + }, + { + "epoch": 4.84, + "learning_rate": 1.8919556033232556e-06, + "loss": 1.0155, + "step": 65522 + }, + { + "epoch": 4.84, + "learning_rate": 1.8917221433752275e-06, + "loss": 0.9865, + "step": 65523 + }, + { + "epoch": 4.84, + "learning_rate": 1.8914886963273538e-06, + "loss": 0.9674, + "step": 65524 + }, + { + "epoch": 4.84, + "learning_rate": 1.8912552621800029e-06, + "loss": 0.9075, + "step": 65525 + }, + { + "epoch": 4.84, + "learning_rate": 1.8910218409335512e-06, + "loss": 0.9644, + "step": 65526 + }, + { + "epoch": 4.84, + "learning_rate": 1.8907884325883674e-06, + "loss": 0.9782, + "step": 65527 + }, + { + "epoch": 4.84, + "learning_rate": 1.8905550371448244e-06, + "loss": 1.019, + "step": 65528 + }, + { + "epoch": 4.84, + "learning_rate": 1.8903216546032888e-06, + "loss": 0.9387, + "step": 65529 + }, + { + "epoch": 4.84, + "learning_rate": 1.8900882849641378e-06, + "loss": 0.9976, + "step": 65530 + }, + { + "epoch": 4.84, + "learning_rate": 1.8898549282277412e-06, + "loss": 1.003, + "step": 65531 + }, + { + "epoch": 4.84, + "learning_rate": 1.8896215843944688e-06, + "loss": 1.0963, + "step": 65532 + }, + { + "epoch": 4.84, + "learning_rate": 1.8893882534646924e-06, + "loss": 0.8971, + "step": 65533 + }, + { + "epoch": 4.84, + "learning_rate": 1.8891549354387806e-06, + "loss": 1.0899, + "step": 65534 + }, + { + "epoch": 4.84, + "learning_rate": 1.8889216303171132e-06, + "loss": 1.08, + "step": 65535 + }, + { + "epoch": 4.84, + "learning_rate": 1.8886883381000486e-06, + "loss": 1.0071, + "step": 65536 + }, + { + "epoch": 4.84, + "learning_rate": 1.8884550587879692e-06, + "loss": 0.9996, + "step": 65537 + }, + { + "epoch": 4.84, + "learning_rate": 1.8882217923812407e-06, + "loss": 0.9931, + "step": 65538 + }, + { + "epoch": 4.84, + "learning_rate": 1.8879885388802355e-06, + "loss": 0.9803, + "step": 65539 + }, + { + "epoch": 4.84, + "learning_rate": 1.887755298285322e-06, + "loss": 0.9149, + "step": 65540 + }, + { + "epoch": 4.84, + "learning_rate": 1.8875220705968767e-06, + "loss": 0.8957, + "step": 65541 + }, + { + "epoch": 4.84, + "learning_rate": 1.887288855815267e-06, + "loss": 1.065, + "step": 65542 + }, + { + "epoch": 4.84, + "learning_rate": 1.8870556539408646e-06, + "loss": 0.9443, + "step": 65543 + }, + { + "epoch": 4.84, + "learning_rate": 1.8868224649740407e-06, + "loss": 0.9735, + "step": 65544 + }, + { + "epoch": 4.84, + "learning_rate": 1.8865892889151638e-06, + "loss": 1.0824, + "step": 65545 + }, + { + "epoch": 4.84, + "learning_rate": 1.8863561257646101e-06, + "loss": 0.9251, + "step": 65546 + }, + { + "epoch": 4.84, + "learning_rate": 1.8861229755227473e-06, + "loss": 0.924, + "step": 65547 + }, + { + "epoch": 4.84, + "learning_rate": 1.8858898381899471e-06, + "loss": 1.0385, + "step": 65548 + }, + { + "epoch": 4.84, + "learning_rate": 1.885656713766577e-06, + "loss": 0.9778, + "step": 65549 + }, + { + "epoch": 4.84, + "learning_rate": 1.8854236022530137e-06, + "loss": 1.0769, + "step": 65550 + }, + { + "epoch": 4.84, + "learning_rate": 1.8851905036496287e-06, + "loss": 1.0004, + "step": 65551 + }, + { + "epoch": 4.84, + "learning_rate": 1.8849574179567832e-06, + "loss": 0.9384, + "step": 65552 + }, + { + "epoch": 4.84, + "learning_rate": 1.8847243451748575e-06, + "loss": 0.9635, + "step": 65553 + }, + { + "epoch": 4.84, + "learning_rate": 1.8844912853042173e-06, + "loss": 0.9565, + "step": 65554 + }, + { + "epoch": 4.84, + "learning_rate": 1.884258238345239e-06, + "loss": 1.0263, + "step": 65555 + }, + { + "epoch": 4.84, + "learning_rate": 1.8840252042982854e-06, + "loss": 0.9684, + "step": 65556 + }, + { + "epoch": 4.84, + "learning_rate": 1.883792183163735e-06, + "loss": 1.0067, + "step": 65557 + }, + { + "epoch": 4.84, + "learning_rate": 1.8835591749419512e-06, + "loss": 0.9327, + "step": 65558 + }, + { + "epoch": 4.84, + "learning_rate": 1.8833261796333147e-06, + "loss": 1.046, + "step": 65559 + }, + { + "epoch": 4.84, + "learning_rate": 1.8830931972381872e-06, + "loss": 1.0215, + "step": 65560 + }, + { + "epoch": 4.84, + "learning_rate": 1.8828602277569396e-06, + "loss": 0.9798, + "step": 65561 + }, + { + "epoch": 4.84, + "learning_rate": 1.8826272711899475e-06, + "loss": 0.9355, + "step": 65562 + }, + { + "epoch": 4.84, + "learning_rate": 1.88239432753758e-06, + "loss": 0.9951, + "step": 65563 + }, + { + "epoch": 4.84, + "learning_rate": 1.882161396800206e-06, + "loss": 1.11, + "step": 65564 + }, + { + "epoch": 4.84, + "learning_rate": 1.8819284789781955e-06, + "loss": 0.9158, + "step": 65565 + }, + { + "epoch": 4.84, + "learning_rate": 1.8816955740719224e-06, + "loss": 1.0861, + "step": 65566 + }, + { + "epoch": 4.84, + "learning_rate": 1.881462682081756e-06, + "loss": 1.0521, + "step": 65567 + }, + { + "epoch": 4.84, + "learning_rate": 1.8812298030080666e-06, + "loss": 1.0605, + "step": 65568 + }, + { + "epoch": 4.84, + "learning_rate": 1.8809969368512238e-06, + "loss": 1.0102, + "step": 65569 + }, + { + "epoch": 4.84, + "learning_rate": 1.8807640836115958e-06, + "loss": 0.9253, + "step": 65570 + }, + { + "epoch": 4.84, + "learning_rate": 1.8805312432895595e-06, + "loss": 1.0517, + "step": 65571 + }, + { + "epoch": 4.85, + "learning_rate": 1.880298415885482e-06, + "loss": 0.9703, + "step": 65572 + }, + { + "epoch": 4.85, + "learning_rate": 1.880065601399732e-06, + "loss": 0.9318, + "step": 65573 + }, + { + "epoch": 4.85, + "learning_rate": 1.8798327998326805e-06, + "loss": 1.057, + "step": 65574 + }, + { + "epoch": 4.85, + "learning_rate": 1.8796000111847e-06, + "loss": 0.9241, + "step": 65575 + }, + { + "epoch": 4.85, + "learning_rate": 1.879367235456161e-06, + "loss": 0.9817, + "step": 65576 + }, + { + "epoch": 4.85, + "learning_rate": 1.8791344726474326e-06, + "loss": 0.9395, + "step": 65577 + }, + { + "epoch": 4.85, + "learning_rate": 1.8789017227588812e-06, + "loss": 1.1428, + "step": 65578 + }, + { + "epoch": 4.85, + "learning_rate": 1.8786689857908846e-06, + "loss": 1.0604, + "step": 65579 + }, + { + "epoch": 4.85, + "learning_rate": 1.8784362617438112e-06, + "loss": 0.9302, + "step": 65580 + }, + { + "epoch": 4.85, + "learning_rate": 1.8782035506180251e-06, + "loss": 0.9967, + "step": 65581 + }, + { + "epoch": 4.85, + "learning_rate": 1.8779708524139029e-06, + "loss": 0.9642, + "step": 65582 + }, + { + "epoch": 4.85, + "learning_rate": 1.877738167131814e-06, + "loss": 0.9681, + "step": 65583 + }, + { + "epoch": 4.85, + "learning_rate": 1.877505494772126e-06, + "loss": 1.0266, + "step": 65584 + }, + { + "epoch": 4.85, + "learning_rate": 1.8772728353352087e-06, + "loss": 0.9127, + "step": 65585 + }, + { + "epoch": 4.85, + "learning_rate": 1.8770401888214363e-06, + "loss": 1.0408, + "step": 65586 + }, + { + "epoch": 4.85, + "learning_rate": 1.876807555231177e-06, + "loss": 1.0377, + "step": 65587 + }, + { + "epoch": 4.85, + "learning_rate": 1.876574934564801e-06, + "loss": 0.9716, + "step": 65588 + }, + { + "epoch": 4.85, + "learning_rate": 1.8763423268226777e-06, + "loss": 0.9745, + "step": 65589 + }, + { + "epoch": 4.85, + "learning_rate": 1.8761097320051759e-06, + "loss": 1.021, + "step": 65590 + }, + { + "epoch": 4.85, + "learning_rate": 1.8758771501126683e-06, + "loss": 1.0074, + "step": 65591 + }, + { + "epoch": 4.85, + "learning_rate": 1.8756445811455248e-06, + "loss": 1.0459, + "step": 65592 + }, + { + "epoch": 4.85, + "learning_rate": 1.875412025104114e-06, + "loss": 1.0424, + "step": 65593 + }, + { + "epoch": 4.85, + "learning_rate": 1.8751794819888047e-06, + "loss": 0.9555, + "step": 65594 + }, + { + "epoch": 4.85, + "learning_rate": 1.8749469517999707e-06, + "loss": 1.0163, + "step": 65595 + }, + { + "epoch": 4.85, + "learning_rate": 1.8747144345379797e-06, + "loss": 0.8613, + "step": 65596 + }, + { + "epoch": 4.85, + "learning_rate": 1.8744819302032014e-06, + "loss": 0.9289, + "step": 65597 + }, + { + "epoch": 4.85, + "learning_rate": 1.8742494387960064e-06, + "loss": 1.0798, + "step": 65598 + }, + { + "epoch": 4.85, + "learning_rate": 1.8740169603167625e-06, + "loss": 0.9424, + "step": 65599 + }, + { + "epoch": 4.85, + "learning_rate": 1.8737844947658457e-06, + "loss": 0.9687, + "step": 65600 + }, + { + "epoch": 4.85, + "learning_rate": 1.873552042143616e-06, + "loss": 0.9751, + "step": 65601 + }, + { + "epoch": 4.85, + "learning_rate": 1.8733196024504518e-06, + "loss": 1.0556, + "step": 65602 + }, + { + "epoch": 4.85, + "learning_rate": 1.8730871756867174e-06, + "loss": 0.9849, + "step": 65603 + }, + { + "epoch": 4.85, + "learning_rate": 1.8728547618527904e-06, + "loss": 1.0187, + "step": 65604 + }, + { + "epoch": 4.85, + "learning_rate": 1.8726223609490302e-06, + "loss": 1.0023, + "step": 65605 + }, + { + "epoch": 4.85, + "learning_rate": 1.8723899729758132e-06, + "loss": 0.8986, + "step": 65606 + }, + { + "epoch": 4.85, + "learning_rate": 1.8721575979335083e-06, + "loss": 1.0379, + "step": 65607 + }, + { + "epoch": 4.85, + "learning_rate": 1.8719252358224838e-06, + "loss": 0.9608, + "step": 65608 + }, + { + "epoch": 4.85, + "learning_rate": 1.8716928866431094e-06, + "loss": 0.9719, + "step": 65609 + }, + { + "epoch": 4.85, + "learning_rate": 1.871460550395754e-06, + "loss": 0.9915, + "step": 65610 + }, + { + "epoch": 4.85, + "learning_rate": 1.8712282270807902e-06, + "loss": 1.0413, + "step": 65611 + }, + { + "epoch": 4.85, + "learning_rate": 1.870995916698587e-06, + "loss": 1.0451, + "step": 65612 + }, + { + "epoch": 4.85, + "learning_rate": 1.8707636192495126e-06, + "loss": 0.945, + "step": 65613 + }, + { + "epoch": 4.85, + "learning_rate": 1.8705313347339338e-06, + "loss": 0.9418, + "step": 65614 + }, + { + "epoch": 4.85, + "learning_rate": 1.8702990631522266e-06, + "loss": 0.9325, + "step": 65615 + }, + { + "epoch": 4.85, + "learning_rate": 1.8700668045047564e-06, + "loss": 0.8934, + "step": 65616 + }, + { + "epoch": 4.85, + "learning_rate": 1.8698345587918941e-06, + "loss": 1.0324, + "step": 65617 + }, + { + "epoch": 4.85, + "learning_rate": 1.8696023260140095e-06, + "loss": 1.0152, + "step": 65618 + }, + { + "epoch": 4.85, + "learning_rate": 1.8693701061714675e-06, + "loss": 1.0419, + "step": 65619 + }, + { + "epoch": 4.85, + "learning_rate": 1.8691378992646448e-06, + "loss": 1.0519, + "step": 65620 + }, + { + "epoch": 4.85, + "learning_rate": 1.8689057052939063e-06, + "loss": 0.9842, + "step": 65621 + }, + { + "epoch": 4.85, + "learning_rate": 1.8686735242596233e-06, + "loss": 1.0242, + "step": 65622 + }, + { + "epoch": 4.85, + "learning_rate": 1.8684413561621618e-06, + "loss": 0.9151, + "step": 65623 + }, + { + "epoch": 4.85, + "learning_rate": 1.8682092010018959e-06, + "loss": 0.9949, + "step": 65624 + }, + { + "epoch": 4.85, + "learning_rate": 1.8679770587791934e-06, + "loss": 0.9791, + "step": 65625 + }, + { + "epoch": 4.85, + "learning_rate": 1.8677449294944228e-06, + "loss": 1.0227, + "step": 65626 + }, + { + "epoch": 4.85, + "learning_rate": 1.8675128131479537e-06, + "loss": 0.8125, + "step": 65627 + }, + { + "epoch": 4.85, + "learning_rate": 1.8672807097401547e-06, + "loss": 0.9685, + "step": 65628 + }, + { + "epoch": 4.85, + "learning_rate": 1.8670486192713965e-06, + "loss": 1.099, + "step": 65629 + }, + { + "epoch": 4.85, + "learning_rate": 1.8668165417420447e-06, + "loss": 1.0487, + "step": 65630 + }, + { + "epoch": 4.85, + "learning_rate": 1.8665844771524744e-06, + "loss": 1.0485, + "step": 65631 + }, + { + "epoch": 4.85, + "learning_rate": 1.8663524255030507e-06, + "loss": 1.0152, + "step": 65632 + }, + { + "epoch": 4.85, + "learning_rate": 1.8661203867941447e-06, + "loss": 1.0205, + "step": 65633 + }, + { + "epoch": 4.85, + "learning_rate": 1.8658883610261225e-06, + "loss": 0.9933, + "step": 65634 + }, + { + "epoch": 4.85, + "learning_rate": 1.8656563481993583e-06, + "loss": 0.9995, + "step": 65635 + }, + { + "epoch": 4.85, + "learning_rate": 1.8654243483142177e-06, + "loss": 0.9849, + "step": 65636 + }, + { + "epoch": 4.85, + "learning_rate": 1.865192361371071e-06, + "loss": 1.0067, + "step": 65637 + }, + { + "epoch": 4.85, + "learning_rate": 1.8649603873702871e-06, + "loss": 0.9857, + "step": 65638 + }, + { + "epoch": 4.85, + "learning_rate": 1.864728426312231e-06, + "loss": 1.0634, + "step": 65639 + }, + { + "epoch": 4.85, + "learning_rate": 1.8644964781972797e-06, + "loss": 0.978, + "step": 65640 + }, + { + "epoch": 4.85, + "learning_rate": 1.8642645430257967e-06, + "loss": 0.9161, + "step": 65641 + }, + { + "epoch": 4.85, + "learning_rate": 1.864032620798154e-06, + "loss": 1.0591, + "step": 65642 + }, + { + "epoch": 4.85, + "learning_rate": 1.863800711514715e-06, + "loss": 1.0392, + "step": 65643 + }, + { + "epoch": 4.85, + "learning_rate": 1.8635688151758558e-06, + "loss": 1.0306, + "step": 65644 + }, + { + "epoch": 4.85, + "learning_rate": 1.8633369317819449e-06, + "loss": 1.1767, + "step": 65645 + }, + { + "epoch": 4.85, + "learning_rate": 1.8631050613333435e-06, + "loss": 0.9406, + "step": 65646 + }, + { + "epoch": 4.85, + "learning_rate": 1.8628732038304276e-06, + "loss": 1.0274, + "step": 65647 + }, + { + "epoch": 4.85, + "learning_rate": 1.8626413592735615e-06, + "loss": 1.0066, + "step": 65648 + }, + { + "epoch": 4.85, + "learning_rate": 1.8624095276631227e-06, + "loss": 0.9595, + "step": 65649 + }, + { + "epoch": 4.85, + "learning_rate": 1.8621777089994675e-06, + "loss": 0.9775, + "step": 65650 + }, + { + "epoch": 4.85, + "learning_rate": 1.8619459032829745e-06, + "loss": 1.0274, + "step": 65651 + }, + { + "epoch": 4.85, + "learning_rate": 1.8617141105140057e-06, + "loss": 0.821, + "step": 65652 + }, + { + "epoch": 4.85, + "learning_rate": 1.8614823306929398e-06, + "loss": 0.9787, + "step": 65653 + }, + { + "epoch": 4.85, + "learning_rate": 1.861250563820135e-06, + "loss": 0.9769, + "step": 65654 + }, + { + "epoch": 4.85, + "learning_rate": 1.8610188098959624e-06, + "loss": 1.11, + "step": 65655 + }, + { + "epoch": 4.85, + "learning_rate": 1.8607870689207953e-06, + "loss": 0.9463, + "step": 65656 + }, + { + "epoch": 4.85, + "learning_rate": 1.8605553408949983e-06, + "loss": 1.0453, + "step": 65657 + }, + { + "epoch": 4.85, + "learning_rate": 1.8603236258189428e-06, + "loss": 0.9627, + "step": 65658 + }, + { + "epoch": 4.85, + "learning_rate": 1.8600919236929926e-06, + "loss": 1.0093, + "step": 65659 + }, + { + "epoch": 4.85, + "learning_rate": 1.859860234517522e-06, + "loss": 0.9804, + "step": 65660 + }, + { + "epoch": 4.85, + "learning_rate": 1.8596285582928974e-06, + "loss": 0.9643, + "step": 65661 + }, + { + "epoch": 4.85, + "learning_rate": 1.8593968950194875e-06, + "loss": 1.0296, + "step": 65662 + }, + { + "epoch": 4.85, + "learning_rate": 1.8591652446976604e-06, + "loss": 1.0681, + "step": 65663 + }, + { + "epoch": 4.85, + "learning_rate": 1.858933607327783e-06, + "loss": 1.0301, + "step": 65664 + }, + { + "epoch": 4.85, + "learning_rate": 1.8587019829102283e-06, + "loss": 1.0702, + "step": 65665 + }, + { + "epoch": 4.85, + "learning_rate": 1.858470371445361e-06, + "loss": 0.8147, + "step": 65666 + }, + { + "epoch": 4.85, + "learning_rate": 1.8582387729335515e-06, + "loss": 0.9183, + "step": 65667 + }, + { + "epoch": 4.85, + "learning_rate": 1.8580071873751659e-06, + "loss": 0.9245, + "step": 65668 + }, + { + "epoch": 4.85, + "learning_rate": 1.857775614770576e-06, + "loss": 1.0584, + "step": 65669 + }, + { + "epoch": 4.85, + "learning_rate": 1.8575440551201495e-06, + "loss": 0.9272, + "step": 65670 + }, + { + "epoch": 4.85, + "learning_rate": 1.8573125084242528e-06, + "loss": 1.1323, + "step": 65671 + }, + { + "epoch": 4.85, + "learning_rate": 1.8570809746832564e-06, + "loss": 1.0249, + "step": 65672 + }, + { + "epoch": 4.85, + "learning_rate": 1.8568494538975268e-06, + "loss": 1.012, + "step": 65673 + }, + { + "epoch": 4.85, + "learning_rate": 1.8566179460674338e-06, + "loss": 1.0112, + "step": 65674 + }, + { + "epoch": 4.85, + "learning_rate": 1.8563864511933427e-06, + "loss": 1.0428, + "step": 65675 + }, + { + "epoch": 4.85, + "learning_rate": 1.8561549692756264e-06, + "loss": 0.9829, + "step": 65676 + }, + { + "epoch": 4.85, + "learning_rate": 1.8559235003146514e-06, + "loss": 0.9964, + "step": 65677 + }, + { + "epoch": 4.85, + "learning_rate": 1.855692044310785e-06, + "loss": 0.9946, + "step": 65678 + }, + { + "epoch": 4.85, + "learning_rate": 1.855460601264394e-06, + "loss": 0.97, + "step": 65679 + }, + { + "epoch": 4.85, + "learning_rate": 1.8552291711758507e-06, + "loss": 1.0152, + "step": 65680 + }, + { + "epoch": 4.85, + "learning_rate": 1.8549977540455212e-06, + "loss": 0.9689, + "step": 65681 + }, + { + "epoch": 4.85, + "learning_rate": 1.8547663498737744e-06, + "loss": 1.1062, + "step": 65682 + }, + { + "epoch": 4.85, + "learning_rate": 1.8545349586609772e-06, + "loss": 1.0366, + "step": 65683 + }, + { + "epoch": 4.85, + "learning_rate": 1.854303580407496e-06, + "loss": 0.9613, + "step": 65684 + }, + { + "epoch": 4.85, + "learning_rate": 1.8540722151137037e-06, + "loss": 0.9283, + "step": 65685 + }, + { + "epoch": 4.85, + "learning_rate": 1.8538408627799665e-06, + "loss": 0.9273, + "step": 65686 + }, + { + "epoch": 4.85, + "learning_rate": 1.853609523406652e-06, + "loss": 0.9358, + "step": 65687 + }, + { + "epoch": 4.85, + "learning_rate": 1.8533781969941245e-06, + "loss": 1.0796, + "step": 65688 + }, + { + "epoch": 4.85, + "learning_rate": 1.853146883542759e-06, + "loss": 1.0207, + "step": 65689 + }, + { + "epoch": 4.85, + "learning_rate": 1.85291558305292e-06, + "loss": 1.0255, + "step": 65690 + }, + { + "epoch": 4.85, + "learning_rate": 1.8526842955249768e-06, + "loss": 1.0123, + "step": 65691 + }, + { + "epoch": 4.85, + "learning_rate": 1.852453020959295e-06, + "loss": 1.0711, + "step": 65692 + }, + { + "epoch": 4.85, + "learning_rate": 1.852221759356242e-06, + "loss": 1.1359, + "step": 65693 + }, + { + "epoch": 4.85, + "learning_rate": 1.851990510716193e-06, + "loss": 1.0293, + "step": 65694 + }, + { + "epoch": 4.85, + "learning_rate": 1.851759275039504e-06, + "loss": 0.9935, + "step": 65695 + }, + { + "epoch": 4.85, + "learning_rate": 1.8515280523265533e-06, + "loss": 1.0063, + "step": 65696 + }, + { + "epoch": 4.85, + "learning_rate": 1.8512968425777022e-06, + "loss": 1.0678, + "step": 65697 + }, + { + "epoch": 4.85, + "learning_rate": 1.8510656457933263e-06, + "loss": 1.0308, + "step": 65698 + }, + { + "epoch": 4.85, + "learning_rate": 1.8508344619737828e-06, + "loss": 1.0764, + "step": 65699 + }, + { + "epoch": 4.85, + "learning_rate": 1.8506032911194472e-06, + "loss": 0.9612, + "step": 65700 + }, + { + "epoch": 4.85, + "learning_rate": 1.8503721332306857e-06, + "loss": 0.8692, + "step": 65701 + }, + { + "epoch": 4.85, + "learning_rate": 1.8501409883078659e-06, + "loss": 1.0706, + "step": 65702 + }, + { + "epoch": 4.85, + "learning_rate": 1.849909856351355e-06, + "loss": 1.1182, + "step": 65703 + }, + { + "epoch": 4.85, + "learning_rate": 1.8496787373615189e-06, + "loss": 1.0702, + "step": 65704 + }, + { + "epoch": 4.85, + "learning_rate": 1.8494476313387288e-06, + "loss": 1.0701, + "step": 65705 + }, + { + "epoch": 4.85, + "learning_rate": 1.8492165382833517e-06, + "loss": 0.9819, + "step": 65706 + }, + { + "epoch": 4.86, + "learning_rate": 1.8489854581957533e-06, + "loss": 0.9512, + "step": 65707 + }, + { + "epoch": 4.86, + "learning_rate": 1.8487543910763005e-06, + "loss": 1.039, + "step": 65708 + }, + { + "epoch": 4.86, + "learning_rate": 1.8485233369253663e-06, + "loss": 1.0131, + "step": 65709 + }, + { + "epoch": 4.86, + "learning_rate": 1.8482922957433136e-06, + "loss": 0.9126, + "step": 65710 + }, + { + "epoch": 4.86, + "learning_rate": 1.8480612675305121e-06, + "loss": 0.9807, + "step": 65711 + }, + { + "epoch": 4.86, + "learning_rate": 1.8478302522873281e-06, + "loss": 0.9075, + "step": 65712 + }, + { + "epoch": 4.86, + "learning_rate": 1.8475992500141272e-06, + "loss": 0.9141, + "step": 65713 + }, + { + "epoch": 4.86, + "learning_rate": 1.8473682607112842e-06, + "loss": 1.0587, + "step": 65714 + }, + { + "epoch": 4.86, + "learning_rate": 1.847137284379157e-06, + "loss": 0.9592, + "step": 65715 + }, + { + "epoch": 4.86, + "learning_rate": 1.8469063210181193e-06, + "loss": 0.9218, + "step": 65716 + }, + { + "epoch": 4.86, + "learning_rate": 1.8466753706285345e-06, + "loss": 0.9235, + "step": 65717 + }, + { + "epoch": 4.86, + "learning_rate": 1.846444433210779e-06, + "loss": 0.9133, + "step": 65718 + }, + { + "epoch": 4.86, + "learning_rate": 1.84621350876521e-06, + "loss": 0.9177, + "step": 65719 + }, + { + "epoch": 4.86, + "learning_rate": 1.8459825972921964e-06, + "loss": 0.8605, + "step": 65720 + }, + { + "epoch": 4.86, + "learning_rate": 1.8457516987921097e-06, + "loss": 1.078, + "step": 65721 + }, + { + "epoch": 4.86, + "learning_rate": 1.845520813265317e-06, + "loss": 1.0963, + "step": 65722 + }, + { + "epoch": 4.86, + "learning_rate": 1.8452899407121826e-06, + "loss": 1.0238, + "step": 65723 + }, + { + "epoch": 4.86, + "learning_rate": 1.8450590811330727e-06, + "loss": 0.9639, + "step": 65724 + }, + { + "epoch": 4.86, + "learning_rate": 1.8448282345283608e-06, + "loss": 1.1128, + "step": 65725 + }, + { + "epoch": 4.86, + "learning_rate": 1.844597400898409e-06, + "loss": 1.0927, + "step": 65726 + }, + { + "epoch": 4.86, + "learning_rate": 1.8443665802435874e-06, + "loss": 0.8944, + "step": 65727 + }, + { + "epoch": 4.86, + "learning_rate": 1.8441357725642605e-06, + "loss": 1.0076, + "step": 65728 + }, + { + "epoch": 4.86, + "learning_rate": 1.8439049778607953e-06, + "loss": 1.0096, + "step": 65729 + }, + { + "epoch": 4.86, + "learning_rate": 1.8436741961335636e-06, + "loss": 1.0161, + "step": 65730 + }, + { + "epoch": 4.86, + "learning_rate": 1.8434434273829293e-06, + "loss": 0.8818, + "step": 65731 + }, + { + "epoch": 4.86, + "learning_rate": 1.843212671609259e-06, + "loss": 1.0221, + "step": 65732 + }, + { + "epoch": 4.86, + "learning_rate": 1.8429819288129181e-06, + "loss": 1.0036, + "step": 65733 + }, + { + "epoch": 4.86, + "learning_rate": 1.8427511989942792e-06, + "loss": 0.9826, + "step": 65734 + }, + { + "epoch": 4.86, + "learning_rate": 1.8425204821537067e-06, + "loss": 0.979, + "step": 65735 + }, + { + "epoch": 4.86, + "learning_rate": 1.8422897782915673e-06, + "loss": 0.9575, + "step": 65736 + }, + { + "epoch": 4.86, + "learning_rate": 1.8420590874082257e-06, + "loss": 1.0639, + "step": 65737 + }, + { + "epoch": 4.86, + "learning_rate": 1.841828409504054e-06, + "loss": 1.0423, + "step": 65738 + }, + { + "epoch": 4.86, + "learning_rate": 1.8415977445794185e-06, + "loss": 0.9716, + "step": 65739 + }, + { + "epoch": 4.86, + "learning_rate": 1.8413670926346794e-06, + "loss": 1.0214, + "step": 65740 + }, + { + "epoch": 4.86, + "learning_rate": 1.8411364536702115e-06, + "loss": 1.0372, + "step": 65741 + }, + { + "epoch": 4.86, + "learning_rate": 1.8409058276863757e-06, + "loss": 0.9842, + "step": 65742 + }, + { + "epoch": 4.86, + "learning_rate": 1.8406752146835472e-06, + "loss": 0.957, + "step": 65743 + }, + { + "epoch": 4.86, + "learning_rate": 1.8404446146620825e-06, + "loss": 0.9535, + "step": 65744 + }, + { + "epoch": 4.86, + "learning_rate": 1.8402140276223568e-06, + "loss": 1.0084, + "step": 65745 + }, + { + "epoch": 4.86, + "learning_rate": 1.839983453564731e-06, + "loss": 1.0229, + "step": 65746 + }, + { + "epoch": 4.86, + "learning_rate": 1.839752892489579e-06, + "loss": 0.8981, + "step": 65747 + }, + { + "epoch": 4.86, + "learning_rate": 1.8395223443972621e-06, + "loss": 0.9779, + "step": 65748 + }, + { + "epoch": 4.86, + "learning_rate": 1.839291809288145e-06, + "loss": 0.96, + "step": 65749 + }, + { + "epoch": 4.86, + "learning_rate": 1.8390612871626013e-06, + "loss": 1.0197, + "step": 65750 + }, + { + "epoch": 4.86, + "learning_rate": 1.8388307780209934e-06, + "loss": 0.8402, + "step": 65751 + }, + { + "epoch": 4.86, + "learning_rate": 1.8386002818636893e-06, + "loss": 1.0134, + "step": 65752 + }, + { + "epoch": 4.86, + "learning_rate": 1.838369798691053e-06, + "loss": 0.9837, + "step": 65753 + }, + { + "epoch": 4.86, + "learning_rate": 1.8381393285034554e-06, + "loss": 1.0072, + "step": 65754 + }, + { + "epoch": 4.86, + "learning_rate": 1.8379088713012626e-06, + "loss": 1.0432, + "step": 65755 + }, + { + "epoch": 4.86, + "learning_rate": 1.837678427084839e-06, + "loss": 1.0253, + "step": 65756 + }, + { + "epoch": 4.86, + "learning_rate": 1.8374479958545522e-06, + "loss": 0.9607, + "step": 65757 + }, + { + "epoch": 4.86, + "learning_rate": 1.837217577610767e-06, + "loss": 1.0153, + "step": 65758 + }, + { + "epoch": 4.86, + "learning_rate": 1.836987172353857e-06, + "loss": 1.0456, + "step": 65759 + }, + { + "epoch": 4.86, + "learning_rate": 1.836756780084178e-06, + "loss": 1.0376, + "step": 65760 + }, + { + "epoch": 4.86, + "learning_rate": 1.8365264008021044e-06, + "loss": 0.994, + "step": 65761 + }, + { + "epoch": 4.86, + "learning_rate": 1.8362960345079984e-06, + "loss": 0.9395, + "step": 65762 + }, + { + "epoch": 4.86, + "learning_rate": 1.8360656812022338e-06, + "loss": 1.0168, + "step": 65763 + }, + { + "epoch": 4.86, + "learning_rate": 1.835835340885167e-06, + "loss": 1.0024, + "step": 65764 + }, + { + "epoch": 4.86, + "learning_rate": 1.8356050135571712e-06, + "loss": 0.9492, + "step": 65765 + }, + { + "epoch": 4.86, + "learning_rate": 1.8353746992186116e-06, + "loss": 0.9791, + "step": 65766 + }, + { + "epoch": 4.86, + "learning_rate": 1.8351443978698546e-06, + "loss": 1.0312, + "step": 65767 + }, + { + "epoch": 4.86, + "learning_rate": 1.8349141095112645e-06, + "loss": 0.9325, + "step": 65768 + }, + { + "epoch": 4.86, + "learning_rate": 1.8346838341432083e-06, + "loss": 1.0368, + "step": 65769 + }, + { + "epoch": 4.86, + "learning_rate": 1.834453571766055e-06, + "loss": 1.0532, + "step": 65770 + }, + { + "epoch": 4.86, + "learning_rate": 1.8342233223801686e-06, + "loss": 0.9944, + "step": 65771 + }, + { + "epoch": 4.86, + "learning_rate": 1.8339930859859166e-06, + "loss": 0.9574, + "step": 65772 + }, + { + "epoch": 4.86, + "learning_rate": 1.8337628625836623e-06, + "loss": 0.9717, + "step": 65773 + }, + { + "epoch": 4.86, + "learning_rate": 1.8335326521737772e-06, + "loss": 0.939, + "step": 65774 + }, + { + "epoch": 4.86, + "learning_rate": 1.8333024547566248e-06, + "loss": 0.9098, + "step": 65775 + }, + { + "epoch": 4.86, + "learning_rate": 1.833072270332571e-06, + "loss": 1.0466, + "step": 65776 + }, + { + "epoch": 4.86, + "learning_rate": 1.8328420989019834e-06, + "loss": 1.0813, + "step": 65777 + }, + { + "epoch": 4.86, + "learning_rate": 1.8326119404652233e-06, + "loss": 1.0081, + "step": 65778 + }, + { + "epoch": 4.86, + "learning_rate": 1.832381795022664e-06, + "loss": 1.0185, + "step": 65779 + }, + { + "epoch": 4.86, + "learning_rate": 1.8321516625746682e-06, + "loss": 1.0217, + "step": 65780 + }, + { + "epoch": 4.86, + "learning_rate": 1.831921543121602e-06, + "loss": 1.049, + "step": 65781 + }, + { + "epoch": 4.86, + "learning_rate": 1.8316914366638295e-06, + "loss": 1.0085, + "step": 65782 + }, + { + "epoch": 4.86, + "learning_rate": 1.8314613432017213e-06, + "loss": 1.0499, + "step": 65783 + }, + { + "epoch": 4.86, + "learning_rate": 1.8312312627356443e-06, + "loss": 0.9424, + "step": 65784 + }, + { + "epoch": 4.86, + "learning_rate": 1.8310011952659556e-06, + "loss": 0.9573, + "step": 65785 + }, + { + "epoch": 4.86, + "learning_rate": 1.8307711407930306e-06, + "loss": 0.9843, + "step": 65786 + }, + { + "epoch": 4.86, + "learning_rate": 1.830541099317229e-06, + "loss": 0.9739, + "step": 65787 + }, + { + "epoch": 4.86, + "learning_rate": 1.830311070838925e-06, + "loss": 0.9263, + "step": 65788 + }, + { + "epoch": 4.86, + "learning_rate": 1.8300810553584735e-06, + "loss": 1.0415, + "step": 65789 + }, + { + "epoch": 4.86, + "learning_rate": 1.8298510528762481e-06, + "loss": 0.9951, + "step": 65790 + }, + { + "epoch": 4.86, + "learning_rate": 1.8296210633926136e-06, + "loss": 0.934, + "step": 65791 + }, + { + "epoch": 4.86, + "learning_rate": 1.8293910869079357e-06, + "loss": 0.9502, + "step": 65792 + }, + { + "epoch": 4.86, + "learning_rate": 1.829161123422576e-06, + "loss": 1.0455, + "step": 65793 + }, + { + "epoch": 4.86, + "learning_rate": 1.8289311729369076e-06, + "loss": 1.0629, + "step": 65794 + }, + { + "epoch": 4.86, + "learning_rate": 1.8287012354512923e-06, + "loss": 0.9951, + "step": 65795 + }, + { + "epoch": 4.86, + "learning_rate": 1.8284713109660956e-06, + "loss": 1.0236, + "step": 65796 + }, + { + "epoch": 4.86, + "learning_rate": 1.828241399481685e-06, + "loss": 0.9456, + "step": 65797 + }, + { + "epoch": 4.86, + "learning_rate": 1.8280115009984235e-06, + "loss": 1.0524, + "step": 65798 + }, + { + "epoch": 4.86, + "learning_rate": 1.8277816155166806e-06, + "loss": 1.062, + "step": 65799 + }, + { + "epoch": 4.86, + "learning_rate": 1.8275517430368206e-06, + "loss": 0.9785, + "step": 65800 + }, + { + "epoch": 4.86, + "learning_rate": 1.8273218835592077e-06, + "loss": 1.0276, + "step": 65801 + }, + { + "epoch": 4.86, + "learning_rate": 1.8270920370842072e-06, + "loss": 0.9872, + "step": 65802 + }, + { + "epoch": 4.86, + "learning_rate": 1.8268622036121887e-06, + "loss": 1.0131, + "step": 65803 + }, + { + "epoch": 4.86, + "learning_rate": 1.8266323831435184e-06, + "loss": 0.9854, + "step": 65804 + }, + { + "epoch": 4.86, + "learning_rate": 1.8264025756785531e-06, + "loss": 0.9915, + "step": 65805 + }, + { + "epoch": 4.86, + "learning_rate": 1.8261727812176678e-06, + "loss": 0.9585, + "step": 65806 + }, + { + "epoch": 4.86, + "learning_rate": 1.825942999761221e-06, + "loss": 0.9988, + "step": 65807 + }, + { + "epoch": 4.86, + "learning_rate": 1.8257132313095882e-06, + "loss": 1.1361, + "step": 65808 + }, + { + "epoch": 4.86, + "learning_rate": 1.8254834758631234e-06, + "loss": 0.8668, + "step": 65809 + }, + { + "epoch": 4.86, + "learning_rate": 1.8252537334221999e-06, + "loss": 1.02, + "step": 65810 + }, + { + "epoch": 4.86, + "learning_rate": 1.825024003987177e-06, + "loss": 0.9876, + "step": 65811 + }, + { + "epoch": 4.86, + "learning_rate": 1.8247942875584302e-06, + "loss": 1.0237, + "step": 65812 + }, + { + "epoch": 4.86, + "learning_rate": 1.8245645841363157e-06, + "loss": 1.0984, + "step": 65813 + }, + { + "epoch": 4.86, + "learning_rate": 1.8243348937212002e-06, + "loss": 1.101, + "step": 65814 + }, + { + "epoch": 4.86, + "learning_rate": 1.8241052163134531e-06, + "loss": 0.981, + "step": 65815 + }, + { + "epoch": 4.86, + "learning_rate": 1.8238755519134377e-06, + "loss": 1.0628, + "step": 65816 + }, + { + "epoch": 4.86, + "learning_rate": 1.823645900521519e-06, + "loss": 1.0574, + "step": 65817 + }, + { + "epoch": 4.86, + "learning_rate": 1.8234162621380601e-06, + "loss": 0.9689, + "step": 65818 + }, + { + "epoch": 4.86, + "learning_rate": 1.823186636763432e-06, + "loss": 1.0762, + "step": 65819 + }, + { + "epoch": 4.86, + "learning_rate": 1.8229570243979966e-06, + "loss": 1.0235, + "step": 65820 + }, + { + "epoch": 4.86, + "learning_rate": 1.822727425042119e-06, + "loss": 1.0592, + "step": 65821 + }, + { + "epoch": 4.86, + "learning_rate": 1.8224978386961655e-06, + "loss": 1.0165, + "step": 65822 + }, + { + "epoch": 4.86, + "learning_rate": 1.8222682653604994e-06, + "loss": 0.9066, + "step": 65823 + }, + { + "epoch": 4.86, + "learning_rate": 1.8220387050354894e-06, + "loss": 1.0339, + "step": 65824 + }, + { + "epoch": 4.86, + "learning_rate": 1.821809157721498e-06, + "loss": 1.1166, + "step": 65825 + }, + { + "epoch": 4.86, + "learning_rate": 1.821579623418892e-06, + "loss": 1.0056, + "step": 65826 + }, + { + "epoch": 4.86, + "learning_rate": 1.8213501021280334e-06, + "loss": 0.9151, + "step": 65827 + }, + { + "epoch": 4.86, + "learning_rate": 1.8211205938492914e-06, + "loss": 1.058, + "step": 65828 + }, + { + "epoch": 4.86, + "learning_rate": 1.8208910985830296e-06, + "loss": 0.9676, + "step": 65829 + }, + { + "epoch": 4.86, + "learning_rate": 1.8206616163296143e-06, + "loss": 0.9784, + "step": 65830 + }, + { + "epoch": 4.86, + "learning_rate": 1.8204321470894081e-06, + "loss": 1.0553, + "step": 65831 + }, + { + "epoch": 4.86, + "learning_rate": 1.8202026908627757e-06, + "loss": 1.0198, + "step": 65832 + }, + { + "epoch": 4.86, + "learning_rate": 1.8199732476500887e-06, + "loss": 1.0069, + "step": 65833 + }, + { + "epoch": 4.86, + "learning_rate": 1.8197438174517012e-06, + "loss": 0.9794, + "step": 65834 + }, + { + "epoch": 4.86, + "learning_rate": 1.8195144002679877e-06, + "loss": 1.0275, + "step": 65835 + }, + { + "epoch": 4.86, + "learning_rate": 1.8192849960993098e-06, + "loss": 0.9297, + "step": 65836 + }, + { + "epoch": 4.86, + "learning_rate": 1.8190556049460318e-06, + "loss": 0.9986, + "step": 65837 + }, + { + "epoch": 4.86, + "learning_rate": 1.8188262268085176e-06, + "loss": 1.1246, + "step": 65838 + }, + { + "epoch": 4.86, + "learning_rate": 1.8185968616871351e-06, + "loss": 0.9386, + "step": 65839 + }, + { + "epoch": 4.86, + "learning_rate": 1.8183675095822494e-06, + "loss": 0.9643, + "step": 65840 + }, + { + "epoch": 4.86, + "learning_rate": 1.8181381704942225e-06, + "loss": 1.1521, + "step": 65841 + }, + { + "epoch": 4.87, + "learning_rate": 1.8179088444234217e-06, + "loss": 0.905, + "step": 65842 + }, + { + "epoch": 4.87, + "learning_rate": 1.8176795313702078e-06, + "loss": 0.942, + "step": 65843 + }, + { + "epoch": 4.87, + "learning_rate": 1.8174502313349517e-06, + "loss": 0.9548, + "step": 65844 + }, + { + "epoch": 4.87, + "learning_rate": 1.8172209443180154e-06, + "loss": 1.0217, + "step": 65845 + }, + { + "epoch": 4.87, + "learning_rate": 1.8169916703197632e-06, + "loss": 0.9837, + "step": 65846 + }, + { + "epoch": 4.87, + "learning_rate": 1.8167624093405566e-06, + "loss": 0.9353, + "step": 65847 + }, + { + "epoch": 4.87, + "learning_rate": 1.816533161380768e-06, + "loss": 0.958, + "step": 65848 + }, + { + "epoch": 4.87, + "learning_rate": 1.8163039264407578e-06, + "loss": 0.9816, + "step": 65849 + }, + { + "epoch": 4.87, + "learning_rate": 1.8160747045208903e-06, + "loss": 0.8869, + "step": 65850 + }, + { + "epoch": 4.87, + "learning_rate": 1.815845495621531e-06, + "loss": 0.9429, + "step": 65851 + }, + { + "epoch": 4.87, + "learning_rate": 1.8156162997430427e-06, + "loss": 1.0076, + "step": 65852 + }, + { + "epoch": 4.87, + "learning_rate": 1.8153871168857962e-06, + "loss": 1.0488, + "step": 65853 + }, + { + "epoch": 4.87, + "learning_rate": 1.8151579470501456e-06, + "loss": 0.9304, + "step": 65854 + }, + { + "epoch": 4.87, + "learning_rate": 1.8149287902364654e-06, + "loss": 0.932, + "step": 65855 + }, + { + "epoch": 4.87, + "learning_rate": 1.8146996464451139e-06, + "loss": 1.0114, + "step": 65856 + }, + { + "epoch": 4.87, + "learning_rate": 1.814470515676463e-06, + "loss": 1.0639, + "step": 65857 + }, + { + "epoch": 4.87, + "learning_rate": 1.814241397930867e-06, + "loss": 1.0057, + "step": 65858 + }, + { + "epoch": 4.87, + "learning_rate": 1.814012293208699e-06, + "loss": 1.1011, + "step": 65859 + }, + { + "epoch": 4.87, + "learning_rate": 1.8137832015103196e-06, + "loss": 1.0291, + "step": 65860 + }, + { + "epoch": 4.87, + "learning_rate": 1.8135541228360943e-06, + "loss": 0.8353, + "step": 65861 + }, + { + "epoch": 4.87, + "learning_rate": 1.8133250571863869e-06, + "loss": 0.933, + "step": 65862 + }, + { + "epoch": 4.87, + "learning_rate": 1.8130960045615608e-06, + "loss": 1.0452, + "step": 65863 + }, + { + "epoch": 4.87, + "learning_rate": 1.8128669649619834e-06, + "loss": 1.0503, + "step": 65864 + }, + { + "epoch": 4.87, + "learning_rate": 1.8126379383880177e-06, + "loss": 1.0554, + "step": 65865 + }, + { + "epoch": 4.87, + "learning_rate": 1.8124089248400277e-06, + "loss": 1.0129, + "step": 65866 + }, + { + "epoch": 4.87, + "learning_rate": 1.8121799243183758e-06, + "loss": 1.0149, + "step": 65867 + }, + { + "epoch": 4.87, + "learning_rate": 1.811950936823431e-06, + "loss": 0.961, + "step": 65868 + }, + { + "epoch": 4.87, + "learning_rate": 1.8117219623555548e-06, + "loss": 1.0196, + "step": 65869 + }, + { + "epoch": 4.87, + "learning_rate": 1.8114930009151121e-06, + "loss": 0.8959, + "step": 65870 + }, + { + "epoch": 4.87, + "learning_rate": 1.811264052502466e-06, + "loss": 0.9114, + "step": 65871 + }, + { + "epoch": 4.87, + "learning_rate": 1.8110351171179808e-06, + "loss": 1.0078, + "step": 65872 + }, + { + "epoch": 4.87, + "learning_rate": 1.8108061947620225e-06, + "loss": 0.9594, + "step": 65873 + }, + { + "epoch": 4.87, + "learning_rate": 1.8105772854349556e-06, + "loss": 1.1028, + "step": 65874 + }, + { + "epoch": 4.87, + "learning_rate": 1.810348389137142e-06, + "loss": 1.0479, + "step": 65875 + }, + { + "epoch": 4.87, + "learning_rate": 1.8101195058689446e-06, + "loss": 1.0213, + "step": 65876 + }, + { + "epoch": 4.87, + "learning_rate": 1.8098906356307334e-06, + "loss": 0.9542, + "step": 65877 + }, + { + "epoch": 4.87, + "learning_rate": 1.8096617784228709e-06, + "loss": 1.0392, + "step": 65878 + }, + { + "epoch": 4.87, + "learning_rate": 1.809432934245714e-06, + "loss": 1.03, + "step": 65879 + }, + { + "epoch": 4.87, + "learning_rate": 1.8092041030996343e-06, + "loss": 1.0103, + "step": 65880 + }, + { + "epoch": 4.87, + "learning_rate": 1.808975284984995e-06, + "loss": 1.0063, + "step": 65881 + }, + { + "epoch": 4.87, + "learning_rate": 1.8087464799021581e-06, + "loss": 0.999, + "step": 65882 + }, + { + "epoch": 4.87, + "learning_rate": 1.8085176878514854e-06, + "loss": 1.0206, + "step": 65883 + }, + { + "epoch": 4.87, + "learning_rate": 1.8082889088333467e-06, + "loss": 0.9663, + "step": 65884 + }, + { + "epoch": 4.87, + "learning_rate": 1.8080601428481037e-06, + "loss": 0.9885, + "step": 65885 + }, + { + "epoch": 4.87, + "learning_rate": 1.8078313898961186e-06, + "loss": 1.0914, + "step": 65886 + }, + { + "epoch": 4.87, + "learning_rate": 1.8076026499777577e-06, + "loss": 0.9669, + "step": 65887 + }, + { + "epoch": 4.87, + "learning_rate": 1.8073739230933807e-06, + "loss": 0.968, + "step": 65888 + }, + { + "epoch": 4.87, + "learning_rate": 1.8071452092433572e-06, + "loss": 0.9274, + "step": 65889 + }, + { + "epoch": 4.87, + "learning_rate": 1.8069165084280483e-06, + "loss": 1.11, + "step": 65890 + }, + { + "epoch": 4.87, + "learning_rate": 1.806687820647819e-06, + "loss": 0.9539, + "step": 65891 + }, + { + "epoch": 4.87, + "learning_rate": 1.8064591459030279e-06, + "loss": 1.0503, + "step": 65892 + }, + { + "epoch": 4.87, + "learning_rate": 1.806230484194047e-06, + "loss": 1.0127, + "step": 65893 + }, + { + "epoch": 4.87, + "learning_rate": 1.806001835521235e-06, + "loss": 1.0546, + "step": 65894 + }, + { + "epoch": 4.87, + "learning_rate": 1.8057731998849582e-06, + "loss": 1.0364, + "step": 65895 + }, + { + "epoch": 4.87, + "learning_rate": 1.8055445772855752e-06, + "loss": 1.0712, + "step": 65896 + }, + { + "epoch": 4.87, + "learning_rate": 1.8053159677234567e-06, + "loss": 1.0859, + "step": 65897 + }, + { + "epoch": 4.87, + "learning_rate": 1.8050873711989658e-06, + "loss": 0.933, + "step": 65898 + }, + { + "epoch": 4.87, + "learning_rate": 1.804858787712459e-06, + "loss": 1.0708, + "step": 65899 + }, + { + "epoch": 4.87, + "learning_rate": 1.804630217264306e-06, + "loss": 1.0552, + "step": 65900 + }, + { + "epoch": 4.87, + "learning_rate": 1.8044016598548674e-06, + "loss": 1.103, + "step": 65901 + }, + { + "epoch": 4.87, + "learning_rate": 1.8041731154845132e-06, + "loss": 0.992, + "step": 65902 + }, + { + "epoch": 4.87, + "learning_rate": 1.8039445841535975e-06, + "loss": 0.9921, + "step": 65903 + }, + { + "epoch": 4.87, + "learning_rate": 1.8037160658624919e-06, + "loss": 1.1133, + "step": 65904 + }, + { + "epoch": 4.87, + "learning_rate": 1.803487560611553e-06, + "loss": 1.0658, + "step": 65905 + }, + { + "epoch": 4.87, + "learning_rate": 1.803259068401154e-06, + "loss": 1.0014, + "step": 65906 + }, + { + "epoch": 4.87, + "learning_rate": 1.8030305892316491e-06, + "loss": 0.9872, + "step": 65907 + }, + { + "epoch": 4.87, + "learning_rate": 1.8028021231034031e-06, + "loss": 0.8925, + "step": 65908 + }, + { + "epoch": 4.87, + "learning_rate": 1.8025736700167851e-06, + "loss": 1.004, + "step": 65909 + }, + { + "epoch": 4.87, + "learning_rate": 1.8023452299721545e-06, + "loss": 0.9667, + "step": 65910 + }, + { + "epoch": 4.87, + "learning_rate": 1.8021168029698744e-06, + "loss": 0.9343, + "step": 65911 + }, + { + "epoch": 4.87, + "learning_rate": 1.8018883890103077e-06, + "loss": 0.8413, + "step": 65912 + }, + { + "epoch": 4.87, + "learning_rate": 1.8016599880938214e-06, + "loss": 0.9809, + "step": 65913 + }, + { + "epoch": 4.87, + "learning_rate": 1.8014316002207777e-06, + "loss": 1.1179, + "step": 65914 + }, + { + "epoch": 4.87, + "learning_rate": 1.801203225391538e-06, + "loss": 0.9858, + "step": 65915 + }, + { + "epoch": 4.87, + "learning_rate": 1.800974863606466e-06, + "loss": 0.9644, + "step": 65916 + }, + { + "epoch": 4.87, + "learning_rate": 1.8007465148659243e-06, + "loss": 1.0413, + "step": 65917 + }, + { + "epoch": 4.87, + "learning_rate": 1.8005181791702809e-06, + "loss": 1.0469, + "step": 65918 + }, + { + "epoch": 4.87, + "learning_rate": 1.8002898565198944e-06, + "loss": 1.0044, + "step": 65919 + }, + { + "epoch": 4.87, + "learning_rate": 1.8000615469151295e-06, + "loss": 1.0447, + "step": 65920 + }, + { + "epoch": 4.87, + "learning_rate": 1.7998332503563466e-06, + "loss": 0.9609, + "step": 65921 + }, + { + "epoch": 4.87, + "learning_rate": 1.7996049668439153e-06, + "loss": 1.0752, + "step": 65922 + }, + { + "epoch": 4.87, + "learning_rate": 1.7993766963781945e-06, + "loss": 0.935, + "step": 65923 + }, + { + "epoch": 4.87, + "learning_rate": 1.7991484389595482e-06, + "loss": 0.9782, + "step": 65924 + }, + { + "epoch": 4.87, + "learning_rate": 1.7989201945883394e-06, + "loss": 0.9189, + "step": 65925 + }, + { + "epoch": 4.87, + "learning_rate": 1.7986919632649314e-06, + "loss": 1.0045, + "step": 65926 + }, + { + "epoch": 4.87, + "learning_rate": 1.7984637449896858e-06, + "loss": 1.0244, + "step": 65927 + }, + { + "epoch": 4.87, + "learning_rate": 1.7982355397629658e-06, + "loss": 1.0504, + "step": 65928 + }, + { + "epoch": 4.87, + "learning_rate": 1.7980073475851378e-06, + "loss": 1.0399, + "step": 65929 + }, + { + "epoch": 4.87, + "learning_rate": 1.7977791684565637e-06, + "loss": 0.992, + "step": 65930 + }, + { + "epoch": 4.87, + "learning_rate": 1.7975510023776044e-06, + "loss": 1.0731, + "step": 65931 + }, + { + "epoch": 4.87, + "learning_rate": 1.7973228493486228e-06, + "loss": 0.98, + "step": 65932 + }, + { + "epoch": 4.87, + "learning_rate": 1.7970947093699843e-06, + "loss": 1.0174, + "step": 65933 + }, + { + "epoch": 4.87, + "learning_rate": 1.7968665824420517e-06, + "loss": 0.9954, + "step": 65934 + }, + { + "epoch": 4.87, + "learning_rate": 1.796638468565186e-06, + "loss": 0.995, + "step": 65935 + }, + { + "epoch": 4.87, + "learning_rate": 1.7964103677397527e-06, + "loss": 1.0772, + "step": 65936 + }, + { + "epoch": 4.87, + "learning_rate": 1.79618227996611e-06, + "loss": 1.0432, + "step": 65937 + }, + { + "epoch": 4.87, + "learning_rate": 1.7959542052446255e-06, + "loss": 0.9449, + "step": 65938 + }, + { + "epoch": 4.87, + "learning_rate": 1.7957261435756613e-06, + "loss": 0.981, + "step": 65939 + }, + { + "epoch": 4.87, + "learning_rate": 1.7954980949595791e-06, + "loss": 1.0388, + "step": 65940 + }, + { + "epoch": 4.87, + "learning_rate": 1.7952700593967398e-06, + "loss": 0.9796, + "step": 65941 + }, + { + "epoch": 4.87, + "learning_rate": 1.795042036887511e-06, + "loss": 0.9651, + "step": 65942 + }, + { + "epoch": 4.87, + "learning_rate": 1.7948140274322556e-06, + "loss": 1.0069, + "step": 65943 + }, + { + "epoch": 4.87, + "learning_rate": 1.794586031031328e-06, + "loss": 1.0714, + "step": 65944 + }, + { + "epoch": 4.87, + "learning_rate": 1.7943580476850996e-06, + "loss": 1.0592, + "step": 65945 + }, + { + "epoch": 4.87, + "learning_rate": 1.7941300773939273e-06, + "loss": 1.0623, + "step": 65946 + }, + { + "epoch": 4.87, + "learning_rate": 1.7939021201581819e-06, + "loss": 0.9762, + "step": 65947 + }, + { + "epoch": 4.87, + "learning_rate": 1.7936741759782162e-06, + "loss": 1.0539, + "step": 65948 + }, + { + "epoch": 4.87, + "learning_rate": 1.7934462448543988e-06, + "loss": 1.0187, + "step": 65949 + }, + { + "epoch": 4.87, + "learning_rate": 1.7932183267870895e-06, + "loss": 0.9832, + "step": 65950 + }, + { + "epoch": 4.87, + "learning_rate": 1.792990421776657e-06, + "loss": 0.9491, + "step": 65951 + }, + { + "epoch": 4.87, + "learning_rate": 1.7927625298234553e-06, + "loss": 1.0015, + "step": 65952 + }, + { + "epoch": 4.87, + "learning_rate": 1.7925346509278519e-06, + "loss": 0.9254, + "step": 65953 + }, + { + "epoch": 4.87, + "learning_rate": 1.7923067850902098e-06, + "loss": 0.9794, + "step": 65954 + }, + { + "epoch": 4.87, + "learning_rate": 1.79207893231089e-06, + "loss": 0.9462, + "step": 65955 + }, + { + "epoch": 4.87, + "learning_rate": 1.7918510925902543e-06, + "loss": 0.9791, + "step": 65956 + }, + { + "epoch": 4.87, + "learning_rate": 1.7916232659286636e-06, + "loss": 0.9805, + "step": 65957 + }, + { + "epoch": 4.87, + "learning_rate": 1.7913954523264864e-06, + "loss": 1.0746, + "step": 65958 + }, + { + "epoch": 4.87, + "learning_rate": 1.7911676517840804e-06, + "loss": 1.056, + "step": 65959 + }, + { + "epoch": 4.87, + "learning_rate": 1.7909398643018106e-06, + "loss": 1.0085, + "step": 65960 + }, + { + "epoch": 4.87, + "learning_rate": 1.7907120898800346e-06, + "loss": 1.067, + "step": 65961 + }, + { + "epoch": 4.87, + "learning_rate": 1.7904843285191208e-06, + "loss": 1.0597, + "step": 65962 + }, + { + "epoch": 4.87, + "learning_rate": 1.7902565802194295e-06, + "loss": 0.9749, + "step": 65963 + }, + { + "epoch": 4.87, + "learning_rate": 1.7900288449813218e-06, + "loss": 1.0183, + "step": 65964 + }, + { + "epoch": 4.87, + "learning_rate": 1.7898011228051603e-06, + "loss": 0.9342, + "step": 65965 + }, + { + "epoch": 4.87, + "learning_rate": 1.7895734136913068e-06, + "loss": 0.9965, + "step": 65966 + }, + { + "epoch": 4.87, + "learning_rate": 1.7893457176401251e-06, + "loss": 1.0686, + "step": 65967 + }, + { + "epoch": 4.87, + "learning_rate": 1.7891180346519788e-06, + "loss": 0.9752, + "step": 65968 + }, + { + "epoch": 4.87, + "learning_rate": 1.788890364727227e-06, + "loss": 1.0129, + "step": 65969 + }, + { + "epoch": 4.87, + "learning_rate": 1.788662707866231e-06, + "loss": 0.9389, + "step": 65970 + }, + { + "epoch": 4.87, + "learning_rate": 1.7884350640693604e-06, + "loss": 0.9519, + "step": 65971 + }, + { + "epoch": 4.87, + "learning_rate": 1.7882074333369692e-06, + "loss": 0.9118, + "step": 65972 + }, + { + "epoch": 4.87, + "learning_rate": 1.7879798156694205e-06, + "loss": 1.0644, + "step": 65973 + }, + { + "epoch": 4.87, + "learning_rate": 1.7877522110670808e-06, + "loss": 1.0904, + "step": 65974 + }, + { + "epoch": 4.87, + "learning_rate": 1.7875246195303086e-06, + "loss": 0.8579, + "step": 65975 + }, + { + "epoch": 4.87, + "learning_rate": 1.7872970410594693e-06, + "loss": 1.0218, + "step": 65976 + }, + { + "epoch": 4.87, + "learning_rate": 1.787069475654919e-06, + "loss": 0.9703, + "step": 65977 + }, + { + "epoch": 4.88, + "learning_rate": 1.7868419233170263e-06, + "loss": 1.037, + "step": 65978 + }, + { + "epoch": 4.88, + "learning_rate": 1.7866143840461515e-06, + "loss": 0.9487, + "step": 65979 + }, + { + "epoch": 4.88, + "learning_rate": 1.7863868578426547e-06, + "loss": 0.9784, + "step": 65980 + }, + { + "epoch": 4.88, + "learning_rate": 1.7861593447068992e-06, + "loss": 1.075, + "step": 65981 + }, + { + "epoch": 4.88, + "learning_rate": 1.7859318446392437e-06, + "loss": 0.9803, + "step": 65982 + }, + { + "epoch": 4.88, + "learning_rate": 1.7857043576400568e-06, + "loss": 1.0378, + "step": 65983 + }, + { + "epoch": 4.88, + "learning_rate": 1.7854768837096959e-06, + "loss": 1.101, + "step": 65984 + }, + { + "epoch": 4.88, + "learning_rate": 1.785249422848524e-06, + "loss": 1.0272, + "step": 65985 + }, + { + "epoch": 4.88, + "learning_rate": 1.7850219750568997e-06, + "loss": 0.9151, + "step": 65986 + }, + { + "epoch": 4.88, + "learning_rate": 1.7847945403351908e-06, + "loss": 1.0022, + "step": 65987 + }, + { + "epoch": 4.88, + "learning_rate": 1.7845671186837565e-06, + "loss": 0.9557, + "step": 65988 + }, + { + "epoch": 4.88, + "learning_rate": 1.784339710102958e-06, + "loss": 1.0535, + "step": 65989 + }, + { + "epoch": 4.88, + "learning_rate": 1.7841123145931573e-06, + "loss": 0.8878, + "step": 65990 + }, + { + "epoch": 4.88, + "learning_rate": 1.7838849321547135e-06, + "loss": 1.003, + "step": 65991 + }, + { + "epoch": 4.88, + "learning_rate": 1.783657562787997e-06, + "loss": 0.9278, + "step": 65992 + }, + { + "epoch": 4.88, + "learning_rate": 1.7834302064933595e-06, + "loss": 0.987, + "step": 65993 + }, + { + "epoch": 4.88, + "learning_rate": 1.7832028632711683e-06, + "loss": 1.0174, + "step": 65994 + }, + { + "epoch": 4.88, + "learning_rate": 1.7829755331217813e-06, + "loss": 1.033, + "step": 65995 + }, + { + "epoch": 4.88, + "learning_rate": 1.7827482160455677e-06, + "loss": 0.9261, + "step": 65996 + }, + { + "epoch": 4.88, + "learning_rate": 1.78252091204288e-06, + "loss": 0.9954, + "step": 65997 + }, + { + "epoch": 4.88, + "learning_rate": 1.782293621114085e-06, + "loss": 0.9394, + "step": 65998 + }, + { + "epoch": 4.88, + "learning_rate": 1.7820663432595443e-06, + "loss": 0.9935, + "step": 65999 + }, + { + "epoch": 4.88, + "learning_rate": 1.7818390784796192e-06, + "loss": 1.0643, + "step": 66000 + }, + { + "epoch": 4.88, + "learning_rate": 1.7816118267746696e-06, + "loss": 1.1083, + "step": 66001 + }, + { + "epoch": 4.88, + "learning_rate": 1.7813845881450552e-06, + "loss": 0.9914, + "step": 66002 + }, + { + "epoch": 4.88, + "learning_rate": 1.7811573625911427e-06, + "loss": 0.9849, + "step": 66003 + }, + { + "epoch": 4.88, + "learning_rate": 1.7809301501132914e-06, + "loss": 0.9665, + "step": 66004 + }, + { + "epoch": 4.88, + "learning_rate": 1.7807029507118633e-06, + "loss": 1.0044, + "step": 66005 + }, + { + "epoch": 4.88, + "learning_rate": 1.7804757643872161e-06, + "loss": 0.9433, + "step": 66006 + }, + { + "epoch": 4.88, + "learning_rate": 1.7802485911397171e-06, + "loss": 1.0193, + "step": 66007 + }, + { + "epoch": 4.88, + "learning_rate": 1.780021430969725e-06, + "loss": 0.9805, + "step": 66008 + }, + { + "epoch": 4.88, + "learning_rate": 1.7797942838776017e-06, + "loss": 1.0378, + "step": 66009 + }, + { + "epoch": 4.88, + "learning_rate": 1.7795671498637068e-06, + "loss": 1.1029, + "step": 66010 + }, + { + "epoch": 4.88, + "learning_rate": 1.7793400289284013e-06, + "loss": 0.9489, + "step": 66011 + }, + { + "epoch": 4.88, + "learning_rate": 1.7791129210720514e-06, + "loss": 0.907, + "step": 66012 + }, + { + "epoch": 4.88, + "learning_rate": 1.7788858262950138e-06, + "loss": 1.0207, + "step": 66013 + }, + { + "epoch": 4.88, + "learning_rate": 1.7786587445976521e-06, + "loss": 1.0683, + "step": 66014 + }, + { + "epoch": 4.88, + "learning_rate": 1.7784316759803244e-06, + "loss": 0.9421, + "step": 66015 + }, + { + "epoch": 4.88, + "learning_rate": 1.7782046204433978e-06, + "loss": 1.0337, + "step": 66016 + }, + { + "epoch": 4.88, + "learning_rate": 1.7779775779872267e-06, + "loss": 1.0166, + "step": 66017 + }, + { + "epoch": 4.88, + "learning_rate": 1.777750548612177e-06, + "loss": 1.0426, + "step": 66018 + }, + { + "epoch": 4.88, + "learning_rate": 1.777523532318609e-06, + "loss": 0.9238, + "step": 66019 + }, + { + "epoch": 4.88, + "learning_rate": 1.7772965291068834e-06, + "loss": 1.0362, + "step": 66020 + }, + { + "epoch": 4.88, + "learning_rate": 1.7770695389773618e-06, + "loss": 1.1213, + "step": 66021 + }, + { + "epoch": 4.88, + "learning_rate": 1.776842561930402e-06, + "loss": 0.9836, + "step": 66022 + }, + { + "epoch": 4.88, + "learning_rate": 1.7766155979663701e-06, + "loss": 0.9923, + "step": 66023 + }, + { + "epoch": 4.88, + "learning_rate": 1.7763886470856262e-06, + "loss": 1.0118, + "step": 66024 + }, + { + "epoch": 4.88, + "learning_rate": 1.7761617092885296e-06, + "loss": 0.9084, + "step": 66025 + }, + { + "epoch": 4.88, + "learning_rate": 1.7759347845754392e-06, + "loss": 1.0318, + "step": 66026 + }, + { + "epoch": 4.88, + "learning_rate": 1.7757078729467226e-06, + "loss": 1.138, + "step": 66027 + }, + { + "epoch": 4.88, + "learning_rate": 1.7754809744027358e-06, + "loss": 0.9385, + "step": 66028 + }, + { + "epoch": 4.88, + "learning_rate": 1.775254088943842e-06, + "loss": 0.9139, + "step": 66029 + }, + { + "epoch": 4.88, + "learning_rate": 1.7750272165704009e-06, + "loss": 0.9956, + "step": 66030 + }, + { + "epoch": 4.88, + "learning_rate": 1.774800357282771e-06, + "loss": 1.0501, + "step": 66031 + }, + { + "epoch": 4.88, + "learning_rate": 1.7745735110813188e-06, + "loss": 1.0301, + "step": 66032 + }, + { + "epoch": 4.88, + "learning_rate": 1.774346677966402e-06, + "loss": 1.0595, + "step": 66033 + }, + { + "epoch": 4.88, + "learning_rate": 1.7741198579383834e-06, + "loss": 1.0874, + "step": 66034 + }, + { + "epoch": 4.88, + "learning_rate": 1.7738930509976183e-06, + "loss": 1.0997, + "step": 66035 + }, + { + "epoch": 4.88, + "learning_rate": 1.7736662571444752e-06, + "loss": 0.9903, + "step": 66036 + }, + { + "epoch": 4.88, + "learning_rate": 1.7734394763793129e-06, + "loss": 0.981, + "step": 66037 + }, + { + "epoch": 4.88, + "learning_rate": 1.7732127087024863e-06, + "loss": 1.0141, + "step": 66038 + }, + { + "epoch": 4.88, + "learning_rate": 1.7729859541143635e-06, + "loss": 1.0098, + "step": 66039 + }, + { + "epoch": 4.88, + "learning_rate": 1.7727592126152992e-06, + "loss": 1.061, + "step": 66040 + }, + { + "epoch": 4.88, + "learning_rate": 1.7725324842056624e-06, + "loss": 1.0916, + "step": 66041 + }, + { + "epoch": 4.88, + "learning_rate": 1.7723057688858037e-06, + "loss": 1.0297, + "step": 66042 + }, + { + "epoch": 4.88, + "learning_rate": 1.7720790666560916e-06, + "loss": 0.9868, + "step": 66043 + }, + { + "epoch": 4.88, + "learning_rate": 1.7718523775168816e-06, + "loss": 1.1466, + "step": 66044 + }, + { + "epoch": 4.88, + "learning_rate": 1.7716257014685412e-06, + "loss": 1.1195, + "step": 66045 + }, + { + "epoch": 4.88, + "learning_rate": 1.7713990385114243e-06, + "loss": 1.0516, + "step": 66046 + }, + { + "epoch": 4.88, + "learning_rate": 1.7711723886458909e-06, + "loss": 1.0053, + "step": 66047 + }, + { + "epoch": 4.88, + "learning_rate": 1.7709457518723083e-06, + "loss": 0.9061, + "step": 66048 + }, + { + "epoch": 4.88, + "learning_rate": 1.770719128191032e-06, + "loss": 0.9026, + "step": 66049 + }, + { + "epoch": 4.88, + "learning_rate": 1.7704925176024245e-06, + "loss": 0.9947, + "step": 66050 + }, + { + "epoch": 4.88, + "learning_rate": 1.7702659201068429e-06, + "loss": 0.9858, + "step": 66051 + }, + { + "epoch": 4.88, + "learning_rate": 1.7700393357046519e-06, + "loss": 0.9229, + "step": 66052 + }, + { + "epoch": 4.88, + "learning_rate": 1.7698127643962127e-06, + "loss": 0.9745, + "step": 66053 + }, + { + "epoch": 4.88, + "learning_rate": 1.7695862061818814e-06, + "loss": 1.0679, + "step": 66054 + }, + { + "epoch": 4.88, + "learning_rate": 1.76935966106202e-06, + "loss": 1.0546, + "step": 66055 + }, + { + "epoch": 4.88, + "learning_rate": 1.7691331290369918e-06, + "loss": 0.9952, + "step": 66056 + }, + { + "epoch": 4.88, + "learning_rate": 1.7689066101071562e-06, + "loss": 0.9863, + "step": 66057 + }, + { + "epoch": 4.88, + "learning_rate": 1.7686801042728684e-06, + "loss": 1.0557, + "step": 66058 + }, + { + "epoch": 4.88, + "learning_rate": 1.7684536115344952e-06, + "loss": 1.1079, + "step": 66059 + }, + { + "epoch": 4.88, + "learning_rate": 1.7682271318923927e-06, + "loss": 0.9467, + "step": 66060 + }, + { + "epoch": 4.88, + "learning_rate": 1.7680006653469274e-06, + "loss": 0.9575, + "step": 66061 + }, + { + "epoch": 4.88, + "learning_rate": 1.7677742118984497e-06, + "loss": 1.0643, + "step": 66062 + }, + { + "epoch": 4.88, + "learning_rate": 1.7675477715473287e-06, + "loss": 0.9248, + "step": 66063 + }, + { + "epoch": 4.88, + "learning_rate": 1.7673213442939185e-06, + "loss": 0.993, + "step": 66064 + }, + { + "epoch": 4.88, + "learning_rate": 1.7670949301385875e-06, + "loss": 1.0457, + "step": 66065 + }, + { + "epoch": 4.88, + "learning_rate": 1.766868529081689e-06, + "loss": 0.9597, + "step": 66066 + }, + { + "epoch": 4.88, + "learning_rate": 1.7666421411235813e-06, + "loss": 1.0217, + "step": 66067 + }, + { + "epoch": 4.88, + "learning_rate": 1.7664157662646308e-06, + "loss": 1.0731, + "step": 66068 + }, + { + "epoch": 4.88, + "learning_rate": 1.7661894045051952e-06, + "loss": 0.986, + "step": 66069 + }, + { + "epoch": 4.88, + "learning_rate": 1.7659630558456341e-06, + "loss": 1.0449, + "step": 66070 + }, + { + "epoch": 4.88, + "learning_rate": 1.7657367202863063e-06, + "loss": 1.053, + "step": 66071 + }, + { + "epoch": 4.88, + "learning_rate": 1.7655103978275755e-06, + "loss": 0.8808, + "step": 66072 + }, + { + "epoch": 4.88, + "learning_rate": 1.7652840884697998e-06, + "loss": 0.9882, + "step": 66073 + }, + { + "epoch": 4.88, + "learning_rate": 1.7650577922133395e-06, + "loss": 0.9504, + "step": 66074 + }, + { + "epoch": 4.88, + "learning_rate": 1.7648315090585544e-06, + "loss": 1.0309, + "step": 66075 + }, + { + "epoch": 4.88, + "learning_rate": 1.764605239005801e-06, + "loss": 1.0885, + "step": 66076 + }, + { + "epoch": 4.88, + "learning_rate": 1.764378982055447e-06, + "loss": 1.012, + "step": 66077 + }, + { + "epoch": 4.88, + "learning_rate": 1.7641527382078471e-06, + "loss": 1.0211, + "step": 66078 + }, + { + "epoch": 4.88, + "learning_rate": 1.7639265074633627e-06, + "loss": 1.0232, + "step": 66079 + }, + { + "epoch": 4.88, + "learning_rate": 1.7637002898223509e-06, + "loss": 0.9785, + "step": 66080 + }, + { + "epoch": 4.88, + "learning_rate": 1.7634740852851761e-06, + "loss": 1.0237, + "step": 66081 + }, + { + "epoch": 4.88, + "learning_rate": 1.7632478938521968e-06, + "loss": 0.9219, + "step": 66082 + }, + { + "epoch": 4.88, + "learning_rate": 1.7630217155237728e-06, + "loss": 1.0599, + "step": 66083 + }, + { + "epoch": 4.88, + "learning_rate": 1.7627955503002624e-06, + "loss": 0.9994, + "step": 66084 + }, + { + "epoch": 4.88, + "learning_rate": 1.7625693981820247e-06, + "loss": 1.0185, + "step": 66085 + }, + { + "epoch": 4.88, + "learning_rate": 1.7623432591694256e-06, + "loss": 0.9943, + "step": 66086 + }, + { + "epoch": 4.88, + "learning_rate": 1.762117133262815e-06, + "loss": 0.9745, + "step": 66087 + }, + { + "epoch": 4.88, + "learning_rate": 1.7618910204625616e-06, + "loss": 0.9986, + "step": 66088 + }, + { + "epoch": 4.88, + "learning_rate": 1.7616649207690195e-06, + "loss": 0.9266, + "step": 66089 + }, + { + "epoch": 4.88, + "learning_rate": 1.7614388341825551e-06, + "loss": 1.0783, + "step": 66090 + }, + { + "epoch": 4.88, + "learning_rate": 1.761212760703519e-06, + "loss": 1.0932, + "step": 66091 + }, + { + "epoch": 4.88, + "learning_rate": 1.7609867003322766e-06, + "loss": 1.0115, + "step": 66092 + }, + { + "epoch": 4.88, + "learning_rate": 1.7607606530691878e-06, + "loss": 0.9011, + "step": 66093 + }, + { + "epoch": 4.88, + "learning_rate": 1.760534618914611e-06, + "loss": 0.9914, + "step": 66094 + }, + { + "epoch": 4.88, + "learning_rate": 1.7603085978689062e-06, + "loss": 1.0355, + "step": 66095 + }, + { + "epoch": 4.88, + "learning_rate": 1.760082589932428e-06, + "loss": 1.0485, + "step": 66096 + }, + { + "epoch": 4.88, + "learning_rate": 1.7598565951055446e-06, + "loss": 1.0466, + "step": 66097 + }, + { + "epoch": 4.88, + "learning_rate": 1.759630613388611e-06, + "loss": 0.8713, + "step": 66098 + }, + { + "epoch": 4.88, + "learning_rate": 1.759404644781988e-06, + "loss": 0.9475, + "step": 66099 + }, + { + "epoch": 4.88, + "learning_rate": 1.7591786892860307e-06, + "loss": 0.9933, + "step": 66100 + }, + { + "epoch": 4.88, + "learning_rate": 1.7589527469011047e-06, + "loss": 0.9428, + "step": 66101 + }, + { + "epoch": 4.88, + "learning_rate": 1.7587268176275707e-06, + "loss": 1.0876, + "step": 66102 + }, + { + "epoch": 4.88, + "learning_rate": 1.7585009014657783e-06, + "loss": 0.9558, + "step": 66103 + }, + { + "epoch": 4.88, + "learning_rate": 1.7582749984160964e-06, + "loss": 1.0247, + "step": 66104 + }, + { + "epoch": 4.88, + "learning_rate": 1.7580491084788775e-06, + "loss": 0.9878, + "step": 66105 + }, + { + "epoch": 4.88, + "learning_rate": 1.7578232316544897e-06, + "loss": 1.0055, + "step": 66106 + }, + { + "epoch": 4.88, + "learning_rate": 1.7575973679432834e-06, + "loss": 0.9861, + "step": 66107 + }, + { + "epoch": 4.88, + "learning_rate": 1.757371517345623e-06, + "loss": 0.9411, + "step": 66108 + }, + { + "epoch": 4.88, + "learning_rate": 1.7571456798618648e-06, + "loss": 1.0312, + "step": 66109 + }, + { + "epoch": 4.88, + "learning_rate": 1.7569198554923739e-06, + "loss": 1.0011, + "step": 66110 + }, + { + "epoch": 4.88, + "learning_rate": 1.7566940442375013e-06, + "loss": 0.9755, + "step": 66111 + }, + { + "epoch": 4.88, + "learning_rate": 1.7564682460976134e-06, + "loss": 0.9862, + "step": 66112 + }, + { + "epoch": 4.89, + "learning_rate": 1.7562424610730665e-06, + "loss": 0.9357, + "step": 66113 + }, + { + "epoch": 4.89, + "learning_rate": 1.7560166891642194e-06, + "loss": 1.0907, + "step": 66114 + }, + { + "epoch": 4.89, + "learning_rate": 1.7557909303714315e-06, + "loss": 0.9623, + "step": 66115 + }, + { + "epoch": 4.89, + "learning_rate": 1.7555651846950605e-06, + "loss": 0.9601, + "step": 66116 + }, + { + "epoch": 4.89, + "learning_rate": 1.7553394521354695e-06, + "loss": 1.0152, + "step": 66117 + }, + { + "epoch": 4.89, + "learning_rate": 1.7551137326930157e-06, + "loss": 1.0976, + "step": 66118 + }, + { + "epoch": 4.89, + "learning_rate": 1.754888026368058e-06, + "loss": 1.0481, + "step": 66119 + }, + { + "epoch": 4.89, + "learning_rate": 1.7546623331609525e-06, + "loss": 1.0727, + "step": 66120 + }, + { + "epoch": 4.89, + "learning_rate": 1.7544366530720647e-06, + "loss": 1.0552, + "step": 66121 + }, + { + "epoch": 4.89, + "learning_rate": 1.7542109861017487e-06, + "loss": 0.9927, + "step": 66122 + }, + { + "epoch": 4.89, + "learning_rate": 1.7539853322503664e-06, + "loss": 1.0224, + "step": 66123 + }, + { + "epoch": 4.89, + "learning_rate": 1.7537596915182754e-06, + "loss": 0.9711, + "step": 66124 + }, + { + "epoch": 4.89, + "learning_rate": 1.7535340639058319e-06, + "loss": 1.0468, + "step": 66125 + }, + { + "epoch": 4.89, + "learning_rate": 1.7533084494134e-06, + "loss": 0.968, + "step": 66126 + }, + { + "epoch": 4.89, + "learning_rate": 1.7530828480413365e-06, + "loss": 0.9775, + "step": 66127 + }, + { + "epoch": 4.89, + "learning_rate": 1.7528572597899995e-06, + "loss": 1.0079, + "step": 66128 + }, + { + "epoch": 4.89, + "learning_rate": 1.7526316846597469e-06, + "loss": 1.0298, + "step": 66129 + }, + { + "epoch": 4.89, + "learning_rate": 1.7524061226509416e-06, + "loss": 1.0353, + "step": 66130 + }, + { + "epoch": 4.89, + "learning_rate": 1.7521805737639431e-06, + "loss": 0.972, + "step": 66131 + }, + { + "epoch": 4.89, + "learning_rate": 1.7519550379991024e-06, + "loss": 1.062, + "step": 66132 + }, + { + "epoch": 4.89, + "learning_rate": 1.7517295153567848e-06, + "loss": 0.9944, + "step": 66133 + }, + { + "epoch": 4.89, + "learning_rate": 1.7515040058373455e-06, + "loss": 1.028, + "step": 66134 + }, + { + "epoch": 4.89, + "learning_rate": 1.75127850944115e-06, + "loss": 0.9838, + "step": 66135 + }, + { + "epoch": 4.89, + "learning_rate": 1.7510530261685476e-06, + "loss": 1.0321, + "step": 66136 + }, + { + "epoch": 4.89, + "learning_rate": 1.750827556019905e-06, + "loss": 0.9469, + "step": 66137 + }, + { + "epoch": 4.89, + "learning_rate": 1.7506020989955773e-06, + "loss": 0.9959, + "step": 66138 + }, + { + "epoch": 4.89, + "learning_rate": 1.7503766550959234e-06, + "loss": 0.964, + "step": 66139 + }, + { + "epoch": 4.89, + "learning_rate": 1.7501512243213026e-06, + "loss": 1.027, + "step": 66140 + }, + { + "epoch": 4.89, + "learning_rate": 1.7499258066720692e-06, + "loss": 0.9005, + "step": 66141 + }, + { + "epoch": 4.89, + "learning_rate": 1.7497004021485896e-06, + "loss": 1.0269, + "step": 66142 + }, + { + "epoch": 4.89, + "learning_rate": 1.7494750107512193e-06, + "loss": 0.9918, + "step": 66143 + }, + { + "epoch": 4.89, + "learning_rate": 1.7492496324803155e-06, + "loss": 0.8715, + "step": 66144 + }, + { + "epoch": 4.89, + "learning_rate": 1.7490242673362345e-06, + "loss": 0.9356, + "step": 66145 + }, + { + "epoch": 4.89, + "learning_rate": 1.7487989153193408e-06, + "loss": 1.0681, + "step": 66146 + }, + { + "epoch": 4.89, + "learning_rate": 1.7485735764299894e-06, + "loss": 0.9815, + "step": 66147 + }, + { + "epoch": 4.89, + "learning_rate": 1.74834825066854e-06, + "loss": 1.0703, + "step": 66148 + }, + { + "epoch": 4.89, + "learning_rate": 1.7481229380353503e-06, + "loss": 0.9085, + "step": 66149 + }, + { + "epoch": 4.89, + "learning_rate": 1.7478976385307766e-06, + "loss": 1.067, + "step": 66150 + }, + { + "epoch": 4.89, + "learning_rate": 1.7476723521551842e-06, + "loss": 1.0786, + "step": 66151 + }, + { + "epoch": 4.89, + "learning_rate": 1.7474470789089226e-06, + "loss": 0.9773, + "step": 66152 + }, + { + "epoch": 4.89, + "learning_rate": 1.7472218187923573e-06, + "loss": 0.9798, + "step": 66153 + }, + { + "epoch": 4.89, + "learning_rate": 1.7469965718058412e-06, + "loss": 1.0391, + "step": 66154 + }, + { + "epoch": 4.89, + "learning_rate": 1.7467713379497398e-06, + "loss": 1.0317, + "step": 66155 + }, + { + "epoch": 4.89, + "learning_rate": 1.7465461172244026e-06, + "loss": 1.1214, + "step": 66156 + }, + { + "epoch": 4.89, + "learning_rate": 1.7463209096301947e-06, + "loss": 1.0393, + "step": 66157 + }, + { + "epoch": 4.89, + "learning_rate": 1.746095715167473e-06, + "loss": 1.1447, + "step": 66158 + }, + { + "epoch": 4.89, + "learning_rate": 1.7458705338365944e-06, + "loss": 1.1541, + "step": 66159 + }, + { + "epoch": 4.89, + "learning_rate": 1.745645365637918e-06, + "loss": 1.0302, + "step": 66160 + }, + { + "epoch": 4.89, + "learning_rate": 1.7454202105717988e-06, + "loss": 0.9042, + "step": 66161 + }, + { + "epoch": 4.89, + "learning_rate": 1.7451950686385999e-06, + "loss": 1.1257, + "step": 66162 + }, + { + "epoch": 4.89, + "learning_rate": 1.7449699398386788e-06, + "loss": 0.9314, + "step": 66163 + }, + { + "epoch": 4.89, + "learning_rate": 1.744744824172392e-06, + "loss": 0.8718, + "step": 66164 + }, + { + "epoch": 4.89, + "learning_rate": 1.7445197216400944e-06, + "loss": 0.9822, + "step": 66165 + }, + { + "epoch": 4.89, + "learning_rate": 1.7442946322421517e-06, + "loss": 1.0967, + "step": 66166 + }, + { + "epoch": 4.89, + "learning_rate": 1.7440695559789178e-06, + "loss": 0.9621, + "step": 66167 + }, + { + "epoch": 4.89, + "learning_rate": 1.7438444928507514e-06, + "loss": 0.8965, + "step": 66168 + }, + { + "epoch": 4.89, + "learning_rate": 1.74361944285801e-06, + "loss": 1.0148, + "step": 66169 + }, + { + "epoch": 4.89, + "learning_rate": 1.7433944060010488e-06, + "loss": 0.9152, + "step": 66170 + }, + { + "epoch": 4.89, + "learning_rate": 1.743169382280232e-06, + "loss": 1.0285, + "step": 66171 + }, + { + "epoch": 4.89, + "learning_rate": 1.7429443716959148e-06, + "loss": 1.0282, + "step": 66172 + }, + { + "epoch": 4.89, + "learning_rate": 1.742719374248455e-06, + "loss": 0.9853, + "step": 66173 + }, + { + "epoch": 4.89, + "learning_rate": 1.7424943899382075e-06, + "loss": 0.9124, + "step": 66174 + }, + { + "epoch": 4.89, + "learning_rate": 1.7422694187655353e-06, + "loss": 1.0508, + "step": 66175 + }, + { + "epoch": 4.89, + "learning_rate": 1.7420444607307952e-06, + "loss": 1.0305, + "step": 66176 + }, + { + "epoch": 4.89, + "learning_rate": 1.7418195158343443e-06, + "loss": 0.9819, + "step": 66177 + }, + { + "epoch": 4.89, + "learning_rate": 1.7415945840765403e-06, + "loss": 1.0026, + "step": 66178 + }, + { + "epoch": 4.89, + "learning_rate": 1.7413696654577416e-06, + "loss": 0.9999, + "step": 66179 + }, + { + "epoch": 4.89, + "learning_rate": 1.7411447599783048e-06, + "loss": 1.0133, + "step": 66180 + }, + { + "epoch": 4.89, + "learning_rate": 1.7409198676385863e-06, + "loss": 1.0311, + "step": 66181 + }, + { + "epoch": 4.89, + "learning_rate": 1.7406949884389479e-06, + "loss": 0.9878, + "step": 66182 + }, + { + "epoch": 4.89, + "learning_rate": 1.740470122379747e-06, + "loss": 0.9304, + "step": 66183 + }, + { + "epoch": 4.89, + "learning_rate": 1.7402452694613391e-06, + "loss": 0.8728, + "step": 66184 + }, + { + "epoch": 4.89, + "learning_rate": 1.7400204296840805e-06, + "loss": 0.9125, + "step": 66185 + }, + { + "epoch": 4.89, + "learning_rate": 1.7397956030483333e-06, + "loss": 0.9427, + "step": 66186 + }, + { + "epoch": 4.89, + "learning_rate": 1.7395707895544534e-06, + "loss": 1.053, + "step": 66187 + }, + { + "epoch": 4.89, + "learning_rate": 1.7393459892027986e-06, + "loss": 1.052, + "step": 66188 + }, + { + "epoch": 4.89, + "learning_rate": 1.7391212019937254e-06, + "loss": 1.0512, + "step": 66189 + }, + { + "epoch": 4.89, + "learning_rate": 1.7388964279275899e-06, + "loss": 0.9677, + "step": 66190 + }, + { + "epoch": 4.89, + "learning_rate": 1.7386716670047555e-06, + "loss": 0.9117, + "step": 66191 + }, + { + "epoch": 4.89, + "learning_rate": 1.7384469192255748e-06, + "loss": 1.0872, + "step": 66192 + }, + { + "epoch": 4.89, + "learning_rate": 1.738222184590408e-06, + "loss": 1.0212, + "step": 66193 + }, + { + "epoch": 4.89, + "learning_rate": 1.7379974630996076e-06, + "loss": 0.9389, + "step": 66194 + }, + { + "epoch": 4.89, + "learning_rate": 1.7377727547535383e-06, + "loss": 1.0214, + "step": 66195 + }, + { + "epoch": 4.89, + "learning_rate": 1.7375480595525573e-06, + "loss": 1.0984, + "step": 66196 + }, + { + "epoch": 4.89, + "learning_rate": 1.7373233774970144e-06, + "loss": 1.0708, + "step": 66197 + }, + { + "epoch": 4.89, + "learning_rate": 1.7370987085872725e-06, + "loss": 0.9164, + "step": 66198 + }, + { + "epoch": 4.89, + "learning_rate": 1.7368740528236872e-06, + "loss": 0.9622, + "step": 66199 + }, + { + "epoch": 4.89, + "learning_rate": 1.7366494102066222e-06, + "loss": 1.0784, + "step": 66200 + }, + { + "epoch": 4.89, + "learning_rate": 1.7364247807364243e-06, + "loss": 1.0565, + "step": 66201 + }, + { + "epoch": 4.89, + "learning_rate": 1.7362001644134585e-06, + "loss": 0.992, + "step": 66202 + }, + { + "epoch": 4.89, + "learning_rate": 1.735975561238078e-06, + "loss": 0.9238, + "step": 66203 + }, + { + "epoch": 4.89, + "learning_rate": 1.7357509712106479e-06, + "loss": 0.9939, + "step": 66204 + }, + { + "epoch": 4.89, + "learning_rate": 1.735526394331516e-06, + "loss": 0.9681, + "step": 66205 + }, + { + "epoch": 4.89, + "learning_rate": 1.7353018306010417e-06, + "loss": 0.9941, + "step": 66206 + }, + { + "epoch": 4.89, + "learning_rate": 1.7350772800195858e-06, + "loss": 1.0106, + "step": 66207 + }, + { + "epoch": 4.89, + "learning_rate": 1.734852742587504e-06, + "loss": 1.0263, + "step": 66208 + }, + { + "epoch": 4.89, + "learning_rate": 1.7346282183051533e-06, + "loss": 1.0747, + "step": 66209 + }, + { + "epoch": 4.89, + "learning_rate": 1.7344037071728882e-06, + "loss": 0.9569, + "step": 66210 + }, + { + "epoch": 4.89, + "learning_rate": 1.7341792091910702e-06, + "loss": 0.9603, + "step": 66211 + }, + { + "epoch": 4.89, + "learning_rate": 1.7339547243600563e-06, + "loss": 1.0909, + "step": 66212 + }, + { + "epoch": 4.89, + "learning_rate": 1.7337302526802014e-06, + "loss": 0.949, + "step": 66213 + }, + { + "epoch": 4.89, + "learning_rate": 1.733505794151864e-06, + "loss": 1.1027, + "step": 66214 + }, + { + "epoch": 4.89, + "learning_rate": 1.7332813487753973e-06, + "loss": 1.0654, + "step": 66215 + }, + { + "epoch": 4.89, + "learning_rate": 1.7330569165511645e-06, + "loss": 1.0372, + "step": 66216 + }, + { + "epoch": 4.89, + "learning_rate": 1.7328324974795197e-06, + "loss": 0.9605, + "step": 66217 + }, + { + "epoch": 4.89, + "learning_rate": 1.73260809156082e-06, + "loss": 1.1242, + "step": 66218 + }, + { + "epoch": 4.89, + "learning_rate": 1.7323836987954212e-06, + "loss": 1.0376, + "step": 66219 + }, + { + "epoch": 4.89, + "learning_rate": 1.7321593191836827e-06, + "loss": 0.9778, + "step": 66220 + }, + { + "epoch": 4.89, + "learning_rate": 1.731934952725961e-06, + "loss": 0.8546, + "step": 66221 + }, + { + "epoch": 4.89, + "learning_rate": 1.7317105994226124e-06, + "loss": 1.1311, + "step": 66222 + }, + { + "epoch": 4.89, + "learning_rate": 1.7314862592739924e-06, + "loss": 1.0513, + "step": 66223 + }, + { + "epoch": 4.89, + "learning_rate": 1.7312619322804636e-06, + "loss": 1.0947, + "step": 66224 + }, + { + "epoch": 4.89, + "learning_rate": 1.731037618442376e-06, + "loss": 1.1255, + "step": 66225 + }, + { + "epoch": 4.89, + "learning_rate": 1.7308133177600873e-06, + "loss": 1.0176, + "step": 66226 + }, + { + "epoch": 4.89, + "learning_rate": 1.730589030233958e-06, + "loss": 1.0721, + "step": 66227 + }, + { + "epoch": 4.89, + "learning_rate": 1.7303647558643433e-06, + "loss": 1.0186, + "step": 66228 + }, + { + "epoch": 4.89, + "learning_rate": 1.730140494651601e-06, + "loss": 1.0348, + "step": 66229 + }, + { + "epoch": 4.89, + "learning_rate": 1.729916246596084e-06, + "loss": 1.0495, + "step": 66230 + }, + { + "epoch": 4.89, + "learning_rate": 1.729692011698154e-06, + "loss": 0.9225, + "step": 66231 + }, + { + "epoch": 4.89, + "learning_rate": 1.7294677899581659e-06, + "loss": 0.9392, + "step": 66232 + }, + { + "epoch": 4.89, + "learning_rate": 1.7292435813764764e-06, + "loss": 1.0411, + "step": 66233 + }, + { + "epoch": 4.89, + "learning_rate": 1.729019385953441e-06, + "loss": 0.9462, + "step": 66234 + }, + { + "epoch": 4.89, + "learning_rate": 1.7287952036894161e-06, + "loss": 0.9506, + "step": 66235 + }, + { + "epoch": 4.89, + "learning_rate": 1.7285710345847617e-06, + "loss": 1.0595, + "step": 66236 + }, + { + "epoch": 4.89, + "learning_rate": 1.7283468786398329e-06, + "loss": 1.0617, + "step": 66237 + }, + { + "epoch": 4.89, + "learning_rate": 1.7281227358549858e-06, + "loss": 0.9089, + "step": 66238 + }, + { + "epoch": 4.89, + "learning_rate": 1.7278986062305736e-06, + "loss": 0.9354, + "step": 66239 + }, + { + "epoch": 4.89, + "learning_rate": 1.7276744897669594e-06, + "loss": 1.0663, + "step": 66240 + }, + { + "epoch": 4.89, + "learning_rate": 1.7274503864644976e-06, + "loss": 0.9702, + "step": 66241 + }, + { + "epoch": 4.89, + "learning_rate": 1.7272262963235431e-06, + "loss": 1.0041, + "step": 66242 + }, + { + "epoch": 4.89, + "learning_rate": 1.7270022193444523e-06, + "loss": 1.0463, + "step": 66243 + }, + { + "epoch": 4.89, + "learning_rate": 1.726778155527582e-06, + "loss": 0.9963, + "step": 66244 + }, + { + "epoch": 4.89, + "learning_rate": 1.7265541048732925e-06, + "loss": 0.9284, + "step": 66245 + }, + { + "epoch": 4.89, + "learning_rate": 1.726330067381934e-06, + "loss": 1.0354, + "step": 66246 + }, + { + "epoch": 4.89, + "learning_rate": 1.7261060430538667e-06, + "loss": 1.0557, + "step": 66247 + }, + { + "epoch": 4.9, + "learning_rate": 1.7258820318894443e-06, + "loss": 0.9065, + "step": 66248 + }, + { + "epoch": 4.9, + "learning_rate": 1.7256580338890305e-06, + "loss": 1.0086, + "step": 66249 + }, + { + "epoch": 4.9, + "learning_rate": 1.7254340490529709e-06, + "loss": 1.0688, + "step": 66250 + }, + { + "epoch": 4.9, + "learning_rate": 1.7252100773816305e-06, + "loss": 1.0075, + "step": 66251 + }, + { + "epoch": 4.9, + "learning_rate": 1.7249861188753624e-06, + "loss": 0.9323, + "step": 66252 + }, + { + "epoch": 4.9, + "learning_rate": 1.7247621735345232e-06, + "loss": 0.9959, + "step": 66253 + }, + { + "epoch": 4.9, + "learning_rate": 1.724538241359468e-06, + "loss": 1.0457, + "step": 66254 + }, + { + "epoch": 4.9, + "learning_rate": 1.724314322350552e-06, + "loss": 1.1152, + "step": 66255 + }, + { + "epoch": 4.9, + "learning_rate": 1.7240904165081362e-06, + "loss": 0.9105, + "step": 66256 + }, + { + "epoch": 4.9, + "learning_rate": 1.7238665238325747e-06, + "loss": 1.0339, + "step": 66257 + }, + { + "epoch": 4.9, + "learning_rate": 1.7236426443242216e-06, + "loss": 1.0603, + "step": 66258 + }, + { + "epoch": 4.9, + "learning_rate": 1.7234187779834333e-06, + "loss": 1.0824, + "step": 66259 + }, + { + "epoch": 4.9, + "learning_rate": 1.7231949248105695e-06, + "loss": 0.9645, + "step": 66260 + }, + { + "epoch": 4.9, + "learning_rate": 1.7229710848059845e-06, + "loss": 0.9627, + "step": 66261 + }, + { + "epoch": 4.9, + "learning_rate": 1.7227472579700332e-06, + "loss": 0.9281, + "step": 66262 + }, + { + "epoch": 4.9, + "learning_rate": 1.7225234443030736e-06, + "loss": 0.9695, + "step": 66263 + }, + { + "epoch": 4.9, + "learning_rate": 1.7222996438054573e-06, + "loss": 0.9729, + "step": 66264 + }, + { + "epoch": 4.9, + "learning_rate": 1.7220758564775475e-06, + "loss": 1.0049, + "step": 66265 + }, + { + "epoch": 4.9, + "learning_rate": 1.721852082319696e-06, + "loss": 1.0164, + "step": 66266 + }, + { + "epoch": 4.9, + "learning_rate": 1.7216283213322594e-06, + "loss": 1.0208, + "step": 66267 + }, + { + "epoch": 4.9, + "learning_rate": 1.7214045735155903e-06, + "loss": 0.9408, + "step": 66268 + }, + { + "epoch": 4.9, + "learning_rate": 1.7211808388700545e-06, + "loss": 1.0267, + "step": 66269 + }, + { + "epoch": 4.9, + "learning_rate": 1.720957117395996e-06, + "loss": 1.0138, + "step": 66270 + }, + { + "epoch": 4.9, + "learning_rate": 1.7207334090937789e-06, + "loss": 1.003, + "step": 66271 + }, + { + "epoch": 4.9, + "learning_rate": 1.720509713963756e-06, + "loss": 0.9646, + "step": 66272 + }, + { + "epoch": 4.9, + "learning_rate": 1.7202860320062842e-06, + "loss": 0.973, + "step": 66273 + }, + { + "epoch": 4.9, + "learning_rate": 1.7200623632217183e-06, + "loss": 0.9892, + "step": 66274 + }, + { + "epoch": 4.9, + "learning_rate": 1.7198387076104128e-06, + "loss": 1.0171, + "step": 66275 + }, + { + "epoch": 4.9, + "learning_rate": 1.7196150651727273e-06, + "loss": 0.9705, + "step": 66276 + }, + { + "epoch": 4.9, + "learning_rate": 1.7193914359090169e-06, + "loss": 0.9645, + "step": 66277 + }, + { + "epoch": 4.9, + "learning_rate": 1.7191678198196349e-06, + "loss": 1.0469, + "step": 66278 + }, + { + "epoch": 4.9, + "learning_rate": 1.7189442169049364e-06, + "loss": 0.9716, + "step": 66279 + }, + { + "epoch": 4.9, + "learning_rate": 1.7187206271652823e-06, + "loss": 1.0621, + "step": 66280 + }, + { + "epoch": 4.9, + "learning_rate": 1.7184970506010245e-06, + "loss": 0.988, + "step": 66281 + }, + { + "epoch": 4.9, + "learning_rate": 1.7182734872125195e-06, + "loss": 0.969, + "step": 66282 + }, + { + "epoch": 4.9, + "learning_rate": 1.7180499370001236e-06, + "loss": 1.1174, + "step": 66283 + }, + { + "epoch": 4.9, + "learning_rate": 1.7178263999641887e-06, + "loss": 1.024, + "step": 66284 + }, + { + "epoch": 4.9, + "learning_rate": 1.7176028761050756e-06, + "loss": 1.0658, + "step": 66285 + }, + { + "epoch": 4.9, + "learning_rate": 1.7173793654231386e-06, + "loss": 0.9555, + "step": 66286 + }, + { + "epoch": 4.9, + "learning_rate": 1.7171558679187317e-06, + "loss": 1.0472, + "step": 66287 + }, + { + "epoch": 4.9, + "learning_rate": 1.716932383592209e-06, + "loss": 0.9981, + "step": 66288 + }, + { + "epoch": 4.9, + "learning_rate": 1.7167089124439308e-06, + "loss": 0.9008, + "step": 66289 + }, + { + "epoch": 4.9, + "learning_rate": 1.7164854544742537e-06, + "loss": 0.9503, + "step": 66290 + }, + { + "epoch": 4.9, + "learning_rate": 1.7162620096835235e-06, + "loss": 1.0235, + "step": 66291 + }, + { + "epoch": 4.9, + "learning_rate": 1.7160385780721045e-06, + "loss": 0.8784, + "step": 66292 + }, + { + "epoch": 4.9, + "learning_rate": 1.7158151596403482e-06, + "loss": 1.01, + "step": 66293 + }, + { + "epoch": 4.9, + "learning_rate": 1.7155917543886158e-06, + "loss": 1.1294, + "step": 66294 + }, + { + "epoch": 4.9, + "learning_rate": 1.7153683623172534e-06, + "loss": 0.9741, + "step": 66295 + }, + { + "epoch": 4.9, + "learning_rate": 1.715144983426623e-06, + "loss": 1.0885, + "step": 66296 + }, + { + "epoch": 4.9, + "learning_rate": 1.714921617717077e-06, + "loss": 0.8877, + "step": 66297 + }, + { + "epoch": 4.9, + "learning_rate": 1.7146982651889777e-06, + "loss": 1.0349, + "step": 66298 + }, + { + "epoch": 4.9, + "learning_rate": 1.7144749258426708e-06, + "loss": 0.9395, + "step": 66299 + }, + { + "epoch": 4.9, + "learning_rate": 1.7142515996785148e-06, + "loss": 1.1023, + "step": 66300 + }, + { + "epoch": 4.9, + "learning_rate": 1.714028286696867e-06, + "loss": 1.101, + "step": 66301 + }, + { + "epoch": 4.9, + "learning_rate": 1.7138049868980833e-06, + "loss": 0.9919, + "step": 66302 + }, + { + "epoch": 4.9, + "learning_rate": 1.713581700282516e-06, + "loss": 1.1075, + "step": 66303 + }, + { + "epoch": 4.9, + "learning_rate": 1.71335842685052e-06, + "loss": 1.0854, + "step": 66304 + }, + { + "epoch": 4.9, + "learning_rate": 1.7131351666024543e-06, + "loss": 0.948, + "step": 66305 + }, + { + "epoch": 4.9, + "learning_rate": 1.7129119195386711e-06, + "loss": 1.1263, + "step": 66306 + }, + { + "epoch": 4.9, + "learning_rate": 1.712688685659527e-06, + "loss": 0.9668, + "step": 66307 + }, + { + "epoch": 4.9, + "learning_rate": 1.7124654649653772e-06, + "loss": 0.9514, + "step": 66308 + }, + { + "epoch": 4.9, + "learning_rate": 1.7122422574565734e-06, + "loss": 0.9859, + "step": 66309 + }, + { + "epoch": 4.9, + "learning_rate": 1.7120190631334766e-06, + "loss": 0.9631, + "step": 66310 + }, + { + "epoch": 4.9, + "learning_rate": 1.7117958819964386e-06, + "loss": 1.1094, + "step": 66311 + }, + { + "epoch": 4.9, + "learning_rate": 1.711572714045815e-06, + "loss": 1.0628, + "step": 66312 + }, + { + "epoch": 4.9, + "learning_rate": 1.7113495592819573e-06, + "loss": 1.0301, + "step": 66313 + }, + { + "epoch": 4.9, + "learning_rate": 1.7111264177052289e-06, + "loss": 1.0333, + "step": 66314 + }, + { + "epoch": 4.9, + "learning_rate": 1.710903289315975e-06, + "loss": 1.0312, + "step": 66315 + }, + { + "epoch": 4.9, + "learning_rate": 1.7106801741145585e-06, + "loss": 0.8857, + "step": 66316 + }, + { + "epoch": 4.9, + "learning_rate": 1.7104570721013302e-06, + "loss": 0.9913, + "step": 66317 + }, + { + "epoch": 4.9, + "learning_rate": 1.7102339832766468e-06, + "loss": 1.0097, + "step": 66318 + }, + { + "epoch": 4.9, + "learning_rate": 1.7100109076408632e-06, + "loss": 1.0608, + "step": 66319 + }, + { + "epoch": 4.9, + "learning_rate": 1.7097878451943305e-06, + "loss": 0.9733, + "step": 66320 + }, + { + "epoch": 4.9, + "learning_rate": 1.7095647959374084e-06, + "loss": 0.9648, + "step": 66321 + }, + { + "epoch": 4.9, + "learning_rate": 1.709341759870451e-06, + "loss": 1.003, + "step": 66322 + }, + { + "epoch": 4.9, + "learning_rate": 1.7091187369938122e-06, + "loss": 1.0934, + "step": 66323 + }, + { + "epoch": 4.9, + "learning_rate": 1.7088957273078443e-06, + "loss": 0.9363, + "step": 66324 + }, + { + "epoch": 4.9, + "learning_rate": 1.708672730812907e-06, + "loss": 1.0193, + "step": 66325 + }, + { + "epoch": 4.9, + "learning_rate": 1.708449747509353e-06, + "loss": 1.0093, + "step": 66326 + }, + { + "epoch": 4.9, + "learning_rate": 1.708226777397536e-06, + "loss": 1.0775, + "step": 66327 + }, + { + "epoch": 4.9, + "learning_rate": 1.7080038204778126e-06, + "loss": 1.0764, + "step": 66328 + }, + { + "epoch": 4.9, + "learning_rate": 1.7077808767505333e-06, + "loss": 0.9443, + "step": 66329 + }, + { + "epoch": 4.9, + "learning_rate": 1.7075579462160585e-06, + "loss": 0.9966, + "step": 66330 + }, + { + "epoch": 4.9, + "learning_rate": 1.7073350288747415e-06, + "loss": 0.9988, + "step": 66331 + }, + { + "epoch": 4.9, + "learning_rate": 1.707112124726934e-06, + "loss": 0.9942, + "step": 66332 + }, + { + "epoch": 4.9, + "learning_rate": 1.7068892337729914e-06, + "loss": 1.0745, + "step": 66333 + }, + { + "epoch": 4.9, + "learning_rate": 1.7066663560132711e-06, + "loss": 1.0568, + "step": 66334 + }, + { + "epoch": 4.9, + "learning_rate": 1.7064434914481266e-06, + "loss": 0.8727, + "step": 66335 + }, + { + "epoch": 4.9, + "learning_rate": 1.7062206400779113e-06, + "loss": 1.0897, + "step": 66336 + }, + { + "epoch": 4.9, + "learning_rate": 1.70599780190298e-06, + "loss": 0.9226, + "step": 66337 + }, + { + "epoch": 4.9, + "learning_rate": 1.7057749769236864e-06, + "loss": 0.9974, + "step": 66338 + }, + { + "epoch": 4.9, + "learning_rate": 1.7055521651403894e-06, + "loss": 1.0862, + "step": 66339 + }, + { + "epoch": 4.9, + "learning_rate": 1.7053293665534365e-06, + "loss": 0.8553, + "step": 66340 + }, + { + "epoch": 4.9, + "learning_rate": 1.7051065811631883e-06, + "loss": 0.9336, + "step": 66341 + }, + { + "epoch": 4.9, + "learning_rate": 1.7048838089699936e-06, + "loss": 1.0056, + "step": 66342 + }, + { + "epoch": 4.9, + "learning_rate": 1.7046610499742156e-06, + "loss": 0.9361, + "step": 66343 + }, + { + "epoch": 4.9, + "learning_rate": 1.704438304176198e-06, + "loss": 0.9792, + "step": 66344 + }, + { + "epoch": 4.9, + "learning_rate": 1.7042155715763032e-06, + "loss": 0.9806, + "step": 66345 + }, + { + "epoch": 4.9, + "learning_rate": 1.7039928521748827e-06, + "loss": 1.0312, + "step": 66346 + }, + { + "epoch": 4.9, + "learning_rate": 1.70377014597229e-06, + "loss": 1.04, + "step": 66347 + }, + { + "epoch": 4.9, + "learning_rate": 1.7035474529688812e-06, + "loss": 0.9513, + "step": 66348 + }, + { + "epoch": 4.9, + "learning_rate": 1.7033247731650071e-06, + "loss": 1.0548, + "step": 66349 + }, + { + "epoch": 4.9, + "learning_rate": 1.7031021065610264e-06, + "loss": 1.0593, + "step": 66350 + }, + { + "epoch": 4.9, + "learning_rate": 1.7028794531572923e-06, + "loss": 0.9175, + "step": 66351 + }, + { + "epoch": 4.9, + "learning_rate": 1.7026568129541587e-06, + "loss": 0.9708, + "step": 66352 + }, + { + "epoch": 4.9, + "learning_rate": 1.7024341859519755e-06, + "loss": 0.9837, + "step": 66353 + }, + { + "epoch": 4.9, + "learning_rate": 1.7022115721511046e-06, + "loss": 1.0106, + "step": 66354 + }, + { + "epoch": 4.9, + "learning_rate": 1.7019889715518956e-06, + "loss": 0.9396, + "step": 66355 + }, + { + "epoch": 4.9, + "learning_rate": 1.7017663841547038e-06, + "loss": 0.9539, + "step": 66356 + }, + { + "epoch": 4.9, + "learning_rate": 1.7015438099598824e-06, + "loss": 0.8955, + "step": 66357 + }, + { + "epoch": 4.9, + "learning_rate": 1.7013212489677844e-06, + "loss": 0.9909, + "step": 66358 + }, + { + "epoch": 4.9, + "learning_rate": 1.7010987011787694e-06, + "loss": 0.9575, + "step": 66359 + }, + { + "epoch": 4.9, + "learning_rate": 1.700876166593184e-06, + "loss": 1.0599, + "step": 66360 + }, + { + "epoch": 4.9, + "learning_rate": 1.7006536452113876e-06, + "loss": 0.8933, + "step": 66361 + }, + { + "epoch": 4.9, + "learning_rate": 1.7004311370337302e-06, + "loss": 0.9557, + "step": 66362 + }, + { + "epoch": 4.9, + "learning_rate": 1.7002086420605735e-06, + "loss": 0.9396, + "step": 66363 + }, + { + "epoch": 4.9, + "learning_rate": 1.699986160292263e-06, + "loss": 1.038, + "step": 66364 + }, + { + "epoch": 4.9, + "learning_rate": 1.6997636917291528e-06, + "loss": 0.9506, + "step": 66365 + }, + { + "epoch": 4.9, + "learning_rate": 1.6995412363716036e-06, + "loss": 0.9797, + "step": 66366 + }, + { + "epoch": 4.9, + "learning_rate": 1.699318794219964e-06, + "loss": 1.0749, + "step": 66367 + }, + { + "epoch": 4.9, + "learning_rate": 1.6990963652745907e-06, + "loss": 0.9948, + "step": 66368 + }, + { + "epoch": 4.9, + "learning_rate": 1.698873949535833e-06, + "loss": 1.1188, + "step": 66369 + }, + { + "epoch": 4.9, + "learning_rate": 1.698651547004051e-06, + "loss": 1.0036, + "step": 66370 + }, + { + "epoch": 4.9, + "learning_rate": 1.6984291576795965e-06, + "loss": 0.997, + "step": 66371 + }, + { + "epoch": 4.9, + "learning_rate": 1.6982067815628212e-06, + "loss": 0.9543, + "step": 66372 + }, + { + "epoch": 4.9, + "learning_rate": 1.6979844186540806e-06, + "loss": 0.9528, + "step": 66373 + }, + { + "epoch": 4.9, + "learning_rate": 1.6977620689537245e-06, + "loss": 1.0146, + "step": 66374 + }, + { + "epoch": 4.9, + "learning_rate": 1.6975397324621134e-06, + "loss": 1.0353, + "step": 66375 + }, + { + "epoch": 4.9, + "learning_rate": 1.6973174091795985e-06, + "loss": 0.9573, + "step": 66376 + }, + { + "epoch": 4.9, + "learning_rate": 1.6970950991065327e-06, + "loss": 0.9648, + "step": 66377 + }, + { + "epoch": 4.9, + "learning_rate": 1.6968728022432668e-06, + "loss": 0.9665, + "step": 66378 + }, + { + "epoch": 4.9, + "learning_rate": 1.6966505185901595e-06, + "loss": 0.983, + "step": 66379 + }, + { + "epoch": 4.9, + "learning_rate": 1.6964282481475637e-06, + "loss": 0.9433, + "step": 66380 + }, + { + "epoch": 4.9, + "learning_rate": 1.6962059909158301e-06, + "loss": 1.0863, + "step": 66381 + }, + { + "epoch": 4.9, + "learning_rate": 1.6959837468953122e-06, + "loss": 0.9918, + "step": 66382 + }, + { + "epoch": 4.9, + "learning_rate": 1.6957615160863682e-06, + "loss": 0.9664, + "step": 66383 + }, + { + "epoch": 4.91, + "learning_rate": 1.6955392984893514e-06, + "loss": 0.9104, + "step": 66384 + }, + { + "epoch": 4.91, + "learning_rate": 1.6953170941046071e-06, + "loss": 0.9248, + "step": 66385 + }, + { + "epoch": 4.91, + "learning_rate": 1.6950949029324982e-06, + "loss": 1.0021, + "step": 66386 + }, + { + "epoch": 4.91, + "learning_rate": 1.6948727249733709e-06, + "loss": 0.9241, + "step": 66387 + }, + { + "epoch": 4.91, + "learning_rate": 1.6946505602275877e-06, + "loss": 1.091, + "step": 66388 + }, + { + "epoch": 4.91, + "learning_rate": 1.6944284086954909e-06, + "loss": 0.9431, + "step": 66389 + }, + { + "epoch": 4.91, + "learning_rate": 1.694206270377443e-06, + "loss": 0.9472, + "step": 66390 + }, + { + "epoch": 4.91, + "learning_rate": 1.6939841452737936e-06, + "loss": 0.9991, + "step": 66391 + }, + { + "epoch": 4.91, + "learning_rate": 1.693762033384897e-06, + "loss": 1.0048, + "step": 66392 + }, + { + "epoch": 4.91, + "learning_rate": 1.693539934711106e-06, + "loss": 0.9429, + "step": 66393 + }, + { + "epoch": 4.91, + "learning_rate": 1.6933178492527714e-06, + "loss": 1.0436, + "step": 66394 + }, + { + "epoch": 4.91, + "learning_rate": 1.6930957770102518e-06, + "loss": 1.0419, + "step": 66395 + }, + { + "epoch": 4.91, + "learning_rate": 1.6928737179838984e-06, + "loss": 1.0156, + "step": 66396 + }, + { + "epoch": 4.91, + "learning_rate": 1.6926516721740637e-06, + "loss": 1.048, + "step": 66397 + }, + { + "epoch": 4.91, + "learning_rate": 1.692429639581098e-06, + "loss": 1.044, + "step": 66398 + }, + { + "epoch": 4.91, + "learning_rate": 1.6922076202053616e-06, + "loss": 1.0182, + "step": 66399 + }, + { + "epoch": 4.91, + "learning_rate": 1.6919856140472034e-06, + "loss": 1.0614, + "step": 66400 + }, + { + "epoch": 4.91, + "learning_rate": 1.6917636211069776e-06, + "loss": 0.9112, + "step": 66401 + }, + { + "epoch": 4.91, + "learning_rate": 1.6915416413850372e-06, + "loss": 1.1056, + "step": 66402 + }, + { + "epoch": 4.91, + "learning_rate": 1.6913196748817317e-06, + "loss": 1.0489, + "step": 66403 + }, + { + "epoch": 4.91, + "learning_rate": 1.6910977215974234e-06, + "loss": 0.9057, + "step": 66404 + }, + { + "epoch": 4.91, + "learning_rate": 1.690875781532455e-06, + "loss": 0.9805, + "step": 66405 + }, + { + "epoch": 4.91, + "learning_rate": 1.6906538546871865e-06, + "loss": 1.0422, + "step": 66406 + }, + { + "epoch": 4.91, + "learning_rate": 1.6904319410619663e-06, + "loss": 1.074, + "step": 66407 + }, + { + "epoch": 4.91, + "learning_rate": 1.6902100406571554e-06, + "loss": 0.9737, + "step": 66408 + }, + { + "epoch": 4.91, + "learning_rate": 1.6899881534730967e-06, + "loss": 0.9946, + "step": 66409 + }, + { + "epoch": 4.91, + "learning_rate": 1.68976627951015e-06, + "loss": 0.9252, + "step": 66410 + }, + { + "epoch": 4.91, + "learning_rate": 1.689544418768666e-06, + "loss": 0.9524, + "step": 66411 + }, + { + "epoch": 4.91, + "learning_rate": 1.689322571248998e-06, + "loss": 0.9502, + "step": 66412 + }, + { + "epoch": 4.91, + "learning_rate": 1.689100736951499e-06, + "loss": 0.9363, + "step": 66413 + }, + { + "epoch": 4.91, + "learning_rate": 1.6888789158765185e-06, + "loss": 1.0683, + "step": 66414 + }, + { + "epoch": 4.91, + "learning_rate": 1.6886571080244162e-06, + "loss": 1.0184, + "step": 66415 + }, + { + "epoch": 4.91, + "learning_rate": 1.6884353133955412e-06, + "loss": 0.9276, + "step": 66416 + }, + { + "epoch": 4.91, + "learning_rate": 1.688213531990247e-06, + "loss": 1.0168, + "step": 66417 + }, + { + "epoch": 4.91, + "learning_rate": 1.687991763808884e-06, + "loss": 1.0957, + "step": 66418 + }, + { + "epoch": 4.91, + "learning_rate": 1.687770008851809e-06, + "loss": 0.9591, + "step": 66419 + }, + { + "epoch": 4.91, + "learning_rate": 1.6875482671193721e-06, + "loss": 0.9836, + "step": 66420 + }, + { + "epoch": 4.91, + "learning_rate": 1.6873265386119286e-06, + "loss": 0.8647, + "step": 66421 + }, + { + "epoch": 4.91, + "learning_rate": 1.6871048233298293e-06, + "loss": 1.0333, + "step": 66422 + }, + { + "epoch": 4.91, + "learning_rate": 1.686883121273424e-06, + "loss": 1.1286, + "step": 66423 + }, + { + "epoch": 4.91, + "learning_rate": 1.6866614324430719e-06, + "loss": 0.8911, + "step": 66424 + }, + { + "epoch": 4.91, + "learning_rate": 1.6864397568391222e-06, + "loss": 1.0617, + "step": 66425 + }, + { + "epoch": 4.91, + "learning_rate": 1.6862180944619278e-06, + "loss": 1.1026, + "step": 66426 + }, + { + "epoch": 4.91, + "learning_rate": 1.6859964453118394e-06, + "loss": 1.0258, + "step": 66427 + }, + { + "epoch": 4.91, + "learning_rate": 1.6857748093892146e-06, + "loss": 1.0204, + "step": 66428 + }, + { + "epoch": 4.91, + "learning_rate": 1.6855531866944053e-06, + "loss": 0.9214, + "step": 66429 + }, + { + "epoch": 4.91, + "learning_rate": 1.6853315772277556e-06, + "loss": 0.9814, + "step": 66430 + }, + { + "epoch": 4.91, + "learning_rate": 1.6851099809896288e-06, + "loss": 0.9194, + "step": 66431 + }, + { + "epoch": 4.91, + "learning_rate": 1.6848883979803698e-06, + "loss": 1.0773, + "step": 66432 + }, + { + "epoch": 4.91, + "learning_rate": 1.6846668282003386e-06, + "loss": 0.9977, + "step": 66433 + }, + { + "epoch": 4.91, + "learning_rate": 1.6844452716498794e-06, + "loss": 0.9897, + "step": 66434 + }, + { + "epoch": 4.91, + "learning_rate": 1.6842237283293516e-06, + "loss": 0.8827, + "step": 66435 + }, + { + "epoch": 4.91, + "learning_rate": 1.684002198239104e-06, + "loss": 0.9641, + "step": 66436 + }, + { + "epoch": 4.91, + "learning_rate": 1.6837806813794898e-06, + "loss": 0.9309, + "step": 66437 + }, + { + "epoch": 4.91, + "learning_rate": 1.6835591777508597e-06, + "loss": 1.0326, + "step": 66438 + }, + { + "epoch": 4.91, + "learning_rate": 1.68333768735357e-06, + "loss": 0.9976, + "step": 66439 + }, + { + "epoch": 4.91, + "learning_rate": 1.6831162101879716e-06, + "loss": 0.9964, + "step": 66440 + }, + { + "epoch": 4.91, + "learning_rate": 1.6828947462544165e-06, + "loss": 0.9911, + "step": 66441 + }, + { + "epoch": 4.91, + "learning_rate": 1.6826732955532565e-06, + "loss": 0.8917, + "step": 66442 + }, + { + "epoch": 4.91, + "learning_rate": 1.6824518580848414e-06, + "loss": 1.0922, + "step": 66443 + }, + { + "epoch": 4.91, + "learning_rate": 1.6822304338495289e-06, + "loss": 0.9896, + "step": 66444 + }, + { + "epoch": 4.91, + "learning_rate": 1.6820090228476693e-06, + "loss": 1.0032, + "step": 66445 + }, + { + "epoch": 4.91, + "learning_rate": 1.681787625079614e-06, + "loss": 0.813, + "step": 66446 + }, + { + "epoch": 4.91, + "learning_rate": 1.6815662405457133e-06, + "loss": 1.0269, + "step": 66447 + }, + { + "epoch": 4.91, + "learning_rate": 1.681344869246324e-06, + "loss": 1.01, + "step": 66448 + }, + { + "epoch": 4.91, + "learning_rate": 1.6811235111817992e-06, + "loss": 1.0566, + "step": 66449 + }, + { + "epoch": 4.91, + "learning_rate": 1.6809021663524816e-06, + "loss": 1.0715, + "step": 66450 + }, + { + "epoch": 4.91, + "learning_rate": 1.6806808347587334e-06, + "loss": 0.8503, + "step": 66451 + }, + { + "epoch": 4.91, + "learning_rate": 1.6804595164008996e-06, + "loss": 0.9373, + "step": 66452 + }, + { + "epoch": 4.91, + "learning_rate": 1.6802382112793403e-06, + "loss": 0.9667, + "step": 66453 + }, + { + "epoch": 4.91, + "learning_rate": 1.6800169193943994e-06, + "loss": 1.09, + "step": 66454 + }, + { + "epoch": 4.91, + "learning_rate": 1.6797956407464332e-06, + "loss": 0.8977, + "step": 66455 + }, + { + "epoch": 4.91, + "learning_rate": 1.679574375335792e-06, + "loss": 0.9886, + "step": 66456 + }, + { + "epoch": 4.91, + "learning_rate": 1.6793531231628335e-06, + "loss": 1.0844, + "step": 66457 + }, + { + "epoch": 4.91, + "learning_rate": 1.6791318842279025e-06, + "loss": 1.0549, + "step": 66458 + }, + { + "epoch": 4.91, + "learning_rate": 1.6789106585313519e-06, + "loss": 1.0293, + "step": 66459 + }, + { + "epoch": 4.91, + "learning_rate": 1.678689446073537e-06, + "loss": 0.9357, + "step": 66460 + }, + { + "epoch": 4.91, + "learning_rate": 1.6784682468548086e-06, + "loss": 1.0037, + "step": 66461 + }, + { + "epoch": 4.91, + "learning_rate": 1.6782470608755175e-06, + "loss": 0.9734, + "step": 66462 + }, + { + "epoch": 4.91, + "learning_rate": 1.6780258881360146e-06, + "loss": 0.7774, + "step": 66463 + }, + { + "epoch": 4.91, + "learning_rate": 1.677804728636655e-06, + "loss": 1.0085, + "step": 66464 + }, + { + "epoch": 4.91, + "learning_rate": 1.6775835823777898e-06, + "loss": 0.9681, + "step": 66465 + }, + { + "epoch": 4.91, + "learning_rate": 1.6773624493597695e-06, + "loss": 1.0132, + "step": 66466 + }, + { + "epoch": 4.91, + "learning_rate": 1.6771413295829474e-06, + "loss": 1.0955, + "step": 66467 + }, + { + "epoch": 4.91, + "learning_rate": 1.6769202230476712e-06, + "loss": 0.9751, + "step": 66468 + }, + { + "epoch": 4.91, + "learning_rate": 1.6766991297542978e-06, + "loss": 1.0096, + "step": 66469 + }, + { + "epoch": 4.91, + "learning_rate": 1.6764780497031784e-06, + "loss": 0.9571, + "step": 66470 + }, + { + "epoch": 4.91, + "learning_rate": 1.6762569828946618e-06, + "loss": 1.0078, + "step": 66471 + }, + { + "epoch": 4.91, + "learning_rate": 1.6760359293290995e-06, + "loss": 1.0464, + "step": 66472 + }, + { + "epoch": 4.91, + "learning_rate": 1.6758148890068483e-06, + "loss": 0.9585, + "step": 66473 + }, + { + "epoch": 4.91, + "learning_rate": 1.6755938619282553e-06, + "loss": 0.9914, + "step": 66474 + }, + { + "epoch": 4.91, + "learning_rate": 1.6753728480936736e-06, + "loss": 0.9666, + "step": 66475 + }, + { + "epoch": 4.91, + "learning_rate": 1.6751518475034546e-06, + "loss": 1.0615, + "step": 66476 + }, + { + "epoch": 4.91, + "learning_rate": 1.6749308601579473e-06, + "loss": 0.9189, + "step": 66477 + }, + { + "epoch": 4.91, + "learning_rate": 1.6747098860575107e-06, + "loss": 0.9331, + "step": 66478 + }, + { + "epoch": 4.91, + "learning_rate": 1.6744889252024876e-06, + "loss": 1.1003, + "step": 66479 + }, + { + "epoch": 4.91, + "learning_rate": 1.6742679775932348e-06, + "loss": 1.0891, + "step": 66480 + }, + { + "epoch": 4.91, + "learning_rate": 1.6740470432301038e-06, + "loss": 0.9354, + "step": 66481 + }, + { + "epoch": 4.91, + "learning_rate": 1.6738261221134434e-06, + "loss": 0.9478, + "step": 66482 + }, + { + "epoch": 4.91, + "learning_rate": 1.6736052142436054e-06, + "loss": 1.0047, + "step": 66483 + }, + { + "epoch": 4.91, + "learning_rate": 1.6733843196209432e-06, + "loss": 0.9474, + "step": 66484 + }, + { + "epoch": 4.91, + "learning_rate": 1.6731634382458084e-06, + "loss": 0.9461, + "step": 66485 + }, + { + "epoch": 4.91, + "learning_rate": 1.672942570118552e-06, + "loss": 1.0882, + "step": 66486 + }, + { + "epoch": 4.91, + "learning_rate": 1.6727217152395238e-06, + "loss": 1.0716, + "step": 66487 + }, + { + "epoch": 4.91, + "learning_rate": 1.6725008736090743e-06, + "loss": 0.9546, + "step": 66488 + }, + { + "epoch": 4.91, + "learning_rate": 1.672280045227559e-06, + "loss": 1.0248, + "step": 66489 + }, + { + "epoch": 4.91, + "learning_rate": 1.6720592300953276e-06, + "loss": 1.0511, + "step": 66490 + }, + { + "epoch": 4.91, + "learning_rate": 1.6718384282127299e-06, + "loss": 1.0284, + "step": 66491 + }, + { + "epoch": 4.91, + "learning_rate": 1.6716176395801165e-06, + "loss": 1.0006, + "step": 66492 + }, + { + "epoch": 4.91, + "learning_rate": 1.6713968641978418e-06, + "loss": 1.1698, + "step": 66493 + }, + { + "epoch": 4.91, + "learning_rate": 1.6711761020662565e-06, + "loss": 1.0179, + "step": 66494 + }, + { + "epoch": 4.91, + "learning_rate": 1.6709553531857104e-06, + "loss": 1.047, + "step": 66495 + }, + { + "epoch": 4.91, + "learning_rate": 1.6707346175565553e-06, + "loss": 0.9574, + "step": 66496 + }, + { + "epoch": 4.91, + "learning_rate": 1.6705138951791388e-06, + "loss": 0.9338, + "step": 66497 + }, + { + "epoch": 4.91, + "learning_rate": 1.670293186053822e-06, + "loss": 0.8973, + "step": 66498 + }, + { + "epoch": 4.91, + "learning_rate": 1.6700724901809439e-06, + "loss": 1.1002, + "step": 66499 + }, + { + "epoch": 4.91, + "learning_rate": 1.6698518075608638e-06, + "loss": 1.081, + "step": 66500 + }, + { + "epoch": 4.91, + "learning_rate": 1.6696311381939267e-06, + "loss": 0.9618, + "step": 66501 + }, + { + "epoch": 4.91, + "learning_rate": 1.6694104820804935e-06, + "loss": 0.9968, + "step": 66502 + }, + { + "epoch": 4.91, + "learning_rate": 1.6691898392209038e-06, + "loss": 0.899, + "step": 66503 + }, + { + "epoch": 4.91, + "learning_rate": 1.6689692096155164e-06, + "loss": 0.9182, + "step": 66504 + }, + { + "epoch": 4.91, + "learning_rate": 1.6687485932646808e-06, + "loss": 1.0575, + "step": 66505 + }, + { + "epoch": 4.91, + "learning_rate": 1.6685279901687456e-06, + "loss": 1.0437, + "step": 66506 + }, + { + "epoch": 4.91, + "learning_rate": 1.668307400328063e-06, + "loss": 1.0368, + "step": 66507 + }, + { + "epoch": 4.91, + "learning_rate": 1.6680868237429825e-06, + "loss": 1.0199, + "step": 66508 + }, + { + "epoch": 4.91, + "learning_rate": 1.6678662604138586e-06, + "loss": 0.8627, + "step": 66509 + }, + { + "epoch": 4.91, + "learning_rate": 1.6676457103410404e-06, + "loss": 1.0416, + "step": 66510 + }, + { + "epoch": 4.91, + "learning_rate": 1.6674251735248793e-06, + "loss": 0.966, + "step": 66511 + }, + { + "epoch": 4.91, + "learning_rate": 1.6672046499657223e-06, + "loss": 0.9211, + "step": 66512 + }, + { + "epoch": 4.91, + "learning_rate": 1.6669841396639263e-06, + "loss": 1.0303, + "step": 66513 + }, + { + "epoch": 4.91, + "learning_rate": 1.6667636426198398e-06, + "loss": 0.9674, + "step": 66514 + }, + { + "epoch": 4.91, + "learning_rate": 1.6665431588338122e-06, + "loss": 1.0117, + "step": 66515 + }, + { + "epoch": 4.91, + "learning_rate": 1.6663226883061956e-06, + "loss": 1.0241, + "step": 66516 + }, + { + "epoch": 4.91, + "learning_rate": 1.6661022310373376e-06, + "loss": 0.9293, + "step": 66517 + }, + { + "epoch": 4.91, + "learning_rate": 1.6658817870275945e-06, + "loss": 1.0299, + "step": 66518 + }, + { + "epoch": 4.92, + "learning_rate": 1.665661356277314e-06, + "loss": 0.9022, + "step": 66519 + }, + { + "epoch": 4.92, + "learning_rate": 1.6654409387868464e-06, + "loss": 0.9979, + "step": 66520 + }, + { + "epoch": 4.92, + "learning_rate": 1.665220534556542e-06, + "loss": 1.045, + "step": 66521 + }, + { + "epoch": 4.92, + "learning_rate": 1.6650001435867568e-06, + "loss": 0.9602, + "step": 66522 + }, + { + "epoch": 4.92, + "learning_rate": 1.6647797658778341e-06, + "loss": 0.9786, + "step": 66523 + }, + { + "epoch": 4.92, + "learning_rate": 1.6645594014301258e-06, + "loss": 0.9814, + "step": 66524 + }, + { + "epoch": 4.92, + "learning_rate": 1.664339050243986e-06, + "loss": 0.9062, + "step": 66525 + }, + { + "epoch": 4.92, + "learning_rate": 1.6641187123197645e-06, + "loss": 0.9888, + "step": 66526 + }, + { + "epoch": 4.92, + "learning_rate": 1.6638983876578096e-06, + "loss": 1.0083, + "step": 66527 + }, + { + "epoch": 4.92, + "learning_rate": 1.6636780762584714e-06, + "loss": 1.0315, + "step": 66528 + }, + { + "epoch": 4.92, + "learning_rate": 1.663457778122105e-06, + "loss": 0.9495, + "step": 66529 + }, + { + "epoch": 4.92, + "learning_rate": 1.6632374932490569e-06, + "loss": 1.0461, + "step": 66530 + }, + { + "epoch": 4.92, + "learning_rate": 1.66301722163968e-06, + "loss": 0.9746, + "step": 66531 + }, + { + "epoch": 4.92, + "learning_rate": 1.662796963294322e-06, + "loss": 0.9815, + "step": 66532 + }, + { + "epoch": 4.92, + "learning_rate": 1.6625767182133335e-06, + "loss": 0.9852, + "step": 66533 + }, + { + "epoch": 4.92, + "learning_rate": 1.6623564863970675e-06, + "loss": 0.9812, + "step": 66534 + }, + { + "epoch": 4.92, + "learning_rate": 1.662136267845874e-06, + "loss": 1.0868, + "step": 66535 + }, + { + "epoch": 4.92, + "learning_rate": 1.6619160625601015e-06, + "loss": 0.929, + "step": 66536 + }, + { + "epoch": 4.92, + "learning_rate": 1.6616958705401e-06, + "loss": 1.0059, + "step": 66537 + }, + { + "epoch": 4.92, + "learning_rate": 1.6614756917862229e-06, + "loss": 0.9473, + "step": 66538 + }, + { + "epoch": 4.92, + "learning_rate": 1.6612555262988183e-06, + "loss": 1.0031, + "step": 66539 + }, + { + "epoch": 4.92, + "learning_rate": 1.661035374078237e-06, + "loss": 1.0342, + "step": 66540 + }, + { + "epoch": 4.92, + "learning_rate": 1.6608152351248274e-06, + "loss": 1.0756, + "step": 66541 + }, + { + "epoch": 4.92, + "learning_rate": 1.6605951094389428e-06, + "loss": 1.0652, + "step": 66542 + }, + { + "epoch": 4.92, + "learning_rate": 1.660374997020935e-06, + "loss": 0.9587, + "step": 66543 + }, + { + "epoch": 4.92, + "learning_rate": 1.6601548978711469e-06, + "loss": 0.9801, + "step": 66544 + }, + { + "epoch": 4.92, + "learning_rate": 1.659934811989935e-06, + "loss": 0.969, + "step": 66545 + }, + { + "epoch": 4.92, + "learning_rate": 1.6597147393776447e-06, + "loss": 0.9437, + "step": 66546 + }, + { + "epoch": 4.92, + "learning_rate": 1.6594946800346334e-06, + "loss": 1.0273, + "step": 66547 + }, + { + "epoch": 4.92, + "learning_rate": 1.6592746339612432e-06, + "loss": 0.9126, + "step": 66548 + }, + { + "epoch": 4.92, + "learning_rate": 1.659054601157829e-06, + "loss": 1.0356, + "step": 66549 + }, + { + "epoch": 4.92, + "learning_rate": 1.6588345816247375e-06, + "loss": 0.9551, + "step": 66550 + }, + { + "epoch": 4.92, + "learning_rate": 1.658614575362326e-06, + "loss": 0.9409, + "step": 66551 + }, + { + "epoch": 4.92, + "learning_rate": 1.6583945823709368e-06, + "loss": 0.9746, + "step": 66552 + }, + { + "epoch": 4.92, + "learning_rate": 1.6581746026509205e-06, + "loss": 1.0427, + "step": 66553 + }, + { + "epoch": 4.92, + "learning_rate": 1.657954636202631e-06, + "loss": 0.9164, + "step": 66554 + }, + { + "epoch": 4.92, + "learning_rate": 1.6577346830264164e-06, + "loss": 1.1142, + "step": 66555 + }, + { + "epoch": 4.92, + "learning_rate": 1.6575147431226269e-06, + "loss": 0.8864, + "step": 66556 + }, + { + "epoch": 4.92, + "learning_rate": 1.657294816491609e-06, + "loss": 1.0712, + "step": 66557 + }, + { + "epoch": 4.92, + "learning_rate": 1.6570749031337184e-06, + "loss": 1.0247, + "step": 66558 + }, + { + "epoch": 4.92, + "learning_rate": 1.6568550030493025e-06, + "loss": 0.9984, + "step": 66559 + }, + { + "epoch": 4.92, + "learning_rate": 1.6566351162387107e-06, + "loss": 1.0683, + "step": 66560 + }, + { + "epoch": 4.92, + "learning_rate": 1.6564152427022928e-06, + "loss": 0.999, + "step": 66561 + }, + { + "epoch": 4.92, + "learning_rate": 1.6561953824403965e-06, + "loss": 0.9583, + "step": 66562 + }, + { + "epoch": 4.92, + "learning_rate": 1.655975535453377e-06, + "loss": 0.9383, + "step": 66563 + }, + { + "epoch": 4.92, + "learning_rate": 1.6557557017415803e-06, + "loss": 0.8954, + "step": 66564 + }, + { + "epoch": 4.92, + "learning_rate": 1.6555358813053568e-06, + "loss": 0.9998, + "step": 66565 + }, + { + "epoch": 4.92, + "learning_rate": 1.6553160741450547e-06, + "loss": 0.9513, + "step": 66566 + }, + { + "epoch": 4.92, + "learning_rate": 1.6550962802610294e-06, + "loss": 0.9483, + "step": 66567 + }, + { + "epoch": 4.92, + "learning_rate": 1.6548764996536227e-06, + "loss": 0.9095, + "step": 66568 + }, + { + "epoch": 4.92, + "learning_rate": 1.6546567323231888e-06, + "loss": 0.9581, + "step": 66569 + }, + { + "epoch": 4.92, + "learning_rate": 1.6544369782700787e-06, + "loss": 1.0664, + "step": 66570 + }, + { + "epoch": 4.92, + "learning_rate": 1.6542172374946386e-06, + "loss": 1.0643, + "step": 66571 + }, + { + "epoch": 4.92, + "learning_rate": 1.6539975099972194e-06, + "loss": 1.1006, + "step": 66572 + }, + { + "epoch": 4.92, + "learning_rate": 1.6537777957781687e-06, + "loss": 0.9965, + "step": 66573 + }, + { + "epoch": 4.92, + "learning_rate": 1.6535580948378405e-06, + "loss": 0.9538, + "step": 66574 + }, + { + "epoch": 4.92, + "learning_rate": 1.6533384071765824e-06, + "loss": 1.0418, + "step": 66575 + }, + { + "epoch": 4.92, + "learning_rate": 1.653118732794743e-06, + "loss": 0.9086, + "step": 66576 + }, + { + "epoch": 4.92, + "learning_rate": 1.6528990716926697e-06, + "loss": 0.9945, + "step": 66577 + }, + { + "epoch": 4.92, + "learning_rate": 1.6526794238707178e-06, + "loss": 0.9279, + "step": 66578 + }, + { + "epoch": 4.92, + "learning_rate": 1.6524597893292326e-06, + "loss": 1.0502, + "step": 66579 + }, + { + "epoch": 4.92, + "learning_rate": 1.652240168068565e-06, + "loss": 0.9977, + "step": 66580 + }, + { + "epoch": 4.92, + "learning_rate": 1.6520205600890637e-06, + "loss": 0.9206, + "step": 66581 + }, + { + "epoch": 4.92, + "learning_rate": 1.651800965391076e-06, + "loss": 0.9572, + "step": 66582 + }, + { + "epoch": 4.92, + "learning_rate": 1.651581383974955e-06, + "loss": 1.0411, + "step": 66583 + }, + { + "epoch": 4.92, + "learning_rate": 1.6513618158410494e-06, + "loss": 1.023, + "step": 66584 + }, + { + "epoch": 4.92, + "learning_rate": 1.6511422609897075e-06, + "loss": 1.0155, + "step": 66585 + }, + { + "epoch": 4.92, + "learning_rate": 1.6509227194212762e-06, + "loss": 1.1288, + "step": 66586 + }, + { + "epoch": 4.92, + "learning_rate": 1.6507031911361105e-06, + "loss": 1.1405, + "step": 66587 + }, + { + "epoch": 4.92, + "learning_rate": 1.6504836761345588e-06, + "loss": 0.9592, + "step": 66588 + }, + { + "epoch": 4.92, + "learning_rate": 1.6502641744169634e-06, + "loss": 1.0042, + "step": 66589 + }, + { + "epoch": 4.92, + "learning_rate": 1.6500446859836793e-06, + "loss": 1.0386, + "step": 66590 + }, + { + "epoch": 4.92, + "learning_rate": 1.6498252108350532e-06, + "loss": 1.0779, + "step": 66591 + }, + { + "epoch": 4.92, + "learning_rate": 1.6496057489714413e-06, + "loss": 1.0751, + "step": 66592 + }, + { + "epoch": 4.92, + "learning_rate": 1.649386300393181e-06, + "loss": 0.9614, + "step": 66593 + }, + { + "epoch": 4.92, + "learning_rate": 1.649166865100631e-06, + "loss": 0.9831, + "step": 66594 + }, + { + "epoch": 4.92, + "learning_rate": 1.6489474430941344e-06, + "loss": 1.0576, + "step": 66595 + }, + { + "epoch": 4.92, + "learning_rate": 1.6487280343740475e-06, + "loss": 0.9524, + "step": 66596 + }, + { + "epoch": 4.92, + "learning_rate": 1.64850863894071e-06, + "loss": 0.96, + "step": 66597 + }, + { + "epoch": 4.92, + "learning_rate": 1.6482892567944785e-06, + "loss": 0.8897, + "step": 66598 + }, + { + "epoch": 4.92, + "learning_rate": 1.6480698879356994e-06, + "loss": 1.0246, + "step": 66599 + }, + { + "epoch": 4.92, + "learning_rate": 1.647850532364721e-06, + "loss": 1.0313, + "step": 66600 + }, + { + "epoch": 4.92, + "learning_rate": 1.647631190081893e-06, + "loss": 0.9761, + "step": 66601 + }, + { + "epoch": 4.92, + "learning_rate": 1.6474118610875623e-06, + "loss": 0.9481, + "step": 66602 + }, + { + "epoch": 4.92, + "learning_rate": 1.6471925453820814e-06, + "loss": 0.9647, + "step": 66603 + }, + { + "epoch": 4.92, + "learning_rate": 1.6469732429657992e-06, + "loss": 0.8365, + "step": 66604 + }, + { + "epoch": 4.92, + "learning_rate": 1.646753953839062e-06, + "loss": 1.0408, + "step": 66605 + }, + { + "epoch": 4.92, + "learning_rate": 1.6465346780022174e-06, + "loss": 1.0471, + "step": 66606 + }, + { + "epoch": 4.92, + "learning_rate": 1.6463154154556183e-06, + "loss": 1.0165, + "step": 66607 + }, + { + "epoch": 4.92, + "learning_rate": 1.6460961661996133e-06, + "loss": 0.9256, + "step": 66608 + }, + { + "epoch": 4.92, + "learning_rate": 1.6458769302345489e-06, + "loss": 0.9981, + "step": 66609 + }, + { + "epoch": 4.92, + "learning_rate": 1.645657707560775e-06, + "loss": 1.0224, + "step": 66610 + }, + { + "epoch": 4.92, + "learning_rate": 1.6454384981786376e-06, + "loss": 0.945, + "step": 66611 + }, + { + "epoch": 4.92, + "learning_rate": 1.6452193020884932e-06, + "loss": 0.9716, + "step": 66612 + }, + { + "epoch": 4.92, + "learning_rate": 1.6450001192906795e-06, + "loss": 1.0346, + "step": 66613 + }, + { + "epoch": 4.92, + "learning_rate": 1.6447809497855548e-06, + "loss": 1.1173, + "step": 66614 + }, + { + "epoch": 4.92, + "learning_rate": 1.6445617935734615e-06, + "loss": 0.9901, + "step": 66615 + }, + { + "epoch": 4.92, + "learning_rate": 1.6443426506547543e-06, + "loss": 0.9018, + "step": 66616 + }, + { + "epoch": 4.92, + "learning_rate": 1.6441235210297778e-06, + "loss": 1.1082, + "step": 66617 + }, + { + "epoch": 4.92, + "learning_rate": 1.6439044046988773e-06, + "loss": 1.0625, + "step": 66618 + }, + { + "epoch": 4.92, + "learning_rate": 1.6436853016624087e-06, + "loss": 1.045, + "step": 66619 + }, + { + "epoch": 4.92, + "learning_rate": 1.6434662119207167e-06, + "loss": 0.9276, + "step": 66620 + }, + { + "epoch": 4.92, + "learning_rate": 1.6432471354741508e-06, + "loss": 0.9879, + "step": 66621 + }, + { + "epoch": 4.92, + "learning_rate": 1.6430280723230562e-06, + "loss": 0.8904, + "step": 66622 + }, + { + "epoch": 4.92, + "learning_rate": 1.6428090224677874e-06, + "loss": 0.9249, + "step": 66623 + }, + { + "epoch": 4.92, + "learning_rate": 1.6425899859086903e-06, + "loss": 0.9603, + "step": 66624 + }, + { + "epoch": 4.92, + "learning_rate": 1.6423709626461115e-06, + "loss": 1.0489, + "step": 66625 + }, + { + "epoch": 4.92, + "learning_rate": 1.642151952680402e-06, + "loss": 1.0601, + "step": 66626 + }, + { + "epoch": 4.92, + "learning_rate": 1.6419329560119068e-06, + "loss": 0.9898, + "step": 66627 + }, + { + "epoch": 4.92, + "learning_rate": 1.6417139726409782e-06, + "loss": 0.9222, + "step": 66628 + }, + { + "epoch": 4.92, + "learning_rate": 1.6414950025679643e-06, + "loss": 1.029, + "step": 66629 + }, + { + "epoch": 4.92, + "learning_rate": 1.6412760457932109e-06, + "loss": 1.0846, + "step": 66630 + }, + { + "epoch": 4.92, + "learning_rate": 1.6410571023170663e-06, + "loss": 1.0024, + "step": 66631 + }, + { + "epoch": 4.92, + "learning_rate": 1.6408381721398824e-06, + "loss": 1.0655, + "step": 66632 + }, + { + "epoch": 4.92, + "learning_rate": 1.6406192552620047e-06, + "loss": 0.9201, + "step": 66633 + }, + { + "epoch": 4.92, + "learning_rate": 1.640400351683783e-06, + "loss": 0.921, + "step": 66634 + }, + { + "epoch": 4.92, + "learning_rate": 1.6401814614055644e-06, + "loss": 0.9824, + "step": 66635 + }, + { + "epoch": 4.92, + "learning_rate": 1.6399625844276957e-06, + "loss": 0.9599, + "step": 66636 + }, + { + "epoch": 4.92, + "learning_rate": 1.6397437207505306e-06, + "loss": 0.9525, + "step": 66637 + }, + { + "epoch": 4.92, + "learning_rate": 1.6395248703744105e-06, + "loss": 0.9873, + "step": 66638 + }, + { + "epoch": 4.92, + "learning_rate": 1.639306033299688e-06, + "loss": 0.9313, + "step": 66639 + }, + { + "epoch": 4.92, + "learning_rate": 1.6390872095267075e-06, + "loss": 0.973, + "step": 66640 + }, + { + "epoch": 4.92, + "learning_rate": 1.638868399055824e-06, + "loss": 0.8753, + "step": 66641 + }, + { + "epoch": 4.92, + "learning_rate": 1.6386496018873777e-06, + "loss": 1.0827, + "step": 66642 + }, + { + "epoch": 4.92, + "learning_rate": 1.6384308180217211e-06, + "loss": 1.0476, + "step": 66643 + }, + { + "epoch": 4.92, + "learning_rate": 1.6382120474592022e-06, + "loss": 0.9794, + "step": 66644 + }, + { + "epoch": 4.92, + "learning_rate": 1.6379932902001694e-06, + "loss": 0.9457, + "step": 66645 + }, + { + "epoch": 4.92, + "learning_rate": 1.6377745462449679e-06, + "loss": 1.0012, + "step": 66646 + }, + { + "epoch": 4.92, + "learning_rate": 1.6375558155939464e-06, + "loss": 0.9151, + "step": 66647 + }, + { + "epoch": 4.92, + "learning_rate": 1.6373370982474557e-06, + "loss": 0.8378, + "step": 66648 + }, + { + "epoch": 4.92, + "learning_rate": 1.6371183942058421e-06, + "loss": 1.0488, + "step": 66649 + }, + { + "epoch": 4.92, + "learning_rate": 1.6368997034694534e-06, + "loss": 1.0253, + "step": 66650 + }, + { + "epoch": 4.92, + "learning_rate": 1.6366810260386346e-06, + "loss": 1.1181, + "step": 66651 + }, + { + "epoch": 4.92, + "learning_rate": 1.6364623619137398e-06, + "loss": 0.875, + "step": 66652 + }, + { + "epoch": 4.92, + "learning_rate": 1.6362437110951135e-06, + "loss": 0.9628, + "step": 66653 + }, + { + "epoch": 4.93, + "learning_rate": 1.636025073583104e-06, + "loss": 0.9658, + "step": 66654 + }, + { + "epoch": 4.93, + "learning_rate": 1.635806449378059e-06, + "loss": 1.0154, + "step": 66655 + }, + { + "epoch": 4.93, + "learning_rate": 1.6355878384803236e-06, + "loss": 0.9985, + "step": 66656 + }, + { + "epoch": 4.93, + "learning_rate": 1.6353692408902533e-06, + "loss": 0.9495, + "step": 66657 + }, + { + "epoch": 4.93, + "learning_rate": 1.6351506566081853e-06, + "loss": 0.9893, + "step": 66658 + }, + { + "epoch": 4.93, + "learning_rate": 1.6349320856344764e-06, + "loss": 1.0429, + "step": 66659 + }, + { + "epoch": 4.93, + "learning_rate": 1.6347135279694682e-06, + "loss": 1.0612, + "step": 66660 + }, + { + "epoch": 4.93, + "learning_rate": 1.634494983613516e-06, + "loss": 0.9079, + "step": 66661 + }, + { + "epoch": 4.93, + "learning_rate": 1.6342764525669574e-06, + "loss": 0.88, + "step": 66662 + }, + { + "epoch": 4.93, + "learning_rate": 1.6340579348301478e-06, + "loss": 1.0579, + "step": 66663 + }, + { + "epoch": 4.93, + "learning_rate": 1.6338394304034323e-06, + "loss": 1.1145, + "step": 66664 + }, + { + "epoch": 4.93, + "learning_rate": 1.6336209392871583e-06, + "loss": 1.0576, + "step": 66665 + }, + { + "epoch": 4.93, + "learning_rate": 1.6334024614816736e-06, + "loss": 0.9798, + "step": 66666 + }, + { + "epoch": 4.93, + "learning_rate": 1.6331839969873243e-06, + "loss": 0.9506, + "step": 66667 + }, + { + "epoch": 4.93, + "learning_rate": 1.6329655458044614e-06, + "loss": 1.0452, + "step": 66668 + }, + { + "epoch": 4.93, + "learning_rate": 1.632747107933431e-06, + "loss": 1.0481, + "step": 66669 + }, + { + "epoch": 4.93, + "learning_rate": 1.632528683374579e-06, + "loss": 1.0387, + "step": 66670 + }, + { + "epoch": 4.93, + "learning_rate": 1.6323102721282535e-06, + "loss": 0.9278, + "step": 66671 + }, + { + "epoch": 4.93, + "learning_rate": 1.6320918741948044e-06, + "loss": 1.0395, + "step": 66672 + }, + { + "epoch": 4.93, + "learning_rate": 1.6318734895745768e-06, + "loss": 0.9888, + "step": 66673 + }, + { + "epoch": 4.93, + "learning_rate": 1.6316551182679187e-06, + "loss": 0.9991, + "step": 66674 + }, + { + "epoch": 4.93, + "learning_rate": 1.631436760275178e-06, + "loss": 0.8992, + "step": 66675 + }, + { + "epoch": 4.93, + "learning_rate": 1.6312184155966993e-06, + "loss": 1.0269, + "step": 66676 + }, + { + "epoch": 4.93, + "learning_rate": 1.6310000842328356e-06, + "loss": 0.9689, + "step": 66677 + }, + { + "epoch": 4.93, + "learning_rate": 1.6307817661839299e-06, + "loss": 1.143, + "step": 66678 + }, + { + "epoch": 4.93, + "learning_rate": 1.6305634614503308e-06, + "loss": 1.0687, + "step": 66679 + }, + { + "epoch": 4.93, + "learning_rate": 1.6303451700323836e-06, + "loss": 0.9012, + "step": 66680 + }, + { + "epoch": 4.93, + "learning_rate": 1.6301268919304391e-06, + "loss": 1.0687, + "step": 66681 + }, + { + "epoch": 4.93, + "learning_rate": 1.6299086271448461e-06, + "loss": 0.8716, + "step": 66682 + }, + { + "epoch": 4.93, + "learning_rate": 1.629690375675943e-06, + "loss": 1.0808, + "step": 66683 + }, + { + "epoch": 4.93, + "learning_rate": 1.6294721375240862e-06, + "loss": 1.0205, + "step": 66684 + }, + { + "epoch": 4.93, + "learning_rate": 1.6292539126896166e-06, + "loss": 0.9068, + "step": 66685 + }, + { + "epoch": 4.93, + "learning_rate": 1.6290357011728896e-06, + "loss": 0.9524, + "step": 66686 + }, + { + "epoch": 4.93, + "learning_rate": 1.6288175029742416e-06, + "loss": 0.9762, + "step": 66687 + }, + { + "epoch": 4.93, + "learning_rate": 1.6285993180940275e-06, + "loss": 0.9712, + "step": 66688 + }, + { + "epoch": 4.93, + "learning_rate": 1.6283811465325916e-06, + "loss": 0.9582, + "step": 66689 + }, + { + "epoch": 4.93, + "learning_rate": 1.6281629882902828e-06, + "loss": 1.0664, + "step": 66690 + }, + { + "epoch": 4.93, + "learning_rate": 1.6279448433674461e-06, + "loss": 0.998, + "step": 66691 + }, + { + "epoch": 4.93, + "learning_rate": 1.6277267117644268e-06, + "loss": 1.0, + "step": 66692 + }, + { + "epoch": 4.93, + "learning_rate": 1.6275085934815783e-06, + "loss": 0.9364, + "step": 66693 + }, + { + "epoch": 4.93, + "learning_rate": 1.627290488519242e-06, + "loss": 1.0198, + "step": 66694 + }, + { + "epoch": 4.93, + "learning_rate": 1.627072396877768e-06, + "loss": 1.0436, + "step": 66695 + }, + { + "epoch": 4.93, + "learning_rate": 1.6268543185574992e-06, + "loss": 0.9888, + "step": 66696 + }, + { + "epoch": 4.93, + "learning_rate": 1.6266362535587876e-06, + "loss": 0.936, + "step": 66697 + }, + { + "epoch": 4.93, + "learning_rate": 1.6264182018819784e-06, + "loss": 1.0525, + "step": 66698 + }, + { + "epoch": 4.93, + "learning_rate": 1.6262001635274182e-06, + "loss": 0.9743, + "step": 66699 + }, + { + "epoch": 4.93, + "learning_rate": 1.6259821384954499e-06, + "loss": 0.9349, + "step": 66700 + }, + { + "epoch": 4.93, + "learning_rate": 1.6257641267864278e-06, + "loss": 1.0202, + "step": 66701 + }, + { + "epoch": 4.93, + "learning_rate": 1.625546128400698e-06, + "loss": 1.0174, + "step": 66702 + }, + { + "epoch": 4.93, + "learning_rate": 1.625328143338598e-06, + "loss": 0.9848, + "step": 66703 + }, + { + "epoch": 4.93, + "learning_rate": 1.6251101716004846e-06, + "loss": 1.0325, + "step": 66704 + }, + { + "epoch": 4.93, + "learning_rate": 1.6248922131866995e-06, + "loss": 0.9358, + "step": 66705 + }, + { + "epoch": 4.93, + "learning_rate": 1.6246742680975946e-06, + "loss": 1.0991, + "step": 66706 + }, + { + "epoch": 4.93, + "learning_rate": 1.6244563363335087e-06, + "loss": 1.0075, + "step": 66707 + }, + { + "epoch": 4.93, + "learning_rate": 1.6242384178947956e-06, + "loss": 0.9577, + "step": 66708 + }, + { + "epoch": 4.93, + "learning_rate": 1.6240205127817965e-06, + "loss": 0.9875, + "step": 66709 + }, + { + "epoch": 4.93, + "learning_rate": 1.6238026209948654e-06, + "loss": 0.9339, + "step": 66710 + }, + { + "epoch": 4.93, + "learning_rate": 1.6235847425343433e-06, + "loss": 1.0573, + "step": 66711 + }, + { + "epoch": 4.93, + "learning_rate": 1.623366877400575e-06, + "loss": 0.918, + "step": 66712 + }, + { + "epoch": 4.93, + "learning_rate": 1.623149025593912e-06, + "loss": 0.9823, + "step": 66713 + }, + { + "epoch": 4.93, + "learning_rate": 1.6229311871146992e-06, + "loss": 1.0177, + "step": 66714 + }, + { + "epoch": 4.93, + "learning_rate": 1.622713361963284e-06, + "loss": 0.97, + "step": 66715 + }, + { + "epoch": 4.93, + "learning_rate": 1.6224955501400085e-06, + "loss": 0.9298, + "step": 66716 + }, + { + "epoch": 4.93, + "learning_rate": 1.6222777516452259e-06, + "loss": 0.9529, + "step": 66717 + }, + { + "epoch": 4.93, + "learning_rate": 1.6220599664792791e-06, + "loss": 0.9475, + "step": 66718 + }, + { + "epoch": 4.93, + "learning_rate": 1.6218421946425146e-06, + "loss": 0.9083, + "step": 66719 + }, + { + "epoch": 4.93, + "learning_rate": 1.6216244361352796e-06, + "loss": 1.0734, + "step": 66720 + }, + { + "epoch": 4.93, + "learning_rate": 1.6214066909579184e-06, + "loss": 0.9608, + "step": 66721 + }, + { + "epoch": 4.93, + "learning_rate": 1.621188959110781e-06, + "loss": 1.004, + "step": 66722 + }, + { + "epoch": 4.93, + "learning_rate": 1.6209712405942123e-06, + "loss": 0.9953, + "step": 66723 + }, + { + "epoch": 4.93, + "learning_rate": 1.620753535408559e-06, + "loss": 0.969, + "step": 66724 + }, + { + "epoch": 4.93, + "learning_rate": 1.620535843554164e-06, + "loss": 1.0034, + "step": 66725 + }, + { + "epoch": 4.93, + "learning_rate": 1.6203181650313793e-06, + "loss": 0.993, + "step": 66726 + }, + { + "epoch": 4.93, + "learning_rate": 1.6201004998405478e-06, + "loss": 1.0123, + "step": 66727 + }, + { + "epoch": 4.93, + "learning_rate": 1.6198828479820173e-06, + "loss": 0.9526, + "step": 66728 + }, + { + "epoch": 4.93, + "learning_rate": 1.6196652094561328e-06, + "loss": 0.9061, + "step": 66729 + }, + { + "epoch": 4.93, + "learning_rate": 1.6194475842632396e-06, + "loss": 0.8785, + "step": 66730 + }, + { + "epoch": 4.93, + "learning_rate": 1.6192299724036887e-06, + "loss": 0.9577, + "step": 66731 + }, + { + "epoch": 4.93, + "learning_rate": 1.6190123738778196e-06, + "loss": 0.9833, + "step": 66732 + }, + { + "epoch": 4.93, + "learning_rate": 1.6187947886859834e-06, + "loss": 0.9583, + "step": 66733 + }, + { + "epoch": 4.93, + "learning_rate": 1.6185772168285253e-06, + "loss": 0.9448, + "step": 66734 + }, + { + "epoch": 4.93, + "learning_rate": 1.6183596583057904e-06, + "loss": 1.0835, + "step": 66735 + }, + { + "epoch": 4.93, + "learning_rate": 1.618142113118123e-06, + "loss": 0.9541, + "step": 66736 + }, + { + "epoch": 4.93, + "learning_rate": 1.617924581265874e-06, + "loss": 0.9446, + "step": 66737 + }, + { + "epoch": 4.93, + "learning_rate": 1.6177070627493874e-06, + "loss": 0.9652, + "step": 66738 + }, + { + "epoch": 4.93, + "learning_rate": 1.6174895575690087e-06, + "loss": 0.9684, + "step": 66739 + }, + { + "epoch": 4.93, + "learning_rate": 1.617272065725084e-06, + "loss": 1.0162, + "step": 66740 + }, + { + "epoch": 4.93, + "learning_rate": 1.6170545872179577e-06, + "loss": 1.0622, + "step": 66741 + }, + { + "epoch": 4.93, + "learning_rate": 1.6168371220479794e-06, + "loss": 0.9839, + "step": 66742 + }, + { + "epoch": 4.93, + "learning_rate": 1.6166196702154935e-06, + "loss": 0.8666, + "step": 66743 + }, + { + "epoch": 4.93, + "learning_rate": 1.616402231720846e-06, + "loss": 1.0904, + "step": 66744 + }, + { + "epoch": 4.93, + "learning_rate": 1.6161848065643804e-06, + "loss": 1.0713, + "step": 66745 + }, + { + "epoch": 4.93, + "learning_rate": 1.615967394746446e-06, + "loss": 0.9357, + "step": 66746 + }, + { + "epoch": 4.93, + "learning_rate": 1.6157499962673907e-06, + "loss": 0.9438, + "step": 66747 + }, + { + "epoch": 4.93, + "learning_rate": 1.6155326111275526e-06, + "loss": 1.0403, + "step": 66748 + }, + { + "epoch": 4.93, + "learning_rate": 1.6153152393272841e-06, + "loss": 1.0432, + "step": 66749 + }, + { + "epoch": 4.93, + "learning_rate": 1.6150978808669271e-06, + "loss": 1.01, + "step": 66750 + }, + { + "epoch": 4.93, + "learning_rate": 1.6148805357468345e-06, + "loss": 0.9951, + "step": 66751 + }, + { + "epoch": 4.93, + "learning_rate": 1.6146632039673415e-06, + "loss": 1.0442, + "step": 66752 + }, + { + "epoch": 4.93, + "learning_rate": 1.6144458855288025e-06, + "loss": 1.0675, + "step": 66753 + }, + { + "epoch": 4.93, + "learning_rate": 1.6142285804315572e-06, + "loss": 1.0808, + "step": 66754 + }, + { + "epoch": 4.93, + "learning_rate": 1.6140112886759596e-06, + "loss": 0.9324, + "step": 66755 + }, + { + "epoch": 4.93, + "learning_rate": 1.6137940102623439e-06, + "loss": 0.8425, + "step": 66756 + }, + { + "epoch": 4.93, + "learning_rate": 1.6135767451910656e-06, + "loss": 1.0679, + "step": 66757 + }, + { + "epoch": 4.93, + "learning_rate": 1.6133594934624664e-06, + "loss": 0.9918, + "step": 66758 + }, + { + "epoch": 4.93, + "learning_rate": 1.6131422550768917e-06, + "loss": 0.96, + "step": 66759 + }, + { + "epoch": 4.93, + "learning_rate": 1.6129250300346888e-06, + "loss": 1.0782, + "step": 66760 + }, + { + "epoch": 4.93, + "learning_rate": 1.6127078183361988e-06, + "loss": 0.8859, + "step": 66761 + }, + { + "epoch": 4.93, + "learning_rate": 1.6124906199817736e-06, + "loss": 0.9703, + "step": 66762 + }, + { + "epoch": 4.93, + "learning_rate": 1.612273434971755e-06, + "loss": 1.067, + "step": 66763 + }, + { + "epoch": 4.93, + "learning_rate": 1.6120562633064907e-06, + "loss": 0.9475, + "step": 66764 + }, + { + "epoch": 4.93, + "learning_rate": 1.6118391049863213e-06, + "loss": 0.9174, + "step": 66765 + }, + { + "epoch": 4.93, + "learning_rate": 1.611621960011599e-06, + "loss": 1.0824, + "step": 66766 + }, + { + "epoch": 4.93, + "learning_rate": 1.6114048283826656e-06, + "loss": 1.0428, + "step": 66767 + }, + { + "epoch": 4.93, + "learning_rate": 1.6111877100998673e-06, + "loss": 1.1051, + "step": 66768 + }, + { + "epoch": 4.93, + "learning_rate": 1.6109706051635499e-06, + "loss": 1.0506, + "step": 66769 + }, + { + "epoch": 4.93, + "learning_rate": 1.610753513574055e-06, + "loss": 0.982, + "step": 66770 + }, + { + "epoch": 4.93, + "learning_rate": 1.6105364353317343e-06, + "loss": 0.9928, + "step": 66771 + }, + { + "epoch": 4.93, + "learning_rate": 1.610319370436929e-06, + "loss": 0.8486, + "step": 66772 + }, + { + "epoch": 4.93, + "learning_rate": 1.6101023188899866e-06, + "loss": 1.0482, + "step": 66773 + }, + { + "epoch": 4.93, + "learning_rate": 1.609885280691249e-06, + "loss": 1.0281, + "step": 66774 + }, + { + "epoch": 4.93, + "learning_rate": 1.609668255841066e-06, + "loss": 0.9725, + "step": 66775 + }, + { + "epoch": 4.93, + "learning_rate": 1.6094512443397826e-06, + "loss": 0.9914, + "step": 66776 + }, + { + "epoch": 4.93, + "learning_rate": 1.6092342461877386e-06, + "loss": 1.0382, + "step": 66777 + }, + { + "epoch": 4.93, + "learning_rate": 1.609017261385284e-06, + "loss": 0.9681, + "step": 66778 + }, + { + "epoch": 4.93, + "learning_rate": 1.6088002899327637e-06, + "loss": 1.0025, + "step": 66779 + }, + { + "epoch": 4.93, + "learning_rate": 1.608583331830522e-06, + "loss": 0.9519, + "step": 66780 + }, + { + "epoch": 4.93, + "learning_rate": 1.6083663870789012e-06, + "loss": 0.9139, + "step": 66781 + }, + { + "epoch": 4.93, + "learning_rate": 1.6081494556782528e-06, + "loss": 0.9749, + "step": 66782 + }, + { + "epoch": 4.93, + "learning_rate": 1.6079325376289178e-06, + "loss": 0.9409, + "step": 66783 + }, + { + "epoch": 4.93, + "learning_rate": 1.6077156329312416e-06, + "loss": 0.9226, + "step": 66784 + }, + { + "epoch": 4.93, + "learning_rate": 1.6074987415855702e-06, + "loss": 1.0159, + "step": 66785 + }, + { + "epoch": 4.93, + "learning_rate": 1.607281863592246e-06, + "loss": 0.9611, + "step": 66786 + }, + { + "epoch": 4.93, + "learning_rate": 1.6070649989516196e-06, + "loss": 0.9506, + "step": 66787 + }, + { + "epoch": 4.93, + "learning_rate": 1.6068481476640318e-06, + "loss": 0.9487, + "step": 66788 + }, + { + "epoch": 4.93, + "learning_rate": 1.6066313097298281e-06, + "loss": 0.9697, + "step": 66789 + }, + { + "epoch": 4.94, + "learning_rate": 1.6064144851493524e-06, + "loss": 1.0134, + "step": 66790 + }, + { + "epoch": 4.94, + "learning_rate": 1.6061976739229524e-06, + "loss": 0.8657, + "step": 66791 + }, + { + "epoch": 4.94, + "learning_rate": 1.6059808760509732e-06, + "loss": 0.9683, + "step": 66792 + }, + { + "epoch": 4.94, + "learning_rate": 1.605764091533757e-06, + "loss": 1.064, + "step": 66793 + }, + { + "epoch": 4.94, + "learning_rate": 1.605547320371651e-06, + "loss": 0.9555, + "step": 66794 + }, + { + "epoch": 4.94, + "learning_rate": 1.6053305625649962e-06, + "loss": 1.0224, + "step": 66795 + }, + { + "epoch": 4.94, + "learning_rate": 1.6051138181141445e-06, + "loss": 1.0978, + "step": 66796 + }, + { + "epoch": 4.94, + "learning_rate": 1.6048970870194337e-06, + "loss": 1.0712, + "step": 66797 + }, + { + "epoch": 4.94, + "learning_rate": 1.6046803692812119e-06, + "loss": 0.9077, + "step": 66798 + }, + { + "epoch": 4.94, + "learning_rate": 1.6044636648998223e-06, + "loss": 0.9912, + "step": 66799 + }, + { + "epoch": 4.94, + "learning_rate": 1.604246973875615e-06, + "loss": 1.0097, + "step": 66800 + }, + { + "epoch": 4.94, + "learning_rate": 1.604030296208927e-06, + "loss": 1.0352, + "step": 66801 + }, + { + "epoch": 4.94, + "learning_rate": 1.6038136319001085e-06, + "loss": 1.0192, + "step": 66802 + }, + { + "epoch": 4.94, + "learning_rate": 1.6035969809495022e-06, + "loss": 0.9961, + "step": 66803 + }, + { + "epoch": 4.94, + "learning_rate": 1.6033803433574525e-06, + "loss": 1.0307, + "step": 66804 + }, + { + "epoch": 4.94, + "learning_rate": 1.6031637191243054e-06, + "loss": 0.965, + "step": 66805 + }, + { + "epoch": 4.94, + "learning_rate": 1.602947108250401e-06, + "loss": 0.9871, + "step": 66806 + }, + { + "epoch": 4.94, + "learning_rate": 1.6027305107360914e-06, + "loss": 1.0163, + "step": 66807 + }, + { + "epoch": 4.94, + "learning_rate": 1.6025139265817168e-06, + "loss": 0.907, + "step": 66808 + }, + { + "epoch": 4.94, + "learning_rate": 1.602297355787623e-06, + "loss": 1.0754, + "step": 66809 + }, + { + "epoch": 4.94, + "learning_rate": 1.602080798354151e-06, + "loss": 0.9864, + "step": 66810 + }, + { + "epoch": 4.94, + "learning_rate": 1.601864254281651e-06, + "loss": 1.0342, + "step": 66811 + }, + { + "epoch": 4.94, + "learning_rate": 1.6016477235704642e-06, + "loss": 1.0071, + "step": 66812 + }, + { + "epoch": 4.94, + "learning_rate": 1.6014312062209358e-06, + "loss": 1.037, + "step": 66813 + }, + { + "epoch": 4.94, + "learning_rate": 1.6012147022334113e-06, + "loss": 1.0123, + "step": 66814 + }, + { + "epoch": 4.94, + "learning_rate": 1.6009982116082302e-06, + "loss": 0.906, + "step": 66815 + }, + { + "epoch": 4.94, + "learning_rate": 1.6007817343457445e-06, + "loss": 1.0205, + "step": 66816 + }, + { + "epoch": 4.94, + "learning_rate": 1.6005652704462938e-06, + "loss": 1.0422, + "step": 66817 + }, + { + "epoch": 4.94, + "learning_rate": 1.6003488199102234e-06, + "loss": 0.9807, + "step": 66818 + }, + { + "epoch": 4.94, + "learning_rate": 1.6001323827378767e-06, + "loss": 1.0338, + "step": 66819 + }, + { + "epoch": 4.94, + "learning_rate": 1.5999159589296009e-06, + "loss": 1.0036, + "step": 66820 + }, + { + "epoch": 4.94, + "learning_rate": 1.599699548485738e-06, + "loss": 1.0418, + "step": 66821 + }, + { + "epoch": 4.94, + "learning_rate": 1.5994831514066345e-06, + "loss": 0.9461, + "step": 66822 + }, + { + "epoch": 4.94, + "learning_rate": 1.599266767692632e-06, + "loss": 1.1364, + "step": 66823 + }, + { + "epoch": 4.94, + "learning_rate": 1.5990503973440752e-06, + "loss": 1.0117, + "step": 66824 + }, + { + "epoch": 4.94, + "learning_rate": 1.59883404036131e-06, + "loss": 0.9708, + "step": 66825 + }, + { + "epoch": 4.94, + "learning_rate": 1.5986176967446775e-06, + "loss": 0.9793, + "step": 66826 + }, + { + "epoch": 4.94, + "learning_rate": 1.5984013664945263e-06, + "loss": 0.9827, + "step": 66827 + }, + { + "epoch": 4.94, + "learning_rate": 1.5981850496111983e-06, + "loss": 1.0911, + "step": 66828 + }, + { + "epoch": 4.94, + "learning_rate": 1.5979687460950366e-06, + "loss": 0.9927, + "step": 66829 + }, + { + "epoch": 4.94, + "learning_rate": 1.5977524559463852e-06, + "loss": 1.063, + "step": 66830 + }, + { + "epoch": 4.94, + "learning_rate": 1.597536179165592e-06, + "loss": 0.9644, + "step": 66831 + }, + { + "epoch": 4.94, + "learning_rate": 1.5973199157529973e-06, + "loss": 0.9713, + "step": 66832 + }, + { + "epoch": 4.94, + "learning_rate": 1.5971036657089477e-06, + "loss": 1.0841, + "step": 66833 + }, + { + "epoch": 4.94, + "learning_rate": 1.5968874290337855e-06, + "loss": 0.8771, + "step": 66834 + }, + { + "epoch": 4.94, + "learning_rate": 1.5966712057278522e-06, + "loss": 0.9524, + "step": 66835 + }, + { + "epoch": 4.94, + "learning_rate": 1.5964549957914976e-06, + "loss": 1.022, + "step": 66836 + }, + { + "epoch": 4.94, + "learning_rate": 1.5962387992250627e-06, + "loss": 1.0164, + "step": 66837 + }, + { + "epoch": 4.94, + "learning_rate": 1.5960226160288916e-06, + "loss": 0.9267, + "step": 66838 + }, + { + "epoch": 4.94, + "learning_rate": 1.5958064462033261e-06, + "loss": 1.098, + "step": 66839 + }, + { + "epoch": 4.94, + "learning_rate": 1.5955902897487151e-06, + "loss": 1.0844, + "step": 66840 + }, + { + "epoch": 4.94, + "learning_rate": 1.5953741466654016e-06, + "loss": 0.9252, + "step": 66841 + }, + { + "epoch": 4.94, + "learning_rate": 1.5951580169537228e-06, + "loss": 1.0961, + "step": 66842 + }, + { + "epoch": 4.94, + "learning_rate": 1.5949419006140298e-06, + "loss": 1.1739, + "step": 66843 + }, + { + "epoch": 4.94, + "learning_rate": 1.5947257976466612e-06, + "loss": 1.0319, + "step": 66844 + }, + { + "epoch": 4.94, + "learning_rate": 1.5945097080519689e-06, + "loss": 1.072, + "step": 66845 + }, + { + "epoch": 4.94, + "learning_rate": 1.5942936318302871e-06, + "loss": 0.868, + "step": 66846 + }, + { + "epoch": 4.94, + "learning_rate": 1.5940775689819654e-06, + "loss": 0.9702, + "step": 66847 + }, + { + "epoch": 4.94, + "learning_rate": 1.5938615195073437e-06, + "loss": 1.0569, + "step": 66848 + }, + { + "epoch": 4.94, + "learning_rate": 1.5936454834067728e-06, + "loss": 1.084, + "step": 66849 + }, + { + "epoch": 4.94, + "learning_rate": 1.59342946068059e-06, + "loss": 0.995, + "step": 66850 + }, + { + "epoch": 4.94, + "learning_rate": 1.5932134513291376e-06, + "loss": 1.0739, + "step": 66851 + }, + { + "epoch": 4.94, + "learning_rate": 1.5929974553527649e-06, + "loss": 0.9994, + "step": 66852 + }, + { + "epoch": 4.94, + "learning_rate": 1.5927814727518133e-06, + "loss": 0.9777, + "step": 66853 + }, + { + "epoch": 4.94, + "learning_rate": 1.5925655035266253e-06, + "loss": 0.8601, + "step": 66854 + }, + { + "epoch": 4.94, + "learning_rate": 1.592349547677543e-06, + "loss": 1.0691, + "step": 66855 + }, + { + "epoch": 4.94, + "learning_rate": 1.5921336052049153e-06, + "loss": 1.0057, + "step": 66856 + }, + { + "epoch": 4.94, + "learning_rate": 1.5919176761090814e-06, + "loss": 0.9912, + "step": 66857 + }, + { + "epoch": 4.94, + "learning_rate": 1.5917017603903873e-06, + "loss": 0.9806, + "step": 66858 + }, + { + "epoch": 4.94, + "learning_rate": 1.5914858580491754e-06, + "loss": 1.0822, + "step": 66859 + }, + { + "epoch": 4.94, + "learning_rate": 1.591269969085787e-06, + "loss": 0.9539, + "step": 66860 + }, + { + "epoch": 4.94, + "learning_rate": 1.5910540935005703e-06, + "loss": 1.0328, + "step": 66861 + }, + { + "epoch": 4.94, + "learning_rate": 1.5908382312938653e-06, + "loss": 1.0112, + "step": 66862 + }, + { + "epoch": 4.94, + "learning_rate": 1.5906223824660171e-06, + "loss": 1.1298, + "step": 66863 + }, + { + "epoch": 4.94, + "learning_rate": 1.5904065470173656e-06, + "loss": 1.0084, + "step": 66864 + }, + { + "epoch": 4.94, + "learning_rate": 1.5901907249482617e-06, + "loss": 1.0346, + "step": 66865 + }, + { + "epoch": 4.94, + "learning_rate": 1.5899749162590395e-06, + "loss": 1.1392, + "step": 66866 + }, + { + "epoch": 4.94, + "learning_rate": 1.5897591209500484e-06, + "loss": 1.1032, + "step": 66867 + }, + { + "epoch": 4.94, + "learning_rate": 1.5895433390216285e-06, + "loss": 1.0894, + "step": 66868 + }, + { + "epoch": 4.94, + "learning_rate": 1.5893275704741307e-06, + "loss": 0.9418, + "step": 66869 + }, + { + "epoch": 4.94, + "learning_rate": 1.5891118153078888e-06, + "loss": 0.9911, + "step": 66870 + }, + { + "epoch": 4.94, + "learning_rate": 1.5888960735232472e-06, + "loss": 0.9347, + "step": 66871 + }, + { + "epoch": 4.94, + "learning_rate": 1.5886803451205545e-06, + "loss": 1.0053, + "step": 66872 + }, + { + "epoch": 4.94, + "learning_rate": 1.5884646301001506e-06, + "loss": 0.903, + "step": 66873 + }, + { + "epoch": 4.94, + "learning_rate": 1.5882489284623791e-06, + "loss": 1.0512, + "step": 66874 + }, + { + "epoch": 4.94, + "learning_rate": 1.5880332402075815e-06, + "loss": 1.0411, + "step": 66875 + }, + { + "epoch": 4.94, + "learning_rate": 1.5878175653361039e-06, + "loss": 1.0413, + "step": 66876 + }, + { + "epoch": 4.94, + "learning_rate": 1.5876019038482892e-06, + "loss": 1.0493, + "step": 66877 + }, + { + "epoch": 4.94, + "learning_rate": 1.5873862557444785e-06, + "loss": 0.957, + "step": 66878 + }, + { + "epoch": 4.94, + "learning_rate": 1.587170621025017e-06, + "loss": 0.8159, + "step": 66879 + }, + { + "epoch": 4.94, + "learning_rate": 1.5869549996902445e-06, + "loss": 1.0788, + "step": 66880 + }, + { + "epoch": 4.94, + "learning_rate": 1.5867393917405072e-06, + "loss": 0.971, + "step": 66881 + }, + { + "epoch": 4.94, + "learning_rate": 1.5865237971761483e-06, + "loss": 0.9907, + "step": 66882 + }, + { + "epoch": 4.94, + "learning_rate": 1.5863082159975096e-06, + "loss": 0.9456, + "step": 66883 + }, + { + "epoch": 4.94, + "learning_rate": 1.586092648204931e-06, + "loss": 0.9756, + "step": 66884 + }, + { + "epoch": 4.94, + "learning_rate": 1.5858770937987611e-06, + "loss": 0.9314, + "step": 66885 + }, + { + "epoch": 4.94, + "learning_rate": 1.5856615527793406e-06, + "loss": 1.035, + "step": 66886 + }, + { + "epoch": 4.94, + "learning_rate": 1.5854460251470128e-06, + "loss": 1.0799, + "step": 66887 + }, + { + "epoch": 4.94, + "learning_rate": 1.5852305109021193e-06, + "loss": 0.8345, + "step": 66888 + }, + { + "epoch": 4.94, + "learning_rate": 1.5850150100450013e-06, + "loss": 0.9626, + "step": 66889 + }, + { + "epoch": 4.94, + "learning_rate": 1.5847995225760093e-06, + "loss": 1.1059, + "step": 66890 + }, + { + "epoch": 4.94, + "learning_rate": 1.5845840484954755e-06, + "loss": 0.9557, + "step": 66891 + }, + { + "epoch": 4.94, + "learning_rate": 1.5843685878037507e-06, + "loss": 0.9127, + "step": 66892 + }, + { + "epoch": 4.94, + "learning_rate": 1.5841531405011723e-06, + "loss": 1.0122, + "step": 66893 + }, + { + "epoch": 4.94, + "learning_rate": 1.5839377065880911e-06, + "loss": 0.9154, + "step": 66894 + }, + { + "epoch": 4.94, + "learning_rate": 1.5837222860648394e-06, + "loss": 0.9287, + "step": 66895 + }, + { + "epoch": 4.94, + "learning_rate": 1.5835068789317688e-06, + "loss": 1.0343, + "step": 66896 + }, + { + "epoch": 4.94, + "learning_rate": 1.5832914851892167e-06, + "loss": 1.1043, + "step": 66897 + }, + { + "epoch": 4.94, + "learning_rate": 1.5830761048375276e-06, + "loss": 1.0888, + "step": 66898 + }, + { + "epoch": 4.94, + "learning_rate": 1.5828607378770445e-06, + "loss": 1.0318, + "step": 66899 + }, + { + "epoch": 4.94, + "learning_rate": 1.582645384308108e-06, + "loss": 1.0169, + "step": 66900 + }, + { + "epoch": 4.94, + "learning_rate": 1.5824300441310635e-06, + "loss": 0.979, + "step": 66901 + }, + { + "epoch": 4.94, + "learning_rate": 1.582214717346252e-06, + "loss": 0.9921, + "step": 66902 + }, + { + "epoch": 4.94, + "learning_rate": 1.5819994039540176e-06, + "loss": 1.0101, + "step": 66903 + }, + { + "epoch": 4.94, + "learning_rate": 1.5817841039546988e-06, + "loss": 1.0619, + "step": 66904 + }, + { + "epoch": 4.94, + "learning_rate": 1.5815688173486433e-06, + "loss": 1.0478, + "step": 66905 + }, + { + "epoch": 4.94, + "learning_rate": 1.5813535441361915e-06, + "loss": 1.0395, + "step": 66906 + }, + { + "epoch": 4.94, + "learning_rate": 1.5811382843176858e-06, + "loss": 0.9659, + "step": 66907 + }, + { + "epoch": 4.94, + "learning_rate": 1.5809230378934681e-06, + "loss": 1.0758, + "step": 66908 + }, + { + "epoch": 4.94, + "learning_rate": 1.5807078048638801e-06, + "loss": 0.9653, + "step": 66909 + }, + { + "epoch": 4.94, + "learning_rate": 1.5804925852292696e-06, + "loss": 0.9643, + "step": 66910 + }, + { + "epoch": 4.94, + "learning_rate": 1.5802773789899694e-06, + "loss": 1.1105, + "step": 66911 + }, + { + "epoch": 4.94, + "learning_rate": 1.5800621861463317e-06, + "loss": 1.0444, + "step": 66912 + }, + { + "epoch": 4.94, + "learning_rate": 1.5798470066986903e-06, + "loss": 0.9019, + "step": 66913 + }, + { + "epoch": 4.94, + "learning_rate": 1.5796318406473976e-06, + "loss": 0.9904, + "step": 66914 + }, + { + "epoch": 4.94, + "learning_rate": 1.5794166879927852e-06, + "loss": 0.9337, + "step": 66915 + }, + { + "epoch": 4.94, + "learning_rate": 1.579201548735203e-06, + "loss": 0.8536, + "step": 66916 + }, + { + "epoch": 4.94, + "learning_rate": 1.5789864228749907e-06, + "loss": 1.0161, + "step": 66917 + }, + { + "epoch": 4.94, + "learning_rate": 1.5787713104124902e-06, + "loss": 1.0142, + "step": 66918 + }, + { + "epoch": 4.94, + "learning_rate": 1.5785562113480434e-06, + "loss": 1.0551, + "step": 66919 + }, + { + "epoch": 4.94, + "learning_rate": 1.5783411256819926e-06, + "loss": 1.0348, + "step": 66920 + }, + { + "epoch": 4.94, + "learning_rate": 1.5781260534146814e-06, + "loss": 0.9986, + "step": 66921 + }, + { + "epoch": 4.94, + "learning_rate": 1.5779109945464521e-06, + "loss": 1.0243, + "step": 66922 + }, + { + "epoch": 4.94, + "learning_rate": 1.5776959490776445e-06, + "loss": 0.9594, + "step": 66923 + }, + { + "epoch": 4.94, + "learning_rate": 1.5774809170086015e-06, + "loss": 0.9675, + "step": 66924 + }, + { + "epoch": 4.95, + "learning_rate": 1.577265898339667e-06, + "loss": 0.9408, + "step": 66925 + }, + { + "epoch": 4.95, + "learning_rate": 1.5770508930711825e-06, + "loss": 0.9389, + "step": 66926 + }, + { + "epoch": 4.95, + "learning_rate": 1.5768359012034895e-06, + "loss": 0.8108, + "step": 66927 + }, + { + "epoch": 4.95, + "learning_rate": 1.57662092273693e-06, + "loss": 1.0526, + "step": 66928 + }, + { + "epoch": 4.95, + "learning_rate": 1.5764059576718438e-06, + "loss": 0.9546, + "step": 66929 + }, + { + "epoch": 4.95, + "learning_rate": 1.5761910060085773e-06, + "loss": 0.9654, + "step": 66930 + }, + { + "epoch": 4.95, + "learning_rate": 1.57597606774747e-06, + "loss": 0.9894, + "step": 66931 + }, + { + "epoch": 4.95, + "learning_rate": 1.5757611428888652e-06, + "loss": 0.8886, + "step": 66932 + }, + { + "epoch": 4.95, + "learning_rate": 1.5755462314331016e-06, + "loss": 0.9424, + "step": 66933 + }, + { + "epoch": 4.95, + "learning_rate": 1.5753313333805253e-06, + "loss": 1.0222, + "step": 66934 + }, + { + "epoch": 4.95, + "learning_rate": 1.5751164487314785e-06, + "loss": 1.0705, + "step": 66935 + }, + { + "epoch": 4.95, + "learning_rate": 1.5749015774862963e-06, + "loss": 1.0509, + "step": 66936 + }, + { + "epoch": 4.95, + "learning_rate": 1.5746867196453274e-06, + "loss": 1.0607, + "step": 66937 + }, + { + "epoch": 4.95, + "learning_rate": 1.5744718752089095e-06, + "loss": 1.0717, + "step": 66938 + }, + { + "epoch": 4.95, + "learning_rate": 1.5742570441773909e-06, + "loss": 1.0206, + "step": 66939 + }, + { + "epoch": 4.95, + "learning_rate": 1.5740422265511035e-06, + "loss": 1.108, + "step": 66940 + }, + { + "epoch": 4.95, + "learning_rate": 1.5738274223303962e-06, + "loss": 0.9784, + "step": 66941 + }, + { + "epoch": 4.95, + "learning_rate": 1.5736126315156074e-06, + "loss": 0.9815, + "step": 66942 + }, + { + "epoch": 4.95, + "learning_rate": 1.573397854107086e-06, + "loss": 1.0525, + "step": 66943 + }, + { + "epoch": 4.95, + "learning_rate": 1.5731830901051647e-06, + "loss": 0.9836, + "step": 66944 + }, + { + "epoch": 4.95, + "learning_rate": 1.5729683395101858e-06, + "loss": 0.9228, + "step": 66945 + }, + { + "epoch": 4.95, + "learning_rate": 1.5727536023224966e-06, + "loss": 0.9796, + "step": 66946 + }, + { + "epoch": 4.95, + "learning_rate": 1.5725388785424356e-06, + "loss": 1.0532, + "step": 66947 + }, + { + "epoch": 4.95, + "learning_rate": 1.5723241681703449e-06, + "loss": 1.0073, + "step": 66948 + }, + { + "epoch": 4.95, + "learning_rate": 1.5721094712065633e-06, + "loss": 1.037, + "step": 66949 + }, + { + "epoch": 4.95, + "learning_rate": 1.571894787651438e-06, + "loss": 1.1158, + "step": 66950 + }, + { + "epoch": 4.95, + "learning_rate": 1.5716801175053065e-06, + "loss": 0.9602, + "step": 66951 + }, + { + "epoch": 4.95, + "learning_rate": 1.5714654607685121e-06, + "loss": 0.915, + "step": 66952 + }, + { + "epoch": 4.95, + "learning_rate": 1.5712508174413953e-06, + "loss": 1.0555, + "step": 66953 + }, + { + "epoch": 4.95, + "learning_rate": 1.5710361875242951e-06, + "loss": 1.0223, + "step": 66954 + }, + { + "epoch": 4.95, + "learning_rate": 1.5708215710175612e-06, + "loss": 1.0396, + "step": 66955 + }, + { + "epoch": 4.95, + "learning_rate": 1.5706069679215242e-06, + "loss": 0.9335, + "step": 66956 + }, + { + "epoch": 4.95, + "learning_rate": 1.5703923782365339e-06, + "loss": 0.9567, + "step": 66957 + }, + { + "epoch": 4.95, + "learning_rate": 1.5701778019629255e-06, + "loss": 0.9109, + "step": 66958 + }, + { + "epoch": 4.95, + "learning_rate": 1.5699632391010489e-06, + "loss": 1.1125, + "step": 66959 + }, + { + "epoch": 4.95, + "learning_rate": 1.5697486896512348e-06, + "loss": 0.994, + "step": 66960 + }, + { + "epoch": 4.95, + "learning_rate": 1.569534153613833e-06, + "loss": 0.9687, + "step": 66961 + }, + { + "epoch": 4.95, + "learning_rate": 1.5693196309891822e-06, + "loss": 1.0841, + "step": 66962 + }, + { + "epoch": 4.95, + "learning_rate": 1.5691051217776222e-06, + "loss": 0.981, + "step": 66963 + }, + { + "epoch": 4.95, + "learning_rate": 1.5688906259794957e-06, + "loss": 1.0337, + "step": 66964 + }, + { + "epoch": 4.95, + "learning_rate": 1.5686761435951415e-06, + "loss": 1.0516, + "step": 66965 + }, + { + "epoch": 4.95, + "learning_rate": 1.568461674624905e-06, + "loss": 1.037, + "step": 66966 + }, + { + "epoch": 4.95, + "learning_rate": 1.5682472190691257e-06, + "loss": 1.0879, + "step": 66967 + }, + { + "epoch": 4.95, + "learning_rate": 1.5680327769281433e-06, + "loss": 1.0526, + "step": 66968 + }, + { + "epoch": 4.95, + "learning_rate": 1.5678183482022991e-06, + "loss": 0.9441, + "step": 66969 + }, + { + "epoch": 4.95, + "learning_rate": 1.5676039328919368e-06, + "loss": 0.9138, + "step": 66970 + }, + { + "epoch": 4.95, + "learning_rate": 1.5673895309973963e-06, + "loss": 0.9962, + "step": 66971 + }, + { + "epoch": 4.95, + "learning_rate": 1.5671751425190184e-06, + "loss": 1.0004, + "step": 66972 + }, + { + "epoch": 4.95, + "learning_rate": 1.5669607674571442e-06, + "loss": 1.0947, + "step": 66973 + }, + { + "epoch": 4.95, + "learning_rate": 1.5667464058121128e-06, + "loss": 1.053, + "step": 66974 + }, + { + "epoch": 4.95, + "learning_rate": 1.566532057584269e-06, + "loss": 0.9623, + "step": 66975 + }, + { + "epoch": 4.95, + "learning_rate": 1.5663177227739523e-06, + "loss": 0.9962, + "step": 66976 + }, + { + "epoch": 4.95, + "learning_rate": 1.5661034013815035e-06, + "loss": 1.0909, + "step": 66977 + }, + { + "epoch": 4.95, + "learning_rate": 1.5658890934072613e-06, + "loss": 1.0966, + "step": 66978 + }, + { + "epoch": 4.95, + "learning_rate": 1.5656747988515709e-06, + "loss": 0.9718, + "step": 66979 + }, + { + "epoch": 4.95, + "learning_rate": 1.5654605177147708e-06, + "loss": 1.0826, + "step": 66980 + }, + { + "epoch": 4.95, + "learning_rate": 1.5652462499972032e-06, + "loss": 0.9633, + "step": 66981 + }, + { + "epoch": 4.95, + "learning_rate": 1.5650319956992077e-06, + "loss": 0.9426, + "step": 66982 + }, + { + "epoch": 4.95, + "learning_rate": 1.564817754821123e-06, + "loss": 1.0333, + "step": 66983 + }, + { + "epoch": 4.95, + "learning_rate": 1.5646035273632975e-06, + "loss": 0.9993, + "step": 66984 + }, + { + "epoch": 4.95, + "learning_rate": 1.5643893133260625e-06, + "loss": 1.0736, + "step": 66985 + }, + { + "epoch": 4.95, + "learning_rate": 1.5641751127097659e-06, + "loss": 0.898, + "step": 66986 + }, + { + "epoch": 4.95, + "learning_rate": 1.563960925514746e-06, + "loss": 1.0882, + "step": 66987 + }, + { + "epoch": 4.95, + "learning_rate": 1.563746751741343e-06, + "loss": 1.0355, + "step": 66988 + }, + { + "epoch": 4.95, + "learning_rate": 1.5635325913898957e-06, + "loss": 0.9705, + "step": 66989 + }, + { + "epoch": 4.95, + "learning_rate": 1.5633184444607507e-06, + "loss": 0.9125, + "step": 66990 + }, + { + "epoch": 4.95, + "learning_rate": 1.5631043109542443e-06, + "loss": 1.0391, + "step": 66991 + }, + { + "epoch": 4.95, + "learning_rate": 1.5628901908707195e-06, + "loss": 0.9416, + "step": 66992 + }, + { + "epoch": 4.95, + "learning_rate": 1.562676084210515e-06, + "loss": 0.9073, + "step": 66993 + }, + { + "epoch": 4.95, + "learning_rate": 1.5624619909739691e-06, + "loss": 1.0536, + "step": 66994 + }, + { + "epoch": 4.95, + "learning_rate": 1.5622479111614287e-06, + "loss": 0.8854, + "step": 66995 + }, + { + "epoch": 4.95, + "learning_rate": 1.5620338447732307e-06, + "loss": 1.048, + "step": 66996 + }, + { + "epoch": 4.95, + "learning_rate": 1.5618197918097166e-06, + "loss": 0.9723, + "step": 66997 + }, + { + "epoch": 4.95, + "learning_rate": 1.5616057522712237e-06, + "loss": 0.9781, + "step": 66998 + }, + { + "epoch": 4.95, + "learning_rate": 1.561391726158098e-06, + "loss": 0.9967, + "step": 66999 + }, + { + "epoch": 4.95, + "learning_rate": 1.5611777134706796e-06, + "loss": 1.0818, + "step": 67000 + }, + { + "epoch": 4.95, + "learning_rate": 1.5609637142093015e-06, + "loss": 0.9772, + "step": 67001 + }, + { + "epoch": 4.95, + "learning_rate": 1.5607497283743122e-06, + "loss": 0.9566, + "step": 67002 + }, + { + "epoch": 4.95, + "learning_rate": 1.5605357559660471e-06, + "loss": 1.0677, + "step": 67003 + }, + { + "epoch": 4.95, + "learning_rate": 1.5603217969848539e-06, + "loss": 0.8956, + "step": 67004 + }, + { + "epoch": 4.95, + "learning_rate": 1.560107851431063e-06, + "loss": 1.028, + "step": 67005 + }, + { + "epoch": 4.95, + "learning_rate": 1.5598939193050212e-06, + "loss": 1.0181, + "step": 67006 + }, + { + "epoch": 4.95, + "learning_rate": 1.5596800006070655e-06, + "loss": 0.9978, + "step": 67007 + }, + { + "epoch": 4.95, + "learning_rate": 1.5594660953375441e-06, + "loss": 0.9589, + "step": 67008 + }, + { + "epoch": 4.95, + "learning_rate": 1.5592522034967882e-06, + "loss": 0.9454, + "step": 67009 + }, + { + "epoch": 4.95, + "learning_rate": 1.5590383250851383e-06, + "loss": 1.0054, + "step": 67010 + }, + { + "epoch": 4.95, + "learning_rate": 1.5588244601029401e-06, + "loss": 1.035, + "step": 67011 + }, + { + "epoch": 4.95, + "learning_rate": 1.5586106085505326e-06, + "loss": 1.0562, + "step": 67012 + }, + { + "epoch": 4.95, + "learning_rate": 1.5583967704282532e-06, + "loss": 0.9852, + "step": 67013 + }, + { + "epoch": 4.95, + "learning_rate": 1.5581829457364429e-06, + "loss": 0.9873, + "step": 67014 + }, + { + "epoch": 4.95, + "learning_rate": 1.5579691344754445e-06, + "loss": 1.055, + "step": 67015 + }, + { + "epoch": 4.95, + "learning_rate": 1.5577553366455966e-06, + "loss": 0.9205, + "step": 67016 + }, + { + "epoch": 4.95, + "learning_rate": 1.5575415522472382e-06, + "loss": 1.0058, + "step": 67017 + }, + { + "epoch": 4.95, + "learning_rate": 1.557327781280712e-06, + "loss": 1.0141, + "step": 67018 + }, + { + "epoch": 4.95, + "learning_rate": 1.5571140237463523e-06, + "loss": 1.0357, + "step": 67019 + }, + { + "epoch": 4.95, + "learning_rate": 1.5569002796445077e-06, + "loss": 1.0754, + "step": 67020 + }, + { + "epoch": 4.95, + "learning_rate": 1.5566865489755124e-06, + "loss": 1.0106, + "step": 67021 + }, + { + "epoch": 4.95, + "learning_rate": 1.5564728317397082e-06, + "loss": 0.9747, + "step": 67022 + }, + { + "epoch": 4.95, + "learning_rate": 1.5562591279374329e-06, + "loss": 0.9877, + "step": 67023 + }, + { + "epoch": 4.95, + "learning_rate": 1.5560454375690314e-06, + "loss": 0.9806, + "step": 67024 + }, + { + "epoch": 4.95, + "learning_rate": 1.5558317606348406e-06, + "loss": 0.999, + "step": 67025 + }, + { + "epoch": 4.95, + "learning_rate": 1.5556180971351997e-06, + "loss": 0.9414, + "step": 67026 + }, + { + "epoch": 4.95, + "learning_rate": 1.555404447070449e-06, + "loss": 1.0236, + "step": 67027 + }, + { + "epoch": 4.95, + "learning_rate": 1.5551908104409296e-06, + "loss": 1.0031, + "step": 67028 + }, + { + "epoch": 4.95, + "learning_rate": 1.5549771872469832e-06, + "loss": 0.901, + "step": 67029 + }, + { + "epoch": 4.95, + "learning_rate": 1.5547635774889437e-06, + "loss": 0.9439, + "step": 67030 + }, + { + "epoch": 4.95, + "learning_rate": 1.554549981167156e-06, + "loss": 1.1415, + "step": 67031 + }, + { + "epoch": 4.95, + "learning_rate": 1.5543363982819593e-06, + "loss": 1.0363, + "step": 67032 + }, + { + "epoch": 4.95, + "learning_rate": 1.5541228288336918e-06, + "loss": 1.0603, + "step": 67033 + }, + { + "epoch": 4.95, + "learning_rate": 1.5539092728226924e-06, + "loss": 1.0115, + "step": 67034 + }, + { + "epoch": 4.95, + "learning_rate": 1.5536957302493038e-06, + "loss": 1.0333, + "step": 67035 + }, + { + "epoch": 4.95, + "learning_rate": 1.553482201113865e-06, + "loss": 0.9788, + "step": 67036 + }, + { + "epoch": 4.95, + "learning_rate": 1.5532686854167157e-06, + "loss": 0.9482, + "step": 67037 + }, + { + "epoch": 4.95, + "learning_rate": 1.5530551831581942e-06, + "loss": 1.0017, + "step": 67038 + }, + { + "epoch": 4.95, + "learning_rate": 1.5528416943386394e-06, + "loss": 1.016, + "step": 67039 + }, + { + "epoch": 4.95, + "learning_rate": 1.5526282189583941e-06, + "loss": 0.9684, + "step": 67040 + }, + { + "epoch": 4.95, + "learning_rate": 1.5524147570177972e-06, + "loss": 0.9145, + "step": 67041 + }, + { + "epoch": 4.95, + "learning_rate": 1.5522013085171871e-06, + "loss": 1.006, + "step": 67042 + }, + { + "epoch": 4.95, + "learning_rate": 1.5519878734569015e-06, + "loss": 0.8871, + "step": 67043 + }, + { + "epoch": 4.95, + "learning_rate": 1.5517744518372858e-06, + "loss": 0.9664, + "step": 67044 + }, + { + "epoch": 4.95, + "learning_rate": 1.5515610436586748e-06, + "loss": 1.0054, + "step": 67045 + }, + { + "epoch": 4.95, + "learning_rate": 1.5513476489214096e-06, + "loss": 0.9453, + "step": 67046 + }, + { + "epoch": 4.95, + "learning_rate": 1.551134267625829e-06, + "loss": 0.9118, + "step": 67047 + }, + { + "epoch": 4.95, + "learning_rate": 1.5509208997722714e-06, + "loss": 0.9947, + "step": 67048 + }, + { + "epoch": 4.95, + "learning_rate": 1.550707545361082e-06, + "loss": 1.0723, + "step": 67049 + }, + { + "epoch": 4.95, + "learning_rate": 1.550494204392592e-06, + "loss": 0.9708, + "step": 67050 + }, + { + "epoch": 4.95, + "learning_rate": 1.5502808768671463e-06, + "loss": 1.0615, + "step": 67051 + }, + { + "epoch": 4.95, + "learning_rate": 1.5500675627850804e-06, + "loss": 0.9163, + "step": 67052 + }, + { + "epoch": 4.95, + "learning_rate": 1.5498542621467415e-06, + "loss": 0.9896, + "step": 67053 + }, + { + "epoch": 4.95, + "learning_rate": 1.5496409749524588e-06, + "loss": 1.0796, + "step": 67054 + }, + { + "epoch": 4.95, + "learning_rate": 1.5494277012025782e-06, + "loss": 1.0096, + "step": 67055 + }, + { + "epoch": 4.95, + "learning_rate": 1.5492144408974375e-06, + "loss": 0.9377, + "step": 67056 + }, + { + "epoch": 4.95, + "learning_rate": 1.5490011940373751e-06, + "loss": 0.8958, + "step": 67057 + }, + { + "epoch": 4.95, + "learning_rate": 1.548787960622732e-06, + "loss": 1.0723, + "step": 67058 + }, + { + "epoch": 4.95, + "learning_rate": 1.548574740653843e-06, + "loss": 0.9252, + "step": 67059 + }, + { + "epoch": 4.96, + "learning_rate": 1.5483615341310543e-06, + "loss": 1.0669, + "step": 67060 + }, + { + "epoch": 4.96, + "learning_rate": 1.5481483410547005e-06, + "loss": 1.0381, + "step": 67061 + }, + { + "epoch": 4.96, + "learning_rate": 1.5479351614251214e-06, + "loss": 0.8843, + "step": 67062 + }, + { + "epoch": 4.96, + "learning_rate": 1.5477219952426547e-06, + "loss": 0.8849, + "step": 67063 + }, + { + "epoch": 4.96, + "learning_rate": 1.5475088425076435e-06, + "loss": 1.0768, + "step": 67064 + }, + { + "epoch": 4.96, + "learning_rate": 1.547295703220425e-06, + "loss": 1.0271, + "step": 67065 + }, + { + "epoch": 4.96, + "learning_rate": 1.547082577381339e-06, + "loss": 0.9256, + "step": 67066 + }, + { + "epoch": 4.96, + "learning_rate": 1.5468694649907223e-06, + "loss": 0.9736, + "step": 67067 + }, + { + "epoch": 4.96, + "learning_rate": 1.546656366048913e-06, + "loss": 1.0395, + "step": 67068 + }, + { + "epoch": 4.96, + "learning_rate": 1.5464432805562557e-06, + "loss": 1.0587, + "step": 67069 + }, + { + "epoch": 4.96, + "learning_rate": 1.5462302085130864e-06, + "loss": 1.0397, + "step": 67070 + }, + { + "epoch": 4.96, + "learning_rate": 1.5460171499197428e-06, + "loss": 0.9288, + "step": 67071 + }, + { + "epoch": 4.96, + "learning_rate": 1.5458041047765638e-06, + "loss": 0.9512, + "step": 67072 + }, + { + "epoch": 4.96, + "learning_rate": 1.5455910730838908e-06, + "loss": 1.0257, + "step": 67073 + }, + { + "epoch": 4.96, + "learning_rate": 1.545378054842064e-06, + "loss": 0.8421, + "step": 67074 + }, + { + "epoch": 4.96, + "learning_rate": 1.5451650500514148e-06, + "loss": 0.9362, + "step": 67075 + }, + { + "epoch": 4.96, + "learning_rate": 1.54495205871229e-06, + "loss": 1.1543, + "step": 67076 + }, + { + "epoch": 4.96, + "learning_rate": 1.544739080825025e-06, + "loss": 0.954, + "step": 67077 + }, + { + "epoch": 4.96, + "learning_rate": 1.5445261163899594e-06, + "loss": 0.8725, + "step": 67078 + }, + { + "epoch": 4.96, + "learning_rate": 1.5443131654074294e-06, + "loss": 1.036, + "step": 67079 + }, + { + "epoch": 4.96, + "learning_rate": 1.5441002278777794e-06, + "loss": 1.0533, + "step": 67080 + }, + { + "epoch": 4.96, + "learning_rate": 1.5438873038013435e-06, + "loss": 0.9936, + "step": 67081 + }, + { + "epoch": 4.96, + "learning_rate": 1.5436743931784625e-06, + "loss": 1.0234, + "step": 67082 + }, + { + "epoch": 4.96, + "learning_rate": 1.543461496009473e-06, + "loss": 1.0411, + "step": 67083 + }, + { + "epoch": 4.96, + "learning_rate": 1.5432486122947165e-06, + "loss": 1.1622, + "step": 67084 + }, + { + "epoch": 4.96, + "learning_rate": 1.543035742034531e-06, + "loss": 1.0223, + "step": 67085 + }, + { + "epoch": 4.96, + "learning_rate": 1.5428228852292547e-06, + "loss": 1.0315, + "step": 67086 + }, + { + "epoch": 4.96, + "learning_rate": 1.5426100418792257e-06, + "loss": 1.0308, + "step": 67087 + }, + { + "epoch": 4.96, + "learning_rate": 1.5423972119847808e-06, + "loss": 1.0315, + "step": 67088 + }, + { + "epoch": 4.96, + "learning_rate": 1.5421843955462635e-06, + "loss": 1.021, + "step": 67089 + }, + { + "epoch": 4.96, + "learning_rate": 1.54197159256401e-06, + "loss": 1.1018, + "step": 67090 + }, + { + "epoch": 4.96, + "learning_rate": 1.5417588030383591e-06, + "loss": 0.9378, + "step": 67091 + }, + { + "epoch": 4.96, + "learning_rate": 1.541546026969646e-06, + "loss": 1.018, + "step": 67092 + }, + { + "epoch": 4.96, + "learning_rate": 1.5413332643582135e-06, + "loss": 0.9991, + "step": 67093 + }, + { + "epoch": 4.96, + "learning_rate": 1.541120515204403e-06, + "loss": 1.0389, + "step": 67094 + }, + { + "epoch": 4.96, + "learning_rate": 1.5409077795085426e-06, + "loss": 0.9511, + "step": 67095 + }, + { + "epoch": 4.96, + "learning_rate": 1.54069505727098e-06, + "loss": 1.0521, + "step": 67096 + }, + { + "epoch": 4.96, + "learning_rate": 1.5404823484920472e-06, + "loss": 1.0053, + "step": 67097 + }, + { + "epoch": 4.96, + "learning_rate": 1.5402696531720917e-06, + "loss": 0.9569, + "step": 67098 + }, + { + "epoch": 4.96, + "learning_rate": 1.540056971311441e-06, + "loss": 0.9275, + "step": 67099 + }, + { + "epoch": 4.96, + "learning_rate": 1.5398443029104415e-06, + "loss": 1.0837, + "step": 67100 + }, + { + "epoch": 4.96, + "learning_rate": 1.5396316479694252e-06, + "loss": 0.9807, + "step": 67101 + }, + { + "epoch": 4.96, + "learning_rate": 1.5394190064887392e-06, + "loss": 0.9042, + "step": 67102 + }, + { + "epoch": 4.96, + "learning_rate": 1.539206378468715e-06, + "loss": 1.0653, + "step": 67103 + }, + { + "epoch": 4.96, + "learning_rate": 1.5389937639096896e-06, + "loss": 0.9235, + "step": 67104 + }, + { + "epoch": 4.96, + "learning_rate": 1.5387811628120052e-06, + "loss": 1.0185, + "step": 67105 + }, + { + "epoch": 4.96, + "learning_rate": 1.5385685751760005e-06, + "loss": 1.0568, + "step": 67106 + }, + { + "epoch": 4.96, + "learning_rate": 1.5383560010020116e-06, + "loss": 0.9606, + "step": 67107 + }, + { + "epoch": 4.96, + "learning_rate": 1.538143440290375e-06, + "loss": 0.9592, + "step": 67108 + }, + { + "epoch": 4.96, + "learning_rate": 1.5379308930414327e-06, + "loss": 1.0688, + "step": 67109 + }, + { + "epoch": 4.96, + "learning_rate": 1.5377183592555222e-06, + "loss": 1.0047, + "step": 67110 + }, + { + "epoch": 4.96, + "learning_rate": 1.53750583893298e-06, + "loss": 1.07, + "step": 67111 + }, + { + "epoch": 4.96, + "learning_rate": 1.5372933320741457e-06, + "loss": 0.9673, + "step": 67112 + }, + { + "epoch": 4.96, + "learning_rate": 1.5370808386793535e-06, + "loss": 0.9721, + "step": 67113 + }, + { + "epoch": 4.96, + "learning_rate": 1.5368683587489474e-06, + "loss": 1.0836, + "step": 67114 + }, + { + "epoch": 4.96, + "learning_rate": 1.536655892283262e-06, + "loss": 1.0242, + "step": 67115 + }, + { + "epoch": 4.96, + "learning_rate": 1.5364434392826366e-06, + "loss": 0.9189, + "step": 67116 + }, + { + "epoch": 4.96, + "learning_rate": 1.5362309997474067e-06, + "loss": 0.9268, + "step": 67117 + }, + { + "epoch": 4.96, + "learning_rate": 1.5360185736779132e-06, + "loss": 1.0193, + "step": 67118 + }, + { + "epoch": 4.96, + "learning_rate": 1.5358061610744946e-06, + "loss": 1.0544, + "step": 67119 + }, + { + "epoch": 4.96, + "learning_rate": 1.5355937619374862e-06, + "loss": 0.9547, + "step": 67120 + }, + { + "epoch": 4.96, + "learning_rate": 1.5353813762672276e-06, + "loss": 1.084, + "step": 67121 + }, + { + "epoch": 4.96, + "learning_rate": 1.5351690040640555e-06, + "loss": 0.9504, + "step": 67122 + }, + { + "epoch": 4.96, + "learning_rate": 1.5349566453283094e-06, + "loss": 1.0435, + "step": 67123 + }, + { + "epoch": 4.96, + "learning_rate": 1.5347443000603224e-06, + "loss": 0.928, + "step": 67124 + }, + { + "epoch": 4.96, + "learning_rate": 1.5345319682604398e-06, + "loss": 1.015, + "step": 67125 + }, + { + "epoch": 4.96, + "learning_rate": 1.5343196499289947e-06, + "loss": 0.9115, + "step": 67126 + }, + { + "epoch": 4.96, + "learning_rate": 1.5341073450663268e-06, + "loss": 0.9761, + "step": 67127 + }, + { + "epoch": 4.96, + "learning_rate": 1.5338950536727703e-06, + "loss": 0.9048, + "step": 67128 + }, + { + "epoch": 4.96, + "learning_rate": 1.5336827757486671e-06, + "loss": 1.0369, + "step": 67129 + }, + { + "epoch": 4.96, + "learning_rate": 1.5334705112943548e-06, + "loss": 1.0896, + "step": 67130 + }, + { + "epoch": 4.96, + "learning_rate": 1.5332582603101697e-06, + "loss": 1.0294, + "step": 67131 + }, + { + "epoch": 4.96, + "learning_rate": 1.5330460227964484e-06, + "loss": 1.0226, + "step": 67132 + }, + { + "epoch": 4.96, + "learning_rate": 1.532833798753528e-06, + "loss": 0.9872, + "step": 67133 + }, + { + "epoch": 4.96, + "learning_rate": 1.5326215881817507e-06, + "loss": 1.031, + "step": 67134 + }, + { + "epoch": 4.96, + "learning_rate": 1.532409391081451e-06, + "loss": 1.0458, + "step": 67135 + }, + { + "epoch": 4.96, + "learning_rate": 1.5321972074529668e-06, + "loss": 1.0481, + "step": 67136 + }, + { + "epoch": 4.96, + "learning_rate": 1.5319850372966328e-06, + "loss": 1.0834, + "step": 67137 + }, + { + "epoch": 4.96, + "learning_rate": 1.531772880612793e-06, + "loss": 1.0575, + "step": 67138 + }, + { + "epoch": 4.96, + "learning_rate": 1.5315607374017804e-06, + "loss": 0.9414, + "step": 67139 + }, + { + "epoch": 4.96, + "learning_rate": 1.531348607663934e-06, + "loss": 0.9211, + "step": 67140 + }, + { + "epoch": 4.96, + "learning_rate": 1.5311364913995907e-06, + "loss": 0.9514, + "step": 67141 + }, + { + "epoch": 4.96, + "learning_rate": 1.5309243886090853e-06, + "loss": 0.9984, + "step": 67142 + }, + { + "epoch": 4.96, + "learning_rate": 1.5307122992927626e-06, + "loss": 1.0275, + "step": 67143 + }, + { + "epoch": 4.96, + "learning_rate": 1.5305002234509513e-06, + "loss": 0.9262, + "step": 67144 + }, + { + "epoch": 4.96, + "learning_rate": 1.530288161083996e-06, + "loss": 1.0068, + "step": 67145 + }, + { + "epoch": 4.96, + "learning_rate": 1.5300761121922269e-06, + "loss": 1.0614, + "step": 67146 + }, + { + "epoch": 4.96, + "learning_rate": 1.529864076775992e-06, + "loss": 0.8789, + "step": 67147 + }, + { + "epoch": 4.96, + "learning_rate": 1.5296520548356165e-06, + "loss": 1.0059, + "step": 67148 + }, + { + "epoch": 4.96, + "learning_rate": 1.5294400463714455e-06, + "loss": 0.978, + "step": 67149 + }, + { + "epoch": 4.96, + "learning_rate": 1.5292280513838153e-06, + "loss": 0.9841, + "step": 67150 + }, + { + "epoch": 4.96, + "learning_rate": 1.5290160698730604e-06, + "loss": 1.0434, + "step": 67151 + }, + { + "epoch": 4.96, + "learning_rate": 1.5288041018395217e-06, + "loss": 1.1016, + "step": 67152 + }, + { + "epoch": 4.96, + "learning_rate": 1.5285921472835308e-06, + "loss": 1.0391, + "step": 67153 + }, + { + "epoch": 4.96, + "learning_rate": 1.528380206205431e-06, + "loss": 1.022, + "step": 67154 + }, + { + "epoch": 4.96, + "learning_rate": 1.5281682786055574e-06, + "loss": 0.9142, + "step": 67155 + }, + { + "epoch": 4.96, + "learning_rate": 1.5279563644842465e-06, + "loss": 1.0206, + "step": 67156 + }, + { + "epoch": 4.96, + "learning_rate": 1.5277444638418347e-06, + "loss": 1.0319, + "step": 67157 + }, + { + "epoch": 4.96, + "learning_rate": 1.5275325766786608e-06, + "loss": 0.9946, + "step": 67158 + }, + { + "epoch": 4.96, + "learning_rate": 1.5273207029950633e-06, + "loss": 0.9111, + "step": 67159 + }, + { + "epoch": 4.96, + "learning_rate": 1.5271088427913762e-06, + "loss": 1.0821, + "step": 67160 + }, + { + "epoch": 4.96, + "learning_rate": 1.526896996067937e-06, + "loss": 1.2287, + "step": 67161 + }, + { + "epoch": 4.96, + "learning_rate": 1.5266851628250822e-06, + "loss": 0.975, + "step": 67162 + }, + { + "epoch": 4.96, + "learning_rate": 1.5264733430631517e-06, + "loss": 1.0453, + "step": 67163 + }, + { + "epoch": 4.96, + "learning_rate": 1.5262615367824818e-06, + "loss": 1.0743, + "step": 67164 + }, + { + "epoch": 4.96, + "learning_rate": 1.5260497439834076e-06, + "loss": 1.0101, + "step": 67165 + }, + { + "epoch": 4.96, + "learning_rate": 1.5258379646662657e-06, + "loss": 1.0654, + "step": 67166 + }, + { + "epoch": 4.96, + "learning_rate": 1.525626198831398e-06, + "loss": 0.9817, + "step": 67167 + }, + { + "epoch": 4.96, + "learning_rate": 1.5254144464791353e-06, + "loss": 1.0849, + "step": 67168 + }, + { + "epoch": 4.96, + "learning_rate": 1.5252027076098152e-06, + "loss": 1.0238, + "step": 67169 + }, + { + "epoch": 4.96, + "learning_rate": 1.5249909822237797e-06, + "loss": 0.9705, + "step": 67170 + }, + { + "epoch": 4.96, + "learning_rate": 1.5247792703213605e-06, + "loss": 0.9469, + "step": 67171 + }, + { + "epoch": 4.96, + "learning_rate": 1.5245675719028973e-06, + "loss": 1.0738, + "step": 67172 + }, + { + "epoch": 4.96, + "learning_rate": 1.5243558869687226e-06, + "loss": 0.8713, + "step": 67173 + }, + { + "epoch": 4.96, + "learning_rate": 1.5241442155191799e-06, + "loss": 0.9702, + "step": 67174 + }, + { + "epoch": 4.96, + "learning_rate": 1.5239325575546016e-06, + "loss": 0.9718, + "step": 67175 + }, + { + "epoch": 4.96, + "learning_rate": 1.523720913075326e-06, + "loss": 0.9989, + "step": 67176 + }, + { + "epoch": 4.96, + "learning_rate": 1.5235092820816888e-06, + "loss": 1.0289, + "step": 67177 + }, + { + "epoch": 4.96, + "learning_rate": 1.523297664574025e-06, + "loss": 1.024, + "step": 67178 + }, + { + "epoch": 4.96, + "learning_rate": 1.5230860605526743e-06, + "loss": 1.0181, + "step": 67179 + }, + { + "epoch": 4.96, + "learning_rate": 1.5228744700179742e-06, + "loss": 0.8975, + "step": 67180 + }, + { + "epoch": 4.96, + "learning_rate": 1.5226628929702581e-06, + "loss": 1.0244, + "step": 67181 + }, + { + "epoch": 4.96, + "learning_rate": 1.5224513294098631e-06, + "loss": 0.9696, + "step": 67182 + }, + { + "epoch": 4.96, + "learning_rate": 1.522239779337128e-06, + "loss": 1.0121, + "step": 67183 + }, + { + "epoch": 4.96, + "learning_rate": 1.522028242752389e-06, + "loss": 0.9155, + "step": 67184 + }, + { + "epoch": 4.96, + "learning_rate": 1.5218167196559808e-06, + "loss": 1.1026, + "step": 67185 + }, + { + "epoch": 4.96, + "learning_rate": 1.5216052100482392e-06, + "loss": 1.0009, + "step": 67186 + }, + { + "epoch": 4.96, + "learning_rate": 1.5213937139295042e-06, + "loss": 1.0203, + "step": 67187 + }, + { + "epoch": 4.96, + "learning_rate": 1.5211822313001135e-06, + "loss": 1.033, + "step": 67188 + }, + { + "epoch": 4.96, + "learning_rate": 1.5209707621603954e-06, + "loss": 0.9587, + "step": 67189 + }, + { + "epoch": 4.96, + "learning_rate": 1.520759306510694e-06, + "loss": 0.9507, + "step": 67190 + }, + { + "epoch": 4.96, + "learning_rate": 1.5205478643513405e-06, + "loss": 1.0636, + "step": 67191 + }, + { + "epoch": 4.96, + "learning_rate": 1.52033643568268e-06, + "loss": 0.9262, + "step": 67192 + }, + { + "epoch": 4.96, + "learning_rate": 1.5201250205050366e-06, + "loss": 0.9564, + "step": 67193 + }, + { + "epoch": 4.96, + "learning_rate": 1.519913618818757e-06, + "loss": 0.9262, + "step": 67194 + }, + { + "epoch": 4.96, + "learning_rate": 1.5197022306241693e-06, + "loss": 1.0605, + "step": 67195 + }, + { + "epoch": 4.97, + "learning_rate": 1.5194908559216193e-06, + "loss": 1.0121, + "step": 67196 + }, + { + "epoch": 4.97, + "learning_rate": 1.5192794947114365e-06, + "loss": 1.0196, + "step": 67197 + }, + { + "epoch": 4.97, + "learning_rate": 1.5190681469939562e-06, + "loss": 0.8806, + "step": 67198 + }, + { + "epoch": 4.97, + "learning_rate": 1.518856812769519e-06, + "loss": 1.078, + "step": 67199 + }, + { + "epoch": 4.97, + "learning_rate": 1.5186454920384585e-06, + "loss": 1.0975, + "step": 67200 + }, + { + "epoch": 4.97, + "learning_rate": 1.518434184801113e-06, + "loss": 1.1021, + "step": 67201 + }, + { + "epoch": 4.97, + "learning_rate": 1.5182228910578145e-06, + "loss": 1.0321, + "step": 67202 + }, + { + "epoch": 4.97, + "learning_rate": 1.518011610808905e-06, + "loss": 1.0006, + "step": 67203 + }, + { + "epoch": 4.97, + "learning_rate": 1.5178003440547163e-06, + "loss": 1.0877, + "step": 67204 + }, + { + "epoch": 4.97, + "learning_rate": 1.517589090795587e-06, + "loss": 0.9094, + "step": 67205 + }, + { + "epoch": 4.97, + "learning_rate": 1.5173778510318515e-06, + "loss": 0.9924, + "step": 67206 + }, + { + "epoch": 4.97, + "learning_rate": 1.5171666247638451e-06, + "loss": 0.9346, + "step": 67207 + }, + { + "epoch": 4.97, + "learning_rate": 1.5169554119919084e-06, + "loss": 0.9951, + "step": 67208 + }, + { + "epoch": 4.97, + "learning_rate": 1.5167442127163712e-06, + "loss": 0.9716, + "step": 67209 + }, + { + "epoch": 4.97, + "learning_rate": 1.5165330269375733e-06, + "loss": 0.9286, + "step": 67210 + }, + { + "epoch": 4.97, + "learning_rate": 1.516321854655849e-06, + "loss": 1.1108, + "step": 67211 + }, + { + "epoch": 4.97, + "learning_rate": 1.5161106958715388e-06, + "loss": 1.0447, + "step": 67212 + }, + { + "epoch": 4.97, + "learning_rate": 1.5158995505849716e-06, + "loss": 1.0942, + "step": 67213 + }, + { + "epoch": 4.97, + "learning_rate": 1.5156884187964882e-06, + "loss": 1.0713, + "step": 67214 + }, + { + "epoch": 4.97, + "learning_rate": 1.5154773005064238e-06, + "loss": 1.0134, + "step": 67215 + }, + { + "epoch": 4.97, + "learning_rate": 1.5152661957151127e-06, + "loss": 0.9969, + "step": 67216 + }, + { + "epoch": 4.97, + "learning_rate": 1.5150551044228923e-06, + "loss": 0.9581, + "step": 67217 + }, + { + "epoch": 4.97, + "learning_rate": 1.5148440266300957e-06, + "loss": 1.0722, + "step": 67218 + }, + { + "epoch": 4.97, + "learning_rate": 1.5146329623370636e-06, + "loss": 0.9561, + "step": 67219 + }, + { + "epoch": 4.97, + "learning_rate": 1.5144219115441283e-06, + "loss": 1.0425, + "step": 67220 + }, + { + "epoch": 4.97, + "learning_rate": 1.5142108742516259e-06, + "loss": 0.9834, + "step": 67221 + }, + { + "epoch": 4.97, + "learning_rate": 1.5139998504598907e-06, + "loss": 1.0362, + "step": 67222 + }, + { + "epoch": 4.97, + "learning_rate": 1.5137888401692636e-06, + "loss": 1.0283, + "step": 67223 + }, + { + "epoch": 4.97, + "learning_rate": 1.5135778433800764e-06, + "loss": 1.0209, + "step": 67224 + }, + { + "epoch": 4.97, + "learning_rate": 1.5133668600926644e-06, + "loss": 0.956, + "step": 67225 + }, + { + "epoch": 4.97, + "learning_rate": 1.513155890307365e-06, + "loss": 0.9308, + "step": 67226 + }, + { + "epoch": 4.97, + "learning_rate": 1.5129449340245118e-06, + "loss": 0.9, + "step": 67227 + }, + { + "epoch": 4.97, + "learning_rate": 1.5127339912444428e-06, + "loss": 1.058, + "step": 67228 + }, + { + "epoch": 4.97, + "learning_rate": 1.5125230619674936e-06, + "loss": 1.0139, + "step": 67229 + }, + { + "epoch": 4.97, + "learning_rate": 1.5123121461939982e-06, + "loss": 1.0201, + "step": 67230 + }, + { + "epoch": 4.97, + "learning_rate": 1.5121012439242898e-06, + "loss": 0.9408, + "step": 67231 + }, + { + "epoch": 4.97, + "learning_rate": 1.5118903551587106e-06, + "loss": 1.0114, + "step": 67232 + }, + { + "epoch": 4.97, + "learning_rate": 1.511679479897593e-06, + "loss": 0.983, + "step": 67233 + }, + { + "epoch": 4.97, + "learning_rate": 1.5114686181412685e-06, + "loss": 0.9372, + "step": 67234 + }, + { + "epoch": 4.97, + "learning_rate": 1.5112577698900788e-06, + "loss": 1.1144, + "step": 67235 + }, + { + "epoch": 4.97, + "learning_rate": 1.5110469351443536e-06, + "loss": 1.031, + "step": 67236 + }, + { + "epoch": 4.97, + "learning_rate": 1.5108361139044347e-06, + "loss": 0.9589, + "step": 67237 + }, + { + "epoch": 4.97, + "learning_rate": 1.510625306170651e-06, + "loss": 1.0059, + "step": 67238 + }, + { + "epoch": 4.97, + "learning_rate": 1.5104145119433433e-06, + "loss": 1.0432, + "step": 67239 + }, + { + "epoch": 4.97, + "learning_rate": 1.5102037312228413e-06, + "loss": 1.0019, + "step": 67240 + }, + { + "epoch": 4.97, + "learning_rate": 1.509992964009489e-06, + "loss": 0.9785, + "step": 67241 + }, + { + "epoch": 4.97, + "learning_rate": 1.509782210303612e-06, + "loss": 0.9083, + "step": 67242 + }, + { + "epoch": 4.97, + "learning_rate": 1.509571470105552e-06, + "loss": 1.064, + "step": 67243 + }, + { + "epoch": 4.97, + "learning_rate": 1.5093607434156421e-06, + "loss": 0.9596, + "step": 67244 + }, + { + "epoch": 4.97, + "learning_rate": 1.5091500302342176e-06, + "loss": 1.0006, + "step": 67245 + }, + { + "epoch": 4.97, + "learning_rate": 1.508939330561614e-06, + "loss": 0.9661, + "step": 67246 + }, + { + "epoch": 4.97, + "learning_rate": 1.5087286443981652e-06, + "loss": 1.0318, + "step": 67247 + }, + { + "epoch": 4.97, + "learning_rate": 1.508517971744209e-06, + "loss": 0.9361, + "step": 67248 + }, + { + "epoch": 4.97, + "learning_rate": 1.508307312600079e-06, + "loss": 1.1104, + "step": 67249 + }, + { + "epoch": 4.97, + "learning_rate": 1.5080966669661112e-06, + "loss": 1.0762, + "step": 67250 + }, + { + "epoch": 4.97, + "learning_rate": 1.5078860348426372e-06, + "loss": 1.0053, + "step": 67251 + }, + { + "epoch": 4.97, + "learning_rate": 1.5076754162299978e-06, + "loss": 1.0145, + "step": 67252 + }, + { + "epoch": 4.97, + "learning_rate": 1.5074648111285272e-06, + "loss": 1.0864, + "step": 67253 + }, + { + "epoch": 4.97, + "learning_rate": 1.5072542195385541e-06, + "loss": 1.0639, + "step": 67254 + }, + { + "epoch": 4.97, + "learning_rate": 1.5070436414604206e-06, + "loss": 1.0294, + "step": 67255 + }, + { + "epoch": 4.97, + "learning_rate": 1.5068330768944573e-06, + "loss": 1.0108, + "step": 67256 + }, + { + "epoch": 4.97, + "learning_rate": 1.506622525841005e-06, + "loss": 1.1024, + "step": 67257 + }, + { + "epoch": 4.97, + "learning_rate": 1.5064119883003902e-06, + "loss": 0.9405, + "step": 67258 + }, + { + "epoch": 4.97, + "learning_rate": 1.506201464272955e-06, + "loss": 0.9608, + "step": 67259 + }, + { + "epoch": 4.97, + "learning_rate": 1.505990953759029e-06, + "loss": 0.9737, + "step": 67260 + }, + { + "epoch": 4.97, + "learning_rate": 1.5057804567589541e-06, + "loss": 0.8925, + "step": 67261 + }, + { + "epoch": 4.97, + "learning_rate": 1.50556997327306e-06, + "loss": 0.9513, + "step": 67262 + }, + { + "epoch": 4.97, + "learning_rate": 1.5053595033016788e-06, + "loss": 1.0035, + "step": 67263 + }, + { + "epoch": 4.97, + "learning_rate": 1.5051490468451524e-06, + "loss": 1.1402, + "step": 67264 + }, + { + "epoch": 4.97, + "learning_rate": 1.5049386039038116e-06, + "loss": 0.9312, + "step": 67265 + }, + { + "epoch": 4.97, + "learning_rate": 1.5047281744779929e-06, + "loss": 1.085, + "step": 67266 + }, + { + "epoch": 4.97, + "learning_rate": 1.504517758568027e-06, + "loss": 0.9503, + "step": 67267 + }, + { + "epoch": 4.97, + "learning_rate": 1.5043073561742539e-06, + "loss": 0.9501, + "step": 67268 + }, + { + "epoch": 4.97, + "learning_rate": 1.5040969672970075e-06, + "loss": 0.9412, + "step": 67269 + }, + { + "epoch": 4.97, + "learning_rate": 1.5038865919366197e-06, + "loss": 0.9803, + "step": 67270 + }, + { + "epoch": 4.97, + "learning_rate": 1.5036762300934272e-06, + "loss": 0.9801, + "step": 67271 + }, + { + "epoch": 4.97, + "learning_rate": 1.5034658817677628e-06, + "loss": 0.9408, + "step": 67272 + }, + { + "epoch": 4.97, + "learning_rate": 1.5032555469599642e-06, + "loss": 0.9389, + "step": 67273 + }, + { + "epoch": 4.97, + "learning_rate": 1.5030452256703642e-06, + "loss": 1.0908, + "step": 67274 + }, + { + "epoch": 4.97, + "learning_rate": 1.5028349178992984e-06, + "loss": 1.0576, + "step": 67275 + }, + { + "epoch": 4.97, + "learning_rate": 1.5026246236470987e-06, + "loss": 0.9376, + "step": 67276 + }, + { + "epoch": 4.97, + "learning_rate": 1.5024143429141024e-06, + "loss": 0.9532, + "step": 67277 + }, + { + "epoch": 4.97, + "learning_rate": 1.5022040757006452e-06, + "loss": 1.0028, + "step": 67278 + }, + { + "epoch": 4.97, + "learning_rate": 1.5019938220070585e-06, + "loss": 1.0395, + "step": 67279 + }, + { + "epoch": 4.97, + "learning_rate": 1.5017835818336778e-06, + "loss": 0.9996, + "step": 67280 + }, + { + "epoch": 4.97, + "learning_rate": 1.5015733551808364e-06, + "loss": 0.9742, + "step": 67281 + }, + { + "epoch": 4.97, + "learning_rate": 1.5013631420488739e-06, + "loss": 0.9502, + "step": 67282 + }, + { + "epoch": 4.97, + "learning_rate": 1.5011529424381177e-06, + "loss": 1.0691, + "step": 67283 + }, + { + "epoch": 4.97, + "learning_rate": 1.5009427563489066e-06, + "loss": 1.0111, + "step": 67284 + }, + { + "epoch": 4.97, + "learning_rate": 1.5007325837815724e-06, + "loss": 0.9668, + "step": 67285 + }, + { + "epoch": 4.97, + "learning_rate": 1.5005224247364559e-06, + "loss": 0.9653, + "step": 67286 + }, + { + "epoch": 4.97, + "learning_rate": 1.5003122792138814e-06, + "loss": 0.9854, + "step": 67287 + }, + { + "epoch": 4.97, + "learning_rate": 1.5001021472141908e-06, + "loss": 0.8636, + "step": 67288 + }, + { + "epoch": 4.97, + "learning_rate": 1.499892028737716e-06, + "loss": 1.0944, + "step": 67289 + }, + { + "epoch": 4.97, + "learning_rate": 1.4996819237847915e-06, + "loss": 0.9494, + "step": 67290 + }, + { + "epoch": 4.97, + "learning_rate": 1.4994718323557512e-06, + "loss": 1.0116, + "step": 67291 + }, + { + "epoch": 4.97, + "learning_rate": 1.499261754450928e-06, + "loss": 0.9227, + "step": 67292 + }, + { + "epoch": 4.97, + "learning_rate": 1.4990516900706598e-06, + "loss": 0.9808, + "step": 67293 + }, + { + "epoch": 4.97, + "learning_rate": 1.4988416392152782e-06, + "loss": 1.0002, + "step": 67294 + }, + { + "epoch": 4.97, + "learning_rate": 1.4986316018851188e-06, + "loss": 0.9753, + "step": 67295 + }, + { + "epoch": 4.97, + "learning_rate": 1.4984215780805122e-06, + "loss": 1.0638, + "step": 67296 + }, + { + "epoch": 4.97, + "learning_rate": 1.4982115678017972e-06, + "loss": 1.0102, + "step": 67297 + }, + { + "epoch": 4.97, + "learning_rate": 1.4980015710493058e-06, + "loss": 0.9704, + "step": 67298 + }, + { + "epoch": 4.97, + "learning_rate": 1.497791587823373e-06, + "loss": 0.882, + "step": 67299 + }, + { + "epoch": 4.97, + "learning_rate": 1.4975816181243318e-06, + "loss": 0.9644, + "step": 67300 + }, + { + "epoch": 4.97, + "learning_rate": 1.4973716619525136e-06, + "loss": 0.8975, + "step": 67301 + }, + { + "epoch": 4.97, + "learning_rate": 1.4971617193082611e-06, + "loss": 0.9725, + "step": 67302 + }, + { + "epoch": 4.97, + "learning_rate": 1.4969517901918973e-06, + "loss": 1.0666, + "step": 67303 + }, + { + "epoch": 4.97, + "learning_rate": 1.4967418746037643e-06, + "loss": 1.0935, + "step": 67304 + }, + { + "epoch": 4.97, + "learning_rate": 1.4965319725441906e-06, + "loss": 0.9705, + "step": 67305 + }, + { + "epoch": 4.97, + "learning_rate": 1.4963220840135185e-06, + "loss": 0.9567, + "step": 67306 + }, + { + "epoch": 4.97, + "learning_rate": 1.4961122090120705e-06, + "loss": 0.8859, + "step": 67307 + }, + { + "epoch": 4.97, + "learning_rate": 1.495902347540189e-06, + "loss": 1.0565, + "step": 67308 + }, + { + "epoch": 4.97, + "learning_rate": 1.495692499598206e-06, + "loss": 0.9189, + "step": 67309 + }, + { + "epoch": 4.97, + "learning_rate": 1.495482665186453e-06, + "loss": 1.0842, + "step": 67310 + }, + { + "epoch": 4.97, + "learning_rate": 1.4952728443052667e-06, + "loss": 1.074, + "step": 67311 + }, + { + "epoch": 4.97, + "learning_rate": 1.4950630369549768e-06, + "loss": 1.0023, + "step": 67312 + }, + { + "epoch": 4.97, + "learning_rate": 1.4948532431359219e-06, + "loss": 0.944, + "step": 67313 + }, + { + "epoch": 4.97, + "learning_rate": 1.494643462848434e-06, + "loss": 1.0783, + "step": 67314 + }, + { + "epoch": 4.97, + "learning_rate": 1.4944336960928474e-06, + "loss": 1.0305, + "step": 67315 + }, + { + "epoch": 4.97, + "learning_rate": 1.4942239428694917e-06, + "loss": 0.8526, + "step": 67316 + }, + { + "epoch": 4.97, + "learning_rate": 1.4940142031787063e-06, + "loss": 1.0082, + "step": 67317 + }, + { + "epoch": 4.97, + "learning_rate": 1.4938044770208237e-06, + "loss": 1.0605, + "step": 67318 + }, + { + "epoch": 4.97, + "learning_rate": 1.4935947643961758e-06, + "loss": 1.0658, + "step": 67319 + }, + { + "epoch": 4.97, + "learning_rate": 1.4933850653050973e-06, + "loss": 0.9651, + "step": 67320 + }, + { + "epoch": 4.97, + "learning_rate": 1.4931753797479188e-06, + "loss": 1.0185, + "step": 67321 + }, + { + "epoch": 4.97, + "learning_rate": 1.4929657077249793e-06, + "loss": 0.9616, + "step": 67322 + }, + { + "epoch": 4.97, + "learning_rate": 1.492756049236609e-06, + "loss": 0.9944, + "step": 67323 + }, + { + "epoch": 4.97, + "learning_rate": 1.4925464042831416e-06, + "loss": 0.9935, + "step": 67324 + }, + { + "epoch": 4.97, + "learning_rate": 1.4923367728649097e-06, + "loss": 1.0608, + "step": 67325 + }, + { + "epoch": 4.97, + "learning_rate": 1.4921271549822513e-06, + "loss": 1.0118, + "step": 67326 + }, + { + "epoch": 4.97, + "learning_rate": 1.4919175506354977e-06, + "loss": 0.9299, + "step": 67327 + }, + { + "epoch": 4.97, + "learning_rate": 1.4917079598249783e-06, + "loss": 1.0524, + "step": 67328 + }, + { + "epoch": 4.97, + "learning_rate": 1.491498382551031e-06, + "loss": 0.9845, + "step": 67329 + }, + { + "epoch": 4.97, + "learning_rate": 1.4912888188139884e-06, + "loss": 1.0547, + "step": 67330 + }, + { + "epoch": 4.98, + "learning_rate": 1.491079268614184e-06, + "loss": 0.9758, + "step": 67331 + }, + { + "epoch": 4.98, + "learning_rate": 1.490869731951947e-06, + "loss": 0.9846, + "step": 67332 + }, + { + "epoch": 4.98, + "learning_rate": 1.490660208827619e-06, + "loss": 1.0705, + "step": 67333 + }, + { + "epoch": 4.98, + "learning_rate": 1.490450699241527e-06, + "loss": 0.9836, + "step": 67334 + }, + { + "epoch": 4.98, + "learning_rate": 1.4902412031940072e-06, + "loss": 1.0061, + "step": 67335 + }, + { + "epoch": 4.98, + "learning_rate": 1.490031720685391e-06, + "loss": 0.9709, + "step": 67336 + }, + { + "epoch": 4.98, + "learning_rate": 1.489822251716011e-06, + "loss": 1.0389, + "step": 67337 + }, + { + "epoch": 4.98, + "learning_rate": 1.489612796286204e-06, + "loss": 0.9188, + "step": 67338 + }, + { + "epoch": 4.98, + "learning_rate": 1.4894033543963015e-06, + "loss": 1.0386, + "step": 67339 + }, + { + "epoch": 4.98, + "learning_rate": 1.489193926046637e-06, + "loss": 1.1542, + "step": 67340 + }, + { + "epoch": 4.98, + "learning_rate": 1.4889845112375401e-06, + "loss": 0.9846, + "step": 67341 + }, + { + "epoch": 4.98, + "learning_rate": 1.4887751099693492e-06, + "loss": 1.0134, + "step": 67342 + }, + { + "epoch": 4.98, + "learning_rate": 1.4885657222423967e-06, + "loss": 0.9362, + "step": 67343 + }, + { + "epoch": 4.98, + "learning_rate": 1.488356348057013e-06, + "loss": 1.051, + "step": 67344 + }, + { + "epoch": 4.98, + "learning_rate": 1.4881469874135313e-06, + "loss": 0.9686, + "step": 67345 + }, + { + "epoch": 4.98, + "learning_rate": 1.487937640312287e-06, + "loss": 1.0122, + "step": 67346 + }, + { + "epoch": 4.98, + "learning_rate": 1.4877283067536153e-06, + "loss": 0.9496, + "step": 67347 + }, + { + "epoch": 4.98, + "learning_rate": 1.4875189867378415e-06, + "loss": 0.982, + "step": 67348 + }, + { + "epoch": 4.98, + "learning_rate": 1.4873096802653052e-06, + "loss": 0.9745, + "step": 67349 + }, + { + "epoch": 4.98, + "learning_rate": 1.4871003873363343e-06, + "loss": 0.9929, + "step": 67350 + }, + { + "epoch": 4.98, + "learning_rate": 1.4868911079512705e-06, + "loss": 1.0349, + "step": 67351 + }, + { + "epoch": 4.98, + "learning_rate": 1.4866818421104367e-06, + "loss": 0.9222, + "step": 67352 + }, + { + "epoch": 4.98, + "learning_rate": 1.486472589814172e-06, + "loss": 0.9426, + "step": 67353 + }, + { + "epoch": 4.98, + "learning_rate": 1.4862633510628056e-06, + "loss": 1.0583, + "step": 67354 + }, + { + "epoch": 4.98, + "learning_rate": 1.4860541258566775e-06, + "loss": 1.122, + "step": 67355 + }, + { + "epoch": 4.98, + "learning_rate": 1.4858449141961117e-06, + "loss": 1.1173, + "step": 67356 + }, + { + "epoch": 4.98, + "learning_rate": 1.4856357160814439e-06, + "loss": 1.0315, + "step": 67357 + }, + { + "epoch": 4.98, + "learning_rate": 1.4854265315130102e-06, + "loss": 0.9537, + "step": 67358 + }, + { + "epoch": 4.98, + "learning_rate": 1.4852173604911403e-06, + "loss": 1.0187, + "step": 67359 + }, + { + "epoch": 4.98, + "learning_rate": 1.4850082030161684e-06, + "loss": 0.9566, + "step": 67360 + }, + { + "epoch": 4.98, + "learning_rate": 1.484799059088423e-06, + "loss": 1.0428, + "step": 67361 + }, + { + "epoch": 4.98, + "learning_rate": 1.4845899287082443e-06, + "loss": 1.0347, + "step": 67362 + }, + { + "epoch": 4.98, + "learning_rate": 1.4843808118759607e-06, + "loss": 1.0213, + "step": 67363 + }, + { + "epoch": 4.98, + "learning_rate": 1.484171708591905e-06, + "loss": 0.9765, + "step": 67364 + }, + { + "epoch": 4.98, + "learning_rate": 1.4839626188564105e-06, + "loss": 0.9932, + "step": 67365 + }, + { + "epoch": 4.98, + "learning_rate": 1.4837535426698069e-06, + "loss": 1.0167, + "step": 67366 + }, + { + "epoch": 4.98, + "learning_rate": 1.4835444800324327e-06, + "loss": 0.9097, + "step": 67367 + }, + { + "epoch": 4.98, + "learning_rate": 1.483335430944617e-06, + "loss": 0.9632, + "step": 67368 + }, + { + "epoch": 4.98, + "learning_rate": 1.4831263954066933e-06, + "loss": 1.0681, + "step": 67369 + }, + { + "epoch": 4.98, + "learning_rate": 1.4829173734189906e-06, + "loss": 1.0847, + "step": 67370 + }, + { + "epoch": 4.98, + "learning_rate": 1.4827083649818474e-06, + "loss": 1.0699, + "step": 67371 + }, + { + "epoch": 4.98, + "learning_rate": 1.4824993700955926e-06, + "loss": 0.9484, + "step": 67372 + }, + { + "epoch": 4.98, + "learning_rate": 1.48229038876056e-06, + "loss": 1.059, + "step": 67373 + }, + { + "epoch": 4.98, + "learning_rate": 1.4820814209770806e-06, + "loss": 0.9572, + "step": 67374 + }, + { + "epoch": 4.98, + "learning_rate": 1.4818724667454888e-06, + "loss": 1.0961, + "step": 67375 + }, + { + "epoch": 4.98, + "learning_rate": 1.481663526066116e-06, + "loss": 0.9578, + "step": 67376 + }, + { + "epoch": 4.98, + "learning_rate": 1.4814545989392926e-06, + "loss": 1.0538, + "step": 67377 + }, + { + "epoch": 4.98, + "learning_rate": 1.4812456853653545e-06, + "loss": 1.0801, + "step": 67378 + }, + { + "epoch": 4.98, + "learning_rate": 1.4810367853446327e-06, + "loss": 0.9951, + "step": 67379 + }, + { + "epoch": 4.98, + "learning_rate": 1.4808278988774593e-06, + "loss": 0.9336, + "step": 67380 + }, + { + "epoch": 4.98, + "learning_rate": 1.480619025964165e-06, + "loss": 1.0335, + "step": 67381 + }, + { + "epoch": 4.98, + "learning_rate": 1.4804101666050863e-06, + "loss": 0.9262, + "step": 67382 + }, + { + "epoch": 4.98, + "learning_rate": 1.4802013208005539e-06, + "loss": 0.9785, + "step": 67383 + }, + { + "epoch": 4.98, + "learning_rate": 1.4799924885508976e-06, + "loss": 0.9484, + "step": 67384 + }, + { + "epoch": 4.98, + "learning_rate": 1.4797836698564527e-06, + "loss": 0.9258, + "step": 67385 + }, + { + "epoch": 4.98, + "learning_rate": 1.479574864717548e-06, + "loss": 0.9436, + "step": 67386 + }, + { + "epoch": 4.98, + "learning_rate": 1.4793660731345184e-06, + "loss": 0.8899, + "step": 67387 + }, + { + "epoch": 4.98, + "learning_rate": 1.4791572951076972e-06, + "loss": 1.0605, + "step": 67388 + }, + { + "epoch": 4.98, + "learning_rate": 1.4789485306374141e-06, + "loss": 1.07, + "step": 67389 + }, + { + "epoch": 4.98, + "learning_rate": 1.4787397797239988e-06, + "loss": 1.0085, + "step": 67390 + }, + { + "epoch": 4.98, + "learning_rate": 1.47853104236779e-06, + "loss": 1.1179, + "step": 67391 + }, + { + "epoch": 4.98, + "learning_rate": 1.4783223185691186e-06, + "loss": 1.0233, + "step": 67392 + }, + { + "epoch": 4.98, + "learning_rate": 1.4781136083283099e-06, + "loss": 1.0274, + "step": 67393 + }, + { + "epoch": 4.98, + "learning_rate": 1.4779049116457023e-06, + "loss": 0.9835, + "step": 67394 + }, + { + "epoch": 4.98, + "learning_rate": 1.4776962285216246e-06, + "loss": 1.0477, + "step": 67395 + }, + { + "epoch": 4.98, + "learning_rate": 1.4774875589564141e-06, + "loss": 0.9036, + "step": 67396 + }, + { + "epoch": 4.98, + "learning_rate": 1.4772789029503942e-06, + "loss": 1.0507, + "step": 67397 + }, + { + "epoch": 4.98, + "learning_rate": 1.4770702605039045e-06, + "loss": 1.0745, + "step": 67398 + }, + { + "epoch": 4.98, + "learning_rate": 1.4768616316172713e-06, + "loss": 1.0708, + "step": 67399 + }, + { + "epoch": 4.98, + "learning_rate": 1.4766530162908343e-06, + "loss": 0.9863, + "step": 67400 + }, + { + "epoch": 4.98, + "learning_rate": 1.4764444145249158e-06, + "loss": 0.9417, + "step": 67401 + }, + { + "epoch": 4.98, + "learning_rate": 1.476235826319855e-06, + "loss": 1.1038, + "step": 67402 + }, + { + "epoch": 4.98, + "learning_rate": 1.47602725167598e-06, + "loss": 0.9826, + "step": 67403 + }, + { + "epoch": 4.98, + "learning_rate": 1.4758186905936246e-06, + "loss": 1.0204, + "step": 67404 + }, + { + "epoch": 4.98, + "learning_rate": 1.4756101430731196e-06, + "loss": 0.9668, + "step": 67405 + }, + { + "epoch": 4.98, + "learning_rate": 1.4754016091147938e-06, + "loss": 0.9405, + "step": 67406 + }, + { + "epoch": 4.98, + "learning_rate": 1.4751930887189846e-06, + "loss": 0.9938, + "step": 67407 + }, + { + "epoch": 4.98, + "learning_rate": 1.474984581886022e-06, + "loss": 0.9643, + "step": 67408 + }, + { + "epoch": 4.98, + "learning_rate": 1.4747760886162365e-06, + "loss": 0.9903, + "step": 67409 + }, + { + "epoch": 4.98, + "learning_rate": 1.474567608909958e-06, + "loss": 0.9839, + "step": 67410 + }, + { + "epoch": 4.98, + "learning_rate": 1.474359142767522e-06, + "loss": 1.0411, + "step": 67411 + }, + { + "epoch": 4.98, + "learning_rate": 1.4741506901892598e-06, + "loss": 1.0562, + "step": 67412 + }, + { + "epoch": 4.98, + "learning_rate": 1.4739422511755008e-06, + "loss": 0.89, + "step": 67413 + }, + { + "epoch": 4.98, + "learning_rate": 1.4737338257265787e-06, + "loss": 1.1034, + "step": 67414 + }, + { + "epoch": 4.98, + "learning_rate": 1.4735254138428212e-06, + "loss": 0.988, + "step": 67415 + }, + { + "epoch": 4.98, + "learning_rate": 1.4733170155245657e-06, + "loss": 0.9339, + "step": 67416 + }, + { + "epoch": 4.98, + "learning_rate": 1.4731086307721399e-06, + "loss": 0.8976, + "step": 67417 + }, + { + "epoch": 4.98, + "learning_rate": 1.4729002595858765e-06, + "loss": 0.9731, + "step": 67418 + }, + { + "epoch": 4.98, + "learning_rate": 1.4726919019661047e-06, + "loss": 1.0065, + "step": 67419 + }, + { + "epoch": 4.98, + "learning_rate": 1.4724835579131636e-06, + "loss": 1.0206, + "step": 67420 + }, + { + "epoch": 4.98, + "learning_rate": 1.4722752274273754e-06, + "loss": 0.9512, + "step": 67421 + }, + { + "epoch": 4.98, + "learning_rate": 1.4720669105090735e-06, + "loss": 1.0252, + "step": 67422 + }, + { + "epoch": 4.98, + "learning_rate": 1.471858607158594e-06, + "loss": 1.0268, + "step": 67423 + }, + { + "epoch": 4.98, + "learning_rate": 1.4716503173762642e-06, + "loss": 0.9564, + "step": 67424 + }, + { + "epoch": 4.98, + "learning_rate": 1.4714420411624174e-06, + "loss": 0.9732, + "step": 67425 + }, + { + "epoch": 4.98, + "learning_rate": 1.4712337785173825e-06, + "loss": 0.9814, + "step": 67426 + }, + { + "epoch": 4.98, + "learning_rate": 1.4710255294414943e-06, + "loss": 0.8774, + "step": 67427 + }, + { + "epoch": 4.98, + "learning_rate": 1.4708172939350828e-06, + "loss": 1.0807, + "step": 67428 + }, + { + "epoch": 4.98, + "learning_rate": 1.470609071998479e-06, + "loss": 0.9988, + "step": 67429 + }, + { + "epoch": 4.98, + "learning_rate": 1.4704008636320132e-06, + "loss": 0.9657, + "step": 67430 + }, + { + "epoch": 4.98, + "learning_rate": 1.4701926688360158e-06, + "loss": 1.0937, + "step": 67431 + }, + { + "epoch": 4.98, + "learning_rate": 1.4699844876108226e-06, + "loss": 1.0421, + "step": 67432 + }, + { + "epoch": 4.98, + "learning_rate": 1.4697763199567617e-06, + "loss": 0.9058, + "step": 67433 + }, + { + "epoch": 4.98, + "learning_rate": 1.4695681658741646e-06, + "loss": 0.9684, + "step": 67434 + }, + { + "epoch": 4.98, + "learning_rate": 1.46936002536336e-06, + "loss": 1.0085, + "step": 67435 + }, + { + "epoch": 4.98, + "learning_rate": 1.4691518984246844e-06, + "loss": 0.948, + "step": 67436 + }, + { + "epoch": 4.98, + "learning_rate": 1.4689437850584664e-06, + "loss": 1.0047, + "step": 67437 + }, + { + "epoch": 4.98, + "learning_rate": 1.468735685265036e-06, + "loss": 1.1107, + "step": 67438 + }, + { + "epoch": 4.98, + "learning_rate": 1.4685275990447245e-06, + "loss": 1.0116, + "step": 67439 + }, + { + "epoch": 4.98, + "learning_rate": 1.4683195263978623e-06, + "loss": 0.9502, + "step": 67440 + }, + { + "epoch": 4.98, + "learning_rate": 1.4681114673247865e-06, + "loss": 1.016, + "step": 67441 + }, + { + "epoch": 4.98, + "learning_rate": 1.4679034218258181e-06, + "loss": 1.0187, + "step": 67442 + }, + { + "epoch": 4.98, + "learning_rate": 1.4676953899012948e-06, + "loss": 1.013, + "step": 67443 + }, + { + "epoch": 4.98, + "learning_rate": 1.4674873715515448e-06, + "loss": 0.9213, + "step": 67444 + }, + { + "epoch": 4.98, + "learning_rate": 1.4672793667769048e-06, + "loss": 1.0355, + "step": 67445 + }, + { + "epoch": 4.98, + "learning_rate": 1.4670713755776967e-06, + "loss": 0.8811, + "step": 67446 + }, + { + "epoch": 4.98, + "learning_rate": 1.466863397954258e-06, + "loss": 1.0591, + "step": 67447 + }, + { + "epoch": 4.98, + "learning_rate": 1.4666554339069172e-06, + "loss": 1.0589, + "step": 67448 + }, + { + "epoch": 4.98, + "learning_rate": 1.4664474834360064e-06, + "loss": 1.1061, + "step": 67449 + }, + { + "epoch": 4.98, + "learning_rate": 1.4662395465418543e-06, + "loss": 0.9773, + "step": 67450 + }, + { + "epoch": 4.98, + "learning_rate": 1.4660316232247918e-06, + "loss": 1.0021, + "step": 67451 + }, + { + "epoch": 4.98, + "learning_rate": 1.4658237134851527e-06, + "loss": 0.8563, + "step": 67452 + }, + { + "epoch": 4.98, + "learning_rate": 1.465615817323266e-06, + "loss": 0.8891, + "step": 67453 + }, + { + "epoch": 4.98, + "learning_rate": 1.4654079347394633e-06, + "loss": 0.9577, + "step": 67454 + }, + { + "epoch": 4.98, + "learning_rate": 1.4652000657340715e-06, + "loss": 1.1077, + "step": 67455 + }, + { + "epoch": 4.98, + "learning_rate": 1.4649922103074267e-06, + "loss": 0.9631, + "step": 67456 + }, + { + "epoch": 4.98, + "learning_rate": 1.464784368459856e-06, + "loss": 0.9737, + "step": 67457 + }, + { + "epoch": 4.98, + "learning_rate": 1.4645765401916922e-06, + "loss": 0.9968, + "step": 67458 + }, + { + "epoch": 4.98, + "learning_rate": 1.4643687255032657e-06, + "loss": 1.003, + "step": 67459 + }, + { + "epoch": 4.98, + "learning_rate": 1.4641609243949029e-06, + "loss": 0.9746, + "step": 67460 + }, + { + "epoch": 4.98, + "learning_rate": 1.4639531368669413e-06, + "loss": 0.9314, + "step": 67461 + }, + { + "epoch": 4.98, + "learning_rate": 1.4637453629197074e-06, + "loss": 0.9613, + "step": 67462 + }, + { + "epoch": 4.98, + "learning_rate": 1.4635376025535331e-06, + "loss": 0.8517, + "step": 67463 + }, + { + "epoch": 4.98, + "learning_rate": 1.4633298557687447e-06, + "loss": 0.9992, + "step": 67464 + }, + { + "epoch": 4.98, + "learning_rate": 1.4631221225656823e-06, + "loss": 0.9942, + "step": 67465 + }, + { + "epoch": 4.99, + "learning_rate": 1.462914402944665e-06, + "loss": 1.0881, + "step": 67466 + }, + { + "epoch": 4.99, + "learning_rate": 1.4627066969060321e-06, + "loss": 1.1209, + "step": 67467 + }, + { + "epoch": 4.99, + "learning_rate": 1.4624990044501097e-06, + "loss": 1.0356, + "step": 67468 + }, + { + "epoch": 4.99, + "learning_rate": 1.4622913255772298e-06, + "loss": 0.9808, + "step": 67469 + }, + { + "epoch": 4.99, + "learning_rate": 1.462083660287722e-06, + "loss": 0.9268, + "step": 67470 + }, + { + "epoch": 4.99, + "learning_rate": 1.461876008581915e-06, + "loss": 0.8839, + "step": 67471 + }, + { + "epoch": 4.99, + "learning_rate": 1.4616683704601432e-06, + "loss": 1.0097, + "step": 67472 + }, + { + "epoch": 4.99, + "learning_rate": 1.4614607459227347e-06, + "loss": 1.089, + "step": 67473 + }, + { + "epoch": 4.99, + "learning_rate": 1.46125313497002e-06, + "loss": 0.8886, + "step": 67474 + }, + { + "epoch": 4.99, + "learning_rate": 1.461045537602327e-06, + "loss": 1.0216, + "step": 67475 + }, + { + "epoch": 4.99, + "learning_rate": 1.4608379538199912e-06, + "loss": 1.0093, + "step": 67476 + }, + { + "epoch": 4.99, + "learning_rate": 1.4606303836233392e-06, + "loss": 0.9524, + "step": 67477 + }, + { + "epoch": 4.99, + "learning_rate": 1.460422827012703e-06, + "loss": 1.1423, + "step": 67478 + }, + { + "epoch": 4.99, + "learning_rate": 1.4602152839884109e-06, + "loss": 1.1019, + "step": 67479 + }, + { + "epoch": 4.99, + "learning_rate": 1.4600077545507929e-06, + "loss": 1.0701, + "step": 67480 + }, + { + "epoch": 4.99, + "learning_rate": 1.459800238700182e-06, + "loss": 0.9199, + "step": 67481 + }, + { + "epoch": 4.99, + "learning_rate": 1.4595927364369067e-06, + "loss": 0.9631, + "step": 67482 + }, + { + "epoch": 4.99, + "learning_rate": 1.4593852477612979e-06, + "loss": 1.1965, + "step": 67483 + }, + { + "epoch": 4.99, + "learning_rate": 1.4591777726736821e-06, + "loss": 1.0947, + "step": 67484 + }, + { + "epoch": 4.99, + "learning_rate": 1.4589703111743947e-06, + "loss": 0.9724, + "step": 67485 + }, + { + "epoch": 4.99, + "learning_rate": 1.4587628632637652e-06, + "loss": 0.9599, + "step": 67486 + }, + { + "epoch": 4.99, + "learning_rate": 1.4585554289421177e-06, + "loss": 1.1429, + "step": 67487 + }, + { + "epoch": 4.99, + "learning_rate": 1.4583480082097878e-06, + "loss": 0.9477, + "step": 67488 + }, + { + "epoch": 4.99, + "learning_rate": 1.4581406010671029e-06, + "loss": 0.9156, + "step": 67489 + }, + { + "epoch": 4.99, + "learning_rate": 1.4579332075143982e-06, + "loss": 0.9991, + "step": 67490 + }, + { + "epoch": 4.99, + "learning_rate": 1.4577258275519945e-06, + "loss": 0.9829, + "step": 67491 + }, + { + "epoch": 4.99, + "learning_rate": 1.4575184611802295e-06, + "loss": 0.9425, + "step": 67492 + }, + { + "epoch": 4.99, + "learning_rate": 1.4573111083994284e-06, + "loss": 0.9612, + "step": 67493 + }, + { + "epoch": 4.99, + "learning_rate": 1.4571037692099265e-06, + "loss": 1.0584, + "step": 67494 + }, + { + "epoch": 4.99, + "learning_rate": 1.4568964436120492e-06, + "loss": 1.0381, + "step": 67495 + }, + { + "epoch": 4.99, + "learning_rate": 1.456689131606125e-06, + "loss": 0.9797, + "step": 67496 + }, + { + "epoch": 4.99, + "learning_rate": 1.4564818331924878e-06, + "loss": 1.0742, + "step": 67497 + }, + { + "epoch": 4.99, + "learning_rate": 1.4562745483714658e-06, + "loss": 1.0205, + "step": 67498 + }, + { + "epoch": 4.99, + "learning_rate": 1.4560672771433892e-06, + "loss": 0.9894, + "step": 67499 + }, + { + "epoch": 4.99, + "learning_rate": 1.4558600195085849e-06, + "loss": 1.0062, + "step": 67500 + }, + { + "epoch": 4.99, + "learning_rate": 1.4556527754673877e-06, + "loss": 1.0002, + "step": 67501 + }, + { + "epoch": 4.99, + "learning_rate": 1.4554455450201245e-06, + "loss": 1.0027, + "step": 67502 + }, + { + "epoch": 4.99, + "learning_rate": 1.4552383281671244e-06, + "loss": 0.8925, + "step": 67503 + }, + { + "epoch": 4.99, + "learning_rate": 1.455031124908719e-06, + "loss": 0.9587, + "step": 67504 + }, + { + "epoch": 4.99, + "learning_rate": 1.4548239352452342e-06, + "loss": 1.1122, + "step": 67505 + }, + { + "epoch": 4.99, + "learning_rate": 1.4546167591770044e-06, + "loss": 1.0564, + "step": 67506 + }, + { + "epoch": 4.99, + "learning_rate": 1.4544095967043569e-06, + "loss": 0.9508, + "step": 67507 + }, + { + "epoch": 4.99, + "learning_rate": 1.4542024478276218e-06, + "loss": 0.9747, + "step": 67508 + }, + { + "epoch": 4.99, + "learning_rate": 1.4539953125471262e-06, + "loss": 1.0952, + "step": 67509 + }, + { + "epoch": 4.99, + "learning_rate": 1.4537881908632057e-06, + "loss": 1.1925, + "step": 67510 + }, + { + "epoch": 4.99, + "learning_rate": 1.453581082776182e-06, + "loss": 1.0549, + "step": 67511 + }, + { + "epoch": 4.99, + "learning_rate": 1.4533739882863907e-06, + "loss": 1.0551, + "step": 67512 + }, + { + "epoch": 4.99, + "learning_rate": 1.4531669073941556e-06, + "loss": 1.0027, + "step": 67513 + }, + { + "epoch": 4.99, + "learning_rate": 1.4529598400998157e-06, + "loss": 0.8712, + "step": 67514 + }, + { + "epoch": 4.99, + "learning_rate": 1.4527527864036916e-06, + "loss": 0.9788, + "step": 67515 + }, + { + "epoch": 4.99, + "learning_rate": 1.4525457463061133e-06, + "loss": 0.962, + "step": 67516 + }, + { + "epoch": 4.99, + "learning_rate": 1.4523387198074147e-06, + "loss": 0.9983, + "step": 67517 + }, + { + "epoch": 4.99, + "learning_rate": 1.4521317069079232e-06, + "loss": 0.9971, + "step": 67518 + }, + { + "epoch": 4.99, + "learning_rate": 1.4519247076079678e-06, + "loss": 1.0266, + "step": 67519 + }, + { + "epoch": 4.99, + "learning_rate": 1.451717721907875e-06, + "loss": 1.0235, + "step": 67520 + }, + { + "epoch": 4.99, + "learning_rate": 1.4515107498079796e-06, + "loss": 1.0471, + "step": 67521 + }, + { + "epoch": 4.99, + "learning_rate": 1.451303791308608e-06, + "loss": 1.0524, + "step": 67522 + }, + { + "epoch": 4.99, + "learning_rate": 1.4510968464100905e-06, + "loss": 0.9848, + "step": 67523 + }, + { + "epoch": 4.99, + "learning_rate": 1.4508899151127554e-06, + "loss": 0.8849, + "step": 67524 + }, + { + "epoch": 4.99, + "learning_rate": 1.4506829974169289e-06, + "loss": 1.0115, + "step": 67525 + }, + { + "epoch": 4.99, + "learning_rate": 1.4504760933229467e-06, + "loss": 1.0208, + "step": 67526 + }, + { + "epoch": 4.99, + "learning_rate": 1.450269202831135e-06, + "loss": 1.0124, + "step": 67527 + }, + { + "epoch": 4.99, + "learning_rate": 1.4500623259418222e-06, + "loss": 0.9541, + "step": 67528 + }, + { + "epoch": 4.99, + "learning_rate": 1.449855462655335e-06, + "loss": 0.974, + "step": 67529 + }, + { + "epoch": 4.99, + "learning_rate": 1.4496486129720089e-06, + "loss": 0.9857, + "step": 67530 + }, + { + "epoch": 4.99, + "learning_rate": 1.4494417768921688e-06, + "loss": 0.9959, + "step": 67531 + }, + { + "epoch": 4.99, + "learning_rate": 1.4492349544161443e-06, + "loss": 1.0173, + "step": 67532 + }, + { + "epoch": 4.99, + "learning_rate": 1.4490281455442646e-06, + "loss": 1.0561, + "step": 67533 + }, + { + "epoch": 4.99, + "learning_rate": 1.4488213502768577e-06, + "loss": 1.0859, + "step": 67534 + }, + { + "epoch": 4.99, + "learning_rate": 1.4486145686142572e-06, + "loss": 1.0703, + "step": 67535 + }, + { + "epoch": 4.99, + "learning_rate": 1.4484078005567837e-06, + "loss": 1.0215, + "step": 67536 + }, + { + "epoch": 4.99, + "learning_rate": 1.4482010461047745e-06, + "loss": 1.0689, + "step": 67537 + }, + { + "epoch": 4.99, + "learning_rate": 1.447994305258552e-06, + "loss": 1.1052, + "step": 67538 + }, + { + "epoch": 4.99, + "learning_rate": 1.4477875780184526e-06, + "loss": 0.9425, + "step": 67539 + }, + { + "epoch": 4.99, + "learning_rate": 1.4475808643847966e-06, + "loss": 1.0278, + "step": 67540 + }, + { + "epoch": 4.99, + "learning_rate": 1.4473741643579186e-06, + "loss": 0.9407, + "step": 67541 + }, + { + "epoch": 4.99, + "learning_rate": 1.4471674779381472e-06, + "loss": 1.0011, + "step": 67542 + }, + { + "epoch": 4.99, + "learning_rate": 1.4469608051258088e-06, + "loss": 0.9171, + "step": 67543 + }, + { + "epoch": 4.99, + "learning_rate": 1.4467541459212342e-06, + "loss": 1.0136, + "step": 67544 + }, + { + "epoch": 4.99, + "learning_rate": 1.4465475003247486e-06, + "loss": 1.0568, + "step": 67545 + }, + { + "epoch": 4.99, + "learning_rate": 1.4463408683366863e-06, + "loss": 0.958, + "step": 67546 + }, + { + "epoch": 4.99, + "learning_rate": 1.4461342499573739e-06, + "loss": 0.9713, + "step": 67547 + }, + { + "epoch": 4.99, + "learning_rate": 1.4459276451871384e-06, + "loss": 1.0954, + "step": 67548 + }, + { + "epoch": 4.99, + "learning_rate": 1.445721054026308e-06, + "loss": 0.9174, + "step": 67549 + }, + { + "epoch": 4.99, + "learning_rate": 1.4455144764752149e-06, + "loss": 1.0601, + "step": 67550 + }, + { + "epoch": 4.99, + "learning_rate": 1.4453079125341873e-06, + "loss": 0.8877, + "step": 67551 + }, + { + "epoch": 4.99, + "learning_rate": 1.4451013622035492e-06, + "loss": 0.9921, + "step": 67552 + }, + { + "epoch": 4.99, + "learning_rate": 1.4448948254836347e-06, + "loss": 0.9789, + "step": 67553 + }, + { + "epoch": 4.99, + "learning_rate": 1.444688302374767e-06, + "loss": 1.0002, + "step": 67554 + }, + { + "epoch": 4.99, + "learning_rate": 1.4444817928772825e-06, + "loss": 1.0446, + "step": 67555 + }, + { + "epoch": 4.99, + "learning_rate": 1.4442752969915008e-06, + "loss": 1.0156, + "step": 67556 + }, + { + "epoch": 4.99, + "learning_rate": 1.4440688147177573e-06, + "loss": 1.0505, + "step": 67557 + }, + { + "epoch": 4.99, + "learning_rate": 1.443862346056375e-06, + "loss": 1.079, + "step": 67558 + }, + { + "epoch": 4.99, + "learning_rate": 1.4436558910076902e-06, + "loss": 0.952, + "step": 67559 + }, + { + "epoch": 4.99, + "learning_rate": 1.443449449572022e-06, + "loss": 1.0337, + "step": 67560 + }, + { + "epoch": 4.99, + "learning_rate": 1.4432430217497052e-06, + "loss": 0.9947, + "step": 67561 + }, + { + "epoch": 4.99, + "learning_rate": 1.4430366075410661e-06, + "loss": 1.0324, + "step": 67562 + }, + { + "epoch": 4.99, + "learning_rate": 1.442830206946434e-06, + "loss": 1.03, + "step": 67563 + }, + { + "epoch": 4.99, + "learning_rate": 1.4426238199661369e-06, + "loss": 1.0079, + "step": 67564 + }, + { + "epoch": 4.99, + "learning_rate": 1.4424174466004993e-06, + "loss": 1.0735, + "step": 67565 + }, + { + "epoch": 4.99, + "learning_rate": 1.4422110868498562e-06, + "loss": 0.9776, + "step": 67566 + }, + { + "epoch": 4.99, + "learning_rate": 1.4420047407145332e-06, + "loss": 1.0378, + "step": 67567 + }, + { + "epoch": 4.99, + "learning_rate": 1.4417984081948577e-06, + "loss": 0.925, + "step": 67568 + }, + { + "epoch": 4.99, + "learning_rate": 1.441592089291156e-06, + "loss": 0.9374, + "step": 67569 + }, + { + "epoch": 4.99, + "learning_rate": 1.4413857840037614e-06, + "loss": 0.9628, + "step": 67570 + }, + { + "epoch": 4.99, + "learning_rate": 1.4411794923330002e-06, + "loss": 0.9208, + "step": 67571 + }, + { + "epoch": 4.99, + "learning_rate": 1.4409732142791989e-06, + "loss": 1.1207, + "step": 67572 + }, + { + "epoch": 4.99, + "learning_rate": 1.440766949842687e-06, + "loss": 0.9457, + "step": 67573 + }, + { + "epoch": 4.99, + "learning_rate": 1.440560699023791e-06, + "loss": 1.0463, + "step": 67574 + }, + { + "epoch": 4.99, + "learning_rate": 1.4403544618228416e-06, + "loss": 0.9954, + "step": 67575 + }, + { + "epoch": 4.99, + "learning_rate": 1.4401482382401665e-06, + "loss": 1.0277, + "step": 67576 + }, + { + "epoch": 4.99, + "learning_rate": 1.4399420282760923e-06, + "loss": 0.9986, + "step": 67577 + }, + { + "epoch": 4.99, + "learning_rate": 1.4397358319309462e-06, + "loss": 1.0173, + "step": 67578 + }, + { + "epoch": 4.99, + "learning_rate": 1.4395296492050603e-06, + "loss": 0.9955, + "step": 67579 + }, + { + "epoch": 4.99, + "learning_rate": 1.4393234800987621e-06, + "loss": 1.0004, + "step": 67580 + }, + { + "epoch": 4.99, + "learning_rate": 1.4391173246123736e-06, + "loss": 0.9355, + "step": 67581 + }, + { + "epoch": 4.99, + "learning_rate": 1.4389111827462288e-06, + "loss": 1.0184, + "step": 67582 + }, + { + "epoch": 4.99, + "learning_rate": 1.438705054500652e-06, + "loss": 0.9385, + "step": 67583 + }, + { + "epoch": 4.99, + "learning_rate": 1.4384989398759775e-06, + "loss": 1.0638, + "step": 67584 + }, + { + "epoch": 4.99, + "learning_rate": 1.4382928388725237e-06, + "loss": 0.984, + "step": 67585 + }, + { + "epoch": 4.99, + "learning_rate": 1.438086751490626e-06, + "loss": 1.005, + "step": 67586 + }, + { + "epoch": 4.99, + "learning_rate": 1.4378806777306099e-06, + "loss": 1.0649, + "step": 67587 + }, + { + "epoch": 4.99, + "learning_rate": 1.4376746175928025e-06, + "loss": 1.0342, + "step": 67588 + }, + { + "epoch": 4.99, + "learning_rate": 1.4374685710775328e-06, + "loss": 0.9626, + "step": 67589 + }, + { + "epoch": 4.99, + "learning_rate": 1.4372625381851257e-06, + "loss": 0.9427, + "step": 67590 + }, + { + "epoch": 4.99, + "learning_rate": 1.4370565189159147e-06, + "loss": 0.8304, + "step": 67591 + }, + { + "epoch": 4.99, + "learning_rate": 1.4368505132702227e-06, + "loss": 1.054, + "step": 67592 + }, + { + "epoch": 4.99, + "learning_rate": 1.4366445212483803e-06, + "loss": 0.9856, + "step": 67593 + }, + { + "epoch": 4.99, + "learning_rate": 1.436438542850711e-06, + "loss": 1.0319, + "step": 67594 + }, + { + "epoch": 4.99, + "learning_rate": 1.4362325780775477e-06, + "loss": 1.0897, + "step": 67595 + }, + { + "epoch": 4.99, + "learning_rate": 1.4360266269292156e-06, + "loss": 0.9404, + "step": 67596 + }, + { + "epoch": 4.99, + "learning_rate": 1.4358206894060433e-06, + "loss": 0.9936, + "step": 67597 + }, + { + "epoch": 4.99, + "learning_rate": 1.4356147655083574e-06, + "loss": 1.0558, + "step": 67598 + }, + { + "epoch": 4.99, + "learning_rate": 1.435408855236483e-06, + "loss": 1.0017, + "step": 67599 + }, + { + "epoch": 4.99, + "learning_rate": 1.4352029585907557e-06, + "loss": 0.9272, + "step": 67600 + }, + { + "epoch": 4.99, + "learning_rate": 1.4349970755714937e-06, + "loss": 0.9324, + "step": 67601 + }, + { + "epoch": 5.0, + "learning_rate": 1.4347912061790302e-06, + "loss": 1.0801, + "step": 67602 + }, + { + "epoch": 5.0, + "learning_rate": 1.4345853504136898e-06, + "loss": 0.9623, + "step": 67603 + }, + { + "epoch": 5.0, + "learning_rate": 1.4343795082758062e-06, + "loss": 1.1967, + "step": 67604 + }, + { + "epoch": 5.0, + "learning_rate": 1.434173679765698e-06, + "loss": 1.123, + "step": 67605 + }, + { + "epoch": 5.0, + "learning_rate": 1.4339678648836985e-06, + "loss": 1.0378, + "step": 67606 + }, + { + "epoch": 5.0, + "learning_rate": 1.433762063630133e-06, + "loss": 1.025, + "step": 67607 + }, + { + "epoch": 5.0, + "learning_rate": 1.43355627600533e-06, + "loss": 0.9774, + "step": 67608 + }, + { + "epoch": 5.0, + "learning_rate": 1.4333505020096171e-06, + "loss": 1.0255, + "step": 67609 + }, + { + "epoch": 5.0, + "learning_rate": 1.4331447416433186e-06, + "loss": 1.0069, + "step": 67610 + }, + { + "epoch": 5.0, + "learning_rate": 1.4329389949067662e-06, + "loss": 1.0547, + "step": 67611 + }, + { + "epoch": 5.0, + "learning_rate": 1.4327332618002843e-06, + "loss": 0.8866, + "step": 67612 + }, + { + "epoch": 5.0, + "learning_rate": 1.4325275423242025e-06, + "loss": 1.0142, + "step": 67613 + }, + { + "epoch": 5.0, + "learning_rate": 1.4323218364788439e-06, + "loss": 0.985, + "step": 67614 + }, + { + "epoch": 5.0, + "learning_rate": 1.4321161442645414e-06, + "loss": 0.8674, + "step": 67615 + }, + { + "epoch": 5.0, + "learning_rate": 1.4319104656816196e-06, + "loss": 1.0055, + "step": 67616 + }, + { + "epoch": 5.0, + "learning_rate": 1.4317048007304057e-06, + "loss": 0.9793, + "step": 67617 + }, + { + "epoch": 5.0, + "learning_rate": 1.4314991494112262e-06, + "loss": 1.03, + "step": 67618 + }, + { + "epoch": 5.0, + "learning_rate": 1.4312935117244065e-06, + "loss": 1.0022, + "step": 67619 + }, + { + "epoch": 5.0, + "learning_rate": 1.4310878876702793e-06, + "loss": 0.9409, + "step": 67620 + }, + { + "epoch": 5.0, + "learning_rate": 1.4308822772491682e-06, + "loss": 0.9155, + "step": 67621 + }, + { + "epoch": 5.0, + "learning_rate": 1.4306766804614013e-06, + "loss": 0.8894, + "step": 67622 + }, + { + "epoch": 5.0, + "learning_rate": 1.4304710973073023e-06, + "loss": 0.9392, + "step": 67623 + }, + { + "epoch": 5.0, + "learning_rate": 1.4302655277872035e-06, + "loss": 1.0119, + "step": 67624 + }, + { + "epoch": 5.0, + "learning_rate": 1.4300599719014307e-06, + "loss": 0.985, + "step": 67625 + }, + { + "epoch": 5.0, + "learning_rate": 1.4298544296503092e-06, + "loss": 0.9673, + "step": 67626 + }, + { + "epoch": 5.0, + "learning_rate": 1.4296489010341662e-06, + "loss": 0.9396, + "step": 67627 + }, + { + "epoch": 5.0, + "learning_rate": 1.4294433860533274e-06, + "loss": 0.9375, + "step": 67628 + }, + { + "epoch": 5.0, + "learning_rate": 1.4292378847081257e-06, + "loss": 0.9567, + "step": 67629 + }, + { + "epoch": 5.0, + "learning_rate": 1.4290323969988795e-06, + "loss": 1.0144, + "step": 67630 + }, + { + "epoch": 5.0, + "learning_rate": 1.4288269229259222e-06, + "loss": 1.0006, + "step": 67631 + }, + { + "epoch": 5.0, + "learning_rate": 1.4286214624895788e-06, + "loss": 1.0015, + "step": 67632 + }, + { + "epoch": 5.0, + "learning_rate": 1.4284160156901772e-06, + "loss": 0.9764, + "step": 67633 + }, + { + "epoch": 5.0, + "learning_rate": 1.428210582528039e-06, + "loss": 1.1028, + "step": 67634 + }, + { + "epoch": 5.0, + "learning_rate": 1.4280051630034986e-06, + "loss": 1.0577, + "step": 67635 + }, + { + "epoch": 5.0, + "learning_rate": 1.4277997571168789e-06, + "loss": 0.8395, + "step": 67636 + }, + { + "epoch": 5.0, + "learning_rate": 1.4275943648685065e-06, + "loss": 0.9257, + "step": 67637 + }, + { + "epoch": 5.0, + "learning_rate": 1.4273889862587099e-06, + "loss": 0.9683, + "step": 67638 + }, + { + "epoch": 5.0, + "learning_rate": 1.4271836212878121e-06, + "loss": 1.0028, + "step": 67639 + }, + { + "epoch": 5.0, + "learning_rate": 1.426978269956144e-06, + "loss": 1.0409, + "step": 67640 + }, + { + "epoch": 5.0, + "learning_rate": 1.4267729322640321e-06, + "loss": 1.0885, + "step": 67641 + }, + { + "epoch": 5.0, + "learning_rate": 1.4265676082118006e-06, + "loss": 1.0572, + "step": 67642 + }, + { + "epoch": 5.0, + "learning_rate": 1.426362297799776e-06, + "loss": 0.9588, + "step": 67643 + }, + { + "epoch": 5.0, + "learning_rate": 1.4261570010282888e-06, + "loss": 0.8408, + "step": 67644 + }, + { + "epoch": 5.0, + "learning_rate": 1.4259517178976644e-06, + "loss": 1.0102, + "step": 67645 + }, + { + "epoch": 5.0, + "learning_rate": 1.425746448408224e-06, + "loss": 1.0293, + "step": 67646 + }, + { + "epoch": 5.0, + "learning_rate": 1.4255411925603014e-06, + "loss": 0.9741, + "step": 67647 + }, + { + "epoch": 5.0, + "learning_rate": 1.4253359503542164e-06, + "loss": 0.967, + "step": 67648 + }, + { + "epoch": 5.0, + "learning_rate": 1.4251307217903053e-06, + "loss": 0.9404, + "step": 67649 + }, + { + "epoch": 5.0, + "learning_rate": 1.4249255068688827e-06, + "loss": 0.9386, + "step": 67650 + }, + { + "epoch": 5.0, + "learning_rate": 1.4247203055902837e-06, + "loss": 1.0275, + "step": 67651 + }, + { + "epoch": 5.0, + "learning_rate": 1.42451511795483e-06, + "loss": 1.02, + "step": 67652 + }, + { + "epoch": 5.0, + "learning_rate": 1.4243099439628537e-06, + "loss": 1.0555, + "step": 67653 + }, + { + "epoch": 5.0, + "learning_rate": 1.424104783614676e-06, + "loss": 0.9649, + "step": 67654 + }, + { + "epoch": 5.0, + "learning_rate": 1.4238996369106229e-06, + "loss": 1.0845, + "step": 67655 + }, + { + "epoch": 5.0, + "learning_rate": 1.423694503851024e-06, + "loss": 1.0288, + "step": 67656 + }, + { + "epoch": 5.0, + "learning_rate": 1.4234893844362063e-06, + "loss": 1.0456, + "step": 67657 + }, + { + "epoch": 5.0, + "learning_rate": 1.4232842786664925e-06, + "loss": 1.0077, + "step": 67658 + }, + { + "epoch": 5.0, + "learning_rate": 1.423079186542209e-06, + "loss": 0.9769, + "step": 67659 + }, + { + "epoch": 5.0, + "learning_rate": 1.4228741080636866e-06, + "loss": 1.0457, + "step": 67660 + }, + { + "epoch": 5.0, + "learning_rate": 1.4226690432312474e-06, + "loss": 1.0435, + "step": 67661 + }, + { + "epoch": 5.0, + "learning_rate": 1.4224639920452199e-06, + "loss": 0.9595, + "step": 67662 + }, + { + "epoch": 5.0, + "learning_rate": 1.4222589545059295e-06, + "loss": 1.0054, + "step": 67663 + }, + { + "epoch": 5.0, + "learning_rate": 1.4220539306137004e-06, + "loss": 0.976, + "step": 67664 + }, + { + "epoch": 5.0, + "learning_rate": 1.4218489203688624e-06, + "loss": 0.9611, + "step": 67665 + }, + { + "epoch": 5.0, + "learning_rate": 1.4216439237717405e-06, + "loss": 0.9824, + "step": 67666 + }, + { + "epoch": 5.0, + "learning_rate": 1.4214389408226603e-06, + "loss": 0.9964, + "step": 67667 + }, + { + "epoch": 5.0, + "learning_rate": 1.4212339715219459e-06, + "loss": 0.9803, + "step": 67668 + }, + { + "epoch": 5.0, + "learning_rate": 1.421029015869927e-06, + "loss": 0.9376, + "step": 67669 + }, + { + "epoch": 5.0, + "learning_rate": 1.4208240738669287e-06, + "loss": 1.0462, + "step": 67670 + }, + { + "epoch": 5.0, + "learning_rate": 1.4206191455132767e-06, + "loss": 1.0554, + "step": 67671 + }, + { + "epoch": 5.0, + "learning_rate": 1.420414230809295e-06, + "loss": 0.9683, + "step": 67672 + }, + { + "epoch": 5.0, + "learning_rate": 1.4202093297553154e-06, + "loss": 0.9339, + "step": 67673 + }, + { + "epoch": 5.0, + "learning_rate": 1.420004442351658e-06, + "loss": 1.0401, + "step": 67674 + }, + { + "epoch": 5.0, + "learning_rate": 1.4197995685986489e-06, + "loss": 0.9553, + "step": 67675 + }, + { + "epoch": 5.0, + "learning_rate": 1.4195947084966167e-06, + "loss": 0.9547, + "step": 67676 + }, + { + "epoch": 5.0, + "learning_rate": 1.419389862045888e-06, + "loss": 0.9158, + "step": 67677 + }, + { + "epoch": 5.0, + "learning_rate": 1.4191850292467868e-06, + "loss": 1.0902, + "step": 67678 + }, + { + "epoch": 5.0, + "learning_rate": 1.4189802100996374e-06, + "loss": 1.0373, + "step": 67679 + }, + { + "epoch": 5.0, + "learning_rate": 1.4187754046047697e-06, + "loss": 1.094, + "step": 67680 + }, + { + "epoch": 5.0, + "learning_rate": 1.4185706127625077e-06, + "loss": 0.9112, + "step": 67681 + }, + { + "epoch": 5.0, + "learning_rate": 1.4183658345731766e-06, + "loss": 0.9285, + "step": 67682 + }, + { + "epoch": 5.0, + "learning_rate": 1.4181610700371029e-06, + "loss": 1.092, + "step": 67683 + }, + { + "epoch": 5.0, + "learning_rate": 1.4179563191546098e-06, + "loss": 0.9403, + "step": 67684 + }, + { + "epoch": 5.0, + "learning_rate": 1.4177515819260279e-06, + "loss": 1.0596, + "step": 67685 + }, + { + "epoch": 5.0, + "learning_rate": 1.4175468583516806e-06, + "loss": 1.0265, + "step": 67686 + }, + { + "epoch": 5.0, + "learning_rate": 1.4173421484318938e-06, + "loss": 1.0601, + "step": 67687 + }, + { + "epoch": 5.0, + "learning_rate": 1.41713745216699e-06, + "loss": 1.081, + "step": 67688 + }, + { + "epoch": 5.0, + "learning_rate": 1.4169327695573e-06, + "loss": 0.9396, + "step": 67689 + }, + { + "epoch": 5.0, + "learning_rate": 1.4167281006031474e-06, + "loss": 1.0177, + "step": 67690 + }, + { + "epoch": 5.0, + "learning_rate": 1.416523445304857e-06, + "loss": 0.9431, + "step": 67691 + }, + { + "epoch": 5.0, + "learning_rate": 1.416318803662755e-06, + "loss": 0.97, + "step": 67692 + }, + { + "epoch": 5.0, + "learning_rate": 1.4161141756771657e-06, + "loss": 0.8534, + "step": 67693 + }, + { + "epoch": 5.0, + "learning_rate": 1.4159095613484197e-06, + "loss": 1.0373, + "step": 67694 + }, + { + "epoch": 5.0, + "learning_rate": 1.4157049606768346e-06, + "loss": 0.9674, + "step": 67695 + }, + { + "epoch": 5.0, + "learning_rate": 1.4155003736627415e-06, + "loss": 0.9594, + "step": 67696 + }, + { + "epoch": 5.0, + "learning_rate": 1.4152958003064631e-06, + "loss": 1.0114, + "step": 67697 + }, + { + "epoch": 5.0, + "learning_rate": 1.4150912406083306e-06, + "loss": 1.0304, + "step": 67698 + }, + { + "epoch": 5.0, + "learning_rate": 1.41488669456866e-06, + "loss": 1.0721, + "step": 67699 + }, + { + "epoch": 5.0, + "learning_rate": 1.414682162187785e-06, + "loss": 1.0525, + "step": 67700 + }, + { + "epoch": 5.0, + "learning_rate": 1.4144776434660279e-06, + "loss": 1.1715, + "step": 67701 + }, + { + "epoch": 5.0, + "learning_rate": 1.4142731384037134e-06, + "loss": 0.998, + "step": 67702 + }, + { + "epoch": 5.0, + "learning_rate": 1.4140686470011677e-06, + "loss": 1.0266, + "step": 67703 + }, + { + "epoch": 5.0, + "learning_rate": 1.413864169258714e-06, + "loss": 1.0392, + "step": 67704 + }, + { + "epoch": 5.0, + "learning_rate": 1.4136597051766809e-06, + "loss": 1.1273, + "step": 67705 + }, + { + "epoch": 5.0, + "learning_rate": 1.4134552547553937e-06, + "loss": 0.8881, + "step": 67706 + }, + { + "epoch": 5.0, + "learning_rate": 1.4132508179951753e-06, + "loss": 0.9781, + "step": 67707 + }, + { + "epoch": 5.0, + "learning_rate": 1.4130463948963502e-06, + "loss": 0.98, + "step": 67708 + }, + { + "epoch": 5.0, + "learning_rate": 1.4128419854592469e-06, + "loss": 1.0283, + "step": 67709 + }, + { + "epoch": 5.0, + "learning_rate": 1.4126375896841904e-06, + "loss": 1.0537, + "step": 67710 + }, + { + "epoch": 5.0, + "learning_rate": 1.412433207571503e-06, + "loss": 0.9478, + "step": 67711 + }, + { + "epoch": 5.0, + "learning_rate": 1.4122288391215133e-06, + "loss": 1.0302, + "step": 67712 + }, + { + "epoch": 5.0, + "learning_rate": 1.412024484334541e-06, + "loss": 0.9647, + "step": 67713 + }, + { + "epoch": 5.0, + "learning_rate": 1.411820143210918e-06, + "loss": 0.9298, + "step": 67714 + }, + { + "epoch": 5.0, + "learning_rate": 1.4116158157509663e-06, + "loss": 1.0385, + "step": 67715 + }, + { + "epoch": 5.0, + "learning_rate": 1.41141150195501e-06, + "loss": 1.0043, + "step": 67716 + }, + { + "epoch": 5.0, + "learning_rate": 1.4112072018233735e-06, + "loss": 0.967, + "step": 67717 + }, + { + "epoch": 5.0, + "learning_rate": 1.4110029153563887e-06, + "loss": 1.0616, + "step": 67718 + }, + { + "epoch": 5.0, + "learning_rate": 1.410798642554372e-06, + "loss": 0.9367, + "step": 67719 + }, + { + "epoch": 5.0, + "learning_rate": 1.4105943834176494e-06, + "loss": 1.0485, + "step": 67720 + }, + { + "epoch": 5.0, + "learning_rate": 1.4103901379465513e-06, + "loss": 0.8833, + "step": 67721 + }, + { + "epoch": 5.0, + "learning_rate": 1.4101859061413992e-06, + "loss": 1.0449, + "step": 67722 + }, + { + "epoch": 5.0, + "learning_rate": 1.4099816880025186e-06, + "loss": 0.9873, + "step": 67723 + }, + { + "epoch": 5.0, + "learning_rate": 1.4097774835302314e-06, + "loss": 1.0138, + "step": 67724 + }, + { + "epoch": 5.0, + "learning_rate": 1.4095732927248685e-06, + "loss": 0.9603, + "step": 67725 + }, + { + "epoch": 5.0, + "learning_rate": 1.4093691155867506e-06, + "loss": 0.9748, + "step": 67726 + }, + { + "epoch": 5.0, + "learning_rate": 1.4091649521162044e-06, + "loss": 0.9977, + "step": 67727 + }, + { + "epoch": 5.0, + "learning_rate": 1.4089608023135516e-06, + "loss": 1.0993, + "step": 67728 + }, + { + "epoch": 5.0, + "learning_rate": 1.4087566661791207e-06, + "loss": 1.0109, + "step": 67729 + }, + { + "epoch": 5.0, + "learning_rate": 1.4085525437132353e-06, + "loss": 1.0559, + "step": 67730 + }, + { + "epoch": 5.0, + "learning_rate": 1.4083484349162202e-06, + "loss": 0.9764, + "step": 67731 + }, + { + "epoch": 5.0, + "learning_rate": 1.4081443397883998e-06, + "loss": 0.9593, + "step": 67732 + }, + { + "epoch": 5.0, + "learning_rate": 1.407940258330096e-06, + "loss": 1.0702, + "step": 67733 + }, + { + "epoch": 5.0, + "learning_rate": 1.4077361905416397e-06, + "loss": 1.0164, + "step": 67734 + }, + { + "epoch": 5.0, + "learning_rate": 1.407532136423352e-06, + "loss": 0.9414, + "step": 67735 + }, + { + "epoch": 5.0, + "learning_rate": 1.4073280959755564e-06, + "loss": 0.9707, + "step": 67736 + }, + { + "epoch": 5.01, + "learning_rate": 1.4071240691985778e-06, + "loss": 1.0528, + "step": 67737 + }, + { + "epoch": 5.01, + "learning_rate": 1.4069200560927433e-06, + "loss": 0.9795, + "step": 67738 + }, + { + "epoch": 5.01, + "learning_rate": 1.4067160566583793e-06, + "loss": 0.9371, + "step": 67739 + }, + { + "epoch": 5.01, + "learning_rate": 1.4065120708958012e-06, + "loss": 0.9934, + "step": 67740 + }, + { + "epoch": 5.01, + "learning_rate": 1.4063080988053434e-06, + "loss": 1.105, + "step": 67741 + }, + { + "epoch": 5.01, + "learning_rate": 1.4061041403873231e-06, + "loss": 0.8825, + "step": 67742 + }, + { + "epoch": 5.01, + "learning_rate": 1.4059001956420725e-06, + "loss": 0.9541, + "step": 67743 + }, + { + "epoch": 5.01, + "learning_rate": 1.4056962645699078e-06, + "loss": 1.0178, + "step": 67744 + }, + { + "epoch": 5.01, + "learning_rate": 1.4054923471711602e-06, + "loss": 0.9484, + "step": 67745 + }, + { + "epoch": 5.01, + "learning_rate": 1.4052884434461477e-06, + "loss": 1.0716, + "step": 67746 + }, + { + "epoch": 5.01, + "learning_rate": 1.405084553395204e-06, + "loss": 1.0744, + "step": 67747 + }, + { + "epoch": 5.01, + "learning_rate": 1.404880677018644e-06, + "loss": 0.9411, + "step": 67748 + }, + { + "epoch": 5.01, + "learning_rate": 1.4046768143167954e-06, + "loss": 1.0311, + "step": 67749 + }, + { + "epoch": 5.01, + "learning_rate": 1.4044729652899846e-06, + "loss": 1.1208, + "step": 67750 + }, + { + "epoch": 5.01, + "learning_rate": 1.4042691299385337e-06, + "loss": 1.0434, + "step": 67751 + }, + { + "epoch": 5.01, + "learning_rate": 1.4040653082627687e-06, + "loss": 1.0267, + "step": 67752 + }, + { + "epoch": 5.01, + "learning_rate": 1.4038615002630096e-06, + "loss": 0.9382, + "step": 67753 + }, + { + "epoch": 5.01, + "learning_rate": 1.4036577059395861e-06, + "loss": 0.9827, + "step": 67754 + }, + { + "epoch": 5.01, + "learning_rate": 1.4034539252928204e-06, + "loss": 1.0312, + "step": 67755 + }, + { + "epoch": 5.01, + "learning_rate": 1.4032501583230374e-06, + "loss": 0.9173, + "step": 67756 + }, + { + "epoch": 5.01, + "learning_rate": 1.4030464050305593e-06, + "loss": 0.989, + "step": 67757 + }, + { + "epoch": 5.01, + "learning_rate": 1.4028426654157102e-06, + "loss": 0.9617, + "step": 67758 + }, + { + "epoch": 5.01, + "learning_rate": 1.4026389394788165e-06, + "loss": 0.9732, + "step": 67759 + }, + { + "epoch": 5.01, + "learning_rate": 1.4024352272202014e-06, + "loss": 1.0584, + "step": 67760 + }, + { + "epoch": 5.01, + "learning_rate": 1.4022315286401889e-06, + "loss": 0.9441, + "step": 67761 + }, + { + "epoch": 5.01, + "learning_rate": 1.402027843739101e-06, + "loss": 1.0236, + "step": 67762 + }, + { + "epoch": 5.01, + "learning_rate": 1.4018241725172687e-06, + "loss": 1.1217, + "step": 67763 + }, + { + "epoch": 5.01, + "learning_rate": 1.4016205149750062e-06, + "loss": 1.0843, + "step": 67764 + }, + { + "epoch": 5.01, + "learning_rate": 1.4014168711126452e-06, + "loss": 0.9563, + "step": 67765 + }, + { + "epoch": 5.01, + "learning_rate": 1.4012132409305057e-06, + "loss": 0.9805, + "step": 67766 + }, + { + "epoch": 5.01, + "learning_rate": 1.401009624428914e-06, + "loss": 1.0506, + "step": 67767 + }, + { + "epoch": 5.01, + "learning_rate": 1.400806021608193e-06, + "loss": 0.9882, + "step": 67768 + }, + { + "epoch": 5.01, + "learning_rate": 1.400602432468664e-06, + "loss": 0.9527, + "step": 67769 + }, + { + "epoch": 5.01, + "learning_rate": 1.4003988570106552e-06, + "loss": 1.0428, + "step": 67770 + }, + { + "epoch": 5.01, + "learning_rate": 1.40019529523449e-06, + "loss": 0.9396, + "step": 67771 + }, + { + "epoch": 5.01, + "learning_rate": 1.3999917471404901e-06, + "loss": 0.9657, + "step": 67772 + }, + { + "epoch": 5.01, + "learning_rate": 1.3997882127289775e-06, + "loss": 0.9363, + "step": 67773 + }, + { + "epoch": 5.01, + "learning_rate": 1.399584692000282e-06, + "loss": 0.9992, + "step": 67774 + }, + { + "epoch": 5.01, + "learning_rate": 1.3993811849547245e-06, + "loss": 0.9952, + "step": 67775 + }, + { + "epoch": 5.01, + "learning_rate": 1.399177691592628e-06, + "loss": 0.9998, + "step": 67776 + }, + { + "epoch": 5.01, + "learning_rate": 1.3989742119143168e-06, + "loss": 0.9613, + "step": 67777 + }, + { + "epoch": 5.01, + "learning_rate": 1.3987707459201117e-06, + "loss": 0.9444, + "step": 67778 + }, + { + "epoch": 5.01, + "learning_rate": 1.3985672936103422e-06, + "loss": 0.9421, + "step": 67779 + }, + { + "epoch": 5.01, + "learning_rate": 1.3983638549853295e-06, + "loss": 0.9641, + "step": 67780 + }, + { + "epoch": 5.01, + "learning_rate": 1.3981604300453955e-06, + "loss": 1.0285, + "step": 67781 + }, + { + "epoch": 5.01, + "learning_rate": 1.3979570187908642e-06, + "loss": 0.8406, + "step": 67782 + }, + { + "epoch": 5.01, + "learning_rate": 1.3977536212220622e-06, + "loss": 1.0618, + "step": 67783 + }, + { + "epoch": 5.01, + "learning_rate": 1.3975502373393102e-06, + "loss": 1.0415, + "step": 67784 + }, + { + "epoch": 5.01, + "learning_rate": 1.3973468671429337e-06, + "loss": 0.8723, + "step": 67785 + }, + { + "epoch": 5.01, + "learning_rate": 1.3971435106332554e-06, + "loss": 1.0047, + "step": 67786 + }, + { + "epoch": 5.01, + "learning_rate": 1.3969401678105953e-06, + "loss": 0.8978, + "step": 67787 + }, + { + "epoch": 5.01, + "learning_rate": 1.3967368386752855e-06, + "loss": 1.0831, + "step": 67788 + }, + { + "epoch": 5.01, + "learning_rate": 1.39653352322764e-06, + "loss": 1.023, + "step": 67789 + }, + { + "epoch": 5.01, + "learning_rate": 1.3963302214679886e-06, + "loss": 0.9862, + "step": 67790 + }, + { + "epoch": 5.01, + "learning_rate": 1.39612693339665e-06, + "loss": 1.0973, + "step": 67791 + }, + { + "epoch": 5.01, + "learning_rate": 1.3959236590139557e-06, + "loss": 0.9259, + "step": 67792 + }, + { + "epoch": 5.01, + "learning_rate": 1.3957203983202184e-06, + "loss": 0.9948, + "step": 67793 + }, + { + "epoch": 5.01, + "learning_rate": 1.3955171513157695e-06, + "loss": 0.9364, + "step": 67794 + }, + { + "epoch": 5.01, + "learning_rate": 1.3953139180009301e-06, + "loss": 1.004, + "step": 67795 + }, + { + "epoch": 5.01, + "learning_rate": 1.3951106983760232e-06, + "loss": 1.0014, + "step": 67796 + }, + { + "epoch": 5.01, + "learning_rate": 1.3949074924413707e-06, + "loss": 1.0845, + "step": 67797 + }, + { + "epoch": 5.01, + "learning_rate": 1.3947043001972972e-06, + "loss": 0.9589, + "step": 67798 + }, + { + "epoch": 5.01, + "learning_rate": 1.3945011216441263e-06, + "loss": 0.9548, + "step": 67799 + }, + { + "epoch": 5.01, + "learning_rate": 1.3942979567821823e-06, + "loss": 1.007, + "step": 67800 + }, + { + "epoch": 5.01, + "learning_rate": 1.3940948056117865e-06, + "loss": 1.1729, + "step": 67801 + }, + { + "epoch": 5.01, + "learning_rate": 1.3938916681332605e-06, + "loss": 0.9338, + "step": 67802 + }, + { + "epoch": 5.01, + "learning_rate": 1.3936885443469316e-06, + "loss": 0.95, + "step": 67803 + }, + { + "epoch": 5.01, + "learning_rate": 1.3934854342531212e-06, + "loss": 0.9676, + "step": 67804 + }, + { + "epoch": 5.01, + "learning_rate": 1.393282337852153e-06, + "loss": 1.0361, + "step": 67805 + }, + { + "epoch": 5.01, + "learning_rate": 1.3930792551443495e-06, + "loss": 1.05, + "step": 67806 + }, + { + "epoch": 5.01, + "learning_rate": 1.392876186130031e-06, + "loss": 1.0697, + "step": 67807 + }, + { + "epoch": 5.01, + "learning_rate": 1.3926731308095276e-06, + "loss": 0.9668, + "step": 67808 + }, + { + "epoch": 5.01, + "learning_rate": 1.3924700891831544e-06, + "loss": 1.1412, + "step": 67809 + }, + { + "epoch": 5.01, + "learning_rate": 1.3922670612512402e-06, + "loss": 1.0143, + "step": 67810 + }, + { + "epoch": 5.01, + "learning_rate": 1.3920640470141034e-06, + "loss": 0.9923, + "step": 67811 + }, + { + "epoch": 5.01, + "learning_rate": 1.3918610464720738e-06, + "loss": 0.9531, + "step": 67812 + }, + { + "epoch": 5.01, + "learning_rate": 1.3916580596254681e-06, + "loss": 0.9542, + "step": 67813 + }, + { + "epoch": 5.01, + "learning_rate": 1.391455086474609e-06, + "loss": 1.011, + "step": 67814 + }, + { + "epoch": 5.01, + "learning_rate": 1.3912521270198242e-06, + "loss": 0.9421, + "step": 67815 + }, + { + "epoch": 5.01, + "learning_rate": 1.3910491812614334e-06, + "loss": 1.0246, + "step": 67816 + }, + { + "epoch": 5.01, + "learning_rate": 1.390846249199761e-06, + "loss": 0.9358, + "step": 67817 + }, + { + "epoch": 5.01, + "learning_rate": 1.3906433308351263e-06, + "loss": 0.9621, + "step": 67818 + }, + { + "epoch": 5.01, + "learning_rate": 1.3904404261678573e-06, + "loss": 1.0534, + "step": 67819 + }, + { + "epoch": 5.01, + "learning_rate": 1.3902375351982745e-06, + "loss": 0.9607, + "step": 67820 + }, + { + "epoch": 5.01, + "learning_rate": 1.3900346579266999e-06, + "loss": 0.9614, + "step": 67821 + }, + { + "epoch": 5.01, + "learning_rate": 1.389831794353458e-06, + "loss": 0.9815, + "step": 67822 + }, + { + "epoch": 5.01, + "learning_rate": 1.389628944478868e-06, + "loss": 1.0167, + "step": 67823 + }, + { + "epoch": 5.01, + "learning_rate": 1.3894261083032567e-06, + "loss": 0.8546, + "step": 67824 + }, + { + "epoch": 5.01, + "learning_rate": 1.389223285826946e-06, + "loss": 1.0274, + "step": 67825 + }, + { + "epoch": 5.01, + "learning_rate": 1.389020477050258e-06, + "loss": 0.9921, + "step": 67826 + }, + { + "epoch": 5.01, + "learning_rate": 1.3888176819735121e-06, + "loss": 1.0451, + "step": 67827 + }, + { + "epoch": 5.01, + "learning_rate": 1.3886149005970373e-06, + "loss": 1.0652, + "step": 67828 + }, + { + "epoch": 5.01, + "learning_rate": 1.388412132921152e-06, + "loss": 1.088, + "step": 67829 + }, + { + "epoch": 5.01, + "learning_rate": 1.3882093789461803e-06, + "loss": 0.8718, + "step": 67830 + }, + { + "epoch": 5.01, + "learning_rate": 1.3880066386724421e-06, + "loss": 0.9188, + "step": 67831 + }, + { + "epoch": 5.01, + "learning_rate": 1.387803912100265e-06, + "loss": 0.9446, + "step": 67832 + }, + { + "epoch": 5.01, + "learning_rate": 1.3876011992299697e-06, + "loss": 1.0043, + "step": 67833 + }, + { + "epoch": 5.01, + "learning_rate": 1.387398500061874e-06, + "loss": 1.004, + "step": 67834 + }, + { + "epoch": 5.01, + "learning_rate": 1.387195814596306e-06, + "loss": 0.9434, + "step": 67835 + }, + { + "epoch": 5.01, + "learning_rate": 1.3869931428335837e-06, + "loss": 0.8809, + "step": 67836 + }, + { + "epoch": 5.01, + "learning_rate": 1.3867904847740365e-06, + "loss": 1.0719, + "step": 67837 + }, + { + "epoch": 5.01, + "learning_rate": 1.386587840417979e-06, + "loss": 1.014, + "step": 67838 + }, + { + "epoch": 5.01, + "learning_rate": 1.3863852097657381e-06, + "loss": 1.0904, + "step": 67839 + }, + { + "epoch": 5.01, + "learning_rate": 1.3861825928176353e-06, + "loss": 1.0279, + "step": 67840 + }, + { + "epoch": 5.01, + "learning_rate": 1.3859799895739935e-06, + "loss": 0.9221, + "step": 67841 + }, + { + "epoch": 5.01, + "learning_rate": 1.3857774000351331e-06, + "loss": 1.1393, + "step": 67842 + }, + { + "epoch": 5.01, + "learning_rate": 1.3855748242013766e-06, + "loss": 1.0461, + "step": 67843 + }, + { + "epoch": 5.01, + "learning_rate": 1.385372262073048e-06, + "loss": 0.9988, + "step": 67844 + }, + { + "epoch": 5.01, + "learning_rate": 1.3851697136504705e-06, + "loss": 1.0401, + "step": 67845 + }, + { + "epoch": 5.01, + "learning_rate": 1.3849671789339635e-06, + "loss": 0.94, + "step": 67846 + }, + { + "epoch": 5.01, + "learning_rate": 1.3847646579238494e-06, + "loss": 0.9452, + "step": 67847 + }, + { + "epoch": 5.01, + "learning_rate": 1.384562150620453e-06, + "loss": 1.0578, + "step": 67848 + }, + { + "epoch": 5.01, + "learning_rate": 1.3843596570240946e-06, + "loss": 1.0058, + "step": 67849 + }, + { + "epoch": 5.01, + "learning_rate": 1.384157177135097e-06, + "loss": 1.003, + "step": 67850 + }, + { + "epoch": 5.01, + "learning_rate": 1.383954710953782e-06, + "loss": 1.0468, + "step": 67851 + }, + { + "epoch": 5.01, + "learning_rate": 1.3837522584804697e-06, + "loss": 1.042, + "step": 67852 + }, + { + "epoch": 5.01, + "learning_rate": 1.3835498197154895e-06, + "loss": 1.0766, + "step": 67853 + }, + { + "epoch": 5.01, + "learning_rate": 1.3833473946591525e-06, + "loss": 0.9716, + "step": 67854 + }, + { + "epoch": 5.01, + "learning_rate": 1.3831449833117894e-06, + "loss": 0.9511, + "step": 67855 + }, + { + "epoch": 5.01, + "learning_rate": 1.3829425856737179e-06, + "loss": 1.018, + "step": 67856 + }, + { + "epoch": 5.01, + "learning_rate": 1.382740201745264e-06, + "loss": 0.9091, + "step": 67857 + }, + { + "epoch": 5.01, + "learning_rate": 1.3825378315267446e-06, + "loss": 1.0825, + "step": 67858 + }, + { + "epoch": 5.01, + "learning_rate": 1.3823354750184847e-06, + "loss": 0.996, + "step": 67859 + }, + { + "epoch": 5.01, + "learning_rate": 1.3821331322208064e-06, + "loss": 0.9305, + "step": 67860 + }, + { + "epoch": 5.01, + "learning_rate": 1.3819308031340316e-06, + "loss": 0.9304, + "step": 67861 + }, + { + "epoch": 5.01, + "learning_rate": 1.3817284877584813e-06, + "loss": 0.9718, + "step": 67862 + }, + { + "epoch": 5.01, + "learning_rate": 1.381526186094475e-06, + "loss": 0.9745, + "step": 67863 + }, + { + "epoch": 5.01, + "learning_rate": 1.3813238981423404e-06, + "loss": 1.0041, + "step": 67864 + }, + { + "epoch": 5.01, + "learning_rate": 1.381121623902395e-06, + "loss": 1.1349, + "step": 67865 + }, + { + "epoch": 5.01, + "learning_rate": 1.3809193633749629e-06, + "loss": 0.913, + "step": 67866 + }, + { + "epoch": 5.01, + "learning_rate": 1.3807171165603616e-06, + "loss": 0.9753, + "step": 67867 + }, + { + "epoch": 5.01, + "learning_rate": 1.380514883458919e-06, + "loss": 1.1818, + "step": 67868 + }, + { + "epoch": 5.01, + "learning_rate": 1.380312664070953e-06, + "loss": 1.0417, + "step": 67869 + }, + { + "epoch": 5.01, + "learning_rate": 1.3801104583967872e-06, + "loss": 1.0333, + "step": 67870 + }, + { + "epoch": 5.01, + "learning_rate": 1.3799082664367425e-06, + "loss": 0.9481, + "step": 67871 + }, + { + "epoch": 5.02, + "learning_rate": 1.3797060881911372e-06, + "loss": 0.8919, + "step": 67872 + }, + { + "epoch": 5.02, + "learning_rate": 1.3795039236602993e-06, + "loss": 0.9408, + "step": 67873 + }, + { + "epoch": 5.02, + "learning_rate": 1.3793017728445458e-06, + "loss": 1.0225, + "step": 67874 + }, + { + "epoch": 5.02, + "learning_rate": 1.379099635744201e-06, + "loss": 1.0052, + "step": 67875 + }, + { + "epoch": 5.02, + "learning_rate": 1.3788975123595828e-06, + "loss": 0.9981, + "step": 67876 + }, + { + "epoch": 5.02, + "learning_rate": 1.3786954026910182e-06, + "loss": 0.9682, + "step": 67877 + }, + { + "epoch": 5.02, + "learning_rate": 1.3784933067388273e-06, + "loss": 0.8981, + "step": 67878 + }, + { + "epoch": 5.02, + "learning_rate": 1.3782912245033264e-06, + "loss": 0.9663, + "step": 67879 + }, + { + "epoch": 5.02, + "learning_rate": 1.378089155984843e-06, + "loss": 0.9879, + "step": 67880 + }, + { + "epoch": 5.02, + "learning_rate": 1.3778871011836936e-06, + "loss": 1.0272, + "step": 67881 + }, + { + "epoch": 5.02, + "learning_rate": 1.3776850601002067e-06, + "loss": 1.0379, + "step": 67882 + }, + { + "epoch": 5.02, + "learning_rate": 1.3774830327346954e-06, + "loss": 0.9484, + "step": 67883 + }, + { + "epoch": 5.02, + "learning_rate": 1.3772810190874875e-06, + "loss": 1.011, + "step": 67884 + }, + { + "epoch": 5.02, + "learning_rate": 1.3770790191589023e-06, + "loss": 0.9128, + "step": 67885 + }, + { + "epoch": 5.02, + "learning_rate": 1.376877032949261e-06, + "loss": 0.9217, + "step": 67886 + }, + { + "epoch": 5.02, + "learning_rate": 1.376675060458883e-06, + "loss": 0.9777, + "step": 67887 + }, + { + "epoch": 5.02, + "learning_rate": 1.3764731016880927e-06, + "loss": 0.9696, + "step": 67888 + }, + { + "epoch": 5.02, + "learning_rate": 1.376271156637211e-06, + "loss": 1.1099, + "step": 67889 + }, + { + "epoch": 5.02, + "learning_rate": 1.376069225306559e-06, + "loss": 1.1033, + "step": 67890 + }, + { + "epoch": 5.02, + "learning_rate": 1.3758673076964569e-06, + "loss": 1.0527, + "step": 67891 + }, + { + "epoch": 5.02, + "learning_rate": 1.3756654038072247e-06, + "loss": 1.0703, + "step": 67892 + }, + { + "epoch": 5.02, + "learning_rate": 1.375463513639188e-06, + "loss": 1.07, + "step": 67893 + }, + { + "epoch": 5.02, + "learning_rate": 1.375261637192664e-06, + "loss": 0.8798, + "step": 67894 + }, + { + "epoch": 5.02, + "learning_rate": 1.3750597744679761e-06, + "loss": 1.0255, + "step": 67895 + }, + { + "epoch": 5.02, + "learning_rate": 1.3748579254654426e-06, + "loss": 1.0162, + "step": 67896 + }, + { + "epoch": 5.02, + "learning_rate": 1.374656090185389e-06, + "loss": 0.9721, + "step": 67897 + }, + { + "epoch": 5.02, + "learning_rate": 1.374454268628137e-06, + "loss": 0.9659, + "step": 67898 + }, + { + "epoch": 5.02, + "learning_rate": 1.374252460794e-06, + "loss": 1.0029, + "step": 67899 + }, + { + "epoch": 5.02, + "learning_rate": 1.3740506666833054e-06, + "loss": 1.0654, + "step": 67900 + }, + { + "epoch": 5.02, + "learning_rate": 1.3738488862963707e-06, + "loss": 0.9346, + "step": 67901 + }, + { + "epoch": 5.02, + "learning_rate": 1.3736471196335243e-06, + "loss": 0.9249, + "step": 67902 + }, + { + "epoch": 5.02, + "learning_rate": 1.3734453666950764e-06, + "loss": 0.9996, + "step": 67903 + }, + { + "epoch": 5.02, + "learning_rate": 1.3732436274813566e-06, + "loss": 1.0679, + "step": 67904 + }, + { + "epoch": 5.02, + "learning_rate": 1.37304190199268e-06, + "loss": 1.0549, + "step": 67905 + }, + { + "epoch": 5.02, + "learning_rate": 1.3728401902293754e-06, + "loss": 0.9634, + "step": 67906 + }, + { + "epoch": 5.02, + "learning_rate": 1.3726384921917556e-06, + "loss": 1.0835, + "step": 67907 + }, + { + "epoch": 5.02, + "learning_rate": 1.372436807880142e-06, + "loss": 1.0344, + "step": 67908 + }, + { + "epoch": 5.02, + "learning_rate": 1.3722351372948618e-06, + "loss": 1.0117, + "step": 67909 + }, + { + "epoch": 5.02, + "learning_rate": 1.37203348043623e-06, + "loss": 0.9901, + "step": 67910 + }, + { + "epoch": 5.02, + "learning_rate": 1.3718318373045714e-06, + "loss": 1.0419, + "step": 67911 + }, + { + "epoch": 5.02, + "learning_rate": 1.371630207900202e-06, + "loss": 0.9538, + "step": 67912 + }, + { + "epoch": 5.02, + "learning_rate": 1.3714285922234472e-06, + "loss": 0.9975, + "step": 67913 + }, + { + "epoch": 5.02, + "learning_rate": 1.3712269902746267e-06, + "loss": 0.9366, + "step": 67914 + }, + { + "epoch": 5.02, + "learning_rate": 1.3710254020540603e-06, + "loss": 1.0101, + "step": 67915 + }, + { + "epoch": 5.02, + "learning_rate": 1.3708238275620688e-06, + "loss": 1.1342, + "step": 67916 + }, + { + "epoch": 5.02, + "learning_rate": 1.370622266798971e-06, + "loss": 1.1072, + "step": 67917 + }, + { + "epoch": 5.02, + "learning_rate": 1.370420719765092e-06, + "loss": 1.0297, + "step": 67918 + }, + { + "epoch": 5.02, + "learning_rate": 1.3702191864607505e-06, + "loss": 0.9912, + "step": 67919 + }, + { + "epoch": 5.02, + "learning_rate": 1.3700176668862664e-06, + "loss": 1.1065, + "step": 67920 + }, + { + "epoch": 5.02, + "learning_rate": 1.3698161610419591e-06, + "loss": 1.071, + "step": 67921 + }, + { + "epoch": 5.02, + "learning_rate": 1.369614668928152e-06, + "loss": 0.9151, + "step": 67922 + }, + { + "epoch": 5.02, + "learning_rate": 1.3694131905451658e-06, + "loss": 1.0305, + "step": 67923 + }, + { + "epoch": 5.02, + "learning_rate": 1.369211725893318e-06, + "loss": 0.9065, + "step": 67924 + }, + { + "epoch": 5.02, + "learning_rate": 1.369010274972933e-06, + "loss": 1.1228, + "step": 67925 + }, + { + "epoch": 5.02, + "learning_rate": 1.3688088377843255e-06, + "loss": 0.9237, + "step": 67926 + }, + { + "epoch": 5.02, + "learning_rate": 1.3686074143278239e-06, + "loss": 0.8845, + "step": 67927 + }, + { + "epoch": 5.02, + "learning_rate": 1.3684060046037406e-06, + "loss": 0.9909, + "step": 67928 + }, + { + "epoch": 5.02, + "learning_rate": 1.3682046086124024e-06, + "loss": 1.0602, + "step": 67929 + }, + { + "epoch": 5.02, + "learning_rate": 1.3680032263541266e-06, + "loss": 0.998, + "step": 67930 + }, + { + "epoch": 5.02, + "learning_rate": 1.367801857829234e-06, + "loss": 0.936, + "step": 67931 + }, + { + "epoch": 5.02, + "learning_rate": 1.3676005030380435e-06, + "loss": 1.0992, + "step": 67932 + }, + { + "epoch": 5.02, + "learning_rate": 1.367399161980878e-06, + "loss": 1.0456, + "step": 67933 + }, + { + "epoch": 5.02, + "learning_rate": 1.3671978346580584e-06, + "loss": 1.0299, + "step": 67934 + }, + { + "epoch": 5.02, + "learning_rate": 1.3669965210699033e-06, + "loss": 1.0039, + "step": 67935 + }, + { + "epoch": 5.02, + "learning_rate": 1.3667952212167323e-06, + "loss": 0.9839, + "step": 67936 + }, + { + "epoch": 5.02, + "learning_rate": 1.3665939350988656e-06, + "loss": 0.9778, + "step": 67937 + }, + { + "epoch": 5.02, + "learning_rate": 1.3663926627166257e-06, + "loss": 1.0693, + "step": 67938 + }, + { + "epoch": 5.02, + "learning_rate": 1.3661914040703318e-06, + "loss": 1.0741, + "step": 67939 + }, + { + "epoch": 5.02, + "learning_rate": 1.3659901591603043e-06, + "loss": 1.0523, + "step": 67940 + }, + { + "epoch": 5.02, + "learning_rate": 1.3657889279868597e-06, + "loss": 1.0199, + "step": 67941 + }, + { + "epoch": 5.02, + "learning_rate": 1.3655877105503245e-06, + "loss": 1.0724, + "step": 67942 + }, + { + "epoch": 5.02, + "learning_rate": 1.3653865068510153e-06, + "loss": 1.0034, + "step": 67943 + }, + { + "epoch": 5.02, + "learning_rate": 1.3651853168892537e-06, + "loss": 0.9915, + "step": 67944 + }, + { + "epoch": 5.02, + "learning_rate": 1.3649841406653574e-06, + "loss": 1.0508, + "step": 67945 + }, + { + "epoch": 5.02, + "learning_rate": 1.3647829781796463e-06, + "loss": 1.0509, + "step": 67946 + }, + { + "epoch": 5.02, + "learning_rate": 1.3645818294324453e-06, + "loss": 1.0098, + "step": 67947 + }, + { + "epoch": 5.02, + "learning_rate": 1.3643806944240678e-06, + "loss": 1.1237, + "step": 67948 + }, + { + "epoch": 5.02, + "learning_rate": 1.364179573154839e-06, + "loss": 1.034, + "step": 67949 + }, + { + "epoch": 5.02, + "learning_rate": 1.3639784656250742e-06, + "loss": 0.9356, + "step": 67950 + }, + { + "epoch": 5.02, + "learning_rate": 1.3637773718351012e-06, + "loss": 0.9311, + "step": 67951 + }, + { + "epoch": 5.02, + "learning_rate": 1.3635762917852303e-06, + "loss": 0.9039, + "step": 67952 + }, + { + "epoch": 5.02, + "learning_rate": 1.363375225475787e-06, + "loss": 0.8632, + "step": 67953 + }, + { + "epoch": 5.02, + "learning_rate": 1.3631741729070912e-06, + "loss": 1.014, + "step": 67954 + }, + { + "epoch": 5.02, + "learning_rate": 1.3629731340794617e-06, + "loss": 0.9534, + "step": 67955 + }, + { + "epoch": 5.02, + "learning_rate": 1.362772108993219e-06, + "loss": 0.9818, + "step": 67956 + }, + { + "epoch": 5.02, + "learning_rate": 1.3625710976486794e-06, + "loss": 0.951, + "step": 67957 + }, + { + "epoch": 5.02, + "learning_rate": 1.3623701000461676e-06, + "loss": 0.9976, + "step": 67958 + }, + { + "epoch": 5.02, + "learning_rate": 1.3621691161860017e-06, + "loss": 1.115, + "step": 67959 + }, + { + "epoch": 5.02, + "learning_rate": 1.3619681460685007e-06, + "loss": 1.0672, + "step": 67960 + }, + { + "epoch": 5.02, + "learning_rate": 1.3617671896939833e-06, + "loss": 1.0049, + "step": 67961 + }, + { + "epoch": 5.02, + "learning_rate": 1.3615662470627722e-06, + "loss": 1.0988, + "step": 67962 + }, + { + "epoch": 5.02, + "learning_rate": 1.3613653181751863e-06, + "loss": 1.0594, + "step": 67963 + }, + { + "epoch": 5.02, + "learning_rate": 1.361164403031544e-06, + "loss": 0.9534, + "step": 67964 + }, + { + "epoch": 5.02, + "learning_rate": 1.3609635016321655e-06, + "loss": 0.9995, + "step": 67965 + }, + { + "epoch": 5.02, + "learning_rate": 1.3607626139773688e-06, + "loss": 1.0384, + "step": 67966 + }, + { + "epoch": 5.02, + "learning_rate": 1.3605617400674764e-06, + "loss": 1.0704, + "step": 67967 + }, + { + "epoch": 5.02, + "learning_rate": 1.360360879902808e-06, + "loss": 1.0019, + "step": 67968 + }, + { + "epoch": 5.02, + "learning_rate": 1.3601600334836807e-06, + "loss": 0.966, + "step": 67969 + }, + { + "epoch": 5.02, + "learning_rate": 1.3599592008104135e-06, + "loss": 1.0082, + "step": 67970 + }, + { + "epoch": 5.02, + "learning_rate": 1.3597583818833304e-06, + "loss": 1.0087, + "step": 67971 + }, + { + "epoch": 5.02, + "learning_rate": 1.359557576702749e-06, + "loss": 0.9578, + "step": 67972 + }, + { + "epoch": 5.02, + "learning_rate": 1.359356785268985e-06, + "loss": 1.0445, + "step": 67973 + }, + { + "epoch": 5.02, + "learning_rate": 1.3591560075823619e-06, + "loss": 0.9544, + "step": 67974 + }, + { + "epoch": 5.02, + "learning_rate": 1.3589552436431974e-06, + "loss": 0.9947, + "step": 67975 + }, + { + "epoch": 5.02, + "learning_rate": 1.3587544934518127e-06, + "loss": 0.9136, + "step": 67976 + }, + { + "epoch": 5.02, + "learning_rate": 1.3585537570085227e-06, + "loss": 1.0607, + "step": 67977 + }, + { + "epoch": 5.02, + "learning_rate": 1.358353034313653e-06, + "loss": 1.0759, + "step": 67978 + }, + { + "epoch": 5.02, + "learning_rate": 1.35815232536752e-06, + "loss": 0.9691, + "step": 67979 + }, + { + "epoch": 5.02, + "learning_rate": 1.3579516301704421e-06, + "loss": 0.9752, + "step": 67980 + }, + { + "epoch": 5.02, + "learning_rate": 1.3577509487227403e-06, + "loss": 1.0289, + "step": 67981 + }, + { + "epoch": 5.02, + "learning_rate": 1.357550281024731e-06, + "loss": 0.9966, + "step": 67982 + }, + { + "epoch": 5.02, + "learning_rate": 1.3573496270767372e-06, + "loss": 0.9137, + "step": 67983 + }, + { + "epoch": 5.02, + "learning_rate": 1.3571489868790765e-06, + "loss": 1.0102, + "step": 67984 + }, + { + "epoch": 5.02, + "learning_rate": 1.3569483604320687e-06, + "loss": 0.8966, + "step": 67985 + }, + { + "epoch": 5.02, + "learning_rate": 1.3567477477360303e-06, + "loss": 0.931, + "step": 67986 + }, + { + "epoch": 5.02, + "learning_rate": 1.356547148791284e-06, + "loss": 1.0471, + "step": 67987 + }, + { + "epoch": 5.02, + "learning_rate": 1.3563465635981476e-06, + "loss": 1.0265, + "step": 67988 + }, + { + "epoch": 5.02, + "learning_rate": 1.3561459921569408e-06, + "loss": 1.1038, + "step": 67989 + }, + { + "epoch": 5.02, + "learning_rate": 1.3559454344679824e-06, + "loss": 1.0153, + "step": 67990 + }, + { + "epoch": 5.02, + "learning_rate": 1.3557448905315873e-06, + "loss": 1.0799, + "step": 67991 + }, + { + "epoch": 5.02, + "learning_rate": 1.3555443603480846e-06, + "loss": 1.0152, + "step": 67992 + }, + { + "epoch": 5.02, + "learning_rate": 1.3553438439177812e-06, + "loss": 1.1049, + "step": 67993 + }, + { + "epoch": 5.02, + "learning_rate": 1.3551433412410054e-06, + "loss": 1.0141, + "step": 67994 + }, + { + "epoch": 5.02, + "learning_rate": 1.3549428523180697e-06, + "loss": 1.045, + "step": 67995 + }, + { + "epoch": 5.02, + "learning_rate": 1.3547423771493019e-06, + "loss": 0.9785, + "step": 67996 + }, + { + "epoch": 5.02, + "learning_rate": 1.3545419157350093e-06, + "loss": 0.941, + "step": 67997 + }, + { + "epoch": 5.02, + "learning_rate": 1.3543414680755195e-06, + "loss": 1.0407, + "step": 67998 + }, + { + "epoch": 5.02, + "learning_rate": 1.3541410341711469e-06, + "loss": 1.0133, + "step": 67999 + }, + { + "epoch": 5.02, + "learning_rate": 1.3539406140222166e-06, + "loss": 1.0694, + "step": 68000 + }, + { + "epoch": 5.02, + "learning_rate": 1.3537402076290407e-06, + "loss": 1.0878, + "step": 68001 + }, + { + "epoch": 5.02, + "learning_rate": 1.3535398149919387e-06, + "loss": 0.8618, + "step": 68002 + }, + { + "epoch": 5.02, + "learning_rate": 1.353339436111233e-06, + "loss": 0.9979, + "step": 68003 + }, + { + "epoch": 5.02, + "learning_rate": 1.3531390709872405e-06, + "loss": 1.0207, + "step": 68004 + }, + { + "epoch": 5.02, + "learning_rate": 1.3529387196202803e-06, + "loss": 0.8938, + "step": 68005 + }, + { + "epoch": 5.02, + "learning_rate": 1.3527383820106687e-06, + "loss": 0.932, + "step": 68006 + }, + { + "epoch": 5.02, + "learning_rate": 1.3525380581587278e-06, + "loss": 1.0021, + "step": 68007 + }, + { + "epoch": 5.03, + "learning_rate": 1.3523377480647759e-06, + "loss": 0.9656, + "step": 68008 + }, + { + "epoch": 5.03, + "learning_rate": 1.3521374517291308e-06, + "loss": 0.9266, + "step": 68009 + }, + { + "epoch": 5.03, + "learning_rate": 1.351937169152112e-06, + "loss": 0.9727, + "step": 68010 + }, + { + "epoch": 5.03, + "learning_rate": 1.3517369003340342e-06, + "loss": 0.9755, + "step": 68011 + }, + { + "epoch": 5.03, + "learning_rate": 1.3515366452752221e-06, + "loss": 1.0043, + "step": 68012 + }, + { + "epoch": 5.03, + "learning_rate": 1.3513364039759913e-06, + "loss": 0.934, + "step": 68013 + }, + { + "epoch": 5.03, + "learning_rate": 1.3511361764366603e-06, + "loss": 0.9814, + "step": 68014 + }, + { + "epoch": 5.03, + "learning_rate": 1.3509359626575458e-06, + "loss": 1.0353, + "step": 68015 + }, + { + "epoch": 5.03, + "learning_rate": 1.3507357626389728e-06, + "loss": 0.9199, + "step": 68016 + }, + { + "epoch": 5.03, + "learning_rate": 1.3505355763812511e-06, + "loss": 0.9677, + "step": 68017 + }, + { + "epoch": 5.03, + "learning_rate": 1.350335403884706e-06, + "loss": 0.9375, + "step": 68018 + }, + { + "epoch": 5.03, + "learning_rate": 1.350135245149653e-06, + "loss": 1.0114, + "step": 68019 + }, + { + "epoch": 5.03, + "learning_rate": 1.3499351001764116e-06, + "loss": 1.0352, + "step": 68020 + }, + { + "epoch": 5.03, + "learning_rate": 1.3497349689652995e-06, + "loss": 0.9142, + "step": 68021 + }, + { + "epoch": 5.03, + "learning_rate": 1.349534851516633e-06, + "loss": 0.9848, + "step": 68022 + }, + { + "epoch": 5.03, + "learning_rate": 1.349334747830735e-06, + "loss": 1.0347, + "step": 68023 + }, + { + "epoch": 5.03, + "learning_rate": 1.3491346579079212e-06, + "loss": 1.0564, + "step": 68024 + }, + { + "epoch": 5.03, + "learning_rate": 1.348934581748511e-06, + "loss": 0.9637, + "step": 68025 + }, + { + "epoch": 5.03, + "learning_rate": 1.34873451935282e-06, + "loss": 1.0292, + "step": 68026 + }, + { + "epoch": 5.03, + "learning_rate": 1.3485344707211712e-06, + "loss": 1.0009, + "step": 68027 + }, + { + "epoch": 5.03, + "learning_rate": 1.3483344358538797e-06, + "loss": 0.9991, + "step": 68028 + }, + { + "epoch": 5.03, + "learning_rate": 1.3481344147512642e-06, + "loss": 0.9261, + "step": 68029 + }, + { + "epoch": 5.03, + "learning_rate": 1.3479344074136424e-06, + "loss": 0.9442, + "step": 68030 + }, + { + "epoch": 5.03, + "learning_rate": 1.3477344138413318e-06, + "loss": 1.1058, + "step": 68031 + }, + { + "epoch": 5.03, + "learning_rate": 1.347534434034654e-06, + "loss": 1.0567, + "step": 68032 + }, + { + "epoch": 5.03, + "learning_rate": 1.347334467993925e-06, + "loss": 1.0152, + "step": 68033 + }, + { + "epoch": 5.03, + "learning_rate": 1.3471345157194638e-06, + "loss": 1.0151, + "step": 68034 + }, + { + "epoch": 5.03, + "learning_rate": 1.346934577211585e-06, + "loss": 0.973, + "step": 68035 + }, + { + "epoch": 5.03, + "learning_rate": 1.3467346524706116e-06, + "loss": 0.938, + "step": 68036 + }, + { + "epoch": 5.03, + "learning_rate": 1.3465347414968622e-06, + "loss": 1.0812, + "step": 68037 + }, + { + "epoch": 5.03, + "learning_rate": 1.3463348442906476e-06, + "loss": 0.9063, + "step": 68038 + }, + { + "epoch": 5.03, + "learning_rate": 1.3461349608522923e-06, + "loss": 1.0311, + "step": 68039 + }, + { + "epoch": 5.03, + "learning_rate": 1.34593509118211e-06, + "loss": 1.0962, + "step": 68040 + }, + { + "epoch": 5.03, + "learning_rate": 1.3457352352804265e-06, + "loss": 0.9644, + "step": 68041 + }, + { + "epoch": 5.03, + "learning_rate": 1.345535393147549e-06, + "loss": 0.9936, + "step": 68042 + }, + { + "epoch": 5.03, + "learning_rate": 1.345335564783804e-06, + "loss": 0.8993, + "step": 68043 + }, + { + "epoch": 5.03, + "learning_rate": 1.3451357501895034e-06, + "loss": 1.092, + "step": 68044 + }, + { + "epoch": 5.03, + "learning_rate": 1.3449359493649728e-06, + "loss": 1.1308, + "step": 68045 + }, + { + "epoch": 5.03, + "learning_rate": 1.3447361623105205e-06, + "loss": 1.077, + "step": 68046 + }, + { + "epoch": 5.03, + "learning_rate": 1.3445363890264719e-06, + "loss": 0.9994, + "step": 68047 + }, + { + "epoch": 5.03, + "learning_rate": 1.344336629513142e-06, + "loss": 0.9976, + "step": 68048 + }, + { + "epoch": 5.03, + "learning_rate": 1.3441368837708479e-06, + "loss": 1.0591, + "step": 68049 + }, + { + "epoch": 5.03, + "learning_rate": 1.3439371517999088e-06, + "loss": 1.0314, + "step": 68050 + }, + { + "epoch": 5.03, + "learning_rate": 1.3437374336006403e-06, + "loss": 0.987, + "step": 68051 + }, + { + "epoch": 5.03, + "learning_rate": 1.3435377291733632e-06, + "loss": 0.9937, + "step": 68052 + }, + { + "epoch": 5.03, + "learning_rate": 1.3433380385183935e-06, + "loss": 1.0257, + "step": 68053 + }, + { + "epoch": 5.03, + "learning_rate": 1.3431383616360494e-06, + "loss": 1.0352, + "step": 68054 + }, + { + "epoch": 5.03, + "learning_rate": 1.3429386985266458e-06, + "loss": 0.9281, + "step": 68055 + }, + { + "epoch": 5.03, + "learning_rate": 1.3427390491905057e-06, + "loss": 1.1122, + "step": 68056 + }, + { + "epoch": 5.03, + "learning_rate": 1.3425394136279424e-06, + "loss": 0.928, + "step": 68057 + }, + { + "epoch": 5.03, + "learning_rate": 1.3423397918392767e-06, + "loss": 0.9847, + "step": 68058 + }, + { + "epoch": 5.03, + "learning_rate": 1.3421401838248237e-06, + "loss": 1.056, + "step": 68059 + }, + { + "epoch": 5.03, + "learning_rate": 1.3419405895849002e-06, + "loss": 0.9911, + "step": 68060 + }, + { + "epoch": 5.03, + "learning_rate": 1.3417410091198291e-06, + "loss": 0.9136, + "step": 68061 + }, + { + "epoch": 5.03, + "learning_rate": 1.34154144242992e-06, + "loss": 1.0412, + "step": 68062 + }, + { + "epoch": 5.03, + "learning_rate": 1.3413418895154973e-06, + "loss": 0.9886, + "step": 68063 + }, + { + "epoch": 5.03, + "learning_rate": 1.3411423503768728e-06, + "loss": 0.9421, + "step": 68064 + }, + { + "epoch": 5.03, + "learning_rate": 1.340942825014372e-06, + "loss": 1.0059, + "step": 68065 + }, + { + "epoch": 5.03, + "learning_rate": 1.3407433134283044e-06, + "loss": 1.0291, + "step": 68066 + }, + { + "epoch": 5.03, + "learning_rate": 1.3405438156189887e-06, + "loss": 1.0249, + "step": 68067 + }, + { + "epoch": 5.03, + "learning_rate": 1.340344331586745e-06, + "loss": 1.0308, + "step": 68068 + }, + { + "epoch": 5.03, + "learning_rate": 1.3401448613318913e-06, + "loss": 0.9886, + "step": 68069 + }, + { + "epoch": 5.03, + "learning_rate": 1.3399454048547423e-06, + "loss": 0.9467, + "step": 68070 + }, + { + "epoch": 5.03, + "learning_rate": 1.339745962155613e-06, + "loss": 1.0336, + "step": 68071 + }, + { + "epoch": 5.03, + "learning_rate": 1.3395465332348278e-06, + "loss": 0.9372, + "step": 68072 + }, + { + "epoch": 5.03, + "learning_rate": 1.3393471180926998e-06, + "loss": 1.0736, + "step": 68073 + }, + { + "epoch": 5.03, + "learning_rate": 1.3391477167295464e-06, + "loss": 0.9942, + "step": 68074 + }, + { + "epoch": 5.03, + "learning_rate": 1.338948329145685e-06, + "loss": 1.0384, + "step": 68075 + }, + { + "epoch": 5.03, + "learning_rate": 1.33874895534143e-06, + "loss": 1.0276, + "step": 68076 + }, + { + "epoch": 5.03, + "learning_rate": 1.3385495953171046e-06, + "loss": 1.0406, + "step": 68077 + }, + { + "epoch": 5.03, + "learning_rate": 1.3383502490730227e-06, + "loss": 0.9928, + "step": 68078 + }, + { + "epoch": 5.03, + "learning_rate": 1.3381509166095018e-06, + "loss": 0.9762, + "step": 68079 + }, + { + "epoch": 5.03, + "learning_rate": 1.3379515979268565e-06, + "loss": 1.0683, + "step": 68080 + }, + { + "epoch": 5.03, + "learning_rate": 1.3377522930254084e-06, + "loss": 0.9947, + "step": 68081 + }, + { + "epoch": 5.03, + "learning_rate": 1.3375530019054727e-06, + "loss": 1.0649, + "step": 68082 + }, + { + "epoch": 5.03, + "learning_rate": 1.3373537245673661e-06, + "loss": 1.0045, + "step": 68083 + }, + { + "epoch": 5.03, + "learning_rate": 1.337154461011405e-06, + "loss": 1.0345, + "step": 68084 + }, + { + "epoch": 5.03, + "learning_rate": 1.3369552112379058e-06, + "loss": 1.0059, + "step": 68085 + }, + { + "epoch": 5.03, + "learning_rate": 1.3367559752471914e-06, + "loss": 0.9721, + "step": 68086 + }, + { + "epoch": 5.03, + "learning_rate": 1.3365567530395695e-06, + "loss": 0.9553, + "step": 68087 + }, + { + "epoch": 5.03, + "learning_rate": 1.3363575446153654e-06, + "loss": 1.0603, + "step": 68088 + }, + { + "epoch": 5.03, + "learning_rate": 1.3361583499748888e-06, + "loss": 0.9942, + "step": 68089 + }, + { + "epoch": 5.03, + "learning_rate": 1.3359591691184648e-06, + "loss": 1.0837, + "step": 68090 + }, + { + "epoch": 5.03, + "learning_rate": 1.3357600020464023e-06, + "loss": 0.9998, + "step": 68091 + }, + { + "epoch": 5.03, + "learning_rate": 1.3355608487590233e-06, + "loss": 0.9424, + "step": 68092 + }, + { + "epoch": 5.03, + "learning_rate": 1.3353617092566428e-06, + "loss": 0.9991, + "step": 68093 + }, + { + "epoch": 5.03, + "learning_rate": 1.3351625835395776e-06, + "loss": 0.9919, + "step": 68094 + }, + { + "epoch": 5.03, + "learning_rate": 1.334963471608145e-06, + "loss": 1.008, + "step": 68095 + }, + { + "epoch": 5.03, + "learning_rate": 1.3347643734626604e-06, + "loss": 0.9314, + "step": 68096 + }, + { + "epoch": 5.03, + "learning_rate": 1.3345652891034423e-06, + "loss": 0.9626, + "step": 68097 + }, + { + "epoch": 5.03, + "learning_rate": 1.3343662185308082e-06, + "loss": 1.1079, + "step": 68098 + }, + { + "epoch": 5.03, + "learning_rate": 1.3341671617450725e-06, + "loss": 0.8527, + "step": 68099 + }, + { + "epoch": 5.03, + "learning_rate": 1.3339681187465514e-06, + "loss": 1.0225, + "step": 68100 + }, + { + "epoch": 5.03, + "learning_rate": 1.3337690895355648e-06, + "loss": 0.9115, + "step": 68101 + }, + { + "epoch": 5.03, + "learning_rate": 1.3335700741124268e-06, + "loss": 1.0301, + "step": 68102 + }, + { + "epoch": 5.03, + "learning_rate": 1.3333710724774563e-06, + "loss": 0.9945, + "step": 68103 + }, + { + "epoch": 5.03, + "learning_rate": 1.333172084630967e-06, + "loss": 1.0657, + "step": 68104 + }, + { + "epoch": 5.03, + "learning_rate": 1.3329731105732758e-06, + "loss": 0.9353, + "step": 68105 + }, + { + "epoch": 5.03, + "learning_rate": 1.3327741503047042e-06, + "loss": 1.0178, + "step": 68106 + }, + { + "epoch": 5.03, + "learning_rate": 1.3325752038255612e-06, + "loss": 1.0434, + "step": 68107 + }, + { + "epoch": 5.03, + "learning_rate": 1.3323762711361687e-06, + "loss": 1.0715, + "step": 68108 + }, + { + "epoch": 5.03, + "learning_rate": 1.3321773522368397e-06, + "loss": 1.0216, + "step": 68109 + }, + { + "epoch": 5.03, + "learning_rate": 1.3319784471278962e-06, + "loss": 1.0113, + "step": 68110 + }, + { + "epoch": 5.03, + "learning_rate": 1.331779555809647e-06, + "loss": 0.8737, + "step": 68111 + }, + { + "epoch": 5.03, + "learning_rate": 1.3315806782824149e-06, + "loss": 0.869, + "step": 68112 + }, + { + "epoch": 5.03, + "learning_rate": 1.331381814546514e-06, + "loss": 1.0604, + "step": 68113 + }, + { + "epoch": 5.03, + "learning_rate": 1.33118296460226e-06, + "loss": 0.9499, + "step": 68114 + }, + { + "epoch": 5.03, + "learning_rate": 1.3309841284499702e-06, + "loss": 1.0496, + "step": 68115 + }, + { + "epoch": 5.03, + "learning_rate": 1.330785306089959e-06, + "loss": 1.0268, + "step": 68116 + }, + { + "epoch": 5.03, + "learning_rate": 1.3305864975225457e-06, + "loss": 0.8679, + "step": 68117 + }, + { + "epoch": 5.03, + "learning_rate": 1.330387702748046e-06, + "loss": 0.9921, + "step": 68118 + }, + { + "epoch": 5.03, + "learning_rate": 1.3301889217667752e-06, + "loss": 1.0451, + "step": 68119 + }, + { + "epoch": 5.03, + "learning_rate": 1.3299901545790472e-06, + "loss": 0.9478, + "step": 68120 + }, + { + "epoch": 5.03, + "learning_rate": 1.3297914011851832e-06, + "loss": 1.0726, + "step": 68121 + }, + { + "epoch": 5.03, + "learning_rate": 1.329592661585497e-06, + "loss": 1.1061, + "step": 68122 + }, + { + "epoch": 5.03, + "learning_rate": 1.3293939357803044e-06, + "loss": 1.0782, + "step": 68123 + }, + { + "epoch": 5.03, + "learning_rate": 1.3291952237699223e-06, + "loss": 0.9562, + "step": 68124 + }, + { + "epoch": 5.03, + "learning_rate": 1.3289965255546644e-06, + "loss": 1.0107, + "step": 68125 + }, + { + "epoch": 5.03, + "learning_rate": 1.3287978411348512e-06, + "loss": 1.0621, + "step": 68126 + }, + { + "epoch": 5.03, + "learning_rate": 1.3285991705107971e-06, + "loss": 0.9617, + "step": 68127 + }, + { + "epoch": 5.03, + "learning_rate": 1.328400513682817e-06, + "loss": 0.9209, + "step": 68128 + }, + { + "epoch": 5.03, + "learning_rate": 1.3282018706512257e-06, + "loss": 1.007, + "step": 68129 + }, + { + "epoch": 5.03, + "learning_rate": 1.3280032414163424e-06, + "loss": 1.0643, + "step": 68130 + }, + { + "epoch": 5.03, + "learning_rate": 1.327804625978485e-06, + "loss": 1.0911, + "step": 68131 + }, + { + "epoch": 5.03, + "learning_rate": 1.3276060243379617e-06, + "loss": 0.969, + "step": 68132 + }, + { + "epoch": 5.03, + "learning_rate": 1.3274074364950961e-06, + "loss": 1.0208, + "step": 68133 + }, + { + "epoch": 5.03, + "learning_rate": 1.3272088624501977e-06, + "loss": 0.9406, + "step": 68134 + }, + { + "epoch": 5.03, + "learning_rate": 1.3270103022035918e-06, + "loss": 1.049, + "step": 68135 + }, + { + "epoch": 5.03, + "learning_rate": 1.3268117557555825e-06, + "loss": 0.9668, + "step": 68136 + }, + { + "epoch": 5.03, + "learning_rate": 1.3266132231064944e-06, + "loss": 0.9976, + "step": 68137 + }, + { + "epoch": 5.03, + "learning_rate": 1.3264147042566411e-06, + "loss": 0.9725, + "step": 68138 + }, + { + "epoch": 5.03, + "learning_rate": 1.3262161992063371e-06, + "loss": 1.1442, + "step": 68139 + }, + { + "epoch": 5.03, + "learning_rate": 1.3260177079559e-06, + "loss": 1.0648, + "step": 68140 + }, + { + "epoch": 5.03, + "learning_rate": 1.3258192305056416e-06, + "loss": 1.066, + "step": 68141 + }, + { + "epoch": 5.03, + "learning_rate": 1.325620766855883e-06, + "loss": 1.0736, + "step": 68142 + }, + { + "epoch": 5.04, + "learning_rate": 1.325422317006938e-06, + "loss": 1.0305, + "step": 68143 + }, + { + "epoch": 5.04, + "learning_rate": 1.325223880959121e-06, + "loss": 1.0856, + "step": 68144 + }, + { + "epoch": 5.04, + "learning_rate": 1.3250254587127476e-06, + "loss": 0.9777, + "step": 68145 + }, + { + "epoch": 5.04, + "learning_rate": 1.324827050268136e-06, + "loss": 1.0054, + "step": 68146 + }, + { + "epoch": 5.04, + "learning_rate": 1.3246286556256017e-06, + "loss": 0.9323, + "step": 68147 + }, + { + "epoch": 5.04, + "learning_rate": 1.3244302747854576e-06, + "loss": 0.7916, + "step": 68148 + }, + { + "epoch": 5.04, + "learning_rate": 1.3242319077480215e-06, + "loss": 1.0113, + "step": 68149 + }, + { + "epoch": 5.04, + "learning_rate": 1.3240335545136062e-06, + "loss": 1.0239, + "step": 68150 + }, + { + "epoch": 5.04, + "learning_rate": 1.323835215082534e-06, + "loss": 1.0212, + "step": 68151 + }, + { + "epoch": 5.04, + "learning_rate": 1.323636889455111e-06, + "loss": 0.9784, + "step": 68152 + }, + { + "epoch": 5.04, + "learning_rate": 1.3234385776316593e-06, + "loss": 1.0147, + "step": 68153 + }, + { + "epoch": 5.04, + "learning_rate": 1.323240279612491e-06, + "loss": 0.9908, + "step": 68154 + }, + { + "epoch": 5.04, + "learning_rate": 1.3230419953979279e-06, + "loss": 1.0603, + "step": 68155 + }, + { + "epoch": 5.04, + "learning_rate": 1.3228437249882764e-06, + "loss": 0.7367, + "step": 68156 + }, + { + "epoch": 5.04, + "learning_rate": 1.3226454683838586e-06, + "loss": 0.979, + "step": 68157 + }, + { + "epoch": 5.04, + "learning_rate": 1.3224472255849852e-06, + "loss": 1.0203, + "step": 68158 + }, + { + "epoch": 5.04, + "learning_rate": 1.3222489965919783e-06, + "loss": 1.1065, + "step": 68159 + }, + { + "epoch": 5.04, + "learning_rate": 1.3220507814051476e-06, + "loss": 1.0074, + "step": 68160 + }, + { + "epoch": 5.04, + "learning_rate": 1.3218525800248072e-06, + "loss": 1.0913, + "step": 68161 + }, + { + "epoch": 5.04, + "learning_rate": 1.321654392451277e-06, + "loss": 0.9956, + "step": 68162 + }, + { + "epoch": 5.04, + "learning_rate": 1.3214562186848712e-06, + "loss": 1.1027, + "step": 68163 + }, + { + "epoch": 5.04, + "learning_rate": 1.321258058725904e-06, + "loss": 0.9818, + "step": 68164 + }, + { + "epoch": 5.04, + "learning_rate": 1.3210599125746892e-06, + "loss": 0.923, + "step": 68165 + }, + { + "epoch": 5.04, + "learning_rate": 1.320861780231546e-06, + "loss": 1.118, + "step": 68166 + }, + { + "epoch": 5.04, + "learning_rate": 1.3206636616967872e-06, + "loss": 0.933, + "step": 68167 + }, + { + "epoch": 5.04, + "learning_rate": 1.3204655569707293e-06, + "loss": 1.0075, + "step": 68168 + }, + { + "epoch": 5.04, + "learning_rate": 1.3202674660536852e-06, + "loss": 0.9386, + "step": 68169 + }, + { + "epoch": 5.04, + "learning_rate": 1.3200693889459704e-06, + "loss": 0.956, + "step": 68170 + }, + { + "epoch": 5.04, + "learning_rate": 1.3198713256479024e-06, + "loss": 1.0096, + "step": 68171 + }, + { + "epoch": 5.04, + "learning_rate": 1.3196732761597952e-06, + "loss": 0.8696, + "step": 68172 + }, + { + "epoch": 5.04, + "learning_rate": 1.3194752404819632e-06, + "loss": 1.0153, + "step": 68173 + }, + { + "epoch": 5.04, + "learning_rate": 1.3192772186147208e-06, + "loss": 0.9545, + "step": 68174 + }, + { + "epoch": 5.04, + "learning_rate": 1.319079210558385e-06, + "loss": 1.0555, + "step": 68175 + }, + { + "epoch": 5.04, + "learning_rate": 1.3188812163132703e-06, + "loss": 1.0198, + "step": 68176 + }, + { + "epoch": 5.04, + "learning_rate": 1.318683235879692e-06, + "loss": 0.9707, + "step": 68177 + }, + { + "epoch": 5.04, + "learning_rate": 1.318485269257964e-06, + "loss": 0.9792, + "step": 68178 + }, + { + "epoch": 5.04, + "learning_rate": 1.3182873164484e-06, + "loss": 0.9981, + "step": 68179 + }, + { + "epoch": 5.04, + "learning_rate": 1.3180893774513204e-06, + "loss": 1.0709, + "step": 68180 + }, + { + "epoch": 5.04, + "learning_rate": 1.3178914522670327e-06, + "loss": 1.0027, + "step": 68181 + }, + { + "epoch": 5.04, + "learning_rate": 1.317693540895857e-06, + "loss": 0.9583, + "step": 68182 + }, + { + "epoch": 5.04, + "learning_rate": 1.317495643338107e-06, + "loss": 0.9467, + "step": 68183 + }, + { + "epoch": 5.04, + "learning_rate": 1.3172977595940972e-06, + "loss": 0.9837, + "step": 68184 + }, + { + "epoch": 5.04, + "learning_rate": 1.3170998896641407e-06, + "loss": 0.9833, + "step": 68185 + }, + { + "epoch": 5.04, + "learning_rate": 1.316902033548555e-06, + "loss": 1.114, + "step": 68186 + }, + { + "epoch": 5.04, + "learning_rate": 1.3167041912476552e-06, + "loss": 1.019, + "step": 68187 + }, + { + "epoch": 5.04, + "learning_rate": 1.3165063627617537e-06, + "loss": 0.9715, + "step": 68188 + }, + { + "epoch": 5.04, + "learning_rate": 1.3163085480911675e-06, + "loss": 1.0433, + "step": 68189 + }, + { + "epoch": 5.04, + "learning_rate": 1.3161107472362067e-06, + "loss": 0.9576, + "step": 68190 + }, + { + "epoch": 5.04, + "learning_rate": 1.315912960197192e-06, + "loss": 1.0, + "step": 68191 + }, + { + "epoch": 5.04, + "learning_rate": 1.3157151869744366e-06, + "loss": 1.0673, + "step": 68192 + }, + { + "epoch": 5.04, + "learning_rate": 1.3155174275682536e-06, + "loss": 1.0148, + "step": 68193 + }, + { + "epoch": 5.04, + "learning_rate": 1.3153196819789548e-06, + "loss": 1.0692, + "step": 68194 + }, + { + "epoch": 5.04, + "learning_rate": 1.3151219502068612e-06, + "loss": 1.0692, + "step": 68195 + }, + { + "epoch": 5.04, + "learning_rate": 1.314924232252287e-06, + "loss": 1.0012, + "step": 68196 + }, + { + "epoch": 5.04, + "learning_rate": 1.3147265281155387e-06, + "loss": 0.9786, + "step": 68197 + }, + { + "epoch": 5.04, + "learning_rate": 1.314528837796939e-06, + "loss": 1.0068, + "step": 68198 + }, + { + "epoch": 5.04, + "learning_rate": 1.314331161296798e-06, + "loss": 0.9157, + "step": 68199 + }, + { + "epoch": 5.04, + "learning_rate": 1.3141334986154353e-06, + "loss": 1.0653, + "step": 68200 + }, + { + "epoch": 5.04, + "learning_rate": 1.3139358497531573e-06, + "loss": 1.0001, + "step": 68201 + }, + { + "epoch": 5.04, + "learning_rate": 1.3137382147102862e-06, + "loss": 0.9156, + "step": 68202 + }, + { + "epoch": 5.04, + "learning_rate": 1.3135405934871303e-06, + "loss": 1.0265, + "step": 68203 + }, + { + "epoch": 5.04, + "learning_rate": 1.313342986084012e-06, + "loss": 1.0197, + "step": 68204 + }, + { + "epoch": 5.04, + "learning_rate": 1.3131453925012382e-06, + "loss": 0.9485, + "step": 68205 + }, + { + "epoch": 5.04, + "learning_rate": 1.3129478127391237e-06, + "loss": 0.9321, + "step": 68206 + }, + { + "epoch": 5.04, + "learning_rate": 1.312750246797988e-06, + "loss": 1.0056, + "step": 68207 + }, + { + "epoch": 5.04, + "learning_rate": 1.3125526946781408e-06, + "loss": 0.984, + "step": 68208 + }, + { + "epoch": 5.04, + "learning_rate": 1.3123551563798997e-06, + "loss": 0.8835, + "step": 68209 + }, + { + "epoch": 5.04, + "learning_rate": 1.3121576319035733e-06, + "loss": 0.9979, + "step": 68210 + }, + { + "epoch": 5.04, + "learning_rate": 1.3119601212494825e-06, + "loss": 1.0254, + "step": 68211 + }, + { + "epoch": 5.04, + "learning_rate": 1.3117626244179394e-06, + "loss": 0.9257, + "step": 68212 + }, + { + "epoch": 5.04, + "learning_rate": 1.3115651414092579e-06, + "loss": 0.9403, + "step": 68213 + }, + { + "epoch": 5.04, + "learning_rate": 1.311367672223749e-06, + "loss": 0.989, + "step": 68214 + }, + { + "epoch": 5.04, + "learning_rate": 1.3111702168617324e-06, + "loss": 0.9528, + "step": 68215 + }, + { + "epoch": 5.04, + "learning_rate": 1.310972775323519e-06, + "loss": 1.0003, + "step": 68216 + }, + { + "epoch": 5.04, + "learning_rate": 1.3107753476094242e-06, + "loss": 1.0345, + "step": 68217 + }, + { + "epoch": 5.04, + "learning_rate": 1.3105779337197621e-06, + "loss": 0.9991, + "step": 68218 + }, + { + "epoch": 5.04, + "learning_rate": 1.3103805336548424e-06, + "loss": 0.954, + "step": 68219 + }, + { + "epoch": 5.04, + "learning_rate": 1.3101831474149861e-06, + "loss": 1.0678, + "step": 68220 + }, + { + "epoch": 5.04, + "learning_rate": 1.3099857750005052e-06, + "loss": 0.9705, + "step": 68221 + }, + { + "epoch": 5.04, + "learning_rate": 1.3097884164117114e-06, + "loss": 1.1161, + "step": 68222 + }, + { + "epoch": 5.04, + "learning_rate": 1.3095910716489168e-06, + "loss": 0.9362, + "step": 68223 + }, + { + "epoch": 5.04, + "learning_rate": 1.3093937407124424e-06, + "loss": 0.956, + "step": 68224 + }, + { + "epoch": 5.04, + "learning_rate": 1.3091964236025989e-06, + "loss": 1.0037, + "step": 68225 + }, + { + "epoch": 5.04, + "learning_rate": 1.3089991203196961e-06, + "loss": 1.0739, + "step": 68226 + }, + { + "epoch": 5.04, + "learning_rate": 1.3088018308640538e-06, + "loss": 1.0616, + "step": 68227 + }, + { + "epoch": 5.04, + "learning_rate": 1.3086045552359827e-06, + "loss": 0.9746, + "step": 68228 + }, + { + "epoch": 5.04, + "learning_rate": 1.308407293435797e-06, + "loss": 0.8347, + "step": 68229 + }, + { + "epoch": 5.04, + "learning_rate": 1.3082100454638102e-06, + "loss": 0.9201, + "step": 68230 + }, + { + "epoch": 5.04, + "learning_rate": 1.3080128113203383e-06, + "loss": 0.9881, + "step": 68231 + }, + { + "epoch": 5.04, + "learning_rate": 1.3078155910056932e-06, + "loss": 0.9793, + "step": 68232 + }, + { + "epoch": 5.04, + "learning_rate": 1.3076183845201896e-06, + "loss": 0.9246, + "step": 68233 + }, + { + "epoch": 5.04, + "learning_rate": 1.307421191864141e-06, + "loss": 0.9559, + "step": 68234 + }, + { + "epoch": 5.04, + "learning_rate": 1.3072240130378576e-06, + "loss": 0.8996, + "step": 68235 + }, + { + "epoch": 5.04, + "learning_rate": 1.3070268480416593e-06, + "loss": 0.8555, + "step": 68236 + }, + { + "epoch": 5.04, + "learning_rate": 1.3068296968758575e-06, + "loss": 1.0378, + "step": 68237 + }, + { + "epoch": 5.04, + "learning_rate": 1.3066325595407646e-06, + "loss": 1.0287, + "step": 68238 + }, + { + "epoch": 5.04, + "learning_rate": 1.3064354360366927e-06, + "loss": 0.8113, + "step": 68239 + }, + { + "epoch": 5.04, + "learning_rate": 1.3062383263639612e-06, + "loss": 1.0215, + "step": 68240 + }, + { + "epoch": 5.04, + "learning_rate": 1.3060412305228787e-06, + "loss": 0.9182, + "step": 68241 + }, + { + "epoch": 5.04, + "learning_rate": 1.3058441485137608e-06, + "loss": 0.9627, + "step": 68242 + }, + { + "epoch": 5.04, + "learning_rate": 1.3056470803369192e-06, + "loss": 1.0626, + "step": 68243 + }, + { + "epoch": 5.04, + "learning_rate": 1.3054500259926683e-06, + "loss": 1.0136, + "step": 68244 + }, + { + "epoch": 5.04, + "learning_rate": 1.3052529854813257e-06, + "loss": 0.9113, + "step": 68245 + }, + { + "epoch": 5.04, + "learning_rate": 1.3050559588031964e-06, + "loss": 1.0585, + "step": 68246 + }, + { + "epoch": 5.04, + "learning_rate": 1.3048589459586014e-06, + "loss": 0.997, + "step": 68247 + }, + { + "epoch": 5.04, + "learning_rate": 1.3046619469478483e-06, + "loss": 0.9643, + "step": 68248 + }, + { + "epoch": 5.04, + "learning_rate": 1.304464961771259e-06, + "loss": 1.0597, + "step": 68249 + }, + { + "epoch": 5.04, + "learning_rate": 1.3042679904291366e-06, + "loss": 1.1138, + "step": 68250 + }, + { + "epoch": 5.04, + "learning_rate": 1.3040710329218021e-06, + "loss": 1.0705, + "step": 68251 + }, + { + "epoch": 5.04, + "learning_rate": 1.303874089249566e-06, + "loss": 1.0495, + "step": 68252 + }, + { + "epoch": 5.04, + "learning_rate": 1.3036771594127417e-06, + "loss": 1.135, + "step": 68253 + }, + { + "epoch": 5.04, + "learning_rate": 1.3034802434116422e-06, + "loss": 0.9295, + "step": 68254 + }, + { + "epoch": 5.04, + "learning_rate": 1.3032833412465794e-06, + "loss": 1.0291, + "step": 68255 + }, + { + "epoch": 5.04, + "learning_rate": 1.303086452917871e-06, + "loss": 1.0094, + "step": 68256 + }, + { + "epoch": 5.04, + "learning_rate": 1.3028895784258267e-06, + "loss": 0.9729, + "step": 68257 + }, + { + "epoch": 5.04, + "learning_rate": 1.3026927177707616e-06, + "loss": 1.102, + "step": 68258 + }, + { + "epoch": 5.04, + "learning_rate": 1.3024958709529856e-06, + "loss": 0.9846, + "step": 68259 + }, + { + "epoch": 5.04, + "learning_rate": 1.3022990379728163e-06, + "loss": 1.0736, + "step": 68260 + }, + { + "epoch": 5.04, + "learning_rate": 1.3021022188305643e-06, + "loss": 0.9583, + "step": 68261 + }, + { + "epoch": 5.04, + "learning_rate": 1.301905413526544e-06, + "loss": 1.0522, + "step": 68262 + }, + { + "epoch": 5.04, + "learning_rate": 1.3017086220610663e-06, + "loss": 1.0286, + "step": 68263 + }, + { + "epoch": 5.04, + "learning_rate": 1.3015118444344455e-06, + "loss": 1.0695, + "step": 68264 + }, + { + "epoch": 5.04, + "learning_rate": 1.3013150806469953e-06, + "loss": 1.1092, + "step": 68265 + }, + { + "epoch": 5.04, + "learning_rate": 1.3011183306990293e-06, + "loss": 0.8888, + "step": 68266 + }, + { + "epoch": 5.04, + "learning_rate": 1.3009215945908594e-06, + "loss": 1.0449, + "step": 68267 + }, + { + "epoch": 5.04, + "learning_rate": 1.3007248723227972e-06, + "loss": 0.9178, + "step": 68268 + }, + { + "epoch": 5.04, + "learning_rate": 1.3005281638951583e-06, + "loss": 0.8892, + "step": 68269 + }, + { + "epoch": 5.04, + "learning_rate": 1.3003314693082558e-06, + "loss": 0.9214, + "step": 68270 + }, + { + "epoch": 5.04, + "learning_rate": 1.3001347885624005e-06, + "loss": 1.0173, + "step": 68271 + }, + { + "epoch": 5.04, + "learning_rate": 1.2999381216579065e-06, + "loss": 0.9309, + "step": 68272 + }, + { + "epoch": 5.04, + "learning_rate": 1.2997414685950871e-06, + "loss": 0.9838, + "step": 68273 + }, + { + "epoch": 5.04, + "learning_rate": 1.299544829374254e-06, + "loss": 0.9934, + "step": 68274 + }, + { + "epoch": 5.04, + "learning_rate": 1.2993482039957184e-06, + "loss": 1.0533, + "step": 68275 + }, + { + "epoch": 5.04, + "learning_rate": 1.2991515924597976e-06, + "loss": 0.9291, + "step": 68276 + }, + { + "epoch": 5.04, + "learning_rate": 1.2989549947668024e-06, + "loss": 1.0517, + "step": 68277 + }, + { + "epoch": 5.05, + "learning_rate": 1.2987584109170447e-06, + "loss": 1.0027, + "step": 68278 + }, + { + "epoch": 5.05, + "learning_rate": 1.2985618409108357e-06, + "loss": 1.0096, + "step": 68279 + }, + { + "epoch": 5.05, + "learning_rate": 1.298365284748493e-06, + "loss": 0.9796, + "step": 68280 + }, + { + "epoch": 5.05, + "learning_rate": 1.2981687424303269e-06, + "loss": 1.1505, + "step": 68281 + }, + { + "epoch": 5.05, + "learning_rate": 1.2979722139566498e-06, + "loss": 1.0013, + "step": 68282 + }, + { + "epoch": 5.05, + "learning_rate": 1.2977756993277735e-06, + "loss": 0.9857, + "step": 68283 + }, + { + "epoch": 5.05, + "learning_rate": 1.29757919854401e-06, + "loss": 1.0658, + "step": 68284 + }, + { + "epoch": 5.05, + "learning_rate": 1.2973827116056749e-06, + "loss": 0.9656, + "step": 68285 + }, + { + "epoch": 5.05, + "learning_rate": 1.2971862385130808e-06, + "loss": 1.0614, + "step": 68286 + }, + { + "epoch": 5.05, + "learning_rate": 1.2969897792665377e-06, + "loss": 0.9996, + "step": 68287 + }, + { + "epoch": 5.05, + "learning_rate": 1.2967933338663574e-06, + "loss": 1.0675, + "step": 68288 + }, + { + "epoch": 5.05, + "learning_rate": 1.2965969023128567e-06, + "loss": 1.0122, + "step": 68289 + }, + { + "epoch": 5.05, + "learning_rate": 1.2964004846063482e-06, + "loss": 0.9892, + "step": 68290 + }, + { + "epoch": 5.05, + "learning_rate": 1.2962040807471376e-06, + "loss": 1.0674, + "step": 68291 + }, + { + "epoch": 5.05, + "learning_rate": 1.2960076907355434e-06, + "loss": 0.9491, + "step": 68292 + }, + { + "epoch": 5.05, + "learning_rate": 1.2958113145718742e-06, + "loss": 1.0209, + "step": 68293 + }, + { + "epoch": 5.05, + "learning_rate": 1.2956149522564498e-06, + "loss": 0.8875, + "step": 68294 + }, + { + "epoch": 5.05, + "learning_rate": 1.2954186037895734e-06, + "loss": 1.0147, + "step": 68295 + }, + { + "epoch": 5.05, + "learning_rate": 1.2952222691715622e-06, + "loss": 0.9744, + "step": 68296 + }, + { + "epoch": 5.05, + "learning_rate": 1.2950259484027273e-06, + "loss": 0.9155, + "step": 68297 + }, + { + "epoch": 5.05, + "learning_rate": 1.294829641483385e-06, + "loss": 0.9839, + "step": 68298 + }, + { + "epoch": 5.05, + "learning_rate": 1.2946333484138418e-06, + "loss": 0.9835, + "step": 68299 + }, + { + "epoch": 5.05, + "learning_rate": 1.2944370691944107e-06, + "loss": 0.9709, + "step": 68300 + }, + { + "epoch": 5.05, + "learning_rate": 1.2942408038254073e-06, + "loss": 1.0077, + "step": 68301 + }, + { + "epoch": 5.05, + "learning_rate": 1.2940445523071433e-06, + "loss": 1.1242, + "step": 68302 + }, + { + "epoch": 5.05, + "learning_rate": 1.2938483146399282e-06, + "loss": 1.0239, + "step": 68303 + }, + { + "epoch": 5.05, + "learning_rate": 1.2936520908240746e-06, + "loss": 0.9719, + "step": 68304 + }, + { + "epoch": 5.05, + "learning_rate": 1.2934558808598986e-06, + "loss": 1.0397, + "step": 68305 + }, + { + "epoch": 5.05, + "learning_rate": 1.2932596847477097e-06, + "loss": 1.0455, + "step": 68306 + }, + { + "epoch": 5.05, + "learning_rate": 1.293063502487819e-06, + "loss": 1.134, + "step": 68307 + }, + { + "epoch": 5.05, + "learning_rate": 1.2928673340805408e-06, + "loss": 0.9337, + "step": 68308 + }, + { + "epoch": 5.05, + "learning_rate": 1.2926711795261837e-06, + "loss": 0.9815, + "step": 68309 + }, + { + "epoch": 5.05, + "learning_rate": 1.292475038825064e-06, + "loss": 1.0112, + "step": 68310 + }, + { + "epoch": 5.05, + "learning_rate": 1.2922789119774914e-06, + "loss": 0.9084, + "step": 68311 + }, + { + "epoch": 5.05, + "learning_rate": 1.2920827989837792e-06, + "loss": 1.0933, + "step": 68312 + }, + { + "epoch": 5.05, + "learning_rate": 1.2918866998442369e-06, + "loss": 0.9869, + "step": 68313 + }, + { + "epoch": 5.05, + "learning_rate": 1.2916906145591801e-06, + "loss": 0.8844, + "step": 68314 + }, + { + "epoch": 5.05, + "learning_rate": 1.2914945431289183e-06, + "loss": 0.9138, + "step": 68315 + }, + { + "epoch": 5.05, + "learning_rate": 1.2912984855537647e-06, + "loss": 0.9852, + "step": 68316 + }, + { + "epoch": 5.05, + "learning_rate": 1.291102441834028e-06, + "loss": 0.9274, + "step": 68317 + }, + { + "epoch": 5.05, + "learning_rate": 1.2909064119700277e-06, + "loss": 0.8537, + "step": 68318 + }, + { + "epoch": 5.05, + "learning_rate": 1.2907103959620682e-06, + "loss": 1.0436, + "step": 68319 + }, + { + "epoch": 5.05, + "learning_rate": 1.2905143938104614e-06, + "loss": 0.907, + "step": 68320 + }, + { + "epoch": 5.05, + "learning_rate": 1.2903184055155239e-06, + "loss": 1.0667, + "step": 68321 + }, + { + "epoch": 5.05, + "learning_rate": 1.290122431077565e-06, + "loss": 1.0531, + "step": 68322 + }, + { + "epoch": 5.05, + "learning_rate": 1.2899264704968973e-06, + "loss": 0.9816, + "step": 68323 + }, + { + "epoch": 5.05, + "learning_rate": 1.28973052377383e-06, + "loss": 0.9644, + "step": 68324 + }, + { + "epoch": 5.05, + "learning_rate": 1.2895345909086777e-06, + "loss": 1.0915, + "step": 68325 + }, + { + "epoch": 5.05, + "learning_rate": 1.289338671901752e-06, + "loss": 1.0101, + "step": 68326 + }, + { + "epoch": 5.05, + "learning_rate": 1.289142766753364e-06, + "loss": 0.9871, + "step": 68327 + }, + { + "epoch": 5.05, + "learning_rate": 1.2889468754638257e-06, + "loss": 1.0079, + "step": 68328 + }, + { + "epoch": 5.05, + "learning_rate": 1.2887509980334457e-06, + "loss": 1.0412, + "step": 68329 + }, + { + "epoch": 5.05, + "learning_rate": 1.2885551344625403e-06, + "loss": 0.9911, + "step": 68330 + }, + { + "epoch": 5.05, + "learning_rate": 1.288359284751418e-06, + "loss": 1.0968, + "step": 68331 + }, + { + "epoch": 5.05, + "learning_rate": 1.2881634489003924e-06, + "loss": 0.9153, + "step": 68332 + }, + { + "epoch": 5.05, + "learning_rate": 1.287967626909772e-06, + "loss": 0.9969, + "step": 68333 + }, + { + "epoch": 5.05, + "learning_rate": 1.2877718187798726e-06, + "loss": 0.901, + "step": 68334 + }, + { + "epoch": 5.05, + "learning_rate": 1.2875760245110036e-06, + "loss": 1.1158, + "step": 68335 + }, + { + "epoch": 5.05, + "learning_rate": 1.2873802441034766e-06, + "loss": 1.0758, + "step": 68336 + }, + { + "epoch": 5.05, + "learning_rate": 1.2871844775576025e-06, + "loss": 1.0162, + "step": 68337 + }, + { + "epoch": 5.05, + "learning_rate": 1.2869887248736913e-06, + "loss": 0.9366, + "step": 68338 + }, + { + "epoch": 5.05, + "learning_rate": 1.286792986052061e-06, + "loss": 1.0469, + "step": 68339 + }, + { + "epoch": 5.05, + "learning_rate": 1.2865972610930144e-06, + "loss": 0.8528, + "step": 68340 + }, + { + "epoch": 5.05, + "learning_rate": 1.2864015499968685e-06, + "loss": 1.013, + "step": 68341 + }, + { + "epoch": 5.05, + "learning_rate": 1.2862058527639298e-06, + "loss": 1.0659, + "step": 68342 + }, + { + "epoch": 5.05, + "learning_rate": 1.2860101693945193e-06, + "loss": 1.0214, + "step": 68343 + }, + { + "epoch": 5.05, + "learning_rate": 1.2858144998889365e-06, + "loss": 1.0011, + "step": 68344 + }, + { + "epoch": 5.05, + "learning_rate": 1.2856188442475005e-06, + "loss": 0.9643, + "step": 68345 + }, + { + "epoch": 5.05, + "learning_rate": 1.2854232024705203e-06, + "loss": 1.0658, + "step": 68346 + }, + { + "epoch": 5.05, + "learning_rate": 1.2852275745583066e-06, + "loss": 1.0881, + "step": 68347 + }, + { + "epoch": 5.05, + "learning_rate": 1.2850319605111715e-06, + "loss": 1.0793, + "step": 68348 + }, + { + "epoch": 5.05, + "learning_rate": 1.2848363603294233e-06, + "loss": 0.9234, + "step": 68349 + }, + { + "epoch": 5.05, + "learning_rate": 1.2846407740133794e-06, + "loss": 1.0457, + "step": 68350 + }, + { + "epoch": 5.05, + "learning_rate": 1.2844452015633458e-06, + "loss": 0.9377, + "step": 68351 + }, + { + "epoch": 5.05, + "learning_rate": 1.2842496429796357e-06, + "loss": 1.0394, + "step": 68352 + }, + { + "epoch": 5.05, + "learning_rate": 1.2840540982625572e-06, + "loss": 1.0805, + "step": 68353 + }, + { + "epoch": 5.05, + "learning_rate": 1.2838585674124272e-06, + "loss": 1.0309, + "step": 68354 + }, + { + "epoch": 5.05, + "learning_rate": 1.2836630504295521e-06, + "loss": 0.9814, + "step": 68355 + }, + { + "epoch": 5.05, + "learning_rate": 1.2834675473142454e-06, + "loss": 1.1053, + "step": 68356 + }, + { + "epoch": 5.05, + "learning_rate": 1.2832720580668168e-06, + "loss": 0.9102, + "step": 68357 + }, + { + "epoch": 5.05, + "learning_rate": 1.2830765826875746e-06, + "loss": 0.9758, + "step": 68358 + }, + { + "epoch": 5.05, + "learning_rate": 1.282881121176839e-06, + "loss": 0.9802, + "step": 68359 + }, + { + "epoch": 5.05, + "learning_rate": 1.2826856735349092e-06, + "loss": 1.0611, + "step": 68360 + }, + { + "epoch": 5.05, + "learning_rate": 1.2824902397621042e-06, + "loss": 0.9916, + "step": 68361 + }, + { + "epoch": 5.05, + "learning_rate": 1.2822948198587304e-06, + "loss": 0.909, + "step": 68362 + }, + { + "epoch": 5.05, + "learning_rate": 1.2820994138251042e-06, + "loss": 1.0184, + "step": 68363 + }, + { + "epoch": 5.05, + "learning_rate": 1.281904021661532e-06, + "loss": 0.9701, + "step": 68364 + }, + { + "epoch": 5.05, + "learning_rate": 1.2817086433683234e-06, + "loss": 1.0369, + "step": 68365 + }, + { + "epoch": 5.05, + "learning_rate": 1.281513278945794e-06, + "loss": 1.0071, + "step": 68366 + }, + { + "epoch": 5.05, + "learning_rate": 1.2813179283942523e-06, + "loss": 1.0069, + "step": 68367 + }, + { + "epoch": 5.05, + "learning_rate": 1.281122591714008e-06, + "loss": 0.9162, + "step": 68368 + }, + { + "epoch": 5.05, + "learning_rate": 1.280927268905372e-06, + "loss": 1.0407, + "step": 68369 + }, + { + "epoch": 5.05, + "learning_rate": 1.2807319599686574e-06, + "loss": 1.0019, + "step": 68370 + }, + { + "epoch": 5.05, + "learning_rate": 1.2805366649041739e-06, + "loss": 0.9974, + "step": 68371 + }, + { + "epoch": 5.05, + "learning_rate": 1.2803413837122314e-06, + "loss": 1.0268, + "step": 68372 + }, + { + "epoch": 5.05, + "learning_rate": 1.2801461163931394e-06, + "loss": 0.8931, + "step": 68373 + }, + { + "epoch": 5.05, + "learning_rate": 1.2799508629472112e-06, + "loss": 0.9444, + "step": 68374 + }, + { + "epoch": 5.05, + "learning_rate": 1.2797556233747577e-06, + "loss": 0.9681, + "step": 68375 + }, + { + "epoch": 5.05, + "learning_rate": 1.2795603976760873e-06, + "loss": 0.9996, + "step": 68376 + }, + { + "epoch": 5.05, + "learning_rate": 1.279365185851512e-06, + "loss": 0.957, + "step": 68377 + }, + { + "epoch": 5.05, + "learning_rate": 1.2791699879013397e-06, + "loss": 0.9874, + "step": 68378 + }, + { + "epoch": 5.05, + "learning_rate": 1.2789748038258854e-06, + "loss": 1.0127, + "step": 68379 + }, + { + "epoch": 5.05, + "learning_rate": 1.2787796336254577e-06, + "loss": 0.9147, + "step": 68380 + }, + { + "epoch": 5.05, + "learning_rate": 1.2785844773003664e-06, + "loss": 0.9756, + "step": 68381 + }, + { + "epoch": 5.05, + "learning_rate": 1.2783893348509202e-06, + "loss": 0.9962, + "step": 68382 + }, + { + "epoch": 5.05, + "learning_rate": 1.2781942062774345e-06, + "loss": 1.0077, + "step": 68383 + }, + { + "epoch": 5.05, + "learning_rate": 1.2779990915802187e-06, + "loss": 0.9922, + "step": 68384 + }, + { + "epoch": 5.05, + "learning_rate": 1.2778039907595774e-06, + "loss": 1.0032, + "step": 68385 + }, + { + "epoch": 5.05, + "learning_rate": 1.2776089038158268e-06, + "loss": 0.9991, + "step": 68386 + }, + { + "epoch": 5.05, + "learning_rate": 1.2774138307492746e-06, + "loss": 1.1539, + "step": 68387 + }, + { + "epoch": 5.05, + "learning_rate": 1.2772187715602357e-06, + "loss": 0.9435, + "step": 68388 + }, + { + "epoch": 5.05, + "learning_rate": 1.2770237262490125e-06, + "loss": 0.9761, + "step": 68389 + }, + { + "epoch": 5.05, + "learning_rate": 1.2768286948159226e-06, + "loss": 0.9593, + "step": 68390 + }, + { + "epoch": 5.05, + "learning_rate": 1.2766336772612708e-06, + "loss": 0.9309, + "step": 68391 + }, + { + "epoch": 5.05, + "learning_rate": 1.276438673585374e-06, + "loss": 0.9166, + "step": 68392 + }, + { + "epoch": 5.05, + "learning_rate": 1.2762436837885372e-06, + "loss": 0.9907, + "step": 68393 + }, + { + "epoch": 5.05, + "learning_rate": 1.2760487078710693e-06, + "loss": 0.8912, + "step": 68394 + }, + { + "epoch": 5.05, + "learning_rate": 1.2758537458332843e-06, + "loss": 0.9369, + "step": 68395 + }, + { + "epoch": 5.05, + "learning_rate": 1.2756587976754919e-06, + "loss": 0.9579, + "step": 68396 + }, + { + "epoch": 5.05, + "learning_rate": 1.2754638633980021e-06, + "loss": 0.9754, + "step": 68397 + }, + { + "epoch": 5.05, + "learning_rate": 1.275268943001121e-06, + "loss": 1.0485, + "step": 68398 + }, + { + "epoch": 5.05, + "learning_rate": 1.2750740364851656e-06, + "loss": 1.0337, + "step": 68399 + }, + { + "epoch": 5.05, + "learning_rate": 1.2748791438504426e-06, + "loss": 1.0669, + "step": 68400 + }, + { + "epoch": 5.05, + "learning_rate": 1.2746842650972612e-06, + "loss": 1.028, + "step": 68401 + }, + { + "epoch": 5.05, + "learning_rate": 1.2744894002259335e-06, + "loss": 0.9434, + "step": 68402 + }, + { + "epoch": 5.05, + "learning_rate": 1.2742945492367642e-06, + "loss": 0.9707, + "step": 68403 + }, + { + "epoch": 5.05, + "learning_rate": 1.2740997121300736e-06, + "loss": 1.0069, + "step": 68404 + }, + { + "epoch": 5.05, + "learning_rate": 1.27390488890616e-06, + "loss": 1.0949, + "step": 68405 + }, + { + "epoch": 5.05, + "learning_rate": 1.2737100795653424e-06, + "loss": 1.0166, + "step": 68406 + }, + { + "epoch": 5.05, + "learning_rate": 1.2735152841079236e-06, + "loss": 0.9602, + "step": 68407 + }, + { + "epoch": 5.05, + "learning_rate": 1.2733205025342222e-06, + "loss": 1.0756, + "step": 68408 + }, + { + "epoch": 5.05, + "learning_rate": 1.2731257348445381e-06, + "loss": 1.0205, + "step": 68409 + }, + { + "epoch": 5.05, + "learning_rate": 1.2729309810391888e-06, + "loss": 0.9722, + "step": 68410 + }, + { + "epoch": 5.05, + "learning_rate": 1.2727362411184819e-06, + "loss": 1.0076, + "step": 68411 + }, + { + "epoch": 5.05, + "learning_rate": 1.272541515082726e-06, + "loss": 0.9363, + "step": 68412 + }, + { + "epoch": 5.05, + "learning_rate": 1.2723468029322316e-06, + "loss": 1.0223, + "step": 68413 + }, + { + "epoch": 5.06, + "learning_rate": 1.272152104667307e-06, + "loss": 1.0781, + "step": 68414 + }, + { + "epoch": 5.06, + "learning_rate": 1.2719574202882646e-06, + "loss": 1.0583, + "step": 68415 + }, + { + "epoch": 5.06, + "learning_rate": 1.2717627497954145e-06, + "loss": 0.936, + "step": 68416 + }, + { + "epoch": 5.06, + "learning_rate": 1.271568093189064e-06, + "loss": 1.022, + "step": 68417 + }, + { + "epoch": 5.06, + "learning_rate": 1.2713734504695219e-06, + "loss": 0.8301, + "step": 68418 + }, + { + "epoch": 5.06, + "learning_rate": 1.2711788216371023e-06, + "loss": 1.0439, + "step": 68419 + }, + { + "epoch": 5.06, + "learning_rate": 1.270984206692112e-06, + "loss": 0.9459, + "step": 68420 + }, + { + "epoch": 5.06, + "learning_rate": 1.2707896056348601e-06, + "loss": 1.1091, + "step": 68421 + }, + { + "epoch": 5.06, + "learning_rate": 1.270595018465658e-06, + "loss": 1.0233, + "step": 68422 + }, + { + "epoch": 5.06, + "learning_rate": 1.2704004451848128e-06, + "loss": 0.9773, + "step": 68423 + }, + { + "epoch": 5.06, + "learning_rate": 1.270205885792637e-06, + "loss": 1.0676, + "step": 68424 + }, + { + "epoch": 5.06, + "learning_rate": 1.2700113402894377e-06, + "loss": 0.9203, + "step": 68425 + }, + { + "epoch": 5.06, + "learning_rate": 1.2698168086755269e-06, + "loss": 1.0854, + "step": 68426 + }, + { + "epoch": 5.06, + "learning_rate": 1.2696222909512102e-06, + "loss": 0.912, + "step": 68427 + }, + { + "epoch": 5.06, + "learning_rate": 1.2694277871168016e-06, + "loss": 1.1524, + "step": 68428 + }, + { + "epoch": 5.06, + "learning_rate": 1.2692332971726084e-06, + "loss": 0.9763, + "step": 68429 + }, + { + "epoch": 5.06, + "learning_rate": 1.2690388211189397e-06, + "loss": 0.9316, + "step": 68430 + }, + { + "epoch": 5.06, + "learning_rate": 1.2688443589561062e-06, + "loss": 0.9766, + "step": 68431 + }, + { + "epoch": 5.06, + "learning_rate": 1.268649910684413e-06, + "loss": 0.9849, + "step": 68432 + }, + { + "epoch": 5.06, + "learning_rate": 1.2684554763041779e-06, + "loss": 1.0002, + "step": 68433 + }, + { + "epoch": 5.06, + "learning_rate": 1.2682610558157015e-06, + "loss": 0.9234, + "step": 68434 + }, + { + "epoch": 5.06, + "learning_rate": 1.2680666492192983e-06, + "loss": 1.1024, + "step": 68435 + }, + { + "epoch": 5.06, + "learning_rate": 1.2678722565152746e-06, + "loss": 0.9288, + "step": 68436 + }, + { + "epoch": 5.06, + "learning_rate": 1.2676778777039444e-06, + "loss": 1.0161, + "step": 68437 + }, + { + "epoch": 5.06, + "learning_rate": 1.26748351278561e-06, + "loss": 0.933, + "step": 68438 + }, + { + "epoch": 5.06, + "learning_rate": 1.2672891617605866e-06, + "loss": 1.0814, + "step": 68439 + }, + { + "epoch": 5.06, + "learning_rate": 1.2670948246291813e-06, + "loss": 1.057, + "step": 68440 + }, + { + "epoch": 5.06, + "learning_rate": 1.2669005013917035e-06, + "loss": 0.9131, + "step": 68441 + }, + { + "epoch": 5.06, + "learning_rate": 1.2667061920484614e-06, + "loss": 0.9502, + "step": 68442 + }, + { + "epoch": 5.06, + "learning_rate": 1.2665118965997625e-06, + "loss": 1.0951, + "step": 68443 + }, + { + "epoch": 5.06, + "learning_rate": 1.2663176150459212e-06, + "loss": 0.9801, + "step": 68444 + }, + { + "epoch": 5.06, + "learning_rate": 1.2661233473872424e-06, + "loss": 0.9887, + "step": 68445 + }, + { + "epoch": 5.06, + "learning_rate": 1.2659290936240375e-06, + "loss": 0.9574, + "step": 68446 + }, + { + "epoch": 5.06, + "learning_rate": 1.2657348537566117e-06, + "loss": 1.0185, + "step": 68447 + }, + { + "epoch": 5.06, + "learning_rate": 1.2655406277852799e-06, + "loss": 1.058, + "step": 68448 + }, + { + "epoch": 5.06, + "learning_rate": 1.265346415710349e-06, + "loss": 1.0303, + "step": 68449 + }, + { + "epoch": 5.06, + "learning_rate": 1.265152217532123e-06, + "loss": 0.9596, + "step": 68450 + }, + { + "epoch": 5.06, + "learning_rate": 1.264958033250917e-06, + "loss": 1.0938, + "step": 68451 + }, + { + "epoch": 5.06, + "learning_rate": 1.2647638628670356e-06, + "loss": 0.9685, + "step": 68452 + }, + { + "epoch": 5.06, + "learning_rate": 1.2645697063807948e-06, + "loss": 0.9888, + "step": 68453 + }, + { + "epoch": 5.06, + "learning_rate": 1.2643755637924937e-06, + "loss": 0.9437, + "step": 68454 + }, + { + "epoch": 5.06, + "learning_rate": 1.2641814351024484e-06, + "loss": 0.884, + "step": 68455 + }, + { + "epoch": 5.06, + "learning_rate": 1.2639873203109632e-06, + "loss": 0.9669, + "step": 68456 + }, + { + "epoch": 5.06, + "learning_rate": 1.2637932194183544e-06, + "loss": 0.9998, + "step": 68457 + }, + { + "epoch": 5.06, + "learning_rate": 1.263599132424922e-06, + "loss": 1.0356, + "step": 68458 + }, + { + "epoch": 5.06, + "learning_rate": 1.2634050593309765e-06, + "loss": 1.0628, + "step": 68459 + }, + { + "epoch": 5.06, + "learning_rate": 1.2632110001368313e-06, + "loss": 1.0668, + "step": 68460 + }, + { + "epoch": 5.06, + "learning_rate": 1.2630169548427928e-06, + "loss": 0.9348, + "step": 68461 + }, + { + "epoch": 5.06, + "learning_rate": 1.2628229234491685e-06, + "loss": 0.8537, + "step": 68462 + }, + { + "epoch": 5.06, + "learning_rate": 1.2626289059562658e-06, + "loss": 1.0055, + "step": 68463 + }, + { + "epoch": 5.06, + "learning_rate": 1.2624349023643978e-06, + "loss": 0.8989, + "step": 68464 + }, + { + "epoch": 5.06, + "learning_rate": 1.262240912673871e-06, + "loss": 0.9809, + "step": 68465 + }, + { + "epoch": 5.06, + "learning_rate": 1.2620469368849942e-06, + "loss": 1.0161, + "step": 68466 + }, + { + "epoch": 5.06, + "learning_rate": 1.2618529749980746e-06, + "loss": 0.9753, + "step": 68467 + }, + { + "epoch": 5.06, + "learning_rate": 1.2616590270134198e-06, + "loss": 1.1137, + "step": 68468 + }, + { + "epoch": 5.06, + "learning_rate": 1.2614650929313444e-06, + "loss": 1.0216, + "step": 68469 + }, + { + "epoch": 5.06, + "learning_rate": 1.261271172752151e-06, + "loss": 0.9702, + "step": 68470 + }, + { + "epoch": 5.06, + "learning_rate": 1.2610772664761517e-06, + "loss": 0.9219, + "step": 68471 + }, + { + "epoch": 5.06, + "learning_rate": 1.2608833741036496e-06, + "loss": 0.9064, + "step": 68472 + }, + { + "epoch": 5.06, + "learning_rate": 1.2606894956349602e-06, + "loss": 1.0045, + "step": 68473 + }, + { + "epoch": 5.06, + "learning_rate": 1.2604956310703886e-06, + "loss": 1.0637, + "step": 68474 + }, + { + "epoch": 5.06, + "learning_rate": 1.2603017804102436e-06, + "loss": 0.969, + "step": 68475 + }, + { + "epoch": 5.06, + "learning_rate": 1.2601079436548314e-06, + "loss": 1.0963, + "step": 68476 + }, + { + "epoch": 5.06, + "learning_rate": 1.259914120804464e-06, + "loss": 0.9385, + "step": 68477 + }, + { + "epoch": 5.06, + "learning_rate": 1.2597203118594514e-06, + "loss": 0.8559, + "step": 68478 + }, + { + "epoch": 5.06, + "learning_rate": 1.2595265168200933e-06, + "loss": 1.0641, + "step": 68479 + }, + { + "epoch": 5.06, + "learning_rate": 1.2593327356867057e-06, + "loss": 0.9513, + "step": 68480 + }, + { + "epoch": 5.06, + "learning_rate": 1.2591389684595957e-06, + "loss": 1.0248, + "step": 68481 + }, + { + "epoch": 5.06, + "learning_rate": 1.2589452151390692e-06, + "loss": 1.0761, + "step": 68482 + }, + { + "epoch": 5.06, + "learning_rate": 1.258751475725435e-06, + "loss": 0.8704, + "step": 68483 + }, + { + "epoch": 5.06, + "learning_rate": 1.2585577502190027e-06, + "loss": 0.9575, + "step": 68484 + }, + { + "epoch": 5.06, + "learning_rate": 1.2583640386200814e-06, + "loss": 1.0314, + "step": 68485 + }, + { + "epoch": 5.06, + "learning_rate": 1.2581703409289769e-06, + "loss": 1.0351, + "step": 68486 + }, + { + "epoch": 5.06, + "learning_rate": 1.2579766571459984e-06, + "loss": 1.0167, + "step": 68487 + }, + { + "epoch": 5.06, + "learning_rate": 1.2577829872714519e-06, + "loss": 1.0505, + "step": 68488 + }, + { + "epoch": 5.06, + "learning_rate": 1.2575893313056497e-06, + "loss": 0.972, + "step": 68489 + }, + { + "epoch": 5.06, + "learning_rate": 1.257395689248898e-06, + "loss": 1.0607, + "step": 68490 + }, + { + "epoch": 5.06, + "learning_rate": 1.2572020611015056e-06, + "loss": 0.9801, + "step": 68491 + }, + { + "epoch": 5.06, + "learning_rate": 1.2570084468637766e-06, + "loss": 1.0357, + "step": 68492 + }, + { + "epoch": 5.06, + "learning_rate": 1.256814846536024e-06, + "loss": 1.0052, + "step": 68493 + }, + { + "epoch": 5.06, + "learning_rate": 1.2566212601185535e-06, + "loss": 0.9164, + "step": 68494 + }, + { + "epoch": 5.06, + "learning_rate": 1.2564276876116744e-06, + "loss": 1.0874, + "step": 68495 + }, + { + "epoch": 5.06, + "learning_rate": 1.2562341290156932e-06, + "loss": 0.8561, + "step": 68496 + }, + { + "epoch": 5.06, + "learning_rate": 1.2560405843309165e-06, + "loss": 1.0097, + "step": 68497 + }, + { + "epoch": 5.06, + "learning_rate": 1.2558470535576583e-06, + "loss": 1.0416, + "step": 68498 + }, + { + "epoch": 5.06, + "learning_rate": 1.2556535366962176e-06, + "loss": 1.0461, + "step": 68499 + }, + { + "epoch": 5.06, + "learning_rate": 1.2554600337469103e-06, + "loss": 0.9482, + "step": 68500 + }, + { + "epoch": 5.06, + "learning_rate": 1.2552665447100377e-06, + "loss": 0.9941, + "step": 68501 + }, + { + "epoch": 5.06, + "learning_rate": 1.2550730695859159e-06, + "loss": 1.0339, + "step": 68502 + }, + { + "epoch": 5.06, + "learning_rate": 1.2548796083748428e-06, + "loss": 1.0442, + "step": 68503 + }, + { + "epoch": 5.06, + "learning_rate": 1.2546861610771343e-06, + "loss": 1.097, + "step": 68504 + }, + { + "epoch": 5.06, + "learning_rate": 1.254492727693094e-06, + "loss": 1.0265, + "step": 68505 + }, + { + "epoch": 5.06, + "learning_rate": 1.2542993082230314e-06, + "loss": 1.068, + "step": 68506 + }, + { + "epoch": 5.06, + "learning_rate": 1.2541059026672532e-06, + "loss": 1.0568, + "step": 68507 + }, + { + "epoch": 5.06, + "learning_rate": 1.2539125110260642e-06, + "loss": 0.9961, + "step": 68508 + }, + { + "epoch": 5.06, + "learning_rate": 1.253719133299779e-06, + "loss": 0.9765, + "step": 68509 + }, + { + "epoch": 5.06, + "learning_rate": 1.2535257694887005e-06, + "loss": 1.0234, + "step": 68510 + }, + { + "epoch": 5.06, + "learning_rate": 1.2533324195931373e-06, + "loss": 0.928, + "step": 68511 + }, + { + "epoch": 5.06, + "learning_rate": 1.2531390836133962e-06, + "loss": 0.8848, + "step": 68512 + }, + { + "epoch": 5.06, + "learning_rate": 1.2529457615497864e-06, + "loss": 0.947, + "step": 68513 + }, + { + "epoch": 5.06, + "learning_rate": 1.2527524534026148e-06, + "loss": 0.9119, + "step": 68514 + }, + { + "epoch": 5.06, + "learning_rate": 1.2525591591721898e-06, + "loss": 0.9855, + "step": 68515 + }, + { + "epoch": 5.06, + "learning_rate": 1.2523658788588177e-06, + "loss": 0.9084, + "step": 68516 + }, + { + "epoch": 5.06, + "learning_rate": 1.2521726124628042e-06, + "loss": 0.9893, + "step": 68517 + }, + { + "epoch": 5.06, + "learning_rate": 1.2519793599844609e-06, + "loss": 0.9438, + "step": 68518 + }, + { + "epoch": 5.06, + "learning_rate": 1.2517861214240935e-06, + "loss": 1.0881, + "step": 68519 + }, + { + "epoch": 5.06, + "learning_rate": 1.251592896782009e-06, + "loss": 0.9971, + "step": 68520 + }, + { + "epoch": 5.06, + "learning_rate": 1.251399686058512e-06, + "loss": 0.9884, + "step": 68521 + }, + { + "epoch": 5.06, + "learning_rate": 1.2512064892539167e-06, + "loss": 1.0397, + "step": 68522 + }, + { + "epoch": 5.06, + "learning_rate": 1.251013306368528e-06, + "loss": 1.0, + "step": 68523 + }, + { + "epoch": 5.06, + "learning_rate": 1.2508201374026473e-06, + "loss": 0.9717, + "step": 68524 + }, + { + "epoch": 5.06, + "learning_rate": 1.2506269823565898e-06, + "loss": 1.0247, + "step": 68525 + }, + { + "epoch": 5.06, + "learning_rate": 1.2504338412306583e-06, + "loss": 1.0305, + "step": 68526 + }, + { + "epoch": 5.06, + "learning_rate": 1.2502407140251626e-06, + "loss": 0.9958, + "step": 68527 + }, + { + "epoch": 5.06, + "learning_rate": 1.250047600740406e-06, + "loss": 0.8864, + "step": 68528 + }, + { + "epoch": 5.06, + "learning_rate": 1.2498545013767004e-06, + "loss": 0.9818, + "step": 68529 + }, + { + "epoch": 5.06, + "learning_rate": 1.2496614159343512e-06, + "loss": 0.9968, + "step": 68530 + }, + { + "epoch": 5.06, + "learning_rate": 1.2494683444136657e-06, + "loss": 0.9636, + "step": 68531 + }, + { + "epoch": 5.06, + "learning_rate": 1.2492752868149494e-06, + "loss": 0.9996, + "step": 68532 + }, + { + "epoch": 5.06, + "learning_rate": 1.2490822431385118e-06, + "loss": 1.1175, + "step": 68533 + }, + { + "epoch": 5.06, + "learning_rate": 1.2488892133846596e-06, + "loss": 0.9716, + "step": 68534 + }, + { + "epoch": 5.06, + "learning_rate": 1.2486961975536994e-06, + "loss": 0.8999, + "step": 68535 + }, + { + "epoch": 5.06, + "learning_rate": 1.2485031956459393e-06, + "loss": 0.9128, + "step": 68536 + }, + { + "epoch": 5.06, + "learning_rate": 1.2483102076616814e-06, + "loss": 0.987, + "step": 68537 + }, + { + "epoch": 5.06, + "learning_rate": 1.2481172336012404e-06, + "loss": 1.1396, + "step": 68538 + }, + { + "epoch": 5.06, + "learning_rate": 1.2479242734649188e-06, + "loss": 0.9973, + "step": 68539 + }, + { + "epoch": 5.06, + "learning_rate": 1.2477313272530256e-06, + "loss": 0.969, + "step": 68540 + }, + { + "epoch": 5.06, + "learning_rate": 1.2475383949658637e-06, + "loss": 1.0189, + "step": 68541 + }, + { + "epoch": 5.06, + "learning_rate": 1.247345476603745e-06, + "loss": 1.0216, + "step": 68542 + }, + { + "epoch": 5.06, + "learning_rate": 1.2471525721669765e-06, + "loss": 0.9647, + "step": 68543 + }, + { + "epoch": 5.06, + "learning_rate": 1.2469596816558594e-06, + "loss": 1.016, + "step": 68544 + }, + { + "epoch": 5.06, + "learning_rate": 1.2467668050707061e-06, + "loss": 1.0559, + "step": 68545 + }, + { + "epoch": 5.06, + "learning_rate": 1.2465739424118183e-06, + "loss": 1.0691, + "step": 68546 + }, + { + "epoch": 5.06, + "learning_rate": 1.2463810936795118e-06, + "loss": 1.0764, + "step": 68547 + }, + { + "epoch": 5.06, + "learning_rate": 1.2461882588740825e-06, + "loss": 1.0838, + "step": 68548 + }, + { + "epoch": 5.07, + "learning_rate": 1.2459954379958461e-06, + "loss": 0.9856, + "step": 68549 + }, + { + "epoch": 5.07, + "learning_rate": 1.245802631045102e-06, + "loss": 1.1185, + "step": 68550 + }, + { + "epoch": 5.07, + "learning_rate": 1.2456098380221659e-06, + "loss": 0.9896, + "step": 68551 + }, + { + "epoch": 5.07, + "learning_rate": 1.2454170589273361e-06, + "loss": 0.9826, + "step": 68552 + }, + { + "epoch": 5.07, + "learning_rate": 1.2452242937609216e-06, + "loss": 1.0111, + "step": 68553 + }, + { + "epoch": 5.07, + "learning_rate": 1.2450315425232317e-06, + "loss": 0.9362, + "step": 68554 + }, + { + "epoch": 5.07, + "learning_rate": 1.244838805214571e-06, + "loss": 0.9833, + "step": 68555 + }, + { + "epoch": 5.07, + "learning_rate": 1.2446460818352468e-06, + "loss": 1.0354, + "step": 68556 + }, + { + "epoch": 5.07, + "learning_rate": 1.2444533723855624e-06, + "loss": 1.0151, + "step": 68557 + }, + { + "epoch": 5.07, + "learning_rate": 1.2442606768658305e-06, + "loss": 1.0201, + "step": 68558 + }, + { + "epoch": 5.07, + "learning_rate": 1.2440679952763546e-06, + "loss": 1.0232, + "step": 68559 + }, + { + "epoch": 5.07, + "learning_rate": 1.2438753276174409e-06, + "loss": 0.9664, + "step": 68560 + }, + { + "epoch": 5.07, + "learning_rate": 1.243682673889397e-06, + "loss": 1.0966, + "step": 68561 + }, + { + "epoch": 5.07, + "learning_rate": 1.243490034092526e-06, + "loss": 0.9992, + "step": 68562 + }, + { + "epoch": 5.07, + "learning_rate": 1.2432974082271388e-06, + "loss": 0.9808, + "step": 68563 + }, + { + "epoch": 5.07, + "learning_rate": 1.2431047962935406e-06, + "loss": 1.0826, + "step": 68564 + }, + { + "epoch": 5.07, + "learning_rate": 1.2429121982920366e-06, + "loss": 1.0612, + "step": 68565 + }, + { + "epoch": 5.07, + "learning_rate": 1.2427196142229325e-06, + "loss": 1.0317, + "step": 68566 + }, + { + "epoch": 5.07, + "learning_rate": 1.2425270440865388e-06, + "loss": 0.9166, + "step": 68567 + }, + { + "epoch": 5.07, + "learning_rate": 1.242334487883159e-06, + "loss": 1.1378, + "step": 68568 + }, + { + "epoch": 5.07, + "learning_rate": 1.2421419456131002e-06, + "loss": 1.0155, + "step": 68569 + }, + { + "epoch": 5.07, + "learning_rate": 1.2419494172766667e-06, + "loss": 0.9493, + "step": 68570 + }, + { + "epoch": 5.07, + "learning_rate": 1.241756902874167e-06, + "loss": 1.0665, + "step": 68571 + }, + { + "epoch": 5.07, + "learning_rate": 1.241564402405908e-06, + "loss": 0.9902, + "step": 68572 + }, + { + "epoch": 5.07, + "learning_rate": 1.2413719158721915e-06, + "loss": 1.0099, + "step": 68573 + }, + { + "epoch": 5.07, + "learning_rate": 1.2411794432733292e-06, + "loss": 0.9736, + "step": 68574 + }, + { + "epoch": 5.07, + "learning_rate": 1.2409869846096256e-06, + "loss": 0.9969, + "step": 68575 + }, + { + "epoch": 5.07, + "learning_rate": 1.2407945398813859e-06, + "loss": 0.9564, + "step": 68576 + }, + { + "epoch": 5.07, + "learning_rate": 1.2406021090889143e-06, + "loss": 0.9335, + "step": 68577 + }, + { + "epoch": 5.07, + "learning_rate": 1.2404096922325216e-06, + "loss": 1.0063, + "step": 68578 + }, + { + "epoch": 5.07, + "learning_rate": 1.2402172893125119e-06, + "loss": 1.018, + "step": 68579 + }, + { + "epoch": 5.07, + "learning_rate": 1.240024900329192e-06, + "loss": 0.9573, + "step": 68580 + }, + { + "epoch": 5.07, + "learning_rate": 1.2398325252828657e-06, + "loss": 0.8728, + "step": 68581 + }, + { + "epoch": 5.07, + "learning_rate": 1.2396401641738397e-06, + "loss": 0.9408, + "step": 68582 + }, + { + "epoch": 5.07, + "learning_rate": 1.2394478170024226e-06, + "loss": 1.0013, + "step": 68583 + }, + { + "epoch": 5.07, + "learning_rate": 1.2392554837689186e-06, + "loss": 0.9708, + "step": 68584 + }, + { + "epoch": 5.07, + "learning_rate": 1.2390631644736338e-06, + "loss": 1.0256, + "step": 68585 + }, + { + "epoch": 5.07, + "learning_rate": 1.2388708591168719e-06, + "loss": 1.011, + "step": 68586 + }, + { + "epoch": 5.07, + "learning_rate": 1.2386785676989432e-06, + "loss": 1.0369, + "step": 68587 + }, + { + "epoch": 5.07, + "learning_rate": 1.2384862902201512e-06, + "loss": 1.0062, + "step": 68588 + }, + { + "epoch": 5.07, + "learning_rate": 1.238294026680803e-06, + "loss": 0.9777, + "step": 68589 + }, + { + "epoch": 5.07, + "learning_rate": 1.2381017770812032e-06, + "loss": 0.9433, + "step": 68590 + }, + { + "epoch": 5.07, + "learning_rate": 1.2379095414216558e-06, + "loss": 0.9481, + "step": 68591 + }, + { + "epoch": 5.07, + "learning_rate": 1.237717319702474e-06, + "loss": 1.0038, + "step": 68592 + }, + { + "epoch": 5.07, + "learning_rate": 1.2375251119239528e-06, + "loss": 0.9716, + "step": 68593 + }, + { + "epoch": 5.07, + "learning_rate": 1.2373329180864069e-06, + "loss": 1.1123, + "step": 68594 + }, + { + "epoch": 5.07, + "learning_rate": 1.2371407381901368e-06, + "loss": 0.9889, + "step": 68595 + }, + { + "epoch": 5.07, + "learning_rate": 1.2369485722354546e-06, + "loss": 0.9887, + "step": 68596 + }, + { + "epoch": 5.07, + "learning_rate": 1.236756420222658e-06, + "loss": 0.9466, + "step": 68597 + }, + { + "epoch": 5.07, + "learning_rate": 1.2365642821520585e-06, + "loss": 1.0111, + "step": 68598 + }, + { + "epoch": 5.07, + "learning_rate": 1.2363721580239595e-06, + "loss": 1.0791, + "step": 68599 + }, + { + "epoch": 5.07, + "learning_rate": 1.2361800478386664e-06, + "loss": 0.9919, + "step": 68600 + }, + { + "epoch": 5.07, + "learning_rate": 1.2359879515964869e-06, + "loss": 1.0144, + "step": 68601 + }, + { + "epoch": 5.07, + "learning_rate": 1.2357958692977224e-06, + "loss": 0.9731, + "step": 68602 + }, + { + "epoch": 5.07, + "learning_rate": 1.235603800942683e-06, + "loss": 0.9891, + "step": 68603 + }, + { + "epoch": 5.07, + "learning_rate": 1.2354117465316727e-06, + "loss": 0.9493, + "step": 68604 + }, + { + "epoch": 5.07, + "learning_rate": 1.2352197060649972e-06, + "loss": 1.0059, + "step": 68605 + }, + { + "epoch": 5.07, + "learning_rate": 1.2350276795429595e-06, + "loss": 1.0307, + "step": 68606 + }, + { + "epoch": 5.07, + "learning_rate": 1.2348356669658702e-06, + "loss": 0.9862, + "step": 68607 + }, + { + "epoch": 5.07, + "learning_rate": 1.2346436683340313e-06, + "loss": 0.9271, + "step": 68608 + }, + { + "epoch": 5.07, + "learning_rate": 1.2344516836477483e-06, + "loss": 1.0832, + "step": 68609 + }, + { + "epoch": 5.07, + "learning_rate": 1.2342597129073285e-06, + "loss": 1.0729, + "step": 68610 + }, + { + "epoch": 5.07, + "learning_rate": 1.234067756113073e-06, + "loss": 1.0621, + "step": 68611 + }, + { + "epoch": 5.07, + "learning_rate": 1.2338758132652928e-06, + "loss": 1.1043, + "step": 68612 + }, + { + "epoch": 5.07, + "learning_rate": 1.2336838843642906e-06, + "loss": 0.9227, + "step": 68613 + }, + { + "epoch": 5.07, + "learning_rate": 1.233491969410372e-06, + "loss": 0.9585, + "step": 68614 + }, + { + "epoch": 5.07, + "learning_rate": 1.2333000684038398e-06, + "loss": 0.8981, + "step": 68615 + }, + { + "epoch": 5.07, + "learning_rate": 1.2331081813450063e-06, + "loss": 0.9824, + "step": 68616 + }, + { + "epoch": 5.07, + "learning_rate": 1.23291630823417e-06, + "loss": 1.0473, + "step": 68617 + }, + { + "epoch": 5.07, + "learning_rate": 1.2327244490716373e-06, + "loss": 1.0973, + "step": 68618 + }, + { + "epoch": 5.07, + "learning_rate": 1.2325326038577156e-06, + "loss": 0.9828, + "step": 68619 + }, + { + "epoch": 5.07, + "learning_rate": 1.2323407725927095e-06, + "loss": 0.9686, + "step": 68620 + }, + { + "epoch": 5.07, + "learning_rate": 1.2321489552769229e-06, + "loss": 1.0415, + "step": 68621 + }, + { + "epoch": 5.07, + "learning_rate": 1.231957151910661e-06, + "loss": 1.0088, + "step": 68622 + }, + { + "epoch": 5.07, + "learning_rate": 1.2317653624942305e-06, + "loss": 0.9977, + "step": 68623 + }, + { + "epoch": 5.07, + "learning_rate": 1.2315735870279377e-06, + "loss": 1.0898, + "step": 68624 + }, + { + "epoch": 5.07, + "learning_rate": 1.2313818255120846e-06, + "loss": 1.0974, + "step": 68625 + }, + { + "epoch": 5.07, + "learning_rate": 1.2311900779469776e-06, + "loss": 1.0713, + "step": 68626 + }, + { + "epoch": 5.07, + "learning_rate": 1.2309983443329198e-06, + "loss": 0.9036, + "step": 68627 + }, + { + "epoch": 5.07, + "learning_rate": 1.23080662467022e-06, + "loss": 0.967, + "step": 68628 + }, + { + "epoch": 5.07, + "learning_rate": 1.230614918959182e-06, + "loss": 0.9785, + "step": 68629 + }, + { + "epoch": 5.07, + "learning_rate": 1.2304232272001105e-06, + "loss": 0.9762, + "step": 68630 + }, + { + "epoch": 5.07, + "learning_rate": 1.230231549393307e-06, + "loss": 1.0074, + "step": 68631 + }, + { + "epoch": 5.07, + "learning_rate": 1.2300398855390827e-06, + "loss": 0.967, + "step": 68632 + }, + { + "epoch": 5.07, + "learning_rate": 1.2298482356377383e-06, + "loss": 1.0951, + "step": 68633 + }, + { + "epoch": 5.07, + "learning_rate": 1.2296565996895804e-06, + "loss": 1.081, + "step": 68634 + }, + { + "epoch": 5.07, + "learning_rate": 1.229464977694914e-06, + "loss": 1.0225, + "step": 68635 + }, + { + "epoch": 5.07, + "learning_rate": 1.2292733696540404e-06, + "loss": 0.9036, + "step": 68636 + }, + { + "epoch": 5.07, + "learning_rate": 1.2290817755672724e-06, + "loss": 1.0445, + "step": 68637 + }, + { + "epoch": 5.07, + "learning_rate": 1.2288901954349054e-06, + "loss": 0.9451, + "step": 68638 + }, + { + "epoch": 5.07, + "learning_rate": 1.22869862925725e-06, + "loss": 1.0283, + "step": 68639 + }, + { + "epoch": 5.07, + "learning_rate": 1.2285070770346086e-06, + "loss": 1.132, + "step": 68640 + }, + { + "epoch": 5.07, + "learning_rate": 1.2283155387672897e-06, + "loss": 0.8813, + "step": 68641 + }, + { + "epoch": 5.07, + "learning_rate": 1.2281240144555917e-06, + "loss": 0.9831, + "step": 68642 + }, + { + "epoch": 5.07, + "learning_rate": 1.2279325040998257e-06, + "loss": 1.1024, + "step": 68643 + }, + { + "epoch": 5.07, + "learning_rate": 1.2277410077002905e-06, + "loss": 1.0699, + "step": 68644 + }, + { + "epoch": 5.07, + "learning_rate": 1.2275495252572988e-06, + "loss": 1.0441, + "step": 68645 + }, + { + "epoch": 5.07, + "learning_rate": 1.2273580567711485e-06, + "loss": 1.037, + "step": 68646 + }, + { + "epoch": 5.07, + "learning_rate": 1.2271666022421424e-06, + "loss": 1.024, + "step": 68647 + }, + { + "epoch": 5.07, + "learning_rate": 1.2269751616705916e-06, + "loss": 0.9813, + "step": 68648 + }, + { + "epoch": 5.07, + "learning_rate": 1.226783735056798e-06, + "loss": 1.0725, + "step": 68649 + }, + { + "epoch": 5.07, + "learning_rate": 1.2265923224010666e-06, + "loss": 0.9484, + "step": 68650 + }, + { + "epoch": 5.07, + "learning_rate": 1.2264009237036977e-06, + "loss": 0.957, + "step": 68651 + }, + { + "epoch": 5.07, + "learning_rate": 1.2262095389650031e-06, + "loss": 1.0951, + "step": 68652 + }, + { + "epoch": 5.07, + "learning_rate": 1.2260181681852824e-06, + "loss": 1.0778, + "step": 68653 + }, + { + "epoch": 5.07, + "learning_rate": 1.225826811364842e-06, + "loss": 1.0476, + "step": 68654 + }, + { + "epoch": 5.07, + "learning_rate": 1.2256354685039851e-06, + "loss": 1.0931, + "step": 68655 + }, + { + "epoch": 5.07, + "learning_rate": 1.2254441396030136e-06, + "loss": 1.0033, + "step": 68656 + }, + { + "epoch": 5.07, + "learning_rate": 1.2252528246622386e-06, + "loss": 0.999, + "step": 68657 + }, + { + "epoch": 5.07, + "learning_rate": 1.2250615236819608e-06, + "loss": 0.996, + "step": 68658 + }, + { + "epoch": 5.07, + "learning_rate": 1.2248702366624833e-06, + "loss": 1.0416, + "step": 68659 + }, + { + "epoch": 5.07, + "learning_rate": 1.2246789636041102e-06, + "loss": 1.0121, + "step": 68660 + }, + { + "epoch": 5.07, + "learning_rate": 1.2244877045071525e-06, + "loss": 0.8621, + "step": 68661 + }, + { + "epoch": 5.07, + "learning_rate": 1.2242964593719043e-06, + "loss": 1.1033, + "step": 68662 + }, + { + "epoch": 5.07, + "learning_rate": 1.2241052281986764e-06, + "loss": 0.9342, + "step": 68663 + }, + { + "epoch": 5.07, + "learning_rate": 1.2239140109877724e-06, + "loss": 1.0238, + "step": 68664 + }, + { + "epoch": 5.07, + "learning_rate": 1.2237228077394948e-06, + "loss": 0.9929, + "step": 68665 + }, + { + "epoch": 5.07, + "learning_rate": 1.223531618454149e-06, + "loss": 0.97, + "step": 68666 + }, + { + "epoch": 5.07, + "learning_rate": 1.2233404431320373e-06, + "loss": 1.0592, + "step": 68667 + }, + { + "epoch": 5.07, + "learning_rate": 1.223149281773467e-06, + "loss": 1.0602, + "step": 68668 + }, + { + "epoch": 5.07, + "learning_rate": 1.2229581343787412e-06, + "loss": 1.0427, + "step": 68669 + }, + { + "epoch": 5.07, + "learning_rate": 1.2227670009481618e-06, + "loss": 0.9645, + "step": 68670 + }, + { + "epoch": 5.07, + "learning_rate": 1.2225758814820344e-06, + "loss": 1.043, + "step": 68671 + }, + { + "epoch": 5.07, + "learning_rate": 1.222384775980664e-06, + "loss": 1.014, + "step": 68672 + }, + { + "epoch": 5.07, + "learning_rate": 1.2221936844443538e-06, + "loss": 1.1014, + "step": 68673 + }, + { + "epoch": 5.07, + "learning_rate": 1.2220026068734092e-06, + "loss": 1.0303, + "step": 68674 + }, + { + "epoch": 5.07, + "learning_rate": 1.221811543268131e-06, + "loss": 1.0511, + "step": 68675 + }, + { + "epoch": 5.07, + "learning_rate": 1.2216204936288245e-06, + "loss": 1.0173, + "step": 68676 + }, + { + "epoch": 5.07, + "learning_rate": 1.221429457955795e-06, + "loss": 1.013, + "step": 68677 + }, + { + "epoch": 5.07, + "learning_rate": 1.2212384362493467e-06, + "loss": 0.9466, + "step": 68678 + }, + { + "epoch": 5.07, + "learning_rate": 1.2210474285097818e-06, + "loss": 0.9705, + "step": 68679 + }, + { + "epoch": 5.07, + "learning_rate": 1.220856434737403e-06, + "loss": 1.0059, + "step": 68680 + }, + { + "epoch": 5.07, + "learning_rate": 1.2206654549325181e-06, + "loss": 0.9765, + "step": 68681 + }, + { + "epoch": 5.07, + "learning_rate": 1.2204744890954313e-06, + "loss": 0.8859, + "step": 68682 + }, + { + "epoch": 5.07, + "learning_rate": 1.22028353722644e-06, + "loss": 1.0648, + "step": 68683 + }, + { + "epoch": 5.08, + "learning_rate": 1.2200925993258529e-06, + "loss": 1.0237, + "step": 68684 + }, + { + "epoch": 5.08, + "learning_rate": 1.219901675393972e-06, + "loss": 1.0711, + "step": 68685 + }, + { + "epoch": 5.08, + "learning_rate": 1.2197107654311048e-06, + "loss": 0.9545, + "step": 68686 + }, + { + "epoch": 5.08, + "learning_rate": 1.2195198694375487e-06, + "loss": 0.9804, + "step": 68687 + }, + { + "epoch": 5.08, + "learning_rate": 1.2193289874136127e-06, + "loss": 1.0524, + "step": 68688 + }, + { + "epoch": 5.08, + "learning_rate": 1.2191381193595975e-06, + "loss": 1.0514, + "step": 68689 + }, + { + "epoch": 5.08, + "learning_rate": 1.2189472652758106e-06, + "loss": 0.9036, + "step": 68690 + }, + { + "epoch": 5.08, + "learning_rate": 1.2187564251625495e-06, + "loss": 1.0129, + "step": 68691 + }, + { + "epoch": 5.08, + "learning_rate": 1.218565599020124e-06, + "loss": 1.0299, + "step": 68692 + }, + { + "epoch": 5.08, + "learning_rate": 1.2183747868488338e-06, + "loss": 0.9454, + "step": 68693 + }, + { + "epoch": 5.08, + "learning_rate": 1.2181839886489843e-06, + "loss": 0.958, + "step": 68694 + }, + { + "epoch": 5.08, + "learning_rate": 1.2179932044208788e-06, + "loss": 1.0306, + "step": 68695 + }, + { + "epoch": 5.08, + "learning_rate": 1.2178024341648175e-06, + "loss": 0.9278, + "step": 68696 + }, + { + "epoch": 5.08, + "learning_rate": 1.2176116778811098e-06, + "loss": 0.9347, + "step": 68697 + }, + { + "epoch": 5.08, + "learning_rate": 1.2174209355700561e-06, + "loss": 1.0813, + "step": 68698 + }, + { + "epoch": 5.08, + "learning_rate": 1.2172302072319597e-06, + "loss": 1.0005, + "step": 68699 + }, + { + "epoch": 5.08, + "learning_rate": 1.2170394928671225e-06, + "loss": 0.9531, + "step": 68700 + }, + { + "epoch": 5.08, + "learning_rate": 1.216848792475851e-06, + "loss": 0.8576, + "step": 68701 + }, + { + "epoch": 5.08, + "learning_rate": 1.2166581060584504e-06, + "loss": 1.1049, + "step": 68702 + }, + { + "epoch": 5.08, + "learning_rate": 1.2164674336152172e-06, + "loss": 1.0828, + "step": 68703 + }, + { + "epoch": 5.08, + "learning_rate": 1.2162767751464599e-06, + "loss": 1.0387, + "step": 68704 + }, + { + "epoch": 5.08, + "learning_rate": 1.2160861306524796e-06, + "loss": 1.0824, + "step": 68705 + }, + { + "epoch": 5.08, + "learning_rate": 1.2158955001335837e-06, + "loss": 0.9773, + "step": 68706 + }, + { + "epoch": 5.08, + "learning_rate": 1.2157048835900687e-06, + "loss": 0.9344, + "step": 68707 + }, + { + "epoch": 5.08, + "learning_rate": 1.2155142810222432e-06, + "loss": 0.989, + "step": 68708 + }, + { + "epoch": 5.08, + "learning_rate": 1.2153236924304069e-06, + "loss": 0.9035, + "step": 68709 + }, + { + "epoch": 5.08, + "learning_rate": 1.2151331178148696e-06, + "loss": 0.8563, + "step": 68710 + }, + { + "epoch": 5.08, + "learning_rate": 1.2149425571759276e-06, + "loss": 1.0354, + "step": 68711 + }, + { + "epoch": 5.08, + "learning_rate": 1.2147520105138832e-06, + "loss": 0.9943, + "step": 68712 + }, + { + "epoch": 5.08, + "learning_rate": 1.2145614778290459e-06, + "loss": 0.884, + "step": 68713 + }, + { + "epoch": 5.08, + "learning_rate": 1.2143709591217156e-06, + "loss": 1.0661, + "step": 68714 + }, + { + "epoch": 5.08, + "learning_rate": 1.2141804543921942e-06, + "loss": 0.8908, + "step": 68715 + }, + { + "epoch": 5.08, + "learning_rate": 1.2139899636407847e-06, + "loss": 0.9628, + "step": 68716 + }, + { + "epoch": 5.08, + "learning_rate": 1.213799486867794e-06, + "loss": 1.0123, + "step": 68717 + }, + { + "epoch": 5.08, + "learning_rate": 1.213609024073522e-06, + "loss": 0.9722, + "step": 68718 + }, + { + "epoch": 5.08, + "learning_rate": 1.2134185752582729e-06, + "loss": 1.088, + "step": 68719 + }, + { + "epoch": 5.08, + "learning_rate": 1.213228140422349e-06, + "loss": 0.9901, + "step": 68720 + }, + { + "epoch": 5.08, + "learning_rate": 1.2130377195660503e-06, + "loss": 1.0414, + "step": 68721 + }, + { + "epoch": 5.08, + "learning_rate": 1.2128473126896855e-06, + "loss": 0.9841, + "step": 68722 + }, + { + "epoch": 5.08, + "learning_rate": 1.2126569197935557e-06, + "loss": 1.0186, + "step": 68723 + }, + { + "epoch": 5.08, + "learning_rate": 1.2124665408779624e-06, + "loss": 0.9884, + "step": 68724 + }, + { + "epoch": 5.08, + "learning_rate": 1.2122761759432067e-06, + "loss": 1.0241, + "step": 68725 + }, + { + "epoch": 5.08, + "learning_rate": 1.212085824989596e-06, + "loss": 0.9631, + "step": 68726 + }, + { + "epoch": 5.08, + "learning_rate": 1.2118954880174317e-06, + "loss": 0.8657, + "step": 68727 + }, + { + "epoch": 5.08, + "learning_rate": 1.211705165027015e-06, + "loss": 1.0527, + "step": 68728 + }, + { + "epoch": 5.08, + "learning_rate": 1.2115148560186507e-06, + "loss": 0.9506, + "step": 68729 + }, + { + "epoch": 5.08, + "learning_rate": 1.2113245609926382e-06, + "loss": 0.9597, + "step": 68730 + }, + { + "epoch": 5.08, + "learning_rate": 1.2111342799492864e-06, + "loss": 1.0763, + "step": 68731 + }, + { + "epoch": 5.08, + "learning_rate": 1.2109440128888905e-06, + "loss": 0.9432, + "step": 68732 + }, + { + "epoch": 5.08, + "learning_rate": 1.210753759811758e-06, + "loss": 0.9848, + "step": 68733 + }, + { + "epoch": 5.08, + "learning_rate": 1.2105635207181888e-06, + "loss": 0.9762, + "step": 68734 + }, + { + "epoch": 5.08, + "learning_rate": 1.2103732956084925e-06, + "loss": 0.9266, + "step": 68735 + }, + { + "epoch": 5.08, + "learning_rate": 1.210183084482961e-06, + "loss": 0.9098, + "step": 68736 + }, + { + "epoch": 5.08, + "learning_rate": 1.2099928873419053e-06, + "loss": 0.9214, + "step": 68737 + }, + { + "epoch": 5.08, + "learning_rate": 1.2098027041856253e-06, + "loss": 0.9978, + "step": 68738 + }, + { + "epoch": 5.08, + "learning_rate": 1.2096125350144227e-06, + "loss": 1.0099, + "step": 68739 + }, + { + "epoch": 5.08, + "learning_rate": 1.2094223798286009e-06, + "loss": 0.9271, + "step": 68740 + }, + { + "epoch": 5.08, + "learning_rate": 1.2092322386284606e-06, + "loss": 0.9813, + "step": 68741 + }, + { + "epoch": 5.08, + "learning_rate": 1.2090421114143081e-06, + "loss": 0.9244, + "step": 68742 + }, + { + "epoch": 5.08, + "learning_rate": 1.2088519981864444e-06, + "loss": 0.9241, + "step": 68743 + }, + { + "epoch": 5.08, + "learning_rate": 1.2086618989451704e-06, + "loss": 1.035, + "step": 68744 + }, + { + "epoch": 5.08, + "learning_rate": 1.208471813690788e-06, + "loss": 1.0089, + "step": 68745 + }, + { + "epoch": 5.08, + "learning_rate": 1.2082817424236026e-06, + "loss": 0.8712, + "step": 68746 + }, + { + "epoch": 5.08, + "learning_rate": 1.2080916851439161e-06, + "loss": 1.1058, + "step": 68747 + }, + { + "epoch": 5.08, + "learning_rate": 1.2079016418520294e-06, + "loss": 1.0035, + "step": 68748 + }, + { + "epoch": 5.08, + "learning_rate": 1.2077116125482447e-06, + "loss": 1.0499, + "step": 68749 + }, + { + "epoch": 5.08, + "learning_rate": 1.2075215972328635e-06, + "loss": 1.027, + "step": 68750 + }, + { + "epoch": 5.08, + "learning_rate": 1.2073315959061948e-06, + "loss": 1.0055, + "step": 68751 + }, + { + "epoch": 5.08, + "learning_rate": 1.2071416085685306e-06, + "loss": 1.0571, + "step": 68752 + }, + { + "epoch": 5.08, + "learning_rate": 1.2069516352201805e-06, + "loss": 0.9622, + "step": 68753 + }, + { + "epoch": 5.08, + "learning_rate": 1.206761675861442e-06, + "loss": 1.0226, + "step": 68754 + }, + { + "epoch": 5.08, + "learning_rate": 1.206571730492625e-06, + "loss": 1.1335, + "step": 68755 + }, + { + "epoch": 5.08, + "learning_rate": 1.2063817991140214e-06, + "loss": 1.1247, + "step": 68756 + }, + { + "epoch": 5.08, + "learning_rate": 1.206191881725941e-06, + "loss": 0.9505, + "step": 68757 + }, + { + "epoch": 5.08, + "learning_rate": 1.2060019783286837e-06, + "loss": 0.9896, + "step": 68758 + }, + { + "epoch": 5.08, + "learning_rate": 1.2058120889225523e-06, + "loss": 1.0634, + "step": 68759 + }, + { + "epoch": 5.08, + "learning_rate": 1.2056222135078466e-06, + "loss": 1.0585, + "step": 68760 + }, + { + "epoch": 5.08, + "learning_rate": 1.2054323520848686e-06, + "loss": 1.0466, + "step": 68761 + }, + { + "epoch": 5.08, + "learning_rate": 1.205242504653924e-06, + "loss": 1.0562, + "step": 68762 + }, + { + "epoch": 5.08, + "learning_rate": 1.205052671215312e-06, + "loss": 0.8565, + "step": 68763 + }, + { + "epoch": 5.08, + "learning_rate": 1.2048628517693362e-06, + "loss": 1.0185, + "step": 68764 + }, + { + "epoch": 5.08, + "learning_rate": 1.2046730463162958e-06, + "loss": 0.8656, + "step": 68765 + }, + { + "epoch": 5.08, + "learning_rate": 1.2044832548564967e-06, + "loss": 1.021, + "step": 68766 + }, + { + "epoch": 5.08, + "learning_rate": 1.2042934773902382e-06, + "loss": 0.9709, + "step": 68767 + }, + { + "epoch": 5.08, + "learning_rate": 1.2041037139178236e-06, + "loss": 1.0838, + "step": 68768 + }, + { + "epoch": 5.08, + "learning_rate": 1.2039139644395536e-06, + "loss": 1.0774, + "step": 68769 + }, + { + "epoch": 5.08, + "learning_rate": 1.2037242289557293e-06, + "loss": 0.961, + "step": 68770 + }, + { + "epoch": 5.08, + "learning_rate": 1.203534507466656e-06, + "loss": 1.0488, + "step": 68771 + }, + { + "epoch": 5.08, + "learning_rate": 1.2033447999726333e-06, + "loss": 1.007, + "step": 68772 + }, + { + "epoch": 5.08, + "learning_rate": 1.2031551064739621e-06, + "loss": 1.0617, + "step": 68773 + }, + { + "epoch": 5.08, + "learning_rate": 1.2029654269709445e-06, + "loss": 0.9327, + "step": 68774 + }, + { + "epoch": 5.08, + "learning_rate": 1.2027757614638846e-06, + "loss": 1.1374, + "step": 68775 + }, + { + "epoch": 5.08, + "learning_rate": 1.2025861099530845e-06, + "loss": 0.9857, + "step": 68776 + }, + { + "epoch": 5.08, + "learning_rate": 1.2023964724388404e-06, + "loss": 0.9494, + "step": 68777 + }, + { + "epoch": 5.08, + "learning_rate": 1.202206848921459e-06, + "loss": 1.0485, + "step": 68778 + }, + { + "epoch": 5.08, + "learning_rate": 1.2020172394012398e-06, + "loss": 1.0336, + "step": 68779 + }, + { + "epoch": 5.08, + "learning_rate": 1.2018276438784882e-06, + "loss": 0.991, + "step": 68780 + }, + { + "epoch": 5.08, + "learning_rate": 1.2016380623534995e-06, + "loss": 0.9206, + "step": 68781 + }, + { + "epoch": 5.08, + "learning_rate": 1.2014484948265803e-06, + "loss": 0.9514, + "step": 68782 + }, + { + "epoch": 5.08, + "learning_rate": 1.2012589412980314e-06, + "loss": 0.927, + "step": 68783 + }, + { + "epoch": 5.08, + "learning_rate": 1.2010694017681523e-06, + "loss": 1.058, + "step": 68784 + }, + { + "epoch": 5.08, + "learning_rate": 1.2008798762372464e-06, + "loss": 1.0915, + "step": 68785 + }, + { + "epoch": 5.08, + "learning_rate": 1.2006903647056134e-06, + "loss": 1.0392, + "step": 68786 + }, + { + "epoch": 5.08, + "learning_rate": 1.2005008671735573e-06, + "loss": 0.9922, + "step": 68787 + }, + { + "epoch": 5.08, + "learning_rate": 1.2003113836413793e-06, + "loss": 0.9087, + "step": 68788 + }, + { + "epoch": 5.08, + "learning_rate": 1.2001219141093802e-06, + "loss": 1.125, + "step": 68789 + }, + { + "epoch": 5.08, + "learning_rate": 1.1999324585778583e-06, + "loss": 0.9984, + "step": 68790 + }, + { + "epoch": 5.08, + "learning_rate": 1.1997430170471202e-06, + "loss": 1.0016, + "step": 68791 + }, + { + "epoch": 5.08, + "learning_rate": 1.1995535895174647e-06, + "loss": 1.0551, + "step": 68792 + }, + { + "epoch": 5.08, + "learning_rate": 1.199364175989194e-06, + "loss": 0.9343, + "step": 68793 + }, + { + "epoch": 5.08, + "learning_rate": 1.1991747764626095e-06, + "loss": 1.0467, + "step": 68794 + }, + { + "epoch": 5.08, + "learning_rate": 1.198985390938009e-06, + "loss": 0.8786, + "step": 68795 + }, + { + "epoch": 5.08, + "learning_rate": 1.198796019415701e-06, + "loss": 0.948, + "step": 68796 + }, + { + "epoch": 5.08, + "learning_rate": 1.1986066618959791e-06, + "loss": 1.1185, + "step": 68797 + }, + { + "epoch": 5.08, + "learning_rate": 1.1984173183791493e-06, + "loss": 1.0383, + "step": 68798 + }, + { + "epoch": 5.08, + "learning_rate": 1.1982279888655102e-06, + "loss": 1.0234, + "step": 68799 + }, + { + "epoch": 5.08, + "learning_rate": 1.1980386733553684e-06, + "loss": 0.9999, + "step": 68800 + }, + { + "epoch": 5.08, + "learning_rate": 1.197849371849017e-06, + "loss": 0.9381, + "step": 68801 + }, + { + "epoch": 5.08, + "learning_rate": 1.1976600843467635e-06, + "loss": 1.029, + "step": 68802 + }, + { + "epoch": 5.08, + "learning_rate": 1.1974708108489053e-06, + "loss": 1.0359, + "step": 68803 + }, + { + "epoch": 5.08, + "learning_rate": 1.197281551355748e-06, + "loss": 0.9985, + "step": 68804 + }, + { + "epoch": 5.08, + "learning_rate": 1.1970923058675888e-06, + "loss": 0.8926, + "step": 68805 + }, + { + "epoch": 5.08, + "learning_rate": 1.1969030743847277e-06, + "loss": 1.0022, + "step": 68806 + }, + { + "epoch": 5.08, + "learning_rate": 1.1967138569074687e-06, + "loss": 0.9961, + "step": 68807 + }, + { + "epoch": 5.08, + "learning_rate": 1.196524653436113e-06, + "loss": 0.9247, + "step": 68808 + }, + { + "epoch": 5.08, + "learning_rate": 1.1963354639709612e-06, + "loss": 1.0169, + "step": 68809 + }, + { + "epoch": 5.08, + "learning_rate": 1.19614628851231e-06, + "loss": 0.8999, + "step": 68810 + }, + { + "epoch": 5.08, + "learning_rate": 1.1959571270604674e-06, + "loss": 0.8493, + "step": 68811 + }, + { + "epoch": 5.08, + "learning_rate": 1.1957679796157307e-06, + "loss": 0.9972, + "step": 68812 + }, + { + "epoch": 5.08, + "learning_rate": 1.1955788461784012e-06, + "loss": 0.9979, + "step": 68813 + }, + { + "epoch": 5.08, + "learning_rate": 1.1953897267487801e-06, + "loss": 1.0033, + "step": 68814 + }, + { + "epoch": 5.08, + "learning_rate": 1.195200621327165e-06, + "loss": 0.9622, + "step": 68815 + }, + { + "epoch": 5.08, + "learning_rate": 1.1950115299138632e-06, + "loss": 1.027, + "step": 68816 + }, + { + "epoch": 5.08, + "learning_rate": 1.194822452509171e-06, + "loss": 1.0753, + "step": 68817 + }, + { + "epoch": 5.08, + "learning_rate": 1.1946333891133898e-06, + "loss": 1.0236, + "step": 68818 + }, + { + "epoch": 5.08, + "learning_rate": 1.19444433972682e-06, + "loss": 1.0197, + "step": 68819 + }, + { + "epoch": 5.09, + "learning_rate": 1.1942553043497651e-06, + "loss": 0.8798, + "step": 68820 + }, + { + "epoch": 5.09, + "learning_rate": 1.1940662829825233e-06, + "loss": 1.0006, + "step": 68821 + }, + { + "epoch": 5.09, + "learning_rate": 1.193877275625397e-06, + "loss": 0.9624, + "step": 68822 + }, + { + "epoch": 5.09, + "learning_rate": 1.1936882822786854e-06, + "loss": 0.9062, + "step": 68823 + }, + { + "epoch": 5.09, + "learning_rate": 1.19349930294269e-06, + "loss": 1.066, + "step": 68824 + }, + { + "epoch": 5.09, + "learning_rate": 1.1933103376177113e-06, + "loss": 0.9417, + "step": 68825 + }, + { + "epoch": 5.09, + "learning_rate": 1.1931213863040468e-06, + "loss": 0.9899, + "step": 68826 + }, + { + "epoch": 5.09, + "learning_rate": 1.1929324490020034e-06, + "loss": 1.0037, + "step": 68827 + }, + { + "epoch": 5.09, + "learning_rate": 1.192743525711878e-06, + "loss": 1.0074, + "step": 68828 + }, + { + "epoch": 5.09, + "learning_rate": 1.1925546164339718e-06, + "loss": 1.0244, + "step": 68829 + }, + { + "epoch": 5.09, + "learning_rate": 1.1923657211685836e-06, + "loss": 1.0777, + "step": 68830 + }, + { + "epoch": 5.09, + "learning_rate": 1.1921768399160172e-06, + "loss": 1.0159, + "step": 68831 + }, + { + "epoch": 5.09, + "learning_rate": 1.1919879726765705e-06, + "loss": 1.0619, + "step": 68832 + }, + { + "epoch": 5.09, + "learning_rate": 1.1917991194505462e-06, + "loss": 0.9608, + "step": 68833 + }, + { + "epoch": 5.09, + "learning_rate": 1.1916102802382424e-06, + "loss": 0.8001, + "step": 68834 + }, + { + "epoch": 5.09, + "learning_rate": 1.1914214550399594e-06, + "loss": 1.0264, + "step": 68835 + }, + { + "epoch": 5.09, + "learning_rate": 1.1912326438560008e-06, + "loss": 0.9518, + "step": 68836 + }, + { + "epoch": 5.09, + "learning_rate": 1.1910438466866648e-06, + "loss": 0.9279, + "step": 68837 + }, + { + "epoch": 5.09, + "learning_rate": 1.1908550635322525e-06, + "loss": 0.9885, + "step": 68838 + }, + { + "epoch": 5.09, + "learning_rate": 1.1906662943930604e-06, + "loss": 1.0296, + "step": 68839 + }, + { + "epoch": 5.09, + "learning_rate": 1.1904775392693957e-06, + "loss": 1.0143, + "step": 68840 + }, + { + "epoch": 5.09, + "learning_rate": 1.1902887981615564e-06, + "loss": 0.9966, + "step": 68841 + }, + { + "epoch": 5.09, + "learning_rate": 1.1901000710698374e-06, + "loss": 1.0769, + "step": 68842 + }, + { + "epoch": 5.09, + "learning_rate": 1.1899113579945455e-06, + "loss": 1.0025, + "step": 68843 + }, + { + "epoch": 5.09, + "learning_rate": 1.1897226589359767e-06, + "loss": 0.9167, + "step": 68844 + }, + { + "epoch": 5.09, + "learning_rate": 1.1895339738944366e-06, + "loss": 0.9187, + "step": 68845 + }, + { + "epoch": 5.09, + "learning_rate": 1.1893453028702184e-06, + "loss": 0.9295, + "step": 68846 + }, + { + "epoch": 5.09, + "learning_rate": 1.1891566458636284e-06, + "loss": 1.1105, + "step": 68847 + }, + { + "epoch": 5.09, + "learning_rate": 1.1889680028749606e-06, + "loss": 0.9953, + "step": 68848 + }, + { + "epoch": 5.09, + "learning_rate": 1.1887793739045238e-06, + "loss": 1.0136, + "step": 68849 + }, + { + "epoch": 5.09, + "learning_rate": 1.1885907589526102e-06, + "loss": 0.8968, + "step": 68850 + }, + { + "epoch": 5.09, + "learning_rate": 1.1884021580195215e-06, + "loss": 0.9362, + "step": 68851 + }, + { + "epoch": 5.09, + "learning_rate": 1.1882135711055597e-06, + "loss": 0.9636, + "step": 68852 + }, + { + "epoch": 5.09, + "learning_rate": 1.1880249982110249e-06, + "loss": 0.9264, + "step": 68853 + }, + { + "epoch": 5.09, + "learning_rate": 1.1878364393362162e-06, + "loss": 0.9883, + "step": 68854 + }, + { + "epoch": 5.09, + "learning_rate": 1.1876478944814306e-06, + "loss": 0.9894, + "step": 68855 + }, + { + "epoch": 5.09, + "learning_rate": 1.1874593636469745e-06, + "loss": 1.0442, + "step": 68856 + }, + { + "epoch": 5.09, + "learning_rate": 1.187270846833144e-06, + "loss": 0.88, + "step": 68857 + }, + { + "epoch": 5.09, + "learning_rate": 1.187082344040239e-06, + "loss": 0.9199, + "step": 68858 + }, + { + "epoch": 5.09, + "learning_rate": 1.1868938552685582e-06, + "loss": 1.0106, + "step": 68859 + }, + { + "epoch": 5.09, + "learning_rate": 1.1867053805184058e-06, + "loss": 0.9976, + "step": 68860 + }, + { + "epoch": 5.09, + "learning_rate": 1.1865169197900783e-06, + "loss": 0.9864, + "step": 68861 + }, + { + "epoch": 5.09, + "learning_rate": 1.1863284730838764e-06, + "loss": 1.0105, + "step": 68862 + }, + { + "epoch": 5.09, + "learning_rate": 1.1861400404001e-06, + "loss": 0.9936, + "step": 68863 + }, + { + "epoch": 5.09, + "learning_rate": 1.1859516217390465e-06, + "loss": 1.027, + "step": 68864 + }, + { + "epoch": 5.09, + "learning_rate": 1.1857632171010203e-06, + "loss": 1.0569, + "step": 68865 + }, + { + "epoch": 5.09, + "learning_rate": 1.1855748264863188e-06, + "loss": 0.9149, + "step": 68866 + }, + { + "epoch": 5.09, + "learning_rate": 1.1853864498952405e-06, + "loss": 0.9847, + "step": 68867 + }, + { + "epoch": 5.09, + "learning_rate": 1.1851980873280856e-06, + "loss": 0.9953, + "step": 68868 + }, + { + "epoch": 5.09, + "learning_rate": 1.1850097387851579e-06, + "loss": 0.9815, + "step": 68869 + }, + { + "epoch": 5.09, + "learning_rate": 1.1848214042667516e-06, + "loss": 0.8743, + "step": 68870 + }, + { + "epoch": 5.09, + "learning_rate": 1.1846330837731656e-06, + "loss": 0.9767, + "step": 68871 + }, + { + "epoch": 5.09, + "learning_rate": 1.1844447773047052e-06, + "loss": 0.9703, + "step": 68872 + }, + { + "epoch": 5.09, + "learning_rate": 1.1842564848616667e-06, + "loss": 0.9667, + "step": 68873 + }, + { + "epoch": 5.09, + "learning_rate": 1.1840682064443498e-06, + "loss": 1.0938, + "step": 68874 + }, + { + "epoch": 5.09, + "learning_rate": 1.1838799420530523e-06, + "loss": 0.9384, + "step": 68875 + }, + { + "epoch": 5.09, + "learning_rate": 1.1836916916880781e-06, + "loss": 0.9677, + "step": 68876 + }, + { + "epoch": 5.09, + "learning_rate": 1.1835034553497237e-06, + "loss": 1.0124, + "step": 68877 + }, + { + "epoch": 5.09, + "learning_rate": 1.1833152330382891e-06, + "loss": 1.0347, + "step": 68878 + }, + { + "epoch": 5.09, + "learning_rate": 1.1831270247540748e-06, + "loss": 0.9448, + "step": 68879 + }, + { + "epoch": 5.09, + "learning_rate": 1.1829388304973767e-06, + "loss": 0.8956, + "step": 68880 + }, + { + "epoch": 5.09, + "learning_rate": 1.1827506502684983e-06, + "loss": 0.9407, + "step": 68881 + }, + { + "epoch": 5.09, + "learning_rate": 1.1825624840677374e-06, + "loss": 1.0931, + "step": 68882 + }, + { + "epoch": 5.09, + "learning_rate": 1.182374331895394e-06, + "loss": 1.0168, + "step": 68883 + }, + { + "epoch": 5.09, + "learning_rate": 1.1821861937517653e-06, + "loss": 0.9288, + "step": 68884 + }, + { + "epoch": 5.09, + "learning_rate": 1.1819980696371536e-06, + "loss": 0.9236, + "step": 68885 + }, + { + "epoch": 5.09, + "learning_rate": 1.1818099595518562e-06, + "loss": 1.0181, + "step": 68886 + }, + { + "epoch": 5.09, + "learning_rate": 1.181621863496174e-06, + "loss": 1.0419, + "step": 68887 + }, + { + "epoch": 5.09, + "learning_rate": 1.1814337814704047e-06, + "loss": 1.0039, + "step": 68888 + }, + { + "epoch": 5.09, + "learning_rate": 1.1812457134748468e-06, + "loss": 0.9504, + "step": 68889 + }, + { + "epoch": 5.09, + "learning_rate": 1.1810576595098044e-06, + "loss": 0.9843, + "step": 68890 + }, + { + "epoch": 5.09, + "learning_rate": 1.1808696195755686e-06, + "loss": 1.0384, + "step": 68891 + }, + { + "epoch": 5.09, + "learning_rate": 1.180681593672446e-06, + "loss": 1.0683, + "step": 68892 + }, + { + "epoch": 5.09, + "learning_rate": 1.1804935818007291e-06, + "loss": 0.9801, + "step": 68893 + }, + { + "epoch": 5.09, + "learning_rate": 1.1803055839607257e-06, + "loss": 0.9649, + "step": 68894 + }, + { + "epoch": 5.09, + "learning_rate": 1.180117600152727e-06, + "loss": 1.0111, + "step": 68895 + }, + { + "epoch": 5.09, + "learning_rate": 1.1799296303770358e-06, + "loss": 1.001, + "step": 68896 + }, + { + "epoch": 5.09, + "learning_rate": 1.1797416746339507e-06, + "loss": 1.0015, + "step": 68897 + }, + { + "epoch": 5.09, + "learning_rate": 1.1795537329237705e-06, + "loss": 1.0131, + "step": 68898 + }, + { + "epoch": 5.09, + "learning_rate": 1.1793658052467939e-06, + "loss": 0.8672, + "step": 68899 + }, + { + "epoch": 5.09, + "learning_rate": 1.179177891603318e-06, + "loss": 0.9993, + "step": 68900 + }, + { + "epoch": 5.09, + "learning_rate": 1.1789899919936453e-06, + "loss": 1.0606, + "step": 68901 + }, + { + "epoch": 5.09, + "learning_rate": 1.1788021064180744e-06, + "loss": 0.9608, + "step": 68902 + }, + { + "epoch": 5.09, + "learning_rate": 1.1786142348769025e-06, + "loss": 0.9902, + "step": 68903 + }, + { + "epoch": 5.09, + "learning_rate": 1.1784263773704275e-06, + "loss": 0.9834, + "step": 68904 + }, + { + "epoch": 5.09, + "learning_rate": 1.1782385338989522e-06, + "loss": 1.0438, + "step": 68905 + }, + { + "epoch": 5.09, + "learning_rate": 1.1780507044627732e-06, + "loss": 1.0192, + "step": 68906 + }, + { + "epoch": 5.09, + "learning_rate": 1.177862889062188e-06, + "loss": 0.9935, + "step": 68907 + }, + { + "epoch": 5.09, + "learning_rate": 1.1776750876974985e-06, + "loss": 1.0737, + "step": 68908 + }, + { + "epoch": 5.09, + "learning_rate": 1.1774873003689979e-06, + "loss": 1.0208, + "step": 68909 + }, + { + "epoch": 5.09, + "learning_rate": 1.1772995270769915e-06, + "loss": 1.0435, + "step": 68910 + }, + { + "epoch": 5.09, + "learning_rate": 1.1771117678217758e-06, + "loss": 1.0592, + "step": 68911 + }, + { + "epoch": 5.09, + "learning_rate": 1.1769240226036493e-06, + "loss": 1.0427, + "step": 68912 + }, + { + "epoch": 5.09, + "learning_rate": 1.1767362914229074e-06, + "loss": 1.0284, + "step": 68913 + }, + { + "epoch": 5.09, + "learning_rate": 1.1765485742798566e-06, + "loss": 0.9799, + "step": 68914 + }, + { + "epoch": 5.09, + "learning_rate": 1.1763608711747864e-06, + "loss": 1.0041, + "step": 68915 + }, + { + "epoch": 5.09, + "learning_rate": 1.1761731821080024e-06, + "loss": 1.0897, + "step": 68916 + }, + { + "epoch": 5.09, + "learning_rate": 1.1759855070798e-06, + "loss": 0.9916, + "step": 68917 + }, + { + "epoch": 5.09, + "learning_rate": 1.1757978460904784e-06, + "loss": 0.991, + "step": 68918 + }, + { + "epoch": 5.09, + "learning_rate": 1.1756101991403356e-06, + "loss": 0.8682, + "step": 68919 + }, + { + "epoch": 5.09, + "learning_rate": 1.1754225662296704e-06, + "loss": 0.9271, + "step": 68920 + }, + { + "epoch": 5.09, + "learning_rate": 1.175234947358782e-06, + "loss": 0.9634, + "step": 68921 + }, + { + "epoch": 5.09, + "learning_rate": 1.1750473425279696e-06, + "loss": 1.0466, + "step": 68922 + }, + { + "epoch": 5.09, + "learning_rate": 1.1748597517375304e-06, + "loss": 0.9236, + "step": 68923 + }, + { + "epoch": 5.09, + "learning_rate": 1.174672174987761e-06, + "loss": 1.0769, + "step": 68924 + }, + { + "epoch": 5.09, + "learning_rate": 1.1744846122789633e-06, + "loss": 0.9757, + "step": 68925 + }, + { + "epoch": 5.09, + "learning_rate": 1.1742970636114347e-06, + "loss": 0.9841, + "step": 68926 + }, + { + "epoch": 5.09, + "learning_rate": 1.1741095289854732e-06, + "loss": 0.8788, + "step": 68927 + }, + { + "epoch": 5.09, + "learning_rate": 1.1739220084013768e-06, + "loss": 0.9441, + "step": 68928 + }, + { + "epoch": 5.09, + "learning_rate": 1.1737345018594437e-06, + "loss": 0.9803, + "step": 68929 + }, + { + "epoch": 5.09, + "learning_rate": 1.1735470093599731e-06, + "loss": 1.0579, + "step": 68930 + }, + { + "epoch": 5.09, + "learning_rate": 1.173359530903264e-06, + "loss": 1.0682, + "step": 68931 + }, + { + "epoch": 5.09, + "learning_rate": 1.173172066489613e-06, + "loss": 1.0242, + "step": 68932 + }, + { + "epoch": 5.09, + "learning_rate": 1.172984616119317e-06, + "loss": 1.0439, + "step": 68933 + }, + { + "epoch": 5.09, + "learning_rate": 1.1727971797926785e-06, + "loss": 0.996, + "step": 68934 + }, + { + "epoch": 5.09, + "learning_rate": 1.1726097575099949e-06, + "loss": 1.0155, + "step": 68935 + }, + { + "epoch": 5.09, + "learning_rate": 1.1724223492715592e-06, + "loss": 1.0064, + "step": 68936 + }, + { + "epoch": 5.09, + "learning_rate": 1.1722349550776747e-06, + "loss": 0.9139, + "step": 68937 + }, + { + "epoch": 5.09, + "learning_rate": 1.1720475749286365e-06, + "loss": 0.9917, + "step": 68938 + }, + { + "epoch": 5.09, + "learning_rate": 1.1718602088247488e-06, + "loss": 1.1336, + "step": 68939 + }, + { + "epoch": 5.09, + "learning_rate": 1.1716728567663005e-06, + "loss": 1.0053, + "step": 68940 + }, + { + "epoch": 5.09, + "learning_rate": 1.1714855187535966e-06, + "loss": 1.0601, + "step": 68941 + }, + { + "epoch": 5.09, + "learning_rate": 1.1712981947869306e-06, + "loss": 1.0617, + "step": 68942 + }, + { + "epoch": 5.09, + "learning_rate": 1.1711108848666074e-06, + "loss": 1.0618, + "step": 68943 + }, + { + "epoch": 5.09, + "learning_rate": 1.1709235889929182e-06, + "loss": 1.0179, + "step": 68944 + }, + { + "epoch": 5.09, + "learning_rate": 1.1707363071661616e-06, + "loss": 0.8951, + "step": 68945 + }, + { + "epoch": 5.09, + "learning_rate": 1.1705490393866381e-06, + "loss": 0.9741, + "step": 68946 + }, + { + "epoch": 5.09, + "learning_rate": 1.1703617856546457e-06, + "loss": 0.9282, + "step": 68947 + }, + { + "epoch": 5.09, + "learning_rate": 1.1701745459704816e-06, + "loss": 0.9534, + "step": 68948 + }, + { + "epoch": 5.09, + "learning_rate": 1.1699873203344403e-06, + "loss": 0.9444, + "step": 68949 + }, + { + "epoch": 5.09, + "learning_rate": 1.1698001087468257e-06, + "loss": 0.9327, + "step": 68950 + }, + { + "epoch": 5.09, + "learning_rate": 1.1696129112079336e-06, + "loss": 0.962, + "step": 68951 + }, + { + "epoch": 5.09, + "learning_rate": 1.16942572771806e-06, + "loss": 1.0014, + "step": 68952 + }, + { + "epoch": 5.09, + "learning_rate": 1.1692385582775035e-06, + "loss": 1.0619, + "step": 68953 + }, + { + "epoch": 5.09, + "learning_rate": 1.1690514028865606e-06, + "loss": 0.9714, + "step": 68954 + }, + { + "epoch": 5.1, + "learning_rate": 1.1688642615455326e-06, + "loss": 0.9273, + "step": 68955 + }, + { + "epoch": 5.1, + "learning_rate": 1.1686771342547142e-06, + "loss": 1.0354, + "step": 68956 + }, + { + "epoch": 5.1, + "learning_rate": 1.1684900210144057e-06, + "loss": 0.8944, + "step": 68957 + }, + { + "epoch": 5.1, + "learning_rate": 1.1683029218248998e-06, + "loss": 0.9228, + "step": 68958 + }, + { + "epoch": 5.1, + "learning_rate": 1.168115836686503e-06, + "loss": 1.1983, + "step": 68959 + }, + { + "epoch": 5.1, + "learning_rate": 1.1679287655995019e-06, + "loss": 1.0017, + "step": 68960 + }, + { + "epoch": 5.1, + "learning_rate": 1.1677417085642029e-06, + "loss": 1.0974, + "step": 68961 + }, + { + "epoch": 5.1, + "learning_rate": 1.1675546655808978e-06, + "loss": 0.9997, + "step": 68962 + }, + { + "epoch": 5.1, + "learning_rate": 1.1673676366498921e-06, + "loss": 0.8643, + "step": 68963 + }, + { + "epoch": 5.1, + "learning_rate": 1.1671806217714754e-06, + "loss": 1.0885, + "step": 68964 + }, + { + "epoch": 5.1, + "learning_rate": 1.1669936209459454e-06, + "loss": 1.0108, + "step": 68965 + }, + { + "epoch": 5.1, + "learning_rate": 1.166806634173605e-06, + "loss": 0.9909, + "step": 68966 + }, + { + "epoch": 5.1, + "learning_rate": 1.1666196614547487e-06, + "loss": 0.9665, + "step": 68967 + }, + { + "epoch": 5.1, + "learning_rate": 1.1664327027896738e-06, + "loss": 1.186, + "step": 68968 + }, + { + "epoch": 5.1, + "learning_rate": 1.1662457581786756e-06, + "loss": 0.8637, + "step": 68969 + }, + { + "epoch": 5.1, + "learning_rate": 1.1660588276220563e-06, + "loss": 1.07, + "step": 68970 + }, + { + "epoch": 5.1, + "learning_rate": 1.1658719111201122e-06, + "loss": 1.0104, + "step": 68971 + }, + { + "epoch": 5.1, + "learning_rate": 1.1656850086731385e-06, + "loss": 0.8503, + "step": 68972 + }, + { + "epoch": 5.1, + "learning_rate": 1.1654981202814331e-06, + "loss": 1.0237, + "step": 68973 + }, + { + "epoch": 5.1, + "learning_rate": 1.1653112459452931e-06, + "loss": 0.8752, + "step": 68974 + }, + { + "epoch": 5.1, + "learning_rate": 1.1651243856650173e-06, + "loss": 1.007, + "step": 68975 + }, + { + "epoch": 5.1, + "learning_rate": 1.1649375394409024e-06, + "loss": 0.9585, + "step": 68976 + }, + { + "epoch": 5.1, + "learning_rate": 1.1647507072732455e-06, + "loss": 1.0382, + "step": 68977 + }, + { + "epoch": 5.1, + "learning_rate": 1.1645638891623424e-06, + "loss": 1.0515, + "step": 68978 + }, + { + "epoch": 5.1, + "learning_rate": 1.1643770851084935e-06, + "loss": 0.966, + "step": 68979 + }, + { + "epoch": 5.1, + "learning_rate": 1.1641902951119943e-06, + "loss": 0.8974, + "step": 68980 + }, + { + "epoch": 5.1, + "learning_rate": 1.1640035191731413e-06, + "loss": 0.9488, + "step": 68981 + }, + { + "epoch": 5.1, + "learning_rate": 1.1638167572922332e-06, + "loss": 0.9849, + "step": 68982 + }, + { + "epoch": 5.1, + "learning_rate": 1.163630009469563e-06, + "loss": 1.0209, + "step": 68983 + }, + { + "epoch": 5.1, + "learning_rate": 1.1634432757054358e-06, + "loss": 1.029, + "step": 68984 + }, + { + "epoch": 5.1, + "learning_rate": 1.1632565560001407e-06, + "loss": 1.0581, + "step": 68985 + }, + { + "epoch": 5.1, + "learning_rate": 1.1630698503539783e-06, + "loss": 1.0013, + "step": 68986 + }, + { + "epoch": 5.1, + "learning_rate": 1.1628831587672452e-06, + "loss": 0.9862, + "step": 68987 + }, + { + "epoch": 5.1, + "learning_rate": 1.1626964812402407e-06, + "loss": 1.0675, + "step": 68988 + }, + { + "epoch": 5.1, + "learning_rate": 1.1625098177732564e-06, + "loss": 1.0807, + "step": 68989 + }, + { + "epoch": 5.1, + "learning_rate": 1.1623231683665949e-06, + "loss": 0.9692, + "step": 68990 + }, + { + "epoch": 5.1, + "learning_rate": 1.1621365330205502e-06, + "loss": 1.0077, + "step": 68991 + }, + { + "epoch": 5.1, + "learning_rate": 1.1619499117354204e-06, + "loss": 0.9935, + "step": 68992 + }, + { + "epoch": 5.1, + "learning_rate": 1.1617633045115017e-06, + "loss": 1.0212, + "step": 68993 + }, + { + "epoch": 5.1, + "learning_rate": 1.1615767113490895e-06, + "loss": 1.0031, + "step": 68994 + }, + { + "epoch": 5.1, + "learning_rate": 1.1613901322484833e-06, + "loss": 0.9882, + "step": 68995 + }, + { + "epoch": 5.1, + "learning_rate": 1.1612035672099798e-06, + "loss": 1.0236, + "step": 68996 + }, + { + "epoch": 5.1, + "learning_rate": 1.1610170162338741e-06, + "loss": 1.0057, + "step": 68997 + }, + { + "epoch": 5.1, + "learning_rate": 1.1608304793204627e-06, + "loss": 0.9807, + "step": 68998 + }, + { + "epoch": 5.1, + "learning_rate": 1.1606439564700456e-06, + "loss": 0.9375, + "step": 68999 + }, + { + "epoch": 5.1, + "learning_rate": 1.1604574476829168e-06, + "loss": 1.0344, + "step": 69000 + }, + { + "epoch": 5.1, + "learning_rate": 1.1602709529593737e-06, + "loss": 0.9965, + "step": 69001 + }, + { + "epoch": 5.1, + "learning_rate": 1.160084472299714e-06, + "loss": 0.9436, + "step": 69002 + }, + { + "epoch": 5.1, + "learning_rate": 1.1598980057042308e-06, + "loss": 1.1322, + "step": 69003 + }, + { + "epoch": 5.1, + "learning_rate": 1.1597115531732273e-06, + "loss": 1.0825, + "step": 69004 + }, + { + "epoch": 5.1, + "learning_rate": 1.1595251147069918e-06, + "loss": 0.9655, + "step": 69005 + }, + { + "epoch": 5.1, + "learning_rate": 1.1593386903058278e-06, + "loss": 1.0519, + "step": 69006 + }, + { + "epoch": 5.1, + "learning_rate": 1.1591522799700283e-06, + "loss": 1.0068, + "step": 69007 + }, + { + "epoch": 5.1, + "learning_rate": 1.158965883699894e-06, + "loss": 1.0978, + "step": 69008 + }, + { + "epoch": 5.1, + "learning_rate": 1.158779501495717e-06, + "loss": 1.0418, + "step": 69009 + }, + { + "epoch": 5.1, + "learning_rate": 1.1585931333577926e-06, + "loss": 1.0476, + "step": 69010 + }, + { + "epoch": 5.1, + "learning_rate": 1.1584067792864228e-06, + "loss": 1.0579, + "step": 69011 + }, + { + "epoch": 5.1, + "learning_rate": 1.1582204392819007e-06, + "loss": 1.0005, + "step": 69012 + }, + { + "epoch": 5.1, + "learning_rate": 1.1580341133445238e-06, + "loss": 0.97, + "step": 69013 + }, + { + "epoch": 5.1, + "learning_rate": 1.1578478014745864e-06, + "loss": 0.9465, + "step": 69014 + }, + { + "epoch": 5.1, + "learning_rate": 1.1576615036723883e-06, + "loss": 1.0019, + "step": 69015 + }, + { + "epoch": 5.1, + "learning_rate": 1.1574752199382244e-06, + "loss": 0.9264, + "step": 69016 + }, + { + "epoch": 5.1, + "learning_rate": 1.1572889502723906e-06, + "loss": 1.0504, + "step": 69017 + }, + { + "epoch": 5.1, + "learning_rate": 1.1571026946751818e-06, + "loss": 1.0525, + "step": 69018 + }, + { + "epoch": 5.1, + "learning_rate": 1.156916453146898e-06, + "loss": 0.874, + "step": 69019 + }, + { + "epoch": 5.1, + "learning_rate": 1.1567302256878332e-06, + "loss": 0.984, + "step": 69020 + }, + { + "epoch": 5.1, + "learning_rate": 1.156544012298284e-06, + "loss": 1.0295, + "step": 69021 + }, + { + "epoch": 5.1, + "learning_rate": 1.1563578129785468e-06, + "loss": 0.9351, + "step": 69022 + }, + { + "epoch": 5.1, + "learning_rate": 1.156171627728917e-06, + "loss": 1.0547, + "step": 69023 + }, + { + "epoch": 5.1, + "learning_rate": 1.155985456549692e-06, + "loss": 1.0828, + "step": 69024 + }, + { + "epoch": 5.1, + "learning_rate": 1.1557992994411694e-06, + "loss": 0.9456, + "step": 69025 + }, + { + "epoch": 5.1, + "learning_rate": 1.1556131564036422e-06, + "loss": 1.0848, + "step": 69026 + }, + { + "epoch": 5.1, + "learning_rate": 1.1554270274374058e-06, + "loss": 1.1917, + "step": 69027 + }, + { + "epoch": 5.1, + "learning_rate": 1.155240912542761e-06, + "loss": 1.0296, + "step": 69028 + }, + { + "epoch": 5.1, + "learning_rate": 1.1550548117200032e-06, + "loss": 1.0208, + "step": 69029 + }, + { + "epoch": 5.1, + "learning_rate": 1.1548687249694235e-06, + "loss": 0.8334, + "step": 69030 + }, + { + "epoch": 5.1, + "learning_rate": 1.1546826522913225e-06, + "loss": 1.0013, + "step": 69031 + }, + { + "epoch": 5.1, + "learning_rate": 1.154496593685992e-06, + "loss": 1.0191, + "step": 69032 + }, + { + "epoch": 5.1, + "learning_rate": 1.1543105491537354e-06, + "loss": 0.9809, + "step": 69033 + }, + { + "epoch": 5.1, + "learning_rate": 1.1541245186948403e-06, + "loss": 0.8726, + "step": 69034 + }, + { + "epoch": 5.1, + "learning_rate": 1.1539385023096083e-06, + "loss": 1.0338, + "step": 69035 + }, + { + "epoch": 5.1, + "learning_rate": 1.153752499998334e-06, + "loss": 0.9511, + "step": 69036 + }, + { + "epoch": 5.1, + "learning_rate": 1.1535665117613126e-06, + "loss": 1.0279, + "step": 69037 + }, + { + "epoch": 5.1, + "learning_rate": 1.1533805375988405e-06, + "loss": 1.0601, + "step": 69038 + }, + { + "epoch": 5.1, + "learning_rate": 1.1531945775112119e-06, + "loss": 0.8497, + "step": 69039 + }, + { + "epoch": 5.1, + "learning_rate": 1.1530086314987255e-06, + "loss": 1.0103, + "step": 69040 + }, + { + "epoch": 5.1, + "learning_rate": 1.1528226995616755e-06, + "loss": 1.0686, + "step": 69041 + }, + { + "epoch": 5.1, + "learning_rate": 1.152636781700358e-06, + "loss": 1.0124, + "step": 69042 + }, + { + "epoch": 5.1, + "learning_rate": 1.1524508779150679e-06, + "loss": 0.9598, + "step": 69043 + }, + { + "epoch": 5.1, + "learning_rate": 1.1522649882061032e-06, + "loss": 1.0537, + "step": 69044 + }, + { + "epoch": 5.1, + "learning_rate": 1.1520791125737574e-06, + "loss": 0.94, + "step": 69045 + }, + { + "epoch": 5.1, + "learning_rate": 1.1518932510183278e-06, + "loss": 1.0648, + "step": 69046 + }, + { + "epoch": 5.1, + "learning_rate": 1.15170740354011e-06, + "loss": 1.0072, + "step": 69047 + }, + { + "epoch": 5.1, + "learning_rate": 1.1515215701393966e-06, + "loss": 1.1055, + "step": 69048 + }, + { + "epoch": 5.1, + "learning_rate": 1.1513357508164902e-06, + "loss": 0.9821, + "step": 69049 + }, + { + "epoch": 5.1, + "learning_rate": 1.1511499455716779e-06, + "loss": 0.8759, + "step": 69050 + }, + { + "epoch": 5.1, + "learning_rate": 1.1509641544052609e-06, + "loss": 0.9794, + "step": 69051 + }, + { + "epoch": 5.1, + "learning_rate": 1.1507783773175318e-06, + "loss": 0.9628, + "step": 69052 + }, + { + "epoch": 5.1, + "learning_rate": 1.150592614308792e-06, + "loss": 1.0834, + "step": 69053 + }, + { + "epoch": 5.1, + "learning_rate": 1.1504068653793277e-06, + "loss": 0.9526, + "step": 69054 + }, + { + "epoch": 5.1, + "learning_rate": 1.150221130529442e-06, + "loss": 1.032, + "step": 69055 + }, + { + "epoch": 5.1, + "learning_rate": 1.150035409759428e-06, + "loss": 1.0141, + "step": 69056 + }, + { + "epoch": 5.1, + "learning_rate": 1.149849703069581e-06, + "loss": 0.9413, + "step": 69057 + }, + { + "epoch": 5.1, + "learning_rate": 1.1496640104601965e-06, + "loss": 0.9979, + "step": 69058 + }, + { + "epoch": 5.1, + "learning_rate": 1.1494783319315683e-06, + "loss": 0.9308, + "step": 69059 + }, + { + "epoch": 5.1, + "learning_rate": 1.1492926674839943e-06, + "loss": 0.9991, + "step": 69060 + }, + { + "epoch": 5.1, + "learning_rate": 1.1491070171177698e-06, + "loss": 0.9178, + "step": 69061 + }, + { + "epoch": 5.1, + "learning_rate": 1.14892138083319e-06, + "loss": 0.9913, + "step": 69062 + }, + { + "epoch": 5.1, + "learning_rate": 1.148735758630547e-06, + "loss": 0.9401, + "step": 69063 + }, + { + "epoch": 5.1, + "learning_rate": 1.1485501505101403e-06, + "loss": 0.9766, + "step": 69064 + }, + { + "epoch": 5.1, + "learning_rate": 1.1483645564722645e-06, + "loss": 1.0108, + "step": 69065 + }, + { + "epoch": 5.1, + "learning_rate": 1.1481789765172136e-06, + "loss": 1.0033, + "step": 69066 + }, + { + "epoch": 5.1, + "learning_rate": 1.1479934106452827e-06, + "loss": 1.074, + "step": 69067 + }, + { + "epoch": 5.1, + "learning_rate": 1.1478078588567666e-06, + "loss": 1.0691, + "step": 69068 + }, + { + "epoch": 5.1, + "learning_rate": 1.1476223211519632e-06, + "loss": 0.9964, + "step": 69069 + }, + { + "epoch": 5.1, + "learning_rate": 1.1474367975311663e-06, + "loss": 0.9589, + "step": 69070 + }, + { + "epoch": 5.1, + "learning_rate": 1.1472512879946706e-06, + "loss": 1.0134, + "step": 69071 + }, + { + "epoch": 5.1, + "learning_rate": 1.1470657925427698e-06, + "loss": 0.9601, + "step": 69072 + }, + { + "epoch": 5.1, + "learning_rate": 1.1468803111757621e-06, + "loss": 1.0991, + "step": 69073 + }, + { + "epoch": 5.1, + "learning_rate": 1.1466948438939418e-06, + "loss": 1.0284, + "step": 69074 + }, + { + "epoch": 5.1, + "learning_rate": 1.146509390697602e-06, + "loss": 0.93, + "step": 69075 + }, + { + "epoch": 5.1, + "learning_rate": 1.1463239515870405e-06, + "loss": 1.0509, + "step": 69076 + }, + { + "epoch": 5.1, + "learning_rate": 1.146138526562548e-06, + "loss": 1.0859, + "step": 69077 + }, + { + "epoch": 5.1, + "learning_rate": 1.1459531156244274e-06, + "loss": 1.0276, + "step": 69078 + }, + { + "epoch": 5.1, + "learning_rate": 1.1457677187729643e-06, + "loss": 0.9052, + "step": 69079 + }, + { + "epoch": 5.1, + "learning_rate": 1.1455823360084595e-06, + "loss": 0.9641, + "step": 69080 + }, + { + "epoch": 5.1, + "learning_rate": 1.145396967331207e-06, + "loss": 0.9682, + "step": 69081 + }, + { + "epoch": 5.1, + "learning_rate": 1.1452116127415014e-06, + "loss": 1.1105, + "step": 69082 + }, + { + "epoch": 5.1, + "learning_rate": 1.1450262722396344e-06, + "loss": 0.9492, + "step": 69083 + }, + { + "epoch": 5.1, + "learning_rate": 1.144840945825907e-06, + "loss": 1.0093, + "step": 69084 + }, + { + "epoch": 5.1, + "learning_rate": 1.144655633500611e-06, + "loss": 0.9222, + "step": 69085 + }, + { + "epoch": 5.1, + "learning_rate": 1.1444703352640408e-06, + "loss": 1.0302, + "step": 69086 + }, + { + "epoch": 5.1, + "learning_rate": 1.1442850511164917e-06, + "loss": 1.0361, + "step": 69087 + }, + { + "epoch": 5.1, + "learning_rate": 1.1440997810582555e-06, + "loss": 0.9857, + "step": 69088 + }, + { + "epoch": 5.1, + "learning_rate": 1.1439145250896333e-06, + "loss": 0.9309, + "step": 69089 + }, + { + "epoch": 5.11, + "learning_rate": 1.1437292832109158e-06, + "loss": 1.0221, + "step": 69090 + }, + { + "epoch": 5.11, + "learning_rate": 1.1435440554223975e-06, + "loss": 1.0075, + "step": 69091 + }, + { + "epoch": 5.11, + "learning_rate": 1.1433588417243735e-06, + "loss": 0.9655, + "step": 69092 + }, + { + "epoch": 5.11, + "learning_rate": 1.1431736421171402e-06, + "loss": 1.0297, + "step": 69093 + }, + { + "epoch": 5.11, + "learning_rate": 1.1429884566009918e-06, + "loss": 1.0117, + "step": 69094 + }, + { + "epoch": 5.11, + "learning_rate": 1.1428032851762193e-06, + "loss": 1.042, + "step": 69095 + }, + { + "epoch": 5.11, + "learning_rate": 1.1426181278431214e-06, + "loss": 0.9942, + "step": 69096 + }, + { + "epoch": 5.11, + "learning_rate": 1.14243298460199e-06, + "loss": 1.2207, + "step": 69097 + }, + { + "epoch": 5.11, + "learning_rate": 1.1422478554531246e-06, + "loss": 0.9704, + "step": 69098 + }, + { + "epoch": 5.11, + "learning_rate": 1.142062740396812e-06, + "loss": 1.0154, + "step": 69099 + }, + { + "epoch": 5.11, + "learning_rate": 1.141877639433353e-06, + "loss": 0.9833, + "step": 69100 + }, + { + "epoch": 5.11, + "learning_rate": 1.1416925525630384e-06, + "loss": 0.9571, + "step": 69101 + }, + { + "epoch": 5.11, + "learning_rate": 1.141507479786167e-06, + "loss": 1.0808, + "step": 69102 + }, + { + "epoch": 5.11, + "learning_rate": 1.1413224211030294e-06, + "loss": 0.9521, + "step": 69103 + }, + { + "epoch": 5.11, + "learning_rate": 1.1411373765139188e-06, + "loss": 1.0605, + "step": 69104 + }, + { + "epoch": 5.11, + "learning_rate": 1.1409523460191342e-06, + "loss": 1.0798, + "step": 69105 + }, + { + "epoch": 5.11, + "learning_rate": 1.140767329618967e-06, + "loss": 1.0143, + "step": 69106 + }, + { + "epoch": 5.11, + "learning_rate": 1.140582327313713e-06, + "loss": 1.0249, + "step": 69107 + }, + { + "epoch": 5.11, + "learning_rate": 1.1403973391036638e-06, + "loss": 0.9398, + "step": 69108 + }, + { + "epoch": 5.11, + "learning_rate": 1.1402123649891173e-06, + "loss": 0.9146, + "step": 69109 + }, + { + "epoch": 5.11, + "learning_rate": 1.1400274049703664e-06, + "loss": 1.0705, + "step": 69110 + }, + { + "epoch": 5.11, + "learning_rate": 1.139842459047704e-06, + "loss": 0.9674, + "step": 69111 + }, + { + "epoch": 5.11, + "learning_rate": 1.1396575272214272e-06, + "loss": 0.9493, + "step": 69112 + }, + { + "epoch": 5.11, + "learning_rate": 1.1394726094918251e-06, + "loss": 1.0084, + "step": 69113 + }, + { + "epoch": 5.11, + "learning_rate": 1.1392877058591978e-06, + "loss": 1.0206, + "step": 69114 + }, + { + "epoch": 5.11, + "learning_rate": 1.139102816323837e-06, + "loss": 0.9395, + "step": 69115 + }, + { + "epoch": 5.11, + "learning_rate": 1.1389179408860374e-06, + "loss": 1.0628, + "step": 69116 + }, + { + "epoch": 5.11, + "learning_rate": 1.1387330795460894e-06, + "loss": 1.0575, + "step": 69117 + }, + { + "epoch": 5.11, + "learning_rate": 1.1385482323042928e-06, + "loss": 1.1069, + "step": 69118 + }, + { + "epoch": 5.11, + "learning_rate": 1.13836339916094e-06, + "loss": 1.0038, + "step": 69119 + }, + { + "epoch": 5.11, + "learning_rate": 1.1381785801163236e-06, + "loss": 0.989, + "step": 69120 + }, + { + "epoch": 5.11, + "learning_rate": 1.137993775170736e-06, + "loss": 0.9463, + "step": 69121 + }, + { + "epoch": 5.11, + "learning_rate": 1.1378089843244756e-06, + "loss": 1.0686, + "step": 69122 + }, + { + "epoch": 5.11, + "learning_rate": 1.1376242075778365e-06, + "loss": 0.9985, + "step": 69123 + }, + { + "epoch": 5.11, + "learning_rate": 1.1374394449311065e-06, + "loss": 0.9485, + "step": 69124 + }, + { + "epoch": 5.11, + "learning_rate": 1.1372546963845854e-06, + "loss": 0.8945, + "step": 69125 + }, + { + "epoch": 5.11, + "learning_rate": 1.1370699619385662e-06, + "loss": 1.0393, + "step": 69126 + }, + { + "epoch": 5.11, + "learning_rate": 1.1368852415933407e-06, + "loss": 1.0461, + "step": 69127 + }, + { + "epoch": 5.11, + "learning_rate": 1.1367005353492022e-06, + "loss": 1.0056, + "step": 69128 + }, + { + "epoch": 5.11, + "learning_rate": 1.1365158432064482e-06, + "loss": 1.0317, + "step": 69129 + }, + { + "epoch": 5.11, + "learning_rate": 1.1363311651653708e-06, + "loss": 1.047, + "step": 69130 + }, + { + "epoch": 5.11, + "learning_rate": 1.136146501226264e-06, + "loss": 1.0527, + "step": 69131 + }, + { + "epoch": 5.11, + "learning_rate": 1.135961851389421e-06, + "loss": 0.9513, + "step": 69132 + }, + { + "epoch": 5.11, + "learning_rate": 1.1357772156551339e-06, + "loss": 1.0311, + "step": 69133 + }, + { + "epoch": 5.11, + "learning_rate": 1.135592594023701e-06, + "loss": 0.9332, + "step": 69134 + }, + { + "epoch": 5.11, + "learning_rate": 1.1354079864954126e-06, + "loss": 0.9965, + "step": 69135 + }, + { + "epoch": 5.11, + "learning_rate": 1.1352233930705636e-06, + "loss": 1.1219, + "step": 69136 + }, + { + "epoch": 5.11, + "learning_rate": 1.135038813749446e-06, + "loss": 1.0452, + "step": 69137 + }, + { + "epoch": 5.11, + "learning_rate": 1.1348542485323555e-06, + "loss": 0.8808, + "step": 69138 + }, + { + "epoch": 5.11, + "learning_rate": 1.1346696974195858e-06, + "loss": 0.9878, + "step": 69139 + }, + { + "epoch": 5.11, + "learning_rate": 1.1344851604114304e-06, + "loss": 1.1789, + "step": 69140 + }, + { + "epoch": 5.11, + "learning_rate": 1.1343006375081812e-06, + "loss": 1.0105, + "step": 69141 + }, + { + "epoch": 5.11, + "learning_rate": 1.134116128710131e-06, + "loss": 0.9072, + "step": 69142 + }, + { + "epoch": 5.11, + "learning_rate": 1.1339316340175798e-06, + "loss": 1.1004, + "step": 69143 + }, + { + "epoch": 5.11, + "learning_rate": 1.133747153430812e-06, + "loss": 1.0513, + "step": 69144 + }, + { + "epoch": 5.11, + "learning_rate": 1.133562686950128e-06, + "loss": 0.957, + "step": 69145 + }, + { + "epoch": 5.11, + "learning_rate": 1.133378234575816e-06, + "loss": 0.981, + "step": 69146 + }, + { + "epoch": 5.11, + "learning_rate": 1.1331937963081774e-06, + "loss": 1.1127, + "step": 69147 + }, + { + "epoch": 5.11, + "learning_rate": 1.1330093721474967e-06, + "loss": 0.9218, + "step": 69148 + }, + { + "epoch": 5.11, + "learning_rate": 1.1328249620940724e-06, + "loss": 0.9521, + "step": 69149 + }, + { + "epoch": 5.11, + "learning_rate": 1.1326405661481977e-06, + "loss": 1.0801, + "step": 69150 + }, + { + "epoch": 5.11, + "learning_rate": 1.1324561843101634e-06, + "loss": 0.9467, + "step": 69151 + }, + { + "epoch": 5.11, + "learning_rate": 1.132271816580266e-06, + "loss": 0.9881, + "step": 69152 + }, + { + "epoch": 5.11, + "learning_rate": 1.132087462958794e-06, + "loss": 0.9916, + "step": 69153 + }, + { + "epoch": 5.11, + "learning_rate": 1.1319031234460477e-06, + "loss": 0.9872, + "step": 69154 + }, + { + "epoch": 5.11, + "learning_rate": 1.131718798042315e-06, + "loss": 0.9822, + "step": 69155 + }, + { + "epoch": 5.11, + "learning_rate": 1.1315344867478916e-06, + "loss": 1.094, + "step": 69156 + }, + { + "epoch": 5.11, + "learning_rate": 1.1313501895630674e-06, + "loss": 1.0277, + "step": 69157 + }, + { + "epoch": 5.11, + "learning_rate": 1.131165906488141e-06, + "loss": 0.9495, + "step": 69158 + }, + { + "epoch": 5.11, + "learning_rate": 1.1309816375234018e-06, + "loss": 0.9314, + "step": 69159 + }, + { + "epoch": 5.11, + "learning_rate": 1.1307973826691442e-06, + "loss": 0.9834, + "step": 69160 + }, + { + "epoch": 5.11, + "learning_rate": 1.1306131419256617e-06, + "loss": 0.9823, + "step": 69161 + }, + { + "epoch": 5.11, + "learning_rate": 1.1304289152932434e-06, + "loss": 1.0104, + "step": 69162 + }, + { + "epoch": 5.11, + "learning_rate": 1.1302447027721897e-06, + "loss": 0.9503, + "step": 69163 + }, + { + "epoch": 5.11, + "learning_rate": 1.1300605043627887e-06, + "loss": 1.1194, + "step": 69164 + }, + { + "epoch": 5.11, + "learning_rate": 1.129876320065334e-06, + "loss": 1.066, + "step": 69165 + }, + { + "epoch": 5.11, + "learning_rate": 1.1296921498801182e-06, + "loss": 1.0057, + "step": 69166 + }, + { + "epoch": 5.11, + "learning_rate": 1.1295079938074393e-06, + "loss": 0.9441, + "step": 69167 + }, + { + "epoch": 5.11, + "learning_rate": 1.1293238518475835e-06, + "loss": 1.0366, + "step": 69168 + }, + { + "epoch": 5.11, + "learning_rate": 1.1291397240008451e-06, + "loss": 1.054, + "step": 69169 + }, + { + "epoch": 5.11, + "learning_rate": 1.1289556102675204e-06, + "loss": 0.911, + "step": 69170 + }, + { + "epoch": 5.11, + "learning_rate": 1.1287715106479003e-06, + "loss": 0.9947, + "step": 69171 + }, + { + "epoch": 5.11, + "learning_rate": 1.1285874251422769e-06, + "loss": 1.0397, + "step": 69172 + }, + { + "epoch": 5.11, + "learning_rate": 1.128403353750943e-06, + "loss": 0.9492, + "step": 69173 + }, + { + "epoch": 5.11, + "learning_rate": 1.1282192964741944e-06, + "loss": 1.1281, + "step": 69174 + }, + { + "epoch": 5.11, + "learning_rate": 1.1280352533123217e-06, + "loss": 1.0244, + "step": 69175 + }, + { + "epoch": 5.11, + "learning_rate": 1.127851224265618e-06, + "loss": 0.9211, + "step": 69176 + }, + { + "epoch": 5.11, + "learning_rate": 1.1276672093343743e-06, + "loss": 1.0507, + "step": 69177 + }, + { + "epoch": 5.11, + "learning_rate": 1.127483208518887e-06, + "loss": 1.0519, + "step": 69178 + }, + { + "epoch": 5.11, + "learning_rate": 1.1272992218194468e-06, + "loss": 0.9835, + "step": 69179 + }, + { + "epoch": 5.11, + "learning_rate": 1.1271152492363469e-06, + "loss": 1.0459, + "step": 69180 + }, + { + "epoch": 5.11, + "learning_rate": 1.1269312907698792e-06, + "loss": 1.0014, + "step": 69181 + }, + { + "epoch": 5.11, + "learning_rate": 1.126747346420335e-06, + "loss": 1.0178, + "step": 69182 + }, + { + "epoch": 5.11, + "learning_rate": 1.1265634161880112e-06, + "loss": 0.9998, + "step": 69183 + }, + { + "epoch": 5.11, + "learning_rate": 1.1263795000731981e-06, + "loss": 0.9939, + "step": 69184 + }, + { + "epoch": 5.11, + "learning_rate": 1.1261955980761885e-06, + "loss": 1.0432, + "step": 69185 + }, + { + "epoch": 5.11, + "learning_rate": 1.1260117101972723e-06, + "loss": 0.9752, + "step": 69186 + }, + { + "epoch": 5.11, + "learning_rate": 1.125827836436746e-06, + "loss": 1.0706, + "step": 69187 + }, + { + "epoch": 5.11, + "learning_rate": 1.1256439767949035e-06, + "loss": 1.0016, + "step": 69188 + }, + { + "epoch": 5.11, + "learning_rate": 1.1254601312720293e-06, + "loss": 0.8696, + "step": 69189 + }, + { + "epoch": 5.11, + "learning_rate": 1.125276299868424e-06, + "loss": 1.0429, + "step": 69190 + }, + { + "epoch": 5.11, + "learning_rate": 1.1250924825843745e-06, + "loss": 0.9841, + "step": 69191 + }, + { + "epoch": 5.11, + "learning_rate": 1.1249086794201813e-06, + "loss": 0.9602, + "step": 69192 + }, + { + "epoch": 5.11, + "learning_rate": 1.1247248903761265e-06, + "loss": 1.0135, + "step": 69193 + }, + { + "epoch": 5.11, + "learning_rate": 1.1245411154525087e-06, + "loss": 0.9329, + "step": 69194 + }, + { + "epoch": 5.11, + "learning_rate": 1.1243573546496178e-06, + "loss": 0.9332, + "step": 69195 + }, + { + "epoch": 5.11, + "learning_rate": 1.1241736079677513e-06, + "loss": 1.0405, + "step": 69196 + }, + { + "epoch": 5.11, + "learning_rate": 1.1239898754071954e-06, + "loss": 1.0664, + "step": 69197 + }, + { + "epoch": 5.11, + "learning_rate": 1.1238061569682423e-06, + "loss": 1.0313, + "step": 69198 + }, + { + "epoch": 5.11, + "learning_rate": 1.1236224526511897e-06, + "loss": 0.9873, + "step": 69199 + }, + { + "epoch": 5.11, + "learning_rate": 1.1234387624563258e-06, + "loss": 0.9498, + "step": 69200 + }, + { + "epoch": 5.11, + "learning_rate": 1.1232550863839442e-06, + "loss": 1.0148, + "step": 69201 + }, + { + "epoch": 5.11, + "learning_rate": 1.1230714244343343e-06, + "loss": 0.967, + "step": 69202 + }, + { + "epoch": 5.11, + "learning_rate": 1.122887776607794e-06, + "loss": 0.9435, + "step": 69203 + }, + { + "epoch": 5.11, + "learning_rate": 1.1227041429046103e-06, + "loss": 1.0962, + "step": 69204 + }, + { + "epoch": 5.11, + "learning_rate": 1.122520523325079e-06, + "loss": 1.0103, + "step": 69205 + }, + { + "epoch": 5.11, + "learning_rate": 1.12233691786949e-06, + "loss": 1.0397, + "step": 69206 + }, + { + "epoch": 5.11, + "learning_rate": 1.1221533265381335e-06, + "loss": 0.9014, + "step": 69207 + }, + { + "epoch": 5.11, + "learning_rate": 1.1219697493313064e-06, + "loss": 1.0273, + "step": 69208 + }, + { + "epoch": 5.11, + "learning_rate": 1.1217861862492984e-06, + "loss": 0.9525, + "step": 69209 + }, + { + "epoch": 5.11, + "learning_rate": 1.1216026372924016e-06, + "loss": 0.9505, + "step": 69210 + }, + { + "epoch": 5.11, + "learning_rate": 1.1214191024609056e-06, + "loss": 0.9815, + "step": 69211 + }, + { + "epoch": 5.11, + "learning_rate": 1.121235581755109e-06, + "loss": 0.96, + "step": 69212 + }, + { + "epoch": 5.11, + "learning_rate": 1.1210520751752963e-06, + "loss": 1.0835, + "step": 69213 + }, + { + "epoch": 5.11, + "learning_rate": 1.1208685827217637e-06, + "loss": 0.9612, + "step": 69214 + }, + { + "epoch": 5.11, + "learning_rate": 1.120685104394803e-06, + "loss": 0.9521, + "step": 69215 + }, + { + "epoch": 5.11, + "learning_rate": 1.1205016401947044e-06, + "loss": 1.0243, + "step": 69216 + }, + { + "epoch": 5.11, + "learning_rate": 1.1203181901217608e-06, + "loss": 0.988, + "step": 69217 + }, + { + "epoch": 5.11, + "learning_rate": 1.120134754176262e-06, + "loss": 0.8743, + "step": 69218 + }, + { + "epoch": 5.11, + "learning_rate": 1.1199513323585043e-06, + "loss": 1.0496, + "step": 69219 + }, + { + "epoch": 5.11, + "learning_rate": 1.1197679246687765e-06, + "loss": 0.9905, + "step": 69220 + }, + { + "epoch": 5.11, + "learning_rate": 1.1195845311073705e-06, + "loss": 0.8193, + "step": 69221 + }, + { + "epoch": 5.11, + "learning_rate": 1.1194011516745773e-06, + "loss": 0.992, + "step": 69222 + }, + { + "epoch": 5.11, + "learning_rate": 1.119217786370691e-06, + "loss": 1.0094, + "step": 69223 + }, + { + "epoch": 5.11, + "learning_rate": 1.1190344351960024e-06, + "loss": 0.9488, + "step": 69224 + }, + { + "epoch": 5.11, + "learning_rate": 1.118851098150804e-06, + "loss": 1.1103, + "step": 69225 + }, + { + "epoch": 5.12, + "learning_rate": 1.1186677752353848e-06, + "loss": 0.9244, + "step": 69226 + }, + { + "epoch": 5.12, + "learning_rate": 1.1184844664500372e-06, + "loss": 1.0512, + "step": 69227 + }, + { + "epoch": 5.12, + "learning_rate": 1.1183011717950555e-06, + "loss": 0.8954, + "step": 69228 + }, + { + "epoch": 5.12, + "learning_rate": 1.118117891270729e-06, + "loss": 0.9417, + "step": 69229 + }, + { + "epoch": 5.12, + "learning_rate": 1.1179346248773503e-06, + "loss": 0.8635, + "step": 69230 + }, + { + "epoch": 5.12, + "learning_rate": 1.117751372615209e-06, + "loss": 0.9701, + "step": 69231 + }, + { + "epoch": 5.12, + "learning_rate": 1.117568134484599e-06, + "loss": 1.0667, + "step": 69232 + }, + { + "epoch": 5.12, + "learning_rate": 1.1173849104858126e-06, + "loss": 1.0015, + "step": 69233 + }, + { + "epoch": 5.12, + "learning_rate": 1.1172017006191382e-06, + "loss": 1.0046, + "step": 69234 + }, + { + "epoch": 5.12, + "learning_rate": 1.1170185048848691e-06, + "loss": 1.0561, + "step": 69235 + }, + { + "epoch": 5.12, + "learning_rate": 1.1168353232832952e-06, + "loss": 0.9689, + "step": 69236 + }, + { + "epoch": 5.12, + "learning_rate": 1.1166521558147125e-06, + "loss": 1.0636, + "step": 69237 + }, + { + "epoch": 5.12, + "learning_rate": 1.1164690024794056e-06, + "loss": 1.0349, + "step": 69238 + }, + { + "epoch": 5.12, + "learning_rate": 1.1162858632776708e-06, + "loss": 1.0716, + "step": 69239 + }, + { + "epoch": 5.12, + "learning_rate": 1.1161027382097967e-06, + "loss": 0.9669, + "step": 69240 + }, + { + "epoch": 5.12, + "learning_rate": 1.1159196272760797e-06, + "loss": 0.9226, + "step": 69241 + }, + { + "epoch": 5.12, + "learning_rate": 1.1157365304768043e-06, + "loss": 0.9957, + "step": 69242 + }, + { + "epoch": 5.12, + "learning_rate": 1.1155534478122664e-06, + "loss": 0.9378, + "step": 69243 + }, + { + "epoch": 5.12, + "learning_rate": 1.1153703792827552e-06, + "loss": 1.0034, + "step": 69244 + }, + { + "epoch": 5.12, + "learning_rate": 1.1151873248885637e-06, + "loss": 0.9897, + "step": 69245 + }, + { + "epoch": 5.12, + "learning_rate": 1.1150042846299814e-06, + "loss": 0.9581, + "step": 69246 + }, + { + "epoch": 5.12, + "learning_rate": 1.1148212585072981e-06, + "loss": 1.0065, + "step": 69247 + }, + { + "epoch": 5.12, + "learning_rate": 1.1146382465208093e-06, + "loss": 0.9447, + "step": 69248 + }, + { + "epoch": 5.12, + "learning_rate": 1.1144552486708039e-06, + "loss": 1.0603, + "step": 69249 + }, + { + "epoch": 5.12, + "learning_rate": 1.114272264957572e-06, + "loss": 0.9795, + "step": 69250 + }, + { + "epoch": 5.12, + "learning_rate": 1.1140892953814053e-06, + "loss": 0.9745, + "step": 69251 + }, + { + "epoch": 5.12, + "learning_rate": 1.1139063399425965e-06, + "loss": 1.1074, + "step": 69252 + }, + { + "epoch": 5.12, + "learning_rate": 1.1137233986414364e-06, + "loss": 1.1262, + "step": 69253 + }, + { + "epoch": 5.12, + "learning_rate": 1.113540471478215e-06, + "loss": 1.0152, + "step": 69254 + }, + { + "epoch": 5.12, + "learning_rate": 1.113357558453223e-06, + "loss": 1.0526, + "step": 69255 + }, + { + "epoch": 5.12, + "learning_rate": 1.1131746595667502e-06, + "loss": 0.9236, + "step": 69256 + }, + { + "epoch": 5.12, + "learning_rate": 1.1129917748190933e-06, + "loss": 1.0015, + "step": 69257 + }, + { + "epoch": 5.12, + "learning_rate": 1.112808904210536e-06, + "loss": 1.0235, + "step": 69258 + }, + { + "epoch": 5.12, + "learning_rate": 1.1126260477413741e-06, + "loss": 0.9662, + "step": 69259 + }, + { + "epoch": 5.12, + "learning_rate": 1.112443205411895e-06, + "loss": 1.0123, + "step": 69260 + }, + { + "epoch": 5.12, + "learning_rate": 1.112260377222395e-06, + "loss": 0.9961, + "step": 69261 + }, + { + "epoch": 5.12, + "learning_rate": 1.1120775631731606e-06, + "loss": 0.9161, + "step": 69262 + }, + { + "epoch": 5.12, + "learning_rate": 1.1118947632644806e-06, + "loss": 1.1083, + "step": 69263 + }, + { + "epoch": 5.12, + "learning_rate": 1.1117119774966523e-06, + "loss": 0.9376, + "step": 69264 + }, + { + "epoch": 5.12, + "learning_rate": 1.1115292058699624e-06, + "loss": 0.9989, + "step": 69265 + }, + { + "epoch": 5.12, + "learning_rate": 1.1113464483847014e-06, + "loss": 1.0267, + "step": 69266 + }, + { + "epoch": 5.12, + "learning_rate": 1.1111637050411594e-06, + "loss": 1.0769, + "step": 69267 + }, + { + "epoch": 5.12, + "learning_rate": 1.1109809758396317e-06, + "loss": 1.0309, + "step": 69268 + }, + { + "epoch": 5.12, + "learning_rate": 1.1107982607804058e-06, + "loss": 0.9988, + "step": 69269 + }, + { + "epoch": 5.12, + "learning_rate": 1.1106155598637724e-06, + "loss": 0.9075, + "step": 69270 + }, + { + "epoch": 5.12, + "learning_rate": 1.1104328730900226e-06, + "loss": 0.9606, + "step": 69271 + }, + { + "epoch": 5.12, + "learning_rate": 1.110250200459445e-06, + "loss": 0.9535, + "step": 69272 + }, + { + "epoch": 5.12, + "learning_rate": 1.110067541972334e-06, + "loss": 1.1248, + "step": 69273 + }, + { + "epoch": 5.12, + "learning_rate": 1.1098848976289778e-06, + "loss": 0.9682, + "step": 69274 + }, + { + "epoch": 5.12, + "learning_rate": 1.1097022674296676e-06, + "loss": 0.9216, + "step": 69275 + }, + { + "epoch": 5.12, + "learning_rate": 1.109519651374693e-06, + "loss": 0.9886, + "step": 69276 + }, + { + "epoch": 5.12, + "learning_rate": 1.1093370494643464e-06, + "loss": 0.9684, + "step": 69277 + }, + { + "epoch": 5.12, + "learning_rate": 1.109154461698918e-06, + "loss": 0.9567, + "step": 69278 + }, + { + "epoch": 5.12, + "learning_rate": 1.108971888078697e-06, + "loss": 0.9408, + "step": 69279 + }, + { + "epoch": 5.12, + "learning_rate": 1.1087893286039753e-06, + "loss": 1.0722, + "step": 69280 + }, + { + "epoch": 5.12, + "learning_rate": 1.1086067832750403e-06, + "loss": 1.0377, + "step": 69281 + }, + { + "epoch": 5.12, + "learning_rate": 1.1084242520921885e-06, + "loss": 1.0321, + "step": 69282 + }, + { + "epoch": 5.12, + "learning_rate": 1.1082417350557019e-06, + "loss": 1.0483, + "step": 69283 + }, + { + "epoch": 5.12, + "learning_rate": 1.1080592321658789e-06, + "loss": 0.9541, + "step": 69284 + }, + { + "epoch": 5.12, + "learning_rate": 1.1078767434230032e-06, + "loss": 0.9551, + "step": 69285 + }, + { + "epoch": 5.12, + "learning_rate": 1.1076942688273728e-06, + "loss": 0.896, + "step": 69286 + }, + { + "epoch": 5.12, + "learning_rate": 1.107511808379269e-06, + "loss": 1.0076, + "step": 69287 + }, + { + "epoch": 5.12, + "learning_rate": 1.1073293620789904e-06, + "loss": 1.0246, + "step": 69288 + }, + { + "epoch": 5.12, + "learning_rate": 1.1071469299268222e-06, + "loss": 1.0113, + "step": 69289 + }, + { + "epoch": 5.12, + "learning_rate": 1.1069645119230565e-06, + "loss": 0.7917, + "step": 69290 + }, + { + "epoch": 5.12, + "learning_rate": 1.1067821080679818e-06, + "loss": 0.9627, + "step": 69291 + }, + { + "epoch": 5.12, + "learning_rate": 1.1065997183618893e-06, + "loss": 1.0107, + "step": 69292 + }, + { + "epoch": 5.12, + "learning_rate": 1.1064173428050706e-06, + "loss": 1.0452, + "step": 69293 + }, + { + "epoch": 5.12, + "learning_rate": 1.1062349813978146e-06, + "loss": 1.0343, + "step": 69294 + }, + { + "epoch": 5.12, + "learning_rate": 1.106052634140411e-06, + "loss": 0.9209, + "step": 69295 + }, + { + "epoch": 5.12, + "learning_rate": 1.1058703010331495e-06, + "loss": 1.0078, + "step": 69296 + }, + { + "epoch": 5.12, + "learning_rate": 1.1056879820763222e-06, + "loss": 1.0397, + "step": 69297 + }, + { + "epoch": 5.12, + "learning_rate": 1.1055056772702187e-06, + "loss": 1.0561, + "step": 69298 + }, + { + "epoch": 5.12, + "learning_rate": 1.105323386615128e-06, + "loss": 1.0383, + "step": 69299 + }, + { + "epoch": 5.12, + "learning_rate": 1.1051411101113406e-06, + "loss": 0.9986, + "step": 69300 + }, + { + "epoch": 5.12, + "learning_rate": 1.1049588477591445e-06, + "loss": 0.9524, + "step": 69301 + }, + { + "epoch": 5.12, + "learning_rate": 1.1047765995588356e-06, + "loss": 0.9734, + "step": 69302 + }, + { + "epoch": 5.12, + "learning_rate": 1.104594365510696e-06, + "loss": 1.0383, + "step": 69303 + }, + { + "epoch": 5.12, + "learning_rate": 1.1044121456150214e-06, + "loss": 0.9299, + "step": 69304 + }, + { + "epoch": 5.12, + "learning_rate": 1.104229939872098e-06, + "loss": 0.9848, + "step": 69305 + }, + { + "epoch": 5.12, + "learning_rate": 1.104047748282221e-06, + "loss": 1.0883, + "step": 69306 + }, + { + "epoch": 5.12, + "learning_rate": 1.1038655708456724e-06, + "loss": 0.9875, + "step": 69307 + }, + { + "epoch": 5.12, + "learning_rate": 1.103683407562749e-06, + "loss": 0.9365, + "step": 69308 + }, + { + "epoch": 5.12, + "learning_rate": 1.1035012584337378e-06, + "loss": 0.9541, + "step": 69309 + }, + { + "epoch": 5.12, + "learning_rate": 1.1033191234589291e-06, + "loss": 1.0601, + "step": 69310 + }, + { + "epoch": 5.12, + "learning_rate": 1.1031370026386123e-06, + "loss": 1.1294, + "step": 69311 + }, + { + "epoch": 5.12, + "learning_rate": 1.1029548959730764e-06, + "loss": 0.9887, + "step": 69312 + }, + { + "epoch": 5.12, + "learning_rate": 1.1027728034626117e-06, + "loss": 1.1162, + "step": 69313 + }, + { + "epoch": 5.12, + "learning_rate": 1.1025907251075096e-06, + "loss": 0.9948, + "step": 69314 + }, + { + "epoch": 5.12, + "learning_rate": 1.1024086609080586e-06, + "loss": 0.9487, + "step": 69315 + }, + { + "epoch": 5.12, + "learning_rate": 1.1022266108645451e-06, + "loss": 1.0185, + "step": 69316 + }, + { + "epoch": 5.12, + "learning_rate": 1.1020445749772646e-06, + "loss": 0.9431, + "step": 69317 + }, + { + "epoch": 5.12, + "learning_rate": 1.1018625532465033e-06, + "loss": 0.9877, + "step": 69318 + }, + { + "epoch": 5.12, + "learning_rate": 1.101680545672551e-06, + "loss": 0.9151, + "step": 69319 + }, + { + "epoch": 5.12, + "learning_rate": 1.1014985522556987e-06, + "loss": 0.9869, + "step": 69320 + }, + { + "epoch": 5.12, + "learning_rate": 1.1013165729962316e-06, + "loss": 0.9653, + "step": 69321 + }, + { + "epoch": 5.12, + "learning_rate": 1.1011346078944462e-06, + "loss": 0.9672, + "step": 69322 + }, + { + "epoch": 5.12, + "learning_rate": 1.100952656950627e-06, + "loss": 1.0528, + "step": 69323 + }, + { + "epoch": 5.12, + "learning_rate": 1.1007707201650653e-06, + "loss": 0.9918, + "step": 69324 + }, + { + "epoch": 5.12, + "learning_rate": 1.100588797538048e-06, + "loss": 1.0479, + "step": 69325 + }, + { + "epoch": 5.12, + "learning_rate": 1.1004068890698683e-06, + "loss": 0.9242, + "step": 69326 + }, + { + "epoch": 5.12, + "learning_rate": 1.1002249947608157e-06, + "loss": 0.8663, + "step": 69327 + }, + { + "epoch": 5.12, + "learning_rate": 1.1000431146111745e-06, + "loss": 0.9706, + "step": 69328 + }, + { + "epoch": 5.12, + "learning_rate": 1.099861248621239e-06, + "loss": 1.0144, + "step": 69329 + }, + { + "epoch": 5.12, + "learning_rate": 1.0996793967912955e-06, + "loss": 0.9561, + "step": 69330 + }, + { + "epoch": 5.12, + "learning_rate": 1.0994975591216383e-06, + "loss": 0.9315, + "step": 69331 + }, + { + "epoch": 5.12, + "learning_rate": 1.0993157356125484e-06, + "loss": 1.0569, + "step": 69332 + }, + { + "epoch": 5.12, + "learning_rate": 1.099133926264323e-06, + "loss": 1.1062, + "step": 69333 + }, + { + "epoch": 5.12, + "learning_rate": 1.0989521310772478e-06, + "loss": 0.9053, + "step": 69334 + }, + { + "epoch": 5.12, + "learning_rate": 1.0987703500516122e-06, + "loss": 1.038, + "step": 69335 + }, + { + "epoch": 5.12, + "learning_rate": 1.0985885831877041e-06, + "loss": 1.1555, + "step": 69336 + }, + { + "epoch": 5.12, + "learning_rate": 1.0984068304858153e-06, + "loss": 0.9992, + "step": 69337 + }, + { + "epoch": 5.12, + "learning_rate": 1.0982250919462345e-06, + "loss": 1.0934, + "step": 69338 + }, + { + "epoch": 5.12, + "learning_rate": 1.0980433675692504e-06, + "loss": 1.1096, + "step": 69339 + }, + { + "epoch": 5.12, + "learning_rate": 1.0978616573551514e-06, + "loss": 0.9985, + "step": 69340 + }, + { + "epoch": 5.12, + "learning_rate": 1.0976799613042256e-06, + "loss": 0.9791, + "step": 69341 + }, + { + "epoch": 5.12, + "learning_rate": 1.0974982794167655e-06, + "loss": 0.8269, + "step": 69342 + }, + { + "epoch": 5.12, + "learning_rate": 1.0973166116930579e-06, + "loss": 1.0188, + "step": 69343 + }, + { + "epoch": 5.12, + "learning_rate": 1.0971349581333922e-06, + "loss": 0.9266, + "step": 69344 + }, + { + "epoch": 5.12, + "learning_rate": 1.0969533187380555e-06, + "loss": 1.0442, + "step": 69345 + }, + { + "epoch": 5.12, + "learning_rate": 1.0967716935073413e-06, + "loss": 0.9732, + "step": 69346 + }, + { + "epoch": 5.12, + "learning_rate": 1.0965900824415376e-06, + "loss": 0.9904, + "step": 69347 + }, + { + "epoch": 5.12, + "learning_rate": 1.0964084855409273e-06, + "loss": 0.9742, + "step": 69348 + }, + { + "epoch": 5.12, + "learning_rate": 1.096226902805806e-06, + "loss": 1.0293, + "step": 69349 + }, + { + "epoch": 5.12, + "learning_rate": 1.0960453342364584e-06, + "loss": 0.9604, + "step": 69350 + }, + { + "epoch": 5.12, + "learning_rate": 1.0958637798331795e-06, + "loss": 0.9755, + "step": 69351 + }, + { + "epoch": 5.12, + "learning_rate": 1.0956822395962496e-06, + "loss": 0.974, + "step": 69352 + }, + { + "epoch": 5.12, + "learning_rate": 1.0955007135259633e-06, + "loss": 1.0164, + "step": 69353 + }, + { + "epoch": 5.12, + "learning_rate": 1.0953192016226067e-06, + "loss": 0.9532, + "step": 69354 + }, + { + "epoch": 5.12, + "learning_rate": 1.0951377038864731e-06, + "loss": 0.8711, + "step": 69355 + }, + { + "epoch": 5.12, + "learning_rate": 1.0949562203178465e-06, + "loss": 1.0588, + "step": 69356 + }, + { + "epoch": 5.12, + "learning_rate": 1.0947747509170158e-06, + "loss": 0.906, + "step": 69357 + }, + { + "epoch": 5.12, + "learning_rate": 1.094593295684272e-06, + "loss": 1.015, + "step": 69358 + }, + { + "epoch": 5.12, + "learning_rate": 1.0944118546199035e-06, + "loss": 1.0267, + "step": 69359 + }, + { + "epoch": 5.12, + "learning_rate": 1.0942304277241988e-06, + "loss": 0.9081, + "step": 69360 + }, + { + "epoch": 5.13, + "learning_rate": 1.0940490149974424e-06, + "loss": 1.0722, + "step": 69361 + }, + { + "epoch": 5.13, + "learning_rate": 1.0938676164399297e-06, + "loss": 1.0108, + "step": 69362 + }, + { + "epoch": 5.13, + "learning_rate": 1.093686232051947e-06, + "loss": 0.998, + "step": 69363 + }, + { + "epoch": 5.13, + "learning_rate": 1.0935048618337806e-06, + "loss": 1.0192, + "step": 69364 + }, + { + "epoch": 5.13, + "learning_rate": 1.0933235057857217e-06, + "loss": 0.9461, + "step": 69365 + }, + { + "epoch": 5.13, + "learning_rate": 1.0931421639080542e-06, + "loss": 0.9427, + "step": 69366 + }, + { + "epoch": 5.13, + "learning_rate": 1.0929608362010735e-06, + "loss": 0.9267, + "step": 69367 + }, + { + "epoch": 5.13, + "learning_rate": 1.092779522665064e-06, + "loss": 0.9362, + "step": 69368 + }, + { + "epoch": 5.13, + "learning_rate": 1.0925982233003153e-06, + "loss": 0.9048, + "step": 69369 + }, + { + "epoch": 5.13, + "learning_rate": 1.0924169381071126e-06, + "loss": 0.9474, + "step": 69370 + }, + { + "epoch": 5.13, + "learning_rate": 1.0922356670857492e-06, + "loss": 1.1845, + "step": 69371 + }, + { + "epoch": 5.13, + "learning_rate": 1.0920544102365127e-06, + "loss": 0.9686, + "step": 69372 + }, + { + "epoch": 5.13, + "learning_rate": 1.0918731675596883e-06, + "loss": 0.968, + "step": 69373 + }, + { + "epoch": 5.13, + "learning_rate": 1.091691939055568e-06, + "loss": 1.1369, + "step": 69374 + }, + { + "epoch": 5.13, + "learning_rate": 1.091510724724435e-06, + "loss": 1.0133, + "step": 69375 + }, + { + "epoch": 5.13, + "learning_rate": 1.0913295245665856e-06, + "loss": 1.028, + "step": 69376 + }, + { + "epoch": 5.13, + "learning_rate": 1.0911483385822996e-06, + "loss": 1.0115, + "step": 69377 + }, + { + "epoch": 5.13, + "learning_rate": 1.0909671667718714e-06, + "loss": 0.9974, + "step": 69378 + }, + { + "epoch": 5.13, + "learning_rate": 1.090786009135586e-06, + "loss": 1.1226, + "step": 69379 + }, + { + "epoch": 5.13, + "learning_rate": 1.0906048656737333e-06, + "loss": 1.0847, + "step": 69380 + }, + { + "epoch": 5.13, + "learning_rate": 1.0904237363865989e-06, + "loss": 0.9799, + "step": 69381 + }, + { + "epoch": 5.13, + "learning_rate": 1.0902426212744743e-06, + "loss": 1.0024, + "step": 69382 + }, + { + "epoch": 5.13, + "learning_rate": 1.0900615203376463e-06, + "loss": 0.9313, + "step": 69383 + }, + { + "epoch": 5.13, + "learning_rate": 1.0898804335764023e-06, + "loss": 0.9669, + "step": 69384 + }, + { + "epoch": 5.13, + "learning_rate": 1.0896993609910323e-06, + "loss": 0.9412, + "step": 69385 + }, + { + "epoch": 5.13, + "learning_rate": 1.0895183025818212e-06, + "loss": 1.0613, + "step": 69386 + }, + { + "epoch": 5.13, + "learning_rate": 1.0893372583490603e-06, + "loss": 1.0503, + "step": 69387 + }, + { + "epoch": 5.13, + "learning_rate": 1.0891562282930356e-06, + "loss": 1.0565, + "step": 69388 + }, + { + "epoch": 5.13, + "learning_rate": 1.0889752124140373e-06, + "loss": 1.0316, + "step": 69389 + }, + { + "epoch": 5.13, + "learning_rate": 1.0887942107123494e-06, + "loss": 1.1159, + "step": 69390 + }, + { + "epoch": 5.13, + "learning_rate": 1.088613223188264e-06, + "loss": 0.9352, + "step": 69391 + }, + { + "epoch": 5.13, + "learning_rate": 1.0884322498420685e-06, + "loss": 1.0078, + "step": 69392 + }, + { + "epoch": 5.13, + "learning_rate": 1.0882512906740484e-06, + "loss": 0.9236, + "step": 69393 + }, + { + "epoch": 5.13, + "learning_rate": 1.0880703456844944e-06, + "loss": 0.9609, + "step": 69394 + }, + { + "epoch": 5.13, + "learning_rate": 1.0878894148736896e-06, + "loss": 0.9667, + "step": 69395 + }, + { + "epoch": 5.13, + "learning_rate": 1.0877084982419305e-06, + "loss": 1.0157, + "step": 69396 + }, + { + "epoch": 5.13, + "learning_rate": 1.0875275957894949e-06, + "loss": 1.0428, + "step": 69397 + }, + { + "epoch": 5.13, + "learning_rate": 1.0873467075166777e-06, + "loss": 0.9939, + "step": 69398 + }, + { + "epoch": 5.13, + "learning_rate": 1.0871658334237623e-06, + "loss": 0.9898, + "step": 69399 + }, + { + "epoch": 5.13, + "learning_rate": 1.0869849735110427e-06, + "loss": 1.052, + "step": 69400 + }, + { + "epoch": 5.13, + "learning_rate": 1.0868041277787978e-06, + "loss": 0.9155, + "step": 69401 + }, + { + "epoch": 5.13, + "learning_rate": 1.0866232962273226e-06, + "loss": 1.0492, + "step": 69402 + }, + { + "epoch": 5.13, + "learning_rate": 1.0864424788569017e-06, + "loss": 1.1188, + "step": 69403 + }, + { + "epoch": 5.13, + "learning_rate": 1.0862616756678235e-06, + "loss": 1.0614, + "step": 69404 + }, + { + "epoch": 5.13, + "learning_rate": 1.0860808866603757e-06, + "loss": 0.9949, + "step": 69405 + }, + { + "epoch": 5.13, + "learning_rate": 1.0859001118348445e-06, + "loss": 0.9689, + "step": 69406 + }, + { + "epoch": 5.13, + "learning_rate": 1.0857193511915186e-06, + "loss": 1.0416, + "step": 69407 + }, + { + "epoch": 5.13, + "learning_rate": 1.0855386047306871e-06, + "loss": 1.0478, + "step": 69408 + }, + { + "epoch": 5.13, + "learning_rate": 1.0853578724526359e-06, + "loss": 0.9259, + "step": 69409 + }, + { + "epoch": 5.13, + "learning_rate": 1.0851771543576505e-06, + "loss": 0.9144, + "step": 69410 + }, + { + "epoch": 5.13, + "learning_rate": 1.084996450446023e-06, + "loss": 0.9709, + "step": 69411 + }, + { + "epoch": 5.13, + "learning_rate": 1.0848157607180387e-06, + "loss": 0.9973, + "step": 69412 + }, + { + "epoch": 5.13, + "learning_rate": 1.084635085173985e-06, + "loss": 0.9788, + "step": 69413 + }, + { + "epoch": 5.13, + "learning_rate": 1.0844544238141498e-06, + "loss": 1.0496, + "step": 69414 + }, + { + "epoch": 5.13, + "learning_rate": 1.0842737766388167e-06, + "loss": 0.9867, + "step": 69415 + }, + { + "epoch": 5.13, + "learning_rate": 1.0840931436482793e-06, + "loss": 0.9123, + "step": 69416 + }, + { + "epoch": 5.13, + "learning_rate": 1.0839125248428217e-06, + "loss": 0.9846, + "step": 69417 + }, + { + "epoch": 5.13, + "learning_rate": 1.0837319202227325e-06, + "loss": 0.9741, + "step": 69418 + }, + { + "epoch": 5.13, + "learning_rate": 1.0835513297882949e-06, + "loss": 1.0696, + "step": 69419 + }, + { + "epoch": 5.13, + "learning_rate": 1.083370753539803e-06, + "loss": 0.9764, + "step": 69420 + }, + { + "epoch": 5.13, + "learning_rate": 1.0831901914775423e-06, + "loss": 1.0343, + "step": 69421 + }, + { + "epoch": 5.13, + "learning_rate": 1.0830096436017934e-06, + "loss": 0.9128, + "step": 69422 + }, + { + "epoch": 5.13, + "learning_rate": 1.082829109912852e-06, + "loss": 1.0097, + "step": 69423 + }, + { + "epoch": 5.13, + "learning_rate": 1.0826485904110006e-06, + "loss": 1.1143, + "step": 69424 + }, + { + "epoch": 5.13, + "learning_rate": 1.0824680850965285e-06, + "loss": 1.0253, + "step": 69425 + }, + { + "epoch": 5.13, + "learning_rate": 1.0822875939697197e-06, + "loss": 0.9246, + "step": 69426 + }, + { + "epoch": 5.13, + "learning_rate": 1.0821071170308662e-06, + "loss": 1.0904, + "step": 69427 + }, + { + "epoch": 5.13, + "learning_rate": 1.081926654280252e-06, + "loss": 0.9595, + "step": 69428 + }, + { + "epoch": 5.13, + "learning_rate": 1.081746205718166e-06, + "loss": 0.9048, + "step": 69429 + }, + { + "epoch": 5.13, + "learning_rate": 1.0815657713448936e-06, + "loss": 0.8938, + "step": 69430 + }, + { + "epoch": 5.13, + "learning_rate": 1.0813853511607197e-06, + "loss": 1.0002, + "step": 69431 + }, + { + "epoch": 5.13, + "learning_rate": 1.0812049451659367e-06, + "loss": 0.9132, + "step": 69432 + }, + { + "epoch": 5.13, + "learning_rate": 1.0810245533608298e-06, + "loss": 0.9752, + "step": 69433 + }, + { + "epoch": 5.13, + "learning_rate": 1.0808441757456845e-06, + "loss": 1.0499, + "step": 69434 + }, + { + "epoch": 5.13, + "learning_rate": 1.080663812320787e-06, + "loss": 0.9661, + "step": 69435 + }, + { + "epoch": 5.13, + "learning_rate": 1.080483463086427e-06, + "loss": 0.9267, + "step": 69436 + }, + { + "epoch": 5.13, + "learning_rate": 1.08030312804289e-06, + "loss": 0.9545, + "step": 69437 + }, + { + "epoch": 5.13, + "learning_rate": 1.0801228071904635e-06, + "loss": 1.0339, + "step": 69438 + }, + { + "epoch": 5.13, + "learning_rate": 1.079942500529435e-06, + "loss": 0.8007, + "step": 69439 + }, + { + "epoch": 5.13, + "learning_rate": 1.0797622080600868e-06, + "loss": 1.0793, + "step": 69440 + }, + { + "epoch": 5.13, + "learning_rate": 1.0795819297827137e-06, + "loss": 0.9556, + "step": 69441 + }, + { + "epoch": 5.13, + "learning_rate": 1.0794016656975947e-06, + "loss": 0.9072, + "step": 69442 + }, + { + "epoch": 5.13, + "learning_rate": 1.0792214158050219e-06, + "loss": 0.9993, + "step": 69443 + }, + { + "epoch": 5.13, + "learning_rate": 1.079041180105278e-06, + "loss": 1.0622, + "step": 69444 + }, + { + "epoch": 5.13, + "learning_rate": 1.0788609585986554e-06, + "loss": 0.9292, + "step": 69445 + }, + { + "epoch": 5.13, + "learning_rate": 1.0786807512854347e-06, + "loss": 1.1195, + "step": 69446 + }, + { + "epoch": 5.13, + "learning_rate": 1.0785005581659058e-06, + "loss": 0.9808, + "step": 69447 + }, + { + "epoch": 5.13, + "learning_rate": 1.078320379240354e-06, + "loss": 1.0008, + "step": 69448 + }, + { + "epoch": 5.13, + "learning_rate": 1.0781402145090703e-06, + "loss": 1.0225, + "step": 69449 + }, + { + "epoch": 5.13, + "learning_rate": 1.0779600639723353e-06, + "loss": 1.0029, + "step": 69450 + }, + { + "epoch": 5.13, + "learning_rate": 1.077779927630438e-06, + "loss": 1.0166, + "step": 69451 + }, + { + "epoch": 5.13, + "learning_rate": 1.0775998054836657e-06, + "loss": 1.0248, + "step": 69452 + }, + { + "epoch": 5.13, + "learning_rate": 1.0774196975323048e-06, + "loss": 1.0475, + "step": 69453 + }, + { + "epoch": 5.13, + "learning_rate": 1.077239603776641e-06, + "loss": 1.0467, + "step": 69454 + }, + { + "epoch": 5.13, + "learning_rate": 1.0770595242169602e-06, + "loss": 1.0654, + "step": 69455 + }, + { + "epoch": 5.13, + "learning_rate": 1.0768794588535513e-06, + "loss": 1.0034, + "step": 69456 + }, + { + "epoch": 5.13, + "learning_rate": 1.0766994076866998e-06, + "loss": 1.0504, + "step": 69457 + }, + { + "epoch": 5.13, + "learning_rate": 1.076519370716692e-06, + "loss": 0.9472, + "step": 69458 + }, + { + "epoch": 5.13, + "learning_rate": 1.0763393479438144e-06, + "loss": 0.8979, + "step": 69459 + }, + { + "epoch": 5.13, + "learning_rate": 1.076159339368351e-06, + "loss": 1.0389, + "step": 69460 + }, + { + "epoch": 5.13, + "learning_rate": 1.0759793449905919e-06, + "loss": 1.0253, + "step": 69461 + }, + { + "epoch": 5.13, + "learning_rate": 1.0757993648108222e-06, + "loss": 1.0232, + "step": 69462 + }, + { + "epoch": 5.13, + "learning_rate": 1.0756193988293285e-06, + "loss": 1.0055, + "step": 69463 + }, + { + "epoch": 5.13, + "learning_rate": 1.0754394470463947e-06, + "loss": 0.953, + "step": 69464 + }, + { + "epoch": 5.13, + "learning_rate": 1.0752595094623109e-06, + "loss": 1.0899, + "step": 69465 + }, + { + "epoch": 5.13, + "learning_rate": 1.0750795860773612e-06, + "loss": 0.9392, + "step": 69466 + }, + { + "epoch": 5.13, + "learning_rate": 1.0748996768918318e-06, + "loss": 1.0636, + "step": 69467 + }, + { + "epoch": 5.13, + "learning_rate": 1.0747197819060107e-06, + "loss": 1.0644, + "step": 69468 + }, + { + "epoch": 5.13, + "learning_rate": 1.0745399011201817e-06, + "loss": 0.9781, + "step": 69469 + }, + { + "epoch": 5.13, + "learning_rate": 1.0743600345346316e-06, + "loss": 0.8586, + "step": 69470 + }, + { + "epoch": 5.13, + "learning_rate": 1.0741801821496457e-06, + "loss": 1.1168, + "step": 69471 + }, + { + "epoch": 5.13, + "learning_rate": 1.0740003439655133e-06, + "loss": 0.9168, + "step": 69472 + }, + { + "epoch": 5.13, + "learning_rate": 1.073820519982519e-06, + "loss": 1.0402, + "step": 69473 + }, + { + "epoch": 5.13, + "learning_rate": 1.0736407102009482e-06, + "loss": 1.0438, + "step": 69474 + }, + { + "epoch": 5.13, + "learning_rate": 1.073460914621085e-06, + "loss": 0.9875, + "step": 69475 + }, + { + "epoch": 5.13, + "learning_rate": 1.0732811332432202e-06, + "loss": 1.008, + "step": 69476 + }, + { + "epoch": 5.13, + "learning_rate": 1.073101366067637e-06, + "loss": 0.9032, + "step": 69477 + }, + { + "epoch": 5.13, + "learning_rate": 1.072921613094623e-06, + "loss": 1.0056, + "step": 69478 + }, + { + "epoch": 5.13, + "learning_rate": 1.072741874324461e-06, + "loss": 1.0526, + "step": 69479 + }, + { + "epoch": 5.13, + "learning_rate": 1.072562149757439e-06, + "loss": 0.9152, + "step": 69480 + }, + { + "epoch": 5.13, + "learning_rate": 1.072382439393843e-06, + "loss": 0.995, + "step": 69481 + }, + { + "epoch": 5.13, + "learning_rate": 1.0722027432339598e-06, + "loss": 1.0319, + "step": 69482 + }, + { + "epoch": 5.13, + "learning_rate": 1.0720230612780747e-06, + "loss": 1.0713, + "step": 69483 + }, + { + "epoch": 5.13, + "learning_rate": 1.0718433935264704e-06, + "loss": 1.0452, + "step": 69484 + }, + { + "epoch": 5.13, + "learning_rate": 1.0716637399794383e-06, + "loss": 1.1225, + "step": 69485 + }, + { + "epoch": 5.13, + "learning_rate": 1.0714841006372622e-06, + "loss": 1.0295, + "step": 69486 + }, + { + "epoch": 5.13, + "learning_rate": 1.0713044755002245e-06, + "loss": 1.0896, + "step": 69487 + }, + { + "epoch": 5.13, + "learning_rate": 1.0711248645686145e-06, + "loss": 1.0856, + "step": 69488 + }, + { + "epoch": 5.13, + "learning_rate": 1.0709452678427156e-06, + "loss": 1.116, + "step": 69489 + }, + { + "epoch": 5.13, + "learning_rate": 1.0707656853228198e-06, + "loss": 1.0262, + "step": 69490 + }, + { + "epoch": 5.13, + "learning_rate": 1.0705861170092035e-06, + "loss": 0.9402, + "step": 69491 + }, + { + "epoch": 5.13, + "learning_rate": 1.0704065629021587e-06, + "loss": 1.0057, + "step": 69492 + }, + { + "epoch": 5.13, + "learning_rate": 1.0702270230019673e-06, + "loss": 0.8668, + "step": 69493 + }, + { + "epoch": 5.13, + "learning_rate": 1.0700474973089203e-06, + "loss": 1.0327, + "step": 69494 + }, + { + "epoch": 5.13, + "learning_rate": 1.0698679858232986e-06, + "loss": 1.0529, + "step": 69495 + }, + { + "epoch": 5.14, + "learning_rate": 1.0696884885453872e-06, + "loss": 0.9545, + "step": 69496 + }, + { + "epoch": 5.14, + "learning_rate": 1.0695090054754765e-06, + "loss": 0.9362, + "step": 69497 + }, + { + "epoch": 5.14, + "learning_rate": 1.0693295366138478e-06, + "loss": 0.8998, + "step": 69498 + }, + { + "epoch": 5.14, + "learning_rate": 1.0691500819607881e-06, + "loss": 1.0802, + "step": 69499 + }, + { + "epoch": 5.14, + "learning_rate": 1.0689706415165812e-06, + "loss": 1.03, + "step": 69500 + }, + { + "epoch": 5.14, + "learning_rate": 1.0687912152815172e-06, + "loss": 1.0438, + "step": 69501 + }, + { + "epoch": 5.14, + "learning_rate": 1.0686118032558778e-06, + "loss": 1.0075, + "step": 69502 + }, + { + "epoch": 5.14, + "learning_rate": 1.0684324054399486e-06, + "loss": 1.0661, + "step": 69503 + }, + { + "epoch": 5.14, + "learning_rate": 1.0682530218340148e-06, + "loss": 1.0423, + "step": 69504 + }, + { + "epoch": 5.14, + "learning_rate": 1.0680736524383638e-06, + "loss": 1.0262, + "step": 69505 + }, + { + "epoch": 5.14, + "learning_rate": 1.0678942972532801e-06, + "loss": 1.0251, + "step": 69506 + }, + { + "epoch": 5.14, + "learning_rate": 1.067714956279049e-06, + "loss": 1.0591, + "step": 69507 + }, + { + "epoch": 5.14, + "learning_rate": 1.0675356295159544e-06, + "loss": 1.1799, + "step": 69508 + }, + { + "epoch": 5.14, + "learning_rate": 1.067356316964282e-06, + "loss": 0.9532, + "step": 69509 + }, + { + "epoch": 5.14, + "learning_rate": 1.0671770186243223e-06, + "loss": 1.0198, + "step": 69510 + }, + { + "epoch": 5.14, + "learning_rate": 1.066997734496351e-06, + "loss": 1.072, + "step": 69511 + }, + { + "epoch": 5.14, + "learning_rate": 1.0668184645806612e-06, + "loss": 1.0553, + "step": 69512 + }, + { + "epoch": 5.14, + "learning_rate": 1.0666392088775336e-06, + "loss": 0.9321, + "step": 69513 + }, + { + "epoch": 5.14, + "learning_rate": 1.0664599673872588e-06, + "loss": 1.0617, + "step": 69514 + }, + { + "epoch": 5.14, + "learning_rate": 1.066280740110116e-06, + "loss": 1.0084, + "step": 69515 + }, + { + "epoch": 5.14, + "learning_rate": 1.0661015270463915e-06, + "loss": 1.051, + "step": 69516 + }, + { + "epoch": 5.14, + "learning_rate": 1.0659223281963727e-06, + "loss": 0.9197, + "step": 69517 + }, + { + "epoch": 5.14, + "learning_rate": 1.0657431435603448e-06, + "loss": 1.0303, + "step": 69518 + }, + { + "epoch": 5.14, + "learning_rate": 1.0655639731385903e-06, + "loss": 0.9931, + "step": 69519 + }, + { + "epoch": 5.14, + "learning_rate": 1.065384816931395e-06, + "loss": 1.0357, + "step": 69520 + }, + { + "epoch": 5.14, + "learning_rate": 1.065205674939046e-06, + "loss": 0.9239, + "step": 69521 + }, + { + "epoch": 5.14, + "learning_rate": 1.0650265471618271e-06, + "loss": 0.9786, + "step": 69522 + }, + { + "epoch": 5.14, + "learning_rate": 1.0648474336000237e-06, + "loss": 0.9947, + "step": 69523 + }, + { + "epoch": 5.14, + "learning_rate": 1.0646683342539187e-06, + "loss": 0.8923, + "step": 69524 + }, + { + "epoch": 5.14, + "learning_rate": 1.0644892491237979e-06, + "loss": 1.0029, + "step": 69525 + }, + { + "epoch": 5.14, + "learning_rate": 1.0643101782099485e-06, + "loss": 1.0479, + "step": 69526 + }, + { + "epoch": 5.14, + "learning_rate": 1.0641311215126538e-06, + "loss": 0.9851, + "step": 69527 + }, + { + "epoch": 5.14, + "learning_rate": 1.063952079032199e-06, + "loss": 0.9602, + "step": 69528 + }, + { + "epoch": 5.14, + "learning_rate": 1.063773050768866e-06, + "loss": 0.9995, + "step": 69529 + }, + { + "epoch": 5.14, + "learning_rate": 1.0635940367229435e-06, + "loss": 1.0096, + "step": 69530 + }, + { + "epoch": 5.14, + "learning_rate": 1.063415036894716e-06, + "loss": 0.9926, + "step": 69531 + }, + { + "epoch": 5.14, + "learning_rate": 1.0632360512844674e-06, + "loss": 1.0593, + "step": 69532 + }, + { + "epoch": 5.14, + "learning_rate": 1.0630570798924823e-06, + "loss": 0.9736, + "step": 69533 + }, + { + "epoch": 5.14, + "learning_rate": 1.0628781227190432e-06, + "loss": 0.9955, + "step": 69534 + }, + { + "epoch": 5.14, + "learning_rate": 1.0626991797644414e-06, + "loss": 1.1064, + "step": 69535 + }, + { + "epoch": 5.14, + "learning_rate": 1.0625202510289534e-06, + "loss": 0.9832, + "step": 69536 + }, + { + "epoch": 5.14, + "learning_rate": 1.0623413365128687e-06, + "loss": 1.007, + "step": 69537 + }, + { + "epoch": 5.14, + "learning_rate": 1.0621624362164706e-06, + "loss": 0.9914, + "step": 69538 + }, + { + "epoch": 5.14, + "learning_rate": 1.0619835501400478e-06, + "loss": 0.9995, + "step": 69539 + }, + { + "epoch": 5.14, + "learning_rate": 1.0618046782838765e-06, + "loss": 0.9736, + "step": 69540 + }, + { + "epoch": 5.14, + "learning_rate": 1.061625820648249e-06, + "loss": 0.9875, + "step": 69541 + }, + { + "epoch": 5.14, + "learning_rate": 1.061446977233447e-06, + "loss": 0.93, + "step": 69542 + }, + { + "epoch": 5.14, + "learning_rate": 1.0612681480397536e-06, + "loss": 1.0163, + "step": 69543 + }, + { + "epoch": 5.14, + "learning_rate": 1.0610893330674565e-06, + "loss": 1.0398, + "step": 69544 + }, + { + "epoch": 5.14, + "learning_rate": 1.0609105323168345e-06, + "loss": 0.9841, + "step": 69545 + }, + { + "epoch": 5.14, + "learning_rate": 1.0607317457881793e-06, + "loss": 1.0609, + "step": 69546 + }, + { + "epoch": 5.14, + "learning_rate": 1.060552973481772e-06, + "loss": 0.8993, + "step": 69547 + }, + { + "epoch": 5.14, + "learning_rate": 1.060374215397897e-06, + "loss": 1.0611, + "step": 69548 + }, + { + "epoch": 5.14, + "learning_rate": 1.0601954715368357e-06, + "loss": 0.8343, + "step": 69549 + }, + { + "epoch": 5.14, + "learning_rate": 1.0600167418988782e-06, + "loss": 1.0231, + "step": 69550 + }, + { + "epoch": 5.14, + "learning_rate": 1.0598380264843066e-06, + "loss": 0.8961, + "step": 69551 + }, + { + "epoch": 5.14, + "learning_rate": 1.0596593252934039e-06, + "loss": 0.9892, + "step": 69552 + }, + { + "epoch": 5.14, + "learning_rate": 1.0594806383264554e-06, + "loss": 1.0611, + "step": 69553 + }, + { + "epoch": 5.14, + "learning_rate": 1.0593019655837444e-06, + "loss": 0.9487, + "step": 69554 + }, + { + "epoch": 5.14, + "learning_rate": 1.0591233070655581e-06, + "loss": 1.0217, + "step": 69555 + }, + { + "epoch": 5.14, + "learning_rate": 1.0589446627721766e-06, + "loss": 1.0041, + "step": 69556 + }, + { + "epoch": 5.14, + "learning_rate": 1.0587660327038873e-06, + "loss": 1.0307, + "step": 69557 + }, + { + "epoch": 5.14, + "learning_rate": 1.058587416860971e-06, + "loss": 1.0468, + "step": 69558 + }, + { + "epoch": 5.14, + "learning_rate": 1.058408815243719e-06, + "loss": 0.9627, + "step": 69559 + }, + { + "epoch": 5.14, + "learning_rate": 1.0582302278524058e-06, + "loss": 1.0942, + "step": 69560 + }, + { + "epoch": 5.14, + "learning_rate": 1.0580516546873232e-06, + "loss": 1.0112, + "step": 69561 + }, + { + "epoch": 5.14, + "learning_rate": 1.0578730957487525e-06, + "loss": 1.0741, + "step": 69562 + }, + { + "epoch": 5.14, + "learning_rate": 1.0576945510369773e-06, + "loss": 1.0436, + "step": 69563 + }, + { + "epoch": 5.14, + "learning_rate": 1.0575160205522828e-06, + "loss": 0.9699, + "step": 69564 + }, + { + "epoch": 5.14, + "learning_rate": 1.0573375042949497e-06, + "loss": 1.0003, + "step": 69565 + }, + { + "epoch": 5.14, + "learning_rate": 1.0571590022652667e-06, + "loss": 0.852, + "step": 69566 + }, + { + "epoch": 5.14, + "learning_rate": 1.056980514463516e-06, + "loss": 1.0437, + "step": 69567 + }, + { + "epoch": 5.14, + "learning_rate": 1.0568020408899805e-06, + "loss": 0.9505, + "step": 69568 + }, + { + "epoch": 5.14, + "learning_rate": 1.0566235815449444e-06, + "loss": 0.9838, + "step": 69569 + }, + { + "epoch": 5.14, + "learning_rate": 1.0564451364286931e-06, + "loss": 1.0936, + "step": 69570 + }, + { + "epoch": 5.14, + "learning_rate": 1.0562667055415098e-06, + "loss": 1.0387, + "step": 69571 + }, + { + "epoch": 5.14, + "learning_rate": 1.0560882888836788e-06, + "loss": 0.9509, + "step": 69572 + }, + { + "epoch": 5.14, + "learning_rate": 1.0559098864554829e-06, + "loss": 1.0335, + "step": 69573 + }, + { + "epoch": 5.14, + "learning_rate": 1.0557314982572041e-06, + "loss": 0.9329, + "step": 69574 + }, + { + "epoch": 5.14, + "learning_rate": 1.0555531242891314e-06, + "loss": 1.1446, + "step": 69575 + }, + { + "epoch": 5.14, + "learning_rate": 1.0553747645515455e-06, + "loss": 1.055, + "step": 69576 + }, + { + "epoch": 5.14, + "learning_rate": 1.0551964190447294e-06, + "loss": 0.9223, + "step": 69577 + }, + { + "epoch": 5.14, + "learning_rate": 1.0550180877689676e-06, + "loss": 0.8849, + "step": 69578 + }, + { + "epoch": 5.14, + "learning_rate": 1.054839770724545e-06, + "loss": 1.0573, + "step": 69579 + }, + { + "epoch": 5.14, + "learning_rate": 1.0546614679117462e-06, + "loss": 1.0891, + "step": 69580 + }, + { + "epoch": 5.14, + "learning_rate": 1.0544831793308486e-06, + "loss": 0.9162, + "step": 69581 + }, + { + "epoch": 5.14, + "learning_rate": 1.054304904982143e-06, + "loss": 0.8319, + "step": 69582 + }, + { + "epoch": 5.14, + "learning_rate": 1.0541266448659083e-06, + "loss": 1.1003, + "step": 69583 + }, + { + "epoch": 5.14, + "learning_rate": 1.053948398982434e-06, + "loss": 0.9725, + "step": 69584 + }, + { + "epoch": 5.14, + "learning_rate": 1.0537701673319955e-06, + "loss": 0.8783, + "step": 69585 + }, + { + "epoch": 5.14, + "learning_rate": 1.0535919499148827e-06, + "loss": 1.0479, + "step": 69586 + }, + { + "epoch": 5.14, + "learning_rate": 1.0534137467313755e-06, + "loss": 0.9767, + "step": 69587 + }, + { + "epoch": 5.14, + "learning_rate": 1.0532355577817632e-06, + "loss": 1.0087, + "step": 69588 + }, + { + "epoch": 5.14, + "learning_rate": 1.0530573830663227e-06, + "loss": 1.0303, + "step": 69589 + }, + { + "epoch": 5.14, + "learning_rate": 1.052879222585338e-06, + "loss": 1.1282, + "step": 69590 + }, + { + "epoch": 5.14, + "learning_rate": 1.0527010763390965e-06, + "loss": 1.0003, + "step": 69591 + }, + { + "epoch": 5.14, + "learning_rate": 1.0525229443278796e-06, + "loss": 1.0881, + "step": 69592 + }, + { + "epoch": 5.14, + "learning_rate": 1.0523448265519697e-06, + "loss": 0.9034, + "step": 69593 + }, + { + "epoch": 5.14, + "learning_rate": 1.0521667230116507e-06, + "loss": 1.0114, + "step": 69594 + }, + { + "epoch": 5.14, + "learning_rate": 1.051988633707207e-06, + "loss": 0.8636, + "step": 69595 + }, + { + "epoch": 5.14, + "learning_rate": 1.0518105586389227e-06, + "loss": 1.0745, + "step": 69596 + }, + { + "epoch": 5.14, + "learning_rate": 1.051632497807079e-06, + "loss": 0.9227, + "step": 69597 + }, + { + "epoch": 5.14, + "learning_rate": 1.0514544512119595e-06, + "loss": 0.9328, + "step": 69598 + }, + { + "epoch": 5.14, + "learning_rate": 1.0512764188538471e-06, + "loss": 1.0407, + "step": 69599 + }, + { + "epoch": 5.14, + "learning_rate": 1.0510984007330294e-06, + "loss": 1.036, + "step": 69600 + }, + { + "epoch": 5.14, + "learning_rate": 1.0509203968497816e-06, + "loss": 1.0503, + "step": 69601 + }, + { + "epoch": 5.14, + "learning_rate": 1.0507424072043936e-06, + "loss": 1.0, + "step": 69602 + }, + { + "epoch": 5.14, + "learning_rate": 1.0505644317971452e-06, + "loss": 1.0482, + "step": 69603 + }, + { + "epoch": 5.14, + "learning_rate": 1.050386470628324e-06, + "loss": 0.8888, + "step": 69604 + }, + { + "epoch": 5.14, + "learning_rate": 1.0502085236982063e-06, + "loss": 1.0003, + "step": 69605 + }, + { + "epoch": 5.14, + "learning_rate": 1.0500305910070808e-06, + "loss": 0.9097, + "step": 69606 + }, + { + "epoch": 5.14, + "learning_rate": 1.049852672555226e-06, + "loss": 1.0414, + "step": 69607 + }, + { + "epoch": 5.14, + "learning_rate": 1.0496747683429321e-06, + "loss": 0.9275, + "step": 69608 + }, + { + "epoch": 5.14, + "learning_rate": 1.049496878370475e-06, + "loss": 0.8699, + "step": 69609 + }, + { + "epoch": 5.14, + "learning_rate": 1.0493190026381383e-06, + "loss": 0.9774, + "step": 69610 + }, + { + "epoch": 5.14, + "learning_rate": 1.0491411411462094e-06, + "loss": 0.9678, + "step": 69611 + }, + { + "epoch": 5.14, + "learning_rate": 1.048963293894969e-06, + "loss": 1.0487, + "step": 69612 + }, + { + "epoch": 5.14, + "learning_rate": 1.0487854608847003e-06, + "loss": 0.9864, + "step": 69613 + }, + { + "epoch": 5.14, + "learning_rate": 1.0486076421156833e-06, + "loss": 0.998, + "step": 69614 + }, + { + "epoch": 5.14, + "learning_rate": 1.0484298375882052e-06, + "loss": 1.1333, + "step": 69615 + }, + { + "epoch": 5.14, + "learning_rate": 1.048252047302547e-06, + "loss": 1.0186, + "step": 69616 + }, + { + "epoch": 5.14, + "learning_rate": 1.0480742712589932e-06, + "loss": 0.9751, + "step": 69617 + }, + { + "epoch": 5.14, + "learning_rate": 1.0478965094578242e-06, + "loss": 1.0291, + "step": 69618 + }, + { + "epoch": 5.14, + "learning_rate": 1.0477187618993213e-06, + "loss": 0.9485, + "step": 69619 + }, + { + "epoch": 5.14, + "learning_rate": 1.0475410285837718e-06, + "loss": 1.0404, + "step": 69620 + }, + { + "epoch": 5.14, + "learning_rate": 1.0473633095114566e-06, + "loss": 0.9166, + "step": 69621 + }, + { + "epoch": 5.14, + "learning_rate": 1.047185604682659e-06, + "loss": 0.9872, + "step": 69622 + }, + { + "epoch": 5.14, + "learning_rate": 1.0470079140976585e-06, + "loss": 1.0448, + "step": 69623 + }, + { + "epoch": 5.14, + "learning_rate": 1.046830237756742e-06, + "loss": 1.0797, + "step": 69624 + }, + { + "epoch": 5.14, + "learning_rate": 1.046652575660191e-06, + "loss": 1.0729, + "step": 69625 + }, + { + "epoch": 5.14, + "learning_rate": 1.0464749278082876e-06, + "loss": 0.998, + "step": 69626 + }, + { + "epoch": 5.14, + "learning_rate": 1.0462972942013139e-06, + "loss": 1.0588, + "step": 69627 + }, + { + "epoch": 5.14, + "learning_rate": 1.046119674839552e-06, + "loss": 1.0816, + "step": 69628 + }, + { + "epoch": 5.14, + "learning_rate": 1.045942069723289e-06, + "loss": 1.03, + "step": 69629 + }, + { + "epoch": 5.14, + "learning_rate": 1.0457644788527999e-06, + "loss": 1.0936, + "step": 69630 + }, + { + "epoch": 5.14, + "learning_rate": 1.0455869022283738e-06, + "loss": 0.9416, + "step": 69631 + }, + { + "epoch": 5.15, + "learning_rate": 1.0454093398502907e-06, + "loss": 1.0322, + "step": 69632 + }, + { + "epoch": 5.15, + "learning_rate": 1.0452317917188336e-06, + "loss": 1.0385, + "step": 69633 + }, + { + "epoch": 5.15, + "learning_rate": 1.0450542578342814e-06, + "loss": 1.0298, + "step": 69634 + }, + { + "epoch": 5.15, + "learning_rate": 1.0448767381969227e-06, + "loss": 0.984, + "step": 69635 + }, + { + "epoch": 5.15, + "learning_rate": 1.0446992328070372e-06, + "loss": 1.0369, + "step": 69636 + }, + { + "epoch": 5.15, + "learning_rate": 1.0445217416649068e-06, + "loss": 1.05, + "step": 69637 + }, + { + "epoch": 5.15, + "learning_rate": 1.0443442647708135e-06, + "loss": 1.013, + "step": 69638 + }, + { + "epoch": 5.15, + "learning_rate": 1.0441668021250396e-06, + "loss": 0.9985, + "step": 69639 + }, + { + "epoch": 5.15, + "learning_rate": 1.0439893537278689e-06, + "loss": 0.93, + "step": 69640 + }, + { + "epoch": 5.15, + "learning_rate": 1.0438119195795838e-06, + "loss": 0.9724, + "step": 69641 + }, + { + "epoch": 5.15, + "learning_rate": 1.043634499680466e-06, + "loss": 1.0824, + "step": 69642 + }, + { + "epoch": 5.15, + "learning_rate": 1.0434570940307954e-06, + "loss": 1.0073, + "step": 69643 + }, + { + "epoch": 5.15, + "learning_rate": 1.0432797026308584e-06, + "loss": 0.9861, + "step": 69644 + }, + { + "epoch": 5.15, + "learning_rate": 1.043102325480937e-06, + "loss": 1.0371, + "step": 69645 + }, + { + "epoch": 5.15, + "learning_rate": 1.0429249625813076e-06, + "loss": 0.9939, + "step": 69646 + }, + { + "epoch": 5.15, + "learning_rate": 1.0427476139322578e-06, + "loss": 1.0345, + "step": 69647 + }, + { + "epoch": 5.15, + "learning_rate": 1.0425702795340675e-06, + "loss": 1.0155, + "step": 69648 + }, + { + "epoch": 5.15, + "learning_rate": 1.042392959387023e-06, + "loss": 0.9758, + "step": 69649 + }, + { + "epoch": 5.15, + "learning_rate": 1.0422156534913997e-06, + "loss": 1.0424, + "step": 69650 + }, + { + "epoch": 5.15, + "learning_rate": 1.042038361847485e-06, + "loss": 0.962, + "step": 69651 + }, + { + "epoch": 5.15, + "learning_rate": 1.0418610844555576e-06, + "loss": 0.913, + "step": 69652 + }, + { + "epoch": 5.15, + "learning_rate": 1.041683821315904e-06, + "loss": 1.1344, + "step": 69653 + }, + { + "epoch": 5.15, + "learning_rate": 1.041506572428802e-06, + "loss": 1.0301, + "step": 69654 + }, + { + "epoch": 5.15, + "learning_rate": 1.0413293377945323e-06, + "loss": 0.9583, + "step": 69655 + }, + { + "epoch": 5.15, + "learning_rate": 1.041152117413382e-06, + "loss": 0.9091, + "step": 69656 + }, + { + "epoch": 5.15, + "learning_rate": 1.0409749112856305e-06, + "loss": 0.9313, + "step": 69657 + }, + { + "epoch": 5.15, + "learning_rate": 1.0407977194115593e-06, + "loss": 1.0379, + "step": 69658 + }, + { + "epoch": 5.15, + "learning_rate": 1.0406205417914484e-06, + "loss": 0.966, + "step": 69659 + }, + { + "epoch": 5.15, + "learning_rate": 1.0404433784255852e-06, + "loss": 0.9265, + "step": 69660 + }, + { + "epoch": 5.15, + "learning_rate": 1.0402662293142473e-06, + "loss": 1.0828, + "step": 69661 + }, + { + "epoch": 5.15, + "learning_rate": 1.040089094457718e-06, + "loss": 1.0128, + "step": 69662 + }, + { + "epoch": 5.15, + "learning_rate": 1.039911973856277e-06, + "loss": 1.0955, + "step": 69663 + }, + { + "epoch": 5.15, + "learning_rate": 1.0397348675102092e-06, + "loss": 0.9564, + "step": 69664 + }, + { + "epoch": 5.15, + "learning_rate": 1.039557775419796e-06, + "loss": 0.934, + "step": 69665 + }, + { + "epoch": 5.15, + "learning_rate": 1.039380697585317e-06, + "loss": 1.0773, + "step": 69666 + }, + { + "epoch": 5.15, + "learning_rate": 1.0392036340070566e-06, + "loss": 0.9073, + "step": 69667 + }, + { + "epoch": 5.15, + "learning_rate": 1.039026584685292e-06, + "loss": 1.0618, + "step": 69668 + }, + { + "epoch": 5.15, + "learning_rate": 1.03884954962031e-06, + "loss": 0.951, + "step": 69669 + }, + { + "epoch": 5.15, + "learning_rate": 1.03867252881239e-06, + "loss": 0.8993, + "step": 69670 + }, + { + "epoch": 5.15, + "learning_rate": 1.0384955222618142e-06, + "loss": 1.0603, + "step": 69671 + }, + { + "epoch": 5.15, + "learning_rate": 1.0383185299688613e-06, + "loss": 1.1, + "step": 69672 + }, + { + "epoch": 5.15, + "learning_rate": 1.0381415519338177e-06, + "loss": 0.9202, + "step": 69673 + }, + { + "epoch": 5.15, + "learning_rate": 1.0379645881569655e-06, + "loss": 1.0875, + "step": 69674 + }, + { + "epoch": 5.15, + "learning_rate": 1.0377876386385789e-06, + "loss": 0.8521, + "step": 69675 + }, + { + "epoch": 5.15, + "learning_rate": 1.0376107033789452e-06, + "loss": 0.9287, + "step": 69676 + }, + { + "epoch": 5.15, + "learning_rate": 1.0374337823783454e-06, + "loss": 1.1117, + "step": 69677 + }, + { + "epoch": 5.15, + "learning_rate": 1.0372568756370594e-06, + "loss": 1.0244, + "step": 69678 + }, + { + "epoch": 5.15, + "learning_rate": 1.037079983155368e-06, + "loss": 0.9554, + "step": 69679 + }, + { + "epoch": 5.15, + "learning_rate": 1.0369031049335565e-06, + "loss": 0.9713, + "step": 69680 + }, + { + "epoch": 5.15, + "learning_rate": 1.0367262409719037e-06, + "loss": 1.0497, + "step": 69681 + }, + { + "epoch": 5.15, + "learning_rate": 1.0365493912706904e-06, + "loss": 0.9508, + "step": 69682 + }, + { + "epoch": 5.15, + "learning_rate": 1.0363725558301997e-06, + "loss": 0.9335, + "step": 69683 + }, + { + "epoch": 5.15, + "learning_rate": 1.03619573465071e-06, + "loss": 0.9212, + "step": 69684 + }, + { + "epoch": 5.15, + "learning_rate": 1.0360189277325061e-06, + "loss": 0.898, + "step": 69685 + }, + { + "epoch": 5.15, + "learning_rate": 1.0358421350758685e-06, + "loss": 1.0373, + "step": 69686 + }, + { + "epoch": 5.15, + "learning_rate": 1.035665356681077e-06, + "loss": 0.995, + "step": 69687 + }, + { + "epoch": 5.15, + "learning_rate": 1.0354885925484125e-06, + "loss": 0.9575, + "step": 69688 + }, + { + "epoch": 5.15, + "learning_rate": 1.0353118426781583e-06, + "loss": 0.8659, + "step": 69689 + }, + { + "epoch": 5.15, + "learning_rate": 1.0351351070705951e-06, + "loss": 1.0926, + "step": 69690 + }, + { + "epoch": 5.15, + "learning_rate": 1.0349583857260048e-06, + "loss": 0.9722, + "step": 69691 + }, + { + "epoch": 5.15, + "learning_rate": 1.0347816786446662e-06, + "loss": 1.1134, + "step": 69692 + }, + { + "epoch": 5.15, + "learning_rate": 1.0346049858268602e-06, + "loss": 0.9861, + "step": 69693 + }, + { + "epoch": 5.15, + "learning_rate": 1.0344283072728734e-06, + "loss": 1.0298, + "step": 69694 + }, + { + "epoch": 5.15, + "learning_rate": 1.0342516429829785e-06, + "loss": 1.0126, + "step": 69695 + }, + { + "epoch": 5.15, + "learning_rate": 1.0340749929574635e-06, + "loss": 1.0023, + "step": 69696 + }, + { + "epoch": 5.15, + "learning_rate": 1.0338983571966044e-06, + "loss": 0.9664, + "step": 69697 + }, + { + "epoch": 5.15, + "learning_rate": 1.033721735700689e-06, + "loss": 1.0418, + "step": 69698 + }, + { + "epoch": 5.15, + "learning_rate": 1.0335451284699893e-06, + "loss": 1.0419, + "step": 69699 + }, + { + "epoch": 5.15, + "learning_rate": 1.033368535504794e-06, + "loss": 0.9875, + "step": 69700 + }, + { + "epoch": 5.15, + "learning_rate": 1.0331919568053805e-06, + "loss": 1.0173, + "step": 69701 + }, + { + "epoch": 5.15, + "learning_rate": 1.0330153923720309e-06, + "loss": 0.9978, + "step": 69702 + }, + { + "epoch": 5.15, + "learning_rate": 1.0328388422050251e-06, + "loss": 1.0695, + "step": 69703 + }, + { + "epoch": 5.15, + "learning_rate": 1.0326623063046425e-06, + "loss": 1.0199, + "step": 69704 + }, + { + "epoch": 5.15, + "learning_rate": 1.0324857846711677e-06, + "loss": 0.9702, + "step": 69705 + }, + { + "epoch": 5.15, + "learning_rate": 1.0323092773048793e-06, + "loss": 0.9232, + "step": 69706 + }, + { + "epoch": 5.15, + "learning_rate": 1.0321327842060591e-06, + "loss": 0.9908, + "step": 69707 + }, + { + "epoch": 5.15, + "learning_rate": 1.031956305374986e-06, + "loss": 1.1102, + "step": 69708 + }, + { + "epoch": 5.15, + "learning_rate": 1.0317798408119429e-06, + "loss": 0.9032, + "step": 69709 + }, + { + "epoch": 5.15, + "learning_rate": 1.0316033905172106e-06, + "loss": 1.0934, + "step": 69710 + }, + { + "epoch": 5.15, + "learning_rate": 1.0314269544910681e-06, + "loss": 1.0687, + "step": 69711 + }, + { + "epoch": 5.15, + "learning_rate": 1.0312505327337975e-06, + "loss": 1.0708, + "step": 69712 + }, + { + "epoch": 5.15, + "learning_rate": 1.0310741252456768e-06, + "loss": 1.0012, + "step": 69713 + }, + { + "epoch": 5.15, + "learning_rate": 1.030897732026992e-06, + "loss": 1.0345, + "step": 69714 + }, + { + "epoch": 5.15, + "learning_rate": 1.0307213530780191e-06, + "loss": 0.8884, + "step": 69715 + }, + { + "epoch": 5.15, + "learning_rate": 1.0305449883990405e-06, + "loss": 0.9786, + "step": 69716 + }, + { + "epoch": 5.15, + "learning_rate": 1.0303686379903354e-06, + "loss": 1.0859, + "step": 69717 + }, + { + "epoch": 5.15, + "learning_rate": 1.0301923018521865e-06, + "loss": 0.9404, + "step": 69718 + }, + { + "epoch": 5.15, + "learning_rate": 1.0300159799848741e-06, + "loss": 0.9594, + "step": 69719 + }, + { + "epoch": 5.15, + "learning_rate": 1.0298396723886772e-06, + "loss": 1.056, + "step": 69720 + }, + { + "epoch": 5.15, + "learning_rate": 1.0296633790638766e-06, + "loss": 0.9502, + "step": 69721 + }, + { + "epoch": 5.15, + "learning_rate": 1.0294871000107542e-06, + "loss": 0.9804, + "step": 69722 + }, + { + "epoch": 5.15, + "learning_rate": 1.0293108352295888e-06, + "loss": 0.9828, + "step": 69723 + }, + { + "epoch": 5.15, + "learning_rate": 1.0291345847206602e-06, + "loss": 0.9931, + "step": 69724 + }, + { + "epoch": 5.15, + "learning_rate": 1.0289583484842525e-06, + "loss": 0.8654, + "step": 69725 + }, + { + "epoch": 5.15, + "learning_rate": 1.028782126520642e-06, + "loss": 1.1397, + "step": 69726 + }, + { + "epoch": 5.15, + "learning_rate": 1.0286059188301123e-06, + "loss": 1.1508, + "step": 69727 + }, + { + "epoch": 5.15, + "learning_rate": 1.0284297254129395e-06, + "loss": 1.1112, + "step": 69728 + }, + { + "epoch": 5.15, + "learning_rate": 1.0282535462694088e-06, + "loss": 1.0838, + "step": 69729 + }, + { + "epoch": 5.15, + "learning_rate": 1.028077381399799e-06, + "loss": 1.0225, + "step": 69730 + }, + { + "epoch": 5.15, + "learning_rate": 1.027901230804389e-06, + "loss": 0.9308, + "step": 69731 + }, + { + "epoch": 5.15, + "learning_rate": 1.0277250944834594e-06, + "loss": 0.9651, + "step": 69732 + }, + { + "epoch": 5.15, + "learning_rate": 1.02754897243729e-06, + "loss": 1.038, + "step": 69733 + }, + { + "epoch": 5.15, + "learning_rate": 1.027372864666163e-06, + "loss": 0.9722, + "step": 69734 + }, + { + "epoch": 5.15, + "learning_rate": 1.027196771170358e-06, + "loss": 1.0061, + "step": 69735 + }, + { + "epoch": 5.15, + "learning_rate": 1.0270206919501536e-06, + "loss": 0.9437, + "step": 69736 + }, + { + "epoch": 5.15, + "learning_rate": 1.0268446270058295e-06, + "loss": 1.0414, + "step": 69737 + }, + { + "epoch": 5.15, + "learning_rate": 1.0266685763376695e-06, + "loss": 0.8751, + "step": 69738 + }, + { + "epoch": 5.15, + "learning_rate": 1.0264925399459524e-06, + "loss": 0.9471, + "step": 69739 + }, + { + "epoch": 5.15, + "learning_rate": 1.0263165178309543e-06, + "loss": 0.9884, + "step": 69740 + }, + { + "epoch": 5.15, + "learning_rate": 1.0261405099929588e-06, + "loss": 0.996, + "step": 69741 + }, + { + "epoch": 5.15, + "learning_rate": 1.0259645164322452e-06, + "loss": 1.0157, + "step": 69742 + }, + { + "epoch": 5.15, + "learning_rate": 1.025788537149096e-06, + "loss": 1.0198, + "step": 69743 + }, + { + "epoch": 5.15, + "learning_rate": 1.025612572143786e-06, + "loss": 1.0936, + "step": 69744 + }, + { + "epoch": 5.15, + "learning_rate": 1.0254366214166e-06, + "loss": 1.0336, + "step": 69745 + }, + { + "epoch": 5.15, + "learning_rate": 1.0252606849678138e-06, + "loss": 0.8497, + "step": 69746 + }, + { + "epoch": 5.15, + "learning_rate": 1.0250847627977123e-06, + "loss": 1.0237, + "step": 69747 + }, + { + "epoch": 5.15, + "learning_rate": 1.0249088549065722e-06, + "loss": 0.8712, + "step": 69748 + }, + { + "epoch": 5.15, + "learning_rate": 1.024732961294671e-06, + "loss": 1.0405, + "step": 69749 + }, + { + "epoch": 5.15, + "learning_rate": 1.024557081962293e-06, + "loss": 1.0567, + "step": 69750 + }, + { + "epoch": 5.15, + "learning_rate": 1.0243812169097167e-06, + "loss": 1.0151, + "step": 69751 + }, + { + "epoch": 5.15, + "learning_rate": 1.0242053661372209e-06, + "loss": 1.0458, + "step": 69752 + }, + { + "epoch": 5.15, + "learning_rate": 1.0240295296450841e-06, + "loss": 1.0225, + "step": 69753 + }, + { + "epoch": 5.15, + "learning_rate": 1.0238537074335908e-06, + "loss": 0.9184, + "step": 69754 + }, + { + "epoch": 5.15, + "learning_rate": 1.023677899503016e-06, + "loss": 1.0621, + "step": 69755 + }, + { + "epoch": 5.15, + "learning_rate": 1.0235021058536431e-06, + "loss": 1.0068, + "step": 69756 + }, + { + "epoch": 5.15, + "learning_rate": 1.0233263264857484e-06, + "loss": 1.0481, + "step": 69757 + }, + { + "epoch": 5.15, + "learning_rate": 1.0231505613996119e-06, + "loss": 1.0541, + "step": 69758 + }, + { + "epoch": 5.15, + "learning_rate": 1.0229748105955161e-06, + "loss": 0.9344, + "step": 69759 + }, + { + "epoch": 5.15, + "learning_rate": 1.0227990740737393e-06, + "loss": 0.9796, + "step": 69760 + }, + { + "epoch": 5.15, + "learning_rate": 1.0226233518345607e-06, + "loss": 1.0058, + "step": 69761 + }, + { + "epoch": 5.15, + "learning_rate": 1.0224476438782583e-06, + "loss": 0.9609, + "step": 69762 + }, + { + "epoch": 5.15, + "learning_rate": 1.0222719502051148e-06, + "loss": 0.9799, + "step": 69763 + }, + { + "epoch": 5.15, + "learning_rate": 1.022096270815408e-06, + "loss": 1.0143, + "step": 69764 + }, + { + "epoch": 5.15, + "learning_rate": 1.0219206057094178e-06, + "loss": 1.0346, + "step": 69765 + }, + { + "epoch": 5.15, + "learning_rate": 1.0217449548874215e-06, + "loss": 0.9788, + "step": 69766 + }, + { + "epoch": 5.16, + "learning_rate": 1.0215693183497045e-06, + "loss": 0.9436, + "step": 69767 + }, + { + "epoch": 5.16, + "learning_rate": 1.0213936960965398e-06, + "loss": 1.0213, + "step": 69768 + }, + { + "epoch": 5.16, + "learning_rate": 1.0212180881282074e-06, + "loss": 1.0301, + "step": 69769 + }, + { + "epoch": 5.16, + "learning_rate": 1.0210424944449903e-06, + "loss": 1.0086, + "step": 69770 + }, + { + "epoch": 5.16, + "learning_rate": 1.0208669150471672e-06, + "loss": 1.022, + "step": 69771 + }, + { + "epoch": 5.16, + "learning_rate": 1.0206913499350158e-06, + "loss": 0.9702, + "step": 69772 + }, + { + "epoch": 5.16, + "learning_rate": 1.0205157991088143e-06, + "loss": 0.937, + "step": 69773 + }, + { + "epoch": 5.16, + "learning_rate": 1.0203402625688452e-06, + "loss": 1.1082, + "step": 69774 + }, + { + "epoch": 5.16, + "learning_rate": 1.0201647403153857e-06, + "loss": 1.1404, + "step": 69775 + }, + { + "epoch": 5.16, + "learning_rate": 1.019989232348717e-06, + "loss": 1.065, + "step": 69776 + }, + { + "epoch": 5.16, + "learning_rate": 1.0198137386691165e-06, + "loss": 0.9391, + "step": 69777 + }, + { + "epoch": 5.16, + "learning_rate": 1.0196382592768617e-06, + "loss": 1.116, + "step": 69778 + }, + { + "epoch": 5.16, + "learning_rate": 1.0194627941722358e-06, + "loss": 1.0824, + "step": 69779 + }, + { + "epoch": 5.16, + "learning_rate": 1.019287343355516e-06, + "loss": 1.209, + "step": 69780 + }, + { + "epoch": 5.16, + "learning_rate": 1.0191119068269817e-06, + "loss": 1.0343, + "step": 69781 + }, + { + "epoch": 5.16, + "learning_rate": 1.0189364845869099e-06, + "loss": 0.9506, + "step": 69782 + }, + { + "epoch": 5.16, + "learning_rate": 1.0187610766355838e-06, + "loss": 0.9939, + "step": 69783 + }, + { + "epoch": 5.16, + "learning_rate": 1.0185856829732798e-06, + "loss": 0.9881, + "step": 69784 + }, + { + "epoch": 5.16, + "learning_rate": 1.018410303600278e-06, + "loss": 1.0621, + "step": 69785 + }, + { + "epoch": 5.16, + "learning_rate": 1.0182349385168567e-06, + "loss": 1.0561, + "step": 69786 + }, + { + "epoch": 5.16, + "learning_rate": 1.0180595877232923e-06, + "loss": 0.9708, + "step": 69787 + }, + { + "epoch": 5.16, + "learning_rate": 1.0178842512198717e-06, + "loss": 0.9015, + "step": 69788 + }, + { + "epoch": 5.16, + "learning_rate": 1.0177089290068643e-06, + "loss": 1.0017, + "step": 69789 + }, + { + "epoch": 5.16, + "learning_rate": 1.0175336210845566e-06, + "loss": 0.9886, + "step": 69790 + }, + { + "epoch": 5.16, + "learning_rate": 1.0173583274532206e-06, + "loss": 1.0446, + "step": 69791 + }, + { + "epoch": 5.16, + "learning_rate": 1.017183048113144e-06, + "loss": 1.0707, + "step": 69792 + }, + { + "epoch": 5.16, + "learning_rate": 1.0170077830645964e-06, + "loss": 0.9667, + "step": 69793 + }, + { + "epoch": 5.16, + "learning_rate": 1.0168325323078632e-06, + "loss": 0.985, + "step": 69794 + }, + { + "epoch": 5.16, + "learning_rate": 1.0166572958432208e-06, + "loss": 1.0224, + "step": 69795 + }, + { + "epoch": 5.16, + "learning_rate": 1.016482073670948e-06, + "loss": 1.0334, + "step": 69796 + }, + { + "epoch": 5.16, + "learning_rate": 1.0163068657913233e-06, + "loss": 0.9958, + "step": 69797 + }, + { + "epoch": 5.16, + "learning_rate": 1.0161316722046243e-06, + "loss": 1.0119, + "step": 69798 + }, + { + "epoch": 5.16, + "learning_rate": 1.0159564929111333e-06, + "loss": 1.0147, + "step": 69799 + }, + { + "epoch": 5.16, + "learning_rate": 1.0157813279111262e-06, + "loss": 1.0161, + "step": 69800 + }, + { + "epoch": 5.16, + "learning_rate": 1.015606177204883e-06, + "loss": 0.9092, + "step": 69801 + }, + { + "epoch": 5.16, + "learning_rate": 1.0154310407926803e-06, + "loss": 0.9007, + "step": 69802 + }, + { + "epoch": 5.16, + "learning_rate": 1.0152559186748e-06, + "loss": 0.9113, + "step": 69803 + }, + { + "epoch": 5.16, + "learning_rate": 1.0150808108515187e-06, + "loss": 0.9663, + "step": 69804 + }, + { + "epoch": 5.16, + "learning_rate": 1.0149057173231159e-06, + "loss": 1.0424, + "step": 69805 + }, + { + "epoch": 5.16, + "learning_rate": 1.014730638089869e-06, + "loss": 1.0302, + "step": 69806 + }, + { + "epoch": 5.16, + "learning_rate": 1.0145555731520552e-06, + "loss": 1.0025, + "step": 69807 + }, + { + "epoch": 5.16, + "learning_rate": 1.0143805225099569e-06, + "loss": 1.0944, + "step": 69808 + }, + { + "epoch": 5.16, + "learning_rate": 1.0142054861638506e-06, + "loss": 1.0962, + "step": 69809 + }, + { + "epoch": 5.16, + "learning_rate": 1.0140304641140152e-06, + "loss": 0.9515, + "step": 69810 + }, + { + "epoch": 5.16, + "learning_rate": 1.013855456360726e-06, + "loss": 0.948, + "step": 69811 + }, + { + "epoch": 5.16, + "learning_rate": 1.0136804629042684e-06, + "loss": 1.0774, + "step": 69812 + }, + { + "epoch": 5.16, + "learning_rate": 1.0135054837449155e-06, + "loss": 1.0664, + "step": 69813 + }, + { + "epoch": 5.16, + "learning_rate": 1.0133305188829445e-06, + "loss": 1.0759, + "step": 69814 + }, + { + "epoch": 5.16, + "learning_rate": 1.0131555683186367e-06, + "loss": 1.0417, + "step": 69815 + }, + { + "epoch": 5.16, + "learning_rate": 1.0129806320522718e-06, + "loss": 0.9711, + "step": 69816 + }, + { + "epoch": 5.16, + "learning_rate": 1.0128057100841248e-06, + "loss": 1.0086, + "step": 69817 + }, + { + "epoch": 5.16, + "learning_rate": 1.0126308024144737e-06, + "loss": 1.0689, + "step": 69818 + }, + { + "epoch": 5.16, + "learning_rate": 1.0124559090436003e-06, + "loss": 1.0041, + "step": 69819 + }, + { + "epoch": 5.16, + "learning_rate": 1.0122810299717811e-06, + "loss": 0.9977, + "step": 69820 + }, + { + "epoch": 5.16, + "learning_rate": 1.0121061651992936e-06, + "loss": 1.069, + "step": 69821 + }, + { + "epoch": 5.16, + "learning_rate": 1.0119313147264153e-06, + "loss": 1.0459, + "step": 69822 + }, + { + "epoch": 5.16, + "learning_rate": 1.0117564785534273e-06, + "loss": 0.8277, + "step": 69823 + }, + { + "epoch": 5.16, + "learning_rate": 1.011581656680607e-06, + "loss": 0.9324, + "step": 69824 + }, + { + "epoch": 5.16, + "learning_rate": 1.0114068491082306e-06, + "loss": 0.9716, + "step": 69825 + }, + { + "epoch": 5.16, + "learning_rate": 1.0112320558365785e-06, + "loss": 1.0887, + "step": 69826 + }, + { + "epoch": 5.16, + "learning_rate": 1.0110572768659243e-06, + "loss": 0.9231, + "step": 69827 + }, + { + "epoch": 5.16, + "learning_rate": 1.0108825121965515e-06, + "loss": 1.0305, + "step": 69828 + }, + { + "epoch": 5.16, + "learning_rate": 1.0107077618287364e-06, + "loss": 0.947, + "step": 69829 + }, + { + "epoch": 5.16, + "learning_rate": 1.0105330257627565e-06, + "loss": 0.992, + "step": 69830 + }, + { + "epoch": 5.16, + "learning_rate": 1.0103583039988885e-06, + "loss": 1.0356, + "step": 69831 + }, + { + "epoch": 5.16, + "learning_rate": 1.0101835965374128e-06, + "loss": 1.0066, + "step": 69832 + }, + { + "epoch": 5.16, + "learning_rate": 1.0100089033786098e-06, + "loss": 0.994, + "step": 69833 + }, + { + "epoch": 5.16, + "learning_rate": 1.0098342245227488e-06, + "loss": 1.0744, + "step": 69834 + }, + { + "epoch": 5.16, + "learning_rate": 1.0096595599701153e-06, + "loss": 0.9097, + "step": 69835 + }, + { + "epoch": 5.16, + "learning_rate": 1.0094849097209824e-06, + "loss": 0.9844, + "step": 69836 + }, + { + "epoch": 5.16, + "learning_rate": 1.0093102737756356e-06, + "loss": 0.9568, + "step": 69837 + }, + { + "epoch": 5.16, + "learning_rate": 1.0091356521343421e-06, + "loss": 0.9217, + "step": 69838 + }, + { + "epoch": 5.16, + "learning_rate": 1.0089610447973885e-06, + "loss": 1.0034, + "step": 69839 + }, + { + "epoch": 5.16, + "learning_rate": 1.008786451765046e-06, + "loss": 0.9653, + "step": 69840 + }, + { + "epoch": 5.16, + "learning_rate": 1.0086118730376004e-06, + "loss": 1.0111, + "step": 69841 + }, + { + "epoch": 5.16, + "learning_rate": 1.008437308615322e-06, + "loss": 0.8138, + "step": 69842 + }, + { + "epoch": 5.16, + "learning_rate": 1.0082627584984894e-06, + "loss": 1.0582, + "step": 69843 + }, + { + "epoch": 5.16, + "learning_rate": 1.0080882226873846e-06, + "loss": 0.9502, + "step": 69844 + }, + { + "epoch": 5.16, + "learning_rate": 1.0079137011822825e-06, + "loss": 1.0509, + "step": 69845 + }, + { + "epoch": 5.16, + "learning_rate": 1.0077391939834602e-06, + "loss": 0.999, + "step": 69846 + }, + { + "epoch": 5.16, + "learning_rate": 1.007564701091196e-06, + "loss": 1.0374, + "step": 69847 + }, + { + "epoch": 5.16, + "learning_rate": 1.0073902225057686e-06, + "loss": 0.9646, + "step": 69848 + }, + { + "epoch": 5.16, + "learning_rate": 1.0072157582274545e-06, + "loss": 1.0591, + "step": 69849 + }, + { + "epoch": 5.16, + "learning_rate": 1.007041308256531e-06, + "loss": 0.9259, + "step": 69850 + }, + { + "epoch": 5.16, + "learning_rate": 1.0068668725932773e-06, + "loss": 1.007, + "step": 69851 + }, + { + "epoch": 5.16, + "learning_rate": 1.0066924512379672e-06, + "loss": 0.9875, + "step": 69852 + }, + { + "epoch": 5.16, + "learning_rate": 1.006518044190883e-06, + "loss": 1.0674, + "step": 69853 + }, + { + "epoch": 5.16, + "learning_rate": 1.0063436514522996e-06, + "loss": 0.9808, + "step": 69854 + }, + { + "epoch": 5.16, + "learning_rate": 1.0061692730224949e-06, + "loss": 1.0142, + "step": 69855 + }, + { + "epoch": 5.16, + "learning_rate": 1.005994908901744e-06, + "loss": 1.0147, + "step": 69856 + }, + { + "epoch": 5.16, + "learning_rate": 1.0058205590903302e-06, + "loss": 0.9999, + "step": 69857 + }, + { + "epoch": 5.16, + "learning_rate": 1.0056462235885244e-06, + "loss": 1.0591, + "step": 69858 + }, + { + "epoch": 5.16, + "learning_rate": 1.0054719023966076e-06, + "loss": 1.0585, + "step": 69859 + }, + { + "epoch": 5.16, + "learning_rate": 1.005297595514857e-06, + "loss": 1.0214, + "step": 69860 + }, + { + "epoch": 5.16, + "learning_rate": 1.0051233029435491e-06, + "loss": 0.978, + "step": 69861 + }, + { + "epoch": 5.16, + "learning_rate": 1.0049490246829619e-06, + "loss": 1.0017, + "step": 69862 + }, + { + "epoch": 5.16, + "learning_rate": 1.0047747607333692e-06, + "loss": 1.1004, + "step": 69863 + }, + { + "epoch": 5.16, + "learning_rate": 1.004600511095053e-06, + "loss": 0.9793, + "step": 69864 + }, + { + "epoch": 5.16, + "learning_rate": 1.0044262757682887e-06, + "loss": 0.9329, + "step": 69865 + }, + { + "epoch": 5.16, + "learning_rate": 1.004252054753354e-06, + "loss": 1.0044, + "step": 69866 + }, + { + "epoch": 5.16, + "learning_rate": 1.0040778480505242e-06, + "loss": 0.9362, + "step": 69867 + }, + { + "epoch": 5.16, + "learning_rate": 1.003903655660079e-06, + "loss": 0.8545, + "step": 69868 + }, + { + "epoch": 5.16, + "learning_rate": 1.0037294775822938e-06, + "loss": 1.0598, + "step": 69869 + }, + { + "epoch": 5.16, + "learning_rate": 1.003555313817447e-06, + "loss": 0.9831, + "step": 69870 + }, + { + "epoch": 5.16, + "learning_rate": 1.0033811643658154e-06, + "loss": 0.9856, + "step": 69871 + }, + { + "epoch": 5.16, + "learning_rate": 1.0032070292276729e-06, + "loss": 1.0653, + "step": 69872 + }, + { + "epoch": 5.16, + "learning_rate": 1.003032908403302e-06, + "loss": 0.8984, + "step": 69873 + }, + { + "epoch": 5.16, + "learning_rate": 1.0028588018929763e-06, + "loss": 0.8979, + "step": 69874 + }, + { + "epoch": 5.16, + "learning_rate": 1.002684709696974e-06, + "loss": 1.0681, + "step": 69875 + }, + { + "epoch": 5.16, + "learning_rate": 1.0025106318155698e-06, + "loss": 0.9925, + "step": 69876 + }, + { + "epoch": 5.16, + "learning_rate": 1.0023365682490438e-06, + "loss": 1.0528, + "step": 69877 + }, + { + "epoch": 5.16, + "learning_rate": 1.0021625189976713e-06, + "loss": 0.9262, + "step": 69878 + }, + { + "epoch": 5.16, + "learning_rate": 1.00198848406173e-06, + "loss": 1.0334, + "step": 69879 + }, + { + "epoch": 5.16, + "learning_rate": 1.0018144634414962e-06, + "loss": 1.132, + "step": 69880 + }, + { + "epoch": 5.16, + "learning_rate": 1.0016404571372452e-06, + "loss": 1.035, + "step": 69881 + }, + { + "epoch": 5.16, + "learning_rate": 1.001466465149259e-06, + "loss": 1.0113, + "step": 69882 + }, + { + "epoch": 5.16, + "learning_rate": 1.0012924874778075e-06, + "loss": 1.0264, + "step": 69883 + }, + { + "epoch": 5.16, + "learning_rate": 1.0011185241231725e-06, + "loss": 1.1326, + "step": 69884 + }, + { + "epoch": 5.16, + "learning_rate": 1.0009445750856272e-06, + "loss": 0.8921, + "step": 69885 + }, + { + "epoch": 5.16, + "learning_rate": 1.000770640365455e-06, + "loss": 1.1337, + "step": 69886 + }, + { + "epoch": 5.16, + "learning_rate": 1.000596719962923e-06, + "loss": 1.0523, + "step": 69887 + }, + { + "epoch": 5.16, + "learning_rate": 1.0004228138783146e-06, + "loss": 1.0788, + "step": 69888 + }, + { + "epoch": 5.16, + "learning_rate": 1.000248922111906e-06, + "loss": 0.9372, + "step": 69889 + }, + { + "epoch": 5.16, + "learning_rate": 1.0000750446639718e-06, + "loss": 1.0572, + "step": 69890 + }, + { + "epoch": 5.16, + "learning_rate": 9.999011815347892e-07, + "loss": 1.0851, + "step": 69891 + }, + { + "epoch": 5.16, + "learning_rate": 9.997273327246337e-07, + "loss": 0.9369, + "step": 69892 + }, + { + "epoch": 5.16, + "learning_rate": 9.995534982337851e-07, + "loss": 0.9377, + "step": 69893 + }, + { + "epoch": 5.16, + "learning_rate": 9.993796780625175e-07, + "loss": 0.938, + "step": 69894 + }, + { + "epoch": 5.16, + "learning_rate": 9.992058722111086e-07, + "loss": 0.9343, + "step": 69895 + }, + { + "epoch": 5.16, + "learning_rate": 9.990320806798327e-07, + "loss": 1.1074, + "step": 69896 + }, + { + "epoch": 5.16, + "learning_rate": 9.988583034689692e-07, + "loss": 1.0147, + "step": 69897 + }, + { + "epoch": 5.16, + "learning_rate": 9.98684540578796e-07, + "loss": 1.0857, + "step": 69898 + }, + { + "epoch": 5.16, + "learning_rate": 9.985107920095817e-07, + "loss": 1.0335, + "step": 69899 + }, + { + "epoch": 5.16, + "learning_rate": 9.983370577616115e-07, + "loss": 0.9703, + "step": 69900 + }, + { + "epoch": 5.16, + "learning_rate": 9.981633378351551e-07, + "loss": 1.058, + "step": 69901 + }, + { + "epoch": 5.17, + "learning_rate": 9.979896322304972e-07, + "loss": 0.9608, + "step": 69902 + }, + { + "epoch": 5.17, + "learning_rate": 9.978159409479038e-07, + "loss": 0.9456, + "step": 69903 + }, + { + "epoch": 5.17, + "learning_rate": 9.976422639876594e-07, + "loss": 0.9748, + "step": 69904 + }, + { + "epoch": 5.17, + "learning_rate": 9.974686013500346e-07, + "loss": 0.9365, + "step": 69905 + }, + { + "epoch": 5.17, + "learning_rate": 9.972949530353126e-07, + "loss": 0.9923, + "step": 69906 + }, + { + "epoch": 5.17, + "learning_rate": 9.971213190437633e-07, + "loss": 1.0749, + "step": 69907 + }, + { + "epoch": 5.17, + "learning_rate": 9.969476993756645e-07, + "loss": 1.0239, + "step": 69908 + }, + { + "epoch": 5.17, + "learning_rate": 9.967740940312942e-07, + "loss": 0.8953, + "step": 69909 + }, + { + "epoch": 5.17, + "learning_rate": 9.96600503010927e-07, + "loss": 0.9573, + "step": 69910 + }, + { + "epoch": 5.17, + "learning_rate": 9.964269263148396e-07, + "loss": 1.0371, + "step": 69911 + }, + { + "epoch": 5.17, + "learning_rate": 9.96253363943307e-07, + "loss": 1.113, + "step": 69912 + }, + { + "epoch": 5.17, + "learning_rate": 9.96079815896608e-07, + "loss": 0.9534, + "step": 69913 + }, + { + "epoch": 5.17, + "learning_rate": 9.959062821750175e-07, + "loss": 1.0996, + "step": 69914 + }, + { + "epoch": 5.17, + "learning_rate": 9.957327627788105e-07, + "loss": 0.958, + "step": 69915 + }, + { + "epoch": 5.17, + "learning_rate": 9.955592577082652e-07, + "loss": 1.0807, + "step": 69916 + }, + { + "epoch": 5.17, + "learning_rate": 9.953857669636536e-07, + "loss": 1.081, + "step": 69917 + }, + { + "epoch": 5.17, + "learning_rate": 9.952122905452556e-07, + "loss": 0.8913, + "step": 69918 + }, + { + "epoch": 5.17, + "learning_rate": 9.950388284533463e-07, + "loss": 1.0475, + "step": 69919 + }, + { + "epoch": 5.17, + "learning_rate": 9.948653806882024e-07, + "loss": 1.0179, + "step": 69920 + }, + { + "epoch": 5.17, + "learning_rate": 9.946919472500959e-07, + "loss": 1.0022, + "step": 69921 + }, + { + "epoch": 5.17, + "learning_rate": 9.945185281393077e-07, + "loss": 0.9654, + "step": 69922 + }, + { + "epoch": 5.17, + "learning_rate": 9.94345123356113e-07, + "loss": 1.0448, + "step": 69923 + }, + { + "epoch": 5.17, + "learning_rate": 9.94171732900785e-07, + "loss": 1.0491, + "step": 69924 + }, + { + "epoch": 5.17, + "learning_rate": 9.939983567736012e-07, + "loss": 1.0885, + "step": 69925 + }, + { + "epoch": 5.17, + "learning_rate": 9.938249949748347e-07, + "loss": 1.0232, + "step": 69926 + }, + { + "epoch": 5.17, + "learning_rate": 9.936516475047687e-07, + "loss": 0.9352, + "step": 69927 + }, + { + "epoch": 5.17, + "learning_rate": 9.934783143636696e-07, + "loss": 0.9049, + "step": 69928 + }, + { + "epoch": 5.17, + "learning_rate": 9.933049955518192e-07, + "loss": 0.9905, + "step": 69929 + }, + { + "epoch": 5.17, + "learning_rate": 9.93131691069491e-07, + "loss": 0.9218, + "step": 69930 + }, + { + "epoch": 5.17, + "learning_rate": 9.929584009169635e-07, + "loss": 0.9878, + "step": 69931 + }, + { + "epoch": 5.17, + "learning_rate": 9.927851250945075e-07, + "loss": 0.9521, + "step": 69932 + }, + { + "epoch": 5.17, + "learning_rate": 9.92611863602404e-07, + "loss": 1.1712, + "step": 69933 + }, + { + "epoch": 5.17, + "learning_rate": 9.92438616440925e-07, + "loss": 0.9183, + "step": 69934 + }, + { + "epoch": 5.17, + "learning_rate": 9.92265383610348e-07, + "loss": 0.953, + "step": 69935 + }, + { + "epoch": 5.17, + "learning_rate": 9.920921651109482e-07, + "loss": 1.0103, + "step": 69936 + }, + { + "epoch": 5.17, + "learning_rate": 9.91918960942998e-07, + "loss": 0.9822, + "step": 69937 + }, + { + "epoch": 5.17, + "learning_rate": 9.91745771106778e-07, + "loss": 0.9229, + "step": 69938 + }, + { + "epoch": 5.17, + "learning_rate": 9.915725956025623e-07, + "loss": 0.9605, + "step": 69939 + }, + { + "epoch": 5.17, + "learning_rate": 9.913994344306254e-07, + "loss": 1.1003, + "step": 69940 + }, + { + "epoch": 5.17, + "learning_rate": 9.912262875912403e-07, + "loss": 1.0138, + "step": 69941 + }, + { + "epoch": 5.17, + "learning_rate": 9.91053155084688e-07, + "loss": 1.0586, + "step": 69942 + }, + { + "epoch": 5.17, + "learning_rate": 9.908800369112416e-07, + "loss": 1.0799, + "step": 69943 + }, + { + "epoch": 5.17, + "learning_rate": 9.907069330711761e-07, + "loss": 0.9232, + "step": 69944 + }, + { + "epoch": 5.17, + "learning_rate": 9.90533843564766e-07, + "loss": 0.9686, + "step": 69945 + }, + { + "epoch": 5.17, + "learning_rate": 9.903607683922866e-07, + "loss": 0.9973, + "step": 69946 + }, + { + "epoch": 5.17, + "learning_rate": 9.901877075540178e-07, + "loss": 1.1029, + "step": 69947 + }, + { + "epoch": 5.17, + "learning_rate": 9.900146610502271e-07, + "loss": 0.9575, + "step": 69948 + }, + { + "epoch": 5.17, + "learning_rate": 9.898416288811974e-07, + "loss": 1.0777, + "step": 69949 + }, + { + "epoch": 5.17, + "learning_rate": 9.896686110471975e-07, + "loss": 1.0512, + "step": 69950 + }, + { + "epoch": 5.17, + "learning_rate": 9.894956075485106e-07, + "loss": 0.9425, + "step": 69951 + }, + { + "epoch": 5.17, + "learning_rate": 9.893226183854033e-07, + "loss": 0.9681, + "step": 69952 + }, + { + "epoch": 5.17, + "learning_rate": 9.89149643558156e-07, + "loss": 1.0361, + "step": 69953 + }, + { + "epoch": 5.17, + "learning_rate": 9.889766830670433e-07, + "loss": 0.8844, + "step": 69954 + }, + { + "epoch": 5.17, + "learning_rate": 9.888037369123404e-07, + "loss": 0.9276, + "step": 69955 + }, + { + "epoch": 5.17, + "learning_rate": 9.886308050943205e-07, + "loss": 0.9271, + "step": 69956 + }, + { + "epoch": 5.17, + "learning_rate": 9.884578876132588e-07, + "loss": 1.1049, + "step": 69957 + }, + { + "epoch": 5.17, + "learning_rate": 9.882849844694342e-07, + "loss": 1.0644, + "step": 69958 + }, + { + "epoch": 5.17, + "learning_rate": 9.881120956631184e-07, + "loss": 0.9822, + "step": 69959 + }, + { + "epoch": 5.17, + "learning_rate": 9.879392211945882e-07, + "loss": 0.9878, + "step": 69960 + }, + { + "epoch": 5.17, + "learning_rate": 9.877663610641142e-07, + "loss": 1.0352, + "step": 69961 + }, + { + "epoch": 5.17, + "learning_rate": 9.875935152719784e-07, + "loss": 0.9938, + "step": 69962 + }, + { + "epoch": 5.17, + "learning_rate": 9.87420683818452e-07, + "loss": 1.0143, + "step": 69963 + }, + { + "epoch": 5.17, + "learning_rate": 9.872478667038099e-07, + "loss": 1.0243, + "step": 69964 + }, + { + "epoch": 5.17, + "learning_rate": 9.870750639283277e-07, + "loss": 1.0589, + "step": 69965 + }, + { + "epoch": 5.17, + "learning_rate": 9.869022754922785e-07, + "loss": 0.9377, + "step": 69966 + }, + { + "epoch": 5.17, + "learning_rate": 9.86729501395941e-07, + "loss": 0.9763, + "step": 69967 + }, + { + "epoch": 5.17, + "learning_rate": 9.86556741639587e-07, + "loss": 0.9762, + "step": 69968 + }, + { + "epoch": 5.17, + "learning_rate": 9.863839962234934e-07, + "loss": 1.1054, + "step": 69969 + }, + { + "epoch": 5.17, + "learning_rate": 9.862112651479306e-07, + "loss": 0.9986, + "step": 69970 + }, + { + "epoch": 5.17, + "learning_rate": 9.860385484131795e-07, + "loss": 1.0065, + "step": 69971 + }, + { + "epoch": 5.17, + "learning_rate": 9.858658460195137e-07, + "loss": 0.9082, + "step": 69972 + }, + { + "epoch": 5.17, + "learning_rate": 9.856931579672024e-07, + "loss": 0.9966, + "step": 69973 + }, + { + "epoch": 5.17, + "learning_rate": 9.85520484256527e-07, + "loss": 1.099, + "step": 69974 + }, + { + "epoch": 5.17, + "learning_rate": 9.85347824887759e-07, + "loss": 1.0013, + "step": 69975 + }, + { + "epoch": 5.17, + "learning_rate": 9.851751798611741e-07, + "loss": 1.0346, + "step": 69976 + }, + { + "epoch": 5.17, + "learning_rate": 9.850025491770431e-07, + "loss": 0.8909, + "step": 69977 + }, + { + "epoch": 5.17, + "learning_rate": 9.84829932835648e-07, + "loss": 0.9034, + "step": 69978 + }, + { + "epoch": 5.17, + "learning_rate": 9.846573308372587e-07, + "loss": 0.9549, + "step": 69979 + }, + { + "epoch": 5.17, + "learning_rate": 9.844847431821503e-07, + "loss": 0.9912, + "step": 69980 + }, + { + "epoch": 5.17, + "learning_rate": 9.84312169870596e-07, + "loss": 0.9568, + "step": 69981 + }, + { + "epoch": 5.17, + "learning_rate": 9.841396109028733e-07, + "loss": 0.9633, + "step": 69982 + }, + { + "epoch": 5.17, + "learning_rate": 9.839670662792566e-07, + "loss": 1.0427, + "step": 69983 + }, + { + "epoch": 5.17, + "learning_rate": 9.837945360000179e-07, + "loss": 0.8094, + "step": 69984 + }, + { + "epoch": 5.17, + "learning_rate": 9.836220200654346e-07, + "loss": 0.9821, + "step": 69985 + }, + { + "epoch": 5.17, + "learning_rate": 9.834495184757763e-07, + "loss": 1.107, + "step": 69986 + }, + { + "epoch": 5.17, + "learning_rate": 9.832770312313245e-07, + "loss": 1.0191, + "step": 69987 + }, + { + "epoch": 5.17, + "learning_rate": 9.831045583323484e-07, + "loss": 1.1172, + "step": 69988 + }, + { + "epoch": 5.17, + "learning_rate": 9.829320997791246e-07, + "loss": 0.9603, + "step": 69989 + }, + { + "epoch": 5.17, + "learning_rate": 9.827596555719243e-07, + "loss": 0.9591, + "step": 69990 + }, + { + "epoch": 5.17, + "learning_rate": 9.82587225711027e-07, + "loss": 0.8893, + "step": 69991 + }, + { + "epoch": 5.17, + "learning_rate": 9.824148101967067e-07, + "loss": 1.0034, + "step": 69992 + }, + { + "epoch": 5.17, + "learning_rate": 9.822424090292304e-07, + "loss": 1.1217, + "step": 69993 + }, + { + "epoch": 5.17, + "learning_rate": 9.820700222088798e-07, + "loss": 1.0337, + "step": 69994 + }, + { + "epoch": 5.17, + "learning_rate": 9.818976497359256e-07, + "loss": 1.1079, + "step": 69995 + }, + { + "epoch": 5.17, + "learning_rate": 9.817252916106458e-07, + "loss": 1.0507, + "step": 69996 + }, + { + "epoch": 5.17, + "learning_rate": 9.815529478333086e-07, + "loss": 0.9441, + "step": 69997 + }, + { + "epoch": 5.17, + "learning_rate": 9.81380618404194e-07, + "loss": 1.0623, + "step": 69998 + }, + { + "epoch": 5.17, + "learning_rate": 9.81208303323572e-07, + "loss": 1.0428, + "step": 69999 + }, + { + "epoch": 5.17, + "learning_rate": 9.810360025917221e-07, + "loss": 0.984, + "step": 70000 + }, + { + "epoch": 5.17, + "learning_rate": 9.80863716208913e-07, + "loss": 1.0677, + "step": 70001 + }, + { + "epoch": 5.17, + "learning_rate": 9.806914441754178e-07, + "loss": 1.0766, + "step": 70002 + }, + { + "epoch": 5.17, + "learning_rate": 9.805191864915164e-07, + "loss": 0.8372, + "step": 70003 + }, + { + "epoch": 5.17, + "learning_rate": 9.803469431574796e-07, + "loss": 1.0128, + "step": 70004 + }, + { + "epoch": 5.17, + "learning_rate": 9.801747141735817e-07, + "loss": 1.0411, + "step": 70005 + }, + { + "epoch": 5.17, + "learning_rate": 9.80002499540096e-07, + "loss": 1.0089, + "step": 70006 + }, + { + "epoch": 5.17, + "learning_rate": 9.798302992572984e-07, + "loss": 0.9988, + "step": 70007 + }, + { + "epoch": 5.17, + "learning_rate": 9.796581133254613e-07, + "loss": 1.0105, + "step": 70008 + }, + { + "epoch": 5.17, + "learning_rate": 9.794859417448588e-07, + "loss": 0.9488, + "step": 70009 + }, + { + "epoch": 5.17, + "learning_rate": 9.793137845157663e-07, + "loss": 1.0662, + "step": 70010 + }, + { + "epoch": 5.17, + "learning_rate": 9.791416416384537e-07, + "loss": 0.8876, + "step": 70011 + }, + { + "epoch": 5.17, + "learning_rate": 9.789695131132004e-07, + "loss": 1.0407, + "step": 70012 + }, + { + "epoch": 5.17, + "learning_rate": 9.787973989402766e-07, + "loss": 0.9598, + "step": 70013 + }, + { + "epoch": 5.17, + "learning_rate": 9.786252991199575e-07, + "loss": 1.0351, + "step": 70014 + }, + { + "epoch": 5.17, + "learning_rate": 9.784532136525148e-07, + "loss": 0.9593, + "step": 70015 + }, + { + "epoch": 5.17, + "learning_rate": 9.782811425382255e-07, + "loss": 0.9346, + "step": 70016 + }, + { + "epoch": 5.17, + "learning_rate": 9.781090857773611e-07, + "loss": 1.0732, + "step": 70017 + }, + { + "epoch": 5.17, + "learning_rate": 9.779370433701974e-07, + "loss": 1.0021, + "step": 70018 + }, + { + "epoch": 5.17, + "learning_rate": 9.77765015317006e-07, + "loss": 1.1168, + "step": 70019 + }, + { + "epoch": 5.17, + "learning_rate": 9.775930016180625e-07, + "loss": 1.11, + "step": 70020 + }, + { + "epoch": 5.17, + "learning_rate": 9.774210022736386e-07, + "loss": 0.99, + "step": 70021 + }, + { + "epoch": 5.17, + "learning_rate": 9.772490172840065e-07, + "loss": 1.0262, + "step": 70022 + }, + { + "epoch": 5.17, + "learning_rate": 9.77077046649444e-07, + "loss": 1.0256, + "step": 70023 + }, + { + "epoch": 5.17, + "learning_rate": 9.769050903702237e-07, + "loss": 1.0164, + "step": 70024 + }, + { + "epoch": 5.17, + "learning_rate": 9.76733148446618e-07, + "loss": 0.9676, + "step": 70025 + }, + { + "epoch": 5.17, + "learning_rate": 9.765612208788977e-07, + "loss": 1.0512, + "step": 70026 + }, + { + "epoch": 5.17, + "learning_rate": 9.763893076673425e-07, + "loss": 1.0257, + "step": 70027 + }, + { + "epoch": 5.17, + "learning_rate": 9.762174088122234e-07, + "loss": 0.9101, + "step": 70028 + }, + { + "epoch": 5.17, + "learning_rate": 9.760455243138122e-07, + "loss": 1.0251, + "step": 70029 + }, + { + "epoch": 5.17, + "learning_rate": 9.758736541723835e-07, + "loss": 1.0474, + "step": 70030 + }, + { + "epoch": 5.17, + "learning_rate": 9.75701798388209e-07, + "loss": 0.9606, + "step": 70031 + }, + { + "epoch": 5.17, + "learning_rate": 9.755299569615663e-07, + "loss": 1.0442, + "step": 70032 + }, + { + "epoch": 5.17, + "learning_rate": 9.753581298927251e-07, + "loss": 1.0921, + "step": 70033 + }, + { + "epoch": 5.17, + "learning_rate": 9.75186317181961e-07, + "loss": 0.9526, + "step": 70034 + }, + { + "epoch": 5.17, + "learning_rate": 9.750145188295435e-07, + "loss": 0.9227, + "step": 70035 + }, + { + "epoch": 5.17, + "learning_rate": 9.748427348357515e-07, + "loss": 1.0728, + "step": 70036 + }, + { + "epoch": 5.17, + "learning_rate": 9.746709652008556e-07, + "loss": 1.0126, + "step": 70037 + }, + { + "epoch": 5.18, + "learning_rate": 9.744992099251283e-07, + "loss": 1.0708, + "step": 70038 + }, + { + "epoch": 5.18, + "learning_rate": 9.743274690088444e-07, + "loss": 1.0355, + "step": 70039 + }, + { + "epoch": 5.18, + "learning_rate": 9.74155742452274e-07, + "loss": 1.0017, + "step": 70040 + }, + { + "epoch": 5.18, + "learning_rate": 9.739840302556968e-07, + "loss": 1.0458, + "step": 70041 + }, + { + "epoch": 5.18, + "learning_rate": 9.738123324193782e-07, + "loss": 1.0136, + "step": 70042 + }, + { + "epoch": 5.18, + "learning_rate": 9.736406489435968e-07, + "loss": 0.9512, + "step": 70043 + }, + { + "epoch": 5.18, + "learning_rate": 9.734689798286224e-07, + "loss": 1.018, + "step": 70044 + }, + { + "epoch": 5.18, + "learning_rate": 9.732973250747324e-07, + "loss": 0.9215, + "step": 70045 + }, + { + "epoch": 5.18, + "learning_rate": 9.731256846821947e-07, + "loss": 0.8673, + "step": 70046 + }, + { + "epoch": 5.18, + "learning_rate": 9.729540586512854e-07, + "loss": 1.0008, + "step": 70047 + }, + { + "epoch": 5.18, + "learning_rate": 9.727824469822777e-07, + "loss": 0.9858, + "step": 70048 + }, + { + "epoch": 5.18, + "learning_rate": 9.72610849675445e-07, + "loss": 1.0466, + "step": 70049 + }, + { + "epoch": 5.18, + "learning_rate": 9.724392667310578e-07, + "loss": 0.9393, + "step": 70050 + }, + { + "epoch": 5.18, + "learning_rate": 9.722676981493894e-07, + "loss": 1.0073, + "step": 70051 + }, + { + "epoch": 5.18, + "learning_rate": 9.720961439307152e-07, + "loss": 0.887, + "step": 70052 + }, + { + "epoch": 5.18, + "learning_rate": 9.719246040753072e-07, + "loss": 1.0045, + "step": 70053 + }, + { + "epoch": 5.18, + "learning_rate": 9.717530785834384e-07, + "loss": 1.0196, + "step": 70054 + }, + { + "epoch": 5.18, + "learning_rate": 9.715815674553796e-07, + "loss": 0.9577, + "step": 70055 + }, + { + "epoch": 5.18, + "learning_rate": 9.714100706914066e-07, + "loss": 0.8573, + "step": 70056 + }, + { + "epoch": 5.18, + "learning_rate": 9.712385882917907e-07, + "loss": 1.015, + "step": 70057 + }, + { + "epoch": 5.18, + "learning_rate": 9.710671202568056e-07, + "loss": 1.0202, + "step": 70058 + }, + { + "epoch": 5.18, + "learning_rate": 9.70895666586723e-07, + "loss": 1.1123, + "step": 70059 + }, + { + "epoch": 5.18, + "learning_rate": 9.70724227281815e-07, + "loss": 1.0815, + "step": 70060 + }, + { + "epoch": 5.18, + "learning_rate": 9.705528023423571e-07, + "loss": 0.8865, + "step": 70061 + }, + { + "epoch": 5.18, + "learning_rate": 9.703813917686212e-07, + "loss": 0.9225, + "step": 70062 + }, + { + "epoch": 5.18, + "learning_rate": 9.702099955608779e-07, + "loss": 0.9961, + "step": 70063 + }, + { + "epoch": 5.18, + "learning_rate": 9.700386137194008e-07, + "loss": 0.9892, + "step": 70064 + }, + { + "epoch": 5.18, + "learning_rate": 9.698672462444658e-07, + "loss": 0.9859, + "step": 70065 + }, + { + "epoch": 5.18, + "learning_rate": 9.69695893136341e-07, + "loss": 1.0355, + "step": 70066 + }, + { + "epoch": 5.18, + "learning_rate": 9.695245543952992e-07, + "loss": 1.0209, + "step": 70067 + }, + { + "epoch": 5.18, + "learning_rate": 9.693532300216169e-07, + "loss": 1.0221, + "step": 70068 + }, + { + "epoch": 5.18, + "learning_rate": 9.691819200155638e-07, + "loss": 0.9239, + "step": 70069 + }, + { + "epoch": 5.18, + "learning_rate": 9.690106243774133e-07, + "loss": 1.0354, + "step": 70070 + }, + { + "epoch": 5.18, + "learning_rate": 9.68839343107436e-07, + "loss": 1.0625, + "step": 70071 + }, + { + "epoch": 5.18, + "learning_rate": 9.686680762059086e-07, + "loss": 1.0021, + "step": 70072 + }, + { + "epoch": 5.18, + "learning_rate": 9.684968236730995e-07, + "loss": 1.0731, + "step": 70073 + }, + { + "epoch": 5.18, + "learning_rate": 9.683255855092843e-07, + "loss": 1.0055, + "step": 70074 + }, + { + "epoch": 5.18, + "learning_rate": 9.681543617147326e-07, + "loss": 0.8965, + "step": 70075 + }, + { + "epoch": 5.18, + "learning_rate": 9.679831522897166e-07, + "loss": 1.0395, + "step": 70076 + }, + { + "epoch": 5.18, + "learning_rate": 9.678119572345124e-07, + "loss": 0.9473, + "step": 70077 + }, + { + "epoch": 5.18, + "learning_rate": 9.6764077654939e-07, + "loss": 0.9635, + "step": 70078 + }, + { + "epoch": 5.18, + "learning_rate": 9.674696102346215e-07, + "loss": 0.9512, + "step": 70079 + }, + { + "epoch": 5.18, + "learning_rate": 9.672984582904787e-07, + "loss": 0.8986, + "step": 70080 + }, + { + "epoch": 5.18, + "learning_rate": 9.67127320717236e-07, + "loss": 1.0121, + "step": 70081 + }, + { + "epoch": 5.18, + "learning_rate": 9.669561975151642e-07, + "loss": 0.9635, + "step": 70082 + }, + { + "epoch": 5.18, + "learning_rate": 9.667850886845366e-07, + "loss": 0.9596, + "step": 70083 + }, + { + "epoch": 5.18, + "learning_rate": 9.666139942256237e-07, + "loss": 1.0349, + "step": 70084 + }, + { + "epoch": 5.18, + "learning_rate": 9.664429141386978e-07, + "loss": 0.9944, + "step": 70085 + }, + { + "epoch": 5.18, + "learning_rate": 9.662718484240352e-07, + "loss": 1.0258, + "step": 70086 + }, + { + "epoch": 5.18, + "learning_rate": 9.661007970819015e-07, + "loss": 0.9693, + "step": 70087 + }, + { + "epoch": 5.18, + "learning_rate": 9.659297601125727e-07, + "loss": 1.0938, + "step": 70088 + }, + { + "epoch": 5.18, + "learning_rate": 9.657587375163202e-07, + "loss": 0.9726, + "step": 70089 + }, + { + "epoch": 5.18, + "learning_rate": 9.65587729293419e-07, + "loss": 1.1209, + "step": 70090 + }, + { + "epoch": 5.18, + "learning_rate": 9.654167354441357e-07, + "loss": 1.0194, + "step": 70091 + }, + { + "epoch": 5.18, + "learning_rate": 9.652457559687467e-07, + "loss": 1.0627, + "step": 70092 + }, + { + "epoch": 5.18, + "learning_rate": 9.650747908675196e-07, + "loss": 1.0227, + "step": 70093 + }, + { + "epoch": 5.18, + "learning_rate": 9.64903840140734e-07, + "loss": 0.8765, + "step": 70094 + }, + { + "epoch": 5.18, + "learning_rate": 9.647329037886545e-07, + "loss": 1.0401, + "step": 70095 + }, + { + "epoch": 5.18, + "learning_rate": 9.645619818115549e-07, + "loss": 1.0476, + "step": 70096 + }, + { + "epoch": 5.18, + "learning_rate": 9.643910742097085e-07, + "loss": 0.9648, + "step": 70097 + }, + { + "epoch": 5.18, + "learning_rate": 9.642201809833874e-07, + "loss": 1.0415, + "step": 70098 + }, + { + "epoch": 5.18, + "learning_rate": 9.640493021328623e-07, + "loss": 0.989, + "step": 70099 + }, + { + "epoch": 5.18, + "learning_rate": 9.638784376584042e-07, + "loss": 1.0174, + "step": 70100 + }, + { + "epoch": 5.18, + "learning_rate": 9.637075875602874e-07, + "loss": 1.0345, + "step": 70101 + }, + { + "epoch": 5.18, + "learning_rate": 9.635367518387828e-07, + "loss": 1.0066, + "step": 70102 + }, + { + "epoch": 5.18, + "learning_rate": 9.633659304941622e-07, + "loss": 1.0221, + "step": 70103 + }, + { + "epoch": 5.18, + "learning_rate": 9.631951235266967e-07, + "loss": 0.998, + "step": 70104 + }, + { + "epoch": 5.18, + "learning_rate": 9.630243309366572e-07, + "loss": 1.0568, + "step": 70105 + }, + { + "epoch": 5.18, + "learning_rate": 9.628535527243178e-07, + "loss": 0.9724, + "step": 70106 + }, + { + "epoch": 5.18, + "learning_rate": 9.626827888899493e-07, + "loss": 0.9565, + "step": 70107 + }, + { + "epoch": 5.18, + "learning_rate": 9.625120394338239e-07, + "loss": 0.9441, + "step": 70108 + }, + { + "epoch": 5.18, + "learning_rate": 9.623413043562102e-07, + "loss": 1.0258, + "step": 70109 + }, + { + "epoch": 5.18, + "learning_rate": 9.621705836573859e-07, + "loss": 0.9328, + "step": 70110 + }, + { + "epoch": 5.18, + "learning_rate": 9.61999877337615e-07, + "loss": 1.0225, + "step": 70111 + }, + { + "epoch": 5.18, + "learning_rate": 9.618291853971752e-07, + "loss": 1.0108, + "step": 70112 + }, + { + "epoch": 5.18, + "learning_rate": 9.61658507836335e-07, + "loss": 0.9518, + "step": 70113 + }, + { + "epoch": 5.18, + "learning_rate": 9.614878446553678e-07, + "loss": 0.9248, + "step": 70114 + }, + { + "epoch": 5.18, + "learning_rate": 9.613171958545432e-07, + "loss": 0.931, + "step": 70115 + }, + { + "epoch": 5.18, + "learning_rate": 9.61146561434132e-07, + "loss": 0.978, + "step": 70116 + }, + { + "epoch": 5.18, + "learning_rate": 9.609759413944097e-07, + "loss": 1.0016, + "step": 70117 + }, + { + "epoch": 5.18, + "learning_rate": 9.60805335735645e-07, + "loss": 1.1582, + "step": 70118 + }, + { + "epoch": 5.18, + "learning_rate": 9.6063474445811e-07, + "loss": 1.089, + "step": 70119 + }, + { + "epoch": 5.18, + "learning_rate": 9.60464167562073e-07, + "loss": 0.9754, + "step": 70120 + }, + { + "epoch": 5.18, + "learning_rate": 9.602936050478107e-07, + "loss": 1.0409, + "step": 70121 + }, + { + "epoch": 5.18, + "learning_rate": 9.601230569155928e-07, + "loss": 1.0055, + "step": 70122 + }, + { + "epoch": 5.18, + "learning_rate": 9.59952523165688e-07, + "loss": 0.9808, + "step": 70123 + }, + { + "epoch": 5.18, + "learning_rate": 9.597820037983706e-07, + "loss": 0.8708, + "step": 70124 + }, + { + "epoch": 5.18, + "learning_rate": 9.596114988139083e-07, + "loss": 1.0825, + "step": 70125 + }, + { + "epoch": 5.18, + "learning_rate": 9.59441008212577e-07, + "loss": 1.0036, + "step": 70126 + }, + { + "epoch": 5.18, + "learning_rate": 9.59270531994646e-07, + "loss": 1.0179, + "step": 70127 + }, + { + "epoch": 5.18, + "learning_rate": 9.591000701603858e-07, + "loss": 1.0682, + "step": 70128 + }, + { + "epoch": 5.18, + "learning_rate": 9.589296227100664e-07, + "loss": 1.0229, + "step": 70129 + }, + { + "epoch": 5.18, + "learning_rate": 9.587591896439618e-07, + "loss": 0.8998, + "step": 70130 + }, + { + "epoch": 5.18, + "learning_rate": 9.585887709623444e-07, + "loss": 1.1015, + "step": 70131 + }, + { + "epoch": 5.18, + "learning_rate": 9.58418366665479e-07, + "loss": 0.93, + "step": 70132 + }, + { + "epoch": 5.18, + "learning_rate": 9.582479767536424e-07, + "loss": 1.0706, + "step": 70133 + }, + { + "epoch": 5.18, + "learning_rate": 9.58077601227102e-07, + "loss": 1.014, + "step": 70134 + }, + { + "epoch": 5.18, + "learning_rate": 9.579072400861356e-07, + "loss": 0.9443, + "step": 70135 + }, + { + "epoch": 5.18, + "learning_rate": 9.577368933310038e-07, + "loss": 0.9034, + "step": 70136 + }, + { + "epoch": 5.18, + "learning_rate": 9.575665609619867e-07, + "loss": 1.1421, + "step": 70137 + }, + { + "epoch": 5.18, + "learning_rate": 9.573962429793493e-07, + "loss": 1.0975, + "step": 70138 + }, + { + "epoch": 5.18, + "learning_rate": 9.572259393833683e-07, + "loss": 0.9617, + "step": 70139 + }, + { + "epoch": 5.18, + "learning_rate": 9.5705565017431e-07, + "loss": 0.9749, + "step": 70140 + }, + { + "epoch": 5.18, + "learning_rate": 9.568853753524454e-07, + "loss": 1.0172, + "step": 70141 + }, + { + "epoch": 5.18, + "learning_rate": 9.567151149180487e-07, + "loss": 1.0214, + "step": 70142 + }, + { + "epoch": 5.18, + "learning_rate": 9.565448688713885e-07, + "loss": 1.005, + "step": 70143 + }, + { + "epoch": 5.18, + "learning_rate": 9.563746372127358e-07, + "loss": 1.021, + "step": 70144 + }, + { + "epoch": 5.18, + "learning_rate": 9.562044199423592e-07, + "loss": 1.0424, + "step": 70145 + }, + { + "epoch": 5.18, + "learning_rate": 9.560342170605353e-07, + "loss": 0.9674, + "step": 70146 + }, + { + "epoch": 5.18, + "learning_rate": 9.558640285675303e-07, + "loss": 0.8882, + "step": 70147 + }, + { + "epoch": 5.18, + "learning_rate": 9.556938544636173e-07, + "loss": 0.9501, + "step": 70148 + }, + { + "epoch": 5.18, + "learning_rate": 9.555236947490632e-07, + "loss": 0.8519, + "step": 70149 + }, + { + "epoch": 5.18, + "learning_rate": 9.553535494241439e-07, + "loss": 1.1331, + "step": 70150 + }, + { + "epoch": 5.18, + "learning_rate": 9.551834184891274e-07, + "loss": 1.0036, + "step": 70151 + }, + { + "epoch": 5.18, + "learning_rate": 9.550133019442853e-07, + "loss": 1.015, + "step": 70152 + }, + { + "epoch": 5.18, + "learning_rate": 9.548431997898867e-07, + "loss": 0.901, + "step": 70153 + }, + { + "epoch": 5.18, + "learning_rate": 9.546731120262021e-07, + "loss": 1.0399, + "step": 70154 + }, + { + "epoch": 5.18, + "learning_rate": 9.54503038653506e-07, + "loss": 1.0741, + "step": 70155 + }, + { + "epoch": 5.18, + "learning_rate": 9.543329796720635e-07, + "loss": 0.945, + "step": 70156 + }, + { + "epoch": 5.18, + "learning_rate": 9.54162935082149e-07, + "loss": 0.9523, + "step": 70157 + }, + { + "epoch": 5.18, + "learning_rate": 9.539929048840303e-07, + "loss": 0.9048, + "step": 70158 + }, + { + "epoch": 5.18, + "learning_rate": 9.538228890779834e-07, + "loss": 1.016, + "step": 70159 + }, + { + "epoch": 5.18, + "learning_rate": 9.536528876642715e-07, + "loss": 1.0156, + "step": 70160 + }, + { + "epoch": 5.18, + "learning_rate": 9.534829006431679e-07, + "loss": 0.9957, + "step": 70161 + }, + { + "epoch": 5.18, + "learning_rate": 9.533129280149456e-07, + "loss": 0.9336, + "step": 70162 + }, + { + "epoch": 5.18, + "learning_rate": 9.531429697798722e-07, + "loss": 0.9521, + "step": 70163 + }, + { + "epoch": 5.18, + "learning_rate": 9.529730259382198e-07, + "loss": 0.9968, + "step": 70164 + }, + { + "epoch": 5.18, + "learning_rate": 9.528030964902546e-07, + "loss": 1.0844, + "step": 70165 + }, + { + "epoch": 5.18, + "learning_rate": 9.526331814362533e-07, + "loss": 0.9906, + "step": 70166 + }, + { + "epoch": 5.18, + "learning_rate": 9.524632807764833e-07, + "loss": 1.0119, + "step": 70167 + }, + { + "epoch": 5.18, + "learning_rate": 9.522933945112145e-07, + "loss": 1.0898, + "step": 70168 + }, + { + "epoch": 5.18, + "learning_rate": 9.521235226407166e-07, + "loss": 0.9486, + "step": 70169 + }, + { + "epoch": 5.18, + "learning_rate": 9.519536651652594e-07, + "loss": 1.0921, + "step": 70170 + }, + { + "epoch": 5.18, + "learning_rate": 9.517838220851172e-07, + "loss": 0.9484, + "step": 70171 + }, + { + "epoch": 5.18, + "learning_rate": 9.516139934005563e-07, + "loss": 0.9769, + "step": 70172 + }, + { + "epoch": 5.19, + "learning_rate": 9.51444179111849e-07, + "loss": 0.8368, + "step": 70173 + }, + { + "epoch": 5.19, + "learning_rate": 9.512743792192614e-07, + "loss": 1.0943, + "step": 70174 + }, + { + "epoch": 5.19, + "learning_rate": 9.5110459372307e-07, + "loss": 1.0029, + "step": 70175 + }, + { + "epoch": 5.19, + "learning_rate": 9.509348226235404e-07, + "loss": 1.0252, + "step": 70176 + }, + { + "epoch": 5.19, + "learning_rate": 9.507650659209455e-07, + "loss": 1.0976, + "step": 70177 + }, + { + "epoch": 5.19, + "learning_rate": 9.50595323615553e-07, + "loss": 1.0355, + "step": 70178 + }, + { + "epoch": 5.19, + "learning_rate": 9.504255957076325e-07, + "loss": 1.1838, + "step": 70179 + }, + { + "epoch": 5.19, + "learning_rate": 9.502558821974584e-07, + "loss": 0.9959, + "step": 70180 + }, + { + "epoch": 5.19, + "learning_rate": 9.500861830852948e-07, + "loss": 0.9706, + "step": 70181 + }, + { + "epoch": 5.19, + "learning_rate": 9.499164983714159e-07, + "loss": 1.0336, + "step": 70182 + }, + { + "epoch": 5.19, + "learning_rate": 9.497468280560884e-07, + "loss": 0.906, + "step": 70183 + }, + { + "epoch": 5.19, + "learning_rate": 9.495771721395886e-07, + "loss": 1.0978, + "step": 70184 + }, + { + "epoch": 5.19, + "learning_rate": 9.494075306221773e-07, + "loss": 0.9698, + "step": 70185 + }, + { + "epoch": 5.19, + "learning_rate": 9.492379035041321e-07, + "loss": 1.0397, + "step": 70186 + }, + { + "epoch": 5.19, + "learning_rate": 9.490682907857196e-07, + "loss": 1.0502, + "step": 70187 + }, + { + "epoch": 5.19, + "learning_rate": 9.488986924672095e-07, + "loss": 0.9945, + "step": 70188 + }, + { + "epoch": 5.19, + "learning_rate": 9.487291085488725e-07, + "loss": 1.0968, + "step": 70189 + }, + { + "epoch": 5.19, + "learning_rate": 9.485595390309755e-07, + "loss": 1.0532, + "step": 70190 + }, + { + "epoch": 5.19, + "learning_rate": 9.483899839137933e-07, + "loss": 1.1139, + "step": 70191 + }, + { + "epoch": 5.19, + "learning_rate": 9.482204431975928e-07, + "loss": 0.9554, + "step": 70192 + }, + { + "epoch": 5.19, + "learning_rate": 9.480509168826447e-07, + "loss": 0.9025, + "step": 70193 + }, + { + "epoch": 5.19, + "learning_rate": 9.478814049692153e-07, + "loss": 1.0787, + "step": 70194 + }, + { + "epoch": 5.19, + "learning_rate": 9.477119074575791e-07, + "loss": 1.0765, + "step": 70195 + }, + { + "epoch": 5.19, + "learning_rate": 9.475424243480047e-07, + "loss": 0.9062, + "step": 70196 + }, + { + "epoch": 5.19, + "learning_rate": 9.473729556407596e-07, + "loss": 1.0995, + "step": 70197 + }, + { + "epoch": 5.19, + "learning_rate": 9.472035013361147e-07, + "loss": 1.0334, + "step": 70198 + }, + { + "epoch": 5.19, + "learning_rate": 9.470340614343387e-07, + "loss": 1.1232, + "step": 70199 + }, + { + "epoch": 5.19, + "learning_rate": 9.468646359357047e-07, + "loss": 1.0466, + "step": 70200 + }, + { + "epoch": 5.19, + "learning_rate": 9.46695224840476e-07, + "loss": 0.8982, + "step": 70201 + }, + { + "epoch": 5.19, + "learning_rate": 9.465258281489276e-07, + "loss": 0.9544, + "step": 70202 + }, + { + "epoch": 5.19, + "learning_rate": 9.463564458613251e-07, + "loss": 1.0732, + "step": 70203 + }, + { + "epoch": 5.19, + "learning_rate": 9.461870779779447e-07, + "loss": 1.0756, + "step": 70204 + }, + { + "epoch": 5.19, + "learning_rate": 9.460177244990465e-07, + "loss": 1.008, + "step": 70205 + }, + { + "epoch": 5.19, + "learning_rate": 9.458483854249057e-07, + "loss": 1.0246, + "step": 70206 + }, + { + "epoch": 5.19, + "learning_rate": 9.45679060755792e-07, + "loss": 1.0031, + "step": 70207 + }, + { + "epoch": 5.19, + "learning_rate": 9.455097504919719e-07, + "loss": 1.0276, + "step": 70208 + }, + { + "epoch": 5.19, + "learning_rate": 9.453404546337175e-07, + "loss": 0.9333, + "step": 70209 + }, + { + "epoch": 5.19, + "learning_rate": 9.451711731812941e-07, + "loss": 0.9856, + "step": 70210 + }, + { + "epoch": 5.19, + "learning_rate": 9.450019061349769e-07, + "loss": 0.9901, + "step": 70211 + }, + { + "epoch": 5.19, + "learning_rate": 9.448326534950314e-07, + "loss": 0.9617, + "step": 70212 + }, + { + "epoch": 5.19, + "learning_rate": 9.446634152617273e-07, + "loss": 0.9875, + "step": 70213 + }, + { + "epoch": 5.19, + "learning_rate": 9.444941914353323e-07, + "loss": 0.9323, + "step": 70214 + }, + { + "epoch": 5.19, + "learning_rate": 9.443249820161193e-07, + "loss": 1.0727, + "step": 70215 + }, + { + "epoch": 5.19, + "learning_rate": 9.441557870043561e-07, + "loss": 0.9319, + "step": 70216 + }, + { + "epoch": 5.19, + "learning_rate": 9.439866064003112e-07, + "loss": 1.0004, + "step": 70217 + }, + { + "epoch": 5.19, + "learning_rate": 9.438174402042544e-07, + "loss": 0.9998, + "step": 70218 + }, + { + "epoch": 5.19, + "learning_rate": 9.436482884164522e-07, + "loss": 1.0181, + "step": 70219 + }, + { + "epoch": 5.19, + "learning_rate": 9.434791510371788e-07, + "loss": 0.9997, + "step": 70220 + }, + { + "epoch": 5.19, + "learning_rate": 9.433100280666996e-07, + "loss": 0.9981, + "step": 70221 + }, + { + "epoch": 5.19, + "learning_rate": 9.431409195052843e-07, + "loss": 0.8995, + "step": 70222 + }, + { + "epoch": 5.19, + "learning_rate": 9.429718253532005e-07, + "loss": 1.006, + "step": 70223 + }, + { + "epoch": 5.19, + "learning_rate": 9.428027456107213e-07, + "loss": 1.0699, + "step": 70224 + }, + { + "epoch": 5.19, + "learning_rate": 9.426336802781155e-07, + "loss": 1.0423, + "step": 70225 + }, + { + "epoch": 5.19, + "learning_rate": 9.42464629355645e-07, + "loss": 1.0401, + "step": 70226 + }, + { + "epoch": 5.19, + "learning_rate": 9.422955928435862e-07, + "loss": 0.9401, + "step": 70227 + }, + { + "epoch": 5.19, + "learning_rate": 9.421265707422044e-07, + "loss": 0.9451, + "step": 70228 + }, + { + "epoch": 5.19, + "learning_rate": 9.419575630517719e-07, + "loss": 0.9545, + "step": 70229 + }, + { + "epoch": 5.19, + "learning_rate": 9.417885697725527e-07, + "loss": 1.0189, + "step": 70230 + }, + { + "epoch": 5.19, + "learning_rate": 9.416195909048198e-07, + "loss": 1.0079, + "step": 70231 + }, + { + "epoch": 5.19, + "learning_rate": 9.414506264488398e-07, + "loss": 1.007, + "step": 70232 + }, + { + "epoch": 5.19, + "learning_rate": 9.412816764048838e-07, + "loss": 0.9504, + "step": 70233 + }, + { + "epoch": 5.19, + "learning_rate": 9.411127407732179e-07, + "loss": 0.9986, + "step": 70234 + }, + { + "epoch": 5.19, + "learning_rate": 9.409438195541099e-07, + "loss": 1.077, + "step": 70235 + }, + { + "epoch": 5.19, + "learning_rate": 9.407749127478339e-07, + "loss": 1.0666, + "step": 70236 + }, + { + "epoch": 5.19, + "learning_rate": 9.406060203546541e-07, + "loss": 1.0229, + "step": 70237 + }, + { + "epoch": 5.19, + "learning_rate": 9.404371423748415e-07, + "loss": 1.0392, + "step": 70238 + }, + { + "epoch": 5.19, + "learning_rate": 9.402682788086614e-07, + "loss": 1.0214, + "step": 70239 + }, + { + "epoch": 5.19, + "learning_rate": 9.400994296563859e-07, + "loss": 0.9739, + "step": 70240 + }, + { + "epoch": 5.19, + "learning_rate": 9.399305949182836e-07, + "loss": 0.9769, + "step": 70241 + }, + { + "epoch": 5.19, + "learning_rate": 9.397617745946219e-07, + "loss": 0.9445, + "step": 70242 + }, + { + "epoch": 5.19, + "learning_rate": 9.395929686856698e-07, + "loss": 0.9956, + "step": 70243 + }, + { + "epoch": 5.19, + "learning_rate": 9.394241771916934e-07, + "loss": 1.0166, + "step": 70244 + }, + { + "epoch": 5.19, + "learning_rate": 9.392554001129672e-07, + "loss": 1.0563, + "step": 70245 + }, + { + "epoch": 5.19, + "learning_rate": 9.390866374497532e-07, + "loss": 1.0667, + "step": 70246 + }, + { + "epoch": 5.19, + "learning_rate": 9.389178892023231e-07, + "loss": 0.9853, + "step": 70247 + }, + { + "epoch": 5.19, + "learning_rate": 9.387491553709438e-07, + "loss": 1.0468, + "step": 70248 + }, + { + "epoch": 5.19, + "learning_rate": 9.385804359558892e-07, + "loss": 1.0455, + "step": 70249 + }, + { + "epoch": 5.19, + "learning_rate": 9.384117309574192e-07, + "loss": 0.9637, + "step": 70250 + }, + { + "epoch": 5.19, + "learning_rate": 9.38243040375808e-07, + "loss": 1.1193, + "step": 70251 + }, + { + "epoch": 5.19, + "learning_rate": 9.38074364211321e-07, + "loss": 1.0542, + "step": 70252 + }, + { + "epoch": 5.19, + "learning_rate": 9.379057024642324e-07, + "loss": 0.957, + "step": 70253 + }, + { + "epoch": 5.19, + "learning_rate": 9.377370551348042e-07, + "loss": 1.024, + "step": 70254 + }, + { + "epoch": 5.19, + "learning_rate": 9.375684222233039e-07, + "loss": 0.9251, + "step": 70255 + }, + { + "epoch": 5.19, + "learning_rate": 9.373998037300047e-07, + "loss": 1.0457, + "step": 70256 + }, + { + "epoch": 5.19, + "learning_rate": 9.372311996551731e-07, + "loss": 0.9649, + "step": 70257 + }, + { + "epoch": 5.19, + "learning_rate": 9.370626099990765e-07, + "loss": 1.03, + "step": 70258 + }, + { + "epoch": 5.19, + "learning_rate": 9.368940347619826e-07, + "loss": 1.0322, + "step": 70259 + }, + { + "epoch": 5.19, + "learning_rate": 9.367254739441623e-07, + "loss": 0.9915, + "step": 70260 + }, + { + "epoch": 5.19, + "learning_rate": 9.365569275458807e-07, + "loss": 0.9576, + "step": 70261 + }, + { + "epoch": 5.19, + "learning_rate": 9.363883955674091e-07, + "loss": 0.9298, + "step": 70262 + }, + { + "epoch": 5.19, + "learning_rate": 9.362198780090126e-07, + "loss": 1.0837, + "step": 70263 + }, + { + "epoch": 5.19, + "learning_rate": 9.360513748709599e-07, + "loss": 0.9982, + "step": 70264 + }, + { + "epoch": 5.19, + "learning_rate": 9.358828861535208e-07, + "loss": 1.0388, + "step": 70265 + }, + { + "epoch": 5.19, + "learning_rate": 9.357144118569628e-07, + "loss": 0.9725, + "step": 70266 + }, + { + "epoch": 5.19, + "learning_rate": 9.355459519815535e-07, + "loss": 1.0023, + "step": 70267 + }, + { + "epoch": 5.19, + "learning_rate": 9.353775065275583e-07, + "loss": 0.9714, + "step": 70268 + }, + { + "epoch": 5.19, + "learning_rate": 9.352090754952503e-07, + "loss": 1.055, + "step": 70269 + }, + { + "epoch": 5.19, + "learning_rate": 9.350406588848948e-07, + "loss": 1.0151, + "step": 70270 + }, + { + "epoch": 5.19, + "learning_rate": 9.348722566967605e-07, + "loss": 0.957, + "step": 70271 + }, + { + "epoch": 5.19, + "learning_rate": 9.347038689311139e-07, + "loss": 0.9167, + "step": 70272 + }, + { + "epoch": 5.19, + "learning_rate": 9.345354955882225e-07, + "loss": 1.0669, + "step": 70273 + }, + { + "epoch": 5.19, + "learning_rate": 9.343671366683582e-07, + "loss": 0.9839, + "step": 70274 + }, + { + "epoch": 5.19, + "learning_rate": 9.341987921717832e-07, + "loss": 1.0637, + "step": 70275 + }, + { + "epoch": 5.19, + "learning_rate": 9.340304620987705e-07, + "loss": 1.0274, + "step": 70276 + }, + { + "epoch": 5.19, + "learning_rate": 9.338621464495845e-07, + "loss": 1.0326, + "step": 70277 + }, + { + "epoch": 5.19, + "learning_rate": 9.336938452244948e-07, + "loss": 0.9172, + "step": 70278 + }, + { + "epoch": 5.19, + "learning_rate": 9.335255584237668e-07, + "loss": 1.0735, + "step": 70279 + }, + { + "epoch": 5.19, + "learning_rate": 9.333572860476714e-07, + "loss": 1.0346, + "step": 70280 + }, + { + "epoch": 5.19, + "learning_rate": 9.331890280964739e-07, + "loss": 0.9986, + "step": 70281 + }, + { + "epoch": 5.19, + "learning_rate": 9.330207845704442e-07, + "loss": 1.1059, + "step": 70282 + }, + { + "epoch": 5.19, + "learning_rate": 9.328525554698487e-07, + "loss": 1.1213, + "step": 70283 + }, + { + "epoch": 5.19, + "learning_rate": 9.326843407949526e-07, + "loss": 1.1032, + "step": 70284 + }, + { + "epoch": 5.19, + "learning_rate": 9.325161405460281e-07, + "loss": 1.0868, + "step": 70285 + }, + { + "epoch": 5.19, + "learning_rate": 9.323479547233405e-07, + "loss": 0.911, + "step": 70286 + }, + { + "epoch": 5.19, + "learning_rate": 9.321797833271584e-07, + "loss": 0.9106, + "step": 70287 + }, + { + "epoch": 5.19, + "learning_rate": 9.32011626357745e-07, + "loss": 1.0527, + "step": 70288 + }, + { + "epoch": 5.19, + "learning_rate": 9.318434838153744e-07, + "loss": 1.067, + "step": 70289 + }, + { + "epoch": 5.19, + "learning_rate": 9.316753557003133e-07, + "loss": 0.9665, + "step": 70290 + }, + { + "epoch": 5.19, + "learning_rate": 9.315072420128224e-07, + "loss": 1.023, + "step": 70291 + }, + { + "epoch": 5.19, + "learning_rate": 9.31339142753176e-07, + "loss": 0.9677, + "step": 70292 + }, + { + "epoch": 5.19, + "learning_rate": 9.311710579216371e-07, + "loss": 1.019, + "step": 70293 + }, + { + "epoch": 5.19, + "learning_rate": 9.310029875184789e-07, + "loss": 1.0522, + "step": 70294 + }, + { + "epoch": 5.19, + "learning_rate": 9.308349315439625e-07, + "loss": 1.0637, + "step": 70295 + }, + { + "epoch": 5.19, + "learning_rate": 9.306668899983595e-07, + "loss": 1.0442, + "step": 70296 + }, + { + "epoch": 5.19, + "learning_rate": 9.304988628819334e-07, + "loss": 0.9948, + "step": 70297 + }, + { + "epoch": 5.19, + "learning_rate": 9.30330850194957e-07, + "loss": 0.9884, + "step": 70298 + }, + { + "epoch": 5.19, + "learning_rate": 9.301628519376938e-07, + "loss": 1.0418, + "step": 70299 + }, + { + "epoch": 5.19, + "learning_rate": 9.299948681104087e-07, + "loss": 0.9523, + "step": 70300 + }, + { + "epoch": 5.19, + "learning_rate": 9.298268987133752e-07, + "loss": 0.9779, + "step": 70301 + }, + { + "epoch": 5.19, + "learning_rate": 9.296589437468562e-07, + "loss": 0.9398, + "step": 70302 + }, + { + "epoch": 5.19, + "learning_rate": 9.294910032111205e-07, + "loss": 1.0067, + "step": 70303 + }, + { + "epoch": 5.19, + "learning_rate": 9.293230771064332e-07, + "loss": 0.9178, + "step": 70304 + }, + { + "epoch": 5.19, + "learning_rate": 9.291551654330644e-07, + "loss": 0.9865, + "step": 70305 + }, + { + "epoch": 5.19, + "learning_rate": 9.289872681912804e-07, + "loss": 0.9895, + "step": 70306 + }, + { + "epoch": 5.19, + "learning_rate": 9.288193853813476e-07, + "loss": 1.0892, + "step": 70307 + }, + { + "epoch": 5.2, + "learning_rate": 9.286515170035315e-07, + "loss": 1.0127, + "step": 70308 + }, + { + "epoch": 5.2, + "learning_rate": 9.284836630581029e-07, + "loss": 0.9953, + "step": 70309 + }, + { + "epoch": 5.2, + "learning_rate": 9.283158235453271e-07, + "loss": 1.0985, + "step": 70310 + }, + { + "epoch": 5.2, + "learning_rate": 9.281479984654707e-07, + "loss": 0.9845, + "step": 70311 + }, + { + "epoch": 5.2, + "learning_rate": 9.279801878188022e-07, + "loss": 0.9714, + "step": 70312 + }, + { + "epoch": 5.2, + "learning_rate": 9.278123916055848e-07, + "loss": 1.0306, + "step": 70313 + }, + { + "epoch": 5.2, + "learning_rate": 9.276446098260894e-07, + "loss": 0.8915, + "step": 70314 + }, + { + "epoch": 5.2, + "learning_rate": 9.274768424805824e-07, + "loss": 1.1442, + "step": 70315 + }, + { + "epoch": 5.2, + "learning_rate": 9.273090895693293e-07, + "loss": 1.0284, + "step": 70316 + }, + { + "epoch": 5.2, + "learning_rate": 9.271413510925964e-07, + "loss": 1.0685, + "step": 70317 + }, + { + "epoch": 5.2, + "learning_rate": 9.269736270506546e-07, + "loss": 0.9472, + "step": 70318 + }, + { + "epoch": 5.2, + "learning_rate": 9.268059174437672e-07, + "loss": 0.9177, + "step": 70319 + }, + { + "epoch": 5.2, + "learning_rate": 9.266382222721992e-07, + "loss": 1.0335, + "step": 70320 + }, + { + "epoch": 5.2, + "learning_rate": 9.264705415362218e-07, + "loss": 0.9152, + "step": 70321 + }, + { + "epoch": 5.2, + "learning_rate": 9.263028752361002e-07, + "loss": 1.0105, + "step": 70322 + }, + { + "epoch": 5.2, + "learning_rate": 9.26135223372101e-07, + "loss": 0.8983, + "step": 70323 + }, + { + "epoch": 5.2, + "learning_rate": 9.259675859444894e-07, + "loss": 1.0466, + "step": 70324 + }, + { + "epoch": 5.2, + "learning_rate": 9.257999629535352e-07, + "loss": 0.9239, + "step": 70325 + }, + { + "epoch": 5.2, + "learning_rate": 9.256323543995027e-07, + "loss": 0.9577, + "step": 70326 + }, + { + "epoch": 5.2, + "learning_rate": 9.254647602826594e-07, + "loss": 0.9923, + "step": 70327 + }, + { + "epoch": 5.2, + "learning_rate": 9.252971806032728e-07, + "loss": 1.0508, + "step": 70328 + }, + { + "epoch": 5.2, + "learning_rate": 9.251296153616063e-07, + "loss": 0.9521, + "step": 70329 + }, + { + "epoch": 5.2, + "learning_rate": 9.249620645579293e-07, + "loss": 0.9796, + "step": 70330 + }, + { + "epoch": 5.2, + "learning_rate": 9.247945281925097e-07, + "loss": 1.0252, + "step": 70331 + }, + { + "epoch": 5.2, + "learning_rate": 9.246270062656104e-07, + "loss": 0.9349, + "step": 70332 + }, + { + "epoch": 5.2, + "learning_rate": 9.244594987774991e-07, + "loss": 1.0657, + "step": 70333 + }, + { + "epoch": 5.2, + "learning_rate": 9.242920057284443e-07, + "loss": 0.9443, + "step": 70334 + }, + { + "epoch": 5.2, + "learning_rate": 9.241245271187116e-07, + "loss": 1.0099, + "step": 70335 + }, + { + "epoch": 5.2, + "learning_rate": 9.239570629485661e-07, + "loss": 1.084, + "step": 70336 + }, + { + "epoch": 5.2, + "learning_rate": 9.237896132182756e-07, + "loss": 0.9827, + "step": 70337 + }, + { + "epoch": 5.2, + "learning_rate": 9.236221779281041e-07, + "loss": 1.1005, + "step": 70338 + }, + { + "epoch": 5.2, + "learning_rate": 9.234547570783237e-07, + "loss": 1.0659, + "step": 70339 + }, + { + "epoch": 5.2, + "learning_rate": 9.23287350669192e-07, + "loss": 0.9266, + "step": 70340 + }, + { + "epoch": 5.2, + "learning_rate": 9.231199587009832e-07, + "loss": 1.011, + "step": 70341 + }, + { + "epoch": 5.2, + "learning_rate": 9.229525811739592e-07, + "loss": 0.956, + "step": 70342 + }, + { + "epoch": 5.2, + "learning_rate": 9.227852180883912e-07, + "loss": 0.9814, + "step": 70343 + }, + { + "epoch": 5.2, + "learning_rate": 9.226178694445376e-07, + "loss": 0.8938, + "step": 70344 + }, + { + "epoch": 5.2, + "learning_rate": 9.224505352426715e-07, + "loss": 1.0464, + "step": 70345 + }, + { + "epoch": 5.2, + "learning_rate": 9.222832154830564e-07, + "loss": 0.9331, + "step": 70346 + }, + { + "epoch": 5.2, + "learning_rate": 9.221159101659593e-07, + "loss": 0.9722, + "step": 70347 + }, + { + "epoch": 5.2, + "learning_rate": 9.21948619291645e-07, + "loss": 1.0143, + "step": 70348 + }, + { + "epoch": 5.2, + "learning_rate": 9.217813428603794e-07, + "loss": 0.9947, + "step": 70349 + }, + { + "epoch": 5.2, + "learning_rate": 9.216140808724317e-07, + "loss": 1.0139, + "step": 70350 + }, + { + "epoch": 5.2, + "learning_rate": 9.214468333280668e-07, + "loss": 1.0496, + "step": 70351 + }, + { + "epoch": 5.2, + "learning_rate": 9.212796002275492e-07, + "loss": 0.936, + "step": 70352 + }, + { + "epoch": 5.2, + "learning_rate": 9.211123815711442e-07, + "loss": 0.934, + "step": 70353 + }, + { + "epoch": 5.2, + "learning_rate": 9.209451773591205e-07, + "loss": 0.9756, + "step": 70354 + }, + { + "epoch": 5.2, + "learning_rate": 9.207779875917444e-07, + "loss": 0.9351, + "step": 70355 + }, + { + "epoch": 5.2, + "learning_rate": 9.206108122692803e-07, + "loss": 1.0508, + "step": 70356 + }, + { + "epoch": 5.2, + "learning_rate": 9.204436513919935e-07, + "loss": 0.9488, + "step": 70357 + }, + { + "epoch": 5.2, + "learning_rate": 9.202765049601503e-07, + "loss": 1.0683, + "step": 70358 + }, + { + "epoch": 5.2, + "learning_rate": 9.201093729740184e-07, + "loss": 0.9704, + "step": 70359 + }, + { + "epoch": 5.2, + "learning_rate": 9.199422554338632e-07, + "loss": 1.1264, + "step": 70360 + }, + { + "epoch": 5.2, + "learning_rate": 9.1977515233995e-07, + "loss": 1.0048, + "step": 70361 + }, + { + "epoch": 5.2, + "learning_rate": 9.19608063692542e-07, + "loss": 0.8734, + "step": 70362 + }, + { + "epoch": 5.2, + "learning_rate": 9.19440989491912e-07, + "loss": 1.0556, + "step": 70363 + }, + { + "epoch": 5.2, + "learning_rate": 9.19273929738318e-07, + "loss": 0.9689, + "step": 70364 + }, + { + "epoch": 5.2, + "learning_rate": 9.191068844320305e-07, + "loss": 1.0493, + "step": 70365 + }, + { + "epoch": 5.2, + "learning_rate": 9.189398535733141e-07, + "loss": 1.125, + "step": 70366 + }, + { + "epoch": 5.2, + "learning_rate": 9.187728371624338e-07, + "loss": 1.1904, + "step": 70367 + }, + { + "epoch": 5.2, + "learning_rate": 9.186058351996574e-07, + "loss": 1.0981, + "step": 70368 + }, + { + "epoch": 5.2, + "learning_rate": 9.184388476852469e-07, + "loss": 1.0665, + "step": 70369 + }, + { + "epoch": 5.2, + "learning_rate": 9.182718746194718e-07, + "loss": 0.9111, + "step": 70370 + }, + { + "epoch": 5.2, + "learning_rate": 9.181049160025956e-07, + "loss": 0.9729, + "step": 70371 + }, + { + "epoch": 5.2, + "learning_rate": 9.179379718348858e-07, + "loss": 1.0099, + "step": 70372 + }, + { + "epoch": 5.2, + "learning_rate": 9.177710421166042e-07, + "loss": 1.0064, + "step": 70373 + }, + { + "epoch": 5.2, + "learning_rate": 9.176041268480207e-07, + "loss": 0.9942, + "step": 70374 + }, + { + "epoch": 5.2, + "learning_rate": 9.174372260293996e-07, + "loss": 0.8886, + "step": 70375 + }, + { + "epoch": 5.2, + "learning_rate": 9.17270339661005e-07, + "loss": 0.9912, + "step": 70376 + }, + { + "epoch": 5.2, + "learning_rate": 9.171034677431045e-07, + "loss": 1.0813, + "step": 70377 + }, + { + "epoch": 5.2, + "learning_rate": 9.169366102759592e-07, + "loss": 0.9767, + "step": 70378 + }, + { + "epoch": 5.2, + "learning_rate": 9.167697672598408e-07, + "loss": 1.0756, + "step": 70379 + }, + { + "epoch": 5.2, + "learning_rate": 9.166029386950115e-07, + "loss": 0.9642, + "step": 70380 + }, + { + "epoch": 5.2, + "learning_rate": 9.164361245817366e-07, + "loss": 0.9855, + "step": 70381 + }, + { + "epoch": 5.2, + "learning_rate": 9.162693249202804e-07, + "loss": 0.9582, + "step": 70382 + }, + { + "epoch": 5.2, + "learning_rate": 9.161025397109114e-07, + "loss": 0.959, + "step": 70383 + }, + { + "epoch": 5.2, + "learning_rate": 9.15935768953895e-07, + "loss": 0.966, + "step": 70384 + }, + { + "epoch": 5.2, + "learning_rate": 9.157690126494911e-07, + "loss": 0.9035, + "step": 70385 + }, + { + "epoch": 5.2, + "learning_rate": 9.156022707979717e-07, + "loss": 1.0255, + "step": 70386 + }, + { + "epoch": 5.2, + "learning_rate": 9.154355433995954e-07, + "loss": 0.8447, + "step": 70387 + }, + { + "epoch": 5.2, + "learning_rate": 9.152688304546364e-07, + "loss": 1.0662, + "step": 70388 + }, + { + "epoch": 5.2, + "learning_rate": 9.151021319633502e-07, + "loss": 1.0264, + "step": 70389 + }, + { + "epoch": 5.2, + "learning_rate": 9.149354479260087e-07, + "loss": 1.0755, + "step": 70390 + }, + { + "epoch": 5.2, + "learning_rate": 9.147687783428727e-07, + "loss": 1.0197, + "step": 70391 + }, + { + "epoch": 5.2, + "learning_rate": 9.146021232142144e-07, + "loss": 0.9293, + "step": 70392 + }, + { + "epoch": 5.2, + "learning_rate": 9.144354825402912e-07, + "loss": 0.9002, + "step": 70393 + }, + { + "epoch": 5.2, + "learning_rate": 9.142688563213698e-07, + "loss": 1.0371, + "step": 70394 + }, + { + "epoch": 5.2, + "learning_rate": 9.141022445577175e-07, + "loss": 0.9766, + "step": 70395 + }, + { + "epoch": 5.2, + "learning_rate": 9.139356472495997e-07, + "loss": 0.9779, + "step": 70396 + }, + { + "epoch": 5.2, + "learning_rate": 9.137690643972797e-07, + "loss": 1.0363, + "step": 70397 + }, + { + "epoch": 5.2, + "learning_rate": 9.136024960010215e-07, + "loss": 0.8818, + "step": 70398 + }, + { + "epoch": 5.2, + "learning_rate": 9.134359420610939e-07, + "loss": 0.9769, + "step": 70399 + }, + { + "epoch": 5.2, + "learning_rate": 9.132694025777588e-07, + "loss": 1.0877, + "step": 70400 + }, + { + "epoch": 5.2, + "learning_rate": 9.131028775512829e-07, + "loss": 1.0357, + "step": 70401 + }, + { + "epoch": 5.2, + "learning_rate": 9.129363669819292e-07, + "loss": 1.0061, + "step": 70402 + }, + { + "epoch": 5.2, + "learning_rate": 9.127698708699628e-07, + "loss": 1.0014, + "step": 70403 + }, + { + "epoch": 5.2, + "learning_rate": 9.126033892156516e-07, + "loss": 0.9431, + "step": 70404 + }, + { + "epoch": 5.2, + "learning_rate": 9.124369220192574e-07, + "loss": 0.9318, + "step": 70405 + }, + { + "epoch": 5.2, + "learning_rate": 9.122704692810469e-07, + "loss": 1.0932, + "step": 70406 + }, + { + "epoch": 5.2, + "learning_rate": 9.121040310012819e-07, + "loss": 0.985, + "step": 70407 + }, + { + "epoch": 5.2, + "learning_rate": 9.119376071802333e-07, + "loss": 1.0363, + "step": 70408 + }, + { + "epoch": 5.2, + "learning_rate": 9.117711978181587e-07, + "loss": 1.0757, + "step": 70409 + }, + { + "epoch": 5.2, + "learning_rate": 9.116048029153268e-07, + "loss": 1.0045, + "step": 70410 + }, + { + "epoch": 5.2, + "learning_rate": 9.114384224720019e-07, + "loss": 0.8935, + "step": 70411 + }, + { + "epoch": 5.2, + "learning_rate": 9.112720564884492e-07, + "loss": 0.9075, + "step": 70412 + }, + { + "epoch": 5.2, + "learning_rate": 9.111057049649319e-07, + "loss": 0.9216, + "step": 70413 + }, + { + "epoch": 5.2, + "learning_rate": 9.109393679017142e-07, + "loss": 0.9742, + "step": 70414 + }, + { + "epoch": 5.2, + "learning_rate": 9.107730452990626e-07, + "loss": 0.9308, + "step": 70415 + }, + { + "epoch": 5.2, + "learning_rate": 9.106067371572414e-07, + "loss": 0.9181, + "step": 70416 + }, + { + "epoch": 5.2, + "learning_rate": 9.104404434765157e-07, + "loss": 0.9018, + "step": 70417 + }, + { + "epoch": 5.2, + "learning_rate": 9.102741642571467e-07, + "loss": 1.0186, + "step": 70418 + }, + { + "epoch": 5.2, + "learning_rate": 9.101078994994039e-07, + "loss": 0.9485, + "step": 70419 + }, + { + "epoch": 5.2, + "learning_rate": 9.099416492035484e-07, + "loss": 1.036, + "step": 70420 + }, + { + "epoch": 5.2, + "learning_rate": 9.097754133698467e-07, + "loss": 0.9998, + "step": 70421 + }, + { + "epoch": 5.2, + "learning_rate": 9.096091919985617e-07, + "loss": 0.993, + "step": 70422 + }, + { + "epoch": 5.2, + "learning_rate": 9.094429850899567e-07, + "loss": 0.9892, + "step": 70423 + }, + { + "epoch": 5.2, + "learning_rate": 9.092767926442991e-07, + "loss": 1.0448, + "step": 70424 + }, + { + "epoch": 5.2, + "learning_rate": 9.091106146618522e-07, + "loss": 0.9698, + "step": 70425 + }, + { + "epoch": 5.2, + "learning_rate": 9.089444511428813e-07, + "loss": 1.0282, + "step": 70426 + }, + { + "epoch": 5.2, + "learning_rate": 9.087783020876473e-07, + "loss": 1.0426, + "step": 70427 + }, + { + "epoch": 5.2, + "learning_rate": 9.086121674964177e-07, + "loss": 1.0508, + "step": 70428 + }, + { + "epoch": 5.2, + "learning_rate": 9.084460473694579e-07, + "loss": 1.0811, + "step": 70429 + }, + { + "epoch": 5.2, + "learning_rate": 9.082799417070287e-07, + "loss": 1.0968, + "step": 70430 + }, + { + "epoch": 5.2, + "learning_rate": 9.081138505093967e-07, + "loss": 1.0177, + "step": 70431 + }, + { + "epoch": 5.2, + "learning_rate": 9.079477737768239e-07, + "loss": 1.0519, + "step": 70432 + }, + { + "epoch": 5.2, + "learning_rate": 9.0778171150958e-07, + "loss": 1.0004, + "step": 70433 + }, + { + "epoch": 5.2, + "learning_rate": 9.076156637079214e-07, + "loss": 1.0147, + "step": 70434 + }, + { + "epoch": 5.2, + "learning_rate": 9.074496303721181e-07, + "loss": 0.9357, + "step": 70435 + }, + { + "epoch": 5.2, + "learning_rate": 9.072836115024297e-07, + "loss": 0.9719, + "step": 70436 + }, + { + "epoch": 5.2, + "learning_rate": 9.071176070991272e-07, + "loss": 0.9378, + "step": 70437 + }, + { + "epoch": 5.2, + "learning_rate": 9.069516171624671e-07, + "loss": 0.9726, + "step": 70438 + }, + { + "epoch": 5.2, + "learning_rate": 9.067856416927179e-07, + "loss": 0.9109, + "step": 70439 + }, + { + "epoch": 5.2, + "learning_rate": 9.06619680690144e-07, + "loss": 1.0039, + "step": 70440 + }, + { + "epoch": 5.2, + "learning_rate": 9.064537341550073e-07, + "loss": 0.946, + "step": 70441 + }, + { + "epoch": 5.2, + "learning_rate": 9.062878020875721e-07, + "loss": 0.9575, + "step": 70442 + }, + { + "epoch": 5.2, + "learning_rate": 9.061218844881014e-07, + "loss": 0.9493, + "step": 70443 + }, + { + "epoch": 5.21, + "learning_rate": 9.05955981356863e-07, + "loss": 1.0571, + "step": 70444 + }, + { + "epoch": 5.21, + "learning_rate": 9.057900926941177e-07, + "loss": 1.0221, + "step": 70445 + }, + { + "epoch": 5.21, + "learning_rate": 9.056242185001307e-07, + "loss": 0.9734, + "step": 70446 + }, + { + "epoch": 5.21, + "learning_rate": 9.054583587751642e-07, + "loss": 0.8988, + "step": 70447 + }, + { + "epoch": 5.21, + "learning_rate": 9.052925135194846e-07, + "loss": 0.9354, + "step": 70448 + }, + { + "epoch": 5.21, + "learning_rate": 9.051266827333549e-07, + "loss": 0.9351, + "step": 70449 + }, + { + "epoch": 5.21, + "learning_rate": 9.049608664170372e-07, + "loss": 1.0334, + "step": 70450 + }, + { + "epoch": 5.21, + "learning_rate": 9.047950645707981e-07, + "loss": 1.0111, + "step": 70451 + }, + { + "epoch": 5.21, + "learning_rate": 9.046292771948972e-07, + "loss": 1.0868, + "step": 70452 + }, + { + "epoch": 5.21, + "learning_rate": 9.044635042896055e-07, + "loss": 1.0253, + "step": 70453 + }, + { + "epoch": 5.21, + "learning_rate": 9.042977458551783e-07, + "loss": 1.0828, + "step": 70454 + }, + { + "epoch": 5.21, + "learning_rate": 9.041320018918842e-07, + "loss": 1.0671, + "step": 70455 + }, + { + "epoch": 5.21, + "learning_rate": 9.039662723999854e-07, + "loss": 1.1136, + "step": 70456 + }, + { + "epoch": 5.21, + "learning_rate": 9.038005573797481e-07, + "loss": 0.8782, + "step": 70457 + }, + { + "epoch": 5.21, + "learning_rate": 9.036348568314335e-07, + "loss": 1.0869, + "step": 70458 + }, + { + "epoch": 5.21, + "learning_rate": 9.034691707553034e-07, + "loss": 0.9799, + "step": 70459 + }, + { + "epoch": 5.21, + "learning_rate": 9.033034991516265e-07, + "loss": 0.9643, + "step": 70460 + }, + { + "epoch": 5.21, + "learning_rate": 9.031378420206627e-07, + "loss": 0.9805, + "step": 70461 + }, + { + "epoch": 5.21, + "learning_rate": 9.029721993626761e-07, + "loss": 0.953, + "step": 70462 + }, + { + "epoch": 5.21, + "learning_rate": 9.028065711779288e-07, + "loss": 1.0115, + "step": 70463 + }, + { + "epoch": 5.21, + "learning_rate": 9.026409574666883e-07, + "loss": 1.0215, + "step": 70464 + }, + { + "epoch": 5.21, + "learning_rate": 9.024753582292156e-07, + "loss": 1.0987, + "step": 70465 + }, + { + "epoch": 5.21, + "learning_rate": 9.023097734657748e-07, + "loss": 0.9683, + "step": 70466 + }, + { + "epoch": 5.21, + "learning_rate": 9.02144203176627e-07, + "loss": 1.0228, + "step": 70467 + }, + { + "epoch": 5.21, + "learning_rate": 9.019786473620395e-07, + "loss": 0.9619, + "step": 70468 + }, + { + "epoch": 5.21, + "learning_rate": 9.018131060222735e-07, + "loss": 1.021, + "step": 70469 + }, + { + "epoch": 5.21, + "learning_rate": 9.016475791575929e-07, + "loss": 0.837, + "step": 70470 + }, + { + "epoch": 5.21, + "learning_rate": 9.014820667682611e-07, + "loss": 0.9629, + "step": 70471 + }, + { + "epoch": 5.21, + "learning_rate": 9.013165688545388e-07, + "loss": 1.0274, + "step": 70472 + }, + { + "epoch": 5.21, + "learning_rate": 9.011510854166938e-07, + "loss": 0.9533, + "step": 70473 + }, + { + "epoch": 5.21, + "learning_rate": 9.009856164549879e-07, + "loss": 1.0937, + "step": 70474 + }, + { + "epoch": 5.21, + "learning_rate": 9.008201619696832e-07, + "loss": 1.0688, + "step": 70475 + }, + { + "epoch": 5.21, + "learning_rate": 9.006547219610406e-07, + "loss": 1.0791, + "step": 70476 + }, + { + "epoch": 5.21, + "learning_rate": 9.004892964293288e-07, + "loss": 1.0823, + "step": 70477 + }, + { + "epoch": 5.21, + "learning_rate": 9.003238853748108e-07, + "loss": 1.035, + "step": 70478 + }, + { + "epoch": 5.21, + "learning_rate": 9.001584887977421e-07, + "loss": 1.0081, + "step": 70479 + }, + { + "epoch": 5.21, + "learning_rate": 8.999931066983947e-07, + "loss": 0.9867, + "step": 70480 + }, + { + "epoch": 5.21, + "learning_rate": 8.998277390770249e-07, + "loss": 1.0183, + "step": 70481 + }, + { + "epoch": 5.21, + "learning_rate": 8.996623859339026e-07, + "loss": 1.0601, + "step": 70482 + }, + { + "epoch": 5.21, + "learning_rate": 8.994970472692843e-07, + "loss": 1.0756, + "step": 70483 + }, + { + "epoch": 5.21, + "learning_rate": 8.993317230834375e-07, + "loss": 0.9868, + "step": 70484 + }, + { + "epoch": 5.21, + "learning_rate": 8.991664133766242e-07, + "loss": 0.9884, + "step": 70485 + }, + { + "epoch": 5.21, + "learning_rate": 8.990011181491065e-07, + "loss": 1.0635, + "step": 70486 + }, + { + "epoch": 5.21, + "learning_rate": 8.988358374011475e-07, + "loss": 0.9706, + "step": 70487 + }, + { + "epoch": 5.21, + "learning_rate": 8.98670571133009e-07, + "loss": 0.9968, + "step": 70488 + }, + { + "epoch": 5.21, + "learning_rate": 8.985053193449567e-07, + "loss": 0.9838, + "step": 70489 + }, + { + "epoch": 5.21, + "learning_rate": 8.983400820372534e-07, + "loss": 0.9405, + "step": 70490 + }, + { + "epoch": 5.21, + "learning_rate": 8.981748592101591e-07, + "loss": 1.0336, + "step": 70491 + }, + { + "epoch": 5.21, + "learning_rate": 8.98009650863938e-07, + "loss": 0.9857, + "step": 70492 + }, + { + "epoch": 5.21, + "learning_rate": 8.978444569988543e-07, + "loss": 1.0053, + "step": 70493 + }, + { + "epoch": 5.21, + "learning_rate": 8.97679277615171e-07, + "loss": 0.9922, + "step": 70494 + }, + { + "epoch": 5.21, + "learning_rate": 8.975141127131481e-07, + "loss": 0.9822, + "step": 70495 + }, + { + "epoch": 5.21, + "learning_rate": 8.973489622930509e-07, + "loss": 0.9908, + "step": 70496 + }, + { + "epoch": 5.21, + "learning_rate": 8.971838263551402e-07, + "loss": 1.0277, + "step": 70497 + }, + { + "epoch": 5.21, + "learning_rate": 8.970187048996825e-07, + "loss": 0.9992, + "step": 70498 + }, + { + "epoch": 5.21, + "learning_rate": 8.968535979269344e-07, + "loss": 0.9122, + "step": 70499 + }, + { + "epoch": 5.21, + "learning_rate": 8.966885054371633e-07, + "loss": 0.9644, + "step": 70500 + }, + { + "epoch": 5.21, + "learning_rate": 8.965234274306289e-07, + "loss": 0.9729, + "step": 70501 + }, + { + "epoch": 5.21, + "learning_rate": 8.96358363907599e-07, + "loss": 0.8971, + "step": 70502 + }, + { + "epoch": 5.21, + "learning_rate": 8.9619331486833e-07, + "loss": 1.0985, + "step": 70503 + }, + { + "epoch": 5.21, + "learning_rate": 8.960282803130882e-07, + "loss": 0.9661, + "step": 70504 + }, + { + "epoch": 5.21, + "learning_rate": 8.958632602421358e-07, + "loss": 1.0321, + "step": 70505 + }, + { + "epoch": 5.21, + "learning_rate": 8.956982546557336e-07, + "loss": 1.0657, + "step": 70506 + }, + { + "epoch": 5.21, + "learning_rate": 8.955332635541458e-07, + "loss": 0.9887, + "step": 70507 + }, + { + "epoch": 5.21, + "learning_rate": 8.953682869376323e-07, + "loss": 0.9673, + "step": 70508 + }, + { + "epoch": 5.21, + "learning_rate": 8.952033248064595e-07, + "loss": 1.0135, + "step": 70509 + }, + { + "epoch": 5.21, + "learning_rate": 8.950383771608884e-07, + "loss": 1.0886, + "step": 70510 + }, + { + "epoch": 5.21, + "learning_rate": 8.948734440011808e-07, + "loss": 0.967, + "step": 70511 + }, + { + "epoch": 5.21, + "learning_rate": 8.947085253275978e-07, + "loss": 0.9762, + "step": 70512 + }, + { + "epoch": 5.21, + "learning_rate": 8.945436211404046e-07, + "loss": 0.9551, + "step": 70513 + }, + { + "epoch": 5.21, + "learning_rate": 8.943787314398622e-07, + "loss": 1.0082, + "step": 70514 + }, + { + "epoch": 5.21, + "learning_rate": 8.942138562262326e-07, + "loss": 1.0141, + "step": 70515 + }, + { + "epoch": 5.21, + "learning_rate": 8.940489954997789e-07, + "loss": 0.9401, + "step": 70516 + }, + { + "epoch": 5.21, + "learning_rate": 8.938841492607608e-07, + "loss": 0.9988, + "step": 70517 + }, + { + "epoch": 5.21, + "learning_rate": 8.93719317509445e-07, + "loss": 1.0069, + "step": 70518 + }, + { + "epoch": 5.21, + "learning_rate": 8.93554500246091e-07, + "loss": 0.9761, + "step": 70519 + }, + { + "epoch": 5.21, + "learning_rate": 8.933896974709622e-07, + "loss": 1.0127, + "step": 70520 + }, + { + "epoch": 5.21, + "learning_rate": 8.932249091843181e-07, + "loss": 1.0069, + "step": 70521 + }, + { + "epoch": 5.21, + "learning_rate": 8.930601353864244e-07, + "loss": 0.979, + "step": 70522 + }, + { + "epoch": 5.21, + "learning_rate": 8.928953760775416e-07, + "loss": 1.0531, + "step": 70523 + }, + { + "epoch": 5.21, + "learning_rate": 8.92730631257932e-07, + "loss": 0.966, + "step": 70524 + }, + { + "epoch": 5.21, + "learning_rate": 8.925659009278576e-07, + "loss": 0.9214, + "step": 70525 + }, + { + "epoch": 5.21, + "learning_rate": 8.924011850875792e-07, + "loss": 0.9847, + "step": 70526 + }, + { + "epoch": 5.21, + "learning_rate": 8.922364837373631e-07, + "loss": 0.9374, + "step": 70527 + }, + { + "epoch": 5.21, + "learning_rate": 8.920717968774662e-07, + "loss": 0.9328, + "step": 70528 + }, + { + "epoch": 5.21, + "learning_rate": 8.919071245081534e-07, + "loss": 0.9259, + "step": 70529 + }, + { + "epoch": 5.21, + "learning_rate": 8.917424666296858e-07, + "loss": 1.1007, + "step": 70530 + }, + { + "epoch": 5.21, + "learning_rate": 8.915778232423267e-07, + "loss": 1.0091, + "step": 70531 + }, + { + "epoch": 5.21, + "learning_rate": 8.914131943463345e-07, + "loss": 1.0647, + "step": 70532 + }, + { + "epoch": 5.21, + "learning_rate": 8.912485799419757e-07, + "loss": 0.9981, + "step": 70533 + }, + { + "epoch": 5.21, + "learning_rate": 8.910839800295101e-07, + "loss": 0.9685, + "step": 70534 + }, + { + "epoch": 5.21, + "learning_rate": 8.909193946091988e-07, + "loss": 0.9441, + "step": 70535 + }, + { + "epoch": 5.21, + "learning_rate": 8.907548236813046e-07, + "loss": 1.0209, + "step": 70536 + }, + { + "epoch": 5.21, + "learning_rate": 8.905902672460876e-07, + "loss": 1.055, + "step": 70537 + }, + { + "epoch": 5.21, + "learning_rate": 8.90425725303814e-07, + "loss": 1.0184, + "step": 70538 + }, + { + "epoch": 5.21, + "learning_rate": 8.902611978547415e-07, + "loss": 0.9823, + "step": 70539 + }, + { + "epoch": 5.21, + "learning_rate": 8.900966848991333e-07, + "loss": 1.0049, + "step": 70540 + }, + { + "epoch": 5.21, + "learning_rate": 8.89932186437249e-07, + "loss": 0.9978, + "step": 70541 + }, + { + "epoch": 5.21, + "learning_rate": 8.897677024693541e-07, + "loss": 1.0003, + "step": 70542 + }, + { + "epoch": 5.21, + "learning_rate": 8.896032329957105e-07, + "loss": 0.957, + "step": 70543 + }, + { + "epoch": 5.21, + "learning_rate": 8.894387780165736e-07, + "loss": 0.9454, + "step": 70544 + }, + { + "epoch": 5.21, + "learning_rate": 8.892743375322122e-07, + "loss": 0.9879, + "step": 70545 + }, + { + "epoch": 5.21, + "learning_rate": 8.891099115428825e-07, + "loss": 1.0145, + "step": 70546 + }, + { + "epoch": 5.21, + "learning_rate": 8.889455000488523e-07, + "loss": 0.9846, + "step": 70547 + }, + { + "epoch": 5.21, + "learning_rate": 8.887811030503757e-07, + "loss": 1.0301, + "step": 70548 + }, + { + "epoch": 5.21, + "learning_rate": 8.886167205477192e-07, + "loss": 1.0634, + "step": 70549 + }, + { + "epoch": 5.21, + "learning_rate": 8.884523525411426e-07, + "loss": 1.0201, + "step": 70550 + }, + { + "epoch": 5.21, + "learning_rate": 8.882879990309112e-07, + "loss": 1.003, + "step": 70551 + }, + { + "epoch": 5.21, + "learning_rate": 8.881236600172804e-07, + "loss": 0.8862, + "step": 70552 + }, + { + "epoch": 5.21, + "learning_rate": 8.879593355005133e-07, + "loss": 1.0678, + "step": 70553 + }, + { + "epoch": 5.21, + "learning_rate": 8.877950254808754e-07, + "loss": 1.07, + "step": 70554 + }, + { + "epoch": 5.21, + "learning_rate": 8.87630729958624e-07, + "loss": 1.0892, + "step": 70555 + }, + { + "epoch": 5.21, + "learning_rate": 8.874664489340223e-07, + "loss": 0.9742, + "step": 70556 + }, + { + "epoch": 5.21, + "learning_rate": 8.873021824073291e-07, + "loss": 0.9526, + "step": 70557 + }, + { + "epoch": 5.21, + "learning_rate": 8.871379303788096e-07, + "loss": 0.9984, + "step": 70558 + }, + { + "epoch": 5.21, + "learning_rate": 8.869736928487227e-07, + "loss": 0.9757, + "step": 70559 + }, + { + "epoch": 5.21, + "learning_rate": 8.868094698173313e-07, + "loss": 0.8552, + "step": 70560 + }, + { + "epoch": 5.21, + "learning_rate": 8.866452612848941e-07, + "loss": 1.0626, + "step": 70561 + }, + { + "epoch": 5.21, + "learning_rate": 8.864810672516732e-07, + "loss": 1.0288, + "step": 70562 + }, + { + "epoch": 5.21, + "learning_rate": 8.863168877179329e-07, + "loss": 1.0315, + "step": 70563 + }, + { + "epoch": 5.21, + "learning_rate": 8.861527226839306e-07, + "loss": 0.999, + "step": 70564 + }, + { + "epoch": 5.21, + "learning_rate": 8.859885721499295e-07, + "loss": 1.0132, + "step": 70565 + }, + { + "epoch": 5.21, + "learning_rate": 8.858244361161894e-07, + "loss": 1.045, + "step": 70566 + }, + { + "epoch": 5.21, + "learning_rate": 8.856603145829723e-07, + "loss": 1.0061, + "step": 70567 + }, + { + "epoch": 5.21, + "learning_rate": 8.854962075505402e-07, + "loss": 1.0142, + "step": 70568 + }, + { + "epoch": 5.21, + "learning_rate": 8.853321150191541e-07, + "loss": 0.9714, + "step": 70569 + }, + { + "epoch": 5.21, + "learning_rate": 8.851680369890725e-07, + "loss": 0.971, + "step": 70570 + }, + { + "epoch": 5.21, + "learning_rate": 8.850039734605576e-07, + "loss": 0.8604, + "step": 70571 + }, + { + "epoch": 5.21, + "learning_rate": 8.848399244338734e-07, + "loss": 0.9482, + "step": 70572 + }, + { + "epoch": 5.21, + "learning_rate": 8.846758899092755e-07, + "loss": 1.0374, + "step": 70573 + }, + { + "epoch": 5.21, + "learning_rate": 8.845118698870303e-07, + "loss": 1.006, + "step": 70574 + }, + { + "epoch": 5.21, + "learning_rate": 8.843478643673953e-07, + "loss": 1.0225, + "step": 70575 + }, + { + "epoch": 5.21, + "learning_rate": 8.841838733506325e-07, + "loss": 1.0748, + "step": 70576 + }, + { + "epoch": 5.21, + "learning_rate": 8.840198968370017e-07, + "loss": 1.0564, + "step": 70577 + }, + { + "epoch": 5.21, + "learning_rate": 8.838559348267661e-07, + "loss": 0.9603, + "step": 70578 + }, + { + "epoch": 5.22, + "learning_rate": 8.836919873201843e-07, + "loss": 0.959, + "step": 70579 + }, + { + "epoch": 5.22, + "learning_rate": 8.835280543175195e-07, + "loss": 1.019, + "step": 70580 + }, + { + "epoch": 5.22, + "learning_rate": 8.833641358190304e-07, + "loss": 0.9771, + "step": 70581 + }, + { + "epoch": 5.22, + "learning_rate": 8.832002318249778e-07, + "loss": 1.0255, + "step": 70582 + }, + { + "epoch": 5.22, + "learning_rate": 8.830363423356236e-07, + "loss": 1.0163, + "step": 70583 + }, + { + "epoch": 5.22, + "learning_rate": 8.82872467351229e-07, + "loss": 0.9694, + "step": 70584 + }, + { + "epoch": 5.22, + "learning_rate": 8.827086068720536e-07, + "loss": 1.0826, + "step": 70585 + }, + { + "epoch": 5.22, + "learning_rate": 8.825447608983561e-07, + "loss": 1.0319, + "step": 70586 + }, + { + "epoch": 5.22, + "learning_rate": 8.82380929430402e-07, + "loss": 0.9907, + "step": 70587 + }, + { + "epoch": 5.22, + "learning_rate": 8.822171124684498e-07, + "loss": 1.0465, + "step": 70588 + }, + { + "epoch": 5.22, + "learning_rate": 8.820533100127581e-07, + "loss": 1.0233, + "step": 70589 + }, + { + "epoch": 5.22, + "learning_rate": 8.818895220635903e-07, + "loss": 1.0006, + "step": 70590 + }, + { + "epoch": 5.22, + "learning_rate": 8.817257486212039e-07, + "loss": 1.0415, + "step": 70591 + }, + { + "epoch": 5.22, + "learning_rate": 8.815619896858652e-07, + "loss": 1.0554, + "step": 70592 + }, + { + "epoch": 5.22, + "learning_rate": 8.813982452578263e-07, + "loss": 1.1384, + "step": 70593 + }, + { + "epoch": 5.22, + "learning_rate": 8.812345153373547e-07, + "loss": 0.9333, + "step": 70594 + }, + { + "epoch": 5.22, + "learning_rate": 8.81070799924707e-07, + "loss": 0.9351, + "step": 70595 + }, + { + "epoch": 5.22, + "learning_rate": 8.809070990201485e-07, + "loss": 0.9974, + "step": 70596 + }, + { + "epoch": 5.22, + "learning_rate": 8.807434126239334e-07, + "loss": 1.0223, + "step": 70597 + }, + { + "epoch": 5.22, + "learning_rate": 8.805797407363259e-07, + "loss": 0.9607, + "step": 70598 + }, + { + "epoch": 5.22, + "learning_rate": 8.804160833575858e-07, + "loss": 1.0124, + "step": 70599 + }, + { + "epoch": 5.22, + "learning_rate": 8.802524404879742e-07, + "loss": 0.9207, + "step": 70600 + }, + { + "epoch": 5.22, + "learning_rate": 8.800888121277495e-07, + "loss": 1.0329, + "step": 70601 + }, + { + "epoch": 5.22, + "learning_rate": 8.799251982771706e-07, + "loss": 1.0433, + "step": 70602 + }, + { + "epoch": 5.22, + "learning_rate": 8.797615989365027e-07, + "loss": 0.9252, + "step": 70603 + }, + { + "epoch": 5.22, + "learning_rate": 8.795980141060045e-07, + "loss": 0.9006, + "step": 70604 + }, + { + "epoch": 5.22, + "learning_rate": 8.794344437859348e-07, + "loss": 0.893, + "step": 70605 + }, + { + "epoch": 5.22, + "learning_rate": 8.792708879765522e-07, + "loss": 1.1535, + "step": 70606 + }, + { + "epoch": 5.22, + "learning_rate": 8.791073466781209e-07, + "loss": 1.021, + "step": 70607 + }, + { + "epoch": 5.22, + "learning_rate": 8.789438198908995e-07, + "loss": 1.0181, + "step": 70608 + }, + { + "epoch": 5.22, + "learning_rate": 8.78780307615148e-07, + "loss": 0.8934, + "step": 70609 + }, + { + "epoch": 5.22, + "learning_rate": 8.786168098511261e-07, + "loss": 0.995, + "step": 70610 + }, + { + "epoch": 5.22, + "learning_rate": 8.784533265990935e-07, + "loss": 0.8978, + "step": 70611 + }, + { + "epoch": 5.22, + "learning_rate": 8.782898578593125e-07, + "loss": 0.9951, + "step": 70612 + }, + { + "epoch": 5.22, + "learning_rate": 8.781264036320413e-07, + "loss": 1.0147, + "step": 70613 + }, + { + "epoch": 5.22, + "learning_rate": 8.779629639175414e-07, + "loss": 0.9909, + "step": 70614 + }, + { + "epoch": 5.22, + "learning_rate": 8.777995387160687e-07, + "loss": 0.9463, + "step": 70615 + }, + { + "epoch": 5.22, + "learning_rate": 8.7763612802789e-07, + "loss": 0.996, + "step": 70616 + }, + { + "epoch": 5.22, + "learning_rate": 8.774727318532628e-07, + "loss": 0.9993, + "step": 70617 + }, + { + "epoch": 5.22, + "learning_rate": 8.773093501924412e-07, + "loss": 1.0098, + "step": 70618 + }, + { + "epoch": 5.22, + "learning_rate": 8.77145983045693e-07, + "loss": 1.0214, + "step": 70619 + }, + { + "epoch": 5.22, + "learning_rate": 8.769826304132745e-07, + "loss": 0.9283, + "step": 70620 + }, + { + "epoch": 5.22, + "learning_rate": 8.768192922954466e-07, + "loss": 0.9575, + "step": 70621 + }, + { + "epoch": 5.22, + "learning_rate": 8.76655968692467e-07, + "loss": 1.0259, + "step": 70622 + }, + { + "epoch": 5.22, + "learning_rate": 8.764926596045986e-07, + "loss": 0.891, + "step": 70623 + }, + { + "epoch": 5.22, + "learning_rate": 8.763293650320993e-07, + "loss": 0.9814, + "step": 70624 + }, + { + "epoch": 5.22, + "learning_rate": 8.761660849752307e-07, + "loss": 1.0087, + "step": 70625 + }, + { + "epoch": 5.22, + "learning_rate": 8.760028194342496e-07, + "loss": 1.0437, + "step": 70626 + }, + { + "epoch": 5.22, + "learning_rate": 8.758395684094168e-07, + "loss": 0.9585, + "step": 70627 + }, + { + "epoch": 5.22, + "learning_rate": 8.756763319009953e-07, + "loss": 0.9822, + "step": 70628 + }, + { + "epoch": 5.22, + "learning_rate": 8.755131099092407e-07, + "loss": 1.0994, + "step": 70629 + }, + { + "epoch": 5.22, + "learning_rate": 8.753499024344148e-07, + "loss": 0.9126, + "step": 70630 + }, + { + "epoch": 5.22, + "learning_rate": 8.751867094767752e-07, + "loss": 1.0729, + "step": 70631 + }, + { + "epoch": 5.22, + "learning_rate": 8.75023531036584e-07, + "loss": 0.9666, + "step": 70632 + }, + { + "epoch": 5.22, + "learning_rate": 8.748603671141009e-07, + "loss": 1.0429, + "step": 70633 + }, + { + "epoch": 5.22, + "learning_rate": 8.746972177095836e-07, + "loss": 0.9717, + "step": 70634 + }, + { + "epoch": 5.22, + "learning_rate": 8.745340828232906e-07, + "loss": 0.9816, + "step": 70635 + }, + { + "epoch": 5.22, + "learning_rate": 8.743709624554853e-07, + "loss": 1.0995, + "step": 70636 + }, + { + "epoch": 5.22, + "learning_rate": 8.742078566064271e-07, + "loss": 1.0481, + "step": 70637 + }, + { + "epoch": 5.22, + "learning_rate": 8.740447652763695e-07, + "loss": 1.0802, + "step": 70638 + }, + { + "epoch": 5.22, + "learning_rate": 8.738816884655787e-07, + "loss": 1.0304, + "step": 70639 + }, + { + "epoch": 5.22, + "learning_rate": 8.737186261743091e-07, + "loss": 0.9265, + "step": 70640 + }, + { + "epoch": 5.22, + "learning_rate": 8.735555784028272e-07, + "loss": 0.991, + "step": 70641 + }, + { + "epoch": 5.22, + "learning_rate": 8.733925451513836e-07, + "loss": 1.1061, + "step": 70642 + }, + { + "epoch": 5.22, + "learning_rate": 8.732295264202428e-07, + "loss": 0.9798, + "step": 70643 + }, + { + "epoch": 5.22, + "learning_rate": 8.730665222096624e-07, + "loss": 0.9195, + "step": 70644 + }, + { + "epoch": 5.22, + "learning_rate": 8.729035325199064e-07, + "loss": 0.9753, + "step": 70645 + }, + { + "epoch": 5.22, + "learning_rate": 8.727405573512282e-07, + "loss": 1.0392, + "step": 70646 + }, + { + "epoch": 5.22, + "learning_rate": 8.725775967038874e-07, + "loss": 1.0118, + "step": 70647 + }, + { + "epoch": 5.22, + "learning_rate": 8.724146505781461e-07, + "loss": 0.9745, + "step": 70648 + }, + { + "epoch": 5.22, + "learning_rate": 8.72251718974263e-07, + "loss": 0.9365, + "step": 70649 + }, + { + "epoch": 5.22, + "learning_rate": 8.720888018924978e-07, + "loss": 1.0161, + "step": 70650 + }, + { + "epoch": 5.22, + "learning_rate": 8.719258993331059e-07, + "loss": 0.9476, + "step": 70651 + }, + { + "epoch": 5.22, + "learning_rate": 8.717630112963516e-07, + "loss": 0.9473, + "step": 70652 + }, + { + "epoch": 5.22, + "learning_rate": 8.716001377824912e-07, + "loss": 0.9815, + "step": 70653 + }, + { + "epoch": 5.22, + "learning_rate": 8.714372787917846e-07, + "loss": 1.0701, + "step": 70654 + }, + { + "epoch": 5.22, + "learning_rate": 8.712744343244905e-07, + "loss": 1.0502, + "step": 70655 + }, + { + "epoch": 5.22, + "learning_rate": 8.711116043808665e-07, + "loss": 1.0044, + "step": 70656 + }, + { + "epoch": 5.22, + "learning_rate": 8.709487889611757e-07, + "loss": 0.9858, + "step": 70657 + }, + { + "epoch": 5.22, + "learning_rate": 8.707859880656733e-07, + "loss": 0.9757, + "step": 70658 + }, + { + "epoch": 5.22, + "learning_rate": 8.706232016946214e-07, + "loss": 1.0841, + "step": 70659 + }, + { + "epoch": 5.22, + "learning_rate": 8.704604298482744e-07, + "loss": 0.8574, + "step": 70660 + }, + { + "epoch": 5.22, + "learning_rate": 8.702976725268985e-07, + "loss": 1.0251, + "step": 70661 + }, + { + "epoch": 5.22, + "learning_rate": 8.701349297307449e-07, + "loss": 1.01, + "step": 70662 + }, + { + "epoch": 5.22, + "learning_rate": 8.699722014600776e-07, + "loss": 0.9471, + "step": 70663 + }, + { + "epoch": 5.22, + "learning_rate": 8.69809487715153e-07, + "loss": 0.9946, + "step": 70664 + }, + { + "epoch": 5.22, + "learning_rate": 8.696467884962323e-07, + "loss": 0.9481, + "step": 70665 + }, + { + "epoch": 5.22, + "learning_rate": 8.694841038035717e-07, + "loss": 1.0258, + "step": 70666 + }, + { + "epoch": 5.22, + "learning_rate": 8.6932143363743e-07, + "loss": 1.0986, + "step": 70667 + }, + { + "epoch": 5.22, + "learning_rate": 8.691587779980692e-07, + "loss": 1.0025, + "step": 70668 + }, + { + "epoch": 5.22, + "learning_rate": 8.689961368857458e-07, + "loss": 1.0177, + "step": 70669 + }, + { + "epoch": 5.22, + "learning_rate": 8.688335103007195e-07, + "loss": 0.9861, + "step": 70670 + }, + { + "epoch": 5.22, + "learning_rate": 8.686708982432468e-07, + "loss": 1.13, + "step": 70671 + }, + { + "epoch": 5.22, + "learning_rate": 8.685083007135886e-07, + "loss": 1.1287, + "step": 70672 + }, + { + "epoch": 5.22, + "learning_rate": 8.683457177120047e-07, + "loss": 1.0159, + "step": 70673 + }, + { + "epoch": 5.22, + "learning_rate": 8.681831492387505e-07, + "loss": 1.089, + "step": 70674 + }, + { + "epoch": 5.22, + "learning_rate": 8.680205952940868e-07, + "loss": 1.0739, + "step": 70675 + }, + { + "epoch": 5.22, + "learning_rate": 8.678580558782701e-07, + "loss": 0.9892, + "step": 70676 + }, + { + "epoch": 5.22, + "learning_rate": 8.676955309915624e-07, + "loss": 0.9745, + "step": 70677 + }, + { + "epoch": 5.22, + "learning_rate": 8.675330206342202e-07, + "loss": 1.0524, + "step": 70678 + }, + { + "epoch": 5.22, + "learning_rate": 8.673705248065023e-07, + "loss": 0.9009, + "step": 70679 + }, + { + "epoch": 5.22, + "learning_rate": 8.67208043508666e-07, + "loss": 1.0202, + "step": 70680 + }, + { + "epoch": 5.22, + "learning_rate": 8.670455767409724e-07, + "loss": 0.9475, + "step": 70681 + }, + { + "epoch": 5.22, + "learning_rate": 8.66883124503679e-07, + "loss": 1.0933, + "step": 70682 + }, + { + "epoch": 5.22, + "learning_rate": 8.667206867970435e-07, + "loss": 1.0977, + "step": 70683 + }, + { + "epoch": 5.22, + "learning_rate": 8.665582636213243e-07, + "loss": 0.9762, + "step": 70684 + }, + { + "epoch": 5.22, + "learning_rate": 8.663958549767792e-07, + "loss": 0.9825, + "step": 70685 + }, + { + "epoch": 5.22, + "learning_rate": 8.662334608636703e-07, + "loss": 1.0271, + "step": 70686 + }, + { + "epoch": 5.22, + "learning_rate": 8.660710812822504e-07, + "loss": 1.0597, + "step": 70687 + }, + { + "epoch": 5.22, + "learning_rate": 8.659087162327829e-07, + "loss": 0.9205, + "step": 70688 + }, + { + "epoch": 5.22, + "learning_rate": 8.657463657155207e-07, + "loss": 1.1251, + "step": 70689 + }, + { + "epoch": 5.22, + "learning_rate": 8.655840297307306e-07, + "loss": 0.8827, + "step": 70690 + }, + { + "epoch": 5.22, + "learning_rate": 8.654217082786609e-07, + "loss": 0.9612, + "step": 70691 + }, + { + "epoch": 5.22, + "learning_rate": 8.652594013595772e-07, + "loss": 1.0464, + "step": 70692 + }, + { + "epoch": 5.22, + "learning_rate": 8.650971089737337e-07, + "loss": 0.9099, + "step": 70693 + }, + { + "epoch": 5.22, + "learning_rate": 8.649348311213912e-07, + "loss": 1.0184, + "step": 70694 + }, + { + "epoch": 5.22, + "learning_rate": 8.647725678028063e-07, + "loss": 0.9521, + "step": 70695 + }, + { + "epoch": 5.22, + "learning_rate": 8.646103190182364e-07, + "loss": 1.0813, + "step": 70696 + }, + { + "epoch": 5.22, + "learning_rate": 8.644480847679415e-07, + "loss": 1.0295, + "step": 70697 + }, + { + "epoch": 5.22, + "learning_rate": 8.642858650521802e-07, + "loss": 1.0363, + "step": 70698 + }, + { + "epoch": 5.22, + "learning_rate": 8.641236598712088e-07, + "loss": 0.9596, + "step": 70699 + }, + { + "epoch": 5.22, + "learning_rate": 8.63961469225284e-07, + "loss": 0.9114, + "step": 70700 + }, + { + "epoch": 5.22, + "learning_rate": 8.637992931146677e-07, + "loss": 0.9782, + "step": 70701 + }, + { + "epoch": 5.22, + "learning_rate": 8.636371315396152e-07, + "loss": 0.9664, + "step": 70702 + }, + { + "epoch": 5.22, + "learning_rate": 8.634749845003865e-07, + "loss": 0.8793, + "step": 70703 + }, + { + "epoch": 5.22, + "learning_rate": 8.633128519972378e-07, + "loss": 0.9745, + "step": 70704 + }, + { + "epoch": 5.22, + "learning_rate": 8.631507340304257e-07, + "loss": 0.8955, + "step": 70705 + }, + { + "epoch": 5.22, + "learning_rate": 8.629886306002144e-07, + "loss": 1.0543, + "step": 70706 + }, + { + "epoch": 5.22, + "learning_rate": 8.628265417068527e-07, + "loss": 0.9511, + "step": 70707 + }, + { + "epoch": 5.22, + "learning_rate": 8.626644673506058e-07, + "loss": 0.9566, + "step": 70708 + }, + { + "epoch": 5.22, + "learning_rate": 8.625024075317268e-07, + "loss": 1.0167, + "step": 70709 + }, + { + "epoch": 5.22, + "learning_rate": 8.6234036225048e-07, + "loss": 1.0184, + "step": 70710 + }, + { + "epoch": 5.22, + "learning_rate": 8.621783315071163e-07, + "loss": 1.0627, + "step": 70711 + }, + { + "epoch": 5.22, + "learning_rate": 8.620163153018956e-07, + "loss": 1.0344, + "step": 70712 + }, + { + "epoch": 5.22, + "learning_rate": 8.618543136350765e-07, + "loss": 0.9184, + "step": 70713 + }, + { + "epoch": 5.23, + "learning_rate": 8.616923265069177e-07, + "loss": 1.0024, + "step": 70714 + }, + { + "epoch": 5.23, + "learning_rate": 8.615303539176745e-07, + "loss": 0.988, + "step": 70715 + }, + { + "epoch": 5.23, + "learning_rate": 8.613683958676055e-07, + "loss": 0.9136, + "step": 70716 + }, + { + "epoch": 5.23, + "learning_rate": 8.612064523569685e-07, + "loss": 1.0135, + "step": 70717 + }, + { + "epoch": 5.23, + "learning_rate": 8.610445233860232e-07, + "loss": 1.1921, + "step": 70718 + }, + { + "epoch": 5.23, + "learning_rate": 8.608826089550237e-07, + "loss": 0.9367, + "step": 70719 + }, + { + "epoch": 5.23, + "learning_rate": 8.6072070906423e-07, + "loss": 1.0449, + "step": 70720 + }, + { + "epoch": 5.23, + "learning_rate": 8.605588237138974e-07, + "loss": 1.0305, + "step": 70721 + }, + { + "epoch": 5.23, + "learning_rate": 8.603969529042866e-07, + "loss": 1.016, + "step": 70722 + }, + { + "epoch": 5.23, + "learning_rate": 8.602350966356532e-07, + "loss": 0.9536, + "step": 70723 + }, + { + "epoch": 5.23, + "learning_rate": 8.600732549082547e-07, + "loss": 1.1023, + "step": 70724 + }, + { + "epoch": 5.23, + "learning_rate": 8.599114277223474e-07, + "loss": 1.0026, + "step": 70725 + }, + { + "epoch": 5.23, + "learning_rate": 8.597496150781914e-07, + "loss": 1.0403, + "step": 70726 + }, + { + "epoch": 5.23, + "learning_rate": 8.595878169760441e-07, + "loss": 0.8773, + "step": 70727 + }, + { + "epoch": 5.23, + "learning_rate": 8.594260334161608e-07, + "loss": 1.0114, + "step": 70728 + }, + { + "epoch": 5.23, + "learning_rate": 8.592642643988014e-07, + "loss": 1.0811, + "step": 70729 + }, + { + "epoch": 5.23, + "learning_rate": 8.59102509924219e-07, + "loss": 0.9889, + "step": 70730 + }, + { + "epoch": 5.23, + "learning_rate": 8.589407699926766e-07, + "loss": 1.0146, + "step": 70731 + }, + { + "epoch": 5.23, + "learning_rate": 8.587790446044264e-07, + "loss": 0.9619, + "step": 70732 + }, + { + "epoch": 5.23, + "learning_rate": 8.586173337597293e-07, + "loss": 1.0442, + "step": 70733 + }, + { + "epoch": 5.23, + "learning_rate": 8.584556374588382e-07, + "loss": 0.9553, + "step": 70734 + }, + { + "epoch": 5.23, + "learning_rate": 8.582939557020187e-07, + "loss": 0.9482, + "step": 70735 + }, + { + "epoch": 5.23, + "learning_rate": 8.581322884895182e-07, + "loss": 0.9146, + "step": 70736 + }, + { + "epoch": 5.23, + "learning_rate": 8.579706358216e-07, + "loss": 1.0601, + "step": 70737 + }, + { + "epoch": 5.23, + "learning_rate": 8.578089976985194e-07, + "loss": 1.0328, + "step": 70738 + }, + { + "epoch": 5.23, + "learning_rate": 8.57647374120536e-07, + "loss": 0.9747, + "step": 70739 + }, + { + "epoch": 5.23, + "learning_rate": 8.574857650879031e-07, + "loss": 1.0209, + "step": 70740 + }, + { + "epoch": 5.23, + "learning_rate": 8.573241706008783e-07, + "loss": 1.0531, + "step": 70741 + }, + { + "epoch": 5.23, + "learning_rate": 8.571625906597214e-07, + "loss": 0.9509, + "step": 70742 + }, + { + "epoch": 5.23, + "learning_rate": 8.570010252646887e-07, + "loss": 1.0339, + "step": 70743 + }, + { + "epoch": 5.23, + "learning_rate": 8.568394744160357e-07, + "loss": 1.1815, + "step": 70744 + }, + { + "epoch": 5.23, + "learning_rate": 8.566779381140189e-07, + "loss": 1.1082, + "step": 70745 + }, + { + "epoch": 5.23, + "learning_rate": 8.56516416358899e-07, + "loss": 0.9718, + "step": 70746 + }, + { + "epoch": 5.23, + "learning_rate": 8.563549091509293e-07, + "loss": 1.0381, + "step": 70747 + }, + { + "epoch": 5.23, + "learning_rate": 8.561934164903696e-07, + "loss": 0.9922, + "step": 70748 + }, + { + "epoch": 5.23, + "learning_rate": 8.56031938377474e-07, + "loss": 1.0575, + "step": 70749 + }, + { + "epoch": 5.23, + "learning_rate": 8.558704748125002e-07, + "loss": 1.0029, + "step": 70750 + }, + { + "epoch": 5.23, + "learning_rate": 8.557090257957079e-07, + "loss": 1.0187, + "step": 70751 + }, + { + "epoch": 5.23, + "learning_rate": 8.555475913273492e-07, + "loss": 0.9558, + "step": 70752 + }, + { + "epoch": 5.23, + "learning_rate": 8.55386171407685e-07, + "loss": 0.9071, + "step": 70753 + }, + { + "epoch": 5.23, + "learning_rate": 8.552247660369684e-07, + "loss": 1.0961, + "step": 70754 + }, + { + "epoch": 5.23, + "learning_rate": 8.550633752154624e-07, + "loss": 1.1112, + "step": 70755 + }, + { + "epoch": 5.23, + "learning_rate": 8.54901998943416e-07, + "loss": 1.1227, + "step": 70756 + }, + { + "epoch": 5.23, + "learning_rate": 8.547406372210909e-07, + "loss": 0.9945, + "step": 70757 + }, + { + "epoch": 5.23, + "learning_rate": 8.545792900487438e-07, + "loss": 1.0757, + "step": 70758 + }, + { + "epoch": 5.23, + "learning_rate": 8.544179574266287e-07, + "loss": 1.0311, + "step": 70759 + }, + { + "epoch": 5.23, + "learning_rate": 8.542566393550045e-07, + "loss": 0.9224, + "step": 70760 + }, + { + "epoch": 5.23, + "learning_rate": 8.540953358341264e-07, + "loss": 0.9389, + "step": 70761 + }, + { + "epoch": 5.23, + "learning_rate": 8.539340468642521e-07, + "loss": 0.9767, + "step": 70762 + }, + { + "epoch": 5.23, + "learning_rate": 8.537727724456379e-07, + "loss": 1.0198, + "step": 70763 + }, + { + "epoch": 5.23, + "learning_rate": 8.536115125785416e-07, + "loss": 1.02, + "step": 70764 + }, + { + "epoch": 5.23, + "learning_rate": 8.53450267263215e-07, + "loss": 1.0692, + "step": 70765 + }, + { + "epoch": 5.23, + "learning_rate": 8.532890364999214e-07, + "loss": 1.0539, + "step": 70766 + }, + { + "epoch": 5.23, + "learning_rate": 8.531278202889137e-07, + "loss": 1.0618, + "step": 70767 + }, + { + "epoch": 5.23, + "learning_rate": 8.529666186304486e-07, + "loss": 1.0799, + "step": 70768 + }, + { + "epoch": 5.23, + "learning_rate": 8.528054315247825e-07, + "loss": 1.0229, + "step": 70769 + }, + { + "epoch": 5.23, + "learning_rate": 8.526442589721707e-07, + "loss": 1.0489, + "step": 70770 + }, + { + "epoch": 5.23, + "learning_rate": 8.524831009728729e-07, + "loss": 1.0648, + "step": 70771 + }, + { + "epoch": 5.23, + "learning_rate": 8.523219575271435e-07, + "loss": 1.0394, + "step": 70772 + }, + { + "epoch": 5.23, + "learning_rate": 8.52160828635239e-07, + "loss": 0.981, + "step": 70773 + }, + { + "epoch": 5.23, + "learning_rate": 8.519997142974135e-07, + "loss": 0.9399, + "step": 70774 + }, + { + "epoch": 5.23, + "learning_rate": 8.51838614513928e-07, + "loss": 0.9963, + "step": 70775 + }, + { + "epoch": 5.23, + "learning_rate": 8.516775292850376e-07, + "loss": 1.0678, + "step": 70776 + }, + { + "epoch": 5.23, + "learning_rate": 8.515164586109937e-07, + "loss": 1.003, + "step": 70777 + }, + { + "epoch": 5.23, + "learning_rate": 8.513554024920579e-07, + "loss": 0.8638, + "step": 70778 + }, + { + "epoch": 5.23, + "learning_rate": 8.511943609284834e-07, + "loss": 1.025, + "step": 70779 + }, + { + "epoch": 5.23, + "learning_rate": 8.510333339205313e-07, + "loss": 1.1285, + "step": 70780 + }, + { + "epoch": 5.23, + "learning_rate": 8.508723214684511e-07, + "loss": 1.0597, + "step": 70781 + }, + { + "epoch": 5.23, + "learning_rate": 8.50711323572504e-07, + "loss": 1.0293, + "step": 70782 + }, + { + "epoch": 5.23, + "learning_rate": 8.505503402329429e-07, + "loss": 0.9006, + "step": 70783 + }, + { + "epoch": 5.23, + "learning_rate": 8.503893714500266e-07, + "loss": 0.9536, + "step": 70784 + }, + { + "epoch": 5.23, + "learning_rate": 8.502284172240094e-07, + "loss": 0.9578, + "step": 70785 + }, + { + "epoch": 5.23, + "learning_rate": 8.500674775551466e-07, + "loss": 0.945, + "step": 70786 + }, + { + "epoch": 5.23, + "learning_rate": 8.499065524436967e-07, + "loss": 0.9687, + "step": 70787 + }, + { + "epoch": 5.23, + "learning_rate": 8.497456418899152e-07, + "loss": 1.0294, + "step": 70788 + }, + { + "epoch": 5.23, + "learning_rate": 8.495847458940565e-07, + "loss": 0.9691, + "step": 70789 + }, + { + "epoch": 5.23, + "learning_rate": 8.494238644563768e-07, + "loss": 0.9085, + "step": 70790 + }, + { + "epoch": 5.23, + "learning_rate": 8.492629975771338e-07, + "loss": 1.0142, + "step": 70791 + }, + { + "epoch": 5.23, + "learning_rate": 8.491021452565829e-07, + "loss": 0.9808, + "step": 70792 + }, + { + "epoch": 5.23, + "learning_rate": 8.489413074949792e-07, + "loss": 0.9788, + "step": 70793 + }, + { + "epoch": 5.23, + "learning_rate": 8.487804842925773e-07, + "loss": 0.9074, + "step": 70794 + }, + { + "epoch": 5.23, + "learning_rate": 8.486196756496368e-07, + "loss": 1.0028, + "step": 70795 + }, + { + "epoch": 5.23, + "learning_rate": 8.48458881566413e-07, + "loss": 0.9422, + "step": 70796 + }, + { + "epoch": 5.23, + "learning_rate": 8.482981020431568e-07, + "loss": 0.7949, + "step": 70797 + }, + { + "epoch": 5.23, + "learning_rate": 8.481373370801283e-07, + "loss": 1.1518, + "step": 70798 + }, + { + "epoch": 5.23, + "learning_rate": 8.479765866775813e-07, + "loss": 0.964, + "step": 70799 + }, + { + "epoch": 5.23, + "learning_rate": 8.47815850835777e-07, + "loss": 0.8688, + "step": 70800 + }, + { + "epoch": 5.23, + "learning_rate": 8.476551295549617e-07, + "loss": 1.0714, + "step": 70801 + }, + { + "epoch": 5.23, + "learning_rate": 8.474944228353988e-07, + "loss": 0.952, + "step": 70802 + }, + { + "epoch": 5.23, + "learning_rate": 8.473337306773388e-07, + "loss": 1.0657, + "step": 70803 + }, + { + "epoch": 5.23, + "learning_rate": 8.47173053081044e-07, + "loss": 0.9705, + "step": 70804 + }, + { + "epoch": 5.23, + "learning_rate": 8.470123900467642e-07, + "loss": 1.0395, + "step": 70805 + }, + { + "epoch": 5.23, + "learning_rate": 8.468517415747546e-07, + "loss": 0.9908, + "step": 70806 + }, + { + "epoch": 5.23, + "learning_rate": 8.46691107665275e-07, + "loss": 1.0071, + "step": 70807 + }, + { + "epoch": 5.23, + "learning_rate": 8.465304883185787e-07, + "loss": 0.9614, + "step": 70808 + }, + { + "epoch": 5.23, + "learning_rate": 8.463698835349221e-07, + "loss": 1.0199, + "step": 70809 + }, + { + "epoch": 5.23, + "learning_rate": 8.462092933145571e-07, + "loss": 0.9955, + "step": 70810 + }, + { + "epoch": 5.23, + "learning_rate": 8.460487176577447e-07, + "loss": 0.9828, + "step": 70811 + }, + { + "epoch": 5.23, + "learning_rate": 8.458881565647381e-07, + "loss": 1.0707, + "step": 70812 + }, + { + "epoch": 5.23, + "learning_rate": 8.457276100357914e-07, + "loss": 0.9953, + "step": 70813 + }, + { + "epoch": 5.23, + "learning_rate": 8.455670780711622e-07, + "loss": 1.1296, + "step": 70814 + }, + { + "epoch": 5.23, + "learning_rate": 8.454065606711026e-07, + "loss": 0.978, + "step": 70815 + }, + { + "epoch": 5.23, + "learning_rate": 8.452460578358712e-07, + "loss": 1.0721, + "step": 70816 + }, + { + "epoch": 5.23, + "learning_rate": 8.450855695657234e-07, + "loss": 0.9988, + "step": 70817 + }, + { + "epoch": 5.23, + "learning_rate": 8.449250958609134e-07, + "loss": 0.9361, + "step": 70818 + }, + { + "epoch": 5.23, + "learning_rate": 8.447646367216944e-07, + "loss": 1.037, + "step": 70819 + }, + { + "epoch": 5.23, + "learning_rate": 8.44604192148325e-07, + "loss": 0.9868, + "step": 70820 + }, + { + "epoch": 5.23, + "learning_rate": 8.444437621410606e-07, + "loss": 0.9862, + "step": 70821 + }, + { + "epoch": 5.23, + "learning_rate": 8.442833467001543e-07, + "loss": 0.9915, + "step": 70822 + }, + { + "epoch": 5.23, + "learning_rate": 8.441229458258626e-07, + "loss": 0.9167, + "step": 70823 + }, + { + "epoch": 5.23, + "learning_rate": 8.439625595184386e-07, + "loss": 1.0659, + "step": 70824 + }, + { + "epoch": 5.23, + "learning_rate": 8.438021877781433e-07, + "loss": 0.9063, + "step": 70825 + }, + { + "epoch": 5.23, + "learning_rate": 8.43641830605223e-07, + "loss": 0.9665, + "step": 70826 + }, + { + "epoch": 5.23, + "learning_rate": 8.434814879999398e-07, + "loss": 0.9178, + "step": 70827 + }, + { + "epoch": 5.23, + "learning_rate": 8.433211599625468e-07, + "loss": 0.9691, + "step": 70828 + }, + { + "epoch": 5.23, + "learning_rate": 8.431608464932983e-07, + "loss": 0.9146, + "step": 70829 + }, + { + "epoch": 5.23, + "learning_rate": 8.430005475924485e-07, + "loss": 0.9811, + "step": 70830 + }, + { + "epoch": 5.23, + "learning_rate": 8.42840263260255e-07, + "loss": 0.9113, + "step": 70831 + }, + { + "epoch": 5.23, + "learning_rate": 8.426799934969721e-07, + "loss": 0.856, + "step": 70832 + }, + { + "epoch": 5.23, + "learning_rate": 8.425197383028549e-07, + "loss": 0.8758, + "step": 70833 + }, + { + "epoch": 5.23, + "learning_rate": 8.423594976781568e-07, + "loss": 1.0245, + "step": 70834 + }, + { + "epoch": 5.23, + "learning_rate": 8.421992716231319e-07, + "loss": 0.9943, + "step": 70835 + }, + { + "epoch": 5.23, + "learning_rate": 8.42039060138039e-07, + "loss": 1.0161, + "step": 70836 + }, + { + "epoch": 5.23, + "learning_rate": 8.418788632231312e-07, + "loss": 1.1514, + "step": 70837 + }, + { + "epoch": 5.23, + "learning_rate": 8.417186808786626e-07, + "loss": 0.92, + "step": 70838 + }, + { + "epoch": 5.23, + "learning_rate": 8.415585131048875e-07, + "loss": 1.0231, + "step": 70839 + }, + { + "epoch": 5.23, + "learning_rate": 8.413983599020636e-07, + "loss": 0.9524, + "step": 70840 + }, + { + "epoch": 5.23, + "learning_rate": 8.412382212704451e-07, + "loss": 1.0046, + "step": 70841 + }, + { + "epoch": 5.23, + "learning_rate": 8.410780972102817e-07, + "loss": 1.0379, + "step": 70842 + }, + { + "epoch": 5.23, + "learning_rate": 8.409179877218343e-07, + "loss": 1.0464, + "step": 70843 + }, + { + "epoch": 5.23, + "learning_rate": 8.407578928053539e-07, + "loss": 0.9894, + "step": 70844 + }, + { + "epoch": 5.23, + "learning_rate": 8.405978124610992e-07, + "loss": 1.0416, + "step": 70845 + }, + { + "epoch": 5.23, + "learning_rate": 8.404377466893188e-07, + "loss": 1.0597, + "step": 70846 + }, + { + "epoch": 5.23, + "learning_rate": 8.402776954902736e-07, + "loss": 1.0993, + "step": 70847 + }, + { + "epoch": 5.23, + "learning_rate": 8.401176588642134e-07, + "loss": 0.9946, + "step": 70848 + }, + { + "epoch": 5.23, + "learning_rate": 8.399576368113982e-07, + "loss": 1.0554, + "step": 70849 + }, + { + "epoch": 5.24, + "learning_rate": 8.397976293320764e-07, + "loss": 1.0102, + "step": 70850 + }, + { + "epoch": 5.24, + "learning_rate": 8.396376364265068e-07, + "loss": 1.048, + "step": 70851 + }, + { + "epoch": 5.24, + "learning_rate": 8.394776580949427e-07, + "loss": 1.0115, + "step": 70852 + }, + { + "epoch": 5.24, + "learning_rate": 8.393176943376391e-07, + "loss": 0.8836, + "step": 70853 + }, + { + "epoch": 5.24, + "learning_rate": 8.391577451548494e-07, + "loss": 1.0841, + "step": 70854 + }, + { + "epoch": 5.24, + "learning_rate": 8.389978105468266e-07, + "loss": 1.0391, + "step": 70855 + }, + { + "epoch": 5.24, + "learning_rate": 8.388378905138295e-07, + "loss": 1.0888, + "step": 70856 + }, + { + "epoch": 5.24, + "learning_rate": 8.3867798505611e-07, + "loss": 0.9821, + "step": 70857 + }, + { + "epoch": 5.24, + "learning_rate": 8.385180941739235e-07, + "loss": 0.9774, + "step": 70858 + }, + { + "epoch": 5.24, + "learning_rate": 8.383582178675209e-07, + "loss": 1.0474, + "step": 70859 + }, + { + "epoch": 5.24, + "learning_rate": 8.381983561371609e-07, + "loss": 0.9608, + "step": 70860 + }, + { + "epoch": 5.24, + "learning_rate": 8.380385089830967e-07, + "loss": 1.036, + "step": 70861 + }, + { + "epoch": 5.24, + "learning_rate": 8.378786764055825e-07, + "loss": 0.913, + "step": 70862 + }, + { + "epoch": 5.24, + "learning_rate": 8.377188584048724e-07, + "loss": 1.0284, + "step": 70863 + }, + { + "epoch": 5.24, + "learning_rate": 8.375590549812174e-07, + "loss": 1.0362, + "step": 70864 + }, + { + "epoch": 5.24, + "learning_rate": 8.373992661348785e-07, + "loss": 0.9919, + "step": 70865 + }, + { + "epoch": 5.24, + "learning_rate": 8.372394918661042e-07, + "loss": 0.9797, + "step": 70866 + }, + { + "epoch": 5.24, + "learning_rate": 8.370797321751523e-07, + "loss": 0.9942, + "step": 70867 + }, + { + "epoch": 5.24, + "learning_rate": 8.369199870622735e-07, + "loss": 1.051, + "step": 70868 + }, + { + "epoch": 5.24, + "learning_rate": 8.367602565277256e-07, + "loss": 0.9614, + "step": 70869 + }, + { + "epoch": 5.24, + "learning_rate": 8.366005405717614e-07, + "loss": 0.9697, + "step": 70870 + }, + { + "epoch": 5.24, + "learning_rate": 8.364408391946332e-07, + "loss": 0.9833, + "step": 70871 + }, + { + "epoch": 5.24, + "learning_rate": 8.362811523965974e-07, + "loss": 0.9573, + "step": 70872 + }, + { + "epoch": 5.24, + "learning_rate": 8.361214801779072e-07, + "loss": 0.9829, + "step": 70873 + }, + { + "epoch": 5.24, + "learning_rate": 8.359618225388156e-07, + "loss": 0.9948, + "step": 70874 + }, + { + "epoch": 5.24, + "learning_rate": 8.358021794795768e-07, + "loss": 0.9846, + "step": 70875 + }, + { + "epoch": 5.24, + "learning_rate": 8.356425510004473e-07, + "loss": 1.0711, + "step": 70876 + }, + { + "epoch": 5.24, + "learning_rate": 8.354829371016804e-07, + "loss": 0.9614, + "step": 70877 + }, + { + "epoch": 5.24, + "learning_rate": 8.35323337783528e-07, + "loss": 0.9555, + "step": 70878 + }, + { + "epoch": 5.24, + "learning_rate": 8.351637530462442e-07, + "loss": 1.0006, + "step": 70879 + }, + { + "epoch": 5.24, + "learning_rate": 8.350041828900835e-07, + "loss": 1.0175, + "step": 70880 + }, + { + "epoch": 5.24, + "learning_rate": 8.348446273153011e-07, + "loss": 1.014, + "step": 70881 + }, + { + "epoch": 5.24, + "learning_rate": 8.346850863221501e-07, + "loss": 0.9567, + "step": 70882 + }, + { + "epoch": 5.24, + "learning_rate": 8.345255599108836e-07, + "loss": 0.969, + "step": 70883 + }, + { + "epoch": 5.24, + "learning_rate": 8.343660480817539e-07, + "loss": 0.9949, + "step": 70884 + }, + { + "epoch": 5.24, + "learning_rate": 8.342065508350183e-07, + "loss": 1.0209, + "step": 70885 + }, + { + "epoch": 5.24, + "learning_rate": 8.3404706817093e-07, + "loss": 0.9754, + "step": 70886 + }, + { + "epoch": 5.24, + "learning_rate": 8.338876000897411e-07, + "loss": 0.9511, + "step": 70887 + }, + { + "epoch": 5.24, + "learning_rate": 8.337281465917058e-07, + "loss": 0.9558, + "step": 70888 + }, + { + "epoch": 5.24, + "learning_rate": 8.335687076770759e-07, + "loss": 1.0711, + "step": 70889 + }, + { + "epoch": 5.24, + "learning_rate": 8.334092833461105e-07, + "loss": 0.9485, + "step": 70890 + }, + { + "epoch": 5.24, + "learning_rate": 8.332498735990568e-07, + "loss": 0.9675, + "step": 70891 + }, + { + "epoch": 5.24, + "learning_rate": 8.330904784361726e-07, + "loss": 0.8855, + "step": 70892 + }, + { + "epoch": 5.24, + "learning_rate": 8.329310978577088e-07, + "loss": 0.9808, + "step": 70893 + }, + { + "epoch": 5.24, + "learning_rate": 8.327717318639239e-07, + "loss": 0.9534, + "step": 70894 + }, + { + "epoch": 5.24, + "learning_rate": 8.326123804550645e-07, + "loss": 0.8833, + "step": 70895 + }, + { + "epoch": 5.24, + "learning_rate": 8.324530436313904e-07, + "loss": 0.9826, + "step": 70896 + }, + { + "epoch": 5.24, + "learning_rate": 8.322937213931492e-07, + "loss": 1.0432, + "step": 70897 + }, + { + "epoch": 5.24, + "learning_rate": 8.321344137406018e-07, + "loss": 0.8235, + "step": 70898 + }, + { + "epoch": 5.24, + "learning_rate": 8.319751206739946e-07, + "loss": 1.0375, + "step": 70899 + }, + { + "epoch": 5.24, + "learning_rate": 8.318158421935829e-07, + "loss": 1.0956, + "step": 70900 + }, + { + "epoch": 5.24, + "learning_rate": 8.316565782996233e-07, + "loss": 1.0316, + "step": 70901 + }, + { + "epoch": 5.24, + "learning_rate": 8.314973289923655e-07, + "loss": 1.0135, + "step": 70902 + }, + { + "epoch": 5.24, + "learning_rate": 8.31338094272065e-07, + "loss": 0.931, + "step": 70903 + }, + { + "epoch": 5.24, + "learning_rate": 8.311788741389736e-07, + "loss": 0.9933, + "step": 70904 + }, + { + "epoch": 5.24, + "learning_rate": 8.310196685933458e-07, + "loss": 0.9649, + "step": 70905 + }, + { + "epoch": 5.24, + "learning_rate": 8.308604776354345e-07, + "loss": 1.0376, + "step": 70906 + }, + { + "epoch": 5.24, + "learning_rate": 8.307013012654941e-07, + "loss": 0.9847, + "step": 70907 + }, + { + "epoch": 5.24, + "learning_rate": 8.305421394837765e-07, + "loss": 0.9646, + "step": 70908 + }, + { + "epoch": 5.24, + "learning_rate": 8.303829922905327e-07, + "loss": 1.058, + "step": 70909 + }, + { + "epoch": 5.24, + "learning_rate": 8.302238596860202e-07, + "loss": 1.084, + "step": 70910 + }, + { + "epoch": 5.24, + "learning_rate": 8.30064741670491e-07, + "loss": 1.0423, + "step": 70911 + }, + { + "epoch": 5.24, + "learning_rate": 8.299056382441973e-07, + "loss": 0.9799, + "step": 70912 + }, + { + "epoch": 5.24, + "learning_rate": 8.297465494073898e-07, + "loss": 0.9929, + "step": 70913 + }, + { + "epoch": 5.24, + "learning_rate": 8.295874751603272e-07, + "loss": 1.0441, + "step": 70914 + }, + { + "epoch": 5.24, + "learning_rate": 8.294284155032594e-07, + "loss": 0.9529, + "step": 70915 + }, + { + "epoch": 5.24, + "learning_rate": 8.292693704364397e-07, + "loss": 0.9752, + "step": 70916 + }, + { + "epoch": 5.24, + "learning_rate": 8.29110339960122e-07, + "loss": 1.062, + "step": 70917 + }, + { + "epoch": 5.24, + "learning_rate": 8.289513240745573e-07, + "loss": 1.0589, + "step": 70918 + }, + { + "epoch": 5.24, + "learning_rate": 8.28792322780001e-07, + "loss": 1.0052, + "step": 70919 + }, + { + "epoch": 5.24, + "learning_rate": 8.286333360767018e-07, + "loss": 0.9496, + "step": 70920 + }, + { + "epoch": 5.24, + "learning_rate": 8.284743639649184e-07, + "loss": 1.0991, + "step": 70921 + }, + { + "epoch": 5.24, + "learning_rate": 8.283154064449017e-07, + "loss": 0.9462, + "step": 70922 + }, + { + "epoch": 5.24, + "learning_rate": 8.281564635169026e-07, + "loss": 1.1132, + "step": 70923 + }, + { + "epoch": 5.24, + "learning_rate": 8.279975351811742e-07, + "loss": 0.8835, + "step": 70924 + }, + { + "epoch": 5.24, + "learning_rate": 8.278386214379719e-07, + "loss": 0.975, + "step": 70925 + }, + { + "epoch": 5.24, + "learning_rate": 8.276797222875476e-07, + "loss": 1.0869, + "step": 70926 + }, + { + "epoch": 5.24, + "learning_rate": 8.275208377301535e-07, + "loss": 0.9324, + "step": 70927 + }, + { + "epoch": 5.24, + "learning_rate": 8.273619677660427e-07, + "loss": 1.0317, + "step": 70928 + }, + { + "epoch": 5.24, + "learning_rate": 8.27203112395466e-07, + "loss": 1.0461, + "step": 70929 + }, + { + "epoch": 5.24, + "learning_rate": 8.270442716186799e-07, + "loss": 0.8522, + "step": 70930 + }, + { + "epoch": 5.24, + "learning_rate": 8.268854454359343e-07, + "loss": 0.8886, + "step": 70931 + }, + { + "epoch": 5.24, + "learning_rate": 8.267266338474833e-07, + "loss": 1.0774, + "step": 70932 + }, + { + "epoch": 5.24, + "learning_rate": 8.265678368535768e-07, + "loss": 1.0009, + "step": 70933 + }, + { + "epoch": 5.24, + "learning_rate": 8.264090544544711e-07, + "loss": 0.9838, + "step": 70934 + }, + { + "epoch": 5.24, + "learning_rate": 8.262502866504197e-07, + "loss": 0.847, + "step": 70935 + }, + { + "epoch": 5.24, + "learning_rate": 8.260915334416697e-07, + "loss": 1.045, + "step": 70936 + }, + { + "epoch": 5.24, + "learning_rate": 8.259327948284779e-07, + "loss": 0.9517, + "step": 70937 + }, + { + "epoch": 5.24, + "learning_rate": 8.257740708110939e-07, + "loss": 1.0231, + "step": 70938 + }, + { + "epoch": 5.24, + "learning_rate": 8.256153613897766e-07, + "loss": 0.9776, + "step": 70939 + }, + { + "epoch": 5.24, + "learning_rate": 8.25456666564769e-07, + "loss": 1.1654, + "step": 70940 + }, + { + "epoch": 5.24, + "learning_rate": 8.25297986336332e-07, + "loss": 1.0043, + "step": 70941 + }, + { + "epoch": 5.24, + "learning_rate": 8.251393207047109e-07, + "loss": 0.931, + "step": 70942 + }, + { + "epoch": 5.24, + "learning_rate": 8.249806696701668e-07, + "loss": 1.1183, + "step": 70943 + }, + { + "epoch": 5.24, + "learning_rate": 8.248220332329449e-07, + "loss": 1.0439, + "step": 70944 + }, + { + "epoch": 5.24, + "learning_rate": 8.246634113932983e-07, + "loss": 0.9899, + "step": 70945 + }, + { + "epoch": 5.24, + "learning_rate": 8.245048041514835e-07, + "loss": 1.0344, + "step": 70946 + }, + { + "epoch": 5.24, + "learning_rate": 8.243462115077494e-07, + "loss": 0.9971, + "step": 70947 + }, + { + "epoch": 5.24, + "learning_rate": 8.241876334623488e-07, + "loss": 0.9697, + "step": 70948 + }, + { + "epoch": 5.24, + "learning_rate": 8.240290700155329e-07, + "loss": 0.9403, + "step": 70949 + }, + { + "epoch": 5.24, + "learning_rate": 8.238705211675579e-07, + "loss": 1.0798, + "step": 70950 + }, + { + "epoch": 5.24, + "learning_rate": 8.237119869186727e-07, + "loss": 1.0907, + "step": 70951 + }, + { + "epoch": 5.24, + "learning_rate": 8.235534672691314e-07, + "loss": 1.0947, + "step": 70952 + }, + { + "epoch": 5.24, + "learning_rate": 8.233949622191828e-07, + "loss": 1.0108, + "step": 70953 + }, + { + "epoch": 5.24, + "learning_rate": 8.232364717690833e-07, + "loss": 1.0487, + "step": 70954 + }, + { + "epoch": 5.24, + "learning_rate": 8.230779959190838e-07, + "loss": 0.9669, + "step": 70955 + }, + { + "epoch": 5.24, + "learning_rate": 8.229195346694351e-07, + "loss": 1.0086, + "step": 70956 + }, + { + "epoch": 5.24, + "learning_rate": 8.227610880203907e-07, + "loss": 0.9634, + "step": 70957 + }, + { + "epoch": 5.24, + "learning_rate": 8.226026559722e-07, + "loss": 1.0144, + "step": 70958 + }, + { + "epoch": 5.24, + "learning_rate": 8.224442385251197e-07, + "loss": 1.1415, + "step": 70959 + }, + { + "epoch": 5.24, + "learning_rate": 8.222858356793984e-07, + "loss": 0.9986, + "step": 70960 + }, + { + "epoch": 5.24, + "learning_rate": 8.221274474352891e-07, + "loss": 1.0774, + "step": 70961 + }, + { + "epoch": 5.24, + "learning_rate": 8.21969073793043e-07, + "loss": 0.9126, + "step": 70962 + }, + { + "epoch": 5.24, + "learning_rate": 8.218107147529153e-07, + "loss": 1.0321, + "step": 70963 + }, + { + "epoch": 5.24, + "learning_rate": 8.216523703151536e-07, + "loss": 0.9618, + "step": 70964 + }, + { + "epoch": 5.24, + "learning_rate": 8.214940404800098e-07, + "loss": 0.9871, + "step": 70965 + }, + { + "epoch": 5.24, + "learning_rate": 8.213357252477405e-07, + "loss": 1.049, + "step": 70966 + }, + { + "epoch": 5.24, + "learning_rate": 8.211774246185933e-07, + "loss": 0.9399, + "step": 70967 + }, + { + "epoch": 5.24, + "learning_rate": 8.210191385928223e-07, + "loss": 0.9768, + "step": 70968 + }, + { + "epoch": 5.24, + "learning_rate": 8.208608671706774e-07, + "loss": 0.9896, + "step": 70969 + }, + { + "epoch": 5.24, + "learning_rate": 8.207026103524118e-07, + "loss": 0.888, + "step": 70970 + }, + { + "epoch": 5.24, + "learning_rate": 8.205443681382786e-07, + "loss": 1.0898, + "step": 70971 + }, + { + "epoch": 5.24, + "learning_rate": 8.203861405285262e-07, + "loss": 1.0389, + "step": 70972 + }, + { + "epoch": 5.24, + "learning_rate": 8.202279275234093e-07, + "loss": 0.9408, + "step": 70973 + }, + { + "epoch": 5.24, + "learning_rate": 8.200697291231763e-07, + "loss": 0.9473, + "step": 70974 + }, + { + "epoch": 5.24, + "learning_rate": 8.199115453280837e-07, + "loss": 1.0136, + "step": 70975 + }, + { + "epoch": 5.24, + "learning_rate": 8.197533761383792e-07, + "loss": 0.9627, + "step": 70976 + }, + { + "epoch": 5.24, + "learning_rate": 8.195952215543168e-07, + "loss": 0.9627, + "step": 70977 + }, + { + "epoch": 5.24, + "learning_rate": 8.194370815761443e-07, + "loss": 1.0465, + "step": 70978 + }, + { + "epoch": 5.24, + "learning_rate": 8.192789562041181e-07, + "loss": 0.9654, + "step": 70979 + }, + { + "epoch": 5.24, + "learning_rate": 8.19120845438488e-07, + "loss": 1.0594, + "step": 70980 + }, + { + "epoch": 5.24, + "learning_rate": 8.189627492795049e-07, + "loss": 0.912, + "step": 70981 + }, + { + "epoch": 5.24, + "learning_rate": 8.188046677274209e-07, + "loss": 1.0152, + "step": 70982 + }, + { + "epoch": 5.24, + "learning_rate": 8.186466007824856e-07, + "loss": 1.0863, + "step": 70983 + }, + { + "epoch": 5.24, + "learning_rate": 8.184885484449557e-07, + "loss": 0.9967, + "step": 70984 + }, + { + "epoch": 5.25, + "learning_rate": 8.183305107150763e-07, + "loss": 1.0272, + "step": 70985 + }, + { + "epoch": 5.25, + "learning_rate": 8.18172487593103e-07, + "loss": 1.0329, + "step": 70986 + }, + { + "epoch": 5.25, + "learning_rate": 8.180144790792832e-07, + "loss": 1.0129, + "step": 70987 + }, + { + "epoch": 5.25, + "learning_rate": 8.178564851738757e-07, + "loss": 1.0426, + "step": 70988 + }, + { + "epoch": 5.25, + "learning_rate": 8.176985058771236e-07, + "loss": 1.0676, + "step": 70989 + }, + { + "epoch": 5.25, + "learning_rate": 8.175405411892823e-07, + "loss": 1.0152, + "step": 70990 + }, + { + "epoch": 5.25, + "learning_rate": 8.173825911106037e-07, + "loss": 0.9263, + "step": 70991 + }, + { + "epoch": 5.25, + "learning_rate": 8.172246556413377e-07, + "loss": 0.8954, + "step": 70992 + }, + { + "epoch": 5.25, + "learning_rate": 8.170667347817362e-07, + "loss": 0.9878, + "step": 70993 + }, + { + "epoch": 5.25, + "learning_rate": 8.169088285320481e-07, + "loss": 1.0501, + "step": 70994 + }, + { + "epoch": 5.25, + "learning_rate": 8.167509368925286e-07, + "loss": 0.9487, + "step": 70995 + }, + { + "epoch": 5.25, + "learning_rate": 8.165930598634264e-07, + "loss": 1.0215, + "step": 70996 + }, + { + "epoch": 5.25, + "learning_rate": 8.164351974449947e-07, + "loss": 0.9158, + "step": 70997 + }, + { + "epoch": 5.25, + "learning_rate": 8.162773496374799e-07, + "loss": 0.9273, + "step": 70998 + }, + { + "epoch": 5.25, + "learning_rate": 8.161195164411384e-07, + "loss": 1.0235, + "step": 70999 + }, + { + "epoch": 5.25, + "learning_rate": 8.159616978562202e-07, + "loss": 0.8392, + "step": 71000 + }, + { + "epoch": 5.25, + "learning_rate": 8.158038938829749e-07, + "loss": 1.0606, + "step": 71001 + }, + { + "epoch": 5.25, + "learning_rate": 8.156461045216546e-07, + "loss": 1.0226, + "step": 71002 + }, + { + "epoch": 5.25, + "learning_rate": 8.15488329772508e-07, + "loss": 0.9421, + "step": 71003 + }, + { + "epoch": 5.25, + "learning_rate": 8.153305696357894e-07, + "loss": 0.9561, + "step": 71004 + }, + { + "epoch": 5.25, + "learning_rate": 8.151728241117485e-07, + "loss": 1.0269, + "step": 71005 + }, + { + "epoch": 5.25, + "learning_rate": 8.150150932006362e-07, + "loss": 1.0058, + "step": 71006 + }, + { + "epoch": 5.25, + "learning_rate": 8.148573769027024e-07, + "loss": 0.9786, + "step": 71007 + }, + { + "epoch": 5.25, + "learning_rate": 8.146996752182024e-07, + "loss": 1.0488, + "step": 71008 + }, + { + "epoch": 5.25, + "learning_rate": 8.145419881473804e-07, + "loss": 0.9057, + "step": 71009 + }, + { + "epoch": 5.25, + "learning_rate": 8.143843156904918e-07, + "loss": 1.0212, + "step": 71010 + }, + { + "epoch": 5.25, + "learning_rate": 8.142266578477864e-07, + "loss": 1.1186, + "step": 71011 + }, + { + "epoch": 5.25, + "learning_rate": 8.140690146195151e-07, + "loss": 0.9589, + "step": 71012 + }, + { + "epoch": 5.25, + "learning_rate": 8.139113860059289e-07, + "loss": 0.9863, + "step": 71013 + }, + { + "epoch": 5.25, + "learning_rate": 8.137537720072763e-07, + "loss": 1.0115, + "step": 71014 + }, + { + "epoch": 5.25, + "learning_rate": 8.135961726238117e-07, + "loss": 1.0406, + "step": 71015 + }, + { + "epoch": 5.25, + "learning_rate": 8.134385878557849e-07, + "loss": 0.9658, + "step": 71016 + }, + { + "epoch": 5.25, + "learning_rate": 8.132810177034445e-07, + "loss": 1.0156, + "step": 71017 + }, + { + "epoch": 5.25, + "learning_rate": 8.131234621670414e-07, + "loss": 1.1077, + "step": 71018 + }, + { + "epoch": 5.25, + "learning_rate": 8.1296592124683e-07, + "loss": 1.0345, + "step": 71019 + }, + { + "epoch": 5.25, + "learning_rate": 8.128083949430576e-07, + "loss": 1.0038, + "step": 71020 + }, + { + "epoch": 5.25, + "learning_rate": 8.126508832559754e-07, + "loss": 1.0292, + "step": 71021 + }, + { + "epoch": 5.25, + "learning_rate": 8.124933861858342e-07, + "loss": 1.015, + "step": 71022 + }, + { + "epoch": 5.25, + "learning_rate": 8.123359037328838e-07, + "loss": 1.0168, + "step": 71023 + }, + { + "epoch": 5.25, + "learning_rate": 8.121784358973761e-07, + "loss": 1.1055, + "step": 71024 + }, + { + "epoch": 5.25, + "learning_rate": 8.120209826795611e-07, + "loss": 1.0845, + "step": 71025 + }, + { + "epoch": 5.25, + "learning_rate": 8.118635440796896e-07, + "loss": 0.9988, + "step": 71026 + }, + { + "epoch": 5.25, + "learning_rate": 8.117061200980104e-07, + "loss": 1.0857, + "step": 71027 + }, + { + "epoch": 5.25, + "learning_rate": 8.115487107347764e-07, + "loss": 1.0709, + "step": 71028 + }, + { + "epoch": 5.25, + "learning_rate": 8.113913159902398e-07, + "loss": 1.0736, + "step": 71029 + }, + { + "epoch": 5.25, + "learning_rate": 8.112339358646437e-07, + "loss": 0.9523, + "step": 71030 + }, + { + "epoch": 5.25, + "learning_rate": 8.110765703582446e-07, + "loss": 1.0403, + "step": 71031 + }, + { + "epoch": 5.25, + "learning_rate": 8.109192194712901e-07, + "loss": 0.9445, + "step": 71032 + }, + { + "epoch": 5.25, + "learning_rate": 8.107618832040343e-07, + "loss": 1.0741, + "step": 71033 + }, + { + "epoch": 5.25, + "learning_rate": 8.106045615567216e-07, + "loss": 1.1066, + "step": 71034 + }, + { + "epoch": 5.25, + "learning_rate": 8.104472545296071e-07, + "loss": 0.9273, + "step": 71035 + }, + { + "epoch": 5.25, + "learning_rate": 8.102899621229388e-07, + "loss": 1.1073, + "step": 71036 + }, + { + "epoch": 5.25, + "learning_rate": 8.101326843369705e-07, + "loss": 0.9169, + "step": 71037 + }, + { + "epoch": 5.25, + "learning_rate": 8.09975421171948e-07, + "loss": 0.9923, + "step": 71038 + }, + { + "epoch": 5.25, + "learning_rate": 8.098181726281207e-07, + "loss": 0.9457, + "step": 71039 + }, + { + "epoch": 5.25, + "learning_rate": 8.096609387057442e-07, + "loss": 0.994, + "step": 71040 + }, + { + "epoch": 5.25, + "learning_rate": 8.095037194050648e-07, + "loss": 1.1589, + "step": 71041 + }, + { + "epoch": 5.25, + "learning_rate": 8.093465147263335e-07, + "loss": 0.9549, + "step": 71042 + }, + { + "epoch": 5.25, + "learning_rate": 8.091893246697991e-07, + "loss": 1.0119, + "step": 71043 + }, + { + "epoch": 5.25, + "learning_rate": 8.090321492357145e-07, + "loss": 0.955, + "step": 71044 + }, + { + "epoch": 5.25, + "learning_rate": 8.088749884243286e-07, + "loss": 1.0481, + "step": 71045 + }, + { + "epoch": 5.25, + "learning_rate": 8.087178422358899e-07, + "loss": 1.0705, + "step": 71046 + }, + { + "epoch": 5.25, + "learning_rate": 8.085607106706506e-07, + "loss": 1.0519, + "step": 71047 + }, + { + "epoch": 5.25, + "learning_rate": 8.084035937288582e-07, + "loss": 1.081, + "step": 71048 + }, + { + "epoch": 5.25, + "learning_rate": 8.082464914107669e-07, + "loss": 0.9966, + "step": 71049 + }, + { + "epoch": 5.25, + "learning_rate": 8.080894037166209e-07, + "loss": 0.9978, + "step": 71050 + }, + { + "epoch": 5.25, + "learning_rate": 8.079323306466746e-07, + "loss": 1.0328, + "step": 71051 + }, + { + "epoch": 5.25, + "learning_rate": 8.077752722011756e-07, + "loss": 0.9244, + "step": 71052 + }, + { + "epoch": 5.25, + "learning_rate": 8.076182283803768e-07, + "loss": 0.8966, + "step": 71053 + }, + { + "epoch": 5.25, + "learning_rate": 8.074611991845227e-07, + "loss": 1.1332, + "step": 71054 + }, + { + "epoch": 5.25, + "learning_rate": 8.073041846138685e-07, + "loss": 1.0408, + "step": 71055 + }, + { + "epoch": 5.25, + "learning_rate": 8.071471846686618e-07, + "loss": 1.0545, + "step": 71056 + }, + { + "epoch": 5.25, + "learning_rate": 8.069901993491525e-07, + "loss": 0.9166, + "step": 71057 + }, + { + "epoch": 5.25, + "learning_rate": 8.068332286555903e-07, + "loss": 1.016, + "step": 71058 + }, + { + "epoch": 5.25, + "learning_rate": 8.066762725882227e-07, + "loss": 0.9762, + "step": 71059 + }, + { + "epoch": 5.25, + "learning_rate": 8.06519331147303e-07, + "loss": 1.0651, + "step": 71060 + }, + { + "epoch": 5.25, + "learning_rate": 8.063624043330797e-07, + "loss": 0.9679, + "step": 71061 + }, + { + "epoch": 5.25, + "learning_rate": 8.062054921458029e-07, + "loss": 1.0333, + "step": 71062 + }, + { + "epoch": 5.25, + "learning_rate": 8.0604859458572e-07, + "loss": 0.8879, + "step": 71063 + }, + { + "epoch": 5.25, + "learning_rate": 8.058917116530829e-07, + "loss": 1.0099, + "step": 71064 + }, + { + "epoch": 5.25, + "learning_rate": 8.057348433481416e-07, + "loss": 0.991, + "step": 71065 + }, + { + "epoch": 5.25, + "learning_rate": 8.055779896711446e-07, + "loss": 1.0399, + "step": 71066 + }, + { + "epoch": 5.25, + "learning_rate": 8.054211506223408e-07, + "loss": 1.0452, + "step": 71067 + }, + { + "epoch": 5.25, + "learning_rate": 8.052643262019788e-07, + "loss": 1.1081, + "step": 71068 + }, + { + "epoch": 5.25, + "learning_rate": 8.051075164103117e-07, + "loss": 0.9401, + "step": 71069 + }, + { + "epoch": 5.25, + "learning_rate": 8.049507212475871e-07, + "loss": 0.9801, + "step": 71070 + }, + { + "epoch": 5.25, + "learning_rate": 8.047939407140537e-07, + "loss": 0.9856, + "step": 71071 + }, + { + "epoch": 5.25, + "learning_rate": 8.046371748099602e-07, + "loss": 1.0078, + "step": 71072 + }, + { + "epoch": 5.25, + "learning_rate": 8.044804235355597e-07, + "loss": 1.078, + "step": 71073 + }, + { + "epoch": 5.25, + "learning_rate": 8.043236868910986e-07, + "loss": 0.8899, + "step": 71074 + }, + { + "epoch": 5.25, + "learning_rate": 8.041669648768269e-07, + "loss": 0.9317, + "step": 71075 + }, + { + "epoch": 5.25, + "learning_rate": 8.040102574929942e-07, + "loss": 0.9995, + "step": 71076 + }, + { + "epoch": 5.25, + "learning_rate": 8.038535647398482e-07, + "loss": 0.948, + "step": 71077 + }, + { + "epoch": 5.25, + "learning_rate": 8.03696886617642e-07, + "loss": 1.0376, + "step": 71078 + }, + { + "epoch": 5.25, + "learning_rate": 8.035402231266198e-07, + "loss": 1.0327, + "step": 71079 + }, + { + "epoch": 5.25, + "learning_rate": 8.03383574267036e-07, + "loss": 1.0286, + "step": 71080 + }, + { + "epoch": 5.25, + "learning_rate": 8.032269400391346e-07, + "loss": 0.8122, + "step": 71081 + }, + { + "epoch": 5.25, + "learning_rate": 8.030703204431711e-07, + "loss": 0.9812, + "step": 71082 + }, + { + "epoch": 5.25, + "learning_rate": 8.029137154793876e-07, + "loss": 1.0666, + "step": 71083 + }, + { + "epoch": 5.25, + "learning_rate": 8.027571251480381e-07, + "loss": 0.9789, + "step": 71084 + }, + { + "epoch": 5.25, + "learning_rate": 8.026005494493716e-07, + "loss": 0.8944, + "step": 71085 + }, + { + "epoch": 5.25, + "learning_rate": 8.024439883836355e-07, + "loss": 1.0105, + "step": 71086 + }, + { + "epoch": 5.25, + "learning_rate": 8.022874419510795e-07, + "loss": 1.0922, + "step": 71087 + }, + { + "epoch": 5.25, + "learning_rate": 8.021309101519503e-07, + "loss": 1.0061, + "step": 71088 + }, + { + "epoch": 5.25, + "learning_rate": 8.019743929865021e-07, + "loss": 0.9904, + "step": 71089 + }, + { + "epoch": 5.25, + "learning_rate": 8.0181789045498e-07, + "loss": 1.0394, + "step": 71090 + }, + { + "epoch": 5.25, + "learning_rate": 8.016614025576341e-07, + "loss": 1.0413, + "step": 71091 + }, + { + "epoch": 5.25, + "learning_rate": 8.015049292947119e-07, + "loss": 0.9813, + "step": 71092 + }, + { + "epoch": 5.25, + "learning_rate": 8.013484706664653e-07, + "loss": 1.0085, + "step": 71093 + }, + { + "epoch": 5.25, + "learning_rate": 8.011920266731443e-07, + "loss": 0.9566, + "step": 71094 + }, + { + "epoch": 5.25, + "learning_rate": 8.010355973149908e-07, + "loss": 1.0361, + "step": 71095 + }, + { + "epoch": 5.25, + "learning_rate": 8.008791825922602e-07, + "loss": 0.9773, + "step": 71096 + }, + { + "epoch": 5.25, + "learning_rate": 8.007227825051967e-07, + "loss": 1.0527, + "step": 71097 + }, + { + "epoch": 5.25, + "learning_rate": 8.005663970540556e-07, + "loss": 1.0065, + "step": 71098 + }, + { + "epoch": 5.25, + "learning_rate": 8.00410026239078e-07, + "loss": 1.0875, + "step": 71099 + }, + { + "epoch": 5.25, + "learning_rate": 8.00253670060519e-07, + "loss": 1.0989, + "step": 71100 + }, + { + "epoch": 5.25, + "learning_rate": 8.00097328518622e-07, + "loss": 1.0253, + "step": 71101 + }, + { + "epoch": 5.25, + "learning_rate": 7.999410016136422e-07, + "loss": 0.9779, + "step": 71102 + }, + { + "epoch": 5.25, + "learning_rate": 7.997846893458239e-07, + "loss": 0.9435, + "step": 71103 + }, + { + "epoch": 5.25, + "learning_rate": 7.996283917154135e-07, + "loss": 1.0303, + "step": 71104 + }, + { + "epoch": 5.25, + "learning_rate": 7.994721087226653e-07, + "loss": 0.9853, + "step": 71105 + }, + { + "epoch": 5.25, + "learning_rate": 7.993158403678247e-07, + "loss": 0.9173, + "step": 71106 + }, + { + "epoch": 5.25, + "learning_rate": 7.991595866511415e-07, + "loss": 0.9575, + "step": 71107 + }, + { + "epoch": 5.25, + "learning_rate": 7.99003347572862e-07, + "loss": 0.9998, + "step": 71108 + }, + { + "epoch": 5.25, + "learning_rate": 7.988471231332384e-07, + "loss": 0.9942, + "step": 71109 + }, + { + "epoch": 5.25, + "learning_rate": 7.986909133325161e-07, + "loss": 0.9369, + "step": 71110 + }, + { + "epoch": 5.25, + "learning_rate": 7.985347181709468e-07, + "loss": 1.0563, + "step": 71111 + }, + { + "epoch": 5.25, + "learning_rate": 7.98378537648774e-07, + "loss": 1.0782, + "step": 71112 + }, + { + "epoch": 5.25, + "learning_rate": 7.982223717662518e-07, + "loss": 1.0664, + "step": 71113 + }, + { + "epoch": 5.25, + "learning_rate": 7.980662205236256e-07, + "loss": 0.9739, + "step": 71114 + }, + { + "epoch": 5.25, + "learning_rate": 7.97910083921144e-07, + "loss": 0.9757, + "step": 71115 + }, + { + "epoch": 5.25, + "learning_rate": 7.977539619590568e-07, + "loss": 0.9964, + "step": 71116 + }, + { + "epoch": 5.25, + "learning_rate": 7.975978546376095e-07, + "loss": 1.0521, + "step": 71117 + }, + { + "epoch": 5.25, + "learning_rate": 7.97441761957054e-07, + "loss": 0.9353, + "step": 71118 + }, + { + "epoch": 5.25, + "learning_rate": 7.972856839176369e-07, + "loss": 1.0952, + "step": 71119 + }, + { + "epoch": 5.26, + "learning_rate": 7.971296205196055e-07, + "loss": 0.8683, + "step": 71120 + }, + { + "epoch": 5.26, + "learning_rate": 7.969735717632088e-07, + "loss": 1.0592, + "step": 71121 + }, + { + "epoch": 5.26, + "learning_rate": 7.968175376486976e-07, + "loss": 1.0431, + "step": 71122 + }, + { + "epoch": 5.26, + "learning_rate": 7.966615181763182e-07, + "loss": 0.9799, + "step": 71123 + }, + { + "epoch": 5.26, + "learning_rate": 7.965055133463162e-07, + "loss": 0.9513, + "step": 71124 + }, + { + "epoch": 5.26, + "learning_rate": 7.963495231589436e-07, + "loss": 0.9653, + "step": 71125 + }, + { + "epoch": 5.26, + "learning_rate": 7.961935476144477e-07, + "loss": 0.9021, + "step": 71126 + }, + { + "epoch": 5.26, + "learning_rate": 7.960375867130765e-07, + "loss": 1.0179, + "step": 71127 + }, + { + "epoch": 5.26, + "learning_rate": 7.958816404550751e-07, + "loss": 1.095, + "step": 71128 + }, + { + "epoch": 5.26, + "learning_rate": 7.957257088406967e-07, + "loss": 1.0553, + "step": 71129 + }, + { + "epoch": 5.26, + "learning_rate": 7.955697918701865e-07, + "loss": 1.0889, + "step": 71130 + }, + { + "epoch": 5.26, + "learning_rate": 7.954138895437935e-07, + "loss": 1.0124, + "step": 71131 + }, + { + "epoch": 5.26, + "learning_rate": 7.95258001861765e-07, + "loss": 0.9185, + "step": 71132 + }, + { + "epoch": 5.26, + "learning_rate": 7.951021288243477e-07, + "loss": 0.801, + "step": 71133 + }, + { + "epoch": 5.26, + "learning_rate": 7.949462704317923e-07, + "loss": 1.0386, + "step": 71134 + }, + { + "epoch": 5.26, + "learning_rate": 7.947904266843465e-07, + "loss": 1.0293, + "step": 71135 + }, + { + "epoch": 5.26, + "learning_rate": 7.946345975822578e-07, + "loss": 1.0251, + "step": 71136 + }, + { + "epoch": 5.26, + "learning_rate": 7.944787831257717e-07, + "loss": 1.1739, + "step": 71137 + }, + { + "epoch": 5.26, + "learning_rate": 7.943229833151389e-07, + "loss": 0.9887, + "step": 71138 + }, + { + "epoch": 5.26, + "learning_rate": 7.941671981506072e-07, + "loss": 1.0058, + "step": 71139 + }, + { + "epoch": 5.26, + "learning_rate": 7.94011427632424e-07, + "loss": 1.0705, + "step": 71140 + }, + { + "epoch": 5.26, + "learning_rate": 7.93855671760837e-07, + "loss": 1.0403, + "step": 71141 + }, + { + "epoch": 5.26, + "learning_rate": 7.936999305360915e-07, + "loss": 0.9775, + "step": 71142 + }, + { + "epoch": 5.26, + "learning_rate": 7.935442039584418e-07, + "loss": 1.0124, + "step": 71143 + }, + { + "epoch": 5.26, + "learning_rate": 7.933884920281288e-07, + "loss": 1.0828, + "step": 71144 + }, + { + "epoch": 5.26, + "learning_rate": 7.932327947454032e-07, + "loss": 0.8778, + "step": 71145 + }, + { + "epoch": 5.26, + "learning_rate": 7.930771121105118e-07, + "loss": 0.9796, + "step": 71146 + }, + { + "epoch": 5.26, + "learning_rate": 7.929214441237054e-07, + "loss": 1.0214, + "step": 71147 + }, + { + "epoch": 5.26, + "learning_rate": 7.927657907852271e-07, + "loss": 0.9698, + "step": 71148 + }, + { + "epoch": 5.26, + "learning_rate": 7.926101520953267e-07, + "loss": 0.9563, + "step": 71149 + }, + { + "epoch": 5.26, + "learning_rate": 7.924545280542528e-07, + "loss": 1.0398, + "step": 71150 + }, + { + "epoch": 5.26, + "learning_rate": 7.922989186622521e-07, + "loss": 0.9995, + "step": 71151 + }, + { + "epoch": 5.26, + "learning_rate": 7.921433239195719e-07, + "loss": 0.9278, + "step": 71152 + }, + { + "epoch": 5.26, + "learning_rate": 7.919877438264589e-07, + "loss": 0.9951, + "step": 71153 + }, + { + "epoch": 5.26, + "learning_rate": 7.918321783831628e-07, + "loss": 1.0213, + "step": 71154 + }, + { + "epoch": 5.26, + "learning_rate": 7.916766275899302e-07, + "loss": 1.0012, + "step": 71155 + }, + { + "epoch": 5.26, + "learning_rate": 7.915210914470073e-07, + "loss": 0.8709, + "step": 71156 + }, + { + "epoch": 5.26, + "learning_rate": 7.913655699546419e-07, + "loss": 0.9858, + "step": 71157 + }, + { + "epoch": 5.26, + "learning_rate": 7.912100631130837e-07, + "loss": 1.0173, + "step": 71158 + }, + { + "epoch": 5.26, + "learning_rate": 7.910545709225792e-07, + "loss": 0.9101, + "step": 71159 + }, + { + "epoch": 5.26, + "learning_rate": 7.908990933833738e-07, + "loss": 0.9972, + "step": 71160 + }, + { + "epoch": 5.26, + "learning_rate": 7.907436304957172e-07, + "loss": 1.047, + "step": 71161 + }, + { + "epoch": 5.26, + "learning_rate": 7.905881822598527e-07, + "loss": 0.9939, + "step": 71162 + }, + { + "epoch": 5.26, + "learning_rate": 7.904327486760333e-07, + "loss": 0.9812, + "step": 71163 + }, + { + "epoch": 5.26, + "learning_rate": 7.902773297445044e-07, + "loss": 0.8759, + "step": 71164 + }, + { + "epoch": 5.26, + "learning_rate": 7.901219254655113e-07, + "loss": 1.0585, + "step": 71165 + }, + { + "epoch": 5.26, + "learning_rate": 7.899665358393005e-07, + "loss": 1.0101, + "step": 71166 + }, + { + "epoch": 5.26, + "learning_rate": 7.898111608661241e-07, + "loss": 0.9966, + "step": 71167 + }, + { + "epoch": 5.26, + "learning_rate": 7.896558005462252e-07, + "loss": 0.9581, + "step": 71168 + }, + { + "epoch": 5.26, + "learning_rate": 7.895004548798524e-07, + "loss": 0.9319, + "step": 71169 + }, + { + "epoch": 5.26, + "learning_rate": 7.893451238672533e-07, + "loss": 1.1012, + "step": 71170 + }, + { + "epoch": 5.26, + "learning_rate": 7.891898075086734e-07, + "loss": 0.9421, + "step": 71171 + }, + { + "epoch": 5.26, + "learning_rate": 7.890345058043613e-07, + "loss": 0.9051, + "step": 71172 + }, + { + "epoch": 5.26, + "learning_rate": 7.888792187545624e-07, + "loss": 1.098, + "step": 71173 + }, + { + "epoch": 5.26, + "learning_rate": 7.887239463595254e-07, + "loss": 1.0325, + "step": 71174 + }, + { + "epoch": 5.26, + "learning_rate": 7.885686886194976e-07, + "loss": 1.0301, + "step": 71175 + }, + { + "epoch": 5.26, + "learning_rate": 7.884134455347248e-07, + "loss": 1.0029, + "step": 71176 + }, + { + "epoch": 5.26, + "learning_rate": 7.882582171054531e-07, + "loss": 1.1344, + "step": 71177 + }, + { + "epoch": 5.26, + "learning_rate": 7.881030033319315e-07, + "loss": 1.0154, + "step": 71178 + }, + { + "epoch": 5.26, + "learning_rate": 7.879478042144073e-07, + "loss": 0.9481, + "step": 71179 + }, + { + "epoch": 5.26, + "learning_rate": 7.87792619753126e-07, + "loss": 1.0339, + "step": 71180 + }, + { + "epoch": 5.26, + "learning_rate": 7.876374499483352e-07, + "loss": 0.9465, + "step": 71181 + }, + { + "epoch": 5.26, + "learning_rate": 7.874822948002792e-07, + "loss": 0.9261, + "step": 71182 + }, + { + "epoch": 5.26, + "learning_rate": 7.873271543092087e-07, + "loss": 1.0609, + "step": 71183 + }, + { + "epoch": 5.26, + "learning_rate": 7.871720284753692e-07, + "loss": 0.964, + "step": 71184 + }, + { + "epoch": 5.26, + "learning_rate": 7.870169172990072e-07, + "loss": 0.8659, + "step": 71185 + }, + { + "epoch": 5.26, + "learning_rate": 7.868618207803669e-07, + "loss": 0.9297, + "step": 71186 + }, + { + "epoch": 5.26, + "learning_rate": 7.867067389197003e-07, + "loss": 1.1192, + "step": 71187 + }, + { + "epoch": 5.26, + "learning_rate": 7.865516717172528e-07, + "loss": 1.0187, + "step": 71188 + }, + { + "epoch": 5.26, + "learning_rate": 7.863966191732663e-07, + "loss": 0.9928, + "step": 71189 + }, + { + "epoch": 5.26, + "learning_rate": 7.862415812879932e-07, + "loss": 1.0557, + "step": 71190 + }, + { + "epoch": 5.26, + "learning_rate": 7.860865580616761e-07, + "loss": 0.9334, + "step": 71191 + }, + { + "epoch": 5.26, + "learning_rate": 7.859315494945663e-07, + "loss": 0.9655, + "step": 71192 + }, + { + "epoch": 5.26, + "learning_rate": 7.857765555869046e-07, + "loss": 0.9728, + "step": 71193 + }, + { + "epoch": 5.26, + "learning_rate": 7.856215763389419e-07, + "loss": 1.0185, + "step": 71194 + }, + { + "epoch": 5.26, + "learning_rate": 7.854666117509224e-07, + "loss": 0.9893, + "step": 71195 + }, + { + "epoch": 5.26, + "learning_rate": 7.853116618230961e-07, + "loss": 0.947, + "step": 71196 + }, + { + "epoch": 5.26, + "learning_rate": 7.851567265557059e-07, + "loss": 1.0024, + "step": 71197 + }, + { + "epoch": 5.26, + "learning_rate": 7.850018059489983e-07, + "loss": 0.9425, + "step": 71198 + }, + { + "epoch": 5.26, + "learning_rate": 7.848469000032222e-07, + "loss": 0.9403, + "step": 71199 + }, + { + "epoch": 5.26, + "learning_rate": 7.846920087186228e-07, + "loss": 1.0977, + "step": 71200 + }, + { + "epoch": 5.26, + "learning_rate": 7.845371320954465e-07, + "loss": 1.0704, + "step": 71201 + }, + { + "epoch": 5.26, + "learning_rate": 7.843822701339388e-07, + "loss": 0.9682, + "step": 71202 + }, + { + "epoch": 5.26, + "learning_rate": 7.842274228343483e-07, + "loss": 0.9383, + "step": 71203 + }, + { + "epoch": 5.26, + "learning_rate": 7.840725901969193e-07, + "loss": 1.0276, + "step": 71204 + }, + { + "epoch": 5.26, + "learning_rate": 7.839177722218994e-07, + "loss": 0.9318, + "step": 71205 + }, + { + "epoch": 5.26, + "learning_rate": 7.837629689095349e-07, + "loss": 1.0725, + "step": 71206 + }, + { + "epoch": 5.26, + "learning_rate": 7.836081802600704e-07, + "loss": 1.1875, + "step": 71207 + }, + { + "epoch": 5.26, + "learning_rate": 7.834534062737542e-07, + "loss": 1.0126, + "step": 71208 + }, + { + "epoch": 5.26, + "learning_rate": 7.832986469508319e-07, + "loss": 0.9973, + "step": 71209 + }, + { + "epoch": 5.26, + "learning_rate": 7.831439022915499e-07, + "loss": 1.0483, + "step": 71210 + }, + { + "epoch": 5.26, + "learning_rate": 7.829891722961524e-07, + "loss": 1.0674, + "step": 71211 + }, + { + "epoch": 5.26, + "learning_rate": 7.828344569648883e-07, + "loss": 0.9682, + "step": 71212 + }, + { + "epoch": 5.26, + "learning_rate": 7.826797562980037e-07, + "loss": 0.9125, + "step": 71213 + }, + { + "epoch": 5.26, + "learning_rate": 7.825250702957432e-07, + "loss": 1.1068, + "step": 71214 + }, + { + "epoch": 5.26, + "learning_rate": 7.823703989583531e-07, + "loss": 1.0008, + "step": 71215 + }, + { + "epoch": 5.26, + "learning_rate": 7.822157422860799e-07, + "loss": 0.9032, + "step": 71216 + }, + { + "epoch": 5.26, + "learning_rate": 7.8206110027917e-07, + "loss": 0.9157, + "step": 71217 + }, + { + "epoch": 5.26, + "learning_rate": 7.819064729378667e-07, + "loss": 0.9365, + "step": 71218 + }, + { + "epoch": 5.26, + "learning_rate": 7.817518602624208e-07, + "loss": 0.9543, + "step": 71219 + }, + { + "epoch": 5.26, + "learning_rate": 7.815972622530754e-07, + "loss": 1.0124, + "step": 71220 + }, + { + "epoch": 5.26, + "learning_rate": 7.81442678910077e-07, + "loss": 0.9965, + "step": 71221 + }, + { + "epoch": 5.26, + "learning_rate": 7.812881102336689e-07, + "loss": 1.0182, + "step": 71222 + }, + { + "epoch": 5.26, + "learning_rate": 7.811335562241018e-07, + "loss": 1.0227, + "step": 71223 + }, + { + "epoch": 5.26, + "learning_rate": 7.809790168816201e-07, + "loss": 0.9895, + "step": 71224 + }, + { + "epoch": 5.26, + "learning_rate": 7.808244922064678e-07, + "loss": 1.0345, + "step": 71225 + }, + { + "epoch": 5.26, + "learning_rate": 7.806699821988928e-07, + "loss": 1.0585, + "step": 71226 + }, + { + "epoch": 5.26, + "learning_rate": 7.805154868591369e-07, + "loss": 0.9571, + "step": 71227 + }, + { + "epoch": 5.26, + "learning_rate": 7.803610061874523e-07, + "loss": 0.9857, + "step": 71228 + }, + { + "epoch": 5.26, + "learning_rate": 7.802065401840797e-07, + "loss": 1.0124, + "step": 71229 + }, + { + "epoch": 5.26, + "learning_rate": 7.80052088849268e-07, + "loss": 1.1057, + "step": 71230 + }, + { + "epoch": 5.26, + "learning_rate": 7.798976521832591e-07, + "loss": 0.965, + "step": 71231 + }, + { + "epoch": 5.26, + "learning_rate": 7.797432301863028e-07, + "loss": 1.1493, + "step": 71232 + }, + { + "epoch": 5.26, + "learning_rate": 7.795888228586435e-07, + "loss": 1.0367, + "step": 71233 + }, + { + "epoch": 5.26, + "learning_rate": 7.794344302005263e-07, + "loss": 0.8607, + "step": 71234 + }, + { + "epoch": 5.26, + "learning_rate": 7.792800522121968e-07, + "loss": 0.9716, + "step": 71235 + }, + { + "epoch": 5.26, + "learning_rate": 7.791256888938992e-07, + "loss": 1.0365, + "step": 71236 + }, + { + "epoch": 5.26, + "learning_rate": 7.789713402458842e-07, + "loss": 0.947, + "step": 71237 + }, + { + "epoch": 5.26, + "learning_rate": 7.788170062683909e-07, + "loss": 1.0068, + "step": 71238 + }, + { + "epoch": 5.26, + "learning_rate": 7.786626869616687e-07, + "loss": 1.1369, + "step": 71239 + }, + { + "epoch": 5.26, + "learning_rate": 7.785083823259609e-07, + "loss": 1.0633, + "step": 71240 + }, + { + "epoch": 5.26, + "learning_rate": 7.783540923615174e-07, + "loss": 1.0221, + "step": 71241 + }, + { + "epoch": 5.26, + "learning_rate": 7.781998170685778e-07, + "loss": 0.9346, + "step": 71242 + }, + { + "epoch": 5.26, + "learning_rate": 7.78045556447391e-07, + "loss": 0.9902, + "step": 71243 + }, + { + "epoch": 5.26, + "learning_rate": 7.778913104982033e-07, + "loss": 0.9744, + "step": 71244 + }, + { + "epoch": 5.26, + "learning_rate": 7.777370792212569e-07, + "loss": 0.9793, + "step": 71245 + }, + { + "epoch": 5.26, + "learning_rate": 7.775828626168002e-07, + "loss": 1.0144, + "step": 71246 + }, + { + "epoch": 5.26, + "learning_rate": 7.774286606850745e-07, + "loss": 1.1005, + "step": 71247 + }, + { + "epoch": 5.26, + "learning_rate": 7.772744734263304e-07, + "loss": 1.0589, + "step": 71248 + }, + { + "epoch": 5.26, + "learning_rate": 7.771203008408101e-07, + "loss": 0.9857, + "step": 71249 + }, + { + "epoch": 5.26, + "learning_rate": 7.7696614292876e-07, + "loss": 1.0046, + "step": 71250 + }, + { + "epoch": 5.26, + "learning_rate": 7.768119996904234e-07, + "loss": 0.9927, + "step": 71251 + }, + { + "epoch": 5.26, + "learning_rate": 7.766578711260475e-07, + "loss": 1.0202, + "step": 71252 + }, + { + "epoch": 5.26, + "learning_rate": 7.765037572358769e-07, + "loss": 1.0309, + "step": 71253 + }, + { + "epoch": 5.26, + "learning_rate": 7.763496580201579e-07, + "loss": 0.9831, + "step": 71254 + }, + { + "epoch": 5.26, + "learning_rate": 7.761955734791337e-07, + "loss": 0.9618, + "step": 71255 + }, + { + "epoch": 5.27, + "learning_rate": 7.760415036130498e-07, + "loss": 1.0251, + "step": 71256 + }, + { + "epoch": 5.27, + "learning_rate": 7.758874484221524e-07, + "loss": 0.9108, + "step": 71257 + }, + { + "epoch": 5.27, + "learning_rate": 7.757334079066869e-07, + "loss": 0.9837, + "step": 71258 + }, + { + "epoch": 5.27, + "learning_rate": 7.755793820668977e-07, + "loss": 1.0302, + "step": 71259 + }, + { + "epoch": 5.27, + "learning_rate": 7.754253709030279e-07, + "loss": 0.9871, + "step": 71260 + }, + { + "epoch": 5.27, + "learning_rate": 7.752713744153273e-07, + "loss": 0.899, + "step": 71261 + }, + { + "epoch": 5.27, + "learning_rate": 7.751173926040368e-07, + "loss": 0.9328, + "step": 71262 + }, + { + "epoch": 5.27, + "learning_rate": 7.749634254693994e-07, + "loss": 1.0408, + "step": 71263 + }, + { + "epoch": 5.27, + "learning_rate": 7.748094730116662e-07, + "loss": 1.0223, + "step": 71264 + }, + { + "epoch": 5.27, + "learning_rate": 7.746555352310792e-07, + "loss": 1.0574, + "step": 71265 + }, + { + "epoch": 5.27, + "learning_rate": 7.745016121278826e-07, + "loss": 0.9196, + "step": 71266 + }, + { + "epoch": 5.27, + "learning_rate": 7.743477037023206e-07, + "loss": 0.9231, + "step": 71267 + }, + { + "epoch": 5.27, + "learning_rate": 7.741938099546419e-07, + "loss": 1.0139, + "step": 71268 + }, + { + "epoch": 5.27, + "learning_rate": 7.740399308850876e-07, + "loss": 0.9772, + "step": 71269 + }, + { + "epoch": 5.27, + "learning_rate": 7.73886066493904e-07, + "loss": 1.113, + "step": 71270 + }, + { + "epoch": 5.27, + "learning_rate": 7.737322167813366e-07, + "loss": 1.0572, + "step": 71271 + }, + { + "epoch": 5.27, + "learning_rate": 7.735783817476261e-07, + "loss": 1.0208, + "step": 71272 + }, + { + "epoch": 5.27, + "learning_rate": 7.734245613930235e-07, + "loss": 1.0702, + "step": 71273 + }, + { + "epoch": 5.27, + "learning_rate": 7.732707557177699e-07, + "loss": 1.0665, + "step": 71274 + }, + { + "epoch": 5.27, + "learning_rate": 7.731169647221104e-07, + "loss": 1.0735, + "step": 71275 + }, + { + "epoch": 5.27, + "learning_rate": 7.729631884062871e-07, + "loss": 1.1363, + "step": 71276 + }, + { + "epoch": 5.27, + "learning_rate": 7.728094267705499e-07, + "loss": 1.0256, + "step": 71277 + }, + { + "epoch": 5.27, + "learning_rate": 7.726556798151419e-07, + "loss": 1.0539, + "step": 71278 + }, + { + "epoch": 5.27, + "learning_rate": 7.72501947540305e-07, + "loss": 0.9973, + "step": 71279 + }, + { + "epoch": 5.27, + "learning_rate": 7.723482299462848e-07, + "loss": 0.9242, + "step": 71280 + }, + { + "epoch": 5.27, + "learning_rate": 7.721945270333287e-07, + "loss": 0.9299, + "step": 71281 + }, + { + "epoch": 5.27, + "learning_rate": 7.720408388016798e-07, + "loss": 1.0092, + "step": 71282 + }, + { + "epoch": 5.27, + "learning_rate": 7.71887165251578e-07, + "loss": 0.9992, + "step": 71283 + }, + { + "epoch": 5.27, + "learning_rate": 7.717335063832743e-07, + "loss": 0.8534, + "step": 71284 + }, + { + "epoch": 5.27, + "learning_rate": 7.715798621970094e-07, + "loss": 0.9752, + "step": 71285 + }, + { + "epoch": 5.27, + "learning_rate": 7.71426232693031e-07, + "loss": 1.0973, + "step": 71286 + }, + { + "epoch": 5.27, + "learning_rate": 7.71272617871579e-07, + "loss": 0.9786, + "step": 71287 + }, + { + "epoch": 5.27, + "learning_rate": 7.711190177329008e-07, + "loss": 1.0523, + "step": 71288 + }, + { + "epoch": 5.27, + "learning_rate": 7.709654322772398e-07, + "loss": 1.1676, + "step": 71289 + }, + { + "epoch": 5.27, + "learning_rate": 7.708118615048432e-07, + "loss": 1.0641, + "step": 71290 + }, + { + "epoch": 5.27, + "learning_rate": 7.706583054159522e-07, + "loss": 0.9831, + "step": 71291 + }, + { + "epoch": 5.27, + "learning_rate": 7.705047640108098e-07, + "loss": 0.9298, + "step": 71292 + }, + { + "epoch": 5.27, + "learning_rate": 7.703512372896638e-07, + "loss": 0.9675, + "step": 71293 + }, + { + "epoch": 5.27, + "learning_rate": 7.70197725252756e-07, + "loss": 1.0572, + "step": 71294 + }, + { + "epoch": 5.27, + "learning_rate": 7.700442279003329e-07, + "loss": 0.941, + "step": 71295 + }, + { + "epoch": 5.27, + "learning_rate": 7.698907452326343e-07, + "loss": 0.9593, + "step": 71296 + }, + { + "epoch": 5.27, + "learning_rate": 7.697372772499101e-07, + "loss": 0.9563, + "step": 71297 + }, + { + "epoch": 5.27, + "learning_rate": 7.695838239524022e-07, + "loss": 0.8687, + "step": 71298 + }, + { + "epoch": 5.27, + "learning_rate": 7.694303853403539e-07, + "loss": 0.9974, + "step": 71299 + }, + { + "epoch": 5.27, + "learning_rate": 7.692769614140094e-07, + "loss": 0.94, + "step": 71300 + }, + { + "epoch": 5.27, + "learning_rate": 7.691235521736118e-07, + "loss": 0.9136, + "step": 71301 + }, + { + "epoch": 5.27, + "learning_rate": 7.689701576194086e-07, + "loss": 1.0786, + "step": 71302 + }, + { + "epoch": 5.27, + "learning_rate": 7.688167777516409e-07, + "loss": 0.9843, + "step": 71303 + }, + { + "epoch": 5.27, + "learning_rate": 7.68663412570555e-07, + "loss": 1.1005, + "step": 71304 + }, + { + "epoch": 5.27, + "learning_rate": 7.685100620763908e-07, + "loss": 0.9731, + "step": 71305 + }, + { + "epoch": 5.27, + "learning_rate": 7.683567262693992e-07, + "loss": 1.046, + "step": 71306 + }, + { + "epoch": 5.27, + "learning_rate": 7.682034051498155e-07, + "loss": 1.0935, + "step": 71307 + }, + { + "epoch": 5.27, + "learning_rate": 7.680500987178907e-07, + "loss": 0.9489, + "step": 71308 + }, + { + "epoch": 5.27, + "learning_rate": 7.678968069738657e-07, + "loss": 1.042, + "step": 71309 + }, + { + "epoch": 5.27, + "learning_rate": 7.677435299179848e-07, + "loss": 1.0047, + "step": 71310 + }, + { + "epoch": 5.27, + "learning_rate": 7.67590267550492e-07, + "loss": 0.9126, + "step": 71311 + }, + { + "epoch": 5.27, + "learning_rate": 7.674370198716285e-07, + "loss": 1.0641, + "step": 71312 + }, + { + "epoch": 5.27, + "learning_rate": 7.672837868816429e-07, + "loss": 1.0576, + "step": 71313 + }, + { + "epoch": 5.27, + "learning_rate": 7.671305685807773e-07, + "loss": 1.0117, + "step": 71314 + }, + { + "epoch": 5.27, + "learning_rate": 7.669773649692747e-07, + "loss": 1.029, + "step": 71315 + }, + { + "epoch": 5.27, + "learning_rate": 7.668241760473761e-07, + "loss": 1.1147, + "step": 71316 + }, + { + "epoch": 5.27, + "learning_rate": 7.666710018153312e-07, + "loss": 0.8744, + "step": 71317 + }, + { + "epoch": 5.27, + "learning_rate": 7.665178422733799e-07, + "loss": 1.0404, + "step": 71318 + }, + { + "epoch": 5.27, + "learning_rate": 7.663646974217664e-07, + "loss": 0.9506, + "step": 71319 + }, + { + "epoch": 5.27, + "learning_rate": 7.662115672607351e-07, + "loss": 1.0204, + "step": 71320 + }, + { + "epoch": 5.27, + "learning_rate": 7.660584517905278e-07, + "loss": 1.0305, + "step": 71321 + }, + { + "epoch": 5.27, + "learning_rate": 7.6590535101139e-07, + "loss": 0.9985, + "step": 71322 + }, + { + "epoch": 5.27, + "learning_rate": 7.657522649235661e-07, + "loss": 1.0344, + "step": 71323 + }, + { + "epoch": 5.27, + "learning_rate": 7.655991935272966e-07, + "loss": 1.0212, + "step": 71324 + }, + { + "epoch": 5.27, + "learning_rate": 7.654461368228261e-07, + "loss": 0.9813, + "step": 71325 + }, + { + "epoch": 5.27, + "learning_rate": 7.652930948103998e-07, + "loss": 0.9432, + "step": 71326 + }, + { + "epoch": 5.27, + "learning_rate": 7.651400674902598e-07, + "loss": 0.937, + "step": 71327 + }, + { + "epoch": 5.27, + "learning_rate": 7.649870548626504e-07, + "loss": 1.0494, + "step": 71328 + }, + { + "epoch": 5.27, + "learning_rate": 7.648340569278145e-07, + "loss": 1.0281, + "step": 71329 + }, + { + "epoch": 5.27, + "learning_rate": 7.646810736859934e-07, + "loss": 0.9723, + "step": 71330 + }, + { + "epoch": 5.27, + "learning_rate": 7.645281051374364e-07, + "loss": 1.0096, + "step": 71331 + }, + { + "epoch": 5.27, + "learning_rate": 7.643751512823794e-07, + "loss": 1.1081, + "step": 71332 + }, + { + "epoch": 5.27, + "learning_rate": 7.642222121210707e-07, + "loss": 1.0434, + "step": 71333 + }, + { + "epoch": 5.27, + "learning_rate": 7.640692876537515e-07, + "loss": 1.0762, + "step": 71334 + }, + { + "epoch": 5.27, + "learning_rate": 7.639163778806691e-07, + "loss": 1.002, + "step": 71335 + }, + { + "epoch": 5.27, + "learning_rate": 7.637634828020601e-07, + "loss": 0.985, + "step": 71336 + }, + { + "epoch": 5.27, + "learning_rate": 7.636106024181722e-07, + "loss": 0.9821, + "step": 71337 + }, + { + "epoch": 5.27, + "learning_rate": 7.634577367292484e-07, + "loss": 0.9613, + "step": 71338 + }, + { + "epoch": 5.27, + "learning_rate": 7.633048857355307e-07, + "loss": 0.9245, + "step": 71339 + }, + { + "epoch": 5.27, + "learning_rate": 7.631520494372635e-07, + "loss": 0.978, + "step": 71340 + }, + { + "epoch": 5.27, + "learning_rate": 7.629992278346876e-07, + "loss": 1.004, + "step": 71341 + }, + { + "epoch": 5.27, + "learning_rate": 7.628464209280484e-07, + "loss": 0.9933, + "step": 71342 + }, + { + "epoch": 5.27, + "learning_rate": 7.62693628717589e-07, + "loss": 0.9634, + "step": 71343 + }, + { + "epoch": 5.27, + "learning_rate": 7.625408512035526e-07, + "loss": 0.9602, + "step": 71344 + }, + { + "epoch": 5.27, + "learning_rate": 7.623880883861789e-07, + "loss": 0.9517, + "step": 71345 + }, + { + "epoch": 5.27, + "learning_rate": 7.622353402657157e-07, + "loss": 0.8818, + "step": 71346 + }, + { + "epoch": 5.27, + "learning_rate": 7.620826068424037e-07, + "loss": 0.9044, + "step": 71347 + }, + { + "epoch": 5.27, + "learning_rate": 7.61929888116486e-07, + "loss": 0.9097, + "step": 71348 + }, + { + "epoch": 5.27, + "learning_rate": 7.61777184088206e-07, + "loss": 1.0032, + "step": 71349 + }, + { + "epoch": 5.27, + "learning_rate": 7.616244947578044e-07, + "loss": 1.031, + "step": 71350 + }, + { + "epoch": 5.27, + "learning_rate": 7.614718201255289e-07, + "loss": 1.0462, + "step": 71351 + }, + { + "epoch": 5.27, + "learning_rate": 7.61319160191617e-07, + "loss": 1.0042, + "step": 71352 + }, + { + "epoch": 5.27, + "learning_rate": 7.611665149563152e-07, + "loss": 1.0125, + "step": 71353 + }, + { + "epoch": 5.27, + "learning_rate": 7.610138844198645e-07, + "loss": 1.0987, + "step": 71354 + }, + { + "epoch": 5.27, + "learning_rate": 7.608612685825112e-07, + "loss": 0.9755, + "step": 71355 + }, + { + "epoch": 5.27, + "learning_rate": 7.607086674444942e-07, + "loss": 1.0248, + "step": 71356 + }, + { + "epoch": 5.27, + "learning_rate": 7.605560810060552e-07, + "loss": 1.0093, + "step": 71357 + }, + { + "epoch": 5.27, + "learning_rate": 7.604035092674411e-07, + "loss": 1.0167, + "step": 71358 + }, + { + "epoch": 5.27, + "learning_rate": 7.602509522288926e-07, + "loss": 0.9644, + "step": 71359 + }, + { + "epoch": 5.27, + "learning_rate": 7.600984098906539e-07, + "loss": 1.024, + "step": 71360 + }, + { + "epoch": 5.27, + "learning_rate": 7.599458822529638e-07, + "loss": 0.9414, + "step": 71361 + }, + { + "epoch": 5.27, + "learning_rate": 7.597933693160686e-07, + "loss": 1.0032, + "step": 71362 + }, + { + "epoch": 5.27, + "learning_rate": 7.596408710802117e-07, + "loss": 0.9416, + "step": 71363 + }, + { + "epoch": 5.27, + "learning_rate": 7.594883875456327e-07, + "loss": 0.8587, + "step": 71364 + }, + { + "epoch": 5.27, + "learning_rate": 7.593359187125759e-07, + "loss": 0.9372, + "step": 71365 + }, + { + "epoch": 5.27, + "learning_rate": 7.591834645812812e-07, + "loss": 1.0735, + "step": 71366 + }, + { + "epoch": 5.27, + "learning_rate": 7.59031025151995e-07, + "loss": 0.9678, + "step": 71367 + }, + { + "epoch": 5.27, + "learning_rate": 7.588786004249593e-07, + "loss": 0.959, + "step": 71368 + }, + { + "epoch": 5.27, + "learning_rate": 7.587261904004151e-07, + "loss": 1.0742, + "step": 71369 + }, + { + "epoch": 5.27, + "learning_rate": 7.585737950786032e-07, + "loss": 1.0476, + "step": 71370 + }, + { + "epoch": 5.27, + "learning_rate": 7.584214144597701e-07, + "loss": 0.9743, + "step": 71371 + }, + { + "epoch": 5.27, + "learning_rate": 7.582690485441557e-07, + "loss": 1.0688, + "step": 71372 + }, + { + "epoch": 5.27, + "learning_rate": 7.581166973320042e-07, + "loss": 0.9463, + "step": 71373 + }, + { + "epoch": 5.27, + "learning_rate": 7.579643608235566e-07, + "loss": 0.9255, + "step": 71374 + }, + { + "epoch": 5.27, + "learning_rate": 7.578120390190535e-07, + "loss": 0.9515, + "step": 71375 + }, + { + "epoch": 5.27, + "learning_rate": 7.576597319187418e-07, + "loss": 0.9236, + "step": 71376 + }, + { + "epoch": 5.27, + "learning_rate": 7.575074395228588e-07, + "loss": 0.9594, + "step": 71377 + }, + { + "epoch": 5.27, + "learning_rate": 7.573551618316511e-07, + "loss": 0.912, + "step": 71378 + }, + { + "epoch": 5.27, + "learning_rate": 7.572028988453561e-07, + "loss": 1.0068, + "step": 71379 + }, + { + "epoch": 5.27, + "learning_rate": 7.570506505642239e-07, + "loss": 0.9873, + "step": 71380 + }, + { + "epoch": 5.27, + "learning_rate": 7.568984169884874e-07, + "loss": 1.061, + "step": 71381 + }, + { + "epoch": 5.27, + "learning_rate": 7.567461981183955e-07, + "loss": 0.9396, + "step": 71382 + }, + { + "epoch": 5.27, + "learning_rate": 7.565939939541889e-07, + "loss": 0.9621, + "step": 71383 + }, + { + "epoch": 5.27, + "learning_rate": 7.564418044961075e-07, + "loss": 0.9071, + "step": 71384 + }, + { + "epoch": 5.27, + "learning_rate": 7.562896297443956e-07, + "loss": 0.9372, + "step": 71385 + }, + { + "epoch": 5.27, + "learning_rate": 7.56137469699293e-07, + "loss": 0.9801, + "step": 71386 + }, + { + "epoch": 5.27, + "learning_rate": 7.559853243610449e-07, + "loss": 1.0141, + "step": 71387 + }, + { + "epoch": 5.27, + "learning_rate": 7.558331937298924e-07, + "loss": 0.992, + "step": 71388 + }, + { + "epoch": 5.27, + "learning_rate": 7.556810778060764e-07, + "loss": 1.0421, + "step": 71389 + }, + { + "epoch": 5.27, + "learning_rate": 7.555289765898388e-07, + "loss": 1.033, + "step": 71390 + }, + { + "epoch": 5.28, + "learning_rate": 7.553768900814229e-07, + "loss": 0.9795, + "step": 71391 + }, + { + "epoch": 5.28, + "learning_rate": 7.552248182810706e-07, + "loss": 0.9019, + "step": 71392 + }, + { + "epoch": 5.28, + "learning_rate": 7.55072761189024e-07, + "loss": 0.9525, + "step": 71393 + }, + { + "epoch": 5.28, + "learning_rate": 7.54920718805523e-07, + "loss": 1.0329, + "step": 71394 + }, + { + "epoch": 5.28, + "learning_rate": 7.547686911308106e-07, + "loss": 0.9534, + "step": 71395 + }, + { + "epoch": 5.28, + "learning_rate": 7.54616678165131e-07, + "loss": 0.9506, + "step": 71396 + }, + { + "epoch": 5.28, + "learning_rate": 7.544646799087219e-07, + "loss": 1.0255, + "step": 71397 + }, + { + "epoch": 5.28, + "learning_rate": 7.543126963618275e-07, + "loss": 1.0134, + "step": 71398 + }, + { + "epoch": 5.28, + "learning_rate": 7.541607275246887e-07, + "loss": 1.0146, + "step": 71399 + }, + { + "epoch": 5.28, + "learning_rate": 7.54008773397551e-07, + "loss": 1.0024, + "step": 71400 + }, + { + "epoch": 5.28, + "learning_rate": 7.538568339806495e-07, + "loss": 1.004, + "step": 71401 + }, + { + "epoch": 5.28, + "learning_rate": 7.53704909274231e-07, + "loss": 0.9719, + "step": 71402 + }, + { + "epoch": 5.28, + "learning_rate": 7.535529992785362e-07, + "loss": 1.1912, + "step": 71403 + }, + { + "epoch": 5.28, + "learning_rate": 7.53401103993806e-07, + "loss": 0.9137, + "step": 71404 + }, + { + "epoch": 5.28, + "learning_rate": 7.532492234202826e-07, + "loss": 0.9659, + "step": 71405 + }, + { + "epoch": 5.28, + "learning_rate": 7.530973575582057e-07, + "loss": 0.9801, + "step": 71406 + }, + { + "epoch": 5.28, + "learning_rate": 7.529455064078195e-07, + "loss": 1.0353, + "step": 71407 + }, + { + "epoch": 5.28, + "learning_rate": 7.52793669969365e-07, + "loss": 1.0599, + "step": 71408 + }, + { + "epoch": 5.28, + "learning_rate": 7.526418482430841e-07, + "loss": 0.9729, + "step": 71409 + }, + { + "epoch": 5.28, + "learning_rate": 7.524900412292158e-07, + "loss": 0.9691, + "step": 71410 + }, + { + "epoch": 5.28, + "learning_rate": 7.523382489280051e-07, + "loss": 1.0703, + "step": 71411 + }, + { + "epoch": 5.28, + "learning_rate": 7.52186471339692e-07, + "loss": 1.0471, + "step": 71412 + }, + { + "epoch": 5.28, + "learning_rate": 7.520347084645174e-07, + "loss": 0.9181, + "step": 71413 + }, + { + "epoch": 5.28, + "learning_rate": 7.518829603027233e-07, + "loss": 1.0352, + "step": 71414 + }, + { + "epoch": 5.28, + "learning_rate": 7.517312268545507e-07, + "loss": 0.9456, + "step": 71415 + }, + { + "epoch": 5.28, + "learning_rate": 7.515795081202415e-07, + "loss": 1.0508, + "step": 71416 + }, + { + "epoch": 5.28, + "learning_rate": 7.514278041000377e-07, + "loss": 0.9437, + "step": 71417 + }, + { + "epoch": 5.28, + "learning_rate": 7.512761147941805e-07, + "loss": 1.0159, + "step": 71418 + }, + { + "epoch": 5.28, + "learning_rate": 7.511244402029084e-07, + "loss": 0.9669, + "step": 71419 + }, + { + "epoch": 5.28, + "learning_rate": 7.509727803264665e-07, + "loss": 0.8735, + "step": 71420 + }, + { + "epoch": 5.28, + "learning_rate": 7.508211351650962e-07, + "loss": 1.0176, + "step": 71421 + }, + { + "epoch": 5.28, + "learning_rate": 7.506695047190338e-07, + "loss": 1.0017, + "step": 71422 + }, + { + "epoch": 5.28, + "learning_rate": 7.505178889885256e-07, + "loss": 1.0009, + "step": 71423 + }, + { + "epoch": 5.28, + "learning_rate": 7.503662879738094e-07, + "loss": 1.1062, + "step": 71424 + }, + { + "epoch": 5.28, + "learning_rate": 7.502147016751315e-07, + "loss": 1.0295, + "step": 71425 + }, + { + "epoch": 5.28, + "learning_rate": 7.500631300927275e-07, + "loss": 1.0236, + "step": 71426 + }, + { + "epoch": 5.28, + "learning_rate": 7.499115732268403e-07, + "loss": 1.0992, + "step": 71427 + }, + { + "epoch": 5.28, + "learning_rate": 7.497600310777131e-07, + "loss": 0.9814, + "step": 71428 + }, + { + "epoch": 5.28, + "learning_rate": 7.496085036455847e-07, + "loss": 0.9769, + "step": 71429 + }, + { + "epoch": 5.28, + "learning_rate": 7.494569909306971e-07, + "loss": 0.9878, + "step": 71430 + }, + { + "epoch": 5.28, + "learning_rate": 7.493054929332888e-07, + "loss": 0.9818, + "step": 71431 + }, + { + "epoch": 5.28, + "learning_rate": 7.491540096536054e-07, + "loss": 1.0734, + "step": 71432 + }, + { + "epoch": 5.28, + "learning_rate": 7.490025410918844e-07, + "loss": 1.0553, + "step": 71433 + }, + { + "epoch": 5.28, + "learning_rate": 7.48851087248369e-07, + "loss": 1.0453, + "step": 71434 + }, + { + "epoch": 5.28, + "learning_rate": 7.486996481232977e-07, + "loss": 0.9985, + "step": 71435 + }, + { + "epoch": 5.28, + "learning_rate": 7.485482237169139e-07, + "loss": 1.0113, + "step": 71436 + }, + { + "epoch": 5.28, + "learning_rate": 7.483968140294584e-07, + "loss": 1.0571, + "step": 71437 + }, + { + "epoch": 5.28, + "learning_rate": 7.48245419061171e-07, + "loss": 1.0123, + "step": 71438 + }, + { + "epoch": 5.28, + "learning_rate": 7.480940388122904e-07, + "loss": 1.0469, + "step": 71439 + }, + { + "epoch": 5.28, + "learning_rate": 7.47942673283062e-07, + "loss": 1.0849, + "step": 71440 + }, + { + "epoch": 5.28, + "learning_rate": 7.477913224737254e-07, + "loss": 1.1041, + "step": 71441 + }, + { + "epoch": 5.28, + "learning_rate": 7.476399863845185e-07, + "loss": 1.0108, + "step": 71442 + }, + { + "epoch": 5.28, + "learning_rate": 7.474886650156843e-07, + "loss": 1.0186, + "step": 71443 + }, + { + "epoch": 5.28, + "learning_rate": 7.473373583674626e-07, + "loss": 0.8661, + "step": 71444 + }, + { + "epoch": 5.28, + "learning_rate": 7.471860664400977e-07, + "loss": 0.942, + "step": 71445 + }, + { + "epoch": 5.28, + "learning_rate": 7.470347892338237e-07, + "loss": 1.0258, + "step": 71446 + }, + { + "epoch": 5.28, + "learning_rate": 7.468835267488861e-07, + "loss": 1.0959, + "step": 71447 + }, + { + "epoch": 5.28, + "learning_rate": 7.467322789855235e-07, + "loss": 1.0902, + "step": 71448 + }, + { + "epoch": 5.28, + "learning_rate": 7.465810459439815e-07, + "loss": 0.9506, + "step": 71449 + }, + { + "epoch": 5.28, + "learning_rate": 7.464298276244942e-07, + "loss": 0.9775, + "step": 71450 + }, + { + "epoch": 5.28, + "learning_rate": 7.462786240273035e-07, + "loss": 0.9574, + "step": 71451 + }, + { + "epoch": 5.28, + "learning_rate": 7.461274351526515e-07, + "loss": 1.0175, + "step": 71452 + }, + { + "epoch": 5.28, + "learning_rate": 7.459762610007793e-07, + "loss": 0.9415, + "step": 71453 + }, + { + "epoch": 5.28, + "learning_rate": 7.458251015719264e-07, + "loss": 0.9445, + "step": 71454 + }, + { + "epoch": 5.28, + "learning_rate": 7.456739568663318e-07, + "loss": 0.9512, + "step": 71455 + }, + { + "epoch": 5.28, + "learning_rate": 7.455228268842396e-07, + "loss": 1.0085, + "step": 71456 + }, + { + "epoch": 5.28, + "learning_rate": 7.453717116258874e-07, + "loss": 0.9954, + "step": 71457 + }, + { + "epoch": 5.28, + "learning_rate": 7.452206110915161e-07, + "loss": 1.1462, + "step": 71458 + }, + { + "epoch": 5.28, + "learning_rate": 7.450695252813678e-07, + "loss": 0.9638, + "step": 71459 + }, + { + "epoch": 5.28, + "learning_rate": 7.449184541956789e-07, + "loss": 0.9542, + "step": 71460 + }, + { + "epoch": 5.28, + "learning_rate": 7.447673978346936e-07, + "loss": 1.0296, + "step": 71461 + }, + { + "epoch": 5.28, + "learning_rate": 7.44616356198652e-07, + "loss": 1.0531, + "step": 71462 + }, + { + "epoch": 5.28, + "learning_rate": 7.444653292877924e-07, + "loss": 1.0173, + "step": 71463 + }, + { + "epoch": 5.28, + "learning_rate": 7.443143171023548e-07, + "loss": 0.994, + "step": 71464 + }, + { + "epoch": 5.28, + "learning_rate": 7.441633196425824e-07, + "loss": 0.9385, + "step": 71465 + }, + { + "epoch": 5.28, + "learning_rate": 7.440123369087127e-07, + "loss": 0.9274, + "step": 71466 + }, + { + "epoch": 5.28, + "learning_rate": 7.438613689009888e-07, + "loss": 0.9101, + "step": 71467 + }, + { + "epoch": 5.28, + "learning_rate": 7.437104156196473e-07, + "loss": 1.0902, + "step": 71468 + }, + { + "epoch": 5.28, + "learning_rate": 7.435594770649312e-07, + "loss": 1.0071, + "step": 71469 + }, + { + "epoch": 5.28, + "learning_rate": 7.434085532370794e-07, + "loss": 1.0423, + "step": 71470 + }, + { + "epoch": 5.28, + "learning_rate": 7.432576441363304e-07, + "loss": 0.9004, + "step": 71471 + }, + { + "epoch": 5.28, + "learning_rate": 7.431067497629274e-07, + "loss": 1.0023, + "step": 71472 + }, + { + "epoch": 5.28, + "learning_rate": 7.429558701171091e-07, + "loss": 0.915, + "step": 71473 + }, + { + "epoch": 5.28, + "learning_rate": 7.428050051991154e-07, + "loss": 0.9702, + "step": 71474 + }, + { + "epoch": 5.28, + "learning_rate": 7.426541550091859e-07, + "loss": 1.0357, + "step": 71475 + }, + { + "epoch": 5.28, + "learning_rate": 7.425033195475618e-07, + "loss": 1.0379, + "step": 71476 + }, + { + "epoch": 5.28, + "learning_rate": 7.423524988144826e-07, + "loss": 0.999, + "step": 71477 + }, + { + "epoch": 5.28, + "learning_rate": 7.422016928101882e-07, + "loss": 1.0145, + "step": 71478 + }, + { + "epoch": 5.28, + "learning_rate": 7.420509015349186e-07, + "loss": 0.9421, + "step": 71479 + }, + { + "epoch": 5.28, + "learning_rate": 7.419001249889124e-07, + "loss": 1.0736, + "step": 71480 + }, + { + "epoch": 5.28, + "learning_rate": 7.417493631724115e-07, + "loss": 1.0707, + "step": 71481 + }, + { + "epoch": 5.28, + "learning_rate": 7.415986160856547e-07, + "loss": 0.9408, + "step": 71482 + }, + { + "epoch": 5.28, + "learning_rate": 7.41447883728883e-07, + "loss": 1.0226, + "step": 71483 + }, + { + "epoch": 5.28, + "learning_rate": 7.412971661023327e-07, + "loss": 0.9316, + "step": 71484 + }, + { + "epoch": 5.28, + "learning_rate": 7.411464632062482e-07, + "loss": 0.9804, + "step": 71485 + }, + { + "epoch": 5.28, + "learning_rate": 7.409957750408691e-07, + "loss": 1.1158, + "step": 71486 + }, + { + "epoch": 5.28, + "learning_rate": 7.408451016064299e-07, + "loss": 0.9762, + "step": 71487 + }, + { + "epoch": 5.28, + "learning_rate": 7.406944429031748e-07, + "loss": 0.9213, + "step": 71488 + }, + { + "epoch": 5.28, + "learning_rate": 7.405437989313413e-07, + "loss": 0.9393, + "step": 71489 + }, + { + "epoch": 5.28, + "learning_rate": 7.403931696911726e-07, + "loss": 0.94, + "step": 71490 + }, + { + "epoch": 5.28, + "learning_rate": 7.40242555182904e-07, + "loss": 1.0812, + "step": 71491 + }, + { + "epoch": 5.28, + "learning_rate": 7.400919554067776e-07, + "loss": 0.9054, + "step": 71492 + }, + { + "epoch": 5.28, + "learning_rate": 7.39941370363031e-07, + "loss": 1.0737, + "step": 71493 + }, + { + "epoch": 5.28, + "learning_rate": 7.397908000519083e-07, + "loss": 1.0023, + "step": 71494 + }, + { + "epoch": 5.28, + "learning_rate": 7.396402444736428e-07, + "loss": 0.9285, + "step": 71495 + }, + { + "epoch": 5.28, + "learning_rate": 7.394897036284787e-07, + "loss": 1.103, + "step": 71496 + }, + { + "epoch": 5.28, + "learning_rate": 7.393391775166537e-07, + "loss": 0.9509, + "step": 71497 + }, + { + "epoch": 5.28, + "learning_rate": 7.391886661384085e-07, + "loss": 0.882, + "step": 71498 + }, + { + "epoch": 5.28, + "learning_rate": 7.390381694939807e-07, + "loss": 0.9676, + "step": 71499 + }, + { + "epoch": 5.28, + "learning_rate": 7.388876875836092e-07, + "loss": 1.0514, + "step": 71500 + }, + { + "epoch": 5.28, + "learning_rate": 7.387372204075371e-07, + "loss": 1.1128, + "step": 71501 + }, + { + "epoch": 5.28, + "learning_rate": 7.38586767966002e-07, + "loss": 0.9097, + "step": 71502 + }, + { + "epoch": 5.28, + "learning_rate": 7.384363302592412e-07, + "loss": 1.0476, + "step": 71503 + }, + { + "epoch": 5.28, + "learning_rate": 7.382859072874948e-07, + "loss": 1.0421, + "step": 71504 + }, + { + "epoch": 5.28, + "learning_rate": 7.381354990510048e-07, + "loss": 1.0246, + "step": 71505 + }, + { + "epoch": 5.28, + "learning_rate": 7.379851055500087e-07, + "loss": 0.9962, + "step": 71506 + }, + { + "epoch": 5.28, + "learning_rate": 7.378347267847463e-07, + "loss": 1.0083, + "step": 71507 + }, + { + "epoch": 5.28, + "learning_rate": 7.376843627554553e-07, + "loss": 1.0071, + "step": 71508 + }, + { + "epoch": 5.28, + "learning_rate": 7.375340134623754e-07, + "loss": 1.0272, + "step": 71509 + }, + { + "epoch": 5.28, + "learning_rate": 7.373836789057476e-07, + "loss": 0.8984, + "step": 71510 + }, + { + "epoch": 5.28, + "learning_rate": 7.372333590858094e-07, + "loss": 0.9092, + "step": 71511 + }, + { + "epoch": 5.28, + "learning_rate": 7.370830540028007e-07, + "loss": 1.0217, + "step": 71512 + }, + { + "epoch": 5.28, + "learning_rate": 7.36932763656959e-07, + "loss": 0.8957, + "step": 71513 + }, + { + "epoch": 5.28, + "learning_rate": 7.367824880485275e-07, + "loss": 0.838, + "step": 71514 + }, + { + "epoch": 5.28, + "learning_rate": 7.366322271777404e-07, + "loss": 1.1036, + "step": 71515 + }, + { + "epoch": 5.28, + "learning_rate": 7.364819810448387e-07, + "loss": 0.9125, + "step": 71516 + }, + { + "epoch": 5.28, + "learning_rate": 7.363317496500621e-07, + "loss": 1.103, + "step": 71517 + }, + { + "epoch": 5.28, + "learning_rate": 7.361815329936494e-07, + "loss": 1.0296, + "step": 71518 + }, + { + "epoch": 5.28, + "learning_rate": 7.360313310758394e-07, + "loss": 1.022, + "step": 71519 + }, + { + "epoch": 5.28, + "learning_rate": 7.358811438968694e-07, + "loss": 1.0421, + "step": 71520 + }, + { + "epoch": 5.28, + "learning_rate": 7.357309714569816e-07, + "loss": 1.0624, + "step": 71521 + }, + { + "epoch": 5.28, + "learning_rate": 7.355808137564135e-07, + "loss": 0.8597, + "step": 71522 + }, + { + "epoch": 5.28, + "learning_rate": 7.354306707954029e-07, + "loss": 0.9559, + "step": 71523 + }, + { + "epoch": 5.28, + "learning_rate": 7.352805425741904e-07, + "loss": 0.9912, + "step": 71524 + }, + { + "epoch": 5.28, + "learning_rate": 7.351304290930128e-07, + "loss": 1.0267, + "step": 71525 + }, + { + "epoch": 5.29, + "learning_rate": 7.349803303521108e-07, + "loss": 1.0557, + "step": 71526 + }, + { + "epoch": 5.29, + "learning_rate": 7.348302463517232e-07, + "loss": 1.0998, + "step": 71527 + }, + { + "epoch": 5.29, + "learning_rate": 7.346801770920875e-07, + "loss": 0.8822, + "step": 71528 + }, + { + "epoch": 5.29, + "learning_rate": 7.345301225734413e-07, + "loss": 0.9877, + "step": 71529 + }, + { + "epoch": 5.29, + "learning_rate": 7.343800827960279e-07, + "loss": 0.9249, + "step": 71530 + }, + { + "epoch": 5.29, + "learning_rate": 7.342300577600824e-07, + "loss": 0.9472, + "step": 71531 + }, + { + "epoch": 5.29, + "learning_rate": 7.340800474658438e-07, + "loss": 0.9349, + "step": 71532 + }, + { + "epoch": 5.29, + "learning_rate": 7.339300519135517e-07, + "loss": 0.964, + "step": 71533 + }, + { + "epoch": 5.29, + "learning_rate": 7.337800711034438e-07, + "loss": 0.9384, + "step": 71534 + }, + { + "epoch": 5.29, + "learning_rate": 7.336301050357608e-07, + "loss": 1.0558, + "step": 71535 + }, + { + "epoch": 5.29, + "learning_rate": 7.334801537107372e-07, + "loss": 0.9583, + "step": 71536 + }, + { + "epoch": 5.29, + "learning_rate": 7.333302171286161e-07, + "loss": 1.0807, + "step": 71537 + }, + { + "epoch": 5.29, + "learning_rate": 7.331802952896316e-07, + "loss": 0.9876, + "step": 71538 + }, + { + "epoch": 5.29, + "learning_rate": 7.330303881940282e-07, + "loss": 1.0563, + "step": 71539 + }, + { + "epoch": 5.29, + "learning_rate": 7.328804958420377e-07, + "loss": 1.0051, + "step": 71540 + }, + { + "epoch": 5.29, + "learning_rate": 7.327306182339034e-07, + "loss": 1.0507, + "step": 71541 + }, + { + "epoch": 5.29, + "learning_rate": 7.325807553698605e-07, + "loss": 1.0891, + "step": 71542 + }, + { + "epoch": 5.29, + "learning_rate": 7.324309072501523e-07, + "loss": 0.9928, + "step": 71543 + }, + { + "epoch": 5.29, + "learning_rate": 7.32281073875013e-07, + "loss": 1.0473, + "step": 71544 + }, + { + "epoch": 5.29, + "learning_rate": 7.32131255244679e-07, + "loss": 0.983, + "step": 71545 + }, + { + "epoch": 5.29, + "learning_rate": 7.319814513593937e-07, + "loss": 0.8979, + "step": 71546 + }, + { + "epoch": 5.29, + "learning_rate": 7.318316622193933e-07, + "loss": 1.0383, + "step": 71547 + }, + { + "epoch": 5.29, + "learning_rate": 7.316818878249166e-07, + "loss": 0.9619, + "step": 71548 + }, + { + "epoch": 5.29, + "learning_rate": 7.315321281761989e-07, + "loss": 1.0099, + "step": 71549 + }, + { + "epoch": 5.29, + "learning_rate": 7.313823832734835e-07, + "loss": 0.9402, + "step": 71550 + }, + { + "epoch": 5.29, + "learning_rate": 7.312326531170055e-07, + "loss": 1.0333, + "step": 71551 + }, + { + "epoch": 5.29, + "learning_rate": 7.310829377070039e-07, + "loss": 0.989, + "step": 71552 + }, + { + "epoch": 5.29, + "learning_rate": 7.309332370437172e-07, + "loss": 0.9357, + "step": 71553 + }, + { + "epoch": 5.29, + "learning_rate": 7.307835511273808e-07, + "loss": 0.9924, + "step": 71554 + }, + { + "epoch": 5.29, + "learning_rate": 7.306338799582368e-07, + "loss": 0.8684, + "step": 71555 + }, + { + "epoch": 5.29, + "learning_rate": 7.304842235365217e-07, + "loss": 0.9944, + "step": 71556 + }, + { + "epoch": 5.29, + "learning_rate": 7.30334581862474e-07, + "loss": 0.9712, + "step": 71557 + }, + { + "epoch": 5.29, + "learning_rate": 7.301849549363293e-07, + "loss": 0.9563, + "step": 71558 + }, + { + "epoch": 5.29, + "learning_rate": 7.300353427583317e-07, + "loss": 0.9819, + "step": 71559 + }, + { + "epoch": 5.29, + "learning_rate": 7.298857453287111e-07, + "loss": 1.0097, + "step": 71560 + }, + { + "epoch": 5.29, + "learning_rate": 7.297361626477118e-07, + "loss": 1.0918, + "step": 71561 + }, + { + "epoch": 5.29, + "learning_rate": 7.295865947155689e-07, + "loss": 1.0181, + "step": 71562 + }, + { + "epoch": 5.29, + "learning_rate": 7.294370415325214e-07, + "loss": 1.0902, + "step": 71563 + }, + { + "epoch": 5.29, + "learning_rate": 7.292875030988078e-07, + "loss": 1.0616, + "step": 71564 + }, + { + "epoch": 5.29, + "learning_rate": 7.291379794146625e-07, + "loss": 0.9265, + "step": 71565 + }, + { + "epoch": 5.29, + "learning_rate": 7.289884704803285e-07, + "loss": 1.0907, + "step": 71566 + }, + { + "epoch": 5.29, + "learning_rate": 7.288389762960402e-07, + "loss": 0.9053, + "step": 71567 + }, + { + "epoch": 5.29, + "learning_rate": 7.286894968620373e-07, + "loss": 1.0487, + "step": 71568 + }, + { + "epoch": 5.29, + "learning_rate": 7.285400321785541e-07, + "loss": 1.058, + "step": 71569 + }, + { + "epoch": 5.29, + "learning_rate": 7.283905822458337e-07, + "loss": 0.9593, + "step": 71570 + }, + { + "epoch": 5.29, + "learning_rate": 7.282411470641104e-07, + "loss": 1.0442, + "step": 71571 + }, + { + "epoch": 5.29, + "learning_rate": 7.280917266336241e-07, + "loss": 0.917, + "step": 71572 + }, + { + "epoch": 5.29, + "learning_rate": 7.279423209546099e-07, + "loss": 0.9137, + "step": 71573 + }, + { + "epoch": 5.29, + "learning_rate": 7.277929300273057e-07, + "loss": 1.0108, + "step": 71574 + }, + { + "epoch": 5.29, + "learning_rate": 7.276435538519521e-07, + "loss": 0.9226, + "step": 71575 + }, + { + "epoch": 5.29, + "learning_rate": 7.274941924287848e-07, + "loss": 0.8419, + "step": 71576 + }, + { + "epoch": 5.29, + "learning_rate": 7.273448457580412e-07, + "loss": 1.08, + "step": 71577 + }, + { + "epoch": 5.29, + "learning_rate": 7.271955138399578e-07, + "loss": 1.0789, + "step": 71578 + }, + { + "epoch": 5.29, + "learning_rate": 7.270461966747755e-07, + "loss": 0.9381, + "step": 71579 + }, + { + "epoch": 5.29, + "learning_rate": 7.268968942627319e-07, + "loss": 1.0343, + "step": 71580 + }, + { + "epoch": 5.29, + "learning_rate": 7.267476066040591e-07, + "loss": 1.087, + "step": 71581 + }, + { + "epoch": 5.29, + "learning_rate": 7.265983336989991e-07, + "loss": 1.0215, + "step": 71582 + }, + { + "epoch": 5.29, + "learning_rate": 7.264490755477883e-07, + "loss": 1.0101, + "step": 71583 + }, + { + "epoch": 5.29, + "learning_rate": 7.262998321506665e-07, + "loss": 0.9376, + "step": 71584 + }, + { + "epoch": 5.29, + "learning_rate": 7.261506035078658e-07, + "loss": 1.0235, + "step": 71585 + }, + { + "epoch": 5.29, + "learning_rate": 7.260013896196283e-07, + "loss": 0.9475, + "step": 71586 + }, + { + "epoch": 5.29, + "learning_rate": 7.258521904861893e-07, + "loss": 0.9034, + "step": 71587 + }, + { + "epoch": 5.29, + "learning_rate": 7.257030061077897e-07, + "loss": 1.0908, + "step": 71588 + }, + { + "epoch": 5.29, + "learning_rate": 7.255538364846615e-07, + "loss": 0.9805, + "step": 71589 + }, + { + "epoch": 5.29, + "learning_rate": 7.254046816170435e-07, + "loss": 1.0661, + "step": 71590 + }, + { + "epoch": 5.29, + "learning_rate": 7.252555415051754e-07, + "loss": 0.9867, + "step": 71591 + }, + { + "epoch": 5.29, + "learning_rate": 7.251064161492927e-07, + "loss": 1.0268, + "step": 71592 + }, + { + "epoch": 5.29, + "learning_rate": 7.249573055496339e-07, + "loss": 1.0717, + "step": 71593 + }, + { + "epoch": 5.29, + "learning_rate": 7.248082097064324e-07, + "loss": 0.9497, + "step": 71594 + }, + { + "epoch": 5.29, + "learning_rate": 7.246591286199311e-07, + "loss": 0.9405, + "step": 71595 + }, + { + "epoch": 5.29, + "learning_rate": 7.245100622903634e-07, + "loss": 0.9072, + "step": 71596 + }, + { + "epoch": 5.29, + "learning_rate": 7.243610107179689e-07, + "loss": 0.9271, + "step": 71597 + }, + { + "epoch": 5.29, + "learning_rate": 7.242119739029808e-07, + "loss": 0.9471, + "step": 71598 + }, + { + "epoch": 5.29, + "learning_rate": 7.240629518456399e-07, + "loss": 0.9664, + "step": 71599 + }, + { + "epoch": 5.29, + "learning_rate": 7.239139445461829e-07, + "loss": 0.9493, + "step": 71600 + }, + { + "epoch": 5.29, + "learning_rate": 7.237649520048462e-07, + "loss": 0.966, + "step": 71601 + }, + { + "epoch": 5.29, + "learning_rate": 7.236159742218663e-07, + "loss": 1.0882, + "step": 71602 + }, + { + "epoch": 5.29, + "learning_rate": 7.234670111974784e-07, + "loss": 0.9847, + "step": 71603 + }, + { + "epoch": 5.29, + "learning_rate": 7.233180629319259e-07, + "loss": 1.0285, + "step": 71604 + }, + { + "epoch": 5.29, + "learning_rate": 7.231691294254383e-07, + "loss": 1.0791, + "step": 71605 + }, + { + "epoch": 5.29, + "learning_rate": 7.230202106782569e-07, + "loss": 1.0487, + "step": 71606 + }, + { + "epoch": 5.29, + "learning_rate": 7.228713066906168e-07, + "loss": 1.042, + "step": 71607 + }, + { + "epoch": 5.29, + "learning_rate": 7.22722417462759e-07, + "loss": 0.9658, + "step": 71608 + }, + { + "epoch": 5.29, + "learning_rate": 7.225735429949154e-07, + "loss": 0.9832, + "step": 71609 + }, + { + "epoch": 5.29, + "learning_rate": 7.224246832873227e-07, + "loss": 1.0448, + "step": 71610 + }, + { + "epoch": 5.29, + "learning_rate": 7.222758383402206e-07, + "loss": 0.9498, + "step": 71611 + }, + { + "epoch": 5.29, + "learning_rate": 7.221270081538456e-07, + "loss": 1.0915, + "step": 71612 + }, + { + "epoch": 5.29, + "learning_rate": 7.219781927284342e-07, + "loss": 0.9466, + "step": 71613 + }, + { + "epoch": 5.29, + "learning_rate": 7.218293920642206e-07, + "loss": 0.9713, + "step": 71614 + }, + { + "epoch": 5.29, + "learning_rate": 7.216806061614445e-07, + "loss": 0.9717, + "step": 71615 + }, + { + "epoch": 5.29, + "learning_rate": 7.215318350203426e-07, + "loss": 1.0852, + "step": 71616 + }, + { + "epoch": 5.29, + "learning_rate": 7.213830786411513e-07, + "loss": 0.9817, + "step": 71617 + }, + { + "epoch": 5.29, + "learning_rate": 7.212343370241059e-07, + "loss": 1.0619, + "step": 71618 + }, + { + "epoch": 5.29, + "learning_rate": 7.21085610169443e-07, + "loss": 0.945, + "step": 71619 + }, + { + "epoch": 5.29, + "learning_rate": 7.209368980774012e-07, + "loss": 0.9401, + "step": 71620 + }, + { + "epoch": 5.29, + "learning_rate": 7.207882007482159e-07, + "loss": 0.9488, + "step": 71621 + }, + { + "epoch": 5.29, + "learning_rate": 7.206395181821246e-07, + "loss": 0.9722, + "step": 71622 + }, + { + "epoch": 5.29, + "learning_rate": 7.204908503793606e-07, + "loss": 0.9891, + "step": 71623 + }, + { + "epoch": 5.29, + "learning_rate": 7.203421973401659e-07, + "loss": 0.9963, + "step": 71624 + }, + { + "epoch": 5.29, + "learning_rate": 7.201935590647724e-07, + "loss": 0.9148, + "step": 71625 + }, + { + "epoch": 5.29, + "learning_rate": 7.20044935553419e-07, + "loss": 0.9141, + "step": 71626 + }, + { + "epoch": 5.29, + "learning_rate": 7.198963268063408e-07, + "loss": 1.0363, + "step": 71627 + }, + { + "epoch": 5.29, + "learning_rate": 7.197477328237734e-07, + "loss": 1.0447, + "step": 71628 + }, + { + "epoch": 5.29, + "learning_rate": 7.195991536059577e-07, + "loss": 1.0206, + "step": 71629 + }, + { + "epoch": 5.29, + "learning_rate": 7.194505891531245e-07, + "loss": 1.022, + "step": 71630 + }, + { + "epoch": 5.29, + "learning_rate": 7.193020394655137e-07, + "loss": 0.9085, + "step": 71631 + }, + { + "epoch": 5.29, + "learning_rate": 7.191535045433584e-07, + "loss": 0.9333, + "step": 71632 + }, + { + "epoch": 5.29, + "learning_rate": 7.190049843869007e-07, + "loss": 1.0775, + "step": 71633 + }, + { + "epoch": 5.29, + "learning_rate": 7.188564789963704e-07, + "loss": 1.0036, + "step": 71634 + }, + { + "epoch": 5.29, + "learning_rate": 7.187079883720083e-07, + "loss": 0.91, + "step": 71635 + }, + { + "epoch": 5.29, + "learning_rate": 7.185595125140488e-07, + "loss": 0.9917, + "step": 71636 + }, + { + "epoch": 5.29, + "learning_rate": 7.184110514227282e-07, + "loss": 1.0133, + "step": 71637 + }, + { + "epoch": 5.29, + "learning_rate": 7.182626050982833e-07, + "loss": 0.9899, + "step": 71638 + }, + { + "epoch": 5.29, + "learning_rate": 7.181141735409469e-07, + "loss": 0.9819, + "step": 71639 + }, + { + "epoch": 5.29, + "learning_rate": 7.179657567509612e-07, + "loss": 1.0749, + "step": 71640 + }, + { + "epoch": 5.29, + "learning_rate": 7.178173547285583e-07, + "loss": 1.0113, + "step": 71641 + }, + { + "epoch": 5.29, + "learning_rate": 7.176689674739756e-07, + "loss": 1.0526, + "step": 71642 + }, + { + "epoch": 5.29, + "learning_rate": 7.175205949874464e-07, + "loss": 1.0238, + "step": 71643 + }, + { + "epoch": 5.29, + "learning_rate": 7.173722372692116e-07, + "loss": 1.0369, + "step": 71644 + }, + { + "epoch": 5.29, + "learning_rate": 7.172238943195043e-07, + "loss": 0.9449, + "step": 71645 + }, + { + "epoch": 5.29, + "learning_rate": 7.170755661385609e-07, + "loss": 0.9969, + "step": 71646 + }, + { + "epoch": 5.29, + "learning_rate": 7.16927252726618e-07, + "loss": 1.1037, + "step": 71647 + }, + { + "epoch": 5.29, + "learning_rate": 7.167789540839087e-07, + "loss": 0.9956, + "step": 71648 + }, + { + "epoch": 5.29, + "learning_rate": 7.166306702106751e-07, + "loss": 0.9081, + "step": 71649 + }, + { + "epoch": 5.29, + "learning_rate": 7.164824011071458e-07, + "loss": 1.0447, + "step": 71650 + }, + { + "epoch": 5.29, + "learning_rate": 7.163341467735619e-07, + "loss": 0.9366, + "step": 71651 + }, + { + "epoch": 5.29, + "learning_rate": 7.161859072101562e-07, + "loss": 0.9134, + "step": 71652 + }, + { + "epoch": 5.29, + "learning_rate": 7.160376824171688e-07, + "loss": 0.9921, + "step": 71653 + }, + { + "epoch": 5.29, + "learning_rate": 7.158894723948285e-07, + "loss": 0.9723, + "step": 71654 + }, + { + "epoch": 5.29, + "learning_rate": 7.15741277143378e-07, + "loss": 0.9613, + "step": 71655 + }, + { + "epoch": 5.29, + "learning_rate": 7.155930966630498e-07, + "loss": 1.0484, + "step": 71656 + }, + { + "epoch": 5.29, + "learning_rate": 7.154449309540801e-07, + "loss": 1.0128, + "step": 71657 + }, + { + "epoch": 5.29, + "learning_rate": 7.152967800167054e-07, + "loss": 1.0151, + "step": 71658 + }, + { + "epoch": 5.29, + "learning_rate": 7.15148643851159e-07, + "loss": 1.0003, + "step": 71659 + }, + { + "epoch": 5.29, + "learning_rate": 7.150005224576795e-07, + "loss": 0.9624, + "step": 71660 + }, + { + "epoch": 5.29, + "learning_rate": 7.148524158365011e-07, + "loss": 1.0093, + "step": 71661 + }, + { + "epoch": 5.3, + "learning_rate": 7.147043239878593e-07, + "loss": 0.9354, + "step": 71662 + }, + { + "epoch": 5.3, + "learning_rate": 7.145562469119894e-07, + "loss": 0.9623, + "step": 71663 + }, + { + "epoch": 5.3, + "learning_rate": 7.144081846091288e-07, + "loss": 0.9905, + "step": 71664 + }, + { + "epoch": 5.3, + "learning_rate": 7.142601370795122e-07, + "loss": 0.9938, + "step": 71665 + }, + { + "epoch": 5.3, + "learning_rate": 7.141121043233746e-07, + "loss": 1.0177, + "step": 71666 + }, + { + "epoch": 5.3, + "learning_rate": 7.139640863409514e-07, + "loss": 0.9468, + "step": 71667 + }, + { + "epoch": 5.3, + "learning_rate": 7.13816083132477e-07, + "loss": 0.95, + "step": 71668 + }, + { + "epoch": 5.3, + "learning_rate": 7.1366809469819e-07, + "loss": 0.9512, + "step": 71669 + }, + { + "epoch": 5.3, + "learning_rate": 7.135201210383247e-07, + "loss": 1.0439, + "step": 71670 + }, + { + "epoch": 5.3, + "learning_rate": 7.133721621531153e-07, + "loss": 0.9822, + "step": 71671 + }, + { + "epoch": 5.3, + "learning_rate": 7.132242180427973e-07, + "loss": 0.9439, + "step": 71672 + }, + { + "epoch": 5.3, + "learning_rate": 7.130762887076081e-07, + "loss": 0.9957, + "step": 71673 + }, + { + "epoch": 5.3, + "learning_rate": 7.129283741477822e-07, + "loss": 0.9574, + "step": 71674 + }, + { + "epoch": 5.3, + "learning_rate": 7.127804743635524e-07, + "loss": 0.8873, + "step": 71675 + }, + { + "epoch": 5.3, + "learning_rate": 7.126325893551567e-07, + "loss": 0.969, + "step": 71676 + }, + { + "epoch": 5.3, + "learning_rate": 7.12484719122829e-07, + "loss": 1.0464, + "step": 71677 + }, + { + "epoch": 5.3, + "learning_rate": 7.123368636668071e-07, + "loss": 0.9374, + "step": 71678 + }, + { + "epoch": 5.3, + "learning_rate": 7.121890229873219e-07, + "loss": 1.0761, + "step": 71679 + }, + { + "epoch": 5.3, + "learning_rate": 7.12041197084613e-07, + "loss": 0.9594, + "step": 71680 + }, + { + "epoch": 5.3, + "learning_rate": 7.118933859589139e-07, + "loss": 0.9233, + "step": 71681 + }, + { + "epoch": 5.3, + "learning_rate": 7.117455896104585e-07, + "loss": 1.0465, + "step": 71682 + }, + { + "epoch": 5.3, + "learning_rate": 7.115978080394837e-07, + "loss": 0.9252, + "step": 71683 + }, + { + "epoch": 5.3, + "learning_rate": 7.114500412462222e-07, + "loss": 1.0569, + "step": 71684 + }, + { + "epoch": 5.3, + "learning_rate": 7.113022892309118e-07, + "loss": 1.0387, + "step": 71685 + }, + { + "epoch": 5.3, + "learning_rate": 7.111545519937879e-07, + "loss": 1.0776, + "step": 71686 + }, + { + "epoch": 5.3, + "learning_rate": 7.110068295350825e-07, + "loss": 1.0358, + "step": 71687 + }, + { + "epoch": 5.3, + "learning_rate": 7.10859121855032e-07, + "loss": 1.0313, + "step": 71688 + }, + { + "epoch": 5.3, + "learning_rate": 7.10711428953873e-07, + "loss": 0.9249, + "step": 71689 + }, + { + "epoch": 5.3, + "learning_rate": 7.105637508318386e-07, + "loss": 0.9803, + "step": 71690 + }, + { + "epoch": 5.3, + "learning_rate": 7.104160874891653e-07, + "loss": 0.9573, + "step": 71691 + }, + { + "epoch": 5.3, + "learning_rate": 7.102684389260861e-07, + "loss": 1.0201, + "step": 71692 + }, + { + "epoch": 5.3, + "learning_rate": 7.101208051428365e-07, + "loss": 1.0656, + "step": 71693 + }, + { + "epoch": 5.3, + "learning_rate": 7.099731861396541e-07, + "loss": 0.9967, + "step": 71694 + }, + { + "epoch": 5.3, + "learning_rate": 7.098255819167688e-07, + "loss": 0.9008, + "step": 71695 + }, + { + "epoch": 5.3, + "learning_rate": 7.096779924744191e-07, + "loss": 1.0447, + "step": 71696 + }, + { + "epoch": 5.3, + "learning_rate": 7.09530417812837e-07, + "loss": 1.1016, + "step": 71697 + }, + { + "epoch": 5.3, + "learning_rate": 7.093828579322637e-07, + "loss": 0.9144, + "step": 71698 + }, + { + "epoch": 5.3, + "learning_rate": 7.092353128329243e-07, + "loss": 1.1294, + "step": 71699 + }, + { + "epoch": 5.3, + "learning_rate": 7.090877825150611e-07, + "loss": 0.9837, + "step": 71700 + }, + { + "epoch": 5.3, + "learning_rate": 7.089402669789069e-07, + "loss": 0.9261, + "step": 71701 + }, + { + "epoch": 5.3, + "learning_rate": 7.087927662246952e-07, + "loss": 1.0166, + "step": 71702 + }, + { + "epoch": 5.3, + "learning_rate": 7.086452802526612e-07, + "loss": 1.0534, + "step": 71703 + }, + { + "epoch": 5.3, + "learning_rate": 7.084978090630379e-07, + "loss": 1.1115, + "step": 71704 + }, + { + "epoch": 5.3, + "learning_rate": 7.083503526560631e-07, + "loss": 0.974, + "step": 71705 + }, + { + "epoch": 5.3, + "learning_rate": 7.082029110319711e-07, + "loss": 1.0097, + "step": 71706 + }, + { + "epoch": 5.3, + "learning_rate": 7.080554841909937e-07, + "loss": 0.9, + "step": 71707 + }, + { + "epoch": 5.3, + "learning_rate": 7.079080721333664e-07, + "loss": 1.0075, + "step": 71708 + }, + { + "epoch": 5.3, + "learning_rate": 7.077606748593257e-07, + "loss": 1.0026, + "step": 71709 + }, + { + "epoch": 5.3, + "learning_rate": 7.076132923691048e-07, + "loss": 1.0258, + "step": 71710 + }, + { + "epoch": 5.3, + "learning_rate": 7.07465924662939e-07, + "loss": 1.0049, + "step": 71711 + }, + { + "epoch": 5.3, + "learning_rate": 7.073185717410613e-07, + "loss": 1.0599, + "step": 71712 + }, + { + "epoch": 5.3, + "learning_rate": 7.071712336037051e-07, + "loss": 0.9445, + "step": 71713 + }, + { + "epoch": 5.3, + "learning_rate": 7.070239102511079e-07, + "loss": 1.0568, + "step": 71714 + }, + { + "epoch": 5.3, + "learning_rate": 7.068766016835038e-07, + "loss": 1.0068, + "step": 71715 + }, + { + "epoch": 5.3, + "learning_rate": 7.067293079011251e-07, + "loss": 0.934, + "step": 71716 + }, + { + "epoch": 5.3, + "learning_rate": 7.065820289042058e-07, + "loss": 1.0149, + "step": 71717 + }, + { + "epoch": 5.3, + "learning_rate": 7.064347646929826e-07, + "loss": 0.9014, + "step": 71718 + }, + { + "epoch": 5.3, + "learning_rate": 7.062875152676896e-07, + "loss": 0.8754, + "step": 71719 + }, + { + "epoch": 5.3, + "learning_rate": 7.0614028062856e-07, + "loss": 1.0452, + "step": 71720 + }, + { + "epoch": 5.3, + "learning_rate": 7.059930607758281e-07, + "loss": 0.9554, + "step": 71721 + }, + { + "epoch": 5.3, + "learning_rate": 7.058458557097258e-07, + "loss": 1.0846, + "step": 71722 + }, + { + "epoch": 5.3, + "learning_rate": 7.056986654304942e-07, + "loss": 1.0318, + "step": 71723 + }, + { + "epoch": 5.3, + "learning_rate": 7.055514899383598e-07, + "loss": 1.0721, + "step": 71724 + }, + { + "epoch": 5.3, + "learning_rate": 7.054043292335611e-07, + "loss": 0.9541, + "step": 71725 + }, + { + "epoch": 5.3, + "learning_rate": 7.052571833163302e-07, + "loss": 1.0453, + "step": 71726 + }, + { + "epoch": 5.3, + "learning_rate": 7.051100521869036e-07, + "loss": 1.045, + "step": 71727 + }, + { + "epoch": 5.3, + "learning_rate": 7.049629358455124e-07, + "loss": 0.8781, + "step": 71728 + }, + { + "epoch": 5.3, + "learning_rate": 7.048158342923928e-07, + "loss": 1.0057, + "step": 71729 + }, + { + "epoch": 5.3, + "learning_rate": 7.046687475277792e-07, + "loss": 1.1221, + "step": 71730 + }, + { + "epoch": 5.3, + "learning_rate": 7.045216755519035e-07, + "loss": 1.0114, + "step": 71731 + }, + { + "epoch": 5.3, + "learning_rate": 7.043746183650013e-07, + "loss": 0.9884, + "step": 71732 + }, + { + "epoch": 5.3, + "learning_rate": 7.042275759673045e-07, + "loss": 0.9597, + "step": 71733 + }, + { + "epoch": 5.3, + "learning_rate": 7.040805483590507e-07, + "loss": 1.0268, + "step": 71734 + }, + { + "epoch": 5.3, + "learning_rate": 7.039335355404708e-07, + "loss": 0.9361, + "step": 71735 + }, + { + "epoch": 5.3, + "learning_rate": 7.037865375118003e-07, + "loss": 1.0058, + "step": 71736 + }, + { + "epoch": 5.3, + "learning_rate": 7.03639554273271e-07, + "loss": 0.9996, + "step": 71737 + }, + { + "epoch": 5.3, + "learning_rate": 7.034925858251196e-07, + "loss": 0.9067, + "step": 71738 + }, + { + "epoch": 5.3, + "learning_rate": 7.033456321675802e-07, + "loss": 0.9753, + "step": 71739 + }, + { + "epoch": 5.3, + "learning_rate": 7.031986933008805e-07, + "loss": 1.0184, + "step": 71740 + }, + { + "epoch": 5.3, + "learning_rate": 7.030517692252614e-07, + "loss": 0.9433, + "step": 71741 + }, + { + "epoch": 5.3, + "learning_rate": 7.029048599409527e-07, + "loss": 1.0116, + "step": 71742 + }, + { + "epoch": 5.3, + "learning_rate": 7.02757965448192e-07, + "loss": 0.9837, + "step": 71743 + }, + { + "epoch": 5.3, + "learning_rate": 7.026110857472069e-07, + "loss": 0.9466, + "step": 71744 + }, + { + "epoch": 5.3, + "learning_rate": 7.024642208382359e-07, + "loss": 0.8965, + "step": 71745 + }, + { + "epoch": 5.3, + "learning_rate": 7.023173707215103e-07, + "loss": 1.0223, + "step": 71746 + }, + { + "epoch": 5.3, + "learning_rate": 7.021705353972685e-07, + "loss": 0.9832, + "step": 71747 + }, + { + "epoch": 5.3, + "learning_rate": 7.020237148657372e-07, + "loss": 1.0475, + "step": 71748 + }, + { + "epoch": 5.3, + "learning_rate": 7.018769091271527e-07, + "loss": 0.9447, + "step": 71749 + }, + { + "epoch": 5.3, + "learning_rate": 7.017301181817504e-07, + "loss": 1.0453, + "step": 71750 + }, + { + "epoch": 5.3, + "learning_rate": 7.015833420297624e-07, + "loss": 0.9494, + "step": 71751 + }, + { + "epoch": 5.3, + "learning_rate": 7.014365806714218e-07, + "loss": 0.9868, + "step": 71752 + }, + { + "epoch": 5.3, + "learning_rate": 7.012898341069619e-07, + "loss": 0.9582, + "step": 71753 + }, + { + "epoch": 5.3, + "learning_rate": 7.011431023366178e-07, + "loss": 0.9869, + "step": 71754 + }, + { + "epoch": 5.3, + "learning_rate": 7.009963853606217e-07, + "loss": 1.0087, + "step": 71755 + }, + { + "epoch": 5.3, + "learning_rate": 7.008496831792078e-07, + "loss": 0.9686, + "step": 71756 + }, + { + "epoch": 5.3, + "learning_rate": 7.00702995792607e-07, + "loss": 0.8262, + "step": 71757 + }, + { + "epoch": 5.3, + "learning_rate": 7.00556323201057e-07, + "loss": 1.1817, + "step": 71758 + }, + { + "epoch": 5.3, + "learning_rate": 7.004096654047887e-07, + "loss": 0.9886, + "step": 71759 + }, + { + "epoch": 5.3, + "learning_rate": 7.002630224040352e-07, + "loss": 0.9857, + "step": 71760 + }, + { + "epoch": 5.3, + "learning_rate": 7.001163941990297e-07, + "loss": 1.0921, + "step": 71761 + }, + { + "epoch": 5.3, + "learning_rate": 6.999697807900052e-07, + "loss": 0.9891, + "step": 71762 + }, + { + "epoch": 5.3, + "learning_rate": 6.998231821771973e-07, + "loss": 0.8945, + "step": 71763 + }, + { + "epoch": 5.3, + "learning_rate": 6.996765983608366e-07, + "loss": 1.0088, + "step": 71764 + }, + { + "epoch": 5.3, + "learning_rate": 6.995300293411588e-07, + "loss": 0.9454, + "step": 71765 + }, + { + "epoch": 5.3, + "learning_rate": 6.993834751183936e-07, + "loss": 1.0949, + "step": 71766 + }, + { + "epoch": 5.3, + "learning_rate": 6.992369356927775e-07, + "loss": 1.1636, + "step": 71767 + }, + { + "epoch": 5.3, + "learning_rate": 6.990904110645435e-07, + "loss": 1.0484, + "step": 71768 + }, + { + "epoch": 5.3, + "learning_rate": 6.989439012339216e-07, + "loss": 1.0026, + "step": 71769 + }, + { + "epoch": 5.3, + "learning_rate": 6.987974062011471e-07, + "loss": 0.9866, + "step": 71770 + }, + { + "epoch": 5.3, + "learning_rate": 6.986509259664543e-07, + "loss": 0.9131, + "step": 71771 + }, + { + "epoch": 5.3, + "learning_rate": 6.985044605300739e-07, + "loss": 1.0702, + "step": 71772 + }, + { + "epoch": 5.3, + "learning_rate": 6.983580098922382e-07, + "loss": 0.9094, + "step": 71773 + }, + { + "epoch": 5.3, + "learning_rate": 6.982115740531836e-07, + "loss": 1.0373, + "step": 71774 + }, + { + "epoch": 5.3, + "learning_rate": 6.98065153013141e-07, + "loss": 0.9017, + "step": 71775 + }, + { + "epoch": 5.3, + "learning_rate": 6.979187467723447e-07, + "loss": 0.9698, + "step": 71776 + }, + { + "epoch": 5.3, + "learning_rate": 6.977723553310256e-07, + "loss": 1.0581, + "step": 71777 + }, + { + "epoch": 5.3, + "learning_rate": 6.976259786894157e-07, + "loss": 1.0453, + "step": 71778 + }, + { + "epoch": 5.3, + "learning_rate": 6.974796168477527e-07, + "loss": 1.0425, + "step": 71779 + }, + { + "epoch": 5.3, + "learning_rate": 6.973332698062651e-07, + "loss": 0.9969, + "step": 71780 + }, + { + "epoch": 5.3, + "learning_rate": 6.971869375651885e-07, + "loss": 1.0947, + "step": 71781 + }, + { + "epoch": 5.3, + "learning_rate": 6.970406201247515e-07, + "loss": 0.9801, + "step": 71782 + }, + { + "epoch": 5.3, + "learning_rate": 6.968943174851917e-07, + "loss": 1.1229, + "step": 71783 + }, + { + "epoch": 5.3, + "learning_rate": 6.967480296467411e-07, + "loss": 0.9368, + "step": 71784 + }, + { + "epoch": 5.3, + "learning_rate": 6.966017566096295e-07, + "loss": 0.9299, + "step": 71785 + }, + { + "epoch": 5.3, + "learning_rate": 6.964554983740924e-07, + "loss": 0.9778, + "step": 71786 + }, + { + "epoch": 5.3, + "learning_rate": 6.963092549403605e-07, + "loss": 1.0043, + "step": 71787 + }, + { + "epoch": 5.3, + "learning_rate": 6.961630263086694e-07, + "loss": 0.9861, + "step": 71788 + }, + { + "epoch": 5.3, + "learning_rate": 6.960168124792477e-07, + "loss": 1.0907, + "step": 71789 + }, + { + "epoch": 5.3, + "learning_rate": 6.958706134523307e-07, + "loss": 1.0002, + "step": 71790 + }, + { + "epoch": 5.3, + "learning_rate": 6.957244292281496e-07, + "loss": 0.9661, + "step": 71791 + }, + { + "epoch": 5.3, + "learning_rate": 6.955782598069405e-07, + "loss": 1.0119, + "step": 71792 + }, + { + "epoch": 5.3, + "learning_rate": 6.954321051889301e-07, + "loss": 0.9321, + "step": 71793 + }, + { + "epoch": 5.3, + "learning_rate": 6.95285965374356e-07, + "loss": 0.9737, + "step": 71794 + }, + { + "epoch": 5.3, + "learning_rate": 6.95139840363449e-07, + "loss": 1.0164, + "step": 71795 + }, + { + "epoch": 5.3, + "learning_rate": 6.949937301564402e-07, + "loss": 0.9447, + "step": 71796 + }, + { + "epoch": 5.31, + "learning_rate": 6.948476347535649e-07, + "loss": 1.0306, + "step": 71797 + }, + { + "epoch": 5.31, + "learning_rate": 6.947015541550506e-07, + "loss": 1.032, + "step": 71798 + }, + { + "epoch": 5.31, + "learning_rate": 6.945554883611361e-07, + "loss": 0.9233, + "step": 71799 + }, + { + "epoch": 5.31, + "learning_rate": 6.9440943737205e-07, + "loss": 1.0229, + "step": 71800 + }, + { + "epoch": 5.31, + "learning_rate": 6.942634011880256e-07, + "loss": 0.9527, + "step": 71801 + }, + { + "epoch": 5.31, + "learning_rate": 6.941173798092926e-07, + "loss": 0.9209, + "step": 71802 + }, + { + "epoch": 5.31, + "learning_rate": 6.939713732360887e-07, + "loss": 1.1284, + "step": 71803 + }, + { + "epoch": 5.31, + "learning_rate": 6.938253814686413e-07, + "loss": 1.0163, + "step": 71804 + }, + { + "epoch": 5.31, + "learning_rate": 6.936794045071859e-07, + "loss": 0.8892, + "step": 71805 + }, + { + "epoch": 5.31, + "learning_rate": 6.935334423519535e-07, + "loss": 1.0555, + "step": 71806 + }, + { + "epoch": 5.31, + "learning_rate": 6.933874950031738e-07, + "loss": 1.0108, + "step": 71807 + }, + { + "epoch": 5.31, + "learning_rate": 6.932415624610844e-07, + "loss": 0.9386, + "step": 71808 + }, + { + "epoch": 5.31, + "learning_rate": 6.930956447259129e-07, + "loss": 0.9607, + "step": 71809 + }, + { + "epoch": 5.31, + "learning_rate": 6.929497417978937e-07, + "loss": 0.9774, + "step": 71810 + }, + { + "epoch": 5.31, + "learning_rate": 6.928038536772574e-07, + "loss": 0.9903, + "step": 71811 + }, + { + "epoch": 5.31, + "learning_rate": 6.926579803642375e-07, + "loss": 0.9909, + "step": 71812 + }, + { + "epoch": 5.31, + "learning_rate": 6.925121218590669e-07, + "loss": 1.1504, + "step": 71813 + }, + { + "epoch": 5.31, + "learning_rate": 6.923662781619756e-07, + "loss": 1.0878, + "step": 71814 + }, + { + "epoch": 5.31, + "learning_rate": 6.922204492731954e-07, + "loss": 1.0067, + "step": 71815 + }, + { + "epoch": 5.31, + "learning_rate": 6.920746351929608e-07, + "loss": 0.8453, + "step": 71816 + }, + { + "epoch": 5.31, + "learning_rate": 6.919288359215015e-07, + "loss": 0.9266, + "step": 71817 + }, + { + "epoch": 5.31, + "learning_rate": 6.917830514590485e-07, + "loss": 1.0062, + "step": 71818 + }, + { + "epoch": 5.31, + "learning_rate": 6.916372818058381e-07, + "loss": 1.0423, + "step": 71819 + }, + { + "epoch": 5.31, + "learning_rate": 6.914915269620992e-07, + "loss": 0.996, + "step": 71820 + }, + { + "epoch": 5.31, + "learning_rate": 6.913457869280637e-07, + "loss": 0.9509, + "step": 71821 + }, + { + "epoch": 5.31, + "learning_rate": 6.912000617039616e-07, + "loss": 1.0253, + "step": 71822 + }, + { + "epoch": 5.31, + "learning_rate": 6.910543512900303e-07, + "loss": 0.8892, + "step": 71823 + }, + { + "epoch": 5.31, + "learning_rate": 6.909086556864974e-07, + "loss": 1.081, + "step": 71824 + }, + { + "epoch": 5.31, + "learning_rate": 6.907629748935952e-07, + "loss": 0.9323, + "step": 71825 + }, + { + "epoch": 5.31, + "learning_rate": 6.906173089115564e-07, + "loss": 0.9902, + "step": 71826 + }, + { + "epoch": 5.31, + "learning_rate": 6.904716577406112e-07, + "loss": 1.1043, + "step": 71827 + }, + { + "epoch": 5.31, + "learning_rate": 6.903260213809937e-07, + "loss": 0.9196, + "step": 71828 + }, + { + "epoch": 5.31, + "learning_rate": 6.901803998329337e-07, + "loss": 0.9829, + "step": 71829 + }, + { + "epoch": 5.31, + "learning_rate": 6.900347930966634e-07, + "loss": 0.9856, + "step": 71830 + }, + { + "epoch": 5.31, + "learning_rate": 6.898892011724134e-07, + "loss": 0.9819, + "step": 71831 + }, + { + "epoch": 5.31, + "learning_rate": 6.897436240604183e-07, + "loss": 0.9926, + "step": 71832 + }, + { + "epoch": 5.31, + "learning_rate": 6.895980617609099e-07, + "loss": 0.8662, + "step": 71833 + }, + { + "epoch": 5.31, + "learning_rate": 6.894525142741138e-07, + "loss": 0.8823, + "step": 71834 + }, + { + "epoch": 5.31, + "learning_rate": 6.893069816002673e-07, + "loss": 0.9605, + "step": 71835 + }, + { + "epoch": 5.31, + "learning_rate": 6.891614637395982e-07, + "loss": 0.9686, + "step": 71836 + }, + { + "epoch": 5.31, + "learning_rate": 6.890159606923442e-07, + "loss": 0.9834, + "step": 71837 + }, + { + "epoch": 5.31, + "learning_rate": 6.888704724587281e-07, + "loss": 0.944, + "step": 71838 + }, + { + "epoch": 5.31, + "learning_rate": 6.887249990389877e-07, + "loss": 1.0608, + "step": 71839 + }, + { + "epoch": 5.31, + "learning_rate": 6.885795404333518e-07, + "loss": 1.0662, + "step": 71840 + }, + { + "epoch": 5.31, + "learning_rate": 6.884340966420555e-07, + "loss": 1.0871, + "step": 71841 + }, + { + "epoch": 5.31, + "learning_rate": 6.882886676653255e-07, + "loss": 0.8574, + "step": 71842 + }, + { + "epoch": 5.31, + "learning_rate": 6.881432535033938e-07, + "loss": 1.0595, + "step": 71843 + }, + { + "epoch": 5.31, + "learning_rate": 6.879978541564947e-07, + "loss": 0.9426, + "step": 71844 + }, + { + "epoch": 5.31, + "learning_rate": 6.878524696248567e-07, + "loss": 1.0552, + "step": 71845 + }, + { + "epoch": 5.31, + "learning_rate": 6.877070999087132e-07, + "loss": 1.0202, + "step": 71846 + }, + { + "epoch": 5.31, + "learning_rate": 6.875617450082928e-07, + "loss": 1.0292, + "step": 71847 + }, + { + "epoch": 5.31, + "learning_rate": 6.874164049238297e-07, + "loss": 0.9847, + "step": 71848 + }, + { + "epoch": 5.31, + "learning_rate": 6.872710796555537e-07, + "loss": 0.9922, + "step": 71849 + }, + { + "epoch": 5.31, + "learning_rate": 6.871257692036959e-07, + "loss": 1.0488, + "step": 71850 + }, + { + "epoch": 5.31, + "learning_rate": 6.869804735684882e-07, + "loss": 0.9997, + "step": 71851 + }, + { + "epoch": 5.31, + "learning_rate": 6.868351927501604e-07, + "loss": 1.0493, + "step": 71852 + }, + { + "epoch": 5.31, + "learning_rate": 6.866899267489457e-07, + "loss": 0.9446, + "step": 71853 + }, + { + "epoch": 5.31, + "learning_rate": 6.86544675565074e-07, + "loss": 1.014, + "step": 71854 + }, + { + "epoch": 5.31, + "learning_rate": 6.863994391987761e-07, + "loss": 0.99, + "step": 71855 + }, + { + "epoch": 5.31, + "learning_rate": 6.862542176502817e-07, + "loss": 1.0893, + "step": 71856 + }, + { + "epoch": 5.31, + "learning_rate": 6.861090109198276e-07, + "loss": 1.0354, + "step": 71857 + }, + { + "epoch": 5.31, + "learning_rate": 6.859638190076368e-07, + "loss": 1.0752, + "step": 71858 + }, + { + "epoch": 5.31, + "learning_rate": 6.858186419139468e-07, + "loss": 1.1408, + "step": 71859 + }, + { + "epoch": 5.31, + "learning_rate": 6.856734796389852e-07, + "loss": 1.0235, + "step": 71860 + }, + { + "epoch": 5.31, + "learning_rate": 6.855283321829842e-07, + "loss": 0.9443, + "step": 71861 + }, + { + "epoch": 5.31, + "learning_rate": 6.853831995461747e-07, + "loss": 1.0626, + "step": 71862 + }, + { + "epoch": 5.31, + "learning_rate": 6.852380817287851e-07, + "loss": 1.0117, + "step": 71863 + }, + { + "epoch": 5.31, + "learning_rate": 6.850929787310501e-07, + "loss": 0.894, + "step": 71864 + }, + { + "epoch": 5.31, + "learning_rate": 6.849478905531992e-07, + "loss": 1.0632, + "step": 71865 + }, + { + "epoch": 5.31, + "learning_rate": 6.848028171954624e-07, + "loss": 1.0589, + "step": 71866 + }, + { + "epoch": 5.31, + "learning_rate": 6.846577586580704e-07, + "loss": 0.9391, + "step": 71867 + }, + { + "epoch": 5.31, + "learning_rate": 6.845127149412555e-07, + "loss": 0.9529, + "step": 71868 + }, + { + "epoch": 5.31, + "learning_rate": 6.843676860452475e-07, + "loss": 0.9362, + "step": 71869 + }, + { + "epoch": 5.31, + "learning_rate": 6.842226719702771e-07, + "loss": 0.9435, + "step": 71870 + }, + { + "epoch": 5.31, + "learning_rate": 6.840776727165754e-07, + "loss": 1.1108, + "step": 71871 + }, + { + "epoch": 5.31, + "learning_rate": 6.839326882843722e-07, + "loss": 1.0484, + "step": 71872 + }, + { + "epoch": 5.31, + "learning_rate": 6.837877186738995e-07, + "loss": 0.9803, + "step": 71873 + }, + { + "epoch": 5.31, + "learning_rate": 6.83642763885387e-07, + "loss": 1.0141, + "step": 71874 + }, + { + "epoch": 5.31, + "learning_rate": 6.834978239190659e-07, + "loss": 1.0236, + "step": 71875 + }, + { + "epoch": 5.31, + "learning_rate": 6.833528987751648e-07, + "loss": 1.0385, + "step": 71876 + }, + { + "epoch": 5.31, + "learning_rate": 6.832079884539178e-07, + "loss": 1.0844, + "step": 71877 + }, + { + "epoch": 5.31, + "learning_rate": 6.830630929555526e-07, + "loss": 1.0031, + "step": 71878 + }, + { + "epoch": 5.31, + "learning_rate": 6.829182122803024e-07, + "loss": 0.9802, + "step": 71879 + }, + { + "epoch": 5.31, + "learning_rate": 6.827733464283948e-07, + "loss": 1.0023, + "step": 71880 + }, + { + "epoch": 5.31, + "learning_rate": 6.826284954000594e-07, + "loss": 1.0848, + "step": 71881 + }, + { + "epoch": 5.31, + "learning_rate": 6.82483659195533e-07, + "loss": 1.0201, + "step": 71882 + }, + { + "epoch": 5.31, + "learning_rate": 6.823388378150386e-07, + "loss": 1.0119, + "step": 71883 + }, + { + "epoch": 5.31, + "learning_rate": 6.821940312588116e-07, + "loss": 0.9318, + "step": 71884 + }, + { + "epoch": 5.31, + "learning_rate": 6.820492395270783e-07, + "loss": 0.9655, + "step": 71885 + }, + { + "epoch": 5.31, + "learning_rate": 6.819044626200744e-07, + "loss": 1.0086, + "step": 71886 + }, + { + "epoch": 5.31, + "learning_rate": 6.817597005380239e-07, + "loss": 0.883, + "step": 71887 + }, + { + "epoch": 5.31, + "learning_rate": 6.816149532811634e-07, + "loss": 0.9906, + "step": 71888 + }, + { + "epoch": 5.31, + "learning_rate": 6.814702208497182e-07, + "loss": 1.0254, + "step": 71889 + }, + { + "epoch": 5.31, + "learning_rate": 6.813255032439215e-07, + "loss": 0.9266, + "step": 71890 + }, + { + "epoch": 5.31, + "learning_rate": 6.81180800464003e-07, + "loss": 1.0502, + "step": 71891 + }, + { + "epoch": 5.31, + "learning_rate": 6.810361125101905e-07, + "loss": 0.9841, + "step": 71892 + }, + { + "epoch": 5.31, + "learning_rate": 6.808914393827182e-07, + "loss": 1.0325, + "step": 71893 + }, + { + "epoch": 5.31, + "learning_rate": 6.807467810818136e-07, + "loss": 1.0717, + "step": 71894 + }, + { + "epoch": 5.31, + "learning_rate": 6.806021376077076e-07, + "loss": 1.1028, + "step": 71895 + }, + { + "epoch": 5.31, + "learning_rate": 6.804575089606291e-07, + "loss": 0.907, + "step": 71896 + }, + { + "epoch": 5.31, + "learning_rate": 6.803128951408111e-07, + "loss": 1.0872, + "step": 71897 + }, + { + "epoch": 5.31, + "learning_rate": 6.801682961484812e-07, + "loss": 0.9925, + "step": 71898 + }, + { + "epoch": 5.31, + "learning_rate": 6.800237119838715e-07, + "loss": 0.997, + "step": 71899 + }, + { + "epoch": 5.31, + "learning_rate": 6.798791426472096e-07, + "loss": 1.0109, + "step": 71900 + }, + { + "epoch": 5.31, + "learning_rate": 6.797345881387252e-07, + "loss": 1.1218, + "step": 71901 + }, + { + "epoch": 5.31, + "learning_rate": 6.795900484586526e-07, + "loss": 1.0142, + "step": 71902 + }, + { + "epoch": 5.31, + "learning_rate": 6.794455236072162e-07, + "loss": 0.8937, + "step": 71903 + }, + { + "epoch": 5.31, + "learning_rate": 6.793010135846501e-07, + "loss": 0.9304, + "step": 71904 + }, + { + "epoch": 5.31, + "learning_rate": 6.791565183911819e-07, + "loss": 0.9051, + "step": 71905 + }, + { + "epoch": 5.31, + "learning_rate": 6.790120380270448e-07, + "loss": 0.9271, + "step": 71906 + }, + { + "epoch": 5.31, + "learning_rate": 6.788675724924643e-07, + "loss": 0.9413, + "step": 71907 + }, + { + "epoch": 5.31, + "learning_rate": 6.78723121787671e-07, + "loss": 0.9932, + "step": 71908 + }, + { + "epoch": 5.31, + "learning_rate": 6.785786859128973e-07, + "loss": 1.0475, + "step": 71909 + }, + { + "epoch": 5.31, + "learning_rate": 6.784342648683706e-07, + "loss": 0.9305, + "step": 71910 + }, + { + "epoch": 5.31, + "learning_rate": 6.78289858654323e-07, + "loss": 1.1154, + "step": 71911 + }, + { + "epoch": 5.31, + "learning_rate": 6.781454672709809e-07, + "loss": 1.022, + "step": 71912 + }, + { + "epoch": 5.31, + "learning_rate": 6.780010907185775e-07, + "loss": 1.0304, + "step": 71913 + }, + { + "epoch": 5.31, + "learning_rate": 6.778567289973404e-07, + "loss": 1.0561, + "step": 71914 + }, + { + "epoch": 5.31, + "learning_rate": 6.777123821075005e-07, + "loss": 0.8713, + "step": 71915 + }, + { + "epoch": 5.31, + "learning_rate": 6.775680500492865e-07, + "loss": 1.0557, + "step": 71916 + }, + { + "epoch": 5.31, + "learning_rate": 6.774237328229272e-07, + "loss": 0.9635, + "step": 71917 + }, + { + "epoch": 5.31, + "learning_rate": 6.772794304286546e-07, + "loss": 0.9323, + "step": 71918 + }, + { + "epoch": 5.31, + "learning_rate": 6.771351428666973e-07, + "loss": 0.971, + "step": 71919 + }, + { + "epoch": 5.31, + "learning_rate": 6.769908701372851e-07, + "loss": 1.0848, + "step": 71920 + }, + { + "epoch": 5.31, + "learning_rate": 6.768466122406447e-07, + "loss": 1.0369, + "step": 71921 + }, + { + "epoch": 5.31, + "learning_rate": 6.767023691770092e-07, + "loss": 0.9118, + "step": 71922 + }, + { + "epoch": 5.31, + "learning_rate": 6.765581409466071e-07, + "loss": 0.9508, + "step": 71923 + }, + { + "epoch": 5.31, + "learning_rate": 6.764139275496684e-07, + "loss": 0.9895, + "step": 71924 + }, + { + "epoch": 5.31, + "learning_rate": 6.762697289864195e-07, + "loss": 0.963, + "step": 71925 + }, + { + "epoch": 5.31, + "learning_rate": 6.761255452570937e-07, + "loss": 1.0221, + "step": 71926 + }, + { + "epoch": 5.31, + "learning_rate": 6.759813763619205e-07, + "loss": 0.9562, + "step": 71927 + }, + { + "epoch": 5.31, + "learning_rate": 6.758372223011234e-07, + "loss": 0.9686, + "step": 71928 + }, + { + "epoch": 5.31, + "learning_rate": 6.756930830749386e-07, + "loss": 0.9357, + "step": 71929 + }, + { + "epoch": 5.31, + "learning_rate": 6.755489586835906e-07, + "loss": 0.9881, + "step": 71930 + }, + { + "epoch": 5.31, + "learning_rate": 6.754048491273146e-07, + "loss": 1.0061, + "step": 71931 + }, + { + "epoch": 5.32, + "learning_rate": 6.752607544063316e-07, + "loss": 1.0683, + "step": 71932 + }, + { + "epoch": 5.32, + "learning_rate": 6.751166745208782e-07, + "loss": 0.9117, + "step": 71933 + }, + { + "epoch": 5.32, + "learning_rate": 6.749726094711806e-07, + "loss": 0.9625, + "step": 71934 + }, + { + "epoch": 5.32, + "learning_rate": 6.748285592574677e-07, + "loss": 1.001, + "step": 71935 + }, + { + "epoch": 5.32, + "learning_rate": 6.746845238799693e-07, + "loss": 0.9777, + "step": 71936 + }, + { + "epoch": 5.32, + "learning_rate": 6.745405033389129e-07, + "loss": 1.0118, + "step": 71937 + }, + { + "epoch": 5.32, + "learning_rate": 6.743964976345308e-07, + "loss": 1.0809, + "step": 71938 + }, + { + "epoch": 5.32, + "learning_rate": 6.742525067670502e-07, + "loss": 1.117, + "step": 71939 + }, + { + "epoch": 5.32, + "learning_rate": 6.741085307367001e-07, + "loss": 0.9661, + "step": 71940 + }, + { + "epoch": 5.32, + "learning_rate": 6.739645695437091e-07, + "loss": 1.0341, + "step": 71941 + }, + { + "epoch": 5.32, + "learning_rate": 6.738206231883082e-07, + "loss": 0.9177, + "step": 71942 + }, + { + "epoch": 5.32, + "learning_rate": 6.736766916707249e-07, + "loss": 0.975, + "step": 71943 + }, + { + "epoch": 5.32, + "learning_rate": 6.73532774991189e-07, + "loss": 1.0473, + "step": 71944 + }, + { + "epoch": 5.32, + "learning_rate": 6.733888731499294e-07, + "loss": 1.0284, + "step": 71945 + }, + { + "epoch": 5.32, + "learning_rate": 6.732449861471724e-07, + "loss": 1.0899, + "step": 71946 + }, + { + "epoch": 5.32, + "learning_rate": 6.731011139831523e-07, + "loss": 0.9475, + "step": 71947 + }, + { + "epoch": 5.32, + "learning_rate": 6.729572566580922e-07, + "loss": 1.0247, + "step": 71948 + }, + { + "epoch": 5.32, + "learning_rate": 6.728134141722253e-07, + "loss": 1.0281, + "step": 71949 + }, + { + "epoch": 5.32, + "learning_rate": 6.726695865257759e-07, + "loss": 1.0022, + "step": 71950 + }, + { + "epoch": 5.32, + "learning_rate": 6.725257737189805e-07, + "loss": 0.9535, + "step": 71951 + }, + { + "epoch": 5.32, + "learning_rate": 6.7238197575206e-07, + "loss": 1.0449, + "step": 71952 + }, + { + "epoch": 5.32, + "learning_rate": 6.722381926252464e-07, + "loss": 0.9709, + "step": 71953 + }, + { + "epoch": 5.32, + "learning_rate": 6.720944243387694e-07, + "loss": 1.0545, + "step": 71954 + }, + { + "epoch": 5.32, + "learning_rate": 6.719506708928569e-07, + "loss": 0.9611, + "step": 71955 + }, + { + "epoch": 5.32, + "learning_rate": 6.718069322877374e-07, + "loss": 0.8842, + "step": 71956 + }, + { + "epoch": 5.32, + "learning_rate": 6.716632085236375e-07, + "loss": 0.8743, + "step": 71957 + }, + { + "epoch": 5.32, + "learning_rate": 6.715194996007901e-07, + "loss": 0.9538, + "step": 71958 + }, + { + "epoch": 5.32, + "learning_rate": 6.713758055194209e-07, + "loss": 0.9821, + "step": 71959 + }, + { + "epoch": 5.32, + "learning_rate": 6.712321262797606e-07, + "loss": 1.0609, + "step": 71960 + }, + { + "epoch": 5.32, + "learning_rate": 6.710884618820335e-07, + "loss": 0.8894, + "step": 71961 + }, + { + "epoch": 5.32, + "learning_rate": 6.70944812326474e-07, + "loss": 0.9798, + "step": 71962 + }, + { + "epoch": 5.32, + "learning_rate": 6.708011776133073e-07, + "loss": 0.8595, + "step": 71963 + }, + { + "epoch": 5.32, + "learning_rate": 6.706575577427621e-07, + "loss": 0.8998, + "step": 71964 + }, + { + "epoch": 5.32, + "learning_rate": 6.705139527150672e-07, + "loss": 1.0411, + "step": 71965 + }, + { + "epoch": 5.32, + "learning_rate": 6.703703625304503e-07, + "loss": 0.9267, + "step": 71966 + }, + { + "epoch": 5.32, + "learning_rate": 6.702267871891422e-07, + "loss": 1.1198, + "step": 71967 + }, + { + "epoch": 5.32, + "learning_rate": 6.700832266913692e-07, + "loss": 1.073, + "step": 71968 + }, + { + "epoch": 5.32, + "learning_rate": 6.699396810373604e-07, + "loss": 1.0224, + "step": 71969 + }, + { + "epoch": 5.32, + "learning_rate": 6.69796150227342e-07, + "loss": 1.0094, + "step": 71970 + }, + { + "epoch": 5.32, + "learning_rate": 6.696526342615473e-07, + "loss": 1.0193, + "step": 71971 + }, + { + "epoch": 5.32, + "learning_rate": 6.695091331402004e-07, + "loss": 0.8902, + "step": 71972 + }, + { + "epoch": 5.32, + "learning_rate": 6.693656468635323e-07, + "loss": 0.9454, + "step": 71973 + }, + { + "epoch": 5.32, + "learning_rate": 6.692221754317696e-07, + "loss": 0.9311, + "step": 71974 + }, + { + "epoch": 5.32, + "learning_rate": 6.690787188451386e-07, + "loss": 0.8405, + "step": 71975 + }, + { + "epoch": 5.32, + "learning_rate": 6.689352771038737e-07, + "loss": 1.0011, + "step": 71976 + }, + { + "epoch": 5.32, + "learning_rate": 6.687918502081959e-07, + "loss": 1.1732, + "step": 71977 + }, + { + "epoch": 5.32, + "learning_rate": 6.68648438158338e-07, + "loss": 1.0105, + "step": 71978 + }, + { + "epoch": 5.32, + "learning_rate": 6.68505040954528e-07, + "loss": 0.848, + "step": 71979 + }, + { + "epoch": 5.32, + "learning_rate": 6.683616585969921e-07, + "loss": 1.0119, + "step": 71980 + }, + { + "epoch": 5.32, + "learning_rate": 6.682182910859569e-07, + "loss": 1.0066, + "step": 71981 + }, + { + "epoch": 5.32, + "learning_rate": 6.680749384216556e-07, + "loss": 1.0043, + "step": 71982 + }, + { + "epoch": 5.32, + "learning_rate": 6.679316006043135e-07, + "loss": 0.9204, + "step": 71983 + }, + { + "epoch": 5.32, + "learning_rate": 6.677882776341582e-07, + "loss": 1.0102, + "step": 71984 + }, + { + "epoch": 5.32, + "learning_rate": 6.676449695114196e-07, + "loss": 0.9136, + "step": 71985 + }, + { + "epoch": 5.32, + "learning_rate": 6.675016762363218e-07, + "loss": 0.8435, + "step": 71986 + }, + { + "epoch": 5.32, + "learning_rate": 6.673583978090958e-07, + "loss": 0.9294, + "step": 71987 + }, + { + "epoch": 5.32, + "learning_rate": 6.672151342299704e-07, + "loss": 1.0222, + "step": 71988 + }, + { + "epoch": 5.32, + "learning_rate": 6.67071885499172e-07, + "loss": 1.0554, + "step": 71989 + }, + { + "epoch": 5.32, + "learning_rate": 6.669286516169271e-07, + "loss": 1.0642, + "step": 71990 + }, + { + "epoch": 5.32, + "learning_rate": 6.667854325834666e-07, + "loss": 0.9551, + "step": 71991 + }, + { + "epoch": 5.32, + "learning_rate": 6.666422283990181e-07, + "loss": 1.0438, + "step": 71992 + }, + { + "epoch": 5.32, + "learning_rate": 6.664990390638059e-07, + "loss": 0.9893, + "step": 71993 + }, + { + "epoch": 5.32, + "learning_rate": 6.663558645780622e-07, + "loss": 1.0561, + "step": 71994 + }, + { + "epoch": 5.32, + "learning_rate": 6.662127049420097e-07, + "loss": 0.9941, + "step": 71995 + }, + { + "epoch": 5.32, + "learning_rate": 6.66069560155883e-07, + "loss": 0.9647, + "step": 71996 + }, + { + "epoch": 5.32, + "learning_rate": 6.659264302199031e-07, + "loss": 1.0705, + "step": 71997 + }, + { + "epoch": 5.32, + "learning_rate": 6.65783315134303e-07, + "loss": 1.1359, + "step": 71998 + }, + { + "epoch": 5.32, + "learning_rate": 6.656402148993058e-07, + "loss": 0.9587, + "step": 71999 + }, + { + "epoch": 5.32, + "learning_rate": 6.654971295151435e-07, + "loss": 1.0518, + "step": 72000 + }, + { + "epoch": 5.32, + "learning_rate": 6.653540589820418e-07, + "loss": 0.9903, + "step": 72001 + }, + { + "epoch": 5.32, + "learning_rate": 6.652110033002257e-07, + "loss": 1.0163, + "step": 72002 + }, + { + "epoch": 5.32, + "learning_rate": 6.650679624699275e-07, + "loss": 0.9928, + "step": 72003 + }, + { + "epoch": 5.32, + "learning_rate": 6.649249364913724e-07, + "loss": 0.9254, + "step": 72004 + }, + { + "epoch": 5.32, + "learning_rate": 6.64781925364788e-07, + "loss": 1.0668, + "step": 72005 + }, + { + "epoch": 5.32, + "learning_rate": 6.64638929090401e-07, + "loss": 1.1737, + "step": 72006 + }, + { + "epoch": 5.32, + "learning_rate": 6.64495947668441e-07, + "loss": 1.0036, + "step": 72007 + }, + { + "epoch": 5.32, + "learning_rate": 6.643529810991345e-07, + "loss": 1.0388, + "step": 72008 + }, + { + "epoch": 5.32, + "learning_rate": 6.642100293827092e-07, + "loss": 1.0049, + "step": 72009 + }, + { + "epoch": 5.32, + "learning_rate": 6.640670925193915e-07, + "loss": 1.0887, + "step": 72010 + }, + { + "epoch": 5.32, + "learning_rate": 6.63924170509408e-07, + "loss": 1.0064, + "step": 72011 + }, + { + "epoch": 5.32, + "learning_rate": 6.637812633529894e-07, + "loss": 0.9274, + "step": 72012 + }, + { + "epoch": 5.32, + "learning_rate": 6.636383710503613e-07, + "loss": 0.9931, + "step": 72013 + }, + { + "epoch": 5.32, + "learning_rate": 6.634954936017501e-07, + "loss": 0.8738, + "step": 72014 + }, + { + "epoch": 5.32, + "learning_rate": 6.633526310073835e-07, + "loss": 1.0631, + "step": 72015 + }, + { + "epoch": 5.32, + "learning_rate": 6.6320978326749e-07, + "loss": 0.9073, + "step": 72016 + }, + { + "epoch": 5.32, + "learning_rate": 6.630669503822973e-07, + "loss": 0.9787, + "step": 72017 + }, + { + "epoch": 5.32, + "learning_rate": 6.629241323520308e-07, + "loss": 0.9783, + "step": 72018 + }, + { + "epoch": 5.32, + "learning_rate": 6.62781329176918e-07, + "loss": 1.048, + "step": 72019 + }, + { + "epoch": 5.32, + "learning_rate": 6.626385408571855e-07, + "loss": 0.9871, + "step": 72020 + }, + { + "epoch": 5.32, + "learning_rate": 6.624957673930643e-07, + "loss": 1.054, + "step": 72021 + }, + { + "epoch": 5.32, + "learning_rate": 6.623530087847762e-07, + "loss": 1.0016, + "step": 72022 + }, + { + "epoch": 5.32, + "learning_rate": 6.622102650325512e-07, + "loss": 1.0619, + "step": 72023 + }, + { + "epoch": 5.32, + "learning_rate": 6.620675361366169e-07, + "loss": 0.9362, + "step": 72024 + }, + { + "epoch": 5.32, + "learning_rate": 6.619248220971997e-07, + "loss": 1.0213, + "step": 72025 + }, + { + "epoch": 5.32, + "learning_rate": 6.61782122914525e-07, + "loss": 1.0347, + "step": 72026 + }, + { + "epoch": 5.32, + "learning_rate": 6.616394385888225e-07, + "loss": 0.9246, + "step": 72027 + }, + { + "epoch": 5.32, + "learning_rate": 6.614967691203189e-07, + "loss": 1.0021, + "step": 72028 + }, + { + "epoch": 5.32, + "learning_rate": 6.613541145092395e-07, + "loss": 1.0543, + "step": 72029 + }, + { + "epoch": 5.32, + "learning_rate": 6.612114747558119e-07, + "loss": 0.9177, + "step": 72030 + }, + { + "epoch": 5.32, + "learning_rate": 6.610688498602624e-07, + "loss": 0.9798, + "step": 72031 + }, + { + "epoch": 5.32, + "learning_rate": 6.6092623982282e-07, + "loss": 1.0411, + "step": 72032 + }, + { + "epoch": 5.32, + "learning_rate": 6.607836446437099e-07, + "loss": 0.9759, + "step": 72033 + }, + { + "epoch": 5.32, + "learning_rate": 6.606410643231598e-07, + "loss": 1.1034, + "step": 72034 + }, + { + "epoch": 5.32, + "learning_rate": 6.604984988613939e-07, + "loss": 0.967, + "step": 72035 + }, + { + "epoch": 5.32, + "learning_rate": 6.603559482586441e-07, + "loss": 0.9457, + "step": 72036 + }, + { + "epoch": 5.32, + "learning_rate": 6.602134125151338e-07, + "loss": 0.8362, + "step": 72037 + }, + { + "epoch": 5.32, + "learning_rate": 6.600708916310894e-07, + "loss": 1.0717, + "step": 72038 + }, + { + "epoch": 5.32, + "learning_rate": 6.599283856067395e-07, + "loss": 0.9432, + "step": 72039 + }, + { + "epoch": 5.32, + "learning_rate": 6.597858944423085e-07, + "loss": 0.9659, + "step": 72040 + }, + { + "epoch": 5.32, + "learning_rate": 6.596434181380274e-07, + "loss": 1.1031, + "step": 72041 + }, + { + "epoch": 5.32, + "learning_rate": 6.595009566941169e-07, + "loss": 0.9433, + "step": 72042 + }, + { + "epoch": 5.32, + "learning_rate": 6.593585101108079e-07, + "loss": 0.9958, + "step": 72043 + }, + { + "epoch": 5.32, + "learning_rate": 6.592160783883251e-07, + "loss": 0.9721, + "step": 72044 + }, + { + "epoch": 5.32, + "learning_rate": 6.590736615268978e-07, + "loss": 0.9772, + "step": 72045 + }, + { + "epoch": 5.32, + "learning_rate": 6.589312595267483e-07, + "loss": 1.0247, + "step": 72046 + }, + { + "epoch": 5.32, + "learning_rate": 6.587888723881076e-07, + "loss": 1.0164, + "step": 72047 + }, + { + "epoch": 5.32, + "learning_rate": 6.586465001111997e-07, + "loss": 1.0768, + "step": 72048 + }, + { + "epoch": 5.32, + "learning_rate": 6.585041426962513e-07, + "loss": 0.964, + "step": 72049 + }, + { + "epoch": 5.32, + "learning_rate": 6.583618001434888e-07, + "loss": 1.1636, + "step": 72050 + }, + { + "epoch": 5.32, + "learning_rate": 6.582194724531388e-07, + "loss": 0.9249, + "step": 72051 + }, + { + "epoch": 5.32, + "learning_rate": 6.580771596254287e-07, + "loss": 0.9404, + "step": 72052 + }, + { + "epoch": 5.32, + "learning_rate": 6.579348616605841e-07, + "loss": 1.0357, + "step": 72053 + }, + { + "epoch": 5.32, + "learning_rate": 6.577925785588313e-07, + "loss": 0.9663, + "step": 72054 + }, + { + "epoch": 5.32, + "learning_rate": 6.576503103203957e-07, + "loss": 1.0037, + "step": 72055 + }, + { + "epoch": 5.32, + "learning_rate": 6.575080569455061e-07, + "loss": 1.0933, + "step": 72056 + }, + { + "epoch": 5.32, + "learning_rate": 6.573658184343879e-07, + "loss": 1.0544, + "step": 72057 + }, + { + "epoch": 5.32, + "learning_rate": 6.572235947872663e-07, + "loss": 1.0102, + "step": 72058 + }, + { + "epoch": 5.32, + "learning_rate": 6.570813860043689e-07, + "loss": 1.1371, + "step": 72059 + }, + { + "epoch": 5.32, + "learning_rate": 6.569391920859192e-07, + "loss": 0.9881, + "step": 72060 + }, + { + "epoch": 5.32, + "learning_rate": 6.567970130321477e-07, + "loss": 0.9814, + "step": 72061 + }, + { + "epoch": 5.32, + "learning_rate": 6.566548488432778e-07, + "loss": 1.0039, + "step": 72062 + }, + { + "epoch": 5.32, + "learning_rate": 6.56512699519537e-07, + "loss": 1.0147, + "step": 72063 + }, + { + "epoch": 5.32, + "learning_rate": 6.563705650611496e-07, + "loss": 0.9603, + "step": 72064 + }, + { + "epoch": 5.32, + "learning_rate": 6.562284454683432e-07, + "loss": 1.0193, + "step": 72065 + }, + { + "epoch": 5.32, + "learning_rate": 6.560863407413465e-07, + "loss": 1.0384, + "step": 72066 + }, + { + "epoch": 5.32, + "learning_rate": 6.559442508803793e-07, + "loss": 1.1258, + "step": 72067 + }, + { + "epoch": 5.33, + "learning_rate": 6.558021758856726e-07, + "loss": 0.962, + "step": 72068 + }, + { + "epoch": 5.33, + "learning_rate": 6.556601157574505e-07, + "loss": 1.0475, + "step": 72069 + }, + { + "epoch": 5.33, + "learning_rate": 6.555180704959396e-07, + "loss": 1.0003, + "step": 72070 + }, + { + "epoch": 5.33, + "learning_rate": 6.553760401013642e-07, + "loss": 0.9949, + "step": 72071 + }, + { + "epoch": 5.33, + "learning_rate": 6.552340245739541e-07, + "loss": 1.1003, + "step": 72072 + }, + { + "epoch": 5.33, + "learning_rate": 6.550920239139324e-07, + "loss": 0.9732, + "step": 72073 + }, + { + "epoch": 5.33, + "learning_rate": 6.549500381215257e-07, + "loss": 1.0215, + "step": 72074 + }, + { + "epoch": 5.33, + "learning_rate": 6.548080671969592e-07, + "loss": 0.9848, + "step": 72075 + }, + { + "epoch": 5.33, + "learning_rate": 6.546661111404584e-07, + "loss": 1.0512, + "step": 72076 + }, + { + "epoch": 5.33, + "learning_rate": 6.54524169952252e-07, + "loss": 0.9851, + "step": 72077 + }, + { + "epoch": 5.33, + "learning_rate": 6.54382243632563e-07, + "loss": 1.0231, + "step": 72078 + }, + { + "epoch": 5.33, + "learning_rate": 6.542403321816182e-07, + "loss": 1.0295, + "step": 72079 + }, + { + "epoch": 5.33, + "learning_rate": 6.540984355996427e-07, + "loss": 1.0418, + "step": 72080 + }, + { + "epoch": 5.33, + "learning_rate": 6.539565538868642e-07, + "loss": 1.0424, + "step": 72081 + }, + { + "epoch": 5.33, + "learning_rate": 6.538146870435069e-07, + "loss": 0.9894, + "step": 72082 + }, + { + "epoch": 5.33, + "learning_rate": 6.536728350697963e-07, + "loss": 1.071, + "step": 72083 + }, + { + "epoch": 5.33, + "learning_rate": 6.535309979659576e-07, + "loss": 1.0739, + "step": 72084 + }, + { + "epoch": 5.33, + "learning_rate": 6.533891757322186e-07, + "loss": 1.0205, + "step": 72085 + }, + { + "epoch": 5.33, + "learning_rate": 6.532473683688045e-07, + "loss": 1.161, + "step": 72086 + }, + { + "epoch": 5.33, + "learning_rate": 6.531055758759375e-07, + "loss": 0.9757, + "step": 72087 + }, + { + "epoch": 5.33, + "learning_rate": 6.529637982538484e-07, + "loss": 0.9339, + "step": 72088 + }, + { + "epoch": 5.33, + "learning_rate": 6.528220355027571e-07, + "loss": 1.0268, + "step": 72089 + }, + { + "epoch": 5.33, + "learning_rate": 6.526802876228966e-07, + "loss": 1.0711, + "step": 72090 + }, + { + "epoch": 5.33, + "learning_rate": 6.525385546144835e-07, + "loss": 0.9342, + "step": 72091 + }, + { + "epoch": 5.33, + "learning_rate": 6.523968364777511e-07, + "loss": 0.9277, + "step": 72092 + }, + { + "epoch": 5.33, + "learning_rate": 6.52255133212919e-07, + "loss": 0.9581, + "step": 72093 + }, + { + "epoch": 5.33, + "learning_rate": 6.521134448202182e-07, + "loss": 0.9789, + "step": 72094 + }, + { + "epoch": 5.33, + "learning_rate": 6.519717712998696e-07, + "loss": 0.9082, + "step": 72095 + }, + { + "epoch": 5.33, + "learning_rate": 6.518301126520987e-07, + "loss": 0.8966, + "step": 72096 + }, + { + "epoch": 5.33, + "learning_rate": 6.516884688771341e-07, + "loss": 1.035, + "step": 72097 + }, + { + "epoch": 5.33, + "learning_rate": 6.515468399752e-07, + "loss": 1.0535, + "step": 72098 + }, + { + "epoch": 5.33, + "learning_rate": 6.514052259465209e-07, + "loss": 1.1189, + "step": 72099 + }, + { + "epoch": 5.33, + "learning_rate": 6.512636267913197e-07, + "loss": 1.0232, + "step": 72100 + }, + { + "epoch": 5.33, + "learning_rate": 6.511220425098264e-07, + "loss": 0.9584, + "step": 72101 + }, + { + "epoch": 5.33, + "learning_rate": 6.509804731022651e-07, + "loss": 1.0204, + "step": 72102 + }, + { + "epoch": 5.33, + "learning_rate": 6.508389185688591e-07, + "loss": 1.0582, + "step": 72103 + }, + { + "epoch": 5.33, + "learning_rate": 6.506973789098347e-07, + "loss": 1.0128, + "step": 72104 + }, + { + "epoch": 5.33, + "learning_rate": 6.505558541254153e-07, + "loss": 1.0085, + "step": 72105 + }, + { + "epoch": 5.33, + "learning_rate": 6.504143442158295e-07, + "loss": 0.9814, + "step": 72106 + }, + { + "epoch": 5.33, + "learning_rate": 6.502728491813015e-07, + "loss": 1.1097, + "step": 72107 + }, + { + "epoch": 5.33, + "learning_rate": 6.501313690220545e-07, + "loss": 0.9133, + "step": 72108 + }, + { + "epoch": 5.33, + "learning_rate": 6.49989903738314e-07, + "loss": 1.0729, + "step": 72109 + }, + { + "epoch": 5.33, + "learning_rate": 6.498484533303062e-07, + "loss": 1.1183, + "step": 72110 + }, + { + "epoch": 5.33, + "learning_rate": 6.497070177982567e-07, + "loss": 0.978, + "step": 72111 + }, + { + "epoch": 5.33, + "learning_rate": 6.495655971423898e-07, + "loss": 0.9561, + "step": 72112 + }, + { + "epoch": 5.33, + "learning_rate": 6.494241913629295e-07, + "loss": 1.0277, + "step": 72113 + }, + { + "epoch": 5.33, + "learning_rate": 6.492828004601015e-07, + "loss": 0.9474, + "step": 72114 + }, + { + "epoch": 5.33, + "learning_rate": 6.491414244341309e-07, + "loss": 0.9437, + "step": 72115 + }, + { + "epoch": 5.33, + "learning_rate": 6.490000632852412e-07, + "loss": 1.0005, + "step": 72116 + }, + { + "epoch": 5.33, + "learning_rate": 6.488587170136606e-07, + "loss": 1.0597, + "step": 72117 + }, + { + "epoch": 5.33, + "learning_rate": 6.487173856196116e-07, + "loss": 1.064, + "step": 72118 + }, + { + "epoch": 5.33, + "learning_rate": 6.485760691033194e-07, + "loss": 1.0463, + "step": 72119 + }, + { + "epoch": 5.33, + "learning_rate": 6.484347674650071e-07, + "loss": 1.1085, + "step": 72120 + }, + { + "epoch": 5.33, + "learning_rate": 6.482934807049024e-07, + "loss": 0.9658, + "step": 72121 + }, + { + "epoch": 5.33, + "learning_rate": 6.481522088232295e-07, + "loss": 1.0744, + "step": 72122 + }, + { + "epoch": 5.33, + "learning_rate": 6.480109518202127e-07, + "loss": 1.1685, + "step": 72123 + }, + { + "epoch": 5.33, + "learning_rate": 6.478697096960773e-07, + "loss": 1.0788, + "step": 72124 + }, + { + "epoch": 5.33, + "learning_rate": 6.477284824510455e-07, + "loss": 0.9607, + "step": 72125 + }, + { + "epoch": 5.33, + "learning_rate": 6.475872700853448e-07, + "loss": 1.089, + "step": 72126 + }, + { + "epoch": 5.33, + "learning_rate": 6.474460725991993e-07, + "loss": 1.121, + "step": 72127 + }, + { + "epoch": 5.33, + "learning_rate": 6.473048899928325e-07, + "loss": 0.9606, + "step": 72128 + }, + { + "epoch": 5.33, + "learning_rate": 6.471637222664695e-07, + "loss": 1.1561, + "step": 72129 + }, + { + "epoch": 5.33, + "learning_rate": 6.470225694203358e-07, + "loss": 0.8937, + "step": 72130 + }, + { + "epoch": 5.33, + "learning_rate": 6.468814314546578e-07, + "loss": 0.9687, + "step": 72131 + }, + { + "epoch": 5.33, + "learning_rate": 6.467403083696544e-07, + "loss": 1.126, + "step": 72132 + }, + { + "epoch": 5.33, + "learning_rate": 6.465992001655541e-07, + "loss": 1.0066, + "step": 72133 + }, + { + "epoch": 5.33, + "learning_rate": 6.464581068425801e-07, + "loss": 1.045, + "step": 72134 + }, + { + "epoch": 5.33, + "learning_rate": 6.463170284009601e-07, + "loss": 0.9352, + "step": 72135 + }, + { + "epoch": 5.33, + "learning_rate": 6.461759648409127e-07, + "loss": 1.0272, + "step": 72136 + }, + { + "epoch": 5.33, + "learning_rate": 6.460349161626667e-07, + "loss": 0.943, + "step": 72137 + }, + { + "epoch": 5.33, + "learning_rate": 6.45893882366444e-07, + "loss": 1.105, + "step": 72138 + }, + { + "epoch": 5.33, + "learning_rate": 6.457528634524735e-07, + "loss": 0.9807, + "step": 72139 + }, + { + "epoch": 5.33, + "learning_rate": 6.456118594209737e-07, + "loss": 1.0008, + "step": 72140 + }, + { + "epoch": 5.33, + "learning_rate": 6.454708702721723e-07, + "loss": 1.0541, + "step": 72141 + }, + { + "epoch": 5.33, + "learning_rate": 6.453298960062937e-07, + "loss": 0.8932, + "step": 72142 + }, + { + "epoch": 5.33, + "learning_rate": 6.451889366235609e-07, + "loss": 1.0517, + "step": 72143 + }, + { + "epoch": 5.33, + "learning_rate": 6.450479921241993e-07, + "loss": 0.9827, + "step": 72144 + }, + { + "epoch": 5.33, + "learning_rate": 6.449070625084297e-07, + "loss": 0.9487, + "step": 72145 + }, + { + "epoch": 5.33, + "learning_rate": 6.447661477764811e-07, + "loss": 1.0049, + "step": 72146 + }, + { + "epoch": 5.33, + "learning_rate": 6.446252479285764e-07, + "loss": 1.1025, + "step": 72147 + }, + { + "epoch": 5.33, + "learning_rate": 6.444843629649389e-07, + "loss": 0.9413, + "step": 72148 + }, + { + "epoch": 5.33, + "learning_rate": 6.443434928857906e-07, + "loss": 1.0293, + "step": 72149 + }, + { + "epoch": 5.33, + "learning_rate": 6.442026376913602e-07, + "loss": 1.0536, + "step": 72150 + }, + { + "epoch": 5.33, + "learning_rate": 6.440617973818697e-07, + "loss": 1.0232, + "step": 72151 + }, + { + "epoch": 5.33, + "learning_rate": 6.439209719575423e-07, + "loss": 1.0485, + "step": 72152 + }, + { + "epoch": 5.33, + "learning_rate": 6.437801614186035e-07, + "loss": 1.0651, + "step": 72153 + }, + { + "epoch": 5.33, + "learning_rate": 6.43639365765275e-07, + "loss": 1.0494, + "step": 72154 + }, + { + "epoch": 5.33, + "learning_rate": 6.434985849977838e-07, + "loss": 0.9735, + "step": 72155 + }, + { + "epoch": 5.33, + "learning_rate": 6.433578191163515e-07, + "loss": 0.9267, + "step": 72156 + }, + { + "epoch": 5.33, + "learning_rate": 6.432170681212046e-07, + "loss": 1.0173, + "step": 72157 + }, + { + "epoch": 5.33, + "learning_rate": 6.430763320125632e-07, + "loss": 0.8826, + "step": 72158 + }, + { + "epoch": 5.33, + "learning_rate": 6.429356107906559e-07, + "loss": 1.1359, + "step": 72159 + }, + { + "epoch": 5.33, + "learning_rate": 6.427949044557036e-07, + "loss": 1.0409, + "step": 72160 + }, + { + "epoch": 5.33, + "learning_rate": 6.426542130079294e-07, + "loss": 0.9276, + "step": 72161 + }, + { + "epoch": 5.33, + "learning_rate": 6.425135364475587e-07, + "loss": 1.1185, + "step": 72162 + }, + { + "epoch": 5.33, + "learning_rate": 6.423728747748159e-07, + "loss": 1.0301, + "step": 72163 + }, + { + "epoch": 5.33, + "learning_rate": 6.42232227989924e-07, + "loss": 1.1354, + "step": 72164 + }, + { + "epoch": 5.33, + "learning_rate": 6.420915960931051e-07, + "loss": 1.0161, + "step": 72165 + }, + { + "epoch": 5.33, + "learning_rate": 6.419509790845868e-07, + "loss": 0.9589, + "step": 72166 + }, + { + "epoch": 5.33, + "learning_rate": 6.4181037696459e-07, + "loss": 0.9581, + "step": 72167 + }, + { + "epoch": 5.33, + "learning_rate": 6.41669789733339e-07, + "loss": 0.8939, + "step": 72168 + }, + { + "epoch": 5.33, + "learning_rate": 6.41529217391057e-07, + "loss": 1.0127, + "step": 72169 + }, + { + "epoch": 5.33, + "learning_rate": 6.413886599379671e-07, + "loss": 1.0291, + "step": 72170 + }, + { + "epoch": 5.33, + "learning_rate": 6.412481173742958e-07, + "loss": 1.0106, + "step": 72171 + }, + { + "epoch": 5.33, + "learning_rate": 6.411075897002638e-07, + "loss": 1.0542, + "step": 72172 + }, + { + "epoch": 5.33, + "learning_rate": 6.409670769160969e-07, + "loss": 0.9664, + "step": 72173 + }, + { + "epoch": 5.33, + "learning_rate": 6.408265790220147e-07, + "loss": 0.9407, + "step": 72174 + }, + { + "epoch": 5.33, + "learning_rate": 6.40686096018246e-07, + "loss": 0.9989, + "step": 72175 + }, + { + "epoch": 5.33, + "learning_rate": 6.405456279050115e-07, + "loss": 1.0059, + "step": 72176 + }, + { + "epoch": 5.33, + "learning_rate": 6.404051746825357e-07, + "loss": 1.0759, + "step": 72177 + }, + { + "epoch": 5.33, + "learning_rate": 6.402647363510405e-07, + "loss": 0.9929, + "step": 72178 + }, + { + "epoch": 5.33, + "learning_rate": 6.401243129107482e-07, + "loss": 1.0614, + "step": 72179 + }, + { + "epoch": 5.33, + "learning_rate": 6.399839043618883e-07, + "loss": 1.0543, + "step": 72180 + }, + { + "epoch": 5.33, + "learning_rate": 6.398435107046763e-07, + "loss": 1.0522, + "step": 72181 + }, + { + "epoch": 5.33, + "learning_rate": 6.397031319393409e-07, + "loss": 0.8947, + "step": 72182 + }, + { + "epoch": 5.33, + "learning_rate": 6.395627680661032e-07, + "loss": 1.0529, + "step": 72183 + }, + { + "epoch": 5.33, + "learning_rate": 6.394224190851895e-07, + "loss": 0.9067, + "step": 72184 + }, + { + "epoch": 5.33, + "learning_rate": 6.392820849968184e-07, + "loss": 1.0039, + "step": 72185 + }, + { + "epoch": 5.33, + "learning_rate": 6.391417658012167e-07, + "loss": 1.0266, + "step": 72186 + }, + { + "epoch": 5.33, + "learning_rate": 6.39001461498605e-07, + "loss": 0.9847, + "step": 72187 + }, + { + "epoch": 5.33, + "learning_rate": 6.388611720892113e-07, + "loss": 0.9911, + "step": 72188 + }, + { + "epoch": 5.33, + "learning_rate": 6.387208975732539e-07, + "loss": 0.9739, + "step": 72189 + }, + { + "epoch": 5.33, + "learning_rate": 6.385806379509552e-07, + "loss": 1.0268, + "step": 72190 + }, + { + "epoch": 5.33, + "learning_rate": 6.384403932225436e-07, + "loss": 0.9391, + "step": 72191 + }, + { + "epoch": 5.33, + "learning_rate": 6.383001633882391e-07, + "loss": 1.0162, + "step": 72192 + }, + { + "epoch": 5.33, + "learning_rate": 6.381599484482648e-07, + "loss": 0.9342, + "step": 72193 + }, + { + "epoch": 5.33, + "learning_rate": 6.380197484028428e-07, + "loss": 0.9709, + "step": 72194 + }, + { + "epoch": 5.33, + "learning_rate": 6.378795632521984e-07, + "loss": 0.9292, + "step": 72195 + }, + { + "epoch": 5.33, + "learning_rate": 6.377393929965536e-07, + "loss": 1.0406, + "step": 72196 + }, + { + "epoch": 5.33, + "learning_rate": 6.375992376361329e-07, + "loss": 0.8766, + "step": 72197 + }, + { + "epoch": 5.33, + "learning_rate": 6.374590971711558e-07, + "loss": 0.906, + "step": 72198 + }, + { + "epoch": 5.33, + "learning_rate": 6.373189716018469e-07, + "loss": 0.9248, + "step": 72199 + }, + { + "epoch": 5.33, + "learning_rate": 6.371788609284335e-07, + "loss": 0.9817, + "step": 72200 + }, + { + "epoch": 5.33, + "learning_rate": 6.370387651511301e-07, + "loss": 1.0038, + "step": 72201 + }, + { + "epoch": 5.33, + "learning_rate": 6.368986842701663e-07, + "loss": 1.0332, + "step": 72202 + }, + { + "epoch": 5.34, + "learning_rate": 6.36758618285761e-07, + "loss": 1.0262, + "step": 72203 + }, + { + "epoch": 5.34, + "learning_rate": 6.366185671981417e-07, + "loss": 1.0431, + "step": 72204 + }, + { + "epoch": 5.34, + "learning_rate": 6.364785310075249e-07, + "loss": 0.9912, + "step": 72205 + }, + { + "epoch": 5.34, + "learning_rate": 6.363385097141395e-07, + "loss": 0.9431, + "step": 72206 + }, + { + "epoch": 5.34, + "learning_rate": 6.36198503318205e-07, + "loss": 0.8838, + "step": 72207 + }, + { + "epoch": 5.34, + "learning_rate": 6.360585118199436e-07, + "loss": 0.9938, + "step": 72208 + }, + { + "epoch": 5.34, + "learning_rate": 6.359185352195807e-07, + "loss": 1.0117, + "step": 72209 + }, + { + "epoch": 5.34, + "learning_rate": 6.35778573517335e-07, + "loss": 1.0014, + "step": 72210 + }, + { + "epoch": 5.34, + "learning_rate": 6.35638626713434e-07, + "loss": 0.9749, + "step": 72211 + }, + { + "epoch": 5.34, + "learning_rate": 6.354986948080966e-07, + "loss": 0.9857, + "step": 72212 + }, + { + "epoch": 5.34, + "learning_rate": 6.35358777801548e-07, + "loss": 0.9798, + "step": 72213 + }, + { + "epoch": 5.34, + "learning_rate": 6.35218875694007e-07, + "loss": 0.9405, + "step": 72214 + }, + { + "epoch": 5.34, + "learning_rate": 6.350789884857012e-07, + "loss": 0.9643, + "step": 72215 + }, + { + "epoch": 5.34, + "learning_rate": 6.349391161768504e-07, + "loss": 0.9756, + "step": 72216 + }, + { + "epoch": 5.34, + "learning_rate": 6.347992587676777e-07, + "loss": 0.9717, + "step": 72217 + }, + { + "epoch": 5.34, + "learning_rate": 6.346594162584052e-07, + "loss": 1.1016, + "step": 72218 + }, + { + "epoch": 5.34, + "learning_rate": 6.345195886492539e-07, + "loss": 0.991, + "step": 72219 + }, + { + "epoch": 5.34, + "learning_rate": 6.343797759404502e-07, + "loss": 1.0223, + "step": 72220 + }, + { + "epoch": 5.34, + "learning_rate": 6.34239978132214e-07, + "loss": 0.953, + "step": 72221 + }, + { + "epoch": 5.34, + "learning_rate": 6.341001952247671e-07, + "loss": 1.0031, + "step": 72222 + }, + { + "epoch": 5.34, + "learning_rate": 6.33960427218332e-07, + "loss": 0.9903, + "step": 72223 + }, + { + "epoch": 5.34, + "learning_rate": 6.338206741131336e-07, + "loss": 1.0765, + "step": 72224 + }, + { + "epoch": 5.34, + "learning_rate": 6.336809359093932e-07, + "loss": 1.0323, + "step": 72225 + }, + { + "epoch": 5.34, + "learning_rate": 6.335412126073304e-07, + "loss": 1.059, + "step": 72226 + }, + { + "epoch": 5.34, + "learning_rate": 6.334015042071706e-07, + "loss": 0.9649, + "step": 72227 + }, + { + "epoch": 5.34, + "learning_rate": 6.332618107091326e-07, + "loss": 1.0656, + "step": 72228 + }, + { + "epoch": 5.34, + "learning_rate": 6.331221321134451e-07, + "loss": 0.9631, + "step": 72229 + }, + { + "epoch": 5.34, + "learning_rate": 6.329824684203234e-07, + "loss": 1.1555, + "step": 72230 + }, + { + "epoch": 5.34, + "learning_rate": 6.328428196299929e-07, + "loss": 0.9438, + "step": 72231 + }, + { + "epoch": 5.34, + "learning_rate": 6.327031857426747e-07, + "loss": 1.0497, + "step": 72232 + }, + { + "epoch": 5.34, + "learning_rate": 6.325635667585949e-07, + "loss": 1.0355, + "step": 72233 + }, + { + "epoch": 5.34, + "learning_rate": 6.324239626779716e-07, + "loss": 0.9496, + "step": 72234 + }, + { + "epoch": 5.34, + "learning_rate": 6.322843735010253e-07, + "loss": 0.9881, + "step": 72235 + }, + { + "epoch": 5.34, + "learning_rate": 6.321447992279817e-07, + "loss": 0.9835, + "step": 72236 + }, + { + "epoch": 5.34, + "learning_rate": 6.320052398590626e-07, + "loss": 1.0771, + "step": 72237 + }, + { + "epoch": 5.34, + "learning_rate": 6.318656953944879e-07, + "loss": 0.8926, + "step": 72238 + }, + { + "epoch": 5.34, + "learning_rate": 6.317261658344808e-07, + "loss": 1.0362, + "step": 72239 + }, + { + "epoch": 5.34, + "learning_rate": 6.315866511792634e-07, + "loss": 0.9998, + "step": 72240 + }, + { + "epoch": 5.34, + "learning_rate": 6.314471514290588e-07, + "loss": 0.9504, + "step": 72241 + }, + { + "epoch": 5.34, + "learning_rate": 6.313076665840878e-07, + "loss": 0.962, + "step": 72242 + }, + { + "epoch": 5.34, + "learning_rate": 6.311681966445693e-07, + "loss": 1.0864, + "step": 72243 + }, + { + "epoch": 5.34, + "learning_rate": 6.310287416107308e-07, + "loss": 1.0057, + "step": 72244 + }, + { + "epoch": 5.34, + "learning_rate": 6.308893014827932e-07, + "loss": 0.949, + "step": 72245 + }, + { + "epoch": 5.34, + "learning_rate": 6.307498762609732e-07, + "loss": 0.9136, + "step": 72246 + }, + { + "epoch": 5.34, + "learning_rate": 6.306104659454981e-07, + "loss": 1.051, + "step": 72247 + }, + { + "epoch": 5.34, + "learning_rate": 6.304710705365858e-07, + "loss": 0.9968, + "step": 72248 + }, + { + "epoch": 5.34, + "learning_rate": 6.303316900344625e-07, + "loss": 1.026, + "step": 72249 + }, + { + "epoch": 5.34, + "learning_rate": 6.301923244393449e-07, + "loss": 1.0859, + "step": 72250 + }, + { + "epoch": 5.34, + "learning_rate": 6.300529737514582e-07, + "loss": 0.9546, + "step": 72251 + }, + { + "epoch": 5.34, + "learning_rate": 6.299136379710225e-07, + "loss": 1.0266, + "step": 72252 + }, + { + "epoch": 5.34, + "learning_rate": 6.297743170982629e-07, + "loss": 0.9523, + "step": 72253 + }, + { + "epoch": 5.34, + "learning_rate": 6.296350111333971e-07, + "loss": 1.0125, + "step": 72254 + }, + { + "epoch": 5.34, + "learning_rate": 6.29495720076646e-07, + "loss": 0.9401, + "step": 72255 + }, + { + "epoch": 5.34, + "learning_rate": 6.29356443928234e-07, + "loss": 0.9589, + "step": 72256 + }, + { + "epoch": 5.34, + "learning_rate": 6.292171826883831e-07, + "loss": 1.0207, + "step": 72257 + }, + { + "epoch": 5.34, + "learning_rate": 6.29077936357313e-07, + "loss": 1.2137, + "step": 72258 + }, + { + "epoch": 5.34, + "learning_rate": 6.289387049352447e-07, + "loss": 0.9268, + "step": 72259 + }, + { + "epoch": 5.34, + "learning_rate": 6.287994884224013e-07, + "loss": 0.9434, + "step": 72260 + }, + { + "epoch": 5.34, + "learning_rate": 6.28660286819005e-07, + "loss": 0.9151, + "step": 72261 + }, + { + "epoch": 5.34, + "learning_rate": 6.285211001252755e-07, + "loss": 0.9086, + "step": 72262 + }, + { + "epoch": 5.34, + "learning_rate": 6.283819283414349e-07, + "loss": 1.0281, + "step": 72263 + }, + { + "epoch": 5.34, + "learning_rate": 6.282427714677031e-07, + "loss": 0.9538, + "step": 72264 + }, + { + "epoch": 5.34, + "learning_rate": 6.281036295043041e-07, + "loss": 0.9953, + "step": 72265 + }, + { + "epoch": 5.34, + "learning_rate": 6.279645024514591e-07, + "loss": 0.9284, + "step": 72266 + }, + { + "epoch": 5.34, + "learning_rate": 6.278253903093878e-07, + "loss": 0.9911, + "step": 72267 + }, + { + "epoch": 5.34, + "learning_rate": 6.276862930783112e-07, + "loss": 1.0147, + "step": 72268 + }, + { + "epoch": 5.34, + "learning_rate": 6.275472107584523e-07, + "loss": 0.9849, + "step": 72269 + }, + { + "epoch": 5.34, + "learning_rate": 6.274081433500323e-07, + "loss": 1.1176, + "step": 72270 + }, + { + "epoch": 5.34, + "learning_rate": 6.272690908532719e-07, + "loss": 1.018, + "step": 72271 + }, + { + "epoch": 5.34, + "learning_rate": 6.27130053268391e-07, + "loss": 1.066, + "step": 72272 + }, + { + "epoch": 5.34, + "learning_rate": 6.269910305956117e-07, + "loss": 0.9802, + "step": 72273 + }, + { + "epoch": 5.34, + "learning_rate": 6.268520228351594e-07, + "loss": 0.9888, + "step": 72274 + }, + { + "epoch": 5.34, + "learning_rate": 6.267130299872471e-07, + "loss": 1.0592, + "step": 72275 + }, + { + "epoch": 5.34, + "learning_rate": 6.265740520521024e-07, + "loss": 1.0212, + "step": 72276 + }, + { + "epoch": 5.34, + "learning_rate": 6.264350890299442e-07, + "loss": 0.9977, + "step": 72277 + }, + { + "epoch": 5.34, + "learning_rate": 6.262961409209933e-07, + "loss": 0.9224, + "step": 72278 + }, + { + "epoch": 5.34, + "learning_rate": 6.261572077254697e-07, + "loss": 0.925, + "step": 72279 + }, + { + "epoch": 5.34, + "learning_rate": 6.260182894435974e-07, + "loss": 1.0485, + "step": 72280 + }, + { + "epoch": 5.34, + "learning_rate": 6.258793860755963e-07, + "loss": 1.0636, + "step": 72281 + }, + { + "epoch": 5.34, + "learning_rate": 6.257404976216863e-07, + "loss": 0.9962, + "step": 72282 + }, + { + "epoch": 5.34, + "learning_rate": 6.256016240820895e-07, + "loss": 0.9485, + "step": 72283 + }, + { + "epoch": 5.34, + "learning_rate": 6.254627654570245e-07, + "loss": 1.0137, + "step": 72284 + }, + { + "epoch": 5.34, + "learning_rate": 6.253239217467155e-07, + "loss": 1.059, + "step": 72285 + }, + { + "epoch": 5.34, + "learning_rate": 6.251850929513825e-07, + "loss": 0.992, + "step": 72286 + }, + { + "epoch": 5.34, + "learning_rate": 6.250462790712464e-07, + "loss": 1.0313, + "step": 72287 + }, + { + "epoch": 5.34, + "learning_rate": 6.249074801065247e-07, + "loss": 0.9703, + "step": 72288 + }, + { + "epoch": 5.34, + "learning_rate": 6.24768696057444e-07, + "loss": 1.0044, + "step": 72289 + }, + { + "epoch": 5.34, + "learning_rate": 6.246299269242229e-07, + "loss": 0.9576, + "step": 72290 + }, + { + "epoch": 5.34, + "learning_rate": 6.24491172707078e-07, + "loss": 0.8571, + "step": 72291 + }, + { + "epoch": 5.34, + "learning_rate": 6.243524334062357e-07, + "loss": 1.0282, + "step": 72292 + }, + { + "epoch": 5.34, + "learning_rate": 6.242137090219125e-07, + "loss": 0.9444, + "step": 72293 + }, + { + "epoch": 5.34, + "learning_rate": 6.240749995543349e-07, + "loss": 1.0477, + "step": 72294 + }, + { + "epoch": 5.34, + "learning_rate": 6.239363050037162e-07, + "loss": 0.9625, + "step": 72295 + }, + { + "epoch": 5.34, + "learning_rate": 6.237976253702827e-07, + "loss": 0.9947, + "step": 72296 + }, + { + "epoch": 5.34, + "learning_rate": 6.23658960654251e-07, + "loss": 0.9234, + "step": 72297 + }, + { + "epoch": 5.34, + "learning_rate": 6.235203108558474e-07, + "loss": 1.0129, + "step": 72298 + }, + { + "epoch": 5.34, + "learning_rate": 6.233816759752853e-07, + "loss": 1.0217, + "step": 72299 + }, + { + "epoch": 5.34, + "learning_rate": 6.2324305601279e-07, + "loss": 1.0608, + "step": 72300 + }, + { + "epoch": 5.34, + "learning_rate": 6.231044509685813e-07, + "loss": 1.0372, + "step": 72301 + }, + { + "epoch": 5.34, + "learning_rate": 6.22965860842879e-07, + "loss": 0.9412, + "step": 72302 + }, + { + "epoch": 5.34, + "learning_rate": 6.228272856359041e-07, + "loss": 0.9946, + "step": 72303 + }, + { + "epoch": 5.34, + "learning_rate": 6.226887253478753e-07, + "loss": 0.9796, + "step": 72304 + }, + { + "epoch": 5.34, + "learning_rate": 6.225501799790168e-07, + "loss": 1.0193, + "step": 72305 + }, + { + "epoch": 5.34, + "learning_rate": 6.224116495295452e-07, + "loss": 1.0874, + "step": 72306 + }, + { + "epoch": 5.34, + "learning_rate": 6.222731339996834e-07, + "loss": 0.8987, + "step": 72307 + }, + { + "epoch": 5.34, + "learning_rate": 6.221346333896494e-07, + "loss": 0.8996, + "step": 72308 + }, + { + "epoch": 5.34, + "learning_rate": 6.219961476996661e-07, + "loss": 0.8564, + "step": 72309 + }, + { + "epoch": 5.34, + "learning_rate": 6.218576769299534e-07, + "loss": 0.9648, + "step": 72310 + }, + { + "epoch": 5.34, + "learning_rate": 6.21719221080731e-07, + "loss": 1.0077, + "step": 72311 + }, + { + "epoch": 5.34, + "learning_rate": 6.215807801522189e-07, + "loss": 0.952, + "step": 72312 + }, + { + "epoch": 5.34, + "learning_rate": 6.214423541446357e-07, + "loss": 0.9898, + "step": 72313 + }, + { + "epoch": 5.34, + "learning_rate": 6.213039430582058e-07, + "loss": 0.95, + "step": 72314 + }, + { + "epoch": 5.34, + "learning_rate": 6.211655468931465e-07, + "loss": 1.0919, + "step": 72315 + }, + { + "epoch": 5.34, + "learning_rate": 6.21027165649678e-07, + "loss": 1.0417, + "step": 72316 + }, + { + "epoch": 5.34, + "learning_rate": 6.20888799328021e-07, + "loss": 0.9497, + "step": 72317 + }, + { + "epoch": 5.34, + "learning_rate": 6.207504479283966e-07, + "loss": 1.064, + "step": 72318 + }, + { + "epoch": 5.34, + "learning_rate": 6.206121114510255e-07, + "loss": 1.0386, + "step": 72319 + }, + { + "epoch": 5.34, + "learning_rate": 6.204737898961233e-07, + "loss": 1.0165, + "step": 72320 + }, + { + "epoch": 5.34, + "learning_rate": 6.203354832639142e-07, + "loss": 1.0004, + "step": 72321 + }, + { + "epoch": 5.34, + "learning_rate": 6.201971915546179e-07, + "loss": 1.0563, + "step": 72322 + }, + { + "epoch": 5.34, + "learning_rate": 6.200589147684532e-07, + "loss": 0.9844, + "step": 72323 + }, + { + "epoch": 5.34, + "learning_rate": 6.19920652905639e-07, + "loss": 0.939, + "step": 72324 + }, + { + "epoch": 5.34, + "learning_rate": 6.197824059663993e-07, + "loss": 0.9776, + "step": 72325 + }, + { + "epoch": 5.34, + "learning_rate": 6.196441739509507e-07, + "loss": 0.9752, + "step": 72326 + }, + { + "epoch": 5.34, + "learning_rate": 6.19505956859514e-07, + "loss": 0.9313, + "step": 72327 + }, + { + "epoch": 5.34, + "learning_rate": 6.193677546923093e-07, + "loss": 0.9814, + "step": 72328 + }, + { + "epoch": 5.34, + "learning_rate": 6.192295674495541e-07, + "loss": 0.9319, + "step": 72329 + }, + { + "epoch": 5.34, + "learning_rate": 6.190913951314726e-07, + "loss": 1.0523, + "step": 72330 + }, + { + "epoch": 5.34, + "learning_rate": 6.189532377382823e-07, + "loss": 0.9644, + "step": 72331 + }, + { + "epoch": 5.34, + "learning_rate": 6.188150952702044e-07, + "loss": 0.9622, + "step": 72332 + }, + { + "epoch": 5.34, + "learning_rate": 6.18676967727454e-07, + "loss": 0.9381, + "step": 72333 + }, + { + "epoch": 5.34, + "learning_rate": 6.185388551102567e-07, + "loss": 1.0463, + "step": 72334 + }, + { + "epoch": 5.34, + "learning_rate": 6.1840075741883e-07, + "loss": 0.9638, + "step": 72335 + }, + { + "epoch": 5.34, + "learning_rate": 6.182626746533937e-07, + "loss": 0.986, + "step": 72336 + }, + { + "epoch": 5.34, + "learning_rate": 6.181246068141667e-07, + "loss": 0.9, + "step": 72337 + }, + { + "epoch": 5.35, + "learning_rate": 6.179865539013674e-07, + "loss": 0.9393, + "step": 72338 + }, + { + "epoch": 5.35, + "learning_rate": 6.178485159152215e-07, + "loss": 0.9374, + "step": 72339 + }, + { + "epoch": 5.35, + "learning_rate": 6.177104928559407e-07, + "loss": 1.0379, + "step": 72340 + }, + { + "epoch": 5.35, + "learning_rate": 6.175724847237497e-07, + "loss": 1.0631, + "step": 72341 + }, + { + "epoch": 5.35, + "learning_rate": 6.174344915188646e-07, + "loss": 0.9087, + "step": 72342 + }, + { + "epoch": 5.35, + "learning_rate": 6.17296513241511e-07, + "loss": 1.0862, + "step": 72343 + }, + { + "epoch": 5.35, + "learning_rate": 6.171585498919008e-07, + "loss": 1.0449, + "step": 72344 + }, + { + "epoch": 5.35, + "learning_rate": 6.170206014702585e-07, + "loss": 1.0541, + "step": 72345 + }, + { + "epoch": 5.35, + "learning_rate": 6.168826679768025e-07, + "loss": 1.061, + "step": 72346 + }, + { + "epoch": 5.35, + "learning_rate": 6.167447494117518e-07, + "loss": 0.9021, + "step": 72347 + }, + { + "epoch": 5.35, + "learning_rate": 6.16606845775326e-07, + "loss": 1.0545, + "step": 72348 + }, + { + "epoch": 5.35, + "learning_rate": 6.164689570677429e-07, + "loss": 0.9149, + "step": 72349 + }, + { + "epoch": 5.35, + "learning_rate": 6.163310832892256e-07, + "loss": 1.0338, + "step": 72350 + }, + { + "epoch": 5.35, + "learning_rate": 6.161932244399904e-07, + "loss": 1.082, + "step": 72351 + }, + { + "epoch": 5.35, + "learning_rate": 6.160553805202574e-07, + "loss": 1.0803, + "step": 72352 + }, + { + "epoch": 5.35, + "learning_rate": 6.159175515302451e-07, + "loss": 1.0497, + "step": 72353 + }, + { + "epoch": 5.35, + "learning_rate": 6.157797374701746e-07, + "loss": 1.0108, + "step": 72354 + }, + { + "epoch": 5.35, + "learning_rate": 6.156419383402645e-07, + "loss": 0.8887, + "step": 72355 + }, + { + "epoch": 5.35, + "learning_rate": 6.155041541407347e-07, + "loss": 0.9591, + "step": 72356 + }, + { + "epoch": 5.35, + "learning_rate": 6.153663848718027e-07, + "loss": 1.0088, + "step": 72357 + }, + { + "epoch": 5.35, + "learning_rate": 6.152286305336874e-07, + "loss": 0.8685, + "step": 72358 + }, + { + "epoch": 5.35, + "learning_rate": 6.150908911266095e-07, + "loss": 0.8978, + "step": 72359 + }, + { + "epoch": 5.35, + "learning_rate": 6.14953166650789e-07, + "loss": 0.9735, + "step": 72360 + }, + { + "epoch": 5.35, + "learning_rate": 6.148154571064435e-07, + "loss": 0.9528, + "step": 72361 + }, + { + "epoch": 5.35, + "learning_rate": 6.146777624937905e-07, + "loss": 0.9843, + "step": 72362 + }, + { + "epoch": 5.35, + "learning_rate": 6.145400828130521e-07, + "loss": 1.0053, + "step": 72363 + }, + { + "epoch": 5.35, + "learning_rate": 6.14402418064447e-07, + "loss": 0.9923, + "step": 72364 + }, + { + "epoch": 5.35, + "learning_rate": 6.142647682481928e-07, + "loss": 1.0419, + "step": 72365 + }, + { + "epoch": 5.35, + "learning_rate": 6.141271333645093e-07, + "loss": 1.0269, + "step": 72366 + }, + { + "epoch": 5.35, + "learning_rate": 6.139895134136143e-07, + "loss": 1.0183, + "step": 72367 + }, + { + "epoch": 5.35, + "learning_rate": 6.138519083957284e-07, + "loss": 0.9815, + "step": 72368 + }, + { + "epoch": 5.35, + "learning_rate": 6.137143183110683e-07, + "loss": 1.0325, + "step": 72369 + }, + { + "epoch": 5.35, + "learning_rate": 6.13576743159856e-07, + "loss": 1.0852, + "step": 72370 + }, + { + "epoch": 5.35, + "learning_rate": 6.134391829423092e-07, + "loss": 1.0352, + "step": 72371 + }, + { + "epoch": 5.35, + "learning_rate": 6.133016376586465e-07, + "loss": 0.9411, + "step": 72372 + }, + { + "epoch": 5.35, + "learning_rate": 6.131641073090844e-07, + "loss": 0.907, + "step": 72373 + }, + { + "epoch": 5.35, + "learning_rate": 6.13026591893845e-07, + "loss": 0.9782, + "step": 72374 + }, + { + "epoch": 5.35, + "learning_rate": 6.128890914131469e-07, + "loss": 1.0194, + "step": 72375 + }, + { + "epoch": 5.35, + "learning_rate": 6.127516058672079e-07, + "loss": 1.0939, + "step": 72376 + }, + { + "epoch": 5.35, + "learning_rate": 6.126141352562464e-07, + "loss": 1.0089, + "step": 72377 + }, + { + "epoch": 5.35, + "learning_rate": 6.124766795804804e-07, + "loss": 0.9863, + "step": 72378 + }, + { + "epoch": 5.35, + "learning_rate": 6.123392388401306e-07, + "loss": 1.0173, + "step": 72379 + }, + { + "epoch": 5.35, + "learning_rate": 6.122018130354146e-07, + "loss": 1.0176, + "step": 72380 + }, + { + "epoch": 5.35, + "learning_rate": 6.120644021665523e-07, + "loss": 1.0928, + "step": 72381 + }, + { + "epoch": 5.35, + "learning_rate": 6.11927006233759e-07, + "loss": 0.9241, + "step": 72382 + }, + { + "epoch": 5.35, + "learning_rate": 6.117896252372569e-07, + "loss": 1.0295, + "step": 72383 + }, + { + "epoch": 5.35, + "learning_rate": 6.116522591772645e-07, + "loss": 0.9445, + "step": 72384 + }, + { + "epoch": 5.35, + "learning_rate": 6.115149080539973e-07, + "loss": 0.9083, + "step": 72385 + }, + { + "epoch": 5.35, + "learning_rate": 6.113775718676762e-07, + "loss": 1.0268, + "step": 72386 + }, + { + "epoch": 5.35, + "learning_rate": 6.112402506185167e-07, + "loss": 0.9258, + "step": 72387 + }, + { + "epoch": 5.35, + "learning_rate": 6.11102944306744e-07, + "loss": 0.8873, + "step": 72388 + }, + { + "epoch": 5.35, + "learning_rate": 6.10965652932568e-07, + "loss": 0.9424, + "step": 72389 + }, + { + "epoch": 5.35, + "learning_rate": 6.108283764962142e-07, + "loss": 1.0246, + "step": 72390 + }, + { + "epoch": 5.35, + "learning_rate": 6.106911149978955e-07, + "loss": 1.0001, + "step": 72391 + }, + { + "epoch": 5.35, + "learning_rate": 6.105538684378365e-07, + "loss": 0.9442, + "step": 72392 + }, + { + "epoch": 5.35, + "learning_rate": 6.104166368162501e-07, + "loss": 1.0503, + "step": 72393 + }, + { + "epoch": 5.35, + "learning_rate": 6.102794201333551e-07, + "loss": 0.9856, + "step": 72394 + }, + { + "epoch": 5.35, + "learning_rate": 6.101422183893724e-07, + "loss": 0.8513, + "step": 72395 + }, + { + "epoch": 5.35, + "learning_rate": 6.100050315845197e-07, + "loss": 1.0229, + "step": 72396 + }, + { + "epoch": 5.35, + "learning_rate": 6.098678597190145e-07, + "loss": 1.0201, + "step": 72397 + }, + { + "epoch": 5.35, + "learning_rate": 6.097307027930732e-07, + "loss": 0.9967, + "step": 72398 + }, + { + "epoch": 5.35, + "learning_rate": 6.09593560806917e-07, + "loss": 1.0797, + "step": 72399 + }, + { + "epoch": 5.35, + "learning_rate": 6.094564337607644e-07, + "loss": 1.0378, + "step": 72400 + }, + { + "epoch": 5.35, + "learning_rate": 6.093193216548321e-07, + "loss": 0.9358, + "step": 72401 + }, + { + "epoch": 5.35, + "learning_rate": 6.091822244893364e-07, + "loss": 0.968, + "step": 72402 + }, + { + "epoch": 5.35, + "learning_rate": 6.090451422644993e-07, + "loss": 1.028, + "step": 72403 + }, + { + "epoch": 5.35, + "learning_rate": 6.089080749805365e-07, + "loss": 0.9839, + "step": 72404 + }, + { + "epoch": 5.35, + "learning_rate": 6.087710226376664e-07, + "loss": 0.9961, + "step": 72405 + }, + { + "epoch": 5.35, + "learning_rate": 6.086339852361078e-07, + "loss": 0.8957, + "step": 72406 + }, + { + "epoch": 5.35, + "learning_rate": 6.084969627760773e-07, + "loss": 0.8151, + "step": 72407 + }, + { + "epoch": 5.35, + "learning_rate": 6.083599552577935e-07, + "loss": 1.057, + "step": 72408 + }, + { + "epoch": 5.35, + "learning_rate": 6.082229626814761e-07, + "loss": 1.018, + "step": 72409 + }, + { + "epoch": 5.35, + "learning_rate": 6.080859850473408e-07, + "loss": 0.9734, + "step": 72410 + }, + { + "epoch": 5.35, + "learning_rate": 6.079490223556051e-07, + "loss": 1.0402, + "step": 72411 + }, + { + "epoch": 5.35, + "learning_rate": 6.078120746064919e-07, + "loss": 1.0295, + "step": 72412 + }, + { + "epoch": 5.35, + "learning_rate": 6.076751418002125e-07, + "loss": 0.9607, + "step": 72413 + }, + { + "epoch": 5.35, + "learning_rate": 6.075382239369876e-07, + "loss": 1.0435, + "step": 72414 + }, + { + "epoch": 5.35, + "learning_rate": 6.074013210170348e-07, + "loss": 0.9763, + "step": 72415 + }, + { + "epoch": 5.35, + "learning_rate": 6.07264433040573e-07, + "loss": 1.0825, + "step": 72416 + }, + { + "epoch": 5.35, + "learning_rate": 6.071275600078197e-07, + "loss": 1.0829, + "step": 72417 + }, + { + "epoch": 5.35, + "learning_rate": 6.069907019189892e-07, + "loss": 1.0412, + "step": 72418 + }, + { + "epoch": 5.35, + "learning_rate": 6.068538587743045e-07, + "loss": 0.938, + "step": 72419 + }, + { + "epoch": 5.35, + "learning_rate": 6.0671703057398e-07, + "loss": 0.9996, + "step": 72420 + }, + { + "epoch": 5.35, + "learning_rate": 6.065802173182355e-07, + "loss": 0.9381, + "step": 72421 + }, + { + "epoch": 5.35, + "learning_rate": 6.064434190072866e-07, + "loss": 1.0789, + "step": 72422 + }, + { + "epoch": 5.35, + "learning_rate": 6.063066356413505e-07, + "loss": 1.0621, + "step": 72423 + }, + { + "epoch": 5.35, + "learning_rate": 6.061698672206485e-07, + "loss": 1.023, + "step": 72424 + }, + { + "epoch": 5.35, + "learning_rate": 6.060331137453945e-07, + "loss": 0.9142, + "step": 72425 + }, + { + "epoch": 5.35, + "learning_rate": 6.058963752158087e-07, + "loss": 1.0018, + "step": 72426 + }, + { + "epoch": 5.35, + "learning_rate": 6.057596516321052e-07, + "loss": 1.0129, + "step": 72427 + }, + { + "epoch": 5.35, + "learning_rate": 6.056229429945048e-07, + "loss": 1.058, + "step": 72428 + }, + { + "epoch": 5.35, + "learning_rate": 6.054862493032243e-07, + "loss": 0.8781, + "step": 72429 + }, + { + "epoch": 5.35, + "learning_rate": 6.053495705584811e-07, + "loss": 1.0516, + "step": 72430 + }, + { + "epoch": 5.35, + "learning_rate": 6.052129067604918e-07, + "loss": 1.0232, + "step": 72431 + }, + { + "epoch": 5.35, + "learning_rate": 6.050762579094727e-07, + "loss": 1.0183, + "step": 72432 + }, + { + "epoch": 5.35, + "learning_rate": 6.049396240056471e-07, + "loss": 1.0122, + "step": 72433 + }, + { + "epoch": 5.35, + "learning_rate": 6.048030050492237e-07, + "loss": 0.9987, + "step": 72434 + }, + { + "epoch": 5.35, + "learning_rate": 6.046664010404269e-07, + "loss": 1.0983, + "step": 72435 + }, + { + "epoch": 5.35, + "learning_rate": 6.045298119794685e-07, + "loss": 1.0048, + "step": 72436 + }, + { + "epoch": 5.35, + "learning_rate": 6.043932378665728e-07, + "loss": 1.0233, + "step": 72437 + }, + { + "epoch": 5.35, + "learning_rate": 6.042566787019499e-07, + "loss": 0.9366, + "step": 72438 + }, + { + "epoch": 5.35, + "learning_rate": 6.041201344858216e-07, + "loss": 1.0845, + "step": 72439 + }, + { + "epoch": 5.35, + "learning_rate": 6.039836052184034e-07, + "loss": 0.9251, + "step": 72440 + }, + { + "epoch": 5.35, + "learning_rate": 6.038470908999139e-07, + "loss": 0.9504, + "step": 72441 + }, + { + "epoch": 5.35, + "learning_rate": 6.037105915305685e-07, + "loss": 1.0003, + "step": 72442 + }, + { + "epoch": 5.35, + "learning_rate": 6.035741071105839e-07, + "loss": 0.9847, + "step": 72443 + }, + { + "epoch": 5.35, + "learning_rate": 6.034376376401796e-07, + "loss": 1.0523, + "step": 72444 + }, + { + "epoch": 5.35, + "learning_rate": 6.033011831195712e-07, + "loss": 0.9624, + "step": 72445 + }, + { + "epoch": 5.35, + "learning_rate": 6.031647435489762e-07, + "loss": 0.9967, + "step": 72446 + }, + { + "epoch": 5.35, + "learning_rate": 6.030283189286112e-07, + "loss": 1.021, + "step": 72447 + }, + { + "epoch": 5.35, + "learning_rate": 6.028919092586949e-07, + "loss": 0.9135, + "step": 72448 + }, + { + "epoch": 5.35, + "learning_rate": 6.027555145394426e-07, + "loss": 1.0345, + "step": 72449 + }, + { + "epoch": 5.35, + "learning_rate": 6.026191347710719e-07, + "loss": 1.0632, + "step": 72450 + }, + { + "epoch": 5.35, + "learning_rate": 6.024827699537994e-07, + "loss": 1.0485, + "step": 72451 + }, + { + "epoch": 5.35, + "learning_rate": 6.023464200878415e-07, + "loss": 0.996, + "step": 72452 + }, + { + "epoch": 5.35, + "learning_rate": 6.02210085173418e-07, + "loss": 1.1083, + "step": 72453 + }, + { + "epoch": 5.35, + "learning_rate": 6.020737652107423e-07, + "loss": 0.9491, + "step": 72454 + }, + { + "epoch": 5.35, + "learning_rate": 6.019374602000339e-07, + "loss": 1.0398, + "step": 72455 + }, + { + "epoch": 5.35, + "learning_rate": 6.018011701415071e-07, + "loss": 0.9962, + "step": 72456 + }, + { + "epoch": 5.35, + "learning_rate": 6.016648950353821e-07, + "loss": 1.0545, + "step": 72457 + }, + { + "epoch": 5.35, + "learning_rate": 6.015286348818716e-07, + "loss": 1.0376, + "step": 72458 + }, + { + "epoch": 5.35, + "learning_rate": 6.013923896811958e-07, + "loss": 0.9525, + "step": 72459 + }, + { + "epoch": 5.35, + "learning_rate": 6.012561594335708e-07, + "loss": 1.0236, + "step": 72460 + }, + { + "epoch": 5.35, + "learning_rate": 6.011199441392124e-07, + "loss": 0.9479, + "step": 72461 + }, + { + "epoch": 5.35, + "learning_rate": 6.009837437983368e-07, + "loss": 1.0262, + "step": 72462 + }, + { + "epoch": 5.35, + "learning_rate": 6.008475584111617e-07, + "loss": 1.0169, + "step": 72463 + }, + { + "epoch": 5.35, + "learning_rate": 6.007113879779036e-07, + "loss": 1.0465, + "step": 72464 + }, + { + "epoch": 5.35, + "learning_rate": 6.005752324987802e-07, + "loss": 1.0026, + "step": 72465 + }, + { + "epoch": 5.35, + "learning_rate": 6.004390919740066e-07, + "loss": 1.0995, + "step": 72466 + }, + { + "epoch": 5.35, + "learning_rate": 6.003029664037996e-07, + "loss": 0.9021, + "step": 72467 + }, + { + "epoch": 5.35, + "learning_rate": 6.001668557883766e-07, + "loss": 0.943, + "step": 72468 + }, + { + "epoch": 5.35, + "learning_rate": 6.000307601279532e-07, + "loss": 0.9288, + "step": 72469 + }, + { + "epoch": 5.35, + "learning_rate": 5.998946794227467e-07, + "loss": 0.9932, + "step": 72470 + }, + { + "epoch": 5.35, + "learning_rate": 5.997586136729738e-07, + "loss": 0.9919, + "step": 72471 + }, + { + "epoch": 5.35, + "learning_rate": 5.996225628788488e-07, + "loss": 1.0414, + "step": 72472 + }, + { + "epoch": 5.35, + "learning_rate": 5.994865270405914e-07, + "loss": 0.9386, + "step": 72473 + }, + { + "epoch": 5.36, + "learning_rate": 5.99350506158416e-07, + "loss": 1.0002, + "step": 72474 + }, + { + "epoch": 5.36, + "learning_rate": 5.9921450023254e-07, + "loss": 1.0385, + "step": 72475 + }, + { + "epoch": 5.36, + "learning_rate": 5.990785092631768e-07, + "loss": 1.1024, + "step": 72476 + }, + { + "epoch": 5.36, + "learning_rate": 5.989425332505472e-07, + "loss": 0.9721, + "step": 72477 + }, + { + "epoch": 5.36, + "learning_rate": 5.988065721948677e-07, + "loss": 0.9602, + "step": 72478 + }, + { + "epoch": 5.36, + "learning_rate": 5.986706260963481e-07, + "loss": 0.9605, + "step": 72479 + }, + { + "epoch": 5.36, + "learning_rate": 5.985346949552118e-07, + "loss": 0.9397, + "step": 72480 + }, + { + "epoch": 5.36, + "learning_rate": 5.983987787716694e-07, + "loss": 0.9656, + "step": 72481 + }, + { + "epoch": 5.36, + "learning_rate": 5.982628775459443e-07, + "loss": 0.9283, + "step": 72482 + }, + { + "epoch": 5.36, + "learning_rate": 5.98126991278245e-07, + "loss": 0.9677, + "step": 72483 + }, + { + "epoch": 5.36, + "learning_rate": 5.979911199687927e-07, + "loss": 1.0496, + "step": 72484 + }, + { + "epoch": 5.36, + "learning_rate": 5.978552636178004e-07, + "loss": 1.1049, + "step": 72485 + }, + { + "epoch": 5.36, + "learning_rate": 5.977194222254889e-07, + "loss": 0.8438, + "step": 72486 + }, + { + "epoch": 5.36, + "learning_rate": 5.975835957920695e-07, + "loss": 1.0189, + "step": 72487 + }, + { + "epoch": 5.36, + "learning_rate": 5.974477843177584e-07, + "loss": 1.04, + "step": 72488 + }, + { + "epoch": 5.36, + "learning_rate": 5.973119878027755e-07, + "loss": 1.0995, + "step": 72489 + }, + { + "epoch": 5.36, + "learning_rate": 5.971762062473352e-07, + "loss": 0.9533, + "step": 72490 + }, + { + "epoch": 5.36, + "learning_rate": 5.970404396516516e-07, + "loss": 1.0303, + "step": 72491 + }, + { + "epoch": 5.36, + "learning_rate": 5.969046880159413e-07, + "loss": 0.9349, + "step": 72492 + }, + { + "epoch": 5.36, + "learning_rate": 5.967689513404229e-07, + "loss": 1.0672, + "step": 72493 + }, + { + "epoch": 5.36, + "learning_rate": 5.966332296253097e-07, + "loss": 1.0887, + "step": 72494 + }, + { + "epoch": 5.36, + "learning_rate": 5.964975228708192e-07, + "loss": 0.8656, + "step": 72495 + }, + { + "epoch": 5.36, + "learning_rate": 5.963618310771669e-07, + "loss": 0.9772, + "step": 72496 + }, + { + "epoch": 5.36, + "learning_rate": 5.962261542445669e-07, + "loss": 0.9302, + "step": 72497 + }, + { + "epoch": 5.36, + "learning_rate": 5.960904923732369e-07, + "loss": 0.9837, + "step": 72498 + }, + { + "epoch": 5.36, + "learning_rate": 5.959548454633934e-07, + "loss": 1.0901, + "step": 72499 + }, + { + "epoch": 5.36, + "learning_rate": 5.958192135152508e-07, + "loss": 1.1159, + "step": 72500 + }, + { + "epoch": 5.36, + "learning_rate": 5.956835965290242e-07, + "loss": 0.9617, + "step": 72501 + }, + { + "epoch": 5.36, + "learning_rate": 5.955479945049325e-07, + "loss": 0.9678, + "step": 72502 + }, + { + "epoch": 5.36, + "learning_rate": 5.954124074431866e-07, + "loss": 0.9948, + "step": 72503 + }, + { + "epoch": 5.36, + "learning_rate": 5.952768353440075e-07, + "loss": 1.0077, + "step": 72504 + }, + { + "epoch": 5.36, + "learning_rate": 5.951412782076072e-07, + "loss": 1.0237, + "step": 72505 + }, + { + "epoch": 5.36, + "learning_rate": 5.950057360342032e-07, + "loss": 0.9874, + "step": 72506 + }, + { + "epoch": 5.36, + "learning_rate": 5.948702088240111e-07, + "loss": 1.013, + "step": 72507 + }, + { + "epoch": 5.36, + "learning_rate": 5.947346965772438e-07, + "loss": 0.9684, + "step": 72508 + }, + { + "epoch": 5.36, + "learning_rate": 5.945991992941203e-07, + "loss": 1.0212, + "step": 72509 + }, + { + "epoch": 5.36, + "learning_rate": 5.944637169748546e-07, + "loss": 1.0385, + "step": 72510 + }, + { + "epoch": 5.36, + "learning_rate": 5.943282496196634e-07, + "loss": 1.1085, + "step": 72511 + }, + { + "epoch": 5.36, + "learning_rate": 5.941927972287598e-07, + "loss": 1.0249, + "step": 72512 + }, + { + "epoch": 5.36, + "learning_rate": 5.940573598023624e-07, + "loss": 0.9558, + "step": 72513 + }, + { + "epoch": 5.36, + "learning_rate": 5.939219373406846e-07, + "loss": 1.1468, + "step": 72514 + }, + { + "epoch": 5.36, + "learning_rate": 5.937865298439427e-07, + "loss": 0.9629, + "step": 72515 + }, + { + "epoch": 5.36, + "learning_rate": 5.936511373123521e-07, + "loss": 0.9644, + "step": 72516 + }, + { + "epoch": 5.36, + "learning_rate": 5.93515759746126e-07, + "loss": 1.0911, + "step": 72517 + }, + { + "epoch": 5.36, + "learning_rate": 5.933803971454833e-07, + "loss": 1.0846, + "step": 72518 + }, + { + "epoch": 5.36, + "learning_rate": 5.932450495106379e-07, + "loss": 0.9811, + "step": 72519 + }, + { + "epoch": 5.36, + "learning_rate": 5.931097168418055e-07, + "loss": 1.0481, + "step": 72520 + }, + { + "epoch": 5.36, + "learning_rate": 5.929743991391989e-07, + "loss": 0.8871, + "step": 72521 + }, + { + "epoch": 5.36, + "learning_rate": 5.928390964030372e-07, + "loss": 1.0433, + "step": 72522 + }, + { + "epoch": 5.36, + "learning_rate": 5.927038086335335e-07, + "loss": 1.0011, + "step": 72523 + }, + { + "epoch": 5.36, + "learning_rate": 5.925685358309041e-07, + "loss": 1.019, + "step": 72524 + }, + { + "epoch": 5.36, + "learning_rate": 5.924332779953634e-07, + "loss": 0.9828, + "step": 72525 + }, + { + "epoch": 5.36, + "learning_rate": 5.922980351271246e-07, + "loss": 1.0263, + "step": 72526 + }, + { + "epoch": 5.36, + "learning_rate": 5.921628072264085e-07, + "loss": 1.0161, + "step": 72527 + }, + { + "epoch": 5.36, + "learning_rate": 5.920275942934239e-07, + "loss": 0.9759, + "step": 72528 + }, + { + "epoch": 5.36, + "learning_rate": 5.918923963283895e-07, + "loss": 0.9463, + "step": 72529 + }, + { + "epoch": 5.36, + "learning_rate": 5.917572133315197e-07, + "loss": 1.064, + "step": 72530 + }, + { + "epoch": 5.36, + "learning_rate": 5.916220453030308e-07, + "loss": 1.137, + "step": 72531 + }, + { + "epoch": 5.36, + "learning_rate": 5.914868922431349e-07, + "loss": 1.1117, + "step": 72532 + }, + { + "epoch": 5.36, + "learning_rate": 5.913517541520497e-07, + "loss": 1.02, + "step": 72533 + }, + { + "epoch": 5.36, + "learning_rate": 5.912166310299893e-07, + "loss": 1.0735, + "step": 72534 + }, + { + "epoch": 5.36, + "learning_rate": 5.910815228771694e-07, + "loss": 1.019, + "step": 72535 + }, + { + "epoch": 5.36, + "learning_rate": 5.909464296938028e-07, + "loss": 0.9772, + "step": 72536 + }, + { + "epoch": 5.36, + "learning_rate": 5.908113514801051e-07, + "loss": 1.0921, + "step": 72537 + }, + { + "epoch": 5.36, + "learning_rate": 5.906762882362938e-07, + "loss": 0.9426, + "step": 72538 + }, + { + "epoch": 5.36, + "learning_rate": 5.905412399625809e-07, + "loss": 0.9301, + "step": 72539 + }, + { + "epoch": 5.36, + "learning_rate": 5.904062066591831e-07, + "loss": 0.9812, + "step": 72540 + }, + { + "epoch": 5.36, + "learning_rate": 5.902711883263123e-07, + "loss": 0.985, + "step": 72541 + }, + { + "epoch": 5.36, + "learning_rate": 5.901361849641874e-07, + "loss": 1.0163, + "step": 72542 + }, + { + "epoch": 5.36, + "learning_rate": 5.900011965730223e-07, + "loss": 0.9275, + "step": 72543 + }, + { + "epoch": 5.36, + "learning_rate": 5.898662231530283e-07, + "loss": 0.9347, + "step": 72544 + }, + { + "epoch": 5.36, + "learning_rate": 5.897312647044228e-07, + "loss": 1.0164, + "step": 72545 + }, + { + "epoch": 5.36, + "learning_rate": 5.895963212274202e-07, + "loss": 0.9706, + "step": 72546 + }, + { + "epoch": 5.36, + "learning_rate": 5.89461392722237e-07, + "loss": 1.0662, + "step": 72547 + }, + { + "epoch": 5.36, + "learning_rate": 5.89326479189084e-07, + "loss": 0.991, + "step": 72548 + }, + { + "epoch": 5.36, + "learning_rate": 5.89191580628179e-07, + "loss": 1.0081, + "step": 72549 + }, + { + "epoch": 5.36, + "learning_rate": 5.890566970397338e-07, + "loss": 0.9917, + "step": 72550 + }, + { + "epoch": 5.36, + "learning_rate": 5.889218284239684e-07, + "loss": 1.0394, + "step": 72551 + }, + { + "epoch": 5.36, + "learning_rate": 5.887869747810915e-07, + "loss": 1.1102, + "step": 72552 + }, + { + "epoch": 5.36, + "learning_rate": 5.886521361113184e-07, + "loss": 0.9503, + "step": 72553 + }, + { + "epoch": 5.36, + "learning_rate": 5.885173124148669e-07, + "loss": 0.8883, + "step": 72554 + }, + { + "epoch": 5.36, + "learning_rate": 5.883825036919499e-07, + "loss": 1.0565, + "step": 72555 + }, + { + "epoch": 5.36, + "learning_rate": 5.882477099427819e-07, + "loss": 1.0251, + "step": 72556 + }, + { + "epoch": 5.36, + "learning_rate": 5.881129311675748e-07, + "loss": 0.9969, + "step": 72557 + }, + { + "epoch": 5.36, + "learning_rate": 5.879781673665475e-07, + "loss": 0.9077, + "step": 72558 + }, + { + "epoch": 5.36, + "learning_rate": 5.878434185399118e-07, + "loss": 1.0215, + "step": 72559 + }, + { + "epoch": 5.36, + "learning_rate": 5.877086846878821e-07, + "loss": 1.044, + "step": 72560 + }, + { + "epoch": 5.36, + "learning_rate": 5.875739658106738e-07, + "loss": 0.9448, + "step": 72561 + }, + { + "epoch": 5.36, + "learning_rate": 5.874392619085001e-07, + "loss": 1.0199, + "step": 72562 + }, + { + "epoch": 5.36, + "learning_rate": 5.873045729815762e-07, + "loss": 0.9506, + "step": 72563 + }, + { + "epoch": 5.36, + "learning_rate": 5.871698990301167e-07, + "loss": 0.9334, + "step": 72564 + }, + { + "epoch": 5.36, + "learning_rate": 5.870352400543344e-07, + "loss": 1.0942, + "step": 72565 + }, + { + "epoch": 5.36, + "learning_rate": 5.869005960544439e-07, + "loss": 0.9948, + "step": 72566 + }, + { + "epoch": 5.36, + "learning_rate": 5.867659670306602e-07, + "loss": 0.9977, + "step": 72567 + }, + { + "epoch": 5.36, + "learning_rate": 5.86631352983198e-07, + "loss": 0.9942, + "step": 72568 + }, + { + "epoch": 5.36, + "learning_rate": 5.864967539122701e-07, + "loss": 0.9069, + "step": 72569 + }, + { + "epoch": 5.36, + "learning_rate": 5.863621698180899e-07, + "loss": 0.9403, + "step": 72570 + }, + { + "epoch": 5.36, + "learning_rate": 5.862276007008749e-07, + "loss": 1.0046, + "step": 72571 + }, + { + "epoch": 5.36, + "learning_rate": 5.860930465608384e-07, + "loss": 0.9924, + "step": 72572 + }, + { + "epoch": 5.36, + "learning_rate": 5.859585073981888e-07, + "loss": 1.0631, + "step": 72573 + }, + { + "epoch": 5.36, + "learning_rate": 5.858239832131474e-07, + "loss": 1.0199, + "step": 72574 + }, + { + "epoch": 5.36, + "learning_rate": 5.856894740059238e-07, + "loss": 0.9342, + "step": 72575 + }, + { + "epoch": 5.36, + "learning_rate": 5.855549797767357e-07, + "loss": 0.9697, + "step": 72576 + }, + { + "epoch": 5.36, + "learning_rate": 5.854205005257929e-07, + "loss": 1.1116, + "step": 72577 + }, + { + "epoch": 5.36, + "learning_rate": 5.85286036253313e-07, + "loss": 0.9666, + "step": 72578 + }, + { + "epoch": 5.36, + "learning_rate": 5.85151586959507e-07, + "loss": 1.0045, + "step": 72579 + }, + { + "epoch": 5.36, + "learning_rate": 5.850171526445914e-07, + "loss": 0.9907, + "step": 72580 + }, + { + "epoch": 5.36, + "learning_rate": 5.848827333087781e-07, + "loss": 0.937, + "step": 72581 + }, + { + "epoch": 5.36, + "learning_rate": 5.847483289522804e-07, + "loss": 0.9424, + "step": 72582 + }, + { + "epoch": 5.36, + "learning_rate": 5.846139395753148e-07, + "loss": 1.0362, + "step": 72583 + }, + { + "epoch": 5.36, + "learning_rate": 5.844795651780944e-07, + "loss": 0.988, + "step": 72584 + }, + { + "epoch": 5.36, + "learning_rate": 5.843452057608312e-07, + "loss": 1.1005, + "step": 72585 + }, + { + "epoch": 5.36, + "learning_rate": 5.842108613237385e-07, + "loss": 0.9917, + "step": 72586 + }, + { + "epoch": 5.36, + "learning_rate": 5.840765318670339e-07, + "loss": 0.963, + "step": 72587 + }, + { + "epoch": 5.36, + "learning_rate": 5.839422173909293e-07, + "loss": 1.0271, + "step": 72588 + }, + { + "epoch": 5.36, + "learning_rate": 5.838079178956368e-07, + "loss": 1.0802, + "step": 72589 + }, + { + "epoch": 5.36, + "learning_rate": 5.836736333813719e-07, + "loss": 1.0529, + "step": 72590 + }, + { + "epoch": 5.36, + "learning_rate": 5.835393638483455e-07, + "loss": 1.0866, + "step": 72591 + }, + { + "epoch": 5.36, + "learning_rate": 5.834051092967763e-07, + "loss": 1.0605, + "step": 72592 + }, + { + "epoch": 5.36, + "learning_rate": 5.832708697268719e-07, + "loss": 1.0513, + "step": 72593 + }, + { + "epoch": 5.36, + "learning_rate": 5.831366451388498e-07, + "loss": 0.9526, + "step": 72594 + }, + { + "epoch": 5.36, + "learning_rate": 5.830024355329222e-07, + "loss": 1.1749, + "step": 72595 + }, + { + "epoch": 5.36, + "learning_rate": 5.828682409093056e-07, + "loss": 0.9494, + "step": 72596 + }, + { + "epoch": 5.36, + "learning_rate": 5.827340612682075e-07, + "loss": 0.9988, + "step": 72597 + }, + { + "epoch": 5.36, + "learning_rate": 5.825998966098468e-07, + "loss": 0.8881, + "step": 72598 + }, + { + "epoch": 5.36, + "learning_rate": 5.824657469344341e-07, + "loss": 0.901, + "step": 72599 + }, + { + "epoch": 5.36, + "learning_rate": 5.82331612242184e-07, + "loss": 0.9874, + "step": 72600 + }, + { + "epoch": 5.36, + "learning_rate": 5.821974925333096e-07, + "loss": 0.9738, + "step": 72601 + }, + { + "epoch": 5.36, + "learning_rate": 5.820633878080228e-07, + "loss": 1.1141, + "step": 72602 + }, + { + "epoch": 5.36, + "learning_rate": 5.819292980665403e-07, + "loss": 0.9891, + "step": 72603 + }, + { + "epoch": 5.36, + "learning_rate": 5.81795223309074e-07, + "loss": 1.0532, + "step": 72604 + }, + { + "epoch": 5.36, + "learning_rate": 5.816611635358349e-07, + "loss": 0.9839, + "step": 72605 + }, + { + "epoch": 5.36, + "learning_rate": 5.815271187470383e-07, + "loss": 0.9496, + "step": 72606 + }, + { + "epoch": 5.36, + "learning_rate": 5.813930889428976e-07, + "loss": 1.0568, + "step": 72607 + }, + { + "epoch": 5.36, + "learning_rate": 5.812590741236268e-07, + "loss": 1.0912, + "step": 72608 + }, + { + "epoch": 5.37, + "learning_rate": 5.811250742894381e-07, + "loss": 1.0213, + "step": 72609 + }, + { + "epoch": 5.37, + "learning_rate": 5.809910894405435e-07, + "loss": 0.9208, + "step": 72610 + }, + { + "epoch": 5.37, + "learning_rate": 5.808571195771562e-07, + "loss": 0.8821, + "step": 72611 + }, + { + "epoch": 5.37, + "learning_rate": 5.807231646994926e-07, + "loss": 0.9861, + "step": 72612 + }, + { + "epoch": 5.37, + "learning_rate": 5.805892248077638e-07, + "loss": 1.0085, + "step": 72613 + }, + { + "epoch": 5.37, + "learning_rate": 5.804552999021817e-07, + "loss": 1.0687, + "step": 72614 + }, + { + "epoch": 5.37, + "learning_rate": 5.803213899829597e-07, + "loss": 0.9734, + "step": 72615 + }, + { + "epoch": 5.37, + "learning_rate": 5.801874950503128e-07, + "loss": 1.0084, + "step": 72616 + }, + { + "epoch": 5.37, + "learning_rate": 5.800536151044545e-07, + "loss": 1.0052, + "step": 72617 + }, + { + "epoch": 5.37, + "learning_rate": 5.799197501455933e-07, + "loss": 0.9085, + "step": 72618 + }, + { + "epoch": 5.37, + "learning_rate": 5.797859001739459e-07, + "loss": 0.9136, + "step": 72619 + }, + { + "epoch": 5.37, + "learning_rate": 5.796520651897231e-07, + "loss": 1.0166, + "step": 72620 + }, + { + "epoch": 5.37, + "learning_rate": 5.795182451931425e-07, + "loss": 0.9014, + "step": 72621 + }, + { + "epoch": 5.37, + "learning_rate": 5.793844401844096e-07, + "loss": 0.9561, + "step": 72622 + }, + { + "epoch": 5.37, + "learning_rate": 5.79250650163744e-07, + "loss": 1.0852, + "step": 72623 + }, + { + "epoch": 5.37, + "learning_rate": 5.791168751313547e-07, + "loss": 1.0265, + "step": 72624 + }, + { + "epoch": 5.37, + "learning_rate": 5.789831150874569e-07, + "loss": 1.0297, + "step": 72625 + }, + { + "epoch": 5.37, + "learning_rate": 5.788493700322595e-07, + "loss": 1.0739, + "step": 72626 + }, + { + "epoch": 5.37, + "learning_rate": 5.787156399659799e-07, + "loss": 0.9908, + "step": 72627 + }, + { + "epoch": 5.37, + "learning_rate": 5.785819248888292e-07, + "loss": 1.045, + "step": 72628 + }, + { + "epoch": 5.37, + "learning_rate": 5.784482248010192e-07, + "loss": 1.0976, + "step": 72629 + }, + { + "epoch": 5.37, + "learning_rate": 5.783145397027645e-07, + "loss": 0.9869, + "step": 72630 + }, + { + "epoch": 5.37, + "learning_rate": 5.781808695942736e-07, + "loss": 0.9166, + "step": 72631 + }, + { + "epoch": 5.37, + "learning_rate": 5.780472144757643e-07, + "loss": 1.0019, + "step": 72632 + }, + { + "epoch": 5.37, + "learning_rate": 5.779135743474473e-07, + "loss": 1.0935, + "step": 72633 + }, + { + "epoch": 5.37, + "learning_rate": 5.777799492095349e-07, + "loss": 0.992, + "step": 72634 + }, + { + "epoch": 5.37, + "learning_rate": 5.776463390622378e-07, + "loss": 0.9564, + "step": 72635 + }, + { + "epoch": 5.37, + "learning_rate": 5.775127439057726e-07, + "loss": 1.1177, + "step": 72636 + }, + { + "epoch": 5.37, + "learning_rate": 5.773791637403515e-07, + "loss": 1.0811, + "step": 72637 + }, + { + "epoch": 5.37, + "learning_rate": 5.772455985661818e-07, + "loss": 0.9588, + "step": 72638 + }, + { + "epoch": 5.37, + "learning_rate": 5.771120483834814e-07, + "loss": 0.8534, + "step": 72639 + }, + { + "epoch": 5.37, + "learning_rate": 5.769785131924588e-07, + "loss": 1.0295, + "step": 72640 + }, + { + "epoch": 5.37, + "learning_rate": 5.768449929933317e-07, + "loss": 1.0045, + "step": 72641 + }, + { + "epoch": 5.37, + "learning_rate": 5.767114877863067e-07, + "loss": 1.0185, + "step": 72642 + }, + { + "epoch": 5.37, + "learning_rate": 5.765779975716002e-07, + "loss": 1.0122, + "step": 72643 + }, + { + "epoch": 5.37, + "learning_rate": 5.76444522349422e-07, + "loss": 1.1181, + "step": 72644 + }, + { + "epoch": 5.37, + "learning_rate": 5.763110621199886e-07, + "loss": 0.8956, + "step": 72645 + }, + { + "epoch": 5.37, + "learning_rate": 5.761776168835087e-07, + "loss": 1.092, + "step": 72646 + }, + { + "epoch": 5.37, + "learning_rate": 5.760441866401934e-07, + "loss": 1.0584, + "step": 72647 + }, + { + "epoch": 5.37, + "learning_rate": 5.759107713902579e-07, + "loss": 1.0201, + "step": 72648 + }, + { + "epoch": 5.37, + "learning_rate": 5.757773711339142e-07, + "loss": 1.0478, + "step": 72649 + }, + { + "epoch": 5.37, + "learning_rate": 5.756439858713736e-07, + "loss": 0.9809, + "step": 72650 + }, + { + "epoch": 5.37, + "learning_rate": 5.755106156028467e-07, + "loss": 0.9648, + "step": 72651 + }, + { + "epoch": 5.37, + "learning_rate": 5.753772603285491e-07, + "loss": 1.0447, + "step": 72652 + }, + { + "epoch": 5.37, + "learning_rate": 5.752439200486915e-07, + "loss": 1.1031, + "step": 72653 + }, + { + "epoch": 5.37, + "learning_rate": 5.751105947634872e-07, + "loss": 0.9538, + "step": 72654 + }, + { + "epoch": 5.37, + "learning_rate": 5.74977284473146e-07, + "loss": 1.0304, + "step": 72655 + }, + { + "epoch": 5.37, + "learning_rate": 5.74843989177879e-07, + "loss": 0.9401, + "step": 72656 + }, + { + "epoch": 5.37, + "learning_rate": 5.747107088779025e-07, + "loss": 0.9225, + "step": 72657 + }, + { + "epoch": 5.37, + "learning_rate": 5.745774435734275e-07, + "loss": 1.0534, + "step": 72658 + }, + { + "epoch": 5.37, + "learning_rate": 5.744441932646639e-07, + "loss": 1.0576, + "step": 72659 + }, + { + "epoch": 5.37, + "learning_rate": 5.743109579518225e-07, + "loss": 1.0743, + "step": 72660 + }, + { + "epoch": 5.37, + "learning_rate": 5.741777376351199e-07, + "loss": 1.021, + "step": 72661 + }, + { + "epoch": 5.37, + "learning_rate": 5.740445323147659e-07, + "loss": 0.9481, + "step": 72662 + }, + { + "epoch": 5.37, + "learning_rate": 5.739113419909715e-07, + "loss": 0.9128, + "step": 72663 + }, + { + "epoch": 5.37, + "learning_rate": 5.737781666639496e-07, + "loss": 0.9321, + "step": 72664 + }, + { + "epoch": 5.37, + "learning_rate": 5.736450063339116e-07, + "loss": 0.9701, + "step": 72665 + }, + { + "epoch": 5.37, + "learning_rate": 5.735118610010704e-07, + "loss": 1.0431, + "step": 72666 + }, + { + "epoch": 5.37, + "learning_rate": 5.733787306656346e-07, + "loss": 1.0493, + "step": 72667 + }, + { + "epoch": 5.37, + "learning_rate": 5.732456153278199e-07, + "loss": 0.889, + "step": 72668 + }, + { + "epoch": 5.37, + "learning_rate": 5.73112514987837e-07, + "loss": 0.9161, + "step": 72669 + }, + { + "epoch": 5.37, + "learning_rate": 5.729794296458968e-07, + "loss": 0.9141, + "step": 72670 + }, + { + "epoch": 5.37, + "learning_rate": 5.728463593022094e-07, + "loss": 0.975, + "step": 72671 + }, + { + "epoch": 5.37, + "learning_rate": 5.72713303956991e-07, + "loss": 1.0406, + "step": 72672 + }, + { + "epoch": 5.37, + "learning_rate": 5.725802636104505e-07, + "loss": 1.0518, + "step": 72673 + }, + { + "epoch": 5.37, + "learning_rate": 5.724472382627999e-07, + "loss": 0.9959, + "step": 72674 + }, + { + "epoch": 5.37, + "learning_rate": 5.723142279142513e-07, + "loss": 0.9155, + "step": 72675 + }, + { + "epoch": 5.37, + "learning_rate": 5.721812325650133e-07, + "loss": 1.0372, + "step": 72676 + }, + { + "epoch": 5.37, + "learning_rate": 5.720482522153025e-07, + "loss": 0.9531, + "step": 72677 + }, + { + "epoch": 5.37, + "learning_rate": 5.719152868653277e-07, + "loss": 1.0099, + "step": 72678 + }, + { + "epoch": 5.37, + "learning_rate": 5.717823365153008e-07, + "loss": 1.0279, + "step": 72679 + }, + { + "epoch": 5.37, + "learning_rate": 5.716494011654317e-07, + "loss": 0.9848, + "step": 72680 + }, + { + "epoch": 5.37, + "learning_rate": 5.715164808159357e-07, + "loss": 1.116, + "step": 72681 + }, + { + "epoch": 5.37, + "learning_rate": 5.713835754670216e-07, + "loss": 0.9812, + "step": 72682 + }, + { + "epoch": 5.37, + "learning_rate": 5.712506851189015e-07, + "loss": 0.9293, + "step": 72683 + }, + { + "epoch": 5.37, + "learning_rate": 5.711178097717862e-07, + "loss": 1.0217, + "step": 72684 + }, + { + "epoch": 5.37, + "learning_rate": 5.709849494258857e-07, + "loss": 1.0158, + "step": 72685 + }, + { + "epoch": 5.37, + "learning_rate": 5.708521040814175e-07, + "loss": 0.999, + "step": 72686 + }, + { + "epoch": 5.37, + "learning_rate": 5.707192737385847e-07, + "loss": 0.9586, + "step": 72687 + }, + { + "epoch": 5.37, + "learning_rate": 5.705864583976051e-07, + "loss": 0.9582, + "step": 72688 + }, + { + "epoch": 5.37, + "learning_rate": 5.70453658058685e-07, + "loss": 1.0284, + "step": 72689 + }, + { + "epoch": 5.37, + "learning_rate": 5.703208727220422e-07, + "loss": 1.1189, + "step": 72690 + }, + { + "epoch": 5.37, + "learning_rate": 5.701881023878797e-07, + "loss": 1.0435, + "step": 72691 + }, + { + "epoch": 5.37, + "learning_rate": 5.700553470564152e-07, + "loss": 0.9361, + "step": 72692 + }, + { + "epoch": 5.37, + "learning_rate": 5.699226067278574e-07, + "loss": 1.0844, + "step": 72693 + }, + { + "epoch": 5.37, + "learning_rate": 5.697898814024183e-07, + "loss": 0.8928, + "step": 72694 + }, + { + "epoch": 5.37, + "learning_rate": 5.69657171080309e-07, + "loss": 0.9572, + "step": 72695 + }, + { + "epoch": 5.37, + "learning_rate": 5.69524475761738e-07, + "loss": 0.972, + "step": 72696 + }, + { + "epoch": 5.37, + "learning_rate": 5.693917954469197e-07, + "loss": 0.9514, + "step": 72697 + }, + { + "epoch": 5.37, + "learning_rate": 5.69259130136065e-07, + "loss": 1.0546, + "step": 72698 + }, + { + "epoch": 5.37, + "learning_rate": 5.691264798293839e-07, + "loss": 1.0564, + "step": 72699 + }, + { + "epoch": 5.37, + "learning_rate": 5.689938445270859e-07, + "loss": 0.8574, + "step": 72700 + }, + { + "epoch": 5.37, + "learning_rate": 5.688612242293856e-07, + "loss": 1.084, + "step": 72701 + }, + { + "epoch": 5.37, + "learning_rate": 5.687286189364916e-07, + "loss": 0.9655, + "step": 72702 + }, + { + "epoch": 5.37, + "learning_rate": 5.685960286486158e-07, + "loss": 0.9793, + "step": 72703 + }, + { + "epoch": 5.37, + "learning_rate": 5.684634533659694e-07, + "loss": 0.9972, + "step": 72704 + }, + { + "epoch": 5.37, + "learning_rate": 5.683308930887599e-07, + "loss": 0.9361, + "step": 72705 + }, + { + "epoch": 5.37, + "learning_rate": 5.681983478172037e-07, + "loss": 0.9979, + "step": 72706 + }, + { + "epoch": 5.37, + "learning_rate": 5.680658175515085e-07, + "loss": 0.9649, + "step": 72707 + }, + { + "epoch": 5.37, + "learning_rate": 5.679333022918853e-07, + "loss": 1.0043, + "step": 72708 + }, + { + "epoch": 5.37, + "learning_rate": 5.67800802038544e-07, + "loss": 0.9722, + "step": 72709 + }, + { + "epoch": 5.37, + "learning_rate": 5.676683167916996e-07, + "loss": 0.9915, + "step": 72710 + }, + { + "epoch": 5.37, + "learning_rate": 5.675358465515579e-07, + "loss": 1.0288, + "step": 72711 + }, + { + "epoch": 5.37, + "learning_rate": 5.674033913183319e-07, + "loss": 1.0445, + "step": 72712 + }, + { + "epoch": 5.37, + "learning_rate": 5.672709510922325e-07, + "loss": 0.9473, + "step": 72713 + }, + { + "epoch": 5.37, + "learning_rate": 5.671385258734696e-07, + "loss": 1.151, + "step": 72714 + }, + { + "epoch": 5.37, + "learning_rate": 5.670061156622552e-07, + "loss": 1.0508, + "step": 72715 + }, + { + "epoch": 5.37, + "learning_rate": 5.668737204587971e-07, + "loss": 0.9246, + "step": 72716 + }, + { + "epoch": 5.37, + "learning_rate": 5.667413402633092e-07, + "loss": 0.9592, + "step": 72717 + }, + { + "epoch": 5.37, + "learning_rate": 5.666089750760018e-07, + "loss": 0.9826, + "step": 72718 + }, + { + "epoch": 5.37, + "learning_rate": 5.664766248970843e-07, + "loss": 0.9972, + "step": 72719 + }, + { + "epoch": 5.37, + "learning_rate": 5.663442897267668e-07, + "loss": 1.0003, + "step": 72720 + }, + { + "epoch": 5.37, + "learning_rate": 5.662119695652601e-07, + "loss": 0.901, + "step": 72721 + }, + { + "epoch": 5.37, + "learning_rate": 5.660796644127764e-07, + "loss": 0.9637, + "step": 72722 + }, + { + "epoch": 5.37, + "learning_rate": 5.659473742695243e-07, + "loss": 0.9247, + "step": 72723 + }, + { + "epoch": 5.37, + "learning_rate": 5.658150991357148e-07, + "loss": 0.891, + "step": 72724 + }, + { + "epoch": 5.37, + "learning_rate": 5.656828390115576e-07, + "loss": 1.0842, + "step": 72725 + }, + { + "epoch": 5.37, + "learning_rate": 5.655505938972661e-07, + "loss": 0.9702, + "step": 72726 + }, + { + "epoch": 5.37, + "learning_rate": 5.654183637930478e-07, + "loss": 0.9823, + "step": 72727 + }, + { + "epoch": 5.37, + "learning_rate": 5.652861486991146e-07, + "loss": 0.9894, + "step": 72728 + }, + { + "epoch": 5.37, + "learning_rate": 5.651539486156732e-07, + "loss": 1.1004, + "step": 72729 + }, + { + "epoch": 5.37, + "learning_rate": 5.6502176354294e-07, + "loss": 1.0302, + "step": 72730 + }, + { + "epoch": 5.37, + "learning_rate": 5.648895934811227e-07, + "loss": 1.0306, + "step": 72731 + }, + { + "epoch": 5.37, + "learning_rate": 5.647574384304289e-07, + "loss": 1.0622, + "step": 72732 + }, + { + "epoch": 5.37, + "learning_rate": 5.646252983910717e-07, + "loss": 0.9366, + "step": 72733 + }, + { + "epoch": 5.37, + "learning_rate": 5.644931733632597e-07, + "loss": 1.0037, + "step": 72734 + }, + { + "epoch": 5.37, + "learning_rate": 5.643610633472063e-07, + "loss": 1.0252, + "step": 72735 + }, + { + "epoch": 5.37, + "learning_rate": 5.642289683431168e-07, + "loss": 1.1049, + "step": 72736 + }, + { + "epoch": 5.37, + "learning_rate": 5.640968883512054e-07, + "loss": 1.0746, + "step": 72737 + }, + { + "epoch": 5.37, + "learning_rate": 5.639648233716799e-07, + "loss": 0.949, + "step": 72738 + }, + { + "epoch": 5.37, + "learning_rate": 5.638327734047544e-07, + "loss": 1.1275, + "step": 72739 + }, + { + "epoch": 5.37, + "learning_rate": 5.637007384506344e-07, + "loss": 1.0647, + "step": 72740 + }, + { + "epoch": 5.37, + "learning_rate": 5.635687185095295e-07, + "loss": 1.0108, + "step": 72741 + }, + { + "epoch": 5.37, + "learning_rate": 5.634367135816532e-07, + "loss": 1.0141, + "step": 72742 + }, + { + "epoch": 5.37, + "learning_rate": 5.63304723667214e-07, + "loss": 0.9925, + "step": 72743 + }, + { + "epoch": 5.38, + "learning_rate": 5.631727487664229e-07, + "loss": 0.9516, + "step": 72744 + }, + { + "epoch": 5.38, + "learning_rate": 5.630407888794865e-07, + "loss": 1.0616, + "step": 72745 + }, + { + "epoch": 5.38, + "learning_rate": 5.629088440066189e-07, + "loss": 0.9925, + "step": 72746 + }, + { + "epoch": 5.38, + "learning_rate": 5.627769141480288e-07, + "loss": 0.9749, + "step": 72747 + }, + { + "epoch": 5.38, + "learning_rate": 5.626449993039252e-07, + "loss": 0.9346, + "step": 72748 + }, + { + "epoch": 5.38, + "learning_rate": 5.625130994745188e-07, + "loss": 0.9348, + "step": 72749 + }, + { + "epoch": 5.38, + "learning_rate": 5.623812146600161e-07, + "loss": 0.9191, + "step": 72750 + }, + { + "epoch": 5.38, + "learning_rate": 5.622493448606325e-07, + "loss": 0.8737, + "step": 72751 + }, + { + "epoch": 5.38, + "learning_rate": 5.621174900765758e-07, + "loss": 1.0194, + "step": 72752 + }, + { + "epoch": 5.38, + "learning_rate": 5.619856503080546e-07, + "loss": 0.9796, + "step": 72753 + }, + { + "epoch": 5.38, + "learning_rate": 5.618538255552775e-07, + "loss": 1.0076, + "step": 72754 + }, + { + "epoch": 5.38, + "learning_rate": 5.617220158184589e-07, + "loss": 0.9908, + "step": 72755 + }, + { + "epoch": 5.38, + "learning_rate": 5.61590221097802e-07, + "loss": 0.9437, + "step": 72756 + }, + { + "epoch": 5.38, + "learning_rate": 5.61458441393522e-07, + "loss": 1.076, + "step": 72757 + }, + { + "epoch": 5.38, + "learning_rate": 5.613266767058268e-07, + "loss": 0.9086, + "step": 72758 + }, + { + "epoch": 5.38, + "learning_rate": 5.61194927034926e-07, + "loss": 0.9665, + "step": 72759 + }, + { + "epoch": 5.38, + "learning_rate": 5.610631923810284e-07, + "loss": 0.9554, + "step": 72760 + }, + { + "epoch": 5.38, + "learning_rate": 5.609314727443427e-07, + "loss": 0.9723, + "step": 72761 + }, + { + "epoch": 5.38, + "learning_rate": 5.607997681250821e-07, + "loss": 1.0065, + "step": 72762 + }, + { + "epoch": 5.38, + "learning_rate": 5.606680785234542e-07, + "loss": 0.9258, + "step": 72763 + }, + { + "epoch": 5.38, + "learning_rate": 5.605364039396676e-07, + "loss": 1.1223, + "step": 72764 + }, + { + "epoch": 5.38, + "learning_rate": 5.604047443739313e-07, + "loss": 1.0546, + "step": 72765 + }, + { + "epoch": 5.38, + "learning_rate": 5.602730998264571e-07, + "loss": 0.9733, + "step": 72766 + }, + { + "epoch": 5.38, + "learning_rate": 5.60141470297454e-07, + "loss": 0.9389, + "step": 72767 + }, + { + "epoch": 5.38, + "learning_rate": 5.600098557871302e-07, + "loss": 1.1025, + "step": 72768 + }, + { + "epoch": 5.38, + "learning_rate": 5.598782562956962e-07, + "loss": 1.0352, + "step": 72769 + }, + { + "epoch": 5.38, + "learning_rate": 5.59746671823359e-07, + "loss": 0.9742, + "step": 72770 + }, + { + "epoch": 5.38, + "learning_rate": 5.596151023703311e-07, + "loss": 0.9001, + "step": 72771 + }, + { + "epoch": 5.38, + "learning_rate": 5.59483547936821e-07, + "loss": 1.11, + "step": 72772 + }, + { + "epoch": 5.38, + "learning_rate": 5.593520085230364e-07, + "loss": 0.8971, + "step": 72773 + }, + { + "epoch": 5.38, + "learning_rate": 5.592204841291871e-07, + "loss": 1.0431, + "step": 72774 + }, + { + "epoch": 5.38, + "learning_rate": 5.590889747554851e-07, + "loss": 1.128, + "step": 72775 + }, + { + "epoch": 5.38, + "learning_rate": 5.589574804021381e-07, + "loss": 1.0614, + "step": 72776 + }, + { + "epoch": 5.38, + "learning_rate": 5.588260010693514e-07, + "loss": 1.0032, + "step": 72777 + }, + { + "epoch": 5.38, + "learning_rate": 5.586945367573393e-07, + "loss": 1.0925, + "step": 72778 + }, + { + "epoch": 5.38, + "learning_rate": 5.585630874663073e-07, + "loss": 0.95, + "step": 72779 + }, + { + "epoch": 5.38, + "learning_rate": 5.584316531964696e-07, + "loss": 0.9489, + "step": 72780 + }, + { + "epoch": 5.38, + "learning_rate": 5.583002339480293e-07, + "loss": 0.9515, + "step": 72781 + }, + { + "epoch": 5.38, + "learning_rate": 5.581688297211995e-07, + "loss": 1.0585, + "step": 72782 + }, + { + "epoch": 5.38, + "learning_rate": 5.58037440516187e-07, + "loss": 1.0386, + "step": 72783 + }, + { + "epoch": 5.38, + "learning_rate": 5.579060663332037e-07, + "loss": 1.1417, + "step": 72784 + }, + { + "epoch": 5.38, + "learning_rate": 5.577747071724549e-07, + "loss": 0.9632, + "step": 72785 + }, + { + "epoch": 5.38, + "learning_rate": 5.576433630341527e-07, + "loss": 0.9235, + "step": 72786 + }, + { + "epoch": 5.38, + "learning_rate": 5.575120339185047e-07, + "loss": 1.0629, + "step": 72787 + }, + { + "epoch": 5.38, + "learning_rate": 5.573807198257197e-07, + "loss": 0.938, + "step": 72788 + }, + { + "epoch": 5.38, + "learning_rate": 5.572494207560064e-07, + "loss": 0.9426, + "step": 72789 + }, + { + "epoch": 5.38, + "learning_rate": 5.571181367095734e-07, + "loss": 1.0189, + "step": 72790 + }, + { + "epoch": 5.38, + "learning_rate": 5.569868676866319e-07, + "loss": 0.99, + "step": 72791 + }, + { + "epoch": 5.38, + "learning_rate": 5.568556136873893e-07, + "loss": 1.0026, + "step": 72792 + }, + { + "epoch": 5.38, + "learning_rate": 5.567243747120543e-07, + "loss": 0.9696, + "step": 72793 + }, + { + "epoch": 5.38, + "learning_rate": 5.565931507608335e-07, + "loss": 0.9784, + "step": 72794 + }, + { + "epoch": 5.38, + "learning_rate": 5.564619418339401e-07, + "loss": 0.937, + "step": 72795 + }, + { + "epoch": 5.38, + "learning_rate": 5.563307479315805e-07, + "loss": 1.0085, + "step": 72796 + }, + { + "epoch": 5.38, + "learning_rate": 5.561995690539634e-07, + "loss": 0.9948, + "step": 72797 + }, + { + "epoch": 5.38, + "learning_rate": 5.560684052012977e-07, + "loss": 0.9754, + "step": 72798 + }, + { + "epoch": 5.38, + "learning_rate": 5.559372563737909e-07, + "loss": 1.0991, + "step": 72799 + }, + { + "epoch": 5.38, + "learning_rate": 5.558061225716549e-07, + "loss": 0.9593, + "step": 72800 + }, + { + "epoch": 5.38, + "learning_rate": 5.556750037950943e-07, + "loss": 0.9501, + "step": 72801 + }, + { + "epoch": 5.38, + "learning_rate": 5.555439000443196e-07, + "loss": 1.096, + "step": 72802 + }, + { + "epoch": 5.38, + "learning_rate": 5.554128113195389e-07, + "loss": 0.8522, + "step": 72803 + }, + { + "epoch": 5.38, + "learning_rate": 5.552817376209652e-07, + "loss": 0.9828, + "step": 72804 + }, + { + "epoch": 5.38, + "learning_rate": 5.551506789488004e-07, + "loss": 0.9822, + "step": 72805 + }, + { + "epoch": 5.38, + "learning_rate": 5.550196353032545e-07, + "loss": 1.0857, + "step": 72806 + }, + { + "epoch": 5.38, + "learning_rate": 5.548886066845394e-07, + "loss": 1.008, + "step": 72807 + }, + { + "epoch": 5.38, + "learning_rate": 5.547575930928606e-07, + "loss": 1.0243, + "step": 72808 + }, + { + "epoch": 5.38, + "learning_rate": 5.546265945284269e-07, + "loss": 0.9744, + "step": 72809 + }, + { + "epoch": 5.38, + "learning_rate": 5.544956109914468e-07, + "loss": 0.9442, + "step": 72810 + }, + { + "epoch": 5.38, + "learning_rate": 5.543646424821303e-07, + "loss": 0.9918, + "step": 72811 + }, + { + "epoch": 5.38, + "learning_rate": 5.542336890006838e-07, + "loss": 0.9123, + "step": 72812 + }, + { + "epoch": 5.38, + "learning_rate": 5.541027505473174e-07, + "loss": 1.0139, + "step": 72813 + }, + { + "epoch": 5.38, + "learning_rate": 5.539718271222372e-07, + "loss": 1.056, + "step": 72814 + }, + { + "epoch": 5.38, + "learning_rate": 5.538409187256522e-07, + "loss": 0.9873, + "step": 72815 + }, + { + "epoch": 5.38, + "learning_rate": 5.53710025357772e-07, + "loss": 1.0355, + "step": 72816 + }, + { + "epoch": 5.38, + "learning_rate": 5.535791470188034e-07, + "loss": 0.9519, + "step": 72817 + }, + { + "epoch": 5.38, + "learning_rate": 5.534482837089561e-07, + "loss": 0.9779, + "step": 72818 + }, + { + "epoch": 5.38, + "learning_rate": 5.533174354284366e-07, + "loss": 1.0487, + "step": 72819 + }, + { + "epoch": 5.38, + "learning_rate": 5.531866021774545e-07, + "loss": 1.0011, + "step": 72820 + }, + { + "epoch": 5.38, + "learning_rate": 5.530557839562167e-07, + "loss": 1.0173, + "step": 72821 + }, + { + "epoch": 5.38, + "learning_rate": 5.529249807649328e-07, + "loss": 0.9892, + "step": 72822 + }, + { + "epoch": 5.38, + "learning_rate": 5.527941926038105e-07, + "loss": 0.9201, + "step": 72823 + }, + { + "epoch": 5.38, + "learning_rate": 5.526634194730551e-07, + "loss": 1.0437, + "step": 72824 + }, + { + "epoch": 5.38, + "learning_rate": 5.52532661372881e-07, + "loss": 1.0397, + "step": 72825 + }, + { + "epoch": 5.38, + "learning_rate": 5.52401918303489e-07, + "loss": 0.9779, + "step": 72826 + }, + { + "epoch": 5.38, + "learning_rate": 5.522711902650912e-07, + "loss": 1.0643, + "step": 72827 + }, + { + "epoch": 5.38, + "learning_rate": 5.521404772578942e-07, + "loss": 0.9272, + "step": 72828 + }, + { + "epoch": 5.38, + "learning_rate": 5.520097792821088e-07, + "loss": 1.0847, + "step": 72829 + }, + { + "epoch": 5.38, + "learning_rate": 5.518790963379373e-07, + "loss": 1.0619, + "step": 72830 + }, + { + "epoch": 5.38, + "learning_rate": 5.517484284255936e-07, + "loss": 0.9409, + "step": 72831 + }, + { + "epoch": 5.38, + "learning_rate": 5.516177755452823e-07, + "loss": 0.9328, + "step": 72832 + }, + { + "epoch": 5.38, + "learning_rate": 5.514871376972119e-07, + "loss": 1.0515, + "step": 72833 + }, + { + "epoch": 5.38, + "learning_rate": 5.513565148815902e-07, + "loss": 1.055, + "step": 72834 + }, + { + "epoch": 5.38, + "learning_rate": 5.512259070986237e-07, + "loss": 0.9575, + "step": 72835 + }, + { + "epoch": 5.38, + "learning_rate": 5.510953143485232e-07, + "loss": 1.1033, + "step": 72836 + }, + { + "epoch": 5.38, + "learning_rate": 5.509647366314952e-07, + "loss": 0.9369, + "step": 72837 + }, + { + "epoch": 5.38, + "learning_rate": 5.508341739477463e-07, + "loss": 1.0275, + "step": 72838 + }, + { + "epoch": 5.38, + "learning_rate": 5.507036262974841e-07, + "loss": 0.9455, + "step": 72839 + }, + { + "epoch": 5.38, + "learning_rate": 5.505730936809183e-07, + "loss": 1.0165, + "step": 72840 + }, + { + "epoch": 5.38, + "learning_rate": 5.504425760982568e-07, + "loss": 0.9343, + "step": 72841 + }, + { + "epoch": 5.38, + "learning_rate": 5.503120735497047e-07, + "loss": 1.1313, + "step": 72842 + }, + { + "epoch": 5.38, + "learning_rate": 5.501815860354709e-07, + "loss": 1.0771, + "step": 72843 + }, + { + "epoch": 5.38, + "learning_rate": 5.500511135557618e-07, + "loss": 1.0516, + "step": 72844 + }, + { + "epoch": 5.38, + "learning_rate": 5.499206561107895e-07, + "loss": 0.9769, + "step": 72845 + }, + { + "epoch": 5.38, + "learning_rate": 5.497902137007549e-07, + "loss": 1.0678, + "step": 72846 + }, + { + "epoch": 5.38, + "learning_rate": 5.496597863258701e-07, + "loss": 0.9104, + "step": 72847 + }, + { + "epoch": 5.38, + "learning_rate": 5.495293739863406e-07, + "loss": 1.021, + "step": 72848 + }, + { + "epoch": 5.38, + "learning_rate": 5.49398976682377e-07, + "loss": 1.0129, + "step": 72849 + }, + { + "epoch": 5.38, + "learning_rate": 5.492685944141818e-07, + "loss": 1.0374, + "step": 72850 + }, + { + "epoch": 5.38, + "learning_rate": 5.491382271819657e-07, + "loss": 1.0477, + "step": 72851 + }, + { + "epoch": 5.38, + "learning_rate": 5.490078749859362e-07, + "loss": 1.0289, + "step": 72852 + }, + { + "epoch": 5.38, + "learning_rate": 5.488775378263e-07, + "loss": 1.0176, + "step": 72853 + }, + { + "epoch": 5.38, + "learning_rate": 5.487472157032636e-07, + "loss": 1.0236, + "step": 72854 + }, + { + "epoch": 5.38, + "learning_rate": 5.486169086170345e-07, + "loss": 0.9644, + "step": 72855 + }, + { + "epoch": 5.38, + "learning_rate": 5.484866165678227e-07, + "loss": 0.9748, + "step": 72856 + }, + { + "epoch": 5.38, + "learning_rate": 5.483563395558323e-07, + "loss": 0.9934, + "step": 72857 + }, + { + "epoch": 5.38, + "learning_rate": 5.48226077581272e-07, + "loss": 1.1043, + "step": 72858 + }, + { + "epoch": 5.38, + "learning_rate": 5.480958306443474e-07, + "loss": 1.0077, + "step": 72859 + }, + { + "epoch": 5.38, + "learning_rate": 5.479655987452693e-07, + "loss": 0.9782, + "step": 72860 + }, + { + "epoch": 5.38, + "learning_rate": 5.478353818842419e-07, + "loss": 0.9782, + "step": 72861 + }, + { + "epoch": 5.38, + "learning_rate": 5.47705180061473e-07, + "loss": 0.9509, + "step": 72862 + }, + { + "epoch": 5.38, + "learning_rate": 5.475749932771701e-07, + "loss": 1.0491, + "step": 72863 + }, + { + "epoch": 5.38, + "learning_rate": 5.474448215315387e-07, + "loss": 0.9591, + "step": 72864 + }, + { + "epoch": 5.38, + "learning_rate": 5.473146648247895e-07, + "loss": 0.9112, + "step": 72865 + }, + { + "epoch": 5.38, + "learning_rate": 5.47184523157126e-07, + "loss": 1.0006, + "step": 72866 + }, + { + "epoch": 5.38, + "learning_rate": 5.470543965287578e-07, + "loss": 0.9639, + "step": 72867 + }, + { + "epoch": 5.38, + "learning_rate": 5.469242849398893e-07, + "loss": 1.0413, + "step": 72868 + }, + { + "epoch": 5.38, + "learning_rate": 5.467941883907301e-07, + "loss": 1.0243, + "step": 72869 + }, + { + "epoch": 5.38, + "learning_rate": 5.466641068814881e-07, + "loss": 0.9862, + "step": 72870 + }, + { + "epoch": 5.38, + "learning_rate": 5.465340404123642e-07, + "loss": 1.0172, + "step": 72871 + }, + { + "epoch": 5.38, + "learning_rate": 5.464039889835715e-07, + "loss": 0.9544, + "step": 72872 + }, + { + "epoch": 5.38, + "learning_rate": 5.462739525953143e-07, + "loss": 0.9427, + "step": 72873 + }, + { + "epoch": 5.38, + "learning_rate": 5.461439312478011e-07, + "loss": 1.0026, + "step": 72874 + }, + { + "epoch": 5.38, + "learning_rate": 5.460139249412356e-07, + "loss": 0.9913, + "step": 72875 + }, + { + "epoch": 5.38, + "learning_rate": 5.458839336758282e-07, + "loss": 1.1072, + "step": 72876 + }, + { + "epoch": 5.38, + "learning_rate": 5.457539574517845e-07, + "loss": 1.0102, + "step": 72877 + }, + { + "epoch": 5.38, + "learning_rate": 5.456239962693111e-07, + "loss": 0.8967, + "step": 72878 + }, + { + "epoch": 5.38, + "learning_rate": 5.454940501286132e-07, + "loss": 1.0284, + "step": 72879 + }, + { + "epoch": 5.39, + "learning_rate": 5.453641190298986e-07, + "loss": 1.0086, + "step": 72880 + }, + { + "epoch": 5.39, + "learning_rate": 5.452342029733759e-07, + "loss": 1.0357, + "step": 72881 + }, + { + "epoch": 5.39, + "learning_rate": 5.451043019592505e-07, + "loss": 1.0088, + "step": 72882 + }, + { + "epoch": 5.39, + "learning_rate": 5.449744159877291e-07, + "loss": 0.9374, + "step": 72883 + }, + { + "epoch": 5.39, + "learning_rate": 5.448445450590156e-07, + "loss": 1.0572, + "step": 72884 + }, + { + "epoch": 5.39, + "learning_rate": 5.447146891733212e-07, + "loss": 0.9738, + "step": 72885 + }, + { + "epoch": 5.39, + "learning_rate": 5.445848483308503e-07, + "loss": 1.0395, + "step": 72886 + }, + { + "epoch": 5.39, + "learning_rate": 5.444550225318101e-07, + "loss": 0.8594, + "step": 72887 + }, + { + "epoch": 5.39, + "learning_rate": 5.443252117764042e-07, + "loss": 0.9669, + "step": 72888 + }, + { + "epoch": 5.39, + "learning_rate": 5.441954160648444e-07, + "loss": 0.9569, + "step": 72889 + }, + { + "epoch": 5.39, + "learning_rate": 5.440656353973351e-07, + "loss": 0.8082, + "step": 72890 + }, + { + "epoch": 5.39, + "learning_rate": 5.439358697740794e-07, + "loss": 1.0658, + "step": 72891 + }, + { + "epoch": 5.39, + "learning_rate": 5.43806119195287e-07, + "loss": 0.9781, + "step": 72892 + }, + { + "epoch": 5.39, + "learning_rate": 5.436763836611636e-07, + "loss": 0.9049, + "step": 72893 + }, + { + "epoch": 5.39, + "learning_rate": 5.435466631719178e-07, + "loss": 1.0566, + "step": 72894 + }, + { + "epoch": 5.39, + "learning_rate": 5.434169577277515e-07, + "loss": 0.9892, + "step": 72895 + }, + { + "epoch": 5.39, + "learning_rate": 5.432872673288748e-07, + "loss": 1.0173, + "step": 72896 + }, + { + "epoch": 5.39, + "learning_rate": 5.431575919754917e-07, + "loss": 0.9541, + "step": 72897 + }, + { + "epoch": 5.39, + "learning_rate": 5.430279316678122e-07, + "loss": 1.0162, + "step": 72898 + }, + { + "epoch": 5.39, + "learning_rate": 5.428982864060383e-07, + "loss": 0.9864, + "step": 72899 + }, + { + "epoch": 5.39, + "learning_rate": 5.427686561903766e-07, + "loss": 0.892, + "step": 72900 + }, + { + "epoch": 5.39, + "learning_rate": 5.426390410210358e-07, + "loss": 0.9793, + "step": 72901 + }, + { + "epoch": 5.39, + "learning_rate": 5.425094408982212e-07, + "loss": 0.9786, + "step": 72902 + }, + { + "epoch": 5.39, + "learning_rate": 5.423798558221383e-07, + "loss": 0.9786, + "step": 72903 + }, + { + "epoch": 5.39, + "learning_rate": 5.422502857929935e-07, + "loss": 1.0741, + "step": 72904 + }, + { + "epoch": 5.39, + "learning_rate": 5.421207308109933e-07, + "loss": 1.0583, + "step": 72905 + }, + { + "epoch": 5.39, + "learning_rate": 5.419911908763442e-07, + "loss": 0.9756, + "step": 72906 + }, + { + "epoch": 5.39, + "learning_rate": 5.418616659892529e-07, + "loss": 1.067, + "step": 72907 + }, + { + "epoch": 5.39, + "learning_rate": 5.417321561499234e-07, + "loss": 1.0892, + "step": 72908 + }, + { + "epoch": 5.39, + "learning_rate": 5.416026613585612e-07, + "loss": 0.8996, + "step": 72909 + }, + { + "epoch": 5.39, + "learning_rate": 5.414731816153751e-07, + "loss": 1.0596, + "step": 72910 + }, + { + "epoch": 5.39, + "learning_rate": 5.413437169205704e-07, + "loss": 0.8835, + "step": 72911 + }, + { + "epoch": 5.39, + "learning_rate": 5.412142672743514e-07, + "loss": 0.9528, + "step": 72912 + }, + { + "epoch": 5.39, + "learning_rate": 5.410848326769247e-07, + "loss": 0.9484, + "step": 72913 + }, + { + "epoch": 5.39, + "learning_rate": 5.409554131284978e-07, + "loss": 0.9664, + "step": 72914 + }, + { + "epoch": 5.39, + "learning_rate": 5.40826008629276e-07, + "loss": 0.9785, + "step": 72915 + }, + { + "epoch": 5.39, + "learning_rate": 5.406966191794638e-07, + "loss": 1.1202, + "step": 72916 + }, + { + "epoch": 5.39, + "learning_rate": 5.405672447792687e-07, + "loss": 0.8715, + "step": 72917 + }, + { + "epoch": 5.39, + "learning_rate": 5.404378854288939e-07, + "loss": 0.9402, + "step": 72918 + }, + { + "epoch": 5.39, + "learning_rate": 5.403085411285503e-07, + "loss": 1.0821, + "step": 72919 + }, + { + "epoch": 5.39, + "learning_rate": 5.401792118784377e-07, + "loss": 1.078, + "step": 72920 + }, + { + "epoch": 5.39, + "learning_rate": 5.40049897678766e-07, + "loss": 1.0837, + "step": 72921 + }, + { + "epoch": 5.39, + "learning_rate": 5.399205985297396e-07, + "loss": 0.949, + "step": 72922 + }, + { + "epoch": 5.39, + "learning_rate": 5.397913144315637e-07, + "loss": 1.0428, + "step": 72923 + }, + { + "epoch": 5.39, + "learning_rate": 5.396620453844436e-07, + "loss": 0.8383, + "step": 72924 + }, + { + "epoch": 5.39, + "learning_rate": 5.395327913885861e-07, + "loss": 0.976, + "step": 72925 + }, + { + "epoch": 5.39, + "learning_rate": 5.394035524441976e-07, + "loss": 1.0268, + "step": 72926 + }, + { + "epoch": 5.39, + "learning_rate": 5.392743285514823e-07, + "loss": 0.9095, + "step": 72927 + }, + { + "epoch": 5.39, + "learning_rate": 5.391451197106468e-07, + "loss": 0.945, + "step": 72928 + }, + { + "epoch": 5.39, + "learning_rate": 5.390159259218941e-07, + "loss": 1.0082, + "step": 72929 + }, + { + "epoch": 5.39, + "learning_rate": 5.388867471854331e-07, + "loss": 0.9698, + "step": 72930 + }, + { + "epoch": 5.39, + "learning_rate": 5.38757583501468e-07, + "loss": 0.9539, + "step": 72931 + }, + { + "epoch": 5.39, + "learning_rate": 5.386284348702054e-07, + "loss": 0.9194, + "step": 72932 + }, + { + "epoch": 5.39, + "learning_rate": 5.384993012918461e-07, + "loss": 0.9375, + "step": 72933 + }, + { + "epoch": 5.39, + "learning_rate": 5.383701827666021e-07, + "loss": 0.985, + "step": 72934 + }, + { + "epoch": 5.39, + "learning_rate": 5.382410792946757e-07, + "loss": 1.0649, + "step": 72935 + }, + { + "epoch": 5.39, + "learning_rate": 5.381119908762711e-07, + "loss": 1.0653, + "step": 72936 + }, + { + "epoch": 5.39, + "learning_rate": 5.379829175115958e-07, + "loss": 0.9844, + "step": 72937 + }, + { + "epoch": 5.39, + "learning_rate": 5.378538592008519e-07, + "loss": 0.9805, + "step": 72938 + }, + { + "epoch": 5.39, + "learning_rate": 5.377248159442505e-07, + "loss": 0.9424, + "step": 72939 + }, + { + "epoch": 5.39, + "learning_rate": 5.375957877419912e-07, + "loss": 1.0843, + "step": 72940 + }, + { + "epoch": 5.39, + "learning_rate": 5.37466774594283e-07, + "loss": 1.0856, + "step": 72941 + }, + { + "epoch": 5.39, + "learning_rate": 5.373377765013277e-07, + "loss": 1.029, + "step": 72942 + }, + { + "epoch": 5.39, + "learning_rate": 5.372087934633363e-07, + "loss": 0.9667, + "step": 72943 + }, + { + "epoch": 5.39, + "learning_rate": 5.370798254805065e-07, + "loss": 1.0722, + "step": 72944 + }, + { + "epoch": 5.39, + "learning_rate": 5.369508725530504e-07, + "loss": 1.0217, + "step": 72945 + }, + { + "epoch": 5.39, + "learning_rate": 5.368219346811687e-07, + "loss": 0.9739, + "step": 72946 + }, + { + "epoch": 5.39, + "learning_rate": 5.366930118650681e-07, + "loss": 1.1163, + "step": 72947 + }, + { + "epoch": 5.39, + "learning_rate": 5.36564104104953e-07, + "loss": 1.0021, + "step": 72948 + }, + { + "epoch": 5.39, + "learning_rate": 5.364352114010285e-07, + "loss": 0.9686, + "step": 72949 + }, + { + "epoch": 5.39, + "learning_rate": 5.363063337535013e-07, + "loss": 0.979, + "step": 72950 + }, + { + "epoch": 5.39, + "learning_rate": 5.361774711625755e-07, + "loss": 1.0733, + "step": 72951 + }, + { + "epoch": 5.39, + "learning_rate": 5.360486236284557e-07, + "loss": 0.9349, + "step": 72952 + }, + { + "epoch": 5.39, + "learning_rate": 5.359197911513459e-07, + "loss": 1.0114, + "step": 72953 + }, + { + "epoch": 5.39, + "learning_rate": 5.357909737314526e-07, + "loss": 0.9768, + "step": 72954 + }, + { + "epoch": 5.39, + "learning_rate": 5.356621713689813e-07, + "loss": 1.0258, + "step": 72955 + }, + { + "epoch": 5.39, + "learning_rate": 5.355333840641364e-07, + "loss": 0.9022, + "step": 72956 + }, + { + "epoch": 5.39, + "learning_rate": 5.35404611817123e-07, + "loss": 1.1061, + "step": 72957 + }, + { + "epoch": 5.39, + "learning_rate": 5.352758546281434e-07, + "loss": 1.0101, + "step": 72958 + }, + { + "epoch": 5.39, + "learning_rate": 5.35147112497405e-07, + "loss": 1.0057, + "step": 72959 + }, + { + "epoch": 5.39, + "learning_rate": 5.350183854251134e-07, + "loss": 0.9795, + "step": 72960 + }, + { + "epoch": 5.39, + "learning_rate": 5.348896734114716e-07, + "loss": 1.0097, + "step": 72961 + }, + { + "epoch": 5.39, + "learning_rate": 5.347609764566831e-07, + "loss": 0.9726, + "step": 72962 + }, + { + "epoch": 5.39, + "learning_rate": 5.346322945609561e-07, + "loss": 1.0072, + "step": 72963 + }, + { + "epoch": 5.39, + "learning_rate": 5.345036277244963e-07, + "loss": 1.0747, + "step": 72964 + }, + { + "epoch": 5.39, + "learning_rate": 5.343749759475014e-07, + "loss": 0.9429, + "step": 72965 + }, + { + "epoch": 5.39, + "learning_rate": 5.342463392301822e-07, + "loss": 1.0078, + "step": 72966 + }, + { + "epoch": 5.39, + "learning_rate": 5.341177175727419e-07, + "loss": 1.0349, + "step": 72967 + }, + { + "epoch": 5.39, + "learning_rate": 5.339891109753858e-07, + "loss": 0.9599, + "step": 72968 + }, + { + "epoch": 5.39, + "learning_rate": 5.338605194383151e-07, + "loss": 0.8703, + "step": 72969 + }, + { + "epoch": 5.39, + "learning_rate": 5.337319429617383e-07, + "loss": 0.9349, + "step": 72970 + }, + { + "epoch": 5.39, + "learning_rate": 5.336033815458586e-07, + "loss": 1.025, + "step": 72971 + }, + { + "epoch": 5.39, + "learning_rate": 5.334748351908803e-07, + "loss": 1.1231, + "step": 72972 + }, + { + "epoch": 5.39, + "learning_rate": 5.333463038970088e-07, + "loss": 1.0008, + "step": 72973 + }, + { + "epoch": 5.39, + "learning_rate": 5.332177876644462e-07, + "loss": 1.097, + "step": 72974 + }, + { + "epoch": 5.39, + "learning_rate": 5.330892864934001e-07, + "loss": 0.9762, + "step": 72975 + }, + { + "epoch": 5.39, + "learning_rate": 5.329608003840725e-07, + "loss": 0.9374, + "step": 72976 + }, + { + "epoch": 5.39, + "learning_rate": 5.328323293366699e-07, + "loss": 0.9813, + "step": 72977 + }, + { + "epoch": 5.39, + "learning_rate": 5.327038733513945e-07, + "loss": 0.8694, + "step": 72978 + }, + { + "epoch": 5.39, + "learning_rate": 5.325754324284527e-07, + "loss": 0.9401, + "step": 72979 + }, + { + "epoch": 5.39, + "learning_rate": 5.324470065680476e-07, + "loss": 0.9391, + "step": 72980 + }, + { + "epoch": 5.39, + "learning_rate": 5.323185957703835e-07, + "loss": 0.9835, + "step": 72981 + }, + { + "epoch": 5.39, + "learning_rate": 5.32190200035666e-07, + "loss": 0.8808, + "step": 72982 + }, + { + "epoch": 5.39, + "learning_rate": 5.32061819364097e-07, + "loss": 1.0758, + "step": 72983 + }, + { + "epoch": 5.39, + "learning_rate": 5.31933453755884e-07, + "loss": 1.0174, + "step": 72984 + }, + { + "epoch": 5.39, + "learning_rate": 5.318051032112271e-07, + "loss": 0.9903, + "step": 72985 + }, + { + "epoch": 5.39, + "learning_rate": 5.316767677303347e-07, + "loss": 1.0324, + "step": 72986 + }, + { + "epoch": 5.39, + "learning_rate": 5.31548447313408e-07, + "loss": 1.1308, + "step": 72987 + }, + { + "epoch": 5.39, + "learning_rate": 5.314201419606535e-07, + "loss": 1.018, + "step": 72988 + }, + { + "epoch": 5.39, + "learning_rate": 5.312918516722732e-07, + "loss": 1.0596, + "step": 72989 + }, + { + "epoch": 5.39, + "learning_rate": 5.311635764484724e-07, + "loss": 0.9253, + "step": 72990 + }, + { + "epoch": 5.39, + "learning_rate": 5.310353162894544e-07, + "loss": 0.9712, + "step": 72991 + }, + { + "epoch": 5.39, + "learning_rate": 5.309070711954245e-07, + "loss": 1.0178, + "step": 72992 + }, + { + "epoch": 5.39, + "learning_rate": 5.307788411665859e-07, + "loss": 0.9945, + "step": 72993 + }, + { + "epoch": 5.39, + "learning_rate": 5.306506262031407e-07, + "loss": 1.0038, + "step": 72994 + }, + { + "epoch": 5.39, + "learning_rate": 5.305224263052966e-07, + "loss": 1.0021, + "step": 72995 + }, + { + "epoch": 5.39, + "learning_rate": 5.303942414732566e-07, + "loss": 1.0612, + "step": 72996 + }, + { + "epoch": 5.39, + "learning_rate": 5.302660717072227e-07, + "loss": 1.0505, + "step": 72997 + }, + { + "epoch": 5.39, + "learning_rate": 5.301379170073995e-07, + "loss": 1.017, + "step": 72998 + }, + { + "epoch": 5.39, + "learning_rate": 5.30009777373992e-07, + "loss": 1.1218, + "step": 72999 + }, + { + "epoch": 5.39, + "learning_rate": 5.298816528072037e-07, + "loss": 1.0373, + "step": 73000 + }, + { + "epoch": 5.39, + "learning_rate": 5.297535433072387e-07, + "loss": 0.9497, + "step": 73001 + }, + { + "epoch": 5.39, + "learning_rate": 5.296254488743003e-07, + "loss": 1.1148, + "step": 73002 + }, + { + "epoch": 5.39, + "learning_rate": 5.294973695085903e-07, + "loss": 1.0918, + "step": 73003 + }, + { + "epoch": 5.39, + "learning_rate": 5.293693052103166e-07, + "loss": 1.0234, + "step": 73004 + }, + { + "epoch": 5.39, + "learning_rate": 5.2924125597968e-07, + "loss": 1.0498, + "step": 73005 + }, + { + "epoch": 5.39, + "learning_rate": 5.291132218168859e-07, + "loss": 1.0636, + "step": 73006 + }, + { + "epoch": 5.39, + "learning_rate": 5.289852027221353e-07, + "loss": 1.0662, + "step": 73007 + }, + { + "epoch": 5.39, + "learning_rate": 5.288571986956359e-07, + "loss": 0.9387, + "step": 73008 + }, + { + "epoch": 5.39, + "learning_rate": 5.287292097375874e-07, + "loss": 0.919, + "step": 73009 + }, + { + "epoch": 5.39, + "learning_rate": 5.286012358481962e-07, + "loss": 0.9567, + "step": 73010 + }, + { + "epoch": 5.39, + "learning_rate": 5.284732770276646e-07, + "loss": 1.068, + "step": 73011 + }, + { + "epoch": 5.39, + "learning_rate": 5.283453332761968e-07, + "loss": 1.0054, + "step": 73012 + }, + { + "epoch": 5.39, + "learning_rate": 5.28217404593997e-07, + "loss": 1.0316, + "step": 73013 + }, + { + "epoch": 5.39, + "learning_rate": 5.280894909812651e-07, + "loss": 0.9905, + "step": 73014 + }, + { + "epoch": 5.4, + "learning_rate": 5.279615924382087e-07, + "loss": 0.9952, + "step": 73015 + }, + { + "epoch": 5.4, + "learning_rate": 5.278337089650309e-07, + "loss": 1.0292, + "step": 73016 + }, + { + "epoch": 5.4, + "learning_rate": 5.27705840561934e-07, + "loss": 0.9358, + "step": 73017 + }, + { + "epoch": 5.4, + "learning_rate": 5.275779872291187e-07, + "loss": 0.9277, + "step": 73018 + }, + { + "epoch": 5.4, + "learning_rate": 5.274501489667938e-07, + "loss": 0.974, + "step": 73019 + }, + { + "epoch": 5.4, + "learning_rate": 5.273223257751603e-07, + "loss": 1.0179, + "step": 73020 + }, + { + "epoch": 5.4, + "learning_rate": 5.271945176544213e-07, + "loss": 1.0279, + "step": 73021 + }, + { + "epoch": 5.4, + "learning_rate": 5.2706672460478e-07, + "loss": 1.0437, + "step": 73022 + }, + { + "epoch": 5.4, + "learning_rate": 5.269389466264396e-07, + "loss": 1.0245, + "step": 73023 + }, + { + "epoch": 5.4, + "learning_rate": 5.268111837196044e-07, + "loss": 1.0492, + "step": 73024 + }, + { + "epoch": 5.4, + "learning_rate": 5.266834358844764e-07, + "loss": 1.0257, + "step": 73025 + }, + { + "epoch": 5.4, + "learning_rate": 5.265557031212609e-07, + "loss": 0.9965, + "step": 73026 + }, + { + "epoch": 5.4, + "learning_rate": 5.264279854301568e-07, + "loss": 0.9897, + "step": 73027 + }, + { + "epoch": 5.4, + "learning_rate": 5.263002828113728e-07, + "loss": 1.021, + "step": 73028 + }, + { + "epoch": 5.4, + "learning_rate": 5.261725952651109e-07, + "loss": 1.0335, + "step": 73029 + }, + { + "epoch": 5.4, + "learning_rate": 5.260449227915698e-07, + "loss": 1.0818, + "step": 73030 + }, + { + "epoch": 5.4, + "learning_rate": 5.259172653909572e-07, + "loss": 0.9451, + "step": 73031 + }, + { + "epoch": 5.4, + "learning_rate": 5.257896230634729e-07, + "loss": 0.9621, + "step": 73032 + }, + { + "epoch": 5.4, + "learning_rate": 5.256619958093245e-07, + "loss": 0.9803, + "step": 73033 + }, + { + "epoch": 5.4, + "learning_rate": 5.255343836287097e-07, + "loss": 1.0351, + "step": 73034 + }, + { + "epoch": 5.4, + "learning_rate": 5.25406786521836e-07, + "loss": 1.0552, + "step": 73035 + }, + { + "epoch": 5.4, + "learning_rate": 5.252792044889021e-07, + "loss": 1.0229, + "step": 73036 + }, + { + "epoch": 5.4, + "learning_rate": 5.25151637530118e-07, + "loss": 1.0454, + "step": 73037 + }, + { + "epoch": 5.4, + "learning_rate": 5.25024085645679e-07, + "loss": 1.0028, + "step": 73038 + }, + { + "epoch": 5.4, + "learning_rate": 5.248965488357904e-07, + "loss": 1.0514, + "step": 73039 + }, + { + "epoch": 5.4, + "learning_rate": 5.247690271006567e-07, + "loss": 0.852, + "step": 73040 + }, + { + "epoch": 5.4, + "learning_rate": 5.246415204404809e-07, + "loss": 0.8959, + "step": 73041 + }, + { + "epoch": 5.4, + "learning_rate": 5.24514028855464e-07, + "loss": 0.8953, + "step": 73042 + }, + { + "epoch": 5.4, + "learning_rate": 5.24386552345808e-07, + "loss": 0.9704, + "step": 73043 + }, + { + "epoch": 5.4, + "learning_rate": 5.242590909117195e-07, + "loss": 0.9353, + "step": 73044 + }, + { + "epoch": 5.4, + "learning_rate": 5.241316445533995e-07, + "loss": 0.9645, + "step": 73045 + }, + { + "epoch": 5.4, + "learning_rate": 5.240042132710499e-07, + "loss": 1.0707, + "step": 73046 + }, + { + "epoch": 5.4, + "learning_rate": 5.238767970648751e-07, + "loss": 0.9905, + "step": 73047 + }, + { + "epoch": 5.4, + "learning_rate": 5.237493959350736e-07, + "loss": 1.0165, + "step": 73048 + }, + { + "epoch": 5.4, + "learning_rate": 5.236220098818534e-07, + "loss": 1.1058, + "step": 73049 + }, + { + "epoch": 5.4, + "learning_rate": 5.234946389054152e-07, + "loss": 1.0373, + "step": 73050 + }, + { + "epoch": 5.4, + "learning_rate": 5.23367283005961e-07, + "loss": 0.9628, + "step": 73051 + }, + { + "epoch": 5.4, + "learning_rate": 5.232399421836931e-07, + "loss": 1.0892, + "step": 73052 + }, + { + "epoch": 5.4, + "learning_rate": 5.231126164388178e-07, + "loss": 0.9486, + "step": 73053 + }, + { + "epoch": 5.4, + "learning_rate": 5.229853057715317e-07, + "loss": 0.9313, + "step": 73054 + }, + { + "epoch": 5.4, + "learning_rate": 5.228580101820413e-07, + "loss": 0.9642, + "step": 73055 + }, + { + "epoch": 5.4, + "learning_rate": 5.227307296705464e-07, + "loss": 1.0125, + "step": 73056 + }, + { + "epoch": 5.4, + "learning_rate": 5.226034642372557e-07, + "loss": 0.9615, + "step": 73057 + }, + { + "epoch": 5.4, + "learning_rate": 5.224762138823647e-07, + "loss": 0.9017, + "step": 73058 + }, + { + "epoch": 5.4, + "learning_rate": 5.223489786060776e-07, + "loss": 0.9093, + "step": 73059 + }, + { + "epoch": 5.4, + "learning_rate": 5.222217584085986e-07, + "loss": 1.0349, + "step": 73060 + }, + { + "epoch": 5.4, + "learning_rate": 5.220945532901289e-07, + "loss": 1.0353, + "step": 73061 + }, + { + "epoch": 5.4, + "learning_rate": 5.219673632508715e-07, + "loss": 1.1139, + "step": 73062 + }, + { + "epoch": 5.4, + "learning_rate": 5.218401882910273e-07, + "loss": 0.9978, + "step": 73063 + }, + { + "epoch": 5.4, + "learning_rate": 5.217130284108008e-07, + "loss": 1.0632, + "step": 73064 + }, + { + "epoch": 5.4, + "learning_rate": 5.215858836103938e-07, + "loss": 0.9646, + "step": 73065 + }, + { + "epoch": 5.4, + "learning_rate": 5.214587538900074e-07, + "loss": 0.9652, + "step": 73066 + }, + { + "epoch": 5.4, + "learning_rate": 5.213316392498446e-07, + "loss": 1.0163, + "step": 73067 + }, + { + "epoch": 5.4, + "learning_rate": 5.212045396901067e-07, + "loss": 1.0551, + "step": 73068 + }, + { + "epoch": 5.4, + "learning_rate": 5.210774552109976e-07, + "loss": 0.9725, + "step": 73069 + }, + { + "epoch": 5.4, + "learning_rate": 5.209503858127196e-07, + "loss": 1.0535, + "step": 73070 + }, + { + "epoch": 5.4, + "learning_rate": 5.208233314954724e-07, + "loss": 0.9651, + "step": 73071 + }, + { + "epoch": 5.4, + "learning_rate": 5.206962922594594e-07, + "loss": 1.0018, + "step": 73072 + }, + { + "epoch": 5.4, + "learning_rate": 5.205692681048846e-07, + "loss": 0.9967, + "step": 73073 + }, + { + "epoch": 5.4, + "learning_rate": 5.204422590319469e-07, + "loss": 1.0393, + "step": 73074 + }, + { + "epoch": 5.4, + "learning_rate": 5.203152650408516e-07, + "loss": 1.0305, + "step": 73075 + }, + { + "epoch": 5.4, + "learning_rate": 5.201882861317986e-07, + "loss": 0.9691, + "step": 73076 + }, + { + "epoch": 5.4, + "learning_rate": 5.200613223049888e-07, + "loss": 1.0248, + "step": 73077 + }, + { + "epoch": 5.4, + "learning_rate": 5.199343735606277e-07, + "loss": 1.01, + "step": 73078 + }, + { + "epoch": 5.4, + "learning_rate": 5.198074398989139e-07, + "loss": 0.9615, + "step": 73079 + }, + { + "epoch": 5.4, + "learning_rate": 5.196805213200518e-07, + "loss": 1.0794, + "step": 73080 + }, + { + "epoch": 5.4, + "learning_rate": 5.19553617824241e-07, + "loss": 1.0419, + "step": 73081 + }, + { + "epoch": 5.4, + "learning_rate": 5.194267294116873e-07, + "loss": 0.9532, + "step": 73082 + }, + { + "epoch": 5.4, + "learning_rate": 5.192998560825879e-07, + "loss": 1.0606, + "step": 73083 + }, + { + "epoch": 5.4, + "learning_rate": 5.191729978371474e-07, + "loss": 0.8683, + "step": 73084 + }, + { + "epoch": 5.4, + "learning_rate": 5.190461546755676e-07, + "loss": 0.9463, + "step": 73085 + }, + { + "epoch": 5.4, + "learning_rate": 5.189193265980486e-07, + "loss": 0.9971, + "step": 73086 + }, + { + "epoch": 5.4, + "learning_rate": 5.187925136047944e-07, + "loss": 1.1087, + "step": 73087 + }, + { + "epoch": 5.4, + "learning_rate": 5.18665715696004e-07, + "loss": 0.9522, + "step": 73088 + }, + { + "epoch": 5.4, + "learning_rate": 5.185389328718826e-07, + "loss": 0.9549, + "step": 73089 + }, + { + "epoch": 5.4, + "learning_rate": 5.18412165132629e-07, + "loss": 0.9534, + "step": 73090 + }, + { + "epoch": 5.4, + "learning_rate": 5.182854124784475e-07, + "loss": 1.0241, + "step": 73091 + }, + { + "epoch": 5.4, + "learning_rate": 5.181586749095357e-07, + "loss": 1.0352, + "step": 73092 + }, + { + "epoch": 5.4, + "learning_rate": 5.18031952426099e-07, + "loss": 1.0284, + "step": 73093 + }, + { + "epoch": 5.4, + "learning_rate": 5.179052450283384e-07, + "loss": 0.9465, + "step": 73094 + }, + { + "epoch": 5.4, + "learning_rate": 5.177785527164547e-07, + "loss": 1.1249, + "step": 73095 + }, + { + "epoch": 5.4, + "learning_rate": 5.176518754906501e-07, + "loss": 0.89, + "step": 73096 + }, + { + "epoch": 5.4, + "learning_rate": 5.175252133511232e-07, + "loss": 1.017, + "step": 73097 + }, + { + "epoch": 5.4, + "learning_rate": 5.173985662980818e-07, + "loss": 1.0318, + "step": 73098 + }, + { + "epoch": 5.4, + "learning_rate": 5.172719343317211e-07, + "loss": 1.0012, + "step": 73099 + }, + { + "epoch": 5.4, + "learning_rate": 5.171453174522457e-07, + "loss": 0.9163, + "step": 73100 + }, + { + "epoch": 5.4, + "learning_rate": 5.17018715659855e-07, + "loss": 1.0632, + "step": 73101 + }, + { + "epoch": 5.4, + "learning_rate": 5.168921289547557e-07, + "loss": 0.9426, + "step": 73102 + }, + { + "epoch": 5.4, + "learning_rate": 5.167655573371411e-07, + "loss": 0.918, + "step": 73103 + }, + { + "epoch": 5.4, + "learning_rate": 5.166390008072187e-07, + "loss": 0.9656, + "step": 73104 + }, + { + "epoch": 5.4, + "learning_rate": 5.165124593651882e-07, + "loss": 1.0643, + "step": 73105 + }, + { + "epoch": 5.4, + "learning_rate": 5.163859330112509e-07, + "loss": 0.9793, + "step": 73106 + }, + { + "epoch": 5.4, + "learning_rate": 5.162594217456074e-07, + "loss": 0.9189, + "step": 73107 + }, + { + "epoch": 5.4, + "learning_rate": 5.161329255684589e-07, + "loss": 1.0488, + "step": 73108 + }, + { + "epoch": 5.4, + "learning_rate": 5.160064444800072e-07, + "loss": 1.0983, + "step": 73109 + }, + { + "epoch": 5.4, + "learning_rate": 5.158799784804547e-07, + "loss": 0.9437, + "step": 73110 + }, + { + "epoch": 5.4, + "learning_rate": 5.15753527570001e-07, + "loss": 0.9649, + "step": 73111 + }, + { + "epoch": 5.4, + "learning_rate": 5.15627091748846e-07, + "loss": 0.9541, + "step": 73112 + }, + { + "epoch": 5.4, + "learning_rate": 5.155006710171951e-07, + "loss": 0.9911, + "step": 73113 + }, + { + "epoch": 5.4, + "learning_rate": 5.153742653752459e-07, + "loss": 1.0451, + "step": 73114 + }, + { + "epoch": 5.4, + "learning_rate": 5.152478748232014e-07, + "loss": 0.9375, + "step": 73115 + }, + { + "epoch": 5.4, + "learning_rate": 5.151214993612607e-07, + "loss": 1.0604, + "step": 73116 + }, + { + "epoch": 5.4, + "learning_rate": 5.149951389896257e-07, + "loss": 0.9978, + "step": 73117 + }, + { + "epoch": 5.4, + "learning_rate": 5.148687937084984e-07, + "loss": 1.0356, + "step": 73118 + }, + { + "epoch": 5.4, + "learning_rate": 5.147424635180798e-07, + "loss": 0.8918, + "step": 73119 + }, + { + "epoch": 5.4, + "learning_rate": 5.146161484185696e-07, + "loss": 0.9597, + "step": 73120 + }, + { + "epoch": 5.4, + "learning_rate": 5.144898484101679e-07, + "loss": 0.9708, + "step": 73121 + }, + { + "epoch": 5.4, + "learning_rate": 5.143635634930788e-07, + "loss": 0.9452, + "step": 73122 + }, + { + "epoch": 5.4, + "learning_rate": 5.142372936675033e-07, + "loss": 0.9404, + "step": 73123 + }, + { + "epoch": 5.4, + "learning_rate": 5.141110389336379e-07, + "loss": 0.9937, + "step": 73124 + }, + { + "epoch": 5.4, + "learning_rate": 5.139847992916869e-07, + "loss": 1.0666, + "step": 73125 + }, + { + "epoch": 5.4, + "learning_rate": 5.13858574741849e-07, + "loss": 1.0909, + "step": 73126 + }, + { + "epoch": 5.4, + "learning_rate": 5.137323652843296e-07, + "loss": 0.9761, + "step": 73127 + }, + { + "epoch": 5.4, + "learning_rate": 5.13606170919323e-07, + "loss": 0.8734, + "step": 73128 + }, + { + "epoch": 5.4, + "learning_rate": 5.134799916470346e-07, + "loss": 0.9336, + "step": 73129 + }, + { + "epoch": 5.4, + "learning_rate": 5.133538274676642e-07, + "loss": 0.9462, + "step": 73130 + }, + { + "epoch": 5.4, + "learning_rate": 5.132276783814117e-07, + "loss": 1.0, + "step": 73131 + }, + { + "epoch": 5.4, + "learning_rate": 5.131015443884791e-07, + "loss": 1.0066, + "step": 73132 + }, + { + "epoch": 5.4, + "learning_rate": 5.12975425489064e-07, + "loss": 1.0421, + "step": 73133 + }, + { + "epoch": 5.4, + "learning_rate": 5.128493216833708e-07, + "loss": 0.9364, + "step": 73134 + }, + { + "epoch": 5.4, + "learning_rate": 5.127232329715981e-07, + "loss": 0.9837, + "step": 73135 + }, + { + "epoch": 5.4, + "learning_rate": 5.12597159353948e-07, + "loss": 1.0342, + "step": 73136 + }, + { + "epoch": 5.4, + "learning_rate": 5.124711008306182e-07, + "loss": 1.0053, + "step": 73137 + }, + { + "epoch": 5.4, + "learning_rate": 5.12345057401813e-07, + "loss": 1.0891, + "step": 73138 + }, + { + "epoch": 5.4, + "learning_rate": 5.122190290677297e-07, + "loss": 0.9855, + "step": 73139 + }, + { + "epoch": 5.4, + "learning_rate": 5.12093015828572e-07, + "loss": 1.0052, + "step": 73140 + }, + { + "epoch": 5.4, + "learning_rate": 5.11967017684537e-07, + "loss": 1.0308, + "step": 73141 + }, + { + "epoch": 5.4, + "learning_rate": 5.11841034635826e-07, + "loss": 1.0251, + "step": 73142 + }, + { + "epoch": 5.4, + "learning_rate": 5.117150666826431e-07, + "loss": 0.9162, + "step": 73143 + }, + { + "epoch": 5.4, + "learning_rate": 5.115891138251827e-07, + "loss": 0.9742, + "step": 73144 + }, + { + "epoch": 5.4, + "learning_rate": 5.114631760636501e-07, + "loss": 1.1578, + "step": 73145 + }, + { + "epoch": 5.4, + "learning_rate": 5.113372533982419e-07, + "loss": 0.9882, + "step": 73146 + }, + { + "epoch": 5.4, + "learning_rate": 5.112113458291635e-07, + "loss": 0.9329, + "step": 73147 + }, + { + "epoch": 5.4, + "learning_rate": 5.110854533566089e-07, + "loss": 0.913, + "step": 73148 + }, + { + "epoch": 5.4, + "learning_rate": 5.10959575980784e-07, + "loss": 1.1074, + "step": 73149 + }, + { + "epoch": 5.41, + "learning_rate": 5.10833713701886e-07, + "loss": 0.9684, + "step": 73150 + }, + { + "epoch": 5.41, + "learning_rate": 5.107078665201159e-07, + "loss": 1.0475, + "step": 73151 + }, + { + "epoch": 5.41, + "learning_rate": 5.105820344356726e-07, + "loss": 1.0109, + "step": 73152 + }, + { + "epoch": 5.41, + "learning_rate": 5.104562174487582e-07, + "loss": 1.0507, + "step": 73153 + }, + { + "epoch": 5.41, + "learning_rate": 5.103304155595723e-07, + "loss": 0.9407, + "step": 73154 + }, + { + "epoch": 5.41, + "learning_rate": 5.102046287683148e-07, + "loss": 0.9889, + "step": 73155 + }, + { + "epoch": 5.41, + "learning_rate": 5.100788570751869e-07, + "loss": 1.0566, + "step": 73156 + }, + { + "epoch": 5.41, + "learning_rate": 5.099531004803848e-07, + "loss": 1.0263, + "step": 73157 + }, + { + "epoch": 5.41, + "learning_rate": 5.09827358984114e-07, + "loss": 1.0282, + "step": 73158 + }, + { + "epoch": 5.41, + "learning_rate": 5.097016325865723e-07, + "loss": 1.0573, + "step": 73159 + }, + { + "epoch": 5.41, + "learning_rate": 5.095759212879592e-07, + "loss": 1.2037, + "step": 73160 + }, + { + "epoch": 5.41, + "learning_rate": 5.094502250884747e-07, + "loss": 1.0746, + "step": 73161 + }, + { + "epoch": 5.41, + "learning_rate": 5.093245439883188e-07, + "loss": 1.1341, + "step": 73162 + }, + { + "epoch": 5.41, + "learning_rate": 5.091988779876922e-07, + "loss": 1.0062, + "step": 73163 + }, + { + "epoch": 5.41, + "learning_rate": 5.090732270867948e-07, + "loss": 1.0678, + "step": 73164 + }, + { + "epoch": 5.41, + "learning_rate": 5.089475912858255e-07, + "loss": 1.0404, + "step": 73165 + }, + { + "epoch": 5.41, + "learning_rate": 5.088219705849839e-07, + "loss": 1.0191, + "step": 73166 + }, + { + "epoch": 5.41, + "learning_rate": 5.086963649844722e-07, + "loss": 1.1677, + "step": 73167 + }, + { + "epoch": 5.41, + "learning_rate": 5.085707744844881e-07, + "loss": 1.0084, + "step": 73168 + }, + { + "epoch": 5.41, + "learning_rate": 5.084451990852335e-07, + "loss": 1.0724, + "step": 73169 + }, + { + "epoch": 5.41, + "learning_rate": 5.08319638786906e-07, + "loss": 1.0621, + "step": 73170 + }, + { + "epoch": 5.41, + "learning_rate": 5.081940935897045e-07, + "loss": 0.9652, + "step": 73171 + }, + { + "epoch": 5.41, + "learning_rate": 5.080685634938332e-07, + "loss": 0.9549, + "step": 73172 + }, + { + "epoch": 5.41, + "learning_rate": 5.079430484994862e-07, + "loss": 0.9924, + "step": 73173 + }, + { + "epoch": 5.41, + "learning_rate": 5.078175486068671e-07, + "loss": 0.9344, + "step": 73174 + }, + { + "epoch": 5.41, + "learning_rate": 5.076920638161754e-07, + "loss": 0.9001, + "step": 73175 + }, + { + "epoch": 5.41, + "learning_rate": 5.075665941276098e-07, + "loss": 1.0171, + "step": 73176 + }, + { + "epoch": 5.41, + "learning_rate": 5.074411395413681e-07, + "loss": 1.0431, + "step": 73177 + }, + { + "epoch": 5.41, + "learning_rate": 5.073157000576523e-07, + "loss": 0.9385, + "step": 73178 + }, + { + "epoch": 5.41, + "learning_rate": 5.071902756766622e-07, + "loss": 0.9392, + "step": 73179 + }, + { + "epoch": 5.41, + "learning_rate": 5.070648663985967e-07, + "loss": 1.0313, + "step": 73180 + }, + { + "epoch": 5.41, + "learning_rate": 5.069394722236553e-07, + "loss": 1.01, + "step": 73181 + }, + { + "epoch": 5.41, + "learning_rate": 5.06814093152036e-07, + "loss": 0.9459, + "step": 73182 + }, + { + "epoch": 5.41, + "learning_rate": 5.066887291839407e-07, + "loss": 0.9261, + "step": 73183 + }, + { + "epoch": 5.41, + "learning_rate": 5.065633803195679e-07, + "loss": 1.0084, + "step": 73184 + }, + { + "epoch": 5.41, + "learning_rate": 5.064380465591168e-07, + "loss": 0.979, + "step": 73185 + }, + { + "epoch": 5.41, + "learning_rate": 5.063127279027857e-07, + "loss": 0.8917, + "step": 73186 + }, + { + "epoch": 5.41, + "learning_rate": 5.061874243507769e-07, + "loss": 0.9707, + "step": 73187 + }, + { + "epoch": 5.41, + "learning_rate": 5.060621359032892e-07, + "loss": 1.0571, + "step": 73188 + }, + { + "epoch": 5.41, + "learning_rate": 5.059368625605177e-07, + "loss": 0.9931, + "step": 73189 + }, + { + "epoch": 5.41, + "learning_rate": 5.058116043226657e-07, + "loss": 1.0189, + "step": 73190 + }, + { + "epoch": 5.41, + "learning_rate": 5.05686361189931e-07, + "loss": 0.9869, + "step": 73191 + }, + { + "epoch": 5.41, + "learning_rate": 5.055611331625165e-07, + "loss": 1.0819, + "step": 73192 + }, + { + "epoch": 5.41, + "learning_rate": 5.054359202406156e-07, + "loss": 0.9985, + "step": 73193 + }, + { + "epoch": 5.41, + "learning_rate": 5.053107224244314e-07, + "loss": 0.9609, + "step": 73194 + }, + { + "epoch": 5.41, + "learning_rate": 5.051855397141603e-07, + "loss": 1.0633, + "step": 73195 + }, + { + "epoch": 5.41, + "learning_rate": 5.050603721100067e-07, + "loss": 0.9586, + "step": 73196 + }, + { + "epoch": 5.41, + "learning_rate": 5.049352196121637e-07, + "loss": 0.9306, + "step": 73197 + }, + { + "epoch": 5.41, + "learning_rate": 5.048100822208324e-07, + "loss": 1.0162, + "step": 73198 + }, + { + "epoch": 5.41, + "learning_rate": 5.046849599362135e-07, + "loss": 1.0665, + "step": 73199 + }, + { + "epoch": 5.41, + "learning_rate": 5.045598527585049e-07, + "loss": 1.0822, + "step": 73200 + }, + { + "epoch": 5.41, + "learning_rate": 5.044347606879064e-07, + "loss": 0.9751, + "step": 73201 + }, + { + "epoch": 5.41, + "learning_rate": 5.043096837246142e-07, + "loss": 1.0046, + "step": 73202 + }, + { + "epoch": 5.41, + "learning_rate": 5.041846218688307e-07, + "loss": 0.9441, + "step": 73203 + }, + { + "epoch": 5.41, + "learning_rate": 5.040595751207545e-07, + "loss": 1.0185, + "step": 73204 + }, + { + "epoch": 5.41, + "learning_rate": 5.039345434805831e-07, + "loss": 0.9695, + "step": 73205 + }, + { + "epoch": 5.41, + "learning_rate": 5.038095269485155e-07, + "loss": 0.8873, + "step": 73206 + }, + { + "epoch": 5.41, + "learning_rate": 5.036845255247503e-07, + "loss": 1.0614, + "step": 73207 + }, + { + "epoch": 5.41, + "learning_rate": 5.035595392094883e-07, + "loss": 1.0673, + "step": 73208 + }, + { + "epoch": 5.41, + "learning_rate": 5.034345680029273e-07, + "loss": 0.9943, + "step": 73209 + }, + { + "epoch": 5.41, + "learning_rate": 5.033096119052661e-07, + "loss": 1.0927, + "step": 73210 + }, + { + "epoch": 5.41, + "learning_rate": 5.031846709167021e-07, + "loss": 1.0144, + "step": 73211 + }, + { + "epoch": 5.41, + "learning_rate": 5.030597450374364e-07, + "loss": 0.9972, + "step": 73212 + }, + { + "epoch": 5.41, + "learning_rate": 5.029348342676676e-07, + "loss": 1.0245, + "step": 73213 + }, + { + "epoch": 5.41, + "learning_rate": 5.028099386075924e-07, + "loss": 1.0072, + "step": 73214 + }, + { + "epoch": 5.41, + "learning_rate": 5.026850580574105e-07, + "loss": 1.0387, + "step": 73215 + }, + { + "epoch": 5.41, + "learning_rate": 5.025601926173218e-07, + "loss": 1.0234, + "step": 73216 + }, + { + "epoch": 5.41, + "learning_rate": 5.024353422875249e-07, + "loss": 1.1239, + "step": 73217 + }, + { + "epoch": 5.41, + "learning_rate": 5.023105070682155e-07, + "loss": 1.0059, + "step": 73218 + }, + { + "epoch": 5.41, + "learning_rate": 5.021856869595942e-07, + "loss": 0.9391, + "step": 73219 + }, + { + "epoch": 5.41, + "learning_rate": 5.020608819618611e-07, + "loss": 0.9672, + "step": 73220 + }, + { + "epoch": 5.41, + "learning_rate": 5.019360920752126e-07, + "loss": 1.0587, + "step": 73221 + }, + { + "epoch": 5.41, + "learning_rate": 5.018113172998463e-07, + "loss": 0.9752, + "step": 73222 + }, + { + "epoch": 5.41, + "learning_rate": 5.016865576359642e-07, + "loss": 0.9852, + "step": 73223 + }, + { + "epoch": 5.41, + "learning_rate": 5.015618130837619e-07, + "loss": 1.0075, + "step": 73224 + }, + { + "epoch": 5.41, + "learning_rate": 5.014370836434401e-07, + "loss": 1.0402, + "step": 73225 + }, + { + "epoch": 5.41, + "learning_rate": 5.013123693151945e-07, + "loss": 1.0929, + "step": 73226 + }, + { + "epoch": 5.41, + "learning_rate": 5.011876700992246e-07, + "loss": 1.0086, + "step": 73227 + }, + { + "epoch": 5.41, + "learning_rate": 5.010629859957306e-07, + "loss": 1.0363, + "step": 73228 + }, + { + "epoch": 5.41, + "learning_rate": 5.009383170049098e-07, + "loss": 1.0486, + "step": 73229 + }, + { + "epoch": 5.41, + "learning_rate": 5.0081366312696e-07, + "loss": 0.9729, + "step": 73230 + }, + { + "epoch": 5.41, + "learning_rate": 5.006890243620777e-07, + "loss": 0.9977, + "step": 73231 + }, + { + "epoch": 5.41, + "learning_rate": 5.005644007104648e-07, + "loss": 1.0688, + "step": 73232 + }, + { + "epoch": 5.41, + "learning_rate": 5.004397921723192e-07, + "loss": 0.9854, + "step": 73233 + }, + { + "epoch": 5.41, + "learning_rate": 5.003151987478372e-07, + "loss": 1.0481, + "step": 73234 + }, + { + "epoch": 5.41, + "learning_rate": 5.001906204372175e-07, + "loss": 1.1032, + "step": 73235 + }, + { + "epoch": 5.41, + "learning_rate": 5.000660572406568e-07, + "loss": 0.9366, + "step": 73236 + }, + { + "epoch": 5.41, + "learning_rate": 4.999415091583593e-07, + "loss": 0.9161, + "step": 73237 + }, + { + "epoch": 5.41, + "learning_rate": 4.998169761905147e-07, + "loss": 0.9886, + "step": 73238 + }, + { + "epoch": 5.41, + "learning_rate": 4.996924583373275e-07, + "loss": 1.0178, + "step": 73239 + }, + { + "epoch": 5.41, + "learning_rate": 4.99567955598993e-07, + "loss": 0.9356, + "step": 73240 + }, + { + "epoch": 5.41, + "learning_rate": 4.99443467975711e-07, + "loss": 1.0082, + "step": 73241 + }, + { + "epoch": 5.41, + "learning_rate": 4.99318995467677e-07, + "loss": 0.9367, + "step": 73242 + }, + { + "epoch": 5.41, + "learning_rate": 4.991945380750918e-07, + "loss": 0.9106, + "step": 73243 + }, + { + "epoch": 5.41, + "learning_rate": 4.99070095798152e-07, + "loss": 1.0501, + "step": 73244 + }, + { + "epoch": 5.41, + "learning_rate": 4.989456686370552e-07, + "loss": 0.9986, + "step": 73245 + }, + { + "epoch": 5.41, + "learning_rate": 4.988212565920003e-07, + "loss": 1.0353, + "step": 73246 + }, + { + "epoch": 5.41, + "learning_rate": 4.986968596631836e-07, + "loss": 0.9475, + "step": 73247 + }, + { + "epoch": 5.41, + "learning_rate": 4.985724778508061e-07, + "loss": 0.8904, + "step": 73248 + }, + { + "epoch": 5.41, + "learning_rate": 4.984481111550632e-07, + "loss": 0.9803, + "step": 73249 + }, + { + "epoch": 5.41, + "learning_rate": 4.983237595761537e-07, + "loss": 1.049, + "step": 73250 + }, + { + "epoch": 5.41, + "learning_rate": 4.981994231142739e-07, + "loss": 0.9596, + "step": 73251 + }, + { + "epoch": 5.41, + "learning_rate": 4.980751017696239e-07, + "loss": 1.0213, + "step": 73252 + }, + { + "epoch": 5.41, + "learning_rate": 4.979507955424001e-07, + "loss": 1.0125, + "step": 73253 + }, + { + "epoch": 5.41, + "learning_rate": 4.978265044328013e-07, + "loss": 1.0383, + "step": 73254 + }, + { + "epoch": 5.41, + "learning_rate": 4.977022284410237e-07, + "loss": 1.0632, + "step": 73255 + }, + { + "epoch": 5.41, + "learning_rate": 4.975779675672654e-07, + "loss": 1.0317, + "step": 73256 + }, + { + "epoch": 5.41, + "learning_rate": 4.974537218117259e-07, + "loss": 1.0842, + "step": 73257 + }, + { + "epoch": 5.41, + "learning_rate": 4.973294911746018e-07, + "loss": 0.9708, + "step": 73258 + }, + { + "epoch": 5.41, + "learning_rate": 4.972052756560896e-07, + "loss": 0.9626, + "step": 73259 + }, + { + "epoch": 5.41, + "learning_rate": 4.970810752563871e-07, + "loss": 0.9892, + "step": 73260 + }, + { + "epoch": 5.41, + "learning_rate": 4.969568899756938e-07, + "loss": 0.9328, + "step": 73261 + }, + { + "epoch": 5.41, + "learning_rate": 4.968327198142075e-07, + "loss": 0.9494, + "step": 73262 + }, + { + "epoch": 5.41, + "learning_rate": 4.967085647721226e-07, + "loss": 1.0346, + "step": 73263 + }, + { + "epoch": 5.41, + "learning_rate": 4.965844248496387e-07, + "loss": 1.0253, + "step": 73264 + }, + { + "epoch": 5.41, + "learning_rate": 4.964603000469525e-07, + "loss": 0.9842, + "step": 73265 + }, + { + "epoch": 5.41, + "learning_rate": 4.963361903642639e-07, + "loss": 1.0599, + "step": 73266 + }, + { + "epoch": 5.41, + "learning_rate": 4.962120958017658e-07, + "loss": 1.0453, + "step": 73267 + }, + { + "epoch": 5.41, + "learning_rate": 4.960880163596593e-07, + "loss": 1.0932, + "step": 73268 + }, + { + "epoch": 5.41, + "learning_rate": 4.959639520381409e-07, + "loss": 1.0407, + "step": 73269 + }, + { + "epoch": 5.41, + "learning_rate": 4.958399028374083e-07, + "loss": 0.9307, + "step": 73270 + }, + { + "epoch": 5.41, + "learning_rate": 4.957158687576568e-07, + "loss": 0.9402, + "step": 73271 + }, + { + "epoch": 5.41, + "learning_rate": 4.955918497990874e-07, + "loss": 1.0246, + "step": 73272 + }, + { + "epoch": 5.41, + "learning_rate": 4.954678459618944e-07, + "loss": 1.0307, + "step": 73273 + }, + { + "epoch": 5.41, + "learning_rate": 4.953438572462765e-07, + "loss": 0.9757, + "step": 73274 + }, + { + "epoch": 5.41, + "learning_rate": 4.952198836524313e-07, + "loss": 0.9499, + "step": 73275 + }, + { + "epoch": 5.41, + "learning_rate": 4.950959251805532e-07, + "loss": 0.9995, + "step": 73276 + }, + { + "epoch": 5.41, + "learning_rate": 4.949719818308429e-07, + "loss": 1.0098, + "step": 73277 + }, + { + "epoch": 5.41, + "learning_rate": 4.948480536034961e-07, + "loss": 0.9653, + "step": 73278 + }, + { + "epoch": 5.41, + "learning_rate": 4.947241404987102e-07, + "loss": 1.0152, + "step": 73279 + }, + { + "epoch": 5.41, + "learning_rate": 4.946002425166807e-07, + "loss": 0.9882, + "step": 73280 + }, + { + "epoch": 5.41, + "learning_rate": 4.944763596576075e-07, + "loss": 1.096, + "step": 73281 + }, + { + "epoch": 5.41, + "learning_rate": 4.94352491921688e-07, + "loss": 1.0116, + "step": 73282 + }, + { + "epoch": 5.41, + "learning_rate": 4.942286393091145e-07, + "loss": 0.9239, + "step": 73283 + }, + { + "epoch": 5.41, + "learning_rate": 4.941048018200889e-07, + "loss": 0.9564, + "step": 73284 + }, + { + "epoch": 5.41, + "learning_rate": 4.939809794548045e-07, + "loss": 0.9785, + "step": 73285 + }, + { + "epoch": 5.42, + "learning_rate": 4.938571722134644e-07, + "loss": 0.9307, + "step": 73286 + }, + { + "epoch": 5.42, + "learning_rate": 4.937333800962585e-07, + "loss": 1.0537, + "step": 73287 + }, + { + "epoch": 5.42, + "learning_rate": 4.936096031033877e-07, + "loss": 1.0036, + "step": 73288 + }, + { + "epoch": 5.42, + "learning_rate": 4.934858412350463e-07, + "loss": 0.9203, + "step": 73289 + }, + { + "epoch": 5.42, + "learning_rate": 4.933620944914364e-07, + "loss": 1.0569, + "step": 73290 + }, + { + "epoch": 5.42, + "learning_rate": 4.9323836287275e-07, + "loss": 1.0334, + "step": 73291 + }, + { + "epoch": 5.42, + "learning_rate": 4.931146463791836e-07, + "loss": 0.9956, + "step": 73292 + }, + { + "epoch": 5.42, + "learning_rate": 4.929909450109372e-07, + "loss": 0.9931, + "step": 73293 + }, + { + "epoch": 5.42, + "learning_rate": 4.928672587682059e-07, + "loss": 0.9468, + "step": 73294 + }, + { + "epoch": 5.42, + "learning_rate": 4.927435876511877e-07, + "loss": 0.8988, + "step": 73295 + }, + { + "epoch": 5.42, + "learning_rate": 4.926199316600755e-07, + "loss": 0.9277, + "step": 73296 + }, + { + "epoch": 5.42, + "learning_rate": 4.924962907950714e-07, + "loss": 1.0924, + "step": 73297 + }, + { + "epoch": 5.42, + "learning_rate": 4.923726650563698e-07, + "loss": 1.0792, + "step": 73298 + }, + { + "epoch": 5.42, + "learning_rate": 4.922490544441671e-07, + "loss": 1.0298, + "step": 73299 + }, + { + "epoch": 5.42, + "learning_rate": 4.9212545895866e-07, + "loss": 1.0775, + "step": 73300 + }, + { + "epoch": 5.42, + "learning_rate": 4.920018786000436e-07, + "loss": 1.0065, + "step": 73301 + }, + { + "epoch": 5.42, + "learning_rate": 4.918783133685179e-07, + "loss": 1.0646, + "step": 73302 + }, + { + "epoch": 5.42, + "learning_rate": 4.917547632642783e-07, + "loss": 0.9792, + "step": 73303 + }, + { + "epoch": 5.42, + "learning_rate": 4.916312282875202e-07, + "loss": 1.0452, + "step": 73304 + }, + { + "epoch": 5.42, + "learning_rate": 4.91507708438439e-07, + "loss": 0.9913, + "step": 73305 + }, + { + "epoch": 5.42, + "learning_rate": 4.913842037172356e-07, + "loss": 0.9775, + "step": 73306 + }, + { + "epoch": 5.42, + "learning_rate": 4.912607141241032e-07, + "loss": 0.9793, + "step": 73307 + }, + { + "epoch": 5.42, + "learning_rate": 4.911372396592385e-07, + "loss": 1.0601, + "step": 73308 + }, + { + "epoch": 5.42, + "learning_rate": 4.910137803228387e-07, + "loss": 1.0945, + "step": 73309 + }, + { + "epoch": 5.42, + "learning_rate": 4.908903361150996e-07, + "loss": 1.0304, + "step": 73310 + }, + { + "epoch": 5.42, + "learning_rate": 4.907669070362186e-07, + "loss": 0.9942, + "step": 73311 + }, + { + "epoch": 5.42, + "learning_rate": 4.906434930863902e-07, + "loss": 1.0564, + "step": 73312 + }, + { + "epoch": 5.42, + "learning_rate": 4.905200942658129e-07, + "loss": 0.9254, + "step": 73313 + }, + { + "epoch": 5.42, + "learning_rate": 4.903967105746821e-07, + "loss": 0.9909, + "step": 73314 + }, + { + "epoch": 5.42, + "learning_rate": 4.902733420131933e-07, + "loss": 0.9491, + "step": 73315 + }, + { + "epoch": 5.42, + "learning_rate": 4.90149988581543e-07, + "loss": 0.9991, + "step": 73316 + }, + { + "epoch": 5.42, + "learning_rate": 4.900266502799301e-07, + "loss": 1.0676, + "step": 73317 + }, + { + "epoch": 5.42, + "learning_rate": 4.899033271085473e-07, + "loss": 1.0403, + "step": 73318 + }, + { + "epoch": 5.42, + "learning_rate": 4.897800190675928e-07, + "loss": 0.9981, + "step": 73319 + }, + { + "epoch": 5.42, + "learning_rate": 4.896567261572627e-07, + "loss": 0.9635, + "step": 73320 + }, + { + "epoch": 5.42, + "learning_rate": 4.895334483777503e-07, + "loss": 0.9909, + "step": 73321 + }, + { + "epoch": 5.42, + "learning_rate": 4.894101857292565e-07, + "loss": 1.0361, + "step": 73322 + }, + { + "epoch": 5.42, + "learning_rate": 4.892869382119747e-07, + "loss": 1.0117, + "step": 73323 + }, + { + "epoch": 5.42, + "learning_rate": 4.891637058261011e-07, + "loss": 0.9374, + "step": 73324 + }, + { + "epoch": 5.42, + "learning_rate": 4.890404885718303e-07, + "loss": 1.1173, + "step": 73325 + }, + { + "epoch": 5.42, + "learning_rate": 4.889172864493619e-07, + "loss": 0.9873, + "step": 73326 + }, + { + "epoch": 5.42, + "learning_rate": 4.887940994588902e-07, + "loss": 1.0375, + "step": 73327 + }, + { + "epoch": 5.42, + "learning_rate": 4.886709276006107e-07, + "loss": 1.0272, + "step": 73328 + }, + { + "epoch": 5.42, + "learning_rate": 4.885477708747189e-07, + "loss": 0.9847, + "step": 73329 + }, + { + "epoch": 5.42, + "learning_rate": 4.88424629281411e-07, + "loss": 0.9416, + "step": 73330 + }, + { + "epoch": 5.42, + "learning_rate": 4.88301502820886e-07, + "loss": 0.9624, + "step": 73331 + }, + { + "epoch": 5.42, + "learning_rate": 4.881783914933347e-07, + "loss": 0.9, + "step": 73332 + }, + { + "epoch": 5.42, + "learning_rate": 4.88055295298957e-07, + "loss": 1.0424, + "step": 73333 + }, + { + "epoch": 5.42, + "learning_rate": 4.87932214237945e-07, + "loss": 0.9587, + "step": 73334 + }, + { + "epoch": 5.42, + "learning_rate": 4.878091483105007e-07, + "loss": 0.9503, + "step": 73335 + }, + { + "epoch": 5.42, + "learning_rate": 4.876860975168118e-07, + "loss": 0.9678, + "step": 73336 + }, + { + "epoch": 5.42, + "learning_rate": 4.875630618570804e-07, + "loss": 1.0454, + "step": 73337 + }, + { + "epoch": 5.42, + "learning_rate": 4.874400413315006e-07, + "loss": 0.9531, + "step": 73338 + }, + { + "epoch": 5.42, + "learning_rate": 4.873170359402668e-07, + "loss": 1.0399, + "step": 73339 + }, + { + "epoch": 5.42, + "learning_rate": 4.871940456835756e-07, + "loss": 1.0388, + "step": 73340 + }, + { + "epoch": 5.42, + "learning_rate": 4.87071070561621e-07, + "loss": 0.9481, + "step": 73341 + }, + { + "epoch": 5.42, + "learning_rate": 4.869481105746021e-07, + "loss": 0.9848, + "step": 73342 + }, + { + "epoch": 5.42, + "learning_rate": 4.86825165722713e-07, + "loss": 0.9762, + "step": 73343 + }, + { + "epoch": 5.42, + "learning_rate": 4.86702236006148e-07, + "loss": 1.0406, + "step": 73344 + }, + { + "epoch": 5.42, + "learning_rate": 4.865793214251024e-07, + "loss": 0.9736, + "step": 73345 + }, + { + "epoch": 5.42, + "learning_rate": 4.864564219797752e-07, + "loss": 1.0242, + "step": 73346 + }, + { + "epoch": 5.42, + "learning_rate": 4.863335376703582e-07, + "loss": 0.9207, + "step": 73347 + }, + { + "epoch": 5.42, + "learning_rate": 4.862106684970491e-07, + "loss": 1.0681, + "step": 73348 + }, + { + "epoch": 5.42, + "learning_rate": 4.860878144600433e-07, + "loss": 1.0016, + "step": 73349 + }, + { + "epoch": 5.42, + "learning_rate": 4.859649755595342e-07, + "loss": 1.0331, + "step": 73350 + }, + { + "epoch": 5.42, + "learning_rate": 4.858421517957213e-07, + "loss": 1.0329, + "step": 73351 + }, + { + "epoch": 5.42, + "learning_rate": 4.857193431687945e-07, + "loss": 0.9561, + "step": 73352 + }, + { + "epoch": 5.42, + "learning_rate": 4.855965496789539e-07, + "loss": 0.9898, + "step": 73353 + }, + { + "epoch": 5.42, + "learning_rate": 4.854737713263913e-07, + "loss": 1.0612, + "step": 73354 + }, + { + "epoch": 5.42, + "learning_rate": 4.853510081113055e-07, + "loss": 0.9505, + "step": 73355 + }, + { + "epoch": 5.42, + "learning_rate": 4.852282600338897e-07, + "loss": 0.9461, + "step": 73356 + }, + { + "epoch": 5.42, + "learning_rate": 4.851055270943384e-07, + "loss": 0.9609, + "step": 73357 + }, + { + "epoch": 5.42, + "learning_rate": 4.849828092928499e-07, + "loss": 1.0521, + "step": 73358 + }, + { + "epoch": 5.42, + "learning_rate": 4.848601066296166e-07, + "loss": 1.0576, + "step": 73359 + }, + { + "epoch": 5.42, + "learning_rate": 4.84737419104836e-07, + "loss": 1.0013, + "step": 73360 + }, + { + "epoch": 5.42, + "learning_rate": 4.84614746718699e-07, + "loss": 0.9982, + "step": 73361 + }, + { + "epoch": 5.42, + "learning_rate": 4.844920894714067e-07, + "loss": 0.9701, + "step": 73362 + }, + { + "epoch": 5.42, + "learning_rate": 4.843694473631499e-07, + "loss": 0.9937, + "step": 73363 + }, + { + "epoch": 5.42, + "learning_rate": 4.842468203941264e-07, + "loss": 1.1624, + "step": 73364 + }, + { + "epoch": 5.42, + "learning_rate": 4.841242085645292e-07, + "loss": 0.959, + "step": 73365 + }, + { + "epoch": 5.42, + "learning_rate": 4.840016118745539e-07, + "loss": 0.9882, + "step": 73366 + }, + { + "epoch": 5.42, + "learning_rate": 4.838790303243968e-07, + "loss": 0.9245, + "step": 73367 + }, + { + "epoch": 5.42, + "learning_rate": 4.837564639142511e-07, + "loss": 1.0708, + "step": 73368 + }, + { + "epoch": 5.42, + "learning_rate": 4.836339126443135e-07, + "loss": 0.9745, + "step": 73369 + }, + { + "epoch": 5.42, + "learning_rate": 4.83511376514777e-07, + "loss": 0.9977, + "step": 73370 + }, + { + "epoch": 5.42, + "learning_rate": 4.833888555258392e-07, + "loss": 0.9915, + "step": 73371 + }, + { + "epoch": 5.42, + "learning_rate": 4.832663496776923e-07, + "loss": 0.9182, + "step": 73372 + }, + { + "epoch": 5.42, + "learning_rate": 4.83143858970534e-07, + "loss": 1.052, + "step": 73373 + }, + { + "epoch": 5.42, + "learning_rate": 4.830213834045561e-07, + "loss": 1.0113, + "step": 73374 + }, + { + "epoch": 5.42, + "learning_rate": 4.828989229799553e-07, + "loss": 1.0016, + "step": 73375 + }, + { + "epoch": 5.42, + "learning_rate": 4.827764776969279e-07, + "loss": 1.0618, + "step": 73376 + }, + { + "epoch": 5.42, + "learning_rate": 4.826540475556652e-07, + "loss": 0.9795, + "step": 73377 + }, + { + "epoch": 5.42, + "learning_rate": 4.825316325563645e-07, + "loss": 0.9645, + "step": 73378 + }, + { + "epoch": 5.42, + "learning_rate": 4.82409232699218e-07, + "loss": 1.0536, + "step": 73379 + }, + { + "epoch": 5.42, + "learning_rate": 4.822868479844256e-07, + "loss": 0.9334, + "step": 73380 + }, + { + "epoch": 5.42, + "learning_rate": 4.821644784121748e-07, + "loss": 1.0819, + "step": 73381 + }, + { + "epoch": 5.42, + "learning_rate": 4.820421239826667e-07, + "loss": 1.0833, + "step": 73382 + }, + { + "epoch": 5.42, + "learning_rate": 4.819197846960922e-07, + "loss": 0.9917, + "step": 73383 + }, + { + "epoch": 5.42, + "learning_rate": 4.817974605526487e-07, + "loss": 0.9734, + "step": 73384 + }, + { + "epoch": 5.42, + "learning_rate": 4.816751515525286e-07, + "loss": 0.9607, + "step": 73385 + }, + { + "epoch": 5.42, + "learning_rate": 4.81552857695926e-07, + "loss": 1.0756, + "step": 73386 + }, + { + "epoch": 5.42, + "learning_rate": 4.814305789830376e-07, + "loss": 0.966, + "step": 73387 + }, + { + "epoch": 5.42, + "learning_rate": 4.813083154140574e-07, + "loss": 0.9564, + "step": 73388 + }, + { + "epoch": 5.42, + "learning_rate": 4.811860669891788e-07, + "loss": 1.0204, + "step": 73389 + }, + { + "epoch": 5.42, + "learning_rate": 4.81063833708596e-07, + "loss": 0.9059, + "step": 73390 + }, + { + "epoch": 5.42, + "learning_rate": 4.809416155725055e-07, + "loss": 0.9437, + "step": 73391 + }, + { + "epoch": 5.42, + "learning_rate": 4.808194125811017e-07, + "loss": 1.0244, + "step": 73392 + }, + { + "epoch": 5.42, + "learning_rate": 4.806972247345765e-07, + "loss": 1.0031, + "step": 73393 + }, + { + "epoch": 5.42, + "learning_rate": 4.805750520331265e-07, + "loss": 0.9262, + "step": 73394 + }, + { + "epoch": 5.42, + "learning_rate": 4.804528944769449e-07, + "loss": 0.9659, + "step": 73395 + }, + { + "epoch": 5.42, + "learning_rate": 4.803307520662282e-07, + "loss": 0.9859, + "step": 73396 + }, + { + "epoch": 5.42, + "learning_rate": 4.802086248011661e-07, + "loss": 1.0382, + "step": 73397 + }, + { + "epoch": 5.42, + "learning_rate": 4.800865126819576e-07, + "loss": 0.9388, + "step": 73398 + }, + { + "epoch": 5.42, + "learning_rate": 4.799644157087946e-07, + "loss": 1.0191, + "step": 73399 + }, + { + "epoch": 5.42, + "learning_rate": 4.798423338818736e-07, + "loss": 0.9548, + "step": 73400 + }, + { + "epoch": 5.42, + "learning_rate": 4.797202672013857e-07, + "loss": 1.0162, + "step": 73401 + }, + { + "epoch": 5.42, + "learning_rate": 4.795982156675272e-07, + "loss": 0.9975, + "step": 73402 + }, + { + "epoch": 5.42, + "learning_rate": 4.794761792804925e-07, + "loss": 1.0079, + "step": 73403 + }, + { + "epoch": 5.42, + "learning_rate": 4.793541580404737e-07, + "loss": 1.0609, + "step": 73404 + }, + { + "epoch": 5.42, + "learning_rate": 4.792321519476673e-07, + "loss": 1.044, + "step": 73405 + }, + { + "epoch": 5.42, + "learning_rate": 4.791101610022641e-07, + "loss": 1.0544, + "step": 73406 + }, + { + "epoch": 5.42, + "learning_rate": 4.789881852044631e-07, + "loss": 0.9613, + "step": 73407 + }, + { + "epoch": 5.42, + "learning_rate": 4.78866224554454e-07, + "loss": 1.0198, + "step": 73408 + }, + { + "epoch": 5.42, + "learning_rate": 4.787442790524332e-07, + "loss": 0.9295, + "step": 73409 + }, + { + "epoch": 5.42, + "learning_rate": 4.78622348698593e-07, + "loss": 0.985, + "step": 73410 + }, + { + "epoch": 5.42, + "learning_rate": 4.785004334931298e-07, + "loss": 0.9861, + "step": 73411 + }, + { + "epoch": 5.42, + "learning_rate": 4.783785334362356e-07, + "loss": 0.9719, + "step": 73412 + }, + { + "epoch": 5.42, + "learning_rate": 4.782566485281059e-07, + "loss": 0.9482, + "step": 73413 + }, + { + "epoch": 5.42, + "learning_rate": 4.781347787689328e-07, + "loss": 0.9982, + "step": 73414 + }, + { + "epoch": 5.42, + "learning_rate": 4.780129241589105e-07, + "loss": 0.9539, + "step": 73415 + }, + { + "epoch": 5.42, + "learning_rate": 4.778910846982331e-07, + "loss": 1.0319, + "step": 73416 + }, + { + "epoch": 5.42, + "learning_rate": 4.777692603870965e-07, + "loss": 1.0315, + "step": 73417 + }, + { + "epoch": 5.42, + "learning_rate": 4.776474512256923e-07, + "loss": 0.9623, + "step": 73418 + }, + { + "epoch": 5.42, + "learning_rate": 4.775256572142128e-07, + "loss": 1.1198, + "step": 73419 + }, + { + "epoch": 5.42, + "learning_rate": 4.774038783528557e-07, + "loss": 0.9424, + "step": 73420 + }, + { + "epoch": 5.43, + "learning_rate": 4.772821146418138e-07, + "loss": 1.0331, + "step": 73421 + }, + { + "epoch": 5.43, + "learning_rate": 4.771603660812773e-07, + "loss": 1.0122, + "step": 73422 + }, + { + "epoch": 5.43, + "learning_rate": 4.770386326714438e-07, + "loss": 0.9694, + "step": 73423 + }, + { + "epoch": 5.43, + "learning_rate": 4.769169144125041e-07, + "loss": 0.9187, + "step": 73424 + }, + { + "epoch": 5.43, + "learning_rate": 4.767952113046559e-07, + "loss": 0.9899, + "step": 73425 + }, + { + "epoch": 5.43, + "learning_rate": 4.7667352334808683e-07, + "loss": 1.0024, + "step": 73426 + }, + { + "epoch": 5.43, + "learning_rate": 4.7655185054299667e-07, + "loss": 0.896, + "step": 73427 + }, + { + "epoch": 5.43, + "learning_rate": 4.7643019288957427e-07, + "loss": 0.9997, + "step": 73428 + }, + { + "epoch": 5.43, + "learning_rate": 4.7630855038801826e-07, + "loss": 1.065, + "step": 73429 + }, + { + "epoch": 5.43, + "learning_rate": 4.761869230385163e-07, + "loss": 0.9866, + "step": 73430 + }, + { + "epoch": 5.43, + "learning_rate": 4.7606531084126495e-07, + "loss": 1.1068, + "step": 73431 + }, + { + "epoch": 5.43, + "learning_rate": 4.7594371379645844e-07, + "loss": 1.0054, + "step": 73432 + }, + { + "epoch": 5.43, + "learning_rate": 4.758221319042899e-07, + "loss": 1.0433, + "step": 73433 + }, + { + "epoch": 5.43, + "learning_rate": 4.7570056516495044e-07, + "loss": 0.9741, + "step": 73434 + }, + { + "epoch": 5.43, + "learning_rate": 4.7557901357863536e-07, + "loss": 1.0527, + "step": 73435 + }, + { + "epoch": 5.43, + "learning_rate": 4.754574771455378e-07, + "loss": 1.0183, + "step": 73436 + }, + { + "epoch": 5.43, + "learning_rate": 4.753359558658521e-07, + "loss": 1.0592, + "step": 73437 + }, + { + "epoch": 5.43, + "learning_rate": 4.752144497397704e-07, + "loss": 1.1305, + "step": 73438 + }, + { + "epoch": 5.43, + "learning_rate": 4.7509295876748463e-07, + "loss": 0.9434, + "step": 73439 + }, + { + "epoch": 5.43, + "learning_rate": 4.7497148294919026e-07, + "loss": 0.9971, + "step": 73440 + }, + { + "epoch": 5.43, + "learning_rate": 4.748500222850827e-07, + "loss": 0.989, + "step": 73441 + }, + { + "epoch": 5.43, + "learning_rate": 4.7472857677534954e-07, + "loss": 0.9652, + "step": 73442 + }, + { + "epoch": 5.43, + "learning_rate": 4.746071464201873e-07, + "loss": 0.9115, + "step": 73443 + }, + { + "epoch": 5.43, + "learning_rate": 4.7448573121978814e-07, + "loss": 0.9261, + "step": 73444 + }, + { + "epoch": 5.43, + "learning_rate": 4.7436433117434845e-07, + "loss": 1.1217, + "step": 73445 + }, + { + "epoch": 5.43, + "learning_rate": 4.74242946284057e-07, + "loss": 1.0237, + "step": 73446 + }, + { + "epoch": 5.43, + "learning_rate": 4.741215765491092e-07, + "loss": 0.9494, + "step": 73447 + }, + { + "epoch": 5.43, + "learning_rate": 4.7400022196969596e-07, + "loss": 1.0677, + "step": 73448 + }, + { + "epoch": 5.43, + "learning_rate": 4.738788825460161e-07, + "loss": 1.0014, + "step": 73449 + }, + { + "epoch": 5.43, + "learning_rate": 4.737575582782561e-07, + "loss": 0.9673, + "step": 73450 + }, + { + "epoch": 5.43, + "learning_rate": 4.736362491666102e-07, + "loss": 0.9687, + "step": 73451 + }, + { + "epoch": 5.43, + "learning_rate": 4.73514955211275e-07, + "loss": 0.9807, + "step": 73452 + }, + { + "epoch": 5.43, + "learning_rate": 4.7339367641244027e-07, + "loss": 0.9878, + "step": 73453 + }, + { + "epoch": 5.43, + "learning_rate": 4.7327241277030035e-07, + "loss": 0.9171, + "step": 73454 + }, + { + "epoch": 5.43, + "learning_rate": 4.7315116428504613e-07, + "loss": 1.0474, + "step": 73455 + }, + { + "epoch": 5.43, + "learning_rate": 4.7302993095687423e-07, + "loss": 1.0894, + "step": 73456 + }, + { + "epoch": 5.43, + "learning_rate": 4.7290871278597437e-07, + "loss": 0.9817, + "step": 73457 + }, + { + "epoch": 5.43, + "learning_rate": 4.7278750977254096e-07, + "loss": 0.9526, + "step": 73458 + }, + { + "epoch": 5.43, + "learning_rate": 4.72666321916766e-07, + "loss": 0.9798, + "step": 73459 + }, + { + "epoch": 5.43, + "learning_rate": 4.7254514921884156e-07, + "loss": 1.0232, + "step": 73460 + }, + { + "epoch": 5.43, + "learning_rate": 4.7242399167896304e-07, + "loss": 1.0375, + "step": 73461 + }, + { + "epoch": 5.43, + "learning_rate": 4.723028492973214e-07, + "loss": 1.025, + "step": 73462 + }, + { + "epoch": 5.43, + "learning_rate": 4.7218172207410875e-07, + "loss": 1.0415, + "step": 73463 + }, + { + "epoch": 5.43, + "learning_rate": 4.720606100095182e-07, + "loss": 1.078, + "step": 73464 + }, + { + "epoch": 5.43, + "learning_rate": 4.719395131037441e-07, + "loss": 1.0048, + "step": 73465 + }, + { + "epoch": 5.43, + "learning_rate": 4.7181843135697846e-07, + "loss": 0.9749, + "step": 73466 + }, + { + "epoch": 5.43, + "learning_rate": 4.716973647694123e-07, + "loss": 1.0105, + "step": 73467 + }, + { + "epoch": 5.43, + "learning_rate": 4.715763133412399e-07, + "loss": 0.9521, + "step": 73468 + }, + { + "epoch": 5.43, + "learning_rate": 4.7145527707265217e-07, + "loss": 1.0919, + "step": 73469 + }, + { + "epoch": 5.43, + "learning_rate": 4.713342559638445e-07, + "loss": 0.9188, + "step": 73470 + }, + { + "epoch": 5.43, + "learning_rate": 4.7121325001500575e-07, + "loss": 0.8514, + "step": 73471 + }, + { + "epoch": 5.43, + "learning_rate": 4.710922592263323e-07, + "loss": 1.042, + "step": 73472 + }, + { + "epoch": 5.43, + "learning_rate": 4.7097128359801403e-07, + "loss": 1.0125, + "step": 73473 + }, + { + "epoch": 5.43, + "learning_rate": 4.7085032313024414e-07, + "loss": 1.1134, + "step": 73474 + }, + { + "epoch": 5.43, + "learning_rate": 4.707293778232136e-07, + "loss": 0.9574, + "step": 73475 + }, + { + "epoch": 5.43, + "learning_rate": 4.7060844767711775e-07, + "loss": 1.053, + "step": 73476 + }, + { + "epoch": 5.43, + "learning_rate": 4.7048753269214765e-07, + "loss": 1.0596, + "step": 73477 + }, + { + "epoch": 5.43, + "learning_rate": 4.703666328684953e-07, + "loss": 1.0893, + "step": 73478 + }, + { + "epoch": 5.43, + "learning_rate": 4.7024574820635274e-07, + "loss": 1.1072, + "step": 73479 + }, + { + "epoch": 5.43, + "learning_rate": 4.7012487870591206e-07, + "loss": 0.9326, + "step": 73480 + }, + { + "epoch": 5.43, + "learning_rate": 4.700040243673676e-07, + "loss": 1.0265, + "step": 73481 + }, + { + "epoch": 5.43, + "learning_rate": 4.6988318519091136e-07, + "loss": 1.0837, + "step": 73482 + }, + { + "epoch": 5.43, + "learning_rate": 4.6976236117673325e-07, + "loss": 0.9145, + "step": 73483 + }, + { + "epoch": 5.43, + "learning_rate": 4.696415523250264e-07, + "loss": 0.9997, + "step": 73484 + }, + { + "epoch": 5.43, + "learning_rate": 4.6952075863598403e-07, + "loss": 1.08, + "step": 73485 + }, + { + "epoch": 5.43, + "learning_rate": 4.6939998010979817e-07, + "loss": 1.0147, + "step": 73486 + }, + { + "epoch": 5.43, + "learning_rate": 4.6927921674666086e-07, + "loss": 0.9951, + "step": 73487 + }, + { + "epoch": 5.43, + "learning_rate": 4.6915846854676427e-07, + "loss": 0.8954, + "step": 73488 + }, + { + "epoch": 5.43, + "learning_rate": 4.690377355102982e-07, + "loss": 0.9656, + "step": 73489 + }, + { + "epoch": 5.43, + "learning_rate": 4.6891701763745913e-07, + "loss": 1.0018, + "step": 73490 + }, + { + "epoch": 5.43, + "learning_rate": 4.687963149284347e-07, + "loss": 1.0004, + "step": 73491 + }, + { + "epoch": 5.43, + "learning_rate": 4.686756273834203e-07, + "loss": 1.0857, + "step": 73492 + }, + { + "epoch": 5.43, + "learning_rate": 4.685549550026047e-07, + "loss": 0.8948, + "step": 73493 + }, + { + "epoch": 5.43, + "learning_rate": 4.684342977861844e-07, + "loss": 0.9762, + "step": 73494 + }, + { + "epoch": 5.43, + "learning_rate": 4.683136557343471e-07, + "loss": 0.9864, + "step": 73495 + }, + { + "epoch": 5.43, + "learning_rate": 4.681930288472869e-07, + "loss": 0.9927, + "step": 73496 + }, + { + "epoch": 5.43, + "learning_rate": 4.680724171251949e-07, + "loss": 1.0657, + "step": 73497 + }, + { + "epoch": 5.43, + "learning_rate": 4.679518205682643e-07, + "loss": 0.9402, + "step": 73498 + }, + { + "epoch": 5.43, + "learning_rate": 4.6783123917668596e-07, + "loss": 1.0222, + "step": 73499 + }, + { + "epoch": 5.43, + "learning_rate": 4.6771067295064975e-07, + "loss": 0.9897, + "step": 73500 + }, + { + "epoch": 5.43, + "learning_rate": 4.6759012189035115e-07, + "loss": 1.0129, + "step": 73501 + }, + { + "epoch": 5.43, + "learning_rate": 4.674695859959799e-07, + "loss": 1.0169, + "step": 73502 + }, + { + "epoch": 5.43, + "learning_rate": 4.6734906526772817e-07, + "loss": 0.979, + "step": 73503 + }, + { + "epoch": 5.43, + "learning_rate": 4.6722855970578686e-07, + "loss": 1.0676, + "step": 73504 + }, + { + "epoch": 5.43, + "learning_rate": 4.6710806931034915e-07, + "loss": 0.9646, + "step": 73505 + }, + { + "epoch": 5.43, + "learning_rate": 4.669875940816071e-07, + "loss": 1.1537, + "step": 73506 + }, + { + "epoch": 5.43, + "learning_rate": 4.668671340197506e-07, + "loss": 0.9399, + "step": 73507 + }, + { + "epoch": 5.43, + "learning_rate": 4.667466891249717e-07, + "loss": 0.9791, + "step": 73508 + }, + { + "epoch": 5.43, + "learning_rate": 4.666262593974624e-07, + "loss": 0.9666, + "step": 73509 + }, + { + "epoch": 5.43, + "learning_rate": 4.665058448374149e-07, + "loss": 1.0222, + "step": 73510 + }, + { + "epoch": 5.43, + "learning_rate": 4.663854454450201e-07, + "loss": 1.0222, + "step": 73511 + }, + { + "epoch": 5.43, + "learning_rate": 4.6626506122047e-07, + "loss": 0.9841, + "step": 73512 + }, + { + "epoch": 5.43, + "learning_rate": 4.6614469216395343e-07, + "loss": 0.9591, + "step": 73513 + }, + { + "epoch": 5.43, + "learning_rate": 4.660243382756668e-07, + "loss": 1.0197, + "step": 73514 + }, + { + "epoch": 5.43, + "learning_rate": 4.659039995558001e-07, + "loss": 0.9166, + "step": 73515 + }, + { + "epoch": 5.43, + "learning_rate": 4.6578367600454086e-07, + "loss": 1.1151, + "step": 73516 + }, + { + "epoch": 5.43, + "learning_rate": 4.6566336762208454e-07, + "loss": 0.9801, + "step": 73517 + }, + { + "epoch": 5.43, + "learning_rate": 4.65543074408622e-07, + "loss": 1.1607, + "step": 73518 + }, + { + "epoch": 5.43, + "learning_rate": 4.654227963643432e-07, + "loss": 0.9786, + "step": 73519 + }, + { + "epoch": 5.43, + "learning_rate": 4.653025334894401e-07, + "loss": 0.9956, + "step": 73520 + }, + { + "epoch": 5.43, + "learning_rate": 4.651822857841048e-07, + "loss": 0.9795, + "step": 73521 + }, + { + "epoch": 5.43, + "learning_rate": 4.6506205324852724e-07, + "loss": 1.0771, + "step": 73522 + }, + { + "epoch": 5.43, + "learning_rate": 4.649418358829005e-07, + "loss": 0.9219, + "step": 73523 + }, + { + "epoch": 5.43, + "learning_rate": 4.648216336874156e-07, + "loss": 0.9699, + "step": 73524 + }, + { + "epoch": 5.43, + "learning_rate": 4.6470144666226015e-07, + "loss": 0.9337, + "step": 73525 + }, + { + "epoch": 5.43, + "learning_rate": 4.645812748076306e-07, + "loss": 1.0336, + "step": 73526 + }, + { + "epoch": 5.43, + "learning_rate": 4.644611181237157e-07, + "loss": 0.971, + "step": 73527 + }, + { + "epoch": 5.43, + "learning_rate": 4.6434097661070655e-07, + "loss": 1.009, + "step": 73528 + }, + { + "epoch": 5.43, + "learning_rate": 4.642208502687939e-07, + "loss": 0.9532, + "step": 73529 + }, + { + "epoch": 5.43, + "learning_rate": 4.6410073909817e-07, + "loss": 1.0148, + "step": 73530 + }, + { + "epoch": 5.43, + "learning_rate": 4.639806430990257e-07, + "loss": 0.9898, + "step": 73531 + }, + { + "epoch": 5.43, + "learning_rate": 4.638605622715509e-07, + "loss": 1.085, + "step": 73532 + }, + { + "epoch": 5.43, + "learning_rate": 4.6374049661593647e-07, + "loss": 0.9388, + "step": 73533 + }, + { + "epoch": 5.43, + "learning_rate": 4.636204461323768e-07, + "loss": 1.0303, + "step": 73534 + }, + { + "epoch": 5.43, + "learning_rate": 4.635004108210617e-07, + "loss": 1.0836, + "step": 73535 + }, + { + "epoch": 5.43, + "learning_rate": 4.633803906821777e-07, + "loss": 0.9526, + "step": 73536 + }, + { + "epoch": 5.43, + "learning_rate": 4.6326038571592014e-07, + "loss": 0.904, + "step": 73537 + }, + { + "epoch": 5.43, + "learning_rate": 4.6314039592247786e-07, + "loss": 1.1011, + "step": 73538 + }, + { + "epoch": 5.43, + "learning_rate": 4.6302042130204614e-07, + "loss": 0.9999, + "step": 73539 + }, + { + "epoch": 5.43, + "learning_rate": 4.629004618548083e-07, + "loss": 0.9415, + "step": 73540 + }, + { + "epoch": 5.43, + "learning_rate": 4.627805175809619e-07, + "loss": 0.9595, + "step": 73541 + }, + { + "epoch": 5.43, + "learning_rate": 4.6266058848069337e-07, + "loss": 1.0081, + "step": 73542 + }, + { + "epoch": 5.43, + "learning_rate": 4.6254067455419827e-07, + "loss": 1.1188, + "step": 73543 + }, + { + "epoch": 5.43, + "learning_rate": 4.624207758016641e-07, + "loss": 0.9474, + "step": 73544 + }, + { + "epoch": 5.43, + "learning_rate": 4.623008922232797e-07, + "loss": 0.9431, + "step": 73545 + }, + { + "epoch": 5.43, + "learning_rate": 4.6218102381923923e-07, + "loss": 0.8656, + "step": 73546 + }, + { + "epoch": 5.43, + "learning_rate": 4.6206117058973267e-07, + "loss": 1.0416, + "step": 73547 + }, + { + "epoch": 5.43, + "learning_rate": 4.619413325349498e-07, + "loss": 1.0876, + "step": 73548 + }, + { + "epoch": 5.43, + "learning_rate": 4.6182150965508153e-07, + "loss": 0.9955, + "step": 73549 + }, + { + "epoch": 5.43, + "learning_rate": 4.617017019503189e-07, + "loss": 0.9038, + "step": 73550 + }, + { + "epoch": 5.43, + "learning_rate": 4.615819094208529e-07, + "loss": 0.8521, + "step": 73551 + }, + { + "epoch": 5.43, + "learning_rate": 4.614621320668733e-07, + "loss": 0.9507, + "step": 73552 + }, + { + "epoch": 5.43, + "learning_rate": 4.613423698885722e-07, + "loss": 0.9223, + "step": 73553 + }, + { + "epoch": 5.43, + "learning_rate": 4.6122262288613604e-07, + "loss": 0.9923, + "step": 73554 + }, + { + "epoch": 5.43, + "learning_rate": 4.6110289105976037e-07, + "loss": 1.0214, + "step": 73555 + }, + { + "epoch": 5.44, + "learning_rate": 4.6098317440963383e-07, + "loss": 1.0917, + "step": 73556 + }, + { + "epoch": 5.44, + "learning_rate": 4.6086347293594627e-07, + "loss": 0.9732, + "step": 73557 + }, + { + "epoch": 5.44, + "learning_rate": 4.607437866388864e-07, + "loss": 0.9196, + "step": 73558 + }, + { + "epoch": 5.44, + "learning_rate": 4.606241155186486e-07, + "loss": 1.047, + "step": 73559 + }, + { + "epoch": 5.44, + "learning_rate": 4.6050445957542157e-07, + "loss": 0.9246, + "step": 73560 + }, + { + "epoch": 5.44, + "learning_rate": 4.603848188093951e-07, + "loss": 1.0986, + "step": 73561 + }, + { + "epoch": 5.44, + "learning_rate": 4.6026519322076025e-07, + "loss": 1.0069, + "step": 73562 + }, + { + "epoch": 5.44, + "learning_rate": 4.6014558280970676e-07, + "loss": 1.0875, + "step": 73563 + }, + { + "epoch": 5.44, + "learning_rate": 4.600259875764246e-07, + "loss": 1.1102, + "step": 73564 + }, + { + "epoch": 5.44, + "learning_rate": 4.599064075211046e-07, + "loss": 1.0198, + "step": 73565 + }, + { + "epoch": 5.44, + "learning_rate": 4.597868426439367e-07, + "loss": 0.9158, + "step": 73566 + }, + { + "epoch": 5.44, + "learning_rate": 4.596672929451129e-07, + "loss": 0.9934, + "step": 73567 + }, + { + "epoch": 5.44, + "learning_rate": 4.5954775842482093e-07, + "loss": 1.0601, + "step": 73568 + }, + { + "epoch": 5.44, + "learning_rate": 4.5942823908325164e-07, + "loss": 1.0395, + "step": 73569 + }, + { + "epoch": 5.44, + "learning_rate": 4.5930873492059603e-07, + "loss": 0.9692, + "step": 73570 + }, + { + "epoch": 5.44, + "learning_rate": 4.5918924593704395e-07, + "loss": 1.0051, + "step": 73571 + }, + { + "epoch": 5.44, + "learning_rate": 4.5906977213278524e-07, + "loss": 1.0212, + "step": 73572 + }, + { + "epoch": 5.44, + "learning_rate": 4.589503135080098e-07, + "loss": 1.0282, + "step": 73573 + }, + { + "epoch": 5.44, + "learning_rate": 4.588308700629063e-07, + "loss": 0.9561, + "step": 73574 + }, + { + "epoch": 5.44, + "learning_rate": 4.5871144179766903e-07, + "loss": 0.859, + "step": 73575 + }, + { + "epoch": 5.44, + "learning_rate": 4.5859202871248455e-07, + "loss": 1.0006, + "step": 73576 + }, + { + "epoch": 5.44, + "learning_rate": 4.584726308075438e-07, + "loss": 1.0172, + "step": 73577 + }, + { + "epoch": 5.44, + "learning_rate": 4.583532480830344e-07, + "loss": 0.9275, + "step": 73578 + }, + { + "epoch": 5.44, + "learning_rate": 4.5823388053915065e-07, + "loss": 0.9829, + "step": 73579 + }, + { + "epoch": 5.44, + "learning_rate": 4.5811452817608124e-07, + "loss": 0.8942, + "step": 73580 + }, + { + "epoch": 5.44, + "learning_rate": 4.5799519099401277e-07, + "loss": 1.1028, + "step": 73581 + }, + { + "epoch": 5.44, + "learning_rate": 4.578758689931384e-07, + "loss": 1.105, + "step": 73582 + }, + { + "epoch": 5.44, + "learning_rate": 4.5775656217364574e-07, + "loss": 0.9868, + "step": 73583 + }, + { + "epoch": 5.44, + "learning_rate": 4.5763727053572903e-07, + "loss": 1.0771, + "step": 73584 + }, + { + "epoch": 5.44, + "learning_rate": 4.5751799407957156e-07, + "loss": 1.0191, + "step": 73585 + }, + { + "epoch": 5.44, + "learning_rate": 4.573987328053686e-07, + "loss": 1.0779, + "step": 73586 + }, + { + "epoch": 5.44, + "learning_rate": 4.572794867133057e-07, + "loss": 0.8957, + "step": 73587 + }, + { + "epoch": 5.44, + "learning_rate": 4.5716025580357706e-07, + "loss": 1.0271, + "step": 73588 + }, + { + "epoch": 5.44, + "learning_rate": 4.5704104007636805e-07, + "loss": 1.0638, + "step": 73589 + }, + { + "epoch": 5.44, + "learning_rate": 4.569218395318709e-07, + "loss": 1.0022, + "step": 73590 + }, + { + "epoch": 5.44, + "learning_rate": 4.568026541702741e-07, + "loss": 0.8941, + "step": 73591 + }, + { + "epoch": 5.44, + "learning_rate": 4.566834839917689e-07, + "loss": 0.9764, + "step": 73592 + }, + { + "epoch": 5.44, + "learning_rate": 4.5656432899654266e-07, + "loss": 0.9978, + "step": 73593 + }, + { + "epoch": 5.44, + "learning_rate": 4.5644518918478543e-07, + "loss": 1.0281, + "step": 73594 + }, + { + "epoch": 5.44, + "learning_rate": 4.5632606455668914e-07, + "loss": 1.0825, + "step": 73595 + }, + { + "epoch": 5.44, + "learning_rate": 4.562069551124404e-07, + "loss": 1.0383, + "step": 73596 + }, + { + "epoch": 5.44, + "learning_rate": 4.560878608522301e-07, + "loss": 0.9508, + "step": 73597 + }, + { + "epoch": 5.44, + "learning_rate": 4.5596878177624704e-07, + "loss": 0.9902, + "step": 73598 + }, + { + "epoch": 5.44, + "learning_rate": 4.55849717884681e-07, + "loss": 1.0668, + "step": 73599 + }, + { + "epoch": 5.44, + "learning_rate": 4.5573066917772303e-07, + "loss": 1.0107, + "step": 73600 + }, + { + "epoch": 5.44, + "learning_rate": 4.5561163565555955e-07, + "loss": 1.0269, + "step": 73601 + }, + { + "epoch": 5.44, + "learning_rate": 4.5549261731838267e-07, + "loss": 0.9954, + "step": 73602 + }, + { + "epoch": 5.44, + "learning_rate": 4.553736141663778e-07, + "loss": 1.0632, + "step": 73603 + }, + { + "epoch": 5.44, + "learning_rate": 4.5525462619973927e-07, + "loss": 1.0932, + "step": 73604 + }, + { + "epoch": 5.44, + "learning_rate": 4.551356534186546e-07, + "loss": 0.9327, + "step": 73605 + }, + { + "epoch": 5.44, + "learning_rate": 4.5501669582331156e-07, + "loss": 0.9898, + "step": 73606 + }, + { + "epoch": 5.44, + "learning_rate": 4.5489775341389876e-07, + "loss": 1.1283, + "step": 73607 + }, + { + "epoch": 5.44, + "learning_rate": 4.5477882619061055e-07, + "loss": 1.0819, + "step": 73608 + }, + { + "epoch": 5.44, + "learning_rate": 4.546599141536312e-07, + "loss": 1.0381, + "step": 73609 + }, + { + "epoch": 5.44, + "learning_rate": 4.5454101730314947e-07, + "loss": 1.0075, + "step": 73610 + }, + { + "epoch": 5.44, + "learning_rate": 4.5442213563935855e-07, + "loss": 1.0246, + "step": 73611 + }, + { + "epoch": 5.44, + "learning_rate": 4.543032691624449e-07, + "loss": 1.0382, + "step": 73612 + }, + { + "epoch": 5.44, + "learning_rate": 4.5418441787259846e-07, + "loss": 0.9127, + "step": 73613 + }, + { + "epoch": 5.44, + "learning_rate": 4.540655817700068e-07, + "loss": 1.0238, + "step": 73614 + }, + { + "epoch": 5.44, + "learning_rate": 4.5394676085486087e-07, + "loss": 0.9295, + "step": 73615 + }, + { + "epoch": 5.44, + "learning_rate": 4.5382795512734947e-07, + "loss": 0.9893, + "step": 73616 + }, + { + "epoch": 5.44, + "learning_rate": 4.5370916458766123e-07, + "loss": 0.9684, + "step": 73617 + }, + { + "epoch": 5.44, + "learning_rate": 4.53590389235985e-07, + "loss": 0.9373, + "step": 73618 + }, + { + "epoch": 5.44, + "learning_rate": 4.534716290725083e-07, + "loss": 1.0311, + "step": 73619 + }, + { + "epoch": 5.44, + "learning_rate": 4.533528840974233e-07, + "loss": 0.8761, + "step": 73620 + }, + { + "epoch": 5.44, + "learning_rate": 4.5323415431091755e-07, + "loss": 0.9819, + "step": 73621 + }, + { + "epoch": 5.44, + "learning_rate": 4.531154397131798e-07, + "loss": 1.0377, + "step": 73622 + }, + { + "epoch": 5.44, + "learning_rate": 4.5299674030439667e-07, + "loss": 1.078, + "step": 73623 + }, + { + "epoch": 5.44, + "learning_rate": 4.528780560847612e-07, + "loss": 0.9983, + "step": 73624 + }, + { + "epoch": 5.44, + "learning_rate": 4.5275938705445997e-07, + "loss": 0.9106, + "step": 73625 + }, + { + "epoch": 5.44, + "learning_rate": 4.5264073321368176e-07, + "loss": 0.98, + "step": 73626 + }, + { + "epoch": 5.44, + "learning_rate": 4.525220945626152e-07, + "loss": 1.052, + "step": 73627 + }, + { + "epoch": 5.44, + "learning_rate": 4.5240347110144913e-07, + "loss": 0.966, + "step": 73628 + }, + { + "epoch": 5.44, + "learning_rate": 4.5228486283037444e-07, + "loss": 1.0125, + "step": 73629 + }, + { + "epoch": 5.44, + "learning_rate": 4.521662697495754e-07, + "loss": 1.0548, + "step": 73630 + }, + { + "epoch": 5.44, + "learning_rate": 4.5204769185924427e-07, + "loss": 1.0526, + "step": 73631 + }, + { + "epoch": 5.44, + "learning_rate": 4.519291291595673e-07, + "loss": 0.9525, + "step": 73632 + }, + { + "epoch": 5.44, + "learning_rate": 4.518105816507379e-07, + "loss": 1.0029, + "step": 73633 + }, + { + "epoch": 5.44, + "learning_rate": 4.5169204933293796e-07, + "loss": 1.025, + "step": 73634 + }, + { + "epoch": 5.44, + "learning_rate": 4.515735322063608e-07, + "loss": 0.9916, + "step": 73635 + }, + { + "epoch": 5.44, + "learning_rate": 4.5145503027119286e-07, + "loss": 0.9628, + "step": 73636 + }, + { + "epoch": 5.44, + "learning_rate": 4.51336543527624e-07, + "loss": 1.0488, + "step": 73637 + }, + { + "epoch": 5.44, + "learning_rate": 4.512180719758419e-07, + "loss": 0.9078, + "step": 73638 + }, + { + "epoch": 5.44, + "learning_rate": 4.510996156160341e-07, + "loss": 0.9156, + "step": 73639 + }, + { + "epoch": 5.44, + "learning_rate": 4.509811744483916e-07, + "loss": 1.0573, + "step": 73640 + }, + { + "epoch": 5.44, + "learning_rate": 4.5086274847310206e-07, + "loss": 0.904, + "step": 73641 + }, + { + "epoch": 5.44, + "learning_rate": 4.5074433769035195e-07, + "loss": 0.976, + "step": 73642 + }, + { + "epoch": 5.44, + "learning_rate": 4.5062594210033005e-07, + "loss": 0.992, + "step": 73643 + }, + { + "epoch": 5.44, + "learning_rate": 4.505075617032273e-07, + "loss": 0.8902, + "step": 73644 + }, + { + "epoch": 5.44, + "learning_rate": 4.503891964992302e-07, + "loss": 0.9137, + "step": 73645 + }, + { + "epoch": 5.44, + "learning_rate": 4.502708464885275e-07, + "loss": 1.0547, + "step": 73646 + }, + { + "epoch": 5.44, + "learning_rate": 4.501525116713068e-07, + "loss": 0.9655, + "step": 73647 + }, + { + "epoch": 5.44, + "learning_rate": 4.500341920477558e-07, + "loss": 1.0189, + "step": 73648 + }, + { + "epoch": 5.44, + "learning_rate": 4.499158876180654e-07, + "loss": 1.0035, + "step": 73649 + }, + { + "epoch": 5.44, + "learning_rate": 4.497975983824221e-07, + "loss": 0.93, + "step": 73650 + }, + { + "epoch": 5.44, + "learning_rate": 4.496793243410136e-07, + "loss": 1.0539, + "step": 73651 + }, + { + "epoch": 5.44, + "learning_rate": 4.495610654940274e-07, + "loss": 0.9924, + "step": 73652 + }, + { + "epoch": 5.44, + "learning_rate": 4.4944282184165577e-07, + "loss": 1.0717, + "step": 73653 + }, + { + "epoch": 5.44, + "learning_rate": 4.4932459338408173e-07, + "loss": 0.9268, + "step": 73654 + }, + { + "epoch": 5.44, + "learning_rate": 4.492063801214963e-07, + "loss": 1.0373, + "step": 73655 + }, + { + "epoch": 5.44, + "learning_rate": 4.490881820540871e-07, + "loss": 1.0049, + "step": 73656 + }, + { + "epoch": 5.44, + "learning_rate": 4.4896999918204175e-07, + "loss": 1.0992, + "step": 73657 + }, + { + "epoch": 5.44, + "learning_rate": 4.48851831505549e-07, + "loss": 0.9235, + "step": 73658 + }, + { + "epoch": 5.44, + "learning_rate": 4.4873367902479427e-07, + "loss": 0.9598, + "step": 73659 + }, + { + "epoch": 5.44, + "learning_rate": 4.486155417399696e-07, + "loss": 1.0347, + "step": 73660 + }, + { + "epoch": 5.44, + "learning_rate": 4.4849741965126037e-07, + "loss": 0.9341, + "step": 73661 + }, + { + "epoch": 5.44, + "learning_rate": 4.483793127588554e-07, + "loss": 0.9961, + "step": 73662 + }, + { + "epoch": 5.44, + "learning_rate": 4.482612210629411e-07, + "loss": 0.8991, + "step": 73663 + }, + { + "epoch": 5.44, + "learning_rate": 4.4814314456370854e-07, + "loss": 0.8769, + "step": 73664 + }, + { + "epoch": 5.44, + "learning_rate": 4.48025083261342e-07, + "loss": 1.0469, + "step": 73665 + }, + { + "epoch": 5.44, + "learning_rate": 4.4790703715603235e-07, + "loss": 0.9551, + "step": 73666 + }, + { + "epoch": 5.44, + "learning_rate": 4.4778900624796504e-07, + "loss": 1.0392, + "step": 73667 + }, + { + "epoch": 5.44, + "learning_rate": 4.4767099053732777e-07, + "loss": 1.0007, + "step": 73668 + }, + { + "epoch": 5.44, + "learning_rate": 4.475529900243103e-07, + "loss": 0.9677, + "step": 73669 + }, + { + "epoch": 5.44, + "learning_rate": 4.4743500470909916e-07, + "loss": 1.012, + "step": 73670 + }, + { + "epoch": 5.44, + "learning_rate": 4.473170345918831e-07, + "loss": 1.0241, + "step": 73671 + }, + { + "epoch": 5.44, + "learning_rate": 4.4719907967284647e-07, + "loss": 0.9318, + "step": 73672 + }, + { + "epoch": 5.44, + "learning_rate": 4.470811399521813e-07, + "loss": 0.9745, + "step": 73673 + }, + { + "epoch": 5.44, + "learning_rate": 4.4696321543007514e-07, + "loss": 0.9906, + "step": 73674 + }, + { + "epoch": 5.44, + "learning_rate": 4.468453061067113e-07, + "loss": 1.0505, + "step": 73675 + }, + { + "epoch": 5.44, + "learning_rate": 4.4672741198228067e-07, + "loss": 0.9517, + "step": 73676 + }, + { + "epoch": 5.44, + "learning_rate": 4.4660953305696863e-07, + "loss": 0.9949, + "step": 73677 + }, + { + "epoch": 5.44, + "learning_rate": 4.464916693309662e-07, + "loss": 1.0056, + "step": 73678 + }, + { + "epoch": 5.44, + "learning_rate": 4.4637382080445766e-07, + "loss": 0.9686, + "step": 73679 + }, + { + "epoch": 5.44, + "learning_rate": 4.4625598747763174e-07, + "loss": 0.9168, + "step": 73680 + }, + { + "epoch": 5.44, + "learning_rate": 4.46138169350675e-07, + "loss": 1.0989, + "step": 73681 + }, + { + "epoch": 5.44, + "learning_rate": 4.4602036642377834e-07, + "loss": 1.108, + "step": 73682 + }, + { + "epoch": 5.44, + "learning_rate": 4.4590257869712603e-07, + "loss": 1.0248, + "step": 73683 + }, + { + "epoch": 5.44, + "learning_rate": 4.457848061709036e-07, + "loss": 1.0452, + "step": 73684 + }, + { + "epoch": 5.44, + "learning_rate": 4.4566704884530297e-07, + "loss": 1.1273, + "step": 73685 + }, + { + "epoch": 5.44, + "learning_rate": 4.4554930672050855e-07, + "loss": 0.9707, + "step": 73686 + }, + { + "epoch": 5.44, + "learning_rate": 4.45431579796709e-07, + "loss": 0.91, + "step": 73687 + }, + { + "epoch": 5.44, + "learning_rate": 4.453138680740898e-07, + "loss": 1.0333, + "step": 73688 + }, + { + "epoch": 5.44, + "learning_rate": 4.451961715528408e-07, + "loss": 1.0436, + "step": 73689 + }, + { + "epoch": 5.44, + "learning_rate": 4.450784902331484e-07, + "loss": 1.0348, + "step": 73690 + }, + { + "epoch": 5.44, + "learning_rate": 4.4496082411519924e-07, + "loss": 0.979, + "step": 73691 + }, + { + "epoch": 5.45, + "learning_rate": 4.448431731991798e-07, + "loss": 0.9564, + "step": 73692 + }, + { + "epoch": 5.45, + "learning_rate": 4.447255374852777e-07, + "loss": 0.9625, + "step": 73693 + }, + { + "epoch": 5.45, + "learning_rate": 4.4460791697368276e-07, + "loss": 1.0131, + "step": 73694 + }, + { + "epoch": 5.45, + "learning_rate": 4.444903116645771e-07, + "loss": 0.9789, + "step": 73695 + }, + { + "epoch": 5.45, + "learning_rate": 4.443727215581528e-07, + "loss": 0.9432, + "step": 73696 + }, + { + "epoch": 5.45, + "learning_rate": 4.44255146654593e-07, + "loss": 1.0218, + "step": 73697 + }, + { + "epoch": 5.45, + "learning_rate": 4.441375869540887e-07, + "loss": 0.9241, + "step": 73698 + }, + { + "epoch": 5.45, + "learning_rate": 4.4402004245682305e-07, + "loss": 1.0102, + "step": 73699 + }, + { + "epoch": 5.45, + "learning_rate": 4.4390251316298593e-07, + "loss": 1.0378, + "step": 73700 + }, + { + "epoch": 5.45, + "learning_rate": 4.437849990727616e-07, + "loss": 0.8945, + "step": 73701 + }, + { + "epoch": 5.45, + "learning_rate": 4.436675001863411e-07, + "loss": 1.0115, + "step": 73702 + }, + { + "epoch": 5.45, + "learning_rate": 4.435500165039075e-07, + "loss": 1.0103, + "step": 73703 + }, + { + "epoch": 5.45, + "learning_rate": 4.434325480256485e-07, + "loss": 0.9456, + "step": 73704 + }, + { + "epoch": 5.45, + "learning_rate": 4.433150947517517e-07, + "loss": 1.1494, + "step": 73705 + }, + { + "epoch": 5.45, + "learning_rate": 4.4319765668240475e-07, + "loss": 0.948, + "step": 73706 + }, + { + "epoch": 5.45, + "learning_rate": 4.4308023381779307e-07, + "loss": 0.9857, + "step": 73707 + }, + { + "epoch": 5.45, + "learning_rate": 4.4296282615810317e-07, + "loss": 0.9473, + "step": 73708 + }, + { + "epoch": 5.45, + "learning_rate": 4.4284543370352373e-07, + "loss": 0.9004, + "step": 73709 + }, + { + "epoch": 5.45, + "learning_rate": 4.427280564542413e-07, + "loss": 0.9101, + "step": 73710 + }, + { + "epoch": 5.45, + "learning_rate": 4.4261069441044024e-07, + "loss": 1.1326, + "step": 73711 + }, + { + "epoch": 5.45, + "learning_rate": 4.424933475723103e-07, + "loss": 0.9915, + "step": 73712 + }, + { + "epoch": 5.45, + "learning_rate": 4.4237601594003475e-07, + "loss": 1.0451, + "step": 73713 + }, + { + "epoch": 5.45, + "learning_rate": 4.422586995138034e-07, + "loss": 1.0053, + "step": 73714 + }, + { + "epoch": 5.45, + "learning_rate": 4.4214139829380164e-07, + "loss": 0.9223, + "step": 73715 + }, + { + "epoch": 5.45, + "learning_rate": 4.4202411228021605e-07, + "loss": 1.0064, + "step": 73716 + }, + { + "epoch": 5.45, + "learning_rate": 4.4190684147323195e-07, + "loss": 1.0703, + "step": 73717 + }, + { + "epoch": 5.45, + "learning_rate": 4.417895858730392e-07, + "loss": 1.0078, + "step": 73718 + }, + { + "epoch": 5.45, + "learning_rate": 4.416723454798222e-07, + "loss": 0.912, + "step": 73719 + }, + { + "epoch": 5.45, + "learning_rate": 4.4155512029376847e-07, + "loss": 1.0455, + "step": 73720 + }, + { + "epoch": 5.45, + "learning_rate": 4.414379103150623e-07, + "loss": 0.949, + "step": 73721 + }, + { + "epoch": 5.45, + "learning_rate": 4.4132071554389143e-07, + "loss": 0.9208, + "step": 73722 + }, + { + "epoch": 5.45, + "learning_rate": 4.412035359804445e-07, + "loss": 1.0445, + "step": 73723 + }, + { + "epoch": 5.45, + "learning_rate": 4.4108637162490364e-07, + "loss": 0.9859, + "step": 73724 + }, + { + "epoch": 5.45, + "learning_rate": 4.4096922247745976e-07, + "loss": 1.0431, + "step": 73725 + }, + { + "epoch": 5.45, + "learning_rate": 4.40852088538295e-07, + "loss": 0.9852, + "step": 73726 + }, + { + "epoch": 5.45, + "learning_rate": 4.407349698076002e-07, + "loss": 1.0623, + "step": 73727 + }, + { + "epoch": 5.45, + "learning_rate": 4.406178662855565e-07, + "loss": 1.098, + "step": 73728 + }, + { + "epoch": 5.45, + "learning_rate": 4.4050077797235467e-07, + "loss": 1.0197, + "step": 73729 + }, + { + "epoch": 5.45, + "learning_rate": 4.4038370486818026e-07, + "loss": 0.8967, + "step": 73730 + }, + { + "epoch": 5.45, + "learning_rate": 4.4026664697321753e-07, + "loss": 0.9523, + "step": 73731 + }, + { + "epoch": 5.45, + "learning_rate": 4.4014960428765406e-07, + "loss": 1.096, + "step": 73732 + }, + { + "epoch": 5.45, + "learning_rate": 4.400325768116742e-07, + "loss": 1.0275, + "step": 73733 + }, + { + "epoch": 5.45, + "learning_rate": 4.399155645454667e-07, + "loss": 1.05, + "step": 73734 + }, + { + "epoch": 5.45, + "learning_rate": 4.39798567489218e-07, + "loss": 0.9647, + "step": 73735 + }, + { + "epoch": 5.45, + "learning_rate": 4.3968158564311136e-07, + "loss": 0.9909, + "step": 73736 + }, + { + "epoch": 5.45, + "learning_rate": 4.395646190073344e-07, + "loss": 1.0211, + "step": 73737 + }, + { + "epoch": 5.45, + "learning_rate": 4.394476675820747e-07, + "loss": 0.9415, + "step": 73738 + }, + { + "epoch": 5.45, + "learning_rate": 4.393307313675177e-07, + "loss": 1.157, + "step": 73739 + }, + { + "epoch": 5.45, + "learning_rate": 4.392138103638466e-07, + "loss": 1.0027, + "step": 73740 + }, + { + "epoch": 5.45, + "learning_rate": 4.390969045712501e-07, + "loss": 0.8588, + "step": 73741 + }, + { + "epoch": 5.45, + "learning_rate": 4.389800139899136e-07, + "loss": 0.9593, + "step": 73742 + }, + { + "epoch": 5.45, + "learning_rate": 4.3886313862002484e-07, + "loss": 1.0549, + "step": 73743 + }, + { + "epoch": 5.45, + "learning_rate": 4.3874627846176576e-07, + "loss": 0.9352, + "step": 73744 + }, + { + "epoch": 5.45, + "learning_rate": 4.3862943351532517e-07, + "loss": 1.0165, + "step": 73745 + }, + { + "epoch": 5.45, + "learning_rate": 4.3851260378088734e-07, + "loss": 0.9672, + "step": 73746 + }, + { + "epoch": 5.45, + "learning_rate": 4.383957892586421e-07, + "loss": 1.0105, + "step": 73747 + }, + { + "epoch": 5.45, + "learning_rate": 4.382789899487694e-07, + "loss": 1.1257, + "step": 73748 + }, + { + "epoch": 5.45, + "learning_rate": 4.381622058514601e-07, + "loss": 1.0826, + "step": 73749 + }, + { + "epoch": 5.45, + "learning_rate": 4.3804543696689737e-07, + "loss": 1.1016, + "step": 73750 + }, + { + "epoch": 5.45, + "learning_rate": 4.3792868329526673e-07, + "loss": 1.1105, + "step": 73751 + }, + { + "epoch": 5.45, + "learning_rate": 4.378119448367557e-07, + "loss": 0.8416, + "step": 73752 + }, + { + "epoch": 5.45, + "learning_rate": 4.376952215915475e-07, + "loss": 0.9668, + "step": 73753 + }, + { + "epoch": 5.45, + "learning_rate": 4.3757851355983095e-07, + "loss": 0.9435, + "step": 73754 + }, + { + "epoch": 5.45, + "learning_rate": 4.374618207417902e-07, + "loss": 1.0857, + "step": 73755 + }, + { + "epoch": 5.45, + "learning_rate": 4.3734514313761076e-07, + "loss": 0.9841, + "step": 73756 + }, + { + "epoch": 5.45, + "learning_rate": 4.3722848074747693e-07, + "loss": 0.9314, + "step": 73757 + }, + { + "epoch": 5.45, + "learning_rate": 4.3711183357157625e-07, + "loss": 0.9491, + "step": 73758 + }, + { + "epoch": 5.45, + "learning_rate": 4.3699520161009535e-07, + "loss": 0.9021, + "step": 73759 + }, + { + "epoch": 5.45, + "learning_rate": 4.3687858486321734e-07, + "loss": 1.0995, + "step": 73760 + }, + { + "epoch": 5.45, + "learning_rate": 4.3676198333112873e-07, + "loss": 1.0459, + "step": 73761 + }, + { + "epoch": 5.45, + "learning_rate": 4.3664539701401386e-07, + "loss": 1.0311, + "step": 73762 + }, + { + "epoch": 5.45, + "learning_rate": 4.365288259120604e-07, + "loss": 1.0888, + "step": 73763 + }, + { + "epoch": 5.45, + "learning_rate": 4.364122700254536e-07, + "loss": 0.9431, + "step": 73764 + }, + { + "epoch": 5.45, + "learning_rate": 4.362957293543768e-07, + "loss": 0.9085, + "step": 73765 + }, + { + "epoch": 5.45, + "learning_rate": 4.3617920389901647e-07, + "loss": 1.0102, + "step": 73766 + }, + { + "epoch": 5.45, + "learning_rate": 4.360626936595591e-07, + "loss": 0.9777, + "step": 73767 + }, + { + "epoch": 5.45, + "learning_rate": 4.3594619863619015e-07, + "loss": 1.009, + "step": 73768 + }, + { + "epoch": 5.45, + "learning_rate": 4.3582971882909167e-07, + "loss": 1.0344, + "step": 73769 + }, + { + "epoch": 5.45, + "learning_rate": 4.3571325423845235e-07, + "loss": 1.0417, + "step": 73770 + }, + { + "epoch": 5.45, + "learning_rate": 4.355968048644554e-07, + "loss": 0.9899, + "step": 73771 + }, + { + "epoch": 5.45, + "learning_rate": 4.3548037070728966e-07, + "loss": 1.0415, + "step": 73772 + }, + { + "epoch": 5.45, + "learning_rate": 4.3536395176713486e-07, + "loss": 1.0095, + "step": 73773 + }, + { + "epoch": 5.45, + "learning_rate": 4.352475480441809e-07, + "loss": 0.9012, + "step": 73774 + }, + { + "epoch": 5.45, + "learning_rate": 4.3513115953861097e-07, + "loss": 1.0398, + "step": 73775 + }, + { + "epoch": 5.45, + "learning_rate": 4.3501478625061045e-07, + "loss": 1.0961, + "step": 73776 + }, + { + "epoch": 5.45, + "learning_rate": 4.348984281803648e-07, + "loss": 0.9825, + "step": 73777 + }, + { + "epoch": 5.45, + "learning_rate": 4.347820853280571e-07, + "loss": 1.0055, + "step": 73778 + }, + { + "epoch": 5.45, + "learning_rate": 4.346657576938762e-07, + "loss": 1.0604, + "step": 73779 + }, + { + "epoch": 5.45, + "learning_rate": 4.3454944527800414e-07, + "loss": 1.0209, + "step": 73780 + }, + { + "epoch": 5.45, + "learning_rate": 4.3443314808062853e-07, + "loss": 0.9015, + "step": 73781 + }, + { + "epoch": 5.45, + "learning_rate": 4.3431686610193036e-07, + "loss": 1.0347, + "step": 73782 + }, + { + "epoch": 5.45, + "learning_rate": 4.3420059934209947e-07, + "loss": 0.9104, + "step": 73783 + }, + { + "epoch": 5.45, + "learning_rate": 4.340843478013179e-07, + "loss": 1.0246, + "step": 73784 + }, + { + "epoch": 5.45, + "learning_rate": 4.3396811147977115e-07, + "loss": 1.0488, + "step": 73785 + }, + { + "epoch": 5.45, + "learning_rate": 4.338518903776445e-07, + "loss": 0.8416, + "step": 73786 + }, + { + "epoch": 5.45, + "learning_rate": 4.337356844951213e-07, + "loss": 0.9817, + "step": 73787 + }, + { + "epoch": 5.45, + "learning_rate": 4.336194938323912e-07, + "loss": 1.0537, + "step": 73788 + }, + { + "epoch": 5.45, + "learning_rate": 4.33503318389632e-07, + "loss": 0.9079, + "step": 73789 + }, + { + "epoch": 5.45, + "learning_rate": 4.3338715816703347e-07, + "loss": 1.148, + "step": 73790 + }, + { + "epoch": 5.45, + "learning_rate": 4.332710131647788e-07, + "loss": 1.0502, + "step": 73791 + }, + { + "epoch": 5.45, + "learning_rate": 4.3315488338305457e-07, + "loss": 1.0609, + "step": 73792 + }, + { + "epoch": 5.45, + "learning_rate": 4.3303876882204166e-07, + "loss": 0.9031, + "step": 73793 + }, + { + "epoch": 5.45, + "learning_rate": 4.3292266948192885e-07, + "loss": 0.8921, + "step": 73794 + }, + { + "epoch": 5.45, + "learning_rate": 4.328065853628993e-07, + "loss": 1.0384, + "step": 73795 + }, + { + "epoch": 5.45, + "learning_rate": 4.3269051646513735e-07, + "loss": 0.9818, + "step": 73796 + }, + { + "epoch": 5.45, + "learning_rate": 4.3257446278882733e-07, + "loss": 1.0444, + "step": 73797 + }, + { + "epoch": 5.45, + "learning_rate": 4.3245842433415344e-07, + "loss": 0.968, + "step": 73798 + }, + { + "epoch": 5.45, + "learning_rate": 4.3234240110130333e-07, + "loss": 1.0753, + "step": 73799 + }, + { + "epoch": 5.45, + "learning_rate": 4.322263930904591e-07, + "loss": 1.1338, + "step": 73800 + }, + { + "epoch": 5.45, + "learning_rate": 4.321104003018051e-07, + "loss": 1.0294, + "step": 73801 + }, + { + "epoch": 5.45, + "learning_rate": 4.319944227355266e-07, + "loss": 0.9659, + "step": 73802 + }, + { + "epoch": 5.45, + "learning_rate": 4.31878460391808e-07, + "loss": 1.036, + "step": 73803 + }, + { + "epoch": 5.45, + "learning_rate": 4.317625132708347e-07, + "loss": 0.9326, + "step": 73804 + }, + { + "epoch": 5.45, + "learning_rate": 4.316465813727899e-07, + "loss": 1.0271, + "step": 73805 + }, + { + "epoch": 5.45, + "learning_rate": 4.31530664697859e-07, + "loss": 0.9965, + "step": 73806 + }, + { + "epoch": 5.45, + "learning_rate": 4.314147632462251e-07, + "loss": 1.0178, + "step": 73807 + }, + { + "epoch": 5.45, + "learning_rate": 4.3129887701807373e-07, + "loss": 1.0566, + "step": 73808 + }, + { + "epoch": 5.45, + "learning_rate": 4.311830060135891e-07, + "loss": 1.0218, + "step": 73809 + }, + { + "epoch": 5.45, + "learning_rate": 4.3106715023295555e-07, + "loss": 1.0536, + "step": 73810 + }, + { + "epoch": 5.45, + "learning_rate": 4.3095130967635625e-07, + "loss": 1.0665, + "step": 73811 + }, + { + "epoch": 5.45, + "learning_rate": 4.308354843439766e-07, + "loss": 0.9627, + "step": 73812 + }, + { + "epoch": 5.45, + "learning_rate": 4.3071967423600203e-07, + "loss": 1.0258, + "step": 73813 + }, + { + "epoch": 5.45, + "learning_rate": 4.3060387935261573e-07, + "loss": 1.0688, + "step": 73814 + }, + { + "epoch": 5.45, + "learning_rate": 4.3048809969400085e-07, + "loss": 1.0572, + "step": 73815 + }, + { + "epoch": 5.45, + "learning_rate": 4.3037233526034284e-07, + "loss": 1.0431, + "step": 73816 + }, + { + "epoch": 5.45, + "learning_rate": 4.3025658605182486e-07, + "loss": 1.0348, + "step": 73817 + }, + { + "epoch": 5.45, + "learning_rate": 4.301408520686312e-07, + "loss": 0.9439, + "step": 73818 + }, + { + "epoch": 5.45, + "learning_rate": 4.300251333109484e-07, + "loss": 1.0403, + "step": 73819 + }, + { + "epoch": 5.45, + "learning_rate": 4.2990942977895745e-07, + "loss": 1.0347, + "step": 73820 + }, + { + "epoch": 5.45, + "learning_rate": 4.297937414728448e-07, + "loss": 1.0686, + "step": 73821 + }, + { + "epoch": 5.45, + "learning_rate": 4.2967806839279146e-07, + "loss": 1.0805, + "step": 73822 + }, + { + "epoch": 5.45, + "learning_rate": 4.2956241053898395e-07, + "loss": 0.9312, + "step": 73823 + }, + { + "epoch": 5.45, + "learning_rate": 4.2944676791160655e-07, + "loss": 1.0045, + "step": 73824 + }, + { + "epoch": 5.45, + "learning_rate": 4.2933114051084245e-07, + "loss": 1.0184, + "step": 73825 + }, + { + "epoch": 5.45, + "learning_rate": 4.2921552833687597e-07, + "loss": 1.0453, + "step": 73826 + }, + { + "epoch": 5.46, + "learning_rate": 4.2909993138988806e-07, + "loss": 0.928, + "step": 73827 + }, + { + "epoch": 5.46, + "learning_rate": 4.289843496700674e-07, + "loss": 1.0263, + "step": 73828 + }, + { + "epoch": 5.46, + "learning_rate": 4.2886878317759616e-07, + "loss": 0.9651, + "step": 73829 + }, + { + "epoch": 5.46, + "learning_rate": 4.287532319126575e-07, + "loss": 1.1618, + "step": 73830 + }, + { + "epoch": 5.46, + "learning_rate": 4.286376958754335e-07, + "loss": 1.0653, + "step": 73831 + }, + { + "epoch": 5.46, + "learning_rate": 4.285221750661117e-07, + "loss": 0.9895, + "step": 73832 + }, + { + "epoch": 5.46, + "learning_rate": 4.2840666948487543e-07, + "loss": 0.9996, + "step": 73833 + }, + { + "epoch": 5.46, + "learning_rate": 4.2829117913190444e-07, + "loss": 1.0094, + "step": 73834 + }, + { + "epoch": 5.46, + "learning_rate": 4.281757040073864e-07, + "loss": 0.9655, + "step": 73835 + }, + { + "epoch": 5.46, + "learning_rate": 4.280602441115034e-07, + "loss": 0.9575, + "step": 73836 + }, + { + "epoch": 5.46, + "learning_rate": 4.279447994444419e-07, + "loss": 0.977, + "step": 73837 + }, + { + "epoch": 5.46, + "learning_rate": 4.278293700063807e-07, + "loss": 1.0012, + "step": 73838 + }, + { + "epoch": 5.46, + "learning_rate": 4.277139557975063e-07, + "loss": 1.1293, + "step": 73839 + }, + { + "epoch": 5.46, + "learning_rate": 4.275985568180019e-07, + "loss": 0.8843, + "step": 73840 + }, + { + "epoch": 5.46, + "learning_rate": 4.274831730680529e-07, + "loss": 0.9366, + "step": 73841 + }, + { + "epoch": 5.46, + "learning_rate": 4.2736780454784135e-07, + "loss": 0.987, + "step": 73842 + }, + { + "epoch": 5.46, + "learning_rate": 4.272524512575482e-07, + "loss": 0.9614, + "step": 73843 + }, + { + "epoch": 5.46, + "learning_rate": 4.2713711319736117e-07, + "loss": 1.1025, + "step": 73844 + }, + { + "epoch": 5.46, + "learning_rate": 4.2702179036746116e-07, + "loss": 1.0493, + "step": 73845 + }, + { + "epoch": 5.46, + "learning_rate": 4.2690648276803357e-07, + "loss": 0.9384, + "step": 73846 + }, + { + "epoch": 5.46, + "learning_rate": 4.267911903992594e-07, + "loss": 1.0818, + "step": 73847 + }, + { + "epoch": 5.46, + "learning_rate": 4.2667591326132407e-07, + "loss": 1.0086, + "step": 73848 + }, + { + "epoch": 5.46, + "learning_rate": 4.265606513544107e-07, + "loss": 1.0434, + "step": 73849 + }, + { + "epoch": 5.46, + "learning_rate": 4.2644540467870253e-07, + "loss": 1.0231, + "step": 73850 + }, + { + "epoch": 5.46, + "learning_rate": 4.263301732343827e-07, + "loss": 1.0565, + "step": 73851 + }, + { + "epoch": 5.46, + "learning_rate": 4.262149570216334e-07, + "loss": 0.9663, + "step": 73852 + }, + { + "epoch": 5.46, + "learning_rate": 4.26099756040641e-07, + "loss": 1.0751, + "step": 73853 + }, + { + "epoch": 5.46, + "learning_rate": 4.2598457029158544e-07, + "loss": 0.9698, + "step": 73854 + }, + { + "epoch": 5.46, + "learning_rate": 4.2586939977465216e-07, + "loss": 0.9883, + "step": 73855 + }, + { + "epoch": 5.46, + "learning_rate": 4.257542444900231e-07, + "loss": 0.9083, + "step": 73856 + }, + { + "epoch": 5.46, + "learning_rate": 4.256391044378827e-07, + "loss": 0.9692, + "step": 73857 + }, + { + "epoch": 5.46, + "learning_rate": 4.25523979618413e-07, + "loss": 1.0212, + "step": 73858 + }, + { + "epoch": 5.46, + "learning_rate": 4.254088700317982e-07, + "loss": 1.0408, + "step": 73859 + }, + { + "epoch": 5.46, + "learning_rate": 4.252937756782194e-07, + "loss": 0.9262, + "step": 73860 + }, + { + "epoch": 5.46, + "learning_rate": 4.2517869655786414e-07, + "loss": 1.0215, + "step": 73861 + }, + { + "epoch": 5.46, + "learning_rate": 4.250636326709112e-07, + "loss": 1.0051, + "step": 73862 + }, + { + "epoch": 5.46, + "learning_rate": 4.249485840175438e-07, + "loss": 0.9826, + "step": 73863 + }, + { + "epoch": 5.46, + "learning_rate": 4.248335505979473e-07, + "loss": 0.9655, + "step": 73864 + }, + { + "epoch": 5.46, + "learning_rate": 4.2471853241230375e-07, + "loss": 1.0688, + "step": 73865 + }, + { + "epoch": 5.46, + "learning_rate": 4.246035294607964e-07, + "loss": 1.0423, + "step": 73866 + }, + { + "epoch": 5.46, + "learning_rate": 4.244885417436062e-07, + "loss": 1.0081, + "step": 73867 + }, + { + "epoch": 5.46, + "learning_rate": 4.2437356926091856e-07, + "loss": 0.9486, + "step": 73868 + }, + { + "epoch": 5.46, + "learning_rate": 4.242586120129155e-07, + "loss": 1.0998, + "step": 73869 + }, + { + "epoch": 5.46, + "learning_rate": 4.241436699997803e-07, + "loss": 1.1117, + "step": 73870 + }, + { + "epoch": 5.46, + "learning_rate": 4.2402874322169606e-07, + "loss": 1.0193, + "step": 73871 + }, + { + "epoch": 5.46, + "learning_rate": 4.239138316788427e-07, + "loss": 0.9661, + "step": 73872 + }, + { + "epoch": 5.46, + "learning_rate": 4.237989353714078e-07, + "loss": 1.0007, + "step": 73873 + }, + { + "epoch": 5.46, + "learning_rate": 4.236840542995713e-07, + "loss": 1.0487, + "step": 73874 + }, + { + "epoch": 5.46, + "learning_rate": 4.235691884635162e-07, + "loss": 1.1037, + "step": 73875 + }, + { + "epoch": 5.46, + "learning_rate": 4.2345433786342484e-07, + "loss": 1.0118, + "step": 73876 + }, + { + "epoch": 5.46, + "learning_rate": 4.233395024994813e-07, + "loss": 0.9398, + "step": 73877 + }, + { + "epoch": 5.46, + "learning_rate": 4.2322468237186773e-07, + "loss": 1.0706, + "step": 73878 + }, + { + "epoch": 5.46, + "learning_rate": 4.231098774807674e-07, + "loss": 0.9353, + "step": 73879 + }, + { + "epoch": 5.46, + "learning_rate": 4.2299508782636114e-07, + "loss": 1.0186, + "step": 73880 + }, + { + "epoch": 5.46, + "learning_rate": 4.2288031340883216e-07, + "loss": 0.9341, + "step": 73881 + }, + { + "epoch": 5.46, + "learning_rate": 4.2276555422836594e-07, + "loss": 1.019, + "step": 73882 + }, + { + "epoch": 5.46, + "learning_rate": 4.226508102851401e-07, + "loss": 1.0626, + "step": 73883 + }, + { + "epoch": 5.46, + "learning_rate": 4.225360815793411e-07, + "loss": 1.061, + "step": 73884 + }, + { + "epoch": 5.46, + "learning_rate": 4.2242136811114886e-07, + "loss": 0.9602, + "step": 73885 + }, + { + "epoch": 5.46, + "learning_rate": 4.2230666988074985e-07, + "loss": 0.9915, + "step": 73886 + }, + { + "epoch": 5.46, + "learning_rate": 4.221919868883206e-07, + "loss": 1.0114, + "step": 73887 + }, + { + "epoch": 5.46, + "learning_rate": 4.2207731913404883e-07, + "loss": 1.0446, + "step": 73888 + }, + { + "epoch": 5.46, + "learning_rate": 4.2196266661811534e-07, + "loss": 0.9854, + "step": 73889 + }, + { + "epoch": 5.46, + "learning_rate": 4.2184802934070125e-07, + "loss": 1.0399, + "step": 73890 + }, + { + "epoch": 5.46, + "learning_rate": 4.2173340730199073e-07, + "loss": 1.0242, + "step": 73891 + }, + { + "epoch": 5.46, + "learning_rate": 4.216188005021638e-07, + "loss": 1.1192, + "step": 73892 + }, + { + "epoch": 5.46, + "learning_rate": 4.215042089414056e-07, + "loss": 1.0428, + "step": 73893 + }, + { + "epoch": 5.46, + "learning_rate": 4.213896326198963e-07, + "loss": 0.9942, + "step": 73894 + }, + { + "epoch": 5.46, + "learning_rate": 4.2127507153782e-07, + "loss": 0.9764, + "step": 73895 + }, + { + "epoch": 5.46, + "learning_rate": 4.211605256953566e-07, + "loss": 0.948, + "step": 73896 + }, + { + "epoch": 5.46, + "learning_rate": 4.210459950926915e-07, + "loss": 0.9143, + "step": 73897 + }, + { + "epoch": 5.46, + "learning_rate": 4.209314797300046e-07, + "loss": 0.9901, + "step": 73898 + }, + { + "epoch": 5.46, + "learning_rate": 4.2081697960747793e-07, + "loss": 1.0769, + "step": 73899 + }, + { + "epoch": 5.46, + "learning_rate": 4.207024947252958e-07, + "loss": 1.0385, + "step": 73900 + }, + { + "epoch": 5.46, + "learning_rate": 4.2058802508363695e-07, + "loss": 0.8484, + "step": 73901 + }, + { + "epoch": 5.46, + "learning_rate": 4.204735706826868e-07, + "loss": 1.0694, + "step": 73902 + }, + { + "epoch": 5.46, + "learning_rate": 4.2035913152262633e-07, + "loss": 1.0364, + "step": 73903 + }, + { + "epoch": 5.46, + "learning_rate": 4.2024470760363754e-07, + "loss": 0.974, + "step": 73904 + }, + { + "epoch": 5.46, + "learning_rate": 4.201302989259004e-07, + "loss": 0.9192, + "step": 73905 + }, + { + "epoch": 5.46, + "learning_rate": 4.2001590548960245e-07, + "loss": 1.0373, + "step": 73906 + }, + { + "epoch": 5.46, + "learning_rate": 4.199015272949203e-07, + "loss": 1.0656, + "step": 73907 + }, + { + "epoch": 5.46, + "learning_rate": 4.1978716434203594e-07, + "loss": 0.9485, + "step": 73908 + }, + { + "epoch": 5.46, + "learning_rate": 4.196728166311348e-07, + "loss": 0.9183, + "step": 73909 + }, + { + "epoch": 5.46, + "learning_rate": 4.1955848416239785e-07, + "loss": 1.1037, + "step": 73910 + }, + { + "epoch": 5.46, + "learning_rate": 4.19444166936005e-07, + "loss": 0.9676, + "step": 73911 + }, + { + "epoch": 5.46, + "learning_rate": 4.1932986495213933e-07, + "loss": 0.8957, + "step": 73912 + }, + { + "epoch": 5.46, + "learning_rate": 4.19215578210983e-07, + "loss": 1.0168, + "step": 73913 + }, + { + "epoch": 5.46, + "learning_rate": 4.191013067127181e-07, + "loss": 0.9764, + "step": 73914 + }, + { + "epoch": 5.46, + "learning_rate": 4.189870504575266e-07, + "loss": 0.9545, + "step": 73915 + }, + { + "epoch": 5.46, + "learning_rate": 4.188728094455874e-07, + "loss": 0.9989, + "step": 73916 + }, + { + "epoch": 5.46, + "learning_rate": 4.1875858367708576e-07, + "loss": 1.0092, + "step": 73917 + }, + { + "epoch": 5.46, + "learning_rate": 4.1864437315220274e-07, + "loss": 1.013, + "step": 73918 + }, + { + "epoch": 5.46, + "learning_rate": 4.1853017787111927e-07, + "loss": 1.0432, + "step": 73919 + }, + { + "epoch": 5.46, + "learning_rate": 4.1841599783401743e-07, + "loss": 1.0744, + "step": 73920 + }, + { + "epoch": 5.46, + "learning_rate": 4.18301833041076e-07, + "loss": 1.0826, + "step": 73921 + }, + { + "epoch": 5.46, + "learning_rate": 4.181876834924825e-07, + "loss": 1.0734, + "step": 73922 + }, + { + "epoch": 5.46, + "learning_rate": 4.180735491884136e-07, + "loss": 0.9657, + "step": 73923 + }, + { + "epoch": 5.46, + "learning_rate": 4.179594301290535e-07, + "loss": 0.9799, + "step": 73924 + }, + { + "epoch": 5.46, + "learning_rate": 4.17845326314581e-07, + "loss": 1.0404, + "step": 73925 + }, + { + "epoch": 5.46, + "learning_rate": 4.1773123774518143e-07, + "loss": 0.9833, + "step": 73926 + }, + { + "epoch": 5.46, + "learning_rate": 4.176171644210358e-07, + "loss": 0.9619, + "step": 73927 + }, + { + "epoch": 5.46, + "learning_rate": 4.175031063423207e-07, + "loss": 0.9116, + "step": 73928 + }, + { + "epoch": 5.46, + "learning_rate": 4.173890635092226e-07, + "loss": 1.0636, + "step": 73929 + }, + { + "epoch": 5.46, + "learning_rate": 4.1727503592192133e-07, + "loss": 1.0183, + "step": 73930 + }, + { + "epoch": 5.46, + "learning_rate": 4.1716102358059896e-07, + "loss": 0.9372, + "step": 73931 + }, + { + "epoch": 5.46, + "learning_rate": 4.170470264854354e-07, + "loss": 0.9561, + "step": 73932 + }, + { + "epoch": 5.46, + "learning_rate": 4.1693304463661376e-07, + "loss": 1.0554, + "step": 73933 + }, + { + "epoch": 5.46, + "learning_rate": 4.1681907803431287e-07, + "loss": 0.9551, + "step": 73934 + }, + { + "epoch": 5.46, + "learning_rate": 4.167051266787181e-07, + "loss": 1.0731, + "step": 73935 + }, + { + "epoch": 5.46, + "learning_rate": 4.165911905700071e-07, + "loss": 1.0387, + "step": 73936 + }, + { + "epoch": 5.46, + "learning_rate": 4.164772697083619e-07, + "loss": 0.9445, + "step": 73937 + }, + { + "epoch": 5.46, + "learning_rate": 4.163633640939646e-07, + "loss": 0.832, + "step": 73938 + }, + { + "epoch": 5.46, + "learning_rate": 4.162494737269962e-07, + "loss": 1.0541, + "step": 73939 + }, + { + "epoch": 5.46, + "learning_rate": 4.1613559860763876e-07, + "loss": 1.0855, + "step": 73940 + }, + { + "epoch": 5.46, + "learning_rate": 4.160217387360699e-07, + "loss": 1.0328, + "step": 73941 + }, + { + "epoch": 5.46, + "learning_rate": 4.15907894112475e-07, + "loss": 1.0915, + "step": 73942 + }, + { + "epoch": 5.46, + "learning_rate": 4.15794064737034e-07, + "loss": 1.064, + "step": 73943 + }, + { + "epoch": 5.46, + "learning_rate": 4.1568025060992776e-07, + "loss": 0.9541, + "step": 73944 + }, + { + "epoch": 5.46, + "learning_rate": 4.1556645173133624e-07, + "loss": 1.0212, + "step": 73945 + }, + { + "epoch": 5.46, + "learning_rate": 4.1545266810144037e-07, + "loss": 0.9275, + "step": 73946 + }, + { + "epoch": 5.46, + "learning_rate": 4.1533889972042327e-07, + "loss": 0.9571, + "step": 73947 + }, + { + "epoch": 5.46, + "learning_rate": 4.1522514658846493e-07, + "loss": 0.9066, + "step": 73948 + }, + { + "epoch": 5.46, + "learning_rate": 4.151114087057473e-07, + "loss": 1.0307, + "step": 73949 + }, + { + "epoch": 5.46, + "learning_rate": 4.1499768607244807e-07, + "loss": 1.0291, + "step": 73950 + }, + { + "epoch": 5.46, + "learning_rate": 4.148839786887526e-07, + "loss": 0.9218, + "step": 73951 + }, + { + "epoch": 5.46, + "learning_rate": 4.1477028655483753e-07, + "loss": 0.9239, + "step": 73952 + }, + { + "epoch": 5.46, + "learning_rate": 4.1465660967088706e-07, + "loss": 0.8783, + "step": 73953 + }, + { + "epoch": 5.46, + "learning_rate": 4.1454294803707993e-07, + "loss": 0.9394, + "step": 73954 + }, + { + "epoch": 5.46, + "learning_rate": 4.144293016535994e-07, + "loss": 1.0666, + "step": 73955 + }, + { + "epoch": 5.46, + "learning_rate": 4.1431567052062307e-07, + "loss": 0.9341, + "step": 73956 + }, + { + "epoch": 5.46, + "learning_rate": 4.1420205463833295e-07, + "loss": 1.046, + "step": 73957 + }, + { + "epoch": 5.46, + "learning_rate": 4.1408845400691123e-07, + "loss": 1.035, + "step": 73958 + }, + { + "epoch": 5.46, + "learning_rate": 4.139748686265377e-07, + "loss": 1.0514, + "step": 73959 + }, + { + "epoch": 5.46, + "learning_rate": 4.1386129849739334e-07, + "loss": 1.0542, + "step": 73960 + }, + { + "epoch": 5.46, + "learning_rate": 4.137477436196569e-07, + "loss": 1.0803, + "step": 73961 + }, + { + "epoch": 5.47, + "learning_rate": 4.136342039935126e-07, + "loss": 1.0647, + "step": 73962 + }, + { + "epoch": 5.47, + "learning_rate": 4.1352067961913824e-07, + "loss": 1.0647, + "step": 73963 + }, + { + "epoch": 5.47, + "learning_rate": 4.1340717049671574e-07, + "loss": 0.9935, + "step": 73964 + }, + { + "epoch": 5.47, + "learning_rate": 4.132936766264251e-07, + "loss": 0.9935, + "step": 73965 + }, + { + "epoch": 5.47, + "learning_rate": 4.13180198008446e-07, + "loss": 1.0157, + "step": 73966 + }, + { + "epoch": 5.47, + "learning_rate": 4.130667346429607e-07, + "loss": 1.1467, + "step": 73967 + }, + { + "epoch": 5.47, + "learning_rate": 4.1295328653015e-07, + "loss": 1.0392, + "step": 73968 + }, + { + "epoch": 5.47, + "learning_rate": 4.128398536701927e-07, + "loss": 0.9759, + "step": 73969 + }, + { + "epoch": 5.47, + "learning_rate": 4.1272643606326876e-07, + "loss": 1.0087, + "step": 73970 + }, + { + "epoch": 5.47, + "learning_rate": 4.126130337095613e-07, + "loss": 1.0637, + "step": 73971 + }, + { + "epoch": 5.47, + "learning_rate": 4.12499646609249e-07, + "loss": 0.9149, + "step": 73972 + }, + { + "epoch": 5.47, + "learning_rate": 4.1238627476251403e-07, + "loss": 0.9927, + "step": 73973 + }, + { + "epoch": 5.47, + "learning_rate": 4.12272918169534e-07, + "loss": 1.1122, + "step": 73974 + }, + { + "epoch": 5.47, + "learning_rate": 4.121595768304898e-07, + "loss": 0.9941, + "step": 73975 + }, + { + "epoch": 5.47, + "learning_rate": 4.120462507455647e-07, + "loss": 1.0107, + "step": 73976 + }, + { + "epoch": 5.47, + "learning_rate": 4.119329399149341e-07, + "loss": 0.9256, + "step": 73977 + }, + { + "epoch": 5.47, + "learning_rate": 4.118196443387823e-07, + "loss": 0.9378, + "step": 73978 + }, + { + "epoch": 5.47, + "learning_rate": 4.11706364017288e-07, + "loss": 1.0213, + "step": 73979 + }, + { + "epoch": 5.47, + "learning_rate": 4.115930989506334e-07, + "loss": 1.0316, + "step": 73980 + }, + { + "epoch": 5.47, + "learning_rate": 4.1147984913899484e-07, + "loss": 0.9055, + "step": 73981 + }, + { + "epoch": 5.47, + "learning_rate": 4.1136661458255565e-07, + "loss": 1.0893, + "step": 73982 + }, + { + "epoch": 5.47, + "learning_rate": 4.112533952814945e-07, + "loss": 1.0149, + "step": 73983 + }, + { + "epoch": 5.47, + "learning_rate": 4.1114019123599245e-07, + "loss": 1.0591, + "step": 73984 + }, + { + "epoch": 5.47, + "learning_rate": 4.1102700244622926e-07, + "loss": 1.0186, + "step": 73985 + }, + { + "epoch": 5.47, + "learning_rate": 4.1091382891238374e-07, + "loss": 1.0115, + "step": 73986 + }, + { + "epoch": 5.47, + "learning_rate": 4.10800670634639e-07, + "loss": 0.8995, + "step": 73987 + }, + { + "epoch": 5.47, + "learning_rate": 4.1068752761317277e-07, + "loss": 0.9036, + "step": 73988 + }, + { + "epoch": 5.47, + "learning_rate": 4.105743998481648e-07, + "loss": 0.9953, + "step": 73989 + }, + { + "epoch": 5.47, + "learning_rate": 4.1046128733979504e-07, + "loss": 0.9675, + "step": 73990 + }, + { + "epoch": 5.47, + "learning_rate": 4.1034819008824444e-07, + "loss": 0.8753, + "step": 73991 + }, + { + "epoch": 5.47, + "learning_rate": 4.1023510809369396e-07, + "loss": 0.9777, + "step": 73992 + }, + { + "epoch": 5.47, + "learning_rate": 4.1012204135632226e-07, + "loss": 0.9692, + "step": 73993 + }, + { + "epoch": 5.47, + "learning_rate": 4.100089898763082e-07, + "loss": 0.9249, + "step": 73994 + }, + { + "epoch": 5.47, + "learning_rate": 4.0989595365383164e-07, + "loss": 1.04, + "step": 73995 + }, + { + "epoch": 5.47, + "learning_rate": 4.0978293268907676e-07, + "loss": 1.0246, + "step": 73996 + }, + { + "epoch": 5.47, + "learning_rate": 4.0966992698221685e-07, + "loss": 1.0406, + "step": 73997 + }, + { + "epoch": 5.47, + "learning_rate": 4.095569365334362e-07, + "loss": 1.0009, + "step": 73998 + }, + { + "epoch": 5.47, + "learning_rate": 4.094439613429113e-07, + "loss": 0.9576, + "step": 73999 + }, + { + "epoch": 5.47, + "learning_rate": 4.093310014108276e-07, + "loss": 1.0048, + "step": 74000 + }, + { + "epoch": 5.47, + "learning_rate": 4.092180567373594e-07, + "loss": 1.0377, + "step": 74001 + }, + { + "epoch": 5.47, + "learning_rate": 4.0910512732268646e-07, + "loss": 0.9453, + "step": 74002 + }, + { + "epoch": 5.47, + "learning_rate": 4.08992213166991e-07, + "loss": 1.0167, + "step": 74003 + }, + { + "epoch": 5.47, + "learning_rate": 4.088793142704528e-07, + "loss": 0.9465, + "step": 74004 + }, + { + "epoch": 5.47, + "learning_rate": 4.087664306332495e-07, + "loss": 0.9916, + "step": 74005 + }, + { + "epoch": 5.47, + "learning_rate": 4.0865356225555985e-07, + "loss": 1.0336, + "step": 74006 + }, + { + "epoch": 5.47, + "learning_rate": 4.0854070913756705e-07, + "loss": 1.0678, + "step": 74007 + }, + { + "epoch": 5.47, + "learning_rate": 4.0842787127944873e-07, + "loss": 1.0231, + "step": 74008 + }, + { + "epoch": 5.47, + "learning_rate": 4.083150486813847e-07, + "loss": 1.0154, + "step": 74009 + }, + { + "epoch": 5.47, + "learning_rate": 4.082022413435538e-07, + "loss": 0.9557, + "step": 74010 + }, + { + "epoch": 5.47, + "learning_rate": 4.080894492661358e-07, + "loss": 0.9951, + "step": 74011 + }, + { + "epoch": 5.47, + "learning_rate": 4.079766724493106e-07, + "loss": 0.9601, + "step": 74012 + }, + { + "epoch": 5.47, + "learning_rate": 4.0786391089325696e-07, + "loss": 1.0238, + "step": 74013 + }, + { + "epoch": 5.47, + "learning_rate": 4.077511645981547e-07, + "loss": 1.0104, + "step": 74014 + }, + { + "epoch": 5.47, + "learning_rate": 4.0763843356418143e-07, + "loss": 0.9318, + "step": 74015 + }, + { + "epoch": 5.47, + "learning_rate": 4.075257177915204e-07, + "loss": 1.0793, + "step": 74016 + }, + { + "epoch": 5.47, + "learning_rate": 4.074130172803481e-07, + "loss": 0.957, + "step": 74017 + }, + { + "epoch": 5.47, + "learning_rate": 4.0730033203084554e-07, + "loss": 0.9915, + "step": 74018 + }, + { + "epoch": 5.47, + "learning_rate": 4.0718766204318806e-07, + "loss": 0.9593, + "step": 74019 + }, + { + "epoch": 5.47, + "learning_rate": 4.0707500731756e-07, + "loss": 1.0751, + "step": 74020 + }, + { + "epoch": 5.47, + "learning_rate": 4.06962367854139e-07, + "loss": 0.9687, + "step": 74021 + }, + { + "epoch": 5.47, + "learning_rate": 4.068497436531016e-07, + "loss": 1.0224, + "step": 74022 + }, + { + "epoch": 5.47, + "learning_rate": 4.067371347146298e-07, + "loss": 0.9729, + "step": 74023 + }, + { + "epoch": 5.47, + "learning_rate": 4.0662454103890134e-07, + "loss": 1.0497, + "step": 74024 + }, + { + "epoch": 5.47, + "learning_rate": 4.0651196262609825e-07, + "loss": 1.0785, + "step": 74025 + }, + { + "epoch": 5.47, + "learning_rate": 4.0639939947639486e-07, + "loss": 1.0355, + "step": 74026 + }, + { + "epoch": 5.47, + "learning_rate": 4.0628685158997314e-07, + "loss": 1.0469, + "step": 74027 + }, + { + "epoch": 5.47, + "learning_rate": 4.061743189670131e-07, + "loss": 0.9837, + "step": 74028 + }, + { + "epoch": 5.47, + "learning_rate": 4.0606180160769224e-07, + "loss": 0.9403, + "step": 74029 + }, + { + "epoch": 5.47, + "learning_rate": 4.0594929951218944e-07, + "loss": 0.9805, + "step": 74030 + }, + { + "epoch": 5.47, + "learning_rate": 4.058368126806822e-07, + "loss": 0.8984, + "step": 74031 + }, + { + "epoch": 5.47, + "learning_rate": 4.057243411133527e-07, + "loss": 0.9575, + "step": 74032 + }, + { + "epoch": 5.47, + "learning_rate": 4.056118848103796e-07, + "loss": 0.9908, + "step": 74033 + }, + { + "epoch": 5.47, + "learning_rate": 4.054994437719395e-07, + "loss": 0.8291, + "step": 74034 + }, + { + "epoch": 5.47, + "learning_rate": 4.053870179982122e-07, + "loss": 1.1683, + "step": 74035 + }, + { + "epoch": 5.47, + "learning_rate": 4.0527460748937764e-07, + "loss": 1.0855, + "step": 74036 + }, + { + "epoch": 5.47, + "learning_rate": 4.0516221224561445e-07, + "loss": 1.1401, + "step": 74037 + }, + { + "epoch": 5.47, + "learning_rate": 4.0504983226710037e-07, + "loss": 1.0783, + "step": 74038 + }, + { + "epoch": 5.47, + "learning_rate": 4.049374675540152e-07, + "loss": 1.0873, + "step": 74039 + }, + { + "epoch": 5.47, + "learning_rate": 4.0482511810653544e-07, + "loss": 0.9334, + "step": 74040 + }, + { + "epoch": 5.47, + "learning_rate": 4.0471278392484437e-07, + "loss": 0.9546, + "step": 74041 + }, + { + "epoch": 5.47, + "learning_rate": 4.046004650091151e-07, + "loss": 0.9855, + "step": 74042 + }, + { + "epoch": 5.47, + "learning_rate": 4.0448816135953085e-07, + "loss": 1.0743, + "step": 74043 + }, + { + "epoch": 5.47, + "learning_rate": 4.04375872976267e-07, + "loss": 1.0377, + "step": 74044 + }, + { + "epoch": 5.47, + "learning_rate": 4.042635998595068e-07, + "loss": 1.0722, + "step": 74045 + }, + { + "epoch": 5.47, + "learning_rate": 4.0415134200942343e-07, + "loss": 0.9016, + "step": 74046 + }, + { + "epoch": 5.47, + "learning_rate": 4.040390994261989e-07, + "loss": 0.9924, + "step": 74047 + }, + { + "epoch": 5.47, + "learning_rate": 4.039268721100109e-07, + "loss": 1.0248, + "step": 74048 + }, + { + "epoch": 5.47, + "learning_rate": 4.038146600610382e-07, + "loss": 0.9922, + "step": 74049 + }, + { + "epoch": 5.47, + "learning_rate": 4.037024632794595e-07, + "loss": 0.9327, + "step": 74050 + }, + { + "epoch": 5.47, + "learning_rate": 4.035902817654502e-07, + "loss": 1.0161, + "step": 74051 + }, + { + "epoch": 5.47, + "learning_rate": 4.0347811551919356e-07, + "loss": 1.0279, + "step": 74052 + }, + { + "epoch": 5.47, + "learning_rate": 4.033659645408661e-07, + "loss": 1.0265, + "step": 74053 + }, + { + "epoch": 5.47, + "learning_rate": 4.0325382883064644e-07, + "loss": 1.0164, + "step": 74054 + }, + { + "epoch": 5.47, + "learning_rate": 4.0314170838871125e-07, + "loss": 1.0355, + "step": 74055 + }, + { + "epoch": 5.47, + "learning_rate": 4.030296032152403e-07, + "loss": 1.068, + "step": 74056 + }, + { + "epoch": 5.47, + "learning_rate": 4.0291751331041353e-07, + "loss": 1.0548, + "step": 74057 + }, + { + "epoch": 5.47, + "learning_rate": 4.0280543867440627e-07, + "loss": 1.0325, + "step": 74058 + }, + { + "epoch": 5.47, + "learning_rate": 4.0269337930739947e-07, + "loss": 1.0101, + "step": 74059 + }, + { + "epoch": 5.47, + "learning_rate": 4.025813352095687e-07, + "loss": 0.9018, + "step": 74060 + }, + { + "epoch": 5.47, + "learning_rate": 4.0246930638109473e-07, + "loss": 1.0285, + "step": 74061 + }, + { + "epoch": 5.47, + "learning_rate": 4.023572928221542e-07, + "loss": 0.9381, + "step": 74062 + }, + { + "epoch": 5.47, + "learning_rate": 4.0224529453292695e-07, + "loss": 0.8922, + "step": 74063 + }, + { + "epoch": 5.47, + "learning_rate": 4.021333115135884e-07, + "loss": 1.0512, + "step": 74064 + }, + { + "epoch": 5.47, + "learning_rate": 4.0202134376431944e-07, + "loss": 1.0403, + "step": 74065 + }, + { + "epoch": 5.47, + "learning_rate": 4.0190939128529895e-07, + "loss": 0.9554, + "step": 74066 + }, + { + "epoch": 5.47, + "learning_rate": 4.017974540767e-07, + "loss": 1.0884, + "step": 74067 + }, + { + "epoch": 5.47, + "learning_rate": 4.0168553213870586e-07, + "loss": 0.9752, + "step": 74068 + }, + { + "epoch": 5.47, + "learning_rate": 4.015736254714908e-07, + "loss": 1.0054, + "step": 74069 + }, + { + "epoch": 5.47, + "learning_rate": 4.01461734075238e-07, + "loss": 0.9892, + "step": 74070 + }, + { + "epoch": 5.47, + "learning_rate": 4.013498579501196e-07, + "loss": 1.0568, + "step": 74071 + }, + { + "epoch": 5.47, + "learning_rate": 4.012379970963176e-07, + "loss": 1.0422, + "step": 74072 + }, + { + "epoch": 5.47, + "learning_rate": 4.0112615151400856e-07, + "loss": 0.989, + "step": 74073 + }, + { + "epoch": 5.47, + "learning_rate": 4.0101432120337014e-07, + "loss": 0.9533, + "step": 74074 + }, + { + "epoch": 5.47, + "learning_rate": 4.009025061645799e-07, + "loss": 1.0157, + "step": 74075 + }, + { + "epoch": 5.47, + "learning_rate": 4.007907063978167e-07, + "loss": 1.0147, + "step": 74076 + }, + { + "epoch": 5.47, + "learning_rate": 4.0067892190325917e-07, + "loss": 0.9703, + "step": 74077 + }, + { + "epoch": 5.47, + "learning_rate": 4.0056715268108393e-07, + "loss": 0.9987, + "step": 74078 + }, + { + "epoch": 5.47, + "learning_rate": 4.0045539873146855e-07, + "loss": 1.12, + "step": 74079 + }, + { + "epoch": 5.47, + "learning_rate": 4.003436600545907e-07, + "loss": 1.0135, + "step": 74080 + }, + { + "epoch": 5.47, + "learning_rate": 4.002319366506291e-07, + "loss": 1.0564, + "step": 74081 + }, + { + "epoch": 5.47, + "learning_rate": 4.001202285197614e-07, + "loss": 1.0545, + "step": 74082 + }, + { + "epoch": 5.47, + "learning_rate": 4.0000853566216523e-07, + "loss": 1.0491, + "step": 74083 + }, + { + "epoch": 5.47, + "learning_rate": 3.9989685807801715e-07, + "loss": 1.0487, + "step": 74084 + }, + { + "epoch": 5.47, + "learning_rate": 3.9978519576749697e-07, + "loss": 1.027, + "step": 74085 + }, + { + "epoch": 5.47, + "learning_rate": 3.9967354873078237e-07, + "loss": 0.8522, + "step": 74086 + }, + { + "epoch": 5.47, + "learning_rate": 3.9956191696804757e-07, + "loss": 0.9755, + "step": 74087 + }, + { + "epoch": 5.47, + "learning_rate": 3.9945030047947255e-07, + "loss": 0.9505, + "step": 74088 + }, + { + "epoch": 5.47, + "learning_rate": 3.9933869926523483e-07, + "loss": 1.0333, + "step": 74089 + }, + { + "epoch": 5.47, + "learning_rate": 3.9922711332551324e-07, + "loss": 1.0232, + "step": 74090 + }, + { + "epoch": 5.47, + "learning_rate": 3.99115542660482e-07, + "loss": 0.8797, + "step": 74091 + }, + { + "epoch": 5.47, + "learning_rate": 3.9900398727032217e-07, + "loss": 0.9393, + "step": 74092 + }, + { + "epoch": 5.47, + "learning_rate": 3.98892447155208e-07, + "loss": 1.0124, + "step": 74093 + }, + { + "epoch": 5.47, + "learning_rate": 3.987809223153216e-07, + "loss": 0.9052, + "step": 74094 + }, + { + "epoch": 5.47, + "learning_rate": 3.9866941275083504e-07, + "loss": 0.9424, + "step": 74095 + }, + { + "epoch": 5.47, + "learning_rate": 3.9855791846192816e-07, + "loss": 0.9759, + "step": 74096 + }, + { + "epoch": 5.47, + "learning_rate": 3.9844643944877857e-07, + "loss": 0.9982, + "step": 74097 + }, + { + "epoch": 5.48, + "learning_rate": 3.983349757115629e-07, + "loss": 0.9903, + "step": 74098 + }, + { + "epoch": 5.48, + "learning_rate": 3.9822352725045977e-07, + "loss": 1.0975, + "step": 74099 + }, + { + "epoch": 5.48, + "learning_rate": 3.9811209406564466e-07, + "loss": 1.0064, + "step": 74100 + }, + { + "epoch": 5.48, + "learning_rate": 3.980006761572963e-07, + "loss": 0.9603, + "step": 74101 + }, + { + "epoch": 5.48, + "learning_rate": 3.978892735255913e-07, + "loss": 0.9346, + "step": 74102 + }, + { + "epoch": 5.48, + "learning_rate": 3.9777788617070823e-07, + "loss": 0.9049, + "step": 74103 + }, + { + "epoch": 5.48, + "learning_rate": 3.9766651409282263e-07, + "loss": 1.0386, + "step": 74104 + }, + { + "epoch": 5.48, + "learning_rate": 3.9755515729210993e-07, + "loss": 1.0239, + "step": 74105 + }, + { + "epoch": 5.48, + "learning_rate": 3.9744381576875213e-07, + "loss": 0.9631, + "step": 74106 + }, + { + "epoch": 5.48, + "learning_rate": 3.9733248952292357e-07, + "loss": 1.0135, + "step": 74107 + }, + { + "epoch": 5.48, + "learning_rate": 3.972211785548008e-07, + "loss": 0.9735, + "step": 74108 + }, + { + "epoch": 5.48, + "learning_rate": 3.9710988286456143e-07, + "loss": 0.9728, + "step": 74109 + }, + { + "epoch": 5.48, + "learning_rate": 3.969986024523842e-07, + "loss": 1.0217, + "step": 74110 + }, + { + "epoch": 5.48, + "learning_rate": 3.9688733731844455e-07, + "loss": 0.9761, + "step": 74111 + }, + { + "epoch": 5.48, + "learning_rate": 3.9677608746291894e-07, + "loss": 0.9895, + "step": 74112 + }, + { + "epoch": 5.48, + "learning_rate": 3.966648528859862e-07, + "loss": 0.9867, + "step": 74113 + }, + { + "epoch": 5.48, + "learning_rate": 3.9655363358782063e-07, + "loss": 0.9708, + "step": 74114 + }, + { + "epoch": 5.48, + "learning_rate": 3.9644242956860315e-07, + "loss": 1.024, + "step": 74115 + }, + { + "epoch": 5.48, + "learning_rate": 3.9633124082850585e-07, + "loss": 1.0208, + "step": 74116 + }, + { + "epoch": 5.48, + "learning_rate": 3.9622006736770967e-07, + "loss": 1.0485, + "step": 74117 + }, + { + "epoch": 5.48, + "learning_rate": 3.961089091863901e-07, + "loss": 1.0292, + "step": 74118 + }, + { + "epoch": 5.48, + "learning_rate": 3.9599776628472366e-07, + "loss": 0.8769, + "step": 74119 + }, + { + "epoch": 5.48, + "learning_rate": 3.958866386628868e-07, + "loss": 1.0474, + "step": 74120 + }, + { + "epoch": 5.48, + "learning_rate": 3.957755263210572e-07, + "loss": 1.051, + "step": 74121 + }, + { + "epoch": 5.48, + "learning_rate": 3.956644292594114e-07, + "loss": 0.8712, + "step": 74122 + }, + { + "epoch": 5.48, + "learning_rate": 3.955533474781259e-07, + "loss": 1.0905, + "step": 74123 + }, + { + "epoch": 5.48, + "learning_rate": 3.9544228097737836e-07, + "loss": 0.9038, + "step": 74124 + }, + { + "epoch": 5.48, + "learning_rate": 3.9533122975734306e-07, + "loss": 1.0749, + "step": 74125 + }, + { + "epoch": 5.48, + "learning_rate": 3.9522019381819987e-07, + "loss": 0.9747, + "step": 74126 + }, + { + "epoch": 5.48, + "learning_rate": 3.9510917316012307e-07, + "loss": 1.0743, + "step": 74127 + }, + { + "epoch": 5.48, + "learning_rate": 3.949981677832915e-07, + "loss": 1.0372, + "step": 74128 + }, + { + "epoch": 5.48, + "learning_rate": 3.948871776878782e-07, + "loss": 1.0192, + "step": 74129 + }, + { + "epoch": 5.48, + "learning_rate": 3.947762028740632e-07, + "loss": 1.031, + "step": 74130 + }, + { + "epoch": 5.48, + "learning_rate": 3.946652433420217e-07, + "loss": 0.9806, + "step": 74131 + }, + { + "epoch": 5.48, + "learning_rate": 3.9455429909193153e-07, + "loss": 0.9766, + "step": 74132 + }, + { + "epoch": 5.48, + "learning_rate": 3.9444337012396695e-07, + "loss": 0.9848, + "step": 74133 + }, + { + "epoch": 5.48, + "learning_rate": 3.943324564383044e-07, + "loss": 1.0017, + "step": 74134 + }, + { + "epoch": 5.48, + "learning_rate": 3.9422155803512384e-07, + "loss": 1.0274, + "step": 74135 + }, + { + "epoch": 5.48, + "learning_rate": 3.941106749145973e-07, + "loss": 0.9236, + "step": 74136 + }, + { + "epoch": 5.48, + "learning_rate": 3.9399980707690466e-07, + "loss": 1.1245, + "step": 74137 + }, + { + "epoch": 5.48, + "learning_rate": 3.938889545222191e-07, + "loss": 0.9301, + "step": 74138 + }, + { + "epoch": 5.48, + "learning_rate": 3.9377811725072156e-07, + "loss": 1.0279, + "step": 74139 + }, + { + "epoch": 5.48, + "learning_rate": 3.9366729526258306e-07, + "loss": 1.0018, + "step": 74140 + }, + { + "epoch": 5.48, + "learning_rate": 3.9355648855798343e-07, + "loss": 1.0106, + "step": 74141 + }, + { + "epoch": 5.48, + "learning_rate": 3.9344569713709813e-07, + "loss": 0.9001, + "step": 74142 + }, + { + "epoch": 5.48, + "learning_rate": 3.933349210001036e-07, + "loss": 0.9513, + "step": 74143 + }, + { + "epoch": 5.48, + "learning_rate": 3.9322416014717534e-07, + "loss": 0.9288, + "step": 74144 + }, + { + "epoch": 5.48, + "learning_rate": 3.9311341457848983e-07, + "loss": 0.916, + "step": 74145 + }, + { + "epoch": 5.48, + "learning_rate": 3.930026842942236e-07, + "loss": 0.9828, + "step": 74146 + }, + { + "epoch": 5.48, + "learning_rate": 3.928919692945532e-07, + "loss": 0.9069, + "step": 74147 + }, + { + "epoch": 5.48, + "learning_rate": 3.927812695796529e-07, + "loss": 0.9058, + "step": 74148 + }, + { + "epoch": 5.48, + "learning_rate": 3.9267058514970035e-07, + "loss": 0.9157, + "step": 74149 + }, + { + "epoch": 5.48, + "learning_rate": 3.925599160048721e-07, + "loss": 1.0466, + "step": 74150 + }, + { + "epoch": 5.48, + "learning_rate": 3.924492621453435e-07, + "loss": 1.0116, + "step": 74151 + }, + { + "epoch": 5.48, + "learning_rate": 3.923386235712912e-07, + "loss": 1.0366, + "step": 74152 + }, + { + "epoch": 5.48, + "learning_rate": 3.9222800028289043e-07, + "loss": 1.0327, + "step": 74153 + }, + { + "epoch": 5.48, + "learning_rate": 3.9211739228031566e-07, + "loss": 0.9877, + "step": 74154 + }, + { + "epoch": 5.48, + "learning_rate": 3.9200679956374665e-07, + "loss": 1.0023, + "step": 74155 + }, + { + "epoch": 5.48, + "learning_rate": 3.9189622213335667e-07, + "loss": 1.0577, + "step": 74156 + }, + { + "epoch": 5.48, + "learning_rate": 3.917856599893233e-07, + "loss": 0.8721, + "step": 74157 + }, + { + "epoch": 5.48, + "learning_rate": 3.9167511313181863e-07, + "loss": 1.0535, + "step": 74158 + }, + { + "epoch": 5.48, + "learning_rate": 3.9156458156102586e-07, + "loss": 0.9964, + "step": 74159 + }, + { + "epoch": 5.48, + "learning_rate": 3.914540652771137e-07, + "loss": 0.946, + "step": 74160 + }, + { + "epoch": 5.48, + "learning_rate": 3.913435642802599e-07, + "loss": 1.0094, + "step": 74161 + }, + { + "epoch": 5.48, + "learning_rate": 3.9123307857064195e-07, + "loss": 0.9483, + "step": 74162 + }, + { + "epoch": 5.48, + "learning_rate": 3.911226081484354e-07, + "loss": 1.0244, + "step": 74163 + }, + { + "epoch": 5.48, + "learning_rate": 3.9101215301381445e-07, + "loss": 1.1663, + "step": 74164 + }, + { + "epoch": 5.48, + "learning_rate": 3.909017131669546e-07, + "loss": 0.9643, + "step": 74165 + }, + { + "epoch": 5.48, + "learning_rate": 3.907912886080345e-07, + "loss": 0.9739, + "step": 74166 + }, + { + "epoch": 5.48, + "learning_rate": 3.9068087933722744e-07, + "loss": 1.0164, + "step": 74167 + }, + { + "epoch": 5.48, + "learning_rate": 3.9057048535470986e-07, + "loss": 0.967, + "step": 74168 + }, + { + "epoch": 5.48, + "learning_rate": 3.9046010666065725e-07, + "loss": 0.9628, + "step": 74169 + }, + { + "epoch": 5.48, + "learning_rate": 3.9034974325524276e-07, + "loss": 0.8951, + "step": 74170 + }, + { + "epoch": 5.48, + "learning_rate": 3.902393951386474e-07, + "loss": 0.9344, + "step": 74171 + }, + { + "epoch": 5.48, + "learning_rate": 3.901290623110421e-07, + "loss": 0.9686, + "step": 74172 + }, + { + "epoch": 5.48, + "learning_rate": 3.9001874477260447e-07, + "loss": 1.0221, + "step": 74173 + }, + { + "epoch": 5.48, + "learning_rate": 3.899084425235089e-07, + "loss": 0.9414, + "step": 74174 + }, + { + "epoch": 5.48, + "learning_rate": 3.8979815556393187e-07, + "loss": 1.059, + "step": 74175 + }, + { + "epoch": 5.48, + "learning_rate": 3.8968788389404877e-07, + "loss": 1.0732, + "step": 74176 + }, + { + "epoch": 5.48, + "learning_rate": 3.8957762751403507e-07, + "loss": 0.9516, + "step": 74177 + }, + { + "epoch": 5.48, + "learning_rate": 3.8946738642406503e-07, + "loss": 1.1543, + "step": 74178 + }, + { + "epoch": 5.48, + "learning_rate": 3.8935716062431517e-07, + "loss": 0.9834, + "step": 74179 + }, + { + "epoch": 5.48, + "learning_rate": 3.8924695011496203e-07, + "loss": 0.9825, + "step": 74180 + }, + { + "epoch": 5.48, + "learning_rate": 3.891367548961766e-07, + "loss": 1.0232, + "step": 74181 + }, + { + "epoch": 5.48, + "learning_rate": 3.890265749681388e-07, + "loss": 0.9455, + "step": 74182 + }, + { + "epoch": 5.48, + "learning_rate": 3.8891641033102166e-07, + "loss": 0.9557, + "step": 74183 + }, + { + "epoch": 5.48, + "learning_rate": 3.8880626098500295e-07, + "loss": 1.0091, + "step": 74184 + }, + { + "epoch": 5.48, + "learning_rate": 3.886961269302525e-07, + "loss": 1.0648, + "step": 74185 + }, + { + "epoch": 5.48, + "learning_rate": 3.8858600816695124e-07, + "loss": 0.9956, + "step": 74186 + }, + { + "epoch": 5.48, + "learning_rate": 3.884759046952713e-07, + "loss": 0.9846, + "step": 74187 + }, + { + "epoch": 5.48, + "learning_rate": 3.883658165153903e-07, + "loss": 0.8452, + "step": 74188 + }, + { + "epoch": 5.48, + "learning_rate": 3.882557436274803e-07, + "loss": 1.0159, + "step": 74189 + }, + { + "epoch": 5.48, + "learning_rate": 3.8814568603171785e-07, + "loss": 0.9823, + "step": 74190 + }, + { + "epoch": 5.48, + "learning_rate": 3.880356437282784e-07, + "loss": 0.9809, + "step": 74191 + }, + { + "epoch": 5.48, + "learning_rate": 3.879256167173373e-07, + "loss": 1.0402, + "step": 74192 + }, + { + "epoch": 5.48, + "learning_rate": 3.8781560499906777e-07, + "loss": 1.0137, + "step": 74193 + }, + { + "epoch": 5.48, + "learning_rate": 3.8770560857364637e-07, + "loss": 0.9437, + "step": 74194 + }, + { + "epoch": 5.48, + "learning_rate": 3.8759562744124844e-07, + "loss": 0.9397, + "step": 74195 + }, + { + "epoch": 5.48, + "learning_rate": 3.874856616020484e-07, + "loss": 1.0025, + "step": 74196 + }, + { + "epoch": 5.48, + "learning_rate": 3.873757110562204e-07, + "loss": 0.9307, + "step": 74197 + }, + { + "epoch": 5.48, + "learning_rate": 3.8726577580394107e-07, + "loss": 0.8546, + "step": 74198 + }, + { + "epoch": 5.48, + "learning_rate": 3.8715585584538254e-07, + "loss": 1.0461, + "step": 74199 + }, + { + "epoch": 5.48, + "learning_rate": 3.870459511807234e-07, + "loss": 0.8538, + "step": 74200 + }, + { + "epoch": 5.48, + "learning_rate": 3.869360618101359e-07, + "loss": 0.9452, + "step": 74201 + }, + { + "epoch": 5.48, + "learning_rate": 3.868261877337964e-07, + "loss": 0.9696, + "step": 74202 + }, + { + "epoch": 5.48, + "learning_rate": 3.867163289518772e-07, + "loss": 1.0576, + "step": 74203 + }, + { + "epoch": 5.48, + "learning_rate": 3.866064854645568e-07, + "loss": 0.9242, + "step": 74204 + }, + { + "epoch": 5.48, + "learning_rate": 3.8649665727200634e-07, + "loss": 0.8797, + "step": 74205 + }, + { + "epoch": 5.48, + "learning_rate": 3.863868443744023e-07, + "loss": 1.0454, + "step": 74206 + }, + { + "epoch": 5.48, + "learning_rate": 3.8627704677192015e-07, + "loss": 0.9971, + "step": 74207 + }, + { + "epoch": 5.48, + "learning_rate": 3.8616726446473405e-07, + "loss": 0.9861, + "step": 74208 + }, + { + "epoch": 5.48, + "learning_rate": 3.8605749745301734e-07, + "loss": 0.8865, + "step": 74209 + }, + { + "epoch": 5.48, + "learning_rate": 3.8594774573694424e-07, + "loss": 0.9866, + "step": 74210 + }, + { + "epoch": 5.48, + "learning_rate": 3.858380093166925e-07, + "loss": 1.0977, + "step": 74211 + }, + { + "epoch": 5.48, + "learning_rate": 3.857282881924351e-07, + "loss": 1.0417, + "step": 74212 + }, + { + "epoch": 5.48, + "learning_rate": 3.856185823643455e-07, + "loss": 1.121, + "step": 74213 + }, + { + "epoch": 5.48, + "learning_rate": 3.855088918325989e-07, + "loss": 0.9574, + "step": 74214 + }, + { + "epoch": 5.48, + "learning_rate": 3.853992165973708e-07, + "loss": 1.0423, + "step": 74215 + }, + { + "epoch": 5.48, + "learning_rate": 3.8528955665883547e-07, + "loss": 1.1634, + "step": 74216 + }, + { + "epoch": 5.48, + "learning_rate": 3.851799120171662e-07, + "loss": 0.9531, + "step": 74217 + }, + { + "epoch": 5.48, + "learning_rate": 3.850702826725394e-07, + "loss": 1.018, + "step": 74218 + }, + { + "epoch": 5.48, + "learning_rate": 3.8496066862512616e-07, + "loss": 1.0147, + "step": 74219 + }, + { + "epoch": 5.48, + "learning_rate": 3.84851069875104e-07, + "loss": 0.8799, + "step": 74220 + }, + { + "epoch": 5.48, + "learning_rate": 3.847414864226462e-07, + "loss": 0.9936, + "step": 74221 + }, + { + "epoch": 5.48, + "learning_rate": 3.84631918267927e-07, + "loss": 1.001, + "step": 74222 + }, + { + "epoch": 5.48, + "learning_rate": 3.845223654111196e-07, + "loss": 1.0391, + "step": 74223 + }, + { + "epoch": 5.48, + "learning_rate": 3.8441282785240173e-07, + "loss": 0.9484, + "step": 74224 + }, + { + "epoch": 5.48, + "learning_rate": 3.843033055919454e-07, + "loss": 1.0321, + "step": 74225 + }, + { + "epoch": 5.48, + "learning_rate": 3.841937986299238e-07, + "loss": 0.9681, + "step": 74226 + }, + { + "epoch": 5.48, + "learning_rate": 3.8408430696651233e-07, + "loss": 0.9751, + "step": 74227 + }, + { + "epoch": 5.48, + "learning_rate": 3.839748306018842e-07, + "loss": 0.9957, + "step": 74228 + }, + { + "epoch": 5.48, + "learning_rate": 3.838653695362171e-07, + "loss": 1.0318, + "step": 74229 + }, + { + "epoch": 5.48, + "learning_rate": 3.8375592376968086e-07, + "loss": 0.9677, + "step": 74230 + }, + { + "epoch": 5.48, + "learning_rate": 3.8364649330245197e-07, + "loss": 0.9637, + "step": 74231 + }, + { + "epoch": 5.48, + "learning_rate": 3.8353707813470253e-07, + "loss": 0.9363, + "step": 74232 + }, + { + "epoch": 5.49, + "learning_rate": 3.834276782666102e-07, + "loss": 0.9772, + "step": 74233 + }, + { + "epoch": 5.49, + "learning_rate": 3.8331829369834486e-07, + "loss": 1.0266, + "step": 74234 + }, + { + "epoch": 5.49, + "learning_rate": 3.8320892443008407e-07, + "loss": 1.1187, + "step": 74235 + }, + { + "epoch": 5.49, + "learning_rate": 3.8309957046199994e-07, + "loss": 1.0484, + "step": 74236 + }, + { + "epoch": 5.49, + "learning_rate": 3.829902317942669e-07, + "loss": 1.0547, + "step": 74237 + }, + { + "epoch": 5.49, + "learning_rate": 3.8288090842705793e-07, + "loss": 1.071, + "step": 74238 + }, + { + "epoch": 5.49, + "learning_rate": 3.8277160036054753e-07, + "loss": 0.9992, + "step": 74239 + }, + { + "epoch": 5.49, + "learning_rate": 3.8266230759491096e-07, + "loss": 0.9662, + "step": 74240 + }, + { + "epoch": 5.49, + "learning_rate": 3.825530301303215e-07, + "loss": 0.9801, + "step": 74241 + }, + { + "epoch": 5.49, + "learning_rate": 3.8244376796695236e-07, + "loss": 1.0477, + "step": 74242 + }, + { + "epoch": 5.49, + "learning_rate": 3.823345211049756e-07, + "loss": 0.9984, + "step": 74243 + }, + { + "epoch": 5.49, + "learning_rate": 3.822252895445688e-07, + "loss": 1.0619, + "step": 74244 + }, + { + "epoch": 5.49, + "learning_rate": 3.8211607328590414e-07, + "loss": 1.0449, + "step": 74245 + }, + { + "epoch": 5.49, + "learning_rate": 3.820068723291548e-07, + "loss": 1.1282, + "step": 74246 + }, + { + "epoch": 5.49, + "learning_rate": 3.81897686674495e-07, + "loss": 0.9348, + "step": 74247 + }, + { + "epoch": 5.49, + "learning_rate": 3.8178851632209804e-07, + "loss": 1.0017, + "step": 74248 + }, + { + "epoch": 5.49, + "learning_rate": 3.816793612721392e-07, + "loss": 1.0182, + "step": 74249 + }, + { + "epoch": 5.49, + "learning_rate": 3.8157022152478853e-07, + "loss": 0.9465, + "step": 74250 + }, + { + "epoch": 5.49, + "learning_rate": 3.814610970802235e-07, + "loss": 1.0576, + "step": 74251 + }, + { + "epoch": 5.49, + "learning_rate": 3.8135198793861517e-07, + "loss": 0.9014, + "step": 74252 + }, + { + "epoch": 5.49, + "learning_rate": 3.8124289410014005e-07, + "loss": 0.9979, + "step": 74253 + }, + { + "epoch": 5.49, + "learning_rate": 3.811338155649702e-07, + "loss": 0.9414, + "step": 74254 + }, + { + "epoch": 5.49, + "learning_rate": 3.8102475233327553e-07, + "loss": 1.0892, + "step": 74255 + }, + { + "epoch": 5.49, + "learning_rate": 3.8091570440523584e-07, + "loss": 1.0309, + "step": 74256 + }, + { + "epoch": 5.49, + "learning_rate": 3.8080667178102105e-07, + "loss": 1.0088, + "step": 74257 + }, + { + "epoch": 5.49, + "learning_rate": 3.806976544608043e-07, + "loss": 1.1075, + "step": 74258 + }, + { + "epoch": 5.49, + "learning_rate": 3.8058865244475993e-07, + "loss": 0.9969, + "step": 74259 + }, + { + "epoch": 5.49, + "learning_rate": 3.8047966573306226e-07, + "loss": 0.9942, + "step": 74260 + }, + { + "epoch": 5.49, + "learning_rate": 3.8037069432588337e-07, + "loss": 1.0658, + "step": 74261 + }, + { + "epoch": 5.49, + "learning_rate": 3.802617382233975e-07, + "loss": 0.9446, + "step": 74262 + }, + { + "epoch": 5.49, + "learning_rate": 3.8015279742577795e-07, + "loss": 1.0288, + "step": 74263 + }, + { + "epoch": 5.49, + "learning_rate": 3.800438719331956e-07, + "loss": 1.0301, + "step": 74264 + }, + { + "epoch": 5.49, + "learning_rate": 3.79934961745827e-07, + "loss": 0.9353, + "step": 74265 + }, + { + "epoch": 5.49, + "learning_rate": 3.798260668638454e-07, + "loss": 1.0718, + "step": 74266 + }, + { + "epoch": 5.49, + "learning_rate": 3.797171872874217e-07, + "loss": 1.0184, + "step": 74267 + }, + { + "epoch": 5.49, + "learning_rate": 3.796083230167291e-07, + "loss": 1.091, + "step": 74268 + }, + { + "epoch": 5.49, + "learning_rate": 3.794994740519442e-07, + "loss": 1.0314, + "step": 74269 + }, + { + "epoch": 5.49, + "learning_rate": 3.793906403932368e-07, + "loss": 1.1113, + "step": 74270 + }, + { + "epoch": 5.49, + "learning_rate": 3.7928182204078234e-07, + "loss": 1.0107, + "step": 74271 + }, + { + "epoch": 5.49, + "learning_rate": 3.791730189947518e-07, + "loss": 0.9301, + "step": 74272 + }, + { + "epoch": 5.49, + "learning_rate": 3.7906423125531945e-07, + "loss": 0.9904, + "step": 74273 + }, + { + "epoch": 5.49, + "learning_rate": 3.789554588226596e-07, + "loss": 1.0085, + "step": 74274 + }, + { + "epoch": 5.49, + "learning_rate": 3.7884670169694213e-07, + "loss": 1.0649, + "step": 74275 + }, + { + "epoch": 5.49, + "learning_rate": 3.787379598783425e-07, + "loss": 1.0016, + "step": 74276 + }, + { + "epoch": 5.49, + "learning_rate": 3.7862923336703274e-07, + "loss": 1.0042, + "step": 74277 + }, + { + "epoch": 5.49, + "learning_rate": 3.785205221631882e-07, + "loss": 0.9414, + "step": 74278 + }, + { + "epoch": 5.49, + "learning_rate": 3.784118262669767e-07, + "loss": 1.0196, + "step": 74279 + }, + { + "epoch": 5.49, + "learning_rate": 3.7830314567857685e-07, + "loss": 0.9489, + "step": 74280 + }, + { + "epoch": 5.49, + "learning_rate": 3.7819448039815856e-07, + "loss": 0.9927, + "step": 74281 + }, + { + "epoch": 5.49, + "learning_rate": 3.78085830425895e-07, + "loss": 0.9152, + "step": 74282 + }, + { + "epoch": 5.49, + "learning_rate": 3.779771957619582e-07, + "loss": 1.0263, + "step": 74283 + }, + { + "epoch": 5.49, + "learning_rate": 3.778685764065226e-07, + "loss": 0.9175, + "step": 74284 + }, + { + "epoch": 5.49, + "learning_rate": 3.7775997235976027e-07, + "loss": 1.019, + "step": 74285 + }, + { + "epoch": 5.49, + "learning_rate": 3.776513836218454e-07, + "loss": 0.9635, + "step": 74286 + }, + { + "epoch": 5.49, + "learning_rate": 3.7754281019294903e-07, + "loss": 1.0014, + "step": 74287 + }, + { + "epoch": 5.49, + "learning_rate": 3.774342520732421e-07, + "loss": 1.0273, + "step": 74288 + }, + { + "epoch": 5.49, + "learning_rate": 3.7732570926290235e-07, + "loss": 1.0338, + "step": 74289 + }, + { + "epoch": 5.49, + "learning_rate": 3.7721718176209844e-07, + "loss": 1.0536, + "step": 74290 + }, + { + "epoch": 5.49, + "learning_rate": 3.771086695710058e-07, + "loss": 1.074, + "step": 74291 + }, + { + "epoch": 5.49, + "learning_rate": 3.770001726897943e-07, + "loss": 0.923, + "step": 74292 + }, + { + "epoch": 5.49, + "learning_rate": 3.768916911186371e-07, + "loss": 1.0853, + "step": 74293 + }, + { + "epoch": 5.49, + "learning_rate": 3.767832248577097e-07, + "loss": 0.9884, + "step": 74294 + }, + { + "epoch": 5.49, + "learning_rate": 3.7667477390718076e-07, + "loss": 0.9947, + "step": 74295 + }, + { + "epoch": 5.49, + "learning_rate": 3.7656633826722465e-07, + "loss": 1.0567, + "step": 74296 + }, + { + "epoch": 5.49, + "learning_rate": 3.7645791793801343e-07, + "loss": 0.8624, + "step": 74297 + }, + { + "epoch": 5.49, + "learning_rate": 3.7634951291972254e-07, + "loss": 0.9211, + "step": 74298 + }, + { + "epoch": 5.49, + "learning_rate": 3.762411232125196e-07, + "loss": 0.9854, + "step": 74299 + }, + { + "epoch": 5.49, + "learning_rate": 3.7613274881658003e-07, + "loss": 0.9251, + "step": 74300 + }, + { + "epoch": 5.49, + "learning_rate": 3.760243897320748e-07, + "loss": 1.0345, + "step": 74301 + }, + { + "epoch": 5.49, + "learning_rate": 3.759160459591782e-07, + "loss": 1.0206, + "step": 74302 + }, + { + "epoch": 5.49, + "learning_rate": 3.758077174980612e-07, + "loss": 0.9877, + "step": 74303 + }, + { + "epoch": 5.49, + "learning_rate": 3.7569940434889485e-07, + "loss": 1.0369, + "step": 74304 + }, + { + "epoch": 5.49, + "learning_rate": 3.755911065118556e-07, + "loss": 1.0791, + "step": 74305 + }, + { + "epoch": 5.49, + "learning_rate": 3.754828239871111e-07, + "loss": 0.88, + "step": 74306 + }, + { + "epoch": 5.49, + "learning_rate": 3.753745567748368e-07, + "loss": 0.9214, + "step": 74307 + }, + { + "epoch": 5.49, + "learning_rate": 3.7526630487520366e-07, + "loss": 1.0591, + "step": 74308 + }, + { + "epoch": 5.49, + "learning_rate": 3.7515806828838376e-07, + "loss": 1.0489, + "step": 74309 + }, + { + "epoch": 5.49, + "learning_rate": 3.7504984701455027e-07, + "loss": 0.9523, + "step": 74310 + }, + { + "epoch": 5.49, + "learning_rate": 3.7494164105387534e-07, + "loss": 1.0223, + "step": 74311 + }, + { + "epoch": 5.49, + "learning_rate": 3.7483345040652986e-07, + "loss": 0.9398, + "step": 74312 + }, + { + "epoch": 5.49, + "learning_rate": 3.747252750726849e-07, + "loss": 1.0451, + "step": 74313 + }, + { + "epoch": 5.49, + "learning_rate": 3.7461711505251686e-07, + "loss": 1.0383, + "step": 74314 + }, + { + "epoch": 5.49, + "learning_rate": 3.745089703461946e-07, + "loss": 1.1072, + "step": 74315 + }, + { + "epoch": 5.49, + "learning_rate": 3.7440084095389017e-07, + "loss": 1.0166, + "step": 74316 + }, + { + "epoch": 5.49, + "learning_rate": 3.742927268757757e-07, + "loss": 1.0035, + "step": 74317 + }, + { + "epoch": 5.49, + "learning_rate": 3.741846281120254e-07, + "loss": 0.9591, + "step": 74318 + }, + { + "epoch": 5.49, + "learning_rate": 3.740765446628103e-07, + "loss": 1.1016, + "step": 74319 + }, + { + "epoch": 5.49, + "learning_rate": 3.7396847652829914e-07, + "loss": 0.9071, + "step": 74320 + }, + { + "epoch": 5.49, + "learning_rate": 3.7386042370866846e-07, + "loss": 0.9411, + "step": 74321 + }, + { + "epoch": 5.49, + "learning_rate": 3.737523862040859e-07, + "loss": 0.9583, + "step": 74322 + }, + { + "epoch": 5.49, + "learning_rate": 3.7364436401472917e-07, + "loss": 0.8847, + "step": 74323 + }, + { + "epoch": 5.49, + "learning_rate": 3.7353635714076244e-07, + "loss": 1.026, + "step": 74324 + }, + { + "epoch": 5.49, + "learning_rate": 3.734283655823645e-07, + "loss": 0.9909, + "step": 74325 + }, + { + "epoch": 5.49, + "learning_rate": 3.7332038933970303e-07, + "loss": 1.0154, + "step": 74326 + }, + { + "epoch": 5.49, + "learning_rate": 3.7321242841295234e-07, + "loss": 0.977, + "step": 74327 + }, + { + "epoch": 5.49, + "learning_rate": 3.7310448280228227e-07, + "loss": 1.0176, + "step": 74328 + }, + { + "epoch": 5.49, + "learning_rate": 3.729965525078638e-07, + "loss": 0.961, + "step": 74329 + }, + { + "epoch": 5.49, + "learning_rate": 3.7288863752987123e-07, + "loss": 1.035, + "step": 74330 + }, + { + "epoch": 5.49, + "learning_rate": 3.727807378684756e-07, + "loss": 1.1705, + "step": 74331 + }, + { + "epoch": 5.49, + "learning_rate": 3.726728535238477e-07, + "loss": 1.0417, + "step": 74332 + }, + { + "epoch": 5.49, + "learning_rate": 3.7256498449615873e-07, + "loss": 0.9465, + "step": 74333 + }, + { + "epoch": 5.49, + "learning_rate": 3.7245713078558177e-07, + "loss": 0.8928, + "step": 74334 + }, + { + "epoch": 5.49, + "learning_rate": 3.723492923922878e-07, + "loss": 1.0573, + "step": 74335 + }, + { + "epoch": 5.49, + "learning_rate": 3.722414693164489e-07, + "loss": 1.0159, + "step": 74336 + }, + { + "epoch": 5.49, + "learning_rate": 3.721336615582349e-07, + "loss": 0.9805, + "step": 74337 + }, + { + "epoch": 5.49, + "learning_rate": 3.7202586911781913e-07, + "loss": 0.8777, + "step": 74338 + }, + { + "epoch": 5.49, + "learning_rate": 3.7191809199537353e-07, + "loss": 0.9701, + "step": 74339 + }, + { + "epoch": 5.49, + "learning_rate": 3.7181033019106694e-07, + "loss": 0.9882, + "step": 74340 + }, + { + "epoch": 5.49, + "learning_rate": 3.7170258370507253e-07, + "loss": 1.0064, + "step": 74341 + }, + { + "epoch": 5.49, + "learning_rate": 3.7159485253756123e-07, + "loss": 0.9464, + "step": 74342 + }, + { + "epoch": 5.49, + "learning_rate": 3.714871366887063e-07, + "loss": 0.9324, + "step": 74343 + }, + { + "epoch": 5.49, + "learning_rate": 3.713794361586753e-07, + "loss": 1.0197, + "step": 74344 + }, + { + "epoch": 5.49, + "learning_rate": 3.712717509476438e-07, + "loss": 0.9275, + "step": 74345 + }, + { + "epoch": 5.49, + "learning_rate": 3.7116408105577927e-07, + "loss": 0.8874, + "step": 74346 + }, + { + "epoch": 5.49, + "learning_rate": 3.7105642648325834e-07, + "loss": 1.035, + "step": 74347 + }, + { + "epoch": 5.49, + "learning_rate": 3.709487872302464e-07, + "loss": 1.0245, + "step": 74348 + }, + { + "epoch": 5.49, + "learning_rate": 3.7084116329691664e-07, + "loss": 0.931, + "step": 74349 + }, + { + "epoch": 5.49, + "learning_rate": 3.707335546834423e-07, + "loss": 1.0564, + "step": 74350 + }, + { + "epoch": 5.49, + "learning_rate": 3.706259613899932e-07, + "loss": 1.0744, + "step": 74351 + }, + { + "epoch": 5.49, + "learning_rate": 3.705183834167392e-07, + "loss": 1.0583, + "step": 74352 + }, + { + "epoch": 5.49, + "learning_rate": 3.7041082076385236e-07, + "loss": 0.9785, + "step": 74353 + }, + { + "epoch": 5.49, + "learning_rate": 3.7030327343150594e-07, + "loss": 0.9843, + "step": 74354 + }, + { + "epoch": 5.49, + "learning_rate": 3.701957414198676e-07, + "loss": 1.0338, + "step": 74355 + }, + { + "epoch": 5.49, + "learning_rate": 3.700882247291115e-07, + "loss": 0.99, + "step": 74356 + }, + { + "epoch": 5.49, + "learning_rate": 3.699807233594066e-07, + "loss": 0.9478, + "step": 74357 + }, + { + "epoch": 5.49, + "learning_rate": 3.698732373109237e-07, + "loss": 0.9824, + "step": 74358 + }, + { + "epoch": 5.49, + "learning_rate": 3.69765766583835e-07, + "loss": 0.8935, + "step": 74359 + }, + { + "epoch": 5.49, + "learning_rate": 3.696583111783125e-07, + "loss": 0.9584, + "step": 74360 + }, + { + "epoch": 5.49, + "learning_rate": 3.69550871094525e-07, + "loss": 0.9374, + "step": 74361 + }, + { + "epoch": 5.49, + "learning_rate": 3.694434463326435e-07, + "loss": 1.0592, + "step": 74362 + }, + { + "epoch": 5.49, + "learning_rate": 3.6933603689284003e-07, + "loss": 1.0773, + "step": 74363 + }, + { + "epoch": 5.49, + "learning_rate": 3.692286427752867e-07, + "loss": 0.9731, + "step": 74364 + }, + { + "epoch": 5.49, + "learning_rate": 3.6912126398015115e-07, + "loss": 1.0566, + "step": 74365 + }, + { + "epoch": 5.49, + "learning_rate": 3.690139005076065e-07, + "loss": 0.9885, + "step": 74366 + }, + { + "epoch": 5.49, + "learning_rate": 3.689065523578228e-07, + "loss": 0.9543, + "step": 74367 + }, + { + "epoch": 5.5, + "learning_rate": 3.6879921953097197e-07, + "loss": 1.0384, + "step": 74368 + }, + { + "epoch": 5.5, + "learning_rate": 3.686919020272217e-07, + "loss": 1.0598, + "step": 74369 + }, + { + "epoch": 5.5, + "learning_rate": 3.685845998467463e-07, + "loss": 1.0485, + "step": 74370 + }, + { + "epoch": 5.5, + "learning_rate": 3.6847731298971455e-07, + "loss": 0.9837, + "step": 74371 + }, + { + "epoch": 5.5, + "learning_rate": 3.683700414562985e-07, + "loss": 1.0379, + "step": 74372 + }, + { + "epoch": 5.5, + "learning_rate": 3.682627852466658e-07, + "loss": 1.0262, + "step": 74373 + }, + { + "epoch": 5.5, + "learning_rate": 3.6815554436099074e-07, + "loss": 1.0237, + "step": 74374 + }, + { + "epoch": 5.5, + "learning_rate": 3.6804831879944214e-07, + "loss": 1.0317, + "step": 74375 + }, + { + "epoch": 5.5, + "learning_rate": 3.679411085621909e-07, + "loss": 0.9317, + "step": 74376 + }, + { + "epoch": 5.5, + "learning_rate": 3.67833913649408e-07, + "loss": 0.9849, + "step": 74377 + }, + { + "epoch": 5.5, + "learning_rate": 3.677267340612622e-07, + "loss": 0.9475, + "step": 74378 + }, + { + "epoch": 5.5, + "learning_rate": 3.6761956979792677e-07, + "loss": 1.0746, + "step": 74379 + }, + { + "epoch": 5.5, + "learning_rate": 3.675124208595704e-07, + "loss": 1.0176, + "step": 74380 + }, + { + "epoch": 5.5, + "learning_rate": 3.674052872463629e-07, + "loss": 0.9949, + "step": 74381 + }, + { + "epoch": 5.5, + "learning_rate": 3.672981689584765e-07, + "loss": 1.0845, + "step": 74382 + }, + { + "epoch": 5.5, + "learning_rate": 3.671910659960809e-07, + "loss": 0.9352, + "step": 74383 + }, + { + "epoch": 5.5, + "learning_rate": 3.670839783593483e-07, + "loss": 0.9901, + "step": 74384 + }, + { + "epoch": 5.5, + "learning_rate": 3.6697690604844406e-07, + "loss": 1.1831, + "step": 74385 + }, + { + "epoch": 5.5, + "learning_rate": 3.668698490635436e-07, + "loss": 0.8373, + "step": 74386 + }, + { + "epoch": 5.5, + "learning_rate": 3.667628074048146e-07, + "loss": 0.9137, + "step": 74387 + }, + { + "epoch": 5.5, + "learning_rate": 3.666557810724292e-07, + "loss": 0.9417, + "step": 74388 + }, + { + "epoch": 5.5, + "learning_rate": 3.6654877006655487e-07, + "loss": 0.9683, + "step": 74389 + }, + { + "epoch": 5.5, + "learning_rate": 3.6644177438736497e-07, + "loss": 0.9621, + "step": 74390 + }, + { + "epoch": 5.5, + "learning_rate": 3.6633479403502815e-07, + "loss": 1.0553, + "step": 74391 + }, + { + "epoch": 5.5, + "learning_rate": 3.6622782900971545e-07, + "loss": 0.9275, + "step": 74392 + }, + { + "epoch": 5.5, + "learning_rate": 3.661208793115956e-07, + "loss": 0.9951, + "step": 74393 + }, + { + "epoch": 5.5, + "learning_rate": 3.660139449408395e-07, + "loss": 1.0512, + "step": 74394 + }, + { + "epoch": 5.5, + "learning_rate": 3.6590702589761826e-07, + "loss": 1.041, + "step": 74395 + }, + { + "epoch": 5.5, + "learning_rate": 3.6580012218210057e-07, + "loss": 0.8972, + "step": 74396 + }, + { + "epoch": 5.5, + "learning_rate": 3.656932337944574e-07, + "loss": 0.9319, + "step": 74397 + }, + { + "epoch": 5.5, + "learning_rate": 3.655863607348564e-07, + "loss": 0.9242, + "step": 74398 + }, + { + "epoch": 5.5, + "learning_rate": 3.654795030034719e-07, + "loss": 1.0057, + "step": 74399 + }, + { + "epoch": 5.5, + "learning_rate": 3.653726606004715e-07, + "loss": 0.9445, + "step": 74400 + }, + { + "epoch": 5.5, + "learning_rate": 3.6526583352602505e-07, + "loss": 0.9321, + "step": 74401 + }, + { + "epoch": 5.5, + "learning_rate": 3.651590217803014e-07, + "loss": 1.0586, + "step": 74402 + }, + { + "epoch": 5.5, + "learning_rate": 3.650522253634736e-07, + "loss": 1.0265, + "step": 74403 + }, + { + "epoch": 5.5, + "learning_rate": 3.649454442757094e-07, + "loss": 1.0184, + "step": 74404 + }, + { + "epoch": 5.5, + "learning_rate": 3.648386785171798e-07, + "loss": 1.0168, + "step": 74405 + }, + { + "epoch": 5.5, + "learning_rate": 3.647319280880535e-07, + "loss": 1.0309, + "step": 74406 + }, + { + "epoch": 5.5, + "learning_rate": 3.646251929884992e-07, + "loss": 0.8887, + "step": 74407 + }, + { + "epoch": 5.5, + "learning_rate": 3.6451847321869015e-07, + "loss": 0.9555, + "step": 74408 + }, + { + "epoch": 5.5, + "learning_rate": 3.6441176877879404e-07, + "loss": 0.8983, + "step": 74409 + }, + { + "epoch": 5.5, + "learning_rate": 3.6430507966898175e-07, + "loss": 0.9153, + "step": 74410 + }, + { + "epoch": 5.5, + "learning_rate": 3.641984058894199e-07, + "loss": 1.0053, + "step": 74411 + }, + { + "epoch": 5.5, + "learning_rate": 3.6409174744028276e-07, + "loss": 1.0706, + "step": 74412 + }, + { + "epoch": 5.5, + "learning_rate": 3.6398510432173794e-07, + "loss": 1.0114, + "step": 74413 + }, + { + "epoch": 5.5, + "learning_rate": 3.6387847653395314e-07, + "loss": 1.0022, + "step": 74414 + }, + { + "epoch": 5.5, + "learning_rate": 3.637718640771015e-07, + "loss": 1.0481, + "step": 74415 + }, + { + "epoch": 5.5, + "learning_rate": 3.636652669513507e-07, + "loss": 1.0503, + "step": 74416 + }, + { + "epoch": 5.5, + "learning_rate": 3.6355868515686955e-07, + "loss": 0.9202, + "step": 74417 + }, + { + "epoch": 5.5, + "learning_rate": 3.6345211869382893e-07, + "loss": 1.0712, + "step": 74418 + }, + { + "epoch": 5.5, + "learning_rate": 3.633455675623987e-07, + "loss": 1.0973, + "step": 74419 + }, + { + "epoch": 5.5, + "learning_rate": 3.6323903176274766e-07, + "loss": 1.0141, + "step": 74420 + }, + { + "epoch": 5.5, + "learning_rate": 3.6313251129504566e-07, + "loss": 1.0523, + "step": 74421 + }, + { + "epoch": 5.5, + "learning_rate": 3.630260061594615e-07, + "loss": 0.8626, + "step": 74422 + }, + { + "epoch": 5.5, + "learning_rate": 3.6291951635616387e-07, + "loss": 1.0075, + "step": 74423 + }, + { + "epoch": 5.5, + "learning_rate": 3.6281304188532484e-07, + "loss": 0.9941, + "step": 74424 + }, + { + "epoch": 5.5, + "learning_rate": 3.627065827471121e-07, + "loss": 0.8081, + "step": 74425 + }, + { + "epoch": 5.5, + "learning_rate": 3.6260013894169666e-07, + "loss": 0.8504, + "step": 74426 + }, + { + "epoch": 5.5, + "learning_rate": 3.624937104692439e-07, + "loss": 0.9623, + "step": 74427 + }, + { + "epoch": 5.5, + "learning_rate": 3.6238729732992695e-07, + "loss": 1.0796, + "step": 74428 + }, + { + "epoch": 5.5, + "learning_rate": 3.6228089952391467e-07, + "loss": 0.9351, + "step": 74429 + }, + { + "epoch": 5.5, + "learning_rate": 3.621745170513746e-07, + "loss": 0.8943, + "step": 74430 + }, + { + "epoch": 5.5, + "learning_rate": 3.6206814991247786e-07, + "loss": 1.1603, + "step": 74431 + }, + { + "epoch": 5.5, + "learning_rate": 3.61961798107392e-07, + "loss": 0.9582, + "step": 74432 + }, + { + "epoch": 5.5, + "learning_rate": 3.618554616362879e-07, + "loss": 0.9825, + "step": 74433 + }, + { + "epoch": 5.5, + "learning_rate": 3.617491404993323e-07, + "loss": 0.9157, + "step": 74434 + }, + { + "epoch": 5.5, + "learning_rate": 3.6164283469669716e-07, + "loss": 1.0706, + "step": 74435 + }, + { + "epoch": 5.5, + "learning_rate": 3.615365442285501e-07, + "loss": 0.9524, + "step": 74436 + }, + { + "epoch": 5.5, + "learning_rate": 3.614302690950611e-07, + "loss": 1.0063, + "step": 74437 + }, + { + "epoch": 5.5, + "learning_rate": 3.613240092963977e-07, + "loss": 1.1361, + "step": 74438 + }, + { + "epoch": 5.5, + "learning_rate": 3.6121776483272973e-07, + "loss": 1.0838, + "step": 74439 + }, + { + "epoch": 5.5, + "learning_rate": 3.611115357042272e-07, + "loss": 0.9141, + "step": 74440 + }, + { + "epoch": 5.5, + "learning_rate": 3.6100532191105873e-07, + "loss": 1.0063, + "step": 74441 + }, + { + "epoch": 5.5, + "learning_rate": 3.6089912345339207e-07, + "loss": 0.8727, + "step": 74442 + }, + { + "epoch": 5.5, + "learning_rate": 3.607929403313959e-07, + "loss": 1.13, + "step": 74443 + }, + { + "epoch": 5.5, + "learning_rate": 3.6068677254524233e-07, + "loss": 0.9963, + "step": 74444 + }, + { + "epoch": 5.5, + "learning_rate": 3.6058062009509675e-07, + "loss": 0.9185, + "step": 74445 + }, + { + "epoch": 5.5, + "learning_rate": 3.604744829811302e-07, + "loss": 0.9718, + "step": 74446 + }, + { + "epoch": 5.5, + "learning_rate": 3.603683612035103e-07, + "loss": 0.9853, + "step": 74447 + }, + { + "epoch": 5.5, + "learning_rate": 3.6026225476240683e-07, + "loss": 0.8531, + "step": 74448 + }, + { + "epoch": 5.5, + "learning_rate": 3.6015616365798866e-07, + "loss": 0.9784, + "step": 74449 + }, + { + "epoch": 5.5, + "learning_rate": 3.6005008789042337e-07, + "loss": 0.9116, + "step": 74450 + }, + { + "epoch": 5.5, + "learning_rate": 3.599440274598809e-07, + "loss": 0.9736, + "step": 74451 + }, + { + "epoch": 5.5, + "learning_rate": 3.598379823665288e-07, + "loss": 1.0067, + "step": 74452 + }, + { + "epoch": 5.5, + "learning_rate": 3.5973195261053695e-07, + "loss": 1.0913, + "step": 74453 + }, + { + "epoch": 5.5, + "learning_rate": 3.596259381920741e-07, + "loss": 0.9265, + "step": 74454 + }, + { + "epoch": 5.5, + "learning_rate": 3.59519939111308e-07, + "loss": 0.8547, + "step": 74455 + }, + { + "epoch": 5.5, + "learning_rate": 3.594139553684073e-07, + "loss": 0.9984, + "step": 74456 + }, + { + "epoch": 5.5, + "learning_rate": 3.5930798696354186e-07, + "loss": 1.0175, + "step": 74457 + }, + { + "epoch": 5.5, + "learning_rate": 3.5920203389687936e-07, + "loss": 1.0021, + "step": 74458 + }, + { + "epoch": 5.5, + "learning_rate": 3.590960961685885e-07, + "loss": 1.026, + "step": 74459 + }, + { + "epoch": 5.5, + "learning_rate": 3.5899017377883817e-07, + "loss": 1.112, + "step": 74460 + }, + { + "epoch": 5.5, + "learning_rate": 3.5888426672779587e-07, + "loss": 1.0209, + "step": 74461 + }, + { + "epoch": 5.5, + "learning_rate": 3.587783750156315e-07, + "loss": 0.8668, + "step": 74462 + }, + { + "epoch": 5.5, + "learning_rate": 3.5867249864251163e-07, + "loss": 1.046, + "step": 74463 + }, + { + "epoch": 5.5, + "learning_rate": 3.585666376086061e-07, + "loss": 1.032, + "step": 74464 + }, + { + "epoch": 5.5, + "learning_rate": 3.5846079191408366e-07, + "loss": 1.0434, + "step": 74465 + }, + { + "epoch": 5.5, + "learning_rate": 3.5835496155911197e-07, + "loss": 0.9532, + "step": 74466 + }, + { + "epoch": 5.5, + "learning_rate": 3.582491465438587e-07, + "loss": 0.9629, + "step": 74467 + }, + { + "epoch": 5.5, + "learning_rate": 3.581433468684947e-07, + "loss": 1.1534, + "step": 74468 + }, + { + "epoch": 5.5, + "learning_rate": 3.580375625331856e-07, + "loss": 0.9764, + "step": 74469 + }, + { + "epoch": 5.5, + "learning_rate": 3.5793179353810216e-07, + "loss": 0.8714, + "step": 74470 + }, + { + "epoch": 5.5, + "learning_rate": 3.578260398834099e-07, + "loss": 0.9812, + "step": 74471 + }, + { + "epoch": 5.5, + "learning_rate": 3.577203015692776e-07, + "loss": 1.0558, + "step": 74472 + }, + { + "epoch": 5.5, + "learning_rate": 3.576145785958762e-07, + "loss": 1.0877, + "step": 74473 + }, + { + "epoch": 5.5, + "learning_rate": 3.575088709633712e-07, + "loss": 1.0292, + "step": 74474 + }, + { + "epoch": 5.5, + "learning_rate": 3.574031786719323e-07, + "loss": 0.991, + "step": 74475 + }, + { + "epoch": 5.5, + "learning_rate": 3.5729750172172505e-07, + "loss": 0.9505, + "step": 74476 + }, + { + "epoch": 5.5, + "learning_rate": 3.571918401129215e-07, + "loss": 0.9798, + "step": 74477 + }, + { + "epoch": 5.5, + "learning_rate": 3.570861938456882e-07, + "loss": 0.9921, + "step": 74478 + }, + { + "epoch": 5.5, + "learning_rate": 3.569805629201917e-07, + "loss": 0.8811, + "step": 74479 + }, + { + "epoch": 5.5, + "learning_rate": 3.5687494733660067e-07, + "loss": 0.9198, + "step": 74480 + }, + { + "epoch": 5.5, + "learning_rate": 3.5676934709508395e-07, + "loss": 1.0564, + "step": 74481 + }, + { + "epoch": 5.5, + "learning_rate": 3.566637621958102e-07, + "loss": 0.927, + "step": 74482 + }, + { + "epoch": 5.5, + "learning_rate": 3.56558192638945e-07, + "loss": 1.0233, + "step": 74483 + }, + { + "epoch": 5.5, + "learning_rate": 3.5645263842465915e-07, + "loss": 0.9705, + "step": 74484 + }, + { + "epoch": 5.5, + "learning_rate": 3.5634709955311706e-07, + "loss": 1.028, + "step": 74485 + }, + { + "epoch": 5.5, + "learning_rate": 3.562415760244919e-07, + "loss": 1.0166, + "step": 74486 + }, + { + "epoch": 5.5, + "learning_rate": 3.5613606783894694e-07, + "loss": 1.034, + "step": 74487 + }, + { + "epoch": 5.5, + "learning_rate": 3.5603057499665085e-07, + "loss": 0.9068, + "step": 74488 + }, + { + "epoch": 5.5, + "learning_rate": 3.559250974977724e-07, + "loss": 0.9805, + "step": 74489 + }, + { + "epoch": 5.5, + "learning_rate": 3.558196353424792e-07, + "loss": 0.9853, + "step": 74490 + }, + { + "epoch": 5.5, + "learning_rate": 3.557141885309401e-07, + "loss": 0.9506, + "step": 74491 + }, + { + "epoch": 5.5, + "learning_rate": 3.556087570633193e-07, + "loss": 0.9916, + "step": 74492 + }, + { + "epoch": 5.5, + "learning_rate": 3.55503340939789e-07, + "loss": 0.981, + "step": 74493 + }, + { + "epoch": 5.5, + "learning_rate": 3.5539794016051455e-07, + "loss": 1.0371, + "step": 74494 + }, + { + "epoch": 5.5, + "learning_rate": 3.552925547256636e-07, + "loss": 1.0662, + "step": 74495 + }, + { + "epoch": 5.5, + "learning_rate": 3.5518718463540494e-07, + "loss": 0.9045, + "step": 74496 + }, + { + "epoch": 5.5, + "learning_rate": 3.550818298899039e-07, + "loss": 1.0926, + "step": 74497 + }, + { + "epoch": 5.5, + "learning_rate": 3.5497649048933045e-07, + "loss": 0.9092, + "step": 74498 + }, + { + "epoch": 5.5, + "learning_rate": 3.5487116643385223e-07, + "loss": 0.9611, + "step": 74499 + }, + { + "epoch": 5.5, + "learning_rate": 3.547658577236346e-07, + "loss": 1.1106, + "step": 74500 + }, + { + "epoch": 5.5, + "learning_rate": 3.546605643588463e-07, + "loss": 1.0666, + "step": 74501 + }, + { + "epoch": 5.5, + "learning_rate": 3.5455528633965843e-07, + "loss": 0.9606, + "step": 74502 + }, + { + "epoch": 5.5, + "learning_rate": 3.544500236662307e-07, + "loss": 0.909, + "step": 74503 + }, + { + "epoch": 5.51, + "learning_rate": 3.543447763387375e-07, + "loss": 1.0205, + "step": 74504 + }, + { + "epoch": 5.51, + "learning_rate": 3.542395443573421e-07, + "loss": 0.9937, + "step": 74505 + }, + { + "epoch": 5.51, + "learning_rate": 3.5413432772221643e-07, + "loss": 0.9374, + "step": 74506 + }, + { + "epoch": 5.51, + "learning_rate": 3.540291264335238e-07, + "loss": 1.0172, + "step": 74507 + }, + { + "epoch": 5.51, + "learning_rate": 3.5392394049143076e-07, + "loss": 1.0431, + "step": 74508 + }, + { + "epoch": 5.51, + "learning_rate": 3.538187698961082e-07, + "loss": 1.0852, + "step": 74509 + }, + { + "epoch": 5.51, + "learning_rate": 3.537136146477227e-07, + "loss": 0.9533, + "step": 74510 + }, + { + "epoch": 5.51, + "learning_rate": 3.5360847474643966e-07, + "loss": 0.9308, + "step": 74511 + }, + { + "epoch": 5.51, + "learning_rate": 3.535033501924279e-07, + "loss": 1.0398, + "step": 74512 + }, + { + "epoch": 5.51, + "learning_rate": 3.5339824098585386e-07, + "loss": 0.962, + "step": 74513 + }, + { + "epoch": 5.51, + "learning_rate": 3.532931471268863e-07, + "loss": 1.0001, + "step": 74514 + }, + { + "epoch": 5.51, + "learning_rate": 3.5318806861569076e-07, + "loss": 1.0769, + "step": 74515 + }, + { + "epoch": 5.51, + "learning_rate": 3.530830054524359e-07, + "loss": 0.9678, + "step": 74516 + }, + { + "epoch": 5.51, + "learning_rate": 3.52977957637286e-07, + "loss": 1.0078, + "step": 74517 + }, + { + "epoch": 5.51, + "learning_rate": 3.5287292517041104e-07, + "loss": 1.0407, + "step": 74518 + }, + { + "epoch": 5.51, + "learning_rate": 3.5276790805197747e-07, + "loss": 0.942, + "step": 74519 + }, + { + "epoch": 5.51, + "learning_rate": 3.5266290628215185e-07, + "loss": 0.9413, + "step": 74520 + }, + { + "epoch": 5.51, + "learning_rate": 3.525579198611007e-07, + "loss": 1.0431, + "step": 74521 + }, + { + "epoch": 5.51, + "learning_rate": 3.5245294878899384e-07, + "loss": 1.079, + "step": 74522 + }, + { + "epoch": 5.51, + "learning_rate": 3.5234799306599453e-07, + "loss": 0.9835, + "step": 74523 + }, + { + "epoch": 5.51, + "learning_rate": 3.5224305269227266e-07, + "loss": 1.0737, + "step": 74524 + }, + { + "epoch": 5.51, + "learning_rate": 3.521381276679947e-07, + "loss": 0.9169, + "step": 74525 + }, + { + "epoch": 5.51, + "learning_rate": 3.52033217993325e-07, + "loss": 1.0476, + "step": 74526 + }, + { + "epoch": 5.51, + "learning_rate": 3.519283236684345e-07, + "loss": 0.9172, + "step": 74527 + }, + { + "epoch": 5.51, + "learning_rate": 3.518234446934854e-07, + "loss": 1.0022, + "step": 74528 + }, + { + "epoch": 5.51, + "learning_rate": 3.5171858106864965e-07, + "loss": 0.9598, + "step": 74529 + }, + { + "epoch": 5.51, + "learning_rate": 3.516137327940894e-07, + "loss": 0.9286, + "step": 74530 + }, + { + "epoch": 5.51, + "learning_rate": 3.515088998699756e-07, + "loss": 1.0932, + "step": 74531 + }, + { + "epoch": 5.51, + "learning_rate": 3.514040822964715e-07, + "loss": 1.0101, + "step": 74532 + }, + { + "epoch": 5.51, + "learning_rate": 3.5129928007374693e-07, + "loss": 1.0471, + "step": 74533 + }, + { + "epoch": 5.51, + "learning_rate": 3.5119449320196617e-07, + "loss": 1.0419, + "step": 74534 + }, + { + "epoch": 5.51, + "learning_rate": 3.510897216812981e-07, + "loss": 0.9427, + "step": 74535 + }, + { + "epoch": 5.51, + "learning_rate": 3.5098496551190684e-07, + "loss": 1.1522, + "step": 74536 + }, + { + "epoch": 5.51, + "learning_rate": 3.508802246939602e-07, + "loss": 1.0958, + "step": 74537 + }, + { + "epoch": 5.51, + "learning_rate": 3.5077549922762576e-07, + "loss": 1.1317, + "step": 74538 + }, + { + "epoch": 5.51, + "learning_rate": 3.5067078911307006e-07, + "loss": 1.0992, + "step": 74539 + }, + { + "epoch": 5.51, + "learning_rate": 3.505660943504585e-07, + "loss": 1.0261, + "step": 74540 + }, + { + "epoch": 5.51, + "learning_rate": 3.5046141493995767e-07, + "loss": 0.9659, + "step": 74541 + }, + { + "epoch": 5.51, + "learning_rate": 3.5035675088173513e-07, + "loss": 1.1241, + "step": 74542 + }, + { + "epoch": 5.51, + "learning_rate": 3.502521021759575e-07, + "loss": 0.9205, + "step": 74543 + }, + { + "epoch": 5.51, + "learning_rate": 3.5014746882279015e-07, + "loss": 1.0886, + "step": 74544 + }, + { + "epoch": 5.51, + "learning_rate": 3.5004285082240074e-07, + "loss": 1.0412, + "step": 74545 + }, + { + "epoch": 5.51, + "learning_rate": 3.4993824817495356e-07, + "loss": 0.9459, + "step": 74546 + }, + { + "epoch": 5.51, + "learning_rate": 3.4983366088061854e-07, + "loss": 1.0802, + "step": 74547 + }, + { + "epoch": 5.51, + "learning_rate": 3.497290889395588e-07, + "loss": 0.9631, + "step": 74548 + }, + { + "epoch": 5.51, + "learning_rate": 3.4962453235194204e-07, + "loss": 0.9329, + "step": 74549 + }, + { + "epoch": 5.51, + "learning_rate": 3.4951999111793366e-07, + "loss": 1.0009, + "step": 74550 + }, + { + "epoch": 5.51, + "learning_rate": 3.494154652377035e-07, + "loss": 1.0029, + "step": 74551 + }, + { + "epoch": 5.51, + "learning_rate": 3.4931095471141484e-07, + "loss": 1.0847, + "step": 74552 + }, + { + "epoch": 5.51, + "learning_rate": 3.4920645953923193e-07, + "loss": 1.0626, + "step": 74553 + }, + { + "epoch": 5.51, + "learning_rate": 3.4910197972132574e-07, + "loss": 1.0233, + "step": 74554 + }, + { + "epoch": 5.51, + "learning_rate": 3.4899751525785954e-07, + "loss": 0.9676, + "step": 74555 + }, + { + "epoch": 5.51, + "learning_rate": 3.4889306614899977e-07, + "loss": 1.0594, + "step": 74556 + }, + { + "epoch": 5.51, + "learning_rate": 3.48788632394913e-07, + "loss": 1.0497, + "step": 74557 + }, + { + "epoch": 5.51, + "learning_rate": 3.486842139957658e-07, + "loss": 1.007, + "step": 74558 + }, + { + "epoch": 5.51, + "learning_rate": 3.4857981095172353e-07, + "loss": 1.0311, + "step": 74559 + }, + { + "epoch": 5.51, + "learning_rate": 3.4847542326295394e-07, + "loss": 0.9932, + "step": 74560 + }, + { + "epoch": 5.51, + "learning_rate": 3.483710509296201e-07, + "loss": 0.9199, + "step": 74561 + }, + { + "epoch": 5.51, + "learning_rate": 3.482666939518908e-07, + "loss": 0.9502, + "step": 74562 + }, + { + "epoch": 5.51, + "learning_rate": 3.481623523299305e-07, + "loss": 0.9953, + "step": 74563 + }, + { + "epoch": 5.51, + "learning_rate": 3.480580260639066e-07, + "loss": 0.9736, + "step": 74564 + }, + { + "epoch": 5.51, + "learning_rate": 3.479537151539847e-07, + "loss": 0.9874, + "step": 74565 + }, + { + "epoch": 5.51, + "learning_rate": 3.478494196003279e-07, + "loss": 1.0079, + "step": 74566 + }, + { + "epoch": 5.51, + "learning_rate": 3.477451394031073e-07, + "loss": 1.0657, + "step": 74567 + }, + { + "epoch": 5.51, + "learning_rate": 3.4764087456248483e-07, + "loss": 1.081, + "step": 74568 + }, + { + "epoch": 5.51, + "learning_rate": 3.475366250786283e-07, + "loss": 1.027, + "step": 74569 + }, + { + "epoch": 5.51, + "learning_rate": 3.474323909517019e-07, + "loss": 1.0021, + "step": 74570 + }, + { + "epoch": 5.51, + "learning_rate": 3.473281721818733e-07, + "loss": 1.0919, + "step": 74571 + }, + { + "epoch": 5.51, + "learning_rate": 3.47223968769308e-07, + "loss": 0.9041, + "step": 74572 + }, + { + "epoch": 5.51, + "learning_rate": 3.4711978071417016e-07, + "loss": 1.0302, + "step": 74573 + }, + { + "epoch": 5.51, + "learning_rate": 3.4701560801662646e-07, + "loss": 1.0051, + "step": 74574 + }, + { + "epoch": 5.51, + "learning_rate": 3.4691145067684226e-07, + "loss": 0.9611, + "step": 74575 + }, + { + "epoch": 5.51, + "learning_rate": 3.4680730869498635e-07, + "loss": 0.9684, + "step": 74576 + }, + { + "epoch": 5.51, + "learning_rate": 3.4670318207121964e-07, + "loss": 1.1069, + "step": 74577 + }, + { + "epoch": 5.51, + "learning_rate": 3.46599070805711e-07, + "loss": 0.9928, + "step": 74578 + }, + { + "epoch": 5.51, + "learning_rate": 3.464949748986246e-07, + "loss": 1.026, + "step": 74579 + }, + { + "epoch": 5.51, + "learning_rate": 3.463908943501282e-07, + "loss": 0.9978, + "step": 74580 + }, + { + "epoch": 5.51, + "learning_rate": 3.4628682916038383e-07, + "loss": 1.063, + "step": 74581 + }, + { + "epoch": 5.51, + "learning_rate": 3.4618277932955915e-07, + "loss": 0.9671, + "step": 74582 + }, + { + "epoch": 5.51, + "learning_rate": 3.460787448578207e-07, + "loss": 1.0463, + "step": 74583 + }, + { + "epoch": 5.51, + "learning_rate": 3.459747257453316e-07, + "loss": 0.9791, + "step": 74584 + }, + { + "epoch": 5.51, + "learning_rate": 3.458707219922597e-07, + "loss": 1.0088, + "step": 74585 + }, + { + "epoch": 5.51, + "learning_rate": 3.4576673359876803e-07, + "loss": 0.9249, + "step": 74586 + }, + { + "epoch": 5.51, + "learning_rate": 3.456627605650242e-07, + "loss": 0.9405, + "step": 74587 + }, + { + "epoch": 5.51, + "learning_rate": 3.455588028911927e-07, + "loss": 0.9756, + "step": 74588 + }, + { + "epoch": 5.51, + "learning_rate": 3.4545486057743995e-07, + "loss": 1.0695, + "step": 74589 + }, + { + "epoch": 5.51, + "learning_rate": 3.4535093362392913e-07, + "loss": 0.8782, + "step": 74590 + }, + { + "epoch": 5.51, + "learning_rate": 3.452470220308268e-07, + "loss": 1.0437, + "step": 74591 + }, + { + "epoch": 5.51, + "learning_rate": 3.451431257982996e-07, + "loss": 1.0304, + "step": 74592 + }, + { + "epoch": 5.51, + "learning_rate": 3.450392449265105e-07, + "loss": 1.0486, + "step": 74593 + }, + { + "epoch": 5.51, + "learning_rate": 3.449353794156252e-07, + "loss": 0.9576, + "step": 74594 + }, + { + "epoch": 5.51, + "learning_rate": 3.4483152926581e-07, + "loss": 1.0096, + "step": 74595 + }, + { + "epoch": 5.51, + "learning_rate": 3.4472769447723153e-07, + "loss": 1.057, + "step": 74596 + }, + { + "epoch": 5.51, + "learning_rate": 3.446238750500508e-07, + "loss": 1.0061, + "step": 74597 + }, + { + "epoch": 5.51, + "learning_rate": 3.445200709844365e-07, + "loss": 1.0983, + "step": 74598 + }, + { + "epoch": 5.51, + "learning_rate": 3.444162822805519e-07, + "loss": 0.9913, + "step": 74599 + }, + { + "epoch": 5.51, + "learning_rate": 3.4431250893856347e-07, + "loss": 0.9728, + "step": 74600 + }, + { + "epoch": 5.51, + "learning_rate": 3.442087509586356e-07, + "loss": 1.0632, + "step": 74601 + }, + { + "epoch": 5.51, + "learning_rate": 3.4410500834093253e-07, + "loss": 0.8371, + "step": 74602 + }, + { + "epoch": 5.51, + "learning_rate": 3.440012810856219e-07, + "loss": 0.9912, + "step": 74603 + }, + { + "epoch": 5.51, + "learning_rate": 3.438975691928659e-07, + "loss": 1.0593, + "step": 74604 + }, + { + "epoch": 5.51, + "learning_rate": 3.4379387266283203e-07, + "loss": 0.9528, + "step": 74605 + }, + { + "epoch": 5.51, + "learning_rate": 3.436901914956814e-07, + "loss": 0.9487, + "step": 74606 + }, + { + "epoch": 5.51, + "learning_rate": 3.435865256915838e-07, + "loss": 1.0021, + "step": 74607 + }, + { + "epoch": 5.51, + "learning_rate": 3.434828752507013e-07, + "loss": 1.0551, + "step": 74608 + }, + { + "epoch": 5.51, + "learning_rate": 3.4337924017319944e-07, + "loss": 1.1142, + "step": 74609 + }, + { + "epoch": 5.51, + "learning_rate": 3.4327562045924245e-07, + "loss": 1.1094, + "step": 74610 + }, + { + "epoch": 5.51, + "learning_rate": 3.4317201610899575e-07, + "loss": 1.0024, + "step": 74611 + }, + { + "epoch": 5.51, + "learning_rate": 3.430684271226248e-07, + "loss": 1.0651, + "step": 74612 + }, + { + "epoch": 5.51, + "learning_rate": 3.4296485350029384e-07, + "loss": 0.9964, + "step": 74613 + }, + { + "epoch": 5.51, + "learning_rate": 3.428612952421673e-07, + "loss": 0.9447, + "step": 74614 + }, + { + "epoch": 5.51, + "learning_rate": 3.4275775234840935e-07, + "loss": 0.9953, + "step": 74615 + }, + { + "epoch": 5.51, + "learning_rate": 3.4265422481918773e-07, + "loss": 1.022, + "step": 74616 + }, + { + "epoch": 5.51, + "learning_rate": 3.4255071265466345e-07, + "loss": 0.9737, + "step": 74617 + }, + { + "epoch": 5.51, + "learning_rate": 3.4244721585500406e-07, + "loss": 0.9571, + "step": 74618 + }, + { + "epoch": 5.51, + "learning_rate": 3.4234373442037284e-07, + "loss": 0.8947, + "step": 74619 + }, + { + "epoch": 5.51, + "learning_rate": 3.4224026835093294e-07, + "loss": 1.0428, + "step": 74620 + }, + { + "epoch": 5.51, + "learning_rate": 3.4213681764685313e-07, + "loss": 0.9357, + "step": 74621 + }, + { + "epoch": 5.51, + "learning_rate": 3.420333823082933e-07, + "loss": 1.0088, + "step": 74622 + }, + { + "epoch": 5.51, + "learning_rate": 3.419299623354211e-07, + "loss": 1.0476, + "step": 74623 + }, + { + "epoch": 5.51, + "learning_rate": 3.4182655772839967e-07, + "loss": 1.0182, + "step": 74624 + }, + { + "epoch": 5.51, + "learning_rate": 3.4172316848739453e-07, + "loss": 1.0355, + "step": 74625 + }, + { + "epoch": 5.51, + "learning_rate": 3.416197946125688e-07, + "loss": 1.0092, + "step": 74626 + }, + { + "epoch": 5.51, + "learning_rate": 3.415164361040879e-07, + "loss": 0.9886, + "step": 74627 + }, + { + "epoch": 5.51, + "learning_rate": 3.4141309296211733e-07, + "loss": 0.9107, + "step": 74628 + }, + { + "epoch": 5.51, + "learning_rate": 3.4130976518682024e-07, + "loss": 1.0529, + "step": 74629 + }, + { + "epoch": 5.51, + "learning_rate": 3.412064527783598e-07, + "loss": 0.9935, + "step": 74630 + }, + { + "epoch": 5.51, + "learning_rate": 3.411031557369016e-07, + "loss": 0.989, + "step": 74631 + }, + { + "epoch": 5.51, + "learning_rate": 3.4099987406261195e-07, + "loss": 1.0226, + "step": 74632 + }, + { + "epoch": 5.51, + "learning_rate": 3.408966077556519e-07, + "loss": 0.9064, + "step": 74633 + }, + { + "epoch": 5.51, + "learning_rate": 3.407933568161881e-07, + "loss": 0.9073, + "step": 74634 + }, + { + "epoch": 5.51, + "learning_rate": 3.406901212443825e-07, + "loss": 0.9557, + "step": 74635 + }, + { + "epoch": 5.51, + "learning_rate": 3.405869010404017e-07, + "loss": 1.0232, + "step": 74636 + }, + { + "epoch": 5.51, + "learning_rate": 3.4048369620441e-07, + "loss": 1.0173, + "step": 74637 + }, + { + "epoch": 5.51, + "learning_rate": 3.4038050673656843e-07, + "loss": 1.0772, + "step": 74638 + }, + { + "epoch": 5.52, + "learning_rate": 3.4027733263704346e-07, + "loss": 1.0096, + "step": 74639 + }, + { + "epoch": 5.52, + "learning_rate": 3.4017417390599936e-07, + "loss": 0.9366, + "step": 74640 + }, + { + "epoch": 5.52, + "learning_rate": 3.400710305436017e-07, + "loss": 1.0419, + "step": 74641 + }, + { + "epoch": 5.52, + "learning_rate": 3.399679025500102e-07, + "loss": 1.0856, + "step": 74642 + }, + { + "epoch": 5.52, + "learning_rate": 3.3986478992539264e-07, + "loss": 0.9354, + "step": 74643 + }, + { + "epoch": 5.52, + "learning_rate": 3.39761692669911e-07, + "loss": 1.0481, + "step": 74644 + }, + { + "epoch": 5.52, + "learning_rate": 3.3965861078373296e-07, + "loss": 1.0384, + "step": 74645 + }, + { + "epoch": 5.52, + "learning_rate": 3.3955554426701844e-07, + "loss": 0.9157, + "step": 74646 + }, + { + "epoch": 5.52, + "learning_rate": 3.394524931199317e-07, + "loss": 1.0508, + "step": 74647 + }, + { + "epoch": 5.52, + "learning_rate": 3.3934945734263925e-07, + "loss": 1.1024, + "step": 74648 + }, + { + "epoch": 5.52, + "learning_rate": 3.392464369353032e-07, + "loss": 0.9497, + "step": 74649 + }, + { + "epoch": 5.52, + "learning_rate": 3.39143431898088e-07, + "loss": 0.9055, + "step": 74650 + }, + { + "epoch": 5.52, + "learning_rate": 3.3904044223115663e-07, + "loss": 0.9547, + "step": 74651 + }, + { + "epoch": 5.52, + "learning_rate": 3.389374679346746e-07, + "loss": 1.0318, + "step": 74652 + }, + { + "epoch": 5.52, + "learning_rate": 3.388345090088041e-07, + "loss": 0.8939, + "step": 74653 + }, + { + "epoch": 5.52, + "learning_rate": 3.387315654537104e-07, + "loss": 1.025, + "step": 74654 + }, + { + "epoch": 5.52, + "learning_rate": 3.3862863726955684e-07, + "loss": 0.91, + "step": 74655 + }, + { + "epoch": 5.52, + "learning_rate": 3.3852572445650433e-07, + "loss": 0.9135, + "step": 74656 + }, + { + "epoch": 5.52, + "learning_rate": 3.384228270147216e-07, + "loss": 1.0518, + "step": 74657 + }, + { + "epoch": 5.52, + "learning_rate": 3.383199449443697e-07, + "loss": 1.157, + "step": 74658 + }, + { + "epoch": 5.52, + "learning_rate": 3.3821707824561177e-07, + "loss": 1.0341, + "step": 74659 + }, + { + "epoch": 5.52, + "learning_rate": 3.381142269186122e-07, + "loss": 1.0084, + "step": 74660 + }, + { + "epoch": 5.52, + "learning_rate": 3.380113909635352e-07, + "loss": 1.0155, + "step": 74661 + }, + { + "epoch": 5.52, + "learning_rate": 3.3790857038054404e-07, + "loss": 1.0581, + "step": 74662 + }, + { + "epoch": 5.52, + "learning_rate": 3.378057651698019e-07, + "loss": 1.0223, + "step": 74663 + }, + { + "epoch": 5.52, + "learning_rate": 3.37702975331472e-07, + "loss": 0.9696, + "step": 74664 + }, + { + "epoch": 5.52, + "learning_rate": 3.3760020086571864e-07, + "loss": 1.035, + "step": 74665 + }, + { + "epoch": 5.52, + "learning_rate": 3.3749744177270727e-07, + "loss": 0.9904, + "step": 74666 + }, + { + "epoch": 5.52, + "learning_rate": 3.373946980525955e-07, + "loss": 0.9968, + "step": 74667 + }, + { + "epoch": 5.52, + "learning_rate": 3.372919697055532e-07, + "loss": 1.0083, + "step": 74668 + }, + { + "epoch": 5.52, + "learning_rate": 3.371892567317403e-07, + "loss": 1.0361, + "step": 74669 + }, + { + "epoch": 5.52, + "learning_rate": 3.3708655913132104e-07, + "loss": 0.9567, + "step": 74670 + }, + { + "epoch": 5.52, + "learning_rate": 3.3698387690445757e-07, + "loss": 1.0294, + "step": 74671 + }, + { + "epoch": 5.52, + "learning_rate": 3.3688121005131636e-07, + "loss": 0.9674, + "step": 74672 + }, + { + "epoch": 5.52, + "learning_rate": 3.3677855857205846e-07, + "loss": 1.0288, + "step": 74673 + }, + { + "epoch": 5.52, + "learning_rate": 3.3667592246684697e-07, + "loss": 0.9153, + "step": 74674 + }, + { + "epoch": 5.52, + "learning_rate": 3.3657330173584633e-07, + "loss": 1.0368, + "step": 74675 + }, + { + "epoch": 5.52, + "learning_rate": 3.364706963792186e-07, + "loss": 0.9454, + "step": 74676 + }, + { + "epoch": 5.52, + "learning_rate": 3.363681063971291e-07, + "loss": 1.0099, + "step": 74677 + }, + { + "epoch": 5.52, + "learning_rate": 3.3626553178973897e-07, + "loss": 1.0063, + "step": 74678 + }, + { + "epoch": 5.52, + "learning_rate": 3.361629725572124e-07, + "loss": 1.0267, + "step": 74679 + }, + { + "epoch": 5.52, + "learning_rate": 3.3606042869971044e-07, + "loss": 1.0435, + "step": 74680 + }, + { + "epoch": 5.52, + "learning_rate": 3.359579002173996e-07, + "loss": 1.0228, + "step": 74681 + }, + { + "epoch": 5.52, + "learning_rate": 3.3585538711044197e-07, + "loss": 1.0692, + "step": 74682 + }, + { + "epoch": 5.52, + "learning_rate": 3.357528893789996e-07, + "loss": 0.9863, + "step": 74683 + }, + { + "epoch": 5.52, + "learning_rate": 3.3565040702323693e-07, + "loss": 1.0863, + "step": 74684 + }, + { + "epoch": 5.52, + "learning_rate": 3.355479400433137e-07, + "loss": 1.0355, + "step": 74685 + }, + { + "epoch": 5.52, + "learning_rate": 3.354454884393987e-07, + "loss": 0.9646, + "step": 74686 + }, + { + "epoch": 5.52, + "learning_rate": 3.3534305221164854e-07, + "loss": 0.9642, + "step": 74687 + }, + { + "epoch": 5.52, + "learning_rate": 3.352406313602319e-07, + "loss": 0.9603, + "step": 74688 + }, + { + "epoch": 5.52, + "learning_rate": 3.3513822588530644e-07, + "loss": 1.134, + "step": 74689 + }, + { + "epoch": 5.52, + "learning_rate": 3.3503583578704093e-07, + "loss": 0.9383, + "step": 74690 + }, + { + "epoch": 5.52, + "learning_rate": 3.3493346106559296e-07, + "loss": 1.0273, + "step": 74691 + }, + { + "epoch": 5.52, + "learning_rate": 3.348311017211281e-07, + "loss": 0.8689, + "step": 74692 + }, + { + "epoch": 5.52, + "learning_rate": 3.347287577538094e-07, + "loss": 0.9533, + "step": 74693 + }, + { + "epoch": 5.52, + "learning_rate": 3.3462642916379795e-07, + "loss": 1.0455, + "step": 74694 + }, + { + "epoch": 5.52, + "learning_rate": 3.34524115951258e-07, + "loss": 1.0326, + "step": 74695 + }, + { + "epoch": 5.52, + "learning_rate": 3.3442181811635165e-07, + "loss": 1.0336, + "step": 74696 + }, + { + "epoch": 5.52, + "learning_rate": 3.3431953565924214e-07, + "loss": 0.9601, + "step": 74697 + }, + { + "epoch": 5.52, + "learning_rate": 3.342172685800915e-07, + "loss": 0.7947, + "step": 74698 + }, + { + "epoch": 5.52, + "learning_rate": 3.3411501687906414e-07, + "loss": 0.8697, + "step": 74699 + }, + { + "epoch": 5.52, + "learning_rate": 3.340127805563198e-07, + "loss": 0.9796, + "step": 74700 + }, + { + "epoch": 5.52, + "learning_rate": 3.339105596120229e-07, + "loss": 1.0021, + "step": 74701 + }, + { + "epoch": 5.52, + "learning_rate": 3.3380835404633774e-07, + "loss": 1.1052, + "step": 74702 + }, + { + "epoch": 5.52, + "learning_rate": 3.3370616385942413e-07, + "loss": 0.9745, + "step": 74703 + }, + { + "epoch": 5.52, + "learning_rate": 3.3360398905144644e-07, + "loss": 1.0408, + "step": 74704 + }, + { + "epoch": 5.52, + "learning_rate": 3.335018296225645e-07, + "loss": 0.9626, + "step": 74705 + }, + { + "epoch": 5.52, + "learning_rate": 3.3339968557294376e-07, + "loss": 1.0181, + "step": 74706 + }, + { + "epoch": 5.52, + "learning_rate": 3.3329755690274747e-07, + "loss": 1.1454, + "step": 74707 + }, + { + "epoch": 5.52, + "learning_rate": 3.331954436121354e-07, + "loss": 1.0895, + "step": 74708 + }, + { + "epoch": 5.52, + "learning_rate": 3.3309334570126973e-07, + "loss": 1.0119, + "step": 74709 + }, + { + "epoch": 5.52, + "learning_rate": 3.329912631703158e-07, + "loss": 1.0854, + "step": 74710 + }, + { + "epoch": 5.52, + "learning_rate": 3.328891960194347e-07, + "loss": 0.9927, + "step": 74711 + }, + { + "epoch": 5.52, + "learning_rate": 3.327871442487873e-07, + "loss": 0.9239, + "step": 74712 + }, + { + "epoch": 5.52, + "learning_rate": 3.32685107858538e-07, + "loss": 0.9699, + "step": 74713 + }, + { + "epoch": 5.52, + "learning_rate": 3.325830868488489e-07, + "loss": 1.0587, + "step": 74714 + }, + { + "epoch": 5.52, + "learning_rate": 3.324810812198809e-07, + "loss": 0.926, + "step": 74715 + }, + { + "epoch": 5.52, + "learning_rate": 3.323790909717961e-07, + "loss": 1.0176, + "step": 74716 + }, + { + "epoch": 5.52, + "learning_rate": 3.3227711610476e-07, + "loss": 1.1064, + "step": 74717 + }, + { + "epoch": 5.52, + "learning_rate": 3.321751566189324e-07, + "loss": 0.9793, + "step": 74718 + }, + { + "epoch": 5.52, + "learning_rate": 3.3207321251447546e-07, + "loss": 0.9073, + "step": 74719 + }, + { + "epoch": 5.52, + "learning_rate": 3.319712837915501e-07, + "loss": 0.8867, + "step": 74720 + }, + { + "epoch": 5.52, + "learning_rate": 3.3186937045032174e-07, + "loss": 1.0408, + "step": 74721 + }, + { + "epoch": 5.52, + "learning_rate": 3.317674724909514e-07, + "loss": 1.0533, + "step": 74722 + }, + { + "epoch": 5.52, + "learning_rate": 3.316655899136001e-07, + "loss": 0.9555, + "step": 74723 + }, + { + "epoch": 5.52, + "learning_rate": 3.3156372271842983e-07, + "loss": 1.0594, + "step": 74724 + }, + { + "epoch": 5.52, + "learning_rate": 3.3146187090560277e-07, + "loss": 0.9742, + "step": 74725 + }, + { + "epoch": 5.52, + "learning_rate": 3.313600344752832e-07, + "loss": 1.0607, + "step": 74726 + }, + { + "epoch": 5.52, + "learning_rate": 3.31258213427631e-07, + "loss": 0.9963, + "step": 74727 + }, + { + "epoch": 5.52, + "learning_rate": 3.3115640776280934e-07, + "loss": 0.9494, + "step": 74728 + }, + { + "epoch": 5.52, + "learning_rate": 3.3105461748097703e-07, + "loss": 1.0022, + "step": 74729 + }, + { + "epoch": 5.52, + "learning_rate": 3.3095284258230055e-07, + "loss": 1.0121, + "step": 74730 + }, + { + "epoch": 5.52, + "learning_rate": 3.308510830669409e-07, + "loss": 0.8932, + "step": 74731 + }, + { + "epoch": 5.52, + "learning_rate": 3.307493389350569e-07, + "loss": 1.0412, + "step": 74732 + }, + { + "epoch": 5.52, + "learning_rate": 3.3064761018681277e-07, + "loss": 1.0557, + "step": 74733 + }, + { + "epoch": 5.52, + "learning_rate": 3.305458968223685e-07, + "loss": 0.9596, + "step": 74734 + }, + { + "epoch": 5.52, + "learning_rate": 3.3044419884189047e-07, + "loss": 1.0069, + "step": 74735 + }, + { + "epoch": 5.52, + "learning_rate": 3.3034251624553426e-07, + "loss": 0.9769, + "step": 74736 + }, + { + "epoch": 5.52, + "learning_rate": 3.302408490334663e-07, + "loss": 1.0075, + "step": 74737 + }, + { + "epoch": 5.52, + "learning_rate": 3.3013919720584654e-07, + "loss": 0.8956, + "step": 74738 + }, + { + "epoch": 5.52, + "learning_rate": 3.300375607628381e-07, + "loss": 1.0505, + "step": 74739 + }, + { + "epoch": 5.52, + "learning_rate": 3.299359397046009e-07, + "loss": 0.9385, + "step": 74740 + }, + { + "epoch": 5.52, + "learning_rate": 3.2983433403129594e-07, + "loss": 0.9504, + "step": 74741 + }, + { + "epoch": 5.52, + "learning_rate": 3.297327437430875e-07, + "loss": 0.976, + "step": 74742 + }, + { + "epoch": 5.52, + "learning_rate": 3.296311688401355e-07, + "loss": 1.1146, + "step": 74743 + }, + { + "epoch": 5.52, + "learning_rate": 3.29529609322603e-07, + "loss": 0.9947, + "step": 74744 + }, + { + "epoch": 5.52, + "learning_rate": 3.294280651906479e-07, + "loss": 1.0443, + "step": 74745 + }, + { + "epoch": 5.52, + "learning_rate": 3.2932653644443647e-07, + "loss": 1.0045, + "step": 74746 + }, + { + "epoch": 5.52, + "learning_rate": 3.292250230841276e-07, + "loss": 1.0592, + "step": 74747 + }, + { + "epoch": 5.52, + "learning_rate": 3.291235251098845e-07, + "loss": 1.1119, + "step": 74748 + }, + { + "epoch": 5.52, + "learning_rate": 3.2902204252186596e-07, + "loss": 0.9275, + "step": 74749 + }, + { + "epoch": 5.52, + "learning_rate": 3.2892057532023513e-07, + "loss": 0.9563, + "step": 74750 + }, + { + "epoch": 5.52, + "learning_rate": 3.288191235051541e-07, + "loss": 1.0042, + "step": 74751 + }, + { + "epoch": 5.52, + "learning_rate": 3.2871768707678276e-07, + "loss": 1.0195, + "step": 74752 + }, + { + "epoch": 5.52, + "learning_rate": 3.286162660352843e-07, + "loss": 1.0511, + "step": 74753 + }, + { + "epoch": 5.52, + "learning_rate": 3.2851486038081635e-07, + "loss": 0.9325, + "step": 74754 + }, + { + "epoch": 5.52, + "learning_rate": 3.284134701135455e-07, + "loss": 1.0691, + "step": 74755 + }, + { + "epoch": 5.52, + "learning_rate": 3.283120952336294e-07, + "loss": 1.0358, + "step": 74756 + }, + { + "epoch": 5.52, + "learning_rate": 3.282107357412312e-07, + "loss": 0.899, + "step": 74757 + }, + { + "epoch": 5.52, + "learning_rate": 3.281093916365108e-07, + "loss": 1.0577, + "step": 74758 + }, + { + "epoch": 5.52, + "learning_rate": 3.280080629196303e-07, + "loss": 1.1525, + "step": 74759 + }, + { + "epoch": 5.52, + "learning_rate": 3.279067495907506e-07, + "loss": 1.0772, + "step": 74760 + }, + { + "epoch": 5.52, + "learning_rate": 3.2780545165003176e-07, + "loss": 1.0995, + "step": 74761 + }, + { + "epoch": 5.52, + "learning_rate": 3.277041690976368e-07, + "loss": 1.1049, + "step": 74762 + }, + { + "epoch": 5.52, + "learning_rate": 3.276029019337268e-07, + "loss": 0.9879, + "step": 74763 + }, + { + "epoch": 5.52, + "learning_rate": 3.275016501584616e-07, + "loss": 0.9069, + "step": 74764 + }, + { + "epoch": 5.52, + "learning_rate": 3.274004137720022e-07, + "loss": 1.0105, + "step": 74765 + }, + { + "epoch": 5.52, + "learning_rate": 3.2729919277451063e-07, + "loss": 0.9605, + "step": 74766 + }, + { + "epoch": 5.52, + "learning_rate": 3.271979871661479e-07, + "loss": 0.9533, + "step": 74767 + }, + { + "epoch": 5.52, + "learning_rate": 3.270967969470751e-07, + "loss": 1.0565, + "step": 74768 + }, + { + "epoch": 5.52, + "learning_rate": 3.2699562211745303e-07, + "loss": 0.9353, + "step": 74769 + }, + { + "epoch": 5.52, + "learning_rate": 3.2689446267744173e-07, + "loss": 0.9531, + "step": 74770 + }, + { + "epoch": 5.52, + "learning_rate": 3.267933186272032e-07, + "loss": 0.9181, + "step": 74771 + }, + { + "epoch": 5.52, + "learning_rate": 3.2669218996689845e-07, + "loss": 1.0179, + "step": 74772 + }, + { + "epoch": 5.52, + "learning_rate": 3.265910766966884e-07, + "loss": 1.0733, + "step": 74773 + }, + { + "epoch": 5.53, + "learning_rate": 3.2648997881673197e-07, + "loss": 1.0002, + "step": 74774 + }, + { + "epoch": 5.53, + "learning_rate": 3.2638889632719217e-07, + "loss": 1.0312, + "step": 74775 + }, + { + "epoch": 5.53, + "learning_rate": 3.262878292282301e-07, + "loss": 1.0515, + "step": 74776 + }, + { + "epoch": 5.53, + "learning_rate": 3.261867775200056e-07, + "loss": 0.9621, + "step": 74777 + }, + { + "epoch": 5.53, + "learning_rate": 3.2608574120267967e-07, + "loss": 1.0412, + "step": 74778 + }, + { + "epoch": 5.53, + "learning_rate": 3.25984720276411e-07, + "loss": 1.0958, + "step": 74779 + }, + { + "epoch": 5.53, + "learning_rate": 3.2588371474136514e-07, + "loss": 1.0438, + "step": 74780 + }, + { + "epoch": 5.53, + "learning_rate": 3.2578272459769745e-07, + "loss": 0.9296, + "step": 74781 + }, + { + "epoch": 5.53, + "learning_rate": 3.256817498455722e-07, + "loss": 0.8974, + "step": 74782 + }, + { + "epoch": 5.53, + "learning_rate": 3.255807904851482e-07, + "loss": 1.0382, + "step": 74783 + }, + { + "epoch": 5.53, + "learning_rate": 3.254798465165876e-07, + "loss": 0.9936, + "step": 74784 + }, + { + "epoch": 5.53, + "learning_rate": 3.2537891794004903e-07, + "loss": 1.029, + "step": 74785 + }, + { + "epoch": 5.53, + "learning_rate": 3.252780047556958e-07, + "loss": 1.0157, + "step": 74786 + }, + { + "epoch": 5.53, + "learning_rate": 3.251771069636855e-07, + "loss": 0.9985, + "step": 74787 + }, + { + "epoch": 5.53, + "learning_rate": 3.2507622456418033e-07, + "loss": 0.9934, + "step": 74788 + }, + { + "epoch": 5.53, + "learning_rate": 3.249753575573411e-07, + "loss": 1.054, + "step": 74789 + }, + { + "epoch": 5.53, + "learning_rate": 3.248745059433267e-07, + "loss": 0.9151, + "step": 74790 + }, + { + "epoch": 5.53, + "learning_rate": 3.247736697222992e-07, + "loss": 0.9707, + "step": 74791 + }, + { + "epoch": 5.53, + "learning_rate": 3.246728488944173e-07, + "loss": 1.0874, + "step": 74792 + }, + { + "epoch": 5.53, + "learning_rate": 3.245720434598443e-07, + "loss": 0.9962, + "step": 74793 + }, + { + "epoch": 5.53, + "learning_rate": 3.2447125341873665e-07, + "loss": 0.9581, + "step": 74794 + }, + { + "epoch": 5.53, + "learning_rate": 3.243704787712576e-07, + "loss": 0.919, + "step": 74795 + }, + { + "epoch": 5.53, + "learning_rate": 3.24269719517567e-07, + "loss": 0.9129, + "step": 74796 + }, + { + "epoch": 5.53, + "learning_rate": 3.241689756578259e-07, + "loss": 0.9095, + "step": 74797 + }, + { + "epoch": 5.53, + "learning_rate": 3.240682471921919e-07, + "loss": 1.0401, + "step": 74798 + }, + { + "epoch": 5.53, + "learning_rate": 3.2396753412082704e-07, + "loss": 0.8943, + "step": 74799 + }, + { + "epoch": 5.53, + "learning_rate": 3.238668364438913e-07, + "loss": 1.0053, + "step": 74800 + }, + { + "epoch": 5.53, + "learning_rate": 3.2376615416154556e-07, + "loss": 0.9514, + "step": 74801 + }, + { + "epoch": 5.53, + "learning_rate": 3.2366548727394975e-07, + "loss": 0.927, + "step": 74802 + }, + { + "epoch": 5.53, + "learning_rate": 3.235648357812626e-07, + "loss": 0.9024, + "step": 74803 + }, + { + "epoch": 5.53, + "learning_rate": 3.234641996836474e-07, + "loss": 1.0023, + "step": 74804 + }, + { + "epoch": 5.53, + "learning_rate": 3.233635789812606e-07, + "loss": 0.959, + "step": 74805 + }, + { + "epoch": 5.53, + "learning_rate": 3.232629736742632e-07, + "loss": 1.0311, + "step": 74806 + }, + { + "epoch": 5.53, + "learning_rate": 3.231623837628173e-07, + "loss": 0.9992, + "step": 74807 + }, + { + "epoch": 5.53, + "learning_rate": 3.2306180924708055e-07, + "loss": 0.9446, + "step": 74808 + }, + { + "epoch": 5.53, + "learning_rate": 3.2296125012721504e-07, + "loss": 0.9964, + "step": 74809 + }, + { + "epoch": 5.53, + "learning_rate": 3.228607064033784e-07, + "loss": 0.9367, + "step": 74810 + }, + { + "epoch": 5.53, + "learning_rate": 3.2276017807573276e-07, + "loss": 0.8627, + "step": 74811 + }, + { + "epoch": 5.53, + "learning_rate": 3.22659665144438e-07, + "loss": 0.9938, + "step": 74812 + }, + { + "epoch": 5.53, + "learning_rate": 3.225591676096529e-07, + "loss": 1.0434, + "step": 74813 + }, + { + "epoch": 5.53, + "learning_rate": 3.2245868547153724e-07, + "loss": 1.0158, + "step": 74814 + }, + { + "epoch": 5.53, + "learning_rate": 3.2235821873025096e-07, + "loss": 0.9505, + "step": 74815 + }, + { + "epoch": 5.53, + "learning_rate": 3.222577673859539e-07, + "loss": 1.0737, + "step": 74816 + }, + { + "epoch": 5.53, + "learning_rate": 3.2215733143880825e-07, + "loss": 0.9141, + "step": 74817 + }, + { + "epoch": 5.53, + "learning_rate": 3.220569108889704e-07, + "loss": 0.9461, + "step": 74818 + }, + { + "epoch": 5.53, + "learning_rate": 3.2195650573660143e-07, + "loss": 0.9056, + "step": 74819 + }, + { + "epoch": 5.53, + "learning_rate": 3.2185611598186115e-07, + "loss": 1.0526, + "step": 74820 + }, + { + "epoch": 5.53, + "learning_rate": 3.2175574162491063e-07, + "loss": 1.0141, + "step": 74821 + }, + { + "epoch": 5.53, + "learning_rate": 3.2165538266590746e-07, + "loss": 1.1267, + "step": 74822 + }, + { + "epoch": 5.53, + "learning_rate": 3.215550391050104e-07, + "loss": 0.9407, + "step": 74823 + }, + { + "epoch": 5.53, + "learning_rate": 3.214547109423827e-07, + "loss": 0.97, + "step": 74824 + }, + { + "epoch": 5.53, + "learning_rate": 3.2135439817818304e-07, + "loss": 1.0033, + "step": 74825 + }, + { + "epoch": 5.53, + "learning_rate": 3.21254100812568e-07, + "loss": 1.0037, + "step": 74826 + }, + { + "epoch": 5.53, + "learning_rate": 3.2115381884570085e-07, + "loss": 0.9398, + "step": 74827 + }, + { + "epoch": 5.53, + "learning_rate": 3.210535522777369e-07, + "loss": 0.9346, + "step": 74828 + }, + { + "epoch": 5.53, + "learning_rate": 3.209533011088417e-07, + "loss": 1.0408, + "step": 74829 + }, + { + "epoch": 5.53, + "learning_rate": 3.2085306533916837e-07, + "loss": 0.9819, + "step": 74830 + }, + { + "epoch": 5.53, + "learning_rate": 3.2075284496888123e-07, + "loss": 0.9287, + "step": 74831 + }, + { + "epoch": 5.53, + "learning_rate": 3.206526399981358e-07, + "loss": 1.013, + "step": 74832 + }, + { + "epoch": 5.53, + "learning_rate": 3.205524504270963e-07, + "loss": 0.9235, + "step": 74833 + }, + { + "epoch": 5.53, + "learning_rate": 3.204522762559181e-07, + "loss": 0.978, + "step": 74834 + }, + { + "epoch": 5.53, + "learning_rate": 3.2035211748476123e-07, + "loss": 1.0026, + "step": 74835 + }, + { + "epoch": 5.53, + "learning_rate": 3.2025197411378663e-07, + "loss": 1.0409, + "step": 74836 + }, + { + "epoch": 5.53, + "learning_rate": 3.2015184614315184e-07, + "loss": 0.9883, + "step": 74837 + }, + { + "epoch": 5.53, + "learning_rate": 3.20051733573018e-07, + "loss": 1.0428, + "step": 74838 + }, + { + "epoch": 5.53, + "learning_rate": 3.1995163640354154e-07, + "loss": 0.9656, + "step": 74839 + }, + { + "epoch": 5.53, + "learning_rate": 3.198515546348846e-07, + "loss": 0.9849, + "step": 74840 + }, + { + "epoch": 5.53, + "learning_rate": 3.197514882672059e-07, + "loss": 1.0364, + "step": 74841 + }, + { + "epoch": 5.53, + "learning_rate": 3.196514373006632e-07, + "loss": 0.9092, + "step": 74842 + }, + { + "epoch": 5.53, + "learning_rate": 3.195514017354173e-07, + "loss": 1.0141, + "step": 74843 + }, + { + "epoch": 5.53, + "learning_rate": 3.194513815716249e-07, + "loss": 1.0919, + "step": 74844 + }, + { + "epoch": 5.53, + "learning_rate": 3.193513768094503e-07, + "loss": 1.0036, + "step": 74845 + }, + { + "epoch": 5.53, + "learning_rate": 3.1925138744904547e-07, + "loss": 0.9887, + "step": 74846 + }, + { + "epoch": 5.53, + "learning_rate": 3.191514134905749e-07, + "loss": 1.0372, + "step": 74847 + }, + { + "epoch": 5.53, + "learning_rate": 3.19051454934195e-07, + "loss": 0.9972, + "step": 74848 + }, + { + "epoch": 5.53, + "learning_rate": 3.189515117800679e-07, + "loss": 0.9933, + "step": 74849 + }, + { + "epoch": 5.53, + "learning_rate": 3.188515840283479e-07, + "loss": 1.0751, + "step": 74850 + }, + { + "epoch": 5.53, + "learning_rate": 3.187516716791983e-07, + "loss": 1.0104, + "step": 74851 + }, + { + "epoch": 5.53, + "learning_rate": 3.1865177473277555e-07, + "loss": 1.0346, + "step": 74852 + }, + { + "epoch": 5.53, + "learning_rate": 3.1855189318923954e-07, + "loss": 1.0308, + "step": 74853 + }, + { + "epoch": 5.53, + "learning_rate": 3.18452027048749e-07, + "loss": 0.9986, + "step": 74854 + }, + { + "epoch": 5.53, + "learning_rate": 3.1835217631146163e-07, + "loss": 1.0141, + "step": 74855 + }, + { + "epoch": 5.53, + "learning_rate": 3.1825234097753844e-07, + "loss": 1.0791, + "step": 74856 + }, + { + "epoch": 5.53, + "learning_rate": 3.18152521047137e-07, + "loss": 0.964, + "step": 74857 + }, + { + "epoch": 5.53, + "learning_rate": 3.180527165204161e-07, + "loss": 1.0096, + "step": 74858 + }, + { + "epoch": 5.53, + "learning_rate": 3.1795292739753457e-07, + "loss": 1.0351, + "step": 74859 + }, + { + "epoch": 5.53, + "learning_rate": 3.1785315367865223e-07, + "loss": 1.0494, + "step": 74860 + }, + { + "epoch": 5.53, + "learning_rate": 3.177533953639278e-07, + "loss": 0.8934, + "step": 74861 + }, + { + "epoch": 5.53, + "learning_rate": 3.1765365245351785e-07, + "loss": 0.9573, + "step": 74862 + }, + { + "epoch": 5.53, + "learning_rate": 3.175539249475834e-07, + "loss": 0.9675, + "step": 74863 + }, + { + "epoch": 5.53, + "learning_rate": 3.174542128462799e-07, + "loss": 1.0362, + "step": 74864 + }, + { + "epoch": 5.53, + "learning_rate": 3.173545161497704e-07, + "loss": 0.9368, + "step": 74865 + }, + { + "epoch": 5.53, + "learning_rate": 3.172548348582105e-07, + "loss": 1.0059, + "step": 74866 + }, + { + "epoch": 5.53, + "learning_rate": 3.1715516897176004e-07, + "loss": 0.9776, + "step": 74867 + }, + { + "epoch": 5.53, + "learning_rate": 3.170555184905766e-07, + "loss": 1.003, + "step": 74868 + }, + { + "epoch": 5.53, + "learning_rate": 3.16955883414819e-07, + "loss": 0.9637, + "step": 74869 + }, + { + "epoch": 5.53, + "learning_rate": 3.168562637446482e-07, + "loss": 0.9703, + "step": 74870 + }, + { + "epoch": 5.53, + "learning_rate": 3.167566594802185e-07, + "loss": 1.0049, + "step": 74871 + }, + { + "epoch": 5.53, + "learning_rate": 3.1665707062169094e-07, + "loss": 1.0446, + "step": 74872 + }, + { + "epoch": 5.53, + "learning_rate": 3.1655749716922204e-07, + "loss": 0.9892, + "step": 74873 + }, + { + "epoch": 5.53, + "learning_rate": 3.1645793912297495e-07, + "loss": 0.9304, + "step": 74874 + }, + { + "epoch": 5.53, + "learning_rate": 3.1635839648310187e-07, + "loss": 0.9919, + "step": 74875 + }, + { + "epoch": 5.53, + "learning_rate": 3.162588692497648e-07, + "loss": 1.0316, + "step": 74876 + }, + { + "epoch": 5.53, + "learning_rate": 3.161593574231203e-07, + "loss": 0.9079, + "step": 74877 + }, + { + "epoch": 5.53, + "learning_rate": 3.1605986100333054e-07, + "loss": 0.8383, + "step": 74878 + }, + { + "epoch": 5.53, + "learning_rate": 3.1596037999054863e-07, + "loss": 1.0294, + "step": 74879 + }, + { + "epoch": 5.53, + "learning_rate": 3.158609143849356e-07, + "loss": 1.0052, + "step": 74880 + }, + { + "epoch": 5.53, + "learning_rate": 3.157614641866502e-07, + "loss": 1.0068, + "step": 74881 + }, + { + "epoch": 5.53, + "learning_rate": 3.15662029395849e-07, + "loss": 1.0033, + "step": 74882 + }, + { + "epoch": 5.53, + "learning_rate": 3.1556261001269183e-07, + "loss": 1.0291, + "step": 74883 + }, + { + "epoch": 5.53, + "learning_rate": 3.1546320603733417e-07, + "loss": 1.1257, + "step": 74884 + }, + { + "epoch": 5.53, + "learning_rate": 3.1536381746993696e-07, + "loss": 0.8001, + "step": 74885 + }, + { + "epoch": 5.53, + "learning_rate": 3.152644443106578e-07, + "loss": 0.9852, + "step": 74886 + }, + { + "epoch": 5.53, + "learning_rate": 3.151650865596545e-07, + "loss": 1.0084, + "step": 74887 + }, + { + "epoch": 5.53, + "learning_rate": 3.150657442170835e-07, + "loss": 1.0448, + "step": 74888 + }, + { + "epoch": 5.53, + "learning_rate": 3.149664172831057e-07, + "loss": 1.0287, + "step": 74889 + }, + { + "epoch": 5.53, + "learning_rate": 3.148671057578778e-07, + "loss": 1.0199, + "step": 74890 + }, + { + "epoch": 5.53, + "learning_rate": 3.147678096415574e-07, + "loss": 1.0268, + "step": 74891 + }, + { + "epoch": 5.53, + "learning_rate": 3.146685289343021e-07, + "loss": 1.1198, + "step": 74892 + }, + { + "epoch": 5.53, + "learning_rate": 3.1456926363627073e-07, + "loss": 1.0328, + "step": 74893 + }, + { + "epoch": 5.53, + "learning_rate": 3.1447001374762307e-07, + "loss": 1.0539, + "step": 74894 + }, + { + "epoch": 5.53, + "learning_rate": 3.1437077926851246e-07, + "loss": 1.0049, + "step": 74895 + }, + { + "epoch": 5.53, + "learning_rate": 3.142715601991009e-07, + "loss": 0.982, + "step": 74896 + }, + { + "epoch": 5.53, + "learning_rate": 3.141723565395438e-07, + "loss": 0.9758, + "step": 74897 + }, + { + "epoch": 5.53, + "learning_rate": 3.140731682900011e-07, + "loss": 1.0214, + "step": 74898 + }, + { + "epoch": 5.53, + "learning_rate": 3.1397399545062935e-07, + "loss": 0.8069, + "step": 74899 + }, + { + "epoch": 5.53, + "learning_rate": 3.13874838021585e-07, + "loss": 0.8777, + "step": 74900 + }, + { + "epoch": 5.53, + "learning_rate": 3.1377569600302804e-07, + "loss": 1.0268, + "step": 74901 + }, + { + "epoch": 5.53, + "learning_rate": 3.136765693951149e-07, + "loss": 0.9762, + "step": 74902 + }, + { + "epoch": 5.53, + "learning_rate": 3.1357745819800446e-07, + "loss": 1.0619, + "step": 74903 + }, + { + "epoch": 5.53, + "learning_rate": 3.1347836241185205e-07, + "loss": 1.0012, + "step": 74904 + }, + { + "epoch": 5.53, + "learning_rate": 3.133792820368187e-07, + "loss": 0.991, + "step": 74905 + }, + { + "epoch": 5.53, + "learning_rate": 3.1328021707305977e-07, + "loss": 0.9648, + "step": 74906 + }, + { + "epoch": 5.53, + "learning_rate": 3.1318116752073415e-07, + "loss": 0.9431, + "step": 74907 + }, + { + "epoch": 5.53, + "learning_rate": 3.1308213337999825e-07, + "loss": 0.9109, + "step": 74908 + }, + { + "epoch": 5.53, + "learning_rate": 3.129831146510087e-07, + "loss": 0.9581, + "step": 74909 + }, + { + "epoch": 5.54, + "learning_rate": 3.128841113339265e-07, + "loss": 1.0115, + "step": 74910 + }, + { + "epoch": 5.54, + "learning_rate": 3.1278512342890585e-07, + "loss": 0.9677, + "step": 74911 + }, + { + "epoch": 5.54, + "learning_rate": 3.1268615093610676e-07, + "loss": 1.0357, + "step": 74912 + }, + { + "epoch": 5.54, + "learning_rate": 3.125871938556835e-07, + "loss": 1.0614, + "step": 74913 + }, + { + "epoch": 5.54, + "learning_rate": 3.1248825218779586e-07, + "loss": 1.0621, + "step": 74914 + }, + { + "epoch": 5.54, + "learning_rate": 3.1238932593260164e-07, + "loss": 1.0326, + "step": 74915 + }, + { + "epoch": 5.54, + "learning_rate": 3.1229041509025726e-07, + "loss": 1.0463, + "step": 74916 + }, + { + "epoch": 5.54, + "learning_rate": 3.1219151966092043e-07, + "loss": 1.0324, + "step": 74917 + }, + { + "epoch": 5.54, + "learning_rate": 3.120926396447477e-07, + "loss": 0.9714, + "step": 74918 + }, + { + "epoch": 5.54, + "learning_rate": 3.1199377504189776e-07, + "loss": 1.0249, + "step": 74919 + }, + { + "epoch": 5.54, + "learning_rate": 3.118949258525261e-07, + "loss": 0.9895, + "step": 74920 + }, + { + "epoch": 5.54, + "learning_rate": 3.117960920767915e-07, + "loss": 0.9945, + "step": 74921 + }, + { + "epoch": 5.54, + "learning_rate": 3.1169727371484935e-07, + "loss": 1.0086, + "step": 74922 + }, + { + "epoch": 5.54, + "learning_rate": 3.1159847076686066e-07, + "loss": 0.9657, + "step": 74923 + }, + { + "epoch": 5.54, + "learning_rate": 3.1149968323297865e-07, + "loss": 0.9748, + "step": 74924 + }, + { + "epoch": 5.54, + "learning_rate": 3.1140091111336204e-07, + "loss": 0.9011, + "step": 74925 + }, + { + "epoch": 5.54, + "learning_rate": 3.1130215440816847e-07, + "loss": 1.0586, + "step": 74926 + }, + { + "epoch": 5.54, + "learning_rate": 3.1120341311755454e-07, + "loss": 0.9818, + "step": 74927 + }, + { + "epoch": 5.54, + "learning_rate": 3.1110468724167786e-07, + "loss": 1.0082, + "step": 74928 + }, + { + "epoch": 5.54, + "learning_rate": 3.110059767806939e-07, + "loss": 1.0721, + "step": 74929 + }, + { + "epoch": 5.54, + "learning_rate": 3.109072817347614e-07, + "loss": 0.9845, + "step": 74930 + }, + { + "epoch": 5.54, + "learning_rate": 3.1080860210403685e-07, + "loss": 0.9764, + "step": 74931 + }, + { + "epoch": 5.54, + "learning_rate": 3.107099378886769e-07, + "loss": 0.8726, + "step": 74932 + }, + { + "epoch": 5.54, + "learning_rate": 3.10611289088838e-07, + "loss": 0.9754, + "step": 74933 + }, + { + "epoch": 5.54, + "learning_rate": 3.1051265570467894e-07, + "loss": 0.9995, + "step": 74934 + }, + { + "epoch": 5.54, + "learning_rate": 3.1041403773635514e-07, + "loss": 0.9853, + "step": 74935 + }, + { + "epoch": 5.54, + "learning_rate": 3.103154351840254e-07, + "loss": 1.0196, + "step": 74936 + }, + { + "epoch": 5.54, + "learning_rate": 3.10216848047844e-07, + "loss": 1.0829, + "step": 74937 + }, + { + "epoch": 5.54, + "learning_rate": 3.101182763279675e-07, + "loss": 1.0201, + "step": 74938 + }, + { + "epoch": 5.54, + "learning_rate": 3.100197200245569e-07, + "loss": 1.0259, + "step": 74939 + }, + { + "epoch": 5.54, + "learning_rate": 3.099211791377643e-07, + "loss": 1.0517, + "step": 74940 + }, + { + "epoch": 5.54, + "learning_rate": 3.098226536677496e-07, + "loss": 0.9206, + "step": 74941 + }, + { + "epoch": 5.54, + "learning_rate": 3.0972414361466697e-07, + "loss": 0.9097, + "step": 74942 + }, + { + "epoch": 5.54, + "learning_rate": 3.0962564897867753e-07, + "loss": 0.9381, + "step": 74943 + }, + { + "epoch": 5.54, + "learning_rate": 3.095271697599322e-07, + "loss": 1.0023, + "step": 74944 + }, + { + "epoch": 5.54, + "learning_rate": 3.09428705958591e-07, + "loss": 1.0471, + "step": 74945 + }, + { + "epoch": 5.54, + "learning_rate": 3.093302575748114e-07, + "loss": 1.0678, + "step": 74946 + }, + { + "epoch": 5.54, + "learning_rate": 3.092318246087478e-07, + "loss": 0.9527, + "step": 74947 + }, + { + "epoch": 5.54, + "learning_rate": 3.091334070605578e-07, + "loss": 1.0063, + "step": 74948 + }, + { + "epoch": 5.54, + "learning_rate": 3.09035004930397e-07, + "loss": 0.9114, + "step": 74949 + }, + { + "epoch": 5.54, + "learning_rate": 3.0893661821842394e-07, + "loss": 1.084, + "step": 74950 + }, + { + "epoch": 5.54, + "learning_rate": 3.088382469247941e-07, + "loss": 0.9418, + "step": 74951 + }, + { + "epoch": 5.54, + "learning_rate": 3.0873989104966416e-07, + "loss": 1.0337, + "step": 74952 + }, + { + "epoch": 5.54, + "learning_rate": 3.0864155059318834e-07, + "loss": 0.933, + "step": 74953 + }, + { + "epoch": 5.54, + "learning_rate": 3.0854322555552765e-07, + "loss": 0.8422, + "step": 74954 + }, + { + "epoch": 5.54, + "learning_rate": 3.0844491593683524e-07, + "loss": 0.9547, + "step": 74955 + }, + { + "epoch": 5.54, + "learning_rate": 3.083466217372688e-07, + "loss": 1.0905, + "step": 74956 + }, + { + "epoch": 5.54, + "learning_rate": 3.082483429569838e-07, + "loss": 1.0531, + "step": 74957 + }, + { + "epoch": 5.54, + "learning_rate": 3.0815007959613563e-07, + "loss": 0.9124, + "step": 74958 + }, + { + "epoch": 5.54, + "learning_rate": 3.080518316548831e-07, + "loss": 0.9136, + "step": 74959 + }, + { + "epoch": 5.54, + "learning_rate": 3.079535991333815e-07, + "loss": 1.0168, + "step": 74960 + }, + { + "epoch": 5.54, + "learning_rate": 3.0785538203178754e-07, + "loss": 1.0455, + "step": 74961 + }, + { + "epoch": 5.54, + "learning_rate": 3.0775718035025547e-07, + "loss": 0.9692, + "step": 74962 + }, + { + "epoch": 5.54, + "learning_rate": 3.0765899408894406e-07, + "loss": 0.988, + "step": 74963 + }, + { + "epoch": 5.54, + "learning_rate": 3.0756082324801093e-07, + "loss": 1.0084, + "step": 74964 + }, + { + "epoch": 5.54, + "learning_rate": 3.0746266782760605e-07, + "loss": 0.9337, + "step": 74965 + }, + { + "epoch": 5.54, + "learning_rate": 3.073645278278914e-07, + "loss": 1.0287, + "step": 74966 + }, + { + "epoch": 5.54, + "learning_rate": 3.072664032490202e-07, + "loss": 0.9593, + "step": 74967 + }, + { + "epoch": 5.54, + "learning_rate": 3.0716829409115024e-07, + "loss": 0.9691, + "step": 74968 + }, + { + "epoch": 5.54, + "learning_rate": 3.070702003544357e-07, + "loss": 0.9496, + "step": 74969 + }, + { + "epoch": 5.54, + "learning_rate": 3.0697212203903427e-07, + "loss": 1.0588, + "step": 74970 + }, + { + "epoch": 5.54, + "learning_rate": 3.0687405914510246e-07, + "loss": 1.1334, + "step": 74971 + }, + { + "epoch": 5.54, + "learning_rate": 3.067760116727947e-07, + "loss": 1.1086, + "step": 74972 + }, + { + "epoch": 5.54, + "learning_rate": 3.066779796222674e-07, + "loss": 1.0238, + "step": 74973 + }, + { + "epoch": 5.54, + "learning_rate": 3.065799629936761e-07, + "loss": 0.9846, + "step": 74974 + }, + { + "epoch": 5.54, + "learning_rate": 3.0648196178717837e-07, + "loss": 0.9893, + "step": 74975 + }, + { + "epoch": 5.54, + "learning_rate": 3.0638397600292967e-07, + "loss": 1.0224, + "step": 74976 + }, + { + "epoch": 5.54, + "learning_rate": 3.0628600564108434e-07, + "loss": 1.034, + "step": 74977 + }, + { + "epoch": 5.54, + "learning_rate": 3.061880507017989e-07, + "loss": 0.9348, + "step": 74978 + }, + { + "epoch": 5.54, + "learning_rate": 3.06090111185231e-07, + "loss": 1.1007, + "step": 74979 + }, + { + "epoch": 5.54, + "learning_rate": 3.0599218709153387e-07, + "loss": 0.9402, + "step": 74980 + }, + { + "epoch": 5.54, + "learning_rate": 3.0589427842086516e-07, + "loss": 0.9657, + "step": 74981 + }, + { + "epoch": 5.54, + "learning_rate": 3.057963851733803e-07, + "loss": 1.0113, + "step": 74982 + }, + { + "epoch": 5.54, + "learning_rate": 3.0569850734923357e-07, + "loss": 1.1832, + "step": 74983 + }, + { + "epoch": 5.54, + "learning_rate": 3.0560064494858266e-07, + "loss": 1.0616, + "step": 74984 + }, + { + "epoch": 5.54, + "learning_rate": 3.055027979715819e-07, + "loss": 1.053, + "step": 74985 + }, + { + "epoch": 5.54, + "learning_rate": 3.054049664183878e-07, + "loss": 1.0215, + "step": 74986 + }, + { + "epoch": 5.54, + "learning_rate": 3.053071502891547e-07, + "loss": 0.9511, + "step": 74987 + }, + { + "epoch": 5.54, + "learning_rate": 3.052093495840414e-07, + "loss": 0.9971, + "step": 74988 + }, + { + "epoch": 5.54, + "learning_rate": 3.0511156430319874e-07, + "loss": 1.025, + "step": 74989 + }, + { + "epoch": 5.54, + "learning_rate": 3.050137944467868e-07, + "loss": 0.9944, + "step": 74990 + }, + { + "epoch": 5.54, + "learning_rate": 3.049160400149576e-07, + "loss": 1.0229, + "step": 74991 + }, + { + "epoch": 5.54, + "learning_rate": 3.048183010078698e-07, + "loss": 1.0029, + "step": 74992 + }, + { + "epoch": 5.54, + "learning_rate": 3.0472057742567786e-07, + "loss": 1.0132, + "step": 74993 + }, + { + "epoch": 5.54, + "learning_rate": 3.046228692685349e-07, + "loss": 0.9998, + "step": 74994 + }, + { + "epoch": 5.54, + "learning_rate": 3.045251765365986e-07, + "loss": 0.9628, + "step": 74995 + }, + { + "epoch": 5.54, + "learning_rate": 3.0442749923002446e-07, + "loss": 1.0499, + "step": 74996 + }, + { + "epoch": 5.54, + "learning_rate": 3.043298373489678e-07, + "loss": 0.977, + "step": 74997 + }, + { + "epoch": 5.54, + "learning_rate": 3.0423219089358304e-07, + "loss": 0.9651, + "step": 74998 + }, + { + "epoch": 5.54, + "learning_rate": 3.041345598640255e-07, + "loss": 1.0561, + "step": 74999 + }, + { + "epoch": 5.54, + "learning_rate": 3.0403694426045294e-07, + "loss": 0.9562, + "step": 75000 + }, + { + "epoch": 5.54, + "learning_rate": 3.039393440830185e-07, + "loss": 1.0682, + "step": 75001 + }, + { + "epoch": 5.54, + "learning_rate": 3.0384175933187654e-07, + "loss": 0.9378, + "step": 75002 + }, + { + "epoch": 5.54, + "learning_rate": 3.037441900071836e-07, + "loss": 0.9925, + "step": 75003 + }, + { + "epoch": 5.54, + "learning_rate": 3.0364663610909617e-07, + "loss": 0.9395, + "step": 75004 + }, + { + "epoch": 5.54, + "learning_rate": 3.035490976377675e-07, + "loss": 1.0478, + "step": 75005 + }, + { + "epoch": 5.54, + "learning_rate": 3.034515745933542e-07, + "loss": 0.9028, + "step": 75006 + }, + { + "epoch": 5.54, + "learning_rate": 3.0335406697600934e-07, + "loss": 0.8794, + "step": 75007 + }, + { + "epoch": 5.54, + "learning_rate": 3.0325657478589066e-07, + "loss": 1.0787, + "step": 75008 + }, + { + "epoch": 5.54, + "learning_rate": 3.0315909802315246e-07, + "loss": 0.8647, + "step": 75009 + }, + { + "epoch": 5.54, + "learning_rate": 3.0306163668794795e-07, + "loss": 0.967, + "step": 75010 + }, + { + "epoch": 5.54, + "learning_rate": 3.0296419078043484e-07, + "loss": 1.0094, + "step": 75011 + }, + { + "epoch": 5.54, + "learning_rate": 3.028667603007662e-07, + "loss": 1.1217, + "step": 75012 + }, + { + "epoch": 5.54, + "learning_rate": 3.0276934524909874e-07, + "loss": 0.9738, + "step": 75013 + }, + { + "epoch": 5.54, + "learning_rate": 3.0267194562558445e-07, + "loss": 1.1203, + "step": 75014 + }, + { + "epoch": 5.54, + "learning_rate": 3.025745614303821e-07, + "loss": 0.998, + "step": 75015 + }, + { + "epoch": 5.54, + "learning_rate": 3.0247719266364494e-07, + "loss": 0.9227, + "step": 75016 + }, + { + "epoch": 5.54, + "learning_rate": 3.023798393255273e-07, + "loss": 0.9707, + "step": 75017 + }, + { + "epoch": 5.54, + "learning_rate": 3.0228250141618344e-07, + "loss": 0.9488, + "step": 75018 + }, + { + "epoch": 5.54, + "learning_rate": 3.0218517893577104e-07, + "loss": 0.9185, + "step": 75019 + }, + { + "epoch": 5.54, + "learning_rate": 3.0208787188444336e-07, + "loss": 1.0957, + "step": 75020 + }, + { + "epoch": 5.54, + "learning_rate": 3.019905802623546e-07, + "loss": 1.0585, + "step": 75021 + }, + { + "epoch": 5.54, + "learning_rate": 3.0189330406966034e-07, + "loss": 1.0108, + "step": 75022 + }, + { + "epoch": 5.54, + "learning_rate": 3.017960433065137e-07, + "loss": 0.9718, + "step": 75023 + }, + { + "epoch": 5.54, + "learning_rate": 3.0169879797307234e-07, + "loss": 0.9238, + "step": 75024 + }, + { + "epoch": 5.54, + "learning_rate": 3.016015680694884e-07, + "loss": 0.9899, + "step": 75025 + }, + { + "epoch": 5.54, + "learning_rate": 3.0150435359591834e-07, + "loss": 1.0247, + "step": 75026 + }, + { + "epoch": 5.54, + "learning_rate": 3.0140715455251546e-07, + "loss": 1.041, + "step": 75027 + }, + { + "epoch": 5.54, + "learning_rate": 3.013099709394351e-07, + "loss": 1.0069, + "step": 75028 + }, + { + "epoch": 5.54, + "learning_rate": 3.01212802756834e-07, + "loss": 0.9635, + "step": 75029 + }, + { + "epoch": 5.54, + "learning_rate": 3.0111565000486174e-07, + "loss": 0.9833, + "step": 75030 + }, + { + "epoch": 5.54, + "learning_rate": 3.0101851268367733e-07, + "loss": 1.0596, + "step": 75031 + }, + { + "epoch": 5.54, + "learning_rate": 3.0092139079343163e-07, + "loss": 0.9263, + "step": 75032 + }, + { + "epoch": 5.54, + "learning_rate": 3.0082428433428456e-07, + "loss": 1.0214, + "step": 75033 + }, + { + "epoch": 5.54, + "learning_rate": 3.007271933063838e-07, + "loss": 1.0915, + "step": 75034 + }, + { + "epoch": 5.54, + "learning_rate": 3.0063011770988917e-07, + "loss": 0.9478, + "step": 75035 + }, + { + "epoch": 5.54, + "learning_rate": 3.0053305754495276e-07, + "loss": 1.0259, + "step": 75036 + }, + { + "epoch": 5.54, + "learning_rate": 3.004360128117312e-07, + "loss": 1.1317, + "step": 75037 + }, + { + "epoch": 5.54, + "learning_rate": 3.0033898351037427e-07, + "loss": 1.053, + "step": 75038 + }, + { + "epoch": 5.54, + "learning_rate": 3.002419696410408e-07, + "loss": 1.071, + "step": 75039 + }, + { + "epoch": 5.54, + "learning_rate": 3.0014497120388397e-07, + "loss": 0.9485, + "step": 75040 + }, + { + "epoch": 5.54, + "learning_rate": 3.0004798819905703e-07, + "loss": 1.041, + "step": 75041 + }, + { + "epoch": 5.54, + "learning_rate": 2.999510206267153e-07, + "loss": 1.0128, + "step": 75042 + }, + { + "epoch": 5.54, + "learning_rate": 2.9985406848701103e-07, + "loss": 0.9732, + "step": 75043 + }, + { + "epoch": 5.54, + "learning_rate": 2.997571317801018e-07, + "loss": 1.084, + "step": 75044 + }, + { + "epoch": 5.55, + "learning_rate": 2.996602105061408e-07, + "loss": 0.9431, + "step": 75045 + }, + { + "epoch": 5.55, + "learning_rate": 2.9956330466528126e-07, + "loss": 0.9764, + "step": 75046 + }, + { + "epoch": 5.55, + "learning_rate": 2.994664142576764e-07, + "loss": 1.0382, + "step": 75047 + }, + { + "epoch": 5.55, + "learning_rate": 2.993695392834828e-07, + "loss": 0.9648, + "step": 75048 + }, + { + "epoch": 5.55, + "learning_rate": 2.9927267974285243e-07, + "loss": 1.1489, + "step": 75049 + }, + { + "epoch": 5.55, + "learning_rate": 2.99175835635942e-07, + "loss": 1.1124, + "step": 75050 + }, + { + "epoch": 5.55, + "learning_rate": 2.990790069629035e-07, + "loss": 0.9694, + "step": 75051 + }, + { + "epoch": 5.55, + "learning_rate": 2.9898219372389014e-07, + "loss": 1.0921, + "step": 75052 + }, + { + "epoch": 5.55, + "learning_rate": 2.988853959190585e-07, + "loss": 1.0921, + "step": 75053 + }, + { + "epoch": 5.55, + "learning_rate": 2.987886135485618e-07, + "loss": 1.0313, + "step": 75054 + }, + { + "epoch": 5.55, + "learning_rate": 2.986918466125532e-07, + "loss": 0.9878, + "step": 75055 + }, + { + "epoch": 5.55, + "learning_rate": 2.9859509511118603e-07, + "loss": 1.118, + "step": 75056 + }, + { + "epoch": 5.55, + "learning_rate": 2.984983590446178e-07, + "loss": 0.9885, + "step": 75057 + }, + { + "epoch": 5.55, + "learning_rate": 2.9840163841299843e-07, + "loss": 1.0447, + "step": 75058 + }, + { + "epoch": 5.55, + "learning_rate": 2.9830493321648226e-07, + "loss": 1.0851, + "step": 75059 + }, + { + "epoch": 5.55, + "learning_rate": 2.982082434552247e-07, + "loss": 0.9912, + "step": 75060 + }, + { + "epoch": 5.55, + "learning_rate": 2.981115691293801e-07, + "loss": 0.9932, + "step": 75061 + }, + { + "epoch": 5.55, + "learning_rate": 2.9801491023910054e-07, + "loss": 1.0231, + "step": 75062 + }, + { + "epoch": 5.55, + "learning_rate": 2.979182667845393e-07, + "loss": 1.0211, + "step": 75063 + }, + { + "epoch": 5.55, + "learning_rate": 2.978216387658528e-07, + "loss": 1.0179, + "step": 75064 + }, + { + "epoch": 5.55, + "learning_rate": 2.977250261831921e-07, + "loss": 0.9379, + "step": 75065 + }, + { + "epoch": 5.55, + "learning_rate": 2.9762842903671376e-07, + "loss": 0.8829, + "step": 75066 + }, + { + "epoch": 5.55, + "learning_rate": 2.975318473265687e-07, + "loss": 1.0151, + "step": 75067 + }, + { + "epoch": 5.55, + "learning_rate": 2.974352810529102e-07, + "loss": 1.0014, + "step": 75068 + }, + { + "epoch": 5.55, + "learning_rate": 2.9733873021589476e-07, + "loss": 0.9668, + "step": 75069 + }, + { + "epoch": 5.55, + "learning_rate": 2.9724219481567563e-07, + "loss": 0.9171, + "step": 75070 + }, + { + "epoch": 5.55, + "learning_rate": 2.9714567485240373e-07, + "loss": 1.0391, + "step": 75071 + }, + { + "epoch": 5.55, + "learning_rate": 2.970491703262346e-07, + "loss": 0.9899, + "step": 75072 + }, + { + "epoch": 5.55, + "learning_rate": 2.9695268123732133e-07, + "loss": 1.0643, + "step": 75073 + }, + { + "epoch": 5.55, + "learning_rate": 2.9685620758581834e-07, + "loss": 0.9017, + "step": 75074 + }, + { + "epoch": 5.55, + "learning_rate": 2.967597493718777e-07, + "loss": 0.9143, + "step": 75075 + }, + { + "epoch": 5.55, + "learning_rate": 2.9666330659565367e-07, + "loss": 0.9698, + "step": 75076 + }, + { + "epoch": 5.55, + "learning_rate": 2.9656687925729846e-07, + "loss": 1.095, + "step": 75077 + }, + { + "epoch": 5.55, + "learning_rate": 2.964704673569696e-07, + "loss": 1.0818, + "step": 75078 + }, + { + "epoch": 5.55, + "learning_rate": 2.9637407089481375e-07, + "loss": 0.917, + "step": 75079 + }, + { + "epoch": 5.55, + "learning_rate": 2.962776898709896e-07, + "loss": 1.0311, + "step": 75080 + }, + { + "epoch": 5.55, + "learning_rate": 2.961813242856482e-07, + "loss": 0.9771, + "step": 75081 + }, + { + "epoch": 5.55, + "learning_rate": 2.960849741389449e-07, + "loss": 1.023, + "step": 75082 + }, + { + "epoch": 5.55, + "learning_rate": 2.9598863943102963e-07, + "loss": 1.0472, + "step": 75083 + }, + { + "epoch": 5.55, + "learning_rate": 2.9589232016205893e-07, + "loss": 1.0502, + "step": 75084 + }, + { + "epoch": 5.55, + "learning_rate": 2.957960163321838e-07, + "loss": 0.9754, + "step": 75085 + }, + { + "epoch": 5.55, + "learning_rate": 2.9569972794155965e-07, + "loss": 1.0912, + "step": 75086 + }, + { + "epoch": 5.55, + "learning_rate": 2.956034549903375e-07, + "loss": 1.1375, + "step": 75087 + }, + { + "epoch": 5.55, + "learning_rate": 2.955071974786705e-07, + "loss": 0.9577, + "step": 75088 + }, + { + "epoch": 5.55, + "learning_rate": 2.9541095540671414e-07, + "loss": 1.0776, + "step": 75089 + }, + { + "epoch": 5.55, + "learning_rate": 2.953147287746194e-07, + "loss": 1.0576, + "step": 75090 + }, + { + "epoch": 5.55, + "learning_rate": 2.952185175825406e-07, + "loss": 1.0041, + "step": 75091 + }, + { + "epoch": 5.55, + "learning_rate": 2.9512232183062985e-07, + "loss": 1.0937, + "step": 75092 + }, + { + "epoch": 5.55, + "learning_rate": 2.950261415190403e-07, + "loss": 0.9268, + "step": 75093 + }, + { + "epoch": 5.55, + "learning_rate": 2.9492997664792634e-07, + "loss": 1.0067, + "step": 75094 + }, + { + "epoch": 5.55, + "learning_rate": 2.9483382721743893e-07, + "loss": 0.9499, + "step": 75095 + }, + { + "epoch": 5.55, + "learning_rate": 2.947376932277335e-07, + "loss": 0.9486, + "step": 75096 + }, + { + "epoch": 5.55, + "learning_rate": 2.9464157467895995e-07, + "loss": 0.9764, + "step": 75097 + }, + { + "epoch": 5.55, + "learning_rate": 2.9454547157127257e-07, + "loss": 0.9779, + "step": 75098 + }, + { + "epoch": 5.55, + "learning_rate": 2.9444938390482567e-07, + "loss": 0.9737, + "step": 75099 + }, + { + "epoch": 5.55, + "learning_rate": 2.9435331167977145e-07, + "loss": 0.9788, + "step": 75100 + }, + { + "epoch": 5.55, + "learning_rate": 2.942572548962597e-07, + "loss": 1.109, + "step": 75101 + }, + { + "epoch": 5.55, + "learning_rate": 2.9416121355444917e-07, + "loss": 0.9516, + "step": 75102 + }, + { + "epoch": 5.55, + "learning_rate": 2.940651876544864e-07, + "loss": 1.0864, + "step": 75103 + }, + { + "epoch": 5.55, + "learning_rate": 2.9396917719652807e-07, + "loss": 0.9164, + "step": 75104 + }, + { + "epoch": 5.55, + "learning_rate": 2.9387318218072726e-07, + "loss": 0.9, + "step": 75105 + }, + { + "epoch": 5.55, + "learning_rate": 2.9377720260723383e-07, + "loss": 1.0274, + "step": 75106 + }, + { + "epoch": 5.55, + "learning_rate": 2.936812384762022e-07, + "loss": 0.9952, + "step": 75107 + }, + { + "epoch": 5.55, + "learning_rate": 2.9358528978778446e-07, + "loss": 0.9646, + "step": 75108 + }, + { + "epoch": 5.55, + "learning_rate": 2.934893565421337e-07, + "loss": 1.0572, + "step": 75109 + }, + { + "epoch": 5.55, + "learning_rate": 2.933934387394033e-07, + "loss": 0.9318, + "step": 75110 + }, + { + "epoch": 5.55, + "learning_rate": 2.932975363797452e-07, + "loss": 1.0058, + "step": 75111 + }, + { + "epoch": 5.55, + "learning_rate": 2.932016494633105e-07, + "loss": 1.0054, + "step": 75112 + }, + { + "epoch": 5.55, + "learning_rate": 2.931057779902535e-07, + "loss": 0.9721, + "step": 75113 + }, + { + "epoch": 5.55, + "learning_rate": 2.930099219607274e-07, + "loss": 0.9284, + "step": 75114 + }, + { + "epoch": 5.55, + "learning_rate": 2.929140813748832e-07, + "loss": 0.9864, + "step": 75115 + }, + { + "epoch": 5.55, + "learning_rate": 2.9281825623287405e-07, + "loss": 1.007, + "step": 75116 + }, + { + "epoch": 5.55, + "learning_rate": 2.9272244653485104e-07, + "loss": 0.932, + "step": 75117 + }, + { + "epoch": 5.55, + "learning_rate": 2.926266522809684e-07, + "loss": 0.9025, + "step": 75118 + }, + { + "epoch": 5.55, + "learning_rate": 2.925308734713783e-07, + "loss": 0.9903, + "step": 75119 + }, + { + "epoch": 5.55, + "learning_rate": 2.9243511010623174e-07, + "loss": 0.9216, + "step": 75120 + }, + { + "epoch": 5.55, + "learning_rate": 2.923393621856818e-07, + "loss": 0.9378, + "step": 75121 + }, + { + "epoch": 5.55, + "learning_rate": 2.9224362970988186e-07, + "loss": 1.029, + "step": 75122 + }, + { + "epoch": 5.55, + "learning_rate": 2.9214791267898503e-07, + "loss": 1.0446, + "step": 75123 + }, + { + "epoch": 5.55, + "learning_rate": 2.9205221109313895e-07, + "loss": 0.9878, + "step": 75124 + }, + { + "epoch": 5.55, + "learning_rate": 2.9195652495250027e-07, + "loss": 0.9862, + "step": 75125 + }, + { + "epoch": 5.55, + "learning_rate": 2.9186085425721875e-07, + "loss": 1.0571, + "step": 75126 + }, + { + "epoch": 5.55, + "learning_rate": 2.9176519900745106e-07, + "loss": 1.0093, + "step": 75127 + }, + { + "epoch": 5.55, + "learning_rate": 2.916695592033425e-07, + "loss": 0.8915, + "step": 75128 + }, + { + "epoch": 5.55, + "learning_rate": 2.915739348450497e-07, + "loss": 0.9369, + "step": 75129 + }, + { + "epoch": 5.55, + "learning_rate": 2.9147832593272364e-07, + "loss": 1.0063, + "step": 75130 + }, + { + "epoch": 5.55, + "learning_rate": 2.913827324665186e-07, + "loss": 1.1002, + "step": 75131 + }, + { + "epoch": 5.55, + "learning_rate": 2.912871544465834e-07, + "loss": 1.0628, + "step": 75132 + }, + { + "epoch": 5.55, + "learning_rate": 2.911915918730701e-07, + "loss": 1.0429, + "step": 75133 + }, + { + "epoch": 5.55, + "learning_rate": 2.9109604474613417e-07, + "loss": 0.9814, + "step": 75134 + }, + { + "epoch": 5.55, + "learning_rate": 2.9100051306592437e-07, + "loss": 1.0278, + "step": 75135 + }, + { + "epoch": 5.55, + "learning_rate": 2.9090499683259496e-07, + "loss": 0.9487, + "step": 75136 + }, + { + "epoch": 5.55, + "learning_rate": 2.908094960462948e-07, + "loss": 1.0102, + "step": 75137 + }, + { + "epoch": 5.55, + "learning_rate": 2.907140107071793e-07, + "loss": 1.0022, + "step": 75138 + }, + { + "epoch": 5.55, + "learning_rate": 2.906185408153983e-07, + "loss": 0.9248, + "step": 75139 + }, + { + "epoch": 5.55, + "learning_rate": 2.9052308637110505e-07, + "loss": 1.0087, + "step": 75140 + }, + { + "epoch": 5.55, + "learning_rate": 2.9042764737445053e-07, + "loss": 1.0919, + "step": 75141 + }, + { + "epoch": 5.55, + "learning_rate": 2.903322238255857e-07, + "loss": 0.9483, + "step": 75142 + }, + { + "epoch": 5.55, + "learning_rate": 2.9023681572466376e-07, + "loss": 0.9841, + "step": 75143 + }, + { + "epoch": 5.55, + "learning_rate": 2.901414230718369e-07, + "loss": 1.2135, + "step": 75144 + }, + { + "epoch": 5.55, + "learning_rate": 2.9004604586725494e-07, + "loss": 1.0479, + "step": 75145 + }, + { + "epoch": 5.55, + "learning_rate": 2.899506841110711e-07, + "loss": 1.0847, + "step": 75146 + }, + { + "epoch": 5.55, + "learning_rate": 2.8985533780343855e-07, + "loss": 0.9671, + "step": 75147 + }, + { + "epoch": 5.55, + "learning_rate": 2.89760006944505e-07, + "loss": 0.9783, + "step": 75148 + }, + { + "epoch": 5.55, + "learning_rate": 2.896646915344248e-07, + "loss": 0.9546, + "step": 75149 + }, + { + "epoch": 5.55, + "learning_rate": 2.8956939157334885e-07, + "loss": 1.1041, + "step": 75150 + }, + { + "epoch": 5.55, + "learning_rate": 2.894741070614315e-07, + "loss": 1.0332, + "step": 75151 + }, + { + "epoch": 5.55, + "learning_rate": 2.893788379988194e-07, + "loss": 1.0156, + "step": 75152 + }, + { + "epoch": 5.55, + "learning_rate": 2.8928358438566674e-07, + "loss": 1.0138, + "step": 75153 + }, + { + "epoch": 5.55, + "learning_rate": 2.8918834622212565e-07, + "loss": 1.0173, + "step": 75154 + }, + { + "epoch": 5.55, + "learning_rate": 2.890931235083472e-07, + "loss": 0.9123, + "step": 75155 + }, + { + "epoch": 5.55, + "learning_rate": 2.8899791624448226e-07, + "loss": 1.0654, + "step": 75156 + }, + { + "epoch": 5.55, + "learning_rate": 2.889027244306819e-07, + "loss": 0.9994, + "step": 75157 + }, + { + "epoch": 5.55, + "learning_rate": 2.888075480671004e-07, + "loss": 0.8778, + "step": 75158 + }, + { + "epoch": 5.55, + "learning_rate": 2.887123871538855e-07, + "loss": 1.1041, + "step": 75159 + }, + { + "epoch": 5.55, + "learning_rate": 2.8861724169119143e-07, + "loss": 1.0318, + "step": 75160 + }, + { + "epoch": 5.55, + "learning_rate": 2.885221116791681e-07, + "loss": 0.9188, + "step": 75161 + }, + { + "epoch": 5.55, + "learning_rate": 2.884269971179654e-07, + "loss": 1.0154, + "step": 75162 + }, + { + "epoch": 5.55, + "learning_rate": 2.8833189800773764e-07, + "loss": 1.1622, + "step": 75163 + }, + { + "epoch": 5.55, + "learning_rate": 2.882368143486358e-07, + "loss": 1.0707, + "step": 75164 + }, + { + "epoch": 5.55, + "learning_rate": 2.881417461408087e-07, + "loss": 0.9686, + "step": 75165 + }, + { + "epoch": 5.55, + "learning_rate": 2.880466933844095e-07, + "loss": 0.9707, + "step": 75166 + }, + { + "epoch": 5.55, + "learning_rate": 2.879516560795892e-07, + "loss": 1.0021, + "step": 75167 + }, + { + "epoch": 5.55, + "learning_rate": 2.878566342264999e-07, + "loss": 0.8899, + "step": 75168 + }, + { + "epoch": 5.55, + "learning_rate": 2.8776162782529036e-07, + "loss": 1.0652, + "step": 75169 + }, + { + "epoch": 5.55, + "learning_rate": 2.8766663687611387e-07, + "loss": 0.9902, + "step": 75170 + }, + { + "epoch": 5.55, + "learning_rate": 2.8757166137912016e-07, + "loss": 0.9808, + "step": 75171 + }, + { + "epoch": 5.55, + "learning_rate": 2.874767013344626e-07, + "loss": 0.9973, + "step": 75172 + }, + { + "epoch": 5.55, + "learning_rate": 2.8738175674228873e-07, + "loss": 0.9567, + "step": 75173 + }, + { + "epoch": 5.55, + "learning_rate": 2.872868276027518e-07, + "loss": 0.9572, + "step": 75174 + }, + { + "epoch": 5.55, + "learning_rate": 2.871919139160029e-07, + "loss": 1.1448, + "step": 75175 + }, + { + "epoch": 5.55, + "learning_rate": 2.870970156821939e-07, + "loss": 0.9546, + "step": 75176 + }, + { + "epoch": 5.55, + "learning_rate": 2.8700213290147273e-07, + "loss": 0.9846, + "step": 75177 + }, + { + "epoch": 5.55, + "learning_rate": 2.869072655739924e-07, + "loss": 1.0213, + "step": 75178 + }, + { + "epoch": 5.55, + "learning_rate": 2.86812413699904e-07, + "loss": 0.9357, + "step": 75179 + }, + { + "epoch": 5.56, + "learning_rate": 2.8671757727935847e-07, + "loss": 1.023, + "step": 75180 + }, + { + "epoch": 5.56, + "learning_rate": 2.866227563125068e-07, + "loss": 1.0572, + "step": 75181 + }, + { + "epoch": 5.56, + "learning_rate": 2.865279507994978e-07, + "loss": 1.0691, + "step": 75182 + }, + { + "epoch": 5.56, + "learning_rate": 2.8643316074048465e-07, + "loss": 1.1268, + "step": 75183 + }, + { + "epoch": 5.56, + "learning_rate": 2.8633838613561726e-07, + "loss": 0.9386, + "step": 75184 + }, + { + "epoch": 5.56, + "learning_rate": 2.8624362698504773e-07, + "loss": 1.0242, + "step": 75185 + }, + { + "epoch": 5.56, + "learning_rate": 2.8614888328892254e-07, + "loss": 1.0001, + "step": 75186 + }, + { + "epoch": 5.56, + "learning_rate": 2.860541550473983e-07, + "loss": 1.0064, + "step": 75187 + }, + { + "epoch": 5.56, + "learning_rate": 2.8595944226062377e-07, + "loss": 0.881, + "step": 75188 + }, + { + "epoch": 5.56, + "learning_rate": 2.8586474492874547e-07, + "loss": 0.9862, + "step": 75189 + }, + { + "epoch": 5.56, + "learning_rate": 2.857700630519189e-07, + "loss": 1.0749, + "step": 75190 + }, + { + "epoch": 5.56, + "learning_rate": 2.8567539663029276e-07, + "loss": 0.9859, + "step": 75191 + }, + { + "epoch": 5.56, + "learning_rate": 2.8558074566402025e-07, + "loss": 0.9777, + "step": 75192 + }, + { + "epoch": 5.56, + "learning_rate": 2.854861101532469e-07, + "loss": 0.9247, + "step": 75193 + }, + { + "epoch": 5.56, + "learning_rate": 2.853914900981281e-07, + "loss": 0.9475, + "step": 75194 + }, + { + "epoch": 5.56, + "learning_rate": 2.852968854988103e-07, + "loss": 0.9808, + "step": 75195 + }, + { + "epoch": 5.56, + "learning_rate": 2.8520229635544907e-07, + "loss": 1.0352, + "step": 75196 + }, + { + "epoch": 5.56, + "learning_rate": 2.8510772266819087e-07, + "loss": 1.0223, + "step": 75197 + }, + { + "epoch": 5.56, + "learning_rate": 2.8501316443718565e-07, + "loss": 1.0712, + "step": 75198 + }, + { + "epoch": 5.56, + "learning_rate": 2.8491862166258656e-07, + "loss": 0.9413, + "step": 75199 + }, + { + "epoch": 5.56, + "learning_rate": 2.848240943445435e-07, + "loss": 1.0448, + "step": 75200 + }, + { + "epoch": 5.56, + "learning_rate": 2.8472958248320636e-07, + "loss": 0.9141, + "step": 75201 + }, + { + "epoch": 5.56, + "learning_rate": 2.8463508607872393e-07, + "loss": 0.9757, + "step": 75202 + }, + { + "epoch": 5.56, + "learning_rate": 2.8454060513124935e-07, + "loss": 1.078, + "step": 75203 + }, + { + "epoch": 5.56, + "learning_rate": 2.8444613964093147e-07, + "loss": 0.9378, + "step": 75204 + }, + { + "epoch": 5.56, + "learning_rate": 2.843516896079213e-07, + "loss": 0.9721, + "step": 75205 + }, + { + "epoch": 5.56, + "learning_rate": 2.8425725503236743e-07, + "loss": 0.9616, + "step": 75206 + }, + { + "epoch": 5.56, + "learning_rate": 2.841628359144222e-07, + "loss": 1.0318, + "step": 75207 + }, + { + "epoch": 5.56, + "learning_rate": 2.840684322542342e-07, + "loss": 1.0523, + "step": 75208 + }, + { + "epoch": 5.56, + "learning_rate": 2.839740440519556e-07, + "loss": 0.9954, + "step": 75209 + }, + { + "epoch": 5.56, + "learning_rate": 2.838796713077341e-07, + "loss": 1.0057, + "step": 75210 + }, + { + "epoch": 5.56, + "learning_rate": 2.837853140217206e-07, + "loss": 0.9359, + "step": 75211 + }, + { + "epoch": 5.56, + "learning_rate": 2.8369097219406615e-07, + "loss": 0.9876, + "step": 75212 + }, + { + "epoch": 5.56, + "learning_rate": 2.8359664582492177e-07, + "loss": 0.9395, + "step": 75213 + }, + { + "epoch": 5.56, + "learning_rate": 2.83502334914435e-07, + "loss": 1.0266, + "step": 75214 + }, + { + "epoch": 5.56, + "learning_rate": 2.8340803946275586e-07, + "loss": 1.0455, + "step": 75215 + }, + { + "epoch": 5.56, + "learning_rate": 2.8331375947003636e-07, + "loss": 1.0393, + "step": 75216 + }, + { + "epoch": 5.56, + "learning_rate": 2.832194949364275e-07, + "loss": 1.0659, + "step": 75217 + }, + { + "epoch": 5.56, + "learning_rate": 2.831252458620748e-07, + "loss": 1.0187, + "step": 75218 + }, + { + "epoch": 5.56, + "learning_rate": 2.8303101224713137e-07, + "loss": 0.9355, + "step": 75219 + }, + { + "epoch": 5.56, + "learning_rate": 2.8293679409174604e-07, + "loss": 1.0843, + "step": 75220 + }, + { + "epoch": 5.56, + "learning_rate": 2.8284259139607196e-07, + "loss": 1.0225, + "step": 75221 + }, + { + "epoch": 5.56, + "learning_rate": 2.827484041602535e-07, + "loss": 0.9316, + "step": 75222 + }, + { + "epoch": 5.56, + "learning_rate": 2.826542323844439e-07, + "loss": 1.0795, + "step": 75223 + }, + { + "epoch": 5.56, + "learning_rate": 2.8256007606879297e-07, + "loss": 1.0403, + "step": 75224 + }, + { + "epoch": 5.56, + "learning_rate": 2.8246593521345065e-07, + "loss": 0.966, + "step": 75225 + }, + { + "epoch": 5.56, + "learning_rate": 2.823718098185646e-07, + "loss": 0.9054, + "step": 75226 + }, + { + "epoch": 5.56, + "learning_rate": 2.8227769988428575e-07, + "loss": 0.9956, + "step": 75227 + }, + { + "epoch": 5.56, + "learning_rate": 2.821836054107652e-07, + "loss": 1.0195, + "step": 75228 + }, + { + "epoch": 5.56, + "learning_rate": 2.8208952639815047e-07, + "loss": 0.9591, + "step": 75229 + }, + { + "epoch": 5.56, + "learning_rate": 2.819954628465937e-07, + "loss": 1.0935, + "step": 75230 + }, + { + "epoch": 5.56, + "learning_rate": 2.8190141475624155e-07, + "loss": 0.9714, + "step": 75231 + }, + { + "epoch": 5.56, + "learning_rate": 2.81807382127246e-07, + "loss": 1.0171, + "step": 75232 + }, + { + "epoch": 5.56, + "learning_rate": 2.8171336495975585e-07, + "loss": 0.9335, + "step": 75233 + }, + { + "epoch": 5.56, + "learning_rate": 2.816193632539199e-07, + "loss": 0.9081, + "step": 75234 + }, + { + "epoch": 5.56, + "learning_rate": 2.8152537700989024e-07, + "loss": 1.0657, + "step": 75235 + }, + { + "epoch": 5.56, + "learning_rate": 2.814314062278123e-07, + "loss": 0.933, + "step": 75236 + }, + { + "epoch": 5.56, + "learning_rate": 2.8133745090784035e-07, + "loss": 0.9942, + "step": 75237 + }, + { + "epoch": 5.56, + "learning_rate": 2.812435110501199e-07, + "loss": 1.0113, + "step": 75238 + }, + { + "epoch": 5.56, + "learning_rate": 2.81149586654802e-07, + "loss": 0.9607, + "step": 75239 + }, + { + "epoch": 5.56, + "learning_rate": 2.8105567772203634e-07, + "loss": 1.0744, + "step": 75240 + }, + { + "epoch": 5.56, + "learning_rate": 2.8096178425197293e-07, + "loss": 1.0178, + "step": 75241 + }, + { + "epoch": 5.56, + "learning_rate": 2.8086790624475835e-07, + "loss": 1.0372, + "step": 75242 + }, + { + "epoch": 5.56, + "learning_rate": 2.8077404370054574e-07, + "loss": 1.0949, + "step": 75243 + }, + { + "epoch": 5.56, + "learning_rate": 2.8068019661948164e-07, + "loss": 0.8506, + "step": 75244 + }, + { + "epoch": 5.56, + "learning_rate": 2.805863650017171e-07, + "loss": 0.9508, + "step": 75245 + }, + { + "epoch": 5.56, + "learning_rate": 2.8049254884739974e-07, + "loss": 1.0601, + "step": 75246 + }, + { + "epoch": 5.56, + "learning_rate": 2.8039874815667945e-07, + "loss": 1.0273, + "step": 75247 + }, + { + "epoch": 5.56, + "learning_rate": 2.803049629297061e-07, + "loss": 0.9499, + "step": 75248 + }, + { + "epoch": 5.56, + "learning_rate": 2.8021119316662846e-07, + "loss": 0.9952, + "step": 75249 + }, + { + "epoch": 5.56, + "learning_rate": 2.8011743886759644e-07, + "loss": 1.0494, + "step": 75250 + }, + { + "epoch": 5.56, + "learning_rate": 2.800237000327566e-07, + "loss": 1.0746, + "step": 75251 + }, + { + "epoch": 5.56, + "learning_rate": 2.799299766622621e-07, + "loss": 1.0668, + "step": 75252 + }, + { + "epoch": 5.56, + "learning_rate": 2.7983626875625834e-07, + "loss": 1.0989, + "step": 75253 + }, + { + "epoch": 5.56, + "learning_rate": 2.7974257631489754e-07, + "loss": 1.1087, + "step": 75254 + }, + { + "epoch": 5.56, + "learning_rate": 2.7964889933832616e-07, + "loss": 1.0293, + "step": 75255 + }, + { + "epoch": 5.56, + "learning_rate": 2.7955523782669413e-07, + "loss": 0.9365, + "step": 75256 + }, + { + "epoch": 5.56, + "learning_rate": 2.794615917801513e-07, + "loss": 0.9617, + "step": 75257 + }, + { + "epoch": 5.56, + "learning_rate": 2.7936796119884533e-07, + "loss": 0.92, + "step": 75258 + }, + { + "epoch": 5.56, + "learning_rate": 2.7927434608292725e-07, + "loss": 0.9934, + "step": 75259 + }, + { + "epoch": 5.56, + "learning_rate": 2.7918074643254245e-07, + "loss": 0.9441, + "step": 75260 + }, + { + "epoch": 5.56, + "learning_rate": 2.7908716224784414e-07, + "loss": 1.1237, + "step": 75261 + }, + { + "epoch": 5.56, + "learning_rate": 2.789935935289778e-07, + "loss": 1.071, + "step": 75262 + }, + { + "epoch": 5.56, + "learning_rate": 2.789000402760944e-07, + "loss": 1.0798, + "step": 75263 + }, + { + "epoch": 5.56, + "learning_rate": 2.788065024893416e-07, + "loss": 1.1305, + "step": 75264 + }, + { + "epoch": 5.56, + "learning_rate": 2.787129801688671e-07, + "loss": 0.9093, + "step": 75265 + }, + { + "epoch": 5.56, + "learning_rate": 2.7861947331482397e-07, + "loss": 1.0159, + "step": 75266 + }, + { + "epoch": 5.56, + "learning_rate": 2.7852598192735557e-07, + "loss": 1.0504, + "step": 75267 + }, + { + "epoch": 5.56, + "learning_rate": 2.784325060066151e-07, + "loss": 1.0848, + "step": 75268 + }, + { + "epoch": 5.56, + "learning_rate": 2.7833904555274905e-07, + "loss": 0.9252, + "step": 75269 + }, + { + "epoch": 5.56, + "learning_rate": 2.782456005659062e-07, + "loss": 0.9914, + "step": 75270 + }, + { + "epoch": 5.56, + "learning_rate": 2.7815217104623426e-07, + "loss": 1.0299, + "step": 75271 + }, + { + "epoch": 5.56, + "learning_rate": 2.780587569938842e-07, + "loss": 1.0695, + "step": 75272 + }, + { + "epoch": 5.56, + "learning_rate": 2.779653584090047e-07, + "loss": 0.8836, + "step": 75273 + }, + { + "epoch": 5.56, + "learning_rate": 2.778719752917414e-07, + "loss": 1.0526, + "step": 75274 + }, + { + "epoch": 5.56, + "learning_rate": 2.777786076422462e-07, + "loss": 0.8819, + "step": 75275 + }, + { + "epoch": 5.56, + "learning_rate": 2.776852554606646e-07, + "loss": 0.9345, + "step": 75276 + }, + { + "epoch": 5.56, + "learning_rate": 2.7759191874714765e-07, + "loss": 1.0457, + "step": 75277 + }, + { + "epoch": 5.56, + "learning_rate": 2.774985975018429e-07, + "loss": 1.0023, + "step": 75278 + }, + { + "epoch": 5.56, + "learning_rate": 2.7740529172489817e-07, + "loss": 0.8533, + "step": 75279 + }, + { + "epoch": 5.56, + "learning_rate": 2.773120014164621e-07, + "loss": 0.9503, + "step": 75280 + }, + { + "epoch": 5.56, + "learning_rate": 2.7721872657668457e-07, + "loss": 0.9762, + "step": 75281 + }, + { + "epoch": 5.56, + "learning_rate": 2.771254672057144e-07, + "loss": 1.1448, + "step": 75282 + }, + { + "epoch": 5.56, + "learning_rate": 2.7703222330369594e-07, + "loss": 0.9834, + "step": 75283 + }, + { + "epoch": 5.56, + "learning_rate": 2.7693899487078126e-07, + "loss": 1.0533, + "step": 75284 + }, + { + "epoch": 5.56, + "learning_rate": 2.7684578190711574e-07, + "loss": 1.1085, + "step": 75285 + }, + { + "epoch": 5.56, + "learning_rate": 2.767525844128527e-07, + "loss": 1.0602, + "step": 75286 + }, + { + "epoch": 5.56, + "learning_rate": 2.7665940238813415e-07, + "loss": 1.0402, + "step": 75287 + }, + { + "epoch": 5.56, + "learning_rate": 2.765662358331123e-07, + "loss": 1.0057, + "step": 75288 + }, + { + "epoch": 5.56, + "learning_rate": 2.7647308474793465e-07, + "loss": 0.9242, + "step": 75289 + }, + { + "epoch": 5.56, + "learning_rate": 2.7637994913275014e-07, + "loss": 0.932, + "step": 75290 + }, + { + "epoch": 5.56, + "learning_rate": 2.762868289877052e-07, + "loss": 1.0715, + "step": 75291 + }, + { + "epoch": 5.56, + "learning_rate": 2.7619372431294865e-07, + "loss": 0.9389, + "step": 75292 + }, + { + "epoch": 5.56, + "learning_rate": 2.761006351086282e-07, + "loss": 0.9407, + "step": 75293 + }, + { + "epoch": 5.56, + "learning_rate": 2.760075613748936e-07, + "loss": 1.0449, + "step": 75294 + }, + { + "epoch": 5.56, + "learning_rate": 2.759145031118915e-07, + "loss": 0.9778, + "step": 75295 + }, + { + "epoch": 5.56, + "learning_rate": 2.7582146031976954e-07, + "loss": 0.9135, + "step": 75296 + }, + { + "epoch": 5.56, + "learning_rate": 2.757284329986776e-07, + "loss": 1.0718, + "step": 75297 + }, + { + "epoch": 5.56, + "learning_rate": 2.756354211487622e-07, + "loss": 0.9482, + "step": 75298 + }, + { + "epoch": 5.56, + "learning_rate": 2.755424247701721e-07, + "loss": 0.9591, + "step": 75299 + }, + { + "epoch": 5.56, + "learning_rate": 2.7544944386305504e-07, + "loss": 0.9013, + "step": 75300 + }, + { + "epoch": 5.56, + "learning_rate": 2.7535647842755643e-07, + "loss": 1.0386, + "step": 75301 + }, + { + "epoch": 5.56, + "learning_rate": 2.752635284638294e-07, + "loss": 0.9764, + "step": 75302 + }, + { + "epoch": 5.56, + "learning_rate": 2.751705939720173e-07, + "loss": 0.9703, + "step": 75303 + }, + { + "epoch": 5.56, + "learning_rate": 2.75077674952271e-07, + "loss": 1.0301, + "step": 75304 + }, + { + "epoch": 5.56, + "learning_rate": 2.749847714047349e-07, + "loss": 1.0729, + "step": 75305 + }, + { + "epoch": 5.56, + "learning_rate": 2.748918833295611e-07, + "loss": 0.9909, + "step": 75306 + }, + { + "epoch": 5.56, + "learning_rate": 2.747990107268939e-07, + "loss": 1.0384, + "step": 75307 + }, + { + "epoch": 5.56, + "learning_rate": 2.747061535968831e-07, + "loss": 0.9418, + "step": 75308 + }, + { + "epoch": 5.56, + "learning_rate": 2.746133119396743e-07, + "loss": 0.8166, + "step": 75309 + }, + { + "epoch": 5.56, + "learning_rate": 2.7452048575541846e-07, + "loss": 1.0121, + "step": 75310 + }, + { + "epoch": 5.56, + "learning_rate": 2.7442767504426095e-07, + "loss": 1.012, + "step": 75311 + }, + { + "epoch": 5.56, + "learning_rate": 2.743348798063483e-07, + "loss": 0.9919, + "step": 75312 + }, + { + "epoch": 5.56, + "learning_rate": 2.742421000418305e-07, + "loss": 1.0285, + "step": 75313 + }, + { + "epoch": 5.56, + "learning_rate": 2.7414933575085513e-07, + "loss": 1.0707, + "step": 75314 + }, + { + "epoch": 5.56, + "learning_rate": 2.7405658693356876e-07, + "loss": 0.983, + "step": 75315 + }, + { + "epoch": 5.57, + "learning_rate": 2.739638535901179e-07, + "loss": 1.0706, + "step": 75316 + }, + { + "epoch": 5.57, + "learning_rate": 2.738711357206525e-07, + "loss": 1.0043, + "step": 75317 + }, + { + "epoch": 5.57, + "learning_rate": 2.737784333253191e-07, + "loss": 1.0669, + "step": 75318 + }, + { + "epoch": 5.57, + "learning_rate": 2.7368574640426525e-07, + "loss": 0.9739, + "step": 75319 + }, + { + "epoch": 5.57, + "learning_rate": 2.7359307495763765e-07, + "loss": 0.9865, + "step": 75320 + }, + { + "epoch": 5.57, + "learning_rate": 2.7350041898558386e-07, + "loss": 1.0862, + "step": 75321 + }, + { + "epoch": 5.57, + "learning_rate": 2.734077784882527e-07, + "loss": 1.0732, + "step": 75322 + }, + { + "epoch": 5.57, + "learning_rate": 2.7331515346579073e-07, + "loss": 0.9536, + "step": 75323 + }, + { + "epoch": 5.57, + "learning_rate": 2.7322254391834446e-07, + "loss": 1.1108, + "step": 75324 + }, + { + "epoch": 5.57, + "learning_rate": 2.7312994984606154e-07, + "loss": 1.1278, + "step": 75325 + }, + { + "epoch": 5.57, + "learning_rate": 2.730373712490908e-07, + "loss": 0.9674, + "step": 75326 + }, + { + "epoch": 5.57, + "learning_rate": 2.7294480812757765e-07, + "loss": 1.0288, + "step": 75327 + }, + { + "epoch": 5.57, + "learning_rate": 2.728522604816708e-07, + "loss": 1.0, + "step": 75328 + }, + { + "epoch": 5.57, + "learning_rate": 2.727597283115169e-07, + "loss": 1.0002, + "step": 75329 + }, + { + "epoch": 5.57, + "learning_rate": 2.726672116172624e-07, + "loss": 0.9634, + "step": 75330 + }, + { + "epoch": 5.57, + "learning_rate": 2.725747103990561e-07, + "loss": 0.9993, + "step": 75331 + }, + { + "epoch": 5.57, + "learning_rate": 2.7248222465704243e-07, + "loss": 1.0521, + "step": 75332 + }, + { + "epoch": 5.57, + "learning_rate": 2.723897543913723e-07, + "loss": 0.9855, + "step": 75333 + }, + { + "epoch": 5.57, + "learning_rate": 2.722972996021889e-07, + "loss": 1.1009, + "step": 75334 + }, + { + "epoch": 5.57, + "learning_rate": 2.722048602896443e-07, + "loss": 0.9897, + "step": 75335 + }, + { + "epoch": 5.57, + "learning_rate": 2.7211243645387966e-07, + "loss": 0.9086, + "step": 75336 + }, + { + "epoch": 5.57, + "learning_rate": 2.720200280950469e-07, + "loss": 1.094, + "step": 75337 + }, + { + "epoch": 5.57, + "learning_rate": 2.7192763521329044e-07, + "loss": 0.9752, + "step": 75338 + }, + { + "epoch": 5.57, + "learning_rate": 2.71835257808758e-07, + "loss": 0.8509, + "step": 75339 + }, + { + "epoch": 5.57, + "learning_rate": 2.7174289588159596e-07, + "loss": 0.9315, + "step": 75340 + }, + { + "epoch": 5.57, + "learning_rate": 2.7165054943195213e-07, + "loss": 1.0646, + "step": 75341 + }, + { + "epoch": 5.57, + "learning_rate": 2.71558218459973e-07, + "loss": 0.9166, + "step": 75342 + }, + { + "epoch": 5.57, + "learning_rate": 2.7146590296580623e-07, + "loss": 0.8863, + "step": 75343 + }, + { + "epoch": 5.57, + "learning_rate": 2.7137360294959727e-07, + "loss": 0.9785, + "step": 75344 + }, + { + "epoch": 5.57, + "learning_rate": 2.712813184114926e-07, + "loss": 0.9201, + "step": 75345 + }, + { + "epoch": 5.57, + "learning_rate": 2.7118904935164226e-07, + "loss": 0.9781, + "step": 75346 + }, + { + "epoch": 5.57, + "learning_rate": 2.7109679577019044e-07, + "loss": 0.9532, + "step": 75347 + }, + { + "epoch": 5.57, + "learning_rate": 2.710045576672837e-07, + "loss": 0.9933, + "step": 75348 + }, + { + "epoch": 5.57, + "learning_rate": 2.709123350430709e-07, + "loss": 1.0105, + "step": 75349 + }, + { + "epoch": 5.57, + "learning_rate": 2.708201278976952e-07, + "loss": 0.9958, + "step": 75350 + }, + { + "epoch": 5.57, + "learning_rate": 2.7072793623130645e-07, + "loss": 1.0551, + "step": 75351 + }, + { + "epoch": 5.57, + "learning_rate": 2.7063576004405124e-07, + "loss": 0.8709, + "step": 75352 + }, + { + "epoch": 5.57, + "learning_rate": 2.70543599336075e-07, + "loss": 1.1284, + "step": 75353 + }, + { + "epoch": 5.57, + "learning_rate": 2.704514541075232e-07, + "loss": 0.9092, + "step": 75354 + }, + { + "epoch": 5.57, + "learning_rate": 2.7035932435854675e-07, + "loss": 0.962, + "step": 75355 + }, + { + "epoch": 5.57, + "learning_rate": 2.702672100892878e-07, + "loss": 0.7838, + "step": 75356 + }, + { + "epoch": 5.57, + "learning_rate": 2.701751112998929e-07, + "loss": 1.0201, + "step": 75357 + }, + { + "epoch": 5.57, + "learning_rate": 2.70083027990512e-07, + "loss": 0.9341, + "step": 75358 + }, + { + "epoch": 5.57, + "learning_rate": 2.6999096016129046e-07, + "loss": 0.9264, + "step": 75359 + }, + { + "epoch": 5.57, + "learning_rate": 2.698989078123726e-07, + "loss": 1.0153, + "step": 75360 + }, + { + "epoch": 5.57, + "learning_rate": 2.6980687094390613e-07, + "loss": 0.968, + "step": 75361 + }, + { + "epoch": 5.57, + "learning_rate": 2.6971484955603865e-07, + "loss": 1.0316, + "step": 75362 + }, + { + "epoch": 5.57, + "learning_rate": 2.696228436489157e-07, + "loss": 1.0309, + "step": 75363 + }, + { + "epoch": 5.57, + "learning_rate": 2.6953085322268367e-07, + "loss": 1.0395, + "step": 75364 + }, + { + "epoch": 5.57, + "learning_rate": 2.6943887827748817e-07, + "loss": 0.9912, + "step": 75365 + }, + { + "epoch": 5.57, + "learning_rate": 2.6934691881347673e-07, + "loss": 0.9101, + "step": 75366 + }, + { + "epoch": 5.57, + "learning_rate": 2.692549748307949e-07, + "loss": 0.9278, + "step": 75367 + }, + { + "epoch": 5.57, + "learning_rate": 2.6916304632958914e-07, + "loss": 1.095, + "step": 75368 + }, + { + "epoch": 5.57, + "learning_rate": 2.69071133310006e-07, + "loss": 1.0219, + "step": 75369 + }, + { + "epoch": 5.57, + "learning_rate": 2.6897923577219096e-07, + "loss": 1.0051, + "step": 75370 + }, + { + "epoch": 5.57, + "learning_rate": 2.688873537162906e-07, + "loss": 1.041, + "step": 75371 + }, + { + "epoch": 5.57, + "learning_rate": 2.6879548714245253e-07, + "loss": 1.1445, + "step": 75372 + }, + { + "epoch": 5.57, + "learning_rate": 2.6870363605082105e-07, + "loss": 0.9739, + "step": 75373 + }, + { + "epoch": 5.57, + "learning_rate": 2.686118004415417e-07, + "loss": 0.9881, + "step": 75374 + }, + { + "epoch": 5.57, + "learning_rate": 2.685199803147631e-07, + "loss": 1.0101, + "step": 75375 + }, + { + "epoch": 5.57, + "learning_rate": 2.6842817567062973e-07, + "loss": 0.9817, + "step": 75376 + }, + { + "epoch": 5.57, + "learning_rate": 2.68336386509287e-07, + "loss": 1.0544, + "step": 75377 + }, + { + "epoch": 5.57, + "learning_rate": 2.6824461283088355e-07, + "loss": 1.0775, + "step": 75378 + }, + { + "epoch": 5.57, + "learning_rate": 2.681528546355616e-07, + "loss": 1.0726, + "step": 75379 + }, + { + "epoch": 5.57, + "learning_rate": 2.6806111192347104e-07, + "loss": 0.9038, + "step": 75380 + }, + { + "epoch": 5.57, + "learning_rate": 2.6796938469475397e-07, + "loss": 0.9811, + "step": 75381 + }, + { + "epoch": 5.57, + "learning_rate": 2.6787767294956026e-07, + "loss": 1.0449, + "step": 75382 + }, + { + "epoch": 5.57, + "learning_rate": 2.6778597668803306e-07, + "loss": 1.0364, + "step": 75383 + }, + { + "epoch": 5.57, + "learning_rate": 2.676942959103201e-07, + "loss": 0.9461, + "step": 75384 + }, + { + "epoch": 5.57, + "learning_rate": 2.6760263061656576e-07, + "loss": 1.0922, + "step": 75385 + }, + { + "epoch": 5.57, + "learning_rate": 2.675109808069154e-07, + "loss": 0.9238, + "step": 75386 + }, + { + "epoch": 5.57, + "learning_rate": 2.674193464815167e-07, + "loss": 1.0183, + "step": 75387 + }, + { + "epoch": 5.57, + "learning_rate": 2.6732772764051394e-07, + "loss": 0.9698, + "step": 75388 + }, + { + "epoch": 5.57, + "learning_rate": 2.6723612428405485e-07, + "loss": 1.0587, + "step": 75389 + }, + { + "epoch": 5.57, + "learning_rate": 2.6714453641228157e-07, + "loss": 1.082, + "step": 75390 + }, + { + "epoch": 5.57, + "learning_rate": 2.6705296402534387e-07, + "loss": 0.9865, + "step": 75391 + }, + { + "epoch": 5.57, + "learning_rate": 2.6696140712338615e-07, + "loss": 0.9847, + "step": 75392 + }, + { + "epoch": 5.57, + "learning_rate": 2.668698657065527e-07, + "loss": 1.0772, + "step": 75393 + }, + { + "epoch": 5.57, + "learning_rate": 2.667783397749901e-07, + "loss": 1.0814, + "step": 75394 + }, + { + "epoch": 5.57, + "learning_rate": 2.666868293288427e-07, + "loss": 0.9173, + "step": 75395 + }, + { + "epoch": 5.57, + "learning_rate": 2.6659533436825925e-07, + "loss": 1.0063, + "step": 75396 + }, + { + "epoch": 5.57, + "learning_rate": 2.665038548933829e-07, + "loss": 1.0902, + "step": 75397 + }, + { + "epoch": 5.57, + "learning_rate": 2.6641239090435924e-07, + "loss": 0.878, + "step": 75398 + }, + { + "epoch": 5.57, + "learning_rate": 2.663209424013336e-07, + "loss": 0.9223, + "step": 75399 + }, + { + "epoch": 5.57, + "learning_rate": 2.662295093844536e-07, + "loss": 0.9961, + "step": 75400 + }, + { + "epoch": 5.57, + "learning_rate": 2.661380918538614e-07, + "loss": 0.8631, + "step": 75401 + }, + { + "epoch": 5.57, + "learning_rate": 2.6604668980970583e-07, + "loss": 0.9693, + "step": 75402 + }, + { + "epoch": 5.57, + "learning_rate": 2.6595530325213004e-07, + "loss": 0.9957, + "step": 75403 + }, + { + "epoch": 5.57, + "learning_rate": 2.658639321812795e-07, + "loss": 0.8881, + "step": 75404 + }, + { + "epoch": 5.57, + "learning_rate": 2.657725765973018e-07, + "loss": 0.9044, + "step": 75405 + }, + { + "epoch": 5.57, + "learning_rate": 2.6568123650033804e-07, + "loss": 1.0171, + "step": 75406 + }, + { + "epoch": 5.57, + "learning_rate": 2.655899118905381e-07, + "loss": 0.9715, + "step": 75407 + }, + { + "epoch": 5.57, + "learning_rate": 2.6549860276804506e-07, + "loss": 0.9859, + "step": 75408 + }, + { + "epoch": 5.57, + "learning_rate": 2.6540730913300447e-07, + "loss": 0.9299, + "step": 75409 + }, + { + "epoch": 5.57, + "learning_rate": 2.653160309855607e-07, + "loss": 1.1015, + "step": 75410 + }, + { + "epoch": 5.57, + "learning_rate": 2.652247683258613e-07, + "loss": 0.94, + "step": 75411 + }, + { + "epoch": 5.57, + "learning_rate": 2.6513352115404957e-07, + "loss": 1.0296, + "step": 75412 + }, + { + "epoch": 5.57, + "learning_rate": 2.650422894702709e-07, + "loss": 0.9698, + "step": 75413 + }, + { + "epoch": 5.57, + "learning_rate": 2.6495107327467074e-07, + "loss": 1.0162, + "step": 75414 + }, + { + "epoch": 5.57, + "learning_rate": 2.6485987256739345e-07, + "loss": 1.0113, + "step": 75415 + }, + { + "epoch": 5.57, + "learning_rate": 2.647686873485855e-07, + "loss": 0.8952, + "step": 75416 + }, + { + "epoch": 5.57, + "learning_rate": 2.6467751761839134e-07, + "loss": 1.0054, + "step": 75417 + }, + { + "epoch": 5.57, + "learning_rate": 2.645863633769552e-07, + "loss": 1.1391, + "step": 75418 + }, + { + "epoch": 5.57, + "learning_rate": 2.644952246244237e-07, + "loss": 0.9186, + "step": 75419 + }, + { + "epoch": 5.57, + "learning_rate": 2.6440410136093997e-07, + "loss": 0.8798, + "step": 75420 + }, + { + "epoch": 5.57, + "learning_rate": 2.6431299358665176e-07, + "loss": 1.0344, + "step": 75421 + }, + { + "epoch": 5.57, + "learning_rate": 2.642219013017011e-07, + "loss": 1.0263, + "step": 75422 + }, + { + "epoch": 5.57, + "learning_rate": 2.6413082450623464e-07, + "loss": 0.9911, + "step": 75423 + }, + { + "epoch": 5.57, + "learning_rate": 2.6403976320039546e-07, + "loss": 1.0168, + "step": 75424 + }, + { + "epoch": 5.57, + "learning_rate": 2.639487173843325e-07, + "loss": 1.0206, + "step": 75425 + }, + { + "epoch": 5.57, + "learning_rate": 2.638576870581844e-07, + "loss": 0.9262, + "step": 75426 + }, + { + "epoch": 5.57, + "learning_rate": 2.637666722221022e-07, + "loss": 0.9773, + "step": 75427 + }, + { + "epoch": 5.57, + "learning_rate": 2.6367567287622574e-07, + "loss": 0.938, + "step": 75428 + }, + { + "epoch": 5.57, + "learning_rate": 2.6358468902070387e-07, + "loss": 1.048, + "step": 75429 + }, + { + "epoch": 5.57, + "learning_rate": 2.634937206556776e-07, + "loss": 0.9412, + "step": 75430 + }, + { + "epoch": 5.57, + "learning_rate": 2.634027677812945e-07, + "loss": 0.9845, + "step": 75431 + }, + { + "epoch": 5.57, + "learning_rate": 2.633118303976989e-07, + "loss": 1.0213, + "step": 75432 + }, + { + "epoch": 5.57, + "learning_rate": 2.6322090850503414e-07, + "loss": 1.0544, + "step": 75433 + }, + { + "epoch": 5.57, + "learning_rate": 2.6313000210344666e-07, + "loss": 1.0553, + "step": 75434 + }, + { + "epoch": 5.57, + "learning_rate": 2.6303911119307746e-07, + "loss": 0.9372, + "step": 75435 + }, + { + "epoch": 5.57, + "learning_rate": 2.629482357740765e-07, + "loss": 0.9747, + "step": 75436 + }, + { + "epoch": 5.57, + "learning_rate": 2.6285737584658356e-07, + "loss": 0.9643, + "step": 75437 + }, + { + "epoch": 5.57, + "learning_rate": 2.6276653141074637e-07, + "loss": 1.0868, + "step": 75438 + }, + { + "epoch": 5.57, + "learning_rate": 2.6267570246670816e-07, + "loss": 1.0057, + "step": 75439 + }, + { + "epoch": 5.57, + "learning_rate": 2.625848890146132e-07, + "loss": 1.0824, + "step": 75440 + }, + { + "epoch": 5.57, + "learning_rate": 2.62494091054607e-07, + "loss": 1.0792, + "step": 75441 + }, + { + "epoch": 5.57, + "learning_rate": 2.6240330858683274e-07, + "loss": 1.0245, + "step": 75442 + }, + { + "epoch": 5.57, + "learning_rate": 2.6231254161143585e-07, + "loss": 0.9679, + "step": 75443 + }, + { + "epoch": 5.57, + "learning_rate": 2.6222179012855955e-07, + "loss": 0.9005, + "step": 75444 + }, + { + "epoch": 5.57, + "learning_rate": 2.621310541383515e-07, + "loss": 0.969, + "step": 75445 + }, + { + "epoch": 5.57, + "learning_rate": 2.620403336409505e-07, + "loss": 0.9836, + "step": 75446 + }, + { + "epoch": 5.57, + "learning_rate": 2.619496286365064e-07, + "loss": 1.0107, + "step": 75447 + }, + { + "epoch": 5.57, + "learning_rate": 2.6185893912515913e-07, + "loss": 1.0707, + "step": 75448 + }, + { + "epoch": 5.57, + "learning_rate": 2.6176826510705745e-07, + "loss": 0.9737, + "step": 75449 + }, + { + "epoch": 5.57, + "learning_rate": 2.616776065823412e-07, + "loss": 0.989, + "step": 75450 + }, + { + "epoch": 5.58, + "learning_rate": 2.6158696355115586e-07, + "loss": 1.0504, + "step": 75451 + }, + { + "epoch": 5.58, + "learning_rate": 2.61496336013648e-07, + "loss": 0.9684, + "step": 75452 + }, + { + "epoch": 5.58, + "learning_rate": 2.6140572396995967e-07, + "loss": 1.0645, + "step": 75453 + }, + { + "epoch": 5.58, + "learning_rate": 2.6131512742023636e-07, + "loss": 0.9482, + "step": 75454 + }, + { + "epoch": 5.58, + "learning_rate": 2.61224546364619e-07, + "loss": 1.026, + "step": 75455 + }, + { + "epoch": 5.58, + "learning_rate": 2.6113398080325537e-07, + "loss": 0.9363, + "step": 75456 + }, + { + "epoch": 5.58, + "learning_rate": 2.610434307362886e-07, + "loss": 1.0964, + "step": 75457 + }, + { + "epoch": 5.58, + "learning_rate": 2.60952896163863e-07, + "loss": 1.0487, + "step": 75458 + }, + { + "epoch": 5.58, + "learning_rate": 2.608623770861207e-07, + "loss": 1.0022, + "step": 75459 + }, + { + "epoch": 5.58, + "learning_rate": 2.6077187350320723e-07, + "loss": 0.9018, + "step": 75460 + }, + { + "epoch": 5.58, + "learning_rate": 2.6068138541526677e-07, + "loss": 1.0584, + "step": 75461 + }, + { + "epoch": 5.58, + "learning_rate": 2.6059091282244263e-07, + "loss": 0.8974, + "step": 75462 + }, + { + "epoch": 5.58, + "learning_rate": 2.6050045572487914e-07, + "loss": 1.0439, + "step": 75463 + }, + { + "epoch": 5.58, + "learning_rate": 2.604100141227195e-07, + "loss": 0.9928, + "step": 75464 + }, + { + "epoch": 5.58, + "learning_rate": 2.60319588016108e-07, + "loss": 0.9938, + "step": 75465 + }, + { + "epoch": 5.58, + "learning_rate": 2.6022917740519015e-07, + "loss": 0.8858, + "step": 75466 + }, + { + "epoch": 5.58, + "learning_rate": 2.60138782290108e-07, + "loss": 0.9475, + "step": 75467 + }, + { + "epoch": 5.58, + "learning_rate": 2.6004840267100487e-07, + "loss": 1.0549, + "step": 75468 + }, + { + "epoch": 5.58, + "learning_rate": 2.5995803854802495e-07, + "loss": 1.0651, + "step": 75469 + }, + { + "epoch": 5.58, + "learning_rate": 2.598676899213148e-07, + "loss": 1.0093, + "step": 75470 + }, + { + "epoch": 5.58, + "learning_rate": 2.5977735679101337e-07, + "loss": 1.0667, + "step": 75471 + }, + { + "epoch": 5.58, + "learning_rate": 2.59687039157267e-07, + "loss": 0.9281, + "step": 75472 + }, + { + "epoch": 5.58, + "learning_rate": 2.5959673702021904e-07, + "loss": 0.9423, + "step": 75473 + }, + { + "epoch": 5.58, + "learning_rate": 2.59506450380016e-07, + "loss": 1.0194, + "step": 75474 + }, + { + "epoch": 5.58, + "learning_rate": 2.594161792367955e-07, + "loss": 0.9685, + "step": 75475 + }, + { + "epoch": 5.58, + "learning_rate": 2.593259235907064e-07, + "loss": 1.046, + "step": 75476 + }, + { + "epoch": 5.58, + "learning_rate": 2.5923568344188967e-07, + "loss": 0.9001, + "step": 75477 + }, + { + "epoch": 5.58, + "learning_rate": 2.5914545879049067e-07, + "loss": 1.0464, + "step": 75478 + }, + { + "epoch": 5.58, + "learning_rate": 2.590552496366505e-07, + "loss": 1.0028, + "step": 75479 + }, + { + "epoch": 5.58, + "learning_rate": 2.589650559805146e-07, + "loss": 1.1043, + "step": 75480 + }, + { + "epoch": 5.58, + "learning_rate": 2.588748778222261e-07, + "loss": 0.994, + "step": 75481 + }, + { + "epoch": 5.58, + "learning_rate": 2.587847151619283e-07, + "loss": 1.0692, + "step": 75482 + }, + { + "epoch": 5.58, + "learning_rate": 2.586945679997643e-07, + "loss": 1.0196, + "step": 75483 + }, + { + "epoch": 5.58, + "learning_rate": 2.586044363358775e-07, + "loss": 0.9443, + "step": 75484 + }, + { + "epoch": 5.58, + "learning_rate": 2.5851432017041213e-07, + "loss": 1.0167, + "step": 75485 + }, + { + "epoch": 5.58, + "learning_rate": 2.584242195035114e-07, + "loss": 1.0107, + "step": 75486 + }, + { + "epoch": 5.58, + "learning_rate": 2.583341343353185e-07, + "loss": 0.968, + "step": 75487 + }, + { + "epoch": 5.58, + "learning_rate": 2.582440646659756e-07, + "loss": 0.9157, + "step": 75488 + }, + { + "epoch": 5.58, + "learning_rate": 2.5815401049562704e-07, + "loss": 0.9266, + "step": 75489 + }, + { + "epoch": 5.58, + "learning_rate": 2.580639718244182e-07, + "loss": 0.9891, + "step": 75490 + }, + { + "epoch": 5.58, + "learning_rate": 2.579739486524879e-07, + "loss": 0.9652, + "step": 75491 + }, + { + "epoch": 5.58, + "learning_rate": 2.578839409799827e-07, + "loss": 0.9875, + "step": 75492 + }, + { + "epoch": 5.58, + "learning_rate": 2.577939488070436e-07, + "loss": 1.0501, + "step": 75493 + }, + { + "epoch": 5.58, + "learning_rate": 2.57703972133817e-07, + "loss": 0.8635, + "step": 75494 + }, + { + "epoch": 5.58, + "learning_rate": 2.576140109604419e-07, + "loss": 0.945, + "step": 75495 + }, + { + "epoch": 5.58, + "learning_rate": 2.5752406528706473e-07, + "loss": 0.9893, + "step": 75496 + }, + { + "epoch": 5.58, + "learning_rate": 2.5743413511382764e-07, + "loss": 0.9939, + "step": 75497 + }, + { + "epoch": 5.58, + "learning_rate": 2.573442204408727e-07, + "loss": 1.0375, + "step": 75498 + }, + { + "epoch": 5.58, + "learning_rate": 2.572543212683443e-07, + "loss": 0.9919, + "step": 75499 + }, + { + "epoch": 5.58, + "learning_rate": 2.571644375963833e-07, + "loss": 0.9894, + "step": 75500 + }, + { + "epoch": 5.58, + "learning_rate": 2.5707456942513533e-07, + "loss": 0.9669, + "step": 75501 + }, + { + "epoch": 5.58, + "learning_rate": 2.5698471675474344e-07, + "loss": 1.0038, + "step": 75502 + }, + { + "epoch": 5.58, + "learning_rate": 2.5689487958534877e-07, + "loss": 1.0493, + "step": 75503 + }, + { + "epoch": 5.58, + "learning_rate": 2.568050579170933e-07, + "loss": 1.0227, + "step": 75504 + }, + { + "epoch": 5.58, + "learning_rate": 2.5671525175012366e-07, + "loss": 0.9994, + "step": 75505 + }, + { + "epoch": 5.58, + "learning_rate": 2.566254610845798e-07, + "loss": 0.8837, + "step": 75506 + }, + { + "epoch": 5.58, + "learning_rate": 2.5653568592060587e-07, + "loss": 1.0376, + "step": 75507 + }, + { + "epoch": 5.58, + "learning_rate": 2.564459262583441e-07, + "loss": 0.9378, + "step": 75508 + }, + { + "epoch": 5.58, + "learning_rate": 2.5635618209793656e-07, + "loss": 0.9801, + "step": 75509 + }, + { + "epoch": 5.58, + "learning_rate": 2.562664534395276e-07, + "loss": 0.9361, + "step": 75510 + }, + { + "epoch": 5.58, + "learning_rate": 2.5617674028325935e-07, + "loss": 1.0683, + "step": 75511 + }, + { + "epoch": 5.58, + "learning_rate": 2.56087042629275e-07, + "loss": 1.0106, + "step": 75512 + }, + { + "epoch": 5.58, + "learning_rate": 2.559973604777155e-07, + "loss": 0.976, + "step": 75513 + }, + { + "epoch": 5.58, + "learning_rate": 2.5590769382872525e-07, + "loss": 0.9947, + "step": 75514 + }, + { + "epoch": 5.58, + "learning_rate": 2.558180426824486e-07, + "loss": 1.0606, + "step": 75515 + }, + { + "epoch": 5.58, + "learning_rate": 2.5572840703902314e-07, + "loss": 0.9809, + "step": 75516 + }, + { + "epoch": 5.58, + "learning_rate": 2.5563878689859546e-07, + "loss": 0.9756, + "step": 75517 + }, + { + "epoch": 5.58, + "learning_rate": 2.555491822613065e-07, + "loss": 0.9611, + "step": 75518 + }, + { + "epoch": 5.58, + "learning_rate": 2.5545959312730073e-07, + "loss": 0.926, + "step": 75519 + }, + { + "epoch": 5.58, + "learning_rate": 2.553700194967179e-07, + "loss": 0.9556, + "step": 75520 + }, + { + "epoch": 5.58, + "learning_rate": 2.5528046136970243e-07, + "loss": 0.9679, + "step": 75521 + }, + { + "epoch": 5.58, + "learning_rate": 2.551909187463963e-07, + "loss": 1.0672, + "step": 75522 + }, + { + "epoch": 5.58, + "learning_rate": 2.5510139162694294e-07, + "loss": 1.0707, + "step": 75523 + }, + { + "epoch": 5.58, + "learning_rate": 2.55011880011482e-07, + "loss": 0.9276, + "step": 75524 + }, + { + "epoch": 5.58, + "learning_rate": 2.549223839001591e-07, + "loss": 0.9272, + "step": 75525 + }, + { + "epoch": 5.58, + "learning_rate": 2.548329032931152e-07, + "loss": 0.9648, + "step": 75526 + }, + { + "epoch": 5.58, + "learning_rate": 2.547434381904923e-07, + "loss": 0.8517, + "step": 75527 + }, + { + "epoch": 5.58, + "learning_rate": 2.546539885924337e-07, + "loss": 0.975, + "step": 75528 + }, + { + "epoch": 5.58, + "learning_rate": 2.5456455449907936e-07, + "loss": 1.0057, + "step": 75529 + }, + { + "epoch": 5.58, + "learning_rate": 2.544751359105757e-07, + "loss": 0.9571, + "step": 75530 + }, + { + "epoch": 5.58, + "learning_rate": 2.5438573282706157e-07, + "loss": 1.0325, + "step": 75531 + }, + { + "epoch": 5.58, + "learning_rate": 2.5429634524868016e-07, + "loss": 1.0263, + "step": 75532 + }, + { + "epoch": 5.58, + "learning_rate": 2.5420697317557365e-07, + "loss": 1.0216, + "step": 75533 + }, + { + "epoch": 5.58, + "learning_rate": 2.5411761660788514e-07, + "loss": 0.9971, + "step": 75534 + }, + { + "epoch": 5.58, + "learning_rate": 2.540282755457568e-07, + "loss": 0.9898, + "step": 75535 + }, + { + "epoch": 5.58, + "learning_rate": 2.5393894998932856e-07, + "loss": 0.987, + "step": 75536 + }, + { + "epoch": 5.58, + "learning_rate": 2.5384963993874467e-07, + "loss": 1.0538, + "step": 75537 + }, + { + "epoch": 5.58, + "learning_rate": 2.537603453941451e-07, + "loss": 1.0333, + "step": 75538 + }, + { + "epoch": 5.58, + "learning_rate": 2.536710663556763e-07, + "loss": 0.967, + "step": 75539 + }, + { + "epoch": 5.58, + "learning_rate": 2.535818028234749e-07, + "loss": 1.0746, + "step": 75540 + }, + { + "epoch": 5.58, + "learning_rate": 2.534925547976863e-07, + "loss": 1.0244, + "step": 75541 + }, + { + "epoch": 5.58, + "learning_rate": 2.5340332227845045e-07, + "loss": 1.035, + "step": 75542 + }, + { + "epoch": 5.58, + "learning_rate": 2.533141052659127e-07, + "loss": 0.9062, + "step": 75543 + }, + { + "epoch": 5.58, + "learning_rate": 2.532249037602119e-07, + "loss": 0.958, + "step": 75544 + }, + { + "epoch": 5.58, + "learning_rate": 2.5313571776149017e-07, + "loss": 0.9412, + "step": 75545 + }, + { + "epoch": 5.58, + "learning_rate": 2.5304654726989066e-07, + "loss": 1.1136, + "step": 75546 + }, + { + "epoch": 5.58, + "learning_rate": 2.5295739228555437e-07, + "loss": 1.0413, + "step": 75547 + }, + { + "epoch": 5.58, + "learning_rate": 2.5286825280862347e-07, + "loss": 0.9515, + "step": 75548 + }, + { + "epoch": 5.58, + "learning_rate": 2.5277912883923895e-07, + "loss": 0.9707, + "step": 75549 + }, + { + "epoch": 5.58, + "learning_rate": 2.5269002037754396e-07, + "loss": 1.0208, + "step": 75550 + }, + { + "epoch": 5.58, + "learning_rate": 2.526009274236796e-07, + "loss": 0.9885, + "step": 75551 + }, + { + "epoch": 5.58, + "learning_rate": 2.52511849977789e-07, + "loss": 0.8297, + "step": 75552 + }, + { + "epoch": 5.58, + "learning_rate": 2.5242278804001096e-07, + "loss": 1.0166, + "step": 75553 + }, + { + "epoch": 5.58, + "learning_rate": 2.5233374161048873e-07, + "loss": 0.9437, + "step": 75554 + }, + { + "epoch": 5.58, + "learning_rate": 2.522447106893655e-07, + "loss": 0.9286, + "step": 75555 + }, + { + "epoch": 5.58, + "learning_rate": 2.521556952767801e-07, + "loss": 1.001, + "step": 75556 + }, + { + "epoch": 5.58, + "learning_rate": 2.5206669537287675e-07, + "loss": 1.0347, + "step": 75557 + }, + { + "epoch": 5.58, + "learning_rate": 2.519777109777943e-07, + "loss": 1.1842, + "step": 75558 + }, + { + "epoch": 5.58, + "learning_rate": 2.5188874209167713e-07, + "loss": 0.9627, + "step": 75559 + }, + { + "epoch": 5.58, + "learning_rate": 2.5179978871466503e-07, + "loss": 0.9784, + "step": 75560 + }, + { + "epoch": 5.58, + "learning_rate": 2.5171085084690016e-07, + "loss": 1.0555, + "step": 75561 + }, + { + "epoch": 5.58, + "learning_rate": 2.5162192848852353e-07, + "loss": 1.0146, + "step": 75562 + }, + { + "epoch": 5.58, + "learning_rate": 2.515330216396761e-07, + "loss": 1.1192, + "step": 75563 + }, + { + "epoch": 5.58, + "learning_rate": 2.514441303005022e-07, + "loss": 0.9704, + "step": 75564 + }, + { + "epoch": 5.58, + "learning_rate": 2.513552544711395e-07, + "loss": 1.0143, + "step": 75565 + }, + { + "epoch": 5.58, + "learning_rate": 2.512663941517324e-07, + "loss": 0.9657, + "step": 75566 + }, + { + "epoch": 5.58, + "learning_rate": 2.5117754934242065e-07, + "loss": 1.1737, + "step": 75567 + }, + { + "epoch": 5.58, + "learning_rate": 2.510887200433454e-07, + "loss": 0.956, + "step": 75568 + }, + { + "epoch": 5.58, + "learning_rate": 2.509999062546475e-07, + "loss": 1.0851, + "step": 75569 + }, + { + "epoch": 5.58, + "learning_rate": 2.5091110797647145e-07, + "loss": 0.9776, + "step": 75570 + }, + { + "epoch": 5.58, + "learning_rate": 2.5082232520895586e-07, + "loss": 1.0581, + "step": 75571 + }, + { + "epoch": 5.58, + "learning_rate": 2.5073355795224186e-07, + "loss": 1.0308, + "step": 75572 + }, + { + "epoch": 5.58, + "learning_rate": 2.5064480620647147e-07, + "loss": 1.0963, + "step": 75573 + }, + { + "epoch": 5.58, + "learning_rate": 2.505560699717846e-07, + "loss": 0.9861, + "step": 75574 + }, + { + "epoch": 5.58, + "learning_rate": 2.504673492483245e-07, + "loss": 1.0216, + "step": 75575 + }, + { + "epoch": 5.58, + "learning_rate": 2.5037864403623215e-07, + "loss": 1.0049, + "step": 75576 + }, + { + "epoch": 5.58, + "learning_rate": 2.502899543356474e-07, + "loss": 0.8927, + "step": 75577 + }, + { + "epoch": 5.58, + "learning_rate": 2.502012801467113e-07, + "loss": 1.0669, + "step": 75578 + }, + { + "epoch": 5.58, + "learning_rate": 2.5011262146956594e-07, + "loss": 0.9876, + "step": 75579 + }, + { + "epoch": 5.58, + "learning_rate": 2.500239783043512e-07, + "loss": 0.9435, + "step": 75580 + }, + { + "epoch": 5.58, + "learning_rate": 2.499353506512103e-07, + "loss": 1.0624, + "step": 75581 + }, + { + "epoch": 5.58, + "learning_rate": 2.4984673851028097e-07, + "loss": 1.0605, + "step": 75582 + }, + { + "epoch": 5.58, + "learning_rate": 2.497581418817063e-07, + "loss": 0.972, + "step": 75583 + }, + { + "epoch": 5.58, + "learning_rate": 2.4966956076562856e-07, + "loss": 1.091, + "step": 75584 + }, + { + "epoch": 5.58, + "learning_rate": 2.4958099516218526e-07, + "loss": 1.0536, + "step": 75585 + }, + { + "epoch": 5.59, + "learning_rate": 2.494924450715197e-07, + "loss": 1.0335, + "step": 75586 + }, + { + "epoch": 5.59, + "learning_rate": 2.4940391049377176e-07, + "loss": 1.0148, + "step": 75587 + }, + { + "epoch": 5.59, + "learning_rate": 2.493153914290847e-07, + "loss": 1.024, + "step": 75588 + }, + { + "epoch": 5.59, + "learning_rate": 2.4922688787759495e-07, + "loss": 1.0562, + "step": 75589 + }, + { + "epoch": 5.59, + "learning_rate": 2.49138399839447e-07, + "loss": 1.0426, + "step": 75590 + }, + { + "epoch": 5.59, + "learning_rate": 2.490499273147795e-07, + "loss": 0.8894, + "step": 75591 + }, + { + "epoch": 5.59, + "learning_rate": 2.4896147030373465e-07, + "loss": 1.0683, + "step": 75592 + }, + { + "epoch": 5.59, + "learning_rate": 2.4887302880645224e-07, + "loss": 0.9071, + "step": 75593 + }, + { + "epoch": 5.59, + "learning_rate": 2.4878460282307335e-07, + "loss": 0.914, + "step": 75594 + }, + { + "epoch": 5.59, + "learning_rate": 2.48696192353739e-07, + "loss": 1.0248, + "step": 75595 + }, + { + "epoch": 5.59, + "learning_rate": 2.48607797398589e-07, + "loss": 0.9591, + "step": 75596 + }, + { + "epoch": 5.59, + "learning_rate": 2.4851941795776547e-07, + "loss": 0.9823, + "step": 75597 + }, + { + "epoch": 5.59, + "learning_rate": 2.484310540314061e-07, + "loss": 1.1613, + "step": 75598 + }, + { + "epoch": 5.59, + "learning_rate": 2.4834270561965413e-07, + "loss": 1.018, + "step": 75599 + }, + { + "epoch": 5.59, + "learning_rate": 2.482543727226505e-07, + "loss": 1.0312, + "step": 75600 + }, + { + "epoch": 5.59, + "learning_rate": 2.4816605534053405e-07, + "loss": 1.0651, + "step": 75601 + }, + { + "epoch": 5.59, + "learning_rate": 2.4807775347344577e-07, + "loss": 0.9864, + "step": 75602 + }, + { + "epoch": 5.59, + "learning_rate": 2.4798946712152547e-07, + "loss": 0.9892, + "step": 75603 + }, + { + "epoch": 5.59, + "learning_rate": 2.479011962849154e-07, + "loss": 0.9923, + "step": 75604 + }, + { + "epoch": 5.59, + "learning_rate": 2.478129409637553e-07, + "loss": 1.0716, + "step": 75605 + }, + { + "epoch": 5.59, + "learning_rate": 2.4772470115818515e-07, + "loss": 1.068, + "step": 75606 + }, + { + "epoch": 5.59, + "learning_rate": 2.476364768683448e-07, + "loss": 1.0755, + "step": 75607 + }, + { + "epoch": 5.59, + "learning_rate": 2.475482680943753e-07, + "loss": 1.0522, + "step": 75608 + }, + { + "epoch": 5.59, + "learning_rate": 2.4746007483641864e-07, + "loss": 0.9709, + "step": 75609 + }, + { + "epoch": 5.59, + "learning_rate": 2.4737189709461154e-07, + "loss": 1.0055, + "step": 75610 + }, + { + "epoch": 5.59, + "learning_rate": 2.4728373486909595e-07, + "loss": 1.0603, + "step": 75611 + }, + { + "epoch": 5.59, + "learning_rate": 2.471955881600141e-07, + "loss": 1.075, + "step": 75612 + }, + { + "epoch": 5.59, + "learning_rate": 2.471074569675036e-07, + "loss": 1.0183, + "step": 75613 + }, + { + "epoch": 5.59, + "learning_rate": 2.4701934129170437e-07, + "loss": 0.954, + "step": 75614 + }, + { + "epoch": 5.59, + "learning_rate": 2.4693124113275955e-07, + "loss": 1.0078, + "step": 75615 + }, + { + "epoch": 5.59, + "learning_rate": 2.46843156490808e-07, + "loss": 1.0232, + "step": 75616 + }, + { + "epoch": 5.59, + "learning_rate": 2.467550873659885e-07, + "loss": 0.9553, + "step": 75617 + }, + { + "epoch": 5.59, + "learning_rate": 2.4666703375844204e-07, + "loss": 0.906, + "step": 75618 + }, + { + "epoch": 5.59, + "learning_rate": 2.4657899566830845e-07, + "loss": 0.9885, + "step": 75619 + }, + { + "epoch": 5.59, + "learning_rate": 2.4649097309572877e-07, + "loss": 0.9829, + "step": 75620 + }, + { + "epoch": 5.59, + "learning_rate": 2.464029660408418e-07, + "loss": 1.0307, + "step": 75621 + }, + { + "epoch": 5.59, + "learning_rate": 2.463149745037896e-07, + "loss": 1.1412, + "step": 75622 + }, + { + "epoch": 5.59, + "learning_rate": 2.4622699848470875e-07, + "loss": 1.1343, + "step": 75623 + }, + { + "epoch": 5.59, + "learning_rate": 2.461390379837425e-07, + "loss": 1.0753, + "step": 75624 + }, + { + "epoch": 5.59, + "learning_rate": 2.4605109300102957e-07, + "loss": 1.0215, + "step": 75625 + }, + { + "epoch": 5.59, + "learning_rate": 2.4596316353670877e-07, + "loss": 0.9562, + "step": 75626 + }, + { + "epoch": 5.59, + "learning_rate": 2.458752495909222e-07, + "loss": 0.9893, + "step": 75627 + }, + { + "epoch": 5.59, + "learning_rate": 2.457873511638076e-07, + "loss": 1.1044, + "step": 75628 + }, + { + "epoch": 5.59, + "learning_rate": 2.4569946825550697e-07, + "loss": 1.0197, + "step": 75629 + }, + { + "epoch": 5.59, + "learning_rate": 2.4561160086615685e-07, + "loss": 0.9388, + "step": 75630 + }, + { + "epoch": 5.59, + "learning_rate": 2.455237489959006e-07, + "loss": 0.9732, + "step": 75631 + }, + { + "epoch": 5.59, + "learning_rate": 2.454359126448758e-07, + "loss": 0.8723, + "step": 75632 + }, + { + "epoch": 5.59, + "learning_rate": 2.453480918132245e-07, + "loss": 1.0536, + "step": 75633 + }, + { + "epoch": 5.59, + "learning_rate": 2.4526028650108224e-07, + "loss": 1.06, + "step": 75634 + }, + { + "epoch": 5.59, + "learning_rate": 2.451724967085933e-07, + "loss": 0.9853, + "step": 75635 + }, + { + "epoch": 5.59, + "learning_rate": 2.450847224358943e-07, + "loss": 1.0268, + "step": 75636 + }, + { + "epoch": 5.59, + "learning_rate": 2.449969636831273e-07, + "loss": 1.0602, + "step": 75637 + }, + { + "epoch": 5.59, + "learning_rate": 2.4490922045042996e-07, + "loss": 1.0321, + "step": 75638 + }, + { + "epoch": 5.59, + "learning_rate": 2.448214927379411e-07, + "loss": 0.9822, + "step": 75639 + }, + { + "epoch": 5.59, + "learning_rate": 2.447337805458028e-07, + "loss": 1.0585, + "step": 75640 + }, + { + "epoch": 5.59, + "learning_rate": 2.4464608387415377e-07, + "loss": 0.969, + "step": 75641 + }, + { + "epoch": 5.59, + "learning_rate": 2.445584027231329e-07, + "loss": 1.0145, + "step": 75642 + }, + { + "epoch": 5.59, + "learning_rate": 2.4447073709288007e-07, + "loss": 0.9785, + "step": 75643 + }, + { + "epoch": 5.59, + "learning_rate": 2.4438308698353505e-07, + "loss": 0.9992, + "step": 75644 + }, + { + "epoch": 5.59, + "learning_rate": 2.442954523952368e-07, + "loss": 1.0129, + "step": 75645 + }, + { + "epoch": 5.59, + "learning_rate": 2.442078333281239e-07, + "loss": 0.947, + "step": 75646 + }, + { + "epoch": 5.59, + "learning_rate": 2.441202297823386e-07, + "loss": 0.9321, + "step": 75647 + }, + { + "epoch": 5.59, + "learning_rate": 2.4403264175801634e-07, + "loss": 0.9695, + "step": 75648 + }, + { + "epoch": 5.59, + "learning_rate": 2.4394506925530025e-07, + "loss": 1.0509, + "step": 75649 + }, + { + "epoch": 5.59, + "learning_rate": 2.4385751227432696e-07, + "loss": 1.025, + "step": 75650 + }, + { + "epoch": 5.59, + "learning_rate": 2.4376997081523746e-07, + "loss": 0.9866, + "step": 75651 + }, + { + "epoch": 5.59, + "learning_rate": 2.436824448781694e-07, + "loss": 0.9808, + "step": 75652 + }, + { + "epoch": 5.59, + "learning_rate": 2.435949344632649e-07, + "loss": 0.8774, + "step": 75653 + }, + { + "epoch": 5.59, + "learning_rate": 2.4350743957065935e-07, + "loss": 1.0659, + "step": 75654 + }, + { + "epoch": 5.59, + "learning_rate": 2.4341996020049387e-07, + "loss": 0.9774, + "step": 75655 + }, + { + "epoch": 5.59, + "learning_rate": 2.4333249635290934e-07, + "loss": 1.0456, + "step": 75656 + }, + { + "epoch": 5.59, + "learning_rate": 2.432450480280424e-07, + "loss": 1.12, + "step": 75657 + }, + { + "epoch": 5.59, + "learning_rate": 2.4315761522603285e-07, + "loss": 1.0588, + "step": 75658 + }, + { + "epoch": 5.59, + "learning_rate": 2.4307019794701846e-07, + "loss": 0.8567, + "step": 75659 + }, + { + "epoch": 5.59, + "learning_rate": 2.429827961911413e-07, + "loss": 0.9609, + "step": 75660 + }, + { + "epoch": 5.59, + "learning_rate": 2.42895409958539e-07, + "loss": 1.0814, + "step": 75661 + }, + { + "epoch": 5.59, + "learning_rate": 2.428080392493504e-07, + "loss": 0.9305, + "step": 75662 + }, + { + "epoch": 5.59, + "learning_rate": 2.427206840637142e-07, + "loss": 1.0679, + "step": 75663 + }, + { + "epoch": 5.59, + "learning_rate": 2.4263334440177034e-07, + "loss": 1.0231, + "step": 75664 + }, + { + "epoch": 5.59, + "learning_rate": 2.425460202636565e-07, + "loss": 1.0389, + "step": 75665 + }, + { + "epoch": 5.59, + "learning_rate": 2.424587116495136e-07, + "loss": 0.881, + "step": 75666 + }, + { + "epoch": 5.59, + "learning_rate": 2.423714185594783e-07, + "loss": 1.0288, + "step": 75667 + }, + { + "epoch": 5.59, + "learning_rate": 2.4228414099368935e-07, + "loss": 0.9595, + "step": 75668 + }, + { + "epoch": 5.59, + "learning_rate": 2.4219687895228885e-07, + "loss": 1.0353, + "step": 75669 + }, + { + "epoch": 5.59, + "learning_rate": 2.421096324354122e-07, + "loss": 1.0069, + "step": 75670 + }, + { + "epoch": 5.59, + "learning_rate": 2.4202240144320043e-07, + "loss": 0.9445, + "step": 75671 + }, + { + "epoch": 5.59, + "learning_rate": 2.419351859757901e-07, + "loss": 0.9266, + "step": 75672 + }, + { + "epoch": 5.59, + "learning_rate": 2.418479860333223e-07, + "loss": 0.9995, + "step": 75673 + }, + { + "epoch": 5.59, + "learning_rate": 2.4176080161593563e-07, + "loss": 0.9813, + "step": 75674 + }, + { + "epoch": 5.59, + "learning_rate": 2.4167363272376674e-07, + "loss": 1.0368, + "step": 75675 + }, + { + "epoch": 5.59, + "learning_rate": 2.415864793569556e-07, + "loss": 1.0135, + "step": 75676 + }, + { + "epoch": 5.59, + "learning_rate": 2.414993415156397e-07, + "loss": 0.9108, + "step": 75677 + }, + { + "epoch": 5.59, + "learning_rate": 2.414122191999613e-07, + "loss": 1.0179, + "step": 75678 + }, + { + "epoch": 5.59, + "learning_rate": 2.4132511241005353e-07, + "loss": 1.0025, + "step": 75679 + }, + { + "epoch": 5.59, + "learning_rate": 2.4123802114605964e-07, + "loss": 1.0237, + "step": 75680 + }, + { + "epoch": 5.59, + "learning_rate": 2.4115094540811514e-07, + "loss": 1.054, + "step": 75681 + }, + { + "epoch": 5.59, + "learning_rate": 2.41063885196362e-07, + "loss": 1.0791, + "step": 75682 + }, + { + "epoch": 5.59, + "learning_rate": 2.4097684051093584e-07, + "loss": 0.9013, + "step": 75683 + }, + { + "epoch": 5.59, + "learning_rate": 2.408898113519742e-07, + "loss": 0.9684, + "step": 75684 + }, + { + "epoch": 5.59, + "learning_rate": 2.4080279771961925e-07, + "loss": 0.9324, + "step": 75685 + }, + { + "epoch": 5.59, + "learning_rate": 2.4071579961400636e-07, + "loss": 1.0503, + "step": 75686 + }, + { + "epoch": 5.59, + "learning_rate": 2.4062881703527554e-07, + "loss": 1.0141, + "step": 75687 + }, + { + "epoch": 5.59, + "learning_rate": 2.405418499835632e-07, + "loss": 1.0046, + "step": 75688 + }, + { + "epoch": 5.59, + "learning_rate": 2.404548984590105e-07, + "loss": 0.9321, + "step": 75689 + }, + { + "epoch": 5.59, + "learning_rate": 2.40367962461755e-07, + "loss": 0.9971, + "step": 75690 + }, + { + "epoch": 5.59, + "learning_rate": 2.4028104199193323e-07, + "loss": 0.8807, + "step": 75691 + }, + { + "epoch": 5.59, + "learning_rate": 2.401941370496852e-07, + "loss": 0.931, + "step": 75692 + }, + { + "epoch": 5.59, + "learning_rate": 2.4010724763514847e-07, + "loss": 0.9819, + "step": 75693 + }, + { + "epoch": 5.59, + "learning_rate": 2.400203737484619e-07, + "loss": 0.9895, + "step": 75694 + }, + { + "epoch": 5.59, + "learning_rate": 2.3993351538976305e-07, + "loss": 0.9329, + "step": 75695 + }, + { + "epoch": 5.59, + "learning_rate": 2.398466725591897e-07, + "loss": 0.9938, + "step": 75696 + }, + { + "epoch": 5.59, + "learning_rate": 2.3975984525688057e-07, + "loss": 1.0506, + "step": 75697 + }, + { + "epoch": 5.59, + "learning_rate": 2.3967303348297556e-07, + "loss": 0.9925, + "step": 75698 + }, + { + "epoch": 5.59, + "learning_rate": 2.395862372376101e-07, + "loss": 1.0547, + "step": 75699 + }, + { + "epoch": 5.59, + "learning_rate": 2.3949945652092297e-07, + "loss": 0.9463, + "step": 75700 + }, + { + "epoch": 5.59, + "learning_rate": 2.394126913330519e-07, + "loss": 0.9153, + "step": 75701 + }, + { + "epoch": 5.59, + "learning_rate": 2.393259416741378e-07, + "loss": 0.9913, + "step": 75702 + }, + { + "epoch": 5.59, + "learning_rate": 2.3923920754431507e-07, + "loss": 0.9875, + "step": 75703 + }, + { + "epoch": 5.59, + "learning_rate": 2.3915248894372244e-07, + "loss": 0.9024, + "step": 75704 + }, + { + "epoch": 5.59, + "learning_rate": 2.3906578587249983e-07, + "loss": 1.0408, + "step": 75705 + }, + { + "epoch": 5.59, + "learning_rate": 2.3897909833078383e-07, + "loss": 0.8571, + "step": 75706 + }, + { + "epoch": 5.59, + "learning_rate": 2.3889242631871203e-07, + "loss": 1.0529, + "step": 75707 + }, + { + "epoch": 5.59, + "learning_rate": 2.388057698364221e-07, + "loss": 0.9309, + "step": 75708 + }, + { + "epoch": 5.59, + "learning_rate": 2.387191288840529e-07, + "loss": 0.9913, + "step": 75709 + }, + { + "epoch": 5.59, + "learning_rate": 2.3863250346174205e-07, + "loss": 1.0888, + "step": 75710 + }, + { + "epoch": 5.59, + "learning_rate": 2.3854589356962723e-07, + "loss": 0.943, + "step": 75711 + }, + { + "epoch": 5.59, + "learning_rate": 2.3845929920784715e-07, + "loss": 1.0668, + "step": 75712 + }, + { + "epoch": 5.59, + "learning_rate": 2.3837272037653624e-07, + "loss": 1.0683, + "step": 75713 + }, + { + "epoch": 5.59, + "learning_rate": 2.382861570758366e-07, + "loss": 0.9239, + "step": 75714 + }, + { + "epoch": 5.59, + "learning_rate": 2.381996093058836e-07, + "loss": 1.03, + "step": 75715 + }, + { + "epoch": 5.59, + "learning_rate": 2.3811307706681607e-07, + "loss": 0.881, + "step": 75716 + }, + { + "epoch": 5.59, + "learning_rate": 2.3802656035876947e-07, + "loss": 0.911, + "step": 75717 + }, + { + "epoch": 5.59, + "learning_rate": 2.3794005918188368e-07, + "loss": 0.9041, + "step": 75718 + }, + { + "epoch": 5.59, + "learning_rate": 2.3785357353629634e-07, + "loss": 0.9987, + "step": 75719 + }, + { + "epoch": 5.59, + "learning_rate": 2.3776710342214292e-07, + "loss": 1.082, + "step": 75720 + }, + { + "epoch": 5.59, + "learning_rate": 2.3768064883956332e-07, + "loss": 0.9239, + "step": 75721 + }, + { + "epoch": 5.6, + "learning_rate": 2.3759420978869296e-07, + "loss": 0.9262, + "step": 75722 + }, + { + "epoch": 5.6, + "learning_rate": 2.3750778626967285e-07, + "loss": 0.8238, + "step": 75723 + }, + { + "epoch": 5.6, + "learning_rate": 2.3742137828263622e-07, + "loss": 0.9751, + "step": 75724 + }, + { + "epoch": 5.6, + "learning_rate": 2.3733498582772295e-07, + "loss": 0.9429, + "step": 75725 + }, + { + "epoch": 5.6, + "learning_rate": 2.3724860890506962e-07, + "loss": 1.0165, + "step": 75726 + }, + { + "epoch": 5.6, + "learning_rate": 2.371622475148161e-07, + "loss": 0.9385, + "step": 75727 + }, + { + "epoch": 5.6, + "learning_rate": 2.3707590165709559e-07, + "loss": 1.0264, + "step": 75728 + }, + { + "epoch": 5.6, + "learning_rate": 2.3698957133204802e-07, + "loss": 1.0771, + "step": 75729 + }, + { + "epoch": 5.6, + "learning_rate": 2.3690325653981104e-07, + "loss": 0.9226, + "step": 75730 + }, + { + "epoch": 5.6, + "learning_rate": 2.3681695728052124e-07, + "loss": 1.0473, + "step": 75731 + }, + { + "epoch": 5.6, + "learning_rate": 2.3673067355431513e-07, + "loss": 0.8409, + "step": 75732 + }, + { + "epoch": 5.6, + "learning_rate": 2.3664440536133038e-07, + "loss": 1.0887, + "step": 75733 + }, + { + "epoch": 5.6, + "learning_rate": 2.365581527017058e-07, + "loss": 1.0922, + "step": 75734 + }, + { + "epoch": 5.6, + "learning_rate": 2.364719155755779e-07, + "loss": 1.0281, + "step": 75735 + }, + { + "epoch": 5.6, + "learning_rate": 2.363856939830833e-07, + "loss": 0.9842, + "step": 75736 + }, + { + "epoch": 5.6, + "learning_rate": 2.3629948792435743e-07, + "loss": 0.9531, + "step": 75737 + }, + { + "epoch": 5.6, + "learning_rate": 2.3621329739954125e-07, + "loss": 0.9422, + "step": 75738 + }, + { + "epoch": 5.6, + "learning_rate": 2.3612712240877024e-07, + "loss": 0.9356, + "step": 75739 + }, + { + "epoch": 5.6, + "learning_rate": 2.3604096295218093e-07, + "loss": 0.9128, + "step": 75740 + }, + { + "epoch": 5.6, + "learning_rate": 2.35954819029911e-07, + "loss": 0.9527, + "step": 75741 + }, + { + "epoch": 5.6, + "learning_rate": 2.3586869064209594e-07, + "loss": 1.0281, + "step": 75742 + }, + { + "epoch": 5.6, + "learning_rate": 2.357825777888756e-07, + "loss": 1.0402, + "step": 75743 + }, + { + "epoch": 5.6, + "learning_rate": 2.356964804703843e-07, + "loss": 0.9794, + "step": 75744 + }, + { + "epoch": 5.6, + "learning_rate": 2.3561039868676083e-07, + "loss": 1.0061, + "step": 75745 + }, + { + "epoch": 5.6, + "learning_rate": 2.3552433243814065e-07, + "loss": 1.0795, + "step": 75746 + }, + { + "epoch": 5.6, + "learning_rate": 2.3543828172466367e-07, + "loss": 0.8864, + "step": 75747 + }, + { + "epoch": 5.6, + "learning_rate": 2.3535224654646195e-07, + "loss": 1.0475, + "step": 75748 + }, + { + "epoch": 5.6, + "learning_rate": 2.352662269036765e-07, + "loss": 0.9291, + "step": 75749 + }, + { + "epoch": 5.6, + "learning_rate": 2.3518022279644393e-07, + "loss": 1.1269, + "step": 75750 + }, + { + "epoch": 5.6, + "learning_rate": 2.3509423422489853e-07, + "loss": 0.9717, + "step": 75751 + }, + { + "epoch": 5.6, + "learning_rate": 2.3500826118917907e-07, + "loss": 0.9733, + "step": 75752 + }, + { + "epoch": 5.6, + "learning_rate": 2.3492230368941992e-07, + "loss": 1.0719, + "step": 75753 + }, + { + "epoch": 5.6, + "learning_rate": 2.348363617257621e-07, + "loss": 0.9681, + "step": 75754 + }, + { + "epoch": 5.6, + "learning_rate": 2.3475043529833874e-07, + "loss": 1.0253, + "step": 75755 + }, + { + "epoch": 5.6, + "learning_rate": 2.346645244072876e-07, + "loss": 0.9952, + "step": 75756 + }, + { + "epoch": 5.6, + "learning_rate": 2.345786290527452e-07, + "loss": 1.1267, + "step": 75757 + }, + { + "epoch": 5.6, + "learning_rate": 2.3449274923484923e-07, + "loss": 0.8841, + "step": 75758 + }, + { + "epoch": 5.6, + "learning_rate": 2.344068849537362e-07, + "loss": 1.1133, + "step": 75759 + }, + { + "epoch": 5.6, + "learning_rate": 2.343210362095405e-07, + "loss": 0.9786, + "step": 75760 + }, + { + "epoch": 5.6, + "learning_rate": 2.34235203002402e-07, + "loss": 1.0606, + "step": 75761 + }, + { + "epoch": 5.6, + "learning_rate": 2.341493853324539e-07, + "loss": 1.0388, + "step": 75762 + }, + { + "epoch": 5.6, + "learning_rate": 2.3406358319983502e-07, + "loss": 1.0749, + "step": 75763 + }, + { + "epoch": 5.6, + "learning_rate": 2.339777966046819e-07, + "loss": 0.9942, + "step": 75764 + }, + { + "epoch": 5.6, + "learning_rate": 2.3389202554712997e-07, + "loss": 0.9862, + "step": 75765 + }, + { + "epoch": 5.6, + "learning_rate": 2.3380627002731472e-07, + "loss": 0.9618, + "step": 75766 + }, + { + "epoch": 5.6, + "learning_rate": 2.3372053004537597e-07, + "loss": 0.9672, + "step": 75767 + }, + { + "epoch": 5.6, + "learning_rate": 2.3363480560144813e-07, + "loss": 0.9697, + "step": 75768 + }, + { + "epoch": 5.6, + "learning_rate": 2.335490966956666e-07, + "loss": 0.9712, + "step": 75769 + }, + { + "epoch": 5.6, + "learning_rate": 2.3346340332816908e-07, + "loss": 0.9134, + "step": 75770 + }, + { + "epoch": 5.6, + "learning_rate": 2.33377725499091e-07, + "loss": 1.0454, + "step": 75771 + }, + { + "epoch": 5.6, + "learning_rate": 2.3329206320857002e-07, + "loss": 1.0629, + "step": 75772 + }, + { + "epoch": 5.6, + "learning_rate": 2.3320641645674047e-07, + "loss": 0.9629, + "step": 75773 + }, + { + "epoch": 5.6, + "learning_rate": 2.3312078524374114e-07, + "loss": 1.0053, + "step": 75774 + }, + { + "epoch": 5.6, + "learning_rate": 2.330351695697064e-07, + "loss": 0.9866, + "step": 75775 + }, + { + "epoch": 5.6, + "learning_rate": 2.3294956943477275e-07, + "loss": 0.9759, + "step": 75776 + }, + { + "epoch": 5.6, + "learning_rate": 2.328639848390768e-07, + "loss": 0.993, + "step": 75777 + }, + { + "epoch": 5.6, + "learning_rate": 2.3277841578275396e-07, + "loss": 0.9264, + "step": 75778 + }, + { + "epoch": 5.6, + "learning_rate": 2.326928622659419e-07, + "loss": 1.065, + "step": 75779 + }, + { + "epoch": 5.6, + "learning_rate": 2.3260732428877497e-07, + "loss": 0.9041, + "step": 75780 + }, + { + "epoch": 5.6, + "learning_rate": 2.3252180185139084e-07, + "loss": 1.0399, + "step": 75781 + }, + { + "epoch": 5.6, + "learning_rate": 2.3243629495392383e-07, + "loss": 1.0978, + "step": 75782 + }, + { + "epoch": 5.6, + "learning_rate": 2.3235080359651164e-07, + "loss": 1.1482, + "step": 75783 + }, + { + "epoch": 5.6, + "learning_rate": 2.3226532777928857e-07, + "loss": 0.9688, + "step": 75784 + }, + { + "epoch": 5.6, + "learning_rate": 2.321798675023934e-07, + "loss": 0.8764, + "step": 75785 + }, + { + "epoch": 5.6, + "learning_rate": 2.3209442276595938e-07, + "loss": 0.9605, + "step": 75786 + }, + { + "epoch": 5.6, + "learning_rate": 2.3200899357012197e-07, + "loss": 0.8509, + "step": 75787 + }, + { + "epoch": 5.6, + "learning_rate": 2.3192357991502101e-07, + "loss": 1.0721, + "step": 75788 + }, + { + "epoch": 5.6, + "learning_rate": 2.3183818180078755e-07, + "loss": 0.9329, + "step": 75789 + }, + { + "epoch": 5.6, + "learning_rate": 2.3175279922756145e-07, + "loss": 1.0309, + "step": 75790 + }, + { + "epoch": 5.6, + "learning_rate": 2.3166743219547483e-07, + "loss": 0.9924, + "step": 75791 + }, + { + "epoch": 5.6, + "learning_rate": 2.315820807046687e-07, + "loss": 0.9248, + "step": 75792 + }, + { + "epoch": 5.6, + "learning_rate": 2.3149674475527294e-07, + "loss": 0.9986, + "step": 75793 + }, + { + "epoch": 5.6, + "learning_rate": 2.3141142434742746e-07, + "loss": 0.9973, + "step": 75794 + }, + { + "epoch": 5.6, + "learning_rate": 2.3132611948126548e-07, + "loss": 1.0232, + "step": 75795 + }, + { + "epoch": 5.6, + "learning_rate": 2.312408301569258e-07, + "loss": 1.005, + "step": 75796 + }, + { + "epoch": 5.6, + "learning_rate": 2.3115555637454157e-07, + "loss": 0.9541, + "step": 75797 + }, + { + "epoch": 5.6, + "learning_rate": 2.310702981342483e-07, + "loss": 0.9713, + "step": 75798 + }, + { + "epoch": 5.6, + "learning_rate": 2.3098505543618254e-07, + "loss": 0.9127, + "step": 75799 + }, + { + "epoch": 5.6, + "learning_rate": 2.3089982828048085e-07, + "loss": 1.0704, + "step": 75800 + }, + { + "epoch": 5.6, + "learning_rate": 2.3081461666727644e-07, + "loss": 1.0111, + "step": 75801 + }, + { + "epoch": 5.6, + "learning_rate": 2.307294205967059e-07, + "loss": 1.0018, + "step": 75802 + }, + { + "epoch": 5.6, + "learning_rate": 2.3064424006890574e-07, + "loss": 1.0, + "step": 75803 + }, + { + "epoch": 5.6, + "learning_rate": 2.3055907508401143e-07, + "loss": 0.8598, + "step": 75804 + }, + { + "epoch": 5.6, + "learning_rate": 2.3047392564215733e-07, + "loss": 1.034, + "step": 75805 + }, + { + "epoch": 5.6, + "learning_rate": 2.3038879174347994e-07, + "loss": 1.0242, + "step": 75806 + }, + { + "epoch": 5.6, + "learning_rate": 2.3030367338811254e-07, + "loss": 0.971, + "step": 75807 + }, + { + "epoch": 5.6, + "learning_rate": 2.302185705761939e-07, + "loss": 0.9263, + "step": 75808 + }, + { + "epoch": 5.6, + "learning_rate": 2.3013348330785724e-07, + "loss": 1.0335, + "step": 75809 + }, + { + "epoch": 5.6, + "learning_rate": 2.300484115832391e-07, + "loss": 0.9509, + "step": 75810 + }, + { + "epoch": 5.6, + "learning_rate": 2.2996335540247273e-07, + "loss": 1.0335, + "step": 75811 + }, + { + "epoch": 5.6, + "learning_rate": 2.2987831476569578e-07, + "loss": 0.9374, + "step": 75812 + }, + { + "epoch": 5.6, + "learning_rate": 2.297932896730426e-07, + "loss": 1.1688, + "step": 75813 + }, + { + "epoch": 5.6, + "learning_rate": 2.2970828012464862e-07, + "loss": 0.9301, + "step": 75814 + }, + { + "epoch": 5.6, + "learning_rate": 2.2962328612064933e-07, + "loss": 0.9178, + "step": 75815 + }, + { + "epoch": 5.6, + "learning_rate": 2.295383076611779e-07, + "loss": 0.8906, + "step": 75816 + }, + { + "epoch": 5.6, + "learning_rate": 2.2945334474637315e-07, + "loss": 1.0352, + "step": 75817 + }, + { + "epoch": 5.6, + "learning_rate": 2.2936839737636718e-07, + "loss": 0.9946, + "step": 75818 + }, + { + "epoch": 5.6, + "learning_rate": 2.2928346555129655e-07, + "loss": 1.0442, + "step": 75819 + }, + { + "epoch": 5.6, + "learning_rate": 2.291985492712967e-07, + "loss": 0.9575, + "step": 75820 + }, + { + "epoch": 5.6, + "learning_rate": 2.2911364853650198e-07, + "loss": 0.9353, + "step": 75821 + }, + { + "epoch": 5.6, + "learning_rate": 2.290287633470467e-07, + "loss": 0.9913, + "step": 75822 + }, + { + "epoch": 5.6, + "learning_rate": 2.2894389370306747e-07, + "loss": 1.0506, + "step": 75823 + }, + { + "epoch": 5.6, + "learning_rate": 2.2885903960469858e-07, + "loss": 0.9875, + "step": 75824 + }, + { + "epoch": 5.6, + "learning_rate": 2.2877420105207548e-07, + "loss": 1.0651, + "step": 75825 + }, + { + "epoch": 5.6, + "learning_rate": 2.2868937804533364e-07, + "loss": 1.0439, + "step": 75826 + }, + { + "epoch": 5.6, + "learning_rate": 2.2860457058460518e-07, + "loss": 1.0691, + "step": 75827 + }, + { + "epoch": 5.6, + "learning_rate": 2.2851977867002773e-07, + "loss": 0.9703, + "step": 75828 + }, + { + "epoch": 5.6, + "learning_rate": 2.2843500230173677e-07, + "loss": 1.0266, + "step": 75829 + }, + { + "epoch": 5.6, + "learning_rate": 2.283502414798644e-07, + "loss": 1.0584, + "step": 75830 + }, + { + "epoch": 5.6, + "learning_rate": 2.282654962045472e-07, + "loss": 0.9523, + "step": 75831 + }, + { + "epoch": 5.6, + "learning_rate": 2.2818076647592057e-07, + "loss": 1.0586, + "step": 75832 + }, + { + "epoch": 5.6, + "learning_rate": 2.280960522941189e-07, + "loss": 1.0235, + "step": 75833 + }, + { + "epoch": 5.6, + "learning_rate": 2.2801135365927541e-07, + "loss": 0.957, + "step": 75834 + }, + { + "epoch": 5.6, + "learning_rate": 2.279266705715266e-07, + "loss": 1.0035, + "step": 75835 + }, + { + "epoch": 5.6, + "learning_rate": 2.278420030310058e-07, + "loss": 0.9697, + "step": 75836 + }, + { + "epoch": 5.6, + "learning_rate": 2.2775735103785056e-07, + "loss": 0.9605, + "step": 75837 + }, + { + "epoch": 5.6, + "learning_rate": 2.2767271459219088e-07, + "loss": 1.059, + "step": 75838 + }, + { + "epoch": 5.6, + "learning_rate": 2.2758809369416545e-07, + "loss": 1.0699, + "step": 75839 + }, + { + "epoch": 5.6, + "learning_rate": 2.2750348834390757e-07, + "loss": 1.0132, + "step": 75840 + }, + { + "epoch": 5.6, + "learning_rate": 2.2741889854155264e-07, + "loss": 0.9796, + "step": 75841 + }, + { + "epoch": 5.6, + "learning_rate": 2.273343242872339e-07, + "loss": 1.0312, + "step": 75842 + }, + { + "epoch": 5.6, + "learning_rate": 2.272497655810857e-07, + "loss": 1.0484, + "step": 75843 + }, + { + "epoch": 5.6, + "learning_rate": 2.2716522242324346e-07, + "loss": 0.9532, + "step": 75844 + }, + { + "epoch": 5.6, + "learning_rate": 2.2708069481384153e-07, + "loss": 0.9586, + "step": 75845 + }, + { + "epoch": 5.6, + "learning_rate": 2.269961827530154e-07, + "loss": 1.0464, + "step": 75846 + }, + { + "epoch": 5.6, + "learning_rate": 2.269116862408971e-07, + "loss": 0.8407, + "step": 75847 + }, + { + "epoch": 5.6, + "learning_rate": 2.2682720527762325e-07, + "loss": 1.0034, + "step": 75848 + }, + { + "epoch": 5.6, + "learning_rate": 2.2674273986332707e-07, + "loss": 0.9839, + "step": 75849 + }, + { + "epoch": 5.6, + "learning_rate": 2.2665828999814398e-07, + "loss": 1.0645, + "step": 75850 + }, + { + "epoch": 5.6, + "learning_rate": 2.2657385568220725e-07, + "loss": 1.1356, + "step": 75851 + }, + { + "epoch": 5.6, + "learning_rate": 2.2648943691565117e-07, + "loss": 0.9953, + "step": 75852 + }, + { + "epoch": 5.6, + "learning_rate": 2.264050336986112e-07, + "loss": 0.9858, + "step": 75853 + }, + { + "epoch": 5.6, + "learning_rate": 2.2632064603122173e-07, + "loss": 1.0309, + "step": 75854 + }, + { + "epoch": 5.6, + "learning_rate": 2.262362739136159e-07, + "loss": 0.9663, + "step": 75855 + }, + { + "epoch": 5.6, + "learning_rate": 2.26151917345927e-07, + "loss": 1.0178, + "step": 75856 + }, + { + "epoch": 5.61, + "learning_rate": 2.2606757632829158e-07, + "loss": 0.8959, + "step": 75857 + }, + { + "epoch": 5.61, + "learning_rate": 2.2598325086084284e-07, + "loss": 1.0433, + "step": 75858 + }, + { + "epoch": 5.61, + "learning_rate": 2.258989409437151e-07, + "loss": 0.9982, + "step": 75859 + }, + { + "epoch": 5.61, + "learning_rate": 2.2581464657704167e-07, + "loss": 1.0149, + "step": 75860 + }, + { + "epoch": 5.61, + "learning_rate": 2.2573036776095791e-07, + "loss": 1.042, + "step": 75861 + }, + { + "epoch": 5.61, + "learning_rate": 2.256461044955982e-07, + "loss": 1.0761, + "step": 75862 + }, + { + "epoch": 5.61, + "learning_rate": 2.2556185678109355e-07, + "loss": 1.0193, + "step": 75863 + }, + { + "epoch": 5.61, + "learning_rate": 2.254776246175805e-07, + "loss": 1.0064, + "step": 75864 + }, + { + "epoch": 5.61, + "learning_rate": 2.253934080051945e-07, + "loss": 0.9937, + "step": 75865 + }, + { + "epoch": 5.61, + "learning_rate": 2.2530920694406656e-07, + "loss": 0.997, + "step": 75866 + }, + { + "epoch": 5.61, + "learning_rate": 2.25225021434331e-07, + "loss": 0.8975, + "step": 75867 + }, + { + "epoch": 5.61, + "learning_rate": 2.2514085147612331e-07, + "loss": 0.8362, + "step": 75868 + }, + { + "epoch": 5.61, + "learning_rate": 2.2505669706957668e-07, + "loss": 0.976, + "step": 75869 + }, + { + "epoch": 5.61, + "learning_rate": 2.2497255821482545e-07, + "loss": 0.9898, + "step": 75870 + }, + { + "epoch": 5.61, + "learning_rate": 2.2488843491200284e-07, + "loss": 0.9556, + "step": 75871 + }, + { + "epoch": 5.61, + "learning_rate": 2.2480432716124212e-07, + "loss": 1.097, + "step": 75872 + }, + { + "epoch": 5.61, + "learning_rate": 2.2472023496267758e-07, + "loss": 1.0184, + "step": 75873 + }, + { + "epoch": 5.61, + "learning_rate": 2.246361583164447e-07, + "loss": 0.9459, + "step": 75874 + }, + { + "epoch": 5.61, + "learning_rate": 2.2455209722267556e-07, + "loss": 0.962, + "step": 75875 + }, + { + "epoch": 5.61, + "learning_rate": 2.2446805168150343e-07, + "loss": 0.9022, + "step": 75876 + }, + { + "epoch": 5.61, + "learning_rate": 2.243840216930626e-07, + "loss": 0.899, + "step": 75877 + }, + { + "epoch": 5.61, + "learning_rate": 2.2430000725748747e-07, + "loss": 0.9962, + "step": 75878 + }, + { + "epoch": 5.61, + "learning_rate": 2.2421600837491119e-07, + "loss": 0.9567, + "step": 75879 + }, + { + "epoch": 5.61, + "learning_rate": 2.2413202504546817e-07, + "loss": 1.0356, + "step": 75880 + }, + { + "epoch": 5.61, + "learning_rate": 2.2404805726928937e-07, + "loss": 1.108, + "step": 75881 + }, + { + "epoch": 5.61, + "learning_rate": 2.2396410504651245e-07, + "loss": 0.8749, + "step": 75882 + }, + { + "epoch": 5.61, + "learning_rate": 2.2388016837726733e-07, + "loss": 1.0525, + "step": 75883 + }, + { + "epoch": 5.61, + "learning_rate": 2.2379624726168948e-07, + "loss": 0.9682, + "step": 75884 + }, + { + "epoch": 5.61, + "learning_rate": 2.2371234169991095e-07, + "loss": 0.9796, + "step": 75885 + }, + { + "epoch": 5.61, + "learning_rate": 2.2362845169206725e-07, + "loss": 0.9278, + "step": 75886 + }, + { + "epoch": 5.61, + "learning_rate": 2.235445772382905e-07, + "loss": 0.9925, + "step": 75887 + }, + { + "epoch": 5.61, + "learning_rate": 2.2346071833871387e-07, + "loss": 1.0513, + "step": 75888 + }, + { + "epoch": 5.61, + "learning_rate": 2.2337687499347282e-07, + "loss": 0.9221, + "step": 75889 + }, + { + "epoch": 5.61, + "learning_rate": 2.232930472026984e-07, + "loss": 0.9478, + "step": 75890 + }, + { + "epoch": 5.61, + "learning_rate": 2.2320923496652492e-07, + "loss": 1.0098, + "step": 75891 + }, + { + "epoch": 5.61, + "learning_rate": 2.2312543828508448e-07, + "loss": 0.9799, + "step": 75892 + }, + { + "epoch": 5.61, + "learning_rate": 2.2304165715851366e-07, + "loss": 1.0425, + "step": 75893 + }, + { + "epoch": 5.61, + "learning_rate": 2.2295789158694235e-07, + "loss": 0.9721, + "step": 75894 + }, + { + "epoch": 5.61, + "learning_rate": 2.2287414157050601e-07, + "loss": 1.0021, + "step": 75895 + }, + { + "epoch": 5.61, + "learning_rate": 2.2279040710933564e-07, + "loss": 0.9732, + "step": 75896 + }, + { + "epoch": 5.61, + "learning_rate": 2.2270668820356668e-07, + "loss": 1.0133, + "step": 75897 + }, + { + "epoch": 5.61, + "learning_rate": 2.2262298485333235e-07, + "loss": 1.0134, + "step": 75898 + }, + { + "epoch": 5.61, + "learning_rate": 2.2253929705876365e-07, + "loss": 1.0361, + "step": 75899 + }, + { + "epoch": 5.61, + "learning_rate": 2.2245562481999606e-07, + "loss": 0.9318, + "step": 75900 + }, + { + "epoch": 5.61, + "learning_rate": 2.2237196813716057e-07, + "loss": 1.0145, + "step": 75901 + }, + { + "epoch": 5.61, + "learning_rate": 2.2228832701039149e-07, + "loss": 0.9734, + "step": 75902 + }, + { + "epoch": 5.61, + "learning_rate": 2.222047014398232e-07, + "loss": 0.9104, + "step": 75903 + }, + { + "epoch": 5.61, + "learning_rate": 2.221210914255867e-07, + "loss": 0.9982, + "step": 75904 + }, + { + "epoch": 5.61, + "learning_rate": 2.2203749696781407e-07, + "loss": 0.9432, + "step": 75905 + }, + { + "epoch": 5.61, + "learning_rate": 2.219539180666419e-07, + "loss": 1.0033, + "step": 75906 + }, + { + "epoch": 5.61, + "learning_rate": 2.218703547222001e-07, + "loss": 1.0104, + "step": 75907 + }, + { + "epoch": 5.61, + "learning_rate": 2.2178680693462297e-07, + "loss": 0.9289, + "step": 75908 + }, + { + "epoch": 5.61, + "learning_rate": 2.2170327470404373e-07, + "loss": 1.0882, + "step": 75909 + }, + { + "epoch": 5.61, + "learning_rate": 2.2161975803059456e-07, + "loss": 1.0614, + "step": 75910 + }, + { + "epoch": 5.61, + "learning_rate": 2.215362569144075e-07, + "loss": 1.0924, + "step": 75911 + }, + { + "epoch": 5.61, + "learning_rate": 2.2145277135561583e-07, + "loss": 0.9613, + "step": 75912 + }, + { + "epoch": 5.61, + "learning_rate": 2.2136930135435497e-07, + "loss": 0.9581, + "step": 75913 + }, + { + "epoch": 5.61, + "learning_rate": 2.2128584691075482e-07, + "loss": 0.8659, + "step": 75914 + }, + { + "epoch": 5.61, + "learning_rate": 2.2120240802494864e-07, + "loss": 1.0515, + "step": 75915 + }, + { + "epoch": 5.61, + "learning_rate": 2.211189846970685e-07, + "loss": 0.9912, + "step": 75916 + }, + { + "epoch": 5.61, + "learning_rate": 2.2103557692724987e-07, + "loss": 0.9553, + "step": 75917 + }, + { + "epoch": 5.61, + "learning_rate": 2.2095218471562263e-07, + "loss": 1.0308, + "step": 75918 + }, + { + "epoch": 5.61, + "learning_rate": 2.2086880806232115e-07, + "loss": 0.9102, + "step": 75919 + }, + { + "epoch": 5.61, + "learning_rate": 2.207854469674775e-07, + "loss": 0.9607, + "step": 75920 + }, + { + "epoch": 5.61, + "learning_rate": 2.2070210143122385e-07, + "loss": 1.005, + "step": 75921 + }, + { + "epoch": 5.61, + "learning_rate": 2.2061877145369336e-07, + "loss": 0.9958, + "step": 75922 + }, + { + "epoch": 5.61, + "learning_rate": 2.2053545703501933e-07, + "loss": 0.9731, + "step": 75923 + }, + { + "epoch": 5.61, + "learning_rate": 2.204521581753327e-07, + "loss": 1.0091, + "step": 75924 + }, + { + "epoch": 5.61, + "learning_rate": 2.2036887487476565e-07, + "loss": 1.0161, + "step": 75925 + }, + { + "epoch": 5.61, + "learning_rate": 2.2028560713345358e-07, + "loss": 0.9914, + "step": 75926 + }, + { + "epoch": 5.61, + "learning_rate": 2.2020235495152753e-07, + "loss": 0.9705, + "step": 75927 + }, + { + "epoch": 5.61, + "learning_rate": 2.2011911832911848e-07, + "loss": 0.9325, + "step": 75928 + }, + { + "epoch": 5.61, + "learning_rate": 2.2003589726635964e-07, + "loss": 0.9505, + "step": 75929 + }, + { + "epoch": 5.61, + "learning_rate": 2.1995269176338318e-07, + "loss": 1.0417, + "step": 75930 + }, + { + "epoch": 5.61, + "learning_rate": 2.1986950182032453e-07, + "loss": 0.9942, + "step": 75931 + }, + { + "epoch": 5.61, + "learning_rate": 2.197863274373113e-07, + "loss": 1.0126, + "step": 75932 + }, + { + "epoch": 5.61, + "learning_rate": 2.1970316861448015e-07, + "loss": 1.0088, + "step": 75933 + }, + { + "epoch": 5.61, + "learning_rate": 2.196200253519587e-07, + "loss": 0.9668, + "step": 75934 + }, + { + "epoch": 5.61, + "learning_rate": 2.195368976498846e-07, + "loss": 1.1065, + "step": 75935 + }, + { + "epoch": 5.61, + "learning_rate": 2.194537855083867e-07, + "loss": 0.9821, + "step": 75936 + }, + { + "epoch": 5.61, + "learning_rate": 2.1937068892759594e-07, + "loss": 0.9659, + "step": 75937 + }, + { + "epoch": 5.61, + "learning_rate": 2.192876079076478e-07, + "loss": 0.9606, + "step": 75938 + }, + { + "epoch": 5.61, + "learning_rate": 2.192045424486733e-07, + "loss": 1.0976, + "step": 75939 + }, + { + "epoch": 5.61, + "learning_rate": 2.1912149255080451e-07, + "loss": 1.0489, + "step": 75940 + }, + { + "epoch": 5.61, + "learning_rate": 2.1903845821417247e-07, + "loss": 1.0957, + "step": 75941 + }, + { + "epoch": 5.61, + "learning_rate": 2.189554394389115e-07, + "loss": 0.9386, + "step": 75942 + }, + { + "epoch": 5.61, + "learning_rate": 2.1887243622515263e-07, + "loss": 0.9773, + "step": 75943 + }, + { + "epoch": 5.61, + "learning_rate": 2.1878944857302687e-07, + "loss": 1.0029, + "step": 75944 + }, + { + "epoch": 5.61, + "learning_rate": 2.187064764826674e-07, + "loss": 0.9816, + "step": 75945 + }, + { + "epoch": 5.61, + "learning_rate": 2.186235199542064e-07, + "loss": 1.042, + "step": 75946 + }, + { + "epoch": 5.61, + "learning_rate": 2.185405789877748e-07, + "loss": 0.9192, + "step": 75947 + }, + { + "epoch": 5.61, + "learning_rate": 2.184576535835059e-07, + "loss": 0.9556, + "step": 75948 + }, + { + "epoch": 5.61, + "learning_rate": 2.1837474374153067e-07, + "loss": 1.0828, + "step": 75949 + }, + { + "epoch": 5.61, + "learning_rate": 2.1829184946198012e-07, + "loss": 1.0098, + "step": 75950 + }, + { + "epoch": 5.61, + "learning_rate": 2.182089707449886e-07, + "loss": 0.9904, + "step": 75951 + }, + { + "epoch": 5.61, + "learning_rate": 2.1812610759068596e-07, + "loss": 1.0073, + "step": 75952 + }, + { + "epoch": 5.61, + "learning_rate": 2.1804325999920552e-07, + "loss": 0.9149, + "step": 75953 + }, + { + "epoch": 5.61, + "learning_rate": 2.1796042797067706e-07, + "loss": 1.0695, + "step": 75954 + }, + { + "epoch": 5.61, + "learning_rate": 2.1787761150523613e-07, + "loss": 0.9416, + "step": 75955 + }, + { + "epoch": 5.61, + "learning_rate": 2.1779481060301032e-07, + "loss": 0.94, + "step": 75956 + }, + { + "epoch": 5.61, + "learning_rate": 2.1771202526413181e-07, + "loss": 0.9238, + "step": 75957 + }, + { + "epoch": 5.61, + "learning_rate": 2.1762925548873493e-07, + "loss": 1.0316, + "step": 75958 + }, + { + "epoch": 5.61, + "learning_rate": 2.1754650127694954e-07, + "loss": 1.005, + "step": 75959 + }, + { + "epoch": 5.61, + "learning_rate": 2.1746376262890778e-07, + "loss": 1.0439, + "step": 75960 + }, + { + "epoch": 5.61, + "learning_rate": 2.1738103954474067e-07, + "loss": 0.9946, + "step": 75961 + }, + { + "epoch": 5.61, + "learning_rate": 2.172983320245814e-07, + "loss": 0.9599, + "step": 75962 + }, + { + "epoch": 5.61, + "learning_rate": 2.1721564006855987e-07, + "loss": 0.9599, + "step": 75963 + }, + { + "epoch": 5.61, + "learning_rate": 2.1713296367680936e-07, + "loss": 0.9248, + "step": 75964 + }, + { + "epoch": 5.61, + "learning_rate": 2.170503028494597e-07, + "loss": 1.0987, + "step": 75965 + }, + { + "epoch": 5.61, + "learning_rate": 2.1696765758664196e-07, + "loss": 0.9815, + "step": 75966 + }, + { + "epoch": 5.61, + "learning_rate": 2.168850278884893e-07, + "loss": 1.0221, + "step": 75967 + }, + { + "epoch": 5.61, + "learning_rate": 2.1680241375513278e-07, + "loss": 1.0377, + "step": 75968 + }, + { + "epoch": 5.61, + "learning_rate": 2.1671981518670447e-07, + "loss": 0.9326, + "step": 75969 + }, + { + "epoch": 5.61, + "learning_rate": 2.1663723218333322e-07, + "loss": 0.9452, + "step": 75970 + }, + { + "epoch": 5.61, + "learning_rate": 2.165546647451533e-07, + "loss": 1.0302, + "step": 75971 + }, + { + "epoch": 5.61, + "learning_rate": 2.1647211287229464e-07, + "loss": 1.0016, + "step": 75972 + }, + { + "epoch": 5.61, + "learning_rate": 2.1638957656488936e-07, + "loss": 0.9393, + "step": 75973 + }, + { + "epoch": 5.61, + "learning_rate": 2.1630705582306733e-07, + "loss": 1.059, + "step": 75974 + }, + { + "epoch": 5.61, + "learning_rate": 2.162245506469607e-07, + "loss": 1.0218, + "step": 75975 + }, + { + "epoch": 5.61, + "learning_rate": 2.1614206103670267e-07, + "loss": 0.9182, + "step": 75976 + }, + { + "epoch": 5.61, + "learning_rate": 2.160595869924209e-07, + "loss": 0.9794, + "step": 75977 + }, + { + "epoch": 5.61, + "learning_rate": 2.1597712851424867e-07, + "loss": 0.971, + "step": 75978 + }, + { + "epoch": 5.61, + "learning_rate": 2.1589468560231696e-07, + "loss": 0.9065, + "step": 75979 + }, + { + "epoch": 5.61, + "learning_rate": 2.1581225825675788e-07, + "loss": 1.0567, + "step": 75980 + }, + { + "epoch": 5.61, + "learning_rate": 2.1572984647769912e-07, + "loss": 1.0485, + "step": 75981 + }, + { + "epoch": 5.61, + "learning_rate": 2.1564745026527612e-07, + "loss": 1.0165, + "step": 75982 + }, + { + "epoch": 5.61, + "learning_rate": 2.1556506961961765e-07, + "loss": 0.9579, + "step": 75983 + }, + { + "epoch": 5.61, + "learning_rate": 2.1548270454085585e-07, + "loss": 1.0894, + "step": 75984 + }, + { + "epoch": 5.61, + "learning_rate": 2.1540035502911948e-07, + "loss": 0.9728, + "step": 75985 + }, + { + "epoch": 5.61, + "learning_rate": 2.153180210845418e-07, + "loss": 0.8859, + "step": 75986 + }, + { + "epoch": 5.61, + "learning_rate": 2.152357027072527e-07, + "loss": 1.0765, + "step": 75987 + }, + { + "epoch": 5.61, + "learning_rate": 2.1515339989738536e-07, + "loss": 0.9777, + "step": 75988 + }, + { + "epoch": 5.61, + "learning_rate": 2.1507111265506753e-07, + "loss": 1.0879, + "step": 75989 + }, + { + "epoch": 5.61, + "learning_rate": 2.1498884098043126e-07, + "loss": 1.0437, + "step": 75990 + }, + { + "epoch": 5.61, + "learning_rate": 2.1490658487360872e-07, + "loss": 1.0194, + "step": 75991 + }, + { + "epoch": 5.61, + "learning_rate": 2.1482434433472865e-07, + "loss": 0.9333, + "step": 75992 + }, + { + "epoch": 5.62, + "learning_rate": 2.147421193639243e-07, + "loss": 1.0025, + "step": 75993 + }, + { + "epoch": 5.62, + "learning_rate": 2.1465990996132446e-07, + "loss": 1.027, + "step": 75994 + }, + { + "epoch": 5.62, + "learning_rate": 2.1457771612706014e-07, + "loss": 1.0406, + "step": 75995 + }, + { + "epoch": 5.62, + "learning_rate": 2.1449553786126454e-07, + "loss": 1.0074, + "step": 75996 + }, + { + "epoch": 5.62, + "learning_rate": 2.1441337516406534e-07, + "loss": 0.9312, + "step": 75997 + }, + { + "epoch": 5.62, + "learning_rate": 2.143312280355947e-07, + "loss": 0.9131, + "step": 75998 + }, + { + "epoch": 5.62, + "learning_rate": 2.1424909647598247e-07, + "loss": 0.9526, + "step": 75999 + }, + { + "epoch": 5.62, + "learning_rate": 2.1416698048536078e-07, + "loss": 0.9534, + "step": 76000 + }, + { + "epoch": 5.62, + "learning_rate": 2.1408488006385952e-07, + "loss": 1.0542, + "step": 76001 + }, + { + "epoch": 5.62, + "learning_rate": 2.140027952116075e-07, + "loss": 1.0448, + "step": 76002 + }, + { + "epoch": 5.62, + "learning_rate": 2.1392072592873902e-07, + "loss": 0.9493, + "step": 76003 + }, + { + "epoch": 5.62, + "learning_rate": 2.138386722153818e-07, + "loss": 0.9694, + "step": 76004 + }, + { + "epoch": 5.62, + "learning_rate": 2.137566340716679e-07, + "loss": 0.9907, + "step": 76005 + }, + { + "epoch": 5.62, + "learning_rate": 2.1367461149772506e-07, + "loss": 1.0129, + "step": 76006 + }, + { + "epoch": 5.62, + "learning_rate": 2.1359260449368756e-07, + "loss": 1.0215, + "step": 76007 + }, + { + "epoch": 5.62, + "learning_rate": 2.1351061305968423e-07, + "loss": 1.0308, + "step": 76008 + }, + { + "epoch": 5.62, + "learning_rate": 2.1342863719584607e-07, + "loss": 0.995, + "step": 76009 + }, + { + "epoch": 5.62, + "learning_rate": 2.133466769023007e-07, + "loss": 0.9147, + "step": 76010 + }, + { + "epoch": 5.62, + "learning_rate": 2.1326473217918252e-07, + "loss": 0.9979, + "step": 76011 + }, + { + "epoch": 5.62, + "learning_rate": 2.1318280302662032e-07, + "loss": 1.029, + "step": 76012 + }, + { + "epoch": 5.62, + "learning_rate": 2.1310088944474393e-07, + "loss": 0.9138, + "step": 76013 + }, + { + "epoch": 5.62, + "learning_rate": 2.130189914336833e-07, + "loss": 1.0681, + "step": 76014 + }, + { + "epoch": 5.62, + "learning_rate": 2.1293710899356945e-07, + "loss": 1.0369, + "step": 76015 + }, + { + "epoch": 5.62, + "learning_rate": 2.1285524212453334e-07, + "loss": 1.0758, + "step": 76016 + }, + { + "epoch": 5.62, + "learning_rate": 2.127733908267049e-07, + "loss": 0.934, + "step": 76017 + }, + { + "epoch": 5.62, + "learning_rate": 2.1269155510021288e-07, + "loss": 0.9802, + "step": 76018 + }, + { + "epoch": 5.62, + "learning_rate": 2.1260973494518833e-07, + "loss": 0.8935, + "step": 76019 + }, + { + "epoch": 5.62, + "learning_rate": 2.1252793036176334e-07, + "loss": 0.9686, + "step": 76020 + }, + { + "epoch": 5.62, + "learning_rate": 2.124461413500656e-07, + "loss": 0.9349, + "step": 76021 + }, + { + "epoch": 5.62, + "learning_rate": 2.1236436791022496e-07, + "loss": 0.9599, + "step": 76022 + }, + { + "epoch": 5.62, + "learning_rate": 2.122826100423736e-07, + "loss": 0.956, + "step": 76023 + }, + { + "epoch": 5.62, + "learning_rate": 2.1220086774664028e-07, + "loss": 0.9862, + "step": 76024 + }, + { + "epoch": 5.62, + "learning_rate": 2.12119141023156e-07, + "loss": 1.1074, + "step": 76025 + }, + { + "epoch": 5.62, + "learning_rate": 2.1203742987204956e-07, + "loss": 1.0541, + "step": 76026 + }, + { + "epoch": 5.62, + "learning_rate": 2.1195573429345084e-07, + "loss": 0.9566, + "step": 76027 + }, + { + "epoch": 5.62, + "learning_rate": 2.1187405428749086e-07, + "loss": 0.9621, + "step": 76028 + }, + { + "epoch": 5.62, + "learning_rate": 2.1179238985430063e-07, + "loss": 1.1056, + "step": 76029 + }, + { + "epoch": 5.62, + "learning_rate": 2.117107409940078e-07, + "loss": 0.9618, + "step": 76030 + }, + { + "epoch": 5.62, + "learning_rate": 2.1162910770674228e-07, + "loss": 0.955, + "step": 76031 + }, + { + "epoch": 5.62, + "learning_rate": 2.115474899926351e-07, + "loss": 1.0087, + "step": 76032 + }, + { + "epoch": 5.62, + "learning_rate": 2.114658878518172e-07, + "loss": 1.0707, + "step": 76033 + }, + { + "epoch": 5.62, + "learning_rate": 2.1138430128441523e-07, + "loss": 1.0203, + "step": 76034 + }, + { + "epoch": 5.62, + "learning_rate": 2.1130273029056126e-07, + "loss": 1.0227, + "step": 76035 + }, + { + "epoch": 5.62, + "learning_rate": 2.1122117487038517e-07, + "loss": 1.0314, + "step": 76036 + }, + { + "epoch": 5.62, + "learning_rate": 2.111396350240158e-07, + "loss": 1.0806, + "step": 76037 + }, + { + "epoch": 5.62, + "learning_rate": 2.11058110751583e-07, + "loss": 1.0437, + "step": 76038 + }, + { + "epoch": 5.62, + "learning_rate": 2.109766020532178e-07, + "loss": 1.0912, + "step": 76039 + }, + { + "epoch": 5.62, + "learning_rate": 2.1089510892904673e-07, + "loss": 0.895, + "step": 76040 + }, + { + "epoch": 5.62, + "learning_rate": 2.108136313792042e-07, + "loss": 1.0593, + "step": 76041 + }, + { + "epoch": 5.62, + "learning_rate": 2.1073216940381448e-07, + "loss": 0.9703, + "step": 76042 + }, + { + "epoch": 5.62, + "learning_rate": 2.106507230030108e-07, + "loss": 0.9671, + "step": 76043 + }, + { + "epoch": 5.62, + "learning_rate": 2.105692921769209e-07, + "loss": 1.002, + "step": 76044 + }, + { + "epoch": 5.62, + "learning_rate": 2.1048787692567797e-07, + "loss": 1.023, + "step": 76045 + }, + { + "epoch": 5.62, + "learning_rate": 2.1040647724940633e-07, + "loss": 0.9977, + "step": 76046 + }, + { + "epoch": 5.62, + "learning_rate": 2.1032509314823812e-07, + "loss": 0.9733, + "step": 76047 + }, + { + "epoch": 5.62, + "learning_rate": 2.1024372462230324e-07, + "loss": 0.9228, + "step": 76048 + }, + { + "epoch": 5.62, + "learning_rate": 2.1016237167173047e-07, + "loss": 0.9526, + "step": 76049 + }, + { + "epoch": 5.62, + "learning_rate": 2.100810342966497e-07, + "loss": 1.0018, + "step": 76050 + }, + { + "epoch": 5.62, + "learning_rate": 2.0999971249718864e-07, + "loss": 1.0334, + "step": 76051 + }, + { + "epoch": 5.62, + "learning_rate": 2.0991840627347825e-07, + "loss": 1.103, + "step": 76052 + }, + { + "epoch": 5.62, + "learning_rate": 2.0983711562564847e-07, + "loss": 1.0038, + "step": 76053 + }, + { + "epoch": 5.62, + "learning_rate": 2.0975584055382804e-07, + "loss": 0.9485, + "step": 76054 + }, + { + "epoch": 5.62, + "learning_rate": 2.0967458105814465e-07, + "loss": 0.9583, + "step": 76055 + }, + { + "epoch": 5.62, + "learning_rate": 2.095933371387293e-07, + "loss": 0.9986, + "step": 76056 + }, + { + "epoch": 5.62, + "learning_rate": 2.0951210879571193e-07, + "loss": 1.0105, + "step": 76057 + }, + { + "epoch": 5.62, + "learning_rate": 2.0943089602922019e-07, + "loss": 0.9919, + "step": 76058 + }, + { + "epoch": 5.62, + "learning_rate": 2.0934969883938283e-07, + "loss": 1.0747, + "step": 76059 + }, + { + "epoch": 5.62, + "learning_rate": 2.092685172263309e-07, + "loss": 1.0181, + "step": 76060 + }, + { + "epoch": 5.62, + "learning_rate": 2.0918735119019207e-07, + "loss": 0.9656, + "step": 76061 + }, + { + "epoch": 5.62, + "learning_rate": 2.0910620073109733e-07, + "loss": 1.0079, + "step": 76062 + }, + { + "epoch": 5.62, + "learning_rate": 2.0902506584917325e-07, + "loss": 0.8635, + "step": 76063 + }, + { + "epoch": 5.62, + "learning_rate": 2.0894394654454974e-07, + "loss": 1.0359, + "step": 76064 + }, + { + "epoch": 5.62, + "learning_rate": 2.088628428173578e-07, + "loss": 0.926, + "step": 76065 + }, + { + "epoch": 5.62, + "learning_rate": 2.087817546677251e-07, + "loss": 0.9889, + "step": 76066 + }, + { + "epoch": 5.62, + "learning_rate": 2.087006820957793e-07, + "loss": 1.0209, + "step": 76067 + }, + { + "epoch": 5.62, + "learning_rate": 2.0861962510165146e-07, + "loss": 1.0286, + "step": 76068 + }, + { + "epoch": 5.62, + "learning_rate": 2.085385836854692e-07, + "loss": 0.9182, + "step": 76069 + }, + { + "epoch": 5.62, + "learning_rate": 2.0845755784736354e-07, + "loss": 1.0009, + "step": 76070 + }, + { + "epoch": 5.62, + "learning_rate": 2.0837654758745995e-07, + "loss": 0.9278, + "step": 76071 + }, + { + "epoch": 5.62, + "learning_rate": 2.0829555290588942e-07, + "loss": 1.0846, + "step": 76072 + }, + { + "epoch": 5.62, + "learning_rate": 2.0821457380277966e-07, + "loss": 0.9527, + "step": 76073 + }, + { + "epoch": 5.62, + "learning_rate": 2.0813361027826273e-07, + "loss": 0.9077, + "step": 76074 + }, + { + "epoch": 5.62, + "learning_rate": 2.0805266233246412e-07, + "loss": 0.9755, + "step": 76075 + }, + { + "epoch": 5.62, + "learning_rate": 2.0797172996551375e-07, + "loss": 1.0429, + "step": 76076 + }, + { + "epoch": 5.62, + "learning_rate": 2.0789081317754033e-07, + "loss": 0.9537, + "step": 76077 + }, + { + "epoch": 5.62, + "learning_rate": 2.0780991196867162e-07, + "loss": 1.037, + "step": 76078 + }, + { + "epoch": 5.62, + "learning_rate": 2.0772902633903858e-07, + "loss": 0.9643, + "step": 76079 + }, + { + "epoch": 5.62, + "learning_rate": 2.0764815628876668e-07, + "loss": 0.9897, + "step": 76080 + }, + { + "epoch": 5.62, + "learning_rate": 2.07567301817988e-07, + "loss": 1.0154, + "step": 76081 + }, + { + "epoch": 5.62, + "learning_rate": 2.0748646292682916e-07, + "loss": 0.9433, + "step": 76082 + }, + { + "epoch": 5.62, + "learning_rate": 2.0740563961542003e-07, + "loss": 0.9652, + "step": 76083 + }, + { + "epoch": 5.62, + "learning_rate": 2.073248318838872e-07, + "loss": 1.0289, + "step": 76084 + }, + { + "epoch": 5.62, + "learning_rate": 2.0724403973236052e-07, + "loss": 0.9934, + "step": 76085 + }, + { + "epoch": 5.62, + "learning_rate": 2.0716326316096992e-07, + "loss": 1.086, + "step": 76086 + }, + { + "epoch": 5.62, + "learning_rate": 2.0708250216984082e-07, + "loss": 1.0428, + "step": 76087 + }, + { + "epoch": 5.62, + "learning_rate": 2.070017567591043e-07, + "loss": 1.1881, + "step": 76088 + }, + { + "epoch": 5.62, + "learning_rate": 2.0692102692888682e-07, + "loss": 0.976, + "step": 76089 + }, + { + "epoch": 5.62, + "learning_rate": 2.0684031267931948e-07, + "loss": 0.961, + "step": 76090 + }, + { + "epoch": 5.62, + "learning_rate": 2.0675961401052768e-07, + "loss": 1.0405, + "step": 76091 + }, + { + "epoch": 5.62, + "learning_rate": 2.0667893092264247e-07, + "loss": 1.0949, + "step": 76092 + }, + { + "epoch": 5.62, + "learning_rate": 2.0659826341578927e-07, + "loss": 0.9917, + "step": 76093 + }, + { + "epoch": 5.62, + "learning_rate": 2.065176114901002e-07, + "loss": 0.9964, + "step": 76094 + }, + { + "epoch": 5.62, + "learning_rate": 2.064369751457007e-07, + "loss": 1.0864, + "step": 76095 + }, + { + "epoch": 5.62, + "learning_rate": 2.063563543827185e-07, + "loss": 0.8742, + "step": 76096 + }, + { + "epoch": 5.62, + "learning_rate": 2.0627574920128457e-07, + "loss": 1.0992, + "step": 76097 + }, + { + "epoch": 5.62, + "learning_rate": 2.0619515960152547e-07, + "loss": 0.9644, + "step": 76098 + }, + { + "epoch": 5.62, + "learning_rate": 2.0611458558357e-07, + "loss": 1.0436, + "step": 76099 + }, + { + "epoch": 5.62, + "learning_rate": 2.0603402714754472e-07, + "loss": 0.9411, + "step": 76100 + }, + { + "epoch": 5.62, + "learning_rate": 2.0595348429358064e-07, + "loss": 1.0423, + "step": 76101 + }, + { + "epoch": 5.62, + "learning_rate": 2.0587295702180433e-07, + "loss": 0.9435, + "step": 76102 + }, + { + "epoch": 5.62, + "learning_rate": 2.0579244533234454e-07, + "loss": 1.0476, + "step": 76103 + }, + { + "epoch": 5.62, + "learning_rate": 2.057119492253279e-07, + "loss": 0.8602, + "step": 76104 + }, + { + "epoch": 5.62, + "learning_rate": 2.0563146870088313e-07, + "loss": 0.9379, + "step": 76105 + }, + { + "epoch": 5.62, + "learning_rate": 2.0555100375913906e-07, + "loss": 1.0644, + "step": 76106 + }, + { + "epoch": 5.62, + "learning_rate": 2.0547055440022336e-07, + "loss": 1.0917, + "step": 76107 + }, + { + "epoch": 5.62, + "learning_rate": 2.0539012062426368e-07, + "loss": 1.0178, + "step": 76108 + }, + { + "epoch": 5.62, + "learning_rate": 2.053097024313877e-07, + "loss": 1.034, + "step": 76109 + }, + { + "epoch": 5.62, + "learning_rate": 2.0522929982172423e-07, + "loss": 0.9661, + "step": 76110 + }, + { + "epoch": 5.62, + "learning_rate": 2.0514891279540095e-07, + "loss": 0.9722, + "step": 76111 + }, + { + "epoch": 5.62, + "learning_rate": 2.050685413525455e-07, + "loss": 1.1155, + "step": 76112 + }, + { + "epoch": 5.62, + "learning_rate": 2.049881854932867e-07, + "loss": 0.9534, + "step": 76113 + }, + { + "epoch": 5.62, + "learning_rate": 2.0490784521774998e-07, + "loss": 1.0247, + "step": 76114 + }, + { + "epoch": 5.62, + "learning_rate": 2.0482752052606635e-07, + "loss": 1.0589, + "step": 76115 + }, + { + "epoch": 5.62, + "learning_rate": 2.0474721141836019e-07, + "loss": 1.0521, + "step": 76116 + }, + { + "epoch": 5.62, + "learning_rate": 2.0466691789476244e-07, + "loss": 1.0052, + "step": 76117 + }, + { + "epoch": 5.62, + "learning_rate": 2.0458663995539863e-07, + "loss": 0.94, + "step": 76118 + }, + { + "epoch": 5.62, + "learning_rate": 2.045063776003986e-07, + "loss": 1.0287, + "step": 76119 + }, + { + "epoch": 5.62, + "learning_rate": 2.0442613082988673e-07, + "loss": 1.0309, + "step": 76120 + }, + { + "epoch": 5.62, + "learning_rate": 2.04345899643994e-07, + "loss": 0.9621, + "step": 76121 + }, + { + "epoch": 5.62, + "learning_rate": 2.0426568404284698e-07, + "loss": 0.8997, + "step": 76122 + }, + { + "epoch": 5.62, + "learning_rate": 2.0418548402657334e-07, + "loss": 0.894, + "step": 76123 + }, + { + "epoch": 5.62, + "learning_rate": 2.0410529959529967e-07, + "loss": 1.0189, + "step": 76124 + }, + { + "epoch": 5.62, + "learning_rate": 2.0402513074915477e-07, + "loss": 0.9868, + "step": 76125 + }, + { + "epoch": 5.62, + "learning_rate": 2.0394497748826514e-07, + "loss": 1.0329, + "step": 76126 + }, + { + "epoch": 5.62, + "learning_rate": 2.0386483981275962e-07, + "loss": 1.0375, + "step": 76127 + }, + { + "epoch": 5.63, + "learning_rate": 2.0378471772276477e-07, + "loss": 1.0644, + "step": 76128 + }, + { + "epoch": 5.63, + "learning_rate": 2.0370461121840713e-07, + "loss": 0.8749, + "step": 76129 + }, + { + "epoch": 5.63, + "learning_rate": 2.036245202998166e-07, + "loss": 0.9536, + "step": 76130 + }, + { + "epoch": 5.63, + "learning_rate": 2.0354444496711978e-07, + "loss": 0.9785, + "step": 76131 + }, + { + "epoch": 5.63, + "learning_rate": 2.034643852204432e-07, + "loss": 0.9228, + "step": 76132 + }, + { + "epoch": 5.63, + "learning_rate": 2.0338434105991457e-07, + "loss": 0.9518, + "step": 76133 + }, + { + "epoch": 5.63, + "learning_rate": 2.0330431248565928e-07, + "loss": 0.9548, + "step": 76134 + }, + { + "epoch": 5.63, + "learning_rate": 2.0322429949780954e-07, + "loss": 0.9738, + "step": 76135 + }, + { + "epoch": 5.63, + "learning_rate": 2.031443020964885e-07, + "loss": 1.0324, + "step": 76136 + }, + { + "epoch": 5.63, + "learning_rate": 2.0306432028182498e-07, + "loss": 0.8996, + "step": 76137 + }, + { + "epoch": 5.63, + "learning_rate": 2.0298435405394446e-07, + "loss": 1.1048, + "step": 76138 + }, + { + "epoch": 5.63, + "learning_rate": 2.029044034129779e-07, + "loss": 0.9689, + "step": 76139 + }, + { + "epoch": 5.63, + "learning_rate": 2.0282446835904858e-07, + "loss": 0.9936, + "step": 76140 + }, + { + "epoch": 5.63, + "learning_rate": 2.0274454889228635e-07, + "loss": 0.8303, + "step": 76141 + }, + { + "epoch": 5.63, + "learning_rate": 2.0266464501281778e-07, + "loss": 1.0334, + "step": 76142 + }, + { + "epoch": 5.63, + "learning_rate": 2.025847567207684e-07, + "loss": 0.963, + "step": 76143 + }, + { + "epoch": 5.63, + "learning_rate": 2.02504884016268e-07, + "loss": 0.9681, + "step": 76144 + }, + { + "epoch": 5.63, + "learning_rate": 2.0242502689943988e-07, + "loss": 0.9373, + "step": 76145 + }, + { + "epoch": 5.63, + "learning_rate": 2.0234518537041504e-07, + "loss": 0.8611, + "step": 76146 + }, + { + "epoch": 5.63, + "learning_rate": 2.022653594293189e-07, + "loss": 0.9182, + "step": 76147 + }, + { + "epoch": 5.63, + "learning_rate": 2.0218554907627808e-07, + "loss": 1.0304, + "step": 76148 + }, + { + "epoch": 5.63, + "learning_rate": 2.021057543114191e-07, + "loss": 1.0727, + "step": 76149 + }, + { + "epoch": 5.63, + "learning_rate": 2.0202597513486967e-07, + "loss": 1.0496, + "step": 76150 + }, + { + "epoch": 5.63, + "learning_rate": 2.019462115467574e-07, + "loss": 0.9961, + "step": 76151 + }, + { + "epoch": 5.63, + "learning_rate": 2.0186646354720896e-07, + "loss": 1.0422, + "step": 76152 + }, + { + "epoch": 5.63, + "learning_rate": 2.0178673113635082e-07, + "loss": 0.8964, + "step": 76153 + }, + { + "epoch": 5.63, + "learning_rate": 2.017070143143085e-07, + "loss": 1.0446, + "step": 76154 + }, + { + "epoch": 5.63, + "learning_rate": 2.0162731308121075e-07, + "loss": 0.9552, + "step": 76155 + }, + { + "epoch": 5.63, + "learning_rate": 2.0154762743718414e-07, + "loss": 1.052, + "step": 76156 + }, + { + "epoch": 5.63, + "learning_rate": 2.0146795738235524e-07, + "loss": 0.9664, + "step": 76157 + }, + { + "epoch": 5.63, + "learning_rate": 2.0138830291684952e-07, + "loss": 1.0859, + "step": 76158 + }, + { + "epoch": 5.63, + "learning_rate": 2.0130866404079575e-07, + "loss": 0.9818, + "step": 76159 + }, + { + "epoch": 5.63, + "learning_rate": 2.0122904075432047e-07, + "loss": 1.0009, + "step": 76160 + }, + { + "epoch": 5.63, + "learning_rate": 2.0114943305754809e-07, + "loss": 1.0824, + "step": 76161 + }, + { + "epoch": 5.63, + "learning_rate": 2.0106984095060733e-07, + "loss": 1.007, + "step": 76162 + }, + { + "epoch": 5.63, + "learning_rate": 2.0099026443362367e-07, + "loss": 1.036, + "step": 76163 + }, + { + "epoch": 5.63, + "learning_rate": 2.0091070350672592e-07, + "loss": 1.0093, + "step": 76164 + }, + { + "epoch": 5.63, + "learning_rate": 2.0083115817003728e-07, + "loss": 0.9961, + "step": 76165 + }, + { + "epoch": 5.63, + "learning_rate": 2.0075162842368657e-07, + "loss": 0.9445, + "step": 76166 + }, + { + "epoch": 5.63, + "learning_rate": 2.0067211426780032e-07, + "loss": 0.9078, + "step": 76167 + }, + { + "epoch": 5.63, + "learning_rate": 2.00592615702504e-07, + "loss": 0.9319, + "step": 76168 + }, + { + "epoch": 5.63, + "learning_rate": 2.0051313272792417e-07, + "loss": 1.051, + "step": 76169 + }, + { + "epoch": 5.63, + "learning_rate": 2.004336653441885e-07, + "loss": 0.9362, + "step": 76170 + }, + { + "epoch": 5.63, + "learning_rate": 2.0035421355142247e-07, + "loss": 0.8902, + "step": 76171 + }, + { + "epoch": 5.63, + "learning_rate": 2.0027477734975266e-07, + "loss": 1.0947, + "step": 76172 + }, + { + "epoch": 5.63, + "learning_rate": 2.0019535673930557e-07, + "loss": 0.9999, + "step": 76173 + }, + { + "epoch": 5.63, + "learning_rate": 2.001159517202067e-07, + "loss": 0.9114, + "step": 76174 + }, + { + "epoch": 5.63, + "learning_rate": 2.0003656229258483e-07, + "loss": 0.9932, + "step": 76175 + }, + { + "epoch": 5.63, + "learning_rate": 1.999571884565632e-07, + "loss": 1.0962, + "step": 76176 + }, + { + "epoch": 5.63, + "learning_rate": 1.998778302122706e-07, + "loss": 1.0206, + "step": 76177 + }, + { + "epoch": 5.63, + "learning_rate": 1.9979848755983022e-07, + "loss": 0.9148, + "step": 76178 + }, + { + "epoch": 5.63, + "learning_rate": 1.9971916049937202e-07, + "loss": 0.9027, + "step": 76179 + }, + { + "epoch": 5.63, + "learning_rate": 1.996398490310203e-07, + "loss": 1.0418, + "step": 76180 + }, + { + "epoch": 5.63, + "learning_rate": 1.9956055315490053e-07, + "loss": 1.0371, + "step": 76181 + }, + { + "epoch": 5.63, + "learning_rate": 1.994812728711404e-07, + "loss": 0.9712, + "step": 76182 + }, + { + "epoch": 5.63, + "learning_rate": 1.9940200817986422e-07, + "loss": 0.9625, + "step": 76183 + }, + { + "epoch": 5.63, + "learning_rate": 1.993227590812008e-07, + "loss": 0.9028, + "step": 76184 + }, + { + "epoch": 5.63, + "learning_rate": 1.992435255752734e-07, + "loss": 0.7968, + "step": 76185 + }, + { + "epoch": 5.63, + "learning_rate": 1.9916430766220961e-07, + "loss": 1.0523, + "step": 76186 + }, + { + "epoch": 5.63, + "learning_rate": 1.99085105342135e-07, + "loss": 0.9389, + "step": 76187 + }, + { + "epoch": 5.63, + "learning_rate": 1.9900591861517716e-07, + "loss": 1.0898, + "step": 76188 + }, + { + "epoch": 5.63, + "learning_rate": 1.9892674748145935e-07, + "loss": 1.0584, + "step": 76189 + }, + { + "epoch": 5.63, + "learning_rate": 1.9884759194110925e-07, + "loss": 1.0347, + "step": 76190 + }, + { + "epoch": 5.63, + "learning_rate": 1.9876845199425232e-07, + "loss": 0.9648, + "step": 76191 + }, + { + "epoch": 5.63, + "learning_rate": 1.98689327641014e-07, + "loss": 0.929, + "step": 76192 + }, + { + "epoch": 5.63, + "learning_rate": 1.9861021888152198e-07, + "loss": 0.9671, + "step": 76193 + }, + { + "epoch": 5.63, + "learning_rate": 1.9853112571589948e-07, + "loss": 1.0018, + "step": 76194 + }, + { + "epoch": 5.63, + "learning_rate": 1.9845204814427422e-07, + "loss": 1.0582, + "step": 76195 + }, + { + "epoch": 5.63, + "learning_rate": 1.9837298616677158e-07, + "loss": 0.9131, + "step": 76196 + }, + { + "epoch": 5.63, + "learning_rate": 1.982939397835182e-07, + "loss": 0.9621, + "step": 76197 + }, + { + "epoch": 5.63, + "learning_rate": 1.9821490899463724e-07, + "loss": 0.9227, + "step": 76198 + }, + { + "epoch": 5.63, + "learning_rate": 1.9813589380025643e-07, + "loss": 1.1073, + "step": 76199 + }, + { + "epoch": 5.63, + "learning_rate": 1.9805689420050233e-07, + "loss": 0.9534, + "step": 76200 + }, + { + "epoch": 5.63, + "learning_rate": 1.9797791019549927e-07, + "loss": 1.0182, + "step": 76201 + }, + { + "epoch": 5.63, + "learning_rate": 1.9789894178537272e-07, + "loss": 0.9831, + "step": 76202 + }, + { + "epoch": 5.63, + "learning_rate": 1.9781998897024812e-07, + "loss": 1.0001, + "step": 76203 + }, + { + "epoch": 5.63, + "learning_rate": 1.9774105175025204e-07, + "loss": 0.9803, + "step": 76204 + }, + { + "epoch": 5.63, + "learning_rate": 1.9766213012550994e-07, + "loss": 1.0023, + "step": 76205 + }, + { + "epoch": 5.63, + "learning_rate": 1.975832240961473e-07, + "loss": 1.0067, + "step": 76206 + }, + { + "epoch": 5.63, + "learning_rate": 1.9750433366228949e-07, + "loss": 0.9859, + "step": 76207 + }, + { + "epoch": 5.63, + "learning_rate": 1.9742545882406205e-07, + "loss": 1.1195, + "step": 76208 + }, + { + "epoch": 5.63, + "learning_rate": 1.9734659958159043e-07, + "loss": 1.0295, + "step": 76209 + }, + { + "epoch": 5.63, + "learning_rate": 1.972677559349989e-07, + "loss": 0.8354, + "step": 76210 + }, + { + "epoch": 5.63, + "learning_rate": 1.9718892788441524e-07, + "loss": 0.9878, + "step": 76211 + }, + { + "epoch": 5.63, + "learning_rate": 1.9711011542996373e-07, + "loss": 0.9979, + "step": 76212 + }, + { + "epoch": 5.63, + "learning_rate": 1.9703131857176982e-07, + "loss": 0.894, + "step": 76213 + }, + { + "epoch": 5.63, + "learning_rate": 1.9695253730995788e-07, + "loss": 1.0445, + "step": 76214 + }, + { + "epoch": 5.63, + "learning_rate": 1.968737716446556e-07, + "loss": 0.9768, + "step": 76215 + }, + { + "epoch": 5.63, + "learning_rate": 1.9679502157598617e-07, + "loss": 0.9715, + "step": 76216 + }, + { + "epoch": 5.63, + "learning_rate": 1.967162871040762e-07, + "loss": 1.0029, + "step": 76217 + }, + { + "epoch": 5.63, + "learning_rate": 1.9663756822905e-07, + "loss": 0.9803, + "step": 76218 + }, + { + "epoch": 5.63, + "learning_rate": 1.9655886495103193e-07, + "loss": 1.0657, + "step": 76219 + }, + { + "epoch": 5.63, + "learning_rate": 1.9648017727014967e-07, + "loss": 1.0266, + "step": 76220 + }, + { + "epoch": 5.63, + "learning_rate": 1.9640150518652755e-07, + "loss": 0.9186, + "step": 76221 + }, + { + "epoch": 5.63, + "learning_rate": 1.9632284870028995e-07, + "loss": 1.0314, + "step": 76222 + }, + { + "epoch": 5.63, + "learning_rate": 1.9624420781156117e-07, + "loss": 0.9382, + "step": 76223 + }, + { + "epoch": 5.63, + "learning_rate": 1.961655825204689e-07, + "loss": 0.9678, + "step": 76224 + }, + { + "epoch": 5.63, + "learning_rate": 1.9608697282713752e-07, + "loss": 1.0748, + "step": 76225 + }, + { + "epoch": 5.63, + "learning_rate": 1.9600837873169132e-07, + "loss": 1.0733, + "step": 76226 + }, + { + "epoch": 5.63, + "learning_rate": 1.9592980023425466e-07, + "loss": 1.0779, + "step": 76227 + }, + { + "epoch": 5.63, + "learning_rate": 1.9585123733495305e-07, + "loss": 1.0046, + "step": 76228 + }, + { + "epoch": 5.63, + "learning_rate": 1.957726900339141e-07, + "loss": 1.0112, + "step": 76229 + }, + { + "epoch": 5.63, + "learning_rate": 1.9569415833125772e-07, + "loss": 0.974, + "step": 76230 + }, + { + "epoch": 5.63, + "learning_rate": 1.9561564222711272e-07, + "loss": 0.9834, + "step": 76231 + }, + { + "epoch": 5.63, + "learning_rate": 1.9553714172160232e-07, + "loss": 1.1044, + "step": 76232 + }, + { + "epoch": 5.63, + "learning_rate": 1.9545865681485422e-07, + "loss": 1.041, + "step": 76233 + }, + { + "epoch": 5.63, + "learning_rate": 1.9538018750698938e-07, + "loss": 0.9314, + "step": 76234 + }, + { + "epoch": 5.63, + "learning_rate": 1.9530173379813444e-07, + "loss": 1.1167, + "step": 76235 + }, + { + "epoch": 5.63, + "learning_rate": 1.9522329568841482e-07, + "loss": 0.8853, + "step": 76236 + }, + { + "epoch": 5.63, + "learning_rate": 1.9514487317795483e-07, + "loss": 1.035, + "step": 76237 + }, + { + "epoch": 5.63, + "learning_rate": 1.9506646626687775e-07, + "loss": 0.9642, + "step": 76238 + }, + { + "epoch": 5.63, + "learning_rate": 1.9498807495531012e-07, + "loss": 1.003, + "step": 76239 + }, + { + "epoch": 5.63, + "learning_rate": 1.9490969924337632e-07, + "loss": 0.9903, + "step": 76240 + }, + { + "epoch": 5.63, + "learning_rate": 1.9483133913120177e-07, + "loss": 1.0648, + "step": 76241 + }, + { + "epoch": 5.63, + "learning_rate": 1.9475299461890972e-07, + "loss": 0.9782, + "step": 76242 + }, + { + "epoch": 5.63, + "learning_rate": 1.9467466570662452e-07, + "loss": 1.075, + "step": 76243 + }, + { + "epoch": 5.63, + "learning_rate": 1.9459635239447273e-07, + "loss": 1.0135, + "step": 76244 + }, + { + "epoch": 5.63, + "learning_rate": 1.9451805468257756e-07, + "loss": 0.8777, + "step": 76245 + }, + { + "epoch": 5.63, + "learning_rate": 1.944397725710634e-07, + "loss": 0.893, + "step": 76246 + }, + { + "epoch": 5.63, + "learning_rate": 1.9436150606005565e-07, + "loss": 0.9054, + "step": 76247 + }, + { + "epoch": 5.63, + "learning_rate": 1.942832551496776e-07, + "loss": 0.9921, + "step": 76248 + }, + { + "epoch": 5.63, + "learning_rate": 1.942050198400558e-07, + "loss": 1.0028, + "step": 76249 + }, + { + "epoch": 5.63, + "learning_rate": 1.9412680013131347e-07, + "loss": 0.9429, + "step": 76250 + }, + { + "epoch": 5.63, + "learning_rate": 1.9404859602357495e-07, + "loss": 1.0446, + "step": 76251 + }, + { + "epoch": 5.63, + "learning_rate": 1.939704075169635e-07, + "loss": 0.9354, + "step": 76252 + }, + { + "epoch": 5.63, + "learning_rate": 1.9389223461160679e-07, + "loss": 1.0256, + "step": 76253 + }, + { + "epoch": 5.63, + "learning_rate": 1.9381407730762691e-07, + "loss": 0.8782, + "step": 76254 + }, + { + "epoch": 5.63, + "learning_rate": 1.9373593560514714e-07, + "loss": 1.1625, + "step": 76255 + }, + { + "epoch": 5.63, + "learning_rate": 1.9365780950429403e-07, + "loss": 1.0861, + "step": 76256 + }, + { + "epoch": 5.63, + "learning_rate": 1.9357969900519193e-07, + "loss": 0.9968, + "step": 76257 + }, + { + "epoch": 5.63, + "learning_rate": 1.9350160410796293e-07, + "loss": 0.991, + "step": 76258 + }, + { + "epoch": 5.63, + "learning_rate": 1.9342352481273253e-07, + "loss": 1.0635, + "step": 76259 + }, + { + "epoch": 5.63, + "learning_rate": 1.9334546111962617e-07, + "loss": 0.9762, + "step": 76260 + }, + { + "epoch": 5.63, + "learning_rate": 1.9326741302876595e-07, + "loss": 1.0144, + "step": 76261 + }, + { + "epoch": 5.63, + "learning_rate": 1.9318938054027737e-07, + "loss": 0.9919, + "step": 76262 + }, + { + "epoch": 5.64, + "learning_rate": 1.9311136365428362e-07, + "loss": 0.9684, + "step": 76263 + }, + { + "epoch": 5.64, + "learning_rate": 1.9303336237090908e-07, + "loss": 0.9922, + "step": 76264 + }, + { + "epoch": 5.64, + "learning_rate": 1.929553766902792e-07, + "loss": 0.984, + "step": 76265 + }, + { + "epoch": 5.64, + "learning_rate": 1.928774066125172e-07, + "loss": 1.0181, + "step": 76266 + }, + { + "epoch": 5.64, + "learning_rate": 1.927994521377463e-07, + "loss": 0.9718, + "step": 76267 + }, + { + "epoch": 5.64, + "learning_rate": 1.927215132660909e-07, + "loss": 0.9886, + "step": 76268 + }, + { + "epoch": 5.64, + "learning_rate": 1.926435899976753e-07, + "loss": 0.8765, + "step": 76269 + }, + { + "epoch": 5.64, + "learning_rate": 1.9256568233262385e-07, + "loss": 1.0948, + "step": 76270 + }, + { + "epoch": 5.64, + "learning_rate": 1.9248779027105978e-07, + "loss": 0.9297, + "step": 76271 + }, + { + "epoch": 5.64, + "learning_rate": 1.9240991381310747e-07, + "loss": 1.0161, + "step": 76272 + }, + { + "epoch": 5.64, + "learning_rate": 1.9233205295889011e-07, + "loss": 1.036, + "step": 76273 + }, + { + "epoch": 5.64, + "learning_rate": 1.922542077085343e-07, + "loss": 0.9847, + "step": 76274 + }, + { + "epoch": 5.64, + "learning_rate": 1.9217637806215883e-07, + "loss": 1.0383, + "step": 76275 + }, + { + "epoch": 5.64, + "learning_rate": 1.9209856401989136e-07, + "loss": 1.0051, + "step": 76276 + }, + { + "epoch": 5.64, + "learning_rate": 1.920207655818551e-07, + "loss": 1.0115, + "step": 76277 + }, + { + "epoch": 5.64, + "learning_rate": 1.9194298274817446e-07, + "loss": 1.0238, + "step": 76278 + }, + { + "epoch": 5.64, + "learning_rate": 1.918652155189704e-07, + "loss": 1.0835, + "step": 76279 + }, + { + "epoch": 5.64, + "learning_rate": 1.9178746389436952e-07, + "loss": 0.9945, + "step": 76280 + }, + { + "epoch": 5.64, + "learning_rate": 1.9170972787449283e-07, + "loss": 1.0377, + "step": 76281 + }, + { + "epoch": 5.64, + "learning_rate": 1.9163200745946797e-07, + "loss": 0.9835, + "step": 76282 + }, + { + "epoch": 5.64, + "learning_rate": 1.9155430264941598e-07, + "loss": 0.933, + "step": 76283 + }, + { + "epoch": 5.64, + "learning_rate": 1.91476613444459e-07, + "loss": 0.8648, + "step": 76284 + }, + { + "epoch": 5.64, + "learning_rate": 1.913989398447236e-07, + "loss": 1.0405, + "step": 76285 + }, + { + "epoch": 5.64, + "learning_rate": 1.9132128185033183e-07, + "loss": 0.9848, + "step": 76286 + }, + { + "epoch": 5.64, + "learning_rate": 1.9124363946140812e-07, + "loss": 1.0508, + "step": 76287 + }, + { + "epoch": 5.64, + "learning_rate": 1.9116601267807454e-07, + "loss": 0.9566, + "step": 76288 + }, + { + "epoch": 5.64, + "learning_rate": 1.910884015004555e-07, + "loss": 0.9516, + "step": 76289 + }, + { + "epoch": 5.64, + "learning_rate": 1.9101080592867527e-07, + "loss": 1.0842, + "step": 76290 + }, + { + "epoch": 5.64, + "learning_rate": 1.9093322596285602e-07, + "loss": 0.9992, + "step": 76291 + }, + { + "epoch": 5.64, + "learning_rate": 1.9085566160312207e-07, + "loss": 0.9835, + "step": 76292 + }, + { + "epoch": 5.64, + "learning_rate": 1.9077811284959556e-07, + "loss": 0.9911, + "step": 76293 + }, + { + "epoch": 5.64, + "learning_rate": 1.9070057970240084e-07, + "loss": 1.1504, + "step": 76294 + }, + { + "epoch": 5.64, + "learning_rate": 1.9062306216166226e-07, + "loss": 0.9909, + "step": 76295 + }, + { + "epoch": 5.64, + "learning_rate": 1.9054556022750192e-07, + "loss": 0.9397, + "step": 76296 + }, + { + "epoch": 5.64, + "learning_rate": 1.9046807390004197e-07, + "loss": 1.0513, + "step": 76297 + }, + { + "epoch": 5.64, + "learning_rate": 1.9039060317940893e-07, + "loss": 1.0577, + "step": 76298 + }, + { + "epoch": 5.64, + "learning_rate": 1.9031314806572166e-07, + "loss": 1.0192, + "step": 76299 + }, + { + "epoch": 5.64, + "learning_rate": 1.9023570855910666e-07, + "loss": 1.0914, + "step": 76300 + }, + { + "epoch": 5.64, + "learning_rate": 1.9015828465968723e-07, + "loss": 0.98, + "step": 76301 + }, + { + "epoch": 5.64, + "learning_rate": 1.9008087636758543e-07, + "loss": 1.0223, + "step": 76302 + }, + { + "epoch": 5.64, + "learning_rate": 1.9000348368292454e-07, + "loss": 0.9603, + "step": 76303 + }, + { + "epoch": 5.64, + "learning_rate": 1.8992610660582667e-07, + "loss": 1.0442, + "step": 76304 + }, + { + "epoch": 5.64, + "learning_rate": 1.8984874513641616e-07, + "loss": 1.1027, + "step": 76305 + }, + { + "epoch": 5.64, + "learning_rate": 1.8977139927481736e-07, + "loss": 0.9647, + "step": 76306 + }, + { + "epoch": 5.64, + "learning_rate": 1.8969406902115128e-07, + "loss": 1.0101, + "step": 76307 + }, + { + "epoch": 5.64, + "learning_rate": 1.8961675437554006e-07, + "loss": 1.0001, + "step": 76308 + }, + { + "epoch": 5.64, + "learning_rate": 1.8953945533811024e-07, + "loss": 1.0868, + "step": 76309 + }, + { + "epoch": 5.64, + "learning_rate": 1.8946217190898175e-07, + "loss": 0.8953, + "step": 76310 + }, + { + "epoch": 5.64, + "learning_rate": 1.8938490408827893e-07, + "loss": 1.0801, + "step": 76311 + }, + { + "epoch": 5.64, + "learning_rate": 1.8930765187612498e-07, + "loss": 0.9313, + "step": 76312 + }, + { + "epoch": 5.64, + "learning_rate": 1.8923041527264096e-07, + "loss": 0.9814, + "step": 76313 + }, + { + "epoch": 5.64, + "learning_rate": 1.891531942779512e-07, + "loss": 0.9632, + "step": 76314 + }, + { + "epoch": 5.64, + "learning_rate": 1.8907598889217893e-07, + "loss": 1.0087, + "step": 76315 + }, + { + "epoch": 5.64, + "learning_rate": 1.8899879911544627e-07, + "loss": 0.926, + "step": 76316 + }, + { + "epoch": 5.64, + "learning_rate": 1.8892162494787536e-07, + "loss": 0.9027, + "step": 76317 + }, + { + "epoch": 5.64, + "learning_rate": 1.8884446638959052e-07, + "loss": 0.8832, + "step": 76318 + }, + { + "epoch": 5.64, + "learning_rate": 1.8876732344071502e-07, + "loss": 0.8953, + "step": 76319 + }, + { + "epoch": 5.64, + "learning_rate": 1.8869019610136875e-07, + "loss": 1.0457, + "step": 76320 + }, + { + "epoch": 5.64, + "learning_rate": 1.8861308437167603e-07, + "loss": 1.0107, + "step": 76321 + }, + { + "epoch": 5.64, + "learning_rate": 1.88535988251759e-07, + "loss": 1.0909, + "step": 76322 + }, + { + "epoch": 5.64, + "learning_rate": 1.8845890774174203e-07, + "loss": 0.9115, + "step": 76323 + }, + { + "epoch": 5.64, + "learning_rate": 1.883818428417461e-07, + "loss": 0.9809, + "step": 76324 + }, + { + "epoch": 5.64, + "learning_rate": 1.8830479355189446e-07, + "loss": 1.0115, + "step": 76325 + }, + { + "epoch": 5.64, + "learning_rate": 1.8822775987230813e-07, + "loss": 0.9754, + "step": 76326 + }, + { + "epoch": 5.64, + "learning_rate": 1.8815074180311367e-07, + "loss": 0.9763, + "step": 76327 + }, + { + "epoch": 5.64, + "learning_rate": 1.8807373934442985e-07, + "loss": 0.9759, + "step": 76328 + }, + { + "epoch": 5.64, + "learning_rate": 1.8799675249637884e-07, + "loss": 1.0878, + "step": 76329 + }, + { + "epoch": 5.64, + "learning_rate": 1.8791978125908605e-07, + "loss": 1.103, + "step": 76330 + }, + { + "epoch": 5.64, + "learning_rate": 1.8784282563267254e-07, + "loss": 0.9329, + "step": 76331 + }, + { + "epoch": 5.64, + "learning_rate": 1.8776588561726038e-07, + "loss": 0.993, + "step": 76332 + }, + { + "epoch": 5.64, + "learning_rate": 1.8768896121297174e-07, + "loss": 0.9286, + "step": 76333 + }, + { + "epoch": 5.64, + "learning_rate": 1.8761205241992985e-07, + "loss": 0.9817, + "step": 76334 + }, + { + "epoch": 5.64, + "learning_rate": 1.8753515923825794e-07, + "loss": 1.0436, + "step": 76335 + }, + { + "epoch": 5.64, + "learning_rate": 1.874582816680759e-07, + "loss": 0.8845, + "step": 76336 + }, + { + "epoch": 5.64, + "learning_rate": 1.8738141970950808e-07, + "loss": 1.0062, + "step": 76337 + }, + { + "epoch": 5.64, + "learning_rate": 1.8730457336267548e-07, + "loss": 1.0648, + "step": 76338 + }, + { + "epoch": 5.64, + "learning_rate": 1.872277426277014e-07, + "loss": 0.913, + "step": 76339 + }, + { + "epoch": 5.64, + "learning_rate": 1.8715092750470787e-07, + "loss": 0.878, + "step": 76340 + }, + { + "epoch": 5.64, + "learning_rate": 1.8707412799381596e-07, + "loss": 0.9911, + "step": 76341 + }, + { + "epoch": 5.64, + "learning_rate": 1.8699734409514893e-07, + "loss": 0.9496, + "step": 76342 + }, + { + "epoch": 5.64, + "learning_rate": 1.8692057580882994e-07, + "loss": 1.0365, + "step": 76343 + }, + { + "epoch": 5.64, + "learning_rate": 1.8684382313497785e-07, + "loss": 1.0422, + "step": 76344 + }, + { + "epoch": 5.64, + "learning_rate": 1.8676708607371809e-07, + "loss": 0.9963, + "step": 76345 + }, + { + "epoch": 5.64, + "learning_rate": 1.8669036462517054e-07, + "loss": 1.0042, + "step": 76346 + }, + { + "epoch": 5.64, + "learning_rate": 1.8661365878945957e-07, + "loss": 1.0305, + "step": 76347 + }, + { + "epoch": 5.64, + "learning_rate": 1.865369685667051e-07, + "loss": 1.0463, + "step": 76348 + }, + { + "epoch": 5.64, + "learning_rate": 1.8646029395702925e-07, + "loss": 0.9688, + "step": 76349 + }, + { + "epoch": 5.64, + "learning_rate": 1.8638363496055633e-07, + "loss": 1.0711, + "step": 76350 + }, + { + "epoch": 5.64, + "learning_rate": 1.8630699157740518e-07, + "loss": 1.0527, + "step": 76351 + }, + { + "epoch": 5.64, + "learning_rate": 1.8623036380770009e-07, + "loss": 1.0033, + "step": 76352 + }, + { + "epoch": 5.64, + "learning_rate": 1.8615375165156102e-07, + "loss": 0.9042, + "step": 76353 + }, + { + "epoch": 5.64, + "learning_rate": 1.8607715510911228e-07, + "loss": 0.9301, + "step": 76354 + }, + { + "epoch": 5.64, + "learning_rate": 1.860005741804738e-07, + "loss": 1.0483, + "step": 76355 + }, + { + "epoch": 5.64, + "learning_rate": 1.8592400886576767e-07, + "loss": 0.9588, + "step": 76356 + }, + { + "epoch": 5.64, + "learning_rate": 1.8584745916511603e-07, + "loss": 0.9326, + "step": 76357 + }, + { + "epoch": 5.64, + "learning_rate": 1.8577092507864102e-07, + "loss": 1.0573, + "step": 76358 + }, + { + "epoch": 5.64, + "learning_rate": 1.8569440660646364e-07, + "loss": 1.1259, + "step": 76359 + }, + { + "epoch": 5.64, + "learning_rate": 1.8561790374870715e-07, + "loss": 1.1181, + "step": 76360 + }, + { + "epoch": 5.64, + "learning_rate": 1.855414165054914e-07, + "loss": 0.9571, + "step": 76361 + }, + { + "epoch": 5.64, + "learning_rate": 1.8546494487693745e-07, + "loss": 0.9675, + "step": 76362 + }, + { + "epoch": 5.64, + "learning_rate": 1.8538848886316963e-07, + "loss": 0.9961, + "step": 76363 + }, + { + "epoch": 5.64, + "learning_rate": 1.8531204846430895e-07, + "loss": 0.9211, + "step": 76364 + }, + { + "epoch": 5.64, + "learning_rate": 1.8523562368047531e-07, + "loss": 0.9976, + "step": 76365 + }, + { + "epoch": 5.64, + "learning_rate": 1.851592145117931e-07, + "loss": 0.9427, + "step": 76366 + }, + { + "epoch": 5.64, + "learning_rate": 1.8508282095837993e-07, + "loss": 1.0793, + "step": 76367 + }, + { + "epoch": 5.64, + "learning_rate": 1.850064430203613e-07, + "loss": 0.9044, + "step": 76368 + }, + { + "epoch": 5.64, + "learning_rate": 1.849300806978549e-07, + "loss": 0.9503, + "step": 76369 + }, + { + "epoch": 5.64, + "learning_rate": 1.8485373399098615e-07, + "loss": 0.9631, + "step": 76370 + }, + { + "epoch": 5.64, + "learning_rate": 1.847774028998739e-07, + "loss": 0.9852, + "step": 76371 + }, + { + "epoch": 5.64, + "learning_rate": 1.847010874246413e-07, + "loss": 1.0387, + "step": 76372 + }, + { + "epoch": 5.64, + "learning_rate": 1.8462478756540725e-07, + "loss": 1.0597, + "step": 76373 + }, + { + "epoch": 5.64, + "learning_rate": 1.84548503322296e-07, + "loss": 0.935, + "step": 76374 + }, + { + "epoch": 5.64, + "learning_rate": 1.844722346954275e-07, + "loss": 1.0381, + "step": 76375 + }, + { + "epoch": 5.64, + "learning_rate": 1.843959816849228e-07, + "loss": 0.8616, + "step": 76376 + }, + { + "epoch": 5.64, + "learning_rate": 1.8431974429090393e-07, + "loss": 0.904, + "step": 76377 + }, + { + "epoch": 5.64, + "learning_rate": 1.8424352251349086e-07, + "loss": 1.0702, + "step": 76378 + }, + { + "epoch": 5.64, + "learning_rate": 1.8416731635280793e-07, + "loss": 1.0654, + "step": 76379 + }, + { + "epoch": 5.64, + "learning_rate": 1.8409112580897283e-07, + "loss": 0.9817, + "step": 76380 + }, + { + "epoch": 5.64, + "learning_rate": 1.8401495088210874e-07, + "loss": 1.0167, + "step": 76381 + }, + { + "epoch": 5.64, + "learning_rate": 1.8393879157233675e-07, + "loss": 1.0781, + "step": 76382 + }, + { + "epoch": 5.64, + "learning_rate": 1.838626478797767e-07, + "loss": 1.0924, + "step": 76383 + }, + { + "epoch": 5.64, + "learning_rate": 1.8378651980455187e-07, + "loss": 1.0195, + "step": 76384 + }, + { + "epoch": 5.64, + "learning_rate": 1.8371040734678213e-07, + "loss": 0.9589, + "step": 76385 + }, + { + "epoch": 5.64, + "learning_rate": 1.8363431050658854e-07, + "loss": 1.0225, + "step": 76386 + }, + { + "epoch": 5.64, + "learning_rate": 1.8355822928409095e-07, + "loss": 1.047, + "step": 76387 + }, + { + "epoch": 5.64, + "learning_rate": 1.8348216367941485e-07, + "loss": 0.905, + "step": 76388 + }, + { + "epoch": 5.64, + "learning_rate": 1.834061136926746e-07, + "loss": 0.9738, + "step": 76389 + }, + { + "epoch": 5.64, + "learning_rate": 1.833300793239967e-07, + "loss": 1.1073, + "step": 76390 + }, + { + "epoch": 5.64, + "learning_rate": 1.8325406057350003e-07, + "loss": 0.9474, + "step": 76391 + }, + { + "epoch": 5.64, + "learning_rate": 1.8317805744130668e-07, + "loss": 0.9202, + "step": 76392 + }, + { + "epoch": 5.64, + "learning_rate": 1.8310206992753433e-07, + "loss": 1.1066, + "step": 76393 + }, + { + "epoch": 5.64, + "learning_rate": 1.8302609803230732e-07, + "loss": 0.971, + "step": 76394 + }, + { + "epoch": 5.64, + "learning_rate": 1.8295014175574555e-07, + "loss": 0.9126, + "step": 76395 + }, + { + "epoch": 5.64, + "learning_rate": 1.8287420109797004e-07, + "loss": 1.0118, + "step": 76396 + }, + { + "epoch": 5.64, + "learning_rate": 1.8279827605910073e-07, + "loss": 1.0364, + "step": 76397 + }, + { + "epoch": 5.64, + "learning_rate": 1.8272236663925748e-07, + "loss": 1.0502, + "step": 76398 + }, + { + "epoch": 5.65, + "learning_rate": 1.8264647283856463e-07, + "loss": 1.0359, + "step": 76399 + }, + { + "epoch": 5.65, + "learning_rate": 1.825705946571399e-07, + "loss": 0.9864, + "step": 76400 + }, + { + "epoch": 5.65, + "learning_rate": 1.8249473209510428e-07, + "loss": 1.0292, + "step": 76401 + }, + { + "epoch": 5.65, + "learning_rate": 1.8241888515257988e-07, + "loss": 0.9818, + "step": 76402 + }, + { + "epoch": 5.65, + "learning_rate": 1.8234305382968665e-07, + "loss": 0.9863, + "step": 76403 + }, + { + "epoch": 5.65, + "learning_rate": 1.8226723812654446e-07, + "loss": 1.1056, + "step": 76404 + }, + { + "epoch": 5.65, + "learning_rate": 1.8219143804327544e-07, + "loss": 1.0075, + "step": 76405 + }, + { + "epoch": 5.65, + "learning_rate": 1.8211565357999948e-07, + "loss": 0.9545, + "step": 76406 + }, + { + "epoch": 5.65, + "learning_rate": 1.8203988473683543e-07, + "loss": 0.9101, + "step": 76407 + }, + { + "epoch": 5.65, + "learning_rate": 1.8196413151390646e-07, + "loss": 1.1195, + "step": 76408 + }, + { + "epoch": 5.65, + "learning_rate": 1.818883939113325e-07, + "loss": 1.0036, + "step": 76409 + }, + { + "epoch": 5.65, + "learning_rate": 1.8181267192923346e-07, + "loss": 1.112, + "step": 76410 + }, + { + "epoch": 5.65, + "learning_rate": 1.8173696556772925e-07, + "loss": 1.0433, + "step": 76411 + }, + { + "epoch": 5.65, + "learning_rate": 1.8166127482694195e-07, + "loss": 1.1356, + "step": 76412 + }, + { + "epoch": 5.65, + "learning_rate": 1.8158559970699153e-07, + "loss": 0.9963, + "step": 76413 + }, + { + "epoch": 5.65, + "learning_rate": 1.8150994020799673e-07, + "loss": 0.9259, + "step": 76414 + }, + { + "epoch": 5.65, + "learning_rate": 1.8143429633007968e-07, + "loss": 0.9251, + "step": 76415 + }, + { + "epoch": 5.65, + "learning_rate": 1.813586680733592e-07, + "loss": 1.0463, + "step": 76416 + }, + { + "epoch": 5.65, + "learning_rate": 1.8128305543795855e-07, + "loss": 1.1096, + "step": 76417 + }, + { + "epoch": 5.65, + "learning_rate": 1.8120745842399424e-07, + "loss": 0.9987, + "step": 76418 + }, + { + "epoch": 5.65, + "learning_rate": 1.8113187703158953e-07, + "loss": 0.9892, + "step": 76419 + }, + { + "epoch": 5.65, + "learning_rate": 1.810563112608632e-07, + "loss": 0.9868, + "step": 76420 + }, + { + "epoch": 5.65, + "learning_rate": 1.809807611119363e-07, + "loss": 1.0368, + "step": 76421 + }, + { + "epoch": 5.65, + "learning_rate": 1.809052265849276e-07, + "loss": 0.9552, + "step": 76422 + }, + { + "epoch": 5.65, + "learning_rate": 1.8082970767995811e-07, + "loss": 0.978, + "step": 76423 + }, + { + "epoch": 5.65, + "learning_rate": 1.8075420439714884e-07, + "loss": 1.0075, + "step": 76424 + }, + { + "epoch": 5.65, + "learning_rate": 1.8067871673661863e-07, + "loss": 1.0044, + "step": 76425 + }, + { + "epoch": 5.65, + "learning_rate": 1.8060324469848844e-07, + "loss": 0.9372, + "step": 76426 + }, + { + "epoch": 5.65, + "learning_rate": 1.805277882828771e-07, + "loss": 1.1612, + "step": 76427 + }, + { + "epoch": 5.65, + "learning_rate": 1.804523474899067e-07, + "loss": 1.0696, + "step": 76428 + }, + { + "epoch": 5.65, + "learning_rate": 1.8037692231969605e-07, + "loss": 1.0155, + "step": 76429 + }, + { + "epoch": 5.65, + "learning_rate": 1.8030151277236509e-07, + "loss": 0.9658, + "step": 76430 + }, + { + "epoch": 5.65, + "learning_rate": 1.8022611884803366e-07, + "loss": 1.0381, + "step": 76431 + }, + { + "epoch": 5.65, + "learning_rate": 1.801507405468217e-07, + "loss": 1.0842, + "step": 76432 + }, + { + "epoch": 5.65, + "learning_rate": 1.800753778688502e-07, + "loss": 0.9653, + "step": 76433 + }, + { + "epoch": 5.65, + "learning_rate": 1.80000030814238e-07, + "loss": 0.9377, + "step": 76434 + }, + { + "epoch": 5.65, + "learning_rate": 1.7992469938310497e-07, + "loss": 0.9751, + "step": 76435 + }, + { + "epoch": 5.65, + "learning_rate": 1.7984938357557103e-07, + "loss": 1.1064, + "step": 76436 + }, + { + "epoch": 5.65, + "learning_rate": 1.7977408339175717e-07, + "loss": 0.9533, + "step": 76437 + }, + { + "epoch": 5.65, + "learning_rate": 1.796987988317811e-07, + "loss": 1.0043, + "step": 76438 + }, + { + "epoch": 5.65, + "learning_rate": 1.7962352989576382e-07, + "loss": 1.049, + "step": 76439 + }, + { + "epoch": 5.65, + "learning_rate": 1.7954827658382522e-07, + "loss": 1.0881, + "step": 76440 + }, + { + "epoch": 5.65, + "learning_rate": 1.7947303889608636e-07, + "loss": 0.9915, + "step": 76441 + }, + { + "epoch": 5.65, + "learning_rate": 1.7939781683266377e-07, + "loss": 0.974, + "step": 76442 + }, + { + "epoch": 5.65, + "learning_rate": 1.7932261039367848e-07, + "loss": 1.0045, + "step": 76443 + }, + { + "epoch": 5.65, + "learning_rate": 1.792474195792504e-07, + "loss": 1.0423, + "step": 76444 + }, + { + "epoch": 5.65, + "learning_rate": 1.7917224438950053e-07, + "loss": 1.0378, + "step": 76445 + }, + { + "epoch": 5.65, + "learning_rate": 1.7909708482454657e-07, + "loss": 0.9802, + "step": 76446 + }, + { + "epoch": 5.65, + "learning_rate": 1.790219408845073e-07, + "loss": 0.9896, + "step": 76447 + }, + { + "epoch": 5.65, + "learning_rate": 1.7894681256950485e-07, + "loss": 1.0911, + "step": 76448 + }, + { + "epoch": 5.65, + "learning_rate": 1.788716998796569e-07, + "loss": 0.913, + "step": 76449 + }, + { + "epoch": 5.65, + "learning_rate": 1.7879660281508447e-07, + "loss": 0.9962, + "step": 76450 + }, + { + "epoch": 5.65, + "learning_rate": 1.7872152137590638e-07, + "loss": 0.9683, + "step": 76451 + }, + { + "epoch": 5.65, + "learning_rate": 1.7864645556224025e-07, + "loss": 0.9542, + "step": 76452 + }, + { + "epoch": 5.65, + "learning_rate": 1.7857140537420826e-07, + "loss": 0.8985, + "step": 76453 + }, + { + "epoch": 5.65, + "learning_rate": 1.7849637081192804e-07, + "loss": 1.093, + "step": 76454 + }, + { + "epoch": 5.65, + "learning_rate": 1.7842135187551957e-07, + "loss": 0.9575, + "step": 76455 + }, + { + "epoch": 5.65, + "learning_rate": 1.7834634856510157e-07, + "loss": 0.915, + "step": 76456 + }, + { + "epoch": 5.65, + "learning_rate": 1.782713608807951e-07, + "loss": 0.9738, + "step": 76457 + }, + { + "epoch": 5.65, + "learning_rate": 1.7819638882271894e-07, + "loss": 1.0765, + "step": 76458 + }, + { + "epoch": 5.65, + "learning_rate": 1.7812143239099077e-07, + "loss": 1.1114, + "step": 76459 + }, + { + "epoch": 5.65, + "learning_rate": 1.780464915857316e-07, + "loss": 1.0499, + "step": 76460 + }, + { + "epoch": 5.65, + "learning_rate": 1.7797156640705914e-07, + "loss": 1.004, + "step": 76461 + }, + { + "epoch": 5.65, + "learning_rate": 1.778966568550944e-07, + "loss": 1.0722, + "step": 76462 + }, + { + "epoch": 5.65, + "learning_rate": 1.778217629299539e-07, + "loss": 0.9346, + "step": 76463 + }, + { + "epoch": 5.65, + "learning_rate": 1.7774688463175983e-07, + "loss": 0.9503, + "step": 76464 + }, + { + "epoch": 5.65, + "learning_rate": 1.7767202196062983e-07, + "loss": 1.0656, + "step": 76465 + }, + { + "epoch": 5.65, + "learning_rate": 1.7759717491668271e-07, + "loss": 0.9347, + "step": 76466 + }, + { + "epoch": 5.65, + "learning_rate": 1.7752234350003726e-07, + "loss": 0.8941, + "step": 76467 + }, + { + "epoch": 5.65, + "learning_rate": 1.7744752771081453e-07, + "loss": 0.9797, + "step": 76468 + }, + { + "epoch": 5.65, + "learning_rate": 1.7737272754913103e-07, + "loss": 0.9884, + "step": 76469 + }, + { + "epoch": 5.65, + "learning_rate": 1.7729794301510782e-07, + "loss": 1.0061, + "step": 76470 + }, + { + "epoch": 5.65, + "learning_rate": 1.7722317410886369e-07, + "loss": 0.8708, + "step": 76471 + }, + { + "epoch": 5.65, + "learning_rate": 1.771484208305152e-07, + "loss": 0.9485, + "step": 76472 + }, + { + "epoch": 5.65, + "learning_rate": 1.7707368318018448e-07, + "loss": 1.0062, + "step": 76473 + }, + { + "epoch": 5.65, + "learning_rate": 1.7699896115798808e-07, + "loss": 0.9372, + "step": 76474 + }, + { + "epoch": 5.65, + "learning_rate": 1.7692425476404595e-07, + "loss": 1.0831, + "step": 76475 + }, + { + "epoch": 5.65, + "learning_rate": 1.7684956399847685e-07, + "loss": 0.9671, + "step": 76476 + }, + { + "epoch": 5.65, + "learning_rate": 1.767748888613996e-07, + "loss": 1.0109, + "step": 76477 + }, + { + "epoch": 5.65, + "learning_rate": 1.7670022935293406e-07, + "loss": 1.0586, + "step": 76478 + }, + { + "epoch": 5.65, + "learning_rate": 1.7662558547319685e-07, + "loss": 1.0046, + "step": 76479 + }, + { + "epoch": 5.65, + "learning_rate": 1.7655095722230785e-07, + "loss": 0.9229, + "step": 76480 + }, + { + "epoch": 5.65, + "learning_rate": 1.7647634460038476e-07, + "loss": 1.0851, + "step": 76481 + }, + { + "epoch": 5.65, + "learning_rate": 1.7640174760754857e-07, + "loss": 1.0144, + "step": 76482 + }, + { + "epoch": 5.65, + "learning_rate": 1.7632716624391588e-07, + "loss": 1.0603, + "step": 76483 + }, + { + "epoch": 5.65, + "learning_rate": 1.762526005096066e-07, + "loss": 0.9932, + "step": 76484 + }, + { + "epoch": 5.65, + "learning_rate": 1.7617805040473834e-07, + "loss": 0.9995, + "step": 76485 + }, + { + "epoch": 5.65, + "learning_rate": 1.761035159294322e-07, + "loss": 0.9893, + "step": 76486 + }, + { + "epoch": 5.65, + "learning_rate": 1.7602899708380362e-07, + "loss": 0.9374, + "step": 76487 + }, + { + "epoch": 5.65, + "learning_rate": 1.7595449386797026e-07, + "loss": 1.0345, + "step": 76488 + }, + { + "epoch": 5.65, + "learning_rate": 1.7588000628205536e-07, + "loss": 1.0104, + "step": 76489 + }, + { + "epoch": 5.65, + "learning_rate": 1.7580553432617442e-07, + "loss": 0.9486, + "step": 76490 + }, + { + "epoch": 5.65, + "learning_rate": 1.7573107800044508e-07, + "loss": 0.9198, + "step": 76491 + }, + { + "epoch": 5.65, + "learning_rate": 1.7565663730498726e-07, + "loss": 0.9159, + "step": 76492 + }, + { + "epoch": 5.65, + "learning_rate": 1.7558221223991977e-07, + "loss": 0.9584, + "step": 76493 + }, + { + "epoch": 5.65, + "learning_rate": 1.755078028053614e-07, + "loss": 1.0426, + "step": 76494 + }, + { + "epoch": 5.65, + "learning_rate": 1.754334090014287e-07, + "loss": 1.0153, + "step": 76495 + }, + { + "epoch": 5.65, + "learning_rate": 1.7535903082823936e-07, + "loss": 1.0229, + "step": 76496 + }, + { + "epoch": 5.65, + "learning_rate": 1.752846682859155e-07, + "loss": 0.9424, + "step": 76497 + }, + { + "epoch": 5.65, + "learning_rate": 1.7521032137457151e-07, + "loss": 0.9489, + "step": 76498 + }, + { + "epoch": 5.65, + "learning_rate": 1.7513599009432835e-07, + "loss": 1.0734, + "step": 76499 + }, + { + "epoch": 5.65, + "learning_rate": 1.7506167444530374e-07, + "loss": 1.0033, + "step": 76500 + }, + { + "epoch": 5.65, + "learning_rate": 1.7498737442761316e-07, + "loss": 1.0438, + "step": 76501 + }, + { + "epoch": 5.65, + "learning_rate": 1.7491309004137868e-07, + "loss": 1.0024, + "step": 76502 + }, + { + "epoch": 5.65, + "learning_rate": 1.7483882128671804e-07, + "loss": 0.8854, + "step": 76503 + }, + { + "epoch": 5.65, + "learning_rate": 1.7476456816374665e-07, + "loss": 1.009, + "step": 76504 + }, + { + "epoch": 5.65, + "learning_rate": 1.7469033067258446e-07, + "loss": 1.0131, + "step": 76505 + }, + { + "epoch": 5.65, + "learning_rate": 1.7461610881335023e-07, + "loss": 1.0059, + "step": 76506 + }, + { + "epoch": 5.65, + "learning_rate": 1.7454190258616055e-07, + "loss": 0.9366, + "step": 76507 + }, + { + "epoch": 5.65, + "learning_rate": 1.744677119911331e-07, + "loss": 0.9572, + "step": 76508 + }, + { + "epoch": 5.65, + "learning_rate": 1.743935370283889e-07, + "loss": 0.9848, + "step": 76509 + }, + { + "epoch": 5.65, + "learning_rate": 1.743193776980423e-07, + "loss": 0.9447, + "step": 76510 + }, + { + "epoch": 5.65, + "learning_rate": 1.742452340002143e-07, + "loss": 0.958, + "step": 76511 + }, + { + "epoch": 5.65, + "learning_rate": 1.741711059350204e-07, + "loss": 0.9975, + "step": 76512 + }, + { + "epoch": 5.65, + "learning_rate": 1.7409699350258046e-07, + "loss": 0.9831, + "step": 76513 + }, + { + "epoch": 5.65, + "learning_rate": 1.7402289670301108e-07, + "loss": 1.0468, + "step": 76514 + }, + { + "epoch": 5.65, + "learning_rate": 1.7394881553643106e-07, + "loss": 1.0283, + "step": 76515 + }, + { + "epoch": 5.65, + "learning_rate": 1.7387475000295806e-07, + "loss": 1.0178, + "step": 76516 + }, + { + "epoch": 5.65, + "learning_rate": 1.7380070010270977e-07, + "loss": 1.0168, + "step": 76517 + }, + { + "epoch": 5.65, + "learning_rate": 1.737266658358039e-07, + "loss": 0.9608, + "step": 76518 + }, + { + "epoch": 5.65, + "learning_rate": 1.736526472023581e-07, + "loss": 1.0153, + "step": 76519 + }, + { + "epoch": 5.65, + "learning_rate": 1.7357864420249006e-07, + "loss": 1.0682, + "step": 76520 + }, + { + "epoch": 5.65, + "learning_rate": 1.7350465683631746e-07, + "loss": 0.9898, + "step": 76521 + }, + { + "epoch": 5.65, + "learning_rate": 1.7343068510395911e-07, + "loss": 0.9898, + "step": 76522 + }, + { + "epoch": 5.65, + "learning_rate": 1.733567290055327e-07, + "loss": 1.0879, + "step": 76523 + }, + { + "epoch": 5.65, + "learning_rate": 1.7328278854115366e-07, + "loss": 0.9595, + "step": 76524 + }, + { + "epoch": 5.65, + "learning_rate": 1.7320886371094192e-07, + "loss": 1.0186, + "step": 76525 + }, + { + "epoch": 5.65, + "learning_rate": 1.7313495451501404e-07, + "loss": 0.9454, + "step": 76526 + }, + { + "epoch": 5.65, + "learning_rate": 1.7306106095348884e-07, + "loss": 0.9692, + "step": 76527 + }, + { + "epoch": 5.65, + "learning_rate": 1.7298718302648176e-07, + "loss": 1.0353, + "step": 76528 + }, + { + "epoch": 5.65, + "learning_rate": 1.729133207341116e-07, + "loss": 1.0516, + "step": 76529 + }, + { + "epoch": 5.65, + "learning_rate": 1.7283947407649494e-07, + "loss": 0.8433, + "step": 76530 + }, + { + "epoch": 5.65, + "learning_rate": 1.7276564305375165e-07, + "loss": 0.9645, + "step": 76531 + }, + { + "epoch": 5.65, + "learning_rate": 1.7269182766599612e-07, + "loss": 0.8968, + "step": 76532 + }, + { + "epoch": 5.65, + "learning_rate": 1.7261802791334825e-07, + "loss": 1.0286, + "step": 76533 + }, + { + "epoch": 5.66, + "learning_rate": 1.7254424379592461e-07, + "loss": 0.9946, + "step": 76534 + }, + { + "epoch": 5.66, + "learning_rate": 1.7247047531384287e-07, + "loss": 0.9314, + "step": 76535 + }, + { + "epoch": 5.66, + "learning_rate": 1.7239672246721962e-07, + "loss": 0.9804, + "step": 76536 + }, + { + "epoch": 5.66, + "learning_rate": 1.723229852561714e-07, + "loss": 0.8715, + "step": 76537 + }, + { + "epoch": 5.66, + "learning_rate": 1.7224926368081708e-07, + "loss": 0.9984, + "step": 76538 + }, + { + "epoch": 5.66, + "learning_rate": 1.7217555774127427e-07, + "loss": 0.9469, + "step": 76539 + }, + { + "epoch": 5.66, + "learning_rate": 1.7210186743765955e-07, + "loss": 1.0192, + "step": 76540 + }, + { + "epoch": 5.66, + "learning_rate": 1.7202819277008953e-07, + "loss": 1.0529, + "step": 76541 + }, + { + "epoch": 5.66, + "learning_rate": 1.71954533738683e-07, + "loss": 0.9448, + "step": 76542 + }, + { + "epoch": 5.66, + "learning_rate": 1.7188089034355537e-07, + "loss": 0.9566, + "step": 76543 + }, + { + "epoch": 5.66, + "learning_rate": 1.718072625848255e-07, + "loss": 0.9991, + "step": 76544 + }, + { + "epoch": 5.66, + "learning_rate": 1.717336504626088e-07, + "loss": 1.0242, + "step": 76545 + }, + { + "epoch": 5.66, + "learning_rate": 1.7166005397702301e-07, + "loss": 0.9715, + "step": 76546 + }, + { + "epoch": 5.66, + "learning_rate": 1.7158647312818576e-07, + "loss": 0.8725, + "step": 76547 + }, + { + "epoch": 5.66, + "learning_rate": 1.7151290791621478e-07, + "loss": 0.9314, + "step": 76548 + }, + { + "epoch": 5.66, + "learning_rate": 1.7143935834122548e-07, + "loss": 1.0419, + "step": 76549 + }, + { + "epoch": 5.66, + "learning_rate": 1.7136582440333448e-07, + "loss": 1.0205, + "step": 76550 + }, + { + "epoch": 5.66, + "learning_rate": 1.7129230610266168e-07, + "loss": 0.9122, + "step": 76551 + }, + { + "epoch": 5.66, + "learning_rate": 1.712188034393203e-07, + "loss": 0.9552, + "step": 76552 + }, + { + "epoch": 5.66, + "learning_rate": 1.7114531641343024e-07, + "loss": 0.9643, + "step": 76553 + }, + { + "epoch": 5.66, + "learning_rate": 1.710718450251081e-07, + "loss": 1.0826, + "step": 76554 + }, + { + "epoch": 5.66, + "learning_rate": 1.7099838927446932e-07, + "loss": 0.8247, + "step": 76555 + }, + { + "epoch": 5.66, + "learning_rate": 1.7092494916163049e-07, + "loss": 0.9629, + "step": 76556 + }, + { + "epoch": 5.66, + "learning_rate": 1.708515246867104e-07, + "loss": 1.0762, + "step": 76557 + }, + { + "epoch": 5.66, + "learning_rate": 1.7077811584982452e-07, + "loss": 0.9647, + "step": 76558 + }, + { + "epoch": 5.66, + "learning_rate": 1.707047226510905e-07, + "loss": 0.9954, + "step": 76559 + }, + { + "epoch": 5.66, + "learning_rate": 1.7063134509062495e-07, + "loss": 0.9994, + "step": 76560 + }, + { + "epoch": 5.66, + "learning_rate": 1.7055798316854333e-07, + "loss": 1.0442, + "step": 76561 + }, + { + "epoch": 5.66, + "learning_rate": 1.704846368849633e-07, + "loss": 1.0278, + "step": 76562 + }, + { + "epoch": 5.66, + "learning_rate": 1.7041130624000258e-07, + "loss": 1.0404, + "step": 76563 + }, + { + "epoch": 5.66, + "learning_rate": 1.703379912337766e-07, + "loss": 0.8775, + "step": 76564 + }, + { + "epoch": 5.66, + "learning_rate": 1.7026469186640193e-07, + "loss": 0.8895, + "step": 76565 + }, + { + "epoch": 5.66, + "learning_rate": 1.7019140813799407e-07, + "loss": 0.9643, + "step": 76566 + }, + { + "epoch": 5.66, + "learning_rate": 1.701181400486729e-07, + "loss": 0.9736, + "step": 76567 + }, + { + "epoch": 5.66, + "learning_rate": 1.7004488759855276e-07, + "loss": 1.029, + "step": 76568 + }, + { + "epoch": 5.66, + "learning_rate": 1.6997165078775134e-07, + "loss": 0.9461, + "step": 76569 + }, + { + "epoch": 5.66, + "learning_rate": 1.6989842961638193e-07, + "loss": 0.9184, + "step": 76570 + }, + { + "epoch": 5.66, + "learning_rate": 1.6982522408456549e-07, + "loss": 1.0302, + "step": 76571 + }, + { + "epoch": 5.66, + "learning_rate": 1.6975203419241747e-07, + "loss": 0.9052, + "step": 76572 + }, + { + "epoch": 5.66, + "learning_rate": 1.6967885994005118e-07, + "loss": 1.0711, + "step": 76573 + }, + { + "epoch": 5.66, + "learning_rate": 1.6960570132758535e-07, + "loss": 1.0715, + "step": 76574 + }, + { + "epoch": 5.66, + "learning_rate": 1.6953255835513661e-07, + "loss": 1.0407, + "step": 76575 + }, + { + "epoch": 5.66, + "learning_rate": 1.694594310228215e-07, + "loss": 1.0139, + "step": 76576 + }, + { + "epoch": 5.66, + "learning_rate": 1.6938631933075543e-07, + "loss": 1.0547, + "step": 76577 + }, + { + "epoch": 5.66, + "learning_rate": 1.6931322327905507e-07, + "loss": 1.0337, + "step": 76578 + }, + { + "epoch": 5.66, + "learning_rate": 1.6924014286783585e-07, + "loss": 1.0204, + "step": 76579 + }, + { + "epoch": 5.66, + "learning_rate": 1.6916707809721655e-07, + "loss": 1.0364, + "step": 76580 + }, + { + "epoch": 5.66, + "learning_rate": 1.6909402896731042e-07, + "loss": 0.9502, + "step": 76581 + }, + { + "epoch": 5.66, + "learning_rate": 1.6902099547823514e-07, + "loss": 1.074, + "step": 76582 + }, + { + "epoch": 5.66, + "learning_rate": 1.689479776301073e-07, + "loss": 0.9793, + "step": 76583 + }, + { + "epoch": 5.66, + "learning_rate": 1.6887497542304232e-07, + "loss": 1.0179, + "step": 76584 + }, + { + "epoch": 5.66, + "learning_rate": 1.688019888571568e-07, + "loss": 1.042, + "step": 76585 + }, + { + "epoch": 5.66, + "learning_rate": 1.6872901793256624e-07, + "loss": 0.9727, + "step": 76586 + }, + { + "epoch": 5.66, + "learning_rate": 1.6865606264938828e-07, + "loss": 1.0674, + "step": 76587 + }, + { + "epoch": 5.66, + "learning_rate": 1.6858312300773727e-07, + "loss": 0.9131, + "step": 76588 + }, + { + "epoch": 5.66, + "learning_rate": 1.685101990077298e-07, + "loss": 0.9909, + "step": 76589 + }, + { + "epoch": 5.66, + "learning_rate": 1.6843729064948134e-07, + "loss": 1.0501, + "step": 76590 + }, + { + "epoch": 5.66, + "learning_rate": 1.6836439793310843e-07, + "loss": 1.0494, + "step": 76591 + }, + { + "epoch": 5.66, + "learning_rate": 1.6829152085872768e-07, + "loss": 1.142, + "step": 76592 + }, + { + "epoch": 5.66, + "learning_rate": 1.6821865942645455e-07, + "loss": 0.9436, + "step": 76593 + }, + { + "epoch": 5.66, + "learning_rate": 1.6814581363640448e-07, + "loss": 0.9866, + "step": 76594 + }, + { + "epoch": 5.66, + "learning_rate": 1.6807298348869295e-07, + "loss": 1.0392, + "step": 76595 + }, + { + "epoch": 5.66, + "learning_rate": 1.6800016898343873e-07, + "loss": 0.9324, + "step": 76596 + }, + { + "epoch": 5.66, + "learning_rate": 1.67927370120754e-07, + "loss": 1.1039, + "step": 76597 + }, + { + "epoch": 5.66, + "learning_rate": 1.678545869007564e-07, + "loss": 0.9464, + "step": 76598 + }, + { + "epoch": 5.66, + "learning_rate": 1.6778181932356142e-07, + "loss": 1.0557, + "step": 76599 + }, + { + "epoch": 5.66, + "learning_rate": 1.6770906738928672e-07, + "loss": 1.0184, + "step": 76600 + }, + { + "epoch": 5.66, + "learning_rate": 1.6763633109804445e-07, + "loss": 0.9593, + "step": 76601 + }, + { + "epoch": 5.66, + "learning_rate": 1.6756361044995117e-07, + "loss": 1.0814, + "step": 76602 + }, + { + "epoch": 5.66, + "learning_rate": 1.6749090544512458e-07, + "loss": 1.0163, + "step": 76603 + }, + { + "epoch": 5.66, + "learning_rate": 1.6741821608368015e-07, + "loss": 0.9276, + "step": 76604 + }, + { + "epoch": 5.66, + "learning_rate": 1.673455423657322e-07, + "loss": 1.0119, + "step": 76605 + }, + { + "epoch": 5.66, + "learning_rate": 1.672728842913951e-07, + "loss": 1.0925, + "step": 76606 + }, + { + "epoch": 5.66, + "learning_rate": 1.6720024186078764e-07, + "loss": 1.0452, + "step": 76607 + }, + { + "epoch": 5.66, + "learning_rate": 1.671276150740242e-07, + "loss": 0.9892, + "step": 76608 + }, + { + "epoch": 5.66, + "learning_rate": 1.670550039312202e-07, + "loss": 0.9827, + "step": 76609 + }, + { + "epoch": 5.66, + "learning_rate": 1.6698240843249002e-07, + "loss": 1.0036, + "step": 76610 + }, + { + "epoch": 5.66, + "learning_rate": 1.6690982857795024e-07, + "loss": 0.9223, + "step": 76611 + }, + { + "epoch": 5.66, + "learning_rate": 1.668372643677163e-07, + "loss": 0.9387, + "step": 76612 + }, + { + "epoch": 5.66, + "learning_rate": 1.6676471580190366e-07, + "loss": 1.0634, + "step": 76613 + }, + { + "epoch": 5.66, + "learning_rate": 1.6669218288062894e-07, + "loss": 1.0025, + "step": 76614 + }, + { + "epoch": 5.66, + "learning_rate": 1.6661966560400423e-07, + "loss": 1.0357, + "step": 76615 + }, + { + "epoch": 5.66, + "learning_rate": 1.6654716397214831e-07, + "loss": 1.047, + "step": 76616 + }, + { + "epoch": 5.66, + "learning_rate": 1.6647467798517447e-07, + "loss": 0.8623, + "step": 76617 + }, + { + "epoch": 5.66, + "learning_rate": 1.6640220764319925e-07, + "loss": 1.1145, + "step": 76618 + }, + { + "epoch": 5.66, + "learning_rate": 1.66329752946337e-07, + "loss": 0.9737, + "step": 76619 + }, + { + "epoch": 5.66, + "learning_rate": 1.6625731389470322e-07, + "loss": 1.151, + "step": 76620 + }, + { + "epoch": 5.66, + "learning_rate": 1.6618489048841447e-07, + "loss": 0.9829, + "step": 76621 + }, + { + "epoch": 5.66, + "learning_rate": 1.6611248272758285e-07, + "loss": 1.0257, + "step": 76622 + }, + { + "epoch": 5.66, + "learning_rate": 1.6604009061232716e-07, + "loss": 1.0412, + "step": 76623 + }, + { + "epoch": 5.66, + "learning_rate": 1.659677141427607e-07, + "loss": 1.0285, + "step": 76624 + }, + { + "epoch": 5.66, + "learning_rate": 1.6589535331899886e-07, + "loss": 1.0492, + "step": 76625 + }, + { + "epoch": 5.66, + "learning_rate": 1.6582300814115716e-07, + "loss": 0.901, + "step": 76626 + }, + { + "epoch": 5.66, + "learning_rate": 1.657506786093499e-07, + "loss": 0.9203, + "step": 76627 + }, + { + "epoch": 5.66, + "learning_rate": 1.656783647236926e-07, + "loss": 0.9863, + "step": 76628 + }, + { + "epoch": 5.66, + "learning_rate": 1.6560606648430065e-07, + "loss": 1.0199, + "step": 76629 + }, + { + "epoch": 5.66, + "learning_rate": 1.655337838912885e-07, + "loss": 0.9517, + "step": 76630 + }, + { + "epoch": 5.66, + "learning_rate": 1.654615169447704e-07, + "loss": 0.953, + "step": 76631 + }, + { + "epoch": 5.66, + "learning_rate": 1.6538926564486414e-07, + "loss": 1.0372, + "step": 76632 + }, + { + "epoch": 5.66, + "learning_rate": 1.6531702999168177e-07, + "loss": 0.9516, + "step": 76633 + }, + { + "epoch": 5.66, + "learning_rate": 1.6524480998533987e-07, + "loss": 0.9853, + "step": 76634 + }, + { + "epoch": 5.66, + "learning_rate": 1.6517260562595173e-07, + "loss": 1.058, + "step": 76635 + }, + { + "epoch": 5.66, + "learning_rate": 1.6510041691363387e-07, + "loss": 1.0111, + "step": 76636 + }, + { + "epoch": 5.66, + "learning_rate": 1.650282438485007e-07, + "loss": 0.9893, + "step": 76637 + }, + { + "epoch": 5.66, + "learning_rate": 1.6495608643066652e-07, + "loss": 1.1086, + "step": 76638 + }, + { + "epoch": 5.66, + "learning_rate": 1.6488394466024682e-07, + "loss": 1.0141, + "step": 76639 + }, + { + "epoch": 5.66, + "learning_rate": 1.6481181853735594e-07, + "loss": 1.0482, + "step": 76640 + }, + { + "epoch": 5.66, + "learning_rate": 1.6473970806210938e-07, + "loss": 1.0503, + "step": 76641 + }, + { + "epoch": 5.66, + "learning_rate": 1.6466761323462032e-07, + "loss": 1.0471, + "step": 76642 + }, + { + "epoch": 5.66, + "learning_rate": 1.6459553405500429e-07, + "loss": 1.0827, + "step": 76643 + }, + { + "epoch": 5.66, + "learning_rate": 1.645234705233767e-07, + "loss": 0.9507, + "step": 76644 + }, + { + "epoch": 5.66, + "learning_rate": 1.644514226398519e-07, + "loss": 0.9917, + "step": 76645 + }, + { + "epoch": 5.66, + "learning_rate": 1.643793904045432e-07, + "loss": 0.9344, + "step": 76646 + }, + { + "epoch": 5.66, + "learning_rate": 1.6430737381756601e-07, + "loss": 0.986, + "step": 76647 + }, + { + "epoch": 5.66, + "learning_rate": 1.642353728790358e-07, + "loss": 1.0393, + "step": 76648 + }, + { + "epoch": 5.66, + "learning_rate": 1.641633875890658e-07, + "loss": 1.1353, + "step": 76649 + }, + { + "epoch": 5.66, + "learning_rate": 1.640914179477726e-07, + "loss": 0.963, + "step": 76650 + }, + { + "epoch": 5.66, + "learning_rate": 1.6401946395526723e-07, + "loss": 0.9768, + "step": 76651 + }, + { + "epoch": 5.66, + "learning_rate": 1.6394752561166738e-07, + "loss": 0.9823, + "step": 76652 + }, + { + "epoch": 5.66, + "learning_rate": 1.6387560291708516e-07, + "loss": 1.0421, + "step": 76653 + }, + { + "epoch": 5.66, + "learning_rate": 1.6380369587163714e-07, + "loss": 0.8851, + "step": 76654 + }, + { + "epoch": 5.66, + "learning_rate": 1.6373180447543658e-07, + "loss": 1.0493, + "step": 76655 + }, + { + "epoch": 5.66, + "learning_rate": 1.636599287285978e-07, + "loss": 1.0507, + "step": 76656 + }, + { + "epoch": 5.66, + "learning_rate": 1.6358806863123521e-07, + "loss": 1.0525, + "step": 76657 + }, + { + "epoch": 5.66, + "learning_rate": 1.6351622418346425e-07, + "loss": 1.0001, + "step": 76658 + }, + { + "epoch": 5.66, + "learning_rate": 1.6344439538539702e-07, + "loss": 0.9294, + "step": 76659 + }, + { + "epoch": 5.66, + "learning_rate": 1.63372582237149e-07, + "loss": 1.0432, + "step": 76660 + }, + { + "epoch": 5.66, + "learning_rate": 1.633007847388346e-07, + "loss": 0.9625, + "step": 76661 + }, + { + "epoch": 5.66, + "learning_rate": 1.632290028905692e-07, + "loss": 0.9096, + "step": 76662 + }, + { + "epoch": 5.66, + "learning_rate": 1.6315723669246385e-07, + "loss": 0.9863, + "step": 76663 + }, + { + "epoch": 5.66, + "learning_rate": 1.6308548614463516e-07, + "loss": 1.0115, + "step": 76664 + }, + { + "epoch": 5.66, + "learning_rate": 1.6301375124719744e-07, + "loss": 1.0712, + "step": 76665 + }, + { + "epoch": 5.66, + "learning_rate": 1.6294203200026503e-07, + "loss": 1.0292, + "step": 76666 + }, + { + "epoch": 5.66, + "learning_rate": 1.62870328403949e-07, + "loss": 0.9348, + "step": 76667 + }, + { + "epoch": 5.66, + "learning_rate": 1.6279864045836592e-07, + "loss": 0.9314, + "step": 76668 + }, + { + "epoch": 5.67, + "learning_rate": 1.62726968163629e-07, + "loss": 0.9704, + "step": 76669 + }, + { + "epoch": 5.67, + "learning_rate": 1.626553115198548e-07, + "loss": 1.0115, + "step": 76670 + }, + { + "epoch": 5.67, + "learning_rate": 1.625836705271533e-07, + "loss": 1.0398, + "step": 76671 + }, + { + "epoch": 5.67, + "learning_rate": 1.62512045185641e-07, + "loss": 0.9749, + "step": 76672 + }, + { + "epoch": 5.67, + "learning_rate": 1.624404354954323e-07, + "loss": 0.9524, + "step": 76673 + }, + { + "epoch": 5.67, + "learning_rate": 1.6236884145663935e-07, + "loss": 1.0811, + "step": 76674 + }, + { + "epoch": 5.67, + "learning_rate": 1.6229726306937643e-07, + "loss": 1.0241, + "step": 76675 + }, + { + "epoch": 5.67, + "learning_rate": 1.6222570033375796e-07, + "loss": 1.0315, + "step": 76676 + }, + { + "epoch": 5.67, + "learning_rate": 1.6215415324989715e-07, + "loss": 1.0528, + "step": 76677 + }, + { + "epoch": 5.67, + "learning_rate": 1.6208262181790945e-07, + "loss": 0.9154, + "step": 76678 + }, + { + "epoch": 5.67, + "learning_rate": 1.6201110603790705e-07, + "loss": 1.1031, + "step": 76679 + }, + { + "epoch": 5.67, + "learning_rate": 1.6193960591000424e-07, + "loss": 0.9599, + "step": 76680 + }, + { + "epoch": 5.67, + "learning_rate": 1.6186812143431428e-07, + "loss": 1.0595, + "step": 76681 + }, + { + "epoch": 5.67, + "learning_rate": 1.6179665261095267e-07, + "loss": 0.9276, + "step": 76682 + }, + { + "epoch": 5.67, + "learning_rate": 1.617251994400304e-07, + "loss": 1.0469, + "step": 76683 + }, + { + "epoch": 5.67, + "learning_rate": 1.6165376192166405e-07, + "loss": 0.9708, + "step": 76684 + }, + { + "epoch": 5.67, + "learning_rate": 1.615823400559635e-07, + "loss": 0.9624, + "step": 76685 + }, + { + "epoch": 5.67, + "learning_rate": 1.6151093384304762e-07, + "loss": 0.9998, + "step": 76686 + }, + { + "epoch": 5.67, + "learning_rate": 1.6143954328302513e-07, + "loss": 0.9604, + "step": 76687 + }, + { + "epoch": 5.67, + "learning_rate": 1.6136816837601155e-07, + "loss": 0.9148, + "step": 76688 + }, + { + "epoch": 5.67, + "learning_rate": 1.612968091221201e-07, + "loss": 1.1227, + "step": 76689 + }, + { + "epoch": 5.67, + "learning_rate": 1.6122546552146623e-07, + "loss": 1.0188, + "step": 76690 + }, + { + "epoch": 5.67, + "learning_rate": 1.6115413757416097e-07, + "loss": 1.1664, + "step": 76691 + }, + { + "epoch": 5.67, + "learning_rate": 1.610828252803187e-07, + "loss": 1.0049, + "step": 76692 + }, + { + "epoch": 5.67, + "learning_rate": 1.6101152864005377e-07, + "loss": 1.0326, + "step": 76693 + }, + { + "epoch": 5.67, + "learning_rate": 1.6094024765347716e-07, + "loss": 1.0134, + "step": 76694 + }, + { + "epoch": 5.67, + "learning_rate": 1.6086898232070546e-07, + "loss": 0.9389, + "step": 76695 + }, + { + "epoch": 5.67, + "learning_rate": 1.607977326418486e-07, + "loss": 1.0336, + "step": 76696 + }, + { + "epoch": 5.67, + "learning_rate": 1.6072649861702317e-07, + "loss": 1.0887, + "step": 76697 + }, + { + "epoch": 5.67, + "learning_rate": 1.6065528024634126e-07, + "loss": 1.0577, + "step": 76698 + }, + { + "epoch": 5.67, + "learning_rate": 1.60584077529915e-07, + "loss": 0.9185, + "step": 76699 + }, + { + "epoch": 5.67, + "learning_rate": 1.6051289046785878e-07, + "loss": 0.9613, + "step": 76700 + }, + { + "epoch": 5.67, + "learning_rate": 1.6044171906028584e-07, + "loss": 0.9637, + "step": 76701 + }, + { + "epoch": 5.67, + "learning_rate": 1.603705633073105e-07, + "loss": 0.9388, + "step": 76702 + }, + { + "epoch": 5.67, + "learning_rate": 1.6029942320904378e-07, + "loss": 1.0485, + "step": 76703 + }, + { + "epoch": 5.67, + "learning_rate": 1.602282987656001e-07, + "loss": 0.8642, + "step": 76704 + }, + { + "epoch": 5.67, + "learning_rate": 1.6015718997709151e-07, + "loss": 1.0425, + "step": 76705 + }, + { + "epoch": 5.67, + "learning_rate": 1.600860968436335e-07, + "loss": 1.0011, + "step": 76706 + }, + { + "epoch": 5.67, + "learning_rate": 1.6001501936533715e-07, + "loss": 1.0007, + "step": 76707 + }, + { + "epoch": 5.67, + "learning_rate": 1.5994395754231673e-07, + "loss": 0.8797, + "step": 76708 + }, + { + "epoch": 5.67, + "learning_rate": 1.5987291137468332e-07, + "loss": 1.0074, + "step": 76709 + }, + { + "epoch": 5.67, + "learning_rate": 1.5980188086255232e-07, + "loss": 1.0826, + "step": 76710 + }, + { + "epoch": 5.67, + "learning_rate": 1.5973086600603482e-07, + "loss": 1.1259, + "step": 76711 + }, + { + "epoch": 5.67, + "learning_rate": 1.5965986680524625e-07, + "loss": 1.1543, + "step": 76712 + }, + { + "epoch": 5.67, + "learning_rate": 1.5958888326029763e-07, + "loss": 1.0257, + "step": 76713 + }, + { + "epoch": 5.67, + "learning_rate": 1.5951791537130112e-07, + "loss": 0.9559, + "step": 76714 + }, + { + "epoch": 5.67, + "learning_rate": 1.5944696313837216e-07, + "loss": 0.9753, + "step": 76715 + }, + { + "epoch": 5.67, + "learning_rate": 1.5937602656162067e-07, + "loss": 0.9018, + "step": 76716 + }, + { + "epoch": 5.67, + "learning_rate": 1.593051056411632e-07, + "loss": 0.8884, + "step": 76717 + }, + { + "epoch": 5.67, + "learning_rate": 1.5923420037710856e-07, + "loss": 1.0163, + "step": 76718 + }, + { + "epoch": 5.67, + "learning_rate": 1.5916331076957337e-07, + "loss": 0.9573, + "step": 76719 + }, + { + "epoch": 5.67, + "learning_rate": 1.5909243681866637e-07, + "loss": 1.0588, + "step": 76720 + }, + { + "epoch": 5.67, + "learning_rate": 1.5902157852450417e-07, + "loss": 0.9705, + "step": 76721 + }, + { + "epoch": 5.67, + "learning_rate": 1.589507358871978e-07, + "loss": 0.9779, + "step": 76722 + }, + { + "epoch": 5.67, + "learning_rate": 1.5887990890685935e-07, + "loss": 1.0639, + "step": 76723 + }, + { + "epoch": 5.67, + "learning_rate": 1.5880909758360208e-07, + "loss": 1.0291, + "step": 76724 + }, + { + "epoch": 5.67, + "learning_rate": 1.5873830191753926e-07, + "loss": 0.9585, + "step": 76725 + }, + { + "epoch": 5.67, + "learning_rate": 1.5866752190878298e-07, + "loss": 1.0042, + "step": 76726 + }, + { + "epoch": 5.67, + "learning_rate": 1.5859675755744542e-07, + "loss": 0.9775, + "step": 76727 + }, + { + "epoch": 5.67, + "learning_rate": 1.5852600886363977e-07, + "loss": 1.0587, + "step": 76728 + }, + { + "epoch": 5.67, + "learning_rate": 1.5845527582747823e-07, + "loss": 0.9601, + "step": 76729 + }, + { + "epoch": 5.67, + "learning_rate": 1.5838455844907396e-07, + "loss": 0.9091, + "step": 76730 + }, + { + "epoch": 5.67, + "learning_rate": 1.5831385672854027e-07, + "loss": 1.0506, + "step": 76731 + }, + { + "epoch": 5.67, + "learning_rate": 1.5824317066598593e-07, + "loss": 1.0544, + "step": 76732 + }, + { + "epoch": 5.67, + "learning_rate": 1.5817250026152752e-07, + "loss": 1.027, + "step": 76733 + }, + { + "epoch": 5.67, + "learning_rate": 1.5810184551527496e-07, + "loss": 0.9626, + "step": 76734 + }, + { + "epoch": 5.67, + "learning_rate": 1.580312064273426e-07, + "loss": 1.0621, + "step": 76735 + }, + { + "epoch": 5.67, + "learning_rate": 1.5796058299784035e-07, + "loss": 0.9624, + "step": 76736 + }, + { + "epoch": 5.67, + "learning_rate": 1.5788997522688255e-07, + "loss": 1.0448, + "step": 76737 + }, + { + "epoch": 5.67, + "learning_rate": 1.5781938311458133e-07, + "loss": 0.973, + "step": 76738 + }, + { + "epoch": 5.67, + "learning_rate": 1.5774880666104887e-07, + "loss": 1.0652, + "step": 76739 + }, + { + "epoch": 5.67, + "learning_rate": 1.5767824586639724e-07, + "loss": 1.0649, + "step": 76740 + }, + { + "epoch": 5.67, + "learning_rate": 1.576077007307375e-07, + "loss": 1.0005, + "step": 76741 + }, + { + "epoch": 5.67, + "learning_rate": 1.57537171254184e-07, + "loss": 0.9746, + "step": 76742 + }, + { + "epoch": 5.67, + "learning_rate": 1.5746665743684887e-07, + "loss": 0.9673, + "step": 76743 + }, + { + "epoch": 5.67, + "learning_rate": 1.5739615927884199e-07, + "loss": 1.0439, + "step": 76744 + }, + { + "epoch": 5.67, + "learning_rate": 1.5732567678027778e-07, + "loss": 0.9731, + "step": 76745 + }, + { + "epoch": 5.67, + "learning_rate": 1.572552099412672e-07, + "loss": 0.9452, + "step": 76746 + }, + { + "epoch": 5.67, + "learning_rate": 1.571847587619235e-07, + "loss": 1.0156, + "step": 76747 + }, + { + "epoch": 5.67, + "learning_rate": 1.5711432324235666e-07, + "loss": 1.0107, + "step": 76748 + }, + { + "epoch": 5.67, + "learning_rate": 1.5704390338268205e-07, + "loss": 1.0548, + "step": 76749 + }, + { + "epoch": 5.67, + "learning_rate": 1.5697349918300742e-07, + "loss": 1.0235, + "step": 76750 + }, + { + "epoch": 5.67, + "learning_rate": 1.5690311064344822e-07, + "loss": 0.9428, + "step": 76751 + }, + { + "epoch": 5.67, + "learning_rate": 1.5683273776411545e-07, + "loss": 1.0888, + "step": 76752 + }, + { + "epoch": 5.67, + "learning_rate": 1.5676238054512127e-07, + "loss": 1.045, + "step": 76753 + }, + { + "epoch": 5.67, + "learning_rate": 1.566920389865767e-07, + "loss": 1.029, + "step": 76754 + }, + { + "epoch": 5.67, + "learning_rate": 1.5662171308859496e-07, + "loss": 0.9351, + "step": 76755 + }, + { + "epoch": 5.67, + "learning_rate": 1.565514028512871e-07, + "loss": 1.0502, + "step": 76756 + }, + { + "epoch": 5.67, + "learning_rate": 1.5648110827476415e-07, + "loss": 0.9798, + "step": 76757 + }, + { + "epoch": 5.67, + "learning_rate": 1.5641082935914043e-07, + "loss": 0.8266, + "step": 76758 + }, + { + "epoch": 5.67, + "learning_rate": 1.5634056610452474e-07, + "loss": 0.9424, + "step": 76759 + }, + { + "epoch": 5.67, + "learning_rate": 1.5627031851103147e-07, + "loss": 0.9347, + "step": 76760 + }, + { + "epoch": 5.67, + "learning_rate": 1.562000865787705e-07, + "loss": 0.9431, + "step": 76761 + }, + { + "epoch": 5.67, + "learning_rate": 1.561298703078551e-07, + "loss": 0.9788, + "step": 76762 + }, + { + "epoch": 5.67, + "learning_rate": 1.5605966969839627e-07, + "loss": 0.9339, + "step": 76763 + }, + { + "epoch": 5.67, + "learning_rate": 1.55989484750505e-07, + "loss": 1.1345, + "step": 76764 + }, + { + "epoch": 5.67, + "learning_rate": 1.559193154642935e-07, + "loss": 0.9706, + "step": 76765 + }, + { + "epoch": 5.67, + "learning_rate": 1.5584916183987386e-07, + "loss": 1.0124, + "step": 76766 + }, + { + "epoch": 5.67, + "learning_rate": 1.557790238773571e-07, + "loss": 1.0303, + "step": 76767 + }, + { + "epoch": 5.67, + "learning_rate": 1.5570890157685648e-07, + "loss": 1.104, + "step": 76768 + }, + { + "epoch": 5.67, + "learning_rate": 1.5563879493848077e-07, + "loss": 0.9321, + "step": 76769 + }, + { + "epoch": 5.67, + "learning_rate": 1.5556870396234213e-07, + "loss": 0.8982, + "step": 76770 + }, + { + "epoch": 5.67, + "learning_rate": 1.5549862864855381e-07, + "loss": 0.8665, + "step": 76771 + }, + { + "epoch": 5.67, + "learning_rate": 1.554285689972268e-07, + "loss": 1.0254, + "step": 76772 + }, + { + "epoch": 5.67, + "learning_rate": 1.5535852500847216e-07, + "loss": 1.0354, + "step": 76773 + }, + { + "epoch": 5.67, + "learning_rate": 1.5528849668239977e-07, + "loss": 0.9996, + "step": 76774 + }, + { + "epoch": 5.67, + "learning_rate": 1.5521848401912398e-07, + "loss": 0.9838, + "step": 76775 + }, + { + "epoch": 5.67, + "learning_rate": 1.5514848701875474e-07, + "loss": 0.9154, + "step": 76776 + }, + { + "epoch": 5.67, + "learning_rate": 1.5507850568140303e-07, + "loss": 0.9295, + "step": 76777 + }, + { + "epoch": 5.67, + "learning_rate": 1.550085400071799e-07, + "loss": 0.9595, + "step": 76778 + }, + { + "epoch": 5.67, + "learning_rate": 1.549385899961975e-07, + "loss": 0.9862, + "step": 76779 + }, + { + "epoch": 5.67, + "learning_rate": 1.548686556485679e-07, + "loss": 1.0844, + "step": 76780 + }, + { + "epoch": 5.67, + "learning_rate": 1.5479873696439995e-07, + "loss": 0.9507, + "step": 76781 + }, + { + "epoch": 5.67, + "learning_rate": 1.54728833943808e-07, + "loss": 1.0584, + "step": 76782 + }, + { + "epoch": 5.67, + "learning_rate": 1.5465894658689973e-07, + "loss": 1.1168, + "step": 76783 + }, + { + "epoch": 5.67, + "learning_rate": 1.545890748937906e-07, + "loss": 0.9533, + "step": 76784 + }, + { + "epoch": 5.67, + "learning_rate": 1.545192188645872e-07, + "loss": 1.0372, + "step": 76785 + }, + { + "epoch": 5.67, + "learning_rate": 1.5444937849940388e-07, + "loss": 0.9945, + "step": 76786 + }, + { + "epoch": 5.67, + "learning_rate": 1.5437955379835058e-07, + "loss": 1.018, + "step": 76787 + }, + { + "epoch": 5.67, + "learning_rate": 1.5430974476153938e-07, + "loss": 1.0252, + "step": 76788 + }, + { + "epoch": 5.67, + "learning_rate": 1.542399513890791e-07, + "loss": 1.0832, + "step": 76789 + }, + { + "epoch": 5.67, + "learning_rate": 1.54170173681083e-07, + "loss": 1.0233, + "step": 76790 + }, + { + "epoch": 5.67, + "learning_rate": 1.5410041163766098e-07, + "loss": 0.9438, + "step": 76791 + }, + { + "epoch": 5.67, + "learning_rate": 1.5403066525892406e-07, + "loss": 0.9721, + "step": 76792 + }, + { + "epoch": 5.67, + "learning_rate": 1.539609345449844e-07, + "loss": 1.0394, + "step": 76793 + }, + { + "epoch": 5.67, + "learning_rate": 1.5389121949595077e-07, + "loss": 1.0476, + "step": 76794 + }, + { + "epoch": 5.67, + "learning_rate": 1.538215201119364e-07, + "loss": 1.1053, + "step": 76795 + }, + { + "epoch": 5.67, + "learning_rate": 1.5375183639305013e-07, + "loss": 1.0205, + "step": 76796 + }, + { + "epoch": 5.67, + "learning_rate": 1.536821683394052e-07, + "loss": 0.9869, + "step": 76797 + }, + { + "epoch": 5.67, + "learning_rate": 1.5361251595111038e-07, + "loss": 1.0426, + "step": 76798 + }, + { + "epoch": 5.67, + "learning_rate": 1.535428792282767e-07, + "loss": 0.9518, + "step": 76799 + }, + { + "epoch": 5.67, + "learning_rate": 1.534732581710152e-07, + "loss": 0.9637, + "step": 76800 + }, + { + "epoch": 5.67, + "learning_rate": 1.5340365277943802e-07, + "loss": 1.0481, + "step": 76801 + }, + { + "epoch": 5.67, + "learning_rate": 1.5333406305365396e-07, + "loss": 0.9719, + "step": 76802 + }, + { + "epoch": 5.67, + "learning_rate": 1.53264488993774e-07, + "loss": 0.9525, + "step": 76803 + }, + { + "epoch": 5.67, + "learning_rate": 1.5319493059991143e-07, + "loss": 0.9141, + "step": 76804 + }, + { + "epoch": 5.68, + "learning_rate": 1.531253878721739e-07, + "loss": 0.8805, + "step": 76805 + }, + { + "epoch": 5.68, + "learning_rate": 1.5305586081067137e-07, + "loss": 1.0438, + "step": 76806 + }, + { + "epoch": 5.68, + "learning_rate": 1.5298634941551817e-07, + "loss": 1.0238, + "step": 76807 + }, + { + "epoch": 5.68, + "learning_rate": 1.5291685368682197e-07, + "loss": 0.9876, + "step": 76808 + }, + { + "epoch": 5.68, + "learning_rate": 1.528473736246938e-07, + "loss": 1.0758, + "step": 76809 + }, + { + "epoch": 5.68, + "learning_rate": 1.5277790922924474e-07, + "loss": 0.9533, + "step": 76810 + }, + { + "epoch": 5.68, + "learning_rate": 1.5270846050058463e-07, + "loss": 1.0179, + "step": 76811 + }, + { + "epoch": 5.68, + "learning_rate": 1.5263902743882563e-07, + "loss": 0.9486, + "step": 76812 + }, + { + "epoch": 5.68, + "learning_rate": 1.5256961004407656e-07, + "loss": 0.9735, + "step": 76813 + }, + { + "epoch": 5.68, + "learning_rate": 1.525002083164484e-07, + "loss": 1.0212, + "step": 76814 + }, + { + "epoch": 5.68, + "learning_rate": 1.524308222560511e-07, + "loss": 1.0034, + "step": 76815 + }, + { + "epoch": 5.68, + "learning_rate": 1.5236145186299567e-07, + "loss": 1.0356, + "step": 76816 + }, + { + "epoch": 5.68, + "learning_rate": 1.5229209713739312e-07, + "loss": 0.9339, + "step": 76817 + }, + { + "epoch": 5.68, + "learning_rate": 1.5222275807935226e-07, + "loss": 1.0074, + "step": 76818 + }, + { + "epoch": 5.68, + "learning_rate": 1.5215343468898302e-07, + "loss": 0.9851, + "step": 76819 + }, + { + "epoch": 5.68, + "learning_rate": 1.5208412696639864e-07, + "loss": 1.05, + "step": 76820 + }, + { + "epoch": 5.68, + "learning_rate": 1.5201483491170675e-07, + "loss": 1.1084, + "step": 76821 + }, + { + "epoch": 5.68, + "learning_rate": 1.5194555852501846e-07, + "loss": 0.8948, + "step": 76822 + }, + { + "epoch": 5.68, + "learning_rate": 1.5187629780644254e-07, + "loss": 0.9288, + "step": 76823 + }, + { + "epoch": 5.68, + "learning_rate": 1.518070527560922e-07, + "loss": 1.0043, + "step": 76824 + }, + { + "epoch": 5.68, + "learning_rate": 1.5173782337407627e-07, + "loss": 1.1232, + "step": 76825 + }, + { + "epoch": 5.68, + "learning_rate": 1.5166860966050355e-07, + "loss": 1.0168, + "step": 76826 + }, + { + "epoch": 5.68, + "learning_rate": 1.5159941161548507e-07, + "loss": 0.8942, + "step": 76827 + }, + { + "epoch": 5.68, + "learning_rate": 1.5153022923913073e-07, + "loss": 0.9484, + "step": 76828 + }, + { + "epoch": 5.68, + "learning_rate": 1.514610625315527e-07, + "loss": 0.9837, + "step": 76829 + }, + { + "epoch": 5.68, + "learning_rate": 1.5139191149285748e-07, + "loss": 1.0967, + "step": 76830 + }, + { + "epoch": 5.68, + "learning_rate": 1.5132277612315726e-07, + "loss": 1.019, + "step": 76831 + }, + { + "epoch": 5.68, + "learning_rate": 1.5125365642256084e-07, + "loss": 1.0438, + "step": 76832 + }, + { + "epoch": 5.68, + "learning_rate": 1.5118455239118036e-07, + "loss": 1.1178, + "step": 76833 + }, + { + "epoch": 5.68, + "learning_rate": 1.5111546402912457e-07, + "loss": 0.9868, + "step": 76834 + }, + { + "epoch": 5.68, + "learning_rate": 1.5104639133650123e-07, + "loss": 0.9952, + "step": 76835 + }, + { + "epoch": 5.68, + "learning_rate": 1.5097733431342353e-07, + "loss": 0.9858, + "step": 76836 + }, + { + "epoch": 5.68, + "learning_rate": 1.509082929599992e-07, + "loss": 1.0487, + "step": 76837 + }, + { + "epoch": 5.68, + "learning_rate": 1.5083926727634035e-07, + "loss": 0.8705, + "step": 76838 + }, + { + "epoch": 5.68, + "learning_rate": 1.5077025726255356e-07, + "loss": 0.9869, + "step": 76839 + }, + { + "epoch": 5.68, + "learning_rate": 1.5070126291875098e-07, + "loss": 0.8972, + "step": 76840 + }, + { + "epoch": 5.68, + "learning_rate": 1.506322842450425e-07, + "loss": 0.9713, + "step": 76841 + }, + { + "epoch": 5.68, + "learning_rate": 1.5056332124153583e-07, + "loss": 1.0405, + "step": 76842 + }, + { + "epoch": 5.68, + "learning_rate": 1.504943739083431e-07, + "loss": 0.9554, + "step": 76843 + }, + { + "epoch": 5.68, + "learning_rate": 1.50425442245572e-07, + "loss": 0.997, + "step": 76844 + }, + { + "epoch": 5.68, + "learning_rate": 1.5035652625333353e-07, + "loss": 0.9879, + "step": 76845 + }, + { + "epoch": 5.68, + "learning_rate": 1.5028762593173764e-07, + "loss": 0.9841, + "step": 76846 + }, + { + "epoch": 5.68, + "learning_rate": 1.5021874128089308e-07, + "loss": 1.101, + "step": 76847 + }, + { + "epoch": 5.68, + "learning_rate": 1.501498723009076e-07, + "loss": 1.0269, + "step": 76848 + }, + { + "epoch": 5.68, + "learning_rate": 1.5008101899189553e-07, + "loss": 0.8812, + "step": 76849 + }, + { + "epoch": 5.68, + "learning_rate": 1.5001218135396123e-07, + "loss": 1.0061, + "step": 76850 + }, + { + "epoch": 5.68, + "learning_rate": 1.4994335938721792e-07, + "loss": 0.9989, + "step": 76851 + }, + { + "epoch": 5.68, + "learning_rate": 1.4987455309177446e-07, + "loss": 0.9173, + "step": 76852 + }, + { + "epoch": 5.68, + "learning_rate": 1.4980576246773847e-07, + "loss": 0.9425, + "step": 76853 + }, + { + "epoch": 5.68, + "learning_rate": 1.4973698751522104e-07, + "loss": 0.9414, + "step": 76854 + }, + { + "epoch": 5.68, + "learning_rate": 1.496682282343298e-07, + "loss": 1.059, + "step": 76855 + }, + { + "epoch": 5.68, + "learning_rate": 1.4959948462517694e-07, + "loss": 0.9841, + "step": 76856 + }, + { + "epoch": 5.68, + "learning_rate": 1.495307566878701e-07, + "loss": 1.0533, + "step": 76857 + }, + { + "epoch": 5.68, + "learning_rate": 1.4946204442251922e-07, + "loss": 1.1074, + "step": 76858 + }, + { + "epoch": 5.68, + "learning_rate": 1.4939334782923198e-07, + "loss": 1.0395, + "step": 76859 + }, + { + "epoch": 5.68, + "learning_rate": 1.4932466690812054e-07, + "loss": 0.9923, + "step": 76860 + }, + { + "epoch": 5.68, + "learning_rate": 1.4925600165929143e-07, + "loss": 1.0331, + "step": 76861 + }, + { + "epoch": 5.68, + "learning_rate": 1.4918735208285685e-07, + "loss": 1.0313, + "step": 76862 + }, + { + "epoch": 5.68, + "learning_rate": 1.491187181789222e-07, + "loss": 1.0244, + "step": 76863 + }, + { + "epoch": 5.68, + "learning_rate": 1.4905009994759968e-07, + "loss": 1.0866, + "step": 76864 + }, + { + "epoch": 5.68, + "learning_rate": 1.4898149738899692e-07, + "loss": 0.9905, + "step": 76865 + }, + { + "epoch": 5.68, + "learning_rate": 1.48912910503225e-07, + "loss": 0.9819, + "step": 76866 + }, + { + "epoch": 5.68, + "learning_rate": 1.4884433929039044e-07, + "loss": 1.0194, + "step": 76867 + }, + { + "epoch": 5.68, + "learning_rate": 1.4877578375060318e-07, + "loss": 1.036, + "step": 76868 + }, + { + "epoch": 5.68, + "learning_rate": 1.487072438839743e-07, + "loss": 0.9564, + "step": 76869 + }, + { + "epoch": 5.68, + "learning_rate": 1.4863871969061029e-07, + "loss": 0.9039, + "step": 76870 + }, + { + "epoch": 5.68, + "learning_rate": 1.4857021117062109e-07, + "loss": 0.9995, + "step": 76871 + }, + { + "epoch": 5.68, + "learning_rate": 1.4850171832411665e-07, + "loss": 1.0779, + "step": 76872 + }, + { + "epoch": 5.68, + "learning_rate": 1.484332411512035e-07, + "loss": 1.0322, + "step": 76873 + }, + { + "epoch": 5.68, + "learning_rate": 1.483647796519938e-07, + "loss": 1.0213, + "step": 76874 + }, + { + "epoch": 5.68, + "learning_rate": 1.4829633382659302e-07, + "loss": 0.9001, + "step": 76875 + }, + { + "epoch": 5.68, + "learning_rate": 1.482279036751133e-07, + "loss": 1.081, + "step": 76876 + }, + { + "epoch": 5.68, + "learning_rate": 1.481594891976601e-07, + "loss": 1.0546, + "step": 76877 + }, + { + "epoch": 5.68, + "learning_rate": 1.4809109039434665e-07, + "loss": 0.98, + "step": 76878 + }, + { + "epoch": 5.68, + "learning_rate": 1.4802270726527735e-07, + "loss": 1.0947, + "step": 76879 + }, + { + "epoch": 5.68, + "learning_rate": 1.4795433981056428e-07, + "loss": 1.0081, + "step": 76880 + }, + { + "epoch": 5.68, + "learning_rate": 1.4788598803031405e-07, + "loss": 1.0365, + "step": 76881 + }, + { + "epoch": 5.68, + "learning_rate": 1.478176519246366e-07, + "loss": 1.0094, + "step": 76882 + }, + { + "epoch": 5.68, + "learning_rate": 1.4774933149364068e-07, + "loss": 1.0404, + "step": 76883 + }, + { + "epoch": 5.68, + "learning_rate": 1.47681026737434e-07, + "loss": 0.9231, + "step": 76884 + }, + { + "epoch": 5.68, + "learning_rate": 1.476127376561265e-07, + "loss": 1.0509, + "step": 76885 + }, + { + "epoch": 5.68, + "learning_rate": 1.4754446424982583e-07, + "loss": 1.0484, + "step": 76886 + }, + { + "epoch": 5.68, + "learning_rate": 1.474762065186408e-07, + "loss": 1.0588, + "step": 76887 + }, + { + "epoch": 5.68, + "learning_rate": 1.4740796446267914e-07, + "loss": 1.0043, + "step": 76888 + }, + { + "epoch": 5.68, + "learning_rate": 1.4733973808205181e-07, + "loss": 0.8995, + "step": 76889 + }, + { + "epoch": 5.68, + "learning_rate": 1.4727152737686545e-07, + "loss": 1.0631, + "step": 76890 + }, + { + "epoch": 5.68, + "learning_rate": 1.4720333234722994e-07, + "loss": 0.9818, + "step": 76891 + }, + { + "epoch": 5.68, + "learning_rate": 1.4713515299325298e-07, + "loss": 0.971, + "step": 76892 + }, + { + "epoch": 5.68, + "learning_rate": 1.4706698931504116e-07, + "loss": 1.0608, + "step": 76893 + }, + { + "epoch": 5.68, + "learning_rate": 1.469988413127077e-07, + "loss": 0.8703, + "step": 76894 + }, + { + "epoch": 5.68, + "learning_rate": 1.4693070898635587e-07, + "loss": 0.9792, + "step": 76895 + }, + { + "epoch": 5.68, + "learning_rate": 1.4686259233609668e-07, + "loss": 0.9781, + "step": 76896 + }, + { + "epoch": 5.68, + "learning_rate": 1.4679449136203783e-07, + "loss": 1.0477, + "step": 76897 + }, + { + "epoch": 5.68, + "learning_rate": 1.4672640606428922e-07, + "loss": 1.0355, + "step": 76898 + }, + { + "epoch": 5.68, + "learning_rate": 1.4665833644295746e-07, + "loss": 0.8615, + "step": 76899 + }, + { + "epoch": 5.68, + "learning_rate": 1.4659028249815022e-07, + "loss": 1.0803, + "step": 76900 + }, + { + "epoch": 5.68, + "learning_rate": 1.4652224422997742e-07, + "loss": 0.9833, + "step": 76901 + }, + { + "epoch": 5.68, + "learning_rate": 1.4645422163854783e-07, + "loss": 0.9037, + "step": 76902 + }, + { + "epoch": 5.68, + "learning_rate": 1.4638621472396807e-07, + "loss": 0.9182, + "step": 76903 + }, + { + "epoch": 5.68, + "learning_rate": 1.4631822348634585e-07, + "loss": 1.0942, + "step": 76904 + }, + { + "epoch": 5.68, + "learning_rate": 1.4625024792579102e-07, + "loss": 0.968, + "step": 76905 + }, + { + "epoch": 5.68, + "learning_rate": 1.461822880424113e-07, + "loss": 1.02, + "step": 76906 + }, + { + "epoch": 5.68, + "learning_rate": 1.461143438363144e-07, + "loss": 1.0115, + "step": 76907 + }, + { + "epoch": 5.68, + "learning_rate": 1.460464153076091e-07, + "loss": 0.9699, + "step": 76908 + }, + { + "epoch": 5.68, + "learning_rate": 1.45978502456402e-07, + "loss": 1.0124, + "step": 76909 + }, + { + "epoch": 5.68, + "learning_rate": 1.4591060528280187e-07, + "loss": 0.9977, + "step": 76910 + }, + { + "epoch": 5.68, + "learning_rate": 1.4584272378691867e-07, + "loss": 1.0485, + "step": 76911 + }, + { + "epoch": 5.68, + "learning_rate": 1.457748579688578e-07, + "loss": 0.9516, + "step": 76912 + }, + { + "epoch": 5.68, + "learning_rate": 1.4570700782872705e-07, + "loss": 0.9896, + "step": 76913 + }, + { + "epoch": 5.68, + "learning_rate": 1.4563917336663624e-07, + "loss": 0.945, + "step": 76914 + }, + { + "epoch": 5.68, + "learning_rate": 1.4557135458269312e-07, + "loss": 0.8958, + "step": 76915 + }, + { + "epoch": 5.68, + "learning_rate": 1.4550355147700535e-07, + "loss": 0.9588, + "step": 76916 + }, + { + "epoch": 5.68, + "learning_rate": 1.454357640496795e-07, + "loss": 1.0635, + "step": 76917 + }, + { + "epoch": 5.68, + "learning_rate": 1.4536799230082333e-07, + "loss": 1.0861, + "step": 76918 + }, + { + "epoch": 5.68, + "learning_rate": 1.4530023623054778e-07, + "loss": 1.0204, + "step": 76919 + }, + { + "epoch": 5.68, + "learning_rate": 1.4523249583895726e-07, + "loss": 1.0465, + "step": 76920 + }, + { + "epoch": 5.68, + "learning_rate": 1.4516477112616167e-07, + "loss": 0.9375, + "step": 76921 + }, + { + "epoch": 5.68, + "learning_rate": 1.4509706209226648e-07, + "loss": 0.9756, + "step": 76922 + }, + { + "epoch": 5.68, + "learning_rate": 1.450293687373827e-07, + "loss": 1.0577, + "step": 76923 + }, + { + "epoch": 5.68, + "learning_rate": 1.4496169106161473e-07, + "loss": 1.0049, + "step": 76924 + }, + { + "epoch": 5.68, + "learning_rate": 1.4489402906507243e-07, + "loss": 0.9327, + "step": 76925 + }, + { + "epoch": 5.68, + "learning_rate": 1.4482638274786353e-07, + "loss": 1.0036, + "step": 76926 + }, + { + "epoch": 5.68, + "learning_rate": 1.4475875211009348e-07, + "loss": 0.9443, + "step": 76927 + }, + { + "epoch": 5.68, + "learning_rate": 1.4469113715187222e-07, + "loss": 0.9788, + "step": 76928 + }, + { + "epoch": 5.68, + "learning_rate": 1.4462353787330518e-07, + "loss": 0.941, + "step": 76929 + }, + { + "epoch": 5.68, + "learning_rate": 1.4455595427450231e-07, + "loss": 0.8635, + "step": 76930 + }, + { + "epoch": 5.68, + "learning_rate": 1.4448838635556907e-07, + "loss": 1.2049, + "step": 76931 + }, + { + "epoch": 5.68, + "learning_rate": 1.4442083411661423e-07, + "loss": 1.0285, + "step": 76932 + }, + { + "epoch": 5.68, + "learning_rate": 1.4435329755774442e-07, + "loss": 1.0123, + "step": 76933 + }, + { + "epoch": 5.68, + "learning_rate": 1.442857766790684e-07, + "loss": 0.9091, + "step": 76934 + }, + { + "epoch": 5.68, + "learning_rate": 1.4421827148069168e-07, + "loss": 1.0734, + "step": 76935 + }, + { + "epoch": 5.68, + "learning_rate": 1.4415078196272304e-07, + "loss": 0.9887, + "step": 76936 + }, + { + "epoch": 5.68, + "learning_rate": 1.4408330812527015e-07, + "loss": 1.0068, + "step": 76937 + }, + { + "epoch": 5.68, + "learning_rate": 1.4401584996843854e-07, + "loss": 1.0077, + "step": 76938 + }, + { + "epoch": 5.68, + "learning_rate": 1.4394840749233807e-07, + "loss": 1.0784, + "step": 76939 + }, + { + "epoch": 5.69, + "learning_rate": 1.4388098069707312e-07, + "loss": 0.9448, + "step": 76940 + }, + { + "epoch": 5.69, + "learning_rate": 1.438135695827536e-07, + "loss": 0.9172, + "step": 76941 + }, + { + "epoch": 5.69, + "learning_rate": 1.437461741494839e-07, + "loss": 0.9394, + "step": 76942 + }, + { + "epoch": 5.69, + "learning_rate": 1.4367879439737608e-07, + "loss": 1.0413, + "step": 76943 + }, + { + "epoch": 5.69, + "learning_rate": 1.4361143032653125e-07, + "loss": 0.9348, + "step": 76944 + }, + { + "epoch": 5.69, + "learning_rate": 1.435440819370615e-07, + "loss": 1.0318, + "step": 76945 + }, + { + "epoch": 5.69, + "learning_rate": 1.4347674922907228e-07, + "loss": 0.923, + "step": 76946 + }, + { + "epoch": 5.69, + "learning_rate": 1.434094322026691e-07, + "loss": 0.9569, + "step": 76947 + }, + { + "epoch": 5.69, + "learning_rate": 1.433421308579619e-07, + "loss": 0.9485, + "step": 76948 + }, + { + "epoch": 5.69, + "learning_rate": 1.4327484519505497e-07, + "loss": 0.9944, + "step": 76949 + }, + { + "epoch": 5.69, + "learning_rate": 1.4320757521405715e-07, + "loss": 1.0261, + "step": 76950 + }, + { + "epoch": 5.69, + "learning_rate": 1.4314032091507613e-07, + "loss": 1.0356, + "step": 76951 + }, + { + "epoch": 5.69, + "learning_rate": 1.4307308229821736e-07, + "loss": 1.0276, + "step": 76952 + }, + { + "epoch": 5.69, + "learning_rate": 1.4300585936358635e-07, + "loss": 1.1378, + "step": 76953 + }, + { + "epoch": 5.69, + "learning_rate": 1.429386521112941e-07, + "loss": 0.9481, + "step": 76954 + }, + { + "epoch": 5.69, + "learning_rate": 1.4287146054144496e-07, + "loss": 1.0907, + "step": 76955 + }, + { + "epoch": 5.69, + "learning_rate": 1.4280428465414664e-07, + "loss": 1.0944, + "step": 76956 + }, + { + "epoch": 5.69, + "learning_rate": 1.427371244495057e-07, + "loss": 1.0631, + "step": 76957 + }, + { + "epoch": 5.69, + "learning_rate": 1.4266997992762765e-07, + "loss": 1.0028, + "step": 76958 + }, + { + "epoch": 5.69, + "learning_rate": 1.4260285108862126e-07, + "loss": 1.0151, + "step": 76959 + }, + { + "epoch": 5.69, + "learning_rate": 1.4253573793259312e-07, + "loss": 1.0792, + "step": 76960 + }, + { + "epoch": 5.69, + "learning_rate": 1.424686404596498e-07, + "loss": 0.9706, + "step": 76961 + }, + { + "epoch": 5.69, + "learning_rate": 1.424015586698968e-07, + "loss": 0.9383, + "step": 76962 + }, + { + "epoch": 5.69, + "learning_rate": 1.423344925634429e-07, + "loss": 1.1074, + "step": 76963 + }, + { + "epoch": 5.69, + "learning_rate": 1.4226744214039468e-07, + "loss": 0.9823, + "step": 76964 + }, + { + "epoch": 5.69, + "learning_rate": 1.422004074008554e-07, + "loss": 0.876, + "step": 76965 + }, + { + "epoch": 5.69, + "learning_rate": 1.4213338834493606e-07, + "loss": 1.0251, + "step": 76966 + }, + { + "epoch": 5.69, + "learning_rate": 1.4206638497273996e-07, + "loss": 1.0591, + "step": 76967 + }, + { + "epoch": 5.69, + "learning_rate": 1.4199939728437695e-07, + "loss": 0.9119, + "step": 76968 + }, + { + "epoch": 5.69, + "learning_rate": 1.4193242527995033e-07, + "loss": 1.0385, + "step": 76969 + }, + { + "epoch": 5.69, + "learning_rate": 1.4186546895956888e-07, + "loss": 0.9148, + "step": 76970 + }, + { + "epoch": 5.69, + "learning_rate": 1.4179852832333918e-07, + "loss": 0.8831, + "step": 76971 + }, + { + "epoch": 5.69, + "learning_rate": 1.417316033713656e-07, + "loss": 0.9713, + "step": 76972 + }, + { + "epoch": 5.69, + "learning_rate": 1.4166469410375695e-07, + "loss": 0.9143, + "step": 76973 + }, + { + "epoch": 5.69, + "learning_rate": 1.4159780052061756e-07, + "loss": 0.9797, + "step": 76974 + }, + { + "epoch": 5.69, + "learning_rate": 1.4153092262205627e-07, + "loss": 1.0861, + "step": 76975 + }, + { + "epoch": 5.69, + "learning_rate": 1.4146406040817851e-07, + "loss": 1.0526, + "step": 76976 + }, + { + "epoch": 5.69, + "learning_rate": 1.4139721387908977e-07, + "loss": 0.9362, + "step": 76977 + }, + { + "epoch": 5.69, + "learning_rate": 1.4133038303489556e-07, + "loss": 1.0634, + "step": 76978 + }, + { + "epoch": 5.69, + "learning_rate": 1.4126356787570573e-07, + "loss": 0.9316, + "step": 76979 + }, + { + "epoch": 5.69, + "learning_rate": 1.4119676840162355e-07, + "loss": 0.9423, + "step": 76980 + }, + { + "epoch": 5.69, + "learning_rate": 1.4112998461275673e-07, + "loss": 0.9297, + "step": 76981 + }, + { + "epoch": 5.69, + "learning_rate": 1.4106321650921073e-07, + "loss": 0.9289, + "step": 76982 + }, + { + "epoch": 5.69, + "learning_rate": 1.4099646409109214e-07, + "loss": 1.0128, + "step": 76983 + }, + { + "epoch": 5.69, + "learning_rate": 1.409297273585075e-07, + "loss": 1.0106, + "step": 76984 + }, + { + "epoch": 5.69, + "learning_rate": 1.4086300631156125e-07, + "loss": 1.0045, + "step": 76985 + }, + { + "epoch": 5.69, + "learning_rate": 1.4079630095036211e-07, + "loss": 0.9666, + "step": 76986 + }, + { + "epoch": 5.69, + "learning_rate": 1.407296112750145e-07, + "loss": 0.9107, + "step": 76987 + }, + { + "epoch": 5.69, + "learning_rate": 1.4066293728562608e-07, + "loss": 0.9593, + "step": 76988 + }, + { + "epoch": 5.69, + "learning_rate": 1.405962789823001e-07, + "loss": 0.8289, + "step": 76989 + }, + { + "epoch": 5.69, + "learning_rate": 1.4052963636514649e-07, + "loss": 0.998, + "step": 76990 + }, + { + "epoch": 5.69, + "learning_rate": 1.4046300943426737e-07, + "loss": 0.9457, + "step": 76991 + }, + { + "epoch": 5.69, + "learning_rate": 1.403963981897716e-07, + "loss": 1.0491, + "step": 76992 + }, + { + "epoch": 5.69, + "learning_rate": 1.4032980263176342e-07, + "loss": 1.0001, + "step": 76993 + }, + { + "epoch": 5.69, + "learning_rate": 1.4026322276034953e-07, + "loss": 1.0074, + "step": 76994 + }, + { + "epoch": 5.69, + "learning_rate": 1.4019665857563647e-07, + "loss": 0.9972, + "step": 76995 + }, + { + "epoch": 5.69, + "learning_rate": 1.401301100777286e-07, + "loss": 1.1158, + "step": 76996 + }, + { + "epoch": 5.69, + "learning_rate": 1.400635772667336e-07, + "loss": 1.0088, + "step": 76997 + }, + { + "epoch": 5.69, + "learning_rate": 1.399970601427547e-07, + "loss": 0.9528, + "step": 76998 + }, + { + "epoch": 5.69, + "learning_rate": 1.399305587059008e-07, + "loss": 0.9504, + "step": 76999 + }, + { + "epoch": 5.69, + "learning_rate": 1.3986407295627612e-07, + "loss": 1.0555, + "step": 77000 + }, + { + "epoch": 5.69, + "learning_rate": 1.3979760289398624e-07, + "loss": 0.9989, + "step": 77001 + }, + { + "epoch": 5.69, + "learning_rate": 1.397311485191377e-07, + "loss": 0.9757, + "step": 77002 + }, + { + "epoch": 5.69, + "learning_rate": 1.3966470983183488e-07, + "loss": 1.0306, + "step": 77003 + }, + { + "epoch": 5.69, + "learning_rate": 1.3959828683218545e-07, + "loss": 1.0608, + "step": 77004 + }, + { + "epoch": 5.69, + "learning_rate": 1.395318795202949e-07, + "loss": 1.0434, + "step": 77005 + }, + { + "epoch": 5.69, + "learning_rate": 1.3946548789626645e-07, + "loss": 1.0633, + "step": 77006 + }, + { + "epoch": 5.69, + "learning_rate": 1.3939911196020782e-07, + "loss": 1.0002, + "step": 77007 + }, + { + "epoch": 5.69, + "learning_rate": 1.3933275171222338e-07, + "loss": 0.9649, + "step": 77008 + }, + { + "epoch": 5.69, + "learning_rate": 1.392664071524208e-07, + "loss": 0.9159, + "step": 77009 + }, + { + "epoch": 5.69, + "learning_rate": 1.3920007828090442e-07, + "loss": 1.0356, + "step": 77010 + }, + { + "epoch": 5.69, + "learning_rate": 1.3913376509777865e-07, + "loss": 0.9718, + "step": 77011 + }, + { + "epoch": 5.69, + "learning_rate": 1.3906746760315004e-07, + "loss": 0.8852, + "step": 77012 + }, + { + "epoch": 5.69, + "learning_rate": 1.3900118579712407e-07, + "loss": 0.9112, + "step": 77013 + }, + { + "epoch": 5.69, + "learning_rate": 1.3893491967980621e-07, + "loss": 1.117, + "step": 77014 + }, + { + "epoch": 5.69, + "learning_rate": 1.3886866925130193e-07, + "loss": 1.1244, + "step": 77015 + }, + { + "epoch": 5.69, + "learning_rate": 1.388024345117167e-07, + "loss": 0.9287, + "step": 77016 + }, + { + "epoch": 5.69, + "learning_rate": 1.3873621546115602e-07, + "loss": 1.0981, + "step": 77017 + }, + { + "epoch": 5.69, + "learning_rate": 1.386700120997231e-07, + "loss": 1.0418, + "step": 77018 + }, + { + "epoch": 5.69, + "learning_rate": 1.3860382442752674e-07, + "loss": 0.8863, + "step": 77019 + }, + { + "epoch": 5.69, + "learning_rate": 1.3853765244467022e-07, + "loss": 1.0456, + "step": 77020 + }, + { + "epoch": 5.69, + "learning_rate": 1.3847149615125898e-07, + "loss": 1.0288, + "step": 77021 + }, + { + "epoch": 5.69, + "learning_rate": 1.3840535554739964e-07, + "loss": 0.905, + "step": 77022 + }, + { + "epoch": 5.69, + "learning_rate": 1.383392306331943e-07, + "loss": 0.9685, + "step": 77023 + }, + { + "epoch": 5.69, + "learning_rate": 1.3827312140875181e-07, + "loss": 1.0209, + "step": 77024 + }, + { + "epoch": 5.69, + "learning_rate": 1.3820702787417427e-07, + "loss": 0.952, + "step": 77025 + }, + { + "epoch": 5.69, + "learning_rate": 1.381409500295694e-07, + "loss": 1.0229, + "step": 77026 + }, + { + "epoch": 5.69, + "learning_rate": 1.3807488787504042e-07, + "loss": 1.0247, + "step": 77027 + }, + { + "epoch": 5.69, + "learning_rate": 1.3800884141069394e-07, + "loss": 1.0081, + "step": 77028 + }, + { + "epoch": 5.69, + "learning_rate": 1.3794281063663428e-07, + "loss": 0.9303, + "step": 77029 + }, + { + "epoch": 5.69, + "learning_rate": 1.3787679555296585e-07, + "loss": 1.1112, + "step": 77030 + }, + { + "epoch": 5.69, + "learning_rate": 1.378107961597952e-07, + "loss": 0.9506, + "step": 77031 + }, + { + "epoch": 5.69, + "learning_rate": 1.377448124572256e-07, + "loss": 0.9641, + "step": 77032 + }, + { + "epoch": 5.69, + "learning_rate": 1.376788444453636e-07, + "loss": 0.9589, + "step": 77033 + }, + { + "epoch": 5.69, + "learning_rate": 1.376128921243125e-07, + "loss": 1.1113, + "step": 77034 + }, + { + "epoch": 5.69, + "learning_rate": 1.3754695549417885e-07, + "loss": 0.9864, + "step": 77035 + }, + { + "epoch": 5.69, + "learning_rate": 1.374810345550659e-07, + "loss": 0.9787, + "step": 77036 + }, + { + "epoch": 5.69, + "learning_rate": 1.3741512930708135e-07, + "loss": 1.0641, + "step": 77037 + }, + { + "epoch": 5.69, + "learning_rate": 1.3734923975032732e-07, + "loss": 1.0672, + "step": 77038 + }, + { + "epoch": 5.69, + "learning_rate": 1.3728336588490932e-07, + "loss": 0.9228, + "step": 77039 + }, + { + "epoch": 5.69, + "learning_rate": 1.372175077109328e-07, + "loss": 1.0874, + "step": 77040 + }, + { + "epoch": 5.69, + "learning_rate": 1.371516652285021e-07, + "loss": 1.0326, + "step": 77041 + }, + { + "epoch": 5.69, + "learning_rate": 1.370858384377216e-07, + "loss": 0.9948, + "step": 77042 + }, + { + "epoch": 5.69, + "learning_rate": 1.370200273386957e-07, + "loss": 0.8765, + "step": 77043 + }, + { + "epoch": 5.69, + "learning_rate": 1.369542319315309e-07, + "loss": 1.053, + "step": 77044 + }, + { + "epoch": 5.69, + "learning_rate": 1.3688845221633052e-07, + "loss": 0.9651, + "step": 77045 + }, + { + "epoch": 5.69, + "learning_rate": 1.368226881932e-07, + "loss": 0.9977, + "step": 77046 + }, + { + "epoch": 5.69, + "learning_rate": 1.367569398622426e-07, + "loss": 1.0565, + "step": 77047 + }, + { + "epoch": 5.69, + "learning_rate": 1.3669120722356487e-07, + "loss": 1.0124, + "step": 77048 + }, + { + "epoch": 5.69, + "learning_rate": 1.36625490277269e-07, + "loss": 0.9894, + "step": 77049 + }, + { + "epoch": 5.69, + "learning_rate": 1.3655978902346157e-07, + "loss": 0.9685, + "step": 77050 + }, + { + "epoch": 5.69, + "learning_rate": 1.364941034622469e-07, + "loss": 0.8269, + "step": 77051 + }, + { + "epoch": 5.69, + "learning_rate": 1.3642843359372827e-07, + "loss": 1.1109, + "step": 77052 + }, + { + "epoch": 5.69, + "learning_rate": 1.3636277941801114e-07, + "loss": 0.9442, + "step": 77053 + }, + { + "epoch": 5.69, + "learning_rate": 1.3629714093519986e-07, + "loss": 0.9666, + "step": 77054 + }, + { + "epoch": 5.69, + "learning_rate": 1.362315181453977e-07, + "loss": 0.9962, + "step": 77055 + }, + { + "epoch": 5.69, + "learning_rate": 1.3616591104871014e-07, + "loss": 0.9499, + "step": 77056 + }, + { + "epoch": 5.69, + "learning_rate": 1.3610031964524262e-07, + "loss": 1.0777, + "step": 77057 + }, + { + "epoch": 5.69, + "learning_rate": 1.3603474393509841e-07, + "loss": 1.0026, + "step": 77058 + }, + { + "epoch": 5.69, + "learning_rate": 1.3596918391838078e-07, + "loss": 0.9655, + "step": 77059 + }, + { + "epoch": 5.69, + "learning_rate": 1.3590363959519625e-07, + "loss": 1.0046, + "step": 77060 + }, + { + "epoch": 5.69, + "learning_rate": 1.3583811096564702e-07, + "loss": 0.9525, + "step": 77061 + }, + { + "epoch": 5.69, + "learning_rate": 1.3577259802983854e-07, + "loss": 0.947, + "step": 77062 + }, + { + "epoch": 5.69, + "learning_rate": 1.3570710078787408e-07, + "loss": 1.0447, + "step": 77063 + }, + { + "epoch": 5.69, + "learning_rate": 1.3564161923985907e-07, + "loss": 0.9994, + "step": 77064 + }, + { + "epoch": 5.69, + "learning_rate": 1.355761533858968e-07, + "loss": 0.9818, + "step": 77065 + }, + { + "epoch": 5.69, + "learning_rate": 1.3551070322609272e-07, + "loss": 0.987, + "step": 77066 + }, + { + "epoch": 5.69, + "learning_rate": 1.3544526876054898e-07, + "loss": 0.9953, + "step": 77067 + }, + { + "epoch": 5.69, + "learning_rate": 1.3537984998936992e-07, + "loss": 0.8676, + "step": 77068 + }, + { + "epoch": 5.69, + "learning_rate": 1.3531444691266216e-07, + "loss": 0.9914, + "step": 77069 + }, + { + "epoch": 5.69, + "learning_rate": 1.352490595305267e-07, + "loss": 1.0178, + "step": 77070 + }, + { + "epoch": 5.69, + "learning_rate": 1.3518368784307013e-07, + "loss": 0.9176, + "step": 77071 + }, + { + "epoch": 5.69, + "learning_rate": 1.3511833185039346e-07, + "loss": 1.0131, + "step": 77072 + }, + { + "epoch": 5.69, + "learning_rate": 1.3505299155260332e-07, + "loss": 1.0325, + "step": 77073 + }, + { + "epoch": 5.69, + "learning_rate": 1.3498766694980292e-07, + "loss": 1.0603, + "step": 77074 + }, + { + "epoch": 5.7, + "learning_rate": 1.3492235804209552e-07, + "loss": 1.1456, + "step": 77075 + }, + { + "epoch": 5.7, + "learning_rate": 1.348570648295866e-07, + "loss": 0.9937, + "step": 77076 + }, + { + "epoch": 5.7, + "learning_rate": 1.3479178731237718e-07, + "loss": 0.9246, + "step": 77077 + }, + { + "epoch": 5.7, + "learning_rate": 1.3472652549057497e-07, + "loss": 0.9682, + "step": 77078 + }, + { + "epoch": 5.7, + "learning_rate": 1.3466127936427986e-07, + "loss": 1.0333, + "step": 77079 + }, + { + "epoch": 5.7, + "learning_rate": 1.3459604893359846e-07, + "loss": 0.9616, + "step": 77080 + }, + { + "epoch": 5.7, + "learning_rate": 1.345308341986329e-07, + "loss": 0.9107, + "step": 77081 + }, + { + "epoch": 5.7, + "learning_rate": 1.3446563515948863e-07, + "loss": 0.9927, + "step": 77082 + }, + { + "epoch": 5.7, + "learning_rate": 1.344004518162667e-07, + "loss": 0.9979, + "step": 77083 + }, + { + "epoch": 5.7, + "learning_rate": 1.3433528416907372e-07, + "loss": 0.9856, + "step": 77084 + }, + { + "epoch": 5.7, + "learning_rate": 1.3427013221801177e-07, + "loss": 0.9878, + "step": 77085 + }, + { + "epoch": 5.7, + "learning_rate": 1.3420499596318637e-07, + "loss": 0.8795, + "step": 77086 + }, + { + "epoch": 5.7, + "learning_rate": 1.3413987540469853e-07, + "loss": 1.0045, + "step": 77087 + }, + { + "epoch": 5.7, + "learning_rate": 1.340747705426515e-07, + "loss": 0.9821, + "step": 77088 + }, + { + "epoch": 5.7, + "learning_rate": 1.3400968137715188e-07, + "loss": 0.9618, + "step": 77089 + }, + { + "epoch": 5.7, + "learning_rate": 1.3394460790830178e-07, + "loss": 1.0191, + "step": 77090 + }, + { + "epoch": 5.7, + "learning_rate": 1.3387955013620447e-07, + "loss": 0.9943, + "step": 77091 + }, + { + "epoch": 5.7, + "learning_rate": 1.3381450806096318e-07, + "loss": 1.042, + "step": 77092 + }, + { + "epoch": 5.7, + "learning_rate": 1.337494816826823e-07, + "loss": 1.0612, + "step": 77093 + }, + { + "epoch": 5.7, + "learning_rate": 1.3368447100146508e-07, + "loss": 1.0256, + "step": 77094 + }, + { + "epoch": 5.7, + "learning_rate": 1.3361947601741477e-07, + "loss": 0.9936, + "step": 77095 + }, + { + "epoch": 5.7, + "learning_rate": 1.3355449673063458e-07, + "loss": 1.0302, + "step": 77096 + }, + { + "epoch": 5.7, + "learning_rate": 1.334895331412267e-07, + "loss": 1.0082, + "step": 77097 + }, + { + "epoch": 5.7, + "learning_rate": 1.334245852492977e-07, + "loss": 1.1, + "step": 77098 + }, + { + "epoch": 5.7, + "learning_rate": 1.3335965305494746e-07, + "loss": 1.0104, + "step": 77099 + }, + { + "epoch": 5.7, + "learning_rate": 1.3329473655828262e-07, + "loss": 1.0798, + "step": 77100 + }, + { + "epoch": 5.7, + "learning_rate": 1.3322983575940307e-07, + "loss": 1.0528, + "step": 77101 + }, + { + "epoch": 5.7, + "learning_rate": 1.331649506584143e-07, + "loss": 1.0156, + "step": 77102 + }, + { + "epoch": 5.7, + "learning_rate": 1.331000812554184e-07, + "loss": 1.0411, + "step": 77103 + }, + { + "epoch": 5.7, + "learning_rate": 1.3303522755051978e-07, + "loss": 1.0297, + "step": 77104 + }, + { + "epoch": 5.7, + "learning_rate": 1.329703895438217e-07, + "loss": 1.0123, + "step": 77105 + }, + { + "epoch": 5.7, + "learning_rate": 1.3290556723542513e-07, + "loss": 0.9464, + "step": 77106 + }, + { + "epoch": 5.7, + "learning_rate": 1.328407606254356e-07, + "loss": 1.0466, + "step": 77107 + }, + { + "epoch": 5.7, + "learning_rate": 1.3277596971395414e-07, + "loss": 1.0808, + "step": 77108 + }, + { + "epoch": 5.7, + "learning_rate": 1.3271119450108504e-07, + "loss": 1.0567, + "step": 77109 + }, + { + "epoch": 5.7, + "learning_rate": 1.3264643498693274e-07, + "loss": 0.9661, + "step": 77110 + }, + { + "epoch": 5.7, + "learning_rate": 1.3258169117159713e-07, + "loss": 0.8457, + "step": 77111 + }, + { + "epoch": 5.7, + "learning_rate": 1.3251696305518369e-07, + "loss": 1.1041, + "step": 77112 + }, + { + "epoch": 5.7, + "learning_rate": 1.3245225063779453e-07, + "loss": 1.0, + "step": 77113 + }, + { + "epoch": 5.7, + "learning_rate": 1.3238755391953185e-07, + "loss": 0.9931, + "step": 77114 + }, + { + "epoch": 5.7, + "learning_rate": 1.3232287290050105e-07, + "loss": 1.0199, + "step": 77115 + }, + { + "epoch": 5.7, + "learning_rate": 1.322582075808021e-07, + "loss": 1.0247, + "step": 77116 + }, + { + "epoch": 5.7, + "learning_rate": 1.3219355796053934e-07, + "loss": 1.0563, + "step": 77117 + }, + { + "epoch": 5.7, + "learning_rate": 1.321289240398149e-07, + "loss": 1.0446, + "step": 77118 + }, + { + "epoch": 5.7, + "learning_rate": 1.3206430581873318e-07, + "loss": 0.9137, + "step": 77119 + }, + { + "epoch": 5.7, + "learning_rate": 1.319997032973963e-07, + "loss": 1.0638, + "step": 77120 + }, + { + "epoch": 5.7, + "learning_rate": 1.319351164759053e-07, + "loss": 1.0259, + "step": 77121 + }, + { + "epoch": 5.7, + "learning_rate": 1.318705453543656e-07, + "loss": 0.8942, + "step": 77122 + }, + { + "epoch": 5.7, + "learning_rate": 1.318059899328783e-07, + "loss": 0.9665, + "step": 77123 + }, + { + "epoch": 5.7, + "learning_rate": 1.317414502115455e-07, + "loss": 0.9939, + "step": 77124 + }, + { + "epoch": 5.7, + "learning_rate": 1.316769261904727e-07, + "loss": 1.1152, + "step": 77125 + }, + { + "epoch": 5.7, + "learning_rate": 1.3161241786975863e-07, + "loss": 0.8523, + "step": 77126 + }, + { + "epoch": 5.7, + "learning_rate": 1.3154792524951e-07, + "loss": 0.9553, + "step": 77127 + }, + { + "epoch": 5.7, + "learning_rate": 1.3148344832982552e-07, + "loss": 1.0111, + "step": 77128 + }, + { + "epoch": 5.7, + "learning_rate": 1.3141898711081181e-07, + "loss": 1.0146, + "step": 77129 + }, + { + "epoch": 5.7, + "learning_rate": 1.3135454159256654e-07, + "loss": 1.0437, + "step": 77130 + }, + { + "epoch": 5.7, + "learning_rate": 1.3129011177519747e-07, + "loss": 0.9446, + "step": 77131 + }, + { + "epoch": 5.7, + "learning_rate": 1.3122569765880444e-07, + "loss": 0.9749, + "step": 77132 + }, + { + "epoch": 5.7, + "learning_rate": 1.3116129924348853e-07, + "loss": 1.0091, + "step": 77133 + }, + { + "epoch": 5.7, + "learning_rate": 1.3109691652935518e-07, + "loss": 1.0492, + "step": 77134 + }, + { + "epoch": 5.7, + "learning_rate": 1.3103254951650546e-07, + "loss": 1.0006, + "step": 77135 + }, + { + "epoch": 5.7, + "learning_rate": 1.3096819820504147e-07, + "loss": 1.0422, + "step": 77136 + }, + { + "epoch": 5.7, + "learning_rate": 1.3090386259506428e-07, + "loss": 0.9591, + "step": 77137 + }, + { + "epoch": 5.7, + "learning_rate": 1.3083954268667932e-07, + "loss": 0.9591, + "step": 77138 + }, + { + "epoch": 5.7, + "learning_rate": 1.3077523847998763e-07, + "loss": 0.9179, + "step": 77139 + }, + { + "epoch": 5.7, + "learning_rate": 1.3071094997509137e-07, + "loss": 0.9119, + "step": 77140 + }, + { + "epoch": 5.7, + "learning_rate": 1.3064667717209156e-07, + "loss": 1.005, + "step": 77141 + }, + { + "epoch": 5.7, + "learning_rate": 1.3058242007109368e-07, + "loss": 0.955, + "step": 77142 + }, + { + "epoch": 5.7, + "learning_rate": 1.3051817867219652e-07, + "loss": 0.9965, + "step": 77143 + }, + { + "epoch": 5.7, + "learning_rate": 1.3045395297550444e-07, + "loss": 1.0254, + "step": 77144 + }, + { + "epoch": 5.7, + "learning_rate": 1.303897429811185e-07, + "loss": 0.9665, + "step": 77145 + }, + { + "epoch": 5.7, + "learning_rate": 1.3032554868914082e-07, + "loss": 1.0341, + "step": 77146 + }, + { + "epoch": 5.7, + "learning_rate": 1.3026137009967465e-07, + "loss": 0.9886, + "step": 77147 + }, + { + "epoch": 5.7, + "learning_rate": 1.3019720721282102e-07, + "loss": 0.9692, + "step": 77148 + }, + { + "epoch": 5.7, + "learning_rate": 1.301330600286832e-07, + "loss": 1.0156, + "step": 77149 + }, + { + "epoch": 5.7, + "learning_rate": 1.3006892854736108e-07, + "loss": 0.9095, + "step": 77150 + }, + { + "epoch": 5.7, + "learning_rate": 1.3000481276896016e-07, + "loss": 0.9466, + "step": 77151 + }, + { + "epoch": 5.7, + "learning_rate": 1.2994071269357922e-07, + "loss": 1.0638, + "step": 77152 + }, + { + "epoch": 5.7, + "learning_rate": 1.298766283213204e-07, + "loss": 1.0411, + "step": 77153 + }, + { + "epoch": 5.7, + "learning_rate": 1.2981255965228812e-07, + "loss": 1.0095, + "step": 77154 + }, + { + "epoch": 5.7, + "learning_rate": 1.2974850668658224e-07, + "loss": 1.1102, + "step": 77155 + }, + { + "epoch": 5.7, + "learning_rate": 1.296844694243049e-07, + "loss": 1.0412, + "step": 77156 + }, + { + "epoch": 5.7, + "learning_rate": 1.2962044786555827e-07, + "loss": 0.9925, + "step": 77157 + }, + { + "epoch": 5.7, + "learning_rate": 1.2955644201044449e-07, + "loss": 0.9771, + "step": 77158 + }, + { + "epoch": 5.7, + "learning_rate": 1.2949245185906566e-07, + "loss": 0.9888, + "step": 77159 + }, + { + "epoch": 5.7, + "learning_rate": 1.2942847741152397e-07, + "loss": 1.0672, + "step": 77160 + }, + { + "epoch": 5.7, + "learning_rate": 1.2936451866791933e-07, + "loss": 0.9088, + "step": 77161 + }, + { + "epoch": 5.7, + "learning_rate": 1.2930057562835386e-07, + "loss": 1.0155, + "step": 77162 + }, + { + "epoch": 5.7, + "learning_rate": 1.292366482929308e-07, + "loss": 0.9102, + "step": 77163 + }, + { + "epoch": 5.7, + "learning_rate": 1.291727366617501e-07, + "loss": 1.0837, + "step": 77164 + }, + { + "epoch": 5.7, + "learning_rate": 1.2910884073491503e-07, + "loss": 0.9419, + "step": 77165 + }, + { + "epoch": 5.7, + "learning_rate": 1.2904496051252658e-07, + "loss": 1.0184, + "step": 77166 + }, + { + "epoch": 5.7, + "learning_rate": 1.2898109599468578e-07, + "loss": 0.9929, + "step": 77167 + }, + { + "epoch": 5.7, + "learning_rate": 1.289172471814959e-07, + "loss": 1.0566, + "step": 77168 + }, + { + "epoch": 5.7, + "learning_rate": 1.2885341407305685e-07, + "loss": 0.9765, + "step": 77169 + }, + { + "epoch": 5.7, + "learning_rate": 1.2878959666947077e-07, + "loss": 0.9925, + "step": 77170 + }, + { + "epoch": 5.7, + "learning_rate": 1.287257949708387e-07, + "loss": 1.0708, + "step": 77171 + }, + { + "epoch": 5.7, + "learning_rate": 1.28662008977265e-07, + "loss": 0.9193, + "step": 77172 + }, + { + "epoch": 5.7, + "learning_rate": 1.2859823868884625e-07, + "loss": 1.0378, + "step": 77173 + }, + { + "epoch": 5.7, + "learning_rate": 1.285344841056868e-07, + "loss": 0.9604, + "step": 77174 + }, + { + "epoch": 5.7, + "learning_rate": 1.2847074522788772e-07, + "loss": 1.0347, + "step": 77175 + }, + { + "epoch": 5.7, + "learning_rate": 1.2840702205555223e-07, + "loss": 0.9687, + "step": 77176 + }, + { + "epoch": 5.7, + "learning_rate": 1.2834331458877803e-07, + "loss": 1.0991, + "step": 77177 + }, + { + "epoch": 5.7, + "learning_rate": 1.2827962282766947e-07, + "loss": 0.9815, + "step": 77178 + }, + { + "epoch": 5.7, + "learning_rate": 1.2821594677232652e-07, + "loss": 0.9681, + "step": 77179 + }, + { + "epoch": 5.7, + "learning_rate": 1.2815228642285015e-07, + "loss": 1.0091, + "step": 77180 + }, + { + "epoch": 5.7, + "learning_rate": 1.2808864177934255e-07, + "loss": 0.9761, + "step": 77181 + }, + { + "epoch": 5.7, + "learning_rate": 1.2802501284190472e-07, + "loss": 1.0794, + "step": 77182 + }, + { + "epoch": 5.7, + "learning_rate": 1.2796139961063768e-07, + "loss": 1.0544, + "step": 77183 + }, + { + "epoch": 5.7, + "learning_rate": 1.2789780208564363e-07, + "loss": 0.9864, + "step": 77184 + }, + { + "epoch": 5.7, + "learning_rate": 1.278342202670224e-07, + "loss": 0.882, + "step": 77185 + }, + { + "epoch": 5.7, + "learning_rate": 1.277706541548751e-07, + "loss": 1.0556, + "step": 77186 + }, + { + "epoch": 5.7, + "learning_rate": 1.2770710374930384e-07, + "loss": 0.933, + "step": 77187 + }, + { + "epoch": 5.7, + "learning_rate": 1.2764356905040966e-07, + "loss": 1.015, + "step": 77188 + }, + { + "epoch": 5.7, + "learning_rate": 1.2758005005829355e-07, + "loss": 0.9132, + "step": 77189 + }, + { + "epoch": 5.7, + "learning_rate": 1.2751654677305548e-07, + "loss": 0.9163, + "step": 77190 + }, + { + "epoch": 5.7, + "learning_rate": 1.2745305919479757e-07, + "loss": 0.9226, + "step": 77191 + }, + { + "epoch": 5.7, + "learning_rate": 1.2738958732362084e-07, + "loss": 0.9707, + "step": 77192 + }, + { + "epoch": 5.7, + "learning_rate": 1.2732613115962633e-07, + "loss": 0.9837, + "step": 77193 + }, + { + "epoch": 5.7, + "learning_rate": 1.2726269070291398e-07, + "loss": 0.9917, + "step": 77194 + }, + { + "epoch": 5.7, + "learning_rate": 1.2719926595358478e-07, + "loss": 0.9974, + "step": 77195 + }, + { + "epoch": 5.7, + "learning_rate": 1.2713585691174202e-07, + "loss": 0.9422, + "step": 77196 + }, + { + "epoch": 5.7, + "learning_rate": 1.2707246357748337e-07, + "loss": 1.0086, + "step": 77197 + }, + { + "epoch": 5.7, + "learning_rate": 1.270090859509121e-07, + "loss": 0.9731, + "step": 77198 + }, + { + "epoch": 5.7, + "learning_rate": 1.2694572403212813e-07, + "loss": 0.9542, + "step": 77199 + }, + { + "epoch": 5.7, + "learning_rate": 1.2688237782123248e-07, + "loss": 0.9991, + "step": 77200 + }, + { + "epoch": 5.7, + "learning_rate": 1.2681904731832507e-07, + "loss": 1.0128, + "step": 77201 + }, + { + "epoch": 5.7, + "learning_rate": 1.2675573252350692e-07, + "loss": 1.0641, + "step": 77202 + }, + { + "epoch": 5.7, + "learning_rate": 1.266924334368802e-07, + "loss": 1.0154, + "step": 77203 + }, + { + "epoch": 5.7, + "learning_rate": 1.266291500585437e-07, + "loss": 1.0694, + "step": 77204 + }, + { + "epoch": 5.7, + "learning_rate": 1.2656588238859957e-07, + "loss": 0.9148, + "step": 77205 + }, + { + "epoch": 5.7, + "learning_rate": 1.2650263042714772e-07, + "loss": 0.9221, + "step": 77206 + }, + { + "epoch": 5.7, + "learning_rate": 1.2643939417428808e-07, + "loss": 0.8362, + "step": 77207 + }, + { + "epoch": 5.7, + "learning_rate": 1.263761736301239e-07, + "loss": 0.8953, + "step": 77208 + }, + { + "epoch": 5.7, + "learning_rate": 1.2631296879475285e-07, + "loss": 1.0292, + "step": 77209 + }, + { + "epoch": 5.7, + "learning_rate": 1.2624977966827602e-07, + "loss": 0.9284, + "step": 77210 + }, + { + "epoch": 5.71, + "learning_rate": 1.2618660625079438e-07, + "loss": 0.9544, + "step": 77211 + }, + { + "epoch": 5.71, + "learning_rate": 1.261234485424101e-07, + "loss": 1.0777, + "step": 77212 + }, + { + "epoch": 5.71, + "learning_rate": 1.2606030654322087e-07, + "loss": 1.0572, + "step": 77213 + }, + { + "epoch": 5.71, + "learning_rate": 1.2599718025332996e-07, + "loss": 0.8536, + "step": 77214 + }, + { + "epoch": 5.71, + "learning_rate": 1.2593406967283396e-07, + "loss": 0.9129, + "step": 77215 + }, + { + "epoch": 5.71, + "learning_rate": 1.258709748018372e-07, + "loss": 1.0153, + "step": 77216 + }, + { + "epoch": 5.71, + "learning_rate": 1.258078956404385e-07, + "loss": 1.0094, + "step": 77217 + }, + { + "epoch": 5.71, + "learning_rate": 1.2574483218873778e-07, + "loss": 1.0072, + "step": 77218 + }, + { + "epoch": 5.71, + "learning_rate": 1.256817844468361e-07, + "loss": 0.9443, + "step": 77219 + }, + { + "epoch": 5.71, + "learning_rate": 1.2561875241483222e-07, + "loss": 1.1276, + "step": 77220 + }, + { + "epoch": 5.71, + "learning_rate": 1.2555573609282945e-07, + "loss": 0.9965, + "step": 77221 + }, + { + "epoch": 5.71, + "learning_rate": 1.2549273548092433e-07, + "loss": 1.0218, + "step": 77222 + }, + { + "epoch": 5.71, + "learning_rate": 1.2542975057922013e-07, + "loss": 0.9055, + "step": 77223 + }, + { + "epoch": 5.71, + "learning_rate": 1.2536678138781455e-07, + "loss": 0.9624, + "step": 77224 + }, + { + "epoch": 5.71, + "learning_rate": 1.2530382790681084e-07, + "loss": 0.9853, + "step": 77225 + }, + { + "epoch": 5.71, + "learning_rate": 1.252408901363078e-07, + "loss": 0.9812, + "step": 77226 + }, + { + "epoch": 5.71, + "learning_rate": 1.2517796807640314e-07, + "loss": 0.8793, + "step": 77227 + }, + { + "epoch": 5.71, + "learning_rate": 1.2511506172720013e-07, + "loss": 1.018, + "step": 77228 + }, + { + "epoch": 5.71, + "learning_rate": 1.2505217108879863e-07, + "loss": 0.9396, + "step": 77229 + }, + { + "epoch": 5.71, + "learning_rate": 1.249892961612964e-07, + "loss": 1.0141, + "step": 77230 + }, + { + "epoch": 5.71, + "learning_rate": 1.2492643694479557e-07, + "loss": 0.9701, + "step": 77231 + }, + { + "epoch": 5.71, + "learning_rate": 1.2486359343939602e-07, + "loss": 1.136, + "step": 77232 + }, + { + "epoch": 5.71, + "learning_rate": 1.2480076564519662e-07, + "loss": 1.0525, + "step": 77233 + }, + { + "epoch": 5.71, + "learning_rate": 1.2473795356229724e-07, + "loss": 0.9511, + "step": 77234 + }, + { + "epoch": 5.71, + "learning_rate": 1.2467515719080003e-07, + "loss": 0.9643, + "step": 77235 + }, + { + "epoch": 5.71, + "learning_rate": 1.246123765308016e-07, + "loss": 1.1025, + "step": 77236 + }, + { + "epoch": 5.71, + "learning_rate": 1.245496115824052e-07, + "loss": 1.0834, + "step": 77237 + }, + { + "epoch": 5.71, + "learning_rate": 1.244868623457085e-07, + "loss": 0.9587, + "step": 77238 + }, + { + "epoch": 5.71, + "learning_rate": 1.2442412882081144e-07, + "loss": 1.0296, + "step": 77239 + }, + { + "epoch": 5.71, + "learning_rate": 1.2436141100781395e-07, + "loss": 0.9966, + "step": 77240 + }, + { + "epoch": 5.71, + "learning_rate": 1.2429870890681817e-07, + "loss": 0.9706, + "step": 77241 + }, + { + "epoch": 5.71, + "learning_rate": 1.2423602251791954e-07, + "loss": 1.0394, + "step": 77242 + }, + { + "epoch": 5.71, + "learning_rate": 1.2417335184122136e-07, + "loss": 0.9534, + "step": 77243 + }, + { + "epoch": 5.71, + "learning_rate": 1.2411069687682132e-07, + "loss": 0.9252, + "step": 77244 + }, + { + "epoch": 5.71, + "learning_rate": 1.240480576248204e-07, + "loss": 1.0366, + "step": 77245 + }, + { + "epoch": 5.71, + "learning_rate": 1.239854340853186e-07, + "loss": 0.9756, + "step": 77246 + }, + { + "epoch": 5.71, + "learning_rate": 1.2392282625841246e-07, + "loss": 0.9021, + "step": 77247 + }, + { + "epoch": 5.71, + "learning_rate": 1.2386023414420522e-07, + "loss": 0.9831, + "step": 77248 + }, + { + "epoch": 5.71, + "learning_rate": 1.2379765774279573e-07, + "loss": 0.8969, + "step": 77249 + }, + { + "epoch": 5.71, + "learning_rate": 1.2373509705428165e-07, + "loss": 0.9578, + "step": 77250 + }, + { + "epoch": 5.71, + "learning_rate": 1.2367255207876405e-07, + "loss": 1.0758, + "step": 77251 + }, + { + "epoch": 5.71, + "learning_rate": 1.2361002281634172e-07, + "loss": 0.9565, + "step": 77252 + }, + { + "epoch": 5.71, + "learning_rate": 1.2354750926711457e-07, + "loss": 1.0453, + "step": 77253 + }, + { + "epoch": 5.71, + "learning_rate": 1.2348501143118253e-07, + "loss": 1.0859, + "step": 77254 + }, + { + "epoch": 5.71, + "learning_rate": 1.234225293086455e-07, + "loss": 0.933, + "step": 77255 + }, + { + "epoch": 5.71, + "learning_rate": 1.2336006289960013e-07, + "loss": 1.0183, + "step": 77256 + }, + { + "epoch": 5.71, + "learning_rate": 1.2329761220414848e-07, + "loss": 0.9114, + "step": 77257 + }, + { + "epoch": 5.71, + "learning_rate": 1.2323517722238833e-07, + "loss": 0.9896, + "step": 77258 + }, + { + "epoch": 5.71, + "learning_rate": 1.2317275795442064e-07, + "loss": 0.9872, + "step": 77259 + }, + { + "epoch": 5.71, + "learning_rate": 1.2311035440034314e-07, + "loss": 1.0657, + "step": 77260 + }, + { + "epoch": 5.71, + "learning_rate": 1.2304796656025574e-07, + "loss": 1.0708, + "step": 77261 + }, + { + "epoch": 5.71, + "learning_rate": 1.2298559443425838e-07, + "loss": 1.0156, + "step": 77262 + }, + { + "epoch": 5.71, + "learning_rate": 1.2292323802244877e-07, + "loss": 0.968, + "step": 77263 + }, + { + "epoch": 5.71, + "learning_rate": 1.228608973249279e-07, + "loss": 0.9996, + "step": 77264 + }, + { + "epoch": 5.71, + "learning_rate": 1.2279857234179237e-07, + "loss": 0.8821, + "step": 77265 + }, + { + "epoch": 5.71, + "learning_rate": 1.2273626307314545e-07, + "loss": 1.0168, + "step": 77266 + }, + { + "epoch": 5.71, + "learning_rate": 1.226739695190815e-07, + "loss": 0.8983, + "step": 77267 + }, + { + "epoch": 5.71, + "learning_rate": 1.2261169167970265e-07, + "loss": 1.157, + "step": 77268 + }, + { + "epoch": 5.71, + "learning_rate": 1.2254942955510774e-07, + "loss": 1.0087, + "step": 77269 + }, + { + "epoch": 5.71, + "learning_rate": 1.2248718314539554e-07, + "loss": 1.0075, + "step": 77270 + }, + { + "epoch": 5.71, + "learning_rate": 1.2242495245066377e-07, + "loss": 0.8825, + "step": 77271 + }, + { + "epoch": 5.71, + "learning_rate": 1.2236273747101347e-07, + "loss": 1.0425, + "step": 77272 + }, + { + "epoch": 5.71, + "learning_rate": 1.2230053820654342e-07, + "loss": 1.0652, + "step": 77273 + }, + { + "epoch": 5.71, + "learning_rate": 1.2223835465735024e-07, + "loss": 1.097, + "step": 77274 + }, + { + "epoch": 5.71, + "learning_rate": 1.2217618682353605e-07, + "loss": 0.8524, + "step": 77275 + }, + { + "epoch": 5.71, + "learning_rate": 1.2211403470519635e-07, + "loss": 1.123, + "step": 77276 + }, + { + "epoch": 5.71, + "learning_rate": 1.2205189830243326e-07, + "loss": 0.9326, + "step": 77277 + }, + { + "epoch": 5.71, + "learning_rate": 1.2198977761534448e-07, + "loss": 1.0768, + "step": 77278 + }, + { + "epoch": 5.71, + "learning_rate": 1.2192767264402883e-07, + "loss": 0.9788, + "step": 77279 + }, + { + "epoch": 5.71, + "learning_rate": 1.2186558338858289e-07, + "loss": 0.9378, + "step": 77280 + }, + { + "epoch": 5.71, + "learning_rate": 1.2180350984910994e-07, + "loss": 1.0135, + "step": 77281 + }, + { + "epoch": 5.71, + "learning_rate": 1.2174145202570541e-07, + "loss": 1.0327, + "step": 77282 + }, + { + "epoch": 5.71, + "learning_rate": 1.2167940991846817e-07, + "loss": 0.9554, + "step": 77283 + }, + { + "epoch": 5.71, + "learning_rate": 1.216173835274981e-07, + "loss": 1.01, + "step": 77284 + }, + { + "epoch": 5.71, + "learning_rate": 1.2155537285289288e-07, + "loss": 1.1387, + "step": 77285 + }, + { + "epoch": 5.71, + "learning_rate": 1.2149337789475357e-07, + "loss": 1.0127, + "step": 77286 + }, + { + "epoch": 5.71, + "learning_rate": 1.2143139865317565e-07, + "loss": 1.0682, + "step": 77287 + }, + { + "epoch": 5.71, + "learning_rate": 1.2136943512825905e-07, + "loss": 1.0341, + "step": 77288 + }, + { + "epoch": 5.71, + "learning_rate": 1.2130748732010144e-07, + "loss": 0.8617, + "step": 77289 + }, + { + "epoch": 5.71, + "learning_rate": 1.2124555522880387e-07, + "loss": 1.0165, + "step": 77290 + }, + { + "epoch": 5.71, + "learning_rate": 1.2118363885446293e-07, + "loss": 1.0237, + "step": 77291 + }, + { + "epoch": 5.71, + "learning_rate": 1.2112173819717632e-07, + "loss": 1.0355, + "step": 77292 + }, + { + "epoch": 5.71, + "learning_rate": 1.2105985325704506e-07, + "loss": 0.964, + "step": 77293 + }, + { + "epoch": 5.71, + "learning_rate": 1.2099798403416574e-07, + "loss": 0.9119, + "step": 77294 + }, + { + "epoch": 5.71, + "learning_rate": 1.2093613052863717e-07, + "loss": 0.9543, + "step": 77295 + }, + { + "epoch": 5.71, + "learning_rate": 1.2087429274055818e-07, + "loss": 1.0716, + "step": 77296 + }, + { + "epoch": 5.71, + "learning_rate": 1.2081247067002643e-07, + "loss": 0.9931, + "step": 77297 + }, + { + "epoch": 5.71, + "learning_rate": 1.2075066431714077e-07, + "loss": 1.0888, + "step": 77298 + }, + { + "epoch": 5.71, + "learning_rate": 1.20688873682e-07, + "loss": 0.9835, + "step": 77299 + }, + { + "epoch": 5.71, + "learning_rate": 1.2062709876470068e-07, + "loss": 1.1161, + "step": 77300 + }, + { + "epoch": 5.71, + "learning_rate": 1.205653395653439e-07, + "loss": 0.9552, + "step": 77301 + }, + { + "epoch": 5.71, + "learning_rate": 1.2050359608402505e-07, + "loss": 1.0244, + "step": 77302 + }, + { + "epoch": 5.71, + "learning_rate": 1.2044186832084415e-07, + "loss": 0.9856, + "step": 77303 + }, + { + "epoch": 5.71, + "learning_rate": 1.2038015627589883e-07, + "loss": 1.0125, + "step": 77304 + }, + { + "epoch": 5.71, + "learning_rate": 1.2031845994928682e-07, + "loss": 0.9896, + "step": 77305 + }, + { + "epoch": 5.71, + "learning_rate": 1.20256779341108e-07, + "loss": 0.8183, + "step": 77306 + }, + { + "epoch": 5.71, + "learning_rate": 1.2019511445145794e-07, + "loss": 0.8571, + "step": 77307 + }, + { + "epoch": 5.71, + "learning_rate": 1.2013346528043757e-07, + "loss": 0.9567, + "step": 77308 + }, + { + "epoch": 5.71, + "learning_rate": 1.200718318281424e-07, + "loss": 1.0777, + "step": 77309 + }, + { + "epoch": 5.71, + "learning_rate": 1.200102140946713e-07, + "loss": 1.2442, + "step": 77310 + }, + { + "epoch": 5.71, + "learning_rate": 1.1994861208012408e-07, + "loss": 1.0511, + "step": 77311 + }, + { + "epoch": 5.71, + "learning_rate": 1.1988702578459522e-07, + "loss": 0.9441, + "step": 77312 + }, + { + "epoch": 5.71, + "learning_rate": 1.198254552081868e-07, + "loss": 0.922, + "step": 77313 + }, + { + "epoch": 5.71, + "learning_rate": 1.1976390035099316e-07, + "loss": 0.9845, + "step": 77314 + }, + { + "epoch": 5.71, + "learning_rate": 1.197023612131154e-07, + "loss": 1.086, + "step": 77315 + }, + { + "epoch": 5.71, + "learning_rate": 1.1964083779464785e-07, + "loss": 1.0769, + "step": 77316 + }, + { + "epoch": 5.71, + "learning_rate": 1.1957933009569156e-07, + "loss": 0.9237, + "step": 77317 + }, + { + "epoch": 5.71, + "learning_rate": 1.195178381163431e-07, + "loss": 1.024, + "step": 77318 + }, + { + "epoch": 5.71, + "learning_rate": 1.1945636185670018e-07, + "loss": 0.9838, + "step": 77319 + }, + { + "epoch": 5.71, + "learning_rate": 1.193949013168605e-07, + "loss": 1.0416, + "step": 77320 + }, + { + "epoch": 5.71, + "learning_rate": 1.1933345649692285e-07, + "loss": 0.8361, + "step": 77321 + }, + { + "epoch": 5.71, + "learning_rate": 1.1927202739698386e-07, + "loss": 1.0314, + "step": 77322 + }, + { + "epoch": 5.71, + "learning_rate": 1.192106140171412e-07, + "loss": 1.0547, + "step": 77323 + }, + { + "epoch": 5.71, + "learning_rate": 1.1914921635749476e-07, + "loss": 0.9469, + "step": 77324 + }, + { + "epoch": 5.71, + "learning_rate": 1.1908783441813899e-07, + "loss": 0.926, + "step": 77325 + }, + { + "epoch": 5.71, + "learning_rate": 1.1902646819917374e-07, + "loss": 0.9597, + "step": 77326 + }, + { + "epoch": 5.71, + "learning_rate": 1.1896511770069563e-07, + "loss": 1.0886, + "step": 77327 + }, + { + "epoch": 5.71, + "learning_rate": 1.1890378292280347e-07, + "loss": 0.9634, + "step": 77328 + }, + { + "epoch": 5.71, + "learning_rate": 1.1884246386559272e-07, + "loss": 0.97, + "step": 77329 + }, + { + "epoch": 5.71, + "learning_rate": 1.1878116052916222e-07, + "loss": 1.0575, + "step": 77330 + }, + { + "epoch": 5.71, + "learning_rate": 1.1871987291361186e-07, + "loss": 1.0558, + "step": 77331 + }, + { + "epoch": 5.71, + "learning_rate": 1.1865860101903381e-07, + "loss": 0.9563, + "step": 77332 + }, + { + "epoch": 5.71, + "learning_rate": 1.1859734484553021e-07, + "loss": 1.0292, + "step": 77333 + }, + { + "epoch": 5.71, + "learning_rate": 1.1853610439319652e-07, + "loss": 1.0634, + "step": 77334 + }, + { + "epoch": 5.71, + "learning_rate": 1.1847487966213156e-07, + "loss": 1.011, + "step": 77335 + }, + { + "epoch": 5.71, + "learning_rate": 1.184136706524297e-07, + "loss": 0.9466, + "step": 77336 + }, + { + "epoch": 5.71, + "learning_rate": 1.1835247736419198e-07, + "loss": 1.0615, + "step": 77337 + }, + { + "epoch": 5.71, + "learning_rate": 1.1829129979751275e-07, + "loss": 0.9749, + "step": 77338 + }, + { + "epoch": 5.71, + "learning_rate": 1.1823013795249194e-07, + "loss": 1.0872, + "step": 77339 + }, + { + "epoch": 5.71, + "learning_rate": 1.1816899182922503e-07, + "loss": 0.9665, + "step": 77340 + }, + { + "epoch": 5.71, + "learning_rate": 1.1810786142780973e-07, + "loss": 1.0145, + "step": 77341 + }, + { + "epoch": 5.71, + "learning_rate": 1.180467467483426e-07, + "loss": 1.0065, + "step": 77342 + }, + { + "epoch": 5.71, + "learning_rate": 1.1798564779092359e-07, + "loss": 1.0399, + "step": 77343 + }, + { + "epoch": 5.71, + "learning_rate": 1.1792456455564705e-07, + "loss": 0.9728, + "step": 77344 + }, + { + "epoch": 5.71, + "learning_rate": 1.1786349704261068e-07, + "loss": 0.8946, + "step": 77345 + }, + { + "epoch": 5.72, + "learning_rate": 1.178024452519122e-07, + "loss": 1.008, + "step": 77346 + }, + { + "epoch": 5.72, + "learning_rate": 1.1774140918364929e-07, + "loss": 0.9524, + "step": 77347 + }, + { + "epoch": 5.72, + "learning_rate": 1.1768038883791855e-07, + "loss": 1.0532, + "step": 77348 + }, + { + "epoch": 5.72, + "learning_rate": 1.1761938421481656e-07, + "loss": 1.0753, + "step": 77349 + }, + { + "epoch": 5.72, + "learning_rate": 1.1755839531443991e-07, + "loss": 1.0534, + "step": 77350 + }, + { + "epoch": 5.72, + "learning_rate": 1.1749742213688741e-07, + "loss": 0.9301, + "step": 77351 + }, + { + "epoch": 5.72, + "learning_rate": 1.1743646468225455e-07, + "loss": 1.0977, + "step": 77352 + }, + { + "epoch": 5.72, + "learning_rate": 1.1737552295064014e-07, + "loss": 1.008, + "step": 77353 + }, + { + "epoch": 5.72, + "learning_rate": 1.1731459694213854e-07, + "loss": 0.9987, + "step": 77354 + }, + { + "epoch": 5.72, + "learning_rate": 1.1725368665684856e-07, + "loss": 0.9703, + "step": 77355 + }, + { + "epoch": 5.72, + "learning_rate": 1.1719279209486679e-07, + "loss": 0.963, + "step": 77356 + }, + { + "epoch": 5.72, + "learning_rate": 1.171319132562887e-07, + "loss": 1.0663, + "step": 77357 + }, + { + "epoch": 5.72, + "learning_rate": 1.1707105014121422e-07, + "loss": 1.0418, + "step": 77358 + }, + { + "epoch": 5.72, + "learning_rate": 1.1701020274973662e-07, + "loss": 1.0202, + "step": 77359 + }, + { + "epoch": 5.72, + "learning_rate": 1.169493710819558e-07, + "loss": 1.0033, + "step": 77360 + }, + { + "epoch": 5.72, + "learning_rate": 1.1688855513796615e-07, + "loss": 1.0036, + "step": 77361 + }, + { + "epoch": 5.72, + "learning_rate": 1.1682775491786535e-07, + "loss": 1.0431, + "step": 77362 + }, + { + "epoch": 5.72, + "learning_rate": 1.1676697042175e-07, + "loss": 0.9384, + "step": 77363 + }, + { + "epoch": 5.72, + "learning_rate": 1.167062016497178e-07, + "loss": 0.9135, + "step": 77364 + }, + { + "epoch": 5.72, + "learning_rate": 1.1664544860186422e-07, + "loss": 0.9465, + "step": 77365 + }, + { + "epoch": 5.72, + "learning_rate": 1.1658471127828585e-07, + "loss": 0.949, + "step": 77366 + }, + { + "epoch": 5.72, + "learning_rate": 1.165239896790804e-07, + "loss": 1.0192, + "step": 77367 + }, + { + "epoch": 5.72, + "learning_rate": 1.1646328380434447e-07, + "loss": 1.0514, + "step": 77368 + }, + { + "epoch": 5.72, + "learning_rate": 1.164025936541735e-07, + "loss": 1.0504, + "step": 77369 + }, + { + "epoch": 5.72, + "learning_rate": 1.1634191922866411e-07, + "loss": 0.997, + "step": 77370 + }, + { + "epoch": 5.72, + "learning_rate": 1.16281260527914e-07, + "loss": 0.9348, + "step": 77371 + }, + { + "epoch": 5.72, + "learning_rate": 1.1622061755201863e-07, + "loss": 0.9389, + "step": 77372 + }, + { + "epoch": 5.72, + "learning_rate": 1.1615999030107461e-07, + "loss": 1.0499, + "step": 77373 + }, + { + "epoch": 5.72, + "learning_rate": 1.1609937877517851e-07, + "loss": 1.0676, + "step": 77374 + }, + { + "epoch": 5.72, + "learning_rate": 1.1603878297442805e-07, + "loss": 0.9653, + "step": 77375 + }, + { + "epoch": 5.72, + "learning_rate": 1.1597820289891759e-07, + "loss": 1.0497, + "step": 77376 + }, + { + "epoch": 5.72, + "learning_rate": 1.1591763854874483e-07, + "loss": 0.9938, + "step": 77377 + }, + { + "epoch": 5.72, + "learning_rate": 1.1585708992400523e-07, + "loss": 1.1134, + "step": 77378 + }, + { + "epoch": 5.72, + "learning_rate": 1.1579655702479542e-07, + "loss": 1.0138, + "step": 77379 + }, + { + "epoch": 5.72, + "learning_rate": 1.1573603985121306e-07, + "loss": 0.9963, + "step": 77380 + }, + { + "epoch": 5.72, + "learning_rate": 1.1567553840335144e-07, + "loss": 1.0866, + "step": 77381 + }, + { + "epoch": 5.72, + "learning_rate": 1.1561505268130935e-07, + "loss": 0.9725, + "step": 77382 + }, + { + "epoch": 5.72, + "learning_rate": 1.1555458268518227e-07, + "loss": 1.0253, + "step": 77383 + }, + { + "epoch": 5.72, + "learning_rate": 1.154941284150668e-07, + "loss": 1.13, + "step": 77384 + }, + { + "epoch": 5.72, + "learning_rate": 1.1543368987105952e-07, + "loss": 0.9758, + "step": 77385 + }, + { + "epoch": 5.72, + "learning_rate": 1.1537326705325369e-07, + "loss": 0.8646, + "step": 77386 + }, + { + "epoch": 5.72, + "learning_rate": 1.1531285996174924e-07, + "loss": 1.1232, + "step": 77387 + }, + { + "epoch": 5.72, + "learning_rate": 1.1525246859664053e-07, + "loss": 0.9001, + "step": 77388 + }, + { + "epoch": 5.72, + "learning_rate": 1.1519209295802303e-07, + "loss": 1.0164, + "step": 77389 + }, + { + "epoch": 5.72, + "learning_rate": 1.1513173304599335e-07, + "loss": 0.9571, + "step": 77390 + }, + { + "epoch": 5.72, + "learning_rate": 1.1507138886064807e-07, + "loss": 0.8875, + "step": 77391 + }, + { + "epoch": 5.72, + "learning_rate": 1.1501106040208265e-07, + "loss": 1.0162, + "step": 77392 + }, + { + "epoch": 5.72, + "learning_rate": 1.1495074767039372e-07, + "loss": 0.9645, + "step": 77393 + }, + { + "epoch": 5.72, + "learning_rate": 1.1489045066567672e-07, + "loss": 0.8992, + "step": 77394 + }, + { + "epoch": 5.72, + "learning_rate": 1.1483016938802605e-07, + "loss": 1.0452, + "step": 77395 + }, + { + "epoch": 5.72, + "learning_rate": 1.147699038375405e-07, + "loss": 1.1571, + "step": 77396 + }, + { + "epoch": 5.72, + "learning_rate": 1.1470965401431445e-07, + "loss": 1.0196, + "step": 77397 + }, + { + "epoch": 5.72, + "learning_rate": 1.1464941991844336e-07, + "loss": 0.9676, + "step": 77398 + }, + { + "epoch": 5.72, + "learning_rate": 1.1458920155002384e-07, + "loss": 1.0011, + "step": 77399 + }, + { + "epoch": 5.72, + "learning_rate": 1.1452899890915248e-07, + "loss": 1.1084, + "step": 77400 + }, + { + "epoch": 5.72, + "learning_rate": 1.1446881199592253e-07, + "loss": 0.9962, + "step": 77401 + }, + { + "epoch": 5.72, + "learning_rate": 1.1440864081043279e-07, + "loss": 0.9099, + "step": 77402 + }, + { + "epoch": 5.72, + "learning_rate": 1.1434848535277654e-07, + "loss": 1.0217, + "step": 77403 + }, + { + "epoch": 5.72, + "learning_rate": 1.1428834562305036e-07, + "loss": 1.039, + "step": 77404 + }, + { + "epoch": 5.72, + "learning_rate": 1.1422822162135083e-07, + "loss": 0.9157, + "step": 77405 + }, + { + "epoch": 5.72, + "learning_rate": 1.1416811334777122e-07, + "loss": 0.9868, + "step": 77406 + }, + { + "epoch": 5.72, + "learning_rate": 1.1410802080240924e-07, + "loss": 1.0067, + "step": 77407 + }, + { + "epoch": 5.72, + "learning_rate": 1.1404794398536034e-07, + "loss": 0.9676, + "step": 77408 + }, + { + "epoch": 5.72, + "learning_rate": 1.1398788289672003e-07, + "loss": 0.863, + "step": 77409 + }, + { + "epoch": 5.72, + "learning_rate": 1.1392783753658266e-07, + "loss": 1.0062, + "step": 77410 + }, + { + "epoch": 5.72, + "learning_rate": 1.1386780790504593e-07, + "loss": 1.0038, + "step": 77411 + }, + { + "epoch": 5.72, + "learning_rate": 1.138077940022031e-07, + "loss": 0.959, + "step": 77412 + }, + { + "epoch": 5.72, + "learning_rate": 1.1374779582815076e-07, + "loss": 1.0182, + "step": 77413 + }, + { + "epoch": 5.72, + "learning_rate": 1.136878133829844e-07, + "loss": 0.9702, + "step": 77414 + }, + { + "epoch": 5.72, + "learning_rate": 1.1362784666679838e-07, + "loss": 1.0896, + "step": 77415 + }, + { + "epoch": 5.72, + "learning_rate": 1.135678956796904e-07, + "loss": 1.0891, + "step": 77416 + }, + { + "epoch": 5.72, + "learning_rate": 1.1350796042175372e-07, + "loss": 1.0618, + "step": 77417 + }, + { + "epoch": 5.72, + "learning_rate": 1.1344804089308492e-07, + "loss": 1.0997, + "step": 77418 + }, + { + "epoch": 5.72, + "learning_rate": 1.1338813709377727e-07, + "loss": 1.0439, + "step": 77419 + }, + { + "epoch": 5.72, + "learning_rate": 1.1332824902392958e-07, + "loss": 1.0198, + "step": 77420 + }, + { + "epoch": 5.72, + "learning_rate": 1.13268376683634e-07, + "loss": 1.0727, + "step": 77421 + }, + { + "epoch": 5.72, + "learning_rate": 1.1320852007298822e-07, + "loss": 0.9659, + "step": 77422 + }, + { + "epoch": 5.72, + "learning_rate": 1.131486791920855e-07, + "loss": 1.0962, + "step": 77423 + }, + { + "epoch": 5.72, + "learning_rate": 1.1308885404102132e-07, + "loss": 0.8979, + "step": 77424 + }, + { + "epoch": 5.72, + "learning_rate": 1.1302904461989227e-07, + "loss": 0.8567, + "step": 77425 + }, + { + "epoch": 5.72, + "learning_rate": 1.1296925092879163e-07, + "loss": 1.0091, + "step": 77426 + }, + { + "epoch": 5.72, + "learning_rate": 1.1290947296781484e-07, + "loss": 0.9509, + "step": 77427 + }, + { + "epoch": 5.72, + "learning_rate": 1.1284971073705853e-07, + "loss": 1.0116, + "step": 77428 + }, + { + "epoch": 5.72, + "learning_rate": 1.1278996423661815e-07, + "loss": 1.0003, + "step": 77429 + }, + { + "epoch": 5.72, + "learning_rate": 1.1273023346658474e-07, + "loss": 1.0588, + "step": 77430 + }, + { + "epoch": 5.72, + "learning_rate": 1.1267051842705822e-07, + "loss": 1.0482, + "step": 77431 + }, + { + "epoch": 5.72, + "learning_rate": 1.1261081911813077e-07, + "loss": 1.0494, + "step": 77432 + }, + { + "epoch": 5.72, + "learning_rate": 1.1255113553989782e-07, + "loss": 1.0706, + "step": 77433 + }, + { + "epoch": 5.72, + "learning_rate": 1.124914676924549e-07, + "loss": 0.9382, + "step": 77434 + }, + { + "epoch": 5.72, + "learning_rate": 1.1243181557589633e-07, + "loss": 1.0418, + "step": 77435 + }, + { + "epoch": 5.72, + "learning_rate": 1.1237217919031762e-07, + "loss": 0.9682, + "step": 77436 + }, + { + "epoch": 5.72, + "learning_rate": 1.1231255853581313e-07, + "loss": 1.1052, + "step": 77437 + }, + { + "epoch": 5.72, + "learning_rate": 1.1225295361247835e-07, + "loss": 1.0368, + "step": 77438 + }, + { + "epoch": 5.72, + "learning_rate": 1.1219336442040651e-07, + "loss": 1.1194, + "step": 77439 + }, + { + "epoch": 5.72, + "learning_rate": 1.1213379095969423e-07, + "loss": 0.9732, + "step": 77440 + }, + { + "epoch": 5.72, + "learning_rate": 1.1207423323043587e-07, + "loss": 0.9913, + "step": 77441 + }, + { + "epoch": 5.72, + "learning_rate": 1.1201469123272579e-07, + "loss": 1.0873, + "step": 77442 + }, + { + "epoch": 5.72, + "learning_rate": 1.1195516496665836e-07, + "loss": 0.9995, + "step": 77443 + }, + { + "epoch": 5.72, + "learning_rate": 1.1189565443232908e-07, + "loss": 1.0955, + "step": 77444 + }, + { + "epoch": 5.72, + "learning_rate": 1.118361596298323e-07, + "loss": 1.0409, + "step": 77445 + }, + { + "epoch": 5.72, + "learning_rate": 1.1177668055926238e-07, + "loss": 1.0821, + "step": 77446 + }, + { + "epoch": 5.72, + "learning_rate": 1.1171721722071483e-07, + "loss": 1.0378, + "step": 77447 + }, + { + "epoch": 5.72, + "learning_rate": 1.1165776961428398e-07, + "loss": 1.1126, + "step": 77448 + }, + { + "epoch": 5.72, + "learning_rate": 1.1159833774006424e-07, + "loss": 0.9481, + "step": 77449 + }, + { + "epoch": 5.72, + "learning_rate": 1.1153892159814994e-07, + "loss": 1.046, + "step": 77450 + }, + { + "epoch": 5.72, + "learning_rate": 1.1147952118863437e-07, + "loss": 0.9776, + "step": 77451 + }, + { + "epoch": 5.72, + "learning_rate": 1.114201365116141e-07, + "loss": 1.055, + "step": 77452 + }, + { + "epoch": 5.72, + "learning_rate": 1.1136076756718351e-07, + "loss": 0.9746, + "step": 77453 + }, + { + "epoch": 5.72, + "learning_rate": 1.1130141435543696e-07, + "loss": 0.9605, + "step": 77454 + }, + { + "epoch": 5.72, + "learning_rate": 1.112420768764666e-07, + "loss": 0.9401, + "step": 77455 + }, + { + "epoch": 5.72, + "learning_rate": 1.1118275513037013e-07, + "loss": 1.0371, + "step": 77456 + }, + { + "epoch": 5.72, + "learning_rate": 1.1112344911723972e-07, + "loss": 1.0557, + "step": 77457 + }, + { + "epoch": 5.72, + "learning_rate": 1.1106415883717081e-07, + "loss": 0.8998, + "step": 77458 + }, + { + "epoch": 5.72, + "learning_rate": 1.1100488429025668e-07, + "loss": 0.976, + "step": 77459 + }, + { + "epoch": 5.72, + "learning_rate": 1.1094562547659282e-07, + "loss": 0.9407, + "step": 77460 + }, + { + "epoch": 5.72, + "learning_rate": 1.1088638239627248e-07, + "loss": 0.9228, + "step": 77461 + }, + { + "epoch": 5.72, + "learning_rate": 1.1082715504939112e-07, + "loss": 0.9438, + "step": 77462 + }, + { + "epoch": 5.72, + "learning_rate": 1.1076794343604202e-07, + "loss": 1.0298, + "step": 77463 + }, + { + "epoch": 5.72, + "learning_rate": 1.1070874755631955e-07, + "loss": 0.9725, + "step": 77464 + }, + { + "epoch": 5.72, + "learning_rate": 1.1064956741031806e-07, + "loss": 0.929, + "step": 77465 + }, + { + "epoch": 5.72, + "learning_rate": 1.1059040299813084e-07, + "loss": 0.9841, + "step": 77466 + }, + { + "epoch": 5.72, + "learning_rate": 1.1053125431985446e-07, + "loss": 0.9442, + "step": 77467 + }, + { + "epoch": 5.72, + "learning_rate": 1.1047212137557994e-07, + "loss": 1.0088, + "step": 77468 + }, + { + "epoch": 5.72, + "learning_rate": 1.104130041654028e-07, + "loss": 0.8903, + "step": 77469 + }, + { + "epoch": 5.72, + "learning_rate": 1.1035390268941848e-07, + "loss": 1.0363, + "step": 77470 + }, + { + "epoch": 5.72, + "learning_rate": 1.1029481694771804e-07, + "loss": 1.0445, + "step": 77471 + }, + { + "epoch": 5.72, + "learning_rate": 1.1023574694039807e-07, + "loss": 0.9726, + "step": 77472 + }, + { + "epoch": 5.72, + "learning_rate": 1.101766926675496e-07, + "loss": 0.9964, + "step": 77473 + }, + { + "epoch": 5.72, + "learning_rate": 1.1011765412927145e-07, + "loss": 0.9286, + "step": 77474 + }, + { + "epoch": 5.72, + "learning_rate": 1.1005863132565242e-07, + "loss": 1.0179, + "step": 77475 + }, + { + "epoch": 5.72, + "learning_rate": 1.0999962425678912e-07, + "loss": 1.1142, + "step": 77476 + }, + { + "epoch": 5.72, + "learning_rate": 1.099406329227748e-07, + "loss": 0.9807, + "step": 77477 + }, + { + "epoch": 5.72, + "learning_rate": 1.0988165732370492e-07, + "loss": 1.0074, + "step": 77478 + }, + { + "epoch": 5.72, + "learning_rate": 1.0982269745967056e-07, + "loss": 1.0035, + "step": 77479 + }, + { + "epoch": 5.72, + "learning_rate": 1.0976375333076605e-07, + "loss": 0.8929, + "step": 77480 + }, + { + "epoch": 5.73, + "learning_rate": 1.097048249370869e-07, + "loss": 1.0053, + "step": 77481 + }, + { + "epoch": 5.73, + "learning_rate": 1.0964591227872634e-07, + "loss": 1.0319, + "step": 77482 + }, + { + "epoch": 5.73, + "learning_rate": 1.0958701535577654e-07, + "loss": 0.947, + "step": 77483 + }, + { + "epoch": 5.73, + "learning_rate": 1.0952813416833297e-07, + "loss": 0.9964, + "step": 77484 + }, + { + "epoch": 5.73, + "learning_rate": 1.0946926871648777e-07, + "loss": 0.957, + "step": 77485 + }, + { + "epoch": 5.73, + "learning_rate": 1.0941041900033644e-07, + "loss": 1.0862, + "step": 77486 + }, + { + "epoch": 5.73, + "learning_rate": 1.0935158501997112e-07, + "loss": 0.9684, + "step": 77487 + }, + { + "epoch": 5.73, + "learning_rate": 1.0929276677548617e-07, + "loss": 0.9802, + "step": 77488 + }, + { + "epoch": 5.73, + "learning_rate": 1.0923396426697486e-07, + "loss": 0.9959, + "step": 77489 + }, + { + "epoch": 5.73, + "learning_rate": 1.0917517749453044e-07, + "loss": 0.9638, + "step": 77490 + }, + { + "epoch": 5.73, + "learning_rate": 1.0911640645824729e-07, + "loss": 1.0385, + "step": 77491 + }, + { + "epoch": 5.73, + "learning_rate": 1.0905765115821865e-07, + "loss": 0.9138, + "step": 77492 + }, + { + "epoch": 5.73, + "learning_rate": 1.0899891159453668e-07, + "loss": 0.9705, + "step": 77493 + }, + { + "epoch": 5.73, + "learning_rate": 1.0894018776729798e-07, + "loss": 1.0862, + "step": 77494 + }, + { + "epoch": 5.73, + "learning_rate": 1.0888147967659135e-07, + "loss": 0.994, + "step": 77495 + }, + { + "epoch": 5.73, + "learning_rate": 1.088227873225145e-07, + "loss": 0.9753, + "step": 77496 + }, + { + "epoch": 5.73, + "learning_rate": 1.0876411070515847e-07, + "loss": 0.9854, + "step": 77497 + }, + { + "epoch": 5.73, + "learning_rate": 1.0870544982461762e-07, + "loss": 0.9558, + "step": 77498 + }, + { + "epoch": 5.73, + "learning_rate": 1.086468046809841e-07, + "loss": 0.9525, + "step": 77499 + }, + { + "epoch": 5.73, + "learning_rate": 1.0858817527435228e-07, + "loss": 0.9139, + "step": 77500 + }, + { + "epoch": 5.73, + "learning_rate": 1.0852956160481542e-07, + "loss": 1.074, + "step": 77501 + }, + { + "epoch": 5.73, + "learning_rate": 1.0847096367246679e-07, + "loss": 0.9988, + "step": 77502 + }, + { + "epoch": 5.73, + "learning_rate": 1.0841238147739853e-07, + "loss": 0.8942, + "step": 77503 + }, + { + "epoch": 5.73, + "learning_rate": 1.0835381501970499e-07, + "loss": 1.0766, + "step": 77504 + }, + { + "epoch": 5.73, + "learning_rate": 1.0829526429947945e-07, + "loss": 0.9333, + "step": 77505 + }, + { + "epoch": 5.73, + "learning_rate": 1.0823672931681405e-07, + "loss": 0.9639, + "step": 77506 + }, + { + "epoch": 5.73, + "learning_rate": 1.0817821007180318e-07, + "loss": 0.9642, + "step": 77507 + }, + { + "epoch": 5.73, + "learning_rate": 1.0811970656453785e-07, + "loss": 1.0604, + "step": 77508 + }, + { + "epoch": 5.73, + "learning_rate": 1.0806121879511355e-07, + "loss": 0.983, + "step": 77509 + }, + { + "epoch": 5.73, + "learning_rate": 1.0800274676362243e-07, + "loss": 1.0266, + "step": 77510 + }, + { + "epoch": 5.73, + "learning_rate": 1.0794429047015665e-07, + "loss": 0.9636, + "step": 77511 + }, + { + "epoch": 5.73, + "learning_rate": 1.0788584991481055e-07, + "loss": 1.0653, + "step": 77512 + }, + { + "epoch": 5.73, + "learning_rate": 1.078274250976763e-07, + "loss": 0.9763, + "step": 77513 + }, + { + "epoch": 5.73, + "learning_rate": 1.0776901601884715e-07, + "loss": 1.1183, + "step": 77514 + }, + { + "epoch": 5.73, + "learning_rate": 1.0771062267841637e-07, + "loss": 1.0336, + "step": 77515 + }, + { + "epoch": 5.73, + "learning_rate": 1.076522450764761e-07, + "loss": 1.0969, + "step": 77516 + }, + { + "epoch": 5.73, + "learning_rate": 1.0759388321312069e-07, + "loss": 1.0794, + "step": 77517 + }, + { + "epoch": 5.73, + "learning_rate": 1.075355370884401e-07, + "loss": 0.9708, + "step": 77518 + }, + { + "epoch": 5.73, + "learning_rate": 1.0747720670252981e-07, + "loss": 0.9458, + "step": 77519 + }, + { + "epoch": 5.73, + "learning_rate": 1.0741889205548084e-07, + "loss": 0.9244, + "step": 77520 + }, + { + "epoch": 5.73, + "learning_rate": 1.0736059314738867e-07, + "loss": 1.0381, + "step": 77521 + }, + { + "epoch": 5.73, + "learning_rate": 1.0730230997834213e-07, + "loss": 1.0378, + "step": 77522 + }, + { + "epoch": 5.73, + "learning_rate": 1.0724404254843779e-07, + "loss": 0.9237, + "step": 77523 + }, + { + "epoch": 5.73, + "learning_rate": 1.0718579085776559e-07, + "loss": 1.0196, + "step": 77524 + }, + { + "epoch": 5.73, + "learning_rate": 1.0712755490641991e-07, + "loss": 1.1451, + "step": 77525 + }, + { + "epoch": 5.73, + "learning_rate": 1.0706933469449287e-07, + "loss": 0.8712, + "step": 77526 + }, + { + "epoch": 5.73, + "learning_rate": 1.0701113022207666e-07, + "loss": 0.9816, + "step": 77527 + }, + { + "epoch": 5.73, + "learning_rate": 1.069529414892645e-07, + "loss": 1.0033, + "step": 77528 + }, + { + "epoch": 5.73, + "learning_rate": 1.0689476849614744e-07, + "loss": 0.8921, + "step": 77529 + }, + { + "epoch": 5.73, + "learning_rate": 1.0683661124281986e-07, + "loss": 0.9826, + "step": 77530 + }, + { + "epoch": 5.73, + "learning_rate": 1.0677846972937389e-07, + "loss": 0.8969, + "step": 77531 + }, + { + "epoch": 5.73, + "learning_rate": 1.067203439559028e-07, + "loss": 1.0012, + "step": 77532 + }, + { + "epoch": 5.73, + "learning_rate": 1.0666223392249653e-07, + "loss": 0.9979, + "step": 77533 + }, + { + "epoch": 5.73, + "learning_rate": 1.0660413962924943e-07, + "loss": 0.9807, + "step": 77534 + }, + { + "epoch": 5.73, + "learning_rate": 1.0654606107625475e-07, + "loss": 1.0378, + "step": 77535 + }, + { + "epoch": 5.73, + "learning_rate": 1.0648799826360245e-07, + "loss": 0.9335, + "step": 77536 + }, + { + "epoch": 5.73, + "learning_rate": 1.0642995119138688e-07, + "loss": 1.0142, + "step": 77537 + }, + { + "epoch": 5.73, + "learning_rate": 1.0637191985969909e-07, + "loss": 1.046, + "step": 77538 + }, + { + "epoch": 5.73, + "learning_rate": 1.0631390426863231e-07, + "loss": 0.9126, + "step": 77539 + }, + { + "epoch": 5.73, + "learning_rate": 1.062559044182787e-07, + "loss": 0.9822, + "step": 77540 + }, + { + "epoch": 5.73, + "learning_rate": 1.0619792030873044e-07, + "loss": 1.0968, + "step": 77541 + }, + { + "epoch": 5.73, + "learning_rate": 1.0613995194007853e-07, + "loss": 0.9519, + "step": 77542 + }, + { + "epoch": 5.73, + "learning_rate": 1.0608199931241847e-07, + "loss": 1.0699, + "step": 77543 + }, + { + "epoch": 5.73, + "learning_rate": 1.0602406242584017e-07, + "loss": 1.0118, + "step": 77544 + }, + { + "epoch": 5.73, + "learning_rate": 1.059661412804347e-07, + "loss": 0.9231, + "step": 77545 + }, + { + "epoch": 5.73, + "learning_rate": 1.0590823587629528e-07, + "loss": 1.0008, + "step": 77546 + }, + { + "epoch": 5.73, + "learning_rate": 1.058503462135152e-07, + "loss": 1.0179, + "step": 77547 + }, + { + "epoch": 5.73, + "learning_rate": 1.0579247229218659e-07, + "loss": 1.0074, + "step": 77548 + }, + { + "epoch": 5.73, + "learning_rate": 1.0573461411239827e-07, + "loss": 1.1586, + "step": 77549 + }, + { + "epoch": 5.73, + "learning_rate": 1.0567677167424683e-07, + "loss": 0.968, + "step": 77550 + }, + { + "epoch": 5.73, + "learning_rate": 1.056189449778211e-07, + "loss": 0.9879, + "step": 77551 + }, + { + "epoch": 5.73, + "learning_rate": 1.0556113402321433e-07, + "loss": 0.9809, + "step": 77552 + }, + { + "epoch": 5.73, + "learning_rate": 1.0550333881051755e-07, + "loss": 1.0663, + "step": 77553 + }, + { + "epoch": 5.73, + "learning_rate": 1.0544555933982404e-07, + "loss": 0.9578, + "step": 77554 + }, + { + "epoch": 5.73, + "learning_rate": 1.0538779561122481e-07, + "loss": 1.0633, + "step": 77555 + }, + { + "epoch": 5.73, + "learning_rate": 1.0533004762481203e-07, + "loss": 1.0587, + "step": 77556 + }, + { + "epoch": 5.73, + "learning_rate": 1.0527231538067783e-07, + "loss": 0.8855, + "step": 77557 + }, + { + "epoch": 5.73, + "learning_rate": 1.0521459887891327e-07, + "loss": 1.1345, + "step": 77558 + }, + { + "epoch": 5.73, + "learning_rate": 1.051568981196105e-07, + "loss": 1.0906, + "step": 77559 + }, + { + "epoch": 5.73, + "learning_rate": 1.0509921310286275e-07, + "loss": 0.9993, + "step": 77560 + }, + { + "epoch": 5.73, + "learning_rate": 1.0504154382875998e-07, + "loss": 1.0429, + "step": 77561 + }, + { + "epoch": 5.73, + "learning_rate": 1.0498389029739431e-07, + "loss": 1.0766, + "step": 77562 + }, + { + "epoch": 5.73, + "learning_rate": 1.0492625250885791e-07, + "loss": 1.0009, + "step": 77563 + }, + { + "epoch": 5.73, + "learning_rate": 1.0486863046324292e-07, + "loss": 1.1801, + "step": 77564 + }, + { + "epoch": 5.73, + "learning_rate": 1.0481102416063926e-07, + "loss": 1.0485, + "step": 77565 + }, + { + "epoch": 5.73, + "learning_rate": 1.047534336011402e-07, + "loss": 1.0247, + "step": 77566 + }, + { + "epoch": 5.73, + "learning_rate": 1.0469585878483568e-07, + "loss": 0.9805, + "step": 77567 + }, + { + "epoch": 5.73, + "learning_rate": 1.0463829971182116e-07, + "loss": 1.0708, + "step": 77568 + }, + { + "epoch": 5.73, + "learning_rate": 1.0458075638218324e-07, + "loss": 1.0072, + "step": 77569 + }, + { + "epoch": 5.73, + "learning_rate": 1.0452322879601628e-07, + "loss": 1.052, + "step": 77570 + }, + { + "epoch": 5.73, + "learning_rate": 1.0446571695341134e-07, + "loss": 0.9441, + "step": 77571 + }, + { + "epoch": 5.73, + "learning_rate": 1.0440822085446056e-07, + "loss": 0.9649, + "step": 77572 + }, + { + "epoch": 5.73, + "learning_rate": 1.0435074049925386e-07, + "loss": 0.9914, + "step": 77573 + }, + { + "epoch": 5.73, + "learning_rate": 1.042932758878834e-07, + "loss": 1.0936, + "step": 77574 + }, + { + "epoch": 5.73, + "learning_rate": 1.0423582702044133e-07, + "loss": 1.01, + "step": 77575 + }, + { + "epoch": 5.73, + "learning_rate": 1.0417839389701867e-07, + "loss": 0.9207, + "step": 77576 + }, + { + "epoch": 5.73, + "learning_rate": 1.0412097651770647e-07, + "loss": 1.0428, + "step": 77577 + }, + { + "epoch": 5.73, + "learning_rate": 1.0406357488259466e-07, + "loss": 1.0376, + "step": 77578 + }, + { + "epoch": 5.73, + "learning_rate": 1.0400618899177761e-07, + "loss": 0.9922, + "step": 77579 + }, + { + "epoch": 5.73, + "learning_rate": 1.0394881884534525e-07, + "loss": 1.1177, + "step": 77580 + }, + { + "epoch": 5.73, + "learning_rate": 1.038914644433886e-07, + "loss": 1.0225, + "step": 77581 + }, + { + "epoch": 5.73, + "learning_rate": 1.0383412578599872e-07, + "loss": 1.0446, + "step": 77582 + }, + { + "epoch": 5.73, + "learning_rate": 1.0377680287326663e-07, + "loss": 1.0381, + "step": 77583 + }, + { + "epoch": 5.73, + "learning_rate": 1.0371949570528561e-07, + "loss": 0.993, + "step": 77584 + }, + { + "epoch": 5.73, + "learning_rate": 1.0366220428214447e-07, + "loss": 0.889, + "step": 77585 + }, + { + "epoch": 5.73, + "learning_rate": 1.0360492860393534e-07, + "loss": 1.0044, + "step": 77586 + }, + { + "epoch": 5.73, + "learning_rate": 1.0354766867074817e-07, + "loss": 1.0686, + "step": 77587 + }, + { + "epoch": 5.73, + "learning_rate": 1.034904244826762e-07, + "loss": 1.0018, + "step": 77588 + }, + { + "epoch": 5.73, + "learning_rate": 1.0343319603980828e-07, + "loss": 0.9607, + "step": 77589 + }, + { + "epoch": 5.73, + "learning_rate": 1.0337598334223764e-07, + "loss": 1.0361, + "step": 77590 + }, + { + "epoch": 5.73, + "learning_rate": 1.0331878639005422e-07, + "loss": 1.0267, + "step": 77591 + }, + { + "epoch": 5.73, + "learning_rate": 1.0326160518334904e-07, + "loss": 1.0246, + "step": 77592 + }, + { + "epoch": 5.73, + "learning_rate": 1.0320443972221206e-07, + "loss": 0.9839, + "step": 77593 + }, + { + "epoch": 5.73, + "learning_rate": 1.031472900067354e-07, + "loss": 0.9546, + "step": 77594 + }, + { + "epoch": 5.73, + "learning_rate": 1.0309015603701011e-07, + "loss": 0.981, + "step": 77595 + }, + { + "epoch": 5.73, + "learning_rate": 1.0303303781312723e-07, + "loss": 0.9603, + "step": 77596 + }, + { + "epoch": 5.73, + "learning_rate": 1.0297593533517669e-07, + "loss": 0.9836, + "step": 77597 + }, + { + "epoch": 5.73, + "learning_rate": 1.0291884860324952e-07, + "loss": 1.0303, + "step": 77598 + }, + { + "epoch": 5.73, + "learning_rate": 1.0286177761743787e-07, + "loss": 0.9468, + "step": 77599 + }, + { + "epoch": 5.73, + "learning_rate": 1.0280472237783057e-07, + "loss": 0.9842, + "step": 77600 + }, + { + "epoch": 5.73, + "learning_rate": 1.0274768288451975e-07, + "loss": 1.0648, + "step": 77601 + }, + { + "epoch": 5.73, + "learning_rate": 1.0269065913759646e-07, + "loss": 1.0146, + "step": 77602 + }, + { + "epoch": 5.73, + "learning_rate": 1.0263365113714952e-07, + "loss": 0.9927, + "step": 77603 + }, + { + "epoch": 5.73, + "learning_rate": 1.0257665888327106e-07, + "loss": 0.8897, + "step": 77604 + }, + { + "epoch": 5.73, + "learning_rate": 1.0251968237605214e-07, + "loss": 1.0079, + "step": 77605 + }, + { + "epoch": 5.73, + "learning_rate": 1.0246272161558268e-07, + "loss": 1.0584, + "step": 77606 + }, + { + "epoch": 5.73, + "learning_rate": 1.0240577660195261e-07, + "loss": 1.0118, + "step": 77607 + }, + { + "epoch": 5.73, + "learning_rate": 1.0234884733525408e-07, + "loss": 1.0344, + "step": 77608 + }, + { + "epoch": 5.73, + "learning_rate": 1.0229193381557701e-07, + "loss": 1.068, + "step": 77609 + }, + { + "epoch": 5.73, + "learning_rate": 1.0223503604301133e-07, + "loss": 0.9212, + "step": 77610 + }, + { + "epoch": 5.73, + "learning_rate": 1.0217815401764808e-07, + "loss": 1.0634, + "step": 77611 + }, + { + "epoch": 5.73, + "learning_rate": 1.021212877395783e-07, + "loss": 0.96, + "step": 77612 + }, + { + "epoch": 5.73, + "learning_rate": 1.0206443720889192e-07, + "loss": 1.0168, + "step": 77613 + }, + { + "epoch": 5.73, + "learning_rate": 1.0200760242567886e-07, + "loss": 1.0613, + "step": 77614 + }, + { + "epoch": 5.73, + "learning_rate": 1.0195078339003017e-07, + "loss": 1.0438, + "step": 77615 + }, + { + "epoch": 5.73, + "learning_rate": 1.0189398010203688e-07, + "loss": 0.9566, + "step": 77616 + }, + { + "epoch": 5.74, + "learning_rate": 1.0183719256178781e-07, + "loss": 1.1254, + "step": 77617 + }, + { + "epoch": 5.74, + "learning_rate": 1.017804207693751e-07, + "loss": 0.9858, + "step": 77618 + }, + { + "epoch": 5.74, + "learning_rate": 1.0172366472488649e-07, + "loss": 0.9256, + "step": 77619 + }, + { + "epoch": 5.74, + "learning_rate": 1.0166692442841519e-07, + "loss": 0.9475, + "step": 77620 + }, + { + "epoch": 5.74, + "learning_rate": 1.0161019988005006e-07, + "loss": 1.0811, + "step": 77621 + }, + { + "epoch": 5.74, + "learning_rate": 1.015534910798821e-07, + "loss": 1.0159, + "step": 77622 + }, + { + "epoch": 5.74, + "learning_rate": 1.0149679802799905e-07, + "loss": 1.1472, + "step": 77623 + }, + { + "epoch": 5.74, + "learning_rate": 1.0144012072449416e-07, + "loss": 0.9084, + "step": 77624 + }, + { + "epoch": 5.74, + "learning_rate": 1.0138345916945736e-07, + "loss": 1.0315, + "step": 77625 + }, + { + "epoch": 5.74, + "learning_rate": 1.0132681336297634e-07, + "loss": 0.9552, + "step": 77626 + }, + { + "epoch": 5.74, + "learning_rate": 1.0127018330514327e-07, + "loss": 1.0293, + "step": 77627 + }, + { + "epoch": 5.74, + "learning_rate": 1.0121356899604696e-07, + "loss": 0.9837, + "step": 77628 + }, + { + "epoch": 5.74, + "learning_rate": 1.0115697043577955e-07, + "loss": 0.9993, + "step": 77629 + }, + { + "epoch": 5.74, + "learning_rate": 1.0110038762442876e-07, + "loss": 0.9394, + "step": 77630 + }, + { + "epoch": 5.74, + "learning_rate": 1.0104382056208672e-07, + "loss": 0.9834, + "step": 77631 + }, + { + "epoch": 5.74, + "learning_rate": 1.0098726924884117e-07, + "loss": 1.0593, + "step": 77632 + }, + { + "epoch": 5.74, + "learning_rate": 1.0093073368478423e-07, + "loss": 1.0214, + "step": 77633 + }, + { + "epoch": 5.74, + "learning_rate": 1.0087421387000362e-07, + "loss": 1.0069, + "step": 77634 + }, + { + "epoch": 5.74, + "learning_rate": 1.0081770980459149e-07, + "loss": 1.0224, + "step": 77635 + }, + { + "epoch": 5.74, + "learning_rate": 1.0076122148863665e-07, + "loss": 1.07, + "step": 77636 + }, + { + "epoch": 5.74, + "learning_rate": 1.0070474892222904e-07, + "loss": 1.0382, + "step": 77637 + }, + { + "epoch": 5.74, + "learning_rate": 1.0064829210545857e-07, + "loss": 1.1029, + "step": 77638 + }, + { + "epoch": 5.74, + "learning_rate": 1.0059185103841406e-07, + "loss": 1.0235, + "step": 77639 + }, + { + "epoch": 5.74, + "learning_rate": 1.0053542572118769e-07, + "loss": 1.0445, + "step": 77640 + }, + { + "epoch": 5.74, + "learning_rate": 1.0047901615386712e-07, + "loss": 0.9827, + "step": 77641 + }, + { + "epoch": 5.74, + "learning_rate": 1.0042262233654343e-07, + "loss": 0.9364, + "step": 77642 + }, + { + "epoch": 5.74, + "learning_rate": 1.003662442693043e-07, + "loss": 0.9489, + "step": 77643 + }, + { + "epoch": 5.74, + "learning_rate": 1.003098819522419e-07, + "loss": 1.0451, + "step": 77644 + }, + { + "epoch": 5.74, + "learning_rate": 1.0025353538544502e-07, + "loss": 0.9225, + "step": 77645 + }, + { + "epoch": 5.74, + "learning_rate": 1.001972045690025e-07, + "loss": 1.032, + "step": 77646 + }, + { + "epoch": 5.74, + "learning_rate": 1.0014088950300538e-07, + "loss": 1.0062, + "step": 77647 + }, + { + "epoch": 5.74, + "learning_rate": 1.0008459018754135e-07, + "loss": 1.058, + "step": 77648 + }, + { + "epoch": 5.74, + "learning_rate": 1.0002830662270147e-07, + "loss": 1.1084, + "step": 77649 + }, + { + "epoch": 5.74, + "learning_rate": 9.997203880857565e-08, + "loss": 0.9783, + "step": 77650 + }, + { + "epoch": 5.74, + "learning_rate": 9.99157867452516e-08, + "loss": 1.0622, + "step": 77651 + }, + { + "epoch": 5.74, + "learning_rate": 9.985955043282037e-08, + "loss": 1.1108, + "step": 77652 + }, + { + "epoch": 5.74, + "learning_rate": 9.980332987137076e-08, + "loss": 0.9654, + "step": 77653 + }, + { + "epoch": 5.74, + "learning_rate": 9.974712506099271e-08, + "loss": 1.0552, + "step": 77654 + }, + { + "epoch": 5.74, + "learning_rate": 9.969093600177505e-08, + "loss": 1.0423, + "step": 77655 + }, + { + "epoch": 5.74, + "learning_rate": 9.963476269380768e-08, + "loss": 1.0419, + "step": 77656 + }, + { + "epoch": 5.74, + "learning_rate": 9.957860513718054e-08, + "loss": 1.1508, + "step": 77657 + }, + { + "epoch": 5.74, + "learning_rate": 9.952246333198135e-08, + "loss": 1.019, + "step": 77658 + }, + { + "epoch": 5.74, + "learning_rate": 9.946633727830002e-08, + "loss": 1.063, + "step": 77659 + }, + { + "epoch": 5.74, + "learning_rate": 9.941022697622649e-08, + "loss": 0.9783, + "step": 77660 + }, + { + "epoch": 5.74, + "learning_rate": 9.935413242584957e-08, + "loss": 1.0186, + "step": 77661 + }, + { + "epoch": 5.74, + "learning_rate": 9.929805362725809e-08, + "loss": 0.9406, + "step": 77662 + }, + { + "epoch": 5.74, + "learning_rate": 9.924199058054196e-08, + "loss": 1.0652, + "step": 77663 + }, + { + "epoch": 5.74, + "learning_rate": 9.918594328579e-08, + "loss": 1.0339, + "step": 77664 + }, + { + "epoch": 5.74, + "learning_rate": 9.912991174309217e-08, + "loss": 0.9935, + "step": 77665 + }, + { + "epoch": 5.74, + "learning_rate": 9.907389595253725e-08, + "loss": 0.9581, + "step": 77666 + }, + { + "epoch": 5.74, + "learning_rate": 9.901789591421407e-08, + "loss": 0.9921, + "step": 77667 + }, + { + "epoch": 5.74, + "learning_rate": 9.896191162821034e-08, + "loss": 0.906, + "step": 77668 + }, + { + "epoch": 5.74, + "learning_rate": 9.89059430946182e-08, + "loss": 0.952, + "step": 77669 + }, + { + "epoch": 5.74, + "learning_rate": 9.884999031352427e-08, + "loss": 1.1053, + "step": 77670 + }, + { + "epoch": 5.74, + "learning_rate": 9.879405328501846e-08, + "loss": 0.9846, + "step": 77671 + }, + { + "epoch": 5.74, + "learning_rate": 9.873813200918846e-08, + "loss": 0.9594, + "step": 77672 + }, + { + "epoch": 5.74, + "learning_rate": 9.868222648612646e-08, + "loss": 0.9923, + "step": 77673 + }, + { + "epoch": 5.74, + "learning_rate": 9.862633671591904e-08, + "loss": 1.076, + "step": 77674 + }, + { + "epoch": 5.74, + "learning_rate": 9.85704626986539e-08, + "loss": 1.0377, + "step": 77675 + }, + { + "epoch": 5.74, + "learning_rate": 9.851460443442317e-08, + "loss": 1.0156, + "step": 77676 + }, + { + "epoch": 5.74, + "learning_rate": 9.845876192331238e-08, + "loss": 1.149, + "step": 77677 + }, + { + "epoch": 5.74, + "learning_rate": 9.840293516541366e-08, + "loss": 1.0487, + "step": 77678 + }, + { + "epoch": 5.74, + "learning_rate": 9.83471241608136e-08, + "loss": 0.9233, + "step": 77679 + }, + { + "epoch": 5.74, + "learning_rate": 9.829132890960102e-08, + "loss": 0.9539, + "step": 77680 + }, + { + "epoch": 5.74, + "learning_rate": 9.823554941186587e-08, + "loss": 0.972, + "step": 77681 + }, + { + "epoch": 5.74, + "learning_rate": 9.817978566769692e-08, + "loss": 0.9103, + "step": 77682 + }, + { + "epoch": 5.74, + "learning_rate": 9.812403767718082e-08, + "loss": 1.0366, + "step": 77683 + }, + { + "epoch": 5.74, + "learning_rate": 9.806830544040857e-08, + "loss": 0.8978, + "step": 77684 + }, + { + "epoch": 5.74, + "learning_rate": 9.80125889574679e-08, + "loss": 1.003, + "step": 77685 + }, + { + "epoch": 5.74, + "learning_rate": 9.795688822844762e-08, + "loss": 1.055, + "step": 77686 + }, + { + "epoch": 5.74, + "learning_rate": 9.790120325343655e-08, + "loss": 0.9792, + "step": 77687 + }, + { + "epoch": 5.74, + "learning_rate": 9.784553403252128e-08, + "loss": 1.0043, + "step": 77688 + }, + { + "epoch": 5.74, + "learning_rate": 9.778988056579397e-08, + "loss": 0.9341, + "step": 77689 + }, + { + "epoch": 5.74, + "learning_rate": 9.77342428533401e-08, + "loss": 1.0367, + "step": 77690 + }, + { + "epoch": 5.74, + "learning_rate": 9.76786208952507e-08, + "loss": 0.8673, + "step": 77691 + }, + { + "epoch": 5.74, + "learning_rate": 9.762301469161128e-08, + "loss": 0.9099, + "step": 77692 + }, + { + "epoch": 5.74, + "learning_rate": 9.756742424251287e-08, + "loss": 1.0983, + "step": 77693 + }, + { + "epoch": 5.74, + "learning_rate": 9.751184954804316e-08, + "loss": 1.0779, + "step": 77694 + }, + { + "epoch": 5.74, + "learning_rate": 9.745629060829098e-08, + "loss": 0.9796, + "step": 77695 + }, + { + "epoch": 5.74, + "learning_rate": 9.740074742334293e-08, + "loss": 1.0125, + "step": 77696 + }, + { + "epoch": 5.74, + "learning_rate": 9.734521999328894e-08, + "loss": 0.9475, + "step": 77697 + }, + { + "epoch": 5.74, + "learning_rate": 9.728970831821672e-08, + "loss": 1.0035, + "step": 77698 + }, + { + "epoch": 5.74, + "learning_rate": 9.723421239821618e-08, + "loss": 1.0492, + "step": 77699 + }, + { + "epoch": 5.74, + "learning_rate": 9.717873223337393e-08, + "loss": 1.0047, + "step": 77700 + }, + { + "epoch": 5.74, + "learning_rate": 9.712326782377768e-08, + "loss": 0.9326, + "step": 77701 + }, + { + "epoch": 5.74, + "learning_rate": 9.706781916951735e-08, + "loss": 1.0835, + "step": 77702 + }, + { + "epoch": 5.74, + "learning_rate": 9.701238627068066e-08, + "loss": 1.1328, + "step": 77703 + }, + { + "epoch": 5.74, + "learning_rate": 9.695696912735419e-08, + "loss": 0.9816, + "step": 77704 + }, + { + "epoch": 5.74, + "learning_rate": 9.6901567739629e-08, + "loss": 1.0071, + "step": 77705 + }, + { + "epoch": 5.74, + "learning_rate": 9.684618210759055e-08, + "loss": 1.0143, + "step": 77706 + }, + { + "epoch": 5.74, + "learning_rate": 9.67908122313288e-08, + "loss": 0.9849, + "step": 77707 + }, + { + "epoch": 5.74, + "learning_rate": 9.673545811093144e-08, + "loss": 1.0316, + "step": 77708 + }, + { + "epoch": 5.74, + "learning_rate": 9.668011974648616e-08, + "loss": 0.9836, + "step": 77709 + }, + { + "epoch": 5.74, + "learning_rate": 9.66247971380807e-08, + "loss": 0.9677, + "step": 77710 + }, + { + "epoch": 5.74, + "learning_rate": 9.656949028580387e-08, + "loss": 1.0429, + "step": 77711 + }, + { + "epoch": 5.74, + "learning_rate": 9.651419918974337e-08, + "loss": 1.0019, + "step": 77712 + }, + { + "epoch": 5.74, + "learning_rate": 9.645892384998579e-08, + "loss": 1.0427, + "step": 77713 + }, + { + "epoch": 5.74, + "learning_rate": 9.640366426662218e-08, + "loss": 0.959, + "step": 77714 + }, + { + "epoch": 5.74, + "learning_rate": 9.634842043973803e-08, + "loss": 0.9803, + "step": 77715 + }, + { + "epoch": 5.74, + "learning_rate": 9.629319236942103e-08, + "loss": 0.9892, + "step": 77716 + }, + { + "epoch": 5.74, + "learning_rate": 9.623798005576002e-08, + "loss": 1.0332, + "step": 77717 + }, + { + "epoch": 5.74, + "learning_rate": 9.61827834988438e-08, + "loss": 1.069, + "step": 77718 + }, + { + "epoch": 5.74, + "learning_rate": 9.612760269875787e-08, + "loss": 1.0489, + "step": 77719 + }, + { + "epoch": 5.74, + "learning_rate": 9.607243765559216e-08, + "loss": 1.0072, + "step": 77720 + }, + { + "epoch": 5.74, + "learning_rate": 9.601728836943325e-08, + "loss": 0.8964, + "step": 77721 + }, + { + "epoch": 5.74, + "learning_rate": 9.596215484036886e-08, + "loss": 0.9569, + "step": 77722 + }, + { + "epoch": 5.74, + "learning_rate": 9.59070370684878e-08, + "loss": 1.0875, + "step": 77723 + }, + { + "epoch": 5.74, + "learning_rate": 9.585193505387558e-08, + "loss": 0.9023, + "step": 77724 + }, + { + "epoch": 5.74, + "learning_rate": 9.57968487966221e-08, + "loss": 0.936, + "step": 77725 + }, + { + "epoch": 5.74, + "learning_rate": 9.574177829681396e-08, + "loss": 1.1189, + "step": 77726 + }, + { + "epoch": 5.74, + "learning_rate": 9.56867235545389e-08, + "loss": 0.9802, + "step": 77727 + }, + { + "epoch": 5.74, + "learning_rate": 9.563168456988458e-08, + "loss": 1.1464, + "step": 77728 + }, + { + "epoch": 5.74, + "learning_rate": 9.557666134293874e-08, + "loss": 0.9779, + "step": 77729 + }, + { + "epoch": 5.74, + "learning_rate": 9.552165387378798e-08, + "loss": 1.0115, + "step": 77730 + }, + { + "epoch": 5.74, + "learning_rate": 9.54666621625222e-08, + "loss": 0.9445, + "step": 77731 + }, + { + "epoch": 5.74, + "learning_rate": 9.54116862092258e-08, + "loss": 0.9818, + "step": 77732 + }, + { + "epoch": 5.74, + "learning_rate": 9.53567260139876e-08, + "loss": 1.0486, + "step": 77733 + }, + { + "epoch": 5.74, + "learning_rate": 9.530178157689641e-08, + "loss": 1.0534, + "step": 77734 + }, + { + "epoch": 5.74, + "learning_rate": 9.52468528980377e-08, + "loss": 0.9779, + "step": 77735 + }, + { + "epoch": 5.74, + "learning_rate": 9.519193997749921e-08, + "loss": 1.021, + "step": 77736 + }, + { + "epoch": 5.74, + "learning_rate": 9.513704281536861e-08, + "loss": 0.9856, + "step": 77737 + }, + { + "epoch": 5.74, + "learning_rate": 9.508216141173254e-08, + "loss": 1.1719, + "step": 77738 + }, + { + "epoch": 5.74, + "learning_rate": 9.50272957666809e-08, + "loss": 0.921, + "step": 77739 + }, + { + "epoch": 5.74, + "learning_rate": 9.497244588029809e-08, + "loss": 1.0465, + "step": 77740 + }, + { + "epoch": 5.74, + "learning_rate": 9.491761175267289e-08, + "loss": 0.9536, + "step": 77741 + }, + { + "epoch": 5.74, + "learning_rate": 9.486279338389081e-08, + "loss": 0.9895, + "step": 77742 + }, + { + "epoch": 5.74, + "learning_rate": 9.480799077404179e-08, + "loss": 1.0531, + "step": 77743 + }, + { + "epoch": 5.74, + "learning_rate": 9.475320392321019e-08, + "loss": 0.8157, + "step": 77744 + }, + { + "epoch": 5.74, + "learning_rate": 9.469843283148595e-08, + "loss": 0.9859, + "step": 77745 + }, + { + "epoch": 5.74, + "learning_rate": 9.464367749895343e-08, + "loss": 1.0034, + "step": 77746 + }, + { + "epoch": 5.74, + "learning_rate": 9.458893792570256e-08, + "loss": 0.9693, + "step": 77747 + }, + { + "epoch": 5.74, + "learning_rate": 9.453421411181773e-08, + "loss": 0.9917, + "step": 77748 + }, + { + "epoch": 5.74, + "learning_rate": 9.447950605738776e-08, + "loss": 1.0828, + "step": 77749 + }, + { + "epoch": 5.74, + "learning_rate": 9.442481376249923e-08, + "loss": 0.9166, + "step": 77750 + }, + { + "epoch": 5.74, + "learning_rate": 9.437013722723875e-08, + "loss": 0.9221, + "step": 77751 + }, + { + "epoch": 5.75, + "learning_rate": 9.431547645169403e-08, + "loss": 0.9364, + "step": 77752 + }, + { + "epoch": 5.75, + "learning_rate": 9.426083143595056e-08, + "loss": 0.9198, + "step": 77753 + }, + { + "epoch": 5.75, + "learning_rate": 9.420620218009713e-08, + "loss": 1.0399, + "step": 77754 + }, + { + "epoch": 5.75, + "learning_rate": 9.415158868422036e-08, + "loss": 1.0892, + "step": 77755 + }, + { + "epoch": 5.75, + "learning_rate": 9.409699094840686e-08, + "loss": 1.0745, + "step": 77756 + }, + { + "epoch": 5.75, + "learning_rate": 9.404240897274209e-08, + "loss": 1.0713, + "step": 77757 + }, + { + "epoch": 5.75, + "learning_rate": 9.398784275731488e-08, + "loss": 0.9489, + "step": 77758 + }, + { + "epoch": 5.75, + "learning_rate": 9.393329230221183e-08, + "loss": 1.0549, + "step": 77759 + }, + { + "epoch": 5.75, + "learning_rate": 9.387875760751842e-08, + "loss": 1.0325, + "step": 77760 + }, + { + "epoch": 5.75, + "learning_rate": 9.382423867332236e-08, + "loss": 0.9015, + "step": 77761 + }, + { + "epoch": 5.75, + "learning_rate": 9.376973549971024e-08, + "loss": 1.0084, + "step": 77762 + }, + { + "epoch": 5.75, + "learning_rate": 9.371524808676868e-08, + "loss": 0.9573, + "step": 77763 + }, + { + "epoch": 5.75, + "learning_rate": 9.366077643458427e-08, + "loss": 1.0511, + "step": 77764 + }, + { + "epoch": 5.75, + "learning_rate": 9.36063205432447e-08, + "loss": 1.1287, + "step": 77765 + }, + { + "epoch": 5.75, + "learning_rate": 9.355188041283436e-08, + "loss": 0.943, + "step": 77766 + }, + { + "epoch": 5.75, + "learning_rate": 9.349745604344206e-08, + "loss": 1.0658, + "step": 77767 + }, + { + "epoch": 5.75, + "learning_rate": 9.344304743515331e-08, + "loss": 1.0014, + "step": 77768 + }, + { + "epoch": 5.75, + "learning_rate": 9.338865458805468e-08, + "loss": 0.9735, + "step": 77769 + }, + { + "epoch": 5.75, + "learning_rate": 9.33342775022339e-08, + "loss": 1.0185, + "step": 77770 + }, + { + "epoch": 5.75, + "learning_rate": 9.327991617777533e-08, + "loss": 1.1121, + "step": 77771 + }, + { + "epoch": 5.75, + "learning_rate": 9.322557061476778e-08, + "loss": 1.0006, + "step": 77772 + }, + { + "epoch": 5.75, + "learning_rate": 9.317124081329564e-08, + "loss": 0.9912, + "step": 77773 + }, + { + "epoch": 5.75, + "learning_rate": 9.311692677344775e-08, + "loss": 1.0631, + "step": 77774 + }, + { + "epoch": 5.75, + "learning_rate": 9.306262849530734e-08, + "loss": 1.1053, + "step": 77775 + }, + { + "epoch": 5.75, + "learning_rate": 9.300834597896435e-08, + "loss": 0.9398, + "step": 77776 + }, + { + "epoch": 5.75, + "learning_rate": 9.295407922450205e-08, + "loss": 0.9704, + "step": 77777 + }, + { + "epoch": 5.75, + "learning_rate": 9.289982823200816e-08, + "loss": 1.0658, + "step": 77778 + }, + { + "epoch": 5.75, + "learning_rate": 9.284559300156925e-08, + "loss": 1.0133, + "step": 77779 + }, + { + "epoch": 5.75, + "learning_rate": 9.279137353327084e-08, + "loss": 1.0363, + "step": 77780 + }, + { + "epoch": 5.75, + "learning_rate": 9.27371698271995e-08, + "loss": 1.0115, + "step": 77781 + }, + { + "epoch": 5.75, + "learning_rate": 9.268298188344182e-08, + "loss": 1.0157, + "step": 77782 + }, + { + "epoch": 5.75, + "learning_rate": 9.262880970208443e-08, + "loss": 1.2013, + "step": 77783 + }, + { + "epoch": 5.75, + "learning_rate": 9.257465328321169e-08, + "loss": 0.9799, + "step": 77784 + }, + { + "epoch": 5.75, + "learning_rate": 9.252051262691242e-08, + "loss": 0.9918, + "step": 77785 + }, + { + "epoch": 5.75, + "learning_rate": 9.246638773326987e-08, + "loss": 0.9173, + "step": 77786 + }, + { + "epoch": 5.75, + "learning_rate": 9.241227860237179e-08, + "loss": 0.9421, + "step": 77787 + }, + { + "epoch": 5.75, + "learning_rate": 9.235818523430473e-08, + "loss": 0.9732, + "step": 77788 + }, + { + "epoch": 5.75, + "learning_rate": 9.230410762915309e-08, + "loss": 0.8716, + "step": 77789 + }, + { + "epoch": 5.75, + "learning_rate": 9.225004578700458e-08, + "loss": 0.9127, + "step": 77790 + }, + { + "epoch": 5.75, + "learning_rate": 9.219599970794357e-08, + "loss": 0.9993, + "step": 77791 + }, + { + "epoch": 5.75, + "learning_rate": 9.214196939205889e-08, + "loss": 1.0065, + "step": 77792 + }, + { + "epoch": 5.75, + "learning_rate": 9.208795483943267e-08, + "loss": 1.0498, + "step": 77793 + }, + { + "epoch": 5.75, + "learning_rate": 9.203395605015375e-08, + "loss": 1.0883, + "step": 77794 + }, + { + "epoch": 5.75, + "learning_rate": 9.19799730243065e-08, + "loss": 0.9654, + "step": 77795 + }, + { + "epoch": 5.75, + "learning_rate": 9.192600576197752e-08, + "loss": 0.9522, + "step": 77796 + }, + { + "epoch": 5.75, + "learning_rate": 9.187205426325341e-08, + "loss": 0.8926, + "step": 77797 + }, + { + "epoch": 5.75, + "learning_rate": 9.181811852821742e-08, + "loss": 0.9854, + "step": 77798 + }, + { + "epoch": 5.75, + "learning_rate": 9.176419855695839e-08, + "loss": 0.9201, + "step": 77799 + }, + { + "epoch": 5.75, + "learning_rate": 9.171029434955958e-08, + "loss": 0.9685, + "step": 77800 + }, + { + "epoch": 5.75, + "learning_rate": 9.165640590610869e-08, + "loss": 1.0184, + "step": 77801 + }, + { + "epoch": 5.75, + "learning_rate": 9.160253322669011e-08, + "loss": 0.8304, + "step": 77802 + }, + { + "epoch": 5.75, + "learning_rate": 9.154867631139153e-08, + "loss": 0.9445, + "step": 77803 + }, + { + "epoch": 5.75, + "learning_rate": 9.149483516029623e-08, + "loss": 1.0586, + "step": 77804 + }, + { + "epoch": 5.75, + "learning_rate": 9.144100977349079e-08, + "loss": 1.1207, + "step": 77805 + }, + { + "epoch": 5.75, + "learning_rate": 9.138720015106183e-08, + "loss": 1.0766, + "step": 77806 + }, + { + "epoch": 5.75, + "learning_rate": 9.133340629309262e-08, + "loss": 0.9752, + "step": 77807 + }, + { + "epoch": 5.75, + "learning_rate": 9.127962819967085e-08, + "loss": 0.988, + "step": 77808 + }, + { + "epoch": 5.75, + "learning_rate": 9.1225865870882e-08, + "loss": 0.9646, + "step": 77809 + }, + { + "epoch": 5.75, + "learning_rate": 9.117211930681047e-08, + "loss": 1.1434, + "step": 77810 + }, + { + "epoch": 5.75, + "learning_rate": 9.111838850754173e-08, + "loss": 0.9644, + "step": 77811 + }, + { + "epoch": 5.75, + "learning_rate": 9.106467347316239e-08, + "loss": 0.9671, + "step": 77812 + }, + { + "epoch": 5.75, + "learning_rate": 9.101097420375792e-08, + "loss": 0.9035, + "step": 77813 + }, + { + "epoch": 5.75, + "learning_rate": 9.095729069941273e-08, + "loss": 1.0427, + "step": 77814 + }, + { + "epoch": 5.75, + "learning_rate": 9.090362296021338e-08, + "loss": 1.0116, + "step": 77815 + }, + { + "epoch": 5.75, + "learning_rate": 9.084997098624315e-08, + "loss": 1.0164, + "step": 77816 + }, + { + "epoch": 5.75, + "learning_rate": 9.079633477758975e-08, + "loss": 1.0516, + "step": 77817 + }, + { + "epoch": 5.75, + "learning_rate": 9.074271433433756e-08, + "loss": 1.067, + "step": 77818 + }, + { + "epoch": 5.75, + "learning_rate": 9.068910965657207e-08, + "loss": 0.9438, + "step": 77819 + }, + { + "epoch": 5.75, + "learning_rate": 9.063552074437765e-08, + "loss": 0.9805, + "step": 77820 + }, + { + "epoch": 5.75, + "learning_rate": 9.0581947597842e-08, + "loss": 0.9354, + "step": 77821 + }, + { + "epoch": 5.75, + "learning_rate": 9.05283902170473e-08, + "loss": 1.007, + "step": 77822 + }, + { + "epoch": 5.75, + "learning_rate": 9.047484860208011e-08, + "loss": 1.038, + "step": 77823 + }, + { + "epoch": 5.75, + "learning_rate": 9.042132275302595e-08, + "loss": 0.9671, + "step": 77824 + }, + { + "epoch": 5.75, + "learning_rate": 9.03678126699703e-08, + "loss": 1.0472, + "step": 77825 + }, + { + "epoch": 5.75, + "learning_rate": 9.031431835299753e-08, + "loss": 0.9993, + "step": 77826 + }, + { + "epoch": 5.75, + "learning_rate": 9.026083980219203e-08, + "loss": 1.0569, + "step": 77827 + }, + { + "epoch": 5.75, + "learning_rate": 9.020737701764037e-08, + "loss": 1.0164, + "step": 77828 + }, + { + "epoch": 5.75, + "learning_rate": 9.015392999942695e-08, + "loss": 1.0798, + "step": 77829 + }, + { + "epoch": 5.75, + "learning_rate": 9.010049874763616e-08, + "loss": 1.1117, + "step": 77830 + }, + { + "epoch": 5.75, + "learning_rate": 9.004708326235457e-08, + "loss": 0.9555, + "step": 77831 + }, + { + "epoch": 5.75, + "learning_rate": 8.999368354366544e-08, + "loss": 0.9219, + "step": 77832 + }, + { + "epoch": 5.75, + "learning_rate": 8.99402995916554e-08, + "loss": 0.977, + "step": 77833 + }, + { + "epoch": 5.75, + "learning_rate": 8.988693140640881e-08, + "loss": 1.02, + "step": 77834 + }, + { + "epoch": 5.75, + "learning_rate": 8.983357898801004e-08, + "loss": 0.9495, + "step": 77835 + }, + { + "epoch": 5.75, + "learning_rate": 8.978024233654348e-08, + "loss": 0.9728, + "step": 77836 + }, + { + "epoch": 5.75, + "learning_rate": 8.972692145209572e-08, + "loss": 1.0202, + "step": 77837 + }, + { + "epoch": 5.75, + "learning_rate": 8.967361633475002e-08, + "loss": 0.9156, + "step": 77838 + }, + { + "epoch": 5.75, + "learning_rate": 8.962032698459189e-08, + "loss": 1.0215, + "step": 77839 + }, + { + "epoch": 5.75, + "learning_rate": 8.956705340170568e-08, + "loss": 1.0857, + "step": 77840 + }, + { + "epoch": 5.75, + "learning_rate": 8.95137955861769e-08, + "loss": 1.1043, + "step": 77841 + }, + { + "epoch": 5.75, + "learning_rate": 8.946055353808991e-08, + "loss": 0.992, + "step": 77842 + }, + { + "epoch": 5.75, + "learning_rate": 8.94073272575291e-08, + "loss": 0.9885, + "step": 77843 + }, + { + "epoch": 5.75, + "learning_rate": 8.935411674457995e-08, + "loss": 0.9863, + "step": 77844 + }, + { + "epoch": 5.75, + "learning_rate": 8.930092199932683e-08, + "loss": 1.058, + "step": 77845 + }, + { + "epoch": 5.75, + "learning_rate": 8.924774302185413e-08, + "loss": 1.0918, + "step": 77846 + }, + { + "epoch": 5.75, + "learning_rate": 8.919457981224511e-08, + "loss": 1.0003, + "step": 77847 + }, + { + "epoch": 5.75, + "learning_rate": 8.914143237058637e-08, + "loss": 0.9392, + "step": 77848 + }, + { + "epoch": 5.75, + "learning_rate": 8.908830069696228e-08, + "loss": 1.0496, + "step": 77849 + }, + { + "epoch": 5.75, + "learning_rate": 8.903518479145723e-08, + "loss": 1.0974, + "step": 77850 + }, + { + "epoch": 5.75, + "learning_rate": 8.898208465415448e-08, + "loss": 1.0022, + "step": 77851 + }, + { + "epoch": 5.75, + "learning_rate": 8.892900028513951e-08, + "loss": 0.9247, + "step": 77852 + }, + { + "epoch": 5.75, + "learning_rate": 8.88759316844967e-08, + "loss": 0.9374, + "step": 77853 + }, + { + "epoch": 5.75, + "learning_rate": 8.882287885231156e-08, + "loss": 1.0566, + "step": 77854 + }, + { + "epoch": 5.75, + "learning_rate": 8.876984178866621e-08, + "loss": 1.0291, + "step": 77855 + }, + { + "epoch": 5.75, + "learning_rate": 8.871682049364616e-08, + "loss": 0.9107, + "step": 77856 + }, + { + "epoch": 5.75, + "learning_rate": 8.86638149673369e-08, + "loss": 0.971, + "step": 77857 + }, + { + "epoch": 5.75, + "learning_rate": 8.861082520982056e-08, + "loss": 0.9378, + "step": 77858 + }, + { + "epoch": 5.75, + "learning_rate": 8.855785122118376e-08, + "loss": 1.0463, + "step": 77859 + }, + { + "epoch": 5.75, + "learning_rate": 8.850489300150866e-08, + "loss": 1.0165, + "step": 77860 + }, + { + "epoch": 5.75, + "learning_rate": 8.845195055088074e-08, + "loss": 0.8952, + "step": 77861 + }, + { + "epoch": 5.75, + "learning_rate": 8.839902386938437e-08, + "loss": 0.9117, + "step": 77862 + }, + { + "epoch": 5.75, + "learning_rate": 8.834611295710283e-08, + "loss": 1.042, + "step": 77863 + }, + { + "epoch": 5.75, + "learning_rate": 8.829321781412159e-08, + "loss": 1.048, + "step": 77864 + }, + { + "epoch": 5.75, + "learning_rate": 8.824033844052283e-08, + "loss": 0.9499, + "step": 77865 + }, + { + "epoch": 5.75, + "learning_rate": 8.818747483639312e-08, + "loss": 1.096, + "step": 77866 + }, + { + "epoch": 5.75, + "learning_rate": 8.813462700181352e-08, + "loss": 1.0025, + "step": 77867 + }, + { + "epoch": 5.75, + "learning_rate": 8.808179493687175e-08, + "loss": 1.0077, + "step": 77868 + }, + { + "epoch": 5.75, + "learning_rate": 8.802897864164883e-08, + "loss": 0.9568, + "step": 77869 + }, + { + "epoch": 5.75, + "learning_rate": 8.797617811623139e-08, + "loss": 1.0007, + "step": 77870 + }, + { + "epoch": 5.75, + "learning_rate": 8.792339336070044e-08, + "loss": 0.9767, + "step": 77871 + }, + { + "epoch": 5.75, + "learning_rate": 8.78706243751426e-08, + "loss": 1.0018, + "step": 77872 + }, + { + "epoch": 5.75, + "learning_rate": 8.781787115964002e-08, + "loss": 0.9686, + "step": 77873 + }, + { + "epoch": 5.75, + "learning_rate": 8.77651337142782e-08, + "loss": 1.053, + "step": 77874 + }, + { + "epoch": 5.75, + "learning_rate": 8.771241203914038e-08, + "loss": 0.9971, + "step": 77875 + }, + { + "epoch": 5.75, + "learning_rate": 8.765970613430874e-08, + "loss": 1.0096, + "step": 77876 + }, + { + "epoch": 5.75, + "learning_rate": 8.760701599986986e-08, + "loss": 1.0208, + "step": 77877 + }, + { + "epoch": 5.75, + "learning_rate": 8.755434163590593e-08, + "loss": 0.8675, + "step": 77878 + }, + { + "epoch": 5.75, + "learning_rate": 8.750168304250239e-08, + "loss": 1.0411, + "step": 77879 + }, + { + "epoch": 5.75, + "learning_rate": 8.744904021974032e-08, + "loss": 1.0835, + "step": 77880 + }, + { + "epoch": 5.75, + "learning_rate": 8.739641316770519e-08, + "loss": 0.9838, + "step": 77881 + }, + { + "epoch": 5.75, + "learning_rate": 8.73438018864825e-08, + "loss": 0.9802, + "step": 77882 + }, + { + "epoch": 5.75, + "learning_rate": 8.729120637615107e-08, + "loss": 1.0333, + "step": 77883 + }, + { + "epoch": 5.75, + "learning_rate": 8.723862663679971e-08, + "loss": 1.0052, + "step": 77884 + }, + { + "epoch": 5.75, + "learning_rate": 8.718606266850838e-08, + "loss": 1.009, + "step": 77885 + }, + { + "epoch": 5.75, + "learning_rate": 8.713351447136365e-08, + "loss": 0.9901, + "step": 77886 + }, + { + "epoch": 5.76, + "learning_rate": 8.708098204544657e-08, + "loss": 0.9928, + "step": 77887 + }, + { + "epoch": 5.76, + "learning_rate": 8.702846539084264e-08, + "loss": 0.9152, + "step": 77888 + }, + { + "epoch": 5.76, + "learning_rate": 8.697596450763402e-08, + "loss": 0.9678, + "step": 77889 + }, + { + "epoch": 5.76, + "learning_rate": 8.692347939590506e-08, + "loss": 0.9807, + "step": 77890 + }, + { + "epoch": 5.76, + "learning_rate": 8.687101005573906e-08, + "loss": 1.0131, + "step": 77891 + }, + { + "epoch": 5.76, + "learning_rate": 8.681855648721815e-08, + "loss": 0.919, + "step": 77892 + }, + { + "epoch": 5.76, + "learning_rate": 8.676611869042895e-08, + "loss": 1.0008, + "step": 77893 + }, + { + "epoch": 5.76, + "learning_rate": 8.671369666545137e-08, + "loss": 0.9611, + "step": 77894 + }, + { + "epoch": 5.76, + "learning_rate": 8.666129041237204e-08, + "loss": 0.9429, + "step": 77895 + }, + { + "epoch": 5.76, + "learning_rate": 8.660889993127197e-08, + "loss": 1.027, + "step": 77896 + }, + { + "epoch": 5.76, + "learning_rate": 8.655652522223446e-08, + "loss": 1.0181, + "step": 77897 + }, + { + "epoch": 5.76, + "learning_rate": 8.650416628534498e-08, + "loss": 1.0422, + "step": 77898 + }, + { + "epoch": 5.76, + "learning_rate": 8.645182312068567e-08, + "loss": 0.9597, + "step": 77899 + }, + { + "epoch": 5.76, + "learning_rate": 8.639949572833872e-08, + "loss": 1.009, + "step": 77900 + }, + { + "epoch": 5.76, + "learning_rate": 8.63471841083885e-08, + "loss": 1.0774, + "step": 77901 + }, + { + "epoch": 5.76, + "learning_rate": 8.629488826091825e-08, + "loss": 0.9448, + "step": 77902 + }, + { + "epoch": 5.76, + "learning_rate": 8.624260818601126e-08, + "loss": 1.0349, + "step": 77903 + }, + { + "epoch": 5.76, + "learning_rate": 8.61903438837508e-08, + "loss": 0.949, + "step": 77904 + }, + { + "epoch": 5.76, + "learning_rate": 8.613809535421901e-08, + "loss": 0.8759, + "step": 77905 + }, + { + "epoch": 5.76, + "learning_rate": 8.608586259749918e-08, + "loss": 1.0052, + "step": 77906 + }, + { + "epoch": 5.76, + "learning_rate": 8.603364561367566e-08, + "loss": 1.0017, + "step": 77907 + }, + { + "epoch": 5.76, + "learning_rate": 8.598144440283062e-08, + "loss": 0.9988, + "step": 77908 + }, + { + "epoch": 5.76, + "learning_rate": 8.592925896504733e-08, + "loss": 0.9782, + "step": 77909 + }, + { + "epoch": 5.76, + "learning_rate": 8.587708930040905e-08, + "loss": 1.0581, + "step": 77910 + }, + { + "epoch": 5.76, + "learning_rate": 8.582493540899795e-08, + "loss": 1.0374, + "step": 77911 + }, + { + "epoch": 5.76, + "learning_rate": 8.577279729089727e-08, + "loss": 1.0737, + "step": 77912 + }, + { + "epoch": 5.76, + "learning_rate": 8.57206749461914e-08, + "loss": 1.0967, + "step": 77913 + }, + { + "epoch": 5.76, + "learning_rate": 8.56685683749614e-08, + "loss": 1.0337, + "step": 77914 + }, + { + "epoch": 5.76, + "learning_rate": 8.561647757729053e-08, + "loss": 0.937, + "step": 77915 + }, + { + "epoch": 5.76, + "learning_rate": 8.556440255326204e-08, + "loss": 0.9761, + "step": 77916 + }, + { + "epoch": 5.76, + "learning_rate": 8.551234330295811e-08, + "loss": 1.0117, + "step": 77917 + }, + { + "epoch": 5.76, + "learning_rate": 8.54602998264631e-08, + "loss": 1.0329, + "step": 77918 + }, + { + "epoch": 5.76, + "learning_rate": 8.540827212385915e-08, + "loss": 1.0964, + "step": 77919 + }, + { + "epoch": 5.76, + "learning_rate": 8.535626019522847e-08, + "loss": 1.179, + "step": 77920 + }, + { + "epoch": 5.76, + "learning_rate": 8.530426404065428e-08, + "loss": 1.0543, + "step": 77921 + }, + { + "epoch": 5.76, + "learning_rate": 8.525228366021875e-08, + "loss": 1.0426, + "step": 77922 + }, + { + "epoch": 5.76, + "learning_rate": 8.520031905400517e-08, + "loss": 0.9798, + "step": 77923 + }, + { + "epoch": 5.76, + "learning_rate": 8.514837022209566e-08, + "loss": 0.9629, + "step": 77924 + }, + { + "epoch": 5.76, + "learning_rate": 8.509643716457349e-08, + "loss": 0.9672, + "step": 77925 + }, + { + "epoch": 5.76, + "learning_rate": 8.504451988152085e-08, + "loss": 1.1086, + "step": 77926 + }, + { + "epoch": 5.76, + "learning_rate": 8.499261837302209e-08, + "loss": 1.0836, + "step": 77927 + }, + { + "epoch": 5.76, + "learning_rate": 8.494073263915603e-08, + "loss": 1.0108, + "step": 77928 + }, + { + "epoch": 5.76, + "learning_rate": 8.488886268000818e-08, + "loss": 0.9368, + "step": 77929 + }, + { + "epoch": 5.76, + "learning_rate": 8.483700849566068e-08, + "loss": 0.8815, + "step": 77930 + }, + { + "epoch": 5.76, + "learning_rate": 8.478517008619568e-08, + "loss": 0.9561, + "step": 77931 + }, + { + "epoch": 5.76, + "learning_rate": 8.473334745169536e-08, + "loss": 0.9539, + "step": 77932 + }, + { + "epoch": 5.76, + "learning_rate": 8.468154059224188e-08, + "loss": 0.9759, + "step": 77933 + }, + { + "epoch": 5.76, + "learning_rate": 8.462974950791846e-08, + "loss": 1.026, + "step": 77934 + }, + { + "epoch": 5.76, + "learning_rate": 8.457797419880842e-08, + "loss": 1.0133, + "step": 77935 + }, + { + "epoch": 5.76, + "learning_rate": 8.452621466499166e-08, + "loss": 0.9347, + "step": 77936 + }, + { + "epoch": 5.76, + "learning_rate": 8.447447090655148e-08, + "loss": 0.9485, + "step": 77937 + }, + { + "epoch": 5.76, + "learning_rate": 8.442274292357223e-08, + "loss": 1.166, + "step": 77938 + }, + { + "epoch": 5.76, + "learning_rate": 8.437103071613272e-08, + "loss": 0.9987, + "step": 77939 + }, + { + "epoch": 5.76, + "learning_rate": 8.431933428431849e-08, + "loss": 1.0817, + "step": 77940 + }, + { + "epoch": 5.76, + "learning_rate": 8.426765362820943e-08, + "loss": 1.0157, + "step": 77941 + }, + { + "epoch": 5.76, + "learning_rate": 8.421598874788883e-08, + "loss": 1.1, + "step": 77942 + }, + { + "epoch": 5.76, + "learning_rate": 8.416433964343885e-08, + "loss": 0.9727, + "step": 77943 + }, + { + "epoch": 5.76, + "learning_rate": 8.411270631494162e-08, + "loss": 0.8995, + "step": 77944 + }, + { + "epoch": 5.76, + "learning_rate": 8.40610887624782e-08, + "loss": 0.9321, + "step": 77945 + }, + { + "epoch": 5.76, + "learning_rate": 8.400948698613299e-08, + "loss": 1.0361, + "step": 77946 + }, + { + "epoch": 5.76, + "learning_rate": 8.395790098598699e-08, + "loss": 1.0342, + "step": 77947 + }, + { + "epoch": 5.76, + "learning_rate": 8.390633076212129e-08, + "loss": 0.9553, + "step": 77948 + }, + { + "epoch": 5.76, + "learning_rate": 8.385477631461914e-08, + "loss": 1.0754, + "step": 77949 + }, + { + "epoch": 5.76, + "learning_rate": 8.380323764356158e-08, + "loss": 1.025, + "step": 77950 + }, + { + "epoch": 5.76, + "learning_rate": 8.375171474903187e-08, + "loss": 0.9927, + "step": 77951 + }, + { + "epoch": 5.76, + "learning_rate": 8.370020763111109e-08, + "loss": 0.9933, + "step": 77952 + }, + { + "epoch": 5.76, + "learning_rate": 8.364871628988136e-08, + "loss": 0.9633, + "step": 77953 + }, + { + "epoch": 5.76, + "learning_rate": 8.359724072542374e-08, + "loss": 0.9513, + "step": 77954 + }, + { + "epoch": 5.76, + "learning_rate": 8.354578093782262e-08, + "loss": 0.9986, + "step": 77955 + }, + { + "epoch": 5.76, + "learning_rate": 8.349433692715792e-08, + "loss": 1.0784, + "step": 77956 + }, + { + "epoch": 5.76, + "learning_rate": 8.344290869351069e-08, + "loss": 0.9376, + "step": 77957 + }, + { + "epoch": 5.76, + "learning_rate": 8.33914962369653e-08, + "loss": 0.8862, + "step": 77958 + }, + { + "epoch": 5.76, + "learning_rate": 8.334009955760169e-08, + "loss": 0.8045, + "step": 77959 + }, + { + "epoch": 5.76, + "learning_rate": 8.328871865550204e-08, + "loss": 0.978, + "step": 77960 + }, + { + "epoch": 5.76, + "learning_rate": 8.323735353074735e-08, + "loss": 1.0218, + "step": 77961 + }, + { + "epoch": 5.76, + "learning_rate": 8.318600418342094e-08, + "loss": 0.9357, + "step": 77962 + }, + { + "epoch": 5.76, + "learning_rate": 8.313467061360381e-08, + "loss": 0.9683, + "step": 77963 + }, + { + "epoch": 5.76, + "learning_rate": 8.308335282137814e-08, + "loss": 1.0494, + "step": 77964 + }, + { + "epoch": 5.76, + "learning_rate": 8.303205080682386e-08, + "loss": 1.0781, + "step": 77965 + }, + { + "epoch": 5.76, + "learning_rate": 8.298076457002313e-08, + "loss": 1.0291, + "step": 77966 + }, + { + "epoch": 5.76, + "learning_rate": 8.292949411105922e-08, + "loss": 0.9439, + "step": 77967 + }, + { + "epoch": 5.76, + "learning_rate": 8.287823943001317e-08, + "loss": 1.0687, + "step": 77968 + }, + { + "epoch": 5.76, + "learning_rate": 8.28270005269649e-08, + "loss": 1.051, + "step": 77969 + }, + { + "epoch": 5.76, + "learning_rate": 8.277577740199661e-08, + "loss": 1.0225, + "step": 77970 + }, + { + "epoch": 5.76, + "learning_rate": 8.272457005519153e-08, + "loss": 0.9594, + "step": 77971 + }, + { + "epoch": 5.76, + "learning_rate": 8.267337848662848e-08, + "loss": 0.9612, + "step": 77972 + }, + { + "epoch": 5.76, + "learning_rate": 8.262220269639076e-08, + "loss": 0.9309, + "step": 77973 + }, + { + "epoch": 5.76, + "learning_rate": 8.25710426845594e-08, + "loss": 0.9242, + "step": 77974 + }, + { + "epoch": 5.76, + "learning_rate": 8.251989845121544e-08, + "loss": 0.9996, + "step": 77975 + }, + { + "epoch": 5.76, + "learning_rate": 8.246876999644104e-08, + "loss": 1.0513, + "step": 77976 + }, + { + "epoch": 5.76, + "learning_rate": 8.241765732031614e-08, + "loss": 0.8745, + "step": 77977 + }, + { + "epoch": 5.76, + "learning_rate": 8.23665604229229e-08, + "loss": 1.0169, + "step": 77978 + }, + { + "epoch": 5.76, + "learning_rate": 8.231547930434236e-08, + "loss": 1.044, + "step": 77979 + }, + { + "epoch": 5.76, + "learning_rate": 8.226441396465667e-08, + "loss": 1.1162, + "step": 77980 + }, + { + "epoch": 5.76, + "learning_rate": 8.221336440394578e-08, + "loss": 0.9868, + "step": 77981 + }, + { + "epoch": 5.76, + "learning_rate": 8.216233062229184e-08, + "loss": 0.9866, + "step": 77982 + }, + { + "epoch": 5.76, + "learning_rate": 8.21113126197759e-08, + "loss": 0.9464, + "step": 77983 + }, + { + "epoch": 5.76, + "learning_rate": 8.206031039647899e-08, + "loss": 0.8802, + "step": 77984 + }, + { + "epoch": 5.76, + "learning_rate": 8.200932395248218e-08, + "loss": 0.9586, + "step": 77985 + }, + { + "epoch": 5.76, + "learning_rate": 8.19583532878665e-08, + "loss": 1.1627, + "step": 77986 + }, + { + "epoch": 5.76, + "learning_rate": 8.1907398402713e-08, + "loss": 1.0034, + "step": 77987 + }, + { + "epoch": 5.76, + "learning_rate": 8.185645929710273e-08, + "loss": 1.0404, + "step": 77988 + }, + { + "epoch": 5.76, + "learning_rate": 8.180553597111784e-08, + "loss": 0.952, + "step": 77989 + }, + { + "epoch": 5.76, + "learning_rate": 8.175462842483716e-08, + "loss": 0.995, + "step": 77990 + }, + { + "epoch": 5.76, + "learning_rate": 8.170373665834397e-08, + "loss": 0.8724, + "step": 77991 + }, + { + "epoch": 5.76, + "learning_rate": 8.165286067171818e-08, + "loss": 0.9701, + "step": 77992 + }, + { + "epoch": 5.76, + "learning_rate": 8.160200046503974e-08, + "loss": 0.9388, + "step": 77993 + }, + { + "epoch": 5.76, + "learning_rate": 8.155115603839192e-08, + "loss": 0.9854, + "step": 77994 + }, + { + "epoch": 5.76, + "learning_rate": 8.150032739185354e-08, + "loss": 1.044, + "step": 77995 + }, + { + "epoch": 5.76, + "learning_rate": 8.144951452550676e-08, + "loss": 0.9668, + "step": 77996 + }, + { + "epoch": 5.76, + "learning_rate": 8.139871743943151e-08, + "loss": 1.1257, + "step": 77997 + }, + { + "epoch": 5.76, + "learning_rate": 8.134793613370884e-08, + "loss": 1.0521, + "step": 77998 + }, + { + "epoch": 5.76, + "learning_rate": 8.12971706084198e-08, + "loss": 0.9593, + "step": 77999 + }, + { + "epoch": 5.76, + "learning_rate": 8.124642086364542e-08, + "loss": 1.0397, + "step": 78000 + }, + { + "epoch": 5.76, + "learning_rate": 8.119568689946567e-08, + "loss": 1.019, + "step": 78001 + }, + { + "epoch": 5.76, + "learning_rate": 8.114496871596155e-08, + "loss": 1.0254, + "step": 78002 + }, + { + "epoch": 5.76, + "learning_rate": 8.109426631321416e-08, + "loss": 1.0855, + "step": 78003 + }, + { + "epoch": 5.76, + "learning_rate": 8.10435796913045e-08, + "loss": 1.0664, + "step": 78004 + }, + { + "epoch": 5.76, + "learning_rate": 8.099290885031141e-08, + "loss": 1.0293, + "step": 78005 + }, + { + "epoch": 5.76, + "learning_rate": 8.094225379031707e-08, + "loss": 1.0877, + "step": 78006 + }, + { + "epoch": 5.76, + "learning_rate": 8.089161451140137e-08, + "loss": 1.0448, + "step": 78007 + }, + { + "epoch": 5.76, + "learning_rate": 8.08409910136465e-08, + "loss": 1.0065, + "step": 78008 + }, + { + "epoch": 5.76, + "learning_rate": 8.079038329713018e-08, + "loss": 0.9786, + "step": 78009 + }, + { + "epoch": 5.76, + "learning_rate": 8.073979136193455e-08, + "loss": 0.9819, + "step": 78010 + }, + { + "epoch": 5.76, + "learning_rate": 8.068921520814065e-08, + "loss": 1.0731, + "step": 78011 + }, + { + "epoch": 5.76, + "learning_rate": 8.063865483582845e-08, + "loss": 0.9766, + "step": 78012 + }, + { + "epoch": 5.76, + "learning_rate": 8.058811024507785e-08, + "loss": 1.0516, + "step": 78013 + }, + { + "epoch": 5.76, + "learning_rate": 8.05375814359699e-08, + "loss": 0.7964, + "step": 78014 + }, + { + "epoch": 5.76, + "learning_rate": 8.048706840858456e-08, + "loss": 0.9758, + "step": 78015 + }, + { + "epoch": 5.76, + "learning_rate": 8.043657116300285e-08, + "loss": 0.9744, + "step": 78016 + }, + { + "epoch": 5.76, + "learning_rate": 8.038608969930362e-08, + "loss": 0.9947, + "step": 78017 + }, + { + "epoch": 5.76, + "learning_rate": 8.0335624017569e-08, + "loss": 1.0918, + "step": 78018 + }, + { + "epoch": 5.76, + "learning_rate": 8.028517411787785e-08, + "loss": 0.9851, + "step": 78019 + }, + { + "epoch": 5.76, + "learning_rate": 8.023474000031118e-08, + "loss": 0.9954, + "step": 78020 + }, + { + "epoch": 5.76, + "learning_rate": 8.018432166495004e-08, + "loss": 0.9802, + "step": 78021 + }, + { + "epoch": 5.76, + "learning_rate": 8.01339191118733e-08, + "loss": 0.9152, + "step": 78022 + }, + { + "epoch": 5.77, + "learning_rate": 8.008353234116084e-08, + "loss": 1.0337, + "step": 78023 + }, + { + "epoch": 5.77, + "learning_rate": 8.003316135289374e-08, + "loss": 1.1059, + "step": 78024 + }, + { + "epoch": 5.77, + "learning_rate": 7.998280614715304e-08, + "loss": 0.9468, + "step": 78025 + }, + { + "epoch": 5.77, + "learning_rate": 7.993246672401645e-08, + "loss": 1.0722, + "step": 78026 + }, + { + "epoch": 5.77, + "learning_rate": 7.988214308356612e-08, + "loss": 1.0056, + "step": 78027 + }, + { + "epoch": 5.77, + "learning_rate": 7.983183522587978e-08, + "loss": 0.8575, + "step": 78028 + }, + { + "epoch": 5.77, + "learning_rate": 7.97815431510418e-08, + "loss": 0.9261, + "step": 78029 + }, + { + "epoch": 5.77, + "learning_rate": 7.973126685912769e-08, + "loss": 1.0514, + "step": 78030 + }, + { + "epoch": 5.77, + "learning_rate": 7.968100635021958e-08, + "loss": 0.8721, + "step": 78031 + }, + { + "epoch": 5.77, + "learning_rate": 7.963076162439632e-08, + "loss": 1.006, + "step": 78032 + }, + { + "epoch": 5.77, + "learning_rate": 7.958053268174004e-08, + "loss": 0.9466, + "step": 78033 + }, + { + "epoch": 5.77, + "learning_rate": 7.953031952232848e-08, + "loss": 1.1142, + "step": 78034 + }, + { + "epoch": 5.77, + "learning_rate": 7.948012214624156e-08, + "loss": 1.015, + "step": 78035 + }, + { + "epoch": 5.77, + "learning_rate": 7.942994055356034e-08, + "loss": 0.9821, + "step": 78036 + }, + { + "epoch": 5.77, + "learning_rate": 7.937977474436476e-08, + "loss": 0.9926, + "step": 78037 + }, + { + "epoch": 5.77, + "learning_rate": 7.932962471873362e-08, + "loss": 1.0621, + "step": 78038 + }, + { + "epoch": 5.77, + "learning_rate": 7.927949047674688e-08, + "loss": 1.0289, + "step": 78039 + }, + { + "epoch": 5.77, + "learning_rate": 7.922937201848446e-08, + "loss": 1.0308, + "step": 78040 + }, + { + "epoch": 5.77, + "learning_rate": 7.917926934402632e-08, + "loss": 1.0445, + "step": 78041 + }, + { + "epoch": 5.77, + "learning_rate": 7.912918245345236e-08, + "loss": 1.032, + "step": 78042 + }, + { + "epoch": 5.77, + "learning_rate": 7.907911134684143e-08, + "loss": 1.0674, + "step": 78043 + }, + { + "epoch": 5.77, + "learning_rate": 7.902905602427347e-08, + "loss": 1.1435, + "step": 78044 + }, + { + "epoch": 5.77, + "learning_rate": 7.897901648582951e-08, + "loss": 1.0999, + "step": 78045 + }, + { + "epoch": 5.77, + "learning_rate": 7.892899273158616e-08, + "loss": 0.9992, + "step": 78046 + }, + { + "epoch": 5.77, + "learning_rate": 7.88789847616267e-08, + "loss": 0.8509, + "step": 78047 + }, + { + "epoch": 5.77, + "learning_rate": 7.882899257602772e-08, + "loss": 0.8834, + "step": 78048 + }, + { + "epoch": 5.77, + "learning_rate": 7.877901617487027e-08, + "loss": 1.053, + "step": 78049 + }, + { + "epoch": 5.77, + "learning_rate": 7.872905555823318e-08, + "loss": 0.9967, + "step": 78050 + }, + { + "epoch": 5.77, + "learning_rate": 7.867911072619638e-08, + "loss": 0.901, + "step": 78051 + }, + { + "epoch": 5.77, + "learning_rate": 7.862918167883982e-08, + "loss": 1.036, + "step": 78052 + }, + { + "epoch": 5.77, + "learning_rate": 7.85792684162412e-08, + "loss": 0.9444, + "step": 78053 + }, + { + "epoch": 5.77, + "learning_rate": 7.852937093848157e-08, + "loss": 0.9747, + "step": 78054 + }, + { + "epoch": 5.77, + "learning_rate": 7.847948924563975e-08, + "loss": 0.9414, + "step": 78055 + }, + { + "epoch": 5.77, + "learning_rate": 7.842962333779459e-08, + "loss": 1.0004, + "step": 78056 + }, + { + "epoch": 5.77, + "learning_rate": 7.837977321502599e-08, + "loss": 0.9454, + "step": 78057 + }, + { + "epoch": 5.77, + "learning_rate": 7.832993887741392e-08, + "loss": 1.157, + "step": 78058 + }, + { + "epoch": 5.77, + "learning_rate": 7.828012032503718e-08, + "loss": 0.9478, + "step": 78059 + }, + { + "epoch": 5.77, + "learning_rate": 7.82303175579735e-08, + "loss": 0.9658, + "step": 78060 + }, + { + "epoch": 5.77, + "learning_rate": 7.818053057630393e-08, + "loss": 0.9764, + "step": 78061 + }, + { + "epoch": 5.77, + "learning_rate": 7.813075938010839e-08, + "loss": 1.0039, + "step": 78062 + }, + { + "epoch": 5.77, + "learning_rate": 7.80810039694635e-08, + "loss": 0.9946, + "step": 78063 + }, + { + "epoch": 5.77, + "learning_rate": 7.803126434444919e-08, + "loss": 0.9843, + "step": 78064 + }, + { + "epoch": 5.77, + "learning_rate": 7.79815405051465e-08, + "loss": 0.9507, + "step": 78065 + }, + { + "epoch": 5.77, + "learning_rate": 7.793183245163316e-08, + "loss": 0.9285, + "step": 78066 + }, + { + "epoch": 5.77, + "learning_rate": 7.788214018398798e-08, + "loss": 0.9284, + "step": 78067 + }, + { + "epoch": 5.77, + "learning_rate": 7.783246370229092e-08, + "loss": 1.0299, + "step": 78068 + }, + { + "epoch": 5.77, + "learning_rate": 7.778280300661967e-08, + "loss": 1.0228, + "step": 78069 + }, + { + "epoch": 5.77, + "learning_rate": 7.773315809705418e-08, + "loss": 0.9993, + "step": 78070 + }, + { + "epoch": 5.77, + "learning_rate": 7.768352897367326e-08, + "loss": 1.005, + "step": 78071 + }, + { + "epoch": 5.77, + "learning_rate": 7.763391563655576e-08, + "loss": 0.9136, + "step": 78072 + }, + { + "epoch": 5.77, + "learning_rate": 7.758431808578049e-08, + "loss": 1.0319, + "step": 78073 + }, + { + "epoch": 5.77, + "learning_rate": 7.75347363214285e-08, + "loss": 1.0614, + "step": 78074 + }, + { + "epoch": 5.77, + "learning_rate": 7.748517034357417e-08, + "loss": 1.0181, + "step": 78075 + }, + { + "epoch": 5.77, + "learning_rate": 7.743562015230078e-08, + "loss": 0.9843, + "step": 78076 + }, + { + "epoch": 5.77, + "learning_rate": 7.738608574768491e-08, + "loss": 0.9706, + "step": 78077 + }, + { + "epoch": 5.77, + "learning_rate": 7.733656712980542e-08, + "loss": 0.8794, + "step": 78078 + }, + { + "epoch": 5.77, + "learning_rate": 7.728706429874222e-08, + "loss": 0.8562, + "step": 78079 + }, + { + "epoch": 5.77, + "learning_rate": 7.723757725457192e-08, + "loss": 1.0655, + "step": 78080 + }, + { + "epoch": 5.77, + "learning_rate": 7.718810599737558e-08, + "loss": 0.9974, + "step": 78081 + }, + { + "epoch": 5.77, + "learning_rate": 7.713865052723091e-08, + "loss": 1.0462, + "step": 78082 + }, + { + "epoch": 5.77, + "learning_rate": 7.708921084421673e-08, + "loss": 0.9391, + "step": 78083 + }, + { + "epoch": 5.77, + "learning_rate": 7.703978694841075e-08, + "loss": 1.0404, + "step": 78084 + }, + { + "epoch": 5.77, + "learning_rate": 7.699037883989402e-08, + "loss": 1.0409, + "step": 78085 + }, + { + "epoch": 5.77, + "learning_rate": 7.694098651874315e-08, + "loss": 0.9883, + "step": 78086 + }, + { + "epoch": 5.77, + "learning_rate": 7.689160998503698e-08, + "loss": 0.9333, + "step": 78087 + }, + { + "epoch": 5.77, + "learning_rate": 7.68422492388543e-08, + "loss": 1.0003, + "step": 78088 + }, + { + "epoch": 5.77, + "learning_rate": 7.679290428027287e-08, + "loss": 0.8609, + "step": 78089 + }, + { + "epoch": 5.77, + "learning_rate": 7.67435751093737e-08, + "loss": 0.9793, + "step": 78090 + }, + { + "epoch": 5.77, + "learning_rate": 7.669426172623229e-08, + "loss": 1.0312, + "step": 78091 + }, + { + "epoch": 5.77, + "learning_rate": 7.664496413092859e-08, + "loss": 1.0145, + "step": 78092 + }, + { + "epoch": 5.77, + "learning_rate": 7.65956823235403e-08, + "loss": 1.1453, + "step": 78093 + }, + { + "epoch": 5.77, + "learning_rate": 7.654641630414739e-08, + "loss": 1.0355, + "step": 78094 + }, + { + "epoch": 5.77, + "learning_rate": 7.649716607282642e-08, + "loss": 0.9882, + "step": 78095 + }, + { + "epoch": 5.77, + "learning_rate": 7.644793162965625e-08, + "loss": 0.9819, + "step": 78096 + }, + { + "epoch": 5.77, + "learning_rate": 7.639871297471568e-08, + "loss": 0.9579, + "step": 78097 + }, + { + "epoch": 5.77, + "learning_rate": 7.634951010808356e-08, + "loss": 1.008, + "step": 78098 + }, + { + "epoch": 5.77, + "learning_rate": 7.630032302983648e-08, + "loss": 0.975, + "step": 78099 + }, + { + "epoch": 5.77, + "learning_rate": 7.625115174005326e-08, + "loss": 0.9212, + "step": 78100 + }, + { + "epoch": 5.77, + "learning_rate": 7.620199623881274e-08, + "loss": 0.8986, + "step": 78101 + }, + { + "epoch": 5.77, + "learning_rate": 7.615285652619376e-08, + "loss": 0.9506, + "step": 78102 + }, + { + "epoch": 5.77, + "learning_rate": 7.610373260227289e-08, + "loss": 1.0035, + "step": 78103 + }, + { + "epoch": 5.77, + "learning_rate": 7.605462446712897e-08, + "loss": 0.9985, + "step": 78104 + }, + { + "epoch": 5.77, + "learning_rate": 7.600553212083971e-08, + "loss": 1.0119, + "step": 78105 + }, + { + "epoch": 5.77, + "learning_rate": 7.595645556348507e-08, + "loss": 0.9585, + "step": 78106 + }, + { + "epoch": 5.77, + "learning_rate": 7.590739479514054e-08, + "loss": 1.0513, + "step": 78107 + }, + { + "epoch": 5.77, + "learning_rate": 7.585834981588602e-08, + "loss": 1.0132, + "step": 78108 + }, + { + "epoch": 5.77, + "learning_rate": 7.580932062579816e-08, + "loss": 0.9733, + "step": 78109 + }, + { + "epoch": 5.77, + "learning_rate": 7.576030722495575e-08, + "loss": 0.9338, + "step": 78110 + }, + { + "epoch": 5.77, + "learning_rate": 7.571130961343765e-08, + "loss": 0.9525, + "step": 78111 + }, + { + "epoch": 5.77, + "learning_rate": 7.566232779131933e-08, + "loss": 0.9509, + "step": 78112 + }, + { + "epoch": 5.77, + "learning_rate": 7.561336175868073e-08, + "loss": 1.0389, + "step": 78113 + }, + { + "epoch": 5.77, + "learning_rate": 7.556441151560068e-08, + "loss": 0.9366, + "step": 78114 + }, + { + "epoch": 5.77, + "learning_rate": 7.551547706215468e-08, + "loss": 0.9851, + "step": 78115 + }, + { + "epoch": 5.77, + "learning_rate": 7.546655839842154e-08, + "loss": 0.9076, + "step": 78116 + }, + { + "epoch": 5.77, + "learning_rate": 7.541765552447899e-08, + "loss": 1.0624, + "step": 78117 + }, + { + "epoch": 5.77, + "learning_rate": 7.536876844040475e-08, + "loss": 0.9987, + "step": 78118 + }, + { + "epoch": 5.77, + "learning_rate": 7.531989714627763e-08, + "loss": 0.9551, + "step": 78119 + }, + { + "epoch": 5.77, + "learning_rate": 7.527104164217313e-08, + "loss": 1.0123, + "step": 78120 + }, + { + "epoch": 5.77, + "learning_rate": 7.522220192817231e-08, + "loss": 1.0583, + "step": 78121 + }, + { + "epoch": 5.77, + "learning_rate": 7.517337800434954e-08, + "loss": 0.9377, + "step": 78122 + }, + { + "epoch": 5.77, + "learning_rate": 7.512456987078476e-08, + "loss": 1.1715, + "step": 78123 + }, + { + "epoch": 5.77, + "learning_rate": 7.507577752755457e-08, + "loss": 0.9564, + "step": 78124 + }, + { + "epoch": 5.77, + "learning_rate": 7.502700097473559e-08, + "loss": 1.0428, + "step": 78125 + }, + { + "epoch": 5.77, + "learning_rate": 7.497824021240773e-08, + "loss": 0.9799, + "step": 78126 + }, + { + "epoch": 5.77, + "learning_rate": 7.492949524064764e-08, + "loss": 0.9234, + "step": 78127 + }, + { + "epoch": 5.77, + "learning_rate": 7.488076605953299e-08, + "loss": 1.0129, + "step": 78128 + }, + { + "epoch": 5.77, + "learning_rate": 7.483205266914039e-08, + "loss": 0.9175, + "step": 78129 + }, + { + "epoch": 5.77, + "learning_rate": 7.478335506954759e-08, + "loss": 0.9021, + "step": 78130 + }, + { + "epoch": 5.77, + "learning_rate": 7.473467326083339e-08, + "loss": 0.9445, + "step": 78131 + }, + { + "epoch": 5.77, + "learning_rate": 7.46860072430744e-08, + "loss": 0.9981, + "step": 78132 + }, + { + "epoch": 5.77, + "learning_rate": 7.463735701634723e-08, + "loss": 1.0012, + "step": 78133 + }, + { + "epoch": 5.77, + "learning_rate": 7.458872258072958e-08, + "loss": 0.9908, + "step": 78134 + }, + { + "epoch": 5.77, + "learning_rate": 7.45401039363014e-08, + "loss": 1.0607, + "step": 78135 + }, + { + "epoch": 5.77, + "learning_rate": 7.449150108313596e-08, + "loss": 0.9462, + "step": 78136 + }, + { + "epoch": 5.77, + "learning_rate": 7.444291402131321e-08, + "loss": 1.0465, + "step": 78137 + }, + { + "epoch": 5.77, + "learning_rate": 7.439434275090863e-08, + "loss": 1.0296, + "step": 78138 + }, + { + "epoch": 5.77, + "learning_rate": 7.434578727200215e-08, + "loss": 1.0616, + "step": 78139 + }, + { + "epoch": 5.77, + "learning_rate": 7.429724758466927e-08, + "loss": 1.0283, + "step": 78140 + }, + { + "epoch": 5.77, + "learning_rate": 7.424872368898661e-08, + "loss": 0.9813, + "step": 78141 + }, + { + "epoch": 5.77, + "learning_rate": 7.420021558503299e-08, + "loss": 1.0448, + "step": 78142 + }, + { + "epoch": 5.77, + "learning_rate": 7.415172327288501e-08, + "loss": 1.0287, + "step": 78143 + }, + { + "epoch": 5.77, + "learning_rate": 7.410324675261926e-08, + "loss": 0.9246, + "step": 78144 + }, + { + "epoch": 5.77, + "learning_rate": 7.405478602431238e-08, + "loss": 0.9634, + "step": 78145 + }, + { + "epoch": 5.77, + "learning_rate": 7.400634108804316e-08, + "loss": 1.0175, + "step": 78146 + }, + { + "epoch": 5.77, + "learning_rate": 7.395791194388824e-08, + "loss": 1.1096, + "step": 78147 + }, + { + "epoch": 5.77, + "learning_rate": 7.39094985919242e-08, + "loss": 1.0649, + "step": 78148 + }, + { + "epoch": 5.77, + "learning_rate": 7.386110103222766e-08, + "loss": 0.9919, + "step": 78149 + }, + { + "epoch": 5.77, + "learning_rate": 7.381271926487632e-08, + "loss": 1.0406, + "step": 78150 + }, + { + "epoch": 5.77, + "learning_rate": 7.37643532899468e-08, + "loss": 1.1068, + "step": 78151 + }, + { + "epoch": 5.77, + "learning_rate": 7.37160031075157e-08, + "loss": 0.9539, + "step": 78152 + }, + { + "epoch": 5.77, + "learning_rate": 7.366766871766073e-08, + "loss": 0.9922, + "step": 78153 + }, + { + "epoch": 5.77, + "learning_rate": 7.36193501204585e-08, + "loss": 1.015, + "step": 78154 + }, + { + "epoch": 5.77, + "learning_rate": 7.357104731598674e-08, + "loss": 0.9696, + "step": 78155 + }, + { + "epoch": 5.77, + "learning_rate": 7.352276030431982e-08, + "loss": 1.0126, + "step": 78156 + }, + { + "epoch": 5.77, + "learning_rate": 7.347448908553768e-08, + "loss": 0.9393, + "step": 78157 + }, + { + "epoch": 5.78, + "learning_rate": 7.342623365971469e-08, + "loss": 1.0935, + "step": 78158 + }, + { + "epoch": 5.78, + "learning_rate": 7.337799402692858e-08, + "loss": 0.9126, + "step": 78159 + }, + { + "epoch": 5.78, + "learning_rate": 7.332977018725595e-08, + "loss": 0.9867, + "step": 78160 + }, + { + "epoch": 5.78, + "learning_rate": 7.328156214077453e-08, + "loss": 0.8783, + "step": 78161 + }, + { + "epoch": 5.78, + "learning_rate": 7.32333698875598e-08, + "loss": 0.7995, + "step": 78162 + }, + { + "epoch": 5.78, + "learning_rate": 7.318519342768727e-08, + "loss": 0.9217, + "step": 78163 + }, + { + "epoch": 5.78, + "learning_rate": 7.313703276123796e-08, + "loss": 1.0493, + "step": 78164 + }, + { + "epoch": 5.78, + "learning_rate": 7.308888788828295e-08, + "loss": 1.0239, + "step": 78165 + }, + { + "epoch": 5.78, + "learning_rate": 7.304075880890215e-08, + "loss": 0.921, + "step": 78166 + }, + { + "epoch": 5.78, + "learning_rate": 7.29926455231722e-08, + "loss": 0.9861, + "step": 78167 + }, + { + "epoch": 5.78, + "learning_rate": 7.294454803116857e-08, + "loss": 0.9209, + "step": 78168 + }, + { + "epoch": 5.78, + "learning_rate": 7.289646633296787e-08, + "loss": 0.9959, + "step": 78169 + }, + { + "epoch": 5.78, + "learning_rate": 7.284840042864671e-08, + "loss": 0.9242, + "step": 78170 + }, + { + "epoch": 5.78, + "learning_rate": 7.280035031828281e-08, + "loss": 0.957, + "step": 78171 + }, + { + "epoch": 5.78, + "learning_rate": 7.275231600195054e-08, + "loss": 1.078, + "step": 78172 + }, + { + "epoch": 5.78, + "learning_rate": 7.270429747972762e-08, + "loss": 0.9113, + "step": 78173 + }, + { + "epoch": 5.78, + "learning_rate": 7.265629475168956e-08, + "loss": 1.1287, + "step": 78174 + }, + { + "epoch": 5.78, + "learning_rate": 7.260830781791406e-08, + "loss": 1.0934, + "step": 78175 + }, + { + "epoch": 5.78, + "learning_rate": 7.256033667847662e-08, + "loss": 1.09, + "step": 78176 + }, + { + "epoch": 5.78, + "learning_rate": 7.251238133345273e-08, + "loss": 0.9704, + "step": 78177 + }, + { + "epoch": 5.78, + "learning_rate": 7.246444178292011e-08, + "loss": 0.9192, + "step": 78178 + }, + { + "epoch": 5.78, + "learning_rate": 7.241651802695537e-08, + "loss": 1.0566, + "step": 78179 + }, + { + "epoch": 5.78, + "learning_rate": 7.236861006563401e-08, + "loss": 1.0314, + "step": 78180 + }, + { + "epoch": 5.78, + "learning_rate": 7.232071789903039e-08, + "loss": 0.9768, + "step": 78181 + }, + { + "epoch": 5.78, + "learning_rate": 7.227284152722336e-08, + "loss": 0.9295, + "step": 78182 + }, + { + "epoch": 5.78, + "learning_rate": 7.22249809502884e-08, + "loss": 0.9644, + "step": 78183 + }, + { + "epoch": 5.78, + "learning_rate": 7.217713616830213e-08, + "loss": 0.9918, + "step": 78184 + }, + { + "epoch": 5.78, + "learning_rate": 7.212930718133893e-08, + "loss": 1.0037, + "step": 78185 + }, + { + "epoch": 5.78, + "learning_rate": 7.208149398947539e-08, + "loss": 0.9334, + "step": 78186 + }, + { + "epoch": 5.78, + "learning_rate": 7.203369659278925e-08, + "loss": 1.0863, + "step": 78187 + }, + { + "epoch": 5.78, + "learning_rate": 7.198591499135599e-08, + "loss": 0.994, + "step": 78188 + }, + { + "epoch": 5.78, + "learning_rate": 7.193814918525e-08, + "loss": 0.9685, + "step": 78189 + }, + { + "epoch": 5.78, + "learning_rate": 7.189039917454788e-08, + "loss": 1.0621, + "step": 78190 + }, + { + "epoch": 5.78, + "learning_rate": 7.184266495932735e-08, + "loss": 1.0531, + "step": 78191 + }, + { + "epoch": 5.78, + "learning_rate": 7.179494653966279e-08, + "loss": 1.0399, + "step": 78192 + }, + { + "epoch": 5.78, + "learning_rate": 7.174724391563082e-08, + "loss": 1.0055, + "step": 78193 + }, + { + "epoch": 5.78, + "learning_rate": 7.16995570873058e-08, + "loss": 1.0843, + "step": 78194 + }, + { + "epoch": 5.78, + "learning_rate": 7.165188605476548e-08, + "loss": 1.0514, + "step": 78195 + }, + { + "epoch": 5.78, + "learning_rate": 7.16042308180842e-08, + "loss": 1.0651, + "step": 78196 + }, + { + "epoch": 5.78, + "learning_rate": 7.155659137733973e-08, + "loss": 1.0854, + "step": 78197 + }, + { + "epoch": 5.78, + "learning_rate": 7.15089677326064e-08, + "loss": 1.0643, + "step": 78198 + }, + { + "epoch": 5.78, + "learning_rate": 7.146135988395863e-08, + "loss": 1.0104, + "step": 78199 + }, + { + "epoch": 5.78, + "learning_rate": 7.141376783147525e-08, + "loss": 0.9316, + "step": 78200 + }, + { + "epoch": 5.78, + "learning_rate": 7.136619157523062e-08, + "loss": 0.9697, + "step": 78201 + }, + { + "epoch": 5.78, + "learning_rate": 7.131863111530025e-08, + "loss": 0.8746, + "step": 78202 + }, + { + "epoch": 5.78, + "learning_rate": 7.127108645175962e-08, + "loss": 0.9805, + "step": 78203 + }, + { + "epoch": 5.78, + "learning_rate": 7.122355758468425e-08, + "loss": 0.9956, + "step": 78204 + }, + { + "epoch": 5.78, + "learning_rate": 7.117604451415072e-08, + "loss": 0.9462, + "step": 78205 + }, + { + "epoch": 5.78, + "learning_rate": 7.112854724023454e-08, + "loss": 0.946, + "step": 78206 + }, + { + "epoch": 5.78, + "learning_rate": 7.108106576301011e-08, + "loss": 1.0394, + "step": 78207 + }, + { + "epoch": 5.78, + "learning_rate": 7.10336000825529e-08, + "loss": 0.9583, + "step": 78208 + }, + { + "epoch": 5.78, + "learning_rate": 7.098615019894062e-08, + "loss": 0.9832, + "step": 78209 + }, + { + "epoch": 5.78, + "learning_rate": 7.093871611224657e-08, + "loss": 0.9663, + "step": 78210 + }, + { + "epoch": 5.78, + "learning_rate": 7.089129782254622e-08, + "loss": 1.1049, + "step": 78211 + }, + { + "epoch": 5.78, + "learning_rate": 7.08438953299173e-08, + "loss": 0.9605, + "step": 78212 + }, + { + "epoch": 5.78, + "learning_rate": 7.07965086344331e-08, + "loss": 0.9999, + "step": 78213 + }, + { + "epoch": 5.78, + "learning_rate": 7.07491377361691e-08, + "loss": 1.0612, + "step": 78214 + }, + { + "epoch": 5.78, + "learning_rate": 7.070178263520078e-08, + "loss": 0.9947, + "step": 78215 + }, + { + "epoch": 5.78, + "learning_rate": 7.065444333160477e-08, + "loss": 0.9486, + "step": 78216 + }, + { + "epoch": 5.78, + "learning_rate": 7.060711982545543e-08, + "loss": 1.0764, + "step": 78217 + }, + { + "epoch": 5.78, + "learning_rate": 7.055981211682827e-08, + "loss": 0.9099, + "step": 78218 + }, + { + "epoch": 5.78, + "learning_rate": 7.051252020579769e-08, + "loss": 1.0741, + "step": 78219 + }, + { + "epoch": 5.78, + "learning_rate": 7.046524409244027e-08, + "loss": 0.9808, + "step": 78220 + }, + { + "epoch": 5.78, + "learning_rate": 7.04179837768304e-08, + "loss": 0.9294, + "step": 78221 + }, + { + "epoch": 5.78, + "learning_rate": 7.037073925904358e-08, + "loss": 0.944, + "step": 78222 + }, + { + "epoch": 5.78, + "learning_rate": 7.03235105391542e-08, + "loss": 0.9557, + "step": 78223 + }, + { + "epoch": 5.78, + "learning_rate": 7.027629761723776e-08, + "loss": 1.1209, + "step": 78224 + }, + { + "epoch": 5.78, + "learning_rate": 7.022910049337084e-08, + "loss": 0.9683, + "step": 78225 + }, + { + "epoch": 5.78, + "learning_rate": 7.018191916762673e-08, + "loss": 1.0241, + "step": 78226 + }, + { + "epoch": 5.78, + "learning_rate": 7.013475364008205e-08, + "loss": 1.0742, + "step": 78227 + }, + { + "epoch": 5.78, + "learning_rate": 7.008760391080893e-08, + "loss": 1.0307, + "step": 78228 + }, + { + "epoch": 5.78, + "learning_rate": 7.004046997988623e-08, + "loss": 1.0249, + "step": 78229 + }, + { + "epoch": 5.78, + "learning_rate": 6.999335184738609e-08, + "loss": 0.9041, + "step": 78230 + }, + { + "epoch": 5.78, + "learning_rate": 6.994624951338513e-08, + "loss": 1.0244, + "step": 78231 + }, + { + "epoch": 5.78, + "learning_rate": 6.989916297795662e-08, + "loss": 0.979, + "step": 78232 + }, + { + "epoch": 5.78, + "learning_rate": 6.985209224117717e-08, + "loss": 1.0839, + "step": 78233 + }, + { + "epoch": 5.78, + "learning_rate": 6.980503730312115e-08, + "loss": 0.9804, + "step": 78234 + }, + { + "epoch": 5.78, + "learning_rate": 6.975799816386297e-08, + "loss": 1.0864, + "step": 78235 + }, + { + "epoch": 5.78, + "learning_rate": 6.9710974823477e-08, + "loss": 1.0458, + "step": 78236 + }, + { + "epoch": 5.78, + "learning_rate": 6.966396728203984e-08, + "loss": 1.0157, + "step": 78237 + }, + { + "epoch": 5.78, + "learning_rate": 6.961697553962477e-08, + "loss": 0.9784, + "step": 78238 + }, + { + "epoch": 5.78, + "learning_rate": 6.956999959630728e-08, + "loss": 0.9847, + "step": 78239 + }, + { + "epoch": 5.78, + "learning_rate": 6.952303945216177e-08, + "loss": 1.0454, + "step": 78240 + }, + { + "epoch": 5.78, + "learning_rate": 6.947609510726261e-08, + "loss": 0.9595, + "step": 78241 + }, + { + "epoch": 5.78, + "learning_rate": 6.942916656168531e-08, + "loss": 0.8677, + "step": 78242 + }, + { + "epoch": 5.78, + "learning_rate": 6.938225381550423e-08, + "loss": 1.0019, + "step": 78243 + }, + { + "epoch": 5.78, + "learning_rate": 6.933535686879378e-08, + "loss": 1.0226, + "step": 78244 + }, + { + "epoch": 5.78, + "learning_rate": 6.928847572162834e-08, + "loss": 1.0723, + "step": 78245 + }, + { + "epoch": 5.78, + "learning_rate": 6.924161037408339e-08, + "loss": 0.9006, + "step": 78246 + }, + { + "epoch": 5.78, + "learning_rate": 6.919476082623222e-08, + "loss": 0.929, + "step": 78247 + }, + { + "epoch": 5.78, + "learning_rate": 6.914792707815032e-08, + "loss": 1.0961, + "step": 78248 + }, + { + "epoch": 5.78, + "learning_rate": 6.910110912991319e-08, + "loss": 1.028, + "step": 78249 + }, + { + "epoch": 5.78, + "learning_rate": 6.905430698159299e-08, + "loss": 1.1106, + "step": 78250 + }, + { + "epoch": 5.78, + "learning_rate": 6.900752063326521e-08, + "loss": 0.97, + "step": 78251 + }, + { + "epoch": 5.78, + "learning_rate": 6.896075008500424e-08, + "loss": 1.0458, + "step": 78252 + }, + { + "epoch": 5.78, + "learning_rate": 6.891399533688558e-08, + "loss": 1.0004, + "step": 78253 + }, + { + "epoch": 5.78, + "learning_rate": 6.886725638898251e-08, + "loss": 0.9536, + "step": 78254 + }, + { + "epoch": 5.78, + "learning_rate": 6.882053324136828e-08, + "loss": 1.0005, + "step": 78255 + }, + { + "epoch": 5.78, + "learning_rate": 6.877382589411951e-08, + "loss": 0.9536, + "step": 78256 + }, + { + "epoch": 5.78, + "learning_rate": 6.872713434730949e-08, + "loss": 1.0259, + "step": 78257 + }, + { + "epoch": 5.78, + "learning_rate": 6.868045860101258e-08, + "loss": 1.0508, + "step": 78258 + }, + { + "epoch": 5.78, + "learning_rate": 6.863379865530206e-08, + "loss": 0.9331, + "step": 78259 + }, + { + "epoch": 5.78, + "learning_rate": 6.858715451025344e-08, + "loss": 0.8919, + "step": 78260 + }, + { + "epoch": 5.78, + "learning_rate": 6.854052616594109e-08, + "loss": 1.0258, + "step": 78261 + }, + { + "epoch": 5.78, + "learning_rate": 6.84939136224394e-08, + "loss": 1.0647, + "step": 78262 + }, + { + "epoch": 5.78, + "learning_rate": 6.844731687982053e-08, + "loss": 0.9251, + "step": 78263 + }, + { + "epoch": 5.78, + "learning_rate": 6.840073593815999e-08, + "loss": 1.071, + "step": 78264 + }, + { + "epoch": 5.78, + "learning_rate": 6.835417079753214e-08, + "loss": 0.9518, + "step": 78265 + }, + { + "epoch": 5.78, + "learning_rate": 6.830762145801028e-08, + "loss": 0.9778, + "step": 78266 + }, + { + "epoch": 5.78, + "learning_rate": 6.826108791966879e-08, + "loss": 1.0278, + "step": 78267 + }, + { + "epoch": 5.78, + "learning_rate": 6.821457018258204e-08, + "loss": 1.0563, + "step": 78268 + }, + { + "epoch": 5.78, + "learning_rate": 6.816806824682442e-08, + "loss": 0.9426, + "step": 78269 + }, + { + "epoch": 5.78, + "learning_rate": 6.812158211246811e-08, + "loss": 1.0021, + "step": 78270 + }, + { + "epoch": 5.78, + "learning_rate": 6.80751117795897e-08, + "loss": 0.7853, + "step": 78271 + }, + { + "epoch": 5.78, + "learning_rate": 6.802865724826024e-08, + "loss": 0.9938, + "step": 78272 + }, + { + "epoch": 5.78, + "learning_rate": 6.798221851855635e-08, + "loss": 1.0513, + "step": 78273 + }, + { + "epoch": 5.78, + "learning_rate": 6.79357955905502e-08, + "loss": 0.9257, + "step": 78274 + }, + { + "epoch": 5.78, + "learning_rate": 6.788938846431614e-08, + "loss": 0.9559, + "step": 78275 + }, + { + "epoch": 5.78, + "learning_rate": 6.784299713992748e-08, + "loss": 1.0461, + "step": 78276 + }, + { + "epoch": 5.78, + "learning_rate": 6.779662161745859e-08, + "loss": 0.9551, + "step": 78277 + }, + { + "epoch": 5.78, + "learning_rate": 6.775026189698386e-08, + "loss": 1.0186, + "step": 78278 + }, + { + "epoch": 5.78, + "learning_rate": 6.770391797857546e-08, + "loss": 0.9968, + "step": 78279 + }, + { + "epoch": 5.78, + "learning_rate": 6.765758986230886e-08, + "loss": 0.8401, + "step": 78280 + }, + { + "epoch": 5.78, + "learning_rate": 6.761127754825624e-08, + "loss": 1.0454, + "step": 78281 + }, + { + "epoch": 5.78, + "learning_rate": 6.7564981036492e-08, + "loss": 0.9601, + "step": 78282 + }, + { + "epoch": 5.78, + "learning_rate": 6.751870032709052e-08, + "loss": 0.9084, + "step": 78283 + }, + { + "epoch": 5.78, + "learning_rate": 6.747243542012393e-08, + "loss": 1.106, + "step": 78284 + }, + { + "epoch": 5.78, + "learning_rate": 6.742618631566666e-08, + "loss": 0.9839, + "step": 78285 + }, + { + "epoch": 5.78, + "learning_rate": 6.737995301379307e-08, + "loss": 0.9803, + "step": 78286 + }, + { + "epoch": 5.78, + "learning_rate": 6.733373551457534e-08, + "loss": 0.9224, + "step": 78287 + }, + { + "epoch": 5.78, + "learning_rate": 6.728753381808672e-08, + "loss": 0.9421, + "step": 78288 + }, + { + "epoch": 5.78, + "learning_rate": 6.724134792440163e-08, + "loss": 1.0177, + "step": 78289 + }, + { + "epoch": 5.78, + "learning_rate": 6.719517783359442e-08, + "loss": 1.0559, + "step": 78290 + }, + { + "epoch": 5.78, + "learning_rate": 6.714902354573727e-08, + "loss": 0.962, + "step": 78291 + }, + { + "epoch": 5.78, + "learning_rate": 6.710288506090345e-08, + "loss": 1.0627, + "step": 78292 + }, + { + "epoch": 5.79, + "learning_rate": 6.705676237916736e-08, + "loss": 0.97, + "step": 78293 + }, + { + "epoch": 5.79, + "learning_rate": 6.701065550060115e-08, + "loss": 0.9249, + "step": 78294 + }, + { + "epoch": 5.79, + "learning_rate": 6.696456442528032e-08, + "loss": 0.8887, + "step": 78295 + }, + { + "epoch": 5.79, + "learning_rate": 6.691848915327592e-08, + "loss": 0.9229, + "step": 78296 + }, + { + "epoch": 5.79, + "learning_rate": 6.687242968466123e-08, + "loss": 0.9784, + "step": 78297 + }, + { + "epoch": 5.79, + "learning_rate": 6.682638601951064e-08, + "loss": 0.8913, + "step": 78298 + }, + { + "epoch": 5.79, + "learning_rate": 6.678035815789852e-08, + "loss": 0.99, + "step": 78299 + }, + { + "epoch": 5.79, + "learning_rate": 6.673434609989593e-08, + "loss": 1.0608, + "step": 78300 + }, + { + "epoch": 5.79, + "learning_rate": 6.668834984557615e-08, + "loss": 0.9715, + "step": 78301 + }, + { + "epoch": 5.79, + "learning_rate": 6.664236939501356e-08, + "loss": 0.999, + "step": 78302 + }, + { + "epoch": 5.79, + "learning_rate": 6.659640474828144e-08, + "loss": 0.9676, + "step": 78303 + }, + { + "epoch": 5.79, + "learning_rate": 6.655045590545196e-08, + "loss": 1.0745, + "step": 78304 + }, + { + "epoch": 5.79, + "learning_rate": 6.650452286659836e-08, + "loss": 0.9502, + "step": 78305 + }, + { + "epoch": 5.79, + "learning_rate": 6.645860563179396e-08, + "loss": 0.9562, + "step": 78306 + }, + { + "epoch": 5.79, + "learning_rate": 6.641270420111201e-08, + "loss": 0.873, + "step": 78307 + }, + { + "epoch": 5.79, + "learning_rate": 6.636681857462468e-08, + "loss": 0.9559, + "step": 78308 + }, + { + "epoch": 5.79, + "learning_rate": 6.632094875240636e-08, + "loss": 1.0154, + "step": 78309 + }, + { + "epoch": 5.79, + "learning_rate": 6.62750947345292e-08, + "loss": 1.0542, + "step": 78310 + }, + { + "epoch": 5.79, + "learning_rate": 6.622925652106537e-08, + "loss": 1.1285, + "step": 78311 + }, + { + "epoch": 5.79, + "learning_rate": 6.618343411209039e-08, + "loss": 1.053, + "step": 78312 + }, + { + "epoch": 5.79, + "learning_rate": 6.613762750767416e-08, + "loss": 1.0599, + "step": 78313 + }, + { + "epoch": 5.79, + "learning_rate": 6.60918367078911e-08, + "loss": 0.9645, + "step": 78314 + }, + { + "epoch": 5.79, + "learning_rate": 6.604606171281335e-08, + "loss": 0.9974, + "step": 78315 + }, + { + "epoch": 5.79, + "learning_rate": 6.600030252251532e-08, + "loss": 0.9698, + "step": 78316 + }, + { + "epoch": 5.79, + "learning_rate": 6.595455913706694e-08, + "loss": 0.8906, + "step": 78317 + }, + { + "epoch": 5.79, + "learning_rate": 6.59088315565437e-08, + "loss": 0.9408, + "step": 78318 + }, + { + "epoch": 5.79, + "learning_rate": 6.586311978101779e-08, + "loss": 1.0968, + "step": 78319 + }, + { + "epoch": 5.79, + "learning_rate": 6.581742381056134e-08, + "loss": 1.0288, + "step": 78320 + }, + { + "epoch": 5.79, + "learning_rate": 6.577174364524763e-08, + "loss": 0.9508, + "step": 78321 + }, + { + "epoch": 5.79, + "learning_rate": 6.572607928514774e-08, + "loss": 0.9954, + "step": 78322 + }, + { + "epoch": 5.79, + "learning_rate": 6.568043073033603e-08, + "loss": 0.853, + "step": 78323 + }, + { + "epoch": 5.79, + "learning_rate": 6.563479798088468e-08, + "loss": 1.1199, + "step": 78324 + }, + { + "epoch": 5.79, + "learning_rate": 6.558918103686695e-08, + "loss": 1.0785, + "step": 78325 + }, + { + "epoch": 5.79, + "learning_rate": 6.554357989835281e-08, + "loss": 0.8491, + "step": 78326 + }, + { + "epoch": 5.79, + "learning_rate": 6.549799456541883e-08, + "loss": 0.9956, + "step": 78327 + }, + { + "epoch": 5.79, + "learning_rate": 6.545242503813388e-08, + "loss": 0.9854, + "step": 78328 + }, + { + "epoch": 5.79, + "learning_rate": 6.540687131657342e-08, + "loss": 0.9867, + "step": 78329 + }, + { + "epoch": 5.79, + "learning_rate": 6.536133340080742e-08, + "loss": 1.0562, + "step": 78330 + }, + { + "epoch": 5.79, + "learning_rate": 6.531581129090914e-08, + "loss": 1.0866, + "step": 78331 + }, + { + "epoch": 5.79, + "learning_rate": 6.527030498695186e-08, + "loss": 1.0522, + "step": 78332 + }, + { + "epoch": 5.79, + "learning_rate": 6.522481448900775e-08, + "loss": 1.0636, + "step": 78333 + }, + { + "epoch": 5.79, + "learning_rate": 6.517933979714786e-08, + "loss": 0.9994, + "step": 78334 + }, + { + "epoch": 5.79, + "learning_rate": 6.513388091144657e-08, + "loss": 1.0426, + "step": 78335 + }, + { + "epoch": 5.79, + "learning_rate": 6.508843783197383e-08, + "loss": 0.9087, + "step": 78336 + }, + { + "epoch": 5.79, + "learning_rate": 6.504301055880403e-08, + "loss": 1.051, + "step": 78337 + }, + { + "epoch": 5.79, + "learning_rate": 6.49975990920082e-08, + "loss": 0.9388, + "step": 78338 + }, + { + "epoch": 5.79, + "learning_rate": 6.495220343165964e-08, + "loss": 1.1084, + "step": 78339 + }, + { + "epoch": 5.79, + "learning_rate": 6.490682357783052e-08, + "loss": 0.9224, + "step": 78340 + }, + { + "epoch": 5.79, + "learning_rate": 6.486145953059187e-08, + "loss": 0.9475, + "step": 78341 + }, + { + "epoch": 5.79, + "learning_rate": 6.481611129001586e-08, + "loss": 0.9743, + "step": 78342 + }, + { + "epoch": 5.79, + "learning_rate": 6.477077885617577e-08, + "loss": 0.9431, + "step": 78343 + }, + { + "epoch": 5.79, + "learning_rate": 6.472546222914266e-08, + "loss": 1.04, + "step": 78344 + }, + { + "epoch": 5.79, + "learning_rate": 6.468016140899091e-08, + "loss": 1.0047, + "step": 78345 + }, + { + "epoch": 5.79, + "learning_rate": 6.463487639578825e-08, + "loss": 0.8544, + "step": 78346 + }, + { + "epoch": 5.79, + "learning_rate": 6.458960718961238e-08, + "loss": 1.0638, + "step": 78347 + }, + { + "epoch": 5.79, + "learning_rate": 6.454435379052993e-08, + "loss": 1.0472, + "step": 78348 + }, + { + "epoch": 5.79, + "learning_rate": 6.449911619861637e-08, + "loss": 0.7745, + "step": 78349 + }, + { + "epoch": 5.79, + "learning_rate": 6.445389441394168e-08, + "loss": 0.9261, + "step": 78350 + }, + { + "epoch": 5.79, + "learning_rate": 6.440868843658021e-08, + "loss": 0.9764, + "step": 78351 + }, + { + "epoch": 5.79, + "learning_rate": 6.436349826660083e-08, + "loss": 0.9131, + "step": 78352 + }, + { + "epoch": 5.79, + "learning_rate": 6.431832390407678e-08, + "loss": 0.9873, + "step": 78353 + }, + { + "epoch": 5.79, + "learning_rate": 6.427316534908135e-08, + "loss": 1.0589, + "step": 78354 + }, + { + "epoch": 5.79, + "learning_rate": 6.422802260168448e-08, + "loss": 0.9916, + "step": 78355 + }, + { + "epoch": 5.79, + "learning_rate": 6.418289566195945e-08, + "loss": 1.0218, + "step": 78356 + }, + { + "epoch": 5.79, + "learning_rate": 6.413778452997732e-08, + "loss": 1.0482, + "step": 78357 + }, + { + "epoch": 5.79, + "learning_rate": 6.409268920580803e-08, + "loss": 0.8987, + "step": 78358 + }, + { + "epoch": 5.79, + "learning_rate": 6.404760968952705e-08, + "loss": 0.9082, + "step": 78359 + }, + { + "epoch": 5.79, + "learning_rate": 6.400254598120326e-08, + "loss": 0.9841, + "step": 78360 + }, + { + "epoch": 5.79, + "learning_rate": 6.395749808090879e-08, + "loss": 0.9328, + "step": 78361 + }, + { + "epoch": 5.79, + "learning_rate": 6.391246598871692e-08, + "loss": 0.9896, + "step": 78362 + }, + { + "epoch": 5.79, + "learning_rate": 6.386744970469761e-08, + "loss": 0.9275, + "step": 78363 + }, + { + "epoch": 5.79, + "learning_rate": 6.3822449228923e-08, + "loss": 0.9166, + "step": 78364 + }, + { + "epoch": 5.79, + "learning_rate": 6.377746456146416e-08, + "loss": 1.1279, + "step": 78365 + }, + { + "epoch": 5.79, + "learning_rate": 6.373249570239437e-08, + "loss": 1.0076, + "step": 78366 + }, + { + "epoch": 5.79, + "learning_rate": 6.368754265178357e-08, + "loss": 0.985, + "step": 78367 + }, + { + "epoch": 5.79, + "learning_rate": 6.36426054097039e-08, + "loss": 1.0415, + "step": 78368 + }, + { + "epoch": 5.79, + "learning_rate": 6.359768397622535e-08, + "loss": 0.9645, + "step": 78369 + }, + { + "epoch": 5.79, + "learning_rate": 6.355277835142226e-08, + "loss": 1.0249, + "step": 78370 + }, + { + "epoch": 5.79, + "learning_rate": 6.350788853536349e-08, + "loss": 1.0045, + "step": 78371 + }, + { + "epoch": 5.79, + "learning_rate": 6.346301452812232e-08, + "loss": 0.894, + "step": 78372 + }, + { + "epoch": 5.79, + "learning_rate": 6.341815632976867e-08, + "loss": 0.9523, + "step": 78373 + }, + { + "epoch": 5.79, + "learning_rate": 6.337331394037582e-08, + "loss": 1.0592, + "step": 78374 + }, + { + "epoch": 5.79, + "learning_rate": 6.332848736001263e-08, + "loss": 0.9813, + "step": 78375 + }, + { + "epoch": 5.79, + "learning_rate": 6.328367658875234e-08, + "loss": 1.0888, + "step": 78376 + }, + { + "epoch": 5.79, + "learning_rate": 6.323888162666492e-08, + "loss": 1.0315, + "step": 78377 + }, + { + "epoch": 5.79, + "learning_rate": 6.319410247382252e-08, + "loss": 0.9661, + "step": 78378 + }, + { + "epoch": 5.79, + "learning_rate": 6.31493391302962e-08, + "loss": 0.9835, + "step": 78379 + }, + { + "epoch": 5.79, + "learning_rate": 6.3104591596157e-08, + "loss": 0.926, + "step": 78380 + }, + { + "epoch": 5.79, + "learning_rate": 6.3059859871476e-08, + "loss": 0.9851, + "step": 78381 + }, + { + "epoch": 5.79, + "learning_rate": 6.301514395632425e-08, + "loss": 1.1256, + "step": 78382 + }, + { + "epoch": 5.79, + "learning_rate": 6.297044385077389e-08, + "loss": 1.0397, + "step": 78383 + }, + { + "epoch": 5.79, + "learning_rate": 6.2925759554896e-08, + "loss": 1.1292, + "step": 78384 + }, + { + "epoch": 5.79, + "learning_rate": 6.28810910687605e-08, + "loss": 1.0056, + "step": 78385 + }, + { + "epoch": 5.79, + "learning_rate": 6.283643839243847e-08, + "loss": 0.9458, + "step": 78386 + }, + { + "epoch": 5.79, + "learning_rate": 6.279180152600206e-08, + "loss": 1.0089, + "step": 78387 + }, + { + "epoch": 5.79, + "learning_rate": 6.274718046952233e-08, + "loss": 0.872, + "step": 78388 + }, + { + "epoch": 5.79, + "learning_rate": 6.270257522306811e-08, + "loss": 1.0229, + "step": 78389 + }, + { + "epoch": 5.79, + "learning_rate": 6.265798578671268e-08, + "loss": 0.9237, + "step": 78390 + }, + { + "epoch": 5.79, + "learning_rate": 6.261341216052597e-08, + "loss": 1.0107, + "step": 78391 + }, + { + "epoch": 5.79, + "learning_rate": 6.256885434458016e-08, + "loss": 0.9849, + "step": 78392 + }, + { + "epoch": 5.79, + "learning_rate": 6.252431233894408e-08, + "loss": 1.0221, + "step": 78393 + }, + { + "epoch": 5.79, + "learning_rate": 6.247978614368988e-08, + "loss": 0.9659, + "step": 78394 + }, + { + "epoch": 5.79, + "learning_rate": 6.243527575888864e-08, + "loss": 1.0715, + "step": 78395 + }, + { + "epoch": 5.79, + "learning_rate": 6.239078118461028e-08, + "loss": 0.9642, + "step": 78396 + }, + { + "epoch": 5.79, + "learning_rate": 6.234630242092588e-08, + "loss": 1.0221, + "step": 78397 + }, + { + "epoch": 5.79, + "learning_rate": 6.230183946790535e-08, + "loss": 0.9645, + "step": 78398 + }, + { + "epoch": 5.79, + "learning_rate": 6.2257392325622e-08, + "loss": 0.9672, + "step": 78399 + }, + { + "epoch": 5.79, + "learning_rate": 6.221296099414464e-08, + "loss": 0.8446, + "step": 78400 + }, + { + "epoch": 5.79, + "learning_rate": 6.216854547354434e-08, + "loss": 0.9512, + "step": 78401 + }, + { + "epoch": 5.79, + "learning_rate": 6.212414576389103e-08, + "loss": 1.0563, + "step": 78402 + }, + { + "epoch": 5.79, + "learning_rate": 6.207976186525577e-08, + "loss": 1.0353, + "step": 78403 + }, + { + "epoch": 5.79, + "learning_rate": 6.203539377771073e-08, + "loss": 1.0606, + "step": 78404 + }, + { + "epoch": 5.79, + "learning_rate": 6.199104150132473e-08, + "loss": 0.9789, + "step": 78405 + }, + { + "epoch": 5.79, + "learning_rate": 6.194670503616884e-08, + "loss": 1.0602, + "step": 78406 + }, + { + "epoch": 5.79, + "learning_rate": 6.190238438231299e-08, + "loss": 1.1044, + "step": 78407 + }, + { + "epoch": 5.79, + "learning_rate": 6.185807953982825e-08, + "loss": 0.9714, + "step": 78408 + }, + { + "epoch": 5.79, + "learning_rate": 6.181379050878566e-08, + "loss": 1.0301, + "step": 78409 + }, + { + "epoch": 5.79, + "learning_rate": 6.176951728925517e-08, + "loss": 0.9745, + "step": 78410 + }, + { + "epoch": 5.79, + "learning_rate": 6.172525988130563e-08, + "loss": 0.948, + "step": 78411 + }, + { + "epoch": 5.79, + "learning_rate": 6.168101828501139e-08, + "loss": 0.9633, + "step": 78412 + }, + { + "epoch": 5.79, + "learning_rate": 6.163679250043908e-08, + "loss": 0.9535, + "step": 78413 + }, + { + "epoch": 5.79, + "learning_rate": 6.159258252766088e-08, + "loss": 0.9278, + "step": 78414 + }, + { + "epoch": 5.79, + "learning_rate": 6.15483883667467e-08, + "loss": 0.9991, + "step": 78415 + }, + { + "epoch": 5.79, + "learning_rate": 6.150421001776652e-08, + "loss": 1.0214, + "step": 78416 + }, + { + "epoch": 5.79, + "learning_rate": 6.146004748079138e-08, + "loss": 1.0281, + "step": 78417 + }, + { + "epoch": 5.79, + "learning_rate": 6.14159007558901e-08, + "loss": 0.9829, + "step": 78418 + }, + { + "epoch": 5.79, + "learning_rate": 6.137176984313486e-08, + "loss": 0.9943, + "step": 78419 + }, + { + "epoch": 5.79, + "learning_rate": 6.132765474259451e-08, + "loss": 1.0047, + "step": 78420 + }, + { + "epoch": 5.79, + "learning_rate": 6.128355545434006e-08, + "loss": 0.9695, + "step": 78421 + }, + { + "epoch": 5.79, + "learning_rate": 6.123947197844149e-08, + "loss": 0.9804, + "step": 78422 + }, + { + "epoch": 5.79, + "learning_rate": 6.119540431496762e-08, + "loss": 1.0685, + "step": 78423 + }, + { + "epoch": 5.79, + "learning_rate": 6.115135246399062e-08, + "loss": 0.951, + "step": 78424 + }, + { + "epoch": 5.79, + "learning_rate": 6.110731642557932e-08, + "loss": 1.0434, + "step": 78425 + }, + { + "epoch": 5.79, + "learning_rate": 6.106329619980477e-08, + "loss": 1.1311, + "step": 78426 + }, + { + "epoch": 5.79, + "learning_rate": 6.101929178673472e-08, + "loss": 0.8862, + "step": 78427 + }, + { + "epoch": 5.79, + "learning_rate": 6.09753031864424e-08, + "loss": 1.007, + "step": 78428 + }, + { + "epoch": 5.8, + "learning_rate": 6.093133039899557e-08, + "loss": 1.0154, + "step": 78429 + }, + { + "epoch": 5.8, + "learning_rate": 6.088737342446416e-08, + "loss": 0.9835, + "step": 78430 + }, + { + "epoch": 5.8, + "learning_rate": 6.084343226291923e-08, + "loss": 1.0175, + "step": 78431 + }, + { + "epoch": 5.8, + "learning_rate": 6.07995069144296e-08, + "loss": 0.8465, + "step": 78432 + }, + { + "epoch": 5.8, + "learning_rate": 6.075559737906744e-08, + "loss": 0.8673, + "step": 78433 + }, + { + "epoch": 5.8, + "learning_rate": 6.071170365689938e-08, + "loss": 0.996, + "step": 78434 + }, + { + "epoch": 5.8, + "learning_rate": 6.066782574799756e-08, + "loss": 0.9565, + "step": 78435 + }, + { + "epoch": 5.8, + "learning_rate": 6.062396365242973e-08, + "loss": 1.0119, + "step": 78436 + }, + { + "epoch": 5.8, + "learning_rate": 6.058011737026914e-08, + "loss": 1.0046, + "step": 78437 + }, + { + "epoch": 5.8, + "learning_rate": 6.053628690158131e-08, + "loss": 1.0789, + "step": 78438 + }, + { + "epoch": 5.8, + "learning_rate": 6.04924722464395e-08, + "loss": 0.8741, + "step": 78439 + }, + { + "epoch": 5.8, + "learning_rate": 6.044867340491034e-08, + "loss": 1.0417, + "step": 78440 + }, + { + "epoch": 5.8, + "learning_rate": 6.04048903770671e-08, + "loss": 1.0008, + "step": 78441 + }, + { + "epoch": 5.8, + "learning_rate": 6.036112316297638e-08, + "loss": 0.9899, + "step": 78442 + }, + { + "epoch": 5.8, + "learning_rate": 6.031737176270925e-08, + "loss": 1.0619, + "step": 78443 + }, + { + "epoch": 5.8, + "learning_rate": 6.027363617633453e-08, + "loss": 0.9724, + "step": 78444 + }, + { + "epoch": 5.8, + "learning_rate": 6.022991640392217e-08, + "loss": 0.9561, + "step": 78445 + }, + { + "epoch": 5.8, + "learning_rate": 6.018621244554323e-08, + "loss": 0.997, + "step": 78446 + }, + { + "epoch": 5.8, + "learning_rate": 6.014252430126433e-08, + "loss": 1.0142, + "step": 78447 + }, + { + "epoch": 5.8, + "learning_rate": 6.00988519711565e-08, + "loss": 0.9478, + "step": 78448 + }, + { + "epoch": 5.8, + "learning_rate": 6.005519545528971e-08, + "loss": 1.1256, + "step": 78449 + }, + { + "epoch": 5.8, + "learning_rate": 6.001155475373278e-08, + "loss": 0.9928, + "step": 78450 + }, + { + "epoch": 5.8, + "learning_rate": 5.996792986655453e-08, + "loss": 1.0355, + "step": 78451 + }, + { + "epoch": 5.8, + "learning_rate": 5.992432079382604e-08, + "loss": 0.9679, + "step": 78452 + }, + { + "epoch": 5.8, + "learning_rate": 5.988072753561502e-08, + "loss": 0.9941, + "step": 78453 + }, + { + "epoch": 5.8, + "learning_rate": 5.983715009199254e-08, + "loss": 1.0808, + "step": 78454 + }, + { + "epoch": 5.8, + "learning_rate": 5.97935884630263e-08, + "loss": 0.9158, + "step": 78455 + }, + { + "epoch": 5.8, + "learning_rate": 5.975004264878515e-08, + "loss": 1.0417, + "step": 78456 + }, + { + "epoch": 5.8, + "learning_rate": 5.970651264934124e-08, + "loss": 0.97, + "step": 78457 + }, + { + "epoch": 5.8, + "learning_rate": 5.96629984647612e-08, + "loss": 0.9134, + "step": 78458 + }, + { + "epoch": 5.8, + "learning_rate": 5.961950009511497e-08, + "loss": 1.023, + "step": 78459 + }, + { + "epoch": 5.8, + "learning_rate": 5.9576017540472486e-08, + "loss": 0.9894, + "step": 78460 + }, + { + "epoch": 5.8, + "learning_rate": 5.953255080090148e-08, + "loss": 0.9573, + "step": 78461 + }, + { + "epoch": 5.8, + "learning_rate": 5.948909987647411e-08, + "loss": 1.0442, + "step": 78462 + }, + { + "epoch": 5.8, + "learning_rate": 5.944566476725477e-08, + "loss": 0.9861, + "step": 78463 + }, + { + "epoch": 5.8, + "learning_rate": 5.9402245473316744e-08, + "loss": 1.0326, + "step": 78464 + }, + { + "epoch": 5.8, + "learning_rate": 5.9358841994726636e-08, + "loss": 0.8985, + "step": 78465 + }, + { + "epoch": 5.8, + "learning_rate": 5.931545433155439e-08, + "loss": 0.9261, + "step": 78466 + }, + { + "epoch": 5.8, + "learning_rate": 5.9272082483868844e-08, + "loss": 1.0255, + "step": 78467 + }, + { + "epoch": 5.8, + "learning_rate": 5.9228726451739935e-08, + "loss": 0.9542, + "step": 78468 + }, + { + "epoch": 5.8, + "learning_rate": 5.918538623523651e-08, + "loss": 1.0162, + "step": 78469 + }, + { + "epoch": 5.8, + "learning_rate": 5.914206183442517e-08, + "loss": 1.0497, + "step": 78470 + }, + { + "epoch": 5.8, + "learning_rate": 5.909875324937808e-08, + "loss": 0.8943, + "step": 78471 + }, + { + "epoch": 5.8, + "learning_rate": 5.9055460480161864e-08, + "loss": 0.982, + "step": 78472 + }, + { + "epoch": 5.8, + "learning_rate": 5.9012183526846454e-08, + "loss": 1.0751, + "step": 78473 + }, + { + "epoch": 5.8, + "learning_rate": 5.896892238949958e-08, + "loss": 0.9642, + "step": 78474 + }, + { + "epoch": 5.8, + "learning_rate": 5.89256770681923e-08, + "loss": 1.0118, + "step": 78475 + }, + { + "epoch": 5.8, + "learning_rate": 5.888244756299011e-08, + "loss": 1.0266, + "step": 78476 + }, + { + "epoch": 5.8, + "learning_rate": 5.8839233873965174e-08, + "loss": 0.9969, + "step": 78477 + }, + { + "epoch": 5.8, + "learning_rate": 5.879603600118522e-08, + "loss": 0.9464, + "step": 78478 + }, + { + "epoch": 5.8, + "learning_rate": 5.875285394471797e-08, + "loss": 1.1589, + "step": 78479 + }, + { + "epoch": 5.8, + "learning_rate": 5.8709687704632253e-08, + "loss": 1.0129, + "step": 78480 + }, + { + "epoch": 5.8, + "learning_rate": 5.8666537280996913e-08, + "loss": 1.019, + "step": 78481 + }, + { + "epoch": 5.8, + "learning_rate": 5.8623402673883e-08, + "loss": 1.044, + "step": 78482 + }, + { + "epoch": 5.8, + "learning_rate": 5.8580283883354905e-08, + "loss": 1.0136, + "step": 78483 + }, + { + "epoch": 5.8, + "learning_rate": 5.853718090948368e-08, + "loss": 1.0487, + "step": 78484 + }, + { + "epoch": 5.8, + "learning_rate": 5.8494093752337053e-08, + "loss": 1.0761, + "step": 78485 + }, + { + "epoch": 5.8, + "learning_rate": 5.845102241198608e-08, + "loss": 0.9722, + "step": 78486 + }, + { + "epoch": 5.8, + "learning_rate": 5.840796688849515e-08, + "loss": 1.0023, + "step": 78487 + }, + { + "epoch": 5.8, + "learning_rate": 5.8364927181936424e-08, + "loss": 0.9677, + "step": 78488 + }, + { + "epoch": 5.8, + "learning_rate": 5.8321903292376526e-08, + "loss": 0.913, + "step": 78489 + }, + { + "epoch": 5.8, + "learning_rate": 5.827889521988317e-08, + "loss": 1.0453, + "step": 78490 + }, + { + "epoch": 5.8, + "learning_rate": 5.823590296452741e-08, + "loss": 1.0125, + "step": 78491 + }, + { + "epoch": 5.8, + "learning_rate": 5.819292652637476e-08, + "loss": 1.0199, + "step": 78492 + }, + { + "epoch": 5.8, + "learning_rate": 5.8149965905495155e-08, + "loss": 1.1139, + "step": 78493 + }, + { + "epoch": 5.8, + "learning_rate": 5.810702110195743e-08, + "loss": 1.0356, + "step": 78494 + }, + { + "epoch": 5.8, + "learning_rate": 5.8064092115829306e-08, + "loss": 0.9959, + "step": 78495 + }, + { + "epoch": 5.8, + "learning_rate": 5.8021178947177405e-08, + "loss": 0.9634, + "step": 78496 + }, + { + "epoch": 5.8, + "learning_rate": 5.797828159607277e-08, + "loss": 1.0606, + "step": 78497 + }, + { + "epoch": 5.8, + "learning_rate": 5.793540006258313e-08, + "loss": 1.031, + "step": 78498 + }, + { + "epoch": 5.8, + "learning_rate": 5.789253434677511e-08, + "loss": 0.9377, + "step": 78499 + }, + { + "epoch": 5.8, + "learning_rate": 5.784968444871753e-08, + "loss": 1.0195, + "step": 78500 + }, + { + "epoch": 5.8, + "learning_rate": 5.780685036847922e-08, + "loss": 0.9457, + "step": 78501 + }, + { + "epoch": 5.8, + "learning_rate": 5.776403210612791e-08, + "loss": 0.9487, + "step": 78502 + }, + { + "epoch": 5.8, + "learning_rate": 5.772122966173133e-08, + "loss": 1.0806, + "step": 78503 + }, + { + "epoch": 5.8, + "learning_rate": 5.767844303535941e-08, + "loss": 1.0499, + "step": 78504 + }, + { + "epoch": 5.8, + "learning_rate": 5.763567222707656e-08, + "loss": 0.9513, + "step": 78505 + }, + { + "epoch": 5.8, + "learning_rate": 5.759291723695493e-08, + "loss": 1.1062, + "step": 78506 + }, + { + "epoch": 5.8, + "learning_rate": 5.7550178065061136e-08, + "loss": 0.8969, + "step": 78507 + }, + { + "epoch": 5.8, + "learning_rate": 5.7507454711461796e-08, + "loss": 1.0252, + "step": 78508 + }, + { + "epoch": 5.8, + "learning_rate": 5.746474717622574e-08, + "loss": 1.0365, + "step": 78509 + }, + { + "epoch": 5.8, + "learning_rate": 5.74220554594207e-08, + "loss": 0.9872, + "step": 78510 + }, + { + "epoch": 5.8, + "learning_rate": 5.73793795611155e-08, + "loss": 1.0012, + "step": 78511 + }, + { + "epoch": 5.8, + "learning_rate": 5.733671948137787e-08, + "loss": 1.0151, + "step": 78512 + }, + { + "epoch": 5.8, + "learning_rate": 5.7294075220274416e-08, + "loss": 0.9883, + "step": 78513 + }, + { + "epoch": 5.8, + "learning_rate": 5.7251446777873975e-08, + "loss": 0.9828, + "step": 78514 + }, + { + "epoch": 5.8, + "learning_rate": 5.7208834154244276e-08, + "loss": 1.0075, + "step": 78515 + }, + { + "epoch": 5.8, + "learning_rate": 5.7166237349454146e-08, + "loss": 0.9419, + "step": 78516 + }, + { + "epoch": 5.8, + "learning_rate": 5.712365636356909e-08, + "loss": 0.9328, + "step": 78517 + }, + { + "epoch": 5.8, + "learning_rate": 5.708109119665795e-08, + "loss": 0.9729, + "step": 78518 + }, + { + "epoch": 5.8, + "learning_rate": 5.703854184878843e-08, + "loss": 1.129, + "step": 78519 + }, + { + "epoch": 5.8, + "learning_rate": 5.699600832002938e-08, + "loss": 1.0091, + "step": 78520 + }, + { + "epoch": 5.8, + "learning_rate": 5.6953490610446307e-08, + "loss": 1.0553, + "step": 78521 + }, + { + "epoch": 5.8, + "learning_rate": 5.691098872010914e-08, + "loss": 0.9404, + "step": 78522 + }, + { + "epoch": 5.8, + "learning_rate": 5.686850264908339e-08, + "loss": 0.9759, + "step": 78523 + }, + { + "epoch": 5.8, + "learning_rate": 5.682603239743789e-08, + "loss": 1.0747, + "step": 78524 + }, + { + "epoch": 5.8, + "learning_rate": 5.678357796524037e-08, + "loss": 1.0955, + "step": 78525 + }, + { + "epoch": 5.8, + "learning_rate": 5.674113935255632e-08, + "loss": 1.0312, + "step": 78526 + }, + { + "epoch": 5.8, + "learning_rate": 5.6698716559456804e-08, + "loss": 0.9471, + "step": 78527 + }, + { + "epoch": 5.8, + "learning_rate": 5.665630958600621e-08, + "loss": 0.9537, + "step": 78528 + }, + { + "epoch": 5.8, + "learning_rate": 5.661391843227337e-08, + "loss": 0.9266, + "step": 78529 + }, + { + "epoch": 5.8, + "learning_rate": 5.657154309832491e-08, + "loss": 0.9553, + "step": 78530 + }, + { + "epoch": 5.8, + "learning_rate": 5.652918358423076e-08, + "loss": 0.983, + "step": 78531 + }, + { + "epoch": 5.8, + "learning_rate": 5.6486839890054214e-08, + "loss": 1.1455, + "step": 78532 + }, + { + "epoch": 5.8, + "learning_rate": 5.6444512015866314e-08, + "loss": 0.9625, + "step": 78533 + }, + { + "epoch": 5.8, + "learning_rate": 5.6402199961732575e-08, + "loss": 0.9843, + "step": 78534 + }, + { + "epoch": 5.8, + "learning_rate": 5.635990372772071e-08, + "loss": 0.9669, + "step": 78535 + }, + { + "epoch": 5.8, + "learning_rate": 5.631762331389734e-08, + "loss": 1.042, + "step": 78536 + }, + { + "epoch": 5.8, + "learning_rate": 5.62753587203313e-08, + "loss": 0.9455, + "step": 78537 + }, + { + "epoch": 5.8, + "learning_rate": 5.623310994708808e-08, + "loss": 0.9416, + "step": 78538 + }, + { + "epoch": 5.8, + "learning_rate": 5.619087699423653e-08, + "loss": 1.0502, + "step": 78539 + }, + { + "epoch": 5.8, + "learning_rate": 5.614865986184215e-08, + "loss": 1.0447, + "step": 78540 + }, + { + "epoch": 5.8, + "learning_rate": 5.6106458549972656e-08, + "loss": 1.0406, + "step": 78541 + }, + { + "epoch": 5.8, + "learning_rate": 5.6064273058696884e-08, + "loss": 0.9433, + "step": 78542 + }, + { + "epoch": 5.8, + "learning_rate": 5.602210338808034e-08, + "loss": 0.9307, + "step": 78543 + }, + { + "epoch": 5.8, + "learning_rate": 5.5979949538189636e-08, + "loss": 1.0952, + "step": 78544 + }, + { + "epoch": 5.8, + "learning_rate": 5.59378115090925e-08, + "loss": 0.9919, + "step": 78545 + }, + { + "epoch": 5.8, + "learning_rate": 5.589568930085665e-08, + "loss": 1.0106, + "step": 78546 + }, + { + "epoch": 5.8, + "learning_rate": 5.585358291354759e-08, + "loss": 0.9474, + "step": 78547 + }, + { + "epoch": 5.8, + "learning_rate": 5.5811492347234156e-08, + "loss": 1.0449, + "step": 78548 + }, + { + "epoch": 5.8, + "learning_rate": 5.576941760198185e-08, + "loss": 0.9619, + "step": 78549 + }, + { + "epoch": 5.8, + "learning_rate": 5.572735867785839e-08, + "loss": 0.9276, + "step": 78550 + }, + { + "epoch": 5.8, + "learning_rate": 5.56853155749304e-08, + "loss": 0.9927, + "step": 78551 + }, + { + "epoch": 5.8, + "learning_rate": 5.564328829326449e-08, + "loss": 0.9946, + "step": 78552 + }, + { + "epoch": 5.8, + "learning_rate": 5.560127683292838e-08, + "loss": 1.0425, + "step": 78553 + }, + { + "epoch": 5.8, + "learning_rate": 5.555928119398868e-08, + "loss": 1.037, + "step": 78554 + }, + { + "epoch": 5.8, + "learning_rate": 5.55173013765109e-08, + "loss": 0.996, + "step": 78555 + }, + { + "epoch": 5.8, + "learning_rate": 5.547533738056277e-08, + "loss": 1.0456, + "step": 78556 + }, + { + "epoch": 5.8, + "learning_rate": 5.5433389206211996e-08, + "loss": 1.1154, + "step": 78557 + }, + { + "epoch": 5.8, + "learning_rate": 5.5391456853524094e-08, + "loss": 0.9296, + "step": 78558 + }, + { + "epoch": 5.8, + "learning_rate": 5.534954032256568e-08, + "loss": 1.0195, + "step": 78559 + }, + { + "epoch": 5.8, + "learning_rate": 5.530763961340446e-08, + "loss": 1.092, + "step": 78560 + }, + { + "epoch": 5.8, + "learning_rate": 5.526575472610485e-08, + "loss": 1.0905, + "step": 78561 + }, + { + "epoch": 5.8, + "learning_rate": 5.522388566073678e-08, + "loss": 0.9235, + "step": 78562 + }, + { + "epoch": 5.8, + "learning_rate": 5.518203241736464e-08, + "loss": 0.9258, + "step": 78563 + }, + { + "epoch": 5.81, + "learning_rate": 5.5140194996056164e-08, + "loss": 0.9979, + "step": 78564 + }, + { + "epoch": 5.81, + "learning_rate": 5.509837339687685e-08, + "loss": 1.0249, + "step": 78565 + }, + { + "epoch": 5.81, + "learning_rate": 5.505656761989442e-08, + "loss": 0.9601, + "step": 78566 + }, + { + "epoch": 5.81, + "learning_rate": 5.5014777665174376e-08, + "loss": 0.9933, + "step": 78567 + }, + { + "epoch": 5.81, + "learning_rate": 5.497300353278334e-08, + "loss": 0.9989, + "step": 78568 + }, + { + "epoch": 5.81, + "learning_rate": 5.493124522278792e-08, + "loss": 0.9854, + "step": 78569 + }, + { + "epoch": 5.81, + "learning_rate": 5.488950273525362e-08, + "loss": 1.0953, + "step": 78570 + }, + { + "epoch": 5.81, + "learning_rate": 5.484777607024927e-08, + "loss": 0.9756, + "step": 78571 + }, + { + "epoch": 5.81, + "learning_rate": 5.480606522784038e-08, + "loss": 1.0723, + "step": 78572 + }, + { + "epoch": 5.81, + "learning_rate": 5.476437020809133e-08, + "loss": 1.1027, + "step": 78573 + }, + { + "epoch": 5.81, + "learning_rate": 5.4722691011069864e-08, + "loss": 1.0663, + "step": 78574 + }, + { + "epoch": 5.81, + "learning_rate": 5.468102763684258e-08, + "loss": 1.067, + "step": 78575 + }, + { + "epoch": 5.81, + "learning_rate": 5.463938008547609e-08, + "loss": 1.0011, + "step": 78576 + }, + { + "epoch": 5.81, + "learning_rate": 5.45977483570348e-08, + "loss": 0.9743, + "step": 78577 + }, + { + "epoch": 5.81, + "learning_rate": 5.455613245158753e-08, + "loss": 1.0558, + "step": 78578 + }, + { + "epoch": 5.81, + "learning_rate": 5.451453236919868e-08, + "loss": 1.0234, + "step": 78579 + }, + { + "epoch": 5.81, + "learning_rate": 5.447294810993486e-08, + "loss": 1.1185, + "step": 78580 + }, + { + "epoch": 5.81, + "learning_rate": 5.44313796738627e-08, + "loss": 0.9781, + "step": 78581 + }, + { + "epoch": 5.81, + "learning_rate": 5.4389827061046565e-08, + "loss": 1.1248, + "step": 78582 + }, + { + "epoch": 5.81, + "learning_rate": 5.4348290271555306e-08, + "loss": 1.0145, + "step": 78583 + }, + { + "epoch": 5.81, + "learning_rate": 5.430676930545331e-08, + "loss": 1.04, + "step": 78584 + }, + { + "epoch": 5.81, + "learning_rate": 5.42652641628072e-08, + "loss": 1.1082, + "step": 78585 + }, + { + "epoch": 5.81, + "learning_rate": 5.4223774843681354e-08, + "loss": 1.0385, + "step": 78586 + }, + { + "epoch": 5.81, + "learning_rate": 5.418230134814462e-08, + "loss": 0.9606, + "step": 78587 + }, + { + "epoch": 5.81, + "learning_rate": 5.414084367626138e-08, + "loss": 0.9572, + "step": 78588 + }, + { + "epoch": 5.81, + "learning_rate": 5.409940182809714e-08, + "loss": 1.104, + "step": 78589 + }, + { + "epoch": 5.81, + "learning_rate": 5.405797580371852e-08, + "loss": 1.0027, + "step": 78590 + }, + { + "epoch": 5.81, + "learning_rate": 5.401656560319213e-08, + "loss": 0.9512, + "step": 78591 + }, + { + "epoch": 5.81, + "learning_rate": 5.3975171226582356e-08, + "loss": 1.0327, + "step": 78592 + }, + { + "epoch": 5.81, + "learning_rate": 5.393379267395582e-08, + "loss": 1.0884, + "step": 78593 + }, + { + "epoch": 5.81, + "learning_rate": 5.3892429945379135e-08, + "loss": 0.9457, + "step": 78594 + }, + { + "epoch": 5.81, + "learning_rate": 5.3851083040915576e-08, + "loss": 0.944, + "step": 78595 + }, + { + "epoch": 5.81, + "learning_rate": 5.380975196063398e-08, + "loss": 1.0045, + "step": 78596 + }, + { + "epoch": 5.81, + "learning_rate": 5.376843670459875e-08, + "loss": 0.9338, + "step": 78597 + }, + { + "epoch": 5.81, + "learning_rate": 5.372713727287537e-08, + "loss": 0.9802, + "step": 78598 + }, + { + "epoch": 5.81, + "learning_rate": 5.368585366552936e-08, + "loss": 1.0052, + "step": 78599 + }, + { + "epoch": 5.81, + "learning_rate": 5.364458588262844e-08, + "loss": 1.0061, + "step": 78600 + }, + { + "epoch": 5.81, + "learning_rate": 5.360333392423589e-08, + "loss": 1.0665, + "step": 78601 + }, + { + "epoch": 5.81, + "learning_rate": 5.3562097790417213e-08, + "loss": 0.9268, + "step": 78602 + }, + { + "epoch": 5.81, + "learning_rate": 5.352087748123902e-08, + "loss": 1.0662, + "step": 78603 + }, + { + "epoch": 5.81, + "learning_rate": 5.3479672996767926e-08, + "loss": 1.0534, + "step": 78604 + }, + { + "epoch": 5.81, + "learning_rate": 5.343848433706722e-08, + "loss": 1.0088, + "step": 78605 + }, + { + "epoch": 5.81, + "learning_rate": 5.3397311502203507e-08, + "loss": 0.933, + "step": 78606 + }, + { + "epoch": 5.81, + "learning_rate": 5.3356154492243404e-08, + "loss": 0.9839, + "step": 78607 + }, + { + "epoch": 5.81, + "learning_rate": 5.331501330725131e-08, + "loss": 1.0302, + "step": 78608 + }, + { + "epoch": 5.81, + "learning_rate": 5.32738879472916e-08, + "loss": 0.9856, + "step": 78609 + }, + { + "epoch": 5.81, + "learning_rate": 5.3232778412432015e-08, + "loss": 0.9314, + "step": 78610 + }, + { + "epoch": 5.81, + "learning_rate": 5.319168470273584e-08, + "loss": 1.093, + "step": 78611 + }, + { + "epoch": 5.81, + "learning_rate": 5.3150606818269665e-08, + "loss": 1.0196, + "step": 78612 + }, + { + "epoch": 5.81, + "learning_rate": 5.310954475909902e-08, + "loss": 0.971, + "step": 78613 + }, + { + "epoch": 5.81, + "learning_rate": 5.306849852528828e-08, + "loss": 1.0266, + "step": 78614 + }, + { + "epoch": 5.81, + "learning_rate": 5.302746811690296e-08, + "loss": 0.9372, + "step": 78615 + }, + { + "epoch": 5.81, + "learning_rate": 5.298645353400966e-08, + "loss": 0.9703, + "step": 78616 + }, + { + "epoch": 5.81, + "learning_rate": 5.294545477667168e-08, + "loss": 0.9241, + "step": 78617 + }, + { + "epoch": 5.81, + "learning_rate": 5.2904471844955615e-08, + "loss": 0.9542, + "step": 78618 + }, + { + "epoch": 5.81, + "learning_rate": 5.286350473892588e-08, + "loss": 0.9701, + "step": 78619 + }, + { + "epoch": 5.81, + "learning_rate": 5.282255345864795e-08, + "loss": 1.0185, + "step": 78620 + }, + { + "epoch": 5.81, + "learning_rate": 5.2781618004187353e-08, + "loss": 1.0448, + "step": 78621 + }, + { + "epoch": 5.81, + "learning_rate": 5.274069837560847e-08, + "loss": 0.9688, + "step": 78622 + }, + { + "epoch": 5.81, + "learning_rate": 5.2699794572977915e-08, + "loss": 1.0934, + "step": 78623 + }, + { + "epoch": 5.81, + "learning_rate": 5.265890659635786e-08, + "loss": 1.0031, + "step": 78624 + }, + { + "epoch": 5.81, + "learning_rate": 5.261803444581714e-08, + "loss": 1.0455, + "step": 78625 + }, + { + "epoch": 5.81, + "learning_rate": 5.2577178121416825e-08, + "loss": 1.0367, + "step": 78626 + }, + { + "epoch": 5.81, + "learning_rate": 5.253633762322574e-08, + "loss": 0.9362, + "step": 78627 + }, + { + "epoch": 5.81, + "learning_rate": 5.249551295130606e-08, + "loss": 1.0352, + "step": 78628 + }, + { + "epoch": 5.81, + "learning_rate": 5.2454704105723285e-08, + "loss": 0.8738, + "step": 78629 + }, + { + "epoch": 5.81, + "learning_rate": 5.241391108654403e-08, + "loss": 0.9047, + "step": 78630 + }, + { + "epoch": 5.81, + "learning_rate": 5.237313389383047e-08, + "loss": 1.0996, + "step": 78631 + }, + { + "epoch": 5.81, + "learning_rate": 5.2332372527650335e-08, + "loss": 0.9552, + "step": 78632 + }, + { + "epoch": 5.81, + "learning_rate": 5.229162698806578e-08, + "loss": 1.0025, + "step": 78633 + }, + { + "epoch": 5.81, + "learning_rate": 5.225089727514343e-08, + "loss": 1.0583, + "step": 78634 + }, + { + "epoch": 5.81, + "learning_rate": 5.221018338894768e-08, + "loss": 0.8874, + "step": 78635 + }, + { + "epoch": 5.81, + "learning_rate": 5.2169485329542914e-08, + "loss": 0.9902, + "step": 78636 + }, + { + "epoch": 5.81, + "learning_rate": 5.212880309699464e-08, + "loss": 0.9716, + "step": 78637 + }, + { + "epoch": 5.81, + "learning_rate": 5.208813669136725e-08, + "loss": 0.9636, + "step": 78638 + }, + { + "epoch": 5.81, + "learning_rate": 5.204748611272514e-08, + "loss": 0.9378, + "step": 78639 + }, + { + "epoch": 5.81, + "learning_rate": 5.20068513611327e-08, + "loss": 1.0207, + "step": 78640 + }, + { + "epoch": 5.81, + "learning_rate": 5.196623243665433e-08, + "loss": 1.0133, + "step": 78641 + }, + { + "epoch": 5.81, + "learning_rate": 5.192562933935663e-08, + "loss": 1.0693, + "step": 78642 + }, + { + "epoch": 5.81, + "learning_rate": 5.188504206930289e-08, + "loss": 1.0595, + "step": 78643 + }, + { + "epoch": 5.81, + "learning_rate": 5.1844470626557506e-08, + "loss": 1.0256, + "step": 78644 + }, + { + "epoch": 5.81, + "learning_rate": 5.1803915011184866e-08, + "loss": 0.9911, + "step": 78645 + }, + { + "epoch": 5.81, + "learning_rate": 5.176337522325048e-08, + "loss": 0.9955, + "step": 78646 + }, + { + "epoch": 5.81, + "learning_rate": 5.172285126281762e-08, + "loss": 1.02, + "step": 78647 + }, + { + "epoch": 5.81, + "learning_rate": 5.1682343129951794e-08, + "loss": 1.084, + "step": 78648 + }, + { + "epoch": 5.81, + "learning_rate": 5.164185082471629e-08, + "loss": 1.0638, + "step": 78649 + }, + { + "epoch": 5.81, + "learning_rate": 5.1601374347177715e-08, + "loss": 1.0175, + "step": 78650 + }, + { + "epoch": 5.81, + "learning_rate": 5.156091369739713e-08, + "loss": 0.9292, + "step": 78651 + }, + { + "epoch": 5.81, + "learning_rate": 5.1520468875442265e-08, + "loss": 0.9837, + "step": 78652 + }, + { + "epoch": 5.81, + "learning_rate": 5.148003988137529e-08, + "loss": 1.014, + "step": 78653 + }, + { + "epoch": 5.81, + "learning_rate": 5.14396267152617e-08, + "loss": 0.9593, + "step": 78654 + }, + { + "epoch": 5.81, + "learning_rate": 5.139922937716479e-08, + "loss": 1.0416, + "step": 78655 + }, + { + "epoch": 5.81, + "learning_rate": 5.1358847867150064e-08, + "loss": 0.831, + "step": 78656 + }, + { + "epoch": 5.81, + "learning_rate": 5.131848218527968e-08, + "loss": 0.9331, + "step": 78657 + }, + { + "epoch": 5.81, + "learning_rate": 5.127813233162027e-08, + "loss": 1.0085, + "step": 78658 + }, + { + "epoch": 5.81, + "learning_rate": 5.12377983062351e-08, + "loss": 1.0398, + "step": 78659 + }, + { + "epoch": 5.81, + "learning_rate": 5.119748010918857e-08, + "loss": 0.9884, + "step": 78660 + }, + { + "epoch": 5.81, + "learning_rate": 5.1157177740543965e-08, + "loss": 0.9921, + "step": 78661 + }, + { + "epoch": 5.81, + "learning_rate": 5.111689120036567e-08, + "loss": 0.9451, + "step": 78662 + }, + { + "epoch": 5.81, + "learning_rate": 5.1076620488719197e-08, + "loss": 1.0382, + "step": 78663 + }, + { + "epoch": 5.81, + "learning_rate": 5.1036365605665603e-08, + "loss": 0.9788, + "step": 78664 + }, + { + "epoch": 5.81, + "learning_rate": 5.0996126551272616e-08, + "loss": 0.9587, + "step": 78665 + }, + { + "epoch": 5.81, + "learning_rate": 5.09559033256013e-08, + "loss": 1.0062, + "step": 78666 + }, + { + "epoch": 5.81, + "learning_rate": 5.091569592871715e-08, + "loss": 0.9836, + "step": 78667 + }, + { + "epoch": 5.81, + "learning_rate": 5.0875504360684555e-08, + "loss": 0.98, + "step": 78668 + }, + { + "epoch": 5.81, + "learning_rate": 5.0835328621565704e-08, + "loss": 0.974, + "step": 78669 + }, + { + "epoch": 5.81, + "learning_rate": 5.079516871142609e-08, + "loss": 0.9806, + "step": 78670 + }, + { + "epoch": 5.81, + "learning_rate": 5.075502463032789e-08, + "loss": 0.927, + "step": 78671 + }, + { + "epoch": 5.81, + "learning_rate": 5.07148963783366e-08, + "loss": 1.0487, + "step": 78672 + }, + { + "epoch": 5.81, + "learning_rate": 5.06747839555155e-08, + "loss": 0.8822, + "step": 78673 + }, + { + "epoch": 5.81, + "learning_rate": 5.063468736193011e-08, + "loss": 1.0406, + "step": 78674 + }, + { + "epoch": 5.81, + "learning_rate": 5.0594606597641484e-08, + "loss": 1.1171, + "step": 78675 + }, + { + "epoch": 5.81, + "learning_rate": 5.0554541662714006e-08, + "loss": 1.0029, + "step": 78676 + }, + { + "epoch": 5.81, + "learning_rate": 5.051449255721208e-08, + "loss": 1.0221, + "step": 78677 + }, + { + "epoch": 5.81, + "learning_rate": 5.047445928119899e-08, + "loss": 0.9573, + "step": 78678 + }, + { + "epoch": 5.81, + "learning_rate": 5.043444183473911e-08, + "loss": 1.1002, + "step": 78679 + }, + { + "epoch": 5.81, + "learning_rate": 5.039444021789575e-08, + "loss": 0.9994, + "step": 78680 + }, + { + "epoch": 5.81, + "learning_rate": 5.035445443073217e-08, + "loss": 1.1427, + "step": 78681 + }, + { + "epoch": 5.81, + "learning_rate": 5.031448447331277e-08, + "loss": 0.9508, + "step": 78682 + }, + { + "epoch": 5.81, + "learning_rate": 5.0274530345699736e-08, + "loss": 1.07, + "step": 78683 + }, + { + "epoch": 5.81, + "learning_rate": 5.023459204795855e-08, + "loss": 1.0457, + "step": 78684 + }, + { + "epoch": 5.81, + "learning_rate": 5.01946695801514e-08, + "loss": 1.0913, + "step": 78685 + }, + { + "epoch": 5.81, + "learning_rate": 5.0154762942342674e-08, + "loss": 0.9704, + "step": 78686 + }, + { + "epoch": 5.81, + "learning_rate": 5.0114872134594536e-08, + "loss": 0.945, + "step": 78687 + }, + { + "epoch": 5.81, + "learning_rate": 5.00749971569714e-08, + "loss": 0.961, + "step": 78688 + }, + { + "epoch": 5.81, + "learning_rate": 5.003513800953652e-08, + "loss": 0.9536, + "step": 78689 + }, + { + "epoch": 5.81, + "learning_rate": 4.9995294692354315e-08, + "loss": 1.0694, + "step": 78690 + }, + { + "epoch": 5.81, + "learning_rate": 4.995546720548583e-08, + "loss": 0.9375, + "step": 78691 + }, + { + "epoch": 5.81, + "learning_rate": 4.991565554899658e-08, + "loss": 0.9844, + "step": 78692 + }, + { + "epoch": 5.81, + "learning_rate": 4.9875859722949834e-08, + "loss": 0.9597, + "step": 78693 + }, + { + "epoch": 5.81, + "learning_rate": 4.983607972740778e-08, + "loss": 0.9441, + "step": 78694 + }, + { + "epoch": 5.81, + "learning_rate": 4.97963155624337e-08, + "loss": 1.0827, + "step": 78695 + }, + { + "epoch": 5.81, + "learning_rate": 4.9756567228090856e-08, + "loss": 1.0294, + "step": 78696 + }, + { + "epoch": 5.81, + "learning_rate": 4.971683472444366e-08, + "loss": 0.8259, + "step": 78697 + }, + { + "epoch": 5.81, + "learning_rate": 4.9677118051555395e-08, + "loss": 0.9545, + "step": 78698 + }, + { + "epoch": 5.82, + "learning_rate": 4.963741720948712e-08, + "loss": 1.0371, + "step": 78699 + }, + { + "epoch": 5.82, + "learning_rate": 4.959773219830322e-08, + "loss": 0.9507, + "step": 78700 + }, + { + "epoch": 5.82, + "learning_rate": 4.955806301806809e-08, + "loss": 0.9983, + "step": 78701 + }, + { + "epoch": 5.82, + "learning_rate": 4.951840966884391e-08, + "loss": 0.9353, + "step": 78702 + }, + { + "epoch": 5.82, + "learning_rate": 4.9478772150692855e-08, + "loss": 0.9566, + "step": 78703 + }, + { + "epoch": 5.82, + "learning_rate": 4.9439150463678195e-08, + "loss": 0.9467, + "step": 78704 + }, + { + "epoch": 5.82, + "learning_rate": 4.939954460786433e-08, + "loss": 0.8889, + "step": 78705 + }, + { + "epoch": 5.82, + "learning_rate": 4.935995458331344e-08, + "loss": 0.9946, + "step": 78706 + }, + { + "epoch": 5.82, + "learning_rate": 4.9320380390087686e-08, + "loss": 1.0335, + "step": 78707 + }, + { + "epoch": 5.82, + "learning_rate": 4.928082202825146e-08, + "loss": 0.9731, + "step": 78708 + }, + { + "epoch": 5.82, + "learning_rate": 4.9241279497866945e-08, + "loss": 1.0698, + "step": 78709 + }, + { + "epoch": 5.82, + "learning_rate": 4.920175279899742e-08, + "loss": 0.9256, + "step": 78710 + }, + { + "epoch": 5.82, + "learning_rate": 4.916224193170505e-08, + "loss": 0.9895, + "step": 78711 + }, + { + "epoch": 5.82, + "learning_rate": 4.912274689605312e-08, + "loss": 1.0457, + "step": 78712 + }, + { + "epoch": 5.82, + "learning_rate": 4.9083267692104916e-08, + "loss": 1.0351, + "step": 78713 + }, + { + "epoch": 5.82, + "learning_rate": 4.904380431992262e-08, + "loss": 0.884, + "step": 78714 + }, + { + "epoch": 5.82, + "learning_rate": 4.9004356779570605e-08, + "loss": 0.9899, + "step": 78715 + }, + { + "epoch": 5.82, + "learning_rate": 4.896492507110884e-08, + "loss": 0.9765, + "step": 78716 + }, + { + "epoch": 5.82, + "learning_rate": 4.892550919460171e-08, + "loss": 1.0062, + "step": 78717 + }, + { + "epoch": 5.82, + "learning_rate": 4.888610915011138e-08, + "loss": 1.0551, + "step": 78718 + }, + { + "epoch": 5.82, + "learning_rate": 4.884672493770115e-08, + "loss": 0.9364, + "step": 78719 + }, + { + "epoch": 5.82, + "learning_rate": 4.8807356557433186e-08, + "loss": 0.9825, + "step": 78720 + }, + { + "epoch": 5.82, + "learning_rate": 4.8768004009370764e-08, + "loss": 0.9637, + "step": 78721 + }, + { + "epoch": 5.82, + "learning_rate": 4.8728667293576056e-08, + "loss": 0.9548, + "step": 78722 + }, + { + "epoch": 5.82, + "learning_rate": 4.868934641011236e-08, + "loss": 1.0124, + "step": 78723 + }, + { + "epoch": 5.82, + "learning_rate": 4.865004135904072e-08, + "loss": 0.9539, + "step": 78724 + }, + { + "epoch": 5.82, + "learning_rate": 4.8610752140424435e-08, + "loss": 1.0832, + "step": 78725 + }, + { + "epoch": 5.82, + "learning_rate": 4.857147875432566e-08, + "loss": 1.0963, + "step": 78726 + }, + { + "epoch": 5.82, + "learning_rate": 4.85322212008077e-08, + "loss": 0.9316, + "step": 78727 + }, + { + "epoch": 5.82, + "learning_rate": 4.84929794799327e-08, + "loss": 0.939, + "step": 78728 + }, + { + "epoch": 5.82, + "learning_rate": 4.8453753591762855e-08, + "loss": 1.0566, + "step": 78729 + }, + { + "epoch": 5.82, + "learning_rate": 4.8414543536360324e-08, + "loss": 0.9169, + "step": 78730 + }, + { + "epoch": 5.82, + "learning_rate": 4.837534931378951e-08, + "loss": 1.0461, + "step": 78731 + }, + { + "epoch": 5.82, + "learning_rate": 4.833617092410925e-08, + "loss": 0.881, + "step": 78732 + }, + { + "epoch": 5.82, + "learning_rate": 4.8297008367385044e-08, + "loss": 0.9689, + "step": 78733 + }, + { + "epoch": 5.82, + "learning_rate": 4.8257861643676854e-08, + "loss": 0.9908, + "step": 78734 + }, + { + "epoch": 5.82, + "learning_rate": 4.821873075304906e-08, + "loss": 0.9513, + "step": 78735 + }, + { + "epoch": 5.82, + "learning_rate": 4.8179615695561624e-08, + "loss": 1.013, + "step": 78736 + }, + { + "epoch": 5.82, + "learning_rate": 4.814051647127893e-08, + "loss": 1.0584, + "step": 78737 + }, + { + "epoch": 5.82, + "learning_rate": 4.810143308026205e-08, + "loss": 1.0591, + "step": 78738 + }, + { + "epoch": 5.82, + "learning_rate": 4.806236552257426e-08, + "loss": 1.0429, + "step": 78739 + }, + { + "epoch": 5.82, + "learning_rate": 4.802331379827552e-08, + "loss": 1.1132, + "step": 78740 + }, + { + "epoch": 5.82, + "learning_rate": 4.7984277907429097e-08, + "loss": 0.9709, + "step": 78741 + }, + { + "epoch": 5.82, + "learning_rate": 4.794525785009829e-08, + "loss": 0.8796, + "step": 78742 + }, + { + "epoch": 5.82, + "learning_rate": 4.7906253626344154e-08, + "loss": 0.9627, + "step": 78743 + }, + { + "epoch": 5.82, + "learning_rate": 4.786726523622886e-08, + "loss": 0.9605, + "step": 78744 + }, + { + "epoch": 5.82, + "learning_rate": 4.782829267981348e-08, + "loss": 0.9546, + "step": 78745 + }, + { + "epoch": 5.82, + "learning_rate": 4.778933595716129e-08, + "loss": 1.0745, + "step": 78746 + }, + { + "epoch": 5.82, + "learning_rate": 4.7750395068334453e-08, + "loss": 0.9662, + "step": 78747 + }, + { + "epoch": 5.82, + "learning_rate": 4.7711470013394046e-08, + "loss": 1.0249, + "step": 78748 + }, + { + "epoch": 5.82, + "learning_rate": 4.767256079240223e-08, + "loss": 0.9381, + "step": 78749 + }, + { + "epoch": 5.82, + "learning_rate": 4.7633667405421194e-08, + "loss": 1.0623, + "step": 78750 + }, + { + "epoch": 5.82, + "learning_rate": 4.759478985251198e-08, + "loss": 0.8181, + "step": 78751 + }, + { + "epoch": 5.82, + "learning_rate": 4.755592813373788e-08, + "loss": 1.0653, + "step": 78752 + }, + { + "epoch": 5.82, + "learning_rate": 4.751708224915885e-08, + "loss": 1.0727, + "step": 78753 + }, + { + "epoch": 5.82, + "learning_rate": 4.747825219883817e-08, + "loss": 0.9051, + "step": 78754 + }, + { + "epoch": 5.82, + "learning_rate": 4.74394379828369e-08, + "loss": 0.9179, + "step": 78755 + }, + { + "epoch": 5.82, + "learning_rate": 4.7400639601218324e-08, + "loss": 0.9412, + "step": 78756 + }, + { + "epoch": 5.82, + "learning_rate": 4.736185705404239e-08, + "loss": 0.9948, + "step": 78757 + }, + { + "epoch": 5.82, + "learning_rate": 4.732309034137017e-08, + "loss": 0.996, + "step": 78758 + }, + { + "epoch": 5.82, + "learning_rate": 4.728433946326605e-08, + "loss": 0.9032, + "step": 78759 + }, + { + "epoch": 5.82, + "learning_rate": 4.724560441978998e-08, + "loss": 0.908, + "step": 78760 + }, + { + "epoch": 5.82, + "learning_rate": 4.720688521100303e-08, + "loss": 1.0376, + "step": 78761 + }, + { + "epoch": 5.82, + "learning_rate": 4.716818183696736e-08, + "loss": 0.9597, + "step": 78762 + }, + { + "epoch": 5.82, + "learning_rate": 4.712949429774516e-08, + "loss": 1.0113, + "step": 78763 + }, + { + "epoch": 5.82, + "learning_rate": 4.709082259339748e-08, + "loss": 1.0832, + "step": 78764 + }, + { + "epoch": 5.82, + "learning_rate": 4.705216672398649e-08, + "loss": 1.0493, + "step": 78765 + }, + { + "epoch": 5.82, + "learning_rate": 4.701352668957215e-08, + "loss": 0.9622, + "step": 78766 + }, + { + "epoch": 5.82, + "learning_rate": 4.697490249021774e-08, + "loss": 1.0315, + "step": 78767 + }, + { + "epoch": 5.82, + "learning_rate": 4.693629412598322e-08, + "loss": 0.9356, + "step": 78768 + }, + { + "epoch": 5.82, + "learning_rate": 4.689770159693075e-08, + "loss": 1.0204, + "step": 78769 + }, + { + "epoch": 5.82, + "learning_rate": 4.685912490312139e-08, + "loss": 0.8883, + "step": 78770 + }, + { + "epoch": 5.82, + "learning_rate": 4.682056404461732e-08, + "loss": 1.0012, + "step": 78771 + }, + { + "epoch": 5.82, + "learning_rate": 4.678201902147961e-08, + "loss": 0.9986, + "step": 78772 + }, + { + "epoch": 5.82, + "learning_rate": 4.67434898337682e-08, + "loss": 0.984, + "step": 78773 + }, + { + "epoch": 5.82, + "learning_rate": 4.670497648154637e-08, + "loss": 0.9443, + "step": 78774 + }, + { + "epoch": 5.82, + "learning_rate": 4.6666478964874086e-08, + "loss": 1.0351, + "step": 78775 + }, + { + "epoch": 5.82, + "learning_rate": 4.662799728381351e-08, + "loss": 1.0518, + "step": 78776 + }, + { + "epoch": 5.82, + "learning_rate": 4.658953143842459e-08, + "loss": 0.8989, + "step": 78777 + }, + { + "epoch": 5.82, + "learning_rate": 4.655108142876952e-08, + "loss": 1.0686, + "step": 78778 + }, + { + "epoch": 5.82, + "learning_rate": 4.6512647254909335e-08, + "loss": 1.0103, + "step": 78779 + }, + { + "epoch": 5.82, + "learning_rate": 4.6474228916906227e-08, + "loss": 1.034, + "step": 78780 + }, + { + "epoch": 5.82, + "learning_rate": 4.643582641481903e-08, + "loss": 0.9964, + "step": 78781 + }, + { + "epoch": 5.82, + "learning_rate": 4.639743974870992e-08, + "loss": 0.961, + "step": 78782 + }, + { + "epoch": 5.82, + "learning_rate": 4.635906891863995e-08, + "loss": 1.0374, + "step": 78783 + }, + { + "epoch": 5.82, + "learning_rate": 4.63207139246713e-08, + "loss": 0.9469, + "step": 78784 + }, + { + "epoch": 5.82, + "learning_rate": 4.628237476686281e-08, + "loss": 1.0354, + "step": 78785 + }, + { + "epoch": 5.82, + "learning_rate": 4.624405144527777e-08, + "loss": 0.9371, + "step": 78786 + }, + { + "epoch": 5.82, + "learning_rate": 4.6205743959975015e-08, + "loss": 1.1072, + "step": 78787 + }, + { + "epoch": 5.82, + "learning_rate": 4.6167452311017823e-08, + "loss": 1.0432, + "step": 78788 + }, + { + "epoch": 5.82, + "learning_rate": 4.6129176498465045e-08, + "loss": 0.9512, + "step": 78789 + }, + { + "epoch": 5.82, + "learning_rate": 4.609091652237774e-08, + "loss": 1.0047, + "step": 78790 + }, + { + "epoch": 5.82, + "learning_rate": 4.605267238281807e-08, + "loss": 0.9148, + "step": 78791 + }, + { + "epoch": 5.82, + "learning_rate": 4.601444407984601e-08, + "loss": 0.9737, + "step": 78792 + }, + { + "epoch": 5.82, + "learning_rate": 4.5976231613522605e-08, + "loss": 0.9586, + "step": 78793 + }, + { + "epoch": 5.82, + "learning_rate": 4.593803498390781e-08, + "loss": 0.9625, + "step": 78794 + }, + { + "epoch": 5.82, + "learning_rate": 4.5899854191064905e-08, + "loss": 0.9731, + "step": 78795 + }, + { + "epoch": 5.82, + "learning_rate": 4.586168923505163e-08, + "loss": 0.8881, + "step": 78796 + }, + { + "epoch": 5.82, + "learning_rate": 4.5823540115930156e-08, + "loss": 1.0111, + "step": 78797 + }, + { + "epoch": 5.82, + "learning_rate": 4.5785406833761536e-08, + "loss": 1.0392, + "step": 78798 + }, + { + "epoch": 5.82, + "learning_rate": 4.574728938860462e-08, + "loss": 0.9074, + "step": 78799 + }, + { + "epoch": 5.82, + "learning_rate": 4.570918778052269e-08, + "loss": 0.966, + "step": 78800 + }, + { + "epoch": 5.82, + "learning_rate": 4.567110200957459e-08, + "loss": 0.9225, + "step": 78801 + }, + { + "epoch": 5.82, + "learning_rate": 4.563303207582137e-08, + "loss": 1.0908, + "step": 78802 + }, + { + "epoch": 5.82, + "learning_rate": 4.5594977979323e-08, + "loss": 1.0082, + "step": 78803 + }, + { + "epoch": 5.82, + "learning_rate": 4.5556939720141634e-08, + "loss": 0.9828, + "step": 78804 + }, + { + "epoch": 5.82, + "learning_rate": 4.551891729833613e-08, + "loss": 0.9315, + "step": 78805 + }, + { + "epoch": 5.82, + "learning_rate": 4.548091071396754e-08, + "loss": 1.0518, + "step": 78806 + }, + { + "epoch": 5.82, + "learning_rate": 4.544291996709693e-08, + "loss": 1.071, + "step": 78807 + }, + { + "epoch": 5.82, + "learning_rate": 4.540494505778426e-08, + "loss": 1.0094, + "step": 78808 + }, + { + "epoch": 5.82, + "learning_rate": 4.536698598608946e-08, + "loss": 1.0753, + "step": 78809 + }, + { + "epoch": 5.82, + "learning_rate": 4.53290427520725e-08, + "loss": 0.9662, + "step": 78810 + }, + { + "epoch": 5.82, + "learning_rate": 4.529111535579556e-08, + "loss": 0.9467, + "step": 78811 + }, + { + "epoch": 5.82, + "learning_rate": 4.525320379731857e-08, + "loss": 1.0543, + "step": 78812 + }, + { + "epoch": 5.82, + "learning_rate": 4.521530807670038e-08, + "loss": 1.0013, + "step": 78813 + }, + { + "epoch": 5.82, + "learning_rate": 4.517742819400206e-08, + "loss": 0.9385, + "step": 78814 + }, + { + "epoch": 5.82, + "learning_rate": 4.5139564149284666e-08, + "loss": 0.9015, + "step": 78815 + }, + { + "epoch": 5.82, + "learning_rate": 4.510171594260815e-08, + "loss": 1.0011, + "step": 78816 + }, + { + "epoch": 5.82, + "learning_rate": 4.5063883574031355e-08, + "loss": 0.9456, + "step": 78817 + }, + { + "epoch": 5.82, + "learning_rate": 4.5026067043616453e-08, + "loss": 1.0175, + "step": 78818 + }, + { + "epoch": 5.82, + "learning_rate": 4.498826635142117e-08, + "loss": 0.9376, + "step": 78819 + }, + { + "epoch": 5.82, + "learning_rate": 4.4950481497508805e-08, + "loss": 0.9729, + "step": 78820 + }, + { + "epoch": 5.82, + "learning_rate": 4.491271248193707e-08, + "loss": 0.9912, + "step": 78821 + }, + { + "epoch": 5.82, + "learning_rate": 4.487495930476704e-08, + "loss": 0.9991, + "step": 78822 + }, + { + "epoch": 5.82, + "learning_rate": 4.4837221966057554e-08, + "loss": 1.0591, + "step": 78823 + }, + { + "epoch": 5.82, + "learning_rate": 4.4799500465870785e-08, + "loss": 1.0082, + "step": 78824 + }, + { + "epoch": 5.82, + "learning_rate": 4.476179480426557e-08, + "loss": 1.0749, + "step": 78825 + }, + { + "epoch": 5.82, + "learning_rate": 4.4724104981300756e-08, + "loss": 0.9801, + "step": 78826 + }, + { + "epoch": 5.82, + "learning_rate": 4.468643099703851e-08, + "loss": 0.9299, + "step": 78827 + }, + { + "epoch": 5.82, + "learning_rate": 4.4648772851536574e-08, + "loss": 0.9592, + "step": 78828 + }, + { + "epoch": 5.82, + "learning_rate": 4.46111305448571e-08, + "loss": 1.0175, + "step": 78829 + }, + { + "epoch": 5.82, + "learning_rate": 4.457350407705896e-08, + "loss": 0.9703, + "step": 78830 + }, + { + "epoch": 5.82, + "learning_rate": 4.453589344820097e-08, + "loss": 1.0399, + "step": 78831 + }, + { + "epoch": 5.82, + "learning_rate": 4.449829865834421e-08, + "loss": 0.9982, + "step": 78832 + }, + { + "epoch": 5.82, + "learning_rate": 4.4460719707549726e-08, + "loss": 1.134, + "step": 78833 + }, + { + "epoch": 5.82, + "learning_rate": 4.442315659587415e-08, + "loss": 1.0251, + "step": 78834 + }, + { + "epoch": 5.83, + "learning_rate": 4.438560932337965e-08, + "loss": 0.96, + "step": 78835 + }, + { + "epoch": 5.83, + "learning_rate": 4.4348077890123964e-08, + "loss": 1.0649, + "step": 78836 + }, + { + "epoch": 5.83, + "learning_rate": 4.431056229616926e-08, + "loss": 1.0068, + "step": 78837 + }, + { + "epoch": 5.83, + "learning_rate": 4.427306254157437e-08, + "loss": 1.058, + "step": 78838 + }, + { + "epoch": 5.83, + "learning_rate": 4.423557862639705e-08, + "loss": 0.9772, + "step": 78839 + }, + { + "epoch": 5.83, + "learning_rate": 4.419811055069945e-08, + "loss": 1.0325, + "step": 78840 + }, + { + "epoch": 5.83, + "learning_rate": 4.4160658314540415e-08, + "loss": 0.9271, + "step": 78841 + }, + { + "epoch": 5.83, + "learning_rate": 4.41232219179788e-08, + "loss": 0.9856, + "step": 78842 + }, + { + "epoch": 5.83, + "learning_rate": 4.4085801361075654e-08, + "loss": 1.0595, + "step": 78843 + }, + { + "epoch": 5.83, + "learning_rate": 4.404839664388871e-08, + "loss": 1.1565, + "step": 78844 + }, + { + "epoch": 5.83, + "learning_rate": 4.401100776647793e-08, + "loss": 0.9913, + "step": 78845 + }, + { + "epoch": 5.83, + "learning_rate": 4.3973634728904366e-08, + "loss": 0.9734, + "step": 78846 + }, + { + "epoch": 5.83, + "learning_rate": 4.3936277531225755e-08, + "loss": 1.0107, + "step": 78847 + }, + { + "epoch": 5.83, + "learning_rate": 4.3898936173500936e-08, + "loss": 1.0034, + "step": 78848 + }, + { + "epoch": 5.83, + "learning_rate": 4.3861610655792086e-08, + "loss": 0.9807, + "step": 78849 + }, + { + "epoch": 5.83, + "learning_rate": 4.3824300978155824e-08, + "loss": 0.9226, + "step": 78850 + }, + { + "epoch": 5.83, + "learning_rate": 4.37870071406532e-08, + "loss": 1.0261, + "step": 78851 + }, + { + "epoch": 5.83, + "learning_rate": 4.3749729143343076e-08, + "loss": 1.0099, + "step": 78852 + }, + { + "epoch": 5.83, + "learning_rate": 4.371246698628539e-08, + "loss": 1.0562, + "step": 78853 + }, + { + "epoch": 5.83, + "learning_rate": 4.367522066953789e-08, + "loss": 0.9239, + "step": 78854 + }, + { + "epoch": 5.83, + "learning_rate": 4.3637990193160504e-08, + "loss": 1.0043, + "step": 78855 + }, + { + "epoch": 5.83, + "learning_rate": 4.3600775557212093e-08, + "loss": 0.9764, + "step": 78856 + }, + { + "epoch": 5.83, + "learning_rate": 4.356357676175371e-08, + "loss": 0.9273, + "step": 78857 + }, + { + "epoch": 5.83, + "learning_rate": 4.352639380684198e-08, + "loss": 0.9952, + "step": 78858 + }, + { + "epoch": 5.83, + "learning_rate": 4.348922669253797e-08, + "loss": 1.0017, + "step": 78859 + }, + { + "epoch": 5.83, + "learning_rate": 4.3452075418900506e-08, + "loss": 0.9863, + "step": 78860 + }, + { + "epoch": 5.83, + "learning_rate": 4.341493998598845e-08, + "loss": 1.0079, + "step": 78861 + }, + { + "epoch": 5.83, + "learning_rate": 4.3377820393859516e-08, + "loss": 1.068, + "step": 78862 + }, + { + "epoch": 5.83, + "learning_rate": 4.334071664257478e-08, + "loss": 0.9873, + "step": 78863 + }, + { + "epoch": 5.83, + "learning_rate": 4.330362873219307e-08, + "loss": 0.9423, + "step": 78864 + }, + { + "epoch": 5.83, + "learning_rate": 4.326655666277213e-08, + "loss": 0.9342, + "step": 78865 + }, + { + "epoch": 5.83, + "learning_rate": 4.3229500434371906e-08, + "loss": 0.9136, + "step": 78866 + }, + { + "epoch": 5.83, + "learning_rate": 4.3192460047050134e-08, + "loss": 1.1025, + "step": 78867 + }, + { + "epoch": 5.83, + "learning_rate": 4.315543550086787e-08, + "loss": 1.037, + "step": 78868 + }, + { + "epoch": 5.83, + "learning_rate": 4.311842679588285e-08, + "loss": 1.0264, + "step": 78869 + }, + { + "epoch": 5.83, + "learning_rate": 4.3081433932152806e-08, + "loss": 1.0854, + "step": 78870 + }, + { + "epoch": 5.83, + "learning_rate": 4.3044456909738794e-08, + "loss": 1.0106, + "step": 78871 + }, + { + "epoch": 5.83, + "learning_rate": 4.300749572869856e-08, + "loss": 0.8994, + "step": 78872 + }, + { + "epoch": 5.83, + "learning_rate": 4.297055038909093e-08, + "loss": 1.0146, + "step": 78873 + }, + { + "epoch": 5.83, + "learning_rate": 4.293362089097475e-08, + "loss": 0.9844, + "step": 78874 + }, + { + "epoch": 5.83, + "learning_rate": 4.289670723440775e-08, + "loss": 0.9606, + "step": 78875 + }, + { + "epoch": 5.83, + "learning_rate": 4.2859809419451003e-08, + "loss": 0.9213, + "step": 78876 + }, + { + "epoch": 5.83, + "learning_rate": 4.282292744616001e-08, + "loss": 0.9985, + "step": 78877 + }, + { + "epoch": 5.83, + "learning_rate": 4.278606131459695e-08, + "loss": 1.016, + "step": 78878 + }, + { + "epoch": 5.83, + "learning_rate": 4.2749211024818435e-08, + "loss": 1.0088, + "step": 78879 + }, + { + "epoch": 5.83, + "learning_rate": 4.2712376576883316e-08, + "loss": 1.0463, + "step": 78880 + }, + { + "epoch": 5.83, + "learning_rate": 4.2675557970850436e-08, + "loss": 0.9664, + "step": 78881 + }, + { + "epoch": 5.83, + "learning_rate": 4.263875520677752e-08, + "loss": 1.0795, + "step": 78882 + }, + { + "epoch": 5.83, + "learning_rate": 4.260196828472563e-08, + "loss": 0.966, + "step": 78883 + }, + { + "epoch": 5.83, + "learning_rate": 4.256519720474916e-08, + "loss": 0.9087, + "step": 78884 + }, + { + "epoch": 5.83, + "learning_rate": 4.2528441966910305e-08, + "loss": 1.0825, + "step": 78885 + }, + { + "epoch": 5.83, + "learning_rate": 4.249170257126678e-08, + "loss": 0.977, + "step": 78886 + }, + { + "epoch": 5.83, + "learning_rate": 4.245497901787632e-08, + "loss": 0.9975, + "step": 78887 + }, + { + "epoch": 5.83, + "learning_rate": 4.241827130679666e-08, + "loss": 0.9771, + "step": 78888 + }, + { + "epoch": 5.83, + "learning_rate": 4.238157943808774e-08, + "loss": 1.0525, + "step": 78889 + }, + { + "epoch": 5.83, + "learning_rate": 4.2344903411806195e-08, + "loss": 0.9505, + "step": 78890 + }, + { + "epoch": 5.83, + "learning_rate": 4.230824322801308e-08, + "loss": 0.927, + "step": 78891 + }, + { + "epoch": 5.83, + "learning_rate": 4.227159888676391e-08, + "loss": 0.8558, + "step": 78892 + }, + { + "epoch": 5.83, + "learning_rate": 4.2234970388118634e-08, + "loss": 1.009, + "step": 78893 + }, + { + "epoch": 5.83, + "learning_rate": 4.219835773213499e-08, + "loss": 0.9245, + "step": 78894 + }, + { + "epoch": 5.83, + "learning_rate": 4.21617609188707e-08, + "loss": 0.9007, + "step": 78895 + }, + { + "epoch": 5.83, + "learning_rate": 4.212517994838572e-08, + "loss": 0.9157, + "step": 78896 + }, + { + "epoch": 5.83, + "learning_rate": 4.208861482073556e-08, + "loss": 1.0139, + "step": 78897 + }, + { + "epoch": 5.83, + "learning_rate": 4.205206553598129e-08, + "loss": 1.0118, + "step": 78898 + }, + { + "epoch": 5.83, + "learning_rate": 4.201553209417952e-08, + "loss": 0.9482, + "step": 78899 + }, + { + "epoch": 5.83, + "learning_rate": 4.197901449538799e-08, + "loss": 0.9685, + "step": 78900 + }, + { + "epoch": 5.83, + "learning_rate": 4.194251273966554e-08, + "loss": 0.9573, + "step": 78901 + }, + { + "epoch": 5.83, + "learning_rate": 4.1906026827071014e-08, + "loss": 1.04, + "step": 78902 + }, + { + "epoch": 5.83, + "learning_rate": 4.186955675766102e-08, + "loss": 1.1378, + "step": 78903 + }, + { + "epoch": 5.83, + "learning_rate": 4.183310253149331e-08, + "loss": 1.0427, + "step": 78904 + }, + { + "epoch": 5.83, + "learning_rate": 4.1796664148626706e-08, + "loss": 0.9902, + "step": 78905 + }, + { + "epoch": 5.83, + "learning_rate": 4.1760241609120066e-08, + "loss": 0.9646, + "step": 78906 + }, + { + "epoch": 5.83, + "learning_rate": 4.1723834913030006e-08, + "loss": 0.9956, + "step": 78907 + }, + { + "epoch": 5.83, + "learning_rate": 4.168744406041425e-08, + "loss": 1.0778, + "step": 78908 + }, + { + "epoch": 5.83, + "learning_rate": 4.165106905133165e-08, + "loss": 1.0642, + "step": 78909 + }, + { + "epoch": 5.83, + "learning_rate": 4.161470988583993e-08, + "loss": 0.9428, + "step": 78910 + }, + { + "epoch": 5.83, + "learning_rate": 4.157836656399683e-08, + "loss": 0.9701, + "step": 78911 + }, + { + "epoch": 5.83, + "learning_rate": 4.1542039085860075e-08, + "loss": 1.0339, + "step": 78912 + }, + { + "epoch": 5.83, + "learning_rate": 4.1505727451486286e-08, + "loss": 0.9749, + "step": 78913 + }, + { + "epoch": 5.83, + "learning_rate": 4.1469431660936534e-08, + "loss": 1.0602, + "step": 78914 + }, + { + "epoch": 5.83, + "learning_rate": 4.143315171426521e-08, + "loss": 1.049, + "step": 78915 + }, + { + "epoch": 5.83, + "learning_rate": 4.1396887611531156e-08, + "loss": 0.9133, + "step": 78916 + }, + { + "epoch": 5.83, + "learning_rate": 4.136063935279211e-08, + "loss": 1.0049, + "step": 78917 + }, + { + "epoch": 5.83, + "learning_rate": 4.13244069381058e-08, + "loss": 1.0386, + "step": 78918 + }, + { + "epoch": 5.83, + "learning_rate": 4.128819036753107e-08, + "loss": 1.0956, + "step": 78919 + }, + { + "epoch": 5.83, + "learning_rate": 4.125198964112231e-08, + "loss": 1.0833, + "step": 78920 + }, + { + "epoch": 5.83, + "learning_rate": 4.12158047589406e-08, + "loss": 1.0565, + "step": 78921 + }, + { + "epoch": 5.83, + "learning_rate": 4.1179635721041443e-08, + "loss": 1.0234, + "step": 78922 + }, + { + "epoch": 5.83, + "learning_rate": 4.114348252748368e-08, + "loss": 0.8275, + "step": 78923 + }, + { + "epoch": 5.83, + "learning_rate": 4.1107345178323934e-08, + "loss": 0.9879, + "step": 78924 + }, + { + "epoch": 5.83, + "learning_rate": 4.107122367361882e-08, + "loss": 0.9627, + "step": 78925 + }, + { + "epoch": 5.83, + "learning_rate": 4.10351180134283e-08, + "loss": 0.9429, + "step": 78926 + }, + { + "epoch": 5.83, + "learning_rate": 4.099902819780788e-08, + "loss": 1.0733, + "step": 78927 + }, + { + "epoch": 5.83, + "learning_rate": 4.0962954226815286e-08, + "loss": 1.1006, + "step": 78928 + }, + { + "epoch": 5.83, + "learning_rate": 4.092689610050826e-08, + "loss": 0.9703, + "step": 78929 + }, + { + "epoch": 5.83, + "learning_rate": 4.089085381894453e-08, + "loss": 0.9713, + "step": 78930 + }, + { + "epoch": 5.83, + "learning_rate": 4.0854827382180715e-08, + "loss": 1.0104, + "step": 78931 + }, + { + "epoch": 5.83, + "learning_rate": 4.081881679027455e-08, + "loss": 1.0113, + "step": 78932 + }, + { + "epoch": 5.83, + "learning_rate": 4.078282204328154e-08, + "loss": 0.8822, + "step": 78933 + }, + { + "epoch": 5.83, + "learning_rate": 4.074684314126276e-08, + "loss": 1.0395, + "step": 78934 + }, + { + "epoch": 5.83, + "learning_rate": 4.07108800842726e-08, + "loss": 0.9757, + "step": 78935 + }, + { + "epoch": 5.83, + "learning_rate": 4.067493287236879e-08, + "loss": 0.9265, + "step": 78936 + }, + { + "epoch": 5.83, + "learning_rate": 4.0639001505607957e-08, + "loss": 1.0063, + "step": 78937 + }, + { + "epoch": 5.83, + "learning_rate": 4.060308598404894e-08, + "loss": 0.9978, + "step": 78938 + }, + { + "epoch": 5.83, + "learning_rate": 4.0567186307747255e-08, + "loss": 0.9232, + "step": 78939 + }, + { + "epoch": 5.83, + "learning_rate": 4.053130247676174e-08, + "loss": 1.049, + "step": 78940 + }, + { + "epoch": 5.83, + "learning_rate": 4.049543449114679e-08, + "loss": 1.0723, + "step": 78941 + }, + { + "epoch": 5.83, + "learning_rate": 4.045958235096237e-08, + "loss": 1.0515, + "step": 78942 + }, + { + "epoch": 5.83, + "learning_rate": 4.042374605626398e-08, + "loss": 0.9492, + "step": 78943 + }, + { + "epoch": 5.83, + "learning_rate": 4.038792560710825e-08, + "loss": 1.068, + "step": 78944 + }, + { + "epoch": 5.83, + "learning_rate": 4.035212100355401e-08, + "loss": 0.9411, + "step": 78945 + }, + { + "epoch": 5.83, + "learning_rate": 4.031633224565679e-08, + "loss": 0.9196, + "step": 78946 + }, + { + "epoch": 5.83, + "learning_rate": 4.028055933347319e-08, + "loss": 0.9591, + "step": 78947 + }, + { + "epoch": 5.83, + "learning_rate": 4.0244802267060956e-08, + "loss": 0.9516, + "step": 78948 + }, + { + "epoch": 5.83, + "learning_rate": 4.020906104647671e-08, + "loss": 1.058, + "step": 78949 + }, + { + "epoch": 5.83, + "learning_rate": 4.0173335671777055e-08, + "loss": 1.0008, + "step": 78950 + }, + { + "epoch": 5.83, + "learning_rate": 4.0137626143019745e-08, + "loss": 1.1298, + "step": 78951 + }, + { + "epoch": 5.83, + "learning_rate": 4.0101932460261394e-08, + "loss": 1.0328, + "step": 78952 + }, + { + "epoch": 5.83, + "learning_rate": 4.00662546235564e-08, + "loss": 1.0124, + "step": 78953 + }, + { + "epoch": 5.83, + "learning_rate": 4.003059263296582e-08, + "loss": 1.082, + "step": 78954 + }, + { + "epoch": 5.83, + "learning_rate": 3.9994946488542966e-08, + "loss": 1.0041, + "step": 78955 + }, + { + "epoch": 5.83, + "learning_rate": 3.9959316190345544e-08, + "loss": 0.9219, + "step": 78956 + }, + { + "epoch": 5.83, + "learning_rate": 3.992370173843019e-08, + "loss": 1.0973, + "step": 78957 + }, + { + "epoch": 5.83, + "learning_rate": 3.988810313285352e-08, + "loss": 1.0155, + "step": 78958 + }, + { + "epoch": 5.83, + "learning_rate": 3.985252037367326e-08, + "loss": 0.9437, + "step": 78959 + }, + { + "epoch": 5.83, + "learning_rate": 3.981695346094494e-08, + "loss": 1.0513, + "step": 78960 + }, + { + "epoch": 5.83, + "learning_rate": 3.978140239472517e-08, + "loss": 1.0035, + "step": 78961 + }, + { + "epoch": 5.83, + "learning_rate": 3.9745867175070564e-08, + "loss": 1.0309, + "step": 78962 + }, + { + "epoch": 5.83, + "learning_rate": 3.9710347802037754e-08, + "loss": 0.8513, + "step": 78963 + }, + { + "epoch": 5.83, + "learning_rate": 3.967484427568447e-08, + "loss": 0.904, + "step": 78964 + }, + { + "epoch": 5.83, + "learning_rate": 3.9639356596063996e-08, + "loss": 1.0268, + "step": 78965 + }, + { + "epoch": 5.83, + "learning_rate": 3.960388476323629e-08, + "loss": 0.9097, + "step": 78966 + }, + { + "epoch": 5.83, + "learning_rate": 3.956842877725464e-08, + "loss": 0.9264, + "step": 78967 + }, + { + "epoch": 5.83, + "learning_rate": 3.9532988638179005e-08, + "loss": 1.0106, + "step": 78968 + }, + { + "epoch": 5.83, + "learning_rate": 3.949756434606267e-08, + "loss": 0.9971, + "step": 78969 + }, + { + "epoch": 5.84, + "learning_rate": 3.946215590096336e-08, + "loss": 1.0132, + "step": 78970 + }, + { + "epoch": 5.84, + "learning_rate": 3.942676330293771e-08, + "loss": 0.9583, + "step": 78971 + }, + { + "epoch": 5.84, + "learning_rate": 3.939138655204122e-08, + "loss": 0.9894, + "step": 78972 + }, + { + "epoch": 5.84, + "learning_rate": 3.935602564833052e-08, + "loss": 0.9608, + "step": 78973 + }, + { + "epoch": 5.84, + "learning_rate": 3.932068059186222e-08, + "loss": 0.9748, + "step": 78974 + }, + { + "epoch": 5.84, + "learning_rate": 3.928535138269185e-08, + "loss": 0.9926, + "step": 78975 + }, + { + "epoch": 5.84, + "learning_rate": 3.9250038020876017e-08, + "loss": 1.0493, + "step": 78976 + }, + { + "epoch": 5.84, + "learning_rate": 3.921474050647134e-08, + "loss": 0.9888, + "step": 78977 + }, + { + "epoch": 5.84, + "learning_rate": 3.9179458839533334e-08, + "loss": 1.0617, + "step": 78978 + }, + { + "epoch": 5.84, + "learning_rate": 3.914419302011752e-08, + "loss": 1.0482, + "step": 78979 + }, + { + "epoch": 5.84, + "learning_rate": 3.9108943048281614e-08, + "loss": 1.0602, + "step": 78980 + }, + { + "epoch": 5.84, + "learning_rate": 3.907370892408002e-08, + "loss": 1.057, + "step": 78981 + }, + { + "epoch": 5.84, + "learning_rate": 3.9038490647570485e-08, + "loss": 0.9948, + "step": 78982 + }, + { + "epoch": 5.84, + "learning_rate": 3.900328821880739e-08, + "loss": 0.8909, + "step": 78983 + }, + { + "epoch": 5.84, + "learning_rate": 3.896810163784848e-08, + "loss": 0.9353, + "step": 78984 + }, + { + "epoch": 5.84, + "learning_rate": 3.8932930904748144e-08, + "loss": 1.0345, + "step": 78985 + }, + { + "epoch": 5.84, + "learning_rate": 3.889777601956413e-08, + "loss": 0.9138, + "step": 78986 + }, + { + "epoch": 5.84, + "learning_rate": 3.886263698234971e-08, + "loss": 0.9358, + "step": 78987 + }, + { + "epoch": 5.84, + "learning_rate": 3.882751379316374e-08, + "loss": 1.0141, + "step": 78988 + }, + { + "epoch": 5.84, + "learning_rate": 3.87924064520595e-08, + "loss": 0.9857, + "step": 78989 + }, + { + "epoch": 5.84, + "learning_rate": 3.875731495909363e-08, + "loss": 0.9826, + "step": 78990 + }, + { + "epoch": 5.84, + "learning_rate": 3.872223931432273e-08, + "loss": 0.9393, + "step": 78991 + }, + { + "epoch": 5.84, + "learning_rate": 3.868717951780343e-08, + "loss": 1.1338, + "step": 78992 + }, + { + "epoch": 5.84, + "learning_rate": 3.865213556958902e-08, + "loss": 1.0444, + "step": 78993 + }, + { + "epoch": 5.84, + "learning_rate": 3.8617107469737234e-08, + "loss": 0.9972, + "step": 78994 + }, + { + "epoch": 5.84, + "learning_rate": 3.8582095218302474e-08, + "loss": 1.0531, + "step": 78995 + }, + { + "epoch": 5.84, + "learning_rate": 3.8547098815342466e-08, + "loss": 0.9586, + "step": 78996 + }, + { + "epoch": 5.84, + "learning_rate": 3.8512118260910505e-08, + "loss": 0.9503, + "step": 78997 + }, + { + "epoch": 5.84, + "learning_rate": 3.8477153555062093e-08, + "loss": 1.0539, + "step": 78998 + }, + { + "epoch": 5.84, + "learning_rate": 3.844220469785498e-08, + "loss": 1.0431, + "step": 78999 + }, + { + "epoch": 5.84, + "learning_rate": 3.840727168934466e-08, + "loss": 0.9991, + "step": 79000 + }, + { + "epoch": 5.84, + "learning_rate": 3.837235452958554e-08, + "loss": 1.1249, + "step": 79001 + }, + { + "epoch": 5.84, + "learning_rate": 3.8337453218633137e-08, + "loss": 1.0107, + "step": 79002 + }, + { + "epoch": 5.84, + "learning_rate": 3.8302567756542955e-08, + "loss": 1.0632, + "step": 79003 + }, + { + "epoch": 5.84, + "learning_rate": 3.82676981433705e-08, + "loss": 1.0863, + "step": 79004 + }, + { + "epoch": 5.84, + "learning_rate": 3.8232844379172405e-08, + "loss": 0.8887, + "step": 79005 + }, + { + "epoch": 5.84, + "learning_rate": 3.819800646400307e-08, + "loss": 0.9949, + "step": 79006 + }, + { + "epoch": 5.84, + "learning_rate": 3.8163184397917995e-08, + "loss": 0.9459, + "step": 79007 + }, + { + "epoch": 5.84, + "learning_rate": 3.8128378180973814e-08, + "loss": 1.0381, + "step": 79008 + }, + { + "epoch": 5.84, + "learning_rate": 3.809358781322381e-08, + "loss": 0.9871, + "step": 79009 + }, + { + "epoch": 5.84, + "learning_rate": 3.8058813294724605e-08, + "loss": 0.9001, + "step": 79010 + }, + { + "epoch": 5.84, + "learning_rate": 3.8024054625531715e-08, + "loss": 1.0076, + "step": 79011 + }, + { + "epoch": 5.84, + "learning_rate": 3.798931180569953e-08, + "loss": 0.895, + "step": 79012 + }, + { + "epoch": 5.84, + "learning_rate": 3.795458483528358e-08, + "loss": 1.0955, + "step": 79013 + }, + { + "epoch": 5.84, + "learning_rate": 3.791987371433936e-08, + "loss": 1.0323, + "step": 79014 + }, + { + "epoch": 5.84, + "learning_rate": 3.7885178442923496e-08, + "loss": 0.905, + "step": 79015 + }, + { + "epoch": 5.84, + "learning_rate": 3.785049902108817e-08, + "loss": 1.074, + "step": 79016 + }, + { + "epoch": 5.84, + "learning_rate": 3.7815835448891115e-08, + "loss": 1.0147, + "step": 79017 + }, + { + "epoch": 5.84, + "learning_rate": 3.7781187726385614e-08, + "loss": 0.9936, + "step": 79018 + }, + { + "epoch": 5.84, + "learning_rate": 3.77465558536283e-08, + "loss": 0.9498, + "step": 79019 + }, + { + "epoch": 5.84, + "learning_rate": 3.771193983067356e-08, + "loss": 0.9819, + "step": 79020 + }, + { + "epoch": 5.84, + "learning_rate": 3.7677339657576915e-08, + "loss": 1.0272, + "step": 79021 + }, + { + "epoch": 5.84, + "learning_rate": 3.7642755334392765e-08, + "loss": 1.0002, + "step": 79022 + }, + { + "epoch": 5.84, + "learning_rate": 3.760818686117662e-08, + "loss": 1.0435, + "step": 79023 + }, + { + "epoch": 5.84, + "learning_rate": 3.757363423798288e-08, + "loss": 0.9529, + "step": 79024 + }, + { + "epoch": 5.84, + "learning_rate": 3.753909746486706e-08, + "loss": 0.9442, + "step": 79025 + }, + { + "epoch": 5.84, + "learning_rate": 3.7504576541884664e-08, + "loss": 0.9664, + "step": 79026 + }, + { + "epoch": 5.84, + "learning_rate": 3.747007146908899e-08, + "loss": 1.1299, + "step": 79027 + }, + { + "epoch": 5.84, + "learning_rate": 3.743558224653665e-08, + "loss": 0.9993, + "step": 79028 + }, + { + "epoch": 5.84, + "learning_rate": 3.7401108874280945e-08, + "loss": 0.9796, + "step": 79029 + }, + { + "epoch": 5.84, + "learning_rate": 3.736665135237849e-08, + "loss": 1.0031, + "step": 79030 + }, + { + "epoch": 5.84, + "learning_rate": 3.733220968088258e-08, + "loss": 0.9021, + "step": 79031 + }, + { + "epoch": 5.84, + "learning_rate": 3.729778385984872e-08, + "loss": 0.9524, + "step": 79032 + }, + { + "epoch": 5.84, + "learning_rate": 3.726337388933243e-08, + "loss": 0.9711, + "step": 79033 + }, + { + "epoch": 5.84, + "learning_rate": 3.7228979769385886e-08, + "loss": 1.0251, + "step": 79034 + }, + { + "epoch": 5.84, + "learning_rate": 3.7194601500066817e-08, + "loss": 1.1528, + "step": 79035 + }, + { + "epoch": 5.84, + "learning_rate": 3.7160239081427405e-08, + "loss": 0.9525, + "step": 79036 + }, + { + "epoch": 5.84, + "learning_rate": 3.7125892513524275e-08, + "loss": 1.0788, + "step": 79037 + }, + { + "epoch": 5.84, + "learning_rate": 3.7091561796411824e-08, + "loss": 1.0102, + "step": 79038 + }, + { + "epoch": 5.84, + "learning_rate": 3.705724693014334e-08, + "loss": 1.006, + "step": 79039 + }, + { + "epoch": 5.84, + "learning_rate": 3.702294791477434e-08, + "loss": 0.895, + "step": 79040 + }, + { + "epoch": 5.84, + "learning_rate": 3.6988664750359224e-08, + "loss": 1.0313, + "step": 79041 + }, + { + "epoch": 5.84, + "learning_rate": 3.6954397436953505e-08, + "loss": 0.9416, + "step": 79042 + }, + { + "epoch": 5.84, + "learning_rate": 3.6920145974609356e-08, + "loss": 1.0711, + "step": 79043 + }, + { + "epoch": 5.84, + "learning_rate": 3.68859103633834e-08, + "loss": 0.9636, + "step": 79044 + }, + { + "epoch": 5.84, + "learning_rate": 3.685169060332894e-08, + "loss": 1.1354, + "step": 79045 + }, + { + "epoch": 5.84, + "learning_rate": 3.681748669450147e-08, + "loss": 0.7925, + "step": 79046 + }, + { + "epoch": 5.84, + "learning_rate": 3.678329863695429e-08, + "loss": 0.9569, + "step": 79047 + }, + { + "epoch": 5.84, + "learning_rate": 3.674912643074291e-08, + "loss": 1.0969, + "step": 79048 + }, + { + "epoch": 5.84, + "learning_rate": 3.671497007592173e-08, + "loss": 0.923, + "step": 79049 + }, + { + "epoch": 5.84, + "learning_rate": 3.668082957254293e-08, + "loss": 0.9182, + "step": 79050 + }, + { + "epoch": 5.84, + "learning_rate": 3.6646704920664246e-08, + "loss": 0.8609, + "step": 79051 + }, + { + "epoch": 5.84, + "learning_rate": 3.661259612033674e-08, + "loss": 0.9551, + "step": 79052 + }, + { + "epoch": 5.84, + "learning_rate": 3.657850317161593e-08, + "loss": 1.0142, + "step": 79053 + }, + { + "epoch": 5.84, + "learning_rate": 3.6544426074556214e-08, + "loss": 0.9823, + "step": 79054 + }, + { + "epoch": 5.84, + "learning_rate": 3.6510364829212e-08, + "loss": 1.0128, + "step": 79055 + }, + { + "epoch": 5.84, + "learning_rate": 3.6476319435637675e-08, + "loss": 1.1323, + "step": 79056 + }, + { + "epoch": 5.84, + "learning_rate": 3.644228989388654e-08, + "loss": 1.0795, + "step": 79057 + }, + { + "epoch": 5.84, + "learning_rate": 3.6408276204014105e-08, + "loss": 1.0551, + "step": 79058 + }, + { + "epoch": 5.84, + "learning_rate": 3.6374278366072547e-08, + "loss": 1.0376, + "step": 79059 + }, + { + "epoch": 5.84, + "learning_rate": 3.634029638011627e-08, + "loss": 1.0849, + "step": 79060 + }, + { + "epoch": 5.84, + "learning_rate": 3.630633024620078e-08, + "loss": 1.026, + "step": 79061 + }, + { + "epoch": 5.84, + "learning_rate": 3.627237996438049e-08, + "loss": 0.9532, + "step": 79062 + }, + { + "epoch": 5.84, + "learning_rate": 3.623844553470646e-08, + "loss": 0.9897, + "step": 79063 + }, + { + "epoch": 5.84, + "learning_rate": 3.620452695723531e-08, + "loss": 0.9879, + "step": 79064 + }, + { + "epoch": 5.84, + "learning_rate": 3.6170624232020336e-08, + "loss": 0.9978, + "step": 79065 + }, + { + "epoch": 5.84, + "learning_rate": 3.6136737359114825e-08, + "loss": 1.0247, + "step": 79066 + }, + { + "epoch": 5.84, + "learning_rate": 3.6102866338574295e-08, + "loss": 0.9038, + "step": 79067 + }, + { + "epoch": 5.84, + "learning_rate": 3.606901117044981e-08, + "loss": 0.923, + "step": 79068 + }, + { + "epoch": 5.84, + "learning_rate": 3.60351718547991e-08, + "loss": 0.9628, + "step": 79069 + }, + { + "epoch": 5.84, + "learning_rate": 3.600134839167213e-08, + "loss": 0.9897, + "step": 79070 + }, + { + "epoch": 5.84, + "learning_rate": 3.596754078112552e-08, + "loss": 1.0078, + "step": 79071 + }, + { + "epoch": 5.84, + "learning_rate": 3.593374902321145e-08, + "loss": 0.9345, + "step": 79072 + }, + { + "epoch": 5.84, + "learning_rate": 3.589997311798432e-08, + "loss": 1.0114, + "step": 79073 + }, + { + "epoch": 5.84, + "learning_rate": 3.586621306549742e-08, + "loss": 0.9585, + "step": 79074 + }, + { + "epoch": 5.84, + "learning_rate": 3.583246886580516e-08, + "loss": 0.9721, + "step": 79075 + }, + { + "epoch": 5.84, + "learning_rate": 3.579874051896082e-08, + "loss": 1.057, + "step": 79076 + }, + { + "epoch": 5.84, + "learning_rate": 3.576502802501769e-08, + "loss": 1.003, + "step": 79077 + }, + { + "epoch": 5.84, + "learning_rate": 3.5731331384031285e-08, + "loss": 1.0212, + "step": 79078 + }, + { + "epoch": 5.84, + "learning_rate": 3.569765059605268e-08, + "loss": 1.0322, + "step": 79079 + }, + { + "epoch": 5.84, + "learning_rate": 3.566398566113627e-08, + "loss": 0.9245, + "step": 79080 + }, + { + "epoch": 5.84, + "learning_rate": 3.563033657933645e-08, + "loss": 0.9897, + "step": 79081 + }, + { + "epoch": 5.84, + "learning_rate": 3.5596703350706526e-08, + "loss": 1.0457, + "step": 79082 + }, + { + "epoch": 5.84, + "learning_rate": 3.556308597529867e-08, + "loss": 0.961, + "step": 79083 + }, + { + "epoch": 5.84, + "learning_rate": 3.552948445316729e-08, + "loss": 0.9634, + "step": 79084 + }, + { + "epoch": 5.84, + "learning_rate": 3.5495898784366766e-08, + "loss": 0.9374, + "step": 79085 + }, + { + "epoch": 5.84, + "learning_rate": 3.546232896895041e-08, + "loss": 0.8892, + "step": 79086 + }, + { + "epoch": 5.84, + "learning_rate": 3.5428775006969285e-08, + "loss": 1.0818, + "step": 79087 + }, + { + "epoch": 5.84, + "learning_rate": 3.53952368984789e-08, + "loss": 0.9611, + "step": 79088 + }, + { + "epoch": 5.84, + "learning_rate": 3.536171464353255e-08, + "loss": 1.0475, + "step": 79089 + }, + { + "epoch": 5.84, + "learning_rate": 3.5328208242182413e-08, + "loss": 0.8852, + "step": 79090 + }, + { + "epoch": 5.84, + "learning_rate": 3.5294717694484006e-08, + "loss": 0.9752, + "step": 79091 + }, + { + "epoch": 5.84, + "learning_rate": 3.526124300048728e-08, + "loss": 0.9138, + "step": 79092 + }, + { + "epoch": 5.84, + "learning_rate": 3.522778416024886e-08, + "loss": 1.0336, + "step": 79093 + }, + { + "epoch": 5.84, + "learning_rate": 3.519434117381981e-08, + "loss": 0.9586, + "step": 79094 + }, + { + "epoch": 5.84, + "learning_rate": 3.5160914041253435e-08, + "loss": 0.994, + "step": 79095 + }, + { + "epoch": 5.84, + "learning_rate": 3.5127502762604126e-08, + "loss": 1.0002, + "step": 79096 + }, + { + "epoch": 5.84, + "learning_rate": 3.5094107337924066e-08, + "loss": 1.0467, + "step": 79097 + }, + { + "epoch": 5.84, + "learning_rate": 3.506072776726766e-08, + "loss": 1.0298, + "step": 79098 + }, + { + "epoch": 5.84, + "learning_rate": 3.5027364050685965e-08, + "loss": 0.9208, + "step": 79099 + }, + { + "epoch": 5.84, + "learning_rate": 3.499401618823339e-08, + "loss": 1.1467, + "step": 79100 + }, + { + "epoch": 5.84, + "learning_rate": 3.496068417996323e-08, + "loss": 0.911, + "step": 79101 + }, + { + "epoch": 5.84, + "learning_rate": 3.492736802592767e-08, + "loss": 1.0549, + "step": 79102 + }, + { + "epoch": 5.84, + "learning_rate": 3.489406772617998e-08, + "loss": 0.9544, + "step": 79103 + }, + { + "epoch": 5.84, + "learning_rate": 3.4860783280773466e-08, + "loss": 1.1206, + "step": 79104 + }, + { + "epoch": 5.85, + "learning_rate": 3.482751468976142e-08, + "loss": 1.0935, + "step": 79105 + }, + { + "epoch": 5.85, + "learning_rate": 3.479426195319491e-08, + "loss": 1.1553, + "step": 79106 + }, + { + "epoch": 5.85, + "learning_rate": 3.476102507112944e-08, + "loss": 0.965, + "step": 79107 + }, + { + "epoch": 5.85, + "learning_rate": 3.472780404361498e-08, + "loss": 0.9879, + "step": 79108 + }, + { + "epoch": 5.85, + "learning_rate": 3.469459887070703e-08, + "loss": 0.9983, + "step": 79109 + }, + { + "epoch": 5.85, + "learning_rate": 3.466140955245778e-08, + "loss": 1.0245, + "step": 79110 + }, + { + "epoch": 5.85, + "learning_rate": 3.462823608891941e-08, + "loss": 0.9332, + "step": 79111 + }, + { + "epoch": 5.85, + "learning_rate": 3.4595078480145204e-08, + "loss": 1.0701, + "step": 79112 + }, + { + "epoch": 5.85, + "learning_rate": 3.456193672618624e-08, + "loss": 0.9875, + "step": 79113 + }, + { + "epoch": 5.85, + "learning_rate": 3.4528810827098024e-08, + "loss": 1.0485, + "step": 79114 + }, + { + "epoch": 5.85, + "learning_rate": 3.449570078293163e-08, + "loss": 0.9188, + "step": 79115 + }, + { + "epoch": 5.85, + "learning_rate": 3.4462606593740346e-08, + "loss": 0.9578, + "step": 79116 + }, + { + "epoch": 5.85, + "learning_rate": 3.442952825957524e-08, + "loss": 1.0025, + "step": 79117 + }, + { + "epoch": 5.85, + "learning_rate": 3.439646578049072e-08, + "loss": 0.9315, + "step": 79118 + }, + { + "epoch": 5.85, + "learning_rate": 3.4363419156540066e-08, + "loss": 1.133, + "step": 79119 + }, + { + "epoch": 5.85, + "learning_rate": 3.433038838777325e-08, + "loss": 0.9811, + "step": 79120 + }, + { + "epoch": 5.85, + "learning_rate": 3.4297373474244664e-08, + "loss": 0.857, + "step": 79121 + }, + { + "epoch": 5.85, + "learning_rate": 3.4264374416005384e-08, + "loss": 0.9964, + "step": 79122 + }, + { + "epoch": 5.85, + "learning_rate": 3.423139121311092e-08, + "loss": 1.0847, + "step": 79123 + }, + { + "epoch": 5.85, + "learning_rate": 3.419842386561012e-08, + "loss": 1.0401, + "step": 79124 + }, + { + "epoch": 5.85, + "learning_rate": 3.4165472373557385e-08, + "loss": 1.0704, + "step": 79125 + }, + { + "epoch": 5.85, + "learning_rate": 3.413253673700489e-08, + "loss": 0.9843, + "step": 79126 + }, + { + "epoch": 5.85, + "learning_rate": 3.409961695600483e-08, + "loss": 1.0022, + "step": 79127 + }, + { + "epoch": 5.85, + "learning_rate": 3.406671303061049e-08, + "loss": 0.9592, + "step": 79128 + }, + { + "epoch": 5.85, + "learning_rate": 3.403382496087293e-08, + "loss": 1.0564, + "step": 79129 + }, + { + "epoch": 5.85, + "learning_rate": 3.400095274684434e-08, + "loss": 0.8975, + "step": 79130 + }, + { + "epoch": 5.85, + "learning_rate": 3.3968096388578007e-08, + "loss": 1.0886, + "step": 79131 + }, + { + "epoch": 5.85, + "learning_rate": 3.3935255886126116e-08, + "loss": 1.0875, + "step": 79132 + }, + { + "epoch": 5.85, + "learning_rate": 3.3902431239539736e-08, + "loss": 0.9027, + "step": 79133 + }, + { + "epoch": 5.85, + "learning_rate": 3.386962244887326e-08, + "loss": 1.1318, + "step": 79134 + }, + { + "epoch": 5.85, + "learning_rate": 3.383682951417666e-08, + "loss": 1.0138, + "step": 79135 + }, + { + "epoch": 5.85, + "learning_rate": 3.380405243550433e-08, + "loss": 0.9782, + "step": 79136 + }, + { + "epoch": 5.85, + "learning_rate": 3.377129121290623e-08, + "loss": 1.0524, + "step": 79137 + }, + { + "epoch": 5.85, + "learning_rate": 3.3738545846435656e-08, + "loss": 0.931, + "step": 79138 + }, + { + "epoch": 5.85, + "learning_rate": 3.3705816336143674e-08, + "loss": 0.9836, + "step": 79139 + }, + { + "epoch": 5.85, + "learning_rate": 3.367310268208357e-08, + "loss": 0.9453, + "step": 79140 + }, + { + "epoch": 5.85, + "learning_rate": 3.364040488430753e-08, + "loss": 1.1386, + "step": 79141 + }, + { + "epoch": 5.85, + "learning_rate": 3.360772294286663e-08, + "loss": 1.0093, + "step": 79142 + }, + { + "epoch": 5.85, + "learning_rate": 3.3575056857813036e-08, + "loss": 1.0074, + "step": 79143 + }, + { + "epoch": 5.85, + "learning_rate": 3.3542406629198944e-08, + "loss": 0.9458, + "step": 79144 + }, + { + "epoch": 5.85, + "learning_rate": 3.350977225707763e-08, + "loss": 0.9593, + "step": 79145 + }, + { + "epoch": 5.85, + "learning_rate": 3.347715374149796e-08, + "loss": 0.9793, + "step": 79146 + }, + { + "epoch": 5.85, + "learning_rate": 3.344455108251432e-08, + "loss": 0.9687, + "step": 79147 + }, + { + "epoch": 5.85, + "learning_rate": 3.341196428017779e-08, + "loss": 0.9852, + "step": 79148 + }, + { + "epoch": 5.85, + "learning_rate": 3.3379393334539435e-08, + "loss": 0.9805, + "step": 79149 + }, + { + "epoch": 5.85, + "learning_rate": 3.334683824565255e-08, + "loss": 0.9623, + "step": 79150 + }, + { + "epoch": 5.85, + "learning_rate": 3.33142990135682e-08, + "loss": 0.9748, + "step": 79151 + }, + { + "epoch": 5.85, + "learning_rate": 3.3281775638338565e-08, + "loss": 1.0378, + "step": 79152 + }, + { + "epoch": 5.85, + "learning_rate": 3.3249268120014723e-08, + "loss": 0.9809, + "step": 79153 + }, + { + "epoch": 5.85, + "learning_rate": 3.3216776458648846e-08, + "loss": 0.971, + "step": 79154 + }, + { + "epoch": 5.85, + "learning_rate": 3.3184300654292014e-08, + "loss": 0.8913, + "step": 79155 + }, + { + "epoch": 5.85, + "learning_rate": 3.31518407069964e-08, + "loss": 1.0204, + "step": 79156 + }, + { + "epoch": 5.85, + "learning_rate": 3.311939661681307e-08, + "loss": 0.9852, + "step": 79157 + }, + { + "epoch": 5.85, + "learning_rate": 3.308696838379533e-08, + "loss": 1.1197, + "step": 79158 + }, + { + "epoch": 5.85, + "learning_rate": 3.3054556007993124e-08, + "loss": 1.1177, + "step": 79159 + }, + { + "epoch": 5.85, + "learning_rate": 3.302215948945864e-08, + "loss": 0.9838, + "step": 79160 + }, + { + "epoch": 5.85, + "learning_rate": 3.298977882824406e-08, + "loss": 0.9341, + "step": 79161 + }, + { + "epoch": 5.85, + "learning_rate": 3.295741402439823e-08, + "loss": 0.8984, + "step": 79162 + }, + { + "epoch": 5.85, + "learning_rate": 3.2925065077975547e-08, + "loss": 1.007, + "step": 79163 + }, + { + "epoch": 5.85, + "learning_rate": 3.2892731989027095e-08, + "loss": 0.9475, + "step": 79164 + }, + { + "epoch": 5.85, + "learning_rate": 3.286041475760282e-08, + "loss": 0.9675, + "step": 79165 + }, + { + "epoch": 5.85, + "learning_rate": 3.282811338375602e-08, + "loss": 1.0001, + "step": 79166 + }, + { + "epoch": 5.85, + "learning_rate": 3.2795827867535544e-08, + "loss": 0.998, + "step": 79167 + }, + { + "epoch": 5.85, + "learning_rate": 3.276355820899579e-08, + "loss": 0.9968, + "step": 79168 + }, + { + "epoch": 5.85, + "learning_rate": 3.2731304408185615e-08, + "loss": 1.0573, + "step": 79169 + }, + { + "epoch": 5.85, + "learning_rate": 3.26990664651583e-08, + "loss": 0.9281, + "step": 79170 + }, + { + "epoch": 5.85, + "learning_rate": 3.26668443799627e-08, + "loss": 0.9191, + "step": 79171 + }, + { + "epoch": 5.85, + "learning_rate": 3.263463815265322e-08, + "loss": 0.927, + "step": 79172 + }, + { + "epoch": 5.85, + "learning_rate": 3.26024477832787e-08, + "loss": 1.0117, + "step": 79173 + }, + { + "epoch": 5.85, + "learning_rate": 3.257027327189133e-08, + "loss": 0.9567, + "step": 79174 + }, + { + "epoch": 5.85, + "learning_rate": 3.253811461854106e-08, + "loss": 0.982, + "step": 79175 + }, + { + "epoch": 5.85, + "learning_rate": 3.250597182328119e-08, + "loss": 1.0679, + "step": 79176 + }, + { + "epoch": 5.85, + "learning_rate": 3.247384488616057e-08, + "loss": 0.9797, + "step": 79177 + }, + { + "epoch": 5.85, + "learning_rate": 3.244173380723248e-08, + "loss": 0.9883, + "step": 79178 + }, + { + "epoch": 5.85, + "learning_rate": 3.2409638586546884e-08, + "loss": 1.0237, + "step": 79179 + }, + { + "epoch": 5.85, + "learning_rate": 3.237755922415375e-08, + "loss": 1.0003, + "step": 79180 + }, + { + "epoch": 5.85, + "learning_rate": 3.234549572010637e-08, + "loss": 0.9645, + "step": 79181 + }, + { + "epoch": 5.85, + "learning_rate": 3.231344807445469e-08, + "loss": 1.0191, + "step": 79182 + }, + { + "epoch": 5.85, + "learning_rate": 3.228141628724979e-08, + "loss": 1.0125, + "step": 79183 + }, + { + "epoch": 5.85, + "learning_rate": 3.224940035854163e-08, + "loss": 1.0388, + "step": 79184 + }, + { + "epoch": 5.85, + "learning_rate": 3.2217400288383496e-08, + "loss": 1.0417, + "step": 79185 + }, + { + "epoch": 5.85, + "learning_rate": 3.218541607682313e-08, + "loss": 1.0257, + "step": 79186 + }, + { + "epoch": 5.85, + "learning_rate": 3.215344772391493e-08, + "loss": 0.9697, + "step": 79187 + }, + { + "epoch": 5.85, + "learning_rate": 3.212149522970664e-08, + "loss": 1.0365, + "step": 79188 + }, + { + "epoch": 5.85, + "learning_rate": 3.208955859425156e-08, + "loss": 0.9467, + "step": 79189 + }, + { + "epoch": 5.85, + "learning_rate": 3.205763781759852e-08, + "loss": 0.9102, + "step": 79190 + }, + { + "epoch": 5.85, + "learning_rate": 3.202573289979971e-08, + "loss": 1.0798, + "step": 79191 + }, + { + "epoch": 5.85, + "learning_rate": 3.1993843840905095e-08, + "loss": 0.9087, + "step": 79192 + }, + { + "epoch": 5.85, + "learning_rate": 3.1961970640964623e-08, + "loss": 0.9728, + "step": 79193 + }, + { + "epoch": 5.85, + "learning_rate": 3.19301133000316e-08, + "loss": 1.0961, + "step": 79194 + }, + { + "epoch": 5.85, + "learning_rate": 3.1898271818153745e-08, + "loss": 1.0244, + "step": 79195 + }, + { + "epoch": 5.85, + "learning_rate": 3.186644619538437e-08, + "loss": 1.0536, + "step": 79196 + }, + { + "epoch": 5.85, + "learning_rate": 3.1834636431772315e-08, + "loss": 1.0666, + "step": 79197 + }, + { + "epoch": 5.85, + "learning_rate": 3.180284252736976e-08, + "loss": 1.0146, + "step": 79198 + }, + { + "epoch": 5.85, + "learning_rate": 3.177106448222445e-08, + "loss": 1.0061, + "step": 79199 + }, + { + "epoch": 5.85, + "learning_rate": 3.173930229638966e-08, + "loss": 0.8939, + "step": 79200 + }, + { + "epoch": 5.85, + "learning_rate": 3.1707555969915374e-08, + "loss": 0.9428, + "step": 79201 + }, + { + "epoch": 5.85, + "learning_rate": 3.167582550285153e-08, + "loss": 1.0296, + "step": 79202 + }, + { + "epoch": 5.85, + "learning_rate": 3.1644110895248105e-08, + "loss": 0.9538, + "step": 79203 + }, + { + "epoch": 5.85, + "learning_rate": 3.161241214715727e-08, + "loss": 0.9943, + "step": 79204 + }, + { + "epoch": 5.85, + "learning_rate": 3.158072925862787e-08, + "loss": 1.0777, + "step": 79205 + }, + { + "epoch": 5.85, + "learning_rate": 3.1549062229710994e-08, + "loss": 0.9539, + "step": 79206 + }, + { + "epoch": 5.85, + "learning_rate": 3.1517411060456585e-08, + "loss": 0.8849, + "step": 79207 + }, + { + "epoch": 5.85, + "learning_rate": 3.148577575091682e-08, + "loss": 1.1154, + "step": 79208 + }, + { + "epoch": 5.85, + "learning_rate": 3.145415630113946e-08, + "loss": 1.0768, + "step": 79209 + }, + { + "epoch": 5.85, + "learning_rate": 3.142255271117556e-08, + "loss": 0.941, + "step": 79210 + }, + { + "epoch": 5.85, + "learning_rate": 3.139096498107619e-08, + "loss": 1.0658, + "step": 79211 + }, + { + "epoch": 5.85, + "learning_rate": 3.135939311089131e-08, + "loss": 0.9802, + "step": 79212 + }, + { + "epoch": 5.85, + "learning_rate": 3.132783710067089e-08, + "loss": 0.9672, + "step": 79213 + }, + { + "epoch": 5.85, + "learning_rate": 3.1296296950464874e-08, + "loss": 1.0111, + "step": 79214 + }, + { + "epoch": 5.85, + "learning_rate": 3.126477266032435e-08, + "loss": 1.068, + "step": 79215 + }, + { + "epoch": 5.85, + "learning_rate": 3.123326423029816e-08, + "loss": 0.9681, + "step": 79216 + }, + { + "epoch": 5.85, + "learning_rate": 3.120177166043847e-08, + "loss": 1.0637, + "step": 79217 + }, + { + "epoch": 5.85, + "learning_rate": 3.117029495079304e-08, + "loss": 0.9744, + "step": 79218 + }, + { + "epoch": 5.85, + "learning_rate": 3.113883410141294e-08, + "loss": 0.8837, + "step": 79219 + }, + { + "epoch": 5.85, + "learning_rate": 3.110738911234923e-08, + "loss": 1.0568, + "step": 79220 + }, + { + "epoch": 5.85, + "learning_rate": 3.107595998365076e-08, + "loss": 0.9737, + "step": 79221 + }, + { + "epoch": 5.85, + "learning_rate": 3.10445467153675e-08, + "loss": 1.0381, + "step": 79222 + }, + { + "epoch": 5.85, + "learning_rate": 3.10131493075505e-08, + "loss": 0.9601, + "step": 79223 + }, + { + "epoch": 5.85, + "learning_rate": 3.098176776024753e-08, + "loss": 0.9972, + "step": 79224 + }, + { + "epoch": 5.85, + "learning_rate": 3.0950402073510745e-08, + "loss": 1.0808, + "step": 79225 + }, + { + "epoch": 5.85, + "learning_rate": 3.091905224739011e-08, + "loss": 0.992, + "step": 79226 + }, + { + "epoch": 5.85, + "learning_rate": 3.0887718281933376e-08, + "loss": 0.8618, + "step": 79227 + }, + { + "epoch": 5.85, + "learning_rate": 3.08564001771916e-08, + "loss": 0.9182, + "step": 79228 + }, + { + "epoch": 5.85, + "learning_rate": 3.0825097933215864e-08, + "loss": 1.0782, + "step": 79229 + }, + { + "epoch": 5.85, + "learning_rate": 3.07938115500539e-08, + "loss": 0.9404, + "step": 79230 + }, + { + "epoch": 5.85, + "learning_rate": 3.0762541027755666e-08, + "loss": 0.9844, + "step": 79231 + }, + { + "epoch": 5.85, + "learning_rate": 3.073128636637224e-08, + "loss": 0.9084, + "step": 79232 + }, + { + "epoch": 5.85, + "learning_rate": 3.070004756595357e-08, + "loss": 1.0017, + "step": 79233 + }, + { + "epoch": 5.85, + "learning_rate": 3.066882462654741e-08, + "loss": 0.9445, + "step": 79234 + }, + { + "epoch": 5.85, + "learning_rate": 3.0637617548204825e-08, + "loss": 1.0419, + "step": 79235 + }, + { + "epoch": 5.85, + "learning_rate": 3.060642633097577e-08, + "loss": 1.067, + "step": 79236 + }, + { + "epoch": 5.85, + "learning_rate": 3.057525097490799e-08, + "loss": 1.0574, + "step": 79237 + }, + { + "epoch": 5.85, + "learning_rate": 3.0544091480053664e-08, + "loss": 1.0454, + "step": 79238 + }, + { + "epoch": 5.85, + "learning_rate": 3.051294784646053e-08, + "loss": 0.8919, + "step": 79239 + }, + { + "epoch": 5.85, + "learning_rate": 3.048182007417744e-08, + "loss": 0.9005, + "step": 79240 + }, + { + "epoch": 5.86, + "learning_rate": 3.045070816325657e-08, + "loss": 1.0935, + "step": 79241 + }, + { + "epoch": 5.86, + "learning_rate": 3.0419612113745665e-08, + "loss": 1.0869, + "step": 79242 + }, + { + "epoch": 5.86, + "learning_rate": 3.038853192569469e-08, + "loss": 1.0661, + "step": 79243 + }, + { + "epoch": 5.86, + "learning_rate": 3.035746759915248e-08, + "loss": 0.9414, + "step": 79244 + }, + { + "epoch": 5.86, + "learning_rate": 3.032641913416901e-08, + "loss": 0.9602, + "step": 79245 + }, + { + "epoch": 5.86, + "learning_rate": 3.029538653079423e-08, + "loss": 0.907, + "step": 79246 + }, + { + "epoch": 5.86, + "learning_rate": 3.0264369789075877e-08, + "loss": 0.9031, + "step": 79247 + }, + { + "epoch": 5.86, + "learning_rate": 3.023336890906503e-08, + "loss": 0.8942, + "step": 79248 + }, + { + "epoch": 5.86, + "learning_rate": 3.020238389080943e-08, + "loss": 1.0225, + "step": 79249 + }, + { + "epoch": 5.86, + "learning_rate": 3.017141473436014e-08, + "loss": 1.0519, + "step": 79250 + }, + { + "epoch": 5.86, + "learning_rate": 3.0140461439764903e-08, + "loss": 1.0827, + "step": 79251 + }, + { + "epoch": 5.86, + "learning_rate": 3.0109524007073675e-08, + "loss": 1.0692, + "step": 79252 + }, + { + "epoch": 5.86, + "learning_rate": 3.007860243633531e-08, + "loss": 0.9235, + "step": 79253 + }, + { + "epoch": 5.86, + "learning_rate": 3.0047696727599774e-08, + "loss": 1.0093, + "step": 79254 + }, + { + "epoch": 5.86, + "learning_rate": 3.001680688091591e-08, + "loss": 1.0728, + "step": 79255 + }, + { + "epoch": 5.86, + "learning_rate": 2.998593289633145e-08, + "loss": 1.0249, + "step": 79256 + }, + { + "epoch": 5.86, + "learning_rate": 2.995507477389747e-08, + "loss": 0.8959, + "step": 79257 + }, + { + "epoch": 5.86, + "learning_rate": 2.9924232513662833e-08, + "loss": 0.9054, + "step": 79258 + }, + { + "epoch": 5.86, + "learning_rate": 2.989340611567526e-08, + "loss": 0.9675, + "step": 79259 + }, + { + "epoch": 5.86, + "learning_rate": 2.986259557998472e-08, + "loss": 0.8342, + "step": 79260 + }, + { + "epoch": 5.86, + "learning_rate": 2.983180090664006e-08, + "loss": 0.9276, + "step": 79261 + }, + { + "epoch": 5.86, + "learning_rate": 2.9801022095691245e-08, + "loss": 0.933, + "step": 79262 + }, + { + "epoch": 5.86, + "learning_rate": 2.97702591471849e-08, + "loss": 0.9942, + "step": 79263 + }, + { + "epoch": 5.86, + "learning_rate": 2.9739512061173203e-08, + "loss": 1.0414, + "step": 79264 + }, + { + "epoch": 5.86, + "learning_rate": 2.970878083770168e-08, + "loss": 0.9642, + "step": 79265 + }, + { + "epoch": 5.86, + "learning_rate": 2.9678065476821392e-08, + "loss": 0.9832, + "step": 79266 + }, + { + "epoch": 5.86, + "learning_rate": 2.96473659785812e-08, + "loss": 0.9928, + "step": 79267 + }, + { + "epoch": 5.86, + "learning_rate": 2.9616682343028836e-08, + "loss": 0.9259, + "step": 79268 + }, + { + "epoch": 5.86, + "learning_rate": 2.958601457021315e-08, + "loss": 1.0548, + "step": 79269 + }, + { + "epoch": 5.86, + "learning_rate": 2.9555362660185217e-08, + "loss": 0.994, + "step": 79270 + }, + { + "epoch": 5.86, + "learning_rate": 2.952472661299055e-08, + "loss": 0.9295, + "step": 79271 + }, + { + "epoch": 5.86, + "learning_rate": 2.9494106428679113e-08, + "loss": 0.9467, + "step": 79272 + }, + { + "epoch": 5.86, + "learning_rate": 2.9463502107300868e-08, + "loss": 1.0075, + "step": 79273 + }, + { + "epoch": 5.86, + "learning_rate": 2.9432913648903548e-08, + "loss": 1.0069, + "step": 79274 + }, + { + "epoch": 5.86, + "learning_rate": 2.94023410535349e-08, + "loss": 0.9329, + "step": 79275 + }, + { + "epoch": 5.86, + "learning_rate": 2.937178432124488e-08, + "loss": 0.9309, + "step": 79276 + }, + { + "epoch": 5.86, + "learning_rate": 2.9341243452082335e-08, + "loss": 0.9602, + "step": 79277 + }, + { + "epoch": 5.86, + "learning_rate": 2.931071844609501e-08, + "loss": 1.0155, + "step": 79278 + }, + { + "epoch": 5.86, + "learning_rate": 2.928020930333175e-08, + "loss": 0.9819, + "step": 79279 + }, + { + "epoch": 5.86, + "learning_rate": 2.92497160238403e-08, + "loss": 1.0214, + "step": 79280 + }, + { + "epoch": 5.86, + "learning_rate": 2.9219238607669508e-08, + "loss": 0.98, + "step": 79281 + }, + { + "epoch": 5.86, + "learning_rate": 2.9188777054869334e-08, + "loss": 1.0751, + "step": 79282 + }, + { + "epoch": 5.86, + "learning_rate": 2.9158331365487512e-08, + "loss": 0.9713, + "step": 79283 + }, + { + "epoch": 5.86, + "learning_rate": 2.912790153957068e-08, + "loss": 0.9267, + "step": 79284 + }, + { + "epoch": 5.86, + "learning_rate": 2.9097487577169903e-08, + "loss": 0.8784, + "step": 79285 + }, + { + "epoch": 5.86, + "learning_rate": 2.906708947833181e-08, + "loss": 1.097, + "step": 79286 + }, + { + "epoch": 5.86, + "learning_rate": 2.9036707243105255e-08, + "loss": 0.9001, + "step": 79287 + }, + { + "epoch": 5.86, + "learning_rate": 2.9006340871539084e-08, + "loss": 1.0419, + "step": 79288 + }, + { + "epoch": 5.86, + "learning_rate": 2.8975990363681038e-08, + "loss": 0.9847, + "step": 79289 + }, + { + "epoch": 5.86, + "learning_rate": 2.8945655719578857e-08, + "loss": 0.904, + "step": 79290 + }, + { + "epoch": 5.86, + "learning_rate": 2.89153369392825e-08, + "loss": 0.9198, + "step": 79291 + }, + { + "epoch": 5.86, + "learning_rate": 2.8885034022838597e-08, + "loss": 1.0406, + "step": 79292 + }, + { + "epoch": 5.86, + "learning_rate": 2.8854746970296e-08, + "loss": 0.94, + "step": 79293 + }, + { + "epoch": 5.86, + "learning_rate": 2.8824475781702444e-08, + "loss": 0.9241, + "step": 79294 + }, + { + "epoch": 5.86, + "learning_rate": 2.879422045710789e-08, + "loss": 0.8975, + "step": 79295 + }, + { + "epoch": 5.86, + "learning_rate": 2.8763980996557862e-08, + "loss": 0.9735, + "step": 79296 + }, + { + "epoch": 5.86, + "learning_rate": 2.8733757400101204e-08, + "loss": 0.9089, + "step": 79297 + }, + { + "epoch": 5.86, + "learning_rate": 2.8703549667787877e-08, + "loss": 1.0501, + "step": 79298 + }, + { + "epoch": 5.86, + "learning_rate": 2.8673357799663403e-08, + "loss": 1.045, + "step": 79299 + }, + { + "epoch": 5.86, + "learning_rate": 2.864318179577774e-08, + "loss": 1.0058, + "step": 79300 + }, + { + "epoch": 5.86, + "learning_rate": 2.861302165617752e-08, + "loss": 0.9839, + "step": 79301 + }, + { + "epoch": 5.86, + "learning_rate": 2.858287738091159e-08, + "loss": 0.979, + "step": 79302 + }, + { + "epoch": 5.86, + "learning_rate": 2.855274897002769e-08, + "loss": 1.086, + "step": 79303 + }, + { + "epoch": 5.86, + "learning_rate": 2.852263642357356e-08, + "loss": 1.0139, + "step": 79304 + }, + { + "epoch": 5.86, + "learning_rate": 2.8492539741596937e-08, + "loss": 1.0266, + "step": 79305 + }, + { + "epoch": 5.86, + "learning_rate": 2.8462458924146675e-08, + "loss": 1.0133, + "step": 79306 + }, + { + "epoch": 5.86, + "learning_rate": 2.84323939712694e-08, + "loss": 0.9655, + "step": 79307 + }, + { + "epoch": 5.86, + "learning_rate": 2.8402344883013966e-08, + "loss": 0.9838, + "step": 79308 + }, + { + "epoch": 5.86, + "learning_rate": 2.8372311659427e-08, + "loss": 1.0384, + "step": 79309 + }, + { + "epoch": 5.86, + "learning_rate": 2.834229430055735e-08, + "loss": 1.0764, + "step": 79310 + }, + { + "epoch": 5.86, + "learning_rate": 2.8312292806453868e-08, + "loss": 0.9659, + "step": 79311 + }, + { + "epoch": 5.86, + "learning_rate": 2.828230717716096e-08, + "loss": 0.8791, + "step": 79312 + }, + { + "epoch": 5.86, + "learning_rate": 2.825233741272859e-08, + "loss": 1.0385, + "step": 79313 + }, + { + "epoch": 5.86, + "learning_rate": 2.82223835132045e-08, + "loss": 1.0149, + "step": 79314 + }, + { + "epoch": 5.86, + "learning_rate": 2.819244547863642e-08, + "loss": 0.9433, + "step": 79315 + }, + { + "epoch": 5.86, + "learning_rate": 2.8162523309069877e-08, + "loss": 1.0724, + "step": 79316 + }, + { + "epoch": 5.86, + "learning_rate": 2.813261700455594e-08, + "loss": 0.9957, + "step": 79317 + }, + { + "epoch": 5.86, + "learning_rate": 2.8102726565139015e-08, + "loss": 0.9058, + "step": 79318 + }, + { + "epoch": 5.86, + "learning_rate": 2.8072851990869066e-08, + "loss": 1.0087, + "step": 79319 + }, + { + "epoch": 5.86, + "learning_rate": 2.804299328179161e-08, + "loss": 1.0147, + "step": 79320 + }, + { + "epoch": 5.86, + "learning_rate": 2.8013150437955493e-08, + "loss": 0.8761, + "step": 79321 + }, + { + "epoch": 5.86, + "learning_rate": 2.798332345940735e-08, + "loss": 1.0731, + "step": 79322 + }, + { + "epoch": 5.86, + "learning_rate": 2.7953512346196034e-08, + "loss": 0.9929, + "step": 79323 + }, + { + "epoch": 5.86, + "learning_rate": 2.7923717098367053e-08, + "loss": 0.9672, + "step": 79324 + }, + { + "epoch": 5.86, + "learning_rate": 2.7893937715968156e-08, + "loss": 0.8754, + "step": 79325 + }, + { + "epoch": 5.86, + "learning_rate": 2.7864174199049298e-08, + "loss": 1.1229, + "step": 79326 + }, + { + "epoch": 5.86, + "learning_rate": 2.7834426547654892e-08, + "loss": 0.8639, + "step": 79327 + }, + { + "epoch": 5.86, + "learning_rate": 2.7804694761832674e-08, + "loss": 1.0391, + "step": 79328 + }, + { + "epoch": 5.86, + "learning_rate": 2.7774978841631496e-08, + "loss": 1.0717, + "step": 79329 + }, + { + "epoch": 5.86, + "learning_rate": 2.7745278787096873e-08, + "loss": 0.9895, + "step": 79330 + }, + { + "epoch": 5.86, + "learning_rate": 2.771559459827655e-08, + "loss": 1.0884, + "step": 79331 + }, + { + "epoch": 5.86, + "learning_rate": 2.7685926275218266e-08, + "loss": 1.0487, + "step": 79332 + }, + { + "epoch": 5.86, + "learning_rate": 2.765627381796976e-08, + "loss": 0.9272, + "step": 79333 + }, + { + "epoch": 5.86, + "learning_rate": 2.7626637226576545e-08, + "loss": 1.0408, + "step": 79334 + }, + { + "epoch": 5.86, + "learning_rate": 2.759701650108748e-08, + "loss": 1.0279, + "step": 79335 + }, + { + "epoch": 5.86, + "learning_rate": 2.756741164154808e-08, + "loss": 0.9533, + "step": 79336 + }, + { + "epoch": 5.86, + "learning_rate": 2.7537822648007195e-08, + "loss": 1.0544, + "step": 79337 + }, + { + "epoch": 5.86, + "learning_rate": 2.7508249520510343e-08, + "loss": 0.8957, + "step": 79338 + }, + { + "epoch": 5.86, + "learning_rate": 2.7478692259105266e-08, + "loss": 0.9711, + "step": 79339 + }, + { + "epoch": 5.86, + "learning_rate": 2.7449150863838592e-08, + "loss": 1.0083, + "step": 79340 + }, + { + "epoch": 5.86, + "learning_rate": 2.741962533475806e-08, + "loss": 1.0343, + "step": 79341 + }, + { + "epoch": 5.86, + "learning_rate": 2.73901156719103e-08, + "loss": 0.9642, + "step": 79342 + }, + { + "epoch": 5.86, + "learning_rate": 2.736062187534194e-08, + "loss": 0.9422, + "step": 79343 + }, + { + "epoch": 5.86, + "learning_rate": 2.7331143945100725e-08, + "loss": 1.1165, + "step": 79344 + }, + { + "epoch": 5.86, + "learning_rate": 2.7301681881232167e-08, + "loss": 1.0284, + "step": 79345 + }, + { + "epoch": 5.86, + "learning_rate": 2.727223568378401e-08, + "loss": 1.0396, + "step": 79346 + }, + { + "epoch": 5.86, + "learning_rate": 2.7242805352802883e-08, + "loss": 0.9721, + "step": 79347 + }, + { + "epoch": 5.86, + "learning_rate": 2.7213390888336523e-08, + "loss": 0.8694, + "step": 79348 + }, + { + "epoch": 5.86, + "learning_rate": 2.718399229043045e-08, + "loss": 1.0231, + "step": 79349 + }, + { + "epoch": 5.86, + "learning_rate": 2.7154609559132406e-08, + "loss": 1.0067, + "step": 79350 + }, + { + "epoch": 5.86, + "learning_rate": 2.712524269448791e-08, + "loss": 0.9838, + "step": 79351 + }, + { + "epoch": 5.86, + "learning_rate": 2.7095891696544695e-08, + "loss": 0.9436, + "step": 79352 + }, + { + "epoch": 5.86, + "learning_rate": 2.70665565653494e-08, + "loss": 1.0026, + "step": 79353 + }, + { + "epoch": 5.86, + "learning_rate": 2.7037237300948647e-08, + "loss": 0.9668, + "step": 79354 + }, + { + "epoch": 5.86, + "learning_rate": 2.700793390338907e-08, + "loss": 0.9558, + "step": 79355 + }, + { + "epoch": 5.86, + "learning_rate": 2.69786463727173e-08, + "loss": 1.1099, + "step": 79356 + }, + { + "epoch": 5.86, + "learning_rate": 2.694937470897996e-08, + "loss": 0.9239, + "step": 79357 + }, + { + "epoch": 5.86, + "learning_rate": 2.692011891222257e-08, + "loss": 0.9866, + "step": 79358 + }, + { + "epoch": 5.86, + "learning_rate": 2.6890878982492875e-08, + "loss": 1.0451, + "step": 79359 + }, + { + "epoch": 5.86, + "learning_rate": 2.68616549198375e-08, + "loss": 1.0761, + "step": 79360 + }, + { + "epoch": 5.86, + "learning_rate": 2.6832446724301964e-08, + "loss": 1.0329, + "step": 79361 + }, + { + "epoch": 5.86, + "learning_rate": 2.6803254395934007e-08, + "loss": 1.0461, + "step": 79362 + }, + { + "epoch": 5.86, + "learning_rate": 2.6774077934779152e-08, + "loss": 1.0782, + "step": 79363 + }, + { + "epoch": 5.86, + "learning_rate": 2.6744917340884025e-08, + "loss": 1.0017, + "step": 79364 + }, + { + "epoch": 5.86, + "learning_rate": 2.6715772614294144e-08, + "loss": 1.0418, + "step": 79365 + }, + { + "epoch": 5.86, + "learning_rate": 2.6686643755058362e-08, + "loss": 0.9093, + "step": 79366 + }, + { + "epoch": 5.86, + "learning_rate": 2.6657530763219975e-08, + "loss": 1.05, + "step": 79367 + }, + { + "epoch": 5.86, + "learning_rate": 2.6628433638827833e-08, + "loss": 0.8975, + "step": 79368 + }, + { + "epoch": 5.86, + "learning_rate": 2.6599352381926347e-08, + "loss": 0.9699, + "step": 79369 + }, + { + "epoch": 5.86, + "learning_rate": 2.6570286992563254e-08, + "loss": 1.0896, + "step": 79370 + }, + { + "epoch": 5.86, + "learning_rate": 2.6541237470784077e-08, + "loss": 1.0424, + "step": 79371 + }, + { + "epoch": 5.86, + "learning_rate": 2.651220381663433e-08, + "loss": 1.0653, + "step": 79372 + }, + { + "epoch": 5.86, + "learning_rate": 2.6483186030161757e-08, + "loss": 0.9853, + "step": 79373 + }, + { + "epoch": 5.86, + "learning_rate": 2.6454184111411874e-08, + "loss": 1.0054, + "step": 79374 + }, + { + "epoch": 5.86, + "learning_rate": 2.6425198060430203e-08, + "loss": 1.0263, + "step": 79375 + }, + { + "epoch": 5.87, + "learning_rate": 2.639622787726337e-08, + "loss": 1.109, + "step": 79376 + }, + { + "epoch": 5.87, + "learning_rate": 2.6367273561958008e-08, + "loss": 1.0527, + "step": 79377 + }, + { + "epoch": 5.87, + "learning_rate": 2.633833511455852e-08, + "loss": 1.047, + "step": 79378 + }, + { + "epoch": 5.87, + "learning_rate": 2.6309412535112654e-08, + "loss": 0.967, + "step": 79379 + }, + { + "epoch": 5.87, + "learning_rate": 2.6280505823665925e-08, + "loss": 0.9111, + "step": 79380 + }, + { + "epoch": 5.87, + "learning_rate": 2.625161498026385e-08, + "loss": 0.9877, + "step": 79381 + }, + { + "epoch": 5.87, + "learning_rate": 2.622274000495306e-08, + "loss": 0.9656, + "step": 79382 + }, + { + "epoch": 5.87, + "learning_rate": 2.619388089777908e-08, + "loss": 0.9207, + "step": 79383 + }, + { + "epoch": 5.87, + "learning_rate": 2.616503765878853e-08, + "loss": 0.963, + "step": 79384 + }, + { + "epoch": 5.87, + "learning_rate": 2.6136210288025822e-08, + "loss": 1.018, + "step": 79385 + }, + { + "epoch": 5.87, + "learning_rate": 2.6107398785538694e-08, + "loss": 0.9767, + "step": 79386 + }, + { + "epoch": 5.87, + "learning_rate": 2.6078603151370453e-08, + "loss": 0.9097, + "step": 79387 + }, + { + "epoch": 5.87, + "learning_rate": 2.604982338556994e-08, + "loss": 1.0198, + "step": 79388 + }, + { + "epoch": 5.87, + "learning_rate": 2.6021059488180455e-08, + "loss": 1.0204, + "step": 79389 + }, + { + "epoch": 5.87, + "learning_rate": 2.5992311459249742e-08, + "loss": 1.0266, + "step": 79390 + }, + { + "epoch": 5.87, + "learning_rate": 2.5963579298822207e-08, + "loss": 1.0575, + "step": 79391 + }, + { + "epoch": 5.87, + "learning_rate": 2.5934863006943367e-08, + "loss": 0.9915, + "step": 79392 + }, + { + "epoch": 5.87, + "learning_rate": 2.5906162583659854e-08, + "loss": 0.9086, + "step": 79393 + }, + { + "epoch": 5.87, + "learning_rate": 2.587747802901719e-08, + "loss": 1.0095, + "step": 79394 + }, + { + "epoch": 5.87, + "learning_rate": 2.5848809343059776e-08, + "loss": 1.0529, + "step": 79395 + }, + { + "epoch": 5.87, + "learning_rate": 2.582015652583536e-08, + "loss": 0.9918, + "step": 79396 + }, + { + "epoch": 5.87, + "learning_rate": 2.5791519577387237e-08, + "loss": 1.041, + "step": 79397 + }, + { + "epoch": 5.87, + "learning_rate": 2.5762898497763144e-08, + "loss": 1.0193, + "step": 79398 + }, + { + "epoch": 5.87, + "learning_rate": 2.5734293287006385e-08, + "loss": 0.9592, + "step": 79399 + }, + { + "epoch": 5.87, + "learning_rate": 2.5705703945164695e-08, + "loss": 1.0345, + "step": 79400 + }, + { + "epoch": 5.87, + "learning_rate": 2.5677130472282487e-08, + "loss": 0.9541, + "step": 79401 + }, + { + "epoch": 5.87, + "learning_rate": 2.5648572868405275e-08, + "loss": 0.9794, + "step": 79402 + }, + { + "epoch": 5.87, + "learning_rate": 2.5620031133577473e-08, + "loss": 0.9712, + "step": 79403 + }, + { + "epoch": 5.87, + "learning_rate": 2.5591505267846816e-08, + "loss": 0.9211, + "step": 79404 + }, + { + "epoch": 5.87, + "learning_rate": 2.5562995271256607e-08, + "loss": 1.1109, + "step": 79405 + }, + { + "epoch": 5.87, + "learning_rate": 2.5534501143853475e-08, + "loss": 0.9469, + "step": 79406 + }, + { + "epoch": 5.87, + "learning_rate": 2.5506022885681824e-08, + "loss": 0.9853, + "step": 79407 + }, + { + "epoch": 5.87, + "learning_rate": 2.547756049678718e-08, + "loss": 1.0287, + "step": 79408 + }, + { + "epoch": 5.87, + "learning_rate": 2.5449113977215057e-08, + "loss": 0.9991, + "step": 79409 + }, + { + "epoch": 5.87, + "learning_rate": 2.5420683327010974e-08, + "loss": 1.0077, + "step": 79410 + }, + { + "epoch": 5.87, + "learning_rate": 2.5392268546219347e-08, + "loss": 1.0971, + "step": 79411 + }, + { + "epoch": 5.87, + "learning_rate": 2.5363869634886795e-08, + "loss": 0.9896, + "step": 79412 + }, + { + "epoch": 5.87, + "learning_rate": 2.5335486593057736e-08, + "loss": 0.9545, + "step": 79413 + }, + { + "epoch": 5.87, + "learning_rate": 2.530711942077657e-08, + "loss": 0.9866, + "step": 79414 + }, + { + "epoch": 5.87, + "learning_rate": 2.5278768118088827e-08, + "loss": 0.9691, + "step": 79415 + }, + { + "epoch": 5.87, + "learning_rate": 2.5250432685040017e-08, + "loss": 0.974, + "step": 79416 + }, + { + "epoch": 5.87, + "learning_rate": 2.5222113121675663e-08, + "loss": 1.0725, + "step": 79417 + }, + { + "epoch": 5.87, + "learning_rate": 2.5193809428039063e-08, + "loss": 1.116, + "step": 79418 + }, + { + "epoch": 5.87, + "learning_rate": 2.5165521604176847e-08, + "loss": 1.084, + "step": 79419 + }, + { + "epoch": 5.87, + "learning_rate": 2.5137249650134532e-08, + "loss": 1.0195, + "step": 79420 + }, + { + "epoch": 5.87, + "learning_rate": 2.510899356595542e-08, + "loss": 0.969, + "step": 79421 + }, + { + "epoch": 5.87, + "learning_rate": 2.5080753351685028e-08, + "loss": 0.9946, + "step": 79422 + }, + { + "epoch": 5.87, + "learning_rate": 2.5052529007368874e-08, + "loss": 1.0038, + "step": 79423 + }, + { + "epoch": 5.87, + "learning_rate": 2.502432053305026e-08, + "loss": 0.9682, + "step": 79424 + }, + { + "epoch": 5.87, + "learning_rate": 2.4996127928776926e-08, + "loss": 0.9217, + "step": 79425 + }, + { + "epoch": 5.87, + "learning_rate": 2.4967951194591056e-08, + "loss": 1.0573, + "step": 79426 + }, + { + "epoch": 5.87, + "learning_rate": 2.4939790330538173e-08, + "loss": 1.052, + "step": 79427 + }, + { + "epoch": 5.87, + "learning_rate": 2.491164533666379e-08, + "loss": 0.9751, + "step": 79428 + }, + { + "epoch": 5.87, + "learning_rate": 2.4883516213013438e-08, + "loss": 1.0655, + "step": 79429 + }, + { + "epoch": 5.87, + "learning_rate": 2.4855402959629294e-08, + "loss": 1.0047, + "step": 79430 + }, + { + "epoch": 5.87, + "learning_rate": 2.482730557655688e-08, + "loss": 0.9329, + "step": 79431 + }, + { + "epoch": 5.87, + "learning_rate": 2.479922406384283e-08, + "loss": 1.0555, + "step": 79432 + }, + { + "epoch": 5.87, + "learning_rate": 2.4771158421530437e-08, + "loss": 0.9974, + "step": 79433 + }, + { + "epoch": 5.87, + "learning_rate": 2.4743108649664116e-08, + "loss": 0.9766, + "step": 79434 + }, + { + "epoch": 5.87, + "learning_rate": 2.471507474828827e-08, + "loss": 1.0726, + "step": 79435 + }, + { + "epoch": 5.87, + "learning_rate": 2.468705671744842e-08, + "loss": 1.0128, + "step": 79436 + }, + { + "epoch": 5.87, + "learning_rate": 2.4659054557188976e-08, + "loss": 0.9752, + "step": 79437 + }, + { + "epoch": 5.87, + "learning_rate": 2.4631068267553236e-08, + "loss": 0.9619, + "step": 79438 + }, + { + "epoch": 5.87, + "learning_rate": 2.460309784858783e-08, + "loss": 1.1069, + "step": 79439 + }, + { + "epoch": 5.87, + "learning_rate": 2.4575143300336058e-08, + "loss": 1.0196, + "step": 79440 + }, + { + "epoch": 5.87, + "learning_rate": 2.4547204622842325e-08, + "loss": 1.0814, + "step": 79441 + }, + { + "epoch": 5.87, + "learning_rate": 2.4519281816151042e-08, + "loss": 0.9218, + "step": 79442 + }, + { + "epoch": 5.87, + "learning_rate": 2.449137488030662e-08, + "loss": 0.9738, + "step": 79443 + }, + { + "epoch": 5.87, + "learning_rate": 2.446348381535457e-08, + "loss": 1.0083, + "step": 79444 + }, + { + "epoch": 5.87, + "learning_rate": 2.44356086213382e-08, + "loss": 0.9777, + "step": 79445 + }, + { + "epoch": 5.87, + "learning_rate": 2.440774929830081e-08, + "loss": 0.9046, + "step": 79446 + }, + { + "epoch": 5.87, + "learning_rate": 2.437990584628902e-08, + "loss": 1.0816, + "step": 79447 + }, + { + "epoch": 5.87, + "learning_rate": 2.4352078265346134e-08, + "loss": 0.9313, + "step": 79448 + }, + { + "epoch": 5.87, + "learning_rate": 2.432426655551656e-08, + "loss": 1.0553, + "step": 79449 + }, + { + "epoch": 5.87, + "learning_rate": 2.4296470716843602e-08, + "loss": 0.9625, + "step": 79450 + }, + { + "epoch": 5.87, + "learning_rate": 2.426869074937277e-08, + "loss": 1.0784, + "step": 79451 + }, + { + "epoch": 5.87, + "learning_rate": 2.4240926653148477e-08, + "loss": 1.0216, + "step": 79452 + }, + { + "epoch": 5.87, + "learning_rate": 2.4213178428212914e-08, + "loss": 0.9784, + "step": 79453 + }, + { + "epoch": 5.87, + "learning_rate": 2.4185446074611596e-08, + "loss": 0.9906, + "step": 79454 + }, + { + "epoch": 5.87, + "learning_rate": 2.4157729592388936e-08, + "loss": 0.9527, + "step": 79455 + }, + { + "epoch": 5.87, + "learning_rate": 2.4130028981588228e-08, + "loss": 0.9857, + "step": 79456 + }, + { + "epoch": 5.87, + "learning_rate": 2.4102344242253885e-08, + "loss": 0.9595, + "step": 79457 + }, + { + "epoch": 5.87, + "learning_rate": 2.4074675374430313e-08, + "loss": 1.0562, + "step": 79458 + }, + { + "epoch": 5.87, + "learning_rate": 2.404702237816081e-08, + "loss": 1.0664, + "step": 79459 + }, + { + "epoch": 5.87, + "learning_rate": 2.4019385253489792e-08, + "loss": 0.9848, + "step": 79460 + }, + { + "epoch": 5.87, + "learning_rate": 2.3991764000461658e-08, + "loss": 0.9854, + "step": 79461 + }, + { + "epoch": 5.87, + "learning_rate": 2.3964158619119714e-08, + "loss": 0.9475, + "step": 79462 + }, + { + "epoch": 5.87, + "learning_rate": 2.3936569109508366e-08, + "loss": 1.0852, + "step": 79463 + }, + { + "epoch": 5.87, + "learning_rate": 2.3908995471669805e-08, + "loss": 0.9536, + "step": 79464 + }, + { + "epoch": 5.87, + "learning_rate": 2.3881437705650656e-08, + "loss": 0.9688, + "step": 79465 + }, + { + "epoch": 5.87, + "learning_rate": 2.3853895811493112e-08, + "loss": 0.944, + "step": 79466 + }, + { + "epoch": 5.87, + "learning_rate": 2.3826369789241576e-08, + "loss": 0.9876, + "step": 79467 + }, + { + "epoch": 5.87, + "learning_rate": 2.3798859638939354e-08, + "loss": 0.9723, + "step": 79468 + }, + { + "epoch": 5.87, + "learning_rate": 2.377136536062974e-08, + "loss": 0.9794, + "step": 79469 + }, + { + "epoch": 5.87, + "learning_rate": 2.3743886954358253e-08, + "loss": 0.9837, + "step": 79470 + }, + { + "epoch": 5.87, + "learning_rate": 2.3716424420167082e-08, + "loss": 0.9814, + "step": 79471 + }, + { + "epoch": 5.87, + "learning_rate": 2.368897775809953e-08, + "loss": 1.1087, + "step": 79472 + }, + { + "epoch": 5.87, + "learning_rate": 2.366154696820111e-08, + "loss": 1.003, + "step": 79473 + }, + { + "epoch": 5.87, + "learning_rate": 2.3634132050514015e-08, + "loss": 0.9311, + "step": 79474 + }, + { + "epoch": 5.87, + "learning_rate": 2.3606733005081538e-08, + "loss": 1.059, + "step": 79475 + }, + { + "epoch": 5.87, + "learning_rate": 2.3579349831949204e-08, + "loss": 0.8023, + "step": 79476 + }, + { + "epoch": 5.87, + "learning_rate": 2.35519825311592e-08, + "loss": 1.0286, + "step": 79477 + }, + { + "epoch": 5.87, + "learning_rate": 2.3524631102754826e-08, + "loss": 1.0263, + "step": 79478 + }, + { + "epoch": 5.87, + "learning_rate": 2.3497295546780487e-08, + "loss": 0.99, + "step": 79479 + }, + { + "epoch": 5.87, + "learning_rate": 2.346997586327837e-08, + "loss": 0.9892, + "step": 79480 + }, + { + "epoch": 5.87, + "learning_rate": 2.3442672052294003e-08, + "loss": 1.0537, + "step": 79481 + }, + { + "epoch": 5.87, + "learning_rate": 2.3415384113868457e-08, + "loss": 1.0284, + "step": 79482 + }, + { + "epoch": 5.87, + "learning_rate": 2.3388112048047252e-08, + "loss": 0.9092, + "step": 79483 + }, + { + "epoch": 5.87, + "learning_rate": 2.3360855854871466e-08, + "loss": 0.9398, + "step": 79484 + }, + { + "epoch": 5.87, + "learning_rate": 2.333361553438773e-08, + "loss": 1.0959, + "step": 79485 + }, + { + "epoch": 5.87, + "learning_rate": 2.3306391086636015e-08, + "loss": 1.0011, + "step": 79486 + }, + { + "epoch": 5.87, + "learning_rate": 2.3279182511661837e-08, + "loss": 1.0203, + "step": 79487 + }, + { + "epoch": 5.87, + "learning_rate": 2.325198980950738e-08, + "loss": 0.8794, + "step": 79488 + }, + { + "epoch": 5.87, + "learning_rate": 2.322481298021595e-08, + "loss": 0.9735, + "step": 79489 + }, + { + "epoch": 5.87, + "learning_rate": 2.3197652023831952e-08, + "loss": 0.9768, + "step": 79490 + }, + { + "epoch": 5.87, + "learning_rate": 2.3170506940396463e-08, + "loss": 0.9189, + "step": 79491 + }, + { + "epoch": 5.87, + "learning_rate": 2.3143377729955006e-08, + "loss": 0.8939, + "step": 79492 + }, + { + "epoch": 5.87, + "learning_rate": 2.311626439254977e-08, + "loss": 0.94, + "step": 79493 + }, + { + "epoch": 5.87, + "learning_rate": 2.308916692822294e-08, + "loss": 0.9517, + "step": 79494 + }, + { + "epoch": 5.87, + "learning_rate": 2.3062085337018924e-08, + "loss": 1.0117, + "step": 79495 + }, + { + "epoch": 5.87, + "learning_rate": 2.3035019618981026e-08, + "loss": 1.0694, + "step": 79496 + }, + { + "epoch": 5.87, + "learning_rate": 2.300796977415143e-08, + "loss": 1.184, + "step": 79497 + }, + { + "epoch": 5.87, + "learning_rate": 2.2980935802572325e-08, + "loss": 0.9881, + "step": 79498 + }, + { + "epoch": 5.87, + "learning_rate": 2.295391770428923e-08, + "loss": 0.985, + "step": 79499 + }, + { + "epoch": 5.87, + "learning_rate": 2.2926915479342116e-08, + "loss": 1.1173, + "step": 79500 + }, + { + "epoch": 5.87, + "learning_rate": 2.28999291277765e-08, + "loss": 1.0061, + "step": 79501 + }, + { + "epoch": 5.87, + "learning_rate": 2.2872958649634568e-08, + "loss": 1.055, + "step": 79502 + }, + { + "epoch": 5.87, + "learning_rate": 2.2846004044958513e-08, + "loss": 0.8972, + "step": 79503 + }, + { + "epoch": 5.87, + "learning_rate": 2.2819065313792744e-08, + "loss": 0.983, + "step": 79504 + }, + { + "epoch": 5.87, + "learning_rate": 2.2792142456178334e-08, + "loss": 0.9893, + "step": 79505 + }, + { + "epoch": 5.87, + "learning_rate": 2.2765235472158586e-08, + "loss": 0.9796, + "step": 79506 + }, + { + "epoch": 5.87, + "learning_rate": 2.27383443617768e-08, + "loss": 0.8884, + "step": 79507 + }, + { + "epoch": 5.87, + "learning_rate": 2.271146912507516e-08, + "loss": 0.9975, + "step": 79508 + }, + { + "epoch": 5.87, + "learning_rate": 2.2684609762096965e-08, + "loss": 1.0019, + "step": 79509 + }, + { + "epoch": 5.87, + "learning_rate": 2.2657766272885516e-08, + "loss": 0.9283, + "step": 79510 + }, + { + "epoch": 5.88, + "learning_rate": 2.2630938657483004e-08, + "loss": 0.925, + "step": 79511 + }, + { + "epoch": 5.88, + "learning_rate": 2.2604126915930504e-08, + "loss": 0.9844, + "step": 79512 + }, + { + "epoch": 5.88, + "learning_rate": 2.2577331048273533e-08, + "loss": 0.972, + "step": 79513 + }, + { + "epoch": 5.88, + "learning_rate": 2.2550551054552062e-08, + "loss": 1.0884, + "step": 79514 + }, + { + "epoch": 5.88, + "learning_rate": 2.25237869348105e-08, + "loss": 1.0216, + "step": 79515 + }, + { + "epoch": 5.88, + "learning_rate": 2.2497038689091033e-08, + "loss": 0.9517, + "step": 79516 + }, + { + "epoch": 5.88, + "learning_rate": 2.247030631743585e-08, + "loss": 1.002, + "step": 79517 + }, + { + "epoch": 5.88, + "learning_rate": 2.2443589819887146e-08, + "loss": 1.1132, + "step": 79518 + }, + { + "epoch": 5.88, + "learning_rate": 2.2416889196489322e-08, + "loss": 1.0803, + "step": 79519 + }, + { + "epoch": 5.88, + "learning_rate": 2.2390204447282348e-08, + "loss": 1.0902, + "step": 79520 + }, + { + "epoch": 5.88, + "learning_rate": 2.2363535572309525e-08, + "loss": 0.901, + "step": 79521 + }, + { + "epoch": 5.88, + "learning_rate": 2.2336882571614148e-08, + "loss": 1.0285, + "step": 79522 + }, + { + "epoch": 5.88, + "learning_rate": 2.231024544523841e-08, + "loss": 0.9602, + "step": 79523 + }, + { + "epoch": 5.88, + "learning_rate": 2.2283624193223385e-08, + "loss": 0.9332, + "step": 79524 + }, + { + "epoch": 5.88, + "learning_rate": 2.2257018815612374e-08, + "loss": 0.9677, + "step": 79525 + }, + { + "epoch": 5.88, + "learning_rate": 2.2230429312448675e-08, + "loss": 0.9549, + "step": 79526 + }, + { + "epoch": 5.88, + "learning_rate": 2.2203855683773367e-08, + "loss": 0.9131, + "step": 79527 + }, + { + "epoch": 5.88, + "learning_rate": 2.2177297929628638e-08, + "loss": 0.9391, + "step": 79528 + }, + { + "epoch": 5.88, + "learning_rate": 2.2150756050056675e-08, + "loss": 1.0469, + "step": 79529 + }, + { + "epoch": 5.88, + "learning_rate": 2.212423004510078e-08, + "loss": 0.9547, + "step": 79530 + }, + { + "epoch": 5.88, + "learning_rate": 2.209771991480203e-08, + "loss": 0.9544, + "step": 79531 + }, + { + "epoch": 5.88, + "learning_rate": 2.207122565920372e-08, + "loss": 1.0191, + "step": 79532 + }, + { + "epoch": 5.88, + "learning_rate": 2.2044747278345825e-08, + "loss": 1.0537, + "step": 79533 + }, + { + "epoch": 5.88, + "learning_rate": 2.2018284772273856e-08, + "loss": 0.9425, + "step": 79534 + }, + { + "epoch": 5.88, + "learning_rate": 2.199183814102668e-08, + "loss": 1.0463, + "step": 79535 + }, + { + "epoch": 5.88, + "learning_rate": 2.1965407384648695e-08, + "loss": 0.8478, + "step": 79536 + }, + { + "epoch": 5.88, + "learning_rate": 2.193899250317988e-08, + "loss": 1.0592, + "step": 79537 + }, + { + "epoch": 5.88, + "learning_rate": 2.1912593496663524e-08, + "loss": 1.0275, + "step": 79538 + }, + { + "epoch": 5.88, + "learning_rate": 2.1886210365142936e-08, + "loss": 1.0726, + "step": 79539 + }, + { + "epoch": 5.88, + "learning_rate": 2.1859843108656964e-08, + "loss": 1.1071, + "step": 79540 + }, + { + "epoch": 5.88, + "learning_rate": 2.1833491727250023e-08, + "loss": 1.0333, + "step": 79541 + }, + { + "epoch": 5.88, + "learning_rate": 2.1807156220963187e-08, + "loss": 1.1847, + "step": 79542 + }, + { + "epoch": 5.88, + "learning_rate": 2.178083658983865e-08, + "loss": 1.0617, + "step": 79543 + }, + { + "epoch": 5.88, + "learning_rate": 2.1754532833917484e-08, + "loss": 0.9712, + "step": 79544 + }, + { + "epoch": 5.88, + "learning_rate": 2.1728244953241885e-08, + "loss": 0.9628, + "step": 79545 + }, + { + "epoch": 5.88, + "learning_rate": 2.1701972947854034e-08, + "loss": 1.0894, + "step": 79546 + }, + { + "epoch": 5.88, + "learning_rate": 2.1675716817796123e-08, + "loss": 0.9583, + "step": 79547 + }, + { + "epoch": 5.88, + "learning_rate": 2.1649476563109228e-08, + "loss": 0.9644, + "step": 79548 + }, + { + "epoch": 5.88, + "learning_rate": 2.162325218383443e-08, + "loss": 0.9947, + "step": 79549 + }, + { + "epoch": 5.88, + "learning_rate": 2.159704368001503e-08, + "loss": 0.9156, + "step": 79550 + }, + { + "epoch": 5.88, + "learning_rate": 2.15708510516921e-08, + "loss": 0.8522, + "step": 79551 + }, + { + "epoch": 5.88, + "learning_rate": 2.154467429890783e-08, + "loss": 1.0079, + "step": 79552 + }, + { + "epoch": 5.88, + "learning_rate": 2.151851342170219e-08, + "loss": 0.929, + "step": 79553 + }, + { + "epoch": 5.88, + "learning_rate": 2.149236842011737e-08, + "loss": 0.9884, + "step": 79554 + }, + { + "epoch": 5.88, + "learning_rate": 2.1466239294196667e-08, + "loss": 0.9759, + "step": 79555 + }, + { + "epoch": 5.88, + "learning_rate": 2.1440126043978936e-08, + "loss": 1.0268, + "step": 79556 + }, + { + "epoch": 5.88, + "learning_rate": 2.141402866950859e-08, + "loss": 1.0317, + "step": 79557 + }, + { + "epoch": 5.88, + "learning_rate": 2.1387947170824485e-08, + "loss": 0.9121, + "step": 79558 + }, + { + "epoch": 5.88, + "learning_rate": 2.136188154797103e-08, + "loss": 0.8908, + "step": 79559 + }, + { + "epoch": 5.88, + "learning_rate": 2.1335831800987084e-08, + "loss": 1.0172, + "step": 79560 + }, + { + "epoch": 5.88, + "learning_rate": 2.1309797929914832e-08, + "loss": 0.8974, + "step": 79561 + }, + { + "epoch": 5.88, + "learning_rate": 2.1283779934795357e-08, + "loss": 0.943, + "step": 79562 + }, + { + "epoch": 5.88, + "learning_rate": 2.1257777815671955e-08, + "loss": 0.862, + "step": 79563 + }, + { + "epoch": 5.88, + "learning_rate": 2.1231791572584592e-08, + "loss": 0.9841, + "step": 79564 + }, + { + "epoch": 5.88, + "learning_rate": 2.120582120557435e-08, + "loss": 0.9695, + "step": 79565 + }, + { + "epoch": 5.88, + "learning_rate": 2.1179866714682308e-08, + "loss": 0.9708, + "step": 79566 + }, + { + "epoch": 5.88, + "learning_rate": 2.1153928099950648e-08, + "loss": 0.9897, + "step": 79567 + }, + { + "epoch": 5.88, + "learning_rate": 2.1128005361420456e-08, + "loss": 1.0926, + "step": 79568 + }, + { + "epoch": 5.88, + "learning_rate": 2.1102098499132806e-08, + "loss": 1.0329, + "step": 79569 + }, + { + "epoch": 5.88, + "learning_rate": 2.1076207513128777e-08, + "loss": 0.9886, + "step": 79570 + }, + { + "epoch": 5.88, + "learning_rate": 2.1050332403450557e-08, + "loss": 0.9209, + "step": 79571 + }, + { + "epoch": 5.88, + "learning_rate": 2.1024473170137005e-08, + "loss": 1.0315, + "step": 79572 + }, + { + "epoch": 5.88, + "learning_rate": 2.099862981323253e-08, + "loss": 0.8966, + "step": 79573 + }, + { + "epoch": 5.88, + "learning_rate": 2.0972802332774878e-08, + "loss": 1.0284, + "step": 79574 + }, + { + "epoch": 5.88, + "learning_rate": 2.0946990728807347e-08, + "loss": 0.9449, + "step": 79575 + }, + { + "epoch": 5.88, + "learning_rate": 2.092119500136991e-08, + "loss": 0.9485, + "step": 79576 + }, + { + "epoch": 5.88, + "learning_rate": 2.089541515050475e-08, + "loss": 0.9891, + "step": 79577 + }, + { + "epoch": 5.88, + "learning_rate": 2.086965117625184e-08, + "loss": 0.9892, + "step": 79578 + }, + { + "epoch": 5.88, + "learning_rate": 2.084390307865336e-08, + "loss": 0.951, + "step": 79579 + }, + { + "epoch": 5.88, + "learning_rate": 2.0818170857748178e-08, + "loss": 0.9234, + "step": 79580 + }, + { + "epoch": 5.88, + "learning_rate": 2.0792454513579585e-08, + "loss": 1.1419, + "step": 79581 + }, + { + "epoch": 5.88, + "learning_rate": 2.0766754046186444e-08, + "loss": 0.891, + "step": 79582 + }, + { + "epoch": 5.88, + "learning_rate": 2.074106945561094e-08, + "loss": 1.0829, + "step": 79583 + }, + { + "epoch": 5.88, + "learning_rate": 2.0715400741894154e-08, + "loss": 1.0118, + "step": 79584 + }, + { + "epoch": 5.88, + "learning_rate": 2.0689747905076053e-08, + "loss": 0.9103, + "step": 79585 + }, + { + "epoch": 5.88, + "learning_rate": 2.0664110945197714e-08, + "loss": 1.0337, + "step": 79586 + }, + { + "epoch": 5.88, + "learning_rate": 2.0638489862300216e-08, + "loss": 0.9911, + "step": 79587 + }, + { + "epoch": 5.88, + "learning_rate": 2.0612884656423525e-08, + "loss": 0.9457, + "step": 79588 + }, + { + "epoch": 5.88, + "learning_rate": 2.0587295327609834e-08, + "loss": 0.9498, + "step": 79589 + }, + { + "epoch": 5.88, + "learning_rate": 2.0561721875898e-08, + "loss": 0.9313, + "step": 79590 + }, + { + "epoch": 5.88, + "learning_rate": 2.053616430133021e-08, + "loss": 1.0994, + "step": 79591 + }, + { + "epoch": 5.88, + "learning_rate": 2.051062260394643e-08, + "loss": 0.9287, + "step": 79592 + }, + { + "epoch": 5.88, + "learning_rate": 2.048509678378774e-08, + "loss": 1.0987, + "step": 79593 + }, + { + "epoch": 5.88, + "learning_rate": 2.045958684089411e-08, + "loss": 0.9397, + "step": 79594 + }, + { + "epoch": 5.88, + "learning_rate": 2.0434092775306613e-08, + "loss": 1.0177, + "step": 79595 + }, + { + "epoch": 5.88, + "learning_rate": 2.0408614587066334e-08, + "loss": 0.9673, + "step": 79596 + }, + { + "epoch": 5.88, + "learning_rate": 2.0383152276212127e-08, + "loss": 1.0435, + "step": 79597 + }, + { + "epoch": 5.88, + "learning_rate": 2.035770584278618e-08, + "loss": 0.9482, + "step": 79598 + }, + { + "epoch": 5.88, + "learning_rate": 2.0332275286828462e-08, + "loss": 0.8649, + "step": 79599 + }, + { + "epoch": 5.88, + "learning_rate": 2.0306860608378942e-08, + "loss": 0.9568, + "step": 79600 + }, + { + "epoch": 5.88, + "learning_rate": 2.02814618074787e-08, + "loss": 0.9675, + "step": 79601 + }, + { + "epoch": 5.88, + "learning_rate": 2.0256078884167697e-08, + "loss": 1.0729, + "step": 79602 + }, + { + "epoch": 5.88, + "learning_rate": 2.0230711838485907e-08, + "loss": 0.9553, + "step": 79603 + }, + { + "epoch": 5.88, + "learning_rate": 2.0205360670475517e-08, + "loss": 0.9733, + "step": 79604 + }, + { + "epoch": 5.88, + "learning_rate": 2.018002538017427e-08, + "loss": 1.1255, + "step": 79605 + }, + { + "epoch": 5.88, + "learning_rate": 2.0154705967624366e-08, + "loss": 0.7967, + "step": 79606 + }, + { + "epoch": 5.88, + "learning_rate": 2.0129402432865762e-08, + "loss": 1.0118, + "step": 79607 + }, + { + "epoch": 5.88, + "learning_rate": 2.0104114775938434e-08, + "loss": 0.9336, + "step": 79608 + }, + { + "epoch": 5.88, + "learning_rate": 2.0078842996882343e-08, + "loss": 1.0982, + "step": 79609 + }, + { + "epoch": 5.88, + "learning_rate": 2.005358709573857e-08, + "loss": 0.9394, + "step": 79610 + }, + { + "epoch": 5.88, + "learning_rate": 2.0028347072545974e-08, + "loss": 1.006, + "step": 79611 + }, + { + "epoch": 5.88, + "learning_rate": 2.000312292734674e-08, + "loss": 1.0463, + "step": 79612 + }, + { + "epoch": 5.88, + "learning_rate": 1.997791466017862e-08, + "loss": 0.953, + "step": 79613 + }, + { + "epoch": 5.88, + "learning_rate": 1.995272227108269e-08, + "loss": 1.0474, + "step": 79614 + }, + { + "epoch": 5.88, + "learning_rate": 1.9927545760100032e-08, + "loss": 0.9067, + "step": 79615 + }, + { + "epoch": 5.88, + "learning_rate": 1.9902385127269498e-08, + "loss": 0.9593, + "step": 79616 + }, + { + "epoch": 5.88, + "learning_rate": 1.9877240372632167e-08, + "loss": 1.0417, + "step": 79617 + }, + { + "epoch": 5.88, + "learning_rate": 1.985211149622579e-08, + "loss": 0.9827, + "step": 79618 + }, + { + "epoch": 5.88, + "learning_rate": 1.9826998498093664e-08, + "loss": 1.0186, + "step": 79619 + }, + { + "epoch": 5.88, + "learning_rate": 1.9801901378272426e-08, + "loss": 0.9366, + "step": 79620 + }, + { + "epoch": 5.88, + "learning_rate": 1.9776820136804266e-08, + "loss": 0.955, + "step": 79621 + }, + { + "epoch": 5.88, + "learning_rate": 1.975175477372915e-08, + "loss": 1.0108, + "step": 79622 + }, + { + "epoch": 5.88, + "learning_rate": 1.9726705289084825e-08, + "loss": 1.0147, + "step": 79623 + }, + { + "epoch": 5.88, + "learning_rate": 1.9701671682913482e-08, + "loss": 0.9665, + "step": 79624 + }, + { + "epoch": 5.88, + "learning_rate": 1.9676653955253976e-08, + "loss": 0.9511, + "step": 79625 + }, + { + "epoch": 5.88, + "learning_rate": 1.965165210614517e-08, + "loss": 1.0247, + "step": 79626 + }, + { + "epoch": 5.88, + "learning_rate": 1.9626666135628137e-08, + "loss": 0.8528, + "step": 79627 + }, + { + "epoch": 5.88, + "learning_rate": 1.9601696043742847e-08, + "loss": 0.9871, + "step": 79628 + }, + { + "epoch": 5.88, + "learning_rate": 1.9576741830528157e-08, + "loss": 1.0484, + "step": 79629 + }, + { + "epoch": 5.88, + "learning_rate": 1.9551803496024035e-08, + "loss": 0.9023, + "step": 79630 + }, + { + "epoch": 5.88, + "learning_rate": 1.952688104027045e-08, + "loss": 0.9381, + "step": 79631 + }, + { + "epoch": 5.88, + "learning_rate": 1.950197446330626e-08, + "loss": 1.039, + "step": 79632 + }, + { + "epoch": 5.88, + "learning_rate": 1.9477083765172543e-08, + "loss": 1.1035, + "step": 79633 + }, + { + "epoch": 5.88, + "learning_rate": 1.9452208945907048e-08, + "loss": 1.024, + "step": 79634 + }, + { + "epoch": 5.88, + "learning_rate": 1.942735000555085e-08, + "loss": 0.9319, + "step": 79635 + }, + { + "epoch": 5.88, + "learning_rate": 1.9402506944142806e-08, + "loss": 0.9634, + "step": 79636 + }, + { + "epoch": 5.88, + "learning_rate": 1.937767976172289e-08, + "loss": 1.0363, + "step": 79637 + }, + { + "epoch": 5.88, + "learning_rate": 1.9352868458329954e-08, + "loss": 0.9791, + "step": 79638 + }, + { + "epoch": 5.88, + "learning_rate": 1.932807303400286e-08, + "loss": 1.166, + "step": 79639 + }, + { + "epoch": 5.88, + "learning_rate": 1.9303293488783792e-08, + "loss": 1.0107, + "step": 79640 + }, + { + "epoch": 5.88, + "learning_rate": 1.9278529822709392e-08, + "loss": 1.1006, + "step": 79641 + }, + { + "epoch": 5.88, + "learning_rate": 1.9253782035819625e-08, + "loss": 1.0505, + "step": 79642 + }, + { + "epoch": 5.88, + "learning_rate": 1.9229050128155568e-08, + "loss": 1.0452, + "step": 79643 + }, + { + "epoch": 5.88, + "learning_rate": 1.9204334099753864e-08, + "loss": 1.0276, + "step": 79644 + }, + { + "epoch": 5.88, + "learning_rate": 1.9179633950656694e-08, + "loss": 1.0921, + "step": 79645 + }, + { + "epoch": 5.88, + "learning_rate": 1.91549496809007e-08, + "loss": 0.9513, + "step": 79646 + }, + { + "epoch": 5.89, + "learning_rate": 1.9130281290526964e-08, + "loss": 0.9993, + "step": 79647 + }, + { + "epoch": 5.89, + "learning_rate": 1.9105628779574336e-08, + "loss": 0.9777, + "step": 79648 + }, + { + "epoch": 5.89, + "learning_rate": 1.9080992148080568e-08, + "loss": 1.0427, + "step": 79649 + }, + { + "epoch": 5.89, + "learning_rate": 1.9056371396087847e-08, + "loss": 0.9242, + "step": 79650 + }, + { + "epoch": 5.89, + "learning_rate": 1.903176652363281e-08, + "loss": 1.0418, + "step": 79651 + }, + { + "epoch": 5.89, + "learning_rate": 1.9007177530755427e-08, + "loss": 0.9916, + "step": 79652 + }, + { + "epoch": 5.89, + "learning_rate": 1.8982604417494554e-08, + "loss": 1.007, + "step": 79653 + }, + { + "epoch": 5.89, + "learning_rate": 1.895804718389016e-08, + "loss": 0.9841, + "step": 79654 + }, + { + "epoch": 5.89, + "learning_rate": 1.8933505829981102e-08, + "loss": 1.1479, + "step": 79655 + }, + { + "epoch": 5.89, + "learning_rate": 1.8908980355805127e-08, + "loss": 1.0078, + "step": 79656 + }, + { + "epoch": 5.89, + "learning_rate": 1.8884470761402208e-08, + "loss": 1.0062, + "step": 79657 + }, + { + "epoch": 5.89, + "learning_rate": 1.8859977046812306e-08, + "loss": 0.9182, + "step": 79658 + }, + { + "epoch": 5.89, + "learning_rate": 1.8835499212072062e-08, + "loss": 1.1193, + "step": 79659 + }, + { + "epoch": 5.89, + "learning_rate": 1.8811037257222552e-08, + "loss": 1.0916, + "step": 79660 + }, + { + "epoch": 5.89, + "learning_rate": 1.8786591182301528e-08, + "loss": 1.1009, + "step": 79661 + }, + { + "epoch": 5.89, + "learning_rate": 1.8762160987347844e-08, + "loss": 0.9961, + "step": 79662 + }, + { + "epoch": 5.89, + "learning_rate": 1.873774667240147e-08, + "loss": 1.0227, + "step": 79663 + }, + { + "epoch": 5.89, + "learning_rate": 1.8713348237500152e-08, + "loss": 0.9765, + "step": 79664 + }, + { + "epoch": 5.89, + "learning_rate": 1.8688965682682746e-08, + "loss": 1.0223, + "step": 79665 + }, + { + "epoch": 5.89, + "learning_rate": 1.8664599007989225e-08, + "loss": 0.8902, + "step": 79666 + }, + { + "epoch": 5.89, + "learning_rate": 1.8640248213457335e-08, + "loss": 1.0122, + "step": 79667 + }, + { + "epoch": 5.89, + "learning_rate": 1.8615913299124823e-08, + "loss": 0.9911, + "step": 79668 + }, + { + "epoch": 5.89, + "learning_rate": 1.8591594265032763e-08, + "loss": 1.1568, + "step": 79669 + }, + { + "epoch": 5.89, + "learning_rate": 1.856729111121891e-08, + "loss": 1.0379, + "step": 79670 + }, + { + "epoch": 5.89, + "learning_rate": 1.8543003837721008e-08, + "loss": 0.9892, + "step": 79671 + }, + { + "epoch": 5.89, + "learning_rate": 1.8518732444579022e-08, + "loss": 1.0622, + "step": 79672 + }, + { + "epoch": 5.89, + "learning_rate": 1.8494476931829597e-08, + "loss": 1.0578, + "step": 79673 + }, + { + "epoch": 5.89, + "learning_rate": 1.8470237299513804e-08, + "loss": 1.0322, + "step": 79674 + }, + { + "epoch": 5.89, + "learning_rate": 1.8446013547669394e-08, + "loss": 1.0114, + "step": 79675 + }, + { + "epoch": 5.89, + "learning_rate": 1.8421805676334115e-08, + "loss": 0.9985, + "step": 79676 + }, + { + "epoch": 5.89, + "learning_rate": 1.839761368554571e-08, + "loss": 0.9261, + "step": 79677 + }, + { + "epoch": 5.89, + "learning_rate": 1.8373437575345265e-08, + "loss": 0.8956, + "step": 79678 + }, + { + "epoch": 5.89, + "learning_rate": 1.834927734576941e-08, + "loss": 1.0335, + "step": 79679 + }, + { + "epoch": 5.89, + "learning_rate": 1.8325132996857008e-08, + "loss": 1.0517, + "step": 79680 + }, + { + "epoch": 5.89, + "learning_rate": 1.8301004528646915e-08, + "loss": 0.9507, + "step": 79681 + }, + { + "epoch": 5.89, + "learning_rate": 1.8276891941176878e-08, + "loss": 0.8721, + "step": 79682 + }, + { + "epoch": 5.89, + "learning_rate": 1.8252795234485754e-08, + "loss": 1.035, + "step": 79683 + }, + { + "epoch": 5.89, + "learning_rate": 1.8228714408611293e-08, + "loss": 0.9012, + "step": 79684 + }, + { + "epoch": 5.89, + "learning_rate": 1.820464946359235e-08, + "loss": 0.9711, + "step": 79685 + }, + { + "epoch": 5.89, + "learning_rate": 1.8180600399466674e-08, + "loss": 1.069, + "step": 79686 + }, + { + "epoch": 5.89, + "learning_rate": 1.8156567216273124e-08, + "loss": 0.9746, + "step": 79687 + }, + { + "epoch": 5.89, + "learning_rate": 1.8132549914049446e-08, + "loss": 1.004, + "step": 79688 + }, + { + "epoch": 5.89, + "learning_rate": 1.81085484928345e-08, + "loss": 1.0462, + "step": 79689 + }, + { + "epoch": 5.89, + "learning_rate": 1.808456295266492e-08, + "loss": 1.0562, + "step": 79690 + }, + { + "epoch": 5.89, + "learning_rate": 1.8060593293581786e-08, + "loss": 1.0513, + "step": 79691 + }, + { + "epoch": 5.89, + "learning_rate": 1.8036639515619513e-08, + "loss": 1.0895, + "step": 79692 + }, + { + "epoch": 5.89, + "learning_rate": 1.8012701618819184e-08, + "loss": 0.9573, + "step": 79693 + }, + { + "epoch": 5.89, + "learning_rate": 1.7988779603217434e-08, + "loss": 1.0257, + "step": 79694 + }, + { + "epoch": 5.89, + "learning_rate": 1.7964873468853116e-08, + "loss": 1.0485, + "step": 79695 + }, + { + "epoch": 5.89, + "learning_rate": 1.7940983215763984e-08, + "loss": 0.9598, + "step": 79696 + }, + { + "epoch": 5.89, + "learning_rate": 1.7917108843987783e-08, + "loss": 0.9783, + "step": 79697 + }, + { + "epoch": 5.89, + "learning_rate": 1.7893250353562263e-08, + "loss": 1.0395, + "step": 79698 + }, + { + "epoch": 5.89, + "learning_rate": 1.786940774452628e-08, + "loss": 0.9952, + "step": 79699 + }, + { + "epoch": 5.89, + "learning_rate": 1.7845581016916467e-08, + "loss": 1.0315, + "step": 79700 + }, + { + "epoch": 5.89, + "learning_rate": 1.782177017077169e-08, + "loss": 0.9892, + "step": 79701 + }, + { + "epoch": 5.89, + "learning_rate": 1.77979752061308e-08, + "loss": 1.0491, + "step": 79702 + }, + { + "epoch": 5.89, + "learning_rate": 1.777419612302933e-08, + "loss": 1.0015, + "step": 79703 + }, + { + "epoch": 5.89, + "learning_rate": 1.7750432921506134e-08, + "loss": 1.0477, + "step": 79704 + }, + { + "epoch": 5.89, + "learning_rate": 1.7726685601600068e-08, + "loss": 0.9478, + "step": 79705 + }, + { + "epoch": 5.89, + "learning_rate": 1.7702954163347774e-08, + "loss": 1.0261, + "step": 79706 + }, + { + "epoch": 5.89, + "learning_rate": 1.7679238606786996e-08, + "loss": 1.0955, + "step": 79707 + }, + { + "epoch": 5.89, + "learning_rate": 1.7655538931955486e-08, + "loss": 0.9481, + "step": 79708 + }, + { + "epoch": 5.89, + "learning_rate": 1.7631855138892094e-08, + "loss": 1.0815, + "step": 79709 + }, + { + "epoch": 5.89, + "learning_rate": 1.7608187227633465e-08, + "loss": 1.0334, + "step": 79710 + }, + { + "epoch": 5.89, + "learning_rate": 1.7584535198216234e-08, + "loss": 0.9612, + "step": 79711 + }, + { + "epoch": 5.89, + "learning_rate": 1.756089905068037e-08, + "loss": 1.0709, + "step": 79712 + }, + { + "epoch": 5.89, + "learning_rate": 1.7537278785061396e-08, + "loss": 1.0367, + "step": 79713 + }, + { + "epoch": 5.89, + "learning_rate": 1.7513674401398173e-08, + "loss": 1.0434, + "step": 79714 + }, + { + "epoch": 5.89, + "learning_rate": 1.749008589972845e-08, + "loss": 0.9871, + "step": 79715 + }, + { + "epoch": 5.89, + "learning_rate": 1.7466513280088858e-08, + "loss": 1.0501, + "step": 79716 + }, + { + "epoch": 5.89, + "learning_rate": 1.7442956542516042e-08, + "loss": 1.0113, + "step": 79717 + }, + { + "epoch": 5.89, + "learning_rate": 1.741941568704997e-08, + "loss": 1.0168, + "step": 79718 + }, + { + "epoch": 5.89, + "learning_rate": 1.739589071372616e-08, + "loss": 0.902, + "step": 79719 + }, + { + "epoch": 5.89, + "learning_rate": 1.7372381622583477e-08, + "loss": 0.9663, + "step": 79720 + }, + { + "epoch": 5.89, + "learning_rate": 1.734888841365745e-08, + "loss": 0.9987, + "step": 79721 + }, + { + "epoch": 5.89, + "learning_rate": 1.7325411086986932e-08, + "loss": 0.9676, + "step": 79722 + }, + { + "epoch": 5.89, + "learning_rate": 1.7301949642609674e-08, + "loss": 1.0527, + "step": 79723 + }, + { + "epoch": 5.89, + "learning_rate": 1.7278504080561197e-08, + "loss": 0.8998, + "step": 79724 + }, + { + "epoch": 5.89, + "learning_rate": 1.7255074400880367e-08, + "loss": 0.9495, + "step": 79725 + }, + { + "epoch": 5.89, + "learning_rate": 1.7231660603602705e-08, + "loss": 0.8863, + "step": 79726 + }, + { + "epoch": 5.89, + "learning_rate": 1.7208262688768185e-08, + "loss": 0.9193, + "step": 79727 + }, + { + "epoch": 5.89, + "learning_rate": 1.7184880656412328e-08, + "loss": 0.9714, + "step": 79728 + }, + { + "epoch": 5.89, + "learning_rate": 1.7161514506571775e-08, + "loss": 0.9376, + "step": 79729 + }, + { + "epoch": 5.89, + "learning_rate": 1.713816423928427e-08, + "loss": 0.9653, + "step": 79730 + }, + { + "epoch": 5.89, + "learning_rate": 1.7114829854587568e-08, + "loss": 0.941, + "step": 79731 + }, + { + "epoch": 5.89, + "learning_rate": 1.7091511352518296e-08, + "loss": 0.9536, + "step": 79732 + }, + { + "epoch": 5.89, + "learning_rate": 1.7068208733113102e-08, + "loss": 1.0105, + "step": 79733 + }, + { + "epoch": 5.89, + "learning_rate": 1.7044921996410833e-08, + "loss": 1.0795, + "step": 79734 + }, + { + "epoch": 5.89, + "learning_rate": 1.7021651142445918e-08, + "loss": 0.8957, + "step": 79735 + }, + { + "epoch": 5.89, + "learning_rate": 1.6998396171257202e-08, + "loss": 0.9433, + "step": 79736 + }, + { + "epoch": 5.89, + "learning_rate": 1.6975157082881332e-08, + "loss": 0.9976, + "step": 79737 + }, + { + "epoch": 5.89, + "learning_rate": 1.6951933877354944e-08, + "loss": 1.0136, + "step": 79738 + }, + { + "epoch": 5.89, + "learning_rate": 1.692872655471578e-08, + "loss": 0.9799, + "step": 79739 + }, + { + "epoch": 5.89, + "learning_rate": 1.690553511499937e-08, + "loss": 1.0446, + "step": 79740 + }, + { + "epoch": 5.89, + "learning_rate": 1.6882359558243467e-08, + "loss": 0.938, + "step": 79741 + }, + { + "epoch": 5.89, + "learning_rate": 1.6859199884485812e-08, + "loss": 1.0437, + "step": 79742 + }, + { + "epoch": 5.89, + "learning_rate": 1.683605609376082e-08, + "loss": 1.0386, + "step": 79743 + }, + { + "epoch": 5.89, + "learning_rate": 1.6812928186108468e-08, + "loss": 1.1133, + "step": 79744 + }, + { + "epoch": 5.89, + "learning_rate": 1.6789816161563165e-08, + "loss": 1.1158, + "step": 79745 + }, + { + "epoch": 5.89, + "learning_rate": 1.6766720020161552e-08, + "loss": 1.0089, + "step": 79746 + }, + { + "epoch": 5.89, + "learning_rate": 1.6743639761942487e-08, + "loss": 0.9197, + "step": 79747 + }, + { + "epoch": 5.89, + "learning_rate": 1.6720575386941496e-08, + "loss": 0.9054, + "step": 79748 + }, + { + "epoch": 5.89, + "learning_rate": 1.6697526895195214e-08, + "loss": 1.05, + "step": 79749 + }, + { + "epoch": 5.89, + "learning_rate": 1.6674494286739174e-08, + "loss": 1.0728, + "step": 79750 + }, + { + "epoch": 5.89, + "learning_rate": 1.6651477561612227e-08, + "loss": 1.0485, + "step": 79751 + }, + { + "epoch": 5.89, + "learning_rate": 1.6628476719849906e-08, + "loss": 1.033, + "step": 79752 + }, + { + "epoch": 5.89, + "learning_rate": 1.6605491761487735e-08, + "loss": 0.9342, + "step": 79753 + }, + { + "epoch": 5.89, + "learning_rate": 1.6582522686564574e-08, + "loss": 0.9502, + "step": 79754 + }, + { + "epoch": 5.89, + "learning_rate": 1.6559569495115946e-08, + "loss": 1.0287, + "step": 79755 + }, + { + "epoch": 5.89, + "learning_rate": 1.653663218717738e-08, + "loss": 0.9084, + "step": 79756 + }, + { + "epoch": 5.89, + "learning_rate": 1.651371076278663e-08, + "loss": 0.99, + "step": 79757 + }, + { + "epoch": 5.89, + "learning_rate": 1.649080522197921e-08, + "loss": 1.1042, + "step": 79758 + }, + { + "epoch": 5.89, + "learning_rate": 1.646791556479288e-08, + "loss": 0.976, + "step": 79759 + }, + { + "epoch": 5.89, + "learning_rate": 1.6445041791263162e-08, + "loss": 0.9934, + "step": 79760 + }, + { + "epoch": 5.89, + "learning_rate": 1.642218390142558e-08, + "loss": 1.0591, + "step": 79761 + }, + { + "epoch": 5.89, + "learning_rate": 1.6399341895317888e-08, + "loss": 0.9861, + "step": 79762 + }, + { + "epoch": 5.89, + "learning_rate": 1.637651577297672e-08, + "loss": 1.0524, + "step": 79763 + }, + { + "epoch": 5.89, + "learning_rate": 1.6353705534437603e-08, + "loss": 0.9848, + "step": 79764 + }, + { + "epoch": 5.89, + "learning_rate": 1.6330911179736064e-08, + "loss": 1.0116, + "step": 79765 + }, + { + "epoch": 5.89, + "learning_rate": 1.630813270890985e-08, + "loss": 0.8957, + "step": 79766 + }, + { + "epoch": 5.89, + "learning_rate": 1.628537012199449e-08, + "loss": 1.0424, + "step": 79767 + }, + { + "epoch": 5.89, + "learning_rate": 1.6262623419025515e-08, + "loss": 1.0969, + "step": 79768 + }, + { + "epoch": 5.89, + "learning_rate": 1.623989260004066e-08, + "loss": 1.075, + "step": 79769 + }, + { + "epoch": 5.89, + "learning_rate": 1.6217177665075466e-08, + "loss": 1.0359, + "step": 79770 + }, + { + "epoch": 5.89, + "learning_rate": 1.6194478614165454e-08, + "loss": 1.0215, + "step": 79771 + }, + { + "epoch": 5.89, + "learning_rate": 1.617179544734726e-08, + "loss": 1.0406, + "step": 79772 + }, + { + "epoch": 5.89, + "learning_rate": 1.6149128164656414e-08, + "loss": 1.0429, + "step": 79773 + }, + { + "epoch": 5.89, + "learning_rate": 1.612647676612955e-08, + "loss": 0.9541, + "step": 79774 + }, + { + "epoch": 5.89, + "learning_rate": 1.610384125180331e-08, + "loss": 1.087, + "step": 79775 + }, + { + "epoch": 5.89, + "learning_rate": 1.6081221621713218e-08, + "loss": 0.9017, + "step": 79776 + }, + { + "epoch": 5.89, + "learning_rate": 1.6058617875893688e-08, + "loss": 1.0485, + "step": 79777 + }, + { + "epoch": 5.89, + "learning_rate": 1.6036030014383584e-08, + "loss": 1.0437, + "step": 79778 + }, + { + "epoch": 5.89, + "learning_rate": 1.6013458037216212e-08, + "loss": 0.9451, + "step": 79779 + }, + { + "epoch": 5.89, + "learning_rate": 1.5990901944429314e-08, + "loss": 0.9348, + "step": 79780 + }, + { + "epoch": 5.89, + "learning_rate": 1.5968361736057315e-08, + "loss": 0.9043, + "step": 79781 + }, + { + "epoch": 5.9, + "learning_rate": 1.5945837412136844e-08, + "loss": 0.9466, + "step": 79782 + }, + { + "epoch": 5.9, + "learning_rate": 1.5923328972704544e-08, + "loss": 0.9712, + "step": 79783 + }, + { + "epoch": 5.9, + "learning_rate": 1.590083641779483e-08, + "loss": 0.9583, + "step": 79784 + }, + { + "epoch": 5.9, + "learning_rate": 1.587835974744434e-08, + "loss": 0.9531, + "step": 79785 + }, + { + "epoch": 5.9, + "learning_rate": 1.585589896168749e-08, + "loss": 1.0855, + "step": 79786 + }, + { + "epoch": 5.9, + "learning_rate": 1.5833454060562025e-08, + "loss": 1.0534, + "step": 79787 + }, + { + "epoch": 5.9, + "learning_rate": 1.5811025044102368e-08, + "loss": 1.0219, + "step": 79788 + }, + { + "epoch": 5.9, + "learning_rate": 1.578861191234515e-08, + "loss": 1.0403, + "step": 79789 + }, + { + "epoch": 5.9, + "learning_rate": 1.5766214665324798e-08, + "loss": 0.926, + "step": 79790 + }, + { + "epoch": 5.9, + "learning_rate": 1.574383330307794e-08, + "loss": 0.9417, + "step": 79791 + }, + { + "epoch": 5.9, + "learning_rate": 1.5721467825638992e-08, + "loss": 0.9872, + "step": 79792 + }, + { + "epoch": 5.9, + "learning_rate": 1.569911823304571e-08, + "loss": 1.0724, + "step": 79793 + }, + { + "epoch": 5.9, + "learning_rate": 1.5676784525331388e-08, + "loss": 0.9423, + "step": 79794 + }, + { + "epoch": 5.9, + "learning_rate": 1.5654466702531567e-08, + "loss": 1.0435, + "step": 79795 + }, + { + "epoch": 5.9, + "learning_rate": 1.5632164764683987e-08, + "loss": 0.9731, + "step": 79796 + }, + { + "epoch": 5.9, + "learning_rate": 1.5609878711823066e-08, + "loss": 1.0275, + "step": 79797 + }, + { + "epoch": 5.9, + "learning_rate": 1.5587608543982113e-08, + "loss": 1.0241, + "step": 79798 + }, + { + "epoch": 5.9, + "learning_rate": 1.5565354261199984e-08, + "loss": 0.9625, + "step": 79799 + }, + { + "epoch": 5.9, + "learning_rate": 1.5543115863509984e-08, + "loss": 0.9515, + "step": 79800 + }, + { + "epoch": 5.9, + "learning_rate": 1.5520893350948752e-08, + "loss": 1.1776, + "step": 79801 + }, + { + "epoch": 5.9, + "learning_rate": 1.5498686723549594e-08, + "loss": 0.981, + "step": 79802 + }, + { + "epoch": 5.9, + "learning_rate": 1.547649598135026e-08, + "loss": 1.0078, + "step": 79803 + }, + { + "epoch": 5.9, + "learning_rate": 1.5454321124384052e-08, + "loss": 1.042, + "step": 79804 + }, + { + "epoch": 5.9, + "learning_rate": 1.5432162152688724e-08, + "loss": 0.9801, + "step": 79805 + }, + { + "epoch": 5.9, + "learning_rate": 1.5410019066296468e-08, + "loss": 0.9647, + "step": 79806 + }, + { + "epoch": 5.9, + "learning_rate": 1.538789186524392e-08, + "loss": 1.0664, + "step": 79807 + }, + { + "epoch": 5.9, + "learning_rate": 1.536578054956772e-08, + "loss": 0.8885, + "step": 79808 + }, + { + "epoch": 5.9, + "learning_rate": 1.5343685119301176e-08, + "loss": 0.9253, + "step": 79809 + }, + { + "epoch": 5.9, + "learning_rate": 1.5321605574479814e-08, + "loss": 1.1277, + "step": 79810 + }, + { + "epoch": 5.9, + "learning_rate": 1.529954191513805e-08, + "loss": 0.9638, + "step": 79811 + }, + { + "epoch": 5.9, + "learning_rate": 1.5277494141312522e-08, + "loss": 1.0538, + "step": 79812 + }, + { + "epoch": 5.9, + "learning_rate": 1.5255462253037645e-08, + "loss": 1.0759, + "step": 79813 + }, + { + "epoch": 5.9, + "learning_rate": 1.523344625034895e-08, + "loss": 0.9533, + "step": 79814 + }, + { + "epoch": 5.9, + "learning_rate": 1.521144613328085e-08, + "loss": 1.0911, + "step": 79815 + }, + { + "epoch": 5.9, + "learning_rate": 1.5189461901867765e-08, + "loss": 0.9935, + "step": 79816 + }, + { + "epoch": 5.9, + "learning_rate": 1.516749355614633e-08, + "loss": 1.0585, + "step": 79817 + }, + { + "epoch": 5.9, + "learning_rate": 1.514554109614985e-08, + "loss": 0.955, + "step": 79818 + }, + { + "epoch": 5.9, + "learning_rate": 1.512360452191497e-08, + "loss": 1.0655, + "step": 79819 + }, + { + "epoch": 5.9, + "learning_rate": 1.510168383347388e-08, + "loss": 0.9686, + "step": 79820 + }, + { + "epoch": 5.9, + "learning_rate": 1.5079779030864328e-08, + "loss": 1.0308, + "step": 79821 + }, + { + "epoch": 5.9, + "learning_rate": 1.5057890114119624e-08, + "loss": 0.9906, + "step": 79822 + }, + { + "epoch": 5.9, + "learning_rate": 1.503601708327529e-08, + "loss": 0.9821, + "step": 79823 + }, + { + "epoch": 5.9, + "learning_rate": 1.5014159938365746e-08, + "loss": 0.9272, + "step": 79824 + }, + { + "epoch": 5.9, + "learning_rate": 1.499231867942541e-08, + "loss": 0.9921, + "step": 79825 + }, + { + "epoch": 5.9, + "learning_rate": 1.497049330648981e-08, + "loss": 0.9488, + "step": 79826 + }, + { + "epoch": 5.9, + "learning_rate": 1.4948683819593357e-08, + "loss": 0.9339, + "step": 79827 + }, + { + "epoch": 5.9, + "learning_rate": 1.4926890218770472e-08, + "loss": 0.9343, + "step": 79828 + }, + { + "epoch": 5.9, + "learning_rate": 1.4905112504056684e-08, + "loss": 0.9771, + "step": 79829 + }, + { + "epoch": 5.9, + "learning_rate": 1.4883350675485297e-08, + "loss": 1.0803, + "step": 79830 + }, + { + "epoch": 5.9, + "learning_rate": 1.4861604733090729e-08, + "loss": 0.9955, + "step": 79831 + }, + { + "epoch": 5.9, + "learning_rate": 1.4839874676909616e-08, + "loss": 1.1165, + "step": 79832 + }, + { + "epoch": 5.9, + "learning_rate": 1.4818160506975266e-08, + "loss": 0.9052, + "step": 79833 + }, + { + "epoch": 5.9, + "learning_rate": 1.4796462223323205e-08, + "loss": 0.9973, + "step": 79834 + }, + { + "epoch": 5.9, + "learning_rate": 1.477477982598563e-08, + "loss": 1.0487, + "step": 79835 + }, + { + "epoch": 5.9, + "learning_rate": 1.475311331499918e-08, + "loss": 1.0205, + "step": 79836 + }, + { + "epoch": 5.9, + "learning_rate": 1.4731462690398268e-08, + "loss": 0.9955, + "step": 79837 + }, + { + "epoch": 5.9, + "learning_rate": 1.4709827952215095e-08, + "loss": 1.0708, + "step": 79838 + }, + { + "epoch": 5.9, + "learning_rate": 1.4688209100487406e-08, + "loss": 1.0614, + "step": 79839 + }, + { + "epoch": 5.9, + "learning_rate": 1.4666606135246287e-08, + "loss": 0.9519, + "step": 79840 + }, + { + "epoch": 5.9, + "learning_rate": 1.4645019056529486e-08, + "loss": 1.0416, + "step": 79841 + }, + { + "epoch": 5.9, + "learning_rate": 1.462344786436809e-08, + "loss": 1.0241, + "step": 79842 + }, + { + "epoch": 5.9, + "learning_rate": 1.4601892558798735e-08, + "loss": 1.101, + "step": 79843 + }, + { + "epoch": 5.9, + "learning_rate": 1.458035313985473e-08, + "loss": 0.8801, + "step": 79844 + }, + { + "epoch": 5.9, + "learning_rate": 1.4558829607570491e-08, + "loss": 1.0431, + "step": 79845 + }, + { + "epoch": 5.9, + "learning_rate": 1.4537321961980433e-08, + "loss": 0.9898, + "step": 79846 + }, + { + "epoch": 5.9, + "learning_rate": 1.4515830203118975e-08, + "loss": 0.9703, + "step": 79847 + }, + { + "epoch": 5.9, + "learning_rate": 1.4494354331019423e-08, + "loss": 0.9345, + "step": 79848 + }, + { + "epoch": 5.9, + "learning_rate": 1.4472894345716192e-08, + "loss": 1.0219, + "step": 79849 + }, + { + "epoch": 5.9, + "learning_rate": 1.4451450247244814e-08, + "loss": 0.9699, + "step": 79850 + }, + { + "epoch": 5.9, + "learning_rate": 1.4430022035637481e-08, + "loss": 1.0825, + "step": 79851 + }, + { + "epoch": 5.9, + "learning_rate": 1.4408609710928611e-08, + "loss": 1.0204, + "step": 79852 + }, + { + "epoch": 5.9, + "learning_rate": 1.4387213273153733e-08, + "loss": 1.0323, + "step": 79853 + }, + { + "epoch": 5.9, + "learning_rate": 1.4365832722346151e-08, + "loss": 1.0841, + "step": 79854 + }, + { + "epoch": 5.9, + "learning_rate": 1.4344468058539174e-08, + "loss": 0.9181, + "step": 79855 + }, + { + "epoch": 5.9, + "learning_rate": 1.4323119281766107e-08, + "loss": 0.9948, + "step": 79856 + }, + { + "epoch": 5.9, + "learning_rate": 1.4301786392063589e-08, + "loss": 0.9592, + "step": 79857 + }, + { + "epoch": 5.9, + "learning_rate": 1.4280469389463814e-08, + "loss": 0.9049, + "step": 79858 + }, + { + "epoch": 5.9, + "learning_rate": 1.42591682740012e-08, + "loss": 1.0262, + "step": 79859 + }, + { + "epoch": 5.9, + "learning_rate": 1.4237883045707945e-08, + "loss": 1.093, + "step": 79860 + }, + { + "epoch": 5.9, + "learning_rate": 1.4216613704620686e-08, + "loss": 1.082, + "step": 79861 + }, + { + "epoch": 5.9, + "learning_rate": 1.4195360250771618e-08, + "loss": 1.0923, + "step": 79862 + }, + { + "epoch": 5.9, + "learning_rate": 1.4174122684195157e-08, + "loss": 0.9004, + "step": 79863 + }, + { + "epoch": 5.9, + "learning_rate": 1.4152901004924612e-08, + "loss": 1.0473, + "step": 79864 + }, + { + "epoch": 5.9, + "learning_rate": 1.4131695212993291e-08, + "loss": 1.0929, + "step": 79865 + }, + { + "epoch": 5.9, + "learning_rate": 1.4110505308436717e-08, + "loss": 0.9823, + "step": 79866 + }, + { + "epoch": 5.9, + "learning_rate": 1.408933129128598e-08, + "loss": 1.0663, + "step": 79867 + }, + { + "epoch": 5.9, + "learning_rate": 1.4068173161577713e-08, + "loss": 1.0793, + "step": 79868 + }, + { + "epoch": 5.9, + "learning_rate": 1.4047030919343008e-08, + "loss": 1.1015, + "step": 79869 + }, + { + "epoch": 5.9, + "learning_rate": 1.4025904564617388e-08, + "loss": 1.048, + "step": 79870 + }, + { + "epoch": 5.9, + "learning_rate": 1.400479409743305e-08, + "loss": 0.8391, + "step": 79871 + }, + { + "epoch": 5.9, + "learning_rate": 1.3983699517824412e-08, + "loss": 0.9848, + "step": 79872 + }, + { + "epoch": 5.9, + "learning_rate": 1.396262082582478e-08, + "loss": 0.979, + "step": 79873 + }, + { + "epoch": 5.9, + "learning_rate": 1.3941558021468571e-08, + "loss": 1.0693, + "step": 79874 + }, + { + "epoch": 5.9, + "learning_rate": 1.3920511104786872e-08, + "loss": 1.082, + "step": 79875 + }, + { + "epoch": 5.9, + "learning_rate": 1.3899480075816318e-08, + "loss": 0.889, + "step": 79876 + }, + { + "epoch": 5.9, + "learning_rate": 1.3878464934587998e-08, + "loss": 0.9404, + "step": 79877 + }, + { + "epoch": 5.9, + "learning_rate": 1.3857465681136329e-08, + "loss": 1.0458, + "step": 79878 + }, + { + "epoch": 5.9, + "learning_rate": 1.3836482315494615e-08, + "loss": 0.8912, + "step": 79879 + }, + { + "epoch": 5.9, + "learning_rate": 1.3815514837696164e-08, + "loss": 1.0719, + "step": 79880 + }, + { + "epoch": 5.9, + "learning_rate": 1.3794563247774285e-08, + "loss": 0.8967, + "step": 79881 + }, + { + "epoch": 5.9, + "learning_rate": 1.377362754576228e-08, + "loss": 0.9739, + "step": 79882 + }, + { + "epoch": 5.9, + "learning_rate": 1.375270773169346e-08, + "loss": 1.1101, + "step": 79883 + }, + { + "epoch": 5.9, + "learning_rate": 1.373180380560113e-08, + "loss": 0.9475, + "step": 79884 + }, + { + "epoch": 5.9, + "learning_rate": 1.3710915767518595e-08, + "loss": 0.9388, + "step": 79885 + }, + { + "epoch": 5.9, + "learning_rate": 1.3690043617480275e-08, + "loss": 0.9946, + "step": 79886 + }, + { + "epoch": 5.9, + "learning_rate": 1.3669187355517254e-08, + "loss": 0.9884, + "step": 79887 + }, + { + "epoch": 5.9, + "learning_rate": 1.3648346981663951e-08, + "loss": 1.1693, + "step": 79888 + }, + { + "epoch": 5.9, + "learning_rate": 1.362752249595256e-08, + "loss": 1.0554, + "step": 79889 + }, + { + "epoch": 5.9, + "learning_rate": 1.3606713898417501e-08, + "loss": 0.9284, + "step": 79890 + }, + { + "epoch": 5.9, + "learning_rate": 1.3585921189090966e-08, + "loss": 0.9558, + "step": 79891 + }, + { + "epoch": 5.9, + "learning_rate": 1.3565144368006266e-08, + "loss": 0.9371, + "step": 79892 + }, + { + "epoch": 5.9, + "learning_rate": 1.3544383435197816e-08, + "loss": 0.9152, + "step": 79893 + }, + { + "epoch": 5.9, + "learning_rate": 1.3523638390695593e-08, + "loss": 1.0127, + "step": 79894 + }, + { + "epoch": 5.9, + "learning_rate": 1.3502909234535122e-08, + "loss": 0.9247, + "step": 79895 + }, + { + "epoch": 5.9, + "learning_rate": 1.34821959667486e-08, + "loss": 0.9022, + "step": 79896 + }, + { + "epoch": 5.9, + "learning_rate": 1.3461498587369337e-08, + "loss": 0.925, + "step": 79897 + }, + { + "epoch": 5.9, + "learning_rate": 1.3440817096429525e-08, + "loss": 1.0143, + "step": 79898 + }, + { + "epoch": 5.9, + "learning_rate": 1.3420151493963584e-08, + "loss": 1.0823, + "step": 79899 + }, + { + "epoch": 5.9, + "learning_rate": 1.3399501780001488e-08, + "loss": 1.027, + "step": 79900 + }, + { + "epoch": 5.9, + "learning_rate": 1.3378867954578768e-08, + "loss": 1.0707, + "step": 79901 + }, + { + "epoch": 5.9, + "learning_rate": 1.3358250017727614e-08, + "loss": 0.9848, + "step": 79902 + }, + { + "epoch": 5.9, + "learning_rate": 1.3337647969480227e-08, + "loss": 0.9773, + "step": 79903 + }, + { + "epoch": 5.9, + "learning_rate": 1.3317061809869913e-08, + "loss": 1.0404, + "step": 79904 + }, + { + "epoch": 5.9, + "learning_rate": 1.3296491538929978e-08, + "loss": 0.8987, + "step": 79905 + }, + { + "epoch": 5.9, + "learning_rate": 1.3275937156691509e-08, + "loss": 1.0415, + "step": 79906 + }, + { + "epoch": 5.9, + "learning_rate": 1.3255398663187812e-08, + "loss": 1.0641, + "step": 79907 + }, + { + "epoch": 5.9, + "learning_rate": 1.3234876058453306e-08, + "loss": 1.0207, + "step": 79908 + }, + { + "epoch": 5.9, + "learning_rate": 1.3214369342517962e-08, + "loss": 1.0194, + "step": 79909 + }, + { + "epoch": 5.9, + "learning_rate": 1.3193878515416203e-08, + "loss": 0.9934, + "step": 79910 + }, + { + "epoch": 5.9, + "learning_rate": 1.3173403577180222e-08, + "loss": 0.9683, + "step": 79911 + }, + { + "epoch": 5.9, + "learning_rate": 1.3152944527842215e-08, + "loss": 1.0382, + "step": 79912 + }, + { + "epoch": 5.9, + "learning_rate": 1.313250136743438e-08, + "loss": 1.0851, + "step": 79913 + }, + { + "epoch": 5.9, + "learning_rate": 1.3112074095991134e-08, + "loss": 0.9464, + "step": 79914 + }, + { + "epoch": 5.9, + "learning_rate": 1.3091662713542453e-08, + "loss": 0.8939, + "step": 79915 + }, + { + "epoch": 5.9, + "learning_rate": 1.3071267220122752e-08, + "loss": 0.9731, + "step": 79916 + }, + { + "epoch": 5.91, + "learning_rate": 1.3050887615764229e-08, + "loss": 1.0909, + "step": 79917 + }, + { + "epoch": 5.91, + "learning_rate": 1.303052390049797e-08, + "loss": 1.0129, + "step": 79918 + }, + { + "epoch": 5.91, + "learning_rate": 1.3010176074357284e-08, + "loss": 1.1073, + "step": 79919 + }, + { + "epoch": 5.91, + "learning_rate": 1.2989844137375474e-08, + "loss": 1.0905, + "step": 79920 + }, + { + "epoch": 5.91, + "learning_rate": 1.2969528089582517e-08, + "loss": 0.9181, + "step": 79921 + }, + { + "epoch": 5.91, + "learning_rate": 1.2949227931012831e-08, + "loss": 0.9638, + "step": 79922 + }, + { + "epoch": 5.91, + "learning_rate": 1.2928943661698612e-08, + "loss": 0.9229, + "step": 79923 + }, + { + "epoch": 5.91, + "learning_rate": 1.2908675281670946e-08, + "loss": 1.058, + "step": 79924 + }, + { + "epoch": 5.91, + "learning_rate": 1.288842279096203e-08, + "loss": 1.0923, + "step": 79925 + }, + { + "epoch": 5.91, + "learning_rate": 1.286818618960628e-08, + "loss": 1.0122, + "step": 79926 + }, + { + "epoch": 5.91, + "learning_rate": 1.2847965477633672e-08, + "loss": 0.8911, + "step": 79927 + }, + { + "epoch": 5.91, + "learning_rate": 1.2827760655076404e-08, + "loss": 0.8984, + "step": 79928 + }, + { + "epoch": 5.91, + "learning_rate": 1.2807571721967782e-08, + "loss": 0.9813, + "step": 79929 + }, + { + "epoch": 5.91, + "learning_rate": 1.2787398678338891e-08, + "loss": 0.9625, + "step": 79930 + }, + { + "epoch": 5.91, + "learning_rate": 1.276724152422415e-08, + "loss": 0.8868, + "step": 79931 + }, + { + "epoch": 5.91, + "learning_rate": 1.2747100259651313e-08, + "loss": 1.0548, + "step": 79932 + }, + { + "epoch": 5.91, + "learning_rate": 1.2726974884657018e-08, + "loss": 1.0335, + "step": 79933 + }, + { + "epoch": 5.91, + "learning_rate": 1.270686539927013e-08, + "loss": 0.965, + "step": 79934 + }, + { + "epoch": 5.91, + "learning_rate": 1.2686771803523956e-08, + "loss": 1.0432, + "step": 79935 + }, + { + "epoch": 5.91, + "learning_rate": 1.2666694097450694e-08, + "loss": 1.0636, + "step": 79936 + }, + { + "epoch": 5.91, + "learning_rate": 1.2646632281081427e-08, + "loss": 0.9839, + "step": 79937 + }, + { + "epoch": 5.91, + "learning_rate": 1.2626586354448356e-08, + "loss": 0.9789, + "step": 79938 + }, + { + "epoch": 5.91, + "learning_rate": 1.2606556317583674e-08, + "loss": 1.0118, + "step": 79939 + }, + { + "epoch": 5.91, + "learning_rate": 1.2586542170518467e-08, + "loss": 0.9622, + "step": 79940 + }, + { + "epoch": 5.91, + "learning_rate": 1.2566543913286044e-08, + "loss": 1.0476, + "step": 79941 + }, + { + "epoch": 5.91, + "learning_rate": 1.254656154591638e-08, + "loss": 1.0127, + "step": 79942 + }, + { + "epoch": 5.91, + "learning_rate": 1.2526595068442782e-08, + "loss": 1.0626, + "step": 79943 + }, + { + "epoch": 5.91, + "learning_rate": 1.2506644480897445e-08, + "loss": 1.0459, + "step": 79944 + }, + { + "epoch": 5.91, + "learning_rate": 1.2486709783309237e-08, + "loss": 0.8776, + "step": 79945 + }, + { + "epoch": 5.91, + "learning_rate": 1.2466790975713682e-08, + "loss": 0.9771, + "step": 79946 + }, + { + "epoch": 5.91, + "learning_rate": 1.244688805813965e-08, + "loss": 0.9816, + "step": 79947 + }, + { + "epoch": 5.91, + "learning_rate": 1.2427001030619334e-08, + "loss": 0.8888, + "step": 79948 + }, + { + "epoch": 5.91, + "learning_rate": 1.2407129893184933e-08, + "loss": 0.8565, + "step": 79949 + }, + { + "epoch": 5.91, + "learning_rate": 1.2387274645868641e-08, + "loss": 0.9784, + "step": 79950 + }, + { + "epoch": 5.91, + "learning_rate": 1.2367435288700436e-08, + "loss": 0.9589, + "step": 79951 + }, + { + "epoch": 5.91, + "learning_rate": 1.2347611821713623e-08, + "loss": 0.9855, + "step": 79952 + }, + { + "epoch": 5.91, + "learning_rate": 1.232780424493818e-08, + "loss": 0.9847, + "step": 79953 + }, + { + "epoch": 5.91, + "learning_rate": 1.23080125584063e-08, + "loss": 1.0217, + "step": 79954 + }, + { + "epoch": 5.91, + "learning_rate": 1.2288236762149075e-08, + "loss": 0.9751, + "step": 79955 + }, + { + "epoch": 5.91, + "learning_rate": 1.2268476856199806e-08, + "loss": 1.0243, + "step": 79956 + }, + { + "epoch": 5.91, + "learning_rate": 1.2248732840587362e-08, + "loss": 0.936, + "step": 79957 + }, + { + "epoch": 5.91, + "learning_rate": 1.222900471534394e-08, + "loss": 0.9735, + "step": 79958 + }, + { + "epoch": 5.91, + "learning_rate": 1.2209292480501733e-08, + "loss": 0.9697, + "step": 79959 + }, + { + "epoch": 5.91, + "learning_rate": 1.218959613609183e-08, + "loss": 1.012, + "step": 79960 + }, + { + "epoch": 5.91, + "learning_rate": 1.2169915682145317e-08, + "loss": 0.9924, + "step": 79961 + }, + { + "epoch": 5.91, + "learning_rate": 1.215025111869328e-08, + "loss": 0.975, + "step": 79962 + }, + { + "epoch": 5.91, + "learning_rate": 1.2130602445767913e-08, + "loss": 0.9896, + "step": 79963 + }, + { + "epoch": 5.91, + "learning_rate": 1.2110969663399197e-08, + "loss": 1.0366, + "step": 79964 + }, + { + "epoch": 5.91, + "learning_rate": 1.2091352771620435e-08, + "loss": 0.8823, + "step": 79965 + }, + { + "epoch": 5.91, + "learning_rate": 1.2071751770459383e-08, + "loss": 0.8687, + "step": 79966 + }, + { + "epoch": 5.91, + "learning_rate": 1.2052166659950459e-08, + "loss": 0.8746, + "step": 79967 + }, + { + "epoch": 5.91, + "learning_rate": 1.203259744012364e-08, + "loss": 1.0494, + "step": 79968 + }, + { + "epoch": 5.91, + "learning_rate": 1.2013044111010008e-08, + "loss": 1.0152, + "step": 79969 + }, + { + "epoch": 5.91, + "learning_rate": 1.1993506672639543e-08, + "loss": 1.1406, + "step": 79970 + }, + { + "epoch": 5.91, + "learning_rate": 1.197398512504555e-08, + "loss": 0.9744, + "step": 79971 + }, + { + "epoch": 5.91, + "learning_rate": 1.1954479468258007e-08, + "loss": 0.9968, + "step": 79972 + }, + { + "epoch": 5.91, + "learning_rate": 1.1934989702307997e-08, + "loss": 1.0458, + "step": 79973 + }, + { + "epoch": 5.91, + "learning_rate": 1.1915515827226609e-08, + "loss": 0.9995, + "step": 79974 + }, + { + "epoch": 5.91, + "learning_rate": 1.1896057843044929e-08, + "loss": 1.0914, + "step": 79975 + }, + { + "epoch": 5.91, + "learning_rate": 1.187661574979293e-08, + "loss": 1.0296, + "step": 79976 + }, + { + "epoch": 5.91, + "learning_rate": 1.1857189547502813e-08, + "loss": 0.9742, + "step": 79977 + }, + { + "epoch": 5.91, + "learning_rate": 1.183777923620455e-08, + "loss": 0.8894, + "step": 79978 + }, + { + "epoch": 5.91, + "learning_rate": 1.181838481593034e-08, + "loss": 1.0234, + "step": 79979 + }, + { + "epoch": 5.91, + "learning_rate": 1.1799006286710158e-08, + "loss": 0.957, + "step": 79980 + }, + { + "epoch": 5.91, + "learning_rate": 1.177964364857398e-08, + "loss": 0.9704, + "step": 79981 + }, + { + "epoch": 5.91, + "learning_rate": 1.1760296901554002e-08, + "loss": 1.0001, + "step": 79982 + }, + { + "epoch": 5.91, + "learning_rate": 1.1740966045680202e-08, + "loss": 1.0346, + "step": 79983 + }, + { + "epoch": 5.91, + "learning_rate": 1.1721651080983664e-08, + "loss": 0.9465, + "step": 79984 + }, + { + "epoch": 5.91, + "learning_rate": 1.1702352007495476e-08, + "loss": 0.987, + "step": 79985 + }, + { + "epoch": 5.91, + "learning_rate": 1.1683068825245613e-08, + "loss": 0.922, + "step": 79986 + }, + { + "epoch": 5.91, + "learning_rate": 1.1663801534265162e-08, + "loss": 1.0313, + "step": 79987 + }, + { + "epoch": 5.91, + "learning_rate": 1.1644550134584099e-08, + "loss": 0.9739, + "step": 79988 + }, + { + "epoch": 5.91, + "learning_rate": 1.1625314626234618e-08, + "loss": 1.0049, + "step": 79989 + }, + { + "epoch": 5.91, + "learning_rate": 1.1606095009245588e-08, + "loss": 1.0307, + "step": 79990 + }, + { + "epoch": 5.91, + "learning_rate": 1.1586891283649203e-08, + "loss": 1.0082, + "step": 79991 + }, + { + "epoch": 5.91, + "learning_rate": 1.1567703449474332e-08, + "loss": 0.9625, + "step": 79992 + }, + { + "epoch": 5.91, + "learning_rate": 1.1548531506752058e-08, + "loss": 0.9573, + "step": 79993 + }, + { + "epoch": 5.91, + "learning_rate": 1.152937545551458e-08, + "loss": 0.9171, + "step": 79994 + }, + { + "epoch": 5.91, + "learning_rate": 1.1510235295789651e-08, + "loss": 1.0068, + "step": 79995 + }, + { + "epoch": 5.91, + "learning_rate": 1.149111102760947e-08, + "loss": 1.0233, + "step": 79996 + }, + { + "epoch": 5.91, + "learning_rate": 1.147200265100401e-08, + "loss": 1.005, + "step": 79997 + }, + { + "epoch": 5.91, + "learning_rate": 1.145291016600436e-08, + "loss": 0.9792, + "step": 79998 + }, + { + "epoch": 5.91, + "learning_rate": 1.1433833572639385e-08, + "loss": 0.8999, + "step": 79999 + }, + { + "epoch": 5.91, + "learning_rate": 1.1414772870940171e-08, + "loss": 0.9385, + "step": 80000 + }, + { + "epoch": 5.91, + "learning_rate": 1.1395728060937804e-08, + "loss": 1.0148, + "step": 80001 + }, + { + "epoch": 5.91, + "learning_rate": 1.1376699142661152e-08, + "loss": 0.9826, + "step": 80002 + }, + { + "epoch": 5.91, + "learning_rate": 1.1357686116141297e-08, + "loss": 0.9343, + "step": 80003 + }, + { + "epoch": 5.91, + "learning_rate": 1.1338688981409329e-08, + "loss": 0.8974, + "step": 80004 + }, + { + "epoch": 5.91, + "learning_rate": 1.1319707738494113e-08, + "loss": 0.9583, + "step": 80005 + }, + { + "epoch": 5.91, + "learning_rate": 1.1300742387425623e-08, + "loss": 0.9246, + "step": 80006 + }, + { + "epoch": 5.91, + "learning_rate": 1.1281792928236057e-08, + "loss": 0.841, + "step": 80007 + }, + { + "epoch": 5.91, + "learning_rate": 1.126285936095317e-08, + "loss": 1.0622, + "step": 80008 + }, + { + "epoch": 5.91, + "learning_rate": 1.1243941685609161e-08, + "loss": 1.0313, + "step": 80009 + }, + { + "epoch": 5.91, + "learning_rate": 1.1225039902231782e-08, + "loss": 1.0075, + "step": 80010 + }, + { + "epoch": 5.91, + "learning_rate": 1.1206154010853232e-08, + "loss": 0.9854, + "step": 80011 + }, + { + "epoch": 5.91, + "learning_rate": 1.1187284011503486e-08, + "loss": 0.9231, + "step": 80012 + }, + { + "epoch": 5.91, + "learning_rate": 1.1168429904210299e-08, + "loss": 0.8993, + "step": 80013 + }, + { + "epoch": 5.91, + "learning_rate": 1.1149591689005867e-08, + "loss": 1.0544, + "step": 80014 + }, + { + "epoch": 5.91, + "learning_rate": 1.1130769365920169e-08, + "loss": 0.9935, + "step": 80015 + }, + { + "epoch": 5.91, + "learning_rate": 1.1111962934982067e-08, + "loss": 1.1164, + "step": 80016 + }, + { + "epoch": 5.91, + "learning_rate": 1.109317239622154e-08, + "loss": 1.0103, + "step": 80017 + }, + { + "epoch": 5.91, + "learning_rate": 1.1074397749669674e-08, + "loss": 0.9126, + "step": 80018 + }, + { + "epoch": 5.91, + "learning_rate": 1.1055638995354223e-08, + "loss": 1.1097, + "step": 80019 + }, + { + "epoch": 5.91, + "learning_rate": 1.1036896133307385e-08, + "loss": 0.9324, + "step": 80020 + }, + { + "epoch": 5.91, + "learning_rate": 1.1018169163558023e-08, + "loss": 1.0088, + "step": 80021 + }, + { + "epoch": 5.91, + "learning_rate": 1.0999458086135007e-08, + "loss": 0.9588, + "step": 80022 + }, + { + "epoch": 5.91, + "learning_rate": 1.098076290106942e-08, + "loss": 1.032, + "step": 80023 + }, + { + "epoch": 5.91, + "learning_rate": 1.0962083608391239e-08, + "loss": 1.0925, + "step": 80024 + }, + { + "epoch": 5.91, + "learning_rate": 1.094342020812822e-08, + "loss": 1.0566, + "step": 80025 + }, + { + "epoch": 5.91, + "learning_rate": 1.0924772700311447e-08, + "loss": 1.0048, + "step": 80026 + }, + { + "epoch": 5.91, + "learning_rate": 1.09061410849709e-08, + "loss": 0.9726, + "step": 80027 + }, + { + "epoch": 5.91, + "learning_rate": 1.0887525362136552e-08, + "loss": 0.9557, + "step": 80028 + }, + { + "epoch": 5.91, + "learning_rate": 1.086892553183616e-08, + "loss": 1.0789, + "step": 80029 + }, + { + "epoch": 5.91, + "learning_rate": 1.085034159410081e-08, + "loss": 1.0291, + "step": 80030 + }, + { + "epoch": 5.91, + "learning_rate": 1.0831773548959368e-08, + "loss": 1.0563, + "step": 80031 + }, + { + "epoch": 5.91, + "learning_rate": 1.0813221396441808e-08, + "loss": 0.9306, + "step": 80032 + }, + { + "epoch": 5.91, + "learning_rate": 1.0794685136577e-08, + "loss": 0.9323, + "step": 80033 + }, + { + "epoch": 5.91, + "learning_rate": 1.0776164769394914e-08, + "loss": 0.955, + "step": 80034 + }, + { + "epoch": 5.91, + "learning_rate": 1.0757660294925532e-08, + "loss": 0.9014, + "step": 80035 + }, + { + "epoch": 5.91, + "learning_rate": 1.0739171713197716e-08, + "loss": 1.0275, + "step": 80036 + }, + { + "epoch": 5.91, + "learning_rate": 1.0720699024241443e-08, + "loss": 0.9795, + "step": 80037 + }, + { + "epoch": 5.91, + "learning_rate": 1.0702242228084469e-08, + "loss": 1.0149, + "step": 80038 + }, + { + "epoch": 5.91, + "learning_rate": 1.068380132475788e-08, + "loss": 1.0661, + "step": 80039 + }, + { + "epoch": 5.91, + "learning_rate": 1.0665376314290543e-08, + "loss": 1.0262, + "step": 80040 + }, + { + "epoch": 5.91, + "learning_rate": 1.0646967196711322e-08, + "loss": 1.0444, + "step": 80041 + }, + { + "epoch": 5.91, + "learning_rate": 1.0628573972050193e-08, + "loss": 1.0264, + "step": 80042 + }, + { + "epoch": 5.91, + "learning_rate": 1.0610196640336023e-08, + "loss": 0.9558, + "step": 80043 + }, + { + "epoch": 5.91, + "learning_rate": 1.0591835201598788e-08, + "loss": 1.0983, + "step": 80044 + }, + { + "epoch": 5.91, + "learning_rate": 1.0573489655866242e-08, + "loss": 0.9135, + "step": 80045 + }, + { + "epoch": 5.91, + "learning_rate": 1.0555160003168364e-08, + "loss": 1.0208, + "step": 80046 + }, + { + "epoch": 5.91, + "learning_rate": 1.0536846243534015e-08, + "loss": 1.1043, + "step": 80047 + }, + { + "epoch": 5.91, + "learning_rate": 1.0518548376993177e-08, + "loss": 0.9209, + "step": 80048 + }, + { + "epoch": 5.91, + "learning_rate": 1.0500266403574711e-08, + "loss": 0.9539, + "step": 80049 + }, + { + "epoch": 5.91, + "learning_rate": 1.0482000323306374e-08, + "loss": 1.0598, + "step": 80050 + }, + { + "epoch": 5.91, + "learning_rate": 1.0463750136219253e-08, + "loss": 1.0205, + "step": 80051 + }, + { + "epoch": 5.91, + "learning_rate": 1.0445515842341103e-08, + "loss": 1.0206, + "step": 80052 + }, + { + "epoch": 5.92, + "learning_rate": 1.042729744170079e-08, + "loss": 1.0583, + "step": 80053 + }, + { + "epoch": 5.92, + "learning_rate": 1.040909493432718e-08, + "loss": 0.8546, + "step": 80054 + }, + { + "epoch": 5.92, + "learning_rate": 1.0390908320251359e-08, + "loss": 0.9732, + "step": 80055 + }, + { + "epoch": 5.92, + "learning_rate": 1.0372737599499971e-08, + "loss": 1.0579, + "step": 80056 + }, + { + "epoch": 5.92, + "learning_rate": 1.0354582772101885e-08, + "loss": 0.9301, + "step": 80057 + }, + { + "epoch": 5.92, + "learning_rate": 1.0336443838087073e-08, + "loss": 1.0636, + "step": 80058 + }, + { + "epoch": 5.92, + "learning_rate": 1.0318320797484405e-08, + "loss": 0.9772, + "step": 80059 + }, + { + "epoch": 5.92, + "learning_rate": 1.0300213650322743e-08, + "loss": 1.0179, + "step": 80060 + }, + { + "epoch": 5.92, + "learning_rate": 1.0282122396629846e-08, + "loss": 1.0938, + "step": 80061 + }, + { + "epoch": 5.92, + "learning_rate": 1.0264047036435687e-08, + "loss": 1.0565, + "step": 80062 + }, + { + "epoch": 5.92, + "learning_rate": 1.0245987569768023e-08, + "loss": 1.0041, + "step": 80063 + }, + { + "epoch": 5.92, + "learning_rate": 1.022794399665572e-08, + "loss": 1.0322, + "step": 80064 + }, + { + "epoch": 5.92, + "learning_rate": 1.0209916317128755e-08, + "loss": 1.0052, + "step": 80065 + }, + { + "epoch": 5.92, + "learning_rate": 1.019190453121488e-08, + "loss": 1.0858, + "step": 80066 + }, + { + "epoch": 5.92, + "learning_rate": 1.0173908638941855e-08, + "loss": 1.0695, + "step": 80067 + }, + { + "epoch": 5.92, + "learning_rate": 1.0155928640339652e-08, + "loss": 1.1876, + "step": 80068 + }, + { + "epoch": 5.92, + "learning_rate": 1.0137964535436028e-08, + "loss": 0.9821, + "step": 80069 + }, + { + "epoch": 5.92, + "learning_rate": 1.0120016324260962e-08, + "loss": 1.0334, + "step": 80070 + }, + { + "epoch": 5.92, + "learning_rate": 1.0102084006841095e-08, + "loss": 1.1248, + "step": 80071 + }, + { + "epoch": 5.92, + "learning_rate": 1.0084167583206405e-08, + "loss": 0.9538, + "step": 80072 + }, + { + "epoch": 5.92, + "learning_rate": 1.0066267053384648e-08, + "loss": 0.9501, + "step": 80073 + }, + { + "epoch": 5.92, + "learning_rate": 1.004838241740469e-08, + "loss": 1.0562, + "step": 80074 + }, + { + "epoch": 5.92, + "learning_rate": 1.0030513675294285e-08, + "loss": 1.0421, + "step": 80075 + }, + { + "epoch": 5.92, + "learning_rate": 1.001266082708341e-08, + "loss": 1.0119, + "step": 80076 + }, + { + "epoch": 5.92, + "learning_rate": 9.99482387279871e-09, + "loss": 1.0598, + "step": 80077 + }, + { + "epoch": 5.92, + "learning_rate": 9.97700281247016e-09, + "loss": 0.967, + "step": 80078 + }, + { + "epoch": 5.92, + "learning_rate": 9.959197646124408e-09, + "loss": 1.0042, + "step": 80079 + }, + { + "epoch": 5.92, + "learning_rate": 9.941408373790318e-09, + "loss": 0.9917, + "step": 80080 + }, + { + "epoch": 5.92, + "learning_rate": 9.923634995496755e-09, + "loss": 1.0769, + "step": 80081 + }, + { + "epoch": 5.92, + "learning_rate": 9.905877511272588e-09, + "loss": 0.9816, + "step": 80082 + }, + { + "epoch": 5.92, + "learning_rate": 9.88813592114446e-09, + "loss": 0.9475, + "step": 80083 + }, + { + "epoch": 5.92, + "learning_rate": 9.870410225141237e-09, + "loss": 1.025, + "step": 80084 + }, + { + "epoch": 5.92, + "learning_rate": 9.852700423291783e-09, + "loss": 0.9683, + "step": 80085 + }, + { + "epoch": 5.92, + "learning_rate": 9.835006515623857e-09, + "loss": 1.0661, + "step": 80086 + }, + { + "epoch": 5.92, + "learning_rate": 9.817328502165213e-09, + "loss": 1.038, + "step": 80087 + }, + { + "epoch": 5.92, + "learning_rate": 9.799666382944717e-09, + "loss": 0.8907, + "step": 80088 + }, + { + "epoch": 5.92, + "learning_rate": 9.782020157989015e-09, + "loss": 1.0507, + "step": 80089 + }, + { + "epoch": 5.92, + "learning_rate": 9.76438982732808e-09, + "loss": 1.076, + "step": 80090 + }, + { + "epoch": 5.92, + "learning_rate": 9.746775390988561e-09, + "loss": 1.0009, + "step": 80091 + }, + { + "epoch": 5.92, + "learning_rate": 9.729176848999321e-09, + "loss": 0.983, + "step": 80092 + }, + { + "epoch": 5.92, + "learning_rate": 9.711594201387009e-09, + "loss": 1.1254, + "step": 80093 + }, + { + "epoch": 5.92, + "learning_rate": 9.694027448181598e-09, + "loss": 1.0188, + "step": 80094 + }, + { + "epoch": 5.92, + "learning_rate": 9.676476589409734e-09, + "loss": 0.9875, + "step": 80095 + }, + { + "epoch": 5.92, + "learning_rate": 9.658941625100281e-09, + "loss": 0.8293, + "step": 80096 + }, + { + "epoch": 5.92, + "learning_rate": 9.641422555279889e-09, + "loss": 1.0214, + "step": 80097 + }, + { + "epoch": 5.92, + "learning_rate": 9.62391937997742e-09, + "loss": 0.9975, + "step": 80098 + }, + { + "epoch": 5.92, + "learning_rate": 9.606432099220631e-09, + "loss": 1.0957, + "step": 80099 + }, + { + "epoch": 5.92, + "learning_rate": 9.588960713036166e-09, + "loss": 1.0673, + "step": 80100 + }, + { + "epoch": 5.92, + "learning_rate": 9.571505221452892e-09, + "loss": 1.0337, + "step": 80101 + }, + { + "epoch": 5.92, + "learning_rate": 9.554065624498565e-09, + "loss": 0.9517, + "step": 80102 + }, + { + "epoch": 5.92, + "learning_rate": 9.53664192220094e-09, + "loss": 1.0485, + "step": 80103 + }, + { + "epoch": 5.92, + "learning_rate": 9.519234114587773e-09, + "loss": 0.8668, + "step": 80104 + }, + { + "epoch": 5.92, + "learning_rate": 9.501842201685707e-09, + "loss": 1.1212, + "step": 80105 + }, + { + "epoch": 5.92, + "learning_rate": 9.484466183524722e-09, + "loss": 0.9184, + "step": 80106 + }, + { + "epoch": 5.92, + "learning_rate": 9.467106060130348e-09, + "loss": 1.064, + "step": 80107 + }, + { + "epoch": 5.92, + "learning_rate": 9.449761831530347e-09, + "loss": 1.0658, + "step": 80108 + }, + { + "epoch": 5.92, + "learning_rate": 9.43243349775247e-09, + "loss": 0.9376, + "step": 80109 + }, + { + "epoch": 5.92, + "learning_rate": 9.415121058825582e-09, + "loss": 0.9906, + "step": 80110 + }, + { + "epoch": 5.92, + "learning_rate": 9.397824514776332e-09, + "loss": 1.0061, + "step": 80111 + }, + { + "epoch": 5.92, + "learning_rate": 9.380543865632474e-09, + "loss": 1.083, + "step": 80112 + }, + { + "epoch": 5.92, + "learning_rate": 9.363279111419543e-09, + "loss": 0.9899, + "step": 80113 + }, + { + "epoch": 5.92, + "learning_rate": 9.346030252168625e-09, + "loss": 1.0271, + "step": 80114 + }, + { + "epoch": 5.92, + "learning_rate": 9.328797287904145e-09, + "loss": 0.8474, + "step": 80115 + }, + { + "epoch": 5.92, + "learning_rate": 9.31158021865497e-09, + "loss": 1.0048, + "step": 80116 + }, + { + "epoch": 5.92, + "learning_rate": 9.294379044447743e-09, + "loss": 0.9832, + "step": 80117 + }, + { + "epoch": 5.92, + "learning_rate": 9.277193765310222e-09, + "loss": 1.0399, + "step": 80118 + }, + { + "epoch": 5.92, + "learning_rate": 9.26002438127127e-09, + "loss": 1.02, + "step": 80119 + }, + { + "epoch": 5.92, + "learning_rate": 9.242870892355315e-09, + "loss": 1.0793, + "step": 80120 + }, + { + "epoch": 5.92, + "learning_rate": 9.225733298591222e-09, + "loss": 0.9648, + "step": 80121 + }, + { + "epoch": 5.92, + "learning_rate": 9.208611600005635e-09, + "loss": 0.982, + "step": 80122 + }, + { + "epoch": 5.92, + "learning_rate": 9.19150579662631e-09, + "loss": 1.026, + "step": 80123 + }, + { + "epoch": 5.92, + "learning_rate": 9.174415888481002e-09, + "loss": 0.9737, + "step": 80124 + }, + { + "epoch": 5.92, + "learning_rate": 9.157341875595249e-09, + "loss": 0.9898, + "step": 80125 + }, + { + "epoch": 5.92, + "learning_rate": 9.140283757997915e-09, + "loss": 0.9278, + "step": 80126 + }, + { + "epoch": 5.92, + "learning_rate": 9.123241535715643e-09, + "loss": 1.0998, + "step": 80127 + }, + { + "epoch": 5.92, + "learning_rate": 9.106215208775082e-09, + "loss": 0.9168, + "step": 80128 + }, + { + "epoch": 5.92, + "learning_rate": 9.089204777202876e-09, + "loss": 0.9784, + "step": 80129 + }, + { + "epoch": 5.92, + "learning_rate": 9.07221024102789e-09, + "loss": 1.1162, + "step": 80130 + }, + { + "epoch": 5.92, + "learning_rate": 9.05523160027566e-09, + "loss": 1.0126, + "step": 80131 + }, + { + "epoch": 5.92, + "learning_rate": 9.038268854973941e-09, + "loss": 0.9793, + "step": 80132 + }, + { + "epoch": 5.92, + "learning_rate": 9.02132200514827e-09, + "loss": 1.1599, + "step": 80133 + }, + { + "epoch": 5.92, + "learning_rate": 9.00439105082751e-09, + "loss": 1.0411, + "step": 80134 + }, + { + "epoch": 5.92, + "learning_rate": 8.987475992038308e-09, + "loss": 1.0428, + "step": 80135 + }, + { + "epoch": 5.92, + "learning_rate": 8.9705768288062e-09, + "loss": 0.9684, + "step": 80136 + }, + { + "epoch": 5.92, + "learning_rate": 8.953693561160049e-09, + "loss": 0.9453, + "step": 80137 + }, + { + "epoch": 5.92, + "learning_rate": 8.93682618912428e-09, + "loss": 0.8291, + "step": 80138 + }, + { + "epoch": 5.92, + "learning_rate": 8.919974712727764e-09, + "loss": 0.997, + "step": 80139 + }, + { + "epoch": 5.92, + "learning_rate": 8.903139131997141e-09, + "loss": 1.0039, + "step": 80140 + }, + { + "epoch": 5.92, + "learning_rate": 8.886319446957947e-09, + "loss": 0.9769, + "step": 80141 + }, + { + "epoch": 5.92, + "learning_rate": 8.869515657637939e-09, + "loss": 1.0266, + "step": 80142 + }, + { + "epoch": 5.92, + "learning_rate": 8.852727764062651e-09, + "loss": 1.0877, + "step": 80143 + }, + { + "epoch": 5.92, + "learning_rate": 8.83595576626095e-09, + "loss": 1.0694, + "step": 80144 + }, + { + "epoch": 5.92, + "learning_rate": 8.819199664257261e-09, + "loss": 1.0858, + "step": 80145 + }, + { + "epoch": 5.92, + "learning_rate": 8.802459458080448e-09, + "loss": 1.0575, + "step": 80146 + }, + { + "epoch": 5.92, + "learning_rate": 8.785735147754937e-09, + "loss": 0.9458, + "step": 80147 + }, + { + "epoch": 5.92, + "learning_rate": 8.769026733309593e-09, + "loss": 1.0329, + "step": 80148 + }, + { + "epoch": 5.92, + "learning_rate": 8.752334214768844e-09, + "loss": 0.8679, + "step": 80149 + }, + { + "epoch": 5.92, + "learning_rate": 8.73565759216044e-09, + "loss": 0.9607, + "step": 80150 + }, + { + "epoch": 5.92, + "learning_rate": 8.718996865509921e-09, + "loss": 0.9988, + "step": 80151 + }, + { + "epoch": 5.92, + "learning_rate": 8.702352034846152e-09, + "loss": 1.0036, + "step": 80152 + }, + { + "epoch": 5.92, + "learning_rate": 8.685723100192445e-09, + "loss": 0.8829, + "step": 80153 + }, + { + "epoch": 5.92, + "learning_rate": 8.669110061577667e-09, + "loss": 0.9186, + "step": 80154 + }, + { + "epoch": 5.92, + "learning_rate": 8.652512919027356e-09, + "loss": 1.0125, + "step": 80155 + }, + { + "epoch": 5.92, + "learning_rate": 8.635931672567043e-09, + "loss": 0.9002, + "step": 80156 + }, + { + "epoch": 5.92, + "learning_rate": 8.619366322224487e-09, + "loss": 0.8975, + "step": 80157 + }, + { + "epoch": 5.92, + "learning_rate": 8.602816868026332e-09, + "loss": 0.9526, + "step": 80158 + }, + { + "epoch": 5.92, + "learning_rate": 8.586283309997001e-09, + "loss": 1.083, + "step": 80159 + }, + { + "epoch": 5.92, + "learning_rate": 8.569765648164252e-09, + "loss": 0.9891, + "step": 80160 + }, + { + "epoch": 5.92, + "learning_rate": 8.55326388255473e-09, + "loss": 1.006, + "step": 80161 + }, + { + "epoch": 5.92, + "learning_rate": 8.536778013193969e-09, + "loss": 1.0525, + "step": 80162 + }, + { + "epoch": 5.92, + "learning_rate": 8.520308040107505e-09, + "loss": 0.9572, + "step": 80163 + }, + { + "epoch": 5.92, + "learning_rate": 8.503853963323094e-09, + "loss": 1.0649, + "step": 80164 + }, + { + "epoch": 5.92, + "learning_rate": 8.48741578286516e-09, + "loss": 0.9834, + "step": 80165 + }, + { + "epoch": 5.92, + "learning_rate": 8.470993498761459e-09, + "loss": 0.9593, + "step": 80166 + }, + { + "epoch": 5.92, + "learning_rate": 8.454587111036417e-09, + "loss": 0.9552, + "step": 80167 + }, + { + "epoch": 5.92, + "learning_rate": 8.438196619718896e-09, + "loss": 0.9354, + "step": 80168 + }, + { + "epoch": 5.92, + "learning_rate": 8.421822024832216e-09, + "loss": 1.0608, + "step": 80169 + }, + { + "epoch": 5.92, + "learning_rate": 8.405463326403018e-09, + "loss": 0.9966, + "step": 80170 + }, + { + "epoch": 5.92, + "learning_rate": 8.389120524459061e-09, + "loss": 0.954, + "step": 80171 + }, + { + "epoch": 5.92, + "learning_rate": 8.372793619023656e-09, + "loss": 0.9436, + "step": 80172 + }, + { + "epoch": 5.92, + "learning_rate": 8.35648261012567e-09, + "loss": 0.9498, + "step": 80173 + }, + { + "epoch": 5.92, + "learning_rate": 8.340187497788421e-09, + "loss": 1.0638, + "step": 80174 + }, + { + "epoch": 5.92, + "learning_rate": 8.32390828203855e-09, + "loss": 0.9186, + "step": 80175 + }, + { + "epoch": 5.92, + "learning_rate": 8.307644962903817e-09, + "loss": 1.018, + "step": 80176 + }, + { + "epoch": 5.92, + "learning_rate": 8.29139754040753e-09, + "loss": 1.023, + "step": 80177 + }, + { + "epoch": 5.92, + "learning_rate": 8.275166014577452e-09, + "loss": 0.9819, + "step": 80178 + }, + { + "epoch": 5.92, + "learning_rate": 8.258950385438002e-09, + "loss": 1.0502, + "step": 80179 + }, + { + "epoch": 5.92, + "learning_rate": 8.24275065301583e-09, + "loss": 1.0092, + "step": 80180 + }, + { + "epoch": 5.92, + "learning_rate": 8.226566817336467e-09, + "loss": 0.952, + "step": 80181 + }, + { + "epoch": 5.92, + "learning_rate": 8.210398878426562e-09, + "loss": 1.0686, + "step": 80182 + }, + { + "epoch": 5.92, + "learning_rate": 8.194246836309427e-09, + "loss": 1.01, + "step": 80183 + }, + { + "epoch": 5.92, + "learning_rate": 8.17811069101393e-09, + "loss": 0.9791, + "step": 80184 + }, + { + "epoch": 5.92, + "learning_rate": 8.161990442562273e-09, + "loss": 1.0834, + "step": 80185 + }, + { + "epoch": 5.92, + "learning_rate": 8.145886090983323e-09, + "loss": 1.0416, + "step": 80186 + }, + { + "epoch": 5.92, + "learning_rate": 8.129797636300397e-09, + "loss": 0.8832, + "step": 80187 + }, + { + "epoch": 5.93, + "learning_rate": 8.113725078540135e-09, + "loss": 0.9335, + "step": 80188 + }, + { + "epoch": 5.93, + "learning_rate": 8.097668417729188e-09, + "loss": 1.0094, + "step": 80189 + }, + { + "epoch": 5.93, + "learning_rate": 8.081627653890866e-09, + "loss": 1.0777, + "step": 80190 + }, + { + "epoch": 5.93, + "learning_rate": 8.065602787051818e-09, + "loss": 1.009, + "step": 80191 + }, + { + "epoch": 5.93, + "learning_rate": 8.049593817237578e-09, + "loss": 1.0126, + "step": 80192 + }, + { + "epoch": 5.93, + "learning_rate": 8.03360074447257e-09, + "loss": 1.0076, + "step": 80193 + }, + { + "epoch": 5.93, + "learning_rate": 8.01762356878455e-09, + "loss": 0.9728, + "step": 80194 + }, + { + "epoch": 5.93, + "learning_rate": 8.001662290196832e-09, + "loss": 0.9184, + "step": 80195 + }, + { + "epoch": 5.93, + "learning_rate": 7.985716908734953e-09, + "loss": 0.9603, + "step": 80196 + }, + { + "epoch": 5.93, + "learning_rate": 7.969787424425556e-09, + "loss": 1.0165, + "step": 80197 + }, + { + "epoch": 5.93, + "learning_rate": 7.95387383729307e-09, + "loss": 0.9858, + "step": 80198 + }, + { + "epoch": 5.93, + "learning_rate": 7.937976147363024e-09, + "loss": 0.9986, + "step": 80199 + }, + { + "epoch": 5.93, + "learning_rate": 7.922094354660958e-09, + "loss": 1.1093, + "step": 80200 + }, + { + "epoch": 5.93, + "learning_rate": 7.906228459211295e-09, + "loss": 1.0638, + "step": 80201 + }, + { + "epoch": 5.93, + "learning_rate": 7.890378461040681e-09, + "loss": 1.0405, + "step": 80202 + }, + { + "epoch": 5.93, + "learning_rate": 7.874544360173541e-09, + "loss": 0.8609, + "step": 80203 + }, + { + "epoch": 5.93, + "learning_rate": 7.8587261566343e-09, + "loss": 1.0515, + "step": 80204 + }, + { + "epoch": 5.93, + "learning_rate": 7.842923850448492e-09, + "loss": 1.0219, + "step": 80205 + }, + { + "epoch": 5.93, + "learning_rate": 7.827137441642762e-09, + "loss": 1.0558, + "step": 80206 + }, + { + "epoch": 5.93, + "learning_rate": 7.811366930240427e-09, + "loss": 1.032, + "step": 80207 + }, + { + "epoch": 5.93, + "learning_rate": 7.795612316267021e-09, + "loss": 0.9319, + "step": 80208 + }, + { + "epoch": 5.93, + "learning_rate": 7.779873599749188e-09, + "loss": 0.96, + "step": 80209 + }, + { + "epoch": 5.93, + "learning_rate": 7.764150780709135e-09, + "loss": 0.9149, + "step": 80210 + }, + { + "epoch": 5.93, + "learning_rate": 7.748443859173505e-09, + "loss": 0.9733, + "step": 80211 + }, + { + "epoch": 5.93, + "learning_rate": 7.732752835167833e-09, + "loss": 0.9574, + "step": 80212 + }, + { + "epoch": 5.93, + "learning_rate": 7.717077708716547e-09, + "loss": 1.0441, + "step": 80213 + }, + { + "epoch": 5.93, + "learning_rate": 7.701418479842958e-09, + "loss": 1.134, + "step": 80214 + }, + { + "epoch": 5.93, + "learning_rate": 7.685775148574826e-09, + "loss": 0.9935, + "step": 80215 + }, + { + "epoch": 5.93, + "learning_rate": 7.670147714935462e-09, + "loss": 0.9154, + "step": 80216 + }, + { + "epoch": 5.93, + "learning_rate": 7.65453617894929e-09, + "loss": 0.9454, + "step": 80217 + }, + { + "epoch": 5.93, + "learning_rate": 7.63894054064296e-09, + "loss": 0.948, + "step": 80218 + }, + { + "epoch": 5.93, + "learning_rate": 7.62336080003867e-09, + "loss": 0.9867, + "step": 80219 + }, + { + "epoch": 5.93, + "learning_rate": 7.607796957163071e-09, + "loss": 1.0566, + "step": 80220 + }, + { + "epoch": 5.93, + "learning_rate": 7.592249012041698e-09, + "loss": 0.9098, + "step": 80221 + }, + { + "epoch": 5.93, + "learning_rate": 7.57671696469675e-09, + "loss": 1.0162, + "step": 80222 + }, + { + "epoch": 5.93, + "learning_rate": 7.561200815154879e-09, + "loss": 1.0431, + "step": 80223 + }, + { + "epoch": 5.93, + "learning_rate": 7.545700563439395e-09, + "loss": 0.9337, + "step": 80224 + }, + { + "epoch": 5.93, + "learning_rate": 7.530216209576946e-09, + "loss": 1.0144, + "step": 80225 + }, + { + "epoch": 5.93, + "learning_rate": 7.514747753589735e-09, + "loss": 0.9806, + "step": 80226 + }, + { + "epoch": 5.93, + "learning_rate": 7.499295195504409e-09, + "loss": 0.9632, + "step": 80227 + }, + { + "epoch": 5.93, + "learning_rate": 7.48385853534428e-09, + "loss": 0.9959, + "step": 80228 + }, + { + "epoch": 5.93, + "learning_rate": 7.468437773134884e-09, + "loss": 1.082, + "step": 80229 + }, + { + "epoch": 5.93, + "learning_rate": 7.453032908899538e-09, + "loss": 1.0499, + "step": 80230 + }, + { + "epoch": 5.93, + "learning_rate": 7.437643942663775e-09, + "loss": 1.0112, + "step": 80231 + }, + { + "epoch": 5.93, + "learning_rate": 7.42227087445202e-09, + "loss": 1.0884, + "step": 80232 + }, + { + "epoch": 5.93, + "learning_rate": 7.4069137042886986e-09, + "loss": 0.8869, + "step": 80233 + }, + { + "epoch": 5.93, + "learning_rate": 7.391572432197125e-09, + "loss": 1.0314, + "step": 80234 + }, + { + "epoch": 5.93, + "learning_rate": 7.3762470582028345e-09, + "loss": 0.9717, + "step": 80235 + }, + { + "epoch": 5.93, + "learning_rate": 7.360937582330252e-09, + "loss": 1.0533, + "step": 80236 + }, + { + "epoch": 5.93, + "learning_rate": 7.345644004603803e-09, + "loss": 1.0262, + "step": 80237 + }, + { + "epoch": 5.93, + "learning_rate": 7.3303663250468e-09, + "loss": 1.0289, + "step": 80238 + }, + { + "epoch": 5.93, + "learning_rate": 7.315104543684781e-09, + "loss": 0.9826, + "step": 80239 + }, + { + "epoch": 5.93, + "learning_rate": 7.299858660541059e-09, + "loss": 1.0115, + "step": 80240 + }, + { + "epoch": 5.93, + "learning_rate": 7.28462867564006e-09, + "loss": 0.8903, + "step": 80241 + }, + { + "epoch": 5.93, + "learning_rate": 7.2694145890062074e-09, + "loss": 1.1645, + "step": 80242 + }, + { + "epoch": 5.93, + "learning_rate": 7.254216400663927e-09, + "loss": 1.0225, + "step": 80243 + }, + { + "epoch": 5.93, + "learning_rate": 7.239034110637644e-09, + "loss": 0.8972, + "step": 80244 + }, + { + "epoch": 5.93, + "learning_rate": 7.223867718950672e-09, + "loss": 1.0259, + "step": 80245 + }, + { + "epoch": 5.93, + "learning_rate": 7.208717225627437e-09, + "loss": 1.0936, + "step": 80246 + }, + { + "epoch": 5.93, + "learning_rate": 7.193582630692364e-09, + "loss": 1.0172, + "step": 80247 + }, + { + "epoch": 5.93, + "learning_rate": 7.178463934169877e-09, + "loss": 0.9109, + "step": 80248 + }, + { + "epoch": 5.93, + "learning_rate": 7.163361136082181e-09, + "loss": 0.9635, + "step": 80249 + }, + { + "epoch": 5.93, + "learning_rate": 7.148274236455921e-09, + "loss": 1.06, + "step": 80250 + }, + { + "epoch": 5.93, + "learning_rate": 7.133203235313302e-09, + "loss": 0.9611, + "step": 80251 + }, + { + "epoch": 5.93, + "learning_rate": 7.118148132678748e-09, + "loss": 0.9624, + "step": 80252 + }, + { + "epoch": 5.93, + "learning_rate": 7.103108928575575e-09, + "loss": 0.9589, + "step": 80253 + }, + { + "epoch": 5.93, + "learning_rate": 7.088085623029317e-09, + "loss": 0.9333, + "step": 80254 + }, + { + "epoch": 5.93, + "learning_rate": 7.073078216063289e-09, + "loss": 0.9799, + "step": 80255 + }, + { + "epoch": 5.93, + "learning_rate": 7.058086707699696e-09, + "loss": 0.9066, + "step": 80256 + }, + { + "epoch": 5.93, + "learning_rate": 7.043111097964072e-09, + "loss": 0.9673, + "step": 80257 + }, + { + "epoch": 5.93, + "learning_rate": 7.028151386880844e-09, + "loss": 0.9803, + "step": 80258 + }, + { + "epoch": 5.93, + "learning_rate": 7.013207574472214e-09, + "loss": 1.1113, + "step": 80259 + }, + { + "epoch": 5.93, + "learning_rate": 6.998279660762608e-09, + "loss": 0.9567, + "step": 80260 + }, + { + "epoch": 5.93, + "learning_rate": 6.9833676457753406e-09, + "loss": 0.9528, + "step": 80261 + }, + { + "epoch": 5.93, + "learning_rate": 6.9684715295348375e-09, + "loss": 0.9031, + "step": 80262 + }, + { + "epoch": 5.93, + "learning_rate": 6.953591312064411e-09, + "loss": 1.0109, + "step": 80263 + }, + { + "epoch": 5.93, + "learning_rate": 6.938726993388489e-09, + "loss": 1.0726, + "step": 80264 + }, + { + "epoch": 5.93, + "learning_rate": 6.923878573529275e-09, + "loss": 0.9691, + "step": 80265 + }, + { + "epoch": 5.93, + "learning_rate": 6.909046052512303e-09, + "loss": 0.9592, + "step": 80266 + }, + { + "epoch": 5.93, + "learning_rate": 6.894229430358667e-09, + "loss": 1.0129, + "step": 80267 + }, + { + "epoch": 5.93, + "learning_rate": 6.879428707093905e-09, + "loss": 0.9976, + "step": 80268 + }, + { + "epoch": 5.93, + "learning_rate": 6.8646438827413286e-09, + "loss": 0.9745, + "step": 80269 + }, + { + "epoch": 5.93, + "learning_rate": 6.849874957324254e-09, + "loss": 1.0225, + "step": 80270 + }, + { + "epoch": 5.93, + "learning_rate": 6.835121930865995e-09, + "loss": 0.9848, + "step": 80271 + }, + { + "epoch": 5.93, + "learning_rate": 6.820384803388757e-09, + "loss": 0.9593, + "step": 80272 + }, + { + "epoch": 5.93, + "learning_rate": 6.805663574919186e-09, + "loss": 0.9592, + "step": 80273 + }, + { + "epoch": 5.93, + "learning_rate": 6.790958245477264e-09, + "loss": 0.917, + "step": 80274 + }, + { + "epoch": 5.93, + "learning_rate": 6.776268815088527e-09, + "loss": 0.9688, + "step": 80275 + }, + { + "epoch": 5.93, + "learning_rate": 6.7615952837762896e-09, + "loss": 0.96, + "step": 80276 + }, + { + "epoch": 5.93, + "learning_rate": 6.7469376515627575e-09, + "loss": 0.9891, + "step": 80277 + }, + { + "epoch": 5.93, + "learning_rate": 6.732295918472354e-09, + "loss": 1.0322, + "step": 80278 + }, + { + "epoch": 5.93, + "learning_rate": 6.717670084527283e-09, + "loss": 1.138, + "step": 80279 + }, + { + "epoch": 5.93, + "learning_rate": 6.703060149751972e-09, + "loss": 0.9971, + "step": 80280 + }, + { + "epoch": 5.93, + "learning_rate": 6.6884661141686234e-09, + "loss": 0.9979, + "step": 80281 + }, + { + "epoch": 5.93, + "learning_rate": 6.6738879778016626e-09, + "loss": 1.076, + "step": 80282 + }, + { + "epoch": 5.93, + "learning_rate": 6.6593257406732945e-09, + "loss": 1.0325, + "step": 80283 + }, + { + "epoch": 5.93, + "learning_rate": 6.644779402806833e-09, + "loss": 1.0455, + "step": 80284 + }, + { + "epoch": 5.93, + "learning_rate": 6.630248964225594e-09, + "loss": 1.047, + "step": 80285 + }, + { + "epoch": 5.93, + "learning_rate": 6.615734424951781e-09, + "loss": 0.9428, + "step": 80286 + }, + { + "epoch": 5.93, + "learning_rate": 6.601235785010929e-09, + "loss": 0.9693, + "step": 80287 + }, + { + "epoch": 5.93, + "learning_rate": 6.586753044423022e-09, + "loss": 0.9514, + "step": 80288 + }, + { + "epoch": 5.93, + "learning_rate": 6.572286203213596e-09, + "loss": 1.0733, + "step": 80289 + }, + { + "epoch": 5.93, + "learning_rate": 6.557835261403745e-09, + "loss": 0.998, + "step": 80290 + }, + { + "epoch": 5.93, + "learning_rate": 6.543400219019003e-09, + "loss": 1.0014, + "step": 80291 + }, + { + "epoch": 5.93, + "learning_rate": 6.528981076079355e-09, + "loss": 0.9128, + "step": 80292 + }, + { + "epoch": 5.93, + "learning_rate": 6.514577832609226e-09, + "loss": 1.0707, + "step": 80293 + }, + { + "epoch": 5.93, + "learning_rate": 6.50019048863193e-09, + "loss": 1.1416, + "step": 80294 + }, + { + "epoch": 5.93, + "learning_rate": 6.485819044169672e-09, + "loss": 0.9796, + "step": 80295 + }, + { + "epoch": 5.93, + "learning_rate": 6.471463499245767e-09, + "loss": 0.958, + "step": 80296 + }, + { + "epoch": 5.93, + "learning_rate": 6.457123853882419e-09, + "loss": 0.9314, + "step": 80297 + }, + { + "epoch": 5.93, + "learning_rate": 6.442800108101832e-09, + "loss": 0.8792, + "step": 80298 + }, + { + "epoch": 5.93, + "learning_rate": 6.428492261929542e-09, + "loss": 1.0832, + "step": 80299 + }, + { + "epoch": 5.93, + "learning_rate": 6.414200315385533e-09, + "loss": 1.0015, + "step": 80300 + }, + { + "epoch": 5.93, + "learning_rate": 6.399924268494228e-09, + "loss": 1.0326, + "step": 80301 + }, + { + "epoch": 5.93, + "learning_rate": 6.3856641212767245e-09, + "loss": 0.9204, + "step": 80302 + }, + { + "epoch": 5.93, + "learning_rate": 6.371419873757445e-09, + "loss": 1.1121, + "step": 80303 + }, + { + "epoch": 5.93, + "learning_rate": 6.3571915259574844e-09, + "loss": 0.9878, + "step": 80304 + }, + { + "epoch": 5.93, + "learning_rate": 6.342979077901268e-09, + "loss": 0.9926, + "step": 80305 + }, + { + "epoch": 5.93, + "learning_rate": 6.32878252960989e-09, + "loss": 0.9704, + "step": 80306 + }, + { + "epoch": 5.93, + "learning_rate": 6.314601881105553e-09, + "loss": 0.9686, + "step": 80307 + }, + { + "epoch": 5.93, + "learning_rate": 6.300437132412685e-09, + "loss": 0.9856, + "step": 80308 + }, + { + "epoch": 5.93, + "learning_rate": 6.286288283552377e-09, + "loss": 0.9814, + "step": 80309 + }, + { + "epoch": 5.93, + "learning_rate": 6.272155334547947e-09, + "loss": 0.993, + "step": 80310 + }, + { + "epoch": 5.93, + "learning_rate": 6.258038285421597e-09, + "loss": 1.0883, + "step": 80311 + }, + { + "epoch": 5.93, + "learning_rate": 6.243937136195533e-09, + "loss": 0.9003, + "step": 80312 + }, + { + "epoch": 5.93, + "learning_rate": 6.229851886893068e-09, + "loss": 1.0499, + "step": 80313 + }, + { + "epoch": 5.93, + "learning_rate": 6.2157825375352974e-09, + "loss": 1.0072, + "step": 80314 + }, + { + "epoch": 5.93, + "learning_rate": 6.201729088144426e-09, + "loss": 1.0069, + "step": 80315 + }, + { + "epoch": 5.93, + "learning_rate": 6.1876915387448775e-09, + "loss": 1.003, + "step": 80316 + }, + { + "epoch": 5.93, + "learning_rate": 6.1736698893566375e-09, + "loss": 0.9396, + "step": 80317 + }, + { + "epoch": 5.93, + "learning_rate": 6.159664140004129e-09, + "loss": 0.9278, + "step": 80318 + }, + { + "epoch": 5.93, + "learning_rate": 6.145674290708448e-09, + "loss": 0.9622, + "step": 80319 + }, + { + "epoch": 5.93, + "learning_rate": 6.131700341491798e-09, + "loss": 1.0198, + "step": 80320 + }, + { + "epoch": 5.93, + "learning_rate": 6.117742292376383e-09, + "loss": 0.9683, + "step": 80321 + }, + { + "epoch": 5.93, + "learning_rate": 6.103800143384408e-09, + "loss": 0.9312, + "step": 80322 + }, + { + "epoch": 5.94, + "learning_rate": 6.089873894539189e-09, + "loss": 1.0531, + "step": 80323 + }, + { + "epoch": 5.94, + "learning_rate": 6.075963545861818e-09, + "loss": 0.9915, + "step": 80324 + }, + { + "epoch": 5.94, + "learning_rate": 6.062069097374501e-09, + "loss": 0.9793, + "step": 80325 + }, + { + "epoch": 5.94, + "learning_rate": 6.048190549098332e-09, + "loss": 1.0354, + "step": 80326 + }, + { + "epoch": 5.94, + "learning_rate": 6.034327901057735e-09, + "loss": 1.1104, + "step": 80327 + }, + { + "epoch": 5.94, + "learning_rate": 6.020481153273805e-09, + "loss": 1.0342, + "step": 80328 + }, + { + "epoch": 5.94, + "learning_rate": 6.0066503057676366e-09, + "loss": 0.9392, + "step": 80329 + }, + { + "epoch": 5.94, + "learning_rate": 5.9928353585614326e-09, + "loss": 1.0113, + "step": 80330 + }, + { + "epoch": 5.94, + "learning_rate": 5.97903631167851e-09, + "loss": 0.9282, + "step": 80331 + }, + { + "epoch": 5.94, + "learning_rate": 5.965253165139962e-09, + "loss": 1.0811, + "step": 80332 + }, + { + "epoch": 5.94, + "learning_rate": 5.951485918966882e-09, + "loss": 1.0518, + "step": 80333 + }, + { + "epoch": 5.94, + "learning_rate": 5.937734573182585e-09, + "loss": 0.961, + "step": 80334 + }, + { + "epoch": 5.94, + "learning_rate": 5.9239991278081665e-09, + "loss": 0.9615, + "step": 80335 + }, + { + "epoch": 5.94, + "learning_rate": 5.91027958286583e-09, + "loss": 1.0393, + "step": 80336 + }, + { + "epoch": 5.94, + "learning_rate": 5.896575938376669e-09, + "loss": 1.0178, + "step": 80337 + }, + { + "epoch": 5.94, + "learning_rate": 5.882888194364e-09, + "loss": 0.9956, + "step": 80338 + }, + { + "epoch": 5.94, + "learning_rate": 5.869216350847806e-09, + "loss": 1.0655, + "step": 80339 + }, + { + "epoch": 5.94, + "learning_rate": 5.855560407851402e-09, + "loss": 1.0457, + "step": 80340 + }, + { + "epoch": 5.94, + "learning_rate": 5.8419203653958815e-09, + "loss": 1.0004, + "step": 80341 + }, + { + "epoch": 5.94, + "learning_rate": 5.828296223502339e-09, + "loss": 0.9834, + "step": 80342 + }, + { + "epoch": 5.94, + "learning_rate": 5.8146879821929795e-09, + "loss": 1.0317, + "step": 80343 + }, + { + "epoch": 5.94, + "learning_rate": 5.801095641490006e-09, + "loss": 0.9774, + "step": 80344 + }, + { + "epoch": 5.94, + "learning_rate": 5.787519201414515e-09, + "loss": 1.0595, + "step": 80345 + }, + { + "epoch": 5.94, + "learning_rate": 5.7739586619876e-09, + "loss": 1.0174, + "step": 80346 + }, + { + "epoch": 5.94, + "learning_rate": 5.760414023231464e-09, + "loss": 0.9428, + "step": 80347 + }, + { + "epoch": 5.94, + "learning_rate": 5.746885285167203e-09, + "loss": 0.9065, + "step": 80348 + }, + { + "epoch": 5.94, + "learning_rate": 5.73337244781702e-09, + "loss": 1.0437, + "step": 80349 + }, + { + "epoch": 5.94, + "learning_rate": 5.719875511202011e-09, + "loss": 1.0731, + "step": 80350 + }, + { + "epoch": 5.94, + "learning_rate": 5.706394475343269e-09, + "loss": 0.9268, + "step": 80351 + }, + { + "epoch": 5.94, + "learning_rate": 5.692929340262998e-09, + "loss": 1.0884, + "step": 80352 + }, + { + "epoch": 5.94, + "learning_rate": 5.679480105982293e-09, + "loss": 1.0292, + "step": 80353 + }, + { + "epoch": 5.94, + "learning_rate": 5.666046772523359e-09, + "loss": 0.9549, + "step": 80354 + }, + { + "epoch": 5.94, + "learning_rate": 5.652629339905069e-09, + "loss": 1.1232, + "step": 80355 + }, + { + "epoch": 5.94, + "learning_rate": 5.639227808151848e-09, + "loss": 1.0484, + "step": 80356 + }, + { + "epoch": 5.94, + "learning_rate": 5.62584217728257e-09, + "loss": 0.9811, + "step": 80357 + }, + { + "epoch": 5.94, + "learning_rate": 5.6124724473205495e-09, + "loss": 1.0098, + "step": 80358 + }, + { + "epoch": 5.94, + "learning_rate": 5.59911861828577e-09, + "loss": 0.8844, + "step": 80359 + }, + { + "epoch": 5.94, + "learning_rate": 5.5857806901993275e-09, + "loss": 1.068, + "step": 80360 + }, + { + "epoch": 5.94, + "learning_rate": 5.572458663083424e-09, + "loss": 0.9078, + "step": 80361 + }, + { + "epoch": 5.94, + "learning_rate": 5.559152536958045e-09, + "loss": 0.9202, + "step": 80362 + }, + { + "epoch": 5.94, + "learning_rate": 5.545862311846506e-09, + "loss": 0.9933, + "step": 80363 + }, + { + "epoch": 5.94, + "learning_rate": 5.532587987767679e-09, + "loss": 0.9134, + "step": 80364 + }, + { + "epoch": 5.94, + "learning_rate": 5.519329564742659e-09, + "loss": 0.947, + "step": 80365 + }, + { + "epoch": 5.94, + "learning_rate": 5.506087042794761e-09, + "loss": 1.0534, + "step": 80366 + }, + { + "epoch": 5.94, + "learning_rate": 5.4928604219428585e-09, + "loss": 1.0244, + "step": 80367 + }, + { + "epoch": 5.94, + "learning_rate": 5.4796497022091555e-09, + "loss": 1.0963, + "step": 80368 + }, + { + "epoch": 5.94, + "learning_rate": 5.466454883613637e-09, + "loss": 1.0415, + "step": 80369 + }, + { + "epoch": 5.94, + "learning_rate": 5.453275966178506e-09, + "loss": 0.9946, + "step": 80370 + }, + { + "epoch": 5.94, + "learning_rate": 5.440112949924859e-09, + "loss": 0.9025, + "step": 80371 + }, + { + "epoch": 5.94, + "learning_rate": 5.4269658348726775e-09, + "loss": 1.0137, + "step": 80372 + }, + { + "epoch": 5.94, + "learning_rate": 5.413834621041947e-09, + "loss": 1.055, + "step": 80373 + }, + { + "epoch": 5.94, + "learning_rate": 5.400719308455982e-09, + "loss": 1.0456, + "step": 80374 + }, + { + "epoch": 5.94, + "learning_rate": 5.387619897134766e-09, + "loss": 0.9603, + "step": 80375 + }, + { + "epoch": 5.94, + "learning_rate": 5.374536387098283e-09, + "loss": 1.0119, + "step": 80376 + }, + { + "epoch": 5.94, + "learning_rate": 5.3614687783676286e-09, + "loss": 0.8734, + "step": 80377 + }, + { + "epoch": 5.94, + "learning_rate": 5.348417070965007e-09, + "loss": 1.0534, + "step": 80378 + }, + { + "epoch": 5.94, + "learning_rate": 5.33538126490929e-09, + "loss": 0.8876, + "step": 80379 + }, + { + "epoch": 5.94, + "learning_rate": 5.322361360221573e-09, + "loss": 0.9701, + "step": 80380 + }, + { + "epoch": 5.94, + "learning_rate": 5.30935735692295e-09, + "loss": 0.9439, + "step": 80381 + }, + { + "epoch": 5.94, + "learning_rate": 5.296369255034517e-09, + "loss": 0.9082, + "step": 80382 + }, + { + "epoch": 5.94, + "learning_rate": 5.283397054577366e-09, + "loss": 0.9896, + "step": 80383 + }, + { + "epoch": 5.94, + "learning_rate": 5.270440755570372e-09, + "loss": 1.0301, + "step": 80384 + }, + { + "epoch": 5.94, + "learning_rate": 5.257500358034629e-09, + "loss": 0.9558, + "step": 80385 + }, + { + "epoch": 5.94, + "learning_rate": 5.24457586199234e-09, + "loss": 0.9741, + "step": 80386 + }, + { + "epoch": 5.94, + "learning_rate": 5.2316672674623815e-09, + "loss": 1.0383, + "step": 80387 + }, + { + "epoch": 5.94, + "learning_rate": 5.218774574465846e-09, + "loss": 0.9208, + "step": 80388 + }, + { + "epoch": 5.94, + "learning_rate": 5.205897783023828e-09, + "loss": 0.9419, + "step": 80389 + }, + { + "epoch": 5.94, + "learning_rate": 5.193036893155201e-09, + "loss": 0.9676, + "step": 80390 + }, + { + "epoch": 5.94, + "learning_rate": 5.18019190488217e-09, + "loss": 0.9257, + "step": 80391 + }, + { + "epoch": 5.94, + "learning_rate": 5.1673628182247194e-09, + "loss": 1.0592, + "step": 80392 + }, + { + "epoch": 5.94, + "learning_rate": 5.154549633201722e-09, + "loss": 0.9968, + "step": 80393 + }, + { + "epoch": 5.94, + "learning_rate": 5.141752349835383e-09, + "loss": 1.092, + "step": 80394 + }, + { + "epoch": 5.94, + "learning_rate": 5.128970968146796e-09, + "loss": 1.0383, + "step": 80395 + }, + { + "epoch": 5.94, + "learning_rate": 5.116205488153725e-09, + "loss": 0.8773, + "step": 80396 + }, + { + "epoch": 5.94, + "learning_rate": 5.1034559098783745e-09, + "loss": 1.068, + "step": 80397 + }, + { + "epoch": 5.94, + "learning_rate": 5.090722233339618e-09, + "loss": 1.0049, + "step": 80398 + }, + { + "epoch": 5.94, + "learning_rate": 5.07800445855855e-09, + "loss": 1.0395, + "step": 80399 + }, + { + "epoch": 5.94, + "learning_rate": 5.065302585556264e-09, + "loss": 0.9864, + "step": 80400 + }, + { + "epoch": 5.94, + "learning_rate": 5.052616614351635e-09, + "loss": 0.9168, + "step": 80401 + }, + { + "epoch": 5.94, + "learning_rate": 5.039946544964647e-09, + "loss": 1.0574, + "step": 80402 + }, + { + "epoch": 5.94, + "learning_rate": 5.0272923774175034e-09, + "loss": 0.9664, + "step": 80403 + }, + { + "epoch": 5.94, + "learning_rate": 5.014654111727968e-09, + "loss": 0.9131, + "step": 80404 + }, + { + "epoch": 5.94, + "learning_rate": 5.002031747917135e-09, + "loss": 0.9371, + "step": 80405 + }, + { + "epoch": 5.94, + "learning_rate": 4.98942528600499e-09, + "loss": 1.0751, + "step": 80406 + }, + { + "epoch": 5.94, + "learning_rate": 4.976834726011515e-09, + "loss": 1.0489, + "step": 80407 + }, + { + "epoch": 5.94, + "learning_rate": 4.964260067957805e-09, + "loss": 1.0154, + "step": 80408 + }, + { + "epoch": 5.94, + "learning_rate": 4.951701311862733e-09, + "loss": 0.9937, + "step": 80409 + }, + { + "epoch": 5.94, + "learning_rate": 4.939158457746285e-09, + "loss": 1.0112, + "step": 80410 + }, + { + "epoch": 5.94, + "learning_rate": 4.926631505628443e-09, + "loss": 1.0086, + "step": 80411 + }, + { + "epoch": 5.94, + "learning_rate": 4.914120455529192e-09, + "loss": 0.9388, + "step": 80412 + }, + { + "epoch": 5.94, + "learning_rate": 4.9016253074685165e-09, + "loss": 0.9479, + "step": 80413 + }, + { + "epoch": 5.94, + "learning_rate": 4.8891460614663985e-09, + "loss": 0.8998, + "step": 80414 + }, + { + "epoch": 5.94, + "learning_rate": 4.876682717542824e-09, + "loss": 1.0865, + "step": 80415 + }, + { + "epoch": 5.94, + "learning_rate": 4.864235275717777e-09, + "loss": 0.9767, + "step": 80416 + }, + { + "epoch": 5.94, + "learning_rate": 4.85180373601013e-09, + "loss": 0.9765, + "step": 80417 + }, + { + "epoch": 5.94, + "learning_rate": 4.839388098440978e-09, + "loss": 0.9163, + "step": 80418 + }, + { + "epoch": 5.94, + "learning_rate": 4.826988363029194e-09, + "loss": 1.0287, + "step": 80419 + }, + { + "epoch": 5.94, + "learning_rate": 4.814604529793654e-09, + "loss": 0.9749, + "step": 80420 + }, + { + "epoch": 5.94, + "learning_rate": 4.80223659875656e-09, + "loss": 0.9003, + "step": 80421 + }, + { + "epoch": 5.94, + "learning_rate": 4.789884569934566e-09, + "loss": 1.009, + "step": 80422 + }, + { + "epoch": 5.94, + "learning_rate": 4.777548443349877e-09, + "loss": 1.0695, + "step": 80423 + }, + { + "epoch": 5.94, + "learning_rate": 4.765228219021367e-09, + "loss": 1.0271, + "step": 80424 + }, + { + "epoch": 5.94, + "learning_rate": 4.752923896967909e-09, + "loss": 0.8894, + "step": 80425 + }, + { + "epoch": 5.94, + "learning_rate": 4.740635477209487e-09, + "loss": 1.0505, + "step": 80426 + }, + { + "epoch": 5.94, + "learning_rate": 4.7283629597649764e-09, + "loss": 0.9948, + "step": 80427 + }, + { + "epoch": 5.94, + "learning_rate": 4.716106344655469e-09, + "loss": 1.1448, + "step": 80428 + }, + { + "epoch": 5.94, + "learning_rate": 4.70386563189984e-09, + "loss": 0.9635, + "step": 80429 + }, + { + "epoch": 5.94, + "learning_rate": 4.6916408215169625e-09, + "loss": 1.0302, + "step": 80430 + }, + { + "epoch": 5.94, + "learning_rate": 4.679431913526822e-09, + "loss": 1.0372, + "step": 80431 + }, + { + "epoch": 5.94, + "learning_rate": 4.6672389079482906e-09, + "loss": 0.9505, + "step": 80432 + }, + { + "epoch": 5.94, + "learning_rate": 4.655061804801353e-09, + "loss": 1.1241, + "step": 80433 + }, + { + "epoch": 5.94, + "learning_rate": 4.642900604105993e-09, + "loss": 0.9786, + "step": 80434 + }, + { + "epoch": 5.94, + "learning_rate": 4.6307553058799745e-09, + "loss": 0.9922, + "step": 80435 + }, + { + "epoch": 5.94, + "learning_rate": 4.618625910143282e-09, + "loss": 1.033, + "step": 80436 + }, + { + "epoch": 5.94, + "learning_rate": 4.6065124169158985e-09, + "loss": 1.0551, + "step": 80437 + }, + { + "epoch": 5.94, + "learning_rate": 4.594414826215587e-09, + "loss": 0.9626, + "step": 80438 + }, + { + "epoch": 5.94, + "learning_rate": 4.582333138063444e-09, + "loss": 0.9554, + "step": 80439 + }, + { + "epoch": 5.94, + "learning_rate": 4.570267352478341e-09, + "loss": 0.9867, + "step": 80440 + }, + { + "epoch": 5.94, + "learning_rate": 4.558217469478044e-09, + "loss": 0.988, + "step": 80441 + }, + { + "epoch": 5.94, + "learning_rate": 4.546183489082534e-09, + "loss": 0.7896, + "step": 80442 + }, + { + "epoch": 5.94, + "learning_rate": 4.534165411310687e-09, + "loss": 1.0303, + "step": 80443 + }, + { + "epoch": 5.94, + "learning_rate": 4.522163236183596e-09, + "loss": 0.9375, + "step": 80444 + }, + { + "epoch": 5.94, + "learning_rate": 4.510176963716806e-09, + "loss": 1.0465, + "step": 80445 + }, + { + "epoch": 5.94, + "learning_rate": 4.498206593932519e-09, + "loss": 0.9998, + "step": 80446 + }, + { + "epoch": 5.94, + "learning_rate": 4.486252126848501e-09, + "loss": 0.9569, + "step": 80447 + }, + { + "epoch": 5.94, + "learning_rate": 4.474313562483623e-09, + "loss": 1.0329, + "step": 80448 + }, + { + "epoch": 5.94, + "learning_rate": 4.462390900856761e-09, + "loss": 0.9337, + "step": 80449 + }, + { + "epoch": 5.94, + "learning_rate": 4.450484141987899e-09, + "loss": 0.9042, + "step": 80450 + }, + { + "epoch": 5.94, + "learning_rate": 4.438593285894799e-09, + "loss": 1.1091, + "step": 80451 + }, + { + "epoch": 5.94, + "learning_rate": 4.426718332596336e-09, + "loss": 0.8968, + "step": 80452 + }, + { + "epoch": 5.94, + "learning_rate": 4.414859282112494e-09, + "loss": 0.9609, + "step": 80453 + }, + { + "epoch": 5.94, + "learning_rate": 4.403016134462146e-09, + "loss": 0.9435, + "step": 80454 + }, + { + "epoch": 5.94, + "learning_rate": 4.3911888896630565e-09, + "loss": 1.0117, + "step": 80455 + }, + { + "epoch": 5.94, + "learning_rate": 4.379377547735209e-09, + "loss": 1.0098, + "step": 80456 + }, + { + "epoch": 5.94, + "learning_rate": 4.367582108696366e-09, + "loss": 0.9224, + "step": 80457 + }, + { + "epoch": 5.94, + "learning_rate": 4.355802572565404e-09, + "loss": 0.8982, + "step": 80458 + }, + { + "epoch": 5.95, + "learning_rate": 4.344038939361195e-09, + "loss": 0.9927, + "step": 80459 + }, + { + "epoch": 5.95, + "learning_rate": 4.332291209102613e-09, + "loss": 1.0109, + "step": 80460 + }, + { + "epoch": 5.95, + "learning_rate": 4.320559381808531e-09, + "loss": 0.9918, + "step": 80461 + }, + { + "epoch": 5.95, + "learning_rate": 4.3088434574978244e-09, + "loss": 1.0614, + "step": 80462 + }, + { + "epoch": 5.95, + "learning_rate": 4.297143436188256e-09, + "loss": 0.9411, + "step": 80463 + }, + { + "epoch": 5.95, + "learning_rate": 4.28545931789981e-09, + "loss": 0.9252, + "step": 80464 + }, + { + "epoch": 5.95, + "learning_rate": 4.273791102649139e-09, + "loss": 1.0413, + "step": 80465 + }, + { + "epoch": 5.95, + "learning_rate": 4.2621387904562274e-09, + "loss": 1.0036, + "step": 80466 + }, + { + "epoch": 5.95, + "learning_rate": 4.250502381338839e-09, + "loss": 0.8657, + "step": 80467 + }, + { + "epoch": 5.95, + "learning_rate": 4.238881875316958e-09, + "loss": 1.0382, + "step": 80468 + }, + { + "epoch": 5.95, + "learning_rate": 4.227277272407238e-09, + "loss": 0.9738, + "step": 80469 + }, + { + "epoch": 5.95, + "learning_rate": 4.215688572628551e-09, + "loss": 0.8652, + "step": 80470 + }, + { + "epoch": 5.95, + "learning_rate": 4.204115775999773e-09, + "loss": 1.0317, + "step": 80471 + }, + { + "epoch": 5.95, + "learning_rate": 4.192558882539776e-09, + "loss": 0.9832, + "step": 80472 + }, + { + "epoch": 5.95, + "learning_rate": 4.1810178922663256e-09, + "loss": 1.0128, + "step": 80473 + }, + { + "epoch": 5.95, + "learning_rate": 4.169492805197184e-09, + "loss": 1.0347, + "step": 80474 + }, + { + "epoch": 5.95, + "learning_rate": 4.157983621352335e-09, + "loss": 0.9428, + "step": 80475 + }, + { + "epoch": 5.95, + "learning_rate": 4.146490340748433e-09, + "loss": 0.9645, + "step": 80476 + }, + { + "epoch": 5.95, + "learning_rate": 4.1350129634043504e-09, + "loss": 0.9867, + "step": 80477 + }, + { + "epoch": 5.95, + "learning_rate": 4.123551489338962e-09, + "loss": 1.0311, + "step": 80478 + }, + { + "epoch": 5.95, + "learning_rate": 4.112105918570031e-09, + "loss": 1.0269, + "step": 80479 + }, + { + "epoch": 5.95, + "learning_rate": 4.100676251115321e-09, + "loss": 0.9298, + "step": 80480 + }, + { + "epoch": 5.95, + "learning_rate": 4.089262486993706e-09, + "loss": 0.961, + "step": 80481 + }, + { + "epoch": 5.95, + "learning_rate": 4.077864626222949e-09, + "loss": 0.9724, + "step": 80482 + }, + { + "epoch": 5.95, + "learning_rate": 4.066482668820815e-09, + "loss": 1.1412, + "step": 80483 + }, + { + "epoch": 5.95, + "learning_rate": 4.055116614807286e-09, + "loss": 1.0483, + "step": 80484 + }, + { + "epoch": 5.95, + "learning_rate": 4.043766464197907e-09, + "loss": 1.0116, + "step": 80485 + }, + { + "epoch": 5.95, + "learning_rate": 4.0324322170115506e-09, + "loss": 1.0601, + "step": 80486 + }, + { + "epoch": 5.95, + "learning_rate": 4.021113873267091e-09, + "loss": 1.1647, + "step": 80487 + }, + { + "epoch": 5.95, + "learning_rate": 4.009811432982291e-09, + "loss": 1.0872, + "step": 80488 + }, + { + "epoch": 5.95, + "learning_rate": 3.998524896174916e-09, + "loss": 1.0507, + "step": 80489 + }, + { + "epoch": 5.95, + "learning_rate": 3.9872542628627276e-09, + "loss": 1.0125, + "step": 80490 + }, + { + "epoch": 5.95, + "learning_rate": 3.97599953306349e-09, + "loss": 0.9766, + "step": 80491 + }, + { + "epoch": 5.95, + "learning_rate": 3.964760706794968e-09, + "loss": 1.04, + "step": 80492 + }, + { + "epoch": 5.95, + "learning_rate": 3.953537784076034e-09, + "loss": 1.0719, + "step": 80493 + }, + { + "epoch": 5.95, + "learning_rate": 3.942330764924451e-09, + "loss": 0.8981, + "step": 80494 + }, + { + "epoch": 5.95, + "learning_rate": 3.931139649356874e-09, + "loss": 1.0486, + "step": 80495 + }, + { + "epoch": 5.95, + "learning_rate": 3.919964437392177e-09, + "loss": 1.0162, + "step": 80496 + }, + { + "epoch": 5.95, + "learning_rate": 3.908805129048121e-09, + "loss": 1.0064, + "step": 80497 + }, + { + "epoch": 5.95, + "learning_rate": 3.897661724341362e-09, + "loss": 1.0618, + "step": 80498 + }, + { + "epoch": 5.95, + "learning_rate": 3.886534223290772e-09, + "loss": 1.0806, + "step": 80499 + }, + { + "epoch": 5.95, + "learning_rate": 3.875422625914116e-09, + "loss": 0.8997, + "step": 80500 + }, + { + "epoch": 5.95, + "learning_rate": 3.864326932228046e-09, + "loss": 0.9396, + "step": 80501 + }, + { + "epoch": 5.95, + "learning_rate": 3.853247142251437e-09, + "loss": 1.1198, + "step": 80502 + }, + { + "epoch": 5.95, + "learning_rate": 3.842183256000942e-09, + "loss": 1.0292, + "step": 80503 + }, + { + "epoch": 5.95, + "learning_rate": 3.831135273494324e-09, + "loss": 0.9663, + "step": 80504 + }, + { + "epoch": 5.95, + "learning_rate": 3.820103194749347e-09, + "loss": 1.0133, + "step": 80505 + }, + { + "epoch": 5.95, + "learning_rate": 3.809087019783775e-09, + "loss": 1.0156, + "step": 80506 + }, + { + "epoch": 5.95, + "learning_rate": 3.7980867486142606e-09, + "loss": 0.9079, + "step": 80507 + }, + { + "epoch": 5.95, + "learning_rate": 3.787102381259677e-09, + "loss": 1.0054, + "step": 80508 + }, + { + "epoch": 5.95, + "learning_rate": 3.7761339177366795e-09, + "loss": 0.9444, + "step": 80509 + }, + { + "epoch": 5.95, + "learning_rate": 3.76518135806303e-09, + "loss": 1.0301, + "step": 80510 + }, + { + "epoch": 5.95, + "learning_rate": 3.754244702255383e-09, + "loss": 0.9236, + "step": 80511 + }, + { + "epoch": 5.95, + "learning_rate": 3.743323950332612e-09, + "loss": 1.0105, + "step": 80512 + }, + { + "epoch": 5.95, + "learning_rate": 3.732419102310259e-09, + "loss": 1.1143, + "step": 80513 + }, + { + "epoch": 5.95, + "learning_rate": 3.7215301582071985e-09, + "loss": 1.0568, + "step": 80514 + }, + { + "epoch": 5.95, + "learning_rate": 3.710657118040084e-09, + "loss": 1.0533, + "step": 80515 + }, + { + "epoch": 5.95, + "learning_rate": 3.6997999818255693e-09, + "loss": 1.0201, + "step": 80516 + }, + { + "epoch": 5.95, + "learning_rate": 3.6889587495825274e-09, + "loss": 0.9916, + "step": 80517 + }, + { + "epoch": 5.95, + "learning_rate": 3.678133421327612e-09, + "loss": 1.078, + "step": 80518 + }, + { + "epoch": 5.95, + "learning_rate": 3.6673239970763663e-09, + "loss": 1.0043, + "step": 80519 + }, + { + "epoch": 5.95, + "learning_rate": 3.656530476848774e-09, + "loss": 0.9408, + "step": 80520 + }, + { + "epoch": 5.95, + "learning_rate": 3.6457528606592684e-09, + "loss": 1.0194, + "step": 80521 + }, + { + "epoch": 5.95, + "learning_rate": 3.634991148527833e-09, + "loss": 1.0329, + "step": 80522 + }, + { + "epoch": 5.95, + "learning_rate": 3.6242453404689015e-09, + "loss": 1.1137, + "step": 80523 + }, + { + "epoch": 5.95, + "learning_rate": 3.613515436501347e-09, + "loss": 0.9701, + "step": 80524 + }, + { + "epoch": 5.95, + "learning_rate": 3.6028014366418227e-09, + "loss": 0.8935, + "step": 80525 + }, + { + "epoch": 5.95, + "learning_rate": 3.592103340906983e-09, + "loss": 0.95, + "step": 80526 + }, + { + "epoch": 5.95, + "learning_rate": 3.58142114931459e-09, + "loss": 1.0937, + "step": 80527 + }, + { + "epoch": 5.95, + "learning_rate": 3.5707548618801878e-09, + "loss": 1.0405, + "step": 80528 + }, + { + "epoch": 5.95, + "learning_rate": 3.56010447862154e-09, + "loss": 0.9958, + "step": 80529 + }, + { + "epoch": 5.95, + "learning_rate": 3.5494699995564097e-09, + "loss": 1.0026, + "step": 80530 + }, + { + "epoch": 5.95, + "learning_rate": 3.5388514247014503e-09, + "loss": 0.9234, + "step": 80531 + }, + { + "epoch": 5.95, + "learning_rate": 3.5282487540722053e-09, + "loss": 0.9639, + "step": 80532 + }, + { + "epoch": 5.95, + "learning_rate": 3.5176619876875484e-09, + "loss": 1.0309, + "step": 80533 + }, + { + "epoch": 5.95, + "learning_rate": 3.507091125561912e-09, + "loss": 1.0238, + "step": 80534 + }, + { + "epoch": 5.95, + "learning_rate": 3.496536167715281e-09, + "loss": 1.0363, + "step": 80535 + }, + { + "epoch": 5.95, + "learning_rate": 3.4859971141609773e-09, + "loss": 1.0059, + "step": 80536 + }, + { + "epoch": 5.95, + "learning_rate": 3.475473964917875e-09, + "loss": 1.0241, + "step": 80537 + }, + { + "epoch": 5.95, + "learning_rate": 3.4649667200026273e-09, + "loss": 0.982, + "step": 80538 + }, + { + "epoch": 5.95, + "learning_rate": 3.4544753794318873e-09, + "loss": 1.0965, + "step": 80539 + }, + { + "epoch": 5.95, + "learning_rate": 3.443999943221199e-09, + "loss": 0.979, + "step": 80540 + }, + { + "epoch": 5.95, + "learning_rate": 3.433540411388325e-09, + "loss": 1.026, + "step": 80541 + }, + { + "epoch": 5.95, + "learning_rate": 3.42309678394992e-09, + "loss": 1.0607, + "step": 80542 + }, + { + "epoch": 5.95, + "learning_rate": 3.412669060921525e-09, + "loss": 1.044, + "step": 80543 + }, + { + "epoch": 5.95, + "learning_rate": 3.402257242322016e-09, + "loss": 1.0727, + "step": 80544 + }, + { + "epoch": 5.95, + "learning_rate": 3.391861328164714e-09, + "loss": 1.0526, + "step": 80545 + }, + { + "epoch": 5.95, + "learning_rate": 3.3814813184696037e-09, + "loss": 0.9157, + "step": 80546 + }, + { + "epoch": 5.95, + "learning_rate": 3.371117213250008e-09, + "loss": 0.9329, + "step": 80547 + }, + { + "epoch": 5.95, + "learning_rate": 3.3607690125248006e-09, + "loss": 0.7936, + "step": 80548 + }, + { + "epoch": 5.95, + "learning_rate": 3.3504367163095243e-09, + "loss": 0.852, + "step": 80549 + }, + { + "epoch": 5.95, + "learning_rate": 3.3401203246208323e-09, + "loss": 1.0043, + "step": 80550 + }, + { + "epoch": 5.95, + "learning_rate": 3.3298198374753787e-09, + "loss": 0.9741, + "step": 80551 + }, + { + "epoch": 5.95, + "learning_rate": 3.319535254888706e-09, + "loss": 1.1312, + "step": 80552 + }, + { + "epoch": 5.95, + "learning_rate": 3.309266576877468e-09, + "loss": 1.0259, + "step": 80553 + }, + { + "epoch": 5.95, + "learning_rate": 3.2990138034583174e-09, + "loss": 0.8402, + "step": 80554 + }, + { + "epoch": 5.95, + "learning_rate": 3.288776934646798e-09, + "loss": 1.0442, + "step": 80555 + }, + { + "epoch": 5.95, + "learning_rate": 3.278555970460673e-09, + "loss": 1.0183, + "step": 80556 + }, + { + "epoch": 5.95, + "learning_rate": 3.2683509109143753e-09, + "loss": 0.8945, + "step": 80557 + }, + { + "epoch": 5.95, + "learning_rate": 3.2581617560256683e-09, + "loss": 0.9367, + "step": 80558 + }, + { + "epoch": 5.95, + "learning_rate": 3.2479885058100955e-09, + "loss": 1.0014, + "step": 80559 + }, + { + "epoch": 5.95, + "learning_rate": 3.2378311602843106e-09, + "loss": 1.0596, + "step": 80560 + }, + { + "epoch": 5.95, + "learning_rate": 3.2276897194638555e-09, + "loss": 1.0011, + "step": 80561 + }, + { + "epoch": 5.95, + "learning_rate": 3.217564183365385e-09, + "loss": 1.0117, + "step": 80562 + }, + { + "epoch": 5.95, + "learning_rate": 3.207454552004441e-09, + "loss": 1.0421, + "step": 80563 + }, + { + "epoch": 5.95, + "learning_rate": 3.1973608253976775e-09, + "loss": 0.9669, + "step": 80564 + }, + { + "epoch": 5.95, + "learning_rate": 3.1872830035606373e-09, + "loss": 0.8548, + "step": 80565 + }, + { + "epoch": 5.95, + "learning_rate": 3.1772210865088636e-09, + "loss": 0.9747, + "step": 80566 + }, + { + "epoch": 5.95, + "learning_rate": 3.1671750742601203e-09, + "loss": 0.9831, + "step": 80567 + }, + { + "epoch": 5.95, + "learning_rate": 3.15714496682884e-09, + "loss": 0.8971, + "step": 80568 + }, + { + "epoch": 5.95, + "learning_rate": 3.1471307642316762e-09, + "loss": 1.0147, + "step": 80569 + }, + { + "epoch": 5.95, + "learning_rate": 3.1371324664852822e-09, + "loss": 0.9184, + "step": 80570 + }, + { + "epoch": 5.95, + "learning_rate": 3.1271500736029804e-09, + "loss": 1.0911, + "step": 80571 + }, + { + "epoch": 5.95, + "learning_rate": 3.117183585603645e-09, + "loss": 0.9063, + "step": 80572 + }, + { + "epoch": 5.95, + "learning_rate": 3.1072330025017085e-09, + "loss": 0.9843, + "step": 80573 + }, + { + "epoch": 5.95, + "learning_rate": 3.097298324312714e-09, + "loss": 1.098, + "step": 80574 + }, + { + "epoch": 5.95, + "learning_rate": 3.0873795510533157e-09, + "loss": 0.9847, + "step": 80575 + }, + { + "epoch": 5.95, + "learning_rate": 3.0774766827379456e-09, + "loss": 1.0505, + "step": 80576 + }, + { + "epoch": 5.95, + "learning_rate": 3.0675897193843673e-09, + "loss": 0.9251, + "step": 80577 + }, + { + "epoch": 5.95, + "learning_rate": 3.057718661007014e-09, + "loss": 0.9291, + "step": 80578 + }, + { + "epoch": 5.95, + "learning_rate": 3.0478635076214293e-09, + "loss": 0.9559, + "step": 80579 + }, + { + "epoch": 5.95, + "learning_rate": 3.038024259243155e-09, + "loss": 0.9801, + "step": 80580 + }, + { + "epoch": 5.95, + "learning_rate": 3.028200915888846e-09, + "loss": 0.9146, + "step": 80581 + }, + { + "epoch": 5.95, + "learning_rate": 3.018393477574044e-09, + "loss": 0.9861, + "step": 80582 + }, + { + "epoch": 5.95, + "learning_rate": 3.0086019443142933e-09, + "loss": 0.9814, + "step": 80583 + }, + { + "epoch": 5.95, + "learning_rate": 2.9988263161240263e-09, + "loss": 0.8713, + "step": 80584 + }, + { + "epoch": 5.95, + "learning_rate": 2.989066593019896e-09, + "loss": 1.0791, + "step": 80585 + }, + { + "epoch": 5.95, + "learning_rate": 2.9793227750163356e-09, + "loss": 0.9146, + "step": 80586 + }, + { + "epoch": 5.95, + "learning_rate": 2.9695948621299988e-09, + "loss": 0.9575, + "step": 80587 + }, + { + "epoch": 5.95, + "learning_rate": 2.959882854376428e-09, + "loss": 1.0379, + "step": 80588 + }, + { + "epoch": 5.95, + "learning_rate": 2.950186751771167e-09, + "loss": 1.0569, + "step": 80589 + }, + { + "epoch": 5.95, + "learning_rate": 2.940506554328648e-09, + "loss": 1.0662, + "step": 80590 + }, + { + "epoch": 5.95, + "learning_rate": 2.930842262064415e-09, + "loss": 1.0269, + "step": 80591 + }, + { + "epoch": 5.95, + "learning_rate": 2.9211938749940107e-09, + "loss": 1.1643, + "step": 80592 + }, + { + "epoch": 5.95, + "learning_rate": 2.911561393132978e-09, + "loss": 0.9799, + "step": 80593 + }, + { + "epoch": 5.96, + "learning_rate": 2.90194481649686e-09, + "loss": 1.0542, + "step": 80594 + }, + { + "epoch": 5.96, + "learning_rate": 2.8923441451012003e-09, + "loss": 1.0504, + "step": 80595 + }, + { + "epoch": 5.96, + "learning_rate": 2.8827593789604315e-09, + "loss": 1.0083, + "step": 80596 + }, + { + "epoch": 5.96, + "learning_rate": 2.8731905180900965e-09, + "loss": 1.0312, + "step": 80597 + }, + { + "epoch": 5.96, + "learning_rate": 2.863637562505739e-09, + "loss": 0.9832, + "step": 80598 + }, + { + "epoch": 5.96, + "learning_rate": 2.8541005122217913e-09, + "loss": 1.0234, + "step": 80599 + }, + { + "epoch": 5.96, + "learning_rate": 2.8445793672537968e-09, + "loss": 1.0596, + "step": 80600 + }, + { + "epoch": 5.96, + "learning_rate": 2.835074127617299e-09, + "loss": 1.002, + "step": 80601 + }, + { + "epoch": 5.96, + "learning_rate": 2.8255847933278404e-09, + "loss": 1.0784, + "step": 80602 + }, + { + "epoch": 5.96, + "learning_rate": 2.816111364398744e-09, + "loss": 1.1237, + "step": 80603 + }, + { + "epoch": 5.96, + "learning_rate": 2.8066538408466625e-09, + "loss": 1.0758, + "step": 80604 + }, + { + "epoch": 5.96, + "learning_rate": 2.79721222268714e-09, + "loss": 1.0366, + "step": 80605 + }, + { + "epoch": 5.96, + "learning_rate": 2.7877865099323886e-09, + "loss": 1.0539, + "step": 80606 + }, + { + "epoch": 5.96, + "learning_rate": 2.778376702600172e-09, + "loss": 0.9465, + "step": 80607 + }, + { + "epoch": 5.96, + "learning_rate": 2.7689828007049222e-09, + "loss": 0.9912, + "step": 80608 + }, + { + "epoch": 5.96, + "learning_rate": 2.7596048042610734e-09, + "loss": 1.0153, + "step": 80609 + }, + { + "epoch": 5.96, + "learning_rate": 2.7502427132830576e-09, + "loss": 1.0133, + "step": 80610 + }, + { + "epoch": 5.96, + "learning_rate": 2.7408965277864187e-09, + "loss": 0.942, + "step": 80611 + }, + { + "epoch": 5.96, + "learning_rate": 2.7315662477866987e-09, + "loss": 0.9229, + "step": 80612 + }, + { + "epoch": 5.96, + "learning_rate": 2.7222518732972214e-09, + "loss": 1.0281, + "step": 80613 + }, + { + "epoch": 5.96, + "learning_rate": 2.7129534043335294e-09, + "loss": 1.0498, + "step": 80614 + }, + { + "epoch": 5.96, + "learning_rate": 2.7036708409111656e-09, + "loss": 1.0073, + "step": 80615 + }, + { + "epoch": 5.96, + "learning_rate": 2.6944041830434532e-09, + "loss": 0.8358, + "step": 80616 + }, + { + "epoch": 5.96, + "learning_rate": 2.685153430747045e-09, + "loss": 1.0548, + "step": 80617 + }, + { + "epoch": 5.96, + "learning_rate": 2.675918584034154e-09, + "loss": 1.0432, + "step": 80618 + }, + { + "epoch": 5.96, + "learning_rate": 2.666699642922543e-09, + "loss": 1.1835, + "step": 80619 + }, + { + "epoch": 5.96, + "learning_rate": 2.6574966074233154e-09, + "loss": 0.8058, + "step": 80620 + }, + { + "epoch": 5.96, + "learning_rate": 2.648309477554234e-09, + "loss": 1.0864, + "step": 80621 + }, + { + "epoch": 5.96, + "learning_rate": 2.6391382533286214e-09, + "loss": 0.8975, + "step": 80622 + }, + { + "epoch": 5.96, + "learning_rate": 2.6299829347598004e-09, + "loss": 0.9783, + "step": 80623 + }, + { + "epoch": 5.96, + "learning_rate": 2.6208435218644248e-09, + "loss": 1.0949, + "step": 80624 + }, + { + "epoch": 5.96, + "learning_rate": 2.6117200146569268e-09, + "loss": 1.0932, + "step": 80625 + }, + { + "epoch": 5.96, + "learning_rate": 2.6026124131506294e-09, + "loss": 0.9152, + "step": 80626 + }, + { + "epoch": 5.96, + "learning_rate": 2.5935207173599652e-09, + "loss": 1.0592, + "step": 80627 + }, + { + "epoch": 5.96, + "learning_rate": 2.584444927300478e-09, + "loss": 0.9327, + "step": 80628 + }, + { + "epoch": 5.96, + "learning_rate": 2.5753850429866e-09, + "loss": 0.9498, + "step": 80629 + }, + { + "epoch": 5.96, + "learning_rate": 2.5663410644316543e-09, + "loss": 1.0482, + "step": 80630 + }, + { + "epoch": 5.96, + "learning_rate": 2.557312991651184e-09, + "loss": 0.9485, + "step": 80631 + }, + { + "epoch": 5.96, + "learning_rate": 2.5483008246585115e-09, + "loss": 0.8413, + "step": 80632 + }, + { + "epoch": 5.96, + "learning_rate": 2.5393045634680703e-09, + "loss": 0.969, + "step": 80633 + }, + { + "epoch": 5.96, + "learning_rate": 2.5303242080954025e-09, + "loss": 1.0687, + "step": 80634 + }, + { + "epoch": 5.96, + "learning_rate": 2.5213597585527216e-09, + "loss": 1.0414, + "step": 80635 + }, + { + "epoch": 5.96, + "learning_rate": 2.5124112148566804e-09, + "loss": 1.017, + "step": 80636 + }, + { + "epoch": 5.96, + "learning_rate": 2.5034785770206017e-09, + "loss": 0.983, + "step": 80637 + }, + { + "epoch": 5.96, + "learning_rate": 2.494561845057808e-09, + "loss": 0.9085, + "step": 80638 + }, + { + "epoch": 5.96, + "learning_rate": 2.4856610189827323e-09, + "loss": 0.9209, + "step": 80639 + }, + { + "epoch": 5.96, + "learning_rate": 2.4767760988098077e-09, + "loss": 0.8303, + "step": 80640 + }, + { + "epoch": 5.96, + "learning_rate": 2.467907084554577e-09, + "loss": 1.0551, + "step": 80641 + }, + { + "epoch": 5.96, + "learning_rate": 2.459053976229253e-09, + "loss": 1.0256, + "step": 80642 + }, + { + "epoch": 5.96, + "learning_rate": 2.4502167738482685e-09, + "loss": 0.9914, + "step": 80643 + }, + { + "epoch": 5.96, + "learning_rate": 2.4413954774260563e-09, + "loss": 0.9226, + "step": 80644 + }, + { + "epoch": 5.96, + "learning_rate": 2.432590086975939e-09, + "loss": 1.073, + "step": 80645 + }, + { + "epoch": 5.96, + "learning_rate": 2.42380060251346e-09, + "loss": 0.9907, + "step": 80646 + }, + { + "epoch": 5.96, + "learning_rate": 2.4150270240519414e-09, + "loss": 0.895, + "step": 80647 + }, + { + "epoch": 5.96, + "learning_rate": 2.4062693516035963e-09, + "loss": 0.8483, + "step": 80648 + }, + { + "epoch": 5.96, + "learning_rate": 2.3975275851850775e-09, + "loss": 1.0104, + "step": 80649 + }, + { + "epoch": 5.96, + "learning_rate": 2.3888017248085983e-09, + "loss": 0.9897, + "step": 80650 + }, + { + "epoch": 5.96, + "learning_rate": 2.380091770488591e-09, + "loss": 1.018, + "step": 80651 + }, + { + "epoch": 5.96, + "learning_rate": 2.3713977222394878e-09, + "loss": 1.0076, + "step": 80652 + }, + { + "epoch": 5.96, + "learning_rate": 2.3627195800735025e-09, + "loss": 0.9696, + "step": 80653 + }, + { + "epoch": 5.96, + "learning_rate": 2.3540573440050673e-09, + "loss": 1.0742, + "step": 80654 + }, + { + "epoch": 5.96, + "learning_rate": 2.3454110140497255e-09, + "loss": 0.9655, + "step": 80655 + }, + { + "epoch": 5.96, + "learning_rate": 2.336780590218579e-09, + "loss": 1.0398, + "step": 80656 + }, + { + "epoch": 5.96, + "learning_rate": 2.3281660725271715e-09, + "loss": 1.0333, + "step": 80657 + }, + { + "epoch": 5.96, + "learning_rate": 2.3195674609877152e-09, + "loss": 1.1459, + "step": 80658 + }, + { + "epoch": 5.96, + "learning_rate": 2.310984755615753e-09, + "loss": 0.9436, + "step": 80659 + }, + { + "epoch": 5.96, + "learning_rate": 2.3024179564234973e-09, + "loss": 1.192, + "step": 80660 + }, + { + "epoch": 5.96, + "learning_rate": 2.293867063425381e-09, + "loss": 0.9711, + "step": 80661 + }, + { + "epoch": 5.96, + "learning_rate": 2.2853320766347275e-09, + "loss": 1.0066, + "step": 80662 + }, + { + "epoch": 5.96, + "learning_rate": 2.276812996064859e-09, + "loss": 1.0341, + "step": 80663 + }, + { + "epoch": 5.96, + "learning_rate": 2.268309821729098e-09, + "loss": 1.0532, + "step": 80664 + }, + { + "epoch": 5.96, + "learning_rate": 2.259822553641877e-09, + "loss": 1.0237, + "step": 80665 + }, + { + "epoch": 5.96, + "learning_rate": 2.25135119181652e-09, + "loss": 1.0734, + "step": 80666 + }, + { + "epoch": 5.96, + "learning_rate": 2.2428957362663485e-09, + "loss": 1.132, + "step": 80667 + }, + { + "epoch": 5.96, + "learning_rate": 2.2344561870035754e-09, + "loss": 0.9588, + "step": 80668 + }, + { + "epoch": 5.96, + "learning_rate": 2.2260325440437437e-09, + "loss": 1.0527, + "step": 80669 + }, + { + "epoch": 5.96, + "learning_rate": 2.217624807399066e-09, + "loss": 1.0434, + "step": 80670 + }, + { + "epoch": 5.96, + "learning_rate": 2.2092329770828647e-09, + "loss": 1.0791, + "step": 80671 + }, + { + "epoch": 5.96, + "learning_rate": 2.200857053108463e-09, + "loss": 1.0589, + "step": 80672 + }, + { + "epoch": 5.96, + "learning_rate": 2.192497035490293e-09, + "loss": 0.8458, + "step": 80673 + }, + { + "epoch": 5.96, + "learning_rate": 2.1841529242394577e-09, + "loss": 0.9021, + "step": 80674 + }, + { + "epoch": 5.96, + "learning_rate": 2.1758247193715e-09, + "loss": 0.9934, + "step": 80675 + }, + { + "epoch": 5.96, + "learning_rate": 2.167512420898632e-09, + "loss": 0.9675, + "step": 80676 + }, + { + "epoch": 5.96, + "learning_rate": 2.159216028835287e-09, + "loss": 1.0315, + "step": 80677 + }, + { + "epoch": 5.96, + "learning_rate": 2.1509355431925673e-09, + "loss": 0.9333, + "step": 80678 + }, + { + "epoch": 5.96, + "learning_rate": 2.142670963983795e-09, + "loss": 0.9991, + "step": 80679 + }, + { + "epoch": 5.96, + "learning_rate": 2.1344222912245137e-09, + "loss": 0.9787, + "step": 80680 + }, + { + "epoch": 5.96, + "learning_rate": 2.126189524925826e-09, + "loss": 0.8509, + "step": 80681 + }, + { + "epoch": 5.96, + "learning_rate": 2.1179726651010534e-09, + "loss": 1.0845, + "step": 80682 + }, + { + "epoch": 5.96, + "learning_rate": 2.1097717117635195e-09, + "loss": 1.0803, + "step": 80683 + }, + { + "epoch": 5.96, + "learning_rate": 2.101586664926547e-09, + "loss": 1.0026, + "step": 80684 + }, + { + "epoch": 5.96, + "learning_rate": 2.093417524602348e-09, + "loss": 1.0622, + "step": 80685 + }, + { + "epoch": 5.96, + "learning_rate": 2.0852642908053556e-09, + "loss": 0.9124, + "step": 80686 + }, + { + "epoch": 5.96, + "learning_rate": 2.0771269635466716e-09, + "loss": 0.9078, + "step": 80687 + }, + { + "epoch": 5.96, + "learning_rate": 2.0690055428407298e-09, + "loss": 1.0573, + "step": 80688 + }, + { + "epoch": 5.96, + "learning_rate": 2.0609000287008518e-09, + "loss": 0.9605, + "step": 80689 + }, + { + "epoch": 5.96, + "learning_rate": 2.0528104211381404e-09, + "loss": 1.045, + "step": 80690 + }, + { + "epoch": 5.96, + "learning_rate": 2.0447367201659184e-09, + "loss": 1.0582, + "step": 80691 + }, + { + "epoch": 5.96, + "learning_rate": 2.0366789257986186e-09, + "loss": 1.0534, + "step": 80692 + }, + { + "epoch": 5.96, + "learning_rate": 2.0286370380462327e-09, + "loss": 1.0675, + "step": 80693 + }, + { + "epoch": 5.96, + "learning_rate": 2.020611056924304e-09, + "loss": 0.9771, + "step": 80694 + }, + { + "epoch": 5.96, + "learning_rate": 2.0126009824450455e-09, + "loss": 0.9639, + "step": 80695 + }, + { + "epoch": 5.96, + "learning_rate": 2.0046068146195587e-09, + "loss": 0.9188, + "step": 80696 + }, + { + "epoch": 5.96, + "learning_rate": 1.9966285534633865e-09, + "loss": 0.9325, + "step": 80697 + }, + { + "epoch": 5.96, + "learning_rate": 1.9886661989854116e-09, + "loss": 0.9207, + "step": 80698 + }, + { + "epoch": 5.96, + "learning_rate": 1.980719751202287e-09, + "loss": 0.9319, + "step": 80699 + }, + { + "epoch": 5.96, + "learning_rate": 1.972789210124004e-09, + "loss": 0.9985, + "step": 80700 + }, + { + "epoch": 5.96, + "learning_rate": 1.9648745757638867e-09, + "loss": 1.029, + "step": 80701 + }, + { + "epoch": 5.96, + "learning_rate": 1.956975848134146e-09, + "loss": 0.963, + "step": 80702 + }, + { + "epoch": 5.96, + "learning_rate": 1.9490930272481058e-09, + "loss": 0.9861, + "step": 80703 + }, + { + "epoch": 5.96, + "learning_rate": 1.9412261131179777e-09, + "loss": 0.9977, + "step": 80704 + }, + { + "epoch": 5.96, + "learning_rate": 1.9333751057570847e-09, + "loss": 1.0472, + "step": 80705 + }, + { + "epoch": 5.96, + "learning_rate": 1.9255400051765293e-09, + "loss": 1.083, + "step": 80706 + }, + { + "epoch": 5.96, + "learning_rate": 1.917720811389634e-09, + "loss": 1.0055, + "step": 80707 + }, + { + "epoch": 5.96, + "learning_rate": 1.909917524408611e-09, + "loss": 1.0169, + "step": 80708 + }, + { + "epoch": 5.96, + "learning_rate": 1.902130144245673e-09, + "loss": 1.0325, + "step": 80709 + }, + { + "epoch": 5.96, + "learning_rate": 1.8943586709130326e-09, + "loss": 0.9091, + "step": 80710 + }, + { + "epoch": 5.96, + "learning_rate": 1.886603104424012e-09, + "loss": 0.9792, + "step": 80711 + }, + { + "epoch": 5.96, + "learning_rate": 1.8788634447897135e-09, + "loss": 0.9904, + "step": 80712 + }, + { + "epoch": 5.96, + "learning_rate": 1.8711396920234603e-09, + "loss": 0.9496, + "step": 80713 + }, + { + "epoch": 5.96, + "learning_rate": 1.8634318461374646e-09, + "loss": 1.0285, + "step": 80714 + }, + { + "epoch": 5.96, + "learning_rate": 1.8557399071428283e-09, + "loss": 0.9966, + "step": 80715 + }, + { + "epoch": 5.96, + "learning_rate": 1.8480638750539848e-09, + "loss": 1.0934, + "step": 80716 + }, + { + "epoch": 5.96, + "learning_rate": 1.8404037498809258e-09, + "loss": 1.0238, + "step": 80717 + }, + { + "epoch": 5.96, + "learning_rate": 1.832759531636974e-09, + "loss": 0.9953, + "step": 80718 + }, + { + "epoch": 5.96, + "learning_rate": 1.8251312203332317e-09, + "loss": 0.9733, + "step": 80719 + }, + { + "epoch": 5.96, + "learning_rate": 1.8175188159841317e-09, + "loss": 0.9608, + "step": 80720 + }, + { + "epoch": 5.96, + "learning_rate": 1.8099223185996662e-09, + "loss": 1.0479, + "step": 80721 + }, + { + "epoch": 5.96, + "learning_rate": 1.8023417281920475e-09, + "loss": 1.0964, + "step": 80722 + }, + { + "epoch": 5.96, + "learning_rate": 1.7947770447745983e-09, + "loss": 0.8433, + "step": 80723 + }, + { + "epoch": 5.96, + "learning_rate": 1.7872282683584208e-09, + "loss": 0.83, + "step": 80724 + }, + { + "epoch": 5.96, + "learning_rate": 1.7796953989557274e-09, + "loss": 1.0092, + "step": 80725 + }, + { + "epoch": 5.96, + "learning_rate": 1.7721784365776207e-09, + "loss": 0.8673, + "step": 80726 + }, + { + "epoch": 5.96, + "learning_rate": 1.7646773812385333e-09, + "loss": 0.9806, + "step": 80727 + }, + { + "epoch": 5.96, + "learning_rate": 1.7571922329473467e-09, + "loss": 0.9004, + "step": 80728 + }, + { + "epoch": 5.97, + "learning_rate": 1.7497229917184943e-09, + "loss": 0.882, + "step": 80729 + }, + { + "epoch": 5.97, + "learning_rate": 1.742269657561968e-09, + "loss": 0.9578, + "step": 80730 + }, + { + "epoch": 5.97, + "learning_rate": 1.7348322304910902e-09, + "loss": 0.9316, + "step": 80731 + }, + { + "epoch": 5.97, + "learning_rate": 1.7274107105158534e-09, + "loss": 0.8918, + "step": 80732 + }, + { + "epoch": 5.97, + "learning_rate": 1.7200050976506898e-09, + "loss": 1.009, + "step": 80733 + }, + { + "epoch": 5.97, + "learning_rate": 1.712615391904482e-09, + "loss": 0.8876, + "step": 80734 + }, + { + "epoch": 5.97, + "learning_rate": 1.7052415932916621e-09, + "loss": 0.9649, + "step": 80735 + }, + { + "epoch": 5.97, + "learning_rate": 1.6978837018222227e-09, + "loss": 0.9232, + "step": 80736 + }, + { + "epoch": 5.97, + "learning_rate": 1.6905417175094862e-09, + "loss": 1.0471, + "step": 80737 + }, + { + "epoch": 5.97, + "learning_rate": 1.6832156403623346e-09, + "loss": 1.0175, + "step": 80738 + }, + { + "epoch": 5.97, + "learning_rate": 1.6759054703952004e-09, + "loss": 1.015, + "step": 80739 + }, + { + "epoch": 5.97, + "learning_rate": 1.668611207618076e-09, + "loss": 1.1229, + "step": 80740 + }, + { + "epoch": 5.97, + "learning_rate": 1.661332852044284e-09, + "loss": 1.0476, + "step": 80741 + }, + { + "epoch": 5.97, + "learning_rate": 1.654070403682706e-09, + "loss": 0.9999, + "step": 80742 + }, + { + "epoch": 5.97, + "learning_rate": 1.646823862547775e-09, + "loss": 0.9325, + "step": 80743 + }, + { + "epoch": 5.97, + "learning_rate": 1.6395932286483729e-09, + "loss": 0.9191, + "step": 80744 + }, + { + "epoch": 5.97, + "learning_rate": 1.6323785019989324e-09, + "loss": 0.9335, + "step": 80745 + }, + { + "epoch": 5.97, + "learning_rate": 1.6251796826083354e-09, + "loss": 0.9332, + "step": 80746 + }, + { + "epoch": 5.97, + "learning_rate": 1.6179967704887943e-09, + "loss": 0.9123, + "step": 80747 + }, + { + "epoch": 5.97, + "learning_rate": 1.6108297656525218e-09, + "loss": 0.9591, + "step": 80748 + }, + { + "epoch": 5.97, + "learning_rate": 1.6036786681095096e-09, + "loss": 0.93, + "step": 80749 + }, + { + "epoch": 5.97, + "learning_rate": 1.5965434778730804e-09, + "loss": 0.905, + "step": 80750 + }, + { + "epoch": 5.97, + "learning_rate": 1.5894241949532263e-09, + "loss": 0.9195, + "step": 80751 + }, + { + "epoch": 5.97, + "learning_rate": 1.5823208193610496e-09, + "loss": 0.9792, + "step": 80752 + }, + { + "epoch": 5.97, + "learning_rate": 1.5752333511087626e-09, + "loss": 0.9827, + "step": 80753 + }, + { + "epoch": 5.97, + "learning_rate": 1.5681617902074674e-09, + "loss": 1.0395, + "step": 80754 + }, + { + "epoch": 5.97, + "learning_rate": 1.5611061366671565e-09, + "loss": 1.0432, + "step": 80755 + }, + { + "epoch": 5.97, + "learning_rate": 1.5540663905011522e-09, + "loss": 0.9939, + "step": 80756 + }, + { + "epoch": 5.97, + "learning_rate": 1.5470425517194465e-09, + "loss": 0.9837, + "step": 80757 + }, + { + "epoch": 5.97, + "learning_rate": 1.5400346203331418e-09, + "loss": 0.8752, + "step": 80758 + }, + { + "epoch": 5.97, + "learning_rate": 1.5330425963533403e-09, + "loss": 0.9829, + "step": 80759 + }, + { + "epoch": 5.97, + "learning_rate": 1.526066479791144e-09, + "loss": 0.9667, + "step": 80760 + }, + { + "epoch": 5.97, + "learning_rate": 1.5191062706587657e-09, + "loss": 0.9799, + "step": 80761 + }, + { + "epoch": 5.97, + "learning_rate": 1.512161968966197e-09, + "loss": 0.987, + "step": 80762 + }, + { + "epoch": 5.97, + "learning_rate": 1.5052335747245407e-09, + "loss": 0.9031, + "step": 80763 + }, + { + "epoch": 5.97, + "learning_rate": 1.4983210879448985e-09, + "loss": 0.9988, + "step": 80764 + }, + { + "epoch": 5.97, + "learning_rate": 1.4914245086383727e-09, + "loss": 0.9357, + "step": 80765 + }, + { + "epoch": 5.97, + "learning_rate": 1.4845438368160659e-09, + "loss": 0.9426, + "step": 80766 + }, + { + "epoch": 5.97, + "learning_rate": 1.47767907248908e-09, + "loss": 0.9093, + "step": 80767 + }, + { + "epoch": 5.97, + "learning_rate": 1.470830215667407e-09, + "loss": 1.03, + "step": 80768 + }, + { + "epoch": 5.97, + "learning_rate": 1.4639972663632596e-09, + "loss": 0.9065, + "step": 80769 + }, + { + "epoch": 5.97, + "learning_rate": 1.4571802245855192e-09, + "loss": 1.016, + "step": 80770 + }, + { + "epoch": 5.97, + "learning_rate": 1.4503790903475091e-09, + "loss": 0.912, + "step": 80771 + }, + { + "epoch": 5.97, + "learning_rate": 1.4435938636592206e-09, + "loss": 1.0107, + "step": 80772 + }, + { + "epoch": 5.97, + "learning_rate": 1.436824544530646e-09, + "loss": 0.9425, + "step": 80773 + }, + { + "epoch": 5.97, + "learning_rate": 1.4300711329728878e-09, + "loss": 1.086, + "step": 80774 + }, + { + "epoch": 5.97, + "learning_rate": 1.4233336289970478e-09, + "loss": 1.0263, + "step": 80775 + }, + { + "epoch": 5.97, + "learning_rate": 1.4166120326131182e-09, + "loss": 1.0093, + "step": 80776 + }, + { + "epoch": 5.97, + "learning_rate": 1.4099063438333117e-09, + "loss": 0.9661, + "step": 80777 + }, + { + "epoch": 5.97, + "learning_rate": 1.4032165626665096e-09, + "loss": 0.9628, + "step": 80778 + }, + { + "epoch": 5.97, + "learning_rate": 1.3965426891238143e-09, + "loss": 0.9699, + "step": 80779 + }, + { + "epoch": 5.97, + "learning_rate": 1.3898847232174384e-09, + "loss": 0.9651, + "step": 80780 + }, + { + "epoch": 5.97, + "learning_rate": 1.3832426649551534e-09, + "loss": 0.957, + "step": 80781 + }, + { + "epoch": 5.97, + "learning_rate": 1.376616514350282e-09, + "loss": 0.9904, + "step": 80782 + }, + { + "epoch": 5.97, + "learning_rate": 1.3700062714117058e-09, + "loss": 1.049, + "step": 80783 + }, + { + "epoch": 5.97, + "learning_rate": 1.3634119361505272e-09, + "loss": 0.9755, + "step": 80784 + }, + { + "epoch": 5.97, + "learning_rate": 1.356833508576738e-09, + "loss": 0.9222, + "step": 80785 + }, + { + "epoch": 5.97, + "learning_rate": 1.350270988701441e-09, + "loss": 1.0023, + "step": 80786 + }, + { + "epoch": 5.97, + "learning_rate": 1.3437243765346276e-09, + "loss": 1.0427, + "step": 80787 + }, + { + "epoch": 5.97, + "learning_rate": 1.3371936720874002e-09, + "loss": 1.0167, + "step": 80788 + }, + { + "epoch": 5.97, + "learning_rate": 1.3306788753686405e-09, + "loss": 1.0379, + "step": 80789 + }, + { + "epoch": 5.97, + "learning_rate": 1.3241799863905612e-09, + "loss": 1.0187, + "step": 80790 + }, + { + "epoch": 5.97, + "learning_rate": 1.3176970051620442e-09, + "loss": 0.9655, + "step": 80791 + }, + { + "epoch": 5.97, + "learning_rate": 1.3112299316953014e-09, + "loss": 1.0729, + "step": 80792 + }, + { + "epoch": 5.97, + "learning_rate": 1.3047787659981048e-09, + "loss": 0.9196, + "step": 80793 + }, + { + "epoch": 5.97, + "learning_rate": 1.2983435080815566e-09, + "loss": 0.9884, + "step": 80794 + }, + { + "epoch": 5.97, + "learning_rate": 1.291924157957869e-09, + "loss": 1.0618, + "step": 80795 + }, + { + "epoch": 5.97, + "learning_rate": 1.2855207156348137e-09, + "loss": 1.0394, + "step": 80796 + }, + { + "epoch": 5.97, + "learning_rate": 1.279133181123493e-09, + "loss": 0.9643, + "step": 80797 + }, + { + "epoch": 5.97, + "learning_rate": 1.272761554433899e-09, + "loss": 1.1022, + "step": 80798 + }, + { + "epoch": 5.97, + "learning_rate": 1.2664058355760233e-09, + "loss": 0.9834, + "step": 80799 + }, + { + "epoch": 5.97, + "learning_rate": 1.2600660245609687e-09, + "loss": 0.9418, + "step": 80800 + }, + { + "epoch": 5.97, + "learning_rate": 1.2537421213976163e-09, + "loss": 1.081, + "step": 80801 + }, + { + "epoch": 5.97, + "learning_rate": 1.2474341260959589e-09, + "loss": 1.0825, + "step": 80802 + }, + { + "epoch": 5.97, + "learning_rate": 1.241142038667098e-09, + "loss": 0.9183, + "step": 80803 + }, + { + "epoch": 5.97, + "learning_rate": 1.2348658591210262e-09, + "loss": 1.0342, + "step": 80804 + }, + { + "epoch": 5.97, + "learning_rate": 1.2286055874666247e-09, + "loss": 0.9821, + "step": 80805 + }, + { + "epoch": 5.97, + "learning_rate": 1.2223612237149963e-09, + "loss": 0.9862, + "step": 80806 + }, + { + "epoch": 5.97, + "learning_rate": 1.2161327678750224e-09, + "loss": 0.974, + "step": 80807 + }, + { + "epoch": 5.97, + "learning_rate": 1.209920219956695e-09, + "loss": 1.0974, + "step": 80808 + }, + { + "epoch": 5.97, + "learning_rate": 1.2037235799711166e-09, + "loss": 1.0282, + "step": 80809 + }, + { + "epoch": 5.97, + "learning_rate": 1.197542847927169e-09, + "loss": 0.9982, + "step": 80810 + }, + { + "epoch": 5.97, + "learning_rate": 1.1913780238348438e-09, + "loss": 0.9056, + "step": 80811 + }, + { + "epoch": 5.97, + "learning_rate": 1.1852291077041333e-09, + "loss": 1.0098, + "step": 80812 + }, + { + "epoch": 5.97, + "learning_rate": 1.1790960995450296e-09, + "loss": 0.8871, + "step": 80813 + }, + { + "epoch": 5.97, + "learning_rate": 1.172978999365304e-09, + "loss": 1.0343, + "step": 80814 + }, + { + "epoch": 5.97, + "learning_rate": 1.1668778071782793e-09, + "loss": 0.9088, + "step": 80815 + }, + { + "epoch": 5.97, + "learning_rate": 1.160792522990617e-09, + "loss": 1.0701, + "step": 80816 + }, + { + "epoch": 5.97, + "learning_rate": 1.154723146813419e-09, + "loss": 1.0844, + "step": 80817 + }, + { + "epoch": 5.97, + "learning_rate": 1.1486696786555674e-09, + "loss": 1.1173, + "step": 80818 + }, + { + "epoch": 5.97, + "learning_rate": 1.142632118527054e-09, + "loss": 0.9467, + "step": 80819 + }, + { + "epoch": 5.97, + "learning_rate": 1.1366104664378709e-09, + "loss": 0.9379, + "step": 80820 + }, + { + "epoch": 5.97, + "learning_rate": 1.13060472239801e-09, + "loss": 0.9591, + "step": 80821 + }, + { + "epoch": 5.97, + "learning_rate": 1.1246148864152428e-09, + "loss": 1.0177, + "step": 80822 + }, + { + "epoch": 5.97, + "learning_rate": 1.118640958500672e-09, + "loss": 0.9543, + "step": 80823 + }, + { + "epoch": 5.97, + "learning_rate": 1.1126829386631787e-09, + "loss": 0.9216, + "step": 80824 + }, + { + "epoch": 5.97, + "learning_rate": 1.1067408269116453e-09, + "loss": 0.8966, + "step": 80825 + }, + { + "epoch": 5.97, + "learning_rate": 1.1008146232571737e-09, + "loss": 1.0262, + "step": 80826 + }, + { + "epoch": 5.97, + "learning_rate": 1.0949043277086457e-09, + "loss": 0.9937, + "step": 80827 + }, + { + "epoch": 5.97, + "learning_rate": 1.089009940273833e-09, + "loss": 1.0115, + "step": 80828 + }, + { + "epoch": 5.97, + "learning_rate": 1.0831314609638377e-09, + "loss": 1.0345, + "step": 80829 + }, + { + "epoch": 5.97, + "learning_rate": 1.0772688897875417e-09, + "loss": 1.0458, + "step": 80830 + }, + { + "epoch": 5.97, + "learning_rate": 1.0714222267538265e-09, + "loss": 0.9276, + "step": 80831 + }, + { + "epoch": 5.97, + "learning_rate": 1.0655914718726846e-09, + "loss": 1.0163, + "step": 80832 + }, + { + "epoch": 5.97, + "learning_rate": 1.0597766251541075e-09, + "loss": 0.9564, + "step": 80833 + }, + { + "epoch": 5.97, + "learning_rate": 1.053977686605867e-09, + "loss": 1.054, + "step": 80834 + }, + { + "epoch": 5.97, + "learning_rate": 1.048194656236845e-09, + "loss": 1.022, + "step": 80835 + }, + { + "epoch": 5.97, + "learning_rate": 1.0424275340570333e-09, + "loss": 0.9593, + "step": 80836 + }, + { + "epoch": 5.97, + "learning_rate": 1.036676320076424e-09, + "loss": 0.847, + "step": 80837 + }, + { + "epoch": 5.97, + "learning_rate": 1.030941014303899e-09, + "loss": 1.1334, + "step": 80838 + }, + { + "epoch": 5.97, + "learning_rate": 1.0252216167472296e-09, + "loss": 1.0781, + "step": 80839 + }, + { + "epoch": 5.97, + "learning_rate": 1.0195181274164078e-09, + "loss": 1.0051, + "step": 80840 + }, + { + "epoch": 5.97, + "learning_rate": 1.0138305463203158e-09, + "loss": 0.957, + "step": 80841 + }, + { + "epoch": 5.97, + "learning_rate": 1.0081588734689452e-09, + "loss": 0.9029, + "step": 80842 + }, + { + "epoch": 5.97, + "learning_rate": 1.0025031088700676e-09, + "loss": 0.9325, + "step": 80843 + }, + { + "epoch": 5.97, + "learning_rate": 9.96863252532565e-10, + "loss": 0.9413, + "step": 80844 + }, + { + "epoch": 5.97, + "learning_rate": 9.912393044664293e-10, + "loss": 1.0604, + "step": 80845 + }, + { + "epoch": 5.97, + "learning_rate": 9.856312646805422e-10, + "loss": 0.9405, + "step": 80846 + }, + { + "epoch": 5.97, + "learning_rate": 9.800391331826753e-10, + "loss": 0.9051, + "step": 80847 + }, + { + "epoch": 5.97, + "learning_rate": 9.744629099839309e-10, + "loss": 1.0948, + "step": 80848 + }, + { + "epoch": 5.97, + "learning_rate": 9.6890259508986e-10, + "loss": 1.0234, + "step": 80849 + }, + { + "epoch": 5.97, + "learning_rate": 9.633581885126754e-10, + "loss": 1.0562, + "step": 80850 + }, + { + "epoch": 5.97, + "learning_rate": 9.578296902590379e-10, + "loss": 1.1141, + "step": 80851 + }, + { + "epoch": 5.97, + "learning_rate": 9.523171003389397e-10, + "loss": 1.0259, + "step": 80852 + }, + { + "epoch": 5.97, + "learning_rate": 9.468204187601527e-10, + "loss": 0.9981, + "step": 80853 + }, + { + "epoch": 5.97, + "learning_rate": 9.413396455326685e-10, + "loss": 0.8856, + "step": 80854 + }, + { + "epoch": 5.97, + "learning_rate": 9.358747806631486e-10, + "loss": 0.8891, + "step": 80855 + }, + { + "epoch": 5.97, + "learning_rate": 9.304258241626951e-10, + "loss": 1.022, + "step": 80856 + }, + { + "epoch": 5.97, + "learning_rate": 9.249927760379696e-10, + "loss": 0.8943, + "step": 80857 + }, + { + "epoch": 5.97, + "learning_rate": 9.195756362989638e-10, + "loss": 0.9575, + "step": 80858 + }, + { + "epoch": 5.97, + "learning_rate": 9.141744049545598e-10, + "loss": 1.0245, + "step": 80859 + }, + { + "epoch": 5.97, + "learning_rate": 9.087890820114187e-10, + "loss": 0.9136, + "step": 80860 + }, + { + "epoch": 5.97, + "learning_rate": 9.034196674806428e-10, + "loss": 0.8729, + "step": 80861 + }, + { + "epoch": 5.97, + "learning_rate": 8.980661613677832e-10, + "loss": 1.0345, + "step": 80862 + }, + { + "epoch": 5.97, + "learning_rate": 8.927285636839422e-10, + "loss": 1.0685, + "step": 80863 + }, + { + "epoch": 5.97, + "learning_rate": 8.874068744368913e-10, + "loss": 0.9757, + "step": 80864 + }, + { + "epoch": 5.98, + "learning_rate": 8.821010936355123e-10, + "loss": 1.0155, + "step": 80865 + }, + { + "epoch": 5.98, + "learning_rate": 8.768112212864666e-10, + "loss": 1.0278, + "step": 80866 + }, + { + "epoch": 5.98, + "learning_rate": 8.71537257399746e-10, + "loss": 1.0135, + "step": 80867 + }, + { + "epoch": 5.98, + "learning_rate": 8.662792019831223e-10, + "loss": 1.0251, + "step": 80868 + }, + { + "epoch": 5.98, + "learning_rate": 8.610370550454772e-10, + "loss": 1.1233, + "step": 80869 + }, + { + "epoch": 5.98, + "learning_rate": 8.558108165945822e-10, + "loss": 0.9908, + "step": 80870 + }, + { + "epoch": 5.98, + "learning_rate": 8.506004866393192e-10, + "loss": 1.0183, + "step": 80871 + }, + { + "epoch": 5.98, + "learning_rate": 8.454060651885698e-10, + "loss": 1.0072, + "step": 80872 + }, + { + "epoch": 5.98, + "learning_rate": 8.402275522478853e-10, + "loss": 1.0137, + "step": 80873 + }, + { + "epoch": 5.98, + "learning_rate": 8.350649478283679e-10, + "loss": 0.963, + "step": 80874 + }, + { + "epoch": 5.98, + "learning_rate": 8.299182519377891e-10, + "loss": 0.8642, + "step": 80875 + }, + { + "epoch": 5.98, + "learning_rate": 8.247874645828102e-10, + "loss": 0.9874, + "step": 80876 + }, + { + "epoch": 5.98, + "learning_rate": 8.196725857723132e-10, + "loss": 1.066, + "step": 80877 + }, + { + "epoch": 5.98, + "learning_rate": 8.145736155151795e-10, + "loss": 1.0254, + "step": 80878 + }, + { + "epoch": 5.98, + "learning_rate": 8.09490553819181e-10, + "loss": 0.932, + "step": 80879 + }, + { + "epoch": 5.98, + "learning_rate": 8.044234006920892e-10, + "loss": 1.0349, + "step": 80880 + }, + { + "epoch": 5.98, + "learning_rate": 7.993721561416756e-10, + "loss": 1.0988, + "step": 80881 + }, + { + "epoch": 5.98, + "learning_rate": 7.94336820176822e-10, + "loss": 1.0417, + "step": 80882 + }, + { + "epoch": 5.98, + "learning_rate": 7.893173928041897e-10, + "loss": 1.1266, + "step": 80883 + }, + { + "epoch": 5.98, + "learning_rate": 7.843138740337708e-10, + "loss": 0.9954, + "step": 80884 + }, + { + "epoch": 5.98, + "learning_rate": 7.793262638722265e-10, + "loss": 1.093, + "step": 80885 + }, + { + "epoch": 5.98, + "learning_rate": 7.743545623273286e-10, + "loss": 0.9685, + "step": 80886 + }, + { + "epoch": 5.98, + "learning_rate": 7.693987694079586e-10, + "loss": 1.071, + "step": 80887 + }, + { + "epoch": 5.98, + "learning_rate": 7.644588851207779e-10, + "loss": 0.875, + "step": 80888 + }, + { + "epoch": 5.98, + "learning_rate": 7.595349094746685e-10, + "loss": 1.0079, + "step": 80889 + }, + { + "epoch": 5.98, + "learning_rate": 7.546268424774017e-10, + "loss": 0.9687, + "step": 80890 + }, + { + "epoch": 5.98, + "learning_rate": 7.49734684135639e-10, + "loss": 1.0318, + "step": 80891 + }, + { + "epoch": 5.98, + "learning_rate": 7.448584344582622e-10, + "loss": 1.1034, + "step": 80892 + }, + { + "epoch": 5.98, + "learning_rate": 7.399980934530426e-10, + "loss": 0.98, + "step": 80893 + }, + { + "epoch": 5.98, + "learning_rate": 7.351536611277522e-10, + "loss": 1.0107, + "step": 80894 + }, + { + "epoch": 5.98, + "learning_rate": 7.303251374890519e-10, + "loss": 0.9718, + "step": 80895 + }, + { + "epoch": 5.98, + "learning_rate": 7.255125225458238e-10, + "loss": 1.0197, + "step": 80896 + }, + { + "epoch": 5.98, + "learning_rate": 7.20715816304729e-10, + "loss": 0.9863, + "step": 80897 + }, + { + "epoch": 5.98, + "learning_rate": 7.159350187735393e-10, + "loss": 0.9995, + "step": 80898 + }, + { + "epoch": 5.98, + "learning_rate": 7.111701299611363e-10, + "loss": 1.1458, + "step": 80899 + }, + { + "epoch": 5.98, + "learning_rate": 7.064211498741813e-10, + "loss": 0.958, + "step": 80900 + }, + { + "epoch": 5.98, + "learning_rate": 7.016880785193359e-10, + "loss": 1.1085, + "step": 80901 + }, + { + "epoch": 5.98, + "learning_rate": 6.969709159065918e-10, + "loss": 0.9189, + "step": 80902 + }, + { + "epoch": 5.98, + "learning_rate": 6.922696620403901e-10, + "loss": 0.9797, + "step": 80903 + }, + { + "epoch": 5.98, + "learning_rate": 6.875843169307228e-10, + "loss": 1.0248, + "step": 80904 + }, + { + "epoch": 5.98, + "learning_rate": 6.829148805831409e-10, + "loss": 0.9883, + "step": 80905 + }, + { + "epoch": 5.98, + "learning_rate": 6.782613530065263e-10, + "loss": 0.9955, + "step": 80906 + }, + { + "epoch": 5.98, + "learning_rate": 6.736237342075403e-10, + "loss": 1.1016, + "step": 80907 + }, + { + "epoch": 5.98, + "learning_rate": 6.690020241939544e-10, + "loss": 0.9634, + "step": 80908 + }, + { + "epoch": 5.98, + "learning_rate": 6.643962229724299e-10, + "loss": 1.0238, + "step": 80909 + }, + { + "epoch": 5.98, + "learning_rate": 6.598063305518487e-10, + "loss": 0.9334, + "step": 80910 + }, + { + "epoch": 5.98, + "learning_rate": 6.552323469377619e-10, + "loss": 0.8771, + "step": 80911 + }, + { + "epoch": 5.98, + "learning_rate": 6.506742721379411e-10, + "loss": 0.9395, + "step": 80912 + }, + { + "epoch": 5.98, + "learning_rate": 6.461321061601577e-10, + "loss": 0.8908, + "step": 80913 + }, + { + "epoch": 5.98, + "learning_rate": 6.416058490110733e-10, + "loss": 0.9116, + "step": 80914 + }, + { + "epoch": 5.98, + "learning_rate": 6.370955006984591e-10, + "loss": 0.9926, + "step": 80915 + }, + { + "epoch": 5.98, + "learning_rate": 6.326010612289768e-10, + "loss": 1.032, + "step": 80916 + }, + { + "epoch": 5.98, + "learning_rate": 6.281225306103978e-10, + "loss": 1.0511, + "step": 80917 + }, + { + "epoch": 5.98, + "learning_rate": 6.236599088482731e-10, + "loss": 1.0016, + "step": 80918 + }, + { + "epoch": 5.98, + "learning_rate": 6.192131959514847e-10, + "loss": 0.885, + "step": 80919 + }, + { + "epoch": 5.98, + "learning_rate": 6.147823919266937e-10, + "loss": 0.9387, + "step": 80920 + }, + { + "epoch": 5.98, + "learning_rate": 6.103674967805617e-10, + "loss": 1.0594, + "step": 80921 + }, + { + "epoch": 5.98, + "learning_rate": 6.0596851051975e-10, + "loss": 0.9704, + "step": 80922 + }, + { + "epoch": 5.98, + "learning_rate": 6.015854331531401e-10, + "loss": 1.0758, + "step": 80923 + }, + { + "epoch": 5.98, + "learning_rate": 5.972182646851732e-10, + "loss": 0.998, + "step": 80924 + }, + { + "epoch": 5.98, + "learning_rate": 5.928670051236207e-10, + "loss": 0.9653, + "step": 80925 + }, + { + "epoch": 5.98, + "learning_rate": 5.885316544762542e-10, + "loss": 1.1264, + "step": 80926 + }, + { + "epoch": 5.98, + "learning_rate": 5.842122127497352e-10, + "loss": 0.9538, + "step": 80927 + }, + { + "epoch": 5.98, + "learning_rate": 5.799086799507248e-10, + "loss": 1.1648, + "step": 80928 + }, + { + "epoch": 5.98, + "learning_rate": 5.756210560858843e-10, + "loss": 0.923, + "step": 80929 + }, + { + "epoch": 5.98, + "learning_rate": 5.713493411618753e-10, + "loss": 1.0151, + "step": 80930 + }, + { + "epoch": 5.98, + "learning_rate": 5.670935351864692e-10, + "loss": 1.012, + "step": 80931 + }, + { + "epoch": 5.98, + "learning_rate": 5.628536381652172e-10, + "loss": 1.0172, + "step": 80932 + }, + { + "epoch": 5.98, + "learning_rate": 5.586296501047806e-10, + "loss": 0.8806, + "step": 80933 + }, + { + "epoch": 5.98, + "learning_rate": 5.54421571014041e-10, + "loss": 1.1558, + "step": 80934 + }, + { + "epoch": 5.98, + "learning_rate": 5.502294008963294e-10, + "loss": 1.0485, + "step": 80935 + }, + { + "epoch": 5.98, + "learning_rate": 5.460531397616376e-10, + "loss": 0.9884, + "step": 80936 + }, + { + "epoch": 5.98, + "learning_rate": 5.418927876144065e-10, + "loss": 1.0511, + "step": 80937 + }, + { + "epoch": 5.98, + "learning_rate": 5.377483444624077e-10, + "loss": 1.0118, + "step": 80938 + }, + { + "epoch": 5.98, + "learning_rate": 5.336198103111922e-10, + "loss": 0.9798, + "step": 80939 + }, + { + "epoch": 5.98, + "learning_rate": 5.295071851685319e-10, + "loss": 1.0608, + "step": 80940 + }, + { + "epoch": 5.98, + "learning_rate": 5.254104690399775e-10, + "loss": 0.8859, + "step": 80941 + }, + { + "epoch": 5.98, + "learning_rate": 5.213296619321905e-10, + "loss": 1.0499, + "step": 80942 + }, + { + "epoch": 5.98, + "learning_rate": 5.172647638518324e-10, + "loss": 1.0011, + "step": 80943 + }, + { + "epoch": 5.98, + "learning_rate": 5.132157748055644e-10, + "loss": 0.9959, + "step": 80944 + }, + { + "epoch": 5.98, + "learning_rate": 5.091826948000478e-10, + "loss": 0.9299, + "step": 80945 + }, + { + "epoch": 5.98, + "learning_rate": 5.05165523841944e-10, + "loss": 1.1013, + "step": 80946 + }, + { + "epoch": 5.98, + "learning_rate": 5.011642619356937e-10, + "loss": 1.0989, + "step": 80947 + }, + { + "epoch": 5.98, + "learning_rate": 4.971789090901791e-10, + "loss": 0.919, + "step": 80948 + }, + { + "epoch": 5.98, + "learning_rate": 4.932094653098408e-10, + "loss": 0.9249, + "step": 80949 + }, + { + "epoch": 5.98, + "learning_rate": 4.8925593060134e-10, + "loss": 0.9979, + "step": 80950 + }, + { + "epoch": 5.98, + "learning_rate": 4.853183049724486e-10, + "loss": 0.999, + "step": 80951 + }, + { + "epoch": 5.98, + "learning_rate": 4.813965884276073e-10, + "loss": 1.0461, + "step": 80952 + }, + { + "epoch": 5.98, + "learning_rate": 4.774907809745876e-10, + "loss": 1.0334, + "step": 80953 + }, + { + "epoch": 5.98, + "learning_rate": 4.736008826178307e-10, + "loss": 1.0519, + "step": 80954 + }, + { + "epoch": 5.98, + "learning_rate": 4.697268933651078e-10, + "loss": 1.0126, + "step": 80955 + }, + { + "epoch": 5.98, + "learning_rate": 4.658688132219702e-10, + "loss": 1.0064, + "step": 80956 + }, + { + "epoch": 5.98, + "learning_rate": 4.6202664219396896e-10, + "loss": 1.0213, + "step": 80957 + }, + { + "epoch": 5.98, + "learning_rate": 4.582003802888757e-10, + "loss": 0.9543, + "step": 80958 + }, + { + "epoch": 5.98, + "learning_rate": 4.5439002751113125e-10, + "loss": 1.056, + "step": 80959 + }, + { + "epoch": 5.98, + "learning_rate": 4.50595583867397e-10, + "loss": 1.0686, + "step": 80960 + }, + { + "epoch": 5.98, + "learning_rate": 4.46817049363224e-10, + "loss": 1.0378, + "step": 80961 + }, + { + "epoch": 5.98, + "learning_rate": 4.4305442400527364e-10, + "loss": 1.0622, + "step": 80962 + }, + { + "epoch": 5.98, + "learning_rate": 4.3930770780020724e-10, + "loss": 0.9289, + "step": 80963 + }, + { + "epoch": 5.98, + "learning_rate": 4.3557690075246573e-10, + "loss": 0.8873, + "step": 80964 + }, + { + "epoch": 5.98, + "learning_rate": 4.318620028687104e-10, + "loss": 0.9081, + "step": 80965 + }, + { + "epoch": 5.98, + "learning_rate": 4.2816301415449236e-10, + "loss": 1.0213, + "step": 80966 + }, + { + "epoch": 5.98, + "learning_rate": 4.24479934616473e-10, + "loss": 0.9651, + "step": 80967 + }, + { + "epoch": 5.98, + "learning_rate": 4.2081276426020335e-10, + "loss": 0.9079, + "step": 80968 + }, + { + "epoch": 5.98, + "learning_rate": 4.1716150309012436e-10, + "loss": 1.0159, + "step": 80969 + }, + { + "epoch": 5.98, + "learning_rate": 4.135261511151178e-10, + "loss": 1.0247, + "step": 80970 + }, + { + "epoch": 5.98, + "learning_rate": 4.099067083374042e-10, + "loss": 0.9515, + "step": 80971 + }, + { + "epoch": 5.98, + "learning_rate": 4.0630317476586525e-10, + "loss": 0.9848, + "step": 80972 + }, + { + "epoch": 5.98, + "learning_rate": 4.027155504038316e-10, + "loss": 0.9899, + "step": 80973 + }, + { + "epoch": 5.98, + "learning_rate": 3.9914383525796465e-10, + "loss": 1.0694, + "step": 80974 + }, + { + "epoch": 5.98, + "learning_rate": 3.955880293349257e-10, + "loss": 0.9594, + "step": 80975 + }, + { + "epoch": 5.98, + "learning_rate": 3.920481326391556e-10, + "loss": 1.0343, + "step": 80976 + }, + { + "epoch": 5.98, + "learning_rate": 3.8852414517620563e-10, + "loss": 1.2032, + "step": 80977 + }, + { + "epoch": 5.98, + "learning_rate": 3.8501606695162675e-10, + "loss": 1.0104, + "step": 80978 + }, + { + "epoch": 5.98, + "learning_rate": 3.8152389797208033e-10, + "loss": 1.0748, + "step": 80979 + }, + { + "epoch": 5.98, + "learning_rate": 3.780476382420073e-10, + "loss": 0.9073, + "step": 80980 + }, + { + "epoch": 5.98, + "learning_rate": 3.74587287768069e-10, + "loss": 0.9589, + "step": 80981 + }, + { + "epoch": 5.98, + "learning_rate": 3.711428465547062e-10, + "loss": 1.023, + "step": 80982 + }, + { + "epoch": 5.98, + "learning_rate": 3.677143146074702e-10, + "loss": 1.0473, + "step": 80983 + }, + { + "epoch": 5.98, + "learning_rate": 3.6430169193302225e-10, + "loss": 1.1078, + "step": 80984 + }, + { + "epoch": 5.98, + "learning_rate": 3.60904978534693e-10, + "loss": 0.9199, + "step": 80985 + }, + { + "epoch": 5.98, + "learning_rate": 3.57524174420254e-10, + "loss": 0.9827, + "step": 80986 + }, + { + "epoch": 5.98, + "learning_rate": 3.5415927959303597e-10, + "loss": 1.0568, + "step": 80987 + }, + { + "epoch": 5.98, + "learning_rate": 3.5081029405970024e-10, + "loss": 1.0078, + "step": 80988 + }, + { + "epoch": 5.98, + "learning_rate": 3.4747721782468766e-10, + "loss": 1.0565, + "step": 80989 + }, + { + "epoch": 5.98, + "learning_rate": 3.441600508946596e-10, + "loss": 1.0475, + "step": 80990 + }, + { + "epoch": 5.98, + "learning_rate": 3.408587932729468e-10, + "loss": 1.051, + "step": 80991 + }, + { + "epoch": 5.98, + "learning_rate": 3.375734449673207e-10, + "loss": 1.0613, + "step": 80992 + }, + { + "epoch": 5.98, + "learning_rate": 3.343040059800018e-10, + "loss": 0.9394, + "step": 80993 + }, + { + "epoch": 5.98, + "learning_rate": 3.3105047631876165e-10, + "loss": 1.0403, + "step": 80994 + }, + { + "epoch": 5.98, + "learning_rate": 3.2781285598693093e-10, + "loss": 0.9343, + "step": 80995 + }, + { + "epoch": 5.98, + "learning_rate": 3.2459114499117093e-10, + "loss": 0.9751, + "step": 80996 + }, + { + "epoch": 5.98, + "learning_rate": 3.213853433359226e-10, + "loss": 0.9987, + "step": 80997 + }, + { + "epoch": 5.98, + "learning_rate": 3.1819545102562687e-10, + "loss": 1.0958, + "step": 80998 + }, + { + "epoch": 5.98, + "learning_rate": 3.15021468066945e-10, + "loss": 1.0241, + "step": 80999 + }, + { + "epoch": 5.99, + "learning_rate": 3.118633944632077e-10, + "loss": 0.9506, + "step": 81000 + }, + { + "epoch": 5.99, + "learning_rate": 3.0872123021996603e-10, + "loss": 0.8929, + "step": 81001 + }, + { + "epoch": 5.99, + "learning_rate": 3.0559497534277115e-10, + "loss": 0.9786, + "step": 81002 + }, + { + "epoch": 5.99, + "learning_rate": 3.0248462983717417e-10, + "loss": 1.0303, + "step": 81003 + }, + { + "epoch": 5.99, + "learning_rate": 2.993901937065058e-10, + "loss": 0.9903, + "step": 81004 + }, + { + "epoch": 5.99, + "learning_rate": 2.963116669563171e-10, + "loss": 0.8521, + "step": 81005 + }, + { + "epoch": 5.99, + "learning_rate": 2.9324904959215917e-10, + "loss": 0.9978, + "step": 81006 + }, + { + "epoch": 5.99, + "learning_rate": 2.902023416173627e-10, + "loss": 0.9601, + "step": 81007 + }, + { + "epoch": 5.99, + "learning_rate": 2.871715430385891e-10, + "loss": 1.0667, + "step": 81008 + }, + { + "epoch": 5.99, + "learning_rate": 2.8415665386027914e-10, + "loss": 0.9739, + "step": 81009 + }, + { + "epoch": 5.99, + "learning_rate": 2.8115767408576356e-10, + "loss": 1.1017, + "step": 81010 + }, + { + "epoch": 5.99, + "learning_rate": 2.781746037217037e-10, + "loss": 0.9026, + "step": 81011 + }, + { + "epoch": 5.99, + "learning_rate": 2.7520744277143017e-10, + "loss": 1.0622, + "step": 81012 + }, + { + "epoch": 5.99, + "learning_rate": 2.7225619124049416e-10, + "loss": 0.9409, + "step": 81013 + }, + { + "epoch": 5.99, + "learning_rate": 2.693208491333366e-10, + "loss": 1.0042, + "step": 81014 + }, + { + "epoch": 5.99, + "learning_rate": 2.664014164543982e-10, + "loss": 0.9361, + "step": 81015 + }, + { + "epoch": 5.99, + "learning_rate": 2.6349789320812005e-10, + "loss": 0.9601, + "step": 81016 + }, + { + "epoch": 5.99, + "learning_rate": 2.6061027940005314e-10, + "loss": 1.0295, + "step": 81017 + }, + { + "epoch": 5.99, + "learning_rate": 2.577385750346384e-10, + "loss": 1.0022, + "step": 81018 + }, + { + "epoch": 5.99, + "learning_rate": 2.548827801152065e-10, + "loss": 1.1256, + "step": 81019 + }, + { + "epoch": 5.99, + "learning_rate": 2.520428946484188e-10, + "loss": 0.9255, + "step": 81020 + }, + { + "epoch": 5.99, + "learning_rate": 2.4921891863649573e-10, + "loss": 0.9981, + "step": 81021 + }, + { + "epoch": 5.99, + "learning_rate": 2.464108520860986e-10, + "loss": 1.0681, + "step": 81022 + }, + { + "epoch": 5.99, + "learning_rate": 2.436186949994479e-10, + "loss": 0.9603, + "step": 81023 + }, + { + "epoch": 5.99, + "learning_rate": 2.408424473820947e-10, + "loss": 0.9911, + "step": 81024 + }, + { + "epoch": 5.99, + "learning_rate": 2.380821092395902e-10, + "loss": 1.0064, + "step": 81025 + }, + { + "epoch": 5.99, + "learning_rate": 2.3533768057415474e-10, + "loss": 1.0015, + "step": 81026 + }, + { + "epoch": 5.99, + "learning_rate": 2.326091613913395e-10, + "loss": 0.9062, + "step": 81027 + }, + { + "epoch": 5.99, + "learning_rate": 2.2989655169558534e-10, + "loss": 1.0081, + "step": 81028 + }, + { + "epoch": 5.99, + "learning_rate": 2.2719985149133316e-10, + "loss": 0.9623, + "step": 81029 + }, + { + "epoch": 5.99, + "learning_rate": 2.2451906078302387e-10, + "loss": 0.972, + "step": 81030 + }, + { + "epoch": 5.99, + "learning_rate": 2.2185417957287792e-10, + "loss": 0.9956, + "step": 81031 + }, + { + "epoch": 5.99, + "learning_rate": 2.192052078675566e-10, + "loss": 0.9935, + "step": 81032 + }, + { + "epoch": 5.99, + "learning_rate": 2.1657214567039065e-10, + "loss": 0.8808, + "step": 81033 + }, + { + "epoch": 5.99, + "learning_rate": 2.1395499298582089e-10, + "loss": 0.9102, + "step": 81034 + }, + { + "epoch": 5.99, + "learning_rate": 2.1135374981717805e-10, + "loss": 1.032, + "step": 81035 + }, + { + "epoch": 5.99, + "learning_rate": 2.087684161700132e-10, + "loss": 1.0837, + "step": 81036 + }, + { + "epoch": 5.99, + "learning_rate": 2.061989920465468e-10, + "loss": 1.0139, + "step": 81037 + }, + { + "epoch": 5.99, + "learning_rate": 2.0364547745232997e-10, + "loss": 0.9361, + "step": 81038 + }, + { + "epoch": 5.99, + "learning_rate": 2.0110787239180364e-10, + "loss": 1.0196, + "step": 81039 + }, + { + "epoch": 5.99, + "learning_rate": 1.9858617686718818e-10, + "loss": 0.9505, + "step": 81040 + }, + { + "epoch": 5.99, + "learning_rate": 1.9608039088403474e-10, + "loss": 0.9216, + "step": 81041 + }, + { + "epoch": 5.99, + "learning_rate": 1.9359051444678424e-10, + "loss": 1.0651, + "step": 81042 + }, + { + "epoch": 5.99, + "learning_rate": 1.9111654755654685e-10, + "loss": 0.9054, + "step": 81043 + }, + { + "epoch": 5.99, + "learning_rate": 1.8865849022109418e-10, + "loss": 1.0509, + "step": 81044 + }, + { + "epoch": 5.99, + "learning_rate": 1.862163424404262e-10, + "loss": 0.8868, + "step": 81045 + }, + { + "epoch": 5.99, + "learning_rate": 1.8379010422231446e-10, + "loss": 0.9731, + "step": 81046 + }, + { + "epoch": 5.99, + "learning_rate": 1.8137977556786924e-10, + "loss": 0.987, + "step": 81047 + }, + { + "epoch": 5.99, + "learning_rate": 1.7898535648153138e-10, + "loss": 0.9061, + "step": 81048 + }, + { + "epoch": 5.99, + "learning_rate": 1.7660684696774179e-10, + "loss": 0.962, + "step": 81049 + }, + { + "epoch": 5.99, + "learning_rate": 1.7424424702983112e-10, + "loss": 1.0716, + "step": 81050 + }, + { + "epoch": 5.99, + "learning_rate": 1.7189755667113005e-10, + "loss": 0.9635, + "step": 81051 + }, + { + "epoch": 5.99, + "learning_rate": 1.695667758960795e-10, + "loss": 0.981, + "step": 81052 + }, + { + "epoch": 5.99, + "learning_rate": 1.6725190470801013e-10, + "loss": 1.0519, + "step": 81053 + }, + { + "epoch": 5.99, + "learning_rate": 1.649529431113628e-10, + "loss": 1.033, + "step": 81054 + }, + { + "epoch": 5.99, + "learning_rate": 1.626698911094682e-10, + "loss": 0.9038, + "step": 81055 + }, + { + "epoch": 5.99, + "learning_rate": 1.6040274870454676e-10, + "loss": 0.9777, + "step": 81056 + }, + { + "epoch": 5.99, + "learning_rate": 1.5815151590214961e-10, + "loss": 0.9717, + "step": 81057 + }, + { + "epoch": 5.99, + "learning_rate": 1.5591619270560744e-10, + "loss": 0.9315, + "step": 81058 + }, + { + "epoch": 5.99, + "learning_rate": 1.5369677911714064e-10, + "loss": 1.0191, + "step": 81059 + }, + { + "epoch": 5.99, + "learning_rate": 1.5149327514119016e-10, + "loss": 1.1864, + "step": 81060 + }, + { + "epoch": 5.99, + "learning_rate": 1.4930568078108666e-10, + "loss": 0.9496, + "step": 81061 + }, + { + "epoch": 5.99, + "learning_rate": 1.4713399604016078e-10, + "loss": 1.0029, + "step": 81062 + }, + { + "epoch": 5.99, + "learning_rate": 1.4497822092174318e-10, + "loss": 0.9464, + "step": 81063 + }, + { + "epoch": 5.99, + "learning_rate": 1.428383554302748e-10, + "loss": 0.9123, + "step": 81064 + }, + { + "epoch": 5.99, + "learning_rate": 1.4071439956797606e-10, + "loss": 1.0701, + "step": 81065 + }, + { + "epoch": 5.99, + "learning_rate": 1.3860635333928786e-10, + "loss": 0.9719, + "step": 81066 + }, + { + "epoch": 5.99, + "learning_rate": 1.3651421674643063e-10, + "loss": 1.049, + "step": 81067 + }, + { + "epoch": 5.99, + "learning_rate": 1.3443798979384526e-10, + "loss": 0.9374, + "step": 81068 + }, + { + "epoch": 5.99, + "learning_rate": 1.3237767248486245e-10, + "loss": 0.9637, + "step": 81069 + }, + { + "epoch": 5.99, + "learning_rate": 1.3033326482059238e-10, + "loss": 1.0041, + "step": 81070 + }, + { + "epoch": 5.99, + "learning_rate": 1.2830476680769644e-10, + "loss": 0.8631, + "step": 81071 + }, + { + "epoch": 5.99, + "learning_rate": 1.262921784461746e-10, + "loss": 0.9905, + "step": 81072 + }, + { + "epoch": 5.99, + "learning_rate": 1.2429549974157796e-10, + "loss": 0.9172, + "step": 81073 + }, + { + "epoch": 5.99, + "learning_rate": 1.2231473069501676e-10, + "loss": 1.0426, + "step": 81074 + }, + { + "epoch": 5.99, + "learning_rate": 1.203498713120421e-10, + "loss": 1.1143, + "step": 81075 + }, + { + "epoch": 5.99, + "learning_rate": 1.1840092159376426e-10, + "loss": 0.9996, + "step": 81076 + }, + { + "epoch": 5.99, + "learning_rate": 1.1646788154351386e-10, + "loss": 1.0884, + "step": 81077 + }, + { + "epoch": 5.99, + "learning_rate": 1.1455075116573178e-10, + "loss": 0.964, + "step": 81078 + }, + { + "epoch": 5.99, + "learning_rate": 1.1264953046263849e-10, + "loss": 0.9703, + "step": 81079 + }, + { + "epoch": 5.99, + "learning_rate": 1.1076421943756466e-10, + "loss": 1.0835, + "step": 81080 + }, + { + "epoch": 5.99, + "learning_rate": 1.088948180916205e-10, + "loss": 0.9737, + "step": 81081 + }, + { + "epoch": 5.99, + "learning_rate": 1.0704132643146736e-10, + "loss": 0.9298, + "step": 81082 + }, + { + "epoch": 5.99, + "learning_rate": 1.0520374445599502e-10, + "loss": 0.9989, + "step": 81083 + }, + { + "epoch": 5.99, + "learning_rate": 1.0338207217075457e-10, + "loss": 0.9753, + "step": 81084 + }, + { + "epoch": 5.99, + "learning_rate": 1.0157630957907671e-10, + "loss": 1.0465, + "step": 81085 + }, + { + "epoch": 5.99, + "learning_rate": 9.978645668096143e-11, + "loss": 1.0781, + "step": 81086 + }, + { + "epoch": 5.99, + "learning_rate": 9.801251348195984e-11, + "loss": 0.8994, + "step": 81087 + }, + { + "epoch": 5.99, + "learning_rate": 9.625447998429239e-11, + "loss": 0.9119, + "step": 81088 + }, + { + "epoch": 5.99, + "learning_rate": 9.451235619017951e-11, + "loss": 1.0797, + "step": 81089 + }, + { + "epoch": 5.99, + "learning_rate": 9.278614210184167e-11, + "loss": 0.9843, + "step": 81090 + }, + { + "epoch": 5.99, + "learning_rate": 9.107583772371975e-11, + "loss": 1.0132, + "step": 81091 + }, + { + "epoch": 5.99, + "learning_rate": 8.938144305692398e-11, + "loss": 0.9264, + "step": 81092 + }, + { + "epoch": 5.99, + "learning_rate": 8.770295810589524e-11, + "loss": 0.9641, + "step": 81093 + }, + { + "epoch": 5.99, + "learning_rate": 8.604038287063354e-11, + "loss": 0.9929, + "step": 81094 + }, + { + "epoch": 5.99, + "learning_rate": 8.439371735669e-11, + "loss": 1.0546, + "step": 81095 + }, + { + "epoch": 5.99, + "learning_rate": 8.276296156517483e-11, + "loss": 1.065, + "step": 81096 + }, + { + "epoch": 5.99, + "learning_rate": 8.114811549830848e-11, + "loss": 0.9331, + "step": 81097 + }, + { + "epoch": 5.99, + "learning_rate": 7.954917916053184e-11, + "loss": 0.9206, + "step": 81098 + }, + { + "epoch": 5.99, + "learning_rate": 7.796615255184492e-11, + "loss": 1.0557, + "step": 81099 + }, + { + "epoch": 5.99, + "learning_rate": 7.63990356766886e-11, + "loss": 0.8993, + "step": 81100 + }, + { + "epoch": 5.99, + "learning_rate": 7.484782853617311e-11, + "loss": 1.0258, + "step": 81101 + }, + { + "epoch": 5.99, + "learning_rate": 7.331253113473936e-11, + "loss": 1.0594, + "step": 81102 + }, + { + "epoch": 5.99, + "learning_rate": 7.17931434723873e-11, + "loss": 0.8949, + "step": 81103 + }, + { + "epoch": 5.99, + "learning_rate": 7.028966555355788e-11, + "loss": 0.9708, + "step": 81104 + }, + { + "epoch": 5.99, + "learning_rate": 6.880209737936127e-11, + "loss": 0.9058, + "step": 81105 + }, + { + "epoch": 5.99, + "learning_rate": 6.733043895312819e-11, + "loss": 0.9359, + "step": 81106 + }, + { + "epoch": 5.99, + "learning_rate": 6.587469027596882e-11, + "loss": 0.9958, + "step": 81107 + }, + { + "epoch": 5.99, + "learning_rate": 6.443485135121385e-11, + "loss": 0.9338, + "step": 81108 + }, + { + "epoch": 5.99, + "learning_rate": 6.301092218108373e-11, + "loss": 1.0256, + "step": 81109 + }, + { + "epoch": 5.99, + "learning_rate": 6.16029027677989e-11, + "loss": 0.9618, + "step": 81110 + }, + { + "epoch": 5.99, + "learning_rate": 6.021079311246958e-11, + "loss": 1.0208, + "step": 81111 + }, + { + "epoch": 5.99, + "learning_rate": 5.883459321953666e-11, + "loss": 0.9769, + "step": 81112 + }, + { + "epoch": 5.99, + "learning_rate": 5.747430308900015e-11, + "loss": 1.024, + "step": 81113 + }, + { + "epoch": 5.99, + "learning_rate": 5.612992272419071e-11, + "loss": 1.0911, + "step": 81114 + }, + { + "epoch": 5.99, + "learning_rate": 5.480145212732879e-11, + "loss": 1.0211, + "step": 81115 + }, + { + "epoch": 5.99, + "learning_rate": 5.3488891299524614e-11, + "loss": 1.0113, + "step": 81116 + }, + { + "epoch": 5.99, + "learning_rate": 5.219224024410885e-11, + "loss": 1.0104, + "step": 81117 + }, + { + "epoch": 5.99, + "learning_rate": 5.091149896219172e-11, + "loss": 1.0421, + "step": 81118 + }, + { + "epoch": 5.99, + "learning_rate": 4.964666745599367e-11, + "loss": 0.9625, + "step": 81119 + }, + { + "epoch": 5.99, + "learning_rate": 4.839774572884537e-11, + "loss": 1.0164, + "step": 81120 + }, + { + "epoch": 5.99, + "learning_rate": 4.7164733780746816e-11, + "loss": 1.0683, + "step": 81121 + }, + { + "epoch": 5.99, + "learning_rate": 4.5947631615028686e-11, + "loss": 0.8279, + "step": 81122 + }, + { + "epoch": 5.99, + "learning_rate": 4.4746439232801195e-11, + "loss": 1.0193, + "step": 81123 + }, + { + "epoch": 5.99, + "learning_rate": 4.356115663628479e-11, + "loss": 0.998, + "step": 81124 + }, + { + "epoch": 5.99, + "learning_rate": 4.239178382769993e-11, + "loss": 1.0182, + "step": 81125 + }, + { + "epoch": 5.99, + "learning_rate": 4.123832080815682e-11, + "loss": 1.0067, + "step": 81126 + }, + { + "epoch": 5.99, + "learning_rate": 4.010076757987591e-11, + "loss": 1.0715, + "step": 81127 + }, + { + "epoch": 5.99, + "learning_rate": 3.897912414507765e-11, + "loss": 1.0342, + "step": 81128 + }, + { + "epoch": 5.99, + "learning_rate": 3.7873390504872265e-11, + "loss": 0.9784, + "step": 81129 + }, + { + "epoch": 5.99, + "learning_rate": 3.67835666614802e-11, + "loss": 0.9753, + "step": 81130 + }, + { + "epoch": 5.99, + "learning_rate": 3.570965261601167e-11, + "loss": 0.9425, + "step": 81131 + }, + { + "epoch": 5.99, + "learning_rate": 3.4651648371797354e-11, + "loss": 1.0678, + "step": 81132 + }, + { + "epoch": 5.99, + "learning_rate": 3.3609553928837247e-11, + "loss": 1.0093, + "step": 81133 + }, + { + "epoch": 5.99, + "learning_rate": 3.2583369289351796e-11, + "loss": 1.0776, + "step": 81134 + }, + { + "epoch": 6.0, + "learning_rate": 3.157309445445122e-11, + "loss": 0.9593, + "step": 81135 + }, + { + "epoch": 6.0, + "learning_rate": 3.057872942635598e-11, + "loss": 0.9737, + "step": 81136 + }, + { + "epoch": 6.0, + "learning_rate": 2.960027420728651e-11, + "loss": 1.0003, + "step": 81137 + }, + { + "epoch": 6.0, + "learning_rate": 2.8637728797242802e-11, + "loss": 1.0328, + "step": 81138 + }, + { + "epoch": 6.0, + "learning_rate": 2.769109319955554e-11, + "loss": 0.8931, + "step": 81139 + }, + { + "epoch": 6.0, + "learning_rate": 2.6760367413114495e-11, + "loss": 0.9416, + "step": 81140 + }, + { + "epoch": 6.0, + "learning_rate": 2.5845551442360563e-11, + "loss": 0.9764, + "step": 81141 + }, + { + "epoch": 6.0, + "learning_rate": 2.4946645287293737e-11, + "loss": 0.9787, + "step": 81142 + }, + { + "epoch": 6.0, + "learning_rate": 2.4063648949024244e-11, + "loss": 1.0167, + "step": 81143 + }, + { + "epoch": 6.0, + "learning_rate": 2.319656242977253e-11, + "loss": 0.9524, + "step": 81144 + }, + { + "epoch": 6.0, + "learning_rate": 2.2345385730648817e-11, + "loss": 1.0431, + "step": 81145 + }, + { + "epoch": 6.0, + "learning_rate": 2.1510118851653105e-11, + "loss": 0.8667, + "step": 81146 + }, + { + "epoch": 6.0, + "learning_rate": 2.0690761796116066e-11, + "loss": 1.0411, + "step": 81147 + }, + { + "epoch": 6.0, + "learning_rate": 1.988731456514792e-11, + "loss": 0.9311, + "step": 81148 + }, + { + "epoch": 6.0, + "learning_rate": 1.9099777158748666e-11, + "loss": 1.0175, + "step": 81149 + }, + { + "epoch": 6.0, + "learning_rate": 1.8328149579138755e-11, + "loss": 0.9933, + "step": 81150 + }, + { + "epoch": 6.0, + "learning_rate": 1.7572431827428403e-11, + "loss": 1.1249, + "step": 81151 + }, + { + "epoch": 6.0, + "learning_rate": 1.6832623904727842e-11, + "loss": 0.9409, + "step": 81152 + }, + { + "epoch": 6.0, + "learning_rate": 1.6108725812147284e-11, + "loss": 0.9462, + "step": 81153 + }, + { + "epoch": 6.0, + "learning_rate": 1.5400737550796963e-11, + "loss": 1.0006, + "step": 81154 + }, + { + "epoch": 6.0, + "learning_rate": 1.4708659121787096e-11, + "loss": 1.0753, + "step": 81155 + }, + { + "epoch": 6.0, + "learning_rate": 1.4032490526227904e-11, + "loss": 0.9653, + "step": 81156 + }, + { + "epoch": 6.0, + "learning_rate": 1.3372231765229615e-11, + "loss": 0.9316, + "step": 81157 + }, + { + "epoch": 6.0, + "learning_rate": 1.2727882841012673e-11, + "loss": 1.0817, + "step": 81158 + }, + { + "epoch": 6.0, + "learning_rate": 1.2099443752466855e-11, + "loss": 1.0207, + "step": 81159 + }, + { + "epoch": 6.0, + "learning_rate": 1.1486914500702384e-11, + "loss": 0.9102, + "step": 81160 + }, + { + "epoch": 6.0, + "learning_rate": 1.089029508904993e-11, + "loss": 0.9756, + "step": 81161 + }, + { + "epoch": 6.0, + "learning_rate": 1.0309585516399267e-11, + "loss": 0.8998, + "step": 81162 + }, + { + "epoch": 6.0, + "learning_rate": 9.744785783860622e-12, + "loss": 1.0359, + "step": 81163 + }, + { + "epoch": 6.0, + "learning_rate": 9.195895893654438e-12, + "loss": 0.8856, + "step": 81164 + }, + { + "epoch": 6.0, + "learning_rate": 8.662915845780717e-12, + "loss": 1.0012, + "step": 81165 + }, + { + "epoch": 6.0, + "learning_rate": 8.145845640239458e-12, + "loss": 0.9366, + "step": 81166 + }, + { + "epoch": 6.0, + "learning_rate": 7.644685279251107e-12, + "loss": 1.0804, + "step": 81167 + }, + { + "epoch": 6.0, + "learning_rate": 7.159434761705442e-12, + "loss": 1.0397, + "step": 81168 + }, + { + "epoch": 6.0, + "learning_rate": 6.6900940909331305e-12, + "loss": 1.0533, + "step": 81169 + }, + { + "epoch": 6.0, + "learning_rate": 6.236663266934173e-12, + "loss": 1.0624, + "step": 81170 + }, + { + "epoch": 6.0, + "learning_rate": 5.799142288598347e-12, + "loss": 0.9782, + "step": 81171 + }, + { + "epoch": 6.0, + "learning_rate": 5.377531159256321e-12, + "loss": 0.9432, + "step": 81172 + }, + { + "epoch": 6.0, + "learning_rate": 4.9718298777978736e-12, + "loss": 0.9604, + "step": 81173 + }, + { + "epoch": 6.0, + "learning_rate": 4.582038444223003e-12, + "loss": 0.9875, + "step": 81174 + }, + { + "epoch": 6.0, + "learning_rate": 4.208156861862378e-12, + "loss": 1.0173, + "step": 81175 + }, + { + "epoch": 6.0, + "learning_rate": 3.850185128495554e-12, + "loss": 1.0034, + "step": 81176 + }, + { + "epoch": 6.0, + "learning_rate": 3.508123247453199e-12, + "loss": 1.0652, + "step": 81177 + }, + { + "epoch": 6.0, + "learning_rate": 3.1819712165148676e-12, + "loss": 1.0345, + "step": 81178 + }, + { + "epoch": 6.0, + "learning_rate": 2.8717290379010055e-12, + "loss": 1.0635, + "step": 81179 + }, + { + "epoch": 6.0, + "learning_rate": 2.577396711611613e-12, + "loss": 1.1377, + "step": 81180 + }, + { + "epoch": 6.0, + "learning_rate": 2.2989742376466895e-12, + "loss": 1.1298, + "step": 81181 + }, + { + "epoch": 6.0, + "learning_rate": 2.0364616171164588e-12, + "loss": 1.0028, + "step": 81182 + }, + { + "epoch": 6.0, + "learning_rate": 1.7898588511311432e-12, + "loss": 1.0105, + "step": 81183 + }, + { + "epoch": 6.0, + "learning_rate": 1.5591659385805202e-12, + "loss": 0.8762, + "step": 81184 + }, + { + "epoch": 6.0, + "learning_rate": 1.3443828805748126e-12, + "loss": 1.0481, + "step": 81185 + }, + { + "epoch": 6.0, + "learning_rate": 1.1455096771140207e-12, + "loss": 0.9954, + "step": 81186 + }, + { + "epoch": 6.0, + "learning_rate": 9.62546328198144e-13, + "loss": 1.0019, + "step": 81187 + }, + { + "epoch": 6.0, + "learning_rate": 7.954928360476289e-13, + "loss": 0.9019, + "step": 81188 + }, + { + "epoch": 6.0, + "learning_rate": 6.443491984420292e-13, + "loss": 0.9334, + "step": 81189 + }, + { + "epoch": 6.0, + "learning_rate": 5.09115417601791e-13, + "loss": 0.9832, + "step": 81190 + }, + { + "epoch": 6.0, + "learning_rate": 3.8979149241669123e-13, + "loss": 0.9925, + "step": 81191 + }, + { + "epoch": 6.0, + "learning_rate": 2.8637742288673e-13, + "loss": 0.9296, + "step": 81192 + }, + { + "epoch": 6.0, + "learning_rate": 1.9887321123235326e-13, + "loss": 1.0313, + "step": 81193 + }, + { + "epoch": 6.0, + "learning_rate": 1.27278855233115e-13, + "loss": 1.0567, + "step": 81194 + }, + { + "epoch": 6.0, + "learning_rate": 7.159435599923825e-14, + "loss": 1.0045, + "step": 81195 + }, + { + "epoch": 6.0, + "learning_rate": 3.1819713530723e-14, + "loss": 0.8662, + "step": 81196 + }, + { + "epoch": 6.0, + "learning_rate": 7.954928937792262e-15, + "loss": 0.9733, + "step": 81197 + }, + { + "epoch": 6.0, + "learning_rate": 0.0, + "loss": 1.1207, + "step": 81198 + }, + { + "epoch": 6.0, + "step": 81198, + "total_flos": 9.230830326470882e+19, + "train_loss": 1.0133407875950922, + "train_runtime": 262143.0861, + "train_samples_per_second": 4.956, + "train_steps_per_second": 0.31 + } + ], + "max_steps": 81198, + "num_train_epochs": 6, + "total_flos": 9.230830326470882e+19, + "trial_name": null, + "trial_params": null +}